Combinatorial Actions, Orbit Averages, and Sage Implementationtroby/homomesySageOrsay2013.pdf ·...

Combinatorial Actions, Orbit Averages, andSage Implementation

Tom Roby (University of Connecticut)Darij Grinberg (MIT)

Describing joint research with Jim Propp

Sage Days 72 Orsay, France

28 June 2013

Slides for this talk are available online (or will be soon) at

http://www.math.uconn.edu/~troby/research.html


Abstract

We consider a variety of combinatorial actions on finite sets, e.g., cyclicrotation of binary strings, promotion of Young tableaux and rowmotion onorder ideals of partially ordered sets. We identify a particular phenomenoncalled “homomesy” appearing in many unrelated combinatorial contexts:namely that the average value of some natural statistic over each orbit is thesame as the average over the entire set. Viewing these actions as products of“toggle operations” allows us to see how some of these actions are related andto extend much of this picture more broadly. In particular, we can generalizethe operations of rowmotion and promotion (in Striker and Williams’terminology) on order ideals in a poset to (1) the order polytope of a poset(the continuous piecewise-linear category), and (2) to the collection of mapsfrom a poset P to rational functions in |P| variables (the birational category).For this latter category, we have developed Sage code that (1) computes therowmotion operator; (2) checks whether this operator appears to have finiteorder; (3) helps check for homomesy of particular statistics under the action ofthe operator. Although we have proofs of homomesy for a number of specialcases, much remains to be done even at the combinatorial level of order idealson posets. For the two other categories most of what we have discovered viacomputations is still conjectural.

Acknowledgments

This talk largely discusses recent work with Jim Propp, includingideas and results from Arkady Berenstein, David Einstein,Shahrzad Haddadan, Jessica Striker, and Nathan Williams.

Mike LaCroix wrote fantastic postscript code to generateanimations and pictures that illustrate our maps operating on orderideals on products of chains. Jim Propp created many of the otherpictures and slides that are used here.

Thanks also to Omer Angel, Drew Armstrong, Anders Björner,Barry Cipra, Karen Edwards, Robert Edwards, Svante Linusson,Vic Reiner, Richard Stanley, Ralf Schiffler, Hugh Thomas, PeteWinkler, and Ben Young.

Overview

Rotation of bit-strings;

Unexpected averaging properties: homomesic statistics;

Suter’s dihedral symmetries in Young’s lattice;

Rowmotion and Promotion actions on antichains and orderideals of posets;

Homomesic statistics for actions in [a]× [b];Generalizing to other actions on order polytopes and rationalfunctions;

Demonstration of Sage code; and

Wish list for Sage implementation.

Please interrupt with questions!

Example 1: Rotation of bit-strings

Let S denote the set of length n binary strings with exactly k 1’s.and set τ := CR : S → S by b = b1b2 · · · bn 7→ bnb1b2 · · · bn−1(cyclic shift). Let ϕ(b) = #inversions(b) = #{i < j : bi > bj}.Then over any τ -orbit O we have:

1

#O∑s∈O

ϕ(s) =k(n − k)

2=

1

#S

∑s∈S

ϕ(s).

EG: n = 4, k = 2 gives us two orbits:

0011 0101

1001 10101100 010101100011



1

#O∑s∈O

ϕ(s) =k(n − k)

2=

1

#S

∑s∈S

ϕ(s).


0011 0101

1001 7→ 2 1010 7→ 31100 7→ 4 0101 7→ 10110 7→ 20011 7→ 0



1

#O∑s∈O

ϕ(s) =k(n − k)

2=

1

#S

∑s∈S

ϕ(s).


0011 0101

1001 7→ 2 1010 7→ 31100 7→ 4 0101 7→ 10110 7→ 2 AVG = 42 = 20011 7→ 0

AVG = 84 = 2

More rotation

EG: n = 6, k = 2 gives us three orbits:

000011 000101 001001

100001 100010 100100110000 010001 010010011000 101000 001001001100 010100000110 001010000011 000101

More rotation


000011 000101 001001

100001 7→ 4 100010 7→ 5 100100 7→ 6110000 7→ 8 010001 7→ 3 010010 7→ 4011000 7→ 6 101000 7→ 7 001001 7→ 2001100 7→ 4 010100 7→ 5000110 7→ 2 001010 7→ 3000011 7→ 0 000101 7→ 1

More rotation


000011 000101 001001

100001 7→ 4 100010 7→ 5 100100 7→ 6110000 7→ 8 010001 7→ 3 010010 7→ 4011000 7→ 6 101000 7→ 7 001001 7→ 2001100 7→ 4 010100 7→ 5000110 7→ 2 001010 7→ 3000011 7→ 0 000101 7→ 1

AVG = 246 = 4 AVG =246 = 4 AVG =

123 = 4

More rotation


000011 000101 001001

100001 7→ 4 100010 7→ 5 100100 7→ 6110000 7→ 8 010001 7→ 3 010010 7→ 4011000 7→ 6 101000 7→ 7 001001 7→ 2001100 7→ 4 010100 7→ 5000110 7→ 2 001010 7→ 3000011 7→ 0 000101 7→ 1

AVG = 246 = 4 AVG =246 = 4 AVG =

123 = 4

We know two simple ways to prove this: one can show pictoriallythat the value of the sum doesn’t change when you mutate b(replacing a 01 somewhere in b by 10 or vice versa), or one canwrite the number of inversions in b as

∑i

Main definition: Homomesic

MAIN DEF: Given an (invertible) action τ on a finite set ofobjects S , call a statistic ϕ : S → C homomesic with respect to(S , τ) iff the average of ϕ over each τ -orbit O is the same for allO, i.e., 1

#O∑s∈O

ϕ(s) does not depend on the choice of O.

Equivalently: the average of ϕ over each τ -orbit O is the same asthe average over the entire set S :

1

#O∑s∈O

ϕ(s) =1

#S

∑s∈S

ϕ(s).

So in looking for homomesic statistics we can compute what theaverage should be before checking whether a statistic ishomomesic.

