142 IEEE JOURNAL ON MULTISCALE AND MULTIPHYSICS COMPUTATIONAL TECHNIQUES, VOL.

9, 2024

Efficient Iterative Solution of Combined Source

Integral Equation Using Characteristic Basis
Function Method With Initial Guess
Zhiwen Dong , Xinlei Chen , Member, IEEE, Fan Gao , Changqing Gu , Zhuo Li , Member, IEEE,
Wu Yang , and Weibing Lu , Senior Member, IEEE

Abstract—Using only the RWG functions, the combined source just needs to discretize the surface of objects. The electric field
integral equation (CSIE) with weak form combined source con- integral equation (EFIE) is known for its excellent accuracy
dition can achieve fine accuracy and fast iterative convergence for but poor spectrum distribution of the integral operator [2],
conductor objects. However, compared with a conventional integral
equation in the method of moments (MoM), the conventional CSIE [3]. Conversely, the magnetic field integral equation (MFIE)
involves more matrices and more complex numerical processing, generally produces a matrix equation with a fast iterative con-
and these make the CSIE inefficient, especially for multiple ex- vergence speed, but the non-dual testing scheme makes the
citation problems. In this article, a characteristic basis function accuracy of MFIE unsatisfactory [4], [5], [6]. To improve the
(CBF)-based CSIE with initial guess is proposed to mitigate this convergence of EFIE and the accuracy of MFIE, the Calderón
problem. The CBF is employed to reduce the number of unknowns
as well as the storage consumptions and iteration time. In the preconditioner [3], [7], and the dual-tested scheme [4], [5], [8],
meantime, an initial guess especially for CBFs is proposed to reduce [9] are proposed in earlier years. However, these technologies
iterations when solving multiple excitation problems. Numerical need the Buffa–Christiansen (BC) functions [10] and involve
results are given to demonstrate the performance of the proposed a barycentric refinement, which is more complicated than the
method. Rao–Wilton–Glisson (RWG) functions [11] and will result in
Index Terms—Method of moments (MoM), characteristic basis six times meshes. The combined field integral equation (CFIE)
function method (CBFM), combined source integral equation [12] is a convex combination of EFIE and MFIE, with tradeoffs
(CSIE), electromagnetic scattering, initial guess. in accuracy and convergence, while immune to the interior
I. INTRODUCTION resonance problem. Nevertheless, the accuracy of CFIE is some-
times unsatisfactory due to the inclusion of the non-dual testing
OR the electromagnetic (EM) scattering problems, the
F method of moments (MoM) [1] is attractive since it usually
scheme MFIE.
For conductor objects, the combined source integral equation
Manuscript received 8 October 2023; revised 3 December 2023; accepted 28 (CSIE) [13], [14], [15] is an interior resonance–free alternative
December 2023. Date of publication 28 March 2024; date of current version 11 to CFIE. In addition, different types of CSIE have been used
April 2024. This work was supported in part by the National Nature Science
Foundation of China under Grant 62371228, in part by the National Science to analyze homogeneous dielectric targets [16], [17]. In recent
Funds for Distinguished Young Scientists under Grant 61925103, in part by the years, the combined source integral equation (CSIE) with weak
Key Program of National Natural Science Foundation of China under Grant form combined source (CS) condition has been proposed [18],
62231001, in part by the Foundation of State Key Laboratory of Millimeter
Waves, Southeast University, China, under Grant K202415, and in part by the [19]. Its good accuracy is ensured by the dual testing scheme
Project of Key Laboratory of Radar Imaging and Microwave Photonics (Nanjing and it shows fast iterative convergence. Only RWG functions are
University of Aeronautics and Astronautics), Ministry of Education under Grant employed, without any barycentric refinement. However, CSIE
NJ20220004. (Corresponding authors: Xinlei Chen; Weibing Lu.)
Zhiwen Dong, Fan Gao, Changqing Gu, and Zhuo Li are with the Key Lab- involves more matrices and more complex numerical processing
oratory of Radar Imaging and Microwave Photonics, College of Electronic and than conventional integral equations. There are two operator
Information Engineering, Nanjing University of Aeronautics and Astronautics, matrices discretized from the operators and two linear complex-
Nanjing 211106, China (e-mail: [email protected]; [email protected];
[email protected]; [email protected]). ity Gram matrices in CSIE, while just one operator matrix is
Xinlei Chen is with the Key Laboratory of Radar Imaging and Microwave in conventional integral equations. Besides, when dealing with
Photonics, College of Electronic and Information Engineering, Nanjing Uni- multiple incidences, each vector on the right-hand side needs to
versity of Aeronautics and Astronautics, Nanjing 211106, China, and also with
the State Key Laboratory of Millimeter Waves, Southeast University, Nanjing restart the iteration process, leading to a long iteration time.
210096, China (e-mail: [email protected]). In this paper, a characteristic basis function (CBF)-based
Wu Yang and Weibing Lu are with the State Key Laboratory of Millimeter CSIE with initial guess is proposed to enhance the efficiency
waves, School of Information Science and Engineering, Southeast University,
Nanjing 210096, China, also with the Center for Flexible RF Technology and of the conventional CSIE. The CBF [20], [21] is employed to
MOE Frontiers Science Center for Mobile Information Communication and reduce the number of unknowns and memory consumption of the
Security, Southeast University, Nanjing 210096, China, and also with the Purple CSIE while keeping good accuracy. Meanwhile, an initial guess
Mountain Laboratories, Nanjing 210096, China (e-mail: [email protected];
[email protected]). for CBF is proposed to reduce the iteration time of solving the
Digital Object Identifier 10.1109/JMMCT.2024.3382725 multiple incidences problems.

