Nonexistence of immersions

To get back into the flow of posting, I’ll talk about a curiosity in differential topology that is most organically proven using the machinery of characteristic classes.

An between smooth manifolds is a smooth map whose differential on tangent spaces is injective on each fiber. The Whitney Immersion Theorem states that any smooth n dimensional manifold admits an immersion into (and an embedding into , which can be improved nontrivially to ). We may ask whether the dimension is optimal: that is, whether or not every smooth manifold of dimension embeds into for some . This turns out not to be the case if is not a power of , but using Stiefel-Whitney classes, we may show that the number is optimal if .

. For , real projective space does not immerse into .

. Suppose there existed an immersion . Then, is a morphism of vector bundles which covers and which factors through the pullback . Because is an immersion, its differential is fiberwise injective, and the same is true for the induced map . Since is a trivial vector bundle, it follows that , i.e.  is a subbundle of a trivial bundle of rank .

Now consider the normal bundle , defined as the orthogonal complement of in . By this definition, we have , implying that . Moreover, by the axioms for Stiefel-Whitney classes, we have . However, whose multiplicative inverse in the ring is . This contradicts the fact that

Comments

I had to follow along with Milnor-Stasheff since it’s been a while since I did algebraic topology, but really nice writeup! I like your other articles as well. Could you give an example of a manifold for which is not optimal?

Error! Probably caused by unsupported commands in the preamble. Please remove those that are not necessary for your main document and try again.

Thank you, and good question! For a low-hanging class of examples, $S^n$ embeds into $\mathbb{R}^{n+1}$ in the usual way, which beats the general Whitney bound for $n > 2$. I think (and I’ll try to find a reference) that every smooth, closed $3$-manifold immerses into $\mathbb{R}^4$, although augmenting “immersion” to “embedding” is far more subtle and unresolved except for special classes, e.g. for $3$-manifolds fibered over surfaces. I think there are also general results for $\mathbb{CP}^n$ immersing into $\mathbb{R}^{4n-4}$ for suitable constraints on $n$.

Add a comment

You must log in to post a comment.