This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES 1 Determination of Effective Constitutive Parameters of Inhomogeneous Metamaterials With Bianisotropy Ugur Cem Hasar , Member, IEEE, Gul Buldu , Yunus Kaya, and Gokhan Ozturk Abstract— A recursive algorithm is proposed to accurately extract the electromagnetic parameters of isotropic/bianisotropic metamaterial (MM) slabs of inhomogeneous structures. The algorithm is based on the recursive calculations of measured/ simulated forward and backward scattering (S-) parameters from previously extracted wave impedances and refractive indices of each homogeneous layer (except the analyzed intermediate layer) by a suitable retrieval method. There are two main advantages of our proposed method over existing extraction algorithms. First, it could retrieve the electromagnetic parameters of both isotropic and bianisotropic MM slabs. Second, it is applicable for electromagnetic parameter extraction of inhomogeneous structures. For validation of our method and assessing its accuracy over existing extraction methods, we used the simulation results of various inhomogeneous structures constructed by different isotropic/bianisotropic MM slabs (split-ring-resonator (SRR), SRR-wire, omega, and omega-wire). For completeness, the effects of noise present in S-parameters and iterative error (by increasing the value of iteration numbers) are examined to test the robustness of the proposed algorithm. Index Terms— Bianisotropic, constitutive parameters, inhomogeneous, metamaterials (MMs), recursive algorithm, retrieval. I. I NTRODUCTION T HE electromagnetic response of materials is generally identified by the parameters of relative permittivity (εr ) and relative permeability (μr ) [or similar terms such as magnetoelectric coupling coefficient (ξ0 ) and chirality parameter (κ)] or wave impedance (Z ) and refractive index (n). In order to interpret the electromagnetic response of conventional materials (right-handed materials) or engineered materials [lefthanded materials generally coined as metamaterials (MMs)] and apply these materials for some specific applications (e.g., electromagnetic cloak [1] and perfect lens [2]), their electromagnetic parameters should be accurately retrieved. Manuscript received September 25, 2017; revised January 19, 2018; accepted March 6, 2018. This work was supported by the Scientific and Technological Research Council of Turkey under Project 112R032. (Corresponding author: Ugur Cem Hasar.) U. C. Hasar is with the Department of Electrical and Electronics Engineering, Gaziantep University, 27310 Gaziantep, Turkey (e-mail: uchasar@gantep.edu.tr). G. Buldu is with the Department of Electrical and Electronics Engineering, Munzur University, 62000 Tunceli, Turkey. Y. Kaya is with the Department of Electricity and Energy, Bayburt University, 69000 Bayburt, Turkey. G. Ozturk is with the Department of Electrical and Electronics Engineering, Ataturk University, 25240 Erzurum, Turkey. Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TMTT.2018.2846726 In this paper, our focus is to accurately extract electromagnetic properties of MM structures. Generally, homogenization approach is applied to MMs [3], when the operating wavelength is much larger (long wavelength limit) than the lattice periods of MMs (quasi-static case), to make the complex electromagnetic response of MMs to be equivalent to the response of a hypothetical continuous material. Various equivalence models were defined such as external equivalence, dispersion equivalence, and singlemode or double-mode equivalences, considering the completeness and sophistication of the constructed model [4]. Among these models, our primary concern is to utilize the external equivalence approach through which effective electromagnetic properties of MM slabs can be retrieved from scattering (S-) parameters. There are already many techniques available in the literature for extraction of electromagnetic properties of isotropic, bianisotropic, and chiral MM slabs based on measured/simulated free space (normal or oblique incidence) or waveguide S-parameters with/without boundary effects [5]–[38]. All these studies were restricted to retrieval of electromagnetic properties of homogeneous MM slabs. The methods in [39] and [40] can be applied for electromagnetic parameter retrieval of inhomogeneous MM slabs. However, these methods do not consider the effect of magnetoelectric coupling arising from the coupling between electric and magnetic fields and thus not applicable for retrieval of electromagnetic properties of bianisotropic MM slabs. The unique advantages of this type of MM slabs are that their n could have a positive real part for a backward wave propagation [41], they have a wider stopband [42], and they have different forward and backward wave impedances resulting in different forward and backward S-parameters [17], different forward and backward powers [43], and different semi-infinite reflection coefficients [30]. In this paper, we extend the methodologies in [39] and [40] and propose a retrieval procedure for the determination of electromagnetic properties of bianisotropic inhomogeneous MM slabs using measured/simulated S-parameters. To this end, we will consider the whole inhomogeneous bianisotropic MM slab as a cascade connection of piecewise bianisotropic homogeneous MM slabs [44]. The organization of the rest of this paper is as follows. First, we give a theoretical analysis to derive intermediate recursive S-parameters in terms of forward and backward wave impedances and refractive indices of each homogeneous layer (e.g., bianisotropic MM slab) of the whole inhomogeneous structure in Section II. Then, in Section III, we present our 0018-9480 © 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. This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 2 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES where dt = 0 for s = 0 and for s ≥ 1 dt = s = dκ . (3) κ=1 Besides, for the m + 1 layer, electric and magnetic fields are i Ē m+1 = āx E m+1 e+ j k0 nm+1 (z−(d1+d2 +···+dm )) H̄m+1 = ā y Fig. 1. Demonstration of forward and backward problems under analysis (an inhomogeneous bianisotropic MM structure). (a) Recursive S11 and S21 parameters of the forward problem. (b) Recursive S12 and S22 parameters of the backward problem. extraction algorithm for the determination of electromagnetic parameters of any homogeneous layer using two-sided recursive S-parameter expressions. We next performed two different analyses (numerical and simulation) for the validation of our algorithm by considering three different inhomogeneous structures in Section IV. In this section, we also investigated the effect of random noise superposed on simulated S-parameters on the performance of our extraction algorithm. Finally, in Section V, we recapitulate some important points about our extraction process and its application. II. T HEORY Fig. 1(a) and (b) shows the geometrical configurations for the derivation of recursive forward (S11 and S21 ) and backward (S12 and S22 ) S-parameters of a bianisotropic inhomogeneous MM structure, respectively. The forward and backward wave impedances, refractive index, and thickness of the mth layer + , Z − , n , and d , respectively. To make are denoted by Z m m m m the analysis as general as possible, the inhomogeneous MM structure is assumed to be surrounded by two different bianisotropic MM layers (Layers 0 and m + 1). Without loss of generality, we assume that the inhomogeneous structure in Fig. 1 is reciprocal (transmission symmetric) and that a uniform plane wave propagating in +z direction with polarization in +x direction impinges onto the MM structure in Fig. 1(a), which can be separated into m-layer piecewise homogeneous bianisotropic MM slabs. Then, the expressions of electric and magnetic fields in the sth layer (s = 0, 1, . . . , m) can be written for the exp(− j ωt) (1) Ē s = āx E si e+ j k0 ns (z−dt ) + E sr e− j k0 ns (z−dt ) i r E E s + j k0 ns (z−dt ) H̄s = ā y − −s e− j k0 ns (z−dt ) (2) +e Zs Zs i E m+1 + Z m+1 e+ j k0 nm+1 (z−(d1+d2 +···+dm )) . (4) (5) In (1)–(5), the superscripts “i ” and “r ” denote the amplitudes of incident and reflected waves; k0 is the free-space wavenumber; and n s , Z s+ , and Z s− are the refractive index and forward and backward wave impedances of the sth layer. Applying boundary conditions (continuity of tangential components of electric and magnetic fields) at interfaces (z = 0, d1 , d1 +d2 , . . . , d1 +d2 +· · ·+dm−1 ) and following the procedure in [40], we obtain the following recursive relations for forward S-parameters (l = 1, 2, . . . , m − 1): (l ) + + − − + + Z l− S111 Z l1 Z l + Z l1 − Z l1 Z l − Z l1 1 (l) (6) S11 = + (l ) + − − − + − 1 Zl Tl2 S11 Z l1 Z l1 − Z l + Z l1 Z l1 + Z l (l1 ) Z l+1 Z l−1 Z l+ + Z l− S21 1 1 (l) (7) S21 = + − + (l ) + − − − 1 Tl Zl S Z Z −Z +Z Z +Z 11 l1 l1 l l1 l1 l where l1 = l − 1, l2 = l + 1, and (l1 ) S11 = Elr1 Eli1 (l1 ) S21 = i E m+1 Eli1 (0) (0) = S11 , S21 = S21 Tl = e+ j k0 nl dl S11 (8) (9) (l1 ) (l1 ) where S11 and S21 , defined at z = d1 +d2 +· · ·+dl−1 , denote the total forward reflection and transmission S-parameters for the structure involving the layers l, l + 1, . . . , m, respectively. In addition, application of the boundary condition (continuity of tangential components of electric and magnetic fields) at z = d1 + d2 + · · · + dm , we find − Z+ + r Zm − Zm Em (m) S11 = i = + m+1 (10) + − Em Z m Z m+1 + Z m + + i − Zm + Zm E m+1 Z m+1 (m) S21 = (11) = + + − . i Em Zm + Zm Z m+1 Now, we assume that a uniform plane wave propagating in the −z direction with polarization in +x direction impinges onto the MM structure in Fig. 1(b). Then, applying the same boundary conditions for the derivation recursive relations for S11 and S21 at interfaces (z = d1 + d2 + · · · + dm , d1 + d2 + · · · + dm−1 , . . . , d1 ) and following the procedure in [40], the recursive relations for the backward reflection and transmission S-parameters are derived as: (l ) − Z+ + Z− + Z+ Z− − Z− S221 Z m m2 m1 m2 m2 m1 (l) 2 S22 = ψ1 (12) (l1 ) − + + + − + S22 Z m 2 Z m 2 − Z m 1 + Z m 2 Z m 2 + Z m 1 + − S (l1 ) Z m1 + Z m (l) 12 1 S12 = ψ2 (13) (l1 ) − + + + − + S22 Z m 2 Z m 2 − Z m 1 + Z m 2 Z m 2 + Z m 1 This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. HASAR et al.: DETERMINATION OF EFFECTIVE CONSTITUTIVE PARAMETERS OF INHOMOGENEOUS MMs Fig. 2. Extraction of wave impedances (Z u+ and Z u− ) and refractive index (n u ) of the uth layer (1 ≤ u ≤ m) in the m layer inhomogeneous structure in Fig. 1. where m 1 = m − l + 1, m 2 = m − l + 2, and ψ1 = + Zm 1 − 2 Zm 1 Tm 1 ψ2 = + Z− Zm 2 m2 − Zm 1 Tm 1 . (14) Application of the boundary conditions at z = 0 yields Z 1+ Z 0− − Z 1− E 1r (m) (15) S22 = i = − + Z 1 Z 1 + Z 0− E1 Z 0− Z 1+ + Z 1− E 0i (m) S12 = i = − + . (16) Z 1 Z 1 + Z 0− E1 We note that setting Z l+1 = Z l−1 = ηl1 , Z l+ = Z l− = ηl , (l) + = Z − = η , and Z + = Z − = η Zm m1 m 2 reduces S11 , m1 m2 m2 1 (l) (l) (l) S21 , S22 , and S12 in (6), (7), (12), and (13) to the recursive expressions of forward and backward reflection and transmission S-parameters in (10) and (13) [40] for an inhomogeneous structure involving isotropic MM slabs. III. E XTRACTION P ROCESS Using the recursive expressions in (6)–(16), it is possible to calculate intermediate S-parameters of a bianisotropic MM (u 1 ) (u 1 ) , S21 , layer in Fig. 1. For example, as shown in Fig. 2, S11 (m−u) (m−u) S12 , and S22 intermediate S-parameters can be used to retrieve the wave impedances (Z u+ and Z u− ) and the refractive index (n u ) of the uth layer (1 ≤ u ≤ m) in the inhomogeneous structure with an m layer, as shown in Fig. 1. However, such an extraction for the geometry in Fig. 2 requires the knowledge of wave impedances and refractive indices of the first (from left to right) u − 1 layers and the last u + 1 to m layers in the inhomogeneous structure in Fig. 1, since S-parameters (S11 , S21 , S12 , and S22 ) of the whole inhomogeneous MM structure in Fig. 1 are not sufficient to determine the electromagnetic properties of each layer of this inhomogeneous MM structure with m > 1. Therefore, we assume that the wave impedances and refractive indices of the first u −1 layers and the last u +1 to m layers are known or extracted by a retrieval procedure, to be discussed later. In the extraction of Z u+ , Z u− , and n u of the uth layer in Fig. 2, we propose the following steps. We assume without + − = Z m+1 = Z air and loss of generality that Z 0+ = Z 0− = Z m+1 n 0 = n m+1 = n air , where Z air and n air are the impedance and refractive index of air, respectively. 3 1) The effective electromagnetic properties (wave impedances and refractive index) of the first u − 1 layers and the last u + 1 to m layers are extracted by a suitable retrieval procedure such as [17] or [24] using simulated or measured S-parameters of each one of these layers (immersed in air region). These parameters will be used in the following steps in the determination of intermediate S-parameters by recursive relations. (0) (0) = S11 and S21 = S21 and continue if u = 1. 2) Use S11 Replace the integer index l with the dummy integer index p (1 ≤ p ≤ m) in the recursive relations in (6) ( p) ( p) and (7) and then iterate S11 and S21 for p < u. For example, for m = 6 and u = 4, one should obtain (1) (2) (3) (0) (1) (2) S11 , S11 , and S11 in terms of S11 , S11 , and S11 , ( p) respectively. A similar result will be obtained for S21 . (m) 3) If u = m, implement the following. Calculate S11 and (m) (l−1) (l−1) S21 from (10) and (11) and express S11 and S21 in (l) the recursive relations in (6) and (7) in terms of S11 and (l) S21 . Then, replace the integer index l with the dummy ( p) ( p) index p (1 ≤ p ≤ m) and iterate S11 and S21 for u < p < m. For example, for m = 6 and u = 4, one (5) (6) should obtain S11 in terms of S11 . Similar parameters ( p) will be determined for S21 . (0) 4) Use S22 = S22 and if u = m, continue with the following steps. Replace the integer index l with the dummy integer index p (1 ≤ p ≤ m) in the recursive ( p) relation in (12). Then, iterate S22 for 1 ≤ p ≤ m 3 where m 3 = m − u. For example, for m = 6 and u = 4, one (1) (2) (0) (1) should obtain S22 and S22 in terms of S22 and S22 , respectively. (m) 5) If u = 1, implement the following. Calculate S22 (l−1) from (15) and express S22 in the recursive relation (l) in (12) in terms of S22 . Then, replace the integer index ( p) l with the dummy index p (1 ≤ p ≤ m) and iterate S22 for m 4 < p ≤ m where m 4 = m − u + 2. For example, (5) for m = 6 and u = 4, one should obtain S22 in terms (6) of S22 . (m ) 6) By setting Tu21 S22 4 to zero for u = 1, and by setting (u ) (u ) Tu22 S112 to zero and Tu22 S212 to one for u = m, + calculate Z u from 10 [(2 6 −3 5 )10 +(2 4 −1 5 )9 ]Z u+ + [2(2 6 −3 5 )9 10 +(2 4 −1 5 ) 3 × 8 10 +229 +(3 4 −1 6 )×210 ]Z u+ 4 + 39 [(2 4 −1 5 )8 +(3 4 −1 6 )10 ] × Z u+ +[(2 4 −1 5 )28 +2(3 5 −2 6 ) × 8 9 +(3 4 −1 6 ) 8 10 +229 ]Z u+ 2 + 8 [(3 5 −2 6 )8 +(3 4 −1 6 )9 ] = 0 (17) where (u 1 ) + (m 3 ) − Z u 2 Z u+1 Z u−1 Z u 2 − S22 1 = 1 + S11 (m 3 ) (u 1 ) + + 1 + S22 Z u 1 − Z u−1 Z u+2 Z u−2 S11 (18) This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 4 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES (u 1 ) + + (m 3 ) − 2 = Z u−1 − S11 Z u 1 Z u 2 − S22 Z u2 (u ) (m ) 3 = − 1 + S111 Z u+1 Z u−1 1 + S22 3 Z u+2 Z u−2 (u 2 ) 4 = 1 + Tu22 S11 1− (m 4 ) − 1 + Tu21 S22 5 = 1− Z u+2 Z u−1 Z u+1 1− (u ) (20) (m 4 ) Z u+2 Tu21 S22 Z u+2 