Minimal Equitable Partitions

• Minimal equitable partitions are defined as the division of a structured set into the fewest possible parts that are balanced in size and interaction according to specific criteria.
• They are applied in continuous measures, matroid theory, and graph theory, optimizing functionals such as local density and neighbor distributions.
• Efficient algorithms and sharp structural bounds guarantee constructibility, supporting applications in fair division and spectral graph analysis.

A minimal equitable partition is a partitioning of a structured set—typically a measurable space, matroid ground set, or graph vertex set—into a prescribed minimal number of classes or parts such that each part is "equitable" with respect to size, combinatorial structure, or interaction with all possible subsets according to specified criteria. This concept provides a unifying analytic and combinatorial framework for equidistribution problems, fairness constraints, and extremal constructions across continuous, discrete, and algebraic settings.

1. Definitions and Foundational Examples

For a matroid $M=(E,\mathcal I)$, a partition of the ground set $E$ into $k$ bases $B_1,\dots,B_k$ is called a $k$-equitable partition if for any subset $S\subseteq E$,

$$\max_{i\in [k]}|B_i \cap S| - \min_{i\in [k]}|B_i \cap S| \le 1.$$

A minimal equitable partition is a $k(M)$-equitable partition into the fewest possible number $k(M)$ of bases such that the equitability condition is fulfilled, i.e., $$k(M) = \min\{k\geq 1: E \text{ is the disjoint union of } k \text{ bases}\}$$ (Akrami et al., 16 Jul 2025).

In the continuous setting, as seen in the context of measurable partitions of intervals or circles, minimal equitable partitions correspond to decompositions into sets of equal Lebesgue measure minimizing a functional counting "small difference" pairs: $E$0 where $E$1 denotes 2D Lebesgue measure. For a partition $E$2 of an interval $E$3 of length $E$4, the minimal equitable partition minimizes $E$5 (Antoniuk et al., 2024).

In algebraic combinatorics, specifically within graph theory, minimal equitable partitions frequently refer to partitions of vertex sets into two or more classes ("cells") so that each vertex of a class has a prescribed number of neighbors in each class. For Hamming graphs, the minimal equitable 2-partition structure is sharply determined by eigenvalue constraints and quotient matrix parameters (Mogilnykh et al., 2019).

2. Measure-Theoretic Minimal Equitable Partitions

The measure-theoretic formulation, central in (Antoniuk et al., 2024), investigates partitions of $E$6 (or the circle $E$7) into $E$8 Lebesgue-measurable sets of equal measure that minimize the sum

$E$9

where $k$0 captures the "local density" of pairs at distance $k$1. The fundamental result asserts

$k$2

with $k$3 being optimal. Equality is realized when each part has measure $k$4 and the sets are arranged to induce the minimum possible overlap in the sense of the $k$5 functional.

The extremal construction prescribes partitioning the circle into blocks of length $k$6 and distributing these cyclically to achieve equitability and optimality. Passing from the circle to the interval setup introduces an additive error of at most $k$7. The structure showcases a deep connection between analytic and discrete combinatorial equitability, establishing sharp bounds for both continuous and large-scale discrete coloring problems (Antoniuk et al., 2024).

3. Matroidal Minimal Equitable Partitions

In the setting of finite matroids, a minimal equitable partition corresponds to splitting the ground set into the smallest number $k$8 of bases so that, for every subset $k$9, the sizes $B_1,\dots,B_k$0 are as nearly equal as possible: $B_1,\dots,B_k$1 The principal result in (Akrami et al., 16 Jul 2025) is that such a partition exists for every $B_1,\dots,B_k$2 whenever the ground set may be partitioned into $B_1,\dots,B_k$3 (disjoint) bases. The constructive algorithm uses a series of base exchanges—controlled by the exchange graph and chordless directed cycles—guaranteed via the base exchange and strong circuit-exchange properties of matroids, to reach equitability for all subsets in polynomial time.

The extension to equitable partitioning of two disjoint subsets $B_1,\dots,B_k$4 makes it possible to control the discrepancy for each set individually (within 1 and 2, respectively). These results have sharpness examples in graphic matroids, showing the bounds are, in general, the best possible.

4. Minimal Equitable Partitions in Hamming Graphs

Equitable 2-partitions of Hamming graphs $B_1,\dots,B_k$5 with nontrivial eigenvalue $B_1,\dots,B_k$6 exhibit a highly structured taxonomy (Mogilnykh et al., 2019):

1. Most such partitions are "reducible," arising by lifting a minimal equitable 2-partition from a lower-dimensional Hamming graph (notably $B_1,\dots,B_k$7).
2. Two exceptional irreducible constructions persist: permutation-switchings of lifted $B_1,\dots,B_k$8-partitions and alphabet-lifting of paired 8-cycles in $B_1,\dots,B_k$9.

Equitable partitions in this context are classified by their quotient matrix, ensuring regularity and prescribed neighbor distribution across parts. This structure is deeply intertwined with the graph's spectral parameters, with Lloyd's theorem confirming correspondence between quotient matrix eigenvalues and the graph's spectrum.

5. Algorithmic Constructibility and Complexity

Across the surveyed domains, minimal equitable partitions are guaranteed to exist under mild structural assumptions and are constructible via efficient algorithms. In matroids, all necessary rebalancing operations can be performed in polynomial time, relying on independence oracles and digraph algorithms for strongly connected components and cycles (Akrami et al., 16 Jul 2025). In measure-theoretic partitions, the explicit construction follows block subdivision and cyclic assignment, with the possibility of approaching optimality arbitrarily closely via regularization procedures (Antoniuk et al., 2024).

For Hamming graphs, the reduction to lower-dimensional representatives and the explicit construction for exceptional cases are fully explicit and algorithmic, governed by the combinatorics of coordinate projections and alphabet decompositions (Mogilnykh et al., 2019).

6. Connections, Applications, and Discrete-Continuous Analogies

Minimal equitable partitions facilitate the analysis of minimization problems for functionals measuring small differences, local densities, or "fairness" across multiple domains. The discrete-to-continuous correspondence is particularly prominent: continuous analytic partition bounds yield asymptotically sharp results for classic combinatorial coloring problems on long intervals and finite sets. Applied consequences extend to matroid-constrained fair division, where minimal equitable partitions enable the design of envy-free up to one item allocations for identical tri-valued additive valuations and guarantee maximin share fairness for bi-valued additive cases (Akrami et al., 16 Jul 2025).

In spectral graph theory, the study of minimal equitable partitions in Cartesian products and Hamming graphs reveals deep structural implications for eigenfunction localization and quotient matrix theory, informing both extremal combinatorics and algebraic graph theory (Mogilnykh et al., 2019).

---

Key References:

• "Sets and partitions minimising small differences" (Antoniuk et al., 2024)
• "Matroids are Equitable" (Akrami et al., 16 Jul 2025)
• "Equitable 2-partitions of the Hamming graphs with the second eigenvalue" (Mogilnykh et al., 2019)