CERN-TH/98-80

Imperial/TP/97-98/30
hep-th/9803116

BPS BRANES IN SUPERGRAVITY ∗

K.S. STELLE
The Blackett Laboratory, Imperial College,
Prince Consort Road, London SW7 2BZ, UK
and
arXiv:hep-th/9803116v3 23 Sep 2009

TH Division, CERN
CH-1211 Geneva 23, Switzerland

DEDICATED TO THE MEMORY OF ABDUS SALAM

This review considers the properties of classical solutions to supergravity theories

with partially unbroken supersymmetry. These solutions saturate Bogomol’ny-
Prasad-Sommerfield bounds on their energy densities and are the carriers of
the p-form charges that appear in the supersymmetry algebra. The simplest
such solutions have the character of (p + 1)-dimensional Poincaré-invariant hy-
perplanes in spacetime, i.e. p-branes. Topics covered include the relations between
mass densities, charge densities and the preservation of unbroken supersymmetry;
interpolating-soliton structure; diagonal and vertical Kaluza-Klein reduction fam-
ilies; multiple-charge solutions and the four D = 11 elements; duality-symmetry
multiplets; charge quantisation; low-velocity scattering and the geometry of world-
volume supersymmetric σ-models; and the target-space geometry of BPS instanton
solutions obtained by the dimensional reduction of static p-branes.

Contents

1 Introduction 3

2 The p-brane ansatz 8

2.1 Single-charge action and field equations . . . . . . . . . . . . . 8
2.2 Electric and magnetic ansätze . . . . . . . . . . . . . . . . . . . 8
2.3 Curvature components and p-brane equations . . . . . . . . . . 10
2.4 p-brane solutions . . . . . . . . . . . . . . . . . . . . . . . . . . 12

3 D = 11 examples 14
3.1 D = 11 Elementary/electric 2-brane . . . . . . . . . . . . . . . 15
3.2 D = 11 Solitonic/magnetic 5-brane . . . . . . . . . . . . . . . . 19
3.3 Black branes . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

4 Charges, Masses and Supersymmetry 23

4.1 p-form charges . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 p-brane mass densities . . . . . . . . . . . . . . . . . . . . . . . 27
4.3 p-brane charges . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.4 Preserved supersymmetry . . . . . . . . . . . . . . . . . . . . . 28
∗ Based upon lectures given at the 1996 and 1997 ICTP Summer Schools in High Energy
Physics and Cosmology.

1
5 Kaluza-Klein dimensional reduction 32
5.1 Multiple field-strength solutions and the single-charge truncation 35
5.2 Diagonal dimensional reduction of p-branes . . . . . . . . . . . 39
5.3 Multi-center solutions and vertical dimensional reduction . . . 40
5.4 The geometry of (D − 3)-branes . . . . . . . . . . . . . . . . . 44
5.5 Beyond the (D − 3)-brane barrier: Scherk-Schwarz reduction
and domain walls . . . . . . . . . . . . . . . . . . . . . . . . . . 46

6 Intersecting branes, scattering branes 53

6.1 Multiple component solutions . . . . . . . . . . . . . . . . . . . 53
6.2 Intersecting branes and the four elements in D = 11 . . . . . . 55
6.3 Brane probes, scattering branes and modulus σ-model geometry 60

7 Duality symmetries and charge quantisation 65

7.1 An example of duality symmetry: D = 8 supergravity . . . . . 66
7.2 p-form charge quantisation conditions . . . . . . . . . . . . . . 68
7.3 Charge quantisation conditions and dimensional reduction . . . 71
7.4 Counting p-branes . . . . . . . . . . . . . . . . . . . . . . . . . 73
7.5 The charge lattice . . . . . . . . . . . . . . . . . . . . . . . . . 77

8 Local versus active dualities 79

8.1 The symmetries of type IIB supergravity . . . . . . . . . . . . . 81
8.2 Active duality symmetries . . . . . . . . . . . . . . . . . . . . . 84

9 Non-compact σ-models, null geodesics, and harmonic maps 88

10 Concluding remarks 93

2
1 Introduction
Let us begin from the bosonic sector of D = 11 supergravity,1
Z n√ o
I11 = d11 x −g(R − 48 1 2
F[4] ) + 61 F[4] ∧ F[4] ∧ A[3] . (1.1)

In addition to the metric, one has a 3-form antisymmetric-tensor 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 gravitino-dependent
sector is included.
The equation of motion for the A[3] gauge potential is
d ∗F[4] + 12 F[4] ∧ F[4] = 0 ; (1.2)
this equation of motion gives rise to the conservation of an “electric” type
charge 2 Z
U= (∗F[4] + 21 A[3] ∧ F[4] ) , (1.3)
∂M8
where the integral of the 7-form integrand is over the boundary at infinity of
an arbitrary infinite spacelike 8-dimensional subspace of D = 11 spacetime.
Another conserved charge relies on the Bianchi identity dF[4] = 0 for its con-
servation, Z
V = F[4] , (1.4)
e5
∂M
where the surface integral is now taken over the boundary at infinity of a
spacelike 5-dimensional subspace.
Charges such as (1.3, 1.4) can occur on the right-hand side of the super-
symmetry algebra,a3
{Q, Q} = C(ΓA PA + ΓAB UAB + ΓABCDE VABCDE ) , (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 right-hand side have a well-
defined meaning, but for extended objects such as p-branes, 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
left-hand side of (1.5) in such cases. This would then require then an infinite normalization
factor for the algebra, whose removal requires the right-hand 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, PA is the energy-momentum 11-
vector and UAB and VABCDE are 2-form and 5-form 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 32-component Majorana spinor,
the LHS of (1.5) has 528 components. The symmetric spinor matrices CΓA ,
CΓAB and CΓABCDE on the RHS of (1.5) also have a total of 528 independent
components: 11 for the momentum PA , 55 for the “electric” charge UAB and
462 for the “magnetic” charge VABCDE .
Now the question arises as to the relation between the charges U and V in
(1.3, 1.4) and the 2-form and 5-form 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 M8 , M f5 that do not fill out a whole
10-dimensional 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 UAB and VABCDE may be guessed from the
2-fold and 5-fold ways that the corresponding 8 and 5 dimensional integration
volumes may be embedded into a 10-dimensional spacelike hypersurface. We
shall see in Section 4 that this is too naı̈ve, however: it masks an important
topological aspect of both the electric charge UAB and the magnetic charge
VABCDE . 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
∂M8 , ∂ M f5 . 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 + 1-dimensional Poincaré-invariant
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 field-theoretic “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 realisation that supergravity theories do not by themselves constitute
acceptable starting points for a quantum theory of gravity came somewhat
before the realisation 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 long-wavelength effective-field-theory limits of string theories.
To see how this happens, consider the σ-model action 10 that describes a bosonic
string moving in a background “condensate” of its own massless modes (gM N ,
AM N , φ):
Z
1 √
I= d2 z γ [γ ij ∂i xM ∂j xN gM N (x)
4πα′
+iǫij ∂i xM ∂j xN AM N (x) + α′ R(γ)φ(x)] . (1.6)

Every string theory contains a sector described by fields (gM N , AM N , φ); these
are the only fields that couple directly to the string worldsheet. In superstring
theories, this sector is called the Neveu-Schwarz/Neveu-Schwarz (NS–NS) sec-
tor.
The σ-model action (1.6) is classically invariant under the worldsheet Weyl
symmetry γij → Λ2 (z)γij . Requiring cancellation of the anomalies in this
symmetry at the quantum level gives differential-equation restrictions on the
background fields (gM N , AM N , 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 field-theory effective action
Z h
√
Ieff = dD x −ge−2φ (D − 26) − 32 α′ (R + 4∇2 φ − 4(∇φ)2
i
1
− 12 FM N P F M N P + O(α′ )2 , (1.7)

where FM N P = ∂M AN P + ∂N AP M + ∂P AM N is the 3-form field strength for the

AM N 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 σ-model calculations. It is not written in the form generally preferred
by relativists, which has a clean Einstein-Hilbert term free from exponential
prefactors like e−2φ . One may rewrite the effective action in a different frame
by making a Weyl-rescaling field redefinition gM N → eλφ gM N . Ieff as written
in (1.7) is in the string frame; after an integration by parts, it takes the form,
specialising now to D = 10,
Z p h i
I string = d10 x −g (s) e−2φ R(g (s) ) + 4∇M φ∇M φ − 12 1
FM N P F M N P . (1.8)

After making the transformation

(e) (s)
gM N = e−φ/2 gM N , (1.9)
one obtains the Einstein frame action,
Z p h i
1 −φ
I Einstein = d10 x −g (e) R(g (e) )− 21 ∇M φ∇M φ− 12 e FM N P F M N P , (1.10)

(e)
where the indices are now raised and lowered with gM N . To understand how this
Weyl rescaling works, note that under x-independent rescalings, the connection
ΓM N P is invariant. This carries over also to terms with φ undifferentiated,
which emerge from the eλφ 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 positive-energy 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 5) the process of Kaluza-Klein dimensional reduction
that relates theories in different dimensions of spacetime. For the present, we
note that upon specifying the Kaluza-Klein ansatz expressing ds211 in terms of
ds210 , the Kaluza-Klein vector AM and the dilaton φ,
ds211 = e−φ/6 ds210 + e4φ/3 (dz + AM dxM )2 M = 0, 1, . . . , 9, (1.11)
the bosonic D = 11 action (1.1) reduces to the Einstein-frame type IIA bosonic
action 14
Z p n
Einstein

IIIA = d10 x −g (e) R(g (e) ) − 12 ∇M φ∇M φ − 121 −φ
e FM N P F M N P
1 1 o
− eφ/2 FM N P Q F MN P Q − e3φ/2 FM N F M N + LF F A , (1.12)
48 4
6
where FM N is the field strength for the Kaluza-Klein vector AM .

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 Chern-Simons 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

Z p n
string  
IIIA = d10 x −g (s) e−2φ R(g (e) ) + 4∇M φ∇M φ − 1
12 FM N P F
MNP

1 1 o
− FM N P Q F M N P Q − FM N F M N + LF F A . (1.13)
48 4

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 coupling-constant factor for the Yang-Mills ac-
tion, one sees that the asymptotic value eφ∞ plays the rôle of the string-theory
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 non-perturbative in-
formation about string theory. In particular, this will arise in the study of
p-brane 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 zero-form (i.e. a pseudoscalar field),
R
F[3] = dAR NS
[2] , a second 3-form field strength making a pair together with F[3]
from the NS–NS sector, and F[5] = dA[4] , which is a self-dual 5-form 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 7-form ∗F[7] . The upshot is that antisymmetric-tensor gauge field strengths
of diverse ranks need to be taken into account when searching for solutions to
string-theory effective field equations. These field strengths will play an essen-
tial rôle in supporting the p-brane solutions that we shall now describe.

7
2 The p-brane ansatz
2.1 Single-charge action and field equations
We have seen that one needs to consider effective theories containing gravity,
various ranks of antisymmetric-tensor 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
gM N , 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
Z
√ h i
I = DD x −g R − 21 ∇M φ∇M φ − 2n! 1 aφ 2
e F[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 rôle 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 F F A Chern-Simons term
in the action. These are also consistently excluded in the truncation to the
single-charge 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:
1
RMN = 2 ∂M φ∂N φ + SM N (2.2a)
1 n−1
SMN = eaφ (FM ··· FN ··· − F 2 gM N ) (2.2b)
2(n − 1)! n(D − 2)
∇M 1 (eaφ F M 1 ···M n ) = 0 (2.2c)
a aφ 2
φ = e F . (2.2d)
2n!

2.2 Electric and magnetic ansätze

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 translationally-symmetric ones. These restrictions can subse-
quently be relaxed in generalizations of the basic class of p-brane solutions
that we shall discuss here. For this basic class of solutions, we make an ansatz
requiring (Poincaré)d × SO(D − d) symmetry. One may view the sought-for
solutions as flat d = p + 1 dimensional hyperplanes embedded in the ambi-
ent D-dimensional spacetime; these hyperplanes may in turn be viewed as the
histories, or worldvolumes, of p-dimensional spatial surfaces. Accordingly, let
the spacetime coordinates be split into two ranges: xM = (xµ , y m ), where xµ
(µ = 0, 1, · · · , p = d − 1) are coordinates adapted to the (Poincaré)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 (Poincaré)d ×SO(D−
d) symmetry is 12

ds2 = e2A(r) dxµ dxν ηµν + e2B(r) dy m dy n δmn

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 xµ and SO(D − d) symmetry in the transverse
directions y m is guaranteed.
The corresponding ansatz for the scalar field φ(xM ) 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 1-form 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 del = n − 1
dimensional worldvolume. This is what we shall call the “elementary,” or
“electric” ansatz:

Aµ1 ···µn−1 = ǫµ1 ···µn−1 eC(r) , others zero. (2.4)

SO(D − d) isotropicity and (Poincaré)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:
(el)
Fmµ1 ···µn−1 = ǫµ1 ···µn−1 ∂m eC(r) , others zero. (2.5)

The worldvolume dimension for the elementary ansatz (2.4, 2.5) is clearly del =
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 dso = 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:

(mag) yp
Fm1 ···mn = λǫm1 ···mn p , others zero, (2.6)
rn+1
where the magnetic-charge 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 dso = d˜el ≡ D−del −2; note also that this relation is idempotent,
f˜ = d.
i.e. (d)

2.3 Curvature components and p-brane equations

In order to write out the field equations after insertion of the above ansätze,
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 tangent-space
indices denoted by underlined indices:

gM N = eM E eN F ηE F . (2.7)


one finds ∂q Fm1 ···mn = r −(n+1) ǫm1 ···mn q − (n + 1)ǫm1 ···mn p y p yq /r 2 ; upon
c Specifically,

taking the totally antisymmetrised combination [qm1 · · · mn ], the factor

P of (n + 1) is evened
out between the two terms and then one finds from cycling a factor y my 2
m m = r , thus
obtaining cancellation.

10
Next, one constructs the corresponding 1-forms: eE = dxM eM E . Splitting up
the tangent-space indices E = (µ, m) similarly to the world indices M = (µ, m),
we have for our ansätze the vielbein 1-forms
eµ = eA(r) dxµ , em = eB(r) dy m . (2.8)
The corresponding spin connection 1-forms are determined by the condi-
tion that the torsion vanishes, deE + ω E F ∧ eF = 0, which yields

ωµ ν = 0, ω µ n = e−B(r) ∂n A(r)eµ
ωm n = e−B(r) ∂n B(r)em − e−B(r) ∂m B(r)en . (2.9)
The curvature 2-forms are then given by
EF
R[2] = dω E F + ω E D ∧ ωD F . (2.10)
From the curvature components so obtained, one finds the Ricci tensor com-
ponents
˜
Rµν = −ηµν e2(A−B) (A′′ + d(A′ )2 + dA˜ ′ B ′ + (d + 1) A′ )
r
˜
Rmn = −δmn (B ′′ + dA′ B ′ + d(B ˜ ′ )2 + (2d + 1) B ′ + d A′ ) (2.11)
r r
y m y n ˜ ′′ ˜
− 2 (dB ˜ ′ )2 − d B ′ − d A′ ) ,
+ dA′′ − 2dA′ B ′ + d(A′ )2 − d(B
r r r
˜
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:
˜ d˜
˜ ′ B ′ + (d+1) A′ =
A′′ +d(A′ )2 + dA 2
r 2(D−2) S {µν}
˜ ′ )2 + (2d̃+1) B ′ + d A′
B ′′ +dA′ B ′ + d(B d
= − 2(D−2) S2 {δmn }
r r
˜ ′′ ′′ ′ ′
dB +dA −2dA B +d(A ) − d(B ) ′ 2 ˜ ′ 2
˜
− dr B ′ − dr A′ + 21 (φ′ )2 = 1 2
2S {ym yn }
˜
˜ ′ φ′ + (d+1)
φ′′ +dA′ φ′ + dB ′
= − 12 ςaS 2
r φ {φ}
(2.12)
where ς = ±1 for the elementary/solitonic cases and the source appearing on
the RHS of these equations is
( 1
(e 2 aφ−dA+C )C ′ electric: d = n − 1, ς = +1
S= 1 ˜ (2.13)
λ(e 2 aφ−d̃B )r−d−1 magnetic: d = D − n − 1, ς = −1.

11
2.4 p-brane solutions
The p-brane equations (2.12, 2.13) are still rather daunting. Before we embark
on solving these equations, let us first note a generalisation. Although Eqs
(2.12) have been specifically written for an isotropic p-brane ansatz, one may
recognise 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 rôles 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
˜ ′=0.
dA′ + dB (2.15)
After eliminating B using (2.15), the independent equations become 17

∇2 φ = − 21 ςaS 2 (2.16a)
d˜
∇2 A = S2 (2.16b)
2(D − 2)
˜ ′ )2
d(D − 2)(A′ )2 + 21 d(φ = 1 ˜ 2
dS , (2.16c)
2

where, for spherically-symmetric (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 ansätze by
imposing another linearity condition:
−ςa(D − 2) ′
φ′ = A . (2.17)
d˜
At this stage, it is useful to introduce a new piece of notation, letting

2dd˜
a2 = ∆ − . (2.18)
(D − 2)
With this notation, equation (2.16c) gives
∆(φ′ )2
S2 = , (2.19)
a2
12
 ς∆
so that the remaining equation for φ becomes ∇2 φ + ′ 2
2a (φ ) = 0, which can
be re-expressed as a Laplace equation,d
ς∆
∇2 e 2a φ = 0 . (2.20)

Solving this in the transverse (D − d) dimensions with our assumption of

transverse isotropicity (i.e. spherical symmetry) yields
ς∆ k
e 2a φ ≡ H(y) = 1 + k>0, (2.21)
rd˜
where the constant of integration φ| has been set equal to zero here for
r→∞
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 antisymmetric-tensor gauge field
potential. In this case, it follows from (2.13) that S 2 = eaφ−2dA (C ′ eC )2 .
Combining this with (2.19), one finds the relation
√
∂ C − ∆ − 1 aφ+dA ′
(e ) = e 2 φ (2.22)
∂r a
(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] :
˜ ′ − dA′ + aφ′ ) = 0 .
∇2 C + C ′ (C ′ + dB (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 Laplace-equation solution H(y) (2.21), one finds 7,13
−4d̃ 4d
ds2 = H ∆(D−2) dxµ dxν ηµν + H ∆(D−2) dy m dy m (2.24a)

2a +1, elementary/electric
eφ = H ς∆ ς= (2.24b)
−1, solitonic/magnetic
k
H(y) = 1+ ˜ , (2.24c)
rd
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
2
eC = √ 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= λ. (2.26)
2d˜
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:
2
Fmµ1 ...µn−1 = √ ǫµ1 ...µn−1 ∂m (H −1 ) m = d, . . . , D − 1 electric
(2.27a)
∆
2
Fm1 ...mn = − √ ǫm1 ...mn r ∂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 p-brane 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
˜
a2 = ∆ − 2dd/(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 F F A Chern-Simons 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 F F A term in the action (1.1) vanishes and hence this term
does not make any non-vanishing contribution to the metric field equations for
our ansätze. For the antisymmetric tensor field equation, a further check is
necessary, since there one requires the variation of the F F A term to vanish in
order to consistently ignore it. The field equation for A[3] is (1.2), which when
written out explicitly becomes
√
1
∂M −gF M U V W + ǫU V W x1 x2 x3 x4 y1 y2 y3 y4 Fx1 x2 x3 x4 Fy1 y2 y3 y4 = 0 . (3.1)
2(4!)2

By direct inspection, one sees that the second term in this equation vanishes
for both ansätze.
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 p-branes.

