Bounded core partitions and Borel--Weil--Bott
Fern Gossow, Andy Huchala
TL;DR
This work recasts the nonvanishing of cohomology groups $H^i(\mathrm{Gr}(k,n),\Omega^j(t))$ as a combinatorial problem about $t$-core partitions fitting in a $ (n-k)\times k$ rectangle, via the Borel--Weil--Bott theorem. It introduces knijt-partitions and a coarsening method to sharpen existence and vanishing criteria, and proves an elementary Nakano vanishing result for $\Omega^j(t)$ when $t>0$. The paper then develops refined bounds using the $t$-boundary, and completely analyzes the extremal cases $i+j=N$ and $N-1$, including a full description in terms of uniform blocks. In the special case $t=3$, it derives explicit arithmetic criteria tying nonvanishing to $3$-core partitions, including a concrete square-determinant condition and explicit parametrizations, thereby providing concrete, testable cohomology vanishing ranges. The results deepen the bridge between Grassmannian cohomology and the combinatorics of $t$-core partitions and lay groundwork for further connections to Griffiths-type residue constructions on Grassmannians.
Abstract
As a consequence of the Borel--Weil--Bott theorem, the nonvanishing of Hodge numbers of line bundles on a Grassmannian can be reinterpreted in terms of the existence of $t$-core partitions inside a rectangle. We sharpen known conditions for the nonvanishing of Hodge numbers of line bundles on a Grassmannian and give an elementary proof of the Nakano vanishing theorem in this setting. Using the theory of $t$-core partitions, we determine when $H^i(\mathrm{Gr}(k,n),Ω^j(t))$ vanishes in the cases $t=3$ and $i+j\geq k(n-k)-1$.
