## Abstract

We present a mathematical model that allows interpreting the dispersion and attenuation of modes in hollow-core fibers (HCFs) on the basis of single interface reflection, giving rise to analytic and semi-analytic expressions for the complex effective indices in the case where the core diameter is large and the guiding is based on the reflection by a thin layer. Our model includes two core-size independent reflection parameters and shows the universal inverse-cubed core diameter dependence of the modal attenuation of HCFs. It substantially reduces simulation complexity and enables large scale parameter sweeps, which we demonstrate on the example of a HCF with a highly anisotropic metallic nanowire cladding, resembling an indefinite metamaterial at high metal filling fractions. We reveal design rules that allow engineering modal discrimination and show that metamaterial HCFs can principally have low losses at mid-IR wavelengths (< 1 dB/m at 10.6 *µ*m). Our model can be applied to a great variety of HCFs with large core diameters and can be used for advanced HCF design and performance optimization, in particular with regard to dispersion engineering and modal discrimination.

© 2016 Optical Society of America

## 1. Introduction

Solid silica fibers are well established low loss waveguides in the visible and near infrared (IR) range. In the mid-IR, being highly relevant for emerging areas such as environmental and life-science [1] or metrology [2], such fibers are not employed due to exceedingly large material absorption [3]. Other glass systems solve that issue (e.g., chalcogenides, fluorides or tellurites [4, 5]) but are difficult to process, show low damage thresholds and susceptibility to radiation. Furthermore, applications such as gas spectroscopy or biomedical sensing demand strong interaction of the light with the medium under investigation, which is challenging in solid fibers.

Improvements in fabrication technology have recently led to disruptive developments in the field of hollow core fiber (HCFs) – fiber with air cores – solving the aforementioned issues and opening up new application areas for optical fibers. In this case, the guided power is dominantly concentrated in the ‘empty’ core with only a small field overlap with the host glass, strongly reducing the effect of material absorption. Since the average refractive index (RI) of the cladding is larger than that of core, the modes of HCFs are intrinsically leaky with the light guidance relying on photonic band gaps [6], omnidirectional reflection [7, 8], low-density of states [9], anti-resonant guidance [10, 11] and metamaterials [12–14]. The modes in strongly confining HCFs are comparable to those of hollow metal waveguides, and thus reducing the losses demands increasing core diameters. It is commonly accepted that losses scale with 1/*R*^{3} (*R*: core radius) [14–19], however, no analytic expression exists which clarifies this dependency.

Here, we introduce an approximate model describing the complex effective index of various HCFs with comparably large air core diameters by analytic expressions, emphasizing the 1/*R*^{3} dependence (Fig 1(a)). The model reduces the complexity of the cylindrical HCF to a single reflection at a planar interface. The interface properties are included in two dimensionless reflection parameters which are independent of the core diameter and can either be calculated analytically or, in the case of more complicated cladding geometries, obtained numerically. The resulting analytical expressions unambiguously indicate that the 1/*R*^{3} dependence is a universal feature of HCFs. As our model substantially reduces the simulation complexity, it enables large scale parameter sweeps, since simulations of the full HCF geometry are unnecessary. We apply our model to the case of a HCF with a highly anisotropic cladding consisting of a ring of longitudinally invariant metallic nanowires, which at high metal filling fractions resembles an indefinite metamaterial (iMM), and define straightforward design rules allowing for modal discrimination engineering.

This article is structured as follows. In Sec. 2 we derive approximate analytical expressions for the complex effective index. In Sec. 3 our model is applied to an example system, which is a hollow core fiber with iMM cladding. We show that the model accurately reproduces the complex effective index evolution obtained from finite element modelling. We also define design rules for the optimization of modal discrimination and show the optimal structural parameters for a practically relevant geometry.

## 2. Derivation of the analytic equations for the complex effective index

Our reflection model assumes that the modal propagation inside HCFs can be approximated by grazing-incidence rays at the air-cladding interface (Fig. 1). The modal attenuation of HCFs thus originates from the attenuation of the individual reflections, i.e., from the non-unity reflection coefficient of the corresponding planar interface (|*r*|^{2} < 1, *r*: amplitude reflection coefficient). Our model correlates this reflection with the various leaky modes propagating inside the air core. The derivation starts by analyzing the reflection of a single interface under grazing incidence. The resulting boundary conditions, in combination with the fields of the cylindrical geometry, finally yields analytic expressions for the complex effective index for the various modes, which are entirely characterized by two dimensionless and polarization-dependent reflection parameters. The following sections present the main steps of the calculation, with more details in Appendix A.

#### 2.1. Field in the core

The fields of cylindrical waveguides with longitudinal invariance are given by the solutions to the Helmholtz equation [20, 21]:

with the tangential components (*z*- and

*ϕ*-components) given by

Here, *J*_{m} and *J*^{′}_{m} are the Bessel function of *m*-th order and its first derivative, *m* is the azimuthal mode index, *β* and *κ* are the axial and radial wave vector components (Fig. 1(b)), *k*_{0} is the vacuum wave number with
${k}_{0}^{2}={\beta}^{2}+{\kappa}^{2}$, and
${Z}_{0}=\sqrt{{\mu}_{0}/{\u03f5}_{0}}$ is the vacuum impedance. In the following we will derive approximate solutions for the lowest order modes in case of large core diameters (*R* ≫ *λ*), resulting in small radial wave vectors (*κ* ≪ *k*_{0}), which corresponds to the case of grazing incidence *ψ* = arcsin *κ*/*k* ≪ 1 (*ψ*: complementary angle of incidence, Fig. 1(b)).

#### 2.2. Planar model

In our model, we separate the calculation of the modal parameters (*κ* and *β*) from the modeling of the cladding. This is achieved by defining two dimensionless reflection parameters ν* _{TE}* and ν

*characterizing the reflection properties of the cladding. As a first step we consider the field in a small volume element of the size Δ*

_{TM}*r*×

*r*Δ

*ϕ*× Δ

*z*at the core-cladding boundary. For Δ

*r*≪

*R*and Δ

*ϕ*≪ 1 we can neglect the curvature of the boundary and approximate the total field as a superposition of incident and reflected plane waves (Fig. 1(b)). This model requires that the reflection at the core-cladding interface solely occurs within the mentioned volume element. For the situation of a simple capillary this condition is always fulfilled, whereas for a HCF with microstructured cladding the thickness

*t*of the reflecting layer must be small compared to the core radius (

*t*≪

*R*). The coordinates and field components of the planar model are related to the quantities of the cylindrical waveguide by

*x*+

*R*→

*r*,

*y*→

*rϕ*,

*E*

_{y}→

*E*etc. In the case of grazing incidence the reflection coefficients for TE and TM waves are close to −1 and their dependency on

_{ϕ}*ψ*can be approximated by a first order series expansion

For a homogeneous dielectric cladding (i.e., a capillary) Eqs. (7) and (8) have analytic forms and can be derived from the Fresnel Equations (see Appendix A). In the case of more complicated claddings such as the iMM considered here, ν* _{TE}* and ν

*depend on material properties, wavelength*

_{TM}*λ*and geometrical parameters, and must be obtained numerically. For the fields at the boundary we can derive the following relations between the longitudinal and the transversal field components (see Appendix A for details). For TE waves we obtain

In the following, Eqs. (9) and (10) will be used as boundary conditions for the mode fields at *r* = *R* with the substitutions *E _{y}* →

*E*, and

_{ϕ}*H*

_{y}→

*H*.

_{ϕ}#### 2.3. TE modes

TE modes are represented by Eqs. (4) and (5) with *A*_{1} = 0 and *m* = 0. In the limit of a perfectly reflecting boundary (*r _{TE}* = −1) the transverse field component is zero at the boundary since perfect conductors have zero penetration depth. Equation (5) leads to
${E}_{\varphi}\left(R\right)\propto {{J}^{\prime}}_{0}\left(\kappa R\right)=0$, and the values of the radial wave number for perfect reflection

*κ*

_{0}are then defined by

*j*

_{1n}by

In the more general case (*r _{TE}* ≈ −1) the boundary condition in Eq. (9) results in

Using a perturbation ansatz *κ* = *κ*_{0} + *κ*_{1} with |*κ*_{1}| ≪ *κ*_{0} and a first order series expansion with respect to *κ*_{1} we obtain

*κ*

_{1}

Using
${\beta}^{2}={k}_{0}^{2}-{\kappa}^{2}$, the complex effective index *n _{eff}* =

*β*/

*k*

_{0}can be calculated assuming |

*κ*

_{1}| ≪

*κ*

_{0}≪

*k*

_{0}

_{0}

*modes*

_{n}*R*

^{3}dependency with respect to the modal attenuation.

#### 2.4. TM modes

The derivation of the TM-case is similar as for the TE-waves using the correct field distribution and the boundary condition (Eq. 10, see Appendix A.2 for details), resulting in the following expressions for the TM_{0n} modes:

#### 2.5. HE/EH modes

For a non-zero azimuthal order (*m* ≠ 0), the modes have six field components (so-called hybrid modes [22]), leading to a more elaborate derivation. Starting again from the case of a perfectly reflecting interface (*r _{TE}* =

*r*= −1) the transverse field components at the boundary are zero (

_{TM}*E*(

_{ϕ}*R*) = 0,

*H*(

_{ϕ}*R*) = 0). Using Eqs. (5) and (6) we obtain a homogeneous set of equations for the coefficients

*A*

_{1}and

*A*

_{2}requiring the determinant to vanish.

