1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
|
\input{newcommands}
\subsection{Open boundary condition for electromagnetic waves}
For the TE case, the original Berenger's Perfectly Matched Layer (PML) writes
% PML
\begin{eqnarray}
\varepsilon _{0}\frac{\partial E_{x}}{\partial t}+\sigma _{y}E_{x} = & \frac{\partial H_{z}}{\partial y}\label{PML_def_1} \\
\varepsilon _{0}\frac{\partial E_{y}}{\partial t}+\sigma _{x}E_{y} = & -\frac{\partial H_{z}}{\partial x}\label{PML_def_2} \\
\mu _{0}\frac{\partial H_{zx}}{\partial t}+\sigma ^{*}_{x}H_{zx} = & -\frac{\partial E_{y}}{\partial x}\label{PML_def_3} \\
\mu _{0}\frac{\partial H_{zy}}{\partial t}+\sigma ^{*}_{y}H_{zy} = & \frac{\partial E_{x}}{\partial y}\label{PML_def_4} \\
H_{z} = & H_{zx}+H_{zy}\label{PML_def_5}
\end{eqnarray}
This can be generalized to
% APML
\begin{eqnarray}
\varepsilon _{0}\frac{\partial E_{x}}{\partial t}+\sigma _{y}E_{x} = & \frac{c_{y}}{c}\frac{\partial H_{z}}{\partial y}+\overline{\sigma }_{y}H_{z}\label{APML_def_1} \\
\varepsilon _{0}\frac{\partial E_{y}}{\partial t}+\sigma _{x}E_{y} = & -\frac{c_{x}}{c}\frac{\partial H_{z}}{\partial x}+\overline{\sigma }_{x}H_{z}\label{APML_def_2} \\
\mu _{0}\frac{\partial H_{zx}}{\partial t}+\sigma ^{*}_{x}H_{zx} = & -\frac{c^{*}_{x}}{c}\frac{\partial E_{y}}{\partial x}+\overline{\sigma }_{x}^{*}E_{y}\label{APML_def_3} \\
\mu _{0}\frac{\partial H_{zy}}{\partial t}+\sigma ^{*}_{y}H_{zy} = & \frac{c^{*}_{y}}{c}\frac{\partial E_{x}}{\partial y}+\overline{\sigma }_{y}^{*}E_{x}\label{APML_def_4} \\
H_{z} = & H_{zx}+H_{zy}\label{APML_def_5}
\end{eqnarray}
For $c_{x}=c_{y}=c^{*}_{x}=c^{*}_{y}=c$ and $\overline{\sigma }_{x}=\overline{\sigma }_{y}=\overline{\sigma }_{x}^{*}=\overline{\sigma }_{y}^{*}=0$,
this system reduces to the Berenger PML medium, while adding the additional
constraint $\sigma _{x}=\sigma _{y}=\sigma _{x}^{*}=\sigma _{y}^{*}=0$
leads to the system of Maxwell equations in vacuum.
\subsubsection{\label{Sec:analytic theory, propa plane wave}Propagation of a Plane Wave in an APML Medium}
We consider a plane wave of magnitude ($ E_{0},H_{zx0},H_{zy0} $)
and pulsation $\omega$ propagating in the APML medium with an
angle $\varphi$ relative to the x axis
\begin{eqnarray}
E_{x} = & -E_{0}\sin \varphi e^{i\omega \left( t-\alpha x-\beta y\right) }\label{Plane_wave_APML_def_1} \\
E_{y} = & E_{0}\cos \varphi e^{i\omega \left( t-\alpha x-\beta y\right) }\label{Plane_wave_APML_def_2} \\
H_{zx} = & H_{zx0}e^{i\omega \left( t-\alpha x-\beta y\right) }\label{Plane_wave_AMPL_def_3} \\
H_{zy} = & H_{zy0}e^{i\omega \left( t-\alpha x-\beta y\right) }\label{Plane_wave_APML_def_4}
\end{eqnarray}
where $\alpha$ and$\beta$ are two complex constants to
be determined.