Z u−2 1− (u 2 ) Z u−1 (21) Tu22 S11 Z u−1 (m ) Tu21 S22 4 (22) (u 2 ) (m 4 ) 6 = − 1 + Tu22 S11 Z u+2 1 + Tu21 S22 Z u−1 (23) 7 = Z u−2 Tu22 S112 (19) (u 2 ) Z u+2 Tu 2 S21 (u ) Z u+1 Z u+1 Z u−1 S211 (u 1 ) 2 +2 −2 2 − 1 + S11 Z u 1 Z u 1 7 (u 1 ) − (u 1 ) + 9 = 1 + S11 Z u 1 − S11 Z u 1 27 Z u+1 Z u−1 (u 2 ) − 1 + Tu22 S11 10 = 1− Z u+2 Fig. 3. Geometry of the analyzed inhomogeneous structures by CST simulations. (a) Two-layer structure consisting of FR4 and SRR-wire MM slab. (b) Three-layer structure consisting of SRR MM slab (Bian.), FR4, and SRR-wire MM slab (Bian.). (c) Four-layer structure consisting of FR4, omega MM slab (Bian.), marble, and omega-wire MM slab (Bian.). (u 2 ) 2 +2 8 = 1 + Tu22 S11 Z u2 (u ) Tu2 S 2 Z u−2 2 11 1− 2 Z u+2 2 (u 2 ) Tu S Z u−2 2 11 (24) Z u+2 (25) 2 (u ) − Z u−1 − S111 Z u+1 27 . (26) Such a calculation can be performed by a suitable numeric function such as Newton’s method [45] or “roots” function of MATLAB. Once Z u+ is computed, Z u− and Tu2 can be found from (A.1) and (A.11). Details of the derivations of Z u+ , Z u− , and Tu can be found in the Appendix. Finally, n u can be calculated from −i ln(Tu2 ) + 2πm b , m b = 0, ∓1, ∓2, . . . (27) 2k0 du where m b denotes the branch index number. For electrically small layer u, m b is essentially equal to zero. However, for electrically large layer u, m b attains various values. Correct value(s) of m b for the layer u can be found by applying different techniques, such as the phase unwrapping method [23], [32], the stepwise method [19], [24], or the improved NRW algorithm [34]. 7) Reiterate the steps 2 through 6 for any other medium ( p = u) in the inhomogeneous MM slab using newly found electromagnetic parameters (Z u+ , Su− , and n u ) of the uth layer. 8) Repeat the steps 1 through 7 for the layers whose electromagnetic parameters are determined till the calculated electromagnetic parameters converge to the limit of prespecified accuracy or till the end of a maximum iteration number (Nmax ) set priori. We want to highlight three important points regarding to our extraction algorithm. First, the algorithm in [40] is suitable for the extraction of electromagnetic properties of an inhomogeneous MM slab composed of isotropic layers, whereas our extraction algorithm is applicable to electromagnetic parameter nu = retrieval of an inhomogeneous MM slab composed of either isotropic or bianisotropic layers (or their combination). Second, while the algorithm in [40] works only in one direction for electromagnetic parameter extraction, our algorithm is two-sided searching for S11 from left to right and S22 from right to left for the same goal. In other words, whereas the (u−1) (u−1) (m−u) and S21 (or S12 and algorithm in [40] requires S11 (m−u) S22 depending on iteration direction) for extraction of the electromagnetic properties of the layer u in Fig. 2, our method (u−1) (m−u) (u−1) (m−u) , S22 , and S21 (or S12 for isotropic requires S11 materials) for the same purpose. Third, when m = u = 1 and Z 0+ = Z 0− = Z 2+ = Z 2− = Z 0 are considered, our algorithm reduces to the extraction method [17], [24] valid for only a one-layer homogeneous bianisotropic MM slab. IV. R ESULTS AND D ISCUSSION In this section, we first validated and tested our proposed method for three different inhomogeneous MM structures composed of different MM slabs (split-ring-resonator (SRR), SRR-wire, omega, and omega-wire). Then, we continued with the effects of noise and iteration on extracted electromagnetic parameters by our method. A. Simulation Results We applied the CST Microwave Studio with its time-domain solver to obtain simulated S-parameters of the inhomogeneous structures in Fig. 3(a)–(c). Simulation details are as follows. Two waveguide ports were located on x y planes at z = 0 and z = d1 + d2 + · · · + dm for the extraction of electromagnetic properties of inhomogeneous MM structures, two of which can show bianisotropic property. Perfect electric (magnetic) conductors were assumed to be located on yz (x z) planes to simulate uniform plane wave propagation in the +z direction with polarization in the +x direction. Adaptive meshing feature of the CST program was set active so that the CST can arrange optimum meshing. Frequency range was set to 2–18 GHz. This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. HASAR et al.: DETERMINATION OF EFFECTIVE CONSTITUTIVE PARAMETERS OF INHOMOGENEOUS MMs Fig. 4. Extraction results for the two-layer inhomogeneous structure [FR4 material and SRR-wire MM slab (Iso.)]. (a) Normalized (relative to air) forward wave impedance of the SRR-wire MM slab. (b) Refractive index of the SRR-wire MM slab. Magnitudes of (c) S11 and S22 and (d) S21 and phases of (e) S11 and S22 and (f) S21 of the two-layer inhomogeneous structure reconstructed from effective parameters (“Effective”) extracted by our method and obtained from simulations (“Two-layer”). In the figures, “SM” and “PM” refer to the stepwise method [24] and the proposed method. We first considered the two-layer inhomogeneous structure (m = 2) in Fig. 3(a) composed of an FR4 material (d1 = 2.0 mm) and a composite SRR-wire MM slab. The dimensions of this slab were set the same as those of the study [40] in order to compare the results. The cell is cubic with side d2 = 2.5 mm. The SRR is constructed over the front face of the FR4 substrate with thickness 0.25 mm and relative electrical permittivity εr1 = 4.4(1+ j 0.02) ( μr1 = 1). For the SRR slab, the outer ring is 2.2 mm, the separation distance between rings is 0.15 mm, the linewidth of each ring is 0.2 mm, and the gap of both rings is 0.15 mm. Besides, the wire with 0.2-mm linewidth extends in the x-direction at the back of the FR4 substrate. Copper material with a thickness of 35 μm is utilized to create SRR and wire strips. It is known that this SRR-wire MM slab shows isotropic (Iso.) property [11], [46] for the assumed plane wave propagation in the z-direction with polarization vector in the x-direction, i.e., S11 = S22 . Applying the steps given in Section III, effective electromagnetic parameters of the SRR-wire MM slab (Iso.) of this inhomogeneous structure were extracted by our algorithm for Nmax = 3 and with 127 968 hexahedral cells used to discretize the computational domain. For comparison, these parameters were also retrieved by different extraction methods [11], [12], [17], [24], [25], [39] using only the simulated S-parameters of the SRR-wire MM slab. Because extracted electromagnetic properties by these methods are similar, only the results of the method in [24] are presented in Fig. 4(a) and (b) for conciseness. Retrieved z 2− is not demonstrated, since it is almost equal to z 2+ . 5 Fig. 5. Extracted wave impedances of the two-layer inhomogeneous structure [FR4 material and SRR-wire MM slab (Iso.)]