48 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 61, NO.

1, FEBRUARY 2019

Reconstruction of Internal Field of Dielectric Objects

for Noninvasive SAR Measurement Using Boundary
Integral Equation
Shuntaro Omi , Student Member, IEEE, Toru Uno, Senior Member, IEEE, Takuji Arima, Member, IEEE,
and Joe Wiart, Senior Member, IEEE

Abstract—Reconstruction of the electromagnetic (EM) fields frequency range. Accordingly, measurement of the SAR has
inside a dielectric object is investigated in order to develop a non- already been indispensable for developing wireless communi-
invasive specific absorption rate measurement. The proposed re- cation devices and industrial products that radiate EM fields.
construction method is based on the boundary integral equation
(BIE) derived from the surface equivalence theorem that relates The equipment for measuring the SAR using a human-
the equivalent EM currents on the surface enclosing the primary equivalent liquid phantom [2] has been developed and is now
source to the radiated external fields. The EM currents are recon- spreading among the associated institutions. In this system, the
structed by solving the discretized BIE using the field data sampled phantom is placed into a human-shaped tank and is illuminated
on the surface surrounding all of the target objects that consist of by the device under test (DUT), i.e., wireless communication
the dielectric phantom and radiating antenna. The field distribu-
tion inside the dielectric object is obtained from the reconstructed devices and/or industrial products. The induced EM fields are
currents. A probe correction technique is also proposed to enable measured by a probe invasively inserted into the phantom. Thus,
the application of this method to practical probe measurements. this method is straightforward, but it inherently involves cer-
As the first step to the practical applications, the validity and use- tain drawbacks such as the considerably long time consumed
fulness of the proposed method are demonstrated numerically and for volumetric (three-dimensional) field measurements and the
experimentally using lossless and lossy homogeneous dielectric ob-
jects located near a dipole antenna, respectively. It is shown that requirement of highly accurate mechanical positioning of the
the accuracy tends to deteriorate in the case of the lossy phantom, probe. In addition, the use of the liquid phantom causes insta-
but this can easily be improved without significant modification of bility in the density.
the proposed method. Noninvasive SAR estimation is indispensable in order to re-
Index Terms—Antenna measurements, electromagnetic (EM) move these inherent difficulties. In this context, the field dis-
fields, inverse problems, integral equations, specific absorption rate tribution in the dielectric phantom is reconstructed from the
(SAR). on-surface (two-dimensional) field data observed in the outer
region of the phantom. The SAR is calculated from the elec-
I. INTRODUCTION tric field. This approach can be used with any type of phantom,
HE GROWTH of wireless communication systems includ- including the stable dry (solid) type, because the field measure-
T ing cell phones has raised concerns about the possible
health effects of exposure to electromagnetic (EM) fields. The
ments are performed noninvasively. Several works have taken
this approach based on a boundary element method (BEM) for-
International Commission on Non-Ionizing Radiation Protec- mulation [3], [4]. In [3], the field continuity condition is applied
tion identifies the specific absorption rate (SAR) as a measure to connect the fields on the surface of the phantom with the
of exposure to EM fields and recommends the use of the SAR fields outside the phantom. In this method, the DUT is not in-
value as the safety standard [1], especially in the microwave cluded in the measurement surface. The field on the surface of
the phantom is estimated by solving a linear equation derived
from the BEM and is used to evaluate the field in the phantom.
Manuscript received January 26, 2018; accepted February 20, 2018. Date of
publication April 5, 2018; date of current version November 15, 2018. This work In [4], a boundary integral equation (BIE) based on the surface
was supported in part by JSPS KAKENHI Grant JP17J07461. (Corresponding equivalence theorem [5], [6] is extended to a geometry where
author: Shuntaro Omi.) the phantom is placed near the DUT by employing the Paggio–
S. Omi is with the Department of Electrical and Electronic Engineering,
Tokyo University of Agriculture and Technology, Tokyo 184-8588, Japan Miller–Chan–Harrington–Wu (PMCHW) formulation [7], [8].
(e-mail: [email protected]). In this method, the equivalent electric current on the surface en-
T. Uno and T. Arima are with the Tokyo University of Agriculture and closing the DUT is reconstructed from the field observed on the
Technology, Tokyo 184-8588, Japan (e-mail: [email protected]; t-arima@
cc.tuat.ac.jp). surface surrounding only the DUT. The field inside the phantom
J. Wiart is with Telecom ParisTech, LTCI, University Paris Saclay, Paris is then evaluated through the EM current on the surface of the
75103, France (e-mail: [email protected]). phantom, which is calculated from the reconstructed current us-
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. ing the PMCHW formulation. Although these studies would be
Digital Object Identifier 10.1109/TEMC.2018.2813398 quite promising for noninvasive SAR estimation, measurements

0018-9375 © 2018 IEEE. Personal use is permitted, but republication/redistribution requires IEEE permission.
See http://www.ieee.org/publications standards/publications/rights/index.html for more information.

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
OMI et al.: RECONSTRUCTION OF INTERNAL FIELD OF DIELECTRIC OBJECTS FOR NONINVASIVE SAR MEASUREMENT USING BOUNDARY 49

on the surface enclosing only the phantom or DUT are required.

