---
title: Steiner Entire Function Analysis
url: https://www.emergentmind.com/topics/steiner-entire-function
type: topic
---

# Steiner Entire Function Analysis

A Steiner entire function, in the sense introduced in "On Steiner entire function" [2507.11626], is the analytic generating function
\[
F_K(z):=f_K(z):=\sum_{k=0}^\infty V_k(K)\,z^k,\qquad z\in\mathbb{C},
\]
associated with the intrinsic volumes of a convex compact set \(K\) in a separable real Hilbert space. It extends the classical finite-dimensional Steiner polynomial to infinite dimensions and encodes geometric information about \(K\) in the growth, zero set, and canonical product structure of an entire function. The terminology is not uniform across the broader entire-function literature: related papers use "Steiner-type" for ray-distributed value problems rather than intrinsic-volume generating functions [1509.03283].

## 1. Geometric origin and definition

Let \(H\) be a separable real Hilbert space with inner product \(\langle\cdot,\cdot\rangle\) and norm \(\|\cdot\|\), and let \(K\subset H\) be convex and compact. In finite dimensions, the normalized intrinsic volumes \(V_k(K)\) are characterized by the classical Steiner formula for parallel sets, with the normalization chosen so that \(V_0(K)=1\), \(V_d(K)=\operatorname{vol}_d(K)\), \(V_{d-1}(K)\) equals one half of the surface area, and \(V_1(K)\) is proportional to the mean width. These quantities depend only on \(K\) and \(k\), and are independent of the ambient dimension once \(K\) is fixed. In the infinite-dimensional setting, following Sudakov and Chevet, the intrinsic volumes are defined by approximation through finite-dimensional convex compact subsets; if \(K\) has finite affine dimension, this agrees with the classical definition [2507.11626].

The relevant regularity hypothesis is Gaussian boundedness. A convex compact \(K\) is called Gaussian bounded (GB) if \(V_1(K)<\infty\). Equivalently, the isonormal Gaussian process on \(H\),
\[
\mathbb{E}[\xi(t)\xi(t')] = \langle t,t'\rangle,
\]
has a version that is almost surely bounded on \(K\). For GB-sets, Chevet and McMullen proved the sharp bound
\[
V_k(K)\le \frac{(V_1(K))^k}{k!},\qquad k=0,1,2,\dots,
\]
which implies the finiteness of the Wills functional
\[
W(K):=\sum_{k=0}^\infty V_k(K)\le \exp\!\big(V_1(K)\big).
\]
The Steiner entire function is then defined by collecting the intrinsic volumes into a single power series.

This construction is simultaneously geometric and analytic. Geometrically, it packages the intrinsic-volume sequence into one object. Analytically, it places convex-geometric questions into the framework of entire function theory, where order, type, Hadamard factorization, and coefficient asymptotics become directly relevant.

## 2. Entire-function structure and finite-dimensional recovery

The estimate
\[
|F_K(z)|\le \sum_{k=0}^\infty \frac{\big(V_1(K)|z|\big)^k}{k!}=\exp\!\big(V_1(K)\,|z|\big),\qquad z\in\mathbb{C},
\]
shows that \(F_K\) is entire and of order at most \(1\), with exponential type at most \(V_1(K)\). Thus the basic analytic class is fixed by the Gaussian-bounded regime: every Steiner entire function is an entire function of controlled exponential growth [2507.11626].

In finite dimensions, the construction reduces to a polynomial. If \(K\subset\mathbb{R}^d\) is a convex body, only the first \(d+1\) coefficients are nonzero, so
\[
F_K(z)=\sum_{k=0}^d V_k(K)\,z^k,\qquad z\in\mathbb{C}.
\]
This is precisely the "Wills polynomial" of Cifre and Nicol, with \(F_K(1)=W(K)\). The classical Steiner formula shows that the coefficients of this polynomial are exactly the intrinsic volumes. Scaling is encoded analytically by
\[
F_{\alpha K}(z)=F_K(\alpha z),
\]
because \(V_k(\alpha K)=\alpha^kV_k(K)\).

Several elementary convex bodies give explicit closed forms. For the axis-parallel cube \(C_n(L)=[0,L]^n\),
\[
V_k(C_n(L))=\binom{n}{k}L^k,
\qquad
F_{C_n(L)}(z)=(1+L z)^n.
\]
For a rectangular box with side lengths \(\ell_1,\dots,\ell_n\),
\[
F_K(z)=\prod_{j=1}^n (1+\ell_j z),
\quad\text{so}\quad
V_k(K)=\sum_{1\le j_1<\cdots<j_k\le n}\ell_{j_1}\cdots \ell_{j_k}.
\]

For general GB-sets, Hadamard factorization applies. Since the order is at most \(1\), the genus is at most \(1\), and one has
\[
F_K(z)=e^{c z}\prod_{j=1}^\infty\Big(1-\frac{z}{z_j}\Big)\exp\!\Big(\frac{z}{z_j}\Big),
\]
where \(\{z_j\}\) are the nonzero zeros of \(F_K\). If \(\sum_j |z_j|^{-1}<\infty\), the genus drops to \(0\) and the product simplifies to
\[
F_K(z)=e^{c z}\prod_{j=1}^\infty\Big(1-\frac{z}{z_j}\Big).
\]
The convergence exponent of the zero sequence equals the order.

## 3. Order, type, and Gaussian continuity

Let
\[
M_{F_K}(r):=\max_{|z|=r}|F_K(z)|.
\]
The order and type are
\[
\rho(K):=\limsup_{r\to\infty}\frac{\ln\ln M_{F_K}(r)}{\ln r},
\qquad
\sigma(K):=\limsup_{r\to\infty}\frac{\ln M_{F_K}(r)}{r^\rho}
\quad (0<\rho<\infty).
\]
The coefficient-based formulas recorded in the paper are
\[
\rho(K)=\lim_{n\to\infty}\frac{n\ln n}{\ln\!\big(1/|V_n(K)|\big)},
\qquad
(\sigma(K)e\rho(K))^{1/\rho(K)}=\lim_{n\to\infty} n^{1/\rho(K)}\,|V_n(K)|^{1/n}.
\]
The main general theorem is that if \(K\) is GB, then \(\rho(K)\le 1\), and if \(\rho(K)=1\), then \(\sigma(K)\le V_1(K)\). Moreover, the full range \(\rho\in[0,1]\) is realizable by suitable GB-sets [2507.11626].

A central sequence is
\[
m_k(K):=\frac{(k+1)V_{k+1}(K)}{V_k(K)},\qquad k\ge 0.
\]
By ultra-log-concavity,
\[
V_k(K)^2\ge \frac{k+1}{k}\,V_{k-1}(K)\,V_{k+1}(K),
\]
the sequence \(\{m_k(K)\}\) is non-increasing and has a limit. Vitale’s criterion identifies Gaussian continuity (GC) by
\[
K\text{ is GC}\quad\Longleftrightarrow\quad \lim_{k\to\infty} m_k(K)=0.
\]
The Steiner entire function translates this into an analytic growth condition: for a GB-set \(K\), \(K\) is GC if and only if either \(\rho(K)<1\), or \(\rho(K)=1\) and \(\sigma(K)=0\). If \(\rho(K)=1\), then the type equals the oscillation,
\[
\sigma(K)=\operatorname{osc}(K)=\lim_{k\to\infty} m_k(K).
\]

This is the core bridge between convex geometry, Gaussian process regularity, and entire-function theory. The geometric question of sample-path continuity becomes a statement about whether the generating entire function has full exponential order and, if so, whether its type vanishes.

## 4. Asymptotics, conjectures, and obstructions

The coefficient ratios admit a quantitative asymptotic law:
\[
\limsup_{k\to\infty}\frac{\ln m_k(K)}{\ln k} = 1-\frac{1}{\rho(K)}.
\]
By convention, \(1/\rho(K)=\infty\) when \(\rho(K)=0\), so in that case \(m_k(K)\) decays faster than any power. This formula refines the relation between the decay of intrinsic volumes and the order of the Steiner entire function [2507.11626].

An important consequence is the disproof of a conjecture of Gao and Vitale. They conjectured that for any GB-compact \(K\subset H\), either \(\lim_{k\to\infty} m_k(K)>0\), or \(m_k(K)=O(k^{-1/2})\). The asymptotic formula shows that whenever \(\rho(K)>2/3\), the decay exponent \(1-1/\rho(K)\) lies in \((-1/2,0]\), so \(m_k(K)\) decays strictly slower than \(k^{-1/2}\). Since all orders \(\rho\in(2/3,1]\) are realizable, the conjectured dichotomy fails.

The same paper proposes a new structural conjecture. A sequence \((a_k)_{k\ge 0}\) with \(a_0=1\) and \(a_k>0\) is conjectured to be the intrinsic volume sequence of some infinite-dimensional GB-compact \(K\subset H\) if and only if it is ultra-log-concave:
\[
a_k^2 \,\ge\, \frac{k+1}{k}\,a_{k-1}\,a_{k+1},\qquad k=1,2,\dots.
\]
Necessity follows from Alexandrov–Fenchel-type inequalities; sufficiency is conjectural. Within the analytic language of Steiner entire functions, this is a proposed characterization of admissible coefficient sequences.

These results separate two issues that are sometimes conflated. Ultra-log-concavity is necessary for intrinsic-volume sequences, but it is not yet known to be sufficient. Likewise, order bounds are sharp, but fine zero-distribution questions remain largely open outside explicit product cases.

## 5. Explicit models and zero distributions

Two stochastic-geometric examples are the closed convex hull of the Wiener spiral and the closed convex hull of the Wiener spiral bridge. For
\[
K:=\overline{\operatorname{conv}(S)},\qquad S:=\{\mathbf{1}_{[0,t]}:\ t\in[0,1]\}\subset L^2[0,1],
\]
the intrinsic volumes satisfy
\[
V_k(K)=\frac{\kappa_k}{k!},
\qquad
m_k(K)=\sqrt{\pi}\,k^{-1/2}(1+o(1)).
\]
The Steiner entire function admits the hypergeometric representation
\[
F_K(z) ={}_0F_2\!\left(\tfrac{1}{2},1;\tfrac{\pi z^2}{4}\right) +2z\,{}_0F_2\!\left(\tfrac{3}{2},\tfrac{3}{2};\tfrac{\pi z^2}{4}\right),
\]
and Stirling-type asymptotics give
\[
\rho(K)=\frac{2}{3},\qquad \sigma(K)=\frac{3}{2}\,(2\pi)^{1/3}.
\]
For the bridge body \(K^\circ\), one has the same asymptotic growth indicators and a comparable hypergeometric closed form [2507.11626].

The cleanest zero description occurs for infinite-dimensional parallelepipeds
\[
K=\prod_{j=1}^\infty [0,\ell_j]\subset \ell^2,
\]
with \(\sum_{j\ge 1}\ell_j^2<\infty\). Such a set is GB if and only if \(\sum_{j\ge 1}\ell_j<\infty\), and then
\[
V_k(K)=\sum_{1\le j_1<\cdots<j_k}\ell_{j_1}\cdots \ell_{j_k},
\qquad
F_K(z)=\prod_{j=1}^\infty (1+\ell_j z).
\]
The zeros are exactly at \(z=-1/\ell_j\), so they lie on the negative real axis, and the order is the convergence exponent of \(\{1/\ell_j\}\):
\[
\rho(K)=\limsup_{j\to\infty}\frac{\ln j}{\ln(1/\ell_j)}.
\]
This realizes the full range of admissible orders. If \(\ell_j=j^{-\alpha}\) with \(\alpha>1\), then \(\rho=1/\alpha\in(0,1)\); if \(\ell_j=e^{-j}\), then \(\rho=0\); if \(\ell_j=\frac{1}{j(\ln j)^2}\), then \(\rho=1\).

These examples show that the Steiner entire function is not merely a formal encoding device. In explicit families, the decay of geometric side lengths becomes the zero distribution and growth of an entire function in a literal Hadamard-product sense.

## 6. Terminological range and related notions

The expression "Steiner entire function" is not standard across the entire-function literature. In the intrinsic-volume setting it has a precise definition, but several nearby traditions use the word "Steiner" differently or only heuristically.

| Context | Meaning of the function | Source |
|---|---|---|
| Convex geometry in Hilbert space | \(F_K(z)=\sum_{k\ge 0}V_k(K)z^k\), the generating function of intrinsic volumes | [2507.11626] |
| Radially distributed zeros and one-points | "Steiner-type" entire functions with zeros and \(1\)-points on prescribed rays, or close to them | [1509.03283] |
| Entire factorial constructions | Broader "Steiner-type" periodic-factor approach to entire solutions related to \(\Gamma\) and \(F(z+1)=zF(z)\) | [2107.11330] |

In "Entire functions with two radially distributed values" [1509.03283], the term "Steiner entire function" is not used, but the paper explicitly notes that the setting of zeros and \(1\)-points constrained to a finite system of rays is the kind of structure often sought under "Steiner/Steinmetz-type" inquiries. There the key theorem states that if zeros are close to a finite system of rays \(A\) and \(1\)-points are close to a finite system of rays \(B\), with \(A\cap B=\{0\}\), then the order is determined by the geometry:
\[
\rho=\frac{\pi}{\omega},
\]
where \(\omega\) is the largest angle between adjacent rays in \(A\cup B\). The paper also gives ODE/Stokes constructions of such functions.

In "A new entire factorial function" [2107.11330], the phrase appears only in a comparative sense. The paper distinguishes its entire factorial function \(K(z)\) from exact-recurrence entire factorials obtained by writing \(F(z)=\Gamma(z)H(z)\) with \(H\) entire and \(1\)-periodic, a construction described there as part of a broader Steiner-type approach. That usage concerns pole cancellation and interpolation of factorial values, not intrinsic volumes.

Other papers use "Steiner" in still different ways. "Applications of Steiner symmetrization to some extremal problems in geometric function theory" [1607.01674] studies Steiner symmetrization of analytic functions in the unit disk and states explicitly that its framework yields a family of radius-by-radius symmetrized maps rather than a single global entire function. "Entire functions arising from trees" [1803.00963] states that, in that paper’s terminology, the relevant objects are Shabat entire functions, not Steiner entire functions. Finally, "An Explicit Entire Function of Order One with All Zeros on a Line and Bounded in a Half-Plane" [2601.18687] says that "Steiner entire function" is not a standard term, while proposing its explicitly constructed \(\Xi_c(s)\) as a prototypical example only "in this sense."

The consistent conclusion is therefore contextual rather than universal. In current explicit form, the mathematically precise notion introduced under that name is the intrinsic-volume generating entire function \(F_K\) of a GB convex compact set in a Hilbert space. Other occurrences belong to adjacent traditions—radial value distribution, Steiner symmetrization, entire factorials, or analytically designed Hadamard products—and should not be identified with \(F_K\) without further qualification.

Source: https://www.emergentmind.com/topics/steiner-entire-function