Research Article: Open Access

DOI: 10.35840/2631-5092/4524

International Journal of
VIBGYOR Optic and Photonic Engineering

Non-Split Perfectly Matched Layer Boundary

Conditions for Numerical Solution of 2D
Maxwell Equations
ISSN: 2631-5092

Arnold Abramov* , Yutao Yue and Mingming Wang

Institute of Deep Perception Technology, Wuxi, China

Abstract
This paper developed a non-split perfectly matched layer (PML) boundary condition (BC)
for Finite Difference Time Domain (FDTD) simulation of electromagnetic wave propagation
in 2D structure. The point source for electric field has been exploited for propagation of
electromagnetic field through 2D structures to validate developed approach. The identity of
resulted field distribution to that obtained for split PML BC have been demonstrated.
Keywords
Electromagnetic wave, Non-split, PML, FDTD, Cylinder

Introduction el can affect significantly the simulation results,

due to spurious reflection from the boundaries of
Electromagnetic field distribution of any sys-
electromagnetic waves incident on them. To ob-
tem can be known by solving Maxwell’s equations.
tain physically reasonable solutions, the calculation
There are many numerical techniques allowing to
area should be interrupted at the edges via special
find appropriate solutions. In this paper we will
boundary conditions that extend the definite area
deal with the finite difference time domain meth-
to infinity, representing it as unbounded volume.
od introduced [1] in 1966 by Kane S. Yee. It has a
Such a boundary condition could be obtained by a
number of advantages, such as divergence-free (in
reflectionless absorbing layer that surrounds the
contrast to the main methods that require a condi-
calculation area and absorbs all incident waves, re-
tion div B = 0), the grid is very simple and the data
gardless of the angle of incidence and wavelength.
is easily for storage). The FDTD method solves the
Thus, in the implementation of FDTD, we set the
time-dependent Maxwell’s equations in a spatially
layers of a hypothetical material along the bound-
finite computation domain. The electric and mag-
aries of computational area specially designed in
netic fields are then represented by their discrete
such a way to completely absorb the radiation inci-
values on the each node of spatial grid, and are
dent on it. Such a layers have been called perfectly
extended in time. Without any boundary condi-
matching layer (PML) and were first introduced by
tions, Yee’s algorithm assumes that the domain is
J. P. Berenger [2]. The PML conditions have a low
surrounded by a perfect electric conductor (PEC)
reflection coefficient, as well as practical indepen-
layer. The presence of the boundaries in the mod-
dence from the angle of incidence of the wave.

*Corresponding author: Arnold Abramov, Institute of Deep Perception Technology, Wuxi, China
Accepted: June 15, 2020; Published: June 17, 2020
Copyright: © 2020 Abramov A, et al. This is an open-access article distributed under the terms of the Creative
Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium,
provided the original author and source are credited.
Qi et al. Int J Opt Photonic Eng 2020, 5:023
Citation: Abramov A, Yue Y, Wang M (2020) Non-Split Perfectly Matched Layer Boundary Conditions for Numerical Solution of 2D
Maxwell Equations. Int J Opt Photonic Eng 5:024
Abramov et al. Int J Opt Photonic Eng 2020, 5:024 ISSN: 2631-5092 | • Page 2 of 6 •

Unfortunately, there is still a reflection: From instead of the electric field. This allows the PML to
the first layer of PML; between layers of PML, be- be independent of the material of the FDTD com-
cause to save computing resources, losses increase putational domain.
from layer to layer; after the last layer of PML, since
Basic Equations
there is a PEC border. Reflection from the first PML
layer and between the PML layers is caused by fi- The starting point for the construction of an
nite-difference discretization errors, and, first of FDTD algorithm is Maxwell’s time-domain equa-
all, by the fact that the vectors E and H do not coin- tions.

cide in space. To reduce reflection within the PML, ∂D  
= ∇×H
it is necessary to strict the growth rate of losses ∂t
to some reasonable limit. Reflection from the PEC  
D = ε 0ε E (1)
boundary after the last PML layer occurs usually for 
already much attenuated wave. The reflected wave ∂H 1  
= − ∇× E
continues to weaken on the way back. But if there ∂t µ0
are few layers (usually less than 5), then the reflect- 
Where E is the  electric field strength vector in
ed wave can be significant. To reduce the reflection
volts per meter, H is the electric displacement vec-
from the first layer, the value of conductivity σ1 is 
tor in coulombs per square meter, H is the magnet-
specially chosen to be small. To reduce the reflec-
ic field strength vector, ε0 and µ0 - dielectric and
tion between the layers, the loss profile is selected
magnetic constant respectively, ε is the permittiv-
in such a way to have a limited growth rate of loss.
ity of the material. Using new expressions for the
In order to reduce the effect of waves reflected
fields
from the PEC boundary, the number of PML layers
should be increased.  ε0 
E→ E
µ0
Berenger’s PML is typically referred to as a split (2)
field PML because the field components in PML  1 
D→ D
regions are split into two unphysical field compo- ε 0 µ0
nents to obtain additional degrees of freedom. This Leads to

modification allows the creation of a reflectionless ∂D  
interface between a dielectric medium and the = c∇ × H
∂t
PML layer. But extra splitting variables that are in-  
D = εE (3)
troduced in Berenger’s PML BC increase computa- 
∂H  
tional memory time burden in implementation. In = − c∇ × E
addition, it leads also to the difficulties of numeri- ∂t
cal implementation. In order to avoid those issues Where we used the speed of light in free space
scholars have done a lot of researches on develop- c = 1/ ε 0 µ0 .
ing of versions of PML absorbing boundary condi- For specifics, we further consider the case of TE
tion for non-splitting fields: Unsplit-field and mate- polarizations for which electric field (along z axis)
rial independent PML formulations are obtained in is orthogonal to the plane (x,y) of electromagnetic
[3] incorporating the auxiliary differential equation field propagation. Thus the field has next nonzero
method into the PML formulations. The non-split components: Ez, Hx, Hy.
PML BC has been used in seismic wave modeling in-
troducing auxiliary variables [4]. A use of non-split ∂Dz  ∂H y ∂H x 
= c − 
PML boundary condition for second-order elastic ∂t  ∂x ∂y 
wave FETD forward modeling with high calculation Dz = ε Ez
accuracy and great improvement in calculation ef- (4)
∂H x ∂Ez
ficiency have been demonstrated in [5]. Non-split = −c
∂t ∂y
PML BC formulation developed previously [3] for
one dimensional case adopted in the present pa- ∂H y ∂Ez
= c
per for 2D Maxwell’s equations. Another feature of ∂t ∂x
the presented formulations (except non-split PML In order to implement the FDTD method and
BC) is that the use of the electric displacement field calculate the field at all points of the computation-

Citation: Abramov A, Yue Y, Wang M (2020) Non-Split Perfectly Matched Layer Boundary Conditions for Numerical Solution of 2D
Maxwell Equations. Int J Opt Photonic Eng 5:024
Abramov et al. Int J Opt Photonic Eng 2020, 5:024 ISSN: 2631-5092 | • Page 3 of 6 •

al domain we have to replace the derivatives by σ Dm

finite-difference approximations and set the field ε=
*
m εm +
iωε 0
values ​​at the boundaries of the computational do- (7)
σ
main. But before we will modify these equations as- µ=
*
m µ m + Hm
suming the presence of an additional region around iωµ0
the computational domain where the field will be m = x,y, with following selection of parameters
absorbed, excluding its reflection to the back into εm
= µ
= 1
m
computational area. For this purpose, we introduce (8)
σ Dm σ Hm
fictitious values ​​that will be selected in such a way = =
as not to affect the field inside the computation- ε0 µ0 ε0
al domain, and to ensure its absorption outside it. Theoretically, if the condition σi / ε0 = σi* / μ0 is
We first write down the system (4) for the case of satisfied, then the average velocity of the electro-
continuous, time-harmonic fields (i.e. time depen- magnetic waves does not change at the interface
dence are described by factor exp(iωt)): and the reflection is equal to zero. At the same
 ∂H y ∂H x  time, because σi, σi* ≠ 0 the absorption of electro-
iω D z
= c −  magnetic waves takes place within the PML. PML
 ∂x ∂y 
parameters used in this work will be described be-
D z (ω ) = ε (ω ) Ez low.
(5)
∂Ez
iω H x = −c Substituting (7-8) into (6) we get:
∂y
 σ ( x)   σ ( y )   ∂H y ∂H x 
∂Ez iω  1 +  1 +  Dz = c  − 
iω H y = c  iωε 0  iωε 0   ∂x ∂y 
∂x
D z = ε Ez
Non-Split PML Boundary Conditions  σ ( x)   σ ( y ) 
−1
∂Ez (9)
iω  1 +  1 +  Hx =
−c
In order to implement PML we input fictitious  iωε 0   iωε 0  ∂y
constants εx*, εy*, µ x*, µ y*:  σ ( x)   σ ( y )  ∂Ez
−1

iω  1 +  1 +  Hy =
c
 ∂H y ∂H x   iωε 0  iωε 0  ∂x
iω D=
zε x ε y
* *
c − 
 ∂x ∂y  We can rewrite these equations as follows:
D z = ε Ez  σ ( x) + σ ( y ) 1 σ ( x)σ ( y )   ∂H y ∂H x 
(6)  iω +
iωε 0
+
iω ε 02
 Dz =
c
 ∂x
−
∂y 