Since *R* ≫ λ we can set *β* ≈ *k*_{0} resulting in

Using Eq. (36) we obtain

providing two sets of solutions, each based on the*n*-th zeros of the Bessel functions

*J*

_{m−1}and

*J*

_{m+1}:

To be consistent with the mode terminology of step index fibers [23] we label the modes related to *j*_{m−1} (i.e., *s* = −1) as HE and those related to *j*_{m+1} (i.e., *s* = +1) as EH.

In the more general case *r _{TE}* ≈ −1 and

*r*≈ −1 the transverse field components must fulfill both boundary conditions (Eqs. 9 and 10). Inserting Eqs. (3)–(6) into Eqs. (9) and (10) yields a homogeneous set of equations for

_{TM}*A*

_{1}and

*A*

_{2}whose determinant must vanish:

Similar to the derivation of the TE modes, we use the ansatz *κ* = *κ*_{0} + *κ*_{1} and a first order series expansion of Δ. The resulting expression contains first and second derivatives of the Bessel functions which can be eliminated using Eqs. (38) and (39). Finally, the expression obtained for *κ*_{1} is simplified by a second order series expansion with respect to 1/*R*, leading to

Inserting κ_{1} into Eq. (16) results in the following expression for the complex effective index of the HE_{mn} (*s* = −1) and EH_{mn} (*s* = +1) modes, respectively:

Together with Eqs. (17) and (19), Eq. (26) clearly shows the scaling of the modal losses with 1/*R*^{3} which is reported in numerous publications [14–19] and the influence of the reflection parameters. Particularly for hybrid modes (*m* ≠ 0), the loss is determined by a superposition of the two reflection parameters, showing that both TE and TM reflections contribute to the modal attenuation. This interesting feature directly emerges from Eq. (26) and results from the fact that all HE and EH modes are principally composed of skew rays [22]. This is the key difference to TE/TM modes which only include one reflection parameter in Eqs. (17) and (19) and are generally described by meridional rays, i.e., rays crossing the fiber axis between two reflections. Due to the different values of the *b*-coefficients the modal discrimination can be engineered by varying the cladding properties and analyzing the reflection parameters. In the case of lossy cladding materials (e.g., metals) or close to interference resonances, *ν _{TE}* and

*ν*can be complex-valued. In this case the real parts determine the losses whereas the imaginary parts give a small contribution to the real part of

_{TM}*n*. For a homogeneous dielectric cladding (i.e., a capillary) ν

_{eff}*and ν*

_{TE}*can be taken from Eqs. (32) and (35), resulting in full analytic expressions that correspond to those reported by E. A. J. Marcatili and R. A. Schmeltzer [15] and are used for instance to analyze modal dispersions of Kagome HCFs [24]. Our model also yields analytic expressions for homogenous metallic cladding structures. Again the key advantage of our model is that the cladding properties of HCFs can be optimized by analyzing the reflection of a corresponding planar interface only, since the modal properties originate from the reflection parameters ν*

_{TM}*and ν*

_{TE}*defined in Eqs. (17), (19) and (26).*

_{TM}## 3. Example: indefinite metamaterial cladding hollow core fiber

#### 3.1. Introduction to the structure

To demonstrate the capabilities of our reflection model we consider a highly anisotropic HCF with iMM cladding and large core diameter and analyze its properties at two important IR wavelengths (3 *µ*m and 10.6 *µ*m). We recently analyzed this kind of HCF in the case of deep subwavelength wires using an effective medium approach [25]. In contrast to that work, the model presented in this contribution is not restricted to the condition of having deep subwavelength wire diameters and is applicable to a much greater variety of fiber geometries, examples of which include photonic band gap fibers or anti-resonant HCFs. The motivation for choosing a nanowire-based HCF here is to reveal the validity of our model on such an extreme multiscale example containing realistic geometries and materials, including losses. The iMM cladding is composed of a single ring of longitudinally invariant gold wires embedded in silica concentrically surrounding an empty core (Fig. 2(a)). In addition to wire diameter *d* and inter-wire spacing (pitch) Λ, one key parameter to consider is the gap size between wire and the glass wall *g* (Fig. 2(b)). If this distance falls within the range of the operation wavelength, interference effect can emerge, strongly modifying the reflection parameters and thus the modal properties. Due to the complexity of the cladding structure we determine the two reflection parameters from numerical calculations and apply them to our model to analyze the modal discrimination within iMM-HCFs.

The reflectivity parameters ν* _{TE}* and ν

*are obtained by simulating the reflection properties of a single rectangular unit cell incorporating one gold wire (Fig. 3) using finite element modelling (COMSOL). To enforce TE- and TM-polarization (corresponding to s- and p-polarization) we use either perfect electric (PEC) or perfect magnetic conductors (PMC) boundary conditions at the sides of the unit cell (with respect to the*

_{TM}*y*-direction). The simulation includes an excitation and monitor port at

*x*= 0 and a perfectly matched layer (PML) terminating the structure. The simulations were performed using a 2D model (Fig. 3) with a large out-off-plane wave vector component that accounts for grazing incidence (

*ψ*≤ 0.001) and a fixed and experimentally accessible wire diameter of

*d*= 1

*µ*m [26, 27]. The optical properties of gold and silica were taken from [28] and [29] (ϵ

*(3*

_{silica}*µ*m) = 2.012, ϵ

*(3*

_{gold}*µ*m) = −361.9+58.7i, ϵ

*(10.6*

_{silica}*µ*m) = 4.740+0.387i, ϵ

*(10.6*

_{gold}*µ*m) = −3867+1712i). The reflection coefficients are obtained from simulations, and the reflection parameters are given as the respective derivatives with respect to

*ψ*. It is important to note that the two different wavelengths lead to fundamentally different guidance regimes, as the imaginary part of the silica at 10.6

*µ*m is orders of magnitude larger than at 3

*µ*m, and approaches the value of the real part.

#### 3.2. Verification of model on example structure

To validate our reflection model for the case of an iMM-HCF we compare the relative complex effective indices (Δ*n _{eff}* = Re(1 −

*n*) and Im(

_{eff}*n*)) of the lowest order modes (TE

_{eff}_{01}, TM

_{01}, and HE

_{11}) determined by Eqs. (17), (19), and (26) with full numerical calculations of the entire iMM-HCF structure (

*g*= 1

*µ*m,

*d*= 1

*µ*m,

*f*= 0.57). The full simulations of the TE

_{01}and TM

_{01}modes only require circular sectors including one wire with PEC or PMC as boundary conditions. For the HE

_{11}mode we used one quadrant with a PEC and a PMC boundary condition. The reflection parameter retrieval procedure presented in the previous section yields

*ν*= 0.883,

_{TE}*ν*= 0.193, for

_{TM}*λ*= 3.0

*µ*m, and

*ν*= 0.989,

_{TE}*ν*= 0.378, for

_{TM}*λ*= 10.6

*µ*m. This comparison (Figs. 4(a) and (b)) reveals that our model perfectly reproduces both real and imaginary parts of the effective mode index even at core radii as small 50

*µ*m, which are diameters of typically used silica-based HCFs and is substantially smaller compared to commercially available fibers operating at 10.6

*µ*m (e.g., Omniguide [8]). It should be noted here that the expression for the real part results from the field at an ideally reflecting boundary, i.e. the zeroth-order approximation with respect to

*κ*

_{1}, whereas the imaginary part requires the first order approximation. As a result, Eqs. (17), (19), and (26) precisely describe the dispersions of a large variety of HCFs and show that the physical properties of HCF-modes can be understood on the basis of single interface reflection. It is interesting to note that for the geometry discussed here, the TM

_{01}-mode at 3

*µ*m shows the lowest loss of all modes in the system – a situation which is difficult to achieve in dielectric fibers, emphasizing that iMM-HCFs have important polarization properties beyond that of dielectric HCFs.

In addition, we calculate the dependency of Im(*n _{eff}*) on filling factor for a constant radius of

*R*= 100

*µ*m for the two wavelengths using

*d*= 1

*µ*m and

*g*= 1

*µ*m (Figs. 5(a) and (b)). Three different guidance regimes can be identified: For large fill factors (

*f*> 0.5) the nanowire layer acts as an effective medium which reflectivity gradually changes with

*f*resulting in a gradual variation of

*n*. If the filling factor is decreased further the pitch becomes sufficiently large to cause higher diffraction orders into the cladding causing the oscillations in Im(

_{eff}*n*). In the limit of low filling factors (

_{eff}*f*< 0.1) the effect of the nanowires vanishes and the properties of the fiber are similar to those of a capillary (

*f*= 0). For any value of

*f*, we observe good agreement between the analytical expressions and the FEM simulations, particular away from the regime of diffraction.

