betaFeedback

WorksPhysics and Astronomy

Huygens' principle and field equivalence relations for cylindrical enclosing surfaces

Abstract

Frequency-domain Huygens' principle and equivalence relations for an infinitely long cylindrical surface enclosing the field sources or scatterers are addressed. Assuming the harmonic dependence of the fields along the cylinder axis, we derive three equivalent versions of line-integral representations for the fields in free space at any point outside the enclosing cylinder. In one of the forms, similarly to the Helmholtz-Kirchhoff scalar diffraction theory, the integrals contain a longitudinal field component and its normal derivative. Another presented form contains longitudinal and normal field components, similarly to the three-dimensional boundary integral representations by Stratton and Chu. This form eliminates the need to calculate the field derivatives. Furthermore, we present a two-dimensional version of the three-dimensional Schelkunoff-Franz representation, which contains only tangential components of the fields, complies with the field equivalence theorem and can be regarded as a rigorous formulation of Huygens' principle for a cylindrical enclosing surface. All three forms of the line-equivalence relation are exact and describe the fields through equivalent field distributions on a line enclosing the cross section of the source region. A physical interpretation of Huygens' principle in terms of conical waves emanated by virtual linear sources on the cylindrical surface enclosing the true sources is given. Specialization of the relations to the intermediate and far-field zones are presented. The derived expressions are applicable for studies of scattering and radiation in a very general class of structures that are infinite and periodic along one direction.

Introduction

Huygens’ principle states that any external and induced field sources located in a closed spatial region can be replaced with equivalent sources on any surface enclosing the original sources [1 Ch. 1]. The validity of this principle does not depend on the media filling the space and the sources outside the enclosing surface [2]. The equivalent sources are electric and magnetic surface currents that are defined by the tangential components of the electric and magnetic fields on the enclosing surface. The fundamental reason for these properties is the mathematical form of the Maxwell-Faraday and Ampère-Maxwell laws that relate the time variation of the electric field to the curl of the magnetic field and vice versa.

As soon as the Huygens surface currents are found, the fields outside the enclosing surface can be calculated using conventional techniques. In this paper, we assume free space outside of Huygens’ surface. In this case, Huygens’ principle can also be formulated as the field equivalence theorem [3 Sec. 6.14], [4, Sec. 1.7], [5, Sec. 3.5], [6], [7, Sec. 7.8]. In contrast to the concept of Huygens’ sources, these formulations assume a homogeneous medium outside the Huygens surface. The derivations are based on the uniqueness theorem for electromagnetic fields and state that the field of actual sources can be replaced by fictitious (or virtual) sources distributed over the surface enclosing the actual sources. The virtual sources are related to the tangential components of the electric and magnetic fields on the boundary, i.e. to Huygens’ sources.

Practical applications of the equivalence theorem require a corresponding representation of fields at any point through their boundary values. By using Green’s functions the fields in a volume free of sources can be expressed as surface integrals of their boundary values. For scalar wave fields, the corresponding integral relation is known as Helmholtz-Kirchhoff integral theorem (see [8 Ch. 2], [9, Sec. 8.13] and [10 Sec. 8.3.1]). For electromagnetic fields in the frequency domain, two versions of the boundary integral relations are available: the Stratton-Chu formula [9, Sec. 8.14], [11] and the Schelkunoff-Franz formula [3 Sec. 6.1], [12], [13], [14] and [15, Sec. 25].1 The latter contains only the tangential field components and therefore can be seen as one based on true Huygens’ sources.

There is an extensive literature on 3D equivalence principle and related integral formulas, e.g. [16, § 16], [17, Sec. 5.3], [18], [19, § 1.4], [20, Sec. 1.6], [21, 22]. The principle finds applications in many practical engineering problems, including the design of antennas, radomes and metasurfaces, e.g. [23, 24, 25, 26].

In a number of applications, it is convenient to assume that the structure consists of cylindrical bodies infinite in one direction, e.g. a line current, a wire, a transmission line, an open waveguide, a periodic array, a section of an aircraft, etc., thus making the problem two-dimensional. Recently, two-dimensional models have been extensively used in studies of scanning reflectarrays and reconfigurable intelligent services, e.g. [27, 28, 29, 30]. The 3D boundary integral relations assume, however, a compact enclosing surface and do not apply to the case of infinitely long structures. A special treatment of the problems with cylindrical geometry is therefore necessary.

A line-integral representation for a scalar solution of the two-dimensional Helmholtz equation has been derived in the past by using a two-dimensional version of the divergence theorem (also known as Gauss’, Green’s or Ostrogradsky’s theorem) [8 Sec. 6.2], [31]. A closed integration line separates the sources from the observation point, and the integral gives the solution of the Helmholtz equation in the source-free region and vanishes in the complementary region. This result can be seen as an equivalence relation for the electromagnetic fields independent of the zz coordinate.

Generalization to the electromagnetic case, in which the fields depend on zz, can be carried out by considering the dependence in the form of the factor exp⁡(−jkzz)\exp(-\mathrm{j}k_z z), where kzk_z is a constant. For example, kz=kcos⁡βk_z = k\cos\beta for a plane wave incident under the angle β\beta to the zz axis. For a waveguide, kz=∓jγk_z = \mp\mathrm{j}\gamma, where γ\gamma is the propagation constant. The fields with a general dependence on zz can be expanded in a Fourier spectrum with kzk_z being the spectral variable, e.g. in a series of Floquet modes when the cylindrical structures are periodic in the zz direction.

Assuming the translational symmetry of the enclosing surface, homogeneous and isotropic medium and the exponential dependence of the fields in the outside, Maxwell’s equations get split into two polarization cases, TM and TE, each fully described by a single field component, EzE_z or HzH_z, respectively [9 Sec. 6.1 and 6.2]. The components satisfy the two-dimensional Helmholtz equations, which permits deriving a line-equivalence relation for every polarization with the integration line being the directrix of the enclosing cylindrical surface. These relations have been derived in [32 Sec. 2.6.2] by adjusting the approach from [2] to the 2D Helmholtz equation.

The obtained 2D integral relations can be seen as a rigorous formulation of Huygens’ principle for an infinite cylindrical enclosing surface, though such an interpretation has not been put forward in the literature so far. A possible reason can be that in contrast to the 3D electromagnetic equivalence theorem, which can be expressed entirely in terms of the electric and magnetic field components tangential to the integration surface, the 2D version derived in [32 Sec. 2.6.2] involves the normal derivatives of the field components as in the Helmholtz-Kirchhoff integral theorem.

A further problem with the normal derivative is that if the fields on the enclosing contour are obtained from a numerical solution, then numerical evaluation of the derivative seems to be necessary, which could lead to losses in accuracy when using the equivalence formula.

In this paper, three equivalence relations for infinite cylindrical enclosing surfaces and the electromagnetic fields with the dependence exp⁡(−jkzz)\exp(-\mathrm{j}k_z z) are presented. In § 1 the 2D divergence formula is applied to the scalar solutions of the 2D Helmholtz equation, which recovers the result obtained in [32 Sec. 2.6.2] by a slightly different approach. In § 1, a modification of the line-equivalence formula is presented, which does not include the normal derivatives of the fields on the integration line. Section IV demonstrates that the Stratton-Chu integral representation recovers the line-equivalence relation from § 2 in the limit of an infinite cylindrical enclosing surface. In § 4, the Schenkelkunoff-Franz integral representation is transformed to a line-equivalence relation that includes only the tangential field components.

The formulas apply to cylindrical enclosing surfaces with arbitrarily shaped cross sections and can be used for extending the fields obtained by a numerical procedure on the boundary of a computational domain into the exterior of the domain, including the infinitely distant point. The line-integral representations can be significantly simplified when the observation point is moved at a sufficient distance from the integration contour. Section VI presents simplified expressions for an observer located at a distance greater than a wavelength from the integration line but not in the far-field region yet. The specification of the derived line-equivalence formulas to the 2D far-field limit is presented in § 6. The presented formulas describe the far-field characteristics of cylindrical geometries (scattering amplitudes, scattering widths). § 7 gives numerical examples of the application of the presented line-equivalence relations to the field radiated by a line source and the field scattered by a homogeneous dielectric cylinder.

The time factor exp⁡(jωt)\exp(\mathrm{j}\omega t) is supposed and omitted throughout.

Derivation of the equivalence relation in 2D

Consider a cylindrical surface Γ\Gamma of infinite extent oriented along the zz axis of a Cartesian coordinate system, which encloses a body or bodies of arbitrary material constitution and geometric shape, infinite cylindrical or finite (Figure 1). The bodies can be either radiating (e.g. a line current) or scattering (e.g. a metal or dielectric cylinder), in which case they are referred to as primary (or imprinted) and secondary sources, respectively. The medium outside the enclosing surface is assumed homogeneous with constant permittivity ε\varepsilon and permeability μ\mu. Because the fields can be expanded into Fourier series with respect to the zz coordinate, without loss of generality the dependence of the fields on zz can be assumed to be of the form exp⁡(−jkzz)\exp(-\mathrm{j}k_z z), where kzk_z is a constant. Under these conditions, the fields outside of Γ\Gamma split up into two polarization cases, TM or E case with Hz=0H_z = 0, and TE or H case with Ez=0E_z = 0, and can be fully described in terms of a single component EzE_z or HzH_z, respectively (e.g. [9 Sec. 6.1, 6.2], [15 Sec. 8]). The components satisfy the two-dimensional Helmholtz equation

Infinite cylindrical surface $\Gamma$ enclosing sources and scatterers

Figure 1. An infinite cylindrical surface Γ\Gamma along the zz axis enclosing variously shaped sources or/and scatterers (here a wire and patches of an array); the line CC is the directrix of Γ\Gamma.

(∇⊥2+k⊥2)U(ρ)=0,(1)(\nabla_{\perp}^{2}+k_{\perp}^{2})U(\rho)=0, \tag*{(1)}

where U(ρ)=Ez(ρ,z)exp⁡(jkzz)U(\rho)=E_z(\rho,z)\exp(\mathrm{j}k_z z) (TM) or U(ρ)=Hz(ρ,z)exp⁡(jkzz)U(\rho)=H_z(\rho,z)\exp(\mathrm{j}k_z z) (TE), ρ\rho is the position-vector in the plane perpendicular to the zz axis, ∇⊥2\nabla_{\perp}^{2} is the two-dimensional Laplacian, k⊥2=k2−kz2k_{\perp}^{2}=k^{2}-k_z^{2} and k=ωεμk=\omega\sqrt{\varepsilon\mu} is the wave number. The rest components of the fields can be determined from EzE_z and HzH_z, e.g. [32 Sec. 2.5.5] for Cartesian and circular cylindrical coordinates or (24)–(27) for general orthogonal curvilinear coordinates.

The two-dimensional wave function U(ρ)U(\boldsymbol{\rho}) is defined on an infinite plane in the exterior of the geometric cross section of the cylindrical region, in which the field sources are located. For ρ→∞\rho\to\infty, U(ρ)U(\boldsymbol{\rho}) is required to be an outgoing wave, which implies that ∣U(ρ)∣→0\lvert U(\boldsymbol{\rho})\rvert\to0 if Im⁡k<0\operatorname{Im} k < 0 (condition at infinity).

Denote the cross section of the cylindrical region containing all the considered sources by AinA_{\mathrm{in}}, its boundary by CC and the region outside by AexA_{\mathrm{ex}} (Figure 2). The line CC is therefore the directrix of the cylindrical region Γ\Gamma enclosing all sources. In order to express U(ρ)U(\boldsymbol{\rho}) in AexA_{\mathrm{ex}} through its boundary values, the Green function of the infinite plane is required, which is a solution of the equation

Geometry of the equivalence relation

Figure 2. Geometry assumed in the formulation of the equivalence relation: the region AinA_{\mathrm{in}} bounded by the contour CC contains all the sources; AexA_{\mathrm{ex}} complements AinA_{\mathrm{in}} to the full xx-yy plane; CRC_R is a circle of a large radius RR.

