Bivariate Exponential Generating Function

• Bivariate Exponential Generating Function is a two-variable series that encodes doubly-indexed arrays with factorial denominators.
• It organizes combinatorial structures such as Hermite polynomials, exponential sequences of arrangements, and special number sequences like Stirling and Bell numbers.
• The EGF framework facilitates operational, integral, and differential formulations, leading to explicit recurrence relations and asymptotic results.

A bivariate exponential generating function (EGF) is a generating function of two variables that enumerates a doubly-indexed array of quantities, encoding them as coefficients multiplied by monomials, each divided by the corresponding factorials. These EGFs arise in a wide range of mathematical, combinatorial, and analytic contexts, serving as organizing tools for two-parameter families of numbers, polynomials, or structures. Prominent examples include generating functions for bivariate Hermite polynomials, enumeration of faces in exponential sequences of arrangements, edge-bicolored graph counts, as well as families of special numbers such as Bernoulli, Stirling, and Eulerian numbers.

1. Formal Definition and Typical Structure

A general bivariate EGF encodes a family $\{a_{m,n}\}_{m,n \in \mathbb{N}}$ as

$$F(x, y) = \sum_{m, n \geq 0} a_{m, n} \frac{x^m}{m!} \frac{y^n}{n!}$$

where $x, y$ are formal variables (often complex), and the division by factorials distinguishes this exponential form from ordinary generating functions.

The domain of convergence for such series may be $\mathbb{C}^2$ (entire), or restricted by functional constraints in analytic or combinatorial applications. In concrete mathematical structures, $a_{m,n}$ may enumerate objects with two natural parameters (e.g., degree and level, or degree and dimension) or, in algebraic contexts, arise as the coefficients in expansions of doubly-indexed polynomials.

2. Bivariate Exponential Generating Functions in Polynomial Theory

Complex Hermite and Poly-Analytic Hermite Polynomials

For the bivariate poly-analytic Hermite polynomials $H_{m,n}(z, \bar{z})$, the canonical bivariate EGF is given by

$$E(z, \bar{z}; t, s) = \sum_{m,n=0}^\infty H_{m,n}(z, \bar{z}) \frac{t^m}{m!} \frac{s^n}{n!} = \exp(-t s + z t + \bar{z} s)$$

This EGF encodes all $H_{m,n}$ simultaneously and enables derivation of their operational, integral, and differential representations (Ghanmi et al., 2019). Equivalent operational (Rodrigues-type) and integral (Fourier–Wigner transform, Ismail representation) formulas are directly tied to this EGF.

Polynomial Arrays and Sheffer Sequences

For Sheffer or Riordan-type structures, the bivariate EGF can encode entire triangles of combinatorial numbers or sequences of polynomials: $$E_S(t, z) = \sum_{d=0}^{\infty} G_d(z) \frac{t^d}{d!}, \quad G_d(z) = \sum_{n} S(n+d, n) z^n$$ with closed expressions via Lagrange inversion: $$E_S(t, z) = H(x(t, z)) - H(0), \quad x - t f(x) = z$$ where $f,g$ derive from the Sheffer characterization (Lang, 2017).

3. Enumeration of Combinatorial Structures

Faces in Exponential Sequences of Arrangements (ESAs)

The function $F_\ell(x, y)$ captures the enumeration of faces by dimension and "level" in exponential sequences of hyperplane arrangements: $$F_\ell(x, y) = \sum_{n \geq d \geq \ell} f_{d, \ell}(A_n) x^{n-d} \frac{y^n}{n!}$$ where $f_{d, \ell}(A_n)$ is the number of faces of dimension $d$ and level $\ell$. The essential structural result is the factorization

$F_\ell(x, y) = [F_1(x, y)]^\ell$

which links higher-level face counts to product structures of level-1 faces, rooted in Zaslavsky’s convolution and the combinatorics of set partitions (Chen et al., 18 Jan 2026).

Applications include:

• Extraction of explicit enumerative formulas in classical ESAs (e.g., braid, Shi, semiorder arrangements) by specializing $F_1(x, y)$.
• Connection to Stirling numbers via alternating sums:

$$\sum_{d=\ell}^n (-1)^d f_{d, \ell}(A_n) = (-1)^\ell \ell! S(n, \ell)$$

Edge-Bicolored Graphs and Multivariate Expansion

Bivariate exponential integrals can encode the enumeration of labeled edge-bicolored graphs, linking analytic (Laplace-type integral) expansions to combinatorial generating functions: $$I(z) = \frac{z}{2\pi} \int_{\mathbb{R}^2} \exp\left(z\,g(x, y)\right) dx dy$$ where $g(x, y)$ is a polynomial perturbation around a Gaussian core. The polynomials $a_{s, t}$ in the expansion

$$\sum_{s, t \geq 0} a_{s, t}(\lambda) x^s y^t = \exp\left(\sum_{u + w \geq 1} \lambda_{u,w}\frac{x^u y^w}{u!w!}\right)$$

are precisely the exponential generating function coefficients for multivalent structures (vertex degrees) in these graphs (Borinsky et al., 2024).

