DFTT07/23
JINR-E2-2007-145
The arrow of time and the Weyl group:
arXiv:0710.1059v1 [hep-th] 4 Oct 2007
all supergravity billiards are integrable†
Pietro Fréa and Alexander S. Sorinb
a
Dipartimento di Fisica Teorica, Universitá di Torino,
& INFN - Sezione di Torino
via P. Giuria 1, I-10125 Torino, Italy
fre@to.infn.it
b
Bogoliubov Laboratory of Theoretical Physics,
Joint Institute for Nuclear Research,
141980 Dubna, Moscow Region, Russia
sorin@theor.jinr.ru
Abstract
In this paper we show that all supergravity billiards corresponding to σ-models on
any U/H non compact-symmetric space and obtained by compactifying supergravity
to D = 3 are fully integrable. The key point in establishing the integration algorithm
is provided by an upper triangular embedding of the solvable Lie algebra associated
with U/H into sl(N, R) which always exists. In this context we establish a remarkable relation between the arrow of time and the properties of the Weyl group. The
asymptotic states of the developing Universe are in one-to-one correspondence with the
elements of the Weyl group which is a property of the Tits Satake universality classes
and not of their single representatives. Furthermore the Weyl group admits a natural
ordering in terms of ℓT , the number of reflections with respect to the simple roots and
the direction of time flows is always towards increasing ℓT , which plays the unexpected
role of an entropy.
†
This work is supported in part by the European Union RTN contract MRTN-CT-2004-005104 and by the
Italian Ministry of University (MIUR) under contracts PRIN 2005-024045 and PRIN 2005-023102. Furthermore the work of A.S. was partially supported by the RFBR Grant No. 06-01-00627-a, RFBR-DFG Grant
No. 06-02-04012-a, DFG Grant 436 RUS 113/669-3, the Program for Supporting Leading Scientific Schools
(Grant No. NSh-5332.2006.2), and the Heisenberg-Landau Program.
1
Foreword
Notwithstanding its length and its somewhat pedagogical organization, the present one is a
research article and not a review. All the presented material is, up to our knowledge, new.
Due to the combination of several different mathematical results and techniques necessary
to make our point, which is instead physical in spirit and relevant to basic questions in
supergravity and superstring cosmology, we considered it appropriate to choose the present
somewhat unconventional format for our paper. After the theoretical statement of our
result, we have illustrated it with the detailed study of a few examples. These case-studies
were essential for us in order to understand the main point which we have formalized in
mathematical terms in part I and we think that they will be similarly essential for the
physicist reader. The table of contents helps the reader to get a comprehensive view of the
article and of its structure.
Contents
1 Foreword
1
I
2
Theory: Stating the principles
2 Supergravity billiards: a paradigm for cosmology
2
3 The
3.1
3.2
3.3
paint group and the Tits Satake projection
The solvable algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
The paint group and its Lie algebra . . . . . . . . . . . . . . . . . . . . . . .
The subpaint group and the Tits Satake subalgebra . . . . . . . . . . . . . .
4
5
5
6
4 Triangular embedding in SL(N, R)/SO(N) and integrability
4.1 The integration algorithm for the Lax Equation . . . . . . . . . . . . . . . .
6
8
5 Properties of the general integral and the parameter space
5.1 Discussion of the generalized Weyl group . . . . . . . . . . . . . . . . . . . .
5.2 The arrow of time, trapped and critical surfaces . . . . . . . . . . . . . . . .
9
10
12
II
Examples illustrating the principles
16
6 Choice of the examples
17
7 The simplest maximally split case: SL(3, R)/SO(3)
7.1 Discussion of the generalized Weyl group . . . . . . . . . . . . . . . . . . . .
7.2 The flow diagram and the critical surfaces for SL(3, R) . . . . . . . . . . . .
17
22
24
1
8 The
8.1
8.2
8.3
8.4
maximally split case Sp(4, R)/U(2)
The Weyl group and the generalized Weyl group of sp(4, R) . . . . . .
Construction of the sp(4, R) Lie algebra . . . . . . . . . . . . . . . . . .
Parameterization of the compact group U(2) and critical submanifolds .
Examples for sp(4, R) . . . . . . . . . . . . . . . . . . . . . . . . . . . .
8.4.1 An example of flow in the bulk of parameter space: Ω5 ⇒ Ω8 .
8.4.2 An example of flow on the super-critical surface Σ9 : Ω6 ⇒ Ω8 .
8.4.3 An example of flow on the super-critical surface Σ2 : Ω1 ⇒ Ω8 .
.
.
.
.
.
.
.
31
32
36
37
44
44
45
47
9 The case of the so(r, r + 2s) algebra
9.1 The corresponding complex Lie algebra and root system . . . . . . . . . . .
9.2 The real form so(r, r + 2s) of the Dr+s Lie algebra . . . . . . . . . . . . . . .
49
49
51
10 A case study for the Tits Satake projection: SO(2, 4)
10.1 The generalized Weyl group for SO(2, 4) . . . . . . . . . . . . . . . . . . . .
10.2 Vertices, edges and trapped surfaces . . . . . . . . . . . . . . . . . . . . . . .
10.3 Examples of flows for SO(2, 4) . . . . . . . . . . . . . . . . . . . . . . . . . .
56
59
61
62
III
68
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Perspectives
11 Summary of results
68
12 Open problems and directions to be pursued
69
Part I
Theory: Stating the principles
2
Supergravity billiards: a paradigm for cosmology
Cosmological implications of superstring theory have been under attentive consideration in
the last few years from various viewpoints [1]. This involves the classification and the study of
possible time-evolving string backgrounds which amounts to the construction, classification
and analysis of supergravity solutions depending only on time or, more generally, on a low
number of coordinates including time.
In this context a quite challenging and potentially highly relevant phenomenon for the
overall interpretation of extra–dimensions and string dynamics is provided by the so named
cosmic billiard phenomenon [2], [3], [4], [5]. This is based on the relation between the cosmological scale factors and the duality groups U of string theory. The group U appears as isometry group of the scalar manifold Mscalar emerging in compactifications of 10–dimensional
supergravity to lower dimensions D < 10 and depends both on the geometry of the compact
dimensions and on the number of preserved supersymmetries NQ ≤ 32. For NQ > 8 the
scalar manifold is always a homogeneous space U/H. The cosmological scale factors ai (t)
2
associated with the various dimensions of supergravity are interpreted as exponentials of
those scalar fields hi (t) which lie in the Cartan subalgebra of U, while the other scalar fields
in U/H correspond to positive roots α > 0 of the Lie algebra U. The cosmological evolution
is described by a fictitious ball that moves in the CSA of U and occasionally bounces on
the hyperplanes orthogonal to the various roots: the billiard walls. Such bounces represent
inversions in the time evolution of scale factors. Such a scenario was introduced by Damour,
Henneaux, Julia and Nicolai in [2], [3], [4], [5], generalizing classical results obtained in the
context of pure General Relativity [6]. In these papers the billiard phenomenon was mainly
considered as an asymptotic regime near singularieties.
In a series of papers [7], [8, 9, 10] involving both the present authors and other collaborators it was started and developed what can be described as the smooth cosmic billiard
programme. This amounts to the study of the billiard features within the framework of exact analytic solutions of supergravity rather than in asymptotic regimes. Crucial starting
point in this programme was the observation [7] that the fundamental mathematical setup
underlying the appearance of the billiard phenomenon is the so named Solvable Lie algebra
parametrization of supergravity scalar manifolds, pioneered in [11] and later applied to the
solution of a large variety of superstring/supergravity problems [12], [13], [16], [14], [15] (for
a comprehensive review see [17]).
Thanks to the solvable parametrization, one can establish a precise algorithm to implement the following programme:
a Reduce the original supergravity in higher dimensions D ≥ 4 (for instance D = 10, 11)
to a gravity-coupled σ–model in D ≤ 3 where gravity is non–dynamical and can be
eliminated. The target manifold is the non compact coset U/H ∼
= exp [Solv (U/H)]
metrically equivalent to a solvable group manifold.
b Utilize various group theoretical techniques in order to integrate analytically the σ–model
equations.
c Dimensionally oxide the solutions obtained in this way to extract time dependent solutions
of D ≥ 4 supergravity.
In view of the above observation we will use the following definition of supergravity billiards:
Definition 2.1 << A supergravity billiard is a one-dimensional σ-model whose target space
is a non-compact coset manifold U/H, metrically equivalent, in force of a general theorem,
to a solvable group manifold exp [Solv (U/H)]. >>
There exists a complete classification [17, 18, 10] of all non-compact coset manifolds U/H
relevant to the various instances of supergravities in all space-time dimensions D and for all
numbers NQ of supercharges. A general important feature is that maximal supersymmetry
NQ = 32 corresponds to maximally split symmetric cosets.
Definition 2.2 << A symmetric coset manifold U/H is maximally split when the Lie algebra U of U is the maximally non compact real section of its own complexification and H ⊂ U
is the unique maximal compact subalgebra. In this case the Cartan subalgebra C is completely
3
non-compact, namely the non-compact rank rn.c. = r equals the rank and the solvable Lie algebra Solv (U/H) is made by all the Cartan generators Hi plus the step operators E α for all
the positive roots α > 0. >>
In [19] the present authors shew that for maximally split cosets the one-dimensional σmodel is fully integrable and the general integral can be constructed using a well established
algorithm endowed with a series of distinctive and quite inspiring features.
In the present paper we demonstrate that the algorithm of [19] can be actually extended
to all the other cases, also those not maximally split, so that all supergravity billiards are in
fact completely integrable as claimed in the title.
Besides demonstrating the integrability we will illustrate the main features of the general
integral which reveal a very rich and highly interesting geometrical structure of the parameter
space. In this context it will emerge a challenging new concept. The time flows appearing
as exact analytical solutions of supergravity billiards have a preferred orientation which is
intrinsically determined in group theoretical terms. There emerges a similarity between the
second law of thermodynamics and the properties of cosmological evolutions just as there is
such a similarity in the case of black-hole dynamics. We establish the following principle
Principle 2.1 << The asymptotic states of the cosmic billiard at past and future infinity t =
±∞ are in one-to-one correspondence with the elements wi of the duality algebra Weyl group
Weyl(U). The Weyl group, which for suitable choice of N is a subgroup of the symmetric
group SN admits a natural ordering in terms of the minimal number ℓT of reflections with
respect to simple roots αi necessary to reproduce any considered element w. The number
ℓT (w), named the height of w ∈ Weyl(U), is the same as the number of transpositions of the
corresponding permutation when Weyl(U) is embedded in the symmetric group. Time flows
goes always in the direction of increasing ℓT which, therefore, plays the role of entropy. >>
3
The paint group and the Tits Satake projection
In [9] first and then more systematically in [18] it was observed that the Tits-Satake theory
of non-compact cosets, which is a classical chapter of modern differential geometry, provides
a natural frame to discuss the structure of the U/H cosets appearing in supergravity with
particular reference to their role in billiard dynamics. In [9] a new concept was introduced,
that of paint group, which plays a fundamental role in classifying the relevant U/H manifolds and grouping them into universality classes with respect to the Tits Satake projection.
The systematics of these universality classes was developed in [18].
In the present paper we will clarify and illustrate by means of explicit examples the
meaning of these universality classes showing that the essential features of billiard dynamics
are just a property of the class, independently from the choice of the representative, namely
independently from the choice of the paint group. In particular the Weyl group and the
asymptotic states are common to the whole class. On the other hand the notion of the paint
group enters in the precise definition of the parameter space for the general integral. Let us
therefore recall the essential notions relevant to our subsequent discussion.
4
3.1
The solvable algebra
Following the discussion of [9] let us recall that in the case the scalar manifold of supergravity
is a non maximally non-compact manifold U/H the Lie algebra U of the numerator group is
some appropriate real form
U = UR
(3.1)
of a complex Lie algebra U(C) of rank r = rank(U). The Lie algebra H of the denominator
H is the maximal compact subalgebra H ⊂ UR , which has typically rank rc < r. Denoting,
as usual, by K the orthogonal complement of H in UR
UR = H ⊕ K
(3.2)
and defining as non-compact rank, or rank of the coset U/H, the dimension of the noncompact
Cartan subalgebra
\
rnc = rank (U/H) ≡ dim Hnc ; Hnc ≡ CSAU(C)
K,
(3.3)
we obtain that rnc < r.
The manifold UR /H is always metrically equivalent to a solvable group manifold MSolv ≡
exp[Solv(UR /H)] although the form of the solvable Lie algebra Solv(UR /H), whose structure
constants define the Nomizu connection, is more complicated when rnc < r than in the
maximally split case rnc = r. For the details on the construction of the solvable Lie algebra
we refer to the literature [11, 17]. The important thing in our present context is that it
exists. Furthermore, using a general theorem proven in such textbooks like [20] we know
that every linear representation of a solvable Lie algebra can be written in a basis where
all of its elements are given by upper triangular matrices. Hence for any of the U/H cosets
of supergravity we can choose a coset representative L(φ) given by the matrix exponential
of an upper triangular matrix. This is the so named solvable parametrization of the coset
manifold which plays a fundamental role in our subsequent discussion of the general integral.
3.2
The paint group and its Lie algebra
Naming M = U/H the considered coset manifold and SolvM ⊂ U the corresponding solvable
algebra, there exists a compact algebra Gpaint which acts as an algebra of outer automorphisms (i.e. outer derivatives) of the solvable algebra SolvM
Aut [SolvM ]
=
{X ∈ U | ∀ Ψ ∈ SolvM : [X , Ψ] ∈ SolvM } .
(3.4)
By its own definition the algebra Aut [SolvM ] contains SolvM as an ideal. Hence we can
define the algebra of external automorphisms as the quotient
AutExt [SolvM ] ≡
Aut [SolvM ]
,
SolvM
(3.5)
and we identify Gpaint as the maximal compact subalgebra of AutExt [SolvM ]. Actually we
immediately see that
Gpaint = AutExt [SolvM ] .
(3.6)
5
Indeed, as a consequence of its own definition the algebra AutExt [SolvM ] is composed of
isometries which belong to the stabilizer subalgebra H of any point of the manifold, since
SolvM acts transitively. In virtue of the Riemannian structure of M we have H ⊂ so(n)
where n = dim (SolvM ) and hence also AutExt [SolvM ] ⊂ so(n) is a compact Lie algebra.
The paint group is now defined by exponentiation of the paint algebra
Gpaint ≡ exp [ Gpaint ] .
(3.7)
The notion of maximally split algebras can be formulated in terms of the paint algebra by
stating that
U = maximally split ⇔ AutExt SolvU/H = ∅ .
(3.8)
Namely U is maximally split if and only if the paint group is just the trivial identity group.
3.3
The subpaint group and the Tits Satake subalgebra
Making a long story short, once the paint algebra has been defined, the solvable Lie algebra
falls into a linear representation of Gpaint and one can define its little group, generated by
the stability subalgebra of a generic element X ∈ SolvM . In other words, viewed as a
representation of Gpaint , under the subalgebra
Gsubpaint ⊂ Gpaint
(3.9)
the solvable Lie algebra decomposes into a singlet subalgebra SolvT S plus a bunch of non
trivial irreducible representations of Gsubpaint . We name such a Lie subalgebra the subpaint
algebra. Then the Tits Satake subalgebra of the original algebra U is defined as the set of
all elements which are invariant with respect to Gsubpaint :
X ∈ UT S ⊂ U ⇔ ∀Ψ ∈ Gsubpaint
:
[X , Ψ] = 0 .
(3.10)
By construction the Tits Satake subalgebra UTS is maximally split and the Tits Satake
projection is defined as the following mapping of coset manifolds:
ΠTS
:
UTS
U
→
.
H
HTS
(3.11)
In terms of root systems the Tits Satake projection has a natural and simple intepretation.
The root system ∆U of the original algebra is composed by a set of vectors in r-dimension
where r is the rank of U. This system of vectors can be projected onto the rnc -dimensional
subspace dual to the non-compact Cartan subalgebra. Somewhat surprisingly, with just one
exception, the projected set of vectors is a new root system in rank rnc , which we name ∆TS .
Indeed the corresponding Lie algebra is precisely the Tits Satake subalgebra UTS ⊂ U of
the original algebra.
4
Triangular embedding in SL(N, R)/SO(N) and integrability
As a consequence of all the algebraic structures we have described we can conclude with the
following statement.
6
Statement 4.1 << Let N be the real dimension of the fundamental representation of U.
Then there is a canonical embedding
U ֒→ sl(N, R) ,
U ⊃ H ֒→ so(N) ⊂ sl(N, R) .
(4.1)
This embedding is determined by the choice of the basis where Solv (U/H) is made by upper
triangular matrices. In the same basis the elements of K are symmetric matrices while those
of H are antisymmetric ones. >>
The embedding (4.1) defines also a canonical embedding of the relevant Weyl group W eyl(U)
of U into that of sl(N, R) namely into the symmetric group SN .
The existence of (4.1) is the key-point in order to extend the integration algorithm of
supergravity billiards presented in [19] from the case of maximally-split cosets to the generic
case. Indeed that algorithm is defined for SL(N, R)/SO(N) and it has the property that if
initial data are defined in a submanifold U/H where U ⊂ SL(N, R) and H ⊂ SO(N), then
the entire time flow occurs in the same submanifold. Hence the embedding (4.1) suffices to
define explicit integration formulae for all supergravity billiards.
Let us review the steps of the procedure.
1. First one defines a coset representative for U/H in the solvable parametrization as
follows:
I=1
Y
L (φ) =
exp [ϕI E αI ] exp hi Hi
(4.2)
I=m
where the roots pertaining to the solvable Lie algebra are ordered in ascending order
of height (αI ≤ αJ if I < J), Hi denote the non compact Cartan generators and
the product of matrix exponentials appearing in (4.2) goes from the highest on the
left, to lowest root on the right. In this way the parameters {φ} ≡ {ϕI , hi } have a
precise and uniquely defined correspondence with the fields of supergravity by means
of dimensional oxidation [7, 8].
2. Restricting all the fields φ of supergravity to pure time dependence φ = φ(t), the coset
representative becomes also a function of time L (φ(t)) = L(t) and we define the Lax
operator L(t) and the connection W (t) as follows:
X
−1 d
L(t) =
Tr L
L Ki Ki ,
dt
i
X
−1 d
L Hℓ Hℓ
(4.3)
W (t) =
Tr L
dt
ℓ
where Ki and Hℓ denote an orthonormal basis of generators for K and H, respectively.
3. With these definitions the field equations of supergravity, which are just the geodesic
equations for the manifold U/H in the solvable parametrization, reduce to the single
matrix valued Lax equation [19]
d
L = [W , L] .
dt
7
(4.4)
4. If we are able to write the general integral of the Lax equation, depending on p =
dim(U/H) integration constants, then comparison of the definition of the Lax operator
(4.3,4.2) with its explicit form in the integration reduces the differential equations of
supergravity to quadratures
d
φ(t) = F (t)
dt
4.1
=
known function of time.
(4.5)
The integration algorithm for the Lax Equation
Let us assume that we have explicitly constructed the embedding (4.1). In this case, in the
decomposition
U =K⊕H
(4.6)
of the relevant Lie algebra U, the matrices representing the elements of K are all symmetric
while those representing the elements of H are all antisymmetric as we have already pointed
out. Furthermore the matrices representing the solvable Lie algebra Solv(U/H) are all upper
triangular. These are the necessary and sufficient conditions to apply to the relevant Lax
equation (4.4) the integration algorithm originally described in [21] and reviewed in [19]. The
key point is that the connection W (t) appearing in eq.(4.4) is related to the Lax operator
by means of an algebraic projection operator as follows:
W = Π(L) := L>0 − L<0 ,
(4.7)
L>0 (<0) denoting the strictly upper (lower) triangular part of the N × N matrix L. The
relation (4.7) is nothing else but the statement that the coset representative L(φ) from which
the Lax operator is extracted is taken in the solvable parametrization.
This established, we can proceed to apply the integration algorithm. Actually this is
nothing else but an instance of the inverse scattering method. Indeed equation (4.4) represents the compatibility condition for the following linear system exhibiting the iso-spectral
property of L:
LΨ = ΨΛ,
d
Ψ = PΨ
dt
(4.8)
where Ψ(t) is the eigenmatrix, namely the matrix whose i-th row is the eigenvector ϕ(t, λi )
corresponding to the eigenvalue λi of the Lax operator L(t) at time t and Λ is the diagonal
matrix of eigenvalues, which are constant throughout the whole time flow
Ψ = [ϕ(λ1 ), . . . , ϕ(λn )] ≡ [ϕi (λj )]1≤i,j≤n ,
Ψ−1 = ψ(λ1 ), . . . , ψ(λn )]T ≡ [ψj (λi ) 1≤i,j≤n ,
Λ = diag (λ1 , . . . , λn ) .
(4.9)
The solution of (4.8) for the Lax operator is given by the following explicit form of the matrix
elements:
[L(t)]ij =
n
X
λk ϕi (λk , t)ψj (λk , t) .
k=1
8
(4.10)
The eigenvectors of the Lax operator at each instant of time, which define the eigenmatrix
Ψ(t), and the columns of its inverse Ψ−1 (t), are expressed in closed form in terms of the
initial data at some conventional instant of time, say at t = 0.
