arXiv reader

Anti-successors and anti-replicators, Manin products and Koszul duality

Chengming Bai, Li Guo, Yunhe Sheng and You Wang

arXiv
2610.21365v1 · PDF · 15 October 2026
Categories
math.RA, math-ph, math.AT, math.MP
Licence
CC BY 4.0 The site shows the text.

Abstract

We develop an operadic theory of anti-splitting and anti-replication for binary operads. For a binary operad \opd, we define its anti-bisuccessor \ABSu(\opd) and anti-trisuccessor \ATSu(\opd), with the property that an operation of \opd splits into a negative sum of two or three operations, yielding canonical operad morphisms \opd \to\ABSu(\opd) and \opd\to\ATSu(\opd). For binary quadratic operads, the process gives Manin black products,

\ABSu(\opd) \cong \antiprelie \bullet \opd, \qquad \ATSu(\opd)\cong \antipostlie\bullet\opd.

Dually, we introduce the anti-duplicator \ADup(\mathcal P) and anti-triplicator \ATrip(\mathcal P) and prove

\ADup(\opd^!)\cong \ABSu(\opd)^!, \qquad \ATrip(\opd^!)\cong \ATSu(\opd)^!.

Consequently, the anti-duplicator gives a new way to calculate Manin white products. %, in particular, \ADup(\opd)\cong \antiperm\circ\opd. The framework recovers known anti-splittings, including anti-dendriform, anti-pre-Lie, and anti-pre-Poisson structures, in a uniform operadic form and produces further examples and dual pairs.

Items 88

Definition

Definition 2.1

6LX5
  1. (i) Let \calt denote the set of planar binary reduced rooted trees together with the trivial tree \vcenter{\xymatrix@M=4pt@R=8pt@C=4pt{\ar@{-}[d]\\ \\}}. If t\in \calt has n leaves, we call t an n-tree. The trivial tree \vcenter{\xymatrix@M=4pt@R=8pt@C=4pt{\ar@{-}[d]\\ \\}} has one leaf.

  2. (ii) Let \vertset be a set. By a decorated tree we mean a tree t of \calt together with a decoration on the vertices of t by elements of \vertset and a decoration on the leaves of t by distinct positive integers. Let t(\vertset) denote the set of decorated trees for t and denote by

    \calt(\vertset):=\coprod_{ t \in \calt} t(\vertset).

    If \tau\in t(\vertset) for an n-tree t, we call \tau a labeled n-tree.

  3. (iii) For \tau\in \calt(\vertset), we let \vin(\tau) (resp. \lin(\tau)) denote the set (resp. ordered set) of labels of the vertices (resp. leaves) of \tau.

  4. (iv) Let \tau_{l},\tau_{r}\in \calt(\vertset) with disjoint sets of leaf labels and \omega\in \vertset. The grafting of \tau_{l} and \tau_{r} along \omega is denoted by \tau= \tau_{l} \vee_{\omega} \tau_{r}, which is an element in \calt(\vertset).

  5. (v) For \tau\in \calt(\vertset) with |\lin(\tau)|>1, we let \tau= \tau_{l} \vee_{\omega} \tau_{r} denote the unique decomposition of \tau as a grafting of \tau_{l} and \tau_{r} in \calt(\vertset) along \omega \in \vertset.

Definition

Definition 2.2

WHW3

Let \gensp be a vector space with a basis \genbas.

  1. (i) Define a vector space

    \genbi= \gensp \otimes (\mathbf{k} \prec \oplus~ \mathbf{k} \succ),

    where we denote (\omega~\otimes \prec) (resp. (\omega~\otimes \succ)) by \begin{pmatrix} \omega \\ \prec \end{pmatrix} (resp. \begin{pmatrix} \omega \\ \succ \end{pmatrix}) for \omega\in \gensp. Then \genbas \times \{ \prec,\succ\} is a basis of \genbi. \genbi can be viewed as a graded module concentrated in arity 2.

  2. (ii) For a labeled n-tree \tau in \calt(\genbas), define \widetilde{\tau} in \mathcal{T}_{\hspace*{-0.1cm}ns}(\genbi) as follows:

    • \bullet \widetilde{|}=|

    • \bullet when n\geq 2, \widetilde{\tau} is obtained by replacing each decoration \gop \in \vin(\tau) by

      \begin{pmatrix} \gop \\ \ast \end{pmatrix}:=-\begin{pmatrix} \gop \\ \prec \end{pmatrix}-\begin{pmatrix} \gop \\ \succ \end{pmatrix}.

    We extend this definition to \mathcal{T}_{\hspace*{-0.1cm}ns}(\genbi) by linearity.

Definition

Definition 2.3

XDNU

Let \gensp be a vector space with a basis \genbas. Let \tau be a labeled n-tree in \calt(\genbas). The anti-bisuccessor \mathrm{ABSu}_{x}(\tau) of \tau with respect to a leaf x\in \lin(\tau) is an element of \mathcal{T}_{\hspace*{-0.1cm}ns}(\genbi) defined by induction on |\lin(\tau)| as follows:

  • \bullet \mathrm{ABSu}_{x}(~|~)=| ;

  • \bullet Let k\geq 0. Assume that \mathrm{ABSu}_{x}(\tau) have been defined for \tau with |\lin(\tau)|\leq k. Then for a labeled (k+1)-tree \tau\in \calt(\genbas) with its decomposition \tau_{l} \vee_{\gop} \tau_{r}, we define

    \mathrm{ABSu}_{x}(\tau)= \mathrm{ABSu}_{x}(\tau_{l} \vee_{\gop} \tau_{r}) = \left\{\begin{array}{ll} \mathrm{ABSu}_{x}(\tau_{l}) \vee_{\begin{pmatrix} \gop \\ \prec \end{pmatrix}} \widetilde{\tau_{r}},\quad & \text{if}~x \in \lin(\tau_{l}); \\ \widetilde{\tau_{l}} \vee_{\begin{pmatrix} \gop \\ \succ \end{pmatrix}} \mathrm{ABSu}_{x}(\tau_{r}), \quad & \text{if}~x \in \lin(\tau_{r}). \end{array}\right.
Proposition

Proposition 2.4

8ZPA

Let \gensp be a vector space with a basis \genbas, \tau be a labeled n-tree in \calt(\genbas) and x be in \lin(\tau). The anti-bisuccessor \mathrm{ABSu}_{x}(\tau) of \tau is obtained by relabeling each vertex of \tau according to the following rules:

  1. (a) we replace the label \gop of each vertex on the path from the root to the leaf x of \tau by

    • (i) \begin{pmatrix} \gop \\ \prec \end{pmatrix}, if the path turns left at this vertex;

    • (ii) \begin{pmatrix} \gop \\ \succ \end{pmatrix}, if the path turns right at this vertex;

  2. (b) we replace the label \gop of each vertex not on the path from the root to the leaf x of \tau by

    \begin{pmatrix} \gop \\ \ast \end{pmatrix}:=-\begin{pmatrix} \gop \\ \prec \end{pmatrix}-\begin{pmatrix} \gop \\ \succ \end{pmatrix}.
Proof of Proposition 2.4

We can prove it by induction on |\lin(\tau)|\geq 1. We omit the details for brevity.

Example

Example 2.5

BB8L
\mathrm{ABSu}_{x_{2}} \left( \begin{xy} (0,8)*+{x_{1}}="v1";(10,8)*+{x_{2}}="v2";(20,8)*+{x_{3}}="v3"; (30,8)*+{x_{4}}="v4"; (5,0)*+{\omega_{1}}="v5";(25,0)*+{\omega_{2}}="v6"; (15,-6)*+{\omega_{3}}="v7";(15,-12)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy} \right)= \begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\succ}}="v5";(25,0)* +{\svec{\omega_{2}}{\ast}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\prec}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{->}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{->}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{->}"v8";"v7"} \end{xy}
=-\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\succ}}="v5";(25,0)* +{\svec{\omega_{2}}{\prec}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\prec}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy}-\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\succ}}="v5";(25,0)* +{\svec{\omega_{2}}{\succ}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\prec}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy}
Lemma

Lemma 2.6

OR6A

Let \gensp be a vector space with a basis \genbas, \tau be a labeled n-tree in \calt(\genbas) and x be in \lin(\tau). Then the following relation holds

\mathrm{ABSu}_{\sigma^{-1}(x)}(\tau^{\sigma})=\mathrm{ABSu}_{x}(\tau) ^{\sigma},\quad \forall \sigma \in \mathbb{S}_n.
Proof of Lemma 2.6

It follows from inspecting the action of the symmetric group on a tree.

Definition

Definition 2.7

GJOQ

Let \gensp be a vector space with basis \genbas.

  1. (i) Define a vector space

    \begin{aligned} \gentri= \gensp \otimes (\mathbf{k} \prec \oplus~ \mathbf{k}\succ\oplus~\mathbf{k}~\cdot~), \end{aligned}

    where we denote (\omega~\otimes \prec) (resp. (\omega~\otimes \succ), resp. (\omega~\otimes\cdot)) by \begin{pmatrix} \omega \\ \prec \end{pmatrix} (resp. \begin{pmatrix} \omega \\ \succ \end{pmatrix}, resp. \begin{pmatrix} \omega \\ \cdot \end{pmatrix} ) for \omega\in \gensp. Then \genbas \times \{ \prec,\succ,\cdot \} is a basis of \gentri. \gentri can be viewed as an arity graded vector space concentrated in arity 2.

  2. (ii) For a labeled n-tree \tau in \calt(\genbas), define \widehat{\tau} in \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentri) as follows:

    • \bullet \widehat{|}=|

    • \bullet when n\geq 2, \widehat{\tau} is obtained by replacing each decoration \gop \in \vin(\tau) by

      \begin{pmatrix} \gop \\ \star \end{pmatrix}:=-\begin{pmatrix} \gop \\ \prec \end{pmatrix}-\begin{pmatrix} \gop \\ \succ \end{pmatrix}-\begin{pmatrix} \gop \\ \cdot \end{pmatrix}.

    We extend this definition to \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentri) by linearity.

Definition

Definition 2.8

LVAS

Let \gensp be a vector space with a basis \genbas. Let \tau be a labeled n-tree in \calt(\genbas) and J be a nonempty subset of \lin(\tau). The anti-trisuccessor \mathrm{ATSu}_{J}(\tau) of \tau with respect to J is an element of \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentri) defined by induction on |\lin(\tau)| as follows:

  • \bullet \mathrm{ATSu}_{J}(~|~)=| ;

  • \bullet Assume that \mathrm{ATSu}_{J}(\tau) have been defined for \tau with |\lin(\tau)|\leq k. Then for a labeled (k+1)-tree \tau\in \calt(\genbas) with its decomposition \tau_{l} \vee_{\gop} \tau_{r}, we define

    \mathrm{ATSu}_{J}(\tau)= \mathrm{ATSu}_{J}(\tau_{l} \vee_{\gop} \tau_{r}) = \left\{\begin{array}{lll} \mathrm{ATSu}_{J}(\tau_{l}) \vee_{\begin{pmatrix} \gop \\ \prec \end{pmatrix}} \widehat{\tau_{r}},\quad & \text{if}~J \subseteq \lin(\tau_{l}), \\ \widehat{\tau_{l}} \vee_{\begin{pmatrix} \gop \\ \succ \end{pmatrix}} \mathrm{ATSu}_{J}(\tau_{r}), \quad & \text{if}~J \subseteq \lin(\tau_{r}), \\ \mathrm{ATSu}_{J\cap \lin(\tau_{l})}(\tau_{l}) \vee_{\begin{pmatrix} \gop \\ \cdot \end{pmatrix}} \mathrm{ATSu}_{J\cap \lin(\tau_{r})}(\tau_{r}), \quad & {\mathrm{otherwise.}} \end{array}\right.
Proposition

Proposition 2.9

0S8T

Let \gensp be a vector space with a basis \genbas, \tau be a labeled n-tree in \calt(\genbas) and J be a nonempty subset of \lin(\tau). The anti-trisuccessor \mathrm{ATSu}_{J}(\tau) is obtained by relabeling each vertex of \tau according to the following rules:

  1. (a) We replace the label \gop of each vertex on at least one of the paths from the root to the leaves in J by

    • (i) \begin{pmatrix} \gop \\ \prec \end{pmatrix}, if all such paths turn left at this vertex;

    • (ii) \begin{pmatrix} \gop \\ \succ \end{pmatrix}, if all such paths turn right at this vertex;

    • (iii) \begin{pmatrix} \gop \\ \cdot \end{pmatrix}, if some of such paths turn left and some of such paths turn right at this vertex;

  2. (b) We replace the label \gop of each other vertex by

    \begin{pmatrix} \gop \\ \star \end{pmatrix}:=-\begin{pmatrix} \gop \\ \prec \end{pmatrix}-\begin{pmatrix} \gop \\ \succ \end{pmatrix}-\begin{pmatrix} \gop \\ \cdot \end{pmatrix}.
Example

Example 2.10

3KEQ
\mathrm{ATSu}_{\{x_{1},x_{4}\}} \left( \begin{xy} (0,8)*+{x_{1}}="v1";(10,8)*+{x_{2}}="v2";(20,8)*+{x_{3}}="v3"; (30,8)*+{x_{4}}="v4"; (5,0)*+{\omega_{1}}="v5";(25,0)*+{\omega_{2}}="v6"; (15,-6)*+{\omega_{3}}="v7";(15,-12)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy} \right)= \begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\prec}}="v5";(25,0)* +{\svec{\omega_{2}}{\succ}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\cdot}}="v7";(15,-15)*+{}="v8"; {\ar@{->}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{->}"v6";"v4"};{\ar@{->}"v7";"v5"};{\ar@{->}"v7";"v6"};{\ar@{->}"v8";"v7"} \end{xy}
Lemma

Lemma 2.11

4WFA

Let \gensp be a vector space with a basis \genbas, \tau be a labeled n-tree in \calt(\genbas) and J be a nonempty subset of \lin(\tau). Then the following relation holds

\mathrm{ATSu}_{\sigma^{-1}(J)}(\tau^{\sigma})=\mathrm{ATSu}_{J}(\tau) ^{\sigma},\quad \forall \sigma \in \mathbb{S}_n.
Proof of Lemma 2.11

By inspection of the action of the symmetric group on a tree.

Definition

Definition 2.12

FOCX
  1. (a) An \mathbb{S}-module over \mathbf{k} is a family

    M:=\{ M(n)\}_{n \geq 0}=\{M(0),M(1),\dots,M(n),\dots\}

    of right \mathbf{k} \mathbb{S}_{n}-modules M(n), where \mathbb{S}_{n} is the symmetric group for all n \geq 0. If every M(n) is the trivial \mathbf{k} \mathbb{S}_{n}-module, then M is called an \mathbb{N}-module.

