Exact holomorphic differentials on a quotient of the Ree curve (Q405836)

For a nonsingular projective curve \(C\) over a field \(K,\) \(H^{0}(C,\Omega^{1})\) is the \(K\)-vector space of holomorphic differentials on \(C\) over \(K.\) Such a differential is said to be exact if it is of the form \(df\) for some function \(f\in K(C)\). If \(f\) is non-constant then \(f\) necessarily has at least one pole. So in characteristic \(0\), there are no non-zero exact holomorphic differentials. But in characteristic \(p > 0\), where \(p\)-th powers differentiate to \(0,\) a pole whose order is a multiple of \(p\) might disappear upon differentiation, and there can be non-zero exact holomorphic differentials. Inside \(H^{0}(C,\Omega^{1}),\) the subspace of exact holomorphic differentials is the kernel of the Cartier operator \(C\). It seems like a natural problem, given a curve \(C/K\) with \(\mathrm{char}(K) = p > 0\), to calculate this subspace. The authors produce several families of exact holomorphic differentials on a quotient \(X\) of the Ree curve in characteristic 3, defined by \(X : y^{q} - y = x^{q}_{0}(x^{q}- x)/F_{q}\) (where \(q_{0}= 3^{s}, s > 1\) and \(q = 3 q^{2}_{0})\). They conjecture that they span the whole space of exact holomorphic differentials, and prove this in the cases \(s = 1\) and \(s = 2,\) by calculating the kernel of the Cartier operator.