This paper develops a Gr"obner--Shirshov bases theory for commutative dialgebras, providing normal forms, solvability of the word problem, and basis construction methods, extending algebraic computational tools to this class of algebras.
Contribution
It introduces a unique reduced Gr"obner--Shirshov basis theory for commutative dialgebras, including finite basis existence and applications to word problems and ideal equality.
Findings
01
Unique reduced Gr"obner--Shirshov bases exist for ideals in commutative dialgebras.
02
The word problem for finitely presented commutative dialgebras is solvable.
03
Deciding ideal equality is solvable when the generating set is finite.
Abstract
We establish Gr\"obner--Shirshov bases theory for commutative dialgebras. We show that for any ideal I of Di[X], I has a unique reduced Gr\"obner--Shirshov basis, where Di[X] is the free commutative dialgebra generated by a set X, in particular, I has a finite Gr\"obner--Shirshov basis if X is finite. As applications, we give normal forms of elements of an arbitrary commutative disemigroup, prove that the word problem for finitely presented commutative dialgebras (disemigroups) is solvable, and show that if X is finite, then the problem whether two ideals of Di[X] are identical is solvable. We construct a Gr\"obner--Shirshov basis in associative dialgebra Di⟨X⟩ by lifting a Gr\"obner--Shirshov basis in Di[X].
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.
Full text
Gröbner–Shirshov bases for commutative dialgebras∗
Yuqun Chen
School of Mathematical Sciences, South China Normal University, Guangzhou 510631, P. R. China
We establish Gröbner–Shirshov bases theory for commutative dialgebras.
We show that for any ideal I of Di[X], I has a unique reduced Gröbner–Shirshov basis, where Di[X] is the free commutative dialgebra generated by a set X, in particular, I has a finite Gröbner–Shirshov basis if X is finite. As applications,
we give normal forms of elements of an arbitrary commutative disemigroup,
prove that the word problem for finitely presented commutative dialgebras (disemigroups) is solvable, and
show that if X is finite, then the problem
whether two ideals of Di[X] are identical is solvable.
We construct a Gröbner–Shirshov basis in associative dialgebra Di⟨X⟩ by
lifting a Gröbner–Shirshov basis in Di[X].
keywords:
commutative dialgebra, commutative disemigroup, Gröbner–Shirshov basis, normal form, word problem
1 Introduction
Recently the study of algebraic properties of dialgebras has attracted considerable attention.
Dialgebras are vector spaces over a field equipped with two binary bilinear associative operations
satisfying some axioms. Moreover, if operations of a dialgebra coincide,
we obtain an associative algebra and so, dialgebras are a generalization of associative algebras.
The class of dialgebras is rather interesting despite the lack of simple examples distinct from the associative algebras.
It is well-known that for Lie algebras there is a notion of a universal enveloping associative algebra.
By the Poincareˊ-Birkhoff-Witt theorem,
to a given Lie algebra L there exists an associative algebra A
such that L is isomorphic to a subalgebra of the Lie algebra A(−).
The universal enveloping algebra for Leibniz
algebras, which are a non-commutative variation of Lie algebras, was found in Loday (1993, 1995).
Dialgebras serve as these enveloping algebras.
In Bokut et al. (2010), a Composition-Diamond lemma for dialgebras was given to obtain
normal forms for some dialgebras including the universal enveloping algebra for Leibniz
algebras. Pozhidaev (2009) studied the connection of Rota-Baxter algebras and dialgebras with associative bar-unity.
Kolesnikov (2008) proved that each dialgebra may be obtained in turn from an associative conformal algebra.
Lately normal forms of free commutative dialgebras were found by Zhuchok (2010), Zhang and Chen (2017).
Gröbner bases and Gröbner–Shirshov bases were invented independently by A.I. Shirshov
for ideals of free (commutative, anti-commutative) non-associative algebras in Shirshov (1962, 2009),
free Lie algebras in Shirshov (2009) and implicitly free associative algebras in Shirshov (2009) (see also Bergman (1978); Bokut (1976)), by
Hironaka (1964) for ideals of the power series algebras (both formal and convergent),
and by Buchberger (1970) for ideals of the polynomial algebras. Gröbner bases and
Gröbner–Shirshov bases theories have been proved to be very useful in different branches
of mathematics, including commutative algebra and combinatorial algebra. It is a powerful
tool to solve the following classical problems: normal form; word problem; conjugacy
problem; rewriting system; automaton; embedding theorem; PBW theorem; extension;
homology; growth function; Dehn function; complexity; etc. See, for example, the books by
Adams and Loustaunau (1994); Bokut and Kukin (1994); Buchberger et al. (1982); Buchberger and Winkler (1998); Cox et al. (2015); Eisenbud (1995) and the surveys by Bokut and Chen (2008, 2014); Bokut et al. (2000); Bokut and Kolesnikov (2000, 2004); Bokut and Shum (2005).
In Gröbner–Shirshov bases theory for a category of algebras, a key part is to establish
“Composition-Diamond lemma” for such algebras. The name “Composition-Diamond
lemma” combines the Diamond Lemma in Newman (1942), the Composition Lemma in Shirshov (1962)
and the Diamond Lemma in Bergman (1978).
In this paper, we establish Gröbner–Shirshov bases theory for commutative dialgebras.
A Composition Diamond lemma for commutative dialgebras is given, see Theorem 14.
We show that for a given monomial-center ordering, each ideal of Di[X] has a unique reduced Gröbner–Shirshov basis;
if X is finite, then Di[X] is Noetherian and each ideal of Di[X] has a finite Gröbner–Shirshov basis, and an algorithm is given to find such a finite (reduced) Gröbner–Shirshov basis. As applications, we give normal forms of elements of an arbitrary commutative disemigroup and prove that for finitely presented commutative dialgebras (disemigroups), the word problem and the problem
whether two ideals of Di[X] are identical are solvable.
These results will be applied to computer algebra systems,
which are referring to the common use of the Buchberger approach to polynomials.
Moreover, we
prove a theorem on the pair of algebras (Di[X],Di⟨X⟩) following the spirit of Eisenbud-Peeva-Sturmfels’ theorem
on the pair (k[X],k⟨X⟩) in Eisenbud et al. (1998). Namely, we construct a Gröbner–Shirshov
basis in associative dialgebra Di⟨X⟩ by lifting a given Gröbner–Shirshov basis S in commutative dialgebra Di[X].
The paper is organized as follows. In section 2, we review the free commutative dialgebra Di[X] generated by X over a field k.
In section 3, we establish a Composition-Diamond lemma for commutative dialgebras.
In section 4, it is shown that each ideal of a finitely generated polynomial dialgebra has a finite Gröbner–Shirshov basis.
Section 5 gives normal forms of commutative disemigroups
and shows the word problem for finitely presented commutative dialgebras (disemigroups) is solvable.
The main results in this section are similar to ones in Gröbner bases theory for commutative algebras (Buchberger, 1965, 1970; Buchberger et al., 1982; Buchberger and Winkler, 1998).
Section 6 provides a method by which we can lift commutative Gröbner–Shirshov bases to associative ones.
2 Free commutative dialgebras
Throughout the paper, we fix a field k.
Z+ stands for the set of positive integers.
Definition 1**.**
(Loday et al., 2001)*
A disemigroup (dialgebra) is a set (k-linear space) D equipped with two maps*
[TABLE]
where ⊢ and ⊣ are associative and satisfy the following identities:
for all a,b,c∈D,
[TABLE]
A disemigroup (dialgebra) (D,⊢,⊣) is commutative if both ⊢ and ⊣ are commutative.
Let X={xi∣i∈I} be a total-ordered set, X+(X∗) the free semigroup (monoid) generated by X,
[TABLE]
the free commutative semigroup without unit generated by X, and
[TABLE]
the free commutative monoid generated by X, where ε is the empty word.
For any
u=xj1xj2⋯xjn∈X+,xjk∈X, we define
[TABLE]
where {xi1,xi2,…,xin}={xj1,xj2,…,xjn} as multisets.
Write
[TABLE]
where ∣v∣ is the length of v.
For any h=⌊u⌋p∈⌊X+⌋1∪⌊XX⌋2,
we call ⌊u⌋ the associative (commutative) word of h.
For convenience, we denote ⌊u⌋1=u if u∈X.
Lemma 2**.**
(Zhuchok, 2010; Zhang and Chen, 2017)*
Let X={xi∣i∈I} be a total-ordered set and*
[TABLE]
Then Disgp[X] is the free commutative disemigroup generated by X,
where the operations ⊢ and ⊣ are as follows: for any
x,x′∈X,⌊u⌋p1,⌊v⌋p2∈⌊X+⌋1∪⌊XX⌋2 with ∣uv∣>2,
[TABLE]
Let Di[X] be the k-linear space with a k-basis ⌊X+⌋1∪⌊XX⌋2.
Then (Di[X],⊢,⊣) is the
free commutative dialgebra generated by X.
For example, if ⌊u⌋,⌊v⌋∈⌊X+⌋,⌊u⌋=⌊xi1xi2⋯xin⌋,⌊v⌋=⌊xj1xj2⌋, xil,xjk∈X, then with the notation as in Bokut et al. (2010); Zhang and Chen (2017),
[TABLE]
Let X be a well-ordered set. We define the deg-lex ordering on ⌊X+⌋ by the following:
for any ⌊u⌋=⌊xi1xi2⋯xin⌋,⌊v⌋=⌊xj1xj2⋯xjm⌋∈⌊X+⌋,
where xil,xjt∈X,
[TABLE]
An ordering > on ⌊X+⌋ is said to be monomial if > is a well ordering and
for any ⌊u⌋,⌊v⌋,⌊w⌋∈⌊X+⌋,
[TABLE]
Clearly, the deg-lex ordering is monomial.
3 Composition-Diamond lemma for commutative dialgebras
Let > be a monomial ordering on ⌊X+⌋.
We define the monomial-center ordering>d on ⌊X+⌋1∪⌊XX⌋2 as follows.
For any ⌊u⌋m,⌊v⌋n∈⌊X+⌋1∪⌊XX⌋2,
[TABLE]
In particular, if > is the deg-lex ordering on ⌊X+⌋,
we call the ordering defined by (\refequ0) the deg-lex-center ordering on ⌊X+⌋1∪⌊XX⌋2.
For simplicity of notation, we write > instead of >d when no confusion can arise.
It is clear that a monomial-center ordering is a well ordering on ⌊X+⌋1∪⌊XX⌋2. Such an ordering plays an important role in the sequel.
Here and subsequently, > is a monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2 unless otherwise stated.
For any nonzero polynomial f∈Di[X],
[TABLE]
where each 0=αi∈k,⌊ui⌋mi∈⌊X+⌋1∪⌊XX⌋2 and
⌊u1⌋m1>⋯>⌊un⌋mn.
We write
∙supp(f):={⌊u1⌋m1,⋯,⌊un⌋mn};
∙f:=⌊u1⌋m1, the leading monomial of f;
∙lt(f):=α1⌊u1⌋m1, the leading term of f;
∙lc(f):=α1, the coefficient of f;
∙f:=⌊u1⌋, the associative word of f;
∙rf:=f−lt(f).
Note that f∈⌊X+⌋.
f is called monic if lc(f)=1.
For any nonempty subset S of Di[X],
S is* monic* if s is monic for all s∈S.
For convenience we assume that ⌊u⌋>0 for any ⌊u⌋∈⌊X+⌋ and 0=0, and ⌊u⌋m>0 for any ⌊u⌋m∈⌊X+⌋1∪⌊XX⌋2.
Definition 3**.**
A nonzero polynomial f∈Di[X] is strong if f>rf.
The proof of the following proposition follows from the Definition 3.
Proposition 4**.**
A nonzero polynomial f∈Di[X] is not strong if and only if
[TABLE]
where 0=α1,α2∈k,x,x′∈X,g∈Di[X] and g<⌊xx′⌋1.
It is easy to check that
> on ⌊X+⌋1∪⌊XX⌋2 is compatible with operations ⊢ and ⊣ in the following sense:
for any ⌊u⌋m,⌊v⌋n,⌊w⌋k∈⌊X+⌋1∪⌊XX⌋2,
[TABLE]
From this, it follows that
Lemma 5**.**
Let 0=f∈Di[X] and ⌊u⌋m∈⌊X+⌋1∪⌊XX⌋2. Then
[TABLE]
In particular, if f is strong, then (f⊢⌊u⌋m)=f⊢⌊u⌋m, and
(f⊣⌊u⌋m)=f⊣⌊u⌋m.
Example 6**.**
Let X={x1,x2,x3}, x3>x2>x1, Chark=2,3 and > be the deg-lex-center ordering on
⌊X+⌋1∪⌊XX⌋2.
Let f=2⌊x2x3⌋2−2⌊x2x3⌋1+3⌊x1x3⌋2. Then
[TABLE]
The polynomial f is not strong since f=⌊x2x3⌋=rf and
[TABLE]
Here and subsequently, S denotes a monic subset of Di[X] unless otherwise stated.
Definition 7**.**
Let S be a monic subset of Di[X]. A polynomial g∈Di[X] is called a normal S-polynomial in Di[X] if either g∈S or g is one of the following:
[TABLE]
or
[TABLE]
If this is so, we also call g a normal s-polynomial.
Lemma 8**.**
Let s∈S and g be a normal s-polynomial. Then
[TABLE]
In particular, g≥s, and g=s if and only if g=s.
Proof.
The proof of (3) is straightforward.
It remains to prove that g>s if g=s. Suppose that g=s. Then g=⌊sc⌋ where ⌊c⌋∈⌊X+⌋.
We claim that g>s. Otherwise, s>g=⌊sc⌋. We have an infinite descending chain
[TABLE]
which contradicts the fact that > is a monomial ordering on ⌊X+⌋. This clearly forces g>s.
∎
Note that if g is a normal s-polynomial, then g=⌊sb⌋m for some ⌊b⌋∈⌊X∗⌋ and m∈{1,2}.
From now on, we use ⌊sb⌋m to present a normal S-polynomial, where s∈S,⌊b⌋∈⌊X∗⌋ and m∈{1,2}, i.e.
[TABLE]
It follows immediately that ⌊sb⌋m=⌊sb⌋m.
For ⌊a⌋,⌊b⌋∈⌊X+⌋, we denote the least common multiple of ⌊a⌋ and ⌊b⌋
by lcm{⌊a⌋,⌊b⌋}.
Definition 9**.**
Let f,g be two monic polynomials in Di[X].
(i)
If f is not strong, then for any x∈X,
we call both f⊢x and f⊣x the multiplication compositions of f.
2. (ii)
If f is strong, ∣f∣>1 and supp(f)∩X=∅, then for any x∈X,
we call (f⊢x)−(f⊣x) the special composition of f.
3. (iii)
If f=g and f=g, then
we call (f,g)f=f−g the equal composition of f and g.
4. (iv)
Suppose that ∣f∣=2 and ∣g∣=1.
If there exists a normal g-polynomial ⌊gx⌋m for some x∈X such that f=⌊gx⌋m, then
we call (f,g)f=f−⌊gx⌋m the short intersection composition of f and g.
5. (v)
Suppose that f,g are strong, ∣f∣+∣g∣>3 and f=g.
If f=g, then for any x∈X,
we call (f,g)⌊fx⌋1=⌊fx⌋1−⌊gx⌋1 the equal multiplication composition of f and g;
if f=g and
lcm{f,g}=⌊w⌋=⌊fa⌋=⌊gb⌋ for some
⌊a⌋∈⌊X∗⌋,⌊b⌋∈⌊X+⌋
and ∣w∣<∣f∣+∣g∣, then we call
(f,g)⌊w⌋1=⌊fa⌋1−⌊gb⌋1
the long intersection composition of f and g.
Definition 10**.**
Let S be a monic subset of Di[X].
A polynomial h∈Di[X] is called trivial moduloS,
if h=∑iαi⌊sibi⌋mi, where each αi∈k,si∈S,⌊bi⌋∈⌊X∗⌋,
and ⌊sibi⌋mi≤hifαi=0.
A monic set S is called a Gröbner–Shirshov basis in Di[X] if any
composition of polynomials in S is trivial modulo S.
S* is said to be closed under the multiplication and special compositions
if any multiplication and special composition of polynomials in S is trivial modulo S.*
For convenience, for any f,g∈Di[X] and ⌊w⌋m∈⌊X+⌋1∪⌊XX⌋2, we write
[TABLE]
which means thatf−g=∑iαi⌊sibi⌋mi, where each αi∈k,si∈S,⌊bi⌋∈⌊X∗⌋,
and ⌊sibi⌋mi<⌊w⌋m if αi=0.
We set
[TABLE]
and use Id(S) to denote the ideal of Di[X] generated by S.
Lemma 11**.**
Let S be closed under the multiplication and special compositions, ⌊sb⌋n a normal S-polynomial,
⌊a⌋m∈⌊X+⌋1∪⌊XX⌋2 and f∈Id(S).
Then
(i)
⌊sb⌋n⊣⌊a⌋m* is trivial modulo S;*
2. (ii)
⌊sb⌋n⊢⌊a⌋m* is trivial modulo S;*
3. (iii)
f=∑iαi⌊sibi⌋mi,*
where each αi∈k,si∈S,⌊bi⌋∈⌊X∗⌋.*
Proof.
(i) It suffices to show that ⌊sb⌋n⊣⌊a⌋1 is trivial modulo S.
The proof is by induction on ∣a∣.
Suppose that ∣a∣=1. The result holds trivially if s is strong.
Assume that s is not strong. Then ⌊b⌋ is empty and we are done by the triviality of multiplication composition.
Now, let ∣a∣>1 and write ⌊a⌋=y⌊a1⌋ in X∗, where y∈X,⌊a1⌋∈⌊X+⌋.
Then, by the above arguments, ⌊sb⌋n⊣⌊a⌋1=(⌊sb⌋n⊣y)⊣⌊a1⌋1 is a linear
combination of polynomials of the form ⌊s′c⌋l⊣⌊a1⌋1,
where s′∈S,⌊c⌋∈⌊X∗⌋
and ⌊s′c⌋l≤⌊sb⌋n⊣y.
By induction, ⌊s′c⌋l⊣⌊a1⌋1 is a linear
combination of normal S-polynomials ⌊sibi⌋mi
and ⌊sibi⌋mi≤⌊s′c⌋l⊣⌊a1⌋1≤⌊s′c⌋l⊣⌊a1⌋1≤(⌊sb⌋n⊣y)⊣⌊a1⌋1=⌊sb⌋n⊣⌊a⌋1,
which implies that ⌊sb⌋n⊣⌊a⌋1 is trivial modulo S.
(ii) If b=ε or ∣a∣>1, then ⌊sb⌋n⊢⌊a⌋m=⌊sb⌋n⊣⌊a⌋1
and we have done by (i). It remains to prove that s⊢a is trivial modulo S, where a∈X.
Note that s⊢a is a normal S-polynomial if ∣s∣=1. Let ∣s∣>1.
If s is not strong, then we are done by the triviality of multiplication composition.
Otherwise, s⊣a is a normal S-polynomial and s⊣a=s⊢a.
For supp(s)∩X=∅ it is easy to see that s⊢a=s⊣a and we are done.
For supp(s)∩X=∅,
by the triviality of special composition, s⊢a−s⊣a
is a linear combination of normal S-polynomials ⌊sibi⌋mi
and ⌊sibi⌋mi≤s⊢a−s⊣a<s⊢a.
Therefore the result holds.
(iii) It is clear that f is a linear combination of polynomials of the forms
[TABLE]
where s∈S,⌊a⌋m∈⌊X+⌋1∪⌊XX⌋2. Then the result follows from (i) and (ii).
∎
Lemma 12**.**
Let S be a monic subset of Di[X].
Then for any nonzero f∈Di[X],
[TABLE]
where each ⌊ui⌋ni∈Irr(S),αi,βj∈k,sj∈S,⌊bj⌋∈⌊X∗⌋, ⌊ui⌋ni≤f and
⌊sjbj⌋mj≤f.
Proof.
If f∈Irr(S),
then take ⌊u⌋n=f and f1=f−lc(f)⌊u⌋n. If f∈/Irr(S),
then f=⌊sb⌋m for some normal S-polynomial ⌊sb⌋m
and take f1=f−lc(f)⌊sb⌋m.
In both cases, we have f1<f and the result
follows from induction on f.
∎
Lemma 13**.**
Let S be a Gröbner–Shirshov basis in Di[X], ⌊s1b1⌋m1 and ⌊s2b2⌋m2 normal S-polynomials.
If ⌊w⌋m=⌊s1b1⌋m1=⌊s2b2⌋m2, then
[TABLE]
Proof.
Since ⌊w⌋m=⌊s1b1⌋m1=⌊s2b2⌋m2,
it follows that ⌊w⌋=⌊s1b1⌋=⌊s2b2⌋ and m=m1=m2.
If b1=b2=ε, then s1=s2 and we are done by the triviality of equal composition.
Suppose only one of b1 and b2 is empty, say, b1=ε and b2=ε.
Then s1=⌊s2b2⌋m2, ⌊b2⌋∈⌊X+⌋ and ∣s1∣+∣s2∣≥3.
We thus have done by the triviality of short and long intersection composition.
Suppose that b1=ε and b2=ε. Thus s1,s2 are strong. Here we need to consider two cases:
Case 1. s1 and s2 are mutually disjoint. We may assume that ⌊b1⌋=⌊s2c⌋,⌊b2⌋=⌊s1c⌋,
where ⌊c⌋∈⌊X∗⌋.
This splits into two cases depending on whether m=1 or m=2.
By (4) and Lemmas 11 and 5,
[TABLE]
[TABLE]
Case 2. s1 and s2 have a nonempty intersection.
In this case, we need to discuss three sub-cases:
Case 2.1. ∣s1∣+∣s2∣=2.
Thus s1=s2 and ⌊b1⌋=⌊b2⌋.
We have two possibilities depending on whether m=1 or m=2.
By the triviality of equal composition and Lemmas 11 and 5,
[TABLE]
Case 2.2. ∣s1∣+∣s2∣=3. Then m1=m2=1. We may assume that ∣s1∣=2 and ∣s2∣=1, i.e.
s1=⌊s2x⌋ and ⌊b2⌋=⌊xb1⌋, where x∈X.
It follows that s1=⌊s2x⌋l for some l=1,2.
By the triviality of short composition and Lemmas 11 and 5,
[TABLE]
Case 2.3. ∣s1∣+∣s2∣>3. Then m1=m2=1. If s1=s2, then ⌊b1⌋=⌊b2⌋.
We are done by the triviality of equal composition and equal multiplication composition, and Lemmas 11 and 5.
Suppose that s1=s2 and
lcm{s1,s2}=⌊w1⌋=⌊s1a1⌋=⌊s2a2⌋, where ⌊a1⌋,⌊a2⌋∈⌊X∗⌋. Then
⌊b1⌋=⌊a1c⌋,⌊b2⌋=⌊a2c⌋ for some ⌊c⌋∈⌊X∗⌋ and ∣w1∣<∣s1∣+∣s2∣. Since s1=s2, we may assume that ⌊a2⌋∈⌊X+⌋. Then we have the long composition ⌊s1a1⌋1−⌊s2a2⌋1.
By Lemmas 11 and 5,
[TABLE]
The proof is complete.
∎
The following theorem is the main result in this article.
Theorem 14**.**
(Composition-Diamond lemma for commutative dialgebras) Let S be a monic subset of Di[X],
> a monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2 and
Id(S) the ideal of Di[X] generated by S. Then the following statements are equivalent.
(i)
S* is a Gröbner–Shirshov basis in Di[X].*
2. (ii)
0=f∈Id(S)⇒f=⌊sb⌋m* for some normal *S-polynomial ⌊sb⌋m.
3. (iii)
The set
[TABLE]
is a k-basis of the commutative dialgebra Di[X∣S]:=Di[X]/Id(S).
Proof.
(i)⇒(ii). Let 0=f∈Id(S). Then by
Lemma 11f has an expression
[TABLE]
where each 0=αi∈k,⌊bi⌋∈⌊X∗⌋,si∈S. Write
⌊wi⌋mi=⌊sibi⌋mi=⌊sibi⌋mi,i=1,2,⋯.
We may assume without loss of generality that
[TABLE]
If l=1, then
f=⌊s1b1⌋m1 and
the result holds. Suppose that l≥2.
Then
[TABLE]
By Lemma 13, we can rewrite the first two summands of (5) in the form
[TABLE]
where each ⌊sj′dj⌋nj is a normal S-polynomial and ⌊sj′dj⌋nj<⌊w1⌋m1.
Thus the result follows from induction on (⌊w1⌋m1,l).
(ii)⇒(iii). By Lemma 12, the set Irr(S) generates
Di[X∣S] as a linear space. On the other hand, suppose that
h=∑iαi⌊ui⌋li=0 in Di[X∣S], where
each αi∈k, ⌊ui⌋li∈Irr(S). This means that
h∈Id(S).
Then all αi must be equal to zero. Otherwise, h=⌊uj⌋lj for some j which contradicts (ii).
(iii)⇒(i). Suppose that h is a composition
of elements of S. Clearly, h∈Id(S). By Lemma 12,
[TABLE]
where each ⌊ui⌋ni∈Irr(S),αi,βj∈k,⌊bj⌋∈⌊X∗⌋,sj∈S, ⌊ui⌋ni≤h and
⌊sjbj⌋mj≤h.
Then ∑αi⌊ui⌋ni∈Id(S).
By (iii), we have all αi=0 and
h is trivial modulo S.
∎
Buchberger–Shirshov algorithm If a monic subset S of Di[X] is not a Gröbner–Shirshov basis
then one can add to S all nontrivial compositions. Continuing this process
repeatedly, we finally obtain a Gröbner–Shirshov basis Scomp that contains S
and generates the same ideal, Id(Scomp)=Id(S).
Remark 15**.**
In Theorem 14, if ⊣=⊢, then Di[X]=k[X] is free commutative algebra generated by X and Theorem 14
is Buchberger Theorem in Buchberger (1965).
A Gröbner–Shirshov basis S in Di[X] is minimal if for any s∈S, s∈Irr(S\{s}).
A Gröbner–Shirshov basis S in Di[X] is reduced if for any s∈S, supp(s)⊆Irr(S\{s}).
If S is a Gröbner–Shirshov basis in Di[X], then we also call S a Gröbner–Shirshov basis for I=Id(S).
Lemma 16**.**
Let I be an ideal of Di[X] and S a monic subset of I. If, for all nonzero f∈I, there exists a normal S-polynomial ⌊sb⌋m such that
f=⌊sb⌋m, then I=Id(S) and S is a Gröbner–Shirshov basis for I.
Proof.
Clearly, Id(S)⊆I. For any nonzero f∈I, f=⌊sb⌋m
for some normal S-polynomial ⌊sb⌋m. Thus f1=f−lc(f)⌊sb⌋m∈I and f1<f.
By induction on f, f1∈Id(S). Hence f=f1+lc(f)⌊sb⌋m∈Id(S).
This shows that I=Id(S). Now the result follows from Theorem 14.
∎
Lemma 17**.**
Let S be a Gröbner–Shirshov basis in Di[X] and f∈S. If f∈/Irr(S\{f}),
then S\{f} is a Gröbner–Shirshov basis for Id(S).
Proof.
Let S1=S\{f}. By Lemma 16, we only need to show that for any non-zero h∈Id(S)
there exists a g∈S1 and h=⌊gb⌋m for some normal g-polynomial ⌊gb⌋m.
Indeed, h=⌊sa⌋m for some normal S-polynomial ⌊sa⌋m by Theorem 14.
If s=f, then we are done.
Suppose that s=f and f=⌊s1c⌋l for some normal S1-polynomial ⌊s1c⌋l.
When f is not strong or s1 is strong, we conclude that
[TABLE]
and we have done.
When f is strong and s1 is not strong, by Proposition 4, we have
[TABLE]
As f−s1∈Id(S) and f−s1<f, ⌊xixj⌋1=f−s1=⌊s′a′⌋1
some normal S1-polynomial ⌊s′a′⌋1 by Theorem 14.
It follows that s′∈S1 is strong and
[TABLE]
for some normal s′-polynomial ⌊s′a′a⌋m. Therefore the result holds.
∎
Lemma 18**.**
Let S be a monic subset of Di[X] and s∈S.
Then s has an expression s=s′+s′′,
where s′,s′′∈Di[X],supp(s′)⊂Irr(S\{s}),s′′∈Id(S\{s}) and
for any ⌊a⌋m∈supp(s′), ⌊a⌋m≤s.
Moreover, if S is a minimal Gröbner–Shirshov basis in Di[X], then for any s∈S, s=s′ and s′ is strong if s is strong.
Proof.
Analysis similar to that in the proof of Lemma 12 shows the first claim.
If S is a minimal Gröbner–Shirshov basis, then we have s=s′∈Irr(S\{s}) for any s∈S. Recall rs:=s−s.
It follows that rs=rs′+s′′ and rs≥rs′. Thus s′=s>rs≥rs′
and the result holds.
∎
Let S be a subset of Di[X] and ⌊a⌋m∈⌊X+⌋1∪⌊XX⌋2. We set
[TABLE]
Theorem 19**.**
Let I be an ideal of Di[X] and > a monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2.
Then there exists a unique reduced Gröbner–Shirshov basis for I.
Proof.
It is clear that {lc(f)−1f∣0=f∈I} is a Gröbner–Shirshov basis for I.
Let S be an arbitrary Gröbner–Shirshov basis for I.
For any g∈S, we set
[TABLE]
By Lemma 17, we conclude that S1 is a minimal Gröbner–Shirshov basis for I.
For any s∈S1, by Lemma 18, we have s=s′+s′′, where supp(s′)⊂Irr(S1\{s}),s′′∈Id(S1\{s}). Let
[TABLE]
We claim that
S2 is a reduced Gröbner–Shirshov basis for I. Indeed, it is clear that S2⊂Id(S1).
For any f∈Id(S1), by Theorem 14 and Lemma 18, f=⌊sb⌋m=⌊s′b⌋m
for some normal S1-polynomial ⌊sb⌋m and normal S2-polynomial ⌊s′b⌋m. According to Lemma 16,
we have S2 is a Gröbner–Shirshov basis for I.
Suppose there exists s′∈S2
such that supp(s′)⊈Irr(S2\{s′}), i.e. there exists
⌊a⌋m∈supp(s′)⊂Irr(S1\{s}) but ⌊a⌋m∈/Irr(S2\{s′}).
Then ⌊a⌋m=⌊t′b⌋m for some normal t′-polynomial ⌊t′b⌋m,
where ⌊b⌋∈⌊X∗⌋,t′∈S2\{s′}, t′=t−t′′,
t∈S1\{s} and t′′∈Id(S1\{t}).
By Lemma 18, t=t′. We must have m=1, ⌊b⌋∈⌊X+⌋,
t′ is strong and t is not strong. Otherwise
⌊tb⌋m is a normal S1\{s}-polynomial and ⌊a⌋m=⌊t′b⌋m=⌊tb⌋m,
which contradicts our assumption. Then t′=t=⌊c⌋2,
where ⌊c⌋∈⌊X+⌋,∣c∣=2. Recall rt:=t−t.
Hence rt=⌊c⌋1=t′′. There exist f∈S1\{t} and a normal f-polynomial ⌊fd⌋1 such that ⌊fd⌋1=[c]1. It follows that f is strong.
Since ⌊a⌋m=⌊a⌋1∈supp(s′), by Lemma 18, we have
f<⌊fdb⌋1=⌊cb⌋1=⌊t′b⌋1=⌊a⌋1≤s.
Hence f=s and ⌊a⌋1=⌊fdb⌋1, where ⌊fdb⌋1 is a normal S1\{s}-polynomial.
This contradicts the fact that ⌊a⌋1∈Irr(S1\{s}). We thus get supp(s′)⊆Irr(S2\{s′})
for all s′∈S2 and our claim holds.
Suppose that T is a reduced Gröbner–Shirshov basis for I. Let s0=minS2 and
r0=minT, where s0∈S2,r0∈T. By Theorem 14 and Lemma 8,
s0=⌊r′a′⌋p≥r′≥r0 for some r′∈T and
normal T-polynomial ⌊r′a′⌋p.
Similarly, r0≥s0. Then r0=s0. We claim that r0=s0.
Otherwise, 0=r0−s0∈I. We apply the above argument again, with replace s0 by r0−s0,
to obtain that r0>r0−s0≥r′′≥r0 for some r′′∈T, a contradiction.
As both T and S2 are reduced Gröbner–Shirshov bases, we have S2s0={s0}={r0}=Tr0.
Given any ⌊w⌋m∈S2∪T with ⌊w⌋m>r0.
Assume that S2<⌊w⌋m=T<⌊w⌋m.
To prove T=S2, it is sufficient to show that S2⌊w⌋m⊆T⌊w⌋m.
For any s∈S2⌊w⌋m, we can see that
s=⌊ra′′⌋q≥r for some r∈T and normal T-polynomial ⌊ra′′⌋q.
Now, we claim that ⌊w⌋m=s=r. Otherwise, ⌊w⌋m=s>r.
Then r∈T<⌊w⌋m=S2<⌊w⌋m and r∈S2\{s}.
But s=⌊ra′′⌋q, which contradicts the fact that S2 is a reduced Gröbner–Shirshov basis.
We next claim that s=r∈T⌊w⌋m. If s=r, then 0=s−r∈I.
By Theorem 14 and Lemma 8, s−r=⌊r1u⌋n=⌊s1v⌋n for some
normal T-polynomial ⌊r1u⌋n and normal S2-polynomial ⌊s1v⌋n with
r1,s1≤s−r<s=r, where r1∈T,s1∈S2.
This means that s1∈S2\{s} and r1∈T\{r}.
Noting that s−r∈supp(s)∪supp(r), we may assume that s−r∈supp(s).
As S2 is a reduced Gröbner–Shirshov basis, we have s−r∈Irr(S2\{s}),
which contradicts the fact that s−r=⌊s1v⌋n, where s1∈S2\{s}.
Thus s=r. This shows that S2⌊w⌋m⊆T⌊w⌋m.
∎
Remark 20**.**
Theorem 19 together with Buchberger–Shirshov algorithm gives a method to find the unique reduced Gröbner–Shirshov basis Sred for the ideal I=Id(S), where S is a monic subset of Di[X]. One can find Sred by the following steps.
(i)
Buchberger–Shirshov algorithm gives a Gröbner–Shirshov basis S0:=Scomp for I.
2. (ii)
Let
[TABLE]
Then S1 is a minimal Gröbner–Shirshov basis for I.
3. (iii)
For each s∈S1, by the same method as in the proof of Lemma 12, s has an expression s=s′+s′′, where supp(s′)⊂Irr(S1\{s}), s′′∈Id(S1\{s}). Then
Sred={s′∣s∈S1}
is the reduced Gröbner–Shirshov basis for I.
4 Polynomial dialgebra has a finite Gröbner–Shirshov basis
In this section, we use the Hilbert’s basis theorem to prove that the polynomial diring DiR[X] is left Noetherian if R is left Noetherian and X is a finite set.
Furthermore, we show that each ideal of the polynomial dialgebra Di[X] has a finite Gröbner–Shirshov basis, if X is a finite set.
Throughout this section, R is an associative ring with unit.
Definition 21**.**
A diring is a quaternary (T,+,⊢,⊣) such that both
(T,+,⊢) and (T,+,⊣) are associative rings with the identities (1)
in T.
Definition 22**.**
Let (D,⊢,⊣) be a disemigroup, R an associative ring with unit and T the free left R-module with an R-basis D. Then
(T,+,⊢,⊣) is a diring with a natural way: for any f=∑irifi,g=∑jrj′gj∈T,ri,rj′∈R,fi,gj∈D,
[TABLE]
Such a diring is called a disemigroup-diring of D over R.
For the free commutative disemigroup Disgp[X] generated by a set X,
we denote DiR[X] the disemigroup-diring of Disgp[X] over R
which is also called the polynomial diring over R.
In particular, Dik[X](=Di[X]) is the free commutative dialgebra (polynomial dialgebra) generated by X when k is a field.
A left (right, resp.) idealI of DiR[X] is an R-submodule of DiR[X] such that
f⊢g,f⊣g∈I(g⊢f,g⊣f∈I, resp.) for any f∈DiR[X] and g∈I.
We call I an ideal of DiR[X] if I is a left and right ideal.
Clearly, DiR[X] is a commutative disemigroup-diring if and only if R is a commutative ring.
Let U be a semigroup, R an associative ring and RU the semigroup ring of U over R. Note that RU is a free left R-module with an R-basis U.
Thus, R⌊X∗⌋ is the ring of polynomials over R in indeterminates X. It is known that R⌊X∗⌋ is left Noetherian if R is left Noetherian and X is a finite set.
Theorem 23**.**
Let X be a finite set. If R is left Noetherian, then so is DiR[X].
Proof.
Let S={⌊a⌋1∈⌊X+⌋1∣∣a∣≥3} and I=Id(S) the ideal of DiR[X] generated by S. We first prove that DiR[X]/I is left Noetherian.
It suffices to show that any left ideal H of DiR[X]/I has a finite set of generators. Since X is finite,
{⌊a⌋p∈⌊X+⌋1∪⌊XX⌋2∣∣a∣≤2} is finite, say, {⌊a⌋p∈⌊X+⌋1∪⌊XX⌋2∣∣a∣≤2}={⌊a1⌋p1,…,⌊am⌋pm} with
⌊a1⌋p1<⋯<⌊am⌋pm, where < is the deg-lex-center ordering on ⌊X+⌋1∪⌊XX⌋2.
Then DiR[X]/I=R⌊a1⌋p1+R⌊a2⌋p2+⋯+R⌊am⌋pm.
For j=1,2,…,m, let Ij be the set of elements αj∈R such that there exists an element of the form
[TABLE]
It is easy to check that Ij is a left ideal of R.
Hence each Ij has a finite set of generators and
we may assume that
Ij is generated by αj1,⋯,αjnj, where nj∈Z+. Then for 1≤i≤nj,
we have polynomials fji=αji⌊aj⌋pj+hji∈H where hji<⌊aj⌋pj.
Thus we obtain a finite set T:={fji∈H∣1≤j≤m,1≤i≤nj}.
Now we claim that the set T generates H. For any nonzero f∈H, we prove f belongs to the left ideal of DiR[X]/I generated by T, by induction on f. Assume that f=α⌊ak⌋pk+f1, where 1≤k≤m, 0=α∈R
and f1<⌊ak⌋pk.
Thus α∈Ik and α=Σiriαki, where ri∈R.
Clearly, Σirifki∈H and lt(Σirifki)=α⌊ak⌋pk=lt(f).
It follows that f1=f−Σirifki∈H and f1<f=⌊ak⌋pk.
If ⌊ak⌋pk=min{g∣0=g∈H},
then f1=0 and f=Σirifki. Hence our assertion holds.
Suppose that ⌊ak⌋pk>min{g∣g∈H}. By induction on f, we obtain the assertion.
Now, for any ascending chain of left ideals of DiR[X]
[TABLE]
we have an ascending chain of left ideals of DiR[X]/I
[TABLE]
Since DiR[X]/I is left Noetherian, it follows that there is a p∈Z+ such that (Lp+I)/I=(Lp+1+I)/I=⋯. Therefore
Lp+I=Lp+1+I=⋯. On the other hand, note that for any f∈DiR[X],h∈I, we have f⊢h=f⊣h, in particular, in I,‘‘⊢"=‘‘⊣".
Thus, I is also a left ideal of the associative ring (R⌊X∗⌋,+,⊣).
Then for the ascending chain of left ideals of R⌊X∗⌋
[TABLE]
since (R⌊X∗⌋,+,⊣) is left Noetherian, there is an l∈Z+ such that Ll∩I=Ll+1∩I=⋯.
Take n=max{p,l}. We thus get
Ln=Ln+(Ln∩I)=Ln+(Ln+1∩I)=Ln+1∩(Ln+I)=Ln+1∩(Ln+1+I)=Ln+1=….
This shows that DiR[X] is left Noetherian.
∎
Corollary 24**.**
The polynomial dialgebra Di[X] is Noetherian, if X is a finite set.
Theorem 25**.**
Let X be a finite set and I an ideal of Di[X]. Then for any monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2, I has a finite Gröbner–Shirshov basis.
Proof.
By Theorem 19, I has a minimal Gröbner–Shirshov basis S.
The leading monomial of a non-strong polynomial f∈S is ⌊xixj⌋2 for some xi,xj∈X by Proposition 4.
Since X is finite, it follows that the set {⌊xixj⌋2∣xi,xj∈X} is finite.
As S is a minimal Gröbner–Shirshov basis, we conclude that the set {s∈S∣s\mboxisnotstrong} is finite.
If S is infinite, then there exists an infinite sequence s1,s2,…, where each si∈S is strong and si=sj if i=j.
For n=1,2,…,
let In be the ideal of Di[X] generated by Tn={s1,…,sn}. Then sn+1∈In. Otherwise, sn+1 has an expression
[TABLE]
where fi,hi∈Di[X],αi∈k. This implies that either sn+1=si⊢x
for some si∈Tn with ∣si∣=1 and x∈X, or sn+1=si⊣⌊a⌋1 for some si∈Tn and
⌊a⌋∈⌊X+⌋.
Because each si is strong, we can conclude that sn+1=si⊢x=si⊢x
or sn+1=si⊣⌊a⌋1=si⊣⌊a⌋1.
It follows that
sn+1=⌊sib⌋m for some normal si-polynomial ⌊sib⌋m, which contradicts the fact that S is a minimal Gröbner–Shirshov basis.
Then we have an infinite sequence I1⊂I2⊂… which contradicts that Di[X] is Noetherian.
∎
Remark 26**.**
Under the conditions in Theorem 25, in Di[X], every minimal Gröbner–Shirshov basis is finite. In particular, every reduced Gröbner–Shirshov basis is finite.
If X is a finite set, then each congruence on the commutative disemigroup Disgp[X] is finitely generated.
Proof.
Suppose that ρ is a congruence on Disgp[X] and ρ is not finitely generated.
Then we can obtain a strictly increasing infinite chain of congruences on Disgp[X]
[TABLE]
Let Si={⌊u⌋m−⌊v⌋n∣(⌊u⌋m,⌊v⌋n)∈ρi,⌊u⌋m>⌊v⌋n},i=1,2,⋯.
It is easy to check that each Si is a Gröbner–Shirshov basis in Di[X] and
[TABLE]
Assume that ⌊u⌋m−⌊v⌋n∈Si+1⊆Id(Si+1) and ⌊u⌋m−⌊v⌋n∈/Si.
We claim that ⌊u⌋m−⌊v⌋n∈/Id(Si). Otherwise, by Theorem 14, we may assume that
[TABLE]
where each αj∈k,⌊uj⌋mj−⌊vj⌋nj∈Si and
⌊u1⌋m1>⋯>⌊ul⌋ml. This implies that α1=1,⌊u⌋m=⌊u1⌋m1.
Now we prove that ⌊u⌋m−⌊v⌋n∈Si by induction on ⌊u⌋m.
If ⌊u⌋m=min{s∣s∈Id(Si)},
then ⌊u⌋m−⌊v⌋n=⌊u1⌋m1−⌊v1⌋n1∈Si.
Otherwise, we can assume that ⌊v1⌋n1−⌊v⌋n=⌊u⌋m−⌊v⌋n−(⌊u1⌋m1−⌊v1⌋n1)=0 and ⌊v1⌋n1>⌊v⌋n. Then
[TABLE]
By induction on ⌊u⌋m, ⌊v1⌋n1−⌊v⌋n∈Si.
It follows that (⌊v1⌋n1,⌊v⌋n)∈ρi. Since (⌊u1⌋m1,⌊v1⌋n1)∈ρi,
we have (⌊u⌋m,⌊v⌋n)=(⌊u1⌋m1,⌊v⌋n)∈ρi
and ⌊u⌋m−⌊v⌋n∈Si. This contradicts our assumption that ⌊u⌋m−⌊v⌋n∈/Si.
Therefore Id(Si)⊊Id(Si+1),i=1,2,⋯, and we can obtain a infinite
properly ascending chain of ideals, which is contradicts the fact that Di[X] is Noetherian.
∎
5 Applications
5.1 Normal forms of commutative disemigroups
Recall that Disgp[X]=⌊X+⌋1∪⌊XX⌋2 is the free commutative disemigroup generated by a set X. Then
for an arbitrary commutative disemigroup D, D has an expression
[TABLE]
for some set X and S⊆Disgp[X]×Disgp[X], where ρ(S) is the congruence of Disgp[X] generated by S.
It is natural to ask how to find normal forms of elements of the commutative disemigroup D=Disgp[X∣S]?
Let > be a monomial-center ordering on Disgp[X] and
S={(⌊ui⌋mi,⌊vi⌋ni)∣⌊ui⌋mi>⌊vi⌋ni,i∈I}.
Consider the commutative dialgebra Di[X∣S], where S={⌊ui⌋mi−⌊vi⌋ni∣i∈I}.
By Buchberger–Shirshov algorithm, we have a Gröbner–Shirshov basis Scomp in Di[X] and Id(Scomp)=Id(S). It is clear that each element in Scomp is of the form ⌊u⌋m−⌊v⌋n, where ⌊u⌋m>⌊v⌋n,⌊u⌋m,⌊v⌋n∈Disgp[X].
The following theorem is an application of Theorem 14.
Theorem 28**.**
Let > be a monomial-center ordering on Disgp[X] and D=Disgp[X∣S],
where S={(⌊ui⌋mi,⌊vi⌋ni)∣⌊ui⌋mi>⌊vi⌋ni,i∈I}.
Then the set Irr(Scomp) is a set of normal forms of elements of the commutative disemigroup D=Disgp[X∣S], where Scomp is a Gröbner–Shirshov basis in Di[X] obtained from S by Buchberger–Shirshov algorithm.
5.2 Word problem for commutative dialgebras
In this subsection, we present algorithms for computing a (reduced) Gröbner–Shirshov basis.
By using these algorithms and Composition-Diamond lemma for commutative dialgebras,
we prove that for finitely presented commutative dialgebras (disemigroups), the word problem and the problem
whether two ideals of Di[X] are identical are solvable.
Let f∈Di[X] be a nonzero polynomial
and let S be a subset of Di[X]. We say f is reducible by S
if there exist a polynomial g in S and a
diword ⌊u⌋m in supp(f) such that ⌊u⌋m=⌊gb⌋m for some normal g-polynomial ⌊gb⌋m. Moreover, if f is reducible by S and α is the coefficient of ⌊u⌋m, then
f→Sf−lc(g)α⌊gb⌋m is called a one-step-reduction by S. In particular,
we say f is partially reducible by S if there exists a polynomial g in S and f=⌊gb⌋m for some normal g-polynomial ⌊gb⌋m.
If an element r∈Di[X] is
obtained from f by finitely many one-step-reductions by S and r is not reducible by S, then we say
that r is a remainder of f modulo S.
In what follows, we assume that X is a finite set. The following algorithm is an analogue of the Buchberger’s Algorithm for polynomial algebras in Buchberger (1965).
Algorithm 29**.**
(Algorithm for computing Gröbner–Shirshov bases in Di[X])
input:F={f1,⋯,fm}⊆Di[X].
output:* A Gröbner–Shirshov basis S for Id(F).*
S:={lc(f)−1f∣0=f∈F}**
G:={s,(s,g)∣s,g∈S}**
for* u∈Gdo*
if* there exists a composition of uthen*
h:=* a remainder of the composition of u modulo S*
if* h=0then*
h:=lc(h)−1h**
G:=G∪{h}∪{(h,s)∣s∈S}∪{(s,h)∣s∈S}**
S:=S∪{h}**
end* if*
end* if*
end* do*
return* S*
Lemma 30**.**
Let > be a monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2.
Then Algorithm 29 terminates after finitely many steps and returns a Gröbner–Shirshov basis in Di[X].
Proof.
(Correctness.) If Algorithm 29 terminates after finitely many steps, then its correctness follows clearly from
Definition 10.
(Termination.) Suppose to the contrary that Algorithm 29 does not terminate. Then, as the algorithm progress, we construct a set Si strictly larger than Si−1 and obtain a strictly increasing infinite sequence
[TABLE]
Each Si+1 is obtained from Si by adding h∈Id(F) to Si+1, where h is a non-zero remainder of a composition of elements in Si modulo Si. It is easy to see that h∈Irr(Si). By the method similar to that in the proof of Theorem 25, we can obtain a infinite properly ascending chain of ideals, which is contradicts the fact that Di[X] is Noetherian.
∎
By mimicking the method used in Buchberger (1965), we obtain the following algorithm.
Algorithm 31**.**
(Algorithm for computing the reduced Gröbner–Shirshov bases in Di[X])
input:* A Gröbner–Shirshov basis S={g1,⋯,gn} in Di[X].*
output:* The reduced Gröbner–Shirshov basis RG for Id(S).*
C:=S**
for* f∈Cdo*
S:=S\{f}**
if* f is not partially reducible by Sthen*
S:=S∪{f}**
end* if*
end* do*
RG:=∅**
for* h∈Sdo*
G:=S\{h}**
r:=* a remainder of h modulo G*
if* r=0then*
RG:=RG∪{r}**
end* if*
end* do*
return* RG*
The following lemma follows from Theorem 19 directly.
Lemma 32**.**
Let > be a monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2.
Then Algorithm 31 terminates after finitely many steps and returns the reduced Gröbner–Shirshov basis in Di[X].
By Theorems 19 and 25 and Algorithm 31, it follows that
Corollary 33**.**
Let X be a finite set, I an ideal of Di[X] and > a monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2.
Then there exists a unique finite reduced Gröbner–Shirshov basis for I.
Theorem 34**.**
The word problem for any finitely presented commutative dialgebra is algorithmically solvable.
Proof.
Let D=Di[X∣S] be a finitely presented commutative dialgebra. By Algorithm 29, we can obtain
a Gröbner–Shirshov basis G from S. For any f∈Di[X],
by Theorem 14, f∈Id(S) if and only if the remainder of f modulo G is equal to zero.
∎
Theorem 35**.**
Let X be a finite set, S and T two finite subsets in Di[X]. The problem whether Id(S) and Id(T) are identical is algorithmically solvable.
Proof.
By Algorithms 29 and 31, we can obtain the finite reduced Gröbner–Shirshov bases G1 for Id(S)
and G2 for Id(T) from S and T, respectively. By Corollary 33, Id(S)=Id(T) if and only if G1=G2.
∎
6 Lifting commutative Gröbner–Shirshov bases to associative ones
Let Di⟨X⟩ be the free dialgebra over a field k generated by a set X. Recall that
X+ is the free semigroup generated by X without unit. Write
[TABLE]
It is well known that [X+]ω is a k-basis of Di⟨X⟩.
We first recall from the following result
concerning the Gröbner–Shirshov bases theory for dialgebras in Zhang and Chen (2017).
Lemma 36**.**
(Zhang and Chen, 2017, Theorem 4.4)*
Let S be a monic subset of Di⟨X⟩,
> a monomial-center ordering on [X+]ω and
Id(S) the ideal of Di⟨X⟩ generated by S. Then S is a Gröbner–Shirshov basis in Di⟨X⟩ if and only if
the set*
[TABLE]
is a k-basis of the quotient dialgebra Di⟨X∣S⟩:=Di⟨X⟩/Id(S).
Now, we construct a Gröbner–Shirshov basis in Di⟨X⟩ by
lifting a Gröbner–Shirshov basis in Di[X] and adding some relations.
Let X={xi∣i∈I} be a well-ordered set. Note that
⌊X+⌋1∪⌊XX⌋2⊆[X+]ω.
We consider the natural map γ:[X+]ω→⌊X+⌋1∪⌊XX⌋2 defined by, for any [a]m∈[X+]ω,
[TABLE]
and define a k-linear map δ:Di[X]→Di⟨X⟩
which is given by
[TABLE]
Let ≻ be any monomial-center ordering on ⌊X+⌋1∪⌊XX⌋2 and >d the deg-lex-center ordering on [X+]ω.
For any [a]m,[b]n∈[X+]ω, define
[TABLE]
It is clear that > is a monomial-center ordering on [X+]ω.
In what follows, a monomial-center ordering ≻ on ⌊X+⌋1∪⌊XX⌋2 and
the ordering > on [X+]ω defined as (6) will be used unless otherwise stated.
For any ⌊a⌋m,⌊b⌋n∈⌊X+⌋1∪⌊XX⌋2,
if ⌊a⌋m≻⌊b⌋n, then
δ(⌊a⌋m)>δ(⌊b⌋n). Thus, for any f∈Di[X],
δ(f)=δ(f).
Let s∈Di[X], ⌊sb⌋m be a normal s-polynomial in Di[X].
Then δ(⌊sb⌋m)=δ(⌊sb⌋m)=δ(⌊sb⌋m).
Let S be a monic subset of Di[X], ⌊v⌋=⌊xi1xi2⋯xir⌋∈⌊X+⌋,xil∈X,1≤l≤r.
We set
[TABLE]
[TABLE]
For example,
let ⌊u⌋=x1x2x2x3x4x5x6x7x7 and ⌊v⌋=x2x6.
Then ⌊u⌋=x1x2⌊vx3x4x5⌋x7x7, where x3x4x5∈U(⌊v⌋).
That is to say, ⌊u⌋ can split into three parts by ⌊v⌋ and
the center part is ⌊va⌋ for some ⌊a⌋∈U(⌊v⌋).
Throughout the following proof, we follow the notation used in Zhang and Chen (2017).
Lemma 37**.**
Let S be a monic subset of Di[X] and Y=⌊X+⌋1∪⌊XX⌋2.
Then
[TABLE]
Proof.
Let [w]n∈[X+]ω. Then
[TABLE]
∎
Theorem 38**.**
Let X={xi∣i∈I} be a well-ordered set, S be a Gröbner–Shirshov basis in
Di[X], and W consist of the following polynomials in Di⟨X⟩:
[TABLE]
Then with the monomial-center ordering on [X+]ω defined as (6), the set
GS⋃W
is a Gröbner–Shirshov basis in Di⟨X⟩.
Proof.
Let T=GS⋃W and
Y=⌊X+⌋1∪⌊XX⌋2.
It is clear that IrrA(W)=Y.
By Theorem 14, Irr(S) is a k-basis of the commutative dialgebra Di[X∣S].
Noting that Di⟨X∣T⟩ is a commutative dialgebra, we have a homomorphism
[TABLE]
which is induced by X→Di⟨X∣T⟩,x↦x+Id(T). Since θ(S)=0, we have a homomorphism
[TABLE]
On the other hand, the following homomorphism
[TABLE]
is induced by X→Di[X∣S],x↦x+Id(S). Consider the natural homomorphism
[TABLE]
Since ξ(T)=0, we have a homomorphism ζ:Di⟨X∣T⟩→Di[X∣S] such that ζη=ξ and ζ is exactly the inverse of σ.
This shows that σ is an isomorphism.
Thus δ(Irr(S))={δ(⌊a⌋n)∣⌊a⌋n∈Irr(S)} is a k-basis of the dialgebra
Di⟨X∣T⟩. Now, for any [w]n∈[X+]ω, by Lemma 37, we have
[TABLE]
Hence, δ(Irr(S))=IrrA(T). It follows that IrrA(T) is a k-basis of the dialgebra Di⟨X∣T⟩.
By Lemma 36, T is a Gröbner–Shirshov basis in Di⟨X⟩.
∎
Acknowledgement
We wish to express our thanks to the referee for helpful suggestions and comments.
Bibliography29
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1Adams and Loustaunau (1994) W. W. Adams, P. Loustaunau, An introduction to Gröbner bases, Graduate Studies in Mathematics, vol. 3, American Mathematical Society, 1994.
2Bergman (1978) G. M. Bergman, The diamond lemma for ring theory, Adv. in Math. 29(2),178–218, 1978.
3Bokut (1976) L. A. Bokut, Imbeddings into simple associative algebras, Algebra i Logika 15(2),117–142, 245, 1976.
4Bokut and Chen (2008) L. A. Bokut, Y. Chen, Gröbner-Shirshov bases: Some new results, in: Proceedings of the Second International Congress in Algebra and Combinatorics, World Scientific, pp. 35–56, 2008.
5Bokut and Chen (2014) L. A. Bokut, Y. Chen, Gröbner-Shirshov Bases and Shirshov Algorithm, Educational tutorial lecture notes, Novosibirsk State University, 2014.
6Bokut et al. (2010) L. A. Bokut, Y. Chen, C. Liu, Gröbner-Shirshov bases for dialgebras, Internat. J. Algebra Comput. 20(3), 391–415, 2010.
7Bokut et al. (2000) L. A. Bokut, Y. Fong, V.-F. Ke, P. S. Kolesnikov, Gröbner and Gröbner-Shirshov bases in algebra, and conformal algebras, Fundam. Prikl. Mat. 6(3), 669–706, 2000.
8Bokut and Kolesnikov (2000) L. A. Bokut, P. S. Kolesnikov, Gröbner-Shirshov bases: from inception to the present time, Zap. Nauchn. Sem. S.-Peterburg. Otdel. Mat. Inst. Steklov. (POMI) 272(Vopr. Teor. Predst. Algebr i Grupp. 7) 26–67, 345, 2000.