This requirement needs the probe measurements to be performed
in the area sandwiched by the phantom and DUT. This constraint
prevents the above methods from use in practical applications
because the DUT is placed close to the phantom in many cases.
In such cases, the probe and its support structure in the sand-
wiched area seriously perturb the field and are remarkably mutu-
ally coupled with both the DUT and phantom. Moreover, probe
correction is not performed in either of these studies.
In this paper, a novel reconstruction method for the inter-
nal field of dielectric objects that satisfies the major necessary Fig. 1. Geometry of the problem. The field data measured by the probe antenna
conditions for practical measurements is proposed. In our pro- on Smeas contain two contributions, one of which is the radiation field from the
DUT, and the other is the field scattered by the phantom.
posed method, measurements of the fields are performed on
the surface surrounding both of the test objects (the DUT and
phantom). Thus, a sufficient distance between the test objects are observed by a probe antenna on the surface Smeas enclosing
and the probe is able to be maintained in order to minimize field both the phantom and DUT. In this situation, the observed field
perturbation and coupling. The inverse reconstruction technique data contain the sum of the contributions of the direct radia-
proposed here is based on Mitharwal’s method [4] employing tion from the DUT and the scattering by the phantom. Using
a probe correction technique, which is derived by the plane- the surface equivalence theorem or Huygens’ principle [5], [6],
wave expansion [9], [10]. As shown later, the proposed method [13], the primary field source can be replaced by the secondary
works accurately in the case of a lossless phantom, but its accu- equivalent electric and magnetic currents flowing on Spha and
racy tends to degrade in the case of a lossy phantom, especially SDUT , and the electric field on Smeas is expressed by
when the reconstruction surface enclosing the DUT becomes
wider and interrupts the reflection from the phantom. However, E(r) = − Lk0 (η0 JDUT ; r) + Kk0 (MDUT ; r)
a method for improving the accuracy in this case is also pro-
posed in this paper. It employs the solution of the lossless case − Lk0 (η0 Jpha ; r) + Kk0 (Mpha ; r) (1)
as the initial solution of the lossy case.
In the next section, the fundamental principles of the recon- where η0 and k0 are the characteristic impedance and the
struction methods are briefly reviewed in association with the wavenumber in vacuum, respectively. The operators L and K are
surface equivalence theorem. The reconstruction procedure for integral operators to be expressed below. JDUT/pha and MDUT/pha
the proposed method is described in Section III after introducing denote the equivalent electric and magnetic currents on SDUT/pha
Mitharwal’s method, which is the basis of our proposed method. that are expressed by the magnetic/electric field (H/E) on the
Probe correction and a numerical implementation are also de- surface, as follows:
veloped. The validity and effectiveness of the proposed method
are demonstrated numerically using a lossless spherical object J(r ) = n̂ × H(r ), M(r ) = E(r ) × n̂ (2)
located near a dipole antenna in Section IV. The effects of probe
correction are also verified in this section using a horn antenna where n̂ indicates the unit normal vector. The integral operators
as the probe. The validity of the proposed method is demon- in (1) are expressed by
strated experimentally in Section V, showing that our proposed 
method is realizable with a simple measurement setup in an 1
Lk (X; r) =  ∇G k (r, r )∇  · X(r )
anechoic chamber. The case of a lossy object is discussed in S k
Section VI-A. The accuracy degradation problem and improved
+ G k (r, r )X(r )d S  , (3)
method mentioned above are also discussed in this section. 
It is noted that this paper is dedicated to the study of internal-
Kk (X; r) =  X(r ) × ∇G k (r, r )d S  (4)
field reconstruction for dielectric objects. The application of the S
reconstruction method is not limited to SAR measurement but
can potentially be extended to other purposes including antenna where X represents EM currents J or M in (1), and G k (r, r ) =

diagnostics with a dielectric medium and the suppression of e− jk|r−r | /|r − r | is the scalar Green’s function with the
undesired fields for antenna measurements that are discussed in wavenumber k. The superscript for each operator indicates the
[11] and [12]. wavenumber used in the Green’s function.
Thus, the electric field can be expressed by two equivalent
currents using Huygens’ principle, but it has been demonstrated
II. STATEMENT OF THE PROBLEM that only the electric or magnetic current is needed to represent
The problem under consideration for the internal-field recon- the electric field outside the closed surface enclosing the pri-
struction is illustrated in Fig. 1, where the DUT is located near mary source [14], [15]. Although the equivalent current is not
the phantom. SDUT and Spha are the closed surface enclosing the given explicitly as (2) in this representation, one of the unknown
DUT and the surface of the phantom, respectively. EM field data quantities can be reduced. Then, the electric-field representation

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
50 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 61, NO. 1, FEBRUARY 2019

used in this paper is simplified as follows:

E(r) = − Lk0 (η0 JDUT ; r)

− Lk0 (η0 Jpha ; r) + Kk0 (Mpha ; r). (5)

