1: \documentclass[aps,pre,amsmath,showpacs]{revtex4}
2: \usepackage{graphics}
3: \usepackage{epsfig}
4: \begin{document}
5: \title{Embedded solitons in the third-order nonlinear Schr\"{o}dinger equation }
6: \author{Debabrata Pal, Sk. Golam Ali }
7: \author{B. Talukdar}
8: \email{binoyt123@rediffmail.com}
9: \affiliation{Department of Physics, Visva-Bharati University,
10: Santiniketan 731235, India}
11: \begin{abstract}
12: We derive a straightforward variational method to construct embedded soliton solutions of the third-order nonlinear Sch\"odinger equation and analytically demonstrate that these solitons exist as a continuous family.
13: We argue that a particular embedded soliton when perturbed may always relax to the adjacent one so as to make it fully stable.
14: \end{abstract}
15: \pacs{42.65.Tg; 42.81.Dp; 05.45.Yv}
16: \maketitle
17:
18: \section{ Introduction}
19: The distortion of pulses propagating through waveguides arises due to dispersive spreading. However, if the waveguides is made using materials the refractive index of which depends nonlinearly on the intensity of the propagating pulses, it will be possible to compensate the effect of dispersion by a pulse-narrowing effect \cite{1}. A typical example in respect of this is provided by optical fibers. Here the weak nonlinearity
20: of the index of refraction arising due to Kerr effect produces a self modulation which in turn causes steepening of the wave. The dynamical interplay between the dispersive and narrowing effects produces the so called optical solitons. For an isotropic medium with cubic nonlinearity the complex optical wave field $\phi(x,t)$ is governed by the nonlinear Schr\"{o}dinger equation (NLS)
21:
22: \begin{equation}
23: i\phi_x+\phi_{2t}+|\phi|^2\phi=0, \,\,\,\, \phi=\phi(x,t).
24: \end{equation}
25: Here $t$ stands for time and $x$, the coordinate along the fiber. The suffixes $x$ and $t$ of $\phi$
26: denote partial derivatives with respect to these variables. In particular, $\phi_{2t}=\dfrac{\partial^2\phi}{\partial t^2}$.
27: The main reason for the robustness of the solitons described by $(1)$ is that the wave numbers of the solitons lie in a range that is forbidden for the linear dispersive waves \cite{2}. As a result, linear wave
28: can not be in resonance with the soliton to receive energy from the latter. We then have a stable soliton.
29: \par
30: The propagation of picosecond pulses is well described by the NLS equation in $(1)$ which accounts for only the second-order dispersion and self modulation. But for femtosecond pulses another important physical effect namely, the third-order dispersion comes into play. In this case the appropriate evolution equation for the pulse propagation is given by \cite{3}
31: \begin{equation}
32: i\phi_x+\phi_{2t}+|\phi|^2\phi=i\beta\phi_{3t}.
33: \end{equation}
34: Understandably, the parameter $\beta$ is a measure of the perturbation caused by the third-order dispersion. Due to the effect of perturbation a soliton could have wave numbers residing inside the linear spectrum of the system. Such a soliton is called an embedded soliton \cite{4}. Clearly, the embedded solitons can be in resonance with the perturbation to generate continuous wave radiation for subsequent decay. Despite this, in certain parameter regimes these solitons can be stable \cite{5}. In view of this the stability analysis of embedded solitons arising in the solutions of higher-order NLS equation has been subject of considerable current interest. The present work is an effort in this direction.
35: \par
36: Studies in embedded solitons \cite{6} within the framework of a third-order NLS equation indicate that these soliton exits as a continuous family. The members of the family being characterized by their propagation velocities are connected through a gauge transformation \cite{3}. We demonstrate the existance of such a family
37: by using a straightforward analytical model and argue that embedded solitons can be fully stable despite their tendency to shed radiation. We believe
38: that such an approach has distinct advantages to deal with physical problems because many unknown effects are then
39: readily expressed and evaluated. To achieve our goal we shall derive a variational approach involving trial functions in order to describe the main characteristics of the embedded solitons as determined by the third-order
40: NLS equation in $(2)$. Although the chosen trial functions have a specific form , the shape parameters are allowed to evolve as the solitons propagate. With this assumtion the evolution equation under consideration simplifies to a reduced Lagrangian problem \cite{7}. We shall try to study the stability of embedded
41: solitons
42: by taking variations of this Lagrangian with respect to the shape parameters.
