3

Suppose $f:\mathbb{R}^n\setminus\{0\}\to \mathbb{R}^m\setminus\{0\}$ is a linear map inducing a map $g:\mathbb{R}P^{n-1}\to \mathbb{R}P^{m-1}$ and hence a homomorphism $$ g^*:\mathbb{Z}/2\mathbb{Z}[x]/(x^m)\to \mathbb{Z}/2\mathbb{Z}[y]/(y^n) $$ on cohomology rings with $\mathbb{Z}/2\mathbb{Z}$-coefficients. Since this is a graded ring map, we have $x\mapsto \lambda y$.

Why does linearity imply that $\lambda=1$? Is there an argument without going into the calculation of the cohomology of the projective space?

user398460
  • 55

1 Answers1

3

First of all notice that $g: \mathbb{R}P^n \to \mathbb{R}P^m$ is well defined provided that $f$ is injective. Consequently $m=n+k$ for some $k \geq 0$.

Since the cohomology ring $H^*(\mathbb{R}P^n)=\mathbb{Z}/2\mathbb{Z}[x]/(x^{n+1})$ is generated by the Poincaré dual of (the homology class of) an hyperplane, we have that $$g^*(x)= g^*PD[\text{hyperplane}]=PD[g^{-1}(\text{generic hyperplane})]=y,$$ where the last identity is because the image of $g$ intersects the generic hyperplane in a $n-1$ dimensional subspace.

Another possibility is to prove that $g$ is isotopic to the inclusion $\mathbb{R}P^n \subset \mathbb{R}P^m$.

Antonio Alfieri
  • 1,555
  • 7
  • 13