**Computing invariants and semi-invariants by means of Frobenius Lie algebras
**
**Alfons I. Ooms
**
*Mathematics Department, Hasselt University,
Agoralaan, 3590 Diepenbeek, Belgium
E-mail address: [email protected]*
Abstract. Let U(g) be the enveloping algebra of a finite dimensional Lie algebra g over a field k of characteristic zero, Z(U(g)) its center and Sz(U(g)) its semicenter. A sufficient condition is given in order for Sz(U(g)) to be a polynomial algebra over k. Surprisingly, this condition holds for many Lie algebras, especially among those for which the radical is nilpotent, in which case Sz(U(g))=Z(U(g)). In particular, it allows the explicit description of Z(U(g)) for more than half of all complex, indecomposable nilpotent Lie algebras of dimension at most 7.
1. INTRODUCTION
Let U(g) be the enveloping algebra of a nonzero finite dimensional Lie algebra g over a field k of characteristic zero, Z(U(g)) its center and Sz(U(g)) its semicenter, i.e. the subalgebra of U(g) generated by the semi-invariants of U(g). The (Poisson) semicenter Sz(S(g)) of the symmetric algebra S(g), which is defined in a similar fashion, is known to be isomorphic to Sz(U(g)) if k is algebraically closed [RV]. Unlike the center Z(U(g)) [DNOW, p. 321], the semicenter Sz(U(g)) is never reduced to k [D1, 4.9.4] and is always factorial [Moe], [LO].
In some important cases Sz(U(g)) turns out to be a polynomial algebra over k. This happens for instance if g is: semisimple [D1, 7.3.8], Frobenius (i.e. the index i(g)=0) [DNO, p. 339], square integrable (i.e. i(g)=dim Z(L)) [DNOW, p. 323], the Lie algebra of strictly lower triangular matrices [D4], any Borel subalgebra (as well as its nilradical) of a complex semisimple Lie algebra [J2, p. 263 and p. 257] and many (but unfortunately not all [Y]) parabolic (and biparabolic) subalgebras of certain complex semisimple Lie algebras [FJ1,2], [J3,4].
The key result of this paper is the following. Let g be a Lie algebra for which there exists a split torus T\subset Der\g (i.e. a commutative Lie algebra consisting of diagonalizable derivations of g) such that the semi-direct product L=T\ \oplus\g is a Frobenius Lie algebra. Let x1,…,xn be a basis of L and let v1,…,vr be the irreducible factors of the nonzero determinant det([xi,xj])∈S(g). Then v1,…,vr are the only (up to nonzero scalar multipliers) irreducible semi-invariants of S(L) and their weights λ1,…,λr∈L∗ are linearly independent over k [DNO, pp. 339-343]. We will show that S(g) and S(L) have the same semicenter, which coincides with the polynomial algebra k[v1,…,vr]. A similar result holds for the semicenters of U(g) and U(L). Under these circumstances dim T=i(g), which implies that i(g)≤rank g, the latter being the dimension of a maximal torus inside Der g. Also, the sum of the degrees of v1,…,vr is at most
c(g)=(dim g + i(g))/2, an inequality which is known to be valid in the cases mentioned above and which has recently been shown to hold in general [OV].
In section 4 we exhibit examples of different types where this simple method can be employed successfully in order to compute algebraically independent generators for Sz(U(g)).
Section 5 is devoted to complex indecomposable nilpotent Lie algebras of dimension at most seven. We use the same notation as Carles [Ca] and Magnin[Ma2], the latter providing, among other things, a maximal torus in each case. However, we also include the notation of Morozov (M_) [Mor] and Cerezo (C_) [C2] in dimension 6 and of Seeley ((…)) [Se] and Romdhani (R_) [R] in dimension 7. We distinguish two classes:
- i(g)≤rank g
Remarkably, in this situation our condition is satisfied with very few exceptions (namely, in dimension 6: g6.15 and in dimension 7: g7.3.1(iλ),λ=0,1; g7,3.12; g7,3.13 and g7,3.17. For all these g is coregular, i.e. Z(U(g)) is a polynomial algebra, except for g7,3.17).
It turns out that the generators we find for Z(U(g)) coincide with the algebraic independent generators of its quotient field listed in [C2] for dimension 6 and in [R] for dimension 7 (although in the latter quite a few cases are missing [Go, pp. 146-149] and some minor corrections have to be made). Most of the calculations were done with Maple.
As a welcome by-product we obtain all solvable, almost algebraic Frobenius Lie algebras over IC of which the nilradical is indecomposable of dimension at most seven. In each case a split torus T⊂Der g for which L=T\ \oplus\g is Frobenius is given. Sometimes, the possible solutions for T form a family depending on parameters. See also [CV].
- i(g)>rank g
In this situation, the following methods are employed in order to find the generators of S(g)\myfontsg (and hence of Z(U(g))): the procedure Dixmier used for dimension at most 5 [D3, pp. 322-330], Singular [GPS] (in case g has an abelian ideal of codimension one), a result by Panyushev [P2, Theorem 1.2] and an effective algorithm recently constructed by Michel Van den Bergh.
We confine ourselves, for now, to finish the seven remaining cases in dimension 6, of which only two are not coregular, namely g6,6 and g6,16, the latter being the 6-dimensional standard filiform Lie algebra L(6) (also called generic filiform). In fact, the n-dimensional standard filiform Lie algebra L(n) (i.e. with basis x1,…,xn and nonzero Lie brackets [x1,xi]=xi+1, i:2,…,n−1) is not coregular if and only if n≥5 [OV].
2. PRELIMINARIES AND GENERAL RESULTS
**(i) The index i(L)
**Let L be a Lie algebra over a field k of characteristic zero with basis x1,…,xn. Let f∈L∗ and consider the alternating bilinear form Bf on L sending (x,y) into f([x,y]), with kernel:
[TABLE]
We put i(L)=f∈L∗min dim L(f), the index of L. We recall from [D1, 1.14.13] that
[TABLE]
where R(L) is the quotient field of the symmetric algebra S(L) of L. In particular dimL−i(L) is an even number. Furthermore, f is called regular if dimL(f)=i(L). It is well-known that the set L\mboxreg∗ of all regular elements of L∗ is an open dense subset of L∗ for the Zariski topology.
Denote by Z(D(L)) the center of the quotient division ring D(L) of U(L) and let R(L)L be the subfield of the invariants of R(L) under the action of ad x, x∈L.
THEOREM 2.1.
[TABLE]
[O3], [RV]. Moreover, equality occurs if one of the following conditions is satisfied:
L is algebraic [O3], [RV]
k is algebraically closed and L has no proper semi-invariants in S(L) (or equivalently in U(L)) [OV, Proposition 4.1]
**(ii) The semicenter Sz(U(L))
**Let λ∈L∗. We denote by U(L)λ the set of all u∈U(L) such that ad x(u)=λ(x)u for all x∈L. Any nonzero element u∈U(L)λ is called a semi-invariant w.r.t. the weight λ. We call u a proper semi-invariant if λ=0. Clearly, U(L)λU(L)μ⊂U(L)λ+μ for all λ,μ∈L∗. The sum of the U(L)λ is direct and it is a subalgebra Sz(U(L)) of U(L), the semicenter of U(L). Sz(U(L)) is factorial and any semi-invariant can be written uniquely as a product of irreducible semi-invariants. Suppose z∈D(L),z=0. Then, z∈Z(D(L)) if and only if z can be written as a quotient of two semi-invariants of U(L) with the same weight. So, if L has no proper semi-invariants (as it is if its radical is nilpotent) then Z(D(L)) is the quotient field of Z(U(L)). A useful link with S(L) is the symmetrization map, i.e. the canonical linear isomorphism s of S(L) onto U(L), which maps each product y1,…,ym, yi∈L, into (1/m!)p∑yp(1)…yp(m), where p ranges over all permutations of {1,…,m}.
s is known to commute with each derivation of L and hence maps S(L)λ onto U(L)λ and also Sz(S(L)) onto Sz(U(L)). This restriction is an algebra isomorphism in case L is nilpotent [D1, 4.8.12] or if L is Frobenius [DNO, p. 399].
The weights of the semi-invariants of U(L) (or S(L)) form a semigroup Λ(L), which is not necessarily finitely generated [DDV]. However, the additive subgroup ΛD(L) of L∗ generated by Λ(L), is a finitely generated free abelian group [NO]. Perhaps unaware of this a new proof was recently given [FJ, p. 1519].
The centralizer of Sz(U(L)) in U(L) coincides with U(LΛ), where LΛ is the intersection of \mboxker λ, λ∈Λ(L). LΛ is a characteristic ideal of L for which the following hold [DNO, pp. 332, 343]:
THEOREM 2.2.
[TABLE]
and Z(D(LΛ)) is the quotient field of Z(U(LΛ)). Moreover, Sz(U(L))=Z(U(LΛ)) in case k is algebraically closed and either L is almost algebraic or L is Frobenius.
Similar results hold in S(L). See also [FJ, p. 1518].
**(iii) Frobenius Lie algebras
**A Lie algebra L is called Frobenius if there is a linear functional f∈L∗ such that the alternating bilinear form Bf(x,y)=f([x,y]), x,y∈L, is nondegenerate, i.e. i(L)=0. This notion was introduced in [O1] in connection with the problem of Jacobson on the characterization of Lie algebras having a primitive universal enveloping algebra. See [O2,3] for the general solution. In particular, U(L) is primitive if L is Frobenius and the converse holds if L is algebraic. It has become clear over the years that these Lie algebras form a large class. They also play an important role in different areas. For instance, many Lie subalgebras of semisimple Lie algebras, such as some parabolic and biparabolic (sea weed) subalgebras, are Frobenius [E1,2,3], [Ge], [P1], including most Borel subalgebras of simple Lie algebras [EO, p. 146]. Other examples are j-algebras which are real, solvable Lie algebras, which play an essential role in the study of bounded homogeneous domains in ICn, admitting a simply transitive solvable group of analytic automorphisms [PS]. Also, Frobenius (and quasi Frobenius) Lie algebras give rise to constant solutions of the classical Yang-Baxter equation [Dr]. More recently, Frobenius Lie algebras appear naturally in the study of simple hypersurface singularities [EK].
We now collect some useful facts on semi-invariants from [O4], [DNO]. Let L be a Frobenius Lie algebra with basis x1,…,xn. Then n is even and L has a trivial center. In fact, L is highly noncommutative since Z(D(L))=k. The Pfaffian Pf([xi,xj])∈S(L) is homogeneous of degree 21dimL and (Pf([xi,xj]))2=det([xi,xj])=0. This is well determined by L (up to a nonzero scalar multiple) and for all g∈\mboxAutL:
[TABLE]
THEOREM 2.3. Let L be Frobenius and let v1,…,vr∈S(L) be the (pairwise nonassociated) irreducible factors of Pf([xi,xj]). Then,
v1,…,vr are the only (up to nonzero scalar multipliers) irreducible semi-invariants of S(L), say with weights λ1,…,λr∈Λ(L).
Sz(S(L))=k[v1,…,vr], a polynomial algebra over k.
λ1,…,λr are linearly independent over k. They generate the semigroup Λ(L) and LΛ=∩ \mboxker λi, i:1,…,r
r=dimL−dimLΛ=i(LΛ) (see also Lemma 3.1)
COROLLARY 2.4. Let L be Frobenius. Any semi-invariant of S(L) is homogeneous. It is also a semi-invariant under the action of \mboxDer L.
**(iv) The Frobenius semiradical
**Put F(L)=f∈Lreg∗∑L(f). This is a characteristic ideal of L contained in LΛ and F(F(L))=F(L). We have the following characterization: Z(D(L))⊂D(F(L)) and if L is algebraic then F(L) is the smallest Lie subalgebra of L with this property. For example, if L is square integrable then F(L)=Z(L) and so Z(D(L))=D(Z(L)). In particular, Z(U(L))=U(Z(L)), which is a polynomial algebra. This phenomenon occurs quite frequently in section 5. Clearly, F(L)=0 if and only if L is Frobenius. Therefore F(L) is called the Frobenius semiradical of L. At the other end of the spectrum we have the Lie algebras for which F(L)=L, which we call quasi quadratic. These are unimodular and they do not possess any proper semi-invariants. They form a large class, which include all quadratic Lie algebras (and hence all abelian and all semisimple Lie algebras) [O6]. This illustrates how far Frobenius and semisimple Lie algebras are removed from each other.
REMARK. Some time ago Mustapha Rais kindly sent us an unpublished manuscript by André Cerezo [C2], in which the soul (respectively the rational soul) of a Lie algebra L is introduced and studied. This is the smallest Lie subalgebra of L whose enveloping algebra (resp. enveloping quotient division ring) contains Z(U(L)) (resp. Z(D(L))). Clearly, the rational soul of L coincides with F(L) in case L is algebraic.
**(v) The number c(L)=(\mboxdimL+i(L))/2
c(L)**, which is an integer, occurs frequently in the theory of Lie algebras.
For instance:
THEOREM 2.5. [OV] (k algebraically closed)
Assume that Sz(S(L)) is freely generated by homogeneous elements f1,…,fr. Then i=1∑rdegfi≤c(L).
The following, which is a minor extension of [OV, Corollary 1.2], is useful as a first test for coregularity.
COROLLARY 2.6. (k algebraically closed)
Assume that L has no proper semi-invariants in S(L) and that S(L)L=k[f1,…,fr] a polynomial algebra for some homogeneous fi∈S(L). Then
[TABLE]
Moreover, if equality occurs then i=1∑rdegfi=c(L) and degfi≤2, i:1,…,r.
On the other hand c(L) is also the maximum transcendence degree of a Poisson commutative subfield of R(L) [Sa]. See also [JL], [PY].
Similarly, c(L) is an upperbound for the transcendence degree of maximal subfields of D(L) and this bound can be achieved quite often. However a strictly smaller transcendence degree in D(L) (and also in R(L)) is possible as the following example demonstrates.
Let L be the solvable, almost algebraic Lie algebra over k with basis {x0,x1,…xn}, n≥2, with nonzero Lie brackets [x0,xi]=λixi, i:1,…,n, such that λ1,…,λn are linearly independent over IQ. Clearly, i(L)=n−1 and c(L)=(n+1+n−1)/2=n. Then k(x0) is a maximal (resp. Poisson) commutative subfield of D(L) (resp. of R(L)) [O5, Theorems 7, 12] and \mboxtrdegk(k(x0))=1<n=c(L).
Next, let L be an n-dimensional algebraic Lie algebra satisfying the Gelfand-Kirillov conjecture [GK], [BGR], [J1], [AOV1,2], [O7], i.e. D(L) can be generated by elements z1,…,zr, p1,…,ps, q1,…,qs∈D(L) such that [pi,pj]=0, [qi,qj]=0, [pi,qj]=δij, i,j:1,…,s and Z(D(L))=k(z1,…,zr), a purely transcendental extension of k. Then r=i(L), n=r+2s and c(L)=r+s is the transcendence degree of the maximal commutative subfield k(z1,…,zr,p1,…,ps) of D(L).
Finally, let P be a commutative Lie subalgebra of L such that dimP=c(L), i.e. P is a commutative polarization (CP) w.r.t. any f∈Lreg∗. Then D(P) is a maximal subfield of D(L) [O5, p. 706] with \mboxtrdegk(D(P))=dimP=c(L).