This analytic-combinatorial translation enables saddle-point asymptotics, as in the Ising model analysis on random 4-regular graphs, with phase transition loci determined by polynomial maxima.

4. Linear Recurrences and Classification via Bivariate EGFs

Many combinatorial recurrences of the form

$$a_{n+1, k+1} = f(a_{n, k}, a_{n, k+1}, a_{n+1, k}, \dots; \text{parameters})$$

are encoded by bivariate EGFs. This allows translation of the recurrence into first-order linear PDEs for $F(x, y) = \sum_{n, k} a_{n, k} x^k y^n / n!$, whose solution space stratifies according to values of recurrence parameters (e.g., "types" determined by coefficients of $x$ in the PDE). Each type yields explicit, often closed-form, solutions for $F(x, y)$ and hence all $a_{n, k}$ (G. et al., 2013).

This approach unifies the treatment of binomial, Stirling, Eulerian, and related triangles, and identifies parameter degeneracies—distinct recurrences yielding identical arrays due to underlying algebraic identities.

5. Bivariate Exponential Generating Functions in Special Number Sequences

Exponential (Bell) and Related Polynomials

The classical EGF for Bell numbers $b_n$ and exponential polynomials $\Phi_n(x)$ is

$$\sum_{n=0}^\infty \Phi_n(x) \frac{t^n}{n!} = \exp(x(e^t - 1))$$

A generalization is the “shifted” bivariate EGF: $$\sum_{n=0}^{\infty} \Phi_{n+m}(x) \frac{t^n}{n!} = e^{x(e^{t}-1)} \Phi_m(x e^t), \quad m \geq 0$$ which admits direct operational and combinatorial interpretations, and specializes to explicit generating functions for related polynomial families (geometric, Apostol–Bernoulli, Apostol–Euler) via elementary substitutions (Kargın, 2015).

Bernoulli Numbers and Log-Rational EGFs

The bivariate EGF for $p$-Bernoulli numbers $B_{n, p}$ can be written in closed log-rational form involving elementary functions: $$G(x, y) = \sum_{n \geq 0} \sum_{p \geq 0} B_{n, p} \frac{x^n}{n!} y^p = \frac{x - (1-u) \ln(1-u) - u}{(e^{x} - 1)(1-u)^2}$$ with $u = \frac{y e^x}{e^x - 1}$, and all expansion coefficients accessible via standard generating-function manipulations (Prodinger et al., 2018). The bivariate parameter interpolates between classical and extended Bernoulli numbers.

6. Operational, Integral, and Differential Representations

Bivariate EGFs often afford multiple equivalent characterizations:

• Operational: Application of exponential (Rodrigues-type) operators to monomials, as in $$H_{m,n}(z, \bar{z}) = \exp(-\partial_z \partial_{\bar{z}})(z^m \bar{z}^n)$$.
• Integral representations: As in Fourier–Wigner transforms, Laplace or contour integrals representing sums over multivariable polynomial or graph-theoretic structures.
• Differential equations: The EGF satisfies specific first-order PDEs reflecting the underlying combinatorial recursion or algebraic structure. For example, $E(z, \bar{z}; t, s)$ above solves explicit ladder-operator and Bochner-type eigenvalue relations (Ghanmi et al., 2019).

These forms underpin the extraction of recurrence relations, orthogonality, structural symmetries, and spectral properties.

7. Applications, Factorization, and Structural Consequences

Bivariate EGFs serve as organizing centers for enumeration, structural theorems, and asymptotic analysis. Key features include:

• Product Structure: Many bivariate EGFs display multiplicative factorization, as in $F_\ell(x, y) = [F_1(x, y)]^\ell$, reflecting combinatorial decomposability by level or block structure.
• Specializations and Identities: Setting parameters (e.g., $x = 0$ or $y = 0$) recovers univariate EGFs and classical results in the theory of special numbers and polynomials.
• Alternating Sums and Invariants: Alternating-sum identities relate EGF coefficients to classical invariants, such as Stirling numbers of the second kind, independently of specific arrangement or recurrence details.
• Binomial/Whitney Basis Expansion: Expansion in the binomial basis (e.g., for Whitney polynomials) allows translation to characteristic or chromatic polynomial identities and combinatorial invariants relevant in algebraic and geometric combinatorics (Chen et al., 18 Jan 2026).
• Asymptotic Combinatorics: In advanced analytic settings, saddle-point and Laplace asymptotics of bivariate EGFs associated to multidimensional integrals govern phase transitions and substructure enumeration in statistical models (Borinsky et al., 2024).

The bivariate exponential generating function thus acts as a universal object unifying enumeration, polynomial/algebraic structure, analytic asymptotics, and combinatorial symmetries across mathematics.