(∇⊥2+k⊥2)G2(ρ,ρ0,k⊥)=−δ(x−x0)δ(y−y0)(2)\left(\nabla_{\perp}^{2}+k_{\perp}^{2}\right)G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})=-\delta(x-x_{0})\delta(y-y_{0}) \tag*{(2)}

with the delta source at a point ρ0=x0x^+y0y^\boldsymbol{\rho}_{0}=x_{0}\hat{\mathbf{x}}+y_{0}\hat{\mathbf{y}}. The solution of (2), which is compliant with the condition at infinity, is given by

G2(ρ,ρ0,k⊥)=14jH0(2)(k⊥∣ρ−ρ0∣),(3)G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})=\frac{1}{4j}H_{0}^{(2)}(k_{\perp}\lvert\boldsymbol{\rho}-\boldsymbol{\rho}_{0}\rvert), \tag*{(3)}

where H0(2)(⋯ )H_{0}^{(2)}(\cdots) is the Hankel function and ∣ρ−ρ0∣\lvert\boldsymbol{\rho}-\boldsymbol{\rho}_{0}\rvert is the distance between the points. Multiply (1) with G2(ρ,ρ0,k⊥)G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp}), (2) with U(ρ)U(\boldsymbol{\rho}) and subtract from each other to obtain the relation

G2(ρ,ρ0,k⊥)∇⊥2U(ρ)−U(ρ)∇⊥2G2(ρ,ρ0,k⊥)=U(ρ)δ(x−x0)δ(y−y0).(4)G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})\nabla_{\perp}^{2}U(\boldsymbol{\rho})-U(\boldsymbol{\rho})\nabla_{\perp}^{2}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})=U(\boldsymbol{\rho})\delta(x-x_{0})\delta(y-y_{0}). \tag*{(4)}

The left-hand side is the divergence on the two-dimensional vector field

F=G2(ρ,ρ0,k⊥)∇⊥U(ρ)−U(ρ)∇⊥G2(ρ,ρ0,k⊥),(5)\mathbf{F}=G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})\nabla_{\perp}U(\boldsymbol{\rho})-U(\boldsymbol{\rho})\nabla_{\perp}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp}), \tag*{(5)}

implying that one can integrate (4) over AexA_{\mathrm{ex}} and apply the two-dimensional version of the divergence theorem [33 Sec. 1.4]

∬Aex(∇⊥⋅F) dA=−∮C(F⋅n^) dl,(6)\iint_{A_{\mathrm{ex}}}(\nabla_{\perp}\cdot\mathbf{F})\,dA=-\oint_{C}(\mathbf{F}\cdot\hat{\mathbf{n}})\,dl, \tag*{(6)}

where n^\hat{\mathbf{n}} is the unit normal to the boundary CC pointing to AexA_{\mathrm{ex}} (Figure 2), dAdA is the infinitesimal surface element in the surface integral and dldl is the infinitesimal arc length in the curvilinear integral.

A mathematically correct way of applying (6) to an infinite region AexA_{\mathrm{ex}} consists in integrating first over its large but bounded part Aex,RA_{\mathrm{ex},R}

∮C∪CR[U(ρ)∂∂nG2(ρ,ρ0,k⊥)−G2(ρ,ρ0,k⊥)∂∂nU(ρ)] dl={U(ρ0),if ρ0∈Aex,R,0,if ρ0∈Ain,(7)\oint_{C\cup C_{R}}\left[U(\boldsymbol{\rho})\frac{\partial}{\partial n}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})-G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}_{0},k_{\perp})\frac{\partial}{\partial n}U(\boldsymbol{\rho})\right]\,dl= \begin{cases} U(\boldsymbol{\rho}_{0}), & \text{if }\boldsymbol{\rho}_{0}\in A_{\mathrm{ex},R},\\ 0, & \text{if }\boldsymbol{\rho}_{0}\in A_{\mathrm{in}}, \end{cases} \tag*{(7)}

where ∂/∂n\partial/\partial n is the derivative in the direction of the unit normal n^\hat{\mathbf{n}} and CRC_{R} is a circle of a large radius RR so as to enclose the source region AinA_{\mathrm{in}}. Then, letting RR go to infinity and accounting for the condition at infinity transform (7) into

∮C[U(ρ′)∂∂n′G2(ρ,ρ′,k⊥)−G2(ρ,ρ′,k⊥)∂∂n′U(ρ′)] dl′={U(ρ),if ρ∈Aex,0,if ρ∈Ain,(8)\oint_{C}\left[U(\boldsymbol{\rho}^{\prime})\frac{\partial}{\partial n^{\prime}}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})-G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})\frac{\partial}{\partial n^{\prime}}U(\boldsymbol{\rho}^{\prime})\right]\,dl^{\prime}= \begin{cases} U(\boldsymbol{\rho}), & \text{if }\boldsymbol{\rho}\in A_{\mathrm{ex}},\\ 0, & \text{if }\boldsymbol{\rho}\in A_{\mathrm{in}}, \end{cases} \tag*{(8)}

where the observation point is renamed to ρ\boldsymbol{\rho}, the integration point to ρ′\boldsymbol{\rho}^{\prime}, and the dash means the dependence on the position of the integration point, e.g., n^′=n^(ρ′)\hat{\mathbf{n}}^{\prime}=\hat{\mathbf{n}}(\boldsymbol{\rho}^{\prime}). The normal derivative of Green’s function evaluates to

∂∂n′G2(ρ,ρ′,k⊥)=k⊥4jH1(2)(k⊥∣ρ−ρ′∣)n^′⋅s^⊥(ρ,ρ′),(9)\frac{\partial}{\partial n^{\prime}}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})=\frac{k_{\perp}}{4j}H_{1}^{(2)}(k_{\perp}\lvert\boldsymbol{\rho}-\boldsymbol{\rho}^{\prime}\rvert)\hat{\mathbf{n}}^{\prime}\cdot\hat{\mathbf{s}}_{\perp}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime}), \tag*{(9)}

where

s^⊥(ρ,ρ′)=ρ−ρ′∣ρ−ρ′∣(10)\hat{\mathbf{s}}_{\perp}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime})=\frac{\boldsymbol{\rho}-\boldsymbol{\rho}^{\prime}}{\lvert\boldsymbol{\rho}-\boldsymbol{\rho}^{\prime}\rvert} \tag*{(10)}

is the unit vector in the plane perpendicular to the axis of the cylindrical region Γ\Gamma.

Specifying (8) with ρ∈Aex\boldsymbol{\rho}\in A_{\mathrm{ex}} to the zz-components of the fields leads to the relations:

Ez(ρ,z)=∮C[Ez(ρ′,z)∂∂n′G2(ρ,ρ′,k⊥)−G2(ρ,ρ′,k⊥)∂∂n′Ez(ρ′,z)] dl′(11)E_{z}(\boldsymbol{\rho},z)=\oint_{C}\left[E_{z}(\boldsymbol{\rho}^{\prime},z)\frac{\partial}{\partial n^{\prime}}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})-G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})\frac{\partial}{\partial n^{\prime}}E_{z}(\boldsymbol{\rho}^{\prime},z)\right]\,dl^{\prime} \tag*{(11)}
Hz(ρ,z)=∮C[Hz(ρ′,z)∂∂n′G2(ρ,ρ′,k⊥)−G2(ρ,ρ′,k⊥)∂∂n′Hz(ρ′,z)] dl′.(12)H_{z}(\boldsymbol{\rho},z)=\oint_{C}\left[H_{z}(\boldsymbol{\rho}^{\prime},z)\frac{\partial}{\partial n^{\prime}}G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})-G_{2}(\boldsymbol{\rho},\boldsymbol{\rho}^{\prime},k_{\perp})\frac{\partial}{\partial n^{\prime}}H_{z}(\boldsymbol{\rho}^{\prime},z)\right]\,dl^{\prime}. \tag*{(12)}

Equations (11) and (12) hold for every value of zz and are in fact independent of zz as the factors exp⁡(−jkzz)\exp(-jk_z z) in both sides of the equations cancel out. Other field components result from Maxwell’s equations (see (24)–(27) below).

Equations (11) and (12) can be seen as field equivalence relations in the case of infinite cylindrical enclosing surface. They state that the zz component of the field, either EzE_z or HzH_z, at any point outside of the source region can be exactly expressed through its value and the value of its normal derivative on the boundary of the source region at the same value of zz. The boundary values can be seen as the equivalent sources that produce the same field outside AinA_{\rm in} as the real sources. Note that the shape and extent of the source region AinA_{\rm in} can be freely chosen as long as AinA_{\rm in} contains all the sources and the material parameters in AexA_{\rm ex} are position-independent. Furthermore, the boundary CC must not necessarily be smooth and can include corner points, e.g. when AinA_{\rm in} is a square region. At the corner points the normal vector n^\hat{\mathbf{n}} is undefined but the contribution of any single point to the integral is zero as long as the integrand is integrable at that point. Finally, the left-hand sides of (11) and (12) are equal to zero for an observer in AinA_{\rm in}.

A complimentary formulation, in which the sources are located in AexA_{\rm ex}, is also possible. It has the same form as (11) and (12), the only difference being that the line integrals vanish for an observer in AexA_{\rm ex} and are equal to the fields due to the sources (with the sign reversed) when ρ∈Ain\rho\in A_{\rm in}.

A modified form of the equivalence relation

The equivalence formula (8) can be used for extending a solution from a closed line, at which it is known, to any point beyond, including the infinitely distant point (ρ→∞\rho\to\infty). The extension requires, however, the knowledge of UU and ∂U/∂n\partial U/\partial n on the boundary, which can be a problem when the boundary field is available in a numeric form because numerical evaluation of the derivative may result in the loss of accuracy. A way around would be the use of a Green function vanishing on CC, which would eliminate the term with the normal derivative from (8). The problem is however that such Green’s functions are unavailable for arbitrarily shaped boundaries and when available, e.g. for a circle or ellipse, the corresponding analytical expressions are much more complicated compared with the Green function of a full plane (3). Another special case is that in which a scattering problem is considered for a cylindrical scatterer with the Neumann boundary condition ∂U/∂n=0\partial U/\partial n = 0 or a boundary condition of the form ∂U/∂n+δU=0\partial U/\partial n + \delta U = 0, where δ\delta is a known parameter. Then the surface Γ\Gamma can be chosen to coincide with the physical boundary of the scatterer, and the boundary conditions can be used to exclude the normal derivative from the integrand.

In this Section it will be shown that in the most general case, i.e. for arbitrarily shaped line CC, the terms ∂Ez/∂n\partial E_z/\partial n and ∂Hz/∂n\partial H_z/\partial n can be expressed through the field components normal and tangential to CC, which will make the modified 2D equivalence relation suitable for integration of numerical data.

The problem can be solved by writing down Maxwell’s equations in the outside of Γ\Gamma in a suitable curvilinear coordinate system. Let u1(x,y)u_1(x,y), u2(x,y)u_2(x,y) and u3=zu_3=z be a right-handed system of orthogonal curvilinear coordinates, and let u^i\hat{\mathbf{u}}_i be a unit vector normal to the coordinate surface uiu_i, oriented toward increasing uiu_i, and such that u^1×u^2=u^3\hat{\mathbf{u}}_1 \times\hat{\mathbf{u}}_2 = \hat{\mathbf{u}}_3, where u^3=z^\hat{\mathbf{u}}_3=\hat{\mathbf{z}}. For any cylindrical geometry, the metric coefficients h1h_1, h2h_2 and h3h_3 are such that h3=1h_3=1 and h1h_1 and h2h_2 do not depend on zz. The vector fields E\mathbf{E} and H\mathbf{H} are given by E=E1u^1+E2u^2+Ezu^3\mathbf{E}=E_1\hat{\mathbf{u}}_1+E_2\hat{\mathbf{u}}_2+E_z\hat{\mathbf{u}}_3 and H=H1u^1+H2u^2+Hzu^3\mathbf{H}=H_1\hat{\mathbf{u}}_1+H_2\hat{\mathbf{u}}_2+H_z\hat{\mathbf{u}}_3 with the components Ei=u^i⋅EE_i=\hat{\mathbf{u}}_i\cdot\mathbf{E} and Hi=u^i⋅HH_i=\hat{\mathbf{u}}_i\cdot\mathbf{H}.

Choose the coordinates u1(x,y)u_1(x,y) and u2(x,y)u_2(x,y) such that the line u1(x,y)=const⁡u_1(x,y)=\operatorname{const} coincides with the line CC. The coordinate system can be obtained, e.g., by conformal mapping of a circle onto the region AinA_{\rm in}. Now, using ∂/∂u3=−jkz\partial/\partial u_3=-jk_z, write the curls in the chosen curvilinear coordinates as [9 Sec. 1.16, 6.1 and 6.2]

∇×E=1h2(∂Ez∂u2+jkzh2E2)u^1−1h1(jkzh1E1+∂Ez∂u1)u^2+1h1h2[∂∂u1(h2E2)−∂∂u2(h1E1)]z^(13)\nabla\times\mathbf{E} = \frac{1}{h_2}\left(\frac{\partial E_z}{\partial u_2}+jk_z h_2 E_2\right)\hat{\mathbf{u}}_1 -\frac{1}{h_1}\left(jk_z h_1 E_1+\frac{\partial E_z}{\partial u_1}\right)\hat{\mathbf{u}}_2 +\frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}(h_2E_2)-\frac{\partial}{\partial u_2}(h_1E_1)\right]\hat{\mathbf{z}} \tag*{(13)}
∇×H=1h2(∂Hz∂u2+jkzh2H2)u^1−1h1(jkzh1H1+∂Hz∂u1)u^2+1h1h2[∂∂u1(h2H2)−∂∂u2(h1H1)]z^(14)\nabla\times\mathbf{H} = \frac{1}{h_2}\left(\frac{\partial H_z}{\partial u_2}+jk_z h_2 H_2\right)\hat{\mathbf{u}}_1 -\frac{1}{h_1}\left(jk_z h_1 H_1+\frac{\partial H_z}{\partial u_1}\right)\hat{\mathbf{u}}_2 +\frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}(h_2H_2)-\frac{\partial}{\partial u_2}(h_1H_1)\right]\hat{\mathbf{z}} \tag*{(14)}

and insert these into Maxwell’s equations to obtain

jωεE1=1h2(∂Hz∂u2+jkzh2H2)(15)j\omega\varepsilon E_1 = \frac{1}{h_2}\left(\frac{\partial H_z}{\partial u_2}+jk_z h_2H_2\right) \tag*{(15)}
jωεE2=−1h1(jkzh1H1+∂Hz∂u1)(16)j\omega\varepsilon E_2 = -\frac{1}{h_1}\left(jk_z h_1H_1+\frac{\partial H_z}{\partial u_1}\right) \tag*{(16)}
jωεEz=1h1h2[∂∂u1(h2H2)−∂∂u2(h1H1)](17)j\omega\varepsilon E_z = \frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}(h_2H_2)-\frac{\partial}{\partial u_2}(h_1H_1)\right] \tag*{(17)}
−jωμH1=1h2(∂Ez∂u2+jkzh2E2)(18)-j\omega\mu H_1 = \frac{1}{h_2}\left(\frac{\partial E_z}{\partial u_2}+jk_z h_2E_2\right) \tag*{(18)}
−jωμH2=−1h1(jkzh1E1+∂Ez∂u1)(19)-j\omega\mu H_2 = -\frac{1}{h_1}\left(jk_z h_1E_1+\frac{\partial E_z}{\partial u_1}\right) \tag*{(19)}
−jωμHz=1h1h2[∂∂u1(h2E2)−∂∂u2(h1E1)].(20)-j\omega\mu H_z = \frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}(h_2E_2)-\frac{\partial}{\partial u_2}(h_1E_1)\right]. \tag*{(20)}

Noting that

∇⋅E=1h1h2[∂∂u1(h2E1)+∂∂u2(h1E2)]−jkzEz(21)\nabla\cdot\mathbf{E} = \frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}(h_2E_1)+\frac{\partial}{\partial u_2}(h_1E_2)\right]-jk_zE_z \tag*{(21)}

and using the Maxwell equation ∇⋅(εE)=0\nabla\cdot(\varepsilon\mathbf{E})=0 with the position-independent ε\varepsilon, equations (18) and (19) can be used in (17) to obtain the Helmholtz equation for EzE_z

1h1h2[∂∂u1(h2h1∂Ez∂u1)+∂∂u2(h1h2∂Ez∂u2)]+k⊥2Ez=0.(22)\frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}\left(\frac{h_2}{h_1}\frac{\partial E_z}{\partial u_1}\right)+\frac{\partial}{\partial u_2}\left(\frac{h_1}{h_2}\frac{\partial E_z}{\partial u_2}\right)\right]+k_\perp^2E_z=0. \tag*{(22)}

The Helmholtz equation for HzH_z

1h1h2[∂∂u1(h2h1∂Hz∂u1)+∂∂u2(h1h2∂Hz∂u2)]+k⊥2Hz=0(23)\frac{1}{h_1h_2}\left[\frac{\partial}{\partial u_1}\left(\frac{h_2}{h_1}\frac{\partial H_z}{\partial u_1}\right)+\frac{\partial}{\partial u_2}\left(\frac{h_1}{h_2}\frac{\partial H_z}{\partial u_2}\right)\right]+k_\perp^2H_z=0 \tag*{(23)}

results from (15), (16) and (20) in a similar manner.

Consider now (15), (16), (18) and (19). These are four equations involving eight quantities: E1E_1, E2E_2, H1H_1, H2H_2, ∂Ez/∂u1\partial E_z/\partial u_1, ∂Ez/∂u2\partial E_z/\partial u_2, ∂Hz/∂u1\partial H_z/\partial u_1 and ∂Hz/∂u2\partial H_z/\partial u_2. So, on the one hand, one can use them to express the components E1E_1, E2E_2, H1H_1 and H2H_2 through the derivatives of EzE_z and HzH_z as

E1=−jk⊥2(kzh1∂Ez∂u1+ωμh2∂Hz∂u2)(24)E_1=-\frac{j}{k_\perp^2}\left(\frac{k_z}{h_1}\frac{\partial E_z}{\partial u_1}+\frac{\omega\mu}{h_2}\frac{\partial H_z}{\partial u_2}\right) \tag*{(24)}
E2=−jk⊥2(kzh1∂Ez∂u2−ωμh1∂Hz∂u1)(25)E_2=-\frac{j}{k_\perp^2}\left(\frac{k_z}{h_1}\frac{\partial E_z}{\partial u_2}-\frac{\omega\mu}{h_1}\frac{\partial H_z}{\partial u_1}\right) \tag*{(25)}
H1=−jk⊥2(kzh1∂Hz∂u1−ωεh2∂Ez∂u2)(26)H_1=-\frac{j}{k_\perp^2}\left(\frac{k_z}{h_1}\frac{\partial H_z}{\partial u_1}-\frac{\omega\varepsilon}{h_2}\frac{\partial E_z}{\partial u_2}\right) \tag*{(26)}
H2=−jk⊥2(kzh2∂Hz∂u2+ωεh1∂Ez∂u1).(27)H_2=-\frac{j}{k_\perp^2}\left(\frac{k_z}{h_2}\frac{\partial H_z}{\partial u_2}+\frac{\omega\varepsilon}{h_1}\frac{\partial E_z}{\partial u_1}\right). \tag*{(27)}

Equations (24)–(27) permit determining the rest field components from the basic components EzE_z and HzH_z.

On the other hand, (15), (16), (18) and (19) can be used to express the derivatives through the components

1h1∂Ez∂u1=j(ωμH2−kzE1)(28)\frac{1}{h_1}\frac{\partial E_z}{\partial u_1}=j\left(\omega\mu H_2-k_zE_1\right) \tag*{(28)}
1h1∂Hz∂u1=−j(ωεE2+kzH1)(29)\frac{1}{h_1}\frac{\partial H_z}{\partial u_1}=-j\left(\omega\varepsilon E_2+k_zH_1\right) \tag*{(29)}
1h2∂Ez∂u2=−j(ωμH1+kzE2)(30)\frac{1}{h_2}\frac{\partial E_z}{\partial u_2}=-j\left(\omega\mu H_1+k_zE_2\right) \tag*{(30)}
1h2∂Hz∂u2=j(ωεE1−kzH2).(31)\frac{1}{h_2}\frac{\partial H_z}{\partial u_2}=j\left(\omega\varepsilon E_1-k_zH_2\right). \tag*{(31)}

Equations (28)–(31) relate the normal and tangential derivatives of the zz components to the normal and tangential components of the fields on the surface Γ\Gamma. Choose u^1=n^\hat{u}_1=\hat{n} and u^2=t^\hat{u}_2=\hat{t}, where n^\hat{n} is the unit normal pointing to the exterior of Γ\Gamma and t^\hat{t} is the unit tangent such that n^×t^=z^\hat{n}\times\hat{t}=\hat{z}. The normal derivative can now be written as

1h1∂f∂u1=(n^,∇f)=∂f∂n,(32)\frac{1}{h_1}\frac{\partial f}{\partial u_1}=(\hat{n},\nabla f)=\frac{\partial f}{\partial n}, \tag*{(32)}

and it follows from (28) and (29) that

∂Ez∂n=j(ωμHt−kzEn)(33)\frac{\partial E_z}{\partial n}=j\left(\omega\mu H_t-k_zE_n\right) \tag*{(33)}
∂Hz∂n=−j(ωεEt+kzHn).(34)\frac{\partial H_z}{\partial n}=-j\left(\omega\varepsilon E_t+k_zH_n\right). \tag*{(34)}

Here, En=n^⋅EE_n=\hat{n}\cdot E, Et=t^⋅EE_t=\hat{t}\cdot E, Hn=n^⋅HH_n=\hat{n}\cdot H and Ht=t^⋅HH_t=\hat{t}\cdot H. Thus, using (33) and (34) in (11) and (12) leads to the modified equivalence relations:

Ez(ρ,z)=∮C{Ez(ρ′,z)∂∂n′G2(ρ,ρ′,k⊥)−j[ωμHt′(ρ′,z)−kzEn′(ρ′,z)]G2(ρ,ρ′,k⊥)} dl′(35)E_z(\boldsymbol{\rho},z)=\oint_C\left\{E_z(\boldsymbol{\rho}',z)\frac{\partial}{\partial n'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)-j\left[\omega\mu H_t'(\boldsymbol{\rho}',z)-k_zE_n'(\boldsymbol{\rho}',z)\right]G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)\right\}\,dl' \tag*{(35)}
Hz(ρ,z)=∮C{Hz(ρ′,z)∂∂n′G2(ρ,ρ′,k⊥)+j[ωεEt′(ρ′,z)+kzHn′(ρ′,z)]G2(ρ,ρ′,k⊥)} dl′(36)H_z(\boldsymbol{\rho},z)=\oint_C\left\{H_z(\boldsymbol{\rho}',z)\frac{\partial}{\partial n'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)+j\left[\omega\varepsilon E_t'(\boldsymbol{\rho}',z)+k_zH_n'(\boldsymbol{\rho}',z)\right]G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)\right\}\,dl' \tag*{(36)}

for TM and TE polarization respectively. Equations (35) and (36) assume that the observation point ρ\boldsymbol{\rho} is in the region AexA_{\mathrm{ex}}. With ρ∈Ain\boldsymbol{\rho}\in A_{\mathrm{in}} the left-hand sides are zeros. In general, the equivalence relations include both the tangential and the normal components. In the limit kz=0k_z=0 (fields independent of zz), the normal components EnE_n and HnH_n disappear from (35) and (36) and only tangential components remain.

Stratton-Chu’s Representation in the 2D Limit

Another way of deriving the line-equivalence relations is to use the 3D boundary integral representations of electromagnetic fields. The procedure requires two steps: going to the limit of an infinite cylindrical integration region and integration with respect to zz in the surface integrals. There are two fundamental integral representations – the one by Stratton and Chu and another by Schelkunoff and Franz (see § 0.1). In this Section, the Stratton-Chu representation is considered, and it will be shown that the formulation leads to the line-equivalence relations (35) and (36).

Stratton-Chu’s formulas describe the fields radiated by the sources enclosed by a surface Ω\Omega and are given by the relations

E(r)=∯Ω[−jωμG0(r,r′)n^′×H′+(n^′⋅E′)∇′G0(r,r′)+(n^′×E′)×∇′G0(r,r′)] dΩ′(37)\mathbf{E}(\mathbf{r})=\oiint_{\Omega}\left[-j\omega\mu G_0(\mathbf{r},\mathbf{r}')\hat{\mathbf{n}}'\times\mathbf{H}'+\left(\hat{\mathbf{n}}'\cdot\mathbf{E}'\right)\nabla'G_0(\mathbf{r},\mathbf{r}')+\left(\hat{\mathbf{n}}'\times\mathbf{E}'\right)\times\nabla'G_0(\mathbf{r},\mathbf{r}')\right]\,d\Omega' \tag*{(37)}
H(r)=∯Ω[jωεG0(r,r′)n^′×E′+(n^′⋅H′)∇′G0(r,r′)+(n^′×H′)×∇′G0(r,r′)] dΩ′,(38)\mathbf{H}(\mathbf{r})=\oiint_{\Omega}\left[j\omega\varepsilon G_0(\mathbf{r},\mathbf{r}')\hat{\mathbf{n}}'\times\mathbf{E}'+\left(\hat{\mathbf{n}}'\cdot\mathbf{H}'\right)\nabla'G_0(\mathbf{r},\mathbf{r}')+\left(\hat{\mathbf{n}}'\times\mathbf{H}'\right)\times\nabla'G_0(\mathbf{r},\mathbf{r}')\right]\,d\Omega', \tag*{(38)}

where r\mathbf{r} and r′\mathbf{r}' are 3D position vectors of an observation and an integration point, respectively, ∇′\nabla' is the 3D nabla operator acting on the primed coordinates and

G0(r,r′)=e−jk∣r−r′∣4π∣r−r′∣.(39)G_0(\mathbf{r},\mathbf{r}')=\frac{e^{-jk\lvert\mathbf{r}-\mathbf{r}'\rvert}}{4\pi\lvert\mathbf{r}-\mathbf{r}'\rvert}. \tag*{(39)}

The vector products of the fields with the unit normal describe the field components tangential to Ω\Omega and are referred to as equivalent surface currents,

Ke=n^×H,Km=E×n^,(40)\mathbf{K}_e=\hat{\mathbf{n}}\times\mathbf{H},\qquad\mathbf{K}_m=\mathbf{E}\times\hat{\mathbf{n}}, \tag*{(40)}

electric and magnetic, respectively. It is sufficient to consider (37) for the electric field. The corresponding result for the magnetic field will follow from the duality of Maxwell’s equations by the substitutions: ε→μ\varepsilon\to\mu, μ→ε\mu\to\varepsilon, E→HE\to H, H→−EH\to-E, Ke→Km\mathbf{K}_e\to\mathbf{K}_m and Km→−Ke\mathbf{K}_m\to-\mathbf{K}_e.

Equations (37) and (38) assume a compact integration surface. To extend them to the case of an infinite cylindrical surface Γ\Gamma, consider Ω\Omega as a cylindrical region of a finite length LL and let LL go to infinity. In the limit, the contributions from the ends of the cylindrical region are required to vanish, implying that the dependence of the fields on zz can be assumed to be exp⁡(−jkzz)\exp(-jk_z z) for an infinite cylindrical boundary, which permits separated integration with respect to z′z' in (37). In the analytical calculation that follows, it is convenient to assume that kzk_z is real-valued. This limitation is removed in the final result as the factor exp⁡(−jkzz)\exp(-jk_z z) is factored out.

In the limit of the infinite cylindrical structure, by using the relations

E(ρ,z′)=e−jkzz′E(ρ,0),H(ρ,z′)=e−jkzz′H(ρ,0)(41)E(\boldsymbol{\rho},z')=e^{-jk_z z'}E(\boldsymbol{\rho},0),\quad H(\boldsymbol{\rho},z')=e^{-jk_z z'}H(\boldsymbol{\rho},0) \tag*{(41)}

to extract the zz-dependence from the boundary fields, (37) can be represented as

E(ρ,z)=−jωμ∮CI0(r,ρ′)Ke(ρ′,0) dl′−∮CKm(ρ′,0)×I1(r,ρ′) dl′+∮CI1(r,ρ′)n^′⋅E(ρ′,0) dl′(42)\begin{aligned} E(\boldsymbol{\rho},z)={}&-j\omega\mu\oint_C I_0(\mathbf r,\boldsymbol{\rho}')\mathbf K_e(\boldsymbol{\rho}',0)\,dl'\\ &-\oint_C\mathbf K_m(\boldsymbol{\rho}',0)\times I_1(\mathbf r,\boldsymbol{\rho}')\,dl'\\ &+\oint_C I_1(\mathbf r,\boldsymbol{\rho}')\hat{\mathbf n}'\cdot E(\boldsymbol{\rho}',0)\,dl' \tag*{(42)} \end{aligned}

where

I0(r,ρ′)=∫−∞+∞e−jkzz′G0(r,r′) dz′(43)I_0(\mathbf r,\boldsymbol{\rho}')=\int_{-\infty}^{+\infty}e^{-jk_z z'}G_0(\mathbf r,\mathbf r')\,dz' \tag*{(43)}
I1(r,ρ′)=∫−∞+∞e−jkzz′∇′G0(r,r′) dz′.(44)I_1(\mathbf r,\boldsymbol{\rho}')=\int_{-\infty}^{+\infty}e^{-jk_z z'}\nabla'G_0(\mathbf r,\mathbf r')\,dz'. \tag*{(44)}

Integral (43) can be evaluated analytically [32 Eq. (2.412)] as

I0(r,ρ′)=e−jkzzG2(ρ,ρ′,k⊥).(45)I_0(\mathbf r,\boldsymbol{\rho}')=e^{-jk_z z}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp). \tag*{(45)}

Green’s function (39) is symmetric with respect to its arguments, implying that

I1(r,ρ′)=−∇I0(r,ρ′)=e−jkzz[jkzG2(ρ,ρ′,k⊥)z^−∇⊥G2(ρ,ρ′,k⊥)].(46)I_1(\mathbf r,\boldsymbol{\rho}')=-\nabla I_0(\mathbf r,\boldsymbol{\rho}')=e^{-jk_z z}\left[jk_zG_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)\hat{\mathbf z}-\nabla_\perp G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)\right]. \tag*{(46)}

The line-integral representation for the magnetic field results from (42) by using duality,

H(ρ,z)=−jωε∮CI0(r,ρ′)Km(ρ′,0) dl′+∮CKe(ρ′,0)×I1(r,ρ′) dl′+∮CI1(r,ρ′)n^′⋅H(ρ′,0) dl′(47)\begin{aligned} H(\boldsymbol{\rho},z)={}&-j\omega\varepsilon\oint_C I_0(\mathbf r,\boldsymbol{\rho}')\mathbf K_m(\boldsymbol{\rho}',0)\,dl'\\ &+\oint_C\mathbf K_e(\boldsymbol{\rho}',0)\times I_1(\mathbf r,\boldsymbol{\rho}')\,dl'\\ &+\oint_C I_1(\mathbf r,\boldsymbol{\rho}')\hat{\mathbf n}'\cdot H(\boldsymbol{\rho}',0)\,dl' \tag*{(47)} \end{aligned}

Equations (42) and (47) provide a version of the line-equivalence relations for cylindrical geometries. From (45) and (46) it is seen that the factors exp⁡(−jkzz)\exp(-jk_z z) in both sides of (42) and (47) can be canceled out, which makes the derived formulas applicable to complex-valued kzk_z. On the other hand, the exponential factor in (45) and (46) implies that (42) and (47) relate the fields at the same value of zz.

Consider the zz components of (42) and (47). Inserting (45) and (46) into (42) and (47) and using the relations

Ke(ρ′,0)=−Hz(ρ′,0)t^′+Ht′(ρ′,0)z^(48)\mathbf K_e(\boldsymbol{\rho}',0)=-H_z(\boldsymbol{\rho}',0)\hat{\mathbf t}'+H_{t'}(\boldsymbol{\rho}',0)\hat{\mathbf z} \tag*{(48)}
Km(ρ′,0)=Ez(ρ′,0)t^′−Et′(ρ′,0)z^(49)\mathbf K_m(\boldsymbol{\rho}',0)=E_z(\boldsymbol{\rho}',0)\hat{\mathbf t}'-E_{t'}(\boldsymbol{\rho}',0)\hat{\mathbf z} \tag*{(49)}
z^⋅[Km(ρ′,0)×I1(r,ρ′)]=−Ez(ρ′,0)∂∂n′G2(ρ,ρ′,k⊥)(50)\hat{\mathbf z}\cdot[\mathbf K_m(\boldsymbol{\rho}',0)\times I_1(\mathbf r,\boldsymbol{\rho}')]=-E_z(\boldsymbol{\rho}',0)\frac{\partial}{\partial n'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp) \tag*{(50)}
z^⋅[Ke(ρ′,0)×I1(r,ρ′)]=Hz(ρ′,0)∂∂n′G2(ρ,ρ′,k⊥)(51)\hat{\mathbf z}\cdot[\mathbf K_e(\boldsymbol{\rho}',0)\times I_1(\mathbf r,\boldsymbol{\rho}')]=H_z(\boldsymbol{\rho}',0)\frac{\partial}{\partial n'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp) \tag*{(51)}

recover expression (35) and (36) for the zz component of the electric and magnetic fields if relations (41) are used.

Schelkunoff-Franz’ Formula in the 2D Limit

The field equivalence theorem suggests the equivalent sources entirely in terms of the field components tangential to the integration surface, namely n^×H\hat{\mathbf n}\times H and n^×E\hat{\mathbf n}\times E. The Stratton-Chu formulas (37) and (38) include, however, the normal components of the fields n^⋅H\hat{\mathbf n}\cdot H and n^⋅E\hat{\mathbf n}\cdot E, which means that despite of being exact, the formulas are not based on the true Huygens’ sources. The integral representations of fields published in [12] and [14] contain only the tangential components,

E(r)=1jωε[k2+∇(∇⋅)]∬ΩG0(r,r′)n^′×H′ dΩ′+∇×∬ΩG0(r,r′)n^′×E′ dΩ′(52)E(\mathbf r)=\frac{1}{j\omega\varepsilon}\left[k^2+\nabla(\nabla\cdot)\right]\iint_\Omega G_0(\mathbf r,\mathbf r')\hat{\mathbf n}'\times H'\,d\Omega' +\nabla\times\iint_\Omega G_0(\mathbf r,\mathbf r')\hat{\mathbf n}'\times E'\,d\Omega' \tag*{(52)}
H(r)=−1jωμ[k2+∇(∇⋅)]∬ΩG0(r,r′)n^′×E′ dΩ′+∇×∬ΩG0(r,r′)n^′×H′ dΩ′.(53)H(\mathbf r)=-\frac{1}{j\omega\mu}\left[k^2+\nabla(\nabla\cdot)\right]\iint_\Omega G_0(\mathbf r,\mathbf r')\hat{\mathbf n}'\times E'\,d\Omega' +\nabla\times\iint_\Omega G_0(\mathbf r,\mathbf r')\hat{\mathbf n}'\times H'\,d\Omega'. \tag*{(53)}

Here, a compact surface Ω\Omega encloses all sources, and the observation point is located in the complementary source-free region. For an observation point inside Ω\Omega the left-hand sides of (52) and (53) are equal to zero.