2379-8793 © 2024 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See https://www.ieee.org/publications/rights/index.html for more information.

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.
DONG et al.: EFFICIENT ITERATIVE SOLUTION OF CSIE USING CHARACTERISTIC BASIS FUNCTION METHOD WITH INITIAL GUESS 143

where the elements of ZL ,ZK ,Gf nf ,Gf f ,VE are expressed as

 
L
Zmn = − jk fm (r) · G(r, r )fn (r )dr dr
Tm Tn
 
1
− G(r, r )∇ · fm (r)∇ · fn (r )dr dr
jk Tm Tn
(8)
 
K
Zmn = fm (r) · ∇G(r, r ) × fn (r )dr dr (9)
Tm Tn

Fig. 1. A Closed PEC object impacted by an incident field in the free space. S Gfmn
nf
= fm (r) · n̂ × fn (r)dr (10)
denotes the closed surface of the object, with the unit normal vector n̂. J is the Tm
surface electric current density and M is the surface magnetic current density.

(a) Original problem. (b) Equivalence problem: the external field equal to that Gfmn
f
= fm (r) · fn (r)dr (11)
of the original problem, with a non-zero internal field. Tm

VmE = fm (r) · Einc (r)dr (12)
Tm
II. FORMULATION
Tm and Tn are the domain of the mth RWG testing function and
A. CSIE-CBFM the nth RWG basis function, respectively. The EFIE in CSIE
Consider a scattering problem of a PEC object in the free tested in dual space explains its good accuracy.
space, as shown in Fig. 1(a), the EFIE in the CSIE framework Substituting (7) into (6) to eliminate b, the final matrix equa-
[14], [15], [16], [17] shown in Fig. 1(b) can be expressed as tion is obtained
 
  1 f nf  f f −1 f nf
1  L
Z a − Z0 Z − G K
G G a = −VE (13)
Z0 L {J (r )} − KP V {M (r )} + n̂ × I {M (r)} 
 
2
2 tan
 Compared with the conventional EFIE, the CSIE needs to
inc 
= − E (r) tan , r ∈ S (1) store more matrices and more complex numerical processing. In
this paper, the CBFM [21] is employed to mitigate this problem.
with the definition of operators
 In the CBFM, the object is divided into separate subsystems, and
then RWG functions are linearly combined to construct CBFs in
L {X (r )} = − jk G(r, r )X(r )dr
S
each subsystem. These CBFs will be employed as the new basis
 functions and testing functions, eventually reducing the size of
1
+∇ G(r, r )∇ · X(r )dr (2) the system matrix.
jk S Taking the mth subsystem as an example, the CBFs are

constructed as follows. Firstly, solve the subsystem with multiple
KP V {X (r )} = P.V. ∇G(r, r ) × X(r )dr (3)
S incidences
 
