This paper classifies a specific class of 2-step nilpotent Lie algebras associated with Clifford algebra representations, extending previous work to a complete understanding based on module dimensions and signatures.
Contribution
It provides a complete classification of pseudo $H$-type algebras related to Clifford modules, identifying isomorphism classes through module dimensions and signatures.
Findings
01
Complete classification of pseudo $H$-type algebras achieved
02
Isomorphism classes determined by module dimension and signature
03
Extended previous partial classifications to a full classification
Abstract
We classify a class of 2-step nilpotent Lie algebras related to the representations of the Clifford algebras in the following way. Let J:\Cl(Rr,s)\toU be a representation of the Clifford algebra \Cl(Rr,s) generated by the pseudo Euclidean vector space Rr,s. Assume that the Clifford module U is endowed with a bilinear symmetric non-degenerate real form \la⋅,⋅\raU making the linear map Jz skew symmetric for any z∈Rr,s. The Lie algebras and the Clifford algebras are related by \laJzv,w\raU=\laz,[v,w]\raRr,s, z∈Rr,s, v,w∈U. We detect the isomorphic and non-isomorphic Lie algebras according to the dimension of U and the range of the non-negative integers~r,s.
Tables7
Table 1. Table 1. Dimensions of minimal admissible modules
8
64
128
7
16
5
4
3
2
1
0
s/r
0
1
2
3
4
5
6
7
8
Table 2. Table 2. Classification result after the first step
7
d
d
d
6
d
h
5
d
h
4
h
h
h
3
d
d
d
d
2
h
d
h
1
d
d
h
0
h
h
h
h
0
1
2
3
4
5
6
7
8
Table 3. Table 3. Final result of the classification
7
6
5
4
3
2
1
0
0
1
2
3 isotyp
3 nonisotyp
4
5
6
7 isotyp
7 nonisotyp
8
Table 4. Table 4. Systems P I r , 0 𝑃 subscript 𝐼 𝑟 0 PI_{r,0} and C O r , 0 𝐶 subscript 𝑂 𝑟 0 CO_{r,0} , r = 4 , … , 7 𝑟 4 … 7 r=4,\ldots,7
isom
isom
isom
Table 5. Table 5. Systems P I r , 4 𝑃 subscript 𝐼 𝑟 4 PI_{r,4} and C O r , 4 𝐶 subscript 𝑂 𝑟 4 CO_{r,4} , r = 0 , 1 , 2 𝑟 0 1 2 r=0,1,2
isom
anti-isom
isom
Table 6. Table 6. Systems P I 3 , s 𝑃 subscript 𝐼 3 𝑠 PI_{3,s} and C O 3 , s 𝐶 subscript 𝑂 3 𝑠 CO_{3,s} , s = 1 , … , 7 𝑠 1 … 7 s=1,\ldots,7 and P I 7 , s 𝑃 subscript 𝐼 7 𝑠 PI_{7,s} , C O 7 , s 𝐶 subscript 𝑂 7 𝑠 CO_{7,s} , s = 1 , 2 , 3 𝑠 1 2 3 s=1,2,3
isom
isom
isom
anti-isom
Table 7. Table 7. Systems P I r , s 𝑃 subscript 𝐼 𝑟 𝑠 PI_{r,s} and C O r , s 𝐶 subscript 𝑂 𝑟 𝑠 CO_{r,s} for Proposition 2.6.2
⎩⎨⎧P1=Jzi1Jzi2Jzi3Jzi4,where allzikare either positive or negative,P2=Jzi1Jzi2Jzi3Jzi4,where two zil are positive and two are negative,P3=Jzi1Jzi2Jzi3,where all three zik are positive,P4=Jzi1Jzi2Jzi3,where one zil is positive and two are negative.
⎩⎨⎧P1=Jzi1Jzi2Jzi3Jzi4,where allzikare either positive or negative,P2=Jzi1Jzi2Jzi3Jzi4,where two zil are positive and two are negative,P3=Jzi1Jzi2Jzi3,where all three zik are positive,P4=Jzi1Jzi2Jzi3,where one zil is positive and two are negative.
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
TopicsAdvanced Algebra and Geometry · Algebraic and Geometric Analysis · Advanced Topics in Algebra
Full text
Complete classification of pseudo H-type algebras: II
Kenro Furutani, Irina Markina
K. Furutani: Department of Mathematics, Faculty of Science
and Technology, Science University of Tokyo, 2641 Yamazaki, Noda, Chiba (278-8510), Japan
We classify a class of 2-step nilpotent Lie algebras related to the representations of the Clifford algebras in the following way. Let J:Cl(Rr,s)→End(U) be a representation of the Clifford algebra Cl(Rr,s) generated by the pseudo Euclidean vector space Rr,s. Assume that the Clifford module U is endowed with a bilinear symmetric non-degenerate real form ⟨⋅,⋅⟩U making the linear map Jz skew symmetric for any z∈Rr,s. The Lie algebras and the Clifford algebras are related by ⟨Jzv,w⟩U=⟨z,[v,w]⟩Rr,s, z∈Rr,s, v,w∈U. We detect the isomorphic and non-isomorphic Lie algebras according to the dimension of U and the range of the non-negative integers r,s.
The first author was partially
supported by the Grant-in-aid for Scientific Research (C) No. 26400124, JSPS.,
and the second author was partially supported by ISP project 239033/F20 of Norwegian Research Council, as well as EU FP7 IRSES program STREVCOMS, grant no. PIRSES-GA-2013-612669
The present work is a continuation of [18] and studies 2-step nilpotent graded Lie algebras associated to the representations of Clifford algebras. Let Clr,s be the Clifford algebra generated by the pseudo Euclidean vector space Rr,s and let J:Clr,s→End(U) be its representation. Assume that there exists a non-degenerate symmetric bi-linear form ⟨⋅,⋅⟩U on the Clifford module U such that
⟨Jzx,y⟩U+⟨x,Jzy⟩U=0 for all z∈Rr,s and x,y∈U. The pair (U,⟨⋅,⋅⟩U) is called an admissible module. The set U⊕Rr,s, endowed with the Lie bracket defined by ⟨z,[x,y]⟩Rr,s=⟨Jzx,y⟩U for x,y∈U and zero otherwise, is called a pseudo H-type Lie algebra and is denoted by Nr,s(U). The pseudo H-type Lie algebras Nr,0(U) were introduced in [22] and their generalisations Nr,s(U) appeared in [11, 19]. The pseudo H-type Lie algebras were actively studied for instance in [5, 12, 17, 18, 23]. They provide a setting for the study of sub-elliptic, hypo-elliptic and Grushin type operators, see for instance [9, 10] and it is an important particular case of the extended Poincaré Lie super algebras [1, 2]. The Lie groups of pseudo H-type Lie algebras constitute a source of interesting examples of sub-Riemannian manifolds [7, 20], nil-manifolds [14, 25], iso-spectral but non-diffeomorphic manifolds [6], Damek-Ricci harmonic spaces [8], symmetric spaces of rank one, [13, 29].
The authors considered in [18] the classification of pseudo H-type Lie algebras whose constructions are based on the minimal dimensional admissible modules Vminr,s. It has been showen that two Lie algebras Nr,s(Vminr,s) and Nr~,s~(Vminr~,s~) are never isomorphic unless r=r~ and s=s~, or r=s~ and s=r~. Among the couples Nr,s(Vminr,s) and Ns,r(Vmins,r) there are isomorphic and non-isomorphic Lie algebras.
The present paper is a continuation of [18] and it finishes the classification of the pseudo H-type Lie algebras Nr,s(U), where U is not necessary minimal dimensional admissible module. The first step of the classification depends on the fact whether the Clifford algebra Clr,s is simple or not. If the Clifford algebra Clr,s is simple, then the Lie algebra Nr,s(U) for r=3(mod4) is uniquely defined by the dimension of the admissible module U and does not depend on the choice of the scalar product on U. As a consequence in this case we obtain that the Lie algebras Nr,s(U) and Ns,r(U~) are isomorphic if dim(U)=dim(U~). If r=3(mod4) then the Lie algebra Nr,s(U) depends on the choice of the scalar product on each minimal dimensional component (Vminr,s)i in the decomposition U=⊕(Vminr,s)i. If r−s=3(mod4), then Clifford algebras Clr,s are not simple and the classification is more complicate. Recall, that the Lie algebras Nr,0(U) for r=3(mod4) are defined only by number of non-equivalent irreducible terms in the decomposition of U into the direct sum of irreducible submodules and any irreducible Clifford module is actually an admissible module, see [8, 13]. For the pseudo H-type Lie algebras Nr,s(U), s>0, and r=3(mod4) the classification is more subtle and depends not only on the number of different minimal dimensional modules, but also on the choice of the scalar product on them. These phenomena come from the signature of the scalar product
restricted to the “common 1-eigenspace” of a set of maximal number of mutually
commuting symmetric isometric involutions of the Clifford action on the minimal dimensional module. It is also closely related to the existence or non-existence of a special type of an automorphism of the Lie algebra Nr,s(U) which is identity on the centre.
The structure of the paper is the following. We recall basic properties of Clifford algebras, such as, periodicity, the system of involutions, the structure of admissible modules, and other information needed to complete classification of pseudo H-type Lie algebras in Section 2. Section 3 is dedicated to the description of pseudo H-type Lie algebras and the structure of their isomorphisms and automorphisms.
The main result is contained in Theorems 4.1.1–4.1.3, see Section 4. In Section 5, we present Tables 4–7 needed to determine important properties of minimal admissible modules for basic cases (2.8), which are summarised in Table 1 in Section 2.5.
2. Clifford algebras and admissible modules
In the section we collect the information
about Clifford algebras and their admissible modules
that we need for the classification of pseudo H-type Lie algebras.
2.1. Definitions of Clifford algebras
We denote by Rr,s the space Rk, r+s=k, with
the non-degenerate quadratic form
Qr,s(z)=∑i=1rzi2−∑j=1szr+j2, z∈Rn of the signature (r,s).
The non-degenerate bi-linear form obtained from Qr,s by polarization
is denoted by ⟨⋅,⋅⟩r,s.
We call the form
⟨⋅,⋅⟩r,s a scalar product.
A vector z∈Rr,s is called
positive if ⟨z,z⟩r,s>0,
negative if ⟨z,z⟩r,s<0, and null
if ⟨z,z⟩r,s=0.
We use the orthonormal basis
{z1,…,zr,zr+1,…,rr+s}
for Rr,s, where the basis vectors z1,…,zr are
positive and zr+1,…,zr+s are negative.
Let Clr,s be the real Clifford algebra
generated by Rr,s, that is the quotient of the tensor algebra
[TABLE]
divided by the two-sided
ideal Ir,s which is generated by the elements of the form
z⊗z+⟨z,z⟩r,s, z∈Rr+s.
The explicit determination of the Clifford algebras
is given in [3] and they are isomorphic
to matrix algebras
R(n),
R(n)⊕R(n),
C(n),
H(n) or
H(n)⊕H(n) where
the size n is determined by r and s, see [27].
Given an algebra homomorphism J:Clr,s→End(U), we call the space U a Clifford module and
the operator Jϕ a Clifford action or a representaion map
of an element ϕ∈Clr,s.
If there is a map
[TABLE]
satisfying Jz2=−⟨z,z⟩r,sIdU for an
arbitrary z∈Rr,s, then J can be uniquely extended to an algebra homomorphism
J by
the universal property, see, for instance [21, 26, 27].
Even though the representation matrices of the Clifford algebras
Clr,s, and the Clifford modules U are given over the fields R, C or
H, we refer to Clr,s as a real algebra and U as a real vector space.
The dimension of U is a multiple of n over the
corresponding fields R, C or
H.
If r−s≡3(mod4), then Clr,s is a simple algebra.
In this case there is only one
irreducible module U=Virrr,s of dimension n.
If r−s≡3(mod4), then the algebra Clr,s is not simple,
and there are two non-equivalent irreducible modules.
They can be distinguished by the action of the ordered volume form
Ωr,s=∏k=1r+szk. In fact, the elements \frac{1}{2}\big{(}\operatorname{Id}\mp\Omega^{r,s}\big{)}
act as an identity operator on the Clifford module, so
JΩr,s≡±Id.
Thus we denote by Virr;±r,s two non-equivalent irreducible Clifford modules
on which the action of the volume form is given by JΩr,s=∏k=1r+sJzk≡±Id.
Proposition 2.1.1**.**
[27]*
Clifford modules are completely reducible. Namely, let U
be a Clifford module, then it can be
decomposed into irreducible modules:*
[TABLE]
The numbers p, p+,p− are uniquely determined by the dimension of U.
The module U=⊕pVirrr,s is called isotypic and the second one in (2.1) is non-isotypic.
2.2. Admissible modules
Definition 2.2.1**.**
[11]** A module U of the Clifford algebra Clr,s
is called admissible if there is a scalar product ⟨⋅,⋅⟩U on
U such that
[TABLE]
We write (U,⟨⋅,⋅⟩U) for
an admissible module to emphasise the scalar product ⟨⋅,⋅⟩U
and call it an admissible scalar product.
We collect properties of admissible modules in several propositions.
Proposition 2.2.2**.**
Let Clr,s be the Clifford algebra generated by the space Rr,s.
(1)
If ⟨⋅,⋅⟩U
is an admissible scalar product for Clr,s, then K⟨⋅,⋅⟩U
is also admissible for any constant K=0. We can assume that K=±1 by normalisation of the scalar products.
(2)
Let (U,⟨⋅,⋅⟩U) be an admissible module for Clr,s and let (U1,⟨⋅,⋅⟩U1) be such that U1 is a submodule of U and ⟨⋅,⋅⟩U1 is the
restriction of ⟨⋅,⋅⟩U to U1. Then the orthogonal complement
U1⊥={x∈U∣⟨x,y⟩U=0,for ally∈U1}
with the scalar product obtained by the restriction of ⟨⋅,⋅⟩U to U1⊥
is also an admissible module.
(3)
Condition (2.2) and the
property Jz2=−⟨z,z⟩r,sIdU imply
[TABLE]
(4)
Relation (2.3) leads to the following:
if z∈Rr,s is positive, then
[TABLE]
In other words the map Jz:U→U is an isometry for ⟨z,z⟩r,s=1. If z∈Rr,s is negative, then
[TABLE]
and the map Jz:U→U is an anti-isometry for ⟨z,z⟩r,s=−1.
(5)
If s>0, then any admissible module (U,⟨⋅,⋅⟩U) of Clr,s
is neutral, i.e., dimU=2l, l∈N, and U it is isometric to
Rl,l, see [11, Proposition 2.2].
(6)
If s=0, then any Clifford module of Clr,0
can be made into admissible with positive definite or negative
definite scalar product, see [21].
Proposition 2.2.3 describes the relation between irreducible and admissible modules. This relation depends on the signature (r,s) of the generating space Rr,s for the Clifford algebra Clr,s.
An admissible module of the minimal possible dimension is called a minimal admissible module.
Proposition 2.2.3**.**
[11, Theorem 3.1]**[18, Proposition 1]**
Let Clr,s be the Clifford algebra generated by the space Rr,s.
(1)
If s=0, then any irreducible Clifford module is minimal
admissible with respect to a positive definite or a
negative definite scalar product.
(2)
If r−s≡0,1,2(\emmod4),
then a unique irreducible module Virrr,s is not necessary admissible. The following situations are possible:
(2-1)
The irreducible module Virrr,s is minimal admissible or,
(2-2)
The irreducible module Virrr,s is not admissible, but the
direct sum Virrr,s⊕Virrr,s is minimal admissible.
(3)
If r−s≡3(\emmod)4, then for two non-equivalent irreducible modules Virr;±r,s the following can occur:
(3-1)
Each irreducible module Virr;±r,s is minimal admissible. The index s must be even in this case.
(3-2)
None of the irreducible modules Virr;±r,s is
admissible. It can happened for even and odd values of s.
(3-2-1)
If s is even, then Virr;+r,s⊕Virr;+r,s, Virr;−r,s⊕Virr;−r,s are minimal admissible modules, and the module Virr;+r,s⊕Virr;−r,s is not admissible.
(3-2-2)
If s is odd, then the module Virr;+r,s⊕Virr;−r,s is minimal admissible and neither Virr;+r,s⊕Virr;+r,s nor Virr;−r,s⊕Virr;−r,s is admissible.
We emphasize the following corollary, see also Table 1 and the remark after it.
Corollary 2.2.4**.**
There are two minimal admissible modules only if r−s≡3(\emmod4)
and s is even.
We distinguish two cases:
(3-1)
Each irreducible module is minimal admissible:
Vmin;+r,s=Virr;+r,s and
Vmin;−r,s=Virr;−r,s. In this case
r≡3(\emmod4), s≡0(\emmod4), or r≡1(\emmod8), s≡2(\emmod8),
or r≡5(\emmod8), s≡6(\emmod8).
(3-2-1)
Direct sums of irreducible modules are minimal
admissible: Vmin;+r,s=Virr;+r,s⊕Virr;+r,s
and Vmin;−r,s=Virr;−r,s⊕Virr;−r,s. It happens if
r≡1(\emmod8), s≡6(\emmod8),
or r≡5(\emmod8), s≡2(\emmod8).
2.3. Mutually commuting isometric involutions
Recall that a linear map Λ defined on a vector space U with
a scalar
product ⟨⋅,⋅⟩U is called
symmetric with respect to the scalar product ⟨⋅,⋅⟩U,
if ⟨Λx,y⟩U=⟨x,Λy⟩U,
isometric (or positive) if it maps positive vectors to
positive vectors and negative vectors to negative vectors
and anti-isometric (or negative) if it reverses the positivity and negativity of the vectors.
Let Jzi be representation maps for an orthonormal basis
{z1,…,zr+s} of Rr,s.
The simplest isometric involutions, written as a product of the maps
Jzi, are one of the following forms:
[TABLE]
The product involutions of types P3 and P4 is not an involution, meanwhile the product of involutions of other types is again an involution.
For a given minimal admissible module Vminr,s,
we denote by PIr,s a set of the maximal
number of mutually commuting symmetric isometric
involutions of the forms (2.4)
and such that none of them is a product
of other involutions in PIr,s.
The set PIr,s is not unique, while
the number of involutions pr,s=#{PIr,s} in PIr,s is unique for the given signature (r,s).
The set PIr,s can be chosen equal for the modules with the
admissible scalar product of opposite signs, or for the minimal admissible modules, based on the two non-equivalent Clifford modules.
The ordering on the set PIr,s can be made, if necessary, in such a way
that at most one involution of the type P3 or P4 is included in PIr,s and it is the last one, see Sections 2.4 and 5.
We also need
a set COr,s of complementary operators to
PIr,s, which are
products of the maps Jzi and they are
ordered according to the
ordering in PIr,s such that
[TABLE]
The complementary operators can be isometric or anti-isometric. They guarantee that all the involutions from PIr,s have their both eigenspaces as subspaces of Vminr,s. Note that if r−s≡3(mod4) and s is even, then the last involution of type P3 or P4 allows to distinguish the modules Vmin;+r,s and Vmin;−r,s, see for instance the proof of Theorem 2.6.2. Therefore, the number of operators in COr,s is different and is equal to
[TABLE]
We define the subspace Er,s of a minimal admissible module Vminr,s
[TABLE]
and call it the “common 1-eigenspace” for the system of involutionsPIr,s. The complementary opertors show whether the common 1-eigenspace
Er,s is a neutral or sign definite vector space with respect to
the restriction of the admissible scalar product.
The sets PIr,s and COr,s are collected in Section 5 and they will be mentioned precisely when it needs to be done. In the following proposition we explain the possible interaction of involutions with the complementary operators.
Proposition 2.3.1**.**
[17]*
If P is an isometric symmetric involution acting on the
space (U,⟨⋅,⋅⟩U) with a neutral scalar product and E±1 are eigenspaces of P, then*
If I:U→U is an isometry such that PI=−IP, then E±1 are neutral,
2.
If A:U→U is an anti-isometry such that PA=AP, then E±1 are neutral,
3.
If A:U→U is an anti-isometry such that PA=−AP, then E±1 are either neutral or sign definite,
with respect to the restriction of the scalar product ⟨⋅,⋅⟩U to E±1.
Since the involutions in PIr,s are symmetric,
the eigenspaces are orthogonal subspaces.
The involutions commute, therefore, they decompose the eigenspaces of
other involutions into smaller (eigen)-subspaces.
We give an example, that is crucial for the paper.
Example 1**.**
The set PIμ,ν for (μ,ν)=(8,0),(0,8) or (4,4) is given by
[TABLE]
The set of complementary operators
are
[TABLE]
The module Vminμ,ν
is decomposed into 16 one dimensional common eigenspaces of
four involutions Ti. Let
v∈Eμ,ν
and ∣⟨v,v⟩Vminμ,ν∣=1. Then other common
eigenspaces
are spanned by Jziv, i=1,…,8,
and Jz1Jzjv, j=2,…,8. Hence we have
[TABLE]
The value ⟨v,v⟩Vminμ,ν can be ±1
according to the admissible scalar product, however
we may assume ⟨v,v⟩Vminμ,ν=1 by Lemma 3.2.5.
2.4. Periodicity of Clifford algebras and admissible modules
We call the following lower values of signature (r,s):
[TABLE]
the basic cases.
Recall the periodicity of Clifford algebras:
[TABLE]
where the last one follows from Clr,s⊗Cl1,1≅Clr+1,s+1, see [3].
The Clifford algebras
Clμ,ν, (μ,ν)∈{(8,0),(0,8),(4,4)},
are isomorphic to R(16).
The unique irreducible module is minimal admissible Virrμ,ν=Vminμ,ν in all the cases. They are isometric to the following spaces:
Vmin8,0≅R16,0≅R0,16, where we fix the first isomorphism for the constructions of Lie algebras due to Lemma 3.2.5, and we also have
Vmin0,8≅R8,8
and Vmin4,4≅R8,8.
Proposition 2.4.1**.**
[17]*
If Vminr,s=(Vminr,s,⟨⋅,⋅⟩Vminr,s)
is a minimal admissible module, then Vr+μ,s+ν=Vminr,s⊗Vminμ,ν
is the minimal admissible module of Clr+μ,s+ν, where the scalar product on Vr+μ,s+ν is given by
⟨⋅,⋅⟩Vminr,s⟨⋅,⋅⟩Vminμ,ν for (μ,ν)∈{(8,0),(0,8),(4,4)}.*
Let {ζ1,…,ζ8} be an orthonormal basis for
Rμ,ν and {z1,…,zr+s} be an orthonormal basis for Rr,s. Let Jζi, i=1,…,8 and Jzj, j=1,…,r+s be the respective representation maps.
We denote by Ωμ,ν=∏i=18ζi
the ordered volume form for Clμ,ν for (μ,ν)∈{(8,0),(0,8),(4,4)}. Set
[TABLE]
Then the maps J^zj and J^ζα
are representations of an orthonormal basis {zj,ζi} of Rr+μ,s+ν
on the space Vminr,s⊗Vminμ,ν,
as it was shown in [17].
2.5. Dimensions of minimal admissible modules
The dimensions of minimal admissible modules are determined for
basic cases (2.8). Then dim(Vminr+μ,s+ν)=dim(Vminr,s)⋅dim(Vminμ,ν)=16dim(Vminr,s)
for any minimal admissible module Vminr,s,
(μ,ν)∈{(8,0),(0,8),(4,4)}. Moreover dim(Vminr,s)=2r+s−pr,s where pr,s=#{PIr,s}. This follows from the fact that
minimal admissible modules are cyclic modules
and pr,s relations among the 2r+s vectors
{Jzi1Jzi2⋯Jzikv}, v∈Er,s,
allow us to span the space Vminr,s by 2r+s−pr,s number
of linearly independent vectors.
We describe the number and the dimension
of minimal admissible modules Vminr,s in
Table 1. We indicate whether
the scalar product restricted to the common 1-eigenspaces Er,s of the
involutions from PIr,s is neutral or sign definite,
see Section 2.6 for the proof.
We use the black colour when dim(Vminr,s)=2dim(Virrr,s), see Proposition 2.2.3,
items (2-2), (3-2-1), and (3-2-2).
(2)
Writing the subscript ”∗×2” we show that the Clifford algebra has two minimal admissible modules corresponding to the non-equivalent irreducible modules, see Corollary 2.2.4.
(3)
The upper index ”∗N” means that the scalar product
restricted to Er,s is neutral.
The fact that Er,s is a neutral space does not depend on the
choice of the scalar product on Vminr,s, see Section 2.6.
(4)
The upper index ”∗±” shows that the scalar product
restricted to the common 1-eigenspace Er,s of the system
PIr,s is sign definite, see Section 2.6.
The sign of the scalar product on Er,s depends on the choice of the admissible scalar product on the module Vminr,s.
For example, the Clifford algebra Cl3,0 has 4 minimal
admissible modules Vmin;±3,0;±,
that is, each non-equivalent irreducible module Virr;±3,0 can be endowed
with two scalar products: positive definite, giving the minimal admissible modules Vmin;±3,0;+=Virr;±3,0;+ and
negative definite: Vmin;±3,0;−=Virr;±3,0;−.
However, it does not mean that pseudo H-type Lie algebras
corresponding to these four choices are different. We explain details in Theorems 4.1.2 and 4.1.3.
To obtain the Table 1
we determine the sets PIr,s and COr,s, given in Section 5.
2.6. Scalar product on the common 1-eigenspace Er,s
Lemma 2.6.1**.**
Let (Vminr,s,⟨⋅,⋅⟩Vminr,s)
be a minimal admissible module of Clr,s and
(Vminr+μ,s+ν,⟨⋅,⋅⟩Vminr+μ,s+ν)
a minimal admissible module of
Clr+μ,s+ν, where (μ,ν)∈{(8,0), (0,8),(4,4)}.
Let Er,s
and Er+μ,s+ν be the common 1-eigenspaces of the involutions PIr,s and PIr+μ,s+ν, respectively.
Then dimEr,s=dimEr+μ,s+ν. Moreover,
if Er,s is a neutral vector space, then Er+μ,s+ν is also neutral, and
if Er,s is a sign definite, then Er+μ,s+ν is also sign definite.
Proof.
If Vminr+μ,s+ν=Vminr,s⊗Vminμ,ν, then
the assertions follow from Proposition 2.4.1.
Let Vminr+μ,s+ν be an arbitrary minimal admissible module
of Clr+μ,s+ν, where (μ,ν)∈{(8,0),(0,8),(4,4)}. Let {zj,ζα;j=1,…,r+s,α=1,…,8} be
an orthonormal basis of
Rr+μ,s+ν. We assume
{zi,ζα;i=1,…,r,α=1,…,μ}
are positive and
{zr+j,ζμ+β;j=1,…,s,β=1,…,ν}
are negative. We identify Rr,s⊕Rμ,ν=Rr+μ,s+ν by using the above bases. We choose PIr+μ,s+ν=PIr,s⋃{Ti}i=14, where Ti are involutions from Example 1.
The system of complementary operators COμ,ν
shows that the involutions Tj∈PIr+μ,s+ν, j=1,2,3,4 decompose the space
Vminr+μ,s+ν into 16 common eigenspaces
{Vi}i=015 of Tj and
[TABLE]
where V0 is the common 1-eigenspace of Tj, j=1,2,3,4. Since the generators Jzj, j=1,…,r+s, commute with
involutions Ti, i=1,2,3,4, we can regard V0 as a minimal
admissible module Vminr,s of Clr,s. The involutions from PIr,s act on
V0=Vminr,s and decompose it into their common
eigenspaces. Then
by definition Er+μ,s+ν=Er,s. This finishes the proof of the theorem.
∎
Theorem 2.6.2**.**
Let Er,s⊂Vminr,s be a common 1-eigenspace of the system PIr,s. Then
the restriction of the admissible scalar product on Er,s is sign definite
for r≡0,1,2(\emmod4) and
s≡0(\emmod4) or for r≡3(\emmod4) and
arbitrary s.
Otherwise the restriction of the admissible scalar product on the common 1-eigenspaces Er,s is neutral.
Proof.
We find the sign of the scalar products on Er,s for basic cases (2.8) and then apply Lemma 2.6.1.
Case (r,0). The scalar products on the common 1-eigenspaces Er,0 are sign definite because the admissible scalar products on Vminr,0 are sign definite.
Case (r,4), r=0,1,2,4. The system of involutions PIr,4,
r=0,1,2 and their complementary
operators are given in Table 5. The complementary operators gives the dimension of Er,s and the basis shows that the space is sign definite.
The case (4,4) was considered in Example 1.
Cases (3,s), s=0,…,7 and (7,s), s=1,2,3.
The system of involutions and their complementary operators are given in Table 6.
Notice that the involution Jz1Jz2Jz3 belongs to all the
systems. The isometric complementary operators ensures that the common
1-eigenspace E1 for involution from
PIr,s∖{Jz1Jz2Jz3}
is neutral. Let E1,1 and E1,−1 be the eigenspaces of Jz1Jz2Jz3 corresponding to the eigenvalues 1 and -1, respectively. The last complementary operator from COr,s, that is anti-isometry, shows that the spaces E1,1∩E1 and E1,−1∩E1 are sign definite with opposite signs of scalar products on E1,1∩E1 and E1,−1∩E1.
The case E3,4 is special since there are two non-equivalent irreducible modules Virr;+3,4 and Virr;−3,4, where the volume form Ω3,4=P1P3P4 acts as Id and −Id respectively. It shows that P4=Jz1Jz2Jz3=−Id on Virr;+3,4 and P4=Jz1Jz2Jz3=Id on Virr;−3,4.
The proof of the statement concerning the neutral common 1-eigenspace
follows from the systems PIr,s and COr,s for mentioned
values of r and s,
see Table 7 in Section 5.
∎
3. Pseudo H-type Lie algebras and Lie groups
In this section we recall basic facts on isomorphisms between pseudo
H-type algebras and discuss some properties of the automorphism
groups Aut(Nr,s(U)) of pseudo H-type algebras.
The Table 2 contains the classification result for the pseudo H-type Nr,s(Vminr,s) obtained in [18].
3.1. Definitions of
the pseudo H-type Lie algebras and their groups
Let (U,⟨⋅,⋅⟩U) be an admissible module of a Clifford
algebra Clr,s.
We define a vector valued skew-symmetric bi-linear form
[TABLE]
by the relation
[TABLE]
Definition 3.1.1**.**
[11]*
The space U⊕Rr,s endowed with the Lie bracket*
[TABLE]
is called a pseudo H-type Lie algebra and it is denoted by Nr,s(U).
A pseudo H-type Lie algebra Nr,s(U) is 2-step nilpotent, the space Rr,s is the centre, and the direct sum U⊕Rr,s is orthogonal with respect to ⟨⋅,⋅⟩U+⟨⋅,⋅⟩r,s.
The Baker-Campbell-Hausdorff formula allows us to define the Lie group structure on the space
U⊕Rr,s by
[TABLE]
The Lie group is denoted by Gr,s(U) and is called the pseudo H-type Lie group.
Note that the scalar product ⟨⋅,⋅⟩U is implicitly
included in the definitions of the H-type Lie algebra and the
corresponding Lie group.
In general, the Lie algebra structure might change
if we replace the admissible scalar product on U, see [4, 15, 16].
The main purpose of the present paper is to classify the Lie algebras Nr,s(U), whose constructions involve the non-minimal admissible modules U of Clifford algebras Clr,s.
3.2. Isomorphisms of pseudo H-type Lie algebras
Let U and U~
be two vector spaces with scalar products ⟨⋅,⋅⟩U and ⟨⋅,⋅⟩U~ respectively. Let Λ:U→U~ be a linear map.
The operator Λτ:U~→U defined by the relation
[TABLE]
is called the adjoint operator with respect to the scalar products
⟨⋅,⋅⟩U and ⟨⋅,⋅⟩U~. If both scalar products are positive definite, we use the notation tΛ.
Let Nr,s(U) and Nr~,s~(U~)
be two pseudo H-type Lie algebras with r+s=r~+s~=k and dim(U)=dim(U~)=n. A Lie algebra isomorphism Φ:Nr,s(U)→Nr~,s~(U~)
has the form
[TABLE]
B:U→Rk is a linear map, see [24]. The action is defined by
Φ(x,z)=(Ax,Bx+Cz), x∈U, z∈Rr,s.
If we write Jz:U→U and J~w:U~→U~ for the corresponding actions on the Clifford modules, then the matrices A and C satisfy the relation
[TABLE]
by (3.1). The matrices Aτ and Cτ are defined as in (3.2).
The matrix B is arbitrary and we can choose B=0 for
simplicity.
To short the notation we write Φ=A⊕C for the isomorphism Φ=(A00C).
In following propositions we collect the properties of
isomorphisms of H-type Lie algebras Nr,s(U) and Nr~,s~(U~) for different values of signatures (r,s) and (r~,s~) studied in [18], see also [5, 8, 30, 31, 32].
Proposition 3.2.1**.**
If Φ=A⊕C:Nr,s(U)→Nr~,s~(U~)
is a Lie algebra isomorphism, then
(1)
the map Φτ=Aτ⊕Cτ:Nr~,s~(U~)→Nr,s(U) is also a Lie algebra isomorphism and moreover
(2)
r=r~, s=s~, or r=s~, s=r~.
Proposition 3.2.2**.**
If Φ=A⊕C:Nr,s(U)→Ns,r(U~) is a Lie algebra isomorphism and r=s, then
(1)
AτAJzAτA=−Jz, z∈Rr,s, AAτJ~wAAτ=−J~w, w∈Rs,r;
(2)
AJz1Jz2=−J~C(z1)J~C(z2)A, AJ~C(z1)J~C(z2)=−Jz1Jz2Aτ
for z1,z2∈Rr,s with ⟨z1,z2⟩r,s=0;
(3)
the linear transformation C:Rr,s→Rs,r
maps positive vectors to negative vectors and vice versa. We can assume
that ∣detAτA∣=1 and CτC=−Id by multiplying the matrix A by a
constant.
Proposition 3.2.3**.**
If Φ=A⊕C:Nr,s(U)→Nr,s(U~) is a Lie algebra isomorphism and r=s, then
(1)
AτAJzAτA=Jz, AAτJ~wAAτ=J~w for z,w∈Rr,s;
(2)
AJz1Jz2=J~C(z1)J~C(z2)A, AJ~C(z1)J~C(z2)=Jz1Jz2Aτ
for z1,z2∈Rr,s with ⟨z1,z2⟩r,s=0;
(3)
the linear transformation C:Rr,s→Rs,r
maps positive vectors to positive vectors, negative to negative ones, and as in Proposition 3.2.2
we may assume that ∣detAτA∣=1 and
CτC=Id by multiplying the matrix A by a
constant.
Proposition 3.2.4**.**
[18]*
If Φ=A⊕C:Nr,r(U)→Nr,r(U~)
is a Lie algebra isomorphism, then
CτC=Id for r≡0,1,2(\emmod4)
and CτC=−Id for r≡3(\emmod4).
The map A can be normalised such that ∣detAτA∣=1 and it satisfies the conditions of items (1)−(2) of Proposition 3.2.3.*
We explain a possible construction of the map A:Vminr,s→V~minr,s. It was shown in [18, Corollary 5, Theorem 3] that the map A can be obtained by the following procedure. The system of involutions PIr,s acting on the minimal admissible modules decomposes them into the direct sums Vminr,s=⊕iEi, V~minr,s=⊕iE~i of common eigenspaces. We start by constructing a map A1:Er,s→E~r~,s~, where Er,s⊂Vminr,s, E~r~,s~⊂V~minr,s are common 1-eigenspaces of the system of involutions PIr,s and PIr~,s~. Then the map A1 produces the rest of the maps Ai:Ei→E~i between the eigenspaces. Thus the map A has block diagonal form A=⊕iAi written in the basis described in Section 3.4 and satisfying the relations of Propositions 3.2.2 and 3.2.3.
Example 2**.**
Recall decomposition (2.7) of Vminμ,ν
for (μ,ν)∈{(8,0),(0,8),(4,4)}. Let A⊕Id:nr,s(Vminμ,ν)→nr,s(V~minμ,ν) and
[TABLE]
as in (2.7). The condition (3.3) applied to a vector u∈Eμ,ν is equivalent to the statement that Aj+1τJ~ζjA1=Jζj, where A1=A∣Eμ,ν and Aj+1τ=Aτ∣J~ζj(E~μ,ν), j=1,…,8, are the restrictions of the maps on the indicated spaces and the diagram
[TABLE]
commutes. This shows that the maps
[TABLE]
are completely determined by the map A1. The conditions AJζ1Jζj=J~ζ1J~ζjA determine the maps
[TABLE]
Thus the map A:Vminμ,ν→V~minμ,ν is defined by A1:Eμ,ν→E~μ,ν and has the form A=⊕j=1j=16Aj in a suitable basis.
Lemma 3.2.5**.**
Let U+=(U,⟨⋅,⋅⟩U) be an admissible module of Clr,s and Jz:U→U be the action map, then the module U−=(U,−⟨⋅,⋅⟩U)
is an admissible with the same action map
Jz:U−=U→U−=U. Moreover,
the Lie algebras Nr,s(U) and Nr,s(U~) are
isomorphic under the isomorphism Id⊕−Id.
Proof.
By the definition, we can see easily that U−=(U,−⟨⋅,⋅⟩U) is
an admissible module and
the map
[TABLE]
is a Lie algebra isomorphism because of
[TABLE]
∎
3.3. Automorphisms of the pseudo H-type Lie algebras
We discuss here the group Aut(Nr,s(U))
of automorphisms of a Lie algebra Nr,s(U), see also [15, 24, 30]. The group Aut(Nr,s(U)) is
a subgroup of GL(r+s+dim(U),R) and consists of the linear maps
[TABLE]
satisfying the condition (3.3), see also Propositions 3.2.3 and 3.2.4.
The group Aut(Nr,s(U))
is isomorphic to the following product
[TABLE]
Here R+ is the group of non-homogeneous dilations δt:U⊕Rr,s→U⊕Rr,s acting as δt(x,z)=(tx,t2z) for
t∈R+.
The group
Br,s={(IdB0Id)}
is isomorphic to R(r+s)⋅dim(U). The subgroup Aut0(Nr,s(U)),
consisting of the automorphisms of the form
Ψ=(A00C) with
CτC=±Id, is called the group of restricted automorphisms.
The semi-direct product in (3.4)
comes from the action of the subgroup Aut0(Nr,s(U))
on
Br,s
by
[TABLE]
The group of automorphisms of the Lie algebras
Nr,0(U) was studied
in [5, 23, 30, 31].
Now we present an example of elements of
Aut0(Nr,s(U)),
that will be important for the classification of the Lie algebras Nr,s(U).
The map
[TABLE]
is extended to the Clifford algebra automorphism α:Clr,s→Clr,s
by the universal property of the Clifford algebras.
We denote by Clr,s× the group of invertible elements in
Clr,s and in particular Rr,s×={v∈Rr,s∣⟨v,v⟩r,s=0}=Rr,s∩Clr,s×. The representation
Ad:Rr,s×→End(Rr,s),
is defined as
[TABLE]
Then it extends to the, so called, twisted adjoint representation
Ad:Clr,s×→GL(Clr,s) by setting
[TABLE]
The map Adv for v∈Rr,s×, leaving the space
Rr,s⊂Clr,s invariant, is also an isometry:
⟨Adv(z),Adv(z)⟩r,s=⟨z,z⟩r,s.
Note that (Adφ−1)τ=Adφ.
Subgroups of Clr,s× defined by
[TABLE]
are called pin and spin groups, respectively. The reader can find more information about the twisted adjoint representation and the groups Pin and Spin, see [27].
We make the identification Spin(r)×Pin(s)≅Spin(r,0)×Pin(0,s)⊂Pin(r,s) and present a special map
from Aut0(Nr,s(U)).
Proposition 3.3.2**.**
Let J:Clr,s→End(U) be a Clifford algebra representation and φ∈Spin(r)×Pin(s). Then
Jφ−1⊕(Adφ)τ∈Aut0(Nr,s(U)).
The group homomorphism
[TABLE]
is injective and the diagram
[TABLE]
is commutative. The kernel Kr,s(U) consists of automorphisms of the
form
A⊕Id.
Proof.
By the definition of the twisted adjoint representation,
α(φ)zφ−1=Adφ(z)
we have
[TABLE]
If we show that Jα(φ)=Jφ−1τ, or equivalently Jα(φ−1)=Jφτ
for φ∈Pin(r,s), then it will imply that Jφ−1⊕(Adφ)τ∈Aut0(Nr,s) due to the relation AτJzA=JCτ(z).
If v∈Rr,s× is such that ⟨v,v⟩r,s=−1, then
Jv−1τ=Jvτ=−Jv=Jα(v), and hence
Jv−1⊕(Adv)τ∈Aut0(Nr,s(U)). If
v is such that ⟨v,v⟩r,s=1, then
Jv−1τ=J−vτ=Jv=Jα(v), and therefore
the map Jv−1⊕(Adv)τ does not belong to Aut0(Nr,s(U)).
If φ=v1v2 with ⟨vi,vi⟩r,s=±1, i=1,2, then
J(v1v2)−1=Jv2v1=Jα(v1v2)τ. It implies
J(v1v2)−1⊕(Adv1v2)τ∈Aut0(Nr,s(U)).
In general, if φ=x1⋯x2p⋅y1⋯yq∈Pin(r,s)
with ⟨xi,xi⟩r,s=1, i=1,…,2p, and ⟨yj,yj⟩r,s=−1, j=1,…,q, then we obtain
[TABLE]
∎
Corollary 3.3.3**.**
There is an automorphism
A⊕−Id∈Aut0(N2r,s(U)) for any r,s.
Proof.
Observe that the image of the map Aut0(N2r,s)→O(2r,s)
includes the group SO(2r)×O(s). This follows from diagram (3.6) and Proposition 3.3.1. Since −Id∈SO(2r)×O(s) belongs to the image of Aut0(N2r,s(U))→O(2r,s) we conclude that there is A:U→U such that A⊕−Id∈Aut0(N2r,s(U)).
∎
Consider the diagram (3.6) in special cases of N0,8=N0,8(Vmin0,8) and N8,0=N8,0(Vmin8,0). The following two diagrams are exact.
[TABLE]
[TABLE]
Since the Lie algebras N8,0 and N0,8 are isomorphic, see [18],
there exists an automorphism A⊕I1∈Aut0(N8,0) which is not in the image
A(Spin(8)),
where I1 is defined as
I1(ζ1)=−ζ1, I1(ζj)=ζjj=2,…,8.
At the end of the section we formulate the relation between the existence of an automorphism and an isomorphism of a special type.
Lemma 3.3.4**.**
A Lie algebra isomorphism A⊕Id:Nr,s(Vminr,s;+)→Nr,s(Vminr,s;−) exists if and only if there is a Lie algebra automorphism
A⊕−Id:Nr,s(Vminr,s;+)→Nr,s(Vminr,s;+).
Proof.
Let us assume that A⊕IdRr,s:Nr,s(Vminr,s;+)→Nr,s(Vminr,s;−), where
A:Vminr,s;+→Vminr,s;− is a Lie algebra isomorphism. We assume that the module actions on Vminr,s;± coincide, but the admissible scalar
products differ by the sign, that is ⟨⋅,⋅⟩Vminr,s;+=−⟨⋅,⋅⟩Vminr,s;−. We denote the Lie brackets on the corresponding pseudo H-type Lie algebras
by [x,y]± for x,y∈Vminr,s;±. Then
[TABLE]
It shows that A⊕−Id is an automorphism of Nr,s(Vminr,s;+). Now assuming that A\oplus-\operatorname{Id}\in\operatorname{Aut}^{0}\big{(}\mathscr{N}_{r,s}(V_{min}^{r,s;+})\big{)}, we obtain
[TABLE]
Thus A⊕Id:Nr,s(Vminr,s;+)→Nr,s(Vminr,s;−) is an isomorphism.
∎
3.4. Existence of lattices on pseudo H-type Lie groups
To achieve the full description of isomorphic Lie algebras Nr,s(U), where
the admissible module U is not necessarily minimal,
we need a special type of bases for the Clifford modules. These type of bases
also show the existence of lattices on the corresponding Lie groups, see [17].
It is enough to construct the bases only for minimal admissible modules, and then apply Proposition 2.4.1.
Let Vminr,s be a minimal admissible module of
the Clifford algebra Clr,s and Er,s be the common
1-eigenspace for the system PIr,s of involutions.
We fix a vector v∈Er,s such that
∣⟨v,v⟩Vminr,s∣=1. Then a basis {xi}i=1N
of the module Vminr,s can be chosen by setting
[TABLE]
that is a subset of all the 2r+s vectors obtained from v by action of Jzi1…Jzik,
1≤ii<i2<⋯<ik≤r+s.
The vector v∈Er,s can be picked up in such a way that
the basis in Er,s will be orthonormal due to the following proposition.
Proposition 3.4.1**.**
[17, Lemma 2.9, Corollary 2.10]**
Let (V,⟨⋅,⋅⟩V) be an admissible module, Λ1,…,Λl symmetric linear operators on V such that
(1)
Λk2=−IdV, k=1,…,l;
(2)
ΛkΛj=−ΛjΛk* for all k,j=1,…,l.*
Then for any w∈V with ⟨w,w⟩V=1 there is a vector w~ satisfying
⟨w~,Λkw~⟩V=0 and ⟨w~,w~⟩V=1, for k=1,…,l.
Since the involutions are symmetric all the eigenspaces are mutually orthogonal, that implies the orthonormality of the constructed basis. The construction of the basis also shows that
⟨Jzixj,xk⟩Vminr,s=⟨zi,[xj,xk]⟩r,s=±1 or [math].
It follows that the structure constants of the Lie algebra Nr,s(U) are ±1 or 0. The concrete construction of bases for Nr,s(U) can be found in [17], see also [12]. Applying the Malćev criterion [28], we obtain the proposition.
Proposition 3.4.2**.**
[28]**
Let U be an admissible module of a Clifford algebra
Clr,s. Then
there exists a lattice on the pseudo H-type Lie group Gr,s(U).
3.5. Classification of pseudo H-type Lie algebras Nr,s(Vminr,s)
The classification of the pseudo H-type algebras Nr,s(Vminr,s), constructed from the minimal admissible modules was done in [18]. We summarise the results of the classification in Table 2.
Here “d” stands for “double”, meaning that dimVminr,s=2dimVmins,r and
“h” (half) means that dimVminr,s=21dimVmins,r.
The corresponding pairs are trivially non-isomorphic due to the
different dimension of minimal admissible modules.
The symbol ≅ denotes the Lie algebra Nr,s(Vminr,s) having
isomorphic counterpart Ns,r(Vmins,r), the symbol ≅ shows that the Lie algebra nr,s(Vminr,s) is not
isomorphic to Ns,r(Vmins,r). The notation ↻ indicates that
the Lie algebra Nr,r(Vminr,r) admits automorphisms A⊕Id,
and ↻ denotes the Lie algebra Nr,r(Vminr,r) that does not have this type of automorphism.
4. Classification of pseudo H-type algebras
In this section we state and prove the
classification of the pseudo H-type algebras Nr,s(U)
with an arbitrary admissible modules U,
and fixed signature (r,s). Eventually, the classification depends on the decomposition of U on the minimal admissible modules. It is enough to classify basic cases (2.8) due to Theorem 4.2.5.
4.1. Statements of main results on isomorphisms of Lie algebras Nr,s(U)
In the rest of the paper we use the upper index ± to indicate
the scalar products that differ by sign: Vminr,s;+=(Vminr,s,⟨⋅,⋅⟩Vminr,s)
and Vminr,s;−=(Vminr,s,−⟨⋅,⋅⟩Vminr,s).
We also use the lower index ±
to distinguish the minimal admissible modules, corresponding to non equivalent irreducible modules,
Vmin;±r,s;+=(Vmin;±r,s,⟨⋅,⋅⟩Vmin;±r,s)
and Vmin;±r,s;−=(Vmin;±r,s,−⟨⋅,⋅⟩Vmin;±r,s) that were mentioned in Corollary 2.2.4.
Recall that Clifford modules are completely reducible, see
Proposition 2.1.1
and any admissible module can be
decomposed into the orthogonal sum of minimal admissible modules, see
Proposition 2.2.2, item (2). To make the complete classification we decompose an admissible module U of the Clifford algebra Clr,s into the direct sum of, possibly different, minimal admissible modules. We distinguish the following possibilities.
If r−s≡3(mod4) and s is arbitrary or r−s≡3(mod4) and s is odd then
[TABLE]
If r−s≡3(mod4) and s is even, then
[TABLE]
The system of involutions PIr,s does not depend on the scalar product on the admissible modules Vminr,s;± and therefore the common 1-eigenspaces Er,s coincide on Vminr,s;+ and Vminr,s;−. Nevertheless, the restrictions of the admissible scalar products from Vminr,s;+ and Vminr,s;− on the respective Er,s will have opposite signs. The result of the classification essentially depends on the signature of the restriction of the admissible scalar product on Er,s and the parity of the index s. We formulate the main results of the classification.
Theorem 4.1.1**.**
Let U=(U,⟨⋅,⋅⟩U) and
U~=(U~,⟨⋅,⋅⟩U~)
be admissible modules of a Clifford algebra Clr,s.
If r≡0,1,2mod4,
then the pseudo H-type Lie algebra Nr,s(U)
is determined by the dimension
of the admissible module U and does not depend on the choice of an admissible scalar product. Thus Nr,s(U)≅Nr,s(U~), if and only if dim(U)=dim(U~).
If r≡3(mod4), then the pseudo H-type Lie algebra Nr,s(U)
is determined by the dimension of U and by the value of the index s.
Theorem 4.1.2**.**
Let U=(U,⟨⋅,⋅⟩U) and
U~=(U~,⟨⋅,⋅⟩U~)
be admissible modules of a Clifford algebra Clr,s.
Let r≡3(\emmod4) and s≡0(\emmod4)
and let the admissible modules be decomposed into the direct sums:
[TABLE]
Then the Lie algebras Nr,s(U) and Nr,s(U~) are isomorphic, if and only if,
[TABLE]
or
[TABLE]
Theorem 4.1.3**.**
Let r≡3(\emmod4) and
s≡1,2,3(\emmod4)
and let U and U~ be decomposed into the direct sums
[TABLE]
Then Nr,s(U)≅Nr,s(U~), if and only if
p=p+=p~+=p~ and q=p−=p~−=q~, or p=p+=p~−=q~ and q=p−=p~+=p~.
4.2. Periodicity of isomorphisms
We can reduce the proof of the main theorems to basic cases (2.8), due to the following facts. Let Vminμ,ν be a minimal admissible module of the
Clifford algebra Clμ,ν, where (μ,ν)∈{(8,0),(0,8),(4,4)}. It was explained in Example 1 that
Vminμ,ν admits decomposition (2.7). The admissible scalar product restricted to Eμ,ν is
necessarily sign definite and we can fix it to be positive
definite scalar product on Eμ,ν by Lemma 3.2.5.
We summarise the results of Section 2.4 and Lemma 2.6.1.
Proposition 4.2.1**.**
Let (Vminr,s,⟨⋅,⋅⟩Vminr,s)
be a minimal admissible module of Clr,s and
Jzi, i=1,…,r+s
the Clifford actions of the orthonormal basis {zi}. Then
[TABLE]
is a minimal admissible module Vminr+μ,s+ν of the Clifford algebra Clr+μ,s+ν.
Conversely, if Vminr+μ,s+ν is a minimal admissible module of the algebra
Clr+μ,s+ν, then the common 1-eigenspace E0 of the involutions
Ti, i=1,2,3,4 from Example 1 can be considered as a minimal admissible module Vminr,s of the algebra
Clr,s. The action of the Clifford algebra Clr,s on
E0 is the restricted action of Clr+μ,s+ν obtained by the natural inclusion
Clr,s⊂Clr+μ,s+ν.
Proposition 4.2.2**.**
According to the correspondence of minimal admissible modules
stated in Proposition 4.2.1, there
is a natural injective map
[TABLE]
Conversely, automorphisms of the form A⊕C∈Aut0(Nr+μ,s+ν(Vminr+μ,s+ν)) with the property that
C(ζj)=ζj, j=1,…,8,
defines an automorphism A∣E0⊕C∣Rr,s of
the algebra Nr,s(E0), where
the space E0 is the common 1-eigenspace of the
involutions Tj, j=1,2,3,4, viewed as a minimal admissible module of Clr,s.
Proof.
Let Ar,s⊕C∈Aut0(Nr,s(Vminr,s)) with (Ar,s)τJziAr,s=JCτ(zi), i=1,…,r+s,
and let Jζj, j=1,…,8, be the actions on
Vminμ,ν of the Clifford algebra Clμ,ν.
We want to construct Aˉ:Vminr+μ,s+ν=Vminr,s⊗Vminμ,ν→Vminr+μ,s+ν=Vminr,s⊗Vminμ,ν
such that zi↦C(zi) and ζj↦ζj by using
the map Ar,s:Vminr,s→Vminr,s.
The action Jˉ on
Vminr,s⊗Vminμ,ν
defined in
Proposition 2.4.1
corresponds to
[TABLE]
according to the decomposition (4.3).
We define Aˉ:Vminr,s⊗Vminμ,ν→Vminr,s⊗Vminμ,ν
on each component of the decomposition (4.3)
such that it satisfies (3.3) with
Cˉ being Cˉ(zi)=C(zi) and Cˉ(ζj)=ζj.
It can be done in a unique way as in Example 2 and the operator Aˉ⊕Cˉ will satisfy Proposition 3.2.3.
Conversely, let Aˉ⊕Cˉ∈Aut0(Nr+μ,s+ν(Vminr+μ,s+ν))
be such that Cˉ(ζj)=ζj, j=1,…,8.
Then Vminr+μ,s+ν is decomposed into the orthogonal
sum (2.9)
and the commutativity of the
operators Jzi with the involutions Tj allows us to define an
automorphism A∣E0⊕C∣Rr,s of
the pseudo H-type algebra Nr,s(E0).
∎
Note that the construction given in Proposition 2.4.1
can be performed for an arbitrary, not necessary minimal admissible module Ur,s. Thus we obtain that Ur+μ,s+ν=Ur,s⊗Vminμ,ν is admissible for (μ,ν)∈{(8,0),(0,8),(4,4)} if Ur,s is admissible. Denote by Kr,s(Ur,s) the kernel of the map
[TABLE]
Corollary 4.2.3**.**
Let Ur,s and Ur+μ,s+ν=Ur,s⊗Vminμ,ν be admissible modules.
Then
[TABLE]
that is the kernel Kr,s(Ur,s) is invariant under the map B defined in Proposition 4.2.2.
Lemma 4.2.4**.**
If the Lie algebras Nr,s(Vminr,s) and Nr,s(V~minr,s) are isomorphic, then the Lie algebras Nr+μ,s+ν(Vminr+μ,s+ν) and Nr+μ,s+ν(V~minr+μ,s+ν) are also isomorphic for (μ,ν)∈{(8,0),(0,8),(4,4)}.
Proof.
Let Φ=A⊕C:Nr,s(Vminr,s)→Nr,s(V~minr,s) be a Lie algebra isomorphism, then Φ that can be extended to Φˉ=Aˉ⊕Cˉ:Nr+μ,s+ν(Vminr+μ,s+ν)→Nr+μ,s+ν(V~minr+μ,s+ν) by the same procedure as in Lemma 4.2.2. Namely, we set Aˉ=A⊗Id and Cˉ=C⊕Id.
∎
Let now Ur,s and U~r,s be two admissible modules of equal dimensions for the Clifford algebra Clr,s. Then Ur,s=⊕k(Vminr,s)k and U~r,s=⊕k(V~minr,s)k. Then admissible modules Ur+μ,s+ν and U~r+μ,s+ν can be identified with
[TABLE]
and
[TABLE]
Now applying Lemma 4.2.4 we obtain the following result.
Theorem 4.2.5**.**
If the Lie algebras Nr,s(Ur,s) and Nr,s(U~r,s) are isomorphic, then the Lie algebras Nr+μ,s+ν(Ur+μ,s+ν) and Nr+μ,s+ν(U~r+μ,s+ν) are also isomorphic for (μ,ν)∈{(8,0),(0,8),(4,4)}.
In order to prove the classification theorems for the pseudo H-type Lie algebras Nr,s(U), one should be careful about the scalar product
on each minimal admissible component of the decompositions (4.1) and (4.2) of the admissible module U. Let us assume that U=⊕iVi, where Vi are minimal admissible modules. If we find linear maps Aij:Vi→Vj for all i and j such that Aij⊕C:Nr,s(Vi)→Nr,s(Vj) are the Lie algebra isomorphisms with CCτ=IdRr,s, then the Lie algebra Nr,s(U) is unique. Even though a different choice of a scalar product on the vector space U gives different minimal admissible modules Vi in the decomposition U=⊕iVi, the resulting Lie algebras Nr,s(U) can be isomorphic if there is a map C:Rr,s→Rr,s that is the same for all Aij:Vi→Vj. For the simplicity we choose C to be identity on Rr,s.
The construction of maps Aij:Vi→Vj depends on the signature of the restriction of the admissible scalar product on the common 1-eigenspace of each minimal admissible module Vi from the direct sum U=⊕iVi. The proof of Theorem 4.1.1 is given in three lemmas according to whether the common 1-eigenspace is sign definite or neutral space and depends also on the type of the decomposition U=⊕iVi in (4.1) and (4.2). The first lemma concerns with the cases when there are only two types of minimal admissible modules Vminr,s;+ and Vminr,s;− and
the restrictions of the scalar product onto the common 1-eigenspace is sign definite and have different sign.
Lemma 4.3.1**.**
The Lie algebras Nr,s(Vminr,s;+) and
Nr,s(Vminr,s;−) are
isomorphic for r≡0,1,2(\emmod4)s≡0(\emmod4)
under the map A⊕IdRr,s.
Proof.
We consider the case r≡0,2(mod4).
In this case the existence of an
isomorphism A⊕IdRr,s:Nr,s(Vminr,s;+)→Nr,s(Vminr,s;−) is equivalent to the existence of the
automorphism A⊕−IdRr,s:Nr,s(Vminr,s;+)→Nr,s(Vminr,s;+) by
Lemma 3.3.4. The necessary automorphism exists by
Corollary 3.3.3.
Let r≡1(mod4).
We need only to consider the cases (1,0)(5,0) and (1,4) due to periodicity. The case N1,0(Vmin1,0;±) is trivial by the uniqueness of three dimensional Heisenberg algebra.
Let (r,s)=(5,0). We can assume that the minimal admissible module Vmin6,0 carries positive definite scalar product due to the first part of Lemma 4.3.1. Moreover, since dim(Vmin6,0)=dim(Vmin5,0;+)=8 the minimal admissible module Vmin5,0;+ can be thought as the
restriction of
the minimal admissible module
Vmin6,0≅R8,0
by restricting the action of Jz:R6,0→End(Vmin6,0) through the inclusion map
R5,0⊂R6,0 as well as the restriction of the scalar product ⟨⋅,⋅⟩Vmin6,0 onto Vmin5,0;+. Let π:R6,0→R5,0 be the orthogonal projection. Then
[TABLE]
is a Lie algebra homomorphism. Let A⊕−IdR6,0∈Aut0(N6,0(Vmin6,0)), which existence is guaranteed by Corollary 3.3.3. The property
[TABLE]
and the homomorphism (4.4) allow to descend the automorphism A⊕−Id∈Aut0(N6,0(Vmin6,0)) to the
automorphism of N5,0(Vmin5,0;+) with the same map A:Vmin5,0;+→Vmin5,0;+. Now applying Lemma 3.3.4, we conclude that there is an isomorphism A⊕Id:Vmin5,0;+→Vmin5,0;− that finishes the proof.
The existence of an isomorphism A⊕Id:Vmin1,4;+→Vmin1,4;− can be deduced from the existence of an automorphism A⊕−Id:Vmin2,4;+→Vmin2,4;+ as in the previous case.
∎
The following two lemmas give the rest of the proof of
Theorem 4.1.1 and they are concerned with indices
r≡0,1,2(mod4), s≡1,2,3(mod4) for
which the restriction of the admissible scalar product to the common 1-eigenspace is neutral. Lemma 4.3.2 deals with the indices (r,s)∈/{(1,2),(1,6),(5,2)}, because in these cases any admissible module U has decomposition (4.1). If (r,s) belongs to {(1,2),(1,6),(5,2)} then an admissible module U has decomposition (4.2) and the results of Lemma 4.3.2 are extended in Lemma 4.3.3.
Lemma 4.3.2**.**
Let (Vminr,s;+=(V,⟨⋅,⋅⟩Vminr,s;+) and Vminr,s;−=(V,⟨⋅,⋅⟩Vminr,s;−) be two minimal admissible modules of Clr,s with ⟨⋅,⋅⟩Vminr,s;+=−⟨⋅,⋅⟩Vminr,s;−. If the restrictions of both scalar products on the common 1-eigenspace Er,s of involutions from PIr,s
are neutral, then the Lie algebras Nr,s(Vminr,s;+) and Nr,s(Vminr,s;−) are isomorphic under the isomorphism A⊕IdRr,s.
Proof.
Notice that the system of involutions PIr,s does not depend on the scalar product and therefore the common 1-eigenspace Er,s is the same for both modules. The restrictions of the scalar products on Er,s are neutral by hypothesis. We find v,u∈Er,s such that ⟨v,v⟩Vminr,s;+=⟨u,u⟩Vminr,s;−=1. We find the orthonormal bases
[TABLE]
[TABLE]
Then the map A⊕IdRr,s is the isomorphism of the Lie algebras Nr,s(Vminr,s;+) and Nr,s(Vminr,s;−), where we set
A:xi↦yi and then extended it by linearity.
Indeed, let zk∈Rr,s be arbitrary and Jzk:V→V be the Clifford action. Then
[TABLE]
since the calculations depend only on number of permutations in the products.
∎
In the cases (r,s)∈{(1,2),(1,6),(5,2)} there are two irreducible modules Virr;±r,s, but they are not admissible. The minimal admissible modules are Vmin;+r,s=Virr;+r,s⊕Virr;+r,s and Vmin;−r,s=Virr;−r,s⊕Virr;−r,s, see Corollary 2.2.4, item (3-2-1). Thus any admissible module U is decomposed on the direct sum of the type (4.2).
Lemma 4.3.3**.**
The Lie algebras
[TABLE]
for (r,s)∈{(1,2),(1,6),(5,2)} are isomorphic under the maps Φk=Ak⊕C, k=1,2,3 with C=IdRr,s.
Proof.
The existence of the maps Φ1 and Φ3 follows from Lemma 4.3.2. We need only to construct Φ2.
Case N1,2
Let v∈Vmin;+1,2;+ be such that ⟨v,v⟩Vmin;+1,2;+=1 and u∈Vmin;−1,2;− with
⟨u,u⟩Vmin;−1,2;−=−1. It is possible, since both scalar products are neutral on the module. Then the vectors
[TABLE]
form the orthonormal bases for Vmin;+1,2;+ and Vmin;−1,2;−, respectively.
We define the correspondence:
A:xi↦yi and C:zi↦zi and extend it by linearity
It is easy to check that Φ2=A⊕IdR1,2
defines an isomorphism between the Lie algebras
N1,2(Vmin;+1,2;+) and N1,2(Vmin;−1,2;−).
Case N1,6.
This case is similar to N1,2 and
we construct an isomorphism
Φ=A⊕Id:N1,6(Vmin;+1,6;+)→N1,6(Vmin;−1,6;−). We have
JΩ1,6≡Id on Vmin;+1,6;+, that implies
P3≡−Id, see Table 7. It also shows that
E1,6=E1,6(Vmin;+1,6;+)={v∈Vmin;+1,6∣P1(v)=v,P2(v)=v}.
Then if necessary, we apply
Proposition 3.4.1, with
the operators Λ1=Jz2Jz4Jz6 and Λ2=Jz2Jz4Jz7 and
obtain the orthonormal basis of Vmin;+1,6;+
from a suitable vector v∈E1,6(Vmin;+1,6;+)
with ⟨v,v⟩Vmin;+1,6;+=1:
[TABLE]
Let u∈E1,6(Vmin;−1,6;−)⊂Vmin;−1,6;− with ⟨u,u⟩Vmin;−1,6;−=−1. Then
by the same way as for Vmin;+1,6;+ we obtain the
orthonormal basis of Vmin;−1,6;−:
[TABLE]
Then as previously, the correspondence
A:xi↦yi,i=1,…,16,
defines the Lie algebra isomorphism
Φ=A⊕Id:N1,6(Vmin;+1,6;+)→N1,6(Vmin;−1,6;−),
since the map A satisfies relation (3.3).
Case N5,2. In this case we can use bases (4.5) and (4.6), since PI5,2=PI1,6. The table of commutators will differ by signs for z1,…,z7.
∎
Theorem 4.1.2 is concerned with the indices r≡3(mod4)
and s≡0(mod4) and is given in Lemmas 4.4.1 and 4.4.3. The cases with s=0 are classical and the result is known, for instance from [8, 13], however in order
to accomplish the whole classification of pseudo H-type Lie algebras
we must take into account that the Lie algebras Nr,0(U) can admit a negative definite admissible scalar product on U. Thus, we can obtain the opposite sign of the restriction of the admissible scalar product on the common 1-eigenspace even for classical cases.
Proposition 4.4.1**.**
Let r≡3(\emmod4)
and s≡0(\emmod4).
Then an admissible module U is decomposed into the direct sum of type (4.2). Moreover
(1)
There is a Lie algebra isomorphism Φ:Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;−r,s;−) of the form
Φ=A⊕Id. There is no isomorphism of the form Φ=A⊕−Id between these algebras.
Analogous results can be stated for the Lie algebras isomorphisms Nr,s(Vmin;+r,s;−)→Nr,s(Vmin;−r,s;+).
(2)
There is a Lie algebra isomorphism
Φ:Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;+r,s;−)
of the form Φ=A⊕C with detC<0 and there is no isomorphism of the form Φ=A⊕Id.
Analogous results hold for the Lie algebra isomorphisms
Nr,s(Vmin;−r,s;+)→Nr,s(Vmin;−r,s;−), Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;−r,s;+), Nr,s(Vmin;+r,s;−)→Nr,s(Vmin;−r,s;−).
Proof.
We start from the proof of the first part. We restrict the consideration to the basic cases
(r,s)∈{(3,0),(3,4),(7,0)} because of the periodicity Theorem 4.2.5. We also can assume that C=IdRr,s. In order to construct an isomorphism Φ=A⊕IdRr,s:Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;−r,s;−)
we choose v∈Er,s⊂Vmin;+r,s;+ with ⟨v,v⟩Vmin;+r,s;+=1
and J(Ωr,s)v=v and a vector
u∈Er,s⊂Vmin;−r,s;− with ⟨u,u⟩Vmin;−r,s;−=−1 and
J~(Ωr,s)u=−u. Here Ωr,s is the volume form of the Clifford algebra Clr,s with actions J:Clr,s→End(Vmin;+r,s;+) and J~:Clr,s→End(Vmin;−r,s;−).
Let (r,s)=(3,0). The respective orthonormal bases are the following:
[TABLE]
In the case (r,s)=(7,0) the initial vector v∈E7,0⊂Vmin;+7,0;+ satisfies P1v=P2v=P3v=P4v=v, where the involutions are given in Table 4.
Note that J(Ω7,0)=P1P4v=v. The initial vector u∈E7,0⊂Vmin;−7,0;− for the basis has to satisfy
P1u=P2u=P3u=−P4u=u with J~(Ω7,0)u=P1P4u=−u.
The bases are
[TABLE]
Let (r,s)=(3,4). The basis is given by (4.7) and Ω3,4=P1P3P4, where the involutions Pi are presented in Table 6.
The maps Φ=A⊕IdRr,s in all the cases are given
by correspondence A:xi↦yi. The Lie algebra isomorphisms
Φ=A⊕IdRr,s:Nr,s(Vmin;+r,s;−)→Nr,s(Vmin;−r,s;+) are constructed analogously.
Assume that the isomorphism Φ=A⊕C with detC<0 exists, where A:Vmin;+r,s;+→Vmin;−r,s;−. If (r,s)∈{(3,0),(7,0)}, then Aτ=−tA. and Ωr,0=−Ω~r,0=Id since the minimal admissible modules correspond to the non-equivalent irreducible modules.
Then
[TABLE]
by (3.3). This is a contradiction,
since the matrix tAA is positive definite.
Let (r,s)=(3,4). The admissible scalar products restricted to common 1-eigenspace Er,s are sign definite and the
symmetric bi-linear forms
⟨x~,y~⟩Vmin;−r,s;− and
⟨x,y⟩Vmin;+r,s;+ restricted to the common 1-eigenspaces
are related through the equalities
[TABLE]
The signs of the values of two symmetric bi-linear forms
coincide if detC<0 and
opposite if detC>0. We conclude that there is no Lie algebra isomorphism Φ:Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;−r,s;−) of the form Φ=A⊕−IdRr,s. Remind that the map A maps the common
1-eigenspace from Vmin;+r,s;+ to common 1-eigenspace
from Vmin;+r,s;− by the construction described in
Section 3.2 after Proposition 3.2.4.
The results for the Lie algebras Nr,s(Vmin;+r,s;−) and Nr,s(Vmin;−r,s;+) can be shown similarly.
We prove now the second part of the lemma. The isomorphisms
[TABLE]
under the map Φ=Id⊕−IdRr,s are given in
Lemma 3.2.5. The reader can find the isomorphisms of the form Φ=A⊕C with detC<0 for the Lie algebras
[TABLE]
in [18, Theorem 12]. The non existence results are proved in a similar way as for the part 1 of the Lemma.
∎
We state separately a corollary of Proposition 4.4.1 that is core for the proof of Theorem 4.1.2.
Corollary 4.4.2**.**
There exists a Lie algebra isomorphism Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;−r,s;−) of the form
Φ=A⊕Id. There does not exist a Lie algebra isomorphism Nr,s(Vmin;+r,s;+)→Nr,s(Vmin;+r,s;−) of the form Φ=A⊕Id.
Lemma 4.4.3**.**
The Lie algebras Nr,s(U), (r,s)∈{(3,0),(7,0),(3,4)}, are completely determined by the pair of numbers (p=p+++p−−, q=p+−+p−+)
in the decomposition of the admissible module U:
[TABLE]
Proof.
Let U be an arbitrary
admissible module of Clr,s for (r,s)∈{(3,0),(7,0),(3,4)}.
First we decompose U into the sum of irreducible modules:
[TABLE]
Then, Ωr,s=IdU+ and Ωr,s=−IdU−.
We decompose the submodules U+ and U− into the minimal admissible modules
[TABLE]
where Vmin;±r,s;+ are the minimal admissible modules for which the restriction of the admissible scalar product on the common 1-eigenspace Er,s is positive definite and Vmin;±r,s;− are those where the restriction of the scalar product on Er,s is negative definite. It was stated in Proposition 4.4.1, item (1), and Corollary 4.4.2 that
[TABLE]
where Φ=A⊕Id. Thus we can consider only the case when the restrictions on Er,s of the scalar product is positive definite.
Since the isomorphism between Lie algebras Nr,s(Vmin;+r,s;+) and Nr,s(Vmin;−r,s;+) can not admit the form A⊕Id by Corollary 4.4.2, we conclude that two Lie algebras Nr,s(U) and Nr,s(U~) with
[TABLE]
are isomorphic if and only if either (p,q)=(p~,q~) or (p,q)=(q~,p~).
∎
In order to prove Theorem 4.1.3, we consider the low values of the indices:
r=3, s=1,2,3,5,6,7 or r=7, s=1,2,3 and then we apply the periodicity Theorem 4.2.5. An admissible module U has decomposition of the type (4.1). Theorem 2.6.2 shows that the restriction of the scalar product ⟨⋅,⋅⟩Vminr,s on the common 1-eigenspace E3,s for s=1,2,3,4,5,6,7 and E7,s, s=1,2,3 is sign definite. We denote by Vminr,s;+=(Vminr,s;+,⟨⋅,⋅⟩Vminr,s;+) the minimal admissible module with positive definite metric on Er,s. Analogously, we write Vminr,s;−=(Vminr,s;−,⟨⋅,⋅⟩Vminr,s;−) for the minimal admissible module with negative definite metric on Er,s. We note that all the systems PIr,s include the involution P=Jz1Jz2Jz3 for mentioned values of r and s.
Consider pseudo H-type algebras N3,s(Vmin3,s;+) and N3,s(Vmin3,s;−) for s=1,2,3. The natural inclusion R3,0⊂R3,s allows to consider the module Vmin3,s;+ as the admissible module U=Vmin;+3,0;+⊕Vmin;−3,0;− of Cl3,0. The module U have to include both eigenspaces of the involution P=Jz1Jz2Jz3 and therefore U=Vmin3,s;+ includes both irreducible modules Vmin;±3,0. The metric on U=Vmin3,s;+ is neutral and therefore the irreducible modules Vmin;±3,0 have to carry the definite metrics of opposite signs. Analogous considerations can be done with Vmin3,s;−. This and the item (1) of Lemma 4.4.1 imply the existence of the isomorphisms
[TABLE]
for any s=1,2,3. More generally there are Lie algebra isomorphisms
[TABLE]
The projection map
π:R3,s→R3,0, where zi, i=3+1,…3+s, are
mapped to zero, defines a Lie algebra surjective homomorphism
Id⊕π:N3,s(U)→N3,0(U).
We assume now that there exists a Lie algebra isomorphism
[TABLE]
Then it defines a Lie algebra isomorphism
A⊕C′:N3,0(U)→N3,0(U) where C′ is the restriction C∣R3,0 of the map C on R3,0. We define
[TABLE]
Then the diagram
[TABLE]
commuts. We conclude that N3,s(U) and N3,s(U) are isomorphic,
if and only if p+=p~+ and p−=p~− or p+=p~− and p−=p~+ by Theorem 4.1.2, see also
[8].
In order to prove the reduction of N3,s(Vmin3,s;±) to N3,0(Vmin;±3,0;±) for s=5,6,7 we observe that there are Lie algebra isomorphisms
[TABLE]
where α(5)=2, α(6)=4 and α(7)=8. The rest of the proof is made analogously.
Literally the same reduction is made for the Lie algebras N7,s(Vmin7,s;±) to the Lie algebra N7,0(Vmin;±7,0;±) for s=1,2,3. The theorem is proved.
4.6. Final result of the classification
The following statement can be proved analogously to Theorem 4.2.5.
Theorem 4.6.1**.**
If the Lie algebras Nr,s(Ur,s) and Ns,r(U~s,r) are isomorphic, then the Lie algebras Nr+μ,s+ν(Ur+μ,s+ν) and Ns+ν,r+μ(U~s+ν,r+μ) are also isomorphic for (μ,ν)∈{(8,0),(0,8),(4,4)}.
We showed in Theorem 4.1.1 that the Lie algebras
Nr,s(U) for r≡0,1,2(mod4) are defined by the
dimension of the admissible module U.
If r≡3(mod4) the Lie algebra Nr,s(U) depends on the decompositions
[TABLE]
where the numbers p,q are defined in Theorems 4.1.2 and 4.1.3. We call admissible modules with decompositions (4.10) isotypic if one of the numbers p or q vanishes. Otherwise the admissible module is called nonisotypic.
Theorem 4.6.2**.**
Let r≡0,1,2(\emmod4) and
s≡0,1,2(\emmod4).
Then Nr,s(U)≅Ns,r(U~) if dim(U)=dim(U~).
Let r≡3(\emmod8), s≡0,4,5,6(\emmod8)
or
r≡7(\emmod8), s≡0,1,2(\emmod8).
Then Nr,s(U)≅Ns,r(U~) if dim(U)=dim(U~)
and U is an isotypic admissible module.
Let r≡3(\emmod8) and s≡1,2,7(\emmod8).
Then N3,s(U) is never isomorphic to Ns,3(U~).
We summarise the results of Theorem 4.6.2 in Table 3. We distinguish the columns for r=3 and r=7 for isotypic and nonisotypic modules. We write the symbol ≅ in the place (r,s) if Nr,s(U) is isomorphic to Ns,r(U~) and the isomorphism only depends on the dimension of the admissible module. For example, N3,0(U)≅N0,3(U~) if dim(U)=dim(U~) and U is isotypic and N3,0(U)≅N0,3(U~) if U is non-isotypic. We have N3,1(U)≅N1,3(U~) for any admissible modules U and U~ even if dim(U)=dim(U~) and the module U of Cl3,1 is isotypic: U=⊕pVmin3,1;+ or U=⊕qVmin3,1;−.
5. Appendix
We give the collections PIr,s
and COr,s for basic cases (2.8) grouped in four tables. The
dimensions of Er,s and signature of the scalar product restricted to Er,s
are listed.
First we mention trivial cases.
[TABLE]
For the cases (r,s) of r−s≡3(mod4) and s even,
there are no a complementary operator which commutes with all the
involutions in PIr,s except the last involution which is of the form P3 or P4 and anti-commutes with the last involution.
In these cases the operator JΩr,s is a product of
involutions in PIr,s and it commutes with all the
operators Jz. This is the reason for the number of complementary operators to be pr,s−1. The last operator in PIr,s of the form P3 or P4 distinguishes the two different minimal admissible modules.
The signature on the admissible scalar product restricted on
the space Er,s is sign definite in Table 6 and
is neutral for signatures (r,s) in Table 7.
The latter can be seen by finding an additional negative operator other than
operators in COr,s which commutes with
all the involutions in PIr,s.
Bibliography32
The reference list from the paper itself. Each links out to its DOI / PubMed record.
1[1] D. V. Alekseevsky, V. Cortés, Classification of N 𝑁 N -(super)-extended algebras and bilinear invariants of the spinor representation of S p i n ( p , q ) 𝑆 𝑝 𝑖 𝑛 𝑝 𝑞 Spin(p,q) . Comm. Math. Phys. 183 (1997), no. 3, 477–510.
2[2] A. Altomani, A. Santi, Classification of maximal transitive prolongations of super-Poincaré algebras . Adv. Math. 265 (2014), 60–96.
3[3] M. S. Atiyah, R. Bott, A. Shapiro, Clifford modules. Topology 3 (1964), 3–38.
4[4] C. Autenried, K. Furutani, I. Markina, A. Vasiliev, Pseudo-metric 2-step nilpotent Lie algebras , ar Xiv 1508.02882, to appear in Adv. Geom.
5[5] P. E. Barbano, Automorphisms and quasi-conformal mappings of Heisenberg-type groups. J. Lie Theory 8 (1998), no. 2, 255–277.
6[6] W. Bauer, K. Furutani, C. Iwasaki, Spectral zeta function on pseudo H-type nilmanifolds . Indian J. Pure Appl. Math. 46 (2015), no. 4, 539–582.
7[7] A. Bellaïche and J. J. Risler, Sub-Riemannian geometry . Edited by André Bellaïche and Jean-Jacques Risler. Progress in Mathematics, 144. Birkhäuser Verlag, Basel, 1996. 393 pp.
8[8] J. Berndt, F. Tricerri, L. Vanhecke, Generalized Heisenberg groups and Damek-Ricci harmonic spaces. Lecture Notes in Mathematics, 1598 . Springer-Verlag, Berlin, 1995. 125 pp.