. (a) Normalized wave impedance extracted by the methods in [11] and [12]. (b) Normalized wave impedance extracted by the method in [39]. Normalized (c) forward and (d) backward wave impedances extracted by the methods in [17] and [24]. Normalized (e) forward and (f) backward wave impedances extracted by the method in [25]. In addition, as shown in Fig. 4(c)–(f), we also compared the simulated S-parameters of the two-layer inhomogeneous structure with its S-parameters obtained from the effective parameters by our algorithm using the transfer matrix method (TMM) from [43, expressions (5) and (6)]. The following points are noted from dependences in Fig. 4(a)–(f). First, normalized forward (and backward) wave impedance z 2+ (and z 2− ) and the refractive index n 2 of this SRR-wire MM slab extracted by our algorithm and the extraction methods [11], [12], [17], [24], [25], [39] are in good agreement with one another except for some small differences. We think that these differences arise from two main factors. As a first factor, as noted in [40], there might be a coupling between layers of the inhomogeneous structure, which in turn can alter the electromagnetic properties of the SRR-wire (Iso.). As a second factor, there might be some spatial effects as a consequence of near-field boundary effects near the border of the SRR-wire MM slab [30], [36], which can also modify its effective electromagnetic parameters. Second, as seen in Fig. 4(c)–(f), both the magnitudes and phases of simulated and reconstructed (by our extraction technique) S-parameters of the SRR-wire MM slab are in good agreement with each other except for some small discrepances over whole band, validating our extraction method and showing its accurateness for the extraction of electromagnetic properties of an isotropic MM slab of an inhomogeneous MM structure. We note that the methods in [11], [12], [17], [24], [25], and [39] are applicable to retrieval of electromagnetic properties of one homogeneous MM slab. It is instructive to analyze the applicability of these methods for inhomogeneous MM slabs. For example, Fig. 5(a)–(f) shows This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 6 the extracted wave impedances of the two-layer inhomogeneous structure (FR4 material and SRR-wire MM slab) with thickness dt = d1 + d2 = 4.5 mm by different extraction methods [11], [12], [17], [24], [25], [39]. Extracted refractive index is not shown for simplicity. The following points are noted from the dependences in Fig. 5. First, none of the extraction methods [11], [12], [17], [24], [25], [39] could retrieve the correct wave impedances [see Fig. 4(a) and (b)] of the SRR-wire slab. Second, although the SRR-wire slab demonstrates isotropic behavior [see Fig. 3(a)], extracted forward and backward wave impedances by the methods in [17] and [24] and the method in [25] significantly differ from each other due to the presence of the FR4 material. Third, real parts of the forward and backward wave impedances of the two-layer inhomogeneous structure retrieved by the method in [25] could result in nonphysical artifacts (e{z} ≤ 0) around 8 and 10 GHz. Besides, we also employed the extraction algorithm [40] to extract electromagnetic parameters of the SRR-wire MM slab of the two-layer inhomogeneous structure in Fig. 3(a) because it does not possess bianisotropy property for assumed plane wave configuration. Electromagnetic parameters z 2+ , z 2− , and n 2 extracted by the extraction algorithm [40] are so close to those retrieved by our extraction algorithm in Fig. 4(a)–(c) (so they are not presented here), demonstrating concurrency between our algorithm and the algorithm in [40]. It is noted that when the bianisotropy breaks down (z + = z − or S11 = S22 ), our method and the method in [40] become very similar. As a second example, we investigated the three-layer (m = 3) inhomogeneous structure in Fig. 3(b) composed of an SRR MM slab (d1 = 2.5 mm), FR4 material (d2 = 2.0 mm), and the SRR-wire MM slab (d3 = 2.5 mm). Here, the SRR-wire MM slab has the same electrical and geometrical properties of the SRR-wire MM slab in the previous example (isotropic MM slab) except that the rings are rotated in the +y direction by 90°. By this new orientation of the ring gaps, as seen in Fig. 3(b), the magnetic field in the y-direction (normal to the plane of the SRR) will produce currents and thus electric dipoles in the x-direction due to gap. Additionally, the electric field in the x-direction (normal to the slit axis) will induce deposition of charges and thus magnetic dipole in the y-direction. As such, this new topology of the SRR-wire MM slab possesses bianisotropic (Bian.) property (z + = z − and S11 = S22 ) for the wave propagation in the z-direction with polarization vector in the x-direction [17], [37], [38], [46]. The SRR MM slab is identical to the SRR-wire MM slab without the wire at the back of the FR4 substrate. Effective electromagnetic parameters of the SRR-wire (Bian.) of the three-layer inhomogeneous structure were extracted by our algorithm for Nmax = 3 and with 153 216 hexahedral cells used to discretize the computational domain. For comparison, we also applied the extraction methods in [17], [24], and [25] to inverse the electromagnetic properties of the SRR-wire MM slab (Bian.) using its simulated S-parameters only. We note that the extraction methods [11], [12], [17], [24], [25], [39] were not tested because they are not applicable for bianisotropic MM slabs. IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES Fig. 6. Extraction results for the three-layer inhomogeneous structure [SRR (Bian.) MM slab, FR4 material, and SRR-wire MM slab (Bian.)]. (a) Normalized (relative to air) forward wave impedance of the SRR-wire MM slab. (b) Normalized (relative to air) backward wave impedance of the SRR-wire MM slab. (c) Refractive index of the SRR-wire MM slab. (d) Closer view of the extracted m{z 3− }. Magnitudes of (e) S11 and S22 and (f) S21 . Phases of (g) S11 and S22 and (h) S21 of the three-layer inhomogeneous structure reconstructed from effective parameters (“Effective”) extracted by our method and obtained from simulations (“Two-layer”). In the figures, “LAO” and “PM” refer to the retrieval method [17] and the proposed method. Since the methods in [17], [24], and [25] produce similar results, only the results of the method in [17] are presented in Fig. 6 for brevity. For example, Fig. 6(a)–(d) shows the frequency dependences of z 3+ , z 3− , and n 3 of the SRR-wire MM slab extracted from our algorithm and the method in [17]. The following two points are noted from the dependences in Fig. 6(a)–(d). First, z 3+ and z 3− are different from each other over the whole frequency band, demonstrating bianisotropic feature of the analyzed SRR-wire MM slab. Second, the normalized wave impedances z 3+ and z 3− and the refractive index n 3 of the SRR-wire (Bian.) MM slab obtained by the sole extraction procedure [17] and by our algorithm are generally in good agreement over the whole band except for the two frequency regions where rapid variations are notable ( f ∼ = 8.67 GHz where |S11 | attains its minimum and f ∼ 14.9 GHz where n 3 approaches zero). For exam= ple, Fig. 6(d) shows the dependence of the imaginary part of the extracted z 3− (m{z 3− }) over 14–16 GHz, indicating some differences between extracted parameters. Specifically, m{z 3− } values at f = 14.8 GHz extracted by the sole This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. HASAR et al.: DETERMINATION OF EFFECTIVE CONSTITUTIVE PARAMETERS OF INHOMOGENEOUS MMs extraction method [17] and the proposed algorithm are −5.70 and −5.14, respectively—showing an approximately 10% variation. We think that the coupling between MM slabs as well as the spatial effects around the borders of MM slabs are the cause of discrepancy between extracted parameters by the method in [17] and our algorithm. We also analyzed and compared the simulated S-parameters of the three-layer inhomogeneous structure with its S-parameters reconstructed from the effective parameters by our algorithm via the TMM method [43]. The results are shown in Fig. 6(e)–(h). We note the following two points from these figures. First, reconstructed individual S-parameters (S11 , S21 , and S22 ) of this inhomogeneous structure by our algorithm partly differ from its simulated S-parameters specially around the resonance regions f ∼ = 8.67 and f ∼ = 11.07 GHz. Second, the difference between reconstructed and simulated individual S-parameters of this inhomogeneous structure [see Fig. 3(b)] is overall greater than the difference between reconstructed and simulated individual S-parameters of the two-layer inhomogeneous structure [see Fig. 3(a)]. We think that there are two reasons for such an increase. First, coupling between electric and magnetic fields for an individual SRR (or SRR-wire) increases due to the rotation of the gap axis of the SRR rings. Second, there might be an additional coupling between the resonating MM slabs of the inhomogeneous structure in Fig. 3(b). Finally, we note from the dependences in Fig. 6(e)–(h) that our proposed method is applicable for the extraction of the electromagnetic properties of bianisotropic MM slabs of an inhomogeneous MM structure. In order to test the applicability of our proposed method for different MM slabs, as a final example, we examined the four-layer inhomogeneous structure in Fig. 3(c) composed of an FR4 material (d1 = 5.0 mm), an omega MM slab (Bian.), a PVC material (d3 = 5.0 mm, εr3 = 2.71(1 + 0.08), and μr3 = 1), and an omega-wire MM slab (Bian.). The geometrical parameters and the electromagnetic properties of both omega and omega-wire MM slabs are similar to those in [24]. The cell of the omega MM slab has a mean radius of 1.19 mm, a width of 0.45 mm, and the tail length of 1.8 mm, and placed over the front face of the FR4 substrate with a thickness of 1.6 mm. The wire with 1.44-mm linewidth extends in the x-direction at the back of the FR4 substrate. Copper material with a thickness of 35 μm is utilized to create omega patterns and wire strips. Both the omega and omega-wire slabs are in cubical form with a side of d2 = d4 = 5.0 mm, exhibiting bianisotropic behavior since electric field is parallel with the tails and magnetic field is normal to the plane of the omega-shape [24], [47]. Effective electromagnetic parameters of the omega MM slab (Bian.) of the four-layer inhomogeneous structure were retrieved by our algorithm for Nmax = 3 and with 632 016 hexahedral cells used to discretize the computational domain. For comparison, we applied the extraction methods in [17], [24], and [25] to inverse the electromagnetic properties of the omega MM slab (Bian.) using its simulated S-parameters only. Only the results of the method in [17] are presented in Fig. 7 for brevity, because the methods 7 Fig. 7. Extraction results for the four-layer inhomogeneous structure [FR4, omega MM slab (Bian.), PVC sample, and omega-wire MM slab (Bian.)]. (a) Normalized (relative to air) forward wave impedance of the omega MM slab. (b) Refractive index of the omega MM slab. Magnitudes of (c) S11 and S21 and phases of (d) S11 and S21 of the four-layer inhomogeneous structure reconstructed from effective parameters (“Effective”) extracted by our method and obtained from simulations (“Two-layer”). In the figures, “LAO” and “PM” refer to the retrieval method [17] and the proposed method. Fig. 8. Results for the real part of the normalized forward wave impedance of the omega MM slab of the four-layer inhomogeneous structure [FR4, omega MM slab (Bian.), PVC sample, and omega-wire MM slab (Bian.)]. (a) Results for our method and the method in [40] (denoted as “Shi-Liang”) for Nmax = 3. (b) Results for our method and the method in [40] (denoted as “Shi-Liang”) for Nmax = 4. Extraction results by the method in [17] (denoted as “LAO”) are also shown here for comparison purposes. in [17], [24], and [25] output similar results. For instance, Fig. 7(a) and (b) shows the frequency dependences of z 2+ and n 2 of the omega MM slab extracted from our algorithm and the method in [17]. For brevity, the frequency dependence of z 2− is not presented in Fig. 7. Good agreement between retrieved electromagnetic parameters z 2+ and n 2 of the omega MM slab by our method and the methods in [17], [24], and [25] is observed in Fig. 7(a) and (b). Additionally, we also compared the simulated S-parameters of the four-layer inhomogeneous structure with its S-parameters obtained from the effective parameters by our algorithm using the TMM method [43], as shown in Fig. 7(c) and (d). We note from Figs. 5(e)–(h) and 7(c) and (d) that our method is applicable for the extraction of the electromagnetic properties of different bianisotropic MM slabs of various inhomogeneous MM structures. In addition to the application of the extraction methods [17], [24], [25] and our method for obtaining the electromagnetic parameters of the omega MM slab of the four-layer inhomogeneous structure in Fig. 3(c), we also applied the method in [40] for the same purpose. This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 8 Fig. 8(a) and (b) shows the retrieved real part of z 2+ of the omega MM slab for different Nmax values. Extracted z 2− and n 2 are not shown here for conciseness. Results from the extraction method [17] are also shown in Fig. 8 for comparison. While simulated S-parameters of the omega MM slab are used in the application of the method in [17], simulated S-parameters of the four-layer structure are used in the application of the method in [40] and our method. From the dependences in Fig. 8(a) and (b), we note two points. First, extracted real parts of z 2+ by our method are more closer than the retrieved real parts of z 2+ by the method in [40] to the extracted real part of z 2+ by the method in [17] for two different values of Nmax . This is because while our proposed method is applicable for the extraction of electromagnetic properties of bianisotropic materials in inhomogeneous structures, the method in [40] is only applicable for the extraction of electromagnetic properties of isotropic materials of inhomogeneous structures. Second, while extracted real parts of z 2+ by the method in [40] change for different Nmax values, extracted real parts of z 2+ by our method do not change much for different Nmax values. This is because our proposed method uses a two-sided propagation and the method in [40] uses a one-sided propagation in extracting electromagnetic parameters of materials. B. Analysis of Noise Effect and Iterative Error Up to now, we have applied our algorithm for the extraction of electromagnetic properties of different inhomogeneous structures without focusing on the effects of noise and iterative error. However, in a real measurement environment, measured S-parameters have some errors arising from the limits of the measuring instrument (e.g., a vector network analyzer) such as finite directivity and dynamic range, as well as from limited accuracy of calibration kits (such as thrureflect-line [48]) used for the calibration of the measuring instrument. Therefore, their effects on our proposed algorithm should be examined to assess the sensitivity and endurance of our algorithm. Toward this end, in line with the extraction algorithm [40], we superposed a random noise to the simulated S-parameters of the analyzed inhomogeneus structures, then applied our extraction algorithm, and finally compared these simulated S-parameters with S-parameters reconstructed by our method using extracted effective parameters via the TMM method [43]. For example, Fig. 9(a)–(d) shows the extracted electromagnetic parameters (z 3+ , z 3− , and n 3 ) of the SRR-wire MM slab (Bian.) of the three-layer inhomogeneous structure in Fig. 3(b). Before the application of our algorithm, a normally distributed random noise with the mean value ρ = 0 and the standard deviation σ = 0.050 was applied to both the real and imaginary parts of the simulated S11 of this inhomogeneous structure over the frequency ranges 8–9 and 14.5–15.5 GHz, where the sharp variations of z 3+ and z 3− are observable [see Fig. 6(a) and (b)]. It is seen from Fig. 9(a)–(d) that electromagnetic parameters extracted by our extraction algorithm still follow the electromagnetic properties retrieved by the sole extraction method [24] (using simulated S-parameters of the SRR-wire MM slab (Bian.) only) IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES Fig. 9. Extraction results for the three-layer inhomogeneous structure [SRR MM slab (Bian.), FR4 material, and SRR-wire MM slab (Bian.)] in Fig. 3(b) when the normally distributed random noise (mean value ρ = 0 and standard deviation σ = 0.05) is added to S11 of the inhomogeneous structure over 8–9- and 14.5–15.5-GHz ranges. (a) Normalized (relative to air) forward wave impedance of the SRR-wire MM slab. (b) Normalized (relative to air) backward wave impedance of the SRR-wire MM slab. (c) Refractive index of the SRR-wire MM slab. (d) Closer view of the extracted m{z 3− }. Magnitudes of (e) S11 and S21 and (f) S21 of the inhomogeneous structure reconstructed from effective parameters (“Effective”) extracted by our method and obtained from simulations (“Two-layer”). In the figures, “SW” and “PM” refer to the stepwise method [24] and the proposed method. although some random noise is applied over the S11 parameter. However, addition of this noise over simulated S11 increases the difference between electromagnetic parameters extracted by our algorithm and by the sole extraction method [24]. For example, as noted from Figs. 6(d) and 9(d), m{z 3− } values at f = 14.8 GHz extracted by the sole extraction method [24] and the proposed algorithm are −5.73 and −5.14, respectively—showing an around 12% variation, as compared to 10% variation for the dependences obtained by the same methods in Fig. 6(d). On the other hand, we performed some numerical analysis for understanding the effect of iterative errors on electromagnetic properties extracted by our algorithm. As noted in [40], round-off error, the number of iterations, and the electromagnetic parameters of layers initially extracted by the methods in [11], [12], [17], [24], [25], and [39] (refer to the first step of our algorithm given in Section III) may influence the performance of our algorithm. Among these terms, we will focus on the effect of the number of iterations. The interested reader can refer to [40] for the analysis of other terms. Here, we considered a three-layer (m = 3) inhomogeneous structure composed of a bianisotropic MM slab (d1 = 6.0 mm), FR4 material (d2 = 10.0 mm), and Marble material (d3 = 5.0 mm) with relative complex permittivity εr3 = 8.3. For the synthesis of the bianisotropic MM slab, we applied the Lorentz This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. HASAR et al.: DETERMINATION OF EFFECTIVE CONSTITUTIVE PARAMETERS OF INHOMOGENEOUS MMs 9 electromagnetic parameters for the close Nmax values is within the limit of accuracy set priori. V. C ONCLUSION Fig. 10. Extraction results for the synthesized inhomogeneous structure [bianisotropic MM slab (d1 = 6.0 mm), FR4 material (d2 = 10.0 mm), and marble (d3 = 5.0 mm)] for different iteration numbers Nmax . (a) Real and (b) imaginary parts of the normalized (relative to air) forward wave impedance of the bianisotropic MM slab. In the figures, “SW” and “PM” refer to the stepwise method [24] and proposed method. dispersion model [13]. According to this model, εr1 , μr1 , and ξr1 of this slab are calculated from Fe f 2 − f e2 + j δe f Fm f 2 μr1 = 1 − 2 f − f m2 + j δm f Fξ f 2 ξ1 = 1 − 2 f − f ξ2 + j δξ f εr1 = 1 − f2 (28) (29) (30) where f e , f m , and fξ are the electric, magnetic, and magnetoelectric resonance frequencies; f is the operating frequency; δe , δm , and δξ are the electric, magnetic, and magnetoelectric damping frequencies; and Fe , Fm , and Fξ are the coefficients depending on structure of the material [13]. In our analysis, Fe = 0.4, Fm = 0.4, Fξ = 0.15, f e = 6.0 GHz, f m = 5.0 GHz, f ξ = 5.0 GHz, δe = 1.0 GHz, δm = 1.0 GHz, and δξ = 1.0 GHz were utilized for the frequency range 1–10 GHz, producing a passive material with e{z 1+ } ≥ 0, e{z 1− } ≥ 0, and m{n 1 } ≥ 0 [see Fig. 10(a) and (b)]. After we determined forward and backward wave impedances and refractive index of each material using [24, expression (3)]. For FR4 and marble samples, their magnetoelectric coupling coefficients (ξ2 and ξ3 ) and relative permeabilities (μ2 and μ3 ) were set to zero and one, respectively. Next, S-parameters of each material were computed from [24, expressions (1) and (2)]. Finally, S-parameters of the inhomogeneous structure were determined by using the TMM method [43]. Fig. 10(a) and (b) shows the frequency dependence of the extracted real and imaginary parts of z 1+ of the bianisotropic MM slab of the synthesized inhomogeneous structure. It is seen from Fig. 10(a) and (b) that while z 1+ extracted by our algorithm for Nmax = 40 follows the z 1+ extracted by the sole extraction method [24], it is not true for Nmax = 80, especially for frequencies between 4 and 6 GHz, where sharp variation of electromagnetic parameters is notable. The reason for this is that our proposed algorithm, as the iterative algorithm proposed in [40], is iterative, and thus its accuracy is affected by the accuracy of electromagnetic parameters of the layers determined before each iteration process, as discussed in the first step in our algorithm in Section III. Therefore, the value of Nmax should be first set to a smaller number around the resonance region of MM slabs and then can be increased gradually to a point where the difference between extracted We have proposed a recursive retrieval procedure for the accurate determination of electromagnetic properties of bianisotropic inhomogeneous MM slabs using measured/simulated S-parameters. For validation and accuracy analysis of our proposed method, we performed simulation analysis by considering a total of three different inhomogeneous structures. From this analysis, we noted the following points. First, sole extraction methods do not accurately extract electromagnetic properties of inhomogeneous materials, since they are applicable for homogeneous materials only. Second, the extraction method applicable for isotropic inhomogeneous structures fails to extract correct electromagnetic parameters of bianisotropic MM slabs constituting the analyzed inhomogeneous structures. Third, extracted electromagnetic parameters of isotropic/bianisotropic MM slab by our method slightly differ from those obtained by the sole extraction methods (using simulated S-parameters of the isotropic/bianisotropic MM slab only). We think that this difference stems from a coupling effect between the layers of the analyzed inhomogeneous structure, spatial effects as a consequence of near field boundary effects, and coupling between electric and magnetic fields (bianisotropic MM slabs). Fourth, while the accuracy of our extraction method is not much altered by a small change in the number of maximum iteration number (from Nmax = 3 to Nmax = 4), the accuracy of the extraction method feasible for isotropic inhomogeneous MM slabs greatly suffers from this change. This difference comes from the nature of the algorithms of our method (two-sided propagation) and the method for isotropic inhomogeneous MM slabs (one-sided propagation). In addition to validation and accuracy analysis, we also analyzed the effects of noise and iterative error by using the results obtained from our algorithm and the method existing in the literature. We observed that our proposed algorithm is robust in producing accurate electromagnetic parameters even in the presence of an additional noise effect. Besides, we also observed that by increasing maximum iteration number considerably (from Nmax = 40 to Nmax = 80 for the analysis of the synthesized inhomogeneous MM structure), the accuracy of our iterative algorithm may decrease due to any inaccurate determination of electromagnetic parameters of the layers before each iteration process. Therefore, the value of iteration numbers should be first set to a smaller number around the resonance region of MM slabs and then can be increased gradually to a point where the difference between extracted electromagnetic parameters for the close two iteration number values is within the limit of accuracy set priori. A PPENDIX From the S11 expressions obtained from steps 2 and 3 in Section III, one can determine Tu2 in terms of Z u+ and Z u− in (A.1)–(A.5). In a similar manner, from the S22 expressions obtained in steps 4 and 5 in Section III, one can obtain Tu2 in This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. 