Explicitly we have
c11 . . . c1,i−1 ϕ01 (λj )
. .
e−λj t
..
..
,
..
..
p
ϕi (λj , t) =
Det
.
.
Di (t)Di−1 (t)
0
ci1 . . . ci,i−1 ϕi (λj )
c11
. . . c1,j
..
.
..
−λi t
.
.
e
.
.
(4.11)
Det
ψj (λi , t) = p
cj−1,1 . . . cj−1,j
Dj (t)Dj−1 (t)
ψ10 (λi ) . . . ψj0 (λi )
where the time dependent matrix cij (t) is defined below
cij (t) =
N
X
e−2λk t ϕ0i (λk )ψj0 (λk )
(4.12)
k=1
and
ϕ0i (λk ) := ϕi (λk , 0) ,
ψi0 (λk ) := ψi (λk , 0)
(4.13)
are the eigenvectors and their adjoints calculated at t = 0. These constant vectors as well
as eigenvalues λk constitute the initial data of the problem and provide the integration
constants. Finally Dk (t) denotes the determinant of the k × k matrix with entries cij (t)
Dk (t) = Det cij (t)
.
(4.14)
1≤i,j≤k
Note that cij (0) = δij and Dk (0) = 1.
5
Properties of the general integral and the parameter
space
The algorithm we have described in the previous section realizes a map
IK
:
L0 7→ L (t, L0 )
(5.1)
which, starting from the initial data, i.e. the Lax operator L(0) = L0 ∈ K at some
conventional time t = 0, produces a flow, namely a map of the infinite time line into the
subspace K ⊂ U
7→ K .
(5.2)
L (t, L0 ) :
R
|{z}
−∞ ≤ t ≤+∞
It is of the outmost interest to enumerate the properties of the maps (5.1,5.2). A first set of
four fundamental properties are listed below:
9
1. The flow L (t, L0 ) is iso-spectral. This means the following. The Lax operator is a
symmetric matrix and therefore can be diagonalized at every instant of time. Calling
λ1 . . . λN the set of its N eigenvalues, we have that this set is time–independent, namely
the numerical values of the eigenvalues remain the same throughout the entire motion.
2. If the Lax operator L(t) is diagonal at any finite time t 6= ±∞, then it is actually
constant L(t) = L0
3. The asymptotic limits of the Lax operator for t 7→ ±∞ are diagonal matrices L± ∞ .
4. If L0 ∈ KU belongs to the symmetric part of a proper Lie subalgebra U ⊂ sl (N, R),
then the entire motion remains in that subalgebra, namely ∀ t , L(t) ∈ KU .
Relying on this first set of properties we can refine our formulation of the initial conditions
and of the asymptotic limits in terms of the generalized Weyl group and of its Tits Satake
projection. This leads to state further properties of the map (5.1) which are even more
striking.
5.1
Discussion of the generalized Weyl group
Diagonal matrices are just elements of the non-compact Cartan subalgebra C ⊂ K ⊂ U.
The Lax operator at t = 0 can be diagonalized by means of an orthogonal matrix O ∈ SO(N)
which actually lies in the subgroup H ⊂ SO(N). Hence by writing
L0 = OT C0 O
(5.3)
initial data can be given as a pair
C0 ∈ CSA
\
K ;
O ∈ H.
(5.4)
Let us now introduce the notion of generalized Weyl group W(U). To understand its definition let us review the definition of the standard Weyl group. This latter is an intrinsic
attribute of a complex Lie algebra. For a complex Lie algebra UC , the Weyl group Weyl(UC )
is the finite group generated by the reflections σα with respect to all the roots α. Actually as
generators of Weyl(U) it suffices to consider the reflections with respect to the simple roots
σαi . It turns out that if we consider the maximally split real section Usplit of the complex
Lie algebra UC then the Weyl group Weyl(UC ) is realized as a subgroup of the maximal
compact subgroup Hsplit ⊂ Usplit. This isomorphism is realized as follows. Consider the
integer valued elements of H defined below
hπ
i
H ∋ γα ≡ exp
E α − E −α
, α > 0
(5.5)
2
and take them as generators. These generators produce a finite subgroup W(U) which
we name generalized Weyl group. It contains a normal subgroup N(U) ⊂ W(U) whose
adjoint action on any Cartan Lie algebra element is just the identity. The factor group
W(U)/N(U) ∼ Weyl(UC ) is isomorphic to the abstract Weyl group of the complex Lie
algebra.
Imitating such a construction also in the non maximally split cases we can introduce the
following
10
Definition 5.1 << Let U be a not necessarily maximally split real section of the complex
Lie algebra UC and H ⊂ U its maximal compact subalgebra. Let α[K] be the set of positive
roots which are not in the kernel of the Tits Satake projection and which therefore participate
in the construction of the solvable Lie algebra of U/R. The generalized Weyl group W(U) is
the finite subgroup of H generated by the following generators:
hπ
i
α[K]
−α[K]
E
−E
, α[K] > 0
(5.6)
H ∋ γα[K] ≡ exp
2
whose number is dim(U/H) − rank(U/H). >>
As we already noted, the generalized Weyl group is typically bigger and has more elements
than the ordinary Weyl group.
By construction the adjoint action of the generalized Weyl group maps the non-compact
Cartan subalgebra into itself
\
\
∀ Ow ∈ W(U) and ∀ C ∈ CSA
K : OwT C Ow ∈ CSA
K.
(5.7)
This can be verified by means of the same calculation which shows that the ordinary Weyl
group, as defined in eq.(5.5), maps the Cartan subalgebra into itself for the maximally split
case.
This observation shows that giving the initial data as we did in eq.(5.4) actually corresponds to an over-counting. Indeed the generalized Weyl group should be modded out since
it amounts to a redefinition of the Cartan subalgebra data C0 . So we are led to guess that
for each choice of the eigenvalues of the Lax operator, namely at fixed C0 , the parameter
space of the Lax equation is P = H/W(U). This however is not yet the complete truth.
Indeed there is also a continuous group, whose adjoint action on the non-compact Cartan
subalgebra is the identity map. This is the paint group Gpaint . Hence the true parameter
space of the Lax equation is the orbifold with respect to the generalized Weyl group, not of
a group, rather of a compact coset manifold. Indeed we can write
P =
H
Gpaint
/ W(U) .
(5.8)
Furthermore we can consider a normal subgroup NW (U) ⊂ W(U) of the generalized Weyl
group defined by the following condition:
\
γ ∈ NW (U) ⊂ W(U) iff ∀C0 ∈ CSA
K
γ T C0 γ = C0
(5.9)
and we can state the proposition which is true for all non-compact cosets U/H:
Statement 5.1 << The factor group of the generalized Weyl group with respect to its normal subgroup stabilizing all elements of the non-compact Cartan subalgebra is just isomorphic
to the ordinary Weyl group of the Tits Satake subalgebra:
W(U)
≃ Weyl (UT S ) .
NW (U)
>>
11
(5.10)
This shows that the only relevant Weyl group is just the Weyl group of the Tits-Satake
subalgebra UTS .
In view of the iso-spectral property and of the asymptotic property of the Lax operator which becomes diagonal at t = ±∞ we conclude that, once C0 is chosen, the available
end-points of the flows at the remote past and at the remote future are in one-to-one correspondence with the elements of the Weyl group Weyl(UT S ). Indeed diagonal matrix means
an element of the non-compact Cartan subalgebra and, since the eigenvalues are numerically
fixed by the original choice of C0 , the only thing which can happen is a permutation. The
available permutations are on the other hand dictated by the embedding of the Weyl group
into the symmetric group:
Weyl (UTS ) ֒→ SN ≃ Weyl (AN −1 )
(5.11)
which is induced by the embedding (4.1) of the Lie algebra U into sl(N, R). The latter, as
we already stressed, follows by the choice of the upper triangular basis for the solvable Lie
algebra in the fundamental representation of U.
5.2
The arrow of time, trapped and critical surfaces
In view of the above discussion we conclude that the integration algorithm (5.1) realizes a
map of the following type:
TK
:
H/Gpaint
=⇒ Weyl (UTS )− ⊗ Weyl (UTS )+
W(U)
(5.12)
where ∓ refer to the choice of a Weyl group element at ∓∞ realized by the asymptotic limits
of the Lax operator.
It is of the outmost interest to explore the general properties of the map TK .
→
Let −
w I , (I = 1, . . . , N) be the weights of U in its fundamental N − dimensional representation RN and let
→
−
h = h1 , .., hrnc , 0, 0, . . . , 0
(5.13)
| {z } | {z }
−
→
r−rnc
h TS
be the r-vector of parameters identifying the C0 element in the non compact Cartan subalgebra
rnc
X
\
hi Hi .
(5.14)
CSA
K ∋ C0 =
i=1
The r − rnc zeros in eq.(5.13) correspond to the statement that all components of C0 in
→
−
the compact directions of the Cartan subalgebra vanish. The sub-vector h T S is the only
non vanishing one and it is the same as we would have in the Tits Satake projected case.
With these notations the N eigenvalues of the Lax operator λ1 , λ2 , . . . , λN are represented as
follows:
→
−
→
→
λI = −
w I (C0 ) = −
wI · h .
(5.15)
12
Consider now the branching of the fundamental representation of U with respect to the Tits
Satake subalgebra times the Gsubpaint algebra:
RN
UTS × Gsubpaint
=⇒
(RNTS , 1) ⊕ (p, q) .
(5.16)
By definition RNTS is the fundamental representation of the Tits Satake subalgebra of dimension NTS < N which is a singlet under the subpaint algebra Gsubpaint , while the remaining
representation (p, q) is non trivial both with respect to the Tits-Satake and with respect
to the subpaint Lie algebra. Obviously we have p × q = N − NTS . Correspondingly the
i
eigenvalues of the Lax operator organize in the way we are going to describe. Let w−|T
S < 0
0
(i = 1, . . . , m) be the negative weights of the representation RNTS , let wT S = 0 be the
i
null-weight of the same representation (if it exists) and let w+|T
S > 0 (i = 1, . . . , m) be the
positive weights. If there is a null-weight we have NTS = 2m + 1, otherwise NTS = 2m. A
conventional order for the N eigenvalues is given by the following vector :
→
−
→
λ1
= −
w 1−|T S · h T S
→
−
2
2
→
−
λ
=
w
·
h
T
S
−|T
S
...
...
...
→
−
m
m
→
−
λ
=
w
·
h
T
S
−|T
S
m+1
λ
=
0
λ
=
0
→
−
m+2
λ [1] =
(5.17)
.
.
.
.
.
.
.
.
.
m+pq+1
=
0
λ
→
−
1
→
λm+pq+2 = −
w
·
h
TS
+|T S
m+pq+3
→
−
→
λ
= −
w 2+|T S · h T S
...
...
...
→
−
m
N
→
−
λ
= w +|T S · h T S
→
−
where the weights are organized from the lowest to the highest. The vector λ [1] corresponds
to the diagonal entries of the matrix C0 defined in eq.(5.4). All the other possible orders
of the same eigenvalues are obtained from the action of the Weyl group Weyl (UTS ) on the
weights of RNTS . By construction, such an action permutes the positions of the non-vanishing
→
−
eigenvalues while all the zeros stay at their place. In this way starting from λ [1] we obtain
→
−
n = |Weyl (UTS ) | such vectors λ [x] in one-to-one correspondence with the elements of the
Tits Satake Weyl group. Schematically, naming Ωx the elements of Weyl (UTS ), we obtain
→
−
→
−
λ [x] = Ωx λ [1]
(5.18)
with the understanding that Ω1 is the identity element of the Weyl group. Each element
Ωx is represented by a permutation of the non vanishing eigenvalues, hence by an element
→
−
→
−
P (Ωx ) ∈ SNTS ⊂ SN . Among the vectors λ [x] there will be one λ [min] where the eigenvalues
are organized in decreasing order
N −1
λ1[min] ≥ λ2[min] ≥ . . . ≥ λ[min]
≥ λN
[min]
13
(5.19)
→
−
and there will be another one λ [max] where the eigenvalues are instead organized in increasing
order
N −1
λ1[max] ≤ λ2[max] ≤ . . . ≤ λ[max]
≤ λN
(5.20)
[max] .
Name Ωmin/max the corresponding Weyl elements. It follows that equation (5.18) can be
rewritten as
→
−
→
−
λ [x] = Ωx Ω−1
(5.21)
min λ [min] .
The symmetric group admits a partial ordering of its elements given by the number ℓT (P )
of elementary transpositions necessary to obtain a given permutation P starting from the
fundamental one P0 . The embedding of the Weyl group into the symmetric group allows
to transfer this partial ordering to the Weyl group as well. We define the length of a Weyl
element as follows. Taking the permutation of λ[min] as fundamental we set
∀ Ωx ∈ Weyl (UT S ) : ℓT (Ωx ) ≡ ℓT P Ωx Ω−1
.
(5.22)
min
With this definition the Weyl element Ωmin has length ℓT = 0 while the Weyl element
Ωmax has the maximal length ℓT = 12 (NTS − 1) NTS and an element Ωx is higher than an
element Ωy if ℓT (Ωx ) > ℓT (Ωy ). We can observe that the partial ordering induced by the
immersion in the symmetric group is, up to some rearrangement, the intrinsic ordering of the
Weyl group provided by counting the minimal number of reflections with respect to simple
roots necessary to construct the considered element. In our context formalising the precise
correspondence between the two ordering procedures is not necessary since the relevant one
is that with respect to permutations and this is well and uniquely defined.
Having introduced the above ordering of Weyl elements we can now state the main and
most significant property of the map (5.12) and of the Toda flows realized by the integration
algorithm (5.1).
Principle 5.1 << In any flow the arrow of time is so directed that the state at t = −∞
corresponds to the lowest accessible Weyl element and the state at t = +∞ corresponds to
the highest accessible one. >>
To make the principle 5.1 precise we need to define the notion of accessible Weyl elements.
This latter relies on another remarkable and striking property of the Toda flows (5.1) which
we have numerically verified in a large variety of cases never finding any counterexample.
Property 5.1 << At any instant of time the Lax operator L(t) can be diagonalized by a
time dependent orthogonal matrix O(t) ∈ H, writing L(t) = OT (t) C0 O(t). Consider now
the N2 − 1 minors of O(t) obtained by intersecting the first k columns with any set of k-rows,
for k = 1, . . . , N − 1. If any of these minors vanishes at any finite time t 6= ±∞ then it is
constant and vanishes at all times. >>
The remarkable conservation law stated in property 5.1 which has the mathematical status
of a conjecture implies that there are generic initial data, namely points of the parameter
space P defined in eq.(5.8) and N 2 − 1 trapped hypersurfaces Σi ⊂ P defined by the
vanishing of one of the minors. These trapped surfaces can also be intersected creating
trapped sub-varieties of equal or lower dimensions. If the initial data are generic, then
principle 5.1 implies that the flow will necessarily be from Ωmin to Ωmax . On the other hand
14
if we are on a trapped surface we have to see which elements of the generalized Weyl group
W(U) belong to that surface. As we know from (5.10) each element of W(U) is equivalent
to an element of Weyl(UTS ) modulo an element in the normal subgroup. Hence we can
introduce the following definition:
Definition 5.2 << A Weyl group element γ ∈ Weyl (UT S ) is accessible to a trapped surface
Σ if there exists a representative µ ∈ W(U) of its equivalence class in the generalized Weyl
group which belongs to Σ. >>
The set of Weyl elements AΣ accessible to a trapped surface Σ inherits an ordering from the
general ordering of Weyl (UT S ), namely we can write:
AΣ = {Ωx1 , Ωx2 , . . . , Ωxσ }
(5.23)
where σ is the cardinality of the set σ = card AΣ and Ωxi ≤ Ωxj if i < j. Then the flow is
always from the lowest Weyl element of AΣ at t = −∞ (i.e., from Ωx1 ) to the highest one
at t = ∞ (i.e., to Ωxσ ) as stated in principle 5.1.
If we consider lower dimensional trapped surfaces obtained by intersection, then the set
of Weyl elements accessible to the intersection is simply given by the intersection of the
accessible sets:
\
AΣ T Π = AΣ
AΠ
(5.24)
and the flow is from the lowest element of AΣ T Π to its highest one.
We can now introduce a further
Definition 5.3 << A trapped surface Σ is named critical if the set of Weyl elements AΣ
accessible to the surface is a proper subset of the Weyl group, in other words if
card AΣ < |Weyl (UT S ) | .
(5.25)
>>
Note that the property of criticality does not necessarily imply a variance of asymptotics
from the generic case Ωmin → Ωmax . Indeed, although the cardinality of AΣ is lower than
the order of the Weyl group so that some elements are missing, yet it suffices that both
Ωmin ∈ AΣ and Ωmax ∈ AΣ to guarantee that the infinite past and infinite future states of
the Universe will be the same as in the generic case. This observation motivates the further
definition:
Definition 5.4 << A trapped surface Σ is named super-critical if it is critical and moreover either the maximal or the minimal Weyl elements are missing from AΣ :
Ωmin ∈
/ AΣ
and/or Ωmax ∈
/ AΣ .
(5.26)
>>
This discussion shows that the truly relevant concept is that of trapped surface which streams
from the remarkable conservation law given by property 5.1, criticality or super-criticality
being, from the mathematical point of view, just accessory features although of the highest
15
physical relevance. If we just focus our attention on the initial and final states the intermediate concept of critical surface seems to be unmotivated. The reason why it is useful is that
critical surfaces as defined in 5.3 can be computed in an intrinsic way taking a dual point
of view. Rather than computing accessible Weyl elements one can define forbidden ones by
using the embedding of the Weyl group into the symmetric group mentioned in eq.(5.11).
It follows from this that to each element Ωx ∈ Weyl (UT S ) we can associate a permutation
Px ∈ SN , where N is the dimension of the fundamental representation of U and hence of the
orthogonal matrix O we are discussing. This fact allows to associate to Ωx a set of N − 1
minors defined as follows:
(2)
(N −1)
Weyl (UT S ) ∋ Ωx → min(1)
[O]
(5.27)
x [O] , minx [O] , . . . , minx
min(k)
x [O] = Det (O [(Px (1), . . . , Px (k)) , (1, . . . , k)])
(5.28)
where M [(a1 , . . . , ak ) , (1, . . . , k)] denotes the minor of the matrix M obtained by intersecting
the k-rows a1 , . . . , ak with the first k-columns. Using a dual view-point it was shown in [22]
that in any flow, in order for a permutation P of the eigenvalues to be a candidate for
asymptotics (i.e. to be available), its associated minors should all be non zero. Hence
relying on the embedding of the Weyl group into the symmetric group we conclude that
if any minor min(k)
x [O] vanishes then Ωx is excluded from the set AΣx|k of Weyl elements
accessible to the surface Σx|k defined by the vanishing of the minor min(k)
x [O]. We can write:
Σx|k ≡
O ∈ H \ min(k)
x [O] = 0
⇒
Ωx ∈
/ AΣx|k .
(5.29)
The same minor is produced by more than one element and hence identifying all the Ωx for
(k)
which min(k)
x [O] = min[0] we immediately calculate the set of Weyl elements excluded from
AΣ0 and by complement we also know the set AΣ0 .
If all the possible minors considered in property (5.1) could be produced by Weyl elements, then what we have just described would be a quick and efficient way to obtain all
trapped surfaces. In that case all trapped surfaces would also be critical. The fact is that
not all minors can be obtained from Weyl elements and this implies that there are trapped
surfaces which are not critical. The reason of this difference is evident from our discussion. It
is due to the fact that the Weyl group is in general only a proper subgroup of the symmetric
group SN . Therefore there are permutations and therefore minors which do not correspond
to any Weyl element and for that reason they define non-critical trapped surfaces. In the
case of SL(N, R)/SO(N) flows the Weyl group is just the full SN and all trapped surfaces are
critical. In conclusion trapped but not critical surfaces are critical surfaces of the embedding
SL(N) where the missing Weyl elements are in the kernel of the projection SN 7→ Weyl(U).
This concludes the general presentation of our results. By means of some case studies
the next part illustrates the principles formulated in this part.
16
Part II
Examples illustrating the principles
6
Choice of the examples
In this part we make three case studies:
1 We survey the flows on the simplest example SL(3, R)/SO(3) of maximally split coset
manifolds in order to demonstrate the relation between the Weyl group and the arrow
of time by calculating explicitly all the critical surfaces which are two-dimensional and
can be visualized. The parameter space is a three dimensional cube with some vertices
identified and can also be visualized.
2 Next we make a detailed study of the flows on Sp(4, R)/U(2). This manifold is the Tits
Satake projection of an entire universality class of manifolds, SO(2, 2 + 2s)/SO(2) ×
SO(2 + 2s) which, on the other hand, is for r = 2 of the type SO(r, r + 2s)/SO(r) ×
SO(r + 2s). For the latter we make the general construction of the triangular basis of
the solvable algebra illustrating the embedding:
so(r, r + 2s) ֒→ sl(2r + 2s)
(6.1)
which is crucial in order to establish the integration algorithm. In the case of sp(4) ∼
so(2, 3), which is maximally split of rank two, the parameter space is a 4-dimensional
hypercube also with vertices identified.
