Algebras defined by Lyndon words and Artin-Schelter regularity
Tatiana Gateva-Ivanova

TL;DR
This paper explores classes of graded algebras defined by Lyndon words, characterizing when they are Artin-Schelter regular, and classifies certain low-dimensional cases using combinatorial Lyndon pair conditions.
Contribution
It introduces a combinatorial framework using Lyndon words to identify Artin-Schelter regular algebras and classifies specific low-dimensional enveloping algebras.
Findings
Identifies Lyndon pair conditions for Artin-Schelter regularity.
Classifies two-generated Artin-Schelter regular algebras of dimensions 6 and 7.
Provides explicit relations for these classified algebras.
Abstract
Let be a finite alphabet, and let be a field. We study classes of graded -algebras , generated by and with a fixed set of obstructions . Initially we do not impose restrictions on and investigate the case when all algebras in have polynomial growth and finite global dimension . Next we consider classes of algebras whose sets of obstructions are antichains of Lyndon words. The central question is "when a class contains Artin-Schelter regular algebras?" Each class defines a Lyndon pair which determines uniquely the global dimension, , and the Gelfand-Kirillov dimension, , for every . We find a combinatorial condition in terms of , so that…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAlgebraic structures and combinatorial models · Advanced Topics in Algebra · Homotopy and Cohomology in Algebraic Topology
Algebras defined by Lyndon words and Artin-Schelter regularity
Tatiana Gateva-Ivanova
Max Planck Institute for Mathematics
Vivatsgasse 7
53111 Bonn
Germany
and
American University in Bulgaria
2700, Blagoevgrad
Bulgaria
Abstract.
Let be a finite alphabet, and let be a field. We study classes of graded -algebras , generated by and with a fixed set of obstructions . Initially we do not impose restrictions on and investigate the case when the algebras in have polynomial growth and finite global dimension . Next we consider classes of algebras whose sets of obstructions are antichains of Lyndon words. The central question is ”when a class contains Artin-Schelter regular algebras?” Each class defines a Lyndon pair , which, if is finite, determines uniquely the global dimension, , and the Gelfand-Kirillov dimension, , for every . We find a combinatorial condition in terms of , so that the class contains the enveloping algebra , of a Lie algebra . We introduce monomial Lie algebras defined by Lyndon words, and prove results on Gröbner-Shirshov bases of Lie ideals generated by Lyndon-Lie monomials. Finally we classify all two-generated Artin-Schelter regular algebras of global dimension and occurring as enveloping of standard monomial Lie algebras. The classification is made in terms of their Lyndon pairs , each of which determines also the explicit relations of .
Key words and phrases:
Associative algebras, Artin-Schelter regular algebras, Polynomial growth, Global dimension, Lyndon words, Graded Lie algebras, Universal enveloping algebras, Finite presentability, normal forms (diamond lemma, term-rewriting)
1991 Mathematics Subject Classification:
Primary 16E65, 16S38, 16S30, 16S15, 16S37, 16P90, 17B30, 17B35, 17B70
The author was partially supported by the Max Planck Institute for Mathematics (MPIM), Bonn, by the Max Planck Institute for Mathematics in the Sciences, MiS, Leipzig, and by Grant KP-06 N 32/1 of 07.12.2019 of the Bulgarian National Science Fund.
Contents
- 1 Introduction
- 2 Graded algebras with polynomial growth and finite global dimension
- 3 Lyndon words, Lyndon atoms, and Lyndon pairs
- 4 Some combinatorial results on Lyndon pairs
- 5 Classes containing the enveloping algebra of a Lie algebra
- 6 Monomial Lie algebras defined by Lyndon words
- 7 Some classical monomial Lie algebra defined by Lyndon words
- 8 More on Gröbner-Shirshov bases of monomial Lie ideals. Applications
- 9 Two-generated AS-regular algebras of global dimensions and , occurring as enveloping of standard monomial Lie algebras
1. Introduction
Let be a finite alphabet. We denote by the free monoid generated by , and by -the free semigroup generated by , , (the empty word is denoted by ). Throughout the paper will denote the free associative -algebra generated by , where is a field. As usual, the length of a word is denoted by . We shall consider the canonical grading by length on , assuming each has degree .
In the paper we shall use three distinct orderings on (denoted by distinct notation) which will be introduced gradually and used in different contexts.
The first order is the so called ”divisibility order”, ””. This is a partial ordering on the set defined as: iff is a proper subword (segment) of , i.e. , , , or is possible. In the case when , is called a proper left factor (segment) of . Proper right factors (segments) are defined analogously.
Let be a set of monomials in , . A monomial is a minimal element of with respect to the ordering ”” if it does not contain as a proper subword any monomial . Clearly, for any there exists a unique minimal element , with
A nonempty subset is called an antichain of monomials (or shortly an antichain) if any two elements are not comparable with respect to “”. In other words, is an antichain of monomials if every monomial is minimal with respect to “”.
The following is straightforward.
Lemma 1.1**.**
Let be a nonempty set of monomials in . Let
[TABLE]
Then is the unique maximal antichain of monomials in such that every has a subword .
Let , and be as above. A monomial is -normal (-standard) if does not contain as a subword any . Clearly is -normal iff is -normal. Denote by the set of -normal words
[TABLE]
The set is closed under taking subwords, Anick calls such a set an order ideal of monomials, or o.i.m., [3].
Conversely, each nonempty order ideal of monomials determines uniquely an antichain of monomials , so that . Indeed, the complement has a unique maximal antichain of monomials , and the following implications are in force.
[TABLE]
In this case is called the set of obstructions for , [3]. Note that satisfies the following condition
[TABLE]
The natural duality between and is discussed with more details in Section 3. We recall briefly some facts from (noncommutative) Gröbner bases theory. By convention throughout the paper, we consider the degree-lexicographic ordering on , extending the (inverse) ordering
[TABLE]
on . Then is a well-ordered monoid that is ”” is a total ordering on compatible with the multiplication and such that every non-empty subset of has a least element. In all cases when Gröbner basis of an ideal in is considered we shall use the degree-lexicographic ordering on , extending (1.1).
Suppose is a nonzero polynomial, then its highest monomial w.r.t. will be denoted by . One has , where , . Given a set of noncommutative polynomials, denotes the set
[TABLE]
The free associative algebra is naturally graded by length,
[TABLE]
Let be a two sided graded ideal in . We shall consider graded algebras, with a minimal presentation, see Notation-convention 1.5, so without loss of generality we may assume that *is generated by homogeneous polynomials of degree *, hence , with . Then the quotient algebra is finitely generated and inherits its grading from . We shall work with the so called normal -basis of .
A monomial is said to be *normal mod I * iff it is -normal. The set of normal modulo monomials coincides with .
Remark 1.2*.*
Let be the maximal antichain of monomials in , then is normal (mod I) iff is -normal. Hence is the obstruction set for and there are equalities of sets
[TABLE]
In particular, the free monoid splits as a disjoint union
[TABLE]
The free associative algebra splits into a direct sum of - subspaces , and there are isomorphisms of vector spaces
[TABLE]
where is the set of all normal words of length .
Definition 1.3**.**
Given , and as above we shall also refer to as the set of obstructions for , [20]. The monomial algebra is called the monomial algebra associated to .
Definition 1.4**.**
A subset of monic polynomials is a Gröbner basis of (with respect to ) if (1) generates as a two-sided ideal, and (2) There is an equality of sets of normal words .
A Gröbner basis of is called reduced if each is reduced modulo , that is is a linear combination of normal modulo monomials.
It is known that each ideal of has a uniquely determined reduced Gröbner basis (with respect to ). However, may be infinite.
Moreover, the set of obstructions coincides with the set of highest monomials of the reduced Gröbner basis , that is
[TABLE]
So, the reduced Gröbner basis is finite if and only if the set of obstructions is finite. In this case the algebra is called a standard finitely presented algebra, or shortly an s.f.p. algebra.
In [20] we initiate the study of algebraic and homological properties of graded associative algebras for which the set of obstructions consists of Lyndon words. Lyndon words and Lyndon-Shirshov bases are widely used in the context of Lie algebras and their enveloping algebras, and also for PI algebras (see for example the celebrated Shirshov theorem of heights, [9]). It will be interesting to explore the remarkable combinatorial properties of Lyndon words in a more general context of graded associative algebras.
Notation-Convention 1.5**.**
We fix an enumeration on the set of generators, , . By we shall denote the class of all graded -algebras , with a finite set of generators and a set of obstructions of arbitrary cardinality. will denote the set of normal words (mod ). is a -basis for each algebra . By convention we consider only minimal presentations of , so
[TABLE]
Without loss of generality we shall always assume that *is generated by a set of homogeneous polynomials of degree *. This implies that , the unique reduced Gröbner basis of , (w.r.t. the degree-lexicographic ordering on , extending (1.1)), is also a set of homogeneous polynomials of degree . Each class contains unique monomial algebra . Clearly, all algebras share the same associated monomial algebra is , and the same normal -basis . There are equalities of sets
[TABLE]
As a consequence, all have the same Gelfand-Kirillov dimension, , and the same Hilbert series . When we work with Lyndon words we often consider a third order, ”” on the set , namely, the pure lexicographic order which extends the ordering on as follows: for any *iff * either is a proper left factor of (), or , where .
We study classes such that every algebra has polynomial growth and finite global dimension. Moreover, starting from Section 3, throughout the paper we are interested in the special case when the obstructions set consists of Lyndon words, see Definition 3.1. One of the central questions in our study is: ”When a given class contains an Artin-Schelter regular algebra?” .
Remark 1.6*.*
The problem of studying classes , as above was posed first in [20] and in several talks given by the author (Bedlewo, 2013, [19], ICTP 2013). On these talks we reported the results of Theorem II, and Theorem III, (2).
We recall the definition of a regular algebra in the sense of Artin and Schelter, see [4], p.171.
Let be a finitely presented graded algebra over a field . The algebra is called Artin-Schelter regular (or AS regular) if
- (i)
has finite global dimension . 2. (ii)
has finite Gelfand-Kirillov dimension, i.e. has polynomial growth. 3. (iii)
is Gorenstein, meaning that if and .
AS regular algebras were introduced and studied first in [4, 5, 6]. Since then AS regular algebras and their geometry have intensively been studied. When all regular algebras are classified. The problem of classification of regular algebras seems to be difficult and remains open even for regular algebras of global dimension , see for example [27]. The study of Artin-Schelter regular algebras, their classification, and finding new classes of such algebras is one of the basic problems in noncommutative geometry. Numerous works on this topic appeared during the last two decades, see for example [5, 6], [25, 37, 28, 14, 18, 38], and references therein.
The main results of the paper are Theorems I, II, III and IV. Theorem IV is formulated and proven in Section 9. Here we state Theorems I, II and III proven under the following hypothesis:
Hypothesis 1.7**.**
In Notation-convention 1.5. Suppose , and is an antichain of words of arbitrary cardinality. Denote by the class of all graded -algebras , generated by whose set of obstructions is (with respect to the degree-lexicographic ordering on , where ). is the monomial algebra determined by , . Let be the set of normal words modulo , and let be the set of normal Lyndon words.
Theorem I**.**
Assumptions as in Hypothesis 1.7. Suppose there are no -chains on , but there exists a -chain on . Let be an arbitrary algebra in , assume that has polynomial growth. Then the following conditions hold.
- (1)
There are equalities
[TABLE] 2. (2)
There exists a finite set of normal words, called ”atoms”, in the sense of Anick, **[2]**: , such that:
- (a)
The normal -basis of (and ) can be expressed in terms of Anick’s atoms as ”a Poincaré-Birkhoff-Witt” basis
[TABLE] 2. (b)
The Hilbert series of satisfies
[TABLE] 3. (c)
Any word can be written as a product . 3. (3)
* is standard finitely presented- the ideal has a finite reduced Gröbner basis , with*
[TABLE] 4. (4)
The following conditions are equivalent.
- (a)
. 2. (b)
There is no atom whose length is . 3. (c)
* and there exists a, possibly new, ordering of :*
[TABLE] 4. (d)
. 5. (e)
* is a PBW algebra in the sense of Priddy (see [32, 31]), equivalently, the reduced Gröbner basis of consists of quadratic homogeneous polynomials.* 6. (f)
. 5. (5)
Each of the equivalent conditions in part (4) implies that is Koszul.
The proof of the Theorem is given in Subsection 2.2
Corollary 1.8**.**
Given a class , where is an arbitrary chain of monomials in , suppose the monomial algebra has global dimension and polynomial growth. Then every algebra is standard finitely presented and .
Theorem II**.**
Assumptions as in Hypothesis 1.7. Let be an antichain of arbitrary monomials.
- (1)
* is an anticain of Lyndon words if and only if the set of normal words has the shape*
[TABLE]
where the set of normal Lyndon words is not necessarily finite.
Suppose that is an antichain of Lyndon words, so is a Lyndon pair, see Definition 3.3. Let . 2. (2)
* has polynomial growth of degree if and only if has order , so . In this case the following conditions hold.*
- (a)
; 2. (b)
* has a Poincaré-Birkhoff-Witt type -basis*
[TABLE] 3. (c)
The algebra is standard finitely presented, more precisely, the reduced Gröbner basis of consists of exactly polynomials, where
[TABLE] 3. (3)
The class contains Artin-Schelter regular algebras, whenever
[TABLE]
The theorem is proven in Section 4.
Theorem III**.**
Suppose the class is defined by an antichain of Lyndon words , where the set of Lyndon atoms has a finite order .
- (1)
The following conditions are equivalent:
- (a)
* and is a connected set of Lyndon atoms, see Def 5.9;* 2. (b)
, and (up to isomorphism of monomial algebras ) the set is determined uniquely:
[TABLE] 3. (c)
, and (up to isomorphism of monomial algebras ) the set is determined uniquely:
[TABLE]
In this case the class contains the universal enveloping U=U\mbox{\mathcal{L}}_{d-1}, where \mbox{\mathcal{L}}_{d-1} is the standard filiform algebra of dimension and nilpotency class . is an AS-regular algebra with . 2. (2)
The following conditions are equivalent:
- (a)
; 2. (b)
, and ; 3. (c)
.
In this case the class contains an abundance of (non isomorphic) Artin-Schelter regular PBW algebras of global dimension , each of them is presented as a skew polynomial ring with square-free binomial relations, (in the sense of **[17]**) and defines a solution of the Yang-Baxter equation.
The theorem is proven in subsection 7.3.
The paper is organized as follows. In Section 2 we study graded algebras with polynomial growth and finite global dimension and prove Theorem I. In Section 3 we recall the notion of Lyndon words and prove some new results. We introduce the notion of Lyndon pairs, which are central for our theory. In Section 4 we prove Theorem II, and Theorem 4.2 which is a purely combinatorial result on Lyndon pairs. In Section 5 we give some basic information on classes containing the enveloping algebra of a Lie algebra (possibly infinite dimensional). We show that for every such a class the set of obstruction is an antichain of Lyndon words, and the corresponding set of Lyndon atoms must be connected, see Definition 5.9. and Proposition 5.11. In Section 6 we introduce monomial Lie algebras defined by Lyndon words -these algebras and their enveloping algebras are central for the paper. A natural problem arises: ”Classify Lyndon pairs such that the class contains an Artin-Schelter regular algebra occurring as an enveloping of a concrete monomial Lie algebra ”. This motivates our study of Gröbner-Shirshov bases for monomial Lie ideals. In this case we find natural combinatorial conditions in terms of the Lyndon pair . Propositions 6.15 and 6.17, give purely combinatorial conditions which in various cases help to decide easily and without computations, whether a set of bracketed monomials is a Gröbner-Shirshov basis of the Lie ideal of the free Lie algebra . In Section 7 we present some classical Lie algebras as standard monomial Lie algebras, , where is a finite Gröbner-Shirshov basis of the ideal . Our first applications of the general results on GS bases are Corollary 7.3 and Theorem 7.9 which show that the free nilpotent Lie algebra of nilpotency class and the Filiform Lie algebras \mbox{\mathcal{L}}_{m} and \mbox{\mathcal{Q}}_{m} are standard multigraded monomial Lie algebras. In Section 8 we introduce the notions of a regular Lyndon pair, a standard Lyndon pair, and a nontrivial disconnected extension of a standard Lyndon pair, which are essential for our classification. Important general results are Proposition 8.5, and especially Proposition 8.8. In Section 9 we prove one of the main results of the paper, Theorem IV, which gives a classification of the two-generated Artin-Schelter regular algebras of global dimensions and , occurring as enveloping of standard monomial Lie algebras. Our general results on Lyndon pairs from the previous sections are crucial for the proof.
2. Graded algebras with polynomial growth and finite global dimension
2.1. Anick’s results on global dimension and growth of augmented graded algebras
Given a finitely generated augmented graded algebra with a set of obstructions , Anick introduces the notion of an -chain on and uses the -chains to construct a resolution of the field considered as an -module. He obtains important results on augmented graded algebras with finite global dimension, see [3], and [2]. It is known that Anick’s resolution is minimal in the cases when (a) is a monomial algebra, that is and (b) is a PBW algebra, that is the ideal is generated by quadratic homogeneous polynomials which form a Gröbner basis w.r.t. degree-lexicographic ordering on . We recall now some definitions and results from [2] and [3].
By Notation-Convention 1.5 throughout the paper we assume
[TABLE]
Definition 2.1**.**
[2] The set of -chains on is defined recursively. A -chain is the monomial , a [math]-chain is any element of , and a -chain is a word in . An (n+1)-prechain is a word , which can be factored in two different ways such that , is an chain, is an -chain, and is a proper left segment of . An -prechain is an -chain if no proper left segment of it is an -prechain. In this case the monomial is called the tail of the -chain .
In fact the global dimension of a finitely presented monomial algebra can be effectively computed using some combinatorial properties of its defining relations, as shows the following corollary extracted from [2], Theorem 4.
Corollary 2.2**.**
[2]** Suppose is an antichain of monomials. The monomial algebra has global dimension iff there are no -chains on but there exists a chain on .
2.2. Connected graded algebras with polynomial growth and finite global dimension (the general case)
In this section we shall study finitely generated graded algebras with polynomial growth and finite global dimension. We prove Theorem I assuming Hypothesis 1.7. Note that the set of obstructions given in the hypothesis is an arbitrary antichain of monomials in , no restrictions like ” consists of Lyndon words”, or ” is finite” have been imposed.
Proof of Theorem I**.**
The algebra and its associated monomial algebra share the same set of obstructions , hence the same -normal basis, , so , and therefore has polynomial growth. Moreover, by hypothesis there are no -chains on , but there exists a -chain on , hence, by Corollary 2.2 . Now a result of the author, [15], Theorem II, implies that there are equalities
[TABLE]
which proves part (1). It is clear, that does not contain a free-subalgebra on two monomials (since has polynomial growth) and therefore the monomial algebra satisfies the hypothesis of [2], Theorem 5. It follows then that there exists a finite, ordered set consisting of atoms in the sense of Anick, where is a linear ordering on , and satisfies the good properties listed in [2], p. 297, which straightforwardly imply (2) (a), (b), (c). By (c) each obstruction can be presented in terms of atoms as where and since there are such pairs , one has .
We know that the reduced Gröbner basis of satisfies , where is the set of highest monomials of the elements of . Moreover, any two elements satisfy . Hence , and therefore is standard finitely presented. This proves part (3).
(4). We recall some details from Anick’s results, [2]. A partial relation on the set of normal words, is introduced on p. 297, [2], and the set of atoms is defined recursively as follows , where , and given , is defined by
[TABLE]
It is proven that the set of atoms satisfy several properties, among which the following two.
* (Exclusivity): for exactly one of the relations and is valid.*
* (The relations in ): Any word , with is a ”would-be-atom”, i.e. for some , with *
By part (2) the set thus , and we use the restriction of the (new) linear ordering on to ”rename” the elements of . so that
[TABLE]
(4a) (4b). Assume . It follows from , and , that for each pair with (equivalently, ) one has (moreover, , whenever ). In particular, the set
[TABLE]
Note that the inclusion (2.2) is in force if and only if there is an equality
[TABLE]
Moreover, (2.3) is equivalent to (4b). Indeed, by the recursive definition, , and , so there are no atoms of length iff .
(4b) (4c). Assume (4b). Then , hence , and (2.2) is in force. Using induction on one verifies that , for all , and therefore . Moreover which proves (4c).
(4c) (4d). Assume (4c). Then is a quadratic monomial algebra with , and , hence, , and .
(4d) (4e). Assume . Part (1) implies , by part (2) , hence . Now , and , imply . It follows that the set of obstructions satisfies (1.9). In particular, the reduced Gröbner basis of consists of homogeneous polynomials of degree , so is a PBW algebra in the sense of Priddy, [32].
(4e) (4c). Assume is a PBW algebra in the sense of Priddy. Then consists of monomials of length , and is a quadratic monomial algebra. By the hypothesis of the theorem has polynomial growth and finite global dimension. Hence does not contain squares. By [18], Theorem 3.7, p 2163 there is a permutation of , such that has the shape (1.9).
The implications (4c) (4f) and (4c) (4a) are clear.
(4f) (4d). Suppose . Then has polynomial growth of degree , hence, by part (1), . This proves part (4).
It is known that every (quadratic) PBW algebra in the sense of Priddy, is Koszul, see [32], and also [31], Theorem 3.1, p 84. This proves part (5). The theorem has been proved.
Corollary 2.3**.**
Let be an Artin-Shelter regular algebra of global dimension , and with a set of obstructions . Suppose that there are no -chain on . Then conditions (1) through (5) of Theorem I, Theorem I are satisfied. In particular, (i) there are equalities and ; (ii) is standard finitely presented and ; (iii) Moreover, if , then , and is Koszul.
Recall the following conjecture of Anick, [2], p.301.
Conjecture 2.4**.**
Suppose is an augmented connected graded -algebra with polynomial growth and with global dimension Then the Hilbert series of is given by (1.7) for some positive integers .
Anick states in [2] that the conjecture has been verified for the cases when is commutative, is an enveloping algebra of a Lie algebra, when is a monomial algebra, and also in the case when is is a Noetherian PI ring (an unpublished result of Lorenz). Our results verify Anick’s conjecture in the case when is an algebra whose associated monomial algebra has finite global dimension.
3. Lyndon words, Lyndon atoms, and Lyndon pairs
3.1. Lyndon words and Lyndon atoms
We recall first the notion of Lyndon words and some of their basic properties (we refer to [29]). Consider the pure lexicographic ordering ”” on the set which extends the ordering on as follows: for any *iff * either is a proper left factor of (), or , where .
Recall that
[TABLE]
So is a linear ordering on the set compatible with the left multiplication in . However, in the case when , the right multiplication does not necessarily preserve inequalities. (For example, ). Furthermore, the decreasing chain condition is not satisfied on , for if , one has .
The following simple relation between the two (fixed) orderings and on the free semigroup is in force.
[TABLE]
For example, if , one has , but .
Definition-Notation 3.1**.**
[29] A nonperiodic word is a Lyndon word if it is minimal, with respect to , in its conjugacy class. In other words, implies . The set of Lyndon words in will be denoted by , denotes the set of all Lyndon words of length .
We shall use notation, terminology and some results from [20].
Definition-Notation 3.2**.**
[20] Given an antichain of Lyndon words, the set of all -normal Lyndon words is denoted by , we shall refer to it as the set of Lyndon atoms corresponding to . By definition the set of Lyndon atoms satisfies
[TABLE]
Assume that is a graded -algebra such that its set of obstructions consists of Lyndon monomials, then we shall also refer to the normal Lyndon words as the set of Lyndon atoms for .
Definition 3.3**.**
Let be an antichain of Lyndon words, and be the corresponding set of Lyndon atoms. Then the pair of sets will be called a Lyndon pair.
The canonical duality between the notions ”an antichain of Lyndon words” and ”its set of Lyndon atoms, ” is discussed in the next subsection, see also [20].
Fact 3.4**.**
(Lyndon’s Theorem)**, [30] 11.5. Every word can be written uniquely as a non increasing product of Lyndon words, .
The following useful facts are extracted from [29], Section 5.1..
Facts 3.5**.**
- (1)
For all , the equality , with , implies 2. (2)
If are Lyndon words then is a Lyndon word, so . 3. (3)
If is the longest proper right segment of which is a Lyndon word, then , where is a Lyndon word. This is called the (right) standard factorization of and denoted as . 4. (4)
Analogously, suppose is the longest proper left segment of which is a Lyndon word, then , where is also a Lyndon word. This determines the left standard factorisation of denoted by
Notation 3.6**.**
[20] For the monomials we write if , , where , ( or is possible). We write if and overlaps with and so that and .
Lemma 3.7**.**
[20]**
- (1)
Let and be Lyndon words, . Then is a Lyndon word. 2. (2)
If and , then . 3. (3)
Suppose that , with . Then . 4. (4)
Let be Lyndon words. Then are Lyndon words for all . 5. (5)
If is the right standard factorization of , then the (right) standard factorization of is . Analogously, if is the left standard factorization of the Lyndon word then the left standard factorization of is .
Lemma 3.8**.**
[20]** Let be Lyndon words, . If a Lyndon word is a subword of , then is a subword of , for some ,
3.2. More results on Lyndon words
We shall prove some additional results on Lyndon words which will be used in the paper.
Lemma 3.9**.**
Suppose , and is a proper subword of .
- (1)
If is the right standard factorization of , and is not a left segment of , then either , or 2. (2)
If is the left standard factorization of , and is not a right segment of , then either or .
Proof.
We sketch a proof of (1). Assume the contrary. Then overlaps with , so that , , where , and is a proper segment of . By Lemma 3.7 (2) the product is a Lyndon word, moreover, is a proper right segment of with . By assumption is the longest Lyndon word which occurs as a right segment of , a contradiction. The proof of part (2) is analogous. ∎
Proposition 3.10**.**
Let be a set of Lyndon words closed under taking Lyndon subwords, let be the corresponding antichain of Lyndon words, see Remark 3.13 (2).
- (1)
If and no Lyndon atom satisfies , then . 2. (2)
Suppose is a discrete convex in , that is does not contain a word with . Then for every pair . 3. (3)
If is a finite set of order , then .
Proof.
(1). By hypothesis there is no Lyndon atom such that . By Facts 3.5 and , hence . We have to show that Assume the contrary, then (since ) some , is a proper subword of . Let , be the standard factorization of . By Remark 3.11 (1), and by Facts 3.5 . Three cases are possible: (i) overlaps with and , so that ; (ii) overlaps with and , so that ; (iii) , , . Assume (i) is in force. Then Lemma 3.7 (3) implies , which contradicts the hypothesis. Similarly, in case (ii) the relation implies , which is impossible. Suppose (iii) holds. Without loss of generality we may assume that is a proper right segment of , so . Now the inequalities
[TABLE]
where contradict the hypothesis. The case when is a proper left segment of is analogous, which proves (1). Part (2) follows straightforwardly from (1). Clearly, (2) implies (3). ∎
Remark 3.11*.*
[20] Let be a Lyndon pair. Then
- (1)
If is a proper Lyndon subword of some then . In particular, if then the (right) standard factorisation and the left standard factorisation satisfy 2. (2)
If has finite order , and is the maximal length of words in then is also finite with , and the length of each is at most 3. (3)
Assume is finite and let be the maximal length of words in . Then is finite if and only if every word has length .
3.3. Canonical dualities on sets of monomials
For completeness we consider the canonical dualities between (a) the notions ”obstruction set”, and ”order ideal of monomials”, see [3], and (b) the notions ”antichains of Lyndon words” and ”Lyndon atoms, ”, [20].
Let be an ideal in , We use the notation and terminology from the previous sections, in particular (1.2), (1.3), and Definition-notation 3.2. We have seen that the set of monomials contains uniquely determined maximal antichain of monomials (with respect to ). satisfies (i) every element of is minimal with respect to ””, and (ii) each contains as a subword (a segment) some , i.e. . Denote by the two-sided ideal of generated by .
Clearly, the set of normal words is uniquely determined by the set of obstructions and is characterized by the properties:
[TABLE]
We have seen in the introduction that every set of monomials which satisfies conditions N1 and N2 determines uniquely an antichain of monomials , so that condition N3 holds and is the set of obstructions for .
An obstruction set is characterized by the following properties:
[TABLE]
Definition 3.12**.**
- (1)
A set of monomials which satisfies conditions N1 and N2 is called an order ideal of monomials (o.i.m.), see [3]. 2. (2)
The antichain of monomials described above is called the set of obstructions for . When is considered as , it is also referred to as the set of obstructions for the algebra
Remark 3.13*.*
- (1)
Suppose that is an antichain of Lyndon monomials, and . Then determines uniquely a set of Lyndon atoms . It satisfies
[TABLE] 2. (2)
Conversely, each set of Lyndon words satisfying conditions C1 and C2 determines uniquely an antichain of Lyndon monomials , such that condition C3 holds, and is the set of Lyndon atoms corresponding to . In fact is the unique maximal antichain of the complement
We systematize the above discussion and add some results from [20] in the following proposition.
Proposition 3.14**.**
- (1)
There exists a one-to-one correspondence between the set of all antichains in with and the set consisting of all order ideals of monomials , satisfying condition N1 and N2. In notation as above this correspondence is defined as
[TABLE] 2. (2)
There are equalities
[TABLE]
and each pair (respectively obtained via this correspondence satisfies condition N3. 3. (3)
There exists a one-to-one correspondence between the set of all antichains of Lyndon words with and the set consisting of all sets of Lyndon words satisfying conditions C1 and C2. In notation as above this correspondence is defined as
[TABLE] 4. (4)
There are equalities
[TABLE]
and each pair (respectively obtained via this correspondence satisfies condition C3. 5. (5)
[20]** Each finite antichain of Lyndon words determines a monomial algebra of finite global dimension, . 6. (6)
[20*]**
If is a finite set of Lyndon words of order , then the corresponding antichain is also finite with .* 7. (7)
Each with C1 and C2 determines uniquely a monomial algebra , with a set of defining relation and a set of Lyndon atoms precisely . The algebra has polynomial growth of degree iff .
Example 3.15**.**
The algebra is s.f.p.- the defining relations form a Gröbner basis of the ideal . The set of obstructions consists of Lyndon words. The set of Lyndon atoms is , and is the corresponding Lyndon pair. The normal basis of is
[TABLE]
is a classical AS-regular algebra of , type A.
4. Some combinatorial results on Lyndon pairs
In this section , and is a Lyndon pair in .
4.1. Proof of Theorem II
Lemma 4.1**.**
Suppose , and let be an antichain of monomials of arbitrary cardinality. Let be the set of normal words modulo , and let be the set of normal Lyndon words. The following conditions are equivalent.
- (1)
* is an antichain of Lyndon words.* 2. (2)
The set of normal words has the shape given in 1.10.
Proof.
As usual, denotes the monomial algebra with defining relations . Then , and the set is a -basis of Note that the set of normal Lyndon words satisfies conditions C1 and C2, whenever is an arbitrary antichain of monomials, but is not necessarily a set of Lyndon words. Indeed, , so satisfies condition C 1. Suppose , and is a subword of . Then, since , the word is also normal, (see N2) thus , and therefore condition C2 also holds. Use now the canonical duality from Proposition 3.14 and consider the pair Notice that if is an arbitrary antichain of monomials, and , then, in general, an inequality is possible. We claim that holds if and only if (1.10) is in force.
(1) (2). Assume that is an antichain of Lyndon words, so is the set of Lyndon atoms corresponding to . Then is a monomial algebra defined by Lyndon words in the sense of [20], hence, by Theorem A [20], the set of normal words has the shape (1.10).
(2) (1). Assume that the set of normal words is presented via (1.10).
We have seen that the canonical duality from Proposition 3.14 defines uniquely the pair where is an antichain of Lyndon words. Then the corresponding set of Lyndon atoms, . Now is a monomial algebra defined by Lyndon words in the sense of [20], and its set of Lyndon atoms is . Use now the first duality in Proposition 3.14, and consider the pair . The set is the normal basis for the monomial algebra . Hence by Theorem A [20] again, has a ”PBW”-type presentation via the Lyndon atoms in which is exactly the set given in (1.10). Therefore , so . It follows that the obstruction set is an antichain of Lyndon words. ∎
Proof of Theorem II**.**
Part (1) of Theorem II is proven by Lemma 4.1.
Suppose that is an antichain of Lyndon words, so is a Lyndon pair. Let , let be the corresponding monomial algebra. Clearly, .
(2). The two algebras and share the same set of obstructions , the same set of Lyndon atoms , the same normal -basis and the same Hilbert seies. In particular, is a monomial algebra defined by Lyndon words in the sense of [20], so the results in [20] are true for .
There is an obvious equality Therefore by [20], Theorem A, (2) each of the algebras and has a polynomial growth of degree iff . Assume has polynomial growth of degree . Then has order , say, . Then Theorem A, [20] implies . Hence there is a -chain on and there is no -chain on . Now the algebra satisfies the hypothesis of Theorem I, so , which gives (2a).
The two algebras and share the same normal -basis and the same Hilbert series, therefore (2b) is in force. The monomial algebra satisfies the hypothesis of [20], Theorem B, and therefore its set of obstructions satisfies the inequalities (1.12), which implies (2c). Part (3) follows from Theorem III, and is given for completeness. The theorem has been proved.
4.2. Up to isomorphism, for every there exists unique Lyndon pair , where is connected and .
Suppose is a Lyndon pair in , where has finite order . It follows from Theorem II that is a finite set with
[TABLE]
It follows from [20], Theorem B (2) that the upper bound is attained if and only if , or equivalently, and , see Theorem II (2). In this case the class contains all binomial skew polynomial rings with square-free relations generated by , in the sence of [17]. Each such a ring is an AS-regular PBW algebra of global dimension . For example, for there are more than non-isomorphic skew-polynomial rings with binomial square-free solutions, each of which is an AS-regular and defines a set-theoretic solution of YBE. It is interesting to describe the other ”extreme” case of Lyndon pairs, when the order attains the lower bound. Theorem 4.2 shows that if is connected then the pure numerical datum
[TABLE]
determines uniquely (i) the order of , , and (ii) unique Lyndon pair in (up to isomorphisms of monomial algebras ). Corollary 7.8 implies that in this case the class contains the universal enveloping U=U\mbox{\mathcal{L}}_{d-1}, where \mbox{\mathcal{L}}_{d-1} is the standard filiform algebra of dimension and nilpotency class . is an AS-regular algebra with .
Theorem 4.2**.**
For every integer , there exists unique (up to isomorphism) Lyndon pair , where is a connected set of finite order and . Moreover, the following conditions are equivalent:
- (1)
* and is a connected set of Lyndon atoms, see Def 5.9;* 2. (2)
* is a connected set of Lyndon atoms and*
[TABLE] 3. (3)
The set has order , so , and (up to isomorphism of monomial algebras ) the set of atoms has the shape
[TABLE] 4. (4)
[TABLE]
Proof.
(1) (2). By Proposition 3.10 (2) there is an inclusion of sets
[TABLE]
The equivalence of (1) and (2) is straightforward.
(2 ) (3). Assume has the shape given in (4.2). We shall prove (3) in several steps.
We start with an useful lemma.
Lemma 4.3**.**
Suppose (4.2) holds, and contains a subchain . Then
- (a)
* is not in .* 2. (b)
Furthermore, if each proper Lyndon subwords is an atom, , one has
Proof.
We shall prove (a). By hypothesis the set of Lyndon atoms is and by (4.2) is the set of words Assume now that , so for some Note that otherwise and , with , is impossible. Therefore and either (i) is a proper left segment of , or (ii) is a proper left segment of . Suppose (i) holds. Then is a proper left segment of and overlaps with a proper left segment of , that is , see Notation 3.6. Now Lemma 3.7 part (3) implies that
[TABLE]
which is impossible, since . In the case (ii) overlaps with a right segment of , and is a proper right segment of . So , and Lemma 3.7 part (3) implies that
[TABLE]
a contradiction. It follows that is not in . Part (b) is straightforward. ∎
Step 1. . Assume the contrary, then has at least three distinct elements, say . By assumption contains only products of Lyndon atoms which are ”neighbors”, so by Lemma 4.3 , and contains the subset . Now by induction on (using analogous argument) one shows that contains an infinite subchain
[TABLE]
which contradicts the hypothesis. By convention has at least 2 elements, so . We shall describe .
The case is trivial, one has . Assume .
Step 2. We prove that does not contain simultaneously and . Otherwise, contains the subchain , hence and are not neighbours, so is not in , and since every proper Lyndon subword of is in , Lemma 4.3 implies . Therefore contains . Lemma 4.3 again implies , so . Next one proves by induction on (and using Lemma 4.3) that contains an infinite set of Lyndon atoms: , which contradicts the hypothesis again.
From now on, without loss of generality, we may assume (If we assume we shall obtain an isomorphic monomial algebra ). Since is connected, it contains a Lyndon word of length 3, hence . Therefore is a subchain in .
Step 3. We shall prove that satisfies (4.3). Assume the contrary, then contains a Lyndon atom of the shape . Let be the shortest Lyndon atom of such shape. By hypothesis is closed under taking Lyndon subwords, and it is easy to prove that , for some . It follows that
[TABLE]
Note that , in particular . Indeed, by assumption is the shortest atom with . Moreover, since and is connected, one has
[TABLE]
One has , and .
Consider the inequalities of Lyndon atoms
[TABLE]
we claim that the product is in . Indeed, by Lemma 4.3 (a) is not in . The set of proper Lyndon subwords of is , and therefore by Lemma 4.3 (b), The following sub-chain of atoms is contained in :
[TABLE]
A similar argument implies that , and using induction on , Lemma 4.3, and a similar argument as above one proves that contains an infinite increasing chain of Lyndon atoms:
[TABLE]
which contradicts It follows that satisfies (4.3).
The implications (3 ) (4) and (3 ) (2) are clear. This proves the theorem. ∎
5. Classes containing the enveloping algebra of a Lie algebra
5.1. Finitely generated graded Lie algebras and their enveloping algebras, standard Lyndon bases
Our main references are [29], [30] [36], [24], [22].
In this subsection we set notation and recall some classical facts about finitely generated graded Lie algebras and their enveloping algebras which will be used in the sequel. As usual, , is a finite set, is a field of characteristic [math], denotes the free Lie - algebra generated by , and is the universal enveloping algebra of . It is known that and are isomorphic as associative algebras, and we shall identify them. denotes the set of all words of length in . The set of all Lyndon words in is denoted by , and is the set of all Lyndon words of length . The free associative algebra is naturally graded by length. An element is called a Lie element (or a Lie polynomial) if . It is known that if is a Lie element, then each homogeneous component of is also a Lie element. A special case of homogeneous Lie elements are the so called -left nested Lie monomials:
[TABLE]
Clearly, the Lie monomial may also be considered as an element of the free associative algebra , via the well-known equality , so . The free Lie algebra is also naturally graded by length,
[TABLE]
Throughout the paper the right standard factorisation will be called simply ”the standard factorisation of ”. Recall that is the longest proper right segment of which is a Lyndon word, in this case is also a Lyndon word.
Definition 5.1**.**
[29]
Let Consider the unique (right) standard factorization denoted . The (right) standard bracketing on the set of all Lyndon words is defined inductively as follows.
[TABLE]
The left standard bracketing is defined analogously, but using the left standard factorization.
Definition-Convention 5.2**.**
The notation stands for the right standard bracketing, and will be used for the left standard bracketing of . The Lie element is called also a Lyndon-Lie monomial (corresponding to ). Given a nonempty set of Lyndon words, , we denote by the set of standard bracketings of all elements of :
[TABLE]
and refer to it as ”the bracketing of ”.
Remark 5.3*.*
- (1)
For each Lyndon word , the standard bracketing considered as an element of the algebra is a (noncommutative) polynomial which is a linear combination
[TABLE]
where each nonzero coefficient is an integer, all (distinct) monomials in this presentation satisfy , in particular , see [29], Lemma 5.3.2, or [36], Theorem 5.1. 2. (2)
It is not difficult to see that considered as ”commutative terms” all monomials occurring in the right-hand side of (5.4) are equal, that is the monomials and have the same multi-degree as , (see Definition-Notation 6.3). 3. (3)
The elements
[TABLE]
form a basis of , see [29], Theorem 5.3.1, or [36], Theorem 4.9.
In particular, for each the set of Lyndon Lie monomials of degree ,
[TABLE]
is a -basis of the graded component (considered as a vector space). 4. (4)
On the other side, for each the graded component is spanned by all left-nested Lie monomials , , hence there are equalities of sets
[TABLE]
The works [24]and [20] are central for this subsection. These two works use different terminology for equivalent notions and, for convenience, we adapt the terminology from [24] to our equivalent terminology given in [20], and typical for the theory of non-commutative Gröbner bases. The equivalence is given in Remark 5.5. The notion of ”an atom” was introduced by Anick, [2] in a more general context, the special properties of ”the atoms” in the sense of Anick are listed in [2], Theorem 5. It is easy to see that if is a Lyndon pair, then the set of Lyndon words has the properties analogous to the properties of a set of atoms, in the sense of Anick. In fact coincides with the set of atoms for the monomial algebra , see also our comments in [20]. Recall that, by convention, all Gröbner bases of (associative) ideals in and all Lyndon-Shyrshov bases of Lie ideals in are considered with respect to ”” -the degree- lexicographic well-ordering on .
Notation-Convention 5.4**.**
From now on will denote a Lie ideal in generated by homogeneous Lie elements,(see the natural grading of , (5.2). will denote the two-sided ideal in generated by , where is considered as a set of associative polynomials, . Let
[TABLE]
It is known that is the enveloping algebra of the Lie algebra , see for example [24], p. 1823. The algebras and are naturally graded. denotes the corresponding set of obstructions for the associative algebra . As usual, denotes the set of normal words modulo (w.r.t. ). , is the set of all Lyndon words which are normal mod , we may also write . By definition , where is the set of Lyndon words. We shall prove that the obstruction set is an antichain of Lyndon words, see Proposition 5.6 (1), so, by convention, we call the set of Lyndon atoms for . In this case is a Lyndon pair.
The remark given below provides an useful interpretation of the terminology of [24] in terms of our equivalent notions introduced in this paper and in [20]. Note that our terminology is introduced in more general settings not necessarily in the context in Lie algebras, and independently. We use results from [20] and [24], see Theorem (2.1), Corollaries (2.5) and (2.8) to deduce straightforwardly the following.
Remark 5.5*.*
In notation and assumption as above, , , and is the set of obstructions for . (1) A word is standard modulo in the sense of [24] iff is normal modulo , that is no satisfies (equivalently, no , for some ). Moreover, the set of normal (mod ) monomials coincides with the set of ”standard mod monomials” in the sense of [24]. (2) The set of Lyndon atoms, where coincides with the set of ”Lie-standard Lyndon words”, in the sense of [24]. Moreover, the set is a -basis of . (3) The enveloping algebra has a -basis of ”Poincaré-Birkhoff-Witt” type expressed in terms of its Lyndon atoms via (1.10). This basis coincides with the set of normal words (modulo ).
5.2. Basic information on classes containing the enveloping algebra of a graded Lie algebra
Proposition 5.6**.**
Suppose is a graded Lie algebra, , is its enveloping algebra. Let be the set of obstructions for , as usual, denotes the set of normal monomial, and is the set of normal Lyndon words. The following conditions hold.
- (1)
* is an antichain of Lyndon words, so is a Lyndon pair, and is a monomial algebra defined by Lyndon words, in the sense of [20].* 2. (2)
Assume furthermore, that is generated by homogeneous Lie elements, so and the enveloping algebra are canonically graded. Then the algebras , and are in the class . The following conditions are equivalent.
- (a)
* is an Artin-Schelter regular algebra.* 2. (b)
* has polynomial growth.* 3. (c)
The Lie algebra is finite dimensional. 4. (d)
The set of Lyndon atoms is finite.
Each of these equivalent conditions implies that
[TABLE]
Moreover, is a standard finitely presented algebra and
[TABLE]
Proof.
(1). The algebra has a -basis of ”Poincaré-Birkhoff-Witt” type expressed in terms of its Lyndon atoms via (1.10), see Remark 5.5. Hence by Lemma 4.1 the set of obstruction is an antichain of Lyndon words. Clearly, then is a Lyndon pair. Moreover, is a monomial algebra defined by Lyndon words and all results of [20] are applicable.
(2). The algebras and belong to , they have the same -basis, , and the same set of Lyndon atoms , in particular, . The equivalence below follows from [20], Theorem A (2).
[TABLE]
Moreover, [20], Theorem B (2) gives the following relations
[TABLE]
Now [24], Theorem (2.1) implies straightforwardly that the set is a -basis of the Lie algebra , there is an isomorphism of vector spaces . Thus is finite dimensional, if and only if the set of atoms is finite, in this case Assume now that has polynomial growth of degree , so
[TABLE]
It follows from (5.8) that , hence (2b) (2d), and (2b) (2c). By (5.9) , and by [15], Theorem II, The obstruction set is finite (by (5.9) again) therefore is standard finitely presented.
Suppose now is Artin-Schelter regular, then by definition it has polynomial growth of degree, say , so (2a) (2b), and therefore the remaining conditions (c) and (d) are also in force. The discussion above implies .
Conversely, assume that is finite dimensional. It is known by the experts that the enveloping algebra of a finite dimensional positively graded Lie algebra is Artin-Schelter regular. A reference to the original result is difficult to find, and we refer to the proof provided by Fløystad and Vatne, see [14], Theorem 2.1. ∎
We have seen that the enveloping algebra of any Lie algebra generated by has a set of obstructions , consisting of Lyndon words, in general, may be infinite. Moreover, if has finite dimension , then the enveloping algebra is a standard finitely presented Artin-Schelter regular algebra of global dimension . The number of its relations satisfies the inequalities (5.7). The following question arises naturally.
Question 5.7**.**
Let be a Lyndon pair, where is a finite set. Under what combinatorial conditions is the set of obstructions for the enveloping of some Lie algebra ?
Proposition 5.11 gives a necessary condition in terms of , the set of Lyndon atoms. In section 6 we introduce and study the monomial Lie algebras . In this case the enveloping algebra belongs to if and only if is a Gröbner-Shirshov basis of the Lie ideal of .
5.3. If the class contains the enveloping algebra of a graded Lie algebra then the set is connected
Remark 5.8*.*
Due to the duality , , see Proposition 3.14, each class of associative graded algebras , with obstruction set consisting of Lyndon words is also uniquely determined by the set of Lyndon atoms . Conversely, each set of Lyndon words closed under taking Lyndon subwords, and with , determines uniquely a class , where , and is a Lyndon pair. It is interesting to classify all Lyndon pairs such that the monomial algebra has polynomial growth and , where is fixed. We use our result (in this case is always finite). Our method consists of two steps:
- (1)
Classify the sets of Lyndon atoms of a fixed finite order , must satisfy C1, C2. 2. (2)
Write down the corresponding set of obstructions , so that is a Lyndon pair.
In this case (whenever we start with ) we shall often use notation for the class of associative graded algebras with set of obstructions . A Lyndon pair is uniquely determined by each of its component , or , and by convention we shall use both notation
[TABLE]
Till the end of the paper we shall consider only the special case when the ideal of is graded by length, respectively, multi-graded, thus the enveloping algebra is also graded, respectively, multi-graded. Moreover, has a set of obstructions consisting of Lyndon words, and .
Note that every Lyndon word of length in a finite alphabet , contains a Lyndon subword of length . Moreover, if is a finite set of Lyndon monomials in with C1, C2, and contains a word of length then .
Definition 5.9**.**
Suppose is a set of Lyndon words (possibly infinite) with conditions C1 and C2. (i.e. , and is closed under taking Lyndon subwords).
- (1)
We say that is connected if
[TABLE]
In particular, if is a finite set, then is connected iff (5.11) is in force for , and all . 2. (2)
Suppose is not connected, that is for some . Then the connected component of , denoted by is defined as
[TABLE]
For completeness we set , whenever is connected. Note that is the maximal connected subset of , which satisfies C1 and C2.
One can find explicit examples in the last section, where we have classified the Lyndon pairs in the alphabet , with , see Subsec. 9.3. Each of the first 12 pairs , (7.4.1) through (7.6.12), has connected set of atoms . Each of the remaining pairs has a disconnected set of Lyndon atoms, but we have identified explicitly the connected components . We illustrate this with the Lyndon pair given in (7.7.15). The set of Lyndon atoms, is disconnected. One has , and . In this case has a connected component N_{con}=N\setminus\{xy^{2}xy^{3}\}=N(\mbox{\mathcal{L}}_{5}), the set of atoms corresponding to the Filiform Lie algebra \mbox{\mathcal{L}}_{5} of dimension and nilpotency class of degree .
Suppose is finite. Clearly, the property ” is connected” is recognizable. Moreover, this is a necessary condition so that ” contains the enveloping algebra of some Lie algebra ”, as shows Proposition 5.11.
Definition 5.10**.**
Recall that is a nilpotent Lie algebra of nilpotency class if is the minimal value for which , where
[TABLE]
is the lower central series of . In this case we shall also write .
Proposition 5.11**.**
Suppose is a Lyndon pair, such that the class contains the enveloping algebra of a graded Lie algebra . Then the following conditions hold.
- (1)
If , for some then , for all . In particular, is finite. Moreover, if is the minimal positive integer with this property, then is a nilpotent Lie algebra of nilpotency class . 2. (2)
If is an infinite set, then for all , so is connected. 3. (3)
Suppose is a finite set, and let . Then the Lie algebra is nilpotent of nilpotency class 4. (4)
The set of Lyndon atoms is connected, and is a -basis for .
Proof.
It follows from the hypothesis that there exists a Lie algebra , where is a (graded) Lie ideal whose enveloping , has set of obstructions ( is the associative two-sided ideal generated by ). In fact determines uniquely the Lyndon pair , given in the hypothesis.
(1) Assume , for some integer . It follows from the equalities (5.5) that there is an inclusion of subspaces
[TABLE]
This implies that for every nested Lie monomial i.e. is a nilpotent Lie algebra of nilpotency class . Suppose that is the minimal integer with . Then , and , hence is a nilpotent Lie algebra of nilpotency class , see Definition 5.10.
Part (1) implies straightforwardly part (2).
(4). Suppose that is a finite set. It follows from Definition 5.9 and part (1) that is connected. By part (2) is connected, whenever it is an infinite set.
(3) Assume that has finite order and . Without loss of generality we may assume The set is connected, hence . The equality implies
[TABLE]
therefore It follows from part (1) that is a nilpotent Lie algebra of nilpotency class . ∎
Corollary 5.12**.**
Let be a Lyndon pair, where is finite. Suppose is not a connected set of Lyndon words. Then (i) the class does not contain the enveloping algebra of a graded Lie algebra . (ii) In particular, is not a Lyndon-Shirshov basis of the Lie ideal of .
6. Monomial Lie algebras defined by Lyndon words
As usual, is a fnite set and denotes an antichain of Lyndon words in . Our goal is to find explicitly (in terms of generators and relations) classes containing enveloping algebras of prescribed global dimension. As a first step in this direction we introduce and study the monomial Lie algebras defined by Lyndon words.
6.1. Basic definitions and first properties of Monomial Lie algebras
Definition 6.1**.**
We say that is a monomial Lie algebra defined by Lyndon words, or shortly, a monomial Lie algebra, if is a Lie ideal generated by the Lie elements , where is an antichain of Lyndon words. If moreover, the set of Lie monomials is a Gröbner-Shirshov basis of the Lie ideal (see Definition 6.10) we shall refer to as a standard monomial Lie algebra and denote it by .
Definition 6.10 is one of the several equivalent definitions of a Gröbner-Shirshov basis of a Lie ideal in (or shortly a GS basis). Proposition 6.15 gives several equivalent conditions each of which can be used for a definition. For now we shall use the fact that is a Gröbner-Shirshov basis of the Lie ideal iff the set considered as ”associative” elements of is a Gröbner basis of the associative ideal , w.r.t the ordering on , see Proposition 6.15. Our first examples are classical.
Example 6.2**.**
- (1)
The free-nilpotent Lie algebra , , generated by , is a standard monomial Lie algebra defined by Lyndon words, see Corollary 7.3. 2. (2)
The standard Filiform Lie algebra , of dimension and nilpotency class , see Subsec 7.4.
Each monomial Lie algebra defined by Lyndon words, is graded by length, and is also positively -graded. Moreover, the two sided ideal of is generated by homogeneous elements and is also -graded, so the enveloping algebra is (non-negatively) -graded.
Definition-Notation 6.3**.**
A monomial , has (a non-negative)multi-degree , , if , considered as a commutative term, can be written as . In this case we write . In particular, the unit has multi-degree , and For each , denote by the set . Then the free monoid is naturally -graded:
[TABLE]
The set of Lyndon words inherits this grading canonically, it splits as a disjoint union
[TABLE]
The free associative algebra generated by is also canonically -graded:
[TABLE]
For example, if , then
[TABLE]
Recall that a graded Lie algebra is a Lie algebra endowed with a gradation compatible with the Lie bracket. The free Lie algebra is naturally -graded:
[TABLE]
If has multi-degree , then the standard (right) bracketing , considered as an associative polynomial in is a linear combination of words so and its highest monomial w.r.t. is precisely . Sometimes we shall also write
[TABLE]
For example, . Let be an antichain of Lyndon monomials in . Then the Lie ideal is multi-graded, where .
Remark 6.4*.*
Recall that if \;\mbox{\mathcal{I}}=(W) is a monomial ideal in , where is an arbitrary antichain of monomials (), then is -graded. If f=\sum c_{a}a\in\mbox{\mathcal{I}}, with then every monomial in this presentation is also contained in , or equivalently, for some . In contrast, a monomial Lie ideal generated by Lyndon Lie monomials may contain a Lie element , where , such that is not in for some , see Example 6.5. Moreover, the reduced Gröbner-Shirshov Lie basis of a monomial Lie ideal in may contain a Lie element , which is not a Lie monomial, as shows Example 8.4.
Example 6.5**.**
Let , , then the Lie monomial is a Gröbner-Shirshov basis of the Lie ideal (since there are no compositions to resolve). The associative polynomial is a Gröbner basis of the associative ideal , Consider the Lie element . Note that , is the left standard bracketing of . One has:
[TABLE]
The Lie monomial is not in since is normal modulo . One has
[TABLE]
6.2. Gröbner-Shirshov bases of monomial Lie ideals, preliminaries
In this subsection we present in terms of Lyndon words some notions and results from Shirshov’s theory based on Shirshov’s regular words, or as they call them recently ”Lyndon-Shirshov words”. During the last decades numerous works appeared involving Lyndon-Shirshov words in the study of Gröbner-Shirshov bases, normal forms, combinatorial and decision problems, and the interested reader can find details in [9, 10, 11, 12], et al, and the references therein.
However, although the so called Lyndon-Shirshov words are analogous to Lyndon words (in the classical sense), they are defined differently and neither Lyndon-Shirshov words, nor corresponding theory and results can be used directly (without certain modification) for our theory based on the original notion of Lyndon word.
Our goal is to simplify the general procedure for finding Gröbner-Shirshov bases in the special case of -graded monomial Lie ideals in and to apply it for our classification theorem in Section 9.
For the purpose we combine some known facts about Gröbner-Shirshov bases with results proven in [3], [24], [20] and especially with our combinatorial results in this paper. As a first step we yield Proposition 6.15, which can be used in some special cases to decide straightforwardly (avoiding computations) whether is a Gröbner-Shirshov basis of the monomial Lie ideal . We next prove Propositions 6.17, and 8.5 which (in various cases) can be and will be used to find straightforwardly the reduced Gröbner-Shirshov basis of the Lie ideals (and avoid the need of heavy computations prescribed by the standard procedures). As an application, in Section 9 we classify all 2-generated AS-regular algebras of global dimension and which occur as enveloping algebras , of standard Lie monomial algebras , (in this case is a Gröbner-Shirshov Lie basis).
Recall that if is a nonzero polynomial, then its highest monomial with respect to the degree-lexicographic order on (with ) is denoted by . One has , where , .
Let be a Lyndon word of length We know that the right bracketing and the left bracketing are Lie elements which (considered as associative elements) are homogeneous monic polynomials with equal highest terms
[TABLE]
More generally, consider a new bracketing (”rebracketing”) of , that is a Lie monomial, obtained by writing Lie brackets in some way. It is called a regular bracketing, or a regular Lie monomial of if it is a monic (associative) polynomial with highest monomial .
Lemma 6.6**.**
Let be a Lyndon word, , and let where are Lyndon words. If , and are regular bracketings (regular Lie monomials) of and , then is a regular bracketing of respecting and . In particular, is a regular bracketing of respecting .
Proof.
It follows from the hypothesis that Recall that for any two elements of the free associative algebra, one has
[TABLE]
The following relations are considered in the free associative algebra:
[TABLE]
For monomials of the same length in we have iff , so , and therefore
[TABLE]
as desired. Clearly this bracketing of respects both and . ∎
The following is a modification of the original result [34], Lemma 4, see also [10], p.372.
Shirshov’s special bracketing. Suppose , (, or is possible). Then the right standard bracketing satisfies
- (1)
, where , , and possibly . 2. (2)
If , express as a product of Lyndon words , , where . Replacing by we obtain the Lie monomial
[TABLE]
which is called the Shirshov special bracketing of relative to . 3. (3)
There is an equality .
To emphasise that we use Lyndon words (and, in general, not Lyndon-Shirshov words) we slightly change the terminology.
Definition-Notation 6.7**.**
Let , , where , or is possible. We shall refer to as the regular bracketing of respecting and denote it also by .
Remark 6.8*.*
We shall give a sketch of proof of the existence of the regular bracketing in terms of Lyndon words which could be used effectively. (It is not analogous to Shirshov’s proof).
We use induction on the length of . If , then each proper segment is in and obviously .
Suppose the statement is true for every pair where , and .
Assume , and is a proper segment of . Three cases are possible (i) , (ii) , and (iii) where . Clearly, in case (i) . Assume (ii), , with . The left standard factorisation satisfies and , possibly . If , then . If , by the induction hypothesis is well defined and we set . Case (iii). Suppose now where . Consider the right standard factorization of ( is the longest Lyndon right segment of , and ). By Lemma 3.9 either , or We apply the induction hypothesis and in the first case we set , while in the second case .
Suppose , then its standard bracketing , and the regular bracketing is a Lie element satisfying
[TABLE]
Moreover, the regular bracketing is an element of the Lie ideal of , generated by .
Example 6.9**.**
(1) Let . Then , and, by Lemma 6.6, the Lie monomial is a regular bracketing of respecting the bracketings and . Here is the Shyrshov special bracketing of relative to . Note that the standard right factorisation of is which implies , so the standard bracketing does not ”respect” the standard bracketing of the Lyndon subword .
(2) Consider where . Then .
Definition 6.10**.**
Let be an antichain of Lyndon words. The set of Lie monomials is a Gröbner-Shirshov basis of the Lie ideal if every nonzero Lie element , can be written as
[TABLE]
where , and , for some . Note that in this case is the reduced Gröbner-Shirshov basis of , and is the reduced Gröbner basis of the associative ideal in .
Definition 6.11**.**
Let be an antichain of Lyndon words, .
- (1)
[2, 3] A 2-prechain is a word , where , and . A 2-prechain is a 2-chain if no proper left segment of is a 2-prechain. Note that every 2-prechain is a Lyndon word, see for example [20], Proposition 6.4. 2. (2)
Suppose is * a 2-prechain*, where , and . The difference of Lie elements
[TABLE]
is called a Lie composition of overlap, or also a Lie composition of intersection (in the sense of Shirshov). 3. (3)
Suppose , with . The composition of inclusion (in the sense of Shirshov), corresponding to the pair , is defined as
[TABLE]
Note that by assumption is an antichain of Lyndon words, that is no is a proper sub-word of some . Thus compositions of inclusion, in the sense of Shirshov, will not occur in the initial step of the procedure for finding a Gröbner-Shirshov basis of the Lie ideal . However, in the process of resolving a composition of overlap, compositions of inclusion may also occur.
Remark 6.12*.*
Notation as above.
- (1)
The Lie element is in the ideal . Indeed, , thus and . 2. (2)
One has , hence, either or
[TABLE] 3. (3)
Suppose a two-prechain , notation as above. Then each of the Lie elements , and is in the graded component . Therefore is a homogenous Lie element of multi-degree , or . More precisely, . In particular, if is the minimal element of (w.r.t. ) then, clearly, , and the composition is solvable
Definition 6.13**.**
Suppose is a 2-prechain of multi-degree , so . The composition of overlap is solvable mod if either , or there exist an explicit presentation
[TABLE]
where (i) are words in ; (ii) , and is a regular bracketing respecting the right standard bracketing .
One can find effectively whether a presentation of the form (6.4) exists. If not, there exists a presentation
[TABLE]
where , and are Lyndon atoms in . In this case is not the obstruction set of the associative ideal , neither is the set of Lyndon atoms for .
The next corollary follows from Anik’s results, see [3], 2.Diamond Lemma revisited (pp 646-647), part (ii) follows from [24].
Corollary 6.14**.**
If all compositions of overlap on 2-chains are solvable modulo in the free associative algebra ) then the compositions on 2-prechains are also solvable, so (i) is a Gröbner basis of the associative ideal in ; (ii) is a Gröbner-Shirshov basis of the Lie ideal .
The following statement is a consequence from well-known results on Gröbner bases, Corollary 6.14, and our assumption that is an antichain of Lyndon words.
Proposition 6.15**.**
Let be a Lyndon pair, let be the Lie ideal of , generated by , let be the two sided ideal in the free associative algebra . Suppose , so is the enveloping algebra of . The following conditions are equivalent.
- (1)
The set of Lie monomials is a Gröbner-Shirshov basis of the Lie ideal , see Definition 6.10. 2. (2)
For every 2-chain , the Lie composition of overlap is solvable mod . 3. (3)
The set
[TABLE]
is a K-basis of . 4. (4)
The set (considered as a set of ”associative” elements of ) is a Gröbner basis of the (associative) ideal (w.r.t. the ordering on ). 5. (5)
The enveloping algebra is a graded algebra whose set of obstructions is , that is every satisfies for some . 6. (6)
The set of monomials
[TABLE]
is a normal - basis of .
In this case is a connected set of Lyndon atoms.
Definition-Notation 6.16**.**
Let be a Lyndon pair. We introduce the following numerical invariants.
[TABLE]
The following Proposition gives some purely combinatorial conditions sufficient for to be a Gröbner-Shirshov basis of the Lie ideal .
Proposition 6.17**.**
In notation and assumptions of Proposition 6.15. Let be a Lyndon pair.
- (1)
Suppose , and is a 2-chain on of multi-degree , or equivalently, . If , then the composition of overlap is solvable mod . Clearly, this is so, whenever 2. (2)
Suppose , and . The composition of inclusion corresponding to the pair , , is solvable mod , whenever . In particular, this is so if . 3. (3)
Suppose , is connected and some of the following conditions is in force:
- (a)
* that is each -chain on has length ;* 2. (b)
Every 2-chain on is a minimal word in its -component, , that is . 3. (c)
Every , belongs to some -component, , which does not contain any -chain .
Then is a Gröbner-Shirshov basis of the Lie ideal in , the Lie algebra is a standard monomial Lie algebra with and a -basis . In this case the enveloping algebra belongs to the class and is an Artin-Schelter regular algebra of .
7. Some classical monomial Lie algebra defined by Lyndon words
7.1. The free nilpotent Lie algebra of nilpotency class is a standard monomial Lie algebra defined by Lyndon words
Notation 7.1**.**
Denote by the Lie ideal of generated by all -(left) nested Lie monomials (length ), see (7.1).
The definition of and (5.5) imply
[TABLE]
Moreover,
[TABLE]
Remark 7.2*.*
Suppose that is a Lie ideal in and is a graded nilpotent Lie algebra of nilpotency class . Then all -left nested Lie monomials of degree are zero in .
The Lie algebra is the free nilpotent Lie algebra of degree , see [22].
Recall that denotes the set of all Lyndon words of length . Denote by , , the set of all Lyndon words of length . Clearly, is closed under taking Lyndon subword, so C1 and C2 are in force. Let be (the unique) antichain of Lyndon words whose set of Lyndon atoms is , so
[TABLE]
Consider the Lie algebra
[TABLE]
where denotes the Lie ideal of generated by the set of Lyndon-Lie monomials
[TABLE]
[TABLE]
Corollary 7.3**.**
In notation as above the following conditions hold.
- (1)
The set is a Lyndon-Shirshov Lie basis of the Lie ideal , so is a standard monomial Lie algebra, see Definition 6.1 2. (2)
The Lie ideal generated by all -nested Lie monomials (Notation 7.1) coincides with the ideal , generated by the Lyndon-Lie monomials , . 3. (3)
The free nilpotent Lie algebra of nilpotency class is isomorphic to the standard monomial Lie algebra , given in (7.4). 4. (4)
Moreover, the enveloping algebra is standard finitely presented as
[TABLE]
where the finite set is an associative Gröbner basis for the ideal of . 5. (5)
The class contains an AS-regular algebra of global dimension , namely the enveloping algebra .
Proof.
(1) Every 2-chain on has length , and the corresponding composition is solvable, since there are no atoms of length . Hence the set is a Lyndon-Shirshov basis of the Lie ideal . (2) Clearly, every has length , hence (by (7.1) and (7.2)). Since the set of generators for the Lie ideal is contained in , one has
[TABLE]
But , and (5.5) gives
[TABLE]
so , and therefore . (3) The equality of the two ideals implies straightforwardly
[TABLE]
so is isomporphic to the free nilpotent Lie algebra of of degree (or nilpotency class ) see Remark 7.2. (3) implies (4). (5) follows straightforwardly from Proposition 5.6. ∎
Remark 7.4*.*
The free nilpotent algebra with generators and nilpotency class , has a -basis
[TABLE]
and dimension
[TABLE]
where , is the necklace polynomials, or (Moreau’s) necklace-counting function
[TABLE]
One can also use the Witt’s formula
[TABLE]
The numbers of binary Lyndon words of each length, starting with length one, form the integer sequence
[TABLE]
Corollary 7.5**.**
Let be a Lie ideal in , , (as in Notation-Convention 5.4), and let be the corresponding Lyndon pair for the enveloping algebra . Suppose the set of Lyndon atoms is finite, and is the minimal integer, such that contains atoms of length , but there are no atoms of length Then all -nested Lie monomials belong to , so , and therefore the algebra is nilpotent of degree . In particular, .
Remark 7.6*.*
Let , where is a Lie ideal in , and be its enveloping algebra. Suppose is the Lyndon pair for the enveloping algebra , so is the set of obstructions for . In general, this datum does not determine uniquely the Gröbner -Shirshov basis of the Lie ideal . It is possible that is a Gröbner-Shirshov basis of the Lie ideal , and is the obstructions set of , but . However in some ”extreme cases”, see Corollaries 7.7 and 7.11 the set of obstructions determines uniquely the defining relations of , and the ideal .
Corollary 7.7**.**
Let be a Lie ideal in , where is a graded Lie algebra, (as usual, in Notation-Convention 5.4) and suppose the Lyndon pair for the enveloping algebra is exactly , 7.3. Then
7.2. The -graded Filiform Lie algebras, basic facts
Filiform Lie algebras have been actively studied and numerous works on their algebraic and geometric properties can be found in the literature. As an application of our approach (and for completeness) we shall show that each of the classical Filiform algebras \mbox{\mathcal{L}}_{n} and \mbox{\mathcal{Q}}_{n} has a canonical presentation as a naturally -graded standard monomial Lie algebra defined by Lyndon words. Similar presentation is given for the infinite Filiform Lie algebras , but, of course, its set of defining Lyndon-Lie monomial relations is infinite.
Recall that a finite dimensional filiform Lie algebra is a nilpotent Lie algebra whose nil index is maximal, that is has nil index , (nilpotency class ), where . It is well-known that, up to isomorphism, there are two families of filiform Lie algebras of rank 2.
[TABLE]
Both algebras and have dimension . Each of them has a basis , their nonzero brackets of basis elements are given below.
[TABLE]
[TABLE]
Infinite Filiform Lie algebra , [13], is presented via generators , and relations This algebra was first studied in [35], where was shown that its enveloping is an integral domain with a subexponential growth, not bounded by any polynomial.
Note that if is a graded Lie algebra generated by generators of degree , , then the data implies
[TABLE]
At the same time we have a result on ”the minimality of the order of independently of the context of Lie algebras. Recall that if is a Lyndon pair in , with , Theorem I implies that is a finite set with . Theorem 4.2 shows that if is connected then the pure numerical datum
[TABLE]
determines uniquely (up to isomorphisms of monomial algebras ): (i) the order of , , and (ii) the Lyndon pair in with :
[TABLE]
Corollary 7.8**.**
Let , let be an integer, and let be the Lyndon pair (7.9). Then
- (1)
The set of Lie monomials is a Lyndon-Shirshov Lie basis of the ideal , so is a -basis of , and is the set of obstructions of the enveloping algebra . 2. (2)
The 2-generated monomial Lie algebra is isomorphic to the standard filiform Lie algebra \mbox{\mathcal{L}}_{d-1} of dimension and nilpotency class . 3. (3)
In particular, the class contains the enveloping which is an AS-regular algebra of global dimension .
Proof.
Each two-chain on has multi-degree , where . But , hence by Proposition 6.17 part (3c) is a Lyndon-Shirshov basis of the Lie ideal . This implies that is a basis of , and is the set of obstructions of the enveloping algebra . Clearly, \mathfrak{g}\cong\mbox{\mathcal{L}}_{d-1}, the standard filiform Lie algebra of dimension , and nilpotency class . ∎
7.3. Proof of Theorem III
Proof of Theorem III**.**
(1). Theorem 4.2 proves the equivalence of conditions (1.a), (1.b), (1.c). In this case the Lyndon pair is given by (7.9), hence the hypothesis of Corollary 7.8 is satisfied. This implies that the set of Lie monomials is a Lyndon-Shirshov Lie basis of the ideal in so is isomorphic to the standard filiform Lie algebra \mbox{\mathcal{L}}_{d-1} of dimension and nilpotency class . In particular, the class contains the enveloping U=U\mbox{\mathcal{L}}_{d-1} which is an Artin-Schelter regular algebra of global dimension . This proves (1)
(2). The equivalences of conditions (2.a), (2.b) and (2.c) follow from [20], Theorem B (2). In this case and . The class contains all binomial skew polynomial rings introduced and studied by the author, see [17]. Each of them is an Artin-Schelter regular PBW algebra of global dimension , see [21], and also [18], and defines in a natural way a square-free solution of the Yang-Baxter equation. When more than (non-isomorphic) square-free solutions were found by Schedler via a computer program.
7.4. The -graded Filiform Lie algebras presented as standard monomial Lie algebras defined by Lyndon words
In this subsection, as an application of our approach (and for completeness) we show that each of the classical Filiform algebras \mbox{\mathcal{L}}_{n} and \mbox{\mathcal{Q}}_{n} has a canonical presentation as a naturally -graded standard monomial Lie algebra defined by Lyndon words.
Theorem 7.9**.**
Let be a Lyndon pair, in the alphabet . Let , and be the corresponding ideals, and . Assume that is connected, with order , and
[TABLE]
Without loss of generality we may assume that .
The following conditions hold.
- (1)
* is infinite if and only if*
[TABLE]
* is a Gröbner-Shirshov basis of , so is a standard monomial Lie algebra with a -basis , and , the infinite-dimensional Filiform algebra, see [13].* 2. (2)
* is a finite set of order if and only if one of the conditions (a), or (b) is in force.*
- (a)
[TABLE]
In this case is a standard monomial algebra with a -basis , moreover, \mathfrak{g}\cong\mbox{\mathcal{L}}_{d-1}, the standard filiform Lie algebra of dimension and nilpotency class . 2. (b)
One has and , where
[TABLE]
Again, is a standard monomial algebra with a -basis , \mathfrak{g}\cong\mbox{\mathcal{Q}}_{2m-1}, the second type Filiform Lie algebra of dimension and nilpotency class .
Lemma 7.10**.**
In notation and assumptions as in the theorem, let . Let be a Lyndon pair, where satisfies:
[TABLE]
Suppose . The following are equivalent.
- (1)
* is a Lyndon-Shirshov Lie basis of the ideal , and contains an atom different from , for .* 2. (2)
[TABLE]
and is a -basis of . 3. (3)
The Lie algebra is generated by the set , and its Lyndon-Lie monomial relations form a Gröbner-Shirshov basis, where
[TABLE]
Proof.
By hypothesis , so . It is clear that given in (7.15) and (7.16), respectively, is a Lyndon pair. The implication is obvious.
We shall prove that for , the bracketing is a Gröbner-Shirshov basis of the Lie ideal , we use Proposition 6.17. Indeed, there are exactly three-types of two-chains on . (a) , where . Hence , and ; (b) where , then, clearly, ; (c) where , hence again. It follows from Proposition 6.17 part (1) that in each of the cases (a), (b), and (c) the corresponding composition of overlap is solvable, therefore is a Gröbner-Shirshov basis of the Lie ideal . Now the equivalence is clear.
(1) (2). By the hypothesis of (1), is a Gröbner- Shirshov basis of the ideal . Moreover contains an atom of multi-degree , Since , and is closed under taking Lyndon subwords, it follows that contains a subword of the shape , with . Let be the shortest word in with this property. Clearly, , for some Moreover, contains a convex subchain
[TABLE]
In particular , with , and since contains no other atom of length , one has
[TABLE]
We claim that is the longest Lyndon atom in , and . Note first that , and our assumption (7.14) and (7.18) give (a) all elements with have the shape . This implies an inclusions of sets
[TABLE]
We shall prove that . Indeed, is not in , and there are no atoms of length of multi-degree , hence it will be enough to prove that
[TABLE]
It follows from (7.19) that , hence is a two-chain on , \omega\in\mbox{\mathcal{L}}(2,2s). So there is a composition of overlap:
[TABLE]
which must be solvable, since by assumption is a GS-Lie basis. Detailed computation implies
[TABLE]
for some integer . Hence , and therefore is not an atom. All proper Lyndon subwords of are atoms, indeed these are
[TABLE]
therefore
This together with the information we already have (i) and is the shortest Lyndon word in which occurs more than once imply that that there is no Lyndon atom of length . By assumption is a GS-Lie basis, so by Lemma 5.11 is connected, thus .
It follows that , and satisfies(7.15), so , and therefore . Clearly, is a filiform Lie algebra of nilpotency class and its defining Lyndon Lie monomial relations are . This proves (1) (2). ∎
Proof of Theorem**.**
7.9*. Without loss of generality we have assumed that , hence . Two cases are possible: (i) all atoms in have the shape , or (ii) contains an atom different from , for . Assume (i) is in force. Then every atom of length has the shape , where and exactly one of the cases (i.1), or (i.2) given below is in force. (i.1) contains all words , then . Thus and satisfy (7.11). In this case is an (infinite) Gröbner-Shirshov basis of the Lie ideal (this can be proved recursively), is a -basis of and , the infinite Filiform Lie algebra. The case (i.2) is explicitly described by Corollary 7.8. Assume (ii) is in force, then Lemma 7.10 completes the proof of the theorem . *
Corollary 7.11**.**
Let be a (non-negatively) graded Lie ideal in , , (in usual notation), and suppose the Lyndon pair for the enveloping algebra satisfies one of the conditions (a) (N,W)=(N(\mbox{\mathcal{L}}_{d}),W(\mbox{\mathcal{L}}_{d})), is the pair corresponding to the Filiform Lie algebra \mbox{\mathcal{L}}_{d} of nilpotency class and dimension , see (7.12); or (b) , is the Lyndon pair corresponding to Fillifotm Lie algebra of second type, \mbox{\mathcal{Q}}_{2m-1}, see (7.15), and (7.16). Then the set of obstructions determines uniquely the Gröbner-Shirshov basis of and the defining relations of . More precisely, in case (a) J=[W(\mbox{\mathcal{L}}_{d})]_{Lie}], \mathfrak{g}=\mbox{\mathcal{L}}_{d}, and is the enveloping algebra of the Filiform Lie algebra \mbox{\mathcal{L}}_{d}, ; in case (b) J=[W(\mbox{\mathcal{Q}}_{2m-1})]_{Lie}], \mathfrak{g}=\mbox{\mathcal{Q}}_{2m-1}, and is the enveloping algebra of the Filiform Lie algebra \mbox{\mathcal{Q}}_{2m-1}, .
Proof.
Note that in each of the cases (a) and (b) every obstruction is minimal (w.r.t. ) in its -component, , where . Hence the only monic homogeneous Lie element with multi-degree and with highest monomial is . Therefore in the special cases when one of the conditions (a) or (b) holds, the obstruction set determines uniquely the set of defining relations for , one has . ∎
8. More on Gröbner-Shirshov bases of monomial Lie ideals. Applications
8.1. More on Gröbner-Shirshov bases of monomial Lie ideals
We have a natural problem.
Problem 8.1**.**
Given and a Lyndon pair , where Decide whether is a Gröbner-Shirshov basis of the Lie ideal in .
Proposition 6.15 (2) implies an effective method (Procedure) to check whether is a Gröbner-Shirshov basis of the Lie ideal . Moreover, Proposition 6.17 can be and will be used to simplify the procedure by omitting the unnecessary computation of compositions corresponding to ”long” -chains.
Definition 8.2**.**
- (1)
We say that a Lyndon pair is regular if (i) is finite, and (ii) the class contains the enveloping of a graded Lie algebra . In this case (i) is an s.f.p. AS-regular algebra of global dimension and (ii) the set is a -basis of . 2. (2)
We say that the Lyndon pair is standard, if is a Gröbner-Shirshov basis of the Lie ideal in . Every standard Lyndon pair with a finite set is regular, since the class contains the enveloping algebra of the standard monomial Lie algebra .
We are interested in Lyndon pairs , where is a finite set of order .
Remark 8.3*.*
Suppose is a Lyndon pair, and is the corresponding monomial Lie ideal in , , .
- (1)
In general, the reduced Gröbner-Shirshov basis of the Lie ideal may contain a Lie element , which is not a Lie monomial, as shows Example 8.4. 2. (2)
If is a finite set of order , then the monomial Lie ideal has a finite reduced Gröbner-Shirshov basis and determines uniquely the corresponding Lyndon pair , where is the set of obstructions for the enveloping algebra . One has . Moreover, the Lie algebra is finite dimensional with a -basis and . The enveloping is in the class , an AS-regular algebra of . The set is not necessarily a Gröbner-Shirshov basis of the Lie ideal , see Example 8.4.
Example 8.4**.**
Let and let be the Lyndon pair, determined by the set of Lyndon atoms given below, ,notation as usual :
[TABLE]
By Proposition 6.17, if a -chain on has length , then the corresponding composition is solvable. Thus there is only one composition of overlap which needs computation, namely the unique composition corresponding to the two-chain of length . One has
[TABLE]
The Lie element is in , it is presented as a linear combination of Lyndon Lie monomials, each of which is in (i.e. normal modulo ). Its highest monomial (w.r.t. ) is of length . The set generates . Moreover, every composition of overlap is solvable, all compositions of inclusion correspond to certain monomials of length , and therefore are solvable. The unique composition of overlap corresponding to a two-chain of order is now solvable, so is a GS-Lie basis of . Denote
[TABLE]
Then the set
[TABLE]
is the reduced GS Lie basis of the Lie ideal . Note that is not a GS Lie basis of , thus . Moreover the Lyndon pair , where
[TABLE]
corresponds to the enveloping algebra
Using the terminology from Definition 8.2, the Lyndon pair is regular, i.e. is the set of obstructions of the enveloping (equivalently, is a -basis of ). However, the set is not a GS-Lie basis of the Lie ideal , hence by the same definition, part (2) this Lyndon pair is not standard.
Recall that given a Lyndon pair , the connected component of is defined in Definition 5.9 (2).
Proposition 8.5**.**
Let be a Lyndon pair, . Let be the Lie ideal in , generated by , , and let be the corresponding associative ideal in . Denote by the Lyndon pair corresponding to the enveloping algebra . Let be the connected component of , say . Then the following conditions hold.
- (1)
* if and only if is the reduced Gröbner-Shirshov basis of . In this case is connected.* 2. (2)
In the general case, one has . 3. (3)
* is a finite antichain of Lyndon words, so the monomial Lie ideal has a finite reduced Gröbner-Shirshov basis , and the enveloping is standard finitely presented.* 4. (4)
Assume that is disconnected, then
[TABLE]
and is an s.f.p. Artin-Schelter regular algebra of global dimension . 5. (5)
The Lie algebra has nilpotency class .
Proof.
Note first that if is a Gröbner-Shirshov basis of the Lie ideal , then it is a reduced Gröbner-Shirshov basis.
(1). Assume , then , and is the set of obstructions of the ideal , hence (considered as a set of associative polynomials) is the reduced Gröbner basis of the ideal , and therefore is the reduced Gröbner-Shirshov basis of the Lie ideal . Conversely, if is the reduced Gröbner-Shirshov basis of the Lie ideal then (considered as a set of associative polynomials) is the reduced Gröbner basis of the (associative) ideal , hence is the set of obstructions for , and hence .
(2) Suppose is an atom for , then is normal modulo (equivalently, does not contain as a subword the highest term of any element ). In particular, for every we have , and , hence is impossible, so . This proves , but since is connected, one has
(3) By (2) is finite, with , therefore, by Theorem B, is also finite and satisfies
[TABLE]
It follows that is standard finitely presented, and the reduced Gröbner-Shirshov basis of the Lie ideal is also finite.
(4). The set of Lyndon atoms is connected and implies . Part (5) is clear. ∎
We remind that in Section 5 was set Notation-Convention 5.4 valid till the end of the paper, and in particular, for the following corollary.
Corollary 8.6**.**
Suppose is a standard Lyndon pair, that is is a Gröbner-Shirshov basis of the Lie ideal in ), . Suppose . Then the universal enveloping algebra is an AS-regular algebra of global dimension Clearly, then the class contains AS-regular algebras of global dimension .
In particular, every standard Lyndon pair with a finite set is regular.
8.2. Nontrivial disconnected extensions of standard Lyndon pairs
In this subsection has arbitrary finite order . In the cases when Filiform algebras occur, , as usual.
Definition 8.7**.**
Let be a standard Lyndon pair (that is is a Gröbner-Shirshov basis of the Lie ideal in ). A Lyndon pair is called a nontrivial disconnected extension of if is disconnected and .
Proposition 8.8**.**
Suppose is a nontrivial disconnected extension of the standard Lyndon pair , where . In notation as usual, is the corresponding Lie ideal in , , and is the corresponding enveloping algebra. The following conditions hold.
- (1)
* is a -basis of .* 2. (2)
The enveloping algebra is in the class . is an s.f.p. Artin-Schelter regular algebra with . 3. (3)
Furthermore, assume that satisfies one of the following conditions: (a) N^{\star}=N(\mbox{\mathcal{L}}_{d}); (b) N^{\star}=N(\mbox{\mathcal{Q}}_{d}); (c) , (the set of all Lyndon words of length ). Then
- (a)
* is the Gröbner-Shirshov basis of the Lie ideal .* 2. (b)
The monomial Lie algebra coincides with the standard monomial algebra 3. (c)
The enveloping algebra has defining relations which form a (finite) Gröbner basis of the corresponding ideal in .
Proof.
(1) As in Proposition 8.5 denote by the Lyndon pair corresponding to the enveloping algebra . It will be enough to prove . By Proposition 8.5, and by the hypothesis one has
[TABLE]
We shall prove that that is every Lyndon word is normal modulo . We claim that a word can not be affected in the process of resolving a composition of overlap corresponding to some -chain on . Indeed, consider the set of two-chains on . Denote by the maximum of all lengths of words in . Note that if a word has length , then . Suppose is a two-chain on . Two cases are possible (i) has length , then is composed out of the Lyndon words , of lengths . It follows hat , so is a two-chain on . Then the composition of overlap induced by does not affect any of the elements , since is a standard Lyndon pair. (ii) has length . Then the composition of overlap, induced by is either [math], or it is a linear combination of elements with the same multi-degree as , . But , hence . It follows that the composition of overlap induced by can not affect any of the Lyndon words in . We have shown that every Lyndon word in is normal modulo the ideal . Therefore , which together with (8.2) implies the desired equality . This implies, that there is an equality of Lyndon pairs
[TABLE]
Hence the set is a -basis of the Lie algebra . Part (2) is straightforward. In particular .
However, as we have already discussed, in general, the set of obstructions of does not determine the Gröbner-Shirshov basis of the ideal , its elements are not necessarily Lyndon Lie monomials.
Part (3) follows from Corollary 7.7, and Corollary 7.11. We see that under the restrictive hypothesis of part (3) the set of obstructions determines explicitly the the Gröbner-Shirshov basis of the Lie ideal , and the relations of the AS regular algebra . ∎
8.3. The Fibonacci monomial algebras , and the Lie algebras
In this subsection . In [20], section 7, we define the sequence of Fibonacci-Lyndon words recursively, as follows: and, then for
[TABLE]
This give the sequence
[TABLE]
Note that if we let be and be , then the Fibonacci-Lyndon word .
Facts 8.9**.**
[20]** The following holds:
- a.
The word is a Lyndon word and its length is the ’th Fibonacci number.
- b.
For the lexicographic order we have
[TABLE]
Let consist of all Lyndon words and , where . A Lyndon word in and is not in the ideal of if and only if is a Fibonacci-Lyndon word.
Let be the antichain of all minimal elements in , with respect to the divisibility order . (This is a finite set, since is a factor of the Fibonacci-Lyndon words later in the sequence.) The corresponding monomial algebra
[TABLE]
is called the Fibonacci-Lyndon monomial algebra, [20] (or shortly, the Fibonacci algebra).
The set of Lyndon atoms with respect to the ideal in is . We shall call the corresponding Lyndon pair the th Fibonacci Lyndon pair. The monomial algebra has Hilbert series where is the ’th Fibonacci number. Clearly, the global dimension and the Gelfand-Kirillov dimension of are both . Proposition 8.1., [20] states that the Fibonacci-Lyndon monomial algebra does not deform to a bigraded Artin-Schelter regular algebra.
Corollary 8.10**.**
[20]** The class does not contain a bigraded Artin-Schelter regular algebra.
The next corollary is a particular case of Proposition 8.8, part (3), and shows that each Fibonacci pair , with induces different defining relations, , for the corresponding monomial Lie algebra , but the result is isomorphic to the same Lie algebra: \mathfrak{g}_{n}=Lie(X)/([W(F_{n})])_{Lie}\cong\mbox{\mathcal{L}}_{3}, the -dimensional Filiform Lie algebra!
Corollary 8.11**.**
Let . Let be the nth Fibonacci Lyndon pair, . Denote by the Lie ideal generated by in , , denotes the two sided ideal in is the enveloping algebra of . Then
- (1)
The class contains an AS-regular algebra occurring as an enveloping algebra of some graded Lie algebra if and only if . 2. (2)
N(F_{4})=\{x<xy<xyy<y\}=N(\mbox{\mathcal{L}}_{3}), \mathfrak{g}_{4}=Lie(X)/J_{(F_{4})}\cong\mbox{\mathcal{L}}_{3}, the standard Filiform Lie algebra of dimension . 3. (3)
Let , then the connected component of is exactly N(\mbox{\mathcal{L}}_{3}). There are equalities , , and
[TABLE]
In particular, , .
9. Two-generated AS-regular algebras of global dimensions and ,
occurring as enveloping of standard monomial Lie algebras
In this section we prove one of the main results of the paper, Theorem IV.
9.1. More properties of Lyndon pairs
Remark 9.1*.*
Suppose is a Lyndon pair in the alphabet , , let . Then , and the following conditions hold
- (1)
Assume is maximal w.r.t. , that is is not a proper subword of any (this is always so if ). Then the set satisfies conditions C1 and C2. Let be the corresponding antichain so is a Lyndon pair, (). Let , be the corresponding monomial algebra. Then every algebra satisfies . 2. (2)
Conversely, let be a Lyndon pair, with . Suppose that are such that , or, equivalently, is not in , but every Lyndon sub-word of belongs to . Then the set satisfies conditions C1 and C2. Let be the corresponding antichain of Lyndon words, so that is a Lyndon pair, one has , and every algebra satisfies .
Problem 9.2**.**
Given a full list of all Lyndon pairs with , (up to isomorphism of monomial algebras ). Find the list of all Lyndon pairs with , (up to isomorphism of monomial algebras ).
Remark 9.1 gives a base for the strategy, conditions (2) and (1) can be and will be used for inductive classification of all Lyndon pairs in , with a fixed order . The interested reader may find a procedure for an effective solution of Problem 9.2 in [23].
Till the end of the section we assume . Applying directly our strategy, we have found the sets and independently of [23], and without using computer programs. In Subsections 9.2, 9.3 we give lists of Lyndon pairs in , where and . The enumeration (6.m.j), or (7.m.j) stands for a Lyndon pair , where is the maximal length of all word . In each case when is not connected we identify the connected component , and we prove that (in these cases) is a standard Lyndon pair, corresponding to an explicitly given standard monomial Lie algebra . Note that, in general, for large values of this is not true.
9.2. A complete list of Lyndon pairs where
[TABLE]
is the connected component of .
9.3. A complete list of Lyndon pairs where
[TABLE]
The enumeration (7.m.j) indicates that . The connected component of is , denotes the free nilpotent algebra of nilpotency class , and \mbox{\mathcal{L}}_{d} is the Filiform Lie algebra of dimension .
Theorem IV**.**
Let , and be the Lyndon pairs in given in subsections 9.2 and 9.3. For each such pair is the Lie ideal of generated by , , denotes the two sided ideal in so is the enveloping algebra of .
- I
Suppose . Then
- (a)
There are eight Lyndon pairs , up to isomorphism of monomial algebras , see the list 9.2. Suppose is from this list. 2. (b)
The algebra is an Artin-Schelter regular algebra of global dimension if and only if . In this case . 3. (c)
The class does not contain any bigraded Artin-Schelter regular algebra. 4. (d)
The class , does not contain any AS regular algebra presented as an enveloping of a graded Lie algebra. 5. (e)
For the algebra is an AS regular algebra with . 2. II
Suppose . Then
- (a)
There are Lyndon pairs , up to isomorphism of monomial algebras , see the list 9.3. Suppose is a pair from this list. 2. (b)
The algebra is an Artin-Schelter regular algebra of global dimension if and only if . In this case . 3. (c)
The class , , does not contain any AS regular algebra presented as an enveloping of a graded Lie algebra. 4. (d)
For , is a finitely presented AS regular algebra of global dimension . These algebras are identified in Corollary 9.5, and Remark 9.3.
Proof.
I. (b) Consider the Lyndon pairs given in subsection 9.2. We claim that each of the sets is a Gröbner-Shirshov basis of the Lie ideal . Note that for the number is , and the minimal length of a -chain is . Indeed, is the only -chain on with length . All remaining -chains have lengths . It follows from Proposition 6.17 (3.a), that all compositions of overlap are solvable, hence is a Gröbner-Shirshov basis of the Lie ideal . For we use the results from subsections 7.2 and 7.4. The Lyndon pair determines \mathfrak{g}_{3}\simeq\mbox{\mathcal{L}}_{5}, the standard 6-dimensional Filiform Lie algebra of nilpotency class , while the pair determines \mathfrak{g}_{4}\simeq\mbox{\mathcal{Q}}_{5}, the second type Filiform algebra of dimension 6. In particular, each of the sets is a Gröbner-Shirshov basis of the Lie ideal , and the set forms a -basis of , . It follows that for all . Moreover, is an Artin-Schelter reqular algebra of . Part (d) implies that for , is not a Grobner-Shirshov basis, and . This completes the proof of (b).
(c) By Corollary 8.10 the class does not contain any bigraded AS-regular algebra.
(d) Each of the sets , is not connected, hence by Corollary 5.12, the class does not contain the enveloping algebra of any graded Lie algebra . In particular, is not a Lyndon-Shirshov basis of the Lie ideal of . For is a nontrivial disconnected extension of its connected components , which is identified explicitly in the list. We use Proposition 8.8 to identify the Lie algebra , , and . In the case (6.5.5) the connected component -the set of all Lyndon words of length . Proposition 8.8 implies that , satisfies . The set is a Gröbner-Shirshov basis of the Lie ideal . Moreover, , the free nilpotent Lie algebra of nilpotency class . Its enveloping algebra is an Artin-Schelter regular algebra of . In case (6.5.6) N_{6}^{\star}=N(\mbox{\mathcal{L}}_{4}), so is a nontrivial disconnected extension of the standard Lyndon pair (N(\mbox{\mathcal{L}}_{4}),W(\mbox{\mathcal{L}}_{4})), corresponding to the Filiform Lie algebra \mbox{\mathcal{L}}_{4}. Proposition 8.8 again implies that J_{6}=([W(\mbox{\mathcal{L}}_{4}))])_{Lie}, hence \mathfrak{g}_{6}\cong\mbox{\mathcal{L}}_{4}, and U_{6}=U\mathfrak{g}_{6}\cong U\mbox{\mathcal{L}}_{4}, U_{6}\in\mathfrak{C}(X,W(\mbox{\mathcal{L}}_{4})) is an Artin-Schelter regular algebra of In each of the cases (6.7.7) and (6.8.8) \mathfrak{g}_{i}\cong\mbox{\mathcal{L}}_{3}, so U=U\mathfrak{g}_{i}\in\mathfrak{C}(X,W(\mbox{\mathcal{L}}_{3})) is an AS regular algebra with , . Part I of the theorem has been proved.
II. (b) We shall prove that for the Lie algebra is a Gröbner-Shirshov basis of the Lie ideal . We split this in several cases.
(b.1). The cases (7.4.i), . For each of the Lyndon pairs the maximal length of a Lyndon atom is , at the same time the minimal length of a two-chain on is , so Proposition 6.17 implies that all compositions of overlap are solvable, and is a GS-basis of the Lie ideal . Therefore is an AS regular algebra of
(b.2) The Lyndon pairs , , are similar. For the maximal length of Lyndon atoms in is , and the minimal length of a two-chain on is . Proposition 6.17 implies again that is a Gröbner-Shirshov basis of the ideal , and is an AS regular algebra of .
(b.3) The Lyndon pair corresponds to the standard filiform Lie algebra , in particular is a Gröbner-Shirshov basis of the Lie ideal .
(b.4) It remains to consider only two cases: (7.5.5) and (7.5.6.) We have
[TABLE]
[TABLE]
For there is unique 2-chain of length , , on . The 2-chain implies the following composition of overlap
[TABLE]
Hence this composition is solvable. All remaining -chains on have length , i = 5,6, hence the corresponding compositions are solvable. It follows that is a GS basis of the Lie ideal .
The cases (7.x.i), where . Then each of the pairs is a nontrivial disconnected extension of a standard Lyndon pair , therefore by Corollary 5.12, is not a Lyndon-Shirshov basis of the Lie ideal of . Moreover, the class , does not contain the enveloping algebra of any graded Lie algebra . Explicit details for the Lyndon-Shirshov basis of the Lie ideals , the corresponding monomial algebras and their enveloping in these cases are given in Corollary 9.5.
It remains to study the cases (7.5.10), (7.5.11), (7.6.12), where the Lyndon pair has connected set , but is not a Gröbner-Shirshov basis of the Lie ideal . Detailed information for each of these cases is given in Remark 9.3. ∎
Remark 9.3*.*
In notation and assumptions of Theorem IV.
- (1)
The Lyndon pair given in (7.5.10) defines a Lie algebra whose enveloping has Lyndon atoms and relations , where
[TABLE]
This is exactly the Lyndon pair given by (6.4.1). Moreover, the GS- basis of the Lie ideal is , so has basis , and is dimensional, is an AS regular algebra with . 2. (2)
The Lyndon pair given in (7.5.11) defines a Lie algebra whose enveloping has Lyndon atoms and relations , where
[TABLE]
The monomial algebra is isomorphic to , corresponding to case (6.4.1). is an AS-regular algebra of . 3. (3)
For the Lyndon pair given in (7.6.12) one has \mathfrak{g}=\mathop{\text{Lie}}/([W_{12}])\simeq\mbox{\mathcal{Q}}_{5}, the second type filiform Lie algebra of dimension , U=U\mbox{\mathcal{Q}}_{5}\in\mathfrak{C}(X,W(\mbox{\mathcal{Q}}_{5})) is an AS regular algebra with .
Proof.
We give a proof of (1). Consider the pair , given in (7.5.10). For simplicity we set . The two-chain on gives a composition of overlap, , as in (9.1), so the same computations imply that . By assumption thus , and therefore the composition is not solvable. It follows that is not a Gröbner Shirshov basis of the Lie ideal . Consider . It is easy to show that the (actual) Lyndon pair corresponding to the ideal is the pair given in (9.2). But this is exactly the Lyndon pair (6.4.1) . Hence the the Gröbner-Shirshov basis of is , see (6.4.1), is dimensional and is an AS regular algebra of global dimension , .
Parts (2) and (3) are verified by similar arguments. In the case (2), (7.5.11) one shows that ; and in the case (3) (7.6.12) the composition . Analogous argument as in (1) completes the proof. ∎
Open Question 9.4**.**
Determine which of the three non-standard Lyndon pairs and , see (7.5.10), (7.5.11), (7.6.12), respectively, is regular. In other words which of the classes with ( contains the enveloping of some graded Lie algebra . In this case (i) is an s.f.p. AS-regular algebra of global dimension and the set is a -basis of .
More generally, it is a nontrivial question to decide whether given a non-standard Lyndon pair , with connected , the pair is regular.
Corollary 9.5**.**
Notation and assumptions as in Theorem IV, part II. Let be a Lyndon pair listed in subsection 9.3, denote its connected component by , is the corresponding Lyndon pair.
- (1)
In each of the cases (7.8.23), (7.9.25), (7.11.28), (7.12.29), (7.13.30)
[TABLE] 2. (2)
In each of the cases (7.7.20), (7.8.21)
[TABLE] 3. (3)
In each of the cases (7.10.26), (7.11.27)
[TABLE] 4. (4)
In each of the cases (7.7.15), (7.9.24)
[TABLE] 5. (5)
In each of the cases (7.7.16), (7.7.18), (7.8.22)
[TABLE] 6. (6)
Each of the cases (7.6.14), (7.7.19) ”degenerates” to (6.4.2), more precisely
[TABLE] 7. (7)
The case (7.7.17) ”degenerates” to (6.4.1), more precisely
[TABLE]
Proof.
Recall that by Proposition 8.8 a nontrivial disconnected extension of a standard Lyndon pair , (here ) does not ”contribute” new classes of Lie algebras, whenever is one of the following (a) N^{\star}=N(\mbox{\mathcal{L}}_{d}); (b) N^{\star}=N(\mbox{\mathcal{Q}}_{d}); (c) , the set of all Lyndon words of length . Clearly, parts (1) through (5) follow straightforwardly from Proposition 8.8. To prove parts (6) and (7) one uses first Proposition 8.8 to deduce that the correct obstruction set for is exactly . Note that every word is minimal (w.r.t.) in its multi-degree component , where , therefore is the unique Lie element in whose highest monomial is . This implies that the monomial Lie algebra is isomorphic to the standard monomial Lie algebra occurring in the list 9.2. ∎
Remark 9.6*.*
We have seen that each non-standard Lyndon pair with ”degenerates” to a standard Lyndon pair with . Moreover, in these cases the monomial Lie algebra is isomorphic to some (already known) standard monomial Lie algebra with a -basis , and the enveloping is AS-regular with However, Example 8.4 shows that when is large enough, (in this case ) a Lyndon pair , with may define a monomial Lie algebra , such that the reduced Gröbner-Shirshov basis of the Lie ideal contains Lie elements which are not Lyndon Lie monomials. This way we can obtain Artin-Schelter regular algebras with obstruction set , and a set of atoms (so ) and explicitly given more sophisticated relations which are non-trivial linear combinations of Lyndon-Lie monomials. In this case the corresponding Lyndon pair is regular, but it may be non-standard.
It is a nontrivial question to decide whether a given non-standard Lyndon pair , with connected , is regular, see Open Question 9.4.
Acknowledgements
This paper was written during my visit to Max Planck Institute for Mathematics (MPIM), Bonn in 2019. I thank MPIM for the wonderful creative and inspiring atmosphere. My cordial thanks to the referee of my paper for their amazing review with numerous corrections and suggestions for improvements.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1[1] D. Anick , Noncommutative graded algebras and their Hilbert series , J. Algebra 78 (1982), 120–140.
- 2[2] D. Anick , On monomial algebras of finite global dimension , Trans. AMS 291 (1985), 291–310.
- 3[3] D. Anick , On the homology of associative algebras , Trans. AMS 296 (1986), 641–659.
- 4[4] M. Artin and W. Schelter , Graded algebras of global dimension 3 , Adv. in Math. 66 (1987), 171–216.
- 5[5] M. Artin, J. Tate, and M. Van den Bergh , Some algebras associated to automorphisms of elliptic curves , in The Grothendiek Festschrift, Vil I, Progr. Math. 86 , Birkhäuser, Boston, 1990, 33–85.
- 6[6] M. Artin, J. Tate, and M. Van den Bergh , Modules over regular algebras of dimension 3 3 3 , Invent. Math. 106 (1991), 335–388.
- 7[7] Belmans, Pieter , On non-quadratic 4 4 4 -dimensional Artin-Schelter regular algebras and 3-folds (2017) Preprint
- 8[8] G. M. Bergman , The diamond lemma for ring theory , Adv. in Math. 29 (1978), 178–218.