  2. (b) A nonsymmetric operad is an \mathbb{N}-module \opd=\{ \opd(n)\}_{n \geq 0} equipped with partial compositions:

    \circ_{i}:=\circ_{m,n,i}: \opd(m) \otimes \opd(n) \to \opd(m+n-1),\quad 1\leq i\leq m

    such that the following equalities hold for all \eta\in \opd(l),\mu \in \opd(m) and \nu\in \opd(n):

    1. (i) (\eta \circ_i \mu) \circ_{i-1+j} \nu= \eta \circ_i (\mu \circ_j \nu), \quad 1\leq i\leq l,1\leq j\leq m.

    2. (ii) (\eta \circ_i \mu) \circ_{k-1+m} \nu= (\eta \circ_k \nu)\circ_i \mu,\quad 1 \leq i <k \leq l.

    3. (iii) There is an element {\rm{Id}} \in \opd(1) such that \mu \circ {\rm{Id}}=\mu and {\rm{Id}} \circ \mu=\mu.

  3. (c) A nonsymmetric operad \opd=\{ \opd(n)\}_{n \geq 0} is called a (symmetric) operad, if \opd=\{ \opd(n)\}_{n \geq 0} is an \mathbb{S}-module and satisfies the following additional conditions

    1. (iv) For any \sigma \in \mathbb{S}_{m}, we have

      \mu^{\sigma} \circ_i \nu=(\mu \circ_{\sigma(i)} \nu)^{\sigma'},\quad 1 \leq i\leq m,

      where \sigma'\in \mathbb{S}_{m+n-1} restricts to the bijection from \{1,\dots,m+n-1\} \backslash \{i,\dots,i+n-1\} to \{1,\dots,m+n-1\} \backslash \{\sigma(1),\dots,\sigma(i)+n-1\} according to \sigma and restricts to the bijection from \{i,\dots,i+m-1\} to \{\sigma(i),\dots,\sigma(i)+m-1\} according to the identity (in the natural order).

    2. (v) For any \tau \in \mathbb{S}_{n}, we have

      \mu \circ_i \nu^{\tau}=(\mu \circ_{i} \nu)^{\tau'},\quad 1 \leq i\leq m,

      where \tau'\in \mathbb{S}_{m+n-1} is the permutation which acts by the identity, except on the block \{i,\dots,i+n-1\} on which it acts via \tau.

Definition

Definition 2.13

HP9U

Let \opd=\mathcal{T}(\gensp)/ (\relsp) be a binary (symmetric) operad on the \mathbb{S}-module \gensp=\gensp(2) concentrated in arity 2 with a \mathbf{k}[\mathbb{S}_2]-basis \genbas, such that the space of relations \relsp is spanned as an \mathbb{S}-module, by locally homogeneous elements of the form

R:=\Biggl\{ r_s:=\sum_{i} c_{s,i} \tau_{s,i}~|~ c_{s,i} \in \mathbf{k}, \tau_{s,i} \in \{t(\genbas),t\in \mathfrak{R}\}, 1 \leq s\leq k, k\geq 1 \Biggr\},

where \mathfrak{R} is a set of representatives of \calt/\thicksim defined as above. For each fixed s, all trees \tau_{s,i} have the same set of leaves \lin(\tau_{s,i}).

  1. (i) The anti-bisuccessor of \opd is defined to be the binary operad

    \mathrm{ABSu}(\opd):=\mathcal{T}(\genbi)/ (\mathrm{ABSu}(\relsp),\st(\relsp)),

    where the \mathbb{S}_2-action on \genbi is given by

    \begin{pmatrix} \gop \\ \prec \end{pmatrix}^{(12)}:=\begin{pmatrix} \gop^{(12)} \\ \succ \end{pmatrix},\quad \begin{pmatrix} \gop \\ \succ \end{pmatrix}^{(12)}:=\begin{pmatrix} \gop^{(12)} \\ \prec \end{pmatrix}, \quad \forall \gop\in \gensp.

    The space of relations is generated as an \mathbb{S}-module, given by

    \mathrm{ABSu}(\relsp):=\Biggl\{ \mathrm{ABSu}_x(r_s)=\sum_{i} c_{s,i} \mathrm{ABSu}_x(\tau_{s,i})~|~ x\in \lin(\tau_{s,i}), 1 \leq s \leq k \Biggr\}

    and

    \st(\relsp):=\Biggl\{ \sum_{i} c_{s,i} \st(\tau_{s,i})=\sum_{i} c_{s,i} \widetilde{\tau}_{s,i}~|~1 \leq s \leq k \Biggr\}.

    Here \st(\tau_{s,i})=\widetilde{\tau}_{s,i} is obtained by replacing each vertex decoration \gop of \tau_{s,i} by

    \begin{pmatrix} \gop \\ \ast \end{pmatrix}:=-\begin{pmatrix} \gop \\ \prec \end{pmatrix}-\begin{pmatrix} \gop \\ \succ \end{pmatrix}.
  2. (ii) The anti-trisuccessor of \opd is defined to be the binary operad

    \mathrm{ATSu}(\opd):=\mathcal{T}(\gentri)/ (\mathrm{ATSu}(\relsp),\ST(\relsp)),

    where the \mathbb{S}_2-action on \gentri is given by

    \begin{pmatrix} \gop \\ \prec \end{pmatrix}^{(12)}:=\begin{pmatrix} \gop^{(12)} \\ \succ \end{pmatrix},\quad \begin{pmatrix} \gop \\ \succ \end{pmatrix}^{(12)}:=\begin{pmatrix} \gop^{(12)} \\ \prec \end{pmatrix}, \quad \begin{pmatrix} \gop \\ \cdot \end{pmatrix}^{(12)}:=\begin{pmatrix} \gop^{(12)} \\ \cdot \end{pmatrix},\quad \forall \gop\in \gensp.

    The space of relations is generated as an \mathbb{S}-module, given by

    \mathrm{ATSu}(\relsp):=\Biggl\{ \mathrm{ATSu}_{J}(r_s)=\sum_{i} c_{s,i} \mathrm{ATSu}_{J}(\tau_{s,i})~| \varnothing \neq J \subseteq \lin(\tau_{s,i}), 1 \leq s \leq k \Biggr\}

    and

    \ST(\relsp):=\Biggl\{ \sum_{i} c_{s,i} \ST(\tau_{s,i})=\sum_{i} c_{s,i} \widetilde{\tau}_{s,i}~|~1 \leq s \leq k \Biggr\}.

    Here \ST(\tau_{s,i})=\widehat{\tau}_{s,i} is obtained by replacing each vertex decoration \gop of \tau_{s,i} by

    \begin{pmatrix} \gop \\ \star \end{pmatrix}:=-\begin{pmatrix} \gop \\ \prec \end{pmatrix}-\begin{pmatrix} \gop \\ \succ \end{pmatrix}-\begin{pmatrix} \gop \\ \cdot \end{pmatrix}.
Remark

Remark 2.14

SFF0

The symbols \star and \ast record the signed total operations underlying the anti-successor construction. In contrast with an ordinary splitting, the original operation will be recovered from the split operations through the negative total sum. Theorem 2.16 below makes this precise by showing that these signed total operations satisfy the defining relations of the original operad.

Remark

Remark 2.15

EMGI

The above definitions also apply to binary nonsymmetric operads by omitting the symmetric group actions and replacing the free symmetric operad \mathcal{T}(\gensp) with the free nonsymmetric operad \mathcal{T}_{\hspace*{-0.1cm}ns}(\gensp). More precisely, let

\opd=\mathcal{T}_{\hspace*{-0.1cm}ns}(\gensp)/(\relsp),

where \gensp=\gensp(2) is a vector space with a linear basis \genbas, and \relsp is spanned by locally homogeneous elements of the form

r_s=\sum_i c_{s,i}\tau_{s,i},\qquad c_{s,i}\in\mathbf{k},\quad \tau_{s,i}\in\calt(\genbas), \quad 1\leq s\leq k.

For each fixed s, all trees \tau_{s,i} have the same arity n_s. The anti-bisuccessor and anti-trisuccessor are then defined by the same formulas:

\mathrm{ABSu}(\opd):= \mathcal{T}_{\hspace*{-0.1cm}ns}(\genbi)/ (\mathrm{ABSu}(\relsp),\st(\relsp)), \qquad \mathrm{ATSu}(\opd):= \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentri)/ (\mathrm{ATSu}(\relsp),\ST(\relsp)),

where the relations are understood as nonsymmetric operadic relations. In this setting, both constructions are independent of the choice of a basis of the vector space of generating operations, by linearity of the anti-successor constructions and the treewise tensor structure of the free nonsymmetric operad.

Theorem

Theorem 2.16

E6HJ

Let \opd=\mathcal{T}(\gensp)/ (\relsp) be a binary operad.

  1. (i) There is a morphism of operads from \opd to \mathrm{ABSu}(\opd) which extends the linear map from \gensp to \widetilde{V} defined by

    \omega \to \begin{pmatrix} \omega\\ \ast \end{pmatrix},\quad \forall \omega\in \gensp.
  2. (ii) There is a morphism of operads from \opd to \mathrm{ATSu}(\opd) which extends the linear map from \gensp to \widetilde{V} defined by

    \omega \to \begin{pmatrix} \omega\\ \star \end{pmatrix},\quad \forall \omega\in \gensp.
  3. (iii) There is a morphism of operads from \opd to \mathrm{ATSu}(\opd) which extends the linear map from \gensp to \widetilde{V} defined by

    \omega \to \begin{pmatrix} \omega\\ \cdot \end{pmatrix},\quad \forall \omega\in \gensp.
Used by Remark 2.18
Proof of Theorem 2.16

Assume that R is given by Eq. (2.13).

(i) It is easy to check that the linear map defined in Eq. (?) is \mathbb{S}_2-equivariant so it induces a morphism of operads from \mathcal{T}(\gensp) to \mathrm{ABSu}(\opd). Since \sum_i c_{s,i}\widetilde{\tau}_{s,i}\in \st(R), thus there exists a morphism of operads from \opd to \mathrm{ABSu}(\opd).

(ii) Since we have \sum_i c_{s,i}\widehat{\tau}_{s,i}\in \ST(R), the other part of proof is similar to the proof of (i).

(iii) it is straightforward to check that the linear map defined in Eq. (?) is \mathbb{S}_2-equivariant so it induces a morphism of operads from \mathcal{T}(\gensp) to \mathrm{ATSu}(\opd). Moreover, by the definition of an anti-trisuccessor, the following equation holds:

\sum_i c_{s,i}\mathrm{ATSu}_{\lin(\tau_{s,i})} (\tau_{s,i})=0. \quad 1\leq s\leq k.

Note that the labeled tree \mathrm{ATSu}_{\lin(\tau_{s,i})} (\tau_{s,i}) is obtained by replacing the label of each vertex \omega of \tau_{s,i} to \begin{pmatrix} \omega\\ \cdot \end{pmatrix}. Thus the conclusion holds.

Remark

Remark 2.17

G0BT

Note that a morphism of operads f:\opd\to\oqd induces a functor from the category of \oqd-algebras to the category of \opd-algebras. More precisely, if (V,\rho:\oqd\to \mathrm{End}_V) is a \oqd-algebra, then the composite

\opd \xrightarrow{f} \oqd \xrightarrow{\rho} \mathrm{End}_V

makes V into a \opd-algebra.

Remark

Remark 2.18

3TVX

Theorem 2.16 gives an intrinsic explanation of the term anti-splitting. Indeed, the morphism \opd\to\mathrm{ABSu}(\opd) sends each generator \omega to \omega_\star=-(\omega_\prec+\omega_\succ). Hence every \mathrm{ABSu}(\opd)-algebra canonically becomes a \opd-algebra under the negative total operation. Likewise, every \mathrm{ATSu}(\opd)-algebra carries the original \opd-structure through \omega_\star=-(\omega_\prec+\omega_\succ+\omega_\cdot). Thus the construction is a genuine splitting procedure, with the prefix “anti-” reflected precisely in the negative reconstruction of the original operation.

Proposition

Proposition 2.19

DLFE

The operad \antiden encoding anti-dendriform algebras is the anti-bisuccessor of the nonsymmetric operad \ass encoding associative algebras.

Proof of Proposition 2.19

Let \ass be the nonsymmetric operad of the associative algebra with product \cdot. Using the abbreviations \rhd:=\begin{pmatrix} \cdot\\ \succ \end{pmatrix}, \lhd:=\begin{pmatrix} \cdot\\ \prec \end{pmatrix} and \bullet:=\rhd + \lhd, by a direct calculation, we have

\begin{aligned} \mathrm{ABSu}_{x}((x \cdot y)\cdot z- x\cdot (y\cdot z)) &= (x\lhd y)\lhd z+ x\lhd (y \bullet z);\\ \mathrm{ABSu}_{y}((x \cdot y)\cdot z- x\cdot (y\cdot z)) &= (x\rhd y)\lhd z- x\rhd (y \lhd z);\\ \mathrm{ABSu}_{z}((x \cdot y)\cdot z- x\cdot (y\cdot z)) &= -(x \bullet y)\rhd z- x\rhd (y \rhd z);\\ \st((x \cdot y)\cdot z- x\cdot (y\cdot z)) &= (x \bullet y)\bullet z-x\bullet (y \bullet z), \end{aligned}

giving the same relations (?)-(?) of anti-dendriform algebras introduced in [GLB]. Therefore, the operad \antiden is the anti-bisuccessor of the operad \ass.

Proposition

Proposition 2.20

MGF1

The operad \antiprelie encoding anti-pre-Lie algebras is the anti-bisuccessor of the operad \lie encoding Lie algebras.

Uses Definition 2.13 · Used by Proof
Proof of Proposition 2.20

Let \mu be the generating operation of the operad \lie. Set \circ:=\begin{pmatrix} \mu\\ \succ \end{pmatrix}. Since \begin{pmatrix} \mu\\ \prec \end{pmatrix}^{(12)}=\begin{pmatrix} \mu^{(12)}\\ \succ \end{pmatrix}=-\begin{pmatrix} \mu\\ \succ \end{pmatrix}=-~\circ, we have \begin{pmatrix} \mu\\ \prec \end{pmatrix}=-\circ^{(12)}. The space of the relations of \lie is generated as an \mathbb{S}_3-module by

v_1+v_5+v_9= \mu \circ_{\rmi} \mu+\mu \circ_{\rmii} \mu+\mu \circ_{\rmiii} \mu.

Then we have

\begin{aligned} \mathrm{ABSu}_{x}(v_1+v_5+v_9)&= z \circ (y \circ x)-[y,z]_{\circ} \circ x-y \circ (z \circ x);\\ \mathrm{ABSu}_{y}(v_1+v_5+v_9)&= x \circ (z \circ y)-[z,x]_{\circ} \circ y-z \circ (x \circ y);\\ \mathrm{ABSu}_{z}(v_1+v_5+v_9)&= y \circ (x \circ z)-[x,y]_{\circ} \circ z-x \circ (y \circ z);\\ \st(v_1+v_5+v_9)&= [[x,y]_{\circ},z]_{\circ}+[[y,z]_{\circ},x]_{\circ}+[[z,x]_{\circ},y]_{\circ}, \end{aligned}

which corresponds to the relations (?) and (?) of anti-pre-Lie algebras. Therefore, \mathrm{ABSu}(\lie)=\antiprelie.

Definition

Definition 2.21

ZWMF

([LB2023]) An anti-Zinbiel algebra (A,\diamond) is a vector space A equipped with a bilinear operation \diamond:A \otimes A \to A satisfying the following equation

x \diamond (y \diamond z)=-(x \diamond y+ y\diamond x)\diamond z=z \diamond (y \diamond x),\quad \forall x,y,z\in A.
Definition

Definition 2.22

7S15

([LB2023]) An anti-pre-Poisson algebra (A,\diamond,\circ) is a vector space A equipped with two bilinear operations \diamond,\circ:A \otimes A \to A such that (A,\diamond) is an anti-Zinbiel algebra and (A,\circ) is an anti-pre-Lie algebra and for all x,y,z\in A, the following conditions hold:

\begin{aligned} (y \diamond z+z \diamond y)\circ x+z \diamond (y \circ x)+y \diamond (z \circ x)&=0;\\ (x \circ z- z\circ x)\diamond y+x \circ (z \diamond y)- z \diamond (x \circ y)&=0;\\ (y \diamond z+z \diamond y)\circ x+(x \circ z- z\circ x)\diamond y+(x \circ y- y\circ x)\diamond z&=0. \end{aligned}
Uses Definition 2.21 · Used by Proof
Proposition

Proposition 2.23

OBF0
  • (i) The operad \antizinbiel is the anti-bisuccessor of the operad \com;

  • (ii) The operad \antiprepoisson is the anti-bisuccessor of the operad \poisson.

Uses Definition 2.13 · Used by Proof
Proof of Proposition 2.23

We only prove (ii) and (i) can be proved similarly. Recall that the operad \poisson:=\mathcal{T}(V)/(R_{Poisson}) is a binary operad on the \mathbb{S}-module V=V(2) concentrated in arity 2 with a \mathbf{k}[\mathbb{S}_2]-basis \mathcal{V}:=\{\omega,\mu\}, such that the space of relations R_{Poisson} is generated as an \mathbb{S}_3-module by

\left\{ \begin{array}{l} \omega \circ_{\rmi} \omega-\omega \circ_{\rmii} \omega;\\ \mu \circ_{\rmi} \mu+\mu \circ_{\rmii} \mu+\mu \circ_{\rmiii} \mu;\\ \mu' \circ_{\rmii} \omega-\omega \circ_{\rmi} \mu-\omega' \circ_{\rmiii} \mu'. \end{array} \right.

By a direct calculation, we have

\begin{aligned} \mathrm{ABSu}_{x}(\mu' \circ_{\rmii} \omega-\omega \circ_{\rmi} \mu-\omega' \circ_{\rmiii} \mu')&= (y \diamond z+z \diamond y)\circ x+z \diamond (y \circ x)+y \diamond (z \circ x);\\ \mathrm{ABSu}_{y}(\mu' \circ_{\rmii} \omega-\omega \circ_{\rmi} \mu-\omega' \circ_{\rmiii} \mu')&= x \circ (z \diamond y)- z \diamond (x \circ y)+(x \circ z- z\circ x)\diamond y;\\ \mathrm{ABSu}_{z}(\mu' \circ_{\rmii} \omega-\omega \circ_{\rmi} \mu-\omega' \circ_{\rmiii} \mu')&= x \circ (y \diamond z)+(x \circ y- y\circ x)\diamond z- y \diamond (x \circ z);\\ \st(\mu' \circ_{\rmii} \omega-\omega \circ_{\rmi} \mu-\omega' \circ_{\rmiii} \mu') &= \{x, y\cdot z\}-\{x, y\}\cdot z-y \cdot \{x, z\}, \end{aligned}

where \{x, y\}:= x \circ y-y \circ x and x\cdot y:=x \diamond y+y \diamond x. It is easy to check that they correspond to the relations (2.22)-(2.22) in Definition 2.22. Therefore, we have \mathrm{ABSu}(\poisson)=\antiprepoisson.

Definition

Definition 2.24

90VR

An anti-tridendriform algebra (A,<,>,\cdot) is a vector space A equipped with an associative product \cdot and two bilinear operations <,>:A \otimes A \to A satisfying the following equations

\begin{aligned} &&(x < y)< z=-x < (y \star z), \quad (x > y)< z=x > (y < z),\quad (x > y)> z=-(x \star y)> z,\\ && (x \cdot y)< z=x \cdot (y < z), \quad (x > y)\cdot z=x > (y \cdot z), \quad (x < y)\cdot z=x \cdot (y > z),\\ &&(x \star y)\star z=x \star (y \star z), \end{aligned}

for all x,y,z\in A, where the binary operation \star:A \otimes A \to A is given by x \star y=x\cdot y+x > y+x < y.

Remark

Remark 2.25

1IL2

Let (A,<,>,\cdot) be an anti-tridendriform algebra. If the associative algebra (A,\cdot) is trivial, then (A,<,>) becomes an anti-dendriform algebra. Therefore, an anti-tridendriform algebra can be viewed as a generalization of an anti-dendriform algebra.

Proposition

Proposition 2.26

0BDF

Let (A,<,>,\cdot) be an anti-tridendriform algebra. Then we have

  • (i) (A,\star) is an associative algebra, which is called the sub-adjacent associative algebra of (A,<,>,\cdot) and denoted by A^c, where the multiplication \star:A \otimes A \to A is given by

    x \star y=x\cdot y+x > y+x < y, \quad \forall x,y \in A.
  • (ii) (-L_>,-R_<):A \to \mathrm{End}(A) gives an action of the associative algebra A^c on A, where the left and right multiplication operators L_>,R_<:A \to \mathrm{End}(A) are defined by L_>(x)(y)=x > y, R_<(x)(y)=y < x respectively, for all x,y\in A.

Proposition

Proposition 2.27

Z4YF

The operad \antitriden is the anti-trisuccessor of the operad \ass.

Proof of Proposition 2.27

Let \nu be the generating operation of the operad \ass satisfying the relation v_1-v_2=\nu \circ_{\rmi} \nu-\nu' \circ_{\rmii} \nu. Using the abbreviations >:=\begin{pmatrix} \nu\\ \succ \end{pmatrix}, <:=\begin{pmatrix} \nu\\ \prec \end{pmatrix}, \cdot:=\begin{pmatrix} \nu\\ \cdot \end{pmatrix} and \star:=>+<+~\cdot, by a direct calculation, we have

\begin{aligned} \mathrm{ATSu}_{\{x\}}(v_1-v_2)&= (x < y)< z+x < (y \star z);\\ \mathrm{ATSu}_{\{y\}}(v_1-v_2)&= (x > y)< z-x > (y < z);\\ \mathrm{ATSu}_{\{z\}}(v_1-v_2)&= -(x > y)> z-(x \star y)> z;\\ \mathrm{ATSu}_{\{x,y\}}(v_1-v_2)&= (x \cdot y)< z-x \cdot (y < z);\\ \mathrm{ATSu}_{\{x,z\}}(v_1-v_2)&= (x > y)\cdot z-x > (y \cdot z);\\ \mathrm{ATSu}_{\{y,z\}}(v_1-v_2)&= (x < y)\cdot z-x \cdot (y > z);\\ \mathrm{ATSu}_{\{x,y,z\}}(v_1-v_2)&= (x \cdot y)\cdot z-x \cdot (y \cdot z);\\ \ST(v_1-v_2)&= (x \star y)\star z-x \star (y \star z), \end{aligned}

which corresponds to the relation of anti-tridendriform algebras. Thus, the operad \antitriden is the anti-trisuccessor of the operad \ass.

Definition

Definition 2.28

76BC

An anti-post-Lie algebra (\mathfrak g,[\cdot,\cdot]_\mathfrak g,\triangleright) consists of a Lie algebra (\mathfrak g,[\cdot,\cdot]_\mathfrak g) and a binary operation \triangleright:\mathfrak g \otimes \mathfrak g \to \mathfrak g such that

\begin{aligned} &&x\triangleright [y,z]_\mathfrak g=[x\triangleright y,z]_\mathfrak g+[y,x\triangleright z]_\mathfrak g,\\ &&([x,y]_\mathfrak g+x\triangleright y-y\triangleright x) \triangleright z=-x\triangleright(y\triangleright z)+y\triangleright(x\triangleright z),\\ &&([x,y]_\mathfrak g+x\triangleright y-y\triangleright x) \triangleright z+([y,z]_\mathfrak g+y\triangleright z-z\triangleright y) \triangleright x+([z,x]_\mathfrak g+z\triangleright x-x\triangleright z) \triangleright y=0, \end{aligned}

for all x,y,z\in \mathfrak g.

Remark

Remark 2.29

EGGR

Let (\mathfrak g,[\cdot,\cdot]_\mathfrak g,\triangleright) be an anti-post-Lie algebra. If the Lie bracket (\mathfrak g,[\cdot,\cdot]_\mathfrak g) vanishes, then (\mathfrak g,\triangleright) degenerates to an anti-pre-Lie algebra. Thus, an anti-post-Lie algebra can be viewed as a nonabelian generalization of an anti-pre-Lie algebra.

Proposition

Proposition 2.30

NILU

Any anti-post-Lie algebra (\mathfrak g,[\cdot,\cdot]_\mathfrak g,\triangleright) gives rise to a new Lie algebra (\mathfrak g,[\cdot,\cdot]_\triangleright) given by

\begin{aligned} [x,y]_\triangleright:=x \triangleright y-y \triangleright x+[x,y]_\mathfrak g,\quad \forall x,y\in \mathfrak g, \end{aligned}

which is called the sub-adjacent Lie algebra of the anti-post-Lie algebra (\mathfrak g,[\cdot,\cdot]_\mathfrak g,\triangleright) and denoted by \mathfrak g_{\triangleright}. Moreover, -L_{\triangleright}:\mathfrak g \to \mathrm{Der}(\mathfrak g) defined by -L_{\triangleright}(x)(y)=-x \triangleright y is an action of \mathfrak g_{\triangleright} on the Lie algebra (\mathfrak g,[\cdot,\cdot]_\mathfrak g).

Proposition

Proposition 2.31

06I5

The operad \antipostlie encoding anti-post-Lie algebras is the anti-trisuccessor of the operad \lie.

Proof of Proposition 2.31

Let \mu be the generating operation of the operad \lie. Set \triangleright:=\begin{pmatrix} \mu\\ \succ \end{pmatrix} and [\cdot,\cdot]:=\begin{pmatrix} \mu\\ \cdot \end{pmatrix}. Since \begin{pmatrix} \mu\\ \prec \end{pmatrix}^{(12)}=\begin{pmatrix} \mu^{(12)}\\ \succ \end{pmatrix}=-\begin{pmatrix} \mu\\ \succ \end{pmatrix}=-\triangleright, we have \begin{pmatrix} \mu\\ \prec \end{pmatrix}=-\triangleright^{(12)}. The space of the relations of \lie is generated as an \mathbb{S}_3-module by

v_1+v_5+v_9= \mu \circ_{\rmi} \mu+\mu \circ_{\rmii} \mu+\mu \circ_{\rmiii} \mu.

Then we have

\begin{aligned} \mathrm{ATSu}_{\{x\}}(v_1+v_5+v_9)&= z \triangleright (y \triangleright x)-y \triangleright (z \triangleright x)+(z\triangleright y-y \triangleright z-[y,z])\triangleright x;\\ \mathrm{ATSu}_{\{y\}}(v_1+v_5+v_9)&= x \triangleright (z \triangleright y)-z \triangleright (x \triangleright y)+(x\triangleright z-z \triangleright x-[z,x])\triangleright y;\\ \mathrm{ATSu}_{\{z\}}(v_1+v_5+v_9)&= y \triangleright (x \triangleright z)-x \triangleright (y \triangleright z)+(y\triangleright x-x \triangleright y-[x,y])\triangleright z;\\ \mathrm{ATSu}_{\{x,y\}}(v_1+v_5+v_9)&= -z \triangleright [x,y]-[z\triangleright y,x]+[z \triangleright x,y];\\ \mathrm{ATSu}_{\{x,z\}}(v_1+v_5+v_9)&= -y \triangleright [z,x]-[y\triangleright x,z]+[y \triangleright z,x];\\ \mathrm{ATSu}_{\{y,z\}}(v_1+v_5+v_9)&= -x \triangleright [y,z]-[x\triangleright z,y]+[x \triangleright y,z];\\ \mathrm{ATSu}_{\{x,y,z\}}(v_1+v_5+v_9)&= [[x,y],z]+ [[y,z],x]+[[z,x],y];\\ \ST(v_1+v_5+v_9)&= [[x,y]_\triangleright,z]_\triangleright+ [[y,z]_\triangleright,x]_\triangleright+[[z,x]_\triangleright,y]_\triangleright, \end{aligned}

which corresponds to the relations (2.28)-(2.28) of anti-post-Lie algebras. Therefore, \mathrm{ATSu}(\lie)=\antipostlie.

Definition

Definition 2.32

TARB

An anti-post-com algebra (A,\cdot,\rhd) consists of a commutative associative algebra (A,\cdot) and a binary operation \rhd:A \otimes A \to A such that

\begin{aligned} -(x \star y) \rhd z &= x \rhd (y \rhd z);\\ z \rhd (x \cdot y) &= x \cdot (z \rhd y);\\ x \rhd (y \rhd z) &= z \rhd (y \rhd x), \end{aligned}

where the bilinear product \star:A \otimes A \to A is given by x \star y:=x \rhd y+y \rhd x+x \cdot y, for all x,y,z\in A.

Remark

Remark 2.33

GFEH

Let (A,\cdot,\rhd) be an anti-post-com algebra. If the commutative associative algebra (A,\cdot) is trivial, then (A,\rhd) degenerates to an anti-Zinbiel algebra. Thus, an anti-post-com algebra can be viewed as a generalization of an anti-Zinbiel algebra.

Moreover, let (A,<,>,\cdot) be an anti-tridendriform algebra. If the associative algebra (A,\cdot) is commutative and two operations <,> satisfy x < y=y > x, then (A,\cdot,>) is an anti-post-com algebra.

Proposition

Proposition 2.34

7CH8

Any anti-post-com algebra (A,\cdot,\rhd) gives rise to a new commutative associative algebra (A,\star) given by

\begin{aligned} x \star y:=x \rhd y+y \rhd x+x \cdot y, \quad \forall x,y\in A, \end{aligned}

which is called the sub-adjacent commutative (associative) algebra of the anti-post-com algebra (A,\cdot,\rhd) and denoted by A_{\rhd}. Moreover, -L_{\rhd}:A \to \mathrm{End}(A) defined by -L_{\rhd}(x)(y)=-x \rhd y is an action of A_{\rhd} on the commutative associative algebra (A,\cdot).

Proposition

Proposition 2.35

EDZT

The operad \antipostcom is the anti-trisuccessor of the operad \com.

Proof of Proposition 2.35

Let \omega be the generating operation of the operad \com. Set \rhd:=\begin{pmatrix} \omega\\ \succ \end{pmatrix} and \cdot:=\begin{pmatrix} \omega\\ \cdot \end{pmatrix}. Since \begin{pmatrix} \omega\\ \prec \end{pmatrix}^{(12)}=\begin{pmatrix} \omega^{(12)}\\ \succ \end{pmatrix}=\begin{pmatrix} \omega\\ \succ \end{pmatrix}=\rhd, we have \begin{pmatrix} \omega\\ \prec \end{pmatrix}=\rhd^{(12)}. The space of relations of \com is generated as an \mathbb{S}_3-module by

v_1-v_9= \omega \circ_{\rmi} \omega-\omega \circ_{\rmii} \omega.

Then we have

\begin{aligned} \mathrm{ATSu}_{\{x\}}(v_1-v_9)&= z \rhd (y \rhd x)+(y \star z)\rhd x;\\ \mathrm{ATSu}_{\{y\}}(v_1-v_9)&= z \rhd (x \rhd y)-x \rhd (z \rhd y);\\ \mathrm{ATSu}_{\{z\}}(v_1-v_9)&= -(x \star y)\rhd z-x \rhd (y \rhd z);\\ \mathrm{ATSu}_{\{x,y\}}(v_1-v_9)&= z \rhd (x \cdot y)-(z \rhd y) \cdot x;\\ \mathrm{ATSu}_{\{x,z\}}(v_1-v_9)&= (y \rhd x)\cdot z-(y \rhd z)\cdot x;\\ \mathrm{ATSu}_{\{y,z\}}(v_1-v_9)&= (x \rhd y)\cdot z- x \rhd (y \cdot z);\\ \mathrm{ATSu}_{\{x,y,z\}}(v_1-v_9)&= (x \cdot y)\cdot z-(y \cdot z)\cdot x;\\ \ST(v_1-v_9)&= (x \star y)\star z-(y \star z)\star x, \end{aligned}

which corresponds to the relations Eqs. (2.32)-(2.32) of anti-post-com algebras. Thus, the operad \antipostcom is the anti-trisuccessor of the operad \com.

Definition

Definition 3.1

OX0X

([Ginzburg, Val2]) Let \mathcal{P}=\mathcal{T}(V)/(R) and \mathcal{Q}=\mathcal{T}(W)/(S) be two binary quadratic operads with finite-dimensional generating spaces. Define their Manin black product by the formula

\mathcal{P} \bullet \mathcal{Q}:= \mathcal{T}(V \otimes W \otimes \mathbf{k}. \mathrm{sgn}_{\mathbb{S}_2})/(\Psi(R \otimes S)),

where \Psi is defined in [Val2, Section 4.3].

Theorem

Theorem 3.2

I1DT

Let \opd be a binary quadratic operad. We have the isomorphism of operads

\mathrm{ATSu}(\opd) \cong \antipostlie \bullet \opd.
Used by Proof
Proof of Theorem 3.2

Denote the generating operations [\cdot,\cdot] and \triangleright of \antipostlie by \beta and \varepsilon respectively. Then we have \beta'=-\beta. The space of relations of \antipostlie is generated as a vector space by

\begin{aligned} &&\beta \circ_{\rmi} \beta +\beta \circ_{\rmii} \beta +\beta \circ_{\rmiii} \beta,\\ &&-\varepsilon' \circ_{\rmi} \beta+\beta \circ_{\rmiii} \varepsilon+\beta' \circ_{\rmii} \varepsilon',\\ &&-\varepsilon' \circ_{\rmii} \beta+\beta \circ_{\rmi} \varepsilon+\beta' \circ_{\rmiii} \varepsilon',\\ &&-\varepsilon' \circ_{\rmiii} \beta+\beta \circ_{\rmii} \varepsilon+\beta' \circ_{\rmi} \varepsilon',\\ &&-\varepsilon \circ_{\rmii} \beta-\varepsilon \circ_{\rmii} \varepsilon+\varepsilon \circ_{\rmii} \varepsilon'-\varepsilon'\circ_{\rmiii} \varepsilon+\varepsilon'\circ_{\rmi} \varepsilon',\\ &&-\varepsilon \circ_{\rmiii} \beta-\varepsilon \circ_{\rmiii} \varepsilon+\varepsilon \circ_{\rmiii} \varepsilon'-\varepsilon'\circ_{\rmi} \varepsilon+\varepsilon'\circ_{\rmii} \varepsilon',\\ &&-\varepsilon \circ_{\rmi} \beta-\varepsilon \circ_{\rmi} \varepsilon+\varepsilon \circ_{\rmi} \varepsilon'-\varepsilon'\circ_{\rmii} \varepsilon+\varepsilon'\circ_{\rmiii} \varepsilon',\\ &&\varepsilon \circ_{\rmi} \beta+\varepsilon \circ_{\rmi} \varepsilon-\varepsilon \circ_{\rmi} \varepsilon'+\varepsilon \circ_{\rmii} \beta+\varepsilon \circ_{\rmii} \varepsilon-\varepsilon \circ_{\rmii} \varepsilon'+\varepsilon \circ_{\rmiii} \beta+\varepsilon \circ_{\rmiii} \varepsilon-\varepsilon \circ_{\rmiii} \varepsilon'. \end{aligned}

We define an isomorphism of \mathbb{S}_2-modules by

\begin{aligned} \eta: \antipostlie(2) \otimes \opd(2) \otimes \mathbf{k}. \mathrm{sgn}_{\mathbb{S}_2} &\to& \mathrm{ATSu}(\opd)(2)\\ \beta \otimes \omega \otimes 1 &\mapsto& \begin{pmatrix} \omega\\ \cdot \end{pmatrix},\\ \varepsilon \otimes \omega \otimes 1 &\mapsto& \begin{pmatrix} \omega\\ \succ \end{pmatrix}, \end{aligned}

which induces an isomorphism of \mathbb{S}_3-modules:

\overline{\eta}: 3(\antipostlie(2) \otimes \opd(2) \otimes \mathbf{k}. \mathrm{sgn}_{\mathbb{S}_2})^{\otimes 2} \to 3\mathrm{ATSu}(\opd)^{\otimes 2}.

Then we just need to prove that the following equations hold for every relation \gamma of \opd:

\begin{aligned} & \overline{\eta}(\Psi((\beta \circ_{\rmi} \beta +\beta \circ_{\rmii} \beta +\beta \circ_{\rmiii} \beta)\otimes \gamma))=\mathrm{ATSu}_{\{x,y,z\}}(\gamma);\\ & \left\{ \begin{aligned} & \overline{\eta}(\Psi((-\varepsilon' \circ_{\rmi} \beta+\beta \circ_{\rmiii} \varepsilon+\beta' \circ_{\rmii} \varepsilon')\otimes \gamma))=\mathrm{ATSu}_{\{x,y\}}(\gamma);\\ & \overline{\eta}(\Psi((-\varepsilon' \circ_{\rmii} \beta+\beta \circ_{\rmi} \varepsilon+\beta' \circ_{\rmiii} \varepsilon')\otimes \gamma))=\mathrm{ATSu}_{\{y,z\}}(\gamma);\\ & \overline{\eta}(\Psi((-\varepsilon' \circ_{\rmiii} \beta+\beta \circ_{\rmii} \varepsilon+\beta' \circ_{\rmi} \varepsilon')\otimes \gamma))=\mathrm{ATSu}_{\{x,z\}}(\gamma); \end{aligned} \right.\\ & \left\{ \begin{aligned} & \overline{\eta}(\Psi((-\varepsilon \circ_{\rmii} \beta-\varepsilon \circ_{\rmii} \varepsilon+\varepsilon \circ_{\rmii} \varepsilon'-\varepsilon'\circ_{\rmiii} \varepsilon+\varepsilon'\circ_{\rmi} \varepsilon')\otimes \gamma))=\mathrm{ATSu}_{\{x\}}(\gamma);\\ & \overline{\eta}(\Psi((-\varepsilon \circ_{\rmiii} \beta-\varepsilon \circ_{\rmiii} \varepsilon+\varepsilon \circ_{\rmiii} \varepsilon'-\varepsilon'\circ_{\rmi} \varepsilon+\varepsilon'\circ_{\rmii} \varepsilon')\otimes \gamma))=\mathrm{ATSu}_{\{y\}}(\gamma);\\ & \overline{\eta}(\Psi((-\varepsilon \circ_{\rmi} \beta-\varepsilon \circ_{\rmi} \varepsilon+\varepsilon \circ_{\rmi} \varepsilon'-\varepsilon'\circ_{\rmii} \varepsilon+\varepsilon'\circ_{\rmiii} \varepsilon')\otimes \gamma))=\mathrm{ATSu}_{\{z\}}(\gamma); \end{aligned} \right.\\ & \overline{\eta}(\Psi((\varepsilon \circ_{\rmi} \beta+\varepsilon \circ_{\rmi} \varepsilon-\varepsilon \circ_{\rmi} \varepsilon'+\varepsilon \circ_{\rmii} \beta+\varepsilon \circ_{\rmii} \varepsilon-\varepsilon \circ_{\rmii} \varepsilon'+\varepsilon \circ_{\rmiii} \beta+\varepsilon \circ_{\rmiii} \varepsilon-\varepsilon \circ_{\rmiii} \varepsilon')\otimes \gamma))\\ & \quad =\ST(\gamma). \end{aligned}

First we prove that Eq. (Proof) holds for every \gamma\in \mathcal{T}(V)(3) using Eq. (?). Note that for all \omega and \nu in V, we have

\begin{aligned} \mathrm{ATSu}_{\{x,y,z\}}(\omega \circ_{\rmi} \nu)=\begin{pmatrix} \omega\\ \cdot \end{pmatrix} \circ_{\rmi} \begin{pmatrix} \nu\\ \cdot \end{pmatrix},~ \mathrm{ATSu}_{\{x,y,z\}}(\omega \circ_{\rmii} \nu)=\begin{pmatrix} \omega\\ \cdot \end{pmatrix} \circ_{\rmii} \begin{pmatrix} \nu\\ \cdot \end{pmatrix},~ \mathrm{ATSu}_{\{x,y,z\}}(\omega \circ_{\rmiii} \nu)=\begin{pmatrix} \omega\\ \cdot \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} \nu\\ \cdot \end{pmatrix}. \end{aligned}

Then we obtain that

\begin{aligned} \overline{\eta}(\Psi((\beta \circ_{\rmi} \beta +\beta \circ_{\rmii} \beta +\beta \circ_{\rmiii} \beta)\otimes (\omega \circ_{\rmi} \nu)))=\mathrm{ATSu}_{\{x,y,z\}}(\omega \circ_{\rmi} \nu). \end{aligned}

The same argument proves Eq. (Proof) for the monomials \omega \circ_{\rmii} \nu and \omega \circ_{\rmiii} \nu. Since \mathcal{T}(V)(3) is generated by elements of the form \omega \circ_{\rmi} \nu, \omega \circ_{\rmii} \nu and \omega \circ_{\rmiii} \nu, it is straightforward to check that Eq. (Proof) holds for every relation \gamma\in \mathcal{T}(V)(3).

Second we prove that Eq. (Proof) holds for every \gamma\in \mathcal{T}(V)(3) by Eq. (?). Since we have

\begin{aligned} \mathrm{ATSu}_{\{x,y\}}(\omega \circ_{\rmi} \nu)=\begin{pmatrix} \omega\\ \prec \end{pmatrix} \circ_{\rmi} \begin{pmatrix} \nu\\ \cdot \end{pmatrix},~ \mathrm{ATSu}_{\{x,y\}}(\omega \circ_{\rmii} \nu)=\begin{pmatrix} \omega\\ \cdot \end{pmatrix} \circ_{\rmii} \begin{pmatrix} \nu\\ \prec \end{pmatrix},~ \mathrm{ATSu}_{\{x,y\}}(\omega \circ_{\rmiii} \nu)=\begin{pmatrix} \omega\\ \cdot \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} \nu\\ \succ \end{pmatrix}, \end{aligned}

which implies that

\begin{aligned} \overline{\eta}(\Psi((-\varepsilon' \circ_{\rmi} \beta+\beta \circ_{\rmiii} \varepsilon+\beta' \circ_{\rmii} \varepsilon')\otimes (\omega \circ_{\rmi} \nu)))=\mathrm{ATSu}_{\{x,y\}}(\omega \circ_{\rmi} \nu). \end{aligned}

Similarly, Eq. (Proof) also holds for the monomials \omega \circ_{\rmii} \nu and \omega \circ_{\rmiii} \nu. Thus, Eq. (Proof) holds for every relation \gamma\in \mathcal{T}(V)(3).

Third, we prove that Eq. (Proof) holds for every \gamma\in \mathcal{T}(V)(3) according to Eq. (?). Since we have

\begin{aligned} \mathrm{ATSu}_{\{x\}}(\omega \circ_{\rmi} \nu)=\begin{pmatrix} \omega\\ \prec \end{pmatrix} \circ_{\rmi} \begin{pmatrix} \nu\\ \prec \end{pmatrix},~ \mathrm{ATSu}_{\{x\}}(\omega \circ_{\rmii} \nu)=\begin{pmatrix} \omega\\ \succ \end{pmatrix} \circ_{\rmii} \begin{pmatrix} \nu\\ \star \end{pmatrix},~ \mathrm{ATSu}_{\{x\}}(\omega \circ_{\rmiii} \nu)=\begin{pmatrix} \omega\\ \prec \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} \nu\\ \succ \end{pmatrix}, \end{aligned}

which implies that

\begin{aligned} \overline{\eta}(\Psi((-\varepsilon \circ_{\rmii} \beta-\varepsilon \circ_{\rmii} \varepsilon+\varepsilon \circ_{\rmii} \varepsilon'-\varepsilon'\circ_{\rmiii} \varepsilon+\varepsilon'\circ_{\rmi} \varepsilon')\otimes (\omega \circ_{\rmi} \nu)))= \mathrm{ATSu}_{\{x\}}(\omega \circ_{\rmi} \nu). \end{aligned}

Similarly, Eq. (Proof) also holds for the monomials \omega \circ_{\rmii} \nu and \omega \circ_{\rmiii} \nu. Thus, Eq. (Proof) holds for every relation \gamma\in \mathcal{T}(V)(3).

Finally, we prove that Eq. (Proof) holds for every \gamma\in \mathcal{T}(V)(3) by Eq. (?). Since we have

\begin{aligned} \ST(\omega \circ_{\rmi} \nu)=\begin{pmatrix} \omega\\ \star \end{pmatrix} \circ_{\rmi} \begin{pmatrix} \nu\\ \star \end{pmatrix},~ \ST(\omega \circ_{\rmii} \nu)=\begin{pmatrix} \omega\\ \star \end{pmatrix} \circ_{\rmii} \begin{pmatrix} \nu\\ \star \end{pmatrix},~ \ST(\omega \circ_{\rmiii} \nu)=\begin{pmatrix} \omega\\ \star \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} \nu\\ \star \end{pmatrix}, \end{aligned}

which implies that

\begin{aligned} &&\overline{\eta}(\Psi((\varepsilon \circ_{\rmi} \beta+\varepsilon \circ_{\rmi} \varepsilon-\varepsilon \circ_{\rmi} \varepsilon'+\varepsilon \circ_{\rmii} \beta+\varepsilon \circ_{\rmii} \varepsilon-\varepsilon \circ_{\rmii} \varepsilon'+\varepsilon \circ_{\rmiii} \beta+\varepsilon \circ_{\rmiii} \varepsilon-\varepsilon \circ_{\rmiii} \varepsilon')\\ &&\otimes (\omega \circ_{\rmi} \nu)))=\ST(\omega \circ_{\rmi} \nu). \end{aligned}

Similarly, Eq. (Proof) also holds for the monomials \omega \circ_{\rmii} \nu and \omega \circ_{\rmiii} \nu. Thus, Eq. (Proof) holds for every relation \gamma\in \mathcal{T}(V)(3). The proof is finished.

Theorem

Theorem 3.3

HRYH

Let \opd be a binary quadratic operad. We have the isomorphism of operads

\mathrm{ABSu}(\opd) \cong \antiprelie \bullet \opd.
Used by Proof
Proof of Theorem 3.3

It follows from Theorem 3.2 by using the following commutative diagram:

\begin{aligned} \xymatrix{ \mathrm{ATSu}(\opd)\ar[d] \ar[r]^{\cong\quad} & \antipostlie \bullet \opd \ar[d] \\ \mathrm{ABSu}(\opd) \ar[r] & \antiprelie \bullet \opd. } \end{aligned}

Here the two vertical maps are surjective and the top isomorphism preserves their kernels. Thus, the bottom map turns out to be an isomorphism of operads.

Proposition

Proposition 3.4

ANTM

Let \opd be a binary quadratic nonsymmetric operad. There are isomorphisms of nonsymmetric operads

\mathrm{ABSu}(\opd) \cong \antiden~\blacksquare~\opd, \quad \mathrm{ATSu}(\opd) \cong \antitriden~\blacksquare~\opd.

where \blacksquare denotes the black square product in [FG, Val2].

Corollary

Corollary 3.5

40WQ

We have \antiprelie \bullet \com\cong \antizinbiel, \antiprelie \bullet \ass\cong \antiden and \antiprelie \bullet \poisson\cong \antiprepoisson.

Corollary

Corollary 3.6

X538

We have \antipostlie \bullet \com\cong \antipostcom, \antipostlie \bullet \lie\cong \antipostlie and \antitriden~\blacksquare~\ass\cong\antitriden.

Definition

Definition 4.1

5G5O

Let \gensp be a vector space with a basis \genbas.

  1. (i) Define a vector space

    \gendup= \gensp \otimes (\mathbf{k} \dashv \oplus~ \mathbf{k} \vdash),

    where we denote (\omega~\otimes \dashv) (resp. (\omega~\otimes \vdash)) by \begin{pmatrix} \omega \\ \dashv \end{pmatrix} (resp. \begin{pmatrix} \omega \\ \vdash \end{pmatrix}) for \omega\in \gensp. Then \bigcup\limits_{\omega\in \gensp} \Biggl\{\begin{pmatrix} \omega \\ \dashv \end{pmatrix},\begin{pmatrix} \omega \\ \vdash \end{pmatrix}\Biggr\} is a basis of \gendup. \gendup can be viewed as an arity graded vector space concentrated in arity 2.

  2. (ii) For a labeled n-tree \tau in \calt(\genbas), define \overline{\tau} in \mathcal{T}_{\hspace*{-0.1cm}ns}(\gendup) as follows:

    • \bullet \overline{~|~}=|

    • \bullet when n\geq 2, \overline{\tau} is obtained by replacing each decoration \gop \in \vin(\tau) by

      \begin{pmatrix} \gop \\ \dagger \end{pmatrix}:=\Biggl\{-\begin{pmatrix} \gop \\ \dashv \end{pmatrix},-\begin{pmatrix} \gop \\ \vdash \end{pmatrix}\Biggr\}.

    Thus, \overline{\tau} is a set of labeled trees. Then we extend this definition to \mathcal{T}_{\hspace*{-0.1cm}ns}(\gendup) by linearity.

Definition

Definition 4.2

NX47

Let \gensp be a vector space with a basis \genbas. Let \tau be a labeled n-tree in \calt(\genbas). The anti-duplicator \mathrm{ADup}_{x}(\tau) of \tau with respect to a leaf x\in \lin(\tau) is a subset of \mathcal{T}_{\hspace*{-0.1cm}ns}(\gendup) defined by induction on |\lin(\tau)| as follows:

  • \bullet \mathrm{ADup}_{x}(~|~)=\{~|~\} ;

  • \bullet Assume that \mathrm{ADup}_{x}(\tau) have been defined for \tau with |\lin(\tau)|\leq k for any k\geq 1. Then for a labeled (k+1)-tree \tau\in \calt(\genbas) with its decomposition \tau_{l} \vee_{\gop} \tau_{r}, we define

    \mathrm{ADup}_{x}(\tau)= \mathrm{ADup}_{x}(\tau_{l} \vee_{\gop} \tau_{r}) = \left\{\begin{array}{ll} \mathrm{ADup}_{x}(\tau_{l}) \vee_{\begin{pmatrix} \gop \\ \dashv \end{pmatrix}} \overline{\tau_{r}},\quad & \text{if}~x \in \lin(\tau_{l}); \\ \overline{\tau_{l}} \vee_{\begin{pmatrix} \gop \\ \vdash \end{pmatrix}} \mathrm{ADup}_{x}(\tau_{r}), \quad & \text{if}~x \in \lin(\tau_{r}). \end{array}\right.
Proposition

Proposition 4.3

OZRS

Let \gensp be a vector space with a basis \genbas, \tau be a labeled n-tree in \calt(\genbas) and x be in \lin(\tau). The anti-duplicator \mathrm{ADup}_{x}(\tau) of \tau is obtained by relabeling each vertex of \tau according to the following rules:

  1. (a) We replace the label \gop of each vertex on the path from the root to the leaf x of \tau by

    • (i) \begin{pmatrix} \gop \\ \dashv \end{pmatrix}, if the path turns left at this vertex;

    • (ii) \begin{pmatrix} \gop \\ \vdash \end{pmatrix}, if the path turns right at this vertex;

  2. (b) We replace the label \gop of each vertex not on the path from the root to the leaf x of \tau by

    \begin{pmatrix} \gop \\ \dagger \end{pmatrix}:=\Biggl\{-\begin{pmatrix} \gop \\ \dashv \end{pmatrix},-\begin{pmatrix} \gop \\ \vdash \end{pmatrix}\Biggr\}.
Proof of Proposition 4.3

We can prove it by induction on |\lin(\tau)|\geq 1. We omit the details for brevity.

Example

Example 4.4

17KP
\mathrm{ADup}_{x_{2}} \left( \begin{xy} (0,8)*+{x_{1}}="v1";(10,8)*+{x_{2}}="v2";(20,8)*+{x_{3}}="v3"; (30,8)*+{x_{4}}="v4"; (5,0)*+{\omega_{1}}="v5";(25,0)*+{\omega_{2}}="v6"; (15,-6)*+{\omega_{3}}="v7";(15,-12)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy} \right)= \begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\vdash}}="v5";(25,0)* +{\svec{\omega_{2}}{\dagger}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{->}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{->}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{->}"v8";"v7"} \end{xy}
= \left\{-\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\vdash}}="v5";(25,0)* +{\svec{\omega_{2}}{\dashv}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"}\end{xy},\quad -\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\vdash}}="v5";(25,0)* +{\svec{\omega_{2}}{\vdash}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy}\right\}
Definition

Definition 4.5

ZLUH

Let \gensp be a vector space with a basis \genbas.

  1. (i) Define a vector space

    \gentrip= \gensp \otimes (\mathbf{k} \dashv \oplus~ \mathbf{k} \vdash\oplus~ \mathbf{k} \perp),

    where we denote (\omega~\otimes \dashv) (resp. (\omega~\otimes \vdash), resp. (\omega~\otimes \perp)) by \begin{pmatrix} \omega \\ \dashv \end{pmatrix} (resp. \begin{pmatrix} \omega \\ \vdash \end{pmatrix}, resp. \begin{pmatrix} \omega \\ \perp \end{pmatrix}) for \omega\in \gensp. Then \bigcup\limits_{\omega\in \gensp} \Biggl\{\begin{pmatrix} \omega \\ \dashv \end{pmatrix},\begin{pmatrix} \omega \\ \vdash \end{pmatrix},\begin{pmatrix} \omega \\ \perp \end{pmatrix}\Biggr\} is a basis of \gentrip. \gentrip can be viewed as an arity graded vector space concentrated in arity 2.

  2. (ii) For a labeled n-tree \tau in \calt(\genbas), define \overrightarrow{\tau} in \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentrip) as follows:

    • \bullet \overrightarrow{~|~}=|

    • \bullet when n\geq 2, \overrightarrow{\tau} is obtained by replacing each decoration \gop \in \vin(\tau) by

      \begin{pmatrix} \gop \\ \ddagger \end{pmatrix}:=\Biggl\{-\begin{pmatrix} \gop \\ \dashv \end{pmatrix},-\begin{pmatrix} \gop \\ \vdash \end{pmatrix},-\begin{pmatrix} \gop \\ \perp \end{pmatrix}\Biggr\}.

    Thus, \overrightarrow{\tau} is a set of labeled trees. Then we extend this definition to \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentrip) by linearity.

Definition

Definition 4.6

DYFM

Let \gensp be a vector space with a basis \genbas. Let \tau be a labeled n-tree in \calt(\genbas). The anti-triplicator \mathrm{ATrip}_{J}(\tau) of \tau with respect to a subset J \subseteq \lin(\tau) is a subset of \mathcal{T}_{\hspace*{-0.1cm}ns}(\gentrip) defined by induction on |\lin(\tau)| as follows:

  • \bullet \mathrm{ATrip}_{J}(~|~)=\{~|~\} ;

  • \bullet \mathrm{ATrip}_{\emptyset}(\tau)=\overrightarrow{\tau};

  • \bullet Assume that \mathrm{ATrip}_{J}(\tau) have been defined for \tau with |\lin(\tau)|\leq k for any k\geq 1. Then for a labeled (k+1)-tree \tau\in \calt(\genbas) with its decomposition \tau_{l} \vee_{\gop} \tau_{r}, we define

    \mathrm{ATrip}_{J}(\tau)= \mathrm{ATrip}_{J}(\tau_{l} \vee_{\gop} \tau_{r}) = \left\{\begin{array}{ll} \mathrm{ATrip}_{J}(\tau_{l}) \vee_{\begin{pmatrix} \gop \\ \dashv \end{pmatrix}} \overrightarrow{\tau_{r}},\quad & \text{if}~J \subseteq \lin(\tau_l); \\ \overrightarrow{\tau_{l}} \vee_{\begin{pmatrix} \gop \\ \vdash \end{pmatrix}} \mathrm{ATrip}_{J}(\tau_{r}), \quad & \text{if}~J \subseteq \lin(\tau_r);\\ \overrightarrow{\tau_{l}} \vee_{\begin{pmatrix} \gop \\ \ddagger \end{pmatrix}} \overrightarrow{\tau_{r}},\quad & \text{if}~J=\emptyset; \\ \mathrm{ATrip}_{J\cap \lin(\tau_l)}(\tau_{l}) \vee_{\begin{pmatrix} \gop \\ \perp \end{pmatrix}} \mathrm{ATrip}_{J\cap \lin(\tau_r)}(\tau_{r}),\quad & \text{otherwise.} \end{array}\right.
Proposition

Proposition 4.7

TQWG

Let \gensp be a vector space with a basis \genbas, \tau be a labeled n-tree in \calt(\genbas) and J be a nonempty subset of \lin(\tau). The anti-triplicator \mathrm{ATrip}_{J}(\tau) of \tau is obtained by relabeling each vertex of \vin(\tau) by the following rules:

  1. (a) We replace the label \gop of each vertex on the paths from the root of \tau to some leaves (possibly multiple) in J by

    • (i) \begin{pmatrix} \gop \\ \dashv \end{pmatrix}, if all of such paths turn left at this vertex;

    • (ii) \begin{pmatrix} \gop \\ \vdash \end{pmatrix}, if all of such paths turn right at this vertex;

    • (iii) \begin{pmatrix} \gop \\ \perp \end{pmatrix}, if some of such paths turn left and some of such paths turn right.

  2. (b) We replace the label \gop of each vertex not on the path from the root to any leaf x\in J by

    \begin{pmatrix} \gop \\ \ddagger \end{pmatrix}:=\Biggl\{-\begin{pmatrix} \gop \\ \dashv \end{pmatrix},-\begin{pmatrix} \gop \\ \vdash \end{pmatrix},-\begin{pmatrix} \gop \\ \perp \end{pmatrix}\Biggr\}.
Proof of Proposition 4.7

We can prove it by induction on |\lin(\tau)|\geq 1. We omit the details for brevity.

Example

Example 4.8

C3CH
\mathrm{ATrip}_{\{x_{1},x_{2}\}} \left( \begin{xy} (0,8)*+{x_{1}}="v1";(10,8)*+{x_{2}}="v2";(20,8)*+{x_{3}}="v3"; (30,8)*+{x_{4}}="v4"; (5,0)*+{\omega_{1}}="v5";(25,0)*+{\omega_{2}}="v6"; (15,-6)*+{\omega_{3}}="v7";(15,-12)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy} \right)= \begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\perp}}="v5";(25,0)* +{\svec{\omega_{2}}{\ddagger}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{->}"v5";"v1"};{\ar@{->}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{->}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{->}"v8";"v7"} \end{xy}
= \left\{-\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\perp}}="v5";(25,0)* +{\svec{\omega_{2}}{\dashv}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"}\end{xy},\quad -\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\perp}}="v5";(25,0)* +{\svec{\omega_{2}}{\vdash}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy},\quad -\begin{xy} (0,10)*+{x_{1}}="v1";(10,10)*+{x_{2}}="v2";(20,10)*+{x_{3}}="v3"; (30,10)*+{x_{4}}="v4"; (5,0)*+{\svec{\omega_{1}}{\perp}}="v5";(25,0)* +{\svec{\omega_{2}}{\perp}}="v6"; (15,-6)*+{\svec{\omega_{3}}{\dashv}}="v7";(15,-15)*+{}="v8"; {\ar@{-}"v5";"v1"};{\ar@{-}"v5";"v2"};{\ar@{-}"v6";"v3"};{\ar@{-}"v6";"v4"};{\ar@{-}"v7";"v5"};{\ar@{-}"v7";"v6"};{\ar@{-}"v8";"v7"} \end{xy}\right\}
Definition

Definition 4.9

3UMD
  1. (i) The anti-duplicator of \opd is defined to be the binary operad

    \mathrm{ADup}(\opd):=\mathcal{T}(\gendup)/ \bigl((\mathrm{ADup}(\relsp))\cap(\comp(\relsp))\bigr),

    where \gendup=\gensp\otimes(\mathbf{k}\dashv\oplus\mathbf{k}\vdash) is concentrated in arity 2, and the \mathbb{S}_2-action on \gendup is given by

    \begin{pmatrix}\gop\\ \dashv\end{pmatrix}^{(12)} :=\begin{pmatrix}\gop^{(12)}\\ \vdash\end{pmatrix}, \quad \begin{pmatrix}\gop\\ \vdash\end{pmatrix}^{(12)} :=\begin{pmatrix}\gop^{(12)}\\ \dashv\end{pmatrix}, \quad \forall\,\gop\in\gensp.

    The space of relations is generated as an \mathbb{S}-module by

    \mathrm{ADup}(\relsp):= \bigcup_{s=1}^k\Bigl(\bigcup_{x\in\lin(r_s)} \mathrm{ADup}_x(r_s)\Bigr), \quad \text{where }\mathrm{ADup}_x(r_s):=\sum_i c_{s,i}\mathrm{ADup}_x(\tau_{s,i}),

    according to the notation by Eq. (?). Moreover, (\comp(\relsp)) is spanned by

    \comp(\relsp):=\bigcup_{s=1}^k\comp(r_s), \quad \text{where }\comp(r_s):=\sum_i c_{s,i}\overline{\tau_{s,i}}.

    Here \comp(\tau_{s,i})=\overline{\tau_{s,i}} is obtained by replacing each vertex decoration \gop of \tau_{s,i} by

    \begin{pmatrix}\gop\\ \dagger\end{pmatrix} :=\Biggl\{ -\begin{pmatrix}\gop\\ \dashv\end{pmatrix}, -\begin{pmatrix}\gop\\ \vdash\end{pmatrix} \Biggr\}.
  2. (ii) The anti-triplicator of \opd is defined to be the binary operad

    \mathrm{ATrip}(\opd):=\mathcal{T}(\gentrip)/ \bigl((\mathrm{ATrip}(\relsp))\cap(\COMP(\relsp))\bigr),

    where \gentrip=\gensp\otimes (\mathbf{k}\dashv\oplus\mathbf{k}\vdash\oplus\mathbf{k}\perp) is concentrated in arity 2, and the \mathbb{S}_2-action on \gentrip is given by

    \begin{pmatrix}\gop\\ \dashv\end{pmatrix}^{(12)} :=\begin{pmatrix}\gop^{(12)}\\ \vdash\end{pmatrix}, \quad \begin{pmatrix}\gop\\ \vdash\end{pmatrix}^{(12)} :=\begin{pmatrix}\gop^{(12)}\\ \dashv\end{pmatrix}, \quad \begin{pmatrix}\gop\\ \perp\end{pmatrix}^{(12)} :=\begin{pmatrix}\gop^{(12)}\\ \perp\end{pmatrix}, \quad \forall\,\gop\in\gensp.

    The space of relations is generated as an \mathbb{S}-module by

    \mathrm{ATrip}(\relsp):= \bigcup_{s=1}^k\Bigl( \bigcup_{\emptyset\neq J\subseteq\lin(r_s)} \mathrm{ATrip}_J(r_s)\Bigr), \quad \text{where }\mathrm{ATrip}_J(r_s):=\sum_i c_{s,i}\mathrm{ATrip}_J(\tau_{s,i}),

    and

    \COMP(\relsp):=\bigcup_{s=1}^k\COMP(r_s), \quad \text{where }\COMP(r_s):=\sum_i c_{s,i}\overrightarrow{\tau_{s,i}}.

    Here \COMP(\tau_{s,i})=\overrightarrow{\tau_{s,i}} is obtained by replacing each vertex decoration \gop of \tau_{s,i} by

    \begin{pmatrix}\gop\\ \ddagger\end{pmatrix} :=\Biggl\{ -\begin{pmatrix}\gop\\ \dashv\end{pmatrix}, -\begin{pmatrix}\gop\\ \vdash\end{pmatrix}, -\begin{pmatrix}\gop\\ \perp\end{pmatrix} \Biggr\}.
Remark

Remark 4.10

ND22

The above definitions also apply to binary nonsymmetric operads by omitting the symmetric group actions and replacing the free symmetric operad \mathcal{T}(\gensp) with the free nonsymmetric operad \mathcal{T}_{\hspace*{-0.1cm}ns}(\gensp). In this setting, the definitions of the anti-duplicator and the anti-triplicator do not depend on the choice of a basis of the vector space of generating operations. This follows from the linearity of the constructions and the treewise tensor module structure on \mathcal{T}_{\hspace*{-0.1cm}ns}(\gensp).

Definition

Definition 4.11

DWH4

An anti-dupli-com algebra (A,\diamond) is a vector space A equipped with a bilinear operation \diamond:A \otimes A \to A satisfying the following equations

\begin{aligned} (x \diamond y) \diamond z &= (y \diamond x) \diamond z;\\ x \diamond (y \diamond z) &= y \diamond (x \diamond z);\\ (x \diamond y) \diamond z+x \diamond (y \diamond z) &= (x \diamond z) \diamond y+x \diamond (z \diamond y), \end{aligned}

for all x,y,z \in A.

Used by Proof
Proposition

Proposition 4.12

UIT6

The operad \antiperm is the anti-duplicator of the operad \com.

Proof of Proposition 4.12

Let \omega be the generating operation of the operad \com. Set \diamond:=\begin{pmatrix} \omega\\ \vdash \end{pmatrix}. Since \begin{pmatrix} \omega\\ \dashv \end{pmatrix}^{(12)}=\begin{pmatrix} \omega^{(12)}\\ \vdash \end{pmatrix}=\begin{pmatrix} \omega\\ \vdash \end{pmatrix}=\diamond, we have \begin{pmatrix} \omega\\ \dashv \end{pmatrix}=\diamond^{(12)}. The space of relations of \com is generated as an \mathbb{S}_3-module by

v_1-v_9= \omega \circ_{\rmi} \omega-\omega \circ_{\rmii} \omega.

Then we have

\begin{aligned} \mathrm{ADup}_{x}(v_1-v_9)&= \{ z \diamond (y \diamond x)+(y \diamond z) \diamond x,z \diamond (y \diamond x)+(z \diamond y) \diamond x \};\\ \mathrm{ADup}_{y}(v_1-v_9)&= \{ z \diamond (x \diamond y)-x \diamond (z \diamond y)\};\\ \mathrm{ADup}_{z}(v_1-v_9)&= \{(x \diamond y) \diamond z-(y \diamond x) \diamond z, (x \diamond y) \diamond z+x \diamond (y \diamond z)\}, \end{aligned}

and

\begin{aligned} &&\comp(v_1-v_9)= \{(x \diamond y) \diamond z-z \diamond (x \diamond y), (x \diamond y) \diamond z-(y \diamond x) \diamond z, (x \diamond y) \diamond z-z \diamond (y \diamond x),\\ && (x \diamond y) \diamond z-(y \diamond z) \diamond x,(x \diamond y) \diamond z-x \diamond (y \diamond z),(x \diamond y) \diamond z-(z \diamond y) \diamond x,(x \diamond y) \diamond z-x \diamond (z \diamond y)\}. \end{aligned}

By a direct and extensive calculation, we can obtain bases for the relation spaces \mathrm{ADup}(\com) and \comp(\com), respectively, as follows:

\begin{aligned} \mathrm{ADup}(\com)&=\langle (x\diamond y)\diamond z- (y\diamond x)\diamond z,~(x\diamond y)\diamond z+x \diamond (y \diamond z) \rangle;\\ \comp(\com)&=\langle (x\diamond y)\diamond z- (y\diamond x)\diamond z,~ (x\diamond y)\diamond z-x \diamond (y \diamond z),~ (x\diamond y)\diamond z- (x\diamond z)\diamond y \rangle, \end{aligned}

for all x,y,z \in A.

Now we compute the intersection of subspaces \mathrm{ADup}(\com) and \comp(\com). It is easy to see that (x\diamond y)\diamond z- (y\diamond x)\diamond z belongs to the intersection of \mathrm{ADup}(\com) and \comp(\com), which corresponds to the Eq. (4.11). Then we consider the following equation

\begin{aligned} &&a_1 ((x\diamond y)\diamond z+x \diamond (y \diamond z))+a_2 ((x\diamond z)\diamond y+x \diamond (z \diamond y))+ a_3 (y \diamond (x \diamond z)+(y \diamond x)\diamond z)\\ &&+a_4 (y \diamond (z \diamond x)+(y \diamond z)\diamond x)+a_5 ((z \diamond x)\diamond y+ z \diamond (x \diamond y))+a_6 ((z \diamond y)\diamond x+z \diamond (y \diamond x))\\ &= b_1 ((x \diamond y)\diamond z-(y \diamond x)\diamond z)+b_2 ((x \diamond z)\diamond y-(z \diamond x)\diamond y)+ b_3 ((y \diamond z)\diamond x-(z \diamond y)\diamond x)\\ &&+ b_4 ((x \diamond y)\diamond z-x \diamond (y \diamond z))+b_5 ((x \diamond z)\diamond y-x \diamond (z \diamond y))+b_6 ((y \diamond x)\diamond z-y \diamond (x \diamond z))\\ &&+b_7 ((y \diamond z)\diamond x-y \diamond (z \diamond x))+b_8 ((z \diamond x)\diamond y-z \diamond (x \diamond y))+ b_9 ((z \diamond y)\diamond x-z \diamond (y \diamond x))\\ &&+b_{10} ((x \diamond y)\diamond z-(x \diamond z) \diamond y)+b_{11} ((y \diamond z)\diamond x-(y \diamond x)\diamond z)+b_{12} ((z \diamond x)\diamond y-(z \diamond y) \diamond x). \end{aligned}

Since \mathcal{T}(V)(3) is generated by the twelve elements v_i~(1\leq i \leq 12) in the table (?), we obtain the following system of linear equations in the variables a_m, b_n, where 1 \leq m \leq 6 and 1 \leq n \leq 12.

\begin{array}{c@{\quad}c@{\quad}c} \left\{ \begin{aligned} a_1 &= -b_4, \\ a_2 &= -b_5, \\ a_3 &= -b_6, \\ a_4 &= -b_7, \\ a_5 &= -b_8, \\ a_6 &= -b_9, \end{aligned} \right. & \text{and} & \left\{ \begin{aligned} 2a_1-b_1-b_{10} &= 0, \\ 2a_5+b_2-b_{12} &= 0, \\ 2a_4-b_3-b_{11} &= 0, \\ 2a_2-b_2+b_{10} &= 0, \\ 2a_6+b_3+b_{12} &= 0, \\ 2a_3+b_1+b_{11} &= 0. \end{aligned} \right. \end{array}

Since there are 18 variables and 12 equations, the final general solution is 6-dimensional. Thus the general solution of the above system of linear equations can be expressed as

(a_1,\dots,a_6,b_1,\dots,b_{12})^{\mathrm{T}} = t_1\eta_1+t_2\eta_2+t_3\eta_3+t_4\eta_4+t_5\eta_5+t_6\eta_6,

where t_1,\dots,t_6\in \mathbf{k} are arbitrary constants, not all zero.

A fundamental system of solutions is given by (\eta_1,\dots,\eta_6), where

\begin{aligned} \begin{aligned} \eta_1&=(1,0,-1,0,0,0;\ 2,0,0,-1,0,1,0,0,0,0,0,0);\\ \eta_2&=(0,1,0,0,-1,0;\ 0,2,0,0,-1,0,0,1,0,0,0,0);\\ \eta_3&=(0,0,0,1,0,-1;\ 0,0,2,0,0,0,-1,0,1,0,0,0);\\ \eta_4&=(1,-1,0,0,0,0;\ 0,0,0,-1,1,0,0,0,0,2,0,0);\\ \eta_5&=(-1,0,0,1,0,0;\ -2,0,0,1,0,0,-1,0,0,0,2,0);\\ \eta_6&=(0,1,0,-1,0,0;\ 0,2,-2,0,-1,0,1,0,0,0,0,2). \end{aligned} \end{aligned}

Here the coordinate order is (a_1,a_2,a_3,a_4,a_5,a_6;b_1,b_2,b_3,b_4,b_5,b_6,b_7,b_8,b_9,b_{10},b_{11},b_{12}). The six vectors \eta_1,\dots,\eta_6 are linearly independent and hence form a basis of the solution space of the above system.

When we consider the solution \eta_1, we obtain the relation (x \diamond y)\diamond z-(y \diamond x)\diamond z+x \diamond (y \diamond z)-y \diamond (x \diamond z). Since (x\diamond y)\diamond z- (y\diamond x)\diamond z belongs to the intersection of \mathrm{ADup}(\com) and \comp(\com), we deduce that the relation x\diamond (y \diamond z)-y \diamond (x \diamond z) belongs to the intersection of \mathrm{ADup}(\com) and \comp(\com), which corresponds to Eq. (4.11). If we consider the solutions \eta_2 and \eta_3, we obtain analogous relations, with the difference being only up to a permutation.

Considering the solution \eta_4, we have the relation (x \diamond y)\diamond z-(x \diamond z)\diamond y+x\diamond (y \diamond z)-x\diamond (z \diamond y), which corresponds to Eq. (4.11). When the solutions \eta_5 and \eta_6 are considered, relations similar to the above one are obtained, which differ only by a permutation. Therefore, the operad \antiperm is the anti-duplicator of the operad \com.

Definition

Definition 4.13

LWFN

A dual Lie admissible algebra (A,\diamond) is a vector space A equipped with a bilinear operation \diamond:A \otimes A \to A satisfying the following equations

\begin{aligned} (x\diamond y)\diamond z&=(y\diamond x)\diamond z;\\ (x\diamond y)\diamond z&=x \diamond (y \diamond z);\\ (x\diamond y)\diamond z&=(x\diamond z)\diamond y, \end{aligned}

for all x,y,z\in A.

Proposition

Proposition 4.15

769L

The dual Lie admissible algebras are precisely the Koszul dual of the Lie admissible algebras.

Proof of Proposition 4.15

The operad encoding Lie admissible algebras \liead is given as follows

\liead=\mathcal{T}(V)\Big/ (R)=\mathcal{T}(V) \Big/ \Big(v_1-v_2+v_3-v_4+v_5-v_6+v_7-v_8+v_9-v_{10}+v_{11}-v_{12} \Big),

where V is generated by one binary operation \mu and the relation space (R) is generated as an \mathbb{S}_3-module by v_1-v_2+v_3-v_4+v_5-v_6+v_7-v_8+v_9-v_{10}+v_{11}-v_{12} in the table (?). We define its Czech dual V^{\vee}=V^{\ast} \otimes \mathrm{sgn}_2, where V^{\vee} is actually generated by one element \mu^*. There is a natural pairing with respect to the duality given as follows:

\begin{aligned} &&\langle\cdot,\cdot\rangle: \mathcal{T}(\gensp^{\vee})(3) \otimes \mathcal{T}(\gensp)(3) \to \mathbf{k},\\ && \langle \mu^* \circ_{u} \mu^*, \mu \circ_{v} \mu \rangle=\delta_{(u,v)} \in \mathbf{k}. \end{aligned}

We use x_i to denote the equation obtained from v_i~(1\leq i\leq 12) in the table (?), by replacing every occurrence of \mu with \mu^*. For example, x_2 stands for (\mu^*)'\circ_{\rmii} \mu^*. Here x_i is understood as the dual tree of v_i, so that \langle x_i, v_j \rangle = \delta_{ij} for all 1 \le i,j \le 12. Denote by R^{\perp} the annihilator of R with respect to the above natural pairing. That means,

R^{\perp}:=\Biggl\{ x \in \mathcal{T}(\gensp^{\vee})(3)~|~\langle x,u\rangle=0,~\forall u\in (R)\Biggr\}.

Let \sum_{j=1}^{12} k_j x_j be an arbitrary element of R^{\perp}. Then we have

\langle v_1-v_2+v_3-v_4+v_5-v_6+v_7-v_8+v_9-v_{10}+v_{11}-v_{12}, \sum_{j=1}^{12} k_j x_j \rangle=0.

Since \langle x_i,v_i\rangle=1 for all 1\leq i\leq 12, we deduce that

k_1-k_2+k_3-k_4+k_5-k_6+k_7-k_8+k_9-k_{10}+k_{11}-k_{12}=0,

Thus, we have

\begin{aligned} \sum\limits_{j=1}^{12} k_j x_j&=k_2 (x_1+x_2)+k_3 (-x_1+x_3)+k_4 (x_1+x_4)+k_5 (-x_1+x_5)+k_6 (x_1+x_6)+k_7 (-x_1+x_7)\\ &&+k_8 (x_1+x_8)+k_9 (-x_1+x_9)+k_{10} (x_1+x_{10})+k_{11} (-x_1+x_{11})+k_{12} (x_1+x_{12}) \in R^{\perp}. \end{aligned}

Since k_j \in \mathbf{k} is arbitrary for all 2\leq j\leq 12, we deduce that

x_1=-x_2=x_3=-x_4=x_5=-x_6=x_7=-x_8=x_9=-x_{10}=x_{11}=-x_{12},

which implies that the following equations are contained in the Koszul dual algebras of Lie admissible algebras:

\begin{aligned} &&(x \diamond y)\diamond z=x \diamond (y \diamond z), \quad (x \diamond y)\diamond z=(x \diamond z) \diamond y,\quad (x \diamond y)\diamond z=z \diamond (x \diamond y),\\ &&(x \diamond y)\diamond z=x \diamond (z \diamond y), \quad (x \diamond y)\diamond z=(z \diamond y) \diamond x,\quad (x \diamond y)\diamond z=y \diamond (z \diamond x),\\ &&(x \diamond y)\diamond z=z \diamond (y \diamond x), \quad (x \diamond y)\diamond z=(y \diamond x) \diamond z,\quad (x \diamond y)\diamond z=y \diamond (x \diamond z),\\ &&(x \diamond y)\diamond z=(z \diamond x)\diamond y, \quad (x \diamond y)\diamond z=(y \diamond z)\diamond x,\quad \forall x,y,z\in A. \end{aligned}

It is straightforward to check that the above equations can be obtained from the following three equalities:

(x\diamond y)\diamond z=(y\diamond x)\diamond z,\quad (x\diamond y)\diamond z=x \diamond (y \diamond z),\quad (x\diamond y)\diamond z=(x\diamond z)\diamond y,

which corresponds to the Eqs. (4.13)-(4.13). Therefore, we have proved that the dual Lie admissible algebras are precisely the Koszul dual of the Lie admissible algebras.

Definition

Definition 4.16

9QOM

An anti-dupli-Lie algebra (A,\circ) is a vector space A equipped with a bilinear operation \circ:A \otimes A \to A satisfying the following equations

\begin{aligned} \sum_{\text{cyclic}} (x \circ y+y \circ x)\circ z&=0;\\ \sum_{\text{cyclic}} (x \circ (z \circ y)+(x \circ z)\circ y-z \circ (x \circ y))&=0, \quad \forall x,y,z \in A, \end{aligned}

where the sum is taken over cyclic permutations of x,y,z \in A.

Used by Proof
Proposition

Proposition 4.17

NNP9

The operad \antileibniz is the anti-duplicator of the operad \lie.

Proof of Proposition 4.17

Let \mu be the generating operation of the operad \lie. Set \circ:=\begin{pmatrix} \mu\\ \vdash \end{pmatrix}. Since \begin{pmatrix} \mu\\ \dashv \end{pmatrix}^{(12)}=\begin{pmatrix} \mu^{(12)}\\ \vdash \end{pmatrix}=-\begin{pmatrix} \mu\\ \vdash \end{pmatrix}=-~\circ, we have \begin{pmatrix} \mu\\ \dashv \end{pmatrix}=-\circ^{(12)}. The space of the relations of \lie is generated as an \mathbb{S}_3-module by

v_1+v_5+v_9= \mu \circ_{\rmi} \mu+\mu \circ_{\rmii} \mu+\mu \circ_{\rmiii} \mu.

Then we have

\begin{aligned} \mathrm{ADup}_{x}(v_1+v_5+v_9)&= \{z \circ (y \circ x)-y \circ (z \circ x)-(y \circ z)\circ x,z \circ (y \circ x)-y \circ (z \circ x)+(z \circ y) \circ x\};\\ \mathrm{ADup}_{y}(v_1+v_5+v_9)&= \{x \circ (z \circ y)-z \circ (x \circ y)-(z \circ x)\circ y,x \circ (z \circ y)-z \circ (x \circ y)+(x \circ z) \circ y\};\\ \mathrm{ADup}_{z}(v_1+v_5+v_9)&= \{y \circ (x \circ z)-x \circ (y \circ z)-(x \circ y)\circ z,y \circ (x \circ z)-x \circ (y \circ z)+(y \circ x) \circ z\}. \end{aligned}

It is straightforward to check that the relation space \mathrm{ADup}(\lie) is spanned by

\begin{aligned} \mathrm{ADup}(\lie)=\langle x \circ (y \circ z)+(x \circ y)\circ z-y \circ (x \circ z)\rangle, \quad \forall x,y,z \in A. \end{aligned}

Moreover, by a direct and extensive calculation, we deduce that the relation space \comp(\lie) is spanned by

\begin{aligned} \comp(\lie)=\langle (x \circ y) \circ z+(z \circ x) \circ y+ (y \circ z) \circ x, ~(x \circ y+y \circ x)\circ z, ~(x \circ y) \circ z+z \circ (x \circ y) \rangle, \end{aligned}

for all x,y,z \in A.

Now we compute the intersection of subspaces \mathrm{ADup}(\lie) and \comp(\lie). By (Proof), we have

(x\circ y)\circ z+(y\circ x)\circ z=x \circ (y \circ z)+(x \circ y)\circ z-y \circ (x \circ z)+y \circ (x \circ z)+(y \circ x)\circ z-x \circ (y \circ z),

which implies that (x\circ y)\circ z+(y\circ x)\circ z belongs to the intersection of \mathrm{ADup}(\lie) and \comp(\lie). Then we consider the following equation

\begin{aligned} &&a_1 ((x\circ y)\circ z+(z\circ x)\circ y+(y\circ z)\circ x)+a_2 ((y\circ x)\circ z+(z \circ y)\circ x+(x \circ z)\circ y)\\ &&+ a_3 ((x \circ y)\circ z+ z\circ (x \circ y))+a_4 ((y \circ x)\circ z+ z\circ (y \circ x))+a_5 ((z \circ y)\circ x+ x\circ (z \circ y))\\ &&+a_6 ((x \circ z)\circ y+ y \circ (x \circ z))+a_7 ((y \circ z)\circ x+ x \circ (y \circ z))+a_8 ((z \circ x)\circ y+ y \circ (z \circ x))\\ &= b_1 (x \circ (y \circ z)+(x \circ y)\circ z-y \circ (x\circ z))+b_2 (x \circ (z \circ y)+(x \circ z)\circ y-z \circ (x\circ y))\\ &&+b_3 (y \circ (x \circ z)+(y \circ x)\circ z-x \circ (y\circ z))+b_4 (z \circ (y \circ x)+(z \circ y)\circ x-y \circ (z\circ x))\\ &&+b_5 (y \circ (z \circ x)+(y \circ z)\circ x-z \circ (y\circ x))+b_6 (z \circ (x \circ y)+(z \circ x)\circ y-x \circ (z\circ y)). \end{aligned}

Since \mathcal{T}(V)(3) is generated by the twelve elements v_i~(1\leq i \leq 12) in the table (?), we obtain the following system of linear equations in the variables a_m and b_n, where 1 \leq m \leq 8 and 1 \leq n \leq 6,

\begin{array}{c@{\quad}c@{\quad}c} \left\{ \begin{aligned} a_1+a_3 &= b_1, \\ a_1+a_8 &= b_6, \\ a_1+a_7 &= b_5, \\ a_2+a_6 &= b_2, \\ a_2+a_5 &= b_4, \\ a_2+a_4 &= b_3, \end{aligned} \right. & \text{and} & \left\{ \begin{aligned} a_7-b_1+b_3 &= 0, \\ a_3+b_2-b_6 &= 0, \\ a_8+b_4-b_5 &= 0, \\ a_5-b_2+b_6 &= 0, \\ a_4-b_4+b_5 &= 0, \\ a_6+b_1-b_3 &= 0. \end{aligned} \right. \end{array}

Since there are 14 variables and 12 equations, the final general solution is 2-dimensional. Thus the general solution of the above system of linear equations can be expressed as

(a_1,\dots,a_8,b_1,\dots,b_6)^{\mathrm{T}} = s_1\delta_1+s_2\delta_2,

where s_1,s_2\in \mathbf{k} are arbitrary constants, not all zero.

A fundamental system of solutions is given by (\delta_1,\delta_2), where

\begin{aligned} \begin{aligned} \delta_1&=(1,1,0,0,0,0,0,0;\ 1,1,1,1,1,1);\\ \delta_2&=(1,0,-1,1,1,1,-1,-1;\ 0,1,1,1,0,0). \end{aligned} \end{aligned}

Here the coordinate order is (a_1,a_2,a_3,a_4,a_5,a_6,a_7,a_8;b_1,b_2,b_3,b_4,b_5,b_6). The two vectors \delta_1,\delta_2 are linearly independent and hence form a basis of the solution space of the above system.

When we consider the solution \delta_1, we obtain the relation \sum\limits_{\text{cyclic}} (x \circ y+y \circ x)\circ z, which corresponds to Eq. (4.16). Considering the solution \delta_2, we have the relation \sum\limits_{\text{cyclic}} (x \circ (z \circ y)+(x \circ z)\circ y-z \circ (x \circ y)), which corresponds to Eq. (4.16). Therefore, the operad \antileibniz is the anti-duplicator of the operad \lie.

Lemma

Lemma 5.1

T8KV

Let \{f_i\}_{1 \leq i \leq m } be a basis of (R). Then we have

  1. (i) \{\mathrm{ABSu}_x(f_i)~|~x \in \lin(f_i),1 \leq i \leq m \} is a linear spanning set of \mathrm{ABSu}(R);

  2. (ii) \{\st(f_i)~|~1 \leq i \leq m \} is a linear spanning set of \st(R);

  3. (iii) \{\mathrm{ADup}_x(f_i)~|~x \in \lin(f_i),1 \leq i \leq m \} is a linear spanning set of \mathrm{ADup}(R);

  4. (iv) \{\comp(f_i)~|~1 \leq i \leq m \} is a linear spanning set of \comp(R).

Used by Proof
Theorem

Theorem 5.2

J9ZW

Let \opd=\mathcal{T}(\gensp)/ (\relsp) be a binary quadratic operad with \relsp \neq 0. Then

\mathrm{ADup}(\opd^!)=\mathrm{ABSu}(\opd)^!.
Used by Proof, Proof, Proof
Proof of Theorem 5.2

Fix a basis \{e_1,\cdots,e_n\} of V over \mathbf{k}. Let \{f_k\}_{1 \leq k \leq m} be the basis of (\relsp) given by (?).

By (?), we have

\begin{aligned} \mathrm{ABSu}_x(f_k)&=\Biggl\{\sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+ b_{i,j}^k \begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \ast \end{pmatrix}+ c_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \succ \end{pmatrix}\Biggr\},\\ \mathrm{ABSu}_y(f_k)&=\Biggl\{\sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \succ \end{pmatrix}+b_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+c_{i,j}^k \begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \ast \end{pmatrix} \Biggr\},\\ \mathrm{ABSu}_z(f_k)&=\Biggl\{\sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \ast \end{pmatrix}+ b_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \succ \end{pmatrix}+ c_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \prec \end{pmatrix} \Biggr\},\\ \st(f_k)&=\Biggl\{\sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \ast \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \ast \end{pmatrix}+ b_{i,j}^k \begin{pmatrix} e_i \\ \ast \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \ast \end{pmatrix}+ c_{i,j}^k \begin{pmatrix} e_i \\ \ast \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \ast \end{pmatrix} \Biggr\}. \end{aligned}

By (?), we similarly obtain

\begin{aligned} \mathrm{ADup}_x(g_l)&=\Biggl\{ \sum_{i,j} \alpha^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee} \\ \dashv \end{pmatrix}+\beta^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee} \\ \dagger \end{pmatrix}+\gamma^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee} \\ \vdash \end{pmatrix} \Biggr\},\\ \mathrm{ADup}_y(g_l)&=\Biggl\{ \sum_{i,j} \alpha^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee} \\ \vdash \end{pmatrix}+\beta^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee} \\ \dashv \end{pmatrix}+ \gamma^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \vdash \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee} \\ \dagger \end{pmatrix} \Biggr\},\\ \mathrm{ADup}_z(g_l)&=\Biggl\{ \sum_{i,j} \alpha^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \vdash \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee} \\ \dagger \end{pmatrix}+ \beta^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee} \\ \vdash \end{pmatrix}+ \gamma^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee} \\ \dashv \end{pmatrix} \Biggr\}, \\ \comp(g_l)&=\Biggl\{ \sum_{i,j} \alpha^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dagger \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee} \\ \dagger \end{pmatrix}+ \beta^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dagger \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee} \\ \dagger \end{pmatrix}+ \gamma^l_{i,j} \begin{pmatrix} e_i^{\vee} \\ \dagger \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee} \\ \dagger \end{pmatrix} \Biggr\}. \end{aligned}

By Lemma 5.1, we have

\begin{aligned} (\mathrm{ABSu}(R))&= \sum\limits_{k=1}^{m} (\mathbf{k} \mathrm{ABSu}_x(f_k)+\mathbf{k} \mathrm{ABSu}_y(f_k)+\mathbf{k} \mathrm{ABSu}_z(f_k)),\\ (\st(R))&= \sum\limits_{k=1}^{m} \mathbf{k} \st(f_k),\\ (\mathrm{ADup}(R^{\perp})) &= \sum\limits_{l=1}^{3n^2-m} (\mathbf{k} \mathrm{ADup}_x(g_l)+\mathbf{k} \mathrm{ADup}_y(g_l)+\mathbf{k} \mathrm{ADup}_z(g_l)),\\ (\comp(R^{\perp})) &= \sum\limits_{l=1}^{3n^2-m} \mathbf{k} \comp(g_l). \end{aligned}

Since we have

\begin{aligned} \mathrm{ABSu}(\opd)^!&=\Big(\mathcal{T}(\genbi)/ (\mathrm{ABSu}(\relsp)\cup\st(\relsp))\Big)^!=\mathcal{T}(\genbi^{\perp})/ (\mathrm{ABSu}(\relsp)\cup\st(\relsp))^{\perp}\\ &=\mathcal{T}(\genbi^{\perp})/ (\mathrm{ABSu}(\relsp))^{\perp} \cap (\st(\relsp))^{\perp};\\ \mathrm{ADup}(\opd^!)&=\mathcal{T}(\overline{\genbas^{\perp}})/ (\mathrm{ADup}(\relsp^{\perp}))\cap (\comp(\relsp^{\perp})), \end{aligned}

To reach our conclusion, it suffices to show the following equalities under the condition \relsp\neq 0:

  1. (i) \mathcal{T}(\genbi^{\perp}) =\mathcal{T}(\overline{\genbas^{\perp}});

  2. (ii) (\mathrm{ABSu}(\relsp))^{\perp}=(\mathrm{ADup}(\relsp^{\perp}));

  3. (iii) (\st(\relsp))^{\perp}=(\comp(\relsp^{\perp})).

As for (i), we have

\begin{pmatrix} e_i\\ \prec \end{pmatrix}^{\perp}=\begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix},\quad \begin{pmatrix} e_i\\ \succ \end{pmatrix}^{\perp}=\begin{pmatrix} e_i^{\vee} \\ \vdash \end{pmatrix},\quad \forall 1 \leq i \leq n.

That means

\langle \begin{pmatrix} e_i\\ \prec \end{pmatrix}, \begin{pmatrix} e_i^{\vee} \\ \dashv \end{pmatrix} \rangle=0,\quad \langle \begin{pmatrix} e_i\\ \succ \end{pmatrix}, \begin{pmatrix} e_i^{\vee} \\ \vdash \end{pmatrix} \rangle=0,

which implies that \mathcal{T}_{\hspace*{-0.1cm}ns}(\genbi^{\perp}) =\mathcal{T}_{\hspace*{-0.1cm}ns}(\overline{\genbas^{\perp}}).

As for (ii), by (?), for all 1\leq k \leq m and 1\leq l\leq 3n^2-m, we have

\langle \mathrm{ABSu}_p(f_k), \mathrm{ADup}_q(g_l) \rangle=0, \quad \forall p,q\in \{x,y,z\}.

Thus, we have \langle \mathrm{ABSu}(f_k), \mathrm{ADup}(g_l) \rangle=0, which implies that (\mathrm{ABSu}(\relsp))^{\perp}\supseteq (\mathrm{ADup}(\relsp^{\perp})). On the other hand, if

h=\sum_{i,j,u,v} x_{i,j,u,v} \begin{pmatrix} e_i^{\vee}\\ u \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ v \end{pmatrix}+y_{i,j,u,v} \begin{pmatrix} e_i^{\vee}\\ u \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ v \end{pmatrix}+z_{i,j,u,v} \begin{pmatrix} e_i^{\vee}\\ u \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ v \end{pmatrix}

is in (\mathrm{ABSu}(R))^{\perp}, where u,v\in \{\dashv,\vdash\}, then for all 1 \leq k \leq m, we have

\langle h,\mathrm{ABSu}_x(f_k)\rangle=0,\quad \langle h,\mathrm{ABSu}_y(f_k)\rangle=0, \quad \langle h,\mathrm{ABSu}_z(f_k)\rangle=0.

Since R \neq 0, by [PBGN, Proposition 3.1], for any fixed i_0,j_0\in \{1,2,\cdots,n\}, there exists 1 \leq k_0 \leq m, such that b_{i_0,j_0}^{k_0}\neq 0. Then for any k, we deduce that the following relations

\begin{aligned} F_1&:=& \sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+b^{k}_{i_0,j_0}\begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+\sum_{i\neq i_0,j\neq j_0} b_{i,j}^k \begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+\sum_{i,j} c_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \succ \end{pmatrix},\\ F_2&:=& \sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+b^{k}_{i_0,j_0}\begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \prec \end{pmatrix}+\sum_{i\neq i_0,j\neq j_0} b_{i,j}^k \begin{pmatrix} e_i \\ \succ \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \succ \end{pmatrix}+\sum_{i,j} c_{i,j}^k \begin{pmatrix} e_i \\ \prec \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \succ \end{pmatrix}, \end{aligned}

are in \mathrm{ABSu}_x(f_k). Thus for all 1 \leq k \leq m, we obtain

\begin{aligned} \sum_{i,j} a^k_{i,j} x_{i,j,\dashv,\dashv}+b^k_{i_0,j_0} y_{i_0,j_0,\vdash,\dashv}+\sum_{i\neq i_0,j\neq j_0} b^k_{i,j} y_{i,j,\vdash,\dashv}+\sum_{i,j} c^k_{i,j} z_{i,j,\dashv,\vdash}&=\langle h,F_1 \rangle=0;\\ \sum_{i,j} a^k_{i,j} x_{i,j,\dashv,\dashv}+b^k_{i_0,j_0} y_{i_0,j_0,\vdash,\vdash}+\sum_{i\neq i_0,j\neq j_0} b^k_{i,j} y_{i,j,\vdash,\dashv}+\sum_{i,j} c^k_{i,j} z_{i,j,\dashv,\vdash}&=\langle h,F_2 \rangle=0. \end{aligned}

Since b^k_{i_0,j_0} \neq 0, we have y_{i_0,j_0,\vdash,\dashv}=y_{i_0,j_0,\vdash,\vdash} for all 1\leq i_0,j_0 \leq n. By (?), we also have

(x_{i,j,\dashv,\dashv}, y_{i,j,\vdash,\dashv}, z_{i,j,\dashv,\vdash})=\sum\limits_{l=1}^{3n^2-m} d_l(\alpha_{i,j}^l,\beta_{i,j}^l,\gamma_{i,j}^l),

for some d_l \in \mathbf{k}. Thus we obtain

\begin{aligned} h_x &:=& \sum_{i,j} x_{i,j,\dashv,\dashv} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+y_{i,j,\vdash,\dashv} \begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+y_{i,j,\vdash,\vdash} \begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}+z_{i,j,\dashv,\vdash} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix} \\ &= \sum_{i,j} x_{i,j,\dashv,\dashv} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+y_{i,j,\vdash,\dashv} \Big(\begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+\begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}\Big)+z_{i,j,\dashv,\vdash} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix} \\ &= \sum_{l=1}^{3n^2-m} d_l \Bigg(\sum_{i,j} \alpha^l_{i,j} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+\beta^l_{i,j} \Big(\begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+\begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}\Big)+\gamma^l_{i,j} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}\Bigg), \end{aligned}

which is in \sum_{l=1}^{3n^2-m} \mathbf{k} \mathrm{ADup}_x(g_l). Similarly, we also find that

\begin{aligned} h_y := \sum_{i,j} x_{i,j,\dashv,\vdash} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}+y_{i,j,\dashv,\dashv} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+z_{i,j,\vdash,\dashv} \begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+z_{i,j,\vdash,\vdash} \begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}, \end{aligned}

is in \sum_{l=1}^{3n^2-m} \mathbf{k} \mathrm{ADup}_y(g_l) and

\begin{aligned} h_z := \sum_{i,j} x_{i,j,\vdash,\dashv} \begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}+x_{i,j,\vdash,\vdash} \begin{pmatrix} e_i^{\vee}\\ \vdash \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}+y_{i,j,\dashv,\vdash} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \vdash \end{pmatrix}+z_{i,j,\dashv,\dashv} \begin{pmatrix} e_i^{\vee}\\ \dashv \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ \dashv \end{pmatrix}, \end{aligned}

is in \sum_{l=1}^{3n^2-m} \mathbf{k} \mathrm{ADup}_z(g_l). Note that h=h_x+h_y+h_z. Thus we deduce that h is in \sum_l \mathbf{k} \mathrm{ADup}_x(g_l)+\mathbf{k} \mathrm{ADup}_y(g_l)+\mathbf{k} \mathrm{ADup}_z(g_l), which implies that h is in the \mathbb{S}-module generated by \mathrm{ADup}(R^{\perp}). Thus we prove that (\mathrm{ABSu}(\relsp))^{\perp}= (\mathrm{ADup}(\relsp^{\perp})).

As for (iii), by (?), for all 1\leq k \leq m and 1\leq l\leq 3n^2-m, we have \langle \st(f_k), \comp(g_l) \rangle=0, which implies that (\st(\relsp))^{\perp}\supseteq (\comp(\relsp^{\perp})). On the other hand, if

h'=\sum_{i,j,u,v} x'_{i,j,u,v} \begin{pmatrix} e_i^{\vee}\\ u \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ v \end{pmatrix}+y'_{i,j,u,v} \begin{pmatrix} e_i^{\vee}\\ u \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ v \end{pmatrix}+z'_{i,j,u,v} \begin{pmatrix} e_i^{\vee}\\ u \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ v \end{pmatrix}

is in (\st(R))^{\perp}, where u,v\in \{\dashv,\vdash\}, then for all 1 \leq k \leq m, we have \langle h',\st(f_k)\rangle=0. For any k,

\begin{aligned} G &:=& \sum_{i,j} a_{i,j}^k \begin{pmatrix} e_i \\ \ast \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j \\ \ast \end{pmatrix}+b_{i,j}^{k}\begin{pmatrix} e_i \\ \ast \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j \\ \ast \end{pmatrix}+c_{i,j}^k \begin{pmatrix} e_i \\ \ast \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j \\ \ast \end{pmatrix} \end{aligned}

is in \st(f_k). Thus for all 1 \leq k \leq m, we obtain

\begin{aligned} \sum_{i,j,u,v} a^k_{i,j} x'_{i,j,u,v}+b^k_{i,j} y'_{i,j,u,v}+c^k_{i,j} z'_{i,j,u,v}=\langle h',G \rangle=0. \end{aligned}

By (?), we also have

(\sum_{u,v} x'_{i,j,u,v}, \sum_{u,v} y'_{i,j,u,v}, \sum_{u,v} z'_{i,j,u,v})=\sum\limits_{l=1}^{3n^2-m} d'_l(\alpha_{i,j}^l,\beta_{i,j}^l,\gamma_{i,j}^l),

for some d'_l \in \mathbf{k}. Thus we obtain

\begin{aligned} h'= \sum_{l=1}^{3n^2-m} d'_l \Bigg(\sum_{i,j} \alpha^l_{i,j} \begin{pmatrix} e_i^{\vee}\\ \dagger \end{pmatrix} \circ_{\rmi} \begin{pmatrix} e_j^{\vee}\\ \dagger \end{pmatrix}+\beta^l_{i,j} \begin{pmatrix} e_i^{\vee}\\ \dagger \end{pmatrix} \circ_{\rmii} \begin{pmatrix} e_j^{\vee}\\ \dagger \end{pmatrix}+\gamma^l_{i,j} \begin{pmatrix} e_i^{\vee}\\ \dagger \end{pmatrix} \circ_{\rmiii} \begin{pmatrix} e_j^{\vee}\\ \dagger \end{pmatrix}\Bigg), \end{aligned}

which implies that h is in the \mathbb{S}-module generated by \comp(\relsp^{\perp}). Thus we have (\st(\relsp))^{\perp}= (\comp(\relsp^{\perp})). Therefore, we have \mathrm{ADup}(\opd^!)=\mathrm{ABSu}(\opd)^!.

Theorem

Theorem 5.3

KJMO

Let \opd=\mathcal{T}(\gensp)/ (\relsp) be a binary quadratic operad with \relsp \neq 0. Then

\mathrm{ATrip}(\opd^!)=\mathrm{ATSu}(\opd)^!.
Proof of Theorem 5.3

The proof is similar to the proof of Theorem 5.2.

Corollary

Corollary 5.4

5VAK

The operad \antiperm is the Koszul dual of the operad \antiprelie.

Used by Proof
Proof of Corollary 5.4

By Theorem 5.2 and Proposition 2.20, with \opd:=\lie, we have

\antiperm=\mathrm{ADup}(\com)=\mathrm{ADup}(\lie^!)=\mathrm{ABSu}(\lie)^{!}=\antiprelie^{!},

which implies that the operad \antiperm is the Koszul dual of the operad \antiprelie.

Corollary

Corollary 5.5

GKKG

The operad \antileibniz is the Koszul dual of the operad \antizinbiel.

Proof of Corollary 5.5

By Theorem 5.2 and Proposition 2.23, with \opd:=\com, we have

\antileibniz=\mathrm{ADup}(\lie)=\mathrm{ADup}(\com^!)=\mathrm{ABSu}(\com)^{!}=\antizinbiel^{!},

which implies that the operad \antileibniz is the Koszul dual of the operad \antizinbiel.

Definition

Definition 5.6

3XYM

Let \opd=\mathcal{T}(\gensp)/ (\relsp) and \oqd=\mathcal{T}(W)/(S) be two binary quadratic operads with finite-dimensional generating spaces. The Manin white product of \opd and \oqd is defined by

\opd \bigcirc \oqd :=\mathcal{T}(\gensp\otimes W)/ (\Phi^{-1}(R\otimes \mathcal{T}(W)+\mathcal{T}(\gensp)\otimes S)).
Theorem

Theorem 5.7

AF9B

Let \opd=\mathcal{T}(\gensp)/ (\relsp) be a binary quadratic operad with \relsp\neq 0. Then we have the isomorphism of operads

\mathrm{ADup}(\opd)\cong \antiperm \bigcirc \opd.
Proof of Theorem 5.7

By Theorem 3.3, Theorem 5.2 and Corollary 5.4, we have

\mathrm{ADup}(\opd)=\mathrm{ABSu}(\opd^!)^! \cong (\antiprelie \bullet \opd^!)^! \cong \antiprelie^! \bigcirc (\opd^!)^! \cong \antiperm \bigcirc \opd.

Here we use the fact that (\opd \bullet \oqd)^! \cong \opd^! \bigcirc \oqd^! [LV, Theorem 8.8.8]. The proof is finished.

Corollary

Corollary 5.8

LA4Z

The Manin white product of the operad \antiperm and the operad \lie is isomorphic to the operad \antileibniz

\antiperm \bigcirc \lie \cong \antileibniz.

Source and licence

The text on this page is from arXiv:2610.21365v1 by Chengming Bai, Li Guo, Yunhe Sheng and You Wang, under the licence CC BY 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY 4.0.

The record of each item is in analysis/objects/2610/2610.21365.jsonl.