3 Finally we study the case of SO(2, 4) in comparison with that of SO(2, 3) in order to
illustrate the properties of the Tits Satake projection and the meaning of Tits Satake
universality classes.
7
The simplest maximally split case: SL(3, R)/SO(3)
In order to illustrate the general ideas discussed in the previous part and as a preparation
to the study of more general cases, we begin with a detailed analysis of the time flows in the
simplest instance of maximally split coset manifolds, namely for
M5 =
SL(3, R)
.
SO(3)
(7.1)
The sl(3, R) Lie algebra is the maximally split real section of the A2 Lie algebra, encoded in
the Dynkin diagram of fig.1. The root system has rank two and it is composed by the six
17
Figure 1: The Dynkin diagram of the A2 Lie algebra.
A2
✐
✐
α1
α2
vectors displayed below and pictured in fig.2:
α1
α2
α1 + α2
∆A2 =
− α1
− α2
−α − α
1
2
=
=
=
=
=
=
The simple roots are α1 and α2 .
Α2
√
2, 0 ,
q
1
√
− 2 , 32 ,
q
3
√1 ,
,
2
2
√
− 2, 0 ,
q
3
√1 , −
,
2
q
2
− √12 , − 32 .
Α1 + Α2
Α1
Figure 2: The A2 root system.
18
(7.2)
A complete set of generators for the Lie algebra is provided by the following 3×3 matrices:
− √16 0
0
√1
0
0
2
; H2 = 0
,
√1
−
0
√1
H1
=
0
−
0
6 q
2
2
0
0
0 0
0
3
0 0 0
0 1 0
α2
0 0 1 ,
0 0 0
;
E
=
E α1
=
0 0 0
0 0 0
0 0 1
α1 +α2
E
=
0 0 0 ,
0 0 0
T
E −α1 = (E α1 )T ; E −α2 = (E α2 )T ; E −α1 −α2 = E α1 +α2
(7.3)
where H1,2 are the two Cartan generators and E α are the step operators associated to the
corresponding roots. The solvable Lie algebra generating the coset (7.1) is composed by the
following five operators:
SL(3, R)
Solv
= span H1 , H2 , E α1 , E α2 , E α1 +α2
(7.4)
SO(3)
and it is clearly represented by upper triangular matrices. The orthogonal decomposition
G = H⊕K
(7.5)
of the Lie algebra with respect to its maximal compact subalgebra:
so(3) ≡ H ⊂ G ≡ sl(3, R)
(7.6)
is performed by defining the following generators:
K = span {K1 , . . . , K5 }
n
≡
H1 , H2 , √12 E α1 + E −α1 , √12
H = span {J1 , . . . , J3 }
n
√1
≡
E α1 − E −α1 ,
2
√1
2
E α2 + E −α2 ,
E α2 − E −α2 ,
√1
2
√1
2
E α1 + α2 + E − α1 −α2
o
,
(7.7)
E α1 + α2 − E − α1 −α2
o
.
(7.8)
By definition the Lax operator L(t) is a symmetric 3 × 3 matrix which can be decomposed
along the generators of the subspace K:
L(t) =
5
X
i=1
19
k i (t) Ki
(7.9)
and once the functions k i (t) have been determined, by means of an oxidation procedure which
was fully described in [7], the fields of supergravity can be extracted by simple quadratures.
As we explained in [19] and we recalled in the introduction, the initial data for the
integration of the Lax equation are provided by the choice of an element of the Cartan
subalgebra, namely by a diagonal matrix of the form:
λ1 0
0
(7.10)
CSA ∋ C ({λ1 , λ2 }) =
0
0 λ2
0 0 −λ1 − λ2
and by a finite element O ∈ SO(3) of the compact subgroup which together with C determines the value of the Lax operator at time t = 0:
L0 = OT C ({λ1 , λ2 }) O .
(7.11)
We stressed that the choice of the group element O is actually defined modulo multiplication
on the left by any element w ∈ W ⊂ H ≃ exp H of the discrete Weyl subgroup. By
definition the Weyl group maps the Cartan subalgebra into itself, so that we have:
∀ w ∈ W ⊂ SO(3)
:
w T C ({λ1 , λ2 }) w = C (w {λ1 , λ2 }) ∈ CSA
(7.12)
where C (w {λ1 , λ2 }) denotes the diagonal matrix of type (7.10) with eigenvalues λ′1 , λ′2 , −λ′1 −
λ′2 obtained from the action of the Weyl group on the original ones. So the actual moduli
space of the Lax equation is not H but the quotient H/W.
In the case of the Lie algebras An the Weyl group
Pnis the symmetric group Sn+1 and its
action on the eigenvalues λ1 , λ2 , . . . , λn , λn+1 = − i=1 λi is just that of permutations on
these n + 1-eigenvalues. For A2 we have S3 whose order is six. The six group elements can
be enumerated in the following way:
1 0 0
; (λ1 , λ2 , λ3 ) 7→ (λ1 , λ2 , λ3 ) ,
w1 =
0
1
0
0 0 1
0 1 0
(7.13)
w2 = 1 0 0
; (λ1 , λ2 , λ3 ) 7→ (λ2 , λ1 , λ3 ) ,
0 0 1
0 0 1
; (λ1 , λ2 , λ3 ) 7→ (λ3 , λ2 , λ1 ) ,
w3 =
0
1
0
1 0 0
20
ℓT Weyl group
of SL(3, R)
0
w2
1
w1
1
w5
2
w3
2
w4
3
w6
Table 1: Partial ordering of the Weyl group of SL(3, R).
1 0 0
w4 =
0
0
0
w5 = 1
0
0
w6 =
0
1
0 1
; (λ1 , λ2 , λ3 ) 7→ (λ1 , λ3 , λ2 ) ,
1 0
0 1
0 0
; (λ1 , λ2 , λ3 ) 7→ (λ2 , λ3 , λ1 ) ,
1 0
1 0
0 1
; (λ1 , λ2 , λ3 ) 7→ (λ3 , λ1 , λ2 ) .
0 0
(7.14)
Let us now choose as eigenvalues λ1 , λ2 , λ3 defined at the central time t = 0 the conventional
set
λ1 = 1 ; λ2 = 2 ; λ3 = −3 .
(7.15)
In this case the decreasing sorting to be expected at past infinity is given by: 2, 1, −3
which, according to eq.(7.14), corresponds to the Weyl element w2 . Hence we can use w2 as
the fundamental permutation and rate all the other Weyl group elements according to the
number of transpositions ℓT needed to bring their corresponding permutation to that of w2 .
In this way we obtain a partial ordering of the Weyl group where the highest element
is the unique w6 corresponding to the increasing sorting of eigenvalues −3, 1, 2. Indeed we
have the result shown in table 1 and if all the Weyl elements are accessible there is a unique
predetermined process: the state of the universe at past infinity is the Kasner era w2 , while
the state of the Universe at future infinity is the Kasner era w6 . If not all the Weyl elements
are accessible, then we can have different situations. In order to discuss them we have to
study the structure of the orbifold SO(3)/W.
To parametrize the SO(3) compact group we introduce three Euler angles θi (i = 1, 2, 3)
21
and we write
O11 O12 O13
O(θi ) ≡ exp [θ1 J1 ] exp [θ2 J2 ] exp [θ3 J3 ] =
O
O
O
21
22
23
O31 O32 O33
(7.16)
where:
O11 = cos (θ1 ) cos (θ3 ) − sin (θ1 ) sin (θ2 ) sin (θ3 )
;
O12 = cos (θ2 ) sin (θ1 )
;
O13 = cos (θ3 ) sin (θ1 ) sin (θ2 ) + cos (θ1 ) sin (θ3 )
;
O21 = − cos (θ3 ) sin (θ1 ) − cos (θ1 ) sin (θ2 ) sin (θ3 ) ;
O22 = cos (θ1 ) cos (θ2 )
;
O23 = cos (θ1 ) cos (θ3 ) sin (θ2 ) − sin (θ1 ) sin (θ3 )
;
O32 = − sin (θ2 )
;
O31 = − cos (θ2 ) sin (θ3 )
;
O33 = cos (θ2 ) cos (θ3 )
.
In this parametrization, if we introduce the notation
π
π π
Oxyz = O x , y , z
2
2
2
we obtain:
1 0
O000 = 0 1
0 0
1 0
O010 =
0 0
0 −1
0
0
O110 =
−1 0
0
−1
0
0
O011 =
−1 0
0
−1
7.1
0
0
1
0
1
0
1
0
0
1
0
0
; O100
; O001
; O101
; O111
0
1
= −1 0
0
0
0
0
=
1
0
−1 0
0
1
=
0
0
−1 0
−1 0
=
0
0
0
−1
(7.17)
(7.18)
0
0
,
1
1
0
,
0
0
−1
,
0
0
−1
.
0
(7.19)
(7.20)
Discussion of the generalized Weyl group
Let us now construct the generalized Weyl group, according to the definition 5.1. This case
is maximally split and all roots participate in the construction. Hence as generators we take
22
the three matrices
generators = {O100 , O010 , O001 }
(7.21)
as defined above in eq.(7.19). Closing the shell of products we find a group W(sl(3)) containing 24 elements organized in 6 equivalence classes with respect to a normal subgroup
N (sl(3)) ∼ Z2 × Z2 . The four elements of N (sl(3)) are the following four matrices:
1 0
0
1 0 0
0 −1 0 ,
0 1 0
;
n
=
n1 =
2
0 0
−1
0 0 1
(7.22)
−1 0
0
−1 0 0
.
; n4 = 0
n3 =
−1
0
0
1
0
0
0
1
0
0 −1
The factor group is isomorphic to the Weyl group of sl(3)
W (sl(3))
∼ Weyl (sl(3)) ≡ S3
N (sl(3))
and a representative of the six equivalence
1 0 0
N (sl(3)) ;
w1 ∼
0
1
0
0 0 1
0
0 1
w3 ∼ 0
1 0
N (sl(3)) ;
−1 0 0
0 0 1
N (sl(3)) ;
w5 ∼
1
0
0
0 1 0
classes is listed below
0 1 0
N (sl(3)) ,
w2 ∼
1
0
0
0 0 −1
1 0
0
w4 ∼ 0 0
1
N (sl(3)) ,
0 −1 0
0 1 0
N (sl(3)) .
w6 ∼
0
0
1
1 0 0
(7.23)
(7.24)
Hence modulo the normal subgroup the eight matrices listed in eq.s(7.19,7.20) can be identified with the six elements of the Weyl group in the following way:
O000 ∼ w1 ; O100 ∼ w2 ; O010 ∼ w4
O001 ∼ w3 ; O110 ∼ w5 ; O101 ∼ w5
(7.25)
O011 ∼ w6 ; O111 ∼ w4 .
Let us now consider the general form of the SO(3) matrix as given in eq.(7.16) and the
modding by the generalized Weyl group. Precisely, with our conventions this means the
following1 :
∀ γ ∈ W(sl(3)) and ∀ O ∈ SO(3) :
γO ∼ O .
(7.26)
1
Modding is done by left multiplication because, if sitting on the left of O the generalized Weyl group
element γ will act on the Cartan element C by conjugation γ T C γ (see eq.(7.11)).
23
In terms of the matrix entries Oij the operation (7.26) is quite simple, it just implies that all
orthogonal matrices which differ by an arbitrary permutation of the rows accompanied by
overall changes of signs rows by rows are to be identified. On the other hand transferring the
multiplication by γ on the theta angles is a highly non trivial and complicated operation. In
other words the map
θi → fγi (θ)
(7.27)
defined by:
O fγi (θ) = γ O (θ)
(7.28)
is quite involved and not handy. This implies that displaying a fundamental cell in θ-space
is not an easy task and does not lead to any illuminating picture. This is no serious problem,
since it is just a coordinate artifact. Furthermore precisely since we are finally interested in
equivalence classes with respect to the algebraic Weyl group, i.e. in sets of 4-matrices of the
form
N(sl(3)) O
(7.29)
then it just suffices to identify a minimal neighborhood of R3 in the open chart of the group
manifold SO(3) defined by the Euler angle parameterization (7.16) such that it contains
at least one copy of each Weyl group element wi ∈ Weyl(sl(3)). An example of such a
minimal submanifold is provided by the cube 0 ≤ θi ≤ π2 shown in fig.3 whose vertices are
just the required representatives of the Weyl group elements. In all cases we can focus our
attention on the hypercube in Euler angle spaces defined by the vertices which correspond
to the nearest copies of all the Weyl group elements. We stress that these hypercubes are
not fundamental cells for the equivalence classes H/W(U) but are just sufficient for our
purposes, in particular in order to study the flow diagram produced by links, namely flows
on one dimensional trapped surfaces.
7.2
The flow diagram and the critical surfaces for SL(3, R)
We can now explore the behaviour of Lax equation on the vertices, the edges and the interior
of the parameter space we have described in the previous section.
Vertices As we know from the general properties of the integral discussed in section 5 if
the Lax operator lies in the Cartan subalgebra at the initial point t = 0, namely it is diagonal
it will remain constant all the time from −∞ to +∞. Hence on each vertex of the cube,
which corresponds to a Weyl group element, we have constant Lax operators, corresponding
to as many Kasner epochs.
24
5
4
3
6
4
5
1
2
Figure 3: A three-dimensional cube 0 ≤ θi ≤ π2 (i = 1, 2, 3) whose eight vertices are identified
with the six Weyl group elements as shown in the picture. The 12 edges of the cube represent
one parameter submanifolds of SO(3) where just one angle varies while the other two are at
fixed values, either 0 or π/2.
Edges It is interesting to see what happens on the twelve edges of the cube. Let us display
the form of the matrix O on each of these edges.
cos (θ1 )
sin (θ1 ) 0
,
1) (000) ↔ (100) O =
−
sin
(θ
)
cos
(θ
)
0
1
1
0
0
1
1 0
0
(7.30)
2) (000) ↔ (010) O = 0 cos (θ2 )
sin (θ2 )
,
0 − sin (θ2 ) cos (θ2 )
cos (θ3 )
0 sin (θ3 )
,
3) (000) ↔ (001) O =
0
1
0
− sin (θ3 ) 0 cos (θ3 )
25
4) (100) ↔ (110)
5) (100) ↔ (101)
6) (010) ↔ (011)
7) (010) ↔ (110)
8) (001) ↔ (101)
9) (001) ↔ (011)
10) (110) ↔ (111)
11) (011) ↔ (111)
12) (101) ↔ (111)
0
cos (θ2 )
sin (θ2 )
,
O = −1 0
0
0
− sin (θ2 ) cos (θ2 )
0
1 0
,
O =
−
cos
(θ
)
0
−
sin
(θ
)
3
3
− sin (θ3 ) 0 cos (θ3 )
cos (θ3 )
0
sin (θ3 )
,
O =
−
sin
(θ
)
0
cos
(θ
)
3
3
0
−1 0
cos (θ1 )
0
sin (θ1 )
,
O =
−
sin
(θ
)
0
cos
(θ
)
1
1
0
−1 0
0
sin (θ1 ) cos (θ1 )
O = 0
cos (θ1 ) − sin (θ1 )
,
−1 0
0
0
0
1
,
O =
−
sin
(θ
)
cos
(θ
)
0
2
2
− cos (θ2 ) − sin (θ2 ) 0
− sin (θ3 ) 0
cos (θ3 )
,
O =
−
cos
(θ
)
0
−
sin
(θ
)
3
3
0
−1 0
− sin (θ1 ) 0
cos (θ1 )
O =
−
cos
(θ
)
0
−
sin
(θ
)
1
1 ,
0
−1 0
− sin (θ2 ) cos (θ2 )
0
O =
0
−1
.
0
− cos (θ2 ) − sin (θ2 ) 0
(7.31)
(7.32)
(7.33)
On each link we have one of the three one-parameter subgroups respectively generated by
J1,2,3 multiplied on the left or on the right by a Weyl group element. By means of a computer
programme we can then easily evaluate the general integral on each of these links. For
instance on the link number 1 we obtain:
L11 (t) L12 (t)
0
,
L(t) = L12 (t) L22 (t)
0
0
0
−λ1 − λ2
26
e2tλ2 λ1 cos2 (θ1 ) + e2tλ1 sin2 (θ1 ) λ2
,
L11 (t) =
e2tλ2 cos2 (θ1 ) + e2tλ1 sin2 (θ1 )
e2tλ2 λ2 cos2 (θ1 ) + e2tλ1 sin2 (θ1 ) λ1
L22 (t) =
,
e2tλ2 cos2 (θ1 ) + e2tλ1 sin2 (θ1 )
et(λ1 +λ2 ) sin (2θ1 ) (λ1 − λ2 )
.
L12 (t) =
(−e2tλ1 + e2tλ2 ) cos (2θ1 ) + e2tλ1 + e2tλ2
(7.34)
We can also calculate the asymptotic limits of the Lax operator at ±∞ for each of these
flows. As it follows from the properties of the general integral, at remotely early or at
remotely late times the Lax operator is always diagonal and its eigenvalues are organized in
one of the possible six ways corresponding to the six Weyl group elements acting on their
reference order (λ1 , λ2 , −λ1 − λ3 ). If we associate an arrow to each of these twelve links
and we take into account the identification of vertices as displayed in eq.(7.25) we obtain
the flow diagram shown in fig.4. As it is clear from the quoted picture, the flows on the
5
4
w
1 w
2
3
w
3
w
6
6
4
5
1
w
4
w
5
2
Figure 4: The oriented diagram of the SL(3, R)/SO(3) flows. The Lie algebra sl(3, R) is the
maximally split real section of the complex Lie algebra A2 . Its Weyl group is S3 and has six
elements identified by their action on the eigenvalues λ1 , λ2 , λ3 of the Lax operator. Six are
therefore the possible asymptotic states of the universe at ±∞ and each possible motion is
an oriented flow from one lower Weyl element to another higher one. The lines of the
graph on the right represent possible oriented flows along one-dimensional submanifolds of
the parameter space located on the edges of the cubedefined
by restricting the range of the
π
three Euler angles {θ1 , θ2 , θ3 } to the closed interval 0, 2 . On the vertices of the cube we
find SO(3) group elements lying in the Weyl group (modulo the center Z32 ), just as shown
in the three-dimensional picture on the right. In the two-dimensional picture on the left,
by choosing as fundamental eigenvalues λ1 = 1, λ2 = 2, λ3 = −3 the Weyl group element
wi ∈ W is identified by the point in the plane that has coordinates equal to the projections of
wi (λ1 , λ2 , λ3 ) along an orthonormal basis of vectors spanning the plane orthogonal to (1, 1, 1).
edges of the cube relate states of the Universe where there is no complete sorting of the
27
eigenvalues at past and future infinities. Indeed if we use as reference set the eigenvalues
of eq.(7.15) then complete sorting would require w2 ∈ W at −∞, corresponding to the
decreasing ordering 2, 1, −3 and w6 at +∞ corresponding to the increasing ordering −3, 1, 2
as we already observed. As it is evident by inspection, the matrices located on the twelve
edges define one-dimensional critical surfaces. It is a a fundamental property of the Lax
equation that flows touching upon a critical surface are completely constrained on it. Hence
flows touching one link just lie on that link at all instants of time and the asymptotic states
correspond to the vertices located at the endpoints of that link. The orientation of the link
is also decided a priori. Past infinity is the lower of the two end point Weyl elements while
future infinity is the higher one. This is just evident by comparing the graph in fig.(4) with
the ordering of Weyl group elements as displayed in table 1.
Besides one-dimensional critical surfaces (the links) there are also two-dimensional ones
(the faces) and these are obtained by studying the minors of O.
Faces Let us consider the orthogonal matrix O ∈ SO(3) and let us name and parameterize
its entries as in eq.s(7.16,7.17). Then there are in general exactly 6 minors that correspond
to the conditions involved in the definition of trapped surfaces in Section 5.2. Since we are
dealing with sl(3) all trapped surfaces are also critical. Three of the relevant minors are 1 ×1
minors and three of them are 2 × 2 minors. Imposing their vanishing one obtains equations
on the three parameters θ1 , θ2 , θ3 which would define as many critical surfaces, namely six.
Let us enumerate these candidate trapped and critical surfaces
O1,1 = 0 = cos (θ1 ) cos (θ3 ) − sin (θ1 ) sin (θ2 ) sin (θ3 ) ,
Σ1 :
O2,1 = 0 = − cos (θ3 ) sin (θ1 ) − cos (θ1 ) sin (θ2 ) sin (θ3 ) ,
Σ2 :
O3,1 = 0 = − cos (θ2 ) sin (θ3 ) ,
Σ3 :
Σ4 : O1,1 O2,2 − O1,2 O2,1 = 0 = cos (θ2 ) cos (θ3 ) ,
Σ5 : O1,1 O3,2 − O1,2 O3,1 = 0 = sin (θ1 ) sin (θ3 ) − cos (θ1 ) cos (θ3 ) sin (θ2 ) ,
Σ6 : O2,1 O3,2 − O2,2 O3,1 = 0 = cos (θ3 ) sin (θ1 ) sin (θ2 ) + cos (θ1 ) sin (θ3 ) .
(7.35)
It is now fairly simple to verify that, while the equations for Σ1 , Σ3 , Σ4 , and Σ5 can be solved
inside the cube 0 ≤ θi ≤ π2 , all solutions of the equations for Σ2 and Σ6 are located outside
this range. The existing inside the cube critical surfaces are shown in fig.s 5, 6, 7. By
means of a computer programme we can evaluate the flows on all these critical surfaces and
we find the following results for the asymptotic values of the Lax operator:
sin(θ2 ) sin(θ3 )
Σ1
: w2 → w6 (surface equation θ1 = arccos √ 2
),
2
2
cos (θ3 )+sin (θ2 ) sin (θ3 )
Σ3
Σ3
Σ4
T
Σ5
: w2 → w4 (for θ3 = 0) ,
: w5 → w6 (for θ3 = π2 ) ,
Σ4 : w5 → w4 (for θ2 = π2 ) ,
: w2
→ w3 (surface equation θ3 = arccos √
28
sin(θ1 )
sin2 (θ1 )+cos2 (θ1 ) sin2 (θ2 )
).
(7.36)
Π
2
5
Θ2
4
3
Π
2
6
4
Θ3
5
1
0
2
Θ1
Π
2
Figure 5: The picture on the left shows the critical surfaces Σ1 defined by the equation
O1,1 = 0 imposed on the SO(3) group element. The picture on the right reminds the reader
of the Weyl group elements located at the vertices of the parameter space. The vertices
belonging the surface are (in increasing order) w2 , w5 , w3 , w6 so that the flow is w2 7→ w6 .
The fourth case listed in eq.(7.36) needs a comment. When we set θ2 = π2 it happens that
both O3,1 = 0 and O1,1 O2,2 − O1,2 O2,1 = 0. Hence this plaquette of the hypercube is
actually the intersection of two critical surfaces. Altogether the result displayed in eq.(7.36)
could be predicted a priori relying on the notion of accessible vertices. Given the equation
of a critical surface, the accessible vertices are defined as those Weyl elements which have at
least one representative satisfying the defining condition and therefore belong to the surface.
Once the accessible set is defined, the flow is easily singled out. It goes from the lowest Weyl
member of the set to the highest one. This task is easily carried through in the present case.
For the six surfaces defined in equation (7.35) the corresponding accessible sets are rapidly
calculated and we find the result displayed in table 2. Expunging surfaces Σ2 and Σ6 which
fall outside the cube, we find that the available flows on critical two dimensional surfaces
inside the cube are just only four, namely the following ones:
w2
w2
w5
w2
7→
7
→
7
→
7
→
w6
w4
w6
w3
,
,
,
.
(7.37)
T
The only non vanishing intersection of these surfaces is the afore-mentioned plaquette Σ3
Σ4 .
We can easily calculate the intersection of vertices accessible to both Σ3 and Σ4 . We find
\
{w2 , w1 , w5 , w4 }
{w5 , w3 , w4 , w6 } = {w5 , w4 }
(7.38)
29
Π
2
5
Θ2
4
3
Π
2
6
4
Θ3
5
1
0
2
Θ1
Π
2
Figure 6: The picture on the left shows the union of the critical surfaces Σ3 and Σ4 , respectively defined by the equations O3,1 = 0 and O1,1 O2,2 − O1,2 O2,1 = 0 imposed on the minors
of the SO(3) group element. The picture on the right reminds the reader of the Weyl group
elements located at the vertices of the parameter space and shows the possible oriented flows
on the critical surfaces. The vertices belonging to Σ3 are w2 , w1 , w5 , w4 and the flow on this
surface is w2 7→ w4 . The vertices belonging to Σ4 are instead w5 , w3 , w4 , w6 andTthe flow
on this surface is w5 7→ w6 . The plaquette θ2 = π2 is actually the intersection Σ3
Σ4 and
on this surfaces the flow goes from the lowest to the highest of the elements in the set of the
vertices accessible to both surfaces, namely we have w5 7→ w4 .
where all sets are written in ascending order. It follows that on the surface Σ3
oriented flow is
w5 7→ w4
T
Σ4 the
(7.39)
as indeed it is verified by numerical calculation on the computer.
This concludes our discussion of the SL(3, R) which has been instrumental to illustrate
the involved mathematical structures.
We have seen that the topology of the parameter space H/W(U) is indeed complicated
and cannot be easily displayed as an hypercube. Yet it is completely defined by the trapped
hypersurfaces which admit a clear definition in terms of algebraic equations. These surfaces
split the parameter space H/W(U) into convex hulls which are separated from each other.
Indeed the walls are impenetrable according to Toda evolution. Moreover we can relate the
initial and final states of the flows to these critical surfaces defined by the vanishing of the
relevant minors in the orthogonal matrix O.
Our next section is devoted to another maximal split case of rank r = 2 which will
correspond to an entire Tits Satake universality class of cases.
30
Π
2
5
Θ2
4
3
Π
2
6
4
Θ3
5
1
0
2
Θ1
Π
2
Figure 7: The picture on the left shows the critical surfaces Σ5 defined by the equation
O1,1 O3,2 − O1,2 O3,1 = 0 imposed on the SO(3) group element. The identification of the cube
vertices with Weyl group elements is shown on the right. Here the accessible vertices are
w2 , w1 , w5 , w3 and the flow on this surface is w2 7→ w3 .
8
The maximally split case Sp(4, R)/U(2)
As we explained in the introduction our goal is the illustration of the Lax integration formula
in the non-maximally split case SO(r, r + 2s)/SO(r)×SO(r + 2s). The Tits Satake projection
of these manifolds is provided by the maximally split coset
SO(r, r + 1)
.
SO(r) × SO(r + 1)
(8.1)
SO(2, 3)
Sp(4, R)
=
SO(2) × SO(3)
U(2)
(8.2)
MTr S ≡
In the case of rank r = 2 we have
MT2 S ≡
due to the accidental isomorphism between the B2 and C2 Lie algebras whose Dynkin diagram
is displayed in fig. 8. For this reason we can rely on either formulation in terms of 4 × 4
symplectic matrices or 5 × 5 pseudo-orthogonal matrices to obtain the same result.
In the symplectic sp(4) interpretation, the C2 root system can be realized by the following
eight two-dimensional vectors:
∆C2 = ±ǫi ± ǫj , ±ǫi
(8.3)
where ǫi (i = 1, 2) denotes a basis of orthonormal unit vectors. In the pseudo-orthogonal
so(2, 3) interpretation of the same algebra the B2 root system is instead realized by the
following eight vectors:
∆B2 = ±ǫi ± ǫj , ±2 ǫi .
(8.4)
The two root systems are displayed in fig. 9.
31
Surf.
Equation
Σ1
O1,1 = 0
Σ2
O2,1 = 0
Σ3
O3,1 = 0
Σ4
−O1,2 O2,1 + O1,1 O2,2
Σ5
−O1,2 O3,1 + O1,1 O3,2
Σ6
−O2,2 O3,1 + O2,1 O3,2
Accessible Vertex
0 w2
1 w5
2 w3
3 w6
1 w1
2 w3
2 w4
3 w6
0 w2
1 w1
1 w5
2 w4
1 w5
2 w3
= 0
2
w
4
3 w6
0 w2
1 w1
= 0
1
w
5
2 w3
0 w2
1 w1
= 0
2
w
4
3 w6
Flow
w2 7→ w6
w1 7→ w6
w2 7→ w4
w5 7→ w6
w2 7→ w3
w2 7→ w6
Table 2: The accessible vertices for each of the six two-dimensional critical surfaces in the
case SL(3, R)/SO(3).
8.1
The Weyl group and the generalized Weyl group of sp(4, R)
Abstractly the Weyl group Weyl(C2 ) of the Lie algebra sp(4, R) is given by (Z2 × Z2 ) ⋉ S2
and its eight elements wi ∈ Weyl(C2 ) can be described by their action on the two Cartan
32
Figure 8: The Dynkin diagram of the B2 ∼ C2 Lie algebra.
C2
B2
Α1
✐
α1
❅
✐❅
✐
α2
✐
α2
α1
Α2
2 Α1 + Α2
Α1 + Α2
Α1 + Α2
2 Α1 + Α2
Α1
Α2
Figure 9: The C2 and B2 root systems. They are related by the exchange of the long with the
short roots and viceversa.
fields h1 , h2
w1 : (h1 , h2 ) → (h1 , h2 ) ,
w2 : (h1 , h2 ) → (−h1 , −h2 ) ,
w3 : (h1 , h2 ) → (−h1 , h2 ) ,
w4 : (h1 , h2 ) → (h1 , −h2 ) ,
w5 : (h1 , h2 ) → (h2 , h1 ) ,
(8.5)
w6 : (h1 , h2 ) → (h2 , −h1 ) ,
w7 : (h1 , h2 ) → (−h2 , h1 ) ,
w8 : (h1 , h2 ) → (−h2 , −h1 ) .
Just as we did in the previous case study we can introduce a partial ordering of the Weyl
group elements which will govern the orientation of all dynamical flows. The key point is
the embedding
Weyl (sp(4)) ֒→ S4 ≃ Weyl (sl(4))
(8.6)
33
of the Weyl group into the symmetric group S4 induced by the 4 × 4 representation of the
Lie algebra sp(4) in which the solvable Lie algebra is made of upper triangular matrices.
The explicit construction of such a representation is performed in the next section. For
the purpose of the considered issue, namely discovering the structure of the generalized
Weyl group and ordering of Weyl group elements, we anticipate some results. The matrices
corresponding to Cartan subalgebra elements are of the following form:
h1 0
0
0
0 h2
0
0
.
CSA ∋
(8.7)
0
0 0 −h2
0 0
0 −h1
In this way, the action of each Weyl group element as defined in eq.(8.5) can be reinterpreted
as a particular permutation of the set (h1 , h2 , −h2 , −h1 ) and this interpretation provides the
embedding (8.6). To make it precise let us derive the structure of the generalized Weyl group.
Following the definition 5.1 we introduce as generators the operator defined in eq.(8.24) for
the following four choices of the θ-angles:
π
0 0 0
2
0 π 0 0
2
{θ1 , θ2 , θ3 , θ4 } =
(8.8)
0 0 π 0
2
0 0 0 π2
which just corresponds to the 4 roots of sp(4). By closing the shell of products we obtain a
group with 32-elements, W(sp(4)). This group has an order four normal subgroup N(sp(4))
with the structure of Z2 × Z2 whose adjoint action on any of the Cartan matrices (8.7) is
the identity. Explicitly N(sp(4)) is made by the following four symplectic matrices:
N1
−1 0
0
=
0
0
N3
0
0
−1 0
0
0
0
1 0
−1 0 0 0
; N2 ≃ 0
−1 0
0
0
0
−1
0
0
1 0 0
0 1 0
0 0 −1
1 0 0 0
,
0 1 0 0
0 −1 0
0
.
; N4 ≃
≃
0
0
1
0
0
0
−1
0
0 0 0 1
0 0
0
1
(8.9)
(8.10)
As expected the factor group W(sp(4))/N(sp(4)) is isomorphic to the Weyl group Weyl (sp(4))
since the adjoint action of each equivalence class produces the same transformation on
the eigenvalues h1 , h2 as the abstract Weyl elements listed in eq.(8.5). Explicitly the 8-
34
equivalence classes of 4-elements each are displayed below
0
0
1 0 0 0
0 1 0 0
0
N(sp(4)) ; Ω2 ≃ 0
Ω1 ≃
−1
0
0 0 1 0
−1 0
0 0 0 1
0
0
Ω3 ≃
0
−1
0
1
Ω5 ≃
0
0
0
0
Ω7 ≃
−1
0
0 0 1
1 0
0 1
1 0
N(sp(4)) ,
0 0
0 0
0 0
1 0
1 0 0
N(sp(4)) ,
N(sp(4)) ; Ω4 ≃ 0 0
0 1 0
0 −1 0 0
0 0
0 1
0 0 0
1 0 0
0 0
1 0
1 0
0 0 0
0
0
N(sp(4)) ; Ω6 ≃
N(sp(4)) ,
0 0 1
0
0
0
1
0 1 0
0 −1 0 0
0
0
1 0
1 0 0
0
0
0
1
0 0 1
N(sp(4)) .
N(sp(4)) ; Ω8 ≃
0 0
0 0 0
−1 0
0
−1 0 0
0 1 0
(8.11)
(8.12)
(8.13)
(8.14)
Considering now a Cartan Lie algebra element in the fundamental representation of sp(4, R)
as given in eq.(8.7) we have
∀wi ∈ W eyl(C2)
:
ΩTi C ({h1 , h2 }) Ωi = C (wi {h1 , h2 }) .
(8.15)
This being established let us choose as conventional reference set of eigenvalues the following
one:
h1 = 1 ; h2 = 2 ,
(8.16)
then the decreasing sorting to be expected at past infinity is 2, 1, −1, −2 and corresponds
to the Weyl element Ω5 . If we take this as the fundamental permutation, all the other eight
permutations belonging to the Weyl group can be ranked with the number of transpositions
needed to bring them to the fundamental one. This procedure provides the partial ordering
of the Weyl group displayed in table 3.
We can now study the general features of the flows associated with the maximally split
coset manifold (8.2) and see how they follow the general principles and connect past and
future Kasner epochs ordered according to table 3. To realize this study the first essential
step is the construction of the sp(4, R) Lie algebra in a basis which fulfils the condition that
the solvable Lie algebra generating the coset is represented by upper triangular matrices.
The form of the Cartan subalgebra in such a basis was already anticipated in eq.(8.7), the
full construction is presented in the next section.
35
ℓT Weyl group
of sp(4, R)
0
w5
1
w6
2
w1
3
w3
3
w4
4
w2
5
w7
6
w8
Table 3: Partial ordering of the Weyl group of sp(4, R).
8.2
Construction of the sp(4, R) Lie algebra
The most compact way of presenting our basis is the following. Let us begin with the solvable
Lie algebra Solv(Sp(4, R)/U(2)). Abstractly the most general element of this algebra is given
by
T = h1 H1 + h2 H2 + e1 E α1 + e2 E α2 + e3 E α1 +α2 + e4 E 2α1 +α2 .
(8.17)
If we write the explicit form of T as
h1
0
Tsym =
0
a 4 × 4, upper triangular symplectic matrix
√
e1 e3
− 2e4
√
h2
2e2 e3
∈ sp(4, R)
0 −h2 −e1
0 0 0
−h1
which satisfies the condition
T
Tsym
02
12
−12 02
!
02
+
12
−12 02
!
Tsym = 0
(8.18)
(8.19)
all the generators of the solvable algebra are defined in the four dimensional symplectic
representation.
By writing the same Lie algebra element (8.17) as a 5 × 5 matrix
√
√
√
h1 + h2 − 2e2 − 2e3 − 2e4 0
√
√
0
2e4
h1 − h2 − 2e1 0
√
√
(8.20)
Tso = 0
∈ so(2, 3)
2e1
0
0
2e3
√
0
0
0
h2 − h1
2e2
0
0
0
0
36
−h1 − h2
which satisfies the condition
0 0
0 0
TsoT 0 0
0 1
1 0
0 0 1
0 0 0 0 1
0 0
0 1 0
+
0 0
1 0 0
0 1
0 0 0
0 0 0
1 0
0 1 0
1 0 0 Tso = 0
0 0 0
0 0 0
(8.21)
we define the same generators also in the five dimensional pseudo-orthogonal representation.
The choice of the invariant metric displayed in eq.(8.21) is that which guarantees the upper
triangular structure of the solvable Lie algebra generators. We shall come back on this point
in later sections.
Once the generators of the solvable Lie algebra are given the full Lie algebra can be
completed by defining the orthonormal generators of the K subspace as follows:
K1 = H1 ,
K2 = H2 ,
K3 = √12 E α1 + (E α1 )T ,
K4 = √12 E α2 + (E α2 )T ,
K5 =
K6 =
√1
2
√1
2
E α1 +α2 + (E α1 +α2 )T
E α1 +2α2 + (E α1 +2α2 )
,
T
,
and those of the maximal compact subalgebra H = u(2) as follows:
J1 = √12 E α1 − (E α1 )T ,
J2 = √12 E α2 − (E α2 )T ,
J3 = √12 E α1 +α2 − (E α1 +α2 )T ,
J4 = √12 E α1 +2α2 − (E α1 +2α2 )T .
(8.22)
(8.23)
In this way we have constructed all the relevant generators in both representations. The flows
are clearly an intrinsic property of the algebra and will not depend on the representation
chosen.
8.3
Parameterization of the compact group U(2) and critical submanifolds
In a way completely analogous to the previous case-study we can now parameterize the
compact subgroup by writing
i
h√
i
h√
(8.24)
O = exp 2 θ1 J1 exp [θ2 J2 ] exp 2 θ3 J3 exp [θ4 J4 ] .
The square root of two factors have been introduced in equation (8.24) in such a way as to
normalize the theta angles so that the group element O becomes an integer valued matrix
37
at θi = π2 . Obviously we have two instances of O: the 4 × 4 symplectic Osp and the 5 × 5
pseudo-orthogonal Oso . Both of them become integer valued for the same choice of the angles
and when acting by similarity transformation on the Cartan subalgebra they correspond to
Weyl group elements.
For simplicity we use the 4 × 4 representation and we find
O11 O12 O13 O14
O21 O22 O23 O24
(8.25)
O =
O
O
O
O
31
32
33
34
O41 O42 O43 O44
where
O11 = cos θ1 cos θ3 cos θ4 − sin θ1 sin θ3 sin (θ2 − θ4 ) ,
O12 = cos θ2 cos θ3 sin θ1 ,
O13 = cos θ3 sin θ1 sin θ2 + cos θ1 sin θ3 ,
(8.26)
O14 = cos θ2 cos θ4 sin θ1 sin θ3 + (sin θ1 sin θ2 sin θ3 − cos θ1 cos θ3 ) sin θ4 ,
O21 = − cos θ3 cos θ4 sin θ1 − cos θ1 sin θ3 sin (θ2 − θ4 ) ,
O22 = cos θ1 cos θ2 cos θ3 ,
O23 = cos θ1 cos θ3 sin θ2 − sin θ1 sin θ3 ,
(8.27)
O24 = cos θ1 cos θ2 − θ4 sin θ3 + cos θ3 sin θ1 sin θ4 ,
O31 = − cos θ1 cos θ2 − θ4 sin θ3 − cos θ3 sin θ1 sin θ4 ,
O32 = sin θ1 sin θ3 − cos θ1 cos θ3 sin θ2 ,
O32 = cos θ1 cos θ2 cos θ3 ,
(8.28)
O33 = cos θ1 cos θ2 cos θ3 ,
O34 = − cos θ3 cos θ4 sin θ1 − cos θ1 sin θ3 sin (θ2 − θ4 ) ,
O41 = (cos θ1 cos θ3 − sin θ1 sin θ2 sin θ3 ) sin θ4 − cos θ2 cos θ4 sin θ1 sin θ3 ,
O42 = − cos θ3 sin θ1 sin θ2 − cos θ1 sin θ3 ,
O43 = cos θ2 cos θ3 sin θ1 ,
(8.29)
O44 = cos θ1 cos θ3 cos θ4 − sin θ1 sin θ3 sin (θ2 − θ4 ) .
Vertices Having parameterized in this way the U(2) group element with the four Euler
angles θi , in a completely analogous way to the case of SL(3, R), we can check that when all
of the θi take either the 0 or the π2 value then the corresponding matrix O becomes integer
valued and its similarity action on a Cartan subalgebra element (8.7) corresponds to the
action of some Weyl group element on the eigenvalues:
0
If ∀i θi =
, ∃ ω ∈ W eyl(C2 ) / OT C ({h1 , h2 }) O = C (ω{h1 , h2 }) . (8.30)
or
π
2
38
#
vertex
Weyl group element
multiplicity of Weyl elem.
1
{0, 0, 0, 0}
Ω1
1
{1, 0, 0, 0}
Ω5
1
{0, 1, 0, 0}
Ω4
3
{0, 0, 1, 0}
Ω8
3
{0, 0, 0, 1}
Ω3
1
{1, 1, 0, 0}
Ω6
3
{1, 0, 1, 0}
Ω2
3
{1, 0, 0, 1}
Ω7
1
{0, 1, 1, 0}
Ω6
3
{0, 1, 0, 1}
Ω2
3
{0, 0, 1, 1}
Ω6
3
{1, 1, 1, 0}
Ω4
3
{1, 1, 0, 1}
Ω8
3
{1, 0, 1, 1}
Ω4
3
{0, 1, 1, 1}
Ω8
3
{1, 1, 1, 1}
Ω2
3
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Table 4: The 16 vertices of the hypercubic parameter space for Sp(4, R)/U(2) and their
identification with Weyl group elements.
In this way the parameter space U(2)/W is reduced to lie in a four dimensional hypercube
and, using a notation analogous to that of eq.(7.18), the identification of the 16 vertices of
the hypercube with Weyl group elements is displayed in table 4. As the reader can observe,
in the chosen numbering the odd-labeled Weyl group elements appear only once, while the
even-labeled appear three-times.
Edges Using just the same strategy as in the previous case-study we can now construct the
64 oriented links connecting the 16 vertices. These are all the possible segments of straight
lines in parameter space connecting two vertices and by means of a computer programme
we can evaluate the orientation of the link, namely discover which of the end-points (Weyl
group element) corresponds to past infinity t = −∞ and which to future infinity t = +∞. As
expected the orientation of all the links is in the direction from lower to higher Weyl elements,
according to the ordering of table 3. The result of these computations is displayed in table
5 and summarized in the flow diagram of fig.10. A four dimensional hypercube cannot be
drawn in three dimension but a standard way to visualize it is provided by presenting its
stereographic projection. Indeed if we shift the origin of the coordinate system to the point
{ 12 , 21 , 12 , 21 } then all the 16 vertices of the hypercube are located on the standard three-sphere,
namely, as 4-component vectors they have unit norm. So we can consider their stereographic
projection from S3 to R3 and connecting them with segments we obtain the visualization of
the hypercube displayed in fig.11.
Trapped hypersurfaces The study of trapped hypersurfaces can now be performed once
again in complete analogy with the case of SL(3, R). We just have to calculate all the relevant
minors and impose their vanishing. In this way we determine equations on the parameters
that have to be solved within the hypercubic range. If solutions within the hypercube exist,
39
#
Vertex
Vertex
Flow
1
{0, 0, 0, 0}
{0, 0, 0, 1}
Ω1 7→ Ω3
{0, 0, 0, 0}
{1, 0, 0, 0}
Ω5 7→ Ω1
5
Ω3 7→ Ω7
9
2
4
6
8
10
12
14
16
18
20
22
24
26
28
30
32
34
36
38
40
42
44
46
48
50
52
54
56
58
60
62
64
{0, 0, 0, 0}
{0, 0, 0, 1}
{0, 0, 0, 1}
{0, 0, 1, 0}
{0, 0, 1, 0}
{0, 0, 1, 1}
{0, 0, 1, 1}
{0, 1, 0, 0}
{0, 1, 0, 0}
{0, 1, 0, 1}
{0, 1, 0, 1}
{0, 1, 1, 0}
{0, 1, 1, 0}
{0, 1, 1, 1}
{0, 1, 1, 1}
{1, 0, 0, 0}
{1, 0, 0, 0}
{1, 0, 0, 1}
{1, 0, 0, 1}
{1, 0, 1, 0}
{1, 0, 1, 0}
{1, 0, 1, 1}
{1, 0, 1, 1}
{1, 1, 0, 0}
{1, 1, 0, 0}
{1, 1, 0, 1}
{1, 1, 0, 1}
{1, 1, 1, 0}
{1, 1, 1, 0}
{1, 1, 1, 1}
{1, 1, 1, 1}
#
Vertex
Vertex
Flow
Ω1 7→ Ω8
3
{0, 0, 0, 0}
{0, 1, 0, 0}
Ω1 7→ Ω4
Ω6 7→ Ω3
7
{0, 0, 0, 1}
{0, 1, 0, 1}
Ω3 7→ Ω2
Ω6 7→ Ω8
11
Ω6 7→ Ω8
15
Ω4 7→ Ω2
19
Ω4 7→ Ω2
23
Ω6 7→ Ω4
27
{0, 1, 0, 1}
Ω2 ] 7→ Ω8
31
{1, 0, 0, 1}
Ω5 7→ Ω7
35
Ω5 7→ Ω7
39
Ω5 7→ Ω2
43
Ω4 7→ Ω7
47
Ω5 7→ Ω6
51
Ω7 7→ Ω8
55
Ω4 7→ Ω2
59
Ω4 7→ Ω2
63
{0, 0, 1, 0}
{0, 0, 1, 1}
{1, 0, 0, 1}
{0, 0, 1, 1}
{1, 0, 1, 0}
{0, 0, 1, 0}
{1, 0, 1, 1}
{0, 1, 0, 1}
{1, 1, 0, 0}
{0, 1, 0, 0}
{1, 1, 0, 1}
{0, 1, 0, 0}
{1, 1, 1, 0}
{1, 1, 1, 1}
{1, 1, 0, 0}
{1, 0, 0, 0}
{1, 1, 0, 1}
{1, 0, 0, 0}
{1, 1, 1, 0}
{1, 0, 0, 1}
{1, 1, 1, 1}
{1, 0, 0, 0}
{1, 1, 1, 0}
{1, 0, 0, 1}
{1, 1, 1, 1}
{1, 0, 1, 0}
{1, 1, 1, 1}
{1, 0, 1, 1}
{1, 1, 1, 0}
Ω2 7→ Ω8
13
Ω6 7→ Ω4
17
Ω6 7→ Ω4
21
Ω2 7→ Ω8
25
Ω6 7→ Ω4
29
Ω2 7→ Ω8
33
Ω5 7→ Ω6
37
Ω7 7→ Ω8
41
Ω4 7→ Ω2
45
Ω4 7→ Ω2
49
Ω6 7→ Ω4
53
Ω2 7→ Ω8
57
Ω4 7→ Ω2
61
Ω4 7→ Ω2
{0, 0, 0, 1}
{0, 0, 1, 0}
{0, 0, 1, 0}
{0, 0, 1, 1}
{0, 0, 1, 1}
{0, 1, 0, 0}
{0, 1, 0, 0}
{0, 1, 0, 1}
{0, 1, 0, 1}
{0, 1, 1, 0}
{0, 1, 1, 0}
{0, 1, 1, 1}
{0, 1, 1, 1}
{1, 0, 0, 0}
{1, 0, 0, 0}
{1, 0, 0, 1}
{1, 0, 0, 1}
{1, 0, 1, 0}
{1, 0, 1, 0}
{1, 0, 1, 1}
{1, 0, 1, 1}
{1, 1, 0, 0}
{1, 1, 0, 0}
{1, 1, 0, 1}
{1, 1, 0, 1}
{1, 1, 1, 0}
{1, 1, 1, 0}
{1, 1, 1, 1}
{1, 1, 1, 1}
{0, 0, 0, 0}
{0, 0, 0, 0}
{0, 1, 1, 0}
{0, 0, 0, 1}
{0, 1, 1, 1}
{0, 0, 0, 0}
{0, 1, 1, 0}
{0, 0, 0, 1}
{0, 1, 1, 1}
{0, 0, 1, 0}
{0, 1, 1, 1}
{0, 0, 1, 1}
{0, 1, 1, 0}
{0, 0, 0, 0}
{1, 0, 1, 0}
{0, 0, 0, 1}
{1, 0, 1, 1}
{0, 0, 1, 0}
{1, 0, 1, 1}
{0, 0, 1, 1}
{1, 0, 1, 0}
{0, 1, 0, 0}
{1, 1, 0, 1}
{0, 1, 0, 1}
{1, 1, 0, 0}
{0, 1, 1, 0}
{1, 1, 0, 0}
{0, 1, 1, 1}
{1, 1, 0, 1}
Ω1 7→ Ω3
Ω1 7→ Ω8
Ω6 7→ Ω8
Ω6 7→ Ω3
Ω6 7→ Ω8
Ω1 7→ Ω4
Ω6 7→ Ω4
Ω3 7→ Ω2
Ω2 7→ Ω8
Ω6 7→ Ω8
Ω6 7→ Ω8
Ω6 7→ Ω8
Ω6 7→ Ω8
Ω5 7→ Ω1
Ω5 7→ Ω2
Ω3 7→ Ω7
Ω4 7→ Ω7
Ω2 7→ Ω8
Ω4 7→ Ω2
Ω6 7→ Ω4
Ω4 7→ Ω2
Ω6 7→ Ω4
Ω6 7→ Ω8
Ω2 7→ Ω8
Ω6 7→ Ω8
Ω6 7→ Ω4
Ω6 7→ Ω4
Ω2 7→ Ω8
Ω2 7→ Ω8
Table 5: A 4-dimensional hypercube has 32 edges which amount to 64 edges if we consider
also the possible orientation. Here are displayed the 64 oriented links of the hypercubic
parameter space for Sp(4, R)/U(2) flows and the corresponding oriented links from Weyl
group elements to Weyl group elements.
then we have a trapped surface. Otherwise we just have a Weyl replica of an already existing
surface. In our case there are just 14 relevant minors distributed in the following way: 4 of
rank 1, 4 of rank 3 and 6 of rank 2. Explicitly we can define the following candidate trapped
40
W
3
W
1
W
7
W
5
W
8
W
6
W
2
W
4
1
2
3
4
5
6
7
8
1
1
2
3
4
5
6
7
8
2
2
1
4
3
8
7
6
5
3
3
4
1
2
6
5
8
7
4
4
3
2
1
7
8
5
6
5
5
8
7
6
1
4
3
2
6
6
7
8
5
3
2
1
4
7
7
6
5
8
4
1
2
3
8
8
5
6
7
2
3
4
1
Figure 10: The oriented phase diagram of the Sp(4, R)/U(2) flows. The Lie algebra sp(4, R)
is the maximally split real section of the complex Lie algebra C2 ∼ B2 . Its Weyl group is
(Z2 × Z2 ) ⋉ S2 and has eight elements identified by their action on the eigenvalues h1 , h2
of the Lax operator. Eight are therefore the possible asymptotic states of the universe at
t = ±∞ and each possible motion is an oriented flow from one Weyl element to another
one. The orientation follows the ordering of Weyl group elements: it is always from a lower
to a higher one. In this picture, choosing as fundamental eigenvalues h1 = 1, h2 = 2 the Weyl
group element Ωi ∈ W eyl(C2) is identified by the point in the plane that has coordinates
Ωi (h1 , h2 ). Each link is therefore associated with a Weyl group element which multiplying on
the left the past infinity element produces the future infinity one. By comparison we display
below the graph the multiplication table of the Weyl group. Note that in each vertex of the
diagram there meet just four lines. In vertex Ω5 there are only outgoing lines. This is so
because Ω5 is the lowest Weyl element and it corresponds to the universal past infinity point
for generic flows. On the contrary in the vertex Ω8 there only incoming lines. This is so
because Ω8 is the highest Weyl element and it corresponds to the universal future infinity for
generic flows. The other vertices have both incoming and outgoing lines.
surfaces:
Σ1
:
O1,1
=
0,
Σ2
:
O2,1
=
0,
Σ3
:
O3,1
=
0,
Σ4
:
O4,1
=
0,
Σ5
:
=
0,
Σ6
:
O1,3 (O2,1 O3,2 − O2,2 O3,1 ) + O1,2 (O2,3 O3,1 − O2,1 O3,3 ) + O1,1 (O2,2 O3,3 − O2,3 O3,2 )
=
0,
Σ7
:
O1,3 (O3,1 O4,2 − O3,2 O4,1 ) + O1,2 (O3,3 O4,1 − O3,1 O4,3 ) + O1,1 (O3,2 O4,3 − O3,3 O4,2 )
=
0,
O1,3 (O2,1 O4,2 − O2,2 O4,1 ) + O1,2 (O2,3 O4,1 −
41O2,1 O4,3 ) + O1,1 (O2,2 O4,3 − O2,3 O4,2 )
(8.31)
8
2
6
6
4
8
2 4
4
2
6
1
8
5
3
7
Figure 11: Stereographic projection of the hypercubic parameter space for Sp(4, R)/U(2)
motions.
Σ8
:
Σ9
:
Σ10
:
Σ11
:
Σ12
:
Σ13
:
Σ14
:
O2,3 (O3,1 O4,2 − O3,2 O4,1 ) + O2,2 (O3,3 O4,1 − O3,1 O4,3 ) + O2,1 (O3,2 O4,3 − O3,3 O4,2 )
=
0,
O1,1 O2,2 − O1,2 O2,1
=
0,
O1,1 O3,2 − O1,2 O3,1
=
0,
O1,1 O4,2 − O1,2 O4,1
=
0,
O2,1 O3,2 − O2,2 O3,1
=
0,
O2,1 O4,2 − O2,2 O4,1
=
0,
O3,1 O4,2 − O3,2 O4,1
=
0
(8.32)
where the explicit form of the equation can be obtained by substituting the values of the
U(2) matrix elements as given in eq.s (8.25–8.29). A full-fledged analysis of all the trapped
surfaces is beyond the scope of the present paper which aims at illustrating the general
principles and at explaining the method. What we can do without any analytic study of the
trapped surfaces is to determine the accessible Weyl group elements for each of them and in
this way single out the corresponding past infinity and future infinity states. The result is
shown in table 6.
42
Surf Accessible Weyl el.
Σ1
Σ2
Σ3
Σ4
Σ5
Σ6
Σ7
Σ8
Σ9
Σ10
Σ11
Σ12
Σ13
Σ14
Flow
Type
{w5 , w6 , w3 , w2 , w7 , w8}
w5 7→ w8 critical
{w5 , w6 , w1 , w3 , w4 , w2}
w5 7→ w2 super-critical
{w1 , w3 , w4 , w2 , w7 , w8}
w1 7→ w8 super-critical
{w5 , w6 , w1 , w4 , w7 , w8}
w5 7→ w8 critical
{w5 , w6 , w3 , w2 , w7 , w8}
w5 7→ w8 critical
{w1 , w3 , w4 , w2 , w7 , w8}
w1 7→ w8 super-critical
{w5 , w6 , w1 , w3 , w4 , w2}
w5 7→ w2 super-critical
{w5 , w6 , w1 , w4 , w7 , w8}
w5 7→ w8 critical
{w6 , w3 , w4 , w2 , w7 , w8}
w6 7→ w8 super-critical
{w5 , w6 , w1 , w3 , w2 , w8}
w5 7→ w8 critical
{w5 , w6 , w1 , w3 , w4 , w2, w7 , w8 } w5 7→ w8 trapped non crit.
{w5 , w6 , w1 , w3 , w4 , w2, w7 , w8 } w5 7→ w8 trapped non crit.
{w5 , w1 , w4 , w2 , w7 , w8}
w5 7→ w8 critical
{w5 , w6 , w1 , w3 , w4 , w7}
w5 7→ w7 super-critical
Table 6: Accessible Weyl elements on the 14 trapped surfaces of Sp(4, R)/U(2) flows. By
inspecting the list AΣ of accessible Weyl elements we easily deduce the character of the
surface. If there are missing Weyl elements it is critical. It is super-critical if one of the
missing elements is either Ωmin = w5 and/or Ωmax = w8 . When no Weyl element is missing
in AΣ , the surface is just only trapped. It would be critical inside the bigger group SL(4, R).
As it is evident by inspection of this table the possible flows realized on critical hypersurfaces of parameter space are just a very small number, namely the following five:
1 : w1 7→ w8 ,
2 : w5 7→ w2 ,
3 : w5 7→ w7 ,
(8.33)
4 : w5 7→ w8 ,
5 : w6 7→ w8 .
As we are going to see this is a property shared by the entire universality class of manifolds
that have the same Tits Satake projection. The Weyl group, the flow diagram on the links
and the possible flows realized on critical surfaces do not depend on the representative inside
the class but are just a property of the class.
In the next section we just focus on the detailed discussion of a few examples of flows for
this maximally split manifold.
43
8.4
Examples for sp(4, R)
In this section, as we just announced, we consider three examples of Sp(4, R)-flows. One
will be in the bulk the other two will be located on two different critical surfaces. We
analyse these cases in detail both to show the relation between the asymptotic states and
the structure of the orthogonal group element O and to illustrate the billiard phenomenon.
In the plots of the Cartan fields we will be able to observe the multiple bouncing of the
cosmic ball on the hyperplanes orthogonal to some of the roots.
8.4.1
An example of flow in the bulk of parameter space: Ω5 ⇒ Ω8
If, as initial data we choose the following element of the compact subgroup U(2) ⊂ Sp(4, R):
U(2) ∋ O = exp[ π6 J1 ] exp[ π4 J2 ] exp[ π6 J3 ] exp[ π3 J4 ]
0
B
B
B
B
=B
B
B
@
1
16
1
16
1
16
1
16
“
√ ”
√
6− 2+ 6
“ √
√
√ ”
3 2−2 3− 6
“
√
√ ”
−6 − 3 2 − 6
“ √
√
√ ”
− 2+6 3− 6
q
3
2
4
3
√
4 “2
1
2
8
√ ”
−3 2
√ ”
√ “
− 18 3 2 + 2
1
8
1
8
√ ”
√ “
3 2+ 2
“
√ ”
−2 + 3 2
3
√
4 2
q
3
2
1
16
1
16
1
16
1
16
4
√
√ ”
2−6 3+ 6
“
√
√ ”
6+3 2+ 6
“ √
√
√ ”
3 2−2 3− 6
“
√ ”
√
6− 2+ 6
“√
(8.34)
1
C
C
C
C
C
C
C
A
we are just in the bulk. Indeed as it is evident by inspection of eq.(8.34), the 4 × 4 matrix
representing O has no minor with vanishing determinant. Hence according to the advocated
theorems we expect asymptotic sorting of the eigenvalues. Indeed this is what happens.
Implementing by numerical evaluation the integration formula on a computer we discover
that the asymptotic form of the Lax operator at t = −∞ corresponds to the Weyl group
element Ω5
2 0 0
0
0 1 0
0
⇔ Ω5 .
lim L(t) =
(8.35)
t → −∞
0 0 −1 0
0 0 0 −2
Similarly at asymptotically late times the limit of the
the Weyl group element Ω8
−2 0 0
0 −1 0
lim L(t) =
0
t → +∞
0 1
0
0 0
Lax operator is that corresponding to
0
0
⇔ Ω8 .
0
2
(8.36)
Algebraically we have Ω8 = Ω2 Ω5 , so that all generic flows in the bulk that avoid touching
critical surfaces are a smooth realization of the Weyl reflection Ω2 ∈ W. The particular
smooth realization of this reflection provided by the present choice of parameters is illustrated
in fig.12 which displays the motion of the cosmic ball on the two dimensional billiard table
whose axes are the Cartan fields h1 , h2 . This motion involves two bounces as it becomes
→
−
evident by plotting the projection of the Cartan vector h = (h1 , h2 ) along the two roots
α1 = (1, −1) and α3 = (1, 1). These plots are displayed in fig.13.
44
h2
14
12
10
8
6
h1
16
18
20
22
Figure 12: Motion of the cosmic ball on the CSA billiard table of Sp(4, R) in a generic bulk
case. The choice of the angles is θ1 = π6 , θ2 = π4 , θ3 = π6 , θ4 = π3 . According to theory this
motion realizes the smooth reflection Ω2 from the universal primordial Kasner era Ω5 to the
universal remote future Kasner era Ω8 . This motion involves just two bounces on the wall
respectively orthogonal to the root α3 = (1, −1) and α1 = (1, 1). In this picture the straight
lines represent the walls orthogonal to α1 and α3 , respectively.
32
8
30
7
28
6
26
5
24
4
22
-3
-2
-1
1
2
3
-3
-2
-1
1
2
3
Figure 13: Plot of α1,3 · h projections for the Sp(4, R) generic bulk flow generated by the
parameter choice θ1 = π6 , θ2 = π4 , θ3 = π6 , θ4 = π3 which connects the primordial Kasner
era Ω5 to the remote future Kasner era Ω8 . There are two bounces in this flow because the
projections α1,3 · h have maxima at different instant of times.
8.4.2
An example of flow on the super-critical surface Σ9 : Ω6 ⇒ Ω8
As a next example we consider a flow confined on a super-critical surface. If we set θ2 =
the U(2) matrix takes the following form:
0
O
=
cos (θ1 + θ3 ) cos (θ4 )
B
B − cos (θ4 ) sin (θ1 + θ3 )
B
B − sin (θ + θ ) sin (θ )
1
3
4
@
cos (θ1 + θ3 ) sin (θ4 )
0
sin (θ1 + θ3 )
0
cos (θ1 + θ3 )
− cos (θ1 + θ3 )
− sin (θ1 + θ3 )
0
0
− cos (θ1 + θ3 ) sin (θ4 )
π
2
1
C
C
C
− cos (θ4 ) sin (θ1 + θ3 ) C
A
cos (θ1 + θ3 ) cos (θ4 )
sin (θ1 + θ3 ) sin (θ4 )
(8.37)
45
which has two notable properties. The first is that it has vanishing some principal minors,
the second is that it actually depends on two variables only, namely θ1 + θ3 and θ4 .
If we consider the equation for the critical super-surface Σ9 we find
cos (θ2 ) cos2 (θ3 ) cos (θ4 ) = 0
(8.38)
so that the hyperplane θ2 = π2 where we have chosen our group element is just one of the
three components of Σ9 . Furthermore, because of what we just observed, on this hyperplane
all points of the following form have to be identified:
π
π
π
∀ φ ∈ 0, 2 ; θ1 + φ, , θ3 − φ, θ4 ∼ θ1 , , θ3 , θ4 .
(8.39)
2
2
Having chosen initial data on a trapped surface, namely Σ9 , we do not expect full asymptotic
sorting of the eigenvalues: actually, according to table 6 we expect flows from Ω6 to Ω8 .
Indeed this is what happens. We verify it in one example. For instance, as initial data we
choose the following element of the compact subgroup U(2) ⊂ Sp(4, R), which lies in the
considered hypersurface:
U(2) ∋ O = exp[ π3 J1 ] exp[ π2 J2 ] exp[ π3 J3 ] exp[ π3 J4 ]
√
√
3
3
− 41
0
2
4
√3
1
3
−4 0
−
2
4
√ .
=
3
1
0
− 43
−4
2
−
√
3
4
−
√
3
2
0
(8.40)
− 14
Implementing by numerical evaluation the integration formula on a computer we discover
that the asymptotic form of the Lax operator at t = −∞ corresponds to the Weyl group
element Ω6 which implies no decreasing sorting of the eigenvalues
2 0 0 0
0 −1 0 0
(8.41)
lim L(t) =
0 0 1 0 ⇔ Ω6 .
t → −∞
0 0 0 −2
On the other hand the limit of the Lax operator at asymptotically late times is that corresponding to the Weyl group element Ω8 which is the same occurring in generic flows and
yields increasing sorting of the eigenvalues
−2 0 0 0
0 −1 0 0
⇔ Ω8 .
(8.42)
lim L(t) =
t → +∞
0 1 0
0
0
0 0 2
Algebraically we have Ω8 = Ω3 Ω6 , so that the flows occurring on this super-critical surface
are smooth realizations of the Weyl reflection Ω3 ∈ W. The particular smooth realization
of this reflection encoded in this flow is illustrated in fig.14 which displays the motion of the
cosmic ball on the two dimensional billiard table This motion involves just one bounce on
the wall orthogonal to the root α4 as it becomes evident by inspecting fig.15.
46
h2
5
h1
16
18
20
22
-5
-10
-15
Figure 14: Motion of the cosmic ball on the CSA billiard table of Sp(4, R) in a super-critical
surface case. The choice of the angles is θ1 = π3 , θ2 = π2 , θ3 = π3 , θ4 = π3 . This motion
realizes the smooth reflection Ω3 from the Kasner era Ω6 at t = −∞ to the Kasner era Ω8
at t = +∞. The two straight lines appearing in the picture are the walls orthogonal to the
roots α2 = (0, 2) and α4 = (2, 0), respectively. As one sees the cosmic ball just bounces once
the on the α4 wall.
-15
42.5
-17.5
40
-20
37.5
-22.5
35
-25
32.5
-27.5
-3
-3
-2
-1
1
2
-2
-1
1
2
3
3
27.5
Figure 15: Plot of α2,4 ·h projections for the Sp(4, R) flow on a super-critical surface generated
by the parameter choice θ1 = π3 , θ2 = π2 , θ3 = π3 , θ4 = π3 which connects the past Kasner era
Ω6 to the future Kasner era Ω8 . There is just one bounce in this flow and this occurs on the
α4 wall.
8.4.3
An example of flow on the super-critical surface Σ2 : Ω1 ⇒ Ω8
In the example considered below the initial state is different from that appearing in a generic
bulk flow, namely there is not decreasing sorting of the eigenvalues at past infinity which
rather corresponds to the Weyl element Ω1 . Yet the end point at t = +∞ coincides with
the universal one Ω8 , namely there is increasing sorting at future infinity. We realize this
situation by choosing initial data on one of the critical surfaces, namely the surface Σ2 .
47
The equation for the trapped surface Σ2 , as defined in (8.31) reads as follows:
0 = cos (θ3 ) cos (θ4 ) sin (θ1 ) + cos (θ1 ) sin (θ3 ) sin (θ2 − θ4 )
(8.43)
and it can be solved within the hypercube by expressing θ1 in terms of the remaining three
Euler angles as it follows:
!
cos (θ3 ) cos (θ4 )
.
(8.44)
θ1 = arccos p
cos2 (θ3 ) cos2 (θ4 ) + sin2 (θ3 ) sin2 (θ2 − θ4 )
On the hypersurface Σ2 we choose the particularly nice point
nπ π π πo
∈ Σ2
(8.45)
, , ,
3 6 3 3
which is easily seen to verify the defining equation (8.43) and which leads to a quite simple
form of the matrix O. With these values of the Euler angles we obtain the following element
of the maximally compact subgroup U(2) ⊂ Sp(4, R):
U(2) ∋ Hsp = exp[ π3 J1 ] exp[ π6 J2 ] exp[ π3 J3 ] exp[ π3 J4 ]
√
√
1
2
3
8
√
3
0
8
=
3
5
−√
4
8 √
− 43 − 3 8 3
3 3
8
− 58
√
3
8
3
8
3
4
3
4
0
1
2
(8.46)
which indeed has vanishing O2,1 matrix element as it is required by the definition of the
Σ2 trapped surface. Hence according to table 6 we expect a flow from Ω1 to Ω8 . Before
proceeding to the integration of the Lax equation it is interesting to consider the so(2, 3)
5-dimensional representation of the same U(2) group element. It is explicitly given by the
following matrix:
√
√
3
5
3
7 3
3
√
−
−
−
16
16
16
16
√4 2
√
√
3 32
11 3
5 3
3
− 19
−
−
√32
32
8
32
32
√
3 3
3
3
.
1
15
15
2
2
Hso =
(8.47)
√
√
−
16
8
16
16 2
16 2
√
√
√
3
3 2
−3
5 3
11 3
− 19
32√ − 32
8
32
32
√
7 3
3
3
3
5
− 16 − 16
− 4√2 16
16
Since all the properties of the flows are intrinsic properties of the group and cannot depend
on the chosen representation it follows that also the matrix (8.47) should be critical namely
some of its relevant minors (those obtained by intesecting the first k-columns with k arbitrary
rows should vanish. Although not evident at first sight, this is indeed true. Calculating the
minors we find that there are three relevant 2×2 minors whose determinant vanishes, namely
√
!
√
5
3
5
3
16
16
16
16 √
= 0 ; Det
= 0,
(8.48)
Det 3√ 3
3
5 3
15
2
√
−
−
32
32
16
16 2
√
3
Det
3
2
16
3
− 32
15
√
16 √
2
5 3
− 32
= 0.
(8.49)
48
Hence the criticality condition is indeed intrinsic to the choice of the group element and not
to its specific representation as a matrix.
Implementing by numerical evaluation the integration formula on a computer we discover
that the asymptotic form of the Lax operator at t = −∞ corresponds to the Weyl group
element Ω1 as expected:
1 0 0
0
0 2 0
0
⇔ Ω1
(8.50)
lim L(t) =
t → −∞
0
0
−2
0
0 0 0 −12
while the limit at asymptotically late times is that corresponding to the Weyl group element
Ω8 as we also expected
−2 0 0 0
0 −1 0 0
⇔ Ω8 .
lim L(t) =
(8.51)
t → +∞
0 1 0
0
0
0 0 2
Algebraically we have Ω8 = Ω8 Ω1 , so that the flows occurring on this super-critical surface
are smooth realizations of the Weyl reflection Ω8 ∈ W. The smooth realization of this
reflection encoded in the flow with these initial data is illustrated in fig.16 which displays
the motion of the cosmic ball in the h1 , h2 Cartan subalgebra plane. This motion involves
three bounces two on the wall orthogonal to the simple root α1 and one on the wall orthogonal
to α2 . This is clearly visible by inspection of fig.17.
9
The case of the so(r, r + 2s) algebra
We are interested in considering the sigma model on the symmetric non compact coset
manifold
SO(r, r + 2s)
M(r,2s) =
.
(9.1)
SO(r) × SO(r + 2s)
For r = 4 the above manifold is quaternionic and corresponds to the family of special
geometries L(0, P = 2s).
9.1
The corresponding complex Lie algebra and root system
The complex Lie algebra of which so(r, r + 2s) is a non-compact real section is just Dℓ where
ℓ=r+s .
(9.2)
The corresponding Dynkin diagram is displayed in fig 18 and the associated root system is
realized by the following set of vectors in Rℓ :
∆ ≡ ± ǫA ± ǫB
; card ∆ = 2 ℓ2 − ℓ
(9.3)
49
h2
h1
Figure 16: Motion of the cosmic ball on the CSA billiard table of Sp(4, R) in a super-critical
surface case. The choice of the angles is θ1 = π3 , θ2 = π6 , θ3 = π3 , θ4 = π3 which lie on the
trapped and super-critical surface Σ2 . This motion realizes the smooth reflection Ω8 from the
Kasner era Ω1 at t = −∞ to the Kasner era Ω8 at t = +∞. The peculiar knot appearing
in this picture implies the existence of two bounces on the same root wall. The two straight
lines displayed in the figure are the walls orthogonal to the two simple roots α1 = (1, −1) and
α2 = (0, 2). The ball bounces twice on the α1 wall.
-13.8
-14
55
-14.2
-14.4
50
-14.6
-14.8
-1.5
-1
-0.5
45
1
0.5
1.5
-15.2
-3
-2
-1
1
2
3
Figure 17: Plot of α1,2 ·h projections for the Sp(4, R) flow on a super-critical surface generated
by the parameter choice θ1 = π3 , θ2 = π6 , θ3 = π3 , θ4 = π3 which connects the past Kasner
era Ω1 to the future Kasner era Ω8 . The two bounces are clearly visible in the maxima and
minima of the first graph.
where ǫA denotes an orthonormal basis of unit vectors. The set of positive roots is then
easily defined as follows:
α > 0 ⇒ α ∈ ∆+ ≡ ǫA ± ǫB
(A < B) .
(9.4)
A standard basis of simple roots representing the Dynkin diagram 18 is given by
α1
=
ǫ1 − ǫ2 ,
50
Figure 18: The Dynkin diagram of the Dℓ Lie algebra.
✐αℓ−1
Dℓ
✐
✐
α1
α2
α2 =
... ...
αℓ−1 =
αℓ =
✐. . . ✐
α3
αℓ−3
✐
❅
αℓ−2
❅✐
αℓ
ǫ2 − ǫ3 ,
... ,
ǫℓ−1 − ǫℓ ,
ǫℓ−1 + ǫℓ .
(9.5)
The maximally split real form of the Dℓ Lie algebra is so(ℓ, ℓ) and it is explicitly realized
by the following 2ℓ × 2ℓ matrices. Let eA,B denote the 2ℓ × 2ℓ matrix whose entries are all
zero except the entry A, B which is equal to one. Then the Cartan generators HA and the
positive root step operators E α are represented as follows:
HA
ǫA − ǫB
E
E ǫA + ǫB
= eA,A − eA+ℓ,A+ℓ ,
= eB,A − eA+ℓ,B+ℓ ,
= eA+ℓ,B − eB+ℓ,A .
(9.6)
The solvable algebra of the maximally split coset
M(ℓ,0) =
SO(ℓ, ℓ)
SO(ℓ) × SO(ℓ)
(9.7)
has therefore a very simple form in terms of matrices. Following the general constructive
principles Solv(ℓ,ℓ) is just the algebraic span of all the matrices (9.6) so that
!
(
T
B
T = upper triangular ,
;
(9.8)
Solv(ℓ,ℓ) ∋ M ⇔ M =
B = − B T antisymmetric.
0 −T T
The matrices of the form (9.8) clearly form a subalgebra of the so(ℓ, ℓ) algebra which, in this
representation, is defined as the set of matrices Λ fulfilling the following condition:
!
!
0
0
1
1
l
l
ΛT
+
Λ = 0.
(9.9)
1l 0
1l 0
9.2
The real form so(r, r + 2s) of the Dr+s Lie algebra
The main point in order to apply to the coset manifold (9.1) the general integration algorithm
of the Lax equation devised for the case SL(2ℓ)/SO(2ℓ) consists of introducing a convenient
51
basis of generators of the Lie algebra so(r, r + 2s) where, in the fundamental representation,
all elements of the solvable Lie algebra associated with the coset under study turn out to
be given by upper triangular matrices. With some ingenuity such a basis can be found
by defining the so(r, r + 2s) Lie algebra as the set of matrices Λt satisfying the following
constraint:
ΛTt ηt + ηt Λt = 0
(9.10)
where the symmetric invariant metric ηt with r + 2s positive eigenvalues (+1) and r negative
ones (−1) is given by the following matrix.
0 ̟r
0
.
(9.11)
ηt =
0
1
0
2s
̟r 0 0
In the above equation the symbol ̟r denotes the completely anti-diagonal r × r matrix
which follows:
0 0 ... ... 0 1
0 0 . . . . . . 1 0
0 0 . . . 1 0 0
r.
̟r =
(9.12)
. . . . . . . . . . . . . . . . . .
0 1 0 . . . . . . 0
1 0 0 ... ... 0
|
{z
}
r
Obviously there is a simple orthogonal transformation which maps the metric ηt into the
standard block diagonal metric ηb written below
0
1r 0
.
(9.13)
ηb =
0
1
0
2s
0 0 −1r
Indeed we can write
ΩT ηb Ω = ηt
where the explicit form of the matrix Ω is the following:
0
12s
0
1
√1 ̟r
Ω =
√2 1r 0
2
√1 1r
√1 ̟r
0
−
2
2
(9.14)
.
(9.15)
Correspondingly the orthogonal transformation Ω maps the Lie algebra and group elements
of so(r, r + 2s) from the standard basis where the invariant metric is ηb to the basis where
it is ηt
Λt = ΩT Λb Ω .
(9.16)
52
In the t-basis the general form of an element of the solvable Lie algebra which generates the
coset manifold (9.1) has the following appearance:
B
T X
SO(r, r + 2s)
T
∋ Λt =
(9.17)
Solv
0
0
X
̟
r
SO(r) × SO(r + 2s)
0 0 −̟r T T ̟r
where
T =
T1,1 T1,2
0
T2,2
0
0
...
...
...
T1,r−1
...
...
T2,r−1
T3,3 . . .
...
...
...
...
0
0
0
. . . Tr−1,r−1
0
0
0
...
...
...
B = −B T antisymmetric r × r ,
X = arbitrary r × 2s
T1,r
T2,r
T3,r
...
Tr−1,r
Tr,r
upper triangular r × r ,
(9.18)
while an element of the maximal compact subalgebra has instead the following appearance:
Y
C ̟r
Z
T
T
(9.19)
so(r) ⊕ so(r + 2s) ∋ Λt =
−Y
Q
−
Y
̟
r
̟r C ̟r Y −̟r Z T ̟r
where
Z
C
Q
Y
=
=
=
=
− Z T antisymmetric r × r ,
− C T antisymmetric r × r ,
− QT antisymmetric 2s × 2s .
arbitrary r × 2s
(9.20)
Having clarified the structure of the matrices representing Lie algebra elements in this basis
well adapted to the Tits Satake projection, we can now discuss a basis of generators also
well adapted to the same projection. To this effect, let us denote by Iij the r × r matrices
whose only non vanishing entry is the ij-th one which is equal to 1
0 0 ...
...
... 0
0 0 ...
...
... 0
... ... ...
...
... ...
(9.21)
Iij = 0 0 . . .
1
. . . 0 } i-th row
.
0 0 ...
...
... 0
.
.
.
.
.
.
0
0 0 ...
|{z}
j-th column
53
Using this notation the r non-compact Cartan generators are given by
0
Iii 0
; (i = 1, . . . , r) .
Hi = 0 0
0
0 0 −̟r Iii ̟r
(9.22)
Next we introduce the coset generators associated with the long roots of type: α = ǫi − ǫj .
Iij + Iji 0
0
α = ǫi − ǫj
⇒ K−ij = √12 E α + E −α = √12
0
0
0
i < j = 1, . . . , r
0
0 −̟r (Iij + Iji ) ̟r
(9.23)
and the coset generators associated with the long roots of type α = ǫi + ǫj :
0 (Iij − Iji ) ̟r
0
α = ǫi + ǫj
ij
α
−α
1
1
= √2
⇒ K+ = √ 2 E + E
0
0
0
i < j = 1, . . . , r
̟r (Iji − Iij ) 0
0
.
(9.24)
The short roots, after the Tits-Satake projection, are just r, namely ǫi . Each of them,
however, appears with multiplicity 2s, due to the paint group. We introduce a 2s-tuple
of coset generators associated to each of the short roots in such a way that such 2s-tuple
transforms in the fundamental representation of Gpaint = so(2s). To this effect let us define
the rectangular r × 2s matricesJim analogous to the square matrices Iij , namely
0 0 ...
...
... ... ... 0
0 0 ...
...
... ... ... 0
... ... ...
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
Jim = 0 0 . . .
.
(9.25)
1
. . . . . . . . . 0 } i-th row
0 0 ...
...
... ... ... 0
0 0 ...
.
.
.
.
.
.
.
.
.
0
.
.
.
|{z}
m-th column
Then we introduce the following coset generators:
Jim
0
0
α = ǫi
i
T
T
⇒ Km
= √12
i = 1, . . . , r
0
−Jim
̟r
Jim
m = 1, . . . , 2s
0 −̟r Jim
0
.
(9.26)
The remaining generators of the so(r, r +2s) algebra are all compact and span the subalgebra
so(r) ⊕ so(r + 2s) ⊂ so(r, r + 2s). According to the nomenclature of eq.(9.19) we introduce
54
four sets of generators. The first set is associated
is defined as follows:
Iij − Iji
α
−α
ij
1
1
= √2
Z = √2 E − E
0
0
with the long roots of type α = ǫi − ǫj and
0
0
0
0
0 −̟r (Iij − Iji ) ̟r
The second set is associated with the long roots of type
0
α
−α
ij
1
1
= √2
C = √2 E − E
0
−̟r (Iji − Iij )
.
(9.27)
α = ǫi + ǫj and is defined as follows:
0 (Iij − Iji ) ̟r
.
(9.28)
0
0
0
0
The third group of compact generators spans the compact coset
SO(r + 2s)
SO(r) × SO(2s)
and it is given by
Ymi
=
√1
2
(9.29)
0
Jim
0
T
−J T
0
−Jim
̟r
im
.
0
̟r Jim
0
(9.30)
The fourth set of compact generators spans the paint group Lie algebra so(2s) and is given
by
0
0
0
Qmn =
(9.31)
0
Q
−
Q
0
mn
nm
0
0
0
where Qmn denotes the analogue of the Iij in 2s rather than in r dimensions.
By performing the change of basis to the block diagonal form of the matrices we can
verify that Cij − Zij generate the so(r) subalgebra while Cij + Zij together with Qmn and
Yim generate the subalgebra so(r + 2s).
The full set of generators is ordered in the following way:
ij
ij
i
ij
ij
i
(9.32)
TΛ =
Hi , K− , K+ , Km , |{z}
Z
, |{z}
C
, Ym , Qmn
|{z}
|{z} |{z}
|{z} |{z}
|{z}
1
1
r
2rs
2rs
s(2s−1)
1
1
r(r−1)
r(r−1)
2
r(r−1)
2
r(r−1)
2
2
and satisfy the trace relation:
Tr (TΛ TΣ ) = gΛΣ ,
gΛΣ = 2 diag +, +, . . . , +, −, −, . . . , − .
{z
} |
{z
}
|
r(r+2s)
55
r 2 −r+2rs+2s2 −s
(9.33)
In this way we have obtained the needed and detailed construction of the embedding (6.1)
which is necessary to apply the integration algorithm. In the next section we make a detailed
study of the case r = 2, s = 1.
10
A case study for the Tits Satake projection: SO(2, 4)
The simplest example of not maximally split manifold inside the series defined by eq.(9.1)
corresponds to the choice: r = 2, s = 1, namely
SO(2, 4)
.
SO(2) × SO(4)
M2,2 ≡
(10.1)
The Tits Satake projection yields the manifold studied at length in section 8
ΠT S :
SO(2, 4)
SO(2, 3)
Sp(4, R)
7→
∼
SO(2) × SO(4)
SO(2) × SO(3)
U(2)
(10.2)
and the paint group is the simplest possible group
Gpaint = SO(2) .
(10.3)
This manifold will be the target of our case study in order to illustrate the bearing of the
Tits Satake projection and the features of the Tits Satake universality classes.
Following the discussion of section 9 we can organize the roots in a well adapted way for
the Tits Satake projection and introduce a basis where the solvable Lie algebra of the coset
is represented by upper triangular matrices.
The root system associated with so(2, 4) is actually that of D3 ∼ A3 described by the
Dynkin diagram which follows:
D3
✐
✐
✐
α1
α2
α3
(10.4)
There are 6 positive roots that are vectors in R3 and can be organized as it follows:
Π
TS
α1,1 = ǫ2 − ǫ3 −→
ǫ2
ΠT S
α1,2 = ǫ2 + ǫ3 −→ ǫ2
ΠT S
≡ α1 ,
≡ α1 ,
α2 = ǫ1 − ǫ2 −→ ǫ1 − ǫ2 ≡ α2 ,
Π
TS
α3,1 = ǫ1 − ǫ3 −→
ǫ1
Π
TS
α3,2 = ǫ1 + ǫ3 −→
ǫ1
ΠT S
≡ α1 + α2 ,
≡ α1 + α2 ,
α4 = ǫ1 + ǫ2 −→ ǫ1 + ǫ2 ≡ 2α1 + α2 .
56
(10.5)
In the above formulae the last three columns describe the Tits-Satake projection of the root
system which, in this case, is simply given by the geometrical projection of the three–vectors
onto the plane {12}. In this way the correspondence with the sp(4, R) root system becomes
explicit (compare with fig.9). We have 2s = 2 preimages of each of the short roots α1 and
α1 + α2 while the long roots α2 and 2α1 + α2 have a single preimage. In complete analogy
with eq.(8.20) we can define the appropriate basis for the realization of the considered Lie
algebra by giving the explicit expression of the most general element of the solvable Lie
algebra Solv (SO(2, 4)/SO(2) × SO(4)). Abstractly this is given by
T
= h1 H1 + h2 H2 + e1,1 E α1,1 + e1,2 E α1,2
+ e2 E α2 + e3,1 E α3,1 + e3,2 E α3,2 + e4 E α4 .
(10.6)
We define the form of all Cartan and step generators by writing the same Lie algebra element
(10.6) as a 6 × 6 matrix
√
√
√
√
h1 + h2 − 2e2 − 2e3,1 − 2e3,2 − 2e4 0
√
√
√
0
h
−
h
−
2e
−
2e
0
2e
1
2
1,1
1,2
4
√
√
0
0
0
0
2e
2e
1,1
3,1
(10.7)
T =
√
√
0
0
0
0
2e
2e
1,2
3,2
√
0
0
0
0
h
−
h
2e
2
1
2
0
0
0
0
0
−h1 − h2
which satisfies the condition (9.10) with the metric ηt defined in eq.(9.11).
Then in full analogy with eq.s (8.22,8.23) we can construct a basis for the subspace K
and for the subalgebra H by writing
and
K1 = H1 ,
K2 = H2 ,
K3 = √12 E α1,1 + (E α1,1 )T
K4 = √12 E α1,2 + (E α1,2 )T
K5 = √12 E α2 + (E α2 )T ,
K6 = √12 E α3,1 + (E α3,1 )T
K7 = √12 E α3,2 + (E α3,2 )T
K8 = √12 E α4 + (E α4 )T
J1 =
J2 =
J3 =
J4 =
J5 =
J6 =
√1
2
√1
2
√1
2
√1
2
1
√
2
1
√
2
E α1,1 − (E α1,1 )T
E
α1,2
− (E
α1,2 T
E α2 − (E α2 )
)
T
E α3,1 − (E α3,1 )
,
T
T
E α3,2 − (E α3,2 )
E α4 − (E α4 )T .
57
,
,
,
,
(10.8)
,
,
,
(10.9)
In this way we have constructed 8 + 6 = 14 generators. One is still missing to complete a 15dimensional basis for the Lie algebra so(2, 4). The missing item is Q, namely the generator
of the paint group SO(2)
0 0 0 0 0 0
0 0 0 0 0 0
0 0 0 1 0 0
.
(10.10)
Q=
0 0 −1 0 0 0
0 0 0 0 0 0
0 0
0
0 0 0
Naively one might think that the six generators Ji defined in (10.9) close the Lie algebra of so(4), while Q generates the factor so(2) in the denominator group of our manifold
SO(2,4)
. From the discussion of the previous section 9 we know that this is not the
SO(2)×SO(4)
case. Indeed the paint group is inside the factor SO(r + 2s) so that the listed Ji constitute
a tangent basis for the coset manifold
SO(4)
SO(2)paint
Pe = SO(2) ×
(10.11)
which is the universal covering of the true parameter space for the integration of our Lax
e by modding out the generalized Weyl group as
equation. The actual P is obtained from P
stated in equation (5.8).
Having established these notations we can just proceed to the construction of the initial
data in the usual way. The Cartan subalgebra element is given by the following 6 × 6 matrix:
0 0
0
0
h1 + h2
0
0
h
−
h
0
0
0
0
1
2
0
0
0
0
0
0
(10.12)
C =
0
0
0
0
0
0
0
0
0
0
−h
+
h
0
1
2
0
0
0 0
0
−h1 − h2
while the orthogonal matrix O ∈ P can be defined in complete analogy to eq.(8.24):
O (θ1 , . . . , θ6 ) =
i
h√
i
h√
exp 2 θ1 J1 exp 2 θ2 J2 exp [θ3 J3 ]
O
O12 O13
11
O
21 O22 O23
O31 O32 O33
=
O
41 O42 O43
O51 O52 O53
O61 O62 O63
58
exp
O14
O24
O34
O44
O54
h√
i
h√
i
2 θ4 J4 exp 2 θ5 J5 exp [θ6 J6 ]
O15 O16
O25 O26
O35 O36
.
(10.13)
O45 O46
O55 O56
O64 O65 O66
We do not write the explicit functional form of the 36 entries because it takes too much
space yet it is clear that they are uniquely defined by the above equation and by the explicit
form of the generators. We just go over to discuss the Weyl group.
10.1
The generalized Weyl group for SO(2, 4)
Applying the procedure of definition 5.1, we introduce six generators for the generalized Weyl
group corresponding to the reflections with respect to the 6 roots. These can be represented
as the rotation matrices
π
(10.14)
γi = O 0, . . . , 0, , 0, . . . , 0 ; i = 1, . . . , 6
| {z } 2 | {z }
i−1
6−i
which are integer valued. Considering all products and all relations among these generators
we obtain the finite group W (so(2, 4)) which has 32 elements. The group W (so(2, 4)) has
a normal subgroup
Z2 × Z2 ∼ N (so(2, 4)) ⊂ W (so(2, 4))
(10.15)
given by the following
1 0
0 1
0 0
n1 =
0 0
0 0
0 0
n3
=
four diagonal matrices:
0 0 0 0
0 0 0 0
1 0 0 0
; n2 =
0 1 0 0
0 0 1 0
0 0 0 1
1 0
0
0
0 0
0 1
0
0
0 0
0
0 0
0 0 −1
0 0
0
0 0
0
0 0
0
−1 0 0
0
1 0
0
0 1
−1
0
0
0
0 0
0
−1 0 0
0
0
1
0
0
0
0
0
0
0
0
0
0
0
0 0
0
,
1 0
0
0 −1 0
0 0 −1
−1 0
0
0
0
0
0 −1 0
0
0
0
0 −1 0
0
0
; n4 = 0
0
0 −1 0
0
0
0
0
0
0 −1 0
0
0
0
0
0 −1
.
(10.16)
The normal subgroup N (so(2, 4)) when acting by similarity transformation on a Cartan
subalgebra element of the form (10.12) leaves it invariant
∀ n ∈ N (so(2, 4)) : nT C n = C .
(10.17)
W (so(2, 4))
≃ Weyl (sp(4)) .
N (so(2, 4))
(10.18)
The order 8 factor group obtained by modding W (so(2, 4)) with respect to N (so(2, 4)) is
isomorphic to the Weyl group of the Tits Satake subalgebra Weyl (sp(4)) and has the same
action on the eigenvalues h1 , h2
59
A representative for each of the eight equivalence classes
1 0 0 0 0 0
0 1 0 0 0 0
0 0 1 0 0 0
;
Λ
=
Λ1 =
2
0 0 0 1 0 0
0 0 0 0 1 0
0 0 0 0 0 1
1
0 0
0 0
0
1
0
0 −1 0 0
0
0
0 0
1
0
0
0
; Λ4 = 0
Λ3 =
0
0 0
0 −1 0
0
0
0 0
0
0
−1
0
Λ5 =
1 0
Λ7 =
0
0 0
0
0
1 0 0 0 0
0
0
−1
0
0
0
0
0
1
0
−1 0 0 0 0 0
0
0 1 0 0
0
0 0 1 0
0
0 0 0 0
0
0 0 0 1
0 0 0 0
1
1 0 0 0
0
0 0 1 0
0
0 0 0 −1 0
0 0 0 0
0
0 1 0 0
0
; Λ6 =
; Λ8 =
can be easily written. We find
0 0 0 0 0 1
0 0 0 0 1 0
0 0 1 0 0 0
,
0 0 0 1 0 0
0 1 0 0 0 0
1 0 0 0 0 0
0
0 0
0
0 0
0
0
−1 0
0
1 0
0
0
,
0
0 −1 0
0
−1 0 0
0
0
0 0
0 0
0
1
0 1 0 0
0 0
0 0 0 0
0 1
0 0 1 0
0
0
,
0 0 0 −1 0 0
1 0 0 0
0
0
0 0 0 0
1 0
0
0 0 0 1 0
0
0
0
0
0
−1
0
0 1 0 0 0
.
0 0 1 0 0
0
−1 0 0 0 0 0
0
1 0 0 0 0
(10.19)
Assembling the information presented above we come to a stronger conclusion. Not only
the factor group is isomorphic to the Weyl group of the Tits Satake projection but even the
generalized Weyl group is isomorphic. Indeed we have found
W (so(2, 4)) ∼ W (sp(4)) .
(10.20)
We have not proved so far that this is true in general but it is an attractive conjecture to
postulate that
W (U) ∼ W (UTS ) .
(10.21)
We leave the proof of such a conjecture to future publications.
60
10.2
Vertices, edges and trapped surfaces
By means of a computer programme we can now study the vertices, the links, the critical
surfaces and the accessible vertices on each critical surface. We summarize the results.
Vertices Our parameter space is now a 6 dimensional hypercube that has 64 vertices and
192 edges. On each of the 64 vertices we find one of the 8 Weyl elements which obviously
reappears several times. Each of the odd elements Λ1,3,5,7 appears 4 times, while each of the
even elements Λ2,4,6,8 appears 12 times so that we have 4 × 4 + 4 × 12 = 64. The 64 vertices
with their Weyl element correspondence are listed in table 7.
Edges The one dimensional links connecting the 64 vertices are 192 and each of them
represents a flow from one lower Weyl element to a higher one. A priori we might expect
that the lines connecting the 8 Weyl elements could now be more numerous than in the case
of the Tits Satake projected manifold. However calculating all these links on a computer we
find that the independent lines are just 16 and the same 16 appearing in the Tits Satake
projection. This is made evident by the flow diagram displayed in fig. 19
L
6
L
4
L
2
L
5
L
8
L
1
L
3
L
7
Figure 19: The oriented phase diagram of the SO(2, 4)/SO(2) × SO(4) flows. The Lie algebra
so(2, 4) is not maximally split and its Tits Satake subalgebra is sp(4, R) ∼ so(2, 3) ⊂
so(2, 4). The relevant Weyl group is that of the Tits Satake subalgebra and the flow diagram
for the integration of the Lax equation on this space just coincides with that of sp(4, R). At
fixed value of the Cartan fields if h1 , h2 , −h1 , −h2 are the eigenvalues of the sp(4, R) Lax
operator, those of the so(2, 4) Lax operator are h1 + h2 , h1 − h2 , 0, 0, −h1 − h2 , −h1 + h2 .
Using {h1 + h2 , h1 − h2 } as coordinates to identify the Weyl element we obtain the presented
flow graph.
Trapped surfaces and accessible vertices The trapped hypersurfaces in parameter
space are obtained by equating to zero the minors obtained by intersecting the first k columns
61
of O with an equal number of arbitrarily chosen rows. In this way we generate a total of 62
trapped surfaces. They are enumerated as follows:
Order of the minor Number
5
6
4
15
3
20
2
15
1
6
(10.22)
62
We can now calculate the set of accessible Weyl elements for each of these 62 surfaces and
within the accessible set we can single out the lowest and the highest Weyl elements which
will correspond to the initial and final end points of the flows confined on that surface. The
result is displayed in table 8. Inspection of this list reveals that the available flows, although
repeated on many different surfaces are just a small set of five possibilities, exactly the same
five possible flows appearing in the the case of the Tits Satake projection that were shown
in eq.(8.33)
This concludes our discussion. As we have seen the vertices and the possible flows on
critical links or trapped hypersurfaces do not depend on the chosen representative within a
Tits Satake universality class rather they depend only on the class. In other words the study
of the maximally split Tits Satake projection already provides us with a complete picture
of all possible flows. It is only the detailed structure of bouncing which varies from one
representative to the other.
10.3
Examples of flows for SO(2, 4)
We come now to the analysis of two explicit examples of flows aiming at illustrating three
aspects:
a The embedding of the Tits Satake flows within the flows of the bigger coset manifold.
b The role of the extra parameters not contained in the Tits Satake projection.
c The instability of super-critical and in general of trapped surfaces.
To this effect we shall reconsider the case analyzed in section 8.4.3 of a flow on the critical
surface Σ2 for the Tits Satake projection of SO(2, 4), namely Sp(4, R) ∼ SO(2, 3).
The unperturbed super-critical flow The choice of the Euler angles is that of eq.(8.46)
which, in the five dimensional representation SO(2, 3) produces the matrix of eq.(8.47). This
latter has three vanishing minors, as shown in eq.(8.49). It is quite easy to embed this case
and the corresponding flow into the non maximally split representative SO(2, 4) of the same
universality class. It suffices to choose the six θi angles as follows:
nπ
π π
πo
{θ1 , θ2 , θ3 , θ4 , θ5 , θ6 } =
(10.23)
, 0, , , 0 ,
3
6 3
3
62
1)
2)
3)
4)
5)
6)
7)
8)
9)
; {0, 0, 0, 0, 0, 0} = Λ1
33) ; {0, 1, 1, 1, 0, 0} = Λ4
; {0, 1, 0, 0, 0, 0} = Λ5
35) ; {0, 1, 1, 0, 0, 1} = Λ8
; {1, 0, 0, 0, 0, 0} = Λ5
34) ; {0, 1, 1, 0, 1, 0} = Λ4
; {0, 0, 1, 0, 0, 0} = Λ4
36) ; {0, 1, 0, 1, 1, 0} = Λ5
; {0, 0, 0, 1, 0, 0} = Λ8
37) ; {0, 1, 0, 1, 0, 1} = Λ4
; {0, 0, 0, 0, 1, 0} = Λ8
38) ; {0, 1, 0, 0, 1, 1} = Λ4
; {0, 0, 0, 0, 0, 1} = Λ3
39) ; {0, 0, 1, 1, 1, 0} = Λ4
; {1, 1, 0, 0, 0, 0} = Λ1
40) ; {0, 0, 1, 1, 0, 1} = Λ8
; {1, 0, 1, 0, 0, 0} = Λ6
41) ; {0, 0, 1, 0, 1, 1} = Λ8
10) ; {1, 0, 0, 1, 0, 0} = Λ2
42) ; {0, 0, 0, 1, 1, 1} = Λ3
11) ; {1, 0, 0, 0, 1, 0} = Λ2
43) ; {1, 1, 1, 1, 0, 0} = Λ6
12) ; {1, 0, 0, 0, 0, 1} = Λ7
44) ; {1, 1, 1, 0, 1, 0} = Λ6
13) ; {0, 1, 1, 0, 0, 0} = Λ6
45) ; {1, 1, 1, 0, 0, 1} = Λ2
14) ; {0, 1, 0, 1, 0, 0} = Λ2
46) ; {1, 1, 0, 1, 1, 0} = Λ1
15) ; {0, 1, 0, 0, 1, 0} = Λ2
47) ; {1, 1, 0, 1, 0, 1} = Λ6
16) ; {0, 1, 0, 0, 0, 1} = Λ7
48) ; {1, 1, 0, 0, 1, 1} = Λ6
17) ; {0, 0, 1, 1, 0, 0} = Λ6
49) ; {1, 0, 1, 1, 1, 0} = Λ6
18) ; {0, 0, 1, 0, 1, 0} = Λ6
50) ; {1, 0, 1, 1, 0, 1} = Λ2
19) ; {0, 0, 1, 0, 0, 1} = Λ2
51) ; {1, 0, 1, 0, 1, 1} = Λ2
20) ; {0, 0, 0, 1, 1, 0} = Λ1
52) ; {1, 0, 0, 1, 1, 1} = Λ7
21) ; {0, 0, 0, 1, 0, 1} = Λ6
53) ; {0, 1, 1, 1, 1, 0} = Λ6
22) ; {0, 0, 0, 0, 1, 1} = Λ6
54) ; {0, 1, 1, 1, 0, 1} = Λ2
23) ; {1, 1, 1, 0, 0, 0} = Λ4
55) ; {0, 1, 1, 0, 1, 1} = Λ2
24) ; {1, 1, 0, 1, 0, 0} = Λ8
56) ; {0, 1, 0, 1, 1, 1} = Λ7
25) ; {1, 1, 0, 0, 1, 0} = Λ8
57) ; {0, 0, 1, 1, 1, 1} = Λ2
26) ; {1, 1, 0, 0, 0, 1} = Λ3
58) ; {1, 1, 1, 1, 1, 0} = Λ4
27) ; {1, 0, 1, 1, 0, 0} = Λ4
59) ; {1, 1, 1, 1, 0, 1} = Λ8
28) ; {1, 0, 1, 0, 1, 0} = Λ4
60) ; {1, 1, 1, 0, 1, 1} = Λ8
29) ; {1, 0, 1, 0, 0, 1} = Λ8
61) ; {1, 1, 0, 1, 1, 1} = Λ3
30) ; {1, 0, 0, 1, 1, 0} = Λ5
62) ; {1, 0, 1, 1, 1, 1} = Λ8
31) ; {1, 0, 0, 1, 0, 1} = Λ4
63) ; {0, 1, 1, 1, 1, 1} = Λ8
32) ; {1, 0, 0, 0, 1, 1} = Λ4
64) ; {1, 1, 1, 1, 1, 1} = Λ2
Table 7: Vertices/Weyl group correspondence for the case SO(2, 4).
63
Σ1
Σ2
Σ3
Σ4
Σ5
Σ6
Σ7
Σ8
Σ9
{w6 , w8 }
Σ32 {w5 , w8 }
{w5 , w8 }
Σ34 {w5 , w2 }
{w5 , w8 }
Σ33 {w5 , w8 }
{w5 , w8 }
Σ35 {w5 , w8 }
{w5 , w8 }
Σ36 {w5 , w8 }
{w5 , w7 }
Σ37 {w5 , w8 }
{w5 , w8 }
Σ38 {w5 , w8 }
{w5 , w8 }
Σ39 {w5 , w8 }
{w5 , w8 }
Σ40 {w5 , w8 }
Σ10 {w5 , w8 }
Σ41 {w5 , w8 }
Σ11 {w5 , w8 }
Σ42 {w5 , w8 }
Σ13 {w1 , w8 }
Σ44 {w5 , w8 }
Σ12 {w5 , w8 }
Σ43 {w5 , w8 }
Σ14 {w5 , w8 }
Σ45 {w1 , w8 }
Σ15 {w5 , w8 }
Σ46 {w5 , w8 }
Σ16 {w5 , w8 }
Σ47 {w5 , w8 }
Σ17 {w5 , w8 }
Σ48 {w5 , w8 }
Σ18 {w5 , w2 }
Σ49 {w5 , w8 }
Σ19 {w5 , w8 }
Σ50 {w5 , w2 }
Σ20 {w5 , w8 }
Σ51 {w5 , w8 }
Σ21 {w5 , w8 }
Σ52 {w5 , w8 }
Σ22 {w5 , w8 }
Σ53 {w5 , w8 }
Σ23 {w5 , w8 }
Σ54 {w5 , w8 }
Σ24 {w5 , w8 }
Σ55 {w5 , w8 }
Σ25 {w5 , w8 }
Σ56 {w5 , w8 }
Σ26 {w5 , w8 }
Σ57 {w6 , w8 }
Σ27 {w1 , w8 }
Σ58 {w5 , w8 }
Σ28 {w5 , w8 }
Σ59 {w5 , w8 }
Σ29 {w5 , w8 }
Σ60 {w5 , w8 }
Σ30 {w5 , w8 }
Σ61 {w5 , w8 }
Σ31 {w5 , w8 }
Σ62 {w5 , w7 }
Table 8: Initial and final endpoints of flows confined on trapped surfaces for the case SO(2, 4).
64
since, according to eq.(10.5) and (10.9), the angles θ2 and θ5 correspond to the second copy
of the compact generators respectively associated with the first and the third of the sp(4)
roots. The result of this choice is the following matrix in SO(2) × SO(4) ⊂ SO(2, 4):
√
√
5
3
7 3
3
3
√
−
−
0
−
16
16
16
√4 2
16
√
√
3 32
19
11 3
5 3
3
0
−
−
−√32
32
8
32
32
√
3
3
3 2
3 2
15
15
1
√
√
0 16 2
−8
16
16 2
Ounp = 16
(10.24)
.
0
0
0
1
0
0
√3
√
√
3
5 3
11 3
19
−3
2
−
0
−
32√
32
8
32
√ 32
3
3
3
5
7 3
√
− 4 2 0 16
− 16 − 16
16
As one sees, by deleting the 4th row and the 4th column one retrieves the matrix of eq.(8.47).
Indeed the matrix (10.24) is manifestly inside the Tits Satake subgroup SO(2, 3) ⊂ SO(2, 4).
If we use Ounp as initial data for our integration algorithm implemented on a computer we
find that the asymptotic limits are Λ1 at past infinity and Λ8 at future infinity just as in the
original case discussed in section 8.4.3. Consider fig.20. It displays the plot of the Cartan
fields projected along the root α1 and clearly demonstrates that there are just two bounces
on the wall orthogonal to this root. Both of them occur in a narrow time range around t = 0.
At very early and very late times there are no more bounces and the result is the trajectory
of the cosmic ball displayed in fig.21. The two asymptotic lines (incoming and outgoing) are
the Kasner epochs Λ1 and Λ8 , respectively.
-6
-4
-2
2
-6.9
-2
-2
-1
1
2
-7.1
-4
-7.2
-7.3
-6
-7.4
-7.5
-8
-7.6
Figure 20: Plot of the α1 · h projection
for the SO(2, 4) flow generated by the parameter
choice {θ1 , θ2 , θ3 , θ4 , θ5 , θ6 } = π3 , 0, π6 , π3 , 0 , π3 . This is actually a flow in the Tits Satake
submanifold and corresponds to a super-critical surface. This super-critical flow connects the
primordial Kasner era Λ1 to the remote future Kasner era Λ8 . The plot on the left and on
the right are the same. The only difference is that on the right we have an enlargement of
the time region around t = 0, while on the left we consider a time range covering a much
wider portion of the early epochs.
65
h2
14
13
12
11
h1
4
3
5
7
6
Figure 21: Trajectory of the
cosmic ball in the SO(2, 4) flow generated by the parameter choice
{θ1 , θ2 , θ3 , θ4 , θ5 , θ6 } = π3 , 0, π6 , π3 , 0 , π3 . This flow is inside the Tits Satake submanifold
and corresponds to a super-critical surface. It connects the primordial Kasner era Λ1 to the
remote future Kasner era Λ8 .
Perturbing the super-critical flow with painted walls In order to illustrate both the
nature of the Tits Satake projection and the instability of super-critical surfaces we consider
now a small perturbation of the initial data used in the previous example. Keeping all the
other angles unchanged we shift from zero the angle θ2 by a very small amount. Explicitly
we choose
π
π
π
π
1
θ1 = , θ2 = arcsin
, θ3 = , θ4 = , θ5 = 0 , θ6 =
.
(10.25)
3
100
6
3
3
As explained θ2 is associated with the second copy of the root α1 . Hence introducing this
small angle is equivalent to creating a new α1 wall just painted with a different color. This
new wall is very very small and therefore it will produce very little effects at finite times.
Yet it is sufficient to remove us from the super-critical surface and this necessarily changes
asymptotics. Instead of Λ1 we expect now Λ5 at past infinity. It is interesting to analyze in
detail how this happens.
If we name Opert the matrix corresponding to the choice of angles (10.25) we can appreciate the perturbation of initial data by writing Opert in the following way:
0
0
0
0
0
0
0
0
0
−6
0
0
√
0
0
0
6 6 0
0
√
√
Opert = Ounp + ǫ1 15
√
9 3
9 3
15
− 2 −2
−2 − 2 3 6 0
0
0
0
−6 0
0
0
0
66
0
0
0
0
+ ǫ2
0
0
5
− 16
√3
− 3163
5
√
9
√
8 2
2
8
0
0
5
− 16
√
− 3163
0
0
0
√3
0 0
0
√
5
0 − 3163 − 16
√3
5 2
0 8√9 2
8
2
4
− 43
0
√3
2 0
0
2
4
0
0
√
− 3163
0 0
ǫ1 ≃ 1.2 × 10−3 ,
ǫ2 ≃ 1.0 × 10−4 .
5
− 16
0
,
(10.26)
Then we can implement the integration algorithm on our computer and calculate the asymptotic values of the Lax operator. Notwithstanding the smallness of the perturbation, the past
infinity regime jumps from Λ1 to Λ5 as expected, while at future infinity it remains Λ8 which
is already the highest possible Weyl element. We can appreciate the mechanism which realizes this effect by looking at fig.s 22 and 23. There is an extra bounce on the α1 wall as
2
-1.5
1
-1.6
-1.7
-6
-4
-2
2
-1.8
-1
-1.9
-2
-1.5
-1
-0.5
0.5
1
1.5
-2.1
-3
-2.2
Figure 22: Plot of the α1 · h projection
for the
4) flow generated by the parameter
π
SO(2,
1
π π
choice {θ1 , θ2 , θ3 , θ4 , θ5 , θ6 } = 3 , arcsin 100 , 6 , 3 , 0 , π3 . This flow is a perturbation of
a super-critical flow. The shift from Λ1 to Λ5 at past infinity occurs via an extra bump on
the α1 wall which occurs at very early times. This bump is not visible in the plot on the right
which is in the range near t = 0 but it is evident in the plot on the left which goes further
back in time. This picture is to be compared with fig.20.
we expected which corrects the trajectory and directs the cosmic ball to Λ5 rather than Λ1
when we go back in time. Since the perturbation is small this bounce occurs at very early
times so that for most of the time the flow is almost on the critical surface. The smaller the
perturbation, the earlier the occurrence of the primeval bounce. It should also be noted that
we would have obtained exactly the same effect if we had perturbed the θ1 angle instead
of the θ2 . Indeed they are associated with the same root. This is the meaning of the Tits
Satake projection which captures all the essential features of the dynamical processes for the
entire universality class.
67
h2
12
11
10
h1
7
8
9
10
h2
12
10
8
6
4
2
h1
2
4
6
8
10
-2
Figure 23: Comparison of the trajectories of the cosmic
ball in the SO(2, 4) flow generated
by the parameter choice {θ1 , θ2 , θ3 , θ4 , θ5 , θ6 } = π3 , 0, π6 , π3 , 0 , π3 and in its perturbation
1
by a small θ2 = arcsin 100
. The thin line is the unperturbed flow, the fatter line is the
perturbed one. The first plot covers a time range around t = 0 while the second plot extends
much earlier in time. The additional bounce responsible for the changing of asymptotic is
visible in the second plot.
Part III
Perspectives
11
Summary of results
In this paper we have made a few steps forward in developing the general programme of
supergravity billiards as a paradigm for superstring cosmology. Our results are both of
physical and mathematical nature.
On the physical side, which for us means supergravity/superstring theory, the essential
points are the following ones:
68
1) We have shown that all supergravity billiards are completely integrable, irrespectively
whether they are defined on a maximally split coset manifold U/H as it happens in
the case of maximal susy or a non maximally split U/H, as it happens in all lower
supersymmetry cases. We have provided the explicit integration algorithm which just
depends on the triangular embedding of the solvable Lie algebra Solv(U/H) into that
of Solv(SL(N)/SO(N)).
2) We have discovered a new principle of time orientation of the cosmic flow which relies on
the natural ordering of the Weyl group elements (or of the permutations) according to
their length ℓT in terms of transpositions. Cosmic evolution is always in the direction of
increasing ℓT which plays the role of an entropy. There is a fascinating similarity, in this
context between the laws of cosmic evolution and those of black hole thermodynamics.
3) We have clarified the meaning of Tits Satake universality classes, introduced in [18], at
least from the vantage point of cosmic billiards. The asymptotic states, the type of
available flows and the critical surfaces in parameter space are properties of the class
and do not depend on the representative manifold in the class.
On the mathematical side the highlights of our paper are the following ones:
1) We have introduced the notion of generalized Weyl group for a non compact symmetric
space U/H and shown that the factor group with respect to its normal subgroup is
just the Weyl group of the Tits Satake subalgebra UT S ⊂ U. Moreover, we have
demonstrated that not only the factor group is isomorphic to the Weyl group of the
Tits Satake projection but even the generalized Weyl group is also isomorphic W (U) ∼
W (UTS ). At least this is true in the considered examples and we make the conjecture
that it is true in general.
2) We have established a remarkable conjecture encoded in property 5.1 of the main text:
the constraints on minors of the diagonalizing orthogonal matrix for the Lax operator
commute with the Toda flow.
3) We have proposed a very simple efficient method of calculating the Toda flow asymptotics
at t = ±∞ for the Lax operator of a σ-model with target space any non compactsymmetric coset space U/H. Our algorithm requires only the knowledge of the corresponding Weyl group Weyl(U) as well as that of the small group H.
4) We have posed the question how the equations cutting out algebraic loci in compact
group or coset manifolds and defined in terms of vanishing minors in the defining
representation can be lifted to the abstract group level and extended to all irreducible
representations.
12
Open problems and directions to be pursued
The results we have obtained are just steps ahead in a programme to be further developed.
They have solved some standing problems and opened new directions of investigations which
seem to us quite exciting. We just mention, as conclusion, the milestones we would like to
69
attain in the near future, evaluated from the view-point which was generated by our present
results:
1) Construction of all the triangular embeddings for all the solvable Lie algebras of all
supergravity models and corresponding construction of the integration algorithm for
all special homogeneous geometries.
2) Oxidation and physical interpretation of the Toda flows we have just shown how to
construct within supergravity models as those coming from string compactifications
on manifolds of restricted holonomy.
3) Extension of the integration algorithm to sigma models with a potential emerging from
flux compactifications and gauged supergravities in higher dimensions.
4) Study of the integration algorithm in affine and hyperbolic Kač-Moody extensions of
the symmetric space U/H, as they emerge from stepping down to D = 2 and D = 1
dimensions.
5) Comparison between the law of increasing ℓT and the second principle of black hole
mechanics in search of an adequate formulation of cosmological thermodynamics and
of a possible mapping between the attractor mechanism in black hole physics and Toda
flows in cosmic billiards.
6) More in depth study of the topology of parameter space H/W(U) and in particular its
partition in complex hulls that admit the trapped surfaces as walls. In this context an
exciting open question is whether these hulls are completely closed or whether there
is the possibility of going from one to the other avoiding the trapped walls. Clearly
if the answer to this question is no then we have the notion of parallel disconnected
universes.
Acknowledgements The authors acknowledge inspiring discussions and conversations with
their good friend and frequent collaborator Mario Trigiante.
70
References
[1] Linde A. D., 1990, Particle Physics and Inflationary Cosmology (Switzerland: Harwood Academic); S. Kachru, R. Kallosh, A. Linde, J. Maldacena, L. McAllister and
S. P. Trivedi, Towards inflation in string theory, hep-th/0308055; S. Kachru, R. Kallosh,
A. Linde and S. P. Trivedi, De Sitter vacua in string theory, Phys. Rev. D 68 (2003)
046005, hep-th/0301240; C. P. Burgess, R. Kallosh and F. Quevedo, de Sitter String
Vacua from Supersymmetric D-terms, :hep-th/0309187; M. Gutperle and A. Strominger, Spacelike branes, JHEP 0204 (2002) 018, hep-th/0202210; V. D. Ivashchuk and
V. N. Melnikov, Multidimensional classical and quantum cosmology with intersecting pbranes, J. Math. Phys. 39 (1998) 2866, :hep-th/9708157; L. Cornalba, M. S. Costa and
C. Kounnas, A resolution of the cosmological singularity with orientifolds, Nucl. Phys.
B 637 (2002) 378, hep-th/0204261; L. Cornalba and M. S. Costa, On the classical stability of orientifold cosmologies, Class. Quant. Grav. 20 (2003) 3969, hep-th/0302137.
F. Leblond and A. W. Peet, A note on the singularity theorem for supergravity SDbranes, hep-th/0305059; M. Kruczenski, R. C. Myers and A. W. Peet, Supergravity Sbranes, JHEP 0205 (2002) 039, hep-th/0204144; N. Ohta, Accelerating cosmologies from
S-branes, Phys. Rev. Lett. 91 (2003) 061303, hep-th/0303238; R. Emparan and J. Garriga, A note on accelerating cosmologies from compactifications and S-branes, JHEP
0305 (2003) 028, hep-th/0304124; A. Buchel and J. Walcher, Comments on supergravity description of S-branes, JHEP 0305 (2003) 069, hep-th/0305055; G. Papadopoulos,
J. G. Russo and A. A. Tseytlin, Solvable model of strings in a time-dependent plane-wave
background, Class. Quant. Grav. 20 (2003) 969, hep-th/0211289; F. Quevedo, Lectures
on string / brane cosmology, hep-th/0210292; M. Gasperini and G. Veneziano, The
pre-big bang scenario in string cosmology, hep-th/0207130; B. Craps, D. Kutasov and
G. Rajesh, String propagation in the presence of cosmological singularities, JHEP 0206,
053 (2002), hep-th/0205101; T. Banks and W. Fischler, M-theory observables for cosmological space-times, hep-th/0102077; J. Khoury, B. A. Ovrut, N. Seiberg, P. J. Steinhardt and N. Turok, From big crunch to big bang, Phys. Rev. D 65, 086007 (2002),
hep-th/0108187; J. E. Lidsey, D. Wands and E. J. Copeland, Superstring cosmology,
Phys. Rept. 337, 343 (2000), hep-th/9909061; A. E. Lawrence and E. J. Martinec,
String field theory in curved spacetime and the resolution of spacelike singularities, Class.
Quant. Grav. 13, 63 (1996), hep-th/9509149.
[2] J. Demaret, M. Henneaux, P. Spindel, Nonoscillatory behavior in vacuum Kaluza-Klein
cosmologies, Phys.Lett.B164:27-30 (1985); J. Demaret, J.L. Hanquin, M. Henneaux, P.
Spindel, A. Taormina, The fate of the mixmaster behavior in vacuum inhomogeneous
Kaluza-Klein cosmological models, Phys.Lett.B175:129-132 (1986); J. Demaret, Y. De
Rop, M. Henneaux, Chaos in nondiagonal spatially homogeneous cosmological models in
space-time dimensions ¡= 10, Phys.Lett.B211:37-41 (1988); T. Damour, M. Henneaux,
B. Julia, H. Nicolai, Hyperbolic Kac-Moody Algebras and Chaos in Kaluza-Klein Models,
Phys.Lett. B509 (2001) 323-33, hep-th/0103094.
[3] T. Damour, S. de Buyl, M. Henneaux, C. Schomblond, Einstein billiards and overextensions of finite-dimensional simple Lie algebras, JHEP 0208 (2002) 030, hep-th/0206125;
71
T. Damour, M. Henneaux, H. Nicolai, Cosmological Billiards, Class.Quant.Grav. 20
(2003) R145-R200, hep-th/0212256.
[4] S. de Buyl, M. Henneaux, B. Julia, L. Paulot, Cosmological billiards and oxidation,
Fortsch.Phys. 52 (2004) 548-554, hep-th/0312251; J. Brown, O. J. Ganor, C. Helfgott,
M-theory and E10: Billiards, Branes, and Imaginary Roots, JHEP 0408 (2004) 063,
hep-th/0401053; F. Englert, M. Henneaux, L. Houart, From very-extended to overextended gravity and M-theories, JHEP 0502 (2005) 070, hep-th/0412184; T. Damour,
Cosmological Singularities, Einstein Billiards and Lorentzian Kac-Moody Algebras, invited talk at Miami Waves 2004 (Conference on Geometric Analysis, Nonlinear Wave
Equations and General Relativity; Miami, 4-10 January 2004), gr-qc/0501064; T.
Damour, Poincare, Relativity, Billiards and Symmetry, invited talk given at the Solvay
Symposium on Henri Poincare (ULB, Brussels, 8-9 October 2004), hep-th/0501168.
[5] M. Henneaux, B. Julia, Hyperbolic billiards of pure D=4 supergravities, JHEP 0305
(2003) 047, hep-th/0304233.
[6] V.A. Belinsky, I.M. Khalatnikov, E.M. Lifshitz, Oscillatory approach to a singular
point in the relativistic cosmology, Adv.Phys.19:525-573 (1970); V.A. Belinsky, I.M.
Khalatnikov, E.M. Lifshitz, A general solution of the Einstein equations with a time
singularity,Adv.Phys.31:639-667, (1982); J. K. Erickson, D. H. Wesley, P. J. Steinhardt,
N. Turok, Kasner and Mixmaster behavior in universes with equation of state w ≥ 1,
Phys.Rev. D69 (2004) 063514, hep-th/0312009.
[7] P. Frè, V. Gili, F. Gargiulo, A. Sorin, K. Rulik, M. Trigiante, Cosmological backgrounds
of superstring theory and Solvable Algebras: Oxidation and Branes, Nucl.Phys. B685
(2004) 3-64, hep-th/0309237.
[8] P. Frè, K. Rulik, M. Trigiante, Exact solutions for Bianchi type cosmological metrics, Weyl orbits of E8(8) subalgebras and p–branes, Nucl.Phys. B694 (2004) 239-274,
hep-th/0312189.
[9] P. Frè, K. Rulik, F. Gargiulo, Cosmic Billiards with Painted Walls in Non Maximal
Supergravities: a worked out example, arXiV:hep-th/0507256.
[10] P. Frè, K. Rulik, F. Gargiulo, M. Trigiante, The general pattern of Kač-Moody extensions
in supergravity and the issue of cosmic billiards, arXiV:hep-th/0507249.
[11] L. Andrianopoli, R. D’Auria, S. Ferrara, P. Frè, M. Trigiante, R–R Scalars, U–Duality
and Solvable Lie Algebras, Nucl.Phys. B496 (1997) 617-629, hep-th/9611014; L. Andrianopoli, R. D’Auria, S. Ferrara, P. Frè, R. Minasian, M. Trigiante, Solvable Lie
Algebras in Type IIA, Type IIB and M Theories, Nucl.Phys. B493 (1997) 249-280,
hep-th/9612202.
[12] L. Andrianopoli, R. D’Auria, S. Ferrara, P. Frè and M. Trigiante, E(7)(7) duality, BPS
black hole evolution and fixed scalars, Nucl.Phys.B509:463-518,(1998), hep-th/9707087.
72
[13] P. Frè, U Duality, Solvable Lie Algebras and Extremal Black-Holes, Talk given at the III
National Meeting of the Italian Society for General Relativity (SIGRAV) on the occasion
of Prof. Bruno Bertotti’s 65th birthday. Rome September 1996, hep-th/9702167; P. Frè,
Solvable Lie Algebras, BPS Black Holes and Supergravity Gaugings, Fortsch.Phys. 47
(1999) 173-181, hep-th/9802045; G. Arcioni, A. Ceresole, F. Cordaro, R. D’Auria, P. Frè,
L. Gualtieri, M.Trigiante, N=8 BPS Black Holes with 1/2 or 1/4 Supersymmetry and
Solvable Lie Algebra Decompositions, Nucl.Phys. B542 (1999) 273-307, hep-th/9807136;
M. Bertolini, P. Frè, M. Trigiante, N=8 BPS black holes preserving 1/8 supersymmetry, Class.Quant.Grav. 16 (1999) 1519-1543, hep-th/9811251; M. Bertolini, P. Frè, M.
Trigiante, The generating solution of regular N=8 BPS black holes, Class.Quant.Grav.
16 (1999) 2987-3004, hep-th/9905143.
[14] M. Bertolini and M. Trigiante, Regular BPS black holes: Macroscopic and microscopic
description of the generating solution Nucl. Phys. B 582 (2000) 393, hep-th/0002191.
[15] M. Bertolini and M. Trigiante, Regular R-R and NS-NS BPS black holes Int. J. Mod.
Phys. A 15 (2000) 5017, hep-th/9910237.
[16] F. Cordaro, P. Frè, L. Gualtieri, P. Termonia, M.Trigiante, N=8 gaugings revisited:
an exhaustive classification, Nucl.Phys. B532 (1998) 245-279, hep-th/9804056; L. Andrianopoli, F. Cordaro, P. Frè, L. Gualtieri, Non-Semisimple Gaugings of D=5 N=8
Supergravity and FDA.s, Class.Quant.Grav. 18 (2001) 395-414, hep-th/0009048; L. Andrianopoli, F. Cordaro, P. Frè, L. Gualtieri, Non-Semisimple Gaugings of D=5 N=8
Supergravity, Fortsch.Phys. 49 (2001) 511-518, hep-th/0012203.
[17] P. Frè, Gaugings and other supergravity tools for p-brane physics, Lectures given at the
RTN School Recent Advances in M-theory, Paris February 1-8 IHP, hep-th/0102114.
[18] P. Fré, F. Gargiulo, Jan Rosseel, K. Rulik, M. Trigiante and A. Van Proeyen, Tits
Satake projections of homogenoeus special geometries, hep-th/0606173.
[19] P. Fré and A.S. Sorin, Integrability of Supergravity Billiards and the generalized Toda
lattice equations, Nucl. Phys. B733 (2006) 334, hep-th/0510156.
[20] S. Helgason, Differential geometry, Lie groups, and symmetric spaces, American Mathematical Society, 2001.
[21] Y. Kodama and K. T-R. McLaughlin, ”Explicit integration of the full symmetric
Toda hierarchy and the sorting property”, solv-int/9502006; Y. Kodama and J. Ye,
”Iso-spectral deformations of general matrix and their reductions on Lie algebras”,
solv-int/9506005.
[22] R.S. Leite and C. Tomei, Parametrization of polytopes of intersections of orbits by
conjugation, Linear Algebra and its Applications 361 (2003) 223, math.RA/0107048.
73