November 27, 2006
CERN-PH-TH/2005-006
.
Lorentzian homogeneous spaces admitting a homogeneous structure of type
Patrick Meessen
(PM) Physics Department, Theory, CERN, Geneva, Switzerland E-mail address : Patrick.Meessen@cern.ch
-
Abstract.
We show that a Lorentzian homogeneous space admitting a homogeneous structure of type
is either a (locally) symmetric space or a singular homogeneous plane wave.
A theorem by Ambrose and Singer [1] , generalized to arbitrary signature in [2] , states that on a reductive homogeneous space, there exists a metric-compatible connection
, with
the Levi-Cività connection, that parallelizes the Riemann tensor
, and the
-tensor
, i.e.
. Since a
-tensor in
decomposes into 3 irreps of
, one can classify the reductive homogeneous spaces by the occurrence of one of these irreps in the tensor
[3, 4] . This leads to 8 different classes, which range from the maximal, denoted
, to the minimal
. Clearly homogeneous spaces of type
are just symmetric spaces. Moreover, also the homogeneous spaces admitting a homogeneous structure of type
(
or
) have been characterized. For the case at hand it is worth knowing that the homogeneous spaces with a
structure, for which
corresponds to a 3-form, are naturally reductive spaces [3, 4] and that strictly Riemannian homogeneous
spaces are symmetric spaces [3] .
Since a homogeneous structure of type
is defined by an invariant vector field, denoted by
, one must distinguish between two cases in the Lorentz setting: the non-degenerate case, for which
is a spaceor time-like vector, and the degenerate case, when
is a null vector. In the former case, Gadea and Oubin͂a [4] proved that, analogously to the strictly Riemannian case, the space is symmetric. In the degenerate case, Montesinos Amilibia [5] showed that a Lorentzian homogeneous space admitting a degenerate
structure is a time-independent singular homogeneous plane wave [6] . A small calculation shows that a generic, i.e. time-dependent, singular homogeneous plane wave admits a degenerate
structure, see e.g.
Appendix A . (By a (non-)degenerate
structure, we mean that the vector field
characterizing the
contribution has (non-)vanishing norm.) This then automatically leads to the question of whether the singular homogeneous plane waves exhaust the degenerate
class. As we will see, this is actually the case.
In the
case the homogeneous structure is given by [3]
where we have defined
, and
is the
contribution.
Since the metric and
are parallel under
, and
is the contraction of
, it follows that
or, written differently:
This equation, together with the fact that
is a 3-form, implies that
, i.e.
is a geodesic vector.
Given an isometry algebra
of a Lie group acting transitively on a given homogeneous space, with a reductive split
, where
is the isotropy subalgebra, it is possible, and usually done, to identify
with
; the action of
on
can then be given by the vector representation of
[7] . This identification enables one to express the algebra in terms of
and the curvature
as, limiting ourselves to the
commutator,
|
(1)
|
where
and
are evaluated at some point
. In the above formula,
signals the presence of
in
. From now on, we only consider this Lie algebra and all the relevant tensor fields are evaluated at a specific point, even though this is not stated explicitly.
Up to this point not too much has been said about
, and in fact not too much can be said. It is known, however [7] , that a tensor field parallelized by
, when evaluated at a point corresponds to an
-invariant tensor. Since in this article we take
(a
-invariant vector field as
) to be non-vanishing, this means that
when
is light-like,
when
is space-like, and
when
is null.
Let us briefly outline the manner in which we arrive at our conclusion: given a reductive homogeneous space with reductive split
, the subalgebra
is an ideal of
. It is this ideal, which is the Lie algebra of a Lie group still acting transitively, that we will consider; we will say that an element of
appears in the algebra if it is an element of
. Given the homogeneous structure, we can then, following Eq. ( 1 ), write down the maximal form of the algebra compatible with the homogeneous structure. Since we are dealing with a Lie algebra, we can then use the Jacobi identities to constrain the structure constants; after a redefinition of some generators in
, corresponding to the choice of a different reductive split, this leads to a recognizable result. Since the non-degenerate case is far less involved than the degenerate case, and gives a better idea of the straightforward manipulations used, it will be discussed before the degenerate case.
1 The non-degenerate case
Let
be spanned by the generators
and
(
), which in this case we can take to satisfy
| |
| |
where
distinguishes between the time-like (for
) and the space-like (for
) cases and
. As is mentioned in the introduction,
is either contained in
(
) or
(
) and the relevant non-vanishing commutation relations are
| |
| |
Once again, let us stress that not every
needs appear, but the elements of
can be written as combinations of the
's, and their commutation relations are induced by the ones above.
With respect to the chosen basis we can decompose
and
, which allows us to write
| |
| |
Let us then, following the strategy outlined above, check the Jacobi identities.
The first one is the
identity, which leads to
and
|
(2)
|
|
(3)
|
where we expanded
and
. Since
we can redefine
from which we trivially find
which at once implies that
, by Eq. ( 2 ), and also that
thanks to Eq. ( 3 ). So the, quite remarkable, result is that a Lorentzian homogeneous space admitting a non-degenerate homogeneous structure of type
, also admits a non-degenerate
structure. Combining this with the results of Gadea and Oubin͂a [4] , we have proven that:
Proposition 1.
A connected Lorentzian homogeneous space admitting a non-degenerate
structure is a (locally) symmetric space.
2 The degenerate case
In the degenerate case we can choose the generators
,
and
(
) spanning
such that
,
. The invariant norm is then
and
and we decompose the
contribution to
as
| |
| |
where
,
and
are totally antisymmetric. Given these abbreviations we can write the most general
commutators as
| |
| |
| |
| |
where the various
need to be expanded in terms of the generators of
. Since
is null, we see that
, which we take to be spanned by
and
with commutation relations
| |
| |
| |
| |
where it should be kept in mind that not all elements of
need appear.
We can then once again start to recover the information contained in the Jacobi identities: the
Jacobi identity reads
| |
| |
|
(4)
|
Cancellation of the
contribution then means that
, which at once means that
can only be non-zero for those directions for which a
appears. Specifically, should none appear, then
. Let us then split the index
into some indices
and
, such that the
do appear whereas the
do not.
Having made the split, we can investigate the implication of having the null-boosts in the algebra. Let us start by looking at the
Jacobi: a small calculation then shows that this implies
| |
| |
In order for the above to be true we must have that
and that
can be non-zero only if no or only one
appears in
. As was said above, the no-case already implies that
, so we had better have a look at the case of one appearing null boost. For this we are helped by the
-part of the above equation. Clearly in the case when we are dealing with only one
, this amounts to the statement that
, which, since there is no rotation in
that can take
to
, means that
, and hence that
.
This then means that in all cases we have
.
Continuing with the analysis, one can see that the
Jacobi leads to
| |
| |
Then, using the fact that
, one then sees that
and that hence
when
includes some null boost. In the case when there is no
, the relevant information can be obtained by picking out the
component in the
Jacobi: this implies that
, which after contraction leads to
and thus implies that
.
The
-part of Eq. ( 4 ) then implies that
, so that
. In order to then identically satisfy Eq. ( 4 ) we must have
, so that
.
Summarizing the results obtained thus far, we find that the non-trivial
-commutators, scaling
in such a way that
and decomposing the various
's, are
| |
| |
Let us then continue our analysis of the Jacobi identities: the
Jacobi implies
| |
| |
|
(5)
|
|
(6)
|
where we defined
From Eq. ( 6 ) one sees that
must be totally antisymmetric. Denoting by
the sum over the permutations
,
and
, the
Jacobi results in
| |
| |
| |
and also, since
is totally antisymmetric,
Of course, if a
occurs in
, then the
Jacobi implies that
Let us then, as before, split the indices
into
, where the
's occur but the
's do not. This means by assumption that
, which implies
,
and
, which implies that only
is non-zero. Furthermore, we then see that only
is non-vanishing. Together with Eq. ( 7 ), this then implies that
, and we get the extra constraint
This last constraint also follows from the
Jacobi, which also tells us that
.
Eq. ( 8 ) then implies that only
and
are non-vanishing, and from Eq. ( 9 ) we find that only
can be non-zero. We can calculate
from Eq. ( 6 ), which then gives
because of Eq. ( 10 ). The same equation then states
, which by means of Eq. ( 5 ) then also implies that only the
can be non-vanishing.
Let us define the generator
from which we can then derive that the algebra takes on the form
| |
| |
| |
| |
so that the
and the
-sectors decouple from each other.
Restricting ourselves to the
-sector and further defining
we immediately find
; calculating the remaining commutator, we find
where the
stands for terms in
. Using now Eq. ( 6 ), we see that this redefinition trivializes
, and by way of Eq. ( 5 ), also
.
At this point, the only difference between the algebra we deduced and the generic singular homogeneous plane-wave algebra in Eq. ( 11 ) are the null boosts in the
-sector, that is a generator one would call
. It is, however, always possible to extend our algebra to an algebra that does contain them; in fact this follows immediately from the consistency of the singular homogeneous plane-wave algebra. Putting everything together, one sees that we obtain the isometry algebra of a generic singular homogeneous plane-wave in Eq. ( 11 ) by, basically, choosing a different reductive split of the same algebra. Thus we have proven that
Theorem 2.
The underlying geometry of a Lorentzian homogeneous space that admits a degenerate
structure is that of a singular homogeneous plane wave.
Acknowledgments
The author would like to thank J. Figueroa-O'Farrill, R. Hernández and S. Philip for very useful discussions, and S. Vascotto for improving the readability of the text.
A Singular homogeneous plane waves
A global coordinate system for the singular homogeneous plane waves is defined by the data1
| |
where the metric is defined by
and
. This class of metrics admits a homogeneous structure given by the components
which corresponds to a degenerate
structure.
The isometry algebra, apart from possible rotations that appear as automorphisms of the algebra, can be found to be [6]
|
(11)
|
References
-
W. Ambrose, I. Singer: “On homogeneous Riemannian manifolds”, Duke Math. J. 25 (1958), 647–669.
-
P. Gadea, J. Oubin͂a: “Homogeneous pseudo-Riemannian structures and homogeneous almost para-Hermitean structures”, Houston J. Math. 18 (1992), 449–465.
-
F. Tricerri, L. Vanhecke: “Homogeneous structures on Riemannian manifolds”, London Math. Soc. Lecture Note Ser. 83 (1983), 1–125.
-
P. Gadea, J. Oubin͂a: “Reductive homogeneous pseudo-Riemannian manifolds”, Monatsh. Math. 124 (1997), 17–34.
-
A. Montesinos Amilibia: “Degenerate homogeneous structures of type
on pseudo-Riemannian manifolds”, Rocky Mountain J. Math. 31 (2001), 561–579.
-
M. Blau, M. O'Loughlin: “Homogeneous plane waves”, Nuclear Phys. B 654 (2003), 135–176.
-
S. Kobayashi, K. Nomizu, “Foundations of differential geometry”, Wiley (1963 and 1969).
(PM) Physics Department, Theory, CERN, Geneva, Switzerland E-mail address : Patrick.Meessen@cern.ch