On the holomorphy of the curvature of planar webs along an invariant curve


Abstract

Let \(\mathcal{W}=\mathcal{W}_{n}\boxtimes\mathcal{W}_{d-n}\) be a \(d\)-web on \((\mathbb{C}^2,0)\), where \(\mathcal{W}_n\) is an \(n\)-web with a totally invariant irreducible curve \(C\), and \(\mathcal{W}_{d-n}\) is a regular \((d-n)\)-web transverse to \(C\). We show that the curvature of \(\mathcal{W}\) is holomorphic along \(C\) if and only if the curvature of \(\mathcal{W}_n\) is holomorphic along \(C\). When \(\mathcal{W}_n\) is non-degenerate along \(C\), we prove that \(K(\mathcal{W}_n)\), and hence \(K(\mathcal{W})\), is holomorphic along \(C.\) We deduce that, if \(\mathcal{W}_n\) is irreducible and \(\mathrm{mult}\left(\Delta(\mathcal{W}_n),C\right)<3(n-1),\) then \(K(\mathcal{W})\) is holomorphic along \(C.\) This generalizes a result of Marı́n and Pereira, obtained in the case where \(C\) has minimal multiplicity \(n-1\) in the discriminant \(\Delta(\mathcal{W}_n).\) If \(n\) is prime or \(n=4\), the condition \(\mathrm{mult}\left(\Delta(\mathcal{W}_n),C\right)<3(n-1)\) can be weakened to \(\mathrm{mult}\left(\Delta(\mathcal{W}_n),C\right)<n(n-1).\) Moreover, we describe a natural decomposition of \(\mathcal{W}_n\) as the product of two subwebs \(\mathcal{W}_n=\mathcal{W}_{n}^{\rm{str}}\boxtimes\mathcal{W}_{n}^{\rm{wk}}.\) Under the assumption that \(\mathcal{W}_{n}^{\rm{wk}}\) is non-degenerate along \(C\), we show that the holomorphy of \(K(\mathcal{W})\) on \(C\) is equivalent to that of \(K(\mathcal{W}_{n}^{\rm{str}}).\)

1 Introduction↩︎

The holomorphy of the curvature of planar webs has been studied in [1][3]. In this paper, we consider webs on \((\mathbb{C}^2,0)\) with an irreducible invariant curve and study the holomorphy of their curvature along such a curve. For definitions and notations concerning webs on \((\mathbb{C}^2,0)\), we refer to [4].

1.1 Webs↩︎

A (singular) \(d\)-web \(\mathcal{W}\) on \((\mathbb{C}^{2},0)\) is defined by a \(d\)-symmetric \(1\)-form \(\omega \in \mathrm{Sym}^{d}\Omega^{1}(\mathbb{C}^{2},0)\) satisfying the following conditions:

  1. the singular locus \(\mathrm{Sing}\omega=\{p\in(\mathbb{C}^{2},0):\omega(p)=0\}\) consists of isolated points;

  2. for every generic point \(p\in(\mathbb{C}^{2},0),\) \(\omega(p)\) factors as the product of \(d\) pairwise linearly independent \(1\)-forms.

Two such \(d\)-symmetric \(1\)-forms \(\omega\) and \(\omega'\) define the same web if there exists a unit \(u\in\mathcal{O}^*(\mathbb{C}^{2},0)\) such that \(\omega'=u\omega.\)

The discriminant \(\Delta(\mathcal{W})\) of \(\mathcal{W}\) is the divisor defined by \(\Delta(\omega)=0\), where \(\Delta(\omega)\) is the discriminant of the \(d\)-symmetric \(1\)-form \(\omega\in\mathrm{Sym}^{d}\Omega^{1}(\mathbb{C}^2,0)\), cf. [4]. The support of \(\Delta(\mathcal{W})\) consists of the points that do not satisfy condition (2). When \(d=1\) this condition is always satisfied and we recover the usual definition of a holomorphic foliation \(\mathcal{F}\) on \((\mathbb{C}^2,0).\)

We say that \(\mathcal{W}\) is decomposable if there are webs \(\mathcal{W}_1\) and \(\mathcal{W}_2\) on \((\mathbb{C}^2,0)\) sharing no common subwebs such that \(\mathcal{W}\) is the superposition of \(\mathcal{W}_1\) and \(\mathcal{W}_2\); we then write \(\mathcal{W}=\mathcal{W}_1\boxtimes\mathcal{W}_2.\) Otherwise \(\mathcal{W}\) is said to be irreducible. It is said to be completely decomposable if there exist holomorphic foliations \(\mathcal{F}_1,\ldots,\mathcal{F}_d\) on \((\mathbb{C}^2,0)\) such that \(\mathcal{W}=\mathcal{F}_1\boxtimes\cdots\boxtimes\mathcal{F}_d\).

The \(d\)-web \(\mathcal{W}\) is regular if its discriminant \(\Delta(\mathcal{W})\) is empty, or equivalently if it is completely decomposable into \(d\) regular holomorphic foliations on \((\mathbb{C}^2,0)\) that are pairwise transverse at \(0\).

Let \(\Gamma\subset\Delta(\mathcal{W})\) be an irreducible component of the discriminant of \(\mathcal{W}\). We say that \(\Gamma\) is invariant (resp. totally invariant) by \(\mathcal{W}\) if, on the regular part of \(\Gamma\), we have \(\mathrm{T}\Gamma\subset\mathrm{T}\mathcal{W}|_{\Gamma}\) (resp. \(\mathrm{T}\Gamma=\mathrm{T}\mathcal{W}|_{\Gamma}\)). When \(\mathcal{W}\) is irreducible, these two notions coincide. For more details on this subject, see [4].

1.2 Strongly and weakly invariant curves by an irreducible \(\nu\)-web↩︎

Let \(\mathcal{W}_\nu\) be an irreducible \(\nu\)-web on \((\mathbb{C}^2,0)\) admitting an irreducible invariant curve \(C\). The irreducibility of \(\mathcal{W}_\nu\) implies that its monodromy group is cyclic of order \(\nu\). Thus, in a neighborhood \(U\) of a generic point of \(C\), there exists a cyclic ramified Galois covering \(\pi\colon\tilde{U}\to U\) of degree \(\nu\) such that the pull-back \(\pi^*\mathcal{W}_\nu\) decomposes completely into \(\nu\) holomorphic foliations on \(\tilde{U}\), and

  • either the curve \(\tilde{C}=\pi^{-1}(C)\) is totally invariant by \(\pi^*\mathcal{W}_\nu\), in which case we say that \(C\) is strongly invariant by \(\mathcal{W}_\nu\);

  • or \(\pi^*\mathcal{W}_\nu\) is transverse to \(\tilde{C}\), in which case we say that \(C\) is weakly invariant by \(\mathcal{W}_\nu\).

More precisely, let us choose a local coordinate system \((z,w)\) on \(U\) such that \(C\cap U=\{w=0\}.\) In this system, \(\mathcal{W}_\nu\) is defined by a \(\nu\)-symmetric \(1\)-form \(\omega\) of type \[\begin{align} \label{equa:nu-forme-omega} &\omega=\mathrm{d}w^{\nu}+wa_{\nu-1}(z,w)\mathrm{d}w^{\nu-1}\mathrm{d}z+\cdots +wa_1(z,w)\mathrm{d}w\mathrm{d}z^{\nu-1}+w^{\kappa}a_0(z,w)\mathrm{d}z^{\nu}, \end{align}\tag{1}\] with \(\kappa\in\mathbb{N}^*\) and \(a_0(z,0)\neq0.\) The ramified covering \(\pi:\tilde{U}\to U\) is given by \((z,w)=\pi(x,y)=(x,y^\nu),\) and \(\tilde{C}=\pi^{-1}(C)=\{y=0\}.\) The above description of the \(\nu\)-web \(\pi^*\mathcal{W}_\nu\) is ensured by the following dichotomy, which will be established in §2 (Lemma 13):

  • if \(\kappa<\nu,\) then \(\pi^*\mathcal{W}_\nu=\boxtimes_{j=1}^{\nu}\mathcal{F}_j\), where each \(\mathcal{F}_j\) is a foliation transverse to \(\tilde{C}\);

  • if \(\kappa\geq\nu,\) then \(\pi^*\mathcal{W}_\nu=\boxtimes_{j=1}^{\nu}\mathcal{F}_j\), where each \(\mathcal{F}_j\) is a foliation having \(\tilde{C}\) as an invariant curve.

The integer \(\kappa\) also plays a role in determining the multiplicity of the discriminant \(\Delta(\mathcal{W}_{\nu})\) along \(C\): for \(\nu\geq2\), we have \[\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)\geq\kappa(\nu-1),\] with equality if and only if \(\gcd(\nu,\kappa)=1\). In particular, \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)\) is minimal, equal to \(\nu-1\), if and only if \(\kappa=1\) (see Lemma 13).

Remark 1.

  • (i) Saying that \(C\) is strongly (resp. weakly) invariant by \(\mathcal{W}_\nu\) is equivalent to \(\kappa\geq\nu\) (resp. \(\kappa<\nu\)).

  • In the case of a foliation \(\mathcal{F}\), i.e. when \(\nu=1\), every \(\mathcal{F}\)-invariant curve is strongly invariant by \(\mathcal{F}\).

  • For \(\nu\geq2\), if \(C\) has minimal multiplicity \(\nu-1\) in \(\Delta(\mathcal{W}_\nu)\), which is equivalent to \(\kappa=1\), then \(C\) is always weakly invariant by \(\mathcal{W}_\nu.\)

The web \(\mathcal{W}_\nu\) being irreducible, the polynomial defining \(\mathcal{W}_\nu\) in \(\mathbb{C}((z))\{w,\frac{\mathrm{d}w}{\mathrm{d}z}\}\) has the following Puiseux parametrizations: \[\begin{align} \label{equa:puiseux} \frac{\mathrm{d}w}{\mathrm{d}z}=\sum_{l=r}^{\infty}c_{l-r}(z)\left(\zeta^{j}w^{\frac{1}{\nu}}\right)^{l}=\zeta^{jr}c_{0}(z)w^{\frac{r}{\nu}}+\cdots, \qquad c_{0}(z)\not\equiv 0,\quad\zeta=\exp(\tfrac{2\mathrm{i}\pi}{\nu}),\quad j=1,\ldots,\nu. \end{align}\tag{2}\]

As we will see (cf. proof of Lemma 13), the integer \(\kappa\) defined via the \(\nu\)-symmetric \(1\)-form \(\omega\) coincides with \(r\), that is, with the numerator of the first Puiseux exponent in the parametrizations (2 ), yielding \[\begin{align} \label{equa:puiseux-kappa} \frac{\mathrm{d}w}{\mathrm{d}z}=\zeta^{j\kappa}c_0(z)w^{\frac{\kappa}{\nu}}+\cdots,\qquad c_{0}(z)\not\equiv 0,\quad j=1,\ldots,\nu. \end{align}\tag{3}\]

Let us consider another system of local coordinates \((z',w')\) such that \(C={w'=0}\). We can then write \[z=\varphi(z',w'), \qquad w=u(z',w')w',\] for some holomorphic functions \(\varphi\) and \(u\) satisfying \[0\neq\left.\det\left(\frac{\partial(z,w)}{\partial(z',w')}\right)\right|_{w'=0}=\partial_{z'}\varphi(z',0)u(z',0).\]

In this new coordinate system, the Puiseux parametrizations (3 ) take the form \[\begin{align} \label{equa:puiseux-kappa-nouveau-systeme} \frac{\mathrm{d}w'}{\mathrm{d}z'}= \frac{\zeta^{j\kappa}c_0(\varphi(z',0))\partial_{z'}\varphi(z',0)}{u(z',0)^{1-\frac{\kappa}{\nu}}}w'^{\frac{\kappa}{\nu}} -\frac{\partial_{z'}u(z',0)}{u(z',0)}w' +\cdots,\qquad\quad j=1,\ldots,\nu. \end{align}\tag{4}\]

When \(C\) is weakly invariant by \(\mathcal{W}_\nu\), i.e. if \(\kappa<\nu\), we see from (4 ) that the first nonzero exponent in the Puiseux series of \(\frac{\mathrm{d}w'}{\mathrm{d}z'}\) at \(w'=0\) remains equal to \(\frac{\kappa}{\nu}\). It follows that the integer \(\kappa\in\{1,\ldots,\nu-1\}\) does not depend on the choice of local coordinate system and is therefore intrinsically attached to the pair \((\mathcal{W}_\nu,C)\). We call this integer \(\kappa\) the Puiseux index of \(\mathcal{W}_\nu\) relative to \(C\) and we denote it by \(\mathfrak{i}(\mathcal{W}_\nu,C).\) The ratio \(\frac{\kappa}{\nu}\) is called the Puiseux exponent of \(\mathcal{W}_\nu\) relative to \(C\) and is denoted by \(\rho(\mathcal{W}_\nu,C).\)

Remark 2. For a \(\nu\)-web \(\mathcal{W}_\nu\) on \((\mathbb{C}^2,0)\), not necessarily irreducible, admitting a totally invariant irreducible curve \(C\), we have the decomposition \[\mathcal{W}_{\nu}=\mathcal{W}_{\nu}^{\rm{str}}\boxtimes\mathcal{W}_{\nu}^{\rm{wk}},\] where \(\mathcal{W}_\nu^{\mathrm{str}}\) (resp. \(\mathcal{W}_\nu^{\mathrm{wk}}\)) denotes the subweb of \(\mathcal{W}_\nu\) consisting of the irreducible subwebs of \(\mathcal{W}_\nu\) having \(C\) as a strongly (resp. weakly) invariant curve.

1.3 Curvature of webs↩︎

Let us briefly recall the definition of the curvature of a \(d\)-web \(\mathcal{W}\) on \((\mathbb{C}^2,0)\) with \(d\geq3.\) First, assume that \(d=3\) and that \(\mathcal{W}\) is completely decomposable, \(\mathcal{W}=\mathcal{F}_1\boxtimes\mathcal{F}_2\boxtimes\mathcal{F}_3.\) For \(i=1,2,3\), let \(\omega_{i}\) be a \(1\)-form with an isolated singularity at \(0\) defining the foliation \(\mathcal{F}_{i}.\) Without loss of generality, we can assume that the \(1\)-forms \(\omega_i\) satisfy \(\omega_1+\omega_2+\omega_3=0.\) It can be shown that there exists a meromorphic \(1\)-form \(\eta(\mathcal{W})\), called the fundamental form of \(\mathcal{W}\), such that \(\mathrm{d}\omega_i=\eta(\mathcal{W})\wedge\omega_i\) for \(i=1,2,3.\) This \(1\)-form is well-defined up to addition of a closed logarithmic \(1\)-form \(\dfrac{\mathrm{d}g}{g}\) with \(g\in\mathcal{O}^*(\mathbb{C}^{2},0).\) The curvature of \(\mathcal{W}\) is then defined as the \(2\)-form \(K(\mathcal{W})=\mathrm{d}\,\eta(\mathcal{W}).\)

Now, if the \(d\)-web \(\mathcal{W}\) is completely decomposable with \(d>3\), \(\mathcal{W}=\mathcal{F}_1\boxtimes\cdots\boxtimes\mathcal{F}_d\), we define the curvature \(K(\mathcal{W})\) of \(\mathcal{W}\) as the sum of the curvatures of all its \(3\)-subwebs.

It is easy to check that \(K(\mathcal{W})\) is a meromorphic \(2\)-form with poles along the discriminant \(\Delta(\mathcal{W})\) of \(\mathcal{W}\), canonically associated to \(\mathcal{W}.\)

Finally, if \(\mathcal{W}\) is not completely decomposable, then its pull-back by a suitable ramified Galois covering is a completely decomposable web. The invariance of the curvature of this new web by the action of the Galois group allows us to descend it to a meromorphic \(2\)-form \(K(\mathcal{W})\) on \((\mathbb{C}^2,0)\), with poles along the discriminant of \(\mathcal{W}\) (see [2]).

1.4 Non-degenerate webs↩︎

We introduce the following definition.

Definition 3. Let \(\mathcal{W}_\nu\) be a \(\nu\)-web on \((\mathbb{C}^2,0)\) admitting a totally invariant irreducible curve \(C.\) Let \(\mathcal{W}_\nu=\boxtimes_{\alpha=1}^{r}\mathcal{W}_{\nu_{\alpha}}\) be the decomposition of \(\mathcal{W}_\nu\) into its irreducible components. We say that \(\mathcal{W}_\nu\) is non-degenerate along \(C\) if \(C\) is weakly invariant by each \(\mathcal{W}_{\nu_\alpha}\) and if, denoting \(\kappa_\alpha=\mathfrak{i}(\mathcal{W}_{\nu_{\alpha}},C)\) and \(\rho_\alpha=\rho(\mathcal{W}_{\nu_{\alpha}},C)=\frac{\kappa_\alpha}{\nu_\alpha},\) the following conditions are satisfied:

  • for all \(\alpha=1,\ldots,r\), \(\gcd(\nu_\alpha,\kappa_\alpha)\leq2\);

  • for all \(\alpha\neq \beta\), if \(\rho_\alpha=\rho_\beta\), then \(\gcd(\nu_\alpha,\kappa_\alpha)=\gcd(\nu_\beta,\kappa_\beta)=1\);

  • there are no three distinct indices \(\alpha,\beta,\gamma\in\{1,\ldots,r\}\) such that \(\rho_\alpha=\rho_\beta=\rho_\gamma.\)

Remark 4. Let \(\mathcal{W}_\nu\) be an irreducible \(\nu\)-web on \((\mathbb{C}^2,0)\) having an irreducible invariant curve \(C\). Saying that \(\mathcal{W}_\nu\) is non-degenerate along \(C\) amounts to saying that \(C\) is weakly invariant by \(\mathcal{W}_\nu\) and that \(\gcd(\nu,\kappa)\leq2\), where \(\kappa=\mathfrak{i}(\mathcal{W}_{\nu},C).\)

In particular:

  • The web \(\mathcal{W}_\nu\) is always non-degenerate along \(C\) if \(C\) is weakly invariant by \(\mathcal{W}_\nu\) with Puiseux index \(\kappa\in\{1,2,\nu-1,\nu-2\}.\)

  • If \(C\) has minimal multiplicity \(\nu-1\) in \(\Delta(\mathcal{W}_\nu)\), then \(\mathcal{W}_\nu\) is non-degenerate along \(C\) (cf. Remark 1 (iii)).

  • When \(\nu\) is prime or equal to \(4\), \(\mathcal{W}_\nu\) is non-degenerate along every curve which is weakly invariant by \(\mathcal{W}_\nu.\)

Example 5. Let \(\mathcal{W}_{\nu_\alpha},\alpha=1,\ldots,r,\) be irreducible \(\nu_\alpha\)-webs on \((\mathbb{C}^2,0)\), with \(\nu_\alpha\geq2\), sharing a common irreducible invariant curve \(C\). We assume that:

  • \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu_\alpha}),C\right)=\nu_\alpha-1\), for all \(\alpha=1,\ldots,r\);

  • there are no three distinct indices \(\alpha,\beta,\gamma\in\{1,\ldots,r\}\) such that \(\nu_\alpha=\nu_\beta=\nu_\gamma.\)

Then the web \(\mathcal{W}_\nu=\mathcal{W}_{\nu_1}\boxtimes\cdots\boxtimes\mathcal{W}_{\nu_r}\) is non-degenerate along \(C.\)

The definition of a non-degenerate web along a totally invariant curve is a generalization of the situation considered in [2], namely that of irreducible \(\nu\)-webs \(\mathcal{W}_\nu\) admitting an irreducible invariant curve of minimal multiplicity \(\nu-1\) in \(\Delta(\mathcal{W}_\nu).\) It will play an important role in the subsequent study of the holomorphy of the curvature of certain webs on \((\mathbb{C}^2,0).\)

1.5 Statements of the main results↩︎

The central result of this paper is the following theorem.

Theorem 6. Let \(\mathcal{W}_n\) be an \(n\)-web on \((\mathbb{C}^2,0)\) admitting a totally invariant irreducible curve \(C\). Let \(\mathcal{W}_{d-n}\) be a regular \((d-n)\)-web on \((\mathbb{C}^2,0)\) transverse to \(C\). Set \(\mathcal{W}=\mathcal{W}_{n}\boxtimes\mathcal{W}_{d-n}.\) Then \(K(\mathcal{W})-K(\mathcal{W}_n)\) is holomorphic along \(C\). In particular, \(K(\mathcal{W})\) is holomorphic along \(C\) if and only if \(K(\mathcal{W}_n)\) is holomorphic along \(C.\)

For \(n=2\), we recover a special case of [2]: the curvature \(K(\mathcal{W})\) is always holomorphic on \(C\).

In [2], the authors consider a \(d\)-web \(\mathcal{W}\) of the form \(\mathcal{W}=\mathcal{W}_\nu\boxtimes\mathcal{W}_{d-\nu}\), where \(\mathcal{W}_\nu\) is an irreducible \(\nu\)-web having an irreducible invariant curve \(C\) of minimal multiplicity \(\nu-1\) in \(\Delta(\mathcal{W}_\nu)\), and \(\mathcal{W}_{d-\nu}\) is a regular \((d-\nu)\)-web transverse to \(C\). They prove that the curvature of \(\mathcal{W}\) is holomorphic along \(C\).

Theorem 6 allows us to recover this result: by arguing only on the component \(\mathcal{W}_\nu\), while following the arguments of [2], we obtain the holomorphy on \(C\) of \(K(\mathcal{W}_\nu)\), and hence that of \(K(\mathcal{W})\). The web \(\mathcal{W}_\nu\) being a particular case of a non-degenerate web along \(C\), the following proposition extends the holomorphy result of \(K(\mathcal{W}_\nu)\) to the general framework of non-degenerate webs along \(C\).

Proposition 7. Let \(\mathcal{W}_\nu\) be a \(\nu\)-web on \((\mathbb{C}^2,0)\) having a totally invariant irreducible curve \(C\). Assume that \(\mathcal{W}_\nu\) is non-degenerate along \(C\). Then the curvature of \(\mathcal{W}_\nu\) is holomorphic along \(C.\)

Remark 8. The holomorphy of the curvature of \(\mathcal{W}_\nu\) along \(C\) is not ensured by merely requiring that \(C\) be weakly invariant by the irreducible subwebs constituting \(\mathcal{W}_\nu\). Indeed, the non-degeneracy of \(\mathcal{W}_\nu\) along \(C\), and in particular each of the conditions (\(\mathfrak{a}\)), (\(\mathfrak{b}\)) and (\(\mathfrak{c}\)), is essential for the validity of Proposition 7, as shown by Examples 17, 18 and 19 in Section §4.

From Theorem 6 and Proposition 7, we obtain a generalization of Proposition 2.6 of [2] by replacing the minimal multiplicity assumption \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)=\nu-1\) with the condition \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)<3(\nu-1).\)

Corollary 9. Let \(\mathcal{W}_\nu\) be an irreducible \(\nu\)-web on \((\mathbb{C}^2,0)\), with \(\nu\geq 2\). Assume that \(\Delta(\mathcal{W}_\nu)\) admits an irreducible component \(C\) invariant by \(\mathcal{W}_\nu\) and such that \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)<3(\nu-1).\) Let \(\mathcal{W}_{d-\nu}\) be a regular \((d-\nu)\)-web on \((\mathbb{C}^2,0)\) transverse to \(C\). Then the curvature of the \(d\)-web \(\mathcal{W}=\mathcal{W}_{\nu}\boxtimes\mathcal{W}_{d-\nu}\) is holomorphic along \(C\).

When \(\nu\) is prime or \(\nu=4\), we can weaken the condition on the multiplicity of \(C\) in \(\Delta(\mathcal{W}_{\nu})\) by replacing the bound \(3(\nu-1)\) with the bound \(\nu(\nu-1).\)

Corollary 10. Let \(\nu\) be a prime number or \(\nu=4\). Let \(\mathcal{W}_\nu\) be an irreducible \(\nu\)-web on \((\mathbb{C}^{2},0)\). Assume that \(\Delta(\mathcal{W}_\nu)\) has an irreducible component \(C\) invariant by \(\mathcal{W}_\nu\) and such that \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)<\nu(\nu-1).\) Let \(\mathcal{W}_{d-\nu}\) be a regular \((d-\nu)\)-web on \((\mathbb{C}^2,0)\) transverse to \(C\). Then the curvature of the \(d\)-web \(\mathcal{W}=\mathcal{W}_{\nu}\boxtimes\mathcal{W}_{d-\nu}\) is holomorphic along \(C\).

In the setting of Theorem 6, we can ask if it is possible to reduce the study of the holomorphy of \(K(\mathcal{W})\) along \(C\) to that of \(K(\mathcal{W}_0)\) for some subweb \(\mathcal{W}_0\subsetneq\mathcal{W}_n\). The following theorem gives a partial answer to this question.

Theorem 11. Let \(\mathcal{W}_n\) be an \(n\)-web on \((\mathbb{C}^2,0)\) having a totally invariant irreducible curve \(C\). Assume that \(\mathcal{W}_{n}^{\rm{wk}}\) is non-degenerate along \(C\). Let \(\mathcal{W}_{d-n}\) be a regular \((d-n)\)-web on \((\mathbb{C}^2,0)\) transverse to \(C\), and set \(\mathcal{W}=\mathcal{W}_n\boxtimes\mathcal{W}_{d-n}.\) Then \(K(\mathcal{W})-K(\mathcal{W}_{n}^{\rm{str}})\) is holomorphic along \(C\). In particular, \(K(\mathcal{W})\) is holomorphic along \(C\) if and only if \(K(\mathcal{W}_{n}^{\rm{str}})\) is holomorphic along \(C\).

Remark 12. Theorem 11 applies in particular when \(\mathcal{W}_n\) is of the form \(\mathcal{W}_n=\mathcal{W}_{n-\nu}\boxtimes\mathcal{W}_{\nu}\), where \(\mathcal{W}_\nu=\mathcal{W}_{\nu_1}\boxtimes\cdots\boxtimes\mathcal{W}_{\nu_r}\) is the \(\nu\)-web from Example 5, and \(\mathcal{W}_{n-\nu}=\mathcal{F}_1\boxtimes\cdots\boxtimes\mathcal{F}_{n-\nu}\) is a completely decomposable \((n-\nu)\)-web leaving \(C\) totally invariant. In this case, we have \(\mathcal{W}_{n}^{\rm{str}}=\mathcal{W}_{n-\nu}\),  \(\mathcal{W}_{n}^{\rm{wk}}=\mathcal{W}_{\nu}\), and the curvature of the \(d\)-web \(\mathcal{W}=\mathcal{W}_{n-\nu}\boxtimes\mathcal{W}_\nu\boxtimes\mathcal{W}_{d-n-\nu}\) is holomorphic on \(C\) if and only if the curvature of the \((n-\nu)\)-web \(\mathcal{W}_{n-\nu}\) is holomorphic on \(C\). For \(r=1\), we recover Proposition 6.2 of [5].

2 Preliminary results↩︎

In the following lemma, \(\mathcal{W}_\nu\) denotes the irreducible \(\nu\)-web of §1.2, defined in local coordinates \((z,w)\) by the \(\nu\)-symmetric \(1\)-form (1 ) and admitting \(C=\{w=0\}\) as an invariant curve. This lemma establishes the facts mentioned in §1.2 concerning the multiplicity of \(C\) in \(\Delta(\mathcal{W}_\nu)\) as well as the local description of the web \(\pi^*\mathcal{W}_\nu\) near \(\tilde{C}=\pi^{-1}(C),\) where \(\pi\colon(x,y)\mapsto(x,y^\nu).\)

Lemma 13.

****1.*** For \(\nu\geq2\), we have \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)\geq\kappa(\nu-1),\) and equality holds if and only if \(\gcd(\nu,\kappa)=1.\) In particular, \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)=\nu-1\) if and only if \(\kappa=1.\)*

****2.*** We have the following alternative:*

  • If \(\kappa<\nu,\) then \(\pi^*\mathcal{W}_\nu=\boxtimes_{j=1}^{\nu}\mathcal{F}_j\), where \(\mathcal{F}_j\) is a foliation transverse to \(\tilde{C}\) and defined by a \(1\)-form \(\omega_j\) of type \[\omega_j=\mathrm{d}x+\zeta^{-j\kappa}y^{\nu-\kappa-1}f(x,\zeta^jy)\mathrm{d}y, \qquad\qquad f(x,0)\not\equiv0,\zeta=\exp(\tfrac{2\mathrm{i}\pi}{\nu}).\]

  • If \(\kappa\geq\nu,\) then \(\pi^*\mathcal{W}_\nu=\boxtimes_{j=1}^{\nu}\mathcal{F}_j\), where \(\mathcal{F}_j\) is a foliation admitting \(\tilde{C}\) as an invariant curve and given by a \(1\)-form \(\theta_j\) of type \[\theta_j=\mathrm{d}y+\zeta^{j\kappa}y^{\kappa-\nu+1}h(x,\zeta^jy)\mathrm{d}x, \qquad\qquad h(x,0)\not\equiv0.\]

**Proof. From (1 ) and (2 ), we obtain \[\begin{align} \prod\limits_{j=1}^{\nu}(\zeta^{jr}c_{0}(z)w^{\frac{r}{\nu}}+\ldots)=(-1)^{r(\nu+1)}w^{r}c_{0}(z)^{\nu}+\ldots=w^{\kappa}a_{0}(z,w). \end{align}\] Since \(a_0(z,0)\neq0,\) we deduce that \(r=\kappa\). Thus we have \[\begin{align} \Delta(\mathcal{W}_{\nu}) &=\prod\limits_{i\neq j}\Big((\zeta^{i\kappa}c_{0}(z)w^{\frac{\kappa}{\nu}}+\cdots)-(\zeta^{j\kappa}c_{0}(z)w^{\frac{\kappa}{\nu}}+\cdots)\Big)\\ &=w^{\kappa(\nu-1)}c_{0}(z)^{\nu(\nu-1)}\prod\limits_{i\neq j}(\zeta^{i\kappa}-\zeta^{j\kappa})+\cdots. \end{align}\] It follows that \(\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)\geq\kappa(\nu-1),\) with equality if and only if \(\zeta^{i\kappa}\neq\zeta^{j\kappa}\) for all \(i\neq j\), that is, if and only if \(\gcd(\nu,\kappa)=1\), since \(\gcd(\nu,\kappa)=\dfrac{\nu}{\#\{\zeta^{i\kappa}\big\vert i=1,\ldots,\nu\}}.\) Hence the first assertion holds.

The second assertion is obtained by applying \(\pi\) to (3 ). ◻

The following proposition corresponds to Theorem 6 in the case where \(\mathcal{W}_n=\mathcal{W}_{n}^{\rm{wk}}.\)

Proposition 14. Let \(\mathcal{W}_\nu\) be a \(\nu\)-web on \((\mathbb{C}^2,0)\), with \(\nu\geq2\). Assume that \(\Delta(\mathcal{W}_\nu)\) has an irreducible component \(C\) weakly invariant by every irreducible subweb of \(\mathcal{W}_\nu.\) Let \(\mathcal{W}_{d-\nu}\) be a regular \((d-\nu)\)-web on \((\mathbb{C}^2,0)\) transverse to \(C.\) Set \(\mathcal{W}=\mathcal{W}_{\nu}\boxtimes\mathcal{W}_{d-\nu}.\) Then \(K(\mathcal{W})-K(\mathcal{W}_\nu)\) is holomorphic along \(C.\)

The proof of this proposition relies on the following technical lemma, which will also be used to prove Proposition 7.

Lemma 15.

Let \(n_1,n_2,n_3\in\mathbb{N}\) and consider the \(3\)-web \(\mathcal{W}\) defined by the \(1\)-forms \[\begin{align} \omega_i=\mathrm{d}x+y^{n_i}h_i(x,y)\mathrm{d}y,\qquad\quad i=1,2,3. \end{align}\] For \(i\neq j\), we set \[\begin{align} h_{ij}= \left\{ \begin{array}{lcl} h_{i} & \text{if} & n_{i}<n_{j}, \\ h_{i}-h_{j} & \text{if} & n_{i}=n_{j},\\ \hphantom{h_{i}}-h_{j} & \text{if} & n_{i}>n_{j}. \end{array} \right. \end{align}\] We assume that the following conditions are satisfied:

  • \(h_{1},h_{2},h_{23}\) and \(h_{31}\) do not vanish along \(y=0\);

  • if \(n_1=n_2=n_3\) and \(h_{12}(x,0)\equiv0\) (i.e. \(h_1(x,0)=h_2(x,0)\)), then there exists \(\lambda\in\mathbb{C}\setminus\{1\}\) such that \(h_3(x,0)=\lambda\,h_1(x,0).\)

Then the fundamental form of \(\mathcal{W}\) has the following form \[\begin{align} \eta(\mathcal{W})=\left(\frac{g(x)}{y^{n_0+1}}+\cdots\right)\mathrm{d}x+ \left(\frac{c}{y}+\cdots\right)\mathrm{d}y, \end{align}\] where \(n_0=\min(n_1,n_2,n_3)\), \(g\) is a holomorphic function of \(x\), \(c\in\mathbb{C}\) and the dots denote higher order terms in \(y.\)

**Proof. By arguing as in [2], we obtain that \(\eta(\mathcal{W})=A(x,y)\mathrm{d}x+B(x,y)\mathrm{d}y,\) where \[\begin{align} A=\frac{ \left| \begin{matrix} \partial_x(\delta_{12} h_3)y^{n_3}-\partial_y \delta_{12} & -\delta_{12}\\ \partial_x(\delta_{23} h_1)y^{n_1}-\partial_y \delta_{23} & -\delta_{23} \end{matrix} \right|}{\delta}, && B=\frac{ \left| \begin{matrix} \delta_{12}y^{n_3}h_3 & \partial_x(\delta_{12} h_3)y^{n_3}-\partial_y\delta_{12}\\ \delta_{23}y^{n_1}h_1 & \partial_x(\delta_{23} h_1)y^{n_1}-\partial_y\delta_{23} \end{matrix} \right| }{\delta}, \end{align}\] with \(\delta_{ij}=y^{n_j}h_j-y^{n_i}h_i\)  and  \(\delta = \delta_{12} \delta_{23} \delta_{31}.\)

When \(h_{12}(x,0)\not\equiv0\), we have \(y\not|h_{12}h_{23}h_{31}\); if \(n_1\leq n_2\leq n_3\), an argument from [2] shows that \[\begin{align} A(x,y)=\frac{n_2-n_1}{h_{31}(x,0)}\frac{1}{y^{n_1+1}}+\cdots, \qquad B(x,y)=\frac{n_1}{y}+\cdots. \end{align}\]

Assume now that \(h_{12}(x,0)\equiv0\). Then \(n_1=n_2\) and \(h_1(x,0)=h_2(x,0)\). Let us write \(h_2(x,y)=h_1(x,y)+y^ku(x,y)\) with \(k\geq1\) and \(u(x,0)\not\equiv0.\) Consider the case \(n_3\leq n_1=n_2\). A direct computation gives

\[\begin{align} &\delta(x,y)A(x,y)=(n_1+k-n_3)u(x,0)h_{31}(x,0)y^{n_1+k+n_3-1}+\cdots,&& \delta(x,y)=-u(x,0)h_{31}(x,0)^2y^{n_1+k+2n_3}+\cdots,\\ &\delta(x,y)B(x,y)=u(x,0)h_{31}(x,0)\Big(kh_3(x,0)-(n_3+k)h_{31}(x,0)\Big)y^{n_1+k+2n_3-1}+\cdots. \end{align}\]

It follows that \[\begin{align} A(x,y)=\frac{n_3-n_1-k}{h_{31}(x,0)}\frac{1}{y^{n_3+1}}+\cdots, \qquad B(x,y)=\left(n_3+k-\frac{kh_3(x,0)}{h_{31}(x,0)}\right)\frac{1}{y}+\cdots. \end{align}\]

If \(n_3<n_1=n_2\), then \(A(x,y)=\frac{n_3-n_1-k}{h_{3}(x,0)}\frac{1}{y^{n_3+1}}+\cdots\)  and  \(B(x,y)=\frac{n_3}{y}+\cdots.\)

If \(n_1=n_2=n_3\), we have by hypothesis \(h_3(x,0)=\lambda h_1(x,0)\) with \(\lambda\neq1\), hence \[\begin{align} A(x,y)=\frac{k}{(1-\lambda)h_1(x,0)}\frac{1}{y^{n_1+1}}+\cdots, \qquad B(x,y)=\left(n_1+\frac{k}{1-\lambda}\right)\frac{1}{y}+\cdots. \end{align}\]

The case \(n_3>n_1=n_2\) reduces to the case \(n_1=n_2=n_3\) with \(\lambda=0\), noting that \(\omega_3=\mathrm{d}x+y^{n_1}\tilde{h}_3(x,y)\mathrm{d}y\), where \(\tilde{h}_3(x,y)=y^{n_3-n_1}h_3(x,y).\) ◻

**Proof of Proposition 14. We will argue as in the proof of [2]. Consider again the local coordinate system \((U,(z,w))\) where \(C\cap U=\{w=0\}.\) We can then write \[\begin{align} \mathrm{T}\mathcal{W}_{d-\nu}|_{U}=\left\{\prod\limits_{l=1}^{d-\nu}(\mathrm{d}z+\tfrac{1}{\nu}g_{l}(z,w)\mathrm{d}w)=0\right\}. \end{align}\] First assume that \(\mathcal{W}_\nu\) is irreducible. By passing to the ramified covering \(\pi(x,y)=(z,w)=(x,y^{\nu})\) and using the second assertion of Lemma 13, we obtain that \(\pi^{*}\mathcal{W}_{\nu}=\boxtimes_{j=1}^{\nu}\mathcal{F}_j\) and \(\pi^{*}\mathcal{W}_{d-\nu}=\boxtimes_{l=1}^{d-\nu}\mathcal{G}_{l},\) where \[\begin{align} \mathcal{F}_j:\mathrm{d}x+\zeta^{-j\kappa}y^{\nu-\kappa-1}f(x,\zeta^jy)\mathrm{d}y=0,&& \mathcal{G}_{l}:\mathrm{d}x+y^{\nu-1}g_{l}(x,y^{\nu})\mathrm{d}y=0, \end{align}\] with \(1\leq\kappa<\nu\), \(\zeta=\exp(\tfrac{2\mathrm{i}\pi}{\nu})\) and \(f(x,0)\not\equiv0.\)

We then have \[\begin{align} K(\pi^*\mathcal{W})-K(\pi^*\mathcal{W}_\nu)-K(\pi^*\mathcal{W}_{d-\nu}) &=\sum_{\substack{1\le j<j'\le\nu\\ 1\le l\le d-\nu}}K(\mathcal{F}_j\boxtimes\mathcal{F}_{j'}\boxtimes\mathcal{G}_l)+\sum_{\substack{1\le j\le\nu\\ 1\le l<l'\le d-\nu}}K(\mathcal{F}_j\boxtimes\mathcal{G}_l\boxtimes\mathcal{G}_{l'})\\ &=\sum_{\substack{1\le j<j'\le\nu\\ 1\le l\le d-\nu}}\mathrm{d}\eta_{jj'l}+\sum_{\substack{1\le j\le\nu\\ 1\le l<l'\le d-\nu}}\mathrm{d}\eta_{jll'}. \end{align}\] where \(\eta_{jj'l}:=\eta(\mathcal{F}_j\boxtimes\mathcal{F}_{j'}\boxtimes\mathcal{G}_l)\)   and   \(\eta_{jll'}:=\eta(\mathcal{F}_j\boxtimes\mathcal{G}_l\boxtimes\mathcal{G}_{l'}).\)

Denoting by \(\varphi_{\ell}(x,y)=(x,\zeta^{\ell}y)\), \(\ell=1,\ldots,\nu,\) the Deck transformations of \(\pi\), we deduce that \[\begin{align} K(\pi^*\mathcal{W})-K(\pi^*\mathcal{W}_\nu)-K(\pi^*\mathcal{W}_{d-\nu}) &=\pi^*\Big(K(\mathcal{W})-K(\mathcal{W}_\nu)-K(\mathcal{W}_{d-\nu})\Big)\\ &=\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_\ell^*\pi^*\Big(K(\mathcal{W})-K(\mathcal{W}_\nu)-K(\mathcal{W}_{d-\nu})\Big)\\ &=\sum_{\substack{1\le j<j'\le\nu\\1\le l\le d-\nu}}\mathrm{d}\left(\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_\ell^*\eta_{jj'l}\right) +\sum_{\substack{1\le j\le\nu\\ 1\le l<l'\le d-\nu}}\mathrm{d}\left(\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_\ell^*\eta_{jll'}\right). \end{align}\] Note that if \(n\not\equiv0\mod\nu\) then \(\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_{\ell}^*(y^{n}\mathrm{d}x)=0\), and if \(n\equiv -1 \mod \nu\) then \(\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_{\ell}^*(y^{n}\mathrm{d}y)=y^{n}\mathrm{d}y\). Moreover, if we write \[\begin{align} \eta_{jj'l}=A_{jj'l}(x,y)\mathrm{d}x+B_{jj'l}(x,y)\mathrm{d}y &&\text{and}&& \eta_{jll'}=A_{jll'}(x,y)\mathrm{d}x+B_{jll'}(x,y)\mathrm{d}y, \end{align}\] Lemma 15 shows that each of \(A_{jj'l}\) and \(A_{jll'}\) has poles of order \(\leq\nu-\kappa\leq\nu-1\) along \(y=0\), and that each of \(B_{jj'l}\) and \(B_{jll'}\) is logarithmic along \(y=0\) with constant residue. It follows that \(\mathrm{d}\left(\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_{\ell}^*\eta_{jj'l}\right)\) and \(\mathrm{d}\left(\frac{1}{\nu}\sum_{\ell=1}^{\nu}\varphi_{\ell}^*\eta_{jll'}\right)\) are holomorphic along \(\tilde{C}=\{y=0\}\), and hence so is \(K(\pi^*\mathcal{W})-K(\pi^*\mathcal{W}_\nu)-K(\pi^*\mathcal{W}_{d-\nu})\). Since \(\mathcal{W}_{d-\nu}\) is regular, \(K(\mathcal{W}_{d-\nu})\) is holomorphic along \(C\), and we deduce that \(K(\mathcal{W})-K(\mathcal{W}_{\nu})\) is also holomorphic along \(C.\)

Now assume that \(\mathcal{W}_\nu=\boxtimes_{\alpha=1}^{r}\mathcal{W}_{\nu_{\alpha}}\) with \(r\geq2\), each \(\mathcal{W}_{\nu_{\alpha}}\) being an irreducible \(\nu_{\alpha}\)-web having \(C\) as a weakly invariant curve. Consider the ramified covering \(\pi\colon(x,y)\mapsto(x,y^{\tilde{\nu}})\), where \(\tilde{\nu}:=\prod_{\alpha=1}^{r}\nu_{\alpha}.\) The pull-back of \(\mathcal{W}_{d-\nu}\) by \(\pi\) writes as \(\pi^{*}\mathcal{W}_{d-\nu}=\boxtimes_{l=1}^{d-\nu}\mathcal{G}_{l},\) where \[\mathcal{G}_{l}:\mathrm{d}x+y^{\tilde{\nu}-1}g_{l}(x,y^{\tilde{\nu}})\mathrm{d}y=0.\] As for \(\pi^{*}\mathcal{W}_{\nu}\), we have \(\pi=\pi_{\alpha}\circ\pi^{'}_{\alpha}\), where \(\pi_{\alpha}\colon(x,y)\mapsto(x,y^{\nu_{\alpha}})\) and \(\pi^{'}_{\alpha}\colon(x,y)\mapsto(x,y^{m_\alpha})\) with \(m_\alpha:= \frac{\tilde{\nu}}{\nu_\alpha}\). By Lemma 13, \(\pi_\alpha^*\mathcal{W}_{\nu_{\alpha}}=\boxtimes_{j=1}^{\nu_{\alpha}}\mathcal{F}_{j,0}^{\alpha}\), where \[\mathcal{F}_{j,0}^{\alpha}\,:\,\mathrm{d}x+\zeta_{\alpha}^{-j\kappa_{\alpha}}y^{\nu_{\alpha}-\kappa_{\alpha}-1}f_\alpha(x,\zeta_{\alpha}^{j}y)\mathrm{d}y=0,\] with \(1\leq\kappa_{\alpha}<\nu_\alpha\), \(\zeta_{\alpha}=\exp(\tfrac{2\mathrm{i}\pi}{\nu_{\alpha}})\) and \(f_\alpha(x,0)\not\equiv0\). It follows that \(\pi^*\mathcal{W}_{\nu}=\boxtimes_{\alpha=1}^{r}\boxtimes_{j=1}^{\nu_{\alpha}}\mathcal{F}_{j}^{\alpha},\) where \[\mathcal{F}_{j}^{\alpha}\,:\, \mathrm{d}x+m_{\alpha}\zeta_{\alpha}^{-j\kappa_{\alpha}}y^{\tilde{\nu}-\tilde{\kappa}_{\alpha}-1}f_{\alpha}(x,\zeta_{\alpha}^{j}y^{m_{\alpha}})\mathrm{d}y=0,\] with \(\tilde{\kappa}_{\alpha}:=m_{\alpha}\kappa_{\alpha}\),  \(1\leq\tilde{\kappa}_{\alpha}<\tilde{\nu}\).

We are thus reduced to a situation analogous to the case where \(\mathcal{W}_\nu\) was irreducible; the same argument shows that \(K(\mathcal{W})-K(\mathcal{W}_\nu)\) is holomorphic along \(C\). ◻

In Section §3, we will need the following lemma to establish Theorem 11.

Lemma 16. Let \(\mathcal{W}_n=\mathcal{F}_1\boxtimes\cdots\boxtimes\mathcal{F}_n\) be a completely decomposable \(n\)-web having a totally invariant irreducible curve \(C.\) Let \(\mathcal{W}_{d-n}=\mathcal{F}^{'}_{1}\boxtimes\cdots\boxtimes\mathcal{F}^{'}_{d-n}\) be a completely decomposable \((d-n)\)-web transverse to \(C.\) Assume that the curvature of \(\mathcal{W}_{d-n}\) is holomorphic along \(C.\) Then \(K(\mathcal{W})-K(\mathcal{W}_n)\) is holomorphic along \(C.\) In particular, \(K(\mathcal{W})\) is holomorphic along \(C\) if and only if \(K(\mathcal{W}_n)\) is holomorphic along \(C.\)

**Proof. We have \[\begin{align} K(\mathcal{W})-K(\mathcal{W}_n)=K(\mathcal{W}_{d-n}) +\sum_{\substack{1\le i<j\le n\\ 1\le k\le d-n}}K(\mathcal{F}_i\boxtimes\mathcal{F}_j\boxtimes\mathcal{F}^{'}_{k}) +\sum_{\substack{1\le i\le n\\ 1\le k<k'\le d-n}}K(\mathcal{F}_i\boxtimes\mathcal{F}^{'}_{k}\boxtimes\mathcal{F}^{'}_{k'}). \end{align}\] Now, \(K(\mathcal{F}_i\boxtimes\mathcal{F}_j\boxtimes\mathcal{F}^{'}_{k})\) and \(K(\mathcal{F}_i\boxtimes\mathcal{F}^{'}_{k}\boxtimes\mathcal{F}^{'}_{k'})\) are holomorphic along \(C\) by applying [2]. The lemma then follows from the assumption that \(K(\mathcal{W}_{d-n})\) is holomorphic along \(C.\) ◻

3 Proofs of the main results↩︎

**Proof of Theorem 6. In a neighborhood of a generic point of \(C\), we can decompose \(\mathcal{W}_n\) as \(\mathcal{W}_n=\mathcal{W}_{\nu}\boxtimes\mathcal{W}_{\nu'}\), with \(\mathcal{W}_{\nu}=\boxtimes_{\alpha=1}^{r}\mathcal{W}_{\nu_{\alpha}}\) and \(\mathcal{W}_{\nu'}=\boxtimes_{\beta=1}^{s}\mathcal{W}_{\nu^{'}_{\beta}}\), where each \(\mathcal{W}_{\nu_{\alpha}}\) (resp. \(\mathcal{W}_{\nu^{'}_{\beta}}\)) is an irreducible \(\nu_{\alpha}\)-web (resp. \(\nu^{'}_{\beta}\)-web) admitting \(C\) as a strongly (resp. weakly) invariant curve. By choosing local coordinates \((z,w)\) such that \(C=\{w=0\}\) and passing to the ramified covering \(\pi\colon(x,y)\mapsto(z,w)=(x,y^{\tilde{\nu}})\), where \(\tilde{\nu}=\displaystyle\prod_{\alpha=1}^{r}\nu_{\alpha}\displaystyle\prod_{\beta=1}^{s}\nu^{'}_{\beta}\), we obtain that \(\pi^*\mathcal{W}_\nu=\boxtimes_{i=1}^{\nu}\mathcal{F}_i\), \(\pi^*\mathcal{W}_{\nu'}=\boxtimes_{j=1}^{\nu'}\mathcal{G}_j\) and \(\pi^*\mathcal{W}_{d-n}=\boxtimes_{k=1}^{d-n}\mathcal{H}_{k}\), where the \(\mathcal{F}_i\) are foliations having \(\tilde{C}=\{y=0\}\) as an invariant curve, and the \(\mathcal{G}_{j}\) and \(\mathcal{H}_{k}\) are foliations transverse to \(\tilde{C}.\)

We then have

\[\begin{align} K(\pi^*\mathcal{W})-K(\pi^*\mathcal{W}_n)&=K\big(\pi^*(\mathcal{W}_{\nu'}\boxtimes\mathcal{W}_{d-n})\big)-K(\pi^*\mathcal{W}_{\nu'}) +\sum_{\substack{1\le i<i'\le\nu\\ 1\le k\le d-n}}K(\mathcal{F}_i \boxtimes\mathcal{F}_{i'}\boxtimes \mathcal{H}_k) +\sum_{\substack{1\le i\le\nu\\ 1\le j\le\nu'\\ 1\le k\le d-n}}K(\mathcal{F}_i\boxtimes\mathcal{G}_j\boxtimes\mathcal{H}_k)\\ & +\sum_{\substack{1\le i\le\nu\\ 1\le k<k'\le d-n}}K(\mathcal{F}_i \boxtimes\mathcal{H}_k\boxtimes\mathcal{H}_{k'}). \end{align}\]

Note that, by [2], the curvatures \(K(\mathcal{F}_i \boxtimes\mathcal{F}_{i'}\boxtimes \mathcal{H}_k)\), \(K(\mathcal{F}_i\boxtimes\mathcal{G}_j\boxtimes\mathcal{H}_k)\) and \(K(\mathcal{F}_i \boxtimes\mathcal{H}_k\boxtimes\mathcal{H}_{k'})\) are holomorphic along \(\tilde{C}\). Moreover, \(K\big(\pi^*(\mathcal{W}_{\nu'}\boxtimes\mathcal{W}_{d-n})\big)-K(\pi^*\mathcal{W}_{\nu'})=\pi^*\Big(K(\mathcal{W}_{\nu'}\boxtimes\mathcal{W}_{d-n})-K(\mathcal{W}_{\nu'})\Big)\) is holomorphic on \(\tilde{C}\), thanks to Proposition 14. It follows that \(K(\pi^*\mathcal{W})-K(\pi^*\mathcal{W}_n)\) is holomorphic along \(\tilde{C}\), and therefore \(K(\mathcal{W})-K(\mathcal{W}_n)\) is holomorphic along \(C.\) ◻

**Proof of Proposition 7. We argue as in the proof of Proposition 14, using the same notation. Writing \(\mathcal{W}_\nu=\boxtimes_{\alpha=1}^{r}\mathcal{W}_{\nu_\alpha}\) and passing to the ramified covering \(\pi\colon(x,y)\mapsto(z,w)=(x,y^{\tilde{\nu}})\), where \(\tilde{\nu}=\prod_{\alpha=1}^{r}\nu_\alpha\), we obtain that \(\pi^*\mathcal{W}_{\nu}=\boxtimes_{\alpha=1}^{r}\boxtimes_{j=1}^{\nu_{\alpha}}\mathcal{F}_{j}^{\alpha},\) where each \(\mathcal{F}_j^\alpha\) is a foliation transverse to \(\tilde{C}=\{y=0\}\) and defined by \[\omega_j^\alpha=\mathrm{d}x+m_{\alpha}\zeta_{\alpha}^{-j\kappa_{\alpha}}y^{n_\alpha}f_{\alpha}(x,\zeta_{\alpha}^{j}y^{m_{\alpha}})\mathrm{d}y,\] with \(n_\alpha=\tilde{\nu}-\tilde{\kappa}_\alpha-1\), \(\tilde{\kappa}_\alpha=m_\alpha\kappa_\alpha=\tilde{\nu}\rho_\alpha\), \(\rho_\alpha=\frac{\kappa_\alpha}{\nu_\alpha}<1\), \(1\le\tilde{\kappa}_\alpha<\tilde{\nu}\)  and  \(f_\alpha(x,0)\not\equiv0.\)

Let us denote \(\eta^{\alpha\beta\gamma}_{jj'j''}=\eta(\mathcal{F}_j^\alpha\boxtimes\mathcal{F}_{j'}^\beta\boxtimes\mathcal{F}_{j''}^\gamma)\); we can write \[K(\pi^*\mathcal{W}_\nu)=\sum_{1\le\alpha\le\beta\le\gamma\le r} \sum_{\substack{1\le j\le\nu_\alpha\\ 1\le j'\le\nu_\beta\\ 1\le j''\le\nu_\gamma}} \mathrm{d} \eta^{\alpha\beta\gamma}_{jj'j''}.\]

Setting \(\varphi_{\ell}(x,y)=(x,\zeta^{\ell}y)\), \(\ell=1,\ldots,\tilde{\nu},\) where \(\zeta=\exp(\tfrac{2\mathrm{i}\pi}{\tilde{\nu}}),\) we have \[\begin{align} K(\pi^*\mathcal{W}_\nu) =\pi^*K(\mathcal{W}_\nu) =\frac{1}{\tilde{\nu}}\sum_{\ell=1}^{\tilde{\nu}}\varphi_\ell^*\pi^*K(\mathcal{W}_\nu) =\sum_{1\le\alpha\le\beta\le\gamma\le r}\sum_{\substack{1\le j\le\nu_\alpha\\1\le j'\le\nu_\beta\\1\le j''\le \nu_\gamma}}\mathrm{d} \left( \frac{1}{\tilde{\nu}}\sum_{\ell=1}^{\tilde{\nu}}\varphi_\ell^*\eta^{\alpha\beta\gamma}_{jj'j''} \right). \end{align}\]

Using conditions (\(\mathfrak{a}\)), resp. (\(\mathfrak{b}\)), resp. (\(\mathfrak{c}\)), we easily verify that the \(3\)-webs \(\mathcal{F}_j^\alpha\boxtimes\mathcal{F}_{j'}^\alpha\boxtimes\mathcal{F}_{j''}^\alpha\), resp. \(\mathcal{F}_j^\alpha\boxtimes\mathcal{F}_{j'}^\alpha\boxtimes\mathcal{F}_{j''}^\gamma\) with \(\alpha\neq\gamma\), resp. \(\mathcal{F}_j^\alpha\boxtimes\mathcal{F}_{j'}^\beta\boxtimes\mathcal{F}_{j''}^\gamma\) with \(\alpha<\beta<\gamma\), have the form of Lemma 15, so that \[\eta_{jj'j''}^{\alpha\beta\gamma}=\left(\frac{g_{jj'j''}^{\alpha\beta\gamma}(x)}{y^{n_{\alpha\beta\gamma}+1}}+\cdots\right)\mathrm{d}x +\left(\frac{c_{jj'j''}^{\alpha\beta\gamma}}{y}+\cdots\right)\mathrm{d}y,\] where \(n_{\alpha\beta\gamma}=\min(n_\alpha,n_\beta,n_\gamma)\), \(g_{jj'j''}^{\alpha\beta\gamma}\) is a holomorphic function of \(x\) and \(c_{jj'j''}^{\alpha\beta\gamma}\in\mathbb{C}.\) Since \(n_{\alpha\beta\gamma}+1\le\tilde{\nu}-1\), we deduce (cf. proof of Proposition 14) that \(\mathrm{d}\Big(\frac{1}{\tilde{\nu}}\sum_{\ell=1}^{\tilde{\nu}}\varphi_\ell^*\eta_{jj'j''}^{\alpha\beta\gamma}\Big)\) is holomorphic along \(\tilde{C}=\{y=0\}\), and hence so is \(K(\pi^*\mathcal{W}_\nu)\). Consequently, \(K(\mathcal{W}_\nu)\) is holomorphic along \(C=\{w=0\}\). ◻

**Proof of Corollary 9. If \(\nu=2\), [2] ensures that the curvature of \(\mathcal{W}\) is holomorphic along \(C\).

Assume that \(\nu\geq3.\) According to the first assertion of Lemma 13, we have \[\kappa(\nu-1)\leq\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)<3(\nu-1),\] hence \(\kappa<3\), and therefore \(\kappa\in\{1,2\}\subset\{1,\ldots,\nu-1\}\). It follows that \(C\) is weakly invariant by \(\mathcal{W}_\nu\) and that \(\gcd(\nu,\kappa)\leq2.\) The web \(\mathcal{W}_\nu\) is then non-degenerate along \(C.\) By Proposition 7, \(K(\mathcal{W}_\nu)\) is holomorphic on \(C\), and by Theorem 6, the same holds for \(K(\mathcal{W})\). ◻

**Proof of Corollary 10. By Lemma 13, we have \[\kappa(\nu-1)\leq\mathrm{mult}\left(\Delta(\mathcal{W}_{\nu}),C\right)<\nu(\nu-1),\] hence \(\kappa<\nu\). The curve \(C\) is then weakly invariant by \(\mathcal{W}_\nu.\) Furthermore, the assumption on \(\nu\) and the inequality \(\kappa<\nu\) imply that \(\gcd(\nu,\kappa)\leq2.\) It follows that \(\mathcal{W}_\nu\) is non-degenerate along \(C\), so that \(K(\mathcal{W}_\nu)\) is holomorphic along \(C\) by Proposition 7, and so is \(K(\mathcal{W})\) by Theorem 6. ◻

**Proof of Theorem 11. Let us decompose \(\mathcal{W}_{n}^{\rm{str}}\) and \(\mathcal{W}_{n}^{\rm{wk}}\) into \(\mathcal{W}_{n}^{\rm{str}}=\boxtimes_{\alpha=1}^{r}\mathcal{W}_{\nu_\alpha}\) and \(\mathcal{W}_{n}^{\rm{wk}}=\boxtimes_{\beta=1}^{s}\mathcal{W}_{\nu^{'}_{\beta}}\), where each \(\mathcal{W}_{\nu_{\alpha}}\) (resp. \(\mathcal{W}_{\nu^{'}_{\beta}}\)) is an irreducible \(\nu_{\alpha}\)-web (resp. \(\nu^{'}_{\beta}\)-web) having \(C\) as a strongly (resp. weakly) invariant curve. We take local coordinates \((z,w)\) such that \(C=\{w=0\}\). By passing to the ramified covering \(\pi\colon(x,y)\mapsto(z,w)=(x,y^{\nu})\), where \(\nu=\displaystyle\prod_{\alpha=1}^{r}\nu_{\alpha}\displaystyle\prod_{\beta=1}^{s}\nu^{'}_{\beta}\), we obtain that \(\pi^*\mathcal{W}_{n}^{\rm{str}}=\boxtimes_{i=1}^{n_1}\mathcal{F}_i\) and \(\pi^*(\mathcal{W}_{n}^{\rm{wk}}\boxtimes\mathcal{W}_{d-n})=\boxtimes_{j=1}^{d-n_1}\mathcal{F}'_j\), where the \(\mathcal{F}_i\) are foliations having \(\tilde{C}=\{y=0\}\) as an invariant curve, the \(\mathcal{F}'_{j}\) are foliations transverse to \(\tilde{C}\), and \(n_1=\sum_{\alpha=1}^{r}\nu_\alpha.\)

The web \(\mathcal{W}_{n}^{\rm{wk}}\) being non-degenerate along \(C\), Proposition 7 ensures that \(K(\mathcal{W}_{n}^{\rm{wk}})\) is holomorphic on \(C\). Since \(\mathcal{W}_{d-n}\) is regular and transverse to \(C\), Theorem 6 implies that \(K(\mathcal{W}_{n}^{\rm{wk}}\boxtimes\mathcal{W}_{d-n})\) is also holomorphic on \(C.\) It follows that \(K\big(\pi^*(\mathcal{W}_{n}^{\rm{wk}}\boxtimes\mathcal{W}_{d-n})\big)\) is holomorphic along \(\tilde{C}.\) We can then apply Lemma 16 to the web \(\pi^*\mathcal{W}=\pi^*\mathcal{W}_{n}^{\rm{str}}\boxtimes\pi^*(\mathcal{W}_{n}^{\rm{wk}}\boxtimes\mathcal{W}_{d-n})\) and deduce that \(K(\pi^*\mathcal{W})-K(\pi^*\mathcal{W}_{n}^{\rm{str}})\) is holomorphic on \(\tilde{C}.\) Consequently, \(K(\mathcal{W})-K(\mathcal{W}_{n}^{\rm{str}})\) is holomorphic along \(C\). ◻

4 Examples↩︎

We give examples showing that, in Proposition 7, the assumption that \(\mathcal{W}_\nu\) is non-degenerate along \(C\) is indispensable; the condition that \(C\) is weakly invariant by the irreducible subwebs of \(\mathcal{W}_\nu\) is not sufficient on its own to guarantee the holomorphy of \(K(\mathcal{W}_\nu)\) along \(C.\) More precisely, we will see that if one of the conditions (\(\mathfrak{a}\)), (\(\mathfrak{b}\)) or (\(\mathfrak{c}\)) is not satisfied, then \(K(\mathcal{W}_\nu)\) is not necessarily holomorphic along \(C.\)

Example 17. Consider the \(6\)-web \(\mathcal{W}\) on \((\mathbb{C}^2,0)\) defined in local coordinates \((z,w)\) by \[\begin{align} \omega=\mathrm{d}w^6-3w\big(z^2+2w\big)\mathrm{d}z^2\mathrm{d}w^4+3w^2\big(z^4-2zw-2zw^2+3w^2\big)\mathrm{d}z^4\mathrm{d}w^2 -w^3\big(z^3-3zw+w+w^2\big)^2\mathrm{d}z^6. \end{align}\] In a neighborhood of a generic point of \(C=\{w=0\}\), the slopes \(p_j\) \((j=1,\ldots,6)\) of \(\mathrm{T}_{(z,w)}\mathcal{W}\) are given by \[\begin{align} p_j=\zeta^{3j}zw^{\frac{1}{2}}+\zeta^{5j}w^{\frac{5}{6}}+\zeta^{j}w^{\frac{7}{6}}, \qquad \text{where}\zeta=\exp(\tfrac{2\mathrm{i}\pi}{6}). \end{align}\] In particular, \(\mathcal{W}\) is irreducible and admits \(C\) as a weakly invariant curve with Puiseux index \(\kappa=3.\) Hence \(\gcd(\nu,\kappa)=3>2\), so condition (\(\mathfrak{a}\)) is not satisfied. Therefore \(\mathcal{W}\) is degenerate along \(C.\)

After passing to the degree \(6\) cover \(\pi\colon(x,y)\mapsto(z,w)=(x,y^6),\) an explicit computation shows that \[K(\pi^*\mathcal{W})=\left(-\frac{4}{y}+\cdots\right)\mathrm{d}x\wedge\mathrm{d}y.\] Thus \(K(\pi^*\mathcal{W})\) is not holomorphic along \(\tilde{C}=\{y=0\}\), and consequently \(K(\mathcal{W})\) is not holomorphic along \(C=\{w=0\}.\)

Example 18. Let \(\mathcal{W}=\mathcal{W}_1\boxtimes\mathcal{W}_2\) be the \(6\)-web on \((\mathbb{C}^2,0),\) where \(\mathcal{W}_1\), resp. \(\mathcal{W}_2\), is the \(2\)-web, resp. the \(4\)-web, defined by \[\begin{align} \omega_1=\mathrm{d}w^2-(z+1)w\mathrm{d}z^2,&& \text{resp.} \omega_2=\mathrm{d}w^4-2w\mathrm{d}w^2\mathrm{d}z^2-4w^2\mathrm{d}w\mathrm{d}z^3-(w-1)w^2\mathrm{d}z^4. \end{align}\] Then the curve \(C=\{w=0\}\) is totally invariant by each of the webs \(\mathcal{W}_1\) and \(\mathcal{W}_2\). According to [2], \(C\) has minimal multiplicity \(\nu_1-1=1\) in \(\Delta(\mathcal{W}_1)\); moreover, \(\mathcal{W}_1\) is irreducible.

As for \(\mathcal{W}_2\), in a neighborhood of a generic point of \(C\), the slopes \(p_j\) \((j=1,\ldots,4)\) of \(\mathrm{T}_{(z,w)}\mathcal{W}_2\) can be written as \[\begin{align} p_j=\zeta^{2j}w^{\frac{1}{2}}+\zeta^{3j}w^{\frac{3}{4}}, \qquad \text{where}\zeta=\exp(\tfrac{2\mathrm{i}\pi}{4})=\mathrm{i}\,; \end{align}\] it follows that \(\mathcal{W}_2\) is also irreducible.

We observe that each of the webs \(\mathcal{W}_1\) and \(\mathcal{W}_2\) admits \(C\) as a weakly invariant curve, with respective Puiseux indices \(\kappa_1=1\) and \(\kappa_2=2.\) Then condition \((\mathfrak{b})\) fails, since \(\rho_1:=\frac{\kappa_1}{\nu_1}=\frac{1}{2}\) and \(\rho_2:=\frac{\kappa_2}{\nu_2}=\frac{2}{4}=\rho_1\), but \(\gcd(\nu_2,\kappa_2)=2\neq1.\) As a result, \(\mathcal{W}\) is degenerate along \(C.\)

Let us consider the degree \(4\) cover \(\pi\colon(x,y)\mapsto(z,w)=(x,y^4)\). An explicit computation leads to \[K(\pi^*\mathcal{W})=\left(-\frac{4}{x^2y}+\cdots\right)\mathrm{d}x\wedge\mathrm{d}y,\] so that \(K(\pi^*\mathcal{W})\) is not holomorphic on \(\tilde{C}=\{y=0\}\), and hence \(K(\mathcal{W})\) is not holomorphic on \(C=\{w=0\}.\)

Example 19. Consider the \(6\)-web \(\mathcal{W}=\mathcal{W}_1\boxtimes\mathcal{W}_2\boxtimes\mathcal{W}_3\) on \((\mathbb{C}^2,0),\) where \(\mathcal{W}_1\), \(\mathcal{W}_2\) and \(\mathcal{W}_3\) are the \(2\)-webs defined in local coordinates \((z,w)\) respectively by \[\begin{align} \omega_1=\mathrm{d}w^2-w\mathrm{d}z^2,&& \omega_2=\mathrm{d}w^2-(w+1)w\mathrm{d}z^2,&& \omega_3=\mathrm{d}w^2-z^2w\mathrm{d}z^2. \end{align}\] Then, for \(\alpha=1,2,3\), the curve \(C=\{w=0\}\) is totally invariant by \(\mathcal{W}_\alpha\). Lemma 2.5 of [2] ensures that \(C\) has minimal multiplicity \(1\) in \(\Delta(\mathcal{W}_\alpha)\); furthermore, each \(\mathcal{W}_\alpha\) is irreducible.

We remark that \(C\) is weakly invariant by each \(\mathcal{W}_\alpha\), with Puiseux index \(\mathfrak{i}(\mathcal{W}_\alpha,C)=1.\) Setting \(\rho_\alpha=\rho(\mathcal{W}_\alpha,C)\), we have \(\rho_1=\rho_2=\rho_3=\frac{1}{2}\). Hence condition (\(\mathfrak{c}\)) is not satisfied, and \(\mathcal{W}\) is degenerate along \(C.\)

Pulling-back \(\mathcal{W}\) by the double cover \(\pi\colon(x,y)\mapsto(z,w)=(x,y^2)\), an explicit computation gives \[K(\pi^*\mathcal{W})=\left(-\frac{16x}{(x^2-1)^2y}+\cdots\right)\mathrm{d}x\wedge\mathrm{d}y.\] Thus \(K(\pi^*\mathcal{W})\) is not holomorphic on \(\tilde{C}=\{y=0\}\), and therefore \(K(\mathcal{W})\) is not holomorphic on \(C=\{w=0\}.\)

References↩︎

[1]
J. V. Pereira and L. Pirio. Classification of exceptional CDQL webs on compact complex surfaces. Int. Math. Res. Not. IMRN, 12:2169–2282, 2010.
[2]
D. Marı́n and J. V. Pereira. Rigid flat webs on the projective plane. Asian J. Math. 17(1):163–191, 2013.
[3]
S. Bedrouni and D. Marı́n. A criterion for the holomorphy of the curvature of smooth planar webs and applications to dual webs of homogeneous foliations on \(\mathbb{P}^{2}_{\mathbb{C}}\). Math. Nachr. 297(11):3964–3981, 2024.
[4]
J. V. Pereira and L. Pirio. An invitation to web geometry, volume 2 of IMPA Monographs. Springer, Cham, 2015.
[5]
S. Bedrouni. Le tissu dual d’un pré-feuilletage convexe réduit sur \(\mathbb{P}^{2}_{\mathbb{C}}\) est plat, , 2024.