Arxiv:1611.04079V2 [Math.CO] 29 Dec 2020 N H Oﬃinso H Hoai Oyoil[5 3.For 13]

arXiv:1611.04079v2 [math.CO] 29 Dec 2020 n h offiinso h hoai oyoil[5 3.For 13]. [15, polynomial chromatic the of coefficients the ing ooigcmlxΓ( complex coloring n ucmlxΓ( subcomplex a and ercfnto 2] te xmlsicuequasisymmetr include examples c Other Stanley’s for functions is [20]. function function quasisymmetric metric alg resulting Hopf combinatorial the the is where ve example fundamental had A that algebras acters. Hopf combinatorial quasisymmetric from of came examples combinatorial functions well-known terminal many how the showed They form functions quasisymmetric n euesm osqecso htrsl.Ortria ob terminal geometry. Our to connections simplicial the relative result. discuss balanced i that we are paper of structures mo this whose consequences Hopf species of combinatorial some goal linearized deduce The for and result of [9]. type matroids similar to associated function lxcnb rte stestdffrneo h oee complex Coxeter the of difference set the as written be can plex h hoai oyoili h ibr oyoilo nide an Φ( of complex simplicial polynomial relative Hilbert a the from arises is polynomial chromatic the aoilSeis obntra ofAgba,Blne S Balanced Algebras, Hopf Combinatorial Species, natorial OOIGCMLXSADCMIAOILHOPF COMBINATORIAL AND COMPLEXES COLORING nterlnmr ae,Aua,Breo n otl 2 pr [2] Sottile and Bergeron Aguiar, paper, landmark their In 2020 Date h hoai oyoilcnb tde sn emty pt up geometry: using studied be can polynomial chromatic The e od n phrases. and words Key aur ,2021. 1, January : ahmtc ujc Classification. Subject Mathematics ovxcaatr.W losuysvrleape fcombin of examples several study also mo monoids. We Hopf Hopf combinatorial of characters. monoid, category collectio the Hopf convex in the combinatorial object that a terminal show forms the We complexes coloring monoid. invariant such Hopf function combinatorial quasisymmetric the the i to some by discover satisfied and com are a construction, when a that such determine We has monoid graph. Hopf a rial of complex general coloring a of are notion These monoids. Hopf combinatorial linearized Abstract. P priin 1] n h ilr-i-enrquasisymmet Billera-Jia-Reiner the and [11], -partitions egnrlz h oino ooigcmlxo rp to graph a of complex coloring of notion the generalize We G a enue ofidsvrlnwieulte regard- inequalities new several find to used been has ) G ,uulyrfre oa h ooigcmlx The complex. coloring the as to referred usually ), hoai oyoil,Qaiymti ucin,Combi- Functions, Quasisymmetric Polynomials, Chromatic 1. AO .WHITE A. JACOB MONOIDS Introduction 1 81,18D35. 18D10, G .Terltv ipiilcom- simplicial relative The ). mlca Complexes. implicial zto fthe of ization opee,s first so complexes, associated s nequalities od with noids od nspecies, in noids rmtcsym- hromatic bao graphs, of ebra ysml char- simple ry hc is which hsppr we paper, this ofalgebra. Hopf fall of n binato- cgenerating ic atorial l[1 which [21] al oso a show to s generating vdthat oved ftype of shift, a o eti a is ject ric A 2 JACOB A. WHITE refer to the relative simplicial complex Φ(G) as the coloring complex and show that flag f-vector of Φ(G) encodes information about the chromatic symmetric function. We have

(1) χ(G, x)= fS(Φ(G))Mα(S) S⊂X[|V |−1] where Mα(S) is a monomial symmetric function. Thus we can use the relative coloring complex to understand the chromatic symmetric function, similar to how the coloring complex was used to study the chromatic polynomial. We assume the reader is familiar with quasisymmetric functions, integer compositions, the partial order of reﬁnement, and monomial and fundamental quasisymmetric functions. In order to get new information using this perspective, we prove new results about balanced relative simplicial complexes. In Section 2, we study ﬂag f-vectors of balanced relative simplicial complexes, as well as the f- vector of pure relative simplicial complexes. We prove the following two theorems. Theorem 1. Let Φ be a balanced relative simplicial complex of dimension d. Let S ⊆ T ⊆ [d]. Then fS(Φ) ≤ fT (Φ).

Theorem 2. Let (f−1,f0,...,fd) be the f-vector of a pure relative simplicial complex Φ. Then the f-vector satisﬁes the following two systems of inequalities:

(1) f−1 ≤ f0 ≤···≤ f⌊d/2⌋. (2) fi−1 ≤ fd−i−1 for 0 ≤ i< ⌊d/2⌋.

Any sequence (a0, . . . , an) that satisfies the two inequalities of Theorem 2 is a strongly flawless sequence. Since the chromatic symmetric function (chromatic polynomial, respectively) of a graph is related to the flag f-vector (ordinary f-vector, respectively) of Φ(G), we obtain inequalities about chromatic symmetric functions and chromatic polynomials. Given an integer composition α, let [Mα]χ(G, x) be the coefficient of Mα. Then for α ≤ β, we see that [Mα]χ(G, x) ≤ [Mβ]χ(G, x). We express this fact more succinctly by saying that χ(G, x) is n n−1 M-increasing. Similarly, if we write χ(G, x)= i=0 fi i , then (f0,...,fn) is strongly flawless. Both of these results appear to be new, although the fact that χ(G, x) is M-increasing can be deducedP from the known fact that it is F -positive. Our primary goal in this paper is to generalize the construction of the coloring complex of a graph to other combinatorial objects and quasisymmetric functions. Given a collection C of combinatorial objects, and a quasisymmetric function Ψ(O, x) for each O ∈ C, we wish to find balanced relative simplicial complexes Φ(O) such that an analogue of Equation (1) holds, with χ(G, x) replaced by Ψ(O, x). Whenever we can show that such COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 3 a construction exists, then we know that the quasisymmetric functions are M-increasing, and that the corresponding polynomials are strongly flawless. Our focus will be on combinatorial objects that form a combinatorial Hopf monoid, and where the quasisymmetric function is induced by a linearized character. In Section 3, we define linearized combinatorial Hopf monoids in species. Roughly speaking, these are Hopf monoids H in the category of linear species equipped with a character ϕ, where each vector space HN comes with a distinguished basis HN such that: (1) The product of basis elements is a basis element. (2) When the coproduct of a basis element is nonzero, then it is a simple tensor of basis elements. (3) The character takes on values 0 or 1 on basis elements. Our notion of linearization is more general than in [18, 4], because we are linearizing pointed set species, and regarding the base points as being the zero vector. For example, the Hopf monoid of posets discussed in this paper is linearized in our sense, but not linearized from set species, because the coproduct of a poset might be zero. Our decision to study linearized characters is motivated by the fact that many examples of characters studied in the literature are linearized from pointed set species. We also discuss the ϕ- chromatic quasisymmetric function, which is an invariant associated to any H-structure in a combinatorial Hopf monoid. The existence of this invariant comes from the work of Aguiar and Mahajan [3] and Aguiar, Bergeron, and Sottile [2], although we give an alternative combinatorial description of the invariant in terms of ϕ-proper colorings. Our primary focus will be on the chromatic character χ which we introduce in this paper. It is defined for any linearized Hopf monoid, and generalizes well-known characters for graphs, posets, matroids, and generalized permutohedra. Given a linearized Hopf monoid H with character ϕ, we want to associate a relative simplicial complex Φϕ(h) to each H-structure h such that

Ψϕ(h, x)= fS(Φϕ(h))Mα(S). S⊆X[|V |−1] We refer to Φ as the geometric realization of H. In Section 4, we give sufficient conditions on when a linearized combinatorial Hopf monoid has a geometric realization. We call characters which satisfy the conditions convex characters. The precise definition is given in Definition 19. We define the ϕ-coloring complex Φϕ(h) in Definition 23. We show that the coloring complexes do form a geometric realization for H whenever ϕ is a convex character. Theorem 3. Let (H, ϕ) be a linearized combinatorial Hopf monoid with a convex character ϕ. Given a H-structure h, let Φϕ(h) be the ϕ-coloring complex of h. Then

Ψϕ(h, x)= fS(Φϕ(h))Mα(S), S⊆X[|V |−1] 4 JACOB A. WHITE where N is the vertex set of h. If ϕ is a balanced convex character, then Φϕ(h) is balanced for all h. In that case, we know that Ψϕ(h, x) is M-increasing and the corresponding polynomial χϕ(h,x) is strongly ﬂawless.

As an example of this Theorem in action, we show that the chromatic character is always a balanced convex character for every linearized Hopf monoid. Thus, we have a vast generalization of the coloring complex of a graph. A first step towards obtaining new inequalities on flag f-vectors is to classify the possible coloring complexes that can arise from our construction. In Section 5, we define an abstract coloring complex. We give the definition in Definition 30. We show that, if H is a linearized combinatorial Hopf monoid with a balanced convex character, then the ϕ-coloring complex of a H-structure h is an abstract coloring complex. Then we show that the collection of abstract coloring complexes forms a species C. Hence we do classify all the possible coloring complexes. Moreover, in Theorem 39, we show that the resulting species is the terminal object in the category of linearized combinatorial Hopf monoids with convex characters. In particular, the map associating Φ(h) to h is a unique Hopf algebra homomorphism. This will be a foundation for future papers: in order to prove a geometric result about realizations of Hopf monoids, it suffices to prove the result for the Hopf monoid of coloring complexes, or some corresponding Hopf submonoid. Throughout the paper we study familiar examples of graphs, posets, and matroids. These examples have appeared in [3, 4, 1], and the resulting quasisymmetric functions are known to be F -positive. In Section 6, we focus on several other examples. First, we study a new character for graphs that was recently studied by Aval, Bergeron, and Machacek [6]. It forms the only prior example of a character we have found that is convex but not balanced. Then we classify all the balanced convex characters of graphs. Then we study examples of Hopf monoids related to mixed graphs, double posets, rooted connected graphs, antimatroids, and generalized permutohedra. The Hopf algebra of double posets was introduced by Malvenuto and Reetuenauer [17], while the Hopf monoid of generalized Permutohedra was introduced by Aguiar and Ardila [1]. Mixed graphs were introduced in [8] and antimatroids were introduced in [16]. The examples of mixed graphs and double posets are interesting because the resulting ϕ-chromatic quasisymmetric functions are not always F -positive. Thus our results about quasisymmetric functions being M-increasing does yield non-trivial results in some cases. Many combinatorial Hopf monoids in the literature have morphisms to the Hopf monoid of extended generalized permutohedra (examples are discussed in detail in [1]). This is an interesting phenomenon that is addressed in [5]. This motivates us to wonder how coloring complexes compare to generalized permutohedra. We show that the Hopf monoid of antimatroids and rooted COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 5 connected graphs do not have injective morphisms to the Hopf monoid of generalized permutohedra, despite having injective morphisms to the Hopf monoid of coloring complexes. Thus we see that coloring complexes are a more general object that contain interesting submonoids. On the other hand, generalized permutohedra have connections to valuation theory.

2. Balanced Relative Simplicial Complexes In this section, we discuss balanced relative simplicial complexes. We prove some new results about their ﬂag f-vectors, and we show that the f-vector of a pure relative simplicial complex is strongly ﬂawless.

