betaFeedback

WorksPhysics and Astronomy

Non-Hermitian guided modes and exceptional points using loss-free negative-index materials

Abstract

We analyze the guided modes in coupled waveguides made of negative-index materials without gain or loss. We show that it supports non-Hermitian phenomenon on the existence of guided mode versus geometric parameters of the structure. The non-Hermitian effect is different from parity-time (PT) symmetry, and can be explained by a simple coupled-mode theory with an anti-PT symmetry. The existence of exceptional points and slow-light effect are discussed. This work highlights the potential of loss-free negative-index materials in the study of non-Hermitian optics.

Introduction

Recent years the non-Hermitian physics, especially parity-time (PT\mathcal{PT}) symmetry [1, 2, 3, 4, 5, 6, 7], has attracted much attention in the society of optics, because it provides an additional degree of freedom in manipulating the dynamics of optical waves. Nevertheless, the requirement on proper strength of gain and loss effects hinders the realistic applications of PT\mathcal{PT} symmetric optics. The advances of other categories of non-Hermitian optics, e.g. anti-PT\mathcal{PT} symmetry [8, 9, 10, 11], might overcome this drawback because gain and loss are not strictly required.

The simplest configuration in studying PT\mathcal{PT} symmetric optics is two parallel waveguides (WGs). Such a kind configuration can be readily studied by using Maxwell’s equations, and it supports some intriguing optical effects especially the stopped light at exceptional points (EPs) [12, 13, 14, 15] that separate the conserved and broken PT\mathcal{PT} phases [16, 17, 18, 19, 20]. However, for an optical wave propagating inside a WG, it contains multiple important physical parameters especially the wavevector kk charactering the propagation of phase front and the Poynting vector SS presenting the propagation of energy flux. They carry different informations about the wave, and might have different directions of propagation [21, 22]. The extreme situation is that in a negative-index material (NIM) with simultaneous negative magnetic permeability (μNIM<0\mu_{\mathrm{NIM}} < 0) and electric permittivity (ϵNIM<0\epsilon_{\mathrm{NIM}} < 0), where the direction of kk is opposite to that of SS [23, 24, 25, 26, 27]. However, non-Hermitian feature of coupled WGs containing NIM has not been extensively discussed. Only recently do we notice that Mealy etc. al. [28] have showed that EP of degeneracy can be obtained in two coupled WGs by a proper coupling of forward and backward waves, where the backward waves are generated by a proper designed grating.

In this article, we check the optical waves in coupled WGs made of NIM and positive-index material (PIM). A PIM has positive magnetic permeability and electric permittivity, and covers all the dielectric materials in nature. We show that the guided modes in this system also support non-Hermitian feature, even when all the constituent media are loss-free. At a given angular frequency and WG thicknesses, the wavevector kk of the eigenmodes and the associated eigenvectors vary with the distance aa between the two WGs, and they coalesce at a critical value of aca_c below which the wavevector kk becomes complex. This phase transition point is an EP. A simple non-Hermitian coupled-mode theory is utilized to explain our results, which implies that the field dynamics induced by NIM belongs to anti-PT\mathcal{PT} symmetry. Features at EPs including the slow-light effect are discussed.This study highlights the important novelties of NIM, especially its great potential in the study of non-Hermitian optics by bypassing many restrictions of PT\mathcal{PT} symmetry.

Simulation and analysis

Let us consider the simple structure shown in Figure 1. It contains two straight WGs made of lossless media surrounded by air. One WG is PIM, and we assume that it is a dielectric of ϵPIM=4\epsilon_{\mathrm{PIM}} = 4 and μPIM=1\mu_{\mathrm{PIM}} = 1. The other WG is a NIM that has been discussed by various authors [29, 30, 31]. The well documented structure of NIM with ϵNIM=1−ωp2/ω2\epsilon_{\mathrm{NIM}} = 1 - \omega_p^2/\omega^2 and μNIM=1−Fω2/(ω2−ω02)\mu_{\mathrm{NIM}} = 1 - F\omega^2/(\omega^2-\omega_0^2) can be utilized here, where ωp=2π×10\omega_p=2\pi\times10 GHz, ω0=2π×4\omega_0=2\pi\times4 GHz, and F=0.56F=0.56 [29, 30, 31]. An effective NIM is achieved between 4 GHz and 6 GHz.

Schematic of the configuration under investigation

Figure 1. Schematic of the configuration under investigation, which contains two straight WGs made of lossless media. One medium is a dielectric with εPIM>0\varepsilon_{\mathrm{PIM}}>0 and μPIM=1\mu_{\mathrm{PIM}}=1, while the other one is a NIM with εNIM=−3\varepsilon_{\mathrm{NIM}}=-3 and μNIM=−0.556\mu_{\mathrm{NIM}}=-0.556.

Because NIM requires an intrinsic dispersion of ∂(εNIMω)/∂ω>0\partial(\varepsilon_{\mathrm{NIM}}\omega)/\partial\omega>0 and ∂(μNIMω)/∂ω>0\partial(\mu_{\mathrm{NIM}}\omega)/\partial\omega>0 so as to give a positive energy density [21, 22], in our below analysis we would keep the angular frequency ω\omega a constant, and test the variation of the wavevectors kk of the eigenmodes versus a geometric parameter of the coupled WGs. Here we set ω=2π×5\omega=2\pi\times5 GHz (a free-space wavelength of 6cm), and the dispersion gives εNIM=−3\varepsilon_{\mathrm{NIM}}=-3 and μNIM=−0.556\mu_{\mathrm{NIM}}=-0.556. The index of refraction equals nNIM=εNIMμNIM=−1.291n_{\mathrm{NIM}}=\sqrt{\varepsilon_{\mathrm{NIM}}\mu_{\mathrm{NIM}}}=-1.291. Thickness bb of each WG is 4 cm. The distance aa between the two WGs is chosen as the variable.

Assuming the direction of the wavevector kk is zz, the dispersion and distribution of the eigenmodes can be numerically found by using Maxwell’s equations and boundary conditions [14, 21, 22]. For example, considering the transverse-electric (TE) polarized eigenmodes with non-vanishing EyE_y component

Ey=e−jkz+jωt{E1e−β(x−a/2−b),x>a/2+b,E2e−jαNIM(x−a/2)+E3e+jαNIM(x−a/2),a/2+b>x>a/2,E4e+βx+E5e−βx,a/2>x>−a/2,E6e−jαPIM(x+a/2)+E7e+jαPIM(x+a/2),−a/2>x>−a/2−b,E8e+β(x+a/2+b),x<−a/2−b.(1)E_y=e^{-jkz+j\omega t} \begin{cases} E_1e^{-\beta(x-a/2-b)}, & x>a/2+b,\\ E_2e^{-j\alpha_{\mathrm{NIM}}(x-a/2)}+E_3e^{+j\alpha_{\mathrm{NIM}}(x-a/2)}, & a/2+b>x>a/2,\\ E_4e^{+\beta x}+E_5e^{-\beta x}, & a/2>x>-a/2,\\ E_6e^{-j\alpha_{\mathrm{PIM}}(x+a/2)}+E_7e^{+j\alpha_{\mathrm{PIM}}(x+a/2)}, & -a/2>x>-a/2-b,\\ E_8e^{+\beta(x+a/2+b)}, & x<-a/2-b. \end{cases} \tag*{(1)}

with

k2−β2=ω2c2,(2)k^2-\beta^2=\frac{\omega^2}{c^2}, \tag*{(2)}
k2+αm2=εmμmω2c2,(3)k^2+\alpha_m^2=\frac{\varepsilon_m\mu_m\omega^2}{c^2}, \tag*{(3)}

where the subscript mm stands for NIM and PIM. The associated magnetic fields can be found from Eq. (1) using ∇×E⃗=−∂B⃗/∂t\nabla\times\vec{E}=-\partial\vec{B}/\partial t.

By applying the electromagnetic boundary conditions, and defining

Fm=β+jαm/μmβ−jαm/μm,(4)F_m=\frac{\beta+j\alpha_m/\mu_m}{\beta-j\alpha_m/\mu_m}, \tag*{(4)}
Υm=Fmexp⁡(j2αmb)−Fm−1exp⁡(j2αmb)−1,(5)\Upsilon_m=\frac{F_m\exp(j2\alpha_m b)-F_m^{-1}}{\exp(j2\alpha_m b)-1}, \tag*{(5)}

it is readily to show that the eigensolutions kk are given by

ΥNIMΥPIM−exp⁡(−2βa)=0.(6)\Upsilon_{\mathrm{NIM}}\Upsilon_{\mathrm{PIM}}-\exp(-2\beta a)=0. \tag*{(6)}

By substituting the eigensolutions kk back into Eq. (1) we can calculate the distributions of fields. Formula for transverse-magnetic (TM) polarized eigenmodes can be developed similarly.

Equation (6) can be numerically solved, for example, by calculating the left-hand side of it at different kk values to find zeroes. A standard result of kk versus aa is shown in Figure 2(a). We can see when the WG distance aa is large enough so that the two WGs are weakly coupled, two separated eigenmodes can be achieved. Each eigenmode represents a localized mode in a single WG. From the distributions of field EyE_y and Poynting vector SzS_z shown in Figure 2(b) and Figure 2(d) we can conclude that the eigenmode at

Variation of $k$ versus $a$ and distributions of field $E_y$ and Poynting vector $S_z$

Figure 2. (a) Variation of kk versus aa. (b) and (d) are the distributions of field EyE_y and Poynting vector SzS_z at the two dispersion points in (a) where a=3a = 3 cm. (c) shows the results in a PIM-PIM structure. Note that Im⁡(Ey)\operatorname{Im}(E_y) is zero in (b) and (d), and the y-coordinate of Im⁡(k)\operatorname{Im}(k) in (a) is shown at the right side, which is properly chosen so that the origins of the curves Im⁡(k)\operatorname{Im}(k) overlap with EP.

kNIM=1.328 cm−1k_{\mathrm{NIM}} = 1.328\ \mathrm{cm}^{-1} is supported by the NIM WG, where the direction of energy flux is opposite to that of kk. The other one at kPIM=1.186 cm−1k_{\mathrm{PIM}} = 1.186\ \mathrm{cm}^{-1} is localized at the PIM WG. From the number of nodes (Ey=0E_y = 0) inside each WG we can see the field supported by the planar PIM WG is a TE2_2 mode, and that in the NIM WG is a TE1_1 one [30]. Note that the wavevector kk is real here, and we have assumed that the expansion coefficient E1E_1 in Eq. (1) is real. As a result, the imaginary component Im⁡(Ey)\operatorname{Im}(E_y) is zero and is not shown in Figure 2(b) and Figure 2(d). The coefficient exp⁡(−jkz+jωt)\exp(-\mathrm{j}kz+\mathrm{j}\omega t) in Eq. (1) is neglected when plotting the distribution of fields.

Figure 2(a) displays a unique phenomenon that is absent in ordinary dielectric WGs. When aa decreases, the modes in the two WGs can couple together to form hybrid eigenmodes. However, here the kk values of the eigenmodes approach to each other. When aa is smaller than a critical distance aca_c, the eigenmodes coalesce. Below aca_c only complex kk exists (brown line and dots in Figure 2(a)), which can be found by searching for the solutions of Eq. (6) in the space formed by Re⁡(k)\operatorname{Re}(k) and Im⁡(k)\operatorname{Im}(k). This phenomenon is in sharp contrast with the ordinary belief about coupled PIM-PIM WGs, that when aa decreases, the coupling between the two WGs would become stronger and the split in kk should increase rather than decrease. To illustrate this difference, we show an example about kk versus aa in a PIM-PIM configuration in Figure 2(c), where NIM in the structure of Figure 1 is replaced by a conjugate PIM with ϵPIM=3\epsilon_{\mathrm{PIM}} = 3 and μPIM=0.556\mu_{\mathrm{PIM}} = 0.556. We can see in this PIM-PIM configuration all the solutions of kk are real, and the two branches of kk do not coalesce.

The coalescence of dispersion at a critical value of aca_c shown in Figure 2 resembles EPs in PT\mathcal{PT} symmetry very much [12, 13, 14, 15, 16, 17, 18, 19, 20]. We check the eigenmodes around this point (see Figure 3), and find that when approaching this point from a>aca > a_c, the field distributions of the two eigenmodes become more and more similar with each other. So at this critical point not only the eigensolutions kk but also the associated eigenvectors coalesce simultaneously. This point is not a diabolic point in Hermitian system, but a standard EP [16, 17, 18]. The region of a>aca > a_c (a<aca < a_c), where the eigensolutions kk are real (complex), possesses an exact (broken) non-Hermitian phase.

Distributions of fields $E_y$ at different points in the $k$ versus $a$ curves

Figure 3. Distributions of fields EyE_y at some different points in the kk versus aa curves. They confirm that the transition point at aca_c is an EP. Note that Im⁡(Ey)\operatorname{Im}(E_y) is no longer zero when a<aca < a_c where Im⁡(k)≠0\operatorname{Im}(k) \ne0.

Figure 3 also displays the distributions of fields when a<aca < a_c. In this phase-broken region, at each given aa value we can generally find two complex solutions of kk, which are conjugate to each other. The complex solutions of kk render complex values of β\beta and αm\alpha_m given by Eqs. (2) and (3). Consequently, Im⁡(Ey)\operatorname{Im}(E_y) is no longer zero (see the insets of Figure 3).

To check how the curves of kk versus aa shown in Figure 2(a) vary when parameters of the coupled WGs change, we repeat the calculations at different ϵPIM\epsilon_{\mathrm{PIM}} values. It would change the kPIMk_{\mathrm{PIM}} value of the PIM WG. One can also achieve the same effect by changing the thickness of it. As for the NIM WG, we would keep all the parameters constant so that kNIMk_{\mathrm{NIM}} does not vary in this article.

Figure 4 displays the results when ϵPIM\epsilon_{\mathrm{PIM}} increases linearly from 4.3 to 4.6. The increased ϵPIM\epsilon_{\mathrm{PIM}} would push the kPIMk_{\mathrm{PIM}} value in the PIM WG close to or even pass kNIMk_{\mathrm{NIM}} in the NIM WG. From Figure 4 we can see with ϵPIM\epsilon_{\mathrm{PIM}} increasing, the loop of kk versus aa would shrink, and eventually disappear around ϵPIM=4.5\epsilon_{\mathrm{PIM}} = 4.5. When ϵPIM\epsilon_{\mathrm{PIM}} further increases from 4.5, the curve appears again and becomes larger. Since kNIM=1.328 cm−1k_{\mathrm{NIM}} = 1.328\ \mathrm{cm}^{-1} does not change, the scenario of ϵPIM=4.5\epsilon_{\mathrm{PIM}} = 4.5 in fact is around the critical point where the degeneracy of kPIM=kNIMk_{\mathrm{PIM}} = k_{\mathrm{NIM}} takes place.

Variation of $k$ versus $a$ at different $\epsilon_{\mathrm{PIM}}$ values

Figure 4. Variation of kk versus aa at different ϵPIM\epsilon_{\mathrm{PIM}} values, from (a) ϵPIM=4.3\epsilon_{\mathrm{PIM}} = 4.3 to (d) ϵPIM=4.6\epsilon_{\mathrm{PIM}} = 4.6.

A non-Hermitian coupled-mode theory

In the above simulation all the media in the coupled WGs are loss-free. Nevertheless, the coalescence of eigenmodes at a critical value of aca_c hints that this loss-free system can possess non-Hermitian phenomenon. Here we could utilize a non-Hermitian coupled-mode theory [21] in the form of MΨ=k±ΨM\boldsymbol{\Psi}=k_{\pm}\boldsymbol{\Psi} to explain the results, which is

\begin{bmatrix} k_{\mathrm{NIM}} & -\gamma\\ \gamma& k_{\mathrm{PIM}} \end{bmatrix} \begin{bmatrix} \psi_{\mathrm{NIM}} \\ \psi_{\mathrm{PIM}} \end{bmatrix} = k_{\pm} \begin{bmatrix} \psi_{\mathrm{NIM}} \\ \psi_{\mathrm{PIM}} \end{bmatrix}. \tag*{(7)}\end{bmatrix}

Here kNIMk_{\mathrm{NIM}} and kPIMk_{\mathrm{PIM}} are the wavevectors of guided modes in separated NIM and PIM WGs, respectively, at the given angular frequency ω\omega. Both kNIMk_{\mathrm{NIM}} and kPIMk_{\mathrm{PIM}} are real. Parameter γ\gamma is also real and characterizes the mutual coupling between the two WGs. The basic vectors ψNIM\psi_{\mathrm{NIM}} and ψPIM\psi_{\mathrm{PIM}} represent field components of the guided modes in characterizing the eigenvectors. Evidently, Eq. (7) is non-Hermitian because the off-diagonal elements are not conjugates, i.e. M12≠M21∗M_{12}\ne M_{21}^{*}. The choice of M21=−M12=γM_{21}=-M_{12}=\gamma can be explained by simultaneously considering two aspects. One is that the total energy should be conserved in this loss-free system [28, 32]. The other is that the propagating directions of energy in the NIM and PIM WGs are opposite to each other, so the coupling between them is contradirectional [28, 32].

To see whether Eq. (7) can explain the main features in our former analysis, we can first solve it. Defining

kNIM=k0+Δ,(8)k_{\mathrm{NIM}}=k_0+\Delta, \tag*{(8)}
kPIM=k0−Δ,(9)k_{\mathrm{PIM}}=k_0-\Delta, \tag*{(9)}

where k0k_0 is the mean wavevector, and Δ\Delta is the detuning, the eigen-solutions k±k_{\pm} are

k±=k0±Δ2−γ2.(10)k_{\pm}=k_0\pm\sqrt{\Delta^2-\gamma^2}. \tag*{(10)}

By using Eq. (10) we can fit the curves kk versus aa given by Eq. (6) and find how the magnitude of γ\gamma varies with the WG distance aa. Results of εPIM=4\varepsilon_{\mathrm{PIM}}=4 is shown in Figure 5. When performing the best fitting, we firstly choose proper values of kNIMk_{\mathrm{NIM}} and kPIMk_{\mathrm{PIM}}, which are 1.328 cm−11.328\ \mathrm{cm}^{-1} and 1.186 cm−11.186\ \mathrm{cm}^{-1}, respectively, for the results shown in Figure 5. The values of kNIMk_{\mathrm{NIM}} and kPIMk_{\mathrm{PIM}}, as well as those of k0=1.257 cm−1k_0=1.257\ \mathrm{cm}^{-1} and Δ=0.071 cm−1\Delta=0.071\ \mathrm{cm}^{-1}, are kept as constants. Then, based on the k±k_{\pm} values obtained from Eq. (6) we can use Eq. (10) to find the values of γ\gamma. Because the two solutions of Eq. (6) might give different values of γ\gamma, we would only keep the averaged one, which is then substituted back into Eq. (10) to test the discrepancy from Eq. (6). Similar fitting process is also performed in the phase-broken region of complex kk values.

Variations of $k$ and $\gamma$ versus the WG distance $a$

Figure 5. Variations of (a) kk and (b) γ\gamma versus the WG distance aa by using Eq. (10) to fit the results from Eq. (6). Green line in (b) is an exponential fit of γ\gamma using Eq. (11).

From Figure 5(a) we can see the prediction from Eq. (10) fits well with those of Eq. (6). The deviation between Re⁡(k)\operatorname{Re}(k) in the phase-broken region might be fixed by assuming a spatial dependence of kNIMk_{\mathrm{NIM}} and kPIMk_{\mathrm{PIM}} on aa, which would not be discussed here. It is interesting to emphasize that the magnitude of γ\gamma exponentially decreases with increasing aa, and can be approximately expressed by

γ=γ0exp⁡(−a/Lc)(11)\gamma=\gamma_0\exp(-a/L_c) \tag*{(11)}

where γ0=0.18 cm−1\gamma_0=0.18\ \mathrm{cm}^{-1}. The decay length LcL_c equals 1.45 cm1.45\ \mathrm{cm}, very close to the value of 1.438 cm1.438\ \mathrm{cm} given by β−1=(k02−ω2/c2)−1/2\beta^{-1}=(k_0^2-\omega^2/c^2)^{-1/2}. Evidently, when aa is smaller than aca_c, the mutual coupling strength ∣γ∣|\gamma| is stronger than ∣Δ∣|\Delta|, then the phase of the coupled WGs is spontaneously broken and gives complex kk values. When aa is larger than aca_c, ∣γ∣|\gamma| becomes smaller than ∣Δ∣|\Delta|, and the phase is conserved. Position aca_c is the critical phase transition point of EP where ∣γ∣=∣Δ∣|\gamma|=|\Delta|. When the two WGs are far away from each other, γ\gamma is zero, and k±=kNIM,PIMk_{\pm}=k_{\mathrm{NIM,PIM}}. Now the eigenvectors are Ψ=[1,0]T\boldsymbol{\Psi}=[1,0]^{\mathrm{T}} and Ψ=[0,1]T\boldsymbol{\Psi}=[0,1]^{\mathrm{T}}, and the fields are localized in separated WGs.

Furthermore, the detuning Δ\Delta determines the size of the loop in the kk versus aa spectrum. When kNIM=kPIMk_{\mathrm{NIM}}=k_{\mathrm{PIM}}, i.e. the eigen-solutions in the NIM and PIM WGs are degenerated, the phase is always spontaneously broken because now Δ=0\Delta=0 and any nonzero γ\gamma would give complex k±k_{\pm} values. It is the case shown in Figure 4(c) at εPIM=4.5\varepsilon_{\mathrm{PIM}}=4.5. Away from this degeneracy scenario, ∣Δ∣|\Delta| increases and the loop becomes more and more great, which can be observed in other plots in Figure 4.

Now, we can pay attention to the EP where

Δ=±γ(12)\Delta=\pm\gamma \tag*{(12)}

that gives k±=k0k_{\pm}=k_0. The eigenfunction of the eigenmode is

[ψNIMψPIM]EP=[1sign⁡(Δ/γ)].(13)\left[\begin{array}{c} \psi_{\mathrm{NIM}} \\ \psi_{\mathrm{PIM}} \end{array}\right]_{\mathrm{EP}} = \left[\begin{array}{c} 1 \\ \operatorname{sign}(\Delta/\gamma) \end{array}\right]. \tag*{(13)}

where the function sign⁡(x)\operatorname{sign}(x) equals 1 (−1-1) when x>0x>0 (x<0x<0). In other words, at this EP the fields in the two WGs are in-phase (a phase difference of zero) or out-phase (a phase difference of π\pi) with each other. The exact phase difference is determined by the sign of Δ/γ\Delta/\gamma. Generally γ\gamma is determined by the overlap integral of fields in the gap between the two WGs and its sign is fixed [14, 15, 28], but sign⁡(Δ/γ)\operatorname{sign}(\Delta/\gamma) can be tuned by changing the resonant conditions of kNIMk_{\mathrm{NIM}}, kPIMk_{\mathrm{PIM}} in the two WGs, e.g. by changing the index of refraction inside or the thicknesses of WGs. As a consequence, the eigenvector at EP is tunable.

Note that here we could not make any comment on the amplitudes of fields in the two WGs because the refractive indexes in them are not required to be equal. It is in sharp contrast with PT\mathcal{PT} symmetric WGs that generally only symmetric configurations are considered [10, 11, 12]. This drawback hinders us to reveal more unique features about EP. However, the phase-jump, as a signature of sign⁡(Δ/γ)\operatorname{sign}(\Delta/\gamma) at EP, is an observable, e.g. by choosing two sets of ϵPIM\epsilon_{\mathrm{PIM}} values so that in one case Δ>0\Delta>0 while in the other case Δ<0\Delta<0, and then analyzing the field patterns inside the structure. The results about the distributions of EyE_y at EPs when ϵPIM=4.4(Δ>0)\epsilon_{\mathrm{PIM}}=4.4(\Delta>0) and ϵPIM=4.6(Δ<0)\epsilon_{\mathrm{PIM}}=4.6(\Delta<0) are shown in Figure 6. From the two plots we can see the fields in the NIM WG are almost identical, but the fields in the PIM WG are flipped with respect to each other. It is then evident that the phase between the two WGs are shifted by π\pi in the two scenarios. Furthermore, associated with the π\pi-phase jump there exists a node inside the gap between the two WGs (red arrow in Figure 6(b)), which implies that in this scenario ψNIM=−ψPIM\psi_{\mathrm{NIM}}=-\psi_{\mathrm{PIM}}. From the fact of Δ<0\Delta<0 in Figure 6(b) we can also conclude that γ>0\gamma>0.

Distributions of fields $E_y$ at EPs

Figure 6. Distributions of fields EyE_y at EPs when (a) ϵPIM=4.4(Δ>0)\epsilon_{\mathrm{PIM}}=4.4(\Delta>0), and (b) ϵPIM=4.6(Δ<0)\epsilon_{\mathrm{PIM}}=4.6(\Delta<0), respectively. Red arrow represents the node in (b), which is associated with the π\pi-phase jump at EP when Δ\Delta varies its sign.

The group velocities vgv_g of the guided modes can also be analyzed. However, since here the angular frequency ω\omega is kept constant, vgv_g cannot be found from ∂ω/∂k\partial\omega/\partial k. Here we calculate vgv_g by using the Poynting vector Sz\mathbf{S}_z and the energy density WW via vg=Sz/Wv_g=S_z/W. When calculating the energy density WW we have adapted the formula of ϵ0∂(ϵNIMω)/∂ω∣E∣2+μ0∂(μNIMω)/∂ω∣H∣2\epsilon_0\partial(\epsilon_{\mathrm{NIM}}\omega)/\partial\omega|E|^2+\mu_0\partial(\mu_{\mathrm{NIM}}\omega)/\partial\omega|H|^2 [21, 22]. At 5 GHz, the utilized dispersion gives ∂(ϵNIMω)/∂ω=5\partial(\epsilon_{\mathrm{NIM}}\omega)/\partial\omega=5 and ∂(μNIMω)/∂ω=4.975\partial(\mu_{\mathrm{NIM}}\omega)/\partial\omega=4.975.

The curves of vgv_g versus aa at ϵPIM=4.0\epsilon_{\mathrm{PIM}}=4.0 are shown in Figure 7. As expected, the values of vgv_g are limited in the region defined by vg=−0.2cv_g=-0.2c of NIM (it is negative because the energy flux is backward propagating) and vg=0.5cv_g=0.5c of PIM, where cc is the speed of light. When the field is mainly localized in the NIM WG, vgv_g is negative. Otherwise vgv_g is positive. As aa decreases, the split between kk also decreases, and vgv_g of the two branches approach to each other. At EP the group velocity is zero, which can be explained by the balanced positive energy flux in PIM (including the surrounded air) and negative energy flux in NIM, respectively.

Group velocity $v_g$ versus $a$

Figure 7. Group velocity vgv_g versus aa when ϵPIM=4.0\epsilon_{\mathrm{PIM}}=4.0. At EP vgv_g is zero.

Note that the stopped light at EP demonstrated here is different from that in PTPT symmetric WGs [12, 13, 14]. Here the stopped light is associated with the negative flux in NIM, and the whole structure is still Hermitian because no loss or gain is presented. On the contrary, the stopped light in PTPT-symmetric WGs with spatial distributions of gain and loss cannot be explained by the propagating of energy flux associated with real-valued fields [14], but is the consequence of zero cc-product [12, 33] of the non-Hermitian system. The cc-product can be applied when the two WGs are not only geometric but also electromagnetic (except for the loss/gain effect) identical so that ψ\psi in the eigenvector can be unambiguously defined [12, 33]. Here the two WGs are made of different media, possess different indexes of refraction, and can have different thicknesses, so the applicability of cc-product needs further discussion.

Discussion

Above we have shown that non-Hermitian optical effects can be observed in loss-free coupled WGs made of NIMs. Here we would like to make further comments on the category of the non-Hermitian effect. From Eq. (7) we can see it can be classified into the anti-PTPT symmetric one because the diagonal elements of the 2×22 \times2 matrix MM satisfy Im⁡(M11)=Im⁡(M22)\operatorname{Im}(M_{11})=\operatorname{Im}(M_{22}) and M12=−M21∗M_{12}=-M_{21}^{*}. It should be emphasized that when utilizing Eq. (7) we have assumed that the basic vectors ψNIM\psi_{\mathrm{NIM}} and ψPIM\psi_{\mathrm{PIM}} represent field components of the guided modes that characterize the wave functions. The set of analogues of the basic vectors ψNIM=EyNIM\psi_{\mathrm{NIM}}=E_y^{\mathrm{NIM}} and ψPIM=EyPIM\psi_{\mathrm{PIM}}=E_y^{\mathrm{PIM}} is adopted in our analysis. However, since the two WGs are not identical with each other, the choice of the basis vectors is arbitrary. For example, we can also utilize another set of analogues by adding an additional phase of π/2\pi/2 to the basic vector ψPIM\psi_{\mathrm{PIM}}. Now, Eq. (7) becomes

\begin{pmatrix} k_{\mathrm{NIM}} & j\gamma\\ j\gamma& k_{\mathrm{PIM}} \end{pmatrix} \begin{pmatrix} \psi_{\mathrm{NIM}} \\ j\psi_{\mathrm{PIM}} \end{pmatrix} = k_{\pm} \begin{pmatrix} \psi_{\mathrm{NIM}} \\ j\psi_{\mathrm{PIM}} \end{pmatrix}. \tag*{(14)}\end{pmatrix}

Equation (14) is just the standard anti-PTPT symmetric operator that has been discussed in many literatures [8, 9, 10, 11]. It can also explain all the features on the dispersion and distributions of fields. As for which one is preferred in studying NIM, we still prefer Eq. (7) because the node shown in Figure 6(b) can be more intuitively explained.

The importance of this article is that it proves we can access non-Hermitian optics without using gain or loss. It also shows that the negative energy flux in NIM has many realistic impacts on future applications, especially by considering the fact that all the media are loss-free. Experimental investigation of the non-Hermitian effect discussed here can utilize documented NIM design [26, 27], and transfer to other frequency regimes by simply rescaling the geometric parameters of the coupled-WG configuration. Theoretical interest can be paid to discuss the deep-lying physics of the spontaneous phase broken, and the utilization of the anti-PTPT symmetry in loss-free NIM to achieve higher-order EPs and mode switching purpose.

Before ending this article, we would like to emphasize again that the whole structure studied here is loss- and gain-free. Consequently, in principle it is a closed Hermitian system, and the energy should be conserved. The non-Hermitian effects discussed above, especially the existence of complex kk below aca_c, should not violate the principle of energy conservation. These complex-kk eigenmodes might be associated with some effects similar to the photonic bandgap effect and waveguiding effect below cutoff, which forbid the propagation of optical field without absorbing any energy [21, 28]. Detailed discussion deserves our further efforts.

Conclusion

In summary, in this article we check the optical waves in coupled WGs made of lossless NIM and PIM. We show that the guided modes in this kind of gain/loss-free optical system can also support non-Hermitian features. A simple non-Hermitian coupled-mode theory is utilized to explain our results. This theory proves that the critical degeneracy point at aca_c is an EP, and the field dynamics induced by NIM belongs to anti-PT\mathcal{PT} symmetry. Features at EPs including the slow-light effect are discussed. This study highlights the non-negligible novelties of NIM, which can be utilized in studying non-Hermitian optics to bypass many restrictions of PT\mathcal{PT} symmetry.

Funding

Natural National Science Foundation of China (NSFC) (12104227, 12274241); Scientific Research Foundation of Nanjing Institute of Technology (YKJ202021).

Disclosures

The authors declare no conflicts of interest.

Data availability

Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.

References

  1. [1]C. M. Bender, ”Making sense of non-Hermitian Hamiltonians,” Rep. Prog. Phys. 70(6), 947–1018 (2007).
  2. [2]S. Longhi, ”𝒫𝒯-symmetric laser absorber,” Phys. Rev. A 82(3), 031801(R) (2010).DOI
  3. [3]A. A. Zyablovsky, A. P. Vinogradov, A. V. Dorofeenko, A. A. Pukhov, and A. A. Lisyansky, ”Causality and phase transitions in 𝒫𝒯-symmetric optical systems,” Phys. Rev. A 89(3), 033808 (2014).DOI
  4. [4]L. Feng, R. El-Ganainy, and L. Ge, ”Non-Hermitian photonics based on parity-time symmetry,” Nat. Photo. 11(12), 752–762 (2017).DOI
  5. [5]R. El-Ganainy, K. G. Makris, M. Khajavikhan, Z. H. Musslimani, S. Rotter, and D. N. Christodoulides, ”Non-Hermitian physics and 𝒫𝒯 symmetry,” Nat. Phys. 14(1), 11–19 (2018).
  6. [6]J. Gear, R. Liu, S. T. Chu, S. Rotter, and J. Li, ”Parity-time symmetry from stacking purely dielectric and magnetic slabs,” Phys. Rev. A 91(3), 033825 (2015).DOI
  7. [7]M. Lawrence, N. Xu, X. Zhang, L. Cong, J. Han, W. Zhang, and S. Zhang, ”Manifestation of 𝒫𝒯 symmetry breaking in polarization space with terahertz metasurfaces,” Phys. Rev. Lett. 113(9), 093901 (2014).pubmed.ncbi.nlm.nih.gov/25215984
  8. [8]L. Ge and H. E. Tureci, ”Antisymmetric 𝒫𝒯-photonic structures with balanced positive- and negative-index materials,” Phys. Rev. A 88(5), 053810 (2013).DOI
  9. [9]Y. Choi, C. Hahn, J. W. Yoon, and S. H. Song, ”Observation of an anti-𝒫𝒯-symmetric exceptional point and energy-difference conserving dynamics in electrical circuit resonators,” Nat. Comm. 9, 2182–2187 (2018).DOI
  10. [10]Q. Li, C. J. Zhang, Z. D. Cheng, W. Z. Liu, J. F. Wang, F. F. Yan, Z. H. Lin, Y. Xiao, K. Sun, Y. T. Wang, J. S. Tang, J. S. Xu, C. F. Li, and G. C. Guo, ”Experimental simulation of anti-parity-time symmetric Lorentz dynamics,” Optica 6(1), 67–71 (2019).DOI
  11. [11]F. Yang, Y. C. Liu, and L. You, ”Anti-𝒫𝒯 symmetry in dissipatively coupled optical systems,” Phys. Rev. A 96(5), 053845 (2017).DOI
  12. [12]T. Goldzak, A. A. Mailybaev, and N. Moiseyev, ”Light stops at exceptional points,” Phys. Rev. Lett. 120(1), 013901 (2018).DOI
  13. [13]S. A. H. Gangaraj and F. Monticone, ”Topological waveguiding near an exceptional point: defect-immune, slow-light, and loss-immune propagation,” Phys. Rev. Lett. 121(9), 093901 (2018).DOI
  14. [14]X. Z. Zhang, R. Z. Luo, and J. Chen, ”Revisit the Poynting vector in 𝒫𝒯-symmetric coupled waveguides,” Opt. Express 30(21), 38753–38763 (2022).DOI
  15. [15]L. S. Sun, B. Zhao, J. Q. Yuan, and J. Chen, ”Kerr-nonlinearity induced bistable-like parity-time phase transition in coupled waveguides,” Opt. Express 29(5), 7935–7947 (2021).DOI
  16. [16]H. Cao and J. Wiersig, ”Dielectric microcavities: model systems for wave chaos and non-Hermitian physics,” Rev. Mod. Phys. 87(1), 61–111 (2015).DOI
  17. [17]S. K. Ozdemir, S. Rotter, F. Nori, and L. Yang, ”Parity-time symmetry and exceptional points in photonics,” Nat. Mater. 18(8), 783–798 (2019).DOI
  18. [18]M. A. Miri and A. Alù, ”Exceptional points in optics and photonics,” Science 363(6422), eaar7709 (2019).DOI
  19. [19]J. B. Khurgin, Y. Sebbag, E. Edrei, R. Zektzer, K. Shastri, U. Levy, and F. Monticone, ”Emulating exceptional-point encirclements using imperfect (leaky) photonic components: asymmetric mode-switching and omni-polarizer action,” Optica 8(4), 563–569 (2021).DOI
  20. [20]H. Z. Chen, T. Liu, H. Y. Luan, R. J. Liu, X. Y. Wang, X. F. Zhu, Y. B. Li, Z. M. Gu, S. J. Liang, H. Gao, L. Le, G. Se, Z. Zhang, J. Zhu, and R. M. Ma, ”Revealing the missing dimension at an exceptional point,” Nat. Phys. 16(5), 571–578 (2020).DOI
  21. [21]J. D. Jackson, Classical Electrodynamics (John Wiley & Sons, 1998).
  22. [22]L. D. Landau and E. M. Lifshitz, Electrodynamics of Continuous Media (Pergamon, 1984).
  23. [23]V. G. Veselago, ”The electrodynamics of substance with simultaneously negative value of ε and μ,” Sov. Phys. Usp. 10, 509–514 (1968).
  24. [24]R. A. Shelby, D. R. Smith, and S. Schultz, ”Experimental verification of a negative index of refraction,” Science 292(5514), 77–79 (2001).pubmed.ncbi.nlm.nih.gov/11292865
  25. [25]J. B. Pendry, ”Negative refraction makes a perfect lens,” Phys. Rev. Lett. 85(18), 3966–3969 (2000).pubmed.ncbi.nlm.nih.gov/11041972
  26. [26]N. I. Zheludev and Y. S. Kivshar, ”From metamaterials to metadevices,” Nat. Mater. 11(11), 917–924 (2012).DOI
  27. [27]Y. M. Liu and X. Zhang, ”Metamaterials: a new frontier of science and technology,” Chemical Society Reviews 40(5), 2494–2507 (2011).DOI
  28. [28]T. Mealy and F. Capolino, ”Exceptional points of degeneracy with indirect band gap induced by mixing forward and backward propagating waves,” Phys. Rev. A 107(1), 012214 (2023).DOI
  29. [29]A. C. Peacock and N. G. R. Broderick, ”Guided modes in channel waveguides with a negative index of refraction,” Opt. Express 11(20), 2502–2510 (2003).pubmed.ncbi.nlm.nih.gov/19471363
  30. [30]I. V. Shadrivov, A. A. Sukhorukov, and Y. S. Kivshar, ”Guided modes in negative-refractive-index waveguides,” Phys. Rev. E 67(5), 057602 (2003).pubmed.ncbi.nlm.nih.gov/12786330
  31. [31]A. V. Novitsky and L. M. Barkovsky, ”Guided modes in negative-refractive-index fibres,” J. Opt. A: Pure Appl. Opt. 7(2), S51–S56 (2005).DOI
  32. [32]A. Yariv, ”Coupled-mode theory for guided-wave optics,” IEEE J. Quantum Electron. 9(9), 919–933 (1973).DOI
  33. [33]N. Moiseyev, Non-Hermitian Quantum Mechanics (Cambridge, 2011).

Paper details

vlnc:W_5ep7rcp3
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

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

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents