You are on page 1of 26

Journal of Scientific Computing, Vol. 28, Nos.

2/3, September 2006 (© 2006)


DOI: 10.1007/s10915-006-9092-x

Simulating Plasma Microwave Diagnostics


J. S. Shang1

Received January 4, 2005; accepted (in revised form) April 26, 2005; Published online March 13, 2006

Computational simulation of plasma diagnostics via microwave absorption has


been successfully accomplished. This simulation capability is developed from
solutions to a combination of the three-dimensional Maxwell equations and
the generalized Ohm’s law in the time domain. As the simulation procedure
developed, numerical results were obtained for a range of plasma transport
properties including electrical conductivity, permittivity, and plasma frequency.
The present results reveal the wave reflection at the media interface and sub-
stantial distortion of the electromagnetic field within a thin plasma sheet from
a guided microwave. The present numerical simulation also accurately predicts
the microwave blackout phenomenon as the wave propagates through a thick
plasma sheet. The diffractions and refractions occurring at antenna apertures
and passing through a plasma column are captured numerically. Finally, the
numerical simulation has successfully duplicated a plasma diagnostic experi-
ment in a hypersonic magneto-hydrodynamic channel.

KEY WORDS: Computing simulation; microwave; plasma; attenuation.

NOMENCLATURE

B: Magnetic flux density;


c: Speed of light;
D: Electric displacement;
E: Electric field intensity;
H: Magnetic field strength;
J: Electric current density;
ε: Electric permittivity;
µ: Magnetic permeability;
ρ: Surface charge density;

1 Mechanical and Material Engineering Department, Wright State University, Dayton, OH


45435, USA. E-mail: jshang@cs.wright.edu

507
0885-7474/06/0900-0507/0 © 2006 Springer Science+Business Media, Inc.
508 Shang

σ: Electric conductivity;
: Frequency.

1. INTRODUCTION
Magneto-fluid-dynamics interaction has recently reemerged as one of the
few last frontiers for fluid dynamic research [1]. Magneto-fluid-dynamics
interactions have been widely applied and have demonstrated an impres-
sive potential for flow control [2–5]; MHD scramjet bypass engines [6],
innovative radiatively driven hypersonic wind tunnels [7] and combustion
or ignition enhancement [8]. To accurately assess the relative magnitude of
electromagnetic and aerodynamic forces in an interaction, an accurate evalu-
ation of the plasma transport properties, such as the charge particle number
density, temperatures, and electric conductivity, are required.
A widely used non-intrusive plasma diagnostic tool is microwave
probing [9–11]. The microwave system is adopted both for plasma diag-
nostics and deep-space communication. For plasma diagnostics, the num-
ber density of the charge particles and its collision frequency with the
neutral particles are measured based on the microwave attenuation phe-
nomenon [12–15]. This unique microwave behavior in weakly ionized air is
also known for the well-known communication blackout phenomenon in
the reentry phase, either for an aerospace vehicle or for an inter-planetary
voyage. Communication blackout is the consequence of an incident micro-
wave propagating at a frequency equal to or higher than the frequency
of plasma. When the two frequencies are equal, the propagating wave
becomes evanescent and the transmission of the electromagnetic energy
ceases. When the transmission bandwidth is greater than that of the
plasma, the microwave will attenuate as it propagates through the plasma
[16, 17]. The dissipated energy along the wave path is proportional to the
electrical conductivity of the medium.
An electromagnetic wave propagating in an electrically conducting
medium depends strongly on its electrical conductivity and transmission
frequency. The relative magnitude of σ and the product of wave frequency
and electrical permittivity ωε will dominate the behavior of the propa-
gating microwave [17]. Additional complication arises when a microwave
impinges on a boundary of two media. At the interface, a part of the
incident wave is reflected and another part is transmitted into the sec-
ond medium. The intrinsic impedance of the media ultimately controls the
behavior of this electromagnetic field at the interface [17]. In all circum-
stances, there will be a reflected wave from the interface, except when the
two media have the identical impedance, and then the incident wave will
transmit uninterrupted.
Simulating Plasma Microwave Diagnostics 509

In most microwave probing, the wave produced by the generator is


directed by a waveguide and transmitted from an antenna [9, 10]. The
widely used antennas are usually a pair of pyramidal horns whose main
function is to collimate the wave and reduce diffraction. Additional com-
plications arise when a microwave impinges on a boundary of two media.
The portion of the transmitted wave is governed by the continuity of
the tangential components of electric and magnetic field. For this reason,
some fundamental and data accuracy uncertainty for plasma diagnostics
remains. The evidence is clearly demonstrated by independent measure-
ments using Langmuir probe [18] and microwave absorption techniques
[14]. The measurement discrepancy in practical applications can be as
high as one order-of-magnitude. Even for a simple glow discharge simu-
lation, the simplifying assumptions of a weakly ionized gas are seriously
challenged [19]. It is therefore, natural to explore numerical simulation to
obtain a better understanding for the basics of microwave diagnostics.
Simulating transverse electromagnetic wave propagation in plasma or
partially ionized gas requires an interdisciplinary computational capabil-
ity. The transport properties of plasma governing the wave motion must
be accurately determined to describe the pertaining physics. Meanwhile the
numerical procedure not only should resolve the wave motion with low dis-
persive and dissipative errors, but also needs to capture the refraction and
diffraction at the media interface, and finally it needs to suppress spurious
reflections at the far field boundary of the computational domain. Research
in high-resolution algorithm research has met some of these challenges. In
particular, the compact-differencing schemes have provided an option in
describing the high-frequency wave motion with a lower truncation error
[20, 21]. In fact these spectral-like methods have demonstrated their ability
to process data in dynamic range of five orders of magnitude with a sparse
grid-point density of 8.1 per wavelength, which constitutes a factor of 3.125
times reduction from the conventional second-order methods [22, 23]. For
three-dimensional simulation, the computing resource saving is 30.15-fold.
Now it is possible to obtain solutions to problems that just a few years ago
defied theoretical or experimental analyses.
The challenge to an electromagnetic radiating simulation also resides in
the implementation of the far-field boundary conditions. This is the inher-
ent difficulty for computational electromagnetics, because the initial/bound-
ary value computations must be terminated in a finite domain [24]. The
most popular approaches in the computational electromagnetics commu-
nity are the absorbing boundary condition or the perfectly matched-layer
scheme [25, 26]. The latter is actually derived from the characteristics or the
domain of dependence of the hyperbolic partial differential equation sys-
tem. The detailed eigenvector and eigenvalue analyses of the characteristic
510 Shang