I {X (r)} = X (r) (4) 1 f nf  f f −1 f nf
ZLmm mJ −Z 0 Z K
mm − G Gmm Gmm Jm = − Vm E
2 mm
where r and r stands for the field vector and source vector, (14)
respectively. Z0 is the free space impedance, P.V. means the
Cauchy principal value integration, and G(r, r ) is the scalar Secondly, apply the SVD to the solution of multiple
Green’s function of the free space. This EFIE is an underde- incidences
termined system and should be augmented by an additional CS Jm = UΣVH (15)
condition [14], [22]
where U and V are the orthogonal matrices, and Σ is the diagonal
M (r) = Z0 n̂ × J (r) (5) matrix with the diagonal element being the singular value of Jm .
Lastly, select the first k columns of matrix U (corresponding to
to obtain a unique solution. the first k largest singular values) with k determined by the SVD
Both unknown surface currents J(r) and M(r) are expanded threshold as CBFs in the mth subsystem, i.e.,
by the RWG functions [11], and the EFIE and the CS condition
are tested by RWG functions [18], [19]. The MoM equation Cm = U k (16)
system in the CSIE is expressed as
The CBFs for the whole problem are expressed in block diagonal
 
L K 1 f nf matrix form as
Z a− Z − G b = − VE (6)
2 C = diag (C1 , C2 , · · · , CB ) (17)
ff f nf
G b = Z0 G a (7) with B being the number of subsystems.

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.
144 IEEE JOURNAL ON MULTISCALE AND MULTIPHYSICS COMPUTATIONAL TECHNIQUES, VOL. 9, 2024

an RWG matrix into a CBF matrix, for one mutual subsystem

or self-subsystem matrix the time complexity is
2 2 2 2
O(M  N  ) + O(M  N  ) = O(M  N  + M  N  ) (22)
with that N  is the number of RWG in each subsystem and is also
a constant when the subsystem is fixed. Then for all subsystem
matrices, the time cost is
2 2
B × B × O(M  N  + M  N  ) = O(M N + M 2 ) = O(M N )
(23)
The time to generate the CBF matrix costs less than solving
CSIE. The generation of CBFs needs time and the time cost is
Fig. 2. An Object is divided into subsystems according to an octree structure. 3 3 2
For a subsystem, there are 26 near-field subsystems and a self-subsystem. B × O(N  ) + O(N  ) = O(N  N ) = O(N ) (24)

for B LU decompositions of size N  and matrix multiplications.

The compression of impedance matrix and the generation of
Finally, the operator discretized with CBFs can be expressed CBFs make the proposed method take more time than CSIE for
as generating system matrices. Nevertheless, since the complexity
ZCBF = CH ZC (18) of the proposed method is O(M 2 ) and that of CSIE is O(N 2 )
when using an iterative solver to solve the final equation, the
where Z could be ZL , ZK , Gf nf , and Gf f . The final matrix proposed method will show an advantage in a time when there
system with size M, where M is the number of CBFs, is written are many right-hand-side vectors.
as
  C. Initial Guess for CBFM
L,CBF CBF K,CBF 1 f nf ,CBF
Z a − Z0 Z − G
2 When dealing with multiple excitation EM scattering prob-
 f f ,CBF −1 f nf ,CBF CBF lems, each plane wave excitation needs to restart the iteration
× G G a = −VE,CBF (19) process, resulting in a long iteration time. To further reduce the
with iteration time of the CSIE-CBFM, an initial guess based on the
CBF is proposed.
VE,CBF = CH VE (20) If the incident angles of two plane waves are close, the
solutions of the two plane waves are highly correlated, and
The magnetic current is not considered unknown and is cal-
the solution of the previous angle can be utilized to reduce the
culated after solving the electric current.
iteration of the next angle [23]. Due to a phase factor e−jk·r ,
where k is the incident angle direction vector, existing in the
B. Complexity of CSIE-CBFM right-hand side of the matrix equation, it can be assumed that
An octree structure is taken as an example, as shown in Fig. 2. there is also the same factor in the current solution. Removing
Assuming that the number of CBFs in each subsystem is equal the phase factor for the current solution
and is M  , if the number of subsystems is sufficient, the memory J1 (r) = J (r) ejk·r (25)
complexity of a Gram matrix is
2
J1 (r) usually varies more slowly relative to J(r) when the
B × 27 × O(M  ) = O(M  M ) = O(M ) (21) incident angle changes. Next, shifting the phase suitable for the
where M  can be regarded as a constant when the subsystem next incident angle
 
