1. Introduction
Surface plasmon-polaritons (SPPs) are surface localized waves propagating on an interface between a metal and a dielectric. Over a frequency band in which the dielectric constant of metal is less than minus of that of adjacent dielectric, SPPs are excited by coupling between collective oscillations of free electrons and evanescent electromagnetic (EM) waves [
1
H. Raether, Surface plasmons on Smooth and Rough Surfaces and on Gratings , (Springer-Verlag, Berlin, 1988).
,
2
W. L. Barnes, A. Dereux, and T. W. Ebbesen, (Insight review articles) “Surface plasmon subwavelength optics,” Nature
424, 824–830 (2003). [CrossRef] [PubMed]
]. This resonant coupling gives SPPs the peculiar properties of surface localizations and field enhancements, which make a lightwave be able to interact strongly with nanoscopic structures by mediating them. For these natures, SPPs have been widely studied since the first optical observation in the anomalous diffraction from a metallic grating by Wood [
3
R. W. Wood, “On a remarkable case of uneven distribution of light in a diffraction grating spectrum,” Phil. Mag.
4, 396 (1902).
,
4
R. W. Wood, “Anomalous diffraction gratings,” Phys. Rev.
48, 928–936 (1935). [CrossRef]
] and its causal explanation by Fano [
5
U. Fano, “The theory of anomalous diffraction gratings and of quasi-stationary waves on metallic surfaces (Sommerfeld’s waves),” J. Opt. Soc. Am.
31, 213–222 (1941). [CrossRef]
]. Applications of SPPs are ranging from colored glasses [
6
S. Linden, J. Kuhl, and H. Giessen, “Controlling the interaction between light and gold nano particles: Selective suppression of extinction,” Phys. Rev. Lett.
86, 4688 (2001). [CrossRef] [PubMed]
] or attenuated total reflection (ATR) based devices [
7
E. Kretchmann and H. Raether, “Radiative decay of nonradiative surface plasmons excited by light,” Z. Naturkorsch. A
23, 2135–2136 (1968).
–
9
J. Homola, S. S. Yee, and G. Gauglitz, “Surface plasmon resonance sensors: review,” Sensors Actuat. B
54, 3–15 (1999). [CrossRef]
] to even potentials for merging photonic and electronic integrated circuits recently [
2
W. L. Barnes, A. Dereux, and T. W. Ebbesen, (Insight review articles) “Surface plasmon subwavelength optics,” Nature
424, 824–830 (2003). [CrossRef] [PubMed]
,
10
E. Ozbay, (Review) “Plasmonics: Merging photonics and electronics at nanoscale dimensions,” Science
311, 189–193 (2006). [CrossRef] [PubMed]
–
13
J. R. Krenn and J.-C. Weeber, “Surface plasmon polaritons in metal stripes and wires,” Phil. Trans. R. Soc. Lond. A
362, 739–756 (2004). [CrossRef]
].
One of the active fields regarding SPPs is applications of long-range SPPs (LRSPPs). LRSPPs are supported by anti-symmetric charge coupling of SPPs on two surfaces of a thin metal film of which thickness is comparable to or less than the penetration depth of EM wave. For its small propagation constant and nearly transversal field components, internal loss (mainly ohmic loss in the metal) is small and direct end-fire coupling with any fundamental mode of a conventional dielectric based waveguide is very efficient [
1
H. Raether, Surface plasmons on Smooth and Rough Surfaces and on Gratings , (Springer-Verlag, Berlin, 1988).
,
14
D. Sarid, “Long-range surface-plasma waves on very thin metal films,” Phys, Rev. Lett.
47, 1927–1930 (1981). [CrossRef]
,
15
P. Berini, “Plasmon-polariton waves guided by thin lossy metal films of finite width: Bounded modes of symmetric structures,” Phys. Rev. B
61, 10484 (2000). [CrossRef]
]. Though LRSPPs are not so confined modes and the loss is still larger than dielectric waveguides, high cladding sensitivity and parallelism of optical and electrical signals make them useful for integrated optical devices. In addition, LRSPP waveguides consisting of about 20
nm thick and a few
µm wide metal strips are much easier to be fabricated by conventional lithography technique than index-guiding dielectric waveguides which require a wavelength scale depth in index steps. Many integrated passive and active optical devices based on LRSPPs such as channel waveguides [
16
T. Nikolajsen, K. Leosson, I. Salakhutdinov, and S. I. Bozhevolnyi, “Polymer-based surface-plasmon-polariton stripe waveguides at telecommunication wavelengths,” Appl. Phys. Lett.
