Está en la página 1de 8
 
1878 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 4, APRIL 2012
A Frequency-Domain Near-Field-to-Far-FieldTransform for Planar Layered Media
İ
lker R. Çapo
ğ
lu
 , Member, IEEE 
, Allen Ta
ove
 , Fellow, IEEE 
, and Vadim Backman
 Abstract— 
We report a frequency-domain near-
eld-to-far-
eldtransform (NFFFT) for the numerical modeling of radiation andscattering in planar multilayered spaces. Although the results areequivalent to those of Demarest
 et al.
 (1996), the formulation ismore compact, more stable, and applicable to observation anglesin the lower half space. Furthermore, the results are presentedin a vector-potential formalism that is more easily adaptable toexisting free-space implementations. The NFFFT algorithm can beused in any differential-equation-based
 
nite numerical method,including but not limited to the
 
nite-difference time-domain(FDTD) method and the
 fi
nite-element method (FEM).
 Index Terms— 
FDTD methods,
 
nite difference methods,
 
niteelement methods, near-
eld far-
eld transformation.
I. I
 NTRODUCTION
N
EAR-FIELD-TO-FAR-FIELD transforms (NFFFTs) in-volve the formulation of the radiated far 
 fi
eld (or Fraun-hofer 
 
eld) in terms of near 
 fi
elds obtained via some
 fi
nite nu-merical method. A free-space frequency-domain near-
eld-to-far-
eld transform (NFFFT) was developed for the
 fi
nite-differ-ence time-domain (FDTD) method by Umashankar and Ta
ove[1]–[3]. This NFFFT algorithm was later generalized to mul-tilayered spaces in [4]. In [5] and [6], a direct time-domain NFFFT algorithm was introduced for the FDTD analysis of three-layered media using a transmission-line (TL) analogy. Inthispaper,wedescribethefrequencydomaingeneralizationandextension of this TL-based approach to general (possibly lossy)layered media. The resulting NFFFT algorithm can be used inconjunction with any
 
nite-difference numerical scheme, in-cluding the FDTD and
 fi
nite-element (FEM) methods. The re-sults are equivalent to those in [4], with the following additionsand improvements:(a) The results are bounded and numerically stable for arbi-trarily-large electrical conductivities within the layeringstructure. This is achieved by putting the recursive rela-tions in a coordinate-independent form that is free of ex- ponentially-increasing terms.
Manuscript received April 28, 2011; revised August 17, 2011; acceptedSeptember 14, 2011. Date of publication January 31, 2012; date of current ver-sion April 06, 2012. This work was supported by the NIH grants R01EB003682and R01CA128641.
İ
. R. Çapo
ğ
lu and V. Backman are with the Biomedical EngineeringDepartment, Northwestern University, Evanston, IL 60208 USA (e-mail:capoglu@ieee.org; v-backman@northwestern.edu).A. Ta
ove is with the Electrical Engineering and Computer ScienceDepartment, Northwestern University, Evanston, IL 60208 USA (e-mail:ta
ove@eecs.northwestern.edu).Digital Object Identi
er 10.1109/TAP.2012.2186253
(b) The far 
eld isexpressed in auniform manner thatisvalidfor an arbitrary observation angle, whether it falls withinthe upper or lower half space.(c) A clear connection is maintained with the traditionalvector-potential integral formulation. This helps thegeneralization of existing free-space NFFFT imple-mentations which are usually formulated in terms of vector-potentials. The software implementer does notneed to change the formulation of the radiated
 
eld based on the vector potentials , [see (3)–(4)], whichis common in existing free-space NFFFTs. The onlycomponent of the software that needs to be updated is theone that calculates the vector potentials , using thescalar transmission-line Green’s functions pertinent to thespeci
c layering structure. The NFFFT is thus split intotwo independent modules, rendering it more amenable tomodularization and code reuse.(d) Explicit formulas are given for up to three layers inclosed form. From a software-implementation perspec-tive, a closed-form solution for a restricted geometry isa convenient starting point before attempting a full gen-eralization to the most complete and complex solution.Afterimplementingandvalidatingthesimplerthree-laye NFFFT, the implementer is equipped with considerableexperience for a more general NFFFT implementation. Inthe debugging stage, a lot of errors in the general NFFFTcan be caught by checking it against the (already tested)three-layer NFFFT in a three-layered geometry.The rest of the paper is organized as follows. In Section II,the theoretical formulation of the NFFFT is explained. A recur-sive algorithm is described for general layered media, and ex- plicit formulas are noted for the special case of three or fewer layers. In Section III, a validation study is presented to con
rmthe accuracy of the method. In Section IV, an FDTD software package featuring the NFFFT in this paper is brie
y introduced.Section V concludes the paper with a summary.II. T
HEORY
The geometry of the problem is shown in Fig. 1 [7]. Thescattering or radiating structure in Fig. 1(a) is embeddedin a planar -layered medium, with relative permittivities, relative permeabilities , and elec-trical conductivities from top to bottom. For the phasor analysis to be valid, the materials that make up thestructure and the layer materials have to be linear. The half space (upper or lower) containing the direction of observation
0018-926X/$31.00 © 2012 IEEE
 
ÇAPO
Ğ
LU
 et al.
: A FREQUENCY-DOMAIN NEAR-FIELD-TO-FAR-FIELD TRANSFORM FOR PLANAR LAYERED MEDIA 1879
Fig.1. Crosssectionoftheradiation/scatteringgeometryinaplanarlayeredmedium.(a)Theoriginalgeometry.(b)Theequivalentgeometrywithsurfacecurrents.
is assumed to be lossless. In other words, if the observa-tion direction is in the upper half space, ; otherwise,. The axis of the Cartesian coordinates is alignedwith the direction of strati
cation, and the coordinates of the layer interfaces from top to bottom are denoted as.A separate formulation for a NFFFT is rarely necessary for integral-equation-based numerical method such as the methodof moments (MoM) because the Green’s functions pertaining tothe layered medium are already available. We therefore assumethat a differential-equation-based
 
nite method is employed,and the structure within the solution space is enclosed by aclosed NFFFT surface as shown in Fig. 1(a). The dimensionsand positioning of the NFFFT surface are arbitrary, as long asthesurfacecontainsthestructure .Ifthe
nitesolutionmethodoperates directly in time-domain (e.g., FDTD), the time-depen-dent tangential electric and magnetic
 
eld components ,on are converted to phasor values , at frequencyusing on-the-
y discrete Fourier transform [3], [8]:(1)where is the time step and is the number of FDTD itera-tions.Thetrueweightofthedesiredfrequencycomponentinthefrequencyspectrumofthewaveformisobtainedbyapplyingtheextra prefactor to the summation representing the Fourietransform.Thesignofthecomplexexponentin(1)followsfromtheelectricalengineeringconvention fortheharmonictime dependence of every
 
eld and source variable. Using thesurface-equivalence principle [9], these tangential phasor 
 fi
eldcomponents are replaced by equivalent surface currents ,radiating in the layered space
 without 
 as shown in Fig. 1(b).The equivalent surface currents are given by(2)in which is the outward normal shown in Fig. 1(a). If the solu-tion method operates directly in frequency-domain (e.g., FEM),then the equivalent surface currents , are already avail-able in phasor form.
 A. Dyadic Green’s Functions for the Vector Potentials
Let the spherical coordinates be centered around theorigin, and the angle variables and denote the longitudinalangle with respect to and the azimuthal angle in the plane,respectively. The and components of radiated electric
 fi
eldat the observation direction speci
ed by the angles can bedirectly obtained from the vector potentials , as follows [9]:(3)(4)in which is the wave impedance of theobservation half space (index 0 or ), and , are therelative permittivity and permeability in the same half space,respectively. In the radiated-
eld zone, the vector potentials can be written as(5)(6)in which primed coordinates denote the positions on the sur-face , is the unit vector in the direction of observation,is the wavenumber in the same direction, isthe lateral position vector , and is the differen-tial area element. The functions and are the asymp-totic forms of the
 dyadic Green’s functions
 pertainingto the speci
c layered geometry. The operator denotes thedyadic-vector dot product, which produces a vector from an-other vector. The axial coordinate does not appear in the phase term in (5)–(6), since the effect of on the far 
eld is en-tirely absorbed in the dyadic Green’s functions and .For free space, a quick comparison with well-known expres-sions reveals that and are proportional to the iden-tity dyadic :(7)The integrals (5)–(6) with (7) form the basis of the well-knownfree-space NFFFT [3]. Solving for the vector potentials in gen-eral layered media requires much more effort. Our approach is
 
1880 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 4, APRIL 2012
 based on expressing the radiated
 
eld in an entirely differentformandcomparingtheresultswith(3)–(4).InAppendixA,theradiated
 fi
eld in the upper and lower half spaces is expressed intermsofone-dimensionalscalar 
transmission-lineGreen’sfunc-tions
 [10], [11] by constructing an analogy between Maxwell’sequations simpli
ed by planar invariance and transmission-linevoltage/current equations. The interested reader is referred to[7] for a more detailed discussion. The necessary informationfor expressing the vector potentials , in terms of the cur-rents , is contained in Appendix A, (47), (49), (61), and(62). By comparison of these equations with (3)–(6), the dyadicGreens functions and are found to be(8)(9)Here, and denote the complexrelative permittivity and the relative permeability at the source point in the layer. The positive and negative signs in frontof the curly braces correspond to observation directions in theupper and lower half spaces, respectively. The scalar functions, are related in a trivial manner to the
 transmis- sion-line voltage/voltage and voltage/current Green’s functions
V     
 ,
 V     
 (see Appendix A):
V     
 (10)As explained in Appendix A,
 V     
 and
 V     
 are thevoltage responses at to an excitation at in theform of an impulsive voltage and current, respectively. Here,is an arbitrary position that is above (or below, for an obser-vation direction in the lower half space) all the radiators, scat-terers, sources, and layer interfaces. It is a temporary value thatwill be seen to eventually drop out of the equations. The de- pendence of , is omitted from the notation for  brevity. We will refer to and as the
 transmis- sion-line voltage and current responses
, respectively. The pa-rameters , are the only part of the formulationthat depends on the speci
c layering. In the following, we de-scribe a straightforward method for computing these functionsfor an arbitrary strati
cation.
 B. Calculation of the Voltage and Current Responses
From a theoretical standpoint, it turns out to be more conve-nienttocalculatetheresponse
inside
thelayersofatransmissionline to an excitation
 away from
 all the layer interfaces, rather than the converse implied in (10). The former is strictly anal-ogous to plane-wave incidence on a planar layered medium; asubject that has been documented thoroughly in the literature[12]. Using the reciprocity properties of the transmission-lineequations [10], the observation and excitation coordinates can be exchanged as follows:
V V     
 (11)
V I    
 (12)The Green’s functions on the right side now represent the re-sponses at the source point to a impulsive voltage or currentat the auxiliary position away from all the sources andlayer interfaces. The voltage functions in (10) can therefore bewritten as
V     
 (13)
 I    
 (14)From here on, we will omit the polarization superscript , be-cause the formulas are exactly the same for the two polariza-tions and except the de
nitions of and [see (59)]. Thecommonequationsforbothpolarizationswillbepresented,withthe implicit assumption that the calculation should be done for the ( ) and ( ) polarizations separately.Since (13)–(14) are in the form of voltage and current wavestraveling on a transmission line, we can assume that in thelayer(seeFig.1),theyareasumofupwardanddownwardprop-agating waves:(15)(16)wheretheupwardanddownwardpropagatingwavesaremarked by the superscripts and , respectively. Although the in-terface coordinates and do not exist, they cancel outin the
 
nal equations, so their values can be chosen arbitrarily(for example, and ). They are includedin the formulation to preserve uniformity in the notation. The propagation constant and the intrinsic impedance in thelayer are given by (58) and (59) with (53). As argued in thediscussion following (59), the correct root for the propagationconstants is the one with a negative imaginary component;for the exponential term would otherwise grow unphysicallywithout bound in a lossy layer ( ) or in the total-internal-re
ection (TIR) regime ( ). With thischoice for , the amplitudes of the exponentials in (15)–(16)always remain smaller than unity, since inthe layer. This will lead to a numerically stable algorithmfor , thatavoids
oating-pointover 
owsevenfor the lossiest layers.As a
 
rst step toward a recursive relation for , , weintroduce the
 total-wave impedance
 at the interfaces of thetransmission line, de
ned as the ratio of the total voltage andthe total current at :
V      I    
 (17)

Recompense su curiosidad

Todo lo que desea leer.
En cualquier momento. En cualquier lugar. Cualquier dispositivo.
Sin compromisos. Cancele cuando quiera.
576648e32a3d8b82ca71961b7a986505