, , , November 27, 2006

While the original motivation for the introduction of dendriform dialgebra by Loday [Lo1, Lo2] was to study the periodicity of algebraic

K

-groups, it soon became clear that dendriform dialgebras are an interesting subject on its own. This can be seen on one hand by its quite extensive study by several authors in areas related to operads [Lo6] , homology [Fra1, Fra2] , combinatorics [Fo, L-R1, A-S1, A-S2] , arithmetic [Lo5] , quantum field theory [Fo] and especially Hopf algebras [A-S1, Ch, Hol1, Ron, L-R3] . On the other hand it has several generalizations and extensions that share many properties of the original dendriform dialgebra. These new structures include the dendriform trialgebra [L-R2] , the dendriform quadri-algebra [A-L] , the 2-associative algebra [L-R3, Pi] , the magma algebra [G-H] , and very recently the ennea-algebra, the dendriform-Nijenhuis algebra and the octo-algebra [Le1, Le2, Le3] .

It is interesting to note that many of these algebras have a Hopf algebra structure on the free algebras. A recent method to obtain such Hopf algebras is by means of coherent unit actions, a concept introduced by Loday [Lo6] who showed that the existence of a coherent unit action on a binary, quadratic, regular operad with splitting of associativity endows the free objects with a Hopf algebra structure. Since then, this method has been applied to obtain Hopf algebra structures on several operads [Lo6, Le1, Le2, Le3] .

In this paper we strive for a thorough understanding of the notion of coherent unit actions of such operads by working with their relations and generators [E-G, Lo7] .

Together with the above result of Loday, we can explicitly describe a large class of operads that gives rise to Hopf algebras.

After briefly recalling related concepts and results, we first give in Section 2 a simple criterion for a unit action to be coherent which reduces the checking of the coherence condition to the verification of a system of linear equations, called coherence equations.

Then in Section 3 we use the coherence equations to obtain, for each integer

n \geq 1

, universal coherence relation spaces

Λ_{n, h}^{'}

and

Λ_{n, h}^{''}

. The first space contains the relation spaces of all operads of dimension

n

equipped with a coherent unit action in which the two linear maps are distinct. The second space contains those with a coherent unit action in which the linear maps are the same. The cases when

n = 2

and

3

are studied in more detail and are related to examples which appeared in the current literature.

The compatibility of coherent unit actions of operads with taking operad products and duals is studied in Section 4 . We show that the coherence condition is preserved by taking products. Thus the Hopf algebra structure on the product operad follows automatically from those on the factor operads, as long as the factor operads have coherent unit actions. Examples and applications are given. In contrast to products, we show that the coherence condition is never preserved by taking the dual in the operadic sense, except for the trivial case when the operad is the one for associative algebras.

We give a similar study of the notion of compatible unit actions, a concept which was also introduced by Loday in [Lo6] . It is related to, but weaker than the notion of coherent unit actions. Coherent unit actions are also defined for binary, quadratic operads by Loday [Lo6] and are recently extended to general algebraic operads by Holtkamp [Hol2] . In these cases, the free objects of the operad

P

have the structure of a

P

-Hopf algebra, a more general concept than Hopf algebra. Extending our findings to the general case appears to be worth doing.

2 Compatible and coherent unit actions

2.1 ABQR operads

We recall the standard definition of algebraic operads in general before rephrasing it in the special case that we are considering. Since we will not need the general definition in the rest of the paper, we refer the interested reader to the standard references, such as [G-K, Lo3, L-S-V, M-S-S] .

Let

k

be a field of characteristic zero and let

V e c t

be the category of

k

-vector spaces. An algebraic operad over

k

is an analytic functor

P : V e c t \to V e c t

such that

P (0) = 0

, and is equipped with a natural transformation of functors

γ : P \circ P \to P

which is associative and has a unit

1 : i d \to P

By considering free

P

-algebras, an operad gives a sequence

{P (n)}

of finitely generated

k [S_{n}]

-modules that satisfy certain composition axioms. An operad is called binary if

P (1) = k

and

P (2)

generates

P (n), n \geq 3

by composition; is called quadratic if all relations among the binary operations

P (2)

are derived from relations in

P (3)

; is called regular if, moreover, the binary operations have no symmetries and the variables

x, y

and

z

appear in the same order (such as

(x \cdot y) \cdot z = x \cdot (y \cdot z)

, not

(x \cdot y) \cdot z = x \cdot (z \cdot y)

By regularity, the space

P (n)

is of the form

P_{n} \otimes k [S_{n}]

where

P_{n}

is a vector space. So the operad

{P (n)}

is determined by

{P_{n}}

. Then a binary, quadratic, regular operad is determined by a pair

(Ω, Λ)

where

Ω = P_{2}

, called the space of generators, and

Λ

is a subspace of

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

, called the space of relations. So we also write

P = (Ω, Λ)

for the operad.

For such a

P = (Ω, Λ)

, a

k

-vector space

A

is called a

P

-algebra if it has binary operations

Ω

and if, for

(\sum_{i = 1}^{k} ⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, \sum_{j = 1}^{m} ⊙_{j}^{(3)} \otimes ⊙_{j}^{(4)}) \in Λ \subseteq Ω^{\otimes 2} \oplus Ω^{\otimes 2}

with

⊙_{i}^{(1)}, ⊙_{i}^{(2)}, ⊙_{j}^{(3)}, ⊙_{j}^{(4)} \in Ω

1 \leq i \leq k

1 \leq j \leq m

, we have

\begin{matrix} \sum_{i = 1}^{k} (x ⊙_{i}^{(1)} y) ⊙_{i}^{(2)} z = \sum_{j = 1}^{m} x ⊙_{j}^{(3)} (y ⊙_{j}^{(4)} z), \forall x, y, z \in A . \end{matrix}

(1)

When there is no danger of confusion, we will use

Λ

to denote

P

Since

(Ω, Λ)

is determined by

(\vec{ω}, \vec{λ})

where

\vec{ω}

is a basis of

Ω

and

\vec{λ}

is a basis of

Λ

, we also use

(\vec{ω}, \vec{λ})

to denote a binary, quadratic, regular operad, as is usually the case in the literature.

We say that a binary, quadratic, regular operad

(Ω, Λ)

has a splitting associativity if there is a choice of

◆

Ω

such that

(◆ \otimes ◆, ◆ \otimes ◆)

is in

Λ

[Lo6] . As abbreviation, we call such an operad an associative BQR operad, or simply an ABQR operad.

Equivalently, a binary, quadratic, regular operad is ABQR if and only if there is a basis

\vec{ω} = {ω_{i}}

Ω

such that

◆ = \sum_{i} ω_{i}

and there is a basis

\vec{λ} = {λ_{j}}_{j}

Λ

such that the associativity of

◆

is given by the sum of

λ_{j}

(splitting associativity):

(◆ \otimes ◆, ◆ \otimes ◆) = \sum_{j} λ_{j} .

Let

(Ω, Λ)

and

(Ω^{'}, Λ^{'})

be ABQR operads with associative operations

◆

and

◆^{'}

respectively. A morphism

f : (Ω, Λ) \to (Ω^{'}, Λ^{'})

is a linear map

Ω \to Ω^{'}

sending

◆

◆^{'}

and inducing a linear map

Λ \to Λ^{'}

. An invertible morphism is called an isomorphism, and called an automorphism if

(Ω, Λ) = (Ω^{'}, Λ^{'})

The following examples of ABQR operads will be used later in the paper.

Example 2.1.

(1) (Associative algebra) An associative $k$ -algebra is a $k$ -vector space $A$ with an associative product $\cdot$ . The corresponding operad is $P_{A} = ({\vec{ω}}_{A}, {\vec{λ}}_{A})$ with ${\vec{ω}}_{A} = {\cdot}$ and ${\vec{λ}}_{A} = {(\cdot \otimes \cdot, \cdot \otimes \cdot)} .$

(2) (Dendriform dialgebra) The dendriform dialgebra of Loday [Lo4] corresponds to the operad

P_{D} = (Ω_{D}, Λ_{D})

with

{\vec{ω}}_{D} = {≺, ≻}

and

\begin{matrix} {\vec{λ}}_{D} = {(≺ \otimes ≺, ≺ \otimes (≺ + ≻)), (≻ \otimes ≺, ≻ \otimes ≺), ((≺ + ≻) \otimes ≻, ≻ \otimes ≻)} . \end{matrix}

(2)

(3) (Dendriform trialgebra) The dendriform trialgebra of Loday and Ronco [L-R2] corresponds to the operad

P_{T} = ({\vec{ω}}_{T}, {\vec{λ}}_{T})

with

{\vec{ω}}_{T} = {≺, ≻, \circ}

and

\begin{matrix} {\vec{λ}}_{T} = & {(≺ \otimes ≺, ≺ \otimes ◆), (≻ \otimes ≺, ≻ \otimes ≺), (◆ \otimes ≻, ≻ \otimes ≻), (≻ \otimes \circ, ≻ \otimes \circ), \end{matrix}

\begin{matrix} (≺ \otimes \circ, \circ \otimes ≻), (\circ \otimes ≺, \circ \otimes ≺), (\circ \otimes \circ, \circ \otimes \circ)} . \end{matrix}

(3)

(4) (NS-algebra) Leroux's NS-algebra [Le2] corresponds to the operad

({\vec{ω}}_{N}, {\vec{λ}}_{N})

with

{\vec{ω}}_{N} = {≺, ≻, ∙}

and

\begin{matrix} {\vec{λ}}_{N} = & {(≺ \otimes ≺, ≺ \otimes ◆), (≻ \otimes ≺, ≻ \otimes ≺), (◆ \otimes ≻, ≻ \otimes ≻), \end{matrix}

\begin{matrix} (◆ \otimes ∙ + ∙ \otimes ≺, ≻ \otimes ∙ + ∙ \otimes ◆)} . \end{matrix}

(4)

2.2 Compatible and coherent unit actions

The concepts of compatible and coherent unit actions on an operad

P

were introduced by Loday [Lo6] to obtain unitary

P

-algebras. By an important theorem of Loday (Theorem 2.2 ), the existence of a coherent unit action on an operad yields a Hopf algebra structure on the free objects.

We review these concepts and the theorem of Loday. We first provide a list of equations, called the coherence equations, that characterize a compatible or a coherent unit action (Theorem 2.3 ). These equations will allow us to give relations for all possible ABQR operads that have a compatible or coherent unit action, and to determine such actions on products and duals of ABQR operads in later sections.

Let

P

be a binary quadratic operad. An unit action on

P

is a choice of two linear maps

α, β : P_{2} - \to k .

P

is an ABQR operad, we further require that, for the chosen associative operation

◆ \in P_{2}

, we have

α (◆) = β (◆) = 1,

the unit of

k

Let

A

be a

P

-algebra. A unit action

(α, β)

allows us to extend a binary operation

⊙ \in P_{2}

A

to a restricted binary operation on

A_{+} : = k . 1 \oplus A

by defining

\begin{matrix} \tilde{⊙} : A_{+} \otimes A_{+} \to A_{+}, a \tilde{⊙} b : = {\begin{matrix} a ⊙ b, & a, b \in A, \\ α (⊙) a, & a \in A, b = 1, \\ β (⊙) b, & a = 1, b \in A, \\ 1, & a = b = 1, ⊙ = ◆, \\ u n d e f i n e d, & a = b = 1, ⊙ \neq ◆ . \end{matrix} \end{matrix}

(5)

Thus the extended binary operation

\tilde{⊙}

is defined on the subspace

(k . 1 \otimes A) \oplus (A \otimes k . 1) \oplus (A \otimes A)

A_{+} \otimes A_{+}

, and on the full space

A_{+} \otimes A_{+}

when

⊙ = ◆

. The unit action is called compatible with the relations of

P

if the relations of

P

are still valid on

A_{+}

whenever the terms are defined.

Next let

A

B

be two

P

-algebras. Consider the subspace

A ⊠ B : = (A \otimes k . 1) \oplus (k . 1 \otimes B) \oplus (A \otimes B)

. For

⊙ \in P (2)

, define a binary operation, still denoted by

⊙

, on

A ⊠ B

\begin{matrix} (a \otimes b) ⊙ (a^{'} \otimes b^{'}) : = {\begin{matrix} (a \tilde{◆} a^{'}) \otimes (b \tilde{⊙} b^{'}), & if b \otimes b^{'} \neq 1 \otimes 1, \\ (a \tilde{⊙} a^{'}) \otimes 1, & otherwise . \end{matrix} \end{matrix}

(6)

We say that the unit action

(α, β)

is coherent with the relations of

P

A ⊠ B

, equipped with these operations is still a

P

-algebra. We note that

A_{+} \otimes B_{+} = k . 1 \oplus (A ⊠ B) = (A ⊠ B)_{+}

. Thus the associative operation

◆

gives an associative operation

◆

A ⊠ B

and then, as in ( 5 ), extends to an associative operation

\tilde{◆}

A_{+} \otimes B_{+}

It will be clear later that an unit action that is coherent with the relations of

P

is also compatible with the relations of

P

. The converse is not true and examples will be given. The significance of the coherence property can be seen in the following theorem of Loday [Lo6] . We refer the reader to the original paper for further details.

Theorem 2.2 (Loday). Let

P

be an ABQR operad. Let

P (V)_{+}

be the augmented free

P

-algebra on a

k

-vector space

V

. Any unit action that is coherent with the relations of

P

equips

P (V)_{+}

with a connected Hopf algebra structure.

The Hopf algebra structure is in fact a

P

-Hopf algebra structure in the sense that the coproduct is a morphism of augmented

P

-algebras.

For example, Loday showed in [Lo6] that, for the dendriform trialgebra

P_{T} = ({\vec{ω}}_{T}, {\vec{λ}}_{T})

with

{\vec{ω}}_{T} = {≺, ≻, \circ}

, the unit action

α (≺) = β (≻) = 1, α (≻) = α (\circ) = β (≺) = β (\circ) = 0

is coherent with the relations of

P_{T}

, and thus the free

P_{T}

-algebra on a

k

-vector space has a

P_{T}

-Hopf algebra structure. This also recovers the Hopf algebra structure on the free dendriform dialgebra obtained earlier in [Lo4] . The same method was applied to get Hopf algebra structures on several other algebras [Lo6, Le1, Le2, Le3] (see Corollary 4.3 ).

2.3 Coherence equations

We give an equivalent condition of compatibility and coherence in terms of equations among the binary operations in an operad. Since the condition of coherence will play a more prominent role in our further study, we first state its equivalent condition and then give the equivalent condition of compatibility as a special case.

Theorem 2.3. Let

P = (Ω, Λ)

be an ABQR operad.

(1) A unit action ( α , β ) on P is coherent with the relations of P if and only if, for every ∑ i ( ⊙ i ( 1 ) ⊗ ⊙ i ( 2 ) , ⊙ i ( 3 ) ⊗ ⊙ i ( 4 ) ) ∈ Λ , the following coherence equations hold.
- (C1) $\sum_{i} β (⊙_{i}^{(1)}) ⊙_{i}^{(2)} = \sum_{i} β (⊙_{i}^{(3)}) ⊙_{i}^{(4)},$
- (C2) $\sum_{i} α (⊙_{i}^{(1)}) ⊙_{i}^{(2)} = \sum_{i} β (⊙_{i}^{(4)}) ⊙_{i}^{(3)},$
- (C3) $\sum_{i} α (⊙_{i}^{(2)}) ⊙_{i}^{(1)} = \sum_{i} α (⊙_{i}^{(4)}) ⊙_{i}^{(3)},$
- (C4) $\sum_{i} β (⊙_{i}^{(2)}) ⊙_{i}^{(1)} = \sum_{i} β (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) ◆,$
- (C5) $\sum_{i} α (⊙_{i}^{(1)}) α (⊙_{i}^{(2)}) ◆ = \sum_{i} α (⊙_{i}^{(3)}) ⊙_{i}^{(4)} .$
(2) A unit action $(α, β)$ on $P$ is compatible with the relations of $P$ if and only if, for every $\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)}) \in Λ$ , equations (C1), (C2) and (C3) above hold.

Before proving Theorem 2.3 , we give an example to show how it can be used to find compatible and coherent unit actions.

Example 2.4. Consider the dendriform dialgebra

P_{D} = ({\vec{ω}}_{D}, {\vec{λ}}_{D})

in Eq.( 2 ). So

{\vec{ω}}_{D} = {≺, ≻}

and

{\vec{λ}}_{D} = {(≺ \otimes ≺, ≺ \otimes (≺ + ≻)), (≻ \otimes ≺, ≻ \otimes ≺), ((≺ + ≻) \otimes ≻, ≻ \otimes ≻)} .

Suppose

(α, β)

is a coherent unit action of

P_{D}

. Then the three equations in

{\vec{λ}}_{D}

satisfy (C1) – (C5). Applying (C1) to the first relation in

{\vec{λ}}_{D}

, we obtain

β (≺) ≺ = β (≺) (≺ + ≻)

. So

β (≺) ≻ = 0

. Therefore

β (≺) = 0

. Applying (C2) to the first relation, we have

α (≺) ≺ = ≺

since

β (≺ + ≻) = β (◆) = 1

. Thus

α (≺) = 1

. Similarly, applying (C2) to the second equation, we have

α (≻) ≺ = β (≺) ≻

. So

α (≻) = 0

. Applying (C1) to the third equation gives

β (≺ + ≻) ≻ = ≻ = β (≻) ≻

. Thus

β (≻) = 1

. Therefore, the only coherent unit action of

P_{D}

is the one given in [Lo4] :

α (≺) = β (≻) = 1, α (≻) = β (≺)) = 0 .

Note that we have only used (C1)-(C3). So the above is also the only compatible unit action of

P_{D}

We will comment on the trialgebra case later.

Example 2.5. We consider the 2-associative algebra in [L-R3] and [Pi] . It is given by generators

Ω = {*, \cdot}

and relations

Λ = {(* \otimes *, * \otimes *), (\cdot \otimes \cdot, \cdot \otimes \cdot)} .

Consider the unit action

(α, β)

in [Lo6] given by

α (*) = α (\cdot) = β (*) = β (\cdot) = 1

. We show that the action is not coherent regardless of the choice of associative operation

◆

which might or might not be

*

\cdot

. Suppose

(α, β)

is coherent. Then applying (C4) to

(* \otimes *, * \otimes *)

, we have

β (*) * = β (*) β (*) ◆

. So

* = ◆

. Applying (C4) to

(\cdot \otimes \cdot, \cdot, \otimes \cdot)

, we have

β (\cdot) \cdot = β (\cdot) β (\cdot) ◆

. So

\cdot = ◆ .

This is impossible. So Loday's Theorem 2.2 cannot be applied to give a Hopf algebra structure on free 2-associative algebras.

However, by verifying (C1)-(C3), we see that the unit action is compatible with these relations when

◆

is taken to be

*

. Loday and Ronco [L-R3] have equipped a free 2-associative algebra with a Hopf algebra structure. This suggests a possible connection between compatibility and Hopf algebras.

Proof. (1) Let

(α, β)

be an unit action on

P

. Let

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)})

be any given element in the space of relations

Λ

P

. We say that the unit action

(α, β)

is coherent with a relation

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)})

if, for any

P

-algebras

A

and

B

, the operations

⊙ \in Ω

, when extended to

A ⊠ B

by Eq. ( 6 ), still satisfy the same relation. Then to prove the theorem, we only need to prove that

(α, β)

is coherent with a given relation

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)}) \in Λ

if and only if (C1) – (C5) hold for this relation.