is fixed and M is the total number of CBFs. For a subsystem, J2 (r) = J1 (r) e−jk ·r = J (r) ejk·r e−jk ·r (26)
there are 26 near-field subsystems and a self-subsystem, and the 
where k is the incident angle direction vector of the next angle,
Gram matrices of far-field subsystems are zero matrices. Thus
J2 (r) is used as the initial guess for the iteration solution of the
it is still a linear complexity.
next incident angle.
As for the storage complexities of the operators L and K (here
Since the phase varies with the field vector r, the phase
K containing the Cauchy principal value part and the identity
correction should be employed on the coefficients of RWG
part with the cross product), they are both O(M 2 ), while that of
basis functions rather than the center of subsystems. First, the
the operators discretized with RWG functions are O(N 2 ) with
coefficients of CBFs aCBF need to be returned to the coefficients
N being the number of RWG functions. Therefore, the storage
of RWG basis functions a
complexities of operators L and K account for the main part, and
the use of CBFM can achieve a lower memory consumption. a = CaCBF (27)
When iteratively solving the final system, the time cost of
and then employing the phase correction
the proposed method is O(M 2 ), while that of solving CSIE is

O(N 2 ). Since CBF is a block diagonal matrix when compressing ãm = am ejk·rm e−jk ·rm (28)

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.
DONG et al.: EFFICIENT ITERATIVE SOLUTION OF CSIE USING CHARACTERISTIC BASIS FUNCTION METHOD WITH INITIAL GUESS 145

where am is the mth element of a, ãm is the mth element of ã

after phase correction, and rm is the center of the mth RWG.
The new coefficients ãCBF of CBFs should satisfy the following
equation
⎛ CBF ⎞ ⎛ ⎞
ã1 ã1
⎜ãCBF ⎟ ⎜ ã2 ⎟
⎜ 2 ⎟ ⎜ ⎟
CãCBF = (c1 , c2 , · · · , cM ) ⎜ . ⎟ = ã = ⎜ . ⎟ (29)
⎝ .. ⎠ ⎝ .. ⎠
ãCBF
M
ãN
with ci the ith column of C, ãCBF
m the mth element of ã
CBF
, and
ãm the mth element of ã.
A problem existing in the above equation is that ã is a vector
of length N (the number of RWGs) and may not be exactly
expressed in the space expanded by the columns of C, because
the number of total CBFs M is always smaller than N. Hence,
we choose the projection ā of ã onto the column space of C as
the initial guess for the next incidence. One advantage of using Fig. 3. Bistatic RCS of EFIE, MFIE, CFIE, and CSIE for the pyramid at 300
this type of CBFs is that the columns of C are unit orthogonal, MHz.
eliminating the calculation of orthogonalization. The ith element
of the projection āCBF is
ã, ci 
āCBF
i = = ã, ci  = cH
i · ã (30)
ci , ci 
The new initial guess of the reduced matrix equation, āCBF ,
for the next incidence written in matrix form is
āCBF = CH · ã (31)

III. NUMERICAL RESULTS

This section is to show the efficiency and accuracy of the
proposed method through several examples of the conductor
objects. In the proposed method, an octree structure is employed
and the CBFs are generated from the CSIE in subsystems where
LU decomposition is used to solve the matrix equation in each
subsystem. For all the examples, when solving the final system
equation in the CSIE and the CSIE-CBFM, the generalized
minimal residual (GMRES) is employed as the inner iterative
solver and the outer iterative solver. Each subsystem size is 1.0 Fig. 4. Bistatic RCS of CSIE and CSIE-CBFM for the pyramid at 300 MHz.
wavelength and will be extended by 0.1 wavelengths to generate
an extended subsystem. The SVD threshold to generate CBFs is
1e-3. where e includes the magnitudes of the electric fields for differ-
ent angles, the error of MFIE is −11.02 dB, and that of CFIE is
A. Pyramid −16.37 dB. The error of the conventional CSIE and the proposed
A pyramid with a base edge length of 3 m and a height of 6 m is method are −20.69 dB and −20.39 dB, respectively. Including
shown, exhibiting the accuracy and compression performance of MFIE makes CFIE less accurate. Compared with CFIE, the CSIE
the proposed method. An x-polarized incident plane wave with achieves higher accuracy. At the same time, the use of CBFM
a frequency of 300 MHz impacts the sphere from a direction can maintain similar accuracy as the conventional CSIE.
of θ = 0◦ , ϕ = 0◦ . The surface of the sphere is discretized by With the help of CBFM, the number of unknowns is reduced
triangles with an average size of 0.1 m, resulting in 15177 RWG from 15177 to 3988. The memory consumption of CSIE is 3514
unknowns. MB while that of the proposed method is 300 MB, achieving a
The bistatic RCS of different methods is shown in Figs. 3 compression ratio of over 11 times.
and 4, with EFIE results as the reference solution. The iteration Fig. 5 shows the iterative convergence of several methods. In
threshold for all methods is set to 1e-3. With a definition of this example, the convergence of the proposed CSIE-CBFM is
root-mean-square (RMS) error superior to that of EFIE. In short, the proposed CSIE-CBFM
could keep a similar level of accuracy to EFIE while its iterative
e − eref 2 /eref 2 (32) convergence is better than EFIE.