Semi-Standard Young Tableaux

Fix a positive integer N. A semi-standard Young tableau(SSYT) of shape λ and ceiling N is a labeling of the cells of theYoung diagram of a partition λ with numbers from 1, 2, . . . ,Nwhich increases weakly along each row and strictly along eachcolumn. For example,

1 1 2 4

2 3 4

4 4 ,

1 1 1 2 2 5

4 4 4 5

5 5 6 but not

1 1 2 2 3 3

2 3 4 4

3 4 4

The weight vector α(T ) = (α1, α2, . . . , αN) is given byαi := αi (T ) = #occurrences of i in T . EG, the weight vectors forthe two tableaux above are (2, 2, 1, 4) and (3, 2, 0, 3, 4, 1).

We let SSYT (λ,N) denote the set of all semi-standard Youngtableaux whose entries lie within [N] = {1, 2, . . .N}.

Bender-Knuth involutions

A standard method for proving combinatorially that Schurfunctions are symmetric is to use Bender-Knuth involutions.Given T ∈ SSYT (λ,N) and i ∈ [N − 1], consider all the i ’s thatappear above an i + 1 in the same column, and all the i + 1’s thatappear below an i in the same column to be married and theremainder free. Then in each row with r free i ’s and s free i + 1’s,βi replaces these with s free i ’s and r free i + 1’s.

ii i i i︸︷︷︸ i + 1 i + 1 i + 1 i + 1︸︷︷︸ i + 1

i + 1 i + 1 r=2 s=4

i7→ i i i i i i︸︷︷︸ i + 1 i + 1︸︷︷︸ i + 1

i + 1 i + 1 s=4 r=2

Promotion of SSYT via Bender-Knuth involutions

Define the following operator on SSYT (λ,N):

∂ := βN−1 ◦ βN−2 ◦ · · · ◦ β2 ◦ β1 ,

the composition of all BK involutions in order. By a result ofGansner [Gan80], this operator coincides with Schützenberger’spromotion operator. Then for all i ∈ [N], the weight vectorcoordinate αi (T ) is homomesic w.r.t. ∂ acting on SSYT (λ,N).

EG: Let N = 5 and T =1 1 1 2 2 3 3 3 42 2 3 3 4 4 53 4 4 5

Then the content vectors α = [α1, α2, . . . , αN ] that arise as wesuccessively apply βi ’s behave as follows, starting from [3, 4, 6, 5, 2]:

[4, 3, 6, 5, 2] [6, 4, 5, 2, 3] [5, 6, 2, 3, 4] [2, 5, 3, 4, 6] [3, 2, 4, 6, 5][4, 6, 3, 5, 2] [6, 5, 4, 2, 3] [5, 2, 6, 3, 4] [2, 3, 5, 4, 6] [3, 4, 2, 6, 5][4, 6, 5, 3, 2] [6, 5, 2, 4, 3] [5, 2, 3, 6, 4] [2, 3, 4, 5, 6] [3, 4, 6, 2, 5][4, 6, 5, 2, 3] [6, 5, 2, 3, 4] [5, 2, 3, 4, 6] [2, 3, 4, 6, 5] [3, 4, 6, 5, 2]

An easy homomesy for Promotion

From previous slide: The content vectors α = [α1, α2, . . . , αN ]that arise as we successively apply βi ’s behave as follows (workingdown columns), starting from [3, 4, 6, 5, 2]:

[4, 3, 6, 5, 2] [6, 4, 5, 2, 3] [5, 6, 2, 3, 4] [2, 5, 3, 4, 6] [3, 2, 4, 6, 5][4, 6, 3, 5, 2] [6, 5, 4, 2, 3] [5, 2, 6, 3, 4] [2, 3, 5, 4, 6] [3, 4, 2, 6, 5][4, 6, 5, 3, 2] [6, 5, 2, 4, 3] [5, 2, 3, 6, 4] [2, 3, 4, 5, 6] [3, 4, 6, 2, 5][4, 6, 5, 2, 3] [6, 5, 2, 3, 4] [5, 2, 3, 4, 6] [2, 3, 4, 6, 5] [3, 4, 6, 5, 2]

Note that each iteration of promotion (going down each fullcolumn) applies a cyclic shift of the content vector. This is enoughto prove that αi (T ) is homomesic, even though we don’t knowhow long the orbit is. But it must be a multiple of 5 for thecontent vector to return to where it started. So over any orbit,each content value visits each position equally often.

However, we conjecture more interesting homomesies which we areabout to test more thoroughly using Sage.

A small example of promotion

(taken from J. Striker and N. Williams, Promotion andRowmotion, European J. Combin. 33 (2012), no. 8, 1919–1942;http://arxiv.org/abs/1108.1172):

http://arxiv.org/abs/1108.1172

A small example of promotion: centrally symmetric sums

Promotion of Semi-Standard Young Tableaux: homomesies

Conjecture

Let S be the set of Semi-Standard Young Tableaux of rectangularshape λ and ceiling N. If c and c ′ are opposite cells, i.e., c and c ′

are related by 180-degree rotation about the center (note: the casec = c ′ is permitted when λ is odd-by-odd), and ϕ(T ) denotes thesum of the numbers in cells c and c ′, then ϕ is homomesic withrespect to (S , ∂) with average value N + 1.

Rectangular shapes are one of the few shapes where the order ofpromotion on the set of SYT is small, i.e., n or 2n. Striker &Williams note that the order of promotion on SYT of shape (8, 6)is 7,554,844,752.

Basics of partially ordered sets

A partially ordered set (or poset) is a set P with a binary relation≤ that satisfies the following properties for every a, b, c ∈ P:(a) reflexive (a ≤ a),(b) antisymmetric (a ≤ b and b ≤ a =⇒ a = b), and(c) transitive (a ≤ b and b ≤ c =⇒ a ≤ c). All our posets will befinite and can be represented by directed graphs of minimalcovering relations, as I’ll draw on the board.

An antichain A ⊆ P is any set of unrelated elements. Thecollection of all antichains is denoted A(P).An order ideal or down-set I ⊆ P is any set with the propertythat z ∈ I and x ≤ z in P =⇒ x ∈ I . The collection of all orderideals is denoted J(P).

There is an easy bijection between A(P) and J(P), given by takingA ∈ A(P) to the order ideal generated by A, and conversely takingI ∈ J(P) to the maximal elements of I .



An antichain A ⊆ P is any set of unrelated elements. Thecollection of all antichains is denoted A(P).

An order ideal or down-set I ⊆ P is any set with the propertythat z ∈ I and x ≤ z in P =⇒ x ∈ I . The collection of all orderideals is denoted J(P).




An antichain A ⊆ P is any set of unrelated elements. Thecollection of all antichains is denoted A(P).An order ideal or down-set I ⊆ P is any set with the propertythat z ∈ I and x ≤ z in P =⇒ x ∈ I . The collection of all orderideals is denoted J(P).


Rowmotion: an invertible operation on antichains

Given A ∈ A(P), let τ(A) be the set of minimal elements of thecomplement of the order ideal generated by A.τ is invertible since it is a composition of three invertibleoperations:

antichains←→ down-sets←→ up-sets←→ antichains

We can also view this as an invertible action τ on J(P), the set oforder ideals of P, via the above isomorphism between A(P) andJ(P); in other words, perform the above steps in the order 2,3,1.

This map and its inverse have been considered with varyingdegrees of generality, by many people more or less independently(using a variety of nomenclatures and notations): Duchet, Brouwerand Schrijver, Cameron and Fon Der Flaass, Fukuda, Panyushev,Rush and Shi, and Striker and Williams [SW12]. Following[SW12], we call this rowmotion.

An example

1. Saturate downward

2. Complement

3. Take minimal element(s)

1−→ 2−→ 3−→

1

Example in lattice cell form

Viewing the elements of the poset as squares, we would map:

Area = 8

X X

τ−→τ−→

Area = 10

X

X X

Panyushev’s conjecture

Let ∆ be a reduced irreducible root system in Rn. (Picture comingsoon!)Choose a system of positive roots and make it a poset of rank n bydecreeing that y covers x iff y − x is a simple root.Conjecture (Conjecture 2.1(iii) in D.I. Panyushev, On orbits ofantichains of positive roots, European J. Combin. 30 (2009),586-594): Let O be an arbitrary τ -orbit. Then

1

#O∑A∈O

#A =n

2.

In our language, the cardinality statistic is homomesic with respectto the action of rowmotion on antichains in root posets.

Panyushev’s Conjecture 2.1(iii) (along with much else) was provedby Armstrong, Stump, and Thomas in their article A uniformbijection between nonnesting and noncrossing partitions,http://arxiv.org/abs/1101.1277.

http://arxiv.org/abs/1101.1277

Picture of root posets

Here are the classes of posets included in Panyushev’s conjecture.

(Graphic courtesy of Striker-Williams.)

Panyushev’s conjecture: The An case, n = 2

Here we have just an orbit of size 2 and an orbit of size 3:

0 2 1

1 1

1

Within each orbit, the average antichain has cardinality n/2 = 1.

The case A3.

Here’s an example orbit taken from [AST] for the A3 root poset:

For A3 this action has three orbits (sized 2, 4, and 8), and theaverage cardinality of an antichain is

1

8(2 + 1 + 1 + 2 + 2 + 1 + 1 + 2) =

3

2=

n

2

Antichains in [a]× [b]: cardinality is homomesic

A simpler-to-prove phenomenon of this kind concerns the poset[a]× [b] (where [k] denotes the linear ordering of {1, 2, . . . , k}):

Theorem (Propp, R.)

Let O be an arbitrary τ -orbit in A([a]× [b]). Then

1

#O∑A∈O

#A =ab

a + b.

This is an easy consequence of unpublished work of Hugh Thomasbuilding on earlier work of Richard Stanley: see the last paragraphof section 2 of R. Stanley, Promotion and evacuation,http://www.combinatorics.org/ojs/index.php/eljc/

article/view/v16i2r9 .

http://www.combinatorics.org/ojs/index.php/eljc/article/view/v16i2r9http://www.combinatorics.org/ojs/index.php/eljc/article/view/v16i2r9

Antichains in [a]× [b]: the case a = b = 2

Here we have an orbit of size 2 and an orbit of size 4:

Within each orbit, the average antichain has cardinalityab/(a + b) = 1.

0 1 2 1

1 1

1

Antichains in [a]× [b]: fiber-cardinality is homomesic

0 0 0 1 1 1 1 0

1 0 0 1

1

Within each orbit, the average antichain has1/2 a green element and 1/2 a blue element.

Antichains in [a]× [b]: fiber-cardinality is homomesic

For (i , j) ∈ [a]× [b], and A an antichain in [a]× [b], let 1i ,j(A) be1 or 0 according to whether or not A contains (i , j).

Also, let fi (A) =∑

j∈[b] 1i ,j(A) ∈ {0, 1} (the cardinality of theintersection of A with the fiber {(i , 1), (i , 2), . . . , (i , b)} in[a]× [b]), so that #A = ∑i fi (A).Likewise let gj(A) =

∑i∈[a] 1i ,j(A), so that #A =

∑j gj(A).

Theorem (Propp, R.)

For all i , j ,

1

#O∑A∈O

fi (A) =b

a + band

1

#O∑A∈O

gj(A) =a

a + b.

The indicator functions fi and gj are homomesic under τ , eventhough the indicator functions 1i ,j aren’t.

Antichains in [a]× [b]: centrally symmetric homomesies

Theorem (Propp, R.)

In any orbit, the number of A that contain (i , j) equals the numberof A that contain the opposite element(i ′, j ′) = (a + 1− i , b + 1− j).

That is, the function 1i ,j − 1i ′,j ′ is homomesic under τ , withaverage value 0 in each orbit.

Linearity

Useful triviality: every linear combination of homomesies is itselfhomomesic.

E.g., consider the adjusted major index statistic defined byamaj(A) =

∑(i ,j)∈A(i − j).

Propp and R. proved that amaj is homomesic under τby writing it as a linear combination of the functions 1i ,j − 1i ′,j ′ .Haddadan gave a simpler proof,writing amaj as a linear combination of the functions fi and gj .

Question: Are there other homomesic combinations of theindicator functions 1i ,j (with (i , j) ∈ [a]× [b]),linearly independent of the functions fi , gj , and 1i ,j − 1i ′,j ′?

Ideals in [a]× [b]: cardinality is homomesic

As we’ve seen, one can view rowmotion as acting either onantichains (A(P)) or on order ideals (J(P)); we denote the lattermap τ . It turns out that the cardinality of the order ideal is alsohomomesic with respect to rowmotion on [a]× [b].

Theorem (Propp, R.)

Let O be an arbitrary τ -orbit in J([a]× [b]). Then

1

#O∑I∈O

#I =ab

2.

It’s worth noting even though there’s a strong connection betweenthe rowmotion map on antichains and on order ideals, that thehomomesy situation could be quite different.

One action, two vector spaces

The map τ is “the same” as τ in the sense that the standardbijection from A(P) to J(P) (downward saturation) makes thefollowing diagram commute:

A(P) τ−→ A(P)↓ ↓

J(P)τ−→ J(P)

However, the bijection from A(P) to J(P) does not carry thevector space generated by the functions 1i ,j to the vector spacegenerated by the functions 1i ,j in a linear way.

So the homomesy situation for τ : J(P)→ J(P) could be(and, as we’ll see, is) different from the homomesy situation forτ : A(P)→ A(P).

Rowmotion on [4]× [2] A

Rowmotion on [4]× [2] A

1

Area = 0

2

Area = 1

3

Area = 3

4

Area = 5

5

Area = 7

6

Area = 8

(0+1+3+5+7+8) / 6 = 4

Rowmotion on [4]× [2] B

Rowmotion on [4]× [2] B

1

Area = 2

2

Area = 4

3

Area = 6

4

Area = 6

5

Area = 4

6

Area = 2

(2+4+6+6+4+2) / 6 = 4

Rowmotion on [4]× [2] C

Rowmotion on [4]× [2] C

1

Area = 3

2

Area = 5

3

Area = 4

4

Area = 3

5

Area = 5

6

Area = 4

(3+5+4+3+5+4) / 6 = 4

Ideals in [a]× [b]: the case a = b = 2

Again we have an orbit of size 2 and an orbit of size 4:

Within each orbit, the average order ideal has cardinality ab/2 = 2.

0 1 3 4

2 2

1

Ideals in [a]× [b]: file-cardinality is homomesic

We also have homomesies for more refined statistics than #I .

0 0 0 0 1 0 1 1 1 1 2 1

1 1 0 0 1 1

1

Within each orbit, the average order ideal has1/2 a violet element, 1 red element, and 1/2 a brown element.

Ideals in [a]× [b]: file-cardinality is homomesic

For 1− b ≤ k ≤ a− 1, define the kth file of [a]× [b] as

{(i , j) : 1 ≤ i ≤ a, 1 ≤ j ≤ b, i − j = k}.

For 1− b ≤ k ≤ a− 1, let hk(I ) be the number of elements of I inthe kth file of [a]× [b], so that #I = ∑k hk(I ).Theorem (Propp, R.)

For every τ -orbit O in J([a]× [b]),

1

#O∑I∈O

hk(I ) =

{(a−k)ba+b if k ≥ 0

a(b+k)a+b if k ≤ 0.

Ideals in [a]× [b]: centrally symmetric homomesies

Given (i , j) ∈ [a]× [b], and I an ideal in [a]× [b], define theindicator function 1i ,j(I ) to be 1 or 0 according to whether or not Icontains (i , j).

Write (i ′, j ′) = (a + 1− i , b + 1− j), the point opposite (i , j) in theposet.

Theorem (Propp, R.)

1i ,j + 1i ′,j ′ is homomesic under τ .

The two vector spaces, compared

In the space associated with antichains:fiber-cardinalities andcentrally symmetric differences

are homomesic.

In the space associated with order ideals:file-cardinalities andcentrally symmetric sums

are homomesic.

Note that the the discovery of these homomesies was driven bycalculation, and the project of generalizing these results to otherposets will clearly be aided by computer-assisted search. Somework has already been done along these lines in a differentcomputer algebra package. . .

Toggling

In their 1995 article Orbits of antichains revisited , European J.Combin. 16 (1995), 545–554, Cameron and Fon-der-Flaass give analternative description of τ .

Given I ∈ J(P) and x ∈ P, let τx(I ) = I4{x} (symmetricdifference) provided that I4{x} is an order ideal of P; otherwise,let τx(I ) = I .

We call the involution τx “toggling at x”.

The involutions τx and τy commute unless x covers y or y covers x .

An example

1. Toggle the top element

2. Toggle the left element

3. Toggle the right element

4. Toggle the bottom element

1−→ 2−→ 3−→ 4−→

1

Toggling from top to bottom

Theorem ([CF95])

Let x1, x2, . . . , xn be any order-preserving enumeration of theelements of the poset P. Then the action on J(P) given by thecomposition τx1 ◦ τx2 ◦ · · · ◦ τxn coincides with the action of τ .

In the particular case P = [a]× [b], we can enumerate Prank-by-rank; that is, we can list the (i , j)’s in order of increasingi + j .

Note that all the involutions coming from a given rank of Pcommute with one another, since no two of them are in a coveringrelation.

Striker and Williams refer to τ (and τ) as rowmotion, since forthem, “row” means “rank”.

Toggling from side to side

Recall that a file in P = [a]× [b] is the set of all (i , j) ∈ P withi − j equal to some fixed value k .

Note that all the involutions coming from a given file commutewith one another, since no two of them are in a covering relation.

It follows that for any enumeration x1, x2, . . . , xn of the elements ofthe poset [a]× [b] arranged in order of increasing i − j , the actionon J(P) given by τx1 ◦ τx2 ◦ · · · ◦ τxn doesn’t depend on whichenumeration was used.

Striker and Williams call this well-defined composition promotion,and denote it by ∂, since for two-rowed tableaux it can be relatedto Schützenberger’s promotion on SYT, described earlier.

Promoting ideals in [a]× [b]: the case a = b = 2

Again we have an orbit of size 2 and an orbit of size 4:

0 2 4 2

1 3

1

J([a]× [b]): cardinality is homomesic under promotion

Theorem (Propp, R.)

Let O be an arbitrary orbit in J([a]× [b]) under the action ofpromotion ∂. Then

1

#O∑I∈O

#I =ab

2.

The result about cyclic rotation of binary words discussed earlierturns out to be a special case of this.

J([a]× [b]): file-cardinality is homomesic under promotion

For 1− b ≤ k ≤ a− 1, let fk(I ) be the number of elements of I inthe kth file of [a]× [b], so that #I = ∑k fk(I ).Theorem

If O is any ∂-orbit in J([a]× [b]),

1

#O∑I∈O

fk(I ) =

{(a−k)ba+b if k ≥ 0

a(b+k)a+b if k ≤ 0.

A([a]× [b]) under promotion

Cardinality of antichains is not homomesic under promotion.although the antipodal functions 1i ,j − 1i ′,j ′ are.

Root posets of type A: antichains

Recall that, by the Armstrong-Stump-Thomas theorem, thecardinality of antichains is homomesic under the action ofrowmotion, where the poset P is a root poset of type An.E.g., for n = 2:

Antichain-cardinality is homomesic: in each orbit, its average is 1.

0 2 1

1 1

1

Root posets of type A: order ideals

What if instead of antichains we take order ideals?

E.g., n = 2:

What is homomesic here?

1

Root posets of type A: rank-signed cardinality

0 2 1

1 1

+ + + +

+ +

−

1

Root posets of type A: rank-signed cardinality is homomesic

Theorem (Haddadan)

Let P be the root poset of type An. If we assign an element x ∈ Pweight wt(x) = (−1)rank(x), and assign an order ideal I ∈ J(P)weight ϕ(I ) =

∑x∈I wt(x), then ϕ is homomesic under rowmotion

and promotion, with average n/2.

The order polytope of a poset

Let P be a poset, with an extra minimal element 0̂ and an extramaximal element 1̂ adjoined.

The order polytope O(P) (introduced by R. Stanley) is the set offunctions f : P → [0, 1] with f (0̂) = 0, f (1̂) = 1, and f (x) ≤ f (y)whenever x ≤P y .We can generalize our entire setup of toggle operators and“rowmotion” to operate on these functions (the “continuouspiecewise-linear (CPL) category”).

Flipping-maps in the order polytope

For each x ∈ P, define the flip-map σx : O(P)→ O(P) sending fto the unique f ′ satisfying

f ′(y) =

{f (y) if y 6= x ,minz ·>x f (z) + maxwx means z covers x and w< · x means x covers w .

Note that the interval [minz ·>x f (z),maxwx f (z) + maxw

Example of flipping at a node

w1 w2

x

z1 z2

.1 .2

.4

.7 .8

−→

.1 .2

.5

.7 .8

1

minz ·>x

f (z) + maxw

Flipping and toggling

If we associate each order-ideal I with the indicator function f ofP \ I (that is, the function that takes the value 0 on I and thevalue 1 everywhere else), then toggling I at x is tantamount toflipping f at x .

That is, we can identify J(P) with the vertices of the polytopeO(P) in such a way that toggling can be seen to be a special caseof flipping.

This may be clearer if you think of J(P) as being in bijection withthe set of monotone 0,1-valued functions on P.

Flipping

Flipping (at least in special cases) is not new, though it is notwell-studied; the most worked-out example we’ve seen isBerenstein and Kirillov’s article Groups generated by involutions,Gelfand-Tsetlin patterns and combinatorics of Young tableaux (St.Petersburg Math. J. 7 (1996), 77–127); seehttp://pages.uoregon.edu/arkadiy/bk1.pdf.

http://pages.uoregon.edu/arkadiy/bk1.pdf

Composing flips

Just as we can apply toggle-maps from top to bottom, we canapply flip-maps from top to bottom (successively at the North,West, East, and South.) :

.8

8888

.6

8888

.6

8888

.4

��.3

��

σN

→ .4

��.3

��

σW

→ .3

��.3

��

.1

8888

.1

8888

.1

8888

.6

8888

.6

8888

σE

→ .3

��.4

��

σS

→ .3

��.4

��

.1

8888

.2

8888

Two Examples of CPL rowmotion orbits

.8>>>

.6>>>

.8>>>

.9>>>

τ

vv

.4

��.3

��τ→ .3

��.4

��τ→ .7

��.6

��τ→ .6

��.7

��

.1

>>>

.2

>>>

.4

>>>

.2

>>>

1:::

1:::

1:::

1:::

τ

ww

1

��0

��τ→ 0

��1

��τ→ 1

��0

��τ→ 0

��1

��

0

:::

0

:::

0

:::

0

:::

The average at eachnode across therespective orbits is shownat right, along with thefile sums.

.8 AA 1 CC

.5}}

.5}}

0.5{{

0.5{{

.2

AA

0

CC

.5 1 .5 0.5 1 0.5

Conjectures in the CPL category

It appears that all of the aforementioned results on homomesy forrowmotion and promotion on J([a]× [b]) lift to correspondingresults in the order polytope, where instead of composingtoggle-maps to obtain rowmotion and promotion we compose thecorresponding flip-maps to obtain continuous piecewise-linear mapsfrom O([a]× [b]) to itself.

News Flash: By lifting an argument from Propp-R. in thecombinatorial category, Propp has very recently shown thatpromotion, and hence rowmotion, must be homomesic (in aslightly generalized sense). But we still don’t have a proof thatthese maps have finite order.

Order of flipping affects order of the composition!

In the combinatorial category, where A(P) and J(P) are finite, it’sclear that any map defined as a product of toggles has finite order.But we can no longer take this for granted in the CPL category.Let P = [2]× [2]. As we’ll soon see, one can show by brute forcethat the CPL mapsσ(1,1) ◦ σ(1,2) ◦ σ(2,1) ◦ σ(2,2) (“lifted rowmotion”) andσ(2,1) ◦ σ(1,1) ◦ σ(2,2) ◦ σ(1,2) (“lifted promotion”) are of order 4.However, not every composition of flips has finite order.

Proposition (Einstein)

The map σ(1,1) ◦ σ(1,2) ◦ σ(2,2) ◦ σ(2,1) (flipping values in clockwiseorder, as opposed to going by rows or columns of P) is of infiniteorder.

De-tropicalizing to birational maps

In the so-called tropical semiring, one replaces the standard binaryring operations (+, ·) with the tropical operations (max,+). In thecontinuous piecewise-linear (CPL) category of the order polytopestudied above, our flipping-map at x replaced the value of afunction f : P → [0, 1] at a point x ∈ P with f ′, where

f ′(x) := minz ·>x

f (z) + maxw

Example of birational rowmotion

In our running example, P = [2]× [2], applying these new flipoperators from top to bottom creates a new rowmotion operator.(Here we assign f (0̂) = f (1̂) = 1.)

z x+yzx+yz

x y 7→ x y 7→ w(x+y)xz y 7→w w w

x+yz

x+yz

w(x+y)xz

w(x+y)yz 7→

w(x+y)xz

w(x+y)yz

w 1z

Example of birational rowmotion orbit

Here’s an orbit of rowmotion in this category:

z x+yzw(x+y)

xy

x y 7→ w(x+y)xzw(x+y)

yz 7→ 1y 1x 7→w 1z

zx+y

1w z

yzw(x+y)

xzw(x+y) 7→ x y

xyw(x+y) w

Geometric Homomesy

In this category, geometric means replace arithmetic means, solet’s compute the product of the function values at each node.

z x+yzw(x+y)

xy

x y 7→ w(x+y)xzw(x+y)

yz 7→ 1y 1x 7→w 1z

zx+y

1w

(x+y)2

xyyz

w(x+y)xz

w(x+y) PROD = 1 1xy

w(x+y)xy

(x+y)2

Geometric homomesy with boundary variables

If we instead generically assign variables f (0̂) = α and f (1̂) = ω:

ω

z (x+y)ωzw(x+y)ω

xy

x y 7→ w(x+y)ωxzw(x+y)ω

yz 7→ αωy αωxw αωz

αzx+y

α

αωw αω

3 (x+y)2

xyαyz

w(x+y)αxz

w(x+y) PROD = α2ω2 α2ω2

αxyw(x+y) α

3ω xy(x+y)2

So the statistic “multiply opposite nodes” has geometric mean αωacross the orbit.

For what posets does this work?

It’s not hard to see that if a map such as rowmotion is homomesicwith respect to some statistics in the birational category, then thisimplies homomesy at the CPL level, which in turn implies it in thecombinatorial category.

We believe that multiplicative versions of homomesy in thebirational category holds for a large class of posets, often ones thatcome up in representation theory. There are also simple examplesof posets, e.g., the Boolean algebra B3 for which nothing we havetried appears to hold. For example, it appears (conjecturally) thatbirational rowmotion has infinite order on B3.

Sage implementation & Wish List

Broader Goals:

A way to define an action τ on a finite set S of n objects andhave Sage give the permutation π ∈ Sn that it represents invarious forms (1-line notation, cycle notation).

Tools to compute arithmetic and multiplicative averages ofstatistics defined on S in the above category.

Specific actions to implement

Bender-Knuth involutions for SSYT: Done and merged byTravis Scrimshaw & DG.

promotion of SSYT: Ticket #7983 needs to be reviewed.

Suter dihedral map on Young’s Lattice: Ticket #14763needs to be reviewed.

toggles & rowmotion in a general poset: Need to discussimplementation of I ∈ J(P): what data type would work?Can Yan Zhang’s code be put on the Trac server?

More wishlist

birational rowmotion in a general poset: Some code hasbeen written DG, but probably still not in shareable form.

promotion in an arbitrary RC poset? Need input fromothers on how to implement RC posets. Probably DG andJessica Striker (and anyone else interested) should discuss.

Unordered trees: It appears that trees currently implementedin Sage are planar/ordered? We need unordered as well.

Semi-fields: The CPL case is just the birational case over thetropical semiring. As a stopgap measure, we could use therecently implemented Tropical Semiring and pretend it’s afield, since Python won’t bother to check types.

Conclusion

A recently identified phenomenon called homomesy appears tobe lurking in a wide range of combinatorial situations.

We are just beginning to develop tools for studying this, sothere are many interesting open problems.

There are intriguing conjectured generalizations to continuouspiecewise-linear maps on order polytopes and to birationalmaps on {f : P → R(ξ1, ξ2, . . . )}.Useful programs exist in Sage for doing computations, andmore are currently under development. . .

References

[Arm06] D. Armstrong, Generalized noncrossing partitions andcombinatorics of Coxeter groups, Mem. Amer. Math. Soc.202 (2006), no. 949.

[AST11] D. Armstrong, C. Stump and H. Thomas, A Uniformbijection between nonnesting and noncrossing partitions,preprint, available at arXiv:math/1101.1277v2 (2011).

[CF95] P. Cameron and D.G. Fon-Der-Flaass, Orbits of AntichainsRevisited, Europ. J. Comb. 16 (1995), 545–554.

[Gan80] E. Gansner, On the equality of two plane partitioncorrespondences, Discrete Math. 30 (1980), 121–132.

[KB95] A. N. Kirillov, A. D. Berenstein, Groups generated byinvolutions, GelfandTsetlin patterns, and combinatorics ofYoung tableaux, Algebra i Analiz, 7:1 (1995), 92-152.

[Pan08] D.I. Panyushev, On orbits of antichains of positive roots,European J. Combin. 30 (2009), no. 2, 586–594.

References 2

[Rei97] V. Reiner, Non-crossing partitions for classical reflectiongroups, Discrete Math. 177 (1997), 195–222.

[RSW04] V. Reiner, D. Stanton, and D. White, The cyclic sievingphenomenon, J. Combin. Theory Ser. A 108 (2004), 17–50.

[Sta09] R. Stanley, Promotion and Evaculation, Electronic J.Comb. 16(2) (2009), #R9.

[SW12] J. Striker and N. Williams, Promotion and rowmotion,European Journal of Combinatorics 33 (2012), 1919–1942.

The last slide of this talk

Slides for this talk are available online (or will be soon) at


For more information, see:

http://jamespropp.org/ucbcomb12.pdf

http://jamespropp.org/mathfest12a.pdf

http://www.math.uconn.edu/∼troby/combErg2012kizugawa.pdfhttp://jamespropp.org/mitcomb13a.pdf

http://www.math.uconn.edu/~troby/ceFPSAC.pdf

Thanks for your attention!

http://www.math.uconn.edu/~troby/research.htmlhttp://jamespropp.org/ucbcomb12.pdfhttp://jamespropp.org/mathfest12a.pdfhttp://www.math.uconn.edu/~troby/combErg2012kizugawa.pdfhttp://jamespropp.org/mitcomb13a.pdfhttp://www.math.uconn.edu/~troby/ceFPSAC.pdf

0.0: 0.1: 0.2: 0.3: 0.4: 0.5: 0.6: 0.7: 0.8: 0.9: 0.10: 0.11: 0.12: 0.13: 0.14: 0.15: 0.16: 0.17: 0.18: 0.19: 0.20: 0.21: 0.22: 0.23: 0.24: 0.25: 0.26: 0.27: 0.28: 0.29: 0.30: 0.31: 0.32: 0.33: 0.34: 0.35: 0.36: 0.37: 0.38: 0.39: 0.40: 0.41: 0.42: 0.43: 0.44: 0.45: 0.46: 0.47: 0.48: 0.49: 0.50: 0.51: 0.52: 0.53: 0.54: 0.55: 0.56: 0.57: 0.58: 0.59: 0.60: 0.61: 0.62: 0.63: 0.64: 0.65: 0.66: 0.67: 0.68: 0.69: 0.70: 0.71: 0.72: 0.73: 0.74: 0.75: 0.76: 0.77: 0.78: 0.79: 0.80: 0.81: 0.82: 0.83: 0.84: 0.85: 0.86: 0.87: 0.88: 0.89: 0.90: 0.91: 0.92: 0.93: 0.94: 0.95: 0.96: 0.97: 0.98: 0.99: 0.100: 0.101: 0.102: 0.103: 0.104: 0.105: 0.106: 0.107: 0.108: 0.109: 0.110: 0.111: 0.112: 0.113: 0.114: 0.115: 0.116: 0.117: 0.118: 0.119: 0.120: 0.121: 0.122: 0.123: 0.124: 0.125: 0.126: 0.127: 0.128: 0.129: 0.130: 0.131: 0.132: 0.133: 0.134: 0.135: 0.136: 0.137: 0.138: 0.139: 0.140: 0.141: 0.142: 0.143: 0.144: 0.145: 0.146: 0.147: 0.148: 0.149: 0.150: 0.151: 0.152: 0.153: 0.154: 0.155: 0.156: 0.157: 0.158: 0.159: 0.160: 0.161: 0.162: 0.163: 0.164: 0.165: 0.166: 0.167: 0.168: 0.169: 0.170: 0.171: 0.172: 0.173: 0.174: 0.175: 0.176: 0.177: 0.178: 0.179: 0.180: 0.181: 0.182: 0.183: 0.184: 0.185: 0.186: 0.187: 0.188: 0.189: 0.190: 0.191: 0.192: 0.193: 0.194: 0.195: 0.196: 0.197: 0.198: 0.199: 0.200: 0.201: 0.202: 0.203: 0.204: 0.205: 0.206: 0.207: 0.208: 0.209: 0.210: 0.211: 0.212: 0.213: 0.214: 0.215: 0.216: 0.217: 0.218: 0.219: 0.220: 0.221: 0.222: 0.223: 0.224: 0.225: 0.226: 0.227: 0.228: 0.229: 0.230: 0.231: 0.232: 0.233: 0.234: 0.235: 0.236: 0.237: 0.238: 0.239: 0.240: 0.241: 0.242: 0.243: 0.244: 0.245: 0.246: 0.247: 0.248: 0.249: anm0: 0.EndLeft: 0.StepLeft: 0.PlayPauseLeft: 0.PlayPauseRight: 0.StepRight: 0.EndRight: 0.Minus: 0.Reset: 0.Plus: 1.0: 1.1: 1.2: 1.3: 1.4: 1.5: 1.6: 1.7: 1.8: 1.9: 1.10: 1.11: 1.12: 1.13: 1.14: 1.15: 1.16: 1.17: 1.18: 1.19: 1.20: 1.21: 1.22: 1.23: 1.24: 1.25: 1.26: 1.27: 1.28: 1.29: 1.30: 1.31: 1.32: 1.33: 1.34: 1.35: 1.36: 1.37: 1.38: 1.39: 1.40: 1.41: 1.42: 1.43: 1.44: 1.45: 1.46: 1.47: 1.48: 1.49: 1.50: 1.51: 1.52: 1.53: 1.54: 1.55: 1.56: 1.57: 1.58: 1.59: 1.60: 1.61: 1.62: 1.63: 1.64: 1.65: 1.66: 1.67: 1.68: 1.69: 1.70: 1.71: 1.72: 1.73: 1.74: 1.75: 1.76: 1.77: 1.78: 1.79: 1.80: 1.81: 1.82: 1.83: 1.84: 1.85: 1.86: 1.87: 1.88: 1.89: 1.90: 1.91: 1.92: 1.93: 1.94: 1.95: 1.96: 1.97: 1.98: 1.99: 1.100: 1.101: 1.102: 1.103: 1.104: 1.105: 1.106: 1.107: 1.108: 1.109: 1.110: 1.111: 1.112: 1.113: 1.114: 1.115: 1.116: 1.117: 1.118: 1.119: 1.120: 1.121: 1.122: 1.123: 1.124: 1.125: 1.126: 1.127: 1.128: 1.129: 1.130: 1.131: 1.132: 1.133: 1.134: 1.135: 1.136: 1.137: 1.138: 1.139: 1.140: 1.141: 1.142: 1.143: 1.144: 1.145: 1.146: 1.147: 1.148: 1.149: 1.150: 1.151: 1.152: 1.153: 1.154: 1.155: 1.156: 1.157: 1.158: 1.159: 1.160: 1.161: 1.162: 1.163: 1.164: 1.165: 1.166: 1.167: 1.168: 1.169: 1.170: 1.171: 1.172: 1.173: 1.174: 1.175: 1.176: 1.177: 1.178: 1.179: 1.180: 1.181: 1.182: 1.183: 1.184: 1.185: 1.186: 1.187: 1.188: 1.189: 1.190: 1.191: 1.192: 1.193: 1.194: 1.195: 1.196: 1.197: 1.198: 1.199: 1.200: 1.201: 1.202: 1.203: 1.204: 1.205: 1.206: 1.207: 1.208: 1.209: 1.210: 1.211: 1.212: 1.213: 1.214: 1.215: 1.216: 1.217: 1.218: 1.219: 1.220: 1.221: 1.222: 1.223: 1.224: 1.225: 1.226: 1.227: 1.228: 1.229: 1.230: 1.231: 1.232: 1.233: 1.234: 1.235: 1.236: 1.237: 1.238: 1.239: 1.240: 1.241: 1.242: 1.243: 1.244: 1.245: 1.246: 1.247: 1.248: 1.249: anm1: 1.EndLeft: 1.StepLeft: 1.PlayPauseLeft: 1.PlayPauseRight: 1.StepRight: 1.EndRight: 1.Minus: 1.Reset: 1.Plus: 2.0: 2.1: 2.2: 2.3: 2.4: 2.5: 2.6: 2.7: 2.8: 2.9: 2.10: 2.11: 2.12: 2.13: 2.14: 2.15: 2.16: 2.17: 2.18: 2.19: 2.20: 2.21: 2.22: 2.23: 2.24: 2.25: 2.26: 2.27: 2.28: 2.29: 2.30: 2.31: 2.32: 2.33: 2.34: 2.35: 2.36: 2.37: 2.38: 2.39: 2.40: 2.41: 2.42: 2.43: 2.44: 2.45: 2.46: 2.47: 2.48: 2.49: 2.50: 2.51: 2.52: 2.53: 2.54: 2.55: 2.56: 2.57: 2.58: 2.59: 2.60: 2.61: 2.62: 2.63: 2.64: 2.65: 2.66: 2.67: 2.68: 2.69: 2.70: 2.71: 2.72: 2.73: 2.74: 2.75: 2.76: 2.77: 2.78: 2.79: 2.80: 2.81: 2.82: 2.83: 2.84: 2.85: 2.86: 2.87: 2.88: 2.89: 2.90: 2.91: 2.92: 2.93: 2.94: 2.95: 2.96: 2.97: 2.98: 2.99: 2.100: 2.101: 2.102: 2.103: 2.104: 2.105: 2.106: 2.107: 2.108: 2.109: 2.110: 2.111: 2.112: 2.113: 2.114: 2.115: 2.116: 2.117: 2.118: 2.119: 2.120: 2.121: 2.122: 2.123: 2.124: 2.125: 2.126: 2.127: 2.128: 2.129: 2.130: 2.131: 2.132: 2.133: 2.134: 2.135: 2.136: 2.137: 2.138: 2.139: 2.140: 2.141: 2.142: 2.143: 2.144: 2.145: 2.146: 2.147: 2.148: 2.149: 2.150: 2.151: 2.152: 2.153: 2.154: 2.155: 2.156: 2.157: 2.158: 2.159: 2.160: 2.161: 2.162: 2.163: 2.164: 2.165: 2.166: 2.167: 2.168: 2.169: 2.170: 2.171: 2.172: 2.173: 2.174: 2.175: 2.176: 2.177: 2.178: 2.179: 2.180: 2.181: 2.182: 2.183: 2.184: 2.185: 2.186: 2.187: 2.188: 2.189: 2.190: 2.191: 2.192: 2.193: 2.194: 2.195: 2.196: 2.197: 2.198: 2.199: 2.200: 2.201: 2.202: 2.203: 2.204: 2.205: 2.206: 2.207: 2.208: 2.209: 2.210: 2.211: 2.212: 2.213: 2.214: 2.215: 2.216: 2.217: 2.218: 2.219: 2.220: 2.221: 2.222: 2.223: 2.224: 2.225: 2.226: 2.227: 2.228: 2.229: 2.230: 2.231: 2.232: 2.233: 2.234: 2.235: 2.236: 2.237: 2.238: 2.239: 2.240: 2.241: 2.242: 2.243: 2.244: 2.245: 2.246: 2.247: 2.248: 2.249: anm2: 2.EndLeft: 2.StepLeft: 2.PlayPauseLeft: 2.PlayPauseRight: 2.StepRight: 2.EndRight: 2.Minus: 2.Reset: 2.Plus:

Combinatorial Actions, Orbit Averages, and Sage Implementationtroby/homomesySageOrsay2013.pdf ·...

Documents

Transcript of Combinatorial Actions, Orbit Averages, and Sage Implementationtroby/homomesySageOrsay2013.pdf ·...