formulation for the three-dimensional Maxwell equations have been


performed [27–29]. In this formulation, the flux vector of the governing
equations is split according to the signs of the eigenvalue to honor the
domain of dependence. The flux vector splitting formulation is utilized in
the present investigation.
The present effort will detail the systematic development of a numer-
ical simulation capability for electromagnetic wave radiation and propaga-
tion in a weakly ionized gas. Specifically, the microwave beam path from
a pyramidal horn antenna across the interface of air and weakly ionized
gas will be formulated, simulated, and validated by comparing with avail-
able theoretical and experimental results.

2. GOVERNING EQUATIONS
The governing equations for the present effort are built around the
three-dimensional Maxwell equations in the time domain. The closure of
this partial differential equation system requires an additional constitutive
relationship to describe the electrical conduction current, J . The rate of
change for the electrical charge density is derived from the generalized
Ohm’s law [30].

∂B/∂t + ∇ × E = 0, (2.1)
∂D/∂t − ∇ × H = −J, (2.2)
∇ ·B = 0, (2.3)
∇ ·D = ρe , (2.4)
(1/eh )∂J /∂t − σ E + J = 0, (2.5)

where eh is the average collision frequency between electrons and heavy
particles and σ is the electrical conductivity, σ = ne e2 /me eh . ne and me
are the electron number density and mass respectively, and the e is the
electron charge.
For the present investigation, the two Gauss’s laws, Eqs. (2.3) and
(2.4), are automatically satisfied, because the incident microwave is a lin-
ear harmonic transverse electric wave (TE), and the net charged particle
number density ρe , vanishes in the plasma. The global electrically neutral
property of the plasma ensures the second constraint, Eq. (2.4), is satis-
fied. The two divergence equations are therefore, eliminated from the pres-
ent solution scheme.
The governing equations constitute a hyperbolic partial differential
equation system that can be easily cast into the flux vector form with a
generalized curvilinear coordinates [22, 27]:
Simulating Plasma Microwave Diagnostics 511

∂U/∂t + ∂Fε /∂ξ + ∂Fη /∂η + ∂Fζ /∂ζ = −J, (2.6)

where ξx , ηy , ζz , etc, are the metrics of the coordinate transformation from


Cartesian (x, y, z) to a generalized curvilinear, body-oriented coordinate
system (ξ, η, ζ ). By a metric identity, the flux vectors in the transformed
space become.

Fξ = Fξ (ξx Fx , ξy Fy , ξz Fz ),
Fη = Fη (ηx Fx , ηy Fy , ηz Fz ), (2.7)
Fζ = Fζ (ζx Fx , ζy Fy , ζz Fz ),

where the functions Fx , Fy , and Fz contained in the flux vectors Fξ , Fη ,


and Fζ of the transformed space are their counterpart in Cartesian frame
of reference [27];

Fx = [0, −Dz /ε, Dy /ε, 0, Bz /µ, −By /µ]T ,


Fy = [Dz /ε, 0, −Dx /ε, −Bz /µ, 0, Bx /µ]T , (2.8)
Fz = [−Dz /ε, Dx /ε, 0, By /µ, −Bx /µ, 0]T .

In each temporal-spatial plane t − ξ , t − η, and t − ζ , the eigenvalues are


easily found by solving the six-degree characteristic equation associated
with the coefficient matrices,
√ √ √ √
λξ = [−α/√εµ, −α/ √εµ, α/ √εµ, α/ √εµ, 0, 0],
λη = [−β/ √εµ, −β/ √εµ, β/ √εµ, β/ √εµ, 0, 0], (2.9)
λζ = [−γ / εµ, −γ / εµ, γ / εµ, γ / εµ, 0, 0],

where ε is the electric permittivity and µ is the magnetic permeability


of the medium. α, β , and γ are functions of the metrics of coordinate
transformation; α = (ξx2 + ξy2 + ξz2 )1/2 , β = (ηx2 + ηy2 + ηz2 )1/2 , and γ = (ζx2 +
ζy2 + ζz2 )1/2 . The above flux vectors now can be further split into domains
of dependence according to the eigenvector structure and the sign of the
eigenvalues [27, 28].

Fε = Fε+ (UL ) + Fε− (UR ),


Fη = Fη+ (UL ) + Fη− (UR ), (2.10)
Fζ = Fζ+ (UL ) + Fζ− (UR ).

