1: \documentclass [12pt] {article}
2: \usepackage[dvips]{graphicx}
3: \renewcommand{\thesection}{\arabic{section}.}
4: \begin{document}
5:
6: \title{Statistical Approach of Modulational Instability in the Class of
7: Derivative Nonlinear Schr\"odinger Equations}
8: \author{ A.T. Grecu,
9: D. Grecu, Anca Visinescu \\
10: \small{\it Department of Theoretical Physics} \\
11: \small{\it ``Horia Hulubei'' National Institute of Physics and Nuclear
12: Engineering} \\
13: \small{Bucharest, P.O. Box MG-6, Romania}\\
14: \small{e-mail: avisin@theor1.theory.nipne.ro}
15: }
16: \date{}
17: \maketitle
18:
19: \abstract{The modulational instability (MI) in the class of NLS equations
20: is discussed using a statistical approach (SAMI). A kinetic equation for
21: the two-point correlation function is studied in a linear approximation,
22: and an integral stability equation is found. The modulational instability
23: is associated with a positive imaginary part of the frequency. The integral
24: equation is solved for different types of initial distributions ($\delta$-
25: function, Lorentzian) and the results are compared with those obtained using
26: a deterministic approach (DAMI). The differences between MI of
27: the normal NLS equation and derivative NLS equations is emphasized.}
28:
29: Keywords:NLS equations, modulational instability
30:
31: PACS: 05.45
32:
33: \section{Introduction}
34:
35: \indent The modulational instability (MI - also known as Benjamin-Feir
36: instability) is a well known phenomenon encountered, in quite general
37: conditions, almost each time a quasi-monochromatic wave propagates in a
38: weakly nonlinear medium (Dodd et al 1982, Remoissenet 1999, Scott
39: 2003, Benjamin and Feir 1967, Bespalov et al 1966 and Abdullaev et
40: al 2002).
41: It is responsible for the generation of robust, localized excitations,
42: the solitons in completely integrable
43: systems, and solitary waves in non-integrable ones.
44:
45: Two different ways to discuss this phenomenon are possible. The first
46: one is a deterministic approach (DAMI) in which the amplitude of a plane
47: wave, with an amplitude dependent dispersion relation (a Stokes wave),
48: is slowly modulated
49:
50: \begin{equation} \label{eq1}
51: \Psi(x,t)=a\left(1+\varepsilon A(x,t)\right)\exp\left(\mathbf{i}
52: (kx-\omega t)\right)
53: \end{equation}
54: \noindent This approach is well known and can be found in any textbook
55: on nonlinear physics (see Dodd et al 1982, Remoissenet 1999).
56: The second approach is a statistical one (SAMI), and its aim is to see
57: the influence of the statistical properties of the medium on the MI
58: phenomenon. It is a less used approach, but in the last ten years was
59: intensively applied to investigate different physical situations
60: (Alber 1978, Janssen 1983, Fedele et al. 2000, Hall et al 2002,
61: Lisak et al 2002, Fedele et al 2002, Onorato et al 2003), ranging
62: from hydrodynamics (Janssen 1983, Fedele et al. 2000, Hall et al 2002,
63: Lisak et al 2002, Fedele et al 2002), to plasma
64: physics (Fedele 2002), large amplitude wave-envelope propagation in nonlinear
65: media, high-energy accelerators, condensed matter.
66: Recently it was extended also to discrete systems (Visinescu et al
67: 2003, Grecu D. et al 2005) and to
68: systems of coupled nonlinear equations (Manakov's system) (Grecu D. et al
69: 2004 and 2005).
70:
71: In this approach a kinetic equation for the two-point correlation function
72: \begin{equation} \label{eq2}
73: \rho(x_1,x_2)=\left<\Psi(x_1)\,\Psi^*(x_2)\right>
74: \end{equation}
75: \noindent is written down and a linear stability analysis near an initial
76: distribution
77: \begin{equation} \label{eq3}
78: \rho_0(x_1,x_2)=\rho_0(\left|x_1-x_2\right|)
79: \end{equation}
80: \noindent is performed. The initial state is assumed homogenous and
81: consequently $\rho_0$ depends only on
82: $|x_1-x_2|$. In the previous expression of $\rho(x_1,x_2)$ by
83: $\left<\ldots\right>$ we understand a
84: statistical average over the distribution function of the medium.
85:
86: The aim of this paper is to investigate MI, especially using the statistical
87: approach of the following equation
88: \begin{equation}
89: \label{eq4}
90: \mathbf{i} \frac{\partial \Psi}{\partial t} + {1\over 2}\frac{\partial^2
91: \Psi}{\partial x^2}+\lambda |\Psi|^2 \Psi +\mathbf{i} \mu |\Psi|^2
92: \frac{\partial \Psi}{\partial x} +\mathbf{i} \nu \frac{\partial |\Psi|^2}
93: {\partial x}\Psi=0
94: \end{equation}
95: \noindent It is an extended NLS equation, and for different choises of
96: the real parameters $\lambda,\mu,\nu$ several completely integrable equations
97: are recovered. If $\mu=\nu=0$, we get the well known NLS equation
98: \begin{equation} \label{eq5}
99: \mathbf{i} \frac{\partial \Psi}{\partial t} + {1\over 2}\frac{\partial^2
100: \Psi}{\partial x^2} + \lambda|\Psi|^2 \Psi=0
101: \end{equation}
102: while for $\lambda=0,\mu=\nu$ one obtains the completely integrable
103: derivative NLS-I equation (dNLS I) (Kaup 1978, Nakamura et al. 1980,
104: Mio et al 1976, Mj\o lhus 1976)
105: \begin{equation} \label{eq6}
106: \mathbf{i} \frac{\partial \Psi}{\partial t} + {1\over 2}\frac{\partial^2
107: \Psi}{\partial x^2} + \mathbf{i}\mu\frac{\partial}{\partial x}
108: \left(|\Psi|^2\Psi\right)=0
109: \end{equation}
110: and for $\lambda=\nu=0$ we get another integrable derivative NLS II
111: equation (dNLS II) (Chen, Lee 1979)
112: \begin{equation} \label{eq7}
113: \mathbf{i}\frac{\partial \Psi}{\partial t} + {1\over 2}\frac{\partial^2
114: \Psi}{\partial x^2}+\mathbf{i}\mu|\Psi|^2\frac{\partial \Psi}{\partial x}=0
115: \end{equation}
116:
117: Before starting the SAMI study of eq. (\ref{eq4}), let us briefly recall
118: the results of a DAMI analysis. Introducing (\ref{eq1}) in (\ref{eq4}) and
119: keeping only the linear terms in $\varepsilon$ a linear evolution equation for
120: $A(x,t)$ is found. Looking for plane wave solutions of it
121: $$
122: A(x,t)=\alpha_1\mathrm{e}^{\mathbf{i}(Qx-\Omega t)}+\alpha_{2}^{*}
123: \mathrm{e}^{-\mathbf{i}(Qx-\Omega^* t)}
124: $$
125: the compatibility condition of the homogenous system leads us to
126: \begin{equation}
127: \label{eq8}
128: \Omega-Q\left[k+(\mu+\nu)|a|^2\right]=\mathbf{i} Q\sqrt{(\lambda-
129: \nu|a|^2-\mu k)|a|^2-\frac{Q^2}{4}}
130: \end{equation}
131: an the instability is related to $\mathrm{Im}\Omega>0$. For $\lambda\neq 0$
132: and denoting
133: $\lambda_1 = \lambda - \nu |a|^2$ we obtain
134: \begin{equation} \label{eq9}
135: \mathrm{Im}\Omega = Q\sqrt{(\lambda_1-\mu k)|a|^2-\frac{Q^2}{4}}
136: \end{equation}
137: and one sees that the instability regions depend on the sign of $\mu$.
138: In a $(Q^2,\,k)$ representation
139: these are given in Fig. \ref{fig1} (arbitrary units)
140: \begin{figure}[ht]
141: \label{fig1}
142: \hfill
143: \begin{minipage}{.45\textwidth}
144: \begin{center}
145: \includegraphics[scale=0.3]{fig1-a}
146: \end{center}
147: \end{minipage}
148: \hfill
149: \begin{minipage}{.45\textwidth}
150: \begin{center}
151: \includegraphics[scale=0.3]{fig1-b}
152: \end{center}
153: \end{minipage}
154: \hfill
155: \caption{\small Instability regions $(\lambda \neq 0)$; (a) $\mu < 0$,
156: (b) $\mu>0$.}
157: \end{figure}
158:
159: It is interesting to consider the instability regions for the two
160: integrable dNLS equations. For dNLS I an instability arises only if
161: $\mu<0$. We have
162: \begin{equation} \label{eq10}
163: \mathrm{Im} \Omega = Q\sqrt{|\mu||a|^2(k-|\mu||a|^2)-\frac{Q^2}{4}}
164: \end{equation}
165: and in a $(Q^2,\,k)$ representation the instability occurs only if
166: $k>|\mu||a|^2$. For dNLS II the same condition, $\mu<0$, must be
167: fulfilled and
168: \begin{equation} \label{eq11}
169: \mathrm{Im} \Omega=Q\sqrt{|\mu||a|^2 k-\frac{Q^2}{4}}
170: \end{equation}
171: The instability regions in $(Q^2,\,k)$ representation (arbitrary units)
172: are presented in Fig. 2.
173: \begin{figure}[ht]
174: \label{fig2}
175: \hfill
176: \begin{minipage}[t]{.45\textwidth}
177: \begin{center}
178: \includegraphics[scale=0.3]{fig2-a}
179: \end{center}
180: \end{minipage}
181: \hfill
182: \begin{minipage}[t]{.45\textwidth}
183: \begin{center}
184: \includegraphics[scale=0.3]{fig2-b}
185: \end{center}
186: \end{minipage}
187: \hfill
188: \caption{\small Instability regions of the integrable dNLS equations:
189: (a) dNLS I, (b) dNLS II}
190: \end{figure}
191:
192: \noindent These results have to be compared with the similar result
193: obtained for NLS equation (\ref{eq5}), which writes
194: \begin{equation} \label{eq12}
195: \mathrm{Im} \Omega=Q\sqrt{\lambda |a|^2 - \frac{Q^2}{4}}
196: \end{equation}
197: We see that in the NLS case the instability region is independent on
198: the wave vector $k$ of the carrying wave, and restricted to the long
199: wave-length region ($Q^2 < 4\lambda |a|^2$), while for all the situations
200: discussed above they are $k$-dependent, and more over they are not
201: restricted to the long wavelength region (exception is the case (b)
202: of Fig 1). Actually for the dNLS equations the long wavelength region
203: is either forbidden, or highly unfavourable for the development of the MI.
204:
205: In the next section the kinetic equation for the two-point correlation
206: function (\ref{eq2}) is obtained and its stability is studied in the
207: linear approximation. An integral stability equation will be obtained
208: and solved for different initial distributions in section 3
209: ($\delta$-distribution, Lorentzian). Few conclusions and comments are
210: given in the last section. Preliminary results have been presented in
211: Grecu, A.T. 2005.
212:
213: \section{Kinetic equation. Linear stability analysis}
214:
215: The kinetic equation for the two-point correlation function (\ref{eq2}),
216: is obtained easily in the following way:
217: \begin{enumerate}
218: \item write the equation (\ref{eq4}) for $x=x_1$ and multiply by
219: $\Psi^{*}(x_2)$;
220: \item write the complex conjugated of (\ref{eq4}) for $x=x_2$ and
221: multiply by $\Psi(x_1)$;
222: \item add the two equations and take the statistical average;
223: \item finally use a Gaussian approximation to decouple the four-point
224: correlation functions.
225: \end{enumerate}
226: As an illustration we give below two examples of such decouplings
227: $$\left < \Psi(x_1) \Psi^{*}(x_1) \Psi(x_1) \Psi^*(x_2) \right > =
228: 2 \overline{a^{2}}(x_1)\rho(x_1,x_2)$$
229: $$\left < \Psi(x_1) \Psi^*(x_1) \frac{\partial \Psi(x_1)}{\partial x_1}
230: \Psi^*(x_2)\right > =$$
231: $$ =\overline{a^{2}}(x_1) \frac{\partial}{\partial x_1}\rho(x_1,x_2) +
232: \rho(x_1,x_2) \lim_{x_2\to x_1} \frac{\partial}{\partial x_1}\rho(x_1,x_2)$$
233: where
234: \begin{equation} \label{eq13}
235: \overline{a^{2}}(x_1)=\left < \Psi(x_1) \Psi^*(x_1) \right >
236: \end{equation}
237: The kinetic equation found in this way is
238: \begin{eqnarray} \label{eq14}
239: & & \mathbf{i} \frac{\partial \rho}{\partial t} +{1 \over 2}
240: \left(\frac{\partial^2}{\partial x_{1}^{2}} - \frac{\partial}
241: {\partial x_{2}^{2}}\right)\rho + 2 \lambda \left[ \overline{a^{2}}
242: (x_1)-\overline{a^{2}}(x_2)\right]\rho \nonumber \\
243: & & + \mathbf{i} (\mu+\nu)\left[ \overline{a^{2}}(x_1)\frac{\partial}
244: {\partial x_1} + \overline{a^{2}}(x_2)\frac{\partial}{\partial x_2}\right]
245: \rho \\
246: & & + \mathbf{i} (\mu+\nu)\rho\left[ \lim_{x_2 \to x_1}\frac{\partial}
247: {\partial x_1} + \lim_{x_1 \to x_2}\frac{\partial}{\partial x_2}\right]\rho
248: \nonumber \\
249: & & + 2\mathbf{i} \nu \rho\left[ \lim_{x_2 \to x_1}\frac{\partial}
250: {\partial x_1} +\lim_{x_1 \to x_2}\frac{\partial}{\partial x_2}\right]\rho^* = 0 \nonumber
251: \end{eqnarray}
252:
253: The next step is to introduce the new set of variables
254: \begin{eqnarray} \label{eq15}
255: & & x=x_1-x_2 \;\;\; \mathrm{relative\: coordinate} \\
256: & & X={1\over 2}(x_1+x_2) \;\; \mathrm{center\: of\: mass\: coordinate}
257: \nonumber
258: \end{eqnarray}
259: and make a Fourier transform with respect to the relative coordinate.
260: This procedure is known as the Wigner-Moyal transform (Wigner 1932,
261: Moyal 1949) (also
262: known as Klimontovich's statistical average method, Toda et al 1995),
263: and is very useful to deal with non-homogenous evolution equations. The
264: Fourier transform is defined as
265: \begin{equation} \label{eq16}
266: F(k,X,t) = {1\over{2\pi}} \int_{-\infty}^{+\infty}{ \,\mathrm{d}x\,
267: \mathrm{e}^{-\mathbf{i} kx} \rho(x,X,t)}
268: \end{equation}
269: Using the definition (\ref{eq13}) of $\overline{a^{2}}$, it is easily
270: seen that
271: \begin{equation} \label{eq17}
272: \overline{a^{2}}(X,t)=\int_{-\infty}^{+\infty}{\,\mathrm{d}k\, F(k,X,t)}
273: \end{equation}
274: After straightforward calculations the Fourier transform of (\ref{eq14})
275: is given by
276: \begin{eqnarray} \label{eq18}
277: & & \frac{\partial F}{\partial t} + k \frac{\partial F}{\partial X} +
278: 4\lambda \sum_{j=0}^{\infty}\frac{(-1)^j}{(2j+1)\,!\,2^{2j+1}}
279: \frac{\partial^{2j+1} \overline{a^{2}}(X)}{\partial X^{2j+1}}
280: \frac{\partial^{2j+1} F(k,X)}{\partial k^{2j+1}} \nonumber \\
281: & & + (\mu+\nu) \frac{\partial \overline{a^{2}}(X)}{\partial X}F(k,X)+
282: 2\nu \frac{\partial \left[\overline{a^{2}}(X)\right]^*}{\partial X}F(k,X) \\
283: & & + (\mu+\nu) \sum_{j=0}^{\infty}{\frac{(-1)^j}{2j\,! 2^{2j}}
284: \frac{\partial^{2j} \overline{a^{2}}(X)}{\partial X^{2j}}\frac{\partial}
285: {\partial X}\frac{\partial^{2j} F(k,X)}{\partial k^{2j}}} \nonumber \\
286: & & - 2(\mu+\nu) \sum_{j=0}^{\infty}{\frac{(-1)^j}{(2j+1)\,!\,2^{2j+1}}
287: \frac{\partial^{2j+1} \overline{a^{2}}(X)}{\partial X^{2j+1}}
288: \frac{\partial^{2j+1} \left[k F(k,X)\right]}{\partial k^{2j+1}}} = 0 \nonumber
289: \end{eqnarray}
290: In obtaining (\ref{eq18}) we have expanded $\overline{a^{2}}(x_1)=
291: \overline{a^{2}}(X+{x\over 2})$ and
292: $\overline{a^{2}}(x_2)=\overline{a^{2}}(X-{x\over 2})$ in power series
293: around the point $X$, and we have used the relation
294: $$x^j\,\mathrm{e}^{-\mathbf{i} kx}=(i)^j \frac{\partial^j}{\partial k^j}
295: \mathrm{e}^{-\mathbf{i} kx}$$
296: Also we used
297: $$\lim_{x_2\to x_1}\frac{\partial}{\partial x_1}\rho(x_1,x_2)=\lim_{x\to 0}
298: \left({1\over 2}\frac{\partial}{\partial X} +\frac{\partial}{\partial x}\right)
299: \int_{-\infty}^{+\infty}{\,\mathrm{d}k\, \mathrm{e}^{\mathbf{i} kx} F(k,X)}
300: =$$
301: $$={1\over 2}\frac{\partial \overline{a^{2}}(X)}{\partial X}+\mathbf{i} k
302: \overline{a^{2}}(X)$$
303: and similar expressions for the other limits which appear in (\ref{eq14}).
304:
305: For a linear stability analysis we consider
306: \begin{equation} \label{eq19}
307: F(k,X,t)=f(k)+\varepsilon \mathcal{F}(k,X,t)
308: \end{equation}
309: and
310: \begin{equation} \label{eq20}
311: \overline{a^{2}}(X,t)=\overline{a^{2}_{0}}+\varepsilon \overline{a^{2}_{1}}
312: (X,t)
313: \end{equation}
314: where
315: \begin{equation} \label{eq21}
316: \overline{a^{2}_{0}}=\int_{-\infty}^{+\infty}{\,\mathrm{d}k\,f(k)},
317: ~~~~~
318: \overline{a^{2}_{1}}(X,t)=\int_{-\infty}^{+\infty}{\,\mathrm{d}k\,
319: \mathcal{F}(k,X,t)}
320: \end{equation}
321: Here $f(k)$ is the Fourier transform of $\rho_0(|x|)$, which, due to the
322: homogenous assumption of the initial state, depends only on $k$, and
323: moreover is an even function of $k$.
324:
325: The linearized kinetic equation in the $(k,X)$ representation is given by
326: \begin{eqnarray} \label{eq22}
327: & & \frac{\partial \mathcal{F}}{\partial t}+[k+(\mu+\nu)\overline{a^{2}_{0}}]
328: \frac{\partial \mathcal{F}}{\partial X}+(\mu+\nu)f\frac{\partial
329: \overline{a^{2}_{1}}}{\partial X}+2\nu f\left(\frac{\partial
330: \overline{a^{2}_{1}}}{\partial X}\right)^* \nonumber \\
331: & & + 4 \lambda \sum_{j=0}^{\infty}\frac{(-1)^j}{(2j+1)\,!\,2^{2j+1}}
332: \frac{\partial^{2j+1} \overline{a^{2}_{1}}(X)}{\partial X^{2j+1}}
333: \frac{\partial^{2j+1} f(k)}{\partial k^{2j+1}} \\
334: & & - 2(\mu+\nu) \sum_{j=0}^{\infty}\frac{(-1)^j}{(2j+1)\,!\,2^{2j+1}}
335: \frac{\partial^{2j+1} \overline{a^{2}_{1}}(X)}{\partial X^{2j+1}}
336: \frac{\partial^{2j+1} (kf(k))}{\partial k^{2j+1}} = 0 \nonumber
337: \end{eqnarray}
338:
339: The last step is to look for a plane wave solution of (\ref{eq22})
340: \begin{equation} \label{eq23}
341: \mathcal{F}(k,X,t)=g_1(k)\mathrm{e}^{\mathbf{i}(QX-\Omega t)}+
342: g_{2}^{*}(k)\mathrm{e}^{-\mathbf{i}(QX-\Omega^* t)}
343: \end{equation}
344: $$\overline{a^{2}_{1}}(X,t)=G_1\mathrm{e}^{\mathbf{i}(QX-\Omega t)}+
345: G_{2}^{*}\mathrm{e}^{-\mathbf{i}(QX-\Omega^* t)}$$
346: where
347: \begin{equation} \label{eq24}
348: G_i=\int_{-\infty}^{+\infty}{g_i(k)}\,\mathrm{d}k\,,\;\;i=1,2
349: \end{equation}
350: Introducing (\ref{eq23}) into (\ref{eq22}) the following equation is found
351: \begin{eqnarray} \label{eq25}
352: && (k-\omega)g_1 + (\mu+\nu)fG_1+2\nu fG_2 \nonumber \\
353: && +{{2\lambda}\over Q}\left[f(k+{Q\over 2})-f(k-{Q\over 2})\right]G_1 \\
354: && -\frac{\mu+\nu}{Q}\left[h(k+{Q\over 2})-h(k-{Q\over 2})\right]G_1=0 \nonumber
355: \end{eqnarray}
356: Here we denoted \ensuremath{\omega = {\Omega\over Q}-(\mu+\nu)
357: \overline{a^{2}_{0}}} and $h(k)=kf(k)$. Also we used
358: $$2\sum_{j=0}^{\infty}\frac{1}{(2j+1)\,!}\left({Q\over 2}\right)^{2j+1}
359: \frac{\partial^{2j+1} f(k)}{\partial k^{2j+1}}=f(k+{Q\over 2})-
360: f(k-{Q\over 2})$$
361: and a similar relation for $h(k)$.
362:
363: In (\ref{eq25}) we have to divide by $(k-\omega)$ and integrate over $k$.
364: Introducing the notations
365: $$I=\int_{-\infty}^{+\infty}{\frac{f(k)}{k-\omega}\,\mathrm{d}k\,}$$
366: \begin{equation} \label{eq26}
367: J=\int_{-\infty}^{+\infty}{\frac{f(k+{Q\over 2})-f(k-{Q\over 2})}{k-\omega}\,
368: \mathrm{d}k\,}
369: \end{equation}
370: $$K=\int_{-\infty}^{+\infty}{\frac{h(k+{Q\over 2})-h(k-{Q\over 2})}
371: {k-\omega}\,\mathrm{d}k\,}$$
372: the following homogenous algebric linear equation in $G_1$ and $G_2$ is
373: obtained
374: \begin{equation} \label{eq27}
375: G_1+(\mu+\nu)IG_1+2\nu IG_2+{{2\lambda}\over Q}JG_1-\frac{\mu+\nu}{Q}KG_1=0
376: \end{equation}
377: A second equation is found starting from the complex conjugated of
378: (\ref{eq22}), namely
379: \begin{equation} \label{eq28}
380: G_2+(\mu+\nu)IG_2+2\nu IG_1+\frac{2\lambda}{Q}JG_2-\frac{\mu+\nu}{Q}KG_2=0
381: \end{equation}
382: The compatibility condition for the system (\ref{eq27}),(\ref{eq28}) leads
383: us to the following integral stability equations
384: $$ \label{eqA}
385: 1+(\mu-\nu)I+\frac{2\lambda}{Q}J-\frac{\mu+\nu}{Q}K=0
386: \eqno\mathrm{(A)}$$
387: or
388: $$ \label{eqB}
389: 1+(\mu+3\nu)I+\frac{2\lambda}{Q}J-\frac{\mu+\nu}{Q}K=0
390: \eqno\mathrm{(B)}$$
391: In the next section these will be solved for different initial conditions
392: $f(k)$.
393:
394: \section{Solution of the integral stability equation}
395:
396: The first distribution function we shall discuss is a $\delta$-spectrum
397: \begin{equation} \label{eq29}
398: f(k)=\overline{a^{2}_{0}}\delta(k)
399: \end{equation}
400: It doesn't describe a realistic situation as it coresponds to a constant
401: two-point correlation function in the initial state,
402: $\rho_0=\overline{a^{2}_{0}}=const.$. But the calculations are easily
403: done and can be considered as a limit case to which more realistic situations
404: can be compared. The integrals $I,J,K$ are immediatly calculated
405: \begin{equation} \label{eq30}
406: I=-\frac{\overline{a^{2}_{0}}}{\omega},\;J=\frac{\overline{a^{2}_{0}}Q}
407: {\omega^2 - {{Q^2}\over 4}},\;K=0
408: \end{equation}
409: Then both integral equations (A) and (B) become an algebraic equation of
410: third order in $\omega$
411: \begin{equation} \label{eq31}
412: \omega^3 -\bar{\mu} \overline{a^{2}_{0}}\omega^2+
413: (2\lambda\overline{a^{2}_{0}}-{{Q^2}\over 4})\omega +
414: \bar{\mu}\overline{a^{2}_{0}}{{Q^2}\over 4}=0
415: \end{equation}
416: where $\bar{\mu}=\mu_A=\mu-\nu$ for A-equation and
417: $\bar{\mu}=\mu_B=\mu+3\nu$ for B-equation. With
418: $\omega=y+{1\over 3}\bar{\mu}\overline{a^{2}_{0}}$
419: it is reduced to the canonical form
420: \begin{equation} \label{eq32}
421: y^3+py+q=0
422: \end{equation}
423: where
424: $$p=2\lambda\overline{a^{2}_{0}}-{1\over 3}(\bar{\mu}\overline{a^{2}_{0}})^2
425: -{{Q^2}\over 4}$$
426: \begin{equation} \label{eq33}
427: q={2\over 3}\bar{\mu}\overline{a^{2}_{0}}\left[\lambda
428: \overline{a^{2}_{0}}-{1\over 9}(\bar{\mu}\overline{a^{2}_{0}})^2+
429: {{Q^2}\over 4}\right]
430: \end{equation}
431: Complex solutions of (\ref{eq32}) are obtained if the discriminant
432: $\Delta=\left({p\over 3}\right)^3+\left({q\over 2}\right)^2$
433: is positive. The sign of $\Delta$ is controlled by the sign of $p$. For
434: $\lambda\neq 0$ we have $p>0$ if
435: $Q^2/4<2\lambda\overline{a^{2}_{0}}-{1\over 3}(\bar{\mu}
436: \overline{a^{2}_{0}})^2$, corresponding to the long wavelength region.
437: With the notations
438: \begin{equation} \label{eq35}
439: \sqrt{\frac{p^3}{27}}={q\over 2}\tan\alpha,\;\;\tan\varphi=
440: \sqrt[3]{\tan{\alpha\over 2}}
441: \end{equation}
442: we have
443: \begin{equation} \label{eq36}
444: \mathrm{Im}\,\omega = \frac{\sqrt{p}}{\sin2\varphi}
445: \end{equation}
446: But if $p<0$ we can still have $\Delta>0$. Writing
447: \begin{equation} \label{eq37}
448: {{Q^2}\over 4}=2\lambda\overline{a^{2}_{0}}-{1\over 3}(\bar{\mu}
449: \overline{a^{2}_{0}})^2+3\tilde{Q}^2
450: \end{equation}
451: the condition $\Delta>0$ leads us to the inequality
452: \begin{equation} \label{eq38}
453: \tilde{Q}^3-\bar{\mu}\overline{a^{2}_{0}}\tilde{Q}^2<\bar{\mu}
454: \overline{a^{2}_{0}}\left[\lambda\overline{a^{2}_{0}}-{4\over 27}
455: (\bar{\mu}\overline{a^{2}_{0}})^2\right]
456: \end{equation}
457: In this case ($\Delta>0$ and $p<0$) the usual notations are
458: $$ \label{eq35p}
459: \sqrt{-\frac{p^3}{27}}={q\over 2}\sin \alpha,\;\;\tan\varphi=
460: \sqrt[3]{\tan{\alpha\over 2}}
461: \eqno(\ref{eq35}')$$
462: and one obtaines
463: $$ \label{eq36p}
464: \mathrm{Im}\,\omega=\sqrt{-p}\cot 2\varphi
465: \eqno(\ref{eq36}')$$
466: The instability remains in the long wavelength region.
467:
468: When $\lambda=0$ the situation is completely changed. Then
469: $$p=-\left[{{Q^2}\over 4}+{1\over 3}(\bar{\mu}\overline{a^{2}_{0}})^2\right]$$
470: \begin{equation} \label{eq39}
471: q={2\over 3}\mu\overline{a^{2}_{0}}\left[{{Q^2}\over 4}-{1\over 9}
472: (\bar{\mu}\overline{a^{2}_{0}})^2\right]
473: \end{equation}
474: and $\Delta>0$ can be realized only if $Q>{2\over 3}\bar{\mu}
475: \overline{a^{2}_{0}}$. This means that the instability develops for
476: higher wave vectors $Q$, a result in complete agreement to the conclusions
477: of the DAMI analysis of the derivative NLS equations. Using the
478: same notations ($34'$) we get the same result ($35'$), and the requirement
479: $\Delta>0$ reduces to
480: \begin{equation} \label{eq040}
481: \left({q\over 2}\right)^2 >-\left({p\over 3}\right)^3
482: \end{equation}
483: an inequality from which the limit value of ${{Q^2}\over 4}$ can be calculated.
484:
485: A more realistic situation is a Lorentzian distribution
486: \begin{equation} \label{eq40}
487: f(k)={\overline{a^{2}_{0}}\over \pi}\frac{\sigma}{k^2 + \sigma^2}
488: \end{equation}
489: which corresponds to an exponential decaying two-point correlation
490: function in the initial state
491: \begin{equation} \label{eq41}
492: \rho_0(x)=\overline{a^{2}_{0}} \mathrm{e}^{-\sigma x}\; x>0
493: \end{equation}
494: The integrals {\bf I}, {\bf J}, {\bf K} are easily calculated in the
495: $k$-complex plane assuming $\omega$ complex with $\mathrm{Im}\,\omega >0$.
496: We obtain
497: $$
498: I=-\frac{\overline{a^{2}_{0}}}{\tilde \omega}
499: $$
500: \begin{equation} \label{eq42}
501: J=\frac{\overline{a^{2}_{0}} Q}{\tilde \omega^2 -{Q^2 \over 4}}
502: \end{equation}
503: $$
504: K=\frac{-\mathbf{i} \sigma \overline{a^{2}_{0}} Q}{\tilde \omega^2 -
505: {Q^2 \over 4}}
506: $$
507: where $\tilde \omega=\omega+\mathbf{i} \sigma$, and the instability
508: condition $\mathrm{Im}\,\omega >0$ becomes for the new variable
509: $\tilde \omega$, $\mathrm{Im}\,\tilde\omega>\sigma$.
510:
511: To have a point of reference, let us recall the result for the usual
512: NLS equation ($\mu=\nu=0$) (Visinescu et al 2003, Grecu D. et
513: al (2004, 2005), Grecu A.T. 2005)
514: \begin{equation} \label{eq43}
515: \mathrm{Im}\,\omega=\sqrt{2\lambda \overline{a^{2}_{0}} -{Q^2 \over 4}}-
516: \sigma
517: \end{equation}
518: We get $\mathrm{Im}\,\omega>0$ if
519: \begin{equation} \label{eq44}
520: Q<2\sqrt{2\lambda\overline{a^{2}_{0}}-\sigma^2}
521: \end{equation}
522: which shows that the MI is restricted to the long wave-length region and
523: is strongly dependent on the correlation length in the initial state;
524: if $\sigma^2>2\lambda\overline{a^{2}_{0}}$, the initial state is stable
525: to small modulations.
526:
527: When (\ref{eq42}) are introduced in (A) and (B), we get again an algebraic
528: equation of third order in $\tilde \omega$, but now with complex
529: coefficients. It can be reduced to the canonical form with
530: \begin{equation} \label{eq45}
531: p=2\left(\lambda+{\mathbf{i}\over 2}\overline{\nu}\sigma\right)
532: \overline{a^{2}_{0}}-{1\over 3}(\overline{\mu}\overline{a^{2}_{0}})^2-
533: {Q^2\over 4}
534: \end{equation}
535: $$
536: q={2\over 3}\overline{\mu}\overline{a^{2}_{0}}\left[\left(\lambda+
537: {\mathbf{i}\over 2}\overline{\nu}\sigma\right)\overline{a^{2}_{0}}-
538: {1\over 9}(\overline{\mu}\overline{a^{2}_{0}})^2+{Q^2\over 4}\right]
539: $$
540: where $\overline{\mu}$ has the same meaning as in the case of the
541: $\delta$-spectrum, and $\overline{\nu}=\mu+\nu$. As mentioned before the
542: MI region coresponds to $\mathrm{Im}\,\tilde\omega > \sigma$.
543:
544: Instead of working on the general case let us consider only the situations
545: of the dNLS equations. We shall start with the dNLS-II equation
546: ($\lambda=\nu=0$). Then both equations (A) and (B) reduce to a single
547: equation
548: \begin{equation} \label{eq46}
549: 1-\mu\overline{a^{2}_{0}}{1\over \tilde{\omega}}+\mathbf{i}\mu
550: \overline{a^{2}_{0}}\frac{\sigma}{\tilde{\omega^2}-{Q^2\over 4}}=0
551: \end{equation}
552: It is convenient to measure $\tilde\omega,\,\sigma\,\mathrm{and}\,Q$
553: in units of $|\mu|\overline{a^{2}_{0}}$. If $\varepsilon=\mathrm{sgn}\,\mu$,
554: the following algebraic equation of third order is found:
555: \begin{equation} \label{eq47}
556: \tilde \omega^3-\varepsilon\tilde \omega^2 -\left({Q^2\over 4}-
557: \mathbf{i}\varepsilon\sigma\right)\tilde \omega+\varepsilon{Q^2\over 4}=0
558: \end{equation}
559: It has a certain symmetry having solutions of the form $\tilde \omega=
560: \varepsilon\tilde \omega_r+\mathbf{i}\tilde \omega_i$. As we are interested
561: in the imaginary part of these solutions, the result is independent on
562: the sign of $\mu$. The equation (\ref{eq47}) is solved numerically.
563: Measuring $\Omega$ in units of $(\mu\overline{a^{2}_{0}})^2$, one has
564: $\mathrm{Im}\,\Omega=Q(\mathrm{Im}\,\tilde \omega -\sigma)$. A three
565: dimensional plot , $\mathrm{Im}\,\Omega$ function of $\sigma$ and $Q^2/4$
566: is given in Fig. \ref{fig:3} As one sees, the neighbourhood of $Q=0$
567: is excluded from the instability region, which extends on a finite domain
568: of $Q$ and $\sigma$. One can find a limit value for
569: $\sigma\,(\sigma_C\simeq0.194)$ and the MI exists only for $\sigma<\sigma_C$.
570: \begin{figure}[ht]
571: \label{fig:3}
572: \begin{center}
573: \includegraphics[width=5in]{fig3-all}
574: \end{center}
575: \caption{\small Instability region for dNLS I. Three dimensional
576: plot of $\mathrm{Im}\,\Omega$ versus
577: $Q^2/4$ and $\sigma$; Two different orientantions.}
578: \end{figure}
579:
580: Now let us consider the case of dNLS-I equation ($\lambda=0,\,\mu=\nu$).
581: The equation (A) becomes
582: \begin{equation} \label{eq48}
583: 1+2\mathbf{i}\frac{\mu\overline{a^{2}_{0}}}{\tilde \omega^2-{Q^2\over 4}}=0
584: \end{equation}
585: which is easily solved giving
586: \begin{equation} \label{eq49}
587: \mathrm{Im}\,\tilde \omega={1\over\sqrt{2}}\left[\sqrt{\left({Q^2\over 4}
588: \right)^2+4(\mu\overline{a^{2}_{0}})^2}-{Q^2\over 4}\right]^{1/2}
589: \end{equation}
590: and the MI region is determined by
591: \begin{equation} \label{eq50}
592: Q^2<4\left[\left(\frac{\mu\overline{a^{2}_{0}}}{\sigma}\right)^2-
593: \sigma^2\right],\;\sigma^2<\mu\overline{a^{2}_{0}}
594: \end{equation}
595: This instability region is located near the origin in the plane
596: $(Q,\,\sigma)$.
597:
598: Another instability region results from (B), which writes
599: \begin{equation} \label{eq51}
600: \tilde \omega^3-4\varepsilon\tilde \omega^2-\left({Q^2\over 4}-
601: 2\mathbf{i}\varepsilon\sigma\right)\tilde \omega+4\varepsilon{Q^2\over 4}=0
602: \end{equation}
603: where as in the case of dNLS-II, we measure $\tilde \omega,\,\sigma$
604: and $Q$ in units of $|\mu|\overline{a^{2}_{0}}$.
605: This equation has the same form as (\ref{eq46}). The instability region
606: is obtained numerically and is represented in Fig. 4. Again the
607: $Q\simeq 0$ region is excluded and one finds a critical value
608: $\sigma_c\simeq 0.441$ for $\sigma$, which determines the superior limit
609: of the MI region.
610: \begin{figure}[ht]
611: \label{fig:4}
612: \begin{center}
613: \includegraphics[width=5in]{fig4-all}
614: \end{center}
615: \caption{\small Instability region for dNLS II. Three dimensional plot
616: of $\mathrm{Im}\,\Omega$ versus $Q^2/4$ and $\sigma$; Two different
617: orientantions.}
618: \end{figure}
619:
620: \section{Conclusions}
621:
622: Several conclusions result from this analysis. For the extended NLS
623: equation (\ref{eq4}) the instability region in a DAMI analysis is
624: dependent on the wave-vector $\vec{k}$ of the carrying wave. In the
625: case of derivative NLS equations the long wave-length region is either
626: forbidden, or highly unfavourable for the development of MI.
627:
628: In the SAMI instead of the $k$-dependence the instability region is
629: now dependent on the correlation length in the initial 2-point correlation
630: function. Although we got explicit results only for a $\delta-$ and
631: Lorentzian distribution function we conclude that the MI is possible
632: only for long-range correlations in the initial state. In the case of a
633: $\delta$-function initial distribution, in a limit situation corresponding
634: to an infinite correlation length, the results are similar with those
635: obtained from DAMI analysis; for the dNLS equations again the long
636: wave-length region is highly unfavourable. The same conclusions can be
637: drawn for the more realistic case of a Lorentzian initial distribution
638: function, namely that the MI region is restricted to a finite region of
639: ($Q,\;\sigma$) plane containing the axis $\sigma = 0$. A critical value
640: $\sigma_c$ is determined for both dNLS equations and again the $Q \simeq 0$
641: region is unfavourable for the instability development.\\
642: Another realistic initial condition is a Gaussian distribution function
643: $$
644: f(k)=\overline{a^{2}}{0} {1 \over\sqrt{2\pi}\sigma}\mathrm{e}^
645: {-\frac{k^2}{2\sigma^2}}
646: $$
647: Calculations done only for NLS eq. (see Alber 1978) reveal the same
648: conclusions, namely the MI is possible only for long-range
649: correlations in the initial state ($\sigma$ small) and in the
650: long wave-length region. The result is more complicated, being
651: expressed through some special functions (complex integral function
652: $\omega(z)$, Abramowitz and Stegun 1965).
653:
654: Such calculations for dNLS equations are not yet performed.
655:
656: The influence of the statistical properties of the medium on the
657: development of large amplitude waves, is a problem of special interest
658: in hydrodynamics, especially for the generation of freak waves in
659: the ocean. In a linear theory there is no interaction between ocean
660: waves. A focusing phenomenon of wave energy may occur only when
661: constructing interference takes place. The situation is competely
662: different when nonlinear wave-wave interaction is present. Then a
663: wave can borrow energy from its neighbours and due to this extra-focusing
664: phenomenon, waves with very high amplitude can be generated. In this
665: process the knowledge of the initial conditions of the oceanic waves
666: is essential. Accordingly, many experimental works have been conducted.
667: For a long time the Joint North Sea Wave Project (JONSWAP) has studied
668: the power spectrum of the oceanic waves (Komen et al 1994), resulting a
669: non-Gaussian expression. It was used to calculate the probability to
670: generate large amplitude oceanic waves (analytically and numerically)
671: in given realistic conditions (Janssen 1983, Onorato et al 2001,
672: 2003, 2004).
673:
674: Calculations using other initial distribution functions, reflecting
675: different physical situations, would be of real interest
676: for sure, even in the case of derivative NLS equations.
677:
678: \noindent\underline{\bf Acknowledgements}\\
679:
680: The present work was done under the contract CEX-D11-9 with the Ministry
681: of Education and Research from Romania.
682:
683: \section{References}
684:
685:
686: Abdulaev, F.Kh., Darmanyan, S.A., Garnier, J. (2002) in
687: {\it Progress in Optics} {\bf 44}, 303 (ed. E. Wolf, Elsevier)
688: \\
689: Abramowitz, M., Stegun, I., (1965) {\it Handbook of Mathematical
690: Functions} (National Bureau of Stand)
691: \\
692: Alber, I.E. (1978) {\it Proc. Roy. Soc. London A} {\bf 363}, 525
693: \\
694: Benjamin, T.B., Feir, J.E., (1967) {\it J. Fluid Mech.} {\bf 27},
695: 417
696: \\
697: Bespalov, V. I., Talanov, V.I., (1966) {\it Prisma JETP} {\bf 3},
698: 417
699: \\
700: Chen, H.H., Lee, Y.C., (1979) {\it Physica Scripta} {\bf 20}, 490
701: \\
702: Dodd, R.K., Eilbeck, J.C., Gibbon, J.D., Morris, H.C., (1982) {\it
703: Solitons and Nonlinear Wave Equations} (Acad. Press, New York)
704: \\
705: Fedele, R., Anderson, D., (2000) {\it J. Opt. B: Quantum Semiclassical
706: Optics} {\bf 2}, 207
707: \\
708: Fedele, R., Anderson, D., Lisak, M., (2000) {\it Physica Scripta
709: T} {\bf 84}, 27
710: \\
711: Fedele, R., Shukla, P.K., Onorato, M., Anderson, D.,
712: Lisak, M., (2002) {\it Phys. Rev. Lett. A} {\bf 303}, 61
713: \\
714: Grecu, A.T., Grecu, D., Visinescu, A., (2005) {\it Rom. J.
715: Phys.} {\bf 50}, 127
716: \\
717: Grecu, A.T., (2005) {\it Ann. Univ. Craiova, Physics AUC} {\bf 15}
718: (part I), 177
719: \\
720: Grecu, D., Visinescu, A., (2004) in {\it Nonlinear Waves. Classical
721: and Quantum Aspects}, p. 151 (eds. F. Kh. Abdulaev, V. V. Konotop,
722: Kluwer Acad. Publ.)
723: \\
724: Grecu, D., Visinescu, A., (2005) {\it Theor. Math. Phys.} {\bf 144},
725: 927
726: \\
727: Grecu, D., Visinescu, A., (2005) {\it Rom. J. Phys.} {\bf 50}, 137
728: \\
729: Hall, B., Lisak, M., Anderson, D., Fedele, R., Semenov, V.E.,
730: (2002) {\it Phys. Rev. E} {\bf 65}, 035602 R
731: \\
732: Janssen, P.A.E.M., (1983) {\it J. Fluid Mech.} {\bf 133}, 113
733: \\
734: Kaup, D.J., Newell, A.C., (1978) {\it J. Math. Phys.} {\bf 19}, 798
735: \\
736: Komen, J.C., Cavaleri, L., Donelan, M., Hasselman, K.,
737: Hasselman, S., Jansen, P.A.E.M., (1994) {\it Dynamics and Modelling of
738: Ocean Waves} (Cambridge Univ. Press)
739: \\
740: Lisak, M., Hall, B., Anderson, D.. Fedele, R., Semenov, V.E., Shukla,
741: P.K., Hasegawa, A., (2002) {\it Physics Scripta T} {\bf 98}, 12
742: \\
743: Mio, K., Ogino, T., Minami, K., Takeda, S., (1976) {\it J. Phys. Soc.
744: Japan} {\bf 41}, 265, 667
745: \\
746: Mj\o lhus, E., (1976) {\it J. Plasma Physics} {\bf 16}, 321
747: \\
748: Moyal, J.E., (1949) {\it Proc. Cambridge Phyl. Soc.} {\bf 45}, 99
749: \\
750: Nakamura, A., Chen, H.H., (1980) {\it J. Phys. Soc. Japan} {\bf 49}, 813
751: \\
752: Onorato, M., Osborne, A.R., Serio, M., Bertone, S., (2001)
753: {\it Phys. Rev. Lett.} {\bf 86}, 5831
754: \\
755: Onorato, M., Osborne, A., Fedele, R., Serio, M., (2003) {\it Phys.
756: Rev. E} {\bf 67}, 046305
757: \\
758: Onorato, M., Osborne, A.R., Serio, M., Cavaleri, L., Brandini,
759: C., Stansberg C.T., (2004) {\it Phys. Rev. E} {\bf 70}, 067302
760: \\
761: Remoissenet, M., (1999) {\it Waves Called Solitons} (Springer Verlag,
762: Berlin)
763: \\
764: Scott, A., (2003) {\it Nonlinear Science. Emergence and Dynamics of
765: Coherent Structures} (Oxford Univ. Press)
766: \\
767: Visinescu, A., Grecu, D., (2003) {\it Eur. Phys. J. B} {\bf 34}, 225
768: \\
769: Visinescu, A., Grecu, D., (2003) {\it Rom. J. Phys.} {\bf 48}, 787
770: \\
771: Toda, M., Kubo, R., Saito, N., (1995) {\it Statistical
772: Physics} vol. 1 (Springer, Berlin)
773: \\
774: Wigner, E., (1932) {\it Phys. Rev.} {\bf 40}, 749
775:
776: \end{document}
777: