MathOverflow is a question and answer site for professional mathematicians. It's 100% free, no registration required.

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

Let $M$ be a closed connected oriented smooth manifold and $\mathrm{Diff}_{+}(M)$ the group of orientation preserving diffeomorphisms of $M$ endowed with the compact-open topology. Pick a base point $x_{0} \in M$ and consider the evaluation map $$\mathrm{ev}\colon \mathrm{Diff}_{+}(M) \to M, \quad \mathrm{ev}(g) = g(x_{0}).$$

I would like to know when the evaluation map admits a global section, i.e. a continuous map $s\colon M \to \mathrm{Diff}_{+}(M)$ such that $\mathrm{ev} \circ s = \mathrm{id}_{M}$.


  1. If $M = G$ is a Lie group, then a section $s$ exists: for $x\in G$ put $s(x) := \text{left multiplication by }x\cdot x_{0}^{-1}$.
  2. However, for $M = S^{2}$, the 2-sphere, such a section $s$ cannot exist: it would induce an injective map on homotopy groups $s_{*}\colon \pi_{k}(S^{2}) \to \pi_{k}(\mathrm{Diff}_{+}(S^{2}))$, but $\pi_{2}(S^2) \cong \mathbb{Z}$ while $$\pi_{2}(\mathrm{Diff}_{+}(S^{2})) \cong \pi_{2}(\mathrm{SO}(3)) = 0.$$

It is probably hard to decide on the existence of a section for general $M$, so let us restrict to spheres $S^{n}$:

Question: For which $n\in \mathbb{N}$ does the evaluation map $\mathrm{ev}\colon \mathrm{Diff}_{+}(S^{n}) \to S^{n}$ admit a continuous global section?

Since $S^1$ and $S^3$ are naturally Lie groups, the section exists in thsese two cases by the first example above. We can analogously use octonionic multiplication to construct a global section for $S^{7}$. Are there any other cases than $n=1,3,7$?

share|cite|improve this question
up vote 13 down vote accepted

If a section $\sigma : M \rightarrow \text{Diff}_{+}(M)$ to $\text{Diff}_{+}(M) \rightarrow M$ exists, then $M$ must be parallelizable (i.e. the tangent bundle of $M$ must be trivial). Indeed, if $\vec{e}_1,\ldots,\vec{e}_n$ is a basis for the tangent space of $M$ at $x_0$, then $\sigma(x)_{\ast}(e_1),\ldots,\sigma(x)_{\ast}(e_n)$ is a basis for the tangent bundle to $M$ at $x \in M$. These vary continuously with $x$, and thus define a trivialization of the tangent bundle of $M$.

For spheres, a theorem that was proved independently in

Kervaire, Michel A. Non-parallelizability of the n-sphere for n>7. Proc. Natl. Acad. Sci. USA 44 (1958), no. 3, 280–283.


Bott, R.; Milnor, J. On the parallelizability of the spheres. Bull. Amer. Math. Soc. 64 1958 87–89.

says that $S^k$ is only parallelizable when $k=1,3,7$, so your examples are the only ones.

share|cite|improve this answer

To add to Andy Putman's answer: An interesting modification of the question is to replace diffeomorphisms by homotopy equivalences. Thus, for a given based manifold or CW complex $(X,p)$, is there a continuous map $F:X\times X\to X$ such that $F(p,x)=x$ for all $x$ and $x\mapsto F(x,y)$ is a homotopy equivalence for all $y$? If $X$ is connected then it is enough if $x\mapsto F(x,p)$ is a homotopy equivalence, and without loss of generality that homotopy equivalence is the identity. So the question is, is there is a continuous multiplication on $X$ for which $p$ is a two-sided identity element.

The answer is no if $\pi_1(X)$ is not abelian, or if the canonical action of $\pi_1(X)$ on some $\pi_n(X)$ is nontrivial, or if some Whitehead product in $X$ is nontrivial. This rules out some parallelizable manifolds for the original question.

share|cite|improve this answer
Sorry, I could not stop myself from correcting the spelling of my last name :). – Andy Putman Apr 30 '14 at 23:34
Oh, I'm so sorry. I usually know how to spell your name. – Tom Goodwillie May 1 '14 at 0:07
This is very nice! – Oldřich Spáčil May 1 '14 at 12:55

Evaluation maps are studied in rational homotopy theory, see e.g. Evaluation maps in rational homotopy by ‎Félix and Lupton. Here is a sample result (see Corollary 1.9 of the above paper):

Let $\mathrm{Aut}_0(M)$ denote the identity component of $\mathrm{maps}(M,M)$. A section of $\mathrm{Diff}_0\to M$ followed by the inclusion into $\mathrm{Aut}_0(M)$ gives a section of the evaluation map $e: \mathrm{Aut}_0(M)\to M$, so the latter is surjective in rational homology. Suppose $M$ is simply-connected. If $e$ is surjective on rational homology, then $M$ is rationally homotopy equivalent to a product of odd-dimensional spheres.

share|cite|improve this answer

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.