---
title: Chow Functions in Incidence Algebra and Geometry
url: https://www.emergentmind.com/topics/chow-functions
type: topic
---

# Chow Functions in Incidence Algebra and Geometry

Chow functions are polynomial-valued invariants attached to kernels in the incidence algebra of a weakly ranked poset. They were introduced as a theory parallel to Kazhdan–Lusztig–Stanley functions, but with a different structural constraint: instead of enforcing a half-degree bound, Chow functions impose a palindromic or symmetric condition and often encode Hilbert–Poincaré series of Chow rings or cohomology-like objects. In the characteristic-kernel case they recover the Hilbert series of the Chow ring of a matroid, and subsequent work developed dual Chow functions, matrix analogues, and real-rootedness theorems for broad classes of posets and matroids [2411.04070, 2605.28474, 2509.17852].

## 1. Incidence-algebra construction

The basic input is a locally finite weakly ranked poset \(P\) with a weak rank function
\[
\rho:\operatorname{Int}(P)\to \mathbb Z_{\ge 0},
\]
satisfying
\[
\rho_{st}>0 \text{ for } s<t,\qquad \rho_{st}=\rho_{sw}+\rho_{wt}\text{ for } s\le w\le t.
\]
This determines a subalgebra \(I_\rho(P)\) of the incidence algebra, consisting of interval functions \(a_{st}(x)\) with
\[
\deg a_{st}(x)\le \rho_{st}.
\]
A standard involution is defined by coefficient reversal relative to the weak rank:
\[
(a^{\mathrm{rev}})_{st}(x)=x^{\rho_{st}}a_{st}(x^{-1}).
\]

A \((P,\rho)\)-kernel is an element \(\kappa\in I_\rho(P)\) such that \(\kappa_{ss}(x)=1\) for all \(s\in P\) and
\[
\kappa^{-1}=\kappa^{\mathrm{rev}}.
\]
Important examples include the characteristic kernel, the Eulerian kernel, and the \(R\)-polynomial kernel on Bruhat intervals. Given such a kernel, one forms the reduced kernel
\[
\bar\kappa_{st}(x)=
\begin{cases}
\dfrac{\kappa_{st}(x)}{x-1}, & s<t,\\[6pt]
-1, & s=t,
\end{cases}
\]
and defines the Chow function by
\[
H = -\,\bar\kappa^{-1}.
\]
Equivalently, \(H\) is characterized recursively by either convolution identity
\[
H_{st}(x)=\sum_{s< w\le t}\kappa_{sw}(x)\,H_{wt}(x),
\]
or
\[
H_{st}(x)=\sum_{s\le w<t}H_{sw}(x)\,\kappa_{wt}(x).
\]

The 2024 theory also gives an equivalent characterization of \(H\) as the unique element with \(H_{ss}=1\), symmetric interval polynomials, and
\[
\kappa H = H^{\mathrm{rev}}
\qquad\text{equivalently}\qquad
H\kappa=H^{\mathrm{rev}}.
\]
This formulation makes clear that Chow functions are intrinsic incidence-algebra objects rather than ad hoc generating series [2411.04070].

## 2. Relation to Kazhdan–Lusztig–Stanley theory

Chow functions were designed to parallel Kazhdan–Lusztig–Stanley (KLS) functions. For the same kernel \(\kappa\), the right and left KLS functions \(f,g\) satisfy
\[
f^{\mathrm{rev}}=\kappa f,\qquad g^{\mathrm{rev}}=g\kappa,
\]
together with the stronger degree bound
\[
\deg f_{st},\deg g_{st}<\frac{\rho_{st}}{2}.
\]
The conceptual distinction is that KLS functions are “lower-half” objects, whereas Chow functions are “palindromic full-degree” objects. In this sense, Chow functions retain more of the interval rank while replacing the KLS half-degree condition by symmetry [2411.04070].

The same framework defines augmented Chow functions
\[
F = H\cdot f^{\mathrm{rev}},\qquad G = g^{\mathrm{rev}}\cdot H,
\]
which generalize augmented Chow rings in matroid Hodge theory. A major technical tool is the numerical canonical decomposition, which expresses Chow functions recursively in terms of KLS data. Using the left KLS function \(g\), one has
\[
H_{st}(x)=\frac{g^{\mathrm{rev}}_{st}(x)-g_{st}(x)}{x-1}
+\sum_{s< w<t}\frac{g^{\mathrm{rev}}_{sw}(x)-x\,g_{sw}(x)}{x-1}\,H_{wt}(x).
\]
There is a dual formula using \(f\), and also a chain expansion
\[
H_{st}(x)=\sum_{s=p_0<p_1<\cdots<p_m=t} \prod_{i=1}^m \frac{g^{\mathrm{rev}}_{p_{i-1}p_i}(x)-x\,g_{p_{i-1}p_i}(x)}{x-1}.
\]

These identities are significant because they give purely numerical analogues of canonical decompositions that, in geometric settings, come from Chow rings or intersection cohomology. This suggests that the formalism is intended as a polynomial shadow of deeper graded-module structures, even when no such structure is known to exist [2411.04070].

## 3. Standard kernels, model examples, and geometric meaning

The characteristic kernel \(\chi\), the Eulerian kernel
\[
\varepsilon_{st}(x)=(x-1)^{\rho_{st}},
\]
and the \(R\)-polynomial kernel on Bruhat intervals are the principal examples. For a graded bounded poset, the characteristic kernel yields the characteristic Chow polynomial. In the matroid case, where \(P=L(M)\) is the lattice of flats of a loopless matroid \(M\), the characteristic Chow polynomial recovers a geometric invariant exactly:
\[
H_{L(M)}(x)=\operatorname{Hilb}(\mathrm{CH}(M),x).
\]
Likewise, the left augmented Chow polynomial satisfies
\[
G_{L(M)}(x)=\operatorname{Hilb}(\underline{\mathrm{CH}(M)},x).
\]
These identities are the main reason for the name “Chow functions” [2411.04070].

For Eulerian posets, the Eulerian Chow polynomial equals the \(h\)-polynomial of the order complex:
\[
H_P(x)=h(\Delta(P),x).
\]
Thus Eulerian Chow functions encode chain enumeration in a form compatible with barycentric subdivision. For Bruhat intervals in a Coxeter group, the \(R\)-polynomial kernel produces a Chow function with a direct path-counting interpretation:
\[
H_{uv}(x)=\sum_{\Delta\in B(u,v)} x^{\rho_{uv}/2+\mathrm{asc}(\Delta)}
=\sum_{\Delta\in B(u,v)} x^{\rho_{uv}/2+\mathrm{des}(\Delta)}.
\]
This makes Coxeter Chow functions new enumerators of Bruhat graph paths [2411.04070].

A related geometric viewpoint appears in the study of the stack of expanded pairs. There the integral Chow ring of the stack \(\sT\) is identified with the Hopf algebra of quasi-symmetric functions:
\[
CH(\sT)\cong \QSym.
\]
The stack \(\sT\) is presented as a colimit of stacks \(\cA^n=[\A^n/\Gm^n]\), and compatible Chow classes on the finite stages correspond exactly to quasi-symmetric functions. The paper explicitly describes this as a “Chow function” perspective: ordinary polynomial functions on \(\A^n\) become \(CH(\cA^n)\), compatible families across all \(n\) become \(\QSym\), and a gluing map \(\mu:\sT\times \sT\to \sT\) induces the Hopf coproduct on \(\QSym\) [1806.10700]. This suggests a broader geometric interpretation in which Chow-theoretic data behave as function spaces on moduli stacks.

## 4. Positivity, unimodality, \(\gamma\)-positivity, and real-rootedness

One of the main structural results is that Chow functions inherit positivity from KLS theory. If either the right or left KLS function is non-negative, then the Chow function is non-negative and unimodal. The proof is purely combinatorial and does not use Hard Lefschetz-type arguments; instead it relies on symmetry, product-preservation of unimodality for non-negative symmetric polynomials, and induction through the numerical canonical decomposition [2411.04070].