Introducing (\ref{Plane_wave_APML_def_1}), (\ref{Plane_wave_APML_def_2}),
(\ref{Plane_wave_AMPL_def_3}) and (\ref{Plane_wave_APML_def_4})
into (\ref{APML_def_1}), (\ref{APML_def_2}), (\ref{APML_def_3})
and (\ref{APML_def_4}) gives
\begin{eqnarray}
\varepsilon _{0}E_{0}\sin \varphi -i\frac{\sigma _{y}}{\omega }E_{0}\sin \varphi = & \beta \frac{c_{y}}{c}\left( H_{zx0}+H_{zy0}\right) +i\frac{\overline{\sigma }_{y}}{\omega }\left( H_{zx0}+H_{zy0}\right) \label{Plane_wave_APML_1_1} \\
\varepsilon _{0}E_{0}\cos \varphi -i\frac{\sigma _{x}}{\omega }E_{0}\cos \varphi = & \alpha \frac{c_{x}}{c}\left( H_{zx0}+H_{zy0}\right) -i\frac{\overline{\sigma }_{x}}{\omega }\left( H_{zx0}+H_{zy0}\right) \label{Plane_wave_APML_1_2} \\
\mu _{0}H_{zx0}-i\frac{\sigma ^{*}_{x}}{\omega }H_{zx0} = & \alpha \frac{c^{*}_{x}}{c}E_{0}\cos \varphi -i\frac{\overline{\sigma }^{*}_{x}}{\omega }E_{0}\cos \varphi \label{Plane_wave_APML_1_3} \\
\mu _{0}H_{zy0}-i\frac{\sigma ^{*}_{y}}{\omega }H_{zy0} = & \beta \frac{c^{*}_{y}}{c}E_{0}\sin \varphi +i\frac{\overline{\sigma }^{*}_{y}}{\omega }E_{0}\sin \varphi \label{Plane_wave_APML_1_4}
\end{eqnarray}
Defining $Z=E_{0}/\left( H_{zx0}+H_{zy0}\right)$ and using (\ref{Plane_wave_APML_1_1})
and (\ref{Plane_wave_APML_1_2}), we get
\begin{eqnarray}
\beta = & \left[ Z\left( \varepsilon _{0}-i\frac{\sigma _{y}}{\omega }\right) \sin \varphi -i\frac{\overline{\sigma }_{y}}{\omega }\right] \frac{c}{c_{y}}\label{Plane_wave_APML_beta_of_g} \\
\alpha = & \left[ Z\left( \varepsilon _{0}-i\frac{\sigma _{x}}{\omega }\right) \cos \varphi +i\frac{\overline{\sigma }_{x}}{\omega }\right] \frac{c}{c_{x}}\label{Plane_wave_APML_alpha_of_g}
\end{eqnarray}
Adding $H_{zx0}$ and $H_{zy0}$ from (\ref{Plane_wave_APML_1_3})
and (\ref{Plane_wave_APML_1_4}) and substituting the expressions
for $\alpha$ and $\beta$ from (\ref{Plane_wave_APML_beta_of_g})
and (\ref{Plane_wave_APML_alpha_of_g}) yields
\begin{eqnarray}
\frac{1}{Z} = & \frac{Z\left( \varepsilon _{0}-i\frac{\sigma _{x}}{\omega }\right) \cos \varphi \frac{c^{*}_{x}}{c_{x}}+i\frac{\overline{\sigma }_{x}}{\omega }\frac{c^{*}_{x}}{c_{x}}-i\frac{\overline{\sigma }^{*}_{x}}{\omega }}{\mu _{0}-i\frac{\sigma ^{*}_{x}}{\omega }}\cos \varphi \nonumber \\
+ & \frac{Z\left( \varepsilon _{0}-i\frac{\sigma _{y}}{\omega }\right) \sin \varphi \frac{c^{*}_{y}}{c_{y}}-i\frac{\overline{\sigma }_{y}}{\omega }\frac{c^{*}_{y}}{c_{y}}+i\frac{\overline{\sigma }^{*}_{y}}{\omega }}{\mu _{0}-i\frac{\sigma ^{*}_{y}}{\omega }}\sin \varphi
\end{eqnarray}
If $c_{x}=c^{*}_{x}$, $c_{y}=c^{*}_{y}$, $\overline{\sigma }_{x}=\overline{\sigma }^{*}_{x}$, $\overline{\sigma }_{y}=\overline{\sigma }^{*}_{y}$, $\frac{\sigma _{x}}{\varepsilon _{0}}=\frac{\sigma ^{*}_{x}}{\mu _{0}}$ and $\frac{\sigma _{y}}{\varepsilon _{0}}=\frac{\sigma ^{*}_{y}}{\mu _{0}}$ then
\begin{eqnarray}
Z = & \pm \sqrt{\frac{\mu _{0}}{\varepsilon _{0}}}\label{APML_impedance}
\end{eqnarray}
which is the impedance of vacuum. Hence, like the PML, given some
restrictions on the parameters, the APML does not generate any reflection
at any angle and any frequency. As for the PML, this property is not
retained after discretization, as shown subsequently in this paper.
Calling $\psi$ any component of the field and $\psi _{0}$
its magnitude, we get from (\ref{Plane_wave_APML_def_1}), (\ref{Plane_wave_APML_beta_of_g}),
(\ref{Plane_wave_APML_alpha_of_g}) and (\ref{APML_impedance}) that
\begin{equation}
\label{Plane_wave_absorption}
\psi =\psi _{0}e^{i\omega \left( t\mp x\cos \varphi /c_{x}\mp y\sin \varphi /c_{y}\right) }e^{-\left( \pm \frac{\sigma _{x}\cos \varphi }{\varepsilon _{0}c_{x}}+\overline{\sigma }_{x}\frac{c}{c_{x}}\right) x}e^{-\left( \pm \frac{\sigma _{y}\sin \varphi }{\varepsilon _{0}c_{y}}+\overline{\sigma }_{y}\frac{c}{c_{y}}\right) y}
\end{equation}
We assume that we have an APML layer of thickness $\delta$ (measured
along $x$) and that $\sigma _{y}=\overline{\sigma }_{y}=0$
and $c_{y}=c.$ Using (\ref{Plane_wave_absorption}), we determine
that the coefficient of reflection given by this layer is
\begin{eqnarray}
R_{APML}\left( \theta \right) = & e^{-\left( \sigma _{x}\cos \varphi /\varepsilon _{0}c_{x}+\overline{\sigma }_{x}c/c_{x}\right) \delta }e^{-\left( \sigma _{x}\cos \varphi /\varepsilon _{0}c_{x}-\overline{\sigma }_{x}c/c_{x}\right) \delta }\nonumber \\
= & e^{-2\left( \sigma _{x}\cos \varphi /\varepsilon _{0}c_{x}\right) \delta }
\end{eqnarray}
which happens to be the same as the PML theoretical coefficient of
reflection if we assume $c_{x}=c$. Hence, it follows that for
the purpose of wave absorption, the term $\overline{\sigma }_{x}$
seems to be of no interest. However, although this conclusion is true
at the infinitesimal limit, it does not hold for the discretized counterpart.
\subsubsection{Discretization}
%
\begin{subequations}
\begin{align}
\frac{E_x|^{n+1}_{j+1/2,k,l}-E_x|^{n}_{j+1/2,k,l}}{\Delta t} + \sigma_y \frac{E_x|^{n+1}_{j+1/2,k,l}+E_x|^{n}_{j+1/2,k,l}}{2} = & \frac{H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j+1/2,k-1/2,l}}{\Delta y} \\
%
\frac{E_y|^{n+1}_{j,k+1/2,l}-E_y|^{n}_{j,k+1/2,l}}{\Delta t} + \sigma_x \frac{E_y|^{n+1}_{j,k+1/2,l}+E_y|^{n}_{j,k+1/2,l}}{2} = & - \frac{H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j-1/2,k+1/2,l}}{\Delta x} \\
%
\frac{H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l}-H_{zx}|^{n}_{j+1/2,k+1/2,l}}{\Delta t} + \sigma^*_x \frac{H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l}+H_{zx}|^{n}_{j+1/2,k+1/2,l}}{2} = & - \frac{E_y|^{n+1}_{j+1,k+1/2,l}-E_y|^{n+1}_{j,k+1/2,l}}{\Delta x} \\
%
\frac{H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l}-H_{zy}|^{n}_{j+1/2,k+1/2,l}}{\Delta t} + \sigma^*_y \frac{H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l}+H_{zy}|^{n}_{j+1/2,k+1/2,l}}{2} = & \frac{E_x|^{n+1}_{j+1/2,k+1,l}-E_x|^{n+1}_{j+1/2,k,l}}{\Delta y} \\
%
H_z = & H_{zx}+H_{zy}
\end{align}
\end{subequations}
%
\begin{subequations}
\begin{align}
E_x|^{n+1}_{j+1/2,k,l} = & \left(\frac{1-\sigma_y \Delta t/2}{1+\sigma_y \Delta t/2}\right) E_x|^{n}_{j+1/2,k,l} + \frac{\Delta t/\Delta y}{1+\sigma_y \Delta t/2} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j+1/2,k-1/2,l}\right) \\
%
E_y|^{n+1}_{j,k+1/2,l} = & \left(\frac{1-\sigma_x \Delta t/2}{1+\sigma_x \Delta t/2}\right) E_y|^{n}_{j,k+1/2,l} - \frac{\Delta t/\Delta x}{1+\sigma_x \Delta t/2} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j-1/2,k+1/2,l}\right) \\
%
H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l} = & \left(\frac{1-\sigma^*_x \Delta t/2}{1+\sigma^*_x \Delta t/2}\right) H_{zx}|^{n}_{j+1/2,k+1/2,l} - \frac{\Delta t/\Delta x}{1+\sigma^*_x \Delta t/2} \left(E_y|^{n+1}_{j+1,k+1/2,l}-E_y|^{n+1}_{j,k+1/2,l}\right) \\
%
H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l} = & \left(\frac{1-\sigma^*_y \Delta t/2}{1+\sigma^*_y \Delta t/2}\right) H_{zy}|^{n}_{j+1/2,k+1/2,l} + \frac{\Delta t/\Delta y}{1+\sigma^*_y \Delta t/2} \left(E_x|^{n+1}_{j+1/2,k+1,l}-E_x|^{n+1}_{j+1/2,k,l}\right) \\
%
H_z = & H_{zx}+H_{zy}
\end{align}
\end{subequations}
%
\begin{subequations}
\begin{align}
E_x|^{n+1}_{j+1/2,k,l} = & e^{-\sigma_y\Delta t} E_x|^{n}_{j+1/2,k,l} + \frac{1-e^{-\sigma_y\Delta t}}{\sigma_y \Delta y} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j+1/2,k-1/2,l}\right) \\
%
E_y|^{n+1}_{j,k+1/2,l} = & e^{-\sigma_x\Delta t} E_y|^{n}_{j,k+1/2,l} - \frac{1-e^{-\sigma_x\Delta t}}{\sigma_x \Delta x} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j-1/2,k+1/2,l}\right) \\
%
H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l} = & e^{-\sigma^*_x\Delta t} H_{zx}|^{n}_{j+1/2,k+1/2,l} - \frac{1-e^{-\sigma^*_x\Delta t}}{\sigma^*_x \Delta x} \left(E_y|^{n+1}_{j+1,k+1/2,l}-E_y|^{n+1}_{j,k+1/2,l}\right) \\
%
H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l} = & e^{-\sigma^*_y\Delta t} H_{zy}|^{n}_{j+1/2,k+1/2,l} + \frac{1-e^{-\sigma^*_y\Delta t}}{\sigma^*_y \Delta y} \left(E_x|^{n+1}_{j+1/2,k+1,l}-E_x|^{n+1}_{j+1/2,k,l}\right) \\
%
H_z = & H_{zx}+H_{zy}
\end{align}
\end{subequations}
%
\begin{subequations}
\begin{align}
E_x|^{n+1}_{j+1/2,k,l} = & e^{-\sigma_y\Delta t} E_x|^{n}_{j+1/2,k,l} + \frac{1-e^{-\sigma_y\Delta t}}{\sigma_y \Delta y}\frac{c_y}{c} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j+1/2,k-1/2,l}\right) \\
%
E_y|^{n+1}_{j,k+1/2,l} = & e^{-\sigma_x\Delta t} E_y|^{n}_{j,k+1/2,l} - \frac{1-e^{-\sigma_x\Delta t}}{\sigma_x \Delta x}\frac{c_x}{c} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j-1/2,k+1/2,l}\right) \\
