3
$\begingroup$

Let $f:X\to Y$ be a morphism of (nice) topological spaces. consider the basechange along $g:Y'\to Y$. We write $F:X'\to Y'$ and $G:X'\to X$. Consider some class of sheaves on those spaces, e.g. abelian sheaves or sheaves of $k$-vector spaces for a fixed field. We can take this to the unbounded derived categories. Then Spaltenstein showed that there is a natural morphism $Lg^* \circ Rf_!\Rightarrow RF_! \circ LG^*$ and that it is an isomorphism.

However $g$ (and hence $G$) is required to be flat, hence $g^*$ (resp. $G^*$) is exact and we basically only have to right derive, instead of dealing with a left and a right derived functor at the same time.

First question: Is there a way to do this without the flatness condition? If not can somebody provide a counterexample?

Next I would like to bring this together with the formula \begin{equation} Rf_!(-\otimes^L Lf^*-)\cong (Rf_! - )\otimes^L - \end{equation}

To bring them into a single "framework" consider a multimorphism $g:Y'\to Y_1,\ldots Y_n$ consisting of $g_i:Y'\to Y_i$ and define the pullback as $$ g^*(-,\ldots, -):=\bigotimes g_i^*- $$ We get the formula above as base change of

$\require{AMScd}$ \begin{CD} X @>f>> Y\\ @V f,1 V V @VV 1,1 V\\ Y, X@>>1,f> Y,Y \end{CD}

Now I want to prove basechange in the general case, where we have morphisms spaces $X,Y_1,\ldots, Y_n, Y'$ and morphisms $f:X\to Y_j$ for fixed $j$ and $g_i:Y'\to Y_i$ for each $i$ in a cartesian square

\begin{CD} X\times_{Y_j} Y' @>F>> Y'\\ @V G V V @VV g V\\ Y_1,\ldots, X, \ldots, Y_n @>>1,f,1> Y_1,\ldots, Y_j,\ldots, Y_n \end{CD}

where $F=\text{pr}_{Y'}$ and $G_i=\begin{cases} g_i\circ \text{pr}_{Y'} & i\neq j\\ \text{pr}_X & i=j\end{cases}$

and preferably I want to do this without using the other two "base changes", i.e. in one go.

Since $g$ contains a tensor product, $g^*$ will not be exact, even if $g$ is flat.

Second question: How would one construct a natural transformation $Rf_!\circ Lg^*\to LG^*\circ RF_!$ in this generalized case?

$\endgroup$
1
  • $\begingroup$ For nicer base change results, one should use Lurie's enlargement of the bounded derived category instead of Spaltenstein's. (Spaltenstein's is a localization of Lurie's and they agree when there's some finiteness hypothesis on your space.) I believe the base change result holds unconditionally there, and you can put (derived!) in if you like. Try HTT 7.3 (especially 7.3.1.19) to get started- although he's mainly interested in sheaves of spaces. $\endgroup$ Sep 18, 2017 at 12:42

0

Your Answer

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