1: %--------begin lsdiscuss.tex-----------
2: \section{Discussion}
3: \reseteqnos
4: Our computations support the conjecture that for any
5: focusing-defocusing nonlinearity, solitary waves with spin
6: (encapsulated vortices) are spectrally stable for any value of the
7: spin index $m$ if they have no nodes ($n=0$) and if the standing wave frequency
8: $\omega$ is sufficiently close to the kink frequency $\omegas$,
9: which is determined by the property that
10: the graphs of $y=g(x)$ and $y=\omegas x$ enclose two regions of
11: equal area for $0\le x\le \wmaxs$. (Equivalently, the potential
12: $F=F(s)$ has two zero-height maxima, at $s=0$ and $s=\wmaxs$.)
13:
14: To begin to understand why these waves might be stable,
15: we observe that in the regime when $\omega\approx\omegas$ and $|m|$ is large,
16: the wave profiles develop a structure that may be accessible to analysis.
17: This structure is suggested from Figures \ref{FCQVarySpin} and
18: \ref{FCQVaryOmegam3},
19: and a formal analysis is given in Appendix A.
20: In this regime, the wave amplitude $\ws(r)$ is approximately zero
21: in a large central
22: core region, then rises quickly in a transition layer where $r\approx\Rin$
23: to a wide region where $\ws(r)\approx \wmaxs$, then decreases rapidly
24: in a transition layer where $r\approx\Rout$ to zero.
25:
26: Since the wave amplitude remains approximately constant near $\wmaxs$
27: in a wide annular region where $r$ is large and $\theta$ is slowly varying,
28: the solution is slowly varying in space.
29: Now because the nonlinearity is defocusing for amplitudes near $\wmaxs$
30: ($h'<0$)
31: the spatially constant solution $u=\wmaxs e^{i\omegas t}$ is spectrally
32: stable, by remark \ref{const-stab}.
33: The zero solution is also spectrally stable, but spatially constant
34: solutions with amplitude in the focusing range $0<a<a_0$ are
35: not spectrally stable.
36:
37: In the zones of rapid transition where $r$ is large and $\theta$
38: slowly varying, the wave profiles
39: $\ws(r)$ appear to be well-approximated by a one-dimensional kink
40: solution of the nonlinear Schr\"odinger equation \eq{1nls}.
41: Such a solution has the form $\gamma e^{i\omegas t}\phis(x)$,
42: where $\phis$ is as described in section 2 and
43: where $\gamma$ is a complex constant of unit modulus.
44: This kink connects the zero solution to the
45: solution $\gamma\wmaxs e^{i\omegas t}$.
46: It is significant to note that the kink solution itself is
47: a spectrally stable solution of \eq{1nls}.
48: Because of the invariance of \eq{1nls}, we may take $\gamma =1$ in
49: analyzing the stability of $\gamma e^{i\omegas t}\phis$.
50:
51: \begin{proposition}\label{P.kink}
52: Assume that $g(u)=h(|u|^2)u$ where $h$ is $C^1$ with $h(0)=0$,
53: and that properties (i) and (ii) of the introduction hold.
54: Then \eq{1nls} admits a one-dimensional kink solution
55: $u(x,y,t)= e^{i\omegas t}\phis(x)$, where $\phis$ is the unique solution
56: of \eq{2kink} and \eq{2kinkbc}, and this solution is spectrally
57: stable.
58: \end{proposition}
59:
60: \proof
61: The existence and uniqueness of $\phis$ are easy to establish.
62: To analyze the stability, we perform the linearization of section 3 for
63: $u_0(x,y,t) =e^{i\omegas t}\phis(x)$, obtaining
64: $\partial_t \Psi = A\Psi$, where now
65: \be
66: A= J\left(\Delta-\sigma^2 +\alpha(\phis(x)) + \beta(\phis(x))R\right).
67: \ee
68: Taking the Fourier transform in the transverse variable $y$, we obtain
69: $\partial_t \hat\Psi = \hat{A}\hat\Psi$ with
70: $ \hat{A}=-J(L+\ell^2 I)$,
71: where $\ell$ is the Fourier transform variable and
72: \be
73: L=\pmatrix{L_+ & 0 \cr 0 & L_-}
74: =\pmatrix{-\partial_x^2 -\fs'(\phis) & 0 \cr
75: 0 & -\partial_x^2 -\fs(\phis)/\phis} .
76: \ee
77:
78: Consider the spectrum of $L$. We know
79: $\fs'(0)<0$, $\fs'(\wmaxs)<0$, and $\fs(\wmaxs)=0$, therefore
80: \[
81: \lim_{x\to\pm\infty} \fs'(\phis(x)) <0 \quad\hbox{and}\quad
82: \lim_{x\to\pm\infty}{\fs(\phis(x))}/{\phis(x)}\le 0.
83: \]
84: Hence the essential
85: spectra of $L_+$ and $L_-$ are contained in $[0,\infty)$.
86: Each operator $L_+$ and $L_-$ annihilates a function of definite sign:
87: $$
88: L_-\phis=0 \quad\quad {\rm and} \quad\quad L_+\phis'=0,
89: $$
90: so a standard argument based on the maximum principle implies that $L_+$ and
91: $L_-$ have no negative eigenvalues.
92: Thus $L+\ell^2 I$ has no negative eigenvalues, and it is simple to show that
93: then $J(L+\ell^2)$ has no eigenvalues with nonzero real part. It follows
94: that the spectrum of $A$ is purely imaginary. Thus the kink
95: $e^{i\omegas t}\phis(x)$ is spectrally stable, as claimed.
96: \eproof
97:
98: \noindent
99: The focusing-defocusing properties of the
100: nonlinearity thus permit the existence of a spectrally stable
101: one-dimensional kink.
102:
103: A simplified description of the wave amplitude
104: profiles, for spinning solitary waves when $\omega\approx\omegas$ and
105: $|m|$ is large, is that they have two pieces:
106: \begin{itemize}
107: \item[(i)]
108: an outer kink region where $w(r)\approx \phis(r-\Rout)$,
109: \item[(ii)]
110: an inner kink region where $w(r)\approx \phis(\Rin-r)$.
111: \end{itemize}
112: Both the inner and outer kink radii $\Rin$ and $\Rout$ become large,
113: with $\Rout$ much larger that $\Rin$.
114: Since both pieces approximate a spectrally stable one-dimensional
115: kink solution, it becomes plausible that the whole wave can be stable.
116:
117: \endinput
118: