This action might not be possible to undo. Are you sure you want to continue?
K.S. STELLE
The Blackett Laboratory, Imperial College,
Prince Consort Road, London SW7 2BZ, UK
This review covers pbrane solutions to supergravity theories, by which we shall
mean solutions whose lowest energy conﬁguration, in a class determined by
asymptotic boundary conditions, contains a (p+1)dimensional Poincar´einvariant
“worldvolume” submanifold. The most important such solutions also possess par
tially unbroken supersymmetry, i.e. they saturate Bogomol’nyPrasadSommerﬁeld
(BPS) bounds on their energy densities with respect to the pform charges that they
carry. These charges also appear in the supersymmetry algebra and determine the
BPS bounds. Topics covered include the relations between mass densities, charge
densities and the preservation of unbroken supersymmetry; interpolatingsoliton
structure; κsymmetric worldvolume actions; diagonal and vertical KaluzaKlein
reductions; the four elementary solutions of D = 11 supergravity and the multiple
charge solutions derived from combinations of them; dualitysymmetry multiplets;
charge quantization; lowvelocity scattering and the geometry of worldvolume su
persymmetric σmodels; and the targetspace geometry of BPS instanton solutions
obtained by dimensional reduction of static pbranes.
Contents
1 Introduction 3
2 The pbrane ansatz 8
2.1 Singlecharge action and ﬁeld equations . . . . . . . . . . . . . 8
2.2 Electric and magnetic ans¨ atze . . . . . . . . . . . . . . . . . . . 8
2.3 Curvature components and pbrane equations . . . . . . . . . . 10
2.4 pbrane solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3 D = 11 examples 14
3.1 D = 11 Elementary/electric 2brane . . . . . . . . . . . . . . . 15
3.2 D = 11 Solitonic/magnetic 5brane . . . . . . . . . . . . . . . . 19
3.3 Black branes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 Charges, Masses and Supersymmetry 23
4.1 pform charges . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 pbrane mass densities . . . . . . . . . . . . . . . . . . . . . . . 27
4.3 pbrane charges . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4 Preserved supersymmetry . . . . . . . . . . . . . . . . . . . . . 28
1
5 The super pbrane worldvolume action 32
6 KaluzaKlein dimensional reduction 39
6.1 Multiple ﬁeldstrength solutions and the singlecharge truncation 41
6.2 Diagonal dimensional reduction of pbranes . . . . . . . . . . . 46
6.3 Multicenter solutions and vertical dimensional reduction . . . 46
6.4 The geometry of (D −3)branes . . . . . . . . . . . . . . . . . 50
6.5 Beyond the (D − 3)brane barrier: ScherkSchwarz reduction
and domain walls . . . . . . . . . . . . . . . . . . . . . . . . . . 52
7 Intersecting branes and scattering branes 60
7.1 Multiple component solutions . . . . . . . . . . . . . . . . . . . 60
7.2 Intersecting branes and the four elements in D = 11 . . . . . . 61
7.3 Brane probes, scattering branes and modulus σmodel geometry 66
8 Duality symmetries and charge quantisation 71
8.1 An example of duality symmetry: D = 8 supergravity . . . . . 72
8.2 pform charge quantisation conditions . . . . . . . . . . . . . . 74
8.3 Charge quantisation conditions and dimensional reduction . . . 77
8.4 Counting pbranes . . . . . . . . . . . . . . . . . . . . . . . . . 79
8.5 The charge lattice . . . . . . . . . . . . . . . . . . . . . . . . . 83
9 Local versus active dualities 85
9.1 The symmetries of type IIB supergravity . . . . . . . . . . . . . 87
9.2 Active duality symmetries . . . . . . . . . . . . . . . . . . . . . 90
10 Noncompact σmodels, null geodesics, and harmonic maps 94
11 Concluding remarks 99
2
1 Introduction
Let us begin from the bosonic sector of D = 11 supergravity,
1
I
11
=
d
11
x
√
−g(R −
1
48
F
2
[4]
) +
1
6
F
[4]
∧ F
[4]
∧ A
[3]
¸
. (1.1)
In addition to the metric, one has a 3form antisymmetrictensor gauge po
tential A
[3]
with a gauge transformation δA
[3]
= dΛ
[2]
and a ﬁeld strength
F
[4]
= dA
[3]
. The third term in the Lagrangian is invariant under the A
[3]
gauge
transformation only up to a total derivative, so the action (1.1) is invariant
under gauge transformations that are continuously connected to the identity.
This term is required, with the coeﬃcient given in (1.1), by the D = 11 local
supersymmetry that is required of the theory when the gravitinodependent
sector is included.
The equation of motion for the A
[3]
gauge potential is
d
∗
F
[4]
+
1
2
F
[4]
∧ F
[4]
= 0 ; (1.2)
this equation of motion gives rise to the conservation of an “electric” type
charge
2
U =
∂M8
(
∗
F
[4]
+
1
2
A
[3]
∧ F
[4]
) , (1.3)
where the integral of the 7form integrand is over the boundary at inﬁnity of
an arbitrary inﬁnite spacelike 8dimensional subspace of D = 11 spacetime.
Another conserved charge relies on the Bianchi identity dF
[4]
= 0 for its con
servation,
V =
∂
M5
F
[4]
, (1.4)
where the surface integral is now taken over the boundary at inﬁnity of a
spacelike 5dimensional subspace.
Charges such as (1.3, 1.4) can occur on the righthand side of the super
symmetry algebra,
a3
{Q, Q} = C(Γ
A
P
A
+ Γ
AB
U
AB
+ Γ
ABCDE
V
ABCDE
) , (1.5)
a
Although formally reasonable, there is admittedly something strange about this algebra.
For objects such as black holes, the total momentum terms on the righthand side have a well
deﬁned meaning, but for extended objects such as pbranes, the U and V terms on the right
hand side have meaning only as intensive quantities taken per spatial unit worldvolume. This
forces a similar intensive interpretation also for the momentum, requiring it to be considered
as a momentum per spatial unit worldvolume. Clearly, a more careful treatment of this
subject would recognize a corresponding divergence in the [Q, Q] anticommutator on the
lefthand side of (1.5) in such cases. This would then require then an inﬁnite normalization
factor for the algebra, whose removal requires the righthand side to be reinterpreted in an
intensive (i.e. per spatial unit worldvolume) as opposed to an extensive way.
3
where C is the charge conjugation matrix, P
A
is the energymomentum 11
vector and U
AB
and V
ABCDE
are 2form and 5form charges that we shall ﬁnd
to be related to the charges U and V (1.3, 1.4) above. Note that since the
supercharge Q in D = 11 supergravity is a 32component Majorana spinor,
the LHS of (1.5) has 528 components. The symmetric spinor matrices CΓ
A
,
CΓ
AB
and CΓ
ABCDE
on the RHS of (1.5) also have a total of 528 independent
components: 11 for the momentum P
A
, 55 for the “electric” charge U
AB
and
462 for the “magnetic” charge V
ABCDE
.
Now the question arises as to the relation between the charges U and V in
(1.3, 1.4) and the 2form and 5form charges appearing in (1.5). One thing that
immediately stands out is that the Gauss’ law integration surfaces in (1.3, 1.4)
are the boundaries of integration volumes M
8
,
M
5
that do not ﬁll out a whole
10dimensional spacelike hypersurface in spacetime, unlike the more familiar
situation for charges in ordinary electrodynamics. A rough idea about the
origin of the index structures on U
AB
and V
ABCDE
may be guessed from the
2fold and 5fold ways that the corresponding 8 and 5 dimensional integration
volumes may be embedded into a 10dimensional spacelike hypersurface. We
shall see in Section 4 that this is too na¨ıve, however: it masks an important
topological aspect of both the electric charge U
AB
and the magnetic charge
V
ABCDE
. The fact that the integration volume does not ﬁll out a full spacelike
hypersurface does not impede the conservation of the charges (1.3, 1.4); this
only requires that no electric or magnetic currents are present at the boundaries
∂M
8
, ∂
M
5
. Before we can discuss such currents, we shall need to consider in
some detail the supergravity solutions that carry charges like (1.3, 1.4). The
simplest of these have the structure of p + 1dimensional Poincar´einvariant
hyperplanes in the supergravity spacetime, and hence have been termed “p
branes” (see, e.g. Ref.
4
). In Sections 2 and 3, we shall delve in some detail
into the properties of these solutions.
Let us recall at this point some features of the relationship between super
gravity theory and string theory. Supergravity theories originally arose from
the desire to include supersymmetry into the framework of gravitational mod
els, and this was in the hope that the resulting models might solve some of
the outstanding diﬃculties of quantum gravity. One of these diﬃculties was
the ultraviolet problem, on which early enthusiasm for supergravity’s promise
gave way to disenchantment when it became clear that local supersymmetry
is not in fact suﬃcient to tame the notorious ultraviolet divergences that arise
in perturbation theory.
b
Nonetheless, supergravity theories won much admira
tion for their beautiful mathematical structure, which is due to the stringent
b
For a review of ultraviolet behavior in supergravity theories, see Ref.
5
4
constraints of their symmetries. These severely restrict the possible terms that
can occur in the Lagrangian. For the maximal supergravity theories, such as
those descended from the D = 11 theory (1.1), there is simultaneously a great
wealth of ﬁelds present and at the same time an impossibility of coupling any
independent external ﬁeldtheoretic “matter.” It was only occasionally noticed
in this early period that this impossibility of coupling to matter ﬁelds does not,
however, rule out coupling to “relativistic objects” such as black holes, strings
and membranes.
The realization that supergravity theories do not by themselves constitute
acceptable starting points for a quantum theory of gravity came somewhat
before the realization sunk in that string theory might instead be the sought
after perturbative foundation for quantum gravity. But the approaches of
supergravity and of string theory are in fact strongly interrelated: supergravity
theories arise as longwavelength eﬀectiveﬁeldtheory limits of string theories.
To see how this happens, consider the σmodel action
10
that describes a bosonic
string moving in a background “condensate” of its own massless modes (g
MN
,
A
MN
, φ):
I =
1
4πα
′
d
2
z
√
γ [γ
ij
∂
i
x
M
∂
j
x
N
g
MN
(x)
+iǫ
ij
∂
i
x
M
∂
j
x
N
A
MN
(x) +α
′
R(γ)φ(x)] . (1.6)
Every string theory contains a sector described by ﬁelds (g
MN
, A
MN
, φ); these
are the only ﬁelds that couple directly to the string worldsheet. In superstring
theories, this sector is called the NeveuSchwarz/NeveuSchwarz (NS–NS) sec
tor.
The σmodel action (1.6) is classically invariant under the worldsheet Weyl
symmetry γ
ij
→ Λ
2
(z)γ
ij
. Requiring cancellation of the anomalies in this
symmetry at the quantum level gives diﬀerentialequation restrictions on the
background ﬁelds (g
MN
, A
MN
, φ) that may be viewed as eﬀective equations of
motion for these massless modes.
11
This system of eﬀective equations may be
summarized by the corresponding ﬁeldtheory eﬀective action
I
eﬀ
=
d
D
x
√
−ge
−2φ
(D −26) −
3
2
α
′
(R + 4∇
2
φ −4(∇φ)
2
−
1
12
F
MNP
F
MNP
+O(α
′
)
2
, (1.7)
where F
MNP
= ∂
M
A
NP
+∂
N
A
PM
+∂
P
A
MN
is the 3form ﬁeld strength for the
A
MN
gauge potential. The (D−26) term reﬂects the critical dimension for the
bosonic string: ﬂat space is a solution of the above eﬀective theory only for
D = 26. The eﬀective action for the superstring theories that we shall consider
5
in this review contains a similar (NS–NS) sector, but with the substitution of
(D−26) by (D−10), reﬂecting the diﬀerent critical dimension for superstrings.
The eﬀective action (1.7) is written in the form directly obtained from
string σmodel calculations. It is not written in the form generally preferred
by relativists, which has a clean EinsteinHilbert term free from exponential
prefactors like e
−2φ
. One may rewrite the eﬀective action in a diﬀerent frame
by making a Weylrescaling ﬁeld redeﬁnition g
MN
→ e
λφ
g
MN
. I
eﬀ
as written
in (1.7) is in the string frame; after an integration by parts, it takes the form,
specializing now to D = 10,
I
string
=
d
10
x
−g
(s)
e
−2φ
R(g
(s)
) + 4∇
M
φ∇
M
φ −
1
12
F
MNP
F
MNP
. (1.8)
After making the transformation
g
(e)
MN
= e
−φ/2
g
(s)
MN
, (1.9)
one obtains the Einstein frame action,
I
Einstein
=
d
10
x
−g
(e)
R(g
(e)
)−
1
2
∇
M
φ∇
M
φ−
1
12
e
−φ
F
MNP
F
MNP
, (1.10)
where the indices are now raised and lowered with g
(e)
MN
. To understand how this
Weyl rescaling works, note that under xindependent rescalings, the connection
Γ
MN
P
is invariant. This carries over also to terms with φ undiﬀerentiated,
which emerge from the e
λφ
Weyl transformation. One then chooses λ so as to
eliminate the e
−2φ
factor. Terms with φ undiﬀerentiated do change, however.
As one can see in (1.10), the Weyl transformation is just what is needed to
unmask the positiveenergy sign of the kinetic term for the φ ﬁeld, despite the
apparently negative sign of its kinetic term in I
string
.
Now let us return to the maximal supergravities descended from (1.1). We
shall discuss in Section 6) the process of KaluzaKlein dimensional reduction
that relates theories in diﬀerent dimensions of spacetime. For the present, we
note that upon specifying the KaluzaKlein ansatz expressing ds
2
11
in terms of
ds
2
10
, the KaluzaKlein vector A
M
and the dilaton φ,
ds
2
11
= e
−φ/6
ds
2
10
+e
4φ/3
(dz +A
M
dx
M
)
2
M = 0, 1, . . . , 9, (1.11)
the bosonic D = 11 action (1.1) reduces to the Einsteinframe type IIA bosonic
action
14
I
Einstein
IIA
=
d
10
x
−g
(e)
R(g
(e)
) −
1
2
∇
M
φ∇
M
φ −
1
12
e
−φ
F
MNP
F
MNP
−
1
48
e
φ/2
F
MNPQ
F
MNPQ
−
1
4
e
3φ/2
F
MN
F
MN
¸
+L
FFA
, (1.12)
6
where F
MN
is the ﬁeld strength for the KaluzaKlein vector A
M
.
The top line in (1.12) corresponds to the NS–NS sector of the IIA theory;
the bottom line corresponds the R–R sector (plus the ChernSimons terms,
which we have not shown explicitly). In order to understand better the dis
tinction between these two sectors, rewrite (1.12) in string frame using (1.9).
One ﬁnds
I
string
IIA
=
d
10
x
−g
(s)
e
−2φ
R(g
(s)
) + 4∇
M
φ∇
M
φ −
1
12
F
MNP
F
MNP
−
1
48
F
MNPQ
F
MNPQ
−
1
4
F
MN
F
MN
¸
+L
FFA
. (1.13)
Now one may see the distinguishing feature of the NS–NS sector as opposed to
the R–R sector: the dilaton coupling is a uniform e
−2φ
in the NS–NS sector,
and it does not couple (in string frame) to the R–R sector ﬁeld strengths. Com
paring with the familiar g
−2
couplingconstant factor for the YangMills ac
tion, one sees that the asymptotic value e
φ∞
plays the rˆole of the stringtheory
coupling constant. Since in classical supergravity theory, one will encounter
transformations that have the eﬀect of ﬂipping the sign of the dilaton, φ →−φ,
the study of classical supergravity will contain decidedly nonperturbative in
formation about string theory. In particular, this will arise in the study of
pbrane solitons, to which we shall shortly turn.
In this review, we shall mostly consider the descendants of the type IIA
action (1.12). This leaves out one important case that we shall have to consider
separately: the chiral type IIB theory in D = 10. In the type IIB theory,
15
one has F
[1]
= dχ, where χ is a R–R zeroform (i.e. a pseudoscalar ﬁeld),
F
R
[3]
= dA
R
[2]
, a second 3form ﬁeld strength making a pair together with F
NS
[3]
from the NS–NS sector, and F
[5]
= dA
[4]
, which is a selfdual 5form in D = 10,
F
[5]
=
∗
F
[5]
.
Thus one naturally encounters ﬁeld strengths of ranks 1–5 in the super
gravity theories deriving from superstring theories. In addition, one may use
ǫ
[10]
to dualize certain ﬁeld strengths; e.g. the original F
[3]
may be dualized to
the 7form
∗
F
[7]
. The upshot is that antisymmetrictensor gauge ﬁeld strengths
of diverse ranks need to be taken into account when searching for solutions to
stringtheory eﬀective ﬁeld equations. These ﬁeld strengths will play an essen
tial rˆole in supporting the pbrane solutions that we shall now describe.
7
2 The pbrane ansatz
2.1 Singlecharge action and ﬁeld equations
We have seen that one needs to consider eﬀective theories containing gravity,
various ranks of antisymmetrictensor ﬁeld strengths and various scalars. To
obtain a more tractable system to study, we shall make a consistent truncation
of the action down to a simple system in D dimensions comprising the metric
g
MN
, a scalar ﬁeld φ and a single (n − 1)form gauge potential A
[n−1]
with
corresponding ﬁeld strength F
[n]
; the whole is described by the action
I =
d
D
x
√
−g
R −
1
2
∇
M
φ∇
M
φ −
1
2n!
e
aφ
F
2
[n]
. (2.1)
We shall consider later in more detail how (2.1) may be obtained by a consis
tent truncation from a full supergravity theory in D dimensions. The notion
of a consistent truncation will play a central rˆole in our discussion of the BPS
solutions of supergravity theories. A consistent truncation is one for which
solutions of the truncated theory are also perfectly good, albeit speciﬁc, solu
tions of the original untruncated theory. Truncation down to the system (2.1)
with a single scalar φ and a single ﬁeld strength F
[n]
will be consistent except
for certain special cases when n = D/2 that we shall have to consider sepa
rately. In such cases, one can have dyonic solutions, and in such cases it will
generally be necessary to retain an axionic scalar χ as well. Note that in (2.1)
we have not included contributions coming from the FFA ChernSimons term
in the action. These are also consistently excluded in the truncation to the
singlecharge action (2.1). The value of the important parameter a controlling
the interaction of the scalar ﬁeld φ with the ﬁeld strength F
[n]
in (2.1) will
vary according to the cases considered in the following.
Varying the action (2.1) produces the following set of equations of motion:
R
MN
=
1
2
∂
M
φ∂
N
φ +S
MN
(2.2a)
S
MN
=
1
2(n −1)!
e
aφ
(F
M···
F
N
···
−
n −1
n(D−2)
F
2
g
MN
) (2.2b)
∇
M
1
(e
aφ
F
M1···Mn
) = 0 (2.2c)
φ =
a
2n!
e
aφ
F
2
. (2.2d)
2.2 Electric and magnetic ans¨atze
In order to solve the above equations, we shall make a simplifying ansatz. We
shall be looking for solutions preserving certain unbroken supersymmetries,
8
and these will in turn require unbroken translational symmetries as well. For
simplicity, we shall also require isotropic symmetry in the directions “trans
verse” to the translationallysymmetric ones. These restrictions can subse
quently be relaxed in generalizations of the basic class of pbrane solutions
that we shall discuss here. For this basic class of solutions, we make an ansatz
requiring (Poincar´e)
d
× SO(D − d) symmetry. One may view the soughtfor
solutions as ﬂat d = p + 1 dimensional hyperplanes embedded in the ambi
ent Ddimensional spacetime; these hyperplanes may in turn be viewed as the
histories, or worldvolumes, of pdimensional spatial surfaces. Accordingly, let
the spacetime coordinates be split into two ranges: x
M
= (x
µ
, y
m
), where x
µ
(µ = 0, 1, · · · , p = d −1) are coordinates adapted to the (Poincar´e)
d
isometries
on the worldvolume and where y
m
(m = d, · · · , D − 1) are the coordinates
“transverse” to the worldvolume.
An ansatz for the spacetime metric that respects the (Poincar´e)
d
×SO(D−
d) symmetry is
12
ds
2
= e
2A(r)
dx
µ
dx
ν
η
µν
+e
2B(r)
dy
m
dy
n
δ
mn
µ = 0, 1, . . . , p m = p + 1, . . . , D−1 , (2.3)
where r =
√
y
m
y
m
is the isotropic radial coordinate in the transverse space.
Since the metric components depend only on r, translational invariance in
the worldvolume directions x
µ
and SO(D − d) symmetry in the transverse
directions y
m
is guaranteed.
The corresponding ansatz for the scalar ﬁeld φ(x
M
) is simply φ = φ(r).
For the antisymmetric tensor gauge ﬁeld, we face a bifurcation of pos
sibilities for the ansatz, the two possibilities being related by duality. The
ﬁrst possibility is naturally expressed directly in terms of the gauge poten
tial A
[n−1]
. Just as the Maxwell 1form naturally couples to the worldline of
a charged particle, so does A
[n−1]
naturally couple to the worldvolume of a
p = d −1 = (n −1) −1 dimensional “charged” extended object. The “charge”
here will be obtained from Gauss’law surface integrals involving F
[n]
, as we
shall see later. Thus, the ﬁrst possibility for A
[n−1]
is to support a d
el
= n −1
dimensional worldvolume. This is what we shall call the “elementary,” or
“electric” ansatz:
A
µ1···µn−1
= ǫ
µ1···µn−1
e
C(r)
, others zero. (2.4)
SO(D−d) isotropicity and (Poincar´e)
d
symmetry are guaranteed here because
the function C(r) depends only on the transverse radial coordinate r. Instead
of the ansatz (2.4), expressed in terms of A
[n−1]
, we could equivalently have
9
given just the F
[n]
ﬁeld strength:
F
(el)
mµ1···µn−1
= ǫ
µ1···µn−1
∂
m
e
C(r)
, others zero. (2.5)
The worldvolume dimension for the elementary ansatz (2.4, 2.5) is clearly d
el
=
n −1.
The second possible way to relate the rank n of F
[n]
to the worldvolume
dimension d of an extended object is suggested by considering the dualized
ﬁeld strength
∗
F, which is a (D − n) form. If one were to ﬁnd an underlying
gauge potential for
∗
F (locally possible by courtesy of a Bianchi identity), this
would naturally couple to a d
so
= D − n − 1 dimensional worldvolume. Since
such a dualized potential would be nonlocally related to the ﬁelds appearing
in the action (2.1), we shall not explicitly follow this construction, but shall
instead take this reference to the dualized theory as an easy way to identify
the worldvolume dimension for the second type of ansatz. This “solitonic”
or “magnetic” ansatz for the antisymmetric tensor ﬁeld is most conveniently
expressed in terms of the ﬁeld strength F
[n]
, which now has nonvanishing values
only for indices corresponding to the transverse directions:
F
(mag)
m1···mn
= λǫ
m1···mnp
y
p
r
n+1
, others zero, (2.6)
where the magneticcharge parameter λ is a constant of integration, the only
thing left undetermined by this ansatz. The power of r in the solitonic/mag
netic ansatz is determined by requiring F
[n]
to satisfy the Bianchi identity.
c
Note that the worldvolume dimensions of the elementary and solitonic cases
are related by d
so
=
˜
d
el
≡ D−d
el
−2; note also that this relation is idempotent,
i.e.
(
˜
d) = d.
2.3 Curvature components and pbrane equations
In order to write out the ﬁeld equations after insertion of the above ans¨ atze,
one needs to compute the Ricci tensor for the metric.
13
This is most easily
done by introducing vielbeins, i.e., orthonormal frames,
16
with tangentspace
indices denoted by underlined indices:
g
MN
= e
M
E
e
N
F
η
EF
. (2.7)
c
Speciﬁcally, one ﬁnds ∂qFm
1
···mn
= r
−(n+1)
ǫm
1
···mnq −(n+1)ǫm
1
···mnpy
p
yq/r
2
; upon
taking the totally antisymmetrized combination [qm
1
· · · mn], the factor of (n+1) is evened
out between the two terms and then one ﬁnds from cycling a factor
m
y
m
ym = r
2
, thus
obtaining cancellation.
10
Next, one constructs the corresponding 1forms: e
E
= dx
M
e
M
E
. Splitting up
the tangentspace indices E = (µ, m) similarly to the world indices M = (µ, m),
we have for our ans¨ atze the vielbein 1forms
e
µ
= e
A(r)
dx
µ
, e
m
= e
B(r)
dy
m
. (2.8)
The corresponding spin connection 1forms are determined by the condi
tion that the torsion vanishes, de
E
+ω
E
F
∧ e
F
= 0, which yields
ω
µν
= 0 , ω
µn
= e
−B(r)
∂
n
A(r)e
µ
ω
mn
= e
−B(r)
∂
n
B(r)e
m
− e
−B(r)
∂
m
B(r)e
n
. (2.9)
The curvature 2forms are then given by
R
EF
[2]
= dω
EF
+ω
ED
∧ ω
D
F
. (2.10)
From the curvature components so obtained, one ﬁnds the Ricci tensor com
ponents
R
µν
= −η
µν
e
2(A−B)
(A
′′
+d(A
′
)
2
+
˜
dA
′
B
′
+
(
˜
d + 1)
r
A
′
)
R
mn
= −δ
mn
(B
′′
+dA
′
B
′
+
˜
d(B
′
)
2
+
(2
˜
d + 1)
r
B
′
+
d
r
A
′
) (2.11)
−
y
m
y
n
r
2
(
˜
dB
′′
+dA
′′
−2dA
′
B
′
+d(A
′
)
2
−
˜
d(B
′
)
2
−
˜
d
r
B
′
−
d
r
A
′
) ,
where again,
˜
d = D −d −2, and the primes indicate ∂/∂r derivatives.
Substituting the above relations, one ﬁnds the set of equations that we
need to solve to obtain the metric and φ:
A
′′
+d(A
′
)
2
+
˜
dA
′
B
′
+
(
˜
d+1)
r
A
′
=
˜
d
2(D−2)
S
2
{µν}
B
′′
+dA
′
B
′
+
˜
d(B
′
)
2
+
(2
˜
d+1)
r
B
′
+
d
r
A
′
= −
d
2(D−2)
S
2
{δ
mn
}
˜
dB
′′
+dA
′′
−2dA
′
B
′
+d(A
′
)
2
−
˜
d(B
′
)
2
−
˜
d
r
B
′
−
d
r
A
′
+
1
2
(φ
′
)
2
=
1
2
S
2
{y
m
y
n
}
φ
′′
+dA
′
φ
′
+
˜
dB
′
φ
′
+
(
˜
d+1)
r
φ
′
= −
1
2
ςaS
2
{φ}
(2.12)
where ς = ±1 for the elementary/solitonic cases and the source appearing on
the RHS of these equations is
S =
(e
1
2
aφ−dA+C
)C
′
electric: d = n −1, ς = +1
λ(e
1
2
aφ−
˜
dB
)r
−
˜
d−1
magnetic: d = D −n −1, ς = −1.
(2.13)
11
2.4 pbrane solutions
The pbrane equations (2.12, 2.13) are still rather daunting. Before we embark
on solving these equations, let us ﬁrst note a generalization. Although Eqs
(2.12) have been speciﬁcally written for an isotropic pbrane ansatz, one may
recognize more general possibilities by noting the form of the Laplace operator,
which for isotropic scalar functions of r is
∇
2
φ = φ
′′
+ (
˜
d + 1)r
−1
φ
′
. (2.14)
We shall see later that more general solutions of the Laplace equation than
the simple isotropic ones considered here will also play important rˆoles in the
story.
In order to reduce the complexity of Eqs (2.12), we shall reﬁne the p
brane ansatz (2.3, 2.5, 2.6) by looking ahead a bit and taking a hint from the
requirements for supersymmetry preservation, which shall be justiﬁed in more
detail later on in Section 4. Accordingly, we shall look for solutions satisfying
the linearity condition
dA
′
+
˜
dB
′
= 0 . (2.15)
After eliminating B using (2.15), the independent equations become
17
∇
2
φ = −
1
2
ςaS
2
(2.16a)
∇
2
A =
˜
d
2(D−2)
S
2
(2.16b)
d(D −2)(A
′
)
2
+
1
2
˜
d(φ
′
)
2
=
1
2
˜
dS
2
, (2.16c)
where, for sphericallysymmetric (i.e. isotropic) functions in the transverse
(D −d) dimensions, the Laplacian is ∇
2
φ = φ
′′
+ (
˜
d + 1)r
−1
φ
′
.
Equations (2.16a,b) suggest that we now further reﬁne the ans¨ atze by
imposing another linearity condition:
φ
′
=
−ςa(D−2)
˜
d
A
′
. (2.17)
At this stage, it is useful to introduce a new piece of notation, letting
a
2
= ∆−
2d
˜
d
(D−2)
. (2.18)
With this notation, equation (2.16c) gives
S
2
=
∆(φ
′
)
2
a
2
, (2.19)
12
so that the remaining equation for φ becomes ∇
2
φ +
ς∆
2a
(φ
′
)
2
= 0, which can
be reexpressed as a Laplace equation,
d
∇
2
e
ς∆
2a
φ
= 0 . (2.20)
Solving this in the transverse (D − d) dimensions with our assumption of
transverse isotropicity (i.e. spherical symmetry) yields
e
ς∆
2a
φ
≡ H(y) = 1 +
k
r
˜
d
k > 0 , (2.21)
where the constant of integration φ

r→∞
has been set equal to zero here for
simplicity: φ
∞
= 0. The integration constant k in (2.21) sets the mass scale
of the solution; it has been taken to be positive in order to ensure the absence
of naked singularities at ﬁnite r. This positivity restriction is similar to the
usual restriction to a positive mass parameter M in the standard Schwarzschild
solution.
In the case of the elementary/electric ansatz, with ς = +1, it still remains
to ﬁnd the function C(r) that determines the antisymmetrictensor gauge ﬁeld
potential. In this case, it follows from (2.13) that S
2
= e
aφ−2dA
(C
′
e
C
)
2
.
Combining this with (2.19), one ﬁnds the relation
∂
∂r
(e
C
) =
−
√
∆
a
e
−
1
2
aφ+dA
φ
′
(2.22)
(where it should be remembered that a < 0). Finally, it is straightforward
to verify that the relation (2.22) is consistent with the equation of motion for
F
[n]
:
∇
2
C +C
′
(C
′
+
˜
dB
′
−dA
′
+aφ
′
) = 0 . (2.23)
In order to simplify the explicit form of the solution, we now pick values
of the integration constants to make A
∞
= B
∞
= 0, so that the solution tends
to ﬂat empty space at transverse inﬁnity. Assembling the result, starting from
the Laplaceequation solution H(y) (2.21), one ﬁnds
7,13
ds
2
= H
−4
˜
d
∆(D−2)
dx
µ
dx
ν
η
µν
+H
4d
∆(D−2)
dy
m
dy
m
(2.24a)
e
φ
= H
2a
ς∆
ς =
+1, elementary/electric
−1, solitonic/magnetic
(2.24b)
H(y) = 1 +
k
r
˜
d
, (2.24c)
d
Note that Eq. (2.20) can also be more generally derived; for example, it still holds if one
relaxes the assumption of isotropicity in the transverse space.
13
and in the elementary/electric case, C(r) is given by
e
C
=
2
√
∆
H
−1
. (2.25)
In the solitonic/magnetic case, the constant of integration is related to the
magnetic charge parameter λ in the ansatz (2.6) by
k =
√
∆
2
˜
d
λ . (2.26)
In the elementary/electric case, this relation may be taken to deﬁne the pa
rameter λ.
The harmonic function H(y) (2.21) determines all of the features of a p
brane solution (except for the choice of gauge for the A
[n−1]
gauge potential).
It is useful to express the electric and magnetic ﬁeld strengths directly in terms
of H:
F
mµ1...µn−1
=
2
√
∆
ǫ
µ1...µn−1
∂
m
(H
−1
) m = d, . . . , D− 1 electric (2.27a)
F
m1...mn
= −
2
√
∆
ǫ
m1...mnr
∂
r
H m = d, . . . , D −1 magnetic, (2.27b)
with all other independent components vanishing in either case.
3 D = 11 examples
Let us now return to the bosonic sector of D = 11 supergravity, which has
the action (1.1). In searching for pbrane solutions to this action, there are
two particular points to note. The ﬁrst is that no scalar ﬁeld is present in
(1.1). This follows from the supermultiplet structure of the D = 11 theory,
in which all ﬁelds are gauge ﬁelds. In lower dimensions, of course, scalars do
appear; e.g. the dilaton in D = 10 type IIA supergravity emerges out of the
D = 11 metric upon dimensional reduction from D = 11 to D = 10. The
absence of the scalar that we had in our general discussion may be handled
here simply by identifying the scalar coupling parameter a with zero, so that
the scalar may be consistently truncated from our general action (2.1). Since
a
2
= ∆−2d
˜
d/(D −2), we identify ∆ = 2 · 3 · 6/9 = 4 for the D = 11 cases.
Now let us consider the consistency of dropping contributions arising from
the FFA ChernSimons term in (1.1). Note that for n = 4, the F
[4]
antisym
metric tensor ﬁeld strength supports either an elementary/electric solution
with d = n − 1 = 3 (i.e. a p = 2 membrane) or a solitonic/magnetic solution
14
with
˜
d = 11 − 3 − 2 = 6 (i.e. a p = 5 brane). In both these elementary and
solitonic cases, the FFA term in the action (1.1) vanishes and hence this term
does not make any nonvanishing contribution to the metric ﬁeld equations for
our ans¨ atze. For the antisymmetric tensor ﬁeld equation, a further check is
necessary, since there one requires the variation of the FFA term to vanish in
order to consistently ignore it. The ﬁeld equation for A
[3]
is (1.2), which when
written out explicitly becomes
∂
M
√
−gF
MUV W
+
1
2(4!)
2
ǫ
UV Wx1x2x3x4y1y2y3y4
F
x1x2x3x4
F
y1y2y3y4
= 0 . (3.1)
By direct inspection, one sees that the second term in this equation vanishes
for both ans¨ atze.
Next, we shall consider the elementary/electric and the solitonic/magnetic
D = 11 cases in detail. Subsequently, we shall explore how these particular
solutions ﬁt into wider, “black,” families of pbranes.
3.1 D = 11 Elementary/electric 2brane
From our general discussion in Sec. 2, we have the elementaryansatz solution
18
ds
2
= (1 +
k
r
6
)
−
2
/3
dx
µ
dx
ν
η
µν
+ (1 +
k
r
6
)
1
/3
dy
m
dy
m
A
µνλ
= ǫ
µνλ
(1 +
k
r
6
)
−1
, other components zero.
electric 2brane: isotropic coordinates
(3.2)
At ﬁrst glance, this solution looks like it might be singular at r = 0. However, if
one calculates the invariant components of the curvature tensor R
MNPQ
and of
the ﬁeld strength F
mµ1µ2µ3
, subsequently referred to an orthonormal frame by
introducing vielbeins as in (2.8), one ﬁnds these invariants to be nonsingular.
Moreover, although the proper distance to the surface r = 0 along a t = x
0
=
const. geodesic diverges, the surface r = 0 can be reached along null geodesics
in ﬁnite aﬃne parameter.
19
Thus, one may suspect that the metric as given in (3.2) does not in fact
cover the entire spacetime, and so one should look for an analytic extension
of it. Accordingly, one may consider a change to “Schwarzschildtype” coordi
nates by setting r = (˜ r
6
−k)
1
/6
. The solution then becomes:
19
ds
2
= (1 −
k
˜ r
6
)
2
/3
(−dt
2
+dσ
2
+dρ
2
) + (1 −
k
˜ r
6
)
−2
d˜ r
2
+ ˜ r
2
dΩ
2
7
A
µνλ
= ǫ
µνλ
(1 −
k
˜ r
6
) , other components zero,
electric 2brane: Schwarzschildtype coordinates
(3.3)
15
where we have supplied explicit worldvolume coordinates x
µ
= (t, σ, ρ) and
where dΩ
2
7
is the line element on the unit 7sphere, corresponding to the bound
ary ∂M
8T
of the 11 −3 = 8 dimensional transverse space.
The Schwarzschildlike coordinates make the surface ˜ r = k
1
/6
(correspond
ing to r = 0) look like a horizon. One may indeed verify that the normal
to this surface is a null vector, conﬁrming that ˜ r = k
1
/6
is in fact a horizon.
This horizon is degenerate, however. Owing to the
2
/3 exponent in the g
00
component, curves along the t axis for ˜ r < k
1
/6
remain timelike, so that light
cones do not “ﬂip over” inside the horizon, unlike the situation for the classic
Schwarzschild solution.
In order to see the structure of the membrane spacetime more clearly,
let us change coordinates once again, setting ˜ r = k
1
/6
(1 − R
3
)
−
1
/6
. Overall,
the transformation from the original isotropic coordinates to these new ones
is eﬀected by setting r = k
1
/6
R
1
/2
/(1 − R
3
)
1
/6
. In these new coordinates, the
solution becomes
19
ds
2
=
¸
R
2
(−dt
2
+dσ
2
+dρ
2
) +
1
4
k
1
/3
R
−2
dR
2
¸
+k
1
/3
dΩ
2
7
(a)
+
1
4
k
1
/3
[(1 −R
3
)
−
7
/3
−1]R
−2
dR
2
+k
1
/3
[(1 −R
3
)
−
1
/3
−1]dΩ
2
7
(b)
A
µνλ
= R
3
ǫ
µνλ
, other components zero.
electric 2brane: interpolating coordinates
(3.4)
This form of the solution makes it clearer that the lightcones do not
“ﬂip over” in the region inside the horizon (which is now at R = 0, with
R < 0 being the interior). The main usefulness of the third form (3.4) of the
membrane solution, however, is that it reveals how the solution interpolates
between other “vacuum” solutions of D = 11 supergravity.
19
As R → 1, the
solution becomes ﬂat, in the asymptotic exterior transverse region. As one
approaches the horizon at R = 0, line (b) of the metric in (3.4) vanishes
at least linearly in R. The residual metric, given in line (a), may then be
recognized as a standard form of the metric on (AdS)
4
× S
7
, generalizing the
RobinsonBertotti solution on (AdS)
2
× S
2
in D = 4. Thus, the membrane
solution interpolates between ﬂat space as R → 1 and (AdS)
4
× S
7
as R → 0
at the horizon.
Continuing on inside the horizon, one eventually encounters a true singu
larity at ˜ r = 0 (R →−∞). Unlike the singularity in the classic Schwarzschild
solution, which is spacelike and hence unavoidable, the singularity in the mem
brane spacetime is timelike. Generically, geodesics do not intersect the singu
larity at a ﬁnite value of an aﬃne parameter value. Radial null geodesics do
intersect the singularity at ﬁnite aﬃne parameter, however, so the spacetime
is in fact genuinely singular. The timelike nature of this singularity, however,
16
invites one to consider coupling a δfunction source to the solution at ˜ r = 0.
Indeed, the D = 11 supermembrane action,
20
which generalizes the Nambu
Goto action for the string, is the unique “matter” system that can consistently
couple to D = 11 supergravity.
20,22
Analysis of this coupling yields a rela
tion between the parameter k in the solution (3.2) and the tension T of the
supermembrane action:
18
k =
κ
2
T
3Ω
7
, (3.5)
where 1/(2κ
2
) is the coeﬃcient of
√
−gR in the EinsteinHilbert Lagrangian
and Ω
7
is the volume of the unit 7sphere S
7
, i.e. the solid angle subtended by
the boundary at transverse inﬁnity.
The global structure of the membrane spacetime
19
is similar to the extreme
ReissnerNordstrom solution of General Relativity.
24
This global structure is
summarized by a CarterPenrose diagram as shown in Figure 1, in which the
angular coordinates on S
7
and also two ignorable worldsheet coordinates have
been suppressed. As one can see, the region mapped by the isotropic coordi
nates does not cover the whole spacetime. This region, shaded in the diagram,
is geodesically incomplete, since one may reach its boundaries H
+
, H
−
along
radial null geodesics at a ﬁnite aﬃneparameter value. These boundary sur
faces are not singular, but, instead, constitute future and past horizons (one
can see from the form (3.3) of the solution that the normals to these sur
faces are null). The “throat” P in the diagram should be thought of as an
exceptional point at inﬁnity, and not as a part of the central singularity.
The region exterior to the horizon interpolates between ﬂat regions J
±
at future and past null inﬁnities and a geometry that asymptotically tends
to (AdS)
4
× S
7
on the horizon. This interpolating portion of the spacetime,
corresponding to the shaded region of Figure 1 which is covered by the isotropic
coordinates, may be sketched as shown in Figure 2.
17
J
+
H
–
H
+
J
–
“throat” P
R = const. hypersurface
t = const. hypersurface
timelike singularity
at r = 0 (R →− ∞)
i spatial infinity
0
Only the shaded region is covered
by the isotropic coordinates
~
Figure 1: CarterPenrose diagram for the D = 11 elementary/electric 2brane solution.
18
flat M
11
infinite “throat:”
(AdS) × S
4
7
Figure 2: The D = 11 elementary/electric 2brane solution interpolates between ﬂat space
at J
±
and (AdS)
4
×S
7
at the horizon.
3.2 D = 11 Solitonic/magnetic 5brane
Now consider the 5brane solution to the D = 11 theory given by the solitonic
ansatz for F
[4]
. In isotropic coordinates, this solution is a magnetic 5brane:
25
ds
2
= (1+
k
r
3
)
−
1
/3
dx
µ
dx
ν
η
µν
+(1+
k
r
3
)
2
/3
dy
m
dy
m
µ, ν = 0, · · · , 5
F
m1···m4
= 3kǫ
m1···m4p
y
p
r
5
other components zero.
magnetic 5brane: isotropic coordinates
(3.6)
As in the case of the elementary/electric membrane, this solution inter
polates between two “vacua” of D = 11 supergravity. Now, however, these
asymptotic geometries consist of the ﬂat region encountered as r →∞ and of
(AdS)
7
×S
4
as one approaches r = 0, which once again is a degenerate horizon.
Combining two coordinate changes analogous to those of the elementary case,
19
r = (˜ r
3
−k)
1
/3
and ˜ r = k
1
/3
(1 −R
6
)
−
1
/3
, one has an overall transformation
r =
k
1
/3
R
2
(1 −R
6
)
1
/3
. (3.7)
After these coordinate changes, the metric becomes
ds
2
= R
2
dx
µ
dx
ν
η
µν
+k
2
/3
4R
−2
(1−R
6
)
8
/3
dR
2
+
dΩ
2
4
(1−R
6
)
2
/3
.
magnetic 5brane: interpolating coordinates
(3.8)
Once again, the surface r = 0 ↔ R = 0 may be seen from (3.8) to be a
nonsingular degenerate horizon. In this case, however, not only do the light
cones maintain their timelike orientation when crossing the horizon, as already
happened in the electric case (3.4), but now the magnetic solution (3.8) is in
fact fully symmetric
26
under a discrete isometry R →−R.
26
Given this isometry R →−R, one can, if one wishes identify the spacetime
region R ≤ 0 with the region R ≥ 0. This identiﬁcation is analogous to
the identiﬁcation one naturally makes for ﬂat space when written in polar
coordinates, with the metric ds
2
ﬂat
= −dt
2
+ dr
2
+ r
2
d
2
. However, one must
be attentive to the issue of conical singularities in this case. Unlike the case
of doubled ﬂat space, where the identiﬁcation removes a conical singularity
with deﬁcit angle 2π, the R ↔ −R identiﬁcation in the 5brane geometry
introduces a conical singularity; the nonsingular spacetime is the R ↔ −R
symmetric but nonidentiﬁed spacetime. This smoothly continued spacetime
has an inﬁnite “throat,” at the horizon R = 0, and the region covered by
the isotropic coordinates may once again be sketched as in Figure 2, except
now with the asymptotic geometry in the “throat” region being (AdS)
7
× S
4
instead of (AdS)
4
×S
7
as in the case of the elementary/electric solution. The
CarterPenrose diagram for the solitonic/magnetic 5brane solution is given in
Figure 3, where the full diagram extends indeﬁnitely by “tiling” the section
shown.
20
R
=
0
R > 0
I
R < 0
symmetric with I
R < 0
symmetric with I
J
(
R
=
1
)
+
J
(
R
=
1
)
–
H
(
R
=
0
)
+
H
(
R
=
0
)
–
J
(
R
=

1
)
'
J
(
R
=

1
)
"
+
J
(
R
=

1
)
'
J
(
R
=

1
)
"
R
=
0
i
0
spatial
infinity
spatial infinity
'
i
0
"
i
0
–
–
+
Figure 3: CarterPenrose diagram for the solitonic/magnetic 5brane solution.
21
The electric and magnetic D = 11 solutions discussed here and in the
previous subsection are “nondilatonic” in that they do not involve a scalar ﬁeld,
since the bosonic sector of D = 11 supergravity (1.1) does not even contain a
scalar ﬁeld. Similar solutions occur in other situations where the parameter
a (2.18) for a ﬁeld strength supporting a pbrane solution vanishes, in which
cases the scalar ﬁelds may consistently be set to zero; this happens for (D, d) =
(11, 3), (11,5), (10,4), (6,2), (5,1), (5,2) and (4,1). In these special cases, the
solutions are nonsingular at the horizon and so one may analytically continue
through to the other side of the horizon. When d is even for “scalarless”
solutions of this type, there exists a discrete isometry analogous to the R →
−R isometry of the D = 11 5brane solution (3.8), allowing the outer and
inner regions to be identiﬁed.
26
When d is odd in such cases, the analytically
extended metric eventually reaches a timelike curvature singularity at ˜ r = 0.
When a = 0 and the scalar ﬁeld associated to the ﬁeld strength supporting
a solution cannot be consistently set to zero, then the solution has a singularity
at the horizon, as can be seen directly in the scalar solution (2.21) itself (where
we recall that in isotropic coordinates, the horizon occurs at r = 0)
3.3 Black branes
In order to understand better the family of supergravity solutions that we have
been discussing, let us now consider a generalization that lifts the degenerate
nature of the horizon. Written in Schwarzschildtype coordinates, one ﬁnds
the generalized “black brane” solution
27,28
ds
2
= −
Σ+
Σ
{
1−
4
˜
d
∆(D−2)
}
−
dt
2
+ Σ
4
˜
d
∆(D−2)
−
dx
i
dx
i
+
Σ
2a
2
∆
˜
d
−1
−
Σ+
d˜ r
2
+ ˜ r
2
Σ
2a
2
∆
˜
d
−
dΩ
2
D−d−1
e
ς∆
2a
φ
= Σ
−1
−
Σ
±
= 1 −
r±
˜ r
˜
d
.
black brane: Schwarzschildtype coordinates
(3.9)
The antisymmetric tensor ﬁeld strength for this solution corresponds to a
charge parameter λ = 2
˜
d/
√
∆(r
+
r
−
)
˜
d/2
, either electric or magnetic.
The characteristic feature of the above “blackened” pbranes is that they
have a nondegenerate, nonsingular outer horizon at ˜ r = r
+
, at which the light
cones “ﬂip over.” At ˜ r = r
−
, one encounters an inner horizon, which, however,
coincides in general with a curvature singularity. The singular nature of the
solution at ˜ r = r
−
is apparent in the scalar φ in (3.9). For solutions with p ≥ 1,
22
the singularity at the inner horizon persists even in cases where the scalar φ is
absent.
The extremal limit of the black brane solution occurs for r
+
= r
−
. When
a = 0 and scalars may consistently be set to zero, the singularity at the hori
zon r
+
= r
−
disappears and then one may analytically continue through the
horizon. In this case, the light cones do not “ﬂip over” at the horizon because
one is really crossing two coalesced horizons, and the coincident “ﬂips” of the
light cones cancel out.
The generally singular nature of the inner horizon of the nonextreme
solution (3.9) shows that the “location” of the pbrane in spacetime should
normally be thought to coincide with the inner horizon, or with the degenerate
horizon in the extremal case.
4 Charges, Masses and Supersymmetry
The pbrane solutions that we have been studying are supported by anti
symmetric tensor gauge ﬁeld strengths that fall oﬀ at transverse inﬁnity like
r
−(
˜
d+1)
, as one can see from (2.5, 2.25, 2.6). This asymptotic falloﬀ is slow
enough to give a nonvanishing total charge density from a Gauss’ law ﬂux in
tegral at transverse inﬁnity, and we shall see that, for the “extremal” class of
solutions that is our main focus, the mass density of the solution saturates a
“Bogomol’ny bound” with respect to the charge density. In this Section, we
shall ﬁrst make more precise the relation between the geometry of the pbrane
solutions, the pform charges U
AB
and V
ABCDE
and the scalar charge magni
tudes U and V (1.3, 1.4); we shall then discuss the relations between these
charges, the energy density and the preservation of unbroken supersymmetry.
4.1 pform charges
Now let us consider the inclusion of sources into the supergravity equations.
The harmonic function (2.21) has a singularity which has for simplicity been
placed at the origin of the transverse coordinates y
m
. As we have seen in
Sections 3.1 and 3.2, whether or not this gives rise to a physical singularity
in a solution depends on the global structure of that solution. In the electric
2brane case, the solution does in the end have a singularity.
26
This singularity
is unlike the Schwarzschild singularity, however, in that it is a timelike curve,
and thus it may be considered to be the wordvolume of a δfunction source.
The electric source that couples to D = 11 supergravity is the fundamental
23
supermembrane action,
20
whose bosonic part is
I
source
= Q
e
W3
d
3
ξ
−det(∂
µ
x
M
∂
ν
x
N
g
MN
(x))
+
1
3!
ǫ
µνρ
∂
µ
x
M
∂
ν
x
N
∂
ρ
x
R
A
MNR
(x)
. (4.1)
The source strength Q
e
will shortly be found to be equal to the electric charge
U upon solving the coupled equations of motion for the supergravity ﬁelds and
a single source of this type. Varying the source action (4.1) with
δ
/δA
[3]
, one
obtains the δfunction current
J
MNR
(z) = Q
e
W3
δ
3
(z −x(ξ))dx
M
∧ dx
N
∧ dx
R
. (4.2)
This current now stands on the RHS of the A
[3]
equation of motion:
d(
∗
F
[4]
+
1
2
A
[3]
∧ F
[4]
) =
∗
J
[3]
. (4.3)
Thus, instead of the Gauss’ law expression for the charge, one may instead
rewrite the charge as a volume integral of the source,
U =
M8
∗
J
[3]
=
1
3!
M8
J
0MN
d
8
S
MN
, (4.4)
where d
8
S
MN
is the 8volume element on M
8
, speciﬁed within a D = 10
spatial section of the supergravity spacetime by a 2form. The charge derived
in this way from a single 2brane source is thus U = Q
e
as expected.
Now consider the eﬀect of making diﬀerent choices of the M
8
integration
volume within the D = 10 spatial spacetime section, as shown in Figure 4. Let
the diﬀerence between the surfaces M
8
and M
′
8
be inﬁnitesimal and be given
by a vector ﬁeld v
N
(x). The diﬀerence in the electric charges obtained is then
given by
δU =
M8
L
v
∗
J
[3]
=
1
3!
∂M8
J
0MN
v
R
d
7
S
MNR
, (4.5)
where L
v
is the Lie derivative along the vector ﬁeld v. The second equality in
(4.5) follows using Stokes’ theorem and the conservation of the current J
[3]
.
Now a topological nature of the charge integral (1.3) becomes apparent;
similar considerations apply to the magnetic charge (1.4). As long as the
current J
[3]
vanishes on the boundary ∂M
8
, the diﬀerence (4.5) between the
charges calculated using the integration volumes M
8
and M
′
8
will vanish.
This divides the electriccharge integration volumes into two topological classes
24
J
[3]
M
8
M
8
'
v(x)
Figure 4: Diﬀerent choices of charge integration volume “capturing” the current J
[3]
.
distinguishing those for which ∂M
8
“captures” the pbrane current, as shown
in Figure 4 and giving U = Q
e
, from those that do not capture the current,
giving U=0.
The above discussion shows that the orientationdependence of the U
charges (1.3) is essentially topological. The topological classes for the charge
integrals are naturally labeled by the asymptotic orientations of the pbrane
spatial surfaces; an integration volume M
8
extending out to inﬁnity ﬂips from
the “capturing” class into the “noncapturing” class when ∂M
8
crosses the
δfunction surface deﬁned by the current J
[3]
. The charge thus naturally has
a magnitude Q
[p]
 = Q
e
and a unit pform orientation Q
[p]
/Q
[p]
 that is pro
portional to the asymptotic spatial volume form of the pbrane. Both the
magnitude and the orientation of this pform charge are conserved using the
supergravity equations of motion.
The necessity of considering asymptotic pbrane volume forms arises be
cause the notion of a pform charge is not limited to static, ﬂat pbrane solutions
such as (2.3, 2.5, 2.6). Such charges can also be deﬁned for any solution whose
energy diﬀers from that of a ﬂat, static one by a ﬁnite amount. The charges
25
for such solutions will also appear in the supersymmetry algebra (1.5) for such
backgrounds, but the corresponding energy densities will not in general sat
urate the BPS bounds. For a ﬁnite energy diﬀerence with respect to a ﬂat,
static pbrane, the asymptotic orientation of the pbrane volume form must
tend to that of a static ﬂat solution, which plays the rˆole of a “BPS vacuum”
in a given pform charge sector of the theory.
In order to have a nonvanishing value for a charge (1.3) or (1.4) occur
ring in the supersymmetry algebra (1.5), the pbrane must be either inﬁnite or
wrapped around a compact spacetime dimension. The case of a ﬁnite pbrane
is sketched in Figure 5. Since the boundary ∂Mof the inﬁnite integration vol
ume Mdoes not capture the locus where the pbrane current is nonvanishing,
the current calculated using M will vanish as a result. Instead of an inﬁnite
pbrane, one may alternately have a pbrane wrapped around a compact di
mension of spacetime, so that an integrationvolume boundary ∂M
8
is still
capable of capturing the pbrane locus (if one considers this case as an inﬁnite,
but periodic, solution, this case may be considered simultaneously with that
of the inﬁnite pbranes). Only in such cases do the pform charges occurring
in the supersymmetry algebra (1.5) take nonvanishing values.
e
J
M
Figure 5: Finite pbrane not captured by ∂M, giving zero charge.
e
If one considers integration volumes that do not extend out to inﬁnity, then one can con
struct integration surfaces that capture ﬁnite pbranes. Such charges do not occur in the
supersymmetry algebra (1.5), but they are still of importance in determining the possible
intersections of pbranes.
21
26
4.2 pbrane mass densities
Now let us consider the mass density of a pbrane solution. Since the pbrane
solutions have translational symmetry in their p spatial worldvolume direc
tions, the total energy as measured by a surface integral at spatial inﬁnity
diverges, owing to the inﬁnite extent. What is thus more appropriate to con
sider instead is the value of the density, energy/(unit pvolume). Since we
are considering solutions in their rest frames, this will also give the value of
mass/(unit pvolume), or tension of the solution. Instead of the standard spa
tial d
D−2
Σ
a
surface integral, this will be a d
(D−d−1)
Σ
m
surface integral over
the boundary ∂M
T
of the transverse space.
The ADM formula for the energy density written as a Gauss’law integral
(see, e.g., Ref.
16
) is, dropping the divergent spatial dΣ
µ=i
integral,
E =
∂MT
d
D−d−1
Σ
m
(∂
n
h
mn
−∂
m
h
b
b
) , (4.6)
written for g
MN
= η
MN
+h
MN
tending asymptotically to ﬂat space in Cartesian
coordinates, and with a, b spatial indices running over the values µ = i =
1, . . . , d − 1; m = d, . . . , D − 1. For the general pbrane solution (2.24), one
ﬁnds
h
mn
=
4kd
∆(D −2)r
˜
d
δ
mn
, h
b
b
=
8k(d +
1
2
˜
d)
∆(D −2)r
˜
d
, (4.7)
and, since d
(D−d−1)
Σ
m
= r
˜
d
y
m
dΩ
(D−d−1)
, one ﬁnds
E =
4k
˜
dΩ
D−d−1
∆
, (4.8)
where Ω
D−d−1
is the volume of the S
D−d−1
unit sphere. Recalling that k =
√
∆λ/(2
˜
d), we consequently have a relation between the mass per unit p volume
and the charge parameter of the solution
E =
2λΩ
D−d−1
√
∆
. (4.9)
By contrast, the black brane solution (3.9) has E > 2λΩ
D−d−1
/
√
∆, so
the extremal pbrane solution (2.24) is seen to saturate the inequality E ≥
2λΩ
D−d−1
/
√
∆.
4.3 pbrane charges
As one can see from (4.8, 4.9), the relation (2.26) between the integration
constant k in the solution (2.24) and the charge parameter λ implies a deep
27
link between the energy density and certain electric or magnetic charges. In
the electric case, this charge is a quantity conserved by virtue of the equations
of motion for the antisymmetric tensor gauge ﬁeld A
[n−1]
, and has generally
become known as a “Page charge,” after its ﬁrst discussion in Ref.
2
To be
speciﬁc, if we once again consider the bosonic sector of D = 11 supergravity
theory (1.1), for which the antisymmetric tensor ﬁeld equation was given in
(3.1), one ﬁnds the Gauss’law form conserved quantity
2
U (1.3).
For the pbrane solutions (2.24), the
A ∧ F term in (1.3) vanishes. The
∗
F term does, however, give a contribution in the elementary/electric case,
provided one picks M
8
to coincide with the transverse space to the d = 3
membrane worldvolume, M
8T
. The surface element for this transverse space
is dΣ
m
(7)
, so for the p = 2 elementary membrane solution (3.2), one ﬁnds
U =
∂M8T
dΣ
m
(7)
F
m012
= λΩ
7
. (4.10)
Since the D = 11 F
[4]
ﬁeld strength supporting this solution has ∆ = 4, the
mass/charge relation is
E = U = λΩ
7
. (4.11)
Thus, like the classic extreme ReissnerNordstrom blackhole solution to which
it is strongly related (as can be seen from the CarterPenrose diagram given in
Figure 1), the D = 11 membrane solution has equal mass and charge densities,
saturating the inequality E ≥ U.
Now let us consider the charge carried by the solitonic/magnetic 5brane
solution (3.6). The ﬁeld strength in (3.6) is purely transverse, so no electric
charge (1.3) is present. The magnetic charge (1.4) is carried by this solution,
however. Once again, let us choose the integration subsurface so as to coincide
with the transverse space to the d = 6 worldvolume, i.e.
M
5
= M
5T
. Then,
we have
V =
∂M5T
dΣ
m
(4)
ǫ
mnpqr
F
npqr
= λΩ
4
. (4.12)
Thus, in the solitonic/magnetic 5brane case as well, we have a saturation of
the masscharge inequality:
E = V = λΩ
4
. (4.13)
4.4 Preserved supersymmetry
Since the bosonic solutions that we have been considering are consistent trun
cations of D = 11 supergravity, they must also possess another conserved
28
quantity, the supercharge. Admittedly, since the supercharge is a Grassma
nian (anticommuting) quantity, its value will clearly be zero for the class of
purely bosonic solutions that we have been discussing. However, the func
tional form of the supercharge is still important, as it determines the form of
the asymptotic supersymmetry algebra. The Gauss’law form of the super
charge is given as an integral over the boundary of the spatial hypersurface.
For the D = 11 solutions, this surface of integration is the boundary at inﬁnity
∂M
10
of the D = 10 spatial hypersurface; the supercharge is then
1
Q =
∂M10
Γ
0bc
ψ
c
dΣ
(9)b
. (4.14)
One can also rewrite this in fully Lorentzcovariant form, where dΣ
(9)b
=
dΣ
(9)0b
→dΣ
(9)AB
:
Q =
∂M10
Γ
ABC
ψ
C
dΣ
(9)AB
. (4.15)
After appropriate deﬁnitions of Poisson brackets, the D = 11 supersym
metry algebra for the supercharge (4.14, 4.15) is found to be given
29
by (1.5)
Thus, the supersymmetry algebra wraps together all of the conserved Gauss’
law type quantities that we have discussed.
The positivity of the Q
2
operator on the LHS of the algebra (1.5) is at the
root of the Bogomol’ny bounds
30,26,32
E ≥ (2/
√
∆)U electric bound (4.16a)
E ≥ (2/
√
∆)V magnetic bound (4.16b)
that are saturated by the pbrane solutions.
The saturation of the Bogomol’ny inequalities by the pbrane solutions
is an indication that they ﬁt into special types of supermultiplets. All of
these boundsaturating solutions share the important property that they leave
some portion of the supersymmetry unbroken. Within the family of pbrane
solutions that we have been discussing, it turns out
32
that the ∆ values of
such “supersymmetric” pbranes are of the form ∆ = 4/N, where N is the
number of antisymmetric tensor ﬁeld strengths participating in the solution
(distinct, but of the same rank). The diﬀerent charge contributions to the
supersymmetry algebra occurring for diﬀerent values of N (hence diﬀerent ∆)
aﬀect the Bogomol’ny bounds as shown in (4.16).
In order to see how a purely bosonic solution may leave some portion of
the supersymmetry unbroken, consider speciﬁcally once again the membrane
solution of D = 11 supergravity.
18
This theory
1
has just one spinor ﬁeld, the
29
gravitino ψ
M
. Checking for the consistency of setting ψ
M
= 0 with the sup
position of some residual supersymmetry with parameter ǫ(x) requires solving
the equation
δψ
A

ψ=0
=
˜
D
A
ǫ = 0 , (4.17)
where ψ
A
= e
A
M
ψ
M
and
˜
D
A
ǫ = D
A
ǫ −
1
288
(Γ
A
BCDE
−8δ
A
B
Γ
CDE
) F
BCDE
ǫ
D
A
ǫ = (∂
A
+
1
4
ω
A
BC
Γ
BC
)ǫ . (4.18)
Solving the equation
˜
D
A
ǫ = 0 amounts to ﬁnding a Killing spinor ﬁeld in the
presence of the bosonic background. Since the Killing spinor equation (4.17) is
linear in ǫ(x), the Grassmanian (anticommuting) character of this parameter
is irrelevant to the problem at hand, which thus reduces eﬀectively to solving
(4.17) for a commuting quantity.
In order to solve the Killing spinor equation (4.17) in a pbrane background,
it is convenient to adopt an appropriate basis for the D = 11 Γ matrices. For
the d = 3 membrane background, one would like to preserve SO(2, 1) ×SO(8)
covariance. An appropriate basis that does this is
Γ
A
= (γ
µ
⊗Σ
9
, 1l
(2)
⊗Σ
m
) , (4.19)
where γ
µ
and 1l
(2)
are 2 × 2 SO(2, 1) matrices; Σ
9
and Σ
m
are 16 × 16 SO(8)
matrices, with Σ
9
= Σ
3
Σ
4
. . . Σ
10
, so Σ
2
9
= 1l
(16)
. The most general spinor ﬁeld
consistent with (Poincar´e)
3
× SO(8) invariance in this spinor basis is of the
form
ǫ(x, y) = ǫ
2
⊗η(r) , (4.20)
where ǫ
2
is a constant SO(2, 1) spinor and η(r) is an SO(8) spinor depending
only on the isotropic radial coordinate r; η may be further decomposed into
Σ
9
eigenstates by the use of
1
2
(1l ±Σ
9
) projectors.
Analysis of the the Killing spinor condition (4.17) in the above spinor basis
leads to the following requirements
12,18
on the background and on the spinor
ﬁeld η(r):
1) The background must satisfy the conditions 3A
′
+ 6B
′
= 0 and C
′
e
C
=
3A
′
e
3A
. The ﬁrst of these conditions is, however, precisely the linearity
condition reﬁnement (2.15) that we made in the pbrane ansatz; the
second condition follows from the ansatz reﬁnement (2.17) (considered
as a condition on φ
′
/a) and from (2.22). Thus, what appeared previously
to be simplifying specializations in the derivation given in Section 2 turn
out in fact to be conditions required for supersymmetric solutions.
30
2) η(r) = H
−1/6
(y)η
0
= e
C(r)/6
η
0
, where η
0
is a constant SO(8) spinor.
Thus, the surviving local supersymmetry parameter ǫ(x, y) must take
the form ǫ(x, y) = H
−1/6
ǫ
∞
, where ǫ
∞
= ǫ
2
⊗ η
0
. Note that, after
imposing this requirement, at most a ﬁnite number of parameters can
remain unﬁxed in the product spinor ǫ
2
⊗ η
0
; i.e. the local supersym
metry of the D = 11 theory is almost entirely broken by any particular
solution. So far, the requirement (4.17) has cut down the amount of sur
viving supersymmetry from D = 11 local supersymmetry (i.e. eﬀectively
an inﬁnite number of components) to the ﬁnite number of independent
components present in ǫ
2
⊗ η
0
. The maximum number of such rigid un
broken supersymmetry components is achieved for D = 11 ﬂat space,
which has a full set of 32 constant components.
3) (1l + Σ
9
)η
0
= 0, so the constant SO(8) spinor η
0
is also required to
be chiral.
f
This cuts the number of surviving parameters in the product
ǫ
∞
= ǫ
2
⊗η
0
by half: the total number of surviving rigid supersymmetries
in ǫ(x, y) is thus 2·8 = 16 (counting real spinor components). Since this is
half of the maximum rigid number (i.e. half of the 32 for ﬂat space), one
says that the membrane solution preserves “half” of the supersymmetry.
In general, the procedure for checking how much supersymmetry is pre
served by a given BPS solution follows steps analogous to points 1) – 3) above:
ﬁrst a check that the conditions required on the background ﬁelds are satisﬁed,
then a determination of the functional form of the supersymmetry parameter
in terms of some ﬁnite set of spinor components, and ﬁnally the imposition of
projection conditions on that ﬁnite set. In a more telegraphic partial discus
sion, one may jump straight to the projection conditions 3). These must, of
course, also emerge from a full analysis of equations like (4.17). But one can
also see more directly what they will be simply by considering the supersym
metry algebra (1.5), specialized to the BPS background. Thus, for example,
in the case of a D = 11 membrane solution oriented in the {012} directions,
one has, after normalizing to a unit 2volume,
1
2vol
{Q
α
, Q
β
} = −(CΓ
0
)
αβ
E + (CΓ
12
)
αβ
U
12
. (4.21)
Since, as we have seen in (4.11), the membrane solution saturating the Bogo
mol’ny bound (4.16a) with E = U = U
12
, one may rewrite (4.21) as
1
2vol
{Q
α
, Q
β
} = 2EP
012
P
012
=
1
2
(1l + Γ
012
) , (4.22)
f
The speciﬁc chirality indicated here is correlated with the sign choice made in the elemen
tary/electric form ansatz (2.4); one may accordingly observe from (1.1) that a D = 11 parity
transformation requires a sign ﬂip of A
[3]
.
31
where P
012
is a projection operator (i.e. P
2
012
= P
012
) whose trace is trP
012
=
1
2
· 32; thus, half of its eigenvalues are zero, and half are unity. Any surviving
supersymmetry transformation must give zero when acting on the BPS back
ground ﬁelds, and so the anticommutator {Q
α
, Q
β
} of the generators must give
zero when contracted with a surviving supersymmetry parameter ǫ
α
. From
(4.22), this translates to
P
012
ǫ
∞
= 0 , (4.23)
which is equivalent to condition 3) above, (1l + Σ
9
)η
0
= 0. Thus, we once
again see that the D = 11 supermembrane solution (3.2) preserves half of
the maximal rigid D = 11 supersymmetry. When we come to discuss the
cases of “intersecting” pbranes in Section 7, it will be useful to have quick
derivations like this for the projection conditions that must be satisﬁed by
surviving supersymmetry parameters.
More generally, the positive semideﬁniteness of the operator {Q
α
, Q
β
}
is the underlying principle in the derivation
30,26,32
of the Bogomol’ny bounds
(4.16). A consequence of this positive semideﬁniteness is that zero eigenvalues
correspond to solutions that saturate the Bogomol’ny inequalities (4.16), and
these solutions preserve one component of unbroken supersymmetry for each
such zero eigenvalue.
Similar consideration of the solitonic/magnetic 5brane solution
25
(3.6)
shows that it also preserves half the rigid D = 11 supersymmetry. In the 5
brane case, the analogue of condition 2) above is ǫ(x, y) = H
−1/12
(y)ǫ
∞
, and
the projection condition following from the algebra of preserved supersymme
try generators for a 5brane oriented in the {012345} directions is P
012345
ǫ
∞
=
0, where P
012345
=
1
2
(1l + Γ
012345
).
5 The super pbrane worldvolume action
We have already seen the bosonic part of the action for a supermembrane in
background gravitational and 3form ﬁelds in Eq. (4.1). We shall now want to
extend this treatment to the full set of bosonic and fermionic variables of an
important class of super pbranes. This class consists of those branes whose
worldvolume variables are just the bosonic and fermionic coordinates of the
super pbrane in the target superspace. In the early days of research on super
p branes, this was the only class known, but it is now recognized that more
general kinds of worldvolume multiplets can also occur, such as those involving
higher form ﬁelds, requiring a generalization of the formalism that we shall now
present.
Let us reformulate in HoweTucker form the bosonic part of the pbrane
action coupled to gravity alone, by introducing an independent worldvolume
32
metric γ
ij
:
33
I
HT
=
1
2
d
p+1
ξ
−det γ
γ
ij
∂
i
x
m
∂
j
x
n
g
mn
−(p −1)
, (5.1)
where the index ranges are i = 0, 1, . . . , p and m = 0, 1, . . . , D−1. The equation
of motion following from (5.1) for the worldvolume position variables x
m
is
γ
ij
∂
i
∂
j
x
m
−
k
ij
(γ) ∂
k
x
m
+∂
i
x
p
∂
j
x
q
Γ
m
pq
(g)
= 0 , (5.2)
where
¸
k
ij
¸
(γ) and Γ
m
pq
(g) are the Christoﬀel connections for the worldvolume
metric γ
ij
(ξ) and the spacetime metric g
mn
(x) respectively. The γ ﬁeld equa
tion is
γ
ij
= ∂
i
x
m
∂
j
x
n
g
mn
(x) , (5.3)
which states that γ
ij
is equal to the metric induced from the spacetime metric
g
mn
through the embedding x
m
(ξ). Inserting (5.3) into (5.2), one obtains
the same equation as that following from the NambuGoto form action that
generalizes (4.1)
I
NG
=
d
p+1
ξdet
1/2
(∂
i
x
m
∂
j
x
n
g
mn
(x)) , (5.4)
thus demonstrating the classical equivalence of (5.1) and (5.4). Note the es
sential appearance here of a “cosmological term” in the worldvolume action
for p = 1. The absence of this term in the speciﬁc case of the string, p = 1,
is at the origin of the worldvolume Weyl symmetry which is obtained only in
the string case.
Now generalize the target space to superspace Z
M
= (x
m
, θ
α
) and describe
the supergravity background by vielbeins E
A
M
(Z) and a superspace (p+1) form
B = [(p +1)!]
−1
E
A1
. . . E
Ap+1
B
Ap+1...A1
, where E
A
= dZ
M
E
A
M
are superspace
vielbein 1forms. The superspace world indices M and tangent space indices A
run both through bosonic values m, a = 0, 1, . . . (D − 1) and fermionic values
α appropriate for the corresponding spinor dimensionality.
The pbrane worldvolume is a map z
M
(ξ) from the space of the world
volume parameters ξ
i
to the target superspace; this map may be used to pull
back forms to the worldvolume: E
A
= dξ
i
E
A
i
, where E
A
i
= ∂
i
z
M
(ξ)E
A
M
(z(ξ).
Using these, one may write the super pbrane action in GreenSchwarz form:
I =
d
p+1
ξ
1
2
√
−γ[γ
ij
E
a
i
E
b
j
η
ab
−(p −1)]
+
1
(p + 1)!
ǫ
i1···ip+1
E
A1
i1
· · · E
Ap+1
ip+1
B
Ap+1···A1
¸
, (5.5)
33
where γ is the traditional shorthand for det γ
ij
.
Writing the super pbrane action in this manifestly targetspace supersym
metric form raises the question of how the expected supersymmetric balance of
bosonic and fermionic degrees of freedom can be achieved on the worldvolume.
For example, in the case of the D = 11, p = 2 supermembrane
20
, the D = 11
spinor dimensionality is 32, while the bosonic coordinates take only 11 val
ues. Now, in order to compare correctly the worldvolume degrees of freedom,
one should ﬁrst remove the worldvolume gauge degrees of freedom, which thus
subtracts from the above account three worldvolume reparameterizations for
the ξ
i
, leaving 11 −3 = 8 bosonic nongauge degrees of freedom. The fermions
are not in fact expected to match this number, because they are expected
to satisfy ﬁrstorder equations of motion on the worldvolume, as opposed to
the secondorder equations expected for the bosons. But, multiplying by 2 in
order to take account of this diﬀerence, one would still be expecting to have
16 active worldvolume fermionic degrees of freedom, instead of the a priori
32. The diﬀerence can only be accounted for by an additional fermionic gauge
symmetry.
This fermionic gauge symmetry is called “κ symmetry” and its general
implementation remains something of a mystery. In some formalisms, it can
be related to a standard worldvolume supersymmetry,
34
but this introduces
additional twistorlike variables that obscure somewhat the physical content
of the theory. The most physically transparent formalism is the original one
of Ref.
35
, where the κ symmetry parameter has a spacetime spinor index just
like the spinor variable θ, but the κ transformation involves a projector that
reduces the number of degrees of freedom removed from the spectrum by
1
2
.
Let δz
A
= dz
M
E
A
M
and consider a transformation such that
δz
a
= 0 , δz
α
= (1 + Γ)
α
β
κ
β
(ξ) , (5.6)
where κ
β
(ξ) is an anticommuting spacetime spinor parameter and
Γ =
(−1)
(p+1)(p−2)
4
(p + 1)!
√
γ
ǫ
i1...ip+1
E
a1
i1
. . . E
ap+1
ip+1
Γ
a1...ap+1
, (5.7)
where Γ
a1...ap+1
is the antisymmetrized product of (p + 1) gamma matrices,
taken “strength one,” i.e. with a normalization factor
1
(p+1)!
. In order to see
the projection property of
1
2
(1 +Γ), let γ
ij
be given by the solution to its ﬁeld
equation, i.e. γ
ij
= E
a
i
E
b
j
η
ab
, from which one obtains Γ
2
= 1, so that
1
2
(1 +Γ)
is indeed a projector.
Detailed analysis of the conditions for κinvariance
35
show that the super
34
pbrane action (5.5) is κinvariant provided the following conditions hold:
g
i) The ﬁeld strength H = dB =
1
(p+1)!
E
Ap+1
· · · E
A1
H
A1···Ap+1
satisﬁes the
constraints
H
αap+1···a1
= 0 (5.8a)
H
αβγap−1···a1
= 0 (5.8b)
H
αβap···a1
=
(−1)
p+
1
4
(p+1)(p−2)
2p!
(Γ
a1···ap
)
αβ
, (5.8c)
where we are using a notation in which the spinor indices on (Γ
a1···ap
)
αβ
are raised and lowered by the charge conjugation matrix (e.g. (Γ
a
)
αβ
=
(Γ
a
)
α
γ
C
γβ
).
ii) The superspace torsion satisﬁes
η
c(a
T
c
b)α
= 0 (5.9a)
T
a
αβ
= (Γ
a
)
αβ
. (5.9b)
iii) H is closed.
Now observe a remarkable consequence of the conditions (5.8,5.9) in a max
imally supersymmetric theory (e.g. D = 11 supergravity or one of the D = 10
N = 2 superstring theories): in a general background, these constraints im
ply the supergravity equations of motion. This is a dramatic link between the
super pbranes and their parent supergravity theories – these supersymmetric
extended objects are, on the one hand, the natural sources for the correspond
ing supergravities, and on the other hand their consistent propagation (i.e.
preservation of κ symmetry) requires the backgrounds in which they move to
satisfy the supergravity equations of motion. This is a link between supersym
metric objects and the corresponding parent supergravities that is even more
direct than that found in quantized string theories, where the beta function
conditions enforcing the vanishing of worldvolume conformal anomalies impose
a set of eﬀective ﬁeld equations on the background. For the super pbranes of
maximal supergravities, this link arises already at the classical level.
h
g
Strictly speaking, what one learns from the requirements for κ symmetry allows for terms
involving a spinor Λα on the RHS of Eqs. (5.8a) and (5.9a), but this spinor can then be set
to zero by a judicious choice of the conventional superspace constraints.
22
h
This observation in turn poses an unresolved question. The beta function conditions for
vanishing of the string conformal anomalies naturally generate quantum corrections to the
eﬀective ﬁeld equations, but it is not fully understood how the preservation of κ symmetry
is achieved in the presence of quantum corrections.
35
To understand the import of κsymmetry better, note that in ﬂat super
space, conditions i) and ii) apply automatically. Moreover, in ﬂat superspace,
one has
E
A
= {(dx
m
−i
¯
θΓ
m
dΘ)δ
a
m
, dθ
µ
δ
a
µ
} (5.10)
so that condition iii), the closure of H, requires
(d
¯
θΓ
a
dθ)(d
¯
θΓ
ab1···bp−1
dθ) = 0 . (5.11)
Noting that the diﬀerential dθ is commuting and that for consistency with
the H constraints (5.8), one must have (Γ
ab1···bp−1
)
αβ
symmetric in (αβ), it
follows that the second factor d
¯
θΓ
ab1···bp−1
dθ in (5.11) does not vanish of its
own accord. Consequently, what one requires for (5.11) to hold is the gamma
matrix condition
(Γ
a
P)
(αβ
(Γ
ab1···bp−1
P)
γδ)
= 0 , (5.12)
where P is a chirality projector that is required if the spinor coordinate θ is
MajoranaWeyl, but is the unit matrix otherwise. Analysis
36
of this constraint
shows it to hold in the (D, p) spacetime/worldvolume dimensions shown in Fig
ure 6). The (D, p) cases in which (5.12) holds are also related to the existence
of “twocomponent notations” over the various division algebras R, C, H, O.
Examples of this may be seen in the p = 1, D = 3, 4, 6, 10 superstring cases:
in D = 3, the Lorentz group is SL(2, R), while in D = 4 it is SL(2, C) and in
D = 6 it is SL(2, H); in D = 10, there is an analogous relation between the
Lorentz group and the quaternions O, although the nonassociative nature of
the quaternions makes this rather more cumbersome. Writing out the gamma
matrix identity (5.12) in these cases using twocomponent notation makes its
proof relatively transparent, as a moment’s consideration of the D = 3 case
shows, keeping in mind that the dθ
α
fermionic oneforms are commuting, while
the SL(2, R) invariant tensor ǫ
αβ
that would be needed to contract indices is
antisymmetric. Similar considerations apply to the p = 0, superparticle cases.
36
It should be noted that the superparticle and superstring cases allow for min
imal and also extended supersymmetries. This possibility arises because there
exist corresponding extended worldvolume multiplets constructed from spinors
and scalars alone.
An essential import of the κ symmetry for the super pbrane actions is that
it allows one to gauge away half of the spinor variables, allowing thus gauge
choices of the form θ = (S, 0). After this reduction by half in the number
of nongauge fermion worldvolume ﬁelds, one can achieve a balance between
the worldvolume bosonic and fermionic degrees of freedom, as required by
the residual unbroken supersymmetry. The counting of degrees of freedom in
36
D
p
11
10
9
8
7
6
5
4
3
2
0 1 2 3 4 5
♦
♦
♦
♠
♠
♠
♠
♠
♠
♥
♥
♥
♥
♣
♣
♣
♦
♠
♥
♣
H
C
R
O
Figure 6: Sequences of kinvariant pbrane actions involving worldvolume spinors and scalars
only. The four sequences ♣, ♥, ♠, ♦ of (D, p) values correspond to solutions to the gamma
matrix identity (5.12) and are associated to the division algebras R, C, H, O. There also
exist other worldvolume actions, corresponding to the diﬀerent brane types shown in Figure
7, but these involve worldvolume ﬁelds other than the simple scalarspinor multiplets shown
here.
superstring cases needs a little more care, however, because the bosonic world
volume degrees of freedom can be split into chiral left and rightmoving modes.
For example, in the minimal N = 1 superstring in D = 10, the MajoranaWeyl
spinor variable yields superpartners only for one chirality amongst the bosonic
modes, while the other chirality modes occur as supersymmetry singlets. In
the N = 2 superstrings, however, both chiralities of bosonic modes are paired
with spinors.
One might ﬁnd it strange to have an unbroken worldvolume supersymme
try given the fact that the spinor pbrane variables θ
α
are targetspace spinors,
but worldvolume scalars. It would seem that a worldvolume supersymmetry
should require the presence of spinor ﬁelds on the worldvolume. To see how
such worldvolume spinors appear, one needs to ﬁx the worldvolume reparam
eterization symmetry. This may conveniently be achieved using the “static”
gauge choice x
i
(ξ) = ξ
i
, i = 0, 1, . . . , p. If one now makes a Lorentz transfor
mation in the target spacetime, this gauge condition will generally be broken,
so in order to maintain this condition one needs to make a compensating gauge
transformation. Splitting the index range after the static gauge choice accord
ing to the pattern x
m
= (x
i
, x
˜ m
), ˜ m = p +1, . . . , D−1, and making a Lorentz
37
transformation with parameters (L
i
j
, L
i
˜ m
, L
˜ m
˜ n
), a combined Lorentz trans
formation and worldvolume reparameterization (with parameter η
i
(ξ)) on the
x
i
variables takes the form
δx
i
= η
j
(ξ)∂
j
X
i
+L
i
j
x
j
+L
i
˜ m
x
˜ m
. (5.13)
Demanding that δx
i
= 0 for x
i
= ξ
i
requires the compensating reparameteri
zation to be given by η
i
= −L
i
j
ξ
j
−L
i
˜ m
x
˜ m
. Then the combined transformation
for the remaining bosonic variables x
˜ m
is
δx
˜ m
= −L
k
j
ξ
j
∂
k
x
˜ m
+L
˜ m
˜ n
x
˜ n
−L
k
˜ n
x
˜ n
∂
k
x
˜ m
. (5.14)
This combined transformation remains linearly realized (i.e. “unbroken”) for
a subgroup SO(p, 1) ×SO(D−p−1), but the remaining generators of SO(D−
1, 1) become nonlinearly realized. The linearly realized SO(p, 1) becomes the
worldvolume Lorentz group, according to which the x
˜ m
transform as a set of
D − p − 1 scalars (although they transform as a vector with respect to the
remaining SO(D − p − 1) factor, which is an “internal symmetry” from the
worldvolume point of view). Now consider the eﬀect of the linearly realized
SO(p, 1) on the θ(ξ) variables:
δθ = −L
i
j
ξ
j
∂
i
θ +
1
4
L
ij
Γ
ij
θ . (5.15)
The matrices Γ
ij
provide a spinor representation (in general reducible) of the
“worldvolume” SO(p, 1) algebra, so that θ can now be identiﬁed, after gauge
ﬁxing, as a worldvolume spinor.
35
Similar considerations apply to supersymmetry after κ symmetry ﬁxing.
Preserving the κ symmetry gauge requires a similar compensating κ symme
try transformation, causing some of the original targetspace supersymmetry
to become nonlinearly realized, but leaving still a linearly realized subgroup,
which may be identiﬁed as the unbroken worldvolume supersymmetry. For
the simple pbrane actions that we have been considering in this section, the
fraction of unbroken supersymmetry is
1
2
. A classic example of a worldvolume
supermultiplet with respect to such a worldvolume supersymmetry is provided
by the D = 11, p = 2 supermembrane: the x
˜ m
, ˜ m = 3, . . . , 11 and the S
residual spinor (32/2 = 16 components) form a worldvolume scalarspinor
multiplet with respect to the unbroken worldvolume supersymmetry. This
multiplet contains 8 + 8 bosonic + fermionic degrees of freedom, transform
ing under 16 unbroken supercharges (it also has an unbroken SO(8) “internal”
symmetry). The existence of such a large multiplet of scalars and spinors is a
speciﬁc feature of the unbroken d = 3, “N = 8” supersymmetry.
38
6 KaluzaKlein dimensional reduction
Let us return now to the arena of purely bosonic ﬁeld theories, and consider
the relations between various bosonicsector theories and the corresponding
relations between pbrane solutions. It is wellknown that supergravity theories
are related by dimensional reduction from a small set of basic theories, the
largest of which being D = 11 supergravity. The spinor sectors of the theories
are equally well related by dimensional reduction, but in the following, we shall
restrict our attention to the purely bosonic sector.
In order to set up the procedure, let us consider a theory in (D + 1)
dimensions, but break up the metric in Ddimensionally covariant pieces:
dˆ s
2
= e
2αϕ
ds
2
+e
2βφ
(dz +A
M
dx
M
)
2
(6.1)
where carets denote (D+1)dimensional quantities corresponding to the (D+
1)dimensional coordinates x
ˆ M
= (x
M
, z); ds
2
is the line element in D dimen
sions and α and β are constants. The scalar ϕ in D dimensions emerges from
the metric in (D + 1) dimensions as (2β)
−1
ln g
zz
. Adjustment of the con
stants α and β is necessary to obtain desired structures in D dimensions. In
particular, one should pick β = −(D−2)α in order to arrange for the Einstein
frame form of the gravitational action in (D+1) dimensions to go over to the
Einsteinframe form of the action in D dimensions.
The essential step in a KaluzaKlein dimensional reduction is a consistent
truncation of the ﬁeld variables, generally made by choosing them to be inde
pendent of the reduction coordinate z. By a consistent truncation, we always
understand a restriction on the variables that commutes with variation of the
action to produce the ﬁeld equations, i.e. a restriction such that solutions to
the equations for the restricted variables are also solutions to the equations for
the unrestricted variables. This ensures that the lowerdimensional solutions
which we shall obtain are also particular solutions to higherdimensional su
pergravity equations as well. Making the parameter choice β = −(D−2)α to
preserve the Einsteinframe form of the action, one obtains
−ˆ gR(ˆ g) =
√
−g
R(g)−(D−1)(D−2)α
2
∇
M
ϕ∇
M
ϕ−
1
4
e
−2(D−1)αϕ
F
MN
F
MN
(6.2)
where F = dA. If one now chooses α
2
= [2(D − 1)(D − 2)]
−1
, the ϕ kinetic
term becomes conventionally normalized.
Next, one needs to establish the reduction ansatz for the (D + 1)dimen
sional antisymmetric tensor gauge ﬁeld
ˆ
F
[n]
= d
ˆ
A
[n−1]
. Clearly, among the
n − 1 antisymmetrized indices of
ˆ
A
[n−1]
at most one can take the value z, so
39
we have the decomposition
ˆ
A
[n−1]
= B
[n−1]
+B
[n−2]
∧ dz . (6.3)
All of these reduced ﬁelds are to be taken to be functionally independent of z.
For the corresponding ﬁeld strengths, ﬁrst deﬁne
G
[n]
= dB
[n−1]
(6.4a)
G
[n−1]
= dB
[n−2]
. (6.4b)
However, these are not exactly the most convenient quantities to work with,
since a certain “ChernSimons” structure appears upon dimensional reduction.
The metric in (D+1) dimensions couples to all ﬁelds, and, consequently, dimen
sional reduction will produce some terms with undiﬀerentiated KaluzaKlein
vector ﬁelds A
M
coupling to Ddimensional antisymmetric tensors. Accord
ingly, it is useful to introduce
G
′
[n]
= G
[n]
−G
[n−1]
∧ A , (6.5)
where the second term in (6.5) may be viewed as a ChernSimons correction
from the reduced Ddimensional point of view.
At this stage, we are ready to perform the dimensional reduction of our
general action (2.1). We ﬁnd that
ˆ
I =:
d
D+1
x
−ˆ g
R(ˆ g) −
1
2
∇
ˆ M
φ∇
ˆ M
φ −
1
2n!
e
ˆ aφ
ˆ
F
2
n]
(6.6)
reduces to
I =
d
D
x
√
−g
R −
1
2
∇
M
φ∇
M
φ −
1
2
∇
M
ϕ∇
M
ϕ −
1
4
e
−2(D−1)αϕ
F
2
[2]
−
1
2n!
e
−2(n−1)αϕ+ˆ αφ
G
′2
[n]
−
1
2(n −1)!
e
2(D−n)αϕ+ˆ aφ
G
2
[n−1]
. (6.7)
Although the dimensional reduction (6.7) has produced a somewhat compli
cated result, the important point to note is that each of the Ddimensional
antisymmetrictensor ﬁeld strength terms G
′2
[n]
and G
2
[n−1]
has an exponential
prefactor of the form e
ar
˜
φr
, where the
˜
φ
r
, r = (n, n − 1) are SO(2)rotated
combinations of ϕ and φ. Now, keeping just one while setting to zero the other
two of the three gauge ﬁelds (A
[1]
, B
[n−2]
, B
n−1]
), but retaining at the same
time the scalarﬁeld combination appearing in the corresponding exponential
prefactor, is a consistent truncation. Thus, any one of the three ﬁeld strengths
40
(F
[2]
, G
[n−1]
, G
′
[n]
), retained alone together with its corresponding scalarﬁeld
combination, can support pbrane solutions in D dimensions of the form that
we have been discussing.
An important point to note here is that, in each of the e
ar
˜
φ
prefactors, the
coeﬃcient a
r
satisﬁes
a
2
r
= ∆−
2d
r
˜
d
r
(D −2)
= ∆−
2(r −1)(D−r −1)
(D −2)
(6.8)
with the same value of ∆ as for the “parent” coupling parameter ˆ a, satisfying
ˆ a
2
r
= ∆−
2d
(n)
˜
d
(n)
((D + 1) −2)
= ∆−
2(n −1)(D −n)
(D −1
(6.9)
in D+1 dimensions. Thus, although the individual parameters a
r
are both D
and rdependent, the quantity ∆ is preserved under KaluzaKlein reduction for
both of the “descendant” ﬁeldstrength couplings (to G
′2
[n]
or to G
2
[n−1]
) coming
from the original term e
ˆ aφ
ˆ
F
2
[n]
. The 2form ﬁeld strength F
[2]
= dA, on the
other hand, emerges out of the gravitational action in D + 1 dimensions; its
coupling parameter corresponds to ∆ = 4.
If one retains in the reduced theory only one of the ﬁeld strengths (F
[2]
,
G
[n−1]
, G
′
[n]
), together with its corresponding scalarﬁeld combination, then
one ﬁnds oneself back in the situation described by our general action (2.1), and
then the p brane solutions obtained for the general case in Sec. 2 immediately
become applicable. Moreover, since retaining only one ﬁeld strength & scalar
combination in this way eﬀects a consistent truncation of the theory, solutions
to this simple truncated system are also solutions to the untruncated theory,
and indeed are also solutions to the original (D+1)dimensional theory, since
the KaluzaKlein dimensional reduction is also a consistent truncation.
6.1 Multiple ﬁeldstrength solutions and the singlecharge truncation
After repeated single steps of KaluzaKlein dimensional reduction from D = 11
down to D dimensions, the metric takes the form
31,32
ds
2
11
= e
2
3
a·
φ
ds
2
D
+
¸
i
e
(
2
3
a−ai)·
φ
(h
i
)
2
(6.10a)
h
i
= dz
i
+A
i
[1]
+A
ij
[0]
dz
j
, (6.10b)
where the A
i
M
are a set of (11−D) KaluzaKlein vectors generalizing the vector
A
M
in (6.1), emerging from the higherdimensional metric upon dimensional
41
reduction. Once such KaluzaKlein vectors have appeared, subsequent dimen
sional reduction also gives rise to the zeroform gauge potentials A
ij
[0]
appearing
in (6.10b) as a consequence of the usual onestep reduction (6.3) of a 1form
gauge potential.
We shall also need the corresponding reduction of the
ˆ
F
[4]
ﬁeld strength
i
(where hatted quantities refer to the original, higher, dimension) and, for later
reference, we shall also give the reduction of its Hodge dual
ˆ∗
ˆ
F
[4]
:
ˆ
F
[4]
= F
[4]
+F
i
[3]
∧h
i
+
1
2
F
ij
[2]
∧h
i
∧h
j
+
1
6
F
ijk
[1]
∧h
i
∧h
j
∧h
k
(6.11a)
ˆ ∗
ˆ
F
[4]
= e
a·
φ ∗
F
[4]
∧v+e
ai·
φ ∗
F
i
[3]
∧v
i
+
1
2
e
aij·
φ ∗
F
ij
[2]
∧v
ij
+
1
6
e
a
ijk
·
φ ∗
F
ijk
[1]
∧v
ijk
,
(6.11b)
(noting that, since the Hodge dual is a metricdependent construction, expo
nentials of the dilatonic vectors
φ appear in the reduction of
ˆ∗
ˆ
F
[4]
) where the
forms v, v
i
, v
ij
and v
ijk
appearing in (6.11b) are given by
v =
1
(11−D)!
ǫ
i1···i11−D
h
i1
∧ · · · ∧ h
i11−D
v
i
=
1
(10−D)!
ǫ
ii2···i11−D
h
i2
∧ · · · ∧ h
i11−D
v
ij
=
1
(9−D)!
ǫ
iji3···i11−D
h
i3
∧ · · · ∧ h
i11−D
v
ijk
=
1
(8−D)!
ǫ
ijki4···i11−D
h
i4
∧ · · · ∧ h
i11−D
. (6.12)
Using (6.10, 6.11a), the bosonic sector of maximal supergravity (1.1) now
reduces to
31,32
I
D
=
d
D
x
√
−g
R −
1
2
(∂
φ)
2
−
1
48
e
a·
φ
F
2
[4]
−
1
12
¸
i
e
ai·
φ
(F
i
[3]
)
2
−
1
4
¸
i<j
e
aij·
φ
(F
ij
[2]
)
2
−
1
4
¸
i
e
bi·φ
(F
i
[2]
)
2
(6.13)
−
1
2
¸
i<j<k
e
a
ijk
·
φ
(F
ijk
[1]
)
2
−
1
2
¸
ij
e
bij·φ
(F
ij
[1]
)
2
+L
FFA
,
where i, j = 1, . . . , 11−D, and ﬁeld strengths with multiple i, j indices may be
taken to be antisymmetric in those indices since these “internal” indices arise
in the stepwise reduction procedure, and two equal index values never occur
in a multiindex sum. From (6.11), one sees that the “straightbacked” ﬁeld
i
Note that the lowerdimensional ﬁeld strengths F
[n]
include “ChernSimons” corrections
similar to those in (6.5).
42
strengths F
[4]
, F
i
[3]
, F
ij
[2]
and F
ijk
[1]
are descendants from F
[4]
in D = 11. The
“calligraphic” ﬁeld strengths F
i
[2]
, on the other hand, are the ﬁeld strengths for
the KaluzaKlein vectors A
i
M
appearing in (6.10b). Similarly, one also has a set
of 1form ﬁeld strengths F
ij
[1]
for the KaluzaKlein zeroform gauge potentials
A
ij
[0]
appearing in (6.10b).
The nonlinearity of the original D = 11 action (1.1) in the metric ten
sor produces a consequent nonlinearity in the (11 − D) dilatonic scalar ﬁelds
φ appearing in the exponential prefactors of the antisymmetrictensor kinetic
terms in (6.13). For each ﬁeldstrength kinetic term in (6.13), there is a cor
responding “dilaton vector” of coeﬃcients determining the linear combination
of the dilatonic scalars appearing in its exponential prefactor. For the 4, 3,
2 and 1form “straightbacked” ﬁeld strengths emerging from F
[4]
in D = 11,
these coeﬃcients are denoted correspondingly a, a
i
, a
ij
and a
ijk
; for the “cal
ligraphic” ﬁeld strengths corresponding to KaluzaKlein vectors and zeroform
gauge potentials emerging out of the metric, these are denoted
b
i
and
b
ij
. How
ever, not all of these dilaton vectors are independent; in fact, they may all be
expressed in terms of the 4form and 3form dilaton vectors a and a
ij
:
31,32
a
ij
= a
i
+a
j
−a b
i
= −a
i
+a
a
ijk
= a
i
+a
j
+a
k
−2a
b
ij
= −a
i
+a
j
.
(6.14)
Another important feature of the dilaton vectors is that they satisfy the fol
lowing dotproduct relations:
a · a =
2(11 −D)
D−2
a · a
i
=
2(8 −D)
D−2
(6.15)
a
i
· a
j
= 2δ
ij
+
2(6 −D)
D −2
.
Throughout this discussion, we have emphasized consistent truncations in
making simplifying restrictions of complicated systems of equations, so that
the solutions of a simpliﬁed system are nonetheless perfectly valid solutions
of the more complicated untruncated system. With the equations of motion
following from (6.13) we face a complicated system that calls for analysis in
simpliﬁed subsectors. Accordingly, we now seek a consistent truncation down
to a simpliﬁed system of the form (2.1), retaining just one dilatonic scalar
combination φ and one rankn ﬁeld strength combination F
[n]
constructed out
of a certain number N of “retained” ﬁeld strengths F
α[n]
, α = 1, . . . , N, (this
43
could possibly be a straightbacked/calligraphic mixture) selected from those
appearing in (6.13), with all the rest being set to zero.
32
Thus, we let
φ = nφ +
φ
⊥
, (6.16)
where n ·
φ
⊥
= 0; in the truncation we then seek to set consistently
φ
⊥
= 0.
We shall see that consistency for the retained ﬁeld strengths F
α[n]
requires
them all to be proportional.
32
We shall let the dot product matrix for the
dilaton vectors of the retained ﬁeld strengths be denoted M
αβ
=: a
α
· a
β
.
Consistency of the truncation requires that the φ
⊥
ﬁeld equation be satisﬁed:
φ
⊥
−
¸
α
Π
⊥
· a
α
(F
α[n]
)
2
= 0 , (6.17)
where Π
⊥
is the projector into the dilatonvector subspace orthogonal to the
retained dilaton direction n. Setting
φ
⊥
= 0 in (6.17) and letting the retained
F
α[n]
be proportional, one sees that achieving consistency is hopeless unless
all the e
aα·
φ
prefactors are the same, thus requiring
a
α
· n = a ∀α = 1, . . . , N , (6.18)
where the constant a will play the role of the dilatonic scalar coeﬃcient in the
reduced system (2.1). Given a set of dilaton vectors for retained ﬁeld strengths
satisfying (6.18), consistency of (6.17) with the imposition of
φ
⊥
= 0 requires
Π
⊥
·
¸
α
a
α
(F
α[n]
)
2
= 0 . (6.19)
This equation requires, for every point x
M
in spacetime, that the combination
¸
α
a
α
(F
α[n]
)
2
be parallel to n in the dilatonvector space. Combining this
with the requirement (6.18), one has
¸
α
a
α
(F
α[n]
)
2
= an
¸
α
(F
α[n]
)
2
. (6.20)
Taking then a dot product of this with a
β
, one has
¸
α
M
βα
(F
α[n]
)
2
= a
2
¸
α
(F
α[n]
)
2
. (6.21)
Detailed analysis
32
shows it to be suﬃcient to consider the cases where M
αβ
is invertible, so by applying M
−1
αβ
to (6.21), one ﬁnds
(F
α[n]
)
2
= a
2
¸
β
M
−1
αβ
¸
γ
(F
γ [n]
)
2
, (6.22)
44
and, indeed, we ﬁnd that the F
α[n]
must all be proportional. Summing on α,
one has
a
2
= (
¸
α,β
M
−1
αβ
)
−1
; (6.23)
one then deﬁnes the retained ﬁeldstrength combination F
[n]
so that
(F
α[n]
)
2
= a
2
¸
β
M
−1
αβ
(F
[n]
)
2
. (6.24)
The only remaining requirement for consistency of the truncation down to
the simpliﬁed (g
MN
, φ, F
[n]
) system (2.1) arises from the necessity to ensure
that the variation of the L
FFA
term in (6.13) is not inconsistent with setting
to zero the discarded dilatonic scalars and gauge potentials. In general, this
imposes a somewhat complicated requirement. In the present review, however,
we shall concentrate mainly on either purelyelectric cases satisfying the ele
mentary ansatz (2.4) or purelymagnetic cases satisfying the solitonic ansatz
(2.6). As one can see by inspection, for pure electric or magnetic solutions
of these sorts, the terms that are dangerous for consistency arising from the
variation of L
FFA
all vanish. Thus, for such solutions one may safely ignore
the complications of the L
FFA
term. This restriction to pure electric or mag
netic solutions does, however, leave out the very interesting cases of dyonic
solutions that exist in D = 8 and D = 4, upon which we shall comment later
on in Section 8.
After truncating down to the system (2.1), the analysis proceeds as in
Section 2. It turns out
32
that supersymmetric pbrane solutions arise when
the matrix M
αβ
for the retained F
α[n]
satisﬁes
M
αβ
= 4δ
αβ
−
2d
˜
d
D −2
, (6.25)
and the corresponding ∆ value for F
[n]
is
∆ =
4
N
, (6.26)
where we recall that N is the number of retained ﬁeld strengths. A gener
alization of this analysis leads to a classiﬁcation of solutions with more than
one independent retained scalarﬁeld combination.
32
We shall see in Section 7
that the N > 1 solutions to singlecharge truncated systems (2.1) may also
be interpreted as special solutions of the full reduced action (6.13) containing
N constituent ∆ = 4 brane components that just happen to have coincident
charge centers. Consequently, one may consider only the N = 1, ∆ = 4 solu
tions to be fundamental.
45
6.2 Diagonal dimensional reduction of pbranes
The family of pbrane solutions is ideally suited to interpretation as solutions
of KaluzaKlein reduced theories, because they are naturally independent of
the “worldvolume” x
µ
coordinates. Accordingly, one may let the reduction
coordinate z be one of the x
µ
. Consequently, the only thing that needs to be
done to such a solution in order to reinterpret it as a solution of a reduced
system (6.7) is to perform a Weyl rescaling on it in order to be in accordance
with the form of the metric chosen in the KaluzaKlein ansatz (6.10), which
was adjusted so as to maintain the Einsteinframe form of the gravitational
term in the dimensionally reduced action.
Upon making such a reinterpretation, elementary/solitonic pbranes in
(D+1) dimensions give rise to elementary/solitonic (p−1)branes in D dimen
sions, corresponding to the same value of ∆, as one can see from (6.8, 6.9). Note
that in this process, the quantity
˜
d is conserved, since both D and d reduce by
one. Reinterpretation of p brane solutions in this way, corresponding to stan
dard KaluzaKlein reduction on a worldvolume coordinate, proceeds diagonally
on a D versus d plot, and hence is referred to as diagonal dimensional reduc
tion. This procedure is the analogue, for supergravity ﬁeldtheory solutions,
of the procedure of double dimensional reduction
22
for pbrane worldvolume
actions, which can be taken to constitute the δfunction sources for singular
pbrane solutions, coupled in to resolve the singularities, as we discussed in
subsection 4.1.
6.3 Multicenter solutions and vertical dimensional reduction
As we have seen, translational Killing symmetries of pbrane solutions allow a
simultaneous interpretation of these ﬁeld conﬁgurations as solutions belonging
to several diﬀerent supergravity theories, related one to another by Kaluza
Klein dimensional reduction. For the original single pbrane solutions (2.24),
the only available translational Killing symmetries are those in the worldvol
ume directions, which we have exploited in describing diagonal dimensional
reduction above. One may, however, generalize the basic solutions (2.24) by
replacing the harmonic function H(y) in (2.21) by a diﬀerent solution of the
Laplace equation (2.20). Thus, one can easily extend the family of pbrane
solutions to multicenter pbrane solutions by taking the harmonic function to
be
H(y) = 1 +
¸
α
k
α
y −y
α

˜
d
k
α
> 0 . (6.27)
46
Once again, the integration constant has been adjusted to make H

∞
= 1 ↔
φ

∞
= 0. The generalized solution (6.27) corresponds to parallel and similarly
oriented pbranes, with all charge parameters λ
α
= 2
˜
dk
α
/
√
∆ required to be
positive in order to avoid naked singularities. The “centers” of the individual
“leaves” of this solution are at the points y = y
α
, where α ranges over any
number of centers. The metric and the electriccase antisymmetric tensor
gauge potential corresponding to (6.27) are given again in terms of H(y) by
(2.24a,2.25). In the solitonic case, the ansatz (2.6) needs to be modiﬁed so as
to accommodate the multicenter form of the solution:
F
m1...mn
= −
˜
d
−1
ǫ
m1...mnp
∂
p
¸
α
λ
α
y −y
α

˜
d
, (6.28)
which ensures the validity of the Bianchi identity just as well as (2.6) does.
The mass/(unit pvolume) density is now
E =
2Ω
D−d−1
√
∆
¸
α
λ
α
, (6.29)
while the total electric or magnetic charge is given by Ω
D−d−1
¸
λ
α
, so the
Bogomol’ny bounds (4.16) are saturated just as they are for the singlecenter
solutions (2.24). Since the multicenter solutions given by (6.27) satisfy the
same supersymmetrypreservation conditions on the metric and antisymmetric
tensor as (2.24), the multicenter solutions leave the same amount of super
symmetry unbroken as the singlecenter solution.
From a mathematical point of view, the multicenter solutions (6.27) exist
owing to the properties of the Laplace equation (2.20). From a physical point
of view, however, these static solutions exist as a result of cancellation between
attractive gravitational and scalarﬁeld forces against repulsive antisymmetric
tensor forces for the similarlyoriented pbrane “leaves.”
The multicenter solutions given by (6.27) can now be used to prepare
solutions adapted to dimensional reduction in the transverse directions. This
combination of a modiﬁcation of the solution followed by dimensional reduction
on a transverse coordinate is called vertical dimensional reduction
23
because it
relates solutions vertically on a D versus d plot.
j
In order to do this, we need
ﬁrst to develop translation invariance in the transverse reduction coordinate.
This can be done by “stacking” up identical p branes using (6.27) in a periodic
array, i.e. by letting the integration constants k
α
all be equal, and aligning the
j
Similar procedures have been considered in a number of articles in the literature; see, e.g.
Refs.
37
47
“centers” y
α
along some axis, e.g. the z axis. Singling out one “stacking axis”
in this way clearly destroys the overall isotropic symmetry of the solution,
but, provided the centers are all in a line, the solution will nonetheless remain
isotropic in the D−d −1 dimensions orthogonal to the stacking axis. Taking
the limit of a denselypacked inﬁnite stack of this sort, one has
¸
α
k
α
y −y
α

˜
d
−→
+∞
−∞
kdz
(ˆ r
2
+z
2
)
˜
d/2
=
˜
k
˜ r
˜
d−1
(6.30a)
ˆ r
2
=
D−2
¸
m=d
y
m
y
m
(6.30b)
ˆ
k =
√
πkΓ(
˜
d −
1
2
)
2Γ(
˜
d)
, (6.30c)
where ˆ r in (6.30b) is the radial coordinate for the D−d −1 residual isotropic
transverse coordinates. After a conformal rescaling in order to maintain the
Einstein frame for the solution, one can ﬁnally reduce on the coordinate z
along the stacking axis.
After stacking and reduction in this way, one obtains a pbrane solution
with the same worldvolume dimension as the original higherdimensional so
lution that was stacked up. Since the same antisymmetric tensors are used
here to support both the stacked and the unstacked solutions, and since ∆
is preserved under dimensional reduction, it follows that vertical dimensional
reduction from D to D − 1 spacetime dimensions preserves the value of ∆
just like the diagonal reduction discussed in the previous subsection. Note
that under vertical reduction, the worldvolume dimension d is preserved, but
˜
d = D−d −2 is reduced by one with each reduction step.
Combining the diagonal and vertical dimensional reduction trajectories of
“descendant” solutions, one ﬁnds the general picture given in the plot of Fig
ure 7. In this plot of spacetime dimension D versus worldvolume dimension
d, reduction families emerge from certain basic solutions that cannot be “ox
idized” back up to higherdimensional isotropic pbrane solutions, and hence
can be called “stainless” pbranes.
13
In Figure 7, these solutions are indicated
by the large circles, with the corresponding ∆ values shown adjacently. The
indication of elementary or solitonic type relates to solutions of supergravity
theories in versions with the lowest possible choice of rank (n ≤ D/2) for the
supporting ﬁeld strength, obtainable by appropriate dualization. Of course,
every solution to a theory obtained by dimensional reduction from D = 11
supergravity (1.1) may be oxidized back up to some solution in D = 11. We
shall see in Section 7 that what one obtains upon oxidation of the “stainless”
48
1
2
3
4
5
6
7
8
9
10
11
D
d 1 2 3 4 5 6 7 0
particle string membrane 3brane 4brane 5brane 6brane instanton
4 4 4 4 4
4 4
2 2
4/3,1' 4/3,1' 2
4/3 4/3
1,4/5,2/3,4/7
1,4/5, 2/3,4/7
1'
1/2 1/2
elementary
solitonic
KaluzaKlein descendants
4 ∆ values
selfdual
vertical reduction trajectories
diagonal reduction trajectories
}
"stainless"
Figure 7: Branescan of supergravity pbrane solutions (p ≤ (D −3))
49
solutions in Figure 7 falls into the interesting class of “intersecting branes”
built from four basic “elemental” solutions of D = 11 supergravity.
6.4 The geometry of (D −3)branes
The process of vertical dimensional reduction described in the previous sub
section proceeds uneventfully until one makes the reduction from a (D, d =
D − 3) solution to a (D − 1, d = D − 3) solution.
k
In this step, the integral
(6.30) contains an additive divergence and needs to be renormalized. This
is easily handled by putting ﬁnite limits ±L on the integral, which becomes
L
−L
d˜ z(r
2
+ ˜ z
2
)
−1/2
, and then by subtracting a divergent term 2 lnL before
taking the limit L → ∞. Then the integral gives the expected ln ˆ r harmonic
function appropriate to two transverse dimensions.
Before proceeding any further with vertical dimensional reduction, let us
consider some of the speciﬁc properties of (D −3)branes that make the next
vertical step down problematic. Firstly, the asymptotic metric of a (D − 3)
brane is not a globally ﬂat space, but only a locally ﬂat space. This distinction
means that there is in general a deﬁcit solid angle at transverse inﬁnity, which
is related to the total mass density of the (D − 3)brane.
38
This means that
any attempt to stack up (D−3)branes within a standard supergravity theory
will soon consume the entire solid angle at transverse inﬁnity, thus destroying
the asymptotic spacetime in the construction.
In order to understand the global structure of the (D−3)branes in some
more detail, consider the supersymmetric string in D = 4 dimensions.
12
In
D = 4, one may dualize the 2form A
µν
ﬁeld to a pseudoscalar, or axion,
ﬁeld χ, so such strings are also solutions to dilatonaxion gravity. The pbrane
ansatz gives a spacetime of the form M
4
= M
2
× Σ
2
, where M
2
is D = 2
Minkowski space. Supporting this string solution, one has the 2form gauge
ﬁeld A
µν
and the dilaton φ. These ﬁelds give rise to a ﬁeld stress tensor of the
form
T
µν
(A, φ) = −
1
16
(a
2
+ 4)∂
m
K∂
m
Kη
µν
T
mn
(A, φ) =
1
8
(a
2
−4)(∂
m
K∂
n
K −
1
2
δ
mn
(∂
p
K∂
p
K)) , (6.31)
where a is as usual the dilaton coupling parameter and e
−K
= H = 1 −
8GT ln(r), with r =
√
y
m
y
m
, m = 2, 3. If one now puts in an elementary
string source action, with the string aligned along the µ, ν = 0, 1 subspace, so
k
Solutions with worldvolume dimension two less than the spacetime dimension will be re
ferred to generally as (D −3)branes, irrespective of whether the spacetime dimension is D
or not.
50
that T
mn(source)
= 0, then one has the source stress tensor
T
µν (source)
=
−T
√
−g
d
2
ξ
√
−γγ
ij
∂
i
X
µ
∂
j
X
ν
e
−
1
2
aφ
δ(x −X) . (6.32)
By inspection of the ﬁeld solution, one has T
mn
(A, φ) = 0, while the contri
butions to T
µν
from the A
µν
and φ ﬁelds and also from the source (6.32) are
both of the form diag(ρ, −ρ). Thus, the overall stress tensor is of the form
T
MN
= diag(ρ, −ρ, 0, 0).
Consequently, the Einstein equation in the transverse m, n indices becomes
R
mn
−
1
2
g
mn
R = 0, since the transverse stress tensor components vanish.
This equation is naturally satisﬁed for a metric satisfying the pbrane ansatz,
because, as one can see from (2.11) with
˜
d = A
′
= 0, this causes the transverse
components of the Ricci tensor to be equal to the Ricci tensor of a D = 2
spacetime, for which R
mn
−
1
2
g
mn
R ≡ 0 is an identity, corresponding to the
fact that the usual Einstein action,
√
−gR, is a topological invariant in D = 2.
Accordingly, in the transverse directions, the equations are satisﬁed simply by
by 0 = 0.
In the worldsheet directions, the equations become
−
1
2
Rη
µν
= −8πGρη
µν
, (6.33)
or just
R = 16πGρ , (6.34)
and as we have already noted, R = R
mm
. Owing to the fact that the D = 2
Weyl tensor vanishes, the transverse space Σ
2
is conformally ﬂat; Eq. (6.34)
gives its conformal factor. Thus, although there is no sensible Einstein action in
the transverse D = 2, space, a usual form of the Einstein equation nonetheless
applies to that space as a result of the symmetries of the pbrane ansatz.
The above supersymmetric string solution may be compared to the cosmic
strings arising in gauge theories with spontaneous symmetry breaking. There,
the Higgs ﬁelds contributing to the energy density of the string are displaced
from their usual vacuum values to unbrokensymmetry conﬁgurations at a
stationary point of the Higgs potential, within a very small transversespace
region that may be considered to be the string “core.” Approximating this
by a delta function in the transverse space, the Ricci tensor and hence the
full curvature vanish outside the string core, so that one obtains a conical
spacetime, which is ﬂat except at the location of the string core. The total
energy is given by the deﬁcit angle 8πGT of the conical spacetime. In contrast,
the supersymmetric string has a ﬁeld stress tensor T
µν
(A, φ) which is not just
concentrated at the string core but instead is smeared out over spacetime.
51
The diﬀerence arises from the absence of a potential for the ﬁelds A
µν
, φ
supporting the solution in the supersymmetric case. Nonetheless, as one can
see from the behavior of the stress tensor T
mn
in Eq. (6.31), the transverse
space Σ
2
is asymptotically locally ﬂat (ALF), with a total energy density given
by the overall deﬁcit angle measured at inﬁnity. For multiplecentered string
solutions, one has
H = 1 −
¸
i
8GT
i
ln y −y
i
 . (6.35)
Consequently, when considered within the original supergravity theory, the in
deﬁnite stacking of supersymmetric strings leads to a destruction of the trans
verse asymptotic space.
A second problem with any attempt to produce (D − 2)branes in ordi
nary supergravity theories is simply stated: starting from the pbrane ansatz
(2.3, 2.6) and searching for (D − 2) branes in ordinary massless supergravity
theories, one simply doesn’t ﬁnd any such solutions.
6.5 Beyond the (D−3)brane barrier: ScherkSchwarz reduction and domain
walls
Faced with the above puzzles about what sort of (D − 2)brane could result
by vertical reduction from a (D−3)brane, one can simply decide to be brave,
and to just proceed anyway with the established mathematical procedure of
vertical dimensional reduction and see what one gets. In the next step of ver
tical dimensional reduction, one again encounters an additive divergence: the
integral
L
−L
dz ln(y
2
+z
2
) needs to be renormalized by subtracting a divergent
term 4L(lnL − 1). Upon subsequently performing the integral, the harmonic
function H(y) becomes linear in the one remaining transverse coordinate.
While the mathematical procedure of vertical dimensional reduction so as
to produce some sort of (D − 2)brane proceeds apparently without serious
complication, an analysis of the physics of the situation needs some care.
39
Consider the reduction from a (D, d = D −2) solution (a p = (D − 3) brane)
to a (D − 1, d = D − 3) solution (a p = (D − 2) brane). Note that both the
(D−3) brane and its descendant (D−2)brane have harmonic functions H(y)
that blow up at inﬁnity. For the (D − 3)brane, however this is not in itself
particularly remarkable, because, as one can see by inspection of (2.24) for this
case, the metric asymptotically tends to a locally ﬂat space as r → ∞, and
also in this limit the antisymmetrictensor oneform ﬁeld strength
F
m
= −ǫ
mn
∂
n
H (6.36)
52
tends asymptotically to zero, while the dilatonic scalar φ tends to its modulus
value φ
∞
(set to zero for simplicity in (2.24)). The expression (6.36) for the ﬁeld
strength, however, shows that the next reduction step down to the (D−1, d =
D−2) solution has a signiﬁcant new feature: upon stacking up (D−3) branes
prior to the vertical reduction, thus producing a linear harmonic function in
the transverse coordinate y,
H(y) = const. +my , (6.37)
the ﬁeld strength (6.36) acquires a constant component along the stacking axis
↔ reduction direction z,
F
z
= −ǫ
zy
∂
y
H = m , (6.38)
which implies an unavoidable dependence
l
of the corresponding zeroform
gauge potential on the reduction coordinate:
A
[0]
(x, y, z) = mz +χ(x, y) . (6.39)
From a KaluzaKlein point of view, the unavoidable linear dependence
of a gauge potential on the reduction coordinate given in (6.39) appears to
be problematic. Throughout this review, we have dealt only with consistent
KaluzaKlein reductions, for which solutions of the reduced theory are also
solutions of the unreduced theory. Generally, retaining any dependence on a
reduction coordinate will lead to an inconsistent truncation of the theory: at
tempting to impose a z dependence of the form given in (6.39) prior to varying
the Lagrangian will give a result diﬀerent from that obtained by imposing this
dependence in the ﬁeld equations after variation.
The resolution of this diﬃculty is that in performing a KaluzaKlein re
duction with an ansatz like (6.39), one ends up outside the standard set of
massless supergravity theories. In order to understand this, let us again focus
on the problem of consistency of the KaluzaKlein reduction. As we have seen,
consistency of any restriction means that the restriction may either be imposed
on the ﬁeld variables in the original action prior to variation so as to derive
the equations of motion, or instead may be imposed on the ﬁeld variables in
the equations of motion after variation, with an equal eﬀect. In this case, so
lutions obeying the restriction will also be solutions of the general unrestricted
equations of motion.
l
Note that this vertical reduction from a (D − 3)brane to a (D −2)brane is the ﬁrst case
in which one is forced to accept a dependence on the reduction coordinate z; in all higher
dimensional vertical reductions, such z dependence can be removed by a gauge transforma
tion. The zeroform gauge potential in (6.39) does not have the needed gauge symmetry,
however.
53
The most usual guarantee of consistency in KaluzaKlein dimensional re
duction is obtained by restricting the ﬁeld variables to carry zero charge with
respect to some conserved current, e.g. momentum in the reduction dimen
sion. But this is not the only way in which consistency may be achieved. In
the present case, retaining a linear dependence on the reduction coordinate as
in (6.39) would clearly produce an inconsistent truncation if the reduction co
ordinate were to appear explicitly in any of the ﬁeld equations. But this does
not imply that a truncation is necessarily inconsistent just because a gauge po
tential contains a term linear in the reduction coordinate. Inconsistency of a
KaluzaKlein truncation occurs when the original, unrestricted, ﬁeld equations
imply a condition that is inconsistent with the reduction ansatz. If a particular
gauge potential appears in the action only through its derivative, i.e. through
its ﬁeld strength, then a consistent truncation may be achieved provided that
the restriction on the gauge potential implies that the ﬁeld strength is inde
pendent of the reduction coordinate. A zeroform gauge potential on which
such a reduction may be carried out, occurring in the action only through its
derivative, will be referred to as an axion.
Requiring axionic ﬁeld strengths to be independent of the reduction coor
dinate amounts to extending the KaluzaKlein reduction framework so as to
allow for linear dependence of an axionic zeroform potential on the reduction
coordinate, precisely of the form occurring in (6.39). So, provided A
[0]
is an ax
ion, the reduction (6.39) turns out to be consistent after all. This extension of
the KaluzaKlein ansatz is in fact an instance of ScherkSchwarz reduction.
40,41
The basic idea of ScherkSchwarz reduction is to use an Abelian rigid symme
try of a system of equations in order to generalize the reduction ansatz by
allowing a linear dependence on the reduction coordinate in the parameter
of this Abelian symmetry. Consistency is guaranteed by cancellations orches
trated by the Abelian symmetry in ﬁeldequation terms where the parameter
does not get diﬀerentiated. When it does get diﬀerentiated, it contributes only
a term that is itself independent of the reduction coordinate. In the present
case, the Abelian symmetry guaranteeing consistency of (6.39) is a simple shift
symmetry A
[0]
→A
[0]
+ const.
Unlike the original implementation of the ScherkSchwarz reduction idea,
40
which used an Abelian U(1) phase symmetry acting on spinors, the Abelian
shift symmetry used here commutes with supersymmetry, and hence the reduc
tion does not spontaneously break supersymmetry. Instead, gauge symmetries
for some of the antisymmetric tensors will be broken, with a corresponding
appearance of mass terms. As with all examples of vertical dimensional re
duction, the ∆ value corresponding to a given ﬁeld strength is also preserved.
Thus, pbrane solutions related by vertical dimensional reduction, even in the
54
enlarged ScherkSchwarz sense, preserve the same amount of unbroken super
symmetry and have the same value of ∆.
It may be necessary to make several redeﬁnitions and integrations by parts
in order to reveal the axionic property of a given zeroform, and thus to pre
pare the theory for a reduction like (6.39). This is most easily explained by an
example, so let us consider the ﬁrst possible ScherkSchwarz reduction
m
in the
sequence of theories descending from (1.1), starting in D = 9 where the ﬁrst
axion ﬁeld appears.
39
The Lagrangian for massless D = 9 maximal supergrav
ity is obtained by specializing the general dimensionallyreduced action (6.13)
given in Section 2 to this case:
L
9
=
√
−g
R−
1
2
(∂φ
1
)
2
−
1
2
(∂φ
2
)
2
−
1
2
e
−
3
2
φ1+
√
7
2
φ2
(∂χ)
2
−
1
48
e
a·
φ
(F
[4]
)
2
−
1
2
e
a1·
φ
(F
(1)
[3]
)
2
−
1
2
e
a2·
φ
(F
(2)
[3]
)
2
−
1
4
e
a12·
φ
(F
(12)
[2]
)
2
−
1
4
e
b1·
φ
(F
(1)
[2]
)
2
−
1
4
e
b2·
φ
(F
(2)
[2]
)
2
−
1
2
˜
F
[4]
∧
˜
F
[4]
∧A
(12)
[1]
−
˜
F
(1)
[3]
∧
˜
F
(2)
[3]
∧A
[3]
, (6.40)
where χ = A
(12)
[0]
and
φ = (φ
1
, φ
2
).
Within the scalar sector (
φ, χ) of (6.40), the dilaton coupling has been
made explicit; in the rest of the Lagrangian, the dilaton vectors have the general
structure given in (6.14, 6.15). The scalar sector of (6.40) forms a nonlinear σ
model for the manifold
GL(2, R)
/SO(2). This already makes it appear that one
may identify χ as an axion available for ScherkSchwarz reduction. However,
account must still be taken of the ChernSimons structure lurking inside the
ﬁeld strengths in (6.11, 6.40). In detail, the ﬁeld strengths are given by
F
[4]
=
˜
F
[4]
−
˜
F
(1)
[3]
∧ A
(1)
[1]
−
˜
F
(2)
[3]
∧ A
(2)
[1]
+χ
˜
F
(1)
[3]
∧ A
(2)
[1]
−
˜
F
(12)
[2]
∧ A
(1)
[1]
∧ A
(2)
[1]
(6.41a)
F
(1)
[3]
=
˜
F
(1)
[3]
−
˜
F
(12)
[2]
∧ A
(2)
[1]
(6.41b)
F
(2)
[3]
=
˜
F
(2)
[3]
+
˜
F
(12)
[2]
∧ A
(1)
[1]
−χ
˜
F
(1)
[3]
(6.41c)
F
(12)
[2]
=
˜
F
(12)
[2]
F
(1)
[2]
=
˜
F
(1)
[2]
−dχ ∧ A
(2)
[1]
(6.41d)
F
(2)
[2]
=
˜
F
(2)
[2]
F
(12)
[1]
= dχ , (6.41e)
where the ﬁeld strengths carrying tildes are the na¨ıve expressions without
ChernSimons corrections, i.e.
˜
F
n]
= dA
[n−1]
. Now the appearance of undiﬀer
m
A higherdimensional ScherkSchwarz reduction is possible
41
starting from type IIB su
pergravity in D = 10, using the axion appearing in the
SL(2, R)
/SO(2) scalar sector of that
theory.
55
entiated χ factors in (6.41a,c) makes it appear that a ScherkSchwarz reduction
would be inconsistent. However, one may eliminate these undiﬀerentiated fac
tors by making the ﬁeld redeﬁnition
A
(2)
[2]
−→A
(2)
[2]
+χA
(1)
[2]
, (6.42)
after which the ﬁeld strengths (6.41a,c) become
F
[4]
=
˜
F
[4]
−
˜
F
(1)
[3]
∧ A
(1)
[1]
−
˜
F
(2)
[3]
∧ A
(2)
[1]
−dχ ∧ A
(1)
[2]
∧ A
(2)
[1]
−
˜
F
(12)
[2]
∧ A
(1)
[1]
∧ A
(2)
[1]
(6.43a)
F
(2)
[3]
=
˜
F
(2)
[3]
+F
(12)
[2]
∧ A
(1)
[1]
+dχ ∧ A
(1)
[2]
, (6.43c)
the rest of (6.41) remaining unchanged.
After the ﬁeld redeﬁnitions (6.42), the axion ﬁeld χ = A
(12)
[0]
is now ready
for application of the ScherkSchwarz reduction ansatz (6.39). The coeﬃcient
of the term linear in the reduction coordinate z has been denoted m because it
carries the dimensions of mass, and correspondingly its eﬀect on the reduced
action is to cause the appearance of mass terms. Applying (6.39) to the D = 9
Lagrangian, one obtains a D = 8 reduced Lagrangian
n
L
8 ss
=
√
−g
R −
1
2
(∂φ
1
)
2
−
1
2
(∂φ
2
)
2
−
1
2
(∂φ
3
)
2
−
1
2
e
b12·
φ
(∂χ −mA
(3)
[1]
)
2
−
1
2
e
b13·
φ
(∂A
(13)
[0]
−∂χA
(23)
[0]
+mA
(2)
[1]
)
2
−
1
2
e
b23·
φ
(∂A
(23)
[0]
)
2
−
1
2
e
a123·
φ
(∂A
(123)
[0]
)
2
−
1
48
e
a·
φ
(F
[4]
−mA
(1)
[2]
∧ A
(2)
[1]
∧ A
(3)
[1]
)
2
−
1
12
e
a1·
φ
(F
(1)
[3]
)
2
−
1
12
e
a2·
φ
(F
(2)
[3]
−mA
(1)
[2]
∧ A
(3)
[1]
)
2
−
1
12
e
a3·
φ
(F
(3)
[3]
+mA
(1)
[2]
∧ A
(2)
[1]
)
2
−
1
4
e
a12·
φ
(F
(12)
[2]
)
2
−
1
4
e
a13·
φ
(F
(13)
[2]
)
2
−
1
4
e
a23·
φ
(F
(23)
[2]
+mA
(1)
[2]
)
2
−
1
4
e
b1·
φ
(F
(1)
[2]
−mA
(2)
[1]
∧ A
(3)
[1]
)
2
−
1
4
e
b2·
φ
(F
(2)
[2]
)
2
−
1
4
e
b3·
φ
(F
(3)
[2]
)
2
−
1
2
m
2
e
b123·
φ
+mF
(1)
[3]
∧ A
(1)
[2]
∧ A
[3]
+L
FFA
, (6.44)
where the dilaton vectors are now those appropriate for D = 8; the term L
FFA
contains only mindependent terms.
n
I am grateful to Marcus Bremer for help in correcting some errors in the original expression
of Eq. (6.44) given in Ref.
39
56
It is apparent from (6.44) that the ﬁelds A
(3)
[1]
, A
(2)
[1]
and A
(1)
[2]
have become
massive. Moreover, there are ﬁeld redeﬁnitions under which the ﬁelds χ, A
(13)
[0]
and A
(23)
[1]
may be absorbed. One way to see how this absorption happens is to
notice that the action obtained from (6.44) has a set of three Stueckelbergtype
gauge transformations under which A
(3)
[1]
, A
(2)
[1]
and A
(1)
[2]
transform according
to their standard gauge transformation laws. These three transformations are
accompanied, however, by various compensating transformations necessitated
by the ChernSimons corrections present in (6.44) as well as by mdependent
shift transformations of χ, A
(13)
[0]
and A
(23)
[1]
, respectively. Owing to the presence
of these local shift terms in the three Stueckelberg symmetries, the ﬁelds χ,
A
(13)
[0]
and A
(23)
[1]
may be gauged to zero. After gauging these three ﬁelds to
zero, one has a clean set of mass terms in (6.44) for the ﬁelds A
(3)
[1]
, A
(2)
[1]
and
A
(1)
[2]
.
As one descends through the available spacetime dimensions for super
gravity theories, the number of axionic scalars available for a ScherkSchwarz
reduction step increases. The numbers of axions are given in the following
Table:
Table 1: Supergravity axions versus spacetime dimension.
D 9 8 7 6 5 4
N
axions
1 4 10 20 36 63
Each of these axions gives rise to a distinct massive supergravity theory
upon ScherkSchwarz reduction,
39
and each of these reduced theories has its
own pattern of mass generation. In addition, once a ScherkSchwarz reduction
step has been performed, the resulting theory can be further reduced using
ordinary KaluzaKlein reduction. Moreover, the ScherkSchwarz and ordinary
KaluzaKlein processes do not commute, so the number of theories obtained by
the various combinations of ScherkSchwarz and ordinary dimensional reduc
tion is cumulative. In addition, there are numerous possibilities of performing
ScherkSchwarz reduction simultaneously on a number of axions. This can
be done either by arranging to cover a number of axions simultaneously with
derivatives, or by further ScherkSchwarz generalizations of the KaluzaKlein
reduction process.
42
For further details on the panoply of ScherkSchwarz re
duction possibilities, we refer the reader to Refs.
39,42
The singlestep procedure of ScherkSchwarz dimensional reduction de
scribed above may be generalized to a procedure exploiting the various coho
57
mology classes of a multidimensional compactiﬁcation manifold.
43
The key to
this link between the ScherkSchwarz generalized dimensional reduction and
the topology of the internal KaluzaKlein manifold K is to recognize that the
singlestep reduction ansatz (6.39) may be generalized to
A
[n−1]
(x, y, z) = ω
[n−1]
+A
[n−1]
(x, y) , (6.45)
where ω
[n−1]
is an (n−1) form deﬁned locally on K, whose exterior derivative
Ω
[n]
= dω
[n−1]
is an element of the cohomology class H
n
(K, R). For example,
in the case of a singlestep generalized reduction on a circle S
1
, one has Ω
[1]
=
mdz ∈ H
1
(S
1
, R), reproducing our earlier singlestep reduction (6.39).
As another example, consider a generalized reduction on a 4torus T
4
starting in D = 11, setting A
[3]
(x, y, z) = ω
[3]
+ A
[3]
(x, y) with Ω
[4]
= dω
[3]
=
mdz
1
∧dz
2
∧dz
3
∧dz
4
∈ H
4
(T
4
, R). In this example, one may choose to write
ω
[3]
locally as ω
[3]
= mz
1
dz
2
∧ dz
3
∧ dz
4
. All of the other ﬁelds are reduced
using the standard KaluzaKlein ansatz, with no dependence on any of the z
i
coordinates. The theory resulting from this T
4
reduction is a D = 7 massive
supergravity with a cosmological potential, analogous to the D = 8 theory
(6.44). The same theory (up to ﬁeld redeﬁnitions) can also be obtained
39
by
ﬁrst making an ordinary KaluzaKlein reduction from D = 11 down to D = 8
on a 3torus T
3
, then making an S
1
singlestep generalized ScherkSchwarz
reduction (6.39) from D = 8 to D = 7. Although the T
4
reduction example
simply reproduces a massive D = 7 theory that can also be obtained via the
singlestep ansatz (6.39), the recognition that one can use any of the H
n
(K, R)
cohomology classes of the compactiﬁcation manifold K signiﬁcantly extends
the scope of the generalized reduction procedure. For example, it allows one
to make generalized reductions on manifolds such as K3 or on CalabiYau
manifolds.
43
For our present purposes, the important feature of theories obtained by
ScherkSchwarz reduction is the appearance of cosmological potential terms
such as the penultimate term in Eq. (6.44). Such terms may be considered
within the context of our simpliﬁed action (2.1) by letting the rank n of the
ﬁeld strength take the value zero. Accordingly, by consistent truncation of
(6.44) or of one of the many theories obtained by ScherkSchwarz reduction in
lower dimensions, one may arrive at the simple Lagrangian
L =
√
−g
R −
1
2
∇
M
φ∇
M
φ −
1
2
m
2
e
aφ
. (6.46)
Since the rank of the form here is n = 0, the elementary/electric type of solution
would have worldvolume dimension d = −1, which is not very sensible, but
58
the solitonic/magnetic solution has
˜
d = D − 1, corresponding to a p = D − 2
brane, or domain wall, as expected. Relating the parameter a in (6.46) to the
reductioninvariant parameter ∆ by the standard formula (2.18) gives ∆ =
a
2
−2(D−1)/(D−2); taking the corresponding p = D−2 brane solution from
(2.24), one ﬁnds
ds
2
= H
4
∆(D−2)
η
µν
dx
µ
dx
ν
+H
4(D−1)
∆(D−2)
dy
2
(6.47a)
e
φ
= H
−2a/∆
, (6.47b)
where the harmonic function H(y) is now a linear function of the single trans
verse coordinate, in accordance with (6.37).
o
The curvature of the metric
(6.47a) tends to zero at large values of y, but it diverges if H tends to zero.
This latter singularity can be avoided by taking H to be
H = const. +My (6.48)
where M =
1
2
m
√
∆. With the choice (6.48), there is just a deltafunction
singularity at the location of the domain wall at y = 0, corresponding to the
discontinuity in the gradient of H.
The domainwall solution (6.47, 6.48) has the peculiarity of tending asymp
totically to ﬂat space as y → ∞, within a theory that does not naturally
admit ﬂat space as a solution (by “naturally,” we are excluding the case
aφ → −∞). Moreover, the theory (6.46) does not even admit a nonﬂat
maximallysymmetric solution, owing to the complication of the cosmological
potential. The domainwall solution (6.47, 6.48), however, manages to “can
cel” this potential at transverse inﬁnity, allowing at least asymptotic ﬂatness
for this solution.
This brings us back to the other facets of the consistency problem for ver
tical dimensional reduction down to (D−2)branes as discussed in subsection
6.4. There is no inconsistency between the existence of domainwall solutions
like (6.47, 6.48) and the inability to ﬁnd such solutions in standard supergrav
ity theories, or with the conicalspacetime character of (D−3)branes, because
these domain walls exist only in massive supergravity theories like (6.44), with
a vacuum structure diﬀerent from that of standard massless supergravities.
Because the ScherkSchwarz generalized dimensional reduction used to obtain
them was a consistent truncation, such domain walls can be oxidized back
to solutions of higherdimensional massless supergravities, but in that case,
they have the form of stacked solutions prepared for vertical reduction, with
nonzero ﬁeld strengths in the reduction directions, as in our example (6.38).
o
Domain walls solutions such as (6.47) in supergravity theories were found for the D = 4
case in Ref.
44
and a review of them has been given in Ref.
45
59
7 Intersecting branes and scattering branes
7.1 Multiple component solutions
Given the existence of solutions (6.24) with several active ﬁeld strengths F
α
[n]
,
but with coincident charge centers, it is natural to try to ﬁnd solutions where
the charge centers for the diﬀerent F
α
[n]
are separated.
46
This will lead us to
a better understanding of the ∆ = 4 solutions shown in Figure 7. Consider
a number of ﬁeld strengths that individually have ∆ = 4 couplings, but now
look for a solution where ℓ of these ﬁeld strengths are active, with centers y
α
,
α = 1, . . . , ℓ. Let the charge parameter for F
α
be λ
α
. Thus, for example, in
the magnetic case, one sets
F
α
m1,...,mn
= λ
α
ǫ
m1,...,mnp
y
p
y −y
α

n+1
. (7.1)
In both the electric and the magnetic cases, the λ
α
are related to the integration
constants k
α
appearing in the metric by k
α
= λ
α
/
˜
d. Letting ς = ±1 in the
electric/magnetic cases as before, the solution for the metric and the active
dilatonic combinations e
ςaα·
φ
is given by
ds
2
=
ℓ
¸
α=1
H
−
˜
d
D−2
α
dx
µ
dx
µ
+
ℓ
¸
α=1
H
d
D−2
α
dy
m
dy
m
e
ςaα·
φ
= H
2
α
ℓ
¸
β1
H
−d
˜
d
D−2
β
(7.2)
H
α
= 1 +
k
α
y −y
α

˜
d
.
The nontrivial step in verifying the validity of this solution is the check that
the nonlinear terms still cancel in the Einstein equations, even with the mul
tiple centers.
46
Now consider a solution with two ﬁeld strengths (F
1
[n]
, F
2
[n]
) in which the
two charge parameters are taken to be the same, λ
α
= λ, while the charge
centers are allowed to coalesce. When the charge centers have coalesced, the
resulting solution may be viewed as a singleﬁeldstrength solution for a ﬁeld
strength rotated by π/4 in the space of ﬁeld strengths (F
1
[n]
, F
2
[n]
). Since the
charges add vectorially, the net charge parameter in this case will be λ =
√
2λ, and the net charge density will be U =
√
2λΩ
D−d−1
/4. On the other
hand, the total mass density will add as a scalar quantity, so E = E
1
+ E
2
=
2λΩ
D−d−1
/4 =
√
2U. Thus, the coalesced solution satisﬁes E = 2U/
√
∆ with
60
∆ = 2. Direct comparison with our general pbrane solution (2.24) shows that
the coalesced solution agrees precisely with the singleﬁeldstrength ∆ = 2
solution. Generalizing this construction to a case with N separate ∆ = 4
components, one ﬁnds in the coincident limit a ∆ = 4/N supersymmetric
solution from the singleﬁeldstrength analysis. In the next subsection, we
shall see that as one adds new components, each one separately charged with
respect to a diﬀerent ∆ = 4 ﬁeld strength, one progressively breaks more and
more supersymmetry. For example, the above solution (7.2) leaves unbroken
1/4 of the original supersymmetry. Since the ∆ = 4/N solutions may in
this way be separated into ∆ = 4 components while still preserving some
degree of unbroken supersymmetry, and without producing any relative forces
to disturb their equilibrium, they may be considered to be “bound states at
threshold.”
46
We shall shortly see that the zeroforce property of such multiple
component solutions is related to their managing still to preserve unbroken a
certain portion of rigid supersymmetry, even though this portion is reduced
with respect to the halfpreservation characterizing singlecomponent ∆ = 4
solutions.
7.2 Intersecting branes and the four elements in D = 11
The multiplechargecenter solutions (7.2) to the dimensionally reduced the
ory (6.13) may automatically be interpreted as solutions of any one of the
higherdimensional theories descending from the D = 11 theory (1.1). This
automatic “oxidation” is possible because we have insisted throughout on con
sidering only consistent truncations. Although all lowerdimensional solutions
may automatically be oxidized in this way into solutions of higherdimensional
supergravity theories, it is not guaranteed that these oxidized branes always
fall into the class of isotropic pbrane solutions that we have mainly been dis
cussing. For example, in D = 9, one has a twoblackhole solution of the form
(7.2), supported by a 1form gauge potential A
12
[1]
descending from the D = 11
gauge potential A
[3]
and also by another 1form gauge potential, e.g. A
2
[1]
,
emerging from the metric as a KaluzaKlein vector ﬁeld. Upon oxidizing the
twoblackhole solution back to D = 11, one ﬁnds the solution
ds
2
11
= H
1
3
1
(y)
H
−1
1
(y){−dt
2
+dρ
2
+dσ
2
+(H
2
(y)−1)(dt+dρ)
2
}+dy
m
dy
m
A
[3]
= H
−1
1
(y)dt∧dρ∧dσ , m = 3, . . . , 10, wave2brane (7.3)
which depends on two independent harmonic functions H
1
(y) and H
2
(y), where
the y
m
are an 8dimensional set of “overall transverse” coordinates.
Although the solution (7.3) clearly falls outside the class of pbrane or
multiple pbrane solutions that we have considered so far, it nonetheless has
61
two clearly recognizable elements, associated to the two harmonic functions
H
1
(y) and H
2
(y). In order to identify these two elements, we may use the
freedom to trivialize one or the other of these harmonic functions by setting it
equal to unity. Thus, setting H
2
= 1, one recovers
ds
2
11
= H
1
3
(y)
H
−1
(y){−dt
2
+dρ
2
+dσ
2
+dy
m
dy
m
A
[3]
= H
−1
(y)dt ∧ dρ ∧ dσ, m = 3, . . . , 10, 2brane (7.4)
which one may recognized as simply a certain style of organizing the harmonic
function factors in the D = 11 membrane solution
18
(3.2), generalized to
an arbitrary harmonic function H(y) ↔ H
1
(y) in the membrane’s transverse
space.
Setting H
1
= 1 in (7.3), on the other hand, produces a solution of D = 11
supergravity that is not a pbrane (i.e. it is not a Poincar´einvariant hyperplane
solution). What one ﬁnds for H
1
= 1 is a classic solution of General Relativity
found originally in 1923 by Brinkmann,
47
the pp wave:
ds
2
11
= {−dt
2
+dρ
2
+ (H(y) −1)(dt +dρ)
2
} +dy
m
dy
m
A
[3]
= 0 , m = 2, . . . , 10, pp wave (7.5)
where for a general wave solution, H(y) could be harmonic in the 9 dimensions
y
m
transverse to the two lightplane dimensions {t, ρ} in which the wave propa
gates; for the speciﬁc case obtained by setting H
1
= 1 in (7.3), H(y) ↔H
2
(y)
is constant in one of these 9 directions, corresponding to the coordinate σ in
(7.3).
The solution (7.3) thus may be viewed as a D = 11 pp wave superposed on
a membrane. Owing to the fact that the harmonic function H
2
(y) depends only
on the overall transverse coordinates y
m
, m = 3, . . . , 10, the wave is actually
“delocalized” in the third membrane worldvolume direction, i.e. the solution
(7.3) is independent of σ as well as of its own lightplane coordinates. Of course,
this delocalization of the wave in the σ direction is just what makes it possible
to perform a dimensional reduction of (7.3) on the {ρ, σ} coordinates down to
a D = 9 conﬁguration of two particles of the sort considered in (7.2), i.e. the
wave in (7.3) has already been stacked up in the σ direction as is necessary in
preparation for a vertical dimensional reduction. Another point to note about
(7.3) is that the charge centers of the two harmonic functions H
1
and H
2
may
be chosen completely independently in the overall transverse space. Thus,
although this is an example of an “intersecting” brane conﬁguration, it should
be understood that the two components of (7.3) need not actually overlap on
any speciﬁc subspace of spacetime. The term “intersecting” is generally taken
62
to mean that there are shared worldvolume coordinates, in this case the {t, ρ}
overlap between the membrane worldvolume and the lightplane coordinates.
48
A very striking feature of the family of multiplecomponent pbrane solu
tions is that their oxidations up to D = 11 involve combinations of only 4 basic
“elemental” D = 11 solutions. Two of these we have just met in the oxidized
solution (7.3): the membrane and the pp wave. The two others are the “duals”
of these: the 5brane
25
and a solution describing the oxidation to D = 11 of
the “KaluzaKlein monopole.”
49
The 5brane may be written in a style similar
to that of the membrane (7.4):
ds
2
11
= H
2
3
(y)
H
−1
(y){−dt
2
+dx
2
1
+. . . +dx
2
5
} +dy
m
dy
m
F
[4]
=
∗
dH(y), m = 6, . . . , 10, 5brane (7.6)
where the H(y) is a general harmonic function in the 5dimensional transverse
space.
The KaluzaKlein monopole oxidized up to D = 11 is the solution
ds
2
11
= −dt
2
+dx
2
1
+. . . +dx
2
6
+ds
2
TN
(y)
A
[3]
= 0 (7.7a)
ds
2
TN
= H(y)dy
i
dy
i
+H
−1
(y)(dψ +V
i
(y)dy
i
)
2
, i = 1, 2, 3,
∇×
V =
∇H , TaubNUT (7.7b)
where ds
2
TN
is the TaubNUT metric, a familiar fourdimensional Euclidean
gravitational instanton. The harmonic function H in (7.7) is a function only
of the 3 coordinates y
i
, and not of the coordinate ψ, which plays a special
rˆole. Generally, the solution (7.7) has a conical singularity on the hyperplane
y
i
= 0, but this becomes a mere coordinate singularity, similar to that for ﬂat
space in polar coordinates, providing the coordinate ψ is periodically identiﬁed.
For a singlecenter harmonic function H(y) = 1 + k/(y), the appropriate
identiﬁcation period for ψ is 4πk.
Thus, the TaubNUT solution naturally invites interpretation as a com
pactiﬁed solution in one less dimension, after reduction on ψ. In the case of
the original KaluzaKlein monopole,
49
the starting solution had 4+1 dimen
sions, giving rise after compactiﬁcation to a magneticallycharged particle in
D = 4 dimensions. The solution (7.7) has an additional 6 spacelike worldvol
ume dimensions x
1
, . . . , x
6
, so after reduction on the ψ coordinate one has a
magneticallycharged 6brane solution in D = 10.
The relation ∆ψ = 4πk between the compactiﬁcation period of ψ and
the chargedetermining integration constant k in the harmonic function H
of the solution (7.7) gives rise to a quantisation condition at the quantum
63
level involving the magnetic charge of the dimensionallyreduced D = 10 6
brane descending from (7.7) and the electric charge of the extreme black hole
particle obtained by reducing the pp wave (7.5). This quantisation condition
is nothing other than an ordinary quantisation of momentum for Fourier wave
components on a compact space, in this case the compact ψ direction. In terms
of the electric and magnetic charges U and V of the dimensionally reduced
particle and 6brane, one ﬁnds UV = 2πκ
2
10
n, with n ∈ Z (where κ
2
10
occurs
because the charges U and V as deﬁned in (1.3, 1.4) are not dimensionless).
This is precisely of the form expected for a Dirac charge quantisation condition.
In Section 8 we shall return to the subject of charge quantisation conditions
more generally for the charges carried by pbranes.
Let us now return to the question of supersymmetry preservation and
enquire whether intersecting branes like (7.3) can also preserve some portion
of unbroken rigid supersymmetry. All four of the elemental D = 11 solutions
(7.4 – 7.7) preserve half the D = 11 rigid supersymmetry. We have already seen
this for the membrane solution in subsection 4.4. As another example, one may
consider the supersymmetry preservation conditions for the pp wave solution
(7.5). We shall skip over points 1) and 2) of the discussion analogous to that of
subsection 4.4 and shall instead concentrate just on the projection conditions
that must be satisﬁed by the surviving rigid supersymmetry parameter ǫ
∞
.
Analogously to our earlier abbreviated discussion using just the supersymmetry
algebra, consider this algebra in the background of a pp wave solution (7.5)
propagating in the {01} directions of spacetime, with normalization to unit
length along the wave’s propagation direction:
1
length
{Q
α
, Q
β
} = 2EP
01
P
01
=
1
2
(1l + Γ
01
) , (7.8)
where P
01
is again a projection operator with half of its eigenvalues zero, half
unity. Consequently, the pp wave solution (7.5) preserves half of the D = 11
rigid supersymmetry.
Now let us apply the projectionoperator analysis to the wave2brane so
lution (7.3). Supersymmetry preservation in a membrane background oriented
parallel to the {012} hyperplane requires the projection condition P
012
ǫ
∞
= 0
(4.23), while supersymmetry preservation in a pp wave background with a {01}
lightplane requires P
01
ǫ
∞
= 0. Imposing these two conditions simultaneously
is consistent because these projectors commute,
[P
012
, P
01
] = 0 . (7.9)
Since tr(P
012
P
01
) =
1
4
· 32, the imposition of both projection conditions on ǫ
∞
cuts the preserved portion of rigid D = 11 supersymmetry down to
1
4
.
64
Now, let’s consider another example of an intersectingbrane solution, con
taining as elements a D = 11 membrane, 5brane pair. The solution is
ds
2
= H
1
3
1
(y)H
2
3
2
(y)
H
−1
1
(y)H
−1
2
(y)(−dt
2
+dx
2
1
) 2 ⊥ 5(1)
+H
−1
1
(y)(dx
2
2
) +H
−1
2
(y)(dx
2
3
+. . . +dx
2
6
)
+dy
m
dy
m
m = 7, . . . , 10 (7.10a)
F
m012
= ∂
m
(H
−1
1
) F
2mnp
= −ǫ
mnpq
∂
q
H
2
, (7.10b)
where as in the waveonamembrane solution (7.3), the harmonic functions
H
1
(y) and H
2
(y) depend only on the overall transverse coordinates. By con
sidering special cases where H
2
= 1 or H
1
= 1, one identiﬁes the membrane
and 5brane elements of the solution (7.10); as before, these elements are delo
calized in (i.e., independent of) the “relative transverse” directions, by which
one means the directions transverse to one element’s worldvolume but belong
ing to the worldvolume of the other element, i.e. the directions {2; 3, . . . , 6}
for the solution (7.10). Note that both the membrane and 5brane elements
share the worldvolume directions {01}; these are accordingly called “overall
worldvolume” directions. Considering this “intersection” to be a string (but
recall, however, that the overalltransverse charge centers of H
1
and H
2
need
not coincide, so there is not necessarily a true string overlap), the solution
(7.10) is denoted 2 ⊥ 5(1).
The forms of the wave2brane solution (7.3) and the 2 ⊥ 5(1) solution
(7.10) illustrate the general structure of intersectingbrane solutions. For a
twoelement solution, there are four sectors among the coordinates: overall
worldvolume, two relative transverse sectors and the overall transverse sector.
One may make a sketch of these relations for the 2 ⊥ 5(1) solution (7.10):
0 1 2 3 4 5 6 7 8 9 10
x x x
x x x x x x
W25 W2T5 W5T2 T25
The character of each coordinate is indicated in this sketch: W2 and W5 indi
cating worldvolume coordinates with respect to each of the two elements and
T2 and T5 indicating transverse coordinates with respect to each of the two
elements. Thus, the overall worldvolume coordinates are the W25 coordinates
and the overall transverse coordinates are the T25 coordinates. Having estab
lished this coordinate classiﬁcation, the general structure of the intersecting
brane metric is as follows. For each element, one puts an overall conformal
65
factor H
d/(D−2)
i
(y) for the whole metric, and then in addition one puts a fac
tor H
−1
i
(y) in front of each dx
2
term belonging to the worldvolume of the i
th
element. One may verify this pattern in the structure of (7.10). This pattern
has been termed the harmonic function rule.
48
This summary of the structure of intersecting brane solutions does not
replace a full check that the supergravity equations of motion are solved, and
in addition one needs to establish which combinations of the D = 11 elements
may be present in a given solution. For a fuller review on this subject, we refer
the reader to Ref.
50
For now, let us just check point 3) in the supersymmetry
preservation analysis for the 2 ⊥ 5(1) solution (7.10). For each of the two
elements, one has a projection condition on the surviving rigid supersymme
try parameter ǫ
∞
: P
012
ǫ
∞
= 0 for the membrane and P
013456
ǫ
∞
= 0 for
the 5brane. These may be consistently imposed at the same time, because
[P
012
, P
013456
] = 0, similarly to our discussion of the wave2brane solution.
The amount of surviving supersymmetry in the 2 ⊥ 5(1) solution is
1
4
, because
tr(P
012
P
013456
) =
1
4
· 32.
7.3 Brane probes, scattering branes and modulus σmodel geometry
The existence of static conﬁgurations such as the wave2brane solution (7.3)
or the 2 ⊥ 5(1) intersectingbrane solution (7.10) derives from the proper
ties of the transversespace Laplace equation (2.20) arising in the process of
solving the supergravity equations subject to the pbrane ans¨ atze (2.3–2.6).
The Laplace equation has the wellknown property of admitting multicenter
solutions, which we have already encountered in Eq. (6.27). Physically, the
existence of such multicenter solutions corresponds either to a cancellation of
attractive gravitational and dilatonic forces against repulsive antisymmetric
tensor forces, or to the fact that one brane couples to the background super
gravity ﬁelds with a conformal factor that wipes out the eﬀects of the other
brane. In order to see such cancellations more explicitly, one may use a source
coupling analogous to the D = 11 bosonic supermembrane action (4.1) in order
to treat the limiting problem of a light brane probe moving in the background
of a heavy brane.
18,51
In this limit, one may ignore the deformation of the
heavybrane background caused by the light brane. The use of the brane
probe coupling is a simple way to approximately treat timedependent brane
conﬁgurations. For a pbrane probe of this sort coupled to a Ddimensional
supergravity background, the probe action is
I
probe
= −T
α
d
p+1
ξ (−det(∂
µ
x
m
∂
ν
x
n
g
mn
(x))
1
2
e
1
2
ς
pr
aα·
φ
+Q
α
˜
A
α
[p+1]
(7.11a)
66
˜
A
α
[p+1]
= [(p + 1)!]
−1
∂
µ1
x
m1
· · · ∂
µp+1
x
mp+1
A
α
m1···mp+1
dξ
µ1
∧ · · · ∧ dξ
µp+1
.
(7.11b)
The dilaton coupling in (7.11a) occurs because one needs to have the correct
source for the Ddimensional Einstein frame, i.e. the conformal frame in which
the Ddimensional EinsteinHilbert action is free from dilatonic scalar factors.
Requiring that the source match correctly to a pbrane (probe) solution de
mands the presence of the dilaton coupling e
1
2
ς
pr
aα·
φ
, where ς
pr
= ±1 according
to whether the pbrane probe is of electric or magnetic type and where a
α
is
the dilaton vector appearing in the kineticterm dilaton coupling in (6.13) for
the gauge potential A
α
[p+1]
, to which the p brane probe couples.
As a simple initial example of such a brane probe, one may take a light
D = 11 membrane probe in the background of a parallel and similarlyoriented
heavy membrane.
18
In this case, the braneprobe action (7.11) becomes just
the D = 11 supermembrane action (4.1). If one takes the form of the heavy
membrane background from the electric ansatz (2.3, 2.4), and if one chooses
the “static” worldvolume gauge ξ
µ
= x
µ
, µ = 0, 1, 2, then the bosonic probe
action becomes
I
probe
= −T
d
3
ξ
−det(e
2A(y)
η
µν
+e
2B(y)
∂
µ
y
m
∂
ν
y
m
) −e
C(y)
.
(7.12)
Expanding the square root in (7.12), one ﬁnds at order (∂y)
0
the eﬀective
potential
V
probe
= T(e
3A(y)
−e
C
) . (7.13)
Recalling the condition (2.22), which becomes just e
3A(y)
= e
C(y)
for the mem
brane background, one has directly V
probe
= 0, conﬁrming the absence of static
forces between the two membrane components.
Continuing on in the expansion of (7.12) to order (∂y)
2
in the probe ve
locity, one has the eﬀective probe σmodel
I
(2)
probe
= −
T
2
d
3
ξe
3A(y)
e
2(B(y)−A(y)
∂
µ
y
m
∂
ν
y
m
η
µν
, (7.14)
but, recalling the supersymmetrypreservation condition (2.17) characterizing
the heavymembrane background, the probe σmodel metric in (7.14) reduces
simply to
γ
mn
= e
A(y)+2B(y)
δ
mn
= δ
mn
, (7.15)
i.e., the membraneprobe σmodel metric is ﬂat.
67
The ﬂatness of the membraneprobe σmodel metric (7.15) accords pre
cisely with the degree of rigid supersymmetry that survives in the underlying
supergravity solution with two parallel, similarly oriented D = 11 membranes,
which we found in subsection 4.4 to have
1
2
· 32 components, i.e. d = 2 + 1,
N = 8 probeworldvolume supersymmetry. This high degree of surviving su
persymmetry is too restrictive in its constraints on the form of the σmodel
to allow for anything other than a ﬂat metric, precisely as one ﬁnds in (7.15).
Continuing on with the expansion of (7.12), one ﬁrst ﬁnds a nontrivial in
teraction between the probe and the heavy membrane background at order
(∂y)
4
(odd powers being ruled out by timereversal invariance of the D = 11
supergravity equations).
Now consider a braneprobe conﬁguration with less surviving supersym
metry, and with correspondingly weaker constraints on the probe worldvolume
σmodel. Corresponding to the wave2brane solution (7.3), one has, after
dimensional reduction down to D = 9 dimensions, a system of two black holes
supported by diﬀerent ∆ = 4, D = 9 vector ﬁelds: one descending from the
D = 11 3form gauge potential and one descending from the metric.
Now repeat the braneprobe analysis for the twoblackhole conﬁguration,
again choosing a static gauge on the probe worldvolume, which in the present
case just becomes ξ
0
= t. Again expand the determinant of the induced metric
in (7.11). At order (∂y)
0
, this now gives V
probe
= e
A
e
−
3
2
√
7
φ, but this potential
turns out to be just a constant because the heavybrane background satisﬁes
A =
3
2
√
7
φ. Thus, we conﬁrm the expected static zeroforce condition for the
1
4
supersymmetric twoblackhole conﬁguration descending from the wave2
brane solution (7.3). This zeroforce condition arises not so much as a result of
a cancellation between diﬀerent forces but as a result of the probe’s coupling to
the background with a dilatonic factor in (7.11) that wipes out the conformal
factor occurring in the heavy brane background metric.
Proceeding on to (velocity)
2
order, one now obtains a nontrivial probe
σmodel, with metric
γ
mn
= H
back
(y)δ
mn
, (7.16)
where H
back
is the harmonic function controlling the heavy brane’s background
ﬁelds; for the case of two black holes in D = 9, the harmonic function H
back
has the structure (1 +k/r
6
).
The above testbrane analysis for two D = 9 black holes is conﬁrmed by
a more detailed study of the lowvelocity scattering of supersymmetric black
holes performed by Shiraishi.
52
The procedure is a standard one in soliton
physics: one promotes the moduli of a static solution to timedependent func
tions and then substitutes the resulting generalized ﬁeld conﬁguration back
68
into the original ﬁeld equations. This leads to a set of diﬀerential equations on
the modulus variables which may be viewed as eﬀective equations for the mod
uli. In the general case of multiple black hole scattering, the resulting system
of diﬀerential equations may be quite complicated. The system of equations,
however, simpliﬁes dramatically in cases corresponding to the scattering of
supersymmetric black holes, e.g. the above pair of D = 9 black holes, where
the result turns out to involve only 2body forces. These twobody forces may
be derived from an eﬀective action involving the position vectors of the two
black holes. Separating the centerofmass motion from the relative motion,
one obtains the same modulus metric (7.16) as that found in the braneprobe
analysis above, except for a rescaling which replaces the braneprobe mass by
the reduced mass of the twoblackhole system.
Now we should resolve a puzzle of how this nontrivial d = 1 scattering
modulus σmodel turns out to be consistent with the surviving supersymme
try.
53
The modulus variables of the twoblackhole system are ﬁelds in one
dimension, i.e. time. The Nextended supersymmetry algebra in d = 1 is
{Q
I
, Q
J
} = 2δ
IJ
ˆ
H I = 1, . . . , N , (7.17)
where
ˆ
H is the Hamiltonian. A d = 1, N = 1 σmodel is speciﬁed by a
triple (M, γ, A
[3]
), where M is the Riemannian σmodel manifold, γ is the
metric on M and A
[3]
is a 3form on M which plays the rˆole of torsion in
the derivative operator acting on fermions, ∇
(+)
t
= ∂
t
x
i
∇
(+)
i
, where ∇
(+)
i
λ
j
=
∇
i
λ
j
+
1
2
A
j
ik
λ
k
. The σmodel action may be written using N = 1 superﬁelds
x
i
(t, θ) (where x
i
(t) = x
i

θ=0
, λ
i
(t) = Dx
i

θ=0
) as
I = −
1
2
dtdθ(iγ
ij
Dx
i
d
dt
x
j
+
1
3!
A
ijk
Dx
i
Dx
j
Dx
k
) . (7.18)
One may additionally
54
have a set of spinorial N = 1 superﬁelds ψ
a
, with
Lagrangian −
1
2
h
ab
ψ
a
∇
t
ψ
b
, where h
ab
is a ﬁbre metric and ∇
t
is constructed
using an appropriate connection for the ﬁbre corresponding to the ψ
a
. How
ever, in the present case we shall not include this extra superﬁeld. In order
to have extended supersymmetry in (7.18), one starts by positing a second
set of supersymmetry transformations of the form δx
i
= ηI
i
j
Dx
j
, and then
requires these transformations to close to form the N = 2 algebra (7.17); then
one also requires that the action (7.18) be invariant. In this way, one obtains
the equations
I
2
= −1l (7.19a)
N
i
jk
≡ I
i
[j,k]
= 0 (7.19b)
69
γ
kl
I
k
i
I
l
j
= γ
ij
(7.19c)
∇
(+)
(i
I
k
j)
= 0 (7.19d)
∂
[i
(I
m
j
A
mkl]
) −2I
m
[i
∂
[m
A
jkl]]
) = 0 , (7.19e)
where (7.19a,b) follow from requiring the closure of the algebra (7.17) and
(7.19ce) follow from requiring invariance of the action (7.18). Conditions
(7.19a,b) imply that M is a complex manifold, with I
i
j
as its complex struc
ture.
The structure of the conditions (7.19) is more complicated than might have
been expected. Experience with d = 1 + 1 extended supersymmetry
54
might
have lead one to expect, by simple dimensional reduction, just the condition
∇
(+)
i
I
j
k
= 0. Certainly, solutions of this condition also satisfy (7.19c–e), but
the converse is not true, i.e. the d = 1 extended supersymmetry conditions
are “weaker” than those obtained by dimensional reduction from d = 1 + 1,
even though the d = 1 + 1 minimal spinors are, as in d = 1, just real single
component objects. Conversely, the d = 1 + 1 theory implies a “stronger”
condition; the diﬀerence is explained by d = 1 + 1 Lorentz invariance: not
all d = 1 theories can be “oxidized” up to Lorentzinvariant d = 1 + 1 the
ories. In the present case with two D = 9 black holes, this is reﬂected in
the circumstance that after even one dimensional oxidation from D = 9 up
to D = 10, the solution already contains a pp wave element (so that we have
a D = 10 “waveonastring” solution), with a lightplane metric that is not
Poincar´e invariant.
Note also that the d = 1 “torsion” A
[3]
is not required to be closed in
(7.19). d = 1 supersymmetric theories satisfying (7.19) are analogous to (2,0)
chiral supersymmetric theories in d = 1 + 1, but the weaker conditions (7.19)
warrant a diﬀerent notation for this wider class of models; one may call them
2b supersymmetric σmodels.
53
Such models are characterized by a K¨ahler
geometry with torsion.
Continuing on to N = 8, d = 1 supersymmetry, one ﬁnds an 8b general
ization
53
of the conditions (7.19), with 7 independent complex structures built
using the octonionic structure constants
p
ϕ
ab
c
: δx
i
= η
a
I
a
i
j
Dx
j
, a = 1, . . . 7,
with (I
a
)
8
b
= δ
ab
, (I
a
)
b
8
= −δ
b
a
, (I
a
)
b
c
= ϕ
a
b
c
, where the octonion multipli
cation rule is e
a
e
b
= −δ
ab
+ϕ
ab
c
e
c
. Models satisfying such conditions have an
“octonionic K¨ahler geometry with torsion,” and are called OKT models.
53
Now, are there any nontrivial solutions to these conditions? Evidently,
from the braneprobe and Shiraishi analyses, there must be. For our two D = 9
p
We let the conventional octonionic “0” index be replaced by “8” here in order to avoid
confusion with a timelike index; the E
8
transverse space is of Euclidean signature.
70
black holes with a D = 8 transverse space, one may start from the ansatz
ds
2
= H(y)ds
2
(E
8
), A
ijk
= Ω
ijk
ℓ
∂
ℓ
H, where Ω is a 4form on E
8
. Then,
from the 8b generalization of condition (7.19d) one learns Ω
8abc
= ϕ
abc
and
Ω
abcd
= −
∗
ϕ
abcd
; from the 8b generalization of condition (7.19e) one learns
δ
ij
∂
i
∂
j
H = 0. Thus we recover the familiar dependence of pbrane solutions
on transversespace harmonic functions, and we reobtain the braneprobe or
Shiraishi structure of the blackhole modulus scattering metric with
H
relative
= 1 +
k
red
y
1
− y
2

6
, (7.20)
where k
red
determines the reduced mass of the two black holes.
8 Duality symmetries and charge quantisation
As one can see from our discussion of KaluzaKlein dimensional reduction in
Section 6, progression down to lower dimensions D causes the number of dila
tonic scalars
φ and also the number of zeroform potentials of 1form ﬁeld
strengths to proliferate. When one reaches D = 4, for example, a total of 70
such spinzero ﬁelds has accumulated. In D = 4, the maximal (N = 8) super
gravity equations of motion have a linearlyrealized H = SU(8) symmetry; this
is also the automorphism symmetry of the D = 4, N = 8 supersymmetry alge
bra relevant to the (selfconjugate) supergravity multiplet. In formulating this
symmetry, it is necessary to consider complex selfdual and antiselfdual combi
nations of the 2form ﬁeld strengths, which are the highestrank ﬁeld strengths
occurring in D = 4, higher ranks having been eliminated in the reduction or
by dualization. Using twocomponent notation for the D = 4 spinors, these
combinations transform as F
[ij]
αβ
and
¯
F
˙ α
˙
β [ij]
, i, j = 1, . . . , 8, i.e. as a complex
28dimensional dimensional representation of SU(8). Since this complex repre
sentation can be carried only by the complex ﬁeldstrength combinations and
not by the 1form gauge potentials, it cannot be locally formulated at the level
of the gauge potentials or of the action, where only an SO(8) symmetry is
apparent.
Taking all the spinzero ﬁelds together, one ﬁnds that they form a rather
impressive nonlinear σmodel with a 70dimensional manifold. Anticipating
that this manifold must be a coset space with H = SU(8) as the linearly
realized denominator group, Cremmer and Julia
55
deduced that it had to be
the manifold
E
7(+7)/SU(8); since the dimension of E
7
is 133 and that of SU(8)
is 63, this gives a 70dimensional manifold. Correspondingly, a nonlinearly
realized E
7(+7)
symmetry also appears as an invariance of the D = 4, N = 8
maximal supergravity equations of motion. Such nonlinearlyrealized symme
71
tries of supergravity theories have always had a somewhat mysterious charac
ter. They arise in part out of general covariance in the higher dimensions, from
which supergravities arise by dimensional reduction, but this is not enough:
such symmetries act transitively on the σmodel manifolds, mixing both ﬁelds
arising from the metric and also from the reduction of the D = 11 3form
potential A
[3]
in (1.1).
In dimensions 4 ≤ D ≤ 9, maximal supergravity has the sets of σmodel
nonlinear G and linear H symmetries shown in Table 2. In all cases, the spin
zero ﬁelds take their values in “target” manifolds
G
/H. Just as the asymptotic
value at inﬁnity of the metric deﬁnes the reference, or “vacuum” spacetime
with respect to which integrated charges and energy/momentum are deﬁned,
so do the asymptotic values of the spinzero ﬁelds deﬁne the “scalar vacuum.”
These asymptotic values are referred to as the moduli of the solution. In
string theory, these moduli acquire interpretations as the coupling constants
and vacuum θangles of the theory. Once these are determined for a given
“vacuum,” the classiﬁcation symmetry that organizes the distinct solutions of
the theory into multiplets with the same energy must be a subgroup of the little
group, or isotropy group, of the vacuum. In ordinary General Relativity with
asymptotically ﬂat spacetimes, the analogous group is the spacetime Poincar´e
group times the appropriate “internal” classifying symmetry, e.g. the group of
rigid (i.e. constantparameter) YangMills gauge transformations.
Table 2: Supergravity σmodel symmetries.
D G H
9 GL(2, R) SO(2)
8 SL(3, R) ×SL(2, R) SO(3) ×SO(2)
7 SL(5, R) SO(5)
6 SO(5, 5) SO(5) ×SO(5)
5 E
6(+6)
USP(8)
4 E
7(+7)
SU(8)
3 E
8(+8)
SO(16)
The isotropy group of any point on a coset manifold
G
/H is just H, so this
is the classical “internal” classifying symmetry for multiplets of supergravity
solutions.
8.1 An example of duality symmetry: D = 8 supergravity
In maximal D = 8 supergravity, one sees from Table 2 that G = SL(3, R) ×
SL(2, R) and the isotropy group is H = SO(3)×SO(2). We have an (11−3 = 8)
72
vector of dilatonic scalars as well as a singlet F
ijk
[1]
and a triplet F
ij
[1]
(i, j, k =
1, 2, 3) of 1form ﬁeld strengths for zeroform potentials. Taken all together, we
have a manifold of dimension 7, which ﬁts in precisely with the dimension of the
(SL(3, R) ×SL(2, R))
/(SO(3) ×SO(2)) cosetspace manifold: 8+3−(3+1) = 7.
Owing to the directproduct structure, we may for the time being drop
the 5dimensional
SL(3, R)
/SO(3) sector and consider for simplicity just the
2dimensional
SL(2, R)
/SO(2) sector. Here is the relevant part of the action:
56
I
SL(2)
8
=
d
8
x
√
−g
R −
1
2
∇
M
σ∇
M
σ −
1
2
e
−2σ
∇
M
χ∇
M
χ
−
1
2 · 4!
e
σ
(F
[4]
)
2
−
1
2 · 4!
χF
[4]
∗
F
[4]
(8.1)
where
∗
F
MNPQ
= 1/(4!
√
−g)ǫ
MNPQx1x2x3x4
F
x1x2x3x4
(the ǫ
[8]
is a density, hence
purely numerical).
On the scalar ﬁelds (σ, χ), the SL(2, R) symmetry acts as follows: let
λ = χ + ie
σ
; then
Λ =
a b
c d
(8.2)
with ab−cd = 1 is an element of SL(2, R) and acts on λ by the fractionallinear
transformation
λ −→
aλ +b
cλ +d
. (8.3)
The action of the SL(2, R) symmetry on the 4form ﬁeld strength gives us
an example of a symmetry of the equations of motion that is not a symmetry
of the action. The ﬁeld strength F
[4]
forms an SL(2, R) doublet together with
G
[4]
= e
σ∗
F
[4]
−χF
[4]
, (8.4)
i.e.,
F
[4]
G
[4]
−→(Λ
T
)
−1
F
[4]
G
[4]
. (8.5)
One may check that these transform the F
[4]
ﬁeld equation
∇
M
(e
σ
F
MNPQ
+χ
∗
F
MNPQ
) = 0 (8.6)
into the corresponding Bianchi identity,
∇
M
∗
F
MNPQ
= 0 . (8.7)
Since the ﬁeld equations may be expressed purely in terms of F
[4]
, we have a
genuine symmetry of the ﬁeld equations in the transformation (8.5), but since
73
this transformation cannot be expressed locally in terms of the gauge potential
A
[3]
, this is not a local symmetry of the action. The transformation (8.3, 8.5) is
a D = 8 analogue of ordinary Maxwell duality transformation in the presence of
scalar ﬁelds. Accordingly, we shall refer generally to the supergravity σmodel
symmetries as duality symmetries.
The F
[4]
ﬁeld strength of the D = 8 theory supports elementary/electric
pbrane solutions with p = 4 − 2 = 2, i.e. membranes, which have a d = 3
dimensional worldvolume. The corresponding solitonic/magnetic solutions in
D = 8 have worldvolume dimension
˜
d = 8 − 3 − 2 = 3 also. So in this case,
F
[4]
supports both electric and magnetic membranes. It is also possible in this
case to have solutions generalizing the purely electric or magnetic solutions
considered so far to solutions that carry both types of charge, i.e. dyons.
56
This possibility is also reﬂected in the combined Bogomol’ny bound
q
for this
situation, which generalizes the singlecharge bounds (4.16):
E
2
≥ e
−σ∞
(U +χ
∞
V )
2
+e
σ∞
V
2
, (8.8)
where U and V are the electric and magnetic charges and σ
∞
and χ
∞
are the
moduli, i.e. the constant asymptotic values of the scalar ﬁelds σ(x) and χ(x).
The bound (8.8) is itself SL(2, R) invariant, provided that one transforms both
the moduli (σ
∞
, χ
∞
) (according to (8.3)) and also the charges (U, V ). For the
simple case with σ
∞
= χ
∞
= 0 that we have mainly chosen in order to simplify
the writing of explicit solutions, the bound (8.8) reduces to E
2
≥ U
2
+V
2
, which
is invariant under an obvious isotropy group H = SO(2).
8.2 pform charge quantisation conditions
So far, we have discussed the structure of pbrane solutions at a purely classical
level. At the classical level, a given supergravity theory can have a continuous
spectrum of electrically and magnetically charged solutions with respect to any
one of the nform ﬁeld strengths that can support the solution. At the quantum
level, however, an important restriction on this spectrum of solutions enters
into force: the DiracSchwingerZwanziger (DSZ) quantisation conditions for
particles with electric or magnetic or dyonic charges.
57,58
As we have seen,
however, the electric and magnetic charges carried by branes and appearing
in the supersymmetry algebra (1.5) are forms, and the study of their charge
quantisation properties involves some special features not seen in the D = 4
Maxwell case.
59
q
In comparing (8.8) to the singlecharge bounds (4.16), one should take note that for F
[4]
in
(8.1) we have ∆ = 4, so 2/
√
∆ = 1.
74
We shall ﬁrst review a WuYang style of argument,
57
(for a Diracstring
argument, see Ref.
58,60
) considering a closed sequence W of deformations of
one pbrane, say an electric one, in the background ﬁelds set up by a dual,
magnetic, ˆ p = D−p−4 brane. After such a sequence of deformations, one sees
from the supermembrane action (4.1) that the electric pbrane wavefunction
picks up a phase factor
exp
iQ
e
(p + 1)!
W
A
M1...Mp+1
dx
M
1
∧ . . . ∧ dx
Mp+1
, (8.9)
where A
[p+1]
is the gauge potential set up (locally) by the magnetic ˆ pbrane
background.
A number of diﬀerences arise in this problem with respect to the ordinary
Dirac quantisation condition for D = 4 particles. One of these is that, as
we have seen in subsection 4.1, objects carrying pform charges appearing in
the supersymmetry algebra (1.5) are necessarily either inﬁnite or are wrapped
around compact spacetime dimensions. For inﬁnite pbranes, some deformation
sequences W will lead to a divergent integral in the exponent in (8.9); such
deformations would also require an inﬁnite amount of energy, and so should
be excluded from consideration. In particular, this excludes deformations that
involve rigid rotations of an entire inﬁnite brane. Thus, at least the asymptotic
orientation of the electric brane must be preserved throughout the sequence of
deformations. Another way of viewing this restriction on the deformations is
to note that the asymptotic orientation of a brane is encoded into the electric
pform charge, and so one should not consider changing this pform in the
course of the deformation any more than one should consider changing the
magnitude of the electric charge in the ordinary D = 4 Maxwell case.
We shall see shortly that another diﬀerence with respect to the ordinary
D = 4 Dirac quantisation of particles in Maxwell theory will be the existence
of “Diracinsensitive” conﬁgurations, for which the phase in (8.9) vanishes.
Restricting attention to deformations that give nondivergent phases, one
may use Stoke’s theorem to rewrite the integral in (8.9):
Q
e
(p + 1)!
W
A
M1...Mp+1
dx
M
1
∧ . . . ∧ dx
Mp+1
=
Q
e
(p + 2)!
MW
F
M1...Mp+2
dx
M1
∧ . . . ∧ dx
Mp+2
= Q
e
Φ
MW
, (8.10)
where M
W
is any surface “capping” the closed surface W, i.e. a surface such
that ∂M
W
= W; Φ
MW
is then the ﬂux through the cap M
W
. Choosing the
capping surface in two diﬀerent ways, one can ﬁnd a ﬂux discrepancy Φ
M1
−
75
Φ
M2
= Φ
M1∩M2
= Φ
M
total
(taking into account the orientation sensitivity of
the ﬂux integral). Then if M
total
= M
1
∩ M
2
“captures” the magnetic ˆ p
brane, the ﬂux Φ
M
total
will equal the magnetic charge Q
m
of the ˆ pbrane; thus
the discrepancy in the phase factor (8.9) will be simply exp(iQ
e
Q
m
). Requiring
this to equal unity gives,
57
in strict analogy to the ordinary case of electric and
magnetic particles in D = 4, the Dirac quantisation condition
Q
e
Q
m
= 2πn , n ∈ Z . (8.11)
The charge quantisation condition (8.11) is almost, but not quite, the full
story. In deriving (8.11), we have not taken into account the pform character
of the charges. Taking this into account shows that the phase in (8.9) vanishes
for a measurezero set of conﬁgurations of the electric and magnetic branes.
59
This is easiest to explain in a simpliﬁed case where the electric and mag
netic branes are kept in static ﬂat conﬁgurations, with the electric pbrane ori
ented along the directions {x
M1
. . . x
Mp
}. The phase factor (8.9) then becomes
exp(iQ
e
W
A
M1...MpR
∂x
R
/∂σ), where σ is an ordering parameter for the closed
sequence of deformations W. In making this deformation sequence, we recall
from the above discussion that one should restrict the deformations to pre
serve the asymptotic orientation of the deformed pbrane. For simplicity, one
may simply consider moving the electric pbrane by parallel transport around
the magnetic ˆ pbrane in a closed loop. The accrued phase factor is invariant
under gauge transformations of the potential A
[p+1]
. This makes it possible to
simplify the discussion by making use of a specially chosen gauge. Note that
magnetic ˆ pbranes have purely transverse ﬁeld strengths like (2.27b); there is
accordingly a gauge in which the gauge potential A
[p+1]
is also purely trans
verse, i.e. it vanishes whenever any of its indices point along a worldvolume
direction of the magnetic ˆ pbrane. Consideration of more general deformation
sequences yields the same result.
59
Now one can see how the Diracinsensitive conﬁgurations arise: the phase
in (8.9) vanishes whenever there is even a partial alignment between the electric
and the magnetic branes, i.e. when there are shared worldvolume directions
between the two branes. This measurezero set of Diracinsensitive conﬁgura
tions may be simply characterized in terms of the p and ˆ p charges themselves
by the condition Q
el
[p]
∧ Q
mag
[ ˆ p]
= 0. For such conﬁgurations, one obtains no
Dirac quantisation condition. To summarize, one may incorporate this ori
entation restriction into the Dirac quantisation condition (8.11) by writing a
(p + ˆ p)form quantisation condition
Q
el
[p]
∧ Q
mag
[ ˆ p]
= 2πn
Q
el
[p]
∧ Q
mag
[ ˆ p]
Q
el
[p]
 Q
mag
[ ˆ p]

, n ∈ Z , (8.12)
76
which reduces to (8.11) for all except the Diracinsensitive set of conﬁgurations.
8.3 Charge quantisation conditions and dimensional reduction
The existence of Diracinsensitive conﬁgurations may seem to be of only pe
ripheral importance, given that they constitute only a measurezero subset
of the total set of asymptotic brane conﬁgurations. However, their relevance
becomes more clear when one considers the relations existing between the p
form charges under dimensional reduction. Let us recall the relations (6.11)
between the ﬁeld strengths in diﬀerent dimensions. Now, the electric and mag
netic charges carried by branes in D dimensions take the forms
Q
e
=
e
c·
φ ∗
F +κ(A)
(8.13a)
Q
m
=
˜
F , (8.13b)
where
˜
F = dA, F =
˜
F + (ChernSimons modiﬁcations) (i.e. modiﬁcations
involving lowerorder forms arising in the dimensional reduction similar to
those in the D = 9 case (6.41)) and c is the dilaton vector corresponding to
F in the dimensionallyreduced action (6.13). The term κ(A) in (8.13a) is
the analogue of the term
1
2
A
[3]
∧ F
[4]
in (1.3). From the expressions (6.11) for
the reduced ﬁeld strengths and their duals, one obtains the following relations
between the original charges in D = 11 and those in the reduced theory:
Table 3: Relations between Q
11
and Q
D
F
[4]
F
i
[3]
F
ij
[2]
F
ijk
[1]
Electric Q
11
e
= Q
D
e
V Q
D
e
V
Li
Q
D
e
V
Li Lj
Q
D
e
V
Li Lj L
k
Magnetic Q
11
m
= Q
D
m
Q
D
m
L
i
Q
D
m
L
i
L
j
Q
D
m
L
i
L
j
L
k
where L
i
=
dz
i
is the compactiﬁcation period of the reduction coordinate
z
i
and V =
d
11−D
z =
¸
11−D
i=1
L
i
is the total compactiﬁcation volume. Note
that the factors of L
i
cancel out in the various products of electric and magnetic
charges only for charges belonging to the same ﬁeld strength in the reduced
dimension D.
Now consider the quantisation conditions obtained between the various di
mensionally reduced charges shown in Table 3. We need to consider the various
schemes possible for dimensional reduction of dual pairs of (p, ˆ p) branes. We
77
have seen that for singleelement brane solutions, there are two basic schemes,
as explained in Section 6: diagonal, which involves reduction on a worldvolume
coordinate, and vertical, which involves reduction on a transverse coordinate
after preparation by “stacking up” singlecenter solutions so as to generate a
transversespace translation invariance needed for the dimensional reduction.
For the dimensional reduction of a solution containing two elements, there
are then four possible schemes, depending on whether the reduction coordinate
z belongs to the worldvolume or to the transverse space of each brane. For
an electric/magnetic pair, we have the following four reduction possibilities:
diagonal/diagonal, diagonal/vertical, vertical/diagonal and vertical/vertical.
Only the mixed cases will turn out to preserve Dirac sensitivity in the lower
dimension after reduction.
This is most easily illustrated by considering the diagonal/diagonal case,
for which z belongs to the worldvolumes of both branes. With such a shared
worldvolume direction, one has clearly fallen into the measurezero set of Dirac
insensitive conﬁgurations with Q
el
[p]
∧ Q
mag
[ ˆ p]
= 0 in the higher dimension D.
Correspondingly, in (D −1) dimensions one ﬁnds that the diagonally reduced
electric (p − 1) brane is supported by an n = p + 1 form ﬁeld strength, but
the diagonally reduced magnetic (ˆ p − 1) brane is supported by an n = p + 2
form; since only branes supported by the same ﬁeld strength can have a Dirac
quantisation condition, this diagonal/diagonal reduction properly corresponds
to a Diracinsensitive conﬁguration.
Now consider the mixed reductions, e.g. diagonal/vertical. In performing
a vertical reduction of a magnetic ˆ pbrane by stacking up an inﬁnite deck of
singlecenter branes in order to create the R translational invariance necessary
for the reduction, the total magnetic charge will clearly diverge. Thus, in a
vertical reduction it is necessary to reinterpret the magnetic charge Q
m
as
a charge density per unit z compactiﬁcation length. Before obtaining the
Dirac quantisation condition in the lower dimension, it is necessary to restore
a gravitationalconstant factor of κ
2
that should properly have appeared in
the quantisation conditions (8.11, 8.12). As one may verify, the electric and
magnetic charges as deﬁned in (1.3, 1.4) are not dimensionless. Thus, (8.11)
in D = 11 should properly have been written Q
e
Q
m
= 2πκ
2
11
n. If one lets the
compactiﬁcation length be denoted by L in the Ddimensional theory prior
to dimensional reduction, then one obtains a Dirac phase exp(iκ
−2
D−1
Q
e
Q
m
L).
This ﬁts precisely, however, with another aspect of dimensional reduction:
the gravitational constants in dimensions D and D − 1 are related by κ
2
D
=
Lκ
2
D−1
. Thus, in dimension D − 1 one obtains the expected quantisation
condition Q
e
Q
m
= 2πκ
2
D−1
n. Note, correspondingly, that upon making a
mixed diagonal/vertical reduction, the electric and magnetic branes remain
78
dual to each other in the lower dimension, supported by the same n = p−1+2 =
p +1 form ﬁeld strength. The opposite mixed vertical/diagonal reduction case
goes similarly, except that the dual branes are then supported by the same
n = p + 2 form ﬁeld strength.
In the ﬁnal case of vertical/vertical reduction, Dirac sensitivity is lost
in the reduction, not owing to the orientation of the branes, but because in
this case both the electric and the magnetic charges need to be interpreted
as densities per unit compactiﬁcation length, and so one obtains a phase
exp(iκ
−2
D
Q
e
Q
m
L
2
). Only one factor of L is absorbed into κ
2
D−1
, and one has
lim
L→0
L
2
/κ
2
D
= 0. Correspondingly, the two dimensionally reduced branes in
the lower dimension are supported by diﬀerent ﬁeld strengths: an n = p + 2
form for the electric brane and an n = p + 1 form for the magnetic brane.
Thus, there is a perfect accord between the structure of the Dirac quantisa
tion conditions for pform charges in the various supergravity theories related
by dimensional reduction. The existence of Diracinsensitive conﬁgurations
plays a central rˆole in establishing this accord, even though they represent
only a subset of measure zero from the point of view of the higherdimensional
theory.
Another indication of the relevance of the Diracinsensitive conﬁgurations
is the observation
59
that all the intersectingbrane solutions with some degree
of preserved supersymmetry, as considered in Section 7, correspond to Dirac
insensitive conﬁgurations. This may immediately be seen in such solutions as
the 2 ⊥ 5(1) solution (7.10), but it is also true for solutions involving pp wave
and TaubNUT elements.
8.4 Counting pbranes
As we have seen at the classical level, the classifying symmetry for solutions in a
given scalar vacuum, speciﬁed by the values of the scalar moduli, is the linearly
realized isotropy symmetry H given in Table 2. When one takes into account
the Dirac quantisation condition, this classifying symmetry becomes restricted
to a discrete group, which clearly must be a subgroup of the corresponding
G(Z) duality group, so in general one seeks to identify the group G(Z)∩H. The
value of this intersection is modulusdependent, showing that the homogeneity
of the
G
/H coset space is broken at the quantum level by the quantisation
condition. Classically, of course, the particular point on the vacuum manifold
G
/H corresponding to the scalar moduli can be changed by application of
a transitivelyacting G transformation, for example with a group element g.
Correspondingly, the isotropy subgroup H moves by conjugation with g,
H −→gHg
−1
. (8.14)
79
The discretized duality group G(Z), on the other hand, does not depend
upon the moduli. This is because the modulus dependence cancels out in the
“canonical” charges that we have deﬁned in Eq. (8.13). One way to see this is
to use the relations between charges in diﬀerent dimensions given in Table 3,
noting that there are no scalar moduli in D = 11, so the modulusindependent
relations of Table 3 imply that the lowerdimensional charges (8.13) do not
depend on the moduli.
r
Another way to understand this is by comparison with ordinary Maxwell
electrodynamics, where an analogous charge would be that derived from the
action I
Max
= −
1
/(4e
2
)
F
can
µν
F
can µν
, corresponding to a covariant derivative
D
µ
= ∂
µ
+iA
can
µ
. This is analogous to our dimensionally reduced action (6.13)
from which the charges (8.13) are derived, because the modulus factors e
c·
φ∞
appearing in (6.13) (together with the rest of the
φ dilatonic scalar dependence)
play the rˆoles of coupling constant factors like e
−2
. If one wants to compare
this to the “conventional” charges deﬁned with respect to a conventional gauge
potential A
conv
µ
= e
−1
A
can
µ
, for which the action is −
1
/4
F
conv
µν
F
conv µν
, then
the canonical and conventional charges obtained via Gauss’s law surface inte
grals are related by
Q
can
=
1
2e
2
d
2
Σ
ij
ǫ
ijk
F
can0k
=
1
2e
d
2
Σ
ij
ǫ
ijk
F
conv 0k
=
1
e
Q
conv
. (8.15)
Thus, in the Maxwell electrodynamics case, the dependence on the electric
charge unit e drops out in Q
can
, although the conventional charge Q
conv
scales
proportionally to e. The modulus independence of the charges (8.13) works
in a similar fashion. Then, given that the discretized quantum duality group
G(Z) is deﬁned by the requirement that it map the set of Diracallowed charges
onto itself, it is evident that the group G(Z), referred to the canonical charges
(8.13), does not depend on the moduli.
As a consequence of the diﬀerent modulus dependences of H and of G(Z),
it follows that the size of the intersection group G(Z) ∩ H is dependent on
the moduli. The analogous feature in ordinary Maxwell theory is that a true
duality symmetry of the theory only arises when the electric charge takes the
value e = 1 (in appropriate units), since the duality transformation maps
e →e
−1
. Thus, the value e = 1 is a distinguished value.
r
Note that the compactiﬁcation periods L
i
appearing in Table 3 have values that may be
adjusted by convention. These should not be thought of as determining the geometry of the
compactifying internal manifold, which is determined instead by the scalar moduli. Thus,
the relations of Table 3 imply the independence of the canonicallydeﬁned charges from the
physically relevant moduli.
80
The distinguished point on the scalar vacuum manifold for general super
gravity theories is the one where all the scalar moduli vanish. This is the point
where G(Z) ∩ H is maximal. Let us return to our D = 8 example to help
identify what this group is. In that case, for the scalars (σ, χ), we may write
out the transformation in detail using (8.3):
e
−σ
−→ (d +cχ)
2
e
−σ
+c
2
e
σ
χe
−σ
−→ (d +cχ)(b +aχ)e
−σ
+ace
σ
. (8.16)
Requiring a, b, c, d ∈ Z and also that the modulus point σ
∞
= χ
∞
= 0 be left
invariant, we ﬁnd only two transformations: the identity and the transforma
tion a = d = 0, b = −1, c = 1, which maps σ and χ according to
e
−σ
−→ e
σ
+χ
2
e
−σ
χe
−σ
−→ −χe
−σ
. (8.17)
Thus, for our truncated system, we ﬁnd just an S
2
discrete symmetry as the
quantum isotropy subgroup of SL(2, Z) at the distinguished point on the scalar
vacuum manifold. This S
2
is the natural analogue of the S
2
symmetry that
appears in Maxwell theory when e = 1.
In order to aid in identifying the pattern behind this D = 8 example,
suppose that the zeroform gauge potential χ is small, and consider the S
2
transformation to lowest order in χ. To this order, the transformation just
ﬂips the signs of σ and χ. Acting on the ﬁeld strengths (F
[4]
, G
[4]
), one ﬁnds
(F
[4]
, G
[4]
) −→(−G
[4]
, F
[4]
) . (8.18)
One may again check (in fact to all orders, not just to lowest order in χ) that
(8.18) maps the ﬁeld equation for F
[4]
into the corresponding Bianchi identity:
∇
M
(e
σ
F
MNPQ
+χ
∗
F
MNPQ
) −→−∇
M
∗
F
MNPQ
. (8.19)
Considering this S
2
transformation to lowest order in the zeroform χ has the
advantage that the signﬂip of φ may be “impressed” upon the a dilaton vector
for F
[4]
: a →−a. The general structure of such G(Z) ∩H transformations will
be found by considering the impressed action of this group on the dilaton
vectors.
Now consider the
SL(3, R)
/SO(3) sector of the D = 8 scalar manifold,
again with the moduli set to the distinguished point on the scalar manifold.
To lowest order in zeroform gauge potentials, the action of SL(3, Z) ∩ H may
similarly by impressed upon the 3form dilaton vectors, causing in this case a
permutation of the a
i
, generating for the D = 8 case overall the discrete group
81
S
3
×S
2
. Now that we have a bit more structure to contemplate, we can notice
that the G(Z) ∩ H transformations leave the (a, a
i
) dot products invariant.
61
The invariance of the dilaton vectors’ dot products prompts one to return
to the algebra (6.15) of these dot products and see what else we may recognize
in it. Noting that the duality groups given in Table 2 for the higher dimen
sions D involve SL(N, R) groups, we recall that the weight vectors
h
i
of the
fundamental representation of SL(N, R) satisfy
h
i
·
h
j
= δ
ij
−
1
N
,
N
¸
i=1
h
i
= 0 . (8.20)
These relations are precisely those satisﬁed by
±1
√
2
a and
1
√
2
a
i
, corresponding
to the cases N = 2 and N = 3. This suggests that the action of the maximal
G(Z) ∩ H group (i.e. for scalar moduli set to the distinguished point on the
scalar manifold) may be identiﬁed in general with the symmetry group of the
set of fundamental weights for the corresponding supergravity duality group
G as given in Table 2. The symmetry group of the fundamental weights is the
Weyl group
61
of G, so the action of the maximal G(Z) ∩H pbrane classifying
symmetry is identiﬁed with that of the Weyl group of G.
As one proceeds down through the lowerdimensional cases, where the
supergravity symmetry groups shown in Table 2 grow in complexity, the above
pattern persists:
61
in all cases, the action of the maximal classifying symmetry
G(Z) ∩H may be identiﬁed with the Weyl group of G. This is then the group
that counts the distinct pbrane solutions
s
of a given type (4.6), subject to the
Dirac quantisation condition and referred to the distinguished point on the
scalar modulus manifold. For example, in D = 7, where from Table 2 one sees
that G = SL(5, R) and H = SO(5), one ﬁnds that the action of G(Z) ∩ H is
equivalent to that of the discrete group S
5
, which is the Weyl group of SL(5, R).
In the lowerdimensional cases shown in Table 2, the discrete group G(Z) ∩H
becomes less familiar, and is most simply described as the Weyl group of G.
From the analysis of the Weylgroup duality multiplets, one may tabu
late
61
the multiplicities of pbranes residing at each point of the plot given in
Figure 7. For supersymmetric pbranes arising from a set of N participating
ﬁeld strengths F
[n]
, corresponding to ∆ = 4/N for the dilatonic scalar cou
pling, one ﬁnds the multiplicities given in Table 4. By combining these duality
multiplets together with the diagonal and vertical dimensional reduction fam
ilies discussed in Sections 6 and 6.3, the full set of p ≤ (D − 3) branes shown
in Figure 7 becomes “welded” together into one overall symmetrical structure.
s
Of course, these solutions must also fall into supermultiplets with respect to the unbroken
supersymmetry; the corresponding supermultiplet structures have been discussed in Ref.
64
82
Table 4: Examples of pbrane Weylgroup multiplicities
D
F
[n]
∆ 10 9 8 7 6 5 4
F
[4]
4 1 1 2
F
[3]
4 1 2 3 5 10
4 1 1+2 6 10 16 27 56
F
[2]
2 2 6 15 40 135 756
4
/3 45 2520
4 2 8 20 40 72 126
F
[1]
2 12 60 280 1080 3780
4
/3 480 4320 30240+2520
8.5 The charge lattice
For the electric and magnetic BPS brane solutions supported by a given ﬁeld
strength, we have seen above that the Dirac charge quantisation condition
(8.12) implies that, given a certain minimum “electric” charge (8.13a), the
allowed set of magnetic charges is determined. Then, taking the minimum
magnetic charge from this set, the argument may be turned around to show
that the set of allowed electric charges is given by integer multiples of the
minimum electric charge. This argument does not directly establish, however,
what the minimum electric charge is, i.e. the value of the charge unit. This
cannot be established by use of the Dirac quantisation condition alone.
There are other tools, however, that one can use to ﬁx the charge lattice
completely. To do so, we shall need to exploit the existence of certain spe
cial “unitsetting” brane types, and also to exploit fully the consequences of
the assumption that the G(Z) duality symmetry remains exactly valid at the
quantum level. We have already encountered one example of a “unitsetting”
brane in subsection 7.2, where we encountered the pp wave/TaubNUT pair
of D = 11 solutions. We saw there that the TaubNUT solution (7.7) is non
singular provided that the coordinate ψ is periodically identiﬁed with period
L = 4πk, where k is the chargedetermining parameter in the 3dimensional
harmonic function H(y) = 1 + k/(y). Upon dimensional reduction down to
D = 10, one obtains a magnetic 6brane solution, with a charge classically
discretized to take a value in the set
Q
m
= rL , r ∈ Z . (8.21)
Given these values for the magnetic charge, the D = 10 Dirac quantisation
83
condition
Q
e
Q
m
= 2πκ
2
10
n , n ∈ Z , (8.22)
or, equivalently, as we saw in subsection 7.2, the quantisation of D = 11 pp
wave momentum in the compact ψ direction, gives an allowed set of electric
charges
Q
e
=
2πκ
2
10
L
n , n ∈ Z . (8.23)
Thus, the requirement that magnetic D = 10 6branes oxidize up to non
singular TaubNUT solutions in D = 11 fully determines the 6brane electric
and magnetic charge units and not just the product of them which occurs in
the Dirac quantisation condition
If one assumes that the G(Z) duality symmetries remain strictly unbro
ken at the quantum level, then one may relate the 6brane charge units to
those of other BPS brane types.
t
In doing so, one must exploit the fact that
brane solutions with Poincar´e worldvolume symmetries may be dimensionally
reduced down to lower dimensions, where the duality groups shown in Table 2
grow larger. In a given dimension D, the G(Z) duality symmetries only rotate
between pbranes of the same worldvolume dimension, supported by the same
kind of ﬁeld strength, as we have seen from our discussion of the Weylgroup
action on pbranes given in subsection 8.4. Upon reduction down to dimen
sions D
red
< D, however, the solutions descending from an original pbrane
in D dimensions are subject to a larger G(Z) duality symmetry, and this can
be used to rotate a descendant brane into descendants of p
′
branes for various
values of p
′
. Dimensional oxidation back up to D dimensions then completes
the link, establishing relations via the duality symmetries between various BPS
brane types which can be supported by diﬀerent ﬁeld strengths, including ﬁeld
strengths of diﬀerent rank.
65
This link may be used to establish relations be
tween the charge units for the various pform charges of diﬀering rank, even
though the corresponding solutions are Diracinsensitive to each each other.
Another chargeunitsetting BPS brane species occurs in the D = 10 type
IIB theory. This theory has a wellknown diﬃculty with the formulation of a
satisfactory action, although its ﬁeld equations are perfectly welldeﬁned. The
diﬃculty in formulating an action arise from the presence of a selfdual 5form
ﬁeld strength, H
[5]
=
∗
H
[5]
. The corresponding electrically and magnetically
charged BPS solutions are 3branes, and, owing to the selfduality condition,
these solutions are actually dyons, with a charge vector at 45
◦
to the electric
axis. We shall consider the type IIB theory in some more detail in Section
9; for now, it will be suﬃcient for us to note that the dyonic 3branes of
t
For details of the duality relations between charge units for diﬀerent pbranes, see Ref.
59
84
D = 10 type IIB theory are also a unitsetting brane species.
59
The unitsetting
property arises because of a characteristic property of the DiracSchwinger
Zwanziger quantisation condition for dyons in dimensions D = 4r + 2: for
dyons (Q
(1)
e
, Q
(1)
m
), (Q
(2)
e
, Q
(2)
m
), this condition is symmetric:
59
Q
(1)
e
Q
(2)
m
+Q
(2)
e
Q
(1)
m
= 2πκ
2
4r+2
n , n ∈ Z , (8.24)
unlike the more familiar antisymmetric DSZ condition that is obtained in di
mensions D = 4r. The symmetric nature of (8.24) means that dyons may be
Diracsensitive to others of their own type,
u
quite diﬀerently from the anti
symmetric cases in D = 4r dimensions. For the 45
◦
dyonic 3branes, one thus
obtains the quantisation condition
Q
[3]
 = n
√
πκ
IIB
, n ∈ Z (8.25)
where κ
IIB
is the gravitational constant for the type IIB theory. Then, using
duality symmetries, one may relate the
√
πκ
IIB
charge unit to those of other
supergravity R–R charges.
Thus, using duality symmetries together with the pp wave/TaubNUT and
selfdual 3brane charge scales, one may determine the chargelattice units for
all BPS brane types.
65,59
. It is easiest to express the units of the resulting over
all charge lattice by making a speciﬁc choice for the compactiﬁcation periods.
If one lets all the compactiﬁcation periods L
i
be equal,
L
i
= L
IIB
= L = (2πκ
2
11
)
1
9
, (8.26)
then the electric and magnetic chargelattice units for rankn ﬁeld strengths
in dimension D are determined to be
59
∆Q
e
= L
D−n−1
, ∆Q
m
= L
n−1
. (8.27)
9 Local versus active dualities
The proper interpretation of the discretized CremmerJulia G(Z) duality sym
metry at the level of supergravity theory is subject to a certain amount of
debate, but at the level of string theory the situation becomes more clear. In
any dimension D, there is a subgroup of G(Z) that corresponds to T duality,
which is a perturbative symmetry holding orderbyorder in the string loop
expansion. T duality
66
consists of transformations that invert the radii of a
u
They will be Diracsensitive provided one of them is slightly rotated so as to avoid hav
ing any common worldvolume directions with the other, in order to avoid having a Dirac
insensitive conﬁguration as discussed in subsection 8.3.
85
toroidal compactiﬁcation, under which quantized string oscillator modes and
string winding modes become interchanged. Aside from such a relabeling, how
ever, the overall string spectrum remains unchanged. Hence, T duality needs
to be viewed as a local symmetry in string theory, i.e. string conﬁgurations on
compact manifolds related by T duality are identiﬁed. Depending on whether
one considers (D−3) branes to be an unavoidable component of the spectrum,
the same has also been argued to be the case at the level of the supergravity
eﬀective ﬁeld theory.
67
The wellfounded basis, in string theory at least, for a
local interpretation of the T duality subgroup of G(Z) has led subsequently to
the hypothesis
62,63
that the full duality group G(Z) should be given a local
interpretation: sets of string solutions and moduli related by G(Z) transforma
tions are to be treated as equivalent descriptions of a single state. This local
interpretation of the G(Z) duality transformations is similar to that adopted
for general coordinate transformations viewed passively, according to which,
e.g., ﬂat space in Cartesian or in Rindler coordinates is viewed as one and the
same solution.
As with general coordinate transformations, however, duality symmetries
may occur in several diﬀerent guises that are not always clearly distinguished.
As one can see from the charge lattice discussed in subsection 8.5, there is also
a G(Z) covariance of the set of charge vectors for physically inequivalent BPS
brane solutions. In the discussion of subsection 8.5, we did not consider in
detail the action of G(Z) on the moduli, because, as we saw in subsection 8.4,
the canonicallydeﬁned charges (8.13) are in fact modulusindependent.
Since the dilatonic and axionic scalar moduli determine the coupling con
stants and vacuum θangles of the theory, these quantities should be ﬁxed
when quantizing about a given vacuum state of the theory. This is similar to
the treatment of asymptotically ﬂat spacetime in gravity, where the choice of a
particular asymptotic geometry is necessary in order to establish the “vacuum”
with respect to which quantized ﬂuctuations can be considered.
Thus, in considering physicallyinequivalent solutions, one should compare
solutions with the same asymptotic values of the scalar ﬁelds. When this is
done, one ﬁnds that solutions carrying charges (8.13) related by G(Z) transfor
mations generally have diﬀering mass densities. Since the standard Cremmer
Julia duality transformations, such as those of our D = 8 example in subsection
8.1, commute with P
0
time translations and so necessarily preserve mass den
sities, it is clear that the BPS spectrum at ﬁxed scalar moduli cannot form
a multiplet under the standard CremmerJulia G(Z) duality symmetry. This
conclusion is in any case unavoidable, given the local interpretation adopted
for the standard duality transformations as discussed above: once one has
identiﬁed solution/modulus sets under the standard G(Z) duality transforma
86
tions, one cannot then turn around and use the same G(Z) transformations to
generate inequivalent solutions.
Thus, the question arises: is there any spectrumgenerating symmetry
lying behind the apparently G(Z) invariant charge lattices of inequivalent so
lutions that we saw in subsection 8.5? At least at the classical level, and for
singlecharge (i.e. ∆ = 4) solutions, the answer
68
turns out to be ‘yes.’ We
shall illustrate the point using type IIB supergravity as an example.
v
9.1 The symmetries of type IIB supergravity
Aside from the diﬃculties arising from the selfduality condition for the 5
form ﬁeld strength H
[5]
, the equations of motion of the bosonic ﬁelds of the
IIB theory may be derived from the action
I
IIB
10
=
d
10
x[eR+
1
4
e tr(∇
µ
M
−1
∇
µ
M)−
1
12
e H
T
[3]
MH
[3]
−
1
240
e H
2
[5]
−
1
2
√
2
ǫ
ij
∗
(B
[4]
∧dA
(i)
[2]
∧dA
(j)
[2]
)] . (9.1)
The 5form selfduality condition
H
[5]
=
∗
H
[5]
(9.2)
may be handled in the fashion of Ref.
70
, being imposed by hand as an extra
constraint on the ﬁeld equations obtained by varying (9.1). This somewhat
hybrid procedure will be suﬃcient for our present purposes.
The matrix Min (9.1) contains two scalar ﬁelds: a dilatonic scalar φ which
occurs nonlinearly through its exponential, and an axionic scalar χ, which may
also be considered to be a zeroform gauge potential; explicitly, one has
M=
e
−φ
+χ
2
e
φ
χe
φ
χe
φ
e
φ
. (9.3)
The doublet H
[3]
contains the ﬁeld strengths of the 2form gauge potentials
A
[2]
:
H
[3]
=
dA
(1)
[2]
dA
(2)
[2]
. (9.4)
The action (9.1) is invariant under the SL(2, R) transformations
H
[3]
−→(Λ
T
)
−1
H
[3]
, M−→ΛMΛ
T
, (9.5)
v
For a detailed discussion of SL(2, R) duality in type IIB supergravity, see Ref.
69
87
where the SL(2, R) parameter matrix is
Λ =
a b
c d
, (9.6)
and the SL(2, R) constraint is ad − bc = 1. If one deﬁnes the complex scalar
ﬁeld τ = χ + i e
−φ
, then the transformation on M can be rewritten as the
fractional linear transformation
τ −→
aτ +b
cτ +d
. (9.7)
Note that since H
[5]
is a singlet under SL(2, R), the selfduality constraint
(9.2), which is imposed by hand, also preserves the SL(2, R) symmetry. Since
this SL(2, R) transformation rotates the doublet A
2]
of electric 2form poten
tials amongst themselves, this is an “electricelectric” duality, as opposed to
the “electricmagnetic” duality discussed in the D = 8 example of subsection
8.1. Nonetheless, similar issues concerning duality multiplets for a ﬁxed scalar
vacuum arise in both cases.
There is one more symmetry of the equations of motion following from the
action (9.1). This is a rather humble symmetry that is not often remarked
upon, but which will play an important role in constructing active SL(2, R)
duality transformations for the physically distinct BPS string and 5brane mul
tiplets of the theory. As for pure sourcefree Einstein theory, the action (9.1)
transforms homogeneously as λ
3
under the following scaling transformations:
g
µν
−→λ
2
g
µν
, A
(i)
[2]
−→λ
2
A
(i)
[2]
, H
[5]
−→λ
4
H
[5]
; (9.8)
note that the power of λ in each ﬁeld’s transformation is equal to the number
of indices it carries, and, accordingly, the scalars φ and χ are not transformed.
Although the transformation (9.8) does not leave the action (9.1) invariant,
the λ
3
homogeneity of this scaling for all terms in the action is suﬃcient to
produce a symmetry of the IIB equations of motion. It should be noted that
the SL(2, R) electricmagnetic duality of the D = 8 example given in subsection
8.1 shares with the transformation (9.8) the feature of being a symmetry only
of the equations of motion, and not of the action.
The SL(2, R) transformations map solutions of (9.1) into other solutions.
We shall need to consider in particular the action of these transformations on
the charges carried by solutions. From the equations of motion of the 3form
ﬁeld strength H
[3]
in (9.1),
d
∗
(MH
[3]
) = −
1
√
2
H
[5]
∧ ΩH
[3]
, (9.9)
88
where Ω is the SL(2, R)invariant tensor
Ω =
0 1
−1 0
, (9.10)
one ﬁnds that the following twocomponent quantity is conserved:
Q
e
=
∗
(MH
[3]
) +
1
3
√
2
Ω(2B
[4]
∧ H
[3]
−H
[5]
∧ A
[2]
)
. (9.11)
Under an SL(2, R) transformation, Q
e
transforms covariantly as a doublet:
Q
e
→ΛQ
e
.
By virtue of the Bianchi identities for the 3form ﬁeld strength, one has in
addition a topologicallyconserved magnetic charge doublet,
Q
m
=
H
[3]
, (9.12)
which transforms under SL(2, R) as Q
m
→(Λ
T
)
−1
Q
m
, i.e. contragrediently to
Q
e
. The transformation properties of the electric and magnetic charge doublets
are just such as to ensure that the Dirac quantization condition Q
T
m
Q
e
∈
2πκ
2
IIB
Z is SL(2, R) invariant.
The overall eﬀect of this standard SL(2, R) symmetry on type IIB super
gravity solutions may be expressed in terms of its action on the solutions’
charges and on the scalar moduli. This group action may be viewed as an
automorphism of a vector bundle, with the scalar ﬁelds’
SL(2, R)
/SO(2) target
manifold as the base space, and the charge vector space as the ﬁber.
We have seen in our general discussion of charge lattices in subsection
8.5 that the continuous classical CremmerJulia symmetries G break down to
discretized G(Z) symmetries that map between states on the quantum charge
lattice. In the present type IIB case, the classical SL(2, R) symmetry breaks
down to SL(2, Z) at the quantum level. Taking the basis states of the IIB
charge lattice to be
e
1
=
1
0
e
2
=
0
1
, (9.13)
the surviving SL(2, Z) group will be represented by SL(2, R) matrices with
integral entries.
As we have discussed above, the discretized duality symmetries G(Z) are
given a local interpretation in string theory. In the case of the type IIB theory,
this is a hypothesis rather than a demonstrated result, because the SL(2, Z)
89
transformations map between NS–NS and R–R states, and this is a distinctly
nonperturbative transformation. Adopting this hypothesis nonetheless, an
orbit of the standard SL(2, Z) transformation reduces to a single point; after
making the corresponding identiﬁcations, the scalar modulus space becomes
the double coset space
SL(2, Z)
\
SL(2, R)
/
SO(2)
.
9.2 Active duality symmetries
Now let us see how duality multiplets of the physically inequivalent BPS states
can occur, even though they will contain states with diﬀerent mass densities.
This latter fact alone tells us that we must include some transformation that
acts on the metric. We shall continue with our exploration of the continu
ous classical SL(2, R) symmetry of the type IIB theory. Finding the surviving
quantumlevel SL(2, Z) later on will be a straightforward matter of restrict
ing the transformations to a subgroup. The procedure starts with a stan
dard SL(2, R) transformation, which transforms the doublet charges (9.11) in
a straightforwardly linear fashion, but which also transforms in an unwanted
way the scalar moduli. Subsequent compensating transformations will then
have the task of eliminating the unwanted transformation of the scalar mod
uli, but without changing the “already ﬁnal” values of the charges. Let us
suppose that this initial transformation, with parameter Λ, maps the charge
vector and complex scalar modulus (Q, τ
∞
) to new values (Q
′
, τ
′
∞
).
After this initial Λ transformation, one wishes to return the complex scalar
modulus τ
′
∞
to its original value τ
∞
, in order to obtain an overall transforma
tion that does not in the end disturb the complex modulus. To do this, notice
that within SL(2, R) there is a subgroup that leaves a doublet charge vector
Q
′
invariant up to an overall rescaling. This projective stability group of Q
′
is
isomorphic to the Borel subgroup of SL(2, R):
Borel =
a b
0 a
−1
a, b ∈ R
. (9.14)
This standard representation of the SL(2, R) Borel subgroup clearly leaves the
basis charge vector e
1
of Eq. (9.13) invariant up to scaling by a. For a general
charge vector Q
′
, there will exist a corresponding projective stability subgroup
which is isomorphic to (9.14), but obtained by conjugation of (9.14) with an
element of H
∼
= SO(2). The importance of the Borel subgroup for our present
purposes is that it acts transitively on the
G
/H =
SL(2, R)
/SO(2) coset space
in which the scalar ﬁelds take their values, so this transformation may be
used to return the scalar moduli to the original values they had before the Λ
transformation.
90
The next step in the construction is to correct for the unwanted scaling
Q
′
→aQ
′
which occurs as a result of the Borel compensating transformation,
by use of a further compensating scaling of the form (9.8), aQ
′
→ λ
2
aQ
′
,
in which one picks the rigid parameter λ such that λ
2
a = 1. This almost
completes the construction of the active SL(2, R). For the ﬁnal step, note
that the transformation (9.8) also scales the metric, g
µν
→ λ
2
g
µν
= a
−1
g
µν
.
Since one does not want to alter the asymptotic metric at inﬁnity, one needs
to compensate for this scaling by a ﬁnal general coordinate transformation,
x
µ
→x
′µ
= a
−1/2
x
µ
.
The overall active SL(2, R) duality package constructed in this way trans
forms the charges in a linear fashion, Q →λQ
′
, in exactly the same way as the
standard supergravity CremmerJulia SL(2, R) duality, but now leaving the
complex scalar modulus τ
∞
unchanged. This is achieved by a net construction
that acts upon the ﬁeld variables of the theory in a quite nonlinear fashion.
This net transformation may be explicitly written by noting that for SL(2, R)
there is an Iwasawa decomposition
Λ =
˜
bh , (9.15)
where
˜
b ∈ Borel
Q
′ is an element of the projective stability group of the ﬁnal
charge vector Q
′
and where h ∈ H
τ∞
is an element of the stability subgroup
of τ
∞
. Clearly, the Borel transformation that is needed in this construction is
just b = (
˜
b)
−1
, leaving thus a transformation h ∈ H
τ∞
which does not change
the complex modulus τ
∞
. The compensating scaling transformation t of the
form (9.8) and the associated general coordinate transformation also leave the
scalar moduli unchanged. The net active SL(2, R) transformation thus is just
btΛ = th. Speciﬁcally, for τ
∞
= χ
∞
+ie
−φ∞
and a transformation Λ mapping
Q
i
=
p
i
q
i
to Q
f
=
p
f
q
f
= ΛQ
i
, the h ∈ H
τ∞
group element is
h
fi
= V
∞
cos
˜
θ
fi
sin
˜
θ
fi
−sin
˜
θ
fi
cos
˜
θ
fi
V
−1
∞
,
˜
θ
fi
=
˜
θ
f
−
˜
θ
i
tan
˜
θ
i
= e
φ∞
(tan θ
i
−χ
∞
) , tan θ
i
= p
i
/q
i
, (9.16)
where the matrix V
∞
is an element of Borel that has the eﬀect of moving the
scalar modulus from the point τ = i to the point τ
∞
:
V
∞
= e
−φ∞/2
1 e
φ∞
χ
∞
0 e
φ∞
. (9.17)
The matrix V
∞
appearing here is also the asymptotic limit of a matrix V (φ, χ)
that serves to factorize the matrix M given in (9.3), M = V V
T
. This fac
91
torization makes plain the transitive action of the Borel subgroup on the
SL(2, R)/SO(2) coset space in which the scalar ﬁelds take their values. Note
that the matrix M determines both the scalar kinetic terms and also their
interactions with the various antisymmetrictensor gauge ﬁelds appearing in
the action (9.1).
The scalingtransformation part of the net active SL(2, R) construction is
simply expressed as a ratio of mass densities,
t
fi
=
m
f
m
i
, m
2
i
= Q
T
i
M
−1
∞
Q
i
. (9.18)
This expression reﬂects the fact that the scaling symmetry (9.8) acts on the
metric and thus enables the active SL(2, R) transformation to relate solutions
at diﬀerent massdensity levels m
i,f
. Since, by contrast, the massdensity levels
are invariant under the action of the standard SL(2, R), it is clear that the two
realizations of this group are distinctly diﬀerent. Mapping between diﬀerent
mass levels, referred to a given scalar vacuum determined by the complex
modulus τ
∞
, can only be achieved by including the scaling transformation
(9.18).
The group composition property of the active SL(2, R) symmetry needs to
be checked in the same fashion as for nonlinear realizations generally, i.e. one
needs to check that a group operation O(Λ, Q) = th acting on an initial state
characterized by a charge doublet Q combines with a second group operation
according to the rule
O(Λ
2
, Λ
1
Q)O(Λ
1
, Q) = O(Λ
2
Λ
1
, Q) . (9.19)
One may verify directly that the nonlinear realization given by (9.16, 9.18)
does in fact satisfy this composition law, when acting on any of the ﬁelds of
the type IIB theory.
At the quantum level, the Dirac quantization condition restricts the al
lowed states of the theory to a discrete charge lattice, as we have seen. The
standard SL(2, R) symmetry thus becomes restricted to a discrete SL(2, Z) sub
group in order to respect this charge lattice, and the active SL(2, R) constructed
above likewise becomes restricted to an SL(2, Z) subgroup. This quantumlevel
discretized group of active transformations is obtained simply by restricting the
matrix parameters Λ for a classical active SL(2, R) transformation so as to lie
in SL(2, Z).
In lowerdimensional spacetime, the supergravity duality groups G shown
in Table 2 grow in rank and the structure of the charge orbits becomes progres
sively more and more complicated, but the above story is basically repeated
for an important class of pbrane solutions. This is the class of singlecharge
92
solutions, for which the charges Q fall into highestweight representations of
G. The duality groups shown in Table 2 are all maximally noncompact, and
possess an Iwasawa decomposition generalizing the SL(2, R) case (9.15):
Λ =
˜
bh
˜
b ∈ Borel
Q
, h ∈ H
moduli
, (9.20)
where Borel
Q
is isomorphic to the Borel subgroup of G. Once again, this
subgroup acts transitively on the coset space
G
/H in which the scalar ﬁelds
take their values, so this is the correct subgroup to use for a compensating
transformation to restore the moduli to their original values in a given scalar
vacuum. As in the SL(2, R) example of the type IIB theory, one may see
that this group action is transitive by noting that the matrix M (9.3) which
governs the scalar kinetic terms and interactions can be parameterized in the
form M= V V
#
, where V is an element of the Borel subgroup. The operation
# here depends on the groups G and H in question; in spacetime dimensions
D ≥ 4 we have
V
#
=
V
T
, for H orthogonal
V
†
, for H unitary
ΩV
†
, for H a USp group.
(9.21)
(The D = 3 case in which G = E
8(+8)
and H = SO(16) needs to be treated as
a special case.
71
)
Given the above grouptheoretical structure, the construction of active G
symmetry transformations that preserve the scalar moduli proceeds in strict
analogy with the type IIB SL(2, R) example that we have presented. This
construction depends upon the existence of a projective stability group
68,71
of the charge Q that is isomorphic to the Borel subgroup of G. This is the
case whenever Q transforms according to a highestweight representation of G.
The BPS brane solutions with this property are the singlecharge solutions with
∆ = 4. As we have seen in Section 7, BPS brane solutions with ∆ = 4/N can
be interpreted as coincidentchargecenter cases of intersectingbrane solutions
with N elements, each of which would separately be a ∆ = 4 solution on its
own. The construction of active duality symmetries for such multiplecharge
solutions remains an open problem, for they have a larger class of integration
constants, representing relative positions and phases of the charge components.
Only the asymptotic scalar moduli can be moved transitively by the Borel
subgroup of G and, correspondingly, the representations carried by the charges
in such multicharge cases are not of highestweight type.
The active G(Z) duality constructions work straightforwardly enough at
the classical level, but their dependence on symmetries of ﬁeld equations that
are not symmetries of the corresponding actions gives a reason for caution
93
about their quantum durability. This may be a subject where string theory
needs to intervene with its famed “miracles.” Some of these miracles can be
seen in supergravitylevel analyses of the persistence of BPS solutions with
arbitrary mass scales, despite the presence of apparently threatening quantum
corrections,
68
but a systematic way to understand the remarkable identities
making this possible is not known. Thus, there still remain some areas where
string theory appears to be more clever than supergravity.
10 Noncompact σmodels, null geodesics, and harmonic maps
A complementary approach
72,73,74
to the analysis of brane solutions in terms
of the four D = 11 elemental solutions presented in Section 7 is to make
a dimensional reduction until only overalltransverse dimensions remain, and
then to consider the resulting nonlinear σmodel supporting the solution. In
such a reduction, all of the worldvolume and relativetransverse coordinates
are eliminated, including the time coordinate, which is possible because the
BPS solutions are all time independent. The two complementary approaches
to the analysis of BPS brane solutions may thus be characterized as oxidation
up to the top of Figure 7, or reduction down to the left edge Figure 7, i.e.
reduction down to BPS “instantons,” or p = −1 branes, with worldvolume
dimension d = 0.
The d = 0 instanton solutions are supported by 1form ﬁeld strengths,
i.e. the derivatives of axionic scalars, F
[1]
= dχ. Taken together with the
dilatonic scalars accumulated in the process of dimensional reduction, these
form a noncompact nonlinear σmodel with a target manifold
G
/H
′
, where G is
the usual supergravity symmetry group shown in Table 2 for the corresponding
(reduced) dimension D but H
′
is a noncompact form of the modulus little
group H shown in Table 2. The diﬀerence between the groups H
′
and H arises
because dimensional reduction on the time coordinate introduces extra minus
signs, with respect to the usual spatialcoordinate KaluzaKlein reduction,
in “kinetic” terms for scalars descending from vector ﬁelds in the (D + 1)
dimensional theory including the time dimension. Scalars descending from
scalars or from the metric in (D + 1) dimensions do not acquire extra minus
signs. The change to the little group H
′
is also needed for the transformation
of ﬁeld strengths of higher rank, but these need not be considered for our
discussion of the BPS instantons. The relevant groups for the noncompact
σmodels in dimensions 9 ≥ D ≥ 3 are given in Table 5. These should be
compared to the standard CremmerJulia groups given in Table 2.
The sector of dimensionallyreduced supergravity that is relevant for the
instanton solutions consists just of the transversespace Euclideansignature
94
Table 5: Symmetries for BPS instanton σmodels.
D G H
′
9 GL(2, R) SO(1, 1)
8 SL(3, R) ×SL(2, R) SO(2, 1) ×SO(1, 1)
7 SL(5, R) SO(3, 2)
6 SO(5, 5) SO(5, C)
5 E
6(+6)
USP(4, 4)
4 E
7(+7)
SU
∗
(8)
3 E
8(+8)
SO
∗
(16)
metric and the
G
/H
′
σmodel, with an action
I
σ
=
d
D
y
√
g
R −
1
2
G
AB
(φ)∂
i
φ
A
∂
j
φ
B
g
ij
, (10.1)
where the φ
A
are σmodel ﬁelds taking values in the
G
/H
′
target space, G
AB
is the targetspace metric and g
ij
(y) is the Euclideansignature metric for the
σmodel domain space. The equations of motion following from (10.1) are
1
√
g
∇
i
(
√
gg
ij
G
AB
(φ)∂
j
φ
B
) = 0 (10.2a)
R
ij
=
1
2
G
AB
(φ)∂
i
φ
A
∂
j
φ
B
, (10.2b)
where ∇
i
is a covariant derivative; when acting on a targetspace vector V
A
, it
is given by
∇
i
V
A
= ∂
i
V
A
−Γ
D
AE
(G)∂
i
φ
E
V
D
, (10.3)
in which Γ
A
BC
(G) is the Christoﬀel connection for the targetspace metric G
AB
.
The action (10.1) and the ﬁeld equations (10.2) are covariant with respect to
generalcoordinate transformations on the σmodel target manifold
G
/H
′
. The
action (10.1) and the ﬁeld equations (10.2) are also covariant with respect to
general y
i
→ y
′i
coordinate transformations of the domain space. These two
types of general coordinate transformations are quite diﬀerent, however, in
that the domainspace transformations constitute a true gauge symmetry of
the dynamical system (10.1), while the σmodel targetspace transformations
generally change the metric G
AB
(φ
A
) and so correspond to an actual symmetry
of (10.1) only for the ﬁniteparameter group G of targetspace isometries.
As in our original search for pbrane solutions given in Section 2, it is
appropriate to adopt an ansatz in order to focus the search for solutions. In
95
the search for instanton solutions, the metric ansatz can take a particularly
simple form:
g
ij
= δ
ij
, (10.4)
in which the domainspace metric is assumed to be ﬂat. The σmodel equations
and domainspace gravity equations for the ﬂat metric (10.4) then become
∇
i
(G
AB
(φ)∂
i
φ
B
) = 0 (10.5a)
R
ij
=
1
2
G
AB
(φ)∂
i
φ
A
∂
j
φ
B
= 0 . (10.5b)
Now comes the key step
72
in ﬁnding instanton solutions to the specialized
equations (10.5): for singlecharge solutions, one supposes that the σmodel
ﬁelds φ
A
depend on the domainspace coordinates y
i
only through some inter
mediate scalar functions σ(y), i.e.
φ
A
(y) = φ
A
(σ(y)) . (10.6)
After making this assumption, the σmodel φ
A
equations (10.5a) become
∇
2
σ
dφ
A
dσ
+ (∂
i
σ)(∂
i
σ)
¸
d
2
φ
A
dσ
2
+ Γ
A
BC
(G)
dφ
B
dσ
dφ
C
dσ
= 0 , (10.7)
while the gravitational equation (10.5b) becomes the constraint
G
AB
(φ)
dφ
A
dσ
dφ
B
dσ
= 0 . (10.8)
An important class of solutions to (10.7) is obtained by taking
∇
2
σ = 0 (10.9a)
d
2
φ
A
dσ
2
+ Γ
A
BC
(G)
dφ
B
dσ
dφ
C
dσ
= 0 . (10.9b)
At this point, one can give a picture of the σmodel maps involved in the
system of equations (10.8, 10.9), noting that (10.9a) is just Laplace’s equation
and that (10.9b) is the geodesic equation on
G
/H
′
, while the constraint (10.8)
requires the tangent vector to a geodesic to be a null vector. The intermediate
function σ(y) is required by (10.9a) to be a harmonic function mapping from
the ﬂat (10.4) Euclidean domain space onto a null geodesic on the target space
G
/H
′
. Clearly, the harmonic map σ(y) should be identiﬁed with the harmonic
function H(y) that controls the singlecharge brane solutions (2.24). On the
geodesic in
G
/H
′
, on the other hand, σ plays the role of an aﬃne parameter.
The importance of the noncompact structure of the target space manifold
96
G
/H
′
, for the groups G and H
′
given in Table 5, now becomes clear: only
on such a noncompact manifold does one have nontrivial null geodesics as
required by the gravitational constraint (10.8). The σmodel solution (10.6)
oxidizes back up to one of the singlecharge brane solutions shown in Figure
7, and, conversely, any solution shown in Figure 7 may be reduced down to
a corresponding noncompact σmodel solution of this type. This sequence of
σmodel maps is sketched in Figure 8.
y
m
flat E I
D
σ(y)
∇ σ = 0
2
φ
(
σ
)
A
G
H'
/
(
n
u
l
l
)
Figure 8: Harmonic map from E
D
to a null geodesic in
G
/H
′
.
An extension
73,74
of this σmodel picture allows for solutions involving
multiple harmonic maps σ
a
(y). In that case, one deals not with a single
geodesic, but with a totally geodesic submanifold of
G
/H
′
, and, moreover,
the geodesics generated by any curve in the intermediate σ
a
parameter space
must be null. This is the σmodel construction that generates multicharge so
lutions, giving rise to intersectingbrane solutions of the types discussed in Sec
tion 7. As with the intersectingbrane solutions, however, there are important
compatibility conditions that must be satisﬁed in order for such multicharge
solutions to exist. We saw in subsection 7.2 that, in order for some portion
of the rigid supersymmetry to remain unbroken, the projectors constraining
the surviving supersymmetry parameter need to be consistent. In the σmodel
picture, a required condition is expressed in terms of the velocity vectors for
the null geodesics. If one adopts a matrix representation M for points in the
coset manifold
G
/H
′
, the σmodel equations for the matrix ﬁelds M(y
m
) are
simply written
∇
i
(M
−1
∂
i
M) = 0 . (10.10)
97
Points on the geodesic submanifold with aﬃne parameters σ
a
may be writ
ten
M = Aexp(
¸
a
B
a
σ
a
) , (10.11)
where the constant matrices B
a
give the velocities for the various geodesics
parametrized by the σ
a
, while an initial point on these geodesics is speciﬁed by
the constant matrix A. The compatibility condition between these velocities
is given by the doublecommutator condition
74
[[B
a
, B
b
], B
c
] = 0 . (10.12)
This condition allows one to rewrite (10.11) as
M = Aexp
−
1
2
¸
c>b
¸
b
[B
b
, B
c
]σ
b
σ
c
¸
a
exp(B
a
σ
a
) , (10.13)
where the ﬁrst factor commutes with the B
a
as a result of (10.12). The matrix
current then becomes
M
−1
∂
i
M =
¸
a
B
a
∂
i
σ
a
−
1
2
¸
c>b
¸
b
[B
b
, B
c
](σ
b
∂
i
σ
c
−σ
c
∂
i
σ
b
) , (10.14)
and this is then seen to be conserved provided the σ
a
satisfy ∇
2
σ
a
(y) = 0,
i.e. they are harmonic maps from the Euclidean overalltransverse space of
the y
m
into the geodesic submanifold (10.11). The constraint imposed by the
gravitational equation is
R
ij
=
1
4
¸
a,b
tr(B
a
B
b
)∂
i
σ
a
∂
j
σ
b
= 0 , (10.15)
which is satisﬁed provided the geodesics parametrized by the σ
a
are null and
orthogonal, i.e.
tr(B
a
B
b
) = 0 . (10.16)
The general set of stationary multicharge brane solutions is thus ob
tained in the σmodel construction by identifying the set of totally null, totally
geodesic submanifolds of
G
/H
′
such that the velocity vectors satisfy the com
patibility condition (10.12).
Aside from the elegance of the above σmodel picture of the equations gov
erning BPS brane solutions, these constructions make quite clear the places
where assumptions have been made that are more stringent than are really
necessary. One example of this is the assumption that the transversespace
98
geometry is ﬂat, Eq. (10.4). This is clearly more restrictive than is really
necessary; one could just as well have a more general Ricciﬂat domainspace
geometry, with a correspondingly covariantized constraint for the null geodesics
on the noncompact manifold
G
/H
′
. The use of more general Ricciﬂat trans
verse geometries is at the basis of “generalized pbrane” solutions that have
been considered in Refs.
77,78
11 Concluding remarks
In this review, we have discussed principally the structure of classical pbrane
solutions to supergravity theories. Some topics that deserve a fuller treatment
have only been touched upon here. In particular, we have covered only in
a cursory fashion the important topic of κinvariant pbrane worldvolume ac
tions. There are wider classes of worldvolume actions than we have considered,
involving excitations other than worldvolume spinors and scalars. For a fuller
treatment, the reader is referred to Refs
4,7
, or to the discussions of κsymmetric
actions for cases involving R–R sector antisymmetrictensor ﬁelds.
79
An important extension of this subject is the remarkable duality between
worldvolume dynamics and bulk dynamics known as AdS/CFT duality.
80
The
original and potentially the most physically relevant example of this occurs in
the theory of stacks of N D3 branes in type IIB supergravity. A single D3
brane has a worldvolume theory described by D = 4, N = 4 super Maxwell
theory, and example of a worldvolume multiplet involving more than spinors
and scalars. In type IIB string theory, the extra excitations coming from short
strings linking the leaves of the stack extend the corresponding U(1)
N
abelian
gauge symmetry to a full nonabelian U(N). The AdS/CFT duality expresses
the equivalence between the dynamics of these solitonic excitations on the
worldvolume and full type IIB string theory in the spacetime bulk.
Another aspect of the pbrane story, which we have only brieﬂy presented
here in Section 7, is the large family of intersecting branes. These include
78
also
intersections at angles other than 90
◦
, and can involve fractions of preserved
supersymmetry other than inverse powers of 2. For a fuller treatment of some
of these subjects, the reader is referred to Ref.
48,50
, and for the implications
of charge conservation in determining the allowed intersections to Refs.
21
Yet another aspect of this subject that we have not dwelt upon here is
the intrinsically stringtheoretic side, in which some of the BPS supergravity
solutions that we have discussed appear as Dirichlet surfaces on which open
strings can end; for this, we refer the reader to Ref.
9
Of course, the real fascination of this subject lies in its connection to the
emerging picture in string theory/quantum gravity, and in particular to the
99
rˆoles that BPS supergravity solutions play as states stable against the eﬀects of
quantum corrections. In this emerging picture, the duality symmetries that we
have discussed in Section 8 play an essential part, uniting the underlying type
IIA, IIB, E
8
×E
8
and SO(32) heterotic, and also the type I string theories into
one overall theory, which then also has a phase with D = 11 supergravity as
its ﬁeldtheory limit. The usefulness of classical supergravity considerations in
probing the structure of this emerging “M theory” is one of the major surprises
of the subject.
Acknowledgments
The author would like to acknowledge helpful conversations with Marcus Bre
mer, Bernard de Wit, Mike Duﬀ, Fran¸ cois Englert, Gary Gibbons, Hong L¨ u,
George Papadopoulos, Chris Pope, and Paul Townsend.
References
1. E. Cremmer, B. Julia and J. Scherk, Phys. Lett. B 76, 409 (1978).
2. D.N. Page, Phys. Rev. D 28, 2976 (1983).
3. J.W. van Holten and A. van Proeyen, “N = 1 Supersymmetry Algebras
in D = 2, D = 3, D = 4 mod8,” J. Phys. A15, 3763 (1982).
4. K.S. Stelle and P.K. Townsend, “Are 2branes better than 1?” in Proc.
CAP Summer Institute, Edmonton, Alberta, July 1987, KEK library
accession number 8801076.
5. P.S. Howe and K.S. Stelle, “The Ultraviolet Properties of
Supersymmetric Field Theories,” Int. J. Mod. Phys. A 4, 1871 (1989).
6. P.K. Townsend, “Three lectures on supersymmetry and extended
objects,” in Integrable Systems, Quantum Groups and Quantum Field
Theories (23
rd
GIFT Seminar on Theoretical Physics, Salamanca,
June, 1992), eds L.A. Ibort and M.A. Rodriguez (Kluwer, 1993).
7. M.J. Duﬀ, R.R. Khuri and J.X. Lu, “String solitons,” Physics Reports
259, 213 (1995), hepth/9412184;
8. M.J. Duﬀ, “Supermembranes,” hepth/9611203.
9. J. Polchinski, “Tasi Lectures on Dbranes,” hepth/9611050.
10. E.S. Fradkin and A.A. Tseytlin, Phys. Lett. B 158, 316 (1985); Nucl.
Phys. B 261, 1 (1985).
11. C. Callan, D. Friedan, E. Martinec and M. Perry, Nucl. Phys. B 262,
593 (1985).
12. A. Dabholkar, G. Gibbons, J.A. Harvey and F. Ruiz Ruiz,
“Superstrings and Solitons,” Nucl. Phys. B 340, 33 (1990).
100
13. H. L¨ u, C.N. Pope, E. Sezgin and K.S. Stelle, “Stainless Super
pbranes,” Nucl. Phys. B 456, 669 (1996), hepth/9508042.
14. I.C. Campbell and P.C. West, Nucl. Phys. B 243, 112 (1984);
F. Giani and M. Pernici, Phys. Rev. D 30, 325 (1984);
M. Huq and M.A. Namazie, Class. Quantum Grav. 2, 293 (1985); ibid.
2, 597 (1985).
15. M.B. Green and J.H. Schwarz, Phys. Lett. B 122, 143 (1983);
J.H. Schwarz and P.C. West, Phys. Lett. B 126, 301 (1983);
J.H. Schwarz, Nucl. Phys. B 226, 269 (1983);
P.S. Howe and P.C. West, Nucl. Phys. B 238, 181 (1984).
16. C.W. Misner, K.S. Thorne and J.A. Wheeler, Gravitation (W.H.
Freeman and Co., San Francisco, 1973), Box 14.5.
17. H. L¨ u, C.N. Pope, E. Sezgin and K.S. Stelle, “Dilatonic pbrane
solitons,” Nucl. Phys. B 371, 46 (1996), hepth/9511203.
18. M.J. Duﬀ and K.S. Stelle, “Multimembrane solutions of D = 11
Supergravity,” Phys. Lett. B 253, 113 (1991).
19. G.W. Gibbons and P.K. Townsend, Phys. Rev. Lett. 71, 3754 (1993),
hepth/9302049;
M.J. Duﬀ, G.W. Gibbons and P.K. Townsend, Phys. Lett. B 332, 321
(1994), hepth/9405124.
20. E. Bergshoeﬀ, E. Sezgin and P.K. Townsend, Phys. Lett. B 189, 75
(1987).
21. E. Witten, “Bound states of strings and pbranes,” Nucl. Phys. B 460,
335 (1996), hepth/9510135;
M.R. Douglas, “Branes within branes,” hepth/9512077;
P.K. Townsend, “Brane Surgery,”in Proc. European Res. Conf. on
Advanced Quantum Field Theory, La LondelesMaures, Sept. 1996,
hepth/9609217;
G. Papadopoulos, “Brane surgery with Thom classes,” JHEP 9905 020
(1999), hepth/9905073.
22. M.J. Duﬀ, P.S. Howe, T. Inami and K.S. Stelle, “Superstrings in D = 10
from Supermembranes in D = 11,” Phys. Lett. B 191, 70 (1987).
23. H. L¨ u, C.N. Pope and K.S. Stelle, “Vertical Versus Diagonal
Dimensional Reduction for pbranes,” Nucl. Phys. B 481, 313 (1996),
hepth/9605082.
24. S.W. Hawking and G.F.R. Ellis, The LargeScale Structure of
SpaceTime, (Cambridge University Press, 1973).
25. R. G¨ uven, “Black pbrane solutions of D = 11 supergravity theory,”
Phys. Lett. B 276, 49 (1992).
26. G.W. Gibbons, G.T. Horowitz and P.K. Townsend, Class. Quantum
101
Grav. 12, 297 (1995), hepth/9410073.
27. G. Horowitz and A. Strominger, Nucl. Phys. B 360, 197 (1991);
M.J. Duﬀ and J.X. Lu, Nucl. Phys. B 416, 301 (1994),
hepth/9306052.
28. M.J. Duﬀ, H. L¨ u and C.N. Pope, “The Black Branes of Mtheory,”
Phys. Lett. B 382, 73 (1996), hepth 9604052.
29. J.A. de Azcarraga, J.P. Gauntlett, J.M. Izquierdo and P.K. Townsend,
Phys. Rev. Lett. 63, 2443 (1989).
30. G.W. Gibbons and C.M. Hull, Phys. Lett. B 109, 190 (1982).
31. H. L¨ u and C.N. Pope, pbrane solitons in maximal supergravities, Nucl.
Phys. B 465, 127 (1996), hepth/9512012.
32. H. L¨ u and C.N. Pope, “An approach to the classiﬁcation of pbrane
solitons,” hepth/9601089.
33. P.S. Howe and R.W. Tucker, J. Phys. A 10, L155 (1977).
34. I. Bandos, D. Sorokin and and D.V. Volkov, “On the Generalized
Action Principle for Superstrings and Superbranes,” Phys. Lett. B 352,
269 (1995), hepth/9502141;
I. Bandos, P. Pasti, D. Sorokin, M. Tonin and D.V. Volkov,
“Superstrings and Supermembranes in the Doubly Supersymmetric
Geometrical Approach,” Nucl. Phys. B 446, 79 (1995),
hepth/9501113.
35. E. Bergshoeﬀ, E. Sezgin and P.K. Townsend Ann. Phys. 185, 330
(1988).
36. A. Achucarro, J.M. Evans, P.K. Townsend and D.L. Wiltshire, Phys.
Lett. B 198, 441 (1987).
37. R. Khuri, Nucl. Phys. B 387, 315 (1992), hepth/9205081;
J.P. Gauntlett, J.A. Harvey and J.T. Liu, Nucl. Phys. B 409, 363
(1993), hepth/9211056.
38. B.R. Greene, A. Shapere, C. Vafa and ST. Yau, Nucl. Phys. B 337, 1
(1990);
G.W. Gibbons, M.B. Green and M.J. Perry, Phys. Lett. B 370, 37
(1996), hepth/9511080.
39. P.M. Cowdall, H. L¨ u, C.N. Pope, K.S. Stelle and P.K. Townsend,
“Domain Walls in Massive Supergravities,” Nucl. Phys. B 486, 49
(1997), hepth/9608173.
40. J. Scherk and J.H. Schwarz, Phys. Lett. B 82, 60 (1979).
41. E. Bergshoeﬀ, M. de Roo, M.B. Green, G. Papadopoulos and P.K.
Townsend, “Duality of Type II 7branes and 8branes,” Nucl. Phys. B
470, 113 (1996), hepth/9601150.
42. N. Kaloper, R.R. Khuri and R.C. Myers, “On generalized axion
102
reductions,” Phys. Lett. B 428, 297 (1998), hepth/9803066.
43. H. L¨ u and C.N. Pope, “Domain walls from Mbranes,” Mod. Phys.
Lett. A 12, 1087 (1997), hepth/9611079;
I.V. Lavrinenko, H. L¨ u and C.N. Pope, “From topology to generalized
dimensional reduction,” Nucl. Phys. B 492, 278 (1997),
hepth/9611134.
44. M. Cvetic, “Extreme domain wall — black hole complementarity in
N = 1 supergravity with a general dilaton coupling,” Phys. Lett. B
341, 160 (1994).
45. M. Cvetic and H.H. Soleng, “Supergravity domain walls,” Physics
Reports 282, 159 (1997), hepth/9604090.
46. J. Rahmfeld, Phys. Lett. B 372, 198 (1996), hepth/9512089;
N. Khviengia, Z. Khviengia, H. L¨ u and C.N. Pope, “Intersecting
Mbranes and Bound States,” Phys. Lett. B 388, 21 (1996),
hepth/9605077;
M.J. Duﬀ and J. Rahmfeld, “Bound States of Black Holes and Other
pbranes,” Nucl. Phys. B 481, 332 (1996), hepth/9605085.
47. H.W. Brinkmann, Proc. Nat. Acad. Sci. 9, 1 (1923).
48. G. Papadopoulos and P.K. Townsend, “Intersecting Mbranes,” Phys.
Lett. B 380, 273 (1996), hepth/9603087;
A. Tseytlin, “Harmonic superpositions of Mbranes,” Nucl. Phys. B
475, 149 (1996);
I.R. Klebanov and A.A. Tseytlin, “Intersecting Mbranes as
fourdimensional black holes,” Nucl. Phys. B 475, 179 (1996);
K. Berndt, E. Bergshoeﬀ and B. Janssen, “Intersecting Dbranes in ten
dimensions and six dimensions,” Phys. Rev. D 55, 3785 (1997),
hepth/9604168;
J. Gauntlett, D. Kastor and J. Traschen, “Overlapping branes in
Mtheory,” Nucl. Phys. B 478, 544 (1996).
49. R. Sorkin, Phys. Rev. Lett. 51, 87 (1983);
D.J. Gross and M.J. Perry, Nucl. Phys. B 226, 29 (1983).
50. J.P. Gauntlett, “Intersecting Branes,” hepth/9705011.
51. A.A. Tseytlin, “ ‘No force’ condition and BPS combinations of pbranes
in eleven dimensions and ten dimensions,” Nucl. Phys. B 487, 141
(1997), hepth/9609212.
52. K. Shiraishi, Nucl. Phys. B 402, 399 (1993).
53. G.W. Gibbons, G. Papadopoulos and K.S. Stelle, “HKT and OKT
geometries on soliton black hole moduli spaces,” Nucl. Phys. B 508,
623 (1997), hepth/9706207.
54. R. Coles and G. Papadopoulos, “The geometry of onedimensional
103
supersymmetric nonlinear sigma models,” Class. Quantum Grav. 7,
427 (1990).
55. E. Cremmer and B. Julia, Nucl. Phys. B 159, 141 (1979).
56. J.M. Izquierdo, N.D. Lambert, G. Papadopoulos and P.K. Townsend,
“Dyonic Membranes,” Nucl. Phys. B 460, 560 (1996),
hepth/9508177.
57. R. Nepomechie, “Magnetic monopoles from antisymmetric tensor gauge
ﬁelds,” Phys. Rev. D 31, 1921 (1985).
58. C. Teitelboim, Phys. Lett. B 67, 63, 69 (1986).
59. M. Bremer, H. L¨ u, C.N. Pope and K.S. Stelle, “Dirac quantisation
conditions and KaluzaKlein reduction,” Nucl. Phys. B 529, 259
(1998), hepth/9710244.
60. S. Deser, A. Gomberoﬀ, M. Henneaux and C. Teitelboim, “Duality,
selfduality, sources and charge quantisation in Abelian Nform
theories,” Phys. Lett. B 400, 80 (1997), hepth/9702184.
61. H. L¨ u, C.N. Pope and K.S. Stelle, “Weyl Group Invariance and pbrane
Multiplets,” Nucl. Phys. B 476, 89 (1996), hepth/9602140.
62. C.M. Hull and P.K. Townsend, Nucl. Phys. B 438, 109 (1995),
hepth/9410167.
63. E. Witten, “String theory dynamics in various dimensions,” Nucl.
Phys. B 443, 85 (1995).
64. M.J. Duﬀ and J. Rahmfeld, “Bound States of Black Holes and Other
pbranes,” Nucl. Phys. B 481, 332 (1996), hepth/9605085.
65. J.H. Schwarz, “The power of Mtheory,” Phys. Lett. B 367, 97 (1996),
hepth/9510086.
66. T.H. Buscher, “A symmetry of the string background ﬁeld equations,”
Phys. Lett. B 194, 69 (1987);
A. Giveon, M. Porrati and E. Rabinovici, “Target space duality in
string theory,” Physics Reports 244, 77 (1994), hepth/9401139.
67. A. Sen, “Electric and magnetic duality in string theory,” Nucl. Phys. B
404, 109 (1993), hepth/9207053;
“SL(2, Z) duality and magnetically charged strings,” Int. J. Mod.
Phys. A 8, 5079 (1993), hepth/9302038.
68. E. Cremmer, H. L¨ u, C.N. Pope and K.S. Stelle, “Spectrumgenerating
symmetries for BPS Solitons,” Nucl. Phys. B 520, 132 (1998),
hepth/9707207.
69. J.H. Schwarz, “An SL(2, Z) multiplet of type IIB superstrings,” Phys.
Lett. B 360, 13 (1995); Erratum ibid. 364, 252 (1995),
hepth/9508143.
70. E. Bergshoeﬀ, C.M. Hull and T. Ortin, “Duality in the type II
104
superstring eﬀective action,” Nucl. Phys. B 451, 547 (1995),
hepth/9504081.
71. E. Cremmer, B. Julia, H. L¨ u and C.N. Pope, “Dualisation of dualities,
I,” Nucl. Phys. B 523, 73 (1998), hepth/9710119.
72. G. Neugebaur and D. Kramer, Ann. der Physik (Leipzig) 24, 62 (1969).
73. P. Breitenlohner, D. Maison and G. Gibbons, Comm. Math. Phys.
120, 253 (1988).
74. G. Clement and D. Gal’tsov, “Stationary BPS solutions to dilatonaxion
gravity,” Phys. Rev. D 54, 6136 (1996), hepth/9607043;
D.V. Gal’tsov and O.A. Rytchkov, “Generating branes via
sigmamodels,” Phys. Rev. D 58, 122001 (1998), hepth/9801160.
75. H. L¨ u and C.N. Pope, “Multiscalar pbrane solitons,” Int. J. Mod.
Phys. A 12, 437 (1997), hepth/9512153.
76. N. Khviengia, Z. Khviengia, H. L¨ u and C.N. Pope, “Intersecting
Mbranes and bound states,” Phys. Lett. B 388, 21 (1996),
hepth/9605077.
77. K. Becker and M. Becker, “Mtheory on eightmanifolds,” Nucl. Phys.
B 477, 155 (1996), hepth/9605053.
78. M. Berkooz, M.R. Douglas and R.G. Leigh, “Branes intersecting at
angles,” Nucl. Phys. B 480, 265 (1996), hepth/9606139;
J.P. Gauntlett, G.W. Gibbons, G. Papadopoulos and P.K. Townsend,
“HyperKahler manifolds and multiply intersecting branes,” Nucl. Phys.
B 500, 133 (1997), hepth/9702202.
79. P.S. Howe and E. Sezgin, “Superbranes,” Phys. Lett. B 390, 133
(1997), hepth/9607227;
P.S. Howe and E. Sezgin, “D = 11, p = 5,” Phys. Lett. B 394, 62
(1997), hepth/9611008;
M. Cederwall, A. von Gussich, B.E.W. Nilsson and A. Westerberg,
“The Dirichlet Super Three Brane in TenDimensional Type IIB
Supergravity,” Nucl. Phys. B 490, 163 (1997), hepth/9610148;
M. Aganagic, C. Popescu and J.H. Schwarz, “Dbrane actions with local
kappasymmetry,” Phys. Lett. B 393, 311 (1997), hepth/9610249;
M. Cederwall, A. von Gussich, B.E.W. Nilsson, P. Sundell and A.
Westerberg, “The dirichlet super pbranes in tendimensional Type IIA
and IIB supergravity,” Nucl. Phys. B 490, 179 (1997),
hepth/9611159;
E. Bergshoeﬀ and P.K. Townsend, “Super Dbranes,” Nucl. Phys. B
490, 145 (1997), hepth/9611173.
I. Bandos, D. Sorokin and M. Tonin, “Generalized Action Principle and
Superﬁeld Equations of Motion for d = 10 Dpbranes,” Nucl. Phys. B
105
497, 275 (1997), hepth/9701127.
80. I. Klebanov, “Tasi Lectures: Introduction to the AdSCFT
correspondence,” hepth/0009139.
106
5 The super pbrane worldvolume action 6 KaluzaKlein dimensional reduction 6.1 Multiple ﬁeldstrength solutions and the singlecharge truncation 6.2 Diagonal dimensional reduction of pbranes . . . . . . . . . . . 6.3 Multicenter solutions and vertical dimensional reduction . . . 6.4 The geometry of (D − 3)branes . . . . . . . . . . . . . . . . . 6.5 Beyond the (D − 3)brane barrier: ScherkSchwarz reduction and domain walls . . . . . . . . . . . . . . . . . . . . . . . . . .
32 39 41 46 46 50 52
7 Intersecting branes and scattering branes 60 7.1 Multiple component solutions . . . . . . . . . . . . . . . . . . . 60 7.2 Intersecting branes and the four elements in D = 11 . . . . . . 61 7.3 Brane probes, scattering branes and modulus σmodel geometry 66 8 Duality symmetries and charge quantisation 8.1 An example of duality symmetry: D = 8 supergravity . . 8.2 pform charge quantisation conditions . . . . . . . . . . . 8.3 Charge quantisation conditions and dimensional reduction 8.4 Counting pbranes . . . . . . . . . . . . . . . . . . . . . . 8.5 The charge lattice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 72 74 77 79 83 85 87 90 94 99
9 Local versus active dualities 9.1 The symmetries of type IIB supergravity . . . . . . . . . . . . . 9.2 Active duality symmetries . . . . . . . . . . . . . . . . . . . . . 10 Noncompact σmodels, null geodesics, and harmonic maps 11 Concluding remarks
2
1
Introduction
Let us begin from the bosonic sector of D = 11 supergravity,1 √ 2 1 1 I11 = d11 x −g(R − 48 F[4] ) + 6 F[4] ∧ F[4] ∧ A[3]
.
(1.1)
In addition to the metric, one has a 3form antisymmetrictensor gauge potential A[3] with a gauge transformation δA[3] = dΛ[2] and a ﬁeld strength F[4] = dA[3] . The third term in the Lagrangian is invariant under the A[3] gauge transformation only up to a total derivative, so the action (1.1) is invariant under gauge transformations that are continuously connected to the identity. This term is required, with the coeﬃcient given in (1.1), by the D = 11 local supersymmetry that is required of the theory when the gravitinodependent sector is included. The equation of motion for the A[3] gauge potential is
1 d ∗F[4] + 2 F[4] ∧ F[4] = 0 ;
(1.2)
this equation of motion gives rise to the conservation of an “electric” type charge 2 U=
∂M8 1 (∗F[4] + 2 A[3] ∧ F[4] ) ,
(1.3)
where the integral of the 7form integrand is over the boundary at inﬁnity of an arbitrary inﬁnite spacelike 8dimensional subspace of D = 11 spacetime. Another conserved charge relies on the Bianchi identity dF[4] = 0 for its conservation, V =
∂ M5
F[4] ,
(1.4)
where the surface integral is now taken over the boundary at inﬁnity of a spacelike 5dimensional subspace. Charges such as (1.3, 1.4) can occur on the righthand side of the supersymmetry algebra,a3
a Although
{Q, Q} = C(ΓA PA + ΓAB UAB + ΓABCDE VABCDE ) ,
(1.5)
formally reasonable, there is admittedly something strange about this algebra. For objects such as black holes, the total momentum terms on the righthand side have a welldeﬁned meaning, but for extended objects such as pbranes, the U and V terms on the righthand side have meaning only as intensive quantities taken per spatial unit worldvolume. This forces a similar intensive interpretation also for the momentum, requiring it to be considered as a momentum per spatial unit worldvolume. Clearly, a more careful treatment of this subject would recognize a corresponding divergence in the [Q, Q] anticommutator on the lefthand side of (1.5) in such cases. This would then require then an inﬁnite normalization factor for the algebra, whose removal requires the righthand side to be reinterpreted in an intensive (i.e. per spatial unit worldvolume) as opposed to an extensive way.
3
4 ). In Sections 2 and 3.5) also have a total of 528 independent components: 11 for the momentum PA . The simplest of these have the structure of p + 1dimensional Poincar´invariant e hyperplanes in the supergravity spacetime. 1. on which early enthusiasm for supergravity’s promise gave way to disenchantment when it became clear that local supersymmetry is not in fact suﬃcient to tame the notorious ultraviolet divergences that arise in perturbation theory.3.4). and this was in the hope that the resulting models might solve some of the outstanding diﬃculties of quantum gravity.5).5 4 .g. see Ref.4) above. the LHS of (1.4) are the boundaries of integration volumes M8 . PA is the energymomentum 11vector and UAB and VABCDE are 2form and 5form charges that we shall ﬁnd to be related to the charges U and V (1. One of these diﬃculties was the ultraviolet problem. Note that since the supercharge Q in D = 11 supergravity is a 32component Majorana spinor. we shall delve in some detail into the properties of these solutions. The symmetric spinor matrices CΓA . The fact that the integration volume does not ﬁll out a full spacelike hypersurface does not impede the conservation of the charges (1.3. Now the question arises as to the relation between the charges U and V in (1. A rough idea about the origin of the index structures on UAB and VABCDE may be guessed from the 2fold and 5fold ways that the corresponding 8 and 5 dimensional integration volumes may be embedded into a 10dimensional spacelike hypersurface.3. Let us recall at this point some features of the relationship between supergravity theory and string theory.4). CΓAB and CΓABCDE on the RHS of (1. and hence have been termed “pbranes” (see. supergravity theories won much admiration for their beautiful mathematical structure. unlike the more familiar situation for charges in ordinary electrodynamics.3. M5 that do not ﬁll out a whole 10dimensional spacelike hypersurface in spacetime. We shall see in Section 4 that this is too na¨ however: it masks an important ıve.3. this only requires that no electric or magnetic currents are present at the boundaries ∂M8 .where C is the charge conjugation matrix. One thing that immediately stands out is that the Gauss’ law integration surfaces in (1.5) has 528 components. 1.b Nonetheless. Before we can discuss such currents. 1. we shall need to consider in some detail the supergravity solutions that carry charges like (1. ∂ M5 . topological aspect of both the electric charge UAB and the magnetic charge VABCDE . 55 for the “electric” charge UAB and 462 for the “magnetic” charge VABCDE . e. 1. 1. Supergravity theories originally arose from the desire to include supersymmetry into the framework of gravitational models.4) and the 2form and 5form charges appearing in (1. which is due to the stringent b For a review of ultraviolet behavior in supergravity theories. Ref.
φ) that may be viewed as eﬀective equations of motion for these massless modes. Requiring cancellation of the anomalies in this symmetry at the quantum level gives diﬀerentialequation restrictions on the background ﬁelds (gM N . rule out coupling to “relativistic objects” such as black holes. φ): I= 1 4πα′ √ d2 z γ [γ ij ∂i xM ∂j xN gM N (x) +iǫij ∂i xM ∂j xN AM N (x) + α′ R(γ)φ(x)] . For the maximal supergravity theories. AM N .” It was only occasionally noticed in this early period that this impossibility of coupling to matter ﬁelds does not. φ). In superstring theories.7) where FM N P = ∂M AN P + ∂N AP M + ∂P AM N is the 3form ﬁeld strength for the AM N gauge potential.6) Every string theory contains a sector described by ﬁelds (gM N . there is simultaneously a great wealth of ﬁelds present and at the same time an impossibility of coupling any independent external ﬁeldtheoretic “matter. AM N . however. such as those descended from the D = 11 theory (1. These severely restrict the possible terms that can occur in the Lagrangian. The eﬀective action for the superstring theories that we shall consider 5 . (1. The realization that supergravity theories do not by themselves constitute acceptable starting points for a quantum theory of gravity came somewhat before the realization sunk in that string theory might instead be the soughtafter perturbative foundation for quantum gravity. consider the σmodel action 10 that describes a bosonic string moving in a background “condensate” of its own massless modes (gM N . AM N . (1. these are the only ﬁelds that couple directly to the string worldsheet. But the approaches of supergravity and of string theory are in fact strongly interrelated: supergravity theories arise as longwavelength eﬀectiveﬁeldtheory limits of string theories. To see how this happens.1). The σmodel action (1.11 This system of eﬀective equations may be summarized by the corresponding ﬁeldtheory eﬀective action Ieﬀ = √ dD x −ge−2φ (D − 26) − 3 α′ (R + 4∇2 φ − 4(∇φ)2 2 1 − 12 FM N P F M N P + O(α′ )2 . The (D − 26) term reﬂects the critical dimension for the bosonic string: ﬂat space is a solution of the above eﬀective theory only for D = 26.constraints of their symmetries. strings and membranes.6) is classically invariant under the worldsheet Weyl symmetry γij → Λ2 (z)γij . this sector is called the NeveuSchwarz/NeveuSchwarz (NS–NS) sector.
one obtains the Einstein frame action.11) the bosonic D = 11 action (1. (1. it takes the form. It is not written in the form generally preferred by relativists.1). The eﬀective action (1. . One then chooses λ so as to eliminate the e−2φ factor. We shall discuss in Section 6) the process of KaluzaKlein dimensional reduction that relates theories in diﬀerent dimensions of spacetime. despite the apparently negative sign of its kinetic term in I string . the connection ΓM N P is invariant. however. note that under xindependent rescalings. reﬂecting the diﬀerent critical dimension for superstrings. As one can see in (1. but with the substitution of (D −26) by (D −10). One may rewrite the eﬀective action in a diﬀerent frame by making a Weylrescaling ﬁeld redeﬁnition gM N → eλφ gM N . This carries over also to terms with φ undiﬀerentiated.9) where the indices are now raised and lowered with gM N . (1. I Einstein = d10 x 1 1 −g (e) R(g (e) )− 2 ∇M φ∇M φ− 12 e−φ FM N P F M N P . .12) 48 4 6 . 1. after an integration by parts. To understand how this Weyl rescaling works.7) is in the string frame.7) is written in the form directly obtained from string σmodel calculations. .10) (e) (e) (s) MNP 1 12 FM N P F . specializing now to D = 10. I string = d10 x −g (s) e−2φ R(g (s) ) + 4∇M φ∇M φ − gM N = e−φ/2 gM N . we note that upon specifying the KaluzaKlein ansatz expressing ds2 in terms of 11 ds2 . (1. which emerge from the eλφ Weyl transformation. the Weyl transformation is just what is needed to unmask the positiveenergy sign of the kinetic term for the φ ﬁeld. For the present. the KaluzaKlein vector AM and the dilaton φ.10). 9. 10 ds2 = e−φ/6 ds2 + e4φ/3 (dz + AM dxM )2 11 10 M = 0. . Ieﬀ as written in (1.1) reduces to the Einsteinframe type IIA bosonic action 14 Einstein IIIA = − d10 x −g (e) R(g (e) ) − 1 ∇M φ∇M φ − 2 1 −φ FM N P F M N P 12 e 1 φ/2 1 e FM N P Q F MN P Q − e3φ/2 FM N F M N + LF F A . which has a clean EinsteinHilbert term free from exponential prefactors like e−2φ .8) After making the transformation (1. Now let us return to the maximal supergravities descended from (1. (1. Terms with φ undiﬀerentiated do change.in this review contains a similar (NS–NS) sector.
where χ is a R–R zeroform (i. F[5] = ∗F[5] . Thus one naturally encounters ﬁeld strengths of ranks 1–5 in the supergravity theories deriving from superstring theories. to which we shall shortly turn. one will encounter transformations that have the eﬀect of ﬂipping the sign of the dilaton.15 one has F[1] = dχ. Comparing with the familiar g −2 couplingconstant factor for the YangMills aco tion. the study of classical supergravity will contain decidedly nonperturbative information about string theory. one sees that the asymptotic value eφ∞ plays the rˆle of the stringtheory coupling constant. e. In this review. One ﬁnds string IIIA = − d10 x −g (s) e−2φ R(g (s) ) + 4∇M φ∇M φ − MNP 1 12 FM N P F 1 1 FM N P Q F M N P Q − FM N F M N + LF F A . The top line in (1. In order to understand better the distinction between these two sectors. R NS F[3] = dAR . a second 3form ﬁeld strength making a pair together with F[3] [2] from the NS–NS sector.12) in string frame using (1. this will arise in the study of pbrane solitons. Since in classical supergravity theory. which is a selfdual 5form in D = 10. we shall mostly consider the descendants of the type IIA action (1. In addition. the bottom line corresponds the R–R sector (plus the ChernSimons terms. a pseudoscalar ﬁeld).13) Now one may see the distinguishing feature of the NS–NS sector as opposed to the R–R sector: the dilaton coupling is a uniform e−2φ in the NS–NS sector.g. and F[5] = dA[4] . o 7 .where FM N is the ﬁeld strength for the KaluzaKlein vector AM .12).e. In particular. These ﬁeld strengths will play an essential rˆle in supporting the pbrane solutions that we shall now describe. φ → −φ. which we have not shown explicitly). This leaves out one important case that we shall have to consider separately: the chiral type IIB theory in D = 10.12) corresponds to the NS–NS sector of the IIA theory. 48 4 (1. one may use ǫ[10] to dualize certain ﬁeld strengths.9). the original F[3] may be dualized to the 7form ∗F[7] . and it does not couple (in string frame) to the R–R sector ﬁeld strengths. In the type IIB theory. rewrite (1. The upshot is that antisymmetrictensor gauge ﬁeld strengths of diverse ranks need to be taken into account when searching for solutions to stringtheory eﬀective ﬁeld equations.
2c) (2.1) may be obtained by a consistent truncation from a full supergravity theory in D dimensions.1) We shall consider later in more detail how (2. we shall make a simplifying ansatz.1 The pbrane ansatz Singlecharge action and ﬁeld equations We have seen that one needs to consider eﬀective theories containing gravity.2d) In order to solve the above equations. 2n! 2. The value of the important parameter a controlling the interaction of the scalar ﬁeld φ with the ﬁeld strength F[n] in (2.1) with a single scalar φ and a single ﬁeld strength F[n] will be consistent except for certain special cases when n = D/2 that we shall have to consider separately. A consistent truncation is one for which solutions of the truncated theory are also perfectly good.2a) (2. we shall make a consistent truncation of the action down to a simple system in D dimensions comprising the metric gM N .1). the whole is described by the action I= √ dD x −g R − 1 ∇M φ∇M φ − 2 1 aφ 2 2n! e F[n] .1) we have not included contributions coming from the F F A ChernSimons term in the action. (2.2 Electric and magnetic ans¨tze a + SM N n−1 1 eaφ (FM ··· FN ··· − F 2 gM N ) 2(n − 1)! n(D − 2) (2. 8 . Note that in (2. We shall be looking for solutions preserving certain unbroken supersymmetries.2b) (2. various ranks of antisymmetrictensor ﬁeld strengths and various scalars. These are also consistently excluded in the truncation to the singlecharge action (2. a scalar ﬁeld φ and a single (n − 1)form gauge potential A[n−1] with corresponding ﬁeld strength F[n] . and in such cases it will generally be necessary to retain an axionic scalar χ as well. To obtain a more tractable system to study.2 2. one can have dyonic solutions. solutions of the original untruncated theory.1) produces the following set of equations of motion: RMN SMN = = 1 2 ∂M φ∂N φ ∇M 1 (eaφ F M 1 ···M n ) = 0 a aφ 2 φ = e F . The notion of a consistent truncation will play a central rˆle in our discussion of the BPS o solutions of supergravity theories.1) will vary according to the cases considered in the following. Varying the action (2. albeit speciﬁc. In such cases. Truncation down to the system (2.
p m = p + 1. . This is what we shall call the “elementary. For the antisymmetric tensor gauge ﬁeld. For simplicity. others zero. Just as the Maxwell 1form naturally couples to the worldline of a charged particle. the ﬁrst possibility for A[n−1] is to support a del = n − 1 dimensional worldvolume. we face a bifurcation of possibilities for the ansatz. .4) SO(D − d) isotropicity and (Poincar´)d symmetry are guaranteed here because e the function C(r) depends only on the transverse radial coordinate r. we make an ansatz requiring (Poincar´)d × SO(D − d) symmetry. . The corresponding ansatz for the scalar ﬁeld φ(xM ) is simply φ = φ(r). For this basic class of solutions. translational invariance in the worldvolume directions xµ and SO(D − d) symmetry in the transverse directions y m is guaranteed. One may view the soughtfor e solutions as ﬂat d = p + 1 dimensional hyperplanes embedded in the ambient Ddimensional spacetime. . 1. these hyperplanes may in turn be viewed as the histories. The ﬁrst possibility is naturally expressed directly in terms of the gauge potential A[n−1] . Thus.4).and these will in turn require unbroken translational symmetries as well. the two possibilities being related by duality. expressed in terms of A[n−1] . D − 1 . we could equivalently have 9 . D − 1) are the coordinates “transverse” to the worldvolume. · · · . Accordingly. y m ). . Instead of the ansatz (2. (2. (2. Since the metric components depend only on r. p = d − 1) are coordinates adapted to the (Poincar´)d isometries e on the worldvolume and where y m (m = d. of pdimensional spatial surfaces. as we shall see later. so does A[n−1] naturally couple to the worldvolume of a p = d − 1 = (n − 1) − 1 dimensional “charged” extended object. The “charge” here will be obtained from Gauss’law surface integrals involving F[n] . These restrictions can subsequently be relaxed in generalizations of the basic class of pbrane solutions that we shall discuss here. An ansatz for the spacetime metric that respects the (Poincar´)d ×SO(D− e d) symmetry is 12 ds2 = e2A(r) dxµ dxν ηµν + e2B(r) dy m dy n δmn µ = 0. where xµ (µ = 0. 1.3) √ where r = y m y m is the isotropic radial coordinate in the transverse space. . or worldvolumes. let the spacetime coordinates be split into two ranges: xM = (xµ . . · · · .” or “electric” ansatz: Aµ1 ···µn−1 = ǫµ1 ···µn−1 eC(r) . we shall also require isotropic symmetry in the directions “transverse” to the translationallysymmetric ones. .
2. upon 10 . c Speciﬁcally.4. this would naturally couple to a dso = D − n − 1 dimensional worldvolume.6) where the magneticcharge parameter λ is a constant of integration.. The power of r in the solitonic/magnetic ansatz is determined by requiring F[n] to satisfy the Bianchi identity. Since such a dualized potential would be nonlocally related to the ﬁelds appearing in the action (2.5) The worldvolume dimension for the elementary ansatz (2. If one were to ﬁnd an underlying gauge potential for ∗F (locally possible by courtesy of a Bianchi identity). orthonormal frames. thus obtaining cancellation. (d) = d.5) is clearly del = n − 1. others zero. the only thing left undetermined by this ansatz. one ﬁnds ∂q Fm1 ···mn = r −(n+1) ǫm1 ···mn q − (n + 1)ǫm1 ···mn p y p yq /r 2 .1). (2. we shall not explicitly follow this construction. the factor of (n + 1) is evened m 2 out between the two terms and then one ﬁnds from cycling a factor m y ym = r . The second possible way to relate the rank n of F[n] to the worldvolume dimension d of an extended object is suggested by considering the dualized ﬁeld strength ∗F .13 This is most easily done by introducing vielbeins.16 with tangentspace indices denoted by underlined indices: gM N = eM E eN F ηE F .c Note that the worldvolume dimensions of the elementary and solitonic cases ˜ are related by dso = del ≡ D −del −2. This “solitonic” or “magnetic” ansatz for the antisymmetric tensor ﬁeld is most conveniently expressed in terms of the ﬁeld strength F[n] .3 Curvature components and pbrane equations In order to write out the ﬁeld equations after insertion of the above ans¨tze. (2. i. a one needs to compute the Ricci tensor for the metric.given just the F[n] ﬁeld strength: Fmµ1 ···µn−1 = ǫµ1 ···µn−1 ∂m eC(r) . which now has nonvanishing values only for indices corresponding to the transverse directions: Fm1 ···mn = λǫm1 ···mn p (mag) yp rn+1 . note also that this relation is idempotent. ˜ i. which is a (D − n) form. (2.e. 2.e. (el) others zero.7) taking the totally antisymmetrized combination [qm1 · · · mn ]. but shall instead take this reference to the dualized theory as an easy way to identify the worldvolume dimension for the second type of ansatz.
ω µ n = e−B(r) ∂n A(r)eµ (2.9) = e−B(r) ∂n B(r)em − e−B(r) ∂m B(r)en . d Substituting the above relations. where again. 11 (2.8) The corresponding spin connection 1forms are determined by the condition that the torsion vanishes. ς = −1. and the primes indicate ∂/∂r derivatives. ς = +1 magnetic: d = D − n − 1. (2.12) where ς = ±1 for the elementary/solitonic cases and the source appearing on the RHS of these equations is S= (e 2 aφ−dA+C )C ′ 1 ˜ ˜ λ(e 2 aφ−dB )r−d−1 1 ˜ ˜ φ′′ +dA′ φ′ + dB ′ φ′ + (d+1) φ′ r 1 − d B ′ − d A′ + 2 (φ′ )2 = r r = − 1 ςaS 2 2 electric: d = n − 1. m). which yields ωµ ν ωm n = 0.13) . Splitting up the tangentspace indices E = (µ.Next. em = eB(r) dy m . we have for our ans¨tze the vielbein 1forms a eµ = eA(r) dxµ . m) similarly to the world indices M = (µ. r r r ˜ = D − d − 2.11) Rmn = −δmn (B ′′ + dA′ B ′ + d(B ′ )2 + r r ˜ d d ymyn ˜ ˜ − 2 (dB ′′ + dA′′ − 2dA′ B ′ + d(A′ )2 − d(B ′ )2 − B ′ − A′ ) . one ﬁnds the set of equations that we need to solve to obtain the metric and φ: Rµν = ˜ −ηµν e2(A−B) (A′′ + d(A′ )2 + dA′ B ′ + ˜ A′′ +d(A′ )2 + dA′ B ′ + (d+1) A′ = r dB +dA −2dA B +d(A ) − d(B ) ˜ ˜ ˜ B ′′ +dA′ B ′ + d(B ′ )2 + (2d+1) B ′ + d A′ r r ′′ ′ ′ ′ 2 ˜ ′′ ˜ ′ 2 ˜ ˜ d 2 2(D−2) S d = − 2(D−2) S 2 1 2 2S {µν} {δmn } {ym yn } {φ} (2. deE + ω E F ∧ eF = 0. EF R[2] = dω E F + ω E D ∧ ωD F . one constructs the corresponding 1forms: eE = dxM eM E . The curvature 2forms are then given by (2.10) From the curvature components so obtained. one ﬁnds the Ricci tensor components ˜ (d + 1) ′ A) r ˜ (2d + 1) ′ d ′ ˜ B + A) (2.
14) We shall see later that more general solutions of the Laplace equation than the simple isotropic ones considered here will also play important rˆles in the o story.16a) S2 (2.4 pbrane solutions The pbrane equations (2. Accordingly. let us ﬁrst note a generalization.12).b) suggest that we now further reﬁne the ans¨tze by a imposing another linearity condition: φ′ = −ςa(D − 2) ′ A . (2. which shall be justiﬁed in more detail later on in Section 4. it is useful to introduce a new piece of notation. we shall reﬁne the pbrane ansatz (2. ˜ d ˜ 2dd . equation (2.5. 2.19) . which for isotropic scalar functions of r is ˜ ∇2 φ = φ′′ + (d + 1)r−1 φ′ .e. a2 12 (2. one may recognize more general possibilities by noting the form of the Laplace operator.17) At this stage. the independent equations become 17 ∇2 φ = 1 − 2 ςaS 2 ˜ d (2. In order to reduce the complexity of Eqs (2.16c) ∇2 A = 1 ˜ d(D − 2)(A′ )2 + 2 d(φ′ )2 = 2(D − 2) 1 ˜ 2 dS . the Laplacian is ∇2 φ = φ′′ + (d + 1)r−1 φ′ .16c) gives S2 = ∆(φ′ )2 . Before we embark on solving these equations. 2.6) by looking ahead a bit and taking a hint from the requirements for supersymmetry preservation. (2.12) have been speciﬁcally written for an isotropic pbrane ansatz.13) are still rather daunting.18) With this notation. we shall look for solutions satisfying the linearity condition ˜ dA′ + dB ′ = 0 .3. isotropic) functions in the transverse ˜ (D − d) dimensions.15) After eliminating B using (2. for sphericallysymmetric (i. 2. Equations (2.2. (D − 2) (2. letting a2 = ∆ − (2.16b) (2.12.16a. 2 where. Although Eqs (2.15).
which can (2.22) is consistent with the equation of motion for F[n] : ˜ (2. one ﬁnds the relation √ 1 − ∆ − 2 aφ+dA ′ ∂ C φ (2.20) Solving this in the transverse (D − d) dimensions with our assumption of transverse isotropicity (i. one ﬁnds 7.20) can also be more generally derived.d ∇2 e 2a φ = 0 . Assembling the result. ς∆ ς∆ ′ 2 2a (φ ) = 0. 13 .24c) ς= .13) that S 2 = eaφ−2dA (C ′ eC )2 . it has been taken to be positive in order to ensure the absence of naked singularities at ﬁnite r. with ς = +1. spherical symmetry) yields e 2a φ ≡ H(y) = 1 + ς∆ k ˜ rd k>0.24b) (2. we now pick values of the integration constants to make A∞ = B∞ = 0.21).e. Combining this with (2. In order to simplify the explicit form of the solution.13 ds2 eφ = = H ∆(D−2) dxµ dxν ηµν + H ∆(D−2) dy m dy m H ς∆ 1+ k ˜ rd 2a ˜ −4d 4d (2. +1. it still remains to ﬁnd the function C(r) that determines the antisymmetrictensor gauge ﬁeld potential. it is straightforward to verify that the relation (2. In this case. (2.21) sets the mass scale of the solution.24a) (2. it follows from (2. so that the solution tends to ﬂat empty space at transverse inﬁnity. Finally. it still holds if one relaxes the assumption of isotropicity in the transverse space. solitonic/magnetic H(y) = d Note that Eq.23) ∇2 C + C ′ (C ′ + dB ′ − dA′ + aφ′ ) = 0 .19).21) where the constant of integration φ has been set equal to zero here for r→∞ simplicity: φ∞ = 0.22) (e ) = e ∂r a (where it should be remembered that a < 0). starting from the Laplaceequation solution H(y) (2. for example. In the case of the elementary/electric ansatz. The integration constant k in (2. (2.so that the remaining equation for φ becomes ∇2 φ + be reexpressed as a Laplace equation. This positivity restriction is similar to the usual restriction to a positive mass parameter M in the standard Schwarzschild solution. elementary/electric −1.
. 3 D = 11 examples Let us now return to the bosonic sector of D = 11 supergravity.. . e. . there are two particular points to note.. the F[4] antisymmetric tensor ﬁeld strength supports either an elementary/electric solution with d = n − 1 = 3 (i..27a) ∆ 2 ( − √ ǫm1 .21) determines all of the features of a pbrane solution (except for the choice of gauge for the A[n−1] gauge potential). a p = 2 membrane) or a solitonic/magnetic solution 14 .1).2. so that the scalar may be consistently truncated from our general action (2. The harmonic function H(y) (2.mn = = 2 √ ǫµ1 . The ﬁrst is that no scalar ﬁeld is present in (1. ∆ (2.. D − 1 magnetic.. In lower dimensions. . .e. scalars do appear. The absence of the scalar that we had in our general discussion may be handled here simply by identifying the scalar coupling parameter a with zero. D − 1 electric (2. . (2. This follows from the supermultiplet structure of the D = 11 theory.g.6) by √ ∆ k= λ.25) In the solitonic/magnetic case. Note that for n = 4.1). of course. the constant of integration is related to the magnetic charge parameter λ in the ansatz (2. Since ˜ a2 = ∆ − 2dd/(D − 2). Now let us consider the consistency of dropping contributions arising from the F F A ChernSimons term in (1.µn−1 Fm1 . we identify ∆ = 2 · 3 · 6/9 = 4 for the D = 11 cases. It is useful to express the electric and magnetic ﬁeld strengths directly in terms of H: Fmµ1 .. which has the action (1.and in the elementary/electric case. In searching for pbrane solutions to this action.mn r ∂r H m = d. this relation may be taken to deﬁne the parameter λ.. C(r) is given by 2 eC = √ H −1 .26) ˜ 2d In the elementary/electric case. ..27b) ∆ with all other independent components vanishing in either case. . in which all ﬁelds are gauge ﬁelds.1).1). the dilaton in D = 10 type IIA supergravity emerges out of the D = 11 metric upon dimensional reduction from D = 11 to D = 10.µn−1 ∂m (H −1 ) m = d.
Moreover. However. a further check is a necessary. and so one should look for an analytic extension of it. Subsequently.19 Thus.1) vanishes and hence this term does not make any nonvanishing contribution to the metric ﬁeld equations for our ans¨tze. For the antisymmetric tensor ﬁeld equation. one may consider a change to “Schwarzschildtype” coordinates by setting r = (˜6 − k) 1/6 .˜ with d = 11 − 3 − 2 = 6 (i. subsequently referred to an orthonormal frame by introducing vielbeins as in (2.e. (3. one ﬁnds these invariants to be nonsingular. we shall explore how these particular solutions ﬁt into wider.8). 3. one may suspect that the metric as given in (3. the F F A term in the action (1. ǫµνλ (1 + r6 )−1 . we shall consider the elementary/electric and the solitonic/magnetic D = 11 cases in detail. which when written out explicitly becomes ∂M √ −gF M U V W + 1 ǫU V W x1 x2 x3 x4 y1 y2 y3 y4 Fx1 x2 x3 x4 Fy1 y2 y3 y4 = 0 . geodesic diverges. the surface r = 0 can be reached along null geodesics in ﬁnite aﬃne parameter. a p = 5 brane). In both these elementary and solitonic cases. ˜ + dσ 2 + dρ2 ) + (1 − k −2 2 d˜ r r6 ) ˜ + r2 dΩ2 ˜ 7 (3. The ﬁeld equation for A[3] is (1. electric 2brane: Schwarzschildtype coordinates 15 .2).2) does not in fact cover the entire spacetime. this solution looks like it might be singular at r = 0. The solution then becomes:19 r Aµνλ = ds2 = (1 − k 2/ 2 3 r 6 ) (−dt ˜ k ǫµνλ (1 − r6 ) . if one calculates the invariant components of the curvature tensor RM N P Q and of the ﬁeld strength Fmµ1 µ2 µ3 . a Next.3) other components zero. Accordingly. (3. we have the elementaryansatz solution 18 ds2 = (1 + Aµνλ = k k − 2/ 3 dxµ dxν ηµν + (1 + r6 ) 1/3 dy m dy m r6 ) k other components zero.1) 2(4!)2 By direct inspection. 2.2) electric 2brane: isotropic coordinates At ﬁrst glance. since there one requires the variation of the F F A term to vanish in order to consistently ignore it. one sees that the second term in this equation vanishes for both ans¨tze. “black.” families of pbranes. although the proper distance to the surface r = 0 along a t = x0 = const.1 D = 11 Elementary/electric 2brane From our general discussion in Sec.
the membrane solution interpolates between ﬂat space as R → 1 and (AdS)4 × S 7 as R → 0 at the horizon. may then be recognized as a standard form of the metric on (AdS)4 × S 7 . however. As one approaches the horizon at R = 0. however. Continuing on inside the horizon. The timelike nature of this singularity. so that light ˜ cones do not “ﬂip over” inside the horizon. the solution becomes 19 ds2 = 1 R2 (−dt2 + dσ 2 + dρ2 ) + 4 k 1/3 R−2 dR2 + k 1/3 dΩ2 7 1 1/ 3 − 7/ −2 3 3 − 1]R dR2 + k 1/3 [(1 − R3 )− 1/3 − 1]dΩ2 + 4 k [(1 − R ) 7 (a) (b) Aµνλ = R3 ǫµνλ . The residual metric. given in line (a). The Schwarzschildlike coordinates make the surface r = k 1/6 (correspond˜ ing to r = 0) look like a horizon. the singularity in the membrane spacetime is timelike. generalizing the RobinsonBertotti solution on (AdS)2 × S 2 in D = 4.where we have supplied explicit worldvolume coordinates xµ = (t. Unlike the singularity in the classic Schwarzschild ˜ solution. ρ) and where dΩ2 is the line element on the unit 7sphere. however. ˜ the transformation from the original isotropic coordinates to these new ones is eﬀected by setting r = k 1/6 R 1/2 /(1 − R3 ) 1/6 .4) vanishes at least linearly in R. conﬁrming that r = k 1/6 is in fact a horizon. is that it reveals how the solution interpolates between other “vacuum” solutions of D = 11 supergravity. 16 . so the spacetime is in fact genuinely singular. Radial null geodesics do intersect the singularity at ﬁnite aﬃne parameter. One may indeed verify that the normal to this surface is a null vector. with R < 0 being the interior). other components zero.4) of the membrane solution. let us change coordinates once again. Overall. corresponding to the bound7 ary ∂M8T of the 11 − 3 = 8 dimensional transverse space. In order to see the structure of the membrane spacetime more clearly. ˜ This horizon is degenerate. which is spacelike and hence unavoidable. The main usefulness of the third form (3. line (b) of the metric in (3. one eventually encounters a true singularity at r = 0 (R → −∞). electric 2brane: interpolating coordinates (3. geodesics do not intersect the singularity at a ﬁnite value of an aﬃne parameter value. σ. the solution becomes ﬂat. curves along the t axis for r < k 1/6 remain timelike.4) This form of the solution makes it clearer that the lightcones do not “ﬂip over” in the region inside the horizon (which is now at R = 0. In these new coordinates.19 As R → 1. however. Generically. Thus. unlike the situation for the classic Schwarzschild solution. Owing to the 2/3 exponent in the g00 component. in the asymptotic exterior transverse region. setting r = k 1/6 (1 − R3 )− 1/6 .
is the unique “matter” system that can consistently couple to D = 11 supergravity. the solid angle subtended by the boundary at transverse inﬁnity. the region mapped by the isotropic coordinates does not cover the whole spacetime. The global structure of the membrane spacetime 19 is similar to the extreme ReissnerNordstrom solution of General Relativity.20. and not as a part of the central singularity. ˜ Indeed. shaded in the diagram. These boundary surfaces are not singular. This interpolating portion of the spacetime. The “throat” P in the diagram should be thought of as an exceptional point at inﬁnity. since one may reach its boundaries H+ . This region.24 This global structure is summarized by a CarterPenrose diagram as shown in Figure 1.3) of the solution that the normals to these surfaces are null). is geodesically incomplete. (3.e. in which the angular coordinates on S 7 and also two ignorable worldsheet coordinates have been suppressed.5) 3Ω7 √ where 1/(2κ2 ) is the coeﬃcient of −gR in the EinsteinHilbert Lagrangian and Ω7 is the volume of the unit 7sphere S 7 . instead. but. corresponding to the shaded region of Figure 1 which is covered by the isotropic coordinates. H− along radial null geodesics at a ﬁnite aﬃneparameter value. As one can see. The region exterior to the horizon interpolates between ﬂat regions J ± at future and past null inﬁnities and a geometry that asymptotically tends to (AdS)4 × S 7 on the horizon. 17 . may be sketched as shown in Figure 2. i. constitute future and past horizons (one can see from the form (3.22 Analysis of this coupling yields a relation between the parameter k in the solution (3.20 which generalizes the NambuGoto action for the string. the D = 11 supermembrane action.invites one to consider coupling a δfunction source to the solution at r = 0.2) and the tension T of the supermembrane action:18 κ2 T k= .
hypersurface + J + H “throat” P i0 spatial infinity J – Figure 1: CarterPenrose diagram for the D = 11 elementary/electric 2brane solution. hypersurface Only the shaded region is covered by the isotropic coordinates 18 . H – R = const.timelike singularity ~ at r = 0 (R → − ∞) t = const.
6) As in the case of the elementary/electric membrane. 5 ds2 = (1+ r3 )− 1/3 dxµ dxν ηµν +(1+ r3 ) 2/3 dy m dy m yp other components zero. · · · . this solution interpolates between two “vacua” of D = 11 supergravity. 19 . ν = 0. Fm1 ···m4 = 3kǫm1 ···m4 p r5 magnetic 5brane: isotropic coordinates (3. 3. however. Now. this solution is a magnetic 5brane:25 k k µ. In isotropic coordinates.2 D = 11 Solitonic/magnetic 5brane Now consider the 5brane solution to the D = 11 theory given by the solitonic ansatz for F[4] . Combining two coordinate changes analogous to those of the elementary case. these asymptotic geometries consist of the ﬂat region encountered as r → ∞ and of (AdS)7 ×S 4 as one approaches r = 0. which once again is a degenerate horizon.flat M 11 infinite “throat:” 7 (AdS)4 × S Figure 2: The D = 11 elementary/electric 2brane solution interpolates between ﬂat space at J ± and (AdS)4 × S 7 at the horizon.
and the region covered by the isotropic coordinates may once again be sketched as in Figure 2. 20 . Unlike the case of doubled ﬂat space. magnetic 5brane: interpolating coordinates dΩ2 (3.8) to be a nonsingular degenerate horizon. (1 − R6 ) 1/3 −2 (3. one has an overall transformation r ˜ r= k 1/3 R2 . the R ↔ −R identiﬁcation in the 5brane geometry introduces a conical singularity. however. but now the magnetic solution (3. However. In this case. where the identiﬁcation removes a conical singularity with deﬁcit angle 2π.r = (˜3 − k) 1/3 and r = k 1/3 (1 − R6 )− 1/3 .” at the horizon R = 0.26 Given this isometry R → −R. the surface r = 0 ↔ R = 0 may be seen from (3.8) Once again.7) After these coordinate changes. This identiﬁcation is analogous to the identiﬁcation one naturally makes for ﬂat space when written in polar coordinates. with the metric ds2 = −dt2 + dr2 + r2 d2 . if one wishes identify the spacetime region R ≤ 0 with the region R ≥ 0.4). The CarterPenrose diagram for the solitonic/magnetic 5brane solution is given in Figure 3. not only do the light cones maintain their timelike orientation when crossing the horizon. the metric becomes 4R ds2 = R2 dxµ dxν ηµν + k 2/3 (1−R6 ) 8/3 dR2 + (1−R64) 2/3 . except now with the asymptotic geometry in the “throat” region being (AdS)7 × S 4 instead of (AdS)4 × S 7 as in the case of the elementary/electric solution. the nonsingular spacetime is the R ↔ −R symmetric but nonidentiﬁed spacetime. This smoothly continued spacetime has an inﬁnite “throat. as already happened in the electric case (3. one must ﬂat be attentive to the issue of conical singularities in this case. one can. where the full diagram extends indeﬁnitely by “tiling” the section shown.8) is in fact fully symmetric 26 under a discrete isometry R → −R.
1 ) R = = (R 0 i0' J ' + symmetric with I R<0 J + 0) (R (R = – + H spatial infinity 1 ) = (R J "+ = 1) i0" symmetric with I R<0 Figure 3: CarterPenrose diagram for the solitonic/magnetic 5brane solution. 21 R = 0 J – (R J J ' = 1) (R = 1 ) – (R " I R>0 i0 H – (R = 0) spatial infinity = ) 1 .
26 When d is odd in such cases. then the solution has a singularity at the horizon. Written in Schwarzschildtype coordinates. in which cases the scalar ﬁelds may consistently be set to zero. there exists a discrete isometry analogous to the R → −R isometry of the D = 11 5brane solution (3.2). (6.8). ˜ 22 . 3). (10. The characteristic feature of the above “blackened” pbranes is that they have a nondegenerate. ˜ When a = 0 and the scalar ﬁeld associated to the ﬁeld strength supporting a solution cannot be consistently set to zero. In these special cases.21) itself (where we recall that in isotropic coordinates.3 Black branes In order to understand better the family of supergravity solutions that we have been discussing. Similar solutions occur in other situations where the parameter a (2. let us now consider a generalization that lifts the degenerate nature of the horizon. this happens for (D. ˜ coincides in general with a curvature singularity. one encounters an inner horizon. d) = (11. (5. the horizon occurs at r = 0) 3.5). When d is even for “scalarless” solutions of this type.1) does not even contain a scalar ﬁeld. the solutions are nonsingular at the horizon and so one may analytically continue through to the other side of the horizon. The singular nature of the solution at r = r− is apparent in the scalar φ in (3.4).The electric and magnetic D = 11 solutions discussed here and in the previous subsection are “nondilatonic” in that they do not involve a scalar ﬁeld. the analyticallyextended metric eventually reaches a timelike curvature singularity at r = 0.9).28 ds2 = − Σ+ Σ− ˜ 4d {1− ∆(D−2) } 2a2 −1 ˜ ∆(D−2) dt2 + Σ− dxi dxi 2a2 ˜ 4d e ς∆ φ 2a = Σ ∆d + − Σ+ Σ−1 − d˜2 + r2 Σ− d dΩ2 r ˜ ∆ ˜ D−d−1 Σ± = 1 − ˜ r± d r ˜ (3. black brane: Schwarzschildtype coordinates The antisymmetric tensor ﬁeld strength for this solution corresponds to a √ ˜ ˜ charge parameter λ = 2d/ ∆(r+ r− )d/2 . which. since the bosonic sector of D = 11 supergravity (1. however. nonsingular outer horizon at r = r+ .1). (11. allowing the outer and inner regions to be identiﬁed.18) for a ﬁeld strength supporting a pbrane solution vanishes.2) and (4. as can be seen directly in the scalar solution (2. (5. one ﬁnds the generalized “black brane” solution 27.9) . at which the light ˜ cones “ﬂip over. either electric or magnetic.1). For solutions with p ≥ 1.” At r = r− .
Masses and Supersymmetry The pbrane solutions that we have been studying are supported by antisymmetric tensor gauge ﬁeld strengths that fall oﬀ at transverse inﬁnity like ˜ r−(d+1) . whether or not this gives rise to a physical singularity in a solution depends on the global structure of that solution. In the electric 2brane case. 4. When a = 0 and scalars may consistently be set to zero.25. and the coincident “ﬂips” of the light cones cancel out. or with the degenerate horizon in the extremal case. As we have seen in Sections 3. In this case. The generally singular nature of the inner horizon of the nonextreme solution (3.4). In this Section. the singularity at the horizon r+ = r− disappears and then one may analytically continue through the horizon.the singularity at the inner horizon persists even in cases where the scalar φ is absent. 4 Charges. 2.2. the energy density and the preservation of unbroken supersymmetry.6). the solution does in the end have a singularity. the light cones do not “ﬂip over” at the horizon because one is really crossing two coalesced horizons. The harmonic function (2.26 This singularity is unlike the Schwarzschild singularity. The extremal limit of the black brane solution occurs for r+ = r− . the mass density of the solution saturates a “Bogomol’ny bound” with respect to the charge density. the pform charges UAB and VABCDE and the scalar charge magnitudes U and V (1. and we shall see that.3. in that it is a timelike curve. 2. we shall ﬁrst make more precise the relation between the geometry of the pbrane solutions. however. 1.1 and 3.1 pform charges Now let us consider the inclusion of sources into the supergravity equations. This asymptotic falloﬀ is slow enough to give a nonvanishing total charge density from a Gauss’ law ﬂux integral at transverse inﬁnity. we shall then discuss the relations between these charges. The electric source that couples to D = 11 supergravity is the fundamental 23 .5.21) has a singularity which has for simplicity been placed at the origin of the transverse coordinates y m .9) shows that the “location” of the pbrane in spacetime should normally be thought to coincide with the inner horizon. and thus it may be considered to be the wordvolume of a δfunction source. as one can see from (2. for the “extremal” class of solutions that is our main focus.
The second equality in (4. U= M8 ∗ J[3] = 1 3! J 0MN d8 SMN . similar considerations apply to the magnetic charge (1. Varying the source action (4. (4. The charge derived in this way from a single 2brane source is thus U = Qe as expected.supermembrane action.5) follows using Stokes’ theorem and the conservation of the current J[3] .4). The diﬀerence in the electric charges obtained is then given by 1 Lv ∗J[3] = δU = J 0MN v R d7 SMN R . the diﬀerence (4.4) where d8 SMN is the 8volume element on M8 . M8 (4. one obtains the δfunction current J MN R (z) = Qe W3 δ 3 (z − x(ξ))dxM ∧ dxN ∧ dxR . instead of the Gauss’ law expression for the charge.3) becomes apparent.5) 3! ∂M8 M8 where Lv is the Lie derivative along the vector ﬁeld v. (4. Now consider the eﬀect of making diﬀerent choices of the M8 integration volume within the D = 10 spatial spacetime section. one may instead rewrite the charge as a volume integral of the source. Let the diﬀerence between the surfaces M8 and M′ be inﬁnitesimal and be given 8 by a vector ﬁeld v N (x).20 whose bosonic part is Isource = Qe W3 d3 ξ − det(∂µ xM ∂ν xN gMN (x)) + 1 µνρ ǫ ∂µ xM ∂ν xN ∂ρ xR AMN R (x) .5) between the charges calculated using the integration volumes M8 and M′ will vanish.2) This current now stands on the RHS of the A[3] equation of motion: 1 d( ∗F[4] + 2 A[3] ∧ F[4] ) = ∗J[3] . as shown in Figure 4.1) 3! The source strength Qe will shortly be found to be equal to the electric charge U upon solving the coupled equations of motion for the supergravity ﬁelds and a single source of this type. 8 This divides the electriccharge integration volumes into two topological classes 24 . (4.3) Thus.1) with δ/δA[3] . (4. speciﬁed within a D = 10 spatial section of the supergravity spacetime by a 2form. Now a topological nature of the charge integral (1. As long as the current J[3] vanishes on the boundary ∂M8 .
2. ﬂat pbrane solutions such as (2.5. The charges 25 . 2. as shown in Figure 4 and giving U = Qe . distinguishing those for which ∂M8 “captures” the pbrane current. The necessity of considering asymptotic pbrane volume forms arises because the notion of a pform charge is not limited to static. from those that do not capture the current.3) is essentially topological. Such charges can also be deﬁned for any solution whose energy diﬀers from that of a ﬂat. Both the magnitude and the orientation of this pform charge are conserved using the supergravity equations of motion. giving U =0. static one by a ﬁnite amount.3. The charge thus naturally has a magnitude Q[p]  = Qe and a unit pform orientation Q[p] /Q[p]  that is proportional to the asymptotic spatial volume form of the pbrane. an integration volume M8 extending out to inﬁnity ﬂips from the “capturing” class into the “noncapturing” class when ∂M8 crosses the δfunction surface deﬁned by the current J[3] .J [3] ' M8 v(x) M8 Figure 4: Diﬀerent choices of charge integration volume “capturing” the current J[3] . The above discussion shows that the orientationdependence of the U charges (1.6). The topological classes for the charge integrals are naturally labeled by the asymptotic orientations of the pbrane spatial surfaces.
but the corresponding energy densities will not in general saturate the BPS bounds. Since the boundary ∂M of the inﬁnite integration volume M does not capture the locus where the pbrane current is nonvanishing. this case may be considered simultaneously with that of the inﬁnite pbranes). Such charges do not occur in the supersymmetry algebra (1.5) take nonvanishing values. the current calculated using M will vanish as a result. For a ﬁnite energy diﬀerence with respect to a ﬂat. e If one considers integration volumes that do not extend out to inﬁnity. but periodic.5). Instead of an inﬁnite pbrane. static pbrane.21 26 . one may alternately have a pbrane wrapped around a compact dimension of spacetime. the pbrane must be either inﬁnite or wrapped around a compact spacetime dimension.5) for such backgrounds. the asymptotic orientation of the pbrane volume form must tend to that of a static ﬂat solution. so that an integrationvolume boundary ∂M8 is still capable of capturing the pbrane locus (if one considers this case as an inﬁnite. Only in such cases do the pform charges occurring in the supersymmetry algebra (1.4) occurring in the supersymmetry algebra (1.for such solutions will also appear in the supersymmetry algebra (1. The case of a ﬁnite pbrane is sketched in Figure 5. giving zero charge.e J M Figure 5: Finite pbrane not captured by ∂M. which plays the rˆle of a “BPS vacuum” o in a given pform charge sector of the theory.3) or (1. In order to have a nonvanishing value for a charge (1. then one can construct integration surfaces that capture ﬁnite pbranes.5). solution. but they are still of importance in determining the possible intersections of pbranes.
3 pbrane charges As one can see from (4.9) has E > 2λΩD−d−1 / ∆. . and with a.9) √ By contrast. . . ∆ ˜ (4. dropping the divergent spatial dΣµ=i integral. . E= dD−d−1 Σm (∂ n hmn − ∂m hb ) . this will also give the value of mass/(unit pvolume). b (4. . the total energy as measured by a surface integral at spatial inﬁnity diverges.24) and the charge parameter λ implies a deep 27 . ∆ (4. we consequently have a relation between the mass per unit p volume and the charge parameter of the solution E= 2λΩD−d−1 √ .2 pbrane mass densities Now let us consider the mass density of a pbrane solution. Since the pbrane solutions have translational symmetry in their p spatial worldvolume directions. Since we are considering solutions in their rest frames.9).4. e.6) ∂MT written for gM N = ηM N +hM N tending asymptotically to ﬂat space in Cartesian coordinates. The ADM formula for the energy density written as a Gauss’law integral (see.24) is seen to saturate the inequality E ≥ √ 2λΩD−d−1 / ∆. since d(D−d−1) Σm = rd y m dΩ(D−d−1) . the relation (2. What is thus more appropriate to consider instead is the value of the density. For the general pbrane solution (2. this will be a d(D−d−1) Σm surface integral over the boundary ∂MT of the transverse space. or tension of the solution.16 ) is. .7) b ˜ ˜ ∆(D − 2)rd ∆(D − 2)rd and.24). D − 1. one ﬁnds ˜ 8k(d + 1 d) 4kd 2 hmn = δmn . Recalling that k = √ ˜ ∆λ/(2d). (4.26) between the integration constant k in the solution (2.8. one ﬁnds E= ˜ 4k dΩD−d−1 . . d − 1. the black brane solution (3. Instead of the standard spatial dD−2 Σa surface integral. so the extremal pbrane solution (2. energy/(unit pvolume).. 4. b spatial indices running over the values µ = i = 1.g. . owing to the inﬁnite extent.8) where ΩD−d−1 is the volume of the S D−d−1 unit sphere. hb = . 4. Ref. m = d.
they must also possess another conserved 28 . however. the D = 11 membrane solution has equal mass and charge densities. the A ∧ F term in (1. (4) (4. let us choose the integration subsurface so as to coincide with the transverse space to the d = 6 worldvolume.3) vanishes. The ∗ F term does. if we once again consider the bosonic sector of D = 11 supergravity theory (1. give a contribution in the elementary/electric case. M8T . however.11) Thus. The magnetic charge (1.2 To be speciﬁc. M5 = M5T . and has generally become known as a “Page charge. 4. so for the p = 2 elementary membrane solution (3.24).6) is purely transverse.6). (7) (4.2). we have a saturation of the masscharge inequality: E = V = λΩ4 . saturating the inequality E ≥ U . one ﬁnds the Gauss’law form conserved quantity 2 U (1. like the classic extreme ReissnerNordstrom blackhole solution to which it is strongly related (as can be seen from the CarterPenrose diagram given in Figure 1).1). Once again. we have V = ∂M5T dΣm ǫmnpqr F npqr = λΩ4 . Then.12) Thus. For the pbrane solutions (2.4 Preserved supersymmetry (4.3) is present. the mass/charge relation is E = U = λΩ7 . one ﬁnds (7) U= ∂M8T dΣm Fm012 = λΩ7 . so no electric charge (1.10) Since the D = 11 F[4] ﬁeld strength supporting this solution has ∆ = 4. (4. In the electric case.4) is carried by this solution.13) Since the bosonic solutions that we have been considering are consistent truncations of D = 11 supergravity. this charge is a quantity conserved by virtue of the equations of motion for the antisymmetric tensor gauge ﬁeld A[n−1] . in the solitonic/magnetic 5brane case as well.e. Now let us consider the charge carried by the solitonic/magnetic 5brane solution (3. The ﬁeld strength in (3.link between the energy density and certain electric or magnetic charges.” after its ﬁrst discussion in Ref. i.1). The surface element for this transverse space is dΣm . provided one picks M8 to coincide with the transverse space to the d = 3 membrane worldvolume. for which the antisymmetric tensor ﬁeld equation was given in (3.3).
its value will clearly be zero for the class of purely bosonic solutions that we have been discussing.15) is found to be given 29 by (1. The diﬀerent charge contributions to the supersymmetry algebra occurring for diﬀerent values of N (hence diﬀerent ∆) aﬀect the Bogomol’ny bounds as shown in (4. For the D = 11 solutions.5) Thus. All of these boundsaturating solutions share the important property that they leave some portion of the supersymmetry unbroken. In order to see how a purely bosonic solution may leave some portion of the supersymmetry unbroken. since the supercharge is a Grassmanian (anticommuting) quantity.18 This theory 1 has just one spinor ﬁeld. where dΣ(9)b = dΣ(9)0b → dΣ(9)AB : Q= ∂M10 ΓABC ψC dΣ(9)AB .14. Admittedly. where N is the number of antisymmetric tensor ﬁeld strengths participating in the solution (distinct. the 29 .32 √ electric bound (4. However. the supercharge is then 1 Q= ∂M10 Γ0bc ψc dΣ(9)b .14) One can also rewrite this in fully Lorentzcovariant form. The Gauss’law form of the supercharge is given as an integral over the boundary of the spatial hypersurface. but of the same rank). (4. the D = 11 supersymmetry algebra for the supercharge (4. the supercharge. consider speciﬁcally once again the membrane solution of D = 11 supergravity. (4.15) After appropriate deﬁnitions of Poisson brackets. The positivity of the Q2 operator on the LHS of the algebra (1. The saturation of the Bogomol’ny inequalities by the pbrane solutions is an indication that they ﬁt into special types of supermultiplets. the functional form of the supercharge is still important. this surface of integration is the boundary at inﬁnity ∂M10 of the D = 10 spatial hypersurface.16a) E ≥ (2/ ∆)U √ E ≥ (2/ ∆)V magnetic bound (4.quantity.16b) that are saturated by the pbrane solutions.16). Within the family of pbrane solutions that we have been discussing. as it determines the form of the asymptotic supersymmetry algebra. 4. it turns out 32 that the ∆ values of such “supersymmetric” pbranes are of the form ∆ = 4/N .5) is at the root of the Bogomol’ny bounds 30.26. the supersymmetry algebra wraps together all of the conserved Gauss’law type quantities that we have discussed.
Checking for the consistency of setting ψM = 0 with the supposition of some residual supersymmetry with parameter ǫ(x) requires solving the equation ˜ δψA = DA ǫ = 0 . which thus reduces eﬀectively to solving (4. In order to solve the Killing spinor equation (4. 30 . An appropriate basis that does this is ΓA = (γµ ⊗ Σ9 . Σ9 and Σm are 16 × 16 SO(8) l matrices.20) where ǫ2 is a constant SO(2. = DA ǫ − (4. with Σ9 = Σ3 Σ4 .18) ˜ Solving the equation DA ǫ = 0 amounts to ﬁnding a Killing spinor ﬁeld in the presence of the bosonic background.19) where γµ and 1(2) are 2 × 2 SO(2. one would like to preserve SO(2. y) = ǫ2 ⊗ η(r) . Thus. the second condition follows from the ansatz reﬁnement (2. (4. The most general spinor ﬁeld l 9 consistent with (Poincar´)3 × SO(8) invariance in this spinor basis is of the e form ǫ(x.17) in a pbrane background. 2 l Analysis of the the Killing spinor condition (4. precisely the linearitycondition reﬁnement (2. 1) matrices. so Σ2 = 1(16) . Σ10 . Since the Killing spinor equation (4.17) (considered as a condition on φ′ /a) and from (2. . 1(2) ⊗ Σm ) . (4. it is convenient to adopt an appropriate basis for the D = 11 Γ matrices.17) for a commuting quantity. 1) spinor and η(r) is an SO(8) spinor depending only on the isotropic radial coordinate r.15) that we made in the pbrane ansatz. For the d = 3 membrane background. 1) × SO(8) covariance. The ﬁrst of these conditions is. η may be further decomposed into Σ9 eigenstates by the use of 1 (1 ± Σ9 ) projectors. l (4.gravitino ψM . the Grassmanian (anticommuting) character of this parameter is irrelevant to the problem at hand.17) ψ=0 where ψA = eA ψM and M ˜ DA ǫ DA ǫ 1 (ΓA BCDE − 8δA B ΓCDE ) FBCDE ǫ 288 1 = (∂A + 4 ωA BC ΓBC )ǫ .22). however. . what appeared previously to be simplifying specializations in the derivation given in Section 2 turn out in fact to be conditions required for supersymmetric solutions.18 on the background and on the spinor ﬁeld η(r): 1) The background must satisfy the conditions 3A′ + 6B ′ = 0 and C ′ eC = 3A′ e3A .17) in the above spinor basis leads to the following requirements 12.17) is linear in ǫ(x).
In a more telegraphic partial discussion. one may accordingly observe from (1. Qβ } = −(CΓ0 )αβ E + (CΓ12 )αβ U12 . at most a ﬁnite number of parameters can remain unﬁxed in the product spinor ǫ2 ⊗ η0 .4). f The 31 . the requirement (4. y) must take the form ǫ(x. the surviving local supersymmetry parameter ǫ(x. and ﬁnally the imposition of projection conditions on that ﬁnite set. one may jump straight to the projection conditions 3). But one can also see more directly what they will be simply by considering the supersymmetry algebra (1. So far. In general. which has a full set of 32 constant components.11). 1 {Qα . of course. one has. the membrane solution saturating the Bogomol’ny bound (4. These must. Qβ } = 2EP012 2vol P012 = 1 (1 + Γ012 ) . The maximum number of such rigid unbroken supersymmetry components is achieved for D = 11 ﬂat space. 2 l (4. where ǫ∞ = ǫ2 ⊗ η0 . after imposing this requirement. Since this is half of the maximum rigid number (i. so the constant SO(8) spinor η0 is also required to l be chiral. 3) (1 + Σ9 )η0 = 0. Thus. the local supersymmetry of the D = 11 theory is almost entirely broken by any particular solution.16a) with E = U = U12 .21) as 1 {Qα . as we have seen in (4.f This cuts the number of surviving parameters in the product ǫ∞ = ǫ2 ⊗η0 by half: the total number of surviving rigid supersymmetries in ǫ(x. Note that. specialized to the BPS background.1) that a D = 11 parity transformation requires a sign ﬂip of A[3] .5). the procedure for checking how much supersymmetry is preserved by a given BPS solution follows steps analogous to points 1) – 3) above: ﬁrst a check that the conditions required on the background ﬁelds are satisﬁed.e.e. y) is thus 2·8 = 16 (counting real spinor components).17) has cut down the amount of surviving supersymmetry from D = 11 local supersymmetry (i. y) = H −1/6 ǫ∞ . then a determination of the functional form of the supersymmetry parameter in terms of some ﬁnite set of spinor components. one may rewrite (4. half of the 32 for ﬂat space). Thus.21) 2vol Since. for example. eﬀectively an inﬁnite number of components) to the ﬁnite number of independent components present in ǫ2 ⊗ η0 . also emerge from a full analysis of equations like (4. one says that the membrane solution preserves “half” of the supersymmetry.2) η(r) = H −1/6 (y)η0 = eC(r)/6 η0 .e.22) speciﬁc chirality indicated here is correlated with the sign choice made in the elementary/electric form ansatz (2. (4. i. in the case of a D = 11 membrane solution oriented in the {012} directions. after normalizing to a unit 2volume. where η0 is a constant SO(8) spinor.17).
P012 = P012 ) whose trace is trP012 = 1 2 · 32. such as those involving higher form ﬁelds.2) preserves half of the maximal rigid D = 11 supersymmetry. When we come to discuss the cases of “intersecting” pbranes in Section 7. (4. In the early days of research on super p branes. and the projection condition following from the algebra of preserved supersymmetry generators for a 5brane oriented in the {012345} directions is P012345 ǫ∞ = 1 l 0. requiring a generalization of the formalism that we shall now present.16). it will be useful to have quick derivations like this for the projection conditions that must be satisﬁed by surviving supersymmetry parameters. Qβ } is the underlying principle in the derivation 30. This class consists of those branes whose worldvolume variables are just the bosonic and fermionic coordinates of the super pbrane in the target superspace. but it is now recognized that more general kinds of worldvolume multiplets can also occur. In the 5brane case. and these solutions preserve one component of unbroken supersymmetry for each such zero eigenvalue. Any surviving supersymmetry transformation must give zero when acting on the BPS background ﬁelds.16). this translates to P012 ǫ∞ = 0 . the analogue of condition 2) above is ǫ(x.2 where P012 is a projection operator (i.32 of the Bogomol’ny bounds (4. Let us reformulate in HoweTucker form the bosonic part of the pbrane action coupled to gravity alone. Thus. Qβ } of the generators must give zero when contracted with a surviving supersymmetry parameter ǫα . where P012345 = 2 (1 + Γ012345 ).23) which is equivalent to condition 3) above.22). Similar consideration of the solitonic/magnetic 5brane solution 25 (3. A consequence of this positive semideﬁniteness is that zero eigenvalues correspond to solutions that saturate the Bogomol’ny inequalities (4. this was the only class known. (1 + Σ9 )η0 = 0. and half are unity. More generally. the positive semideﬁniteness of the operator {Qα .6) shows that it also preserves half the rigid D = 11 supersymmetry.e.1). We shall now want to extend this treatment to the full set of bosonic and fermionic variables of an important class of super pbranes. 5 The super pbrane worldvolume action We have already seen the bosonic part of the action for a supermembrane in background gravitational and 3form ﬁelds in Eq. thus. y) = H −1/12 (y)ǫ∞ . and so the anticommutator {Qα . From (4. we once l again see that the D = 11 supermembrane solution (3. half of its eigenvalues are zero. by introducing an independent worldvolume 32 .26. (4.
Now generalize the target space to superspace Z M = (xm . one may write the super pbrane action in GreenSchwarz form: I = dp+1 ξ ij a b 1√ 2 −γ[γ Ei Ej ηab − (p − 1)] .A1 .3) which states that γij is equal to the metric induced from the spacetime metric gmn through the embedding xm (ξ).metric γij :33 IHT = 1 2 dp+1 ξ − det γ γ ij ∂i xm ∂j xn gmn − (p − 1) . . where Ei = ∂i z M (ξ)EM (z(ξ). . (5. (5. where E A = dZ M EM are superspace B = [(p + 1)!] E . D−1.. . . p = 1. Note the essential appearance here of a “cosmological term” in the worldvolume action for p = 1. .1) for the worldvolume position variables xm is γ ij ∂i ∂j xm − k (γ) ∂k xm + ∂i xp ∂j xq Γm (g) pq ij =0. The γ ﬁeld equation is γij = ∂i xm ∂j xn gmn (x) . The pbrane worldvolume is a map z M (ξ) from the space of the worldvolume parameters ξ i to the target superspace. .2) k where ij (γ) and Γm (g) are the Christoﬀel connections for the worldvolume pq metric γij (ξ) and the spacetime metric gmn (x) respectively. a = 0. The absence of this term in the speciﬁc case of the string.1) where the index ranges are i = 0.4) thus demonstrating the classical equivalence of (5.4). this map may be used to pull A A A back forms to the worldvolume: E A = dξ i Ei . (5. Using these. p and m = 0. is at the origin of the worldvolume Weyl symmetry which is obtained only in the string case. .5) + 1 Ap+1 A ǫi1 ···ip+1 Ei1 1 · · · Eip+1 BAp+1 ···A1 (p + 1)! 33 . 1. E vielbein 1forms. The equation of motion following from (5. (D − 1) and fermionic values α appropriate for the corresponding spinor dimensionality.2). one obtains the same equation as that following from the NambuGoto form action that generalizes (4. θα ) and describe A the supergravity background by vielbeins EM (Z) and a superspace (p+1) form A Ap+1 −1 A1 BAp+1 .1) ING = dp+1 ξdet1/2 (∂i xm ∂j xn gmn (x)) .1) and (5. . . . Inserting (5. (5.3) into (5.. . 1. . . 1. (5. The superspace world indices M and tangent space indices A run both through bosonic values m.
Writing the super pbrane action in this manifestly targetspace supersymmetric form raises the question of how the expected supersymmetric balance of bosonic and fermionic degrees of freedom can be achieved on the worldvolume. But. 2 A Let δz A = dz M EM and consider a transformation such that δz a = 0 . This fermionic gauge symmetry is called “κ symmetry” and its general implementation remains something of a mystery.. the D = 11 spinor dimensionality is 32.6) where κβ (ξ) is an anticommuting spacetime spinor parameter and Γ= 4 (−1) ap+1 a √ ǫi1 . it can be related to a standard worldvolume supersymmetry. The diﬀerence can only be accounted for by an additional fermionic gauge symmetry.. Eip+1 Γa1 . where the κ symmetry parameter has a spacetime spinor index just like the spinor variable θ. i. which thus subtracts from the above account three worldvolume reparameterizations for the ξ i . In order to see 1 the projection property of 2 (1 + Γ). but the κ transformation involves a projector that reduces the number of degrees of freedom removed from the spectrum by 1 . p = 2 supermembrane 20 . leaving 11 − 3 = 8 bosonic nongauge degrees of freedom. γij = Ei Ej ηab . In some formalisms. The most physically transparent formalism is the original one of Ref. . because they are expected to satisfy ﬁrstorder equations of motion on the worldvolume. one should ﬁrst remove the worldvolume gauge degrees of freedom. from which one obtains Γ2 = 1. as opposed to the secondorder equations expected for the bosons. multiplying by 2 in order to take account of this diﬀerence.7) where Γa1 . in order to compare correctly the worldvolume degrees of freedom.ap+1 (p + 1)! γ (p+1)(p−2) . . α δz α = (1 + Γ)β κβ (ξ) . Now.ap+1 is the antisymmetrized product of (p + 1) gamma matrices. one would still be expecting to have 16 active worldvolume fermionic degrees of freedom.e.. Detailed analysis of the conditions for κinvariance 35 show that the super 34 . let γij be given by the solution to its ﬁeld a b equation. For example.34 but this introduces additional twistorlike variables that obscure somewhat the physical content of the theory.ip+1 Ei11 .where γ is the traditional shorthand for det γij . The fermions are not in fact expected to match this number.” i..e. (5. while the bosonic coordinates take only 11 values.. (5.35 . 1 taken “strength one. instead of the a priori 32.. in the case of the D = 11. so that 1 (1 + Γ) 2 is indeed a projector. with a normalization factor (p+1)! .
on the one hand. this link arises already at the classical level.e.9a) (5. (5. (5. (Γa )αβ = (Γa )α γ Cγβ ). and on the other hand their consistent propagation (i. but it is not fully understood how the preservation of κ symmetry is achieved in the presence of quantum corrections.8c) where we are using a notation in which the spinor indices on (Γa1 ···ap )αβ are raised and lowered by the charge conjugation matrix (e. preservation of κ symmetry) requires the backgrounds in which they move to satisfy the supergravity equations of motion. For the super pbranes of maximal supergravities.5) is κinvariant provided the following conditions hold:g i) The ﬁeld strength H = dB = constraints Hαap+1 ···a1 Hαβγap−1 ···a1 Hαβap ···a1 = = = 1 Ap+1 (p+1)! E · · · E A1 HA1 ···Ap+1 satisﬁes the (5.22 h This observation in turn poses an unresolved question.8. where the beta function conditions enforcing the vanishing of worldvolume conformal anomalies impose a set of eﬀective ﬁeld equations on the background.9) in a maximally supersymmetric theory (e.8b) 0 0 (−1)p+ 4 (p+1)(p−2) (Γa1 ···ap )αβ . these constraints imply the supergravity equations of motion.9a). what one learns from the requirements for κ symmetry allows for terms involving a spinor Λα on the RHS of Eqs. This is a link between supersymmetric objects and the corresponding parent supergravities that is even more direct than that found in quantized string theories. D = 11 supergravity or one of the D = 10 N = 2 superstring theories): in a general background. the natural sources for the corresponding supergravities. This is a dramatic link between the super pbranes and their parent supergravity theories – these supersymmetric extended objects are.g.g.8a) and (5. Now observe a remarkable consequence of the conditions (5. ii) The superspace torsion satisﬁes c ηc(a Tb)α a Tαβ = 0 = (Γa )αβ .5. g Strictly 35 .9b) iii) H is closed.8a) (5. The beta function conditions for vanishing of the string conformal anomalies naturally generate quantum corrections to the eﬀective ﬁeld equations.h speaking. 2p! 1 (5. but this spinor can then be set to zero by a judicious choice of the conventional superspace constraints.pbrane action (5.
although the nonassociative nature of the quaternions makes this rather more cumbersome.12) holds are also related to the existence of “twocomponent notations” over the various division algebras R. it ¯ follows that the second factor dθΓab1 ···bp−1 dθ in (5. the closure of H.11) to hold is the gamma matrix condition (5. C) and in D = 6 it is SL(2.8). where P is a chirality projector that is required if the spinor coordinate θ is MajoranaWeyl. 10 superstring cases: in D = 3. Consequently.36 It should be noted that the superparticle and superstring cases allow for minimal and also extended supersymmetries. (5. The (D. D = 3. R) invariant tensor ǫαβ that would be needed to contract indices is antisymmetric. keeping in mind that the dθα fermionic oneforms are commuting. Moreover. This possibility arises because there exist corresponding extended worldvolume multiplets constructed from spinors and scalars alone. but is the unit matrix otherwise. the Lorentz group is SL(2.11) Noting that the diﬀerential dθ is commuting and that for consistency with the H constraints (5. 0). requires ¯ ¯ (dθΓa dθ)(dθΓab1 ···bp−1 dθ) = 0 .10) so that condition iii).11) does not vanish of its own accord. one can achieve a balance between the worldvolume bosonic and fermionic degrees of freedom. p) spacetime/worldvolume dimensions shown in Figure 6). there is an analogous relation between the Lorentz group and the quaternions O. conditions i) and ii) apply automatically. as a moment’s consideration of the D = 3 case shows. R). allowing thus gauge choices of the form θ = (S. as required by the residual unbroken supersymmetry. O. Analysis 36 of this constraint shows it to hold in the (D.12) in these cases using twocomponent notation makes its proof relatively transparent. 6. while the SL(2. An essential import of the κ symmetry for the super pbrane actions is that it allows one to gauge away half of the spinor variables. dθµ δµ } (5.To understand the import of κsymmetry better. one has a a ¯ E A = {(dxm − iθΓm dΘ)δm . Writing out the gamma matrix identity (5. Similar considerations apply to the p = 0. 4. in D = 10. one must have (Γab1 ···bp−1 )αβ symmetric in (αβ). H. p) cases in which (5. what one requires for (5.12) (Γa P )(αβ (Γab1 ···bp−1 P )γδ) = 0 . note that in ﬂat superspace. H). C. while in D = 4 it is SL(2. Examples of this may be seen in the p = 1. After this reduction by half in the number of nongauge fermion worldvolume ﬁelds. in ﬂat superspace. superparticle cases. The counting of degrees of freedom in 36 .
and making a Lorentz ˜ 37 . ♥. O.12) and are associated to the division algebras R. while the other chirality modes occur as supersymmetry singlets. xm ). but these involve worldvolume ﬁelds other than the simple scalarspinor multiplets shown here. however. The four sequences ♣.and rightmoving modes. 1. the MajoranaWeyl spinor variable yields superpartners only for one chirality amongst the bosonic modes. because the bosonic worldvolume degrees of freedom can be split into chiral left. m = p + 1. one needs to ﬁx the worldvolume reparameterization symmetry. If one now makes a Lorentz transformation in the target spacetime. ♠. There also exist other worldvolume actions. . . It would seem that a worldvolume supersymmetry should require the presence of spinor ﬁelds on the worldvolume. Splitting the index range after the static gauge choice accord˜ ing to the pattern xm = (xi . this gauge condition will generally be broken. i = 0. H. p. so in order to maintain this condition one needs to make a compensating gauge transformation. in the minimal N = 1 superstring in D = 10. This may conveniently be achieved using the “static” gauge choice xi (ξ) = ξ i . ♦ of (D. . superstring cases needs a little more care. To see how such worldvolume spinors appear. but worldvolume scalars. however. For example. p) values correspond to solutions to the gamma matrix identity (5.p D 11 10 9 8 7 6 5 4 3 2 0 1 2 3 4 5 ♦ ♦ ♦ ♠ ♠ ♠ ♠ ♥ ♥ ♣ ♥ ♥ ♣ ♠ ♠ ♠ ♦ O H C R ♥ ♣ ♣ Figure 6: Sequences of kinvariant pbrane actions involving worldvolume spinors and scalars only. . . corresponding to the diﬀerent brane types shown in Figure 7. One might ﬁnd it strange to have an unbroken worldvolume supersymmetry given the fact that the spinor pbrane variables θα are targetspace spinors. . C. . . In the N = 2 superstrings. D − 1. both chiralities of bosonic modes are paired with spinors.
. Now consider the eﬀect of the linearly realized SO(p. 11 and the S ˜ residual spinor (32/2 = 16 components) form a worldvolume scalarspinor multiplet with respect to the unbroken worldvolume supersymmetry. 38 . . A classic example of a worldvolume supermultiplet with respect to such a worldvolume supersymmetry is provided ˜ by the D = 11. but leaving still a linearly realized subgroup. the 1 fraction of unbroken supersymmetry is 2 . as a worldvolume spinor. 1) become nonlinearly realized. . ˜ (5. p = 2 supermembrane: the xm . 1) becomes the ˜ worldvolume Lorentz group. causing some of the original targetspace supersymmetry to become nonlinearly realized.13) Demanding that δxi = 0 for xi = ξ i requires the compensating reparameteri˜ zation to be given by η i = −Li ξ j − Li˜ xm . which may be identiﬁed as the unbroken worldvolume supersymmetry. For the simple pbrane actions that we have been considering in this section.14) This combined transformation remains linearly realized (i. which is an “internal symmetry” from the worldvolume point of view). ˜ ˜ (5. after gauge ﬁxing. 4 (5. The linearly realized SO(p. The existence of such a large multiplet of scalars and spinors is a speciﬁc feature of the unbroken d = 3. “unbroken”) for a subgroup SO(p. according to which the xm transform as a set of D − p − 1 scalars (although they transform as a vector with respect to the remaining SO(D − p − 1) factor. 1) × SO(D − p − 1). transforming under 16 unbroken supercharges (it also has an unbroken SO(8) “internal” symmetry).15) The matrices Γij provide a spinor representation (in general reducible) of the “worldvolume” SO(p.35 Similar considerations apply to supersymmetry after κ symmetry ﬁxing. 1) algebra. Then the combined transformation j m m ˜ for the remaining bosonic variables x is ˜ ˜ ˜ ˜ ˜ ˜ δxm = −Lk j ξ j ∂k xm + Lm n xn − Lk n xn ∂k xm . 1) on the θ(ξ) variables: δθ = −Li j ξ j ∂i θ + 1 Lij Γij θ . Preserving the κ symmetry gauge requires a similar compensating κ symmetry transformation. but the remaining generators of SO(D − 1. “N = 8” supersymmetry. This multiplet contains 8 + 8 bosonic + fermionic degrees of freedom.e. so that θ can now be identiﬁed. Lm n ). . Li m . m = 3.˜ transformation with parameters (Li j . a combined Lorentz trans˜ ˜ formation and worldvolume reparameterization (with parameter η i (ξ)) on the xi variables takes the form ˜ δxi = η j (ξ)∂j X i + Li j xj + Li m xm .
a restriction such that solutions to the equations for the restricted variables are also solutions to the equations for the unrestricted variables.6 KaluzaKlein dimensional reduction Let us return now to the arena of purely bosonic ﬁeld theories. and consider the relations between various bosonicsector theories and the corresponding relations between pbrane solutions. The spinor sectors of the theories are equally well related by dimensional reduction. Next.1) where carets denote (D + 1)dimensional quantities corresponding to the (D + ˆ 1)dimensional coordinates xM = (xM . The essential step in a KaluzaKlein dimensional reduction is a consistent truncation of the ﬁeld variables. generally made by choosing them to be independent of the reduction coordinate z.2) where F = dA. i. but in the following. In order to set up the procedure. but break up the metric in Ddimensionally covariant pieces: dˆ2 = e2αϕ ds2 + e2βφ (dz + AM dxM )2 s (6. This ensures that the lowerdimensional solutions which we shall obtain are also particular solutions to higherdimensional supergravity equations as well. It is wellknown that supergravity theories are related by dimensional reduction from a small set of basic theories. one should pick β = −(D − 2)α in order to arrange for the Einsteinframe form of the gravitational action in (D + 1) dimensions to go over to the Einsteinframe form of the action in D dimensions. If one now chooses α2 = [2(D − 1)(D − 2)]−1 . By a consistent truncation. the largest of which being D = 11 supergravity. Adjustment of the constants α and β is necessary to obtain desired structures in D dimensions. one needs to establish the reduction ansatz for the (D + 1)dimenˆ ˆ sional antisymmetric tensor gauge ﬁeld F[n] = dA[n−1] . the ϕ kinetic term becomes conventionally normalized. z). The scalar ϕ in D dimensions emerges from the metric in (D + 1) dimensions as (2β)−1 ln gzz . let us consider a theory in (D + 1) dimensions. Making the parameter choice β = −(D − 2)α to preserve the Einsteinframe form of the action. we always understand a restriction on the variables that commutes with variation of the action to produce the ﬁeld equations. among the ˆ n − 1 antisymmetrized indices of A[n−1] at most one can take the value z.e. In particular. one obtains √ 1 −g R(g)−(D−1)(D−2)α2 ∇M ϕ∇M ϕ− 4 e−2(D−1)αϕ FM N F M N (6. Clearly. so −ˆR(ˆ) = g g 39 . ds2 is the line element in D dimensions and α and β are constants. we shall restrict our attention to the purely bosonic sector.
5) where the second term in (6.4b) However. n − 1) are SO(2)rotated . Thus. and. r = (n.we have the decomposition ˆ A[n−1] = B[n−1] + B[n−2] ∧ dz . [n] (6. Now.3) All of these reduced ﬁelds are to be taken to be functionally independent of z. (6. it is useful to introduce G′ = G[n] − G[n−1] ∧ A .6) Although the dimensional reduction (6.4a) (6. we are ready to perform the dimensional reduction of our general action (2. The metric in (D+1) dimensions couples to all ﬁelds.1). Accordingly. We ﬁnd that ˆ I =: reduces to I = √ 2 1 1 dD x −g R − 2 ∇M φ∇M φ − 1 ∇M ϕ∇M ϕ − 4 e−2(D−1)αϕ F[2] 2 − 1 1 −2(n−1)αϕ+αφ ′2 ˆ a 2 e G[n] − e2(D−n)αϕ+ˆφ G[n−1] . dimensional reduction will produce some terms with undiﬀerentiated KaluzaKlein vector ﬁelds AM coupling to Ddimensional antisymmetric tensors.7) has produced a somewhat complicated result.7) 2n! 2(n − 1)! dD+1 x ˆ −ˆ R(ˆ) − 1 ∇M φ∇M φ − g g ˆ 2 1 aφ ˆ 2 eˆ Fn] 2n! (6. (6. (6. Bn−1] ). At this stage. the important point to note is that each of the Ddimensional antisymmetrictensor ﬁeld strength terms G′2 and G2 [n] [n−1] has an exponential ˜ ar φ r ˜r . where the φ prefactor of the form e combinations of ϕ and φ. but retaining at the same time the scalarﬁeld combination appearing in the corresponding exponential prefactor. keeping just one while setting to zero the other two of the three gauge ﬁelds (A[1] .5) may be viewed as a ChernSimons correction from the reduced Ddimensional point of view. B[n−2] . ﬁrst deﬁne G[n] G[n−1] = dB[n−1] = dB[n−2] . For the corresponding ﬁeld strengths. is a consistent truncation. since a certain “ChernSimons” structure appears upon dimensional reduction. these are not exactly the most convenient quantities to work with. consequently. any one of the three ﬁeld strengths 40 .
solutions to this simple truncated system are also solutions to the untruncated theory. 2 immediately become applicable. The 2form ﬁeld strength F[2] = dA.(F[2] . emerges out of the gravitational action in D + 1 dimensions.8) with the same value of ∆ as for the “parent” coupling parameter a. emerging from the higherdimensional metric upon dimensional 41 .9) in D + 1 dimensions. the coeﬃcient ar satisﬁes a2 = ∆ − r ˜ 2dr dr 2(r − 1)(D − r − 1) =∆− (D − 2) (D − 2) (6. then [n] one ﬁnds oneself back in the situation described by our general action (2. in each of the ear φ prefactors. although the individual parameters ar are both Dand rdependent. the metric takes the form 31.1). ˜ An important point to note here is that. the quantity ∆ is preserved under KaluzaKlein reduction for both of the “descendant” ﬁeldstrength couplings (to G′2 or to G2 [n] [n−1] ) coming aφ ˆ 2 ˆ from the original term e F[n] . G′ ). G′ ). and indeed are also solutions to the original (D + 1)dimensional theory. G[n−1] .32 ds2 11 hi = = e 3 a·φ ds2 + D i 2 e( 3 a−ai )·φ (hi )2 2 (6. satisfying ˆ a2 = ∆ − ˆr ˜ 2d(n) d(n) 2(n − 1)(D − n) =∆− ((D + 1) − 2) (D − 1 (6. since retaining only one ﬁeld strength & scalar combination in this way eﬀects a consistent truncation of the theory. together with its corresponding scalarﬁeld combination. G[n−1] . its coupling parameter corresponds to ∆ = 4.10a) (6. If one retains in the reduced theory only one of the ﬁeld strengths (F[2] .10b) dz i + Ai + Aij dz j . Moreover. on the other hand. retained alone together with its corresponding scalarﬁeld [n] combination.1).1 Multiple ﬁeldstrength solutions and the singlecharge truncation After repeated single steps of KaluzaKlein dimensional reduction from D = 11 down to D dimensions. can support pbrane solutions in D dimensions of the form that we have been discussing. [1] [0] where the Ai are a set of (11−D) KaluzaKlein vectors generalizing the vector M AM in (6. and then the p brane solutions obtained for the general case in Sec. Thus. 6. since the KaluzaKlein dimensional reduction is also a consistent truncation.
5).11a) ij ijk i 1 = ea·φ ∗F[4] ∧v+eai ·φ ∗F[3] ∧v i + 2 eaij ·φ ∗F[2] ∧v ij + 1 eaijk ·φ ∗F[1] ∧v ijk . 11 − D.12) Using (6.32 ID = √ 1 dD x −g R − 2 (∂ φ)2 − 1 −4 1 −2 1 a·φ 2 F[4] 48 e − e 1 12 i i eai ·φ (F[3] )2 e i<j aij ·φ ij (F[2] )2 − 1 4 i 1 2 bi ·φ i (F[2] )2 (6. for later ∗ˆ reference. 42 . Once such KaluzaKlein vectors have appeared. . higher. (6. and ﬁeld strengths with multiple i. . 6. we shall also give the reduction of its Hodge dual ˆF[4] : ˆ F[4] ˆ ∗ˆ F[4] ij i = F[4] +F[3] ∧hi + 1 F[2] ∧hi ∧hj + 2 i j k 1 ijk 6 F[1] ∧h ∧h ∧h (6.3) of a 1form gauge potential. j = 1.reduction.11).1) now reduces to 31. the bosonic sector of maximal supergravity (1. ˆ We shall also need the corresponding reduction of the F[4] ﬁeld strength i (where hatted quantities refer to the original.10b) as a consequence of the usual onestep reduction (6.11b) are given by v vi vij vijk = = = = 1 (11−D)! ǫi1 ···i11−D 1 (10−D)! ǫii2 ···i11−D 1 (9−D)! ǫiji3 ···i11−D hi1 ∧ · · · ∧ hi11−D 1 (8−D)! ǫijki4 ···i11−D hi3 ∧ · · · ∧ hi11−D hi2 ∧ · · · ∧ hi11−D hi4 ∧ · · · ∧ hi11−D . where i.11a).13) i<j<k ijk eaijk ·φ (F[1] )2 − ij ij ebij ·φ (F[1] )2 + LF F A .11b) (noting that. and two equal index values never occur in a multiindex sum. subsequent dimensional reduction also gives rise to the zeroform gauge potentials Aij appearing [0] in (6. dimension) and. since the Hodge dual is a metricdependent construction.10. j indices may be taken to be antisymmetric in those indices since these “internal” indices arise in the stepwise reduction procedure. 6 (6. . expo∗ˆ nentials of the dilatonic vectors φ appear in the reduction of ˆF[4] ) where the forms v. From (6. . one sees that the “straightbacked” ﬁeld i Note that the lowerdimensional ﬁeld strengths F[n] include “ChernSimons” corrections similar to those in (6. vij and vijk appearing in (6. vi .
ij ijk i strengths F[4] , F[3] , F[2] and F[1] are descendants from F[4] in D = 11. The i “calligraphic” ﬁeld strengths F[2] , on the other hand, are the ﬁeld strengths for the KaluzaKlein vectors Ai appearing in (6.10b). Similarly, one also has a set M ij of 1form ﬁeld strengths F[1] for the KaluzaKlein zeroform gauge potentials
Aij appearing in (6.10b). [0] The nonlinearity of the original D = 11 action (1.1) in the metric tensor produces a consequent nonlinearity in the (11 − D) dilatonic scalar ﬁelds φ appearing in the exponential prefactors of the antisymmetrictensor kinetic terms in (6.13). For each ﬁeldstrength kinetic term in (6.13), there is a corresponding “dilaton vector” of coeﬃcients determining the linear combination of the dilatonic scalars appearing in its exponential prefactor. For the 4, 3, 2 and 1form “straightbacked” ﬁeld strengths emerging from F[4] in D = 11, these coeﬃcients are denoted correspondingly a, ai , aij and aijk ; for the “calligraphic” ﬁeld strengths corresponding to KaluzaKlein vectors and zeroform gauge potentials emerging out of the metric, these are denoted bi and bij . However, not all of these dilaton vectors are independent; in fact, they may all be expressed in terms of the 4form and 3form dilaton vectors a and aij :31,32 aij = ai + aj − a aijk = ai + aj + ak − 2a bi = −ai + a bij = −ai + aj . (6.14)
Another important feature of the dilaton vectors is that they satisfy the following dotproduct relations: a·a a · ai ai · aj = = = 2(11 − D) D−2 2(8 − D) D−2 2(6 − D) 2δij + . D−2
(6.15)
Throughout this discussion, we have emphasized consistent truncations in making simplifying restrictions of complicated systems of equations, so that the solutions of a simpliﬁed system are nonetheless perfectly valid solutions of the more complicated untruncated system. With the equations of motion following from (6.13) we face a complicated system that calls for analysis in simpliﬁed subsectors. Accordingly, we now seek a consistent truncation down to a simpliﬁed system of the form (2.1), retaining just one dilatonic scalar combination φ and one rankn ﬁeld strength combination F[n] constructed out of a certain number N of “retained” ﬁeld strengths Fα [n] , α = 1, . . . , N , (this 43
could possibly be a straightbacked/calligraphic mixture) selected from those appearing in (6.13), with all the rest being set to zero.32 Thus, we let φ = nφ + φ⊥ , (6.16)
where n · φ⊥ = 0; in the truncation we then seek to set consistently φ⊥ = 0. We shall see that consistency for the retained ﬁeld strengths Fα [n] requires them all to be proportional.32 We shall let the dot product matrix for the dilaton vectors of the retained ﬁeld strengths be denoted Mαβ =: aα · aβ . Consistency of the truncation requires that the φ⊥ ﬁeld equation be satisﬁed: φ⊥ −
α
Π⊥ · aα (Fα [n] )2 = 0 ,
(6.17)
where Π⊥ is the projector into the dilatonvector subspace orthogonal to the retained dilaton direction n. Setting φ⊥ = 0 in (6.17) and letting the retained Fα [n] be proportional, one sees that achieving consistency is hopeless unless all the eaα ·φ prefactors are the same, thus requiring aα · n = a ∀α = 1, . . . , N , (6.18)
where the constant a will play the role of the dilatonic scalar coeﬃcient in the reduced system (2.1). Given a set of dilaton vectors for retained ﬁeld strengths satisfying (6.18), consistency of (6.17) with the imposition of φ⊥ = 0 requires Π⊥ · aα (Fα [n] )2 = 0 .
α
(6.19)
This equation requires, for every point xM in spacetime, that the combination 2 α aα (Fα [n] ) be parallel to n in the dilatonvector space. Combining this with the requirement (6.18), one has aα (Fα [n] )2 = an
α α
(Fα [n] )2 .
(6.20)
Taking then a dot product of this with aβ , one has Mβα (Fα [n] )2 = a2
α α
(Fα [n] )2 .
(6.21)
Detailed analysis 32 shows it to be suﬃcient to consider the cases where Mαβ −1 is invertible, so by applying Mαβ to (6.21), one ﬁnds (Fα [n] )2 = a2
β −1 Mαβ γ
(Fγ [n] )2 ,
(6.22)
44
and, indeed, we ﬁnd that the Fα [n] must all be proportional. Summing on α, one has −1 a2 = ( Mαβ )−1 ; (6.23)
α,β
one then deﬁnes the retained ﬁeldstrength combination F[n] so that (Fα [n] )2 = a2
β −1 Mαβ (F[n] )2 .
(6.24)
The only remaining requirement for consistency of the truncation down to the simpliﬁed (gM N , φ, F[n] ) system (2.1) arises from the necessity to ensure that the variation of the LF F A term in (6.13) is not inconsistent with setting to zero the discarded dilatonic scalars and gauge potentials. In general, this imposes a somewhat complicated requirement. In the present review, however, we shall concentrate mainly on either purelyelectric cases satisfying the elementary ansatz (2.4) or purelymagnetic cases satisfying the solitonic ansatz (2.6). As one can see by inspection, for pure electric or magnetic solutions of these sorts, the terms that are dangerous for consistency arising from the variation of LF F A all vanish. Thus, for such solutions one may safely ignore the complications of the LF F A term. This restriction to pure electric or magnetic solutions does, however, leave out the very interesting cases of dyonic solutions that exist in D = 8 and D = 4, upon which we shall comment later on in Section 8. After truncating down to the system (2.1), the analysis proceeds as in Section 2. It turns out 32 that supersymmetric pbrane solutions arise when the matrix Mαβ for the retained Fα [n] satisﬁes Mαβ = 4δαβ − and the corresponding ∆ value for F[n] is ∆= ˜ 2dd , D−2 (6.25)
4 , (6.26) N where we recall that N is the number of retained ﬁeld strengths. A generalization of this analysis leads to a classiﬁcation of solutions with more than one independent retained scalarﬁeld combination.32 We shall see in Section 7 that the N > 1 solutions to singlecharge truncated systems (2.1) may also be interpreted as special solutions of the full reduced action (6.13) containing N constituent ∆ = 4 brane components that just happen to have coincident charge centers. Consequently, one may consider only the N = 1, ∆ = 4 solutions to be fundamental. 45
which we have exploited in describing diagonal dimensional reduction above. (6. for supergravity ﬁeldtheory solutions.8. This procedure is the analogue.24). 6.1. Reinterpretation of p brane solutions in this way.9). Upon making such a reinterpretation.6. however. translational Killing symmetries of pbrane solutions allow a simultaneous interpretation of these ﬁeld conﬁgurations as solutions belonging to several diﬀerent supergravity theories. corresponding to the same value of ∆. as we discussed in subsection 4.3 Multicenter solutions and vertical dimensional reduction As we have seen. which was adjusted so as to maintain the Einsteinframe form of the gravitational term in the dimensionally reduced action. the quantity d is conserved. Thus. For the original single pbrane solutions (2. One may. of the procedure of double dimensional reduction 22 for pbrane worldvolume actions. coupled in to resolve the singularities. the only available translational Killing symmetries are those in the worldvolume directions. and hence is referred to as diagonal dimensional reduction.7) is to perform a Weyl rescaling on it in order to be in accordance with the form of the metric chosen in the KaluzaKlein ansatz (6. proceeds diagonally on a D versus d plot. related one to another by KaluzaKlein dimensional reduction. generalize the basic solutions (2. one may let the reduction coordinate z be one of the xµ . which can be taken to constitute the δfunction sources for singular pbrane solutions. elementary/solitonic pbranes in (D + 1) dimensions give rise to elementary/solitonic (p− 1)branes in D dimensions. corresponding to standard KaluzaKlein reduction on a worldvolume coordinate.27) H(y) = 1 + ˜ y − yα d α 46 . because they are naturally independent of the “worldvolume” xµ coordinates. since both D and d reduce by one.20).21) by a diﬀerent solution of the Laplace equation (2. Note ˜ that in this process. one can easily extend the family of pbrane solutions to multicenter pbrane solutions by taking the harmonic function to be kα kα > 0 . Consequently. Accordingly. as one can see from (6. 6.10). the only thing that needs to be done to such a solution in order to reinterpret it as a solution of a reduced system (6.24) by replacing the harmonic function H(y) in (2.2 Diagonal dimensional reduction of pbranes The family of pbrane solutions is ideally suited to interpretation as solutions of KaluzaKlein reduced theories.
.g. with all charge parameters λα = 2dkα / ∆ required to be positive in order to avoid naked singularities. The metric and the electriccase antisymmetric tensor gauge potential corresponding to (6.j In order to do this. the multicenter solutions (6.” The multicenter solutions given by (6. see. The generalized solution (6. the ansatz (2.24a.27) in a periodic array.Once again.25).28) which ensures the validity of the Bianchi identity just as well as (2.mn p ∂p λα α ˜ y − yα d .20). This can be done by “stacking” up identical p branes using (6.24). The “centers” of the individual “leaves” of this solution are at the points y = yα . by letting the integration constants kα all be equal.27) satisfy the same supersymmetrypreservation conditions on the metric and antisymmetric tensor as (2. i. however. From a mathematical point of view.6) does. the integration constant has been adjusted to make H  = 1 ↔ ∞ φ = 0. and aligning the j Similar procedures have been considered in a number of articles in the literature. From a physical point of view.27) can now be used to prepare solutions adapted to dimensional reduction in the transverse directions.6) needs to be modiﬁed so as to accommodate the multicenter form of the solution: ˜ Fm1 . α (6.29) while the total electric or magnetic charge is given by ΩD−d−1 λα . these static solutions exist as a result of cancellation between attractive gravitational and scalarﬁeld forces against repulsive antisymmetrictensor forces for the similarlyoriented pbrane “leaves.24). the multicenter solutions leave the same amount of supersymmetry unbroken as the singlecenter solution.27) are given again in terms of H(y) by (2. so the Bogomol’ny bounds (4.27) corresponds to parallel and similarly∞ √ ˜ oriented pbranes. Refs. where α ranges over any number of centers. Since the multicenter solutions given by (6.27) exist owing to the properties of the Laplace equation (2..2. The mass/(unit pvolume) density is now E= 2ΩD−d−1 √ ∆ λα . In the solitonic case.. we need ﬁrst to develop translation invariance in the transverse reduction coordinate. (6.e. e..37 47 .16) are saturated just as they are for the singlecenter solutions (2.mn = −d−1 ǫm1 . This combination of a modiﬁcation of the solution followed by dimensional reduction on a transverse coordinate is called vertical dimensional reduction 23 because it relates solutions vertically on a D versus d plot.
but ˜ d = D − d − 2 is reduced by one with each reduction step. The indication of elementary or solitonic type relates to solutions of supergravity theories in versions with the lowest possible choice of rank (n ≤ D/2) for the supporting ﬁeld strength. with the corresponding ∆ values shown adjacently. the solution will nonetheless remain isotropic in the D − d − 1 dimensions orthogonal to the stacking axis. obtainable by appropriate dualization.30c) −∞ D−2 r2 ˆ ˆ k ymym m=d = √ ˜ 1 πkΓ(d − 2 ) .30a) (6. and hence can be called “stainless” pbranes.1) may be oxidized back up to some solution in D = 11. it follows that vertical dimensional reduction from D to D − 1 spacetime dimensions preserves the value of ∆ just like the diagonal reduction discussed in the previous subsection. the z axis.13 In Figure 7. Note that under vertical reduction. one ﬁnds the general picture given in the plot of Figure 7. these solutions are indicated by the large circles. the worldvolume dimension d is preserved. Of course.“centers” yα along some axis. e. We shall see in Section 7 that what one obtains upon oxidation of the “stainless” 48 . Since the same antisymmetric tensors are used here to support both the stacked and the unstacked solutions. After a conformal rescaling in order to maintain the Einstein frame for the solution. Combining the diagonal and vertical dimensional reduction trajectories of “descendant” solutions. one has kα α +∞ ˜ d y − yα −→ = kdz (ˆ2 + r ˜ z 2 )d/2 = ˜ k rd−1 ˜˜ (6. Taking the limit of a denselypacked inﬁnite stack of this sort.30b) (6. one can ﬁnally reduce on the coordinate z along the stacking axis. one obtains a pbrane solution with the same worldvolume dimension as the original higherdimensional solution that was stacked up. ˜ 2Γ(d) where r in (6.g. reduction families emerge from certain basic solutions that cannot be “oxidized” back up to higherdimensional isotropic pbrane solutions. In this plot of spacetime dimension D versus worldvolume dimension d. every solution to a theory obtained by dimensional reduction from D = 11 supergravity (1. provided the centers are all in a line. After stacking and reduction in this way.30b) is the radial coordinate for the D − d − 1 residual isotropic ˆ transverse coordinates. Singling out one “stacking axis” in this way clearly destroys the overall isotropic symmetry of the solution. but. and since ∆ is preserved under dimensional reduction.
4/5.2/3. 2/3.D 11 4 4 4 4 4 4 4 10 9 2 2 8 7 6 4/3.4/5.4/7 selfdual } "stainless" KaluzaKlein descendants 3 1/2 4 ∆ values vertical reduction trajectories diagonal reduction trajectories 2 1 0 1 2 string 3 4 5 6 5brane 7 6brane d instanton particle membrane 3brane 4brane Figure 7: Branescan of supergravity pbrane solutions (p ≤ (D − 3)) 49 .1' 4/3 1' 1/2 2 4/3 4/3.1' elementary 5 solitonic 1.4/7 4 1.
6. If one now puts in an elementary string source action. φ) 1 = − 16 (a2 + 4)∂m K∂m Kηµν 1 2 8 (a = 1 − 4)(∂m K∂n K − 2 δmn (∂p K∂p K)) . or axion. The pbrane ansatz gives a spacetime of the form M4 = M2 × Σ2 . so such strings are also solutions to dilatonaxion gravity. Before proceeding any further with vertical dimensional reduction. consider the supersymmetric string in D = 4 dimensions. m = 2. 50 . 1 subspace. Then the integral gives the expected ln r harmonic ˆ function appropriate to two transverse dimensions. Supporting this string solution. This distinction means that there is in general a deﬁcit solid angle at transverse inﬁnity. d = D − 3) solution to a (D − 1. the integral (6.30) contains an additive divergence and needs to be renormalized.solutions in Figure 7 falls into the interesting class of “intersecting branes” built from four basic “elemental” solutions of D = 11 supergravity.k In this step. Firstly. let us consider some of the speciﬁc properties of (D − 3)branes that make the next vertical step down problematic. This is easily handled by putting ﬁnite limits ±L on the integral. one has the 2form gauge ﬁeld Aµν and the dilaton φ. with r = y m y m . the asymptotic metric of a (D − 3)brane is not a globally ﬂat space. In order to understand the global structure of the (D − 3)branes in some more detail. where M2 is D = 2 Minkowski space. which is related to the total mass density of the (D − 3)brane. These ﬁelds give rise to a ﬁeld stress tensor of the form Tµν (A. so k Solutions with worldvolume dimension two less than the spacetime dimension will be referred to generally as (D − 3)branes.12 In D = 4.38 This means that any attempt to stack up (D − 3)branes within a standard supergravity theory will soon consume the entire solid angle at transverse inﬁnity. but only a locally ﬂat space. and then by subtracting a divergent term 2 ln L before z ˜ −L taking the limit L → ∞. with the string aligned along the µ. (6.31) where a is as usual the dilaton coupling parameter and e−K = H = 1 − √ 8GT ln(r). one may dualize the 2form Aµν ﬁeld to a pseudoscalar. ν = 0. thus destroying the asymptotic spacetime in the construction. φ) Tmn (A. which becomes L d˜(r2 + z 2 )−1/2 . ﬁeld χ. 3. d = D − 3) solution. irrespective of whether the spacetime dimension is D or not.4 The geometry of (D − 3)branes The process of vertical dimensional reduction described in the previous subsection proceeds uneventfully until one makes the reduction from a (D.
φ) which is not just concentrated at the string core but instead is smeared out over spacetime.34) gives its conformal factor. within a very small transversespace region that may be considered to be the string “core. and as we have already noted. 51 . φ) = 0. Thus.32) By inspection of the ﬁeld solution. ˜ because. for which Rmn − 2 gmn R ≡ 0 is an identity. Owing to the fact that the D = 2 Weyl tensor vanishes.that Tmn (source) = 0. Consequently. the Einstein equation in the transverse m. a usual form of the Einstein equation nonetheless applies to that space as a result of the symmetries of the pbrane ansatz. There. In the worldsheet directions. the equations are satisﬁed simply by by 0 = 0. as one can see from (2. Accordingly. −gR. the Ricci tensor and hence the full curvature vanish outside the string core. although there is no sensible Einstein action in the transverse D = 2. the equations become 1 − 2 Rηµν = −8πGρηµν .” Approximating this by a delta function in the transverse space.32) are both of the form diag(ρ. the Higgs ﬁelds contributing to the energy density of the string are displaced from their usual vacuum values to unbrokensymmetry conﬁgurations at a stationary point of the Higgs potential. the transverse space Σ2 is conformally ﬂat. (6.11) with d = A′ = 0. The above supersymmetric string solution may be compared to the cosmic strings arising in gauge theories with spontaneous symmetry breaking. while the contributions to Tµν from the Aµν and φ ﬁelds and also from the source (6. Thus. space.34) or just R = 16πGρ . the supersymmetric string has a ﬁeld stress tensor Tµν (A. (6. Eq. The total energy is given by the deﬁcit angle 8πGT of the conical spacetime. 0. This equation is naturally satisﬁed for a metric satisfying the pbrane ansatz. since the transverse stress tensor components vanish. in the transverse directions. (6. In contrast.33) (6. the overall stress tensor is of the form TMN = diag(ρ. −ρ. which is ﬂat except at the location of the string core. then one has the source stress tensor −T Tµν (source) = √ −g √ 1 d2 ξ −γγ ij ∂i Xµ ∂j Xν e− 2 aφ δ(x − X) . 0). so that one obtains a conical spacetime. corresponding to the √ fact that the usual Einstein action. this causes the transverse components of the Ricci tensor to be equal to the Ricci tensor of a D = 2 1 spacetime. R = Rmm . n indices becomes 1 Rmn − 2 gmn R = 0. −ρ). is a topological invariant in D = 2. one has Tmn (A.
6) and searching for (D − 2) branes in ordinary massless supergravity theories. d = D − 2) solution (a p = (D − 3) brane) to a (D − 1. the harmonic function H(y) becomes linear in the one remaining transverse coordinate.The diﬀerence arises from the absence of a potential for the ﬁelds Aµν . however this is not in itself particularly remarkable.31). one simply doesn’t ﬁnd any such solutions. an analysis of the physics of the situation needs some care. φ supporting the solution in the supersymmetric case. the indeﬁnite stacking of supersymmetric strings leads to a destruction of the transverse asymptotic space. the metric asymptotically tends to a locally ﬂat space as r → ∞.36) .24) for this case. Upon subsequently performing the integral. with a total energy density given by the overall deﬁcit angle measured at inﬁnity. one can simply decide to be brave.3. when considered within the original supergravity theory.39 Consider the reduction from a (D. For the (D − 3)brane. because. Note that both the (D − 3) brane and its descendant (D − 2)brane have harmonic functions H(y) that blow up at inﬁnity. Nonetheless. (6. A second problem with any attempt to produce (D − 2)branes in ordinary supergravity theories is simply stated: starting from the pbrane ansatz (2.5 Beyond the (D − 3)brane barrier: ScherkSchwarz reduction and domain walls Faced with the above puzzles about what sort of (D − 2)brane could result by vertical reduction from a (D − 3)brane. 6. While the mathematical procedure of vertical dimensional reduction so as to produce some sort of (D − 2)brane proceeds apparently without serious complication. 2. the transverse space Σ2 is asymptotically locally ﬂat (ALF). For multiplecentered string solutions. one again encounters an additive divergence: the L integral −L dz ln(y 2 + z 2 ) needs to be renormalized by subtracting a divergent term 4L(ln L − 1). d = D − 3) solution (a p = (D − 2) brane). as one can see from the behavior of the stress tensor Tmn in Eq. as one can see by inspection of (2. one has H =1− i 8GTi ln y − yi  . In the next step of vertical dimensional reduction. and also in this limit the antisymmetrictensor oneform ﬁeld strength Fm = −ǫmn ∂n H 52 (6. (6.35) Consequently. and to just proceed anyway with the established mathematical procedure of vertical dimensional reduction and see what one gets.
or instead may be imposed on the ﬁeld variables in the equations of motion after variation.36) acquires a constant component along the stacking axis ↔ reduction direction z. H(y) = const.tends asymptotically to zero. z) = mz + χ(x.38) which implies an unavoidable dependence l of the corresponding zeroform gauge potential on the reduction coordinate: A[0] (x.39) prior to varying the Lagrangian will give a result diﬀerent from that obtained by imposing this dependence in the ﬁeld equations after variation. retaining any dependence on a reduction coordinate will lead to an inconsistent truncation of the theory: attempting to impose a z dependence of the form given in (6. The expression (6. such z dependence can be removed by a gauge transformation. (6. let us again focus on the problem of consistency of the KaluzaKlein reduction. with an equal eﬀect. shows that the next reduction step down to the (D − 1. for which solutions of the reduced theory are also solutions of the unreduced theory. l Note that this vertical reduction from a (D − 3)brane to a (D − 2)brane is the ﬁrst case in which one is forced to accept a dependence on the reduction coordinate z. Throughout this review. however. In this case. y) . (6. 53 . Generally. one ends up outside the standard set of massless supergravity theories. in all higherdimensional vertical reductions. The resolution of this diﬃculty is that in performing a KaluzaKlein reduction with an ansatz like (6.39) appears to be problematic. Fz = −ǫzy ∂y H = m . y. we have dealt only with consistent KaluzaKlein reductions. solutions obeying the restriction will also be solutions of the general unrestricted equations of motion. d = D − 2) solution has a signiﬁcant new feature: upon stacking up (D − 3) branes prior to the vertical reduction. (6.39) does not have the needed gauge symmetry. while the dilatonic scalar φ tends to its modulus value φ∞ (set to zero for simplicity in (2. The zeroform gauge potential in (6. + my . the unavoidable linear dependence of a gauge potential on the reduction coordinate given in (6. As we have seen.36) for the ﬁeld strength.39).39) From a KaluzaKlein point of view. however. consistency of any restriction means that the restriction may either be imposed on the ﬁeld variables in the original action prior to variation so as to derive the equations of motion. In order to understand this. thus producing a linear harmonic function in the transverse coordinate y.24)).37) the ﬁeld strength (6.
and hence the reduction does not spontaneously break supersymmetry. momentum in the reduction dimension. ﬁeld equations imply a condition that is inconsistent with the reduction ansatz. through its ﬁeld strength. provided A[0] is an axion. i. will be referred to as an axion. even in the 54 .39) would clearly produce an inconsistent truncation if the reduction coordinate were to appear explicitly in any of the ﬁeld equations. So.40 which used an Abelian U (1) phase symmetry acting on spinors. A zeroform gauge potential on which such a reduction may be carried out.39) turns out to be consistent after all. Requiring axionic ﬁeld strengths to be independent of the reduction coordinate amounts to extending the KaluzaKlein reduction framework so as to allow for linear dependence of an axionic zeroform potential on the reduction coordinate.e. Inconsistency of a KaluzaKlein truncation occurs when the original. If a particular gauge potential appears in the action only through its derivative. Thus. In the present case. unrestricted. When it does get diﬀerentiated. As with all examples of vertical dimensional reduction. e. gauge symmetries for some of the antisymmetric tensors will be broken. This extension of the KaluzaKlein ansatz is in fact an instance of ScherkSchwarz reduction. Unlike the original implementation of the ScherkSchwarz reduction idea.The most usual guarantee of consistency in KaluzaKlein dimensional reduction is obtained by restricting the ﬁeld variables to carry zero charge with respect to some conserved current. Instead.40. with a corresponding appearance of mass terms.41 The basic idea of ScherkSchwarz reduction is to use an Abelian rigid symmetry of a system of equations in order to generalize the reduction ansatz by allowing a linear dependence on the reduction coordinate in the parameter of this Abelian symmetry. the ∆ value corresponding to a given ﬁeld strength is also preserved. the Abelian symmetry guaranteeing consistency of (6. precisely of the form occurring in (6. retaining a linear dependence on the reduction coordinate as in (6.39). the reduction (6. Consistency is guaranteed by cancellations orchestrated by the Abelian symmetry in ﬁeldequation terms where the parameter does not get diﬀerentiated.39) is a simple shift symmetry A[0] → A[0] + const. then a consistent truncation may be achieved provided that the restriction on the gauge potential implies that the ﬁeld strength is independent of the reduction coordinate. it contributes only a term that is itself independent of the reduction coordinate. the Abelian shift symmetry used here commutes with supersymmetry.g. But this does not imply that a truncation is necessarily inconsistent just because a gauge potential contains a term linear in the reduction coordinate. occurring in the action only through its derivative. But this is not the only way in which consistency may be achieved. In the present case. pbrane solutions related by vertical dimensional reduction.
i.11. χ) of (6. account must still be taken of the ChernSimons structure lurking inside the ﬁeld strengths in (6.39). R)/SO(2).1). It may be necessary to make several redeﬁnitions and integrations by parts in order to reveal the axionic property of a given zeroform.40) Within the scalar sector (φ. so let us consider the ﬁrst possible ScherkSchwarz reduction m in the sequence of theories descending from (1. This already makes it appear that one may identify χ as an axion available for ScherkSchwarz reduction.41a) (6. (12) (1) (1) A[1] (2) A[1] (1) − ˜ (1) χF[3] (2) (6.15). However.40).41e) F[2] where the ﬁeld strengths carrying tildes are the na¨ expressions without ıve ˜ ChernSimons corrections.41d) (6.41b) − + ˜ (12) F[2] ˜ (12) F[2] ∧ ∧ F[2] (12) (2) ˜ (12) F[2] ˜ F[2] = F[2] − dχ ∧ A[1] F[1] = dχ .13) given in Section 2 to this case: L9 = √ 3 √ 7 1 1 −g R − 1 (∂φ1 )2 − 2 (∂φ2 )2 − 1 e− 2 φ1 + 2 φ2 (∂χ)2 − 48 ea·φ (F[4] )2 2 2 1 1 − 2 ea1 ·φ (F[3] )2 − 1 ea2 ·φ (F[3] )2 − 4 ea12 ·φ (F[2] )2 − 1 eb1 ·φ (F[2] )2 2 4 1 ˜ ˜ ˜ ˜ − 4 eb2 ·φ (F[2] )2 − 1 F[4] ∧ F[4] ∧ A[1] − F[3] ∧ F[3] ∧ A[3] . R)/SO(2) scalar sector of that theory.40). The scalar sector of (6. 2 (2) (12) (1) (2) (1) (2) (12) (1) (6.14. In detail. using the axion appearing in the SL(2.40) forms a nonlinear σmodel for the manifold GL(2.41c) (6. Fn] = dA[n−1] . Now the appearance of undiﬀerm A higherdimensional ScherkSchwarz reduction is possible 41 starting from type IIB supergravity in D = 10. 6. and thus to prepare the theory for a reduction like (6. φ2 ). in the rest of the Lagrangian. the dilaton vectors have the general structure given in (6.enlarged ScherkSchwarz sense. (12) (1) F[3] (2) F[3] = = = = ˜ (1) F[3] ˜ (2) F[3] ˜ (2) F[2] (2) (1) (2) ˜ (12) ˜ (1) +χF[3] ∧ A[1] − F[2] ∧ A[1] ∧ A[1] (6. preserve the same amount of unbroken supersymmetry and have the same value of ∆. This is most easily explained by an example.39 The Lagrangian for massless D = 9 maximal supergravity is obtained by specializing the general dimensionallyreduced action (6. the dilaton coupling has been made explicit. 55 .e. 6. the ﬁeld strengths are given by F[4] = (2) (1) ˜ (2) ˜ ˜ (1) F[4] − F[3] ∧ A[1] − F[3] ∧ A[1] where χ = A[0] and φ = (φ1 . starting in D = 9 where the ﬁrst axion ﬁeld appears.
entiated χ factors in (6.41) remaining unchanged. The coeﬃcient of the term linear in the reduction coordinate z has been denoted m because it carries the dimensions of mass.44) given in Ref. However. one obtains a D = 8 reduced Lagrangian n L8 ss = √ 1 −g R − 1 (∂φ1 )2 − 1 (∂φ2 )2 − 2 (∂φ3 )2 2 2 1 − 2 eb12 ·φ (∂χ − mA[1] )2 − 1 eb13 ·φ (∂A[0] − ∂χA[0] + mA[1] )2 2 1 − 2 eb23 ·φ (∂A[0] )2 − 1 ea123 ·φ (∂A[0] 2 1 − 48 ea·φ (F[4] − (2) (1) mA[2] (23) (123) 2 (3) (13) (23) (2) ) ∧ (2) A[1] ∧ (3) A[1] )2 − (1) 1 a1 · φ (F[3] )2 12 e 1 − 12 ea2 ·φ (F[3] − mA[2] ∧ A[1] )2 − (12) (13) (1) (3) (3) 1 a3 · φ (F[3] 12 e + mA[2] ∧ A[1] )2 (23) (1) (3) (1) (2) 1 1 − 4 ea12 ·φ (F[2] )2 − 1 ea13 ·φ (F[2] )2 − 4 ea23 ·φ (F[2] + mA[2] )2 4 1 1 − 4 eb1 ·φ (F[2] − mA[1] ∧ A[1] )2 − 4 eb2 ·φ (F[2] )2 − 1 eb3 ·φ (F[2] )2 4 1 − 2 m2 eb123 ·φ + mF[3] ∧ A[2] ∧ A[3] + LF F A .44) where the dilaton vectors are now those appropriate for D = 8. the axion ﬁeld χ = A[0] is now ready for application of the ScherkSchwarz reduction ansatz (6.c) become F[4] = (2) (1) ˜ (2) ˜ ˜ (1) F[4] − F[3] ∧ A[1] − F[3] ∧ A[1] (1) (2) (12) (2) (2) (1) (6. (12) After the ﬁeld redeﬁnitions (6. Applying (6.39). (1) (1) (1) (2) (3) (2) (6.43a) (6.39 nI 56 . am grateful to Marcus Bremer for help in correcting some errors in the original expression of Eq.c) makes it appear that a ScherkSchwarz reduction would be inconsistent.42) F[3] (2) = (1) (1) (12) ˜ (2) F[3] + F[2] ∧ A[1] + dχ ∧ A[2] .43c) the rest of (6.42).39) to the D = 9 Lagrangian. and correspondingly its eﬀect on the reduced action is to cause the appearance of mass terms.41a. one may eliminate these undiﬀerentiated factors by making the ﬁeld redeﬁnition A[2] −→ A[2] + χA[2] .41a. after which the ﬁeld strengths (6. ˜ −dχ ∧ A[2] ∧ A[1] − F[2] ∧ A[1] ∧ A[1] (1) (2) (6. (6. the term LF F A contains only mindependent terms.
or by further ScherkSchwarz generalizations of the KaluzaKlein reduction process.44) that the ﬁelds A[1] . A[0] and A[1] . A[1] and (1) (3) (2) D Naxions 9 1 8 4 7 10 6 20 5 36 4 63 Each of these axions gives rise to a distinct massive supergravity theory upon ScherkSchwarz reduction. once a ScherkSchwarz reduction step has been performed.and A[1] may be absorbed. In addition. there are numerous possibilities of performing ScherkSchwarz reduction simultaneously on a number of axions. These three transformations are accompanied. there are ﬁeld redeﬁnitions under which the ﬁelds χ. (23) (13) A[0] and A[1] may be gauged to zero. This can be done either by arranging to cover a number of axions simultaneously with derivatives. In addition.39. we refer the reader to Refs. Owing to the presence of these local shift terms in the three Stueckelberg symmetries. Moreover.39 and each of these reduced theories has its own pattern of mass generation.44) as well as by mdependent (23) (13) shift transformations of χ. The numbers of axions are given in the following Table: Table 1: Supergravity axions versus spacetime dimension. One way to see how this absorption happens is to notice that the action obtained from (6. the resulting theory can be further reduced using ordinary KaluzaKlein reduction. by various compensating transformations necessitated by the ChernSimons corrections present in (6. A[1] and A[2] have become (13) (3) (2) (1) zero.42 For further details on the panoply of ScherkSchwarz reduction possibilities. massive. the ScherkSchwarz and ordinary KaluzaKlein processes do not commute. After gauging these three ﬁelds to A[2] .44) for the ﬁelds A[1] . however. As one descends through the available spacetime dimensions for supergravity theories.42 The singlestep procedure of ScherkSchwarz dimensional reduction described above may be generalized to a procedure exploiting the various coho57 . respectively. the ﬁelds χ. Moreover. A[1] and A[2] transform according to their standard gauge transformation laws. the number of axionic scalars available for a ScherkSchwarz reduction step increases. A[0] (23) It is apparent from (6. so the number of theories obtained by the various combinations of ScherkSchwarz and ordinary dimensional reduction is cumulative. one has a clean set of mass terms in (6.44) has a set of three Stueckelbergtype (1) (2) (3) gauge transformations under which A[1] .
44). the recognition that one can use any of the H n (K. For example. reproducing our earlier singlestep reduction (6. with no dependence on any of the zi coordinates. In this example. R). R) cohomology classes of the compactiﬁcation manifold K signiﬁcantly extends the scope of the generalized reduction procedure.46) Since the rank of the form here is n = 0. it allows one to make generalized reductions on manifolds such as K3 or on CalabiYau manifolds. which is not very sensible. z) = ω[3] + A[3] (x. All of the other ﬁelds are reduced using the standard KaluzaKlein ansatz.mology classes of a multidimensional compactiﬁcation manifold. 2 2 (6.39).44). R). As another example. then making an S 1 singlestep generalized ScherkSchwarz reduction (6. but 58 . The same theory (up to ﬁeld redeﬁnitions) can also be obtained 39 by ﬁrst making an ordinary KaluzaKlein reduction from D = 11 down to D = 8 on a 3torus T 3 . in the case of a singlestep generalized reduction on a circle S 1 . analogous to the D = 8 theory (6. y. R). (6. The theory resulting from this T 4 reduction is a D = 7 massive supergravity with a cosmological potential. whose exterior derivative Ω[n] = dω[n−1] is an element of the cohomology class H n (K. Although the T 4 reduction example simply reproduces a massive D = 7 theory that can also be obtained via the singlestep ansatz (6. one has Ω[1] = mdz ∈ H 1 (S 1 .45) where ω[n−1] is an (n − 1) form deﬁned locally on K.1) by letting the rank n of the ﬁeld strength take the value zero. y) with Ω[4] = dω[3] = mdz1 ∧ dz2 ∧ dz3 ∧ dz4 ∈ H 4 (T 4 .39) may be generalized to A[n−1] (x.39) from D = 8 to D = 7.43 The key to this link between the ScherkSchwarz generalized dimensional reduction and the topology of the internal KaluzaKlein manifold K is to recognize that the singlestep reduction ansatz (6. by consistent truncation of (6.39). setting A[3] (x. (6. z) = ω[n−1] + A[n−1] (x.43 For our present purposes. the important feature of theories obtained by ScherkSchwarz reduction is the appearance of cosmological potential terms such as the penultimate term in Eq. one may arrive at the simple Lagrangian L= √ −g R − 1 ∇M φ∇M φ − 1 m2 eaφ . one may choose to write ω[3] locally as ω[3] = mz1 dz2 ∧ dz3 ∧ dz4 . consider a generalized reduction on a 4torus T 4 starting in D = 11. Such terms may be considered within the context of our simpliﬁed action (2. Accordingly. For example. y) . y.44) or of one of the many theories obtained by ScherkSchwarz reduction in lower dimensions. the elementary/electric type of solution would have worldvolume dimension d = −1.
with nonzero ﬁeld strengths in the reduction directions.” we are excluding the case aφ → −∞).47a) tends to zero at large values of y. Moreover.47.47) in supergravity theories were found for the D = 4 case in Ref. This latter singularity can be avoided by taking H to be H = const. The domainwall solution (6.45 o Domain 59 . With the choice (6.48). 6. as in our example (6. This brings us back to the other facets of the consistency problem for vertical dimensional reduction down to (D − 2)branes as discussed in subsection 6. taking the corresponding p = D − 2 brane solution from (2. but it diverges if H tends to zero. + M y (6. in accordance with (6.48). 6.18) gives ∆ = a2 − 2(D − 1)/(D − 2).o The curvature of the metric (6.˜ the solitonic/magnetic solution has d = D − 1.44 and a review of them has been given in Ref.47. allowing at least asymptotic ﬂatness for this solution. corresponding to the discontinuity in the gradient of H. there is just a deltafunction singularity at the location of the domain wall at y = 0. such domain walls can be oxidized back to solutions of higherdimensional massless supergravities. corresponding to a p = D − 2 brane. because these domain walls exist only in massive supergravity theories like (6. within a theory that does not naturally admit ﬂat space as a solution (by “naturally.44). There is no inconsistency between the existence of domainwall solutions like (6. Because the ScherkSchwarz generalized dimensional reduction used to obtain them was a consistent truncation. as expected. walls solutions such as (6. manages to “cancel” this potential at transverse inﬁnity.48) has the peculiarity of tending asymptotically to ﬂat space as y → ∞. 6. however. owing to the complication of the cosmological potential.4.48) and the inability to ﬁnd such solutions in standard supergravity theories.47b) . they have the form of stacked solutions prepared for vertical reduction. The domainwall solution (6.37).46) does not even admit a nonﬂat maximallysymmetric solution. or domain wall.47a) (6.47.46) to the reductioninvariant parameter ∆ by the standard formula (2. the theory (6. one ﬁnds ds2 = H ∆(D−2) ηµν dxµ dxν + H ∆(D−2) dy 2 e =H φ −2a/∆ 4 4(D−1) (6. or with the conicalspacetime character of (D − 3)branes. where the harmonic function H(y) is now a linear function of the single transverse coordinate.48) √ 1 where M = 2 m ∆.38). with a vacuum structure diﬀerent from that of standard massless supergravities. but in that case. Relating the parameter a in (6.24).
and the net charge density will be U = 2λΩD−d−1 /4.mn p yp . it is natural to try to ﬁnd solutions where α the charge centers for the diﬀerent F[n] are separated.2) ... . When the charge centers have coalesced. Thus... the λα are related to the integration ˜ constants k α appearing in the metric by k α = λα /d. the solution for the metric and the active dilatonic combinations eςaα ·φ is given by ℓ ds 2 = α=1 2 Hα Hα ℓ ˜ −d D−2 ℓ dx dxµ + α=1 ˜ −dd µ D−2 Hα dy m dy m d eςaα ·φ Hα = = D−2 Hβ (7.1) In both the electric and the magnetic cases. so E = E√+ E2 = 1 2λΩD−d−1 /4 = 2U . even with the multiple centers. Thus.. Since the charges add vectorially. in the magnetic case.. y − yα n+1 (7. with centers yα . . β1 1+ kα ˜ y − yα d The nontrivial step in verifying the validity of this solution is the check that the nonlinear terms still cancel in the Einstein equations.7 7. for example. while the charge centers are allowed to coalesce.. λα = λ. F[n] ) in which the two charge parameters are taken to be the same. On the other hand.. .46 This will lead us to a better understanding of the ∆ = 4 solutions shown in Figure 7.1 Intersecting branes and scattering branes Multiple component solutions α Given the existence of solutions (6.mn = λα ǫm1 . one sets α Fm1 . F[n] ). the resulting solution may be viewed as a singleﬁeldstrength solution for a ﬁeld 1 2 strength rotated by π/4 in the space of ﬁeld strengths (F[n] . the net charge parameter in this case will be λ = √ √ 2λ. ℓ. . Letting ς = ±1 in the electric/magnetic cases as before.46 1 2 Now consider a solution with two ﬁeld strengths (F[n] . the total √ mass density will add as a scalar quantity.24) with several active ﬁeld strengths F[n] . but with coincident charge centers. Let the charge parameter for F α be λα . Consider a number of ﬁeld strengths that individually have ∆ = 4 couplings. the coalesced solution satisﬁes E = 2U/ ∆ with 60 . α = 1. but now look for a solution where ℓ of these ﬁeld strengths are active.
A2 .3) which depends on two independent harmonic functions H1 (y) and H2 (y). Upon oxidizing the twoblackhole solution back to D = 11.13) may automatically be interpreted as solutions of any one of the higherdimensional theories descending from the D = 11 theory (1.2) to the dimensionally reduced theory (6. they may be considered to be “bound states at threshold. supported by a 1form gauge potential A12 descending from the D = 11 [1] gauge potential A[3] and also by another 1form gauge potential.2 Intersecting branes and the four elements in D = 11 The multiplechargecenter solutions (7. it is not guaranteed that these oxidized branes always fall into the class of isotropic pbrane solutions that we have mainly been discussing.2). . in D = 9. one ﬁnds the solution −1 ds2 = H13 (y) H1 (y){−dt2 +dρ2 +dσ 2 +(H2 (y)−1)(dt+dρ)2 }+dy mdy m 11 −1 A[3] = H1 (y)dt∧dρ∧dσ . Direct comparison with our general pbrane solution (2.2) leaves unbroken 1/4 of the original supersymmetry.”46 We shall shortly see that the zeroforce property of such multiplecomponent solutions is related to their managing still to preserve unbroken a certain portion of rigid supersymmetry. Although all lowerdimensional solutions may automatically be oxidized in this way into solutions of higherdimensional supergravity theories. For example. For example.24) shows that the coalesced solution agrees precisely with the singleﬁeldstrength ∆ = 2 solution. Since the ∆ = 4/N solutions may in this way be separated into ∆ = 4 components while still preserving some degree of unbroken supersymmetry. 1 m = 3.∆ = 2. [1] emerging from the metric as a KaluzaKlein vector ﬁeld. wave 2brane (7. . one has a twoblackhole solution of the form (7. This automatic “oxidation” is possible because we have insisted throughout on considering only consistent truncations.g. Although the solution (7. even though this portion is reduced with respect to the halfpreservation characterizing singlecomponent ∆ = 4 solutions. one progressively breaks more and more supersymmetry. . and without producing any relative forces to disturb their equilibrium. 10.1). 7. Generalizing this construction to a case with N separate ∆ = 4 components. In the next subsection. . one ﬁnds in the coincident limit a ∆ = 4/N supersymmetric solution from the singleﬁeldstrength analysis. where the y m are an 8dimensional set of “overall transverse” coordinates. it nonetheless has 61 . e. the above solution (7. each one separately charged with respect to a diﬀerent ∆ = 4 ﬁeld strength. we shall see that as one adds new components.3) clearly falls outside the class of pbrane or multiple pbrane solutions that we have considered so far.
one recovers ds2 11 A[3] = = H 3 (y) H −1 (y){−dt2 + dρ2 + dσ 2 + dy m dy m H −1 (y)dt ∧ dρ ∧ dσ.3) is that the charge centers of the two harmonic functions H1 and H2 may be chosen completely independently in the overall transverse space. setting H2 = 1. . The term “intersecting” is generally taken 62 .3). 10.3) need not actually overlap on any speciﬁc subspace of spacetime. it should be understood that the two components of (7. H(y) ↔ H2 (y) is constant in one of these 9 directions. . Of course. . 0.47 the pp wave: ds2 11 A[3] = = {−dt2 + dρ2 + (H(y) − 1)(dt + dρ)2 } + dy m dy m m = 2.3). on the other hand. . it is not a Poincar´invariant hyperplane e solution). Owing to the fact that the harmonic function H2 (y) depends only on the overall transverse coordinates y m .2). 10. the wave is actually “delocalized” in the third membrane worldvolume direction.5) where for a general wave solution. generalized to an arbitrary harmonic function H(y) ↔ H1 (y) in the membrane’s transverse space.two clearly recognizable elements. Another point to note about (7. m = 3. Thus. σ} coordinates down to a D = 9 conﬁguration of two particles of the sort considered in (7.2). 10. the solution (7.3) has already been stacked up in the σ direction as is necessary in preparation for a vertical dimensional reduction. associated to the two harmonic functions H1 (y) and H2 (y).3) on the {ρ.3).3) is independent of σ as well as of its own lightplane coordinates. .e. . .3) thus may be viewed as a D = 11 pp wave superposed on a membrane. . 2brane (7. although this is an example of an “intersecting” brane conﬁguration. . . pp wave (7. m = 3. we may use the freedom to trivialize one or the other of these harmonic functions by setting it equal to unity. produces a solution of D = 11 supergravity that is not a pbrane (i. What one ﬁnds for H1 = 1 is a classic solution of General Relativity found originally in 1923 by Brinkmann.e. Thus. i.e. .4) 1 which one may recognized as simply a certain style of organizing the harmonicfunction factors in the D = 11 membrane solution 18 (3. i. . for the speciﬁc case obtained by setting H1 = 1 in (7. H(y) could be harmonic in the 9 dimensions y m transverse to the two lightplane dimensions {t. Setting H1 = 1 in (7. ρ} in which the wave propagates. In order to identify these two elements. this delocalization of the wave in the σ direction is just what makes it possible to perform a dimensional reduction of (7. corresponding to the coordinate σ in (7. The solution (7. the wave in (7.
3. a familiar fourdimensional Euclidean gravitational instanton. Thus. . giving rise after compactiﬁcation to a magneticallycharged particle in D = 4 dimensions. For a singlecenter harmonic function H(y) = 1 + k/(y). The solution (7. In the case of the original KaluzaKlein monopole. so after reduction on the ψ coordinate one has a magneticallycharged 6brane solution in D = 10. but this becomes a mere coordinate singularity. similar to that for ﬂat space in polar coordinates. . providing the coordinate ψ is periodically identiﬁed. . . + dx5 } + dy m dy m 1 = ∗dH(y).48 A very striking feature of the family of multiplecomponent pbrane solutions is that their oxidations up to D = 11 involve combinations of only 4 basic “elemental” D = 11 solutions. the solution (7. and not of the coordinate ψ. Generally. Two of these we have just met in the oxidized solution (7. m = 6. .3): the membrane and the pp wave. The relation ∆ψ = 4πk between the compactiﬁcation period of ψ and the chargedetermining integration constant k in the harmonic function H of the solution (7.49 the starting solution had 4+1 dimensions. 2. The KaluzaKlein monopole oxidized up to D = 11 is the solution 2 ds11 A[3] 2 dsTN = 0 = −dt2 + dx2 + . after reduction on ψ.7) has a conical singularity on the hyperplane o y i = 0. i = 1. The two others are the “duals” of these: the 5brane 25 and a solution describing the oxidation to D = 11 of the “KaluzaKlein monopole. . .6) where the H(y) is a general harmonic function in the 5dimensional transverse space. x6 . which plays a special rˆle.7) is a function only of the 3 coordinates y i . the TaubNUT solution naturally invites interpretation as a compactiﬁed solution in one less dimension. .to mean that there are shared worldvolume coordinates. . ρ} overlap between the membrane worldvolume and the lightplane coordinates. in this case the {t.4): ds2 11 F[4] 2 = H 3 (y) H −1 (y){−dt2 + dx2 + .7a) i 2 = H(y)dy dy + H = ∇H . 5brane 2 (7. The harmonic function H in (7. (y)(dψ + Vi (y)dy ) . (7.7) gives rise to a quantisation condition at the quantum 63 .”49 The 5brane may be written in a style similar to that of the membrane (7.7) has an additional 6 spacelike worldvolume dimensions x1 .7b) ∇×V TaubNUT 2 where dsTN is the TaubNUT metric. 10. . + dx2 + ds2 (y) 1 6 TN i i −1 (7. . the appropriate identiﬁcation period for ψ is 4πk. .
4. the pp wave solution (7.8) where P01 is again a projection operator with half of its eigenvalues zero. In terms of the electric and magnetic charges U and V of the dimensionally reduced particle and 6brane. Now let us apply the projectionoperator analysis to the wave 2brane solution (7. We have already seen this for the membrane solution in subsection 4.3.4 and shall instead concentrate just on the projection conditions that must be satisﬁed by the surviving rigid supersymmetry parameter ǫ∞ . This is precisely of the form expected for a Dirac charge quantisation condition. This quantisation condition is nothing other than an ordinary quantisation of momentum for Fourier wave components on a compact space. Analogously to our earlier abbreviated discussion using just the supersymmetry algebra.23).level involving the magnetic charge of the dimensionallyreduced D = 10 6brane descending from (7. P01 ] = 0 . Supersymmetry preservation in a membrane background oriented parallel to the {012} hyperplane requires the projection condition P012 ǫ∞ = 0 (4. Imposing these two conditions simultaneously is consistent because these projectors commute.3). We shall skip over points 1) and 2) of the discussion analogous to that of subsection 4. As another example.5) propagating in the {01} directions of spacetime. while supersymmetry preservation in a pp wave background with a {01} lightplane requires P01 ǫ∞ = 0.3) can also preserve some portion of unbroken rigid supersymmetry. one ﬁnds U V = 2πκ2 n.4 – 7.5). 1. one may consider the supersymmetry preservation conditions for the pp wave solution (7. In Section 8 we shall return to the subject of charge quantisation conditions more generally for the charges carried by pbranes.5). half unity. l (7. with normalization to unit length along the wave’s propagation direction: 1 {Qα .9) 1 Since tr(P012 P01 ) = 4 · 32. 64 . the imposition of both projection conditions on ǫ∞ 1 cuts the preserved portion of rigid D = 11 supersymmetry down to 4 . Let us now return to the question of supersymmetry preservation and enquire whether intersecting branes like (7.5) preserves half of the D = 11 rigid supersymmetry.7) and the electric charge of the extreme black hole particle obtained by reducing the pp wave (7. All four of the elemental D = 11 solutions (7. with n ∈ Z (where κ2 occurs 10 10 because the charges U and V as deﬁned in (1. [P012 .7) preserve half the D = 11 rigid supersymmetry. Qβ } = 2EP01 length 1 P01 = 2 (1 + Γ01 ) . (7. Consequently. consider this algebra in the background of a pp wave solution (7. in this case the compact ψ direction.4) are not dimensionless).
. Having established this coordinate classiﬁcation. Thus. The solution is ds2 −1 −1 3 H1 (y)H23 (y) H1 (y)H2 (y)(−dt2 + dx2 ) 1 −1 −1 2 2 +H1 (y)(dx2 ) + H2 (y)(dx2 + . the general structure of the intersecting brane metric is as follows. For each element. By considering special cases where H2 = 1 or H1 = 1. For a twoelement solution. one identiﬁes the membrane and 5brane elements of the solution (7. these are accordingly called “overall worldvolume” directions.10) illustrate the general structure of intersectingbrane solutions. so there is not necessarily a true string overlap). 10 F2mnp = −ǫmnpq ∂q H2 . by which one means the directions transverse to one element’s worldvolume but belonging to the worldvolume of the other element.Now. . that the overalltransverse charge centers of H1 and H2 need not coincide.10). Considering this “intersection” to be a string (but recall. . the directions {2. as before. independent of) the “relative transverse” directions.e. 5brane pair. . one puts an overall conformal 65 . . Note that both the membrane and 5brane elements share the worldvolume directions {01}. . two relative transverse sectors and the overall transverse sector. 6} for the solution (7. the overall worldvolume coordinates are the W25 coordinates and the overall transverse coordinates are the T25 coordinates. 3.10): 0 1 2 3 4 5 6 x x x x x x x x x W25 W2T5 W5T2 7 8 9 10 T25 The character of each coordinate is indicated in this sketch: W2 and W5 indicating worldvolume coordinates with respect to each of the two elements and T2 and T5 indicating transverse coordinates with respect to each of the two elements. The forms of the wave 2brane solution (7. there are four sectors among the coordinates: overall worldvolume.e. containing as elements a D = 11 membrane. the harmonic functions H1 (y) and H2 (y) depend only on the overall transverse coordinates.10a) (7. however. these elements are delocalized in (i. . .10). One may make a sketch of these relations for the 2 ⊥ 5(1) solution (7.3) and the 2 ⊥ 5(1) solution (7. let’s consider another example of an intersectingbrane solution.10b) +dy m dy m Fm012 = −1 ∂m (H1 ) m = 7. . .10) is denoted 2 ⊥ 5(1).3). where as in the waveonamembrane solution (7. i. the solution (7. + dx6 ) 3 1 2 = 2 ⊥ 5(1) (7..
Physically. P013456 ] = 0. and in addition one needs to establish which combinations of the D = 11 elements may be present in a given solution.11a) 66 . scattering branes and modulus σmodel geometry d/(D−2) The existence of static conﬁgurations such as the wave 2brane solution (7.20) arising in the process of solving the supergravity equations subject to the pbrane ans¨tze (2. or to the fact that one brane couples to the background supergravity ﬁelds with a conformal factor that wipes out the eﬀects of the other brane. a The Laplace equation has the wellknown property of admitting multicenter solutions. because 4 1 tr(P012 P013456 ) = 4 · 32.1) in order to treat the limiting problem of a light brane probe moving in the background of a heavy brane.3–2. (6. For each of the two elements. because [P012 .3 Brane probes.10). These may be consistently imposed at the same time. which we have already encountered in Eq. This pattern has been termed the harmonic function rule.3) or the 2 ⊥ 5(1) intersectingbrane solution (7.51 In this limit.6). the probe action is Iprobe = −Tα dp+1 ξ (− det(∂µ xm ∂ν xn gmn (x)) 2 e 2 ς 1 1 pr aα · φ + Qα ˜ Aα [p+1] (7. one may ignore the deformation of the heavybrane background caused by the light brane.50 For now.10). similarly to our discussion of the wave 2brane solution. one has a projection condition on the surviving rigid supersymmetry parameter ǫ∞ : P012 ǫ∞ = 0 for the membrane and P013456 ǫ∞ = 0 for the 5brane. 7. The use of the braneprobe coupling is a simple way to approximately treat timedependent brane conﬁgurations.18. For a fuller review on this subject. The amount of surviving supersymmetry in the 2 ⊥ 5(1) solution is 1 .10) derives from the properties of the transversespace Laplace equation (2. let us just check point 3) in the supersymmetrypreservation analysis for the 2 ⊥ 5(1) solution (7. and then in addition one puts a facfactor Hi tor Hi−1 (y) in front of each dx2 term belonging to the worldvolume of the ith element. One may verify this pattern in the structure of (7.(y) for the whole metric. In order to see such cancellations more explicitly.48 This summary of the structure of intersecting brane solutions does not replace a full check that the supergravity equations of motion are solved.27). we refer the reader to Ref. the existence of such multicenter solutions corresponds either to a cancellation of attractive gravitational and dilatonic forces against repulsive antisymmetrictensor forces. For a pbrane probe of this sort coupled to a Ddimensional supergravity background. one may use a source coupling analogous to the D = 11 bosonic supermembrane action (4.
If one takes the form of the heavymembrane background from the electric ansatz (2. Requiring that the source match correctly to a pbrane (probe) solution de1 pr mands the presence of the dilaton coupling e 2 ς aα ·φ . 2. 1. the conformal frame in which the Ddimensional EinsteinHilbert action is free from dilatonic scalar factors.13) Recalling the condition (2. to which the p brane probe couples. 2. (7. which becomes just e3A(y) = eC(y) for the membrane background. (7.15) i.12) to order (∂y)2 in the probe velocity. the braneprobe action (7. then the bosonic probe action becomes Iprobe = −T d3 ξ − det(e2A(y) ηµν + e2B(y) ∂µ y m ∂ν y m ) − eC(y) .18 In this case. the membraneprobe σmodel metric is ﬂat.1). µ = 0. recalling the supersymmetrypreservation condition (2.11a) occurs because one needs to have the correct source for the Ddimensional Einstein frame. one may take a light D = 11 membrane probe in the background of a parallel and similarlyoriented heavy membrane. the probe σmodel metric in (7.e. (7.3.11) becomes just the D = 11 supermembrane action (4.11b) The dilaton coupling in (7.˜ Aα [p+1] = [(p + 1)!]−1 ∂µ1 xm1 · · · ∂µp+1 xmp+1 Aα 1 ···mp+1 dξ µ1 ∧ · · · ∧ dξ µp+1 . one ﬁnds at order (∂y)0 the eﬀective potential Vprobe = T (e3A(y) − eC ) .14) but. conﬁrming the absence of static forces between the two membrane components. and if one chooses the “static” worldvolume gauge ξ µ = xµ . (7. where ς pr = ±1 according to whether the pbrane probe is of electric or magnetic type and where aα is the dilaton vector appearing in the kineticterm dilaton coupling in (6. one has directly Vprobe = 0.14) reduces simply to γ mn = eA(y)+2B(y) δ mn = δ mn .22). 67 .12).12) Expanding the square root in (7..4). one has the eﬀective probe σmodel Iprobe = − (2) T 2 d3 ξe3A(y) e2(B(y)−A(y) ∂µ y m ∂ν y m ηµν . i. [p+1] As a simple initial example of such a brane probe. Continuing on in the expansion of (7.e.17) characterizing the heavymembrane background. m (7.13) for the gauge potential Aα .
(7. 2 N = 8 probeworldvolume supersymmetry. precisely as one ﬁnds in (7. The above testbrane analysis for two D = 9 black holes is conﬁrmed by a more detailed study of the lowvelocity scattering of supersymmetric black holes performed by Shiraishi. Now repeat the braneprobe analysis for the twoblackhole conﬁguration.11). for the case of two black holes in D = 9.12). which we found in subsection 4. This high degree of surviving supersymmetry is too restrictive in its constraints on the form of the σmodel to allow for anything other than a ﬂat metric. with metric γ mn = Hback (y)δ mn . this now gives Vprobe = eA e 2 7 φ. after dimensional reduction down to D = 9 dimensions. At order (∂y)0 . but this potential turns out to be just a constant because the heavybrane background satisﬁes 3 A = 2√7 φ. Continuing on with the expansion of (7. Thus. Again expand the determinant of the induced metric 3 − √ in (7. we conﬁrm the expected static zeroforce condition for the 1 4 supersymmetric twoblackhole conﬁguration descending from the wave 2brane solution (7.The ﬂatness of the membraneprobe σmodel metric (7.3). one now obtains a nontrivial probe σmodel. Proceeding on to (velocity)2 order. Now consider a braneprobe conﬁguration with less surviving supersymmetry.52 The procedure is a standard one in soliton physics: one promotes the moduli of a static solution to timedependent functions and then substitutes the resulting generalized ﬁeld conﬁguration back 68 .15). one ﬁrst ﬁnds a nontrivial interaction between the probe and the heavy membrane background at order (∂y)4 (odd powers being ruled out by timereversal invariance of the D = 11 supergravity equations). D = 9 vector ﬁelds: one descending from the D = 11 3form gauge potential and one descending from the metric.15) accords precisely with the degree of rigid supersymmetry that survives in the underlying supergravity solution with two parallel. which in the present case just becomes ξ 0 = t.4 to have 1 · 32 components. a system of two black holes supported by diﬀerent ∆ = 4.16) where Hback is the harmonic function controlling the heavy brane’s background ﬁelds.e. again choosing a static gauge on the probe worldvolume. and with correspondingly weaker constraints on the probe worldvolume σmodel. i.11) that wipes out the conformal factor occurring in the heavy brane background metric. This zeroforce condition arises not so much as a result of a cancellation between diﬀerent forces but as a result of the probe’s coupling to the background with a dilatonic factor in (7. Corresponding to the wave 2brane solution (7. one has. d = 2 + 1. the harmonic function Hback has the structure (1 + k/r6 ). similarly oriented D = 11 membranes.3).
16) as that found in the braneprobe analysis above. . Now we should resolve a puzzle of how this nontrivial d = 1 scattering modulus σmodel turns out to be consistent with the surviving supersymmetry.into the original ﬁeld equations. . In order to have extended supersymmetry in (7. γ.19a) (7. where hab is a ﬁbre metric and ∇t is constructed using an appropriate connection for the ﬁbre corresponding to the ψ a . simpliﬁes dramatically in cases corresponding to the scattering of supersymmetric black holes.53 The modulus variables of the twoblackhole system are ﬁelds in one dimension. and then requires these transformations to close to form the N = 2 algebra (7. with 1 Lagrangian − 2 hab ψ a ∇t ψ b . γ is the metric on M and A[3] is a 3form on M which plays the rˆle of torsion in o (+) (+) (+) the derivative operator acting on fermions. one starts by positing a second set of supersymmetry transformations of the form δxi = ηI i j Dxj . the above pair of D = 9 black holes. N . In this way. The system of equations. one obtains the equations I2 i Njk i I[j.18) One may additionally 54 have a set of spinorial N = 1 superﬁelds ψ a .19b) 69 . however. These twobody forces may be derived from an eﬀective action involving the position vectors of the two black holes. QJ } = 2δ IJ H I = 1. This leads to a set of diﬀerential equations on the modulus variables which may be viewed as eﬀective equations for the moduli. θ) (where xi (t) = xi  . . where M is the Riemannian σmodel manifold. .g. time.17) ˆ where H is the Hamiltonian. A[3] ). the resulting system of diﬀerential equations may be quite complicated.17). ∇t = ∂t xi ∇i .e. In the general case of multiple black hole scattering. except for a rescaling which replaces the braneprobe mass by the reduced mass of the twoblackhole system. However. where the result turns out to involve only 2body forces. The N extended supersymmetry algebra in d = 1 is ˆ {QI . i. then one also requires that the action (7. (7. λi (t) = Dxi  ) as θ=0 θ=0 1 I = −2 dtdθ(iγij Dxi 1 d j x + Aijk Dxi Dxj Dxk ) .k] ≡ = −1 l = 0 (7. dt 3! (7. The σmodel action may be written using N = 1 superﬁelds xi (t. one obtains the same modulus metric (7. in the present case we shall not include this extra superﬁeld.18) be invariant. e. A d = 1.18). where ∇i λj = 1 j ∇i λj + 2 A ik λk . N = 1 σmodel is speciﬁed by a triple (M. Separating the centerofmass motion from the relative motion.
where the octonion multiplication rule is ea eb = −δab + ϕab c ec . Continuing on to N = 8. with (Ia )8 b = δab . The structure of the conditions (7.19).19) is more complicated than might have been expected.19) warrant a diﬀerent notation for this wider class of models.18). d = 1 supersymmetric theories satisfying (7.53 Such models are characterized by a K¨hler a geometry with torsion.19d) (7. just real singlecomponent objects.19) are analogous to (2.0) chiral supersymmetric theories in d = 1 + 1. Models satisfying such conditions have an “octonionic K¨hler geometry with torsion.19c) (7. even though the d = 1 + 1 minimal spinors are. the d = 1 + 1 theory implies a “stronger” condition. one ﬁnds an 8b generalization 53 of the conditions (7. a = 1.e. .53 a Now.b) follow from requiring the closure of the algebra (7. Conditions (7.19ce) follow from requiring invariance of the action (7. (7. Certainly.” and are called OKT models. d = 1 supersymmetry. with I i j as its complex structure. .19). the solution already contains a pp wave element (so that we have a D = 10 “waveonastring” solution). (Ia )b c = ϕa b c .γkl I k i I l j ∂[i (I m j Amkl] ) = γij = 0 = 0.b) imply that M is a complex manifold. the d = 1 extended supersymmetry conditions are “weaker” than those obtained by dimensional reduction from d = 1 + 1. Experience with d = 1 + 1 extended supersymmetry 54 might have lead one to expect.19e) − 2I m [i ∂[m Ajkl]] ) (+) ∇(i I k j) where (7. are there any nontrivial solutions to these conditions? Evidently. this is reﬂected in the circumstance that after even one dimensional oxidation from D = 9 up to D = 10. 7. . Conversely.19a.17) and (7.19c–e). one may call them 2b supersymmetric σmodels. from the braneprobe and Shiraishi analyses. by simple dimensional reduction. the diﬀerence is explained by d = 1 + 1 Lorentz invariance: not all d = 1 theories can be “oxidized” up to Lorentzinvariant d = 1 + 1 theories. but the weaker conditions (7. there must be. the E8 transverse space is of Euclidean signature. p We 70 . but the converse is not true. with 7 independent complex structures built using the octonionic structure constants p ϕab c : δxi = η a Ia i j Dxj .19a. with a lightplane metric that is not Poincar´ invariant. e Note also that the d = 1 “torsion” A[3] is not required to be closed in (7. (Ia )b 8 = −δ b a . In the present case with two D = 9 black holes. as in d = 1. solutions of this condition also satisfy (7. just the condition (+) ∇i I j k = 0. For our two D = 9 let the conventional octonionic “0” index be replaced by “8” here in order to avoid confusion with a timelike index. i.
y1 − y2 6 (7. higher ranks having been eliminated in the reduction or by dualization. Such nonlinearlyrealized symme71 . N = 8 supersymmetry algebra relevant to the (selfconjugate) supergravity multiplet. Since this complex representation can be carried only by the complex ﬁeldstrength combinations and not by the 1form gauge potentials.black holes with a D = 8 transverse space. Thus we recover the familiar dependence of pbrane solutions on transversespace harmonic functions. Then.e. for example. Using twocomponent notation for the D = 4 spinors. this is also the automorphism symmetry of the D = 4. where Ω is a 4form on E8 . In D = 4. a nonlinearlyrealized E7(+7) symmetry also appears as an invariance of the D = 4. these [ij] ¯ ˙ combinations transform as Fαβ and Fαβ [ij] . Cremmer and Julia 55 deduced that it had to be the manifold E7(+7)/SU(8). 8 Duality symmetries and charge quantisation As one can see from our discussion of KaluzaKlein dimensional reduction in Section 6. one ﬁnds that they form a rather impressive nonlinear σmodel with a 70dimensional manifold. where only an SO(8) symmetry is apparent. Correspondingly. it cannot be locally formulated at the level of the gauge potentials or of the action. Anticipating that this manifold must be a coset space with H = SU(8) as the linearlyrealized denominator group.19d) one learns Ω8abc = ϕabc and Ωabcd = −∗ ϕabcd . Taking all the spinzero ﬁelds together. since the dimension of E7 is 133 and that of SU(8) is 63. which are the highestrank ﬁeld strengths occurring in D = 4. . When one reaches D = 4. In formulating this symmetry. as a complex ˙ 28dimensional dimensional representation of SU(8). N = 8 maximal supergravity equations of motion. 8. from the 8b generalization of condition (7. a total of 70 such spinzero ﬁelds has accumulated. progression down to lower dimensions D causes the number of dilatonic scalars φ and also the number of zeroform potentials of 1form ﬁeld strengths to proliferate. from the 8b generalization of condition (7. the maximal (N = 8) supergravity equations of motion have a linearlyrealized H = SU(8) symmetry. one may start from the ansatz ds2 = H(y)ds2 (E8 ). and we reobtain the braneprobe or Shiraishi structure of the blackhole modulus scattering metric with Hrelative = 1 + kred . i. j = 1. this gives a 70dimensional manifold.20) where kred determines the reduced mass of the two black holes. it is necessary to consider complex selfdual and antiselfdual combinations of the 2form ﬁeld strengths. Aijk = Ωijk ℓ ∂ℓ H.19e) one learns δ ij ∂i ∂j H = 0. . i. . .
constantparameter) YangMills gauge transformations. In all cases. R) and the isotropy group is H = SO(3)×SO(2). the group of rigid (i.g. these moduli acquire interpretations as the coupling constants and vacuum θangles of the theory. the spinzero ﬁelds take their values in “target” manifolds G/H . maximal supergravity has the sets of σmodel nonlinear G and linear H symmetries shown in Table 2. or isotropy group. In dimensions 4 ≤ D ≤ 9. R) × SL(2. from which supergravities arise by dimensional reduction. In string theory. one sees from Table 2 that G = SL(3. R) SO(5. 8. e. so this is the classical “internal” classifying symmetry for multiplets of supergravity solutions. We have an (11−3 = 8) 72 . Table 2: Supergravity σmodel symmetries. Once these are determined for a given “vacuum. or “vacuum” spacetime with respect to which integrated charges and energy/momentum are deﬁned. Just as the asymptotic value at inﬁnity of the metric deﬁnes the reference. 5) E6(+6) E7(+7) E8(+8) H SO(2) SO(3) × SO(2) SO(5) SO(5) × SO(5) USP(8) SU(8) SO(16) The isotropy group of any point on a coset manifold G/H is just H.” These asymptotic values are referred to as the moduli of the solution. mixing both ﬁelds arising from the metric and also from the reduction of the D = 11 3form potential A[3] in (1. R) SL(3. so do the asymptotic values of the spinzero ﬁelds deﬁne the “scalar vacuum.1).1 An example of duality symmetry: D = 8 supergravity In maximal D = 8 supergravity. but this is not enough: such symmetries act transitively on the σmodel manifolds. In ordinary General Relativity with asymptotically ﬂat spacetimes. of the vacuum. D 9 8 7 6 5 4 3 G GL(2. the analogous group is the spacetime Poincar´ e group times the appropriate “internal” classifying symmetry.” the classiﬁcation symmetry that organizes the distinct solutions of the theory into multiplets with the same energy must be a subgroup of the little group. R) SL(5. R) × SL(2. They arise in part out of general covariance in the higher dimensions.e.tries of supergravity theories have always had a somewhat mysterious character.
R))/(SO(3) × SO(2)) cosetspace manifold: 8+3−(3+1) = 7.1) √ where ∗F M N P Q = 1/(4! −g)ǫM N P Qx1 x2 x3 x4 Fx1 x2 x3 x4 (the ǫ[8] is a density. (8. F[4] G[4] −→ (ΛT )−1 F[4] G[4] .. Taken all together. R) symmetry acts as follows: let λ = χ + ieσ . 2. hence purely numerical). the SL(2. but since 73 .2) c d with ab−cd = 1 is an element of SL(2. j. Owing to the directproduct structure.7) (8. Here is the relevant part of the action:56 I8 SL(2) = √ d8 x −g R − 1 ∇M σ∇M σ − 1 e−2σ ∇M χ∇M χ 2 2 − 1 1 σ e (F[4] )2 − χF[4] ∗F[4] 2 · 4! 2 · 4! (8. On the scalar ﬁelds (σ.5) (8.3) λ −→ cλ + d The action of the SL(2. we may for the time being drop the 5dimensional SL(3.e. R) symmetry on the 4form ﬁeld strength gives us an example of a symmetry of the equations of motion that is not a symmetry of the action. R) × SL(2. R)/SO(3) sector and consider for simplicity just the 2dimensional SL(2. χ). (8. R)/SO(2) sector. 3) of 1form ﬁeld strengths for zeroform potentials.5).ijk ij vector of dilatonic scalars as well as a singlet F[1] and a triplet F[1] (i. (8. R) doublet together with G[4] = eσ ∗F[4] − χF[4] .6) Since the ﬁeld equations may be expressed purely in terms of F[4] . ∇M ∗F M N P Q = 0 . R) and acts on λ by the fractionallinear transformation aλ + b . The ﬁeld strength F[4] forms an SL(2. k = 1.4) One may check that these transform the F[4] ﬁeld equation ∇M (eσ F M N P Q + χ∗F M N P Q ) = 0 into the corresponding Bianchi identity. we have a genuine symmetry of the ﬁeld equations in the transformation (8. i. which ﬁts in precisely with the dimension of the (SL(3. then a b Λ= (8. we have a manifold of dimension 7.
3. which generalizes the singlecharge bounds (4. provided that one transforms both the moduli (σ∞ . i. R) invariant. which is invariant under an obvious isotropy group H = SO(2). V ). The corresponding solitonic/magnetic solutions in ˜ D = 8 have worldvolume dimension d = 8 − 3 − 2 = 3 also.8) to the singlecharge bounds (4. membranes. The transformation (8.16).57.58 As we have seen. this is not a local symmetry of the action. the electric and magnetic charges carried by branes and appearing in the supersymmetry algebra (1.59 q In comparing (8. At the classical level.5) is a D = 8 analogue of ordinary Maxwell duality transformation in the presence of scalar ﬁelds. It is also possible in this case to have solutions generalizing the purely electric or magnetic solutions considered so far to solutions that carry both types of charge. χ∞ ) (according to (8. and the study of their chargequantisation properties involves some special features not seen in the D = 4 Maxwell case. (8.e. 8. i.56 This possibility is also reﬂected in the combined Bogomol’ny bound q for this situation. dyons. F[4] supports both electric and magnetic membranes. however.e.8) is itself SL(2. 74 . For the simple case with σ∞ = χ∞ = 0 that we have mainly chosen in order to simplify the writing of explicit solutions.8) where U and V are the electric and magnetic charges and σ∞ and χ∞ are the moduli.16): E 2 ≥ e−σ∞ (U + χ∞ V )2 + eσ∞ V 2 . we shall refer generally to the supergravity σmodel symmetries as duality symmetries.this transformation cannot be expressed locally in terms of the gauge potential A[3] .5) are forms. i. an important restriction on this spectrum of solutions enters into force: the DiracSchwingerZwanziger (DSZ) quantisation conditions for particles with electric or magnetic or dyonic charges. one should take note that for F [4] in √ (8.8) reduces to E 2 ≥ U 2 +V 2 . Accordingly.2 pform charge quantisation conditions So far. however. The bound (8. the constant asymptotic values of the scalar ﬁelds σ(x) and χ(x). 8.1) we have ∆ = 4. So in this case. so 2/ ∆ = 1. The F[4] ﬁeld strength of the D = 8 theory supports elementary/electric pbrane solutions with p = 4 − 2 = 2.e. which have a d = 3 dimensional worldvolume. At the quantum level. a given supergravity theory can have a continuous spectrum of electrically and magnetically charged solutions with respect to any one of the nform ﬁeld strengths that can support the solution.3)) and also the charges (U. we have discussed the structure of pbrane solutions at a purely classical level. the bound (8.
We shall see shortly that another diﬀerence with respect to the ordinary D = 4 Dirac quantisation of particles in Maxwell theory will be the existence of “Diracinsensitive” conﬁgurations. .. Thus.We shall ﬁrst review a WuYang style of argument. p = D − p − 4 brane.Mp+1 dxM ∧ .9): Qe AM1 . one can ﬁnd a ﬂux discrepancy ΦM1 − 75 . ∧ dxMp+1 1 . . in the background ﬁelds set up by a dual. . this excludes deformations that involve rigid rotations of an entire inﬁnite brane..9).5) are necessarily either inﬁnite or are wrapped around compact spacetime dimensions. In particular.57 (for a Diracstring argument. ΦMW is then the ﬂux through the cap MW . . one sees ˆ from the supermembrane action (4. for which the phase in (8. see Ref. (8.Mp+2 dxM1 ∧ . at least the asymptotic orientation of the electric brane must be preserved throughout the sequence of deformations. as we have seen in subsection 4. One of these is that. i.9) W where A[p+1] is the gauge potential set up (locally) by the magnetic pbrane ˆ background.e. . one may use Stoke’s theorem to rewrite the integral in (8.9) vanishes. say an electric one. some deformation sequences W will lead to a divergent integral in the exponent in (8. . and so one should not consider changing this pform in the course of the deformation any more than one should consider changing the magnitude of the electric charge in the ordinary D = 4 Maxwell case. ∧ dxMp+2 = Qe ΦMW .1) that the electric pbrane wavefunction picks up a phase factor exp iQe (p + 1)! AM1 .10) (p + 2)! MW where MW is any surface “capping” the closed surface W. magnetic.1... A number of diﬀerences arise in this problem with respect to the ordinary Dirac quantisation condition for D = 4 particles. objects carrying pform charges appearing in the supersymmetry algebra (1. Choosing the capping surface in two diﬀerent ways. a surface such that ∂MW = W. For inﬁnite pbranes. (8. After such a sequence of deformations..58. such deformations would also require an inﬁnite amount of energy. Another way of viewing this restriction on the deformations is to note that the asymptotic orientation of a brane is encoded into the electric pform charge. ∧ dxMp+1 = 1 (p + 1)! W Qe FM1 ..60 ) considering a closed sequence W of deformations of one pbrane.Mp+1 dxM ∧ . and so should be excluded from consideration. Restricting attention to deformations that give nondivergent phases.
12) . . For simplicity. with the electric pbrane oriented along the directions {xM1 . (8. the Dirac quantisation condition Qe Qm = 2πn .27b).9) will be simply exp(iQe Qm ). one may simply consider moving the electric pbrane by parallel transport around the magnetic pbrane in a closed loop. Taking this into account shows that the phase in (8. Consideration of more general deformation ˆ sequences yields the same result. Then if Mtotal = M1 ∩ M2 “captures” the magnetic pˆ brane.e.11) is almost. The phase factor (8. n∈Z. Note that magnetic pbranes have purely transverse ﬁeld strengths like (2.Mp R ∂xR/∂σ). the ﬂux ΦMtotal will equal the magnetic charge Qm of the pbrane. This measurezero set of Diracinsensitive conﬁgurations may be simply characterized in terms of the p and p charges themselves ˆ by the condition Qel ∧ Qmag = 0. but not quite. This makes it possible to simplify the discussion by making use of a specially chosen gauge.59 Now one can see how the Diracinsensitive conﬁgurations arise: the phase in (8. In making this deformation sequence. when there are shared worldvolume directions between the two branes. we recall from the above discussion that one should restrict the deformations to preserve the asymptotic orientation of the deformed pbrane. there is ˆ accordingly a gauge in which the gauge potential A[p+1] is also purely transverse. i. In deriving (8. The accrued phase factor is invariant ˆ under gauge transformations of the potential A[p+1] . thus ˆ the discrepancy in the phase factor (8.59 This is easiest to explain in a simpliﬁed case where the electric and magnetic branes are kept in static ﬂat conﬁgurations.11) by writing a (p + p)form quantisation condition ˆ Qel ∧ Qmag = 2πn [p] [p] ˆ Qel  Qmag  [p] [p] ˆ 76 mag Qel ∧ Q[p] [p] ˆ . Requiring this to equal unity gives.9) then becomes exp(iQe W AM1 .11) The charge quantisation condition (8. the full story. n∈Z.9) vanishes whenever there is even a partial alignment between the electric and the magnetic branes.11). For such conﬁgurations. it vanishes whenever any of its indices point along a worldvolume direction of the magnetic pbrane. (8.ΦM2 = ΦM1 ∩M2 = ΦMtotal (taking into account the orientation sensitivity of the ﬂux integral).9) vanishes for a measurezero set of conﬁgurations of the electric and magnetic branes. one may incorporate this orientation restriction into the Dirac quantisation condition (8. we have not taken into account the pform character of the charges. xMp }.57 in strict analogy to the ordinary case of electric and magnetic particles in D = 4. where σ is an ordering parameter for the closed sequence of deformations W. i.. . To summarize..e. one obtains no [p] [p] ˆ Dirac quantisation condition.
Let us recall the relations (6. The term κ(A) in (8.3). We need to consider the various schemes possible for dimensional reduction of dual pairs of (p. their relevance becomes more clear when one considers the relations existing between the pform charges under dimensional reduction.11) for all except the Diracinsensitive set of conﬁgurations.13). one obtains the following relations between the original charges in D = 11 and those in the reduced theory: Table 3: Relations between Q11 and QD F[4] Electric Q11 = e Magnetic Q11 = m QD V e QD m i F[3] ij F[2] ijk F[1] QD e V Li QD e V Li Lj QD e V Li Lj Lk QD L i m QD L i L j m QD L i L j L k m where Li = dz i is the compactiﬁcation period of the reduction coordinate 11−D z i and V = d11−D z = i=1 Li is the total compactiﬁcation volume. From the expressions (6. Now. 8.13a) is the analogue of the term 1 A[3] ∧ F[4] in (1.which reduces to (8. (8.e. Now consider the quantisation conditions obtained between the various dimensionally reduced charges shown in Table 3.ˆ) branes. F = F + (ChernSimons modiﬁcations) (i. given that they constitute only a measurezero subset of the total set of asymptotic brane conﬁgurations.3 Charge quantisation conditions and dimensional reduction The existence of Diracinsensitive conﬁgurations may seem to be of only peripheral importance.41)) and c is the dilaton vector corresponding to F in the dimensionallyreduced action (6. modiﬁcations involving lowerorder forms arising in the dimensional reduction similar to those in the D = 9 case (6.11) between the ﬁeld strengths in diﬀerent dimensions.13b) ˜ ˜ where F = dA. Note that the factors of Li cancel out in the various products of electric and magnetic charges only for charges belonging to the same ﬁeld strength in the reduced dimension D. We p 77 . However. the electric and magnetic charges carried by branes in D dimensions take the forms Qe Qm = = ec·φ ∗F + κ(A) ˜ F .13a) (8.11) for 2 the reduced ﬁeld strengths and their duals.
it is necessary to restore a gravitationalconstant factor of κ2 that should properly have appeared in the quantisation conditions (8. e. Now consider the mixed reductions.11. If one lets the 11 compactiﬁcation length be denoted by L in the Ddimensional theory prior to dimensional reduction. correspondingly. diagonal/vertical. since only branes supported by the same ﬁeld strength can have a Dirac quantisation condition. With such a shared worldvolume direction. Only the mixed cases will turn out to preserve Dirac sensitivity in the lower dimension after reduction. Thus. Thus. vertical/diagonal and vertical/vertical. there are two basic schemes. we have the following four reduction possibilities: diagonal/diagonal. and vertical.12). this diagonal/diagonal reduction properly corresponds to a Diracinsensitive conﬁguration.g. Note. as explained in Section 6: diagonal. For an electric/magnetic pair. As one may verify. which involves reduction on a transverse coordinate after preparation by “stacking up” singlecenter solutions so as to generate a transversespace translation invariance needed for the dimensional reduction. (8. 8. This is most easily illustrated by considering the diagonal/diagonal case.11) in D = 11 should properly have been written Qe Qm = 2πκ2 n. In performing a vertical reduction of a magnetic pbrane by stacking up an inﬁnite deck of ˆ singlecenter branes in order to create the R translational invariance necessary for the reduction. diagonal/vertical.3. Thus.4) are not dimensionless.have seen that for singleelement brane solutions. the electric and magnetic branes remain 78 . [p] ˆ Correspondingly. however. the electric and magnetic charges as deﬁned in (1. which involves reduction on a worldvolume coordinate. there are then four possible schemes. Before obtaining the Dirac quantisation condition in the lower dimension. for which z belongs to the worldvolumes of both branes. in (D − 1) dimensions one ﬁnds that the diagonally reduced electric (p − 1) brane is supported by an n = p + 1 form ﬁeld strength. For the dimensional reduction of a solution containing two elements. 1. with another aspect of dimensional reduction: the gravitational constants in dimensions D and D − 1 are related by κ2 = D Lκ2 . in a vertical reduction it is necessary to reinterpret the magnetic charge Qm as a charge density per unit z compactiﬁcation length. D−1 This ﬁts precisely. the total magnetic charge will clearly diverge. then one obtains a Dirac phase exp(iκ−2 Qe Qm L). but the diagonally reduced magnetic (ˆ − 1) brane is supported by an n = p + 2 p form. in dimension D − 1 one obtains the expected quantisation D−1 condition Qe Qm = 2πκ2 n. one has clearly fallen into the measurezero set of Diracel insensitive conﬁgurations with Q[p] ∧ Qmag = 0 in the higher dimension D. depending on whether the reduction coordinate z belongs to the worldvolume or to the transverse space of each brane. that upon making a D−1 mixed diagonal/vertical reduction.
correspond to Diracinsensitive conﬁgurations. of course. the isotropy subgroup H moves by conjugation with g. as considered in Section 7. This may immediately be seen in such solutions as the 2 ⊥ 5(1) solution (7. H −→ gHg −1 . even though they represent o only a subset of measure zero from the point of view of the higherdimensional theory. supported by the same n = p−1+2 = p + 1 form ﬁeld strength. When one takes into account the Dirac quantisation condition. speciﬁed by the values of the scalar moduli. showing that the homogeneity of the G/H coset space is broken at the quantum level by the quantisation condition. The opposite mixed vertical/diagonal reduction case goes similarly. which clearly must be a subgroup of the corresponding G(Z) duality group. Thus. this classifying symmetry becomes restricted to a discrete group. Dirac sensitivity is lost in the reduction. Correspondingly. is the linearlyrealized isotropy symmetry H given in Table 2. the particular point on the vacuum manifold G/H corresponding to the scalar moduli can be changed by application of a transitivelyacting G transformation. but because in this case both the electric and the magnetic charges need to be interpreted as densities per unit compactiﬁcation length. The existence of Diracinsensitive conﬁgurations plays a central rˆle in establishing this accord. but it is also true for solutions involving pp wave and TaubNUT elements.4 Counting pbranes As we have seen at the classical level. the two dimensionally reduced branes in D the lower dimension are supported by diﬀerent ﬁeld strengths: an n = p + 2 form for the electric brane and an n = p + 1 form for the magnetic brane. and one has D−1 D limL→0 L2 /κ2 = 0. and so one obtains a phase exp(iκ−2 Qe Qm L2 ). except that the dual branes are then supported by the same n = p + 2 form ﬁeld strength. Classically.14) . 79 (8. The value of this intersection is modulusdependent. 8. Only one factor of L is absorbed into κ2 . Correspondingly. not owing to the orientation of the branes.10). Another indication of the relevance of the Diracinsensitive conﬁgurations is the observation 59 that all the intersectingbrane solutions with some degree of preserved supersymmetry.dual to each other in the lower dimension. for example with a group element g. In the ﬁnal case of vertical/vertical reduction. so in general one seeks to identify the group G(Z)∩H. there is a perfect accord between the structure of the Dirac quantisation conditions for pform charges in the various supergravity theories related by dimensional reduction. the classifying symmetry for solutions in a given scalar vacuum.
so the modulusindependent relations of Table 3 imply that the lowerdimensional charges (8. since the duality transformation maps e → e−1 . 80 . These should not be thought of as determining the geometry of the compactifying internal manifold. This is because the modulus dependence cancels out in the “canonical” charges that we have deﬁned in Eq.15) e Thus. (8. One way to see this is to use the relations between charges in diﬀerent dimensions given in Table 3.13) from which the charges (8. Thus.13) (together with the rest of the φ dilatonic scalar dependence) play the rˆles of coupling constant factors like e−2 . corresponding to a covariant derivative can Dµ = ∂µ + iAµ . because the modulus factors ec·φ∞ appearing in (6. although the conventional charge Qconv scales proportionally to e. it is evident that the group G(Z). The analogous feature in ordinary Maxwell theory is that a true duality symmetry of the theory only arises when the electric charge takes the value e = 1 (in appropriate units). does not depend on the moduli.The discretized duality group G(Z). for which the action is − 1/4 Fµν F conv µν .13). r Note that the compactiﬁcation periods L appearing in Table 3 have values that may be i adjusted by convention.13) do not depend on the moduli. The modulus independence of the charges (8.13) are derived. where an analogous charge would be that derived from the can action IMax = − 1/(4e2 ) Fµν F can µν . If one wants to compare o this to the “conventional” charges deﬁned with respect to a conventional gauge conv potential Aconv = e−1 Acan . in the Maxwell electrodynamics case. As a consequence of the diﬀerent modulus dependences of H and of G(Z). This is analogous to our dimensionally reduced action (6. it follows that the size of the intersection group G(Z) ∩ H is dependent on the moduli.13) works in a similar fashion. which is determined instead by the scalar moduli. given that the discretized quantum duality group G(Z) is deﬁned by the requirement that it map the set of Diracallowed charges onto itself. referred to the canonical charges (8. the dependence on the electric charge unit e drops out in Qcan . the value e = 1 is a distinguished value. Then. on the other hand. Thus. does not depend upon the moduli.13). (8. the relations of Table 3 imply the independence of the canonicallydeﬁned charges from the physically relevant moduli.r Another way to understand this is by comparison with ordinary Maxwell electrodynamics. noting that there are no scalar moduli in D = 11. then µ µ the canonical and conventional charges obtained via Gauss’s law surface integrals are related by Qcan = 1 2e2 d2 Σij ǫijk F can 0k = 1 2e d2 Σij ǫijk F conv 0k = 1 Qconv .
16) Requiring a. −→ eσ + χ2 e−σ (8. we may write out the transformation in detail using (8. Z) at the distinguished point on the scalar vacuum manifold. G[4] ) −→ (−G[4] . To lowest order in zeroform gauge potentials. and consider the S2 transformation to lowest order in χ.19) Considering this S2 transformation to lowest order in the zeroform χ has the advantage that the signﬂip of φ may be “impressed” upon the a dilaton vector for F[4] : a → −a. one ﬁnds (F[4] . for the scalars (σ.The distinguished point on the scalar vacuum manifold for general supergravity theories is the one where all the scalar moduli vanish.17) Thus. b = −1. d ∈ Z and also that the modulus point σ∞ = χ∞ = 0 be left invariant. χ). suppose that the zeroform gauge potential χ is small. c = 1. again with the moduli set to the distinguished point on the scalar manifold. (8. the transformation just ﬂips the signs of σ and χ. Acting on the ﬁeld strengths (F[4] . for our truncated system. which maps σ and χ according to e−σ χe −σ −→ −χe−σ . causing in this case a permutation of the ai .18) maps the ﬁeld equation for F[4] into the corresponding Bianchi identity: ∇M (eσ F M N P Q + χ∗F M N P Q ) −→ −∇M ∗F M N P Q . we ﬁnd only two transformations: the identity and the transformation a = d = 0. generating for the D = 8 case overall the discrete group 81 . This is the point where G(Z) ∩ H is maximal.3): e−σ χe −σ −→ (d + cχ)(b + aχ)e−σ + aceσ . −→ (d + cχ)2 e−σ + c2 eσ (8. To this order.18) One may again check (in fact to all orders. G[4] ). b. Now consider the SL(3. (8. F[4] ) . we ﬁnd just an S2 discrete symmetry as the quantum isotropy subgroup of SL(2. This S2 is the natural analogue of the S2 symmetry that appears in Maxwell theory when e = 1. In order to aid in identifying the pattern behind this D = 8 example. not just to lowest order in χ) that (8. c. the action of SL(3. Z) ∩ H may similarly by impressed upon the 3form dilaton vectors. R)/SO(3) sector of the D = 8 scalar manifold. In that case. The general structure of such G(Z) ∩ H transformations will be found by considering the impressed action of this group on the dilaton vectors. Let us return to our D = 8 example to help identify what this group is.
By combining these duality multiplets together with the diagonal and vertical dimensional reduction families discussed in Sections 6 and 6. so the action of the maximal G(Z) ∩ H pbrane classifying symmetry is identiﬁed with that of the Weyl group of G.S3 × S2 . From the analysis of the Weylgroup duality multiplets. Now that we have a bit more structure to contemplate. R) satisfy hi · hj = δij − 1 . For example. corresponding to the cases N = 2 and N = 3. we recall that the weight vectors hi of the fundamental representation of SL(N. the above pattern persists:61 in all cases.20) ±1 1 These relations are precisely those satisﬁed by √2 a and √2 ai . corresponding to ∆ = 4/N for the dilatonic scalar coupling. and is most simply described as the Weyl group of G. R) groups. R). N N hi = 0 . This is then the group that counts the distinct pbrane solutions s of a given type (4.3.64 s Of 82 . where the supergravity symmetry groups shown in Table 2 grow in complexity.61 The invariance of the dilaton vectors’ dot products prompts one to return to the algebra (6. This suggests that the action of the maximal G(Z) ∩ H group (i. course. The symmetry group of the fundamental weights is the Weyl group 61 of G.e. one ﬁnds the multiplicities given in Table 4. i=1 (8. ai ) dot products invariant. subject to the Dirac quantisation condition and referred to the distinguished point on the scalar modulus manifold. As one proceeds down through the lowerdimensional cases. the action of the maximal classifying symmetry G(Z) ∩ H may be identiﬁed with the Weyl group of G. For supersymmetric pbranes arising from a set of N participating ﬁeld strengths F[n] . R) and H = SO(5). in D = 7. one ﬁnds that the action of G(Z) ∩ H is equivalent to that of the discrete group S5 . which is the Weyl group of SL(5. In the lowerdimensional cases shown in Table 2. the discrete group G(Z) ∩ H becomes less familiar. we can notice that the G(Z) ∩ H transformations leave the (a. for scalar moduli set to the distinguished point on the scalar manifold) may be identiﬁed in general with the symmetry group of the set of fundamental weights for the corresponding supergravity duality group G as given in Table 2.15) of these dot products and see what else we may recognize in it. where from Table 2 one sees that G = SL(5.6). one may tabulate 61 the multiplicities of pbranes residing at each point of the plot given in Figure 7. Noting that the duality groups given in Table 2 for the higher dimensions D involve SL(N. the corresponding supermultiplet structures have been discussed in Ref. these solutions must also fall into supermultiplets with respect to the unbroken supersymmetry. the full set of p ≤ (D − 3) branes shown in Figure 7 becomes “welded” together into one overall symmetrical structure.
we shall need to exploit the existence of certain special “unitsetting” brane types.13a). what the minimum electric charge is. the argument may be turned around to show that the set of allowed electric charges is given by integer multiples of the minimum electric charge.5 The charge lattice For the electric and magnetic BPS brane solutions supported by a given ﬁeld strength. (8. given a certain minimum “electric” charge (8. taking the minimum magnetic charge from this set.Table 4: Examples of pbrane Weylgroup multiplicities F[n] F[4] F[3] F[2] F[1] ∆ 4 4 4 2 4/3 4 2 4/3 10 1 1 1 9 1 2 1+2 2 2 8 2 3 6 6 8 12 7 5 10 15 20 60 D 6 10 16 40 40 280 480 5 4 27 135 45 72 1080 4320 56 756 2520 126 3780 30240+2520 8. the allowed set of magnetic charges is determined. This cannot be established by use of the Dirac quantisation condition alone.21) Given these values for the magnetic charge. Then. Upon dimensional reduction down to D = 10. one obtains a magnetic 6brane solution. There are other tools. r∈Z. the D = 10 Dirac quantisation 83 . This argument does not directly establish. we have seen above that the Dirac charge quantisation condition (8. where k is the chargedetermining parameter in the 3dimensional harmonic function H(y) = 1 + k/(y).e. i. We have already encountered one example of a “unitsetting” brane in subsection 7. that one can use to ﬁx the charge lattice completely.2.12) implies that. the value of the charge unit. however.7) is nonsingular provided that the coordinate ψ is periodically identiﬁed with period L = 4πk. where we encountered the pp wave/TaubNUT pair of D = 11 solutions. with a charge classically discretized to take a value in the set Qm = rL . however. To do so. and also to exploit fully the consequences of the assumption that the G(Z) duality symmetry remains exactly valid at the quantum level. We saw there that the TaubNUT solution (7.
65 This link may be used to establish relations between the charge units for the various pform charges of diﬀering rank.2. one must exploit the fact that brane solutions with Poincar´ worldvolume symmetries may be dimensionally e reduced down to lower dimensions.t In doing so. although its ﬁeld equations are perfectly welldeﬁned.4. see Ref. with a charge vector at 45◦ to the electric axis. (8. In a given dimension D.59 84 . as we saw in subsection 7.23) Qe = L Thus. the G(Z) duality symmetries only rotate between pbranes of the same worldvolume dimension. We shall consider the type IIB theory in some more detail in Section 9. The corresponding electrically and magnetically charged BPS solutions are 3branes. the solutions descending from an original pbrane in D dimensions are subject to a larger G(Z) duality symmetry. owing to the selfduality condition. and this can be used to rotate a descendant brane into descendants of p′ branes for various values of p′ . these solutions are actually dyons. then one may relate the 6brane charge units to those of other BPS brane types. and. establishing relations via the duality symmetries between various BPS brane types which can be supported by diﬀerent ﬁeld strengths. n∈Z. gives an allowed set of electric charges 2πκ2 10 n. it will be suﬃcient for us to note that the dyonic 3branes of t For details of the duality relations between charge units for diﬀerent pbranes. Dimensional oxidation back up to D dimensions then completes the link. the requirement that magnetic D = 10 6branes oxidize up to nonsingular TaubNUT solutions in D = 11 fully determines the 6brane electric and magnetic charge units and not just the product of them which occurs in the Dirac quantisation condition If one assumes that the G(Z) duality symmetries remain strictly unbroken at the quantum level. where the duality groups shown in Table 2 grow larger. equivalently. Another chargeunitsetting BPS brane species occurs in the D = 10 type IIB theory. 10 n∈Z.22) or. This theory has a wellknown diﬃculty with the formulation of a satisfactory action.condition Qe Qm = 2πκ2 n . however. supported by the same kind of ﬁeld strength. the quantisation of D = 11 pp wave momentum in the compact ψ direction. for now. (8. Upon reduction down to dimensions Dred < D. including ﬁeld strengths of diﬀerent rank. even though the corresponding solutions are Diracinsensitive to each each other. H[5] = ∗H[5] . The diﬃculty in formulating an action arise from the presence of a selfdual 5form ﬁeld strength. as we have seen from our discussion of the Weylgroup action on pbranes given in subsection 8.
this condition is symmetric:59 (1) Q(1) Q(2) + Q(2) Qm = 2πκ2 n . In any dimension D. (Qe . n∈Z (8.u quite diﬀerently from the antisymmetric cases in D = 4r dimensions.27) The proper interpretation of the discretized CremmerJulia G(Z) duality symmetry at the level of supergravity theory is subject to a certain amount of debate. one may relate the πκIIB charge unit to those of other supergravity R–R charges. one may determine the chargelattice units for all BPS brane types.59 . Qm ). but at the level of string theory the situation becomes more clear. using √ duality symmetries. which is a perturbative symmetry holding orderbyorder in the string loop expansion. Then.3. T duality 66 consists of transformations that invert the radii of a u They will be Diracsensitive provided one of them is slightly rotated so as to avoid having any common worldvolume directions with the other.D = 10 type IIB theory are also a unitsetting brane species. It is easiest to express the units of the resulting overall charge lattice by making a speciﬁc choice for the compactiﬁcation periods. in order to avoid having a Diracinsensitive conﬁguration as discussed in subsection 8. 11 1 (8.24) means that dyons may be Diracsensitive to others of their own type. (8.24) unlike the more familiar antisymmetric DSZ condition that is obtained in dimensions D = 4r. 85 . Thus.65. there is a subgroup of G(Z) that corresponds to T duality. (8. If one lets all the compactiﬁcation periods Li be equal. The symmetric nature of (8. e m e 4r+2 n∈Z. one thus obtains the quantisation condition √ Q[3]  = n πκIIB . Li = LIIB = L = (2πκ2 ) 9 . 9 Local versus active dualities ∆Qm = Ln−1 . Qm ).59 The unitsetting property arises because of a characteristic property of the DiracSchwingerZwanziger quantisation condition for dyons in dimensions D = 4r + 2: for (1) (1) (2) (2) dyons (Qe . using duality symmetries together with the pp wave/TaubNUT and selfdual 3brane charge scales.26) then the electric and magnetic chargelattice units for rankn ﬁeld strengths in dimension D are determined to be 59 ∆Qe = LD−n−1 .25) where κIIB is the gravitational constant for the type IIB theory. For the 45◦ dyonic 3branes.
there is also a G(Z) covariance of the set of charge vectors for physically inequivalent BPS brane solutions. where the choice of a particular asymptotic geometry is necessary in order to establish the “vacuum” with respect to which quantized ﬂuctuations can be considered. As with general coordinate transformations. we did not consider in detail the action of G(Z) on the moduli. duality symmetries may occur in several diﬀerent guises that are not always clearly distinguished. it is clear that the BPS spectrum at ﬁxed scalar moduli cannot form a multiplet under the standard CremmerJulia G(Z) duality symmetry. T duality needs to be viewed as a local symmetry in string theory. one should compare solutions with the same asymptotic values of the scalar ﬁelds. Since the standard CremmerJulia duality transformations. Aside from such a relabeling. the overall string spectrum remains unchanged. these quantities should be ﬁxed when quantizing about a given vacuum state of the theory.13) are in fact modulusindependent.. Hence. because. however. When this is done. given the local interpretation adopted for the standard duality transformations as discussed above: once one has identiﬁed solution/modulus sets under the standard G(Z) duality transforma86 . ﬂat space in Cartesian or in Rindler coordinates is viewed as one and the same solution. the canonicallydeﬁned charges (8. according to which. as we saw in subsection 8.13) related by G(Z) transformations generally have diﬀering mass densities.toroidal compactiﬁcation.5.4. one ﬁnds that solutions carrying charges (8. such as those of our D = 8 example in subsection 8. in considering physicallyinequivalent solutions. This local interpretation of the G(Z) duality transformations is similar to that adopted for general coordinate transformations viewed passively. Depending on whether one considers (D − 3) branes to be an unavoidable component of the spectrum.g. Thus. under which quantized string oscillator modes and string winding modes become interchanged.67 The wellfounded basis. Since the dilatonic and axionic scalar moduli determine the coupling constants and vacuum θangles of the theory. i. the same has also been argued to be the case at the level of the supergravity eﬀective ﬁeld theory.5. This is similar to the treatment of asymptotically ﬂat spacetime in gravity. e.e. commute with P 0 time translations and so necessarily preserve mass densities. string conﬁgurations on compact manifolds related by T duality are identiﬁed.1. This conclusion is in any case unavoidable. In the discussion of subsection 8. in string theory at least. however.63 that the full duality group G(Z) should be given a local interpretation: sets of string solutions and moduli related by G(Z) transformations are to be treated as equivalent descriptions of a single state. As one can see from the charge lattice discussed in subsection 8. for a local interpretation of the T duality subgroup of G(Z) has led subsequently to the hypothesis 62.
v For M −→ Λ M ΛT .1) contains two scalar ﬁelds: a dilatonic scalar φ which occurs nonlinearly through its exponential. (9.4) H[3] = (2) dA[2] The action (9. R) transformations H[3] −→ (ΛT )−1 H[3] .1).5? At least at the classical level.3) The doublet H[3] contains the ﬁeld strengths of the 2form gauge potentials A[2] : (1) dA[2] . The matrix M in (9. (i) (j) (9.e.70 . explicitly. and an axionic scalar χ. and for singlecharge (i.tions. the question arises: is there any spectrumgenerating symmetry lying behind the apparently G(Z) invariant charge lattices of inequivalent solutions that we saw in subsection 8.1 The symmetries of type IIB supergravity Aside from the diﬃculties arising from the selfduality condition for the 5form ﬁeld strength H[5] .5) a detailed discussion of SL(2. This somewhat hybrid procedure will be suﬃcient for our present purposes. see Ref. one cannot then turn around and use the same G(Z) transformations to generate inequivalent solutions.1) is invariant under the SL(2.2) may be handled in the fashion of Ref.69 87 . (9.1) The 5form selfduality condition H[5] = ∗H[5] (9. (9. ∆ = 4) solutions. which may also be considered to be a zeroform gauge potential. R) duality in type IIB supergravity. the answer 68 turns out to be ‘yes. one has M= e−φ + χ2 eφ χ eφ χ eφ eφ . being imposed by hand as an extra constraint on the ﬁeld equations obtained by varying (9.v 9.’ We shall illustrate the point using type IIB supergravity as an example. the equations of motion of the bosonic ﬁelds of the IIB theory may be derived from the action IIB I10 = T 2 1 1 d10 x[eR+ 1 e tr(∇µ M−1 ∇µ M)− 12 e H[3] M H[3] − 240 e H[5] 4 1 − 2√2 ǫij ∗ (B[4] ∧dA[2] ∧dA[2] )] . Thus.
R) duality transformations for the physically distinct BPS string and 5brane multiplets of the theory. the action (9.8) does not leave the action (9.1) invariant. R) transformations map solutions of (9.8) note that the power of λ in each ﬁeld’s transformation is equal to the number of indices it carries. A[2] −→ λ2 A[2] . Although the transformation (9. There is one more symmetry of the equations of motion following from the action (9. this is an “electricelectric” duality. From the equations of motion of the 3form ﬁeld strength H[3] in (9. then the transformation on M can be rewritten as the fractional linear transformation τ −→ aτ + b . similar issues concerning duality multiplets for a ﬁxed scalar vacuum arise in both cases. R) constraint is ad − bc = 1. If one deﬁnes the complex scalar ﬁeld τ = χ + i e−φ .1).1). R). (i) (i) H[5] −→ λ4 H[5] .1) into other solutions.6) and the SL(2.7) Note that since H[5] is a singlet under SL(2.8) the feature of being a symmetry only of the equations of motion. Since this SL(2. (9. and. It should be noted that the SL(2. (9. R) symmetry. as opposed to the “electricmagnetic” duality discussed in the D = 8 example of subsection 8.1) transforms homogeneously as λ3 under the following scaling transformations: gµν −→ λ2 gµν . R) transformation rotates the doublet A2] of electric 2form potentials amongst themselves. R) electricmagnetic duality of the D = 8 example given in subsection 8. This is a rather humble symmetry that is not often remarked upon.9) . The SL(2. Nonetheless. cτ + d (9. but which will play an important role in constructing active SL(2. 2 88 (9.2).1 shares with the transformation (9. also preserves the SL(2. accordingly. As for pure sourcefree Einstein theory. which is imposed by hand. 1 d ∗ (M H[3] ) = − √ H[5] ∧ ΩH[3] . We shall need to consider in particular the action of these transformations on the charges carried by solutions. the λ3 homogeneity of this scaling for all terms in the action is suﬃcient to produce a symmetry of the IIB equations of motion. the selfduality constraint (9. the scalars φ and χ are not transformed. and not of the action.where the SL(2.1. R) parameter matrix is Λ= a c b d .
Taking the basis states of the IIB charge lattice to be e1 = 1 0 e2 = 0 1 . We have seen in our general discussion of charge lattices in subsection 8. In the case of the type IIB theory. Z) 89 .12) which transforms under SL(2. contragrediently to Qe . R) invariant. (9. the discretized duality symmetries G(Z) are given a local interpretation in string theory.5 that the continuous classical CremmerJulia symmetries G break down to discretized G(Z) symmetries that map between states on the quantum charge lattice. R)invariant tensor Ω= 0 1 −1 0 .where Ω is the SL(2. R) symmetry on type IIB supergravity solutions may be expressed in terms of its action on the solutions’ charges and on the scalar moduli. This group action may be viewed as an automorphism of a vector bundle. the classical SL(2. (9. R) transformation.e. and the charge vector space as the ﬁber. As we have discussed above. R)/SO(2) target manifold as the base space. The transformation properties of the electric and magnetic charge doublets are just such as to ensure that the Dirac quantization condition QT Qe ∈ m 2πκ2 Z is SL(2. one has in addition a topologicallyconserved magnetic charge doublet. Qe transforms covariantly as a doublet: Qe → ΛQe . IIB The overall eﬀect of this standard SL(2. with the scalar ﬁelds’ SL(2. In the present type IIB case. i.10) one ﬁnds that the following twocomponent quantity is conserved: Qe = ∗ 1 (M H[3] ) + √ Ω(2B[4] ∧ H[3] − H[5] ∧ A[2] ) 3 2 . this is a hypothesis rather than a demonstrated result.13) the surviving SL(2. R) as Qm → (ΛT )−1 Qm . (9. R) matrices with integral entries.11) Under an SL(2. (9. Z) group will be represented by SL(2. because the SL(2. Qm = H[3] . By virtue of the Bianchi identities for the 3form ﬁeld strength. R) symmetry breaks down to SL(2. Z) at the quantum level.
Z) transformation reduces to a single point. The importance of the Borel subgroup for our present purposes is that it acts transitively on the G/H = SL(2. there will exist a corresponding projective stability subgroup which is isomorphic to (9. τ∞ ) to new values (Q′ . even though they will contain states with diﬀerent mass densities. and this is a distinctly nonperturbative transformation. with parameter Λ. notice that within SL(2. but obtained by conjugation of (9. (9. Finding the surviving quantumlevel SL(2. one wishes to return the complex scalar ′ modulus τ∞ to its original value τ∞ . 90 . To do this. maps the charge ′ vector and complex scalar modulus (Q.14). but without changing the “already ﬁnal” values of the charges. Adopting this hypothesis nonetheless.14) This standard representation of the SL(2. R) there is a subgroup that leaves a doublet charge vector Q′ invariant up to an overall rescaling. b ∈ R . R) symmetry of the type IIB theory. This latter fact alone tells us that we must include some transformation that acts on the metric.transformations map between NS–NS and R–R states. For a general charge vector Q′ . but which also transforms in an unwanted way the scalar moduli. which transforms the doublet charges (9.2 Active duality symmetries Now let us see how duality multiplets of the physically inequivalent BPS states can occur. After this initial Λ transformation. R) Borel subgroup clearly leaves the basis charge vector e1 of Eq. R): Borel = a 0 b a−1 a.13) invariant up to scaling by a. We shall continue with our exploration of the continuous classical SL(2. The procedure starts with a standard SL(2. R)/ SO(2). R) transformation. R)/SO(2) coset space in which the scalar ﬁelds take their values. an orbit of the standard SL(2. Subsequent compensating transformations will then have the task of eliminating the unwanted transformation of the scalar moduli. 9. τ∞ ). in order to obtain an overall transformation that does not in the end disturb the complex modulus.14) with an = element of H ∼ SO(2). Let us suppose that this initial transformation. so this transformation may be used to return the scalar moduli to the original values they had before the Λ transformation. This projective stability group of Q′ is isomorphic to the Borel subgroup of SL(2.11) in a straightforwardly linear fashion. the scalar modulus space becomes the double coset space SL(2. Z) later on will be a straightforward matter of restricting the transformations to a subgroup. Z)\ SL(2. after making the corresponding identiﬁcations. (9.
8) also scales the metric.The next step in the construction is to correct for the unwanted scaling Q′ → aQ′ which occurs as a result of the Borel compensating transformation. R) duality package constructed in this way transforms the charges in a linear fashion. ˜ ˜ ˜ θfi = θf − θi tan θi = pi /qi . The net active SL(2. aQ′ → λ2 aQ′ . by use of a further compensating scaling of the form (9. xµ → x′µ = a−1/2 xµ .3). in which one picks the rigid parameter λ such that λ2 a = 1. The compensating scaling transformation t of the form (9. where the matrix V∞ is an element of Borel that has the eﬀect of moving the scalar modulus from the point τ = i to the point τ∞ : V∞ = e−φ∞ /2 1 0 e φ∞ χ∞ e φ∞ . Speciﬁcally. This fac91 . Q → λQ′ . R) there is an Iwasawa decomposition Λ=˜ .15) where ˜ ∈ BorelQ′ is an element of the projective stability group of the ﬁnal b charge vector Q′ and where h ∈ Hτ∞ is an element of the stability subgroup of τ∞ . note that the transformation (9. for τ∞ = χ∞ + ie−φ∞ and a transformation Λ mapping pi pf Qi = to Qf = = ΛQi .8). the Borel transformation that is needed in this construction is just b = (˜ −1 . M = V V T . For the ﬁnal step. R). This is achieved by a net construction that acts upon the ﬁeld variables of the theory in a quite nonlinear fashion. Clearly. bh (9. but now leaving the complex scalar modulus τ∞ unchanged.17) The matrix V∞ appearing here is also the asymptotic limit of a matrix V (φ. This almost completes the construction of the active SL(2. one needs to compensate for this scaling by a ﬁnal general coordinate transformation. the h ∈ Hτ∞ group element is qi qf hfi = V∞ ˜ cos θfi ˜ − sin θfi ˜ sin θfi ˜ cos θfi −1 V∞ . (9.8) and the associated general coordinate transformation also leave the scalar moduli unchanged. The overall active SL(2. gµν → λ2 gµν = a−1 gµν .16) ˜ tan θi = eφ∞ (tan θi − χ∞ ) . R) transformation thus is just btΛ = th. in exactly the same way as the standard supergravity CremmerJulia SL(2. χ) that serves to factorize the matrix M given in (9. This net transformation may be explicitly written by noting that for SL(2. R) duality. leaving thus a transformation h ∈ Hτ∞ which does not change b) the complex modulus τ∞ . (9. Since one does not want to alter the asymptotic metric at inﬁnity.
The standard SL(2. Note that the matrix M determines both the scalar kinetic terms and also their interactions with the various antisymmetrictensor gauge ﬁelds appearing in the action (9. tfi = mf . Λ1 Q)O(Λ1 . the supergravity duality groups G shown in Table 2 grow in rank and the structure of the charge orbits becomes progressively more and more complicated. but the above story is basically repeated for an important class of pbrane solutions. R) symmetry needs to be checked in the same fashion as for nonlinear realizations generally. R) constructed above likewise becomes restricted to an SL(2. Z) subgroup in order to respect this charge lattice. it is clear that the two realizations of this group are distinctly diﬀerent. At the quantum level. This quantumlevel discretized group of active transformations is obtained simply by restricting the matrix parameters Λ for a classical active SL(2. Since. one needs to check that a group operation O(Λ. the Dirac quantization condition restricts the allowed states of the theory to a discrete charge lattice. R) construction is simply expressed as a ratio of mass densities.torization makes plain the transitive action of the Borel subgroup on the SL(2. i. referred to a given scalar vacuum determined by the complex modulus τ∞ . Z) subgroup. by contrast. Mapping between diﬀerent mass levels. mi m2 = QT M−1 Qi . Q) = O(Λ2 Λ1 .8) acts on the metric and thus enables the active SL(2. as we have seen. The scalingtransformation part of the net active SL(2. R). i i ∞ (9.19) One may verify directly that the nonlinear realization given by (9. This is the class of singlecharge 92 .18) This expression reﬂects the fact that the scaling symmetry (9.18).16.18) does in fact satisfy this composition law. the massdensity levels are invariant under the action of the standard SL(2. when acting on any of the ﬁelds of the type IIB theory. 9.f . Q) = th acting on an initial state characterized by a charge doublet Q combines with a second group operation according to the rule O(Λ2 . R) transformation so as to lie in SL(2. R) transformation to relate solutions at diﬀerent massdensity levels mi. The group composition property of the active SL(2. Z). (9.e. and the active SL(2. In lowerdimensional spacetime.1). R) symmetry thus becomes restricted to a discrete SL(2. R)/SO(2) coset space in which the scalar ﬁelds take their values. can only be achieved by including the scaling transformation (9. Q) .
this subgroup acts transitively on the coset space G/H in which the scalar ﬁelds take their values. BPS brane solutions with ∆ = 4/N can be interpreted as coincidentchargecenter cases of intersectingbrane solutions with N elements. Once again. one may see that this group action is transitive by noting that the matrix M (9. R) example that we have presented. h ∈ Hmoduli . This construction depends upon the existence of a projective stability group 68. The active G(Z) duality constructions work straightforwardly enough at the classical level.71 of the charge Q that is isomorphic to the Borel subgroup of G. . but their dependence on symmetries of ﬁeld equations that are not symmetries of the corresponding actions gives a reason for caution 93 where BorelQ is isomorphic to the Borel subgroup of G. The operation # here depends on the groups G and H in question. each of which would separately be a ∆ = 4 solution on its own. The construction of active duality symmetries for such multiplecharge solutions remains an open problem. the construction of active G symmetry transformations that preserve the scalar moduli proceeds in strict analogy with the type IIB SL(2. for H orthogonal # (9.71 ) Given the above grouptheoretical structure.15): Λ=˜ bh ˜ ∈ BorelQ . Only the asymptotic scalar moduli can be moved transitively by the Borel subgroup of G and. As in the SL(2. and possess an Iwasawa decomposition generalizing the SL(2.20) (The D = 3 case in which G = E8(+8) and H = SO(16) needs to be treated as a special case.solutions. b (9. the representations carried by the charges in such multicharge cases are not of highestweight type.3) which governs the scalar kinetic terms and interactions can be parameterized in the form M = V V # .21) V = V †. for H a USp group. where V is an element of the Borel subgroup. As we have seen in Section 7. correspondingly. R) case (9. for H unitary ΩV † . in spacetime dimensions D ≥ 4 we have V T. for they have a larger class of integration constants. The duality groups shown in Table 2 are all maximally noncompact. This is the case whenever Q transforms according to a highestweight representation of G. R) example of the type IIB theory. for which the charges Q fall into highestweight representations of G. representing relative positions and phases of the charge components. so this is the correct subgroup to use for a compensating transformation to restore the moduli to their original values in a given scalar vacuum. The BPS brane solutions with this property are the singlecharge solutions with ∆ = 4.
about their quantum durability. This may be a subject where string theory needs to intervene with its famed “miracles.” Some of these miracles can be seen in supergravitylevel analyses of the persistence of BPS solutions with arbitrary mass scales, despite the presence of apparently threatening quantum corrections,68 but a systematic way to understand the remarkable identities making this possible is not known. Thus, there still remain some areas where string theory appears to be more clever than supergravity. 10 Noncompact σmodels, null geodesics, and harmonic maps
A complementary approach 72,73,74 to the analysis of brane solutions in terms of the four D = 11 elemental solutions presented in Section 7 is to make a dimensional reduction until only overalltransverse dimensions remain, and then to consider the resulting nonlinear σmodel supporting the solution. In such a reduction, all of the worldvolume and relativetransverse coordinates are eliminated, including the time coordinate, which is possible because the BPS solutions are all time independent. The two complementary approaches to the analysis of BPS brane solutions may thus be characterized as oxidation up to the top of Figure 7, or reduction down to the left edge Figure 7, i.e. reduction down to BPS “instantons,” or p = −1 branes, with worldvolume dimension d = 0. The d = 0 instanton solutions are supported by 1form ﬁeld strengths, i.e. the derivatives of axionic scalars, F[1] = dχ. Taken together with the dilatonic scalars accumulated in the process of dimensional reduction, these form a noncompact nonlinear σmodel with a target manifold G/H ′ , where G is the usual supergravity symmetry group shown in Table 2 for the corresponding (reduced) dimension D but H ′ is a noncompact form of the modulus little group H shown in Table 2. The diﬀerence between the groups H ′ and H arises because dimensional reduction on the time coordinate introduces extra minus signs, with respect to the usual spatialcoordinate KaluzaKlein reduction, in “kinetic” terms for scalars descending from vector ﬁelds in the (D + 1) dimensional theory including the time dimension. Scalars descending from scalars or from the metric in (D + 1) dimensions do not acquire extra minus signs. The change to the little group H ′ is also needed for the transformation of ﬁeld strengths of higher rank, but these need not be considered for our discussion of the BPS instantons. The relevant groups for the noncompact σmodels in dimensions 9 ≥ D ≥ 3 are given in Table 5. These should be compared to the standard CremmerJulia groups given in Table 2. The sector of dimensionallyreduced supergravity that is relevant for the instanton solutions consists just of the transversespace Euclideansignature 94
Table 5: Symmetries for BPS instanton σmodels.
D 9 8 7 6 5 4 3
G GL(2, R) SL(3, R) × SL(2, R) SL(5, R) SO(5, 5) E6(+6) E7(+7) E8(+8)
H′ SO(1, 1) SO(2, 1) × SO(1, 1) SO(3, 2) SO(5, C) USP(4, 4) SU∗ (8) SO∗ (16)
metric and the G/H ′ σmodel, with an action Iσ = √ 1 dD y g R − 2 GAB (φ)∂i φA ∂j φB g ij , (10.1)
where the φA are σmodel ﬁelds taking values in the G/H ′ target space, GAB is the targetspace metric and g ij (y) is the Euclideansignature metric for the σmodel domain space. The equations of motion following from (10.1) are √ 1 √ ∇i ( gg ij GAB (φ)∂j φB ) g =0 (10.2a) (10.2b)
Rij = 1 GAB (φ)∂i φA ∂j φB , 2
where ∇i is a covariant derivative; when acting on a targetspace vector VA , it is given by ∇i VA = ∂i VA − ΓD (G)∂i φE VD , (10.3) AE in which ΓA (G) is the Christoﬀel connection for the targetspace metric GAB . BC The action (10.1) and the ﬁeld equations (10.2) are covariant with respect to generalcoordinate transformations on the σmodel target manifold G/H ′ . The action (10.1) and the ﬁeld equations (10.2) are also covariant with respect to general y i → y ′i coordinate transformations of the domain space. These two types of general coordinate transformations are quite diﬀerent, however, in that the domainspace transformations constitute a true gauge symmetry of the dynamical system (10.1), while the σmodel targetspace transformations generally change the metric GAB (φA ) and so correspond to an actual symmetry of (10.1) only for the ﬁniteparameter group G of targetspace isometries. As in our original search for pbrane solutions given in Section 2, it is appropriate to adopt an ansatz in order to focus the search for solutions. In 95
the search for instanton solutions, the metric ansatz can take a particularly simple form: g ij = δ ij , (10.4) in which the domainspace metric is assumed to be ﬂat. The σmodel equations and domainspace gravity equations for the ﬂat metric (10.4) then become ∇i (GAB (φ)∂i φB ) = 0 Rij = 1 GAB (φ)∂i φA ∂j φB = 0 2 (10.5a) (10.5b)
.
Now comes the key step 72 in ﬁnding instanton solutions to the specialized equations (10.5): for singlecharge solutions, one supposes that the σmodel ﬁelds φA depend on the domainspace coordinates y i only through some intermediate scalar functions σ(y), i.e. φA (y) = φA (σ(y)) . (10.6)
After making this assumption, the σmodel φA equations (10.5a) become ∇2 σ dφB dφC d2 φA dφA + ΓA (G) + (∂i σ)(∂i σ) BC dσ dσ 2 dσ dσ =0, (10.7)
while the gravitational equation (10.5b) becomes the constraint GAB (φ) dφA dφB =0. dσ dσ (10.8)
An important class of solutions to (10.7) is obtained by taking ∇2 σ = 0 d2 φA dφB dφC + ΓA (G) =0. BC dσ 2 dσ dσ (10.9a) (10.9b)
At this point, one can give a picture of the σmodel maps involved in the system of equations (10.8, 10.9), noting that (10.9a) is just Laplace’s equation and that (10.9b) is the geodesic equation on G/H ′ , while the constraint (10.8) requires the tangent vector to a geodesic to be a null vector. The intermediate function σ(y) is required by (10.9a) to be a harmonic function mapping from the ﬂat (10.4) Euclidean domain space onto a null geodesic on the target space G/H ′ . Clearly, the harmonic map σ(y) should be identiﬁed with the harmonic function H(y) that controls the singlecharge brane solutions (2.24). On the geodesic in G/H ′ , on the other hand, σ plays the role of an aﬃne parameter. The importance of the noncompact structure of the target space manifold 96
This is the σmodel construction that generates multicharge solutions. there are important compatibility conditions that must be satisﬁed in order for such multicharge solutions to exist.10) . however. In that case. conversely. but with a totally geodesic submanifold of G/H ′ . The σmodel solution (10. a required condition is expressed in terms of the velocity vectors for the null geodesics. σ(y) ym φ (A σ ( nu ) ll) G/ H' ∇2 σ = 0 flat I D E Figure 8: Harmonic map from ED to a null geodesic in G/H ′ .G/H ′ .2 that. one deals not with a single geodesic. the geodesics generated by any curve in the intermediate σa parameter space must be null. If one adopts a matrix representation M for points in the coset manifold G/H ′ . the projectors constraining the surviving supersymmetry parameter need to be consistent.8). and.6) oxidizes back up to one of the singlecharge brane solutions shown in Figure 7. any solution shown in Figure 7 may be reduced down to a corresponding noncompact σmodel solution of this type. for the groups G and H ′ given in Table 5. We saw in subsection 7. moreover. An extension 73.74 of this σmodel picture allows for solutions involving multiple harmonic maps σa (y). giving rise to intersectingbrane solutions of the types discussed in Section 7. in order for some portion of the rigid supersymmetry to remain unbroken. the σmodel equations for the matrix ﬁelds M (y m ) are simply written ∇i (M −1 ∂i M ) = 0 . In the σmodel picture. 97 (10. and. now becomes clear: only on such a noncompact manifold does one have nontrivial null geodesics as required by the gravitational constraint (10. This sequence of σmodel maps is sketched in Figure 8. As with the intersectingbrane solutions.
The matrix current then becomes M −1 ∂i M = a B a ∂i σa − 1 2 c>b b [B b . these constructions make quite clear the places where assumptions have been made that are more stringent than are really necessary. i. This condition allows one to rewrite (10. tr(B a B b ) = 0 .Points on the geodesic submanifold with aﬃne parameters σa may be written M = A exp( a B a σa ) . (10.b tr(B a B b )∂i σa ∂j σb = 0 .11) as 1 M = A exp − 2 (10. The compatibility condition between these velocities is given by the doublecommutator condition 74 [[B a . The constraint imposed by the gravitational equation is Rij = 1 4 a. while an initial point on these geodesics is speciﬁed by the constant matrix A.16) The general set of stationary multicharge brane solutions is thus obtained in the σmodel construction by identifying the set of totally null. B c ](σb ∂i σc − σc ∂i σb ) .e. B c ] = 0 .12) [B b .11).13) where the ﬁrst factor commutes with the B a as a result of (10. (10. (10. Aside from the elegance of the above σmodel picture of the equations governing BPS brane solutions.e.15) which is satisﬁed provided the geodesics parametrized by the σa are null and orthogonal. (10.12). B c ]σb σc c>b b a exp(B a σa ) .12). (10. they are harmonic maps from the Euclidean overalltransverse space of the y m into the geodesic submanifold (10. One example of this is the assumption that the transversespace 98 . B b ].14) and this is then seen to be conserved provided the σa satisfy ∇2 σa (y) = 0. i.11) where the constant matrices B a give the velocities for the various geodesics parametrized by the σa . totally geodesic submanifolds of G/H ′ such that the velocity vectors satisfy the compatibility condition (10.
geometry is ﬂat. which we have only brieﬂy presented here in Section 7.80 The original and potentially the most physically relevant example of this occurs in the theory of stacks of N D3 branes in type IIB supergravity. 21 Yet another aspect of this subject that we have not dwelt upon here is the intrinsically stringtheoretic side. and can involve fractions of preserved supersymmetry other than inverse powers of 2. In type IIB string theory. with a correspondingly covariantized constraint for the null geodesics on the noncompact manifold G/H ′ . In particular. Another aspect of the pbrane story.78 11 Concluding remarks In this review. for this. and for the implications of charge conservation in determining the allowed intersections to Refs. and in particular to the 99 . one could just as well have a more general Ricciﬂat domainspace geometry. This is clearly more restrictive than is really necessary. we refer the reader to Ref. N = 4 super Maxwell theory. These include 78 also intersections at angles other than 90◦ .79 An important extension of this subject is the remarkable duality between worldvolume dynamics and bulk dynamics known as AdS/CFT duality. the real fascination of this subject lies in its connection to the emerging picture in string theory/quantum gravity. There are wider classes of worldvolume actions than we have considered. For a fuller treatment of some of these subjects.48. the reader is referred to Ref. the extra excitations coming from short strings linking the leaves of the stack extend the corresponding U(1)N abelian gauge symmetry to a full nonabelian U(N ).4). or to the discussions of κsymmetric actions for cases involving R–R sector antisymmetrictensor ﬁelds.77. we have discussed principally the structure of classical pbrane solutions to supergravity theories. For a fuller treatment. Some topics that deserve a fuller treatment have only been touched upon here.7 . is the large family of intersecting branes. in which some of the BPS supergravity solutions that we have discussed appear as Dirichlet surfaces on which open strings can end. Eq. A single D3 brane has a worldvolume theory described by D = 4.9 Of course. The use of more general Ricciﬂat transverse geometries is at the basis of “generalized pbrane” solutions that have been considered in Refs. and example of a worldvolume multiplet involving more than spinors and scalars. involving excitations other than worldvolume spinors and scalars. the reader is referred to Refs 4. (10.50 . we have covered only in a cursory fashion the important topic of κinvariant pbrane worldvolume actions. The AdS/CFT duality expresses the equivalence between the dynamics of these solitonic excitations on the worldvolume and full type IIB string theory in the spacetime bulk.
593 (1985). Alberta. K. 409 (1978). June. 6. B 262. E8 × E8 and SO(32) heterotic. D = 3.N. B 158. Bernard de Wit. M. Duﬀ. Gibbons. IIB.K.W. 11. Friedan. “Are 2branes better than 1?” in Proc. Dabholkar. Phys. Nucl. 1 (1985). “String solitons. Rev. 213 (1995). E. Harvey and F. Edmonton. 3763 (1982).A. Lett. Duﬀ.” Int.” Physics Reports 259. Martinec and M. A 4. C. A15. Phys. Acknowledgments The author would like to acknowledge helpful conversations with Marcus Bremer. J.J. KEK library accession number 8801076. B 261. Phys. Townsend. D. D. M. Ibort and M.A.K. uniting the underlying type IIA. D = 4 mod8. c u George Papadopoulos. B 76. 10. “The Ultraviolet Properties of Supersymmetric Field Theories. Mike Duﬀ.R.S. Scherk. Nucl.A. B. “Tasi Lectures on Dbranes.S. 1871 (1989). 5. Phys. “Three lectures on supersymmetry and extended objects. Chris Pope. 8. Stelle. the duality symmetries that we have discussed in Section 8 play an essential part. and Paul Townsend.A.rˆles that BPS supergravity solutions play as states stable against the eﬀects of o quantum corrections. 1992). Page. Mod. Howe and K. R.S. References 1. Tseytlin. In this emerging picture. 12.” Nucl. 1993). Gary Gibbons. A. Fradkin and A. The usefulness of classical supergravity considerations in probing the structure of this emerging “M theory” is one of the major surprises of the subject. Khuri and J. “Superstrings and Solitons. P. Stelle and P. Phys. B 340. 33 (1990). Phys. eds L. Fran¸ois Englert. Quantum Groups and Quantum Field Theories (23 rd GIFT Seminar on Theoretical Physics. Phys. Cremmer.J.” hepth/9611203. Lu. Polchinski. July 1987. CAP Summer Institute. Perry. Callan. van Holten and A. 9. Lett. Rodriguez (Kluwer.” J. Ruiz Ruiz. E.” hepth/9611050. E.S. 3. “N = 1 Supersymmetry Algebras in D = 2. which then also has a phase with D = 11 supergravity as its ﬁeldtheory limit. Townsend. 7. 2. 2976 (1983). Julia and J. and also the type I string theories into one overall theory. Salamanca. Hong L¨ . “Supermembranes. van Proeyen. 4. J. D 28. hepth/9412184. J.” in Integrable Systems.X. 100 . Phys. G. 316 (1985). P. J.
“Bound states of strings and pbranes. 293 (1985). 75 (1987). I. E.K. 21. K. 16. Pope and K. 301 (1983). S. Lett. hepth/9511203. J. G. B 126. Gibbons.. hepth/9508042. “Brane surgery with Thom classes. Lett.S. 2. hepth/9605082. Sept.N. T. ibid. Gibbons and P. M. “Vertical Versus Diagonal u Dimensional Reduction for pbranes. P. hepth/9905073.R. 143 (1983). Phys. Phys.5. Misner. Phys.” Nucl.C.” JHEP 9905 020 (1999). Phys. 70 (1987). Sezgin and K. 19. “Branes within branes. Nucl.S. Hawking and G. “Dilatonic pbrane u solitons.K.”in Proc. C. Rev. Quantum 101 . Bergshoeﬀ. Huq and M. West. Quantum Grav. hepth/9510135. Phys. 181 (1984). Nucl.N. Duﬀ. 112 (1984). E. 325 (1984). 26. Stelle. Schwarz and P. B 276.J. R.” Nucl. B 456. La LondelesMaures. Gravitation (W. Pope. Rev.S. hepth/9302049. “Brane Surgery.R. Phys.S. Townsend. Schwarz. B 189. Class. Townsend. Ellis. Duﬀ. 321 (1994). H.S. B 226.C. Thorne and J. 46 (1996). European Res.A. Townsend. Sezgin and K. M.B. L¨. 17.K. B 253. 22. 269 (1983).” Nucl. Lett.W. “Stainless Super u pbranes. B 191. B 481.K. Lett. 1973). “Multimembrane solutions of D = 11 Supergravity. Box 14. Witten. Namazie. (Cambridge University Press. 15. Townsend. on Advanced Quantum Field Theory.T.S.A. Lett. 71.” Phys. Lett. Pope. M. 335 (1996).C. G. Wheeler. 49 (1992). Conf. 23. 20. F. C. C. Inami and K. Stelle. Phys. G. 25. Papadopoulos. B 122. Nucl. Class.” Phys. E. P.C. Freeman and Co. 2.” u Phys. Lett. Schwarz. Horowitz and P. G. 1996. H.H.13. 24. B 371. Lett. C. 1973). San Francisco. Phys. M.S. Pernici. D 30. Campbell and P. Phys.K. West. 18.” Nucl. Giani and M. B 332. Phys. H. 3754 (1993). Phys. Gibbons and P. Townsend. 597 (1985). “Black pbrane solutions of D = 11 supergravity theory. G.H. E. Phys.S. 14.W. P. West. Stelle. Stelle. M. B 243.N. Stelle.J. Howe and P.” hepth/9512077. The LargeScale Structure of SpaceTime. E. “Superstrings in D = 10 from Supermembranes in D = 11. B 460. hepth/9609217. Sezgin and P.W.W. hepth/9405124.W. M. L¨. Phys. 669 (1996).H.H. Howe. B 238. 113 (1991). L¨. 313 (1996). Green and J. Duﬀ and K. G¨ven.J. Douglas.F. J.
J. E. 40.K. Achucarro. B 82. Izquierdo and P. Phys.S. L¨ and C. B 370. Pope. hepth/9512012. Pasti. M.M. L¨ and C. Bandos. Lu.L. Duﬀ. Perry. I. H. 79 (1995). Green and M. B 382. D. Lett. Rev. C. D.M. 28. Lett.J.W. 127 (1996). Evans. G. 441 (1987). Phys. Volkov. Duﬀ and J. Lett.V. Papadopoulos and P. Townsend Ann. 36. Myers. J. Nucl. L155 (1977). J. Volkov. Phys. Phys. M. Phys.A. 12. Gibbons. 30.N. Townsend. G. G. H. u “Domain Walls in Massive Supergravities.M. Phys. Yau. Pope. B 109.M.K. Pope. 34. pbrane solitons in maximal supergravities.P. hepth 9604052. E. “Duality of Type II 7branes and 8branes.” u Phys. H. Sorokin. N. Sezgin and P. Nucl.P. Pope. Lett. 197 (1991). M. B 486. 27. B 465. Townsend and D. J. I. hepth/9211056. hepth/9501113. Bergshoeﬀ. B 360. Lett. J. P. 37 (1996). B 470. P. C. Phys. 73 (1996). 185. A. 42. hepth/9410073. Phys. Tucker. Gauntlett. Phys.” hepth/9601089. Townsend. 363 (1993). Cowdall. “Superstrings and Supermembranes in the Doubly Supersymmetric Geometrical Approach.R.T. M. Schwarz. B 387. P. 330 (1988).K. “An approach to the classiﬁcation of pbrane u solitons. hepth/9601150. K. 113 (1996). de Roo. hepth/9511080.B. Phys. 315 (1992).V. 37. Nucl. B 337. Hull. Harvey and J. 29.Grav. Lett.W.H. 63. G.” Nucl.K. Liu.” Nucl. Kaloper. 301 (1994). Stelle and P. Phys. M.R. Townsend. 60 (1979). E. B 198. Nucl. 41. Phys. hepth/9502141. B 446. L¨. 2443 (1989). P. L¨ and C. Khuri and R. B 416. B. 31. u Phys. Phys. Sorokin and and D. hepth/9608173. Nucl.W. 269 (1995). 38. 297 (1995). hepth/9205081. R. J. H. “On generalized axion 102 . 32.N. A 10. Bandos.C. hepth/9306052.N. 33. Green. Gibbons and C.” Nucl. 49 (1997). Bergshoeﬀ. J. Phys.K. Strominger.B. de Azcarraga. A. 190 (1982). Greene.X.” Phys. Nucl.J. R. 39. “The Black Branes of M theory. Gauntlett. Wiltshire. M. Khuri. Vafa and ST. 35. 1 (1990). Shapere. Phys.S. Scherk and J.N. Lett. Howe and R. Tonin and D.J. “On the Generalized Action Principle for Superstrings and Superbranes. J. Horowitz and A. B 409.A. B 352.
E. D 55. A. Klebanov and A. B 372. Gibbons. Shiraishi. Tseytlin. Nat.” Nucl.” Nucl. Coles and G. Rahmfeld. 51. Pope. Khviengia.reductions. H. Phys. B 481. Phys. Sci.” hepth/9705011. “From topology to generalized u dimensional reduction. G. Gauntlett. Sorkin.” Nucl. L¨ and C. Papadopoulos. 3785 (1997). Phys. R. 9.” Nucl. Khviengia. hepth/9803066. Perry. B 508. 87 (1983). 21 (1996). 399 (1993). “Intersecting Dbranes in ten dimensions and six dimensions. H.H. Duﬀ and J. K. Lett. Pope.R.” Phys. hepth/9609212.” Nucl. Lett. Phys. Phys. hepth/9604090. Phys.N. I. Bergshoeﬀ and B.P. hepth/9512089. “HKT and OKT geometries on soliton black hole moduli spaces. Phys. hepth/9605077. Acad. J. I. hepth/9604168. hepth/9611079. 141 (1997). hepth/9706207. “Extreme domain wall — black hole complementarity in N = 1 supergravity with a general dilaton coupling. 159 (1997). 51. “Bound States of Black Holes and Other pbranes. K. Kastor and J.W. 29 (1983). 623 (1997).” Phys.” Nucl. H. J. “Harmonic superpositions of Mbranes. “Domain walls from Mbranes. M. 49. Phys.A. “Intersecting u Mbranes and Bound States. Nucl. 1 (1923).” Phys. Lett. “Overlapping branes in Mtheory. Pope.” Nucl. D. Janssen. J. u Lett. 273 (1996). 297 (1998). 160 (1994). Rev. M.S. Berndt. Tseytlin. B 341. Z. B 226. 179 (1996). 46. 53. Soleng. R. Rahmfeld. Lett.” Phys. Papadopoulos and P. “ ‘No force’ condition and BPS combinations of pbranes in eleven dimensions and ten dimensions.J. 52.J. Nucl. Gross and M. Phys. G. “Intersecting Mbranes as fourdimensional black holes. hepth/9611134. G.” Phys. B 402. 44. hepth/9605085. “The geometry of onedimensional 103 . B 388.N. H. 48. N. Phys. “Supergravity domain walls. Brinkmann. 43.A.J. Stelle. Tseytlin. M. 50. hepth/9603087. 332 (1996). 54. 278 (1997). Papadopoulos and K. 1087 (1997). L¨ and C. Rev. B 380. D. 47. B 475. B 478. Proc. Lavrinenko.” Mod. Cvetic. “Intersecting Branes. Lett. A 12. B 492. L¨ and C. 198 (1996). 149 (1996). Phys. Gauntlett. A. B 428. Townsend.W. Phys. B 475. B 487. Traschen. Lett.” Physics Reports 282.N. “Intersecting Mbranes. Cvetic and H. 45. 544 (1996).K.V.
109 (1995). A.” Nucl.S. 77 (1994). Pope and K. 62. Stelle.” Nucl. Phys. Lett.” Phys. C. E.55. “Weyl Group Invariance and pbrane u Multiplets. 69 (1987). Z) multiplet of type IIB superstrings.S. 69 (1986). Porrati and E. 56. hepth/9605085. Cremmer. Phys. H. S. Henneaux and C. Julia. “Target space duality in string theory. B 194. Pope and K.K.” Phys.” Phys. hepth/9707207. C. T. B 481. H. Schwarz.H.” Phys. Teitelboim. 427 (1990). 61.J. R. 57. “SL(2. Ortin. Bremer. B 159.M. 109 (1993). hepth/9207053. B 520. 66.N. B 360. Nucl. E. hepth/9401139. Phys. hepth/9410167. Phys. M.N. 259 (1998). 5079 (1993). Bergshoeﬀ. “Duality in the type II 104 .M.” Int. Nucl. A 8. Phys. 332 (1996). Buscher. 63. Erratum ibid. 67. M. Quantum Grav. G. 13 (1995). Lett. 59.” Nucl. 64. C. Pope and K. Phys. B 404. B 529. hepth/9710244. E. hepth/9510086. Teitelboim. Phys. 132 (1998). 364. 141 (1979). “Spectrumgenerating u symmetries for BPS Solitons.” Nucl. Deser. C. “Bound States of Black Holes and Other pbranes.” Class. “The power of Mtheory. selfduality.” Phys.K.S.M. B 460. Schwarz. Lett. Rahmfeld. hepth/9602140. “String theory dynamics in various dimensions. B 476. 7. supersymmetric nonlinear sigma models. 68. B 367. A. Cremmer and B. 80 (1997). Z) duality and magnetically charged strings. Lambert. H. 85 (1995). Rabinovici. J. C. Mod. “Duality. Izquierdo. A. “Magnetic monopoles from antisymmetric tensor gauge ﬁelds. M. D 31. Witten. L¨. “A symmetry of the string background ﬁeld equations. Lett. Phys. “An SL(2. 63. E.N. Townsend. Papadopoulos and P. 560 (1996). C. sources and charge quantisation in Abelian Nform theories. Lett. Phys. hepth/9508177. 97 (1996).H.” Physics Reports 244. 70. Hull and P. J. Phys. N. 89 (1996). Rev. L¨. “Dirac quantisation u conditions and KaluzaKlein reduction. L¨. M. Hull and T. 65. B 400. B 438. “Electric and magnetic duality in string theory. 69. Phys. Stelle.” Nucl. Sen. B 443. J. Townsend. Giveon. 252 (1995). J. “Dyonic Membranes.D. hepth/9702184.H. 1921 (1985). hepth/9508143. Stelle. Duﬀ and J. 60. B 67.” Nucl. Nepomechie. hepth/9302038. Gomberoﬀ. 58.” Nucl.
79. J. E. Leigh. Lett. B 390.” Nucl. 145 (1997). Townsend. B 105 .” Phys. Sorokin and M. B 490. I. 120. G. Pope.” Int. H. Julia. 163 (1997). Phys. hepth/9605053. Phys. P. 133 (1997). 265 (1996). Papadopoulos and P. D 58. 437 (1997). 179 (1997). hepth/9710119.A.V.” Phys.S. 62 (1997). C. B. superstring eﬀective action. Bergshoeﬀ and P. hepth/9801160. M. G. Ann. Lett. Nilsson. P. 122001 (1998). B 490. hepth/9611159. Cederwall. “Generalized Action Principle and Superﬁeld Equations of Motion for d = 10 Dpbranes. Sezgin. Khviengia.S. hepth/9611173. N. D. Z. von Gussich. B. “Dbrane actions with local kappasymmetry. Aganagic. Howe and E.R. B 451. Rytchkov.N. 133 (1997). hepth/9607227. “Generating branes via sigmamodels. Sundell and A. L¨ and C. hepth/9512153.” Phys. Sezgin. B. hepth/9607043. “Branes intersecting at angles. hepth/9610249. Maison and G. 73 (1998). Popescu and J. D. der Physik (Leipzig) 24. Lett. E. 72. “Stationary BPS solutions to dilatonaxion gravity. Phys.71. B 480. Nilsson and A. L¨ and C. “HyperKahler manifolds and multiply intersecting branes. Pope. Pope.” Nucl. Cederwall. P. Kramer. Rev. 253 (1988). D 54.” Nucl.K. Phys. “Super Dbranes. 547 (1995).” Phys. Gauntlett. 75. Clement and D.” Nucl. Math.” Nucl. P. 74. B 523. Townsend. G. Phys.” Phys. B 393.E. hepth/9605077. Howe and E. Neugebaur and D. M. “The dirichlet super pbranes in tendimensional Type IIA and IIB supergravity. 155 (1996). H. u I. Phys. A 12. L¨ and C.N. hepth/9611008. 62 (1969). B 477. 78.P.” Nucl. 21 (1996). A. B 394. Gal’tsov and O.E.K. Phys. M.N. 73. D. B 500. G. Becker. Comm. “D = 11. A.” Nucl. Khviengia. “Superbranes.” Nucl. Westerberg. 311 (1997). 6136 (1996). “Mtheory on eightmanifolds.W. Gibbons. B 490. p = 5. “Multiscalar pbrane solitons. Bandos. Cremmer. K. Gal’tsov. Mod. hepth/9504081.” Nucl.G.W. Schwarz. “The Dirichlet Super Three Brane in TenDimensional Type IIB Supergravity. “Dualisation of dualities. Phys.W. hepth/9606139. Phys. Douglas and R.” Phys.H. Rev. Breitenlohner. Berkooz. Tonin. hepth/9610148. Becker and M. u Phys. 77. Gibbons. 76. Phys. B 388. J. hepth/9702202. “Intersecting u Mbranes and bound states. M. Lett. von Gussich. Westerberg. M. H.
497. Klebanov. “Tasi Lectures: Introduction to the AdSCFT correspondence. I. 106 .” hepth/0009139. 275 (1997). hepth/9701127. 80.
This action might not be possible to undo. Are you sure you want to continue?
We've moved you to where you read on your other device.
Get the full title to continue listening from where you left off, or restart the preview.