Preprint typeset in JHEP style - HYPER VERSION

Supergravity
Lectures by Prof Gary W Gibbons
Notes typeset by Ziyang Hu
Abstract: These are the unocial notes for the Part III course given in Easter term
2009 in DAMTP, the University of Cambridge. The (ocial) books and references for this
course are:
P. van Nieuwenhuizen, Physics Report 68 (1981) 189;
P. van Nieuwenhuizen in Supergravity 81, CUP, eds. Ferrara and Taylor;
P. van Nieuwenhuizen in Superstrings and Supergravity, eds. Davies and Sutherland;
D.Z. Freedman, B. de Witt in Supersymmetry, eds. Dietz et al.
Some examples given in the introduction section are omitted from these notes, but this
should not aect understanding of subsequent material. Please direct any spotted errors,
suggestions etc. to [email protected].
Contents
1. Introduction 2
1.1 Conventions 2
1.2 Dimensions and Planck units 2
1.3 Motivation for supergravity 3
1.4 Supersymmetry 4
1.5 The current status of supergravity 4
2. General relativity and the action principle 5
2.1 Moving frames 5
2.2 Connection 1-forms 6
2.3 Poincare gauge theory 7
2.4 Action principles 8
2.5 The 1.5 formalism 11
3. Spinors and the Dirac equation 12
3.1 Cliord algebra and Majorana spinors 12
3.2 Dirac and Weyl spinors, complex structure 13
3.3 Coupling to gravity 14
3.4 Einstein-Cartan-Weyl-Sciama-Kibble theory 17
4. Supergravity lagrangian and super invariance 19
4.1 Rarita-Schwinger equation in at spacetime 19
4.2 ^ = 1 supergravity 21
4.3 ^ = 2 supergravity 23
4.4 Super invariance of the exact theory 24
4.5 Fierz identities 25
4.6 Supersymmetric background 26
4.7 Super Poincare group and gauge theory 29
5. Wittens proof of the positive energy theorem 34
6. Central charges and BPS states 39
1
1. Introduction
1.1 Conventions
Our conventions is mostly that of Hawking-Ellis and Misner-Thorne-Wheeler. For the
metric tensor in general relativity, we will use the west coast signature (, +, +, +). The
covariant derivative is dened as

.
Note our rather unconventional placement of indices for the connection coecients: the
leftmost index is always the dierentiation index. We do not assume that the connection
is symmetric, and hence the torsion may not vanish:

,=

, T

,= 0.
The Riemann curvature tensor is
R

,
the Ricci tensor and Ricci scalar are
R

= R

, R = R

= R

,
and the Einstein eld equation is
R

1
2
Rg

=
8G
c
4
T

.
In places we will be replacing Newtons constant by

2
=
8G
c
3
.
For spinors, our gamma matrices satisfy

= 2

where the Minkowski metric is = diag(1, 1, 1, 1). To translate from the convention of
some other books:

.
1.2 Dimensions and Planck units
Dimensional analysis is a useful tool for working out some equations and coecients in a
physical theory, and we will use it in several places throughout the course. Here we will
give a brief overview and establish notation.
The eects of quantum gravity becomes important when half of the Schwarz radius is
comparable to the Compton radius
Gm
p
c
2
=

m
p
c
.
2
From this we can dene the Planck mass, length and time
m
p
=
_
c
G
10
5
grams 10
19
GeV
l
p
=
_
G
c
3
=
_

2
8
10
33
cm
t
p
=
_
G
c
5
=
_

2
8c
2
10
44
seconds
i.e. huge mass (compared to, say, a proton) and very small distance and intervals of time.
The units of relevant quantities are
[] = L length
[x

] = L
[R

1
2
Rg

] = L
2
[T

] = MT
2
L
1
energy per unit volume
[c
4
/8G] = MLT
2
tension.
In this course, we will set = c = 1.
1.3 Motivation for supergravity
Supergravity (SUGRA) is an extension of Einsteins general relativity to include supersym-
metry (SUSY). General relativity demands extensions since it has shortcomings including
at least the following:
Spacetime singularities. The singularity theorems of Penrose, Hawking and Geroch
shows that general relativity is incomplete: it predicts its own demise.
Failure to unify gravity with the strong and electroweak forces. In the Einstein
equation, the left hand side, i.e. spacetime geometry, is a house of marbles, whereas
the right hand side, i.e. matter elds, is lowly hovel. Historically, Kaluza-Klein
theory addressed this problem. However, it did not give realistic predictions.
Incompatibility with quantum mechanics. Conceptually, the role of time in general
relativity is very dierent from its role in quantum theory. If we think of the relativ-
istic time as an operator, its unitarity, which is required in a consistent quantum
theory, is not obvious. A Hilbert space based on curved geometry is dicult to dene.
Computationally, pure quantum gravity theory is not thought to be renormalizable
and hence has little predictive power.
If we include supersymmetry in a theory of gravity, the situation becomes a little bit better,
since the simplest example of divergences: zero point energy of the vacuum, can potentially
be cancelled by super partners of ordinary particles.
3
1.4 Supersymmetry
We here give a very brief overview of supersymmetry, one of the ingredients of supergravity.
Quantum mechanically, a supersymmetric theory is a theory in which the Hilbert space
can be written as a direct sum
H = H
B
H
F
and there exists self-adjoint operators Q
i
= Q

i
, i = 1, 2, . . . , 4^ acting on H, which satisfy
Q
i
, Q
j
=
ij
H, [Q
i
, H] = 0, Q
i
H
B
H
F
, Q
i
H
F
H
B
where H is a certain hamiltonian operator. In a relativistic theory, the operators Q
i
carry
angular momentum
1
2
. If a state [) has spin s, then Q
i
[) has spin s
1
2
. States fall
into supermultiplets with respect to actions of these operators.
The energy expectation value of a state with superpartners can be calculated
[H[) = 2[Q

i
Q
i
[) 0.
So in a supersymmetric theory, the energy is always non-negative.
The simplest supersymmetric theory is the case ^ =
1
2
. If we dene
a =
1

2
(Q
1
+iQ
2
), a

=
1

2
(Q
1
iQ
2
)
we recover the creation and annihilation operator relations for the harmonic oscillator
[a, a] = 0, a, a

= H.
In this case, we have a single multiplet with 2 states.
In general, the multiplets are of dimension 4
N
. These are called the long multiplets
Q

, Q

= C

, P
0
= H, C =
0
, , = 1, . . . , 4.
For example, it is easy to construct a theory in which
Q
1
, Q
1
= H +P
1
Q
2
, Q
2
= H P
1
Q
3
, Q
3
= H +P
1
Q
4
, Q
4
= H P
1
where H [P[. For the special case where states are lightlike H = [P[ = [P
1
[, half of the
states will vanish, and we are left with a short multiplet.
1.5 The current status of supergravity
Currently, supergravity is generally thought of as
a reliable approximation to M-theory at low energy;
a valuable technical tool (e.g. Wittens proof of the positive energy theorem);
an essential ingredient for supersymmetric phenomenology (minimal supersymmetric
standard model coupled to ^ = 1 supergravity);
an essential ingredient for the AdS/CFT correspondence of Maldacena.
4
2. General relativity and the action principle
2.1 Moving frames
To dene spin structure on spacetime, we will need to formulate general relativity in the
moving frame language. Let
e
a
= e
a

(x)dx

be a basis of 1-forms. The last part of the Greek alphabet denotes world indices, i.e. of
local coordinates, whereas the rst part of the Latin alphabet denotes tangent space indices,
i.e. of moving frames. Then
e
a
= e

a
(x)

x

form a basis of vector elds orthogonal to the basis of 1-forms: e

a
(e
b
) =
a
b
. In coordinates
e
a

(x)e

b
(x) =
a
b
.
We can think of e
a

as components of the matrix e. Then e

a
are just the components of
the inverse matrix e
1
. To carry this analogy with matrices further, we will always write
the upstairs index rst, even though it really does not matter. If in an expression the
contraction is not between adjacent indices, a matrix transpose is technically needed.
If we contract the tangent space index instead of the world index, we get
e

a
e
a

.
We can exchange world index for tangent space index, i.e. translating from a holonomic
frame to a moving frame. For example, in the case of a vector,
V