It is easy to see that F(L)⊂P and hence is commutative. P plays a special role in the construction of the irreducible representations of U(L) and their kernels, the primitive ideals. If in addition, P is an ideal of a completely solvable Lie algebra L, then P is a Vergne polarization of L. Furthermore, Sz(U(L))⊂U(P) and P⊂LΛ [EO, p. 141]. In section 5 we will provide for each nilpotent Lie algebra a CP-ideal (if it exists).
3. THE MAIN RESULT
LEMMA 3.1. Let L be Frobenius and H an ideal of L such that L/H is nilpotent. Then
[TABLE]
PROOF. Let f∈Lreg∗ such that its restriction f∣H ∈Hreg∗. Denote by L∞ the intersection of all terms Ci(L) of the descending central series of L. With respect to the nondegenerate bilinear form Bf(x,y)=f([x,y]), x,y∈L, we have that (L∞)⊥⊂L∞ [MO, Corollary 6.2]. Now, L/H being nilpotent implies that L∞⊂H. Hence,
[TABLE]
Consequently,
[TABLE]
□
PROPOSITION 3.2. Let H be an ideal of L such that L/H is nilpotent. Suppose v∈S(L) is a semi-invariant with weight λ∈L∗. Then S(H) contains a semi-invariant w of S(L) with the same weight.
In particular, if U(L) is primitive then each semi-invariant v of S(L) is contained in S(H) and
[TABLE]
PROOF. Since L/H is nilpotent, we can find ideals Hi of L with dimHi=i, [L,Hi]⊂Hi−1 and such that
[TABLE]
If v∈S(H) then put w=v. So, we may assume that v∈S(Hi)\S(Hi−1) with i>d. Choose x∈Hi\Hi−1. Then S(Hi)=S(Hi−1)[x]. So, v can be written as a polynomial in x with coefficients in S(Hi−1):
[TABLE]
Take any y∈L. Then ad y(x)∈Hi−1 and
λ(y)v=ad y(v)=ad y(am)xm+ terms of lower degree in x.
It follows that ad y(am)=λ(y)am for all y∈L, i.e. am∈S(L)λ∩S(Hi−1). After repeating the same reasoning a number of times, we find a semi-invariant w of S(L) with weight λ∈Λ(L), contained in S(H).
Now, suppose U(L) is primitive, which is equivalent with R(L)L=k. [O3, p. 69]. Since v and w are semi-invariants with the same weight λ∈Λ(L), we observe that
[TABLE]
for all y∈L, i.e. vw−1∈R(L)L=k Hence, for some a∈k we have v=aw∈S(H). □
LEMMA 3.3. Let λ∈L∗ be a weight. Then
[TABLE]
PROOF. The inclusion ⊃ is obvious. For ⊂ we take v∈S(L)λ and let
[TABLE]
be its unique decomposition into homogeneous components (i.e. vi∈Si(L)). Then for any x∈L:
[TABLE]
Since ad x(vi)∈Si(L) for all i we may conclude that ad x(vi)=λ(x)vi for all x∈L, i.e. vi∈Si(L)∩S(L)λ. □
PROPOSITION 3.4. Let g be a finite-dimensional Lie algebra over k. Suppose there exists a split torus T⊂Der g. Consider the semi-direct product L=T\ \oplus\g. Then,
[TABLE]
Moreover, each weight λ∈Λ(\mbox\myfontsg) can be extended to a weight λ′∈Λ(L) i.e. λ′∣\mbox\myfontsg =λ.
PROOF. Take any weight λ∈Λ(\mbox\myfontsg). By the previous lemma and the fact that Sz(S(\mbox\myfontsg))=λ⨁S(\mbox\myfontsg)λ, it suffices to show that for all m Vm,λ=Sm(\mbox\myfontsg)∩S(\mbox\myfontsg)λ is contained in Sz(S(L)). First we notice that S(\mbox\myfontsg)λ is a T-submodule of S(\mbox\myfontsg), since S(\mbox\myfontsg)λ is stable under the derivations of g [Mon, p. 265]. Next, we choose a basis x1,…,xn of g consisting of common eigenvectors for all t∈T. Then for each m Sm(\mbox\myfontsg) is a finite dimensional diagonalizable T-module (indeed the monomials x1m1…xnmn, i∑mi=m, form a basis of Sm(\mbox\myfontsg) consisting of eigenvectors for any t∈T). Hence, the same holds for the T-submodule Vm,λ. Therefore if Vm,λ=0 it contains a basis v1,…,vp such that t(vi)=μi(t)vi for all t∈T for some μi∈T∗. On the other hand, ad x(vi)=λ(x)vi for all x∈\mbox\myfontsg. Hence, each vi is a semi-invariant for L=T⊕\mbox\myfontsg, say with weight λ′∈Λ(L). Note that λ′ is an extension of λ. It follows that Vm,λ⊂Sz(S(L)). Consequently, Sz(S(\mbox\myfontsg))⊂Sz(S(L)), which implies that Sz(U(\mbox\myfontsg))⊂Sz(U(L)) (take the image under the symmetrization s). □
NOTATION. Let L be a Frobenius Lie algebra with basis x1,…,xn. We recall that the Pfaffian Pf([xi,xj])∈S(L) is a homogeneous semi-invariant of degree 21dimL with weight τ∈Λ(L) where τ(x)=\mboxtr(ad x), x∈L. Also,
[TABLE]
We now put
[TABLE]
These are well-defined, up to a nonzero scalar multiple.
THEOREM 3.5. Let g be a finite dimensional Lie algebra over k. Suppose there exists a split torus T⊂\mboxDer \mbox\myfontsg such that the semi-direct product L=T⊕\mbox\myfontsg is a Frobenius Lie algebra. Let v1,…,vr be the irreducible factors of Δ(L).
Then the following hold:
Sz(S(\mbox\myfontsg))=Sz(S(L))=k[v1,…,vr], a polynomial algebra.
Sz(U(\mbox\myfontsg))=Sz(U(L))=k[s(v1),…,s(vr)], a polynomial algebra. This coincides with Z(U(LΛ)) if k is algebraically closed.
dimT=i(\mbox\myfontsg) and r=i(LΛ)=dimL−dimLΛ
Λ(\mbox\myfontsg)={λ∣\mbox\myfontsg ∣ λ∈Λ(L)} and \mbox\myfontsgΛ=\mbox\myfontsg∩LΛ
\mboxdeg v1+…+\mboxdeg vr≤c(\mbox\myfontsg)=c(L)
If P is a CP of g, then P is also a CP of L.
If g has no proper semi-invariants in S(\mbox\myfontsg) (i.e. \mbox\myfontsg=\mbox\myfontsgΛ) then
[TABLE]
If in addition k is algebraically closed or if g is algebraic, then \mbox\myfontsg=LΛ.
PROOF. (1) By the previous proposition Sz(S(\mbox\myfontsg))⊂Sz(S(L)) and each weight λ∈Λ(\mbox\myfontsg) can be extended to a weight λ′∈Λ(L). On the other hand, U(L) is primitive as L is Frobenius. Moreover, g is an ideal of L and L/\mbox\myfontsg is abelian. Then, by Proposition 3.2 each semi-invariant of S(L) is already contained in S(\mbox\myfontsg) and Sz(S(L))⊂Sz(S(\mbox\myfontsg)). Consequently, Sz(S(\mbox\myfontsg))=Sz(S(L)), the latter being a polynomial algebra in the variables v1,…,vr by Theorem 2.3. Note also that if μ∈Λ(L) then its restriction μ∣\mbox\myfontsg ∈Λ(\mbox\myfontsg).
(2) This follows immediately from (1) and the fact that s:Sz(S(L))→Sz(U(L)) is an algebra isomorphism which maps Sz(S(\mbox\myfontsg)) onto Sz(U(\mbox\myfontsg)).
(3) Both quotients L/\mbox\myfontsg and L/LΛ [DNO, p. 330] are abelian. Hence, by Lemma 3.1 and Theorem 2.3:
[TABLE]
(4) From the proof of (1) we deduce that the map Λ(L)→Λ(\mbox\myfontsg), sending λ into its restriction λ∣\mbox\myfontsg, is surjective. It follows that
[TABLE]
(5) v1,…,vr are also the irreducible factors of Pf(L) as Δ(L)=(Pf(L))2. Hence, the sum of their degrees is at most:
[TABLE]
On the other hand, c(L)=21(dimL+i(L))=21dimL=c(\mbox\myfontsg)
(6) Let P be a CP of g, i.e. P is a commutative Lie subalgebra of g for which dimP=c(\mbox\myfontsg). Clearly, P is also a CP of L since c(\mbox\myfontsg)=c(L).
(7) S(\mbox\myfontsg)\mbox\myfontsg=Sz(S(\mbox\myfontsg)) and Z(U(\mbox\myfontsg))=Sz(U(\mbox\myfontsg)) as g does not have proper semi-invariants in S(\mbox\myfontsg) nor in U(\mbox\myfontsg). So, it suffices to apply (1) and (2).
Next, we assume in addition that either k is algebraically closed or that g is algebraic. Using (4), we deduce from \mbox\myfontsg=\mbox\myfontsgΛ that \mbox\myfontsg⊂LΛ. Hence it is enough to show that dim\mbox\myfontsg=dimLΛ. As there are no proper semi-invariants in U(\mbox\myfontsg) we see that Z(D(\mbox\myfontsg)) is the quotient field of Z(U(\mbox\myfontsg)). Therefore,
[TABLE]
In particular, r=\mboxtrdegkZ(D(\mbox\myfontsg))=i(\mbox\myfontsg) by Theorem 2.1. So,
[TABLE]
which yields dimLΛ=dim\mbox\myfontsg. □
REMARK. (1) and (2) of the theorem remain valid under the condition that U(L) is primitive (instead of requiring L to be Frobenius). In that case v1,…,vr are the irreducible factors of Δn(L), a special semi-invariant of S(L) [DNO, p. 337].
COROLLARY 3.6. See also [J2]. Let k be algebraically closed and let L be a simple Lie algebra over k of rank r with triangular decomposition L=N−⊕H⊕N. Consider the Borel subalgebra B=H⊕N. If L is not of one of the following types An, n≥2; D2t+1, t≥2; E6, then
B is Frobenius
i(N)=r, c(B)=c(N), N=BΛ
Sz(U(B))=Z(U(N)), a polynomial algebra in r generators which can be explicitly determined by the method of Theorem 3.5.
PROOF. Clearly, B can be considered as the semi-direct product of the split torus T=adNH⊂\mboxDerN with the nilradical N. Since B is Frobenius [EO, p. 146] the previous theorem can be applied. □
4. EXAMPLES
- Let g be the 5-dimensional solvable Lie algebra over k with basis x1,…,x5 and with nonvanishing Lie brackets: [x1,x3]=x3 [x1,x4]=x4 [x1,x5]=x5 [x2,x3]=x4 [x2,x4]=x5.
We want to apply Theorem 3.5. Since i(\mbox\myfontsg)=1 we need to find a 1-dimensional split torus T⊂Der \mbox\myfontsg such that the semi-direct product L=T⊕\mbox\myfontsg is Frobenius. Put T=<t> with t=\mboxdiag(0,−1,2,1,0). The matrix of Lie brackets of L is:
[TABLE]
Its determinant is Δ(L)=x52(x42−2x3x5)2 which is nonzero. So, L=T⊕\mbox\myfontsg is Frobenius. Clearly, x5 and x42−2x3x5 are the irreducible factors of Δ(L). Hence, by Theorem 3.5:
[TABLE]
which is a polynomial algebra.
Finally, by symmetrization:
[TABLE]
- The following nonsolvable examples, selected from [AOV2], have a Levi decomposition \mbox\myfontsg=S⊕R, where S=sl(2,k), with standard basis h,x,y and nonzero Lie brackets [h,x]=2x, [h,y]=−2y, [x,y]=h.
e0,e1,…,ep will be a basis of R. Wn will be the (n+1)-dimensional irreducible sl(2,k)-module, with standard basis e0,e1,…,en.
(i) \mbox\myfontsg=sl(2,k)⊕W1, with W1 abelian [AOV2, p. 554]. g is a 5-dimensional algebraic Lie algebra of index 1. Put T=<t>, t=\mboxdiag(0,0,0,1,1) and L=T⊕\mbox\myfontsg.
Its matrix of Lie brackets
[TABLE]
has determinant Δ(L)=4(e12x+e0e1h−e02y)2=0.
So, L is Frobenius. On the other hand, g has no proper semi-invariants in S(\mbox\myfontsg) as [\mbox\myfontsg,\mbox\myfontsg]=\mbox\myfontsg. By (7) of Theorem 3.5 we conclude that \mbox\myfontsg=LΛ and
[TABLE]
(ii) \mbox\myfontsg=sl(2,k)⊕R, where R=W2⊕W1 with standard basis e0,e1,e2;e3,e4. [AOV2, p. 567], [O6, p. 908]. The nontrivial action of sl(2,k) on R is given by
[h,e0]=2e0[h,e2]=−2e2[h,e3]=e3[h,e4]=−e4
[x,e1]=2e0[x,e2]=e1[x,e4]=e3 [y,e0]=e1[y,e1]=2e2[y,e3]=e4.
g is an 8-dimensional algebraic Lie algebra of index 2. Consider the split torus T=<t1,t2>; t1=\mboxdiag(0,0,0,1,1,1,0,0), t2=\mboxdiag(0,0,0,0,0,0,1,1) and L=T⊕\mbox\myfontsg.
Then Δ(L)=4(e12−4e0e2)2(e0e42−e1e3e4+e2e32)2=0. Hence, L is a 10-dimensional Frobenius Lie algebra. Clearly, [\mbox\myfontsg,\mbox\myfontsg]=\mbox\myfontsg. By (7) of Theorem 3.5 we obtain that \mbox\myfontsg=LΛ and
[TABLE]
(iii) Let R be the 5-dimensional nilpotent Lie algebra with basis e0,e1,e2,e3,e4 and nonzero Lie brackets [e2,e4]=e0 [e3,e4]=e1. sl(2,k) acts on R as follows: [h,e0]=e0 [h,e1]=−e1 [h,e2]=e2 [h,e3]=−e3 [x,e1]=e0 [x,e3]=e2 [y,e0]=e1 [y,e2]=e3.
Consider the semi-direct product \mbox\myfontsg=sl(2,k)⊕R. g is an 8-dimensional algebraic Lie algebra of index 2 [AOV2, p. 573].
Put T=<t1,t2>; t1=diag(0,0,0,1,1,0,0,1), t2=diag(0,0,0,1,1,1,1,0). Then,
[TABLE]
Hence, L is a 10-dimensional Frobenius Lie algebra. Clearly, g has no proper semi-invariants in S(\mbox\myfontsg) as R is nilpotent. By (7) of Theorem 3.5 we may conclude that \mbox\myfontsg=LΛ and after symmetrization
[TABLE]
REMARK. In the last 3 examples we notice that the algebraically independent generators of Z(D(\mbox\myfontsg)) we obtained in [AOV2] and [O6] (needed for the verification of the GK-conjecture) turn out to be the generators of Z(U(\mbox\myfontsg)).
5. INDECOMPOSABLE NILPOTENT LIE ALGEBRAS OF DIMENSION ≤7 (k=IC)
The main purpose is to describe Z=Z(U(\mbox\myfontsg)) for each Lie algebra g, but also to give the Frobenius semiradical F=F(\mbox\myfontsg) and if they exist a CP-ideal (CPI) and a torus T⊂\mboxDer \mbox\myfontsg for which the semi-direct product T ⊕ \mbox\myfontsg is Frobenius. Sometimes the possible solutions for T form a family depending on parameters ∈k. Other abbreviations are: i=i(\mbox\myfontsg), r=\mboxrank \mbox\myfontsg, c=c(\mbox\myfontsg), SQ.I. = square integrable, Q(Z)= the quotient field of Z(U(\mbox\myfontsg)). For dimension ≤6 all Lie algebras are listed, while in dimension 7 only those for which i(\mbox\myfontsg)≤\mboxrank \mbox\myfontsg.
These Lie algebras are coregular (i.e. Z(U(\mbox\myfontsg)) is a polynomial algebra over k), except \mbox\myfontsg5,5, \mbox\myfontsg6,6, \mbox\myfontsg6,16 and \mbox\myfontsg7,3.17.
Notation: We use the same notation as Magnin [Ma2] and Carles [Ca]. In addition we include the notation of Morozov (M_) [Mor] and Cerezo (C_) [C2] in dimension 6 and of Seeley ((…)) [Se] and Romdhani (R_)[R] in dimension 7.
dim\mbox\myfontsg≤5 (See also [D3, pp. 322-330])
\mbox\myfontsg3 (3-dim Heisenberg Lie algebra)
[x1,x2]=x3.
SQ.I. i=1r=2c=2Z=k[x3]F=<x3>
CPI=<x2,x3>
T=<t>, t=\mboxdiag(α,1−α,1).
\mbox\myfontsg4 (4-dim standard filiform Lie algebra)
[x1,x2]=x3[x1,x3]=x4.
i=2r=2Z=k[x4,x32−2x2x4]F=<x2,x3,x4>=CPI
T=<t1,t2>, t1=\mboxdiag(0,1,1,1), t2=\mboxdiag(1,−2,−1,0).
dim\mbox\myfontsg=5, i(\mbox\myfontsg)≤rank \mbox\myfontsg
\mbox\myfontsg5,1 (5-dim. Heisenberg Lie algebra)
[x1,x3]=x5[x2,x4]=x5.
SQ.I. i=1r=3c=3Z=k[x5]
F=<x5>CPI=<x3,x4,x5>
T=<t>,t=\mboxdiag(α,β,1−α,1−β,1).
\mbox\myfontsg5,2
[x1,x2]=x4[x1,x3]=x5.
i=3r=3c=4Z=k[x4,x5,x2x5−x3x4]
F=<x2,x3,x4,x5>=CPI
T=<t1,t2,t3>, t1=\mboxdiag(1,0,0,1,1),t2=\mboxdiag(0,1,0,1,0),t3=\mboxdiag(1,−1,−1,0,0).
\mbox\myfontsg5,3
[x1,x2]=x4[x1,x4]=x5[x2,x3]=x5
SQ.I. i=1r=2c=3Z=k[x5]F=<x5>CPI=<x3,x4,x5>
T=<t>, t=\mboxdiag(α,1−2α,2α,1−α,1).
\mbox\myfontsg5,6
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x2,x3]=x5
SQ.I. i=1r=1c=3Z=k[x5]F=<x5>CPI=<x3,x4,x5>
T=<t>, t=\mboxdiag(1,2,3,4,5).
dim\mbox\myfontsg=5,i(\mbox\myfontsg)>\mboxrank \mbox\myfontsg
\mbox\myfontsg5,4
[x1,x2]=x3[x1,x3]=x4[x2,x3]=x5.
i=3r=2c=4Z=k[x4,x5,x32+2x1x5−2x2x4]
F=\mbox\myfontsg5,4 No CP’s.
\mbox\myfontsg5,5 (5-dim. standard filiform Lie algebra)
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5.
i=3r=2c=4F=<x2,x3,x4,x5>=CPI
Z=k[x5,f1,f2,f3]f1=2x3x5−x42f2=3x2x52−3x3x4x5+x43
f3=9x22x52−18x2x3x4x5+6x2x43+8x33x5−3x32x42
Relation: f13+f22−x52f3=0 Q(Z)=k(x5,f1,f2).
dim\mbox\myfontsg=6,i(\mbox\myfontsg)≤\mboxrank \mbox\myfontsg
\mbox\myfontsg6,1≅M4≅C24
[x1,x2]=x5[x1,x4]=x6[x2,x3]=x6.
SQ.I. i=2r=3c=4Z=k[x5,x6]
F=<x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>, t1=\mboxdiag(α,1−α,α,1−α,1,1),t2=\mboxdiag(1,0,0,−1,1,0).
\mbox\myfontsg6,2≅M12≅C22
[x1,x2]=x5[x1,x5]=x6[x3,x4]=x6.
i=2r=3c=4Z=k[x6,x52−2x2x6]
F=<x2,x5,x6>CPI=<x2,x4,x5,x6>
T=<t1,t2>, t1=\mboxdiag(0,1,α,1−α,1,1), t2=\mboxdiag(1,0,2,0,1,2).
\mbox\myfontsg6,4≅M7≅C18
[x1,x2]=x4[x1,x3]=x6[x2,x4]=x5.
SQ.I. i=2r=3c=4Z=k[x5,x6]
F=<x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(1−2α,α,2α,1−α,1,1), t2=\mboxdiag(0,1,0,1,2,0).
\mbox\myfontsg6,5≅M8≅C12
[x1,x2]=x4[x1,x4]=x5[x2,x3]=x6[x2,x4]=x6.
SQ.I. i=2r=2c=4Z=k[x5,x6]
F=<x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(1,0,1,1,2,1),t2=\mboxdiag(2,−1,1,1,3,0).
\mbox\myfontsg6,7≅M6≅C19
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x2,x3]=−x6.
SQ.I. i=2r=2c=4Z=k[x5,x6]
F=<x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(0,1,0,1,0,1),t2=\mboxdiag(1,−2,2,−1,3,0).
\mbox\myfontsg6,8≅M9≅C13
[x1,x2]=x4[x1,x4]=x5[x2,x3]=x5[x2,x4]=x6.
SQ.I. i=2r=2c=4Z=k[x5,x6]
F=<x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(1,0,2,1,2,1),t2=\mboxdiag(2,−1,4,1,3,0).
\mbox\myfontsg6,9≅M14≅C16
[x1,x2]=x4[x1,x3]=x5[x2,x5]=x6[x3,x4]=x6.
i=2r=3c=4Z=k[x6,x1x6+x4x5]
F=<x1,x4,x5,x6>=CPI
T=<t1,t2>,t1=\mboxdiag(1,α,−α,1+α,1−α,1),t2=\mboxdiag(1,0,−1,1,0,0).
\mbox\myfontsg6,10≅M13≅C17
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x3,x5]=x6.
i=2r=2c=4Z=k[x6,x42−2x2x6]
F=<x2,x4,x6>CPI=<x2,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(1,1,1,2,2,3),t2=\mboxdiag(2,−4,−1,−2,1,0).
\mbox\myfontsg6,11≅M17≅C11
[x1,x2]=x4[x1,x4]=x5[x1,x5]=x6[x2,x3]=x6.
i=2r=2c=4Z=k[x6,x52−2x4x6]
F=<x4,x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(1,0,3,1,2,3),t2=\mboxdiag(−1,3,−3,2,1,0).
\mbox\myfontsg6,13≅M16≅C9
[x1,x2]=x4[x1,x4]=x5[x1,x5]=x6[x2,x3]=x5[x3,x4]=−x6.
i=2r=2c=4Z=k[x6,x53−3x4x5x6+3x2x62]
F=<x2,x4,x5,x6>=CPI
T=<t1,t2>,t1=\mboxdiag(1,−2,2,−1,0,1),t2=\mboxdiag(1,−3,2,−2,−1,0).
\mbox\myfontsg6,14≅M11≅C8
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x2,x3]=x6.
SQ.I. i=2r=2c=4Z=k[x5,x6]
F=<x5,x6>CPI=<x3,x4,x5,x6>
T=<t1,t2>,t1=\mboxdiag(1,0,1,2,3,1),t2=\mboxdiag(2,−1,1,3,5,0).
\mbox\myfontsg6,15≅M18(−1)≅C7
[x1,x2]=x3[x1,x3]=x4[x1,x5]=x6[x2,x3]=x5[x2,x4]=x6.
i=2r=2c=4Z=k[x6,x3x6−x4x5]
F=<x3,x4,x5,x6>=CPI
\mbox\myfontsg6,18≅M21≅C2
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x2,x5]=x6[x3,x4]=−x6.
i=2r=2c=4Z=k[x6,x42−2x1x6−2x3x5]
F=<x1,x3,x4,x5,x6> No CP’s
T=<t1,t2>,t1=\mboxdiag(1,−1,0,1,2,1),t2=\mboxdiag(2,−3,−1,1,3,0).
dim\mbox\myfontsg=6,i(\mbox\myfontsg)>\mboxrank \mbox\myfontsg
\mbox\myfontsg6,3≅M3≅C21
[x1,x2]=x4[x1,x3]=x5[x2,x3]=x6.
i=4r=3c=5Z=k[x4,x5,x6,x1x6−x2x5+x3x4]
F=\mbox\myfontsg6,3 No CP’s.
\mbox\myfontsg6,6≅M1≅C20
[x1,x2]=x4[x2,x3]=x6[x2,x4]=x5.
i=4r=3c=5F=<x1,x3,x4,x5,x6>=CPI
Z=k[x5,x6,f1,f2,f3],f1=x42+2x1x5,f2=x3x5−x4x6,
f3=2x1x62+2x3x4x6−x32x5,
Relation: x62f1−f22−x5f3=0Q(Z)=k(x5,x6,f1,f2).
\mbox\myfontsg6,12≅M15≅C10
[x1,x2]=x4[x1,x4]=x5[x1,x5]=x6[x2,x3]=x6[x2,x4]=x6.
i=2r=1c=4Z=k[x6,x52+2x3x6−2x4x6]
F=<x3−x4,x5,x6>CPI=<x3,x4,x5,x6>.
\mbox\myfontsg6,16≅M2≅C5 (6-dim standard filiform Lie algebra)
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x5]=x6.
i=4r=2c=5F=<x2,x3,x4,x5,x6>=CPI
Z=k[x6,f1,f2,f3,f4],f1=x52−2x4x6,f2=x53−3x4x5x6+3x3x62,f3=x42+2x2x6−2x3x5,f4=2x43+6x2x52+9x32x6−12x2x4x6−6x3x4x5.
Relation: f13−f22−3x62f1f3+x63f4=0Q(Z)=k(x6,f1,f2,f3).
\mbox\myfontsg6,17≅M19≅C4
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x5]=x6[x2,x3]=x6.i=2r=1c=4Z=k[x6,x52−2x4x6]
F=<x4,x5,x6> CPI=<x3,x4,x5,x6>
\mbox\myfontsg6,19≅M20≅C3
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x5]=x6
[x2,x3]=x5 [x2,x4]=x6.
i=2r=1c=4Z=k[x6,x53−3x4x5x6+3x3x62]F=<x3,x4,x5,x6>=CPI.
\mbox\myfontsg6,20≅M22≅C1
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x2,x3]=x5[x2,x5]=x6
[x3,x4]=−x6.
i=2r=1c=4Z=k[x6,2x53+3x42x6−6x3x5x6−6x1x62]F=<x1,x3,x4,x5,x6> No CP’s.
dim\mbox\myfontsg=7,i(\mbox\myfontsg)≤\mboxrank \mbox\myfontsg
\mbox\myfontsg7,1.03≅(13457G)≅R26
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x6]=x7[x2,x3]=x6
[x2,x4]=x7[x2,x5]=x7[x3,x4]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(0,1,1,1,1,2,2).
\mbox\myfontsg7,1.1(iλ)≅(123457I)≅R1λ,λ=0,1
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x5]=x6[x1,x6]=x7
[x2,x3]=x5[x2,x4]=x6[x2,x5]=λx7[x3,x4]=(1−λ)x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,4,5,6,7).
\mbox\myfontsg7,1.1(ii)≅(123457C)≅R3
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x5]=x6[x1,x6]=x7
[x2,x5]=x7[x3,x4]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,4,5,6,7).
\mbox\myfontsg7,1.1(iv)≅(12457I)≅R22
[x1,x2]=x3[x1,x3]=x4[x1,x5]=x6[x1,x6]=x7[x2,x3]=x5
[x2,x4]=x6[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,4,5,6,7).
\mbox\myfontsg7,1.1(v)≅(12357C)≅R42
[x1,x3]=x4[x1,x4]=x5[x1,x5]=x6[x1,x6]=x7[x2,x3]=x5
[x2,x4]=x6[x2,x5]=x7[x3,x4]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,4,5,6,7).
\mbox\myfontsg7,1.1(vi)≅(13457E)≅R38
[x1,x2]=x3[x1,x3]=x4[x1,x4]=x5[x1,x6]=x7[x2,x3]=x5
[x2,x5]=x7[x3,x4]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,4,5,6,7).
\mbox\myfontsg7,1.2(iλ)≅(1357S)(ξ=0)≅R52λ,λ=1
[x1,x2]=x4[x1,x3]=x6[x1,x4]=x5[x1,x5]=x7[x2,x3]=λx5
[x2,x4]=x6[x2,x6]=x7[x3,x4]=(1−λ)x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,1,2,2,3,3,4).
\mbox\myfontsg7,1.2(ii)≅(1357S)(ξ=0)
[x1,x2]=x4[x1,x4]=x5[x1,x5]=x7[x1,x6]=x7[x2,x3]=x6
[x2,x4]=x6[x2,x5]=x7[x3,x4]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,1,2,2,3,3,4).
\mbox\myfontsg7,1.2(iv)≅(1357H)≅R72
[x1,x2]=x4[x1,x4]=x6[x1,x5]=−x7[x1,x6]=x7[x2,x3]=x5
[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,1,2,2,3,3,4).
\mbox\myfontsg7,1.3(iλ)≅(1357N)(ξ=0)≅R62,λ=0
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x1,x6]=x7[x2,x3]=x6
[x2,x4]=λx7[x2,x5]=x7[x3,x5]=x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,2,3,3,4,5).
\mbox\myfontsg7,1.3(ii)≅(1357L)≅R63
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x1,x6]=x7[x2,x3]=x6
[x2,x4]=x7[x2,x5]=x7/2[x3,x4]=−x7/2.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,2,3,3,4,5).
\mbox\myfontsg7,1.3(iii)≅(1357F)≅R65
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x1,x6]=x7[x2,x4]=x7
[x3,x5]=x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,2,3,3,4,5).
\mbox\myfontsg7,1.3(v)≅(1357C)≅R99
[x1,x2]=x4[x1,x4]=x6[x1,x6]=x7[x2,x3]=x6[x2,x4]=x7
[x3,x4]=−x7[x3,x5]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>, t=\mboxdiag(1,2,2,3,3,4,5).
\mbox\myfontsg7,1.8≅(1357J)≅R70
[x1,x2]=x4[x1,x4]=x6[x1,x6]=x7[x2,x3]=x5[x2,x4]=x7
[x3,x5]=x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(2,4,3,6,7,8,10).
\mbox\myfontsg7,1.11≅(12457E)≅R46
[x1,x2]=x4[x1,x4]=x5[x1,x5]=x6[x1,x6]=x7[x2,x3]=x6
[x2,x4]=x6[x2,x5]=x7[x3,x4]=−x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,3,4,5,6).
\mbox\myfontsg7,1.20≅(12457D)≅R35
[x1,x2]=x3[x1,x5]=x6[x1,x6]=x7[x2,x3]=x4[x2,x4]=x6
[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=1c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1,2,3,5,6,7,8).
\mbox\myfontsg7,2.1(iλ)≅(1357M)≅R64λ,λ=0,1
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x1,x6]=x7[x2,x3]=x6
[x2,x5]=λx7[x3,x4]=(λ−1)x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,1−3α,2α,1−2α,3α,1−α,1).
\mbox\myfontsg7,2.1(ii)≅(1357D)≅R69
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x1,x6]=x7[x2,x5]=x7
[x3,x4]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,1−3α,2α,1−2α,3α,1−α,1).
\mbox\myfontsg7,2.1(iii)≅(1357A)≅R101
[x1,x2]=x4[x1,x4]=x6[x1,x6]=x7[x2,x3]=x6[x2,x5]=x7
[x3,x4]=−x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,1−3α,2α,1−2α,3α,1−α,1).
\mbox\myfontsg7,2.1(iv)≅(137D)
[x1,x3]=x5[x1,x4]=x6[x1,x6]=x7[x2,x3]=x6[x2,x5]=x7
[x3,x4]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,1−3α,2α,1−2α,3α,1−α,1).
\mbox\myfontsg7,2.2≅(147D)≅R95
[x1,x2]=x5[x1,x3]=x6[x1,x4]=2x7[x2,x3]=x4[x2,x6]=x7
[x3,x5]=−x7[x3,x6]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1−2α,α,α,2α,1−α,1−α,1).
\mbox\myfontsg7,2.10≅(13457C)≅R39
[x1,x2]=x4[x1,x3]=x7[x1,x4]=x5[x1,x5]=x6[x2,x6]=x7
[x4,x5]=−x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x3,x5,x6,x7>
T=<t>,t=\mboxdiag(−1+2α,2−3α,2−2α,1−α,α,−1+3α,1).
\mbox\myfontsg7,2.23≅(137B)≅R107
[x1,x4]=x6[x1,x6]=x7[x2,x3]=x5[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,21−α,2α,1−2α,21+α,1−α,1).
\mbox\myfontsg7,2.28≅(147B)≅R112
[x1,x2]=x5[x1,x3]=x6[x1,x6]=x7[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(1−2α,α,−1+4α,2−4α,1−α,2α,1).
\mbox\myfontsg7,2.30≅(1457B)≅R102
[x1,x2]=x5[x1,x5]=x6[x1,x6]=x7[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(51,52,α,1−α,53,54,1).
\mbox\myfontsg7,2.37≅(1357R)
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x5[x1,x6]=x7[x2,x4]=x6
[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=2c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,21−α,21,21,21+α,1−α,1).
\mbox\myfontsg7,3.1(iλ)≅(147E)≅R93λ,λ=0,1
[x1,x2]=x4[x1,x3]=x5[x1,x6]=x7[x2,x3]=x6[x2,x5]=λx7
[x3,x4]=(λ−1)x7.
SQ.I. i=1r=3c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,β,1−α−β,α+β,1−β,1−α,1).
\mbox\myfontsg7,3.1(iλ),λ=0
[x1,x2]=x4[x1,x3]=x5[x1,x6]=x7[x2,x3]=x6[x3,x4]=−x7.
i=3r=3c=5Z=k[x5,x7,x2x7−x4x6]F=<x2,x4,x5,x6,x7>=CPI
\mbox\myfontsg7,3.1(iλ),λ=1
[x1,x2]=x4[x1,x3]=x5[x1,x6]=x7[x2,x3]=x6[x2,x5]=x7.
i=3r=3c=5Z=k[x4,x7,x3x7−x5x6]F=<x3,x4,x5,x6,x7>=CPI
\mbox\myfontsg7,3.1(iii)≅(147A)≅R113
[x1,x2]=x4[x1,x3]=x5[x1,x6]=x7[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=3c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,β,1−α−β,α+β,1−β,1−α,1).
\mbox\myfontsg7,3.3≅(2457B)≅R80
[x1,x2]=x4[x1,x4]=x6[x1,x6]=x7[x2,x3]=x5.
i=3r=3c=5Z=k[x5,x7,x62−2x4x7]F=<x4,x5,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,1,0,1,1,1,1),
t2=\mboxdiag(−1,3,0,2,3,1,0),t3=\mboxdiag(0,0,1,0,1,0,0).
\mbox\myfontsg7,3.4≅(247F)≅R83
[x1,x2]=x4[x1,x3]=x5[x2,x4]=x6[x3,x5]=x7.
i=3r=3c=5Z=k[x6,x7,x42x7+2x1x6x7+x52x6]F=<x1,x4,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=\mboxdiag(1,0,0,1,1,1,1),
t2=\mboxdiag(0,1,0,1,0,2,0),t3=\mboxdiag(2,−1,−1,1,1,0,0).
\mbox\myfontsg7,3.5≅(247I)≅R86
[x1,x2]=x4[x1,x3]=x5[x2,x4]=x6[x2,x5]=x7[x3,x4]=x7.
i=3r=3c=5Z=k[x6,x7,2x1x72+2x4x5x7−x52x6]F=<x1,x4,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=\mboxdiag(1,0,0,1,1,1,1),t2=\mboxdiag(1,−1,0,0,1,−1,0),t3=\mboxdiag(2,−1,−1,1,1,0,0).
\mbox\myfontsg7,3.6≅(357A)≅R98
[x1,x2]=x4[x1,x3]=x5[x1,x5]=x7[x2,x3]=x6.
SQ.I. i=3r=3c=5Z=k[x4,x6,x7]F=<x4,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,0,1,0,1,1,1),t2=\mboxdiag(0,1,0,1,0,1,0),t3=\mboxdiag(1,2,−2,3,−1,0,0).
\mbox\myfontsg7,3.7≅(257H)≅R117
[x1,x2]=x5[x1,x5]=x6[x2,x4]=x6[x3,x4]=−x7.
i=3r=3c=5Z=k[x6,x7,2x3x62+2x2x6x7−x52x7]F=<x2,x3,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=diag(0,0,1,0,0,0,1),t2=\mboxdiag(0,1,0,0,1,1,0),t3=\mboxdiag(1,−2,−2,2,−1,0,0).
\mbox\myfontsg7,3.8≅(257A)≅R121
[x1,x2]=x5[x1,x3]=x6[x1,x5]=x7[x2,x4]=x7.
i=3r=3c=5Z=k[x6,x7,x3x7−x5x6]F=<x3,x5,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(1,0,0,2,1,1,2)t2=\mboxdiag(0,0,1,0,0,1,0),t3=\mboxdiag(1,−2,−1,2,−1,0,0).
\mbox\myfontsg7,3.9≅(257C)≅R123
[x1,x2]=x5[x1,x5]=−x7[x2,x3]=x6[x2,x4]=x7.
i=3r=3c=5Z=k[x6,x7,x3x7−x4x6]F=<x3,x4,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,1,0,0,1,1,1),t2=\mboxdiag(0,0,1,0,0,1,0),t3=\mboxdiag(1,−2,2,2,−1,0,0).
\mbox\myfontsg7,3.10≅(137C)≅R111
[x1,x2]=x5[x1,x3]=x6[x2,x4]=x6[x2,x6]=x7[x3,x5]=x7.
i=3r=3c=5Z=k[x7,x62−2x4x7,x1x7+x5x6]F=<x1,x4,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=\mboxdiag(1,0,0,1,1,1,1),t2=\mboxdiag(0,−1,1,2,−1,1,0),t3=\mboxdiag(1,0,−1,0,1,0,0).
\mbox\myfontsg7,3.11≅(257B)≅R119
[x1,x2]=x5[x1,x3]=x6[x1,x5]=x7[x2,x4]=x6.
i=3r=3c=5Z=k[x6,x7,x3x7−x5x6]F=<x3,x5,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,1,0,−1,1,0,1),t2=\mboxdiag(0,0,1,1,0,1,0),t3=\mboxdiag(−1,2,1,−2,1,0,0).
\mbox\myfontsg7,3.12≅(37D)≅R124
[x1,x2]=x5[x1,x3]=x6[x2,x4]=x6[x3,x4]=x7.
SQ.I. i=3r=3c=5Z=k[x5,x6,x7]F=<x5,x6,x7>CPI=<x1,x4,x5,x6,x7>.
\mbox\myfontsg7,3.13≅(257K)≅R105
[x1,x2]=x5[x1,x5]=x6[x2,x5]=x7[x3,x4]=x7.
i=3r=3c=5Z=k[x6,x7,x52+2x1x7−2x2x6]F=<x1,x2,x5,x6,x7> No CP’s.
\mbox\myfontsg7,3.14≅(257F)≅R120
[x1,x2]=x5[x1,x3]=x6[x1,x6]=x7[x2,x4]=x7.
i=3r=3c=5Z=k[x5,x7,x62−2x3x7]F=<x3,x5,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,0,1,1,0,1,1),t2=\mboxdiag(−1,0,2,0,−1,1,0),t3=\mboxdiag(0,1,0,−1,1,0,0).
\mbox\myfontsg7,3.15≅(257E)≅R122
[x1,x2]=x5[x1,x3]=x6[x2,x5]=x7[x3,x4]=x7.
i=3r=3c=5Z=k[x6,x7,x52+2x1x7+2x4x6]F=<x1,x4,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=\mboxdiag(1,0,0,1,1,1,1),t2=\mboxdiag(0,0,1,−1,0,1,0),t3=\mboxdiag(2,−1,−2,2,1,0,0).
\mbox\myfontsg7,3.16≅(137A)≅R108
[x1,x2]=x5[x1,x5]=x7[x3,x4]=x6[x3,x6]=x7.
i=3r=3c=5Z=k[x7,x52−2x2x7,x62−2x4x7]F=<x2,x4,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=\mboxdiag(0,1,0,1,1,1,1),t2=\mboxdiag(0,0,−1,2,0,1,0),t3=\mboxdiag(1,−2,0,0,−1,0,0).
\mbox\myfontsg7,3.17≅(1457A)≅R103
[x1,x2]=x5[x1,x5]=x6[x1,x6]=x7[x3,x4]=x7.
i=3r=3c=5F=<x2,x5,x6,x7>
CPI=<x2,x4,x5,x6,x7>Z=k[x7,f1,f2,f3]
f1=2x5x7−x62,f2=3x2x72−3x5x6x7+x63,f3=9x22x72−18x2x5x6x7+6x2x63+8x53x7−3x52x62.
Relation: f13+f22−x72f3=0Q(Z)=k(x7,f1,f2). Note that Z is isomorphic with Z(U(\mbox\myfontsg5,5)).
\mbox\myfontsg7,3.18≅(157)≅R129
[x1,x2]=x6[x1,x6]=x7[x2,x5]=x7[x3,x4]=x7.
SQ.I. i=1r=3c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,1−2α,β,1−β,2α,1−α,1).
\mbox\myfontsg7,3.19≅(27B)≅R130
[x1,x2]=x6[x1,x3]=x7[x3,x4]=x6[x4,x5]=x7.
i=3r=3c=5Z=k[x6,x7,x2x72−x3x6x7−x5x62]F=<x2,x3,x5,x6,x7>=CPI
T=<t1,t2,t3>,t1=\mboxdiag(0,0,1,−1,2,0,1),t2=\mboxdiag(0,1,0,1,−1,1,0),t3=\mboxdiag(1,−1,−1,1,−1,0,0).
\mbox\myfontsg7,3.21≅(247B)
[x1,x2]=x4[x1,x3]=x5[x1,x4]=x6[x3,x5]=x7.
i=3r=3c=5Z=k[x6,x7,x42−2x2x6]F=<x2,x4,x6,x7>CPI=<x2,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(1,0,0,1,1,2,1),t2=\mboxdiag(0,1,0,1,0,1,0),t3=\mboxdiag(2,−4,−1,−2,1,0,0).
\mbox\myfontsg7,3.22≅(247D)
[x1,x2]=x4[x1,x3]=x5[x1,x5]=x7[x2,x5]=x6[x3,x4]=x6.
i=3r=3c=5Z=k[x6,x7,x1x6−x2x7+x4x5]F=<x1,x2,x4,x5,x6,x7> No CP’s
T=<t1,t2,t3>,t1=\mboxdiag(0,0,1,0,1,1,1,1),t2=\mboxdiag(0,1,0,1,0,1,0),t3=\mboxdiag(1,1,−2,2,−1,0,0).
\mbox\myfontsg7,3.23≅(357B)
[x1,x2]=x3[x1,x3]=x5[x1,x4]=x7[x2,x3]=x6.
SQ.I. i=3r=3c=5Z=k[x5,x6,x7]F=<x5,x6,x7>CPI=<x3,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,0,0,1,0,0,1),t2=\mboxdiag(0,1,1,0,1,2,0),t3=\mboxdiag(2,−1,1,−2,3,0,0).
\mbox\myfontsg7,3.24≅(37C)
[x1,x2]=x5[x2,x3]=x6[x2,x4]=x7[x3,x4]=x5.
SQ.I. i=3r=3c=5Z=k[x5,x6,x7]F=<x5,x6,x7>CPI=<x1,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(0,0,−1,1,0,−1,1),t2=\mboxdiag(1,0,1,0,1,1,0),t3=\mboxdiag(3,−1,1,1,2,0,0).
\mbox\myfontsg7,4.1≅(37B)≅R126
[x1,x2]=x5[x1,x3]=x6[x3,x4]=x7.
SQ.I. i=3r=4c=5Z=k[x5,x6,x7]F=<x5,x6,x7>CPI=<x2,x4,x5,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(α,1−α,1−α,α,1,1,1),t2=\mboxdiag(1,0,0,0,1,1,0),t3=\mboxdiag(1,0,−1,1,1,0,0).
\mbox\myfontsg7,4.3≅(27A)≅R131
[x1,x2]=x6[x3,x5]=x6[x4,x5]=x7.
i=3r=4c=5Z=k[x6,x7,x3x7−x4x6]F=<x3,x4,x6,x7>CPI=<x1,x3,x4,x6,x7>
T=<t1,t2,t3>,t1=\mboxdiag(α,1−α,1,1,0,1,1),t2=\mboxdiag(1,0,1,0,0,1,0),t3=\mboxdiag(0,0,1,1,−1,0,0).
\mbox\myfontsg7,4.4≅(17)≅R132 (7-dim Heisenberg Lie algebra)
[x1,x4]=x7[x2,x5]=x7[x3,x6]=x7.
SQ.I. i=1r=4c=4Z=k[x7]F=<x7>CPI=<x4,x5,x6,x7>
T=<t>,t=\mboxdiag(α,β,γ,1−α,1−β,1−γ,1).
ACKNOWLEDGMENTS
We like to thank Alexander Elashvili for fruitful discussions on the subject and for his encouragement. We are also very grateful to Michel Van den Bergh for his genuine interest and for his valuable help in calculating the invariants for \mbox\myfontsg6,6 and \mbox\myfontsg6,16. Finally, we wish to thank Mustapha Rais and François Rouvière for sending us some unpublished manuscripts by André Cerezo (1945-2003). The life and work of this remarkable mathematician is commemorated in the interesting website
http://math.unice.fr/$\sim$frou/AC.html
created by François Rouvière.