#### 3.3. Modal discrimination of the three lowest order modes

The key feature of our model is that the simulation of the HCF-modes is reduced to the analysis of a single reflection at the core-cladding interface, allowing for large-scale parameter sweeps, and in particular to analyze modal discrimination. Equations (17), (19), and (26) suggests that losses increase with increasing mode indices *m* and *n*, revealing that the TE_{01}, TM_{01}, and HE_{11} modes have the lowest losses and are thus the most relevant. Our equations allow specifying design rules for desirably tuning the modal attenuation, i.e., choosing the lowest loss mode in the system. Defining the critical discrimination parameter
$C=2{j}_{11}^{2}/{j}_{01}^{2}-1\approx 4$, we find the discrimination conditions shown in Tab. 1 for the three lowest order modes.

To illustrate the capabilities of our model for a modal discrimination analysis, we discuss the variation of the reflection parameters (which determine modal discrimination) of the iMM-HCF at the two wavelengths considered thus far, as function of gap size g and circumferential metal fill factor *f* = *d/*Λ, assuming a constant wire diameter (*d* = 1 *µ*m).

For a wavelength of 3 *µ*m (Fig. 6(a) and (b)), both *ν _{TE}* and

*ν*have low values for large fill factors

_{TM}*f*indicating that low modal attenuation generally appears in the case where the metal wires occupy a large fraction of the unit cell (Fig. 3). This can be explained by the strong reflection of gold at mid-IR wavelengths, resulting from the small penetration of the electromagnetic field into the metal (skin depth < 30 nm). At small filling fractions, however, the reflection of the metal wires becomes less relevant and leakage loss (i.e., transmission through the single interface) dominates. At this wavelength, the small value of Im

*ϵ*(3

_{silica}*µ*m)< 10

^{−5}suggests a plane wave absorption length of

*L*= 0.16 m, imposing constructive and destructive interference between the silica wall and the metal wire, which is visible in Figs. 6(a,b) via the dependence of the reflection parameters on gap size (for a fixed circumferential fill factor

_{abs}*f*). As a result, modal discrimination requires a precise analysis of the individual modes, which as we have shown can be simply achieved by investigating the reflection parameters. One striking feature of the iMM cladding is associated with the large parameter domains in which the TM

_{01}mode has the lowest loss – a situation which is difficult to achieve in pure dielectric HCFs and can lead to new applications, since radially polarized modes offer properties beyond that of HE

_{11}modes (e.g., focusing to extremely small spot sizes [30]). To demonstrate the superior properties of the iMM cladding, it is helpful to compare its reflection parameters with those of a capillary with homogenous silica cladding (

*ν*= 1.98 and

_{TE}*ν*= 3.98, obtained from Eqs. (32) and (35). The values of our iMM-cladding can be up to two orders of magnitude lower particular away from the interferences, revealing the advantage of using wire based iMM-claddings for achieving low modal attenuation in the mid-IR.

_{TM}At *λ* = 10.6 *µ*m, our model reveals a similar magnitude of the reflection parameters but only one interference-induced oscillation (Fig. 7(a) and (b)), which is less pronounced due to the large damping of the wave within the silica (*L _{abs}* = 19

*µ*m). The HE

_{11}mode dominates the discrimination behavior over a wide range of parameters, with the TM

_{01}and TE

_{01}modes being only relevant within small intervals. At very large filling factors, the TE

_{01}-mode shows the lowest attenuation – a characteristic behavior of metallic tube waveguides and nanowire-enhanced fibers [31, 32]. Higher values of the reflection parameters are observed for smaller filling fractions, indicating the emergence of leakage loss. Compared to a homogenous silica capillary the reflection parameters are about ten to one hundred times smaller (as

*ν*= 1.03, and

_{TE}*ν*= 4.88), confirming the improved guidance of iMM-HCFs. It is interesting to note that the reflection parameters suggest that the lowest loss capillary mode is the TE

_{TM}_{01}mode, revealing that a capillary with high material attenuation behaves similar to a metallic tube. For example, to operate the iMM-HCF within the HE

_{11}mode at 10.6

*µ*m with experimentally realistic parameters (

*d*= 1

*µ*m,

*g*= 1

*µ*m,

*f*= 0.7) [26,27], we obtain

*ν*≈

_{TE}*ν*≈ 0.3, leading to Im

_{TM}*n*= 3.5 · 10

_{eff}^{−3}(

*λ*/

*R*)

^{3}. This corresponds to an attenuation of 0.3 dB/m for

*R*= 400

*µ*m, which is less than observed in commercially available Omniguide fibers [8].