43:
44: \section{ Variational description of the third-order NLS equation}
45: For the NLS equation in $(1)$ there exists a well-defined spectral problem \cite{8} such that one can write a closed
46: form analytical solution of it in terms of sech functions. As opposed to this, the third-order NLS equation can not be solved
47: analytically. Keeping this in view we look for a variational treatment of $(2)$ to find an accurate approximation solution. For our treatment we will essentially rely on a Ritz optimization procedure \cite{9}
48: applied to the Lagrangian function for the NLS equation in $(2)$. We have found that the action functional
49: \begin{equation}
50: W=\int\int{\cal L}\left( \phi,\phi^{\star},\phi_x,\phi_x^{\star},\phi_t,\phi_{2t},\phi_t^{\star},\phi_{2t}^{\star}\right) dxdt
51: \end{equation}
52: with the Lagrangian density
53: \begin{equation}
54: {\cal L}=\frac{1}{2}\left( \phi\phi_x^{\star}-\phi^{\star}\phi_x\right)-\frac{1}{2}\phi^2{\phi^{\star}}^{2}
55: +\phi_t\phi_t^{\star}+\frac{i}{2}\beta\left( \phi_t\phi_{2t}^{\star}-\phi_t^{\star}\phi_{2t}\right)
56: \end{equation}
57: reproduces the third-order NLS equation via the Hamilton's variational principle.
58: \par
59: In the Ritz optimization procedure, the first variation of the action functional is made to vanish for a suitable
60: chosen trial function. In analogy with the well-known solution of $(1)$ we introduce a sech type trial function
61: %$\begin{widetext}
62: \begin{eqnarray}
63: \phi(x,t)=\eta(x) \text{sech}\left[ \frac{t-y(x)}{a(x)} \right]\times \,\,\,\,\,\,\nonumber \\\exp i\left[ V(x)(t-y(x))+\frac{b(x)}{2a(x)}\left( t-y(x)\right)^2+\sigma(x) \right] .
64: \end{eqnarray}
65: %$\end{widetext}
66: Here $\eta$, $y$, $V$, $a$, $b$, and $\sigma$ are real functions of $x$. The parameters $\eta$, $y$ and $a$ are related to the three lowest-order moments of the $\phi$ envelope and represent
67: respectively its amplitude, central position and width respectively. The other parameters $\sigma$, $V$ and $b$ stand for the phase, velocity (centre of the soliton) and frequency chirp. Understandably, these parameters will all vary with the distance of propagation. Using $(5)$ in $(4)$ we get
68: \begin{equation}
69: {\cal L}_s=\sum_{i=1}^{4}{\cal L}^{(i)}_{s},
70: \end{equation}
71: where
72:
73: \begin{widetext}
74: \begin{subequations}
75: \begin{equation}
76: {\cal L}_s^{(1)}=\eta^{2}\left\lbrace a\frac{dV}{dx}\left( \frac{t-y}{a}\right)-V\frac{dy}{dx}+\frac{1}{2}\left( a\frac{db}{dx}-b\frac{da}{dx}\right) \left( \frac{t-y}{a}\right)^2-b\frac{dy}{dx} \left(\frac{t-y}{a}\right)+\frac{d\sigma}{dx} \right\rbrace sech^2\left(\frac{t-y}{a}\right),
77: \end{equation}
78: \begin{equation}
79: {\cal L}_s^{(2)}= -\frac{1}{2}\eta^{4}sech^4\left(\frac{t-y}{a}\right),
80: \end{equation}
81: \begin{equation}
82: {\cal L}_s^{(3)}=\eta^{2}\left[ \frac{1}{a^2}tanh^2\left(\frac{t-y}{a}\right)+ \left\lbrace V+\frac{b(t-y)}{a}\right\rbrace ^2\right]sech^2\left(\frac{t-y}{a}\right)
83: \end{equation}
84: and
85: \begin{eqnarray}
86: {\cal L}_s^{(4)}=-\beta\eta^{2}{\bigg\lgroup} \frac{b}{a^2}tanh\left(\frac{t-y}{a}\right)-\frac{1}{a^2}\left\lbrace V+\frac{b(t-y)}{a}\right\rbrace tanh^2\left(\frac{t-y}{a}\right)\nonumber\\-\frac{1}{a^2}\left\lbrace V+\frac{b(t-y)}{a}\right\rbrace sech^2\left(\frac{t-y}{a}\right)-\left\lbrace V+\frac{b(t-y)}{a}\right\rbrace^3 {\bigg\rgroup}sech^2\left(\frac{t-y}{a}\right) .