%
H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l} = & e^{-\sigma^*_x\Delta t} H_{zx}|^{n}_{j+1/2,k+1/2,l} - \frac{1-e^{-\sigma^*_x\Delta t}}{\sigma^*_x \Delta x}\frac{c^*_x}{c} \left(E_y|^{n+1}_{j+1,k+1/2,l}-E_y|^{n+1}_{j,k+1/2,l}\right) \\
%
H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l} = & e^{-\sigma^*_y\Delta t} H_{zy}|^{n}_{j+1/2,k+1/2,l} + \frac{1-e^{-\sigma^*_y\Delta t}}{\sigma^*_y \Delta y}\frac{c^*_y}{c} \left(E_x|^{n+1}_{j+1/2,k+1,l}-E_x|^{n+1}_{j+1/2,k,l}\right) \\
%
H_z = & H_{zx}+H_{zy}
\end{align}
\end{subequations}
%
\begin{subequations}
\begin{align}
c_x = & c e^{-\sigma_x\Delta t} \frac{\sigma_x \Delta x}{1-e^{-\sigma_x\Delta t}} \\
c_y = & c e^{-\sigma_y\Delta t} \frac{\sigma_y \Delta y}{1-e^{-\sigma_y\Delta t}} \\
c^*_x = & c e^{-\sigma^*_x\Delta t} \frac{\sigma^*_x \Delta x}{1-e^{-\sigma^*_x\Delta t}} \\
c^*_y = & c e^{-\sigma^*_y\Delta t} \frac{\sigma^*_y \Delta y}{1-e^{-\sigma^*_y\Delta t}}
\end{align}
\end{subequations}
%
\begin{subequations}
\begin{align}
E_x|^{n+1}_{j+1/2,k,l} = & e^{-\sigma_y\Delta t} \left[ E_x|^{n}_{j+1/2,k,l} + \frac{\Delta t}{\Delta y} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j+1/2,k-1/2,l}\right) \right] \\
%
E_y|^{n+1}_{j,k+1/2,l} = & e^{-\sigma_x\Delta t} \left[ E_y|^{n}_{j,k+1/2,l} - \frac{\Delta t}{\Delta x} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j-1/2,k+1/2,l}\right) \right] \\
%
H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l} = & e^{-\sigma^*_x\Delta t} \left[ H_{zx}|^{n}_{j+1/2,k+1/2,l} - \frac{\Delta t}{\Delta x} \left(E_y|^{n+1}_{j+1,k+1/2,l}-E_y|^{n+1}_{j,k+1/2,l}\right) \right] \\
%
H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l} = & e^{-\sigma^*_y\Delta t} \left[ H_{zy}|^{n}_{j+1/2,k+1/2,l} + \frac{\Delta t}{\Delta y} \left(E_x|^{n+1}_{j+1/2,k+1,l}-E_x|^{n+1}_{j+1/2,k,l}\right) \right] \\
%
H_z = & H_{zx}+H_{zy}
\end{align}
\end{subequations}
%
\begin{subequations}
\begin{align}
E_x|^{n+1}_{j+1/2,k,l} = & E_x|^{n}_{j+1/2,k,l} + \frac{\Delta t}{\Delta y} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j+1/2,k-1/2,l}\right) \\
%
E_y|^{n+1}_{j,k+1/2,l} = & E_y|^{n}_{j,k+1/2,l} - \frac{\Delta t}{\Delta x} \left(H_z|^{n+1/2}_{j+1/2,k+1/2,l}-H_z|^{n+1/2}_{j-1/2,k+1/2,l}\right) \\
%
H_{zx}|^{n+3/2}_{j+1/2,k+1/2,l} = & H_{zx}|^{n}_{j+1/2,k+1/2,l} - \frac{\Delta t}{\Delta x} \left(E_y|^{n+1}_{j+1,k+1/2,l}-E_y|^{n+1}_{j,k+1/2,l}\right) \\
%
H_{zy}|^{n+3/2}_{j+1/2,k+1/2,l} = & H_{zy}|^{n}_{j+1/2,k+1/2,l} + \frac{\Delta t}{\Delta y} \left(E_x|^{n+1}_{j+1/2,k+1,l}-E_x|^{n+1}_{j+1/2,k,l}\right) \\
%
H_z = & H_{zx}+H_{zy}
\end{align}
\end{subequations}
|
