Let $C(n):X^n+Y^n=Z^n$ be the plane projective Fermat curve of degree $n$ over $\mathbb{C}$.

Shorter version of the question: How can I describe explicit representatives for a basis for the singular homology $H_1(C(n),\mathbb{Q})$?

Longer version of the question: Consider the following paths on $C(n)$:

$$ \gamma_{r,s} : [0,1]\to C(n), \quad t \mapsto [\zeta^r(1-t)^{1/n}:\zeta^st^{1/n}:1],$$

where $\zeta$ is a primitive $n-$th root of unity and $r,s \in \mathbb{Z}/n\mathbb{Z}$. One can combine them to obtain cycles

$$ \Delta_{r,s} := [\gamma_{0,s}-\gamma_{r,s}+\gamma_{r,0}-\gamma_{-r,0} + \gamma_{-r,-s}-\gamma_{0,-s}] \in H_1(C(n),\mathbb{Q})^-$$

that are anti-invariant under complex conjugation (that's what the $^-$ stands for), meaning that $\overline{\Delta_{r,s}}=-\Delta_{r,s}$. My aim is to show that the $\Delta_{r,s}$ span $H_1(C(n),\mathbb{Q})^-$, which has dimension equal to the genus


There are in principle enough elements. It's true that $\Delta_{r,0}=\Delta_{0,s}=0$ and that $\Delta_{-r,-s}=-\Delta_{r,s}$ and that $\Delta_{\frac{n}{2},\frac{n}{2}}=0$ for $n$ even, but one could in principle still find up to

$$\left\lfloor \frac{(n-1)^2}{2} \right\rfloor \ge g$$

cycles among the $\Delta_{r,s}$, and based on computer calculations I have reason to believe that they actually span $H_1(C(n),\mathbb{Q})^-$. But I would like to prove it. So I either need a direct argument or, if I can find a basis of $H_1(C(n),\mathbb{Q})$, I could try to express the $\Delta_{r,s}$ in terms of this basis and see what comes out.

The problem is that I do not have a nice way to visualize paths on the curve and describe them with explicit formulas.


  • Of course one can take the standard basis on the compact Riemann surface with $g$ holes, but then to go from there to the concrete case of $C(n)$ probably requires complicated formulas (I honestly would have no idea how to do it).
  • Find a basis of $H_1(C(n),\mathbb{Q})$ by induction. The cases with $n=1,2$ have trivial homology, so we just need the step. But it's difficult to do it without an idea for an explicit formula.
  • Look at the Jacobian $J(n)$. It has the same homology as $C(n)$. It can be contructed as $$ J(n) = \frac{\left(\Omega^{1}_{C(n)}\right)^*}{\Lambda},$$ i.e., as the dual of global holomorphic differential $1-$forms on $C(n)$ quotiented by the span $\Lambda$ of the functionals on $\Omega^{1}_{C(n)}$ of the form $\lambda = \int_{[c]} \cdot$ for some $[c] \in H_1(C(n),\mathbb{Q})$. $\Omega^{1}_{C(n)}$ is a complex vector space of dimension $g$, so if one finds a basis one can view $J(n)$ as $\mathbb{C}^g$ quotient a lattice, and cycles might be easier to describe there. And maybe, if one finds a basis of the homology of $J(n)$, its pullback via the Abel-Jacobi map $C(n) \to J(n)$ to $C(n)$ is again a basis and one can do something. But I am not sure about many points here.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.