Further, by definition,

(α, β)

is coherent with

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)})

means that, for any

P

-algebra

A

and

B

and for any

a, a^{'}, a^{''} \in A_{+}

and

b, b^{'}, b^{''} \in B_{+}

such that at least one of

b, b^{'}, b^{''}

is not 1, we have the equation

\begin{matrix} \sum_{i} ((a \otimes b) ⊙_{i}^{(1)} (a^{'} \otimes b^{'})) ⊙_{i}^{(2)} (a^{''} \otimes b^{''}) = \sum_{i} (a \otimes b) ⊙_{i}^{(3)} ((a^{'} \otimes b^{'}) ⊙_{i}^{(4)} (a^{''} \otimes b^{''})) . \end{matrix}

(7)

Thus there are 7 mutually disjoint cases for the choice of such

b, b^{'}, b^{''}

: the case when none of

b, b^{'}, b^{''}

is 1, the three cases when exactly one of

b, b^{'}, b^{''}

is 1, and the three cases when exactly two of

b, b^{'}, b^{''}

are 1. Note that when none of

b, b^{'}, b^{''}

is 1, Eq. ( 7 ) just means that

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)})

is a relation for

P

, so is automatic true. Thus to prove the theorem we only need to prove

Case 1. Eq. ( 7 ) holds for $b = 1$ , $b^{'} \neq 1 \neq b^{''}$ if and only if (C1) is true;
Case 2. Eq. ( 7 ) holds for $b^{'} = 1$ , $b \neq 1 \neq b^{''}$ if and only if (C2) is true;
Case 3. Eq. ( 7 ) holds for $b^{''} = 1$ , $b \neq 1 \neq b^{'}$ if and only if (C3) is true;
Case 4. Eq. ( 7 ) holds for $b = b^{'} = 1$ , $b^{''} \neq 1$ if and only if (C4) is true;
Case 5. Eq. ( 7 ) holds for $b^{'} = b^{''} = 1$ , $b \neq 1$ if and only if (C5) is true;
Case 6. Eq. ( 7 ) holds for $b = b^{''} = 1$ , $b^{'} \neq 1$ if (C1) is true.

So Case 6 follows from the previous 5 cases. We first consider the three cases when exactly one of

b, b^{'}, b^{''}

is 1. Then by the definition of the operation

⊙

A ⊠ B

in Eq. ( 6 ), we can rewrite Eq. ( 7 ) as

\begin{matrix} \sum_{i} (((a \tilde{◆} a^{'}) \tilde{◆} a^{''}) \otimes ((b {\tilde{⊙}}_{i}^{(1)} b^{'}) {\tilde{⊙}}_{i}^{(2)} b^{''})) = \sum_{i} ((a \tilde{◆} (a^{'} \tilde{◆} a^{''})) \otimes (b {\tilde{⊙}}_{i}^{(3)} (b^{'} {\tilde{⊙}}_{i}^{(4)} b^{''}))), \end{matrix}

(8)

for all

1 \otimes 1 \otimes 1 \neq b \otimes b^{'} \otimes b^{''} \in (B_{+})^{\otimes 3}

. Since

\tilde{◆}

is associative, by the arbitrariness of

A

and

a, a^{'}, a^{''} \in A_{+}

(say by taking

a = a^{'} = a^{''} = 1

), we see that Eq. ( 8 ), and hence Eq. ( 7 ), is equivalent to

\begin{matrix} \sum_{i} ((b {\tilde{⊙}}_{i}^{(1)} b^{'}) {\tilde{⊙}}_{i}^{(2)} b^{''}) = \sum_{i} (b {\tilde{⊙}}_{i}^{(3)} (b^{'} {\tilde{⊙}}_{i}^{(4)} b^{''})) . \end{matrix}

(9)

Case 1. Assume

b = 1

and

b^{'}, b^{''} \neq 1

. Then Eq. ( 9 ) is

\sum_{i} ((1 {\tilde{⊙}}_{i}^{(1)} b^{'}) {\tilde{⊙}}_{i}^{(2)} b^{''}) = \sum_{i} (1 {\tilde{⊙}}_{i}^{(3)} (b^{'} {\tilde{⊙}}_{i}^{(4)} b^{''}))

and, by Eq. ( 5 ), this means

\sum_{i} (β (⊙_{i}^{(1)}) b^{'} ⊙_{i}^{(2)} b^{''}) = \sum_{i} β (⊙_{i}^{(3)}) (b^{'} ⊙_{i}^{(4)} b^{''}) .

That is,

b^{'} (\sum_{i} β (⊙_{i}^{(1)}) ⊙_{i}^{(2)}) b^{''} = b^{'} (\sum_{i} β (⊙_{i}^{(3)}) ⊙_{i}^{(4)}) b^{''} .

This holds for every

P

-algebra

B

and

b^{'}, b^{''} \in B

if and only if

\sum_{i} β (⊙_{i}^{(1)}) ⊙_{i}^{(2)} = \sum_{i} β (⊙_{i}^{(3)}) ⊙_{i}^{(4)},

giving (C1). Here and later in the proof, we use the following elementary fact:

Lemma 2.6. For

⊙_{1}, ⊙_{2} \in Ω

, we have

⊙_{1} = ⊙_{2}

if and only if

a ⊙_{1} a^{'} = a ⊙_{2} a^{'}

for all

P

-algebras

A

and

a, a^{'} \in A

Proof. The only if part is clear. Now suppose $a ⊙_{1} a^{'} = a ⊙_{2} a^{'}$ for all $P$ -algebras $A$ and $a, a^{'} \in A$ . Let $A$ be the free $P$ -algebra on one generator $x$ . Then $A = k x \oplus P_{2} \oplus \dots .$ Here $P_{2} = \oplus_{i} k ω_{i}$ in which ${ω_{i}}$ is a basis of $Ω$ . Also, a binary operation $⊙ \in Ω$ acts on $A \otimes A$ by $x ⊙ x = ⊙ \in P_{2} .$ Thus we have $⊙_{1} = x ⊙_{1} x = x ⊙_{2} x = ⊙_{2} .$ This proves the if part. □

Case 2. Assume

b^{'} = 1

and

b, b^{''} \neq 1

. As in Case 1, we have 9 5

\begin{matrix} E q . () & \overset{}{\Leftrightarrow} \sum_{i} ((b {\tilde{⊙}}_{i}^{(1)} 1) {\tilde{⊙}}_{i}^{(2)} b^{''}) = \sum_{i} (b {\tilde{⊙}}_{i}^{(3)} (1 {\tilde{⊙}}_{i}^{(4)} b^{''})), \forall b, b^{''} \in B, \end{matrix}

\begin{matrix} \overset{()}{\Leftrightarrow} \sum_{i} b α (⊙_{i}^{(1)}) ⊙_{i}^{(2)} b^{''} = \sum_{i} b β (⊙_{i}^{(4)}) ⊙_{i}^{(3)} b^{''}, \forall b, b^{''} \in B, \end{matrix}

\begin{matrix} \overset{}{\Leftrightarrow} \sum_{i} α (⊙_{i}^{(1)}) ⊙_{i}^{(2)} = \sum_{i} β (⊙_{i}^{(4)}) ⊙_{i}^{(3)} \end{matrix}

\begin{matrix} \overset{}{\Leftrightarrow} (C 2) . \end{matrix}

Case 3. Assume

b, b^{'} \neq 1

and

b^{''} = 1

. As in Case 1, we have 9 5

\begin{matrix} E q . () & \overset{()}{\Leftrightarrow} \sum_{i} (b ⊙_{i}^{(1)} b^{'}) α (⊙^{(2)}) = \sum_{i} b ⊙^{(3)} b^{'} α (⊙^{(4)}), \forall b, b^{'} \in B, \end{matrix}

\begin{matrix} \overset{}{\Leftrightarrow} (C 3) . \end{matrix}

We next consider the three cases when exactly two of

b, b^{'}, b^{''}

are the identity.

Case 4. Assume

b = b^{'} = 1

and

b^{''} \neq 1

. Now Eq. ( 8 ) does not apply. Directly from Eq. ( 6 ), we see that Eq. ( 7 ) means

\sum_{i} ((a {\tilde{⊙}}_{i}^{(1)} a^{'}) \tilde{◆} a^{''}) \otimes (1 {\tilde{⊙}}_{i}^{(2)} b^{''}) = \sum_{i} (a \tilde{◆} (a^{'} \tilde{◆} a^{''})) \otimes (1 {\tilde{⊙}}_{i}^{(3)} (1 {\tilde{⊙}}_{i}^{(4)} b^{''})) .

Then by Eq. ( 5 ), we equivalently have

\sum_{i} ((a {\tilde{⊙}}_{i}^{(1)} a^{'}) \tilde{◆} a^{''}) \otimes (β (⊙_{i}^{(2)}) b^{''}) = \sum_{i} (a \tilde{◆} (a^{'} \tilde{◆} a^{''})) \otimes (β (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) b^{''}) .

This means, by moving the scalars

β (⊙_{i}^{(j)})

across the tensor product,

\sum_{i} β (⊙_{i}^{(2)}) ((a {\tilde{⊙}}_{i}^{(1)} a^{'}) \tilde{◆} a^{''}) \otimes b^{''} = \sum_{i} β (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) (a \tilde{◆} (a^{'} \tilde{◆} a^{''})) \otimes b^{''} .

Since this is true for all

B

and

b^{''} \in B

, we equivalently have

\begin{matrix} \sum_{i} β (⊙_{i}^{(2)}) ((a {\tilde{⊙}}_{i}^{(1)} a^{'}) \tilde{◆} a^{''}) = \sum_{i} β (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) (a \tilde{◆} (a^{'} \tilde{◆} a^{''})) . \end{matrix}

(10)

Taking

a^{''} = 1

, we have

\begin{matrix} \sum_{i} β (⊙_{i}^{(2)}) (a {\tilde{⊙}}_{i}^{(1)} a^{'}) = \sum_{i} β (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) (a \tilde{◆} a^{'}) . \end{matrix}

(11)

Conversely, right multiplying

a^{''}

(by

\tilde{◆}

) to this equation and using the associativity of

\tilde{◆}

, we obtain Eq. ( 10 ). So Eq. ( 10 ) and Eq. ( 11 ) are equivalent.

Now by the arbitrariness of

P

-algebra

A

and

a, a^{'} \in A

, Eq. ( 11 ) implies the relation

\begin{matrix} \sum_{i} β (⊙_{i}^{(2)}) ⊙_{i}^{(1)} = \sum_{i} β (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) ◆ \end{matrix}

(12)

by Lemma 2.6 . This is (C4). To go backwards, assuming (C4), then by the linearity of the map

⊙ \mapsto \tilde{⊙}

in Eq. ( 5 ), we get Eq. ( 11 ) for all augmented

P

-algebras

A_{+}

. So we are done with Case 4.

Case 5. Assume

b \neq 1

and

b^{'} = b^{''} = 1

. This case is similar to Case 4. By Eq. ( 6 ) we have

\sum_{i} ((a \tilde{◆} a^{'}) \tilde{◆} a^{''}) \otimes ((b {\tilde{⊙}}_{i}^{(1)} 1) {\tilde{⊙}}_{i}^{(2)} 1) = \sum_{i} (a \tilde{◆} (a^{'} {\tilde{⊙}}_{i}^{(4)} a^{''})) \otimes (b {\tilde{⊙}}_{i}^{(3)} 1) .

Then by Eq ( 5 ), we have

\sum_{i} ((a \tilde{◆} a^{'}) \tilde{◆} a^{''}) \otimes (b α (⊙_{i}^{(1)}) α (⊙_{i}^{(2)})) = \sum_{i} (a \tilde{◆} (a^{'} {\tilde{⊙}}_{i}^{(4)} a^{''})) \otimes (b α (⊙_{i}^{(3)})) .

Taking

a = 1

a^{'} \neq 1 \neq a^{''}

, and moving the scalars

α (⊙_{i}^{(j)})

across the tensor product, we have

\sum_{i} α (⊙_{i}^{(1)}) β (⊙_{i}^{(2)}) (a^{'} ◆ a^{''}) \otimes b = \sum_{i} α (⊙_{i}^{(3)}) (a ⊙_{i}^{(4)} a^{'}) \otimes b .

Since this is true for any

B

and

b \in B

, we have

\sum_{i} α (⊙_{i}^{(1)}) β (⊙_{i}^{(2)}) (a^{'} ◆ a^{''}) = \sum_{i} α (⊙_{i}^{(3)}) (a ⊙_{i}^{(4)} a^{'}),

and by the arbitrariness of

A

and

a^{'}, a^{''} \in A

, we have

\sum_{i} α (⊙_{i}^{(1)}) α (⊙_{i}^{(2)}) ◆ = \sum_{i} α (⊙_{i}^{(3)}) ⊙_{i}^{(4)} .

This is (C5). As in Case 4, all implications here can be reversed.

Case 6. Assume

b = b^{''} = 1

and

b^{'} \neq 1

. Then Eq. ( 8 ) still applies and we get

\sum_{i} (1 {\tilde{⊙}}_{i}^{(1)} b^{'}) {\tilde{⊙}}_{i}^{(2)} 1 = \sum_{i} 1 {\tilde{⊙}}_{i}^{(3)} (b^{'} {\tilde{⊙}}_{i}^{(4)} 1)

and, by Eq. ( 5 ), we get

\sum_{i} (β (⊙_{i}^{(1)}) α (⊙_{i}^{(2)})) b^{'} = \sum_{i} (β (⊙_{i}^{(3)}) α (⊙_{i}^{(4)})) b^{'} .

Thus

\sum_{i} β (⊙_{i}^{(1)}) α (⊙_{i}^{(2)}) = \sum_{i} β (⊙_{i}^{(3)}) α (⊙_{i}^{(4)}) .

We note that this follows from applying

α

to (C1). So we do not get a new relation.

This completes the proof of (1) of Theorem 2.3 .

(2) We note that the precise meaning of compatibility of the unit action

(α, β)

with a relation

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)})

in the space of relations

Λ

P

is the requirement that Eq. ( 9 ) holds for any

P

-algebra

B

in the above proof. Thus the verification of the compatibility condition is the verification of (C1), (C2) and (C3). This proves (2) of Theorem 2.3 . □

3 Operads with coherent unit actions

We now apply Theorem 2.3 to give relations of ABQR operads with coherent unit actions and compatible unit actions. This allows us to easily provide operads with coherent unit actions. We also discuss some special cases.

3.1 Universal coherent relations

We describe the relations of ABQR operads

(Ω, Λ)

that admit a coherent or compatible unit action.

Theorem 3.1. Let

P = (Ω, Λ)

be an ABQR operad of dimension

n

(that is,

dim Ω = n

(1) There is a coherent unit action

(α, β)

P

with

α \neq β

if and only if for some associative operation

◆

, there is a basis

{⊙_{i}}

Ω

with

◆ = \sum_{i} ⊙_{i}

such that

Λ

contains

(◆ \otimes ◆, ◆ \otimes ◆)

and is contained in the subspace

Λ_{n, c o h}^{'}

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

with basis

\begin{matrix} {\vec{λ}}_{n, c o h}^{'} : = {\begin{matrix} (◆ \otimes ⊙_{2}, ⊙_{2} \otimes ⊙_{2}), \\ (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes ◆), \\ (⊙_{i} \otimes ⊙_{1}, ⊙_{i} \otimes ⊙_{1}), 2 \leq i \leq n, \\ (⊙_{2} \otimes ⊙_{j}, ⊙_{2} \otimes ⊙_{j}), 3 \leq j \leq n, \\ (⊙_{1} \otimes ⊙_{i}, ⊙_{i} \otimes ⊙_{2}), 3 \leq i \leq n, \\ (⊙_{i} \otimes ⊙_{j}, 0), (0, ⊙_{i} \otimes ⊙_{j}), 3 \leq i, j \leq n . \end{matrix}} \end{matrix}

(13)

(2) There is a coherent unit action

(α, β)

P

with

α = β

if and only if for some associative operation

◆

, there is a basis

{⊙_{i}}

Ω

with

◆ = \sum_{i} ⊙_{i}

such that

Λ

contains

(◆ \otimes ◆, ◆ \otimes ◆)

and is contained in the subspace

Λ_{n, c o h}^{''}

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

with basis

\begin{matrix} {\vec{λ}}_{n, c o h}^{''} : = {\begin{matrix} (⊙_{1} \otimes ◆, ⊙_{1} \otimes ◆) + (◆ \otimes ⊙_{1}, ◆ \otimes ⊙_{1}) - (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes ⊙_{1}), \\ (⊙_{i} \otimes ⊙_{j}, 0), (0, ⊙_{i} \otimes ⊙_{j}), 2 \leq i, j \leq n . \end{matrix}} . \end{matrix}

(14)

(3) There is a compatible unit action

(α, β)

P

with

α \neq β

if and only if for some associative operation

◆

, there is a basis

{⊙_{i}}

Ω

with

◆ = \sum_{i} ⊙_{i}

such that

Λ

contains

◆

and is contained in the subspace

Λ_{n, c o m p}^{'}

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

with basis

\begin{matrix} {\vec{λ}}_{n, c o m p}^{'} : = {\begin{matrix} ((⊙_{1} + ⊙_{2}) \otimes ⊙_{2}, ⊙_{2} \otimes ⊙_{2}), \\ (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes (⊙_{1} + ⊙_{2})), \\ (⊙_{i} \otimes ⊙_{1}, ⊙_{i} \otimes ⊙_{1}), 2 \leq i \leq n, \\ (⊙_{2} \otimes ⊙_{j}, ⊙_{2} \otimes ⊙_{j}), 3 \leq j \leq n, \\ (⊙_{1} \otimes ⊙_{i}, ⊙_{i} \otimes ⊙_{2}), 3 \leq i \leq n, \\ (⊙_{i} \otimes ⊙_{2}, 0), (0, ⊙_{1} \otimes ⊙_{i}), 3 \leq i \leq n, \\ (⊙_{i} \otimes ⊙_{j}, 0), (0, ⊙_{i} \otimes ⊙_{j}), 3 \leq i, j \leq n . \end{matrix}} \end{matrix}

(15)

(4) There is a compatible unit action

(α, β)

P

with

α = β

if and only if for some associative operation

◆

, there is a basis

{⊙_{i}}

Ω

with

◆ = \sum_{i} ⊙_{i}

such that

Λ

contains

◆

and is contained in the subspace

Λ_{n, c o m p}^{''}

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

with basis

\begin{matrix} {\vec{λ}}_{n, c o m p}^{''} : = {\begin{matrix} (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes ⊙_{1}), \\ (⊙_{1} \otimes ⊙_{i}, ⊙_{1} \otimes ⊙_{i}) + (⊙_{i} \otimes ⊙_{1}, ⊙_{i} \otimes ⊙_{1}), 2 \leq i \leq n, \\ (⊙_{i} \otimes ⊙_{j}, 0), (0, ⊙_{i} \otimes ⊙_{j}), 2 \leq i, j \leq n . \end{matrix}} . \end{matrix}

(16)

Proof. (1). Let

P = (Ω, Λ)

be an ABQR . Suppose for an associative operation

◆

, there is a basis

{⊙_{i}}

with

◆ = \sum_{i} ⊙_{i}

such that

Λ

is contained in the subspace of

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

generated by the relations ( 13 ). Define linear maps

α, β : Ω \to k

α (⊙_{i}) = δ_{1, i} 1, β (⊙_{i}) = δ_{2, i} 1, 1 \leq i \leq n,

where

δ_{i, j}

is the Kronecker delta. Then

α (◆) = 1 = β (◆)

and

α \neq β

. It is straightforward to check that each element in ( 13 ) satisfies the equations (C1)-(C5) in Theorem 2.3 . Therefore, each element in

Λ

satisfies the equations (C1)-(C5). Thus the unit action

(α, β)

is coherent with

Λ

. This proves the “if” part.

To prove the “only if” part, we assume that there is a unit action

(α, β)

that is coherent with

P

and

α \neq β

. In particular, there is an associative operation

◆

Ω

with

α (◆) = β (◆) = 1

. Then there are direct sum decompositions

Ω = k ◆ \oplus ker α = k ◆ \oplus ker β .

This, together with

α (◆) = β (◆) = 1

, implies that

α = β

if and only if

ker α = ker β

. So we have

ker α \neq ker β

. In fact,

ker α ⊊ ker β

and

ker β ⊊ ker α

since

dim ker α = dim ker β = n - 1

. So there are elements

⊙_{1} \in ker β

such that

α (⊙_{1}) \neq 0

and

⊙_{2} \in ker α

such that

β (⊙_{2}) \neq 0

. By rescaling, we can assume that

α (⊙_{1}) = 1 = β (⊙_{2}), α (⊙_{2}) = 0 = β (⊙_{1}) .

Thus

ker α \cap ker β

has dimension

n - 2

and contains

◆ - ⊙_{1} - ⊙_{2}

. So there is a basis

⊙_{3}, \dots, ⊙_{n}

ker α \cap ker β

such that

◆ - ⊙_{1} - ⊙_{2} = ⊙_{3} + \dots + ⊙_{n} .

Therefore

◆ = ⊙_{1} + \dots + ⊙_{n} .

Now any element in

λ \in Λ

is of the form

λ = (\sum_{i, j = 1}^{n} a_{i j} ⊙_{i} \otimes ⊙_{j}, \sum_{i, j = 1}^{n} b_{i j} ⊙_{i} \otimes ⊙_{j}) .

If the unit action

(α, β)

is coherent with this relation, then each of the five coherent equations in Theorem 2.3 holds for it. For (C1), the equation is

\sum_{i j} a_{i j} β (⊙_{i}) ⊙_{j} = \sum_{i j} b_{i j} β (⊙_{i}) ⊙_{j} .

By our choice of the basis

{⊙_{i}}

, this means

\sum_{j} a_{2, j} ⊙_{j} = \sum_{j} b_{2, j} ⊙_{j} .

Thus we have

\begin{matrix} a_{2, j} = b_{2, j}, 1 \leq j \leq n . \end{matrix}

(17)

Similarly, from (C3), we obtain

\begin{matrix} a_{i, 1} = b_{i, 1} \end{matrix}

(18)

Applying (C2), we obtain

\sum_{i j} a_{i j} α (⊙_{i}) ⊙_{j} = \sum_{i, j} b_{i j} β (⊙_{j}) ⊙_{i} .

This gives

\begin{matrix} a_{1, i} = b_{i, 2} . \end{matrix}

(19)

Applying (C4), we have

\sum_{i j} β (⊙_{j}) ⊙_{i} = \sum_{i j} β (⊙_{i}) β (⊙_{j}) ◆

which means

\sum_{i} a_{i, 2} ⊙_{i} = b_{2, 2} ◆ .

b_{2, 2} = 0

, then since

{⊙_{i}}

is a basis, we have

a_{i, 2} = 0, 1 \leq i \leq n .

b_{2, 2} \neq 0

, then again since

{⊙_{i}}

is a basis and

◆ = \sum_{i} ⊙_{i}

by construction, we must have

a_{i, 2} = b_{2, 2}

Thus we always have

\begin{matrix} a_{i, 2} = b_{2, 2}, 1 \leq i \leq n . \end{matrix}

(20)

Thus the equation always holds. As with (C4), applying (C5) gives

\begin{matrix} a_{1, 1} = b_{1, j}, 1 \leq j \leq n . \end{matrix}

(21)

Some of the relations above are duplicated. For examples,

a_{1, 1} = b_{1, 1}

is both in Eq. ( 18 ) and Eq. ( 21 ). To avoid this we list ( 20 ) and ( 21 ) first and list the rest only when needed. Note also that the above relations only involve coefficients with at least one of the subscripts in

{1, 2}

. This means that there are no restrictions among

{(⊙_{i} \otimes ⊙_{j}, 0), (0, ⊙_{i} \otimes ⊙_{j}) | 3 \leq i, j \leq n} .

Then we see that

λ

is of the following linear combination of linearly independent elements in

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

\begin{matrix} λ & = & b_{2, 2} (\sum_{i = 1}^{n} ⊙_{i} \otimes ⊙_{2}, ⊙_{2} \otimes ⊙_{2}) + a_{1, 1} (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes \sum_{j = 1}^{n} ⊙_{j}) \end{matrix}

\begin{matrix} + \sum_{i = 2}^{n} a_{i, 1} (⊙_{i} \otimes ⊙_{1}, ⊙_{i} \otimes ⊙_{1}) + \sum_{j = 3}^{n} a_{2, j} (⊙_{2} \otimes ⊙_{j}, ⊙_{2} \otimes ⊙_{j}) \end{matrix}

\begin{matrix} + \sum_{i = 3}^{n} a_{1, i} (⊙_{1} \otimes ⊙_{i}, ⊙_{i} \otimes ⊙_{2}) \end{matrix}

\begin{matrix} + \sum_{i, j = 3}^{n} a_{i, j} (⊙_{i} \otimes ⊙_{j}, 0) + \sum_{i, j = 3}^{n} b_{i, j} (0, ⊙_{i} \otimes ⊙_{j}) . \end{matrix}

Recall that

◆ = \sum_{i} ⊙_{i}

. We see that

λ

is in

Λ_{n, c o h}^{'}

defined by Eq. ( 13 ).

(2) To prove the “if” direction, suppose for an associative operation

◆

, there is a basis

{⊙_{i}}

with

◆ = \sum_{i} ⊙_{i}

such that

Λ

is contained in the subspace of

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

generated by the relations ( 14 ). Define linear maps

α = β : Ω \to k

α (⊙_{i}) = δ_{1, i} 1, 1 \leq i \leq n .

Then

α (◆) = 1 = β (◆)

. It is straightforward to check that elements in ( 14 ) satisfy the equations (C1)-(C5) in Theorem 2.3 . Therefore, each element in

Λ

satisfies the equations (C1)-(C5). Thus the unit action

(α, β)

is coherent with

Λ

Now we consider the “only if” direction. Since

α = β

, we have

Ω = k ◆ \oplus ker α

. Let

{⊙_{2}, \dots, ⊙_{n}}

be a basis of

ker α = ker β

and define

⊙_{1} = ◆ - (⊙_{2} + \dots + ⊙_{n})

. We have

◆ = \sum_{i} ⊙_{i}

α (⊙_{1}) = 1

and

α (⊙_{i}) = 0, 2 \leq i \leq n .

Let

λ = (\sum_{i, j = 1}^{n} a_{i j} ⊙_{i} \otimes ⊙_{j}, \sum_{i, j = 1}^{n} b_{i j} ⊙_{i} \otimes ⊙_{j})

be in

Λ

. Since

(α, β)

is coherent with

Λ

$∙$ by (C1), $\sum_{j} a_{1, j} ⊙_{j} = \sum_{j} b_{1, j} ⊙_{j}$ , so $a_{1, j} = b_{1, j};$
$∙$ by (C2), $\sum_{j} a_{1, j} ⊙_{j} = \sum_{i} b_{i, 1} ⊙_{i}$ , so $a_{1, j} = b_{j, 1};$
$∙$ by (C3), $\sum_{i} a_{i, 1} ⊙_{i} = \sum_{i} b_{i, 1} ⊙_{i}$ , so $a_{i, 1} = b_{i, 1};$
$∙$ by (C4), $\sum_{i} a_{i, 1} ⊙_{i} = b_{1, 1} ◆$ , so $a_{i, 1} = b_{1, 1};$
$∙$ by (C5), $a_{1, 1} ◆ = \sum_{j} b_{1, j} ⊙_{j}$ , so $a_{1, 1} = b_{1, j}$ .

Therefore,

a_{i, 1} = a_{1, j} = b_{i, 1} = b_{1, j}, 1 \leq i, j \leq n .

Thus

\begin{matrix} (\sum_{i = 1 o r j = 1} a_{i j} ⊙_{i} \otimes ⊙_{j}, \sum_{i = 1 o r j = 1} b_{i j} ⊙_{i} \otimes ⊙_{j}) \end{matrix}

\begin{matrix} = & a_{1, 1} (\sum_{i = 1 o r j = 1} ⊙_{i} \otimes ⊙_{j}, \sum_{i = 1 o r j = 1} ⊙_{i} \otimes ⊙_{j}) \end{matrix}

\begin{matrix} = & a_{1, 1} ((⊙_{1} \otimes ◆, ⊙_{1} \otimes ◆) + ((◆ - ⊙_{1}) \otimes ⊙_{1}, (◆ - ⊙_{1}) \otimes ⊙_{1})) \end{matrix}

\begin{matrix} = & a_{1, 1} ((⊙_{1} \otimes ◆, ⊙_{1} \otimes ◆) + (◆ \otimes ⊙_{1}, ◆ \otimes ⊙_{1}) - (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes ⊙_{1})) . \end{matrix}

On the other hand, the coherence equations impose no restriction on other elements in

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

. So the subspace

Λ_{n, c o h}^{''}

of relations coherent with

(α, β)

has its basis given in Eq. ( 14 ).

The proofs of part (3) of Theorem 3.1 follows from a similar analysis as for part (1).

But we only need to consider (C1)-(C3) which give relations ( 17 ), ( 18 ) and ( 19 ).

Likewise, for the proof of part (4), we only consider (C1)-(C3) in the proof of part (2). Grouping the resulting relations, we obtain Eq. ( 16 ). □

3.2 Special cases

We now consider cases where

Ω

is of dimension 2 and 3.

Suppose

α \neq β

. Then from Theorem 3.1 we easily check that

{\vec{λ}}_{2, c o h}^{'} = Λ_{2, c o m p}^{'} = {(◆ \otimes ⊙_{2}, ⊙_{2} \otimes ⊙_{2}), (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes ◆), (⊙_{2} \otimes ⊙_{1}, ⊙_{2} \otimes ⊙_{1})} .

Replacing

⊙_{1}

≺

and replacing

⊙_{2}

≻

, we obtain the following improvement of Proposition 1.2 in [Lo6] .

Corollary 3.2. Let

dim Ω = 2

. The following statements are equivalent.

(1) There is a coherent unit action $(α, β)$ on an ABQR operad $(Ω, Λ)$ with $α \neq β$ ;
(2) There is a compatible unit action $(α, β)$ on an ABQR operad $(Ω, Λ)$ with $α \neq β$ ;
(3) There is an associative operation $◆$ and a basis $(≺, ≻)$ of $Ω$ with $◆ = ≺ + ≻$ , such that $Λ$ contains $◆$ and is contained in the subspace of $Ω^{\otimes 2} \oplus Ω^{\otimes 2}$ with basis ${(◆ \otimes ≻, ≻ \otimes ≻), (≺ \otimes ≺, ≺ \otimes ◆), (≻ \otimes ≺, ≻ \otimes ≺)} .$

Next suppose

α = β

. Then from Theorem 3.1 , we get

Corollary 3.3. Let

dim Ω = 2

(1) There is a coherent unit action $(α, β)$ on an ABQR operad $(Ω, Λ)$ with $α = β$ if and only there is an associative operation $◆$ and a basis $(⊙_{1}, ⊙_{2})$ of $Ω$ with $◆ = ⊙_{1} + ⊙_{2}$ , such that $Λ$ contains $◆$ and is contained in the subspace of $Ω^{\otimes 2} \oplus Ω^{\otimes 2}$ with basis ${(⊙_{1} \otimes ◆, ⊙_{1} \otimes ◆) + (⊙_{2} \otimes ⊙_{1}, ⊙_{2} \otimes ⊙_{1}), (⊙_{2} \otimes ⊙_{2}, 0), (0, ⊙_{2} \otimes ⊙_{2})} .$
(2) There is a compatible unit action $(α, β)$ on an ABQR operad $(Ω, Λ)$ with $α = β$ if and only there is an associative operation $◆$ and a basis $(⊙_{1}, ⊙_{2})$ of $Ω$ with $◆ = ⊙_{1} + ⊙_{2}$ , such that $Λ$ is contained in the subspace of $Ω^{\otimes 2} \oplus Ω^{\otimes 2}$ with basis ${(⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes ⊙_{1}), (⊙_{1} \otimes ⊙_{2}, ⊙_{1} \otimes ⊙_{2}) + (⊙_{2} \otimes ⊙_{1}, ⊙_{2} \otimes ⊙_{1}), (⊙_{2} \otimes ⊙_{2}, 0), (0, ⊙_{2} \otimes ⊙_{2})} .$

Now we consider dimension three case. From Theorem 3.1 we get

Corollary 3.4. Let

dim Ω = 3

. There is a coherent unit action

(α, β)

on an ABQR operad

(Ω, Λ)

with

α \neq β

if and only if there is a basis

(≺, ≻, \circ)

Ω

such that

Λ

is contained in the subspace with basis

\begin{matrix} {\vec{λ}}_{3, c o h}^{'} = & {(≺ \otimes ≺, ≺ \otimes ◆), (≻ \otimes ≺, ≻ \otimes ≺), (◆ \otimes ≻, ≻ \otimes ≻), (≻ \otimes \circ, ≻ \otimes \circ), \end{matrix}

\begin{matrix} (≺ \otimes \circ, \circ \otimes ≻), (\circ \otimes ≺, \circ \otimes ≺), (\circ \otimes \circ, 0), (0, \circ \otimes \circ)} . \end{matrix}

Proof. It follows from Theorem 3.1 by replacing ${⊙_{1}, ⊙_{2}, ⊙_{3}}$ with ${≺, ≻, \circ}$ . □

Clearly the relations in Eq. ( 3 ) and ( 4 ) of the dendriform trialgebra and NS-algebra, respectively, are subspaces of

{\vec{λ}}_{3, c o h}^{'}

, so have coherent unit actions, as were already shown in [Lo6, Le2] . We also note that when

n \geq 3

, compatibility does not imply coherence for unit actions.

4 Coherent Actions on Products and Duals

We briefly recall the concept of the black square product of ABQR operads [E-G, Lo7] , and show that coherent and compatible unit actions are preserved by the black square product. We then recall the concept of the dual operad and show that, other than a simple case, coherence and compatibility are not preserved by taking the duals.

4.1 Products of operads

For ABQR operads

(Ω_{1}, Λ_{1})

and

(Ω_{2}, Λ_{2})

, and for

⊙^{(i)} \in Ω_{i}, i = 1, 2

, we use a column vector

[\binom{⊙_{1}}{⊙_{2}}]

to denote the tensor product

⊙_{1} \otimes ⊙_{2} \in Ω_{1} \otimes Ω_{2}

. For

f_{i} = (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)}) \in Ω_{i}^{\otimes 2} \oplus Ω_{i}^{\otimes 2}, i = 1, 2,

define

[\binom{f_{1}}{f_{2}}] = ([\binom{⊙_{1}^{(1)}}{⊙_{2}^{(1)}}] \otimes [\binom{⊙_{1}^{(2)}}{⊙_{2}^{(2)}}], [\binom{⊙_{1}^{(3)}}{⊙_{2}^{(3)}}] \otimes [\binom{⊙_{1}^{(4)}}{⊙_{2}^{(4)}}]) \in (Ω_{1} \otimes Ω_{2})^{\otimes 2} \oplus (Ω_{1} \otimes Ω_{2})^{\otimes 2} .

This extends by bilinearity to all

f_{i} \in Ω_{i}^{\otimes 2} \oplus Ω_{i}^{\otimes 2}, i = 1, 2 .

We define a subspace of

(Ω_{1} \otimes Ω_{2})^{\otimes 2} \oplus (Ω_{1} \otimes Ω_{2})^{\otimes 2}

Λ_{1} ■ Λ_{2} = {[\binom{f_{1}}{f_{2}}] | f_{i} \in Λ_{i}, i = 1, 2} .

Λ_{1} ■ Λ_{2}

is a set of relations for the operator set

Ω_{1} \otimes Ω_{2}

The black square product of

(Ω_{1}, Λ_{1})

and

(Ω_{2}, Λ_{2})

is the operad

(Ω_{1} \otimes Ω_{2}, Λ_{1} ■ Λ_{2})

We recall the following results from [E-G] for later reference.

Proposition 4.1.

(1) The quadri-algebra of Aguiar and Loday [A-L] , defined by four binary operations ${↗, ↖, ↘, ↙}$ and 9 relations, is isomorphic to the square product $(Ω_{D}, Λ_{D}) ■ (Ω_{D}, Λ_{D}) = (Ω_{D} \otimes Ω_{D}, Λ_{D} ■ Λ_{D})$ of the dendriform dialgebra $(Ω_{D}, Λ_{D})$ in Eq. ( 2 ).
(2) The ennea-algebra of Leroux [Le1] , defined by 9 binary operations $↖, ↑, ↗, ≺, \circ, ≻, ↙, ↓, ↘$ and 49 relations, is isomorphic to the square product $(Ω_{T}, Λ_{T}) ■ (Ω_{T}, Λ_{T})$ of the dendriform trialgebra in Eq. ( 3 ).
(3) The dendriform-Nijenhuis algebra [Le2] , equipped with 9 binary operations $↗, ↘, ↙, ↖, ↑, ↓, \tilde{≺}, \tilde{≻}, \tilde{∙}$ and 28 relations, is isomorphic to the square product $(Ω_{T}, Λ_{T}) ■ (Ω_{N}, Λ_{N})$ of the dendriform trialgebra and the NS-algebra $(Ω_{N}, Λ_{N})$ [Le2, E-G] .
(4) The octo-algebra [Le3] , defined using 8 operations $↗_{i}, ↖_{i}, ↙_{i}, ↘_{i}, i = 1, 2$ and 27 relations, is isomorphic to the third power of the dendriform dialgebra $(Ω_{D}, Λ_{D})$ .

4.2 Unit actions on products

We now use Theorem 2.3 to show that coherent unit actions on ABQR operads are preserved by the black square product and thus give rise to Hopf algebra structures on the free objects.

For each

i = 1, 2

, let

P_{i} : = (Ω_{i}, Λ_{i})

be an ABQR operad and let

(α_{i}, β_{i})

be a unit action on

P_{i}

with

α_{i} (◆_{i}) = β_{i} (◆_{i})

. It is easy to see [E-G] that the product operad

P : = P_{1} ■ P_{2} : = (Ω_{1} \otimes Ω_{2}, Λ_{1} ■ Λ_{2})

, with the associative operation

[\binom{◆_{1}}{◆_{2}}]

, is also an ABQR operad. Define

\begin{matrix} α : = α_{1} \otimes α_{2} (r e s p . β : = β_{1} \otimes β_{2}) : Ω_{1} \otimes Ω_{2} & \to & k, \end{matrix}

\begin{matrix} [\binom{⊙_{1}}{⊙_{2}}] & \mapsto & α_{1} (⊙_{1}) α_{2} (⊙_{2}) (r e s p . β_{1} (⊙_{1}) β_{2} (⊙_{2})) . \end{matrix}

Then

(α_{1} \otimes α_{1}, β_{1} \otimes β_{2})

defines a unit action on

P

Theorem 4.2. Let

P_{i} : = (Ω_{i}, Λ_{i}), i = 1, 2,

be ABQR operads with coherent unit actions

(α_{i}, β_{i})

. Then the unit action

(α_{1} \otimes α_{2}, β_{1} \otimes β_{2})

on the ABQR operad

P : = P_{1} ■ P_{2} : = (Ω_{1} \otimes Ω_{2}, Λ_{1} ■ Λ_{2})

is also coherent. Therefore, The augmented free

P

-algebra

P (V)_{+}

on a

k

-vector space

V

is a connected Hopf algebra.

It will be clear from the proof that, if

P_{1}

and

P_{2}

have compatible unit actions, then so does their product.

Proof. By Theorem 2.3 , we just need to verify that the relations in

Λ_{1} ■ Λ_{2}

satisfy (C1)–(C5) with the unit action

(α, β)

P

. We recall that a relation in

Λ_{1} ■ Λ_{2}

is of the form

[\binom{f_{1}}{f_{2}}] : = \sum_{i, j} ([\binom{⊙_{i}^{(1)}}{⊙_{j}^{(1)}}] \otimes [\binom{⊙_{i}^{(2)}}{⊙_{j}^{(2)}}], [\binom{⊙_{i}^{(3)}}{⊙_{j}^{(3)}}] \otimes [\binom{⊙_{i}^{(4)}}{⊙_{j}^{(4)}}])

for a

\sum_{i} (⊙_{i}^{(1)} \otimes ⊙_{i}^{(2)}, ⊙_{i}^{(3)} \otimes ⊙_{i}^{(4)}) \in Λ_{1}

and a

\sum_{j} (⊙_{j}^{(1)} \otimes ⊙_{j}^{(2)}, ⊙_{j}^{(3)} \otimes ⊙_{j}^{(4)}) \in Λ_{2}

Then by linearity of the product

[\binom{⊙_{1}}{⊙_{2}}]

and the equation (C1) for

P_{1}

and

P_{2}

, we have

\begin{matrix} \sum_{i, j} β ([\binom{⊙_{i}^{(1)}}{⊙_{j}^{(1)}}]) [\binom{⊙_{i}^{(2)}}{⊙_{j}^{(2)}}] = \sum_{i, j} β_{1} (⊙_{i}^{(1)}) β_{2} (⊙_{j}^{(1)}) [\binom{⊙_{i}^{(2)}}{⊙_{j}^{(2)}}] \end{matrix}

\begin{matrix} = & \sum_{i, j} [\binom{β_{1} (⊙_{i}^{(1)}) ⊙_{i}^{(2)}}{β_{2} (⊙_{j}^{(1)}) ⊙_{j}^{(2)}}] \end{matrix}

\begin{matrix} = & [\binom{\sum_{i} β_{1} (⊙_{i}^{(1)}) ⊙_{i}^{(2)}}{\sum_{j} β_{2} (⊙_{j}^{(1)}) ⊙_{j}^{(2)}}] = [\binom{\sum_{i} β_{1} (⊙_{i}^{(3)}) ⊙_{i}^{(4)}}{\sum_{j} β_{2} (⊙_{j}^{(3)}) ⊙_{j}^{(4)}}] \end{matrix}

\begin{matrix} = & \sum_{i, j} [\binom{β_{1} (⊙_{i}^{(3)}) ⊙_{i}^{(4)}}{β_{2} (⊙_{j}^{(3)}) ⊙_{j}^{(4)}}] = \sum_{i, j} β_{1} (⊙_{i}^{(3)}) β_{2} (⊙_{j}^{(3)}) [\binom{⊙_{i}^{(4)}}{⊙_{j}^{(4)}}] \end{matrix}

\begin{matrix} = & \sum_{i, j} β ([\binom{⊙_{i}^{(3)}}{⊙_{j}^{(3)}}]) [\binom{⊙_{i}^{(4)}}{⊙_{j}^{(4)}}] . \end{matrix}

This gives (C1) for the product operad

P

. Conditions (C2) and (C3) can be verified in the same way.

For (C4), we have

\begin{matrix} \sum_{i, j} β ([\binom{⊙_{i}^{(2)}}{⊙_{j}^{(2)}}]) [\binom{⊙_{i}^{(1)}}{⊙_{j}^{(1)}}] = \sum_{i, j} β_{1} (⊙_{i}^{(2)}) β_{2} (⊙_{j}^{(2)}) [\binom{⊙_{i}^{(1)}}{⊙_{j}^{(1)}}] \end{matrix}

\begin{matrix} = & [\binom{\sum_{i} β_{1} (⊙_{i}^{(2)}) ⊙_{i}^{(1)}}{\sum_{j} β_{2} (⊙_{j}^{(2)}) ⊙_{j}^{(1)}}] \end{matrix}

\begin{matrix} = & [\binom{\sum_{i} β_{1} (⊙_{i}^{(3)}) β_{1} (⊙_{i}^{(4)}) ◆_{1}}{\sum_{j} β_{2} (⊙_{j}^{(3)}) β_{2} (⊙_{j}^{(4)}) ◆_{2}}] = \sum_{i, j} [\binom{β_{1} (⊙_{i}^{(3)}) β (⊙_{i}^{(4)}) ◆_{1}}{β_{2} (⊙_{j}^{(3)}) β (⊙_{j}^{(4)}) ◆_{2}}] \end{matrix}

\begin{matrix} = & \sum_{i, j} β_{1} (⊙_{i}^{(3)}) β_{1} (⊙_{i}^{(4)}) β_{2} (⊙_{j}^{(3)}) β_{2} (⊙_{j}^{(4)}) [\binom{◆_{1}}{◆_{2}}] \end{matrix}

\begin{matrix} = & \sum_{i, j} β ([\binom{⊙_{i}^{(3)}}{⊙_{j}^{(3)}}]) β ([\binom{⊙_{i}^{(4)}}{⊙_{j}^{(4)}}]) [\binom{◆_{1}}{◆_{2}}] . \end{matrix}

This gives (C4) for the product operad

P

. The same works for (C5). □

By Proposition 4.1 and Theorem 4.2 we recover the following results on Hopf algebras.

Corollary 4.3.

(1) (Loday[Lo6] ) Augmented free quadri-algebras are Hopf algebras;
(2) (Leroux[Le1] ) Augmented free ennea-algebras are Hopf algebras;
(3) (Leroux[Le2] ) Augmented free Nijenhuis-dendriform algebras are Hopf algebras;
(4) (Leroux[Le3] ) Augmented free octo-algebras are Hopf algebras;

4.3 Duality of ABQR operads

We now recall the definition of the dual [E-G, Lo4, Lo7] of an ABQR operad before we study the relation between coherent unit actions and taking duals.

For an ABQR operad

P = (Ω, Λ)

, let

\overset{ˇ}{Ω} : = H o m (Ω, k)

be the dual space of

Ω

, giving the natural pairing

〈, 〉_{Ω} : Ω \times \overset{ˇ}{Ω} \to k .

Then

{\overset{ˇ}{Ω}}^{\otimes 2}

is identified with the dual space of

Ω^{\otimes 2}

, giving the natural pairing

〈, 〉^{\otimes 2} : Ω^{\otimes 2} \times {\overset{ˇ}{Ω}}^{\otimes 2} \to k, 〈 x \otimes y, a \otimes b 〉^{\otimes 2} = 〈 x, a 〉_{Ω} 〈 y, b 〉_{Ω} .

We then define a pairing

〈, 〉 : (Ω^{\otimes 2} \oplus Ω^{\otimes 2}) \times ({\overset{ˇ}{Ω}}^{\otimes 2} \oplus {\overset{ˇ}{Ω}}^{\otimes 2}) \to k

〈 (α, β), (γ, δ) 〉 = 〈 α, γ 〉^{\otimes 2} - 〈 β, δ 〉^{\otimes 2}, α, β \in Ω^{\otimes 2}, γ, δ \in {\overset{ˇ}{Ω}}^{\otimes 2} .

We now define

Λ^{⊥}

to be the annihilator of

Λ \subseteq Ω^{\otimes 2} \oplus Ω^{\otimes 2}

{\overset{ˇ}{Ω}}^{\otimes 2} \oplus {\overset{ˇ}{Ω}}^{\otimes 2}

under the pairing

〈, 〉

. We call

P^{!} : = (\overset{ˇ}{Ω}, Λ^{⊥})

the dual operad of

P = (Ω, Λ)

which is the Koszul dual in our special case. It follows from the definition that

(P^{!})^{!} = P

Example 4.4. (Associative dialgebra [Lo4,Proposition8.3] ) Let

(Ω_{D}, Λ_{D})

be the operad for the dendriform dialgebra. Let

{⊣, ⊢} \in {\overset{ˇ}{Ω}}_{D}

be the dual basis of

{≺, ≻}

(in this order). Then

Λ_{D}^{⊥}

is given by

\begin{matrix} {(⊣ \otimes ⊣, ⊣ \otimes ⊣), (⊢ \otimes ⊢, ⊢ \otimes ⊢), (⊣ \otimes ⊣, ⊣ \otimes ⊢), (⊢ \otimes ⊣, ⊢ \otimes ⊣), (⊣ \otimes ⊢, ⊢ \otimes ⊢)} . \end{matrix}

(22)

It is called the associative dialgebra

(Ω_{A D}, Λ_{A D})

Other examples of duals of ABQR operads can be found in [L-R2, E-G] .

4.4 Unit actions on duals

A natural question to ask is whether the dual of an ABQR operad

(Ω, Λ)

with a coherent unit action still has a coherent unit action. When

Ω

has only one element, the answer is positive. This is because the element must be an associative operation

◆

. That is,

(Ω, Λ)

is the operad for associative algebras, with the coherent unit action given by

α (◆) = β (◆) = 1

. The dual operad is isomorphic to itself, so again admits a coherent unit actions. We now show that this is the only case that coherent unit action is preserved by duality.

Theorem 4.5. Suppose that an ABQR operad

P = (Ω, Λ)

with

dim Ω \geq 2

has a compatible unit action. Let

P^{!} = (\overset{ˇ}{Ω}, Λ^{⊥})

be the dual operad.

(1) $P^{!}$ has an associative operation, so is an ABQR operad.
(2) $P^{!}$ does not have a compatible unit action for any associative operation.

The same is true when compatible is replaced by coherent in all statement.

For example the associative dialgebra does not have a coherent unit action even though it has two linearly independent associative binary operations

⊢

and

⊣

Proof. (1) Let

dim Ω = n

. First assume that the compatible unit action on the given ABQR operad

(Ω, Λ)

satisfies

α \neq β

. Let

{⊙_{1}, \dots, ⊙_{n}}

be the bases of

Ω

given in the proof of Theorem 3.1 .(3). So we have

\sum_{i} ⊙_{i} = ◆, α (⊙_{i}) = δ_{1, i}, β (⊙_{i}) = δ_{2, i} .

Then by Theorem 3.1 ,

Λ

is a subspace of the subspace

Λ_{n, c o m p}^{'}

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

with a basis

\begin{matrix} {\vec{λ}}_{n, c o m p}^{'} : = {\begin{matrix} ((⊙_{1} + ⊙_{2}) \otimes ⊙_{2}, ⊙_{2} \otimes ⊙_{2}), \\ (⊙_{1} \otimes ⊙_{1}, ⊙_{1} \otimes (⊙_{1} + ⊙_{2})), \\ (⊙_{i} \otimes ⊙_{1}, ⊙_{i} \otimes ⊙_{1}), 2 \leq i \leq n, \\ (⊙_{2} \otimes ⊙_{j}, ⊙_{2} \otimes ⊙_{j}), 3 \leq j \leq n, \\ (⊙_{1} \otimes ⊙_{i}, ⊙_{i} \otimes ⊙_{2}), 3 \leq i \leq n, \\ ((⊙_{i} \otimes ⊙_{2}, 0), (0, ⊙_{1} \otimes ⊙_{i}), 3 \leq i \leq n, \\ (⊙_{i} \otimes ⊙_{j}, 0), (0, ⊙_{i} \otimes ⊙_{j}), 3 \leq i, j \leq n . \end{matrix}} \end{matrix}

(23)

Let

{{\overset{ˇ}{⊙}}_{i}}

be the dual basis of

{⊙_{i}}

. So

{\overset{ˇ}{⊙}}_{i} (⊙_{j}) = δ_{i, j}, 1 \leq i, j \leq n .

Then the pairing between

Ω^{\otimes 2} \oplus Ω^{\otimes 2}

and

{\overset{ˇ}{Ω}}^{\otimes 2} \oplus {\overset{ˇ}{Ω}}^{\otimes 2}

is given by

〈 (⊙_{i} \otimes ⊙_{j}, ⊙_{k} \otimes ⊙_{ℓ}), ({\overset{ˇ}{⊙}}_{s} \otimes {\overset{ˇ}{⊙}}_{t}, {\overset{ˇ}{⊙}}_{u} \otimes {\overset{ˇ}{⊙}}_{v}) 〉 = δ_{i, s} δ_{j, t} - δ_{k, u} δ_{ℓ, v} .

Consider

x = ({\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{2}, {\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{2})

. Then the pairing between

x

and the first element in

λ_{n, c o m p}^{'}

1 - 1 = 0

. The pairing between

x

and every other element in

λ_{n, c o m p}^{'}

is also 0 since

{\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{2}

does not occur in the element. Thus

{\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{2}

is in

{Λ_{n, c o m p}^{'}}^{⊥}

and hence in

Λ^{⊥}

, the relation space of

P^{!}

. This shows that

P^{!}

has an associative operation.

Next assume that the compatible unit action on the given ABQR operad

(Ω, Λ)

satisfies

α = β

. Let

{⊙_{1}, \dots, ⊙_{n}}

n = dim Ω

, be the bases of

Ω

given in the proof of Theorem 3.1 .(4). Then as in the last case, we check that

({\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{1}, {\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{1})

is in

{Λ_{n, c o m p}^{''}}^{⊥} \subseteq Λ^{⊥}

, again showing that

P^{!}

is ABQR .

(2) Suppose that there is a choice of associative operation

◆

\overset{ˇ}{Ω}

and a unit action

(α, β)

that is compatible with

(\overset{ˇ}{Ω}, Λ^{⊥})

with

α (◆) = β (◆) = 1

. First assume that

α \neq β

We easily verify that the three relations

\begin{matrix} ({\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{1}, {\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{1}), ({\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{1}, {\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{1}), ({\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{2}, {\overset{ˇ}{⊙}}_{2} \otimes {\overset{ˇ}{⊙}}_{2}) \end{matrix}

(24)

are in

{Λ_{n, c o m p}^{'}}^{⊥}

and hence in

Λ^{⊥}

. So they should satisfy the compatible equations (C1)-(C3) in Theorem 3.1 . Applying (C2) to the first relation in Eq. ( 24 ), we obtain

α ({\overset{ˇ}{⊙}}_{2}) {\overset{ˇ}{⊙}}_{1} = β ({\overset{ˇ}{⊙}}_{1}) {\overset{ˇ}{⊙}}_{2}

. So

\begin{matrix} α ({\overset{ˇ}{⊙}}_{2}) = β ({\overset{ˇ}{⊙}}_{1}) = 0 . \end{matrix}

(25)

Applying (C2) to the second relation in Eq. ( 24 ), we have

α ({\overset{ˇ}{⊙}}_{1}) = β ({\overset{ˇ}{⊙}}_{1})

, yielding

α ({\overset{ˇ}{⊙}}_{1}) = 0

by Eq. ( 25 ). Applying (C2) to the third relation in Eq. ( 24 ), we have

α ({\overset{ˇ}{⊙}}_{2}) = β ({\overset{ˇ}{⊙}}_{2})

, giving

β ({\overset{ˇ}{⊙}}_{2}) = 0

by Eq. ( 25 ). For

i \geq 2

, we check that

({\overset{ˇ}{⊙}}_{i} \otimes {\overset{ˇ}{⊙}}_{2}, 0)

and

(0, {\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{i})

are in

Λ^{⊥}

so satisfy (C1)-(C2). Applying (C1) to the first relation gives

β ({\overset{ˇ}{⊙}}_{i}) {\overset{ˇ}{⊙}}_{2} = 0

. So

β ({\overset{ˇ}{⊙}}_{i}) = 0

. Applying (C3) to the second equation gives

α ({\overset{ˇ}{⊙}}_{i}) {\overset{ˇ}{⊙}}_{1} = 0

α ({\overset{ˇ}{⊙}}_{i}) = 0

. Therefore

α

and

β

are identically zero. But this is impossible, since

α (◆)

and

β (◆)

should be 1.

Next assume that

α = β

. For each

i \geq 2

(there is such an

i

since

dim Ω \geq 2

), we check that

({\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{i}, {\overset{ˇ}{⊙}}_{1} \otimes {\overset{ˇ}{⊙}}_{i})

and

({\overset{ˇ}{⊙}}_{i} \otimes {\overset{ˇ}{⊙}}_{1}, {\overset{ˇ}{⊙}}_{i} \otimes {\overset{ˇ}{⊙}}_{1})

are in

{Λ_{n, c o m p}^{''}}^{⊥} \subseteq Λ^{⊥}

. Applying (C2) to them, we get

α ({\overset{ˇ}{⊙}}_{1}) {\overset{ˇ}{⊙}}_{i} = β ({\overset{ˇ}{⊙}}_{i}) {\overset{ˇ}{⊙}}_{1}, α ({\overset{ˇ}{⊙}}_{i}) {\overset{ˇ}{⊙}}_{1} = β ({\overset{ˇ}{⊙}}_{1}) {\overset{ˇ}{⊙}}_{i} .

α ({\overset{ˇ}{⊙}}_{1}) = β ({\overset{ˇ}{⊙}}_{i}) = 0

and

α ({\overset{ˇ}{⊙}}_{i}) = β ({\overset{ˇ}{⊙}}_{1}) = 0, i \geq 2 .

Thus

α

and

β

are identically zero, giving a contradiction.

Finally if

P

has a coherent unit action, then it automatically has a compatible unit action. So by part 1,

P^{!}

is an ABQR operad, but does not have a compatible unit action. It therefore does not have a coherent unit action. □

Acknowledgements: The first author would like to thank the Ev. Studienwerk Villigst and the theory department of the Physikalisches Institut, at Bonn University for generous support. The second author thanks a grant from the Research Council of the Rutgers University. We are grateful to Prof. J.-L. Loday for suggestions and encouragement.

References

M. Aguiar and F. Sottile, Structure of the Loday-Ronco Hopf algebra of trees, preprint, May 2004.
M. Aguiar and F. Sottile, Cocommutative Hopf algebras of permutations and trees, preprint, March 2004, ArXiv:math.QA/0403101.
M. Aguiar and J.-L. Loday, Quadri-algebras, J. Pure Applied Algebra 191 (2004), 205-221. ArXiv:math.QA/03090171.
F. Chapoton, Un théorème de Cartier-Milnor-Moore-Quillen pour les bigèbres dendriformes et les algèbres braces, J. Pure Appl. Alg., 168, (2002), 1-18.
E. Ebrahimi-Fard and L. Guo, On products and duality of binary, quadratic, regular operads, J. Pure Applied Algebra, in press, ArXiv:math.RA/0407162.
L. Foissy, Les algèbres de Hopf des arbres enracinés décorés II, Bull. Sci. Math., 126, (2002), 249-288.
A. Frabetti, Dialgebra homology of associative algebras, C. R. Acad. Sci. Paris, 325, (1997), 135-140.
A. Frabetti, Leibniz homology of dialgebras of matrices, J. Pure Appl. Alg., 129, (1998), 123-141.
L. Gerritzen and R. Holtkamp, Hopf co-addition for free magma algebras and the non-associative Hausdorff series, J. of Algebra, 265 (2003), 264-284.
V. Ginzburg and M. Kapranov, Koszul duality for operads. Duke Math. J., 76 (1994), 203-272.
R. Holtkamp, Comparison of Hopf algebras on trees, Archiv der Mathematik, Vol.80, (2003), 368-383.
R. Holtkamp, On Hopf algebra structures over operads, preprint, July 2004, ArXiv:math.RA/0407074.
P. Leroux, Ennea-algebras, J. Algebra, 281 (2004), 287-302, ArXiv:math.QA/0309213.
P. Leroux, Construction of Nijenhuis operators and dendriform trialgebras, preprint, Nov. 2003, ArXiv:math.QA/0311132.
P. Leroux, On some remarkable operads constructed from Baxter operators, preprint, Nov. 2003, ArXiv:math.QA/0311214.
J.-L. Loday, Une version non commutative des algèbre de Lie: les algèbres de Leibniz, Ens. Math., 39, (1993), 269-293.
J.-L. Loday, Algèbres ayant deux opérations associatives (digèbres), C. R. Acad. Sci. Paris 321 (1995), 141-146.
J.-L. Loday, La renaissance des opérades, Séminaire Bourbaki, Vol. 1994/95. Astérisque No. 237 (1996), Exp. No. 792, 3, 47-74.
J.-L. Loday, Dialgebras, in Dialgebras and related operads, Lecture Notes in Math., 1763, (2002), 7-66. ArXiv:math.QA/0102053.
J.-L. Loday, Arithmetree, J. Algebra, 258 (2002), 275-309.
J.-L. Loday, Scindement d'associativité et algèbres de Hopf. to appear in the Proceedings of the Conference in honor of Jean Leray, Nantes (2002), Séminaire et Congrès (SMF) 9 (2004), 155-172.
J.-L. Loday, Completing the operadic butterfly, preprint, June 2004.
J.-L. Loday and M. O. Ronco, Hopf algebra of the planar binary trees, Adv. Math., 139, (1998), 293-309.
J.-L. Loday and M. Ronco, Trialgebras and families of polytopes, in “Homotopy Theory: Relations with Algebraic Geometry, Group Cohomology, and Algebraic K-theory” Contemporary Mathematics 346 (2004). ArXiv:math.AT/0205043.
J.-L. Loday and M. Ronco, Algèbre de Hopf colibres, C. R. Acad. Sci. Paris, 337, (2003), 153-158.
J.-L. Loday, J. D. Stasheff and A. A. Voronov, Operads: Proceedings of Renaissance Conferences: Special Session and International Conference on Moduli Spaces, Operads, and Representation Theory (Contemporary Mathematics, Vol 202), AMS, (1997).
M. Markl, S. Shnider, J. Stasheff, Operads in Algebra, Topology and Physics, AMS, 2002.
T. Pirashvili, Sets with two associative operations, C.E.J.M. 2 (2003), 169-183.
M. R. Ronco, Eulerian idempotents and Milnor-Moore theorem for certain non-commutative Hopf algebras, J. Algebra, 254, (2002), 152-172.

Institut Henri Poincare, 11, rue Pierre et Marie Curie F-75231 Paris Cedex 05 FRANCE and Universitat Bonn Physikalisches Institut, Nussallee 12, D-53115 Bonn, Germany E-mail address : kurusch@ihes.fr Department of Mathematics and Computer Science, Rutgers University, Newark, NJ 07102, USA E-mail address : liguo@newark.rutgers.edu

November 27, 2006

Kurusch Ebrahimi-Fard

Li Guo