8

Let $X$ be a topological space. If we know that for $\mathbb{F}=\mathbb{Q}$ and $\mathbb{Z}/p$, for any prime $p$, $$ H^*(X;\mathbb{F})=0$$ for any $*\geq n+1$, can we conclude that $$ H^*(X;\mathbb{Z})=0$$ for any $*\geq n+1$? If cannot what can we conclude?

Eric Wofsey
  • 330,363
Shiquan
  • 8,529
  • 2
  • 27
  • 84
  • 2
    The title question is stronger than the body question. The answer to the title question is no: see http://math.stackexchange.com/questions/1129606/homology-over-fields/1129640#1129640 for details. – Qiaochu Yuan Mar 05 '15 at 05:22
  • Crossposted: http://mathoverflow.net/questions/199112/integral-cohomology-derived-from-cohomology-with-coefficients-in-fields – Qiaochu Yuan Mar 05 '15 at 05:23

2 Answers2

5

If you're working with spaces having f.g. cohomology, the result in the post (not the title) holds. All $\text{Ext}(H_n(X, \mathbb{Z}), \mathbb{Z}_p)$ vanish by the universal coefficient theorem, so $H_n(X, \mathbb{Z})$ contains no torsion. It follows from the universal coefficient theorem for homology that $H_{n+1}(X, \mathbb{Z})$ must vanish, and so $H^{n+1}(X, \mathbb{Z})$ must too.

anomaly
  • 25,364
  • 5
  • 44
  • 85
3

Almost but not quite. If $H^*(X;F)=0$ for $*>n$ and all prime fields $F$, then by the universal coefficient theorem, $\operatorname{Hom}(H_*(X;\mathbb{Z}),F)=\operatorname{Ext}(H_*(X;\mathbb{Z}),F)=0$ for $*>n$. But if $\operatorname{Hom}(A,\mathbb{Q})=0$ then $A$ must be torsion, and any injection $\mathbb{Z}/p\to A$ induces a surjection $\operatorname{Ext}(A,\mathbb{Z}/p)\to\operatorname{Ext}(\mathbb{Z}/p,\mathbb{Z}/p)\neq0$.

It follows that we must have $H_*(X;\mathbb{Z})=0$ for all $*>n$, and so $H^*(X;\mathbb{Z})=0$ for $*>n+1$ and $H^{n+1}(X;\mathbb{Z})=\operatorname{Ext}(H_n(X;\mathbb{Z}),\mathbb{Z})$. However, this latter group could be nontrivial: for instance, $H_n(X;\mathbb{Z})$ could be $\mathbb{Q}$, in which case $H^{n+1}(X;\mathbb{Z})=\operatorname{Ext}(\mathbb{Q},\mathbb{Z})$ is uncountable.

Eric Wofsey
  • 330,363