For the characteristic kernel on graded bounded posets, the theory sharpens substantially. If \(P\) is graded and bounded, then \(\chi\)-Chow is unimodal. If \(P\) is Cohen–Macaulay, then \(\chi\)-Chow is \(\gamma\)-positive, and a conjecture is stated that for every Cohen–Macaulay poset, \(\chi\)-Chow is real-rooted. A key formula expresses the \(\gamma\)-polynomial directly in terms of the flag \(h\)-vector:
\[
\gamma_P(x)=\sum_{S\subseteq [r-1]\text{ good}} \beta_P(S)\,x^{|S|}.
\]
This ties Chow-function positivity to classical flag-enumerative positivity [2411.04070].

The matrix extension strengthens these results from unimodality and \(\gamma\)-positivity to real-rootedness in a broad setting. For a lower triangular matrix \(R=(r_{n,k})\) with all diagonal entries equal to \(1\), regarded as an incidence-algebra element on a chain, one defines associated Chow polynomials \(H_n\) and Chow-derangement polynomials \(d_n\) by
\[
H_n=\sum_{k=0}^n r_{n,k}d_k,
\qquad
d_n=tS_{n-1}\left(\sum_{k=0}^{n-1} r_{n,k}d_k\right).
\]
If \(R\) is lower triangular, totally nonnegative, and has diagonal entries equal to \(1\), then the Chow polynomials \(H_n\) and Chow-derangement polynomials \(d_n\) are real-rooted, and moreover
\[
H_n\prec d_n,\qquad H_n\prec H_{n+1},\qquad d_n\prec d_{n+1}.
\]
The same paper proves augmented analogues and applies the theory to projective and affine geometries, dual partition and Dowling lattices, perfect matroid designs and paving matroids [2509.17852].

## 5. Dual Chow functions and matroid deletion theory

Dual Chow functions were introduced by applying the Chow construction not to \(\kappa\) itself but to the sign-twisted reverse kernel. Besides the reverse involution, one uses the sign twist
\[
(a^{\mathrm{sgn}})_{st}(x)=(-1)^{\rho_{st}}a_{st}(x).
\]
The dual Chow function of \(\kappa\) is then
\[
H^* := \text{Chow}\big((\kappa^{\mathrm{rev}})^{\mathrm{sgn}}\big).
\]
If \(\kappa\) satisfies the skew-symmetry relation
\[
\kappa_{st}(x)=(-1)^{\rho_{st}}\kappa_{st}(x^{-1}),
\qquad\text{equivalently}\qquad
\kappa^{\mathrm{rev}}=\kappa^{\mathrm{sgn}},
\]
then \((\kappa^{\mathrm{rev}})^{\mathrm{sgn}}=\kappa\), so dual and ordinary Chow functions coincide. The Eulerian kernel of an Eulerian poset and the \(R\)-polynomial kernel on Bruhat intervals satisfy this condition. By contrast, for the characteristic kernel the dual invariant is genuinely different [2605.28474].

For the characteristic kernel
\[
\chi=\mu\cdot \zeta^{\mathrm{rev}}=\zeta^{-1}\cdot \zeta^{\mathrm{rev}},
\]
the dual theory has a chain formula, positivity criteria, and deletion recursions. A central chain expansion is
\[
H_P^*(x) = (-1)^r \sum_{\widehat 0\le c_0<\cdots<c_m=\widehat 1} \mu_{\widehat 0c_0}
\prod_{i=1}^m \mu_{c_{i-1}c_i}\, \frac{x^{\rho_{c_{i-1}c_i}}-x}{x-1}.
\]
If the Möbius function alternates in sign,
\[
(-1)^{\rho_{st}}\mu_{st}\ge 0 \qquad \text{for all intervals }[s,t],
\]
then every dual Chow polynomial \(H^*_{st}(x)\) has nonnegative and unimodal coefficients. The dual characteristic Chow polynomial also admits an explicit \(\gamma\)-expansion:
\[
H^*_P(x)=\sum_{\substack{S\subseteq [r-2]\\ S\text{ stable}}}
\beta_P([r-1]\setminus S)\,x^{|S|}(1+x)^{r-1-2|S|}.
\]
Hence nonnegativity of the flag \(h\)-vector implies \(\gamma\)-positivity, and this applies in particular to Cohen–Macaulay posets and to all matroids [2605.28474].

For matroids, the theory produces a deletion formula. If \(M\) is a matroid and \(i\in E\) is neither a coloop nor parallel to any element, then
\[
H^*_{M}(x) = H^*_{M\setminus i}(x) +(x+1)H^*_{M/i}(x)
+x\sum_{F\in\underline{\mathscr S_i}} H^*_{M|F}(x)\,H^*_{M/(F\cup\{i\})}(x).
\]
This recursion implies, by induction, that \(H^*_M(x)\) is \(\gamma\)-positive for every matroid. The paper also derives explicit formulas for uniform matroids and proves that for uniform matroids the dual Chow polynomial and the dual augmented Chow polynomial are real-rooted [2605.28474]. A plausible implication is that dual Chow functions are intended to play for characteristic kernels a role analogous to the ordinary Chow function for Eulerian and Bruhat-type kernels, but with different correction terms under basic operations.

## 6. Matrix models, Eulerian specializations, and broader terminology

The matrix formalism makes Chow functions accessible beyond posets. For Toeplitz matrices associated to a sequence \(\{a_n\}\) with generating series
\[
f(z)=\sum_{n\ge0} a_n z^n,
\]
the Chow-type families admit closed generating functions:
\[
D(z,t)=\sum_{n\ge0} d_n(t)z^n = \frac{1-t}{f(tz)-t f(z)},
\]
\[
H(z,t)=\sum_{n\ge0} H_n(t)z^n = \frac{(1-t)f(z)}{f(tz)-t f(z)},
\]
\[
A(z,t)=\sum_{n\ge0} A_n(t)z^n = \frac{(1-t)f(tz)}{f(tz)-t f(z)},
\]
\[
G(z,t)=\sum_{n\ge0} G_n(t)z^n = \frac{(1-t)f(tz)f(z)}{f(tz)-t f(z)}.
\]
In the Boolean case, Chow polynomials recover Eulerian polynomials and Chow-derangement polynomials recover derangement polynomials. For Toeplitz matrices coming from Pólya frequency sequences, all four families are real-rooted. The same framework also recovers generalized Eulerian polynomials studied by Stanley, Brenti, Stembridge, and Shareshian–Wachs [2509.17852].

The term “Chow function” also appears in a different, regulator-theoretic sense. Chow polylogarithms are presented as integrals attached to a smooth complete variety \(Y\) and a wedge of rational functions \(a\in \Lambda^{2m-1}F(Y)^\times\), via
\[
\theta_Y(f_1\wedge\cdots\wedge f_{2m-1})
= \int_{Y(\mathbf C)} \log|f_1|\, d\log|f_2|\cdots d\log|f_{2m-1}|,
\qquad
P_m(Y,a)=\theta_Y(a).
\]
They are described there as a genuine “Chow function”: a regulator-type invariant built from algebraic cycles and rational functions. Their fundamental functional equation is a reciprocity law
\[
\sum_{D\subset Y} P_m\bigl(D,\partial_D(a)\bigr)=0,
\]
derived from a Chow-type complex and a Beilinson–Soulé-type vanishing statement [2411.03889]. This usage is conceptually distinct from the incidence-algebra theory, but it reflects the same general tendency: Chow-theoretic data are organized as function-like objects satisfying formal symmetries, residue identities, and functoriality.

Taken together, these developments place Chow functions at the intersection of incidence algebras, matroid and Coxeter combinatorics, moduli-theoretic Chow rings, and regulator theory. The common pattern is that a kernel, cycle complex, or compatible Chow-class system gives rise to a polynomial or analytic function-like invariant whose formal properties mirror geometric structures that may or may not be explicitly present.

Source: https://www.emergentmind.com/topics/chow-functions