V
a
= e
a

.
For our basis, the metric of the moving frame is pseudo-orthonormal
g
ab
= e

a
g

b
=
ab
= diag(1, 1, 1, 1), g

= e
a

ab
e
b

.
Tangent space index can be raised and lowered with
ab
and
ab
.
The volume form on a manifold is =

g d
4
x where g = det g

. Since det g

=
(det e)
2
, we have
= [e[ d
4
x
= e
0
e
1
e
2
e
3
= e
0

e
1

e
2

e
3

dx

dx

dx

dx

.
The whole expression can be checked using the denition of the determinant and the fact
that dx

dx

dx

dx

. Our convention is that

0123
= 1. Note that our wedge
product includes the appropriate normalization factor.
5
2.2 Connection 1-forms
Let be a metric compatible connection, possibly with torsion. Acting on the basis,
e
a
=
a
b
e
b
,
a
b
= e
c

c
a
b
=

a
b
dx

c
a
b
is called the Ricci rotation coecients (old-fashioned) and
a
b
the spin connection.
Metric compatibility implies
ab
=
ba
, and in this case,
e
a
=

a
b
e
b

dx

dx

=
1
2
(

a
b
e
b

dx

dx

e
b

b
a
dx

dx

)
=
a
b
e
b
.
The covariant derivative acting on a vector V = V
a
e
a
gives
(V )

a
=

V
a
+

a
b
V
b
.
To write this in a more compact way, we think of

V
a
as a vector-valued 1-form

V
a
x

e
a
= dV
a
e
a
,
then
V
a
= dV
a
+
a
b
V
b
.
Cartans rst structural equation is
de
a
+
a
b
e
b
= T
a
, T
c
= T
a
c
b
e
a
e
b
where the torsion form T
c
is a vector-valued 2-form. This is derived as follows: expand
e
a
=
a
b
e
b
in coordinate basis

e
a

e
a

a
b
e
b

= 0
and antisymmetrise. We see that the our connection decomposes

a
b
=

a
b
(e) +K

a
b
where

denotes the Levi-Civita connection in moving frame, and K is the contorsion

K

=
1
2
(T

+T

+T

).
In a holonomic frame, this is

+K

.
For a function f,
(

)f = T

f.
The Riemann tensor for this connection is
R

.
6
We think of it as a 2-form valued matrix
R

= R

dx

dx

and convert indices using frames

R
a
b
= e
a

e
b

dx

dx

then, it easily follows

R
a
b
= d
a
b
+
a
c

c
b
.
The important thing to note is that R
a
b
transforms homogeneously under local Lorentz
transformation (or tetrad rotation) SO(3, 1), while does not. Indeed,
e
a
=
a
b
e
b
d e = de +d e = e +d
1
e
=
1
+d
1

R
a
b
=
a
c
R
c
d
(
1
)
d
b
.
The Ricci tensor need not be symmetric in our theory: R

,= R

. We can think of
it as a vector-valued 1-form R
a

dx

. The Ricci scalar is then R = g

= e
a

R
a

().
2.3 Poincare gauge theory
To make the bundle structure clear and to ease our subsequent introduction of spin struc-
ture, we think of gravity as gauge symmetries. The Poincare group, i.e. the local symmetry
group of general relativity, is the pseudo-Euclidean group E
3,1
= O(3, 1) R
4
, a semidirect
product of rotations with translations. We can write its action on spacetime coordinates
in matrix notation
_

a
b
a
b
0 1
__
x
b
1
_
=
_

a
b
x
b
+a
b
1
_
.
Its Lie algebra is so(3, 1) R
4
. The Lie algebra of R
4
is still itself, while for O(3, 1) it is
given by

a
b
=
a
b
+
a
b
+. . . ,
ab
=
ba
.
So in this gauge theory, innitesimal gauge transformations are generated by translations
and frame rotations. In any gauge eld theory, the gauge eld takes values in the Lie
algebra of the group. So we can write a eld as
=
_

a
b
a
a
0 0
_
(c.f. the Higgs eld). A covariant derivative is needed to make the gauge symmetry a
local one. Following Cartan, a covariant derivative (connection) in this case is just a Lie
algebra-valued 1-form,
/ =
_

a
b
e
a
0 0
_
=
_

a
b
dx

e
a

dx

0 0
_
.
7
This is the simplest example of a Cartan connection. Its meaning is this: for a function f, df
is the change in f under innitesimal displacement. But an innitesimal displacement
can at best be described by giving a tangent vector, therefore df is a 1-form acting on
tangent vectors to give real numbers. In a similar way, this connection we have constructed
measures how the Lie algebra relevant for the gauge theory changes under innitesimal
displacement.
We can now calculate the curvature for this connection, a Lie algebra valued 2-form:
T = d/+/ /
=
_
d
a
b
+
a
c

c
b
de
a
+
a
b
e
b
0 0
_
=
_
R
a
b
T
a
0 0
_
.
This makes sense: T
a
is a vector-valued 2-form, while R
a
b
takes value in so(3, 1). The
torsion T
a
is nothing but the curvature for translation.
We introduce a connection symbol D for this gauge theory:
De
a
= de
a
+
a
b
e
b
= T
a
, D
ab
= 0, D
abcd
= 0.
Note that this does not contradict our previous result e
a
=
a
b
e
b
: is a connection
on the tangent bundle whereas D is the connection for the gauge eld.
To summarise, we have introduced a 4-plane bundle (bres are isomorphic to R
4
as
vector spaces) E over our manifold M, with bre trivialization V
a
and equipped with
the bre metric
ab
. The principal bundle P with bres isomorphic to the Poincare group
acts on E in an ane manner. The frame elds e
a
provides a local unnatural isomorphism
between TM and E. The frame elds e
a
are called the soldering form, and in general
de
a
,= 0. Our point of view has changed: we started by saying that e
a
are 1-forms, so at
a point p they belong to T

p
M, but now they become members of Hom(T
p
M, R
4
). In a
similar fashion, e
a
(p) Hom(T

p
M, (R
4
)

).
2.4 Action principles
To derive the eld equations in general relativity using the action principle, after we have
written down the metric, we vary the metric and its rst derivative
S = S(g

, g
,
).
This is called the second order metric formalism. An equivalent procedure, the rst order
or Palatini procedure, is varying the metric and the connection independently
S = S(g

).
Translating into moving frame language, we have
S = S(e
a

, e
a
,
) second order,
S = S(e
a

a
b
) rst order.
8
For example, the second order action for gravity with cosmological constant is
S =
_
M

g d
4
x
R 2
16G
+. . .
where the dots represent boundary terms. A rst order action in terms of frame elds is
S =
1
2
2
_
M
[e[d
4
xe
a
e
b
R
ab
().
Let us vary this action. We need
e
1
= e
1
e e
1
, [e[ = e

a
e
a

= Tr(e
1
e)
then
S(e, ) =
1

2
_
[e[d
4
xG
a

e
a

+
1
2
2
_
[e[d
4
xe
a
e
b
R
ab
().
We see immediately from the rst integral (variation with respect to e
a

) the Einstein
equation in vacuum G
a

= 0. The second term, variation with respect to , should give

us relations between the connection and the metric (or the soldering form).
To proceed, we adopt the Poincare gauge theory point of view:
a
b
and R
a
b
take
values in
2
(E) = so(3, 1) and
ab
provides an isomorphism between E and E

. The
exterior product e
a
e
b
is a
2
-valued 2-form. We need to do some algebra. First,

abcd
A
ab
B
cd
= Tr(A B) = Tr(B A)
therefore

abcd
e
a
e
b
R
cd
= Tr(e e R)
Tr(e e e e) = 24[e[d
4
x
2[e[Rd
4
x = Tr(e e R).
To verify the second expression above, use e
a
e
b
e
c
e
d
= [e[d
4
x
abcd
and
abcd

abef
=
2(
c
e

d
f

c
f

d
e
) and further contractions. To verify the third expression above, use

abcd
e
a
e
b
R
cd
=
1
2

abcd
e
a
e
b
R
cd
ef
e
e
e
f
. Also note that, by expanding R,
R = d
a
b
+
a
c

c
b
+
a
c

c
b
= D.
Putting everything together, we have

_
Tr(e e R) =
_
Tr(e e R)
=
_
d Tr(e e ) 2
_
Tr(De e )
.
= 2
_
Tr(T e )
9
where in the last line we have thrown away a boundary term having no eect on the
equations of motion (the sign
.
= will be used to denote equality up to boundary terms).
Now dene a tensor 1 by
cd
= e
e
1
e
cd
, and
Tr(T e ) =
1
2

abcd
T
r
a
s
e
r
e
s
e
b
e
e
1
e
cd
=
1
2

abcd

rsbe
T
r
a
s
1
e
cd
[e[d
4
x
= (
r
a

s
c

e
d
+ cyclic permutations)T
r
a
s
1
e
cd
[e[d
4
x
= (T
a
a
c

e
d
+T
d
a
a

e
c
+T
c
e
d
)1
e
cd
[e[d
4
x.
The expression in the parentheses in the last line must vanish. After some further algebra,
this is equivalent to T
a
= 0, our equation of motion. Hence, if T
a
= 0 by assumption, then
is a total derivative in second order formalism and we are only left with the Einstein
equation. Therefore, no torsion can be present in vacuum.
Now we add to our action a matter piece S
m
(e, , ) where represents the matter
elds. Variation gives
S
m
=
_
[e[d
4
xT
a

e
a

+
_
[e[d
4
xS
c
e
d
1
e
cd
where T
a

is the canonical (unsymmetrized) energy momentum tensor: T

,= T

in
general, and S
c
e
d
denotes the spin current. Besides the Einstein equation coupling the
energy momentum tensor to the Einstein tensor G
a