The limit of (52) and (53) for infinite cylindrical regions can be studied along the same lines as in the case of Stratton-Chu representation (sec. IV). Again, because of the duality, it is sufficient to consider only one of the equations (52) and (53). Assuming that Ω\Omega is a cylindrical region of a finite length LL, letting LL go to infinity, requiring vanishing contributions from the ends of the region and using the formula (45) transform (52) into the following equation:

E(ρ,z)=1jωε∮CA(r,ρ′) dl′−∮CB(r,ρ′) dl′(54)E(\boldsymbol{\rho},z)=\frac{1}{j\omega\varepsilon}\oint_C A(\mathbf r,\boldsymbol{\rho}')\,dl'-\oint_C B(\mathbf r,\boldsymbol{\rho}')\,dl' \tag*{(54)}

where

A(r,ρ′)=[k2+∇(∇⋅)][I0(r,ρ′)Ke(ρ′,0)](55)A(\mathbf r,\boldsymbol{\rho}')=[k^2+\nabla(\nabla\cdot)][I_0(\mathbf r,\boldsymbol{\rho}')\mathbf K_e(\boldsymbol{\rho}',0)] \tag*{(55)}
B(r,ρ′)=∇×[I0(r,ρ′)Km(ρ′,0)](56)B(\mathbf r,\boldsymbol{\rho}')=\nabla\times[I_0(\mathbf r,\boldsymbol{\rho}')\mathbf K_m(\boldsymbol{\rho}',0)] \tag*{(56)}

and the definitions (40) for the equivalent electric and magnetic currents are used. In (55) and (56), ∇\nabla operates on ρ\boldsymbol{\rho} and zz but not on ρ′\boldsymbol{\rho}', implying that

∇×[I0(r,ρ′)Km(ρ′,0)]=∇I0(r,ρ′)×Km(ρ′,0)(57)\nabla\times[I_0(\mathbf r,\boldsymbol{\rho}')\mathbf K_m(\boldsymbol{\rho}',0)]=\nabla I_0(\mathbf r,\boldsymbol{\rho}')\times\mathbf K_m(\boldsymbol{\rho}',0) \tag*{(57)}
∇(∇⋅)[I0(r,ρ′)Ke(ρ′,0)]=∇[Ke(ρ′,0)⋅∇I0(r,ρ′)].(58)\nabla(\nabla\cdot)[I_0(\mathbf r,\boldsymbol{\rho}')\mathbf K_e(\boldsymbol{\rho}',0)]=\nabla[\mathbf K_e(\boldsymbol{\rho}',0)\cdot\nabla I_0(\mathbf r,\boldsymbol{\rho}')]. \tag*{(58)}

and (54) takes the form

E(ρ,z)=1jωε∮C{k⊥2I0(r,ρ′)Ke(ρ′,0)−∇[Ke(ρ′,0)⋅I1(r,ρ′)]} dl′+∮CI1(r,ρ′)×Km(ρ′,0) dl′(59)\mathbf{E}(\boldsymbol{\rho},z)=\frac{1}{j\omega\varepsilon}\oint_C\left\{k_\perp^2\mathbf{I}_0(\mathbf{r},\boldsymbol{\rho}')K_{\mathrm{e}}(\boldsymbol{\rho}',0)-\nabla\left[K_{\mathrm{e}}(\boldsymbol{\rho}',0)\cdot\mathbf{I}_1(\mathbf{r},\boldsymbol{\rho}')\right]\right\}\,dl'+\oint_C\mathbf{I}_1(\mathbf{r},\boldsymbol{\rho}')\times\mathbf{K}_{\mathrm{m}}(\boldsymbol{\rho}',0)\,dl' \tag*{(59)}

The line-integral representation for HH results from (59) by using the duality,

H(ρ,z)=1jωμ∮C{k⊥2I0(r,ρ′)Km(ρ′,0)−∇[Km(ρ′,0)⋅I1(r,ρ′)]} dl′−∮CI1(r,ρ′)×Ke(ρ′,0) dl′(60)\mathbf{H}(\boldsymbol{\rho},z)=\frac{1}{j\omega\mu}\oint_C\left\{k_\perp^2\mathbf{I}_0(\mathbf{r},\boldsymbol{\rho}')K_{\mathrm{m}}(\boldsymbol{\rho}',0)-\nabla\left[K_{\mathrm{m}}(\boldsymbol{\rho}',0)\cdot\mathbf{I}_1(\mathbf{r},\boldsymbol{\rho}')\right]\right\}\,dl'-\oint_C\mathbf{I}_1(\mathbf{r},\boldsymbol{\rho}')\times\mathbf{K}_{\mathrm{e}}(\boldsymbol{\rho}',0)\,dl' \tag*{(60)}

In (59) and (60) the factor exp⁡(−jkzz)\exp(-jk_z z) can be moved from I0I_0 and I1I_1 to Ke,m(ρ′,0)K_{\mathrm{e,m}}(\boldsymbol{\rho}',0) to get a form in which the fields on both sides of the equations are taken at the same value of zz. Similarly to (42) and (47), equations (59) and (60) are fully vectorial and describe all components of the fields. To compare with the line-equivalence relations (11), (12), (35) and (36), let us calculate the zz components of (59) and (60). The vector quantities (55) and (56) can be analytically calculated by using (3), (48) and (49) with the following result for their zz components

Az(r,ρ′)=−e−jkzzk⊥4[jk⊥Ht′(ρ′,0)H0(2)(k⊥∣ρ−ρ′∣)+kzHz(ρ′,0)H1(2)(k⊥∣ρ−ρ′∣)t^′⋅s^⊥(ρ,ρ′)](61)A_z(\mathbf{r},\boldsymbol{\rho}')=-e^{-jk_z z}\frac{k_\perp}{4}\left[jk_\perp H_t'(\boldsymbol{\rho}',0)H_0^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)+k_zH_z(\boldsymbol{\rho}',0)H_1^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)\hat{\mathbf{t}}'\cdot\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}')\right] \tag*{(61)}
Bz(r,ρ′)=−e−jkzzk⊥4jEz(ρ′,0)H1(2)(k⊥∣ρ−ρ′∣)n^′⋅s^⊥(ρ,ρ′)(62)B_z(\mathbf{r},\boldsymbol{\rho}')=-e^{-jk_z z}\frac{k_\perp}{4j}E_z(\boldsymbol{\rho}',0)H_1^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)\hat{\mathbf{n}}'\cdot\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}') \tag*{(62)}

where s^⊥(ρ,ρ′)\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}') is defined in (10). The zz projection of (59) is therefore equal to

Ez(ρ,z)=−k⊥4j∮C[jk⊥ωεHt′(ρ′,z)H0(2)(k⊥∣ρ−ρ′∣)+kzωεHz(ρ′,z)H1(2)(k⊥∣ρ−ρ′∣)t^′⋅s^⊥(ρ,ρ′)−Ez(ρ′,z)H1(2)(k⊥∣ρ−ρ′∣)n^′⋅s^⊥(ρ,ρ′)] dl′(63)E_z(\boldsymbol{\rho},z)=-\frac{k_\perp}{4j}\oint_C\left[\frac{jk_\perp}{\omega\varepsilon}H_t'(\boldsymbol{\rho}',z)H_0^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)+\frac{k_z}{\omega\varepsilon}H_z(\boldsymbol{\rho}',z)H_1^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)\hat{\mathbf{t}}'\cdot\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}')-E_z(\boldsymbol{\rho}',z)H_1^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)\hat{\mathbf{n}}'\cdot\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}')\right]\,dl' \tag*{(63)}

Using (9) and

∂∂t′G2(ρ,ρ′,k⊥)=k⊥4jH1(2)(k⊥∣ρ−ρ′∣)t^′⋅s^⊥(ρ,ρ′)(64)\frac{\partial}{\partial t'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)=\frac{k_\perp}{4j}H_1^{(2)}(k_\perp\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert)\hat{\mathbf{t}}'\cdot\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}') \tag*{(64)}

leads to the expressions:

Ez(ρ,z)=∮C[Ez(ρ′,z)∂∂n′G2(ρ,ρ′,k⊥)−jk⊥2ωεHt′(ρ′,z)G2(ρ,ρ′,k⊥)−kzωεHz(ρ′,z)∂∂t′G2(ρ,ρ′,k⊥)] dl′(65)E_z(\boldsymbol{\rho},z)=\oint_C\left[E_z(\boldsymbol{\rho}',z)\frac{\partial}{\partial n'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)-j\frac{k_\perp^2}{\omega\varepsilon}H_t'(\boldsymbol{\rho}',z)G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)-\frac{k_z}{\omega\varepsilon}H_z(\boldsymbol{\rho}',z)\frac{\partial}{\partial t'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)\right]\,dl' \tag*{(65)}
Hz(ρ,z)=∮C[Hz(ρ′,z)∂∂n′G2(ρ,ρ′,k⊥)+jk⊥2ωμEt′(ρ′,z)G2(ρ,ρ′,k⊥)+kzωμEz(ρ′,z)∂∂t′G2(ρ,ρ′,k⊥)] dl′(66)H_z(\boldsymbol{\rho},z)=\oint_C\left[H_z(\boldsymbol{\rho}',z)\frac{\partial}{\partial n'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)+j\frac{k_\perp^2}{\omega\mu}E_t'(\boldsymbol{\rho}',z)G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)+\frac{k_z}{\omega\mu}E_z(\boldsymbol{\rho}',z)\frac{\partial}{\partial t'}G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp)\right]\,dl' \tag*{(66)}

It is seen that (65) and (66) are different from (11), (12), (35) and (36) as they include only tangential field components. Moreover, (65) and (66) contain the tangential derivative of the Green function in addition to the Green function and its normal derivative. Representations (59) and (60) describe the fields outside of the source region entirely in terms of the tangential field components on the enclosing cylindrical surface Γ\Gamma. These relations can be seen as the line-equivalence relations that use truly Huygens’ equivalent sources. For an observation point inside the region CC the left-hand sides of (59), (60), (65) and (66) are equal to zero, which is the characteristic feature of Huygens’ sources.

The intermediate region

The line-equivalence relations derived before simplify as soon as the observation point ρ\boldsymbol{\rho} is removed from the integration contour CC. The simplifications are associated with the following conditions:

d≫1∣k⊥∣(67)d\gg\frac{1}{\lvert k_\perp\rvert} \tag*{(67)}
d≫D(68)d\gg D \tag*{(68)}
d≫∣k⊥∣D2(69)d\gg\lvert k_\perp\rvert D^2 \tag*{(69)}

where d=min⁡ρ′∈C∣ρ−ρ′∣d=\min_{\boldsymbol{\rho}'\in C}\lvert\boldsymbol{\rho}-\boldsymbol{\rho}'\rvert is the shortest distance between the observer and the line CC, and DD is the diameter of the region enclosed by CC. Condition (67) is the most important one since it permits the most significant simplifications. For electrically small regions AinA_{\mathrm{in}} (k⊥D≪1k_\perp D\ll1), conditions (68) and (69) are automatically met once (67) is true. For an electrically large region AinA_{\mathrm{in}} (k⊥D≫1k_\perp D\gg1), conditions (68) and (69) are essential additional conditions that are satisfied at greater distances from CC than required by (67). In the latter case, the approximations that follow describe the fields in the intermediate region, i.e. at the distances between the distance at which (67) is satisfied and the distance at which (68) and (69) are met. Note that (67) and (69) contain k⊥k_\perp which even in the high-frequency case (k≫1k\gg1) can be small if kz≈kk_z\approx k, i.e. for waves propagating along the axis of the cylindrical region. On the other hand, if kzk_z is small or zero, (67) has the meaning of the high-frequency approximation (d≫λ/2πd \gg\lambda/2\pi), while condition (69) is the far-zone condition for large scatterers or antennas (d≫D2/λd \gg D^2/\lambda).