## 4. Conclusions and outlook

We introduce a mathematical model that allows to interpret the dispersion and attenuation of modes in HCFs with large air cores on the basis of single interface reflections, giving rise to analytic and semi-analytic expressions for the complex effective indices when the core diameter is sufficiently large and the guiding is based on reflection by a sufficiently thin layer. We have identified two dimensionless and core-size independent reflection parameters as the key quantities, and show that the commonly-accepted core size dependence of the modal loss (∝ 1/*R*^{3}) is a universal feature of HCFs. Our model significantly reduces the simulation complexity and enables large scale parameter sweeps, as we show on the example of a HCF with a highly anisotropic indefinite metamaterial cladding. We reveal design rules that allow engineering the modal discrimination and show for this multiscale example structure that the model is applicable in cases the core size is comparably large. The investigated metamaterial geometry shows principal losses lower than that of a corresponding dielectric capillary in any desired mode for mid-IR wavelength (e.g., at 10.6 *µ*m losses < 1 dB/m can be achieved). Our calculation approach substantially reduces simulation and computational effort as only the reflection at a single interface needs to be analyzed. The application of our model is not restricted to metamaterial fibers, and can be applied to a wide variety of HCFs, provided that the core sizes are large enough so that grazing incidence can be considered, i.e. the incident angle is sufficiently small to allow for linearly approximating the single interface reflection coefficient. Future work will target Bragg- and Omniguide fibers, potentially reaching full analytic equations for the complex effective indices. Due to its simplicity, we expect that the model will be also useful to analyze the guidance properties of advanced HCFs and to optimize their performance, in particular within the scope of recent developments (e.g. antiresonant fibers with junction nodes [10], negative curvature fiber [33], Kagome fibers [34], and metamaterial fibers [12–14]), since our model can be easily applied to other shapes by adapting the geometry (Fig. 3) in the simulation of the reflection parameters. We also believe that the basic idea to relate the fiber losses to the reflection coefficients of the core-cladding interface might be generalized to noncircular fibers. However, in this case the reflection coefficient will no longer be constant along the boundary requiring a more elaborate calculation.

## Appendix

## A. Relations of the field components at a planar boundary

## A.1. TE waves

The fields at a boundary between air and the cladding medium (Fig. 1(b)) are given by an incident and a reflected plane wave
${E}_{y}^{(i)}(x,z)={E}^{(i)}\mathrm{exp}[i(\kappa x+\beta z)]$ and
${E}_{y}^{(r)}(x,z)={E}^{(r)}\mathrm{exp}[i(-\kappa x+\beta z)]$, respectively. The reflection properties of that boundary are fully defined by the amplitude reflection coefficient *r _{TE}* =

*E*

^{(}

^{r}^{)}

*/E*

^{(}

^{i}^{)}without making any assumptions on the cladding. The longitudinal component of the magnetic field is obtained from the Maxwell equations as

The total fields are superpositions of incident and reflected waves at the boundary (*x* = 0) leading to

If the cladding is a homogeneous dielectric medium with dielectric permittivity *ϵ* the reflection coefficient can be obtained from the Fresnel equations. In the case of grazing incidence (*ψ* ≪ 1) we obtain the following linear approximation of the Fresnel reflection coefficient

For other cladding structures such as the wire-based iMM, we use the linear ansatz for the reflection coefficient given by Eq. (7). The parameter *ν _{TE}* depends on the permittivities of gold and silica and the geometric parameters of the metamaterial, and can be obtained by numerically simulating the reflection of plane waves at the interface (Fig. 3). From Eqs. (31) and (7) and

*κ*=

*k*

_{0}sin

*ψ*≈

*k*

_{0}

*ψ*we obtain the relation between the tangential field components and the reflection parameter given in Eq. (9).

## A.2. TM waves

In the TM case the relations between the field components are similar and can be obtained from the TE relations by substituting *E _{y}* →

*H*,

_{y}*H*→

_{z}*E*,

_{z}*Z*

_{0}→ −1/

*Z*

_{0}

*µ*

_{0}→ −

*ϵ*

_{0},

*ν*→

_{TE}*ν*. In particular, we obtain

_{TM}*ϵ*the Fresnel equations result in

We have defined the reflection coefficient here as the ratio of the transverse magnetic field components in order to get the same mathematical form of expression as for the TE case.

## B. Relations of the Bessel functions

The first derivative of the Bessel function fulfills the following relation (given in [35] for *s* = −1 and *s* = 1 separately).

From the Bessel differential equation [35] and Eq. (36) we obtain

If the argument is a zero of *J _{m}*

_{+}

*we obtain*

_{s}Some values of *j _{mn}* a given in Tab. 2.

## C. Numerical values of the coefficients

Table 3 shows numerical values of the a and b coefficients to be used in Eqs. (17), (19), and (26).

## Acknowledgments

The authors acknowledge funding from the German Science Foundation (DFG) via the grant SCHM 2655/8-1. A.T. acknowledges support from the Alexander Von Humboldt Foundation.

## References and links

**1. **D. Torres, O. Weisberg, G. Shapira, C. Anastassiou, B. Temelkuran, M. Shurgalin, S. A. Jacobs, R. U. Ahmad, T. Wang, U. Kolodny, S. M. Shapshay, Z. Wang, A. K. Devaiah, U. D. Upadhyay, and J. A. Koufman, “OmniGuide photonic bandgap fibers for flexible delivery of CO2 laser energy for laryngeal and airway surgery,” P. Soc. Photo-Opt Ins. **5686**, 310–321 (2005).

**2. **F. Adler, K. C. Cossel, M. J. Thorpe, I. Hartl, M. E. Fermann, and J. Ye, “Phase-stabilized, 1.5 W frequency comb at 2.8–4.8µm,” Opt. Lett. **34**, 1330–1332 (2009). [CrossRef] [PubMed]

**3. **O. Humbach, H. Fabian, U. Grzesik, U. Haken, and W. Heitmann, “Analysis of OH absorption bands in synthetic silica,” J. Non-Cryst. Solids **203**, 19–26 (1996). [CrossRef]

**4. **T. M. Monro and H. Ebendorff-Heidepriem, “Progress in microstructured optical fibers,” Ann. Rev. Mater. Res. **36**, 467–495 (2006). [CrossRef]

**5. **K. F. Lee, N. Granzow, M. A. Schmidt, W. Chang, L. Wang, Q. Coulombier, J. Troles, N. Leindecker, K. L. Vodopyanov, P. G. Schunemann, M. E. Fermann, P. S. J. Russell, and I. Hartl, “Midinfrared frequency combs from coherent supercontinuum in chalcogenide and optical parametric oscillation,” Opt. Lett. **39**, 2056–2059 (2014). [CrossRef] [PubMed]

**6. **J. C. Knight, J. Broeng, T. A. Birks, and P. S. J. Russell, “Photonic band cap guidance in optical fibers,” Science **282**, 1476–1478 (1998). [CrossRef] [PubMed]

**7. **S. D. Hart, G. R. Maskaly, B. Temelkuran, P. H. Prideaux, J. D. Joannopoulos, and Y. Fink, “External reflection from omnidirectional dielectric mirror fibers,” Science **296**, 510–513 (2002). [CrossRef] [PubMed]

**8. **B. Temelkuran, S. D. Hart, G. Benoit, J. D. Joannopoulos, and Y. Fink, “Wavelength-scalable hollow optical fibres with large photonic bandgaps for CO2 laser transmission,” Nature **420**, 650–653 (2002). [CrossRef] [PubMed]

**9. **M. A. Finger, N. Y. Joly, T. Weiss, and P. S. J. Russell, “Accuracy of the capillary approximation for gas-filled kagome-style photonic crystal fibers,” Opt. Lett. **39**, 821–824 (2014). [CrossRef] [PubMed]

**10. **A. Hartung, J. Kobelke, A. Schwuchow, J. Bierlich, J. Popp, M. A. Schmidt, and T. Frosch, “Low-loss single-mode guidance in large-core antiresonant hollow-core fibers,” Opt. Lett. **40**, 3432–3435 (2015). [CrossRef] [PubMed]

**11. **A. Hartung, J. Kobelke, A. Schwuchow, K. Wondraczek, J. Bierlich, J. Popp, T. Frosch, and M. A. Schmidt, “Double antiresonant hollow core fiber-guidance in the deep ultraviolet by modified tunneling leaky modes,” Opt. Express **22**, 19131–19140 (2014). [CrossRef] [PubMed]

**12. **S. Atakaramians, A. Argyros, S. C. Fleming, and B. T. Kuhlmey, “Hollow-core waveguides with uniaxial metamaterial cladding: modal equations and guidance conditions,” J. Opt. Soc. Am. B **29** (9), 2462–2477, (2012). [CrossRef]

**13. **S. Atakaramians, A. Argyros, S. C. Fleming, and B. T. Kuhlmey, “Hollow-core uniaxial metamaterial clad fibers with dispersive metamaterials,” J. Opt. Soc. Am. B **30** (4), 851–867, (2013). [CrossRef]

**14. **M. Yan and N. A. Mortensen, “Hollow-core infrared fiber incorporating metal-wire metamaterial,” Opt. Express **17**, 14851–14864 (2009). [CrossRef] [PubMed]

**15. **E. A. J. Marcatili and R. A. Schmeltzer, “Hollow metallic and dielectric waveguides for long distance optical transmission and lasers,” At&T Tech J **43**, 1783 (1964).

**16. **M. M. Hasan, D. S. Kumar, M. R. C. Mahdy, D. N. Hasan, and M. A. Matin, “Robust optical fiber using single negative metamaterial cladding,” IEEE Photonic Tech. L. **25**, 1043–1046 (2013). [CrossRef]

**17. **M. J. Shawon, G. A. Mahdiraji, M. M. Hasan, B. H. Shakibaei, S. Y. Gang, M. R. C. Mahdy, and F. R. M. Adikan, “Single negative metamaterial-based hollow-core bandgap fiber with multilayer cladding,” IEEE Photonics J. **7**, 4600812 (2015). [CrossRef]

**18. **M. Ibanescu, S. G. Johnson, M. Soljacic, J. D. Joannopoulos, Y. Fink, O. Weisberg, T. D. Engeness, S. A. Jacobs, and M. Skorobogatiy, “Analysis of mode structure in hollow dielectric waveguide fibers,” Phys. Rev. E **67**, 046608 (2003). [CrossRef]

**19. **S. G. Johnson, M. Ibanescu, M. Skorobogatiy, O. Weisberg, T. D. Engeness, M. Soljacic, S. A. Jacobs, J. D. Joannopoulos, and Y. Fink, “Low-loss asymptotically single-mode propagation in large-core OmniGuide fibers,” Opt. Express **9**, 748–779 (2001). [CrossRef] [PubMed]

**20. **M. A. Kaliteevskii, V. V. Nikolaev, and R. A. Abram, “Calculation of the mode structure of multilayer optical fibers based on transfer matrices for cylindrical waves,” Opt. Spectrosc. **88**, 792–795 (2000). [CrossRef]

**21. **P. Yeh, A. Yariv, and E. Marom, “Theory of Bragg fiber,” J. Opt. Soc. Am. **68**, 1196–1201 (1978). [CrossRef]

**22. **J. A. Snyder and J. D. Love, *Optical Waveguide Theory* (Chapman and Hall, 1983).

**23. **T. Okoshi, *Optical Fibers* (Academic, 1982).

**24. **P. S. Russell, P. Holzer, W. Chang, A. Abdolvand, and J. C. Travers, “Hollow -core photonic crystal fibres for gas-based nonlinear optics,” Nat. Photonics **8**, 278–286 (2014). [CrossRef]

**25. **A. Tuniz, M. Zeisberger, and M. A. Schmidt, “Tailored loss discrimination in indefinite metamaterial-clad hollow-core fibers,” Opt. Express **24**, 15702–15709 (2016) [CrossRef] [PubMed]

**26. **P. Uebel, M. A. Schmidt, H. W. Lee, and P. S. J. Russell, “Polarisation-resolved near-field mapping of a coupled gold nanowire array,” Opt. Express **20**, 28409–28417 (2012). [CrossRef] [PubMed]

**27. **H. W. Lee, M. A. Schmidt, and P. S. J. Russell, “Excitation of a nanowire molecule in gold-filled photonic crystal fiber,” Opt. Lett. **37**, 2946 (2012). [CrossRef] [PubMed]

**28. **A. D. Rakic, A. B. Djurisic, J. M. Elazar, and M. L. Majewski, “Optical properties of metallic films for vertical-cavity optoelectronic devices,” Appl. Opt. **37**, 5271–5283 (1998). [CrossRef]

**29. **E. D. Palik, *Handbook of Optical Constants of Solids* (Academic, 1998).

**30. **S. Quabis, R. Dorn, M. Eberler, O. Glockl, and G. Leuchs, “Focusing light to a tighter spot,” Opt. Commun. **179**, 1–7 (2000). [CrossRef]

**31. **J. D. Jackson, *Classical Electrodynamics* (Gruyter, 2006).

**32. **P. Uebel, M. A. Schmidt, M. Scharrer, and P. S. Russell, “An azimuthally polarizing photonic crystal fibre with a central gold nanowire,” New J. Phys. **13**063016 (2011). [CrossRef]

**33. **F. Yu and J. C. Knight, “Spectral attenuation limits of silica hollow core negative curvature fiber,” Opt. Express **21**, 21466–21471 (2013). [CrossRef] [PubMed]

**34. **P. Ghenuche, S. Rammler, N. Y. Joly, M. Scharrer, M. Frosz, J. Wenger, P. S. Russell, and H. Rigneault, “Kagome hollow-core photonic crystal fiber probe for Raman spectroscopy,” Opt. Lett. **37**, 4371–4373 (2012). [CrossRef] [PubMed]

**35. **W. Magnus, F. Oberhettinger, and R. P. Soni, *Formulas and Theorems for the Special Functions of Mathematical Physics* (Springer, 1966). [CrossRef]