This article has been accepted for publication in a future issue of this journal, but has not been
fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JPHOT.2016.2625816, IEEE
Photonics Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) <
Efficient Frequency-Dependent Newmark-Beta-
FDTD Method for Periodic Grating Calculation
Sheng-Bing Shi, Student Member, IEEE, Wei Shao, Member, IEEE, Tu-Lu Liang, Li-Ye Xiao,
Xue-Song Yang, Member, IEEE, and Haiyan Ou
 FP-like resonant peak because of a phase reversal of the
Abstract—In this paper, the Newmark-Beta algorithm is
introduced into the finite-different time-domain (FDTD) method
in dispersion media, resulting in an unconditionally stable method
for periodic metallic grating analysis. The proposed method
eliminates the Courant-Friedrich-Levy constraint and improves
the simulation efficiency for multiscale problems. The dispersion
of the metal, caused by the evanescent waves propagating along
the interface between the metal and dielectric materials in the
visible and near infrared regions, is solved with a generalized
auxiliary differential equation (ADE) technique. The
extraordinary optical transmission through a periodic metallic
grating with different number and different size of the
perpendicular bump is also investigated. Compared with the
traditional ADE-FDTD method and ADE alternating-direction-
implicit FDTD method, the results from the proposed method
show its accuracy and efficiency.
Index Terms—extraordinary optical transmission (EOT),
metallic grating, Newmark-Beta algorithm, surface plasmons.
I. INTRODUCTION
T phenomenon of extraordinary optical transmission
HE
(EOT) through subwavelength metallic structures has
aroused great interest [1], [2] since the phenomenon in a thick
metal film perforated with a two-dimensional (2-D) array of
subwavelength holes was first observed in 1998 [3]. Many
investigations, such as one-dimensional (1-D) periodic arrays
of slits [4], [5] and 2-D periodic arrays of holes with different
shapes [6]-[8], have been discussed actively, not only for the
rich physical mechanisms, but also for their potential
applications in optics and optoelectronics [2]. Though the
physical mechanism under the EOT phenomenon remains
controversial [3], [9], [10], there are two kinds of transmission
resonance theory to explain this phenomenon, namely the
coupled surface plasmon polariton (SPP) resonance and the
Fabry-Pérot (FP)-like resonance. However, a third type of
resonance in periodic metallic grating structures referred to as
phase resonance has been identified and has received much
attention in recent years. It is characterized by the splitting of
Manuscript received September 30, 2016; revised October 18, 2016;
accepted November 1, 2016. This work was supported by the National Natural
Science Foundation of China (61331007, 61471105 and 61271027) and 973
Project (613272).
The authors are with the School of Physical Electronics, University of
Electronic Science and Technology of China, Chengdu, China (e-mail:
shengbing77@163.com; weishao@uestc.edu.cn; liangtul@163.com; liyexiao
@std.uestc.edu.cn; xsyang@uestc.edu.cn; ouhaiyan@uestc.edu.cn).
1943-0655 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
1
magnetic field in adjacent slits within each period.
For the analysis of periodic metallic gratings, the
finite-difference time-domain (FDTD) method is usually used
in simulation due to its easy time-frequency domain transform
characteristic. However, as the SPPs are highly localized along
the metal-dielectric interface, fine spatial mesh should be used
in this interface for high accuracy. This results in an extremely
small time step determined by the Courant-Friedrich-Levy
(CFL) constraint [11], causing a heavy computational burden.
In recent years, many unconditionally stable FDTD methods
have been presented to eliminate the CFL condition and
improve computational efficiency. Three typical
unconditionally stable methods, the alternating-direction
implicit (ADI) FDTD method [12]-[14], split-step (SS) FDTD
method [15]-[17] and locally one-dimensional (LOD) FDTD
method [18]-[20], only need to solve a tridiagonal matrix
equation during the time-marching calculation. However, the
absorbing and periodic boundary conditions in these methods
always break down the tridiagonal property of the matrix.
Moreover, their numerical dispersion will rapidly degrade as
the time step size increases [15], [21], [22]. The
Crank-Nicolson (CN) FDTD method is another unconditionally
stable time-marching method, in which a full time step size in
one marching step is used to solve the discretized Maxwell’s
equations [23], [24]. Besides, the weighted Laguerre
polynomial (WLP) FDTD method is an order-marching scheme
and does not discretize the time variable, leading to an
unconditionally stable scheme [25], [26]. Both CN-FDTD and
WLP-FDTD need to solve a large banded-sparse matrix
equation.
Recently, a new unconditionally stable FDTD method based
on the Newmark-Beta algorithm has been proposed for the
microwave device simulation [27]. This method, in which the
Newmark-Beta algorithm is utilized to discretize the temporal
variables and results in an implicit scheme, shows its high
efficiency for multiscale problems. So far some frequency-time
transform techniques, such as the recursive convolution (RC)
[28], Z transforms [29], shift operator (SO) [30] and auxiliary
differential equations (ADEs) [31], [32], are presented to deal
with the frequency-dependent dispersion media in FDTD
simulations. The ADE technique, which does not have to deal
with the convolution or the Fourier transform, has been widely
used for modeling dispersion materials due to its simplicity and
flexibility. In this work, the ADE technique is introduced into
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JPHOT.2016.2625816, IEEE
Photonics Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 2

the Newmark-Beta FDTD method for SPP simulations. At the  1   n 1   1   n 1

same time, the reverse Cuthill-Mckee (RCM) technique, which    J x   J xn     Jx
 2t 4  2  2t 4  (3d)
compresses the bandwidth of sparse matrices [33], is used to
1 1 1
perform the lower-upper (LU) decomposition efficiently. With   0p Ex   0p Ex   0p Ex
2 n 1 2 n 2 n 1

the ADE-Newmark-Beta-FDTD method, the periodic metallic 4 2 4

gratings with/without perpendicular bumps are investigated.  1   n 1   1   n 1
   J y   J yn     Jy
The numerical examples verify the efficiency and accuracy of  2 t 4  2  2t 4 
the proposed method. . (3e)
1 1 1
  0p E y   0p E y   0p E y
2 n 1 2 n 2 n 1

II. NUMERICAL FORMULATIONS 4 2 4

Using the central difference scheme to discretize the spatial
The frequency-dependent permittivity of the metal is
derivatives, (3a)-(3e) will be transformed into difference
expressed by the Drude model for the dielectric response of the
equations. Substituting (3d) into (3a) and (3e) into (3b), the
metal which is mainly governed by its free electron plasmon
explicit updating expressions of Exn 1 , E yn 1 can be obtained.
p 2
    1  (1) Then substituting these two expressions into (3c), we obtain the
   i 
implicit updating expression of H zn 1 , as shown in (4a), where
where ωp is the plasma frequency, γ is the absorption n 1
coefficient, and ω stands for the operation frequency. Exn 1 , E y are the known intermediate variables involving the
For simplicity, a 2-D TMx wave including Ex, Ey, and Hz nth and (n-1)th components, as shown in (4b) and (4c).
components is considered without loss of generality. The 2-D After the magnetic field is obtained by solving the matrix
Maxwell’s equations and auxiliary differential equations in the equation from (4a), the electric field components and the
dispersive metal can be written as current-densities can be calculated in an explicit way.
E H z Since the metallic grating is a periodic structure, the periodic
 x   Jx (2a)
t y boundary condition (PBC) has to be applied to the simulation. It
is assumed that the PBCs lie on the edges of jmin and jmax along
E y
H z
 
 Jy (2b) the y-direction and the period is p. According to the Floquet
t x theorem
H z Ex E y f ( x, y  p, t )  f ( x, y, t ) (5)
   (2c)
t y x we have
J x H z i  1 , j  1  H z i  1 , j  1 , H z i  1 , j  1  H z i  1 , j  1 (6)
  J x   0p2 Ex (2d) 2
min
2 2
max
2 2
max
2 2
min
2
t and the other electromagnetic field components outside the
J y computational region can be solved in a similar way.
  J y   0p2 E y (2e)
t It is worth noting that the equation in (4a) is a tridiagonal
where ε and ε0 are the electric permittivity of the medium and matrix equation which can be solved by the Thomas algorithm
free space, respectively, and μ is the magnetic permeability. efficiently. However, this property is destroyed by the PML and
Using the Newmark-Beta algorithm to discretize the temporal PBC boundary conditions, resulting in a banded-sparse matrix
derivatives in (2a)-(2e) [27], we obtain equation with a large bandwidth. Here, in order to solve this
matrix equation efficiently, the reverse Cuthill-Mckee (RCM)
 n 1  n 1 1  H zn 1 
Ex  Ex    J xn 1  preconditioning is applied to the lower-upper (LU)
2t 2t 4  y  decomposition to reduce the bandwidth of the coefficient
(3a)
1  H z
n
n  1  H zn 1
n 1 
matrix.
   Jx     Jx 
2  y  4  y  III. NUMERICAL RESULTS AND DISCUSSION
 n 1  n 1 1  H n 1 
E  E    z  J yn 1  With the proposed ADE-Newmark-Beta-FDTD method, the
2t 2t x
y y
4  transmission resonances of the periodic metallic grating
(3b)
1  H n  1  H n 1  structures with sub-wavelength slits are investigated in the
   z  J yn     z  J yn 1  visible and near infrared regions in this section.
4  x  4  x 
1  E n 1 E y 
n 1
 
H zn 1  H zn 1   x  
2t 2t 4  y x 
(3c)
1  E n E y  1  Exn 1 E y 
n n 1

  x      
2  y x  4  y x 

1943-0655 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JPHOT.2016.2625816, IEEE
Photonics Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 3

 1 1 
    Hz n 1
 Hz
n 1
Hz
n 1
 Hz
n 1

 0  2y 2x 2
1 3 1 1 3 1 1 1
 
2 i , j i , j i , j i , j
H n 1

1 2 2 2 2
 2 2 2 2
 2t 2  0 tp2  z 2  
1 1
i , j
2 2 2  t p  4y 2 4x 2 
   
0
 
 t 2  t  t 2  t
 E n 1  E n 1   Ex n 1  Ex n 1   Ex n 11  Ex n 11 
 x i  12 , j 1 x i 1, j
  i  2 , j 1 i , j
2   i  2 , j 1 i , j
2 
 
2
0    y  1 y 
 Hz
n 1

1  y  
1    (4a)
2  0 tp  E n 1  E n 1  2
1 1
2t i , j 2 n n
 4  E n 1  E n 1 
1  Ey
2 2
   Ey
y i 1, j  1 y i, j  1  i 1, j  i, j 
1
  y i 1, j  1 y i, j  1

t 2  t   2 2  
2 2
 
2 2

 x   x   x 

where
 Hz n
 Hz
n
Hz
n 1
 Hz
n 1


1 1 1 1 1 1 1 1
n 1 n 1  i , j i , j i , j i , j

Ex 1  E x i 1, j + 2 2 2 2
 2 2 2 2

i , j
2 2t 2
 2y 4y 
 
(4b)
 
 1 1 1 n 1  1  n 1   1 1 n 1 
  2  0 p E x   0 p2 Ex    Jx    2 Jx  J x i 1, j 
2 n n n
1 1 1 Jx 1 1
 2t 4 
i , j i , j i , j i , j i , j
  
2 2 4 2 2 2 2  2 4 2 

 t 
and
 Hz n
 Hz
n
Hz
n 1
 Hz
n 1


1 1 1 1 1 1 1 1
n 1 n 1  i , j i , j i , j i , j

Ey 1  Ey 1  2 2 2 2
 2 2 2 2

i, j 
2 2t i, j 
2  2x 4x 
 
. (4c)
 
 1 1 n 1 n 1  1  n 1  n  1 n 1 n 1 
  2  0 p E y   0 p2 E y    Jy    2 Jy  Jy 1 
2
1 1 1 Jy 1 1
 2   i, j 
2 4 i, j 
2  2t 4  i, j 
2 2 i, j 
2  i, j 
2 4 i, j  2 
 t 

is gold and the considered wavelength range is 700 nm to 2500

PBC
h = 100 nm
P = 700 nm
PML
PBC w = 1500 nm
Fig. 1. Schematic of a unit cell of the periodic metallic grating without bumps.
The geometric structure of a unit cell of the periodic metallic
grating without bumps is shown in Fig. 1. The calculated region
is truncated by the PML conditions on the left and right sides
and PBC conditions on the top and bottom sides. A Gaussian
pulse is used as the source excitation, which can be written as
F (t )  exp   t  t0  /  2 
2
(7)
 
with the maximum frequency fmax = 430 THz, τ = 1/(2fmax), and
t0 = 3τ. As the SPPs are highly localized along the
metal-dielectric interface, fine spatial cells are used near this
interface for high accuracy. The metal investigated in this paper
nm. Hence, its corresponding parameters are ωp = 1.37×1016
rad/s and γ = 4.08×1013 rad/s [34].
To simulate this model effectively, the graded cell size in the
y-direction is used. The minimum and maximum cell sizes
along the y-direction are 1 nm and 25 nm, respectively, and the
PML cell size along the x-direction is 20 nm. The total cell number in
y the computational domain is 111 × 68. The traditional
ADE-FDTD method, the ADE-ADI-FDTD method, and the
z x proposed ADE-Newmark-Beta- FDTD method are used for this
simulation, where ∆tFDTD = ∆min/2/c = 1.6667×10-3 fs is chosen
for FDTD according to the CFL constraint, ∆tADI-FDTD = 5∆tFDTD
= 8.3333×10-3 fs (CFLN = ∆tADI-FDTD/∆tFDTD = 5) and ∆tNewmark
= 100∆tFDTD = 1.6667×10-1 fs (CFLN = 100) are chosen for
ADI-FDTD and Newmark-Beta-FDTD, respectively. Fig. 2
shows the transmission and reflection spectrums of the periodic
metallic grating from these three methods. It is clear that the
results from FDTD, ADI-FDTD, and Newmark-Beta-FDTD
are in good agreement. In addition, the transmission spectrum
shows transmission peaks at the wavelengths of 1929 nm, 1287
nm, 974 nm and 778 nm, which correspond to the four
waveguide mode resonances [35], [36], labeled as N = 2, N = 3,
N = 4 and N = 5, respectively. The four waveguide resonances
are associated with different standing wave modes of the slit

1943-0655 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JPHOT.2016.2625816, IEEE
Photonics Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 4

which acts as the Fabry-Pérot (F-P) cavity [37]. Table I presents
the required computational resources for FDTD, ADI-FDTD
with CFLN = 5, and Newmark-Beta-FDTD with CFLN = 50,
80, 100, 120, and 150. Because the large time step in
ADI-FDTD leads to large dispersion errors [21], its CFLN is
chosen as 5 in this example. It is worth noting the late-time
calculation in ADI-FDTD will not be convergent if its CFLN is
larger than 5.
The mean absolute percentage error (MAPE) reveals the
computational accuracy of the four resonances wavelengths,
which can be written as
1 4 peak  peak-FDTD
MAPE=  
4 n 1
100% ,
peak-FDTD
where λpeak-FDTD is the transmission peaks from FDTD. To
ensure the accuracy, the results of Newmark-Beta-FDTD with
CFLN = 100 (MAPE < 1%) are chosen in this simulation. Due
to the storage of the sparse matrix and LU decomposition, the
memory requirement of Newmark-Beta-FDTD is larger than
ADI-FDTD and FDTD. However, with the much larger time
step beyond the CFL constraint, the CPU time of
Newmark-Beta-FDTD can be largely reduced. All calculations
in this paper are performed on an Intel i7-4790 3.60-GHz
machine with 16-GB RAM.
1.0 N = 5 N = 4 N=3 N=2
0.8
Transmission
PBC
w2
h2
h = 100 nm
h1 = 25 nm
w1 = 750 nm P = 700 nm
PML h1 = 25 nm PML
h2 y
w2
PBC w = 1500 nm z x
Fig. 3. Schematic of a unit cell of the periodic metallic grating with bumps.
Then the transmission and reflection spectrums of the
(8) gratings with perpendicular bumps are investigated. Two
grating structures with two bumps (h1 = 25 nm and h2 = 0 nm)
and with four bumps (h1 = 25 nm and h2 = 25 nm) are
considered, as shown in Fig. 3. To simulate the effect of SPPs
around the bumps accurately, graded grids are used both in the
x- and y-directions. The minimum cell size near the bumps is 1
nm × 0.1 nm. Fig. 4 shows the transmission spectrum with
ADE-Newmark-Beta-FDTD. It is observed that the
transmission peaks for odd modes (N = 3 and N = 5) in the F-P
cavity exhibit a blue shift, while the even modes (N = 2 and N =
4) show a red shift compared with the transmission peaks of the
grating without the bumps in Fig. 1. This transmission behavior
can be explained by the following localized waveguide
resonance mode [37]. The resonant wavelength of the F-P
mode in a smooth slit array can be obtained by 2kLFP+θ = 2Nπ,
where k = 2nπ/λ is the wave vector (λ is the wavelength of the
Nth order mode, and n is the effective refractive index of the
fundamental Bloch mode propagating in the slit), LFP is the

Reflection

0.6 length of the cavity, and θ is the total phase reflected at the ends
of the slits and a N-dependent value. The existence of
0.4 perpendicular bumps enlarges the effective length of F-P cavity
in even modes resulting in a movement to the greater resonance
0.2 FDTD wavelength. On the contrary, it decreases the effective length of
ADI-FDTD
Newmark-Beta-FDTD F-P cavity in odd modes. Hence, the resonance wavelength of
0.0 the odd modes becomes smaller. In addition, the grating with
800 1000 1200 1400 1600 1800 2000 2200 2400
Wavelength (nm) four bumps has a greater impact than that of two bumps on the
effective wavelength.
Fig. 2. Transmission and reflection spectrums of the grating without bumps
with FDTD, ADI-FDTD (CFLN = 5), and Newmark-Beta-FDTD (CFLN =
100).
1.0 N = 5 N = 4 N = 3 N=2

TABLE I 0.9
COMPARISON OF THE COMPUTATIONAL EFFORTS FOR THE THREE METHODS 0.8
Marching CPU time Memory 0.7
Transmission

Method CFLN MAPE

steps (s) (Mb) 0.6
FDTD 1 40000 - 698.76 2.33 0.5
ADI-FDTD 5 8000 0.03% 281.61 5.71 0.4
0.3 no bump
50 800 0.18% 133.69 35.27 two bumps
0.2 four bumps
80 500 0.49% 87.05 35.27 0.1
Newmark-
100 400 0.75% 69.31 35.27 0.0
Beta-FDTD 800 1000 1200 1400 1600 1800 2000 2200 2400
120 333 1.08% 59.44 35.27 Wavelength(nm)
150 267 1.69% 49.10 35.27
Fig. 4. Transmission spectrum of the grating with no bump, two bumps and four
bumps.

At the same time, the transmission and reflection spectrums

1943-0655 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JPHOT.2016.2625816, IEEE
Photonics Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 5

of the grating structure with four bumps from FDTD,

ADI-FDTD (CFLN = 5), and Newmark-Beta-FDTD (CFLN = 1.1
1100) are given in Fig. 5, It can be seen that the results from the 1.0 N = 5 N = 4 N = 3 N=2

three methods are in quite good agreement. Table II presents 0.9

the required computational resources for the three methods. 0.8
0.7

Transmission
Similarly, CFLN = 5 is chosen for ADI-FDTD to keep the
0.6
convergence of the time-domain calculation, while CFLN =
0.5
1100 (MAPE < 1%) of Newmark-Beta-FDTD is chosen to
0.4 no bump
guaranteed simulation precision. Since the time step of the 0.3 w2 = 100 nm
Newmark-Beta-FDTD is chosen 1100 times of the FDTD, its 0.2 w2 = 150 nm
CPU time can be reduced to about 1.14% of FDTD. 0.1 w2 = 200 nm
Nevertheless, the memory requirement of Newmark-Beta- 0.0
800 1000 1200 1400 1600 1800 2000 2200 2400
FDTD is about 4 times larger than that of FDTD.
Wavelength(nm)
Fig. 6. Transmission spectrum of the grating with different widths of the
bumps.
1.0

0.8 IV. CONCLUSION

Tramsmission

Reflection

This work introduces the ADE technique to the

0.6
unconditionally stable Newmark-Beta-FDTD method for
0.4
periodic metallic grating analysis. The numerical formulation
of the method is derived. With the preconditioning RCM
0.2 FDTD technique to the LU decomposition, which is executed only
ADI-FDTD once at the beginning of the calculation, the implicit updated
Newmark-Beta-FDTD
0.0 equation can be solved quickly. Compared with the traditional
800 1000 1200 1400 1600 1800 2000 2200 2400
FDTD and ADI-FDTD methods, the accuracy and efficiency of
Wavelength(nm)
the proposed method are verified with the numerical examples
Fig. 5. Transmission and reflection spectrums of the grating with four bumps of the periodic metallic grating with different structures. This
from FDTD, ADI-FDTD (CFLN = 5) and Newmark-Beta-FDTD (CFLN = method is very suitable for the problems involving multiscale
1100).
grid division, such as photonic gratings, photonic crystals,
TABLE II microwave devices and antennas. Further application of the
COMPARISON OF THE COMPUTATIONAL EFFORTS FOR THE THREE METHODS proposed method will be studied in our future work.
Marching CPU Memory
Method CFLN MAPE
step time (s) (Mb)
REFERENCES
FDTD 1 1000000 - 49877.94 33.13
[1] W. L. Barnes, A. Dereux, and T. W. Ebbesen, “Surface plasmon
ADI-FDTD 5 200000 0.02% 20739.32 41.58 subwavelength optics,” Nature, vol. 424, no. 6950, pp. 824-830, Aug.
2003.
800 1250 0.47% 761.20 129.41 [2] C. Genet and T. W. Ebbesen, “Light in tiny holes,” Nature, vol. 445, no.
900 1111 0.60% 679.09 129.41 7123, pp. 39-46, Jan. 2007.
Newmark- [3] T. W. Ebbesen, H. J. Lezec, H. F. Ghaemi, T. Thio, and P. A. Wolff,
1000 1000 0.74% 612.13 129.41 “Extraordinary optical transmission through sub-wavelength hole arrays,”
Beta-FDTD
1100 909 0.88% 567.94 129.41 Nature, vol. 391, no. 6668, pp. 667-669, Feb. 1998.
[4] J. A. Porto, F. J. Garcia-Vidal, and J. B. Pendry, “Transmission resonance
1200 833 1.06% 517.56 129.41 on metallic gratings with very narrow slits,” Phys. Rev. Lett., vol. 83, no.
14, pp. 2845-2848, Oct. 1999.
[5] X. Jiao, P. Wang, L. Tang, Y. Lu, Q. Li, D. Zhang, P. Yao, H. Ming, and J.
Finally, in order to understand the dependence of resonance Xie, “Fabry-Pérot-like phenomenon in the surface plasmons resonant
wavelength on the bump width, the structures in Fig. 3 (h1 = 25 transmission of metallic gratings with very narrow slits,” Appl. Phys. B,
vol. 80, no. 3, pp. 301-305, Mar. 2005.
nm and h2 = 25 nm) with different bump widths are investigated. [6] W. L. Barnes, W. A. Murray, J. Dintinger, E. Devaux, and T. W. Ebbesen,
Fig. 6 shows the transmission spectrum of the granting with “Surface plasmon polaritons and their role in the enhances transmission
four bumps from the Newmark-Beta-FDTD method. Similar to of light through periodic arrays of subwavelength holes in a metal film,”
Phys. Rev. Lett., vol. 92, no. 10, pp. 107401-1-107401-4, Mar. 2004.
Fig. 4, the transmission peaks for odd modes of the F-P cavity [7] H.J. Lezec and T. Thio, “Diffracted evanescent wave model for enhanced
(N = 3 and N = 5) exhibit a blue shift, while the even modes (N and suppressed optical transmission through subwavelength hole arrays,”
= 2 and N = 4) show a red shift compared with the transmission Opt. Express, vol. 12, no. 16, pp. 3629-3651, Aug. 2004.
[8] S. Wu, L. Zhou, Y. M. Wang, G. D. Wang, Q. J. Wang, C. P. Huang, and
peaks of the grating without bumps. Besides, the offset of the
Y. Y. Zhu, “Optical properties of a metal film perforated with coaxial
even modes is larger than that of the odd modes because the elliptical hole arrays,” Phys. Rev. E, vol. 81, no. 5, pp.
bumps locate at the place of the even-mode nodes. 057601-1-057601-4, May 2010.
[9] F. Medina, J. A. Ruiz-Cruz, F. Mesa, J. M. Rebollar, J. R. Montejo-Garai,
and R. Marques, “Experimental verification of extraordinary transmission

1943-0655 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.
This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/JPHOT.2016.2625816, IEEE
Photonics Journal
> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 6

without surface plasmons,” Appl. Phys. Lett., vol. 95, no. 7, pp. [32] O. P. Gandhi, B. Q. Gao, and J. Y. Chen, “A frequency-dependent
071102-1-071102-3, Aug. 2009. finite-difference time-domain formulation for general dispersive media,”
[10] Z. C. Ruan and M. Qiu, “Enhanced transmission through periodic arrays IEEE Trans. Microw. Theory Techn., vol. 41, no. 4, pp. 658-665, Apr.
of subwavelength holes: the role of localized waveguide resonances,” 1993.
Phys. Rev. Lett., vol. 96, no. 23, pp. 233901-1-233901-4, Jun. 2006. [33] E. Cuthill and J. McKee, “Reducing the bandwidth of sparse symmetric
[11] A. Taflove and S.C. Hagness, computational Electrodynamics: The matrices,” Proc., 24th Nat. Conf. ACM, pp. 157-172, 1981.
Finite-Difference Time-Domain Method. Norwood, MA: Artech House, [34] E. D. Palik. Handbook of optical constants in solids, Academic, 1982.
2000. [35] X. Zhai, J. Q. Liu, M. D. He, L. L. Wang, S. C. Wen, and D. Y. Fan,
[12] T. Namiki, “A new FDTD algorithm based on alternating-direction “Adjustable phase resonances in a compound metallic grating with
implicit method,” IEEE Trans. Microw. Theory Techn., vol. 47, no. 10, pp. perpendicular cuts,” Opt. Express, vol. 18, no. 7, pp. 6871-6876, Mar.
2003-2007, Oct. 1999. 2010.
[13] T. Namiki, “3-D ADI-FDTD method-unconditionally stable time-domain [36] Z. M. Liu, H. J. Li, S. X. Xie, H. Q. Xu, S. L. Fu, X. Zhou, and C. N. Wu,
algorithm for solving full vector Maxwell’s equations,” IEEE Trans. “Tunable phase resonances in a compound metallic grating with
Microw. Theory Techn., vol. 48, no. 10, pp. 1743-1748, Oct. 2000. perpendicular bumps and cuts,” Opt. Express, vol. 19, no. 5, pp.
[14] F. Zheng, Z. Chen, and J. Zhang, “Towards the development of a 4217-4222, Feb. 2011.
three-dimensional unconditionally stable finite-difference time-domain [37] A. P. Hibbins, M. J. Lockyear, and J. R. Sambles, “The resonant
method,” IEEE Trans. Microw. Theory Techn., vol. 48, no. 9, pp. electromagnetic fields of an array of metallic slits acting as Fabry-Pérot
1550-1558, Sep. 2000. cavities,” J. Appl. Phys., vol. 99, no. 12, pp.124903-1-124903-5, Jun.
[15] J. Lee and B. Fornberg, “A split step approach for the 3-D maxwell’s 2006.
equations,” J. Comput. Appl. Math., vol. 158, no. 5, pp. 485-505, Sep.
2003.
[16] J. Lee and B. Fornberg, “Some unconditionally stable time stepping
methods for the 3-D Maxwell’s equations,” J. Comput. Appl. Math., vol.
166, no. 2, pp. 497-523, Apr. 2004.
[17] L. Gao, B. Zhang, and D. Liang, “The splitting finite difference
time-domain methods for Maxwell’s equations in two dimensions,” J.
Comput. Appl. Math., vol. 205, no. 1, pp. 207-230, Aug. 2007.
[18] J. Shibayama, M. Muraki, J. Yamauchi, and H. Nakano, “Efficient
implicit FDTD algorithm based on locally one-dimensional scheme,”
Electro. Lett., vol. 41, no. 19, pp. 1046-1047, Sep. 2005.
[19] E. L. Tan, “Unconditionally stable LOD-FDTD method for 3-D
Maxwell’s equations,” IEEE Microw. Wireless Compon. Lett., vol. 17, no.
2, pp. 85-87, Feb. 2007.
[20] J. Shibayama, M. Muraki, R. Takahashi, J. Yamauchi, and H. Nakano,
“Performance evaluation of several implicit FDTD methods for optical
waveguide analyses,” Journal of Lightwave Technology, vol. 24, no. 6, pp.
2465-2472, Jun. 2006.
[21] F. Zheng and Z. Chen, “Numerical dispersion analysis of the
unconditionally stable 3-D ADI-FDTD method,” IEEE Trans. Microw.
Theory Techn., vol. 49, no. 5, pp. 1006-1009, May 2001.
[22] I. Ahmed, E. K. Chun, and E. P. Li, “Numerical dispersion analysis of the
unconditionally stable three-dimensional LOD-FDTD method,” IEEE
Trans. Antennas Propag., vol. 58, no. 12, pp. 3983-3989, Dec. 2010.
[23] G. Sun and C. W. Trueman, “Approximate Crank-Nicolson schemes for
the 2-D finite-difference time-domain method for TEz waves,” IEEE
Trans. Antennas Propag., vol. 52, no. 11, pp. 2963-2972, Nov. 2004.
[24] G. Sun and C. W. Truneman, “Efficient implementations of the
Crank-Nicolson scheme for the finite-difference time-domain method,”
IEEE Trans. Microw. Theory Techn., vol. 54, no. 5, pp. 2275-2284, May
2006.
[25] Y. S. Chung, T. K. Sarkar, B. H. Jung, and M. Salazar-Palma, “An
unconditionally stable scheme for the finite-difference time-domain
method,” IEEE Trans. Microw. Theory Techn., vol. 51, no. 3, pp. 697-704,
Mar. 2003.
[26] M. Ha, K. Srinivasan, and M. Swaminathan, “Transient chip-package
co-simulation using the Laguerre-FDTD scheme,” IEEE Trans. Adv.
Packag., vol. 32, no. 4, pp. 816-830, Nov. 2009.
[27] S. B. Shi, W. Shao, X. K. Wei, X. S. Yang and B. Z. Wang, “A new
unconditionally stable FDTD method based on the Newmark-Beta
algorithm,” IEEE Microw. Theory Techn., vol. 65, no. 99, accepted for
publication.
[28] R. Luebbers, F. P. Hunsberger, K. S. Kunz, R. B. Standler, M. Schneider,
“A frequency-denpendent finite-difference time-domain formulation for
dispersive materials,” IEEE Trans. Electromagn. Compat., vol. 32, no. 3,
pp. 222-227, Aug. 1990.
[29] D. M. Sullivan, “Frequency-dependent FDTD methods using z
transforms,” IEEE Trans. Antennas Propaga., vol. 40, no. 10, pp.
1223-1230, Oct. 1992.
[30] X. L. Duan, H. W. Yang, X. K. Kong, and H. Liu, “Analysis on the
calculation of pladma medium with parallel SO-FDTD method,” ETRI,
vol. 31, no. 4, pp. 387-392, Aug. 2009.
[31] L. J. Nichisch, P. M. Franke, “Finite-difference time-domain solution of
maxwell’s equations for the dispersive lonosphere,” IEEE Antennas
Propagat. Mag., vol. 34, no. 5, pp. 33-39, Oct. 1992.

1943-0655 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See
http://www.ieee.org/publications_standards/publications/rights/index.html for more information.