3.1 D = 11 Elementary/electric 2-brane

From our general discussion in Sec. 2, we have the elementary-ansatz solution 18
k − 2/3
ds2 = (1 + r6 ) dxµ dxν ηµν + (1 + rk6 ) 1/3 dy m dy m
Aµνλ = ǫµνλ (1 + rk6 )−1 , other components zero. (3.2)
electric 2-brane: isotropic coordinates

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 RM N P Q and of
the field strength Fmµ1 µ2 µ3 , 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 = x0 =
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 “Schwarzschild-type” coordi-
nates by setting r = (r̃6 − k) 1/6 . The solution then becomes:19
k 2/3 k −2 2
ds2 = (1 − r̃ 6 ) (−dt
2
+ dσ 2 + dρ2 ) + (1 − r̃ 6 ) dr̃ + r̃2 dΩ27
k
Aµνλ = ǫµνλ (1 − r̃6 ) , other components zero, (3.3)
electric 2-brane: Schwarzschild-type coordinates

15
where we have supplied explicit worldvolume coordinates xµ = (t, σ, ρ) and
where dΩ27 is the line element on the unit 7-sphere, corresponding to the bound-
ary ∂M8T of the 11 − 3 = 8 dimensional transverse space.
The Schwarzschild-like coordinates make the surface r̃ = 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 r̃ = k 1/6 is in fact a horizon.
This horizon is degenerate, however. Owing to the 2/3 exponent in the g00
component, curves along the t axis for r̃ < 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 r̃ = k 1/6 (1 − R3 )− 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 − R3 ) 1/6 . In these new coordinates, the
solution becomes 19

ds2 = R2 (−dt2 + dσ 2 + dρ2 ) + 41 k 1/3 R−2 dR2 + k 1/3 dΩ27 (a)
+ 41 k 1/3 [(1 − R3 )− 7/3 − 1]R−2 dR2 + k 1/3 [(1 − R3 )− 1/3 − 1]dΩ27 (b)
Aµνλ = R3 ǫµνλ , other components zero.
electric 2-brane: interpolating coordinates
(3.4)
This form of the solution makes it clearer that the light-cones 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
Robinson-Bertotti 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 r̃ = 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 δ-function source to the solution at r̃ = 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
κ2 T
k= , (3.5)
3Ω7
√
where 1/(2κ2 ) is the coefficient of −gR in the Einstein-Hilbert Lagrangian
and Ω7 is the volume of the unit 7-sphere 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
Reissner-Nordstrom solution of General Relativity.24 This global structure is
summarized by a Carter-Penrose 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 affine-parameter 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
 timelike singularity
at r~ = 0 (R → − ∞)
t = const. hypersurface
+

J
+
H

“throat” P i0 spatial infinity

H

J –
–

R = const. hypersurface

Only the shaded region is covered

by the isotropic coordinates

Figure 1: Carter-Penrose diagram for the D = 11 elementary/electric 2-brane solution.

18
 flat M 11

infinite “throat:”
7
(AdS)4 × S

Figure 2: The D = 11 elementary/electric 2-brane solution interpolates between flat space

at J ± and (AdS)4 × S 7 at the horizon.

3.2 D = 11 Solitonic/magnetic 5-brane

Now consider the 5-brane solution to the D = 11 theory given by the solitonic
ansatz for F[4] . In isotropic coordinates, this solution is a magnetic 5-brane:25

ds2 = (1+ rk3 )− 1/3 dxµ dxν ηµν +(1+ rk3 ) 2/3 dy m dy m µ, ν = 0, · · · , 5
p
Fm1 ···m4 = 3kǫm1 ···m4 p yr5 other components zero.
magnetic 5-brane: 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 = (r̃3 − k) 1/3 and r̃ = k 1/3 (1 − R6 )− 1/3 , one has an overall transformation
k 1/3 R2
r= . (3.7)
(1 − R6 ) 1/3
After these coordinate changes, the metric becomes
h i
4R−2 dΩ2
ds2 = R2 dxµ dxν ηµν + k 2/3 (1−R6 ) 8/3
dR2 + (1−R64) 2/3 .
(3.8)
magnetic 5-brane: interpolating coordinates
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 identify the spacetime region R ≤ 0
with the region R ≥ 0. This identification is analogous e to the identification
one naturally makes for flat space when written in polar coordinates, with the
metric ds2flat = −dt2 + dr2 + r2 d2 . Ostensibly, in these coordinates there ap-
pear to be separate regions of flat space with r >
< 0, but, owing to the existence
of the isometry r → −r, these regions may be identified. Accordingly, in the
solitonic/magnetic 5-brane spacetime, we identify the region −1 < R ≤ 0 with
the region 0 ≤ R < 1. In the asymptotic limit where R → −1, one finds an
asymptotically flat geometry that is indistinguishable from the region where
R → +1, i.e. where r → ∞. Thus, there is no singularity at all in the soli-
tonic/magnetic 5-brane geometry. There is still an infinite “throat,” however,
at the horizon, and the region covered by the isotropic coordinates might again
be sketched as in Figure 2, except now with the asymptotic geometry down
the “throat” being (AdS)7 × S 4 instead of (AdS)4 × S 7 as for the elemen-
tary/electric solution. The Carter-Penrose diagram for the solitonic/magnetic
5-brane solution is given in Figure 3, where the full diagram extends indefi-
nitely by “tiling” the section shown. Upon using the R → −R isometry to
make discrete identifications, however, the whole of the spacetime may be con-
sidered to consist of just region I, which is the region covered by the isotropic
coordinates (3.6).
e In considering this analogy, one should also take into account the possibility of conical
singularities. In the case of flat space, a conical singularity with deficit angle θ = 2π arises
at the origin R = 0 if one chooses not to make a discrete identification of the two regions
>
R < 0. This is most easily seen by considering a combined pair of forward and backward
cones with deficit angle θ < 2π, then taking the limit θ → 2π. In this case, as in the case
of the magnetic 5-brane geometry, the elimination of conical singularities actually requires
making the discrete identification.

20
 )
-1

R
=

=
(R

0
J' +

i0' identified with I

R<0

J
+
0)
=

(R
J
'
–

(R

=
+

1)
(R

H
=
-1
)

spatial infinity,
identified with i0 I
i0
R>0
)
-1

H
–
=

1)
(R

(R
J "+

spatial
(R
=
0)

infinity
J

identified with I
i0"
R<0
J
"
– (R

0
=
=
-1

R
)

Figure 3: Carter-Penrose diagram for the solitonic/magnetic 5-brane solution.

21
 After identification of the R > < 0 regions, the 5-brane spacetime (3.6) is
geodesically complete. Unlike the case of the elementary membrane solution
(3.2, 3.4), one finds in the solitonic/magnetic case that the null geodesics pass-
ing through the horizon at R = 0 continue to evolve in their affine parameters
without bound as R → −1. Thus, the solitonic 5-brane solution is completely
non-singular.
The electric and magnetic D = 11 solutions discussed here and in the
previous subsection are special 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 p-brane 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 5-brane 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 r̃ = 0.
When a 6= 0 and the scalar field associated to the field strength supporting
a solution cannot be consistently set to zero, then the solution is singular 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 Schwarzschild-type coordinates, one finds
the generalized “black brane” solution 27,28
4d̃
Σ+
ds2 = −  4d̃
dt2 + Σ−
∆(D−2)
dxi dxi
1−
∆(D−2)
Σ−
 2a2
−1
∆d̃ 2a2
Σ−
+ dr̃2 + r̃2 Σ−∆d̃ dΩ2D−d−1 (3.9)
Σ+

d̃
Σ± = 1 − rr̃±
ς∆
e 2a φ = Σ−1
− .
black brane: Schwarzschild-type coordinates

The antisymmetric tensor√field strength for this solution corresponds to a

˜ ∆(r+ r− )d̃/2 , either electric or magnetic.
charge parameter λ = 2d/

22
 The characteristic feature of the above “blackened” p-branes is that they
have a nondegenerate, nonsingular outer horizon at r̃ = r+ , at which the light
cones “flip over.” At r̃ = r− , one encounters an inner horizon, which, however,
coincides in general with a curvature singularity. The singular nature of the
solution at r̃ = r− is apparent in the scalar φ in (3.9). For solutions with p ≥ 1,
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 non-extreme
solution (3.9) shows that the “location” of the p-brane 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 p-brane 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 p-brane
solutions, the p-form charges UAB and VABCDE 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 p-form 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
2-brane 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,

23
and thus it may be considered to be the wordvolume of a δ-function source.
The electric source that couples to D = 11 supergravity is the fundamental
supermembrane action,20 whose bosonic part is
Z hq
Isource = Qe d3 ξ − det(∂µ xM ∂ν xN gMN (x))
W3
1 µνρ i
+ ǫ ∂µ xM ∂ν xN ∂ρ xR AMN R (x) . (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. Varying the source action (4.1) with δ/δA[3] , one
obtains the δ-function current
Z
J MN R (z) = Qe δ 3 (z − x(ξ))dxM ∧ dxN ∧ dxR . (4.2)
W3

This current now stands on the RHS of the A[3] equation of motion:

d( ∗F[4] + 21 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,
Z Z
∗ 1
U= J[3] = J 0MN d8 SMN , (4.4)
M8 3! M8

where d8 SMN is the 8-volume element on M8 , specified within a D = 10

spatial section of the supergravity spacetime by a 2-form. The charge derived
in this way from a single 2-brane source is thus U = Qe as expected.
Now consider the effect of making different choices of the M8 integration
volume within the D = 10 spatial spacetime section, as shown in Figure 4. Let
the difference between the surfaces M8 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 Z Z
∗ 1
δU = Lv J[3] = J 0MN v R d7 SMN R , (4.5)
M8 3! ∂M8
where Lv 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 ∂M8 , the difference (4.5) between the

24
charges calculated using the integration volumes M8 and M′8 will vanish.
This divides the electric-charge integration volumes into two topological classes
distinguishing those for which ∂M8 “captures” the p-brane current, as shown
in Figure 4 and giving U = Qe , from those that do not capture the current,
giving U =0.

J
[3]

M 8'

v(x)
M8

Figure 4: Different choices of charge integration volume “capturing” the current J[3] .

The above discussion shows that the orientation-dependence 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 p-brane
spatial surfaces; an integration volume M8 extending out to infinity flips from
the “capturing” class into the “non-capturing” class when ∂M8 crosses the
δ-function surface defined by the current J[3] . The charge thus naturally has
a magnitude |Q[p] | = Qe and a unit p-form orientation Q[p] /|Q[p] | that is pro-
portional to the asymptotic spatial volume form of the p-brane. Both the
magnitude and the orientation of this p-form charge are conserved using the
supergravity equations of motion.
The necessity of considering asymptotic p-brane volume forms arises be-
cause the notion of a p-form charge is not limited to static, flat p-brane solutions

25
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
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 p-brane, the asymptotic orientation of the p-brane volume form must
tend to that of a static flat solution, which plays the rôle of a “BPS vacuum”
in a given p-form charge sector of the theory.
In order to have a non-vanishing value for a charge (1.3) or (1.4) occur-
ring in the supersymmetry algebra (1.5), the p-brane must be either infinite or
wrapped around a compact spacetime dimension. The case of a finite p-brane
is sketched in Figure 5. Since the boundary ∂M of the infinite integration vol-
ume M does not capture the locus where the p-brane current is non-vanishing,
the current calculated using M will vanish as a result. Instead of an infinite
p-brane, one may alternately have a p-brane wrapped around a compact di-
mension of spacetime, so that an integration-volume boundary ∂M8 is still
capable of capturing the p-brane locus (if one considers this case as an infinite,
but periodic, solution, this case may be considered simultaneously with that
of the infinite p-branes). Only in such cases do the p-form charges occurring
in the supersymmetry algebra (1.5) take non-vanishing values.f

Figure 5: Finite p-brane not captured by ∂M, giving zero charge.

f Ifone considers integration volumes that do not extend out to infinity, then one can con-
struct integration surfaces that capture finite p-branes. Such charges do not occur in the
supersymmetry algebra (1.5), but they are still of importance in determining the possible
intersections of p-branes.21

26
4.2 p-brane mass densities
Now let us consider the mass density of a p-brane solution. Since the p-brane
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 p-volume). Since we
are considering solutions in their rest frames, this will also give the value of
mass/(unit p-volume), or tension of the solution. Instead of the standard spa-
tial dD−2 Σa surface integral, this will be a d(D−d−1) Σm surface integral over
the boundary ∂MT 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 dΣµ=i integral,
Z
E= dD−d−1 Σm (∂ n hmn − ∂m hbb ) , (4.6)
∂MT

written for gM N = ηM N +hM N 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 p-brane solution (2.24), one
finds
4kd ˜
8k(d + 12 d)
b
hmn = δ mn , h b = , (4.7)
∆(D − 2)rd˜ ∆(D − 2)rd˜
˜
and, since d(D−d−1) Σm = rd y m dΩ(D−d−1) , one finds
˜ D−d−1
4k dΩ
E= , (4.8)
∆
where
√ ΩD−d−1 is the volume of the S D−d−1 unit sphere. Recalling that k =
˜ we consequently have a relation between the mass per unit p volume
∆λ/(2d),
and the charge parameter of the solution
2λΩD−d−1
E= √ . (4.9)
∆
√
By contrast, the black brane solution (3.9) has E > 2λΩD−d−1 / ∆, so
the extremal
√ p-brane solution (2.24) is seen to saturate the inequality E ≥
2λΩD−d−1 / ∆.

4.3 p-brane 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
R quantity 2 U (1.3).
R ∗ For the p-brane 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 M8 to coincide with the transverse space to the d = 3
membrane worldvolume, M8T . The surface element for this transverse space
is dΣm(7) , so for the p = 2 elementary membrane solution (3.2), one finds
Z
U= dΣm
(7) Fm012 = λΩ7 . (4.10)
∂M8T

Since the D = 11 F[4] field strength supporting this solution has ∆ = 4, the
mass/charge relation is
E = U = λΩ7 . (4.11)
Thus, like the classic extreme Reissner-Nordstrom black-hole solution to which
it is strongly related (as can be seen from the Carter-Penrose 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 5-brane
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 f5 = M5T . Then,
we have Z
V = dΣm(4) ǫmnpqr F
npqr
= λΩ4 . (4.12)
∂M5T

Thus, in the solitonic/magnetic 5-brane case as well, we have a saturation of

the mass-charge inequality:

E = V = λΩ4 . (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
∂M10 of the D = 10 spatial hypersurface; the supercharge is then 1
Z
Q= Γ0bc ψc dΣ(9)b . (4.14)
∂M10

One can also rewrite this in fully Lorentz-covariant form, where dΣ(9)b =
dΣ(9)0b → dΣ(9)AB :
Z
Q= ΓABC ψC dΣ(9)AB . (4.15)
∂M10

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 Q2 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 p-brane solutions.

The saturation of the Bogomol’ny inequalities by the p-brane solutions
is an indication that they fit into special types of supermultiplets. All of
these bound-saturating solutions share the important property that they leave
some portion of the supersymmetry unbroken. Within the family of p-brane
solutions that we have been discussing, it turns out 32 that the ∆ values of
such “supersymmetric” p-branes 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 ψM . Checking for the consistency of setting ψM = 0 with the sup-
position of some residual supersymmetry with parameter ǫ(x) requires solving
the equation
δψA | = D̃A ǫ = 0 , (4.17)
ψ=0

where ψA = eA M ψM and
1
D̃A ǫ = DA ǫ − (ΓA BCDE − 8δA B ΓCDE ) FBCDE ǫ
288
DA ǫ = (∂A + 41 ωA BC ΓBC )ǫ . (4.18)

Solving the equation D̃A 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 p-brane 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

ΓA = (γµ ⊗ Σ9 , 1l(2) ⊗ Σm ) , (4.19)

where γµ and 1l(2) are 2 × 2 SO(2, 1) matrices; Σ9 and Σm are 16 × 16 SO(8)

matrices, with Σ9 = Σ3 Σ4 . . . Σ10 , so Σ29 = 1l(16) . The most general spinor field
consistent with (Poincaré)3 × SO(8) invariance in this spinor basis is of the
form
ǫ(x, y) = ǫ2 ⊗ η(r) , (4.20)
where ǫ2 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
Σ9 eigenstates by the use of 12 (1l ± Σ9 ) 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 ′ eC =
3A′ e3A . The first of these conditions is, however, precisely the linearity-
condition refinement (2.15) that we made in the p-brane 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)η0 = eC(r)/6 η0 , where η0 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∞ = ǫ2 ⊗ η0 . Note that, after
imposing this requirement, at most a finite number of parameters can
remain unfixed in the product spinor ǫ2 ⊗ η0 ; 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 ǫ2 ⊗ η0 . 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 + Σ9 )η0 = 0, so the constant SO(8) spinor η0 is also required to
be chiral.g 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, 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), specialised to the BPS background. Thus, for example, in
the case of a D = 11 membrane solution oriented in the {012} directions, one
has, after normalising to a unit 2-volume,
1
{Qα , Qβ } = −(CΓ0 )αβ E + (CΓ12 )αβ U12 . (4.21)
2-vol
Since, as we have seen in (4.11), the membrane solution saturating the Bogo-
mol’ny bound (4.16a) with E = U = U12 , one may rewrite (4.21) as
1
{Qα , Qβ } = 2EP012 P012 = 12 (1l + Γ012 ) , (4.22)
2-vol
g 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
 2
where P012 is a projection operator (i.e. P012 = P012 ) whose trace is trP012 =
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 {Qα , Qβ } of the generators must give
zero when contracted with a surviving supersymmetry parameter ǫα . From
(4.22), this translates to
P012 ǫ∞ = 0 , (4.23)

which is equivalent to condition 3) above, (1l + Σ9 )η0 = 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” p-branes in Section 6, 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 semi-definiteness of the operator {Qα , Qβ }
is the underlying principle in the derivation 30,26,32 of the Bogomol’ny bounds
(4.16). A consequence of this positive semi-definiteness 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 5-brane 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 5-brane oriented in the {012345} directions is P012345 ǫ∞ =
0, where P012345 = 21 (1l + Γ012345 ).

