Abstract
Let G and X be germs of holomorphic vector fields at \(0\in {\mathbb {C}}^n\). Consider the real analytic map \(\psi _{G,X}:{\mathbb {C}}^n\rightarrow {\mathbb {C}}\) defined by \(\psi _{G,X}(z)=\langle G(z),X(z)\rangle \), where \(\langle \cdot ,\cdot \rangle \) represents the usual Hermitian product. In this paper, we investigate the following question: under which conditions on the germs of holomorphic vector fields G and X is the real analytic hypersurface \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\) Levi-flat? This problem was posed by Maria A. Soares Ruas.
Similar content being viewed by others
Avoid common mistakes on your manuscript.
1 Introduction and Statement of Results
Let M be a real analytic hypersurface at the origin \(0\in {\mathbb {C}}^n\), \(n\ge 2\), defined by the equation \(F(z_1,\ldots ,z_n)=0\), where F is a real analytic function vanishing at 0. We say that M is nondegenerate if the Levi form
is nondegenerate at 0. Otherwise, we say that M is Levi-flat. The purpose of this paper is to study the degeneracy of the Levi form of real analytic hypersurfaces obtained from real singularities with a Milnor fibration. More precisely, Milnor proved in Milnor (1968, Theorem 11.2) that if \(f:({\mathbb {R}}^n,0)\rightarrow ({\mathbb {R}}^p,0)\), \(n>p\), is a real analytic map whose derivative Df has rank p on a punctured neighborhood of \(0\in {\mathbb {R}}^n\), then, for every sufficiently small sphere \(\mathbb {S}_{\epsilon } \subset {\mathbb {R}}^n\), the mapping
is a locally trivial fibration, where \(K=f^{-1}(0) \cap \mathbb {S}_{\epsilon }\) is the singularity link, and \(N_K\) is a tubular neighborhood of K in \(\mathbb {S}_{\epsilon }\). The map \(\psi \) can always be extended to \(\mathbb {S}_{\epsilon }-K\) as the projection of a fibration, but this extension is not necessarily as \(\dfrac{f}{\Vert f\Vert }\). Follows Ruas et al. (2002, Definition 1.1), we will say that \(f:({\mathbb {R}}^n,0)\rightarrow ({\mathbb {R}}^p,0)\), \(n>p\), satisfies the Milnor condition at 0 if Df has rank p on a punctured neighborhood of 0. When f satisfies the Milnor condition at 0, and furthermore, the map \(\dfrac{f}{\Vert f\Vert }: \mathbb {S}_{\epsilon }-K \rightarrow \mathbb {S}^{p-1}\) is a fibration for every sufficiently small sphere \(\mathbb {S}_{\epsilon } \subset {\mathbb {R}}^n\), we say that f satisfies the strong Milnor condition at 0, see for instance (Ruas et al. 2002, Definition 2.5). Maps of this type induce an open book decomposition on the sphere \(\mathbb {S}_{\epsilon }\). Milnor pointed out in his book that is difficult to find examples satisfying the strong Milnor condition, (see Milnor 1968, p. 100). In Seade (1997) and Seade (1996), Seade presented a method for constructing families of nontrivial maps \(f:{\mathbb {R}}^{2n} \rightarrow {\mathbb {R}}^2\) that satisfy the strong Milnor condition at 0. This construction is given as follows: let \(\chi ({\mathbb {C}}^n,0)\) denote the space of all germs of holomorphic vector fields at \(0\in {\mathbb {C}}^n\), and let G, X be elements in \(\chi ({\mathbb {C}}^n,0)\). Consider the real analytic map
defined by \(\psi _{G,X}(z)=\langle G(z),X(z)\rangle \), where
is the usual Hermitian product. Note that the argument of \(i\langle G(z), X(z)\rangle \) is the angle by which we rotate the field G so that it becomes orthogonal to the field X. Thus, the real analytic variety \(\psi _{G,X}^{-1}(0)\), called the polar variety of G and X, is the set of points where G and X are orthogonal. Consequently, on the polar variety, the holomorphic foliations defined by the fields G and X are transversal, and their intersection gives rise to a foliation by real curves in \(\psi _{G,X}^{-1}(0)\). In the particular case where X is the gradient field of a real analytic function \(f:{\mathbb {R}}^{2n} \rightarrow {\mathbb {R}}^2\), the polar variety is the set of points where the foliations defined by the field G and the level curves of f are tangent. Furthermore, \(\psi _{G,X}^{-1}(0)\) is a complete intersection defined by the equations
In Seade (1997), Seade proved that if \(X=(z_1,\ldots ,z_n)\) is the radial field and \(G=(\lambda _1z_1^{a_1}, \ldots ,\lambda _nz_n^{a_n})\), then \(\psi _{G,X}\) satisfies the Milnor condition for any \(\lambda _k \in {\mathbb {C}}^*\) and integers \(a_k>1\). On the other hand, given \(\sigma \in S_n\), a permutation of the set \(\underline{n}:=\{1,\ldots ,n\}\), families of vector fields of the form \(G=(\lambda _1z_{\sigma _1}^{a_1}, \ldots , \lambda _nz_{\sigma _n}^{a_n})\) and \(X=(\beta _1z_1^{b_1}, \ldots , \beta _nz_n^{b_n})\) that satisfy the Milnor condition or the strong Milnor condition at the origin were classified by Ruas–Seade–Verjovsky (Ruas et al. 2002, Theorem 2.7).
In this paper, we consider \(M=\{F(z)=0\}\) defined by
where \(G,X\in \chi ({\mathbb {C}}^n,0)\). A simple example in \(({\mathbb {C}}^2,0)\) is when we consider \(G(z_1,z_2)=(z_1,z_2)\) and \(X=(z_2,-z_1)\). Then
is a real analytic Levi-flat hypersurface at \(0\in {\mathbb {C}}^2\) whose Levi foliation admits as leaves the complex curves \(z_1=c\cdot z_2\), where \(c\in {\mathbb {C}}\) (see Burns and Gong 2003, p. 51). Motivated by this, M.A. Soares Ruas have posed the following problem:
Problem 1
Under which conditions on the germs of holomorphic vector fields G and X is the real analytic hypersurface \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\) Levi-flat?
In order to answer Problem 1, we consider the vector fields G and X explored in Ruas et al. (2002, pp. 203–211). More specifically, our first result is as follows:
Theorem A
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G(z)=(z_1^{a_1},\ldots , z_n^{a_n})\) and \(X(z)=(z_1^{b_1},\ldots , z_n^{b_n})\), where \(a_k > b_k \ge 1\) are positive integers, for all \(k=1,\ldots ,n\). Then \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is nondegenerate at \(0\in {\mathbb {C}}^n\).
Our second and third theorems are motivated by the example given in (5):
Theorem B
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G(z)=(z_1^a,z_2^b)\) and \(X(z)=(z_2^b,z_1^a)\) with a, b positive integers. Then \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is a Levi-flat hypersurface at \(0\in {\mathbb {C}}^2\).
Theorem C
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G=(z_1^{a_1},z_2^{a_2})\) and \(X=(z_2^{b_2},z_1^{b_1})\), where \(a_1\ge b_1\) and \(a_2 \ge b_2\) are positive integers satisfying \(a_1b_2=a_2b_1\). Then \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is Levi-flat if, and only if, \(a_1=b_1\) and \(a_2=b_2\).
Finally, we consider a family of vector fields studied in Ruas et al. (2002, Theorem 2.1).
Theorem D
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G=(z_1^{a_1},\ldots , z_n^{a_n})\) and \(X=(z_{\sigma _1}^{b_{\sigma _1}},\ldots , z_{\sigma _n}^{b_{\sigma _n}})\), where \(a_k\ge b_k\) are positive integers. Let us assume that for some \(\ell \in \underline{n}\), the integers \(a_{\ell }, b_{\ell }, a_{\sigma _{\ell }}, b_{\sigma _{\ell }}\) satisfy the following conditions: \(a_{\ell }>b_{\ell }\) and \(a_{\ell }b_{\sigma _{\ell }}=b_{\ell }a_{\sigma _{\ell }}\). Then, \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is nondegenerate at \(0\in {\mathbb {C}}^n\).
Following our results it seems that the property of M being Levi-flat is related to the property that the function \(\psi _{G,X}\) does not satisfy the Milnor condition, see Ruas et al. (2002, Theorem 2.1).
The paper is organized as follows: In Sect. 2, we introduce the concept of real analytic Levi-flat hypersurfaces at \((\mathbb {C}^n, 0)\), shedding light on essential properties that will play a pivotal role throughout the paper. Section 3 is dedicated to proving Theorem A. In Sect. 4, we establish the validity of Theorems B and C, while Sect. 5 focuses on the proof of Theorem D. Finally, in Sect. 6, we provide examples for further illustration.
2 Levi-Flat Hypersurfaces
In this section, we will discuss real analytic Levi-flat hypersurfaces at \(({\mathbb {C}}^n,0)\). These are real analytic hypersurfaces whose regular part is foliated by immersed complex submanifolds of codimension one. Levi-flat hypersurfaces naturally arise in the theory of foliations as invariant subsets. In general, germs of codimension one holomorphic foliations that leave invariant hypersurfaces of this type admit a meromorphic first integral (see Cerveau and Lins-Neto 2011, Theorem 1). On the other hand, there are examples of holomorphic webs that leave invariant Levi-flat hypersurfaces (see Da Silva and Fernández-Pérez 2023; Fernández-Pérez 2013; Shafikov and Sukhov 2015). Levi-flat hypersurfaces are a central focus of the development in this paper.
Let \(M=\{F(z)=0\}\) be a germ of real analytic hypersurface at \(0\in {\mathbb {C}}^n\), where \(F:({\mathbb {C}}^n,0)\rightarrow (\mathbb {R},0)\) is a real analytic function at \(0\in {\mathbb {C}}^n\). The singular set of M is denoted by \({\textsf {Sing}}(M)\) and defined by
We define the regular part of M as \(M^{*}:=\{F(z)=0\}{\setminus }\{dF(z)=0\}\). In \(M^{*}\), the Levi distribution is given by \(L_p:=Ker(\partial F(p)) \subset T_p M^{*}\), where \(p \in M^{*}\). Note that \(L_p\) is the unique complex hyperplane contained in \(T_p M^{*}\).
Definition 2.1
We say that M is Levi-flat if the Levi distribution on \(M^{*}\) is integrable. In this case, the Levi distribution induces a foliation on \(M^{*}\) called the Levi foliation, denoted by \(\mathcal {L}\).
The Levi distribution can also be given by the 1-form \(\eta =i(\partial {F}-\bar{\partial } F)\) called the Levi 1-form. Thus, the integrability of the Levi distribution is equivalent to the integrability of the form \(\eta \) in the sense of Frobenius, that is, \(\eta \) is integrable if and only if \(\left. \eta \wedge d\eta \right| _{M^*} \equiv 0\).
The simplest example of a Levi-flat hypersurface is given below.
Example 2.1
In \({\mathbb {C}}^n\) with coordinates \((z_1,\ldots ,z_n)\), consider \(M=\{Im(z_n)=0\}.\) Then M is a smooth Levi-flat real analytic hypersurface, meaning \({\textsf {Sing}}(M)=\emptyset \). The Levi distribution on M is given by \(L_p = \textrm{Ker} (dz_n(p))\), \( p \in M^*.\) The leaves of the Levi foliation on M are given by \(\{z_n=c\}\) where \(c\in {\mathbb {R}}\).
Let’s consider a slightly more elaborate example given by Brunella (2007, Example 1.2).
Example 2.2
With coordinates (z, w) in \({\mathbb {C}}^2\) such that \(z=x+iy\) and \(w=s+it\), the real analytic hypersurface M given by
is Levi-flat, with singular set \({\textsf {Sing}}(M)=\{t=y=0\}\). The leaves of the Levi foliation on \(M^*\) are given by \(L_c=\{w=(z+c)^2: Im(z)\ne 0\}\) with \(c \in {\mathbb {R}}\).
The next result provides the local form of a smooth Levi-flat hypersurface. Essentially, it tells us that, at regular points, every Levi-flat hypersurface is locally similar to the Example 2.1.
Theorem 2.1
[Cartan’s theorem (Cartan 1933)] Let \(M \subset {\mathbb {C}}^n\) be a real analytic Levi-flat hypersurface. In a neighborhood of each point \(p \in M^*\), there exists a holomorphic coordinate system \(z=(z_1, \ldots , z_n)\) such that \(M=\{Im(z_n)=0\}.\)
A criterion for the integrability of the Levi form is given in the next proposition.
Proposition 2.2
Let \(M=\{F(z)=0\}\) be a germ of a real analytic hypersurface at \(0\in {\mathbb {C}}^n\). Then, M is Levi-flat if and only if \(\partial F(p) \wedge \overline{\partial }F(p) \wedge \partial \overline{\partial } F(p) = 0\) for all \(p \in M\).
Proof
Let \(\eta =i(\partial F-\bar{\partial }F)\) denote the Levi 1-form of M. Assuming M is Levi-flat, this implies \(\eta \wedge d\eta |_{M^{*}}=0\), which is equivalent to \((\partial {F}-\bar{\partial }F)\wedge \partial \bar{\partial } F|_{M^{*}}=0\). Consequently, we have \(\bar{\partial }F\wedge \partial \bar{\partial } F|_{M^{*}}=\partial F\wedge \partial \bar{\partial } F|_{M^{*}}\). In particular, \(\partial F(p) \wedge \overline{\partial }F(p) \wedge \partial \overline{\partial } F(p) = 0\) for all \(p \in M\).
Conversely, the condition \(\partial F(p) \wedge \overline{\partial }F(p) \wedge \partial \overline{\partial } F(p) = 0\) for all \(p \in M\) is equivalent to \((\partial {F}-\bar{\partial }F)\wedge \partial \bar{\partial } F\wedge dF |_{M^{*}}=0\). Hence
where \(\theta \) is a 4-form in some open subset of \({\mathbb {C}}^n\). Since \(\theta \wedge dF=0\), we can express \(\theta \) as \(\theta =\beta \wedge dF\), where \(\beta \) is a 3-form in some open subset of \({\mathbb {C}}^n\). Substituting this into Eq. (6), we obtain
Thus, there exists a 2-form \(\kappa \) such that
this expression implies that \((\partial {F}-\bar{\partial }F)\wedge \partial \bar{\partial } F|_{M^{*}}=0\), leading to the integrability of \(\eta \). \(\square \)
Now, let’s verify that the regular part of M is mapped to the regular part of \(M'\) for M and \(M'\) being biholomorphic (not necessarily Levi-flat).
Lemma 2.3
Let \(u \in M'\). Then \(z=z(u) \in {\textsf {Sing}}(M)\) if and only if \(u \in {\textsf {Sing}}(M')\).
Proof
Denote \(M'= \{G(u)=0\}\) where \(G:({\mathbb {C}}^n, 0) \rightarrow ({\mathbb {R}}), 0\) is a real analytic function at \(0 \in {\mathbb {C}}^n\).
We have \(dG(u)=\partial G(u)+\overline{\partial }G(u)\), where
A point \(u \in M'\) belongs to the singular set Sing(\(M'\)) if and only if all partial derivatives \(\dfrac{\partial G}{\partial u_j}(u), \dfrac{\partial G}{\partial \bar{u}_j}(u)\) are identically zero. By the chain rule, we obtain
thus, it follows that
Since \(z=z(u)\) is a biholomorphism, the change of coordinates matrix is invertible, and therefore,
\(\square \)
In the next proposition, we will see that the Levi-flat property is invariant under a change of coordinates. More specifically, we establish the following result.
Proposition 2.4
Let \(G: V \subset {\mathbb {C}}^n \rightarrow {\mathbb {R}}\) be a real-analytic function in coordinates \(z=(z_1,\ldots , z_n)\) and let \(z=z(u)\) be a change of coordinates, that is, a biholomorphism from the open set \(V \subset {\mathbb {C}}^n\) to an open set \(U \subset {\mathbb {C}}^n\). Then, the hypersurface \(M'=\{u \in U: G(u)=G(z(u))=0\}\) is Levi-flat if and only if \(M=\{z \in V: G(z)=0\}\) is Levi-flat.
Proof
The partial derivatives of G in the coordinates \(z=(z_1,\ldots ,z_n)\) are given by
Therefore, in coordinates \(z=(z_1, \ldots , z_n)\), we have
Now, by making the change of coordinates \(z=z(u)\), we obtain
from which it follows that
Hence, in coordinates \(u=(u_1,\ldots ,u_n)\),
From the expression above, combined with Eq. (7), it follows that \(M'\) is Levi-flat if and only if M is Levi-flat. This concludes the proof of the Proposition 2.4. \(\square \)
2.1 Complexification
Let \(F:({\mathbb {C}}^n,0)\rightarrow (\mathbb {R},0)\) be a real analytic function at \(0\in {\mathbb {C}}^n\). The complexification of F is defined by
where \(F(z)=\sum _{\mu ,\nu }F_{\mu ,\nu }z^{\mu }\overline{z}^{\nu }\) is a power series of F convergent in a neighborhood of the origin. We observe that \(F_{{\mathbb {C}}}\) is holomorphic at \(({\mathbb {C}}^n\times {\mathbb {C}}^n,0)\). The complexification of M is defined as \(M_{{\mathbb {C}}}=\{F_{{\mathbb {C}}}=0\},\) and the complexification of the Levi 1-form is given by
Given that \(\eta \) is integrable on \(M^{*}\), it follows that \(\eta _{{\mathbb {C}}}\) is also integrable on \(M_{{\mathbb {C}}}^{*}\). It’s worth noting that we can express \(dF_{{\mathbb {C}}}=\alpha + \beta \) and \(\eta _{{\mathbb {C}}}=i(\alpha - \beta )\), where
Furthermore, we observe that \(\alpha \) and \(\beta \) define the same foliation as \(\eta _{{\mathbb {C}}}\) on \(M^*_{{\mathbb {C}}}\). Thus, the integrability of \(\eta _{{\mathbb {C}}}\) is equivalent to
This condition will be used later to verify that the real part of a certain family of polar varieties is not Levi-flat.
3 Proof of Theorem A
Before proving Theorem A, we will make some considerations for the case where the vector fields G and X are given by \(G=(z_1^{a_1},\ldots , z_n^{a_n})\) and \(X=(z_1^{b_1},\ldots , z_n^{b_n})\), with \(a_k, b_k\) positive integers such that \(a_k\ge b_k\), for \(k \in \underline{n}\). In this case, the Hermitian product of G and X is given by
and
Note that, if \(a_k=b_k\) for every \(k \in \underline{n}\), then \(M=\{F=0\}\) is a point. Indeed, if \(a_k=b_k\) for every \(k \in \underline{n}\), we have
and consequently, \(M=\{0\}\). Therefore, in the statement of Theorem A, we do not consider \(a_k=b_k\) for every \(k \in \underline{n}\).
Regarding the singular set of M, we have the following proposition:
Proposition 3.1
Let \(G=(z_1^{a_1}, \ldots , z_n^{a_n})\) and \(X=(z_1^{b_1}, \ldots , z_n^{b_n})\), with \(a_k\ge b_k\ge 1\), and \(M=\{F=0\}\), where \(F(z)=2{\textrm{Re}}\langle G(z), X(z) \rangle \). Then \({\textsf {Sing}}(M)=\{0\}\).
Proof
Taking the partial derivatives of F, we have
and
Then \(dF(z)= \partial F(z) + \overline{\partial } F(z)=0\) if and only if
Clearly, \(0 \in \textrm{Sing}(M)\). Now let’s denote \(I:=\{k \in \underline{n}; a_k=b_k\}\) and \(J:=\{k \in \underline{n}; a_k>b_k\}\). We have \(\underline{n}=I \cup J\) and \(I \cap J=\emptyset \). Consider the following equations from the system above:
If \(k \in I\), the equations above are rewritten as:
and it follows that \(z=(z_1,\ldots ,z_n) \in {\textsf {Sing}}(M)\) implies \(z_k=0\), for all \(k \in I\). Now, let’s assume that \(z=(z_1,\ldots , z_n) \in {\textsf {Sing}}(M)\) and \(z_k \ne 0\), for some \(k \in J\). From (11), we obtain
that yields,
From (12) we get
that is,
Thus, from (11) and (12), we deduce \(\frac{a_k}{b_k}= \frac{b_k}{a_k}\), but this is a contradiction, because \(a_k > b_k\). Hence, \({\textsf {Sing}}(M) = \{0\}\) and \(M^* = M - \{0\}.\) \(\square \)
Now, we prove a technical lemma that will be used in the proof of Theorem A.
Lemma 3.2
Let \(M_{\mathbb {C}}\) be the complexification of \(M=\left\{ F(z)=\displaystyle 2{\textrm{Re}}\left( \sum _{k=1}^{n}z_{k}^{a_k} \overline{z}_{k}^{b_k}\right) =0\right\} \), \(n\ge 3\). Consider the functions
where \(a_m> b_m\ge 1\) are integers for each \(m=1,\ldots ,n\), and \(n\ge 3\). For each triple \((j, k, \ell )\) of indices \(j,k,\ell =1,\ldots ,n\); there exists \((z_0,w_0) \in M_{{\mathbb {C}}}\) such that
Proof
In this case, the complexification of F is given by
For each triple \((j, k, \ell )\), let’s choose \((z_0, w_0) = (z_1,\ldots , z_n, w_1,\ldots , w_n)\) satisfying:
-
1.
\(z_j=z_k=z_{\ell }=w_j=w_k=1\),
-
2.
\(w_{\ell }\) such that \(w_{\ell }^{b_{\ell }}+w_{\ell }^{a_{\ell }}=-4\),
-
3.
\(z_m=w_m=0\) for the remaining indices.
We observe that \((z_0,w_0) \in M_{{\mathbb {C}}}\). In fact, we have
then \(F_{{\mathbb {C}}}(z_0,w_0)=0\) on \(M_{{\mathbb {C}}}^*\). From item (3) above, we obtain
Furthermore, we have
Hence
\(\square \)
Now, let’s restate Theorem A for completeness.
Theorem A
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G(z)=(z_1^{a_1},\ldots , z_n^{a_n})\) and \(X(z)=(z_1^{b_1},\ldots , z_n^{b_n})\), where \(a_k > b_k \ge 1\) are positive integers, for all \(k=1,\ldots ,n\). Then \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is nondegenerate at \(0\in {\mathbb {C}}^n\).
Proof
We divide the proof into two cases \(n=2\) and \(n\ge 3\). For \(n=2\), we examine the Eq. (9) associated to the complexification of \(M=\{F=0\}\), where
Note that, the complexification of F is given by
then \(dF_{{\mathbb {C}}}=\alpha + \beta \), where
Now we have
We use the following notations
So, we have \(\alpha \wedge d\alpha \wedge \beta = (\alpha _1\beta _1 g_2 +\alpha _2 \beta _2 g_1) dz_1\wedge dz_2 \wedge dw_1 \wedge dw_2.\) Therefore,
Now, using the expression (15), we obtain the following relationship on \(M_{{\mathbb {C}}}^*\):
With this, we can rewrite
In this way, we get
Again using (15), we obtain
Hence
that is,
Therefore, by (9), we conclude that \(\eta |_{M^{*}}\) is integrable if and only if \(a_1b_1(a_2-b_2)^2=0\) and \(a_2b_2(a_1-b_1)^2=0\), that is, \(a_2=b_2\) and \(a_1=b_1\). This completes the proof for \(n=2\).
Now, we consider \(n\ge 3\). With the notation introduced in Lemma 3.2, we can rewrite
Then \(d\alpha =\sum _{\ell =1}^n h_{\ell }dw_{\ell }\wedge dz_{\ell }.\) Indeed, we have
Furthermore, we have the following relationships:
Hence, it follows that
Thus, we have
We observe that for \(j \ne k\), the coefficient of the term \(dz_k\wedge dw_{\ell } \wedge dz_{\ell } \wedge dw_j\) in 4-form \(\alpha \wedge d\alpha \wedge \beta \) is exactly \(f_k g_j h_{\ell }\). In other words, if \(j \ne k\),
is the only term in Eq. (16) that is of the form \(\rho \ dz_k\wedge dw_{\ell } \wedge dz_{\ell } \wedge dw_j.\) Thus, given a triple \((j, k, \ell )\) of pairwise distinct indices, by Lemma (3.2) there exists \((z_0,w_0) \in M_{{\mathbb {C}}}\) such that
Therefore, M is not Levi-flat, and we conclude the proof of Theorem A. \(\square \)
4 The Case of Vector Fields \(G=(z_1^{a_1}, z_2^{a_2})\) and \(X=(z_{\sigma _1}^{b_{\sigma _1}}, z_{\sigma _2}^{b_{\sigma _2}})\)
In this section, we will explore some results regarding the singular set of the hypersurface. \(M=\{F(z)=0\}\), where \(F(z)=2{\textrm{Re}}\langle G(z),X(z) \rangle \), \(G=(z_1^{a_1}, z_2^{a_2})\), and \(X=(z_{\sigma _1}^{b_{\sigma _1}}, z_{\sigma _2}^{b_{\sigma _2}})\), where \(\sigma \in \mathcal {S}_n\) is a permutation of the set \(\{1, \dots , n\}\). Let’s start by considering the particular case where \(G=(z_1^a,z_2^b)\) and \(X=(z_2^b,z_1^a)\) with a, b being positive integers. In this case, the Hermitian product of G with X is given by
We observe that \(\langle G,X \rangle \in {\mathbb {R}}\). Therefore, let’s consider
We observe that for these fields, the map \(\psi _{G,X}\) does not satisfy the Milnor condition, as stated in theorem (Ruas et al. 2002, Theorem 2.7). In this case, the hypersurface M will be Levi-flat, as we will show in Theorem B. First, we will verify the following result regarding the singular set of M.
Proposition 4.1
Let M be the hypersurface described in Eq. (17). Then we have the following options for \(\textrm{Sing}(M)\):
-
1.
\(\textrm{Sing}(M)=\{0\}\), if \(b=a=1\),
-
2.
\(\textrm{Sing}(M) =\{z_1=0\}\), if \(b=1\) and \(a>1\),
-
3.
\(\textrm{Sing}(M)=\{z_2=0\}\), if \(a=1\) and \(b>1\),
-
4.
\(\textrm{Sing}(M)=\{z_1=0\} \cup \{z_2=0\}\), if \(a,b>1\).
Proof
The partial derivatives of F are given by
Thus, \((z_1,z_2) \in \textrm{Sing}(M)\) if and only if the following equations are satisfied:
If \(a, b > 1\), it follows that \(\textrm{Sing}(M)=\{z_1=0\} \cup \{z_2=0\}\). If \(a=1\) or \(b=1\), we will have \(\textrm{Sing}(M)=\{z_2=0\}\) or \(\textrm{Sing}(M)=\{z_1=0\}\), respectively. Finally, if \(a=b=1\), we obtain \(\textrm{Sing}(M)=\{0\}\). \(\square \)
Let’s restate Theorem B for completeness.
Theorem B
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G(z)=(z_1^a,z_2^b)\) and \(X(z)=(z_2^b,z_1^a)\) with a, b positive integers. Then \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is a Levi-flat hypersurface at \(0\in {\mathbb {C}}^2\).
Proof
In this case, F is given by \(F(z)=z_1^a \overline{z}_2^b + \overline{z}_1^a z_2^b\) and its complexification will be given by
Therefore, we have on \(M_{{\mathbb {C}}}^*\)
Hence
From (18), we get
which implies that \(\alpha \wedge d\alpha \wedge \beta \) is equal to
that is,
Again, from (18), we have \(z_1^aw_2^b=-w_1^az_2^b\), which yields \(z_1^{2a}w_2^{2b}=w_1^{2a}z_2^{2b}\) in \(M_{{\mathbb {C}}}\). Thus, M is Levi-flat. \(\square \)
Now, we consider
In Proposition 4.1, we saw that \(\textrm{Sing}(M)=\{z_1=0\} \cup \{z_2=0\}\) in the case where \(a_1=b_1>1\) and \(a_2=b_2>1\). Now, let’s verify that this also occurs in the case where \(a_1>b_1>1\) and \(a_2>b_2>1\) (even without the assumption \(a_1b_2=a_2b_1\)).
Proposition 4.2
Let M be the hypersurface described in Eq. (19). Then, we have the following options for \(\textrm{Sing}(M)\):
-
1.
\(\textrm{Sing}(M)=\{0\}\), if \(b_1=b_2=1\);
-
2.
\(\textrm{Sing}(M)=\{z_1=0\}\), if \(b_2=1\) and \(b_1>1\);
-
3.
\(\textrm{Sing}(M)=\{z_2=0\}\), if \(b_1=1\) and \(b_2>1\);
-
4.
\(\textrm{Sing}(M)=\{z_1=0\} \cup \{z_2=0\}\) if \(b_1,b_2>1\).
Proof
We observe that F is given by
and its partial derivatives are given by
Therefore, \((z_1,z_2) \in M\) belongs to the singular set if it satisfies the equations
Let’s assume that \(b_1, b_2 > 1\). Then we see that \(\{z_1=0\} \cup \{z_2=0\} \subset \textrm{Sing}(M)\). Now, suppose by contradiction that \((z_1,z_2) \in \textrm{Sing}(M)\) with \(z_1 \ne 0\) and \(z_2 \ne 0\). From Eq. (20), we obtain
and from Eq. (23), we obtain
which implies
In other words, \(a_1a_2=b_1b_2\). This is absurd, since \(a_1>b_1\) and \(a_2>b_2\) by assumption. We conclude that \({\textsf {Sing}}(M)=\{z_1=0\} \cup \{z_2=0\}\), if \(b_1,b_2 >1\). Now let’s assume \(b_1=1\) and \(b_2>1\). Then, the equations for the singular set are given by
Clearly, \(\{z_2=0\} \subset \textrm{Sing}(M)\). Now suppose \(z_2\ne 0\), then we necessarily have \(z_1\ne 0\) by Eq. (24). Thus, it follows from Eqs. (24) and (27):
which implies \(a_1=\frac{b_2}{a_2}\), that is, \(a_1a_2=b_2\). However, this contradicts \(a_2>b_2\) and \(a_1>1\). Therefore, \({\textsf {Sing}}(M)=\{z_2=0\}\) if \(b_1=1\) and \(b_2>1\). In a similar way we obtain \({\textsf {Sing}}(M)=\{z_1=0\}\) if \(b_2=1\) and \(b_1>1\), and we also obtain \({\textsf {Sing}}(M)=\{ 0\}\) if \(b_1=b_2=1\). \(\square \)
To prove Theorem C, we will use the following lemma
Lemma 4.3
Let \(M_{{\mathbb {C}}}\) be the complexification of
we have \(\alpha \wedge d\alpha \wedge \beta = h\ dz_1 \wedge dw_1 \wedge dz_2 \wedge dw_2\), where
and \(c=a_1b_2=a_2b_1\).
Proof
The complexification of F is given by
Therefore, we have
We will use the notations
so that we can write:
With these notations, we get
In \(M_{{\mathbb {C}}}\), we have
which implies on \(M_{{\mathbb {C}}}^*\)
Taking \(c=a_1b_2=b_1a_2\), we have
and it follows that
Therefore, on \(M_{{\mathbb {C}}}^*\), we obtain
thus
Now, we observe
which implies
Similarly, we obtain
so that \(z_1w_2\alpha _1 \beta _2-w_1z_2\alpha _2\beta _1\) is equal to
Again, from Eq. (28), on \(M_{{\mathbb {C}}}^*\) we have:
that is,
which implies that \(z_1w_2\alpha _1 \beta _2-w_1z_2\alpha _2\beta _1\) is equal to
So, \(z_1w_2\alpha _1 \beta _2-w_1z_2\alpha _2\beta _1\) is equal to
Finally, we conclude
and the lemma follows by substituting the above expression and the expression for \(g_2\) into Eq. (29). \(\square \)
Now, we prove Theorem C.
Theorem C
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G=(z_1^{a_1},z_2^{a_2})\) and \(X=(z_2^{b_2},z_1^{b_1})\), where \(a_1\ge b_1\) and \(a_2 \ge b_2\) are positive integers satisfying \(a_1b_2=a_2b_1\). Then \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is Levi-flat if, and only if, \(a_1=b_1\) and \(a_2=b_2\).
Proof
From Lemma 4.3, we have \(\alpha (z,w)\wedge \beta (z,w)\wedge d\alpha (z,w)=0\) for all \((z,w) \in M_{{\mathbb {C}}}\) if, and only if, \(a_1a_2+b_1b_2-2c=0\), i.e., \(2c=a_1a_2+b_1b_2\). Using the fact \(c=a_1b_2=b_1a_2\), it follows that
Therefore, \(M_{{\mathbb {C}}}\) is Levi-flat if, and only if, \(a_1=b_1\) and \(a_2=b_2\). This concludes the proof of Theorem C. \(\square \)
Remark 4.1
From the above result, we conclude that for \(n=2\), the hypersurface M will not Levi-flat precisely when the map \(\psi _{G,X}\) satisfies the Milnor condition at the origin.
5 Proof of Theorem D
Now, we study the case with permutations in higher dimensions. Consider \(\sigma \in S_n\) a permutation on the set \(\underline{n}:=\{1,\ldots ,n\}\), and let’s use the notation \(\sigma _k:=\sigma (k)\). We employ a transversality argument, along with the dimension 2 case (see Theorem C), to obtain the following result:
Theorem D
Let G and X be elements in \(\chi ({\mathbb {C}}^ n,0)\), \(n\ge 2\), of the form \(G=(z_1^{a_1},\ldots , z_n^{a_n})\) and \(X=(z_{\sigma _1}^{b_{\sigma _1}},\ldots , z_{\sigma _n}^{b_{\sigma _n}})\), where \(a_k\ge b_k\) are positive integers. Let us assume that for some \(\ell \in \underline{n}\), the integers \(a_{\ell }, b_{\ell }, a_{\sigma _{\ell }}, b_{\sigma _{\ell }}\) satisfy the following conditions: \(a_{\ell }>b_{\ell }\) and \(a_{\ell }b_{\sigma _{\ell }}=b_{\ell }a_{\sigma _{\ell }}\). Then, \(M=\{F(z)=2{\textrm{Re}}(\psi _{G,X}(z))=0\}\), i.e.,
is nondegenerate at \(0\in {\mathbb {C}}^n\).
Proof
Let’s assume by contradiction that M is Levi-flat. Therefore, the regular part \(M^*\) is foliated by complex submanifolds of (complex) dimension \(n-1\). Thus, we can choose \(i: {\mathbb {C}}^2 \hookrightarrow {\mathbb {C}}^n\) to be a transversal embedding to M (see Cerveau and Lins-Neto 2011, Corollary 3.3), so that the regular part of \(i^{-1}(M)\) is also foliated by complex submanifolds of dimension 1 (Riemann surfaces). Without loss of generality, we can assume that \(\sigma (1) = 2\) and that \(a_1, b_1, a_2, b_2\) are the integers satisfying the conditions in the statement. Making a change of coordinates and using the fact that the Levi-flat property is invariant under biholomorphisms (Proposition 2.4), we can assume that the embedding \(i:{\mathbb {C}}^2\rightarrow {\mathbb {C}}^n\) is given by \(i: (z_1, z_2) \mapsto (z_1, z_2, 0,\ldots ,0).\) We observe
But by Theorem C, \(i^{-1}(M)\) does not have the regular part foliated by complex submanifolds. Therefore, M is not Levi-flat. \(\square \)
6 Examples
In this section, we will explore examples where the fields G and X do not satisfy the Milnor condition. In these examples, our hypersurfaces are all Levi-flat.
Example 6.1
Given the Pham–Brieskorn polynomial \(f:{\mathbb {C}}^2\rightarrow {\mathbb {C}}\) defined by \(f(z)=z_1^p+z_2^q\), where \(p,q>2\). Consider the holomorphic vector field
whose solutions represent the fibers of f. Also, take the vector field \(X=(a_1,a_2)\). Thus, the Hermitian product of G and X is given by
Clearly, \(\Psi _{G,X}\) does not satisfy the Milnor fibration condition at the origin. Now we take
Since \(\Psi _{G,X}\) is a holomorphic function, then M is Levi-flat.
Example 6.2
Let \(G=(z_1,iz_2,(-1-i)z_3)\) and \(X=(z_1,z_2,z_3)\). According to Seade (1997, Theorem 1), \(\psi _{G,X}\) does not satisfy the Milnor fibration condition at the origin. The Hermitian product of G and X is given by
Consider \(M=\{F(z)=2{\textrm{Re}}\ \langle G(z),X(z) \rangle =z_1\overline{z}_1-z_3\overline{z}_3=0\}\). Let’s us verify that M is Levi-flat. First, we note that \(F(z)=z_1\overline{z}_1-z_3\overline{z}_3\), and the partial derivatives of F are given by
Then, \({\textsf {Sing}}(M)=\{(0,z_2,0) \in {\mathbb {C}}^3:z_2\in {\mathbb {C}}\}\). Clearly, \(\dim _{{\mathbb {R}}}{\textsf {Sing}}(M)=2\). Moreover,
We have \(\partial \overline{\partial }F= dz_1\wedge d\overline{z}_1-dz_3\wedge d\overline{z}_3,\) and
that yields \(\partial F \wedge \overline{\partial } F \wedge \partial \overline{\partial } F=(z_3\overline{z}_3-z_1\overline{z}_1)dz_1\wedge d\overline{z}_1\wedge dz_3\wedge d\overline{z}_3.\) Thus, \(\partial F(p) \wedge \overline{\partial }F (p)\wedge \partial \overline{\partial }F(p)=0\) for all \(p \in M\) and therefore, M is Levi-flat.
Example 6.3
Let \(G(z)=(z_1,z_2)\) and \(X(z)=(-iz_2,iz_1)\). Then
is clearly Levi-flat whose singular set is \({\textsf {Sing}}(M)=\{0\}\).
In general, when \(G(z)=(z_1,z_2)\) is the radial vector field, \(X=(\lambda _1z_1, \lambda _2z_2)\), and if \({\textrm{Re}}(\lambda _k)\ne 0\) for \(k=1,2\), then the singular set of \(M=\{2{\textrm{Re}}\ \langle G(z),X(z)\rangle =0\}\) is just the origin \(0 \in {\mathbb {C}}^2\), and M is Levi-flat. However, in this case, the map \(\psi _{G,X}\) also does not satisfy the Milnor fibration condition at the origin, see Seade (1996, Example 3.4).
References
Brunella, M.: Singular Levi-flat hypersurfaces and codimension one foliations. Ann. Sc. Norm. Super. Pisa Cl. Sci. (5) VI(4), 661–672 (2007)
Burns, D., Gong, X.: Singular Levi-flat real analytic hypersurfaces. Am. J. Math. (2003). https://doi.org/10.1353/ajm.1999.0002
Cartan, E.: Sur la géométrie pseudo-conforme des hypersurfaces de l’espace de deux variables complexes. Ann. Mat. Pura Appl. 11(1), 17–90 (1933)
Cerveau, D., Lins-Neto, A.: Local Levi-flat hypersurfaces invariants by a codimension one holomorphic foliation. Am. J. Math. 133(3), 677–716 (2011). https://doi.org/10.1353/ajm.2011.0018
Da Silva, A.A., Fernández-Pérez, A.: On real-analytic Levi-flat hypersurfaces and holomorphic webs. Expo. Math. 41(4), 18 (2023) (Paper No. 125510)
Fernández-Pérez, A.: On Levi-flat hypersurfaces with generic real singular set. J. Geom. Anal. 23, 2020 (2013). https://doi.org/10.1007/s12220-012-9317-1
Milnor, J.: Singular points of complex hypersurfaces. In: Singular Points of Complex Hypersurfaces, Ann. of Math Studies, No. 61, Princeton University Press, Princeton, NJ; University of Tokyo Press, Tokyo, 1968, iii+122 pp
Ruas, M.A.S., Seade, J., Verjovsky, A.: On real singularities with a Milnor fibration. In: Trends in Singularities. Trends Math., pp. 191–213. Birkhäuser, Basel (2002). ISBN:3-7643-6704-0
Seade, J.: Fibred links and a construction of real singularities via complex geometry. Bol. Soc. Brasil. Mat. (N.S.) 27(2), 199–215 (1996)
Seade, J.: Open book decompositions associated to holomorphic vector fields. Bol. Soc. Mat. Mex. 3, 323–336 (1997)
Shafikov, R., Sukhov, A.: Germs of singular Levi-flat hypersurfaces and holomorphic foliations. Comment. Math. Helv. 90, 479–502 (2015)
Acknowledgements
Ayane A. Da Silva acknowledges support by a Ph.D grant from Capes-Brazil, and the results of this paper are part of her Ph.D thesis. Fernández-Pérez acknowledges support from CNPq Projeto Universal 408687/2023-1 “Geometria das Equações Diferenciais Algébricas” and CNPq Projeto PQ 2023 306011/2023-9. Finally, we also would like to thank the referee for suggestions and pointing out corrections.
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The second author acknowledges support from CNPq Projeto Universal 408687/2023-1 “Geometria das Equações Diferenciais Algébricas” and CNPq Projeto PQ 2023 306011/2023-9.
About this article
Cite this article
da Silva, A.A., Fernández-Pérez, A. On Real Analytic Levi-Flat Hypersurfaces Associated with Milnor Fibrations. Bull Braz Math Soc, New Series 55, 30 (2024). https://doi.org/10.1007/s00574-024-00406-7
Received:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00574-024-00406-7