---
title: Generalized Meixner-type Free Gamma Distributions
url: https://www.emergentmind.com/topics/generalized-meixner-type-free-gamma-distributions
type: topic
---

# Generalized Meixner-type Free Gamma Distributions

Searching arXiv for the relevant papers and related terminology.
The term **generalized Meixner-type free gamma distributions** refers to the three-parameter family
\[
\mu_{t,\theta,\lambda}\qquad (t,\theta>0,\ \lambda\ge 1),
\]
introduced as a class of probability measures on \(\mathbb R_{\ge 0}\) that includes both the free gamma distributions introduced by Anshelevich and certain scaled free beta prime distributions introduced by Yoshida [2508.15585]. In the source introducing this family, these laws are defined through an \(R\)-transform, placed inside the free Meixner framework, and related to Gibbs measures and free entropy maximization. At the same time, the literature contains another important notion of a **free Gamma distribution**, namely the Bercovici–Pata image \(\nu_\alpha=\Lambda(\mu_\alpha)\) of the classical Gamma law, studied as a free selfdecomposable distribution rather than as a Meixner-class law [1302.3738]. This distinction is central: generalized Meixner-type free gamma distributions belong to the Meixner-side theory, whereas the Bercovici–Pata free Gamma laws are analyzed through free infinite divisibility and analytic subordination, not through a generalized Meixner classification.

## 1. Definition and placement within the free Meixner framework

The family \(\mu_{t,\theta,\lambda}\) is defined by its \(R\)-transform as
\[
R_{\mu_{t,\theta,\lambda}}(z)
 =\int_{\mathbb R}\left(\frac{1}{1-zx}-1\right)\frac{t\,k_{\theta,\lambda}(x)}{x}\,dx,
 \qquad z\in \mathbb C^-,
\]
where \(k_{\theta,\lambda}\) is the Marchenko–Pastur density
\[
k_{\theta,\lambda}(x)
 = \frac{\sqrt{(a^+-x)(x-a^-)}}{2\pi\theta x}\mathbf 1_{(a^-,a^+)}(x),
 \qquad a^\pm=\theta(\sqrt\lambda\pm 1)^2.
\]
The same paper derives the closed form
\[
R_{\mu_{t,\theta,\lambda}}(z)
 = t\cdot \frac{1+\theta(1-\lambda)z-\sqrt{(1+\theta(1-\lambda)z)^2-4\theta z}}{2\theta},
\]
which makes the deformation parameter \(\lambda\) explicit [2508.15585].

The family is called **generalized Meixner-type** because it sits inside the free Meixner framework:
\[
\mu_{t,\theta,\lambda}
 = \nu_{t\theta\lambda,\ \theta(\lambda+1),\ \theta^2\lambda}\boxplus \delta_t,
\]
where \(\nu_{s,a,b}\) is the centered free Meixner distribution. In this formulation, \(\mu_{t,\theta,\lambda}\) is a shifted free Meixner law, with parameters belonging to a Meixner regime satisfying \(a^2\ge 4b\) [2508.15585].

This placement is consistent with the broader characterization of free Meixner laws by Jacobi parameters that are constant from level \(2\) onward,
\[
a=(a_1,a_2,a_2,\dots), \qquad B=(\beta_1,\beta_2,\beta_2,\dots),
\]
together with the continued-fraction expansion of the Cauchy transform. Within that broader family, the literature explicitly lists **free Gamma** among the canonical subclasses, alongside the semicircle, Marchenko–Pastur, free Pascal, free binomial, and free hyperbolic secant laws [1305.3470].

A further point of terminology comes from the characterization paper on free Meixner laws: there the normalized family \(\mu_{a,b}\) is classified by the parameter relation \(b>0,\ a^2=4b\) as the **free Gamma law**, identifying free Gamma as a boundary case of the free Meixner class [1211.4260]. This suggests that, in the Meixner literature, “free Gamma” is a structural subclass of free Meixner laws, whereas in the Bercovici–Pata literature the same phrase designates a different construction.