5 Kaluza-Klein dimensional reduction

Let us return now to the arena of purely bosonic field theories, and consider
the relations between various bosonic-sector theories and the corresponding
relations between p-brane solutions. It is well-known 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 D-dimensionally covariant pieces:

dŝ2 = e2αϕ ds2 + e2βφ (dz + AM dxM )2 (5.1)

32
where carets denote (D + 1)-dimensional quantities corresponding to the (D +
1)-dimensional coordinates xM̂ = (xM , z); ds2 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 gzz . 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
Einstein-frame form of the action in D dimensions.
The essential step in a Kaluza-Klein 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 lower-dimensional solutions
which we shall obtain are also particular solutions to higher-dimensional su-
pergravity equations as well. Making the parameter choice β = −(D − 2)α to
preserve the Einstein-frame form of the action, one obtains
p √  
−ĝR(ĝ) = −g R(g)−(D−1)(D−2)α2 ∇M ϕ∇M ϕ− 41 e−2(D−1)αϕ FM N F M N
(5.2)
where F = dA. If one now chooses α2 = [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 F̂[n] = dÂ[n−1] . Clearly, among the
n − 1 antisymmetrised indices of Â[n−1] at most one can take the value z, so
we have the decomposition

Â[n−1] = B[n−1] + B[n−2] ∧ dz . (5.3)

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] (5.4a)

G[n−1] = dB[n−2] . (5.4b)

However, these are not exactly the most convenient quantities to work with,
since a certain “Chern-Simons” 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 Kaluza-Klein

33
vector fields AM coupling to D-dimensional antisymmetric tensors. Accord-
ingly, it is useful to introduce
G′[n] = G[n] − G[n−1] ∧ A , (5.5)
where the second term in (5.5) may be viewed as a Chern-Simons correction
from the reduced D-dimensional point of view.
At this stage, we are ready to perform the dimensional reduction of our
general action (2.1). We find that
Z p h 1 âφ 2 i
Iˆ =: dD+1 x −ĝ R(ĝ) − 12 ∇M̂ φ∇M̂ φ − e F̂n] (5.6)
2n!
reduces to
Z
√ h
I = dD x −g R − 21 ∇M φ∇M φ − 12 ∇M ϕ∇M ϕ − 41 e−2(D−1)αϕ F[2]2

1 −2(n−1)αϕ+α̂φ ′2 1 i
− e G[n] − e2(D−n)αϕ+âφ G2[n−1] . (5.7)
2n! 2(n − 1)!
Although the dimensional reduction (5.7) has produced a somewhat compli-
cated result, the important point to note is that each of the D-dimensional
antisymmetric-tensor field strength terms G′2 2
[n] and G[n−1] has an exponential
prefactor of the form ear φ̃r , where the φ̃r , 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] , Bn−1] ), but retaining at the same
time the scalar-field combination appearing in the corresponding exponential
prefactor, is a consistent truncation. Thus, any one of the three field strengths
(F[2] , G[n−1] , G′[n] ), retained alone together with its corresponding scalar-field
combination, can support p-brane solutions in D dimensions of the form that
we have been discussing.
An important point to note here is that, in each of the ear φ̃ prefactors, the
coefficient ar satisfies
2dr d˜r 2(r − 1)(D − r − 1)
a2r = ∆ − =∆− (5.8)
(D − 2) (D − 2)
with the same value of ∆ as for the “parent” coupling parameter â, satisfying
2d(n) d˜(n) 2(n − 1)(D − n)
â2r = ∆ − =∆− (5.9)
((D + 1) − 2) (D − 1
in D + 1 dimensions. Thus, although the individual parameters ar are both D-
and r-dependent, the quantity ∆ is preserved under Kaluza-Klein reduction for

34
both of the “descendant” field-strength couplings (to G′2 2
[n] or to G[n−1] ) coming
from the original term eâφ F̂[n]
2
. The 2-form 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 scalar-field 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 Kaluza-Klein dimensional reduction is also a consistent truncation.

5.1 Multiple field-strength solutions and the single-charge truncation

After repeated single steps of Kaluza-Klein dimensional reduction from D = 11
down to D dimensions, the metric takes the form 31,32
2 ~ X 2 ~
ds211 = e 3 ~a·φ ds2D + e( 3 ~a−~ai )·φ (hi )2 (5.10a)
i

hi = dz i + Ai[1] + Aij j
[0] dz , (5.10b)

where the AiM are a set of (11−D) Kaluza-Klein vectors generalising the vector
AM in (5.1), emerging from the higher-dimensional metric upon dimensional
reduction. Once such Kaluza-Klein vectors have appeared, subsequent dimen-
sional reduction also gives rise to the zero-form gauge potentials Aij [0] appearing
in (5.10b) as a consequence of the usual one-step reduction (5.3) of a 1-form
gauge potential.
We shall also need the corresponding reduction of the F̂[4] field strength h
(where hatted quantities refer to the original, higher, dimension) and, for later
reference, we shall also give the reduction of its Hodge dual ˆ∗F̂[4] :

i ij 1 ijk
F̂[4] = F[4] +F[3] ∧hi + 21 F[2] ∧hi ∧hj + i j
6 F[1] ∧h ∧h ∧h
k
(5.11a)
ˆ
∗ ~ ~∗
a· φ ~ ~ ij ~ ijk
F̂[4] = e F[4] ∧v+e~ai ·φ ∗F[3]
i
∧v i + 21 e~aij ·φ ∗F[2] ∧v ij + 16 e~aijk ·φ ∗F[1] ∧v ijk ,
(5.11b)
h Note that the lower-dimensional field strengths F[n] include “Chern-Simons” corrections
similar to those in (5.5).

35
(noting that, since the Hodge dual is a metric-dependent construction, expo-
~ appear in the reduction of ˆ∗F̂[4] ) where the
nentials of the dilatonic vectors φ
forms v, vi , vij and vijk appearing in (5.11b) are given by

v = 1
(11−D)! ǫi1 ···i11−D hi1 ∧ · · · ∧ hi11−D
vi = 1
(10−D)! ǫii2 ···i11−D hi2 ∧ · · · ∧ hi11−D
vij = 1
(9−D)! ǫiji3 ···i11−D hi3 ∧ · · · ∧ hi11−D
vijk = 1
(8−D)! ǫijki4 ···i11−D hi4 ∧ · · · ∧ hi11−D . (5.12)

Using (5.10, 5.11a), the bosonic sector of maximal supergravity (1.1) now
reduces to 31,32
Z
√ h ~ 2 − 1 e~a·φ~ F 2 − 1
X ~
ID = dD x −g R − 21 (∂ φ) 48 [4] 12 e~ai ·φ (F[3]
i 2
)
i
X ~ ij 2
X ~bi ·φ
~
aij ·φ i 2
− 41 e (F[2] ) − 1
4 e (F[2] ) (5.13)
i<j i
X X i
~ ijk 2 ~ ij 2
− 21 e~aijk ·φ (F[1] ) − 1
2 ebij ·φ (F[1] ) + LF F A ,
i<j<k ij

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 multi-index sum. From (5.11), one sees that the “straight-backed” field
i ij ijk
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 Kaluza-Klein vectors AiM appearing in (5.10b). Similarly, one also has a set
ij
of 1-form field strengths F[1] for the Kaluza-Klein zero-form gauge potentials
Aij
[0] appearing in (5.10b).
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
~ appearing in the exponential prefactors of the antisymmetric-tensor kinetic
φ
terms in (5.13). For each field-strength kinetic term in (5.13), 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 1-form “straight-backed” field strengths emerging from F[4] in D = 11,
these coefficients are denoted correspondingly ~a, ~ai , ~aij and ~aijk ; for the “cal-
ligraphic” field strengths corresponding to Kaluza-Klein vectors and zero-form

36
gauge potentials emerging out of the metric, these are denoted ~bi and ~bij . How-
ever, not all of these dilaton vectors are independent; in fact, they may all be
expressed in terms of the 4-form and 3-form dilaton vectors ~a and ~aij :31,32

~aij = ~ai + ~aj − ~a bi = −~ai + ~a

~bij = −~ai + ~aj . (5.14)
~aijk = ~ai + ~aj + ~ak − 2~a

Another important feature of the dilaton vectors is that they satisfy the fol-
lowing dot-product relations:
2(11 − D)
~a · ~a =
D−2
2(8 − D)
~a · ~ai = (5.15)
D−2
2(6 − D)
~ai · ~aj = 2δij + .
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 (5.13) 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 rank-n field strength combination F[n] constructed out
of a certain number N of “retained” field strengths Fα [n] , α = 1, . . . , N , (this
could possibly be a straight-backed/calligraphic mixture) selected from those
appearing in (5.13), with all the rest being set to zero.32 Thus, we let

~ = ~nφ + φ
φ ~⊥ , (5.16)

~ ⊥ = 0; in the truncation we then seek to set consistently φ

where ~n · 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:
X
~⊥ −
φ Π⊥ · ~aα (Fα [n] )2 = 0 , (5.17)
α

where Π⊥ is the projector into the dilaton-vector subspace orthogonal to the

~ ⊥ = 0 in (5.17) and letting the retained
retained dilaton direction ~n. Setting φ

37
Fα [n] be proportional, one sees that achieving consistency is hopeless unless
~
all the e~aα ·φ prefactors are the same, thus requiring

~aα · ~n = a ∀α = 1, . . . , N , (5.18)

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 (5.18), consistency of (5.17) with the imposition of φ~⊥ = 0 requires
X
Π⊥ · ~aα (Fα [n] )2 = 0 . (5.19)
α

This
P equation2requires, for every point x in spacetime, that the combination
M

α~aα (Fα [n] ) be parallel to ~n in the dilaton-vector space. Combining this

with the requirement (5.18), one has
X X
~aα (Fα [n] )2 = a~n (Fα [n] )2 . (5.20)
α α

Taking then a dot product of this with ~aβ , one has

X X
Mβα (Fα [n] )2 = a2 (Fα [n] )2 . (5.21)
α α

Detailed analysis 32 shows it to be sufficient to consider the cases where Mαβ

−1
is invertible, so by applying Mαβ to (5.21), one finds
X X
−1
(Fα [n] )2 = a2 Mαβ (Fγ [n] )2 , (5.22)
β γ

and, indeed, we find that the Fα [n] must all be proportional. Summing on α,
one has X
−1 −1
a2 = ( Mαβ ) ; (5.23)
α,β

one then defines the retained field-strength combination F[n] so that

X
−1
(Fα [n] )2 = a2 Mαβ (F[n] )2 . (5.24)
β

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 (5.13) is not inconsistent with setting

38
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 purely-electric cases satisfying the ele-
mentary ansatz (2.4) or purely-magnetic 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 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 7.
After truncating down to the system (2.1), the analysis proceeds as in
Section 2. It turns out 32 that supersymmetric p-brane solutions arise when
the matrix Mαβ for the retained Fα [n] satisfies

2dd˜
Mαβ = 4δαβ − , (5.25)
D−2
and the corresponding ∆ value for F[n] is

4
∆= , (5.26)
N
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 scalar-field combination.32 We shall see in Section 6
that the N > 1 solutions to single-charge truncated systems (2.1) may also
be interpreted as special solutions of the full reduced action (5.13) 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.

5.2 Diagonal dimensional reduction of p-branes

The family of p-brane solutions is ideally suited to interpretation as solutions
of Kaluza-Klein reduced theories, because they are naturally independent of
the “worldvolume” xµ coordinates. Accordingly, one may let the reduction
coordinate z be one of the xµ . 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 (5.7) is to perform a Weyl rescaling on it in order to be in accordance
with the form of the metric chosen in the Kaluza-Klein ansatz (5.10), which

39
was adjusted so as to maintain the Einstein-frame form of the gravitational
term in the dimensionally reduced action.
Upon making such a reinterpretation, elementary/solitonic p-branes 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 (5.8, 5.9). 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 Kaluza-Klein 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 field-theory solutions,
of the procedure of double dimensional reduction 22 for p-brane worldvolume
actions, which can be taken to constitute the δ-function sources for singular
p-brane solutions, coupled in to resolve the singularities, as we discussed in
subsection 4.1.

5.3 Multi-center solutions and vertical dimensional reduction

As we have seen, translational Killing symmetries of p-brane 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 p-brane 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 p-brane
solutions to multi-center p-brane solutions by taking the harmonic function to
be
X kα
H(y) = 1 + kα > 0 . (5.27)
α |~y − ~yα |d̃

Once again, the integration constant has been adjusted to make H | = 1 ↔

∞
φ| = 0. The generalized solution (5.27) corresponds to parallel and similarly-
∞ √
oriented p-branes, with all charge parameters λα = 2dk ˜ 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 = yα , where α ranges over any
number of centers. The metric and the electric-case antisymmetric tensor
gauge potential corresponding to (5.27) 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

40
to accommodate the multi-center form of the solution:
X λα
Fm1 ...mn = −d˜−1 ǫm1 ...mn p ∂p , (5.28)
α |~y − ~yα |d̃

which ensures the validity of the Bianchi identity just as well as (2.6) does.
The mass/(unit p-volume) density is now

2ΩD−d−1 X
E= √ λα , (5.29)
∆ α
P
while the total electric or magnetic charge is given by ΩD−d−1 λα , so the
Bogomol’ny bounds (4.16) are saturated just as they are for the single-center
solutions (2.24). Since the multi-center solutions given by (5.27) satisfy the
same supersymmetry-preservation conditions on the metric and antisymmetric
tensor as (2.24), the multi-center solutions leave the same amount of super-
symmetry unbroken as the single-center solution.
From a mathematical point of view, the multi-center solutions (5.27) 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 scalar-field forces against repulsive antisymmetric-
tensor forces for the similarly-oriented p-brane “leaves.”
The multi-center solutions given by (5.27) 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.i 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 (5.27) in a periodic
array, i.e. by letting the integration constants kα all be equal, and aligning the
“centers” yα 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 densely-packed infinite stack of this sort, one has

X Z +∞
kα kdz k̃
−→ = ˜
(5.30a)
α |~y − ~yα |d̃ −∞ (r̂2 + z 2 )d̃/2 r̃d−1
i Similar procedures have been considered in a number of articles in the literature; see, e.g.
Refs.33

41
 D−2
X
r̂2 = ymym (5.30b)
m=d
√
πkΓ(d˜ − 21 )
k̂ = , (5.30c)
˜
2Γ(d)

where r̂ in (5.30b) 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 p-brane solution
with the same worldvolume dimension as the original higher-dimensional 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 6. 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 higher-dimensional isotropic p-brane solutions, and hence
can be called “stainless” p-branes.13 In Figure 6, 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 oxidised back up to some solution in D = 11. We
shall see in Section 6 that what one obtains upon oxidation of the “stainless”
solutions in Figure 6 falls into the interesting class of “intersecting branes”
built from four basic “elemental” solutions of D = 11 supergravity.

42
 D

4 4
11

4 4 4 4 4
10

9 2 2

6 4/3,1' 2 4/3,1'

}
elementary
5 4/3 4/3
solitonic "stainless"
1,4/5, 2/3,4/7
1' 1,4/5,2/3,4/7 self-dual
4

Kaluza-Klein descendants
1/2 1/2
3
4 ∆ values

vertical reduction trajectories

2
diagonal reduction trajectories
1
0 1 2 3 4 5 6 7 d
instanton particle string membrane 3-brane 4-brane 5-brane 6-brane

Figure 6: Brane-scan of supergravity p-brane solutions (p ≤ (D − 3))

43
5.4 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.j In this step, the integral
(5.30) contains an additive divergence and needs to be renormalized. This
is easily handled by putting finite limits ±L on the integral, which becomes
RL
−L
dz̃(r2 + z̃ 2 )−1/2 , and then by subtracting a divergent term 2 ln L before
taking the limit L → ∞. Then the integral gives the expected ln r̂ 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.34 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 2-form Aµν field to a pseudoscalar, or axion,
field χ, so such strings are also solutions to dilaton-axion gravity. The p-brane
ansatz gives a spacetime of the form M4 = M2 × Σ2 , where M2 is D = 2
Minkowski space. Supporting this string solution, one has the 2-form gauge
field Aµν and the dilaton φ. These fields give rise to a field stress tensor of the
form

Tµν (A, φ) 1
= − 16 (a2 + 4)∂m K∂m Kηµν
1 2
Tmn (A, φ) = 8 (a − 4)(∂m K∂n K − 21 δmn (∂p K∂p K)) , (5.31)

where a is as usual the√ mdilaton coupling parameter and e−K = H = 1 −

m
8GT ln(r), with r = y y , 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
that Tmn (source) = 0, then one has the source stress tensor
Z
−T √ 1
Tµν (source) = √ d2 ξ −γγ ij ∂i Xµ ∂j Xν e− 2 aφ δ(x − X) . (5.32)
−g
j 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.

44
By inspection of the field solution, one has Tmn (A, φ) = 0, while the contri-
butions to Tµν from the Aµν and φ fields and also from the source (5.32) are
both of the form diag(ρ, −ρ). Thus, the overall stress tensor is of the form
TMN = diag(ρ, −ρ, 0, 0).
Consequently, the Einstein equation in the transverse m, n indices becomes
Rmn − 21 gmn R = 0, since the transverse stress tensor components vanish.
This equation is naturally satisfied for a metric satisfying the p-brane 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 Rmn − 12 gmn 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 world-sheet directions, the equations become

− 21 Rηµν = −8πGρηµν , (5.33)

or just
R = 16πGρ , (5.34)
and as we have already noted, R = Rmm . Owing to the fact that the D = 2
Weyl tensor vanishes, the transverse space Σ2 is conformally flat; Eq. (5.34)
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 p-brane 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 unbroken-symmetry configurations at a
stationary point of the Higgs potential, within a very small transverse-space
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 Tµν (A, φ) which is not just
concentrated at the string core but instead is smeared out over spacetime.
The difference arises from the absence of a potential for the fields Aµν , φ
supporting the solution in the supersymmetric case. Nonetheless, as one can
see from the behavior of the stress tensor Tmn in Eq. (5.31), the transverse
space Σ2 is asymptotically locally flat (ALF), with a total energy density given

45
by the overall deficit angle measured at infinity. For multiple-centered string
solutions, one has
X
H =1− 8GTi ln |~y − ~yi | . (5.35)
i

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 p-brane ansatz
(2.3, 2.6) and searching for (D − 2) branes in ordinary massless supergravity
theories, one simply doesn’t find any such solutions.

5.5 Beyond the (D − 3)-brane barrier: Scherk-Schwarz 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
RL
integral −L dz ln(y 2 + z 2 ) needs to be renormalized by subtracting a divergent
term 4L(ln L − 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.35
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 antisymmetric-tensor one-form field strength

Fm = −ǫmn ∂n H (5.36)

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 (5.36) 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

46
prior to the vertical reduction, thus producing a linear harmonic function in
the transverse coordinate y,
H(y) = const. + my , (5.37)
the field strength (5.36) acquires a constant component along the stacking axis
↔ reduction direction z,
Fz = −ǫzy ∂y H = m , (5.38)
which implies an unavoidable dependence k of the corresponding zero-form
gauge potential on the reduction coordinate:
A[0] (x, y, z) = mz + χ(x, y) . (5.39)
From a Kaluza-Klein point of view, the unavoidable linear dependence
of a gauge potential on the reduction coordinate given in (5.39) appears to
be problematic. Throughout this review, we have dealt only with consistent
Kaluza-Klein 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 (5.39) 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 Kaluza-Klein re-
duction with an ansatz like (5.39), 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 Kaluza-Klein 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.
The most usual guarantee of consistency in Kaluza-Klein 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
k 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 zero-form gauge potential in (5.39) does not have the needed gauge symmetry,
however.

47
the present case, retaining a linear dependence on the reduction coordinate as
in (5.39) 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
Kaluza-Klein 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 zero-form 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 Kaluza-Klein reduction framework so as to
allow for linear dependence of an axionic zero-form potential on the reduction
coordinate, precisely of the form occurring in (5.39). So, provided A[0] is an ax-
ion, the reduction (5.39) turns out to be consistent after all. This extension of
the Kaluza-Klein ansatz is in fact an instance of Scherk-Schwarz reduction.36,37
The basic idea of Scherk-Schwarz 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 field-equation 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 (5.39) is a simple shift
symmetry A[0] → A[0] + const.
Unlike the original implementation of the Scherk-Schwarz reduction idea,36
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, p-brane solutions related by vertical dimensional reduction, even in the
enlarged Scherk-Schwarz 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 zero-form, and thus to pre-

48
pare the theory for a reduction like (5.39). This is most easily explained by an
example, so let us consider the first possible Scherk-Schwarz reduction l in the
sequence of theories descending from (1.1), starting in D = 9 where the first
axion field appears.35 The Lagrangian for massless D = 9 maximal supergrav-
ity is obtained by specializing the general dimensionally-reduced action (5.13)
given in Section 2 to this case:
√ h 3 √
7 ~
L9 = −g R − 12 (∂φ1 )2 − 21 (∂φ2 )2 − 12 e− 2 φ1 + 2 φ2 (∂χ)2 − 48
1 ~
ea·φ (F[4] )2
~ (1) ~ (2) ~ (12) ~ ~ (1)
− 21 e~a1 ·φ (F[3] )2 − 12 e~a2 ·φ (F[3] )2 − 41 e~a12 ·φ (F[2] )2 − 14 eb1 ·φ (F[2] )2
i
~ ~ (2) (12) (1) (2)
− 41 eb2 ·φ (F[2] )2 − 12 F̃[4] ∧ F̃[4] ∧ A[1] − F̃[3] ∧ F̃[3] ∧ A[3] , (5.40)
(12) ~ = (φ1 , φ2 ).
where χ = A[0] and φ
Within the scalar sector (φ, ~ χ) of (5.40), the dilaton coupling has been
made explicit; in the rest of the Lagrangian, the dilaton vectors have the general
structure given in (5.14, 5.15). The scalar sector of (5.40) forms a nonlinear σ-
model for the manifold GL(2, IR)/SO(2). This already makes it appear that one
may identify χ as an axion available for Scherk-Schwarz reduction. However,
account must still be taken of the Chern-Simons structure lurking inside the
field strengths in (5.11, 5.40). In detail, the field strengths are given by
(1) (1) (2) (2)
F[4] = F̃[4] − F̃[3] ∧ A[1] − F̃[3] ∧ A[1]
(1) (2) (12) (1) (2)
+χF̃[3] ∧ A[1] − F̃[2] ∧ A[1] ∧ A[1] (5.41a)
(1) (1) (12) (2)
F[3] = F̃[3] − F̃[2] ∧ A[1] (5.41b)
(2) (2) (12) (1) (1)
F[3] = F̃[3] + F̃[2] ∧ A[1] − χF̃[3] (5.41c)
(12) (12) (1) (1) (2)
F[2] = F̃[2] F[2] = F̃[2] − dχ ∧ A[1] (5.41d)
(2) (2) (12)
F[2] = F̃[2] F[1] = dχ , (5.41e)
where the field strengths carrying tildes are the naı̈ve expressions without
Chern-Simons corrections, i.e. F̃n] = dA[n−1] . Now the appearance of undiffer-
entiated χ factors in (5.41a,c) makes it appear that a Scherk-Schwarz reduction
would be inconsistent. However, one may eliminate these undifferentiated fac-
tors by making the field redefinition
(2) (2) (1)
A[2] −→ A[2] + χA[2] , (5.42)
lA higher-dimensional Scherk-Schwarz reduction is possible 37 starting from type IIB super-
gravity in D = 10, using the axion appearing in the SL(2, IR)/SO(2) scalar sector of that
theory.

49
after which the field strengths (5.41a,c) become
(1) (1) (2) (2)
F[4] = F̃[4] − F̃[3] ∧ A[1] − F̃[3] ∧ A[1]
(1) (2) (12) (1) (2)
−dχ ∧ A[2] ∧ A[1] − F̃[2] ∧ A[1] ∧ A[1] (5.43a)
(2) (2) (12) (1) (1)
F[3] = F̃[3] + F[2] ∧ A[1] + dχ ∧ A[2] , (5.43c)

the rest of (5.41) remaining unchanged.

(12)
After the field redefinitions (5.42), the axion field χ = A[0] is now ready
for application of the Scherk-Schwarz reduction ansatz (5.39). 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 (5.39) to the D = 9
Lagrangian, one obtains a D = 8 reduced Lagrangian m

L8 ss =
√ h
−g R − 12 (∂φ1 )2 − 12 (∂φ2 )2 − 21 (∂φ3 )2
~ ~ (3) ~ ~ (13) (23) (2)
− 21 eb12 ·φ (∂χ − mA[1] )2 − 12 eb13 ·φ (∂A[0] − ∂χA[0] + mA[1] )2
~ ~ (23) ~ (123) 2
− 21 eb23 ·φ (∂A[0] )2 − 12 e~a123 ·φ (∂A[0] )
~ ~
a· φ (1) (2) (3) ~ (1)
1
− 48 e (F[4] − mA[2] ∧ A[1] ∧ A[1] )2 − 1 ~
12 e
a1 · φ
(F[3] )2
1 ~ ~ (2) (1) (3) ~ (3) (1) (2)
− 12 ea2 ·φ (F[3] − mA[2] ∧ A[1] )2 − 1 ~
12 e
a3 · φ
(F[3] + mA[2] ∧ A[1] )2
~ (12) ~ (13) ~ (23) (1)
− 41 e~a12 ·φ (F[2] )2 − 14 e~a13 ·φ (F[2] )2 − 41 e~a23 ·φ (F[2] + mA[2] )2
~ ~ (1) (2) (3) ~ ~ (2) ~ ~ (3)
− 41 eb1 ·φ (F[2] − mA[1] ∧ A[1] )2 − 41 eb2 ·φ (F[2] )2 − 14 eb3 ·φ (F[2] )2
i
~ ~ (1) (1)
− 21 m2 eb123 ·φ + mF[3] ∧ A[2] ∧ A[3] + LF F A , (5.44)

where the dilaton vectors are now those appropriate for D = 8; the term LF F A
contains only m-independent terms.
(3) (2) (1)
It is apparent from (5.44) that the fields A[1] , A[1] and A[2] have become
(13)
massive. Moreover, there are field redefinitions under which the fields χ, A[0]
(23)
and A[1] may be absorbed. One way to see how this absorption happens is to
notice that the action obtained from (5.44) has a set of three Stueckelberg-type
mI am grateful to Marcus Bremer for help in correcting some errors in the original expression
of Eq. (5.44) given in Ref.35

50
 (3) (2) (1)
gauge transformations under which A[1] , A[1] and A[2] transform according
to their standard gauge transformation laws. These three transformations are
accompanied, however, by various compensating transformations necessitated
by the Chern-Simons corrections present in (5.44) as well as by m-dependent
(13) (23)
shift transformations of χ, A[0] and A[1] , respectively. Owing to the presence
of these local shift terms in the three Stueckelberg symmetries, the fields χ,
(13) (23)
A[0] and A[1] may be gauged to zero. After gauging these three fields to
(3) (2)
zero, one has a clean set of mass terms in (5.44) for the fields A[1] , A[1] and
(1)
A[2] .
As one descends through the available spacetime dimensions for super-
gravity theories, the number of axionic scalars available for a Scherk-Schwarz
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
Naxions 1 4 10 20 36 63

Each of these axions gives rise to a distinct massive supergravity theory

upon Scherk-Schwarz reduction,35 and each of these reduced theories has its
own pattern of mass generation. In addition, once a Scherk-Schwarz reduction
step has been performed, the resulting theory can be further reduced using
ordinary Kaluza-Klein reduction. Moreover, the Scherk-Schwarz and ordinary
Kaluza-Klein processes do not commute, so the number of theories obtained by
the various combinations of Scherk-Schwarz and ordinary dimensional reduc-
tion is cumulative. In addition, there are numerous possibilities of performing
Scherk-Schwarz 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 Scherk-Schwarz generalisations of the Kaluza-Klein
reduction process.38 For further details on the panoply of Scherk-Schwarz re-
duction possibilities, we refer the reader to Refs.35,38
The single-step procedure of Scherk-Schwarz dimensional reduction de-
scribed above may be generalised to a procedure exploiting the various coho-
mology classes of a multi-dimensional compactification manifold.39 The key to
this link between the Scherk-Schwarz generalised dimensional reduction and
the topology of the internal Kaluza-Klein manifold K is to recognise that the
single-step reduction ansatz (5.39) may be generalised to

A[n−1] (x, y, z) = ω[n−1] + A[n−1] (x, y) , (5.45)

51
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, IR). For example,
in the case of a single-step generalised reduction on a circle S 1 , one has Ω[1] =
mdz ∈ H 1 (S 1 , IR), reproducing our earlier single-step reduction (5.39).
As another example, consider a generalised reduction on a 4-torus T 4 start-
ing in D = 11, setting A[3] (x, y, z) = ω[3] + A[3] (x, y) with Ω[4] = dω[3] =
mdz1 ∧ dz2 ∧ dz3 ∧ dz4 ∈ H 4 (T 4 , IR). In this example, one may choose to
write ω[3] locally as ω[3] = mz1 dz2 ∧ dz3 ∧ dz4 . All of the other fields are
reduced using the standard Kaluza-Klein ansatz, with no dependence on any
of the zi 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 (5.44). The same theory (up to field redefinitions) can also be ob-
tained 35 by first making an ordinary Kaluza-Klein reduction from D = 11
down to D = 8 on a 3-torus T 3 , then making an S 1 single-step generalised
Scherk-Schwarz reduction (5.39) 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 single-step ansatz (5.39), the recognition that one can use
any of the H n (K, IR) cohomology classes of the compactification manifold K
significantly extends the scope of the generalised reduction procedure. For ex-
ample, it allows one to make generalised reductions on manifolds such as K3
or on Calabi-Yau manifolds.39
For our present purposes, the important feature of theories obtained by
Scherk-Schwarz reduction is the appearance of cosmological potential terms
such as the penultimate term in Eq. (5.44). 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
(5.44) or of one of the many theories obtained by Scherk-Schwarz reduction in
lower dimensions, one may arrive at the simple Lagrangian
√ h i
L = −g R − 12 ∇M φ∇M φ − 12 m2 eaφ . (5.46)

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
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 (5.46) to the
reduction-invariant parameter ∆ by the standard formula (2.18) gives ∆ =
a2 − 2(D − 1)/(D − 2); taking the corresponding p = D − 2 brane solution from
(2.24), one finds
4 4(D−1)
ds2 = H ∆(D−2) ηµν dxµ dxν + H ∆(D−2) dy 2 (5.47a)

52
 eφ = H −2a/∆ , (5.47b)

where the harmonic function H(y) is now a linear function of the single trans-
verse coordinate, in accordance with (5.37).n The curvature of the metric
(5.47a) 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. + M |y| (5.48)

√
where M = 21 m ∆. With the choice (5.48), there is just a delta-function
singularity at the location of the domain wall at y = 0, corresponding to the
discontinuity in the gradient of H.
The domain-wall solution (5.47, 5.48) 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 (5.46) does not even admit a non-flat
maximally-symmetric solution, owing to the complication of the cosmological
potential. The domain-wall solution (5.47, 5.48), 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
5.4. There is no inconsistency between the existence of domain-wall solutions
like (5.47, 5.48) and the inability to find such solutions in standard supergrav-
ity theories, or with the conical-spacetime character of (D − 3)-branes, because
these domain walls exist only in massive supergravity theories like (5.44), with
a vacuum structure different from that of standard massless supergravities.
Because the Scherk-Schwarz generalised dimensional reduction used to obtain
them was a consistent truncation, such domain walls can be oxidised back
to solutions of higher-dimensional massless supergravities, but in that case,
they have the form of stacked solutions prepared for vertical reduction, with
non-zero field strengths in the reduction directions, as in our example (5.38).

6 Intersecting branes, scattering branes

6.1 Multiple component solutions
α
Given the existence of solutions (5.24) with several active field strengths F[n] ,
but with coincident charge centers, it is natural to try to find solutions where
n Domain walls solutions such as (5.47) in supergravity theories were found for the D = 4
case in Ref.40 and a review of them has been given in Ref.41

53
 α
the charge centers for the different F[n] are separated.42 This will lead us to
a better understanding of the ∆ 6= 4 solutions shown in Figure 6. 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 ~yα ,
α = 1, . . . , ℓ. Let the charge parameter for F α be λα . Thus, for example, in
the magnetic case, one sets

α yp
Fm 1 ,...,mn
= λα ǫm1 ,...,mn p . (6.1)
|~y − ~yα |n+1

In both the electric and the magnetic cases, the λα are related to the integration
constants k α appearing in the metric by k α = λα /d. ˜ Letting ς = ±1 in the
electric/magnetic cases as before, the solution for the metric and the active
~
dilatonic combinations eς~aα ·φ is given by
ℓ
Y −d̃
ℓ
Y d
ds2 = HαD−2 dxµ dxµ + HαD−2 dy m dy m
α=1 α1
ℓ
X −dd̃
~
eς~aα ·φ = Hα2 HβD−2 (6.2)
β1
kα
Hα = 1+ .
|~y − ~yα |d̃

The non-trivial step in verifying the validity of this solution is the check that
the non-linear terms still cancel in the Einstein equations, even with the mul-
tiple centers.42
1 2
Now consider a solution with two field strengths (F[n] , F[n] ) in which the
two charge parameters are taken to be the same, λα = λ, while the charge
centers are allowed to coalesce. When the charge centers have coalesced, the
resulting solution may be viewed as a single-field-strength solution for a field
1 2
strength rotated by π/4 in the space of field strengths (F[n] , F[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 = 2λΩD−d−1 /4. On the other
hand, the total √ mass density will add as a scalar quantity, so E = E√ 1 + E2 =
2λΩD−d−1 /4 = 2U . Thus, the coalesced solution satisfies E = 2U/ ∆ with
∆ = 2. Direct comparison with our general p-brane solution (2.24) shows that
the coalesced solution agrees precisely with the single-field-strength ∆ = 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 single-field-strength analysis. In the next subsection, we

54
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 (6.2) 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.”42 We shall shortly see that the zero-force 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 half-preservation characterising single-component ∆ = 4
solutions.

6.2 Intersecting branes and the four elements in D = 11

The multiple-charge-center solutions (6.2) to the dimensionally reduced the-

ory (5.13) may automatically be interpreted as solutions of any one of the
higher-dimensional 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 lower-dimensional solutions
may automatically be oxidised in this way into solutions of higher-dimensional
supergravity theories, it is not guaranteed that these oxidised branes always
fall into the class of isotropic p-brane solutions that we have mainly been dis-
cussing. For example, in D = 9, one has a two-black-hole solution of the form
(6.2), supported by a 1-form gauge potential A12 [1] descending from the D = 11
gauge potential A[3] and also by another 1-form gauge potential, e.g. A2[1] ,
emerging from the metric as a Kaluza-Klein vector field. Upon oxidising the
two-black-hole solution back to D = 11, one finds the solution
1  
ds211 = H13 (y) H1−1 (y){−dt2 +dρ2 +dσ 2 +(H2 (y)−1)(dt+dρ)2 }+dy mdy m
A[3] = H1−1 (y)dt∧dρ∧dσ , m = 3, . . . , 10, wavek2-brane (6.3)

which depends on two independent harmonic functions H1 (y) and H2 (y), where
the y m are an 8-dimensional set of “overall transverse” coordinates.
Although the solution (6.3) clearly falls outside the class of p-brane or
multiple p-brane solutions that we have considered so far, it nonetheless has
two clearly recognisable elements, associated to the two harmonic functions
H1 (y) and H2 (y). In order to identify these two elements, we may use the
freedom to trivialise one or the other of these harmonic functions by setting it

55
equal to unity. Thus, setting H2 = 1, one recovers
1  
ds211 = H 3 (y) H −1 (y){−dt2 + dρ2 + dσ 2 + dy m dy m
A[3] = H −1 (y)dt ∧ dρ ∧ dσ, m = 3, . . . , 10, 2-brane (6.4)

which one may recognised as simply a certain style of organising the harmonic-
function factors in the D = 11 membrane solution 18 (3.2), generalised to
an arbitrary harmonic function H(y) ↔ H1 (y) in the membrane’s transverse
space.
Setting H1 = 1 in (6.3), on the other hand, produces a solution of D = 11
supergravity that is not a p-brane (i.e. it is not a Poincaré-invariant hyperplane
solution). What one finds for H1 = 1 is a classic solution of General Relativity
found originally in 1923 by Brinkmann,43 the pp wave:

ds211 = {−dt2 + dρ2 + (H(y) − 1)(dt + dρ)2 } + dy m dy m

A[3] = 0, m = 2, . . . , 10, pp wave (6.5)

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 H1 = 1 in (6.3), H(y) ↔ H2 (y)
is constant in one of these 9 directions, corresponding to the coordinate σ in
(6.3).
The solution (6.3) thus may be viewed as a D = 11 pp wave superposed on
a membrane. Owing to the fact that the harmonic function H2 (y) depends only
on the overall transverse coordinates y m , m = 3, . . . , 10, the wave is actually
“delocalised” in the third membrane worldvolume direction, i.e. the solution
(6.3) is independent of σ as well as of its own lightplane coordinates. Of course,
this delocalisation of the wave in the σ direction is just what makes it possible
to perform a dimensional reduction of (6.3) on the {ρ, σ} coordinates down to
a D = 9 configuration of two particles of the sort considered in (6.2), i.e. the
wave in (6.3) has already been stacked up in the σ direction as is necessary in
preparation for a vertical dimensional reduction. Another point to note about
(6.3) is that the charge centers of the two harmonic functions H1 and H2 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 (6.3) need not actually overlap on
any specific subspace of spacetime. The term “intersecting” is generally taken
to mean that there are shared worldvolume coordinates, in this case the {t, ρ}
overlap between the membrane worldvolume and the lightplane coordinates.44
A very striking feature of the family of multiple-component p-brane solu-
tions is that their oxidations up to D = 11 involve combinations of only 4 basic

56
“elemental” D = 11 solutions. Two of these we have just met in the oxidised
solution (6.3): the membrane and the pp wave. The two others are the “duals”
of these: the 5-brane 25 and a solution describing the oxidation to D = 11 of
the “Kaluza-Klein monopole.”45 The 5-brane may be written in a style similar
to that of the membrane (6.4):
2  
ds211 = H 3 (y) H −1 (y){−dt2 + dx21 + . . . + dx25 } + dy m dy m
∗
F[4] = dH(y), m = 6, . . . , 10, 5-brane (6.6)

where the H(y) is a general harmonic function in the 5-dimensional transverse

space.
The Kaluza-Klein monopole oxidised up to D = 11 is the solution

ds211 = −dt2 + dx21 + . . . + dx26 + ds2TN (y)

A[3] = 0 (6.7a)
ds2TN i
= H(y)dy dy + H i −1 i 2
(y)(dψ + Vi (y)dy ) , i = 1, 2, 3,
~ ×V
∇ ~ ~ ,
= ∇H Taub-NUT (6.7b)

where ds2TN is the Taub-NUT metric, a familiar four-dimensional Euclidean

gravitational instanton. The harmonic function H in (6.7) is a function only
of the 3 coordinates y i , and not of the coordinate ψ, which plays a special
rôle. Generally, the solution (6.7) 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 single-center harmonic function H(y) = 1 + k/(|y|), the appropriate
identification period for ψ is 4πk.
Thus, the Taub-NUT solution naturally invites interpretation as a com-
pactified solution in one less dimension, after reduction on ψ. In the case of
the original Kaluza-Klein monopole,45 the starting solution had 4+1 dimen-
sions, giving rise after compactification to a magnetically-charged particle in
D = 4 dimensions. The solution (6.7) has an additional 6 spacelike worldvol-
ume dimensions x1 , . . . , x6 , so after reduction on the ψ coordinate one has a
magnetically-charged 6-brane solution in D = 10.
The relation ∆ψ = 4πk between the compactification period of ψ and
the charge-determining integration constant k in the harmonic function H
of the solution (6.7) gives rise to a quantisation condition at the quantum
level involving the magnetic charge of the dimensionally-reduced D = 10 6-
brane descending from (6.7) and the electric charge of the extreme black hole
particle obtained by reducing the pp wave (6.5). This quantisation condition
is nothing other than an ordinary quantisation of momentum for Fourier wave

57
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 6-brane, one finds U V = 2πκ210 n, with n ∈ ZZ (where κ210 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 7 we shall return to the subject of charge quantisation conditions
more generally for the charges carried by p-branes.
Let us now return to the question of supersymmetry preservation and
enquire whether intersecting branes like (6.3) can also preserve some portion
of unbroken rigid supersymmetry. All four of the elemental D = 11 solutions
(6.4 – 6.7) 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
(6.5). 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 (6.5)
propagating in the {01} directions of spacetime, with normalisation to unit
length along the wave’s propagation direction:

1
{Qα , Qβ } = 2EP01 P01 = 21 (1l + Γ01 ) , (6.8)
length

where P01 is again a projection operator with half of its eigenvalues zero, half
unity. Consequently, the pp wave solution (6.5) preserves half of the D = 11
rigid supersymmetry.
Now let us apply the projection-operator analysis to the wavek2-brane so-
lution (6.3). Supersymmetry preservation in a membrane background oriented
parallel to the {012} hyperplane requires the projection condition P012 ǫ∞ = 0
(4.23), while supersymmetry preservation in a pp wave background with a {01}
lightplane requires P01 ǫ∞ = 0. Imposing these two conditions simultaneously
is consistent because these projectors commute,

[P012 , P01 ] = 0 . (6.9)

Since tr(P012 P01 ) = 41 · 32, the imposition of both projection conditions on ǫ∞

cuts the preserved portion of rigid D = 11 supersymmetry down to 41 .
Now, let’s consider another example of an intersecting-brane solution, con-
taining as elements a D = 11 membrane, 5-brane pair. The solution is

58
 1 2 
ds2 = H13 (y)H23 (y) H1−1 (y)H2−1 (y)(−dt2 + dx21 ) 2 ⊥ 5(1)
+H1−1 (y)(dx22 ) + H2−1 (y)(dx23 + ... + dx26 )
m m

+dy dy m = 7, . . . , 10 (6.10a)
Fm012 = ∂m (H1−1 ) F2mnp = −ǫmnpq ∂q H2 , (6.10b)

where as in the wave-on-a-membrane solution (6.3), the harmonic functions

H1 (y) and H2 (y) depend only on the overall transverse coordinates. By con-
sidering special cases where H2 = 1 or H1 = 1, one identifies the membrane
and 5-brane elements of the solution (6.10); as before, these elements are delo-
calised 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 (6.10). Note that both the membrane and 5-brane 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 overall-transverse charge centers of H1 and H2 need
not coincide, so there is not necessarily a true string overlap), the solution
(6.10) is denoted 2 ⊥ 5(1).
The forms of the wavek2-brane solution (6.3) and the 2 ⊥ 5(1) solution
(6.10) illustrate the general structure of intersecting-brane solutions. For a
two-element 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 (6.10):
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
d/(D−2)
factor Hi (y) for the whole metric, and then in addition one puts a fac-
−1
tor Hi (y) in front of each dx2 term belonging to the worldvolume of the ith

59
element. One may verify this pattern in the structure of (6.10). This pattern
has been termed the harmonic function rule.44
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.46 For now, let us just check point 3) in the supersymmetry-
preservation analysis for the 2 ⊥ 5(1) solution (6.10). For each of the two
elements, one has a projection condition on the surviving rigid supersymme-
try parameter ǫ∞ : P012 ǫ∞ = 0 for the membrane and P013456 ǫ∞ = 0 for
the 5-brane. These may be consistently imposed at the same time, because
[P012 , P013456 ] = 0, similarly to our discussion of the wavek2-brane solution.
The amount of surviving supersymmetry in the 2 ⊥ 5(1) solution is 14 , because
tr(P012 P013456 ) = 41 · 32.

6.3 Brane probes, scattering branes and modulus σ-model geometry

The existence of static configurations such as the wavek2-brane solution (6.3)
or the 2 ⊥ 5(1) intersecting-brane solution (6.10) derives from the proper-
ties of the transverse-space Laplace equation (2.20) arising in the process of
solving the supergravity equations subject to the p-brane ansätze (2.3–2.6).
The Laplace equation has the well-known property of admitting multi-center
solutions, which we have already encountered in Eq. (5.27). Physically, the
existence of such multi-center 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.18,47 In this limit, one may ignore the deformation of the
heavy-brane background caused by the light brane. The use of the brane-
probe coupling is a simple way to approximately treat time-dependent brane
configurations. For a p-brane probe of this sort coupled to a D-dimensional
supergravity background, the probe action is
Z Z
1 1 pr ~
p+1 m n ς ~aα · φ
Iprobe = −Tα d ξ (− det(∂µ x ∂ν x gmn (x)) e 2 2 + Qα Ãα
[p+1]

(6.11a)

60
Ãα
[p+1] = [(p + 1)!]−1 ∂µ1 xm1 · · · ∂µp+1 xmp+1 Aα
m1 ···mp+1 dξ
µ1
∧ · · · ∧ dξ µp+1 .
(6.11b)
The dilaton coupling in (6.11a) occurs because one needs to have the correct
source for the D-dimensional Einstein frame, i.e. the conformal frame in which
the D-dimensional Einstein-Hilbert action is free from dilatonic scalar factors.
Requiring that the source match correctly to a p-brane (probe) solution de-
1 pr ~
mands the presence of the dilaton coupling e 2 ς ~aα ·φ , where ς pr = ±1 according
to whether the p-brane probe is of electric or magnetic type and where ~aα is
the dilaton vector appearing in the kinetic-term dilaton coupling in (5.13) for
the gauge potential Aα[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 similarly-oriented
heavy membrane.18 In this case, the brane-probe action (6.11) 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 = xµ , µ = 0, 1, 2, then the bosonic probe
action becomes
Z q 
Iprobe = −T d3 ξ − det(e2A(y) ηµν + e2B(y) ∂µ y m ∂ν y m ) − eC(y) .
(6.12)
Expanding the square root in (6.12), one finds at order (∂y)0 the effective
potential
Vprobe = T (e3A(y) − eC ) . (6.13)
Recalling the condition (2.22), which becomes just e3A(y) = eC(y) for the mem-
brane background, one has directly Vprobe = 0, confirming the absence of static
forces between the two membrane components.
Continuing on in the expansion of (6.12) to order (∂y)2 in the probe ve-
locity, one has the effective probe σ-model
Z
(2) T
Iprobe = − d3 ξe3A(y) e2(B(y)−A(y) ∂µ y m ∂ν y m ηµν , (6.14)
2
but, recalling the supersymmetry-preservation condition (2.17) characterising
the heavy-membrane background, the probe σ-model metric in (6.14) reduces
simply to
γ mn = eA(y)+2B(y) δ mn = δ mn , (6.15)
i.e., the membrane-probe σ-model metric is flat.
The flatness of the membrane-probe σ-model metric (6.15) accords pre-
cisely with the degree of rigid supersymmetry that survives in the underlying

61
supergravity solution with two parallel, similarly oriented D = 11 membranes,
which we found in subsection 4.4 to have 12 · 32 components, i.e. d = 2 + 1,
N = 8 probe-worldvolume supersymmetry. This high degree of surviving su-
persymmetry is too restrictive in its constraints on the form of the σ-model
to allow for anything other than a flat metric, precisely as one finds in (6.15).
Continuing on with the expansion of (6.12), 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 time-reversal invariance of the D = 11
supergravity equations).
Now consider a brane-probe configuration with less surviving supersym-
metry, and with correspondingly weaker constraints on the probe worldvolume
σ-model. Corresponding to the wavek2-brane solution (6.3), 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 3-form gauge potential and one descending from the metric.
Now repeat the brane-probe analysis for the two-black-hole 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
3
− √
in (6.11). At order (∂y)0 , this now gives Vprobe = eA e 2 7 φ, but this potential
turns out to be just a constant because the heavy-brane background satisfies
3
A = 2√ 7
φ. Thus, we confirm the expected static zero-force condition for the
1
4 supersymmetric two-black-hole configuration descending from the wavek2-
brane solution (6.3). This zero-force 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 (6.11) that wipes out the conformal
factor occurring in the heavy brane background metric.
Proceeding on to (velocity)2 order, one now obtains a non-trivial probe
σ-model, with metric
γ mn = Hback (y)δ mn , (6.16)

where Hback is the harmonic function controlling the heavy brane’s background
fields; for the case of two black holes in D = 9, the harmonic function Hback
has the structure (1 + k/r6 ).
The above test-brane analysis for two D = 9 black holes is confirmed by
a more detailed study of the low-velocity scattering of supersymmetric black
holes performed by Shiraishi.48 The procedure is a standard one in soliton
physics: one promotes the moduli of a static solution to time-dependent func-
tions and then substitutes the resulting generalized field configuration back
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-

62
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 2-body forces. These two-body forces may
be derived from an effective action involving the position vectors of the two
black holes. Separating the center-of-mass motion from the relative motion,
one obtains the same modulus metric (6.16) as that found in the brane-probe
analysis above, except for a rescaling which replaces the brane-probe mass by
the reduced mass of the two-black-hole system.
Now we should resolve a puzzle of how this non-trivial d = 1 scattering
modulus σ-model turns out to be consistent with the surviving supersymme-
try.49 The modulus variables of the two-black-hole system are fields in one
dimension, i.e. time. The N -extended supersymmetry algebra in d = 1 is

{QI , QJ } = 2δ IJ Ĥ I = 1, . . . , N , (6.17)

where Ĥ is the Hamiltonian. A d = 1, N = 1 σ-model is specified by a

triple (M, γ, A[3] ), where M is the Riemannian σ-model manifold, γ is the
metric on M and A[3] is a 3-form on M which plays the rôle of torsion in
(+) (+) (+)
the derivative operator acting on fermions, ∇t = ∂t xi ∇i , where ∇i λj =
∇i λj + 21 Aj ik λk . The σ-model action may be written using N = 1 superfields
xi (t, θ) (where xi (t) = xi | , λi (t) = Dxi | ) as
θ=0 θ=0
Z
d j 1
I = − 21 dtdθ(iγij Dxi x + Aijk Dxi Dxj Dxk ) . (6.18)
dt 3!

One may additionally 50 have a set of spinorial N = 1 superfields ψ a , with

Lagrangian − 21 hab ψ a ∇t ψ b , where hab 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 (6.18), one starts by positing a second
set of supersymmetry transformations of the form δxi = ηI i j Dxj , and then
requires these transformations to close to form the N = 2 algebra (6.17); then
one also requires that the action (6.18) be invariant. In this way, one obtains
the equations

I2 = −1l (6.19a)
i i
Njk ≡ I[j,k] = 0 (6.19b)
γkl I k i I l j = γij (6.19c)

63
 (+)
∇(i I k j) = 0 (6.19d)
m m
∂[i (I j A|m|kl] ) − 2I [i ∂[m Ajkl]] ) = 0, (6.19e)

where (6.19a,b) follow from requiring the closure of the algebra (6.17) and
(6.19c-e) follow from requiring invariance of the action (6.18). Conditions
(6.19a,b) imply that M is a complex manifold, with I i j as its complex struc-
ture.
The structure of the conditions (6.19) is more complicated than might have
been expected. Experience with d = 1 + 1 extended supersymmetry 50 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 (6.19c–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 Lorentz-invariant 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 “wave-on-a-string” solution), with a lightplane metric that is not
Poincaré invariant.
Note also that the d = 1 “torsion” A[3] is not required to be closed in
(6.19). d = 1 supersymmetric theories satisfying (6.19) are analogous to (2,0)
chiral supersymmetric theories in d = 1 + 1, but the weaker conditions (6.19)
warrant a different notation for this wider class of models; one may call them
2b supersymmetric σ-models.49 Such models are characterized by a Kähler
geometry with torsion.
Continuing on to N = 8, d = 1 supersymmetry, one finds an 8b general-
ization 49 of the conditions (6.19), with 7 independent complex structures built
using the octonionic structure constants o ϕab c : δxi = η a Ia i j Dxj , a = 1, . . . 7,
with (Ia )8 b = δab , (Ia )b 8 = −δ b a , (Ia )b c = ϕa b c , where the octonion multipli-
cation rule is ea eb = −δab + ϕab c ec . Models satisfying such conditions have an
“octonionic Kähler geometry with torsion,” and are called OKT models.49
Now, are there any non-trivial solutions to these conditions? Evidently,
from the brane-probe and Shiraishi analyses, there must be. For our two
D = 9 black holes with a D = 8 transverse space, one may start from the
ansatz ds2 = H(y)ds2 (IE8 ), Aijk = Ωijk ℓ ∂ℓ H, where Ω is a 4-form on IE8 .
o We let the conventional octonionic “0” index be replaced by “8” here in order to avoid
confusion with a timelike index; the IE8 transverse space is of Euclidean signature.

64
Then, from the 8b generalization of condition (6.19d) one learns Ω8abc = ϕabc
and Ωabcd = −∗ ϕabcd ; from the 8b generalization of condition (6.19e) one learns
δ ij ∂i ∂j H = 0. Thus we recover the familiar dependence of p-brane solutions
on transverse-space harmonic functions, and we reobtain the brane-probe or
Shiraishi structure of the black-hole modulus scattering metric with
kred
Hrelative = 1 + , (6.20)
|y1 − y2 |6

where kred determines the reduced mass of the two black holes.