82, 668–670 (2003). [CrossRef]
,
17
P. Berini, R. Charbonneau, N. Lahoud, and G. Mattiussi, “Characterization of long-range surface-plasmon-polariton waveguides,” J. Appl. Phys.
98, Art. No. 043109 (2005). [CrossRef]
], thermo-optic modulators [
18
T. Nikolajsen, K. Leosson, and S. I. Bozhevolnyi, “In-line extinction modulator based on long-range surface plasmon polaritons,” Opt. Commun.
244, 455–459 (2005). [CrossRef]
–
20
G. Gagnon, N. Lahoud, G. A. Mattiussi, and P. Berini, “Thermally activated variable attenuation of long-range surface plasmon-polariton waves,” J. Lightwave Technol.
24, 4391–4402 (2006). [CrossRef]
], directional couplers [
21
A. Boltasseva, T. Nikolajsen, K. Leosson, K. Kjaer, M. S. Larsen, and S. I. Bozhevolnyi, “Integrated optical components utilizing long-range surface plasmon polaritons,” J. Lightwave technol.
23, 413–422 (2005). [CrossRef]
–
23
R. Charbonneau, C. Scales, I. Breukelaar, S. Fafard, N. Lahoud, G. Mattiussi, and P. Berini, “Passive integrated optics elements based on long-range surface plasmon polaritons,” J. Lightwave Technol.
24, 477–494 (2006). [CrossRef]
], and Bragg grating filters [
24
A. Boltasseva, S. I. Bozhevolnyi, T. Søndergaard, T. Nikolajsen, and K. Leosson, “Compact Z-add-drop wavelength filters for long-range surface plasmon polaritons,” Opt. Express
13, 4237–4243 (2005). [CrossRef] [PubMed]
,
25
S. J. Charbonneau, R. Charbonneau, N. Lahoud, G. A. Mattuissi, and P. Berini, “Bragg gratings based on long-range surface plasmon-polariton waveguides: Comparison of theory and experiment,” IEEE J. Quantum Electron.
41, 1480–1491 (2005). [CrossRef]
] were reported recently in this context.
A double-electrode structure which consists of two metallic thin film had been reported as a possible waveguide structure supporting LRSPPs quite long time ago [
26
G. I. Stegeman and J. J. Burke, “Long-range surface-plasmons in electrode structures,” Appl. Phys. Lett.
43, 221–223 (1983). [CrossRef]
]. Stegeman
et al pointed out that LRSPPs on a double-electrode waveguide can be controlled more actively by using electro- or acousto-optic materials as a central dielectric medium because the pair of the metal film itself offers to supply DC electric field in the core dielectric medium. This structure has some other advantages over single-film LRSPP waveguides. As any homogeneous change in the core dielectric material does not break vertical symmetry of the structure, LRSPPs exist over much broader range of index contrast between the core and the cladding dielectric [
26
G. I. Stegeman and J. J. Burke, “Long-range surface-plasmons in electrode structures,” Appl. Phys. Lett.
43, 221–223 (1983). [CrossRef]
]. Therefore, the structure does not require a nearly free standing metal film for LRSPP based fluidic sensors, and it can include dielectric grating or photonic crystal structures in the core dielectric medium. In spite of those advantages, little attention has been paid to a double-electrode LRSPP waveguide. Dispersion relations and attenuation properties of LRSPPs excited on a double-electrode waveguide had been reported [
26
G. I. Stegeman and J. J. Burke, “Long-range surface-plasmons in electrode structures,” Appl. Phys. Lett.
43, 221–223 (1983). [CrossRef]
,
27
E. N. Economou, “Surface plasmons in thin films,” Phys. Rev.
182, 539–554 (1969). [CrossRef]
], but decisive questions for its practical applicability considering dependence of propagation lengths and field identities on operation frequency have not been addressed so far.
In this paper, we study frequency dispersive characteristics of LRSPPs on double-electrode waveguides. Dispersion curves and corresponding field identities of the modes are analyzed according to the structural parameters of core and cladding dielectrics. The analysis allows us to describe double-electrode LRSPPs in a general manner with the help of definition of a characteristic frequency by which different core-thickness dependencies are classified. Several interesting properties such as cut-off core-thickness at the vicinity of which the attenuation is largely reduced without any need of thinning metallic films, transiently saturated attenuation which can be used as a switching operation point, and the thickness-dispersionless flat-top LRSPP which can provide efficient adiabatic mode scaling are discussed on this basis. In section 2, dispersion relation of double-electrode waveguide structures is briefly derived. Dispersion curves in terms of frequency and effective-index of LRSPP modes are presented in Section 3, where dependence of the dispersion relation on the core-thickness is discussed for both low-index-core and high-index-core structures. The characteristic frequency that offers a criterion classifying the characteristic groups of LRSPP modes is described in Section 4. Field identities and attenuation properties at relative frequencies to ωc
are investigated in Section 5. Finally we conclude our analysis on double-electrode waveguide structures in Section 6.
Fig. 1. Schematic of a double-electrode slab waveguide structure and the shape of magnetic field profile for a symmetrically coupled LRSPP.
2. Dispersion relation of double-electrode waveguides
A double-electrode waveguide considered here is illustrated in
Fig. 1. It consists of two identical metal slabs with a dielectric constant
ε2
surrounded by core (
ε1
) and cladding (
ε3
) dielectrics. Thickness of the core dielectric and the metal film is
D and
t, respectively. In frequency domain, the magnetic field of a transverse magnetic (TM) polarized mode with propagation constant
β in this structure can be written in general as
where
in the region
y≥0. The cosh(
α1y
) or sinh(
α1y
) is taken for a symmetric or an anti-symmetric mode respectively. The field in the region
y≤0 is also provided by continuation of
Eq. (2), and the electric field is given by
E=-∇×
H/(
iωεm
) for
m=1, 2, or 3. The decay constant in each layer is given by
Requiring continuity of the tangential fields (
Hz
and
Ex
) at the interfaces leads to a dispersion relation and the amplitudes of a symmetric mode,
A, B, and
C in
Eq. (2), as follows.
Expressions for an anti-symmetric mode can be obtained by exchanging cosh and sinh to each other. The dispersion relation in
Eq. (4) gives a complex propagation constant of
β=βr+iβi
. Frequency dependent characteristics of
βr
provide a dispersion curve according to
D, and 1/(2
βi
) defines propagation length where the mode intensity is decayed down to its 1/
e value after propagation.
It is known that four modal solutions exist implicitly in
Eq. (4) as SPP modes for a double-electrode structure [
26
G. I. Stegeman and J. J. Burke, “Long-range surface-plasmons in electrode structures,” Appl. Phys. Lett.
43, 221–223 (1983). [CrossRef]
,
27
E. N. Economou, “Surface plasmons in thin films,” Phys. Rev.
182, 539–554 (1969). [CrossRef]
]. Each of them represents a combination out of symmetrically or anti-symmetrically coupled (sc- or ac-) long-range SPPs (LRSPPs) or short-range SPPs (SRSPPs). Internal loss is smaller in long-range SPP modes while larger in short-range SPP modes since they possess a node in the metal slabs. Each of the long-range and short-range SPPs splits again into two separate modes according to their symmetry referenced at the center of the core dielectric. Only a symmetrically-coupled long-range SPP, sc-LRSPP, mode whose magnetic field has a profile depicted schematically in
Fig. 1 will be considered hereafter, because an anti-symmetrically coupled long-range SPP may hardly be excited by direct end-fire coupling and even disappears when the core thickness is less than about the decay length of SPPs excited on a single-electrode waveguide [
26
G. I. Stegeman and J. J. Burke, “Long-range surface-plasmons in electrode structures,” Appl. Phys. Lett.
43, 221–223 (1983). [CrossRef]
].
Complex
β which satisfies
Eq. (4) is found by 2 stage processes. On the first stage, an approximate solution is given by observing values of left-hand side of
Eq. (4). Reciprocals of absolute square values of the left-hand side of
Eq. (4) are evaluated along real
β-axis within a specified domain of
β. A range
ε
3
1/2
k0
<
β<2
ε
3
1/2
k
0 was enough to completely contain the sc-LRSPP dispersion curve for an arbitrary core thickness over a frequency domain from DC to visible light. The reciprocal of absolute square value of the equation shows series Lorentzian-shaped peaks centered at Re{
β
sc-LRSPP} and Re{
β
sc-SRSPP}. The peak center position (
β0
) closest to
ε
3
1/2
k0
among them is recognized as an initial guess for
βr
of the sc-LRSPP. Its full-width at half-maximum (
Δβ) is assigned as an approximate value of
2βi
. On the second stage, Newton method starting at
β
(0)=
β
0+
iΔβ/2 is applied to find the exact complex
β for the sc-LRSPP until the propagation constant converges within |
β(
(n+1)-
β(
(n)
|/
k
0=±1.1×10
-16 during consecutive iterations. As
βi/βr
is an order of 10
-3 or less (
β is dominantly real), the first stage gives a well defined initial guess and provides a good repeatability in finding roots.
3. Symmetrically-coupled long-range surface plasmon-polaritons
Two different cases of low-index-core and high-index-core structures are analyzed to find out global features of sc-LRSPPs. Let’s specify, for simplicity, the metal slabs by a 20
nm-thick gold film and the refractive index of the core (cladding) dielectric by 1.45 (1.47) for the low-index-core structure and by 1.47 (1.45) for the high-index-core structure, as shown in the insets of
Fig. 2. Dispersion curves calculated by
Eq. (4) are shown in
Fig. 2 for the two core structures. The effective index of a sc-LRSPP mode is defined as a relative propagation constant of
neff
=
βr/k0
, and the frequency is normalized by the frequency of an electromagnetic wave whose vacuum wavelength is 1
µm. A Drude model fitted to give
ε
2=-118+11.58
i at 1550
nm vacuum wavelength is used for the frequency dependent dielectric constant of Au [
28
E. D. Palik, Handbook of Optical Constants of Solids II , (Academic Press, San Diego, 1998).
].
Figures 2(a) and
2(b) show frequency dispersive characteristics of the effective indices for the sc-LRSPP modes exited at the two double-electrode waveguide structures with low-index-core and high-index-core, respectively. The dispersion curves exist only in the shadow regions as
D increases from 0
µm to infinity, and those depicted in between the two asymptotic limits are of the finite core-thickness ranging from 0.2
µm to 6.0
µm in 0.2
µm steps. It is worth discussing first the two limits of no core dielectric (
D=0
µm) and infinitely thick core dielectric (
D→∞). The dispersion curves and modal properties for any other finite cores will be addressed in between these two limits as shown in
Fig. 2. For the limit case when
D=0
µm, the dispersion curves exist over the continuous range starting from the zero frequency for both core-index structures. This no-core limit is the same as the dispersion of LRSPP modes for a single-electrode waveguide with a metal thickness of 2
t (40
nm). On the other hand, the dispersion curves when
D→∞ have a characteristic frequency
ωc
at the point of 1.47 effective index. In the frequency range smaller than
ωc
there is no LRSPP mode for the infinite core thickness. Since when
D→∞ there is no coupling between the modes on the two metal waveguides, the dispersion curves shown above the
ωc
are the same as the dispersion of a asymmetric single-electrode waveguide composed by a metal film with thickness of
t (20
nm) and two different surrounding dielectrics with
n=1.45 and
n=1.47. In the case of the 20
nm-thick Au film sandwiched by two dielectrics with the refractive indices of 1.45 and 1.47,
ωc
is 1.09×2
πc/µm and the corresponding vacuum wavelength is 917.08
nm.
For a finite core-thickness in the low-index-core structure shown in
Fig. 2(a), there is no low frequency cut-off in the dispersion curve as
D increases up to 1.0
µm. For a core-thickness larger than 1.0
µm, on the other hand, low-frequency cut-off appears and it approaches
ωc
as
D increases. Therefore the core-thickness dependent characteristics of sc-LRSPP modes is quite different according to the relative position of the frequency to
ωc
. For a fixed frequency lower than
ωc
, as
D gets larger, the effective index decreases down to the cladding index of
n
3 (1.47) as
D approaches a particular value, Dcut-off, and the sc-LRSPP modes disappear when
D>
Dcut-off
. That is, there is a range of the core thickness that supports sc-LRSPPs when
ω<
ωc
. This behavior of the effective index is shown in
Fig 2(c). Dashed and dash-dotted line represents
D-dependent effective index at two different frequencies below
ωc
. Corresponding vacuum wavelength is 1550(
ω
1) and 1100(
ω
2)
nm, respectively. Please note that grey lines below 1.47 indicate not true solution but initial guesses given by peak positions of reciprocal value of the dispersion equation. There is no
β satisfying
Eq. (4) below effective index of 1.47. It is confirmed that the effective index decreases for increasing
D and finally disappear for
D>
Dcut-off. Dcut-off
is 1.340
µm for
ω
1 and 1.688
µm for
ω
2. For a core thickness just below
Dcut-off
, the mode size diverges to infinity since the decay constant in the cladding given by
α
3=[
β2-n3
2k02]
1/2 is about to vanish and therefore the decay length of the evanescent tail (1/
α
3) stretched into the cladding diverges to infinity. The energy dissipated as ohmic loss in the metal will vanish too as shown in
Fig 2(d). Therefore, it is expected that the propagation length will be largely extended just before the mode cut-off.
Dcut-off
becomes larger for a higher frequency and infinite when
ω=
ωc
. For
ω>
ωc
there is no cut-off core thickness and therefore no drastic extension in the propagation length is expected over the frequency range higher than
ωc
.
Fig. 2. The dispersion curves of sc-LRSPPs for two different cases of (a) low index core (n1=1.45, n
3=1.47) and (b) high index core (n1=1.47, n3=1.45). Two asymptotic limits of D=0 µm and D=∞ are indicated by two thick curves on the boundaries of grey regions, and the dispersion curves in the grey region are of the core thicknesses ranging from 0.2 µm to 6.0 µm in 0.2 µm step. Please note that the core is thicker as the curve approaches the boundary of the grey region marked by ‘D=∞’. (c) and (d) shows thickness dispersive characteristics of effective index and βi
for several frequencies below ωc
, respectively. Relative locations in the dispersion diagram for ω
1 and ω
2 are indicated by horizontal lines in (a).
On the other case of the high-index-core structure, frequency dependence of the sc-LRSPP modes is quite interesting as shown in
Fig. 2(b). All curves are crossing at a single position at (
n
1,
ωc
) regardless of the value of
D. When
ω=
ωc
the propagation constant of sc-LRSPP mode is non-dispersive on
D. Any cut-off core thickness does not exist over the full frequency range, so the internal loss of the sc-LRSPP modes never decreases extremely low. For
ω <
ωc
the effective index has a value less than
n
1 (1.47) and gradually approaches
n1 as
D increases, while for
ω >
ωc
dependence of the dispersion relation on
D is similar to the low-index-core case. The sc-LRSPP mode with a larger core thickness reduces to a fundamental mode of dielectric slab waveguide which consists of the same core and cladding dielectrics without metal. A physical origin of existence of the crossing point at (
n
1,
ωc
) in the high-index-core structure can be clearly found by requiring
Eq. (4) to be specified at the condition that
β=
n1k0
and
α
1=0. Under this specific condition, the dispersion relation is reduced to (
α3/ε3
)+(
α2/ε2
) tanh(
α2t)=0, and therefore it has no dependence on
D. It can be deduced that eigen frequencies of the sc-LRSPP modes which have the effective index equal to the core index must keep in constant to be
ωc
even for core-thickness variation.
Now we would like to summarize the frequency-dependent characteristics of sc-LRSPP modes excited on double-electrode structures as follows: In low-index-core case shown in
Fig. 2(a), a cut-off core thickness exists for sc-LRSPP modes at a frequency range less than the characteristic frequency
ωc
and mode size and propagation length become extremely extended just before the cut-off. For
ω >
ωc
, on the other hand, a sc-LRSPP mode can always be supported for arbitrary core thickness without a drastic extension of its mode size and propagation length. In low-index-core case shown in
Fig. 2(b), when
ω<
ωc
sc-LRSPP modes approach a fundamental mode of dielectric slab waveguide gradually from a single-electrode LRSPP mode as
D increases. At
ω=
ωc
the propagation constants of sc-LRSPP modes are non-dispersive and equal to
n
1
k
0 even for core-thickness variation. For
ω>
ωc
, the frequency-dependent characteristics are the same as the low-index case.
4. The characteristic frequency
The characteristic frequency ωc
can be derived from the dispersion relation when propagation constant has no dependence on D, (α3/ε3
)+(α2/ε2
) tanh(α2t)=0, as follow.
To extract the value of
ωc
in terms of the three dielectric constants and the metal thickness is not straightforward, but it is plausible to estimate its dependence on the index contrast, Δ
n=
n3-n1
, between the dielectric layers. If
Δn=0(viz.
ε1-ε3
=0),
ωc
=0 obviously. For a non-zero or even a large
Δn, intuitively
ωc
can be estimated from the dispersion diagrams in
Fig. 2(b). In fact, the reduced equation of (
α3/ε3
)+(
α2/ε2
) tanh(
α2t)=0 is exactly the same as the dispersion relation of LRSPPs excited on a single-electrode waveguide consisting of a single metal slab with thickness of 2
t surrounded by dielectric with
ε3. ωc
is the frequency of the LRSPP mode on this single-electrode waveguide when the relation
β=
n
1
k
0 satisfied.
Fig. 3. Dependences of the characteristic frequency on the index difference for several refractive indices of the cladding (a) and some thicknesses of the metal slab (b). The Au thickness is fixed at 20 nm for (a), and the refractive index of the cladding is fixed at 1.45 for (b) respectively.
Therefore, for the structures considered in
Fig. 2, ωc is a point corresponding to the effective index of
n
1. If
n
1 increases,
ωc
increases along the curve for
D=0
µm. For a large
Δn,
ωc
approaches a cut-off frequency of the single-electrode LRSPP, that is
ωc
→
ωp
/(1+
ε3
)
1/2.
Figure 3 shows dependence of
ωc
on the index contrast
Δn and the metal thickness
t. In
Fig. 3(a), the cladding index
n
3 is chosen from some typical dielectric materials like silica (1.45), LiNbO
3 (2.2142), and silicon (3.4764) taken in the Ref. 40 by interpolating at
λ
0=1550
nm, while the metal thickness is fixed at 20
nm. It can be confirmed that
ωc
converges to
ωp/(1+
ε3
)
1/2 for a large index contrast. In the order of
n
3=1.0, 1.33, 1.45, 2.2142, and 3.4764, the values of
ωp
/(1+
ε3
)
1/2 normalized by 2
πc/µm are 5.00, 4.25, 4.01, 2.91, and 1.95, respectively. In the same order of
n
3,
Δn that gives
ωc
corresponding to the vacuum wavelength of 1550
nm is 2.14×10
-3, 5.09×10
-3, 6.62×10
-3, 2.44×10
-2, and 1.30×10
-1, respectively. For
n
3=1.45, the
ωc
dependence on the metal thickness is shown in
Fig. 3(b). As
Δn increases,
ωc
converges to a constant value of 6.62×10
-3 even for the different thickness, but
ωc
grows more rapidly for a smaller
t. It is confirmed that
ωc
is tunable over broad spectrum from an optical communication to whole visible frequency range with acceptable index differences of 10
-3~1.
5. Propagation lengths and field profiles
Figures 4(a) and
5(a) show propagation length defined by 1/(2
βi
) for sc-LRSPP modes excited on the two double-electrode structures with low-index core and high-index core, respectively.
D varies from 0
µm to 8
µm. Three representative vacuum wavelengths (
λ
0) of 667
nm, 917 nm, and 1550
nm are selected to compare the propagation lengths for three different frequency regions of
ω<
ωc
,
ω=
ωc
(=3.27×10
14
rad./s),
ω>
ωc
, respectively. In case of low-index core shown in
Fig. 4(a), frequency characteristics of sc-LRSPP propagation length dependent on
D variation are clearly recognizable with the three black line curves. At
λ0
=1550
nm (
ω <
ωc
, dashed black line), the propagation length increases exponentially for a small core thickness up to 0.7
µm before the transient saturation at the vicinity of 1
µm, and it gradually diverges without any supported sc-LRSPP mode after 1.33
µm. The behaviors of the exponential increase and the transient saturation are also seen at
λ
0=917
nm (
ω=
ωc
, dash-dotted black line) bellow the 1.5
µm core-thickness. The two grey curves near the solid black line of 917
nm represent those for the frequencies adjacent to
ωc
, for instance,
ωc
-0.17×10
6
rad./s for the upper grey and
ωc
+2.07×10
6
rad./s for the lower grey line. At
λ
0=667
nm (
ω>
ωc
), on the other hand, the propagation length keep in constant for large
D variations. It is noted that, at
ω=
ωc
, the propagation length does not go to infinity abruptly after the transient saturation but increases exponentially and that the propagation behavior suffers from a sudden divergence or a bound limit even for a small amount of frequency deviation from
ωc
.
Fig. 4. (a). Dependences of the propagation lengths of sc-LRSPPs on the core thickness for the case of low index core at several wavelengths. The vacuum wavelengths are selected as representatives for ω<ωc
(dashed line), ω=ωc
(black solid line) and ω>ωc
(dash-dotted line). Each of the upper and the lower grey solid curve represents just below and over the characteristic frequency respectively, and the corresponding differences from ωc
are -0.17 and +2.07×106
rad./s. (b) Field profiles (Hz
) of a sc-LRSPP at ω=ωc
for the case of low index core. The inset shows the profiles normalized by its value at the core-metal interface in the metal film. In the inset, grey region indicates the metal film.
A reason exhibiting the transient saturation near the core thickness of about wavelength scale can be found from transition property of the mode profiles as
D varies.
Figure 4(b) shows the normalized
Hz
components of sc-LRSPP modes when
D=0.3, 1.0, and 4.0
µm, where the surface normal position at 0
µm means the center of the core dielectric and those at the peaks indicate the positions of the metal slabs. The inset in
Fig. 4(b) is the magnified
Hz
profiles around the 20
nm-thick metal embedded in between the core (
n
1) and the cladding (
n
3). All the mode profiles reveal necessarily exponential decay in the cladding and hyperbolic-cosine distribution in the core, but the ratio of the field amplitudes at the two metal-dielectric interfaces is quite different with each other. At
D=0.3
µm, before the transient saturation in propagation length, the field amplitude is smaller at the metal-core interface and it grows up inside the metal layer as approaching the metal-cladding interface. At
D=1.0
µm within the transient saturation region or at
D=4.0
µm after it, on the other hand, the field amplitude, decreases down toward the metal-cladding interface. This difference in the field profiles can be understood by a transition property of the mode profiles between two limiting cases of
D=0 and
D→∞: When the core thickness shrinks to be zero the field profiles on two metal slabs combine together to make a single hyperbolic-cosine form, therefore the field amplitudes on both sides of the metal-dielectric interfaces become higher than that in the middle of the metal layer. If the core thickness is large enough not to allow any field coupling between the two metals, the field in each metal slab has a larger amplitude on the metal boundary facing the dielectric with a lower refractive index since the evanescent tail of the field profile is stretched further into the lower dielectric.
In case of the high-index core shown in
Fig. 5, change in the propagation length as
D varies is moderate with no drastic increase for all three wavelengths. (Note that the vertical scale is linear, not logaritmic as in
Fig. 4(a)). For
λ
0=1550
nm the propagation length decreases gradually down to a minimum at
D=3 µm before its increase. The normalized
Hz
profiles for
D=0.3, 3.0, and 8.0 µm depicted in
Fig. 5(b) imply that the energy portion localized in the core region becomes larger as increasing
D and the mode approaches to a fundamental mode of dielectric slab waveguides. Also, the field amplitudes inside the metal layer always grow up as approaching the metal-cladding interface as shown in the inset of
Fig. 5(b).
Fig. 5. (a). Dependences of the propagation lengths of sc-LRSPPs on the core thickness for the case of high index core at several wavelengths. The vacuum wavelengths are selected as representatives for
ω<
ωc
(dotted line),
ω=
ωc
(solid line) and
ω>
ωc
(dash-dotted line). Please note that, differently from
Fig. 4(a), the vertical axis is not logarithmic but linear scale. (b), (c), and (d) shows the field profiles (Hz) of sc-LRSPPs for several core thicknesses at
ω<
ωc
,
ω=
ωc
, and
ω>
ωc
respectively. The inset in each figure shows the profiles normalized by its value at the core-metal interface in the metal film (grey region).
For
λ
0=917
nm (
ω=
ωc
), dependence of the propagation length on
D and evolution of the field profiles are quite interesting as depicted in
Figs. 5(a) and
5(c), respectively. The propagation length is linearly proportional to
D. Regardless of the core thickness, the field amplitude is uniform over the whole core region and the normalized field profiles around the metal layers are exactly matched with each other as shown in the inset of
Fig. 5(c). These distinctive characteristics at
ω=
ωc
are caused by the fact that
β=
n
1
k
0 thus
α
1=0 for arbitrary core thickness. The non-dispersive property of the propagation constants on
D also insists that
α
2 and
α
3 must be constant, therefore, not only the decay lengths to the core and cladding dielectrics but also the amplitudes of the exponents given in
Eq. (5) should be constant. It must be emphasized that this particular property is very desirable for adiabatic coupling between optical modes with different lateral sizes via waveguide tapering. Adiabatic coupling is reasonable under the condition that the eikonal parameter defined by tan(
ϕ)[
d(1/
βr
)/
dD] should be much less than 1, where
ϕ is a taper angle [
29
M. I. Stockman, “Nanofocusing of optical energy in tapered plasmonic waveguides,” Phys. Rev. Lett.
93, Art. No. 137404 (2004). [CrossRef] [PubMed]
]. The eikonal parameter is zero for sc-LRSPP modes with the characteristic frequency, resulting that a very large tapering angle can be acceptable for an efficient mode coupling. Moreover, the flat-top profiles of the field amplitudes in the core are sustained even for a large core-thickness variation. This may offer a noble method for flat-top beam shaping of light radiated from a conventional waveguide or a point-like source.
Finally over a higher frequency range of
ω>
ωc
, core-thickness dependence of the propagation length in
Fig. 5(a) and the mode profiles in
Fig. 5(d) are not so different from those of the low-index-core case. The modes converge to a decoupled LRSPP mode from a sc-LRSPP as increasing
D with no distinct feature in between.
It may be worth considering whether such a crossing in dispersion curves as shown in
Fig. 2(b) also takes place in a two dimensional guiding structure, like a cylindrical plasmonic waveguide consisting of a metal coated high-index dielectric wire surrounded by a low-index cladding [
30
U. Schröter and A. Dereux, “Surface plasmon polaritons on metal cylinders with dielectric core,” Phys. Rev. B
64, Art. No. 125420 (2001). [CrossRef]
]. Actually the SPP mode represented by
n=1 in a higher branch of dispersion curve of the cylindrical plasmonic waveguide, where
n is the number of azimuthal oscillation, also has a crossing point near the core light line. According to Ref. 30, dispersion curves for higher branch of
n=1 modes cross core light line and curves for different core diameter intersect with each other. Therefore, it can be concluded that there exists flat-top modes (
α
1=0) and nearly non-dispersive characteristics on core diameter. But unfortunately, we have found that the crossing point does not always correspond to the core light line after an intuitive consideration as follow: Noted that one of the physical origins of the non-dispersive characteristic at
ω=
ωc
results from the separable property of the exponential functions in
Eqs. (4) and
(5) so as to satisfy exp[±
α
2(
D+t)]=exp(±
α2D)exp(±
α2t). These separable functions guarantee to eliminate the term exp(±
α2D) in
Eqs. (4) and
(5), and only the exponents having ±
α1D argument among the
D dependent terms survive before they disappears when
α
1=0 at
ω=ωc
. However, such elimination is not possible in a cylindrical plasmonic waveguide because its mode expression does not consist of exponential functions any more but modified Bessel functions of second kind. Therefore, a flat-top mode exists in a cylindrical plasmonic waveguide, but it does not always have non-dispersive characteristic on core diameter. Though the two noble properties do not always coincide with each other, it will be interesting to find out other 2-dimensional structures that support efficient flat-top collimation or coupling between optical source and
µm-scale waveguide.
6. Conclusion
For double-electrode plasmonic waveguide structures, modal properties of the sc-LRSPPs have been described comprehensively by investigating frequency dependence of the dispersion relation and the mode profiles consistently. We have found out that the sc-LRSPP modes excited on a double-electrode plasmonic waveguide can be classified by four groups of distinctive characters according to their dependence on the core thickness
D. For the low-index-core structure, the propagation length can be largely extended at a particular core thickness of wavelength scale at a frequency range lower than the characteristic frequency
ωc
which depends on the index difference in the core-cladding layers and the thickness of the metal. The characteristic of drastic increase in the propagation length may be used not only for refractrometric sensors or non-linear ATR devices but also variable optical attenuators by incorporating media for active index modulation as a core dielectric [
19
S. Park and S. H. Song, “Polymeric optical attenuator based on long range surface plasmon polaritons,” Electron. Lett.
42, 402–404 (2006). [CrossRef]
]. For a frequency range higher than
ωc
, the sc-LRSPP mode converts its modal property moderately from symmetrically coupled to decoupled as
D increases. For the high-index-core structure, on the other hand, there is no cut-off core thickness in overall frequency range. For
ω<
ωc
the sc-LRSPPs with a large
D converge to a mode of conventional dielectric slab waveguides, but at
ω=ωc
the sc-LRSPP modes are completely non-dispersive for change in
D. Regardless of
D alteration, the modes at
ω=ωc
consistently maintain a perfect flat-top profile in the core region and identical decay tails in the cladding. 2-dimensional extension of this prominent characteristic may provide us a novel means for uniform dimension-scaling and flat-top profile-shaping of an optical mode tightly guided by plasmonic waveguides. Thus it will be worth to find out conditions for flat-top modes in a cylindrical waveguide with metal coated dielectric core. The flat-top sc-LRSPP modes may excite an active medium sandwiched in between the metal layers very uniformly, therefore it will be interesting to implement such a non-dispersive flat-top mode for nonlinear applications of SPP waveguides.