## Abstract

We present comprehensive case studies on trapping of light in plasmonic waveguides, including the metal-insulator-metal (MIM) and insulator-metal-insulator (IMI) waveguides. Due to the geometrical symmetry, the guided modes are classified into the anti-symmetric and symmetric modes. For the lossless case, where the relative electric permittivity of metal (*ε _{m}*) and dielectric (

*ε*) are purely real, we define

_{d}*ρ*as

*ρ*= -

*ε*/

_{m}*ε*. It is shown that trapping of light occurs in the following cases: the anti-symmetric mode in the MIM waveguide with 1 <

_{d}*ρ*< 1.28, the symmetric mode in the MIM waveguide with

*ρ*≪ 1, and the symmetric mode in the IMI waveguide with

*ρ*< 1. The physical interpretation reveals that these conditions are closely connected with the field distributions in the core and the cladding. Various mode properties such as the number of supported modes and the core width for the mode cut off are also presented.

©2010 Optical Society of America

## 1. Introduction

Slow light is light with a low group velocity, conventionally slower than one-hundredth of the velocity of light in free space [1]. Recently, there has been growing interest in achieving and utilizing slow light due to its various applications. Optical pulse trains can be delayed by a few bits when they are transmitted through slow light waveguides, allowing for optical buffers [2]. Compression of energy density accompanied with slow light can be used to enhance nonlinear properties of materials with a small feature size [3]. In addition, the large group index of slow light usually leads to the large effective refractive index, which is a very promising characteristic in optical switching [4].

The extreme case of slow light is trapping, which indicates that the group velocity of light vanishes. This phenomenon is also referred to as pinning, stopping, or freezing of light. Methods to realize trapping of light in waveguides can be classified into two groups: one is to
use photonic crystal waveguides with resonators and the other is to invoke metamaterial waveguides. The former is associated with multiple reflections of light inside resonators artificially employed in the photonic crystal waveguide, and it is known to have strong points from view points of Q factor and implementation [5–10]. The photon pinning in an optical resonator with the ultrahigh-Q factor of ~5 × 10^{9} was achieved by means of local index tuning in the photonic crystal waveguide [5]. The dynamic release of trapped light was experimentally demonstrated by using adiabatic frequency shift [6].

Research on the usage of metamaterial waveguides for trapping light has been initiated by trailblazing work done by Tsakmakidis *et al* [11]. The key principle is based on the fact that the direction of energy is directly opposite to that of the phase in a negative index material. By adopting positive- and negative-index materials for the core and the cladding, one can obtain a balance point where the optical power flow in the cladding is exactly compensated by that in the core, resulting in trapping of light. In addition, it was reported that trapping is related to the mode degeneracy [12]. Two distinct modes with different propagation properties can be supported simultaneously in metamaterial waveguides i.e., one has the parallel phase and group velocities, whereas the other exhibits an anti-parallel relation. If these modes degenerate into one mode, its group velocity vanishes. Fu and Gan reported various configurations for trapping of light in the THz [13, 14] and telecom [15] regimes. Kim devised a thin cladding composed of a negative index material for temporary trapping of light
[16].

Although considerable research has been devoted to the realization of the negative refractive index materials in the optical frequency regime [17–19], it still remains difficult to implement them. Meanwhile, there has been much interest in the plasmonic waveguide, which consists of only the metal and dielectric and does not require any metamaterial [20–29]. It turned out that the plasmonic waveguide can support a propagating mode with the anti-parallel phase and group velocities [20–22]. The properties of guided modes in the planar plasmonic waveguides were investigated in detail and the existence of solutions with negative group velocities was observed [20]. Dionne and her coworkers also examined the modal dispersion of the metal-insulator-metal (MIM) and insulator-metal-insulator (IMI) waveguides with the relative electric permittivity of the metal that is empirically obtained [21, 23]. In particular, they provided an experimental demonstration for the negative group velocity in the MIM waveguide [24]. However, none of these studies have reported whether light can be trapped inside the plasmonic waveguide. Although Davoyan and his coworkers recently observed that the mode degeneracy can occur in the MIM waveguide under a certain condition, they did not offer physical origin of such degeneracy and the exact condition for its existence [30].

In this study, we seek answers to the following two questions: 1) Is it possible for light to be trapped in the plasmonic waveguide? 2) If possible, what is the condition for that? Full case studies reveal that trapping of light can occur in the plasmonic waveguides under certain conditions. The physical interpretation on the origin of these conditions is provided. In addition, the conditions for the mode to be supported in the plasmonic waveguides are given in detail, which can be used for various design processes of the plasmonic waveguides.

This paper is organized as follows. The basic terms, definitions, conventions are presented in Section 2. Then, the existence of modes and condition for the mode degeneracy in the MIM waveguide are provided in Section 3. The discussion branches into two classes: the anti-symmetric mode in the MIM waveguide (Subsection 3.1) and the symmetric mode in the MIM waveguide (Subsection 3.2). The results of discussion for the anti-symmetric and symmetric modes in the MIM waveguide are provided in Table 2 and Table 3, respectively. The next section offers the case study for the IMI waveguide. Similarly, the discussion branches into two classes: the symmetric mode in the IMI waveguide (Subsection 4.1 and Table 4) and the anti-symmetric mode in the IMI waveguide (Subsection 4.2 and Table 5). Finally, the conclusion is presented.

## 2. Propagation constant, phase and group velocities in the plasmonic waveguide

Let λ, ω, and *k*
_{0} denote the wavelength, the angular frequency, and the wavenumber in free space, respectively. The relative electric permittivity of the metal and dielectric are denoted by *ε _{m}* and

*ε*, respectively. The relative magnetic permeability is assumed to be unity in all material. The dependence of

_{d}*ε*on ω is governed by the Drude model:

_{m}where *ω _{P}* is the bulk plasma frequency. In this paper, we only consider the lossless case i.e., γ = 0 and

*ε*is purely real. In addition, we focus on the case that

_{m}*ε*is negative i.e.,

_{m}*ε*= -∣

_{m}*ε*∣ (

_{m}*ω*<

*ω*). Novel metals such as gold or silver satisfy

_{p}*ε*= -∣

_{m}*ε*∣ in the visible and infrared regimes. For convenience, let us define the ratio between

_{m}*ε*and

_{m}*ε*as follows:

_{d}Note that *ρ*>0 and *ρ* decreases as *ω* increases. By combining Eqs. (1) and (2), we know that *ρ*>1 for *ω*<*ω _{SP}* and

*ρ*<1 for

*ω*>

*ω*, where ${\omega}_{SP}={\omega}_{P}/\sqrt{1+{\epsilon}_{d}}$ is the surface plasma frequency. The bulk plasma wavelength

_{SP}*λ*and wavenumber

_{P}*β*are defined by 2

_{P}*πc*

_{0}/

*ω*and

_{p}*ω*/

_{P}*c*

_{0}, respectively, where

*c*

_{0}denotes the speed of light in free space.

It is assumed that the wavefunction has a factor of exp(-*jωt*). The guided modes propagate along Z -direction with the propagation constant *β* i.e., the Z -dependence of the wavefunction is given by exp(*jβZ*). *β* is also referred to as the longitudinal wavenumber. The effective refractive index of a guided mode *n _{eff}* is defined as

*β*/

*k*

_{0}.