## 2. Relation to earlier notions of free gamma distributions

For \(\lambda=1\), the generalized family reduces to the Meixner-type free gamma law:
\[
\mu_{t\theta,\theta,1}=\eta(t,\theta).
\]
The source describes \(\eta(t,\theta)\) as Anshelevich’s free gamma distribution and records its \(R\)-transform as
\[
R_{\eta(t,\theta)}(z)=\frac{t}{2}\bigl(1-\sqrt{1-4\theta z}\bigr).
\]
It also satisfies the free analogues
\[
\eta(1,\theta)^{\boxplus t}=\eta(t,\theta),\qquad
\eta(t_1,\theta)\boxplus \eta(t_2,\theta)=\eta(t_1+t_2,\theta),\qquad
D_\theta(\eta(t,1))=\eta(t,\theta),
\]
so the generalized family extends a free convolution semigroup already present in the Meixner-type gamma case [2508.15585].

By contrast, the paper “On the free Gamma distributions” defines, for each \(\alpha>0\),
\[
\nu_\alpha=\Lambda(\mu_\alpha),
\]
where \(\Lambda:ID(*)\to ID(\boxplus)\) is the Bercovici–Pata bijection and \(\mu_\alpha\) is the classical Gamma law on \([0,\infty)\) with density
\[
\frac{1}{\Gamma(\alpha)} t^{\alpha-1}e^{-t}\,dt.
\]
That paper explicitly places \(\nu_\alpha\) in the framework of free selfdecomposable distributions and states that it does **not** identify \(\nu_\alpha\) as a Meixner law or as a member of a generalized Meixner family [1302.3738].

This terminological split matters. The generalized Meixner-type family \(\mu_{t,\theta,\lambda}\) extends the **Anshelevich/Meixner-type free gamma law**, not the Bercovici–Pata free Gamma law \(\nu_\alpha\). A plausible implication is that two different research programs use “free gamma” for structurally different objects: one governed by the Meixner \(R\)-transform algebra, the other by the Bercovici–Pata correspondence and free selfdecomposability.

## 3. Explicit transforms, densities, cumulants, and support

The Cauchy transform of \(\mu_{t,\theta,\lambda}\) is computed explicitly as
\[
G_{\mu_{t,\theta,\lambda}}(z)
 = \frac{(t+2\theta)z-t\bigl(t-\theta(\lambda-1)\bigr)-t\sqrt{(z-\alpha^-)(z-\alpha^+)}}{2\theta z(z+t(\lambda-1))},
\]
with
\[
\alpha^\pm=\theta(\lambda+1)+t\pm 2\sqrt{\theta\lambda(\theta+t)}.
\]
From this one obtains the absolutely continuous part
\[
\frac{d\mu_{t,\theta,\lambda}}{dx}(x)
 = \frac{t\sqrt{(x-\alpha^-)(\alpha^+-x)}}{2\pi\theta\,x(x+t(\lambda-1))}
 \mathbf 1_{[\alpha^-,\alpha^+]}(x),
\]
together with a possible atom at \(0\):
\[
\mu_{t,\theta,\lambda}(\{0\}) =
\begin{cases}
0, & 1\le \lambda\le 1+t/\theta,\\[4pt]
1-\dfrac{t}{\theta(\lambda-1)}, & \lambda>1+t/\theta.
\end{cases}
\]
Thus the support changes qualitatively at the threshold \(\lambda=1+t/\theta\): below and at the threshold, the law is purely absolutely continuous on a compact interval; above it, an atom at the origin appears [2508.15585].

The paper also gives a free-cumulant formula. The first cumulant is
\[
\kappa_1(\mu_{t,\theta,\lambda})=t,
\]
and for \(n\ge 1\),
\[
\kappa_{n+1}(\mu_{t,\theta,\lambda})
 = t\,m_n(\pi_{\theta,\lambda})
 = \frac{t\theta^n}{n}\sum_{k=0}^{n-1}\binom{n}{k}\binom{n}{k+1}\lambda^k.
\]
The moments are then obtained by the moment-cumulant formula [2508.15585].

These formulas place the family in direct contact with standard free Meixner descriptions. For normalized free Meixner laws with \(a_1=0\), \(\beta_1=1\), the density is recorded in the free Meixner random-matrix paper as
\[
d\mu(x)=\frac{\sqrt{4\beta_2-(x-a_2)^2}}{2\pi\big((\beta_2-1)x^2+a_2x+1\big)}\,dx
\]
on
\[
[a_2-2\sqrt{\beta_2},\,a_2+2\sqrt{\beta_2}],
\]
with the possibility of one or two atoms outside the absolutely continuous part [1305.3470]. The generalized Meixner-type free gamma densities fit this larger pattern, but with the specific nonnegative-support structure encoded by \((t,\theta,\lambda)\).

## 4. Convolution, scaling, and mixture structure

The parameter \(t\) acts as a free-convolution time:
\[
\mu_{t,\theta,\lambda}=\mu_{1,\theta,\lambda}^{\boxplus t}.
\]
The paper also states the scaling relation
\[
\mu_{t,\theta,\lambda}=D_\theta(\mu_{t,1,\lambda})^{\boxplus 1/\theta},
\]
so \(\theta\) behaves as a scale parameter [2508.15585].

For \(\lambda>1\), the family admits a free multiplicative convolution formula:
\[
\mu_{t,\theta,\lambda}
 = D_{t(\lambda-1)}
 \left(
 \pi_{1,\frac{t}{\theta(\lambda-1)}}
 \boxtimes
 (\pi_{1,1+t/\theta})^{\langle -1\rangle}
 \right)
 = \mu_{t,\theta,1}\boxtimes \pi_{q^{-1},q},
\]
where
\[
q=\frac{t}{\theta(\lambda-1)}.
\]
This exhibits \(\mu_{t,\theta,\lambda}\) as a free multiplicative convolution of the free gamma law \(\mu_{t,\theta,1}\) with a Marchenko–Pastur law [2508.15585].

The same source identifies a beta-prime description:
\[
\mu_{t,\theta,\lambda}
 = D_{t(\lambda-1)}
 \left(
 f\beta'\!\left(\frac{t}{\theta(\lambda-1)},\,1+\frac{t}{\theta}\right)
 \right),
\]
where
\[
f\beta'(a,b):=\pi_{1,a}\boxtimes \pi_{1,b}^{\langle -1\rangle},
\qquad a>0,\ b>1.
\]
Equivalently, \(\mu_{t,\theta,\lambda}\) is a scaled free beta prime law [2508.15585].

At the boundary value
\[
\lambda=1+t/\theta,
\]
one has
\[
\mu_{t,\theta,1+t/\theta}=\mu_{t,\theta,1}\boxtimes \pi_{1,1},
\]
and the paper states that at this boundary the measure becomes a free compound Poisson law [2508.15585]. This boundary is therefore the transition point between purely absolutely continuous generalized gamma behavior and a regime in which a point mass at zero appears.

## 5. Selfdecomposability, unimodality, and entropy–potential correspondence

A sharp structural dichotomy is stated in the introduction paper for the generalized family:
\[
\mu_{t,\theta,\lambda}\text{ is freely selfdecomposable } \iff \lambda=1.
\]
Thus only the undeformed free gamma case remains freely selfdecomposable. Likewise, for fixed \(t,\theta>0\),
\[
\mu_{t,\theta,\lambda}\text{ is unimodal } \iff 1\le \lambda \le 1+t/\theta.
\]
The unimodality threshold therefore coincides with the onset of the atom at zero [2508.15585].

The same work develops a **potential correspondence** based on Gibbs measures rather than the Bercovici–Pata bijection. The associated classical Gibbs law is
\[
\rho_{t,\theta,\lambda}(dx)=\frac{1}{\mathcal Z_{t,\theta,\lambda}}e^{-V_{t,\theta,\lambda}(x)}\,dx,
\]
with
\[
V_{t,\theta,\lambda}(x)=
\begin{cases}
\left(2+\frac{t}{\theta}\right)\log x+\frac{t^2}{\theta x}, & \lambda=1,\\[6pt]
\left(1-\frac{t}{\theta(\lambda-1)}\right)\log x
+\left(1+\frac{t\lambda}{\theta(\lambda-1)}\right)\log(x+t(\lambda-1)), & \lambda>1.
\end{cases}
\]
For \(\lambda>1\),
\[
\rho_{t,\theta,\lambda}
 = D_{t(\lambda-1)}
 \left(
 \gamma\!\left(\frac{t}{\theta(\lambda-1)},1\right)
 \circledast
 \gamma\!\left(1+\frac{t}{\theta},1\right)^{\langle -1\rangle}
 \right)
 = \rho_{t,\theta,1}\circledast
 \gamma\!\left(\frac{t}{\theta(\lambda-1)},\frac{\theta(\lambda-1)}{t}\right).
\]
The paper emphasizes that this classical–free matching is **not** the Bercovici–Pata bijection, but a variational correspondence between equilibrium measures of entropy functionals under matching potentials [2508.15585].

Its main variational theorem states that for \(t,\theta>0\) and
\[
1\le \lambda < 1+t/\theta,
\]
the measure \(\mu_{t,\theta,\lambda}\) is the unique maximizer of Voiculescu’s free entropy functional
\[
\Sigma_{V_{t,\theta,\lambda}}(\mu)
=
\iint \log|x-y|\,\mu(dx)\mu(dy)
-
\int V_{t,\theta,\lambda}(x)\,\mu(dx)
\]
over all probability measures on \(\mathbb R_{>0}\) [2508.15585]. This identifies the generalized Meixner-type free gamma distributions as free equilibrium measures for a natural logarithmic potential family.

For comparison, the Bercovici–Pata free Gamma laws \(\nu_\alpha\) are analyzed by entirely different methods. The central relation there is
\[
G_{\nu_\alpha}(H_\alpha(z))=\frac{1}{z},
\]
with \(H_\alpha\) built from the Cauchy transform of the classical exponential law. The resulting density is analytic on \((s_\alpha,\infty)\), supported on \([s_\alpha,\infty)\), and unimodal, but the classification invoked is free selfdecomposability under \(\Lambda\), not Meixner structure [1302.3738].

## 6. Characterizations, operator models, and conceptual scope

The free Meixner setting supplies both structural characterizations and random-matrix realizations. In the characterization paper, the normalized free Meixner laws \(\mu_{a,b}\) have Cauchy transform
\[
G_{\mu_{a,b}}(z)
 = \frac{(1+2b)z+a-\sqrt{(z-a)^2-4(1+b)}}{2(bz^2+az+1)},
\]
and \(R\)-transform
\[
R_{\mu_{a,b}}(z)
 =\frac{1-az-\sqrt{(1-az)^2-4bz^2}}{2z}.
\]
The classification recorded there includes the free Gamma law as the case
\[
b>0,\qquad a^2=4b.
\]
The central theorem states that free Meixner laws are characterized by a linear regression condition together with a conditional moment identity involving a third-degree polynomial. Since free Gamma is the boundary case \(a^2=4b\), the theorem specializes automatically to the free Gamma case [1211.4260].

Independently, the random-matrix model paper constructs a \(2\times 2\) block Gaussian Hermitian ensemble
\[
Y(n)=
\begin{pmatrix}
A(n) & B(n)\\
C(n) & D(n)
\end{pmatrix},
\qquad
C(n)=B(n)^*,
\]
with a degenerating block regime
\[
d_1=\lim_{n\to\infty}\frac{|N_1|}{n}=0,\qquad
d_2=\lim_{n\to\infty}\frac{|N_2|}{n}=1,
\]
and deterministic diagonal shift
\[
M(n)=Y(n)+a_1 I_1(n)+a_2 I_2(n).
\]
Under the first partial trace, the moments converge to those of the free Meixner law associated with \((a_1,a_2,\beta_1,\beta_2)\); for ensembles of such matrices, the paper proves asymptotic conditional freeness with respect to the pair of partial traces \((T_1(n),T_2(n))\) [1305.3470]. Since free Gamma is explicitly listed as one of the Meixner subclasses, this model provides a random-matrix realization of the encompassing class in which generalized Meixner-type free gamma distributions reside.

Taken together, these works delineate a precise conceptual scope. The generalized distributions \(\mu_{t,\theta,\lambda}\) unify four structures already present in the literature: free gamma distributions in the Meixner sense, scaled free beta prime distributions, shifted centered free Meixner laws, and free equilibrium measures for explicit logarithmic potentials [2508.15585]. At the same time, the phrase “free Gamma distribution” remains non-uniform across free probability. One line of work uses it for Meixner-type laws characterized by \(R\)-transforms, Jacobi parameters, and cubic conditional moments; another uses it for Bercovici–Pata images of classical Gamma laws with analytic densities on half-lines and asymptotic selfdecomposability properties [1302.3738]. A plausible implication is that any discussion of “generalized Meixner-type free gamma distributions” must specify that it concerns the Meixner-type branch of the theory rather than the Bercovici–Pata free Gamma laws.

Source: https://www.emergentmind.com/topics/generalized-meixner-type-free-gamma-distributions