=
2
T
a

, we also have the following

equation of motion:
T
a
a
c

e
d
+T
d
a
a

e
c
+T
c
e
d
= 2
2
S
c
e
d
.
So if matter lagrangian contains explicitly, spin is a source of torsion.
Usually, the spin current vanishes for bosonic elds. For example, the lagrangian for
scalar elds
L =
1
2

g g

has no spin current, neither have the Maxwell lagrangian

L =
1
4
F

, F

(i.e. F = dA).
However, in this case a little care is needed: if we carry out the procedure of minimal
coupling advertised in introductory general relativity courses , we must make
sure that we use the Levi-Civita connection uniquely determined by the metric. Otherwise,
torsion comes into play, and in general
F
wrong

,=

.
For a gauge transformation A

,
(

) = T

,= 0,
hence charge conservation is broken. In this case, we are just unnecessarily asking for
trouble, since the exterior derivative is perfectly well-dened on curved spacetime and is
coordinate-independent.
10
2.5 The 1.5 formalism
The 1.5 formalism is the following: for an action S, we have
S =
S
e
e +
S

rst order formalism;

S =
S
e
e +
S

e
e second order formalism.
However, we can think of = (e) as dened by S/ = 0, then the second term in the
second order formalism can be ignored and we are eectively 1.5.
11
3. Spinors and the Dirac equation
3.1 Cliord algebra and Majorana spinors
To describe fermionic elds we must rst dene spinors. We will label components of the
gamma matrices using the rst part of the Greek alphabet, e.g. , = 1, 2, 3, 4 (the other
part of the Greek alphabet is used for holonomic coordinate indices). Our signature for the
metric is (, +, +, +). Our representation of the Cliord algebra Cli(3, 1) is generated by
the following four real matrices

0
=
_
_
_
_
_
0 +1 0 0
1 0 0 0
0 0 0 1
0 0 +1 0
_
_
_
_
_
,
1
=
_
_
_
_
_
0 +1 0 0
+1 0 0 0
0 0 0 +1
0 0 +1 0
_
_
_
_
_
,

2
=
_
_
_
_
_
+1 0 0 0
0 1 0 0
0 0 +1 0
0 0 0 1
_
_
_
_
_
,
3
=
_
_
_
_
_
0 0 0 +1
0 0 1 0
0 1 0 0
+1 0 0 0
_
_
_
_
_
,
which also generates R
44
. Note

0
=
0
= (
0
)
t
,
i
=
i
= (
i
)
t
. (i = 1, 2, 3)
The gamma matrices relation is

a
,
b
= 2
ab
I.
A basis for the Cliord algebra is obtained by multiplying these matrices together. They
are (the right hand side gives the number independent matrices in each category)
I 1

a
4

ab
=
[a

b]
6

abc
=
[a

c]
4

abcd
=
[a

d]
=
5

abcd
1
where (note everything is real)

5
=
0

3
,
5
= (
5
)
t
, (
5
)
2
= I.
The matrices R
44
act on R
4
by the usual linear action. We will write M for R
4
, denoting
Majorana spinors. As a vector space, Mhas a symplectic formC

= C

, and C
a
C
1
=
(
a
)
t
. This is the representation of the charge conjugation operator exchanging particles
and antiparticles. In our representation, C =
0
. Here are some other useful identities
(C
a
)

= (C
a
)

, (C
ab
)

= (C
ab
)

,
(C
abc
)

= (C
abc
)

, (C
5
)

= (C
5
)

.
12
The Dirac equation is
(
a

a
+m) = 0.
The sign of m is irrelevant: if we apply the operator in the bracket (with the opposite sign)
again, we get the Klein-Gordon equation (+m
2
) = 0. Furthermore, if is a solution,
then
5
is also a solution of the conjugate equation
(
a

a
m)
5
= 0.
Lorentz transformation leaves the equation invariant by the following action on spinors:
let x
a

a
b
x
b
be a Lorentz transformation, and innitesimally
a
b
= exp(
a
b
), then
exp(
1
4

ab

ab
).
3.2 Dirac and Weyl spinors, complex structure
Once we have dened Majorana spinors as a real vector space on which the spin group
acts, Dirac spinors are easy: the complexication of Majorana spinors D = C
4
= M
R
C.
To get Weyl spinors, we claim that Dirac spinor is the direct sum of two Weyl spinors:
D = W

W. We make this decomposition concrete in the following way: take a Dirac
spinor , if
5
= i, an eigenstate with eigenvalue i, then W, whereas if
5
= i,
then

W. Decomposition of D then gives W= C
2
in some basis.
The action of the spin group Spin(3, 1) on a Weyl spinor is via its homomorphism to
SL(2, C): if we write the spacetime coordinates in a matrix
X =
_
t +z x +iy
x iy t z
_
= X

,
then we see that det(X) =
ab
x
a
x
b
. So if S SL(2, C), under the transformation
X SXS

,
det(X) is invariant.
Let us examine in more details the above constructions. Let V
R
be an even-dimensional
real vector space, and J End(V
R
) a linear operator satisfying J
2
= I, then J acts as a
complex structure on V
R
and make it into a complex vector space V
C
= V
R
C. We can
extend J to V
C
by complex linearity, and a basis can be chosen such that J is diagonal on
V
C
. This is just our construction of Dirac spinors from Majorana spinors, with J =
5
in
our basis. The Weyl spinors are then just the eigenspaces of J.
Let us see some examples other than spinors. The electromagnetic eld tensor F

is
completely determined by the electric and magnetic elds E and B in a frame, so F

R
6
.
We have the Hodge star operator that manifests the electromagnetic duality
E = B, B = E, ()
2
= 1,
so the Hodge star can be chosen as the complex structure. A Weyl spinor in this case
can be written as
M = E+iB W
13
and the Maxwell equations become
M = 0, M = i
M
t
.
The symmetry group acting on the usual Maxwell theory SO(3, 1)
R
then becomes SO(3)
C
,
and the group action leaves invariant
M M = E
2
B
2
+ 2iE B = F
2
+F F.
Another example is an even-dimensional manifold equipped with a metric g
ab
= g
(ab)
and a symplectic form
ab
=
[ab]
, both of which are covariantly constant under a connec-
tion g
ab
=
ab
= 0. Then we can form a complex structure
J
a
b
= g
ac

ab
, J
2
= I, J
a
b
= 0.
The complexied tangent space decomposes into Weyl sums T
C
M = WM WM, and
a tangent vector decomposes into the direct sum of a holomorphic tangent vector and an
antiholomorphic tangent vector. The manifold now has the structure of a Kahler manifold.
3.3 Coupling to gravity
Back to spinors, in supergravity it is convenient to work purely over the reals, and despite
the above constructions, consider a Dirac spinor as the direct sum of two Majorana spinors
D = MM instead of complexication. We write a Dirac spinor as
i
, i = 1, 2. Introduce
a complex structure on this space
J =
ij
=
_
0 1
1 0
_
,
ij

ij
=
ik
.
Then J induces R
4
R
4
= R
8
= C
4
= D. Note that
ij
is the generator of SO(2) = U(1),
and hence in this notation a U(1) gauge eld can be written as
A
ij

= A
ji

=
ij
A

.
The complexication of a Majorana spinor can be written as
j
A
ij
.
Now suppose a eld
A
transforms under a representation of Spin(3, 1). Let (
ab
)
A
B
=
(
ba
)
A
B
be a representation of this Lie algebra
[
ab
,
cd
] =
ab

cd

ac

bd

bd

ac
+
ad

bc
,
and the eld transforms as
exp(
1
2

ab

ab
)
A
B

B
= S().
For example, if
a
is a four-vector, then
(
ab
)
e
f
=
ae

b
f

be

a
f
14
reproduces standard action on vectors. But for a spinor, the representation is

ab
=
1
4
[
a
,
b
] =
1
2

[a

b]
=
1
2

ab
, [
ab
,
c
] = 2(
ac

bc

a
).
For coupling to gravity, we need a covariant derivative. We dene it for the general case:

+
1
2

ab

ab
, or = d +
1
2

ab

ab
.
Under the action exp(
1
2

ab

ab
), we have exp(
1
2

ab

ab
). We also need
(

) =
1
2
(R

ab

ab
) T

.
If we specialise to a Dirac spinor
a
, we have
= d +
1
4

ab

ab

and on background with no torsion,

(

) =
1
8
R
ab

ab
.
We will write the gamma matrices in coordinate basis as

= e

a
, and hence

= 0.
The Majorana and Dirac conjugate of a spinor are dened as

M
=
t
C =

,
D
=

and the matrices C and satises

C

C
1
=
t

1
=

, =

= (

.
In our basis, we can choose = C =
0
. Majorana spinors are exactly those that satisfy

M
=
D
, and in our basis, a Majorana is purely real.
Finally, we dene a symplectic linear product

1
C

2
=

2
=
_

1
for commuting spinors,
+

1
for anticommuting spinors.
Now we can write down the lagrangian in at spacetime for a Dirac spinor
L
1/2
=
1
2

(
/
+m).
The second term vanishes if are commuting. Also,

=
1
2

),
so if the spinors are commuting, L is a total derivative. This motivates thinking of

as
taking values in some large (strictly speaking innite dimensional) real Grassmann algebra
(. Bosons are even elements of (, while fermions are odd elements.
15
To calculate the variation, note
(