The computational load including the coding of (5) would be

light compared with that of (1). The original representation of
Jpha , Mpha still remains for the reconstruction of the field inside
the phantom. The electric field at any points inside a uniform
phantom is also calculated from the surface currents using the
equivalent theorem: Fig. 2. Illustration of the scattering by the phantom. The information of the
shadowed area is mainly transmitted through the scattering by the phantom.
E(r) = Lk p (η p Jpha ; r) − Kk p (Mpha ; r) (6)
where the second term on the right-hand side is expressed by
where η p and k p are the characteristic impedance and the  
wavenumber in the phantom, respectively. M(JDUT ; r) = −η0 LkM0 P KkM0 P
  k k
−1
−LkP0 P − L PpP KkP0P + K PpP
III. SOURCE RECONSTRUCTION n̂ × k k
KkP0P + K PpP LkP0 P + L PpP
As described in Section II, the electric field in the phantom
is calculated from the equivalent currents Jpha and Mpha . In this  
−n̂ × η0 LkP0D  
section, the inverse problem for reconstructing these currents is k0
JDUT . (9)
considered. In this problem, the field data on the surface Smeas are n̂ × K P D
assumed to be known by the measurement. Mitharwal’s method The current JDUT is obtained from the electric field on Smeas by
is introduced at first; then, the probe correction and numerical solving (8), and the currents Jpha and Mpha are then given by (7).
implementation are presented. The first and second terms in (8) have important physical
meanings; the first one represents the direct propagation from
A. Mitharwal’s Reconstruction Technique DUT to the probe, whereas the second represents the propa-
gation through the scattering (reflection or refraction) of the
A reconstruction technique applicable to the internal-field phantom. The information of JDUT is transmitted along these
reconstruction problem has been proposed by Mitharwal et al. propagation paths; in particular, the information of the current
in [4]. This method has been established by relating JDUT in (5) on the shadowed area facing the phantom is mainly transmitted
to Jpha and Mpha using the PMCHW formulation [7], [8]: through the scattering of the phantom, as illustrated in Fig. 2.
    k k
−1
Jpha −LkP0 P − L PpP KkP0P + K PpP B. Probe Correction
= n̂ × k k
Mpha KkP0P + K PpP LkP0 P + L PpP In order to solve (8) with respect to JDUT , the electric field
  E at the observation point r on Smeas is required. However, it is
n̂ × EDUT
fundamentally impossible to obtain its net value from the output
n̂ × η0 HDUT voltage at a probe terminal. Consequently, probe correction is
  −1 essential. In this study, a probe correction technique utilizing a
k k
−LkP0 P − L PpP KkP0P + K PpP plane-wave expansion with a diagonal translation operator [9],
= n̂ × k k [10] is adopted and incorporated into the reconstruction method
KkP0P + K PpP LkP0 P + L PpP
described above.
  The measured data, i.e., the output voltages at the probe ter-
−n̂ × η0 LkP0D  
k0
JDUT , (7) minal, are expressed by the weighted integral of the electric
n̂ × K P D field

where n̂ is the unit normal vector on Spha . The symbol (-)−1 U (ri ) = w(r) · E(r)d V 
indicates the inverse operator of the PMCHW formulation Vprobe
that calculates the currents on the phantom from the incident 
EM fields illuminated by the DUT [EDUT and HDUT in (7)]. = w(r) · [−η0 Lk0 (JDUT ; r) + M(JDUT ; r)]d V 
The subscript for each operator indicates the observation or Vprobe
source region; that is, P is for Spha and D for SDUT . Using the
(10)
above relation, (5) is reformulated as
where w is a weighting function of the probe that depends on
E(r) = −η0 Lk0 (JDUT ; r) + M(JDUT ; r) (8) the probe’s receiving characteristics, and ri indicates the probe’s

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
OMI et al.: RECONSTRUCTION OF INTERNAL FIELD OF DIELECTRIC OBJECTS FOR NONINVASIVE SAR MEASUREMENT USING BOUNDARY 51

reference position, i.e., a sampling point on Smeas . Introducing distinguished by the superscripts L and K ; if L is chosen, the
a new vector ri = r − ri , the scalar Green’s function in vac- operator F L calculates (I − k̂k̂) · X̃(k̂) as in (13), whereas if K
uum contained in (10), G 0 (ri + ri , r ), can be expressed using is chosen, k̂ × X̃(k̂) is calculated as in (14). The operator W is
a plane-wave expansion in a diagonal form as the translating and weighting operator as in (13) and (14) for
 the sampling point ri .
 
G 0 (ri + ri , r ) =  T∞ (k, ri )e j k̂·ri e− j k̂·r d k̂2 (11) It is noted that the proposed formulation in (16) can be imple-
mented in the multilevel multipole manner, as described in [9].
where k = k k̂, and T∞ is the diagonal translation operator given This potentially reduces the computational cost for the iterative
by matrix inversion method.
∞
jk
T∞ (k, ri ) = − (− j)l (2l + 1)h l(2) (kri )Pl (k̂ · r̂i ). (12) C. Numerical Implementation
4π l=0
Let us explain how to implement the numerical calculation
In the above equation, h l(2)
is the spherical Hankel function of method for obtaining the EM currents and field distribution of
the second kind and Pl is the Legendre polynomial. With the the phantom. First, the surfaces SDUT and Spha are meshed into
above expansion, the two terms on the right-hand side of (10) small triangular patches. Then, the current distribution JDUT on
are modified to the following forms: SDUT is expanded as the weighted sum of the basis functions in
 the same manner as the method of moments (MoM) [16]
w(r) · Lk0 (X; r)dv N
n 
Vprobe JDUT = In f D (r D ) (18)
 n=0
=  T∞ (k, ri )w̃ k̂ · I − k̂k̂ · X̃ k̂ d k̂2 , (13)
where the Rao–Wilton–Glisson (RWG) vector function was
  chosen as the basis function f Dn
in this study. In is an unknown

w(r) · Kk0 (X; r)dv =  T∞ (k, ri )w̃ k̂ · k̂ × X̃ k̂ d k̂2 expansion coefficient, and r D is a point on the triangle.
Vprobe Substituting (18) into (16) and discretizing the integral opera-
(14) tors, the matrix equation relating the expansion coefficient In in
(18) to the measured data U on Smeas is obtained and expressed
where X̃ is the spatial Fourier transform of X given by by


X̃ k̂ = F X(r ); k̂ =

X(r )e− jk·r d S  (15) U = W F LD + N I (19)
S
where U and I are vectors consisting of the measured data
and can be interpreted as the far-field expression of the source X.
and expansion coefficients, respectively. The matrices W, F LD ,
The vector function w̃ in (13) and (14) is also the inverse Fourier
and N will be indicated below. If the vector I is obtained by
transform of the spatial weighting function w evaluated in the
 solving (19), then the current distribution JDUT on SDUT can be
same manner but replacing the Fourier kernel with e jk·r . It rep-
determined from (18). The electric-field distribution inside the
resents the probe’s receiving characteristics for various incident
phantom is obtained from JDUT through (7) and (6).
plane waves, i.e., the far-field pattern of the probe multiplied by
The i jth element of the matrix W is given by numerically
its antenna factor.
integrating (13) or (14) and is expressed by
Using the above equations, (8) is transformed into a new
expression in which the given quantity is not the electric field Wi j = w j TL (k k̂ j , ri )w̃(k̂ j ) (20)
but the observed voltage U at the sampling point:
where k̂ j = sin θ j cos φ j x̂ + sin θ j sin φ j ŷ + cos θ j ẑ is a unit
U (ri ) = W F DL JDUT ; k̂ + N JDUT ; k̂ ; ri (16) wavenumber vector at jth abscissa with respect to θ and φ,
where the new operator N is expressed by and w j is a weight. They are determined by a numerical in-
  tegration formula. In this study, the Gaussian quadrature rule
N JDUT ; k̂ = −η0 F PL F PK was adopted with an emphasis on precision. TL is the diagonal
  −1 translation operator when the series in (12) is truncated at the
k k
−LkP0 P − L PpP KkP0P + K PpP finite number L. It is known from a numerical analysis [17] that
n̂ × k k the required minimum value of L is given by
KkP0P + K PpP LkP0 P + L PpP
2/3
  L = ka + 1.8d0 (ka)1/3 (21)
−n̂ × η0 LkP0D  
k0
JDUT . (17) where k is the wavenumber and a is the radius of the sphere
n̂ × K P D
enclosing all of the test objects. d0 is the number of digits for
In (16) and (17), F is the Fourier operator expressed in (15), the desired accuracy. The number of abscissas of numerical
and its subscript D or P indicates the source region S in (15); integration in (20) depends on L as well. We set Nθ = αL for
that is, F D means that the integration is performed on SDUT , and θ , Nφ = α(2L + 1) for φ, and α = 2 in this study, where α is
F P is on Spha . The integral form with respect to the vector k̂ is the oversample factor.

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
52 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 61, NO. 1, FEBRUARY 2019

The elements of F correspond to the far field in the direction

determined by the abscissa of numerical integration, as seen
from (15), and are given as follows:
 L 
F D/P i j = I − k̂i k̂i · F f D/P (rD/P ); k̂i
j
(22)
 K 
F D/P i j = k̂i × ·F f D/P (rD/P ); k̂i
j
(23)

where f P (rP ) is the RWG function defined on Spha .

The matrix N corresponding to the discretized version of the
operator N is obtained in the same manner described above Fig. 3. Geometry for numerical demonstration. (a) Simulation model.
employing the Galerkin method for the PMCHW operator; it is (b) Reconstruction surface meshed into 1256 triangles.
expressed by
 
N = −η0 F LP F KP
 k k
−1
−LkP0 P − L PpP KkP0 P + K PpP
k k
KkP0 P + K PpP LkP0 P + L PpP
 
−η0 LkP0D
(24)
KkP0D
where L P(P/D) and K P(P/D) are given by
L P(P/D) = n̂ × f P (rP ), n̂ × Lk0 (f P/D (rP/D ); rP ) Spha
= −f P (rP ), Lk0 (f P/D (rP/D ); rP ) Spha (25)

K P(P/D) = n̂ × f P (rP ), n̂ × Kk0 (f P/D (rP/D ); rP ) Spha

= −f P (rP ), Kk0 (f P/D (rP/D ); rP ) Spha (26)
and •, • Spha indicates the inner product on Spha .
According to the above implementation, all of the elements
for solving the matrix equation in (19) are completed. The re-
sultant rectangular matrix equation is solved in the least-squares
sense by solving the alternative normalized equation Fig. 4. Electric-current distributions on Spha (left) and electric-field distribu-
∗ ∗ ∗ ∗
tions in the sphere (right). The absolute values at the centers of the meshed
F L∗
D +N W U= F L∗
D +N W W F LD +N I (27) triangles are shown. (a) Reference (MoM). (b) The proposed method.

given by multiplying the adjoint matrix. The generalized mini-

mal residual method [18], one of the iterative methods, is em- near field is sampled at discrete points on the spherical surface
ployed for solving the above normalized equation. with a radius of 5λ. The spacing between the sampling points
is constant, and the angular intervals are θ = φ = 5◦ . In
IV. NUMERICAL DEMONSTRATIONS this simulation, only the φ components of the electric fields are
used for reconstruction, assuming the case where measurement
In this section, the validity and effectiveness of the proposed
is performed using a horizontally polarized small dipole probe.
method described above are numerically demonstrated using
In the proposed method with probe correction, the probe is
a lossless dielectric object placed close to a dipole antenna.
assumed to be a Hertzian dipole whose antenna factor is 1.
It is also shown how effectively the reconstructed results are
Using the calculated field data, the currents on the reconstruction
improved by probe correction. In all numerical simulations,
surfaces in Fig. 3(b) are reconstructed, where the cylinder that
the given data for the reconstruction and the reference values
encloses the dipole antenna corresponds to SDUT in Fig. 1 and
are calculated by the MoM.
the sphere is the surface Spha . The cylinder has a height of 2λ/3
and a radius of λ/12 and is centered at the dipole’s feed point.
A. Lossless Dielectric Object The sphere is coincident with the surface of the dielectric object.
The simulation model studied here is illustrated in Fig. 3(a). Fig. 4(b) shows the current distributions on the dielectric
A lossless dielectric sphere with a relative permittivity εr = 40 sphere Spha (left) and the electric-field distributions in the sphere
is illuminated by a half-wave (length: 6 cm) dipole operating at (right) reconstructed using the proposed method. The simulation
2.5 GHz. The diameter of the sphere is 5 cm (2λ/3), and the results obtained by the MoM are also illustrated in Fig. 4(a)
dipole is placed λ/6 away from the sphere’s surface. The electric for comparison with the reconstructed results. All figures are

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
OMI et al.: RECONSTRUCTION OF INTERNAL FIELD OF DIELECTRIC OBJECTS FOR NONINVASIVE SAR MEASUREMENT USING BOUNDARY 53

Fig. 6. Reconstructed electric currents with/without probe correction.

(a) Without correction. (b) With correction.

Fig. 7. Measurement setup. (a) Geometry. (b) Reconstruction surface.

Fig. 5. Horn antenna used as the probe. (a) Model. (b) Horizontal pattern.
(c) The measured voltage normalized by the antenna factor (upper) and the reconstruction method without/with probe correction are shown.
electric field |E φ | (lower). “Without probe correction” implies that the probe is assumed
to be an ideal Hertzian dipole in the reconstruction procedure.
depicted in absolute values. The electric currents are viewed Comparing these results with the reference in Fig. 4(a), we find
from the −x-axis, and the field distribution is displayed on the a significant difference in the accuracy resulting from the dif-
y = 0 cut plane. By comparing the reconstructed results and ference between the voltages of the probe and the field values.
reference, it is clearly found that the proposed method is highly Thus, the effectiveness and importance of probe correction are
accurate. It demonstrates that our proposed method is valid and confirmed from these results.
effective, including the calculation schemes.
V. EXPERIMENTAL CONFIRMATION
B. Effects of Probe Correction
In this section, the validity of proposed method is confirmed
The effects of the probe correction proposed in Section III-B experimentally. The experiment was performed using a rectan-
are specifically studied. In order to show these effects clearly, gular lossless dielectric object made of FR-4 and a half-wave
the horn antenna illustrated Fig. 5(a) is employed as the probe. dipole antenna placed parallel to the plate. The geometry is il-
Its aperture size is 22.5 cm × 21.4 cm (= 3.75λ × 3.57λ at lustrated in Fig. 7(a). The operating frequency is 2.5 GHz, and
2.5 GHz), and the radiation pattern is shown in Fig. 5(b). the distance between the antenna and the dielectric rectangu-
The receiving voltage U is simulated by the MoM for the lar object is 0.5 λ. We use a typical value of 4 for the relative
probe located around the same test objects in the previous section permittivity for the FR-4 dielectric rectangular object. The S21
[see Fig. 3(a)]. The sampling points are determined in cylindrical values between the DUT and a probe are measured using a vec-
coordinates. The radius of the cylinder at which the sampling tor network analyzer, placing the probe at each of the points
points are set is ρ = 60 cm, the angular interval is φ = 5◦ , and equally distributed on the 2.5λ-radius sphere. The sampling in-
the interval along the z-axis is z = 5 cm. The sampling range terval is θ = φ = 5◦ . From the measured S21 values, the
along the z-axis is set as [−60 cm, 60 cm]. The probe is assumed currents on the surface illustrated in Fig. 7(b) are reconstructed.
to be φ-polarized. In this simulation, mutual coupling, i.e., the A horizontally polarized dipole with a length of 5 cm (5λ/12)
multiple reflection between the test objects and the probe, is not is used as the probe; its far-field pattern was calculated by the
taken in consideration. Fig. 3(c) shows the measured voltage MoM for probe correction. Fig. 8 shows a photograph of the
normalized by the antenna factor (upper) and the electric field measurement setup in an anechoic chamber. The probe is fixed
itself (lower). As shown in the figure, the receiving voltage has to an arm rotated about the vertical (θ ) axis. Horizontal rota-
a narrower pattern along the z-axis because of the narrower tion is realized by an azimuth table, upon which the dielectric
pattern of the probe relative to the ideal Hertzian dipole. rectangular object and source dipole are located.
Fig. 6 shows the electric currents reconstructed from the simu- Fig. 9 shows the reconstructed electric current on the front
lated voltages of the probe. The results obtained by the proposed surface (left) and the electric-field distribution (right) with the

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
54 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 61, NO. 1, FEBRUARY 2019

Fig. 8. Measurement setup in an anechoic chamber.

Fig. 9. Electric current and field distribution. (a) Reference obtained by the
MoM. (b) Reconstructed results. Fig. 10. Electric-current distribution on Spha (left) and electric-field distri-
bution (right) for a lossy sphere. (a) Reference. (b) Results reconstructed by
the proposed method. (c) Reconstructed results when the reconstruction surface
references calculated by an MoM simulation. The right-hand SDUT is large.
side is the normalized distribution in a slice on the yz plane
passing through the center of the rectangular dielectric object. the antenna is estimated slightly (12%) smaller. This error in-
Both sets of reconstructed results are in extremely good agree- creases when the reconstruction surface SDUT is enlarged. An
ment with the reference values and experimentally demonstrate example of this is shown in Fig. 10(c). In this case, the radius
the effectiveness of the proposed method. of the cylindrical reconstruction surface shown in Fig. 3(b) was
enlarged from λ/12 (0.5 cm) to λ/4 (1.5 cm); then, the error
VI. LOSSY DIELECTRIC OBJECT in the peak value increased to 23%. Although it is not shown
here, we have confirmed from other numerical tests that the
In this section, the reconstruction method is applied to the
reconstruction error for lossless dielectric objects is almost in-
case where the phantom is lossy. This is a more realistic case
dependent of the size of the reconstruction surface SDUT . Hence,
for dosimetric problems.
the error characteristics described above are caused by the loss
of the phantom.
A. Numerical Demonstration and Accuracy Degradation
Here, it is discussed why the accuracy degradation occurs
For comparison with the lossless case, the same simulation in the case of the lossy phantom with a wider reconstruction
model in Section IV-A is used, except for the conductivity of surface SDUT . As discussed in Section III-A, the information of
the dielectric sphere. The probe and sampling points are also the the current JDUT on the shadowed area is transmitted to the probe
same. The conductivity is set to σ = 2 S/m, which corresponds through the scattering (refraction or reflection) by the phantom.
to tan δ = 0.36 at 2.5 GHz. However, if the phantom is highly lossy, this information is lost
Fig. 10(b) shows the reconstructed results of the electric cur- during propagation in the phantom. In this case, the information
rent on the surface Spha (left) and the electric-field distribu- from the shadowed area is obtained mainly from the reflection
tion in the sphere (right). In comparison with the reference of the phantom surface, but the reflection is interrupted by SDUT ,
shown in Fig. 10(a), it is found that both distributions are well- especially when it has a wider shape. This information loss leads
reconstructed. However, the peak value at the point closest to to the nonuniqueness of solutions for solving (19), resulting in

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
OMI et al.: RECONSTRUCTION OF INTERNAL FIELD OF DIELECTRIC OBJECTS FOR NONINVASIVE SAR MEASUREMENT USING BOUNDARY 55

developed to use general probe antennas. The validity and effec-

tiveness of the proposed method, including the probe correction
technique, have been numerically and experimentally demon-
strated using dielectric objects located near a dipole antenna. It
has also been shown that the accuracy tends to degrade in the
case of the lossy phantom, especially when the reconstruction
surface enclosing the DUT becomes wider. A relatively simple
method to overcome this problem has been proposed, and its
effectiveness has been numerically confirmed. This method can
Fig. 11. Solutions using the method for improving the accuracy. be used in the experimental situation if a good initial guess has
been given (a solution in the lossless case is employed in this
the lower peak solution because the iterative solver tends to paper).
converge to the solution with the smaller norm. Although our method has been validated only for the homo-
geneous case, the reconstruction method proposed in this paper
B. Method for Improving the Accuracy in the Lossy Case can be extended to the piecewise homogeneous case by mod-
ifying the PMCHW operator to impose field continuity at all
Notwithstanding that a small reconstruction surface is needed
medium boundaries [19]. The application of the reconstruction
for accurate reconstruction in the case of lossy phantoms, this
method and improved method (in Section VI-B) for such case
is not always possible depending on the structure of the DUT.
would be further investigated in the future works.
In this section, we propose a simple but effective method for
improving the accuracy to overcome this difficulty. The method
REFERENCES
is to employ the solution of the lossless case as the initial solu-
tion of lossy one. As can be seen by comparing the results of [1] International Commission on Non-Ionizing Radiation Protection, “Guide-
lines for limiting exposure to time-varying electric, magnetic, and electro-
Figs. 4(a) and 10(a), lossless and lossy phantoms with the same magnetic fields (up to 300 GHz),” Health Phys., vol. 74, no. 4, pp. 494–
permittivity produce a similar current distribution in the shad- 521, 1998.
owed area if they are illuminated by the same source. This fact [2] T. Schmid, O. Egger, and N. Kuster, “Automated E-field scanning system
for dosimetric assessments,” IEEE Trans. Microw. Theory Techn., vol. 44,
supports the assumption that the solution for the lossy phantom no. 1, pp. 105–113, Jan. 1996.
ought to be located near that for the lossless case. Therefore, the [3] O. Bottauscio, M. Chiampi, and L. Zilberti, “A boundary element approach
reconstructed results for the lossless phantom can be used as an to relate surface fields with the specific absorption rate (SAR) induced in
3-D human phantoms,” Eng. Anal. Boundary Elements, vol. 35, no. 4,
initial solution of the iterative method. This idea is equivalent pp. 657–666, 2011.
to solving [4] R. Mitharwal and F. P. Andriulli, “A regularised boundary element for-
mulation for contactless SAR evaluations within homogeneous and inho-
b = A(x0 + r ) (28) mogeneous head phantoms,” Comptes Rendus Physique, vol. 16, no. 9,
pp. 776–788, 2015.
for r , where x0 is the initial solution of the lossless phantom, [5] C. A. Balanis, Advanced Engineering Electromagnetics. New York, NY,
and b and A are the known vector and matrix that correspond to USA: Wiley, 1999.
[6] R. F. Harrington, Time-Harmonic Electromagnetic Fields. New York, NY,
(19). By iteratively solving the original equation in (19) with the USA: McGraw-Hill, 1961.
initial solution x0 or equivalently solving (28) for r , the solution [7] K. Umashankar, A. Taflove, and S. Rao, “Electromagnetic scattering
of the lossy phantom x = x0 + r is determined for r with the by arbitrary shaped three-dimensional homogeneous lossy dielectric ob-
jects,” IEEE Trans. Antennas Propag., vol. 34, no. 6, pp. 758–766, Jun.
smaller norm; that is, the solution closer to the initial one is 1986.
determined. Hence, we can determine the solution of a lossy [8] J. R. Mautz and R. F. Harrington, “Electromagnetic scattering from a
phantom that is closer to that of the lossless one; as a result, the homogeneous material body of revolution,” Archiv Elektronik und Ueber-
tragungstechnik, vol. 33, pp. 71–80, 1979.
current in the shadowed area, which could not be reconstructed, [9] T. F. Eibert and C. H. Schmidt, “Multilevel fast multipole accelerated
is corrected to be similar to that in the lossless case. inverse equivalent current method employing Rao–Wilton–Glisson dis-
In order to confirm the validity of the method, the same wider cretization of electric and magnetic surface currents,” IEEE Trans. Anten-
nas Propag., vol. 57, no. 4, pp. 1178–1185, Apr. 2009.
cylinder model as in Fig. 10(c) is used. Fig. 11 shows the recon- [10] T. F. Eibert, Ismatullah, E. Kaliyaperumal, and C. H. Schmidt, “Inverse
structed results of the lossy phantom using the improved method equivalent surface current method with hierarchical higher order basis
described above. A significant improvement in the accuracy in functions, full probe correction and multilevel fast multipole acceleration,”
Progress Electromagn. Res., vol. 106, pp. 377–394, 2010.
the shadowed area is observed. The error in the maximum value [11] L. Foged, L. Scialacqua, F. Saccardi, J. A. Quijano, G. Vecchi, and M.
decreases to 7%. Sabbadini, “Practical application of the equivalent source method as an
antenna diagnostics tool,” IEEE Antennas Propag. Mag., vol. 54, no. 5,
pp. 243–249, Oct. 2012.
VII. CONCLUSION [12] J. L. A. Quijano, L. Scialacqua, J. Zackrisson, L. J. Foged, M. Sabba-
dini, and G. Vecchi, “Suppression of undesired radiated fields based on
In this paper, a reconstruction method for the internal field equivalent currents reconstruction from measured data,” IEEE Antennas
of dielectric objects based on the BIE has been developed for Wireless Propag. Lett., vol. 10, pp. 314–317, 2011.
the purpose of practical noninvasive SAR measurements. The [13] C. A. Balanis, Antenna Theory: Analysis and Design. New York, NY,
USA: Wiley, 2005.
reconstruction technique proposed here is based on Mitharwal’s [14] J. L. A. Quijano and G. Vecchi, “Field and source equivalence in source
method, which employs the PMCHW formulation. A probe reconstruction on 3D surfaces,” Prog. Electromagn. Res. B, vol. 103,
correction technique based on a plane-wave expansion has been pp. 67–100, 2010.

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.
56 IEEE TRANSACTIONS ON ELECTROMAGNETIC COMPATIBILITY, VOL. 61, NO. 1, FEBRUARY 2019

[15] E. Martini, G. Carli, and S. Maci, “An equivalence theorem based on the Takuji Arima (M’04) received the M.E. and D.E.
use of electric currents radiating in free space,” IEEE Antennas Wireless degrees in engineering from the Tokyo University of
Propag. Lett., vol. 7, pp. 421–424, 2008. Agriculture and Technology (TUAT), Tokyo, Japan,
[16] W. C. Gibson, The Method of Moments in Electromagnetics. Boca Raton, in 1999 and 2002, respectively.
FL, USA: CRC Press, 2014. He is currently an Associate Professor with the
[17] J. Song and W. C. Chew, “Error analysis for the truncation of multipole Department of Electrical and Electronics Engineer-
expansion of vector Green’s functions,” IEEE Microw. Compon. Lett., ing, TUAT, and also a Part-Time Researcher with the
vol. 11, no. 7, pp. 311–313, Jul. 2001. National Institute of Information and Communica-
[18] Y. Saad, Iterative Methods for Sparse Linear Systems. Philadelphia, PA, tions Technology, Tokyo, Japan. His current research
USA: SIAM, 2003. interests include computational electromagnetics and
[19] R. Pogorzelski, “On the numerical computation of scattering from inhomo- metamaterials.
geneous penetrable objects,” IEEE Trans. Antennas Propag., vol. AP-26, Dr. Arima was the recipient of the Young Scientist Award from the IEEE
no. 4, pp. 616–618, Jul. 1978. Antennas and Propagation Society Japan Chapter.

Shuntaro Omi (S’16) received the B.E. and M.E.

degrees from the Tokyo University of Agriculture
and Technology (TUAT), Tokyo, Japan, in 2014 and
2016, respectively. He is currently working toward
the Ph.D. degree with the Department of Electronic
and Information Engineering, TUAT.
His current research interests include antenna mea-
surement techniques.
Mr. Omi was the recipient of the Student Award
from the IEEE Antennas and Propagation Society Joe Wiart (SM’02) received the engineering de-
Japan Chapter. gree from telecom ParisTech, Paris, France, in 1992,
and the Ph.D. and H.D.R. degrees from University
Pierre et Marie Curie, Paris France in 1995 and 2015,
Toru Uno (M’85–SM’02) received the B.S.E.E. de- respectively.
gree from the Tokyo University of Agriculture and He is the holder of the Chair C2M (Car-
Technology (TUAT), Tokyo, Japan, in 1980, and the actérisation, Modélisation et Maitrise) of the Institut
M.S. and Ph.D. degrees in electrical engineering from Mines Telecom - Telecom-ParisTech, Paris, France
Tohoku University, Sendai, Japan, in 1982 and 1985, (http://chairec2m.mines-telecom.fr). He was previ-
respectively. ously the Head of the Dosimetry Research Unit of
He was a Research Associate with the Depart- Orange (former France Telecom). He is the present
ment of Electrical Engineering, Tohoku University, Chairman of the TC106x of the European Committee for Electrotechnical Stan-
from 1985 to 1991, and an Associate Professor from dardization and of International Union of Radio Science commission K. He
1991 to 1994. From 1998 to 1999, he was on leave has been the Chairman of the French chapter of URSI and consultant of In-
from the TUAT. He was a Visiting Scholar with the ternational Commission on Non-Ionizing Radiation Protection. He is emeritus
Department of Electrical Engineering, Pennsylvania State University, State Col- member of The Society of Environmental Engineers since 2008. He has led
lege, PA, USA. He is currently a Professor with the Department of Electrical several national and European projects dedicated to dosimetry (e.g., LEXNET
and Electronics Engineering, TUAT. project http://www.lexnet-project.eu/ 2012–2015).
Dr. Uno served as the Chair of the IEEE Antennas and Propagation So- His research interests include RF dosimetry, numerical methods, and statistic
ciety Japan Chapter from 2005 to 2006 and an Associate Editor of the IEEE applied in electromagnetism and dosimetry. His works gave rise to more than
ANTENNAS AND WIRELESS PROPAGATION LETTERS from 2008 to 2013. He was 120 publications in journal papers and more than 150 communications (among
the Chair of the Technical Group on Antennas and Propagation, Institute of which of numerous invited communications).
Electronics, Information and Communications Engineers (IEICE) from 2011
to 2012, and was a General Chair of the 2016 International Symposium on
Antennas and Propagation. He is currently a Fellow of the IEICE. He was the
recipient of the Young Scientist Award, the Distinguished Contributions Award,
and the Paper Award from the IEICE in 1990, 2006, and 2007, respectively.

Authorized licensed use limited to: Telecom ParisTech. Downloaded on March 28,2024 at 11:17:17 UTC from IEEE Xplore. Restrictions apply.