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.
146 IEEE JOURNAL ON MULTISCALE AND MULTIPHYSICS COMPUTATIONAL TECHNIQUES, VOL. 9, 2024

Fig. 5. Convergence of different methods for the pyramid. Fig. 7. Convergence of EFIE, CSIE, and the proposed method in the helicopter
model at a frequency of 300 MHz.

TABLE I
PERFORMANCE COMPARISON OF CSIE AND CSIE-CBFM FOR CALCULATING
THE MONOSTATIC SCATTERING OF THE HELICOPTER

the iterative convergence of the proposed method is similar to

that of the CSIE and is better than that of the EFIE.
The number of iterations for different initial guesses in the
Fig. 6. Monostatic RCS of different methods for the helicopter at 300 MHz. proposed method is shown in Fig. 8. Using a zero-initial value in
the iterations process for all angles requires an average of about
367 iterations. Just using the solution of the previous angle as
an initial value for the next angle needs an average of about 285
B. Helicopter Platform
iterations, while using the solution of the previous angle with
In this part, a multiscale helicopter platform impacted by phase correction as an initial value for the next angle needs an
multiple plane waves is shown. The monostatic RCS results at average of about 165 iterations.
300 MHz of the proposed method compared with that of EFIE, The time consumed is shown in Table I. It can be seen that
CFIE, and CSIE are shown in Fig. 6. The θ angle sample interval when dealing with multiple right-hand-side vectors, iterations
of the plane waves is 1.0°, and there are 361 samples. The size take up most of the time, and the reduction of unknowns reduces
of the helicopter is about 13.91 m × 5.23 m × 3.08 m and the iteration time. Compared with the conventional CSIE, the
the number of RWGs is 46227, as shown in Fig. 7. With the CSIE-CBFM without initial guess (zero initial value) saves
proposed method, the number of unknowns is reduced to 7082 about 91% in total time by reducing the number of unknowns.
and the memory usage is 891 MB, as shown in Table I, realizing a By using the initial guess with phase correction, the iteration
memory compression ratio of over 36 times. The EFIE is chosen time is further reduced to 2.56 h, saving about 55% compared
as the reference solution. The error of CFIE is −11.97 dB. The with the CSIE-CBFM without initial guess.
errors of CSIE and the proposed method are −18.53 dB and In addition, the computational performance of two methods at
−15.83 dB, respectively. several frequencies is shown in Table II. The proposed method
Fig. 7 shows the iterative convergence of different methods in represents the CSIE-CBFM using the initial guess with phase
this model at a frequency of 300 MHz, where can be seen that correction. The “Unknown” in CSIE represents the number of

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.
DONG et al.: EFFICIENT ITERATIVE SOLUTION OF CSIE USING CHARACTERISTIC BASIS FUNCTION METHOD WITH INITIAL GUESS 147

REFERENCES
[1] W. C. Gibson, The Method of Moments in Electromagnetics. Boca Raton,
FL, USA: CRC Press, 2007.
[2] J.-C. Nedéléc, Acoustic and Electromagnetic Equations. New York, NY,
USA: Springer, 2000.
[3] F. P. Andriulli et al., “A multiplicative Calderon preconditioner for the
electric field integral equation,” IEEE Trans. Antennas Propag., vol. 56,
no. 8, pp. 2398–2412, Aug. 2008.
[4] K. Cools, F. Andriulli, F. Olyslager, and E. Michielssen, “Improving the
MFIE’s accuracy by using a mixed discretization,” in Proc. IEEE Antennas
Propag. Soc. Int. Symp., 2009, pp. 1–4.
[5] K. Cools, F. P. Andriulli, D. De Zutter, and E. Michielssen, “Accurate and
conforming mixed discretization of the MFIE,” IEEE Antennas Wireless
Propag. Lett., vol. 10, pp. 528–531, 2011.
[6] P. Ylä-Oijala, S. P. Kiminki, J. Markkanen, and S. Järvenpää, “Error-
controllable and well-conditioned MoM solutions in computational elec-
tromagnetics: Ultimate surface integral-equation formulation,” IEEE An-
tennas Propag. Mag., vol. 55, no. 6, pp. 310–331, Dec. 2013.
[7] G. C. Hsiao and R. E. Kleinman, “Mathematical foundation for error
estimation in numerical solution of integral equations in electromagnetics,”
IEEE Trans. Antennas Propag., vol. 45, no. 3, pp. 316–328, Mar. 1997.
Fig. 8. The number of iterations for different initial guesses to reach 10−3 in [8] S. Yan, J.-M. Jin, and Z. Nie, “Improving the accuracy of the second-
the helicopter model. kind Fredholm integral equations by using the Buffa-Christiansen func-
tions,” IEEE Trans. Antennas Propag., vol. 59, no. 4, pp. 1299–1310,
Apr. 2011.
TABLE II [9] S. Yan, J.-M. Jin, and Z. Nie, “Accuracy improvement of the second-kind
PERFORMANCE COMPARISON OF CSIE AND THE PROPOSED METHOD FOR integral equations for generally shaped objects,” IEEE Trans. Antennas
CALCULATING THE MONOSTATIC SCATTERING OF THE HELICOPTER AT Propag., vol. 61, no. 2, pp. 788–797, Feb. 2013.
SEVERAL FREQUENCIES [10] A. Buffa and S. Christiansen, “A dual finite element complex on
the barycentric refinement,” Math. Computation, vol. 76, no. 260,
pp. 1743–1769, Oct. 2007.
[11] S. Rao, D. Wilton, and A. Glisson, “Electromagnetic scattering by surfaces
of arbitrary shape,” IEEE Trans. Antennas Propag., vol. AP-30, no. 3,
pp. 409–418, May 1982.
[12] J. R. Mautz and R. F. Harrington, “H-field, E-field, and combined-field
solutions for conducting bodies of revolution,” Int. J. Electron. Commun.,
vol. 32, no. 4, pp. 157–164, 1978.
[13] J.-C. Bolomey and W. Tabbara, “Numerical aspects on coupling be-
tween complementary boundary value problems,” IEEE Trans. Antennas
Propag., vol. AP-21, no. 3, pp. 356–363, May 1973.
[14] J. Mautz and R. Harrington, “A combined-source solution for radiation
and scattering from a perfectly conducting body,” IEEE Trans. Antennas
Propag., vol. AP-27, no. 4, pp. 445–454, Jul. 1979.
[15] J. Lee and S. Nam, “Protecting the method of auxiliary sources (MAS)
solutions from the interior resonance problem,” IEEE Microw. Wireless
RWG and the number of CBFs in the proposed method. Similar Compon. Lett., vol. 15, no. 3, pp. 186–188, Mar. 2005.
to Table I, in Table II “Mat.” denotes the matrix generation, “Ite.” [16] P. Ylä-Oijala, S. P. Kiminki, K. Cools, F. P. Andriulli, and S. Järvenpää,
“Stable discretization of combined source integral equation for scattering
denotes the iteration, and “Tot.” denotes the Total. It can be seen by dielectric objects,” IEEE Trans. Antennas Propag., vol. 60, no. 5,
that although the matrix generation time is slightly increased in pp. 2575–2578, May 2012.
the proposed method, its total calculation time is greatly reduced [17] P. Ylä-Oijala, S. P. Kiminki, and S. Järvenpää, “Electric and magnetic
combined source integral equations for electromagnetic transmission prob-
due to the substantial reduction of iteration time. In the 600 MHz lems,” Eng. Anal. Boundary Elements, vol. 37, no. 1, pp. 68–73, Jan. 2013.
example, it failed to solve by using CSIE because there are too [18] J. Kornprobst and T. J. Eibert, “A combined source integral equation with
many unknowns, and the memory usage exceeds the maximum weak form combined source condition,” IEEE Trans. Antennas Propag.,
vol. 66, no. 4, pp. 1251–2155, Apr. 2018.
memory of the computer. However, the 600 MHz example is [19] J. Kornprobst and T. J. Eibert, “A weak-form combined source inte-
solvable through the proposed method, and it has much less gral equation with explicit inversion of the combined-source condition,”
total time than the 300 MHz example in CSIE. IEEE Antennas Wireless Propag. Lett., vol. 17, no. 12, pp. 2174–2178,
Dec. 2018.
[20] V. V. S. Prakash and R. Mittra, “Characteristic basis function method:
IV. CONCLUSION A new technique for efficient solution of method of moments matrix
equations,” Microw. Opt. Technol. Lett., vol. 36, no. 2, pp. 95–100,
A CBF-based CSIE with initial guess is proposed to accu- Jan. 2003.
[21] E. Lucente, A. Monorchio, and R. Mittra, “An iteration-free MoM
rately and efficiently solve the multiple excitation EM scattering approach based on excitation independent characteristic basis func-
problems. The use of CBFM effectively reduces the number of tions for solving large multiscale electromagnetic scattering prob-
unknowns, and the initial guess for CBFs significantly reduces lems,” IEEE Trans. Antennas Propag., vol. 56, no. 4, pp. 999–1007,
Apr. 2008.
the number of iterations. The proposed CSIE-CBFM can keep [22] J. Rogers, “On combined source solutions for bodies with impedance
a similar accuracy as the conventional CSIE because it is still boundary conditions,” IEEE Trans. Antennas Propag., vol. AP-33, no. 4,
tested in the dual space of its range space. The numerical pp. 462–465, Apr. 1985.
[23] J. Song, C.-C. Lu, and W. C. Chew, “Multilevel fast multipole algorithm
experiment shows that the convergence of the proposed method for electromagnetic scattering by large complex objects,” IEEE Trans.
is much better than that of EFIE. Antennas Propag., vol. 45, no. 10, pp. 1488–1493, Oct. 1997.

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.
148 IEEE JOURNAL ON MULTISCALE AND MULTIPHYSICS COMPUTATIONAL TECHNIQUES, VOL. 9, 2024

Zhiwen Dong was born in Shaoguan, Guangdong Zhuo Li (Member, IEEE) received the B.Sc. degree in
Province, China, in 1996. He received the B.S. degree electronic engineering and the M. E. degree in elec-
in electronic information science and technology in tromagnetic fields and microwave techniques from
2019 from the Nanjing University of Aeronautics the Nanjing University of Aeronautics and Astronau-
and Astronautics, Nanjing, China, where he is cur- tics, Nanjing, China, in 2001 and 2004, respectively,
rently working toward the Ph.D. degree in informa- and the Ph.D. degree in radio engineering from the
tion and communication engineering. His research State Key Laboratory of Millimeter Waves, Southeast
interests include integral equation methods in compu- University, Nanjing, in 2009. From 2015 to 2016,
tational electromagnetics, well-conditioned formula- he was a Visiting Scholar with the Department of
tions, and conforming discretization schemes. Physics, Arizona State University, Tempe, AZ, USA.
He is currently a Full Professor with the College of
Electronic and Information Engineering, NUAA. He has authored or coauthored
more than 160 papers in refereed journals and conference proceedings, including
Advanced Science, Advanced Optical Materials, Nano Letters, Physical Review
Applied, IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES,
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, and IEEE TRANS-
ACTIONS ON ELECTROMAGNETIC COMPATIBILITY. He has filed more than 30
Xinlei Chen (Member IEEE) was born in Xuzhou,
patents. His research interests include plasmonic metamaterials, MHz-through-
Jiangsu Province, China, in 1984. He received the
THz technologies and transceivers for wireless sensors and biomedical applica-
B.S. degree in electronic information science and
tions, electromagnetic compatibility in avionics system, auto electrical system,
technology, the M.E. degree in electromagnetic field
and modeling and design of microwave and terahertz photonic circuits and
and microwave technology, and the Ph.D. degree
in communication and information system from the systems. Prof. Li is a Senior Member of Chinese Institute of Electronics and
a Committee Member of Antenna Branch of Chinese Institute of Electronics.
Nanjing University of Aeronautics and Astronautics,
He was the recipient of the Second Prize of Natural Science of China Institute
Nanjing, China, in 2007, 2010, and 2015, respec-
of Electronics in 2019 and NUAA Outstanding Undergraduate Teaching Pro-
tively. From 2015 to 2017, he was a Lecturer with the
College of Electronic and Information Engineering, motion Award in 2019. He is on the Advisory Board for Journal of Physics
D: Applied Physics and on the Review Board for many international journals
Nanjing University of Aeronautics and Astronautics.
including Advanced Materials, ACS Photonics, Nanotechnology, Nano-Micro
He is currently an Associate Professor with the College of Electronic and
Letters, IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, IEEE
Information Engineering, Nanjing University of Aeronautics and Astronautics.
He has authored or coauthored more than 60 papers in journals and confer- TRANSACTIONS ON ANTENNAS AND PROPAGATION, and IEEE TRANSACTIONS
ON ELECTROMAGNETIC COMPATIBILITY. He is also an invited speaker and the
ences, including IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, IEEE
session chair in many international conferences and workshops.
TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, and IEEE Antennas
Wireless Propagation Letter, and IET Microwaves, Antennas & Propagation. His
research interests include integral equation methods in computational electro-
magnetics, and electromagnetic scattering and radiation.

Wu Yang was born in Jiangxi, China, in 1978. He

received the B.S. and Ph.D. degrees from the De-
Fan Gao was born in Taixing, Jiangsu Province, partment of Mathematics, University of Science and
China, in 1996. He received the B.S. degree in elec- Technology of China, Hefei, China, in 1998 and 2003,
tronic information science and technology from the respectively. From 2003 to 2005, he was a Postdoctor-
Huaiyin Institute of Technology, Huaian, China, in ate with the College of Computer Science, Zhejiang
2018. He is currently working toward the Ph.D. de- University. From 2005 to 2017, he was an Associate
gree in information and communication engineer- Professor with the School of Computer Science and
ing with the Nanjing University of Aeronautics and Engineering, Southeast University. He is currently an
Astronautics, Nanjing, China. His research interests Associate Professor with the State Key Laboratory
include computational electromagnetics and its appli- of Millimeter Waves, School of Information Science
cation in scattering problems. and Engineering, Southeast University, Nanjing, China. His research interests
include computational electromagnetic and computer graphics.

Weibing Lu (Senior Member, IEEE) was born in

Changqing Gu was born in Jiangsu, China, in 1958. Jiangsu, China, in 1977. He received the B.S. and
He received the M.E. degree in electromagnetic field M.S. degrees from Northeast Normal University,
and microwave technology from the Huazhong In- Changchun, China, in 1999 and 2002, respectively,
stitute of Technology, Wuhan, China, in 1986. Then, and the Ph.D. degree from Southeast University, Nan-
he joined the Nanjing University of Aeronautics and jing, China, in 2005. Since 2005, he has been with
Astronautics, Nanjing, China, where he is currently the State Key Lab of Millimeter Waves, Southeast
a Professor with the College of Electronic and Infor- University, where he is currently a Chief Professor.
mation Engineering. He has authored or coauthored His research interests include flexible RF technology,
more than 100 papers in refereed journals and con- near zero communication, and computational elec-
ference proceedings, including IEEE TRANSACTIONS tromagnetics. He was the recipient of a New Century
ON ANTENNAS AND PROPAGATION, IEEE TRANSAC- Excellent Scholar Award from the Chinese Ministry of Education in 2008, Award
TIONS ON ELECTROMAGNETIC COMPATIBILITY, Applied Physics Letter, Optics of Six Talent Peaks Project in Jiangsu Province in 2016. He was supported by
Letters, and Optics Express. His research interests include computational elec- the National Science Fund for Distinguished Young Scholars in 2019, and was
tromagnetics, electromagnetic compatibility, antenna design, metamaterials, and selected as Jiangsu Specially-Appointed Professor in 2020.
metasurface.

Authorized licensed use limited to: ANKIT PANDEY. Downloaded on April 23,2024 at 17:02:59 UTC from IEEE Xplore. Restrictions apply.