2.1. Relative Simplicial Complexes and face vectors. Definition 4. A relative simplicial complex on a vertex set V is a collection Φ of subsets of V with the following property: • For every ρ ⊆ σ ⊆ τ, if ρ, τ ∈ Φ, then σ ∈ Φ. The name comes from the fact that there exists simplicial complexes (Γ, Σ) with Γ ⊆ Σ, and Φ = Σ \ Γ. Given Φ with vertices S, we let C[S] be the polynomial ring with indeterminates s1,...,sk, the vertices of S. The Stanley-Reisner ideal for Σ is generated by hσ ⊆ S : σ 6∈ Σi, and the Stanley-Reisner module for (Γ, Σ) is IΓ/IΣ. The module is graded by total degree, and its Hilbert function Hilb(Φ,n) is the number of monomials of degree n in the module. It is known that the Hilbert function is eventually a polynomial in n, and that it is an invariant of Φ, despite being defined in terms of (Γ, Σ): details can be found in [19]. The dimension of a face σ ∈ Φ is |σ|− 1, and the dimension of Φ is that maximum dimension of a face of Φ. A complex is pure if all the maximal faces have the same dimension. Now we discuss inequalities involving the f-vectors of relative simplicial complexes. Our proofs are similar to ones appearing in work of Hibi [14]. Let fi(Φ) be the number of faces of dimension i. The f-vector is defined by f(Φ) = (f−1(Φ),...,fd(Φ)). First, observe that

d n − 1 Hilb(Φ,n)= f . i i − 1 iX=−1 Thus, determining the f-vector determines the Hilbert polynomial. Recall that the f-vector is strongly ﬂawless if the following two systems of inequalities are satisﬁed:

(1) f−1 ≤ f0 ≤···≤ f⌊d/2⌋. (2) fi−1 ≤ fd−i−1 for 0 ≤ i< ⌊d/2⌋. Proposition 5. Suppose that Φ is a pure relative simplicial complex. Then f(Φ) is strongly ﬂawless. 6 JACOB A. WHITE

Proof. Let F(Φ) = {F1,...,Fm} be the set of facets of Σ. If m = 0, then there are no facets, and Φ is the void complex, and both inequalities are trivially true. So we suppose m ≥ 1. Let Σ = hF1,...,Fmi, and let Γ = Φ \ Σ. We prove the first set of inequalities by induction on m. V First, suppose m = 1. Let i < ⌊d/2⌋. Then Σ is a simplex. Let i be the subsets of V (Σ) of size i. It is a known fact (see [14, 10]) that there V V is a bijection σi : i → d−i with the feature that S ⊆ σ(S) for all S. Fix such a bijection. Then if σ(S) ∈ Γ, we have S ∈ Γ. Hence S ∈ Σ \ Γ implies that σ(S) ∈ Σ \ Γ. Therefore σ restricts to an injection on the set of i-faces of Φ, whose image is contained in the set of (d − i)-faces of Φ. Hence fi−1(Φ) ≤ fd−i−1(Φ). Now suppose m > 1. Let Φ1 = hFmi∩ Φ, and let Φ2 = Φ \ Φ1. Clearly Φ1 is a pure relative simplicial complex. Every face in Φ1 is a face of Φ that is contained in Fm. Hence every face in Φ2 must be contained in some facet in {F1,...,Fm−1}, as otherwise it would be a face of Φ1. Thus Φ2 is a pure relative simplicial complex. We see that fi(Φ) = fi(Φ1)+ fi(Φ2). By induction, both f(Φ1) and f(Φ2) are flawless, and the sum of two flawless sequences is also flawless. Thus, f(Φ) is flawless. Now we prove the second set of inequalities. Let i ≤ ⌊d/2⌋, and define Φ2i to be the subset of faces of dimension at most 2i. Then Φ2i is pure, and so f(Φ2i) is flawless. One of those inequalities coming from the definition of flawless is equivalent to fi−1(Φ) ≤ fi(Φ).

2.2. Balanced Relative Simplicial Complexes. A colored relative simplicial complex is a relative simplicial complex with a function κ : S → [k] where [k] is a set of colors, with the property that no two vertices of a face of Φ receive the same color. It is a balanced relative simplicial complex if k = d + 1, where d is the dimension of the complex. For colored relative simplicial complexes, we can define the flag f-vector. Given S ⊆ [n], we let fS(Φ) denote the number of faces {i1,...,ik} such that {κ(i1),...,κ(ik)} = S. We encode the flag f-vector with a quasisymmetric function

Hilb(Φ, x)= fS(Φ)Mα(S) SX⊂[k] where if S = {s1,...,sk}, then α(S) = (s1,s2 − s1,s3 − s2,...,sk−1 − sk−2). We say that a flag f-vector is weakly increasing if fS(Φ) ≤ fT (Φ) for all S ⊆ T ⊆ [d + 1]. Equivalently, a flag f-vector is weakly increasing if Hilb(Φ, x) is M-increasing. Naturally, there is also a flag h-vector, given by:

|S|−|T | fS(Φ) = hT (Φ) or hS (Φ) = (−1) fT (Φ) TX⊆S TX⊆S

We see that Hilb(Φ, x)= S⊆[k] hS(Φ)Fα(S), so the ﬂag h-vector expresses Φ in the basis of fundamental quasisymmetric functions. P COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 7

We say a flag h-vector is positive if hS(Φ) ≥ 0 for all S ⊆ 0. If the flag h-vector of Φ is positive, then the flag f-vector of Φ is weakly increasing. The converse does not always hold. Proposition 6. Let Φ be a balanced relative simplicial complex of dimension d. Then the flag f-vector of Φ is weakly increasing. Proof. We prove the results by induction on the number of facets of Φ. Let F1, F2,...,Fm be the facets of Φ. First, suppose m = 1. Φ has exactly one facet σ. Then f[d+1] = 1. In fact, by replacing vertex i with f(i), we can assume that our simplicial complex has vertex set [d + 1], and σ = [d + 1]. We also see that fS ≤ 1 for all S ⊆ [d + 1]. Let S ⊆ T . If fS = 0, then fS ≤ fT . So suppose fS = 1. Then S ∈ Φ. Since [d + 1] is also an element of Φ and Φ is a relative simplicial complex, we conclude that T ∈ Φ. Hence fT = 1. Now assume m> 1. Let Φ1 =Φ ∩ hFmi, and let Φ2 =Φ \ Φ1. Then the flag f-vector satisfies the following:

fS(Φ) = fS(Φ1)+ fS(Φ2) for every S ⊆ [d+1]. Moreover, both Φ1 and Φ2 are balanced. By induction, the ﬂag f-vectors of Φ1 and Φ2 are both weakly-increasing. Thus the ﬂag f-vector of Φ is also weakly increasing.

Finally, we mention the relationship between Hilb(Φ, x) and Hilb(Φ,x). Recall that the double cone of Φ is the relative simplicial complex dcone(Φ) with vertex set V (Φ) ∪ {−∞, ∞} given by: dcone(Φ) = {σ ∪ τ : σ ∈ Φ,τ ⊆ {−∞, ∞}}

Given a quasisymmetric function f, the principal specialization psn(f) is obtained by setting xi = 1 for i ≤ n, and xi = 0 otherwise. The invariant psn(f) is a polynomial in the variable n. Lemma 7. Let Φ be a colored relative simplicial complex. Then

psn Hilb(Φ, x) = Hilb(dcone(Φ),n + 1).

n+1 Proof. Observe that psn+1 Mα = ℓ(α) . Thus, n + 1 ps Hilb(Φ, x)= f (Φ) n+1 S |S| + 1 SX⊆[d] d n + 1 = f (Φ) . k−1 k + 1 Xk=0 8 JACOB A. WHITE

Applying the recurrence n + 1 n − 1 n − 1 n − 1 = + 2 + k + 1 k + 1 k k − 1 for each k, we obtain d−1 n − 1 (f (Φ) + 2f (Φ) + f (Φ)) . k k−1 k−2 k kX=−1 However, we have

fk(dcone(Φ)) = fk(Φ) + 2fk−1(Φ) + fk−2(Φ), so the result follows.

2.3. Coxeter Complex of type A and Convex Albums. In this section, we discuss relative simplicial complexes Φ that are a subcomplex of the Coxeter complex of type A. These form a key class of relative complexes in this paper. Given a ﬁnite set N, let V (N) be the collection of proper subsets of N. Then we deﬁne the Coxeter complex of the type A,ΣN , as follows:

ΣN = {{F1, F2,...,Fk} : ∅⊂ F1 ⊂ F2 ⊂···⊂ Fk ⊂ N}

The faces of ΣN are flags of proper subsets of N. We denote the faces by F• and given a flag F• = F1 ⊂ F2 ⊂···⊂ Fk, we write ℓ(F•)= k + 1. There is another combinatorial description of a face of the Coxeter complex. Given a finite set N, a set composition is a sequence (S1,...,Sk) of disjoint non-empty subsets whose union is N. We denote set compositions as S1|S2|···|Sk, and refer to the sets Si as blocks. We let C |= N to denote that C is a set composition of N, and let ℓ(C) = k be the length of the composition. To every set composition C |= N, there is an associated flag F (C) = i {F1,...,Fℓ(C)−1}. We define Fi = j=1 Cj. Note that F (C) ∈ ΣN . Simi- larly, if F ∈ Σ , and ℓ(F )= k, then there is an associated set composition • N • S C(F ), defined by:

(1) C1 = F1, (2) Ci = Fi \ Fi−1 for 2 ≤ i ≤ k − 1, and (3) Ck = N \ Fk−1. For example, to the set composition 13|2|45, the associated edge in the complex Σ{1,2,3,4,5} is the flag {1, 3} ⊂ {1, 2, 3}. Hence, we can denote faces of the Coxeter complex by flags of subsets or by set compositions. An album A is a collection of set compositions. To every album A, we can define a collection of flags F (A) = {F (C) : C ∈ A}. Similarly, given a collection F of faces of the Coxeter complex ΣN , we can associate an album A(F) = {C(F ) : F ∈ F}. This defines a correspondence between albums and collections of flags. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 9

Example 8. As an example, if we let F be the collection of all flags in Σ{a, b, c}, then A(F)= {a|b|c, ab|c, b|a|c, b|ac, b|c|a, bc|a, c|b|a, c|ab, c|a|b, ac|b, a|c|b, a|bc}. Now we discuss how to determine if F (A) is a relative simplicial complex. Since there is a partial order on the faces of ΣN given by containment, we also have a partial order on set compositions: we say C ≤ C′ if every block of C′ is a subset of a block of C. This partial order is known as refinement. An album is convex if, whenever C,C′′ ∈A and there exists C′ with C ≤ C ≤ C′′, we have C′ ∈A. Proposition 9. An album A is convex if and only if F (A) is a relative simplicial complex. Proof. Let N be a finite set, and suppose that A is a convex album on N. Then F (A) is a collection of flags. We wish to show F (A) is a relative simplicial complex with vertex set given by proper subsets of N. Let ′′ ′′ ′ F•, F• ∈ F (A) such that F• ⊆ F• . Let F• be another flag such that ′ ′′ ′ ′′ F ⊆ F• ⊆ F• . Then observe that C(F ) ≤ C(F•) ≤ C(F• ) as set composi- ′′ ′ tions, and C(F ),C(F• ) ∈A. Since A is convex, it follows that C(F•) ∈A, ′ and hence F• ∈ F (A). Thus F (A) is a relative simplicial complex. Now suppose that F (A) is a relative simplicial complex. Let C,C′,C′′ be set compositions with C ≤ C′ ≤ C′′ and C,C′′ ∈A. Then F (C) ≤ F (C′) ≤ F (C′′), and F (C), F (C′′) ∈ F (A). Since F (A) is a relative simplicial complex, it follows that F (C′) ∈ F (A), so C′ ∈ A. Therefore A is a convex album.

Example 10. As an example, let A = {a|b|c|d, a|c|b|d, c|a|b|d, c|a|d|b, a|c|d|b, c|d|a|b, ab|c|d, a|b|cd, ab|cd, ac|bd, ac|d|b, a|c|bd, c|a|bd, ac|b|d, c|ab|d, c|ad|b, cd|ab, c|d|ab, cd|a|b}. The reader can check that this is a convex album. Hence F (A) is a relative simplicial complex. In fact, it is the relative simplicial complex given in middle of Figure 6.2.

3. Combinatorial Hopf monoids In this section, we define combinatorial Hopf monoids. First, we define the notion of Hopf monoid in the category of linear species. A linear species is a functor F : Set → V ec from the category of finite sets with bijections to the category of finite dimensional vector spaces over a field K and linear transformations. 3.1. Linearized Hopf Monoids. A Hopf monoid is a Hopf monoid object in the category of linear species [3]. We give some of the structural definition and axioms related to associativity and compatability. There are also unit, 10 JACOB A. WHITE counit morphisms and antipode axioms. We refer to [4, 1] for more details. For every pair of disjoint finite sets M,N, there are linear transformations µM,N : HM ⊗ HN → HM⊔N and ∆M,N : HM⊔N → HM ⊗ HN . We refer to µ as multiplication and ∆ as comultiplication. We focus only on connected species, where dim H∅ = 1. Recall that idX : HX → HX is the identity function. We require several axioms, including: (1) (associativity) for every triple L, M, and N of disjoint sets, we have µL,M⊔N ◦ idL ⊗µM,N = µL⊔M,N ◦ µL,M ⊗ idN . (2) (identity) For every finite set N, and x ∈ HN , we have µ∅,N (1 ⊗ x)= x = µN,∅(x ⊗ 1). (3) (coassociativity) for every triple L, M, and N of disjoint sets, we have idL ⊗∆M,N ◦ ∆L,M⊔N =∆L,M ⊗ idN ◦∆L⊔M,N . (4) (counit) for every finite set N, and x ∈ HN , we have ∆∅,N (x) = 1 ⊗ x and ∆N,∅(x)= x ⊗ 1. (5) (compatability) for every quadruple A,B,C and D of disjoint sets, we have

∆A⊔C,B⊔D ◦ µA⊔B,C⊔D = µA,C ⊗ µB,D ◦ idA ⊗τB,C ⊗ idD ◦∆A,B ⊗ ∆C,D,

where τB,C : HB ⊗ HC → HC ⊗ HB is given by τB,C (x ⊗ y)= y ⊗ x.

