next up previous
Next: 1.6 Harnack Curves Up: 1 Early Study of Previous: 1.4 Curves of Degree

1.5 The Classical Method of Constructing Nonsingular Plane Curves

All of the classical constructions of the topology of nonsingular plane curves are based on a single construction, which I will call classical small perturbation. Some special cases were given in the previous subsection. Here I will give a detailed description of the conditions under which it can be applied and the results.

We say that a real singular point $ \xi=(\xi_0:\xi_1:\xi_2)$ of the curve $ A$ is an intersection point of two real transversal branches, or, more briefly, a crossing,2 if the polynomial $ a$ defining the curve has matrix of second partial derivatives at the point $ (\xi_0,\xi_1,\xi_2)$ with both a positive and a negative eigenvalue, or, equivalently, if the point $ \xi$ is a nondegenerate critical point of index 1 of the functions $ \{x\in \mathbb{R}P^2\vert x_i\ne 0\}\to \mathbb{R}\:x\mapsto a(x)/x_i\deg a$ for $ i$ with $ \xi_i\ne 0$. By Morse lemma (see, e.g. [Mil-69]) in a neighborhood of such a point the curve looks like a union of two real lines. Conversely, if $ \mathbb{R}A_1,\dotsc,\mathbb{R}A_k$ are nonsingular mutually transverse curves no three of which pass through the same point, then all of the singular points of the union $ \mathbb{R}
A_1\cup\cdots\cup \mathbb{R}A_k$ (this is precisely the pairwise intersection points) are crossings.

1.5.A Classical Small Perturbation Theorem (see Figure 5)   Let $ A$ be a plane curve of degree $ m$ all of whose singular points are crossings, and let $ B$ be a plane curve of degree $ m$ which does not pass through the singular points of $ A$. Let $ U$ be a regular neighborhood of the curve $ \mathbb{R}A$ in $ \mathbb{R}P^2$, represented as the union of a neighborhood $ U_0$ of the set of singular points of $ A$ and a tubular neighborhood $ U_1$ of the submanifold $ \mathbb{R}A\smallsetminus U_0$ in $ \mathbb{R}P^2\smallsetminus U_0$.

Then there exists a nonsingular plane curve $ X$ of degree $ m$ such that:

(1) $ \mathbb{R}X\subset U$.

(2) For each component $ V$ of $ U_0$ there exists a homeomorphism $ h\:V\to D^1\times D^1$ such that $ h(\mathbb{R}A\cap V)=D^1\times 0\cup
0\times D^1$ and $ h(\mathbb{R}X\cap V)=\{(x,y)\in D^1\times D^1\vert xy=1/2\}$.

(3) $ \mathbb{R}X\smallsetminus U_0$ is a section of the tubular fibration $ U_1\to \mathbb{R}A\smallsetminus U_0$.

(4) $ \mathbb{R}X\subset\{(x_0:x_1:x_2)\in\mathbb{R}P^2\vert a(x_0,x_1,x_2)b
(x_0,x_1,x_2)\le 0\}$, where $ a$ and $ b$ are polynomials defining the curves $ A$ and $ B$.

(5) $ \mathbb{R}X\cap \mathbb{R}A=\mathbb{R}X\cap \mathbb{R}B=\mathbb{R}A\cap \mathbb{R}B$.

(6) If $ p\in \mathbb{R}A\cap \mathbb{R}B$ is a nonsingular point of $ B$ and $ \mathbb{R}B$ is transversal to $ \mathbb{R}A$ at this point, then $ \mathbb{R}X$ is also transversal to $ \mathbb{R}A$ at the point.

There exists $ \varepsilon>0$ such that for any $ t\in (0,\varepsilon]$ the curve given by the polynomial $ a+tb$ satisfies all of the above requirements imposed on $ X$.

Figure 5:
\begin{figure}\centerline{\epsffile{f5s.eps}}\end{figure}

It follows from (1)-(3) that for fixed $ A$ the isotopy type of the curve $ \mathbb{R}X$ depends on which of two possible ways it behaves in a neighborhood of each of the crossings of the curve $ A$, and this is determined by condition (4). Thus, conditions (1)-(4) characterize the isotopy type of the curve $ \mathbb{R}X$. Conditions (4)-(6) characterize its position relative to $ \mathbb{R}A$.

We say that the curves defined by the polynomials $ a+tb$ with $ t\in (0,\varepsilon]$ are obtained by small perturbations of $ A$ directed to the curve $ B$. It should be noted that the curves $ A$ and $ B$ do not determine the isotopy type of the perturbed curves: since both of the polynomials $ b$ and $ -b$ determine the curve $ B$, it follows that the polynomials $ a-tb$ with small $ t>0$ also give small perturbations of $ A$ directed to $ B$. But these curves are not isotopic to the curves given by $ a+tb$ (at least not in $ U)$, if the curve $ A$ actually has singularities.

Proof. Proof of Theorem 1.5.A We set $ x_t=a+tb$. It is clear that for any $ t\ne 0$ the curve $ X_t$ given by the polynomial $ x_t$ satisfies conditions (5) and (6), and if $ t>0$ it satisfies (4). For small $ \vert t\vert$ we obviously have $ \mathbb{R}
X_t\subset U$. Furthermore, if $ \vert t\vert$ is small, the curve $ \mathbb{R}X_t$ is nonsingular at the points of intersection $ \mathbb{R}X_t\cap \mathbb{R}B=\mathbb{R}A\cap \mathbb{R}B$, since the gradient of $ x_t$ differs very little from the gradient of $ a$ when $ \vert t\vert$ is small, and the latter gradient is nonzero on $ \mathbb{R}A\cap \mathbb{R}B$ (this is because, by assumption, $ B$ does not pass through the singular points of $ A)$. Outside $ \mathbb{R}B$ the curve $ \mathbb{R}X_t$ is a level curve of the function $ a/b$. On $ \mathbb{R}A\smallsetminus \mathbb{R}B$ this level curve has critical points only at the singular points of $ \mathbb{R}A$, and these critical points are nondegenerate. Hence, for small $ t$ the behavior of $ \mathbb{R}X_t$ outside $ \mathbb{R}B$ is described by the implicit function theorem and Morse Lemma (see, for example, [Mil-69]); in particular, for small $ t\ne 0$ this curve is nonsingular and satisfies conditions (2) and (3). Consequently, there exists $ \varepsilon>0$ such that for any $ t\in (0,\varepsilon]$ the curve $ \mathbb{R}X_t$ is nonsingular and satisfies (1)-(6). $ \qedsymbol$


next up previous
Next: 1.6 Harnack Curves Up: 1 Early Study of Previous: 1.4 Curves of Degree
Oleg Viro 2000-12-30