In this Section, we consider the situation, in which condition (67) is satisfied, without any assumption about (68) and (69). The Hankel function in the Green function (3) can be replaced by the large-argument approximation (e.g. [34 Eq. (9.2.4)])

G2(ρ,ρ′,k⊥)≈G2as(ρ,ρ′,k⊥)=e−jk⊥∣ρ−ρ′∣−jπ/48πk⊥∣ρ−ρ′∣(70)G_2(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp) \approx G_2^{\mathrm{as}}(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp) = \frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|-j\pi/4}}{\sqrt{8\pi k_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|}} \tag*{(70)}

implying that

I0(r,ρ′)≈e−jkzzG2as(ρ,ρ′,k⊥)(71)I_0(\mathbf{r},\boldsymbol{\rho}') \approx e^{-jk_z z}G_2^{\mathrm{as}}(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp) \tag*{(71)}

and

I1(r,ρ′)≈jks′e−jkzzG2as(ρ,ρ′,k⊥),(72)I_1(\mathbf{r},\boldsymbol{\rho}') \approx jk'_s e^{-jk_z z}G_2^{\mathrm{as}}(\boldsymbol{\rho},\boldsymbol{\rho}',k_\perp), \tag*{(72)}

where ks′=k⊥s^⊥(ρ,ρ′)+kzz^\mathbf{k}'_s=k_\perp\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}')+k_z\hat{\mathbf{z}} is a vector with the length kk. Using (41) and (70)–(72) in the line-equivalence relations of the Stratton-Chu type, (42) and (47), gives

E(ρ,z)≈ejπ/48πk⊥∮Ce−jk⊥∣ρ−ρ′∣∣ρ−ρ′∣[−ωμKe(ρ′,z)−Km(ρ′,z)×ks′+n^′⋅E(ρ′,z)ks′] dl′(73)\mathbf{E}(\boldsymbol{\rho},z) \approx\frac{e^{j\pi/4}}{\sqrt{8\pi k_\perp}}\oint_C\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}{\sqrt{|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}\left[-\omega\mu\mathbf{K}_e(\boldsymbol{\rho}',z)-\mathbf{K}_m(\boldsymbol{\rho}',z)\times\mathbf{k}'_s+\hat{\mathbf{n}}'\cdot\mathbf{E}(\boldsymbol{\rho}',z)\mathbf{k}'_s\right]\,dl' \tag*{(73)}
H(ρ,z)≈ejπ/48πk⊥∮Ce−jk⊥∣ρ−ρ′∣∣ρ−ρ′∣[−ωεKm(ρ′,z)+Ke(ρ′,z)×ks′+n^′⋅H(ρ′,z)ks′] dl′.(74)\mathbf{H}(\boldsymbol{\rho},z) \approx\frac{e^{j\pi/4}}{\sqrt{8\pi k_\perp}}\oint_C\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}{\sqrt{|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}\left[-\omega\varepsilon\mathbf{K}_m(\boldsymbol{\rho}',z)+\mathbf{K}_e(\boldsymbol{\rho}',z)\times\mathbf{k}'_s+\hat{\mathbf{n}}'\cdot\mathbf{H}(\boldsymbol{\rho}',z)\mathbf{k}'_s\right]\,dl'. \tag*{(74)}

The line-equivalence relations of the Schelkunoff-Franz type (59) and (60) simplify to

E(ρ,z)≈ejπ/48πk⊥∮Ce−jk⊥∣ρ−ρ′∣∣ρ−ρ′∣{ks′×Km(ρ′,z)+1ωεks′×[ks′×Ke(ρ′,z)]} dl′(75)\mathbf{E}(\boldsymbol{\rho},z) \approx\frac{e^{j\pi/4}}{\sqrt{8\pi k_\perp}}\oint_C\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}{\sqrt{|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}\left\{\mathbf{k}'_s\times\mathbf{K}_m(\boldsymbol{\rho}',z)+\frac{1}{\omega\varepsilon}\mathbf{k}'_s\times\left[\mathbf{k}'_s\times\mathbf{K}_e(\boldsymbol{\rho}',z)\right]\right\}\,dl' \tag*{(75)}
H(ρ,z)≈ejπ/48πk⊥∮Ce−jk⊥∣ρ−ρ′∣∣ρ−ρ′∣{−ks′×Ke(ρ′,z)+1ωμks′×[ks′×Km(ρ′,z)]} dl′,(76)\mathbf{H}(\boldsymbol{\rho},z) \approx\frac{e^{j\pi/4}}{\sqrt{8\pi k_\perp}}\oint_C\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}{\sqrt{|\boldsymbol{\rho}-\boldsymbol{\rho}'|}}\left\{-\mathbf{k}'_s\times\mathbf{K}_e(\boldsymbol{\rho}',z)+\frac{1}{\omega\mu}\mathbf{k}'_s\times\left[\mathbf{k}'_s\times\mathbf{K}_m(\boldsymbol{\rho}',z)\right]\right\}\,dl', \tag*{(76)}

where the relation

ks′×[ks′×K]=ks′(ks′⋅K)−k2K(77)\mathbf{k}'_s\times\left[\mathbf{k}'_s\times\mathbf{K}\right]=\mathbf{k}'_s\left(\mathbf{k}'_s\cdot\mathbf{K}\right)-k^2\mathbf{K} \tag*{(77)}

with K\mathbf{K} being either Ke\mathbf{K}_e or Km\mathbf{K}_m has been used.

The approximate equations (73)–(76) represent the fields of the enclosed sources as superpositions of conical waves originated at every point ρ′\boldsymbol{\rho}' of the line CC with the amplitude factor as in a cylindrical wave and the phase function k⊥∣ρ−ρ′∣+kzzk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|+k_z z that describes a conical wave front, e.g. [32 Sec. 2.5]. So, Huygens’ principle in the case of cylindrical geometries sees the outgoing wave as a superposition of conical waves emanated by every element of the enclosing cylindrical surface.

The Far-Field Region

The region, where all conditions (67)–(69) are met, will be referred to as the far-field region. Let us take a look at the line-equivalence relations (73)–(76) in the case that in addition to (67), further conditions are satisfied. If (68) is true, then ∣ρ−ρ′∣|\boldsymbol{\rho}-\boldsymbol{\rho}'| in the denominator can be replaced with ∣ρ−ρc∣|\boldsymbol{\rho}-\boldsymbol{\rho}_c|, where ρc\boldsymbol{\rho}_c is the geometric center of the region, s^⊥(ρ,ρ′)\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}') with s^⊥c=s^⊥(ρ,ρc)\hat{\mathbf{s}}_\perp^c=\hat{\mathbf{s}}_\perp(\boldsymbol{\rho},\boldsymbol{\rho}_c) and ks′\mathbf{k}'_s with ksc=k⊥s^⊥c+kzz^\mathbf{k}_s^c=k_\perp\hat{\mathbf{s}}_\perp^c+k_z\hat{\mathbf{z}}, which do not depend on the position of the integration point.

At distances, at which the condition (69) is satisfied, the approximation

∣ρ−ρ′∣=∣ρ−ρc∣−s^⊥c⋅(ρ′−ρc)+O(D2/d)(78)|\boldsymbol{\rho}-\boldsymbol{\rho}'|=|\boldsymbol{\rho}-\boldsymbol{\rho}_c|-\hat{\mathbf{s}}_\perp^c\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_c)+O(D^2/d) \tag*{(78)}

can be used in the phase factors of (73)–(76), implying that

e−jk⊥∣ρ−ρ′∣≈e−jk⊥∣ρ−ρc∣+jk⊥s^⊥c⋅(ρ′−ρc).(79)e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}'|}\approx e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|+jk_\perp\hat{\mathbf{s}}_\perp^c\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_c)}. \tag*{(79)}

Note that for an electrically large region AinA_{\mathrm{in}} (∣k⊥∣D≫1|k_\perp|D\gg1), the condition (69) is much more severe than (67) and (68).

With (68) and (69) satisfied, (73)–(76) become

E(ρ,z)≈e−jk⊥∣ρ−ρc∣+jπ/48πk⊥∣ρ−ρc∣∮Cejk⊥s^⊥c⋅(ρ′−ρc)[ksc×Km(ρ′,z)−ωμKe(ρ′,z)+n^′⋅E(ρ′,z)ksc] dl′(80)\mathbf{E}(\boldsymbol{\rho},z)\approx\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|+j\pi/4}}{\sqrt{8\pi k_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|}}\oint_C e^{jk_\perp\hat{\mathbf{s}}_\perp^c\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_c)}\left[\mathbf{k}_s^c\times\mathbf{K}_m(\boldsymbol{\rho}',z)-\omega\mu\mathbf{K}_e(\boldsymbol{\rho}',z)+\hat{\mathbf{n}}'\cdot\mathbf{E}(\boldsymbol{\rho}',z)\mathbf{k}_s^c\right]\,dl' \tag*{(80)}
H(ρ,z)≈e−jk⊥∣ρ−ρc∣+jπ/48πk⊥∣ρ−ρc∣∮Cejk⊥s^⊥c⋅(ρ′−ρc)[Ke(ρ′,z)×ksc−ωεKm(ρ′,z)+n^′⋅H(ρ′,z)ksc] dl′(81)\mathbf{H}(\boldsymbol{\rho},z)\approx\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|+j\pi/4}}{\sqrt{8\pi k_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|}}\oint_C e^{jk_\perp\hat{\mathbf{s}}_\perp^c\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_c)}\left[\mathbf{K}_e(\boldsymbol{\rho}',z)\times\mathbf{k}_s^c-\omega\varepsilon\mathbf{K}_m(\boldsymbol{\rho}',z)+\hat{\mathbf{n}}'\cdot\mathbf{H}(\boldsymbol{\rho}',z)\mathbf{k}_s^c\right]\,dl' \tag*{(81)}

and

E(ρ,z)≈e−jk⊥∣ρ−ρc∣+jπ/48πk⊥∣ρ−ρc∣ksc×∮Cejk⊥s^⊥c⋅(ρ′−ρc)[1ωεksc×Ke(ρ′,z)+Km(ρ′,z)] dl′(82)\mathbf{E}(\boldsymbol{\rho},z)\approx\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|+j\pi/4}}{\sqrt{8\pi k_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|}}\mathbf{k}_s^c\times\oint_C e^{jk_\perp\hat{\mathbf{s}}_\perp^c\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_c)}\left[\frac{1}{\omega\varepsilon}\mathbf{k}_s^c\times\mathbf{K}_e(\boldsymbol{\rho}',z)+\mathbf{K}_m(\boldsymbol{\rho}',z)\right]\,dl' \tag*{(82)}
H(ρ,z)≈e−jk⊥∣ρ−ρc∣+jπ/48πk⊥∣ρ−ρc∣ksc×∮Cejk⊥s^⊥c⋅(ρ′−ρc)[1ωμksc×Km(ρ′,z)−Ke(ρ′,z)] dl′.(83)\mathbf{H}(\boldsymbol{\rho},z)\approx\frac{e^{-jk_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|+j\pi/4}}{\sqrt{8\pi k_\perp|\boldsymbol{\rho}-\boldsymbol{\rho}_c|}}\mathbf{k}_s^c\times\oint_C e^{jk_\perp\hat{\mathbf{s}}_\perp^c\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_c)}\left[\frac{1}{\omega\mu}\mathbf{k}_s^c\times\mathbf{K}_m(\boldsymbol{\rho}',z)-\mathbf{K}_e(\boldsymbol{\rho}',z)\right]\,dl'. \tag*{(83)}

It is seen that (80)–(83) describe conical waves with the phase center at ρ=ρc\boldsymbol{\rho}=\boldsymbol{\rho}_c. Furthermore, the fields E\mathbf{E} and H\mathbf{H} have no components in the direction of the vector ksc\mathbf{k}_s^c, which is apparent from the Schelkunoff-Franz line-equivalence relations (82) and (83).

Finally, when the observation point is removed at a distance which is much greater than the distance between the phase center and the origin of the employed coordinate frame, i.e. ρ≫ρc\rho\gg\rho_c, where ρ=∣ρ∣\rho=|\boldsymbol{\rho}| and ρc=∣ρc∣\rho_c=|\boldsymbol{\rho}_c|, the following simplifications become valid:

s^⊥c≈ρρ=ρ^,∣ρ−ρc∣≈ρ−ρ^⋅ρc,(84)\hat{\mathbf{s}}_\perp^c\approx\frac{\boldsymbol{\rho}}{\rho}=\hat{\boldsymbol{\rho}},\qquad|\boldsymbol{\rho}-\boldsymbol{\rho}_c|\approx\rho-\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}_c, \tag*{(84)}

implying that

e−jk⊥∣ρ−ρc∣+jk⊥s^⊥⋅(ρ′−ρc)∣ρ−ρc∣≈e−jk⊥ρ+jk⊥ρ^⋅ρ′ρ,(85)\frac{e^{-jk_{\perp}|\boldsymbol{\rho}-\boldsymbol{\rho}_{c}|+jk_{\perp}\hat{\mathbf{s}}_{\perp}\cdot(\boldsymbol{\rho}'-\boldsymbol{\rho}_{c})}}{\sqrt{|\boldsymbol{\rho}-\boldsymbol{\rho}_{c}|}} \approx \frac{e^{-jk_{\perp}\rho+jk_{\perp}\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}'}}{\sqrt{\rho}}, \tag*{(85)}

which makes (80)–(83) to appear as conical waves emanated from the zz axis of the coordinate system.

The behavior of fields at such distances is fully described by far-field coefficients FE(ϕ)F_E(\phi) and FH(ϕ)F_H(\phi), which are defined by the relations (e.g. [32 Sec. 3.4.4]):

Ez(ρ,z)≈2πk⊥ρFE(ϕ)e−jkzz−jk⊥ρ−j3π/4(86)E_z(\boldsymbol{\rho},z) \approx\sqrt{\frac{2}{\pi k_{\perp}\rho}}F_E(\phi)e^{-jk_z z-jk_{\perp}\rho-j3\pi/4} \tag*{(86)}
Hz(ρ,z)≈2πk⊥ρFH(ϕ)e−jkzz−jk⊥ρ−j3π/4.(87)H_z(\boldsymbol{\rho},z) \approx\sqrt{\frac{2}{\pi k_{\perp}\rho}}F_H(\phi)e^{-jk_z z-jk_{\perp}\rho-j3\pi/4}. \tag*{(87)}

Considering (11) and (12) in the limit ρ→+∞\rho\to+\infty and using (41) and the large-argument approximations of the Hankel functions in (3), (9) and (64) result in the formulas:

FE(φ)=−k⊥4∮Cejk⊥(ρ^⋅ρ′)[(ρ^⋅n^′)Ez(ρ′,0)+jk⊥∂∂n′Ez(ρ′,0)] dl′(88)F_E(\varphi)=-\frac{k_{\perp}}{4}\oint_C e^{jk_{\perp}(\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}')} \left[(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{n}}')E_z(\boldsymbol{\rho}',0)+\frac{j}{k_{\perp}}\frac{\partial}{\partial n'}E_z(\boldsymbol{\rho}',0)\right]\,\mathrm{d}l' \tag*{(88)}
FH(φ)=−k⊥4∮Cejk⊥(ρ^⋅ρ′)[(ρ^⋅n^′)Hz(ρ′,0)+jk⊥∂∂n′Hz(ρ′,0)] dl′.(89)F_H(\varphi)=-\frac{k_{\perp}}{4}\oint_C e^{jk_{\perp}(\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}')} \left[(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{n}}')H_z(\boldsymbol{\rho}',0)+\frac{j}{k_{\perp}}\frac{\partial}{\partial n'}H_z(\boldsymbol{\rho}',0)\right]\,\mathrm{d}l'. \tag*{(89)}

An alternative form follows from (35) and (36) as

FE(φ)=−k⊥4∮Cejk⊥(ρ^⋅ρ′)[(ρ^⋅n^′)Ez(ρ′,0)−kk⊥ZHt′(ρ′,0)+kzk⊥En′(ρ′,0)] dl′(90)F_E(\varphi)=-\frac{k_{\perp}}{4}\oint_C e^{jk_{\perp}(\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}')} \left[(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{n}}')E_z(\boldsymbol{\rho}',0)-\frac{k}{k_{\perp}}ZH_{t'}(\boldsymbol{\rho}',0)+\frac{k_z}{k_{\perp}}E_{n'}(\boldsymbol{\rho}',0)\right]\,\mathrm{d}l' \tag*{(90)}
FH(φ)=−k⊥4∮Cejk⊥(ρ^⋅ρ′)[(ρ^⋅n^′)Hz(ρ′,0)+kk⊥YEt′(ρ′,0)+kzk⊥Hn′(ρ′,0)] dl′,(91)F_H(\varphi)=-\frac{k_{\perp}}{4}\oint_C e^{jk_{\perp}(\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}')} \left[(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{n}}')H_z(\boldsymbol{\rho}',0)+\frac{k}{k_{\perp}}YE_{t'}(\boldsymbol{\rho}',0)+\frac{k_z}{k_{\perp}}H_{n'}(\boldsymbol{\rho}',0)\right]\,\mathrm{d}l', \tag*{(91)}

where Z=μ/ϵZ=\sqrt{\mu/\epsilon} and Y=ϵ/μY=\sqrt{\epsilon/\mu} are the intrinsic impedance and admittance of the medium outside CC.

Yet another form results from formulas (65) and (66) as

FE(φ)=−k⊥4∮Cejk⊥(ρ^⋅ρ′)[(ρ^⋅n^′)Ez(ρ′,0)−k⊥kZHt′(ρ′,0)−kzk(ρ^⋅t^′)ZHz(ρ′,0)] dl′(92)F_E(\varphi)=-\frac{k_{\perp}}{4}\oint_C e^{jk_{\perp}(\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}')} \left[(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{n}}')E_z(\boldsymbol{\rho}',0)-\frac{k_{\perp}}{k}ZH_{t'}(\boldsymbol{\rho}',0)-\frac{k_z}{k}(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{t}}')ZH_z(\boldsymbol{\rho}',0)\right]\,\mathrm{d}l' \tag*{(92)}
FH(φ)=−k⊥4∮Cejk⊥(ρ^⋅ρ′)[(ρ^⋅n^′)Hz(ρ′,0)+k⊥kYEt′(ρ′,0)+kzk(ρ^⋅t^′)YEz(ρ′,0)] dl′.(93)F_H(\varphi)=-\frac{k_{\perp}}{4}\oint_C e^{jk_{\perp}(\hat{\boldsymbol{\rho}}\cdot\boldsymbol{\rho}')} \left[(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{n}}')H_z(\boldsymbol{\rho}',0)+\frac{k_{\perp}}{k}YE_{t'}(\boldsymbol{\rho}',0)+\frac{k_z}{k}(\hat{\boldsymbol{\rho}}\cdot\hat{\mathbf{t}}')YE_z(\boldsymbol{\rho}',0)\right]\,\mathrm{d}l'. \tag*{(93)}

Equations (88)–(93) describe the far field of the sources enclosed by Γ\Gamma. The fields in the integrands are taken at z=0z=0 since their dependence on zz is explicitly included in (86) and (87). Representations (88)–(93) differ in the presence of the normal derivative, normal and tangential field components but despite their different appearance, describe identical far-field coefficients.

In the case of scattering problems, the scattering ability of cylindrical scatterers is described, depending on polarization, by the scattering widths wEw_E and wHw_H. The incident field is assumed to be the plane waves

Ezinc=E0zejk⊥ρcos⁡(φ−φ0)−jkzz(94)E_z^{\mathrm{inc}}=E_{0z}e^{jk_{\perp}\rho\cos(\varphi-\varphi_0)-jk_z z} \tag*{(94)}
Hzinc=H0zejk⊥ρcos⁡(φ−φ0)−jkzz(95)H_z^{\mathrm{inc}}=H_{0z}e^{jk_{\perp}\rho\cos(\varphi-\varphi_0)-jk_z z} \tag*{(95)}

for the TM and TE polarization, respectively. The incidence direction is described by the polar angle φ0\varphi_0 in the xx-yy plane and by the angle β\beta such that kz=kcos⁡βk_z=k\cos\beta, k⊥=ksin⁡βk_{\perp}=k\sin\beta and β=0\beta=0 corresponds to the incidence in the direction of positive zz axis. The far-field coefficients are related to the scattering widths, and depending on the polarization of the incident and scattered waves, four polarization cases are distinguished,

wEE=4kk⊥2∣FEE0z∣2,wHE=4kk⊥2∣FHYE0z∣2(96)w_{EE}=\frac{4k}{k_{\perp}^{2}}\left|\frac{F_E}{E_{0z}}\right|^{2},\qquad w_{HE}=\frac{4k}{k_{\perp}^{2}}\left|\frac{F_H}{YE_{0z}}\right|^{2} \tag*{(96)}

with H0z=0H_{0z}=0 and

wEH=4kk⊥2∣FEZH0z∣2,wHH=4kk⊥2∣FHH0z∣2.(97)w_{EH}=\frac{4k}{k_{\perp}^{2}}\left|\frac{F_E}{ZH_{0z}}\right|^{2},\qquad w_{HH}=\frac{4k}{k_{\perp}^{2}}\left|\frac{F_H}{H_{0z}}\right|^{2}. \tag*{(97)}

with E0z=0E_{0z}=0. The total scattering, extinction and absorption widths can also be expressed through the bistatic scattering widths FEF_E and FHF_H (e.g. [32 Sec. 3.4.4]).

Numerical checks and illustrations

The line-equivalence relations presented in Sections II, III, IV and V as well as the formulas for the far-field coefficients from Section VII are exact analytical results. In order to verify their validity, a number of numerical checks have been carried out by using Wolfram Mathematica. The following aspects have been considered:

  1. radiation problems (primary sources inside CC) and scattering problems (primary sources outside CC);

  2. integration curve CC smooth (circle) and with corner points (square);

  3. fields EE and HH and far-field coefficients FEF_E and FHF_H;

  4. observation point inside and outside CC.

In all test cases the formulas for the equivalence relations have been confirmed within the standard machine precision.

Radiation from a line source

In the first test case, the electromagnetic field

Ez(ρ,z)=E0ze−jkzz⋅H0(2)[k⊥(x−x0)2+(y−y0)2](98)E_z(\boldsymbol{\rho},z)=E_{0z}e^{-jk_z z} \cdot H_0^{(2)}\left[k_{\perp}\sqrt{(x-x_0)^2+(y-y_0)^2}\right] \tag*{(98)}

radiated by an electric line current at x=x0x=x_0 and y=y0y=y_0 with E0z=1E_{0z}=1 V/m and several values of kzk_z has been used. The other components of the fields can be obtained from (24)–(27). The rest parameters have been as follows: x0=0x_0 = 0 m, y0=0.5y_0 = 0.5 m, k=1k = 1 m−1^{-1}. Without loss of generality, the checks have been limited to the case of TM polarization (H0z=0H_{0z} = 0) since all presented relations are electromagnetically dual.

Equations (11) and (12), (42) and (47), and (59) and (60) have been tested. The enclosing line CC have been either a circle of radius ρ′=1\rho^{\prime} = 1 m centered at the origin of the coordinate system or a square contour with the corner points at ρ1=(−1\boldsymbol{\rho}_1 = (-1 m, −1-1 m)), ρ2=(1\boldsymbol{\rho}_2 = (1 m, −1-1 m)), ρ3=(1\boldsymbol{\rho}_3 = (1 m, 11 m)) and ρ4=(−1\boldsymbol{\rho}_4 = (-1 m, 11 m)). All formulations have recovered the fields EE and HH of the line source outside the integration contour while giving zeros inside the integration contour. Figure 3 shows the result of integration over the square contour CC for ∣ReE(x,y)∣|\mathop{\rm Re} E(x,y)| with kz=0k_z = 0 in the region −2-2 m ≤x,y≤2\le x,y \le2 m. The results confirm the applicability of the presented line-equivalence formulas to radiation problems and integration contours with and without corner points.