10 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES terms of Z u+ and Z u− in (A.1) and (A.6)–(A.9) 2 χ1 χ2 2 χ5 χ6 Tu S = Tu S = 11 22 χ3 χ4 χ7 χ8 (A.1) where Z u+ (u ) (u ) χ1 = 1 + Tu22 S112 Z u+2 + 1 − −2 Tu22 S112 Z u− Z u2 (u 1 ) + (u 1 ) Z u 1 − Z u−1 Z u+ χ2 = 1 + S11 Z u+1 Z u−1 + S11 (u ) χ3 = 1 + Tu22 S112 Z u+2 − 1 − Z u+2 2 (u 2 ) Tu S Z u−2 2 11 Z u+ (u 1 ) + (u 1 ) Z u+1 Z u−1 − S11 Z u 1 − Z u−1 Z u− χ4 = 1 + S11 χ5 = 1 + χ6 = 1 + χ7 = 1 + χ8 = 1 + Z u− (m ) (m ) Tu21 S22 2 Z u−1 + 1 − +1 Tu21 S22 2 Z u+ Z u1 (m 1 ) − (m 1 ) Z u+2 Z u−2 + S22 S22 Z u 2 − Z u+2 Z u− Z u− (m 2 ) (m 2 ) Z u− Tu21 S22 Z u−1 − 1 − +1 Tu21 S22 Z u1 (m ) (m ) S22 1 Z u+2 Z u−2 − S22 1 Z u−2 − Z u+2 Z u+ . (A.2) (A.3) (A.4) (A.5) (A.6) (A.7) (A.8) (A.9) Equating both Tu expressions in (A.1), one can find 4 Z u− + 5 Z u− + 6 2 1 Z u− + 2 Z u− + 3 2 −4 Z u+ + 5 Z u+ + 6 2 = −1 Z u+ + 2 Z u+ + 3 2 (A.10) which expresses the relation between Z u− and Z u+ [and electromagnetic parameters of other layers ( p = u)], where 1 , 2 , 3 , 4 , 5 , and 6 are given in (18)–(23). Then, from the S21 expressions obtained from steps 2 and 3 in Section III, Z u− is written by Z u+ and electromagnetic parameters of other layers ( p = u) after substituting Tu2 for S11 (or S22 ) in (A.1) into (A.10) as Z u− = 8 + 9 Z u+ 9 + 10 Z u+ (A.11) where 8 , 9 , and 10 are presented in (24)–(26). Finally, substituting (A.11) into (A.10), the metric function F(Z u+ ) in (17) is derived. R EFERENCES [1] D. Schurig et al., “Metamaterial electromagnetic cloak at microwave frequencies,” Science, vol. 314, no. 5801, pp. 977–980, 2006. [2] J. B. Pendry, “Negative refraction makes a perfect lens,” Phys. Rev. Lett., vol. 85, no. 18, pp. 3966–3969, 2000. [3] C. R. Simovski, “Material parameters of metamaterials: A review,” Opt. Spectrosc., vol. 107, no. 5, pp. 726–753, 2009. [4] E. Martini, G. M. Sardi, and S. Maci, “Homogenization processes and retrieval of equivalent constitutive parameters for multisurfacemetamaterials,” IEEE Trans. Antennas Propag., vol. 62, no. 4, pp. 2081–2092, Apr. 2014. [5] A. M. Nicolson and G. Ross, “Measurement of the intrinsic properties of materials by time–domain techniques,” IEEE Trans. Instrum. Meas., vol. IM-19, no. 4, pp. 377–382, Nov. 1970. [6] W. B. Weir, “Automatic measurement of complex dielectric constant and permeability at microwave frequencies,” Proc. IEEE, vol. JPROC-62, no. 1, pp. 33–36, Jan. 1974. [7] J. Baker-Jarvis, E. J. Vanzura, and W. A. Kissick, “Improved technique for determining complex permittivity with the transmission/reflection method,” IEEE Trans. Microw. Theory Techn., vol. 38, no. 8, pp. 1096–1103, Aug. 1990. [8] A.-H. Boughriet, C. Legrand, and A. Chapoton, “Noniterative stable transmission/reflection method for low-loss material complex permittivity determination,” IEEE Trans. Microw. Theory Techn., vol. 45, no. 1, pp. 52–57, Jan. 1997. [9] V. V. Varadan and R. Ro, “Unique retrieval of complex permittivity and permeability of dispersive materials from reflection and transmitted fields by enforcing causality,” IEEE Trans. Microw. Theory Techn., vol. 55, no. 10, pp. 2224–2230, Oct. 2007. [10] K. Chalapat, K. Sarvala, J. Li, and G. S. Paraoanu, “Wideband reference-plane invariant method for measuring electromagnetic parameters of materials,” IEEE Trans. Microw. Theory Techn., vol. 57, no. 9, pp. 2257–2267, Sep. 2009. [11] D. R. Smith, S. Schultz, P. Markoš, and C. M. Soukoulis, “Determination of effective permittivity and permeability of metamaterials from reflection and transmission coefficients,” Phys. Rev. B, Condens. Matter, vol. 65, Apr. 2002, Art. no. 195104. [12] X. Chen, T. M. Grzegorczyk, B.-I. Wu, J. Pacheco, Jr., and J. A. Kong, “Robust method to retrieve the constitutive effective parameters of metamaterials,” Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat. Interdiscip. Top., vol. 70, Jul. 2004, Art. no. 016608. [13] X. Chen, B.-I. Wu, J. A. Kong, and T. M. Grzegorczyk, “Retrieval of the effective constitutive parameters of bianisotropic metamaterials,” Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat. Interdiscip. Top., vol. 71, Apr. 2005, Art. no. 046610. [14] T. Driscoll, D. N. Basov, W. J. Padilla, J. J. Mock, and D. R. Smith, “Electromagnetic characterization of planar metamaterials by oblique angle spectroscopic measurements,” Phys. Rev. B, Condens. Matter, vol. 75, no. 11, 2007, Art. no. 115114. [15] C. Menzel, C. Rockstuhl, T. Paul, F. Lederer, and T. Pertsch, “Retrieving effective parameters for metamaterials at oblique incidence,” Phys. Rev. B, Condens. Matter, vol. 77, May 2008, Art. no. 195328. [16] C. Menzel, C. Rockstuhl, T. Paul, and F. Lederer, “Retrieving effective parameters for quasiplanar chiral metamaterials,” Appl. Phys. Lett., vol. 93, no. 23, 2008, Art. no. 233106. [17] Z. Li, K. Aydin, and E. Ozbay, “Determination of the effective constitutive parameters of bianisotropic metamaterials from reflection and transmission coefficients,” Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat. Interdiscip. Top., vol. 79, no. 2, 2009, Art. no. 026610. [18] Z. Li et al., “Chiral metamaterials with negative refractive index based on four ‘U’ split ring resonators,” Appl. Phys. Lett., vol. 97, no. 8, 2010, Art. no. 081901. [19] O. Luukkonen, S. I. Maslovski, and S. A. Tretyakov, “A stepwise nicolson–ross–weir-based material parameter extraction method,” IEEE Antennas Wireless Propag. Lett., vol. 10, pp. 1295–1298, Nov. 2011. [20] A. Alù, “First-principles homogenization theory for periodic metamaterials,” Phys. Rev. B, Condens. Matter, vol. 84, Aug. 2011, Art. no. 075153. [21] J. J. Barroso and U. C. Hasar, “Resolving phase ambiguity in the inverse problem of transmission/reflection measurement methods,” J. Infr., Millim., Terahertz Waves, vol. 32, no. 6, pp. 857–866, 2011. [22] S. Kim, E. F. Kuester, C. L. Holloway, A. D. Scher, and J. Baker-Jarvis, “Boundary effects on the determination of metamaterial parameters from normal incidence reflection and transmission measurements,” IEEE Trans. Antennas Propag., vol. 59, no. 6, pp. 2226–2240, Jun. 2011. [23] J. J. Barroso and U. C. Hasar, “Constitutive parameters of a metamaterial slab retrieved by the phase unwrapping method,” J. Infr. Millim. Terahertz Waves, vol. 33, no. 2, pp. 237–244, 2012. [24] U. C. Hasar, J. J. Barroso, C. Sabah, Y. Kaya, and M. Ertugrul, “Stepwise technique for accurate and unique retrieval of electromagnetic properties of bianisotropic metamaterials,” J. Opt. Soc. Amer. B, Opt. Phys., vol. 30, no. 4, pp. 1058–1068, 2013. [25] V. Milosevic, B. Jokanovic, and R. Bojanic, “Effective electromagnetic parameters of metamaterial transmission line loaded with asymmetric unit cells,” IEEE Trans. Microw. Theory Techn., vol. 61, no. 8, pp. 2761–2772, Aug. 2013. [26] V. S. Asadchy, I. A. Faniayeu, Y. Ra’di, and S. A. Tretyakov, “Determining polarizability tensors for an arbitrary small electromagnetic scatterer,” Photon. Nanostruct.—Fundam. Appl., vol. 12, no. 4, pp. 298–304, 2014. [27] T. D. Karamanos, A. I. Dimitriadis, and N. V. Kantartzis, “Robust technique for the polarisability matrix retrieval of bianisotropic scatterers via their reflection and transmission coefficients,” IET Microw., Antennas Propag., vol. 8, no. 15, pp. 1398–1407, 2014. This article has been accepted for inclusion in a future issue of this journal. Content is final as presented, with the exception of pagination. HASAR et al.: DETERMINATION OF EFFECTIVE CONSTITUTIVE PARAMETERS OF INHOMOGENEOUS MMs [28] T. D. Karamanos, S. D. Assimonis, A. I. Dimitriadis, and N. V. Kantartzis, “Effective parameter extraction of 3D metamaterial arrays via first-principles homogenization theory,” Photon. Nanostruct.— Fundam. Appl., vol. 12, no. 4, pp. 291–297, 2014. [29] D. Cohen and R. Shavit, “Bi-anisotropic metamaterials effective constitutive parameters extraction using oblique incidence S-parameters method,” IEEE Trans. Antennas Propag., vol. 63, no. 5, pp. 2071–2078, May 2015. [30] U. C. Hasar, J. J. Barroso, T. Karacali, and M. Ertugrul, “Semi-infinite reflection coefficients of bi-anisotropic metamaterial slabs including boundary effects,” IEEE Microw. Wireless Compon. Lett., vol. 25, no. 5, pp. 283–285, May 2015. [31] U. C. Hasar et al., “Reference-plane-invariant effective thickness and electromagnetic property determination of isotropic metamaterials involving boundary effects,” IEEE J. Sel. Topics Quantum Electron., vol. 21, no. 4, Jul./Aug. 2015, Art. no. 4700211. [32] Y. Shi, Z.-Y. Li, L. Li, and C.-H. Liang, “An electromagnetic parameters extraction method for metamaterials based on phase unwrapping technique,” Waves Random Complex Media, vol. 26, no. 4, pp. 417–433, 2016. [33] A. Andryieuski, A. V. Lavrinenko, M. Petrov, and S. A. Tretyakov, “Homogenization of metasurfaces formed by random resonant particles in periodical lattices,” Phys. Rev. B, Condens. Matter, vol. 93, May 2016, Art. no. 205127. [34] Y. Shi, T. Hao, L. Li, and C.-H. Liang, “An improved NRW method to extract electromagnetic parameters of metamaterials,” Microw. Opt. Technol. Lett., vol. 58, no. 3, pp. 647–652, 2016. [35] X. X. Liu, Y. Zhao, and A. Alù, “Polarizability tensor retrieval for subwavelength particles of arbitrary shape,” IEEE Trans. Antennas Propag., vol. 64, no. 6, pp. 2301–2310, Jun. 2016. [36] U. C. Hasar, J. J. Barroso, M. Bute, A. Muratoglu, and M. Ertugrul, “Boundary effects on the determination of electromagnetic properties of bianisotropic metamaterials from scattering parameters,” IEEE Trans. Antennas Propag., vol. 64, no. 8, pp. 3459–3469, Aug. 2016. [37] U. C. Hasar, A. Muratoglu, M. Bute, J. J. Barroso, and M. Ertugrul, “Effective constitutive parameters retrieval method for bianisotropic metamaterials using waveguide measurements,” IEEE Trans. Microw. Theory Techn., vol. 65, no. 5, pp. 1488–1497, May 2017. [38] U. C. Hasar, G. Buldu, and J. J. Barroso, “Waveguide method for electromagnetic parameter extraction of weakly coupled metamaterials,” IEEE Microw. Wireless Compon. Lett., vol. 27, no. 9, pp. 851–853, Sep. 2017. [39] D. R. Smith, D. C. Vier, T. Koschny, and C. M. Soukoulis, “Electromagnetic parameter retrieval from inhomogeneous metamaterials,” Phys. Rev. E, Stat. Phys. Plasmas Fluids Relat. Interdiscip. Top., vol. 71, Mar. 2005, Art. no. 036617. [40] Y. Shi, Z.-Y. Li, K. Li, L. Li, and C.-H. Liang, “A retrieval method of effective electromagnetic parameters for inhomogeneous metamaterials,” IEEE Trans. Microw. Theory Techn., vol. 65, no. 4, pp. 1160–1178, Apr. 2017. [41] Z.-G. Dong et textit al, “Non-left-handed transmission and bianisotropic effect in a π -shaped metallic metamaterial,” Phys. Rev. B, Condens. Matter, vol. 75, no. 7, 2007, Art. no. 075117. [42] N. Katsarakis, T. Koschny, M. Kafesaki, E. N. Economou, and C. M. Soukoulis, “Electric coupling to the magnetic resonance of split ring resonators,” Appl. Phys. Lett., vol. 84, no. 15, pp. 2943–2945, Apr. 2004. [43] U. C. Hasar, M. Bute, J. J. Barroso, C. Sabah, Y. Kaya, and M. Ertugrul, “Power analysis of multilayer structures composed of conventional materials and bi-anisotropic metamaterial slabs,” J. Opt. Soc. Amer. B, Opt. Phys., vol. 31, no. 5, pp. 939–947, 2014. [44] W. C. Chew, Waves and Fields in Inhomogenous Media. New York, NY, USA: IEEE Press, 1995. [45] W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes: The Art of Scientific Computing. New York, NY, USA: Cambridge Univ. Press, 2007. [46] R. Marqués, F. Medina, and R. Rafii-El-Idrissi, “Role of bianisotropy in negative permeability and left-handed metamaterials,” Phys. Rev. B, Condens. Matter, vol. 65, no. 14, 2002, Art. no. 144440. 11 [47] K. Aydin, Z. Li, M. Hudlička, S. A. Tretyakov, and E. Ozbay, “Transmission characteristics of bianisotropic metamaterials based on omega shaped metallic inclusions,” New J. Phys., vol. 9, p. 326, Sep. 2007. [48] G. F. Engen and C. A. Hoer, “Thru-reflect-line: An improved technique for calibrating the dual six-port automatic network analyzer,” IEEE Trans. Microw. Theory Techn., vol. MTT-27, no. 12, pp. 987–993, Dec. 1979. Ugur Cem Hasar (M’00) received the B.Sc. and M.Sc. degrees (Hons.) in electrical and electronics engineering from Cukurova University, Adana, Turkey, in 2000 and 2002, respectively, and the Ph.D. degree (Hons.) in electrical and computer engineering from The State University of New York at Binghamton, Binghamton, NY, USA, in 2008. From 2000 to 2005, he was a Research and Teaching Assistant with the Department of Electrical and Electronics Engineering, Cukurova University. From 2005 to 2008, he was a Research Assistant with the Department of Electrical and Electronics Engineering, Ataturk University, Erzurum, Turkey, where he was an Assistant Professor from 2009 to 2011 and also an Associate Professor from 2011 to 2013. Since 2017, he has been a full-time Professor with the Department of Electrical and Electronics Engineering, Gaziantep University, Gaziantep, Turkey. His current research interests include nondestructive testing and evaluation of materials using microwaves, novel calibration-dependent, calibration-independent techniques for the electrical and physical characterization of conventional materials at microwaves, millimeter waves, and THz frequencies, high-temperature packaging for high power density applications, porous silicon-based devices and their applications, and metamaterials. Gul Buldu received the B.Sc. degree (Hons.) from Mustafa Kemal University, Hatay, Turkey, in 2008, and the M.Sc. degree from Gaziantep University, Gaziantep, Turkey, in 2017. She is also a Scholar with the Scientific and Technological Research Council of Turkey under the TUBITAK Project 112R032. Her current research interests include the characterization of materials by calibration-independent methods and metamaterials. Yunus Kaya received the B.Sc. degree in electrical and electronics engineering and B.Sc. degree in mechanical engineering from Ataturk University, Erzurum, Turkey, in 2011 and 2012, respectively. He is currently pursuing the Ph.D. degree in electrical and electronics engineering at Ataturk University. Since 2013, he has been an Instructor with the Department of Electricity and Energy, Bayburt University, Bayburt, Turkey. His research interests include characterization of materials by microwaves and metamaterials. Gokhan Ozturk received the B.Sc. degree from Fırat University, Elazığ, Turkey, in 2009, and the M.Sc. degree from Ataturk University, Erzurum, Turkey, in 2014, all in electrical and electronics engineering. He was a Research Assistant with the Department of Electrical and Electronics Engineering, Iğdır University, Iğdır, Turkey, from 2010 to 2012, and Kafkas University, Kars, Turkey, in 2012, respectively. Since 2012, he has been a Research Assistant with the Department of Electrical and Electronics Engineering, Ataturk University. His current research interests include the characterization of material by microwaves, metamaterials, and numerical methods in electromagnetics.