Lower bounds on the number of envy-free divisions
Duško Jojić, Gaiane Panina, Rade Živaljević
TL;DR
This work investigates lower bounds on the number of envy-free divisions in two cake-cutting models. It leverages secretive/expelled variants of envy-free results and a configuration-space/test-map framework to quantify multiplicity: a universal lower bound of at least two envy-free divisions in the classical Woodall-Stormquist setting, and a significantly larger bound $\binom{2p-1}{p-1} 2^{2-p}$ in a prime-$p$ measure-equipartition hybrid. The analysis identifies extremal preferences that realize exactly two divisions, showing the underlying incidence graph forms a single $2r$-cycle, and demonstrates how topological methods yield precise counting in the extended model. Together, these results illuminate how additional structure (secretive/expelled participants, measure constraints) enhances the multiplicity of envy-free divisions and showcases the power of topological combinatorics in fair-division problems.
Abstract
We analyze lower bounds for the number of envy-free divisions, in the classical Woodall-Stormquist setting and in a non-classical case, when envy-freeness is combined with the equipartition of a measure. 1. In the first scenario, there are $r$ hungry players, and the cake (that is, the segment $[0,1]$) is cut into $r$ pieces. Then there exist at least two different envy-free divisions. This bound is sharp: for each $r$, we present an example of preferences such that there are exactly two envy-free divisions. 2. In the second (hybrid) scenario, there are $p$ not necessarily hungry players ($p$ is a prime) and a continuous measure $μ$ on $[0,1]$. The cake is cut into $2p-1$ pieces, the pieces are allocated to $p$ boxes (with some restrictions) and the players choose the boxes. Then there exists at least $\binom{2p-1}{p-1} \cdot 2^{2-p}$ envy-free divisions such that the measure $μ$ is equidistributed among the players.
