1: \documentclass[12pt]{article}
2: \usepackage{amsfonts,psfig,epsfig,latexsym,amscd,multicol,theorem,}
3: \textheight 22.5cm\topmargin -0.4 in
4: \textwidth 16.8cm
5: \oddsidemargin 0in
6: \evensidemargin 0in
7: \def\baselinestretch{1.0}
8: \newcommand{\R}{{\mathbb{R}}}
9: \newcommand{\Z}{{\mathbb{Z}}}
10: \newcommand{\N}{{\mathbb{N}}}
11: \newcommand{\C}{{\mathbb{C}}}
12: \newcommand{\cx}{{{\mathbb{C}}^\times}}
13: \newcommand{\I}{{\mathbb{I}}}
14: \newcommand{\Ll}{{\mathbb{L}}}
15: \newcommand{\CP}{{\mathbb{C}}{{P}}}
16: \newcommand{\RP}{{\mathbb{R}{{P}}}}
17: \newcommand{\pr}{\partial}
18: \newcommand{\beq}{\begin{equation}}
19: \newcommand{\eeq}{\end{equation}}
20: \newcommand{\bea}{\begin{eqnarray}}
21: \newcommand{\eea}{\end{eqnarray}}
22: \newcommand{\ra}{\rightarrow}
23: \newcommand{\rhu}{\rightharpoonup}
24: \newcommand{\hra}{\hookrightarrow}
25: \newcommand{\ds}{\displaystyle}
26: \newcommand{\cd}{\partial}
27: \newcommand{\wt}{\widetilde}
28: \newcommand{\wh}{\widehat}
29: \newcommand{\M}{{\sf M}}
30: \newcommand{\barr}{\overline}
31: \newcommand{\nid}{\noindent}
32: \newcommand{\orot}{{\cal O}}
33: \newcommand{\hess}{{\sf Hess}}
34: \newcommand{\rat}{{\sf Rat}}
35: \newcommand{\req}{{\sf Rat}^{eq}}
36: \newcommand{\half}{\frac{1}{2}}
37: \newcommand{\su}{{\mathfrak{su}}}
38: \newcommand{\so}{{\mathfrak{so}}}
39: \newcommand{\ol}{\overline}
40: \newcommand{\xv}{{\bf x}}
41: \newcommand{\yv}{{\bf y}}
42: \newcommand{\zv}{{\bf z}}
43: \newcommand{\tr}{{\rm tr}}
44: \newcommand{\cosec}{{\rm cosec}\, }
45: \newcommand{\id}{{\rm Id}}
46: \theoremstyle{plain}
47: \newtheorem{thm}{Theorem}
48: \newtheorem{lemma}[thm]{Lemma}
49: \newtheorem{prop}[thm]{Proposition}
50: \newtheorem{cor}[thm]{Corollary}
51: \newtheorem{conj}[thm]{Conjecture} {\theorembodyfont{\rmfamily}
52: \newtheorem{defn}[thm]{Definition}
53: \newtheorem{remark}[thm]{Remark}
54: \newtheorem{rem}[thm]{Reminder}
55: \newtheorem{eg}[thm]{Example}
56: \newtheorem{ceg}[thm]{Counterexample}
57: \newtheorem{fact}[thm]{Fact}
58: \newtheorem{cordef}[thm]{Corollary/Definition}}
59: %\theoremheaderfont{\scshape}
60: %\renewcommand{\theequation}{\arabic{section}.\arabic{equation}}
61: \newcommand{\news}{\setcounter{equation}{0}}
62: %\newcommand{\newss}{\setcounter{subsection}{2}}
63: %\newcommand{\newp}{\setcounter{thm}{16}}
64: \newcommand{\comment}[1]{}
65: \renewcommand{\theequation}{\thesection.\arabic{equation}}
66:
67: \begin{document}
68:
69: \title{Slow equivariant lump dynamics on the two sphere}
70: \author{J.A. McGlade\thanks{E-mail: {\tt jmcglade@maths.leeds.ac.uk}}\,\, and
71: J.M. Speight\thanks{E-mail: {\tt speight@maths.leeds.ac.uk}} \\
72: School of Mathematics, University of Leeds\\
73: Leeds LS2 9JT, England}
74:
75: \date{}
76:
77: \maketitle
78:
79: \begin{abstract}
80: The low-energy, rotationally equivariant dynamics of $n$ $\CP^1$ lumps on $S^2$
81: is studied within the approximation of geodesic motion in the moduli space of
82: static solutions $\req_n$. The volume and curvature properties of $\req_n$ are
83: computed. By lifting the geodesic flow to the completion of an $n$-fold cover
84: of $\req_n$, a good understanding of nearly singular lump dynamics within this
85: approximation is obtained.
86: \end{abstract}
87:
88: \maketitle
89:
90: \section{Introduction}\label{sec:intro}
91: \news
92: The $\CP^1$ model in $2+1$
93: dimensions is a field theory of Bogomol'nyi type, analogous in many
94: respects to the Yang-Mills-Higgs and abelian Higgs models. It has a
95: topological lower bound on energy, saturated by solutions of a first order
96: self-duality equation. These solutions may be interpreted as topological
97: solitons, called lumps, analogous to monopoles and vortices. They have
98: various physical interpretations in theoretical high energy and condensed
99: matter physics. If space is a Riemann surface $\Sigma$, then static lumps
100: are holomorphic maps $\Sigma\ra\CP^1$, the Cauchy-Riemann condition
101: playing the role of the self-duality equation. The most fruitful approach
102: to understanding the dynamics of $n$ moving lumps is, following Ward
103: \cite{war}, to restrict the field dynamics to $\M_n$, the moduli space of
104: degree $n$ static lumps. This is the geodesic approximation originally
105: proposed by Manton for monopole dynamics \cite{man}. It works well for
106: vortex and monopole dynamics \cite{gibman,sam,stu1,stu2}, though it lacks
107: a rigorous underpinning for lumps. As is well known, the reduced dynamics
108: amounts to geodesic motion in $(\M_n,\gamma)$ where $\gamma$ is the $L^2$
109: metric, defined by the restriction to $T\M_n$ of the kinetic energy
110: functional of the field theory. One important difference between lumps and
111: monopoles or vortices is that $(\M_n,\gamma)$ is geodesically incomplete
112: in the lump case \cite{sadspe}, so the approximation predicts that lumps
113: may collapse and form singularities in finite time.
114:
115: In reducing to the
116: geodesic approximation, we replace a nonlinear hyperbolic PDE (the field
117: equation) by a finite system of nonlinear ODEs (the geodesic equation in
118: $\M_n$).This is clearly a much simpler system in principle. It is still
119: highly nontrivial to study its solutions, however, principally because it
120: is usually impossible to obtain explicit formulae for the metric $\gamma$.
121: The same is true for monopoles and vortices. For these systems,
122: interesting progress has been made by imposing extra rotational symmetries
123: on the geodesic problem, so as to reduce it to low-dimensional
124: submanifolds of $\M_n$
125: \cite{housut}. In the present paper, we apply this technique to $\CP^1$
126: lumps moving on $\Sigma=S^2$, concentrating particularly on the behaviour
127: of geodesics close to the singularities where lumps collapse. The $\CP^1$
128: model is more usually formulated on domain $\Sigma=\C$. This is a bad
129: choice from our viewpoint since the $L^2$ metric is undefined due to the
130: presence of non-normalizable zero modes \cite{war} (though one can study
131: geodesic motion on the leaves of a foliation of $\M_n$ on which these bad
132: zero modes are frozen \cite{lee}). This problem is absent when $\Sigma$
133: is a compact Riemann surface. The choice $\Sigma=S^2$ is particularly
134: natural because then $\M_n$ (though not $\gamma$) coincides with the
135: $\Sigma=\C$ moduli space. Noting that $\CP^1\cong S^2$, if we choose
136: stereographic coordinates $z,W$ on domain and codomain respectively, then
137: a degree $n$ holomorphic map is simply
138: \beq
139: \label{*}
140: \phi:z\mapsto
141: W(z)=\frac{a_0+ a_1z+\cdots+a_nz^n}{b_0+b_1z+\cdots+b_nz^n}=
142: \frac{p(z)}{q(z)}
143: \eeq
144: where $p(z)$ and $q(z)$ have no common roots and at
145: least one of $a_n,b_n$ is nonzero. So $\M_n=\rat_n$, the space of degree
146: $n$ rational maps \cite{wood}. There is a natural open inclusion
147: $\rat_n\hra\CP^{2n+1}$, namely
148: \beq
149: W(z)\mapsto [a_0,a_1,\ldots,a_n,b_0,b_1,\ldots,b_n]
150: \eeq
151: whence $\rat_n$ inherits the
152: structure of a complex manifold. $\rat_n$ is noncompact since it omits
153: from $\CP^{2n+1}$ the complex codimension 1 variety on which $p$ and $q$
154: share roots. As $\phi$ approaches this missing set, one or more lumps
155: collapse to infinitely thin spikes and disappear. It is known that
156: $\gamma$ is K\"ahler with respect to this complex structure \cite{spe}.
157: See \cite{bap,spe} for a comprehensive survey of the geometric properties
158: of $(\rat_1,\gamma)$.
159:
160: In the next section we identify in each $\rat_n$ a
161: totally geodesic submanifold $\req_n$, topologically cylindrical,
162: consisting of those $n$-lumps invariant under a certain $SO(2)$ action.
163: We compute the induced metric on $\req_n$, also denoted $\gamma$, and the
164: total volume of $(\req_n, \gamma)$, which turns out to be finite and,
165: somewhat surprisingly, independent of $n$. In
166: section \ref{lifmet} we study the lift of $\gamma$ to the obvious $n$-fold
167: cover of $\req_n$, itself cylindrical. We show that the lifted metric
168: extends to a metric on $S^2$ which is $C^0$ if $n\geq 2$, $C^1$ if $n\geq
169: 3$ and $C^2$ if $n\geq 4$, and deduce the total Gauss curvature of
170: $\req_n$ for $n\geq 1$. There is strong numerical evidence that
171: $(\req_n,\gamma)$ may be isometrically embedded as a surface of revolution
172: in $\R^3$, and we construct this surface numerically for small $n$.
173: Finally, in section \ref{geoflo} we study the geodesic problem on
174: $(\req_n,\gamma)$ by lifting it to the $n$-fold cover. This allows us, in
175: particular, to gain a good understanding of near singular
176: geodesics.
177:
178:
179: \section{The geometry of $\req_n$} \label{georat}
180: \news
181: There is a natural isometric action of $G=SO(3)\times SO(3)$ on
182: $(\rat_n,\gamma)$
183: descending from the usual action of $SO(3)$ on $S^2\subset\R^3$, namely
184: \beq
185: (\orot_1,\orot_2):\phi\mapsto \orot_1\circ \phi\circ\orot_2^{-1}
186: \eeq
187: where we have used $\orot_i$ to denote both an element of $SO(3)$ and its
188: action on $S^2$ \cite{spe}. Given any subgroup (indeed, subset) $K$ of
189: $G$, the fixed point set $\rat_n^K$ of $K$ in $\rat_n$ is, if a
190: submanifold, a totally geodesic submanifold of $(\rat_n,\gamma)$:
191: geodesics which start on and tangential to $\rat_n^K$ remain on $\rat_n^K$
192: for all subsequent time \cite{rom}. Consider the following subgroup
193: $K\cong SO(2)$:
194: \beq K=\{(R(n\alpha),R(\alpha)):\alpha\in\R\},\quad\mbox{where}\quad R(\alpha)
195: =\left(\begin{array}{ccc} \cos\alpha&-\sin\alpha&0\\
196: \sin\alpha&\cos\alpha&0\\ 0&0&1 \end{array}\right).
197: \eeq
198: Let us denote its fixed point set $\req_n$. For later convenience, we also
199: define a $SO(2)$ subgroup of purely spatial rotations:
200: \beq \label{k0} K_0=\{(0,R(\alpha)):\alpha\in\R\}.
201: \eeq
202: In terms of stereographic coordinates, the action of $K$ is
203: \beq
204: \label{A} W(z)\mapsto e^{in\alpha}W(e^{-i\alpha}z).
205: \eeq
206: We may split $\rat_n$ into $U_0$, the
207: subset on which $b_0\neq 0$ and its complement. On $U_0$, we may uniquely
208: write $W(z)$ in the form
209: \beq
210: W(z)=\frac{a_0+a_1z+\cdots+a_nz^n}{1+b_1z+\cdots+b_nz^n}.
211: \eeq
212: If $W(z)\in
213: U_0\cap\req_n$ then for all $z$ and
214: $\alpha$
215: \bea
216: \frac{a_0e^{in\alpha}+a_1e^{(n-1)i\alpha}z+\cdots+a_nz^n}{1+b_1z
217: e^{-i\alpha}+\cdots+b_ne^{-in\alpha}z^n}
218: &=&\frac{a_0+a_1z+\cdots+a_nz^n}{1+b_1z+\cdots+b_nz^n}
219: \eea
220: by (\ref{A}),
221: and hence
222: \beq
223: a_0=a_1=\cdots=a_{n-1}=b_1=b_2=\cdots=b_n=0,\quad a_n\neq 0.
224: \eeq
225: Any rational map in the complement of $U_0$ may be uniquely written
226: \beq
227: W(z)=\frac{1+a_1z+\cdots+a_nz^n}{b_1z+\cdots+b_nz^n},
228: \eeq
229: since $a_0,b_0$ cannot both vanish, by the no common roots condition.
230: Hence if $W(z)\in \req_n$ and $W(z)\notin U_0$,
231: then for all $z$ and $\alpha$
232: \beq
233: \frac{e^{in\alpha}+a_1e^{(n-1)i\alpha}z+\cdots+a_nz^n}{b_1z
234: e^{-i\alpha}+\cdots+b_ne^{-in\alpha}z^n}=\frac{1+a_1z+\cdots
235: +a_nz^n}{b_1z+\cdots+b_nz^n}
236: \eeq
237: which has no solution. Hence
238: $\req_n\subset U_0$:
239: \beq
240: \req_n=\{az^n:a\in\cx\}.
241: \eeq
242: Clearly $\req_n$ is a
243: noncompact complex submanifold of $\rat_n$ of complex dimension 1,
244: biholomorphic to $S^2\backslash\{0,\infty\}$.
245:
246: Physically, $\req_n$ should
247: be thought of as the space of coincident $n$-lumps located at either the
248: north or the south pole of the domain $S^2$. If $a=\chi e^{i\psi}$, then
249: $\chi\in\R^+$ describes the shape of the $n$-lump, while $\psi$ is its
250: internal phase. The case $n=1$ was described in \cite{spe1}, so let us
251: assume $n\geq 2$. The energy density ${\cal E}=\frac{1}{2}|d\phi|^2$
252: is $K_0$ invariant, ${\cal E}(z)={\cal E}(e^{-i\alpha}z)$, hence
253: independent of $\psi$, and is localized in a band centred on a circle
254: of constant latitude, as illustrated in
255: figure \ref{fig1}. Note that if $\chi\gg 1$ ($\chi\ll
256: 1$), the energy
257: accumulates towards the South pole (North pole), although ${\cal E}$ vanishes
258: identically at the poles themselves. One should bear in mind
259: that geodesics in $\req_n$ correspond to $n$-lump motions in which the
260: shape varies in this one-parameter family and the internal phase
261: simultaneously varies. The coincident lump position occupies only the two
262: polar values, though the band of maximum energy density does move up and
263: down smoothly.
264:
265: \begin{figure}[htb]
266: \centering
267: \begin{tabular}{ccc}
268: \includegraphics[scale=0.3333]{dens90.eps}
269: &
270: \includegraphics[scale=0.3333]{dens.eps}
271: &
272: \includegraphics[scale=0.3333]{dens10.eps}
273: \end{tabular}
274: \caption{The energy density ${\cal E}$ of
275: $W(z)=\chi z^5$ with $\chi=1$, $1/50$ and $1/50000$ respectively. Depicted
276: are vertical cross sections through the graph of ${\cal E}$, plotted
277: radially outwards as a non-negative function on $S^2$. The
278: complete graphs are
279: rotationally symmetric about the vertical axis.}
280: \label{fig1}
281: \end{figure}
282:
283:
284: Note that $K$ invariance is an admissible equivariance
285: constraint for the full field equation also. If we let $z=re^{i\varphi}$,
286: then the $\CP^1$ field equation is
287: \beq
288: \frac{4}{(1+r^2)^2}\left[W_{tt}-\frac{2\bar{W}W_t^2}{1+|W|^2}\right]=
289: W_{rr}+\frac{W_r}{r}+\frac{W_{\varphi\varphi}}{r^2}-
290: \frac{2\bar{W}}{1+|W|^2}\left(W_r^2+\frac{W_\varphi^2}{r^2}\right)
291: \eeq
292: which supports solutions within the $K$ invariant
293: ansatz
294: \beq
295: W(r,\theta,t)=r^na(r,t)e^{in\varphi}
296: \eeq
297: for any $n\in\Z$. While
298: the complex valued function $a(r,t)$ is $C^1$, nonvanishing and has limits
299: at $r=0,\infty$ such solutions have degree $n$. We may regard geodesic
300: flow in $(\req_n,\gamma)$ as the geodesic approximation to this symmetry
301: reduced field dynamics, or as a symmetry reduction of the geodesic
302: approximation to the unreduced field dynamics.
303:
304: The metric on $\req_n$ is $K_0$ invariant and hermitian, so
305: \beq
306: \label{jsg1}
307: \gamma=F(\chi)(d\chi^2+\chi^2d\psi^2)
308: \eeq
309: for some smooth positive
310: function $F$. Let $\sigma:S^2\ra S^2$ denote the isometry $z\mapsto
311: z^{-1}$ (rotation by $\pi$ about the $x_1$ axis), and $\hat{\sigma}$
312: denote the corresponding isometry of $\rat_n$, that is, $\phi\mapsto
313: \sigma\circ\phi\circ\sigma^{-1}$. Since $\hat{\sigma}$ preserves $\req_n$,
314: in coordinates $\hat{\sigma}:(\chi,\psi)\mapsto (\chi^{-1},-\psi)$, it is
315: an isometry of $\req_n$, so from equation
316: (\ref{jsg1}),
317: \beq
318: \label{isom}
319: \hat{\sigma}^*\gamma=\chi^{-4}F(\chi^{-1})(d\chi^2
320: +\chi^2d\psi^2)=\gamma\quad \Rightarrow\quad F(\chi^{-1})\equiv \chi^4
321: F(\chi).
322: \eeq
323: It suffices, therefore, to understand the geometry of the ``hemisphere''
324: of $\req_n$ where $0<\chi\leq 1$. To deduce $F(\chi)$, we must compute
325: the squared $L^2$ norm of the zero mode $\cd/\cd\chi \in
326: T_{(\chi,0)}\req_n$, that is, twice the initial kinetic energy of the
327: field $W(z ,t)=(\chi+t)z^n$:
328: \beq
329: \label{B}
330: F(\chi)=\int_\C\frac{dzd\ol{z}}{(1+|z|^2)^2}\frac{|\dot{W}(z,0)|^2}{(1 +
331: |W(z,0)|^2)^2}=2\pi\int_0^\infty\frac{dr}{(1
332: + r^2)^2}\frac{r^{2n+1}}{(1+\chi^2 r^{2n})^2}.
333: \eeq
334: To be consistent with previous work, we have given both domain and
335: codomain the metric $(1+|z|^2)^{-1}dzd\ol{z}$, or equivalently, radius
336: $\frac{1}{2}$. The $L^2$ metric for maps between spheres of radii $R_1$
337: and $R_2$ is easily deduced from this:
338: \beq
339: \gamma'=16R_1^2R_2^2\gamma.
340: \eeq
341: The even function $F:\R\ra \R^+$ defined in (\ref{B}) is smooth by, for
342: example, repeated application of Lemma 2.2 from \cite{spe}. Since the
343: integrand in (\ref{B}) is rational, $F(\chi)$ can be computed explicitly,
344: in principle, for any $n\in\Z^+$, though in practice the expressions
345: become so complicated as to be useless as $n$ increases. The integral
346: formula (\ref{B}) turns out to be far more useful than the explicit
347: expressions in any case. A striking illustration of this
348: is
349:
350: \begin{prop}\label{prop1}
351: $\req_n$ has volume $\pi^2$, independent of $n$.
352: \end{prop}
353:
354: \nid {\it Proof:}\bea {\rm Vol}(\req_n)&=&\int_0^{2\pi}d\psi\int_0^\infty
355: d\chi\, \chi F(\chi)= 4\pi^2\int_0^\infty
356: d\chi\int_0^\infty\frac{dr}{(1+r^2)^2}\frac{\chi r^{2n+1}}{ (1+\chi^2
357: r^{2n})^2}\nonumber\\
358: &=&4\pi^2\int_0^\infty\frac{dr\, r}{(1+r^2)^2}\int_0^\infty d\chi\, r^n\,
359: \frac{r^n\chi}{(1+(r^n\chi)^2)^2}\nonumber \\
360: \label{C} &=&4\pi^2\left[\int_0^\infty\frac{d\alpha\, \alpha}{(1+
361: \alpha^2)^2}\right]^2= \pi^2,\nonumber
362: \eea
363: where we have applied Tonelli's Theorem \cite{priest}.$\quad\Box$\\
364:
365:
366: Of more direct consequence for the geodesic flow on $\req_n$ is an
367: understanding of the singularity of $\gamma$ as $\chi\ra 0$, hence, by the
368: isometry $\hat{\sigma}$, also as $\chi\ra \infty$. Such understanding is
369: obtained by lifting $\gamma$ to the $n$-fold cover of
370: $\req_n$.
371:
372:
373: \section{The lifted metric} \label{lifmet}
374:
375: \news
376:
377: There is a natural $n$-fold cover of $\req_n\cong\cx$ by $\cx$ itself,
378: namely $\pi:c\mapsto
379: c^n$. In terms of polar coordinates $c=\rho e^{i\lambda}$,
380: $\pi:(\rho,\lambda)\mapsto (\chi,\psi)=(\rho^n,n\lambda)$. The lifted
381: metric $\wt{\gamma}=\pi^*\gamma$ on $\cx$ is
382: \beq
383: \wt{\gamma}=\wt{F}(\rho)(d\rho^2+\rho^2d\lambda^2)\quad\mbox{where} \quad
384: \wt{F}(\rho)=n^2\rho^{2n-2}F(\rho^n).
385: \eeq
386: In fact, rather than deduce an integral formula for $\wt{F}$ from that for
387: $F$, it is easier to compute $\wt{F}(\rho)$ directly as the squared
388: $L^2$ norm of the zero mode $\cd/\cd\rho$ in the family $W(z)=(\rho z)^n$,
389: \beq
390: \label{D} \wt{F}(\rho)=\pi
391: n^2\int_0^\infty\frac{ds}{(\rho^2+s)^2}\frac{s^n}{(1+s^n)^2}
392: \eeq
393: where we
394: have used the substitution $s=(\rho|z|)^2$. Note that $\pi(1/c)=1/\pi(c)$,
395: so $\hat{\sigma}:(\rho,\lambda)\mapsto (\rho^{-1},-\lambda)$ is an
396: isometry of $\wt{\gamma}$, and
397: hence
398: \beq
399: \label{jsg2}
400: \wt{F}(\rho^{-1})\equiv\rho^4\wt{F}(\rho)
401: \eeq
402: just as for $F(\chi)$. The integrand in (\ref{D}) is globally bounded on
403: $(0,\infty)$, independent of $\rho$, by $s^n(1+s^n)^{-2}$, which is
404: Lebesgue integrable if $n\geq 2$. Hence, by the Lebesgue dominated
405: convergence theorem (LDCT \cite{ldct})
406: \beq
407: \lim_{\rho\ra 0}\wt{F}(\rho)=\pi
408: n^2\int_0^\infty ds\lim_{\rho\ra\infty}
409: \frac{1}{(\rho^2+s)^2}\frac{s^n}{(1+s^n)^2} =\pi
410: n^2\int_0^\infty\frac{ds\, s^{n-2}}{(1+s^n)^2}
411: \eeq
412: which is finite and positive for $n\geq 2$. It follows that $\wt{\gamma}$
413: extends to a $C^0$ metric $\ol{\gamma}$ on $S^2=\cx\cup\{0,\infty\}$,
414: smooth away from $0$ and $\infty$. We suspect that $\ol{\gamma}$ is never
415: (i.e.\ is for no $n\in\Z^+$) a smooth metric on $S^2$, but is $C^k$ provided $n\geq k+2$.
416: For our purposes it will suffice to prove this for $k=1$ and
417: $k=2$.
418:
419: \begin{prop}\label{prop3}The $C^0$ metric
420: $\ol{\gamma}=\wt{F}(\rho)(d\rho^2 +\rho^2d\lambda^2)$ on $S^2$
421: is $C^1$ if $n\geq
422: 3$ and $C^2$ if $n\geq 4$.\end{prop}
423:
424: \noindent {\it Proof:} Since
425: $\ol{\gamma}$ is smooth away from $\{0,\infty\}$ and $\hat{\sigma}$ is an
426: isometry it suffices to check that $\ol{\gamma}$ is $C^k$, $k=1,2$, at
427: $\rho=0$. So $\ol{\gamma}$ is $C^1$ if $\lim_{\rho\ra 0}\wt{F}'(\rho)=0$,
428: and is $C^2$ if, in addition, $\lim_{\rho\ra
429: 0}\wt{F}''(\rho)-\wt{F}'(\rho)/\rho=0$. Now
430: \beq\wt{F}'(\rho)=-4\pi
431: n^2\rho\int_0^\infty\frac{ds}{(\rho^2+s)^3}\frac{s^n}{(1+ s^n)^2}=:-4\pi
432: n^2\rho f(\rho).
433: \eeq
434: The integrand of $f$ is dominated by
435: $s^{n-3}(1+s^n)^{-2}$ which is integrable if $n\geq 3$. Hence
436: $\lim_{\rho\ra 0}f(\rho)=\int_0^\infty ds\, s^{n-3}(1+s^n)^{-2}<\infty$ by
437: the LDCT, so $\lim_{\rho\ra 0}\wt{F}'(\rho)=0$ as required. Further,
438: \beq
439: \wt{F}''(\rho)-\frac{1}{\rho}\wt{F}'(\rho)=-4\pi n^2\rho f'(\rho)
440: \eeq
441: and
442: \beq \lim_{\rho\ra 0}f'(\rho)=\lim_{\rho\ra
443: 0}-6\rho\int_0^\infty\frac{ds}{ (\rho^2+s)^4}\frac{s^n}{(1+s^n)^2}=0
444: \eeq
445: if $n\geq 4$ by appeal, once again, to the LDCT.\hfill$\Box$
446: \vspace{0.25cm}
447:
448: This $C^2$ lift property has immediate consequences
449: for the curvature properties of $\req_n$. Let $\kappa$ and $\wt{\kappa}$
450: be the Gauss curvatures of $(\req_n,\gamma)$ and $(\cx,\wt{\gamma})$.
451: Since $\pi$ is by definition a local isometry,
452: $\wt{\kappa}=\kappa\circ\pi$. If $n\geq 4$ then $\wt{\gamma}$ extends to a
453: $C^2$ metric on $S^2$, compact, so $\wt{\kappa}$, and hence $\kappa$, must
454: be bounded in this case. This should be contrasted with $(\req_1,\gamma)$
455: whose Gauss curvature is unbounded above. We may also compute the total
456: Gauss curvature of $\req_n$ exactly:
457:
458: \begin{prop}\label{prop4}
459: The total Gauss curvature of $(\req_n,\gamma)$ is, for $n\geq 1$, $$
460: \int_{\req_n}\kappa=\frac{4\pi}{n}. $$
461: \end{prop}
462:
463: \noindent {\it Proof:} Let $\Delta\subset\cx$ be the wedge $
464: \Delta=\{\rho e^{i\lambda}\, :\, \rho\in\R^+,\,
465: 0\leq\lambda<\frac{2\pi}{n}\}. $ Note that the local isometry
466: $\pi:\cx\ra\req_n$ maps $\Delta$ bijectively onto $\req_n$.
467: Hence
468: \beq
469: \int_{\req_n}\kappa=\int_\Delta\wt{\kappa}=\frac{1}{n}
470: \int_\cx\wt{\kappa}
471: \eeq
472: by $SO(2)$ invariance of $\wt{\gamma}$. If $n\geq 4$, the total Gauss
473: curvature of $(S^2,\ol{\gamma})$ is $4\pi$ since $\ol{\gamma}$ is
474: sufficiently regular to apply the Gauss-Bonnet theorem. The total Gauss
475: curvature of $(\cx,\gamma)$ is also $4\pi$ since $S^2\backslash\cx$ has
476: measure $0$, and the result follows. To cover the cases $n=1,2,3$, one
477: must resort to direct computation. Since
478: \beq
479: \wt{\kappa}=-\frac{1}{\rho\wt{F}(\rho)}\frac{d\,\, }{d\rho}\left(
480: \frac{\rho\wt{F}'(\rho)}{2\wt{F}(\rho)}\right)
481: \eeq
482: we have that
483: \beq
484: \int_\cx\wt{\kappa}=4\pi\int_0^1d\rho\, \rho\wt{F}(\rho)\,
485: \wt{\kappa}(\rho)
486: =-4\pi\left[\frac{\rho\wt{F}'(\rho)}{2\wt{F}(\rho)}\right]_0^1
487: \eeq
488: where we have used the isometry $\hat{\sigma}$ to reduce the $\rho$
489: integral to $(0,1]$. Differentiating the identity (\ref{jsg2}) at
490: $\rho=1$ shows that $\wt{F}'(1)=-2\wt{F}(1)$, whence the result follows
491: provided
492: \beq
493: \lim_{\rho\ra 0}\frac{\rho\wt{F}'(\rho)}{2\wt{F}(\rho)}=0.
494: \eeq
495: We have already noted
496: that $\lim_{\rho\ra 0}\wt{F}(\rho)$ exists and is nonzero for $n\geq 2$,
497: so it remains to show that $\lim_{\rho\ra 0}\rho\wt{F}'(\rho)=0$. This
498: follows from the proof of Proposition \ref{prop3} for $n\geq 3$, and may
499: be checked easily for $n=2$ by computing $\rho\wt{F}'(\rho)$ explicitly
500: (using, for example, Maple) and evaluating the limit by hand. The case
501: $n=1$ again requires us to calculate $\rho\wt{F}'(\rho)$ explicitly, but
502: now also $\wt{F}(\rho)$, take the ratio and then take the limit (using,
503: for example, Maple again). \hfill $\Box$
504: \vspace{0.25cm}
505:
506: The qualitative behaviour
507: of geodesic flow on a surface depends crucially on the sign of $\kappa$.
508: In this connexion we make \begin{conj}\label{conj5}For all $n\geq 1$
509: $(\req_n,\gamma)$ has positive Gauss curvature, and may be isometrically
510: embedded as a surface of revolution in $\R^3$.\end{conj}\noindent There
511: is strong numerical evidence for Conjecture \ref{conj5}. Assume that such
512: an embedding $\xv:\req_n\ra\R^3$ does exist
513: \beq
514: \xv(\chi,\psi)=(\alpha(\chi),\beta(\chi)\cos\psi,\beta(\chi)\sin\psi).
515: \eeq
516: We may construct its generating curve by equating $\gamma$ with the
517: induced metric on
518: $\xv(\req_n)\subset\R^3$,
519: \beq
520: (\alpha'(\chi)^2+\beta'(\chi)^2)d\chi^2+\beta(\chi)^2d\psi^2=F(\chi)
521: (d\chi^2+\chi^2d\psi^2).
522: \eeq
523: This fixes $\beta(\chi)=\chi\sqrt{F(\chi)}$. To construct $\alpha(\chi)$
524: we solve the
525: ODE
526: \beq
527: \label{E}
528: \frac{d\alpha}{d\chi}=\sqrt{F(\chi)}\sqrt{1-\left(1+\frac{\chi
529: F'(\chi)}{2 F(\chi)}\right)^2}
530: \eeq
531: with initial data $\alpha(1)=0$. Clearly, the solution exists
532: whilever
533: \beq
534: \label{F}-1\leq 1+\frac{\chi F'(\chi)}{2 F(\chi)}\leq 1,
535: \eeq
536: which we find numerically holds true for all $\chi$ for
537: $n=1,2,\ldots,6$. Inequality (\ref{F}) has a nice geometric
538: interpretation: let $\xi(\chi)$ be the angle between the $x_1$ axis and
539: the tangent to the generating curve at $(\alpha(\chi),\beta(\chi))$. Then
540: $\sin\xi(\chi)$ is precisely the function bounded in (\ref{F}), so the
541: generating curve exists precisely where $-1\leq\sin\xi(\chi)\leq 1$.
542:
543:
544: We
545: have solved (\ref{E}) numerically for $n=2,\ldots,6$, the resulting
546: generating curves being diplayed in figure \ref{fig2}. Note that each
547: curve is concave
548: \begin{figure}[htb]
549: \centering
550: \includegraphics[scale=0.9]{man26.eps}
551: \caption{Generating curves for $\req_n$, $n=2,\ldots 6$}
552: \label{fig2}
553: \end{figure}
554: down indicating that the surface it generates has positive Gauss
555: curvature. If one changes parameters $\chi\mapsto\rho=\chi^\frac{1}{n}$,
556: one finds
557: \beq
558: \sin\xi(\chi)=\frac{\rho\wt{F}'(\rho)}{2n\wt{F}(\rho)}+\frac{1}{n} \,\,
559: \stackrel{\chi\ra 0}{\longrightarrow}\,\, \frac{1}{n}\qquad (n\geq 1)
560: \eeq
561: by the argument used to prove Proposition \ref{prop4}. Hence $\req_n$,
562: $n\geq 2$, has conical singularities of deficit
563: angle
564: \beq
565: 2\pi(1-\sin\xi(0))=2\pi(1-\frac{1}{n})
566: \eeq
567: at $\chi=0$ and
568: $\chi=\infty$. This gives an alternative interpretation of the proof of
569: Proposition \ref{prop4} in terms of the local Gauss-Bonnet theorem
570: applied to the embedded surface of revolution \cite{localgb}:
571: \beq
572: \int_{\req_n}\kappa=2\pi(\sin\xi(0)-\sin\xi(\infty))=\frac{4\pi}{n}.
573: \eeq
574:
575: \section{The geodesic flow} \label{geoflo}
576: \news
577:
578: Consider the one parameter family of geodesics in $(\req_n,\gamma)$ with
579: initial data $a(0)=1$, $\dot{a}(0)=e^{i\alpha}$,
580: $\alpha\in[0,\frac{\pi}{2}]$. This family contains all geodesics, up to
581: isometries and time rescaling. A convenient way to construct such a
582: geodesic is to lift the initial data to the covering space, $c(0)=1$,
583: $\dot{c}(0)=e^{i\alpha/n}$, solve the geodesic equation in
584: $(\cx,\wt{\gamma})$, then project, $a(t)=(\pi\circ c)(t)=c(t)^n$. Since
585: $\pi$ is a local isometry, $a(t)$ is the required geodesic. The advantage
586: of this is that, for $n\geq 4$, $\wt{\gamma}$ extends to a $C^2$ metric
587: $\ol{\gamma}$ on $S^2$, which is just regular enough to ensure that
588: the geodesic in $(S^2,\ol{\gamma})$ exists for all time (by compactness)
589: and depends continuously on the initial data. The point is that the
590: geodesic equations involve only first derivatives of the metric
591: coefficients, so if these coefficients are $C^2$, the flow function for
592: the geodesic equation is $C^1$, hence locally Lipschitz, which is the
593: minimal requirement for local existence, uniqueness and continuous
594: dependence of solutions of an ODE system. So the lifting procedure allows
595: one to construct reliably geodesics in $(\req_n,\gamma)$ which approach
596: arbitrarily close to the singularities at $\chi=0,\infty$, and even to
597: define an unambiguous continuation of the singular geodesic ($\alpha=0$)
598: (which travels along the curve $\psi=0$ from $\chi=0$ to $\chi=\infty$ in
599: finite time by the estimate of \cite{sadspe}) beyond both the future and
600: past singularities. In the lifted picture, the ``singular'' points
601: $\rho=0$ and $\rho=\infty$ are not special, and the geodesic family
602: varies continuously as it approaches and hits them.
603:
604:
605: Let the closest
606: approach of $|c(t)|$ to $0$ for the $\alpha$ geodesic be $\delta >0$,
607: very small. This is easily computed as a function of $\alpha$ using
608: angular momentum and energy conservation. For $\delta$ sufficiently
609: small, $c(t)$, being $C^2$, will be well approximated by a straight line
610: on the $2\delta$ disk centred on $0$. Hence the projected geodesic
611: $a(t)=c(t)^n$ will wind around the singularity $\chi=0$ $(n-1)/2$ times
612: before exiting the $(2\delta)^n$ disk. To describe the corresponding
613: field dynamics $W(z,t)$, we shall think of a configuration as a smooth
614: distribution of classical spins over physical space $S^2$, as in the
615: Heisenberg model of a ferromagnet. While $a(t)$ is in the $(2\delta)^n$
616: disk, the spins are all aligned almost exactly downwards except in a
617: small neighbourhood of the north pole, where they vary rapidly (in space)
618: in a charge $n$ bubble. Their energy is thus highly concentrated towards
619: the north pole. As $c(t)$ traverses the $2\delta$ disk, the spins precess
620: rapidly $(n-1)/2$ times about the north-south axis. The configuration
621: then spreads out before reforming at the south pole and undergoing a
622: similar rapid precession, and so on, indefinitely. In the limit
623: $\delta\ra 0$, one obtains an extended geodesic in which no spin
624: precession occurs, but the configuration pinches to a point singularity
625: at one pole, then spreads out to pinch at the opposite pole. There is a
626: discontinuous phase flip (rotation of each spin by $\pm\pi$ about the
627: north-south axis) associated with each pinch if $n$ is odd, but not if
628: $n$ is even.
629:
630: \begin{figure}[htb]
631: \centering
632: \begin{tabular}{ccc}
633: \includegraphics[scale=0.38]{n2geo.eps}
634: &
635: \includegraphics[scale=0.38]{n3geo.eps}
636: &
637: \includegraphics[scale=0.38]{n4geo.eps}\\
638:
639: \includegraphics[scale=0.38]{n5geo.eps}
640: &
641: \includegraphics[scale=0.38]{n6geo.eps}
642: &
643: \end{tabular}
644: \caption{Geodesics in the disk $\chi\leq 1$ in $\req_n$
645: when $n=2,3,4,5,6$}
646: \label{fig3}
647: \end{figure}
648:
649: The above description is confirmed by numerical solution of
650: the lifted geodesic problem. The equations were solved using a 4th order
651: Runge Kutta method with variable time step. Energy and angular momentum
652: were conserved to within $10^{-5}\,\%$. Figure \ref{fig3} shows
653: the projected geodesics in various cases. Although we can only prove
654: global existence and continuous dependence of all lifted geodesics for
655: $n\geq 4$, the lifting procedure seems to work well also for $n=2$ and
656: $n=3$. This is not surprising given the presence of conical singularities
657: in these cases, as explained in section \ref{lifmet}. Of course, it is
658: questionable whether geodesics which approach the singularities extremely
659: closely really do accurately model the $\CP^1$ field dynamics. In fact,
660: recent numerical \cite{linsad} and analytic \cite{hasspe,str} work gives
661: some grounds for optimism in the equivariant $n\geq 2$ case.
662:
663:
664: Staying
665: within the geodesic approximation, there are many interesting open
666: questions about the $L^2$ geometry of $\rat_n$ which require a good
667: understanding of its boundary at infinity, so far lacking except in the
668: case $n=1$. For example, is the volume and/or diameter finite? Is the
669: spectrum of the Laplacian continuous or discrete (the answer having
670: implications for quantum lump dynamics)? In this paper we have obtained a
671: comprehensive understanding of the boundary at infinity of a (very) low
672: dimensional totally geodesic submanifold of $\rat_n$ which suggests that
673: constructing natural $n$-fold covers of $\rat_n$ may be a productive line
674: of attack.
675:
676: \section*{Acknowledgements} JAM acknowledges the University of
677: Leeds and EPSRC for financial support, and Zen Harper for useful
678: conversations.
679:
680: \begin{thebibliography}{99}
681:
682:
683:
684:
685:
686: \bibitem{bap} J.M. Baptista,
687: ``Some special K\"ahler metrics on $SL(2,\C)$ and their holomorphic
688: quatization''
689: {\sl J. Geom.\ Phys.} {\bf 50} (2004) 1-27.
690:
691: \bibitem{ldct} Y. Choquet-Bruhat, C. DeWitt-Morette and M. Dillard-Bleick,
692: {\sl Analysis, Manifolds and Physics, Part I} (North-Holland, Amsterdam,
693: 1982) pp.\ 43-44.
694:
695: \bibitem{gibman} G.W. Gibbons and N.S. Manton, ``Classical and quantum
696: dynamics of BPS monopoles'' {\sl Nucl.\ Phys.} {\bf B274} (1986)
697: 183-224.
698:
699: \bibitem{hasspe} M. Haskins and J.M. Speight,
700: ``The geodesic approximation for lump dynamics and coercivity of the
701: Hessian for harmonic maps''
702: {\sl J. Math.\ Phys.} {\bf 44} (2003) 3470-3494.
703:
704: \bibitem{housut} C. Houghton and P.M. Sutcliffe,
705: ``$SU(N)$ monopoles and Platonic symmetry''
706: {\sl J. Math.\ Phys.} {\bf 38} (1997) 5576-5589.
707:
708: \bibitem{lee} R.A. Leese, ``Low energy scattering of solitons in the
709: $\CP^1$ model'' {\sl Nucl.\ Phys.} {\bf B344} (1990) 33-72.
710:
711: \bibitem{linsad} J.M. Linhart and L.A. Sadun, ``Fast and slow blowup in
712: the $S^2$ $\sigma$ model and the $(4+1)$-dimensional Yang-Mills
713: model''{\sl Nonlinearity} {\bf 15} (2002) 219-38.
714:
715: \bibitem{man} N.S. Manton, ``A remark on the scattering of BPS
716: monopoles'' {\sl Phys.\ Lett.} {\bf 110B} (1982) 54-6.
717:
718: \bibitem{localgb} J. Oprea,
719: {\sl Differential Geometry and its Applications}
720: (Prentice-Hall, London, UK, 1997) p205.
721:
722: \bibitem{priest} H.A. Priestley, {\sl Introduction to Integration} (Oxford
723: University Press, Oxford, UK, 1997), {p193}
724:
725: \bibitem{rom} N.M. Rom\~ao,
726: ``Dynamics of $\CP^1$ lumps on a cylinder''
727: preprint {\tt math-ph/0404008}
728:
729: \bibitem{sadspe} L. Sadun and J.M. Speight, ``Geodesic incompleteness in
730: the $\CP^1$ model on a compact Riemann surface'' {\sl Lett.\ Math.\
731: Phys.}{\bf 43} (1998) 329-34.
732:
733: \bibitem{sam} T.M. Samols, ``Vortex scattering'' {\sl Commun.\ Math.\
734: Phys.} {\bf 145} (1992) 149-79.
735:
736: \bibitem{spe1} J.M. Speight, ``Low energy dynamics of a $\CP^1$ lump on
737: the sphere'' {\sl J. Math.\ Phys.} {\bf 36} (1995) 796-813.
738:
739: \bibitem{spe} J.M. Speight, ``The $L^2$ geometry of spaces of harmonic
740: maps $S^2\ra S^2$ and $\RP^2\ra\RP^2$''
741: {\sl J. Geom.\
742: Phys.} {\bf 47} (2003) 343-368.
743:
744: \bibitem{str} M. Struwe,
745: ``Equivariant wave maps in two space dimensions''
746: {\sl Comm.\ Pure Appl.\ Math.} {\bf 56} (2003)815-823.
747:
748: \bibitem{stu1} D. Stuart, ``Dynamics of abelian Higgs vortices in the
749: near Bogomolny regime'' {\sl Commun.\ Math.\ Phys.} {\bf 159} (1994) 51-91.
750:
751: \bibitem{stu2} D. Stuart, ``The geodesic approximation for the
752: Yang-Mills-Higgs equations'' {\sl Commun.\ Math.\ Phys.} {\bf 166} (1994)
753: 149-90.
754:
755: \bibitem{war} R.S. Ward, ``Slowly moving lumps in the $\CP^1$ model in
756: $(2+1)$ dimensions'' {\sl Phys.\ Lett.} {\bf 158B} (1985) 424-8.
757:
758: \bibitem{wood} J.C. Wood,
759: ``Harmonic maps and complex analysis'' in {\sl Lectures of the
760: International Seminar on Complex Analysis and its
761: Applications, vol.\ III, Trieste, 1975}, International Atomic Energy Agency,
762: Vienna, 1976, pp.\ 289-308.
763:
764: \end{thebibliography}
765:
766:
767: \end{document}
768:
769: