Suppose we have a Riemannian manifold $(M^m, g_M)$ and a smooth manifold $N^n$ (with dimensions $m > n$). And suppose we have a smooth surjective submersion $F: M^m \to N^n$.

Question: Is there a natural way to construct a Riemannian metric $g_N$ on $N^n$ such that $F$ can actually be "upgraded" to a Riemannian submersion?


In most cases, there is no Riemannian metric on $N$ which makes $F$ a Riemannian submersion. However, if there is such a metric, it is unique. To shed more light on this matter, let us try to construct the desired metric.

Let $p\in N$, and let $q\in F^{-1}(p)$. Let $H_q$ denote the orthogonal complement of $\ker dF_q$ in $T_qM$. Then $dF_q$ restricts to an isomorphism of vector spaces $$dF_q|_{H_q}:H_q\xrightarrow{\sim}T_pN,$$ and there is a unique inner product on $T_pN$ which makes the above isomorphism an isometry. This explains the uniqueness part in the first paragraph. Now, if we repeat this argument for some $q'\in F^{-1}(p)$ other than $q$, there is no reason to expect we get the same inner product on $T_pN$, and this is why the desired Riemannian metric does not exist usually.

  • 1
    $\begingroup$ (+1) It may be worth mentioning that there's an important but very special case where everything does work out: a $G$-invariant Riemannian metric on the total space $P$ of a principal $G$-bundle $P \to B$ canonically induces a unique Riemannian metric on the base space $B$ that makes the quotient map $P \to B$ a Riemannian submersion. $\endgroup$ – Branimir Ćaćić Dec 22 '17 at 23:37
  • $\begingroup$ Clear explanation. Thanks! $\endgroup$ – user32416 Dec 23 '17 at 2:37
  • $\begingroup$ @AmitaiYuval, thanks for good explanation. I am aware that it is not appropriate to ask another question in comments, but I want to know about your last sentence by a concrete example. $\endgroup$ – C.F.G Aug 15 at 18:49

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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