The above differential system consists of a total of five dependent variables


and two explicit constitutive relations.
In a stationary frame of reference, the boundary conditions on media
interface require that the tangential components of the electric field intensity
E, and the normal component of the magnetic flux density B be continuous
across the boundary. The discontinuity of the tangential components of the
512 Shang

magnetic field strength H equals the surface density Js . Finally the surface
charge density ρe balances the difference between the normal components
of the electrical displacement D, across the media interface [17];

n × (E1 − E2 ) = 0, (2.11)
n × (H1 − H2 ) = Js , (2.12)
n · (B1 − B2 ) = 0, (2.13)
n · (D1 − D2 ) = ρe . (2.14)

In the split flux vector formulation, the far-field boundary condition is


implemented by setting the incoming flux vector component to a null
value, Eq. (2.10). In essence, at the boundaries of the computational
domain, all incoming waves are suppressed using the signs of the local
eigenvalue as the discriminator [27, 29].

F−
ζ =0 (2.15)


 −(ζy2 + ζz2 )/2γ √εµ ζx ζy /2γ √εµ √
ζx ζz /2γ εµ
 √ √ √
 ζx ζy /2γ εµ −(ζx2 + ζz2 )/2γ εµ ζy ζz /2γ εµ
 √ √ √
 −(ζx2 + ζy2 )/2γ εµ
Fζ− =  ζx ζz /2γ εµ ζy ζz /2γ εµ
0 ζz /2µ −ζy /2µ
 −ζ /2µ
 z 0 ζx /2µ
 ζy /2µ −ζx /2µ 0

0 −ζz /2ε ζy /2ε 

ζz /2ε 0 −ζx /2ε 

−ζy /2ε ζx /2ε 0 
√ √ √ .
−(ζy2 + ζz2 )/2γ εµ ζx ζy /2γ εµ ζx ζz /2γ εµ 
√ √ √ 
ζx ζy /2γ εµ −(ζx2 + ζz2 )/2γ εµ ζy ζz /2γ εµ 
√ √ √ 
ζx ζz /2γ εµ ζy ζz /2γ εµ −(ζx + ζy )/2γ εµ
2 2

For the initial/boundary value problem, the initial value can be easily pre-
scribed by the theoretical results of the transverse electrical wave in the
fundamental mode, TE1,0 , as entrance conditions for either the waveguide
or the antenna [17].

3. NUMERICAL PROCEDURES
Two finite-volume procedures are adopted in the present analysis.
A high-resolution procedure is derived from the compact differencing
Simulating Plasma Microwave Diagnostics 513

scheme [20]. Only the first-order derivative is required for the present anal-
ysis; the formula for the ξ derivative is given as,

αc Uξ (i − 1, j, k) + Uξ (i, j, k) + αc Uξ (I + 1, j, k)
= C1 [Uξ (i + 1, j, k) − Uξ (i − 1, j, k)] + C2 [Uξ (i + 2, j, k) − Uξ (i − 2, j, k)].
(3.1)

In the above formulation, the coefficient is specified by a single param-


eter αc as, C1 = 2(αc + 2)/3 and C2 = (4αc − 1)/3. By selecting αc = 1/4,
a fourth-order accurate or the classic Pade scheme is recovered. The six-
order scheme is obtained by setting αc = 1/3.
For the finite-volume application, the reconstruction process employs
a primitive function for describing the variation of the cell-average data
array. The difference operator, Eq. (3.1), determines the accuracy of recon-
struction [22, 23]. A fourth-order Runge–Kutta method is adopted for
time integration.

U n+1 = U n + ∆t[(Ut )1 + 2(Ut )2 + 2(Ut )3 + (Ut )4 ]/6. (3.2)

In the high-order scheme, the mesh point stencil involves five points
or more and a boundary transition operator becomes necessary. The
staged one-order-lower approximation for the boundary transition scheme
is derived from the summation-by-parts energy analysis [31]. Therefore, a
formally sixth-order interior scheme must be supported by a fifth-order
boundary transition scheme. In the present effort a fourth-order and a
fifth-order one-sided difference formulation are used. In most applications,
this procedure is the source of numerical instability, manifested by spu-
rious high-frequency Fourier components [22, 23]. To maintain numeri-
cal stability for the compact differencing method, a spatial low-pass filter
is incorporated into the high-resolution scheme [23]. For this reason, the
high-resolution procedure is used solely for dispersive and dissipative error
assessment.
The major portion of the present computations is produced by the
numerical procedure derived from the MUSCL (Monotone Upstream-
Centered Schemes for Conservation Laws) scheme [28, 29]. This solu-
tion scheme has a wide range of options in the basic algorithm selection
through the flux vector reconstruction across the control surface of an ele-
mentary volume.
L
Ui+1/2 = Ui + φ/4[(1 − κ)∇ + (1 + κ)∆]Ui , (3.3)
R
Ui+1/2 = Ui − φ/4[(1 + κ)∇ + (1 − κ)∆]Ui , (3.4)
514 Shang

where the superscript L and R denoting the dependent variables that are
evaluated at the left or the right side of the control surface. The differ-
ence operators ∇ and ∆ represent the forward and backward differencing
approximations, respectively. A third-order spatial resolution is achieved as
a windward biased formula by setting φ = 1 and κ = 1/3. Therefore, all
computed results have at least a third-order spatial and fourth-order tem-
poral accuracy.
The implementation of boundary conditions and initial conditions is
straightforward. The most complex no-reflection, far-field boundary condi-
tion is facilitated by the characteristic formulation: at the computational
domain, the incoming flux vectors along each of the transformed coor-
dinates are set to the null value, Eq. (2.15). At the media interfaces, the
outward normal unit vector can be determined to satisfy the Neumann con-
dition. The present analysis uses a build-up process to simulate the plasma
diagnostics; the computational simulations include a wide range of config-
urations of waveguides and pyramidal horn antennas. In general, the initial
value can be easily prescribed by theoretical results of the transverse electri-
cal wave in the fundamental mode, TE1,0 , as entrance conditions for either
the waveguide or the antenna [17]. However, the detailed initial condition for
each computation will be addressed in every specific section of discussion.
The key element of plasma diagnostics using microwave probing is a
pair of pyramidal horn antennas. Their main function is to collimate the
microwave beam and to transmit and receive a microwave across a plasma
medium [9,10]. The schematic of this arrangement is depicted in Fig. 1. In
the present analysis, the microwave transmission is simulated at a frequency
of 12.5 GHz (wavelength of 2.398 cm). The entrance of the pyramidal horn
has the dimensions in wavelengths of (0.397 × 0.928), the smallest and the

Fig. 1. Schematic of pyramidal-horn antennas for plasma diagnostics.


Simulating Plasma Microwave Diagnostics 515

normalized length scale is 0.397 wavelengths (0.938 cm). The horn increases
its cross section linearly in both the x and y-directions; its aperture has the
dimensions in wavelengths of (3.045 × 3.640). The total length of the horn
is 3.307 wavelengths. For the present investigation, two groups of computa-
tions for a rectangular waveguide and microwave antennas were conducted.
For each group of simulations, a single mesh system was utilized.
Taking advantage of the symmetrical geometry with respect to the
x-coordinate and the invariant property of the fundamental transverse elec-
trical wave (TE1,0 ) in the y-coordinate, only the upper half-space of the
antenna was simulated by a single mesh system of (74 × 5 × 142) for the
entire computational domain. The mesh system of the three-dimensional
computation is generated by a surface-oriented coordinate transformation.
All coordinates are normalized by the minimum height of the aperture
with the dimensions of (15.74 × 4.71 × 26.89). At the simulated microwave
frequency, the grid density in terms of points per wavelength (PPW) is
maintained at a value no less than 14. Over the entire range of the elec-
tric conductivities 0.0 < σ/ωε < 2.0, the agreement between the results by
the high-resolution and the MUSCL scheme for waveguides and pyramidal
horn antenna is 1.22%. This numerical error assessment is consistent with
the previous computations for microwave propagation where comparisons
with theoretical results are available [25].
For the evaluation of the energy transmission, the Poynting vector
P = E × H , and the instantaneous power transmitted by the microwave
through plasma along the beam path are calculated by the following
equations:
P = −Ez Hy i − Ex Hz j + (Ex Hy − Ey Hx )k, (3.5)
  
(E × H ) · ds = −1/2∂/∂t [εE 2 + µH 2 ]dV − σ E 2 dV ,
(3.6)
where i, j , and k are the unit vectors in the x, y, and z-direction, an ds
is the surface vector of a control volume associated with an elementary
volume dV .

4. MICROWAVE ATTENUATION
In the frequency spectrum of experimental microwave diagnostics eh2
2 2
<< p <<  , the measured phase shift depends only upon the electron
number density. The attenuation depends on both density and collision
frequency for momentum transfer [9, 10]. For a transverse plane electro-
magnetic wave in the TE1,0 mode and traveling in the z-direction, the wave
516 Shang

motion is described by:

E = Eo exp[−αz + i(βz −  t)], (4.1)

where α and β are the attenuation and phase constants. The phase con-
stant is given by the relationship of β = 2π/λ, and the phase velocity of
the wave is uph =  /β. In a rectangular waveguide, the attenuation and
phase constants α and β of the transverse wave can be calculated by the
following formula:

α = ( /c)2 (σ/ ε)/2β, (4.2)

β = {(βo2 + [βo4 − (σ/cε)2 ]1/2 )/21/2 }, (4.3)

where βo = ( /c)2 − (nπ/a)2 . In the wavelength range where the incident


microwave has a higher frequency than the plasma, the plasma becomes
a relatively low-loss dielectric medium. The wave amplitude attenuates
exponentially and the phase shifts linearly proportional to the distance
propagated [17]. However, there are serious limitations in applying the
one-dimensional, plane-wave model for plasma diagnostics that have a
finite dimension and fluctuating geometric boundary.
The computational simulation for a traveling microwave in homoge-
neous, isotopic plasma has been developed by earlier research efforts [29].
In Fig. 2, the x-component of the magnetic field intensities Hx over a
range of electrical conductivities at fixed values of microwave frequency
and plasma electric permittivity is given together with theoretical results,
0.0 < σ/ ε < 0.25. The simulated microwave in the transverse electrical
T1,0 mode propagates through the plasma at a frequency of 4.0 GHz and a
wavelength of 7.495 cm. The microwave traveling in plasma that is devoid
of a media interface reveals a monotone attenuation proportional to the
magnitude of the electric conductivity. By doubling the value of electric
conductivity, the wave is completely dissipated in 4.75 wavelengths. The
numerical results of wave attenuation are in excellent agreement with the
theory; the maximum discrepancy is less 1% [17, 29].
Interference at the media interface of the transmitted microwave can
be dominant, particularly when the difference of intrinsic impedances is
large and the plasma is only a few wavelengths in thickness. This type
of plasma dimension is frequently encountered in most plasma chan-
nels [14, 15, 18] and the interference effect can easily obscure the plasma
absorption. This phenomenon is clearly demonstrated by numerical simu-
lation of a guided TE1,0 microwave with a frequency of 4 GHz. All the
following numerical results are obtained on a (25 × 25 × 197) grid system
Simulating Plasma Microwave Diagnostics 517

Fig. 2. Validating microwave attenuation in isotopic plasma TE1,0 wave,  = 4 GHz, 0.0 <
σ/ ε < 0.25.

for both the compact difference and MUSCL scheme. The entire compu-
tational domain consists of 10 wavelengths and the square waveguide has
sidewall dimensions of one wavelength each. The plasma sheet strides the
guided wave at the midpoint of the waveguide and has a thickness of two
wavelengths. The plasma sheet is characterized by a uniform and greater
electrical conductivity, σ = 0.75 ε.
In Fig. 3, the x-component of the magnetic field strength along the
microwave path is presented. For purpose of comparison, the same var-
iable in free space is appended in the figure as a reference. From the
numerical results, the magnetic field is substantially distorted by the reflec-
tions from the media interfaces. This behavior is drastically different from
the numerical results describing the microwave motion that is completely
confined within the plasma. The reflected wave from the incident inter-
face produces a wave cancellation toward its origin. In the plasma domain,
the microwave magnetic component exhibits a significant distortion, and
remerges at the exit interface showing attenuation in wave amplitude. This
interference has been known to obscure plasma absorption measurements
518 Shang

Fig. 3. Magnetic field of microwave in plasma sheet TE1,0 wave,  = 4 GHz, σ = 0.75 ε.

[9, 10], and is demonstrated for the first time in the present numerical
simulation.
The y-component of the electrical field intensity of the TE1,0 wave
is depicted in Fig. 4. Again the reflected wave from the media interfaces
substantially distorts the electrical field of the microwave in direct contrast
to the microwave propagating in the plasma without interface boundaries.
For the electrical field component, the reflected wave from the incident
media interface exhibits a much more pronounced phase shift and ampli-
tude reduction.
The overall computed field structure substantiates the observation
that the reflections from the media interfaces create a very complex
wave structure within the plasma sheet, but the attenuated microwave
still retains most of its incident wave characteristics. Since the microwave
measurement is an integrated result from the beam path, it reflects only
an accumulated effect. The measurement is therefore sampling-condition
dependent and no general conclusion can be drawn to compensate the
effects of reflection.
Simulating Plasma Microwave Diagnostics 519

Fig. 4. Electric field of microwave in plasma sheet TE1,0 wave,  = 4 GHz, σ = 0.75 ε.

5. RADIATION STRUCTURE BETWEEN ANTENNAS


The instantaneous radiating field of a microwave between the trans-
mitting and receiving antennas is presented in Fig. 5. In this figure, the
field is presented in continuous contours of electrical field intensity within
the pyramidal horns, and the wave pattern is given in discrete contour
lines external to the antennas. At the frequency of 12.5 GHz, the micro-
wave is only required to travel 10.71 wavelengths along the z-axis to cover
the entire distance from the entrance of the transmitting horn to the exit
of the receiving pyramidal horn. The radiation field reveals a very complex
wave interference pattern, including the diffractions at the edges of the
transmitting and receiving horns, and the multiple reflecting waves origi-
nating from the receiving horn surface.
Microwave propagation is a time-dependent phenomenon; in free
space it is simply a harmonic motion. The temporal sequence of the prop-
agating microwave as it exits from the waveguide, enters the transmitting
antenna and propagates into the receiving antenna is presented by its y-
component electrical field strength, Ey , in Fig. 5. Three features stand
520 Shang

Fig. 5. Microwave radiating field for plasma diagnostics TE1,0 wave,  = 12.5 GHz.

out; first, the wave radiating from the transmitting horn is well structured
and undisturbed until the wave impinges onto the exiting computational
domain and the reflected wave from the exiting boundary is insignificant.
Second, the wavelength of the microwave is modified from the entrance to
the aperture to reflect the wave speed differential from the waveguide to
the antenna exit to agree with the theory of antenna [32, 33]. Third, strong
wave interference is easily detected in the receiving horn.
Figure 6 presents all electromagnetic field components (Bx , Bz , and
Ey ) of the transverse TE1,0 wave when the wave has traversed the entire
computational domain. The wave interaction resulting from the reflected
waves from the metallic surface of the receiving antenna can also be seen.
However, the edge diffractions from the antennas are not detectable from
the computed results. The behavior of the x-component of the magnetic
flux density Bx appears almost as a mirror image of the y-component of
electrical field strength Ey . In fact, they are completely out of phase from
each other.
In the continuously expanding structure of radiation within the trans-
mitting antenna, the energy flux density increases with the increasing size
Simulating Plasma Microwave Diagnostics 521

Fig. 6. Instantaneous electromagnetic field components between antennas TE1,0 wave,


 = 12.5 GHz.

of the control volume along the beam path. The reverse behavior is also
observed in the receiving antenna; however a part of the reduced power
transmitted is due to the reflected wave from the antenna surface. For
the microwave that propagates through free space without any attenuation
the time-averaged ratio of power received at the apertures of the horns is
around 0.1905. In other words, only one fifth of the microwave energy of
the TE1,0 wave is received for the plasma diagnostic arrangement and the
rest has been propagated and reflected into free space. This finding is ver-
ified by the experimental observation [14].

6. INTERACTION WITH PLASMA COLUMN


The diffraction and refraction of a microwave encountering a plasma
column of a finite physical dimension is simulated by inserting a plasma
column on the centerline of the beam path. The TE1,0 microwave propa-
gates at a frequency of 12.5 GHz (λ = 2.398 cm). The rectangular plasma
column has the dimension of one wavelength in width and 0.455 wave-
lengths in depth, and the uniform plasma is characterized by an electrical
conductivity of 69.5 mho/m. The diffracted and reflected wave pattern is
522 Shang

Fig. 7. Comparison of EM fields in free space and passing through plasma column TE1,0
wave,  = 12.5 GHz, w = λ, d = 0.46λ.