87: \end{eqnarray}
88: \end{subequations}
89: \end{widetext}
90: Here the subscript $s$ on ${\cal L}$ indicates that we have inserted the $sech$ ansatz for $\phi(x,t)$ into the Lagrangian density. In terms of $(6)$ the variational principle implied by $(3)$ leads to
91: \begin{equation}
92: \delta\int\left\langle L\right\rangle dx=0
93: \end{equation}
94: with
95: \begin{equation}
96: \left\langle L\right\rangle =\int_{-\infty}^{+\infty}{\cal L}_s dt.
97: \end{equation}
98: We have found that result for $\left\langle L\right\rangle$ is given by
99: \begin{widetext}
100: \begin{eqnarray}
101: \left\langle L\right\rangle=a\eta^{2}{\bigg\lgroup}-2V\frac{dy}{dx}+\frac{\pi^2}{12}\left( a\frac{db}{dx}-b\frac{da}{dx}\right) +\frac{d\sigma}{dx}-\frac{2}{3}\eta^2+\frac{2}{3a^2}+2V^2+\frac{\pi^2b^2}{6}+\beta V\left( \frac{2}{a^2}+2V^2+\frac{\pi^2b^2}{2}\right) {\bigg\rgroup}.
102: \end{eqnarray}
103: \end{widetext}
104: \section{ Parameters of the trial function}
105: From the vanishing conditions of the variationals $\frac{\delta \left\langle L\right\rangle}{\delta \sigma}$,
106: $\frac{\delta \left\langle L\right\rangle}{\delta \eta}$, $\frac{\delta \left\langle L\right\rangle}{\delta y}$,
107: $\frac{\delta \left\langle L\right\rangle}{\delta a}$, $\frac{\delta \left\langle L\right\rangle}{\delta b}$ and
108: $\frac{\delta \left\langle L\right\rangle}{\delta V}$
109: we obtain the following equations.
110: \begin{widetext}
111: \begin{subequations}
112: \begin{equation}
113: \frac{d}{dx}\left(a\eta^2 \right)=0,
114: \end{equation}
115: \begin{eqnarray}
116: -2V\frac{dy}{dx}+\frac{\pi^2}{12}\left( a\frac{db}{dx}-b\frac{da}{dx}\right) +\frac{d\sigma}{dx}-\frac{4}{3}\eta^2+\frac{2}{3a^2}+2V^2+\frac{\pi^2b^2}{6}+ \beta V\left( \frac{2}{a^2}+2V^2+\frac{\pi^2b^2}{2}\right)=0,
117: \end{eqnarray}
118: \begin{equation}
119: \frac{d}{dx}\left( 2a\eta^2V\right) =0 ,
120: \end{equation}
121: \begin{eqnarray}
122: -2V\frac{dy}{dx}+\frac{\pi^2}{12}\left( a\frac{db}{dx}-b\frac{da}{dx}\right) +\frac{d\sigma}{dx}-\frac{2}{3}\eta^2+\frac{2}{3a^2}+2V^2+\frac{\pi^2b^2}{6}+ \nonumber \\ \beta V\left( \frac{2}{a^2}+2V^2+ \frac{\pi^2b^2}{2}\right) +a\left( \frac{\pi^2}{12}\frac{db}{dx}-\frac{4}{3a^3}-\frac{4\beta V}{a^3}\right) +\frac{\pi^2}{12\eta^2}\frac{d}{dx}\left( a\eta^2b\right) =0,
123: \end{eqnarray}
124: \begin{equation}
125: a\left( -\frac{1}{12}\frac{da}{dx}+\frac{b}{3}+\beta Vb\right)- \frac{1}{12\eta^2}\frac{d}{dx}\left( a^2\eta^2\right) =0
126: \end{equation}
127: and
128: \begin{equation}
129: \frac{dy}{dx}-2V-\beta\left( \frac{1}{a^2}+3V^2+\frac{\pi^2b^2}{4}\right) =0.
130: \end{equation}
131: \end{subequations}
132: From $(11a)$
133: \begin{subequations}
134: \begin{equation}
135: a\eta^2={\rm{constant}}={\rm{E}}_0 ({\rm{say}}) .
136: \end{equation}
137: Use of $(12a)$ in $(11c)$ gives
138: \begin{equation}
139: V={\rm{constant}} .
140: \end{equation}
141: Again $(12a)$ and $(11d)$ give
142: \begin{eqnarray}
143: -2V\frac{dy}{dx}+\frac{\pi^2a}{4}\frac{db}{dx}-\frac{\pi^2b}{12}\frac{da}{dx}+\frac{d\sigma}{dx}-\frac{2}{3}\eta^2-\frac{2}{3}a^2+2V^2+\frac{\pi^2b^2}{6}+\beta V\left( 2V^2+\frac{\pi^2b^2}{2}-\frac{2}{a^2}\right) =0
144: \end{eqnarray}
145: Similarly from $(12a)$ and $(11e)$ we have
146: \begin{equation}
147: \frac{da}{dx}=r b,\,\,r=2+6\beta V.
148: \end{equation}
149: \end{subequations}
150: \end{widetext}
151: Subtracting $(11b)$ from $(12c)$ we write
152: \begin{equation}
153: \frac{db}{dx}=\frac{4}{\pi^2a}\left( \frac{r}{a^2}-\eta^2\right) .
154: \end{equation}
155: From $(12a)$, $(12d)$ and $(13)$ we get a second-order differtial equation for the pulse width
156: \begin{equation}
157: \frac{d^2a}{dx^2}=\frac{4r}{\pi^2}\left( \frac{r}{a^3}-\frac{{\rm{E}}_0}{a^2}\right) .
158: \end{equation}
159: The first integral of $(14)$ gives
160: \begin{equation}
161: \frac{1}{2}\left( \frac{da}{dx}\right)^{2}+\Pi(a)=0
162: \end{equation}
163: where the potential field
164: \begin{equation}
165: \Pi(a)=\frac{2r^2}{\pi^2a^2}-\frac{4r{\rm{E}}_0}{\pi^2a}+K.
166: \end{equation}
167: The constant of integration $K$ is determined by the initial conditions of $(14)$.
168: \section{Potential function formulation for the dynamics of embedded solitons}
169:
170: First from the initial conditions of $(14)$ written as $a(x)|_{x=0}=a_0$ and $\left( \frac{da}{dx}\right)|_{x=0}=0 $,
171: we have
172: \begin{equation}
173: K=-\frac{2r^2}{\pi^2a_{0}^2}+\frac{4r{\rm{E}}_0}{\pi^2a_0}.
174: \end{equation}
175: Using $(17)$ in $(16)$ we write
176: \begin{equation}
177: \Pi(a)=\frac{2r^2}{\pi^2a^2}\left(\frac{a^2_{0}}{a^2}-1\right) -\frac{4r{\rm{E}}_0}{\pi^2a}\left(\frac{a_{0}}{a}-1\right) .
178: \end{equation}
179: We now introduce a normalized pulse width $z(x)=\frac{a(x)}{a_{0}}$ to rewrite $(15)$ as
180:
181: \begin{equation}
182: \frac{1}{2}\left( \frac{dz}{dx}\right)^{2}+\Pi(z)=0
183: \end{equation}
184: with
185:
186: \begin{equation}
187: \Pi(z)=\left( 1+3\beta V\right) \left( \frac{1}{z}-1\right) \left\lbrace \mu\left( 1+3\beta V\right)\left( \frac{1}{z}+1\right) +\nu\right\rbrace
188: \end{equation}
189: and
190:
191: \begin{equation}
192: \mu=\frac{8}{\pi^2a_{0}^4}\,\,\,\,\,\nu=-\frac{8{\rm{E}}_0}{\pi^2a_{0}^2} .
193: \end{equation}
194: The potential function $\Pi(z)$ vanishes for two values of $z$, namely, $z_1=1$ and $z_2=-\frac{\mu \left( 1+3\beta V\right) }{\nu+\mu \left( 1+3\beta V\right)}$
195: implying that there is a minimum in between these points. The point at which $\Pi(z)$ is minimum can be obtained
196: from $\frac{d\Pi(z)}{dz}=0$ as
197: \begin{equation}
198: z_m=-\frac{2\mu \left( 1+3\beta V\right)}{\nu}.
199: \end{equation}
200: In the limit when $\frac{\nu}{\mu}=-2\left( 1+3\beta V\right)$ we have $z_2=1=z_m$ and $\frac{d\Pi(z)}{dz}|_{z=1}=\Pi(z_m)=0$. In this case the potential well degenerates into a single point such that a particle released at that point will stay there. In the context of this work, this signifies that a wave pulse
201: for which $\frac{\nu}{\mu}=-2\left( 1+3\beta V\right)$ propagates with unchanged shape as a cosequence of exact balance between nonlinear and dispersive effects. Interestingly, we now have a continuous family of soliton solutions because $\frac{\nu}{\mu}$ depends on the continuous variable $\beta V$. This implies that in the neighbourhood of each embedded soliton there is another one having a slightly higher or lower energy. A particular embedded soliton when perturbed may always relax to the adjacent one. This makes an embedded soliton fully stable.
202:
203:
204: \begin{figure}
205: \includegraphics[width=0.70\columnwidth, angle=-90] {pot.eps}
206: \caption{${\Pi}(z)$ as a function of $z$ for $\frac{\nu}{\mu}=-2\left( 1+3\beta V\right)$}
207: \end{figure}
208:
209: \par
210: In Fig. 1 we plot the potential function $\Pi(z)$ as a function of $z$ for $\beta V= -0.2, -0.1, 0, 0.1, 0.2$.
211: Each of these potentials has $z_1=z_2=z_m=1$ and is, therefore, associated with an embedded soliton. The solid curve $\beta V=0$ is the potential function corresponding to the usual second-order NLS equation while the
212: dotted curves represent the potential functions of the embedded solitons as found in the solution of third-order NLS equation. For $\beta V >0$, the potential curves for the embedded solitons lie above the solid curve for $z >1$. For $\beta V <0$ the dotted curves fall below the solid curve. We have drawn only a few curves for discrete values of $\beta V $. However, a dense set of curves can be drwan by varying $\beta V $ continuously.
213:
214:
215:
216:
217:
218:
219: \section*{ACKNOWLEDGEMENTS}
220: This work forms the part of a Research Project F.10-10/2003(SR) supported by the University Grants Commission,
221: Govt. of India. One of the authors (SGA) is thankful to the UGC, Govt. of India for a Research Fellowship.
222:
223:
224:
225: \begin{thebibliography}{department}
226: \bibitem{1} G. P. Agrawal, Nonlinear Fiber Optics, 3rd Edition, (Academic Press, 1995, San Diego, California, USA).
227: \bibitem{2} J. Fujioka, A. Espinosa-Cer\'on and R. F. Rodr\'iguez, A Survey of embedded solitons, Rev. Mex. De Fis. 52 (2006) 6-14.
228: \bibitem{3} J. Yang and R. Akylas,Continuous families of embedded solitons in the third-order nonlinear Schr\"{o}dinger equation, Stud. Appl. Math. 111 (2003) 359-375.
229: \bibitem{4} J. Yang, B. A. Malomed and D. J. Kaup, Embedded soliton in second -harmonic-generating systems, Phys. Rev. Lett. 83 (1999) 1958-1961.
230: \bibitem{5} J. Yang, Stable embedded solitons, Phys. Rev. Lett. 91 (2003) 143903(1-4).
231: \bibitem{6} D. C. Calvo and T. R. Akylas, Stability of bound state near the zero-order dispersion wavelength in optical fibers, Phys. Rev. E 56, (1997) 4757-4764; D. E. Pelinovsky and J. Yang, Stability analysis of embedded solitons in the generalized third-order nonlinear Schr\"{o}dinger equation, Chaos 15 (2005) 037115(5-11); A. A. Minzoni, N. F. Smyth and A. L. Worthy, A Variational approach to the stability of an embedded NLS soliton at edge of the continuum, Physica D 206 (2005) 166-179.
232: \bibitem{7} D. Anderson, Variational approach to nonlinear pulse propagation in optical fibers, Phys. Rev. A 27 (1983) 3135-3144.
233: \bibitem{8} V. E. Zakharov and A. B. Shabat, Exact theory of two-dimensional self-focousing and one-dimensional self modulation of waves in nonlinear media, Sov. Phys. JETP 34 (1972) 62-69.
234: \bibitem{9} G. B. Arfken and H. J. Weber, Mathematical Method for Physicists (Elsevier, New Delhi, 2004).
235:
236:
237: \end{thebibliography}
238: \end{document}
239: