1: %%% -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
2: %%% RG.tex
3: %%%
4: %%% RG-flow calculation
5: %%% -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-
6:
7: \section{The renormalization-group-flow approach}
8: \label{sec:textscrg-flow-appr}
9:
10: \subsection{The behaviour of the overall scale in general situations}
11: \label{sec:overall-scale}
12:
13: We would like to study the \textsc{rg}-flow for the
14: coupling of the metric in a system without dilaton, \emph{i.e.} for the
15: sigma model
16: \begin{equation}
17: \label{eq:low-dimensional-sigma}
18: \mathcal{L} (c)= \frac{1}{2} \left( c g_{\mu \nu} + B_{\mu \nu}
19: \right) \partial X^\mu \bar \partial X^\nu ,
20: \end{equation}
21: where $g_{\mu \nu}$ is a $d$-dimensional Euclidean metric and $c$
22: and $\lambda$ are constants. The antisymmetric tensor is tuned so
23: that for $c = 1$ the model is conformal\footnote{One should not
24: confuse $c$ with the central charge.} and for simplicity we
25: require $g_{\mu\nu} $ to be Einstein with constant curvature (that
26: is always true \textit{e.g.} for group manifolds, but this case is
27: not exhaustive). Geometrically this means that $H = \di B$ is the
28: torsion that parallelizes the Riemannian metric $g_{\mu \nu}$, and
29: no dilaton is required for conformal invariance.
30:
31: Two-dimensional sigma models are perturbatively renormalizable.
32: %\cite{Zamolodchikov:1986gt}.
33: We will analyze the evolution of a
34: certain class of deformations with special emphasis to their
35: target-space interpretation, in the mini-superspace approximation.
36: We will perform this analysis at one loop although it is in
37: principle tractable at higher order.
38:
39: Following
40: \cite{Alvarez-Gaume:1981hn,Braaten:1985is,Friedan:1980jm,Osborn:1989bu}
41: we define the connection
42: \begin{equation}
43: {\Gamma^-}\ud{\mu}{\nu \rho} = \Chris{\mu}{\nu \rho} - \frac{1}{2} H\ud{\mu}{\nu \rho}
44: \end{equation}
45: and the corresponding Riemann tensor
46: \begin{equation}
47: {R^-}\ud{\mu}{\nu \rho \sigma} = \left( 1 - \frac{1}{c^2} \right) R\ud{\mu}{ \nu \rho \sigma} .
48: \end{equation}
49: In a dimensional-regularization scheme the one-loop counterterm
50: reads:
51: \begin{eqnarray}
52: \delta\mathcal{L}_{(1)} &=& \frac{\mu^\epsilon }
53: {4\pi \epsilon}R^{-}_{\mu \nu} \partial X^\mu \bar \partial
54: X^\nu \\
55: &=& \frac{\mu^\epsilon }
56: {4\pi \epsilon} \frac{R}{d} \left( 1 - \frac{1 }{c^2} \right)g_{\mu \nu} \partial X^\mu \bar \partial
57: X^\nu,
58: \end{eqnarray}
59: where $R$ is the Ricci scalar for $g_{\mu \nu}$. In particular, if
60: we start with an $SU(2)$ \textsc{wzw} model at level $k$, we have
61: $d=3$ and $R=6/k$.
62:
63: From the above expressions, we can determine the behaviour of $c$
64: with respect to the scale $\mu$, captured in the corresponding
65: beta-function,
66: $\beta(c)=\nicefrac{\mathrm{d}c}{\mathrm{d}\log\mu}$. We find:
67: \begin{equation}
68: \beta(c) = \frac{R}{2 \pi d} \left(1 -
69: \frac{1}{c^2} \right),
70: \end{equation}
71: which exhibits an infra-red fixed point at $c=1$ that we already
72: know. It is possible to go beyond one-loop and resum higher-order
73: corrections. In the particular case of \textsc{wzw} models, where
74: the target space is a group manifold, this amounts to a finite
75: renormalization of the radius, or equivalently, a shift of the
76: level of the affine algebra $k \to k + g^\ast$, where $g^\ast$ is
77: the dual Coxeter number of the algebra.
78:
79: It is interesting to remark that given the form of the one-loop
80: counterterm, the evolution equations take the form of a
81: \emph{Ricci flow} where the effect of the $B$ field is taken into
82: account by the connection $\Gamma^-$. In fact the equations above
83: are equivalent to
84: \begin{equation}
85: \frac{\di c g_{\mu \nu}}{\di \log \mu} = \frac{1}{2\pi} {R^-}_{\mu
86: \nu}.
87: \end{equation}
88: As usual one should pay attention to the direction of the flow and
89: it is hence useful to define an \textsc{rg} time variable $\tilde
90: t = - \log \mu$ so that $g_{\mu \nu} (\tilde t)$ describes the
91: evolution of the system going towards the infra-red.
92:
93: In order to compare the result with what we will find in the
94: following, we introduce
95: \begin{equation}
96: \sigma ( \tilde t) = \frac{1}{2} \log c ( \tilde t) .
97: \end{equation}
98: Then the evolution of $\sigma ( \tilde t)$ going toward the
99: infra-red gives:
100: \begin{equation}
101: \label{eq:energy-evol-sigma}
102: \frac{\di \sigma }{\di \tilde t } =
103: -\frac{ R }{4 \pi d} \mathrm{e}^{-2\sigma (\tilde t)}\left(1 - \mathrm{e}^{-4\sigma (\tilde t)} \right)
104: \equiv - \frac{1}{4\pi} V^\prime (\sigma(\tilde t))
105: \end{equation}
106: with $V(\sigma)$ given in Eq. (\ref{eq:pot}). Equation
107: (\ref{eq:energy-evol-sigma})
108: admits the implicit solution
109: \begin{equation}
110: \tilde t = - \frac{2 \pi d }{R} \left( \mathrm{e}^{2\sigma(\tilde t)} -
111: \arg\tanh \left( \mathrm{e}^{2\sigma(\tilde t)} \right) \right) + \mathrm{cst.}
112: \end{equation}
113:
114: To summarize, in $d$ dimensions and for constant-curvature
115: metrics, the running of the one-parameter deformation that
116: maintains the normalization for the topological term is a Ricci
117: flow, and it does not produce any new non-trivial fixed points.
118:
119: \subsection{Generic perturbations of $S^3$: the Halphen system}
120: \label{sec:three-sphere}
121:
122: The case of \textsc{wzw} is of special interest. The underlying
123: algebraic structure allows to switch on more general
124: perturbations, while it remains possible to analyze their
125: simultaneous evolution under the \textsc{rg} flow. Although the
126: issue of metric perturbation has attracted much attention in the
127: past (\textit{e.g.} for $O(3)$ or $O(4)$ sigma models
128: \cite{Fateev:1992tk,Fateev:1996ea}), we will here investigate a
129: class of perturbations, for which the full analysis of the
130: \textsc{rg} flow can also be carried out in the presence of a
131: Kalb--Ramond field.
132:
133: Let us for the moment consider a \textsc{wzw} model on a generic
134: compact semi-simple group $G$. Its action is as in Eq.
135: (\ref{eq:low-dimensional-sigma}) with $c=1$. The corresponding
136: algebra $\mathfrak{g}$ has generators $\set{T^\alpha }$ and
137: structure constants $f^{\alpha\beta}_{\phantom{\alpha\beta}
138: \gamma}$. The standard Killing form is\footnote{We will absorb the
139: level $k$ of the algebra in the currents. Therefore, in terms of
140: the structure constants, $H = f_{\alpha \beta \gamma}J^\alpha
141: \wedge J^\beta \wedge J^\gamma$, where the first indices in $f$
142: are lowered with the Killing metric ($H$ is the standard field for
143: the unperturbed \textsc{wzw} model). Its square is contracted with
144: the new metric though. Explicitly, $
145: H_{\alpha \beta }^2 =
146: f_{\alpha \gamma \delta} f_{\beta \gamma'
147: \delta'}g^{\gamma \gamma'} g^{\delta \delta'} .
148: $}
149: \begin{equation}
150: \di s^2 = \delta_{\alpha\beta} J^\alpha \otimes J^\beta,
151: \end{equation}
152: where $J^\alpha $ are the left currents $J^\alpha = \sqrt{k} \tr
153: (T^\alpha g^{-1} \di g)$ for $g \in G$. We will consider a
154: deformed \textsc{wzw} model with metric
155: \begin{equation}
156: \di s^2 = g_{\alpha\beta} J^\alpha \otimes J^\beta =
157: \sum_\alpha \gamma_\alpha (\mu) J^\alpha \otimes J^\alpha
158: \end{equation}
159: ($\mu$ is the \textsc{rg}-scale) and the standard $B$ field, which
160: is fixed, being a topological term. It is convenient to go to the
161: vielbein of the currents so that in particular the metric reads:
162: \begin{equation}
163: \left( g_{\alpha \beta}\right) =
164: \begin{pmatrix}
165: \gamma_1 (\mu) \\
166: & \gamma_2 (\mu) \\
167: & & \ddots \\
168: & & & \gamma_n (\mu)
169: \end{pmatrix},
170: \label{eq:defmets}
171: \end{equation}
172: where $n = \dim G$ and $\gamma_\alpha (\mu)$ arbitrary positive
173: functions.
174:
175: The \textsc{rg}-flow equations are here
176: \begin{equation}
177: \frac{\di g_{\alpha \beta }}{\di \log \mu} = \frac{1}{2\pi}\left(R_{\alpha \beta } -
178: \frac{1}{4} H_{\alpha
179: \beta}^2\right) = \frac{1}{2 \pi}{R^{-}}_{\alpha \beta },
180: \end{equation}
181: where $R_{\alpha\beta}$ are the components of the Ricci tensor, $H
182: = \di B$ and $H_{\alpha \beta}^2 = H_{\alpha \gamma \delta}
183: H_{\beta}^{\phantom{\beta} \gamma
184: \delta}$.
185: We observe that the \textsc{rg}-flow equations for the
186: perturbation pattern at hand are again governed by a Ricci flow
187: with torsion, as in Sec. \ref{sec:overall-scale}.
188:
189: Although the full analysis is tractable for any compact group $G$,
190: we will here focus on the $SU(2)$, where, as we will see below,
191: the flow equations can be solved explicitly. The metric
192: (\ref{eq:defmets}) has now only three entries: $\gamma_1 (\mu),
193: \gamma_2 (\mu), \gamma_3 (\mu)$.
194: \begin{comment}
195: \begin{equation}
196: g_{\alpha \beta} = k
197: \begin{pmatrix}
198: \gamma_1 (\mu) \\
199: & \gamma_2 (\mu) \\
200: & & \gamma_3 (\mu)
201: \end{pmatrix}. \label{eq:defmets3}
202: \end{equation}
203: \end{comment}
204: Using (\ref{eq:genric}), we obtain the following Ricci tensor:
205: \begin{equation}
206: \left(R_{\alpha \beta} \right)= \frac{1}{2}
207: \begin{pmatrix}
208: \frac{\gamma_1^2 - \left( \gamma_2 - \gamma_3 \right)^2}{\gamma_2
209: \gamma_3} \\
210: & \frac{\gamma_2^2 - \left( \gamma_1 - \gamma_3
211: \right)^2}{\gamma_1 \gamma_3} \\
212: & & \frac{\gamma_3^2 - \left( \gamma_1 - \gamma_2
213: \right)^2}{\gamma_1 \gamma_2}
214: \end{pmatrix},
215: \end{equation}
216: while the Kalb--Ramond term reads:
217: \begin{equation}
218: \left(H^2_{\alpha \beta}\right)= 2
219: \begin{pmatrix}
220: \frac{1}{\gamma_2 \gamma_3} \\
221: & \frac{1}{\gamma_3 \gamma_1} \\
222: & & \frac{1}{\gamma_1 \gamma_2}
223: \end{pmatrix}.
224: \end{equation}
225: We introduce as before an \textsc{rg}-time pointing towards the
226: infra-red,
227: \begin{equation}
228: \di \tilde t = - \frac{1}{2\pi \gamma_1(\mu) \gamma_2 (\mu) \gamma_3(\mu)} \di \log \mu ,
229: \end{equation}
230: where we have also reabsorbed the product of the three
231: $\gamma_\alpha$'s. These are positive functions and such a
232: rescaling is therefore harmless since it does not spoil the
233: monotonic evolution of $\tilde{t}$ with $\mu$ ($\tilde{t}$
234: increases from minus infinity to plus infinity when flowing to the
235: infra-red from infinite to zero $\mu$). Putting everything
236: together one obtains the following \textsc{rg} equations:
237: \begin{equation}\label{eq:gamtor}
238: \begin{cases}
239: \displaystyle{ 2 \dfrac{\dot{\gamma}_1}{\gamma_1} = \left(\gamma_2 - \gamma_3
240: \right)^2 - \gamma_1^2 + 1} ,\\[1.2em]
241: 2 \dfrac{\dot{\gamma}_2}{\gamma_2} = \left(\gamma_3 - \gamma_1
242: \right)^2 - \gamma_2^2 + 1 , \\[1.2em]
243: 2 \dfrac{\dot{\gamma}_3}{\gamma_3} = \left(\gamma_1 - \gamma_2
244: \right)^2 - \gamma_3^2 + 1 ,
245: \end{cases}
246: \end{equation}
247: where dot denotes the derivative with respect to $\tilde{t}$.
248:
249: The evolution of Ricci flows on general homogeneous or locally
250: homogeneous spaces has been studied from various perspectives (see
251: \textit{e.g.} \cite{ChowKnopf, Isenberg:1992}). In the absence of
252: torsion, the last constant term in (\ref{eq:gamtor}) is missing
253: and the flow converges towards the round sphere, of
254: \emph{vanishing} radius though. The presence of torsion does not
255: alter this behaviour but affects the radius of the sphere which
256: stabilizes to $\sqrt{k}$ because all $\gamma_\alpha$'s now
257: converge to one\footnote{Note that the differential systems under
258: consideration are parabolic, hence not invariant under time
259: reversal.}. This non-trivial infra-red fixed point corresponds to
260: the $SU(2)_k$ \textsc{wzw} model.
261:
262: The above results on the convergence of the flow are based on
263: asymptotic analysis. However, as already advertised, the
264: Ricci-flow equations (\ref{eq:gamtor}) can be solved explicitly in
265: the case at hand. Indeed, setting
266: \begin{equation}
267: \tilde{t} =
268: \log (T + T_0), \label{eq:12}
269: \end{equation}
270: which amounts in identifying $T\in [0, +\infty[$ with a monotonic
271: function of the inverse energy scale $\nicefrac{1}{\mu}$, and
272: \begin{align}
273: \Omega_1 = \frac{\gamma_2 \gamma_3}{T + T_0} , &&
274: \Omega_2 = \frac{\gamma_3 \gamma_1}{T + T_0} , &&
275: \Omega_3 = \frac{\gamma_1 \gamma_2}{T + T_0} , \label{eq:6}
276: \end{align}
277: equations (\ref{eq:gamtor}) are recast as:
278: \begin{equation}
279: \label{eq:halphen}
280: \begin{cases}
281: \dot{\Omega}_1 = \Omega_2 \Omega_3 - \Omega_1 \left(\Omega_2
282: + \Omega_3 \right) , \\
283: \dot{\Omega}_2 = \Omega_3 \Omega_1 - \Omega_2 \left(\Omega_3
284: + \Omega_1 \right) , \\
285: \dot{\Omega}_3 = \Omega_1 \Omega_2 - \Omega_3 \left(\Omega_1 +
286: \Omega_2 \right) ,
287: \end{cases}
288: \end{equation}
289: where the dot now denotes the derivative with respect to the new
290: time\footnote{In the case of a system without torsion one would
291: obtain the same equations by defining $\Omega_1 = \gamma_2\,
292: \gamma_3$ and cyclic and keeping the same time variable $\tilde
293: t$.} $T$. The arbitrary constant $T_0$ does not play any
294: significant role.
295:
296: This is the celebrated Halphen system that was studied in the 19th
297: century, also called Darboux--Halphen because the equations were
298: written by Darboux on the analysis of triply orthogonal surfaces
299: \cite{Darboux} and solved three years later by Halphen
300: \cite{halph1,halph2}. It has appeared since then in several
301: instances in physics\footnote{The connection to Ricci flows was
302: also noticed by K. Sfetsos in unpublished work.}, as \textit{e.g.}
303: in the search for self-dual or anti-self-dual reductions in
304: four-dimensional Euclidean gravity (Bianchi IX class with $SU(2)$
305: isometry) \cite{Gibbons:1979xn} or in the scattering of
306: $SU(2)$-Yang--Mills monopoles \cite{Atiyah:1985dv}. The function $
307: Y= -2\left(\Omega_1 + \Omega_2 + \Omega_3 \right)
308: $
309: satisfies
310: \begin{equation}\label{eq:chaz}
311: \dot{\ddot{Y}}=2Y\, \ddot{Y} -3\dot{Y}^2,
312: \end{equation}
313: known as the Chazy equation \cite{chaz1,chaz2}. Note that
314: integrable ordinary differential equations like Chazy's or
315: Halphen's turn out to be systematically related to self-dual
316: Yang--Mills reductions.
317:
318: The Halphen and the Chazy equations possess remarkable properties.
319: If $\omega_\alpha (z)$ provide a solution to the Halphen system
320: for generic $z \in \mathbb{C}$ (the dot stands then for
321: $\nicefrac{\mathrm{d}}{\mathrm{d}z}$), then so do
322: \begin{equation}
323: \tilde \omega_\alpha (z) = \frac{1}{\left(c z + d \right)^2} \omega_\alpha
324: \left( \tfrac{a z + b}{c z + d} \right) + \frac{c}{c z + d} ,
325: \label{eq:modtr}
326: \end{equation}
327: with $\left(
328: \begin{smallmatrix}
329: a & b \\ c & d
330: \end{smallmatrix}
331: \right) \in PSL(2, \setC )$. Similarly, if $y(z)$ is a solution
332: of the Chazy equation, so is
333: \begin{equation}
334: \tilde y (z) = \frac{1}{\left(c z + d \right)^2} y
335: \left( \tfrac{a z + b}{c z + d} \right) - \frac{6c}{c z + d}.
336: \label{eq:modty}
337: \end{equation}
338:
339: %This property enables one to
340: %generate any solution corresponding to a given set of initial
341: %conditions $\omega_\alpha^0$ from one particular one.
342:
343: A class of solutions are expressed as \cite{Takhtajan:1992qb}
344: \begin{equation}
345: \label{eq:Ealpha}
346: \omega_\alpha (z)= - \frac{1}{2} \frac{\di}{\di z } \log E_\alpha (z) ,
347: \end{equation}
348: where $E_\alpha $ form a triplet of modular forms of weight two
349: for $\Gamma(2) \subset PSL(2,\mathbb{Z})$ (see \textit{e.g.}
350: \cite{ford, serre}), transforming thereby as
351: \begin{equation}
352: z\to-\frac{1}{z}:\quad \begin{pmatrix}
353: E_1 \\
354: E_2 \\
355: E_3
356: \end{pmatrix} \to z^2\begin{pmatrix}
357: 0 & 1& 0\\
358: 1 & 0& 0\\
359: 0 & 0& -1
360: \end{pmatrix}
361: \begin{pmatrix}
362: E_1 \\
363: E_2 \\
364: E_3
365: \end{pmatrix}\label{eq:Sgam}
366: \end{equation}
367: and
368: \begin{equation}
369: z\to z+1 :\quad \begin{pmatrix}
370: E_1 \\
371: E_2 \\
372: E_3
373: \end{pmatrix}\to
374: -\begin{pmatrix}
375: 1 & 0& 0\\
376: 0 & 1& 0\\
377: 0 & 0& 1
378: \end{pmatrix} \begin{pmatrix}
379: E_1 \\
380: E_2 \\
381: E_3
382: \end{pmatrix}.
383: \end{equation}
384: The class of solutions under consideration \emph{does not} capture
385: situations where some $\omega_\alpha$'s are equal. In our physical
386: set-up, this would correspond to deformations of the three-sphere
387: that preserve some extra isometries, whereas all-different
388: $\omega_\alpha$'s restrict the isometry group to $SU(2)$,
389: strictly. We will discuss this possibility towards the end of the
390: present section, but the following should already be stressed:
391: Eqs. (\ref{eq:halphen}) imply that if $\Omega_\alpha=\Omega_\beta$
392: then $\dot{\Omega}_\alpha=\dot{\Omega}_\beta$ and therefore
393: $\Omega_\alpha$ and $\Omega_\beta$ never cross unless they are
394: equal at any time. For the moment, we will focus on the situation
395: where $\Omega_1\neq\Omega_2\neq\Omega_3$. We can express the
396: $E_\alpha$ as
397: \begin{equation}\label{eq:parteisenhal}
398: E_1 = \frac{\nicefrac{\di \lambda}{\di z}}{\lambda},
399: \quad E_2 = \frac{\nicefrac{\di \lambda}{\di z}}{\lambda -1},
400: \quad E_3 =\frac{\nicefrac{\di \lambda}{\di z}}{\lambda(\lambda
401: -1)},
402: \end{equation}
403: and recast the Halphen system in terms of the Schwartz equation
404: for $\lambda$:
405: \begin{equation}\label{eq:schwartz}
406: \frac{\nicefrac{\di ^3\lambda}{\di z^3}}{\nicefrac{\di \lambda}{\di z}}
407: -\frac{3}{2}\left(\frac{\nicefrac{\di ^2\lambda}{\di z^2}}
408: {\nicefrac{\di \lambda}{\di z}}\right)^2 =
409: -\frac{1}{2}\left(\frac{1}{\lambda^2}+\frac{1}{(\lambda-1)^2}-\frac{1}{\lambda(\lambda-1)}\right)
410: \left(\frac{\di \lambda}{\di z}\right)^2.
411: \end{equation}
412: The elliptic modular function
413: \begin{equation}\label{lamell}
414: \lambda = \frac{\vartheta_2^4}{\vartheta_3^4}
415: \end{equation}
416: is a celebrated solution for this equation (see App.
417: \ref{sec:theta} for a reminder on Jacobi theta functions). This
418: provides a particular solution of the Halphen system in terms of
419: the three $\Gamma(2)$ weight-two \emph{Eisenstein series}.
420: Correspondingly,
421: \begin{equation}\label{eq:parteisencha}
422: y = 12 \frac{\di }{\di z } \log \eta,
423: \end{equation}
424: where $\eta$ is the Dedekind function.
425:
426: For our present purposes, we must focus on real solutions
427: $\Omega_\alpha$ for real time $T$ ($z=iT$). These are obtained as
428: \begin{equation}\label{eq:omreal}
429: \Omega_\alpha(T) = i \omega_\alpha(iT) =- \frac{1}{2} \frac{\di }{\di T } \log E_\alpha
430: (iT)
431: \end{equation}
432: and
433: \begin{equation}
434: Y(T) = i y(iT) = \frac{\di }{\di T } \log E_1 \, E_2 \, E_3.
435: \end{equation}
436: The modular properties of the functions under consideration set
437: stringent constraints between the asymptotics of the solution and
438: its initial conditions: large-$T$ and small-$T$ regimes are
439: related by $T \leftrightarrow \nicefrac{1}{T}$. Assuming
440: $\Omega_\alpha^0\equiv \Omega_\alpha^{\vphantom 0}(0)$
441: \emph{finite} and using (\ref{eq:Sgam}) and (\ref{eq:omreal}), the
442: asymptotic behaviour ($T \to \infty$) is found to be
443: \emph{universally}
444: %\footnote{The
445: %Jacobi functions do not follow from positive initial values and
446: %consequently $\Omega_1 <0$ while $\Omega_2, \Omega_3 > 0$. This is
447: %not in contradiction with the existence of well behaved
448: %all-positive solutions.}
449: %\begin{equation}
450: % \log E_\alpha (iT) \sim -2 \log T + \mathrm{const.\ }
451: %\end{equation}
452: %and hence
453: \begin{equation}\label{eq:largeT}
454: \Omega_\alpha = \frac{1}{T} + \mathrm{subleading}.
455: %\mathcal{O}\left(\mathrm{e}^{-T}\right).
456: \end{equation}
457: This provides an elegant proof of the universality of generic
458: $SU(2)$ Ricci flows in the presence of torsion towards the
459: corresponding \textsc{wzw} infra-red fixed point.
460:
461: One can further assume $\Omega_\alpha^0$ \emph{finite and
462: positive}. This is natural for describing the initial deformation
463: of a three-sphere and sufficient to show that \emph{all
464: $\Omega_\alpha$ remain positive} at any later time. Indeed,
465: suppose that $0<\Omega_1^0<\Omega_2^0<\Omega_3^0$ and that
466: $\Omega_1$ has reached at time $t_1$ the value $\Omega_1^1=0$,
467: while $\Omega_2^1, \Omega_3^1>0$. From Eqs.~(\ref{eq:halphen}) we
468: conclude that at time $t_1$,
469: $\dot{\Omega}_2^1=\dot{\Omega}_3^1=-\Omega_2^1\, \Omega_3^1<0$ and
470: $\dot{\Omega}_1^1=\Omega_2^1\, \Omega_3^1>0$. This latter
471: inequality implies that $\Omega_1$ vanishes at $t_1$ while it is
472: increasing, passing therefore from negative to positive values.
473: This could only happen if $\Omega_1^0$ were negative, which
474: contradicts the original assumption. However, if indeed
475: $\Omega_1^0<0$ and $\Omega_2^0,\Omega_3^0>0$, there is a time
476: $t_1$ where $\Omega_1$ becomes positive and remains positive
477: together with $\Omega_2$ and $\Omega_3$ until they reach the
478: asymptotic region, where they all satisfy (\ref{eq:largeT}). The
479: generic behaviour of a positive-initial-value solution is given in
480: Fig.~\ref{fig:gen-sol-halph}.
481:
482: \begin{figure}
483: \begin{center}
484: \includegraphics[width=.6\linewidth]{GenericHalphen}
485: \end{center}
486: \caption{A generic solution of the Halphen system for
487: positive $\Omega_1^0<\Omega_2^0<\Omega_3^0$.}
488: \label{fig:gen-sol-halph}
489: \end{figure}
490:
491: Although the Halphen system can be solved, its deep nature makes
492: it difficult to establish the correspondence between a given set
493: of initial conditions and the modular forms $E_\alpha$ (see Eqs.
494: (\ref{eq:Ealpha}) or (\ref{eq:omreal})) necessary to provide the
495: actual solution. Moreover, it is important to stress that
496: solutions exist, for which the initial conditions
497: $\Omega_\alpha^0$ are not all positive and not even finite. Hence,
498: for these solutions, (\ref{eq:largeT}) does not hold. This happens
499: \emph{e.g.} for the particular solution given in Eqs.
500: (\ref{eq:parteisenhal}) and (\ref{lamell}), where $T=0$ is a
501: simple pole with positive (unit) residue for $\Omega_2, \Omega_3$
502: and a double pole with negative residue ($-\nicefrac{\pi}{2}$) for
503: $\Omega_1$. As a consequence, $\Omega_1$ is negative and
504: increases, at large $T$, exponentially towards zero, while
505: $\Omega_3$ is positive and decreases exponentially towards zero;
506: $\Omega_2$ is positive and decreases exponentially towards
507: $\nicefrac{\pi}{2}$. Solutions with negative $\Omega$'s were
508: considered in Refs. \cite{Atiyah:1985dv,Gibbons:1986df} for the
509: description of the configuration manifold of two $SU(2)$
510: monopoles.
511:
512: The existence of poles is generic for \emph{all} solutions of the
513: Halphen system. For solutions corresponding to a set of finite
514: and different initial values, these poles are pushed behind $T=0$.
515: This is related to the following general property: Halphen's
516: solutions possess a natural movable boundary
517: \cite{Takhtajan:1992qb}. They exist in a domain of $\mathbb{C}$,
518: where they are holomorphic and single valued, and this domain has
519: a boundary that contains a dense set of essential singularities.
520: The precise location of this boundary depends on the initial
521: conditions.
522:
523: We would like finally to mention that the Halphen system admits
524: also solutions which are \emph{not} based on elliptic functions.
525: These solutions have power-like dependence on time:
526: \begin{equation}
527: \label{eq:polsol}
528: \begin{cases}
529: \Omega_1=\frac{1}{T+A}+\frac{C}{(T+A)^2},
530: \\
531: \Omega_2=\Omega_3=\frac{1}{T+A}.
532: \end{cases}
533: \end{equation}
534: They are actually the \emph{most general} solutions with
535: $\Omega_2=\Omega_3$ and describe axisymmetric deformations of the
536: three-sphere\footnote{Notice that the equations at hand being
537: first order, setting $\Omega_2^0=\Omega_3^0$ guarantees that
538: $\Omega_2=\Omega_3$ at any subsequent time.}, namely deformations
539: preserving an $SU(2) \times U(1)$ . The constants $(A,C)$ are
540: arbitrary and determined by the initial conditions. This class is
541: closed under $PSL(2, \setC )$ transformations. Indeed, using
542: (\ref{eq:modtr}), we learn that under $T\to \nicefrac{1}{T}$,
543: $(A,C)\to (\nicefrac{1}{A},\nicefrac{-C}{A^2})$, while for $T\to
544: T+1$, $(A,C)\to(A+1,C)$. Regularity requires $\Omega_{2,3}^0\equiv
545: \nicefrac{1}{A}>0$, which ensures that the pole of $\Omega_\alpha$
546: is located at $T<0$. Furthermore, the asymptotic behaviour is
547: again universal as in Eq. (\ref{eq:largeT}). No new fixed point
548: appears therefore in this case either. The \textsc{rg} flow of
549: perturbations of this type, namely with $\gamma_1 \neq \gamma_2=
550: \gamma_3$, was analyzed in \cite{Fateev:1996ea}, without torsion
551: though. When translated into the language of self-dual
552: four-dimensional Euclidean metrics with $SU(2)$ isometry,
553: solutions (\ref{eq:polsol}) correspond to the general Taub--NUT
554: family, including Eguchi--Hanson metrics \cite{Gibbons:1979xn}.
555:
556: The case of $\Omega_1=\Omega_2=\Omega_3$ corresponds to fully
557: isotropic deformations studied in the previous section. In this
558: case, the solution is unique: $\Omega_\alpha=\nicefrac{1}{T+A}$.
559:
560: To summarize and conclude the present analysis, solutions of the
561: Halphen system fall in three classes:
562: $\Omega_1\neq\Omega_2\neq\Omega_3$,
563: $\Omega_1\neq\Omega_2=\Omega_3$ and $\Omega_1=\Omega_2=\Omega_3$.
564: The corresponding three-manifolds, target spaces of the
565: sigma-model, are homogeneous with isometry group $SU(2)$, $SU(2)
566: \times U(1)$ and $SU(2)\times SU(2)$, respectively. This captures
567: all possible isometry groups for a general three-sphere, the
568: latter, most symmetric case corresponding to the usual round
569: sphere. The Ricci flow describing the renormalization of the
570: sigma-model leads unavoidably to the round sphere, which is
571: therefore the unique perturbative infra-red fixed point, found at
572: one loop.
573:
574: %%% Local Variables:
575: %%% mode: latex
576: %%% TeX-master: "Friction"
577: %%% End:
578: