1: \documentclass[oneside]{amsart}
2: \usepackage{amsfonts,latexsym,amssymb,amsmath,amscd}
3: \usepackage[english]{babel}
4: \usepackage[dvips]{graphicx}
5: \usepackage{enumerate}
6: \usepackage{amscd}
7:
8:
9: %definiciones
10: \newtheorem{exm}{Exemple}
11: \newtheorem{teo}{Theorem}
12:
13: \newtheorem{prop}{Proposition}
14: \newtheorem{lem}{Lemma}
15: \newtheorem{defi}{Definition}
16: \newtheorem{rem}{Remark}
17: \newcommand{\be}{\begin{enumerate}}
18: \newcommand{\ee}{\end{enumerate}}
19:
20: %texto
21: \newcommand{\ds}{\displaystyle}
22: \newcommand{\no}{\noindent}
23: \newcommand{\tx}{\textrm}
24: \newcommand{\cl}[1]{\mathcal #1 }
25: \newenvironment{prueba}{\no \textbf{Proof:}\:\:}{\hfill \mbox{$\square$}\par}
26: %conjuntos
27:
28: \newcommand{\R}{{\mathbb R}}
29: \newcommand{\PC}{{\mathbb P}_{\C}}
30: \newcommand{\Z}{{\mathbb Z}}
31: \newcommand{\N}{{\mathbb N}}
32: \newcommand{\Q}{{\mathbb Q}}
33: \newcommand{\C}{{\mathbb C}}
34: \newcommand{\X}{{\mathcal X}}
35: \newcommand{\eps}{\varepsilon}
36:
37: \DeclareMathOperator{\diff}{Diff}
38:
39: %analisis
40:
41: \newcommand{\parcial}[2]{\ds\frac{\partial #1}{\partial #2}}
42: \newcommand{\parcialf}[1]{\ds \frac{\partial f}{\partial #1}}
43:
44: %foliaciones
45: \newcommand{\F}{\mathcal{F}}
46: \newcommand{\G}{\mathcal{G}}
47: \renewcommand{\H}{\mathcal{H}}
48:
49: \hyphenation{di-fe-ren--ci-a-ble}
50:
51: \begin{document}
52: \title{On the Quasi-ordinary cuspidal foliations in $(\C^3,0)$}
53: \author{Percy Fern\'andez-S\'anchez \and Jorge
54: Mozo-Fern\'{a}ndez}
55: \address{Percy Fern\'{a}ndez S\'{a}nchez: Instituto de Matem\'{a}tica y Ciencias Afines \\ Universidad
56: Nacional de Ingenier\'{\i}a y
57: Pontificia Universidad Cat\'{o}lica de Per\'{u} \\ Casa de las
58: Trece Monedas \\ Jr. Ancash 536 \\ Lima 1 \\ Per\'{u}}
59: \email{pefernan@pucp.edu.pe}
60: \address{Jorge Mozo Fern\'{a}ndez: Depto. Matem\'{a}tica Aplicada \\
61: University of Valladolid \\ETS de Arquitectura \\ Avda. Salamanca, s/n \\
62: 47014 Valladolid \\ Spain} \email{jmozo@maf.uva.es}
63: \thanks{Both authors partially supported by CONCYTEC (Peru) under the research project 551-OAJ (2004), and by Junta de
64: Castilla y Le\'{o}n (Spain) under project VA123/04.}
65: \date{\today}
66: \maketitle
67:
68: %\begin{abstract}
69: %.
70: %\end{abstract}
71:
72: %-------------------------------------------------------------------------------------------
73:
74: % introduction and motivation
75:
76: %-------------------------------------------------------------------------------------------
77:
78: \section{Introduction and motivation}
79:
80: We would like to study the reduction of the singularities and the
81: analytic classification, in some cases that we shall describe, of
82: germs of singular holomorphic foliations in $(\C^3,0)$, with
83: non-zero linear part. Consider, more generally, $\omega$ a germ in
84: $(\C^n,0)$ of an integrable $1$-form, and let
85:
86: \[
87: \omega=\omega_1+\omega_2+\cdots
88: \]
89:
90: \no be its decomposition in homogeneous forms
91: ($\omega_i=\sum_{j=1}^{n}A_{ij}dx_i$, $A_{ij}$ homogeneous
92: polynomial of degree $i$). Suppose, moreover, that $\omega_1
93: \not\equiv 0$. In general, we can write
94:
95: \[
96: \omega_1=\sum_{i,j}^n c_{ij}x_jdx_{i}, \,\,\,\, c_{ij}\in \C.
97: \]
98:
99: \no The integrability condition $\omega \wedge d\omega=0$ implies
100: that $\omega_1\wedge d\omega_1=0$. Let $C$ be the matrix
101: $(c_{ij})_{i,j} \in M_{n\times n}(\C)$. Writing down explicitly the
102: integrability condition, the coefficient of $dx_i \wedge dx_j \wedge
103: dx_k$ ($i<j<k$) in $\omega_1\wedge d\omega_1$ is
104:
105: \[
106: c_i(c_{kj}-c_{jk})-c_j(c_{ki}-c_{ik})+c_k(c_{ji}-c_{ij}),
107: \]
108: where $c_i=\sum_{j=1}^n c_{ij}x_j$. Two cases appear:
109: \begin{enumerate}
110: \item $C$ is a symmetric matrix. \item $C$ is not symmetric. So,
111: it exists $(j,k)$ with $c_{kj}\ne c_{jk}$.
112: \end{enumerate}
113: In the last case, the polynomials $c_i$, $c_j$, $c_k$ are linearly
114: dependent for every $i$, $j$, $k$, and so $rk(C)\leq 2$. Moreover,
115: $d\omega_1(0)=d\omega(0)\ne 0$, so we are in presence of a
116: Kupka-type phenomenon and, in fact, it exists a biholomorphism $f$
117: such that $f^*\omega\wedge \eta=0$, where $\eta$ is a form in
118: $2$-variables. For bidimensional phenomena, lots of work have been
119: done.
120:
121: We then focus on the symmetric case. A linear change of
122: coordinates changes $C$ in $P^tCP$, $P$ invertible, so we can
123: suppose $C$ diagonal and moreover
124:
125: \[
126: \omega_1=\sum_{i=1}^r x_idx_i,\,\,\,\, r\leq n.
127: \]
128:
129: If $r=n$, G. Reeb, in his thesis \cite{R} shows that there always
130: exists a holomorphic first integral. The behaviour of the foliation
131: is, then, the behaviour of a function. Using Malgrange's singular
132: Frobenius theorem \cite{Ma} we recover this result.
133:
134: If $r<n$, some work was done by R. Moussu \cite{Mo} under additional
135: hypothesis. The fundamental paper of Mattei and Moussu \cite{MM}
136: completes the mentioned results. Let us recall in this case,
137: briefly, the 2-dimensional situation. The foliations studied are
138: defined by 1-forms $ydy+\cdots $. Following Takens \cite{T} such a
139: foliation has a formal normal form
140:
141: \[
142: \omega_N=d(y^2+x^n)+x^pU(x)dy
143: \]
144:
145: \no where $n\geq 3$, $p\geq 2$, $U(x)\in \C[[x]]$, $U(0)\ne 0$.
146:
147: The generic case ($n=3$) was studied by Moussu \cite{Mo2} and a
148: generalization ($n\geq 3$, $2p>n$) by Cerveau and Moussu
149: \cite{CM}. In both cases, the reduction of the singularities of
150: $\omega$ (and $\omega_N$) agrees with the reduction of the curve
151: $y^2+x^n=0$. Projective holonomy classifies and, generically,
152: there is a rigidity phenomenon formal / analytic. If $n$ is even
153: and $2p=n$, it has been studied by Meziani \cite{Me} under some
154: restrictions on the values of $U(0)$. If $2p<n$
155: the study was done (not in full
156: generality) by Berthier, Meziani and Sad \cite{BMS}. We shall call
157: ``cuspidal'' to these foliations.
158:
159: The objective of the present work is to generalize this situation to
160: dimension three. We want to study foliations whose linear part is
161: given by $d(x^2+y^2)$ or by $d(z^2)$. In this paper we shall focus
162: in the case $d(z^2)$. A surface that controls the resolutions of the
163: singularities, with an equation $z^2+\cdots=0$ will appear in the
164: considered cases.
165:
166: Let us recall some results about reduction of the singularities of
167: a complex surface, following Hironaka \cite{H}. A surface $X$ in
168: $\C^3$ has an equation
169:
170: \[
171: f=f_{\nu}+f_{\nu+1}+\cdots=0
172: \]
173: where $f_i$ is homogeneous of degree $i$. For such a surface,
174: define, at the origin:
175:
176: \begin{enumerate}
177: \item The tangent cone, $C_{X}$, as the cone $f_{\nu}=0$.
178:
179: \item The Zariski tangent cone $T_X$, as
180: $Spec(\mathcal{M}/\mathcal{M}^2)$, $\mathcal{M}$ being the maximal
181: ideal corresponding to the origin of $\C[[x,y,z]]/(f)$. This is
182: the smallest linear space containing $C_X$.
183:
184: \item The strict tangent cone $S_X$, as the largest linear
185: subspace $T$ of $T_X$ such that $C_X=C_X+T$. The codimension of
186: $S_X$ is the minimum number of variables required to write down the
187: equations of $C_X$.
188: \end{enumerate}
189:
190: The resolution of singularities of an analytic surface $X$ is a
191: problem that may stated as follows: to find a non-singular surface
192: $\widetilde{X}$ and a birational morphism $\widetilde{X}\to X$
193: composed of quadratic (point blow-ups) and monoidal (curve blow-ups)
194: transformations. These must be done in a precise order. The main
195: case to consider is when the three tangent spaces defined above
196: coincide, and the most difficult case is when, moreover, $\dim
197: S_X=2$. In this case, the tangent cone can be written as $z^{\nu}$.
198: The resolution may be controlled by Hironaka's characteristic
199: polyhedra of the singularities \cite{H}. The precise sequence of
200: blow-ups needed can be read in the polyhedra.
201:
202: A kind of surface singularities whose resolution is particularly
203: simple, and combinatorial, are quasi-ordinary singularities. To
204: define them, consider a finite projection $X \stackrel{\pi}{\to}
205: \C^2$ and let $\Delta$ be discriminant locus of $\pi$ (i.e. the
206: projection of the apparent contour). If $\Delta$ has normal
207: crossings the singularities of $X$ are called quasi-ordinary.
208:
209:
210: Quasi-ordinary singularities are studied not only because they are
211: relatively simple, but because they arise in the Jungian approach to
212: desingularization. First of all desingularize the discriminant locus
213: in order to obtain quasi-ordinary singularities. Then, the problem
214: (simpler) is to reduce the singularities of a quasi-ordinary
215: surface. Some good references of this are the articles of Giraud
216: \cite{G} and Cossart \cite{Co}.
217:
218: Quasi-ordinary singularities can be parametrized by fractional
219: power series, as branches of curves:
220: \[
221: \left\{\begin{array}{l}
222: x=x \\
223: y=y \\
224: z=\sum_{i,j}c_{ij}x^{\frac{i}{n}}y^{\frac{j}{n}}
225: \end{array}\right.
226: \]
227: By the condition of the discriminant, it can be seen that the set
228: of points $\{(i,j)\in \R^2:c_{ij}\neq 0\}$ is contained in a
229: quadrant $(a,b)+\R_{+}^2$, where $c_{ab}\neq 0$. Characteristic
230: pairs may be defined for this parametrization, as is the case of
231: curves, and they still determine the local topology of the
232: singularity, while the converse is not known \cite{Li}.
233:
234: Coming back to foliations, this is related with the case we shall
235: study. More precisely, we search a class of foliations in $(\C^3,0)$
236: whose reduction process can be read in a quasi-ordinary surface. For
237: the case considered $\omega_1=d(z^2)$, by Weierstrass preparation
238: theorem and Tschirnhausen transformations we find that, in
239: appropriate coordinates, the surface is $z^2+\varphi(x,y)=0$, that
240: is not necessarily a separatrix. The natural generalization of
241: cuspidal foliations will be those with an equation
242: \[
243: \omega=d(z^2+\varphi(x,y))+A(x,y)dz.
244: \]
245:
246: In fact, in a recent work, Frank Loray \cite{L} finds an analytic
247: normal form as
248:
249: \[
250: \omega=dF+zdG+zdz,
251: \]
252:
253: \no where $F,G\in \C\{x,y\}$, for integrable holomorphic foliations
254: with linear part not tangent to the radial vector field. Note that a
255: coordinate change $z\to z-G(x,y)$ in Loray's form gives an equation
256: like our expression for the foliations. This is integrable if and
257: only if $d\varphi \wedge dA=0$, i.e. if $\varphi$, $A$ are
258: analytically dependent. As are shall restrict to the quasi-ordinary
259: case, we have that $\varphi(x,y)=x^py^qU(x,y)$, with $U$ a unit. A
260: convenient change of variable in $x$, $y$, allows us to suppose that
261: $\varphi(x,y)=x^py^q$. Let $d=\gcd (p,q)$ $p=dp'$ $q=dq'$. The
262: integrability condition $d\varphi\wedge dA=0$ is then that
263: $A(x,y)=L(x^{p'}y^{q'})$ where $L(u)\in \C\{u\}$.
264:
265:
266: The plan of this paper is as follows. In section 2 we shall review
267: the notion of simple singularity of a foliation, in the sense
268: defined by Cano and Cerveau \cite{CC}, and its analytic
269: classification according to Cerveau and Mozo \cite{CeMo}. Section 3
270: is devoted to describe the resolution of singularities of the
271: quasi-ordinary foliations we are going to study, and the topology of
272: the exceptional divisor. In section 4, we construct a Hopf fibration
273: associated to the quasi-ordinary foliations, making a reduction of
274: the separatrix to a canonical form. Finally, section 5 is devoted to
275: present the main result of the paper: In the considered cases, the
276: holonomy of a certain component of the exceptional divisor
277: classifies analytically the foliation. The cases we study, as we
278: shall see, are essentially the sames that are studied in dimension
279: two.
280:
281: Some notations used throughout the paper are presented here.
282: $\diff (\C ,0)$ will denote the group (under composition) of germs
283: of analytic diffeomorphisms of $(\C,0)$. If $\Omega$ denotes a
284: holomorphic integrable 1-form, defining a foliation, and $D$ is a
285: component of the divisor obtained after reduction of
286: singularities, ${\mathcal H}_{\Omega, D} : \pi_{1}(D\setminus
287: {\mathcal S})\longrightarrow \diff (\C,0)$ is the holonomy
288: representation, defined over a transversal to $D$ (omitted from
289: the notation), where ${\mathcal S}$ is the singular set of the
290: reduced foliation.
291:
292: {\bf Acknowledgements.-} This paper was begun during the stay of the
293: first author at the University of Valladolid and was finished at the
294: Instituto de Matem\'{a}tica y Ciencias Afines (IMCA), of Lima, and
295: the Pontificia Universidad Cat\'{o}lica de Per\'{u}. We want to
296: thank these institutions.
297:
298:
299:
300:
301:
302: %------------------------------------------------------------------------------------
303:
304: % Singularidades Simples
305:
306: %-------------------------------------------------------------------------------------------
307:
308: \section{Simple singularities of foliations and analytic classification}
309:
310: The process of reduction of singularities for a holomorphic
311: foliation is well known in dimension two. After a finite number of
312: point blow-ups performed in any order, a germ of analytic space
313: and a foliation are obtained, and around the singular points, the
314: foliation is generated by a one-form
315: \[
316: \omega=(\lambda y+h.o.t.)dx + (\mu x+h.o.t.)dy,
317: \]
318: with $\mu\neq 0$, $\frac{\lambda}{\mu}\not \in \Q_{<0}$.
319:
320: The analytic classification is well studied in a wide variety of
321: cases:
322:
323: \begin{enumerate}
324: \item If $\frac{\lambda}{\mu}\not \in \R_{\geq 0}$, $\omega$ is
325: analytically linearizable, i. e., there exists and analytic
326: diffeomorphism $\phi:(\C^2,0)\to (\C^2,0)$ such that
327: \[
328: \phi^*\omega \wedge (\lambda ydx+\mu xdy)=0
329: \]
330:
331: \item If $\frac{\lambda}{\mu}\in \R_{>0} \setminus \Q$, but it is
332: not ``well-approached" by rational numbers, it is also
333: linearizable. If it is well-approached, we face a problem of small
334: divisors, and the situation becomes more complicated.
335:
336: \item If $\lambda=0$ or $\frac{\lambda}{\mu}\in \Q_{+}$, Martinet
337: and Ramis find a large moduli space formal/analytic. In this case
338: the classification of the foliation agrees with the classification
339: of the holonomy of a strong separatrix (i.e. a separatrix in the
340: direction of a non-zero eigenvalue). Moreover in the resonant case
341: ($\frac{\lambda}{\mu}\in \Q_{+}$) or in the saddle-node case
342: ($\lambda=0$) with analytic center manifold, the conjugation of
343: the foliation is fibered. This means the following: choose
344: coordinates $x$, $y$ such that the axis are the separatrices,
345: $y=0$ being a strong one; the foliations are defined by 1-forms
346: \[
347: \omega_i=yA_i(x,y)dx+\mu /\lambda x(1+B_i(x,y))dy,
348: \]
349: with $i=1,2$. Let $h^{(i)} (x)$ be the holonomies of $y=0$, supposed
350: conjugated. Then the foliations are conjugated by a diffeomorphism
351: $\phi(x,y)=(x,yg(x,y))$.
352: \end{enumerate}
353:
354: The singularities obtained after this reduction process are called
355: simple or reduced. The class of simple singularities is stable
356: under blow-ups. Let us observe that the notion of simple
357: singularity is not only analytic, but formal: if $\omega_1$,
358: $\omega_2$ are analytic $1$-forms, and $\hat{\phi}$ is a local
359: diffeomorphism such that $\hat{\phi}^{*}\omega_1\wedge
360: \omega_2=0$, then $\omega_1$ has a simple singularity if and only
361: if $\omega_2$ has.
362:
363: If the dimension of the ambient space is greater or equal than
364: three, the notion of simple singularity has been developed in
365: \cite{CC}, \cite{Ca}, and its analytic classification studied in
366: \cite{CeMo}. The reduction of singularities is only achieved when
367: the dimension of the ambient space is at most three, and in this
368: case, simple singularities are the final ones obtained after the
369: reduction process. Let us summarize here, for convenience of the
370: reader, the main results in dimension three.
371:
372: First of all, let us recall the notion of ``dimensional type". A
373: foliation has dimensional type $r$ if there exist analytic (resp.
374: formal) coordinates such that the foliation is defined by an
375: integrable $1$-form $\omega$ that can be written in coordinates
376: $x_1,\cdots,x_r$ ($r\leq n$), but not less. So, a
377: three-dimensional singularity of foliation has dimensional type
378: $2$ and $3$. For instance, if we are in presence of a Kupka
379: phenomenon, the dimensional type is $2$. The notions of formal
380: dimensional type or analytic dimensional type are equivalent, as
381: seen in \cite{CeMo}. So, we have simple singularities of
382: dimensional types $2$ and $3$. If the dimensional type is $2$,
383: simple singularities are defined by a simple $2$-dimensional
384: $1$-form. They have $2$ separatrices, of which at most one is
385: formal.
386:
387: If the dimensional type is tree, simple singularities are the ones
388: that admit one of the following formal normal forms:
389: \begin{align}
390: \omega=xyz\left(\alpha \frac{dx}{x}+\beta \frac{dy}{y} +\gamma
391: \frac{dz}{z}\right),
392: \end{align}
393: with $\frac{\alpha}{\gamma},\:\frac{\beta}{\gamma},\:
394: \frac{\alpha}{\beta}\not \in \Q_{-}$ (and $\alpha\beta\gamma\neq 0$,
395: as the dimensional type is $3$). This is the linearizable case. If,
396: for instance, some of the quotients is not real, the linearization
397: is analytic \cite{CL}.
398:
399: \begin{align}
400: \omega_N=xyz\left( x^py^qz^r \right)^s\left[\alpha
401: \frac{dx}{x}+\beta \frac{dy}{y}+
402: \left(\lambda+\frac{1}{(x^py^qz^r)^s}\right)
403: \left(p\frac{dx}{x}+q\frac{dy}{y}+ r\frac{dz}{z}\right)\right],
404: \end{align}
405: where $p,q,r\in \N$, $qr\neq 0$, $s\in \N^*$, $\alpha$, $\beta$
406: constants, not both zero. This is the \emph{resonant case}.
407: Several things can be said about foliations that are formally
408: equivalent to this normal form:
409:
410: \begin{enumerate}
411: \item $\F$ has three separatrices, of which at most one is formal
412: (which, in the preceding coordinates, would be $x=0$). This is a
413: confluence of simple two-dimensional singularities defined along the
414: axis. Saddle-nodes only appear if $p=0$, and only in this case the
415: existence of a formal, non convergent separatrix is possible.
416:
417: \item The holonomy group of $z=0$ (strong separatrix) classifies
418: analytically the foliation. Moreover, the conjugations is fibered
419: if the three separatrices are convergent.
420:
421: \item If $\frac{\alpha}{\beta}\not \in \Q$, there is a rigidity
422: phenomenon: every such foliation is analytically equivalent to
423: $\omega_N$.
424: \end{enumerate}
425:
426: A typical case in which we are in presence of a simple singularity
427: and that will appear in the sequel, is when the foliation is defined
428: by a 1-form
429: \begin{align}\label{simple}
430: \omega=xyz\left[(p+A(x,y,z))\frac{dx}{x}+(q+B(x,y,z))\frac{dy}{y}+
431: (r+C(x,y,z))\frac{dz}{z}\right],
432: \end{align}
433: with $p,q,r\in \N^*$,
434: $\nu(A), \nu(B), \nu(C)>0$.
435:
436: More can be said: the transformation $\phi$ that converts $\omega$
437: in its formal normal form $\omega_N$, even if it is not analytic, it
438: is transversally formal and fibered. This means in particular that
439: such a $\phi$ can be found in the form
440: \[
441: \phi(x,y,z)=(x,y,\varphi(x,y,z)).
442: \]
443:
444: The existence of local holomorphic first integrals, according to
445: Mattei and Moussu \cite{MM}, is equivalent to the periodicity of
446: the holonomy group. Moreover, an integrable 1-form $\omega$, that
447: generates a reduced foliation of dimensional type three, has a
448: holomorphic first integral if and only if there exists analytic
449: coordinates $(x,y,z)$ such that
450: $$
451: \omega \wedge (pyz dx+ qxz dy+r xy dz) =0,
452: $$
453: where $p,q,r\in \N^\ast$.
454:
455: %------------------------------------------------------------------------------------
456:
457: % Desingularizacion
458:
459: %-------------------------------------------------------------------------------------------
460: \section{Reduction of singularities and topology of the
461: divisor}\label{section3}
462:
463: In this paper, we shall study the analytic classification of
464: quasi-ordinary cuspidal foliations in dimension three, i.e.,
465: foliations such that, in appropriate coordinates, can be defined
466: by an integrable 1-form
467: $$
468: \omega= d(z^2+x^py^q)+A(x,y)dz.
469: $$
470: The integrability condition here is equivalent to $d(x^py^q)\wedge
471: dA=0$. So, let $d= gcd (p,q)$, $p=dp'$, $q=dq'$. Such a 1-form can
472: be written as
473: $$
474: \omega= d(z^2+x^py^q)+(x^{p'}y^{q'})^k h(x^{p'}y^{q'})dz,
475: $$
476: where $h(u)\in \C \{ u\}$, $h(0)\neq 0$. Fixing $p$, $q$, we shall call $\Sigma_{pq}$
477: the set of holomorphic foliations that are analitically equivalent to the foliation defined by one of these
478: 1-forms.
479:
480: As it will become clear from the development of the paper, the
481: separatrices of this foliation have the equation
482: $$
483: z^2+x^py^q+ h.o.t. =0,
484: $$
485: and Weierstrass preparation theorem and Tschirnhausen transformation
486: show that this separatrix is analytically equivalent to
487: $z^2+x^py^q=0$.
488:
489: The reduction of singularities for these foliations is quite
490: simple, similar to plane curves, and it is the main objective of
491: this section their detailed analysis. For convenience, we divide
492: the problem in three cases:
493:
494: \be[{Case }1.]
495:
496:
497: \item $p$, $q$ even.
498:
499: \item $p$ even, $q$ odd.
500:
501: \item $p$, $q$ odd.
502:
503: \ee
504:
505:
506: \be[{{\bf Case }}1.] \item Suppose $p$, $q$ are even, and
507: $d=2d'$. If $k>d'$, the reduction of the singularities is obtained
508: after $\dfrac{p+q}{2}$ blow-ups: \be \item First of all, blow up
509: $\dfrac{p}{2}$ times the $y$-axis. We obtain a sequence of
510: divisors $D_1,\ldots ,D_{p/2}$, topologically germs $(\PC^1 \times
511: \C , \PC^1 )$. The intersection of two consecutive components is a germ of a
512: line $(\C, 0)$, $L_i=D_i\cap D_{i+1}$, $1\leq i <\dfrac{p}{2}$. In
513: the appropriate chart, these blow-ups have the equations
514: $$
515: \left\{ \begin{array}{rcl}
516: x & = & x \\
517: y & = & y \\
518: t_{i-1} & = &x\cdot t_i,
519: \end{array}
520: \right.
521: $$
522: where $t_0=z$, $1\leq i \leq \dfrac{p}{2}$. \item Then blow-up
523: $\dfrac{q}{2}$ times the $x$-axis, obtaining again a sequence of
524: divisors $D_{\frac{p}{2}+1},\ldots ,D_{\frac{p+q}{2}}$,
525: topologically equal to $(\PC^1 \times \C , \PC^1 )$. Again, the
526: intersection between two consecutive components is a line $L_i=
527: D_i\cap D_{i+1}$, $\dfrac{p}{2}+1 \leq i < \dfrac{p+q}{2}$. Now,
528: the coordinates of the blow-ups are
529: $$
530: \left\{ \begin{array}{rcl}
531: x & = & x \\
532: y & = & y \\
533: t_{i-1} & = &y\cdot t_i,
534: \end{array}
535: \right.
536: $$
537: with $\dfrac{p}{2}<i <\dfrac{p+q}{2}$.
538:
539: The result of the composition of all the blow-ups in the preceding
540: charts is the map $\pi (x,y,t_{\frac{p+q}{2}})=
541: (x,y,x^{\frac{p}{2}}\cdot y^{\frac{q}{2}}\cdot
542: t_{\frac{p+q}{2}})$. The pull-back of the foliation is given by
543: \begin{eqnarray*}
544: \lefteqn{\pi^\ast \omega = x^{p-1}y^{q-1} \cdot \left[ 2xyt dt
545: +(t^2+1)xy \left( p\frac{dx}{x}+q\dfrac{dy}{y}\right) + \right. }
546: & &
547: \\
548: & + & \left. (x^{p'}y^{q'})^{k-d'} h(x^{p'} y^{q'})xyt\cdot
549: \left( \dfrac{p}{2} \dfrac{dx}{x} +\dfrac{q}{2} \dfrac{dy}{y} +
550: \dfrac{dt}{t} \right) \right]
551: \end{eqnarray*}
552: (here $t= t_{\frac{p+q}{2}}$).
553:
554: The foliation, now, is reduced. Let ${\mathcal S}$ be the singular
555: locus of this reduced foliation. ${\mathcal S}$ is an analytic,
556: normal crossing space of dimension one, composed by:
557: \be
558: \item The
559: lines $L_i$ of intersection of the divisors. These are resonnant
560: singular points of dimensional type two.
561: \item The lines $L$, $L'$
562: in $D_{\frac{p+q}{2}}$ of equations $(y=0, t=i)$, $(y=0, t=-i)$,
563: and also the lines $M'$, $M''$ in $D_{\frac{p}{2}}$ of equations
564: $(x=0, t=i)$, $(x=0, t=-i)$ (in the last chart). These lines are
565: the intersections of the two separatrices $S'$, $S''$ with the
566: divisors.
567: \item The intersection $P_i:=D_{\frac{p}{2}}\cap D_i$,
568: $\dfrac{p}{2}<i\leq \dfrac{p+q}{2}$ is a projective line composed of
569: points of dimensional type two, except at the corners: \be
570: \item
571: $m_i=P_i\cap L_i= D_{\frac{p}{2}}\cap D_i \cap D_{i+1}$,
572: $\dfrac{p}{2}<i<\dfrac{p+q}{2}$. These are the resonnant singular
573: points of dimensional type three, having $D_{\frac{p}{2}}$, $D_i$,
574: $D_{i+1}$ as separatrices.
575: \item
576: \begin{eqnarray*}
577: m' & := & D_{\frac{p}{2}}\cap D_{\frac{p+q}{2}} \cap S' = L' \cap
578: M' \cap P_{\frac{p+q}{2}}, \text{ and } \\
579: m'' & := & D_{\frac{q}{2}}\cap D_{\frac{p+q}{2}}\cap S'' = L''
580: \cap M''\cap P_{\frac{p+q}{2}}.
581: \end{eqnarray*}
582:
583: These are the resonant singular points of dimensional type three
584: corresponding to the separatrices of the foliations.
585: \ee
586: \ee
587: \ee
588: According to the preceding description of the resolution of the
589: singularities, we have all the information about the topology of
590: $D_i\setminus {\mathcal S}$, and more precisely about the
591: fundamental group of these components. We have:
592: \begin{itemize}
593: \item $D_1\setminus {\mathcal S}$ is topologically $\C \times \C$,
594: so simply connected.
595: \item $D_i\setminus {\mathcal S}$
596: ($1<i<\dfrac{p}{2}$) is topologically $\C^\ast \times \C$. The
597: generator of the fundamental group is a loop $\gamma_i$ that turns
598: around $L_i$ (or $\gamma_i^{-1}$ around $L_{i-1}$).
599: \item
600: $D_{\frac{p}{2}+1}\setminus {\mathcal S} \cong \C^\ast \times \C$.
601: The fundamental group is generated by a loop $\alpha_i$ around
602: $P_{\frac{p}{2}+1}$.
603: \item $D_i\setminus {\mathcal S} \cong
604: \C^{\ast }\times \C^\ast$ ($\dfrac{p}{2}+1<i<\dfrac{p+q}{2}$). The
605: fundamental group has generators $\gamma_i$ around $L_i$ and
606: $\alpha_i$ around $P_i$, that commute.
607: \item
608: $D_{\frac{p+q}{2}}\setminus {\mathcal S} \cong (\C\setminus \{
609: m',m''\}) \times \C^\ast$. We have one loop $\alpha_{\frac{p+q}{2}}$
610: around $P_{\frac{p+q}{2}}$ and loops $\gamma'$, $\gamma''$ around
611: the separatrices (i.e., around $m'$, $m''$). \item
612: $D_{\frac{p}{2}}\setminus {\mathcal S} \cong \C^2\setminus {\mathcal
613: C}$, where ${\mathcal C}$ is the curve with coordinates
614: $t^2_{\frac{p}{2}}+y^q=0$, composed of two smooth branches that meet
615: tangentially at the origin. In this case (see \cite{La}), $\pi_{1}(
616: \C^2\setminus {\mathcal C})$ is the group, written in terms of
617: generators and relations as
618: $$
619: \pi_{1}(\C^2\setminus {\mathcal C})= \langle \alpha, \beta;
620: \alpha^{\frac{q}{2}}\beta =\beta\alpha^{\frac{q}{2}} \rangle.
621: $$
622:
623: These loops go as follows. Consider the curve $t^2_{\frac{p}{2}}+ y^q=0$
624: on $\C^2$, and cut by $y=1$. You obtain
625: $\C\setminus \{ m', m'' \}$; then $\alpha$ is a loop in $y=1$ that
626: turns around these two points $m'$, $m''$, and $\beta$ is a loop
627: in $t_{\frac{p}{2}}=0$ that turns around the origin. At the end of
628: the reduction process, $\alpha$ is going to be a loop in
629: $D_{\frac{p}{2}}$ around the two separatrices, and $\beta$ a loop
630: around $P_{\frac{p+q}{2}}$ ``between $S'$ and $S''$".
631: \end{itemize}
632:
633: The case $k=d'$ ($2k=d$) is almost identical, except for some
634: values of the coefficient $h(0)$. More precisely, after
635: $\dfrac{p+q}{2} $ blow-ups, in order to obtain the complete
636: reduction of singularities (i.e., simple singular points) it is
637: necessary and sufficient that
638: $$
639: h(0)^2\neq \frac{(16+r)^2}{16+2r},\ \forall\, r\in \Q_{>0}.
640: $$
641: Moreover, if in the preceding expression we put $r=0$, we have
642: then $h(0)=\pm 4$. In this case, only one separatrix is obtained,
643: but it is a three-dimensional saddle-node, the divisor being the
644: weak separatrix (then convergent). We shall assume that this is
645: not the case, i.e., if $k=d'$ we shall assume that
646: $$
647: h(0)^2 \neq \frac{(16+r)^2}{16+2r},\ \forall\, r\in \Q_{\geq 0}.
648: $$
649: The reader may verify that this condition is equivalent to
650: ${\mathcal P}_2$ property in \cite{Me,Me-tesis} (i.e. $h(0)\neq
651: \pm 2 \left( \sqrt{r}+ \dfrac{1}{\sqrt{r}} \right) $, $\forall\,
652: r\in (0,1] \cap \Q$).
653:
654: Suppose now that $k<d'$. In this case, the reduction of
655: singularities is achieved blowing-up $kp'$ times the $y$-axis and
656: $kq'$ times the $x$-axis. After these, in the last chart we obtain
657: as singularities the sets $L'= (x=t=0)$, $M'=(y=t=0)$, $L''= (x=0,
658: t=1)$, $M''= (y=0, t=1)$. These are also two singular points of
659: dimensional type three, namely $m'= L'\cap M'\cap P_{k(p'+q')}$,
660: $m''=L''\cap M'' \cap P_{k(p'+q')}$ (with analogous notations as
661: before), corresponding respectively to the points $(0,0,0)$ and
662: $(0,0,1)$. But now $m''$ is a saddle-node, so the separatrix $S''$
663: is maybe formal. In this paper, we shall assume that always $S''$
664: is convergent, i.e., there is a center manifold.
665:
666: \item Suppose $p$ even, $q$ odd. If $k>d$, the reduction of
667: singularities is obtained after the following sequence of
668: blow-ups. \be \item First, blow-up $p/2$ times the $y$-axis,
669: obtaining divisors $D_1,\ldots ,D_{\frac{p}{2}}$ linked by lines
670: $L_1,\ldots ,L_{\frac{p}{2}-1}$. The equations of these blow-ups
671: are
672: $$
673: \left\{
674: \begin{array}{rcl}
675: x & = & x \\
676: y & = & y \\
677: t_i & = & x\cdot t_{i+1},
678: \end{array}
679: \right.
680: $$
681: where $t_0:=z$, $i<\dfrac{p}{2}$.
682:
683: \item Blow-up $\dfrac{q-1}{2}$ times the $x$-axis, obtaining
684: $D_{\frac{p}{2}+1},\ldots ,D_{\frac{p+q-1}{2}}$ joined by lines
685: $L_i=D_i\cap D_{i+1}$, and $D_i$ joined to $D_{p/2}$ by a
686: projective $P_i$. The equations are
687: $$
688: \left\{
689: \begin{array}{rcl}
690: x & = & x \\
691: y & = & y \\
692: t_i & = & y\cdot t_{i+1},
693: \end{array}
694: \right.
695: $$
696: $\dfrac{p}{2}\leq i <\dfrac{p+q-1}{2}.$
697:
698: \item It appears a tangency in the singular locus. In order to
699: break it, blow-up again the $x$-axis and take a chart centered in
700: the point corresponding to $t_{\frac{p+q-1}{2}}$. The equations
701: are now
702: $$
703: \left\{
704: \begin{array}{rcl}
705: x & = & x \\
706: y & = & s\cdot t_{\frac{p+q-1}{2}} \\
707: t_{\frac{p+q-1}{2}} & = & t_{\frac{p+q-1}{2}},
708: \end{array}
709: \right.
710: $$
711: and we obtain a new component $D'$ such that $D'\cap
712: D_{\frac{p+q-1}{2}}=L_{\frac{p+q-1}{2}}$, $D'\cap
713: D_{\frac{p}{2}}=P'$.
714: \item Finally, blow-up again the $x$-axis, in
715: order to obtain normal crossings. We obtain a final component $D''$
716: and the only separatrix $S$ of the foliation cuts $D'$
717: transversely in a line $L$ (and $D_{p/2}$ in a line $M$). We have
718: $L'=D'\cap D''$ and $P''=D''\cap D_{p/2}$. \ee
719:
720: The singular points of dimensional type three are
721: $m_i:=D_{p/2}\cap D_i\cap D_{i+1}$
722: ($\dfrac{p}{2}<i<\dfrac{p+q-1}{2}$), $m_{\frac{p+q-1}{2}}:=
723: D_{p/2}\cap D_{\frac{p+q-1}{2}}\cap D'$, $m'=D_{p/2} \cap D'\cap
724: D''$. and $m=D_{p/2}\cap D' \cap S$.
725:
726: The topology of the components is as in Case 1. If ${\mathcal S}$
727: is the singular locus, $D_1\setminus {\mathcal S} \cong \C^2$ is
728: simply connected, $D_i\setminus {\mathcal S} \cong \C^\ast \times
729: \C$ if $1<i<\dfrac{p}{2}$, $D_{\frac{p}{2}+1}\setminus {\mathcal
730: S} \cong \C^\ast \times \C$, $D_i\setminus {\mathcal S} \cong
731: \C^\ast \times \C^\ast$ if $\dfrac{p}{2}+1<i<\dfrac{p+q-1}{2}$,
732: $D'\setminus {\mathcal S} \cong (\C\setminus \{ m,m'' \} ) \times
733: \C^\ast$, $D''\setminus {\mathcal S}\cong \C \times \C^\ast$.
734: Finally, $D_{p/2}\setminus {\mathcal S}\cong \C^2 \setminus
735: {\mathcal C}$, where ${\mathcal C}$ is the curve with coordinates
736: $t_{p/2}^2 +y^q=0$. As before,
737: $$
738: \pi_1 (\C^2 \setminus {\mathcal C}) = \langle \alpha, \beta ;
739: \alpha^q=\beta^2 \rangle .
740: $$
741: \vspace{0.5cm}
742: \begin{figure}[h]
743: \begin{center}
744: \includegraphics[height=6.5cm,width=12cm]{nilpotente7.eps}
745: \caption{\footnotesize The reduction of the surface
746: $z^2+x^2y^3=0$}. \label{reduction}
747: \end{center}
748: \end{figure}
749: When $k<d$, as in dimension two, the situation is as in Case 1, with $k<d'$.
750:
751: \item $p$, $q$ odd. Now, the resolution is something different than before.
752: First, blow-up $\dfrac{p-1}{2}$ times the $y$-axis and
753: $\dfrac{q-1}{2}$ times the $x$-axis obtaining $D_1,\ldots
754: ,D_{\frac{p-1}{2}}, D_{\frac{p+1}{2}},\ldots
755: ,D_{\frac{p+q}{2}-1}$. In the new coordinates $(x,y,
756: t:=t_{\frac{p+q}{2}-1})$ the singular locus is given by the three
757: coordinate axis, that corresponds to the intersection of the
758: divisors and the intersection of the cone $t^2+xy=0$ with the
759: divisors.
760:
761: Now, blow-up the origin, obtaining $P$, a projective $\PC^2$. The
762: three coordinate axis, now transverse to $P$, continue being
763: singular. Over $P$, the singular locus is composed by two
764: projective lines and a conic tangent to both lines. In order to
765: finish, blow-up twice each of the axis $x$ and $y$ transverse to
766: $P$, obtaining $D_{(1)}'$, $D_{(1)}''$, $D_{(2)}'$, $D_{(2)}''$.
767:
768: With respect to the topology of the divisors, the only interesting
769: case (i.e., not similar to the preceding ones) to comment is
770: $P\setminus {\mathcal S}$. As we said before, $P\cap {\mathcal S}$
771: is composed by two lines and a regular conic, so
772: $$
773: \pi_1 (P\setminus {\mathcal S}) \cong \langle \alpha, \beta;
774: \alpha^2\beta =\beta \alpha^2 \rangle.
775: $$
776: \ee
777: \section{Reduction of the separatrix to a canonical form}
778:
779: Let $\F$ be a germ of a singular foliation defined on $(\C^3,0)$,
780: and let $\pi:(M,D)\to (\C^3,0)$ be the minimal reduction of the
781: singularities of $\F$ in Cano-Cerveau sense, as described above
782: \cite{CC}. Let $\tilde{\F}$ be the strict transform of the
783: foliation $\F$ by $\pi$ and let $D_i$ be a component of the
784: exceptional divisor $D$.
785:
786: We recall, that a \emph{Hopf fibration} $\H_{\F_{\Omega}}$ adapted to
787: $\F_{\Omega}\in \sum_{pq}$ is a holomorphic transversal fibration
788: $f:M\to D_i$ to the foliation $\F_{\Omega}$, i.e:
789: \begin{enumerate}
790: \item $f$ is a retraction, more precisely, $f$ is a submersion and
791: $f|_{D_i}=Id_{D_i}$.
792:
793: \item The fibers $f^{-1}(p)$ of $\H_{\F_{\Omega}}$ are contained in
794: the separatrices of $\F_{\Omega}$, for all $p\in D_i\cap Sing
795: (\tilde{\F}_{\Omega}) $.
796:
797: \item The fibers $f^{-1}(p)$ of $\H_{\F_{\Omega}}$ are transversal
798: to the foliation $\F_{\Omega}$, for all $p\in D_i\setminus Sing
799: (\tilde{\F}_{\Omega})$.
800: \end{enumerate}
801:
802: We shall be interested in finding a Hopf fibration adapted to the
803: foliation, relative to a particular component of the exceptional
804: divisor. For, if $p$ is even, call $\tilde{D}:= D_{p/2}$, i.e.,
805: the last component obtained after the first sequence of line
806: blow-ups. If $p$ and $q$ are odd, $\tilde{D}:= P$, i.e., the
807: projective obtained after the (only) point blow-up.
808:
809:
810: The task of finding a Hopf fibration associated to the
811: foliation ${\F}_{\Omega}$ is not easy in the actual coordinates
812: $(x,y,z)$. As it is done in the two-dimensional case, to overcome
813: this obstacle, we analyze the desingularization of
814: ${\F}_{\Omega}$
815: in order to obtain a simple equation for the separatrices.
816:
817: From Section (\ref{section3}) we know that the foliation
818: $\F_{\Omega}\in \Sigma_{pq}$ defined by the one-form
819: \[
820: \Omega=d(z^2+(x^{p'}y^{q'})^d)+(x^{p'}y^{q'})^kh(x^{p'}y^{q'})dz,
821: \]
822: has a separatrix analytically equivalent to $S:
823: z^2+(x^{p'}y^{q'})^r=0$ for some $r\in \N$. In order to find a
824: Hopf fibration $\H_{\F}$ of the foliation $\F$, we need to
825: normalize the one-form $\Omega$ such that the foliation defined by
826: this normal form has exactly $S: z^2+(x^{p'}y^{q'})^r=0$ as
827: separatrix, for certain $r$. So, the strict transformed of $S$ by
828: the desingularization is an hyperplane in these coordinates and
829: invariant by Hopf fibration.
830:
831:
832: \begin{prop}\label{normal} The foliation $\F_{\Omega}$ is analytically
833: equivalent to a foliation defined by the one-form
834: \[
835: d(z^2+(x^{p'}y^{q'})^r)+g(x^{p'}y^{q'},z).x^{p'}y^{q'}z\left(2\frac{dz}{z}-p'\frac{dx}{x}-q'\frac{dy}{y}\right),
836: \]
837: where $r=d$ if $2k\geq d$ and $r=2k$ if $2k< d$. In particular,
838: the separatrix of the foliation $\F_{\Omega}$ is analytically
839: equivalent to $S:\:z^2+(x^{p'}y^{q'})^r=0$.
840: \end{prop}
841:
842: \begin{prueba} The foliation $\F_{\Omega}$ is defined by the
843: $1$-form
844: \[
845: \Omega=d(z^2+(x^{p'}y^{q'})^d)+(x^{p'}y^{q'})^kh(x^{p'}y^{q'})dz,
846: \]
847: where $(p,q)=d$, $p=p'd$, $q=q'd$. That is,
848: $\Omega$ is the pull-back of the 1-form
849: $\Omega_0=d(z^2+u^d)+u^kh(u)dz$ by the ramified fibration
850: \[
851: \begin{array}{ccl}
852: \rho: (\C^3,0)&\to& (\C^2,0) \\
853: (x,y,z) &\to & (x^{p'}y^{q'},z)=(u,z).
854: \end{array}
855: \]
856:
857: The equation of the separatrices of $\Omega$ is of the form $z^2+u^d+h.o.t. =0$ (if $d$ is even, this is a joint equation, i.e.,
858: the product of the two separatrices).
859:
860: Using Weierstrass' preparation theorem, we can assume that the local
861: equation of the separatrix is a polynomial in $z$:
862: $z^2+a(u)z+b(u)=0$, with $a(0)=b(0)=0$. If $\Phi_1
863: (u,z)=(u,z-\frac{a(u)}{2})$ is the Tschirnhausen transformation,
864: then the pull-back $\Phi_1^\ast \Omega_0$ has $z^2+c(u)=0$ as
865: separatrix, with $c(u)=b(u)-\frac{a(u)^2}{4}=u^r f(u)$, $f(0)\neq
866: 0$. If $d>2$ (cuspidal case), we have that $\nu (a)>1$, $\nu (b)>2$,
867: and then $r>2$. In fact, $r=d$ when $2k>d$ or $r=2k$ when $2k\leq d$
868: (see \cite{CM,BMS,Ca}). Similar computations are valid when $d=1$ or
869: $d=2$ (in these cases, $2k\geq d$).
870:
871: Let us write this reduced equation of the separatrices as
872: $$
873: \frac{z^2}{f(u)}+u^r=0,
874: $$
875: and let $f(u)^{1/2}$ be a square root of the unit $f(u)$. If $ \Phi_2 (u,z)= (u,z\cdot f(u)^{1/2})$, and $\Phi:= \Phi_1 \circ \Phi_2$, then $\Phi^\ast \Omega_0$ has $z^2+u^r=0$ as separatrix. This map has the form
876: $$
877: \Phi (u,z)= \left(
878: u,z\cdot f(u)^{1/2}-\frac{a(u)}{2}
879: \right) .
880: $$
881:
882: Consider the diagram
883: $$
884: \begin{CD}
885: \C^3 @>{\rho}>> \C^2 \\
886: @VFVV @VV{\Phi}V \\
887: \C^3 @>{\rho}>> \C^2
888: \end{CD}.
889: $$
890:
891:
892: We want to find a diffeomorphism $F= (F_1,F_2,F_3)$ that makes commutative the diagram, i.e., that
893:
894: \[
895: (F_1^{p'}F_2^{q'},F_3)=\left(x^{p'}y^{q'},zf(x^{p'}y^{q'})^{\frac{1}{2}}-\frac{a(x^{p'}y^{q'})}{2}\right).
896: \]
897: For, we may choose $F_1=x$, $F_2=y$, $F_3=z\cdot f(x^{p'}y^{q'})^{1/2}- \dfrac{a(x^{p'}y^{q'})}{2}$.
898: The form $\Phi^* \Omega_0$, having $z^2+u^r=0$ as a separatrix is,
899: up to a unit, $d(z^2+u^r)+g(u,z)(2udz-dzdu)$, so $F^*\Omega_0$
900: defines the same foliation that
901: \[
902: d(z^2+(x^{p'}y^{q'})^r)+g(x^{p'}y^{q'},z).x^{p'}y^{q'}z\left(2\frac{dz}{z}-p'\frac{dx}{x}-q'\frac{dy}{y}\right).
903: \]
904: We reproduce part of the proof presented in \cite{CM} in order to
905: find the transformation $F$ fibered.
906: \end{prueba}
907:
908: As a consequence of this normal form for $\F_{\Omega}$, there
909: exists coordinates $(x,y,z)$, such that the separatrix $S$ of the
910: normal form is given by the equation: $z^2+(x^{p'}y^{q'})^r=0$,
911: where $r$ is as in the Proposition (\ref{normal}), and not only
912: ``analytically equivalent to". Now, we can find a Hopf fibration,
913: from a holomorphic vector field $X_1$ for which $S$ is an invariant
914: set, that is
915: \[
916: X_1=\left\{\begin{array}{ll}
917: x\parcial{}{x}+\dfrac{p}{2} z\parcial{}{z}, & p \tx{ is even} \\
918: x\parcial{}{x}+y\parcial{}{y}+\left(\frac{p+q}{2}\right)
919: z\parcial{}{z},& p \tx{ and } q \tx{ are odd}.
920: \end{array}\right.
921: \]
922: So, we have that the Hopf fibration $\H_{\F_{\Omega}} \: (f:M\to
923: \tilde{D})$, adapted to the foliation defined by the one-form
924: $\Omega\in \sum_{pq}$ will be determined (not uniquely) by a
925: linearizable singularity of a holomorphic vector field
926: $X=X_1+X_2+\cdots$.
927:
928: Having defined a Hopf fibration adapted to $\F_{\Omega}$, we can
929: define the holonomy of the leaf $\tilde{D}\setminus
930: Sing(\tilde{\F}_{\Omega})$ respect to this fibration. In order to
931: determine it, we fix a point $p_0\in \tilde{D}\setminus
932: Sing(\tilde{\F}_{\Omega})$. Over this point we have a transversal
933: $f^{-1}(p_0)$ and by path lifting construction, a representation of
934: the fundamental group of $\tilde{D}\setminus
935: Sing(\tilde{\F}_{\Omega})$ in $\diff (\C,0)$ is determined, denoted
936: by ${\mathcal H}_{\Omega,\tilde{D}}$
937: \[
938: {\mathcal
939: H}_{\Omega,\tilde{D}}:\pi_1(\tilde{D}\setminus
940: Sing(\tilde{\F}_{\Omega}),p_0)\to Diff(\C,0).
941: \]
942: This representation is independent of $p_0$ modulo conjugacy and its
943: image will
944: be called the \emph{exceptional holonomy} and denoted
945: $H_{\Omega,\tilde{D}}$.
946:
947:
948:
949: \section{Classification of the singularities}
950:
951: From section (\ref{section3}) we know that the homotopy group
952: $\pi_1(\tilde{D}\setminus Sing(\tilde{\F}_{\Omega});t_0)$ can be
953: generated by two elements $\alpha$ and $\beta$ in all the cases
954: considered, with different relations in each case:
955: \begin{enumerate}
956: \item If $p$ and $q$ are even:
957: $\alpha^{\frac{q}{2}}\beta=\beta\alpha^{\frac{q}{2}}$,
958:
959: \item If $p$ is even and $q$ is odd: $\alpha^q=\beta^{2}$,
960:
961: \item If $p$ and $q$ are odd: $\alpha^2\beta=\beta\alpha^2$.
962: \end{enumerate}
963:
964: If $\gamma$ is an element of the homotopy group, let us denote
965: $h_\gamma$ its image by the map ${\mathcal H}_{\Omega, \tilde{D}}$
966: in the exceptional holonomy. This holonomy can be generated by
967: $h_{\alpha}$, $h_{\beta}$, which at least satisfy the same relations
968: than $\alpha$, $\beta$. But in some cases, these relations may be
969: improved. The following proposition collects some of these
970: improvements:
971:
972:
973: \begin{prop}
974: \begin{enumerate}
975: \item If $p$ is even,
976: $h_{\alpha}^{\frac{p}{2}}=id$.
977: \item If $p$ is even and $q$ is odd,
978: $h_{\alpha}^{\frac{p}{2}}=h_{\beta}^{p'}=id$.
979: \end{enumerate}
980:
981: \end{prop}
982:
983: \begin{proof}
984:
985: Consider $p$ even. After $\dfrac{p}{2}$ blow-ups, the
986: strict transform of the separatrix ${\mathcal S}$ is given by a
987: surface analytically equivalent to $t^2_{p/2}+y^q=0$. This singular
988: surface is a cylinder over a curve, that
989: is either a cuspidal curve of characteristic pair $(2,q)$ or a
990: couple
991: of regular curves tangent at the origin at order $\frac{q}{2}$.
992: Applying Picard-Lefschetz techniques, it can be seen that
993: the loop $\alpha$ is a simple curve contained in the plane
994: $y=\varepsilon$, with $|\varepsilon |$ small enough, that turns
995: around the points $(t,y)=(\pm i\cdot
996: \eps^{q/2}, \eps )$. Thus, the holonomy $h_{\alpha}$ is
997: completely determined
998: by the holonomy of a loop that turns around the line
999: $D_{p/2-1}\cap
1000: D_{p/2}$. Along this line, the foliation is a reduced foliation
1001: of dimensional type 2 (in fact, we are in presence of a Kupka
1002: phenomenon), and its analytic type is determined by a
1003: two-dimensional section transversal to the $y$-axis. This
1004: foliation has a linearizable, periodic holonomy, and $h'_{\alpha
1005: } (0)= e^{-2\pi i \cdot \frac{p-2}{p}}$.
1006:
1007: If, moreover, $q$ is odd, the periodicity of $h_{\alpha}$ implies
1008: the periodicity of $h_{\beta}$, and so, $h_{\beta}$ is linearizable,
1009: $h'_{\beta}(0)= e^{2\pi i \frac{q'}{p'}}$. Nevertheless, it does not
1010: mean necessary that the holonomy group $H_{\Omega, \tilde{D}}$ is
1011: linearizable, as in particular we don't know if it is abelian or
1012: not.
1013: \end{proof}
1014:
1015:
1016: %\begin{rem}
1017: %Dar informacion de estos grupos. (Loray).
1018: %\end{rem}
1019:
1020: The following theorem contains the main result of the paper. In the
1021: proof, several techniques from \cite{CeMo,BMS,CM,Me} are frequently
1022: used, and we shall not enter in details about them.
1023:
1024: \begin{teo} Let $\Omega_1$, $\Omega_2$ be elements of
1025: $\sum_{pq}$. Consider the foliations $\F_{\Omega_1}$ and
1026: $\F_{\Omega_2}$, and their exceptional holonomies
1027: $H_{\Omega_i,\tilde{D}}= <h_{\alpha}^i, h_{\beta}^i >$, $i=1,2$,
1028: defined as before. Then, the foliations are analytically
1029: conjugated if and only if the couples $(h_{\alpha}^i, h_{\beta}^i
1030: )$ are also analytically conjugated, i.e., if and only if there
1031: exists $\Psi \in \diff (\C, 0) $ such that $\Psi^\ast
1032: h^1_{\gamma}= h^2_{\gamma }$, where $\gamma =\alpha, \beta$.
1033: \end{teo}
1034:
1035: \begin{prueba} If the foliations are conjugated then clearly
1036: their exceptional holonomies are also conjugated. Conversely,
1037: suppose that the exceptional holonomies are conjugated via $\Psi$,
1038: and let $\tilde{\F}_{\Omega_1}$, $\tilde{\F}_{\Omega_2}$ be the
1039: desingularized, reduced foliations. Because of the existence of
1040: the Hopf fibration relative to $\tilde{D}$, $\Psi$ can be extended
1041: to a neighbourhood of $\tilde{D}$, away from the singular points.
1042: These singular points are the intersections of $\tilde{D}$ with
1043: the other components of the divisor, and with the separatrix (the
1044: separatrices in the even-even case). All these points are singular
1045: points of dimensional types two or three, and for all of them,
1046: $\tilde{D}$ is a strong separatrix. In this situation, the
1047: conjugation of the holonomies of $\tilde{D}$ implies conjugation
1048: of the reduced foliations in a neighbourhood of the singular
1049: points \cite{CeMo}.
1050:
1051: So, we have that $\tilde{\F}_{\Omega_1}$, $\tilde{\F}_{\Omega_2}$
1052: are conjugated in a neighbourhood of $\tilde{D}$. Suppose now that
1053: $p$ is even. We need to conjugate the foliations also in a
1054: neighbourhood of $D_1,\ldots D_{p/2-1}$. As $D_1$ is simply
1055: connected, its holonomy is trivial. So, the holonomy of $D_2$,
1056: generated by one loop around $L_1=D_1\cap D_2$ is periodic (the
1057: argument is the same as in \cite{MM}). The same argument shows
1058: that $D_i$ has a periodic holonomy, $1\leq i<\dfrac{p}{2}$, and
1059: so, the foliations have first integrals in a neighbourhood of each
1060: $L_i$, $1\leq i<\dfrac{p}{2}$. These are points of dimensional
1061: type two. By analogous reasons as in the two-dimensional case,
1062: $\Psi$ can be extended to a neighbourhood of the exceptional
1063: divisor, so, $\F_{\Omega_1}$, $\F_{\Omega_2}$ are conjugated
1064: outside the singular locus, which has codimension two. We conclude
1065: using Hartogs' theorem to extend the conjugation to a
1066: neighbourhood of the origin.
1067:
1068: Suppose now that $p$, $q$ are odd. $\tilde{\F}_{\Omega_1}$ and
1069: $\tilde{\F}_{\Omega_2}$ are conjugated in a neighbourhood of
1070: $\tilde{D}$ (that is a projective ${\mathbb P}^2_\C$ in this case).
1071: The fundamental group of $D_{\frac{p-1}{2}}$ is generated by only
1072: one loop, that, after the resolution, can be seen as a loop around
1073: $D_{\frac{p-1}{2}} \cap D''_{(1)}$. This is one of the loops that
1074: generates the holonomy of $D_{\frac{p-1}{2}}$ locally at the reduced
1075: singular points, and following similar arguments as in the preceding
1076: cases, and as the two-dimensional case, the foliation is
1077: linearizable around these points. Let us detail, in this case, how
1078: the use of first integrals allows the extension of the conjugation.
1079:
1080: Consider, for instance, the singular point $D_{\frac{p-1}{2}}\cap
1081: D_{\frac{p+q}{2}-2}\cap D_{\frac{p+q}{2}-1 }$, with coordinates
1082: $(x',s',t')$ as in picture \ref{coordinates} .
1083: \begin{figure}[h]
1084: \begin{center}
1085: \includegraphics[height=8cm,width=12cm]{nilpotente8.eps}
1086: \caption{\footnotesize Coordinates $(x',s',t')$ the singular point
1087: $D_{\frac{p-1}{2}}\cap D_{\frac{p+q}{2}-2}\cap D_{\frac{p+q}{2}-1
1088: }$}. \label{coordinates}
1089: \end{center}
1090: \end{figure}
1091:
1092:
1093: We have a conjugation $\Psi$ between the foliations
1094: $\F_{\Omega_1}$ and $\F_{\Omega_2}$ defined over an annulus
1095: $$
1096: \{ |x'| <\eps \} \times \{ |s'| <\eps \} \times \{ c_1<|t'|
1097: <c_2 \}
1098: $$
1099: that respects the fibration. In these coordinates, the foliation
1100: is given by $x't'=cst.$, $s't'^2=cst.$, and the first integral of
1101: $\F_{\Omega_j}$ is $x'^{p-1}s'^{q-1}t'^{q-3} \cdot U_j
1102: (x',s',t')$, $U_j(0)=1$. This first integral may be extended to
1103: $$
1104: \{ |x'| <\eps \} \times \{ |s'| <\eps \} \times \{ |t'|
1105: <c \} ,
1106: $$
1107: where $c_1<c$, eventually making $\eps$ small enough. We look
1108: first for a diffeomorphism $\Psi_j$ that transforms this first
1109: integral into $x'^{p-1}s'^{q-1}t'^{q-3}$, respecting the
1110: fibration. This diffeomorphism is
1111: $$
1112: \Psi_{j}(x',s',t')= (x'\cdot V_{1j}, s'\cdot V_{2j}, t'\cdot
1113: V_{3j}),
1114: $$
1115: and the conditions mean that
1116: \begin{eqnarray*}
1117: V_{1}^{p-1}\cdot V_2^{q-1}\cdot V_3^{q-3} & = & U_j \\
1118: V_{1}\cdot V_3 & = & 1 \\
1119: V_2\cdot V_3^2 & = & 1.
1120: \end{eqnarray*}
1121: So, $V_{1j}=U_j^{-(p+q)}$; $V_{2j}= U_j^{-2(p+q)}$;
1122: $V_{3j}=U_j^{p+q}$.
1123:
1124: Consider now the diffeomorphism $\tilde{\Psi}:= \Psi_{1}\circ
1125: \Psi\circ \Psi_2^{-1}$. It respects both the fibration and the
1126: first integral $x'^{p-1}s'^{q-1}t'^{q-3}$. Write $\tilde{\Psi}=
1127: (\theta_1, \theta_2, \theta_3 )$. The conditions above mean that
1128: \begin{eqnarray*}
1129: \theta_1\cdot \theta_3 & = & x't' \\
1130: \theta_2 \cdot \theta_3^2 & = & s't'^2 \\
1131: \theta_1^{p-1} \cdot \theta_2^{q-1} \cdot \theta_3^{q-3} & = &
1132: x'^{p-1}s'^{q-1}t'^{q-3} \cdot g(x'^{p-1}s'^{q-1}t'^{q-3}),
1133: \end{eqnarray*}
1134: with $g(0)\neq 0$. As before, we have that $\theta_1 =x'\cdot
1135: g^{-(p+q)}$; $\theta_2 = s' \cdot g^{-2(p+q)}$; $\theta_3 =
1136: t'\cdot g^{p+q}$.
1137:
1138: This is a map defined, in the considered chart, over a set of the
1139: type $\{ |x'^{p-1}s'^{q-1}t'^{q-3} | <\eps \}$, and this set
1140: intersects the domain of definition of $\Psi$. So, $\Psi=
1141: \Psi_1^{-1}\circ \tilde{\Psi}\circ \Psi_2$ may be extended to a
1142: neighbourhood of $L_{\frac{p+q}{2}-1}= D_{\frac{p-1}{2}}\cap
1143: D_{\frac{p+q}{2}-1}$.
1144:
1145: Repeating the argument, we extend the conjugation to a neighbourhood
1146: of $D_{\frac{p-1}{2}}\cap (D_{\frac{p+1}{2}}\cup \cdots \cup
1147: D_{\frac{p+q}{2}})$. Now, similar arguments as in the preceding
1148: si\-tua\-tions, and as in the two-dimensional case, allow us to
1149: extend $\Psi$ to a neighbourhood of the exceptional divisor, and
1150: again Hartogs' theorem completes the result.
1151:
1152: \end{prueba}
1153:
1154:
1155: %-------------------------------------------------------------------------------------------
1156:
1157: % Referencias
1158:
1159: %-------------------------------------------------------------------------------------------
1160:
1161:
1162:
1163: \begin{thebibliography}{99}
1164:
1165: \bibitem{BMS}
1166: {\sc M. Berthier, R. Meziani and P. Sad}, {\it On the
1167: classification of nilpotent singularities}, Bull. Sci. Math. 123
1168: (1999) p. 351-370.
1169:
1170:
1171: \bibitem{Ca}
1172: {\sc F. Cano}, {\it Reduction of the singularities of foliations
1173: and applications.} Collection: Singularities
1174: Symposium-Lojasiewicz 70 (Krakow, 1996; Warsaw, 1996), 51-71
1175: Series: Banach Center Publ., 44, Polish Acad. Sci., Warsaw, 1998.
1176:
1177: \bibitem{CC}
1178:
1179: {\sc F. Cano, D. Cerveau}, {\it Desingularization of
1180: non-dicritical holomorphic foliations and existence of
1181: separatrices.}, Acta Math., 169 (1992), p. 1-103.
1182:
1183: \bibitem{CL}
1184:
1185: {\sc D. Cerveau, A. Lins-Neto}, {\it Formes tangentes \`{a} des actions commutatives}, Ann. Fac. Sci. Toulouse VI (1984), 51-85.
1186:
1187:
1188: \bibitem{CM}
1189: {\sc D. Cerveau et Moussu}, {\it Groupes d'automorphismes de
1190: $(\C,0)$ et \'equations diff\'erentielles $ydy+\cdots=0$}, Bull.
1191: Soc. Math. France 116 (1988) 459-488.
1192:
1193: \bibitem{CeMo}
1194:
1195: {\sc D. Cerveau, J. Mozo Fern\'{a}ndez}, {\it Classification
1196: analytique des feuilletages singuliers r\'{e}duits de codimension
1197: 1 en dimension $n\geq 3$}, Ergod. Th. Dynam. Sys. (2002), 22,
1198: 1041-1060.
1199:
1200: \bibitem{Co}
1201: {\sc V. Cossart}, {\it Desingularization in dimension two}, in
1202: {\it Resolution of surface singularities}, Springer LNM 1101
1203: (1984), 79-98.
1204:
1205: \bibitem{G}
1206:
1207: {\sc J. Giraud}, {\it Desingularization in low dimension}, in {\it
1208: Resolution of surface singularities}, Springer LNM 1101 (1984),
1209: 51-78.
1210:
1211: \bibitem{H}
1212:
1213: {\sc H. Hironaka}, {\it Desingularization of excellent schemes},
1214: in {\it Resolution of surface singularities}, Springer LNM 1101
1215: (1984), 99-132.
1216:
1217: \bibitem{La}
1218: {\sc F. Lazzeri}, {\it Analytic singularities}, in {\it
1219: Singularities of analytic spaces}, CIME, Bressanone, 1974,
1220: Edizioni Cremonese, Roma (1975).
1221:
1222: \bibitem{Li}
1223:
1224: {\sc J. Lipman}, {\it Quasi-ordinary singularities of surfaces in
1225: $\C^3$}, Proceedings of Symposia in Pure Mathematics, 40 (1983),
1226: Part 2, 161-172.
1227:
1228: \bibitem{L}
1229: {\sc F. Loray}, {\it A preparation theorem for codimension one
1230: foliations}, Preprint ArXiv:math.DG/0402411 v1, Feb 25 2004.
1231:
1232: \bibitem{Ma}
1233:
1234: {\sc B. Malgrange}, {\it Frobenius avec singularit\'{e}s.
1235: Codimension un}.
1236: Inst.-Hautes-Etudes-Sci.-Publ.-Math. (1976), No. 46, 163--173.
1237:
1238:
1239: \bibitem{MM}
1240:
1241: {\sc J.F. Mattei et R. Moussu}, {\it Holonomie et int\'egrales
1242: premi\'eres}, Ann. Sci. \'Ecole Normale Sup. 13 (1980) 469-523.
1243: \bibitem{Me}
1244:
1245: {\sc R. Meziani}, {\it Classification analytique d'\'equations
1246: diff\'{U}rentielles $ydy+\cdots=0$ et espaces de modules}, Bol. da
1247: Soc. Bras. Mat. 27 (1996) 23-53.
1248:
1249: \bibitem{Me-tesis}
1250:
1251: {\sc R. Meziani}, {\it Probl\`{e}me de modules pour des
1252: \'{e}quations di{f}{f}\'{e}rentielles d\'{e}g\'{e}n\'{e}r\'{e}es
1253: de $ (\C^2, 0) $}, PhD. thesis, Univ. Rennes I (1992)
1254:
1255: \bibitem{Mo}
1256:
1257: {\sc R. Moussu}, {\it Sur l'existence d'int\'{e}grales
1258: pr\'{e}mi\`{e}res pour un germe de forme de Pfaff}, Ann. Inst.
1259: Fourier 26 (1976), no. 2, 171-220.
1260:
1261: \bibitem{Mo2}
1262:
1263: {\sc R. Moussu}, {\it Holonomie evanescente des \'{e}quations
1264: di{f}{f}\'{e}rentielles d\'{e}g\'{e}n\'{e}r\'{e}es transverses},
1265: in {\it Singularities and Dynamical Systems}, North-Holland,
1266: (1985),
1267: 151-173.
1268:
1269: \bibitem{R}
1270:
1271: {\sc G. Reeb}, {\it Sur certaines propri\'et\'es topologiques des
1272: vari\'et\'es feuillet\'ees}, Thesis, Hermann (1952).
1273:
1274: \bibitem {T}
1275:
1276: {\sc F. Takens}, {\it Singularities of vectors fields}, Inst.
1277: Hautes \'Etudes Sci. Publ. Math. 43 (1974) 47-100.
1278:
1279:
1280:
1281:
1282: \end{thebibliography}
1283:
1284:
1285: \end{document}
1286: