Nonlinear Waveguides
Nonlinear Waveguides
Nonlinear Waveguides
This article has been downloaded from IOPscience. Please scroll down to see the full text article.
(http://iopscience.iop.org/1464-4258/11/11/114032)
View the table of contents for this issue, or go to the journal homepage for more
Download details:
IP Address: 93.180.53.211
The article was downloaded on 28/06/2013 at 10:49
Abstract
The first comprehensive exact theory of strongly nonlinear guided waves in a double-negative
planar metamaterial waveguide is developed. The theoretical consequences are that novel
surface and guided waves are predicted because of the special relationship of the boundary
fields to each other. The analysis leads smoothly to tunability with power and direct access to
group velocity control.
2
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
(a) (b)
Figure 2. Example shapes of (a) two order 1 amplitude profiles with the same frequency given by the two dispersion curves mentioned in the
text. (b) An example of an order 0 guided wave amplitude profile.
The outcome is where Z 01 = sc−1 [E 0 /b1 | m (1) ], m (1) = (a12 − b12 )/a12 and
C2 = (μ22 κ12 − κ22 + 2 μ2 E 02 )E 02
= (μ22 κ32 − κ22 + 2 μ2 E b2 )E b2 . (4) a12 = (κ22 + κ24 − 4 2 |μ|C2 )/2 2 |μ|,
Note that, since μ2 is the only permeability term to appear
in (4), the subscript will not be used in rest of the paper, in b12 = (κ22 − κ24 − 4 2 |μ|C2 )/2 2 |μ|.
order to make the presentation less cumbersome.
Equation (4) also gives the relationship between the values In order to prevent the solutions being merely reflections
of the amplitudes at the two film boundaries [2, 3]. For a in the Z -axis of the other solutions, in the calculations that
particular value of E 02 , equation (4) indicates that in general follow E 0 is always considered to be positive [3]. Equation (6)
there are two values of E b2 . However to keep E b real only those therefore has a negative slope for all values of and thus
values where E b2 > 0 are considered [2]. gives order 1 surface waves (for example see figure 2). Since
Solutions to (3a ) and (3c) are simply E 1 = E 0 eκ1 Z , these waves are of order 1, the values of E b must be negative.
E 3 = E b eκ3 (D−Z ) . The solutions to (3b ) can be found by, prior Equation (4) then shows that there are two solutions for each
to integration, expressing (3b ) in the form that satisfy the dispersion equation i.e. two dispersion curves
⎡ ⎤ 12 for each value of E 0 (see for example figure 2(a)).
∂ E2 κ 2
2 + κ 2
4
− 4 2 |μ|C
2 (ii) Guided wave solutions. κ22 < 0, C2 > 0.
= |μ| 2 ⎣ E 22 + ⎦
1
In this case for E 2 < b3 the solution is:
∂Z 2 2 |μ|
1
⎡ ⎤ 12 E 2 = b3 sn(Z 03 − a3 |μ| 2 Z | m (3) ), (7)
κ22 − κ24 − 4 2 |μ|C2
× ⎣ E 22 + ⎦ . (5) where Z 03 = sn−1 [E 0 /b3 |m (3) ] or, if E 2 > b3 ,
2 2 |μ|
1
E 2 = a3 ns(Z 03b − a3 |μ| 2 Z | m (3) ), (8)
If the factors on the right-hand side of (5) are to be kept
real, the solutions can be written in terms of the Jacobi where Z 03b = ns−1 [E 0 /a3 |m (3) ]. In both cases, m (3) = b32 /a32
functions sc( p|m), ns( p|m), sn( p|m) and ds( p|m) where and
p is the relevant real argument and m the relevant Jacobi
a32 = (κ̃22 + κ̃24 − 4 2 |μ|C2 )/2 2 |μ|,
parameter [12]. The solutions then fall into three categories
depending on the signs of C2 and κ22 . These categories are
(using m (i) , i = 1, 2, 3 for the Jacobi parameters): b32 = (κ̃22 − κ̃24 − 4 2 |μ|C2 )/2 2 |μ|.
(i) Surface wave solutions. κ22 > 0, C2 > 0.
In this case, a solution to the nonlinear wave equation is Equation (7) gives guided wave solutions of order n =
1
1, 2, . . .. Equation (8) gives an order 0 guided wave solution
E 2 = b1 sc(Z 01 − a1 |μ| 2 Z | m (1) ), (6) (for example see figure 2(b)).
3
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
4
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
(a) (b)
Figure 4. (a) Magnified view of two order 0 nonlinear dispersion curves for D = 1. The intersections of the line = 0.512 on the curve that
is very close to the linear curve ( E 0 = 0.0001) are labelled C and D. Labels on the lines are values of E 0 , and the dashed curve is the cut-off
κ1 = 0. The inset shows the field profiles at the maximum and minimum points on the total power flow in (b). Nonlinear curves with values of
E 0 > 0.425 have no intersection with the line = 0.512. (b) Power curves for the range 0.8 < κx < 1. The heavy curve is the total power
flow, the dashed curve is the power flow in the film and the light curve is the value of E 0 .
respectively. The flows in the substrate and cladding are always μ2 = μ2 (ω) and ε2 = ε2 (ω) that has been widely used in the
positive. In the film the flow is always negative due to the literature [7, 14–18]. Losses are not taken into account, for the
presence of μ in the denominator of the first term in the bracket reasons stated very early on, so that both the linear permittivity
in (14). The total power flow is then either positive, zero or and permeability are real and are
negative depending on the relative values of these three terms.
It should be noted that E 2 , E 0 and E b have an implicit factor ε2 (ω) = 1 − ωp2 /ω2
α . To get the limit of linear power flow this must be cancelled (15)
with the factor α in the denominator of P0 . As α approaches μ(ω) = μ2 (ω) = 1 − Fω2 /(ω2 − ω02 )
zero, E 0 approaches zero, so in the results presented below,
in which ωp is a plasma frequency and ω0 is a frequency
this latter condition is regarded as approaching the linear case.
associated with the magnetic resonance. It is not
The difference in the linear results and the nonlinear results are
expected that adopting a Drude model will produce any
indistinguishable on the scale of the figures when E 0 ≈ 0.001.
qualitative changes to the upcoming results [18]. The
The numerical value of P0 ≈ 8 × 105 /αωp , therefore the
values ω0 /ωp = 0.4, F = 0.56 are also used, without losing
possibility of the nonlinear effects being significant depends
generality. It will also be assumed that the nonlinearity is of
not only on the Kerr coefficient α , but also on the value of ω p .
the Kerr form.
This in turn determines the relationship between the value of z
and that of Z . As an example, taking α = 1.25 × 10−11 [4] and
a wavelength in the visible light range, say 700 nm, together 4.1. The narrow film D = 1
with ω/ωp = 0.5, ωp = 5.39 × 1015 yields P0 ≈ 12 W m−1 , or
about 0.12 W cm−1 . The thicknesses of the film in the results The numerical investigations begin with this thin film case
below would then be approximately 56 × D nm. for two reasons. Firstly, it allows comparison with published
results for the linear case [14] and, secondly, it gives a clear
insight into the effects of increasing D systematically.
4. Numerical investigations
The above analysis is general but in the specific Jacobi 4.1.1. Order 0 waves. The dispersion curves for the order
functions given in equations (6)–(9), the values of μ and ε2 0 waves when D = 1 are shown in figure 3. The curves are
are considered to be negative. Simple solutions exist if μ labelled with values of E 0 . The dotted curve κ2 = 0 shows the
and ε2 are constant but these need not be considered here. curve that is crossed in going from surface modes to guided
For the numerical investigations, a familiar model is used for modes. The figure prompts the following observations.
5
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
(a) (b)
Figure 5. Total power flow (heavy lines), power flow in the film (dashed lines) and the values of E 0 (fine line) for nonlinear order 0 waves
when D = 1. The vertical lines at the left-hand edges are close to the κ1 = 0 cut-off. The lines are labelled with values of frequency .
(a) Above a critical value of E 0 (see text) there is a point where the total power flow is zero, and there is also a linear limit ( E 0 = 0), (b) below
this critical value there is no point where the total power flow is zero and there is no linear limit.
(a) (b)
Figure 6. A sample set of lower order 1 nonlinear dispersion curves for D = 1. The dashed curves are as in figure 3. (a) The linear dispersion
curve and the high E 0 value nonlinear curves. (b) The low value nonlinear curves are each raised by an increment 0.2 referred to the κx -axis to
help distinguish them from the linear curve.
(i) In order to keep the permittivity and permeability nonlinear curves, this bend-back becomes more pronounced
negative, the normalized frequency range 0.4 < < 0.6 must and the value of κx at the turning point increases, i.e. moves
be adopted. The linear mode 0 dispersion curve, which lies in further away from the cut-off line. The bend-back is of major
the upper part of this segment, displays a ‘bend-back’ close importance in the nonlinear case since it suggests that control
to the cut-off line κ1 = 0 (figures 3 and 4(a)) [19]. For the of the direction of power flow may be possible. This is because
6
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
Figure 7. Samples of the upper order 1 nonlinear dispersion curves Figure 9. The three order 1 NL profiles for = 0.46, κx = 2.8. The
for D = 1. The dashed lines are as in figure 3. The dotted arrow letters correspond to the points labelled in figure 8.
shows the path of the curves as E 0 increases.
7
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
(a) (b)
Figure 10. Dispersion curves for D = 3. The dotted lines are as defined in figure 3. (a) The order 0 curves labelled with values of E 0 . The
order 2 guided wave for E 0 = 0.001 is shown, labelled O2. (b) The order 1 curves labelled with values of E 0 . The bracketed letters refer to
lower set curves (L) and upper set curves (U).
(a) (b)
Figure 11. (a) Total power flow for D = 3, = 0.4725. The insets are the profiles at the maximum and minimum of the power curve.
(b) The values of E 0 (dotted) and the power flow in the film (full line). The points where E 0 = 0 are the linear limits.
(ii) At frequencies below that where the linear mode critical value there is always a value of κx where the total power
0 dispersion curve is cut off, nonlinear dispersion curves flow is zero, again indicating the possibility of control.
continue to display a bend-back until another critical value Above this second critical value of E 0 the total power flow
of E 0 is reached (compare the curves for E 0 = 2.4 and 4 in is always positive and the values of where the dispersion
figure 3). As shown in figure 5, when E 0 is below this second equation is satisfied have no linear solutions, i.e. do not cut
8
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
9
J. Opt. A: Pure Appl. Opt. 11 (2009) 114032 A D Boardman and P Egan
(a) (b)
Figure 13. (a) Power flow for D = 10, κx = 1.25. At this value the frequency determines the direction of total power flow. (b) Power flow for
= 0.48 on the nonlinear order 0 dispersion curve labelled E 0 = 0.5 in figure 12. For this value the direction of power flow is determined
by the value of κx . In both figures E 0 is the fine line, the power flow in the film is the dotted line and the total power flow is the heavy line.
10