7 Duality symmetries and charge quantisation

As one can see from our discussion of Kaluza-Klein dimensional reduction in
Section 5, progression down to lower dimensions D causes the number of dila-
tonic scalars φ~ and also the number of zero-form potentials of 1-form field
strengths to proliferate. When one reaches D = 4, for example, a total of 70
such spin-zero fields has accumulated. In D = 4, the maximal (N = 8) super-
gravity equations of motion have a linearly-realized H = SU(8) symmetry; this
is also the automorphism symmetry of the D = 4, N = 8 supersymmetry alge-
bra relevant to the (self-conjugate) supergravity multiplet. In formulating this
symmetry, it is necessary to consider complex self-dual and anti-self-dual combi-
nations of the 2-form field strengths, which are the highest-rank field strengths
occurring in D = 4, higher ranks having been eliminated in the reduction or
by dualization. Using two-component notation for the D = 4 spinors, these
[ij]
combinations transform as Fαβ and F̄α̇β̇ [ij] , i, j = 1, . . . , 8, i.e. as a complex
28-dimensional dimensional representation of SU(8). Since this complex repre-
sentation can be carried only by the complex field-strength combinations and
not by the 1-form 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 spin-zero fields together, one finds that they form a rather
impressive nonlinear σ-model with a 70-dimensional manifold. Anticipating
that this manifold must be a coset space with H = SU(8) as the linearly-
realized denominator group, Cremmer and Julia 51 deduced that it had to be
the manifold E7(+7)/SU(8); since the dimension of E7 is 133 and that of SU(8)
is 63, this gives a 70-dimensional manifold. Correspondingly, a nonlinearly-
realized E7(+7) symmetry also appears as an invariance of the D = 4, N = 8
maximal supergravity equations of motion. Such nonlinearly-realized symme-
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

65
which supergravities arise by dimensional reduction, but this is not enough:
such symmetries act transitively on the σ-model manifolds, mixing both fields
arising from the metric and also from the reduction of the D = 11 3-form
potential A[3] in (1.1).
In dimensions 4 ≤ D ≤ 9, maximal supergravity has the sets of σ-model
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 spin-zero 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 θ-angles 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 Poincaré
group times the appropriate “internal” classifying symmetry, e.g. the group of
rigid (i.e. constant-parameter) Yang-Mills gauge transformations.
Table 2: Supergravity σ-model symmetries.

D G H
9 GL(2, IR) SO(2)
8 SL(3, IR) × SL(2, IR) SO(3) × SO(2)
7 SL(5, IR) SO(5)
6 SO(5, 5) SO(5) × SO(5)
5 E6(+6) USP(8)
4 E7(+7) SU(8)
3 E8(+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.

7.1 An example of duality symmetry: D = 8 supergravity

In maximal D = 8 supergravity, one sees from Table 2 that G = SL(3, IR) ×
SL(2, IR) and the isotropy group is H = SO(3) × SO(2). We have an (11 −
ijk ij
3 = 8) vector of dilatonic scalars as well as a singlet F[1] and a triplet F[1]

66
(i, j, k = 1, 2, 3) of 1-form field strengths for zero-form potentials. Taken all
together, we have a manifold of dimension 7, which fits in precisely with the
dimension of the (SL(3, IR) × SL(2, IR))/(SO(3) × SO(2)) coset-space manifold:
8 + 3 − (3 + 1) = 7.
Owing to the direct-product structure, we may for the time being drop
the 5-dimensional SL(3, IR)/SO(3) sector and consider for simplicity just the
2-dimensional SL(2, IR)/SO(2) sector. Here is the relevant part of the action:52
Z
SL(2) √ h
I8 = d8 x −g R − 12 ∇M σ∇M σ − 12 e−2σ ∇M χ∇M χ
1 σ 1 i
− e (F[4] )2 − χF[4] ∗F[4] (7.1)
2 · 4! 2 · 4!
√
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, hence
purely numerical).
On the scalar fields (σ, χ), the SL(2, IR) symmetry acts as follows: let
λ = χ + ieσ ; then  
a b
Λ= (7.2)
c d
with ab−cd = 1 is an element of SL(2, IR) and acts on λ by the fractional-linear
transformation
aλ + b
λ −→ . (7.3)
cλ + d
The action of the SL(2, IR) symmetry on the 4-form 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, IR) doublet together with

G[4] = eσ ∗F[4] − χF[4] , (7.4)

i.e.,    
F[4] F[4]
−→ (ΛT )−1 . (7.5)
G[4] G[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 (7.6)

into the corresponding Bianchi identity,

∇M ∗F M N P Q = 0 . (7.7)

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 (7.5), but since

67
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 (7.3, 7.5) is
a D = 8 analogue of ordinary Maxwell duality transformation in the presence of
scalar fields. Accordingly, we shall refer generally to the supergravity σ-model
symmetries as duality symmetries.
The F[4] field strength of the D = 8 theory supports elementary/electric
p-brane 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.52
This possibility is also reflected in the combined Bogomol’ny bound p for this
situation, which generalizes the single-charge bounds (4.16):

E 2 ≥ e−σ∞ (U + χ∞ V )2 + eσ∞ V 2 , (7.8)

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 (7.8) is itself SL(2, IR) invariant, provided that one transforms both
the moduli (σ∞ , χ∞ ) (according to (7.3)) 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 (7.8) reduces to E 2 ≥ U 2 +V 2 , which
is invariant under an obvious isotropy group H = SO(2).

7.2 p-form charge quantisation conditions

So far, we have discussed the structure of p-brane 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 n-form field strengths that can support the solution. At the quantum
level, however, an important restriction on this spectrum of solutions enters
into force: the Dirac-Schwinger-Zwanziger (DSZ) quantisation conditions for
particles with electric or magnetic or dyonic charges.53,54 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.55
p In comparing (7.8) to the single-charge bounds (4.16), one should take note that for F
[4] in
√
(7.1) we have ∆ = 4, so 2/ ∆ = 1.

68
 We shall first review a Wu-Yang style of argument,53 (for a Dirac-string
argument, see Ref.54,56 ) considering a closed sequence W of deformations of
one p-brane, say an electric one, in the background fields set up by a dual,
magnetic, p̂ = D − p − 4 brane. After such a sequence of deformations, one sees
from the supermembrane action (4.1) that the electric p-brane wavefunction
picks up a phase factor
 I 
iQe
exp AM1 ...Mp+1 dxM
1 ∧ . . . ∧ dxMp+1
, (7.9)
(p + 1)! W

where A[p+1] is the gauge potential set up (locally) by the magnetic p̂-brane
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 p-form charges appearing in
the supersymmetry algebra (1.5) are necessarily either infinite or are wrapped
around compact spacetime dimensions. For infinite p-branes, some deformation
sequences W will lead to a divergent integral in the exponent in (7.9); 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
p-form charge, and so one should not consider changing this p-form 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 “Dirac-insensitive” configurations, for which the phase in (7.9) vanishes.
Restricting attention to deformations that give non-divergent phases, one
may use Stoke’s theorem to rewrite the integral in (7.9):
I
Qe
AM1 ...Mp+1 dxM1 ∧ . . . ∧ dx
Mp+1
=
(p + 1)! W
Z
Qe
FM1 ...Mp+2 dxM1 ∧ . . . ∧ dxMp+2 = Qe ΦMW , (7.10)
(p + 2)! MW

where MW is any surface “capping” the closed surface W, i.e. a surface such
that ∂MW = W; ΦMW is then the flux through the cap MW . Choosing the
capping surface in two different ways, one can find a flux discrepancy ΦM1 −

69
ΦM2 = ΦM1 ∩M2 = ΦMtotal (taking into account the orientation sensitivity of
the flux integral). Then if Mtotal = M1 ∩ M2 “captures” the magnetic p̂-
brane, the flux ΦMtotal will equal the magnetic charge Qm of the p̂-brane; thus
the discrepancy in the phase factor (7.9) will be simply exp(iQe Qm ). Requiring
this to equal unity gives,53 in strict analogy to the ordinary case of electric and
magnetic particles in D = 4, the Dirac quantisation condition
Qe Qm = 2πn , n ∈ ZZ . (7.11)
The charge quantisation condition (7.11) is almost, but not quite, the full
story. In deriving (7.11), we have not taken into account the p-form character
of the charges. Taking this into account shows that the phase in (7.9) vanishes
for a measure-zero set of configurations of the electric and magnetic branes.55
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 p-brane ori-
ented along the directions {xM1 . . . xMp }. The phase factor (7.9) then becomes
H
exp(iQe W AM1 ...Mp R ∂xR/∂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 p-brane. For simplicity, one
may simply consider moving the electric p-brane by parallel transport around
the magnetic p̂-brane 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 p̂-branes 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 p̂-brane. Consideration of more general deformation
sequences yields the same result.55
Now one can see how the Dirac-insensitive configurations arise: the phase
in (7.9) 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 measure-zero set of Dirac-insensitive configura-
tions may be simply characterised in terms of the p and p̂ charges themselves
mag
by the condition Qel [p] ∧ Q[p̂] = 0. For such configurations, one obtains no
Dirac quantisation condition. To summarise, one may incorporate this ori-
entation restriction into the Dirac quantisation condition (7.11) by writing a
(p + p̂)-form quantisation condition
mag
mag
Qel
[p] ∧ Q[p̂]
Qel
[p] ∧ Q[p̂] = 2πn mag , n ∈ ZZ , (7.12)
|Qel
[p] | |Q[p̂] |

70
which reduces to (7.11) for all except the Dirac-insensitive set of configurations.

7.3 Charge quantisation conditions and dimensional reduction

The existence of Dirac-insensitive configurations may seem to be of only pe-
ripheral importance, given that they constitute only a measure-zero 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 (5.11)
between the field strengths in different dimensions. Now, the electric and mag-
netic charges carried by branes in D dimensions take the forms
Z  
~
Qe = e~c·φ ∗F + κ(A) (7.13a)
Z
Qm = F̃ , (7.13b)

where F̃ = dA, F = F̃ + (Chern-Simons modifications) (i.e. modifications

involving lower-order forms arising in the dimensional reduction similar to
those in the D = 9 case (5.41)) and ~c is the dilaton vector corresponding to
F in the dimensionally-reduced action (5.13). The term κ(A) in (7.13a) is
the analogue of the term 12 A[3] ∧ F[4] in (1.3). From the expressions (5.11) 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 Q11 and QD

i ij ijk
F[4] F[3] F[2] F[1]
V V V
Electric Q11
e = QD
e V QD
e Li QD
e Li Lj QD
e Li Lj Lk

Magnetic Q11
m = QD
m m Li
QD m Li Lj
QD m Li Lj Lk
QD

R
where Li = dz i is the compactification period of the reduction coordinate
R Q11−D
z i and V = d11−D z = i=1 Li is the total compactification volume. 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.
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,p̂) branes. We

71
have seen that for single-element brane solutions, there are two basic schemes,
as explained in Section 5: diagonal, which involves reduction on a worldvolume
coordinate, and vertical, which involves reduction on a transverse coordinate
after preparation by “stacking up” single-center solutions so as to generate a
transverse-space 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 measure-zero set of Dirac-
mag
insensitive configurations with Qel [p] ∧ Q[p̂] = 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 (p̂ − 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 Dirac-insensitive configuration.
Now consider the mixed reductions, e.g. diagonal/vertical. In performing
a vertical reduction of a magnetic p̂-brane by stacking up an infinite deck of
single-center branes in order to create the IR 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 Qm 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 gravitational-constant factor of κ2 that should properly have appeared in
the quantisation conditions (7.11, 7.12). As one may verify, the electric and
magnetic charges as defined in (1.3, 1.4) are not dimensionless. Thus, (7.11)
in D = 11 should properly have been written Qe Qm = 2πκ211 n. If one lets the
compactification length be denoted by L in the D-dimensional theory prior
to dimensional reduction, then one obtains a Dirac phase exp(iκ−2 D−1 Qe Qm L).
This fits precisely, however, with another aspect of dimensional reduction:
the gravitational constants in dimensions D and D − 1 are related by κ2D =
Lκ2D−1 . Thus, in dimension D − 1 one obtains the expected quantisation
condition Qe Qm = 2πκ2D−1 n. Note, correspondingly, that upon making a
mixed diagonal/vertical reduction, the electric and magnetic branes remain

72
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 2 2
D Qe Qm L ). Only one factor of L is absorbed into κD−1 , and one has
2 2
limL→0 L /κD = 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 p-form charges in the various supergravity theories related
by dimensional reduction. The existence of Dirac-insensitive configurations
plays a central rôle in establishing this accord, even though they represent
only a subset of measure zero from the point of view of the higher-dimensional
theory.
Another indication of the relevance of the Dirac-insensitive configurations
is the observation 55 that all the intersecting-brane solutions with some degree
of preserved supersymmetry, as considered in Section 6, correspond to Dirac-
insensitive configurations. This may immediately be seen in such solutions as
the 2 ⊥ 5(1) solution (6.10), but it is also true for solutions involving pp wave
and Taub-NUT elements.

7.4 Counting p-branes

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(ZZ) duality group, so in general one seeks to identify the group G(ZZ)∩H. The
value of this intersection is modulus-dependent, 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 transitively-acting G transformation, for example with a group element g.
Correspondingly, the isotropy subgroup H moves by conjugation with g,

H −→ gHg −1 . (7.14)

73
 The discretized duality group G(ZZ), 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. (7.13). 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 modulus-independent
relations of Table 3 imply that the lower-dimensional charges (7.13) do not
depend on the moduli.q
Another way to understand this is by comparison with ordinary Maxwell
electrodynamics, whereR an analogous charge would be that derived from the
can can µν
action IMax = − 1/(4e2 ) Fµν F , corresponding to a covariant derivative
can
Dµ = ∂µ + iAµ . This is analogous to our dimensionally reduced action (5.13)
~
from which the charges (7.13) are derived, because the modulus factors e~c·φ∞
~ dilatonic scalar dependence)
appearing in (5.13) (together with the rest of the φ
play the rôles of coupling constant factors like e−2 . If one wants to compare
this to the “conventional” charges defined with respect toRa conventional gauge
potential Aconv
µ = e−1 Acan
µ , for which the action is − /4 Fµν F
1 conv conv µν
, then
the canonical and conventional charges obtained via Gauss’s law surface inte-
grals are related by
Z Z
1 1 1
Qcan = d2 Σij ǫijk F can 0k = d2 Σij ǫijk F conv 0k = Qconv . (7.15)
2e2 2e e

Thus, in the Maxwell electrodynamics case, the dependence on the electric

charge unit e drops out in Qcan , although the conventional charge Qconv scales
proportionally to e. The modulus independence of the charges (7.13) works
in a similar fashion. Then, given that the discretised quantum duality group
G(ZZ) is defined by the requirement that it map the set of Dirac-allowed charges
onto itself, it is evident that the group G(ZZ), referred to the canonical charges
(7.13), does not depend on the moduli.
As a consequence of the different modulus dependences of H and of G(ZZ),
it follows that the size of the intersection group G(ZZ) ∩ 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.
q Note that the compactification periods Li 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 canonically-defined charges from the
physically relevant moduli.

74
 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(ZZ) ∩ 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 (7.3):

e−σ −→ (d + cχ)2 e−σ + c2 eσ

−σ
χe −→ (d + cχ)(b + aχ)e−σ + aceσ . (7.16)

Requiring a, b, c, d ∈ ZZ 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 −→ eσ + χ2 e−σ
−σ
χe −→ −χe−σ . (7.17)

Thus, for our truncated system, we find just an S2 discrete symmetry as the
quantum isotropy subgroup of SL(2, ZZ) at the distinguished point on the scalar
vacuum manifold. 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,
suppose that the zero-form gauge potential χ is small, and consider the S2
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] ) . (7.18)

One may again check (in fact to all orders, not just to lowest order in χ) that
(7.18) 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 . (7.19)

Considering this S2 transformation to lowest order in the zero-form χ has the

advantage that the sign-flip of φ may be “impressed” upon the ~a dilaton vector
for F[4] : ~a → −~a. The general structure of such G(ZZ) ∩ H transformations
will be found by considering the impressed action of this group on the dilaton
vectors.
Now consider the SL(3, IR)/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 zero-form gauge potentials, the action of SL(3, ZZ) ∩ H may
similarly by impressed upon the 3-form dilaton vectors, causing in this case a
permutation of the ~ai , generating for the D = 8 case overall the discrete group

75
S3 × S2 . Now that we have a bit more structure to contemplate, we can notice
that the G(ZZ) ∩ H transformations leave the (~a, ~ai ) dot products invariant.57
The invariance of the dilaton vectors’ dot products prompts one to return
to the algebra (5.15) 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, IR) groups, we recall that the weight vectors ~hi of the
fundamental representation of SL(N, IR) satisfy

N
X
~hi · ~hj = δij − 1 , ~hi = 0 . (7.20)
N i=1

±1
These relations are precisely those satisfied by √ 2
~a and √12 ~ai , corresponding
to the cases N = 2 and N = 3. This suggests that the action of the maximal
G(ZZ) ∩ 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 57 of G, so the action of the maximal G(ZZ) ∩ H p-brane classifying
symmetry is identified with that of the Weyl group of G.
As one proceeds down through the lower-dimensional cases, where the
supergravity symmetry groups shown in Table 2 grow in complexity, the above
pattern persists:57 in all cases, the action of the maximal classifying symmetry
G(ZZ) ∩ H may be identified with the Weyl group of G. This is then the group
that counts the distinct p-brane solutions r 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, IR) and H = SO(5), one finds that the action of G(ZZ) ∩ H
is equivalent to that of the discrete group S5 , which is the Weyl group of
SL(5, IR). In the lower-dimensional cases shown in Table 2, the discrete group
G(ZZ) ∩ H becomes less familiar, and is most simply described as the Weyl
group of G.
From the analysis of the Weyl-group duality multiplets, one may tabu-
late 57 the multiplicities of p-branes residing at each point of the plot given in
Figure 6. For supersymmetric p-branes 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-
r Ofcourse, these solutions must also fall into supermultiplets with respect to the unbroken
supersymmetry; the corresponding supermultiplet structures have been discussed in Ref.60

76
ilies discussed in Sections 5 and 5.3, the full set of p ≤ (D − 3) branes shown
in Figure 6 becomes “welded” together into one overall symmetrical structure.

Table 4: Examples of p-brane Weyl-group 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

7.5 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
(7.12) implies that, given a certain minimum “electric” charge (7.13a), 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 “unit-setting” brane types, and also to exploit fully the consequences of
the assumption that the G(ZZ) duality symmetry remains exactly valid at the
quantum level. We have already encountered one example of a “unit-setting”
brane in subsection 6.2, where we encountered the pp wave/Taub-NUT pair
of D = 11 solutions. We saw there that the Taub-NUT solution (6.7) is non-
singular provided that the coordinate ψ is periodically identified with period
L = 4πk, where k is the charge-determining parameter in the 3-dimensional
harmonic function H(y) = 1 + k/(|y|). Upon dimensional reduction down to
D = 10, one obtains a magnetic 6-brane solution, with a charge classically

77
discretised to take a value in the set

Qm = rL , r ∈ ZZ . (7.21)

Given these values for the magnetic charge, the D = 10 Dirac quantisation
condition
Qe Qm = 2πκ210 n , n ∈ ZZ , (7.22)
or, equivalently, as we saw in subsection 6.2, the quantisation of D = 11 pp
wave momentum in the compact ψ direction, gives an allowed set of electric
charges
2πκ210
Qe = n, n ∈ ZZ . (7.23)
L
Thus, the requirement that magnetic D = 10 6-branes oxidise up to non-
singular Taub-NUT solutions in D = 11 fully determines the 6-brane 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(ZZ) duality symmetries remain strictly unbro-
ken at the quantum level, then one may relate the 6-brane charge units to
those of other BPS brane types.s In doing so, one must exploit the fact that
brane solutions with Poincaré 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(ZZ) duality symmetries only
rotate between p-branes of the same worldvolume dimension, supported by
the same kind of field strength, as we have seen from our discussion of the
Weyl-group action on p-branes given in subsection 7.4. Upon reduction down
to dimensions Dred < D, however, the solutions descending from an original
p-brane in D dimensions are subject to a larger G(ZZ) 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.61 This link may be used to establish
relations between the charge units for the various p-form charges of differing
rank, even though the corresponding solutions are Dirac-insensitive to each
each other.
Another charge-unit-setting BPS brane species occurs in the D = 10 type
IIB theory. This theory has a well-known difficulty with the formulation of a
satisfactory action, although its field equations are perfectly well-defined. The
difficulty in formulating an action arise from the presence of a self-dual 5-form
s For details of the duality relations between charge units for different p-branes, see Ref.55

78
field strength, H[5] = ∗H[5] . The corresponding electrically and magnetically
charged BPS solutions are 3-branes, and, owing to the self-duality 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
8; for now, it will be sufficient for us to note that the dyonic 3-branes of
D = 10 type IIB theory are also a unit-setting brane species.55 The unit-setting
property arises because of a characteristic property of the Dirac-Schwinger-
Zwanziger quantisation condition for dyons in dimensions D = 4r + 2: for
(1) (1) (2) (2)
dyons (Qe , Qm ), (Qe , Qm ), this condition is symmetric:55

Q(1) (2) (2) (1) 2

e Qm + Qe Qm = 2πκ4r+2 n , n ∈ ZZ , (7.24)

unlike the more familiar antisymmetric DSZ condition that is obtained in di-
mensions D = 4r. The symmetric nature of (7.24) means that dyons may be
Dirac-sensitive to others of their own type,t quite differently from the anti-
symmetric cases in D = 4r dimensions. For the 45◦ dyonic 3-branes, one thus
obtains the quantisation condition
√
|Q[3] | = n πκIIB , n ∈ ZZ (7.25)

where κIIB is the gravitational constant for √ the type IIB theory. Then, using
duality symmetries, one may relate the πκIIB charge unit to those of other
supergravity R–R charges.
Thus, using duality symmetries together with the pp wave/Taub-NUT and
self-dual 3-brane charge scales, one may determine the charge-lattice units for
all BPS brane types.61,55 . 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 Li be equal,
1
Li = LIIB = L = (2πκ211 ) 9 , (7.26)

then the electric and magnetic charge-lattice units for rank-n field strengths
in dimension D are determined to be 55

∆Qe = LD−n−1 , ∆Qm = Ln−1 . (7.27)

8 Local versus active dualities

The proper interpretation of the discretised Cremmer-Julia G(ZZ) duality sym-
metry at the level of supergravity theory is subject to a certain amount of
t They will be Dirac-sensitive provided one of them is slightly rotated so as to avoid having any
common worldvolume directions with the other, in order to avoid having a Dirac-insensitive
configuration as discussed in subsection 7.3.

79
debate, but at the level of string theory the situation becomes more clear. In
any dimension D, there is a subgroup of G(ZZ) that corresponds to T duality,
which is a perturbative symmetry holding order-by-order in the string loop
expansion. T duality 62 consists of transformations that invert the radii of a
toroidal compactification, under which quantised 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.63 The well-founded basis, in string theory at least, for a
local interpretation of the T duality subgroup of G(ZZ) has led subsequently to
the hypothesis 58,59 that the full duality group G(ZZ) should be given a local
interpretation: sets of string solutions and moduli related by G(ZZ) transforma-
tions are to be treated as equivalent descriptions of a single state. This local
interpretation of the G(ZZ) 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 7.5, there is also
a G(ZZ) covariance of the set of charge vectors for physically inequivalent BPS
brane solutions. In the discussion of subsection 7.5, we did not consider in
detail the action of G(ZZ) on the moduli, because, as we saw in subsection 7.4,
the canonically-defined charges (7.13) are in fact modulus-independent.
Since the dilatonic and axionic scalar moduli determine the coupling con-
stants and vacuum θ-angles of the theory, these quantities should be fixed
when quantising 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 quantised fluctuations can be considered.
Thus, in considering physically-inequivalent solutions, one should compare
solutions with the same asymptotic values of the scalar fields. When this is
done, one finds that solutions carrying charges (7.13) related by G(ZZ) transfor-
mations generally have differing mass densities. Since the standard Cremmer-
Julia duality transformations, such as those of our D = 8 example in subsection
7.1, 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

80
a multiplet under the standard Cremmer-Julia G(ZZ) 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(ZZ) duality transforma-
tions, one cannot then turn around and use the same G(ZZ) transformations to
generate inequivalent solutions.
Thus, the question arises: is there any spectrum-generating symmetry
lying behind the apparently G(ZZ) invariant charge lattices of inequivalent so-
lutions that we saw in subsection 7.5? At least at the classical level, and for
single-charge (i.e. ∆ = 4) solutions, the answer 64 turns out to be ‘yes.’ We
shall illustrate the point using type IIB supergravity as an example.u

8.1 The symmetries of type IIB supergravity

Aside from the difficulties arising from the self-duality 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
Z
IIB
I10 = d10 x[eR+ 14 e tr(∇µ M−1 ∇µ M)− 121 T
e H[3] 1
M H[3] − 240 2
e H[5]

(i) (j)
1
− 2√ ǫ ∗ (B[4] ∧dA[2] ∧dA[2] )] .
2 ij
(8.1)

The 5-form self-duality condition

H[5] = ∗H[5] (8.2)
may be handled in the fashion of Ref.66 , being imposed by hand as an extra
constraint on the field equations obtained by varying (8.1). This somewhat
hybrid procedure will be sufficient for our present purposes.
The matrix M in (8.1) 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 zero-form gauge potential; explicitly, one has
 −φ 
e + χ2 e φ χ e φ
M= . (8.3)
χ eφ eφ
The doublet H[3] contains the field strengths of the 2-form gauge potentials
A[2] :
(1)
!
dA[2]
H[3] = (2) . (8.4)
dA[2]
u For a detailed discussion of SL(2, IR) duality in type IIB supergravity, see Ref.65

81
 The action (8.1) is invariant under the SL(2, IR) transformations

H[3] −→ (ΛT )−1 H[3] , M −→ Λ M ΛT , (8.5)

where the SL(2, IR) parameter matrix is

 
a b
Λ= , (8.6)
c d

and the SL(2, IR) 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
aτ + b
τ −→ . (8.7)
cτ + d
Note that since H[5] is a singlet under SL(2, IR), the self-duality constraint
(8.2), which is imposed by hand, also preserves the SL(2, IR) symmetry. Since
this SL(2, IR) transformation rotates the doublet A2] of electric 2-form poten-
tials amongst themselves, this is an “electric-electric” duality, as opposed to
the “electric-magnetic” duality discussed in the D = 8 example of subsection
7.1. 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 (8.1). This is a rather humble symmetry that is not often remarked
upon, but which will play an important role in constructing active SL(2, IR)
duality transformations for the physically distinct BPS string and 5-brane mul-
tiplets of the theory. As for pure source-free Einstein theory, the action (8.1)
transforms homogeneously as λ3 under the following scaling transformations:
(i) (i)
gµν −→ λ2 gµν , A[2] −→ λ2 A[2] , H[5] −→ λ4 H[5] ; (8.8)

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 (8.8) does not leave the action (8.1) invariant,
the λ3 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, IR) electric-magnetic duality of the D = 8 example given in subsection
7.1 shares with the transformation (8.8) the feature of being a symmetry only
of the equations of motion, and not of the action.
The SL(2, IR) transformations map solutions of (8.1) into other solutions.
We shall need to consider in particular the action of these transformations on

82
the charges carried by solutions. From the equations of motion of the 3-form
field strength H[3] in (8.1),

1
d ∗ (M H[3] ) = − √ H[5] ∧ ΩH[3] , (8.9)
2

where Ω is the SL(2, IR)-invariant tensor

 
0 1
Ω= , (8.10)
−1 0

one finds that the following two-component quantity is conserved:

Z  
∗ 1
Qe = (M H[3] ) + √ Ω(2B[4] ∧ H[3] − H[5] ∧ A[2] ) . (8.11)
3 2

Under an SL(2, IR) transformation, Qe transforms covariantly as a doublet:

Qe → ΛQe .
By virtue of the Bianchi identities for the 3-form field strength, one has in
addition a topologically-conserved magnetic charge doublet,
Z
Qm = H[3] , (8.12)

which transforms under SL(2, IR) as Qm → (ΛT )−1 Qm , i.e. contragrediently to

Qe . The transformation properties of the electric and magnetic charge doublets
are just such as to ensure that the Dirac quantization condition QT m Qe ∈
2πκ2IIB ZZ is SL(2, IR) invariant.
The overall effect of this standard SL(2, IR) symmetry on type IIB su-
pergravity 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 au-
tomorphism of a vector bundle, with the scalar fields’ SL(2, IR)/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
7.5 that the continuous classical Cremmer-Julia symmetries G break down to
discretised G(ZZ) symmetries that map between states on the quantum charge
lattice. In the present type IIB case, the classical SL(2, IR) symmetry breaks
down to SL(2, ZZ) at the quantum level. Taking the basis states of the IIB
charge lattice to be
   
1 0
e1 = e2 = , (8.13)
0 1

83
the surviving SL(2, ZZ) group will be represented by SL(2, IR) matrices with
integral entries.
As we have discussed above, the discretised duality symmetries G(ZZ) 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, ZZ)
transformations map between NS–NS and R–R states, and this is a distinctly
non-perturbative transformation. Adopting this hypothesis nonetheless, an
orbit of the standard SL(2, ZZ) transformation reduces to a single point; after
making the corresponding identifications, the scalar modulus space becomes
the double coset space SL(2, ZZ)\ SL(2, IR)/ SO(2).

8.2 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 continuous
classical SL(2, IR) symmetry of the type IIB theory. Finding the surviving
quantum-level SL(2, ZZ) later on will be a straightforward matter of restricting
the transformations to a subgroup. The procedure starts with a standard
SL(2, IR) transformation, which transforms the doublet charges (8.11) 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 moduli,
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, IR) 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, IR):
  
a b
Borel = a, b ∈ IR . (8.14)
0 a−1

This standard representation of the SL(2, IR) Borel subgroup clearly leaves
the basis charge vector e1 of Eq. (8.13) invariant up to scaling by a. For a
general charge vector Q′ , there will exist a corresponding projective stability
subgroup which is isomorphic to (8.14), but obtained by conjugation of (8.14)

84
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, IR)/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.
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 (8.8), aQ′ → λ2 aQ′ ,
in which one picks the rigid parameter λ such that λ2 a = 1. This almost
completes the construction of the active SL(2, IR). For the final step, note
that the transformation (8.8) also scales the metric, gµν → λ2 gµν = a−1 gµν .
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,
xµ → x′µ = a−1/2 xµ .
The overall active SL(2, IR) 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 Cremmer-Julia SL(2, IR) 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, IR)
there is an Iwasawa decomposition

Λ = b̃h , (8.15)

where b̃ ∈ BorelQ′ is an element of the projective stability group of the final

charge vector Q′ and where h ∈ Hτ∞ is an element of the stability subgroup
of τ∞ . Clearly, the Borel transformation that is needed in this construction is
just b = (b̃)−1 , leaving thus a transformation h ∈ Hτ∞ which does not change
the complex modulus τ∞ . The compensating scaling transformation t of the
form (8.8) and the associated general coordinate transformation also leave the
scalar moduli unchanged. The net active SL(2, IR) transformation thus is just
−φ∞
btΛ =th. Specifically, forτ∞ = χ∞ + ie and a transformation Λ mapping
pi pf
Qi = to Qf = = ΛQi , the h ∈ Hτ∞ group element is
qi qf
 
cos θ̃fi sin θ̃fi −1
hfi = V∞ V∞ , θ̃fi = θ̃f − θ̃i
− sin θ̃fi cos θ̃fi

tan θ̃i = eφ∞ (tan θi − χ∞ ) , tan θi = pi /qi , (8.16)

where the matrix V∞ is an element of Borel that has the effect of moving the

85
scalar modulus from the point τ = i to the point τ∞ :
 
1 e φ ∞ χ∞
V∞ = e−φ∞ /2 . (8.17)
0 e φ∞

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 (8.3), M = V V T . This fac-
torization makes plain the transitive action of the Borel subgroup on the
SL(2, IR)/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 antisymmetric-tensor gauge fields appearing in
the action (8.1).
The scaling-transformation part of the net active SL(2, IR) construction is
simply expressed as a ratio of mass densities,
mf
tfi = , m2i = QT −1
i M∞ Qi . (8.18)
mi
This expression reflects the fact that the scaling symmetry (8.8) acts on the
metric and thus enables the active SL(2, IR) transformation to relate solutions
at different mass-density levels mi,f . Since, by contrast, the mass-density levels
are invariant under the action of the standard SL(2, IR), 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
(8.18).
The group composition property of the active SL(2, IR) 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(Λ2 , Λ1 Q)O(Λ1 , Q) = O(Λ2 Λ1 , Q) . (8.19)

One may verify directly that the nonlinear realization given by (8.16, 8.18)
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, IR) symmetry thus becomes restricted to a discrete SL(2, ZZ)
subgroup in order to respect this charge lattice, and the active SL(2, IR) con-
structed above likewise becomes restricted to an SL(2, ZZ) subgroup. This

86
quantum-level discretised group of active transformations is obtained simply
by restricting the matrix parameters Λ for a classical active SL(2, IR) transfor-
mation so as to lie in SL(2, ZZ).
In lower-dimensional 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 p-brane solutions. This is the class of single-charge
solutions, for which the charges Q fall into highest-weight representations of
G. The duality groups shown in Table 2 are all maximally noncompact, and
possess an Iwasawa decomposition generalizing the SL(2, IR) case (8.15):

Λ = b̃h b̃ ∈ BorelQ , h ∈ Hmoduli , (8.20)

where BorelQ 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, IR) example of the type IIB theory, one may see
that this group action is transitive by noting that the matrix M (8.3) 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 T, for H orthogonal
#
V = V †, for H unitary (8.21)

ΩV † , for H a USp group.
(The D = 3 case in which G = E8(+8) and H = SO(16) needs to be treated as
a special case.67 )
Given the above group-theoretical structure, the construction of active G
symmetry transformations that preserve the scalar moduli proceeds in strict
analogy with the type IIB SL(2, IR) example that we have presented. This
construction depends upon the existence of a projective stability group 64,67
of the charge Q that is isomorphic to the Borel subgroup of G. This is the
case whenever Q transforms according to a highest-weight representation of G.
The BPS brane solutions with this property are the single-charge solutions with
∆ = 4. As we have seen in Section 6, BPS brane solutions with ∆ = 4/N can
be interpreted as coincident-charge-center cases of intersecting-brane solutions
with N elements, each of which would separately be a ∆ = 4 solution on its
own. The construction of active duality symmetries for such multiple-charge
solutions remains an open problem, for they have a larger class of integration

87
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 multi-charge cases are not of highest-weight type.
The active G(ZZ) 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
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 supergravity-level analyses of the persistence of BPS solutions with
arbitrary mass scales, despite the presence of apparently threatening quantum
corrections,64 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.

9 Non-compact σ-models, null geodesics, and harmonic maps

A complementary approach 68,69,70 to the analysis of brane solutions in terms

of the four D = 11 elemental solutions presented in Section 6 is to make
a dimensional reduction until only overall-transverse dimensions remain, and
then to consider the resulting nonlinear σ-model supporting the solution. In
such a reduction, all of the worldvolume and relative-transverse 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 characterised as oxidation
up to the top of Figure 6, or reduction down to the left edge Figure 6, i.e.
reduction down to BPS “instantons,” or p = −1 branes, with worldvolume
dimension d = 0.
The d = 0 instanton solutions are supported by 1-form 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 σ-model 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 spatial-coordinate Kaluza-Klein 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

88
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 v for the noncompact
σ-models in dimensions 9 ≥ D ≥ 3 are given in Table 5. These should be
compared to the standard Cremmer-Julia groups given in Table 2.

Table 5: Symmetries for BPS instanton σ-models.

D G H′
9 GL(2, IR) SO(1, 1)
8 SL(3, IR) × SL(2, IR) SO(2, 1) × SO(1, 1)
7 SL(5, IR) SO(3, 2)
6 SO(5, 5) SO(5, C)
|

5 E6(+6) USP(4, 4)
4 E7(+7) SU∗ (8)
3 E8(+8) SO∗ (16)

The sector of dimensionally-reduced supergravity that is relevant for the

instanton solutions consists just of the transverse-space Euclidean-signature
metric and the G/H ′ σ-model, with an action
Z
√

Iσ = dD y g R − 21 GAB (φ)∂i φA ∂j φB g ij , (9.1)

where the φA are σ-model fields taking values in the G/H ′ target space, GAB
is the target-space metric and g ij (y) is the Euclidean-signature metric for the
σ-model domain space. The equations of motion following from (9.1) are
√
√1 ∇i ( gg ij GAB (φ)∂j φB ) =0 (9.2a)
g

Rij = 12 GAB (φ)∂i φA ∂j φB , (9.2b)

where ∇i is a covariant derivative; when acting on a target-space vector VA , it

is given by
∇i VA = ∂i VA − ΓD E
AE (G)∂i φ VD , (9.3)
in which ΓA
BC
(G) is the Christoffel connection for the target-space metric GAB .
The action (9.1) and the field equations (9.2) are covariant with respect to
general-coordinate transformations on the σ-model target manifold G/H ′ . The
v The author would like to thank Chris Pope and Hong Lü for pointing out an error in an
earlier version of this table.

89
action (9.1) and the field equations (9.2) 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 domain-space transformations constitute a true gauge symmetry of
the dynamical system (9.1), while the σ-model target-space transformations
generally change the metric GAB (φA ) and so correspond to an actual symmetry
of (9.1) only for the finite-parameter group G of target-space isometries.
As in our original search for p-brane solutions given in Section 2, it is
appropriate to adopt an ansatz in order to focus the search for solutions. In
the search for instanton solutions, the metric ansatz can take a particularly
simple form:
g ij = δ ij , (9.4)
in which the domain-space metric is assumed to be flat. The σ-model equations
and domain-space gravity equations for the flat metric (9.4) then become

∇i (GAB (φ)∂i φB ) = 0 (9.5a)

Rij = 12 GAB (φ)∂i φA ∂j φB = 0 . (9.5b)

Now comes the key step 68 in finding instanton solutions to the specialised
equations (9.5): for single-charge solutions, one supposes that the σ-model
fields φA depend on the domain-space coordinates y i only through some inter-
mediate scalar functions σ(y), i.e.

φA (y) = φA (σ(y)) . (9.6)

After making this assumption, the σ-model φA equations (9.5a) become

 2 A 
dφA d φ dφB dφC
∇2 σ + (∂i σ)(∂i σ) + Γ A
(G) =0, (9.7)
dσ dσ 2 BC
dσ dσ
while the gravitational equation (9.5b) becomes the constraint
dφA dφB
GAB (φ) =0. (9.8)
dσ dσ
An important class of solutions to (9.7) is obtained by taking

∇2 σ = 0 (9.9a)
d2 φA A
dφB dφC
+ Γ (G) =0. (9.9b)
dσ 2 BC
dσ dσ
At this point, one can give a picture of the σ-model maps involved in the
system of equations (9.8, 9.9), noting that (9.9a) is just Laplace’s equation

90
and that (9.9b) is the geodesic equation on G/H ′ , while the constraint (9.8)
requires the tangent vector to a geodesic to be a null vector. The intermediate
function σ(y) is required by (9.9a) to be a harmonic function mapping from
the flat (9.4) 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 single-charge 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
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 (9.8). The σ-model solution (9.6)
oxidises back up to one of the single-charge brane solutions shown in Figure
6, and, conversely, any solution shown in Figure 6 may be reduced down to
a corresponding noncompact σ-model solution of this type. This sequence of
σ-model maps is sketched in Figure 7.

σ(y)

∇2 σ = 0
( nu )
σ
φ (A
ll)

ym

G/
H'
flat IE D

Figure 7: Harmonic map from IED to a null geodesic in G/H ′ .

An extension 69,70 of this σ-model picture allows for solutions involving

multiple harmonic maps σa (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 σa parameter space
must be null. This is the σ-model construction that generates multi-charge so-
lutions, giving rise to intersecting-brane solutions of the types discussed in Sec-
tion 6. As with the intersecting-brane solutions, however, there are important
compatibility conditions that must be satisfied in order for such multi-charge

91
solutions to exist. We saw in subsection 6.2 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 σ-model
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 σ-model equations for the matrix fields M (y m ) are
simply written
∇i (M −1 ∂i M ) = 0 . (9.10)
Points on the geodesic submanifold with affine parameters σa may be writ-
ten X
M = A exp( B a σa ) , (9.11)
a

where the constant matrices B a give the velocities for the various geodesics
parametrised by the σa , while an initial point on these geodesics is specified by
the constant matrix A. The compatibility condition between these velocities
is given by the double-commutator condition 70

[[B a , B b ], B c ] = 0 . (9.12)

This condition allows one to rewrite (9.11) as

!
XX Y
M = A exp − 12 [B b , B c ]σb σc exp(B a σa ) , (9.13)
c>b b a

where the first factor commutes with the B a as a result of (9.12). The matrix
current then becomes
X XX
M −1 ∂i M = B a ∂i σa − 12 [B b , B c ](σb ∂i σc − σc ∂i σb ) , (9.14)
a c>b b

and this is then seen to be conserved provided the σa satisfy ∇2 σa (y) = 0,

i.e. they are harmonic maps from the Euclidean overall-transverse space of
the y m into the geodesic submanifold (9.11). The constraint imposed by the
gravitational equation is
X
Rij = 41 tr(B a B b )∂i σa ∂j σb = 0 , (9.15)
a,b

which is satisfied provided the geodesics parametrised by the σa are null and
orthogonal, i.e.
tr(B a B b ) = 0 . (9.16)

92
 The general set of stationary multi-charge brane solutions is thus ob-
tained in the σ-model construction by identifying the set of totally null, totally
geodesic submanifolds of G/H ′ such that the velocity vectors satisfy the com-
patibility condition (9.12).
Aside from the elegance of the above σ-model 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 transverse-space
geometry is flat, Eq. (9.4). This is clearly more restrictive than is really neces-
sary; one could just as well have a more general Ricci-flat domain-space geom-
etry, with a correspondingly covariantised constraint for the null geodesics on
the noncompact manifold G/H ′ . The use of more general Ricci-flat transverse
geometries is at the basis of “generalised p-brane” solutions that have been
considered in Refs.73,74

10 Concluding remarks
In this review, we have discussed principally the structure of classical p-brane
solutions to supergravity theories. Some topics that deserve a fuller treatment
have only been touched upon here. For example, the worldvolume symme-
tries of p-brane sources, and in particular the important subject of κ sym-
metry, which bridges the gap between the full target-space supersymmetry of
the ambient supergravity theory and the fractional supersymmetry surviving
in the BPS brane background, have only been touched upon. For a fuller
treatment, the reader is referred to Refs 4,7 , or to the more recent discussions
of κ-symmetric actions for cases involving R–R sector antisymmetric-tensor
fields.75
Another aspect of the p-brane story, which we have only briefly presented
here in Section 6, is the large family of intersecting branes. These include 74 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.44,46 , 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 string-theoretic 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
rôles that BPS supergravity solutions play as states stable against the effects of

93
quantum corrections. In this emerging picture, the duality symmetries that we
have discussed in Section 7 play an essential part, uniting the underlying type
IIA, IIB, E8 × E8 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 field-theory 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, François Englert, Gary Gibbons, Hong Lü,
George Papadopoulos, Chris Pope, and Paul Townsend. The author would
like to thank The Abdus Salam International Center for Theoretical Physics
for the invitations to give lectures at two successive summer schools, on which
this review is based, and also CERN, SISSA, the Ecole Normale Supérieure,
UCLA and the University of Pennsylvania for hospitality at different periods
during the writing. This work was supported in part by the Commission of
the European Communities under contract ERBFMRX-CT96-0045.

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 mod-8,” J. Phys. A15, 3763 (1982).
4. K.S. Stelle and P.K. Townsend, “Are 2-branes 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), hep-th/9412184;
8. M.J. Duff, “Supermembranes,” hep-th/9611203.
9. J. Polchinski, “Tasi Lectures on D-branes,” hep-th/9611050.
10. E.S. Fradkin and A.A. Tseytlin, Phys. Lett. B 158, 316 (1985); Nucl.
Phys. B 261, 1 (1985).

94
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).
13. H. Lü, C.N. Pope, E. Sezgin and K.S. Stelle, “Stainless Super
p-branes,” Nucl. Phys. B 456, 669 (1996), hep-th/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. Lü, C.N. Pope, E. Sezgin and K.S. Stelle, “Dilatonic p-brane
solitons,” Nucl. Phys. B 371, 46 (1996), hep-th/9511203.
18. M.J. Duff and K.S. Stelle, “Multi-membrane 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),
hep-th/9302049;
M.J. Duff, G.W. Gibbons and P.K. Townsend, Phys. Lett. B 332, 321
(1994), hep-th/9405124.
20. E. Bergshoeff, E. Sezgin and P.K. Townsend, Phys. Lett. B 189, 75
(1987).
21. E. Witten, “Bound states of strings and p-branes,” Nucl. Phys. B 460,
335 (1996), hep-th/9510135;
M.R. Douglas, “Branes within branes,” hep-th/9512077;
P.K. Townsend, “Brane Surgery,”in Proc. European Res. Conf. on
Advanced Quantum Field Theory, La Londe-les-Maures, Sept. 1996,
hep-th/9609217.
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. Lü, C.N. Pope and K.S. Stelle, “Vertical Versus Diagonal
Dimensional Reduction for p-branes,” Nucl. Phys. B 481, 313 (1996),
hep-th/9605082.
24. S.W. Hawking and G.F.R. Ellis, The Large-Scale Structure of
Space-Time, (Cambridge University Press, 1973).
25. R. Güven, “Black p-brane solutions of D = 11 supergravity theory,”

95
 Phys. Lett. B 276, 49 (1992).
26. G.W. Gibbons, G.T. Horowitz and P.K. Townsend, Class. Quantum
Grav. 12, 297 (1995), hep-th/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),
hep-th/9306052.
28. M.J. Duff, H. Lü and C.N. Pope, “The Black Branes of M -theory,”
Phys. Lett. B 382, 73 (1996), hep-th 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. Lü and C.N. Pope, p-brane solitons in maximal supergravities, Nucl.
Phys. B 465, 127 (1996), hep-th/9512012.
32. H. Lü and C.N. Pope, “An approach to the classification of p-brane
solitons,” hep-th/9601089.
33. R. Khuri, Nucl. Phys. B 387, 315 (1992), hep-th/9205081;
J.P. Gauntlett, J.A. Harvey and J.T. Liu, Nucl. Phys. B 409, 363
(1993), hep-th/9211056.
34. B.R. Greene, A. Shapere, C. Vafa and S-T. Yau, Nucl. Phys. B 337, 1
(1990);
G.W. Gibbons, M.B. Green and M.J. Perry, Phys. Lett. B 370, 37
(1996), hep-th/9511080.
35. P.M. Cowdall, H. Lü, C.N. Pope, K.S. Stelle and P.K. Townsend,
“Domain Walls in Massive Supergravities,” Nucl. Phys. B 486, 49
(1997), hep-th/9608173.
36. J. Scherk and J.H. Schwarz, Phys. Lett. B 82, 60 (1979).
37. E. Bergshoeff, M. de Roo, M.B. Green, G. Papadopoulos and P.K.
Townsend, “Duality of Type II 7-branes and 8-branes,” Nucl. Phys. B
470, 113 (1996), hep-th/9601150.
38. N. Kaloper, R.R. Khuri and R.C. Myers, “On generalised axion
reductions,” Phys. Lett. B 428, 297 (1998), hep-th/9803066.
39. H. Lü and C.N. Pope, “Domain walls from M-branes,” Mod. Phys.
Lett. A 12, 1087 (1997), hep-th/9611079;
I.V. Lavrinenko, H. Lü and C.N. Pope, “From topology to generalised
dimensional reduction,” Nucl. Phys. B 492, 278 (1997),
hep-th/9611134.
40. M. Cvetic, “Extreme domain wall — black hole complementarity in
N = 1 supergravity with a general dilaton coupling,” Phys. Lett. B
341, 160 (1994).
41. M. Cvetic and H.H. Soleng, “Supergravity domain walls,” Physics

96
 Reports 282, 159 (1997), hep-th/9604090.
42. J. Rahmfeld, Phys. Lett. B 372, 198 (1996), hep-th/9512089;
N. Khviengia, Z. Khviengia, H. Lü and C.N. Pope, “Intersecting
M-branes and Bound States,” Phys. Lett. B 388, 21 (1996),
hep-th/9605077;
M.J. Duff and J. Rahmfeld, “Bound States of Black Holes and Other
p-branes,” Nucl. Phys. B 481, 332 (1996), hep-th/9605085.
43. H.W. Brinkmann, Proc. Nat. Acad. Sci. 9, 1 (1923).
44. G. Papadopoulos and P.K. Townsend, “Intersecting M-branes,” Phys.
Lett. B 380, 273 (1996), hep-th/9603087;
A. Tseytlin, “Harmonic superpositions of M-branes,” Nucl. Phys. B
475, 149 (1996);
I.R. Klebanov and A.A. Tseytlin, “Intersecting M-branes as
four-dimensional black holes,” Nucl. Phys. B 475, 179 (1996);
K. Berndt, E. Bergshoeff and B. Janssen, “Intersecting D-branes in ten
dimensions and six dimensions,” Phys. Rev. D 55, 3785 (1997),
hep-th/9604168;
J. Gauntlett, D. Kastor and J. Traschen, “Overlapping branes in
M-theory,” Nucl. Phys. B 478, 544 (1996).
45. R. Sorkin, Phys. Rev. Lett. 51, 87 (1983);
D.J. Gross and M.J. Perry, Nucl. Phys. B 226, 29 (1983).
46. J.P. Gauntlett, “Intersecting Branes,” hep-th/9705011.
47. A.A. Tseytlin, “ ‘No force’ condition and BPS combinations of p-branes
in eleven dimensions and ten dimensions,” Nucl. Phys. B 487, 141
(1997), hep-th/9609212.
48. K. Shiraishi, Nucl. Phys. B 402, 399 (1993).
49. 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), hep-th/9706207.
50. R. Coles and G. Papadopoulos, “The geometry of one-dimensional
supersymmetric non-linear sigma models,” Class. Quantum Grav. 7,
427 (1990).
51. E. Cremmer and B. Julia, Nucl. Phys. B 159, 141 (1979).
52. J.M. Izquierdo, N.D. Lambert, G. Papadopoulos and P.K. Townsend,
“Dyonic Membranes,” Nucl. Phys. B 460, 560 (1996),
hep-th/9508177.
53. R. Nepomechie, “Magnetic monopoles from antisymmetric tensor gauge
fields,” Phys. Rev. D 31, 1921 (1985).
54. C. Teitelboim, Phys. Lett. B 67, 63, 69 (1986).
55. M. Bremer, H. Lü, C.N. Pope and K.S. Stelle, “Dirac quantisation

97
 conditions and Kaluza-Klein reduction,” Nucl. Phys. B 529, 259
(1998), hep-th/9710244.
56. S. Deser, A. Gomberoff, M. Henneaux and C. Teitelboim, “Duality,
self-duality, sources and charge quantisation in Abelian N-form
theories,” Phys. Lett. B 400, 80 (1997), hep-th/9702184.
57. H. Lü, C.N. Pope and K.S. Stelle, “Weyl Group Invariance and p-brane
Multiplets,” Nucl. Phys. B 476, 89 (1996), hep-th/9602140.
58. C.M. Hull and P.K. Townsend, Nucl. Phys. B 438, 109 (1995),
hep-th/9410167.
59. E. Witten, “String theory dynamics in various dimensions,” Nucl.
Phys. B 443, 85 (1995).
60. M.J. Duff and J. Rahmfeld, “Bound States of Black Holes and Other
p-branes,” Nucl. Phys. B 481, 332 (1996), hep-th/9605085.
61. J.H. Schwarz, “The power of M-theory,” Phys. Lett. B 367, 97 (1996),
hep-th/9510086.
62. 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), hep-th/9401139.
63. A. Sen, “Electric and magnetic duality in string theory,” Nucl. Phys. B
404, 109 (1993), hep-th/9207053;
“SL(2, ZZ) duality and magnetically charged strings,” Int. J. Mod.
Phys. A 8, 5079 (1993), hep-th/9302038.
64. E. Cremmer, H. Lü, C.N. Pope and K.S. Stelle, “Spectrum-generating
symmetries for BPS Solitons,” Nucl. Phys. B 520, 132 (1998),
hep-th/9707207.
65. J.H. Schwarz, “An SL(2, ZZ) multiplet of type IIB superstrings,” Phys.
Lett. B 360, 13 (1995); Erratum ibid. 364, 252 (1995),
hep-th/9508143.
66. E. Bergshoeff, C.M. Hull and T. Ortin, “Duality in the type II
superstring effective action,” Nucl. Phys. B 451, 547 (1995),
hep-th/9504081.
67. E. Cremmer, B. Julia, H. Lü and C.N. Pope, “Dualisation of dualities,
I,” Nucl. Phys. B 523, 73 (1998), hep-th/9710119.
68. G. Neugebaur and D. Kramer, Ann. der Physik (Leipzig) 24, 62 (1969).
69. P. Breitenlohner, D. Maison and G. Gibbons, Comm. Math. Phys.
120, 253 (1988).
70. G. Clement and D. Gal’tsov, “Stationary BPS solutions to dilaton-axion
gravity,” Phys. Rev. D 54, 6136 (1996), hep-th/9607043;
D.V. Gal’tsov and O.A. Rytchkov, “Generating branes via

98
 sigma-models,” Phys. Rev. D 58, 122001 (1998), hep-th/9801160.
71. H. Lü and C.N. Pope, “Multi-scalar p-brane solitons,” Int. J. Mod.
Phys. A 12, 437 (1997), hep-th/9512153.
72. N. Khviengia, Z. Khviengia, H. Lü and C.N. Pope, “Intersecting
M-branes and bound states,” Phys. Lett. B 388, 21 (1996),
hep-th/9605077.
73. K. Becker and M. Becker, “M-theory on eight-manifolds,” Nucl. Phys.
B 477, 155 (1996), hep-th/9605053.
74. M. Berkooz, M.R. Douglas and R.G. Leigh, “Branes intersecting at
angles,” Nucl. Phys. B 480, 265 (1996), hep-th/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), hep-th/9702202.
75. P.S. Howe and E. Sezgin, “Superbranes,” Phys. Lett. B 390, 133
(1997), hep-th/9607227;
P.S. Howe and E. Sezgin, “D = 11, p = 5,” Phys. Lett. B 394, 62
(1997), hep-th/9611008;
M. Cederwall, A. von Gussich, B.E.W. Nilsson and A. Westerberg,
“The Dirichlet Super Three Brane in Ten-Dimensional Type IIB
Supergravity,” Nucl. Phys. B 490, 163 (1997), hep-th/9610148;
M. Aganagic, C. Popescu and J.H. Schwarz, “D-brane actions with local
kappa-symmetry,” Phys. Lett. B 393, 311 (1997), hep-th/9610249;
M. Cederwall, A. von Gussich, B.E.W. Nilsson, P. Sundell and A.
Westerberg, “The dirichlet super p-branes in ten-dimensional Type IIA
and IIB supergravity,” Nucl. Phys. B 490, 179 (1997),
hep-th/9611159;
E. Bergshoeff and P.K. Townsend, “Super D-branes,” Nucl. Phys. B
490, 145 (1997), hep-th/9611173.
I. Bandos, D. Sorokin and M. Tonin, “Generalized Action Principle and
Superfield Equations of Motion for d = 10 D-p-branes,” Nucl. Phys. B
497, 275 (1997), hep-th/9701127.

99