8

Let $n\ge 3$ be odd and $\mathbb{R}P^n$ be the $n$-dimensional real projective space. I'm studying the degree of maps and I am looking for an example of a degree $1$ map $f:\mathbb{R}P^n \to M$ for a closed, orientable $n$-dimensional manifold $M$. Such maps will preserve fundamental classes. I have been unable to find such examples for $M \ne \mathbb{R}P^n$ and $M \ne S^n$.

I know that any map $f:\mathbb{R}P^n\to S^{n-1}\times S^1$ has degree $0$ (see here). I also know that the inclusion $f: \mathbb{R}P^n \to \mathbb{R}P^n\#\mathbb{R}P^n$ does not have degree $1$. Can someone please suggest me an appropriate (type of) examples of degree $1$ maps from $\mathbb{R}P^n$? More general ideas on constructing such maps $f$ will also be appreciated. Thank you in advance.

  • 2
    Any closed orientable manifold of dimension $n$ admits a degree 1 map to $S^n$ (collapse the complement of a coordinate ball), for a start. – Ben Steffan Jul 29 '24 at 21:12
  • But $\mathbb RP^n$ isn't orientable – Lieven Jul 29 '24 at 21:13
  • 2
    @Lieven $\mathbb{R}\mathrm{P}^n$ is orientable if $n$ is odd. Now read the first sentence of the question again. – Ben Steffan Jul 29 '24 at 21:14
  • Ok, I should learn how to read thanks – Lieven Jul 29 '24 at 21:14
  • @BenSteffan I know that fact but I don't understand how a degree $1$ map $g:M \to S^n$ helps me to get a degree $1$ map $f:\mathbb{R}P^n \to M$? Can you please clarify. – DavidChi Jul 29 '24 at 21:17
  • @DavidChi Take $M = \mathbb{R}\mathrm{P}^n$. – Ben Steffan Jul 29 '24 at 21:19
  • Well yeah, I missed that. Are there any more interesting examples, where $M \ne S^n$? Let me edit my question. – DavidChi Jul 29 '24 at 21:20
  • 2
    @DavidChi I don't quite know, but the fact that a degree 1 map $\mathbb{R}\mathrm{P}^n \to M$ gives rise to a degree 2 map $S^n \to M$ (just precompose with the quotient projection) already enforces the condition that $H_k(M)$ be entirely 2-torsion for $0 < k < n$. Using Poincare duality one can probably strengthen this requirement further to rule out more cases, but explicitly constructing degree 1 maps is likely rather difficult. – Ben Steffan Jul 29 '24 at 22:09
  • 1
    Given @BenSteffan's comment, you may find this useful. – Michael Albanese Jul 29 '24 at 22:28
  • @BenSteffan thank you giving the perspective of a degree $2$ mapping $S^n$ that rules out many cases. – DavidChi Jul 29 '24 at 23:26
  • 1
    @MichaelAlbanese thank you sharing your post. From the attached post, I conclude, in particular, that if there exists a degree $1$ map $f:\mathbb{R}P^n \to M$, then $M$ and its universal cover $\widetilde{M}$ are rational homology spheres. Kindly correct me if I'm wrong. – DavidChi Jul 29 '24 at 23:29
  • 1
    Now, it would be interesting to find a sufficient condition (in particular an example) for a degree $1$ map $f:\mathbb{R}P^n\to M$ to exist! We may note that even if both $M$ and $\widetilde{M}$ are rational homology spheres, then while there is a non-zero degree map $g:S^n \to M$ (see the above post), it does NOT promise a degree $1$ map $f:\mathbb{R}P^n\to M$. – DavidChi Jul 29 '24 at 23:37
  • 2
    @DavidChi Furthermore, since a degree 1 map induces an onto homomorphism at $\pi_1$ level, $\pi_1(M)$ is either $0$ or $\mathbb{Z}_2$. Maybe this rules out more possibilities when combined with the requirement that both $M$ and $\tilde{M}$ are $\mathbb{Q}$-homology spheres. – Dmitry K. Jul 29 '24 at 23:50

1 Answers1

7

This is a very interesting question, and (remarkably) we can decide it entirely using only relatively basic tools from the algebraic topologist's tool case. To be precise, we're going to prove the following:

Proposition: A map $\newcommand{\RP}{\mathbb{R}\mathrm{P}}f\colon \RP^n \to M$ of degree 1 where $n$ is odd and $M$ a closed orientable $n$-manifold exists if and only if $M$ is either $S^n$ or homotopy equivalent to $\RP^n$.

To prove this, first note that since the quotient/covering map $S^n \to \RP^n$ is of degree 2, any such $f$ gives rise to a degree 2 map $S^n \to M$ just by precomposition, so by a standard application of Poincaré duality all the homology groups $H_k(M)$ for $0 < k < n$ must be 2-torsion (see this question for more details). If $H_k(M) = 0$ for all $0 < k < n$, then $M$ is an integral homology sphere, but then $M$ must already be simply connected (see KFJ2611's comment about $\pi_1(M)$ either being 0 or $\mathbb{Z} / 2$) and therefore homeomorphic to $S^n$ (by Whitehead's theorem applied to a degree 1 map $M \to S^n$ and the generalised Poincaré conjecture¹).

Otherwise, we will show that $f_*\colon H_*(\RP^n) \to H_*(M)$ is already an isomorphism.

To this end, note first of all that $f_*\colon H_*(\RP^n) \to H_*(M)$ is surjective since $f$ is of degree 1 (this is a consequence of the projection formula $\alpha \frown f_*(b) = f_*(f^*(\alpha) \frown b)$ for all $\alpha \in H^*(M)$, $b \in H_*(\RP^n)$ applied to $b$ the fundamental class of $\RP^n$ and Poincaré duality, though let me know if you want more details). Thus, $H_k(M)$ is 0 if $k$ is even and either $\newcommand{Z}{\mathbb{Z}}\newcommand{\F}{\mathbb{F}} \Z / 2$ or 0 if $k$ is odd in the range $0 < k < n$. Passing to co/homology with coefficients in $\F_2$, we obtain a map from a graded $\F_2$-algebra $H^*(M; \F_2)$ which is levelwise of dimension at most 1 to $H^*(\RP^n; \F_2) \cong \F_2[x]$ with $x$ the unique class in degree 1.

Let now $0 < k < n$ be the lowest degree in which $H^*(M; \F_2) \neq 0$. If $k = 1$, then since $f^*(y_1^k) = f^*(y_1)^k = x^k$ for $y_1 \in H^1(M; \F_2)$ the generator we see that $f^*$ is an isomorphism, hence $f_*$ is an isomorphism both with $\F_2$ and integral coefficients and we are done². Otherwise, we note that by the universal coefficient theorem we must have $H^{k + 1}(M; \F_2) \neq 0$ as well, so by Poincaré duality $H^{n - k}(M; \F_2)$ and $H^{n - k - 1}(M; \F_2)$, too, must be non-zero, whence $y_k y_{n - k - 1}$ is a nonzero element in degree $n - 1$ (again since $f^*(y_k y_{n - k - 1}) = f^*(y_k) f^*(y_{n - k - 1}) = x^k x^{n - k - 1} = x^{n - 1} \neq 0$) and another round of Poincaré duality then yields that $H^1(M; \F_2) \neq 0$, contradicting our choice of $k$. $\Box$


¹I know I promised "only relatively basic tools from the algebraic topologist's tool case." This is the one "advanced" theorem I need.

²As was pointed out by ShamanR in the comments, it does not directly follow from this that $M$ is homotopy equivalent to $\RP^n$. To conclude, we need the following version of the Whitehead/Hurewicz theorem:

Theorem: Let $g\colon X \to Y$ be a map between CW-complexes. If $g$ induces an isomorphism on $\pi_1$ and moreover a lift $\tilde{g}\colon \tilde{X} \to \tilde{Y}$ of $g$ to the universal covers of $X$ and $Y$ induces an isomorphism on integral homology, then $g$ is a homotopy equivalence.

Proof: By Whitehead's theorem $\tilde{g}_*$ induces an isomorphism on homotopy groups since $\tilde{X}$ and $\tilde{Y}$ are simply connected, and since covering projections induce isomorphisms on $\pi_k$ for all $k > 1$ we see that $g$ must induce an isomorphism on all homotopy groups. $\Box$

Derived Cats was so kind as to fill in the rest of the argument:

Let $\tilde{M}$ be the universal cover of $M$. Since $\pi_1(M) \cong \Z / 2$ it is a double cover, so a lift $\tilde{f}\colon S^n \to \tilde{M}$ of $f$ to universal covers must be a degree 1 map (here we use that $\tilde{M}$ is itself a closed orientable $n$-manifold). By the same argument as in the first paragraph of the proof, this means that $\tilde{M}$ is a simply connected integral homology sphere and therefore homeomorphic to $S^n$. Thus $\tilde{f}_*\colon H_*(S^n) \to H_*(\tilde{M})$ is an isomorphism and we conclude by the theorem after noting that it is applicable since closed manifolds have the homotopy type of a CW-complex.

Ben Steffan
  • 8,325
  • 4
    Nice answer. If I understand correctly, you have shown in the second part that $M$ has the same homology as $\mathbb{R}P^n$. However, you have claimed in the statement of the proposition that $M$ is homotopy equivalent to $S^n$ or $\mathbb{R}P^n$. Here, isomorphic homology with $\mathbb{R}P^n$ does not quite imply a homotopy equivalence with it, right? – Milind Jul 30 '24 at 01:13
  • @ShamanR Ah, that is a very silly mistake. You're right, of course; I let myself be lulled in by the fact that $f$ induces an iso. on homology and on $\pi_1$. Thank you for pointing that out. I'll think about whether this can be remedied; let me put a notice to that effect in the answer in the meantime. – Ben Steffan Jul 30 '24 at 01:58
  • 1
    @Ben Steffan Your argument shows that $\pi_1 M = \mathbb{Z}_2$, so the universal cover $\tilde{M} \to M$ is a double cover. The map $f$ therefore lifts to a degree $1$ map on the universal covers $F:S^n \to \tilde{M}$. By your previous arguments, $F$ is a homotopy equivalence. By Poincaré $\tilde{M} \cong S^n$, so $M$ is the quotient of $S^n$ by a free $\mathbb{Z}_2$ action. Now it's tempting to say that $M$ has to be homotopy equivalent to $\mathbb{R}P^n$, but I'm not sure how to justify this. – Derived Cats Jul 30 '24 at 02:06
  • 1
    @DerivedCats This should be enough, thank you. There is a version of the Whitehead theorem that reads "If $f\colon X \to Y$ induces an iso. on $\pi_1$ and a lift of $f$ to the universal covers induces an isomorphism on $H_*({{-}})$, then $f$ is a (weak) equivalence." – Ben Steffan Jul 30 '24 at 02:12
  • 2
    @BenSteffan thanks for this perfect answer. Thanks also to Michael Albanese, KFJ2611, ShamanR and Derived Cats for participating in this discussion. – DavidChi Jul 30 '24 at 03:39