Abstract: In Duke Math. J. 69, No.2, 335-347 (1993) the author translates a conjecture of Le Dung Trang on the non-existence of injective analytic maps f:(C^n,0)->(C^{n+1},0) with df(0)=0 into the non-existence of a hypersurface germ in (C^{n+1},0) with rather unexpected properties. However, the proof given there contains an apparently fatal error, as we demonstrate with an example.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is whether the proof of **Le Dung Trang's conjecture** is valid. Specifically, the paper discusses whether there exists an **injective analytic map** $ f: (\mathbb{C}^n, 0) \to (\mathbb{C}^{n + 1}, 0) $ whose differential $ df(0) $ has a rank less than $ n - 1 $. In particular, when $ n = 2 $, Le Dung Trang conjectured that such a mapping does not exist.
### Core problems of the paper
The main objective of the paper is to analyze the proof proposed in [Ném93] and point out the key errors therein. Specifically:
1. **Background problem**: [Ném93] transformed Le Dung Trang's conjecture into an equivalent problem: whether there exists a hypersurface germ with "unexpected properties". If such a hypersurface germ does not exist, the conjecture holds.
2. **Error analysis**: The authors Thomas Keilen and David Mond pointed out that there is an unfixable fundamental error in the proof in [Ném93]. They illustrated this problem by constructing a counterexample.
---
### Review of formulas and definitions
#### Definition: Good Subgerm
A two - dimensional subgerm $ (X, 0) \subset (\mathbb{C}^3, 0) $ is called "good" if there exist coordinates $ (w_1, w_2, w_3) $ and a mapping germ $ F: (\mathbb{C}^3, 0) \to (\mathbb{C}, 0) $ such that:
- $ X = F^{-1}(0) $;
- $ W_0 = X\cap \{w_1 = 0\} $ is an isolated plane curve singularity;
- $ \frac{\partial F}{\partial w_1}\notin \langle w_1,\frac{\partial F}{\partial w_2},\frac{\partial F}{\partial w_3}\rangle $.
#### Core theorem (the questioned conclusion)
[Ném93] proposed the following theorem:
> If the image $ (X, 0) $ of an injective analytic mapping germ $ f: (\mathbb{C}^2, 0) \to (\mathbb{C}^3, 0) $ is "good", then the rank of $ df(0) $ is at least 1, and $ (X, 0) $ is an equisingular family of plane curve singularities.
---
### Counterexample and error analysis
Keilen and Mond constructed a specific counterexample to illustrate the problems in the proof of [Ném93]. The following is the specific content of the counterexample:
#### Counterexample mapping
Consider the mapping $ f: (\mathbb{C}^2, 0) \to (\mathbb{C}^3, 0) $ defined as:
$$
f(x,y)=(y^3 + x^2,x,y^2).
$$
Obviously, $ f $ is injective.
#### Related geometric objects
1. $ V_0 = f_1^{-1}(0)=\{y^3 + x^2 = 0\} $ is a cusp, so it is not smooth.
2. $ W_0 = X\cap \{w_1 = 0\}=\{w_2^4 - w_3^3 = 0\} $ is an $ E_6 $ singularity.
Although $ f $ is injective, the Milnor numbers of $ V_0 $ and $ W_0 $ are not equal, which contradicts the conclusion in [Ném93].
---
### Source of the error
The proof in [Ném93] assumes that the vanishing cycles in the Milnor fiber can be related in a certain way.