Note that these are equalities of functions. We let ∆L,M,N = idL ⊗∆M,N ◦ ∆L,M⊔N , and x · y = µM,N (x ⊗ y). Now we discuss the notion of pointed set species and linearization. Most of our Hopf monoids come from this construction. A pointed set consists of a pair (A, x) where x is an element of A. An morphism ϕ : (A, x) → (B,y) between pointed sets consists of a function ϕ : A → B such that ϕ(x) = y. An isomorphism is a bijective morphism. Unless we need to be careful, we will refer to the base point as 0. So all of our pointed sets can be denoted by (A, 0). Given two pointed sets (A, 0) and (B, 0), their wedge product is the pointed (A × B/ ∼, (0, 0)), where ∼ is the equivalence relation induced by requiring (a, 0) ∼ (0, 0) ∼ (0, b) for all a ∈ A, b ∈ B. A pointed set species is a functor F : Set → PSet from the category of finite sets with bijections, to the category of pointed finite sets with isomorphisms. Given a set species F, there is an associated linear species KF called the linearization: we define (KF)N to be the vector space with basis FN \ {0}. We refer to f 6= 0 as an F-structure if there exists a finite set N such that f ∈ FN \ {0}. Let H be a species, and suppose that the linearization KH is a Hopf monoid. We say that H is a linearized Hopf monoid if it is a Hopf monoid, and: (1) For every pair of disjoint finite sets M,N, the multiplication map µM,N on KH is linearized from a morphism µM,N : HM ∧HN → HM⊔N . This means that for every x ∈ HM , y ∈ HN , we have x · y ∈ HM⊔N or x · y = 0. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 11

(2) For every pair of disjoint ﬁnite sets M,N, the comultiplication map ∆M,N on KH is linearized from a morphism ∆M,N : HM⊔N → HM ∧ HN . Thus, for every x ∈ HM⊔N , if ∆M,N (x) 6= 0 then there exists x|M ∈ HM and x/M ∈ HN with ∆M,N (x)= x|M ⊗ x/M. Our notion of linearized is more general than the usual one found in the literature, because we are working with pointed set species. The advantage is that we view the species of posets as a linearized species, despite the fact that the coproduct is sometimes zero. We refer to h|T /S as a minor of h for any S ⊆ T ⊆ N. We will denote the minor h|T /S by h[S,T ]. Note that this is merely a convenience that makes later formulas easier to parse. We do note that if a Hopf monoid is linearized, then its multiplication is linearized over set species. Let H be a linearized Hopf monoid, and let M and N be disjoint ﬁnite sets with x ∈ HM and y ∈ HN . Suppose that x·y = 0. Then x =0or y = 0. This is because we have 0 = ∆M,N (x·y)= x⊗y. Thus, when we work with H-structures, it follows that x · y is also a H-structure. Most Hopf monoids that have been studied are Hadamard products of linearized Hopf monoids and their duals. Examples of linearized Hopf monoids include the Hopf monoid of graphs, posets, matroids, hypergraphs, set partitions, linear orders, and generalized permutohedra. In fact, almost every Hopf monoid studied in [1] is a linearized Hopf monoid. We now list several examples of linearized Hopf monoids. These examples have all appeared in [3].

Example 11. Given a ﬁnite set N, let EN = {1, 0}. This gives rise a pointed set species, called the exponential species.

Example 12. Given a ﬁnite set N, we let SN denote the collection of set compositions with ground set N. This forms the species of set compositions. Given a set decomposition N = S ⊔ T , if C |= S is given by C1|···|Ck ′ ′ ′ ′ and C |= T is given by C1|···|Cr, then their product C · C is the set ′ composition C1|···|Ck|C1|···|Cr. Given a set composition C |= M ⊔ N, we let C|M = C1 ∩M|C2 ∩M|···|Ck ∩M and C/M = C1 ∩N|···|Ck ∩N, where it is understood that we remove any empty blocks from the composition. For example, ∆{1,3,5},{2,4,6}(12|35|46) = 1|35 ⊗ 2|46. Then S is a linearized Hopf monoid.

Example 13. Given a ﬁnite set N, let GN denote the collection of graphs with vertex set N. Given a bijection σ : M → N, and a graph g ∈ GM , −1 −1 deﬁne Gσ(g) to be the graph on N with edges ij if and only if σ (i)σ (j) is an edge of g. Then this gives rise to a pointed set species G, the species of graphs. The species of graphs is a linearized Hopf monoid. The product is given by g · h = g ⊔ h, the disjoint union of graphs. Given a graph g, and S ⊆ N, g|S is the induced subgraph on S, and g/S is the induced subgraph on N −S.

Example 14. Given a ﬁnite set N, let PN denote all partial orders on N. Given a bijection σ : M → N, and a partial order p ∈ PM , deﬁne Pσ(g) to be 12 JACOB A. WHITE

−1 −1 the partial order on N given by x ≤ y if and only if σ (x) ≤p σ (y). Then this gives rise to a pointed set species P, the species of posets. The species of posets is a linearized Hopf monoid. The product is given by p · q = p ⊔ q, the disjoint union of partial orders. Given a partial order p, and S ⊆ N, we deﬁne p|S = p/S =0 if S is not an order ideal of p. If S is an order ideal, then p|S and p/S are the induced subposets on S and N − S respectively.

Example 15. Given a finite set N, let MN denote the collection of matroids with ground set N. Given a bijection σ : M → N, and a matroid m ∈ MM , define Mσ(m) to be the matroid on N where a set S is a basis if and only if σ−1(S) is a basis of m. Then this gives rise to a pointed set species M, the species of matroids. The species of matroids forms a linearized Hopf monoid. The product is given by the direct sum operation. Given a matroid m, and S ⊂ N, we define m|S to be the restriction, and m/S to be the contraction of matroids. 3.2. Characters and Combinatorial Hopf Monoids. A morphism of species ϕ : F → G is a natural transformation. Given a morphism of pointed set species ϕ : F → G, there is a linearization given by K(ϕ) : K(F) → K(F) given by taking ϕN : FN → GN and extending by linearity to induce a linear transformation K(ϕ)N : K(F)N → K(G)N . We refer to a morphism of linear species as linearized if it is the linearization of a morphism of pointed set species. A morphism ϕ : H → E, where H is a Hopf monoid in species, is a character if for all disjoint finite sets M and N, and all x ∈ HM and all y ∈ HY , we have ϕN (x) · ϕM (y)= ϕM⊔N (x · y). By an abuse of notation, we will write ϕ(h) in place of ϕN(h)(h), when no confusion will arise. Now we define a combinatorial Hopf monoid. A combinatorial Hopf monoid is a linearized Hopf monoid H with a linearized character ϕ. This means that ϕ(h)=0 or ϕ(h) = 1 for every H-structure. The motivation is that many combinatorial Hopf algebras are studied where the character ϕ only takes on the values 0 and 1. First, we mention two examples of characters that are defined for every Hopf monoid. Example 16. Let H be a Hopf monoid. We define ζ(h) = 1 for every H-structure h. We refer to ζ as the zeta character.

Example 17. Let H be a Hopf monoid. We say a H-structure h ∈ HN is totally reducible if |N| = 1, or there exists a nontrivial decomposition N = S ⊔ T , and totally reducible elements x ∈ HS and y ∈ HT such that h = x · y. We deﬁne 1 if h is totally reducible χ(h)= (0 otherwise. We call χ the chromatic character, and (H,χ) is always a combinatorial Hopf monoid. For instance, if we let H = G, then a graph g is totally reducible if COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 13 and only if it is edgeless. If we let H = P, then a poset p is totally reducible if and only if it is an antichain. Finally, if H = M, then a matroid m is totally reducible if and only if is a direct sum of loops and coloops, which means m has a unique basis. These characters were studied in context of Hopf algebras in [2], and in the context of Hopf monoids in [1]. These characters give rise to the chromatic symmetric function, Gessel’s P -partition enumerator, and the Billera-Jia-Reiner invariant of a matroid, as we discuss below.

3.3. Hopf submonoids. Finally, given a species P, a species Q is a sub- species of P if the following two conditions are satisﬁed:

(1) For every ﬁnite set N, we have QN ⊆ PN . (2) For every bijection σ : M → N, and every q ∈ QN , we have Qσ(q)= Pσ(q). Suppose that P is a linearized combinatorial Hopf monoid. Then Q is a submonoid if the following two conditions are satisﬁed:

(1) For every pair of disjoint ﬁnite sets M and N, and structures p ∈ QM and q ∈ QN , we have p · q ∈ QM⊔N . (2) For every pair of disjoint ﬁnite sets M and N, and every Q structure q ∈ QM⊔N , if ∆M,N (q) 6= 0, then q|S and q/S are both Q structures.

3.4. The chromatic quasisymmetric function. Recall that a combinatorial Hopf algebra is a graded connected Hopf algebra H with a linear character ϕ : H → K. Aguiar, Bergeron, and Sottile [2] showed that there was a Hopf algebra homomorphism from any Hopf algebra H to the Hopf algebra of quasisymmetric functions. In this section we review a construction that associates a Hopf algebra to a Hopf monoid. The end result is that, given a linearized combinatorial Hopf monoid H, and a H-structure h, there is a quasisymmetric function invariant Ψϕ(h, x) associated to h. The advantage of working with species is that we are able to describe this invariant as enumerating colorings. We shall refer to this invariant as the ϕ-chromatic quasisymmetric function of H. Aguiar and Mahajan [4] deﬁned a Fock functor, which sends species to graded vector spaces, and sends linear Hopf monoids to graded Hopf algebras. Thus, given a Hopf monoid in pointed set species, we have F(K(H)) is a graded Hopf algebra. We review the explicit description of the resulting K Hopf algebra. Write F( (H)) = n≥0 Hn, where Hn is the free vector space with basis indexed by isomorphism classes of H-structures on [n]. The multiplication and comultiplication areL inherited from the Hopf monoid. That is, we multiply and comultiply basis elements by performing the product and coproduct on representatives of the isomorphism classes. Suppose that we have a character ϕ : H → E. Given an isomorphism class [h], we can deﬁne a characterϕ ˆ : F(H) → K byϕ ˆ([h]) = ϕ(h). Thus (F(H), ϕˆ) is a combinatorial Hopf algebra. 14 JACOB A. WHITE

We recall the quasisymmetric function associated to a character on a combinatorial Hopf algebra (H, ϕ). Given h ∈ Hn, n ⊗k k−1 (2) Ψϕ(h, x)= (ϕ ◦(ρα1 ⊗ρα2 ⊗···⊗ραk )◦∆ )(h)Mα α +α +···+α =n Xk=1 1 2 X k where ρn is the projection map from H to Hn, and the summation is over all integer compositions α of n. Recall that ℓ(α) is the length of a composition. Thus, given a linearized combinatorial Hopf monoid H, there is an associated combinatorial Hopf algebra (F(H), ϕˆ), and to every H-structure h there is an associated quasisymmetric function Ψϕ(h, x), which we call the ϕ-chromatic quasisymmetric function. We also deﬁne χϕ(h,n)=psn Ψϕ(h, x), where ps is the principal specialization. Then χϕ(h,n) is a polynomial in n, which we call the ϕ-chromatic polynomial. We give two alternative formulas for Ψϕ(h, x). Given a set composition

C |= N(h), we define ∆C = (idC1 ⊗∆C/C1 ) ◦ ∆C1,C2∪C3∪···∪Ck . We also let k ϕC = ( i=1 ϕCi ) ◦ ∆C . We use this notation for various proofs. We also write ϕS,T = ϕ . We say a set composition is ϕ-proper if ϕC (h) = 1. N S|T Let N be a finite set, and let h ∈ HN . Let f : N → N be a function. There are only finitely many colors that are used: let i1 < · · · < ik be those colors. We let Ni be the set of vertices v such that f(v) ≤ i. We call f a proper ϕ-coloring of h if ϕ(h[Ni,Ni+1]) = 1 for all i. Theorem 18. Let H be a linearized combinatorial Hopf monoid. Fix a finite set N, and h ∈ HN . Then

Ψϕ(h, x)= Mα(C) = xf(n). C is ϕX-proper f is ϕX-proper nY∈N Similarly, χϕ(h,n) is the number of ϕ-proper colorings f such that f(N) ⊆ [n]. For example, consider the combinatorial Hopf monoid (G,χ). Then the resulting invariant Ψϕ(g, x) is the chromatic symmetric function. The invariant Ψζ (p, x) for the combinatorial Hopf monoid (P,ζ) enumerates p- partitions. Finally, for the combinatorial Hopf monoid (M,χ), the invariant Ψχ(m, x) is the Billera-Jia-Reiner quasisymmetric function of a matroid [9].

Proof. The coproduct of the Hopf algebra is given by ∆(h) = S⊆N h|S ⊗ h/S, where we are viewing the elements h, h|S, and h/S up to isomorphism. So P

ϕ|α| ◦ ρα ◦ ∆|α|−1(h) = (ϕ|α| ◦ ρα) h  [Si−1,Si] S1⊆S2X⊆···⊆S|α| O   = ϕ(h[Si−1,Si]) S1⊆S2X⊆···⊆S|α| Y COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 15 where the last sum is over all ﬂags where |Si \ Si−1| = αi for all i.

We say a flag F• = S1 ⊆ S2 ⊆···⊆ S|α| is ϕ-proper if ϕ(h[Si−1,Si]) = 1 for all i. Then ϕ|α| ◦ ρα ◦ ∆|α|(h) is the number of ϕ-proper flags of type α. Thus F• is ϕ-proper if and only if C(F•) is ϕ-proper. Thus we have a bijection between ϕ-proper flags of type α and ϕ-proper compositions of type α. This implies the first equality. Given a coloring f, we let i1, · · · , ik be the only colors that f uses. Then −1 we have a set composition C = C1|C2|···|Ck, where Cj = f (ij). Given a composition C, we let hi = h[C1∪···∪Ci−1,C1∪···∪Ci]. We observe that ϕ(hi) = 1 for all i, since f is a coloring. We associate the composition C to f, denoting it by C(f). Given any set composition C, there exists a coloring f with C(f)= C only when ϕC (h) = 1. Thus C(f) is ϕ-proper. Given a ϕ-proper composition C, we have

xf = Mα f:CX(f)=C where αi = |Ci|. Thus, f xf = α cαMα, where cα is the number of ϕ-proper compositions of type α. This implies the second equality. P P 4. Geometric Realization Now we introduce geometric realizations for a combinatorial Hopf monoid. Given a combinatorial Hopf monoid (H, ϕ), a geometric realization consists of a pair (K, Σϕ), where: (1) The structures of the linearized species K are colored relative simplicial complexes. (2) The natural transformation Σϕ : H → K satisﬁes the additional property that Ψϕ(h, x) = Hilb(Σϕ(h), x) for every H-structure h. Our focus will be to describe a geometric realization that comes from a certain construction which relies on the fact that we already have a description of Ψϕ(h, x) as a summation over set compositions.

4.1. Convex characters. Given a combinatorial Hopf monoid (H, ϕ), and an H-structure h ∈ HN , we let

Aϕ(h)= {C |= N : ϕC (h) = 1} be the album associated to h. We see that this is a natural album which arises when studying Ψϕ(h, x). Moreover, F (Aϕ(h)) is a collection of sets, and if we deﬁne κ : 2N → [|N|] via κ(S) = |S|, then it is not hard to show that κ is a proper coloring for F (Aϕ(h)). So if Aϕ(h) is convex, then F (Aϕ(h)) is a colored relative complex. Our next goal is to determine precisely when this happens. Deﬁnition 19. We say that ϕ is a convex character if, for every H-structure h such that ϕ(h) = 1, and any two set compositions C ≤ C′ such that ′ C,C |= N(h), if ϕC′ (h) = 1, then ϕC (h) = 1. It is a balanced convex 16 JACOB A. WHITE character if, for every H-structure h such that ϕ(h) = 1, we have |N(h)| = 1 or ϕC (h) = 1 for some proper C ⊢ N(h). The word convex comes from the following proposition. Proposition 20. Let (H, ϕ) be a combinatorial Hopf monoid. Then the albums Aϕ(h) are convex for all H-structures h if and only if ϕ is a convex character.

Proof. First, suppose that Aϕ(h) is convex for every H-structure h. Let h be a H-structure such that ϕ(h) = 1. Let C ≤ C′ be two set compositions ′ ′ such that C,C |= N(h) and ϕC′ (h) = 1. Then C ∈Aϕ(h). Since ϕ(h) = 1, ′ N ∈ Aϕ(h). Since N ≤ C ≤ C , it follows that C ∈ Aϕ(h), and hence ϕC (h) = 1. Therefore ϕ is a convex character. Now suppose that ϕ is a convex character. Let N be a ﬁnite set and let ′ ′′ ′′ h ∈ HN . Let C ≤ C ≤ C |= N, with C,C ∈ Aϕ(h). Let F (C) = S1 ⊂

F2 ⊂ ··· ⊂ Sk. Since ϕC (h) = 1, we know ϕ(h[Si−1,Si]) = 1 for all i. We ′ ′′ ′′ observe that ϕC (h[Si−1,Si]) = 1 for all i. Since C[S ,S ] ≤ C[S ,S ] [Si−1,Si] i−1 i i−1 i ′ and ϕ is a convex character, we see that ϕC (h[Si−1,Si]) = 1 for all i. [Si−1,Si] ′ Then ϕC′ (h) = 1, so C ∈Aϕ(h). Now we give sufficient conditions for a character to be convex. These conditions are often easier to verify in practice than the definition of convex. Proposition 21. Let (H, ϕ) be a combinatorial Hopf monoid. Suppose that, for every finite set N, every decomposition N = S ⊔ T , and every h ∈ HN , if ϕ(h) = 1 and ∆S,T (h) 6= 0, then ϕS|T (h) = 1. Then ϕ is a convex character. Moreover, suppose that, for every H- structure h with ϕ(h) = 1, we have ∆S,T (h) 6= 0 for some nontrivial decomposition N(h) = S ⊔ T or |N(h)| = 1. Then ϕ is a balanced convex character. ′ Proof. Fix a finite set N. Let h ∈ HN with ϕ(h) = 1. Let C ≤ C |= N be two compositions with ϕC′ (h) = 1. Hence ∆C′ (h) 6= 0. By properties of coproduct, it follows that ∆C (h) 6= 0. Applying induction to our condition, we conclude that ϕC (h) = 1. For the last part, if we know ∆S,T (h) 6= 0 for some nontrivial decomposition N(h) = S ⊔ T , then ϕS|T (h) = 1, so we have a nontrivial composition C ⊢ N(h) with ϕC (h) = 1. Thus ϕ is a balanced convex character. Theorem 22. Let H be a linearized Hopf monoid. Then ζ is a convex character, and χ is a balanced convex character. Proof. First we focus on ζ. Let h be a H-structure with ζ(h) = 1. Let N(h)= S ⊔ T with ∆S,T (h) 6= 0. Then ζS,T (h) = 1, so ζ is convex. Let h be a H-structure with χ(h) = 1. Then there exists a linear order

ℓ on N(h), and objects hi ∈ H{ℓi} such that h = h1 · h2 · · · h|N|. Since H is COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 17

connected, it follows that ∆∅,{ℓi}(hi) = 1⊗hi and ∆{ℓi},∅(hi)= hi ⊗1. By the compatability with products and coproducts, given N(h)= S ⊔ T, we have ∆S,T (h) = ( i∈S hi) ⊗ ( j∈T hj). This implies that ∆S,T (h) 6= 0. It also implies that h| and h/S are both totally reducible, and hence ϕ (h) = 1. QS Q S,T Thus by Proposition 21, χ is a balanced convex character. 4.2. The geometric realization. Now we give the geometric realization. Definition 23. Let (H, ϕ) be a combinatorial Hopf monoid. Let N be a finite set, and let V be the collection of all proper subsets of N, and define κ : V → [|N|] by κ(S) = |S|. Given h ∈ HN , define the ϕ-coloring complex of h by

Σϕ(h)= F (Aϕ(h)) k

= {C1 ⊂ S2 ⊂···⊂ Sk ⊂ N : ϕSi\Si−1 ◦ ∆S1|S2\S1|···|N\Sk (h) = 1}. ! Oi=1 Lemma 24. Let (H, ϕ) be a combinatorial Hopf monoid. Let N be a finite set, and let V be the collection of all proper subsets of N, and define κ : V → [|N|] by κ(S) = |S|. Given h ∈ HN , let Σϕ(h) be the ϕ-coloring complex. Then Σϕ(h) is a colored relative simplicial complex. Proof. Since ϕ is a convex character, it follows from Proposition 20 that Aϕ(h) is a convex album for all h. By Proposition 9 F (Aϕ(h)) is a relative simplicial complex. Given a flag F• = F1 ⊂ F2 ⊂ ··· ⊂ Fk ∈ Σϕ(h), we see that κ(F•) = {|F1|, |F2|,..., |Fk|}. Hence Σϕ(h) is a colored relative simplicial complex.

Theorem 25. Let (H, ϕ) be a combinatorial Hopf monoid. Let KN be the set of all relative simplicial complexes Φ such that Φ is a subcomplex of the Coxeter complex of type A. Let Σϕ : H → K be given by Σϕ(h)= F (Aϕ(h)). Then (K, Σϕ) is a geometric realization of H. Proof. We see that K is a species, whose elements are relative simplicial complexes. Moreover, given a relative simplicial complex Φ such that Φ ∈ KN , we know that the vertices of Φ are subsets of N. A natural coloring comes from κ : V (Φ) → [|N|] given by κ(S)= |S|. One can check that Φ is a colored relative simplicial complex. Hence K is a species whose structures are colored relative simplicial complexes. Let h ∈ HN for some finite set N. We know from Lemma 24 that Σϕ(h) ∈ KN . The reader can check that Σϕ : H → K is a natural transformation. We show that Ψϕ(h, x) = Hilb(Σϕ(h), x) for every h-structure. We do this by letting S ⊆ [|N|], and considering the coefficient of Mα on both sides. The coefficient of Mα in Hilb(Σϕ(h), x) is the number of ϕ-proper flags of type α. In the proof of Theorem 18, we showed that this was the coefficient of Mα in Ψϕ(h, x). 18 JACOB A. WHITE

Example 26. Let G be the combinatorial Hopf monoid of graphs. Then the character χ is balanced convex. We see that Aχ(g) is the collection of set compositions where each part is an independent set. Then the corresponding relative simplicial complex Σχ(g) consists of chains S1 ⊂ S2 ⊂···⊂ Sk ⊂ N such that Si \ Si−1 is an independent set for all i. One presentation for Φ(g) is as the pair (Σ, Γ(g)), where Σ is the Coxeter complex of type A, and Γ(g) is collection of ﬂags S1 ⊂ S2 ⊂ ··· ⊂ Sk ⊂ N where Si+1 \ Si must contain an edge for some i. The subcomplex Γ(g) is the coloring complex, as introduced by Steingr´ımsson [21]. 4.3. Balanced realizations. We also discuss a condition which ensures all the geometric realizations are balanced complexes. Theorem 27. Let H be a Hopf monoid, and let ϕ be a convex character. Then Σϕ(h) is balanced for every H-structure h if and only if ϕ is a balanced convex character.

Proof. Suppose that Σϕ(h) is balanced for every H-structure h. Let N be a finite, and let h ∈ HN . Suppose that ϕ(h) = 1. This implies that Σϕ(h) is non-empty. However, then the definition of being balanced, with color set [|N|] implies that there must be a face F with color set [|N|]. Then by definition, we have ϕC(F )(h) = 1 and C(F ) |= N. This gives a nontrivial composition unless |N| = 1. Now suppose that ϕ is a balanced convex character. Let F• ∈ Σϕ(h), with F• = S1 ⊂ S2 ⊂ ··· ⊂ Sk. Then ϕ(h|Si /Si−1) = 1 for all i. Since ϕ is balanced and convex, we may assume by induction on |N(h)| that

Σϕ(h|Si /Si−1) is balanced. In particular, there exists a ﬂag i F• = Fi,1 ⊂ Fi,2 ⊂···⊂ Fi,ij ⊂ Si \ Si−1 i with F• ∈ Σϕ(h|Si /Si−1) and such that |Fi,j \ Fi,j−1| = 1 for all j. Then we see that the ﬂag

F1,1 ⊂···⊂ F1,1j ⊂ S1 ∪ F2,1 ⊂···⊂ S1 ∪ F2,2j ⊂···⊂ Sk ∪ Fk+1,(k+1)j is a ﬂag in Σϕ(h) that contains F•. Hence Σϕ(h) is balanced.

4.4. Some consequences. Now we apply knowledge about balanced simplicial complexes to derive inequalities for Ψϕ(h, x) and χϕ(h,x). Lemma 28. Suppose that (H, ϕ) is a combinatorial Hopf monoid, and ϕ is a convex character. Let h be a H-structure. Then

Hilb(dcone(Σϕ(h)),x)= χϕ(h,x + 1).

Proof. First, we observe that χϕ(h,n) = psnΨϕ(h, x). Then the result follows from Theorem 25, and Lemma 7. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 19

Theorem 29. Let (H, ϕ) be a combinatorial Hopf monoid. Suppose that ϕ is a balanced convex character. Let h be a H-structure. Then Ψϕ(h, x) is n x M-increasing. Also, if we write χϕ(h,x) = i=0 fi i , then (f0,...,fn) is strongly ﬂawless. P Proof. Since ϕ is convex, we know that Ψϕ(h, x) = Hilb(Σϕ(h), x), where Σϕ(h) is the ϕ-proper coloring complex of h. Since ϕ is a balanced convex character, Σϕ(h) is balanced. We know by Proposition 6 that Hilb(Σ, x) is M-increasing for any balanced relative simplicial complex. Also, a balanced complex is pure, and so dcone(Σϕ(h)) is also pure. Hence the f-vector of dcone(Σϕ(h)) is strongly ﬂawless. The second result follows from Lemma 28.

This implies that the f-vector for the chromatic polynomial of a graph, and the order polynomial of a poset are both strongly ﬂawless. These both appear to be new results. It also implies that the chromatic symmetric function, the P -partition enumerator, and the Billera-Jia-Reiner invariant of a matroid are all M-increasing. Of course, since these invariants are all known to be F -positive, the fact that they are M-increasing is not surprising.

5. Coloring complexes A natural question is to determine what types of relative simplicial complexes arise from our geometric construction. The aim of this section is to characterize such complexes completely. We shall temporarily refer to the resulting complexes as abstract coloring complexes. Now we define abstract coloring complexes. Let N be a finite set. Given flags E• = E1 ⊂ E2 ⊂···⊂ Ek ⊆ N and F• = F1 ⊂ F2 ⊂···⊂ Fm ⊆ N, such that Ea = Fc for some a and b, then we define their exchange with respect to Ea to be the flag

E• ·a F• = E1 ⊂···⊂ Ea ⊂ Fc+1 ⊂ Fc+2 ⊂···⊂ Fm. Let F be a collection of flags of subsets of N. We say that F satisfies the flag exchange property if whenever we have two flags E•, F• ∈ F such that Ei = Fj for some i and j, then E• ·i F• ∈ F. Definition 30. Let 2N be the set of all flags. This is a simplicial complex. Then a relative simplicial complex Φ is an abstract coloring complex if it is a subcomplex of 2N which satisfies the flag exchange property. Proposition 31. Let (H, ϕ) be a combinatorial Hopf monoid, and suppose that ϕ is a convex character. Let h be an H-structure. Then Σϕ(h) is an abstract coloring complex.

Proof. Let E•, F• ∈ Σϕ(h) such that there exists i with Ei = Fi. Then

ϕ(h[Ei−1,Ei]) = 1 for all i. Similarly, ϕ(h[Fi−1,Fi]) = 1 for all i. This implies that E• ·i F• ∈ Σϕ(h). 20 JACOB A. WHITE

Thus, every ϕ-coloring complex is an abstract coloring complex. For the sequel, we abuse terminology and write coloring complex in place of abstract coloring complex. We show that every coloring complex is a ϕ- coloring complex for some combinatorial Hopf monoid H. We do this by showing that the species of coloring complexes is itself a combinatorial Hopf monoid. For a finite set N, let CN be the set of coloring complexes on N. Note that we are treating the void complex as the base point 0. Let F• be a flag of subsets F1 ⊂ F2 ⊂···⊂ Fk ⊂ N of N. For a subset S ⊂ N, we let F• ∩ S = F1 ∩ S ⊆ F2 ∩ S ⊆···⊆ Fk ∩ S ⊆ S. We view this as a proper flag of S by removing duplicates and deleting S if necessary. Given a relative coloring complex Σ on N, and a subset S ⊆ N, we let

Φ|S = {F• ∩ S : F• ∈ Φ,S ∈ F•} and

Φ/S = {F• ∩ (N \ S) : F• ∈ Φ,S ∈ F•}. These two complexes are the restriction and contraction with respect to S. Example 32. Let Φ be the simplicial complex from the middle of Figure 6.2. We observe that Φ is a coloring complex. We see that Φ|{a,d} = ∅, since {a, d} is not a vertex of Φ. Hence ∆{a,d},{b,c}(Φ) = 0. On the other hand, A(Φ|{a,b,c})= {a|b|c, a|c|b, c|a|b, ab|c, ac|b, c|ab}. Proposition 33. Let Φ be a coloring complex on a ﬁnite set N, and let S ⊆ N. Then Φ|S is a coloring complex on S, and Φ/S is a coloring complex on N \ S. Moreover, if Φ is balanced, and S is an element of some face of Φ, then Φ|S and Φ/S are also balanced.

Proof. Let E•, F• ∈ Φ|S such that there exists i with Ei ∈ F•. Then there exists D• ∈ Φ such that S ∈ D• and D• ∩ S = E•. Similarly, there exists G• ∈ Φ such that S ∈ G• and G• ∩ S = F•. We see that Ei ∈ D• and Ei ∈ G•. Since Φ is a coloring complex, we have D• ·i G• ∈ Φ. Since S ∈ D• ·i G•, we see that E• ·i F• = (D• ·i G•) ∩ S ∈ Φ|S. Thus Φ|S is a coloring complex. Let E•, F• ∈ Φ/S such that there exists i with Ei ∈ F•. Then there exists D• ∈ Φ such that S ∈ D• and D• ∩ S = E•. Similarly, there exists G• ∈ Φ such that S ∈ G• and G• ∩ S = F•. We see that Ei ∈ D• and Ei ∈ G•. Choose j such that Dj = Ei. Since Φ is a coloring complex, we have D• ·j G• ∈ Φ. Since S ∈ D• ·j G•, we see that E• ·i F• = (D• ·j G•)∩S ∈ Φ/S. Thus Φ/S is a coloring complex. Suppose that Φ is balanced, and that there is a face σ ∈ Φ with S ∈ σ. Let E• ∈ Φ|S. Then there exists a ﬂag F• ∈ Φ such that S ∈ F• and F• ∩S = E•. Since Φ is balanced, there exists a reﬁnement G• such that F• ⊂ G• and Gi \Gi−1 is a singleton for all i. Then we see that E• ⊂ G• ∩S ∈ Φ|S. Hence Φ|S is balanced. A similar argument shows that Φ/S is also balanced. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 21

abc a

c acd

Figure 1. A product of two coloring complexes.

Given disjoint ﬁnite sets N and M, and two relative coloring complexes Σ ∈ CM and Γ ∈ CN , we deﬁne Σ · Γ= {F : F ∩ M ∈ Σ and F ∩ N ∈ Γ}. Example 34. Let Φ be given by A(Φ) = {a|b, ab}, and Σ be given by A(Σ) = {c|d}. Then Φ · Σ appears in Figure 1.

Proposition 35. Given disjoint ﬁnite sets N and M, and two relative coloring complexes Σ ∈ CM and Γ ∈ CN , then Σ · Γ is a coloring complex on N ⊔ M. Moreover, if Σ and Γ are balanced, then so is Σ · Γ.

Proof. Let E•, F• ∈ Σ · Γ such that there exists i with Ei ∈ F•. We let C• = E• ∩ M and D• = E• ∩ N. Similarly, we define G• = F• ∩ M and H• = F• ∩ N. We see that there exists an integer j such that Cj = Ei ∩ M. Then Cj ∈ G•, and C• ·j G• ∈ Σ. There also exists and integer k such that Dk = Ei ∩ N. Then Dk ∈ H•, and D• ·k H• ∈ Γ. Finally, we observe that (E• ·i F•)∩M = C• ·j G• and (E• ·i F•)∩N = D• ·k H•. Thus E• ·i F• ∈ Σ·Γ. Therefore Σ · Γ is a coloring complex. Now suppose that Σ and Γ are balanced, and let F• ∈ Σ · Γ. Write F• = F1 ⊂ F2 ⊂···⊂ Fk. Suppose that F• is maximal. If |Fi \ Fi−1| = 1 for all i, then F• has dimension (dimΣ + dimΓ) as required. So suppose |Fi \ Fi−1|≥ 2 for some i. If Fi−1 \ Fi contains elements from both M and N, then it is possible to refine F•, contradicting the fact that it is maximal. Hence, without loss of generality, we may assume that Fi \ Fi−1 ⊂ M. Let E• = F• ∩ M. Choose j such that Ej = Fi ∩ M. Then |Ej \ Ej−1| ≥ 2. Since Σ is balanced, E• is not a facet, and it is possible to refine it further, by adding a new set E between Ej−1 and Ej. However, then we see that we can add E ∪ Fi−1 between Fi−1 and Fi, contradicting the assumption that F• was maximal. Therefore Σ · Γ must be balanced. Theorem 36. The species C of coloring complexes is a linearized Hopf monoid, and the species BC of balanced coloring complexes is a Hopf submonoid. 22 JACOB A. WHITE

Proof. Given finite sets M and N, and coloring complexes Σ ∈ CM and Γ ∈ CN , we see from Proposition 35 that Σ · Γ ∈ CM⊔M . We define µM,N by µM,N (Σ ⊗ Γ)=Σ · Γ, and extend by linearity. Given finite sets L, M and N, and coloring complexes Π ∈ CL,Σ ∈ CM , and Γ ∈ CN , we see that (Π · Σ) · Γ = Π · (Σ · Γ) = {E• : E• ∩ L ∈ Π, E• ∩ M ∈ Σ, E• ∩ N ∈ Γ}. Extending by linearity, we see that multiplication is associative. Note here the that C∅ has two elements. One of the elements is the void complex, which we identify with 0. The other element is {∅}, the collection consisting of only the empty flag (observe that we require flags to consist of proper subsets of N, and in this case there are no proper subsets of N). We observe that {∅} is a multiplicative identity. Given finite sets M and N, and a coloring complex Φ ∈ CM⊔N , let ∆M,N (Φ) = Φ|S ⊗Φ/S. Extend by linearity to obtain a map ∆M,N : CM⊔N → CM ⊗ CN . Let L, M, N be finite sets, and let Φ ∈ CL⊔M⊔N . Define Φ1 = {F• ∩ L : F• ∈ Φ,L,L ∪ M ∈ F•}, Φ2 = {F• ∩ M : F• ∈ Φ,L,L ∪ M ∈ F•}, and Φ3 = {F• ∩ N : F• ∈ Φ,L,L ∪ M ∈ F•}. Then we observe that

((∆L,M ⊗ idN ) ◦ ∆L∪M,N )(Φ) = Φ1 ⊗ Φ2 ⊗ Φ3

= ((idL ⊗∆M,N ) ◦ ∆L,M∪N )(Φ). Thus, extending by linearity, we see that the coproduct is coassociative. We observe that the counit axiom is also satisﬁed. Let A,B,C and D be ﬁnite sets, and let Σ ∈ CA⊔B and Γ ∈ CC⊔D. We show that (Σ · Γ)|A⊔C =Σ|A · Γ|C . Let F• ∈ (Σ · Γ)|A⊔C . Then there exists E• ∈ Σ · Γ such that A ⊔ C ∈ E• and E• ∩ (A ⊔ C)= F•. Since E• ∈ Σ · Γ, we know that E• ∩(A⊔B) ∈ Σ. Since A⊔C ∈ E•, we see that A ∈ E• ∩(A⊔B). Hence F• ∩ A = E• ∩ (A ⊔ C) ∩ A = F• ∩ A ∈ Σ|A. By a similar argument, F• ∩ C ∈ Γ|C . Thus, we see that F• ∈ Σ|A · Γ|C . Let F• ∈ Σ|A · Γ|C . Then F• ∩ A ∈ Σ|A. Thus, there exists D• ∈ Σ with A ∈ D• and D• ∩ A = F• ∩ A. Similarly, there exists E• ∈ Γ with C ∈ E• and E• ∩ C = F• ∩ C. We see that there exists an i such that Di = A, and a j such that Ej = C. Let

G• = F• ∪ {Dk ∪ A ∪ C : k > i} ∪ {Ek ∪ A ∪ B ∪ C : k > j}.

By construction, G• ∩ (A ∪ B) = D•, and G• ∩ (C ∪ D) = E•. Hence G• ∈ Σ · Γ. Moreover, G• ∩ (A ∪ C)= F•. Thus F• ∈ (Σ · Γ)|A∪C . Hence we have shown that (Σ · Γ)|A⊔C = (Σ|A) · (Γ|C ). A similar proof shows that (Σ · Γ)/(A ⊔ C) = (Σ/A) · (Γ/C). Thus, the multiplication and comultiplication are compatible. For the ﬁnal claim, it is enough to notice that the set of balanced coloring complexes BCN is a subset of CN , and that this inclusion induces an inclusion BC ⊂ C of species. We have already shown that the product and coproduct of balanced coloring complexes remains balanced, and hence the result follows. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 23

Let N be a ﬁnite set, and let ϕ : CN → EN be given by 1 if Σ is a simplicial complex ϕ(Σ) = (0 otherwise. ϕ(Σ) = 1 if Σ is a simplicial complex, and 0 otherwise.

Lemma 37. Let Φ be a coloring complex on N. Let C |= N. Then ϕC (Φ) = 1 if and only if F (C) ∈ Φ.

Proof. Let C |= N. Let F• = F (C). Suppose that F• ∈ Φ. Recall that

Φ[Fi,Fi+1] = {E• ∩ Ci+1 : Fi, Fi+1 ∈ E•}. We see that F• ∩ Ci+1 = {Ci+1}, which implies that ∅∈ Φ[Fi,Fi+1]. Thus Φ[Fi,Fi+1] is a simplicial complex for all i. Hence ϕC (Φ) = 1.

Now suppose that ϕC (Φ) = 1. Then we see that ∅ ∈ Φ[Fi,Fi+1] for all i. i i i For each i, there exists E• such that Ej = Fi, Ej+1 = Ei+1 for some j. Since i Φ is closed under exchange operations, then exchanging the diﬀerent E• for various i will result in F•. Hence F• ∈ Φ. Proposition 38. The morphism ϕ : C → E is a convex character, and the morphism ϕ : BC → E is a balanced convex character. Proof. Let Φ be a coloring complex on N and suppose ϕ(Φ) = 1. Let ′ C ≤ C |= N, and suppose that ϕC′ (Φ) = 1. Then we see that Φ is a simplicial complex. Moreover, by Lemma 37, F (C′) is a face of Φ. However, ′ since F (C) ⊆ F (C ), it follows that F (C) ∈ Φ. By Lemma 37, ϕC (Φ) = 1. Therefore ϕ is a convex character. The character ϕ restricts to a convex character on BC as well. Let Φ be a balanced coloring complex on N, and suppose ϕ(Φ) = 1. Then Φ is a balanced simplicial complex. If |N| > 1, it has a facet F which corresponds to a composition C where each block is a singleton. Then ϕC (Φ) = 1, and so ϕ is a balanced convex character. Theorem 39. The pair (C, ϕ) is the terminal object in the category of combinatorial Hopf monoids with convex characters. In particular, for every Hopf monoid H, the map Σϕ : H → C given by Σϕ(h) = F (Aϕ(h)) is a morphism of Hopf monoids. Similarly, the pair (BC, ϕ) is the terminal object in the category of combinatorial Hopf monoids with balanced convex characters. Proof. Let (H, ϕ) be a combinatorial Hopf monoid with a convex character. Given a H-structure h, we see that Σϕ(h) is a coloring complex. Thus we do obtain a map Σϕ,N : HN → CN . The reader can verify that this map is natural in N, and hence we obtain a natural transformation Σϕ : H → C. We show that Σϕ is a morphism of Hopf monoids. ′ ′ Let h and h be H-structures with h ∈ HM and h ∈ HN . Let F• = F1 ⊂ ′ ′ F2 ⊂ ··· ⊂ Fk. Then F• ∈ Σϕ(h · h ) if and only if ϕC(F•)(h · h ) = 1. ′ Then ϕ((h · h )[Fi−1,Fi]) = 1 for all i if and only if ϕ(h[Fi−1∩M,Fi∩M]) = 24 JACOB A. WHITE

ϕ(h′ ) = 1 for all i. Thus F ∩ M ∈ Σ (h) and F ∩ N ∈ Σ (h′), [Fi−1∩N,Fi∩N] • ϕ • ϕ ′ ′ which is equivalent to F• ∈ Σϕ(h) · Σϕ(h ). Thus we have Σϕ(h · h ) = ′ Σϕ(h) · Σϕ(h ). Hence the morphism preserves multiplication. We focus on comultiplication. Let F• ∈ Σϕ(h)|S . Then there exists

E• ∈ Σϕ(h) such that S ∈ E• and E• ∩ S = F•. We see that (h|S)[Fi−1,Fi] = h[Ei−1,Ei] for i ≤ ℓ(F•). Then ϕ(h[Ei−1,Ei]) = 1 for all i, which implies that

ϕ((h|S )[Fi−1,Fi]) = 1 for all i. Thus F• ∈ Σϕ(h|S). Thus we conclude that Σϕ(h)|S ⊆ Σϕ(h|S). By a similar argument, Σϕ(h)/S ⊆ Σϕ(h/S). If Σϕ(h/S) = ∅ or Σϕ(h|S) = ∅, then we have ∆S,T (Σϕ(h)) = 0 = Σϕ(h|S ) ⊗ Σϕ(h/S). So assume both relative complexes are not the void complex. Let E• ∈ Σϕ(h|S) and let F• ∈ Σϕ(h/S). Deﬁne G• by G• = E1 ⊂ E2 ⊂ ··· ⊂ Ek ⊂ S ⊂ S ∪ F1 ⊂ S ∪ F2 ⊂ ··· ⊂ S ∪ Fr. Since

E• ∈ Σϕ(h|S), we have 1 = ϕ((h|S )[Ei−1,Ei]) = ϕ(h[Ei−1,Ei]) for all i. Since

F• ∈ Σϕ(h/S), we have 1 = ϕ((h/S)[Fi−1,Fi])= ϕ(h[S∪Fi−1,S∪Fi]) for all i. We see then that ϕ(h[Gi−1,Gi]) = 1 for all i. Thus G• ∈ Σϕ(h) with S ∈ G•. Then E• = G• ∩S ∈ Σϕ(h)|S , and F• = G• ∩(N \S) ∈ Σϕ(h)/S. Hence Σϕ(h|S)= Σϕ(h)|S and Σϕ(h/S)=Σϕ(h)/S. Therefore ∆S,T (Σϕ = (Σϕ ⊗ Σϕ) ◦ ∆S,T , and Σϕ is a morphism of Hopf monoids. Now we show that Σϕ : H → C is the unique Hopf monoid homomorphism with the property that ϕ(Σϕ(h)) = ϕ(h) for every H-structure h. Let Γ be another such Hopf monoid homomorphism. Let h be a H-structure. Let

F• be a ﬂag. Then by Lemma 37, F• ∈ Σϕ(h) if and only if ϕC(F•)(h) = 1. Since Γ is a morphism of Hopf monoids which commutes with ϕ, then

ϕC(F•)(Γ(h)) = 1. By Lemma 37, this holds if and only if F ∈ Γ(h). Thus, if ϕC(F•)(h) = 1, then F• is a face of both Σϕ(h) and Γ(h). On the other hand, if ϕC(F•)(h) = 0, then ϕC(F•)(Γ(h)) = 0, and F• is not a face of either of complex. Therefore Σϕ(h)=Γ(h). Finally, let H be a combinatorial Hopf monoid with a balanced convex character ϕ. Then we have a morphism Σϕ : H → C. Moreover, Σϕ(h) is balanced for all h, which means we have a morphism Σϕ : H → BC.

6. Applications 6.1. Graphs. We already discussed the standard character that yields the chromatic symmetric function. Here we detail one more interesting character that was ﬁrst studied by Aval, Bergeron, and Machacek [6]. Let

0 if g is a perfect matching ψ(g)= (1 otherwise. It is clear that ψ is a character. The resulting quasisymmetric function, Ψψ(g, x), enumerates functions f : V → N such that the induced subgraph on each color class is a perfect matching. Let ψ(g) = 1. Then g is a perfect matching. Let C ≤ C′ and suppose ′ that ψC′ (g) = 1. Hence, g restricted to each block of C is also a perfect COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 25

matching. This implies that g|Ci is also a perfect matching for each block of C. Hence ψC (g) = 1. Therefore ψ is a convex character. We see that Σψ(g) 6= ∅ only when g has a perfect matching. In those cases, the facets consist of set compositions where each block has size two. Hence, Σψ(g) is pure, but is balanced with respect to the color set 2, 4, 6,.... In particular, [Mα]Ψψ(g, x) =0 if α contains an odd part. If α ≤ β and β contains only even parts, then [Mα]Ψψ(g, x) ≤ [Mβ]Ψψ(g, x). We also get that the corresponding chromatic polynomial is strongly ﬂawless. Aval, Bergeron, and Machacek [6] showed that the polynomial can be computed from chromatic polynomials. They also give a large class of invariants coming from linearized characters. Let C be a collection of isomorphism classes of connected graphs. Then we can deﬁne ϕC by 1 if every component of g is isomorphic to a graph in C ϕC (g)= (0 otherwise.

We see that every linearized character must be of the form ϕC for some C.

In particular, χ = ϕ{K1} and ψ = ϕ{K2}, where Kn is the complete graph on n vertices. We are able to classify the balanced convex characters on graphs. Theorem 40. Let C be a collection of isomorphism classes of graphs. Then ϕC is a balanced convex character if and only if C satisﬁes the following two properties: (1) For every g ∈ C, and every induced connected subgraph h ∈ C, we have h ∈C. (2) The collection C contains K1.

Proof. First, let ϕC be a balanced convex character. Suppose that K1 6∈ C, and let g ∈ C be a graph of minimum size. Then ϕC (g) = 1. Since ϕC is balanced, there must be a nontrivial composition C such that ϕC (g) = 1, which implies that there is a smaller graph in C, a contradiction. Let g ∈ C be a graph on N, and let h be an induced connected graph on vertex set S. Let C = S|v1|···|vk, where v1, . . . , vk are the vertices of ′ N \ S in some order. Let C = w1|···|wm|v1|···|vk, where w1, . . . , wm are the vertices of S in some order. Then we have ϕC (g)=1= ϕC′ (g). Since ϕC is convex, this means that ϕC (g) = 1, which implies that ϕC(g|S) = 1. However, h = g|S, and thus h ∈ C. Thus C is closed under inclusion of induced connected subgraphs. Conversely, suppose that C is closed under inclusion of induced connected subgraphs, and contains K1. Let g be a graph for which ϕC (g) = 1. We see that any composition C |= N(g) into singletons satisﬁes ϕC (g) = 1. Let N(g) = S ⊔ T . Then we see that g|S is a disjoint union of connected components hi. Moreover, each is an induced subgraph of a component of g. Hence ϕ(g|S ) = 1. Similarly, ϕ(g/S) = 1. By Proposition 21, the character ϕC is balanced and convex. 26 JACOB A. WHITE

6.2. Acyclic mixed graphs. First, we define the combinatorial objects. Given a finite set N, a mixed graph is a triple (N, U, D), where U is a set of undirected edges, and D is a set of directed edges. A mixed graph is acyclic if it does not contain a directed cycle. There are two well-known polynomial invariants associated to acyclic acyclic mixed graphs: the weak and strong chromatic polynomial, both introduced in [7], motivated by work in [8]. Given an acyclic mixed graph g, the weak chromatic polynomial χ(g, k) counts the number of functions f : N → [k] subject to: (1) For every uv ∈ U, we have f(u) 6= f(v). (2) For every (u, v) ∈ D, we have f(u) ≤ f(v). The strong chromatic polynomialχ ¯(g, k) counts similar functions, only with strict inequalities for the second condition instead of the weak inequality. Naturally there are quasisymmetric function generalizations of both these invariants, which we introduce. Moreover, both invariants can be understood from the theory of combinatorial Hopf monoids, and come from convex characters, as we now demonstrate. Let MN be the set of all acyclic mixed graphs with vertex set N. Given a bijection σ : N → M, and an acyclic mixed graph g = (N, U, D) ∈ MN , ′ ′ ′ we define Mσ(g) = (M,U , D ), where U = {{σ(u),σ(v)} : {u, v} ∈ U} and ′ D = {(σ(u),σ(v)) : (u, v) ∈ D}. We see that we get a map Mσ : MN → MM . Thus M is a species. Now we turn M into a linearized Hopf monoid. Given disjoint sets M and N, and acyclic mixed graphs g ∈ MM and d ∈ MN , we define g · d by taking disjoint unions of edge sets. Likewise, given an acyclic mixed graph g = (M ⊔ N, U, D) ∈ MM⊔N , we define ∆M,N (g) as follows. First, if there exists (n,m) ∈ D with m ∈ M,n ∈ N, then ∆M,N (g) = 0. Otherwise, we let g|M be the induced subgraph on M, and g/M be the induced subgraph on N, and define ∆M,N (g)= g|M ⊗g/M. There are two interesting examples of characters we will be interested in. The first is χ. We see that χ(g) = 1 if and only if g has no edges. We define

0 if d has at least one undirected edge ϕ(g)= (1 otherwise.

Theorem 41. The pairs (M, ϕ) and (M,χ) are combinatorial Hopf monoids. Moreover, both ϕ and χ are balanced convex characters.

Proof. Let L, M and N be ﬁnite sets. Let g ∈ ML, h ∈ MM , and k ∈ MN . Then (g · h) · k = g · (h · k), as the result in both cases is an acyclic mixed graph that is the disjoint union of the three individual graphs. Let f ∈ ML⊔M⊔N . If there is a directed edge from a vertex in L to a vertex in M, or from a vertex in L to a vertex in N, or from a vertex in M to a COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 27 vertex in N, then

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(f) = 0

= ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(f). If not such directed edges exist, then

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(f)= f|L ⊗ f|M ⊗ f|N

= ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(f), and hence we have coassociativity. Now, we show compatability. Let A,B,C and D be ﬁnite sets, and let g ∈ MA⊔B and h ∈ MC⊔D. If there is a directed edge in g from a vertex in A to a vertex in B, or there is a directed edge in h from a vertex in C to a vertex in D, then

∆A⊔C,B⊔D(g · h) = 0

= ((µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(g ⊗ h). Otherwise,

∆A⊔C,B⊔D(g · h)= g|A · h|C ⊗ g|B · h|D

= ((µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(g ⊗ h). Thus we see that the product and coproduct are compatible. It is not hard to see that ϕ is multiplicative. To show that it is convex, suppose that g is an acyclic mixed graph where ϕ(g) = 1. Then g has no undirected edges. Let V = S ⊔ T , and suppose ∆S,T (g) 6= 0. Then ∆S,T (g)= g|S ⊗ g/S. Both g|S and g/S are induced subgraphs, and hence have no undirected edges. Thus ϕ(g|S )= ϕ(g/S) = 1. Thus by Proposition 21, the character ϕ is convex. It is also clear from the deﬁnition that for every acyclic mixed graph g, there exists a decomposition N(g)= S ⊔ T for which ∆S,T (g) 6= 0. For example, one can choose S to consists of the set of sinks of g with respect to the directed edges of g. Hence ϕ is a balanced convex character. The fact that χ is balanced convex follows from Theorem 22.

Given an acyclic mixed graph g with vertex set V , we see that Ψϕ(g, x) is a summation over all proper colorings of g. Similarly, Ψχ(g, x) is a weighted sum over strong proper colorings of g. As an example, if we let g be the graph on the left in Figure 6.2, then

Ψϕ(g, x) = 3M2,2 + 4M1,1,2 + 2M1,2,1 + 4M2,1,1 + 6M1,1,1,1

= 3F2,2 + F1,1,2 + 2F1,2,1 + F2,1,1 − F1,1,1,1. We have that

Ψχ(g, x)= M2,2 + 2M1,1,2 + 2M2,1,1 + 6M1,1,1,1.

Note that Ψϕ(g, x) is not F -positive. The relative simplicial complex in the middle of Figure 6.2 is the coloring complex of g with respect to ϕ, while 28 JACOB A. WHITE

ab ab abc a abc a b d ac ac

c acd c acd

a c cd cd

Figure 2. an acyclic mixed graph, and its two coloring complexes with respect to diﬀerent characters. the relative simplicial complex on the right is the coloring complex of g with respect toϕ ¯. Corollary 42. For every acyclic acyclic mixed graph g, the quasisymmetric functions Ψϕ(g, x) and Ψϕ¯(g, x) are both M-increasing. Moreover, the corresponding weak and strong chromatic polynomials are both strongly ﬂaw- less.

We show in a future paper that Ψϕ¯(g) is F -positive by showing that the relative simplicial complex is relatively shellable. It is an interesting problem to determine necessary and suﬃcient criteria for when Σϕ(g) is relatively shellable.

6.3. Rooted Connected Graphs. First, we define the combinatorial objects. Given a finite set N, a rooted connected graph is a connected graph on N ⊔ {r}. We view r as an extra vertex, which we call the root. Let RN be the set of all rooted connected graphs with vertex set N ⊔ {r}. Given a bijection σ : N → M, we extend to a bijection σ : N ⊔{r}→ M ⊔{r} by defining σ(r) = r. Given a rooted graph g ∈ RN , we define Rσ(g) by E(Rσ(g)) = {{σ(u),σ(v)} : {u, v} ∈ E(g)}. We see that we get a map Rσ : RN → RM . Thus R is a species. Now we turn R into a linearized Hopf monoid. Given disjoint sets M and N, and rooted connected graphs g ∈ RM and h ∈ RN , we define g·h by taking disjoint unions of edge sets. In this case g · h has vertex set M ⊔ N ⊔ {r}, and we are identifying the roots of g and h. As a result, the product is a rooted connected graph. Likewise, given a rooted connected graph g ∈ RM⊔N , we define∆M,N (g) as follows. First, if g|M⊔{r}, the induced subgraph of M ⊔{r}, is not connected, then ∆M,N (g) = 0. Suppose then that g|M⊔{r} is connected. Let W ⊂ N be the set of vertices of N that are adjacent to some vertex in M ⊔ {r}. Then define g/M by taking the induced subgraph on M, and adding edges between r and every vertex in W . Then let ∆M,N (g) = g|M⊔{r} ⊗ g/M. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 29

Note that g/M can be viewed as contracting all the vertices of M ⊔ {r} into one vertex. Given a rooted connected graph g, we see that χ(g) = 1 if and only if g is a star graph rooted at the only nonleaf vertex. Theorem 43. The pairs (R, ϕ) is a combinatorial Hopf monoid. Moreover, χ is a balanced convex character.

Proof. Let L, M and N be ﬁnite sets. Let g ∈ RL, h ∈ RM , and k ∈ RN . Then (g · h) · k = g · (h · k), as the result in both cases is a rooted connected graph that is the union of the three individual graphs. Let g ∈ RL⊔M⊔N . If g|L∪{r} is disconnected, or g|L⊔M⊔{r} is disconnected, then

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(g) = 0

= ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(g). Otherwise, we can see that

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(g)= g|L⊔{r} ⊗ (g|L⊔M⊔{r})/L ⊗ g/(L ⊔ M). Similarly,

((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(g)= g|L⊔{r} ⊗ (g/L)|M⊔{r} ⊗ g/(L ∪ M).

We see that (g|L⊔M⊔{r})/L = (g/L)|M⊔{r}, as both graphs consist of re- stricting to M and contracting the vertices in L ⊔ {r}. Hence we have coassociativity. Now we show compatability. Let A,B,C and D be ﬁnite sets, and let g ∈ MA⊔B and h ∈ MC⊔D. If g|A∪{r} or h|C∪{r} is disconnected, then

∆A⊔C,B⊔D(g · h) = 0

= ((µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(g ⊗ h). Otherwise, we see that

∆A⊔C,B⊔D(g · h)= g|A · h|C ⊗ g/A · h/C

= (µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(g ⊗ h). Thus we see that the product and coproduct are compatible. By Theorem 22, χ is a balanced convex character.

Given a rooted connected graph g with vertex set V , we see that Ψχ(g, x) is a summation over proper colorings that have the additional feature that, for every vertex v, there exists a path r, v1, . . . , vk, v with the property that f(v1) < f(v2) < · · · < f(vk) < f(v). If there is an edge between r and every vertex of N, then Ψχ(g, x) is just the chromatic symmetric function. However, for other types of rooted connected graphs, we obtain a new quasisymmetric function. As an example, if we let g be the graph on the left in Figure 6.3, then

Ψϕ(g, x)= M2,1 + 4M1,1,1. 30 JACOB A. WHITE

b a, b b, c

a c a, c

r a c

Figure 3. A rooted connected graph, and its corresponding coloring complex.

The relative simplicial complex on the right of Figure 6.3 is the coloring complex of g with respect to χ. Corollary 44. For every rooted connected graph g, the quasisymmetric function Ψχ(g, x) is M-increasing. Moreover, the corresponding chromatic polynomial is strongly ﬂawless.

6.4. Double Posets. Now we will discuss double posets. The Hopf algebra of double posets was introduced by Malvenuto and Reutenauer [17]. How- ever, we show that double posets also form a combinatorial Hopf monoid. The associated quasisymmetric function is a generalization of Gessel’s P - partition enumerator, along with a generalization for labeled posets (P,ω). This quasisymmetric function is studied extensively by Grinberg [12], who shows F -positivity results when the double poset is tertispecial. First, we define the combinatorial objects. Given a finite set N, a double poset on N is a triple (N, ≤1, ≤2) where ≤1 and ≤2 are both partial orders on N. Let DPN be the set of all double posets with vertex set N. Given a bijection σ : N → M, and a double poset d ∈ DPN , we define DPσ(d) to be −1 −1 the double poset given by declaring x ≤i y if and only if σ (x) ≤i σ (y) for all x,y ∈ N, i ∈ {1, 2}. We see that we get a map DPσ : DPN → DPM . Thus DP is a species. Now we turn DP into a linearized Hopf monoid. Given disjoint sets M and N, and double posets d ∈ DPM and h ∈ DPN , we define two partial orders on M ⊔ N. For x,y ∈ M ⊔ N, we say x ≤1 y if one of the following holds: (1) x,y ∈ M and x ≤1 y in d. (2) x,y ∈ N and x ≤1 y in h. We observe that ≤1 is the disjoint union of two partial orders. For x,y ∈ M ⊔ N, we say x ≤2 y if one of the following holds: (1) x,y ∈ M and x ≤2 y in d. (2) x,y ∈ N and x ≤2 y in h. (3) x ∈ M and y ∈ N.

We observe that ≤2 is the ordinal sum of two partial orders. Then we let d · h = (M ⊔ N, ≤1, ≤2). Let DN be the set of double posets on N. Given p ∈ DM and q ∈ DN , where M and N are disjoint sets, we define a double COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 31 poset p · q = (M ⊔ N, ≤1, ≤2). Hence we have a multiplication operation for D. Now we define the comultiplication. Given a double poset d ∈ DPM⊔N , we define d|M by: for x,y ∈ M, we say x ≤i y in p|M if and only if x ≤i y in d. We define

d|M ⊗ d|N if M is an order ideal of ≤1 ∆M,N (d)= (0 otherwise. ′ ′ Given a double poset d, a pair (m,m ) ∈ M is an inversion if m <1 m ′ and m <2 m. Finally, given a double poset d ∈ DPN , we deﬁne 0 if d has an inversion ϕ(d)= (1 otherwise. Theorem 45. The pair (DP, ϕ) is a combinatorial Hopf monoid. Moreover, ϕ is a balanced convex character.

Proof. Let L, M and N be ﬁnite sets. Let d ∈ ML, f ∈ MM , and g ∈ MN . Then (d · f) · g = d · (f · g), as the result in both cases is a double poset such that <1 is the disjoint union of the ﬁrst partial orders for d, f and g, and <2 is the ordinal sum of the second partial orders of d, f and g. Let d ∈ ML⊔M⊔N . If u>1 v for some u, v with u ∈ L and v ∈ M, or u ∈ L and v ∈ N, or u ∈ M and v ∈ N, then

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(f) = 0

= ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(f). If no such pair exists, then we obtain

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(f)= f|L ⊗ f|M ⊗ f|N

= ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(f). Hence we have coassociativity. Now we show compatability. Let A,B,C and D be ﬁnite sets, and let d ∈ MA⊔B and f ∈ MC⊔D. If there exists a ∈ A, b ∈ B with a >1 b in d, or there exists c ∈ C, d ∈ D with c>1 d in f, then

∆A⊔C,B⊔D(d · f) = 0

= ((µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(d ⊗ f).

h ⊗ k = ((µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(d ⊗ f). Now we show that the character ϕ is balanced convex. First, note that if d is a double poset on at least two vertices, then there exists a nontrivial 32 JACOB A. WHITE

a b d c a abc ac

c a c b d acd

Figure 4. A double poset, and its coloring complex. decomposition N(d)= S ⊔ T for which ∆S,T (d) 6= 0. For example, S = {v} for any minimum element v with respect to ≤1 of d. Let d be a double poset such that ϕ(d) = 1. Let N = S ⊔ T such that ∆S,T (d) 6= 0. Then ∆S,T (d) = d|S ⊗ d|T , where each factor is the induced double poset on the corresponding subset. We see that if d has no inversions, then every induced subposet of d also has no inversions. Hence ϕS,T (d) = 1. Thus, by Proposition 21, ϕ is a balanced convex character.

The corresponding quasisymmetric function Ψϕ(d, x) enumerates double poset partitions. A double poset partition is a function σ : N → N subject to:

(1) for x,y ∈ N, if x ≤1 y, the σ(x) ≤ σ(y). (2) for x,y ∈ N, if x ≤1 y and y <2 x, then σ(x) <σ(y).

It is not hard to show that Ψϕ(d, x) = σ m∈M xσ(m). This amounts to showing that a double poset partition is a ϕ-proper function. As an example, if we let d be the doubleP posetQ in Figure 6.4, where the Hasse diagram on the left is for ≤1 and the Hasse diagram in the middle is for ≤2. Then

Ψϕ(d, x)= M2,2 + 2M1,1,2 + 2M1,2,1 + 2M2,1,1 + 4M1,1,1,1

= F2,2 + F1,1,2 + 2F1,2,1 + F2,1,1 − F1,1,1,1. This quasisymmetric function is not F -positive. The relative simplicial complex on the right of Figure 6.4 is the coloring complex of d with respect to ϕ. Corollary 46. For every double poset d, the quasisymmetric functions Ψϕ(p, x) is M-increasing. Moreover, the corresponding order polynomial is strongly flawless. It is an interesting problem to determine necessary and sufficient criteria for when Σϕ(d) is relatively shellable. 6.5. Antimatroids. First, we define the combinatorial objects. Given a finite set N, an antimatroid is a collection a of subsets of N that satisfies the following conditions: COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 33

(1) For every S ∈ a, there exists x ∈ S such that S \ {x}∈ a. (2) For every S, T ∈ a, we have S ∪ T ∈ a. Antimatroids were introduced by Author [16]. Given a partially ordered set P , let J(P ) be the set of order ideals of P . Then J(P ) is an antimatroid. Another example of an antimatroid appears on the left in Figure 6.5. Let AN be the set of all antimatroids with vertex set N. Given a bijection σ : N → M, and an antimatroid a ∈ AN , we define σ(a) = {σ(S) : S ∈ a}. We see that we get a map Aσ : AN → AM . Thus A is a species. Now we turn A into a linearized Hopf monoid. Given disjoint sets M and N, and antimatroids a ∈ AM and b ∈ AN , we define a · b = {X ∪ Y : X ∈ a,Y ∈ b}. Likewise, given an antimatroid a ∈ AM⊔N with M ∈ a, we define a|M = {S : S ∈ a,S ⊆ M} and a/M = {T : T ⊆ N, T ∪ M ∈ a}. For a ∈ AM⊔N , we define a|M ⊗ a/M if M ∈ a ∆M,N (a)= (0 otherwise. Finally, given an antimatroid a ∈ AN , we see that χ(a) = 1 if and only if a = 2N . Theorem 47. The pair (A,χ) is a combinatorial Hopf monoid. Moreover, χ is a balanced convex character.

Proof. Let L, M and N be ﬁnite sets. Let a ∈ AL, b ∈ AM , and c ∈ AN . Then (a · b) · c = a · (b · c), as the result in both cases is the antimatroid a · b · c = {X ∪ Y ∪ Z : X ∈ a,Y ∈ b,Z ∈ c}.

Next, we show coassociativity. Let a ∈ AL⊔M⊔N . If L 6∈ a or L ∪ M 6∈ a, then

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(a) = 0 = ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(a). Otherwise, we have

((∆L,M ⊗ idN ) ◦ ∆L⊔M,N )(a)= a|L ⊗ aL,M ⊗ a/(L ∪ M)

= ((idL ⊗∆M,N ) ◦ ∆L,M⊔N )(a) where aL,M = {Y ⊆ M : Y ∪ L ∈ a}. Hence we have coassociativity. Now we show compatability. Let A,B,C and D be ﬁnite sets, and let a ∈ AA⊔B and b ∈ AC⊔D. If A 6∈ a or C 6∈ b, then

∆A⊔C,B⊔D(a · b) = 0

= ((µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(a ⊗ b). In all other cases, we have

∆A⊔C,B⊔D(a · b)= a|A · b|C ⊗ a/A · b/C

= (µA,C ⊗ µB,C ) ◦ (idA ⊗τB,C ⊗ idD) ◦ (∆A,B ⊗ ∆C,D))(a ⊗ b). Thus we see that the product and coproduct are compatible. By Theorem 22, χ is a balanced convex character. 34 JACOB A. WHITE

{a, b, c}

{a, b} {b, c}

{a, c} {a, b} {b, c}

{a} {c} {a, c}

{a} {c} ∅

Figure 5. An antimatroid, and its corresponding coloring complex.

As an example, in Figure 6.5, we have an antimatroid a and the cor- reponding coloring complex Σϕ(a). The resulting quasisymmetric function is

Ψϕ(a, x) = 2M1,2 + M2,1 + 4M1,1,1.

The resulting quasisymmetric function Ψχ(a, x) is a generalization of Ges- sel’s P -partition quasisymmetric function. Now we enumerate all functions f : I → N such that f −1([i]) ∈ a and f −1(i) is a boolean lattice, for all i. This appears to be a new invariant.

Corollary 48. Let a be an antimatroid. Then the invariant Ψϕ(a, x) is M-increasing.

6.6. Generalized Permutohedra. First, we define the combinatorial objects. The Hopf monoid of generalized Permutohedra can be described in terms of generalized permutohedra, or in terms of submodular functions [1]. We will focus on the latter description. A submodular function is a funcion z : 2N → R which satisfies: (1) z(∅) = 0, (2) For A, B ⊆ N, we have z(A ∪ B)+ z(A ∩ B) ≤ z(A)+ z(B). We say z is modular if we have z(A ∪ B)+ z(A ∩ B)= z(A)+ z(B). Let SFN be the set of all submodular functions with vertex set N. Given a bijection σ : N → M, and a submodular function z ∈ SFN , we define σ(z) −1 by σ(z)= z ◦ σ . We see that we get a map SFσ : SFN → SFM . Thus SF is a species. Now we turn SF into a linearized Hopf monoid. Given disjoint sets M and N, and submodular functions w ∈ SFM and z ∈ SFN , we define w · z by (w · z)(X)= w(X ∩ M)+ z(X ∩ N). COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 35

Likewise, given a submodular function z ∈ AM⊔N , we define z|M by z|M (S)= z(S) for S ⊆ M. We define z/M by z/M(X)= z(M ∪X)−z(M), for S ⊆ N. Then ∆M,N (z)= z|M ⊗ z/M. Finally, given a submodular function z ∈ SFN , we see that χ(z)=1if and only if z is modular. This is the same character studied by Aguiar and Ardila [1], which they refer to as the basic invariant (they define it for generalized permutohedra instead). Theorem 49. The pair (SF,χ) is a combinatorial Hopf monoid. Moreover, χ is a balanced convex character. Hence, for every submodular function z, the quasisymmetric function Ψχ(z, x) is M-increasing, and the polynomial χχ(z, x) is strongly flawless. Proof. The face that SF is a Hopf monoid was proven by Aguiar and Ardila [1]. The fact that χ is balanced and convex follows from Theorem 22.

The polynomial χχ(z, x) was introduced by Aguiar and Ardila [1]. In the context of generalized permutohedra, it counts functions that are maximized on a unique vertex, while Ψχ(z, x) is a natural quasisymmetric function generalization of the polynomial. Finally, Aguiar and Ardila have shown that many Hopf monoids arise as Hopf submonoids of SF, or a related Hopf monoid of extended generalized permutohedra. We show that A does not correspond to just a Hopf submonoid. This demonstrates that there are examples of Hopf monoids in the literature that can be studied via coloring complexes but not generalized permutohedra. An extended generalized permutohedron corresponds to an extended submodular function z : N → R ∪ {∞} which has the property that, for any A, B ⊂ N with z(A) < ∞ and z(B) < ∞, we have z(A ∪ B) + z(A ∩ B) ≤ z(A) + z(B). The corresponding species ESF is also a linearized combinatorial Hopf monoid, with the rule that ∆S,T (z)= z|S ⊗ z/S if z(S) < ∞, and ∆S,T (z) = 0 otherwise. Proposition 50. There is no injective homomorphism ϕ : A → ESF. Proof. First, suppose that ϕ : A → ESF is an injective homomorphism. We show that ϕ is compatible with χ. Clearly, given an antimatroid a, if χ(a) = 1, then a = a1 · · · ak for several antimatroids ai, where each ai has a singleton vertex set. Then ϕ(a)= ϕ(a1) · · · ϕ(ak), and so χ(ϕ(a)) = 1. Suppose then that there exists an antimatroid a with χ(ϕ(a)) = 1. There is a linear order ℓ on N(a), and modular functions z1,..., zn such that

ϕ(a) = z1 · · · zn, where zi ∈ ESF{ℓi}. Let ai = a|{ℓ1,...,ℓi}/{ℓ1,...,ℓi−1}. Applying coproducts, and the fact that ϕ is a homomorphism, we see that ϕ(ai)= zi. Then ϕ(a)= ϕ(a1 · · · an). Since ϕ is injective, a = a1 · · · an, and hence χ(a) = 1. Hence ϕ is compatible with χ. Given z, and sets S, T with z(S) < ∞ and z(T ) < ∞, it follows that z(S ∩ T ) < ∞. In particular, if we let C(z) be the collection of sets S with z(S) < ∞, then C(z) is closed under intersections. We know that C(z) is 36 JACOB A. WHITE also the vertex set of Σχ(z). On the other hand, given an antimatroid a, the vertex set of Σχ(a) are the subsets S ∈ a. Since ϕ is compatible with χ, it follows that Σχ(ϕ(a)) =Σχ(a). This implies that the subsets of a are closed under intersection for every antimatroid. However, this is not true: the antimatroid in Figure 6.5 is a counterexample. Thus we have obtained a contradiction. In particular, there are natural examples of linearized combinatorial Hopf monoids which are not submonoids of the Hopf monoid of extended generalized permutohedra, but which may be viewed as submonoids of the Hopf monoid of coloring complexes. The same result is also true for rooted connected graphs, with a similar proof.

References [1] M. Aguiar and F. Ardila. Hopf monoids and generalized permutahedra. arXiv e-prints, page arXiv:1709.07504, Sept. 2017. [2] M. Aguiar, N. Bergeron, and F. Sottile. Combinatorial Hopf algebras and generalized Dehn-Sommerville relations. Compos. Math., 142(1):1–30, 2006. [3] M. Aguiar and S. Mahajan. Monoidal functors, species and Hopf algebras, volume 29 of CRM Monograph Series. American Mathematical Society, Providence, RI, 2010. With forewords by Kenneth Brown and Stephen Chase and Andr´e Joyal. [4] M. Aguiar and S. Mahajan. Hopf monoids in the category of species. Contemporary Mathematics, 585:17–124, 2013. [5] F. Ardila and M. Sanchez. Valuations and the Hopf Monoid of Generalized Permu- tahedra. arXiv e-prints, page arXiv:2010.11178, Oct. 2020. [6] J.-C. Aval, N. Bergeron, and J. Machacek. New invariants for permutations, orders and graphs. Adv. in Appl. Math., 121:102080, 30, 2020. [7] M. Beck, D. Blado, J. Crawford, T. Jean-Louis, and M. Young. On weak chromatic polynomials of mixed graphs. Graphs Combin., 31(1):91–98, 2015. [8] M. Beck, T. Bogart, and T. Pham. Enumeration of Golomb rulers and acyclic orien- tations of mixed graphs. Electron. J. Combin., 19(3):Paper 42, 13, 2012. [9] L. J. Billera, N. Jia, and V. Reiner. A quasisymmetric function for matroids. European J. Combin., 30(8):1727–1757, 2009. [10] N. G. de Bruijn, C. van Ebbenhorst Tengbergen, and D. Kruyswijk. On the set of divisors of a number. Nieuw Arch. Wiskunde (2), 23:191–193, 1951. [11] I. M. Gessel. Multipartite P -partitions and inner products of skew Schur functions. In Combinatorics and algebra (Boulder, Colo., 1983), volume 34 of Contemp. Math., pages 289–317. Amer. Math. Soc., Providence, RI, 1984. [12] D. Grinberg. Double posets and the antipode of QSym. Electron. J. Combin., 24(2):Paper No. 2.22, 47, 2017. [13] P. Hersh and E. Swartz. Coloring complexes and arrangements. J. Algebraic Combin., 27(2):205–214, 2008. [14] T. Hibi. What can be said about pure O-sequences? J. Combin. Theory Ser. A, 50(2):319–322, 1989. [15] A. Hultman. Link complexes of subspace arrangements. European J. Combin., 28(3):781–790, 2007. [16] R. Jamison. Copoints in antimatroids. Congr. Numer., 29:535–544, 1980. [17] C. Malvenuto and C. Reutenauer. A self paired Hopf algebra on double posets and a Littlewood-Richardson rule. J. Combin. Theory Ser. A, 118(4):1322–1333, 2011. [18] E. Marberg. Strong forms of linearization for Hopf monoids in species. J. Algebraic Combin., 42(2):391–428, 2015. COLORING COMPLEXES AND COMBINATORIAL HOPF MONOIDS 37

[19] R. P. Stanley. An introduction to combinatorial commutative algebra. In Enumeration and design (Waterloo, Ont., 1982), pages 3–18. Academic Press, Toronto, ON, 1984. [20] R. P. Stanley. A symmetric function generalization of the chromatic polynomial of a graph. Adv. Math., 111(1):166–194, 1995. [21] E. Steingr´ımsson. The coloring ideal and coloring complex of a graph. J. Algebraic Combin., 14(1):73–84, 2001.

School of Mathematical and Statistical Sciences, University of Texas - Rio Grande Valley, Edinburg, TX 78539