I'm attempting to calculate the de Rham cohomologies of the $n$-torus: $n \choose k$.

I'd like to use a Mayer-Vietoris sequence relating $H^kT^{n}$ to $H^kT^{n-1}$ and $H^{k-1}T^{n-1}$ so I can use the identity ${n \choose k} = {n-1 \choose k} + {n-1 \choose {k-1}}$ to achieve the result by induction. But I am having trouble choosing an open cover for the sequence. The obvious choice would be to split the torus along one of its circles, i.e. $T^n=T^{n-1} \times S^1=U \cup V$ with $U=T^{n-1} \times (1/4,3/4)$ and $V=T^{n-1} \times [0,1/2)\cup(1/2,1]$ where we are identifying the point $0$ and $1$ to make $[0,1]\cong S^1$.

Here's my beef: By retraction $H^k U \cong H^k V \cong H^k T^{n-1}$ and we'd have $$H^k(U \cap V)\cong H^k (T^{n-1}\sqcup T^{n-1})\cong H^k T^{n-1} \oplus H^k T^{n-1} \cong H^kU \oplus H^kV$$ so that the $T^{n-1}$ terms cancel each other out of the sequence and you just end up with $\sum_{k=0}^n (-1)^k \dim H^k T^n=0$ which is true but only helpful if I cared about the Euler characteristic. Am I mistaken? Is there a better cover for the sequence?


Your covering is fine. Let's take a closer look at the maps defining the Mayer-Vietoris sequence: $$H^{k-1}(T^{n-1})\oplus H^{k-1}(T^{n-1})\xrightarrow{f}H^{k-1}(T^{n-1}\sqcup T^{n-1})\xrightarrow{\partial}H^k(T^n)\xrightarrow{g}H^k(T^{n-1})\oplus H^k(T^{n-1})$$

Although $f$ is a map between two isomorphic spaces, it is not an isomorphism, so the terms do not cancel. Let $([\alpha],[\beta])\in H^{k-1}(T^{n-1})\oplus H^{k-1}(T^{n-1})$. Then we have


I'm being a bit sloppy here with notation, using $\alpha$ instead of $i^*[\alpha]$. Similarly, for $[\alpha]\in H^k(T^n)$, we have $$g([\alpha])=([\alpha],[\alpha])$$ again with shorthand notation.

Finally, for any $[\alpha]\in H^{k-1}(T^{n-1}\sqcup T^{n-1})$, we may write $[\alpha]=[\beta]+[\gamma]$ reflecting the direct sum decomposition into disjoint components. Then $$\partial([\alpha])=[d\beta]=-[d\gamma]$$ where $d$ is the exterior derivative.

Now that we've had a closer look at the maps in the sequence, see if you can use the exactness of Mayer-Vietoris to show that there is a short exact seqence: $$0\rightarrow H^{k-1}(T^{n-1})\rightarrow H^{k}(T^n)\rightarrow H^k(T^{n-1})\to 0$$

From which your result will follow.


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.