Han's Formula: Combinatorics & Info Theory

• Han's Formula is a collection of closed-form identities leveraging algebraic generating functions and hook length techniques to solve enumeration and entropy problems.
• It provides efficient methods for rational function extraction via the Ω-operator and for enumerating trees and partitions through sophisticated hook length formulas.
• The formulas extend to derive fundamental entropy inequalities in information theory, revealing deep links between combinatorial structure and probabilistic measures.

Han's formula refers to several distinctive results in combinatorics, partition theory, and information theory, unified by their reliance on structural properties of combinatorial objects and algebraic generating functions. Notable instances include Han’s closed-form for certain partition generating functions, his hook length formulas for trees and partitions, as well as information inequalities in entropy. These formulas reveal deep algebraic, combinatorial, and information-theoretic phenomena with wide-ranging implications for partition analysis, enumeration, and the structure of entropy in Boolean analysis.

1. Rational Function Extraction and Han’s Formula

Han’s formula in the context of rational function extraction expresses a closed-form identity for extracting nonnegative powers of a Laurent series via MacMahon’s Ω-operator. Let

$$A(\lambda) = \prod_{i=1}^n (1 - x_i \lambda), \qquad B(t) = \prod_{j=1}^m (1 - y_j t)$$

with indeterminates $x_i, y_j$ such that $x_i \ne 1$ and $x_i y_j \ne 1$ for all $i,j$. For $U(\lambda)$ a Laurent polynomial of degree at most $n-1$,

$$\boxed{ \Omega_\ge \frac{U(\lambda)}{A(\lambda) B(1/\lambda)} = \sum_{i=1}^n \frac{ x_i^{n-1} U(1/x_i) }{ (1 - x_i) B(x_i) \prod_{j \ne i} (x_i - x_j) } } \tag{H}$$

where $\Omega_\ge$ extracts the sum of coefficients of $\lambda^k$ for $k \ge 0$ in the expansion. The formula provides an explicit and efficient route to partial fraction decomposition and coefficient extraction, enabling enumeration of symbolic generating functions governed by linear inequalities. This approach leverages the underlined-denominator rule: contributions are summed directly over the poles given by roots of linear factors $(1 - x_i \lambda)$, evaluated after suitable normalization, and is computationally more direct than recursive Gröbner-basis methods or iterative root-finding (Liu et al., 17 Dec 2025).

2. Combinatorial Hook Length Formulas

Han discovered a set of hook length formulas for enumerating trees and partitions, which are now central in algebraic and enumerative combinatorics.

(a) Binary and k-ary Tree Formulas

For binary trees $B_n$ of $n$ vertices, Han’s formulas are: $$\sum_{T \in B_n} \prod_{u \in T} (h_u)^{h_u-1} = \frac{1}{n!}$$

$$\sum_{T \in B_n} \prod_{u \in T} (2 h_u + 1)^{2 h_u - 1} = \frac{1}{(2n+1)!}$$

where $h_u$ is the hook length: the size of the subtree rooted at $u$. These identities generalize to k-ary trees via Yang: $$\sum_{T \in \mathcal{T}_{n,k}} \prod_{u \in T} (h_u)^{k h_u - 1} = \frac{1}{n!}$$ Bijective proofs rely on establishing correspondences between tree labelings and “staircase arrays,” where hook length exponents count the number of staircase labelings. The structural insight is that the total number of such labelings equals both sides, with the exponentiation reflecting lattice-theoretic properties of the tree (Chen et al., 2011).

(b) Partition Hook Generating Function

In integer partition theory, Han’s (also attributed to Nekrasov–Okounkov) identity connects products over the hook lengths in partition diagrams: $$\sum_{\lambda} q^{|\lambda|} \prod_{h \in H(\lambda)} \left(1 - \frac{b}{h^2}\right) = \prod_{k=1}^\infty (1 - q^k)^{b-1}$$ The coefficient of $q^n$ can be expanded as a polynomial in $b$, with intricate congruence properties modulo primes, including symmetric residue distributions proven via combinatorial arguments involving hook complements and binomial sums (Keith, 2011).

3. Applications in Partition Theory

Han’s formula provides a unified analytic toolkit for extracting enumerative information from multivariate generating functions subject to linear inequalities, a core task in partition analysis. The Ω-operator formulation connects to MacMahon’s partition analysis framework: for a system of inequalities, a generating function encodes constraints via exponent patterns, and the Ω-operator projects onto the admissible subspace. The formula (H) enables one-pass computation for a large class of problems, including those that previously required substantial symbolic computation, such as the $k$-gon partitions problem (Liu et al., 17 Dec 2025).

A corollary is the explanation of pattern congruences in $p_n(b)$, the coefficient polynomials in $b$ for the above partition hook formula. For $n=5k+4$, coefficients vanish for $t \le k$, while nonzero residues distribute cyclically in blocks for $t>k$. Analogous equidistribution results exist for other primes and prime powers, revealing a deeply structured modular behavior in partition-generated polynomials (Keith, 2011).

4. Han’s Inequalities in Information Theory

Han’s formula also refers to foundational entropy inequalities characterizing dependencies among random variables.

(a) Classical Han’s Inequality

For random variables $X_1, \dots, X_n$ (discrete or continuous): $$\sum_{i=1}^n H(X_{[n] \setminus \{i\}}) \ge (n-1) H(X_1, \dots, X_n)$$ This submodularity-based inequality (and its generalizations to $k$-sized subsets) is a cornerstone of information theory. The proof exploits the polymatroid structure of the Shannon entropy set function, and leads to strong corollaries such as Shearer’s lemma and various moment inequalities (Sason, 2022).

(b) Strengthened Fourier Entropy-Influence Inequality

In Boolean Fourier analysis, Han proved (initially for $\{-1,1\}$-valued $f$): $$H[\widehat{f}] \le C_1 I(f) + C_2 \sum_{i} I_i(f) \ln \frac{1}{I_i(f)}$$ where $H[\widehat{f}]$ is the Shannon entropy of the Fourier weight distribution, $I(f)$ is the total influence, and $I_i(f)$ are the coordinate influences. An information-theoretic proof yields sharp constants $C_1 = C_2 = 1$ for real-valued $f$ of unit $L^2$-norm, showing that the law of the Fourier spectrum is essentially governed by a Markov chain revealing coordinates sequentially, with entropy increments tightly controlled by influences. This result isolates a structural property of the entropy-influence relationship, distinct from earlier subcube-restriction/mixing arguments (Li et al., 2 Dec 2025).

5. Proof Methods and Structural Insights

The various Han formulas share a thematic reliance on:

• Rational function extraction and residue computations, often optimized using partial fraction decompositions or the underlined-denominator rule.
• Bijections between combinatorial objects (trees, array labelings, partitions) and analytic generating structures, with hook lengths encoding recursive or lattice-theoretic statistics.
• Probabilistic and information-theoretic interpretations, especially in the context of entropy inequalities, where Markov chain constructions and submodularity arguments yield generalized or sharp results.
• Linkage with submodular function theory and polymatroid theory, providing a unifying perspective for both combinatorial and information-theoretic variants of Han's formula.

The generality and efficiency of the Ω-operator-based approach yield new computational paradigms for classical enumeration problems (Liu et al., 17 Dec 2025).

6. Extensions, Congruences, and Open Questions

The general framework of Han’s hook-based formulas extends naturally to general $k$-ary trees, to partitions with various statistics, and to modular and arithmetic settings. In the setting of partition congruences, Han’s formula underlies new Ramanujan-type distribution patterns for the polynomial coefficients, extending to arbitrary primes and revealing hidden algebraic and combinatorial symmetry (Keith, 2011).

Open questions persist regarding the interpretation of polynomial coefficients, potential connections to colored/weighted partitions, further congruence families, and analytic continuations or refinements tracking additional statistics in the generating structure.

A plausible implication is that the Han/MacMahon operator formula may provide new insights across diverse areas including analytic combinatorics, algebraic geometry (via rational function specializations), and information theory, as further generalizations and computational strategies are developed.

7. Summary Table: Prominent Han’s Formulas

Domain	Structural Formulation	Reference
Partition analysis	$$\Omega_\ge \frac{U(\lambda)}{A(\lambda) B(1/\lambda)} = \sum ...$$	(Liu et al., 17 Dec 2025)
Trees (hook formula)	$\sum_T \prod_{u \in T} (h_u)^{k h_u -1} = 1/n!$	(Chen et al., 2011)
Partition hooks	$$\sum_\lambda q^{|\lambda|} \prod_{h \in H(\lambda)} (1-b/h^2) = ...$$	(Keith, 2011)
Entropy inequality	$\sum_i H(X_{[n]\setminus\{i\}}) \ge (n-1) H(\cdot)$	(Sason, 2022)
Fourier entropy-influence	$$H[\widehat{f}] \le I(f) + \sum_i I_i(f) \ln \frac{1}{I_i(f)}$$	(Li et al., 2 Dec 2025)

These results exemplify the breadth and algebraic power of Han’s formulas in combinatorics and related fields.