Let $X$ be a smooth, projective variety and $Y \subset X$ be a hyperplane section (possibly singular) of $X$. Suppose that the dimension of $X$ is $n$. Is it true that for any $k<n-1$, the induced morphism of intersection homologies: $$IH_k(Y) \to IH_k(X)$$ is an isomorphism?

I would think this is true as the version of the Lefschetz hyperplane section theorem I have it states this conclusion under the additional assumption that $Y$ is transverse to every strata of some Whitney stratification of $X$. As $X$ is smooth, in my case, I can take the only non-trivial strata to be $X$ and the other sub-strata to be empty sets. This should answer the above question, but I am not sure if I am making a mistake.


So, the answer is positive and follows from the "usual" Weak Lefschetz Theorem if you have transversality (and it suffices to assume that $Y$ is smooth). On the other hand, if $Y$ is singular then you surely don't have transversality and probably don't have the isomorphism in question (as far as I remember, the intersection homology of $X$ is isomorphic to its "usual" cohomology, whereas for a singular $Y$ the group $IH_k(Y)$ is "usually smaller" than $H^k(Y)$).

  • $\begingroup$ I don't think this answers the question. I would've expected the answer to be "yes" and I don't really see an argument in your answer for why you think the isomorphism fails in general. $\endgroup$ Mar 29 '17 at 20:34
  • $\begingroup$ I am just saying that if you replace the group $H^k(Y)$ in the Lefschetz theorem by the group $IH_k(Y)$ that is not isomorphic to it then you would get a wrong statement. However, it may happen that $H^k(Y)$ is always isomorphic to $IH_k(Y)$ the range in question. Anyway, the (current) argument of user45397 is not correct. $\endgroup$ Mar 29 '17 at 20:59
  • $\begingroup$ Oh, I see. I thought the "usual" Lefschetz theorem had an assumption that $Y$ is smooth, but I see now that it's enough that the singular locus of $X$ is contained in $Y$. Yeah, then I agree with you. $\endgroup$ Mar 29 '17 at 21:03

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.