---
title: 'Period-1 Polytopes: Exact Scaling in Lattice Enumeration'
url: https://www.emergentmind.com/topics/period-1-polytopes
type: topic
---

# Period-1 Polytopes: Exact Scaling in Lattice Enumeration

Period-1 polytopes are polytopes for which a dilation invariant loses its expected periodic fluctuation and becomes exactly linear or polynomial in the dilation parameter. In the Ehrhart-theoretic usage, a period-1 polytope is one whose lattice-point enumerator is an honest polynomial rather than a quasi-polynomial; in the toric-code usage developed for lattice polytopes, a period-1 polytope is one for which Minkowski length scales exactly by dilation, \(L(tP)=tL(P)\) for all \(t\in \mathbb{Z}^+\). These usages are closely related in spirit but concern different invariants, and the modern literature treats both as structurally important instances of “period collapse” or exact linearization under scaling [2002.03257], [2509.21178], [2206.02276].

## 1. Definitions and competing usages

Two technical meanings of “period-1 polytope” are current.

| Notion | Defining condition | Typical setting |
|---|---|---|
| Ehrhart period-1 | The Ehrhart quasi-polynomial has period sequence \((1,\dots,1)\), so the counting function is a polynomial | Rational or irrational lattice-point enumeration |
| Minkowski-length period-1 | \(L(tP)=tL(P)\) for all \(t\in\mathbb Z^+\) | Lattice polytopes and toric surface codes |

For a rational polytope \(P\subset \mathbb{R}^n\), the Ehrhart function is
\[
\mathrm{ehr}_P(k)=|kP\cap \mathbb{Z}^n|,
\]
and Ehrhart’s theorem states that this is a degree-\(n\) quasi-polynomial,
\[
\mathrm{ehr}_P(k)=c_n(k)k^n+c_{n-1}(k)k^{n-1}+\cdots+c_0(k),
\]
with periodic coefficient functions \(c_i(k)\). The associated period sequence is \((p_0,p_1,\ldots,p_{n-1},1)\), where \(p_i\) is the minimal period of \(c_i\) and the leading coefficient is always constant [2002.03257].

In the toric-code literature, Gibbs, Hogan, Jabbusch, Plute, and Toloczko define a lattice polytope \(P\subset \mathbb{R}^m\) to have period 1 when
\[
L(tP)=tL(P)\text{ for all }t\in\mathbb Z^+,
\]
where \(L(P)\) is the Minkowski length, defined as
\[
L(P)=\max \{r: Q_1+\cdots+Q_r\subseteq P,\ \text{each }Q_i\text{ a primitive segment}\}.
\]
Here “period” refers not to Ehrhart coefficients but to the exact scaling law of Minkowski length under dilation [2509.21178].

## 2. Ehrhart period 1 and polynomial lattice-point enumeration

For integral polytopes, all Ehrhart coefficients are constant, the period sequence is \((1,1,\ldots,1)\), and the Ehrhart function is an ordinary polynomial [2601.22992]. This gives a universal sufficient condition for Ehrhart period 1. More generally, McMullen’s bounds state that the minimal period \(p_i\) of \(c_{P,i}(k)\) divides the \(i\)-index \(q_i\), the minimal integer such that every \(i\)-face of \(q_iP\) contains a lattice point in its affine span; in particular, integral polytopes have all indices equal to \(1\) [2601.22992].

A central phenomenon is Ehrhart period collapse: the Ehrhart quasi-polynomial has period strictly smaller than the denominator predicted by the rational vertex data, possibly equal to \(1\) [2206.02276]. Restricted Birkhoff polytopes supply a prominent example. The Birkhoff polytope \(\mathcal{B}_n\) consists of \(n\times n\) nonnegative real matrices with all row sums and column sums equal to \(1\), and the restricted polytope \(\mathcal{B}_n^k\) is obtained by imposing, for every upright lattice path from \((1,1)\) to \((n,n)\), the inequality that the sum along the path is at most \(k\). There are \(\binom{2n-2}{n-1}\) such inequalities [2206.02276].

For \(n=3\) and \(k=2\), \(\mathcal{B}_3^2\) is not integral: besides five permutation-matrix vertices, it has the rational vertex
\[
\begin{pmatrix}
\frac{1}{2} & 0 & \frac{1}{2}\\
0 & 1 & 0\\
\frac{1}{2} & 0 & \frac{1}{2}
\end{pmatrix}.
\]
Nevertheless,
\[
L(\mathcal{B}_3^2;t)=\frac{1}{12}t^4+\frac{1}{2}t^3+\frac{17}{12}t^2+2t+1
=2\binom{t+3}{4}+\binom{t+2}{2},
\]
so the Ehrhart quasi-polynomial is an honest polynomial [2206.02276]. The proof proceeds through a continuous, piecewise-linear, homogeneous bijection
\[
\rho:\mathcal{B}_n^k\to \mathcal{M}_n^k
\]
that is essentially Pak’s numerical version of the Robinson–Schensted–Knuth correspondence. Although \(\rho\) is not an integral equivalence, it gives a bijection between lattice points in \(t\mathcal{B}_n^k\) and \(t\mathcal{M}_n^k\) for every integer \(t\), and \(\mathcal{M}_n^k\) is integrally equivalent to a Gelfand–Tsetlin polytope whose Ehrhart function is always a polynomial [2206.02276].

## 3. Non-integral and irrational period-1 examples

The class of Ehrhart period-1 polytopes extends beyond integral rational polytopes. Cristofaro-Gardiner, Li, and Stanley study triangles
\[
T_{u,v}=\operatorname{conv}\{(0,0),(1/u,0),(0,1/v)\},
\]
with \(u,v>0\), and characterize when the Ehrhart function is quasi-polynomial or polynomial even when \(u/v\) is irrational [1509.01887].

Their theorem states that if \(u/v\) is irrational, then \(T_{u,v}\) is primitive and pseudo-rational if and only if there exist positive integers \(a,b\) such that
\[
u+v=a,\qquad \frac{1}{u}+\frac{1}{v}=b,
\]
and in this case the period of the Ehrhart quasi-polynomial divides \(a\). The triangle is primitive and pseudo-integral, meaning its Ehrhart function is a polynomial, if and only if these conditions hold and, in addition, either \(a=1\) or \((a,b)\in\{(3,3),(2,4)\}\) [1509.01887]. The same source then records under its explicit examples that the case \(a=1\) is “Not possible” for positive irrational \(u\) and \(v\), while the pairs \((3,3)\) and \((2,4)\) do produce pseudo-integral irrational triangles [1509.01887].

The paper also gives higher-dimensional irrational examples. In dimension \(3\), the simplex with vertices
\[
(0,0,0),\ (2,0,0),\ (0,2+\sqrt{2},0),\ (0,0,2-\sqrt{2})
\]
has polynomial Ehrhart function
\[
f(t)=\frac{1}{3}t^3+2t^2+t+1.
\]
These constructions show that period collapse to \(1\) can occur even for polytopes with infinite denominator, and the same work relates several such examples to symplectic embedding problems, especially to four-dimensional ellipsoids [1509.01887].

## 4. Period sequences, arbitrary periods, and the role of convexity

Ehrhart period 1 is the extreme case of a broader period-sequence problem. For a rational polytope, the coefficient periods \(p_i\) of the Ehrhart quasi-polynomial form the period sequence \((p_0,p_1,\ldots,p_{n-1},1)\), and a major open problem is to determine which quasi-polynomials arise as Ehrhart quasi-polynomials of rational polytopes [2601.22992].

A constructive answer for arbitrary prescribed periods is given for nonconvex polytopal balls. For any positive integers \(p_0,p_1,\ldots,p_{n-1}\), there exists an \(n\)-dimensional rational polytopal ball \(Q^*\), not necessarily convex, such that the period of the coefficient of \(t^i\) in the Ehrhart quasi-polynomial of \(Q^*\) is exactly \(p_i\) for \(0\le i\le n-1\) [2002.03257]. The proof starts from building blocks whose periodicity is confined to a single coefficient: a line segment \(\ell=[-\frac1p,0]\), special pentagons in \(\mathbb R^2\), and the \(i\)-fold pyramid over \(\ell\), which yields an \((i+1)\)-simplex with period sequence \((p,1,\dots,1)\). For each coefficient index \(i\), an \(n\)-dimensional polytope
\[
Q_i:=L_i\cup M\cup R_i
\]
is constructed from products of cyclic polytopes and pyramidal blocks, translated into appropriate halfspaces so that only the degree-\(i\) coefficient has period \(p\) [2002.03257].

The cyclic-polytope input uses F. Liu’s recursion
\[
\mathrm{ehr}_{C_i}(t)=\mathrm{Vol}(C_i)t^i+\mathrm{ehr}_{C_{i-1}}(t),
\]
and for each \(Q_i\),
\[
\mathrm{ehr}_{Q_i}(t)=\mathrm{Vol}(C_i)t^i\,\mathrm{ehr}_{A^{n-i-1}(\ell)}(t).
\]
The final object is the union
\[
Q^*=\bigcup_i Q_i,
\]
with sufficiently large translations to keep the glued polytopes in convex position [2002.03257].

This construction removes the divisibility restrictions that occur in the convex case. Earlier work of Beck, Sam, and Woods showed that convex polytopes can realize any sequence \((p_0,\dots,p_{n-1},1)\) subject to
\[
p_{n-1}\mid p_{n-2}\mid \dots \mid p_0,
\]
whereas arbitrary periods become possible only after allowing nonconvex polytopal balls [2002.03257]. Period 1 for all coefficients is therefore one endpoint of a much larger design space.

Some summaries formulate integrality as the defining source of period-1 Ehrhart behavior, while explicit period-collapse families exhibit polynomial Ehrhart functions for non-integral rational and even irrational polytopes [2601.22992], [2206.02276], [1509.01887]. This suggests that integrality is a robust sufficient condition, but that any converse must be read with careful attention to the hypotheses of the specific subclass under discussion.

## 5. Minkowski-length period 1 and toric surface codes

In toric surface code theory, period-1 behavior concerns Minkowski length rather than Ehrhart coefficients. A lattice polytope \(P\subset \mathbb R^m\) has period \(1\) if
\[
L(tP)=tL(P)\quad\text{for all }t\in \mathbb Z^+,
\]
and this property is preserved under lattice equivalence by \(\operatorname{AGL}(m,\mathbb Z)\) [2509.21178]. For general polytopes one has only \(L(tP)\ge tL(P)\); equality is the special feature singled out by the term.

In dimension \(2\), the class of period-1 polytopes is described up to lattice equivalence in terms of maximal Minkowski decompositions. Primitive line segments are period-1 polytopes, simplices \(t\Delta\) are period-1 polytopes, and so are rectangles, parallelograms, and zonotopes built from primitive segments. The unique exception among maximal decomposition summands is the exceptional triangle
\[
\mathfrak 0=\operatorname{conv}\{(0,0),(1,2),(2,1)\},
\]
for which
\[
L(t\mathfrak 0)=t+\left\lfloor \frac{t}{2}\right\rfloor.
\]
The paper summarizes the planar classification by stating that line segments and simplices are period-1 polytopes and that the exceptional triangle is the only polytope which can appear in a maximal decomposition that has period greater than \(1\) [2509.21178].

This exact scaling law has direct coding-theoretic consequences. For toric codes associated to a polytope \(P\subseteq [0,q-2]^2\), the block length is \(n=(q-1)^2\) and the dimension is \(k=\#(P\cap \mathbb Z^2)\). For
\[
P\approx m[0,e_1]+n[0,e_2]+\ell[0,e_1+e_2],\qquad 0<n\le m\le \ell,
\]
and under the additional conditions \(q-1-M\sqrt q>\operatorname{Area}(P)\) and the existence of subgroups of certain order in \(\mathbb F_q^\times\), the minimum distance is
\[
d(C_P)=(q-1)^2-L(P)(q-1)+\ell(m+n),
\]
where \(L(P)=m+n+\ell\) [2509.21178]. The proof uses maximal decompositions into primitive segments, explicit products
\[
f(x,y)=\prod_{i=1}^m(x-a_i)\prod_{j=1}^n(y-b_j)\prod_{k=1}^{\ell}(xy-c_k),
\]
and inclusion-exclusion to count zeros.

A second family,
\[
P\approx \ell\Delta+\ell[0,e_1]+\ell[0,e_2],
\]
is analyzed by the method of Little and Schwarz. Here a length-\(\ell\) staircase configuration and the Vandermonde matrix \(V(P;S)\) are used to show linear independence of evaluations, leading to
\[
d(C_P)=(q-1-2\ell)(q-1-\ell).
\]
In this setting, period-1 behavior makes the minimum-distance formulas tractable because the combinatorics of maximal Minkowski decompositions scale without periodic deviation [2509.21178].

## 6. Graph-associated period-1 polytopes and conceptual scope

Another Ehrhart-theoretic family with period \(1\) arises from \(\{1,3\}\)-graphs. A \(\{1,3\}\)-graph is a graph whose nodes have degree \(1\) or \(3\). Liu and Osserman associated polytopes \(P_G\) and \(Q_G\) to such graphs, and the period behavior depends sharply on the graph class. For connected \(\{1,3\}\)-graphs, the Ehrhart quasi-polynomial \(L_G^Q(t)\) has period \(1\) if and only if \(G\) is a \(\{1,3\}\)-tree; for \(P_G\), period \(1\) occurs only in trivial cases, and for trees \(P_T\) has period exactly \(2\) [2104.11025].

The geometric reason is explicit. For a \(\{1,3\}\)-tree \(T\), \(Q_T\) is a \(0/1\) polytope: its vertices are exactly the characteristic vectors of collections of disjoint leaf-paths in \(T\). Hence \(Q_T\) is integral, and Ehrhart’s theorem implies that \(L_T^Q(t)\) is a polynomial [2104.11025]. For the three-edge tree, the paper gives
\[
L_T^Q(t)=\frac{1}{6}t^3+t^2+\frac{11}{6}t+1,
\]
whereas
\[
L_T^P(t)=\frac{1}{24}t^3+\frac{1}{4}t^2+
\begin{cases}
\frac{5}{6}t+1,& t\text{ even}\\
\frac{11}{24}t+\frac{1}{4},& t\text{ odd}
\end{cases}
\]
has period \(2\) [2104.11025].

Across these examples, period-1 behavior is best understood not as a single rigid class but as a family of exact-scaling phenomena attached to different invariants. In Ehrhart theory it includes integral polytopes, restricted Birkhoff polytopes, certain irrational triangles and higher-dimensional simplices, and the \(Q\)-polytopes of \(\{1,3\}\)-trees [2206.02276], [1509.01887], [2104.11025]. In toric-code theory it identifies lattice polytopes whose Minkowski length scales linearly, with the exceptional triangle as the canonical obstruction in dimension \(2\) [2509.21178]. The common theme is the disappearance of periodic fluctuation under dilation, but the mechanism may be integrality, a piecewise-linear bijection such as RSK, a special irrational parametrization, or the exclusion of particular decomposition summands.

Source: https://www.emergentmind.com/topics/period-1-polytopes