∂Ez  
iω H x µ x* µ *y = −c Dz = ε Ez
∂y  σ ( y)   ∂Ez σ ( x) ∂Ez  (10)
 iω + H =
−c + 
∂Ez  ε0  x  ∂y iωε 0 ∂y 
iω H y µ µ*
x
*
y = c  σ ( x)   ∂Ez σ ( y ) ∂Ez 
∂x  iω + H =c + 
 ε0  y  ∂x iωε 0 ∂x 
These fictitious parameters should be chosen
so both that there are no changes in the medium To go back to time dependence, we will perform
parameters within the main computational domain a formal replacement iω → d/dt. Then instead of
and ensure the absorption of the field in the PML (10) we have:
∂Dz σ ( x) + σ ( y ) σ ( x)σ ( y )  ∂H y ∂H x 
domain. ∂t
+
ε0
Dz +
ε 02
∑ D=
z ∆t c − 
 ∂x ∂y 
n
(11)
By preserving the electric displacement field Dz Dz = ε Ez
∂H x σ ( y )  ∂Ez σ ( x) ∂E 
in the equations (6), we obtain that all dimensional ∂t
+
ε0
Hx = −c + ∑ z ∆t 
ε 0 n ∂y
 ∂y 
parameters, as well as all the detailed information ∂H y σ ( x)  ∂Ez σ ( y ) ∂Ez 
about the structure of the computational domain, ∂t
+
ε0
H y =c  + ∑ ∆t 
ε 0 n ∂x
 ∂x 
are enclosed on the input of the dielectric constant. Now we replace the derivatives with their fi-
It also allows the PML to be independent of the ma- nite-difference approximations and we finally get
terial in the FDTD computing area. the PML-modified FDTD equations:
In [6] it was shown that in order to ensure ab-  σ ( x) + σ ( y )
1 + ∆t +
σ ( x)σ ( y )   σ ( x) + σ ( y )  n-1/2
(∆t ) 2  Dzn +1/2 (i, j )= 1 − ∆t  Dz (i, j ) −
2ε 0 ε 02 2ε 0
sorption in the PML layers, the expressions for the 
σ ( x)σ ( y ) c∆t
  
(∆t ) 2 ∑ Dzn +1/2 (i, j ) + ( H yn (i + 1/ 2, j ) − H yn (i − 1/ 2, j ) − H xn (i, j + 1/ 2) + H xn (i, j − 1/ 2) )
ε* and µ* should be as follows ε0 2
n ∆x

Citation: Abramov A, Yue Y, Wang M (2020) Non-Split Perfectly Matched Layer Boundary Conditions for Numerical Solution of 2D
Maxwell Equations. Int J Opt Photonic Eng 5:024
 Abramov et al. Int J Opt Photonic Eng 2020, 5:024 ISSN: 2631-5092 | • Page 4 of 6 •

 σ ( y )  n +1  σ ( y)  n As mentioned above a propagation of TE-polarized

1 + ∆t  H x (i, j + 1 / 2) =  1 − ∆t  H x (i, j + 1 / 2) −
 2ε 0   2ε 0  wave in 2D plane was simulating via solution of
 Ezn +1/2 (i, j + 1) − Ezn +1/2 (i, j ) σ ( x) ∂Ez  Maxwell’s equations (1) by FDTD method. The di-
c∆t  + ∑ ∆t 
 ∆ x ε 0 n ∂y  rection of wave propagation is along the x-axis (Fig-

ure 1). The origin of the coordinates corresponds
σ ( x)  n +1  σ ( x)  n
1 +
2ε
∆t  H y (i + 1 / 2, j ) = 1 −
2ε 0
∆t  H y (i + 1 / 2, j ) + to the centre of the computational domain. Three
 0   
 E n +1/2 (i + 1, j ) − Ezn +1/2 (i, j ) σ ( y ) ∂E 
cases were considered: 2D homogeneous space
c∆t  z
∆x
+ ∑ z ∆t 
ε 0 n ∂x (Figure 1a) and two kinds of structures placed into
 
centre of 2D computational area: Single cylinder
The selection of the fictitious parameters can (Figure 1b); group of 9 cylinders (Figure 1c). We
be determined by the following expressions intro- choose cylinder as main figure in our simulation
duced in Ref. [6]: due to since it is the study of scattering on the cyl-
σ ( m) inder that led to the discovery of photonic nanojet
= ∆t pm=, m x, y
2ε 0 [7] and a further increase in researchers' attention
to scattering effects on cylindrical objects [8-14].
With pm = 0.333(i/NPML)3 , i = 1.. NPML, NPML - num-
ber of PML. Refraction index of each cylinder are nA = 1.59.
All results below are presented after 1200 time step
Finally the FDTD equations are as follow: simulations. Each time step was equal ∆t = Sc*∆x/c,
=Dzn +1/2 (i, j )
1 − ( px + p y )
Dzn-1/2 (i, j ) − 4 px p y I1 +
where Sc = 0.5 is Courant stability factor.
1 + px + p y + 1 − 4 px p y
1 c∆t Polynomial [12] and power dependence [6] mod-
( H yn (i + 1/ 2, j ) − H yn (i − 1/ 2, j ) − H xn (i, j + 1/ 2) + H xn (i, j − 1/ 2) )
1 + px + p y + 1 − 4 px p y ∆x els for medium parameters have been choused for
I1n=
+1/ 2
I1n −1/ 2 + Dzn −1/ 2 split and non-split PML BC respectively. Effective-
ness of those models demonstrates Figure 2: Pro-
H xn +1 (=
i, j + 1 / 2)
1 − py
H xn (i, j + 1 / 2) −
c∆t
( Ezn +1/2 (i, j + 1) − Ezn +1/2 (i, j ) + 2 px I 2 ) jections of the calculated field distribution E(x,y) on
1 + py (1 + p y )∆x
the plane of two variables (E,x) and (E,y) have been
I 2n +1/=
2
I 2n −1/2 + ( Ezn-1/2 (i, j + 1) − Ezn-1/2 (i, j ) ) drawn. Field absorption occurs within PML layers
(x,y = 0-10 and x,y = 190-200). A maximal value on
1 − px n c∆t
H yn +1 (i + 1=
/ 2, j )
1 + px
H y (i + 1 / 2, j ) +
(1 + px )∆x
( Ezn +1/2 (i + 1, j ) − Ezn +1/2 (i, j ) + 2 p y I3 ) the dependencies corresponds to position of the
point source, where a field magnitude is kept con-
I 3n +1/ 2 = I 3n −1/2 + ( Ezn-1/2 (i + 1, j ) − Ezn-1/2 (i, j ) )
stant.
Results The difference between two approaches (non-
The model we used consists of rectangular area split and split PML BC) was estimated by expression
of size 200 × 200 unit cells. The size of the unit cell R = (Ens-Es)/Es
∆x was chosen in such a way that one wavelength λ
is equal to integer number N of unit cells: λ = N∆x. Where Ens and Es are absolute value of calculated
Point source generating electric field was defined electric field for the cases of non-split and split BC
offset 30 cells from the left edge of the problem respectively.
space along x-axis. Number of PML layers NPML = 10. The Figure 3 demonstrates change of R along

Figure 1: Computational domain of the two dimensional problem spaces: 1- PML; 2- Area of real space param-
eters; 3- Infinite cylinder (top view).

Citation: Abramov A, Yue Y, Wang M (2020) Non-Split Perfectly Matched Layer Boundary Conditions for Numerical Solution of 2D
Maxwell Equations. Int J Opt Photonic Eng 5:024
Abramov et al. Int J Opt Photonic Eng 2020, 5:024 ISSN: 2631-5092 | • Page 5 of 6 •

Figure 2: Projections of the field distribution E(x,y) inside 2D area (Figure 1a) on the plane of two variables
(E, x) (a) and (E, y) (b).
Solid line - Split PML BC; Dashed line - Non-split PML BC.

Figure 3: Relative difference between simulation results for non-split and split PML BC along x (dashed line)
and y (solid line) axis’s for one (a) and nine (b) Cylinders in computational area.

and across field propagation for the cases of (0, 1)
and 9 cylinders inside a computational area. As it
is seen the relative difference is less than 4%. The
cases of 0 (empty computational domain, i.e. free
propagation of electromagnetic wave) and 1 cylin-
der are represented by 1 line because the differ-
ence in R for them occurred less than 1%. A point
where R = 0 corresponds to position of point source.
In fact, all the lines on Figure 3 are oscillating and
short vertical lines shows an amplitude of those os-
cillations while main lines itself represent average
positions for R dependencies. More complex struc-
ture clearly leads to a greater error and oscillations.
However, overall these changes are minor, within
4%, those demonstrating efficiency of approxima-
tion efficiency and applicability of non-split BC for
FDTD method.
magnetic field splitting which does not follow from
Maxwell's equations, and thus it can lead to the
save of the computational time. As result of used
approach, all dimensional parameters, as well as all
the detailed information about the structure of the
computational domain, are enclosed on the input
of the dielectric constant. It is because of using the
electric displacement field instead of the electric
field. This allows also the PML to be independent
of the material of the FDTD computational domain.
The developed calculation scheme verified by
comparison of final results to those obtained for
ordinary (split) version of BC. Calculated field dis-
tributions for different parameters of computation-
al domain examined for both cases and are in the
good agreement.

Conclusion
The FDTD method to solve 2D Maxwell’s equa-
tions has been described. Non-split PML BC has
been used. This approach allows to avoid electro-
References
1. KS Yee (1966) Numerical solution of initial boundary
value problems involving maxwell’s equations in iso-
tropic media. IEEE Trans Antenn Propag 14: 302-307.

Citation: Abramov A, Yue Y, Wang M (2020) Non-Split Perfectly Matched Layer Boundary Conditions for Numerical Solution of 2D
Maxwell Equations. Int J Opt Photonic Eng 5:024
Abramov et al. Int J Opt Photonic Eng 2020, 5:024
2. JP Berenger (1994) A perfectly match layer for the
absorption of electromagnetic waves. Journal of
Computational Physics 114: 185-200.
3. O Ramadan (2003) Auxiliary differential equation
formulation: An efficient implementation of the per-
fectly matched layer. IEEE Microwave and Wireless
Components Lett 13: 69-71.
4. C Zhang, B Sun, H Yang, J Ma (2016) A non-split per-
fectly matched layer absorbing boundary condition
for the second-order wave equation modeling. Jour-
nal of Seismic Exploration 25: 1-9.
5. U Basu, AK Chopra (2004) Perfectly matched lay-
ers for transient elastodynamics of unbounded do-
mains. International Journal for Numerical Methods
in Engineering 59: 1039-1074.
6. DM Sullivan (2000) Electromagnetic simulation using
the FDTD method. IEEE Press, New York, 165.
7. ZA Chen, A Taflove, V Backman (2004) Photonic
nanojet enhancement of backscattering of light by
nanoparticles: A potential novel visible-light ultrami-
croscopy technique. Opt Express 12: 1214-1220.
8. Z Zhishen, H Yin, F Yuanhua, S Yuecheng, L Zhaohui
(2019) An ultra narrow photonic nanojet formed by
an engineered two-layer microcylinder of high re-
fractive-index materials. Opt Express 27: 9178-9188.
DOI: 10.35840/2631-5092/4524
Citation: Abramov A, Yue Y, Wang M (2020) Non-Split Perfectly Matched Layer Boundary Conditions for Numerical Solution of 2D
Maxwell Equations. Int J Opt Photonic Eng 5:024
ISSN: 2631-5092 | • Page 6 of 6 •
9. CY Liu (2013) Photonic nanojet enhancement of di-
electric microcylinders with metallic coating. J Opto-
electron Adv Mater 15: 150-154.
10. J Schaefer, S-C Lee, A Kienle (2012) Calculation of
the near fields for the scattering of electromagnetic
waves by multiple infinite cylinders at perpendicu-
lar incidence. J Quant Spectrosc Radiat Transf 113:
2113-2123.
11. A Abramov, C Ji, R Liu, A Kostikov (2016) Enhance-
ment of the electromagnetic waves intensity by scat-
tering on multiple infinite cylinders. Journal of Pho-
tonic Materials and Technology 2: 1-5.
12. A Abramov, A Kostikov (2017) Formation of whis-
pering gallery modes by scattering of an electromag-
netic plane wave by two cylinders. Phys Lett A 381:
1107-1110.
13. A Abramov, A Kostikov, R Liu, C Ji, X Li, et al. (2017)
Formation of the high electromagnetic waves inten-
sity areas by multiple cylinders scattering: whisper-
ing gallery modes and photonic nanojet. Journal of
Electromagn Waves and Applications 31: 820-827.
14. A Taflove, S Hagness (2005) Computational electro-
dynamics: The finite-difference time-domain meth-
od. (3rd edn), Artech House, Boston, MA, USA.