) =

= 2


and
(
1
2

/
) =

)
so variation gives
L
.
=

(
/
+m) = 0
therefore we obtain the Dirac equation
(
/
+m) = 0.
The spin current for a single spinor is
J

This vanishes for anticommuting objects. Therefore, to have a non-vanishing spin current,
we need two Majorana spinors or one Dirac spinor, in which case
J

j

ij
=

2
,= 0.
It is easy to generalise the construction to a purely bosonic background with action
S
1/2
=
1
2
_
[e[d
4
x

(

+m)
and variation with respect to gives the Dirac equation in curved background
(
/
+m) = 0.
If we iterate the Dirac operator and hope to get a generalisation of the Klein-Gordon
equation, we get
0 = (

m
2
)
= (
(

m
2
+
[

)
= (
2
m
2
+

]
)
= (
2
m
2
+
1
8

)
It may look like that we are getting some spin-curvature coupling in the last step, but this
is false: here is the Levi-Civita connection, and we have
_

2
m
2

1
4
R
_
= 0,
so there is no spin current coupled to the curvature. This calculation is rst done by Perez
and Lichnerowicz, is mostly easily veried by substitution: one needs the following formula

c
=
[abc]
+
ab

cd

b
+
cb

[abc]
=
a

[bc]

ba

c
+
bc

a
16
so
R
abcd

d
= 2R
ab

b
R
abcd

d
= 2R

abc
R
acef

ef
= 4G
e
f

f
and the result follows.
It should be noted that the massless Dirac equation in curved spacetime is conformally
invariant: if (g

, ) is a solution of
/
= 0, then (
2
g

, /
3/2
) is a solution also. We
can use this fact to deduce how R changes under Weyl rescaling.
Let us now investigate the equation under chiral rotations. A chiral rotation is
exp(
5
) = (cos +
5
sin)

exp(
5
) =

(cos +
5
sin).
Since
5
anticommutes with

, the kinetic term is invariant but the mass term is not:

m

exp(2
5
)m. Massless theory has chiral invariance. But consider a theory
where the mass term is replaced by
m m
1
+
5
m
2
where m
1
is called the Dirac mass, and m
2
the Majorana mass. Using chiral rotation we
can eliminate the m
2
term: the quantity
_
m
2
1
+m
2
2
is invariant.
In general, given n Majorana spinors
i
L =
1
2
n

i=1

i
/
+
1
2
n

i,j=1

i
M
ij

j
,
the mass matrix M
ij
= m
1
ij
+
5
m
2
ij
where both m
1
ij
and m
2
ij
are symmetric, a chiral
rotation of the form
exp(
ij
+
5

ij
) U(n),
ij
=
ji
,
ij
=
ji
can make the mass matrix M
ij
diagonal.
3.4 Einstein-Cartan-Weyl-Sciama-Kibble theory
In this theory, one adds torsion to the connection and attempt to couple matter elds
(fermions) to gravity. The lagrangian is
L =
R
2

2
+
1
2

/
.
Since

/
contains

ab
, spin density couples to torsion. The spin density is given by
S
a

b
=
1
8

ab

17
so that S
abc
= S
[abc]
, which implies T
abc
= T
[abc]
. The additional equation of motion is
T
abc
=

2
4

abc
.
In this case, we have a totally antisymmetric torsion. A theory with totally antisymmetric
torsion has the following characterisation. Autoparallels are dened by solution curves
to the following geodesic equation (the connection is not Levi-Civita, and hence the
quotation marks):
d
2
x

d
2
+

dx

d
dx

d
= 0,
while geodesics are dened by the extremal curves as dened by the length functional with
respect to the metric, i.e. solution curves to the same equation but with the Levi-Civita
connection. Now

+K

and
K

=
1
2
(T

+T

+T

) =
1
2
T

since T is totally antisymmetric. So K

(

)
= 0, and hence geodesics and autoparallels
coincide.
18
4. Supergravity lagrangian and super invariance
4.1 Rarita-Schwinger equation in at spacetime
We will write, as our rst attempt at a supergravity lagrangian (a lagrangian for spin-
3
2
elds), the following massless action for Majorana fermions:
S =
1
2
_
d
4
x

abc

c
.
Note that
abc

=
cba

. Variation gives
S
.
=
_
d
4
x

abc

c
and the equation of motion is

abc

c
= 0.
This equation is invariant under chiral rotations, and also fermionic gauge transformations

a

a
+
a
, analogous to the bosoinc gauge transformations for spin-1 elds.
We can simplify the equation further. Start with

abc
= 2
bc
,
a

ab
= 3
b
,
ab
=
a

ab
and hitting the equation with
a
, we get
/

a
/ = 0.
We now attempt to answer the Cauchy question: nd how many (real) functions are
needed to give the Cauchy data to this equation so that it has a unique solution. We rst
choose a gauge, i.e. use gauge invariance to set

i
= 0, t, i = 1, 2, 3.
This amounts to the following. Under a gauge transformation,
i

i

i

i
+
/
, so if we
solve the highly non-local equation =
1
/
(3)
(
i

i
), we can set what we require to zero.
Now write spatial and time part of the equation of motion separately
(
0

0
+ )

(
0

0
+
i

i
) = 0.
The = 0 equation of motion then gives
( )
0
= 0
iteration gives
2

0
= 0, so
0
= 0. Therefore,
i

i
= 0, and we are left with the equation

i
= 0. We conclude that, the equation of motion in this gauge is just
/

i
= 0.
Now
i
, having a spatial vector index, includes 3 4 = 12 functions. The constraints
are
i

i
= 0 and
i

i
= 0, each gives four constraints on functions (these are matrix
19
equations). So we are left with 4 free functions. Four free functions gives two degrees of
freedom for Majorana fermions.
What we have done is similar to the Coulumb gauge in electromagnetism: i.e. we set
A
0
= 0, A = 0
using the gauge transformation A
i
A
i
+
i
.
We can also use a covariant gauge, the Lorenz gauge: set
a

a
= 0, then
a

a
=
0 and the equation of motion is
/

a
= 0. This is analogous to setting

= 0 in
electromagnetism.
In quantum eld theory we complexify
a
and take it to be proportional to the plane
wave solutions e
ikx
. Then we have
k
2
= 0, k
a

a
= 0,
a

a
= 0.
Therefore
a
lies in a null plane with null normal k
a
. We still have freedom to add to
a
any multiples of k
a
. So we can set
k
a
= (1, 1, 0, 0),
a
= (0, 0,
2
,
3
).
So we have as many degrees of freedom as two vectors and two spinors have, minus the
following constraint:

3
+
3

2
= 0.
Hence, the spin content of
i
is the following:
i
transforms as spin 1
1
2
=
3
2

1
2
under
SU(2), but the spin-
1
2
part is killed by the constraints. So the eld
i
is a spin-
3
2
eld.
To add a mass term to our gravitino, we also look for analogies with electromagnetism.
The massive analogue of the Maxwell equation is the Proca equation

= m
2
A

from the following lagrangian

L =
1
4
F

1
2
m
2
A

.
This includes the Klein-Gordon equation: if we hit both sides with

we obtain m
2

=
0, back substitution gives
2
A

= m
2
A

.
Therefore, for gravitino, we use the following lagrangian
L =
1
2

abc

c
+
1
2
m

ab

b
.
Note that
ab

=
[ab]

=
ab
()
. The equation of motion is

abc

c
+m
ab

b
= 0.
Hit with
a
, we get
m
ab

b
= 0.
20
But
ab
=
a

ab
, so
=
/
/ .
Also
a

ab
= 3
b
, which implies 3m
a

a
= 0, or / = 0, so = 0. Now

abc
=
a

bc
+
ac

ab

c
,
expanding our original equation, we get
/

a
/ +m
a
/ m
a
= 0
therefore our nal equation of motion is
(
/
m)
a
= 0 subject to = / = 0.
To analyse further, we again complexify and take
a
e
ikx
, k
a
k
a
= m
2
. Set
k
a
= (m, 0, 0, 0),
a
= (0,
1
,
2
,
3
),
then

1
+
2

2
+
3

3
= 0.
As a representation of SU(2) we have vector together with spinor again, but this time it
is 1
1
2

3
2

1
2
, with the last
1
2
eliminated by the residual gauge freedom. So we have
2
3
2
+ 1 = 4 degrees of freedom in the massive case.
To couple our theory to gravity, we could try the minimal coupling , but it
doesnt work. This procedure actually only works for spin 1 and below. For example, for
spin 2 in at spacetime, we have

2
h

= 0, h

= 0,

= 0,
but letting gives too many conditions on h

.
4.2 ^ = 1 supergravity
We want to generalise the Rarita-Schwinger equation to curved spacetime. Again, we start
with the massless case. Our action is
S
3/2
=
_
d
4
x

g
1
2

a
(
abc
D
b

c
), D

+
1
4

ab

ab

note that we have omitted the Levi-Civita term

in D

, since it vanishes when

we anti-symmetrise with respect to , . We consider =

dx

as a spinor-valued 1-form.
The equation of motion is

abc
D
b

c
= 0.
What about gauge invariance for this theory? Let us start with a pure gauge spinor
= D = d +
1
2

ab

ab

21
then

abc
D
b
D
c
=
abc
D
[b
D
c]
= 0
so

abc
R
bcef

ef
= 0, or G
b
f

f
= 0.
In general det(G
b
f

f
) ,= 0 and det(a
a

a
) = (a
a
a
a
)
2
, so this term do not vanish. Therefore,
spin-
3
2
eld in curved background with G
ab
,= 0 is inconsistent! This is a generic problem
with spin greater than 1, known as the Buchdahl condition.
One method of overcoming this problem is by cancelling this term using the Einstein
action. Let a symmetry transformation be

=
1

where is a spinor eld with dimension [] = L

1/2
. Remember [] = L
3/2
. Then the
variation of our gravitino action gives
S
3/2
=
1

_
d
4
x

abc
D
b
D
c

.
=
1

_
d
4
x

g

G
a
b

b
.
For the Einstein action
S
2
=
_
d
4
x

g
R
2
2
,
we have
S
2
.
=
1

2
_
d
4
x

g G

b
e
b

.
Hence, by choosing
1
2
=
1
2
, i.e. use the specic symmetry variation
e
a

a
, [e] = 0, [] = L
we can show that

__
d
4
x

g
R
2
2
+
1
2

abc
D
b

c
_
= 0.
This is to the lowest order (ignoring 4-Fermi terms) a proof of the invariance of ^ = 1
supergravity lagrangian under supersymmetry variations (and we see why this lagrangian
is called the supergravity lagrangian).
For a proof of the invariance in the exact theory, in second order formalism, we write
the spin connection as

ab
=

ab
(e) +
1
2

2
(

a
+

b
)
where

denotes the Levi-Civita connection. The expression in bracket are the 4-Fermi
terms which needs to be cancelled using Fierz identities discussed later. Note that
T
abc
=
1
2

c
is in general not totally antisymmetric, and hence the Bianchi identities are modied
d + = R, de + e = T
22
which imply
d e de = dT, (R ) e = (T e), R e = DT.
An extension of this ^ = 1 theory is addition of the cosmological constant, rst
considered by Townsend
D

= D

+
1
2
a

, R R 2, =
3
a
2
< 0
where a is a constant.

D

is an example of a super-covariant derivative:

=
1

.
This change would appear to introduce a mass term into the lagrangian
1
2a

abc

c
=
1
2a

ac

c
but this is not really a mass, since the lagrangian is still invariant. We have maintained
gauge invariance at the expense of changing our ground state, i.e.
R

= 0 R

= g

and Minkowski spacetime is no longer a solution. The ground state is now AdS
4
=
SO(3, 2)/SO(3, 1). This is the homogeneous hypersurface dened in E
3,2
by the equation
(x
1
)
2
+ (x
2
)
2
+ (x
3
)
3
(x
4
)
4
(x
5
)
5
= a
2
homogenous meaning that the metric tensor completely determines the Riemann tensor
(no derivatives):
R

=
1
a
2
(g

).
In at spacetime, the covariant derivative commutes: D
[
D
]
= 0. In anti-de Sitter space
it does not, but the super covariant derivative does:

D
[

D
]
= 0.
This only works in anti-de Sitter spacetime, not in de Sitter spacetime, since in that case
an i has to be introduced, rendering the lagrangian complex.
4.3 ^ = 2 supergravity
In a ^ = 2 theory, to the lowest order, our elds are the metric, the gauge eld and two
Majorana spinors
g

, A

,
i

(i = 1, 2).
To the usual gauge lagrangian piece
L
1
=
1
4
F

, F

23
we add the Pauli term
L
Pauli
=

2

(F

+F

5
)
j

ij
, F

=
1
2

.
The supersymmetry variations of the various elds are
e
a

=
1
2

i
, A

=
1

2

i

ij
,
i

=
1

i
where in the last case the super-covariant derivative is

i
= D

i
+
1
2

ij
(F

+F

5
)
j
.
A feature of this theory is that the equation of motion (but not action) is chiral-duality
invariant, i.e. invariant under

exp(
5
)
i
, F

exp(e)F

.
4.4 Super invariance of the exact theory
For easier manipulation in showing the super invariance of the exact theory, we write our
theory using spinor-valued 1-forms =

dx

=
a
e
a
. Note that the wedge product on
spinors now commutes, since the wedge product and Grassmann variables both contribute
a minus sign. We have

1

2
=
2

1
,

= 0, e
a
= e
a
,

abc
= 0
while the combination

a
,= 0. Also recall that
abc
=
5

abcd
and
abcd
[e[d
4
x =
e
1
e
2
e
3
e
4
. The gravitino part of the lagrangian is
L
3/2
=
1
2

abc
D
b

c
[e[d
4
x
=
1
2

5
D
=
1
12

abc
D e
d

abcd
where =

dx

is a matrix-valued 1-form and

D = d +
1
4

ab

ab
.
The total lagrangian is
L =
1
4
2
Tr(e e R) +
1
2

5
D.
If we now do the variation, we get a term proportional to D
ab
for variation of R
ab
with
respect to the connection, and D De
s
for variation of

= D / and integration
by parts,

mnr

kl

mnrs
= 6

mkls
for variation respect to the connection in D
and nally
1
2

a
for variation of the metric e
a
and using the Fierz identities. Putting
everything together, we have
L
.
=
1
2
2
(
mn
e
r

1
6

mnr
D) (De
s

1
4

s
)
mnrs
.
The interpretation of this variation depends on the formalism used: in second order form-
alism,
mn
= 0 since by denition = (e, ); in the rst order formalism, this denes
the variation of
mn
since we really only have e and ; in the 1.5 formalism,
mn
vanishes as well because of the equation of motion coming for
mn
.
24
4.5 Fierz identities
Now we come to the Fierz identities that is referred to previously. Cli(3, 1) has a basis
I,

,
5

,
5
and we write them as
A
for A = 1, . . . , 16. We consider the matrices

as endomorphisms of the spinor vector space. We claim that, except for I, Tr

A
= 0.
This can be shown by, e.g.

5
=

, Tr(
5

5
) = Tr(

) = Tr(

) = 0
due to cyclicity of trace. For those involving
5
, we use
5
=
1

1
. Similarly, we obtain
relations for Tr
A

B
. Putting them together,
Tr
A
= 0
A
,= I, Tr
A

B
= 4
AB

AB
= diag(1).
Let us determine the signs of
AB
: we write (m, n) for m plus signatures and n minus
signatures for generators. We get
I (1, 0)

(3, 1)

(3, 3)

(3, 1)

5
(0, 1)
So adding everything together, we have Cli(3, 1) = R
10,6
.
Note that as a vector space (ignoring the product structure), Cli(3, 1) = (R
3,1
) has
a natural inner product induced from
ab
:
(F
p
, F

p
) =
1
p!
(F

1
...
p
F

1
...
p

1
. . .

p
).
Taking into consideration of the product structure, however, the counting goes
1 (1, 0)
e

(3, 1)
e

(3, 3)
e

(1, 3)
e

(0, 1)
The Fierz identities is used to simplify the following expression
(

1
M
2
)(

3
N
4
) =
1
M

3
N

4
.
Use the complete basis, we write
M

= C
A

multiply by
B

to obtain
C
A

=
1
4
M

25
therefore
M

=
1
4
(M
a
N)

=
1
4

(N
A
M)

.
Using this, for anticommuting spinors
(

1
M
2
)(

3
N
4
) =
1
4
(

1
M
A
N
4
)(

2
) =
1
4
(

4
)(

3
N
A
M
2
).
Therefore,
(

1
M
2
)(

3
N
4
) =
1
4
(

4
)(

3
NM
2
)

1
4
(

4
)(

3
N

M
2
)
+
1
4
(

4
)(

3
N
5

A
M
2
)

1
4
(

4
)(

3
N
5

M
2
)
+
1
8
(

4
)(

NM
2
).
The signs are opposite if the spinors are commuting. Therefore

2
=
1
4
(

1
)I +
1
4
(

1
)
5

1
4
(

1
)

+
1
4
(

1
)
5

+
1
8
(

1
)

so that

1
=
1
2
(

2
)

1
4
(

1
)

.
The Fierz identities are useful in establishing the commutator of 2 supersymmetry trans-
formations and thus checking the algebra.
The following identities

= 4,

= 3

= 0,

= 3
5

= 4,
allow we to simplify expressions and we can obtain interesting consequences, including the
fact that

is lightlike and future-pointing for commuting Majorana spinors:

(

)(

) = 0,
t

0
> 0,
also

= 0,
by setting M =

, N =

and
1
=
2
=
3
=
4
= .
4.6 Supersymmetric background
In at background, schematically a supersymmetry variation gives

B F,

F (

B), [

1
,

2
]
1
2

1

where B is bosonic and F is fermionic. The dimensions are

[F] = L
3/2
, [B] = L
1
, [] = L
1/2
.
26
A purely bosonic background has F = 0, is a solution of the equation of motion, and is
invariant under any supersymmetry transformations satisfying
(

B) = 0.
(A background is called supersymmetric or BPS if it is invariant under action of at least
one supersymmetry. This gives extra constraint on B, e.g. B = constant.)
Let us give a concrete example: the Wess-Zumino model. In at spacetime, consider
free and massless elds
L =
1
2
(A)
2

1
2
(B)
2
+
1
2

/
.
Variation gives
L =
2
AA+
2
BB +

AA+

BB
1
2

)
where
A =
1
2
, B =
1
2

5
, =
1
2

(A+
5
B)).
So the equations of motion are

2
(A+
5
B) = 0,
/
= 0.
And once the equation of motion is satised, we see that L =

, where the conserved

supercurrent J

is
J

=
1
4
(

(A+
5
B)).
We can calculate the commutator of two supersymmetry transformations in this theory:
[

1
,

2
] =
1
2

1

.
This is a translation, and this algebra closes on shell, i.e. we need to impose the equations
of motions to get this.
Let us try to solve the equations of motions for bosonic background. For the wave-like
equation, we write
A+
5
B = exp(
5
k x)(A
0
+B
0

5
).
Substituting into the equation of motion, we see that the wave vector is lightlike
k

= 0.
In order to have bosonic solutions, the variation = 0, so we must also have
(k

) = 0.
This is an equation of constraint on . As det(k

) = k
2
= 0, k

has a non-trivial
kernel. Now let us go into the frame where
k

= (1, 1, 0, 0)
27
we have
(
0
+
1
) = 0
which, using (
0

1
)
2
= I and Tr(
0

1
) = 0, we get

1
= .
Now we can introduce
P
+
=
1
2
(1 +
0

1
), P

=
1
2
(1
0

1
)
which are projection operators
P
2
+
= P
+
, P
2

= P

, P
+
P

= P

P
+
= 0, P
+
+P

= I.
These projection operators can be used to split the Majorana spinors into direct sums:
M = M
+
M

where

if and only if P

and any spinor decomposes

uniquely as =
+
+

. Also note
0

+
=
+
and
0

. In terms of these
operators, our Killing spinor belongs to the minus part of the space: M

. Now write
= a

with a anticommuting and

commuting, we have

= (
t

,
t

,
t

,
t

)
= (
t

,
t

,
t

,
t

)
=
t
(1, 1, 0, 0) k

.
To get the last equality, we need to do some gamma matrices manipulations:

=
t

(
0

1
)

=
t

and use

=
t

= 0.
In a curved spacetime, the supersymmetry variation on the spinor eld is

=
1

where we have used the super covariant derivative. From the previous analysis we see that
there is no harm in assuming to be a commuting (and possibly Dirac) spinor. We can
obtain a vector by setting k

. Typically, k

is covariantly constant

= 0
and Killing
(

)
= 0. Moreover, k
0
=

(remember

=

0
) in a local frame,
hence k
0
> 0 in all Lorentz frames: it is future directed and causal. If is actually
Majorana, then it is lightlike. The probability current for the Dirac equation in this theory
is J

. Note that in such theories we typically get a supply of covariantly constant

tensor elds like

, etc.
In the simplest case of such a theory ^ = 1, = 0, we have

D = , the Levi-Civita
connection. In this case we would have

= 0, i.e. is a covariantly constant Killing

spinor eld. The k

is covariantly constant and null, also

) = 0
where the expression in bracket is a covariantly constant 2-form.
28
From the result in general relativity, if we use lightcone coordinates u = t x
1
and
v = t +x
1
, then in this case we can write the metric as
ds
2
= 2dudv +H(u, x
i
)du
2
and then

v
= k

is a covariantly constant null Killing vector eld. If the spacetime we are considering is
Ricci at, then
i

i
H = 0 for arbitrary u. An example in this case would be
H = (u)((x
1
)
2
(x
2
)
2
) + 2(u)x
2
x
3
for arbitrary (u) and (u). Observe that the and represent polarisations (+) and ()
of a quadruple moment, i.e. describing the behaviour of a classical graviton. Therefore,
that classical gravitons are invariant under 2 of the 4 supersymmetry transformations.
(For ^ = 1, = 0, the vacuum ground state E
3,1
is invariant under all 4 supersymmetry
transformations.)
4.7 Super Poincare group and gauge theory
We now attempt to make the supersymmetry that we have used in previous discussions
systematic. We will do this by introducing a superspace and a super group acting on it,
and nally dene a gauge theory analogous to the general relativity case. The action of
super Poincare group on a superspace is
_
_
_

a
b

1
4

a
a
a
0 S()

0 0 1
_
_
_
_
_
_
x
b

1
_
_
_
where the action on spinors is S() = exp(
1
4

ab

b
) for
a
b
= exp
a
b
. The dimensions
of various quantities are
[] = L
1/2
, [] = L
1/2
, [a] = L.
The superspace (x
b
,

) is identied with E
3,1
M. We have the Grassmann variables
and here, so we need to be careful with signs. As matrix multiplication, this is a
left action, and we can interpret this group action as a semidirect product of Lorentz
transformations with translations and supertranslations, where supertranslations are
+
with the associated translation
x

1
4

.
Translations and supertranslations together form an invariant subgroup of the whole super
Poincare group. The superspace is then the quotient manifold of the super Poincare group
by the Lorentz group in the usual way.
29
We can consider functions on this superspace f(x, ). When we do Taylor expansion,
the series in terminates at nite power: to be exact, at
4
due to the Grassmann nature
of . Consider

f = x

f
x

=
1
4

f +

f
where we have dened
R

+
1
4

.
We need to be careful about position of

in the variation since it is anticommuting.

Actually,

behaves like an inner multiplication. For R

, we think of as right invariant

vector elds generating left translations on superspace and thus commuting with generators
of right translations
L

, R

= 0, L

+
1
4

the analogy with right and left invariant vector elds on a Lie group is exact (note that
left invariant vector elds generate right translations, etc.). To make this analogy more
concrete, let us introduce left invariant 1-forms

= d

= dx

+
1
4

,
and right invariant 1-forms

= d

= dx

1
4

.
Let us recall the following Maurer-Cartan relations from the theory of Lie groups and
Lie algebras, where C
A
B
C
are the structural constants and A, B, . . . = 1, 2, . . . dimg:
d
C
= C
A
C
B

A

B
d
C
= +C
A
C
B

A

B
[L
A
, L
B
] = +C
A
C
B
L
C
[R
A
, R
B
] = C
A
C
B
R
C
[L
A
, R
B
] = 0

A
B
=
A
(R
B
) =
A
(L
B
).
The Maurer-Cartan form for the Lie group can be written g
1
dg =
A
T
A
where T
A
are
the basis for the Lie algebra. Also, dg g
1
=
A
T
A
. In quantum theories, one usually
denes the generators after division by i so they can act as quantum mechanical hermitian
operators.
Now let us write down similar formulae for our super theory:
d

= +
1
4
d

=
1
4
d

, R

= +
1
2

, L

=
1
2

,
30
and now the structural constants are
C

= C

=
1
2

.
Note that for anticommuting objects, d

= +d

. In the literature, the

usual practice is to call our R

by Q

, the generators and our L

by D

, the covariant
derivatives. The anti-commutation relation then reads
Q

, Q

=
1
2

.
Let us also use this notation from now on.
Consider Q(
1
) =

1
T

where T is an R-valued (anti-hermitian) matrix. To write out

this matrix representation in full,
Q(
1
) =
_
_
_
0
1
4

1
0
0 0
1
0 0 0
_
_
_
, T

=
_
_
_
0
1
4

0
0 0

0 0 0
_
_
_
, T

=
_
_
_
0 0

0 0 0
0 0 0
_
_
_
,
with
T

+T

=
1
2

(4.1)
Then one can check
Q(
1
)Q(
2
) Q(
2
)Q(
1
) =

1
T

2
T

2
T

1
T

2
T

, T

.
A supersymmetry transformation acting on a supereld then can be written as
f =

f =

f.
Again, quantum mechanically, we would write
T

i
1/2
, T

i
,

P

i
,
and then

+

Q

=
1
2
(C

)

P

.
which is
1
2

P
0

in the rest frame. Since the left hand side is non-negative, The total
expectation value is then

) 0. Taking the trace, we get

P
0
) > 0, and hence

)
is future directed timelike or null. This is the positive energy property of supersymmetric
theories.
We can now formulate a super Poincare gauge theory, similar to the Poincare case.
The connection 1-form is
/ =
_
_
_

a
b

1
4

a
e
a
0
1
4

ab

ab

0 0 0
_
_
_
where
=

dx

, [] = L
1
.
31
The curvature is
T =
_
_
_
d + de
a
+
a
b
e
b

1
4

0
1
4

ab
(d + ) (d +
1
4

ef

f
)
0 0 0
_
_
_
.
We see that now (d +
1
4

ef

f
) is the curvature associated with supertranslations,
while de
a
+
a
b
e
b

1
4

a
is the curvature associated with translations. We want no
curvature for translations, so setting the appropriate term to zero, we get
de
a
+
a
b
e
b
=
1
4

a
= T
a
fermionic contribution to torsion.
The super Poincare group is relevant for Minkowski spacetime. For AdS
4
, because
Spin(3, 2) = Sp(4, R), the supergroup relevant in this case is the orthosymplectic group
Osp(1[4), i.e. those matrices leaving invariant the quadratic form

ij
x
i
x
j
+C

where C

= C

is the symplectic form. We can pass from Osp(1[4) to the super Poincare
group by a process called Wigner-Inon u contraction, which we will outline below.
The orthosymplectic group Osp(M[N) has a diagonal subgroup SO(M) Sp(N, R),
i.e. matrices of the form
_
SO(M) 0
0 Sp(N, R)
_
where elements of the symplectic group Sp(N, R) preserves the symplectic form C
ab
. The
dimensions are
dim
R
SO(M) =
1
2
M(M 1), dim
R
Sp(N, R) =
1
2
N(N + 1).
The action of Osp(M[N) on superspace is
_

i
j

i
b

a
j

a
b
__
x
j

b
_
where

i
j
=
i
j
+
i
j
+. . .
ij
=
ik

k
j
,
ij
=
ji
,

a
b
=
a
b
+
a
b
+. . .
ab
= C
ac

c
b
,
ab
= +
ba
.
For the isometry group of AdS
4
, Sp(4, R) = Spin(3, 2), we have the generators

A
= (

,
5
), A = 0, 1, 2, 3, 4
and

B
+
B

A
= 2
AB
,
AB
= diag(1, 1, 1, 1, 1).
32
For a real (i.e. Majorana) representation of Cli(4, 1),
C
5
=
0

5
, C
5
= C
t
5
we have the basis which generate Sp(4, R)
C
5
1 skew,
C
5

A
5 skew,
C
5

AB
10 symmetric.
Pass into the supersymmetric case, Osp(1[4) is the super anti-de Sitter group for ^ = 1,
and Osp(^[4) is the ^-extended super anti-de Sitter group.
Now consider the generators of SO(3, 2), which we will call M
AB
. In these, the gener-
ators M

generate Lorentz rotations, while M

4,
generators non-commuting translations:
_
M
4
a
,
M
4
a
_

a
where a is a length scale set by the radius of the anti de Sitter space. Note the dimension
[M

] = 1, [P

] = L
so the above can be written as
[P

, P

] a
2
M

.
To proceed with the Wigner-Inon u contraction, we take the limit a . Then in the
limit [P

, P

] = 0, i.e. Minkowski spacetime. Recall that the anti de Sitter space is dened
by
x
A
x
B

AB
= a
2
,
so the limit a corresponds to setting the radius of curvature to innity, and hence
=
3
a
2
0.
If we do a similar analysis for the de Sitter spacetime, the algebra is anti-unitary
instead of unitary.
33
5. Wittens proof of the positive energy theorem
In general relativity, the positive energy theorem states that, assuming the dominant energy
condition, the mass of an asymptotically at spacetime is non-negative; furthermore, the
mass is zero only for Minkowski spacetime. This theorem is important not only in physics,
but in pure mathematics as well, where it is more commonly known as the positive mass
theorem. By far the simplest and most elegant proof is given by Edward Witten in 1981,
which is a very good illustration of the power of using spinors and supergravity in doing
essentially classical calculations. We will briey outline Wittens proof in this section.
We begin by dening the Nestor 2-form
N

where is a commuting Dirac spinor. To manipulate this, we will use Stokes theorem.
Let

= J

, and suppose is a spacelike hypersurface in spacetime, e.g. a Cauchy

surface. For a domain D with boundary D, Stokes theorem states that
1
2
_
D
=

=
_
D
J

Now, take increasing domains D such that limD = , limD = S

2

. Also assume space-

time (M, g
ab
) is asymptotically at. Then we have

+
1
8
R

+
1
2

+
1
2

.
To proceed, we assume the dominant energy condition: T

0. Consider this
equation for all future directed timelike vectors V

, W

, we see that it implies T

00
0.
An equivalent statement is that T

0
[T

[ in all Lorentz frames, where hatted indices
denote local pseudo orthonormal frames. Hence, T

is past directed and timelike for

any future directed timelike W

. Note that in our sign convention, T

0
0
< 0. Putting all
these together, we see that
1
2

is past directed and timelike.

Now the second term in the expansion of

has been taken care of, let us come to

the rst term. We will use a local pseudo-orthonormal frame e
i
adapted to the hypersurface
, with e
0
orthonormal to . The zeroth component of the rst term in the expansion
reads
(
i
)

ij

j
= (
i
)

(
i

ij
)
j

= (
i

i
)
2
()
2
where

i
= (
i
+
1
4

i rs

s
) +
1
2

i rs

s
.
34
In this expression for the connection, we recognise the rst term as the Levi-Civita con-
nection, whereas the second term can be thought of the second fundamental form of .
Now we impose the following conditions: rst, we require
i

i
= 0 on . This is
called the Witten equation. It is dierent from the Dirac equation since the Dirac one
uses the Levi-Civita connection induced by the metric. Second, we require the boundary
condition =
0
, a constant spinor at spatial innity. These conditions together with the
above calculation implies J

is past directed and timelike.

For further manipulation of the second fundamental form introduced above, we will
work in a Gaussian normal coordinate. Locally, we write the metric as
ds
2
= dt
2
+g
ij
(x, t)dx
i
dx
j
so, to connect with our frame,
e
0
= dt, de
0
= 0 =
0
i
e
i
, de
i
=
e
i
t
dt +d
x
e
i
=
i
0
dt
i
k
e
k
.
We see the emergence of the Levi-Civita connection with respect to g
ij
in the last expression
above.
We write
0
j
= e
i
K
ij
, then
0
j
= K
ji
e
i
. Also note that de
0
= 0 is equivalent to
K
ij
= K
ji
. Then,
e
i
t
= K
ij
e
j
,

t
(e
i

S
e
j
) = 2K
ij
e
i

S
e
j
so in our coordinate system,
K
ij
=
1
2
g
ij
t
=
1
2

n
g
ij
.
The constraint equations that needs to be required in this coordinate are
(3)
R = K
ij
K
ij
K
i
i
K
j
j
, (K
ij
g
ij
K
k
k
)
;l
= 0.
Covariantly, this is
K

= (

where n is the unit normal to the hypersurface, n

= 1, and is the projection onto

using h

+n

.
We now take a detour and outline several applications using this second fundamental
form approach. The rst is that we can see for the Witten equation

i
=
L.C.
i
+
1
2
K
ij

j
,
multiplication by
i
gives
i

i
= 0. The second is for the Einstein-Hilbert action
S =
1
2
2
_
D
R[e[d
4
x + 2
_
D
(Tr K
ij
)

g d
3
x
variation contains only g terms on boundary and no terms form (g
ij
/n). The third
application is for the necessary condition for the existence of Killing spinors in anti-de
Sitter spacetime. Let A = 0, 1, 2, 3, 4 label the coordinates in E
3,2
, on which the anti-de
35
Sitter spacetime is dened using the usual quadric. The super covariant derivatives on AdS
commutes: [

D

,

D

] = 0, and the existence of Killing spinors is the existence of satisfying

= 0. This can be constructed as follows. Let now denotes a constant spinor in E

3,2
,
so
A
= 0. Then it is easy to see

A
= 0,

= 0
and in this case, K

where g

is the induced metric.

Let us return to Wittens proof of the positive mass theorem. We now need to solve the
equation
i

i
= 0 subject to
0
at spatial innity. Since this equation is an elliptical
equation, existence of solution follows from standard results from dierential equations. For
uniqueness, suppose
1
and
2
are both solutions satisfying the same boundary conditions.
Then =
1

2
satises the formula and the boundary term vanishes. Since

is (subject to the Witten condition) past directed and timelike, the volume term has a xed
sign. The boundary term, however, vanishes, since it only depends on
0
. Hence
1
=
0
.
The last step in the proof is to note that
2

2
_
S
2

1
2
N

=
0

0
P
ADM

0 (5.1)
where P
ADM

is the ADM 4-momentum of spacetime. Hence P

ADM

is future directed and

timelike, if (M, g

) is asymptotically at.
We can identify the boundary term with the ADM mass in this theory. First we analyse
in more details the ADM mass/momentum. The metric at innity is:
ds
2

_
1
2MG
r
_
dt
2
+
dr
2
1
2MG
r
+r
2
(d
2
+ sin
2
d
2
),
where M is the ADM mass. We can write this in isotropic coordinates
ds
2
=
(1
GM
2
)
2
(1 +
GM
2
)
2
dt
2
+
_
1 +
GM
2
_
4
d
2
x.
This has a momentum P

= (M, 0, 0, 0). We can boost it such that P

i
,= 0. This will
introduce cross terms in the metric.
We can solve the Witten equation in this metric and evaluate the boundary term. The
metric is
g

+O(r
1
)
so the connection
= O(r
2
)
with a factor of M. For =
0
+. . ., we get
D

=
0

0
M
r
2
+. . . .
36
For the integrand in (5.1), it is of order O(Mr
2
), and d

is of order O(r
2
). The
boundary term therefore is of order M(
0

0
). For its coecient, we can either work it
out (it is complicated), or we can look at the identity

0

0
P
ADM

=
_
T

dV +
2

2
(. . .)dV.
Evaluating near at spacetime, we get the integrand as
_
(T

0
)d
3
x +. . .
Now
0
is a constant spinor in Minkowski spacetime, and hence
0

0
is a Killing vector
of the background. The whole expression is like a total energy. By the linearised Einstein
equation this is the total mass M of the linearised theory. However, the boundary term
depends only on the asymptotic metric, not on the interior. Thus we can always identify
the boundary term with the ADM mass.
We now consider the case where P
ADM

= 0, i.e. the case of equality in equation (5.1).

Then
i
[

= 0, and some components of T

= 0. Consider another Cauchy surface

, with the same boundary at innity

= S
2

(i.e. a nite variation of ), we also

have
i
[

= 0 and T

= 0. Thus is actually covariantly constant,

= 0, and
the dominant energy condition becomes T

= 0. Therefore, k

is a covariantly
constant Killing eld, and R

= 0 (the Einstein equation). Now, if k

is null, and then

we have gravitational waves, the spacetime is not asymptotically at, and hence this case
is not allowed. If k

is timelike, then it is hypersurface orthogonal by

= 0 and has
constant norm g

. Then the metric is ultrastatic:

ds
2
= dt
2
+g
ij
(x)dx

dx

and g
ij
must admit constant spinor
i
= 0. Then
R

= 0
(3)
R
ij
= 0
(3)
R
ijrs
= 0 g
ij
=
ij
,
i.e. the hypersurface is at. This remains true in higher dimensions if we quote the theorem
that asymptotically atness and Ricci atness together imply atness.
There are some global issues with this approach to the proof of the positive energy
theorem: the existence and uniqueness of spin structures. For the moving frames we
use, global frames are not essential but convenient. If a global frame exists there is no
diculty in introducing a spin structure. However, uniqueness is more problematic. If the
fundamental group is non-trivial, H
2
(M, Z
2
) ,= 0, then there exists more than one spin
structure. They can be odd or even as non-trivial closed curves. If no global frame exists
at all, there may not be a spin structure on the manifold. The obstruction in this case is
called the second Stiefel-Witney class W
2
H
2
(M, Z
2
), which vanishes if all 2-cycles are
topologically trivial.
In four spacetime dimensions, a globally hyperbolic spacetime always admit a global
framing: M = R and is spacelike. Also, every 3-manifold is parallelizable, and hence
37
M admits a global framing and at least one spin structure. If it is not simply connected,
it may admit more than one. For our argument of the positive mass theorem to work, we
need a spin structure which allows constant spinors at innity.
If black holes are present in this spacetime, then we can work on the exterior only and
use boundary condition on the horizon. Typically, = 0 on horizon.
The positive mass theorem has extensions: instead of asymptotic atness, we can
require asymptotic AdS
4
behaviour, e.g. the Kottler solution
ds
2
=
_
1
2GM
r

r
2
3
_
dt
2
+
dr
2
(. . .)
+r
2
d
2
.
Here M is called the Abbot-Deser mass. Using

D

as dened with a cosmological constant,

we can prove that M 0. The case M = 0 corresponds to AdS
4
spacetime. The solutions
of Witten equation
i

D
i
= 0 tend to Killing spinors of AdS
4
at innity.
38
6. Central charges and BPS states
BPS stands for Bogomolnyi, Prasad and Sommereld. BPS bounds refers to a series
of inequalities for solutions of eld equations depending only asymptotic behaviour the
solutions at innity (actually, only on the homotopy class of the solution). A BPS state is
a solution which saturates this bound. In supersymmetric theories, the BPS bound usually
is saturated when half of the SUSY generators are unbroken. This happens when the
mass is equal to the central extension, which can be interpreted as a topological conserved
charge. In this section we will investigate BPS states briey.
For ^ 1, the supersymmetry algebra admits a central extension:

Q
i

,

Q
j

=
1
2
(P

ij
+
1
2
C

X
ij
+
1
2
(C
5
)

Y
ij
, X
ij
= X
ji
, Y
ij
= Y
ji
,
where X and Y commutes with all elements of the algebra and can be thought of electric
and magnetic charges respectively. In fact, we can obtain this algebra by contraction of
Osp(^[4). Witten and Olive showed that such conserved central charges can arise as
boundary terms in supersymmetric eld theories.
For example, for ^ = 2, X
ij
= X
ij
, they showed that M

X
2
+Y
2
and the
equality case is invariant under half of the maximal supersymmetry transformations. In
the rest frame, the right hand side of the algebra relation is
M
ij
+ (X
0
+Y
0

5
)
ij
,
which is a 8 8 matrix. Passing into Dirac notation
ij

1 = i, it becomes a 4 4
hermitian matrix
A = M +iR, R = X
0
+Y
0

5
.
Then R
2
= (X
2
+Y
2
) implies
A
2
= M
2
+ 2iMR R
2
= 2MAM
2
+X
2
+Y
2
hence we have the characteristic equation
A
2
2MA+M
2
(X
2
+Y
2
) = 0.
The eigenvalues are obtained from the characteristic equation as

= M
_
X
2
+Y
2
.
Since Tr A = 4M, we see that

are both doubly degenerate. The BPS state has M =

X
2
+Y
2
and A has a kernel the dimension of which is half of the dimension of A,
i.e. half of the maximal supersymmetry. Another way to say this is that there exists 2
linear combinations of the supersymmetry generators which annihilate the BPS states.
Let us now focus on BPS states in ^ = 2 supergravity. The BPS states are invariant
under 4^/2 = 2^ supersymmetries. After diagonalisation,
A = diag(M +
_
X
2
+Y
2
, M +
_
X
2
+Y
2
, M +
_
X
2
Y
2
, M +
_
X
2
Y
2
).
39
We implement it as follows. The bosonic part of the supergravity lagrangian is the Einstein-
Maxwell theory. Dene
L
B
=
R
4M
2

1
2
F

,
then
T

= (F

1
4
g

)
satises the dominant energy condition. We have

+
i
2

2
(F

.
We can now identify boundary terms. N

contains F

, and
X =
Q

4G
, Y =
P

4G
where Q and P are the electric and magnetic charges. Since the electromagnetic potential
is asymptotically
A

dx

Q
4r
+
P cos
4
d,
the rst term can be thought of as the Coulomb potential and the second the Dirac mono-
pole. Because we are ungauged since the charges are central, neither Q or P is quantised.
The BPS bound on the mass is
M
_
2Q
2

2
+
2P
2

2
and for BPS states, the equality holds.
We can compare this with the Reissner-Nordstrom black hole:
ds
2
=
_
1
2M
r
+
Q
2
+P
2
4r
2
_
dt
2
+
dr
2
(. . .)
+r
2
d
2
in which case we have
M
_
Q
2
+P
2
4G
.
Then
r

G
= M
_
M
2

(Q
2
+P
2
)
4G
.
The extremal case is that for r
+
= r

. If we draw Penrose diagrams for the black hole case

and the supergravity case, we see that they complement each other.
The Hawking temperature for this black hole can be calculated
T
H
=
1
2G
_
M
2

1
4G
(P
2
+Q
2
)
_
M +
_
M
2

1
4G
(P
2
+Q
2
)
_
2
the temperature T 0 as M
_
1
4G
(P
2
+Q
2
). For non-extremal cases, k =

t
becomes
spacelike for r

< r < r
+
, so it cannot be a BPS state. Also note that the Hawking temper-
ature is nite, so the theory can never be supersymmetric, since the Fermi-Dirac statistics
40
is dierent from the Bose-Einstein statistics at all nite temperatures. The supersymmetric
generators cannot even act in such a case.
The extremal black hole case is compatible with the BPS condition, for which
T
H
=
_
M
2

1
4G
(P
2
+Q
2
)
Gr
2
+
.
Let us calculate the Killing spinors.

D

= 0 implies the Dirac equation

= 0, hence

= 0.
Let us solve the Dirac equation. If
ds
2
=
2
g

dx

dx

= g

dx

dx

.
Then
/
D =
(n+1)/2
/
D

where =
(n1)/2

. Introduce isotropic coordinates A =
A(x) and B = B(x), we have
ds
2
= A
2
dt
2
+B
2
dx
2
= A
2
_
dt
2
+
B
2
A
2
dx
2
_
,
an ultra static metric. Then
(
0

0
+
i

i
)

= 0.
Assume
t

= 0 on the 3-dimensional surface, we have

i

= 0.
Then

= 0,
for the at Dirac operator on E
3
. Hence
=
1

AB
1

.
Let us try the solution

=
0
a constant. Then

0
.
Now choose
0

0
= i
0
, then

0
,
however, the factor involved here is not a constant.
In general, the solution is not regular near horizon at which A = 0. The exceptional
case occurs when AB = 1. There if
ds
2
= H
2
dt
2
+H
2
dx
2
, H = 1 +
k

a=1
GM
a
[x x
0
[
,
we have the Majumdar-Papeptrou solution. For k = 1, we have the extremal Reissner-
Nordstrom solution again, and

=

t
.
41