order polynomial

{{Distinguish|Order of a polynomial (disambiguation){{!}}Order of a polynomial}}

The order polynomial is a polynomial studied in mathematics, in particular in algebraic graph theory and algebraic combinatorics. The order polynomial counts the number of order-preserving maps from a poset to a chain of length n. These order-preserving maps were first introduced by Richard P. Stanley while studying ordered structures and partitions as a Ph.D. student at Harvard University in 1971 under the guidance of Gian-Carlo Rota.

Definition

Let P be a finite poset with p elements denoted x,y \in P, and let [n]=\{1<2<\ldots be a chain n elements. A map \phi: P \to [n] is order-preserving if x \leq y implies \phi(x) \leq \phi(y). The number of such maps grows polynomially with n, and the function that counts their number is the order polynomial \Omega(n)=\Omega(P,n).

Similarly, we can define an order polynomial that counts the number of strictly order-preserving maps \phi: P \to [n], meaning x < y implies \phi(x) < \phi(y). The number of such maps is the strict order polynomial \Omega^{\circ}\! (n)=\Omega^{\circ}\! (P,n).{{cite book|last1=Stanley|first1=Richard P.|title=Ordered structures and partitions|date=1972|publisher=American Mathematical Society|location=Providence, Rhode Island}}

Both \Omega(n) and \Omega^\circ\!(n) have degree p. The order-preserving maps generalize the linear extensions of P, the order-preserving bijections \phi:P\stackrel{\sim}{\longrightarrow}[p]. In fact, the leading coefficient of \Omega(n) and \Omega^\circ\!(n) is the number of linear extensions divided by p!.

== Examples ==

Letting P be a chain of p elements, we have

\Omega(n) = \binom{n+p-1}{p} = \left(\!\left({n\atop p}\right)\!\right) and \Omega^\circ(n) = \binom{n}{p}.
There is only one linear extension (the identity mapping), and both polynomials have leading term \tfrac 1{p!}n^p.

Letting P be an antichain of p incomparable elements, we have \Omega(n) =\Omega^\circ(n) = n^p . Since any bijection \phi:P\stackrel{\sim}{\longrightarrow}[p] is (strictly) order-preserving, there are p! linear extensions, and both polynomials reduce to the leading term \tfrac {p!}{p!}n^p = n^p.

Reciprocity theorem

There is a relation between strictly order-preserving maps and order-preserving maps:{{cite journal |last1=Stanley |first1=Richard P. |title=A chromatic-like polynomial for ordered sets |date=1970 |journal=Proc. Second Chapel Hill Conference on Combinatorial Mathematics and Its Appl. |pages=421–427}}

:\Omega^\circ(n) = (-1)^

P
\Omega(-n).

In the case that P is a chain, this recovers the negative binomial identity. There are similar results for the chromatic polynomial and Ehrhart polynomial (see below), all special cases of Stanley's general Reciprocity Theorem.{{Cite book|title=Enumerative combinatorics. Volume 1|last1=Stanley | first1=Richard P.|date=2012|publisher=Cambridge University Press|isbn=9781139206549|edition=2nd|location=New York|chapter=4.5.14 Reciprocity theorem for linear homogeneous diophantine equations|oclc=777400915}}

Connections with other counting polynomials

=Chromatic polynomial =

The chromatic polynomial P(G,n)counts the number of proper colorings of a finite graph G with n available colors. For an acyclic orientation \sigma of the edges of G, there is a natural "downstream" partial order on the vertices V(G) implied by the basic relations u > v whenever u \rightarrow v is a directed edge of \sigma. (Thus, the Hasse diagram of the poset is a subgraph of the oriented graph \sigma.) We say \phi: V(G) \rightarrow [n] is compatible with \sigma if \phi is order-preserving. Then we have

: P(G,n) \ =\ \sum_{\sigma} \Omega^\circ\!(\sigma,n),

where \sigma runs over all acyclic orientations of G, considered as poset structures.{{cite journal|last1=Stanley|first1=Richard P.|title=Acyclic orientations of graphs|journal=Discrete Mathematics|date=1973|volume=5|issue=2|pages=171–178|doi=10.1016/0012-365X(73)90108-8|doi-access=}}

= Order polytope and Ehrhart polynomial =

{{main|Order polytope}}

The order polytope associates a polytope with a partial order. For a poset P with p elements, the order polytope O(P) is the set of order-preserving maps f:P\to [0,1], where [0,1] = \{ t\in\mathbb{R}\mid 0\leq t\leq 1\} is the ordered unit interval, a continuous chain poset.{{cite journal |first1=Alexander |last1=Karzanov |first2=Leonid |last2=Khachiyan |title=On the conductance of Order Markov Chains |journal=Order |volume=8 |pages=7–15 |year=1991 |doi=10.1007/BF00385809|doi-access=|s2cid=120532896 }}{{cite journal |first1=Graham |last1=Brightwell |first2=Peter |last2=Winkler |title=Counting linear extensions |journal=Order |volume=8 |pages=225–242 |year=1991 |issue=3 |doi=10.1007/BF00383444|doi-access=|s2cid=119697949 }} More geometrically, we may list the elements P=\{x_1,\ldots,x_p\}, and identify any mapping f:P\to\mathbb R with the point (f(x_1),\ldots,f(x_p))\in \mathbb{R}^{p}; then the order polytope is the set of points (t_1,\ldots,t_p)\in [0,1]^p with t_i \leq t_j if x_i \leq x_j.{{Cite journal|last=Stanley|first=Richard P.|date=1986|title=Two poset polytopes|journal=Discrete & Computational Geometry|volume=1|pages=9–23|doi=10.1007/BF02187680|doi-access=free}}

The Ehrhart polynomial counts the number of integer lattice points inside the dilations of a polytope. Specifically, consider the lattice L=\mathbb{Z}^n and a d-dimensional polytope K\subset\mathbb{R}^d with vertices in L; then we define

:L(K,n) = \#(nK \cap L),

the number of lattice points in nK, the dilation of K by a positive integer scalar n. Ehrhart showed that this is a rational polynomial of degree d in the variable n, provided K has vertices in the lattice.{{Cite book|title=Computing the continuous discretely|title-link= Computing the Continuous Discretely |last1=Beck|first1=Matthias|last2=Robins|first2=Sinai|publisher=Springer|year=2015|isbn=978-1-4939-2968-9|location=New York|pages=64–72}}

In fact, the Ehrhart polynomial of an order polytope is equal to the order polynomial of the original poset (with a shifted argument):{{cite journal|last=Linial|first=Nathan|year=1984|title=The information-theoretic bound is good for merging|journal=SIAM J. Comput.|volume=13|issue=4|pages=795–801|doi=10.1137/0213049}}
{{cite journal

| last1 = Kahn | first1 = Jeff

| last2 = Kim | first2 = Jeong Han

| doi = 10.1006/jcss.1995.1077

| issue = 3

| journal = Journal of Computer and System Sciences

| pages = 390–399

| title = Entropy and sorting.

| volume = 51

| year = 1995| doi-access = free

}}

L(O(P),n)\ =\ \Omega(P,n{+}1).
This is an immediate consequence of the definitions, considering the embedding of the (n{+}1)-chain poset [n{+}1]=\{0<1<\cdots.

References