clearly discernible by the contrast of continuous contours of electrical field


intensity with and without the presence of the plasma column. Figure 7
presents a composite of the contours of the radiating fields. The radiating
field in the free space is depicted in the lower half-plane and the interact-
ing field of the microwave and the uniform plasma column is inserted in
the upper half-plane. The diffraction and reflection by the plasma column
are detectable by the distorted wave fringe pattern and a perceptible phase
shift by comparison.
The distributions of Poynting vector components in free space and
the interacting field of the microwave and plasma column are presented
together in Fig. 8. First of all, the component of the Poynting vec-
tor along the microwave path dominates the component normal to the
propagating direction. The small plasma column only significantly affects
the magnitude of the Poynting vector after their intersection centered at
Z/L = 13.48. The instantaneous microwave attenuation at the aperture of
Simulating Plasma Microwave Diagnostics 523

Fig. 8. Poynting vector of microwave passing through plasma column TE1,0 wave,  =
12.5 GHz, w = λ, d = 0.46λ.

the receiving antenna is 2.7% and the observed phase shift in Fig. 6 is too
small to be quantified.
In short, the diffraction in a microwave beam path by a weakly
ionized gas column of a relatively low electrical conductivity (σ/ ε =
0.25) is detectable by numerical simulation by the distorted wave fringe
pattern. The reflection by the media interface is inseparable from the
plasma absorption by the present computation, but the total wave atten-
uation by the finite-dimension plasma column can be determined. Unfor-
tunately, there is neither theory nor experimental data to facilitate a direct
comparison.

7. INTERACTION WITH PLASMA SHEETS


An electromagnetic wave attenuates in quasi-conducting and conduct-
ing media. The attenuation can be so rapid that the wave penetrates the
medium only a fraction of a wavelength. Traditionally this depth of pen-
etration is defined as the distance by which the amplitude of the field
decreases to 1/e (36.8%) of its initial value and is given by the formula
524 Shang

Fig. 9. Radiating field of microwave pass through plasma sheet,  = 12.5 GHz, T = 1.823λ,
σ = 69.5 mho/m, at t = 28.52.

δ = 1/(π µσ  )1/2 [17]. Therefore, the depth of a plasma sheet is a criti-


cal parameter for the evaluation of microwave attenuation. In the pres-
ent effort four computations of different plasma sheet thicknesses, 0.456,
0.911, 1.367, and 1.823 wavelengths, are simulated by the MUSCL scheme
with a PPW value of 14.
The interacting field of a TE1,0 microwave at 12.5 GHz and a plasma
sheet of uniform electrical conductivity of 69.5 mho/m (σ/ ε = 0.25) can
be best visualized by the contours of the electrical field intensity. Since the
thinner plasma sheets produce only a small difference from the wave propa-
gation in the free space, only the computed result from the maximum plasma
sheet thickness is given in Fig. 9. The instantaneous contour plot is recorded
at t = 28.52 and has 40 equal increments from the null to the maximum
magnitude of the electrical field intensity. The leading and trailing edges of
the plasma sheet are located at z/L = 11.09 and z/L = 15.68, respectively.
Substantial microwave attenuation is clearly visible as it propagates through
the plasma sheet. However the reflected wave from the walls of the receiving
pyramidal horn is barely identifiable.
Simulating Plasma Microwave Diagnostics 525

Fig. 10. Microwave blackout phenomenon TE1,0 wave,  = 12.5 GHz, T = 3λ,
σ = 69.5 mho/m.

The blackout phenomenon occurs for the incident microwave with a


frequency of 12.5 GHz when the plasma sheet is thicker than three wave-
lengths and at an electrical conductivity of 69.5 mho/m. The computed
contours of the electrical field intensity, Ey , clearly reveal this phenome-
non in Fig. 10. In fact, the microwave motion diminishes substantially in
the plasma sheet starting at its leading edge near the aperture of the trans-
mitting antenna, z/L = 7.71.
This observation is clearly supported by the behavior of the instanta-
neous x-components of the magnetic flux density, Bx , and the y-component
of the electrical intensity, Ey . The suppressed amplitudes of both the elec-
tromagnetic components of the TE1,0 wave are clearly displayed in Fig. 11.
Both components essentially vanish at z/L = 14.67 or 1.66 wavelengths from
the antenna aperture. Equally important, the reflected wave at the media
interface is not detected. The microwave blackout phenomenon is further
substantiated by the suppressed absolute value of the Poynting vector along
the beam path in Fig. 12. Note that the Poynting vector essentially goes
526 Shang

Fig. 11. Typical electromagnetic field components in microwave blackout TE1,0 wave,  =
12.5 GHz, T = 3λ, σ = 69.5 mho/m.

to zero at z/L = 15.00. This computed result under predicts the theoretical
depth of penetration by 0.7%.

8. DIAGNOSTICS FOR HYPERSONIC MHD CHANNEL


In practical applications using microwaves for plasma diagnostics, the
plasma is assumed to be a three-species, uniform medium with negligi-
ble thermal particle motion. Under these circumstances, microwave atten-
uation can be studied as a function of plasma frequency, electron-neutral
particle collision frequency, and transmitted frequency. It is common prac-
tice in microwave diagnostics to measure the attenuation at two or more
transmitting frequencies to eliminate the error incurred in estimating the
electron-neutral particle collision frequency [9–13].
In the data reduction process, the attenuation is evaluated by comparing
the microwave power level at the receiving antenna to that in the absence
of plasma. In most experiments, the microwave detector records only the
voltage of the microwave [9, 10, 14, 15]. The voltage must be converted
into an averaged microwave power by time-integrating the converted signal
Simulating Plasma Microwave Diagnostics 527

Fig. 12. Poynting vector in microwave blackout TE1,0 wave,  = 12.5 GHz, T = 3λ,
σ = 69.5 mho/m.

after passing through a logarithmic amplifier. This data reduction process


becomes a source of measurement error.
The diffraction and reflection of the probing beam and plasma are
unaccounted for in most experimental investigations. A major concern is
that all reflected microwaves could be sources of error in determining the
transmitted power, particularly for measurements of lower plasma atten-
uation in a weakly ionized gas in a hypersonic plasma channel [14, 15,
18]. In a ground-based facility, even if the transport property remains
unaltered, the fluctuating plasma still can change its dimension and geo-
metric configuration constantly. In the present simulation, the multiple
wave reflections overwhelm the high-resolution scheme. Therefore, only the
MUSCL scheme is used.
The present simulation duplicates the microwave diagnostics experi-
mental setup in a small plasma channel operating at a Mach number of
5.15. The weakly ionized air is generated by a DC glow discharge and,
from a prior survey by a double Langmuir probe; the plasma stream
has an electron number density and electron temperature of 1012 /cc and
10,000 K, respectively [15, 18]. Microwave attenuation measurements are
528 Shang

Fig. 13. Typical magnetic flux density distributions along the microwave beam path TE1,0
wave,  = 12.5 GHz, 1 < σ < 12 mho/m.

obtained by a previously described set of pyramidal horns across the


plasma channel. The microwave transmits through quartz port windows
with a thickness of 0.722 cm each, and the channel width is 2.89 cm.
Therefore, the separation distance between apertures of the antenna horns
is 4.33 cm. At the probing microwave frequency of 12.5 GHz, this sep-
aration distance is 1.807 wavelengths. A total of five different plasma
conditions have been investigated to bracket all possible experimental
conditions. Since the detailed chemical kinetics are not considered, the
plasma is assigned a range of electrical conductivities from 1 to 12 mho/m
to encompass the experimental conditions. The computed magnetic field
flux densities along the beam path for all plasmas simulated are given
in Fig. 13. At the lowest value of electric conductivity, the mismatched
intrinsic impedance between media is extremely small; thus the reflection
at the media interface becomes negligible.
Figure 14 depicts the comparison of computed and measured micro-
wave power distributions in the receiving horn antenna; the transmitted
microwave power in free space is included as a reference comparison.
At the relatively low electrical conductivity conditions, the electromag-
netic fields exhibit very little difference from the microwave propagation in
Simulating Plasma Microwave Diagnostics 529

Fig. 14. Comparison of microwave power distribution in receiving horn antenna TE1,0
wave,  = 12.5 GHz, σ ≈ 1 mho/m.

free space. It becomes evident by direct contrast to the microwave power


transmission in free space that the power is continuously reflected from
the radiating field and the energy is persistently dissipated in the plasma.
Again, the reflected wave from the media interface is negligible, and the
effects of diffraction at the apertures of antenna are accounted for, but are
not separable from the integrated result.
The attenuation of the microwave increases substantially with increased
electrical conductivity of the plasma within a hypersonic plasma channel.
The computed result of electrical conductivity of 1 mho/m under predicts
measurements by Kurpik et al. [14] and Henderson et al. [18] by nearly
16%. This discrepancy is anticipated, because the plasma is not uniform
across the beam path and the measurement by microwave probing can
yield only an averaged quantity. The computed simulation, on the other
hand, assumes the medium is uniform in the MHD channel. Nevertheless,
the time-averaged microwave energy of 4.2 µW received by the antenna is
530 Shang

comparable to the measured data; therefore it is reasonable to conclude that


the measured data and computed result are complementary.

9. CONCLUSIONS
Solutions of the time-dependent Maxwell equations and the general-
ized Ohms law have successfully described the radiating electromagnetic field
for plasma diagnostics using microwave probing. The numerical simulation
duplicates the commonly adopted pyramidal horn antennas arrangement
used in experiment at a transmitting frequency of 12.5 GHz. The diffrac-
tions at the apertures of the antenna and a uniform plasma column in the
beam path are captured in the numerical results.
For all investigated weakly ionized gases with an electrical conductiv-
ity up to 100 mho/m, the wave reflection from the interface of free space
and plasma is negligible. This numerical result fully supports earlier exper-
imental observations.
In the first application for plasma diagnostics in a hypersonic mag-
neto-hydrodynamic channel, the simulated result reaches a reasonable
affinity with experimental observations.

ACKNOWLEDGMENT
The sponsorship from Dr. J. Schmisseur and Dr. F. Fahroo of the Air
Force Office of Scientific Research is deeply appreciated. The stimulating
discussions with Prof. J. Menart of Wright State University, as well as, Dr.
R. Kimmel and J. Hayes of the Air Force Research laboratory are sin-
cerely acknowledged.

REFERENCES
1. Shang, J. S. (2002). Shared knowledge in computational fluid dynamics, electromagnet-
ics, and magneto-aerodynamics. Prog. Aerospace Sci. 38(6–7), 449–467.
2. Roth, J. R. (2000). Electrohydrodynamic flow control with a glow-discharge surface
plasma. AIAA J. 38(7), 1166–1172.
3. Artana, G., Adamo, J., Leger, L., Moreau, E., and Touchard, G. (2002). Flow control
with electrohydrodyamic actuators. AIAA J. 40(9), 1773–1779.
4. Enloe, C. L., McLaughlin, T. E., Van Dyken, R. D., Kachner, K. D., Jumper, E. J., and
Corke, T. J. (2004). Mechanisms and responses of a single dielectric barrier plasma actu-
ator: plasma morphology. AIAA J. 42, 589–594.
5. Shang, J. S., and Surzhikov, S. T. (2004). Magneto-aerodynamic interaction for hyper-
sonic flow control. AIAA J. 43(8), 1633–1643.
6. Park, C., Mehta, U. B., and Bogdanoff, D. W. (2001). MHD Energy bypass scramjet
performance with real gas effects. J. Propul. power 19(5), 1049–1057.
Simulating Plasma Microwave Diagnostics 531

7. Miles, R. B., Brown, G. L., Lempert, W. R., Yetter, R., Williams, G. L., Bogdonoff, S.
M., Natelson, D., and Guest, J.R. (1995). Radiatively driven hypersonic wind tunnel,
AIAA J. 33(8), 1463–1470.
8. Anikin, N., Kukaev, E., Starikovskaia, S., and Starikovskii, A. (2004). Ignition of hydro-
gen-air and methane-air mixtures at low temperatures by nanosecond high voltage dis-
charge. AIAA 2004–0833, Reno NV, January 2004.
9. Heald, M. A., and Wharton, C. B. (1965). Plasma Diagnostics with Microwaves, Wiley,
New York.
10. Golant, V. E. (1983). Methods of diagnostics based on the interaction of electromagnetic
radiation with a plasma, basic plasma physics I. In Galeev, A., and Sudan, R. (eds.),
North Holland Publishing Co., Amsterdam, 630–680.
11. Raizer, Yu., and Gas, P. (1991). Discharge Physics, Springer-Verlag, Berlin.
12. Smoot, L. D., and Underwood, D. L. (1966). Prediction of microwave attenuation char-
acteristics of rocket exhausts. J. Spacecraft Rockets 3(3), 302–309.
13. Fedrick, R. A., Blevins, J. A., and Coleman, H. W. (1995). Investigation of microwave attenu-
ation measurements in a laboratory-scale motor plume. J. Spacecraft Rockets 32(5), 923–925.
14. Kurpik, A., Menart, J., Shang, J., Kimmel, R., and Hayes, J. (2003). Technique for mak-
ing microwave absorption measurements in a thin plasma discharge. AIAA 2003–3748,
Orlando, FL, 23–26 June 2003.
15. Shang, J. S., Kimmel, R., Hayes, J., Tyler, C., and Menart, J. (2005). Hypersonic mag-
neto-aerodynamic experimental facility for magneto-aerodynamics interaction. J Space-
craft Rockets 42(1).
16. Kennet, H., and Taylor, R. A. (1996). Earth re-entry simulation of planetary entry envi-
ronment. J. Spacecraft Rockets 3(4), 504–512.
17. Krause, J. D. (1953). Electromagnetics, McGraw-Hill, pp.390–481.
18. Henderson, S., Menart, J., Shang, J. S., Kimmel, R., and Hayes, J. (2004). Data reduc-
tion analysis for a cylindrical, double langmuir probe operating in a high speed flow.
AIAA 2004–0360, Reno NV, January 5–8, 2004.
19. Surzhikov, S. T., and Shang, J. S. (2004). Two-Component plasma model for
two-dimensional glow discharge in magnetic field. J. Comp. Phys. 1999, 437–464.
20. Lele, S. K. (1992). Compact difference schemes with spectral-like resolution. J. Comp.
Phys. 103, 16–42.
21. Carpenter, M. H., Gottlieb, D., and Abarbanel, S. (1995). Time-stable boundary condi-
tions for finite-difference schemes solving hyperbolic systems: methodology and applica-
tion to higher-order compact systems. J. Comp. Phys. 117, 300–317.
22. Shang, J. S. (1999). High-order compact-difference schemes for time-dependent maxwell
equations. J. Comput. Phys. 153 312–333.
23. Gaitonde, D. V., Shang, J. S., and Young, J.L. (1999). Practical aspects of higher-order
numerical schemes for wave propagation phenomena. Int. J. Numer. Methods Eng. 45,
1849–1869.
24. Taflove, A. (1997). Numerical issues regarding finite-difference time-domain modeling of
microwave structures, time-domain methods for microwave structure analysis and design.
In Itoh, T., and Houshmand, B. IEEE Press, New York, pp. 59–75.
25. Mur, G. (1981). Absorbing boundary conditions for the finite-difference approximation
of the time-domain electromagnetics field equations. IEEE Trans. Electromagn. Compat-
ibility, 23, 377–382.
26. Berenger, J.-P. (1994). A perfect matched layer for the absorption of electromagnetic
waves. J. Comp. Phys. 114, 185–200.
27. Shang, J. S., and Fithen, R. M. (1996). A Comparative study of characteristic-based
algorithms for the maxwell equations. Comput. Phys. 125, 378–394.
532 Shang

28. Shang, J. S., and Gaitonde, D. (1995). Characteristic-based, time-dependent maxwell


equation solver on a general curvilinear frame. AIAA J. 33(3), 491–498.
29. Shang, J. S. (2003). Computational electromagnetics for microwave attenuation in
weakly ionized air. AIAA 2003–3621, Orlando, FL, 23–26.
30. Mitchner, M., and Kruger, C. H., Jr. (1973). Partially Ionized Gases, Wiley, New York,
126–241.
31. Gustafasson, B. (1975). The Convergence rate for difference approximations to mixed
initial boundary value problems. Math. Comp. 29, 396–414.
32. Balanis, C.A. (1982). Antenna Theory: Analysis and Design, Wiley, NY, 8.3–8.86.
33. Maloney J., Smith, G., Thiele, E., and Gandhi, O. (2002). Modeling of antennas, com-
putational electromagnetics: the-finite-difference, time-domain methods. In Taflove, A.,
and Hagnes, S. (Eds.), 2nd ed., Artech House, Boston, pp. 627–701.
34. Katz, D., Piket-May, M., Taflove, A., and Umashankar, K. (1991). FDTD analysis
of electromagnetic wave radiation from systems containing horn antennas. IEEE Tans.
Antenna Propagation. 39(8), 1203–1212.

You might also like