Amplitude of the real part of the field

Figure 3. Amplitude of the real part of the field (98) due a line source with kz=0k_z = 0 and x0=0x_0 = 0 m and y0=0.5y_0 = 0.5 m (the position is indicated by the red dot) computed by integrating Huygens’ sources over a square enclosing contour. The field is equal to zero inside the contour and coincides with the field radiated by the source in the outside of the contour.

Scattering from a dielectric cylinder

Equivalence relations can be applied to computing the fields scattered by cylindrical objects. In this case, the sources are secondary currents induced by an incident wave coming from outside of the enclosing surface Γ\Gamma. Consider, for example, the case of a dielectric circular cylinder illuminated by a plane wave. The exact analytical solution is available (e.g. [32 Sec. 6.5.1], [35, Sec. 4.1.3]) and can be used for checks of the equivalence formulas. The fields and the scattering amplitudes can be calculated either directly by using the exact series solution or indirectly by using the equivalence relations.

Figures 4 through 7 present the bistatic scattering widths wEEw_{EE}, wHEw_{HE}, wEHw_{EH} and wHHw_{HH} as functions of the scattering angle φ\varphi for a circular cylinder from quartz glass (ϵr′=3.81\epsilon^{\prime}_r = 3.81, ϵr′′=0.003\epsilon^{\prime\prime}_r = 0.003, μr=1\mu_r = 1) with the radius a=0.05a = 0.05 m at 80 GHz. The incidence direction is in the xx-zz plane (φ0=0∘\varphi_0 = 0^\circ) and makes the angle β=45∘\beta= 45^\circ with the cylinder axis. The formulas (88)–(93) have been used.

Log-polar plot of the bistatic scattering width $w_{HE}$

Figure 5. The same as in Fig. 4 but for wHEw_{HE} (wmax⁡=0.11w_{\max} = 0.11 m).

The integration contour CC has been chosen to be a circle of the radius 0.1 m with the center at the axis of the circular cylinder. The suitable orthogonal coordinate system is that of circular cylinder with u1=ρu_1 = \rho, u2=φu_2 = \varphi, h1=1h_1 = 1, h2=ρh_2 = \rho and dl=ρ dφdl = \rho\,d\varphi. As expected, the direct and indirect methods have given identical results, which is illustrated in Figs. 4 through 7. The agreement has also been observed for an observation point located at a finite distance from the cylinder by using the field relations either (35) and (36), or (65) and (66). Furthermore, the checks of the conventional form (11) and (12) have also been conducted by inserting analytically evaluated normal derivatives.

Conclusions

Field equivalence relations for infinite cylindrical enclosing surfaces and the zz-dependence exp⁡(−jkzz)\exp(-j k_z z) have been derived in the form of line integrals with the integration contour being the directrix of the cylindrical surface enclosing the source region. The sources can be radiating and/or scattering bodies, arbitrarily shaped and with arbitrary material constitution. Three versions of the field equivalence relation have been presented, analytically equivalent but differing in the presence of the specific field components (normal and tangential) or their normal derivative under the integration sign.

Polar plot comparing direct and indirect methods for $w_{EH}$

Figure 6. The same as in Figure 4 but for wEHw_{EH} (wmax=0.11 mw_{max}=0.11\,\mathrm{m}). The cross-pol plots coincide with each other, which is an intrinsic property of the solution for isotropic circular cylinders.

Log-polar plot of the bistatic scattering width $w_{EE}$

Figure 4. Log-polar plot of the bistatic scattering width wEEw_{EE} of a quartz glass circular cylinder with a=0.05a = 0.05 m for β=45∘\beta= 45^\circ and φ0=0∘\varphi_0 = 0^\circ at 80 GHz. The maximum value wmax⁡=5.34w_{\max} = 5.34 m is achieved on the outmost circle. The results are presented on the dB scale with the 5 dB step between the grid circles.

Polar plot comparing direct and indirect methods for $w_{HH}$

Figure 7. The same as in Figure 4 but for wHHw_{HH} (wmax=5.34 mw_{max}=5.34\,\mathrm{m}).

The conventional formulation ((11), (12), (88) and (89)) includes the normal derivative of the fields and can be inconvenient in numerical computations. The formulations of the Stratton-Chu type ((35), (36), (42), (47), (73), (74), (88) and (89)) and the Schelkunoff-Franz type ((59), (60), (65), (66), (75), (76), (92) and (93)) do not contain the normal derivative and can be advantageous when the fields are numerically obtained. Moreover, the relations of the Schelkunoff-Franz type include only the tangential components and can be regarded as the rigorous electromagnetic formulation of Huygens’s principle for infinite cylindrical enclosing surfaces.

Furthermore, approximate versions of the line-equivalence relations in the intermediate and far zones have shown how Huygens’ principle is to be understood for a cylindrical surface enclosing the sources: every element of the surface can be seen as a virtual linear source emanating a conical wave.

The relations have been specialized to the far-field limit (ρ→∞\rho\to\infty), and the corresponding expressions for the far-field coefficients have been given, see (88)–(93).

The derived expressions are applicable for a very general class of structures including those that are infinite and periodic along one direction, while having a finite cross section in the transverse plane. The results allow convenient calculations of fields at any point outside a cylindrical surface enclosing the structure by calculating a line integral of field values at the surface. For uniform structures, only one integration is sufficient. For periodic structures, calculations should be done for the fields of all significant Floquet harmonics. If only the far-field values are of interest and the period is smaller than half-wavelength, then also for periodic structures only the zero-order Floquet harmonic fields need to be integrated.

For inhomogeneous and non-periodic structures, the line-equivalence relations are to be applied to all harmonics from the Fourier spectrum with respect to the zz coordinate. If however only the far field is of interest, then the integration over kzk_z only from −k-k to kk, not from −∞-\infty to ∞\infty, is required as only the propagating part will contribute, thus eliminating the need to calculate fast-varying fields (large kzk_z). Note that in the conventional 3D surface integrals it is not easy to see how fine resolution for fields on the surface is enough. For periodic structures, far-field calculations need to be done only for a few (or even one) propagating Floquet harmonics of the field expansion.

Importantly, the theory is also applicable when the propagation constant kzk_z is complex-valued, allowing for studying leaky-wave regimes or active structures. In particular, the formulation entirely in terms of tangential components provides a suitable framework for describing reflection and transmission at frequency-selective surfaces (FSS) and metasurface structures (e.g., anomalous reflectors). The intermediate-zone expressions ((73)–(76)) are applicable, for example, in scenarios of millimeter-wave indoor communications, where the high-frequency approximation is valid, but the far-field conditions are usually not satisfied. The results can be also used in numerical electromagnetics, offering fast means for near- to far-field transformations for solutions of 2D problems based on periodic boundary conditions.

The presented equivalence relations assume a bounded cross-section of the cylindrical geometry. They do not apply to cylindrical bodies with an unbounded cross-section, e.g. wedges.

References

  1. [1]C. Huygens, Traité de la Lumière. Leiden, Netherlands: Pieter van der Aa, 1690.
  2. [2]I. V. Lindell, “Huygens’ principle in electromagnetics,” IEE Proceedings - Science, Measurement and Technology, vol. 143, no. 2, pp. 103–105, 1996.DOI
  3. [3]S. A. Schelkunoff, Electromagnetic Waves. Princeton, New Jersey: D. van Nostrand Company, 1943.
  4. [8]B. B. Baker and E. T. Copson, The Mathematical Theory of Huygens’ Principle, 2nd ed. Oxford: Clarendon Press, 1939.
  5. [9]J. A. Stratton, Electromagnetic Theory. McGraw-Hill, 1941.
  6. [10]M. Born and E. Wolf, Eds., Principles of Optics. Pergamon Press, 1959.
  7. [11]J. A. Stratton and L. J. Chu, “Diffraction theory of electromagnetic waves,” Physical Review, vol. 56, pp. 99–107, Jul 1939.DOI
  8. [12]S. A. Schelkunoff, “Some equivalence theorems of electromagnetics and their application to radiation problems,” Bell System Technical Journal, vol. 15, no. 1, pp. 92–112, 1936.DOI
  9. [13]——, “On diffraction and radiation of electromagnetic waves,” Physical Review, vol. 56, pp. 308–316, Aug 1939.DOI
  10. [14]W. Franz, “Zur Formulierung des Huygensschen Prinzips,” Z. Naturforschung, vol. 3a, pp. 500–506, 1948.DOI
  11. [15]——, Theorie der Beugung elektromagnetischer Wellen. Springer-Verlag, 1957.
  12. [21]S. Renganarajan and Y. Rahmat-Samii, “The field equivalence principle: illustration of the establishment of the non-intuitive null fields,” IEEE Antennas and Propagation Magazine, vol. 42, no. 4, pp. 122–128, 2000.
  13. [22]A. Booysen, “Aperture theory and the equivalence principle,” IEEE Antennas and Propagation Magazine, vol. 45, no. 3, pp. 29–40, 2003.DOI
  14. [23]C. Narendra, T. Brown, and P. Mojabi, “Gradient-based electromagnetic inversion for metasurface design using circuit models,” IEEE Transactions on Antennas and Propagation, vol. 70, no. 3, pp. 2046–2058, 2022.DOI
  15. [24]Y. Sugimoto, H. Arai, K. Sakakibara, and N. Kikumа, “Far-field estimation method of antennas above conductive plane using a partially spherical near-field,” IEEE Transactions on Antennas and Propagation, vol. 72, no. 4, pp. 3662–3669, 2024.
  16. [25]X.-J. Wu, W. Chen, H.-Z. Li, and T. Yuan, “Analysis of TE(z)-mode cluster in a stacked dielectric resonator antenna and its bandwidth diversity application,” IEEE Transactions on Antennas and Propagation, vol. 72, no. 4, pp. 3742–3747, 2024.DOI
  17. [26]C. J. M. Barker and A. K. Iyer, “Conformal metasurface radomes for beamshaping,” IEEE Transactions on Antennas and Propagation, vol. 74, no. 6, pp. 5345–5360, 2026.DOI
  18. [27]V. Popov, F. Boust, and S. N. Burokur, “Beamforming with metagratings at microwave frequencies: Design procedure and experimental demonstration,” IEEE Transactions on Antennas and Propagation, vol. 68, no. 3, pp. 1533–1541, 2020.
  19. [28]G. Xu, V. G. Ataloglou, S. V. Hum, and G. V. Eleftheriades, “Extreme beam-forming with impedance metasurfaces featuring embedded sources and auxiliary surface wave optimization,” IEEE Access, vol. 10, pp. 28 670–28 684, 2022.DOI
  20. [29]Y. Li, X. Ma, X. Wang, G. Pitcyn, M. Movahediqomi, and S. A. Tretyakov, “All-angle scanning perfect anomalous reflection by using passive aperiodic gratings,” IEEE Transactions on Antennas and Propagation, vol. 72, no. 1, pp. 877–889, 2024.DOI
  21. [30]S. Kumar Reddy Vuyyuru, R. Valkonen, S. A. Tretyakov, and D.-H. Kwon, “Efficient synthesis of large finite patch arrays for scanning wide-angle anomalous reflectors,” IEEE Open Journal of Antennas and Propagation, vol. 6, no. 1, pp. 75–87, 2025.
  22. [31]H. Weber, “Über die Integration der partiellen Differentialgleichung: ∂²u/(∂x)² + ∂²u/(∂y)² + k²u = 0,” Mathematische Annalen, vol. 1, pp. 1–36, 1869.
  23. [32]A. V. Osipov and S. A. Tretyakov, Modern Electromagnetic Scattering Theory with Applications. John Wiley & Sons, 2017.DOI
  24. [33]P. M. Morse and H. Feshbach, Methods of Theoretical Physics. McGraw-Hill, 1953.
  25. [34]M. Abramowitz and I. A. Stegun, Eds., Handbook of Mathematical Functions, 5th ed. Dover Publications, 1968.

Paper details

vlnc:W_z73nw5zn
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents