For a compact 4-manifold, no 2-torsion in $H_1(M;\Bbb Z)$ implies no 2-torsion in $H_n(M;\Bbb Z)$ for all $n$

algebraic-topologyhomological-algebrahomology-cohomologymanifoldspoincare-duality

Let $M$ be a topological compact connected oriented 4-manifold with nonempty boundary, and suppose that each boundary component of $M$ is a rational homology 3-sphere. Is it true that if $H_1(M;\Bbb Z)$ has no 2-torsion then $H_n(M;\Bbb Z)$ has no 2-torsion for all $n$?

We only need to check $n=0,2,3,4$. Clearly $H_0(M;\Bbb Z)=\Bbb Z$. Also, since $\partial M\neq \emptyset$, $H_0(M,\partial M;\Bbb Z)=0$, so $H_4(M;\Bbb Z)=H^0(M,\partial M;\Bbb Z)=\text{Hom}(H_0(M,\partial M;\Bbb Z),\Bbb Z)=0$ and $H_3(M;\Bbb Z)=H^1(M,\partial M;\Bbb Z)=\text{Hom}(H_1(M,\partial M;\Bbb Z),\Bbb Z)$ has no 2-torsion. (Here I have used Poincare duality and the universal coefficient theorem for cohomology.)

Finally, $H_2(M;\Bbb Z)= H^2(M,\partial M;\Bbb Z)=\text{Hom}(H_2(M,\partial M;\Bbb Z),\Bbb Z)\oplus \text{Ext}(H_1(M,\partial M;\Bbb Z),\Bbb Z)$ has 2-torsion iff $H_1(M,\partial M;\Bbb Z)$ has 2-torsion. Can $H_1(M,\partial M;\Bbb Z)$ have 2-torsion even if $H_1(M;\Bbb Z)$ does not have 2-torsion?

P.S. The argument given above did not use that $\partial M$ consists of rational homology 3-spheres. This condition is possibly unnecessary.

Best Answer

Consider the long exact sequence of cohomology. $$\cdots\to H^1(\partial M;\Bbb Z)\to H^2(M,\partial M;\Bbb Z)\overset{d^\ast}{\to} H^2(M;\Bbb Z)\to H^2(\partial M;\Bbb Z)\to\cdots$$ We know that each boundary component is a rational homology $3$-sphere, so $0=H_1(\partial M;\Bbb Q)\cong H_1(\partial M;\Bbb Z)\otimes\Bbb Q$. Therefore, $H_1(\partial M;\Bbb Z)$ consists of torsion only, so $H^1(\partial M;\Bbb Z)=0$ by UCT and the coboundary map $d^*:H^2(M,\partial M)\to H^2(M;\Bbb Z)$ is injective by exactness. Now, UCT implies that $H^2(M;\Bbb Z)\cong \operatorname{Hom}(H_2(M;\Bbb Z),\Bbb Z)\oplus \operatorname{Ext}(H_1(M;\Bbb Z),\Bbb Z)$, which contains no $2$-torsion by assumption. Hence, $H_2(M;\Bbb Z)\cong H^2(M;\partial M;\Bbb Z)$ has no $2$-torsion as it's embedded as a subgroup of $H^2(M;\Bbb Z)$ by $d^\ast$.

[See edit] [The argument only requires $H^1(\partial M;\Bbb Z)=0$, so the condition that the boundary consists of rational homology $3$-spheres is unnecessary.]


Edit:

Actually, saying $\partial M$ is the disjoint union of rational homology $3$-spheres is equivalent to saying $H^1(\partial M;\Bbb Z)=0$. One direction is already demonstrated in the argument above. For the other direction, note that $\partial M$ is a closed, orientable $3$-manifold. If $H^1(\partial M;\Bbb Z)=0$, then $H_1(\partial M;\Bbb Z)$ consists of torsion elements only. By Poincare duality (applied to each component of $\partial M$), we see that $H^2(\partial M;\Bbb Z)\cong H_1(\partial M;\Bbb Z)$ also consists of torsion elements. Apply UCT, we see that $H_2(\partial M;\Bbb Z)$ cannot have free factors, i.e., also consists of torsion elements. Tensor product with $\Bbb Q$ kills all torsion elements, so each connected component is a rational homology $3$-sphere.

Related Question