Let $f\colon M\to N$ a smooth surjective map of compact oriented manifolds of the same dimension. Then there is a map $f_!\colon H_i(N)\to H_i(M)$ obtained from the induced map on cohomology combined with Poincaré duality. This map has several names. I have seen it called the transfer, umkehr or wrong-way map. And it is treated in several question on mathoverflow and math.stackexchange:

Pullback map in homology

Reference for push-pull formula in cohomology


I am interested in a geometric interpretation of this map in the following situation. Assume that there is an open subset $U\subset N$ such that, letting $V=f^{-1}(U)$, the restriction $f|_V\colon V\to U$ is an $r$-sheeted covering map ($r\in\mathbb{N}$) and $N'\subset U$ a compact oriented submanifold. If I understood correctly the answers and comments from the questions above, then we have


Is this correct? And is there a reference for this? Several books in the answers and comments to the questions above are mentioned but I have actually not found this statement.

I should also say that I want to use this statement in a paper (if true!) and neither me nor the audience of the paper are experts on algebraic topology. So I would very much appreciate a reference where this fact is explicitly stated or at least a statement which easily implies this fact.

Edit: One can construct $f_!$ also in the case when $M$ and $N$ are oriented manifolds with boundary and $f^{-1}(\partial N)=\partial M$ (and all other assumptions the same). Now the map map $f_!\colon H_i(N)\to H_i(M)$ is obtained from the induced map on cohomology relative to the boundaries combined with Lefschetz duality. Is the conclusion $f_!([N'])=[f^{-1}(N')]$ still true? We still assume $f|_V\colon V\to U$ being a $r$-sheeted covering, neither $U$ nor $V$ intersect the boundaries of $M$ and $N$, and $N'\subset U$ a compact oriented submanifold (without boundary).


1 Answer 1


Let $K\subset N$ be a compact oriented smooth submanifold of codimension $k$ in an oriented smooth manifold $N$, $T\subset N$ a tubular neighbourhood of $K$ and $\tau$ a $k$-form on $N$ supported on $T$ such that the restriction of $\tau$ to the fibre of $T$ over any point of $K$ (which is of course diffeomorphic to $\mathbb R^k$) has compact support and integrates to 1. Then the class of $\tau$ in $H^k_c(N)$ is Poincaré dual to $[K]$. Such a $\tau$ always exists; it is called a Thom form. This is discussed in Bott, Tu, Differential forms in algebraic topology, § 5--6.

If $M$ is an oriented smooth manifold, $f:M\to N$ a smooth map transversal to $K$ and $T$ a sufficiently small tubular neighbourhood of $K$, then $f^{-1}(T)$ is a tubular neighbourhood of $f^{-1}(K)$; one can arrange trivializations in such a way that $f$ maps any fibre of $f^{-1}(T)$ diffeomorphically to a fibre of $T$. So the pullback via $f$ of a Thom form for $K$ will be a Thom form for $f^{-1}(K)$. This proves the desired statement $f_!([K])=[f^{-1}(K)]$.

  • $\begingroup$ Hi igorf, thanks a lot! Does this by any chance also work when $M$ and $N$ are manifolds with boundaries and the $f^{-1}(\partial N)=\partial M$? $\endgroup$
    – Hans
    May 14, 2023 at 11:44
  • $\begingroup$ (see the edits for the precise setup/question) $\endgroup$
    – Hans
    May 14, 2023 at 11:51
  • 3
    $\begingroup$ Hi Hans! Let's remove the boundaries. This doesn't change homology, and the map $\overset\circ f:\overset\circ M\to\overset\circ N$ remains proper. Is your $f_!$ equal to the composition $H_*(N)\simeq H_*(\overset\circ N)\simeq H^*_c(\overset\circ N)\xrightarrow{f^*} H^*_c(\overset\circ M)\simeq H_*(\overset\circ M)\simeq H_*(M)$? If so, then the same reasoning works just as well. $\endgroup$
    – igorf
    May 14, 2023 at 12:55
  • $\begingroup$ Ah, right. Thanks! $\endgroup$
    – Hans
    May 14, 2023 at 15:48

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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