*n*of the surface plasmon polariton (SPP) propagating along the single interface between the metal with

_{eff}*ε*and the dielectric with

_{m}*ε*, called the single interface SPP, is given by ${n}_{SPP}=\sqrt{{\epsilon}_{m}{\epsilon}_{d}/\left({\epsilon}_{m}+{\epsilon}_{d}\right)}.$

_{d}The plasmonic waveguides dealt with in this paper are the two-dimensional planar slab waveguides. Depending on configurations, they can be categorized into two classes. One consists of the dielectric core and the metal cladding, called the MIM waveguides [21]. The other is composed of the metal core and the dielectric cladding, referred to as the IMI waveguide [23]. The plasmonic waveguide symmetric across the line along the center of the geometry has the transverse magnetic field with either symmetric or anti-symmetric distribution. In this paper, the former is called the *symmetric mode* and the latter the *anti-symmetric mode*. For convenience the following acronyms are adopted: MIMa, MIMs, IMIa, and IMIs for the anti-symmetric mode in the MIM waveguide, the symmetric mode in the MIM waveguide, the anti-symmetric mode in the IMI waveguide, and the symmetric mode in the IMI waveguide, respectively. In addition, we adopt a convention that the mode with
*n _{eff}* > √

*ε*is referred to as the

_{d}*plasmonic mode*and that with

*n*< √

_{eff}*ε*the

_{d}*photonic mode*[24].

It is noteworthy that the characteristic equation in a waveguide is a function of *β*
^{2} rather than *β*. Consequently, there can be two solutions for *β*. For example, if *β* = 1.2*k*
_{0} is the solution, then -1.2*k*
_{0} is also the solution to the characteristic equation. In other words, a solution *β* always has its twin -*β* and they are located symmetrically with respect to the
origin. After solving a characteristic equation, we have to choose either *β* or -*β* as a
solution. Consideration on this duality is significant in defining signs of the phase and group velocities, which are given by *ω*/*β* and (*dβ*/ *dω*)^{-1}, respectively. The direction of the phase velocity is associated with the sign of the real part of *β*. The phase propagates toward +Z (-T) direction if the real part of *β* has a positive (negative) value. Since causality in electromagnetics requires that the energy flow outward from a source, it is necessary to choose *β* in such a way that the sign of group velocity, which corresponds to the slope in the *ω* - *β* dispersion relation, is positive. Detail discussions on the aforementioned argument can also be found in Refs. [22, 31]. Throughout this paper, the *positive mode* describes a solution with parallel phase and group velocities (both phase and group velocities are positive), whereas the *negative mode* denotes a solution with anti-parallel phase and group velocities (the negative phase velocity and the positive group velocity).

In this paper, the effect of the material loss is not considered. When considering the material loss, we usually employ the concept of the *spatial* loss, which means that the characteristic equation is solved under assumption that *ω* is real and *β* is complex (*β* = *β _{r}* +

*iβ*), where

_{i}*β*and

_{r}*β*denote the real and imaginary parts of β, respectively [22]. Then the wave function has a factor of exp[

_{i}*j*(

*β*-

_{r}Z*ω*)]exp(-

*β*). The dispersion relation is plotted as

_{i}z*ω*versus

*β*and the propagation length is defined as 1/(2/

_{r}*β*). However, dealing with the trapped light, which does not propagate, the concept of the

_{i}*temporal*loss is more suitable than the spatial loss. The temporal loss means that we calculate the characteristic equation under assumption that

*β*is real and

*ω*is complex (

*ω*=

*ωr*+

*iω*), where

_{i}*ω*and

_{r}*ω*denote the real and imaginary parts of

_{i}*ω*, respectively [22]. The wave function has a factor of exp[

*j*(

*βz*-

*ω*)]exp(-

_{r}t*ω*). From this point of view, the propagation length is not defined, Instead, the life time

_{i}*τ*= 1/

*ω*is used to describe the extent to which the wave function decays as time progresses. From the view point of the temporal loss, the trapping of light discussed below is still feasible. The detailed discussion on the effect of the material loss on the trapping of light is under work and will be reported in a next paper.

_{i}## 3. Metal-insulator-metal waveguide

In this section, it is shown that a certain condition exists that the MIM waveguide has two anti-symmetric modes simultaneously and that the optical power flow vanishes at the degenerate point of them. In addition, the symmetric mode also has the mode degeneracy and thus the zero optical power flow takes place in the symmetric mode.

Figure 1 shows the schematic diagram of the MIM waveguide. Four modes are illustrated: the anti-symmetric plasmonic mode, the anti-symmetric photonic mode, the symmetric plasmonic mode, and the symmetric photonic mode. We will derive conditions that each mode is supported in the MIM plasmonic mode. From the boundary condition that tangential electric and magnetic fields are continuous across the interface between the core and the cladding, the characteristic equations are given by [27]

*κ _{m}* and κ

*d*denote the transverse wavenumber of the plasmonic mode in the metal and dielectric, respectively, and

*k*that of the photonic mode in the dielectric. From the momentum conservation relation, we obtain

_{d}Here, it is assumed that *Κ _{m}*,

*Κ*,

_{d}*k*

_{0}, and β are purely real and positive. For convenience, we introduce auxiliary definitions as follows [12, 20]:

In addition, instead of the actual core width 2*a*, the reduced core width *ak*
_{0} will be used. This is because what matters in the existence of a mode is the ratio between the core width and the wavelength. In this section, the wavelength *λ*
_{0} is set to be 400 nm and it is assumed that various values of *ρ* originate from its corresponding bulk plasma frequency *ω _{P}*.

#### 3.1 Anti-symmetric mode

Let us first consider the condition that the anti-symmetric plasmonic mode can propagate in the MIM waveguide, by using Eqs. (3), (7), and (8) with help of Eqs. (10), (11), and (13). Invoking the graphical method, we want to find the condition that two curves of
*W* = **ρU** coth *U* and $W=\sqrt{{U}^{2}+{V}^{2}}$ have an intersection in the (*U*,*W*) space. Considering the
asymptotic behavior of coth function, we know that *ρU* coth *U* asymptotes to *ρU* for *U* → ∞. The limiting value of *ρU* coth*U* for *U* → 0 is given by *ρ*. Meanwhile, $\sqrt{{U}^{2}+{V}^{2}}$ asymptotes to *U* for *U* → ∞ and the limiting value of $\sqrt{{U}^{2}+{V}^{2}}$ for *U* → 0 is simply *V*.

Depending on whether *ρ*>1 or *ρ*<1, the discussion for the existence of the antisymmetric plasmonic mode can be classified into two cases. Figure 2(a) depicts the aforementioned two curves in the first quadrant of *U* and *W* for the case of *ρ* > 1. It is seen that the asymptotes of *W* = *ρU* coth *U* (the red dash-dotted line), which is given by *W* = *ρU* (the black dashed line), grows faster than that of $W=\sqrt{{U}^{2}+{V}^{2}}$ (the blue solid line),
which is given by *W* = *U* (the green dotted line). Therefore, if the *W* -intersect of *W* = *ρU*coth *U*, which is *ρ* (the point A in Fig. 2(a)), is smaller than that of $W=\sqrt{{U}^{2}+{V}^{2}}$, which is given by *V* (the point B in Fig. 2(a)), then it is sure that there exists an intersection between *W* = *ρU*coth*U* and $W=\sqrt{{U}^{2}+{V}^{2}}$. In other words, for the case *ρ*>1 i.e., *ω* < *ω _{SP}*, it is guaranteed that the anti-symmetric plasmonic modes can propagate in the MIM waveguide, provided that

*ρ*<

*V*i.e.,

*ak*

_{0}> ξ, where

It should be pointed out that putting *β* = *k*
_{0}
*ε _{d}* also results in Eq. (14). In the same manner, we can induce the condition that the anti-symmetric plasmonic mode exists for the case

*ρ*< 1. Figure 2(b) shows the two curves for

*ρ*< 1. Let us consider the asymptotic behaviors of two curves. For

*ρ*< 1, if the

*W*-intersect of

*W*=

*ρU*coth

*U*,

*ρ*(the point A in Fig. 2(b)), is larger than that of $W=\sqrt{{U}^{2}+{V}^{2}}$,

*V*(the point B in Fig. 2(b)), then the anti-symmetric plasmonic mode exists. This relation can be expressed in a form of

*ak*

_{0}<ξ.

Next, we look for the condition that the anti-symmetric photonic mode can propagates in the MIM waveguide by employing Eqs. (4), (7), and (9) with help of Eqs. (10), (12), and (13). Two curves of *W* = *pJ* cot *J* and $W=\sqrt{-{J}^{2}+{V}^{2}}$ will be examined. Owing to the periodic property of the cot function, more than one anti-symmetric photonic mode can be supported, provided that *V* is sufficiently large. Here, we focus on the fundamental anti-symmetric mode whose *n _{eff}* is the largest one. Contrary to the domain of definition of

*U*discussed above,

*U*∈ [0,∞],

*J*has the domain of definition of

*J*∈ [0,

*ak*

_{0}√

*ε*]. This is because

_{d}*β*is purely real and thus

*β*

_{2}should be positive (See Eqs. (9) and (12).).

Depending on whether *ρ* > *V* or *ρ* < *V*, our discussion branches off. In Fig. 3(a), we show *W* = *ρJ* cot *J* (the red dash-dotted line) and $W=\sqrt{-{J}^{2}+{V}^{2}}$ (the blue solid line) for *ρ* > *V*. The bold lines illustrate lines in the domain of definition (*J* ∈ [0, *ak*
_{0}√*ε _{d}*]), whereas thin lines denote lines outside the domain of definition. In the case ρ >

*V*, if

*ρJ*cot

*J*at

*J*=

*ak*

_{0}√

*ε*(the point A in Fig. 3(a)) is smaller than $W=\sqrt{-{J}^{2}+{V}^{2}}$ at

_{d}*J*=

*ak*

_{0}√

*ε*(the point B in Fig. 3(a)) i.e., ρ√

_{d}*ε*cot(

_{d}*ak*

_{0}√

*ε*) < √-

_{d}*ε*, it is guaranteed that the anti-symmetric photonic mode exists i.e.,

_{m}*ak*

_{0}> ζ, where

It should be mentioned that Eq. (15) can also be derived by putting *β* = 0 into Eqs. (4), (7), and (9). This property can be found in Table 1 in Ref. [27]. In the same manner, we can obtain the condition that the anti-symmetric photonic mode can propagate in the case of *ρ* > *V* : ρ√cot(*ak*
_{0}√*ε _{d}*) > √-

*ε*i.e.,

_{m}*ak*

_{0}(see Fig. 3(b)). Let us remind that

*ρ*>

*V*and

*ρ*<

*V*conditions correspond to

*ak*

_{0}<

*E*, and

*ak*

_{0}> ξ, respectively. Hence, depending on whether ξ > ζ or ξ < ζ , the condition for the existence of the anti-symmetric photonic mode can be classified into two cases. Meanwhile, the difference between ξ and ζ i.e., ξ - ζ , is a monotonic increasing function of

*ρ*and becomes zero when

*ρ*=

*ρ*, where

_{c}*ρ*is a solution satisfying

_{c}*ρ _{c}* is calculated as about 1.28. Using this property, we know that

*ξ*>

*ζ*and

*ξ*<

*ζ*correspond to

*ρ*>

*ρ*and

_{c}*ρ*<

*ρ*, respectively. Therefore, the condition for the existence of the anti-symmetric photonic mode is given by

_{c}*ξ*<

*ak*

_{0}< ζ for

*ρ*<

*ρ*and

_{c}*ζ*<

*ak*

_{0}<

*ξ*for ρ >

*ρ*. Table 1 summarizes aforementioned property.

_{c}So far, we have derived the range of the reduced core width *ak*
_{0} where anti-symmetric plasmonic and photonic modes exist for given range of *ρ*. Now we are led to the discussion on the mode degeneracy and the zero optical flow of anti-symmetric modes. Let us examine the dependence of the absolute value of the effective refractive index |*n _{eff}*| on the reduced core width

*ak*

_{0}for each case of

*ρ*<1 , 1 <

*ρ*< 1.28 , and

*ρ*> 1.28 , shown in Figs. 4(a)–(c), respectively. Note that |

*n*|, rather than

_{eff}*n*, is used here, owing to the fact that in some cases a negative-valued

_{eff}*n*is chosen. The vertical dash-dotted and dashed lines in Figs. 4(a)–(c) denote the lines of

_{eff}*ak*

_{0}= ξ, and

*ak*

_{0}= ζ, respectively. The dotted horizontal lines in Figs. 4(a)–(c) indicate the lines of |

*n*|=√

_{eff}*ε*. The dash-dotted horizontal lines in Figs. 4(b) and 4(c) correspond to the lines of |

_{d}*n*| =

_{eff}*n*. In Fig. 4(a), it is seen that |

_{spp}*n*| is decreased monotonically as

_{eff}*ak*

_{0}is increased. The anti-symmetric modes is plasmonic for

*ak*

_{0}<

*ξ*. As

*ak*

_{0}grows and gets larger than

*ξ*, the anti-symmetric mode transits from the plasmonic mode into the photonic mode. As

*ak*

_{0}goes from

*ξ*to

*ζ*, |

*n*| decreases monotonically and the anti-symmetric mode vanishes at

_{eff}*ak*

_{0}=

*ζ*. This mode cut off is in good agreement with the results of previous studies. It has been reported that this property can be used to offer the lateral confinement in the MIM waveguide [32] and to implement a barrier acting as a mirror, resulting in a plasmonic resonator in the nanoscale [27].

Figure 4(b) illustrates |*n _{eff}*| as a function of

*ak*

_{0}for 1 <

*ρ*< 1.28. Since

*ak*

_{0}>

*ξ*, allows the plasmonic mode and

*ξ*<

*ak*

_{0}<

*ζ*supports the photonic mode, it can be concluded that the plasmonic and photonic modes exist simultaneously when

*ξ*<

*ak*

_{0}<

*ζ*. It is also noteworthy that |

*n*| of the plasmonic mode is decreased as

_{eff}*ak*

_{0}is reduced in the domain

*ak*

_{0}>

*ξ*, whereas that of the photonic mode is increased with the decrease of

*ak*

_{0}in the domain

*ξ*<

*ak*

_{0}<

*ζ*. These two modes come closer to each other in the domain

*ak*

_{0}<

*ξ*. It is therefore guaranteed that two modes degenerate in one mode at a specific

*ak*

_{0}smaller than ξ . This is one of the most important findings in this paper. It will be shown below that, at this degenerate point, the optical power flow vanishes. Although this kind of curve has been reported in [30], neither the origin nor exact condition was provided.

With respect to the aforementioned observation, it appears that the result in Fig. 2(a) and summary in Table 1 should be supplemented. The region *ρ* > 1 in 2Fig. 2(a) is classified into two regions: one is the region *ρ* > 1.28 and the other is the region 1 < *ρ* < 1.28 . The former results in *ξ* <ζ , whereas the latter gives rise to ξ < *ζ*. The *U* - *W* relationship shown in Fig. 2(a) is for the case *ρ* = 4 > 1.28. Thus we show in Fig. 5(a) the supplementary result for 1 < *ρ* = 1 < 1.28. Note that there are two intersections between two curves *W* = *ρU* coth*U* and *W* = √*U*
^{2} + *V*
^{2}. Since these curves are too close to each other, it is hard to distinguish one from the other. In Fig. 5(b), the dependence of *ρU* coth *U* -√*U*
^{2} + *V*
^{2} as a function of *U* for various values of *ak*
_{0} can be seen. The zeros, i.e., *U* -intersections of this curve, correspond to the intersections of *W* = *ρU* coth*U* and *W* = √*U*
^{2} + *V*
^{2} . If *ak*
_{0} = *ζ*, there is only one zero. Two zeros are observed when *ak*
_{0} = *ξ*. As *ak*
_{0} is decreased from *ξ*, these zeros come close to each
other. At a certain value, denoted by *h _{c}*, the curve

*ρU*coth

*U*- √

*U*

^{2}+

*V*

^{2}has the degenerated zero. This is in good agreement with the result shown in Fig. 4(b). In Table 2, the supplemental result from Table 1 is provided.

Now let us come back to the discussion on the geometric dispersions in Fig. 4.
Figure 4(c) shows the effect of *ak*
_{0} on |*n _{eff}*| for

*ρ*>1.28 . The functional behavior of |

*n*| for this case is well established [20, 21, 27]. When

_{eff}*ak*

_{0}is sufficiently large, |

*n*| asymptotes to

_{eff}*n*, as expected. As

_{spp}*ak*

_{0}is reduced, |

*n*| is decreased monotonically. When

_{eff}*ak*

_{0}gets small and crosses

*ξ*, the mode transits from the plasmonic mode to the photonic mode. As

*ak*

_{0}reaches

*ζ*, |

*n*| asymptotes to zero and the mode is cut off. Neither the simultaneous presence of two plasmonic and photonic modes nor the degeneracy is observed in this case.

_{eff}Here, a question arises: what happens at the degenerate point? To answer this question, we calculated the normalized optical power flow *P _{norm}* (See Appendix A). Figures 4(d)–(f) show the normalized optical power flow of the anti-symmetric mode for the cases of

*ρ*<1, 1<

*ρ*<1.28 , and

*ρ*> 1.28 , respectively. It is seen in Fig. 4(d) that the anti-symmetric plasmonic and photonic modes for

*ρ*< 1 exhibit a negative normalized optical power flow, indicating that this solution corresponds to the negative mode. Let us remind that the sign of the power flow in the metal is negative due to the negative

*ε*and that in the dielectric is positive owing to the positive

_{m}*ε*. Whether the normalized total optical power flow is positive or negative depends on how much energy is guided through the metal and the dielectric. If the portion of energy guided in the metal cladding is larger than that in the dielectric core, the overall optical power flow becomes negative and vice versa. Note also that the skin depth of the surface mode propagating along the interface between a metal and a dielectric is inversely proportional to the relative electric permittivity i.e., the skin depths into the metal and the dielectric scale with 1|

_{d}*ε*| and 1/

_{m}*ε*. Therefore if

_{d}*ρ*< 1 i.e., |

*ε*| <

_{m}*ε*, then it is expected that more energy is guided through the metal region. The negative power flow of the anti-symmetric mode in the MIM waveguide can be ascribed to this property [24]. Unlike the anti-symmetric mode in the case

_{d}*ρ*< 1, the anti-symmetric mode in the case

*ρ*> 1.28 shows always positive values (Fig. 4(f)). This is because the portion of energy guided through the dielectric core is more than that through the metal cladding. In other words, the anti-symmetric modes for the case

*ρ*> 1.28 are always the positive modes.

The behavior in the case of 1<*ρ*<1.28 is noteworthy (Fig. 4(e)). When the single-interface SPPs are considered, it is expected that the total power flow is positive since the skin depth to the dielectric is longer than that to the metal (*ρ*>1). However, the anti-symmetric mode in the MIM waveguide originates from the coupling between two single-interface SPPs. Note that the anti-symmetric mode has a node line along the center of the dielectric core (*x* = 0), indicating that a certain portion of energy should be pumped out from the dielectric core and guided in the metal. Consequently, there exists chance that the energy guided in the metal cladding is the same as that in the dielectric core. If the reduced core width *ak*
_{0} is large enough, then the normalized optical power flow is positive due to the fact that more energy resides in the dielectric core. As *ak*
_{0} is decreased, the portion of the energy in the metal is increased, resulting in the decrease of the normalized optical power flow. If *ak*
_{0} < *ζ*, then the anti-symmetric photonic mode whose power flow is negative emerges and two anti-symmetric modes are supported simultaneously. As *ak*
_{0} continues to decrease and reaches a certain point, two modes degenerate into one mode whose normalized optical power flow vanishes, as can be shown in Fig. 4(e). In other words, the condition 1<*ρ*<1.28 guarantees that there exists a certain value of the reduced core width at which the normalized optical power flow vanishes. The rigorous verification of this property is presented in Appendix B.

So far, we have focused on the geometrical dispersion i.e., the dependence of |*n _{eff}*| on

*ak*

_{0}. It was shown that, when 1<

*ρ*<1.28, a specific core width exists where the normalized optical power flow vanishes for given values of the dielectric material

*ε*and the angular frequency

_{d}*ω*. The zero power flow can be obtained from another approaches, i.e., varying either

*ε*or

_{d}*ω*. Figures 5(a) and 5(b) illustrate |

*n*| and

_{eff}*P*as a function of

_{norm}*ε*. Here,

_{d}*ak*

_{0}is chosen in such a way that

*P*vanishes for

_{norm}*ε*=1. The anti-symmetric plasmonic mode is supported for 1<

_{d}*ε*<1.09 and the sign of

_{d}*P*is positive. It is observed that, as

_{norm}*ε*grows to 1.09 , |

_{d}*n*| diverges to infinite and the anti-symmetric plasmonic mode is cut off. As

_{eff}*ε*is decreased, |

_{d}*n*| of the anti-symmetric plasmonic mode is also decreased. However, the anti-symmetric photonic mode exists for 1<

_{eff}*ε*<1.17 with the negative

_{d}*P*. |

_{norm}*n*| of the anti-symmetric photonic mode is inversely proportional with

_{eff}*ε*. It is noteworthy that the mode degeneracy takes place for

_{d}*ε*= 1. Figure 5(b) makes sure that the normalized optical power flow vanishes at

_{d}*ε*=1. Meanwhile, it appears that Pnorm of the anti-symmetric plasmonic mode could vanish near

_{d}*ε*= 1.1, which has not been predicted from our analysis. This point seems not to be associated with the mode degeneracy. As can be seen in Fig. 5(a), I

_{d}*n*diverges at this point.

_{eff}This would lead to the diverging of *κ _{d}* and

*κ*(See Eqs. (7) and (8).). Unfortunately, it is not certain how the normalized optical power flow is affected by the diverging. It is thought that more study is required for this phenomenon.

_{m}Figures 5(c) and 5(d) show *β* and *Pnorm* as a function of *ω* according to the convention that *ω* is set along the ordinate. In Fig. 6(c), the ordinate is normalized with the bulk plasma frequency *ω _{P}* and the abscissa is normalized with the bulk plasma wavenumber

*β*. The horizontal dash-dotted line indicates the surface plasma frequency

_{p}*ω*=

*ωSP*i.e.,

*ρ*= 1. The dashed straight line shows the light line

*β*=

*ω*√

*ε*/

_{d}*c*

_{0}. The dotted curve lying at the right side of the light line corresponds to the dispersion relation of the single-interface SPP. Note that

*ε*is also a function of

_{m}*ω*, which is given by Eq. (1). The solid curve denotes the dispersion relation of the anti-symmetric mode in the MIM waveguide. In Fig. 6(c), it is seen that a certain range of

*ω*has two solutions of

*β*. As

*ω*decreases, two branches come close and the mode degeneracy takes place at a certain value of

*ω*. It is remarkable that the group velocity, defined by

*dω*/

*dβ*becomes zero at the degenerate point. To best of our knowledge, the zero-group velocity in the plasmonic waveguide, which does not include any metamaterials for the core nor the cladding, has not been reported. Figure 6(d) confirms that the normalized optical power flow vanishes at the degenerate point.

Before continuing to examine other cases in the plasmonic waveguide, let us make a brief summary for the anti-symmetric mode in the MIM waveguide. It was shown that the condition of 1<*ρ*<1.28 guarantees that two anti-symmetric modes can be supported simultaneously for *ξ* < *ak*
_{0} < *ζ*. These two modes degenerate into one mode at a certain reduced core width of *ak*
_{0} < *ξ*, where the normalized optical power flow vanishes. The dispersion relation shows that the group velocity becomes zero at the degenerate point.

#### 3.2 Symmetric mode

In this section, by employing a similar method used above for the case of the anti-symmetric mode, we investigate the condition for the existence of the symmetric plasmonic and photonic mode. It will be shown that the symmetric mode also has the mode degeneracy where the optical power flux vanishes.

The characteristic equation of the symmetric plasmonic mode is obtained from Eqs. (5), (7), and (8) with help of Eqs. (10), (11), and (13). Figures 7(a) and 7(b) illustrate two curves of *W* = -*ρU*tanh*U* and *W* = √*U*
^{2} + *V*
^{2} for those two cases of *ρ*<1 and *ρ*>1 , respectively. Note that *ρU*tanh*U* → 0 as *U* → 0 and *ρU* tanh *U* → *ρU* as *U* → ∞ , whereas √*U*
^{2} + *V*
^{2} + *V*
^{2} → *V* as *U* → 0 and √*U*
^{2} + *V*
^{2} → as *U* → *ω*. Due to the behavior of the tanh function, *ρU* tanh *U* is always smaller than *ρU*. In addition, The change in the reduced core width *ak*
_{0} affects only *V*, which is the *W* - intersect of the *W* = √*U*
^{2} + *V*
^{2} curve. Consequently, if *ρ*, which is the slope of the asymptote line of *ρU* tanh *U*, is smaller than one, there is no chance that two curves (*W* = *ρU*tanh*U* and *W* = √*U*
^{2} + *V*
^{2}) have an intersection (Fig. 7(b)). However, if *ρ*>1, then it is guaranteed that the symmetric plasmonic mode can always be supported (Fig. 7(a)). Table 3 summarizes the aforementioned discussion.

Next, let us examine the condition for the existence of the symmetric photonic mode in the MIM waveguide. From Eqs. (6), (7), and (9) with help of Eqs. (10), (12), and (13), we obtain two curves of *W* = -*pJ*tan*J* and *W* = √-*j*
^{2} + *V*
^{2} (See Fig. 8(a).). As in the anti-symmetric photonic mode, the domain of definition of *J* is given by *J* ∈ [0,*ak*
_{0}√*ε _{d}*] and we focus on the fundamental symmetric photonic mode. Let us consider the case when p is not so small (Fig. 8(a)). It will be shown that the condition for existence of the symmetric photonic mode gets more complex for

*ρ*≪ 1(Fig. 8(b)). Note that -

*ρJ*tan

*J*is positive only for

*J*∈ (

*nπ*-

*π*/2,

*nπ*), where

*n*is an integer. The fundamental symmetric photonic mode comes from

*J*∈ (

*π*/ 2,

*π*). Hence it is required that

*ak*

_{0}√

*ε*>

_{d}*π*/2, resulting in the condition

*ak*

_{0}>σ with

If -*ρJ* tan *J* at *J* = *ak*
_{0}√*ε _{d}* (the point A in Fig. 8(a)) is smaller than √-

*J*

^{2}+

*V*

^{2}at

*J*=

*ak*

_{0}√

*ε*(the point B in Fig. 8(a)), i.e., -

_{d}*ρ*(

*ak*

_{0}√

*ε*) tan (

_{d}*ak*

_{0}√

*ε*) <

_{d}*ak*

_{0}√-

*ε*, then it is guaranteed that the symmetric photonic mode is supported. This relation reduces to a simple form of

_{m}*ak*

_{0}> χ, where

It should be pointed out that Eq. (18) can also be obtained with the substitution *β* = 0 into Eqs. (6), (7), and (9). Here, let us take the range of the tan^{-1} function as (*π*/2,3*π*,2). Since √*ε _{d}*/(-

*ε*) is positive, χ lies in (π/2√

_{m}*ε*),

_{d}*π*/ √

*ε*). Thus χ is always larger than

_{d}*σ*. Therefore the condition of

*ak*

_{0}>χ is sufficient for the existence of the symmetric photonic mode.

Now we deal with the case of *ρ* ≪ 1. It will be shown that the MIM waveguide can support two symmetric photonic modes for a certain combination (not clearly specified) of *ρ* and *ak*
_{0}. Let us recall that -*ρJ* tan *J* and √-*J*
^{2} + *V*
^{2} are decreasing function of *J*. Thus it is expected that those curves have more than one intersection under a certain criterion. In Fig. 8(b), it is seen that two curves (-*ρJ*tan*J* and √-*J*
^{2} + *V*
^{2}) have two intersections, indicating that two symmetric photonic modes are simultaneously supported. The decrease of *ak*
_{0} gives rise to the decrease of the radius (*V*) of the circle √-*J*
^{2} + *V*
^{2} and two intersects come close to each other. Therefore, it is expected that, at a certain value of *ak*
_{0}, those two intersections degenerate into single point of contact, giving rise to trapped light. Table 3 summarizes the aforementioned discussion.

Now let us consider the dependence of *n _{eff}* and

*P*on

_{norm}*ak*

_{0}for the symmetric mode in the MIM waveguide. Figures 9(a)–(c) show results for the cases of

*ρ*≪ 1,

*ρ*< 1, and

*ρ*> 1, respectively. The vertical dashed and dash-dotted lines in Figs. 9(a)–(f) denote

*ak*

_{0}=

*σ*and

*ak*

_{0}= χ, respectively. We first examine the

*ρ*< 1 case (Fig. 9(b)). The horizontal dotted line indicates the refractive index of the cladding (|

*n*| = √

_{eff}*ε*). No symmetric plasmonic mode is observed, originating from the fact that a single interface between the metal with

_{d}*ε*and the dielectric with

_{m}*ε*does not support the plasmonic mode when

_{d}*ρ*< 1. Note that this property is different from the result of the anti-symmetric plasmonic mode, which can be supported even if

*ρ*< 1 (See Fig. 4(a)). In contrast to the symmetric plasmonic mode, the symmetric photonic mode is supported even for

*ρ*< 1, provided that

*ak*

_{0}> χ. Figure 9(e), illustrating

*P*for the case

_{norm}*ρ*< 1, shows that the optical power flow of the symmetric photonic mode is positive.

Meanwhile, the functional behavior of the symmetric mode for *ρ* > 1 (Figs. 9(c) and 9(f)) is well established [20, 21, 26–28]. Ordinary plasmonic devices that consist of noble metals such as gold or silver and dielectric materials such as oxides or air satisfy -*ε _{m}* >

*ε*in the visible and infrared regimes. As can be seen in Fig. 7(a), the symmetric plasmonic mode exists no matter how narrow the core width is. In other words, the symmetric plasmonic mode does not exhibit the mode cut off associated with the narrow core width [27]. It was also reported that the effective refractive index of the symmetric plasmonic mode in the MIM waveguide reveals an asymptotic behavior as the angular frequency is increased to the infrared regime [28]. Figure 9(f) reveals that its normalized optical power flow is always positive. The symmetric mode for

_{d}*ρ*> 1 has a photonic mode when

*ak*

_{0}> χ with a positive

*P*(See Figs. 9(c) and 9(f).).

_{norm}A notable feature is that the symmetric mode can hold two photonic modes
simultaneously if *ρ* ≪ 1 (Fig. 9(a)). This was expected in the analysis in Fig. 8(b). It turns out in Fig. 9(d) that the upper branch of these two photonic modes corresponds to the positive mode, whereas the lower branch the negative mode. At the degeneracy point, the symmetric mode conveys the optical power flow neither forward nor backward. Note that the reduced core width of the degeneracy is around *ak*
_{0} = χ.

## 4. Insulator-metal-insulator plasmonic waveguide

In this section, we show that a certain condition exists that two symmetric modes can be simultaneously supported in the IMI waveguide. The optical power flow becomes zero at the degenerate point. On the other hand, the anti-symmetric mode has only one solution, and thus the zero optical power flow does not occur.

Figure 10 shows the schematic diagram of the IMI waveguide. There are two kinds of modes: one is the anti-symmetric plasmonic mode and the other is the symmetric plasmonic mode. In contrast to the MIM waveguide, where the photonic mode as well as the plasmonic mode is allowed, the IMI waveguide does not support the photonic mode. This is because the metallic core of the IMI waveguide is opaque. As in the previous section, the wavelength is chosen to be 400 nm. We derive conditions that each mode is guided in the IMI waveguide. From the boundary condition that tangential electric and magnetic fields are continuous across the interface between the core and the cladding, the characteristic equations are given by [23]

#### 4.1 Symmetric mode

Let us first deal with the symmetric mode. Only plasmonic mode is allowed. The characteristic equation of the symmetric plasmonic mode in the IMI waveguide is derived from Eqs. (19), (7), and (8) along with Eqs. (10), (11), and (13). Figures 11(a) and 11(b) depict the graphical method to determine the solution of the characteristic equation. The solution corresponds to the intersection between two curves *U* = *ρ*
^{-1}tanh*W* (the red dash-dotted lines in Figs. 11(a) and 11(b)) and *U* = √*W*
^{2} - *V*
^{2} (the blue solid lines in Figs. 11(a) and 11(b)). Their asymptotic lines are *U* = *ρ*
^{-1}
*W* (the black dashed lines) and *U* = *W* (the green dotted lines), respectively. The lines *W* = *V* are shown as the horizontal dashed lines. For **U** to be real, *W* should be larger than or equal to *V*, i.e., the domain of definition of *W* is *W* ∈ [*V*,∞]. As in the analysis for the MIM waveguide, discussion splits depending on whether *ρ* > 1 or *ρ* < 1.

In Fig. 11(a), we illustrate the aforementioned two curves and their asymptotic lines for the case *ρ* > 1. The curve *U* = *ρ*
^{-1}
*W* tanh *W* grows from zero with the increase of *W* and is present to the right to the straight line *U* = *ρ*
^{-1}
*W* , the slope of which is smaller than one. On the other hand, the curve *U* = √*w*
^{2} - *V*
^{2} asymptotes to the line *U* = *W* , whose slope is one. As a result, if *ρ* > 1 , then two curves *U* = *ρ*
^{-1}
*W*tanh*W* and *U* = √*W*
^{2} - *V*
^{2} always have an intersection. Although the change in the core width, which contributes to the change in *V*, may affect the value of *β* of the mode, the number of the mode is still one.

Figure 11(b) show the two curves *U* = *ρ*
^{-1}
*W*tanh*W* and *U* =√*W*
^{2} - *V*
^{2} for the case *ρ* < 1. Let us first consider two extreme cases: one is the case where the core width is very small (*V* → 0) and the other is the case in which the core width is considerably large (*V* ≫ 1). In the former case, it is expected that there are two intersections at *U* = 0 and *U* = *W* , which corresponds to |*n _{eff}*| = √

*ε*and |

_{d}*n*| → ∞, respectively. In the latter case, where the curve

_{eff}*U*= √

*W*

^{2}-

*V*

^{2}is moved to the right far away, there is no chance that

*U*=

*ρ*

^{-1}

*W*tanh

*W*and

*U*= √

*W*

^{2}-

*V*

^{2}have an intersection (see Fig. 11(b)). Consequently, it can be inferred that, as

*ak*

_{0}is increased from zero, |

*n*| of two modes would come closer to each other, and at a certain value of

_{eff}*ak*

_{0}, they would degenerate into single mode. Unfortunately, we could not extract the exact formulation to get the value of

*ak*

_{0}where the mode degeneracy occurs. Table 4 summarizes the aforementioned properties.

Figures 12(a)-(d) show the effect of *ak*
_{0} on |*n _{eff}*| and

*P*in the case

_{norm}*ρ*< 1 and

*ρ*> 1. As explained above, two symmetric modes are simultaneously allowed for the core width under a certain point (See Fig. 12(a).). The difference of |

*n*| between the upper and lower branches is decreased with the increase of

_{eff}*ak*

_{0}. At a certain point of

*ak*

_{0}, these two modes degenerate into single mode. In Fig. 12(c), it is confirmed that the normalized optical power flow vanishes at the degenerate point, resulting in trapping of light. Contrarily, the results in the case

*ρ*> 1 (Figs. 12(b) and 12(d)) show that the dependences of |

*n*| and

_{eff}*P*on

_{norm}*ak*

_{0}are monotonic. If

*ak*

_{0}is sufficiently large, the functional behavior asymptotes to the result of the single interface SPP (See Fig. 12(b).). As

*ak*

_{0}is decreased, more power is guided through the dielectric, which makes |

*n*| asymptotes to √

_{eff}*ε*. This is in good agreement with results of previous studies [23, 33, 34]. In the presence of material loss, this mode exhibits a long propagation length compared to the anti-symmetric mode and is referred to as the long range SPP [35]. In Fig. 12(d), it is observed that

_{d}*P*→ 1 as

_{norm}*ak*

_{0}→ 0, which indicates that all power is guided through the dielectric cladding, leading to a disadvantage from a view point of confinement.

#### 4.2 Anti-symmetric mode

The analysis on the anti-symmetric mode in the IMI waveguide is quite simple. In this section, it is shown that the anti-symmetric mode in the IMI waveguide exists only when *ρ* > 1 and this mode is a positive mode. In the case *ρ* < 1, there is no anti-symmetric mode regardless of the core width. In addition, the mode degeneracy does not occur for the anti-symmetric mode in the IMI waveguide.

As in the analysis for the symmetric mode in the IMI waveguide, the photonic mode is not supported and only the plasmonic mode is considered. The characteristic equation of the anti-symmetric plasmonic mode in the IMI waveguide is obtained from Eqs. (20), (7), and (8) along with Eqs. (10), (11), and (13). In Figs. 13(a) and 13(b), we illustrate two curves of *U* = *ρ*
^{-1}
*W*coth*W* (the red dash-dotted lines) and *U* =√*W*
^{2} - *V*
^{2} (the blue solid lines). Their asymptotic lines are given by *U* = *ρ*
^{-1}
*W* (the black dashed lines) and *U* = *W* (the green dotted lines), respectively. The horizontal dashed lines denote lines *W* = *V* . The domain of definition of *W* is given by *W* ∈ [*V*,∞].

Let us first consider the case *ρ* > 1 (Fig. 13(a)). As *W* is increased, *U* = √*W*
^{2} - *V*
^{2} grows monotonically from zero and asymptotes to the line *U* = *W*. In contrast, *U* = ρ^{-1}
*W*coth*W* increases with the increase of *W* and it asymptotes to the line *U* = *ρ*
^{-1}
*W*, the slope of which is smaller than that of *U* = *W*. Therefore, two curves *U* = *ρ*
^{-1}
*W*coth*W* and *U* = √*W*
^{2} - *V*
^{2} always have an intersection. The variance in *ak*
_{0} may increase or decrease *V*, which is the *W* - intersect in the graph. However, the change in *V* does not affect the number of intersections, i.e., there is only one intersection regardless of the core width.

Next we move to the discussion for the case *ρ* < 1 (Fig. 13(b)). Note that the curve *U* = *ρ*
^{-1}
*W*coth*W* is always located in the upper region of its asymptotic line *U* = *W*
^{-1}
*W*, while the curve *U* = √*W*
^{2} - *V*
^{2} is under its asymptotic line *U* = *W* . Since the slope of *U* = *ρ*
^{-1}
*W* is larger than that of *U* = *W* , there is no chance that *U* = *ρ*
^{-1}
*W*coth*W* and *U* = √*W*
^{2} - *V*
^{2} have an intersection. Table 5 summarizes the aforementioned properties.

In Figs. 14(a) and 14(b), we illustrate |*n _{eff}*| and

*P3*as a function of

*ak*

_{0}in the case

*ρ*> 1. If

*ak*

_{0}is large, |

*n*| asymptotes to the line |

_{eff}*n*| =

_{eff}*n*(Fig. 14(a)). As

_{spp}*ak*

_{0}dwindles, |

*n*| is increased monotonically and cut off is not observed. This coincides well with published results [23, 33, 34]. In particular, it has been reported that the absence of cut off can be used to confine light in the subwavelength scale [36, 37]. The nanofocusing of light in a metallic nano rod was theoretically expected [36], which was experimentally demonstrated by using the asymmetric cladding and adiabatic tapering [37]. The guided modes explored in these reports are associated with the anti-symmetric mode in the IMI waveguide [33, 37]. If the material loss is taken into account, this mode shows a short propagation length compared to the symmetric mode and is called the short range SPP [35]. The dependence of |

_{eff}*n*| and

_{eff}*P*on

_{norm}*ak*

_{0}in the case

*ρ*< 1 is not provided since no mode exists in that case. Thus the mode degeneration and trapping of light do not occur for the anti-symmetric mode in the IMI waveguide.

## 5. Conclusion

Comprehensive case studies on trapping of light in the plasmonic waveguide are presented. The fundamental principle is simply based on the fact that the direction of the optical power flow in a metal composing a plasmonic waveguide is opposite to that in a dielectric. If amounts of these two optical power flows are the same, then the total optical power flow vanishes. However, the problem on the existence of such a balance point in a plasmonic waveguide is not trivial. The ratio between the optical power flows guided in the metal and the dielectric is governed by two factors: one is a ratio *ρ* between the relative electric permittivity of the metal *ε _{d}* and the dielectric

*ε*i.e.,

_{d}*ρ*= -

*ε*/

_{m}*ε*. The other is a ratio between the core width 2

_{d}*a*and the operating wavelength

*λ*

_{0}= 2

*π*/

*k*

_{0}, and we use the reduced core width

*ak*

_{0}. The rigorous case studies conducted for the lossless two dimensional MIM and IMI waveguides demonstrate that trapping of light occurs in the following cases: 1) the anti-symmetric mode in the MIM waveguide with 1<

*ρ*<1.28 and

*ak*

_{0}<

*ξ*(Figs. 4(b) and 4(e)), 2) the symmetric mode in the MIM waveguide with

*ρ*≪ 1 and

*ak*

_{0}< χ (Figs. 9(a) and 9(d)), and 3) the symmetric mode in the IMI waveguide with

*ρ*< 1 and small

*ak*

_{0}(Figs. 12(a) and 12(c). The explicit condition on the reduced core width is not given). The definitions of

*ξ*, and χ are presented in Eqs. (14) and (18), respectively. It is believed that the aforementioned results of this study can open a way to implement various optical devices with trapped or slow light.

## Appendix A

The *y* -component of the magnetic field of the anti-symmetric plasmonic mode in the MIM waveguide is expressed as

From the Maxwell’s curl equation, the *x* - and *z* - components of the electric field of the anti-symmetric plasmonic mode are given by

The relation between the coupling coefficients *A*
_{1} and *B*
_{1} is obtained from the boundary condition for Eqs. (A1) and (A3) as

which results in the characteristic equation

The *z* -component of the Poynting vector *P _{z}* =

*E*

_{x}H^{*}

_{y}/2 is given from Eqs. (A1), (A2), and (A4) as

Note that *κ _{d}* and

*κ*are assumed to be purely real here. By taking an integral in Eq. (A7), we get the total power propagating through the metal cladding and the dielectric core as follows:

_{m}$$\phantom{\rule{1.2em}{0ex}}=2\left[\left(\frac{\beta {\mid {B}_{1}\mid}^{2}}{\omega {\epsilon}_{0}}\right)\left(\frac{1}{{\epsilon}_{m}}\right){\mathrm{sinh}}^{2}\left({\kappa}_{d}a\right)\right]\underset{0}{\overset{\infty}{\int}}\mathrm{exp}\left[-2{\kappa}_{m}x\right]dx$$

$$\phantom{\rule{1.2em}{0ex}}=\left(\frac{\beta {\mid {B}_{1}\mid}^{2}}{\omega {\epsilon}_{0}}\right)\left(\frac{{\mathrm{sinh}}^{2}\left({\kappa}_{d}a\right)}{{\epsilon}_{m}{\kappa}_{m}}\right),$$

$$\phantom{\rule{1.2em}{0ex}}=2\left(\frac{\beta {\mid {B}_{1}\mid}^{2}}{\omega {\epsilon}_{0}}\right)\left(\frac{1}{{\epsilon}_{d}}\right)\underset{0}{\overset{a}{\int}}{\mathrm{sinh}}^{2}\left({\kappa}_{d}x\right)dx$$

$$\phantom{\rule{1.2em}{0ex}}=2\left(\frac{\beta {\mid {B}_{1}\mid}^{2}}{\omega {\epsilon}_{0}}\right)\left(\frac{1}{{\epsilon}_{d}}\right){\left[\frac{\mathrm{sinh}\left(2{\kappa}_{d}x\right)}{4{\kappa}_{d}}-\frac{x}{2}\right]}_{0}^{a}$$

$$\phantom{\rule{1.2em}{0ex}}=\left(\frac{\beta {\mid {B}_{1}\mid}^{2}}{\omega {\epsilon}_{0}}\right)\left(\frac{1}{{\epsilon}_{d}{\kappa}_{d}}\right)\left[\frac{\mathrm{sinh}\left({2\kappa}_{d}a\right)}{2}-a{\kappa}_{d}\right].$$

The normalized power flow *P _{norm}* = (

*P*+

_{dielectric}*P*)/(

_{metal}*P*+

_{dielectric}*P*) for the anti-symmetric plasmonic mode in the MIM waveguide is defined as

_{metal}The counterparts of Eqs. (A6), (A10) for the anti-symmetric photonic mode, the symmetric plasmonic and photonic modes in the MIM waveguide, and the anti-symmetric and symmetric plasmonic modes in the IMI waveguide can be derived in the same way by replacing Eq. (A1) as

respectively.

## Appendix B

From the characteristic equation for the anti-symmetric plasmonic mode in the MIM waveguide i.e., Eq. (3), we calculate the partial derivative of *ak*
_{0} on *n _{eff}* as follows:

The derivative of *ak*
_{0} on *n _{eff}* is obtained as follows:

which results in

Comparison between Eqs. (A10) to (B3) reveals that the sign of *d*(*ak*
_{0})/*dn _{eff}* is the same as
that of

*P*. Therefore the normalized power flux vanishes at the mode degenerate point, where

_{norm}*d*(

*ak*

_{0})/

*dn*= 0. In addition, it can be shown that positive is the normalized power flux of the upper branch of the anti-symmetric plasmonic mode for 1<

_{eff}*ρ*< 1.28, where

*n*is decreased with the decrease of

_{eff}*ak*

_{0}(

*d*(

*ak*

_{0})/

*d*

*n*> 0), whereas negative is that of the lower branch owing to the fact that

_{eff}*d*(

*ak*

_{0})/

*dn*, < 0 (See Fig. 4(a).). The counterparts of

_{eff}*d*(

*ak*

_{0})/

*d*

*n*for the anti-symmetric photonic mode, the symmetric plasmonic and photonic modes in the MIM waveguide, and the anti-symmetric and symmetric plasmonic modes in the IMI waveguide can be derived in the similar way.

_{eff}## Acknowledgment

This work was supported by the National Research Foundation and the Ministry of Education, Science and Technology of Korea through the Creative Research Initiatives Program (Active Plasmonics Application Systems).

## References and links

**1. **T. F. Krauss, “Why do we need slow light?,” Nature Photon. **2**, 448–450 (2008). [CrossRef]

**2. **L. Thevenaz, “Slow and fast light in optical fibres,” Nature Photon. **2**, 474–481 (2008). [CrossRef]

**3. **J. E. Heebner, R. W. Boyd, and Q. H. Park, “Slow light, induced dispersion, enhanced nonlinearity, and optical solitons in a resonator-array waveguide,” Phys. Rev. E **65**, 036619 (2002). [CrossRef]

**4. **D. M. Beggs, T. P. White, L. O’Faolain, and T. F. Krauss, “Ultracompact and low-power optical switch based on silicon photonic crystals,” Opt. Lett. **33**, 147–149 (2008). [CrossRef] [PubMed]

**5. **M. Notomi and H. Taniyama, “On-demand ultrahigh-Q cavity formation and photon pinning via dynamic waveguide tuning,” Opt. Express **16**, 18657–18666 (2008). [CrossRef]

**6. **T. Tanabe, M. Notomi, H. Taniyama, and E. Kuramochi, “Dynamic release of trapped light from an ultrahigh-Q nanocavity via adiabatic frequency tuning,” Phys. Rev. Lett. **102**, 043907 (2009). [CrossRef] [PubMed]

**7. **T. Baba and D. Mori, “Slow light engineering in photonic crystals,” J. Phys. D **40**, 2659–2665 (2007). [CrossRef]

**8. **T. Kawasaki, D. Mori, and T. Baba, “Experimental observation of slow light in photonic crystal coupled waveguides,” Opt. Express **15**, 10274–10281 (2007). [CrossRef] [PubMed]

**9. **D. Mori, S. Kubo, H. Sasaki, and T. Baba, “Experimental demonstration of wideband dispersion-compensated slow light by a chirped photonic crystal directional coupler,” Opt. Express **15**, 5264–5270 (2007). [CrossRef] [PubMed]

**10. **T. Baba, “Slow light in photonic crystals,” Nature Photon. **2**, 465–473 (2008). [CrossRef]

**11. **K. L. Tsakmakidis, A. D. Boardman, and O. Hess, “Trapped rainbow’ storage of light in metamaterials,” Nature **450**, 397–401 (2007). [CrossRef] [PubMed]

**12. **K. L. Tsakmakidis, C. Hermann, A. Klaedtke, C. Jamois, and O. Hess, “Surface plasmon polaritons in generalized slab heterostructures with negative permittivity and permeability,“ Phys. Rev. B **73**, 085104 (2006). [CrossRef]

**13. **Z. Fu, Q. Q. Gan, Y. J. J. Ding, and F. J. Bartoli, “From waveguiding to spatial localization of THz waves within a plasmonic metallic grating,“ IEEE J. Sel. Topics Quantum Electron. **14**, 486–490 (2008). [CrossRef]

**14. **Q. Q. Gan, Z. Fu, Y. J. Ding, and F. J. Bartoli, “Ultrawide-bandwidth slow-light system based on THz plasmonic graded metallic grating structures,“ Phys. Rev. Lett. **101**, 169903 (2008). [CrossRef]

**15. **Q. Q. Gan, Y. J. Ding, and F. J. Bartoli, ““Rainbow“ trapping and releasing at telecommunication wavelengths,“ Phys. Rev. Lett. **102**, 056801 (2009). [CrossRef] [PubMed]

**16. **K. Y. Kim, “Tunneling-induced temporary light trapping in negative-index-clad slab waveguide,“ Jpn. J. Appl. Phys. **47**, 4843–4845 (2008). [CrossRef]

**17. **V. M. Shalaev, “Optical negative-index metamaterials,“ Nature Photon. **1**, 41–48 (2007). [CrossRef]

**18. **J. Valentine, S. Zhang, T. Zentgraf, E. Ulin-Avila, D. A. Genov, G. Bartal, and X. Zhang, “Three-dimensional optical metamaterial with a negative refractive index,“ Nature **455**, 376–U332 (2008). [CrossRef] [PubMed]

**19. **J. Yao, Z. W. Liu, Y. M. Liu, Y. Wang, C. Sun, G. Bartal, A. M. Stacy, and X. Zhang, “Optical negative refraction in bulk metamaterials of nanowires,“ Science **321**, 930–930 (2008). [CrossRef] [PubMed]

**20. **B. Prade, J. Y. Vinet, and A. Mysyrowicz, “Guided optical waves in planar heterostructures with negative dielectric-constant,“ Phys. Rev. B **44**, 13556–13572 (1991). [CrossRef]

**21. **J. A. Dionne, L. A. Sweatlock, H. A. Atwater, and A. Polman, “Plasmon slot waveguides: Towards chip-scale propagation with subwavelength-scale localization,“ Phys. Rev. B **73**, 035407 (2006). [CrossRef]

**22. **J. A. Dionne, E. Verhagen, A. Polman, and H. A. Atwater, “Are negative index materials achievable with surface plasmon waveguides? A case study of three plasmonic geometries,“ Opt. Express **16**, 19001–19017 (2008). [CrossRef]

**23. **J. A. Dionne, L. A. Sweatlock, H. A. Atwater, and A. Polman, “Planar metal plasmon waveguides: frequency-dependent dispersion, propagation, localization, and loss beyond the free electron model,“ Phys. Rev. B **72**, 075405 (2005). [CrossRef]

**24. **H. J. Lezec, J. A. Dionne, and H. A. Atwater, “Negative refraction at visible frequencies,“ Science **316**, 430–432 (2007). [CrossRef] [PubMed]

**25. **J. A. Dionne, H. J. Lezec, and H. A. Atwater, “Highly confined photon transport in subwavelength metallic slot waveguides,“ Nano Lett. **6**, 1928–1932 (2006). [CrossRef] [PubMed]

**26. **J. Park, H. Kim, and B. Lee, “High order plasmonic Bragg reflection in the metal-insulator-metal waveguide Bragg grating,“ Opt. Express **16**, 413–425 (2008). [CrossRef] [PubMed]

**27. **J. Park, H. Kim, I.-M. Lee, S. Kim, J. Jung, and B. Lee, “Resonant tunneling of surface plasmon polariton in the plasmonic nano-cavity,“ Opt. Express **16**, 16903–16915 (2008). [CrossRef] [PubMed]

**28. **J. Park and B. Lee, “An approximate formula of the effective refractive index of the metal-insulator-metal surface plasmon polariton waveguide in the infrared region,“ Jpn. J. Appl. Phys. **47**, 8449–8451 (2008). [CrossRef]

**29. **E. Feigenbaum and M. Orenstein, “Backward propagating slow light in inverted plasmonic taper,“ Opt. Express **17**, 2465–2469 (2009). [CrossRef] [PubMed]

**30. **A. R. Davoyan, I. V. Shadrivov, and Y. S. Kivshar, “Nonlinear plasmonic slot waveguides,“ Opt. Express **16**, 21209–21214 (2008). [CrossRef] [PubMed]

**31. **E. Feigenbaum, N. Kaminski, and M. Orenstein, “Negative group velocity: Is it a backward wave or fast light?,“ arXiv:0807.4915 (2008).

**32. **F. Kusunoki, T. Yotsuya, and J. Takahara, “Confinement and guiding of two-dimensional optical waves by low-refractive-index cores,“ Opt. Express **14**, 5651–5656 (2006). [CrossRef] [PubMed]

**33. **J. J. Burke, G. I. Stegeman, and T. Tamir, “Surface-polariton-like waves guided by thin, lossy metal-films,“ Phys. Rev. B **33**, 5186–5201 (1986). [CrossRef]

**34. **J. Takahara and T. Kobayashi, “Low-dimensional optical waves and nano-optical circuits,“ Optics and Photonics News **6**, 54–59 (2004). [CrossRef]

**35. **P. Berini, “Plasmon-polariton waves guided by thin lossy metal films of finite width: bound modes of symmetric structures,“ Phys. Rev. B **61**, 10484–10503 (2000). [CrossRef]

**36. **M. I. Stockman, “Nanofocusing of optical energy in tapered plasmonic waveguides,“ Phys. Rev. Lett. **93**, 137404 (2004). [CrossRef] [PubMed]

**37. **E. Verhagen, A. Polman, and L. Kuipers, “Nanofocusing in laterally tapered plasmonic waveguides,“ Opt. Express **16**, 45–57 (2008). [CrossRef] [PubMed]