(Progress in Mathematical Physics 59) Gerardo F. Torres Del Castillo (Auth.) - Spinors in Four-Dimensional Spaces-BirkhÃ User Basel (2010)

Progress in Mathematical Physics
Volume 59
Editors-in-Chief
Anne Boutet de Monvel, Université Paris VII Denis Diderot
Gerald Kaiser, Center for Signals and Waves, Austin, TX
Editorial Board
Sir M. Berry, University of Bristol
C. Berenstein, University of Maryland, College Park
P. Blanchard, Universität Bielefeld
M. Eastwood, University of Adelaide
A.S. Fokas, Imperial College of Science, Technology and Medicine
C. Tracy, University of California, Davis
Gerardo F. Torres del Castillo
Spinors in
Four-Dimensional Spaces
Birkhäuser
Gerardo F. Torres del Castillo
Instituto de Ciencias
Universidad Autónoma de Puebla
Ciudad Universitaria
72570 Puebla, Puebla, México
[email protected]
ISBN 978-0-8176-4983-8 e-ISBN 978-0-8176-4984-5

DOI 10.1007/978-0-8176-4984-5
Springer New York Dordrecht Heidelberg London
Library of Congress Control Number: 2010930605
Mathematics Subject Classification (2010): 15A66, 15B10, 22E43, 53B30, 81Q05, 81R20, 83C50,
83C60
c Springer Science+Business Media, LLC 2010

All rights reserved. This work may not be translated or copied in whole or in part without the written
permission of the publisher (Springer Science+Business Media, LLC, 233 Spring Street, New York,
NY 10013, USA), except for brief excerpts in connection with reviews or scholarly analysis. Use in
connection with any form of information storage and retrieval, electronic adaptation, computer software,
or by similar or dissimilar methodology now known or hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are
not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject
to proprietary rights.
Printed on acid-free paper
www.birkhauser-science.com
Preface
The aim of this book is to present in an elementary manner the spinor formalism
applicable to four-dimensional spaces of any signature, without assuming previous
knowledge about spinors, nor an advanced knowledge about Lie groups. The ap-
proach followed here is not based on the Clifford algebras and does not make use of
the language of fiber bundles.
This book should provide the basic notions for graduate students interested in the
foundations of the two-component spinor formalism or its applications in general
relativity or relativistic quantum mechanics. It is also intended for use as a reference
book, since it contains several results scattered in research papers, as well as some
previously unpublished results.
Throughout the book I have tried to stress the advantages of the two-component
spinor formalism over other formalisms encountered in the literature, and the ex-
amples considered here have been selected with this purpose. Whenever possible,
I have tried to derive the relations given in the book making use of the spinor for-
malism, not simply translating into the spinor notation some relation obtained by
means of another formalism.
My first encounter with the two-component spinor formalism was through Pro-
fessor Jerzy Plebański’s monograph entitled Spinors, Tetrads and Forms (Plebański
1974). This monograph circulated in handwritten form in the 1970s, and when Pro-
fessor Plebański decided to update his work and get it published as a book, he invited
me to collaborate in the task. After several not very successful attempts to complete
Professor Plebański’s manuscript, I decided to start from scratch, rewriting every-
thing. Unfortunately, owing to his much deteriorated health, Professor Plebański
was unable to give his opinion on the first chapter, which was almost finished before
his death in 2005. To a great extent, I have tried to follow the conventions employed
in the works of Professor Plebański and collaborators.
This book differs from the books about spinors already published and also from
Professor Plebański’s monograph in many details. It develops the two-component
spinor formalism for four-dimensional spaces of any signature, giving a detailed
study of the orthogonal groups based on spinors. It also gives a detailed account of
the relationship with the standard treatment of the Dirac bispinors.
v
vi Preface
The concept of the mate of a spinor (which can be defined only when the sig-
nature is Euclidean or ultrahyperbolic) is introduced and applied. Many worked
examples are included in the manuscript; in some of them, the bases of the self-dual
and the anti-self-dual two-forms are employed to find the connection and curvature
of the manifold. The self-dual electromagnetic fields are studied to find the solu-
tions of the source-free Maxwell equations, presenting results and derivations not
previously given in book form. The D(k, 0) Killing spinors and their applications
are discussed in some detail, and the formalism of the H H spaces is also included
with complete derivations. The Killing bispinors are studied making use of the two-
component spinor formalism.
In Chapter 1, which deals with the algebraic aspects of the spinor formalism and
contains a fairly complete discussion about the orthogonal groups, it is assumed
that the reader has some familiarity with linear algebra, elementary group theory,
and the use of tensor indices. Chapter 2 deals with the elementary applications of
the spinor formalism to Riemannian manifolds, assuming some basic knowledge
about differentiable manifolds, Riemannian connections, vector fields, and differen-
tial forms. For the last two chapters, it is convenient to have also some knowledge
of general relativity. Most of the examples considered in the book are taken from
general relativity and differential geometry.
I would like to thank the reviewers of the original version of the manuscript for
their very helpful comments. I am also grateful to Gerald Kaiser and Ann Kostant
at Birkhäuser for their support.
Puebla, Puebla, México Gerardo F. Torres del Castillo

October 2009
Contents
Preface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
1 Spinor Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Orthogonal Groups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Null Tetrads and the Spinor Equivalent of a Tensor . . . . . . . . . . . . . . 4
1.3 Spinorial Representation of the Orthogonal Transformations . . . . . . 17
1.3.1 Euclidean Signature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.2 Lorentzian Signature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
1.3.3 Ultrahyperbolic Signature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
1.4 Reflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
1.5 Clifford Algebra. Dirac Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
1.6 Inner Products. Mate of a Spinor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
1.7 Principal Spinors. Algebraic Classification . . . . . . . . . . . . . . . . . . . . . 59
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64
2 Connection and Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

2.1 Covariant Differentiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
2.2 Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
2.2.1 Curvature Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
2.2.2 Algebraic Classification of the Conformal Curvature . . . . . . . 91
2.3 Conformal Rescalings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
2.4 Killing Vectors. Lie Derivative of Spinors . . . . . . . . . . . . . . . . . . . . . . 95
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3 Applications to General Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

3.1 Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104
3.2 Dirac’s Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115
3.3 Einstein’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
3.3.1 The Goldberg–Sachs Theorem . . . . . . . . . . . . . . . . . . . . . . . . . 123
3.3.2 Space-Times with Symmetries. Ernst Potentials . . . . . . . . . . . 131
3.4 Killing Spinors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144
vii
viii Contents
4 Further Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

4.1 Self-Dual Yang–Mills Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149
4.2 H and H H Spaces . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152
4.3 Killing Bispinors. The Dirac Operator . . . . . . . . . . . . . . . . . . . . . . . . . 162
Exercises . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 165
A Bases Induced by Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
Chapter 1
Spinor Algebra
In this chapter the algebraic background of the two-component spinor formalism is

developed, showing that a four-dimensional real vector space with an inner prod-
uct can be considered as a subspace of the tensor product of two two-dimensional
complex vector spaces (the spin spaces), and several examples of the usefulness of
this identification are given, deriving various properties of the orthogonal transfor-
mations as well as several tensor relations.
Most existing applications of the two-component spinor formalism are related to
the space-time of special or general relativity, and it is sometimes asserted that the
two-component spinor formalism is tied to the signature of the space-time metric.
As shown below in detail, the two-component spinor formalism can be applied to
four-dimensional spaces of any signature.
In the traditional tensor formalism, one is acquainted with the fact that it is pos-
sible to construct tensors with any number of indices by means of the tensor prod-
uct of vectors. As we shall see, in turn, the vectors of a four-dimensional vector
space can be constructed from the tensor product of simpler objects, which are the
two-component spinors. In this sense, it is often said that the spinors are square
roots of vectors and that the spinors are more fundamental objects than the vectors
themselves. Any vector or tensor has a spinor equivalent, and all the operations be-
tween vectors and tensors have a straightforward counterpart in the spinor algebra.
As shown throughout this book, in many cases, the expression of the spinor equiva-
lent of a tensor is much simpler than that of the corresponding tensor, and therefore
it is easier to derive many relations employing spinors instead of tensors. Some of
the advantages of the spinor formalism come from the fact that each spinor index
takes two values only.
Apart from giving the basic rules of the spinor algebra, this chapter contains a
detailed study of the orthogonal groups, as an example of the simplifications that
can be achieved by making use of the spinor formalism. The connection between
the two-component spinors and the Dirac spinors, for all signatures of the metric, is
developed here, and the algebraic classification of totally symmetric spinors is also
discussed.
G.F.T. del Castillo, Spinors in Four-Dimensional Spaces, Progress in Mathematical 1

Physics 59, DOI 10.1007/978-0-8176-4984-5_1,
2 1 Spinor Algebra
1.1 Orthogonal Groups
Let V be a four-dimensional real vector space with a metric tensor g (i.e., g is a

nondegenerate, symmetric, not necessarily definite bilinear form). (Some authors,
e.g., Hall 2004, reserve the name metric tensor for the tensor field defining a Rie-
mannian manifold, reserving inner product for the nondegenerate bilinear form of a
vector space.) It is always possible to find an orthogonal basis of V , {e1 , e2 , e3 , e4 },
such that for a = 1, 2, 3, 4, g(ea , ea ) = 1 or −1; in other words, the 4 × 4 matrix
(gab ) (a, b = 1, 2, 3, 4) that represents the metric tensor g with respect to the basis
{e1 , e2 , e3 , e4 }, defined by gab = g(ea , eb ), is diagonal with 1’s or (−1)’s along the
diagonal. Such a basis will be called orthonormal. We can assume that the basis
vectors e1 , e2 , e3 , e4 are ordered in such a way that the first p entries of the diagonal
are equal to 1 and the last q = 4 − p entries are equal to (−1).
The numbers, p and q, of 1’s and (−1)’s appearing in the diagonal matrix (gab )
do not depend on the orthonormal basis chosen, but are fixed by the metric tensor.
The pair of numbers (p, q) defines the signature of g. Since p + q must be equal
to 4, the knowledge of p or q defines the signature of g; some authors define the sig-
nature as p − q. Thus, (gab ) must be the matrix diag (1, 1, 1, 1), diag (1, 1, 1, −1),
diag (1, 1, −1, −1), diag (1, −1, −1, −1), or diag (−1, −1, −1, −1). Since the last
two cases are obtained from the first two by reversing the sign of g and relabel-
ing the basis vectors, we will consider only the first three cases
⎧
⎨ diag (1, 1, 1, 1) (Euclidean signature),
(gab ) = diag (1, 1, 1, −1) (Lorentzian or hyperbolic signature), (1.1)
⎩
diag (1, 1, −1, −1) (Kleinian or ultrahyperbolic signature).
In what follows it will be assumed that the metric tensor of V has one of the
forms (1.1).
For a given signature, there are infinitely many bases with respect to which (gab )
takes one of the forms (1.1). If {e1 , e2 , e3 , e4 } and {e1 , e2 , e3 , e4 } are two orthonor-
mal bases of V with respect to which g is represented by the same matrix (gab ), i.e.,
g(ea , eb ) = g(ea , eb ), (1.2)

then the fact that {e1 , e2 , e3 , e4 } is a basis of V implies the existence of a 4 × 4 real
matrix (La b ) such that
ea = Lb a e b (1.3)
(with a sum over repeated indices; here and in what follows, lowercase Latin indices
a, b, . . . take values from 1 to 4). Substituting (1.3) into (1.2), using the bilinearity of
g, one obtains
gab = Lc a Ld b gcd . (1.4)
This condition implies that det(La b ) = 1 or −1 (and, hence, (La b ) is invertible).
The matrices (La b ) satisfying (1.4) are called orthogonal, and as can readily be
seen, the set of orthogonal matrices forms a group with the usual matrix multi-
plication as the group operation, which is denoted by O(p, q), where (p, q) is the
1.1 Orthogonal Groups 3
signature of g. The matrices satisfying (1.4) with det(La b ) = 1 form a subgroup of

O(p, q), denoted by SO(p, q). (When q is equal to zero, we simply write O(4) or
SO(4) instead of O(4, 0) or SO(4, 0), respectively.)
If (gab ) is the inverse of the matrix (gab ), i.e., gab gbc = δac , from (1.4) we obtain
δb = gea gab = gea Lc a Ld b gcd = gea Lc a gcd Ld b , which means that the inverse of (La b ),
e
with entries (L−1 )a b , is given by

(L−1 )e d = gea Lc a gcd , (1.5)
or, following the standard rules for raising and lowering tensor indices by means
of the metric tensor (gab ) and its inverse (gab ) (e.g., t a = gabtb , ta = gabt b ), we can
write (1.5) as
(L−1 )e d = Ld e . (1.6)
The inverse of a matrix satisfying (1.4) also belongs to O(p, q) and therefore also
satisfies condition (1.4), i.e., gab = (L−1 )c a (L−1 )d b gcd , and according to (1.6) we
conclude that
gab = La c Lb d gcd . (1.7)
In order to motivate the spinor notation employed in what follows, it is convenient
to consider the space R2,2 , which is the vector space R4 with the ultrahyperbolic
metric tensor (gab ) = diag (1, 1, −1, −1) with respect to the canonical basis. The
mapping

1 −x − z −y − w
(x, y, z, w) → √ (1.8)
2 w−y x−z
is a one-to-one correspondence between
√ R2,2 and the vector space formed by the
2 × 2 real matrices (the factor 1/ 2 is introduced in order to get agreement with
the conventions adopted in the following sections). Denoting by P the matrix on the
right-hand side of (1.8), we see that
1
detP = − (x2 + y2 − z2 − w2 ),
2
i.e., apart from the factor −1/2, det P is the inner product of the vector (x, y, z, w)
with itself.
If K and M are 2×2 matrices, both real or both pure imaginary, the transformation
P → P ≡ KPM (1.9)
is linear and, by means of the correspondence (1.8), is equivalent to some linear
transformation of R2,2 into itself. Since det P = det(KPM) = (detK)(det P)(det M),
it follows that if
(detK)(det M) = 1, (1.10)
then detP = detP ; that is, denoting by (x , y , z , w ) the vector corresponding to P
according to (1.8), the inner product of (x, y, z, w) with itself is equal to the inner
product of (x , y , z , w ) with itself. This implies that the mapping (1.9) corresponds
to an orthogonal transformation of R2,2 , i.e., an element of O(2, 2).
4 1 Spinor Algebra
≡
Assuming that the matrices K, M satisfy the condition (1.10), letting K
(det K)−1/2
K, M ≡ (det K) M, we have
1/2
M,
P = KPM = KP
and det K = 1 = det M; hence, in order for (1.9) to correspond to an orthogonal
transformation, we can assume that the determinants of K and M are equal to 1.
(Note that since K and M are both real or both pure imaginary, (det K)1/2 is real or
pure imaginary and the matrices K and M are also both real or both pure imaginary.)
Hence, by representing the points of R by 2 × 2 matrices, as in (1.8), the simple
2,2
algebraic conditions detK = 1 = det M, allow us to find orthogonal transformations.

If we denote the entries of P by Pi j , where the superscript labels the row and the
subscript labels the column, and similarly for the other 2 × 2 matrices, the transfor-
mation (1.9) is equivalent to
Pi j = K i k Pk l M l j = K i k M l j Pk l . (1.11)
This last equation is similar to the transformation law for the components of a rank-
two tensor, with the difference that in the latter case, the matrix K would be the
inverse of M, while in (1.11), K and M can be two arbitrary unimodular 2 × 2 real
or pure imaginary matrices.
Hence, it is convenient to distinguish the two indices labeling the entries of the
matrix P in a way that explicitly indicates which one of the matrices K and M
appearing in (1.9) is employed in the transformation. Specifically, the entries of K
will be labeled with undotted indices A, B, . . . that take two values only, K = (K A B )
(A, B, . . . = 1, 2), while the entries of M will be labeled with dotted indices Ȧ, Ḃ, . . .,
M = (M Ȧ Ḃ ) (Ȧ, Ḃ, . . . = 1̇, 2̇), and the entries of P and P will be labeled with one
undotted index and one dotted index, P = (PA Ḃ ), so that (1.9) amounts to
PA Ḃ = K AC PC Ḋ M Ḋ Ḃ = K AC M Ḋ Ḃ PC Ḋ . (1.12)
As we shall show in the following sections, all the SO(p, q) transformations can
be expressed in a form analogous to (1.12), though in some cases it will be necessary
or convenient to make use of linear combinations of vectors of V with complex
scalars.
1.2 Null Tetrads and the Spinor Equivalent of a Tensor
Apart from the orthonormal bases considered in the foregoing section, it will be
useful to consider bases, {E1 , E2 , E3 , E4 }, with respect to which the metric tensor
of V is represented by the matrix
⎛ ⎞
0100
⎜1 0 0 0⎟
⎜ ⎟
⎝ 0 0 0 1 ⎠, (1.13)
0010
1.2 Null Tetrads and the Spinor Equivalent of a Tensor 5
i.e., the only nonvanishing inner products among the vectors Ea are given by
g(E1 , E2 ) = 1 = g(E3 , E4 ). Such a basis will be called a null tetrad, since each
basis vector is null (g(Ea , Ea ) = 0, without summation on a). Only when the sig-
nature of the metric of V is ultrahyperbolic will it be possible to find a null tetrad
formed by real vectors (see (1.16) below), and therefore, in most cases we will have
to assume that the vectors Ea belong to the complexification of V (see, e.g., Hirsch
and Smale 1974). For instance, if the signature of V is Euclidean and {e1 , e2 , e3 , e4 }
is an orthonormal basis of V , we can take
1 1
E1 = √ (e1 + ie2 ), E2 = √ (e1 − ie2 ),
2 2
(1.14)
1 1
E3 = √ (e3 − ie4 ), E4 = √ (e3 + ie4 ).
2 2
Similarly, it can be readily verified that if (gab ) = diag (1, 1, 1, −1), we can take
1 1
E1 = √ (e1 + ie2 ), E2 = √ (e1 − ie2 ),
2 2
(1.15)
1 1
E3 = √ (e3 + e4 ), E4 = √ (e3 − e4 ),
2 2
while if (gab ) = diag (1, 1, −1, −1), a convenient choice is
1 1
E1 = √ (e1 − e3), E2 = √ (e1 + e3),
2 2
(1.16)
1 1
E3 = √ (−e2 + e4 ), E4 = √ (−e2 − e4 ).
2 2
Conversely, if one assumes that the only nonzero inner products among the
vectors Ea are given by g(E1 , E2 ) = 1 = g(E3 , E4 ), then the vectors ea given by
(1.14)–(1.16) form an orthonormal basis with respect to which the metric tensor
is represented by one of the forms (1.1), depending on the signature of the met-
ric tensor.
Whereas the components of the metric tensor with respect to an orthonormal
basis {e1 , e2 , e3 , e4 } depend on the signature of the metric tensor [see (1.1)], the
components of the metric tensor with respect to a null tetrad {E1 , E2 , E3 , E4 } will
be given in all cases by (1.13); the signature of the metric tensor is determined by the
behavior of the vectors Ea under complex conjugation. (For instance, the relations
E1 = E2 , E3 = E4 , where the bar denotes complex conjugation, satisfied by the null
tetrad (1.14) imply that the metric has Euclidean signature.)
Instead of a single subscript, a = 1, 2, 3, 4, labeling the vectors, Ea , of a null
tetrad, it is convenient to make use of a pair of subscripts that take two values each.
Letting

E4 E2
(eAḂ ) ≡ (1.17)
E1 −E3
6 1 Spinor Algebra
(A, B = 1, 2, Ȧ, Ḃ = 1̇, 2̇), one finds that the definition of the null tetrad is equivalent to
g(eAḂ , eCḊ ) = −εAC εḂḊ , (1.18)
where
01
(εAB ) = (ε AB
)= = (εȦḂ ) = (ε ȦḂ ). (1.19)
−1 0
(Some works make use of primed indices instead of dotted ones (e.g., Penrose 1960,
Penrose and Rindler 1984, Stewart 1990).)
Since the vectors eAḂ belong to the complexification of V , there exist complex
scalars, σ a AḂ , such that
1
eAḂ = √ σ a AḂ ea , (1.20)
2
and as a consequence of (1.18), they obey
σ a AḂ σ bCḊ gab = −2εAC εḂḊ . (1.21)
The scalars σ a AḂ will be called connection symbols or Infeld–van der Waerden
symbols.
The Levi-Civita symbols (1.19) will be employed to lower or raise the spinor
indices A, B. We shall follow the convention (e.g., Plebański 1974, 1975)
ψA = εAB ψ B , ψ A = ψB ε BA (1.22)
(thus, ψ 2 = ψ1 , ψ 1 = −ψ2 ), and similarly for dotted indices. It should be remarked

that other authors (e.g., Penrose 1960, Penrose and Rindler 1984, Wald 1984,
Stewart 1990) follow the convention ψ A = ε AB ψB . The antisymmetry of εAB implies
that
ψ A φA = ψ A εAB φ B = −ψ A εBA φ B = −ψB φ B = −ψA φ A ; (1.23)
thus,
ψ A ψA = 0. (1.24)
Similar results hold for objects with more than one index (dotted or undotted), e.g.,
αAB β AB = −α A B βA B = α AB βAB , μAḂC ν AḂ = −μ A ḂC νA Ḃ = μ AḂC νAḂ . Furthermore,
from (1.22) it follows that
ε A B = δBA and εA B = −δAB . (1.25)
The definitions (1.19) are consistent with the rules (1.22) in the sense that, for in-
stance, lowering the indices of ε AB , one obtains εAB , that is, εAC εBD ε CD = εAB .
The Levi-Civita symbols (1.19) will appear very often in what follows, in many
cases as a consequence of the fact that any object antisymmetric in two indices must
be proportional to one of the symbols (1.19).
Proposition 1.1. If ψAB = −ψBA , then
ψAB = 12 ψ R R εAB . (1.26)
0a
Proof. Any antisymmetric 2×2 matrix is of the form −a 0 , which can be written as
a(εAB ); therefore, if ψAB is antisymmetric, it must be a multiple of εAB : ψAB = aεAB ,
for some a. Then, by contracting both sides of this equality with ε AB ; one obtains
ψ A A = 2a, which leads to (1.26).

Similarly, ψ AB = 12 ψ R R ε AB , if ψ AB is antisymmetric in A, B, with analogous
identities for objects with dotted indices. Furthermore, for an object with more than
two indices, e.g., hA ḂC D , such that hA ḂC D = −hC Ḃ A D , we have hA ḂC D = 12 hRḂ R D εAC .
As a corollary, noting that (1.26) can be written as ψAB = 12 (ε RS ψRS )εAB , we see
that if ψAB...ĊḊ... is any object with at least two indices of the same type (dotted or
undotted), a difference of the form
ψA...R...S...ĊḊ... − ψA...S...R...ĊḊ... ,
where only two indices of the same type interchange places, is antisymmetric in
R, S, and therefore it is equal to 12 ε PQ (ψ...P...Q... − ψ...Q...P... ) εRS = 12 (ψ... Q ...Q... −
ψ...Q... Q ... ) εRS = ψ... Q ...Q... εRS . Thus,
ψ...R...S... − ψ...S...R... = ψ... Q ...Q... εRS , (1.27)
with analogous relations for the cases where the interchanged indices are dotted or
appear as superscripts (dotted or undotted). By raising, e.g., the index R on both
sides of (1.27), one obtains
ψ... R ...S... − ψ...S... R ... = ψ... Q ...Q... δSR . (1.28)
Note also that (1.27) implies that
ψ...R...S... = ψ...S...R... ⇔ ψ... Q ...Q... = 0. (1.29)
If (M R S ) is a (real or complex) 2 × 2 matrix, then

εAC M A B MC D = det(M R S )εBD . (1.30)
In fact, letting ψBD ≡ εAC M A B MC D , it follows that
ψDB = εAC M A D MC B = −εCA MC B M A D = −ψBD ,

and since ψ R R = ψ 1 1 + ψ 2 2 = −ψ21 + ψ12 = 2ψ12 = 2εAC M A 1 MC 2 = 2(M 1 1 M 2 2 −
M 2 1 M 1 2 ) = 2 det(M R S ), making use of (1.26) one obtains (1.30).
Given a set of connection symbols σ a AḂ , the scalars
σ a AḂ = KC A M Ḋ Ḃ σ aCḊ (1.31)
also satisfy (1.21) (with the same gab ) if and only if det(K A B ) det(M Ȧ Ḃ ) = 1. In fact,
from (1.31) and (1.30) we have
σ a AḂ σ bCḊ gab = K E A M Ḟ Ḃ σ a E Ḟ K GC M Ḣ Ḋ σ b GḢ gab
= K E A M Ḟ Ḃ K GC M Ḣ Ḋ (−2εEG εḞ Ḣ )
= −2 det(K A B ) εAC det(M Ȧ Ḃ ) εḂḊ
= −2εAC εḂḊ .
8 1 Spinor Algebra
Similarly,
σ a AḂ = KĊ A M D Ḃ σ a DĊ (1.32)
satisfy (1.21) if and only if det(K Ȧ B ) det(M A Ḃ ) = 1. As we shall show in the fol-
lowing section, given two sets of Infeld–van der Waerden symbols σ a AḂ and σ a AḂ
(satisfying (1.21) with the same metric tensor gab ), there exist two unimodular 2 × 2
matrices such that (1.31) or (1.32) holds.
A Convenient Choice for the Infeld–van der Waerden Symbols
Comparing (1.14) with (1.20), one finds that when (gab ) = diag (1, 1, 1, 1), the
Infeld–van der Waerden symbols can be chosen as

01 0 −i
(σ AḂ ) =
1 , (σ AḂ ) =
2 ,
10 i 0
(1.33)
1 0 i 0
(σ AḂ ) =
3 , (σ AḂ ) =
4 .
0 −1 0 i
(Note that the first three matrices in (1.33) are the usual Pauli matrices.)
Similarly, when (gab ) = diag (1, 1, 1, −1), from (1.15), one finds that we can take

01 0 −i
(σ 1 AḂ ) = , (σ 2 AḂ ) = ,
10 i 0
(1.34)
1 0 −1 0
(σ 3 AḂ ) = , (σ 4 AḂ ) = ,
0 −1 0 −1
and if (gab ) = diag (1, 1, −1, −1), then

01 −1 0
(σ AḂ ) =
1 , (σ 2 AḂ ) = ,
10 01
(1.35)
01 −1 0
(σ 3 AḂ ) = , (σ 4 AḂ ) = ,
−1 0 0 −1
which are all real [see (1.16)]. By means of the relation (1.31), with the σ a AḂ given
by (1.35) and the unimodular matrices

0 1 1 − i −1 + i 0 1 1 + i −1 − i
(K A B ) ≡ , (M Ȧ Ḃ ) ≡ , (1.36)
2 1+i 1+i 2 1−i 1−i
one obtains another set of connection symbols for the same signature given by (drop-
ping the tilde)

0 −i 01
(σ 1 AḂ ) = , (σ 2 AḂ ) = ,
i 0 10
(1.37)
−i 0 −1 0
(σ AḂ ) =
3
, (σ 4
AḂ ) = .
0 i 0 −1
The choices given in (1.33)–(1.35) and (1.37) are convenient because they satisfy
the relations
⎧
⎪
⎪ −σ aAḂ if the σ a AḂ are given by (1.33),
⎨ a
σ BȦ if the σ a AḂ are given by (1.34),
σ a AḂ = (1.38)
⎪ σ AḂ
⎪
a if the σ a AḂ are given by (1.35),
⎩
ηAC ηḂḊ σ aCḊ if the σ a AḂ are given by (1.37),
where
1 0
(ηAB ) ≡ ≡ (ηȦḂ ). (1.39)
0 −1
By contracting both sides of (1.21) with σcCḊ , one obtains
σ a AḂ σ bCḊ σcCḊ gab = −2εAC εḂḊ σcCḊ = −2σcAḂ,
which is equivalent to
σ a AḂ (σaCḊ σcCḊ + 2gac ) = 0.
Since the Infeld–van der Waerden symbols represent a change of basis, they corre-
spond to an invertible relation, and therefore, from the last equation it follows that
σaAḂ σb AḂ = −2gab. (1.40)

The minus sign appearing in the right-hand sides of and (1.21) and (1.40), as
well as in very many equations in the rest of the book, is necessary in order to
have the simple relation σ a AḂ = σ a BȦ when (gab ) = diag (1, 1, 1, −1); this sign
can be avoided, maintaining the relation σ a AḂ = σ a BȦ , if one chooses (gab ) =
diag (1, −1, −1, −1), as in Penrose (1960), Penrose and Rindler (1984), and
Stewart (1990).
If tab...d are the components of a tensor with respect to the orthonormal basis
{e1 , e2 , e3 , e4 }, then the components of its spinor equivalent are defined by
1 1 1
tAḂCḊ...GḢ ≡ √ σ a AḂ √ σ bCḊ · · · √ σ d GḢ tab...d . (1.41)
2 2 2
It is often convenient to write (1.41) in the form
1 1 1
tAȦBḂ...DḊ ≡ √ σ a AȦ √ σ b BḂ · · · √ σ d DḊ tab...d , (1.42)
2 2 2
where it is understood that the indices A and Ȧ are independent of each other, and
so on; in this manner, the tensor index a is replaced by the pair of indices AȦ.
10 1 Spinor Algebra
Sometimes all the undotted indices are written to the left of all the dotted indices,
maintaining the order in which the indices of each type appear, that is,
1 1 1
tAB...DȦḂ...Ḋ ≡ √ σ a AȦ √ σ b BḂ · · · √ σ d DḊ tab...d . (1.43)
2 2 2
Then, according to (1.40),

1 1 1
tab...d = − √ σa AȦ
− √ σb BḂ
· · · − √ σd DḊ
tAȦBḂ...DḊ . (1.44)
2 2 2
Owing to (1.20), the spinor equivalent of a tensor gives the components of that tensor
with respect to the basis {e11̇ , e12̇ , e21̇ , e22̇ }. For instance, an arbitrary element of V
can be expressed as a linear combination va ea , and by virtue of (1.44) and (1.20),
1
va ea = − √ σ a AȦ vAȦ ea = −vAȦ eAȦ . (1.45)
2
From the definition (1.42) it follows that the spinor equivalent of a sum [resp.
product] of tensors is the sum [resp. product] of their spinor equivalents; however,
the spinor equivalent of the contraction of a tensor is not always equal to the con-
traction of its spinor equivalent. From (1.21) we obtain, for instance,
ta sa = −tAȦ sAȦ . (1.46)
Equation (1.45) allows us to consider the complexification of V as the tensor

product of two complex two-dimensional vector spaces called spin spaces. We iden-
tify each null vector eAḂ with the tensor product eA ⊗ eḂ , where {e1 , e2 } and {e1̇ , e2̇ }
are bases of the spin spaces. The elements of the spin spaces will be called one-index
spinors, since with respect to the bases {e1 , e2 } and {e1̇ , e2̇ } they are represented by
complex components of the form φ A and φ Ȧ , respectively (i.e., an arbitrary element
of each spin space has the form φ A eA or φ Ȧ eȦ ).
By forming tensor products of one-index spinors we obtain spinors with any
number of dotted or undotted indices (the spinor equivalent of a tensor is a special
case for which the number of dotted indices coincides with the number of undotted
indices).
From (1.38) and (1.41) it follows that the components of an n-index tensor tab...d
with respect to the orthonormal basis {e1 , e2 , e3 , e4 } are real if and only if
tAḂCḊ...GḢ
⎧
⎪
⎪ (−1)nt AḂCḊ...GḢ if the σ a AḂ are given by (1.33),
⎨
tBȦDĊ...H Ġ if the σ a AḂ are given by (1.34),
=
⎪
⎪ t if the σ a AḂ are given by (1.35),
⎩ AḂCḊ...GḢ
ηAR ηḂṠ ηCT ηḊU̇ . . . ηGV ηḢẆ t RṠT U̇...V Ẇ if the σ a AḂ are given by (1.37).
(1.47)
Each of the two sets of Infeld–van der Waerden symbols given above for the case
of the ultrahyperbolic signature, (1.35) and (1.37), has a certain advantage over the
other. The set of real Infeld–van der Waerden symbols may seem more natural in
the sense that the complexification of V is not necessary and the spinor equivalent
of a real tensor is real. On the other hand, the use of the complex Infeld–van der
Waerden symbols allows us to reduce the number of independent components or
equations to consider.
The Spinor Equivalents of Symmetric or Antisymmetric Tensors
The symmetries of a tensor are inherited by its spinor equivalent. For example, if tab
is a two-index symmetric tensor, then its spinor equivalent has the symmetry
tAȦBḂ = tBḂAȦ , (1.48)
but tAȦBḂ will not necessarily coincide with tBȦAḂ (which, according to (1.48), is
equal to tAḂBȦ ). In fact, using (1.27), we have
tAȦBḂ − tBȦAḂ = εABt R ȦRḂ ,
but owing to (1.48), t R ȦRḂ is antisymmetric,
t R ȦRḂ = tRḂ R Ȧ = −t R ḂRȦ ,
and therefore t R ȦRḂ is proportional to εȦḂ ; thus
tAȦBḂ − tBȦAḂ = εAB 12 t RṠ RṠ εȦḂ = − 12 t s s εAB εȦḂ , (1.49)
which means that tAȦBḂ = tBȦAḂ = tAḂBȦ if and only if tAȦBḂ is the spinor equiv-
alent of a traceless symmetric tensor. Similarly, one finds that tAȦBḂ...CĊ is com-
pletely symmetric separately on the undotted and on the dotted indices if and only
if tAȦBḂ...CĊ is the spinor equivalent of a traceless totally symmetric tensor.
In the case of a completely symmetric object MAB...L with n spinor indices, the
n + 1 components M111...1 , M211...1 , M221...1 , . . . , M222...2 , are independent; hence an
n-index traceless totally symmetric tensor has (n + 1)2 independent components.
(Note that the components tAȦBḂ...CĊ may be complex, but also in that case, taking
into account the conditions (1.47), one concludes that a real n-index traceless totally
symmetric tensor has (n + 1)2 real independent components.)
The spinor equivalent of an antisymmetric two-index tensor tab = −tba satisfies
tAȦBḂ = −tBḂAȦ ; hence,
tAȦBḂ = 12 (tAȦBḂ + tBȦAḂ ) + 12 (tAȦBḂ − tBȦAḂ )
= 12 (tAȦBḂ − tAḂBȦ ) + 12 (tAȦBḂ − tBȦAḂ )
= 12 tA Ṙ BṘ εȦḂ + 12 t R ȦRḂ εAB
= τAB εȦḂ + τȦḂ εAB , (1.50)
12 1 Spinor Algebra
where
τAB ≡ 12 tA Ṙ BṘ , τȦḂ ≡ 12 t R ȦRḂ . (1.51)
Owing to tAȦBḂ = −tBḂAȦ and (1.23), τAB and τȦḂ are symmetric in their two indices;
e.g., tB Ṙ AṘ = −tAṘB Ṙ = tA Ṙ BṘ .
Note that if tab is real, the components τAB and τȦḂ may be complex or related
to one another. For instance, if the metric of V has Lorentzian signature, then from
(1.47) we have tAḂCḊ = tBȦDĊ and therefore
τAB = 12 tA Ṙ BṘ = 12 t R ȦRḂ = τȦḂ

(for the other two signatures see Section 1.6).
In particular, from (1.50) one finds that the spinor equivalent of the antisym-
metrized product va wb − vb wa of a pair of vectors is
vAȦ wBḂ − vBḂwAȦ = 12 (vA Ṙ wBṘ − vBṘwA Ṙ )εȦḂ + 12 (vR Ȧ wRḂ − vRḂ wR Ȧ )εAB
= 12 (vA Ṙ wBṘ + vBṘ wAṘ )εȦḂ + 12 (vR Ȧ wRḂ + vR Ḃ wRȦ )εAB
= v(A Ṙ wB)Ṙ εȦḂ + vR (Ȧ w|R|Ḃ) εAB , (1.52)
where the parentheses denote symmetrization on the indices enclosed [e.g., ξ(AB) =
2 (ξAB + ξBA )] and the indices between bars are excluded from the symmetrization.
1
If tabc is totally antisymmetric (that is, it changes sign under any transposition of
a pair of its indices), then in particular, it is antisymmetric in the last pair of indices;
therefore, according to the preceding results, the spinor equivalent of tabc must be of
the form
tAȦBḂCĊ = μAȦBC εḂĊ + μAȦḂĊ εBC , (1.53)
with μAȦBC = μAȦ(BC) , μAȦḂĊ = μAȦ(ḂĊ) . The antisymmetry in the first pair of in-
dices of tabc is then equivalent to
μAȦBC εḂĊ + μAȦḂĊ εBC = −μBḂAC εȦĊ − μBḂȦĊ εAC . (1.54)
Contracting the last equation with ε ḂĊ , we obtain [see (1.29)]
2 μAȦBC = −μB Ḃ AC εȦḂ − μBḂ ȦḂ εAC

= −μBȦAC − μBḂ ȦḂ εAC
= −μAȦBC + μAȦBC − μBȦAC − μBṘ ȦṘ εAC ,
and hence, making use of (1.27),
3μAȦBC = μ R ȦRC εAB − μBṘ ȦṘ εAC . (1.55)
Contracting now this last equation with ε BC , we obtain
0 = −μ R ȦRA − μA Ṙ ȦṘ ,
and therefore, defining tBȦ ≡ 23 μ R ȦRB , from (1.55) we obtain
2 μAȦBC = tCȦ εAB + tBȦ εAC . (1.56)
In an entirely similar way, from (1.54) one arrives at

2μAȦḂĊ = −tAĊ εȦḂ − tAḂεȦĊ . (1.57)
Substituting (1.56) and (1.57) into (1.53), it follows that the spinor equivalent of
a three-index antisymmetric tensor is of the form
tAȦBḂCĊ = 12 [(tCȦ εAB + tBȦεAC )εḂĊ − (tAĊ εȦḂ + tAḂ εȦĊ )εBC ]
= 12 [(tCȦ εAB − tBȦεAC + 2tBȦεAC )εḂĊ
− (tAĊ εȦḂ − tAḂεȦĊ + 2tAḂεȦĊ )εBC ]
= 2 [(t Ȧ εAR εCB + 2tBȦεAC )εḂĊ
1 R
− (tA Ṙ εȦṘ εĊḂ + 2tAḂεȦĊ )εBC ]

= tBȦ εAC εḂĊ − tAḂεBC εȦĊ , (1.58)
for some tBȦ .
If now tabcd is a totally antisymmetric four-index tensor, then as a consequence
of the antisymmetry on the first three indices, according to (1.58) we can write
tAȦBḂCĊDḊ = φBȦDḊ εAC εḂĊ − φAḂDḊ εȦĊ εBC , (1.59)
and the antisymmetry of tabcd in the last two indices is then equivalent to
φBȦDḊ εAC εḂĊ − φAḂDḊ εȦĊ εBC = −φBȦCĊ εAD εḂḊ + φAḂCĊ εȦḊ εBD ;
hence, contracting with ε AC , we have
2φBȦDḊ εḂĊ − φBḂDḊ εȦĊ = −φBȦDĊ εḂḊ + φ R ḂRĊ εȦḊ εBD , (1.60)
and contracting this last equation with ε ȦĊ , we obtain
0 = −φB Ṙ DṘ εḂḊ + φ R ḂRḊ εBD ,
which implies that
φB Ṙ DṘ = 2a εBD , φ R ḂRḊ = 2a εḂḊ , (1.61)

for some scalar a.
On the other hand, contracting (1.60) with ε ḂḊ , rearranging, and using (1.61),
one finds that
4φBȦDĊ = φB Ḃ DḂ εȦĊ + φ RḂ RĊ εȦḂ εBD
= 2a(εBD εȦĊ + εȦĊ εBD )
= 4aεBD εȦĊ . (1.62)
14 1 Spinor Algebra
Substituting (1.62) into (1.59), one finds that the spinor equivalent of an antisym-
metric four-index tensor is of the form
tAȦBḂCĊDḊ = a(εBD εȦḊ εAC εḂĊ − εAD εȦĊ εBC εḂḊ ), (1.63)
for some scalar a.
The spinor equivalent εAȦBḂCĊDḊ of the Levi-Civita symbol εabcd , which is anti-
symmetric with ε1234 = 1, must be proportional to εAC εBD εȦḊ εḂĊ − εAD εBC εȦĊ εḂḊ .
Since ε abcd εabcd = (−1)q 24, where q is the number of negative entries in the diag-
onal matrix (gab ), the proportionality factor is real or pure imaginary depending on
the signature of the metric tensor, and for a given signature, this factor is defined
up to sign. From ε1234 = 1 one finds that, with the connection symbols given by
(1.33)–(1.35) and (1.37),
εAȦBḂCĊDḊ = iq (εAC εBD εȦḊ εḂĊ − εAD εBC εȦĊ εḂḊ ). (1.64)
Therefore, the spinor equivalent of the dual of an antisymmetric two-index tensor

tab , defined by ∗tab ≡ 12 εabcd t cd , is [see (1.50)]
∗
tAȦBḂ = iq (−τAB εȦḂ + τȦḂ εAB ), (1.65)
with τAB and τȦḂ defined by (1.51), and from this formula we readily obtain the
well-known fact that
∗ ∗
( tab ) = (−1)q tab . (1.66)
iq
(Note that the factor appears in (1.64) and (1.65) as a consequence of the specific
choice for the connection symbols (1.33)–(1.35) and (1.37); by contrast, (1.66) fol-
lows from the definition of the dual of tab .) The terms τȦḂ εAB and τAB εȦḂ will be
referred to as the self-dual and the anti-self-dual parts of tABȦḂ , respectively, though
in the case of ultrahyperbolic signature, τAB εȦḂ is the spinor equivalent of an anti-
symmetric tensor that coincides with its dual.
From (1.50) and (1.65) we see that for an arbitrary antisymmetric two-index
tensor (or bivector) tab , we have
∗
tabt ab = −iq (τAB εȦḂ − τȦḂ εAB )(τ AB ε ȦḂ + τ ȦḂ ε AB )
= −2iq (τAB τ AB − τȦḂ τ ȦḂ ).
The bivector tab is simple if there exist vectors va , wa such that tab = va wb − vb wa .
Hence, for a simple bivector, ∗tabt ab = 12 εabcd t cd t ab = 2εabcd vc wd va wb = 0. Thus,
we have proved the following result.
Proposition 1.2. If the bivector tab is simple, then
τAB τ AB = τȦḂ τ ȦḂ . (1.67)
The converse of this proposition is also true, and we shall give a constructive
proof in Section 1.7. Combining (1.65) and (1.67), one concludes that a bivector tab
is simple if and only if its dual is simple. (See also Exercise 1.6.)
Spin-Tensors and Other Derived Objects
Substituting (1.50) into (1.44), one finds that an antisymmetric two-index tensor can
be expressed in the form
tab = 12 σa AȦ σb BḂ (τAB εȦḂ + τȦḂ εAB )

= 12 τAB σa (A|Ṙ| σb B) Ṙ + 12 τȦḂ σa R(Ȧ σbR Ḃ)
= 12 τAB Sab AB + 12 τȦḂ Sab ȦḂ , (1.68)
where
Sab AB ≡ σa (A|Ṙ| σb B) Ṙ , Sab ȦḂ ≡ σa R(Ȧ σbR Ḃ) . (1.69)
Making use of (1.23), we obtain the alternative expression
Sab AB = 12 (σa AṘ σb B Ṙ + σa BṘ σb A Ṙ )

= 12 (σa AṘ σb B Ṙ − σb AṘ σa B Ṙ )
= σ[a AṘ σb] B Ṙ , (1.70)
where the brackets denote antisymmetrization on the indices enclosed. In a similar

way, we have
Sab ȦḂ = σ[a RȦ σb]R Ḃ .
There are many relations satisfied by the spin-tensors Sab AB and Sab ȦḂ that follow
directly from (1.21). For example,
Sab AB SabCD = 4(ε AC ε BD + ε AD ε BC ),

Sab AB SabĊḊ = 0, (1.71)
Sab ȦḂ abĊḊ
S = 4(ε ε
ȦĊ ḂḊ
+ε ε
ȦḊ ḂĊ
).
Hence, from (1.68) we see that
τ AB = 14 tab SabAB, τ ȦḂ = 14 tab SabȦḂ . (1.72)
On the other hand, from (1.65) we obtain

∗
tab = −iq ( 12 τAB Sab AB − 12 τȦḂ Sab ȦḂ ), (1.73)
and therefore, combining (1.68) and (1.73), we obtain, for example,

tab − (−i)q ∗tab = τAB Sab AB = 12 gac gbd − gad gbc − (−i)q εabcd t cd ,
and making use of (1.72), we have τAB Sab AB = 14 t cd ScdAB Sab AB . In this manner we
conclude that
Sab AB ScdAB = 2 gac gbd − gad gbc − (−i)q εabcd (1.74)
16 1 Spinor Algebra
and, in a similar way,

Sab ȦḂ Scd ȦḂ = 2 gac gbd − gad gbc + (−i)q εabcd . (1.75)
Substituting (1.68) and (1.73) into the definition ∗tab = 12 εabcd t cd , it follows that
2 εabcd S = −iq Sab AB , 2 εabcd S = iq Sab ȦḂ .

1 cdAB 1 cd ȦḂ
(1.76)
The second equation (1.71) implies that the contraction mabtab of a self-dual bivector
with an anti-self-dual one vanishes. Furthermore, (1.71) implies that if mabtab = 0
for all self-dual [resp. anti-self-dual] bivectors mab , then tab is anti-self-dual [resp.
self-dual].
The complex scalars Sab A B and Sab Ȧ Ḃ can be arranged in 2 × 2 matrices, em-
ploying the spinor indices to label the entries of these matrices. A straightforward
computation, using the explicit expressions (1.33)–(1.35) and (1.37), yields

(S23 A B ), (S31 A B ), (S12 A B ), (S14 A B ), (S24 A B ), (S34 A B )
⎧
⎪ (iσ1 , −iσ2 , iσ3 , −iσ1 , iσ2 , −iσ3 ) if the σ a AḂ are given by (1.33),
⎪ a
⎨
(iσ1 , −iσ2 , iσ3 , −σ1 , σ2 , −σ3 ) if the σ AḂ are given by (1.34),
=
⎪
⎪ (σ1 , −σ3 , −iσ2 , −σ1 , σ3 , −iσ2 ) if the σ a AḂ are given by (1.35),
⎩
(−σ2 , σ1 , −iσ3 , σ2 , −σ1 , −iσ3 ) if the σ a AḂ are given by (1.37),
(1.77)
where the σi are the standard Pauli matrices,

01 0 −i 1 0
σ1 = , σ2 = , σ3 = , (1.78)
10 i 0 0 −1
and similarly,

(S23 Ȧ Ḃ ), (S31 Ȧ Ḃ ), (S12 Ȧ Ḃ ), (S14 Ȧ Ḃ ), (S24 Ȧ Ḃ ), (S34 Ȧ Ḃ )
⎧
⎪
⎪ (−iσ1 , −iσ2 , −iσ3 , −iσ1 , −iσ2 , −iσ3 ) if the σ a AḂ are given by (1.33),
⎨
(−iσ1 , −iσ2 , −iσ3 , −σ1 , −σ2 , −σ3 ) if the σ a AḂ are given by (1.34),
=
⎪
⎪ (−σ1 , σ3 , −iσ2 , −σ1 , σ3 , iσ2 ) if the σ a AḂ are given by (1.35),
⎩
(−σ2 , −σ1 , iσ3 , −σ2 , −σ1 , −iσ3 ) if the σ a AḂ are given by (1.37).
(1.79)
As we shall see in Section 1.3, the matrices (Sab A B ) and (Sab Ȧ Ḃ ) span two irreducible
representations of the Lie algebra of SO(p, q) [see, e.g., (1.124) and (1.158)]. Note,
however, that only when the metric has Lorentzian signature are the sets of 2 × 2
matrices {(Sab A B )} and {(Sab Ȧ Ḃ )}, with 1 a < b 4, linearly independent over R
[see (1.76)].
In the case of a totally antisymmetric three-index tensor tabc , (1.44) and (1.58)
give
1 1
tabc = − √ σa AȦ σb BḂ σcCĊ (tBȦ εAC εḂĊ − tAḂεBC εȦĊ ) = − √ σ̌abc AḂtAḂ ,
2 2 2
1.3 Spinorial Representation of the Orthogonal Transformations 17
where
σ̌abc AḂ ≡ σ[a RḂ σb AṠ σc]RṠ . (1.80)
A convenient alternative expression of the symbols σ̌abc AḂ is obtained as follows.

Making use of (1.40) and (1.64), one obtains
1 abcd RḂ AṠ 1
ε σa σb σcRṠ = − ε abce σa RḂ σb AṠ σcRṠ σe PQ̇ σ d PQ̇
6 12
1
= − ε RḂAṠ RṠ PQ̇ σ d PQ̇
3
1
= − iq (ε R R ε AP ε ḂQ̇ ε Ṡ Ṡ − ε RP ε A R ε Ḃ Ṡ ε ṠQ̇ )σ d PQ̇
3
= −iq σ dAḂ , (1.81)
i.e.,
1 abcd
ε σ̌abc AḂ = −iq σ dAḂ . (1.82)
6
Furthermore, from (1.58) we see that t A ȦB Ḃ AḂ = 3tBȦ . Hence, using (1.42),
1 1
tBȦ = √ σ aA Ȧ σ b B Ḃ σ c AḂtabc = √ σ̌ abc BȦtabc . (1.83)
6 2 6 2
By combining the preceding formulas, one finds that
σ̌ abc BȦ σ̌abc QṖ = −12δBQ δȦṖ ,
which, owing to (1.82), is equivalent to (1.21).
1.3 Spinorial Representation of the Orthogonal Transformations
As we shall show in this section, with the aid of the spinor algebra, one can find
simple explicit expressions for all the orthogonal transformations in terms of 2 × 2
matrices that act on the one-index spinors. We shall also show that any orthogonal
transformation with negative determinant can be expressed in terms of two vectors.
According to the definition (1.42) and (1.21), the spinor equivalent of the metric
tensor gab is
1 a
σ σ b gab = −εAB εȦḂ ;
2 AȦ BḂ
therefore, (1.4) is equivalent to
εAB εȦḂ = LCĊ AȦ LDḊ BḂ εCD εĊḊ , (1.84)

18 1 Spinor Algebra
where LAȦ BḂ is the spinor equivalent of La b . Condition (1.84) implies, in particular,
that
LCĊ 11̇ LDḊ 11̇ εCD εĊḊ = 0,
or equivalently,
LCĊ 11̇ LCĊ11̇ = 0. (1.85)
In order to determine the consequences of this relation we shall establish the fol-
lowing useful result.
Proposition 1.3. If ψAḂ satisfies ψAḂ ψ AḂ = 0, then there exist αA and βȦ such that
ψAḂ = αA βḂ . (1.86)
Proof. Making use of the rules (1.22), one finds that ψAḂ ψ AḂ = 2ψ11̇ ψ22̇ − 2ψ12̇ψ21̇
= 2 det(ψAḂ ); therefore ψAḂ ψ AḂ = 0 if and only if the 2 × 2 matrix (ψAḂ ) has van-
ishing determinant, which means that its rows are linearly dependent and that its
columns are linearly dependent, and hence ψAḂ must be of the form (1.86) (see also
Exercise 1.5).

Thus, from (1.85) we conclude that
LCĊ 11̇ = α C β Ċ , (1.87)
for some α A , β Ȧ . In an entirely analogous manner one finds that
LCĊ 22̇ = γ C δ Ċ , (1.88)
for some γ A , δ Ȧ . Substituting (1.87) and (1.88) into LCĊ 11̇ LDḊ 22̇ εCD εĊḊ = 1, which
follows from (1.84), one obtains the condition
(α A γA )(β Ḃ δḂ ) = 1. (1.89)
This condition implies that the two sets {α A , γ A } and {β Ȧ , δ Ȧ } are linearly indepen-
dent. (Indeed, if one assumes, for instance, that γ A = λ α A , then α A γA = λ α A αA =
0.) Furthermore, making use of (1.27), we see that
(αA γB − αB γA )(βȦ δḂ − βḂδȦ ) = (α R γR )εAB (β Ṡ δṠ )εȦḂ = εAB εȦḂ . (1.90)
On the other hand, the spinor equivalent of (1.7) is given by
εAB εȦḂ = LAȦCĊ LBḂ DḊ εCD εĊḊ . (1.91)
Then substituting (1.90) into the left-hand side of (1.91), and (1.87) and (1.88) into
the right-hand side, after some simplifications one obtains
αA δȦ γB βḂ + γA βȦ αB δḂ = LAȦ21̇ LBḂ12̇ + LAȦ12̇ LBḂ21̇ . (1.92)

Using the fact that the sets {α A , γ A } and {β Ȧ , δ Ȧ } are linearly independent we can
expand LAȦ12̇ and LAȦ21̇ in the form
LAȦ12̇ = a1 αA βȦ + a2αA δȦ + a3 γA βȦ + a4 γA δȦ ,

(1.93)
LAȦ21̇ = b1 αA βȦ + b2αA δȦ + b3 γA βȦ + b4 γA δȦ ,
where a1 , . . . , a4 , b1 , . . . , b4 are some (possibly complex) scalars. Then, substituting

(1.93) into (1.92), one finds that a1 = a4 = b1 = b4 = 0 and that only the following
two cases are possible:
1
(i) LAȦ12̇ = a2 αA δȦ , LAȦ21̇ = γ β , (1.94)
a2 A Ȧ
1
(ii) LAȦ12̇ = a3 γA βȦ , LAȦ21̇ = αA δȦ . (1.95)
a3
In the case (i), by defining

1/2 1/2
a2 1
K A1 ≡ α A, KA2 ≡ γ A,
α R γR a2 α R γR
1/2
α R γR 1/2 Ȧ
M Ȧ 1̇ ≡ β Ȧ , M Ȧ 2̇ ≡ a2 α R γR δ ,
a2
from (1.87), (1.88), and (1.94) we obtain
(i) LAȦ BḂ = K A B M Ȧ Ḃ , (1.96)
and
det(K A B ) = 1, det(M Ȧ Ḃ ) = 1 (1.97)
[see (1.89)]. In the case (ii), making
1/2 1/2
1 a3
K A 1̇ ≡ α A, K A 2̇ ≡ γ A,
a3 α R γR α R γR
1/2
1/2 Ȧ α R γR
M Ȧ 1 ≡ a3 α R γR β , M Ȧ 2 ≡ δ Ȧ ,
a3
we have
(ii) LAȦ BḂ = K A Ḃ M Ȧ B , (1.98)
with
det(K A Ḃ ) = 1, det(M Ȧ B ) = 1. (1.99)
Using (1.96) and (1.30) one finds that (1.84) reduces to the condition
det(K A B ) det(M Ȧ Ḃ ) = 1,
20 1 Spinor Algebra
which is satisfied as a consequence of (1.97), and similarly, substituting (1.98) into

(1.84), it follows that
det(K A Ḃ ) det(M Ȧ B ) = 1.
The determinant of the matrix (La b ) can be computed by means of the general
expression
εabcd La e Lb f Lc g Ld h = det(Lr s ) εe f gh (1.100)
[cf. (1.30)]. Hence, by virtue of (1.64),
(εAC εBD εȦḊ εḂĊ − εAD εBC εȦĊ εḂḊ )LAȦ E Ė LBḂ F Ḟ LCĊ GĠ LDḊ H Ḣ
= det(Lr s ) (εEG εFH εĖ Ḣ εḞ Ġ − εEH εFG εĖ Ġ εḞ Ḣ ),
and making use of (1.30) one finds that (1.96) is the spinor equivalent of an orthog-
onal transformation (La b ) with determinant equal to +1, while (1.98) is the spinor
equivalent of a matrix (La b ) with determinant equal to −1. The orthogonal transfor-
mations with positive [resp. negative] determinant are called proper [resp. improper]
transformations.
If σ a AḂ and σ
a AḂ are two sets of Infeld–van der Waerden symbols (satisfying
(1.21) with the same metric tensor gab ) and {e1 , e2 , e3 , e4 } is an orthonormal basis
of V as above, then
ea = − 12 σa AḂ σ b AḂ eb
is also an orthonormal basis of V (or of its complexification); in fact, using (1.21)
and (1.40),
g(ea , eb ) = 14 σa AḂ σ c AḂ σ
bCḊ σ d CḊ gcd
= − 12 σa AḂ σ
bCḊ εAC εḂḊ
= − 12 σaAḂ σb AḂ
= gab .
Hence, ea = Lb a eb for some 4 × 4 matrix (La b ) satisfying (1.4), and by virtue of
the definition of the ea , we have ea = − 12 Lb a σ
b AḂ σ c AḂ ec , that is, − 12 σ
b AḂ σ c AḂ =
(L−1 )c b = Lb c , or
σb AḂ = Lb c σc AḂ
= − 12 La c σ aCḊ σbCḊ σc AḂ
= −LCḊ AḂ σbCḊ .
Then, making use of (1.96) or (1.98), one proves that σ a AḂ and σ a AḂ must be related
as in (1.31) or (1.32).
According to the preceding results, any orthogonal transformation with positive
determinant, that is, any element of SO(p, q), is of the form
1
La b = σ a AȦ σb BḂ K A B M Ȧ Ḃ , (1.101)
2
where (K A B ) and (M Ȧ Ḃ ) belong to SL(2, C), the group formed by the unimodular
2 × 2 complex matrices with the usual matrix multiplication. Since for a 4 × 4 matrix
(La b ), det(La b ) = det(−La b ), any element of SO(p, q) can also be expressed as
1
La b = − σ a AȦ σb BḂ K A B M Ȧ Ḃ . (1.102)
2
This last expression is more convenient than (1.101) because when (K A B ) and
(M Ȧ Ḃ ) are both the identity 2 × 2 matrix, (La b ) given by (1.102) is the identity
4 × 4 matrix [see (1.40)]. However, in order for (La b ) to be real, the matrices (K A B )
and (M Ȧ Ḃ ) must be suitably restricted. Since the Infeld–van der Waerden symbols
are related to their complex conjugates in a way that depends on the signature of
the metric tensor, we need to deal with each case separately. Note in passing that
according to (1.21), from (1.102) we have
tr (La b ) = tr (K A B ) tr (M Ȧ Ḃ ). (1.103)
1.3.1 Euclidean Signature
When the metric of V has Euclidean signature, the connection symbols have been
chosen in such a way that σ a AḂ = −σ aAḂ [see (1.38)]; therefore, the entries (1.102)
are real if and only if
σ a AȦ σb BḂ K A B M Ȧ Ḃ = σ aAȦ σbBḂ K A B M Ȧ Ḃ .
Writing the left-hand side of this equation in the form σ aAȦ σbBḂ KA B MȦ Ḃ , one ob-
tains the equivalent condition
KA B MȦ Ḃ = K A B M Ȧ Ḃ .
Since the indices A, B, Ȧ, and Ḃ are independent of each other, from this last rela-
tion it follows that KA B = λ K A B and MȦ Ḃ = λ −1 M Ȧ Ḃ , for some scalar λ . Taking
into account the fact that the determinant of (K A B ) and that of its complex conju-
gate are equal to 1, it follows that λ 2 = 1. On the other hand, by virtue of (1.22),
det(K A B ) = K 1 1 K 2 2 − K 1 2 K 2 1 = −K 1 1 K1 1 − K 1 2 K1 2 = −K 1 1 λ K 1 1 − K 1 2 λ K 1 2 =
−λ (|K 1 1 |2 + |K 1 2 |2 ), which implies that λ must be equal to −1. Thus, we are left
with
KA B = −K A B (1.104)
and, similarly, MȦ Ḃ = −M Ȧ Ḃ . These conditions mean that (K A B ) and (M Ȧ Ḃ ) are

unitary. Indeed, from (1.30) and (1.22) we have, for an arbitrary 2 × 2 matrix (M A B ),
M AC MAD = det(M R S ) εCD , (1.105)

22 1 Spinor Algebra
and hence, raising the index D, MA D M AC = − det(M R S ) δCD , which means that the
inverse of (M A B ) has entries
1
(M −1 )D A = − M D (1.106)
det(M R S ) A
(assuming, of course, that det(M R S ) = 0). In particular, for a unimodular 2 × 2
matrix,
(M −1 )D A = −MA D (1.107)
[cf. (1.6)]. Hence, the condition (1.104) amounts to (K −1 )B A = K A B , thus showing

that (K A B ) is unitary.
In summary, any SO(4) transformation is of the form (1.102), where (K A B ) and
(M Ȧ Ḃ ) belong to SU(2), the group of unimodular unitary 2 × 2 matrices with ma-

trix multiplication. However, the two pairs of SU(2) matrices (K A B ), (M Ȧ Ḃ ) and

−(K A B ), −(M Ȧ Ḃ ) correspond to the same SO(4) transformation. The mapping
from SU(2) × SU(2) onto SO(4) given by (1.102) is a group homomorphism whose
kernel is {(I, I), (−I, −I)}, where I denotes the 2 × 2 unit matrix; hence, we can
write
SO(4) SU(2) × SU(2) /Z2 , (1.108)
where Z2 denotes the group with two elements, which is identified here with
{(I, I), (−I, −I)}. In effect, if (La b ) is given by (1.102) and
a b = − 1 σ a σb BḂ K
L A B M
Ȧ Ḃ ,
2 AȦ
then the product of the matrices (

La b ) and (La b ) is given by
A M
La c Lc b = 14 σ a AȦ σcCĊ K Ȧ σ c DḊ σ BḂ K D M Ḋ Ḃ
C Ċ b B
A K C M
= − 12 σ a AȦ σb BḂ K Ȧ MĊ Ḃ ,
C B Ċ
which is the SO(4) transformation corresponding to the pair of SU(2) matrices

MM),
(KK, where K = (K A B ), K = (K A ), M = (M A ), M = (M A ) [recall that
B B B
the group operation in SU(2) × SU(2) is defined by (K, M)(K,
MM)].
M) = (KK,
The group SU(2), whose underlying manifold can be identified with the unit
sphere S3 , is connected and so is SU(2) × SU(2); hence, SU(2) × SU(2) is a double
covering group of SO(4).
The existence of a homomorphism between SO(4) and SU(2) × SU(2) is a well-
known result, which is often obtained by considering the corresponding Lie algebra
homomorphism. If the 4 × 4 matrices (La b ) = exp t(ω a b ), for t ∈ R, are orthogonal,
then the entries of the matrix (ω a b ) must satisfy
ωab = −ωba .
(This condition can be obtained by substituting (La b ) = exp t(ω a b ) into (1.4) and
taking the derivative of both sides of the equation with respect to t at t = 0.) There-
fore, the 4 × 4 real matrices Jab with entries
(Jab )c d ≡ δac gbd − δbc gad (1.109)
generate so(4), the Lie algebra of SO(4) (note that Jab = −Jba ). (Actually, expres-
sions (1.109) generate the Lie algebra of SO(p, q) if the entries gab appearing on
the right-hand side are those of the metric with signature (p, q).) The commutators
between these matrices are readily found to be
[Jab , Jcd ] = gac Jdb − gbc Jda + gad Jbc − gbd Jac , (1.110)
or, making use of the definitions
L1 ≡ −J23, L2 ≡ −J31 , L3 ≡ −J12 ,
and
M1 ≡ J14 , M2 ≡ J24 , M3 ≡ J34 ,
one finds that the commutation relations (1.110) are equivalent to
[Li , L j ] = εi jk Lk , [Li , M j ] = εi jk Mk , [Mi , M j ] = εi jk Lk ,
where the lowercase Latin indices i, j, k range from 1 to 3 and there is summation
over repeated indices. Finally, letting
Ai ≡ 12 (Li + Mi ), Bi ≡ 12 (Li − Mi ),
one finds that
[Ai , A j ] = εi jk Ak , [Ai , B j ] = 0, [Bi , B j ] = εi jk Bk .
Thus, the sets of 4 × 4 matrices {A1 , A2 , A3 } and {B1 , B2 , B3 }, which together form
a basis of so(4), are bases of Lie algebras isomorphic to that of SU(2). Making use
of the definitions above, (1.69) and (1.76), one finds that the spinor equivalents of
the matrices Ai and Bi are
1 1
(Ai )AḂCḊ = − εi jk δḊḂ S jk AC , (Bi )AḂCḊ = − εi jk δCA S jk Ḃ Ḋ .
4 4
That is, each matrix Ai or Bi is the identity transformation on one of the spin spaces.
We shall not follow this approach, based on the Lie algebras, centering our attention
on the groups directly. However, the connection with the Lie algebras will arise
naturally in what follows [see, e.g., (1.124)].
The SU(2) matrices can be parameterized in several convenient ways (see also
(1.156) below). For instance, given (K A B ) ∈ SU(2), we can look for its eigenspinors.
The unitarity of (K A B ) implies that the eigenvalues of (K A B ) are of the form eiα , for
24 1 Spinor Algebra
some α ∈ R, and the fact that det(K A B ) = 1 implies that the product of the two
eigenvalues of (K A B ) must be equal to 1; therefore, there exist αA and βA such that
K A B α B = eiθ /2 α A , K A B β B = e−iθ /2 β A , (1.111)
where the factor 1/2 has been introduced for convenience. Hence [see (1.24) and
(1.28)],
1 iθ /2 A
KAB = e α βB − e−iθ /2 β A αB
α Rβ R
1
= R cos 12 θ (α A βB − β A αB ) + i sin 12 θ (α A βB + β A αB )
α βR
α A βB + β A αB
= cos 12 θ δBA + i sin 12 θ . (1.112)
α R βR
As a consequence of the relation (1.104), the one-index spinors αA and βA are

related to each other. Indeed, taking the complex conjugate of the first equation in
(1.111), we have K A B α B = e−iθ /2 α A , hence −KA B α B = e−iθ /2 α A , and, introducing
the definition
α
C ≡ α C (1.113)
(i.e., α
1 = −α2 , α B = e−iθ /2 α
2 = α1 ) we obtain −KA B α A , or equivalently, K A B α
B =
e−iθ /2 α
A . By comparing with the second equation in (1.111) one concludes that
βA = λα A , for some λ ∈ C. Then, (1.112) can be written in the form
α Aα
B + α
A αB
K A B = cos 12 θ δBA + i sin 12 θ . (1.114)
α Rα
R
Note that α R α
R = |α1 |2 + |α2 |2 0, and therefore, if αA = 0, the set {αA , α
A } is
linearly independent. Equation (1.114) yields
K A A = 2 cos 12 θ . (1.115)
Thus, any SO(4) transformation is given by (1.102) with (K A B ) and (M Ȧ Ḃ ) being

two SU(2) matrices. The matrix (K A B ) can be expressed in the form (1.114), and in
a similar manner, there exists γ Ȧ such that
1
M Ȧ Ḃ = (eiϕ /2 γ Ȧ γḂ − e−iϕ /2 γȦ γḂ )
γ Ṙ γṘ
γ Ȧ γḂ + γȦγḂ
= cos 12 ϕ δḂȦ + i sin 12 ϕ , (1.116)
γ Ṙ γṘ
for some ϕ ∈ R, where

γĊ ≡ γ Ċ . (1.117)
Then, M Ȧ Ḃ γ Ḃ = eiϕ /2 γ Ȧ , M Ȧ Ḃ γḂ = e−iϕ /2 γȦ , and
M Ȧ Ȧ = 2 cos 12 ϕ . (1.118)
The vector equivalents of α A γ Ȧ , α

A γȦ , α A γȦ , and α A γ Ȧ are (complex) eigen-
vectors of (L b ) with eigenvalues e
a i( θ + ϕ )/2 ,e −i( θ + ϕ )/2 , ei(θ −ϕ )/2 , and e−i(θ −ϕ )/2 ,
respectively.
Classification of the SO(4) Transformations
The SO(4) transformations can be classified according to whether the traces K A A

and M Ȧ Ȧ coincide or not. In view of (1.115) and (1.118), K A A = M Ȧ Ȧ if and only
if eiθ /2 = eiϕ /2 or eiθ /2 = e−iϕ /2 . Hence, the traces K A A and M Ȧ Ȧ coincide if and
only if two eigenvalues of (La b ) are equal to 1. As is well known, any SO(3) trans-
formation has at least one eigenvalue equal to 1, which means that there exists a
one-dimensional subspace (the axis of the rotation) pointwise invariant under the
transformation. The SO(4) transformations (La b ) such that K A A = M Ȧ Ȧ (which will
be called simple) are analogous to the SO(3) transformations in the sense that there
exists a proper subspace pointwise invariant under (La b ).
Assuming α R α R = 1 = γ Ṙ γṘ , we define
1 i
vAȦ = √ (α A γ Ȧ − α
A γȦ ), wAȦ = √ (α A γ Ȧ + α
A γȦ ). (1.119)
2 2
Then, one can readily verify that vAȦ and wAȦ are the spinor equivalents of two
mutually orthogonal real unit vectors va , wa . When eiθ /2 = eiϕ /2 we obtain
K A B M Ȧ Ḃ vBḂ = cos θ vAȦ + sin θ wAȦ ,

K A B M Ȧ Ḃ wBḂ = − sin θ vAȦ + cos θ wAȦ ,
which means that the corresponding orthogonal transformation is a rotation through

the angle θ in the plane spanned by {va , wa }. In a similar manner, when eiθ /2 =
e−iϕ /2 , the SO(4) transformation defined by the pair of SU(2) matrices (1.114) and
(1.116) is a rotation through the angle θ in the two-dimensional plane generated by
the vector equivalents of
1 i
√ (α A γȦ + α
A γ Ȧ ), √ (α A γȦ − α
A γ Ȧ )
2 2
(which are real and orthogonal to va and wa ).
Since θ = 12 (θ + ϕ ) + 12 (θ − ϕ ) and ϕ = 12 (θ + ϕ ) − 12 (θ − ϕ ), any SO(4)
transformation, which is defined by a pair of matrices of the form (1.114) and
(1.116), is the composition of two simple SO(4) transformations, one of them rep-
resenting a rotation through 12 (θ + ϕ ) in a two-dimensional plane and the other
26 1 Spinor Algebra
representing a rotation through 12 (θ − ϕ ) in the orthogonal complement of the first

two-dimensional plane.
The SU(2) matrices (K A B ) and (M Ȧ Ḃ ) can be expressed in terms of the unit vec-
tors va , wa , introduced above. Making use of the definitions (1.119) and (1.69),
we have
va wb SabAB = va wb σa(A Ṙ σ|b|B)Ṙ
= 2v(A Ṙ wB)Ṙ
= i(αA α
B + α
A αB ), (1.120)
and therefore (1.114) can be written as
K A B = cos 12 θ δBA + sin 12 θ va wb Sab A B . (1.121)
In a similar way, we have
va wb SabȦḂ = va wb σa R (Ȧ σ|bR|Ḃ)

= 2vR (Ȧ w|R|Ḃ)
= i(γȦ γḂ + γȦ γḂ );
hence, (1.116) is equivalent to
M Ȧ Ḃ = cos 12 ϕ δḂȦ + sin 12 ϕ va wb Sab Ȧ Ḃ .
With α R α
R = 1, the square of the matrix (α A α B + α A αB ) is the 2 × 2 identity
matrix,
(α A α
C + α
A αC )(α C α
B + α
C αB ) = α A α
B − α A αB = δBA ,

and therefore, from (1.120) it follows that the powers of the matrix va wb Sab A B are
given by
a b A 2k a b A 2k+1
v w Sab B = (−1)k (δBA ), v w Sab B = (−1)k va wb Sab A B
for k = 0, 1, 2, . . . , and making use of the usual series expansion for the exponential,
we obtain the matrix equality
1
exp 2θ va wb Sab A B = cos 12 θ δBA + sin 12 θ va wb Sab A B .
Thus, comparing with (1.121), we conclude that

(K A B ) = exp 12 θ va wb Sab A B ,
or equivalently,
(K A B ) = exp iθ 12 (α A α
B + α
A αB ) . (1.122)
In an analogous manner, we find that

1
(M Ȧ Ḃ ) = exp 2ϕ va wb Sab Ȧ Ḃ ,
or
(M Ȧ Ḃ ) = exp iϕ 12 (γ Ȧ γḂ + γȦ γḂ ) . (1.123)
These expressions show that any SU(2) matrix is the exponential of some (traceless)
matrix.
We can get expressions for (K A B ) and (M Ȧ Ḃ ) that show more similarities. Indeed,
making use of (1.76), we have

θ va wb Sab A B = 12 (θ + ϕ ) va wb − 12 (θ − ϕ ) 12 vc wd ε cdab Sab A B
and
1
ϕ va wb Sab Ȧ Ḃ = 2 (θ + ϕ ) va wb − 12 (θ − ϕ ) 12 vc wd ε cdab Sab Ȧ Ḃ ,
and therefore, we can also write
1
(K A B ) = exp 4t
ab S A,
ab B
(1.124)
(M Ȧ Ḃ ) = exp 14 t ab Sab Ȧ Ḃ ,
where we have introduced the bivector
t ab ≡ (θ + ϕ ) v[a wb] − (θ − ϕ ) 12 ε abcd vc wd . (1.125)
Making use of (1.52) and (1.65), one finds that the spinor equivalent of the bivector
tab is
tAȦBḂ = iθ α(A α
B) εȦḂ + iϕ γ(Ȧ γḂ) εAB . (1.126)
As shown below, the spinor equivalent of any real bivector is of the form (1.126)
[see (1.232)]; therefore, any real bivector is of the form (1.125), and for any real
bivector tab , equations (1.124) define an SO(4) transformation.
Any matrix whose spinor equivalent is of the form τ A B ε Ȧ Ḃ commutes with any
matrix whose spinor equivalent is of the form ε A B τ Ȧ Ḃ ; therefore

exp (τ A B ε Ȧ Ḃ ) + (ε A B τ Ȧ Ḃ ) = exp τ A B ε Ȧ Ḃ exp ε A B τ Ȧ Ḃ ,
and taking τAB = iθ α(A α

B) , τȦḂ = iϕ γ(Ȧ γḂ) , from (1.122) and (1.123) we have

exp (τ A B ε Ȧ Ḃ ) + (ε A B τ Ȧ Ḃ ) = M A B K Ȧ Ḃ ,
which implies that the SO(4) transformation corresponding to the pair of SU(2)
matrices (1.124) can be also expressed as
(La b ) = exp(−t a b ). (1.127)

28 1 Spinor Algebra
Thus, any SO(4) transformation can be written as the exponential of some matrix.
Furthermore, making use of Proposition 1.2 and the preceding results one can
verify that the bivector t ab is simple if and only if the SO(4) transformation
(La b ) = exp(−t a b ) is simple.
Improper O(4) Transformations
Following a procedure similar to that leading to (1.104), one concludes that any
O(4) transformation with negative determinant can be expressed in the form
1
La b = − σ a AȦ σb BḂ K A Ḃ M Ȧ B , (1.128)
2
where (K A Ḃ ) and (M Ȧ B ) are SU(2) matrices; given (La b ), the matrices (K A Ḃ ) and
(M Ȧ B ) are defined up to a common sign.
The matrices (K A Ḃ ) and (M Ȧ B ) appearing in (1.128) can be regarded as the spinor
equivalents of two vectors. Owing to (1.104) and (1.47), these two vectors are pure
imaginary. Hence, letting
√ √
KAḂ = i 2 vAḂ , MȦB = i 2 wBȦ , (1.129)
one finds that det(K A Ḃ ) = 12 K AḂ KAḂ = −vAḂvAḂ = 1 and, similarly, −wAḂ wAḂ = 1;
therefore, the vector equivalents va and wa of vAḂ and wAḂ , respectively, are real unit
vectors [see (1.46)]. Then, making use of (1.27) we obtain the identity
vAḂ wBȦ = 12 (vAḂ wBȦ − vAȦwBḂ + vAȦwBḂ

+ vAḂwBȦ − vBḂwAȦ + vBḂwAȦ )
= 12 (−εȦḂ vA Ṙ wBṘ + vAȦ wBḂ + εABvR Ḃ wRȦ + vBḂwAȦ )

= 12 − εȦḂ (v(A Ṙ wB)Ṙ + 12 εAB vSṘ vSṘ ) + vAȦwBḂ

+ εAB (vR (Ḃ w|R|Ȧ) + 12 εḂȦ vRṠ wRṠ ) + vBḂwAȦ
= − 12 (vRṠ wRṠ )εAB εȦḂ + 12 (vAȦ wBḂ + vBḂwAȦ )
− 12 (v(A Ṙ wB)Ṙ εȦḂ − vR (Ȧ w|R|Ḃ) εAB ). (1.130)
According to (1.52) and (1.65), −(v(A Ṙ wB)Ṙ εȦḂ − vR (Ȧ w|R|Ḃ) εAB ), which appears in
the last line of this identity, is the spinor equivalent of the dual of the antisym-
metrized product va wb − vb wa . Hence, substituting (1.129) and (1.130) into (1.128),
we obtain
La b = −vc wc δba + vawb + wa vb + gaeεebcd vc wd , (1.131)
which is, therefore, the tensor equivalent of LAȦ BḂ = 2vA Ḃ wB Ȧ .
Thus, for each improper orthogonal transformation there exist two (not nec-
essarily distinct) unit vectors va , wa , defined up to a common sign, such that
(1.131) holds. Expression (1.131) shows that if {va , wa } is linearly independent,

there are two two-dimensional subspaces of V invariant under the orthogonal trans-
formation (La b ); one of them is the subspace spanned by {va , wa } (note that
La b vb = wa and La b wb = va ), and the other is formed by the vectors simultaneously
orthogonal to va and wa . When {va , wa } is linearly dependent (i.e., wa = ∓va ), the
last term in (1.131) vanishes, and (La b ) corresponds to the reflection on the hyper-
plane orthogonal to va or to the negative of this reflection (see Section 1.4).
Conversely, given two real unit vectors va , wa , equation (1.131) yields an im-
proper orthogonal transformation.
1.3.2 Lorentzian Signature
In the case that the metric has Lorentzian signature, the connection symbols have
been chosen in such a way that σ a AḂ = σ a BȦ ; therefore, the unimodular matrix (La b )
given by (1.102) is real if and only if
σ a AḂ σbCḊ K AC M Ḃ Ḋ = σ a BȦ σb DĊ K AC M Ḃ Ḋ ,
which is equivalent to K AC M Ḃ Ḋ = K B D M ȦĊ . Hence, M ȦĊ = λ K AC and K B D =

λ −1 M Ḃ Ḋ , for some scalar λ . Since both matrices (K A B ) and (M Ȧ Ḃ ) have unit deter-
minant, we have λ 2 = 1, so that finally we obtain the condition
M Ȧ Ḃ = ±K A B . (1.132)
Thus, any SO(3, 1) transformation can be expressed in the form
La b = ∓ 12 σ a AḂ σbCḊ K AC K B D , (1.133)
with (K A B ) ∈ SL(2, C). The SO(3, 1) transformations of the form (1.133) with the
upper sign form a subgroup of SO(3, 1) denoted by SO↑ (3, 1) or SO(3, 1)0 (being
the connected component of SO(3, 1) containing the identity). Furthermore, the
mapping
(K A B ) → (La b ), with La b = − 12 σ a AḂ σbCḊ K AC K B D , (1.134)
is a two-to-one homomorphism of SL(2, C) onto SO↑ (3, 1). In fact, making use of
(1.21), one finds that the product of the SO↑ (3, 1) matrices
La b = − 12 σ a AḂ σbCḊ K AC K B D , M a b = − 12 σ a AḂ σbCḊ QAC QB D
is given by
La b M b c = − 12 σ a AḂ σc GḢ K AC QC G K B D QD H ,
30 1 Spinor Algebra
which is the SO↑ (3, 1) matrix corresponding to the product of the matrices (K A B )
and (QA B ) under the mapping (1.134).
Thus, SO↑ (3, 1) SL(2, C)/Z2 , and since SL(2, C) is connected, SL(2, C) is a
double covering group of SO↑ (3, 1).
Making use of the explicit expressions for the Infeld–van der Waerden symbols
(1.33), from (1.133) one finds that

L4 4 = ± 12 |K 1 1 |2 + |K 1 2 |2 + |K 2 1 |2 + |K 22 |2 .
Therefore, SO↑ (3, 1) is formed by the orthogonal transformations with unit deter-
minant such that L4 4 > 0.
In the context of special relativity, the O(3, 1) transformations are identified with
the Lorentz transformations that can be defined as those relating the Cartesian space-
time coordinates measured by two inertial frames (see, e.g., Jackson 1975, Rindler
1977). The elements of SO↑ (3, 1) are called orthochronous proper Lorentz transfor-
mations; SO↑ (3, 1) is called the restricted Lorentz group. The SO(3, 1) transforma-
tions (1.133) with the lower sign produce a time inversion (L4 4 < 0) and a space
inversion.
Lorentz Boosts
The elementary Lorentz transformation

x = γ (x − vt),
y = y,
(1.135)
z = z,
ct = γ (ct − vx/c),
where c denotes the speed of light in vacuum and γ ≡ (1 − v2 /c2 )−1/2 , gives the
relationship between the Cartesian coordinates of an event measured by two inertial
reference frames, assuming that the primed reference frame S moves with respect
to the unprimed one, S, along the x-axis with velocity v. The Lorentz transformation
(1.135) is equivalent to xa = La b xb if we identify (x, y, z, ct) = (x1 , x2 , x3 , x4 ), and
similarly for the primed coordinates, with
a
v vb γ
L b = δb + (γ − 1) c − T Tb + (va Tb − T a vb ).
a a a
(1.136)
v vc c
Here the T a are the components of a (timelike) vector such that T a Ta = −1, va is
a (spacelike) vector orthogonal to T a , va Ta = 0, which gives the relative velocity
of S with respect to S, and γ = (1 − va va /c2 )−1/2 . In fact, it can readily be seen
that (1.136) reduces to (1.135) when (T a ) = (0, 0, 0, 1) and (va ) = (v, 0, 0, 0). The
matrix (La b ) given by (1.136) represents an orthogonal transformation and therefore
must be of the form (1.133). (It is customary to define x0 = ct, with the tensor indices
a, b, . . . running from 0 to 3; we do not follow here that convention in order to employ
a uniform convention for all signatures of the metric.)
As is well known, it is convenient to write (va va )1/2 = c tanh ζ , for some ζ ∈ R

(called rapidity); then γ = cosh ζ , and (1.136) amounts to

va vb
L b = δb + (cosh ζ − 1)
a a
−T a
+ Tb
(vc vc )1/2 (vc vc )1/2
a
v Tb − T a vb
− (cosh ζ − 1 − sinh ζ )
(vc vc )1/2
= δba − 2(cosh ζ − 1) nalb − (cosh ζ − 1 − sinh ζ )(l a nb − nalb )
= δba − (e−ζ − 1) l a nb − (eζ − 1) nalb , (1.137)
where we have defined

1 va 1 va
l ≡ √ T + c 1/2 ,
a a
n ≡ √ T − c 1/2 .
a a
2 (v vc ) 2 (v vc )
By virtue of the definitions of T a and va , the vectors l a and na satisfy
l a la = 0 = na na , l a na = −1. (1.138)
As can readily be seen, the null vectors l a and na are eigenvectors of (La b ) with
eigenvalues e−ζ and eζ , respectively.
Then, for instance, l AḂ , the spinor equivalent of the null vector l a , satisfies
l lAḂ = 0, and according to (1.86), there exist αA and γȦ such that lAḂ = αA γḂ .
A Ḃ
Since the components l a are real, lAḂ = lBȦ , which implies that γȦ = λ αA , for some
λ ∈ R. Hence, absorbing the absolute value of λ into αA , it follows that
lAḂ = ±αA α Ḃ , (1.139)
where
α Ḃ ≡ αB . (1.140)
(An explicit proof is given below; see (1.161).)
According to the conventions (1.34) and (1.44), the sign in (1.139) is positive
[resp. negative] if the component l 4 is positive [resp. negative]. (From (1.34), (1.44),
(1.139), and (1.140) we have l 4 = 2−1/2 (l 11̇ + l 22̇ ) = ±2−1/2(|α 1 |2 + |α 2 |2 ).) In a
similar manner, there exists βA such that nAḂ = ±βA β Ḃ . Taking
lAḂ = αA α Ḃ , nAḂ = βA β Ḃ , (1.141)
from the last condition in (1.138) we obtain 1 = l AḂ nAḂ = α A α Ḃ βA β Ḃ = |α A βA |2 .

Since expressions (1.141) are left unchanged if we multiply αA or βA by any phase
factor, we can assume that
α A βA = 1. (1.142)
32 1 Spinor Algebra
Making use of (1.141), (1.28), and (1.142), one finds that the spinor equivalent of
(1.137) is given by

LAȦ BḂ = −δBA δḂȦ − e−ζ − 1 α A α Ȧ βB β Ḃ − eζ − 1 β A β Ȧ αB α Ḃ

= − α A βB − β A αB α Ȧ β Ḃ − β Ȧ α Ḃ − e−ζ − 1 α A α Ȧ βB β Ḃ

− eζ − 1 β A β Ȧ αB α Ḃ

= − e−ζ /2 α A βB − eζ /2β A αB e−ζ /2 α Ȧ β Ḃ − eζ /2 β Ȧ α Ḃ
= −K A B K A B , (1.143)
where

K A B = ± e−ζ /2 α A βB − eζ /2 β A αB

= ± cosh 12 ζ δBA − sinh 12 ζ (α A βB + β AαB ) (1.144)
[cf. (1.112)]. One can verify directly that det(K A B ) = 1. The rapidity is determined
by the absolute value of the trace of (K A B ), |K A A | = 2 cosh 12 ζ .
The direction of the velocity of S with respect to S, represented by va , is hidden
in the definitions of l a , na , and of the one-index spinors αA , βA appearing in (1.141)
and (1.144); however, it may be noticed that by virtue of the antisymmetry of SabAB
on the tensor indices a, b, we have [see (1.69), (1.141), and (1.142)]
(vc vc )−1/2 va T b SabAB = l a nb SabAB = l a nb σa(A Ṙ σ|b|B)Ṙ = 2l(A Ṙ nB)Ṙ

= 2α(A α Ṙ βB) β Ṙ = 2α(A βB) ; (1.145)
thus, expression (1.144) is equivalent to

K A B = ± cosh 12 ζ δBA − sinh 12 ζ ua T b Sab A B , (1.146)
where ua ≡ va /(vc vc )1/2 . Hence, with T a = δ4a as above, making use of (1.77),

cosh 12 ζ + u3 sinh 12 ζ (u1 + iu2 ) sinh 12 ζ
(K B ) = ±
A
. (1.147)
(u1 − iu2 ) sinh 12 ζ cosh 12 ζ − u3 sinh 12 ζ
The complex conjugate of the matrix (1.147) can be expressed as

K = (K A B ) = ± (cosh 12 ζ )I + (sinh 12 ζ )u j σ j ,
where I is the 2 × 2 identity matrix, and the σi are the Pauli matrices (i, j, . . . =
1, 2, 3). Then, the components of the unit vector ui are given in terms of K by
tr (K σi ) = ±2(sinh 12 ζ )ui .

As a consequence of (1.142), the square of the matrix α A βB + β A αB =
a b A
u T Sab B is the identity; therefore the matrix (1.146) can be expressed as

(K A B ) = ± exp − 12 ζ ua T b Sab A B = ± exp − 12 ζ l a nb Sab A B . (1.148)
As we shall see below, this last equation implies that the boost (1.136), (1.137) can
be written in the form
(La b ) = exp ζ (l a nb − na lb )
[see (1.159)], which, among other things, shows the additive character of the rapidity
ζ for boosts in the same direction.
Classification of the Orthochronous Proper Lorentz Transformations
The matrix (1.144) does not represent the most general SL(2, C) matrix. If (K A B ) is
an arbitrary SL(2, C) matrix, we can write KAB = 12 (KAB − KBA ) + 12 (KAB + KBA ) =
2 K C εAB + K(AB) , i.e., KAB = νεAB + μAB , where ν is some complex scalar and
1 C
μAB = μBA . As shown below [see (1.227)], the fundamental theorem of algebra im-
plies the existence of αA , βA , such that μAB = α(A βB) . Then α A and β A are eigen-
spinors of (K A B ); in fact,

K A B α B = νδBA + 12 (α A βB + β AαB ) α B = (ν + 12 α B βB )α A ,
and K A B β B = (ν − 12 α B βB )β A ; thus, if {α A , β A } is linearly independent, then
K ABα B = λ α A, K A B β B = λ −1 β A , (1.149)
for some λ ∈ C, since the determinant of (K A B ) is equal to 1. The contraction α A βA
is different from zero; hence, by rescaling α A or β A , we can impose the condition
α A βA = 1; then (1.149) are equivalent to
K A B = λ α A βB − λ −1β A αB (1.150)
[cf. (1.144)].
From (1.134) and (1.149) it follows that the products α A α Ȧ , β A β Ȧ , α A β Ȧ , and
β α Ȧ are the spinor equivalents of eigenvectors of (La b ) with eigenvalues
A
|λ |2 , |λ |−2 , λ /λ , λ /λ , (1.151)
respectively; α A α Ȧ , β A β Ȧ are the spinor equivalents of two real null vectors

[see (1.139)], while α A β Ȧ and β A α Ȧ are the spinor equivalents of two complex-
conjugate null vectors. Thus, if (K A B ) is of the form (1.150), then (La b ) pos-
sesses two real null eigenvectors, and the vector equivalents of α A β Ȧ + β A α Ȧ and
i(α A β Ȧ − β A α Ȧ ), which are real, span a two-dimensional invariant subspace on
which (La b ) is a rotation through the angle arg(λ /λ ).
Two eigenvalues of (La b ) are equal to 1 if and only if |λ | = 1 or λ ∈ R [see
(1.151)]. The orthochronous proper Lorentz transformations of this type, which
leave pointwise invariant a two-dimensional subspace, are called simple (Lounesto
1997, Sec. 9.6). When |λ | = 1, (La b ) represents a rotation; the vector equivalents
of α A α Ȧ and β A β Ȧ span a real two-dimensional subspace pointwise invariant under
34 1 Spinor Algebra
(La b ), the trace K A A is real, and |K A A | < 2. When λ is real, (La b ) represents a
boost, the vector equivalents of α A β Ȧ + β A α Ȧ and i(α A β Ȧ − β A α Ȧ ) span a real
two-dimensional subspace pointwise invariant under (La b ), the trace K A A is real,
and |K A A | > 2.
On the other hand, it can be readily verified that K A A = λ + λ −1 is real if and
only if |λ | = 1 or λ ∈ R. Therefore, the SL(2, C) matrix (1.150) corresponds to a
simple orthochronous proper Lorentz transformation if and only if the trace K A A is
real.
In the remaining case, in which {α A , β A } is linearly dependent, we have
K A B = νδBA + μα A αB ,
where ν and μ are two complex numbers. Then, the condition det(K A B ) = 1 gives
ν = ±1, and by rescaling αA we can write
K A B = ±(δBA − 12 ζ α A αB ), (1.152)
where ζ is a real number. Now α A is the only eigenspinor of (K A B ), and α A α Ȧ

is the spinor equivalent of a real null eigenvector of the corresponding orthogonal
transformation (La b ), with eigenvalue 1 (these Lorentz transformations are called
null rotations).
If γA is a one-index spinor such that α A γA = 1, then the vector equivalent of
i(α A γ Ȧ − γ A α Ȧ ) is a (spacelike) real eigenvector of (La b ) with eigenvalue 1; hence,
for the SO↑ (3, 1) transformation corresponding to (1.152) there is always a real
two-dimensional subspace pointwise invariant under (La b ) (that is, (La b ) is sim-
ple). (Note that γA is defined up to the transformation γA → γA + καA , with κ ∈ C
arbitrary, but the two-dimensional subspace pointwise invariant under (La b ) is not
affected by this ambiguity.) From (1.152) we see that K A A = ±2.
Substituting (1.152) into (1.102), with the aid of (1.28), we obtain

La b = − 12 σ a AȦ σb BḂ δBA − 12 ζ α A αB δḂȦ − 12 ζ α Ȧ α Ḃ

= δba − 12 σ a AȦ σb BḂ − 12 ζ α A α Ȧ (αB γ Ḃ + γBα Ḃ )

+ 12 ζ (α A γ Ȧ + γ A α Ȧ )αB α Ḃ + 14 ζ 2 α A α Ȧ αB α Ḃ
1 1
= δba + √ ζ (l a sb − sa lb ) − ζ 2 l a lb , (1.153)
2 4
√
where la and sa are the vector equivalents of αA α Ȧ and (1/ 2)(αA γ Ȧ + γA α Ȧ ),
respectively. The vectors la and sa are real and orthogonal to each other; la is null
and sa sa = 1. (A similar classification of the proper orthochronous Lorentz transfor-
mations, making use of quaternions, is obtained in Lanczos (1970); see also Penrose
and Rindler 1984, Hall 2004.)
The foregoing discussion proves the following result.
Proposition 1.4. The trace of the SL(2, C) matrix (K A B ) is real if and only if there
exists a real two-dimensional subspace pointwise invariant under the corresponding
SO↑ (3, 1) transformation (La b ). Furthermore, in that case, (La b ) is a boost or a

rotation if |K A A | > 2 or |K A A | < 2, respectively. When |K A A | = 2, with (K A B ) =
±(δBA ), (La b ) has only one real null eigenvector, and the corresponding eigenvalue
is equal to 1.
Since any nonzero complex number λ can be expressed as the product of a
positive real number and a complex number of unit modulus, λ = |λ | eiθ /2 , for
some θ ∈ R, from (1.149) it follows that the SO↑ (3, 1) transformation defined by
(1.150) is the composition of two simple orthochronous Lorentz transformations.
Furthermore, making use of (1.145), one finds that (1.150) can also be expressed in
the form
K A B = 12 (λ + λ −1) δBA + 12 (λ − λ −1) l a nb Sab A B , (1.154)
with la and na defined by (1.141). For instance, taking λ = eiθ /2 and T a = δ4a , we
have (see Exercise 1.9)

cos 12 θ − iu3 sin 12 θ −i(u1 + iu2 ) sin 12 θ
(K B ) =
A
, (1.155)
−i(u1 − iu2) sin 12 θ cos 12 θ + iu3 sin 12 θ
where now ua is a unit vector defining the axis of the rotation [cf. (1.147)]. Alterna-
tively, denoting by ma the vector equivalent of α A β Ȧ , we have
ma mb SabAB = ma mb σa(A Ṙ σ|b|B)Ṙ = 2α(A β Ṙ βB) α Ṙ = −2α(AβB)
[cf. (1.145)]; hence, (1.150) is also equivalent to
K A B = 12 (λ + λ −1) δBA − 12 (λ − λ −1 ) ma mb Sab A B .
With the conventions adopted here, the composition of a rotation about the z-axis
through an angle ϕ followed by a rotation about the y-axis through an angle θ ,
followed finally by a rotation about the z-axis through an angle χ , corresponds to
the matrix

e−iχ /2 0 cos 12 θ sin 12 θ e−iϕ /2 0
(K B ) =
A
0 eiχ /2 − sin 12 θ cos 12 θ 0 eiϕ /2

e−i(ϕ +χ )/2 cos 12 θ ei(ϕ −χ )/2 sin 12 θ
= (1.156)
−ei(χ −ϕ )/2 sin 12 θ ei(ϕ +χ )/2 cos 12 θ
and its negative. (Note the order of the factors.) As one can readily verify, this matrix
also belongs to SU(2), and this last expression gives a parameterization of the SU(2)
matrices by Euler angles (ϕ , θ , χ ).
It is convenient to express the eigenvalue λ of (K A B ) in the form λ = e−z/2 ;
then (K A B ) corresponds to a simple SO↑ (3, 1) transformation when z is real or pure
imaginary, and from (1.154) we also have
K A B = cosh 12 z δBA − sinh 12 z l a nb Sab A B .

36 1 Spinor Algebra
On the other hand, using the fact that the square of the matrix (l a nb Sab A B ) is the
identity [see (1.145)], we find that

(K A B ) = exp − 12 z l a nb Sab A B
and, by virtue of (1.76), letting z = x + iy,
z l a nb Sab A B = (x + iy)l a nb Sab A B

= (xlc nd − y 12 l a nb εabcd ) Scd A B
= − 12 tab SabA B ,
where tab is the real bivector
tab ≡ −2xl[a nb] + y εabcd l c nd . (1.157)
Thus, 1
(K A B ) = exp 4t
ab
Sab A B (1.158)
[cf. (1.124)]. Making use of the criterion given in Proposition 1.2, one can verify
that tab is simple if and only if x or y vanishes; therefore, the bivector (1.157) is
simple if and only if the SL(2, C) matrix (1.158) corresponds to a simple SO↑ (3, 1)
transformation. Since for the Lorentzian signature we have SabAB = SabȦḂ , it follows

that (M Ȧ Ḃ ) = exp 14 t ab Sab Ȧ Ḃ [cf. (1.124)]; hence, as in the foregoing subsection, it
follows that
(La b ) = exp(−t a b ). (1.159)
Finally, in the case of the null rotations,
1 1
√ l a sb SabAB = √ l a sb σa(A Ṙ σ|b|B)Ṙ = αA αB ,
2 2
√
where la and sa are the vector equivalents of αA α Ȧ and (1/ 2)(αA γ Ȧ + γA α Ȧ ),
respectively. Hence (1.152) can also be expressed in the form

1
K B = ± δB − √ ζ l s Sab B ,
A A a b A
2 2
or equivalently,
1
(K A B ) = ± exp − √ ζ l a sb Sab A B , (1.160)
2 2
since the square of the matrix (α A αB ) is equal
√ to zero. This expression is of the form
(1.158), where tab is the simple bivector − 2 ζ l[a sb] . As we have seen, (1.152) cor-
responds to a simple Lorentz transformation,
and therefore, in all cases, the SL(2, C)
matrix (K A B ) = exp 14 t ab Sab A B corresponds to a simple SO↑ (3, 1) transformation
if and only if the bivector tab is simple. (One can directly verify that in this case,
(1.159) reduces to (1.153).)
Note that in order to obtain (1.124), (1.158), and (1.160), the explicit form of the
matrices (Sab A B ) and (Sab Ȧ Ḃ ), given by (1.77) and (1.79), has not been required.
Parameterization of Null Vectors, Aberration of Light
By definition, a (real) null vector l a satisfies the condition l a la = 0 and therefore

1/2
l 4 = ± (l 1 )2 + (l 2 )2 + (l 3 )2 .
This means that apart from the ambiguity in the sign of l 4 , the components of the
null vector l a are determined by the point (l 1 , l 2 , l 3 ) ∈ R3 , which, in turn, can be
parameterized by the usual spherical coordinates (r, θ , ϕ ), according to
(l 1 , l 2 , l 3 ) = (r sin θ cos ϕ , r sin θ sin ϕ , r cos θ ).
Assuming l 4 > 0, the components of the spinor equivalent of l a are then given by
[see (1.34)]

l 11̇ l 12̇ r 1 − cos θ −eiϕ sin θ
= √
l 21̇ l 22̇ 2 −e−iϕ sin θ 1 + cos θ

√ sin2 21 θ −eiϕ sin 12 θ cos 12 θ
= 2r
−e−iϕ sin 12 θ cos 12 θ cos2 21 θ

√ −eiϕ /2 sin 12 θ −iϕ /2
= 2r −e sin 12 θ eiϕ /2 cos 12 θ
e −iϕ /2 cos 2 θ
1
1
α
= α 1̇ α 2̇ , (1.161)
α 2
which shows explicitly that the spinor equivalent of a real null vector with l 4 > 0 is
of the form l AȦ = α A α Ȧ [see (1.139)]. From (1.161) we see that
1
iϕ /2
α −e sin 1
θ
= 21/4 r1/2 e−iχ /2 2
, (1.162)
α2 e−iϕ /2 cos 1 θ 2
where χ is an arbitrary real number.

The wave four-vector of an electromagnetic plane wave ka is a future-pointing
null vector, that is, k4 > 0 (see, e.g., Jackson 1975), and therefore, its spinor equiv-
alent is of the form α A α Ȧ , with α A given by
1
α −eiϕ /2 sin 12 θ
=2 k1/4 1/2
, (1.163)
α2 e−iϕ /2 cos 12 θ
38 1 Spinor Algebra
where k = [(k1 )2 + (k2 )2 + (k3 )2 ]1/2 is the wave number, and θ and ϕ are the stan-
dard polar and azimuth angles of the direction of propagation of the wave with
respect to some inertial reference frame S. The spinor equivalent of ka the wave
four-vector of the wave with respect to a second inertial reference frame S , which
moves with velocity v with respect to S along the z-axis, is also of the form α A α Ȧ ,
with α A = K A B α B , and (K A B ) given by [see (1.147)]

cosh 1
ζ + sinh 1
ζ 0 e ζ /2 0
2 2
(K A B ) = = ,
0 cosh 12 ζ − sinh 12 ζ 0 e−ζ /2
with v = c tanh ζ ; hence, expressing α A in the form (1.163), we immediately obtain

k1/2 sin 12 θ eiϕ /2 = eζ /2 k1/2 sin 12 θ eiϕ /2 ,

k1/2 cos 12 θ e−iϕ /2 = e−ζ /2 k1/2 cos 12 θ e−iϕ /2 ,
which imply ϕ = ϕ and

1 ζ c+v
2θ = e tan 2θ = tan 12 θ
1
tan
c−v
(see also Misner, Thorne, and Wheeler 1973, Penrose and Rindler 1984). (Note that
θ is the angle between the direction of propagation of the wave and the direction of
motion of S with respect to S.) Furthermore, recalling that the angular frequency ω
of the wave is related to k by ω = ck, we have
2 2 v
ω = c eζ /2 k1/2 sin 12 θ + e−ζ /2 k1/2 cos 12 θ = γω 1 − cos θ ,
c
which gives the relativistic Doppler shift.
Improper Lorentz Transformations. Spatial Inversion
Proceeding as at the beginning of this subsection, one finds that the O(3, 1) transfor-
mations with determinant −1 are of the form
La b = ∓ 12 σ a AḂ σbCḊ K A Ḋ K BĊ , (1.164)
where (K A Ḃ ) ∈ SL(2, C). Making use of the explicit expression (1.33), from (1.164)
one finds that
L4 4 = ± 12 |K 1 1̇ |2 + |K 1 2̇ |2 + |K 2 1̇ |2 + |K 22̇ |2 ,
which shows that (La b ) preserves [resp. reverses] the time orientation if we take the
upper [resp. lower] sign in (1.164).
As in the case of Euclidean signature, the entries of the matrix (K A Ḃ ) appearing
in (1.164) can be regarded as the spinor equivalent of a vector, but in the present
case that vector is complex (according to (1.47), KAḂ is the spinor equivalent of a
real vector if and only if KAḂ = KBȦ ). Since det(K A Ḃ ) = 12 K AḂ KAḂ = 1, denoting
by Ka the vector equivalent of KAḂ , we have K a Ka = −2, which means that the real
and the imaginary parts of Ka are orthogonal to each other. Making use of (1.130),
(1.52), and (1.65), from (1.164) we obtain

La b = ∓ 12 − K c Kc δba + K a Kb + K a Kb − igae εebcd K c K d

= ∓ − 12 (vc vc + wc wc )δba + va vb + wa wb − gae εebcd vc wd , (1.165)
where va and wa are the real and the imaginary parts of Ka , respectively [cf. (1.131)]
(see also Penrose and Rindler 1984).
From K a Ka = −2, it follows that va va − wa wa = −2, and therefore, from (1.165)
we see that La b vb = ±va and La b wb = ∓wa . If {va , wa } is linearly independent, this
set spans a two-dimensional subspace invariant under (La b ).
When {va , wa } is linearly dependent, the orthogonality of va and wa implies
that va or wa is equal to zero; then (La b ) corresponds to the reflection in a three-
dimensional hyperplane passing through the origin or to the negative of such a re-
flection (see also Section 1.4, below). If va is equal to zero, wa is spacelike, and
if wa is equal to zero, va is timelike. Thus, in the latter case, (La b ) corresponds to
the reflection in√the three-dimensional hyperplane orthogonal to the unit timelike
vector T a = va / 2, or to the negative of this reflection. Hence, if TAḂ is the spinor
equivalent of Ta , from (1.164) it follows that
La b = −σ a AḂ σbCḊ T A Ḋ TC Ḃ (1.166)
corresponds to the improper Lorentz transformation that leaves invariant the time-
like vector T a and reverses the sign of the vectors orthogonal to T a . This transfor-
mation is the spatial inversion for an observer for which T a (or −T a ) represents its
time axis [cf. also (1.190)].
1.3.3 Ultrahyperbolic Signature
When (gab ) = diag (1, 1, −1, −1), the Infeld–van der Waerden symbols can all be
real [see (1.35)]. If these symbols are real, the matrix (1.102) is real if and only if
the unimodular matrices (K A B ) and (M Ȧ Ḃ ) are both real or both pure imaginary. The
4 × 4 matrices of the form
La b = − 12 σ a AȦ σb BḂ K A B M Ȧ Ḃ , (1.167)
with σ a AḂ given by (1.35) and (K A B ), (M Ȧ Ḃ ) ∈ SL(2, R), the group of unimodu-
lar real 2 × 2 matrices with matrix multiplication, form a connected subgroup of
SO(2, 2) that contains the identity of this group and will be denoted by SO(2, 2)0 .
40 1 Spinor Algebra

Furthermore, the mapping (K A B ), (M Ȧ Ḃ ) → (La b ), with La b given by (1.167), is a
two-to-one homomorphism of SL(2, R) × SL(2, R) onto SO(2, 2)0 , i.e.,

SO(2, 2)0 SL(2, R) × SL(2, R) /Z2 . (1.168)
Following a procedure analogous to those employed in the preceding two sub-

sections, we can find the possible forms of the SL(2, R) matrices. As shown in Sec-
tion 1.7, below, for a given 2 × 2 matrix (K A B ) there exist αA and βB such that
K A B = νδBA + 12 (α A βB + β A αB ), for some scalar ν . Then αA and βA are eigen-
spinors of (K A B ), and, assuming that {αA , βA } is linearly independent, the condition
det(K A B ) = 1 implies that
K ABα B = λ α A, K A B β B = λ −1 β A , (1.169)
for some λ ∈ C [cf. (1.149)]. If we now assume that (K A B ) belongs to SL(2, R), the
complex conjugate of the first equation in (1.169) gives
K ABα
B = λ α
A,
where, in the present case, the mate (conjugate or adjoint) of αA is defined by
α
A ≡ αA (1.170)
[cf. (1.113)]. This means that if αA is an eigenspinor of an SL(2, R) matrix, so
is its mate α
A , and now we have to consider two possible subcases according to
whether α A α
A is different from zero or not. (Note that α A α
A = α1 α2 − α2 α1 is pure
imaginary and that iα A α
A = 2 Re (iα1 α2 ) can take any real value.)
When α A αA = 0, {αA , α
A } is linearly independent, the equations above imply
that βA must be proportional to α A and λ = λ −1 , which means that λ = eiθ /2 , for
some θ ∈ R. Hence, from (1.169) we have
1 iθ /2 A
KAB = e α α B − e−iθ /2 α
A αB
α Rα
R
α Aα
B + α A αB
= cos 12 θ δBA + i sin 12 θ . (1.171)
α Rα
R
When α A αA = 0, α A is proportional to αA , and hence, λ = λ . Writing α

A = μαA ,
with μ ∈ C, then α
A = (μαA ) = μ αA = μμαA , and using the fact that α
A = αA ,
i χ i χ /2 i χ /2
it follows that μ = e , for some χ ∈ R; thus, (e αA )= e αA , and absorbing
the factor eiχ /2 into αA , we have αA = αA . Similarly, βA must be also proportional

to βA and we can assume that βA = βA . Then, expressing λ in the form λ = ±eθ /2 ,
from (1.169) it follows that
1 θ /2 A
K AB = ± e α βB − e−θ /2 β A αB
α R βR

α A βB + β A αB
= ± cosh 12 θ δBA + sinh 12 θ . (1.172)
α R βR
Finally, when {αA , βA } is linearly dependent we have K A B = νδBA + μα A αB , and

the condition det(K A B ) = 1 requires ν = ±1. Since (K A B ) must be real, α A has to
be proportional to αA and we can assume that α A = αA ; hence, in this case,

K A B = ± δBA + 12 θ α A αB ,
with θ ∈ R.
Summarizing, any SL(2, R) matrix different from the unit matrix has one of the
forms
⎧
⎪ cos 1 θ δ A + i sin 1 θ α α
A +α A αB
⎪
⎪
⎪
B
, with α A α
A = 0,
⎪
⎪
2 B 2 α Rα
R
⎪
⎨
1 α βB + β αB
A A
(1.173)
⎪
⎪ ± cosh 2 θ δB + sinh 2 θ
1 A A = αA , βA = βA ,
, with α
⎪
⎪ α Rβ
⎪
⎪
R
⎪
⎩ A 1
± δB + 2 θ α αB ,
A with α
A = αA .
These three cases correspond to |K A A | less than 2, greater than 2, and equal to 2,
respectively.
Similarly, one concludes that any SL(2, R) matrix (M Ȧ Ḃ ) different from the unit
matrix can have one of the forms
⎧
Ȧ + i sin 1 ϕ α αḂ + α αḂ ,
Ȧ Ȧ
⎪
⎪
⎪
⎪ cos 1
ϕ δ with α Ȧ α
Ȧ = 0,
⎪
⎪
2 Ḃ 2
α Ṙ α
Ṙ
⎪
⎨
α Ȧ βḂ + β Ȧ αḂ (1.174)
⎪
⎪ ± cosh 2ϕ
1
δḂȦ + sinh 12 ϕ Ȧ = αȦ , βȦ = βȦ ,
, with α
⎪
⎪ α Ṙ βṘ
⎪
⎪
⎪
⎩ Ȧ 1
± δḂ + 2 ϕ α Ȧ αḂ , with α
Ȧ = αȦ ,
where
α
Ȧ ≡ αȦ . (1.175)
Since the two SL(2, R) matrices (K A B ), (M Ȧ Ḃ ) appearing in (1.167) are independent
of each other and each of these matrices can have one of the forms (1.173) and
(1.174), there is a huge variety of SO(2, 2)0 transformations.
Improper O(2, 2) Transformations
With the Infeld–van der Waerden symbols given by (1.35), all the orthogonal trans-
formations with determinant equal to −1 are given by
1
La b = − σ a AȦ σb BḂ K A Ḃ M Ȧ B , (1.176)
2
42 1 Spinor Algebra
where the unimodular matrices (K A Ḃ ) and (M Ȧ B ) are both real or both pure imagi-
nary. KAḂ and MȦB can be regarded as the spinor equivalents of two real vectors or
two pure imaginary vectors, respectively; in the first case, writing
√ √
KAḂ = 2 vAḂ , MȦB = 2 wBȦ ,
one finds that vAḂ and wAḂ are the spinor equivalents of two real vectors va and wa
such that va va = −1 = wa wa . Making use of (1.52), (1.65), and (1.130), one finds
that (1.176) is equivalent to
La b = vc wc δba − va wb − wa vb + gae εebcd vc wd .
From this last expression one finds that La b vb = wa and La b wb = va .

In the second case, KAḂ and MȦB can be regarded as the spinor equivalents of two
pure imaginary vectors; letting
√ √
KAḂ = i 2 vAḂ , MȦB = i 2 wBȦ ,
one finds that vAḂ and wAḂ are the spinor equivalents of two real vectors va and wa
such that va va = 1 = wa wa , and (1.176) amounts to
La b = −vc wc δba + vawb + wa vb − gaeεebcd vc wd .
Here we also have La b vb = wa and La b wb = va .

As in the cases of the improper O(4) and O(3, 1) transformations, when {va , wa }
is linearly independent, the two-dimensional subspace spanned by {va , wa } is invari-
ant under (La b ) as well as its orthogonal complement, and when {va , wa } is linearly
dependent, (La b ) or −(La b ) is a reflection in a three-dimensional hyperplane passing
through the origin.
Alternative Basis
Apart from the real Infeld–van der Waerden symbols (1.35), another advanta-
geous choice is given by the complex scalars (1.37) because they satisfy σ a AȦ =
ηAB ηȦḂ σ aBḂ , with ηAB and ηȦḂ defined as in (1.39),

1 0
(ηAB ) ≡ ≡ (ηȦḂ ).
0 −1
Then, the entries of the orthogonal transformation (La b ) given by (1.102) are real if
and only if
σ a AȦ σb BḂ K A B M Ȧ Ḃ = ηAC ηȦĊ σ aCĊ η BD η ḂḊ σbDḊ K A B M Ȧ Ḃ ,
where, following the general rules (1.22),


−1 0
(η AB ) = = (η ȦḂ ).
01
Thus
KC D MĊ Ḋ = ηAC ηȦĊ η BD η ḂḊ K A B M Ȧ Ḃ ,
i.e., KC D = λ ηAC η BD K A B and MĊ Ḋ = λ −1 ηȦĊ η ḂḊ M Ȧ Ḃ , for some λ ∈ C. Again,

since det(K A B ) and det(M Ȧ Ḃ ) are equal to 1, λ = ±1. Hence, making use of (1.107),
one obtains the conditions
ηCA KC S K A R = ±ηSR , ηĊȦ MĊ Ṡ M Ȧ Ṙ = ±ηṠṘ . (1.177)
The unimodular 2 × 2 complex matrices (K A B ) or (M Ȧ Ḃ ) satisfying (1.177) with

the positive sign form a group with the usual matrix multiplication that is de-

noted by SU(1, 1). Since the pairs (K A B ), (M Ȧ Ḃ ) and −(K A B ), −(M
Ḃ ) give
Ȧ
the same SO(2, 2)0 transformation, SO(2, 2)0 can be identified with SU(1, 1) ×
SU(1, 1) /Z2 .
The components of the spinor equivalent vAḂ of a vector va computed according
to (1.42) with the aid of the real Infeld–van der Waerden symbols (1.35) are related
to the components of the spinor equivalent vAḂ of the same vector computed us-
ing the complex Infeld–van der Waerden symbols (1.37), which will be denoted by
σ a AḂ , by means of [see (1.31)]
1 a 1 0 0 0 0
vAḂ = √ σ AḂ va = √ K C A M Ḋ Ḃ σ aCḊ va =K C A M Ḋ Ḃ vCḊ , (1.178)
2 2
0 0
with (K A B ) and (M Ȧ Ḃ ) given by (1.36). Consistently with this relationship, we shall
assume that the components of one-index spinors are related by
0 0
φA =K B A φB , ψ
Ȧ =M Ḃ Ȧ ψḂ , (1.179)
where φA , ψ
Ȧ are the components with respect to the spinor basis associated with
the complex Infeld–van der Waerden symbols (1.37), and φA , ψȦ are the compo-
nents with respect to the basis associated with the real Infeld–van der Waerden
symbols (1.35).
Given the components of a one-index spinor φA with respect to the basis associ-
ated with the real Infeld–van der Waerden symbols (1.35), the operation φA → φA
produces another one-index spinor, which has been denoted by φA [see (1.170)].
According to (1.179), the components of φA with respect to the spinor basis as-
sociated with the complex Infeld–van der Waerden symbols (1.37) are given by
0 0 0 0
φ =K B φ =K B φ , but φ = (K −1 )C φ = − K C φ , and hence
A A B A B B B C B C
0 0 0 0
φA =K B A (− K BC φC ) = (K −1 )A B K BC φC .
44 1 Spinor Algebra
On the other hand, an explicit computation using (1.36) shows that
0 0
K A B = −iεBC η CD K A D , (1.180)
and therefore
0 0
φA = (K −1 )A B (−iε CD ηDE K B E )φC = −iηAD φD .
Hence, dropping the tildes, with respect to the basis associated with the complex
Infeld–van der Waerden symbols (1.37), the mate (conjugate or adjoint) of αA has
to be defined by
α
A ≡ −i ηAB α B (1.181)
(i.e., α
1 = iα2 , α
2 = iα1 ) [cf. (1.113)]. One can verify that as a consequence of
(1.177), if αA is an eigenspinor of an SU(1, 1) matrix (K A B ), then so is its mate α
A .
Depending on the value of the trace K A A , an SU(1, 1) matrix (K A B ) has one of
the forms (1.173). (Note that the trace K A A is real.)
0 0
Since (M Ȧ Ḃ ) is the complex conjugate of (K A B ), from (1.179) and (1.180) one
finds that with respect to the basis associated with the complex Infeld–van der Waer-
den symbols (1.37), the definition of the mate (conjugate or adjoint) of a spinor with
one dotted index (1.175) is equivalent to
α
Ȧ ≡ i ηȦḂ α Ḃ . (1.182)
Spin Transformations
Given two orthonormal bases {e1 , e2 , e3 , e4 } and {e1 , e2 , e3 , e4 } satisfying (1.2), any
vector v ∈ V can be expressed in the form v = va ea and v = va e a , where the va and
va are real numbers related by the “transformation law”
va = La b vb , (1.183)
which follows from (1.3), with (La b ) ∈ O(p, q). Hence, if (La b ) ∈ SO(p, q)0 [the
connected component of the identity in SO(p, q)], the spinor equivalents of va and
va , denoted by vAḂ and vAḂ , respectively, are related through [see (1.102)]
vAḂ = K AC M Ḃ Ḋ vCḊ , (1.184)
where
⎧
⎪
⎪ (K B ), (M Ḃ ) ∈ SU(2)
A Ȧ if the σ a AḂ are given by (1.33),
⎪
⎨ A
(K B ) ∈ SL(2, C), (M Ȧ Ḃ ) = (K A B ) if the σ a AḂ are given by (1.34),
(1.185)
⎪ (K A B ), (M Ȧ Ḃ ) ∈ SL(2, R)
⎪ if the σ a AḂ are given by (1.35),
⎪
⎩ A
(K B ), (M Ȧ Ḃ ) ∈ SU(1, 1) if the σ a AḂ are given by (1.37).
1.4 Reflections 45
Similarly, since the components of a tensor with respect to the bases {e1 , e2 , e3 , e4 }
and {e1 , e2 , e3 , e4 } are related by t ab... = La c Lb d · · ·t cd... , their spinor equivalents are
related by means of
t AB...ȦḂ... = K AC M ȦĊ K B D M Ḃ Ḋ · · ·t CD...ĊḊ... . (1.186)
The transformations (1.184) and (1.186) can be seen as consequences of the spin
transformations
ψ A = K AC ψ C , φ Ȧ = M ȦĊ φ Ċ (1.187)
on the one-index spinors.
Whereas in the transformations
(1.184)
and (1.186) the
two pairs of matrices (K A B ), (M Ȧ Ḃ ) and −(K A B ), −(M Ȧ Ḃ ) , corresponding to a
given SO(p, q)0 transformation, yield the same result (as it should), the effect of the
spin transformations (1.187) does depend on which pair of matrices restricted by
(1.185) we choose.
1.4 Reflections
A special class of orthogonal transformations is that of reflections on hyperplanes

passing through the origin. A nonnull vector Na defines a three-dimensional sub-
space of V formed by the vectors orthogonal to Na . The 4 × 4 real matrix given by
2N a Nb
La b = δba − (1.188)
N c Nc
corresponds to the reflection on the hyperplane passing through the origin orthogo-
nal to Na and is an orthogonal transformation with negative determinant, as can be
verified using (1.4) and (1.100) or, in a simpler manner, by comparing (1.190), be-
low, with (1.98). Making use of (1.42), (1.21), and (1.46), one finds that the spinor
equivalent of (1.188) is
2N AȦ NBḂ
LAȦ BḂ = −δBA δḂȦ + , (1.189)
NCĊ NCĊ
and then, by virtue of the identity [see (1.28) and (1.23)]
N AȦ NBḂ = N AȦ NBḂ − N A Ḃ NB Ȧ + N A Ḃ NB Ȧ

= N AṘ NBṘ δḂȦ + N A Ḃ NB Ȧ
= 12 (N AṘ NBṘ − NB Ṙ N A Ṙ )δḂȦ + N A Ḃ NB Ȧ
= 12 δBA δḂȦ N SṘ NSṘ + N A Ḃ NB Ȧ ,
one obtains the simple expression

46 1 Spinor Algebra
2N A Ḃ NB Ȧ
LAȦ BḂ = , (1.190)
NCĊ NCĊ
which is of the form (1.98).

The composition of any two reflections on hyperplanes passing through the origin
(or of any even number of such reflections) is an orthogonal transformation with unit
determinant. In a three-dimensional space, with definite or indefinite metric tensor,
any orthogonal transformation with unit determinant is the composition of two re-
flections (see, e.g., Cartan 1966, Porteous 1995, Lounesto 1997, Torres del Castillo
2003); however, in a four-dimensional space, a similar result does not hold. In fact,
the vectors belonging to the two-dimensional subspace formed by the intersection of
two different hyperplanes passing through the origin are fixed points of the composi-
tion of the reflections on these hyperplanes. Thus, the composition of the reflections
in two hyperplanes passing through the origin is a proper orthogonal transformation
with a double eigenvalue 1, while, as we have seen, not every proper orthogonal
transformation has two eigenvalues equal to 1.
For instance, if the metric tensor of V is positive definite, the spinor equivalent of
the composition of the reflections on the hyperplanes orthogonal to the unit vectors
N a and N a is [see (1.190); an extra minus sign comes from (1.46)]

C Ḃ N
LAȦ BḂ = − − 2N AĊ NC Ȧ − 2N Ċ = −4N A N Ċ Ȧ C
B Ċ B NC N Ḃ ,
which corresponds to an SO(4) transformation of the form (1.102) with K A B =

Ċ and M Ȧ Ḃ = 2N Ȧ N
2N AĊ N C Ḃ . Then, K A = −2N AĊ N
, and M Ȧ = −2NCȦ N
,
B C A AĊ Ȧ CȦ
i.e., K A = M Ȧ , which means that the composition of two reflections is a sim-
A Ȧ
ple SO(4) transformation (see Section 1.3.1). In fact, the converse is also true; we
can see explicitly that any simple SO(4) transformation is the composition of two
reflections in hyperplanes passing through the origin. Letting
1
NAḂ = √ eiθ /4 αA γḂ − e−iθ /4 α
A γḂ ,
2
1
AḂ = √ e−iθ /4 α γḂ − eiθ /4 α
N A γḂ ,
A
2
assuming that α R αR = 1 and γ Ṙ γṘ = 1, one finds that 2N AĊ N Ċ = eiθ /2 α A α
B −
B
e −iθ /2 α
αB and 2NC N
A Ȧ Ḃ = e γ γḂ − e
C iθ /2 Ȧ −iθ /2 γ γḂ , which are of the form (1.114)
Ȧ
and (1.116) with ϕ = θ , and letting

1
NAḂ = √ eiθ /4 αA γḂ + e−iθ /4 α
A γḂ ,
2
1
AḂ = √ eiθ /4 α
N A γḂ + e−iθ /4 αA γḂ ,
2
Ċ = eiθ /2 α A α
one finds that 2N AĊ N B − e−iθ /2 α C Ḃ = e−iθ /2 γ Ȧ γḂ −
A αB and 2NC Ȧ N
B
θ /2
e γ γḂ , which are of the form (1.114) and (1.116) with ϕ = −θ . In both cases,
i Ȧ
1.4 Reflections 47
a of N AḂ and N
the vector equivalents N a and N AḂ , respectively, are real unit vectors
a
and satisfy N Na = cos 2 θ .
1
As shown in Section 1.3.1, any SO(4) transformation is simple or is the composi-

tion of two simple SO(4) transformations, and therefore, any SO(4) transformation
can be obtained through the composition of two or four reflections.
On the other hand, the spinor equivalent of the composition of an arbitrary im-
proper O(4) transformation given by (1.128), followed by a reflection given by
(1.190), is
2N A Ḃ NB Ȧ B Ḃ (−2N AḂ wC Ḃ ) (2NB Ȧ vBĊ )

K Ċ M C = − ,
N RṘ NRṘ −N RṘ N −N RṘ N
RṘ RṘ
where we have made use of the definitions (1.129). This composition corresponds
to a simple SO(4) transformation if, for instance, we choose NAḂ orthogonal to vAḂ
and wAḂ , since the traces
(−2N A Ḃ wA Ḃ ) (2N Ȧ vB )
and B Ċ
−N RṘ NRṘ −N RṘ NRṘ
coincide. In effect,
−2N A Ḃ wA Ḃ = 0 and 2NB Ȧ vB Ȧ = 0.
(Alternatively, if vAḂ = wAḂ , one can choose NAḂ = vAḂ −wAḂ .) Hence, using the fact
that the square of a reflection is equal to the identity, we conclude that any improper
O(4) transformation is the composition of a simple SO(4) transformation followed
by a reflection, or equivalently, any improper O(4) transformation is a reflection or
can be expressed as the composition of three reflections.
Now we shall consider only the SO↑ (3, 1) transformations, showing that any sim-
ple orthochronous proper Lorentz transformation can be expressed as the composi-
tion of two reflections. Hence, any orthochronous proper Lorentz transformation is
the composition of two or four reflections.
The Lorentz boosts, given by (1.136) or (1.137), have a double eigenvalue equal
to 1, and these transformations can be expressed as the composition of two reflec-
tions on hyperplanes passing through the origin. The vector equivalents of
1
NAḂ = √ e−ζ /4 αA α Ḃ + eζ /4 βA β Ḃ ,
2

AḂ = √1 eζ /4 α α Ḃ + e−ζ /4 β β Ḃ ,
N A A
2
with αA and βA defined by (1.141) and (1.142), are two real vectors Na and Na
AḂ = N
(i.e., NAḂ = N and N ) such that N a Na = −1, N a = −1, and N
aN a Na =
BȦ BȦ
48 1 Spinor Algebra
− cosh 12 ζ . Making use of (1.190), one finds that the composition of the reflections
in the hyperplanes orthogonal to N a and N a is determined by
C Ḃ N
LAȦ BḂ = −(2N AĊ NC Ȧ )(2N Ċ )
B
Ċ )(2N Ȧ N
= −(2N AĊ N C Ḃ )
B C
Ċ )(2N A N
= −(2N AĊ N Ċ
B Ċ B ), (1.191)
but
ζ /4
Ċ = e−ζ /4 α A α + eζ /4 β A β
2N AĊ N αB α Ċ + e−ζ /4βB β Ċ
B Ċ Ċ e

= − e−ζ /2 α A βB − eζ /2 β A αB ,
and therefore (1.191) coincides with (1.143).

Similarly, any orthochronous proper Lorentz transformation that represents a ro-
tation is the composition of two reflections. For instance, the vector equivalents of
1
NAḂ = √ eiθ /4 αA β Ḃ + e−iθ /4βA α Ḃ ,
2
1
AḂ = √ e−iθ /4 αA β Ḃ + eiθ /4βA α Ḃ ,
N
2
with α A βA = 1, are two real vectors Na and Na such that N a Na = 1, N a = 1,

aN
a 1 A Ċ iθ /2
and N Na = cos 2 θ ; furthermore, 2N Ċ NB = e α βB − e
A −iθ /2 β αB , which co-
A
incides with (1.150) when λ = eiθ /2 .

The orthochronous proper Lorentz transformation corresponding to the SL(2, C)
matrix (1.152) is equal to the composition of the reflections on the hyperplanes
orthogonal to the vectors given by
1
NAḂ = √ αA β Ḃ + βAα Ḃ ,
2

AḂ = ± √1 − 1 ζ α α Ḃ + α β Ḃ + β α Ḃ ,
N 2 A A A
2
A Ċ
βAA is1a one-index
spinor such that α βA = 1. In fact, one finds that 2N Ċ NB
where A
= ± δB − 2 ζ α αB .
A
Orthogonal Projection on a Hyperplane
Given a vector va , the decomposition into symmetric and antisymmetric parts
NB Ṙ vAṘ = N(B Ṙ vA)Ṙ + N[B Ṙ vA]Ṙ = N(B Ṙ vA)Ṙ + 12 εBA N RṘ vRṘ ,
which, contracted with N B Ȧ , is equivalent to

1.5 Clifford Algebra. Dirac Spinors 49
N B Ȧ NB Ṙ vAṘ = N B Ȧ N(B Ṙ vA)Ṙ + 12 εBA N B Ȧ N RṘ vRṘ
or
1 1 1
− N BḂ NBḂ δȦṘ vAṘ = √ (NCĊ NCĊ )1/2 N B Ȧ v⊥ BA − N RṘ vRṘ NAȦ ,
2 2 2
where √
⊥
2 N(B Ṙ vA)Ṙ
v BA ≡ , (1.192)
(NCĊ NCĊ )1/2
allows us to express (the spinor components of) the vector va as the sum of a vector
orthogonal to N a and a vector proportional to N a ,
√ B ⊥
2 N Ȧ v AB N BḂ vBḂ NAȦ
vAȦ = − + .
(NCĊ NCĊ )1/2 NCĊ NCĊ
(The first term on the right-hand side of the last expression corresponds to a vector
orthogonal to N a , since N AȦ (N B Ȧ v⊥ AB ) = 12 N RȦ NRȦ ε AB v⊥ AB , which is equal to zero
because of the symmetry of v⊥ AB .) Hence, using the fact that v⊥ BA is symmetric, the
orthogonal projection of va on the hyperplane orthogonal to N a is the vector
√ B ⊥ √
2 N Ȧ v AB eAȦ 2 N(B Ȧ eA)Ȧ
= −v⊥AB ,
(NCĊ NCĊ )1/2 (NCĊ NCĊ )1/2
and therefore, a basis of (the complexification of) this hyperplane is formed by the
three vectors √
2 N(A Ȧ eB)Ȧ
eAB ≡ . (1.193)
(NCĊ NCĊ )1/2
The vectors eAB need not be real, and the explicit relationship between the vectors
eAB and their complex conjugates depends on the choice of the Infeld–van der Waer-
den symbols.
The map that sends an arbitrary vector v = −vAḂ eAḂ into its orthogonal projection
on the hyperplane orthogonal to N a , −v⊥AB eAB , can be extended to higher-rank
tensors assuming that this map is linear and preserves tensor products, keeping in
mind that the components of the projected objects are referred to a basis induced
by {eAB} (see also Sommers 1980, Sen 1982, Shaw 1983a,b, Ashtekar 1987, 1991,
Torres del Castillo 2003). (In a three-dimensional space, with a definite or indefinite
metric, only one type of spinor indices is necessary.)
1.5 Clifford Algebra. Dirac Spinors
Spinors are frequently introduced starting from the study of the Clifford algebras
(see, e.g., Penrose and Rindler 1986, Lawson and Michelsohn 1989, Chevalley 1996,
50 1 Spinor Algebra
Lounesto 1997). In this section we will show how the one-index spinors are related
to the Clifford algebras considering only the case of four-dimensional spaces.
As in the preceding sections, gab will denote the components of the metric tensor
of V with respect to an orthonormal basis, and we shall assume that (gab ) is of one
of the forms (1.1). The Clifford algebra associated with V can be defined as the
associative algebra generated by 1, γ1 , γ2 , γ3 , γ4 , which satisfy the relations
γa γb + γb γa = 2gab , (1.194)
and 1 is the unit element. The elements of the algebra are the linear combinations
with complex coefficients of 1, γ1 , γ2 , γ3 , γ4 , and their products.
The anticommutation relations (1.194) appear in connection with the Dirac equa-
tion for the electron. The relation E 2 /c2 − p2 = m20 c2 between the relativistic energy
E and the relativistic momentum p of a particle with rest mass m0 , leads to the
Klein–Gordon equation

ab ∂ ∂ m20 c2
g − 2 ψ = 0, (1.195)
∂ xa ∂ xb h̄
where (gab ) = diag (1, 1, 1, −1) and the xa are Cartesian space-time coordinates as-
sociated with an inertial reference frame, when E and p are replaced by the operators
ih̄∂ /∂ t and −ih̄∇, respectively, following the standard rules of quantum mechanics.
In contrast with the Schrödinger equation, the Klein–Gordon equation involves sec-
ond derivatives with respect to time and does not lead to a conserved positive definite
probability density. The Dirac equation is obtained by looking for a factorization of
the operator appearing in the Klein–Gordon equation (1.195) of the form

ab ∂ ∂ m20 c2 a ∂ m0 c b ∂ m0 c
g − 2 = γ + γ − ,
∂ xa ∂ xb h̄ ∂ xa h̄ ∂ xb h̄
which requires the validity of (1.194). (The index of γa is raised following the usual
rule with the aid of gab .) The Dirac equation can be expressed as

a ∂ m0 c
γ − ψ = 0. (1.196)
∂ xa h̄
We are interested in finding an irreducible representation of the Clifford algebra,
that is, we want to consider the elements of the algebra as linear operators on some
complex vector space D in such a way that there are no proper subspaces of D
that are invariant under all these operators. As we shall show, there is an infinite
number of equivalent irreducible representations of the algebra in a complex four-
dimensional vector space D, and finding each of these representations amounts to
finding a set of Infeld–van der Waerden symbols.
With the components of the metric tensor given by (1.1), the basic relations
(1.194) yield
(γa )2 = ±IN , (1.197)
where IN denotes the identity of D, and

γa γb = −γb γa , for a = b. (1.198)
From (1.197) it follows that each operator γa is invertible, γa −1 = ±γa , and from
(1.198) one obtains det γa det γb = (−1)N det γb det γa , where N denotes the dimen-
sion of D. Hence, N must be even.
Defining
γ5 ≡ iq γ1 γ2 γ3 γ4 , (1.199)
where, as above, q is the number of (−1)’s appearing in the diagonal matrix (gab ),
with the aid of (1.197) and (1.198) one finds that
γ5 γa = −γa γ5 (1.200)
and γ52 = (−1)q (γ1 )2 (γ2 )2 (γ3 )2 (γ4 )2 = IN . Furthermore, from (1.200) we have γ5 =
−γa γ5 γa −1 (without summation on a); hence, using the fact that tr (AB) = tr (BA),
we obtain tr γ5 = −tr (γa γ5 γa −1 ) = −tr γ5 , i.e., tr γ5 = 0. The eigenvalues of γ5 are
equal to +1 or −1 and from the condition tr γ5 = 0 it follows that there are as many
+1’s as −1’s (which also implies that dim D is even). Hence, there is a basis of D
formed by eigenvectors of γ5 , with respect to which γ5 is represented by the matrix

IN/2 0
,
0 −IN/2
where IN/2 is the N/2 × N/2 unit matrix. Writing the matrix corresponding to γa in
block form
Aa Ba
,
Ca Da
where Aa , Ba , Ca , and Da are N/2 × N/2 matrices, from (1.200) one finds that Aa =
Da = 0, and (1.194) amounts to
BaCb + BbCa = 2gab IN/2 , Ca Bb + Cb Ba = 2gab IN/2 . (1.201)
These equations imply that Ca = gaa (Ba )−1 (without summation on a). Furthermore,
for a = b, BaCb = −BbCa , and therefore det Ba detCb = (−1)N/2 det Bb detCa . On the
other hand, detCa = (gaa )N/2 (det Ba )−1 (without summation on a); hence
(gaa )N/2 (gbb )N/2

= (−1)N/2 , for a = b,
(det Ba )2 (det Bb )2
which cannot be satisfied if N/2 is odd, and we conclude that N, the dimension of
D, must be a multiple of 4.
52 1 Spinor Algebra
Assuming N = 4, we have

xa ya
Ba = , (1.202)
za wa
where for a = 1, 2, 3, 4, xa , ya , za , and wa are scalars, and hence

−wa ya
Ca = λa (without summation on a) (1.203)
za −xa
with
gaa
λa ≡ − (without summation on a).
detBa
Substituting (1.202) and (1.203) into (1.201), one finds that all λa must have a com-
mon value, λ (say) (otherwise (gab ) would be equal to a matrix of rank 2 at most),
and
λ (−xa wb − wa xb + ya zb + za yb ) = 2gab . (1.204)
By means of an appropriate change of basis of D we can eliminate the factor λ
appearing in (1.203) and (1.204). In effect, applying the similarity transformation
1/2 −1/2
0 Ba λ I 0 0 Ba λ I 0
→
Ca 0 0 I Ca 0 0 I

0 λ Ba
1/2
= ,
λ −1/2Ca 0
and redefining λ 1/2 xa , λ 1/2 ya , λ 1/2 za , and λ 1/2 wa as xa , ya , za , and wa , respectively,

one concludes that there is a basis of D, formed by eigenvectors of γ5 , with respect
to which the operators γa are represented by

0 Ba
γa = (1.205)
Ca 0
with
xa ya −wa ya
Ba = , Ca = , (1.206)
za wa za −xa
and
2gab = −xa wb − wa xb + ya zb + za yb . (1.207)
This last relation implies that the four (possibly complex) vectors with compo-
nents xa , ya , za , and wa form essentially a null tetrad as defined in Section 1.2;
i.e., they are null vectors, xa xa = ya ya = za za = wa wa = 0, that form a basis of (the
complexification of) V , and the only nonvanishing inner products among them are
xb wb = −2 and yb zb = 2. In fact, from (1.207) it follows that for any vector va ,
va = gab vb = 12 (−vb wb xa − vb xb wa + vb zb ya + vbyb za ),

thus showing that xa , ya , za , and wa span the complexification of V , and from this
last relation one obtains, e.g.,
xa = 12 (−xb wb xa − xb xb wa + xb zb ya + xb yb za ),
which shows that xb xb = xb zb = xb yb = 0 and xb wb = −2.

Letting
σa11̇ ≡ xa , σa12̇ ≡ ya , σa21̇ ≡ za , σa22̇ ≡ wa , (1.208)
one finds that (1.207) is equivalent to (1.40), and therefore, finding a representation
of the Clifford algebra associated with the metric tensor gab in a four-dimensional
complex vector space D is equivalent to finding a null tetrad and also to find-
ing Infeld–van der Waerden symbols for the metric tensor gab . Thus, according to
(1.206) and (1.208) we have

σa11̇ σa12̇ σa 11̇ σa 21̇
Ba = , Ca = − , (1.209)
σa21̇ σa22̇ σa 12̇ σa 22̇
and in order to get agreement with the notation employed in the preceding sections,
the components of an element of D with respect to a basis such that (1.205)–(1.207)
and (1.209) hold will be denoted by

ψA
. (1.210)
φ Ȧ
The elements of D are called Dirac spinors or bispinors. Owing to (1.205) and
(1.209),
ψA σaAȦ φ Ȧ
γa = . (1.211)
φ Ȧ −σa AȦ ψA
In order to verify that the representation of the Clifford algebra (1.194) given
by (1.205)–(1.207) is irreducible, it suffices to show (according to Schur’s lemma)
that any operator that commutes with all the linear combinations of the operators
(1.205) must be a scalar multiple of the identity. Indeed, such an operator, Δ , say,
must commute with γ5 and therefore with respect to the basis of D formed by the
eigenvectors of γ5 previously employed is represented by a matrix of the block form

F 0
,
0G
where F = (FA B ) and G = (GȦ Ḃ ) are 2 × 2 matrices. Then, Δ also commutes with
γa if and only if FBa = Ba G and GCa = Ca F, which amounts to
FA B σaBĊ = σaAḂ GḂĊ

54 1 Spinor Algebra
[see (1.209)]. Then, by contracting with σ aRṠ one finds that
FA R δĊṠ = GṠĊ δAR ,
which implies that F and G are the same scalar multiple of the identity matrix and
therefore Δ is a scalar multiple of the identity.
If (La b ) is an orthogonal transformation and the operators γa satisfy (1.194), then
the operators γa ≡ La b γb also satisfy (1.194) [cf. (1.183)]; in fact, γa γb + γb γa =
La c γc Lb d γd + Lb d γd La c γc = La c Lb d (γc γd + γd γc ) = 2La c Lb d gcd IN = 2gabIN . Thus,
the γa and the γa form representations of the Clifford algebra, and as a consequence
of (1.96), (1.98), and (1.211), these two representations are equivalent to each other
in the sense that there exists a unimodular operator U of D onto itself such that
γa = U −1 γaU. In effect, if (La b ) ∈ SO(p, q), using (1.102) and (1.211), one finds that

ψA La b σbAȦ φ Ȧ KC A MĊ Ȧ σaCĊ φ Ȧ
γa = =
φ Ȧ −La b σb AȦ ψA −KC A MĊ Ȧ σaCĊ ψA

ψA
= U −1 γaU ,
φ Ȧ
with

ψA −KAC ψC −1
ψA KC A ψC
U = , U = , (1.212)
φ Ȧ M ȦĊ φ Ċ φ Ȧ −MĊ Ȧ φ Ċ
and one can verify that U −1 γ5U = γ5 . Recall that the matrices (K A B ) and (M Ȧ Ḃ )
are defined by (La b ) up to a common sign; hence, U is defined by (La b ) up to sign.
Note also that (1.212) coincides with the spin transformations (1.187), and we can
identify D with the direct sum of the spin spaces.
Similarly, in the case of an improper orthogonal transformation given by Lb a =
− 2 σ BḂ σa AȦ K B Ȧ M Ḃ A , we have
1 b

ψA KC Ȧ MĊ A σaCĊ φ Ȧ ψA
γa = =U −1
γaU ,
φ Ȧ −KC Ȧ MĊ A σaCĊ ψA φ Ȧ
with

ψA KAĊ φ Ċ −1
ψA −MĊA φ Ċ
U = , U = , (1.213)
φ Ȧ −M ȦC ψC φ Ȧ KCȦ ψC
and U −1 γ5U = −γ5 . Thus, for any (La b ) ∈ O(p, q) there exists a unimodular opera-
tor U of D onto itself such that
U −1 γ aU = La b γ b . (1.214)
The operator U can be chosen in the explicit form (1.212) or (1.213), for a proper or
improper orthogonal transformation, respectively. In the first case each spinor space
is mapped onto itself, but in the second case one spin space is mapped onto the
other. (In fact, given a unimodular operator U satisfying (1.214), the operators iU,
−U, and −iU are also unimodular and satisfy (1.214). One can verify directly that
the set of unimodular operators U such that (1.214) holds for some (La b ) ∈ O(p, q)
form a group under composition and that the map U → (La b ) given by (1.214) is a
group homomorphism.)
We can prove the covariance of the Dirac equation (1.196) under a Lorentz trans-
formation xa = La b xb , with (La b ) ∈ O(3, 1). Indeed, applying the operator U defined
by (1.212) or (1.213) to the Dirac equation (1.196), we obtain

∂ m0 c
U γ aU −1 a − U ψ = 0,
∂x h̄
but U γ aU −1 = (L−1 )a b γ b and ∂ /∂ xa = Lc a (∂ /∂ xc ), and hence the last equation is

equivalent to
∂ m0 c
γ a a − U ψ = 0,
∂x h̄
which is of the form (1.196), with the coordinates xa in place of xa and ψ replaced
by U ψ .
More generally, if {γ1 , γ2 , γ3 , γ4 } is any other set of operators on D satisfying the
anticommutation relations (1.194), then, according to the foregoing results, there
exists a basis of D formed by eigenvectors of γ5 = iq γ1 γ2 γ3 γ4 . Defining an operator
S by the condition that S send each eigenvector of γ5 into an eigenvector of γ5 corre-
sponding to the same eigenvalue, we have, S−1 γ5 S = γ5 . Then, one can readily verify
that the operators S−1 γa S anticommute with γ5 and satisfy the anticommutation re-
lations (1.194); therefore, with respect to the basis formed by the eigenvectors of γ5 ,
S−1 γa S has the form given by (1.205) and (1.209), for some set of Infeld–van der
Waerden symbols. As shown in Section 1.3, any two sets of Infeld–van der Waerden
symbols are related through (1.31) or (1.32), and one finally concludes that there ex-
ists an operator U such that γa = U −1 γaU. In the context of the Dirac equation, this
result is known as Pauli’s theorem.
When the metric (gab ) has Lorentzian signature, one can define the Hermitian
inner product between bispinors

χA ψA
, ≡ i χ Ȧ φ Ȧ − η A ψA , (1.215)
η Ȧ φ Ȧ
which is indefinite and manifestly invariant under SO↑ (3, 1). It can readily be shown
that this inner product is invariant only under the transformations (1.212) or (1.213)
induced by proper or improper Lorentz transformations that preserve the time orien-
tation. The operators γa are anti-Hermitian with respect to this inner product. The ad-
joint of an operator defined by means of this inner product is the so-called Dirac ad-
joint. (Note that γ5 is also anti-Hermitian but its eigenvalues are real; this is possible
56 1 Spinor Algebra
owing to the fact that the inner product (1.215) is indefinite.) When the metric has
Euclidean or ultrahyperbolic signature, it is also possible to define a Hermitian inner
product between bispinors (see Exercise 1.12).
The inner product , std between bispinors employed in the standard formulation
of the Dirac equation (see, e.g., Messiah 1962, Davydov 1988, Merzbacher 1998)
is positive definite, but it is not invariant under SO↑ (3, 1) transformations. Up to
a constant positive real factor, this inner product is related to the inner product
(1.215) by
Φ , Ψ std = −iΦ , γ 4Ψ .
According to (1.211) and (1.69), the action of the commutator [γa , γb ] on a
bispinor (1.210) is given by

ψA 2SabAC ψC
[γa , γb ] = . (1.216)
φ Ȧ −2Sab ȦĊ φ Ċ
As we have shown, any SO(4) transformation can be written in the form (La b ) =
exp(−t a b ), where tab is an antisymmetric tensor [see (1.127)] and thecorrespond-
ing SU(2) matrices can be chosen as in (1.124), namely (K A B ) = exp 14 t ab Sab A B ,

(M Ȧ Ḃ ) = exp 14 t ab Sab Ȧ Ḃ ; hence, from (1.212) and (1.216) we see that the transfor-
mation induced on the bispinors is given by

U = exp − 18 t ab [γa , γb ] . (1.217)
According to the results of Section 1.3.2, any SO↑ (3, 1) transformation can also be
expressed as (La b ) = exp(−t a b ), where tab is an antisymmetric tensor, and one finds
that (1.217) also holds.
As in the case of any vector space, one can make use of other bases of D (not
formed by eigenvectors of γ5 ). When the metric has Lorentzian signature, the bases
usually employed in connection with the Dirac equation are formed by eigenvectors
of γ4 .
The explicit representation for the γa is useful to obtain some identities that do
not depend on the representation. For instance, from (1.211) and (1.81) it follows
that
1 abcd
ε γa γb γc = iq γ5 γ d ,
6
or equivalently,
γ[a γb γc] = (−i)q γ5 εabcd γ d . (1.218)
Making use repeatedly of the basic relations (1.194), one finds that γ[a γb γc] =
γa γb γc − gbc γa + gac γb − gab γc , hence, from (1.218),
γa γb γc = gbc γa − gac γb + gabγc + (−i)q γ5 εabcd γ d
(see also Exercise 1.14).

1.6 Inner Products. Mate of a Spinor 57
1.6 Inner Products. Mate of a Spinor
As we have seen in Section 1.3, when the signature of the metric is (+ + + +) or

(+ + − −), it is useful to define the mate (conjugate, or adjoint) of a spinor by
⎧
⎪
⎪ψ
AB...ĊḊ...
⎪
⎪
⎨ψ if the σ a AḂ are given by (1.35),
AB...ĊḊ...
ψ
AB...ĊḊ... = (1.219)
⎪
⎪ (−i)η (−i)η · · · η η · · · ψ MN... ṘṠ...
⎪
⎪ AM BN i Ċ Ṙ i Ḋ Ṡ
⎪
⎩
if the σ a AḂ are given by (1.37),
or equivalently,
⎧
⎪
⎪ (−1)m ψAB...ĊḊ... if the σ a AḂ are given by (1.33),
⎪
⎪
⎪
⎨ ψ AB...ĊḊ... if the σ a are given by (1.35),
ψ
AB...ĊḊ... = AḂ
(1.220)
⎪
⎪ (−1)m (−i)η AM (−i)η BN · · · i η ĊṘ i η ḊṠ · · · ψMN...ṘṠ...
⎪
⎪
⎪
⎩
if the σ a AḂ are given by (1.37),
where m is the total number of indices of ψAB...ĊḊ... . Then one finds that
εAB = εAB ,
εȦḂ = εȦḂ ,
(1.221)
and

(−1)m ψAB...ĊḊ... if the signature is (+ + + +),
ψ AB...Ċ Ḋ... = (1.222)
ψAB...ĊḊ... if the signature is (+ + − −).
From (1.222) it follows that when the signature is Euclidean, a nonzero m-index
spinor can be proportional to its mate if and only if m is even; assuming ψ AB...ĊḊ... =
λ ψAB...ĊḊ... , for some scalar λ , we have ψ
AB...ĊḊ... = λ ψ
AB...ĊḊ... = |λ |2 ψAB...ĊḊ... .
Comparison with (1.222) yields |λ | = (−1) , which implies that m must be even,
2 m
and if m is even, λ must be of the form eiθ , for some θ ∈ R.

The mapping ψAB...ĊḊ... → ψ AB...ĊḊ... is antilinear, and its existence is equivalent
to that of a Hermitian inner product given by

φ ψ AB...ṘṠ... if the signature is (+ + + +),
φ , ψ ≡ mAB...ṘṠ... (1.223)
i φAB...ṘṠ... ψ AB...ṘṠ... if the signature is (+ + − −),
for any pair of spinors of the same type, where m is the total number of indices of
each spinor. This inner product is invariant under spin transformations. Furthermore,
the inner product (1.223) is positive definite if and only if the metric tensor of V is
positive definite.
58 1 Spinor Algebra
The mate of the sum [resp. tensor product] of two spinors is equal to the sum
[resp. tensor product] of their mates, and by virtue of (1.221), we have, for instance,
α A βA = α A βA .
It should be remarked that the explicit expression for the components of the mate
of a spinor depends on the choice of the Infeld–van der Waerden symbols; each set
of Infeld–van der Waerden symbols corresponds to a definite choice for the basis of
the spin space.
By comparing (1.47) with (1.219) one finds that tAȦ...DḊ is the spinor equivalent
of a real n-index tensor tab...d if and only if

(−1)ntAȦ...DḊ if the signature is (+ + + +),
tAȦ...DḊ = (1.224)
tAȦ...DḊ if the signature is (+ + − −).
For instance, the spinor equivalent of a real bivector satisfies

tAȦBḂ = tAȦBḂ , which,
by virtue of (1.50) and (1.221), is equivalent to τAB = τAB and τȦḂ = τȦḂ .
The existence of the mate of a spinor in the cases in which the spin group is SU(2)
or SU(1, 1) (or a group isomorphic to one of them) follows from the fact that for each
of these groups, the fundamental representation is equivalent to its conjugate, which
amounts to the existence of a nonsingular 2 × 2 matrix (defined up to a factor) M
such that
K = M −1 KM, (1.225)
for all K in the group. For example, one can readily verify in an explicit manner
αβ
that any 2 × 2 complex matrix K belonging to SU(2) is of the form −β α , with
|α |2 + |β |2 = 1, and that
−1
αβ 0 −1 αβ 0 −1
= .
−β α 1 0 −β α 1 0
0 −1
That is, (1.225) holds with M = or any nonvanishing scalar multiple of this
10
α β
matrix. Similarly, any element of SU(1, 1) is of the form β α , with |α |2 − |β |2 = 1,
and −1
αβ 0 −i αβ 0 −i
= .
β α −i 0 β α −i 0
The matrix M gives (up to a factor) the relation between the complex conjugates
of the components of a spinor and the components of its mate; thus, in the cases of
SU(2) and SU(1, 1) we have

0 −1 ψ1 −ψ 2 0 −i ψ1 −iψ 2
= and = ,
1 0 ψ2 ψ1 −i 0 ψ2 −iψ 1
respectively [cf. (1.219)].

1.7 Principal Spinors. Algebraic Classification 59
1.7 Principal Spinors. Algebraic Classification
Owing to (1.27) and its dotted version, any spinor with two or more undotted or
dotted indices can be expressed as a sum of a totally symmetric spinor in each type
of spinor indices plus totally symmetric spinors multiplied by ε ’s. For example, an
object like μAȦḂĊ introduced in (1.53), which is symmetric in the last two indices,
has a totally symmetric part
μA(ȦḂĊ) = 16 (μAȦḂĊ + μAȦĊḂ + μAḂȦĊ + μAḂĊȦ + μAĊȦḂ + μAĊḂȦ )
= 13 (μAȦḂĊ + μAḂȦĊ + μAĊȦḂ ),
where we have made use of the symmetry of μAȦḂĊ in the last pair of dotted indices.
Thus,
μA(ȦḂĊ) = μAȦḂĊ + 13 (μAḂȦĊ − μAȦḂĊ + μAĊȦḂ − μAȦḂĊ )
= μAȦḂĊ + 13 μA Ṙ ṘĊ εḂȦ + 13 μA Ṙ ṘḂ εĊȦ ,
i.e.,
μAȦḂĊ = μA(ȦḂĊ) + 13 μA Ṙ ṘĊ εȦḂ + 13 μA Ṙ ṘḂ εȦĊ . (1.226)
The image of a totally symmetric spinor under a spin transformation is also
totally symmetric, and since the spin transformations have unit determinant, εAB and
εȦḂ are invariant under the spin transformations in the sense that, e.g., KAC KB D εCD =
εAB [see (1.30)]; hence, the (complex) vector space formed by the spinors totally
symmetric on k undotted indices and totally symmetric on l dotted indices multi-
plied by a fixed number of εAB ’s and εȦḂ ’s is invariant under the spin transforma-
tions. Furthermore, the action of the group of spin transformations on this vector
space is irreducible, in the sense that there is no nontrivial subspace invariant under
all the spin transformations. Thus, a decomposition such as the one given in (1.226)
corresponds to a decomposition into irreducible parts, and the group of the spin
transformations possesses an irreducible representation on the vector space formed
by the spinors totally symmetric on k undotted indices and totally symmetric on
l dotted indices, denoted by D(k/2, l/2). The dimension of this representation is
(k + 1)(l + 1).
A particularly simple and important case corresponds to the totally symmetric
spinors with only one type of indices (undotted or dotted) because they can be ex-
pressed as the symmetrized tensor product of one-index spinors (Penrose 1960).
Proposition 1.5. If φAB...L is a k-index totally symmetric spinor, then there exist k
one-index spinors (not necessarily distinct), αA , βA , . . . , ζA , such that
φAB...L = α(A βB · · · ζL) . (1.227)

60 1 Spinor Algebra
Proof. Let λ 1 , λ 2 be two auxiliary complex variables. Then if λ 2 = 0, the expression

k k−1
λ1 λ1
(λ 2 )−k φAB...L λ A λ B · · · λ L = φ11...11 + k φ11...12
λ2 λ2
k−2
k(k − 1) λ1
+ φ11...22 + · · · + φ22...22
2 λ2
is a polynomial in λ 1 /λ 2 of degree k provided that φ11...11 does not vanish, and

according to the fundamental theorem of algebra, there exist k complex numbers (the
roots of the polynomial) that can be written as ratios of complex numbers α 1 /α 2 ,
β 1 /β 2 , . . . , ζ 1 /ζ 2 such that
(λ 2 )−k φAB...L λ A λ B · · · λ L

λ 1 α1 λ1 β1 λ1 ζ1
= φ11...11 − − ··· −
λ 2 α2 λ2 β2 λ2 ζ2
φ11...11
= (α 2 λ 1 − α 1 λ 2 )(β 2 λ 1 − β 1 λ 2 ) · · · (ζ 2 λ 1 − ζ 1 λ 2 );
(λ 2 )k α 2 β 2 · · · ζ 2
hence
φ11...11
φAB...L λ A λ B · · · λ L = α λ A βB λ B · · · ζL λ L . (1.228)
α1 β1 · · · ζ1 A
Since only the ratios α 1 /α 2 , β 1 /β 2 , . . . , ζ 1 /ζ 2 are fixed, we can make φ11...11 =
α1 β1 · · · ζ1 , and using the fact that λ A is arbitrary one obtains (1.227).
If, for instance, φ11...11 = 0 but φ11...12 = 0, then (λ 2 )−k φAB...L λ A λ B · · · λ L is a
polynomial in λ 1 /λ 2 of degree k − 1; hence, expressing the k − 1 roots of this poly-
nomial in the form β 1 /β 2 , . . . , ζ 1 /ζ 2 ,
1 1
2 −k λ β1 λ ζ1
(λ ) φAB...L λ λ · · · λ = kφ11...12
A B L
− ··· −
λ2 β2 λ2 ζ2
kφ
= 2 k−111...12 (β 2 λ 1 − β 1 λ 2 ) · · · (ζ 2 λ 1 − ζ 1 λ 2 ).
(λ ) β 2 · · · ζ 2
Thus
kφ11...12 2
φAB...L λ A λ B · · · λ L = λ βB λ B · · · ζL λ L = αA λ A βB λ B · · · ζL λ L ,
β1 · · · ζ1
with
kφ11...12
α1 ≡ 0, α2 ≡ ,
β1 · · · ζ1
which leads again to (1.227). In a similar way one can verify the validity of (1.227)
when the degree of the polynomial (λ 2 )−k φAB...L λ A λ B · · · λ L is less than k − 1.

The one-index spinors αA , βA , . . . , ζA appearing in (1.227) are called principal

spinors of φAB...L . As pointed out above, if k 2, each principal spinor is defined up
to a scalar factor. From (1.228) it follows that φAB...L λ A λ B · · · λ L = 0 if and only if
λ A is proportional to one of the principal spinors of φAB...L .
A first algebraic classification of the totally symmetric k-index spinors comes
from the possible multiplicities of the roots of the polynomial appearing in the proof
of Proposition 1.5,
k(k − 1)
φ11...11 zk + k φ11...12 zk−1 + φ11...22 zk−2 + · · · + φ22...22 . (1.229)
2
If one of the roots of this polynomial is repeated m times, e.g., assuming that
precisely the first m roots coincide, α 1 /α 2 = β 1 /β 2 = · · · , we can take αA = βA
= · · · ; hence,
φAB...D...L = α(A αB · · · αD · · · ζL) ,
!" #
m
which is equivalent to the conditions
φAB...D...L α A α B!"· · · α D# = 0, φAB...D...L α A α B!"· · · α D# = 0. (1.230)

k−m k−m+1
It should be clear that all the preceding results, especially (1.227) and (1.230),
also apply if all the indices are dotted.
When the signature of V is Euclidean or ultrahyperbolic, the mate of a totally
symmetric k-index spinor φAB...L can be proportional to φAB...L , that is, φAB...L =
λ φAB...L , and in that case, if αA is an m-fold repeated principal spinor of φAB...L , then
so is α
A . This conclusion follows from the fact that the relations (1.230) would then
be equivalent to
φAB...D...L α
Aα
B · · ·α
!"
D = 0,
# φAB...D...L α
Aα
B · · · α
!"
D = 0.
#
k−m k−m+1
As we have shown (Section 1.6), if the signature of V is Euclidean, αA cannot be

proportional to αA (except in the trivial case αA = 0), and therefore a nonvanishing
totally symmetric k-index spinor φAB...L such that φAB...L is proportional to φAB...L
must be of the form
B βC βD · · · ζL ζM) ,
φABCD...LM = eiθ α(A α (1.231)
for some θ ∈ R, and necessarily k must be an even number (the one-index spinors
αA , βA , . . . need not be distinct). An example is provided by any antisymmetric real
two-index tensor (or bivector) tab whose spinor equivalent is of the form tAȦBḂ =
τAB εȦḂ + τȦḂ εAB , with τAB = τ(AB) , τȦḂ = τ(ȦḂ) and τAB = τAB , τȦḂ = τȦḂ [see (1.50)
and (1.224)]. Thus, there exists γ such that τ = λ γ γ . Then, τ = λ γ
A AB (A B) AB γ =
(A B)
−λ γ(A γB) , which implies that λ = ib, for some b ∈ R. Thus, we have τAB = iα(A α
B) ,
62 1 Spinor Algebra
√ √ √
where αA ≡ b γA if b > 0, and αA ≡ −b γA if b < 0 (then α A = − −b γA ).
Similarly, it follows that there exists βȦ such that τȦḂ = iβ(Ȧ βḂ) and
B) εȦḂ + iβ(ȦβḂ) εAB .

tAȦBḂ = iα(A α (1.232)
As an application of the expression (1.232) we can prove that if a bivector tab

satisfies ∗tabt ab = 0, then tab is simple. Indeed, from (1.67) and (1.232) we see that
∗ t t ab = 0 implies α α (A B) = β β (Ȧ Ḃ)
ab (A B) α α (Ȧ Ḃ) β β , which, in turn, is equivalent to
α Aα = β Ȧ β . Taking
A Ȧ

α β −α A βȦ A βȦ
i αA βȦ + α
vAȦ = A ȦR , wAȦ = ,
(2α α R )1/2 (2α R α
R )1/2
we have [see (1.52)]
vAȦ wBḂ − vBḂwAȦ = v(A Ṙ wB)Ṙ εȦḂ + vR(Ȧ w|R|Ḃ) εAB ,
1
= 2iα(A α B) β Ṙ βṘ εȦḂ + 2iβ(ȦβḂ) α R α
R εAB
2α αS
S
= iα α ε + iβ β ε ,
(A B) ȦḂ (Ȧ Ḃ) AB
i.e., tab = va wb − vb wa , where va and wa are the vector equivalents of vAȦ and wAȦ ,
respectively, thus showing that tab is simple. It may be verified that the vectors va
and wa are real [see (1.224)].
If the signature is ultrahyperbolic, a nonzero one-index spinor can be propor-
tional to its mate, and therefore, in contrast to (1.231), the principal spinors of a
totally symmetric k-index spinor proportional to its mate need not appear in conju-
gate pairs. For instance, when k = 2, if φAB = φAB , then φAB is of one of the forms
⎧
⎨ ±α(A α
⎪ B) with α
A αA = 0,
α(A βB) with α A = αA , βA = βA ,
⎪
⎩ ±α α with α
A B =α , A A
which correspond to φ AB φAB being positive, negative, or equal to zero, respectively.

When V has Lorentzian signature, the spinor equivalent of a real bivector tab is
of the form
tAȦBḂ = α(A βB) εȦḂ + α (Ȧ β Ḃ) εAB . (1.233)
Making use of this expression, we can explicitly show that if ∗tabt ab = 0, then tab is
simple. In fact, from ∗tabt ab = 0 it follows that α(A βB) α (A β B) = α (Ȧ β Ḃ) α (Ȧ β Ḃ) , or
equivalently, α A βA = ±α Ȧ β Ȧ , which means that α A βA is real or pure imaginary. If
α A βA is real and different from zero, we can assume without loss of generality that
α A βA is positive (otherwise, we have only to interchange αA and βA , which leaves
(1.233) unchanged); then, letting
αA α Ȧ − βAβ Ȧ αA α Ȧ + βAβ Ȧ
vAȦ = , wAȦ = ,
(2α R βR )1/2 (2α R βR )1/2
one finds that tab = va wb − vb wa . If α A βA is different from zero and pure imaginary,
we can assume that iα A βA is positive. Then with

i αA β Ȧ − βAα Ȧ αA β Ȧ + βAα Ȧ
vAȦ = , wAȦ = ,
(2iα R βR )1/2 (2iα R βR )1/2
we obtain tab = va wb − vb wa .
Finally, if α A βA = 0, then βA is proportional to αA and we can assume that tAȦBḂ
is of the form
tAȦBḂ = αA αB εȦḂ + α Ȧ α Ḃ εAB . (1.234)
Taking
vAȦ = αA α Ȧ , wAȦ = αA γ Ȧ + γA α Ȧ ,
where γA is a one-index spinor such that α R γR = 1, one finds that tab = va wb − vb wa .
Since tAȦBḂt AȦBḂ = −(α A βA )2 − (α Ȧ β Ȧ )2 [see (1.233)], these three possible cases
correspond to tabt ab negative, positive, or zero, respectively. In all cases, the vectors
va and wa defined above are real (cf. also (3.29) and (3.30)).
Spin
Now we restrict ourselves to the case that V has Lorentzian signature. In elementary
quantum mechanics, the spin of an object is defined by its behavior under spatial
rotations. A spin-s object has 2s + 1 components that form a basis for an irreducible
representation of SU(2), the double covering group of SO(3). As we shall see, the
components ψAB...ĊḊ... of a spinor with n undotted indices and m dotted indices can
be decomposed as the sum of objects of spin 12 (n + m), 12 (n + m) − 1, . . . , 12 |n − m|.
In order to isolate the spatial rotations, we employ a timelike vector ta . The set of
vectors orthogonal to ta form a three-dimensional subspace of V , on which the inner
product is positive definite. The SO↑ (3, 1) transformations that leave ta invariant
(and hence leave invariant the subspace orthogonal to ta ) form a group isomorphic
to SO(3) (the group of rotations of a three-dimensional space with a definite inner
product). Correspondingly, the SL(2, C) matrices (K A B ) satisfying
Ċ
K ABK Ḋ tAĊ = tBḊ , (1.235)
where tAḂ is the spinor equivalent of ta , form a group isomorphic to SU(2). Since
det(tAḂ ) = 12 t AḂtAḂ = − 12 t ata = 0, the 2 × 2 matrix (tAḂ ) is invertible, and therefore
there is an invertible relation between ψAB...ĊḊ... and
φ AB...CD... ≡ t CĊ t D Ḋ · · · ψ AB...ĊḊ... .
Then, under the spin transformation defined by an SL(2, C) matrix (K A B ) that satis-
fies the condition (1.235), we have
64 1 Spinor Algebra
Ċ Ḋ
φ AB...CD... → K A M K B N · · · K Ṗ K Q̇ · · ·t Ċ t Ḋ ψ
C D MN...ṖQ̇...
= K A M K B N · · · KC Rt R Ṗ K D St S Q̇ · · · ψ MN...ṖQ̇...
= K A M K B N · · · KC R K D S · · · φ MN...RS... ,
which means that the components φ AB...CD... form a basis for a (possibly reducible)
representation of SU(2).
By expressing φ AB...CD... as the sum of its totally symmetric part plus totally
symmetric objects (with n + m − 2, n + m − 4, . . . , |n − m| indices) multiplied by
εAB ’s (though some terms may be equal to zero), as at the beginning of this section,
one decomposes φ AB...CD... as a sum of objects of spin 12 (n + m), 12 (n + m) − 1, . . . ,
2 |n − m| (though some values may be absent). Note that this decomposition depends
1
on the timelike vector ta , except in the case that m or n is equal to zero (actually,
when ψȦḂ...L̇ has only dotted indices, the value of the components of φAB...L do
depend on ta , but the number of independent totally symmetric spinors that can be
extracted from φAB...L does not depend on ta ). (See, e.g., the discussion about spin-
3/2 particles in Berestetskii et al. 1982, §31.)
Exercises
1.1. Show that if tAȦBḂ is the spinor equivalent of a symmetric two-index tensor tab ,
then sAȦBḂ = tBȦAḂ is the spinor equivalent of tab − 12 t c c gab .
1.2. Show that if tABĊḊ is the spinor equivalent of a symmetric two-index tensor tab ,
then t(AB)ĊḊ = tAB(ĊḊ) is the spinor equivalent of tab − 14 t c c gab (the traceless part
of tab ).
1.3. Show that the spinor equivalent of a three-index tensor Habc such that
Habc = −Hbac , Hab b = 0, Habc + Hbca + Hcab = 0,
is of the form HAȦBḂCĊ = hABCĊ εȦḂ + hȦḂĊC εAB , with hABCĊ = h(ABC)Ċ and hȦḂĊC =
h(ȦḂĊ)C .
1.4. Using the spinor formalism, show that if the signature is Lorentzian, then two
nonzero real null vectors are orthogonal to each other if and only if they are propor-
tional.
1.5. Show that

tAȦtBḂ = tBȦtAḂ
if and only if tAȦ is the spinor equivalent of a null vector. (If tAȦ is different
from zero, there exist αA and βȦ such that tAȦ α A β Ȧ = 0; hence, contracting both
sides of the equation above with α B β Ḃ , one concludes that if tAȦt AȦ = 0, then
tAȦ = tBȦ α BtAḂ β Ḃ /(tBḂ α B β Ḃ ); this constitutes a proof of Proposition 1.3.)
1.6. Using the spinor formalism, show that if tab and sab are two bivectors, then
tac sb c + (−1)q ∗ sac ∗tb c = 12 tde sde gab
and conclude that tac ∗tb c = 14 tde ∗t de gab . Hence, a bivector tab is simple if and only
if tac ∗tb c = 0, or equivalently, ta[btcd] = 0.
1.7. Using the spinor formalism, show that if tab is a bivector, then
1 ∗
det(t ab ) = 16 ( tab t ) .
ab 2
1.8. Let tab be an anti-self-dual bivector (i.e., ∗tab = −iqtab ) and ua an arbitrary
nonnull vector. Show, using the spinor formalism, that if we let va ≡ tab ub , then
1
tab = va ub − ua vb + (−i)q εabcd uc vd .
uc u c
1.9. Assume that the signature of the metric is Lorentzian. If the axes of the refer-
ence frame S are obtained from those of S by means of a spatial rotation through
an angle θ about the axis defined by the unit vector ua , the Cartesian coordinates of
a point with respect to these frames are related by xa = La b xb , with
La b = cos θ δba + (1 − cos θ )(ua ub − T a Tb ) + sin θ gam εmbcd uc T d ,
where T a is a timelike vector orthogonal to ua such that T a Ta = −1. Show that the
spinor equivalent of La b is of the form −K A B K Ȧ Ḃ with
K A B = ± (eiθ /2 α A βB − e−iθ /2β A αB )
and α A βA =√ 1. Here αA α Ȧ and

√ βA β Ȧ are the spinor equivalents of the null vectors
(T a + ua )/ 2 and (T a − ua )/ 2, respectively. (Note that as in the case of the boost
(1.135), we are considering passive transformations in the sense that the points are
not affected by transformations; only the basis vectors are transformed.)
1.10. Assuming that the signature of the metric is Lorentzian, show that any two-
dimensional spacelike plane orthogonal to a given null vector la is spanned by the
vector equivalents of αA β Ȧ + βA α Ȧ and i(αA β Ȧ − βA α Ȧ ), where ±αA α Ȧ is the
spinor equivalent of la and βA is some one-index spinor such that α A βA = 0.
1.11. Show that each matrix in (1.173) can be expressed in the form
νδBA + μ va wb Sab A B ,
where ν , μ are two real scalars and va , wa are two real vectors.
1.12. Show that if the signature of the metric is Euclidean, the expression
Φ , Ψ = χ
A ψ A + η
Ȧ φ Ȧ ,
66 1 Spinor Algebra
χ ψ
where Φ and Ψ are the bispinors Φ = A , Ψ = A , defines a positive definite
η
Ȧ Ȧ φ
Hermitian inner product. Show that with respect to this inner product, the operators
γa are Hermitian. Also show that if the signature of the metric is ultrahyperbolic,
then
Φ , Ψ = i χA ψ A + η
Ȧ φ Ȧ
defines an indefinite Hermitian inner product. Show that with respect to this inner
product, the operators γa are anti-Hermitian.
1.13. Show that the eigenvalues of each operator γa are ±(gaa )1/2 (without summa-
tion on a) and find the corresponding eigenvectors.
1.14. Show that

2 εabcd [γ , γ ] = −i γ5 [γa , γb ].
1 c d q
1.15. Prove explicitly that if the signature of the metric is ultrahyperbolic, a real
bivector tab such that ∗tabt ab = 0 is simple.
Chapter 2
Connection and Curvature
In this chapter some applications of the spinor formalism to four-dimensional

Riemannian manifolds are considered. By definition, if p is a point of a Riemannian
manifold M, the tangent space to M at p, Tp M, is a vector space with a nondegen-
erate inner product, and therefore, one can construct null tetrads in the complexifi-
cation of Tp M, thus allowing the use of the spinor algebra discussed in the previous
chapter.
It is shown that the connection and curvature of a four-dimensional Riemannian
manifold can be conveniently computed and analyzed by making use of the two-
component spinor formalism. As examples, the connection and curvature of the
standard metric of S4 , the Schwarzschild metric, and the Euclidean Schwarzschild
metric are computed. Some further applications that will be employed in the follow-
ing chapters, such as the conformal rescalings of the metric, the Killing vector fields,
and the algebraic classification of the conformal curvature for all signatures of the
metric, are studied. The Newman–Penrose and the Geroch–Held–Penrose notations
are also introduced.
It is assumed that the reader is acquainted with the basic concepts related to
differentiable manifolds such as tangent vectors, vector fields, differential forms,
and connections (see, e.g., Conlon 2001, Hall 2004).
Throughout this chapter it will be assumed the Infeld–van der Waerden symbols
are given by one of the forms (1.33)–(1.35) and (1.37), depending on the signature
of the metric of M.
2.1 Covariant Differentiation
In what follows, M will denote a four-dimensional Riemannian manifold, that is,

M will be a differentiable manifold of dimension four endowed with a Riemannian
metric (a differentiable tensor field g such that at each point p ∈ M, g p is a non-
degenerate, symmetric, not necessarily definite, bilinear form defined on Tp M, the
tangent space to M at p).

Physics 59, DOI 10.1007/978-0-8176-4984-5_2,
68 2 Connection and Curvature
Each point p ∈ M possesses a neighborhood U in M such that there exist four

differentiable vector fields {∂1 , ∂2 , ∂3 , ∂4 } defined on U, which, evaluated at a point
q ∈ U, form an orthonormal basis of Tq M. Such a set is called an orthonormal
tetrad. Furthermore, we shall assume that the constants gab = g(∂a , ∂b ) have one
of the forms (1.1). (In most cases, the vector fields ∂a will not be the holonomic
basis associated with a coordinate system; that is, in most cases there will not exist
coordinates xa such that ∂a = ∂ /∂ xa . See also Appendix A.)
Given an orthonormal tetrad {∂a }, there exists a unique set of four one-forms
{θ a } such that θ a (∂b ) = δba . Then, the metric tensor can be expressed locally in the
form
g = gab θ a ⊗ θ b , (2.1)
where ⊗ denotes the tensor product, that is, g(X,Y ) = gab θ a (X)θ b (Y ), for any pair
of vector fields X, Y on M. Thus, we have
⎧ 1
⎨ θ ⊗ θ 1 + θ 2 ⊗ θ 2 + θ 3 ⊗ θ 3 + θ 4 ⊗ θ 4,
g = θ 1 ⊗ θ 1 + θ 2 ⊗ θ 2 + θ 3 ⊗ θ 3 − θ 4 ⊗ θ 4,
⎩ 1
θ ⊗ θ 1 + θ 2 ⊗ θ 2 − θ 3 ⊗ θ 3 − θ 4 ⊗ θ 4,
if the signature of the metric is (+ + + +), (+ + + −), or (+ + − −), respectively.

By expressing a given metric in this form, one can identify a set of one-forms {θ a }
corresponding to some orthonormal tetrad {∂a }.
Connection One-Forms
With respect to an orthonormal tetrad {∂a }, a connection ∇ on M is represented by

the set of (real-valued) functions Γ a bc , defined by
∇a ∂b = Γ c ba ∂c , (2.2)
where ∇a denotes the covariant derivative along ∂a , that is, ∇a = ∇∂a [note the
position of the subscripts of Γ c ba in (2.2)]. It follows from (2.2) and the properties
of a connection that the connection one-forms
Γ a b ≡ Γ a bc θ c (2.3)
satisfy the relation

Γ a b (X) = θ a (∇X ∂b ), (2.4)
for any vector field X on M.
With the torsion of the connection ∇ being defined by
T (X,Y ) = ∇X Y − ∇Y X − [X,Y ], (2.5)
for any pair of vector fields X, Y on M, the torsion two-forms with respect to the
tetrad {∂a } will be defined by

T a (X,Y ) ≡ 12 θ a T (X,Y ) , (2.6)
2.1 Covariant Differentiation 69
for any pair of vector fields X, Y on M. Then, from (2.4)–(2.6), one derives the
Cartan first structure equations
dθ a + Γ a b ∧ θ b = T a (2.7)
if the differential (or exterior derivative) of a one-form α is defined by

dα (X,Y ) ≡ 12 X α (Y ) − Y α (X) − α ([X,Y ]) , (2.8)
for any pair of vector fields X, Y on M, and the exterior product of two one-forms
α , β is defined by

(α ∧ β )(X,Y ) ≡ 12 α (X)β (Y ) − α (Y )β (X) . (2.9)
If the connection ∇ is compatible with the metric, that is,
X[g(Y, Z)] = g(∇X Y, Z) + g(Y, ∇X Z),
for all vector fields X, Y , Z on M, we have ∂a gbc = g(∇a ∂b , ∂c ) + g(∂b , ∇a ∂c ) =

Γ d ba gdc + Γ d ca gbd = Γcba + Γbca (lowering the tetrad indices by means of gab in the
usual way). Thus, with respect to an orthonormal tetrad, a connection compatible
with the metric is represented by a set of functions Γabc (the Ricci rotation coeffi-
cients) satisfying
Γabc = −Γbac . (2.10)
As is well known, in a Riemannian manifold there is a unique torsion-free con-
nection compatible with the metric (the Riemannian or Levi-Civita connection) (see,
e.g., Conlon 2001). In what follows we shall consider only Riemannian connections.
The Ricci rotation coefficients can be computed by making use of the Cartan
first structure equations (2.7) or the commutators of the vector fields ∂a . In fact,
combining (2.5) and (2.2), one finds that when the torsion vanishes,
[∂a , ∂b ] = (Γ c ba − Γ c ab )∂c . (2.11)
By virtue of (2.10), the Ricci rotation coefficients are determined by the commuta-
tion relations (2.11). In effect, writing
[∂a , ∂b ] = Cc ba ∂c ,
we have Cc ba = Γ c ba − Γ c ab , and one readily verifies that
Γabc = 12 (Cabc + Cbca − Ccab ). (2.12)
As a consequence of (2.10), the connection one-forms (2.3) satisfy Γab = −Γba ;

therefore, their spinor equivalents ΓAȦBḂ = 12 σ a AȦ σ b BḂΓab can be expressed in the
form
ΓAȦBḂ = IΓAB εȦḂ + IΓȦḂ εAB , (2.13)
where IΓAB = IΓ(AB) and IΓȦḂ = IΓ(ȦḂ) are one-forms [see (1.50)]. (We are employing
the symbol IΓ in order to avoid confusion with the connection one-forms Γ a b , which
also possess two indices.) Then, letting
1
θ AȦ ≡ √ σa AȦ θ a , (2.14)
2
the Cartan first structure equations (2.7), with vanishing torsion, are equivalent to
dθ AȦ = Γ AȦ BḂ ∧ θ BḂ = IΓA B ∧ θ BȦ + IΓȦ Ḃ ∧ θ AḂ. (2.15)
(The one-forms θ√AḂ are related to the one-forms

√ gAḂ employed by Plebański (1974,
1975) by g = 2 θ . The factor 1/ 2 included in (2.14) is dictated by the rule
AḂ AḂ
(1.41).)
The one-forms IΓAB and IΓȦḂ can be expressed in the form
IΓAB = −ΓABCĊ θ CĊ , IΓȦḂ = −ΓȦḂĊC θ CĊ , (2.16)
where the ΓABCĊ and ΓȦḂĊC are some (possibly complex-valued) functions; then,
substituting these expressions into the right-hand side of (2.13) and making use of
(2.3), we obtain
1
ΓABCĊ εȦḂ + ΓȦḂĊC εAB = √ σ a AȦ σ b BḂ σ cCĊ Γabc , (2.17)
2 2
which leads to the expressions
1
ΓABCĊ = √ SabAB σ cCĊ Γabc ,
4 2
(2.18)
1
ΓȦḂĊC = √ SabȦḂ σ cCĊ Γabc
4 2
[see (1.69)]. Following Newman and Penrose (1962), the functions ΓABCĊ and
ΓȦḂĊC will be called spin coefficients. (The spin coefficients can also be expressed
in terms of the Christoffel symbols associated with some coordinate system; see
Appendix A.)
From (2.18), (2.16), and (2.14) it follows that the one-forms IΓAB and IΓȦḂ are
related to the connection one-forms Γab by
IΓAB = 14 SabABΓab , IΓȦḂ = 14 Sab ȦḂΓab .
The right-hand side of (2.17) is the spinor equivalent of the Ricci rotation coeffi-
cients ΓAȦBḂCĊ . Thus
ΓAȦBḂCĊ = ΓABCĊ εȦḂ + ΓȦḂĊC εAB . (2.19)
The fact that the Ricci rotation coefficients are real-valued functions amounts to
certain specific relations between the spin coefficients and their complex conjugates,
depending on the signature of the metric. From (1.224) and (1.221) one finds that if
the signature of the metric is (+ + + +), then
ΓABCḊ = −ΓABCḊ , ΓȦḂĊD = −ΓȦḂĊD (2.20)
(i.e., Γ ABCḊ = −ΓABCḊ and Γ ȦḂĊD = −ΓȦḂĊD ); if the signature is (+ + + −), we

have
ΓABCḊ = ΓȦḂĊD , (2.21)
and if the signature of the metric is (+ + − −), then
ΓABCḊ = ΓABCḊ , ΓȦḂĊD = ΓȦḂĊD . (2.22)
The spin coefficients can be obtained directly from the relations
[∂AḂ , ∂CḊ ] = −Γ RCAḂ ∂RḊ − Γ Ṙ ḊḂA ∂CṘ + Γ R ACḊ ∂RḂ + Γ Ṙ ḂḊC ∂AṘ , (2.23)
which follow from (2.11) and (2.17), with the definition

1
∂AḂ ≡ √ σ a AḂ ∂a (2.24)
2
[cf. (1.20)] (see Example 2.2 on pp. 86–87). In actual computations, it is convenient
to make use of individual symbols for the 24 spin coefficients, owing to the rela-
tively high number of indices of ΓABCḊ and ΓȦḂĊD . Following Newman and Penrose
(1962), we shall denote the spin coefficients ΓABCḊ according to
Γ1111̇ = κ , Γ1112̇ = σ , Γ1121̇ = ρ , Γ1122̇ = τ ,

Γ1211̇ = ε , Γ1212̇ = β , Γ1221̇ = α , Γ1222̇ = γ , (2.25)
Γ2211̇ = π , Γ2212̇ = μ , Γ2221̇ = λ , Γ2222̇ = ν ,
and, in a similar way,
Γ1̇1̇1̇1 = κ, Γ1̇1̇1̇2 = σ , Γ1̇1̇2̇1 = ρ, Γ1̇1̇2̇2 = τ,

Γ1̇2̇1̇1 =
ε, Γ1̇2̇1̇2 = β, Γ1̇2̇2̇1 = α, Γ1̇2̇2̇2 = γ, (2.26)
Γ2̇2̇1̇1 = π, Γ2̇2̇1̇2 = μ, Γ2̇2̇2̇1 =
λ, Γ2̇2̇2̇2 = ν.
Then, the relations (2.23) take the explicit form
[D, δ ] = −(α ε )δ + σ δ,
+ β − π)D − κΔ + (ρ + ε −
− β )Δ − (μ − γ + γ)δ −
[Δ , δ ] = νD − (τ − α λ δ,
ε )Δ + (τ + π )δ + (τ + π)δ,
[D, Δ ] = −(γ + γ)D − (ε +
(2.27)
[δ , δ] = (μ − μ )Δ − (α − β)δ − (β − α
)D + (ρ − ρ )δ,
[D, δ] = −(α + β − π )D − κ ε − ε )δ + σ δ ,
Δ + (ρ +
[Δ , δ] = ν D − (τ − α − β)Δ − (μ
− γ + γ )δ − λ δ ,
with the definitions
D ≡ ∂11̇ , δ ≡ ∂12̇ , δ ≡ ∂21̇ , Δ ≡ ∂22̇ . (2.28)
When the metric has Lorentzian signature, then D and Δ are real, each symbol
with a tilde is the complex conjugate of the corresponding symbol without a tilde
[see (2.21)], and the last two lines of (2.27) are the complex conjugates of the first
two lines. (Despite several differences in the conventions, the commutation relations
(2.27) coincide with those given in Newman and Penrose 1962.)
When the signature of the metric is Euclidean, the spin coefficients are related by
κ = −ν , σ = λ, ρ = μ, τ = −π , ε = γ, β = −α , (2.29)
with identical relations for the spin coefficients with a tilde (e.g., κ = −ν) and
D = −Δ , δ = δ. (2.30)
In the case that the signature of the metric is (+ + − −) and one employs the real
Infeld–van der Waerden symbols, the vector fields D, δ , δ, and Δ are real, and the
24 spin coefficients (2.25) and (2.26) are real and independent.
Using the properties of the covariant derivative, one finds that the covariant
derivative of a vector field X = X a ∂a with respect to a vector field Y = Y a ∂a is
given by
∇Y X = Y a ∇a (X b ∂b )
= Y a [(∂a X b )∂b + X b∇a ∂b ]
= Y a (∂a X c + Γ c ba X b )∂c .
By abuse of notation, the components of the covariant derivative of X along ∂a are

usually denoted by ∇a X c , that is,
∇a X c = ∂a X c + Γ c ba X b . (2.31)
Hence, according to (2.19), the spinor equivalent of ∇a X c is given by
∇AȦ X CĊ = ∂AȦ X CĊ − Γ CĊ BḂAȦ X BḂ

= ∂AȦ X CĊ − Γ C BAȦ X BĊ − Γ Ċ ḂȦA X CḂ . (2.32)
The covariant derivative of an arbitrary tensor field can be defined as usual, as-
suming that the covariant differentiation is linear, satisfies the Leibniz rule, com-
mutes with contraction, and the covariant derivative of a function coincides with the
directional derivative (∇X f = X f , for any vector field X). In this way one finds that
the components of the covariant derivative of a tensor field with tetrad components
ab... are
tcd...
∇atde...
bc...
= ∂atde...
bc...
+ Γ b satde...
sc...
+ Γ c satde...
bs...
+ · · · − Γ s datse...
bc...
− Γ s eatds...
bc...
− · · · . (2.33)
Tetrad Rotations
In the intersection of their domains, any two orthonormal tetrads {∂a } and {∂a }
must be related by means of an expression of the form ∂a = La b ∂b , where the La b are
differentiable functions such that at each point p belonging to the common domain
of {∂a } and {∂a }, (La b (p)) is an orthogonal matrix [see (1.3)]. From (2.2) and the
properties of a connection we have
∇∂a ∂b = La c ∇∂c (Lb d ∂d )

= La c Lb d Γ s dc ∂s + (∂c Lb d )∂d

= La c Lb d Lr sΓ s dc + (∂c Lb d )Lr d ∂r .
This last expression must coincide with Γ r ba ∂r , where Γ a bc are the Ricci rotation
coefficients for the tetrad {∂a }; hence,

Γrba = La c Lb d Lr sΓsdc + La c Lr d ∂c Lbd . (2.34)
The transformation law (2.34) is equivalent to some relation between the corre-
sponding spin coefficients. For instance, if the matrices (La b (p)) have positive de-
terminant, the functions La b can be expressed as
1
La b = − σa AȦ σ b BḂ KA B MȦ Ḃ , (2.35)
2
where the K A B and M Ȧ Ḃ are differentiable functions such that at each point p where
they are defined, the matrices (K A B (p)) and (M Ȧ Ḃ (p)) belong to the appropriate
subgroup of SL(2, C), depending on the signature of the metric. Substituting (2.35)
into (2.34), making use of (2.17) and (1.30) we obtain
Γ ABCĊ = KA R KB S KC D MĊ Ḋ ΓRSDḊ − KA R KC D MĊ Ḋ ∂DḊ KBR (2.36)
and
Γ ȦḂĊC = MȦ Ṙ MḂ Ṡ MĊ Ḋ KC D ΓṘṠḊD − MȦ Ṙ MĊ Ḋ KC D ∂DḊ MḂṘ . (2.37)
These formulas can be expressed in a more compact form in terms of the connection
one-forms IΓAB and IΓȦḂ :
IΓAB = KA R KB S IΓRS − KA R dKBR ,

(2.38)
IΓȦḂ = MȦ Ṙ MḂ Ṡ IΓṘṠ − MȦṘ dMḂṘ .
Spinor Fields
Throughout Chapter 1, and in what follows, spinors are given by means of their
components with respect to some basis, which is induced by an orthonormal basis
of a vector space V , or of the tangent space to M at some point. A simultaneous

rotation through 2π in both spin spaces leaves the corresponding orthonormal basis
invariant, but changes the sign of any spinor with an odd number of indices [see
(1.112), (1.116), (1.155), and (1.171)]. In order to define spinor fields on M, it is
necessary to assign consistently the change of sign under these rotations over M,
which imposes some restrictions on the global structure of M (see, e.g., Geroch
1968, 1970, Penrose and Rindler 1984, Wald 1984, Lawson and Michelsohn 1989,
Huggett and Tod 1994). We shall assume that M admits a spinor structure, and we
shall restrict ourselves to local considerations.
A spinor field will be given by means of its components with respect to bases of
the spin spaces (or spin frames); a basis of one of the spin spaces together with a
basis of the other spin space defines a unique null tetrad, but the converse is not true.
Covariant Derivatives of Spinor Fields
Since the Riemannian connection of M is compatible with the metric tensor, the
parallel transport of tangent vectors to M along a given curve on M is an orthogonal
transformation between the tangent spaces at any two points on the curve. Hence,
with respect to a (differentiable) null tetrad ∂AḂ , this orthogonal transformation must
be of the form K A B MĊ Ḋ (since by continuity, its determinant must be positive),
which means that the parallel transport of spinors along a curve can be defined in
such a way that undotted [resp. dotted] spinors are mapped onto undotted [resp.
dotted] spinors, and therefore, the covariant derivative of any spinor field will be a
spinor field of the same type as the original spinor field.
Equation (2.32) is consistent with the assumption that for one-index spinor fields,
∇AḂ φ C = ∂AḂ φ C − Γ C SAḂ φ S , ∇AḂ φ Ċ = ∂AḂ φ Ċ − Γ Ċ ṠḂA φ Ṡ , (2.39)
and that the covariant derivative of spinor fields also satisfies the Leibniz rule; then,
as a consequence of the symmetry ΓABCḊ = ΓBACḊ , we have
∇AḂ ε CD = 0, ∇AḂ ε ĊḊ = 0,
and therefore, assuming that the covariant derivative commutes with the
contractions,
∇AḂ εCD = 0, ∇AḂ εĊḊ = 0.
Thus, for an arbitrary spinor field ψC...

A...Ḃ... ,
Ḋ...
∇RṠ ψC... Ḋ... = ∂RṠ ψC...Ḋ... − Γ PRṠ ψC...Ḋ... − · · · − Γ ṖṠR ψC...Ḋ... − · · ·

A...Ḃ... A...Ḃ... A P...Ḃ... Ḃ A...Ṗ...
+ Γ PCRṠ ψP...Ḋ... + · · · + Γ ḂṠR ψC...Ṗ... + · · · .

A...Ḃ... Ṗ A...Ḃ...
(2.40)
Weighted Quantities
When the metric has Lorentzian signature, any nonvanishing complex-valued func-
tion λ defines a tetrad rotation (2.35) with the SL(2, C)-valued matrix
−1
λ 0
(K A B ) = (2.41)
0 λ
and M Ȧ Ḃ = K A B . Following Geroch, Held, and Penrose (1973), a quantity η is of

type (p, q) if under the tetrad rotation defined by (2.41), it transforms into
η = λ pλ q η .
Thus, for instance, from ∂ AḂ = KAC MḂ Ḋ ∂CḊ , it follows that the null tetrad vector
fields ∂11̇ , ∂12̇ , ∂21̇ , and ∂22̇ are of type (1, 1), (1, −1), (−1, 1), and (−1, −1), re-
spectively. In a similar way, each component ψAB...ĊḊ... of a spinor field is of type
(p, q), where p is the number of undotted subscripts, A, B, . . . equal to 1 minus the
number of these subscripts equal to 2, while q is the number of dotted subscripts
Ċ, Ḋ, . . . equal to 1̇ minus the number of these subscripts equal to 2̇ [see (1.187)].
According to (2.36) and (2.37), the same rules apply to the spin coefficients Γ11AḂ,
Γ22AḂ, Γ1̇1̇ȦB , and Γ2̇2̇ȦB , but
Γ 12CĊ = KC D MĊ Ḋ (Γ12DḊ + λ −1∂DḊ λ ), (2.42)
which means that the spin coefficients Γ12DḊ are not of a definite type. Similarly,
even though each member of the null tetrad ∂AḂ is of a definite type, the directional
derivative ∂AḂ η of a quantity of type (p, q) is not of a definite type. In fact, under
the tetrad rotation defined by (2.41), ∂AḂ η transforms into

∂ AḂ η = KAC MḂ Ḋ ∂CḊ λ p λ q η . (2.43)
Thus, by combining (2.43) with (2.42) and its conjugate, one finds that
(∂ AḂ − pΓ 12AḂ − qΓ 1̇2̇ḂA )η = KAC MḂ Ḋ λ p λ q (∂CḊ − pΓ12CḊ − qΓ1̇2̇ḊC )η , (2.44)
which means that (∂AḂ − pΓ12AḂ − qΓ1̇2̇ḂA )η has a well-defined type (which depends
on p, q, and the specific values of A and Ḃ). This fact leads to the definition of the
Geroch–Held–Penrose (GHP) operators
Þη ≡ (D − pε − qε )η ,
ðη ≡ (δ − pβ − qα )η ,
(2.45)
ð η ≡ (δ − pα − qβ )η ,
Þ η ≡ (Δ − pγ − qγ )η ,
if η is of type (p, q), where we have made use of the notation (2.25) and (2.28). From
(2.44) it follows that Þη , ðη , Þ η , and ð η are of type (p + 1, q + 1), (p + 1, q − 1),
(p − 1, q + 1), and (p − 1, q − 1), respectively.
From the definition one finds that if η is of type (p, q), then its complex conjugate
η is of type (q, p). Hence,
Þη = Þη , ðη = ð η , Þ η = Þ η .
The spin weight s and the boost weight w of η are defined by

p−q p+q
s= , w= .
2 2
When the signature of the metric is Euclidean, any pair of complex-valued func-
tions λ , μ such that |λ | = 1 = |μ | defines a tetrad rotation (2.35) with the SU(2)-
valued matrices
−1 −1
λ 0 μ 0
(K B ) =
A
, (M Ḃ ) =
Ȧ
. (2.46)
0 λ 0 μ
A quantity η is said to be of type (p, q) if under the tetrad rotation given by (2.46)
it transforms according to η = λ p μ q η . Following a procedure similar to that em-
ployed in the previous case, one finds that if η is of type (p, q), then
Þη ≡ (D − pε − qε )η ,
ðη ≡ (δ − pβ − qα )η ,
(2.47)
ð η ≡ (δ − pα − qβ)η ,

Þ η ≡ (Δ − pγ − qγ)η ,
are of type (p + 1, q + 1), (p + 1, q − 1), (p − 1, q + 1), and (p − 1, q − 1), respectively

[cf. (2.45)].
Now, if η is of type (p, q), then η is of type (−p, −q), and as a consequence of
(2.29) and (2.30),
Þη = −Þ η , ðη = ð η .
Making use of the corresponding definitions, one finds that in all cases, if η is of
type (p, q) and κ is of type (r, s), then ηκ is of type (p + r, q + s) and
O(ηκ ) = η O κ + κ O η ,
where O is any of the weighted operators Þ, ð, ð , and Þ .

The use of weighted quantities allows one to write many explicit expressions in
a compact form, simplifying some computations (see, e.g., Jeffryes 1986). Whereas
the commutators between the vector fields ∂AḂ involve only the spin coefficients
[(2.23)], the presence of the spin coefficients in the weighted operators (2.44) im-
plies that the commutators of the latter contain some components of the curvature
[see (2.91)–(2.94)].
2.2 Curvature 77
2.2 Curvature
The curvature tensor Ω of the connection ∇ is defined by
Ω (X,Y )Z = ∇X ∇Y Z − ∇Y ∇X Z − ∇[X,Y ] Z, (2.48)
for any vector fields X, Y , Z on M. With respect to an orthonormal tetrad {∂a }, the
curvature of a connection is represented by the curvature two-forms Ω a b , defined by

Ω a b (X,Y ) ≡ 12 θ a Ω (X,Y )∂b , (2.49)
for any pair of vector fields X, Y on M. One can verify that the Ω a b are indeed
two-forms and that
Ω a b = 12 Ra bcd θ c ∧ θ d , (2.50)
where the Ra bcd are the components of the curvature tensor with respect to the tetrad
{∂a }, which are defined by Ω (∂a , ∂b )∂c = Rd cab ∂d .
Making use of (2.8), (2.9), (2.48), (2.49), (2.4), and the properties of ∇, one
obtains the Cartan second structure equations
Ω a b = dΓ a b + Γ a c ∧ Γ c b . (2.51)
The curvature two-forms satisfy the relation Ωab = −Ωba , which can be readily
obtained from (2.51). Indeed, from (2.51) and the fact that the connection one-forms
satisfy Γab = −Γba we have
Ωab = dΓab + Γac ∧ Γ c b = −dΓba − Γbc ∧ Γ c a = −Ωba.
Hence, the components of the curvature tensor satisfy Rabcd = −Rbacd .

Applying the exterior derivative operator d to both sides of (2.7), using again
these equations and (2.51) one obtains the identities
dT a + Γ a b ∧ T b = Ω a b ∧ θ b , (2.52)
and therefore, if the torsion of the connection vanishes,
Ω a b ∧ θ b = 0. (2.53)
Substituting (2.50) into (2.53), one finds that when the torsion is equal to zero,
Ra bcd + Ra cdb + Ra dbc = 0.
Similarly, applying the exterior derivative operator to both sides of (2.51), one
obtains the Bianchi identities
dΩ a b + Γ a c ∧ Ω c b − Γ c b ∧ Ω a c = 0. (2.54)
2.2.1 Curvature Spinors
Any two-form can be expressed locally in terms of the exterior products θ a ∧ θ b , or

equivalently, in terms of the exterior products θ AȦ ∧ θ BḂ, which, according to (1.50)
and (1.69), can be written in the form
θ AȦ ∧ θ BḂ = 12 σa AȦ σb BḂ θ a ∧ θ b

= ε ȦḂ 14 σa (A|Ṙ| σb B) Ṙ + ε AB 41 σa R(Ȧ σbR Ḃ) θ a ∧ θ b

= 14 Sab AB ε ȦḂ + 14 Sab ȦḂ ε AB θ a ∧ θ b
≡ SABε ȦḂ + SȦḂ ε AB , (2.55)
where we have introduced the six two-forms
SAB ≡ 14 Sab AB θ a ∧ θ b , SȦḂ ≡ 14 Sab ȦḂ θ a ∧ θ b . (2.56)
Then, from (2.55) we have
SAB = 12 θ AȦ ∧ θ B Ȧ , SȦḂ = 12 θ AȦ ∧ θA Ḃ , (2.57)
or in an explicit form,
S11 = θ 11̇ ∧ θ 12̇ , S12 = 12 (θ 11̇ ∧ θ 22̇ − θ 12̇ ∧ θ 21̇ ), S22 = θ 21̇ ∧ θ 22̇ ,
(2.58)
S1̇1̇ = θ 11̇ ∧ θ 21̇ , S1̇2̇ = 12 (θ 11̇ ∧ θ 22̇ − θ 21̇ ∧ θ 12̇ ), S2̇2̇ = θ 12̇ ∧ θ 22̇ .
From (1.76) we see that the duals of the two-forms SAB and SȦḂ are given by
∗ AB ∗ ȦḂ
S = −iq SAB, S = iq SȦḂ . (2.59)
The two-forms SAB [resp. SȦḂ ] form a local basis for the anti-self-dual [resp. self-
dual] two-forms.
Making use of (2.57), the properties of the exterior derivative, the Cartan first
structure equations (2.15), and the symmetry IΓṘṠ = IΓṠṘ , one finds that
dSAB = 12 d(θ AṘ ∧ θ B Ṙ )

= dθ (A|Ṙ| ∧ θ B) Ṙ
= (IΓ(AC ∧ θ |CṘ| + IΓṘ Ṡ ∧ θ (A|Ṡ| ) ∧ θ B) Ṙ
= 2IΓ(AC ∧ SB)C , (2.60)
and in a similar way one obtains
dSȦḂ = 2IΓ(ȦĊ ∧ SḂ)Ċ . (2.61)

2.2 Curvature 79
As we shall show below, the relations (2.60) and (2.61) are very useful in obtaining
the spin coefficients for a given tetrad. (An equivalent formulation, for the Lorentzian
signature, without the spinor notation is given in Israel 1979.)
Since the components of the curvature tensor of a connection compatible with the
metric are antisymmetric in the first and the last pairs of indices Rabcd = −Rbacd =
−Rabdc , making use of (1.50) it follows that the spinor equivalent of the curvature
tensor can be written as
RAȦBḂCĊDḊ = XABCD εȦḂ εĊḊ + CȦḂCD εAB εĊḊ
+ CABĊḊ εȦḂ εCD + XȦḂĊḊ εAB εCD , (2.62)
where the sets of functions XABCD , CȦḂCD , CABĊḊ , and XȦḂĊḊ are symmetric in the
first and the last pairs of indices, XABCD = X(AB)(CD) , CȦḂCD = C(ȦḂ)(CD) , CABĊḊ =
C(AB)(ĊḊ) , and XȦḂĊḊ = X(ȦḂ)(ĊḊ) . If the Ricci tensor is defined by
Rab ≡ Rc acb , (2.63)
from (2.62) and (1.46) we find that the spinor equivalent of the Ricci tensor is
RBḂDḊ = −RAȦBḂAȦDḊ
= −X ABAD εḂḊ + CḂḊBD + CBDḂḊ − X ȦḂȦḊ εBD , (2.64)
and the scalar curvature R ≡ Ra a is given by
R = −RBḂ BḂ = 2(X AB AB + X ȦḂ ȦḂ ). (2.65)
The antisymmetry of the curvature two-forms Ωab = −Ωba implies that their
spinor equivalents ΩAȦBḂ are of the form
ΩAȦBḂ = RAB εȦḂ + RȦḂ εAB , (2.66)
where RAB = R(AB) and RȦḂ = R(ȦḂ) are two-forms. From (2.50) and (2.55) it
follows that ΩAȦBḂ = 12 RAȦBḂCĊDḊ (SCD ε ĊḊ + SĊḊ ε CD ), and substituting (2.66) and
(2.62), one concludes that the two-forms RAB and RȦḂ are given by
RAB = XABCD SCD + CABĊḊ SĊḊ ,

(2.67)
RȦḂ = CȦḂCD SCD + XȦḂĊḊ SĊḊ .
The Cartan second structure equations (2.51) expressed in terms of the spinor equiv-
alents of the connection one-forms and of the curvature two-forms given by (2.13)
and (2.66), respectively, read
RAB = dIΓAB − IΓAC ∧ IΓC B ,
(2.68)
RȦḂ = dIΓȦḂ − IΓȦĊ ∧ IΓĊ Ḃ
[cf. (2.51)].
Similarly, making use of (2.66) and (2.67), one finds that (2.53) is equivalent to
0 = RAB ∧ θ B Ȧ + RȦḂ ∧ θA Ḃ
= (XABCD SCD + CABĊḊ SĊḊ ) ∧ θ B Ȧ + (CȦḂCD SCD + XȦḂĊḊ SĊḊ ) ∧ θA Ḃ . (2.69)
Owing to the antisymmetry of the exterior product of one-forms, according to (1.58),
θ AȦ ∧ θ BḂ ∧ θ CĊ = ε AC ε ḂĊ θ̆ BȦ − ε BC ε ȦĊ θ̆ AḂ , (2.70)
where the θ̆ AȦ are some three-forms, which implies that
SAB ∧ θ CĊ = ε C(A θ̆ B)Ċ , SȦḂ ∧ θ CĊ = −ε Ċ(Ȧ θ̆ |C|Ḃ) (2.71)
[see (2.57)]; hence, (2.69) can be rewritten as
0 = XA B BD θ̆ D Ȧ + CABȦḊ θ̆ BḊ − CȦḂAD θ̆ DḂ − XȦḂ ḂḊ θ̆A Ḋ .
At each point of its domain, the set of three-forms {θ̆ AȦ} is linearly independent
(see Exercise 2.2), and therefore the last equation is equivalent to
0 = XA B BD εȦḊ + CADȦḊ − CȦḊAD − XȦ Ḃ ḂḊ εAD . (2.72)
Hence, contracting the last equation with ε ȦḊ yields
0 = 2XAB BD − X ḊḂ ḂḊ εAD , (2.73)
while contracting with ε AD gives us
0 = X DB BD εȦḊ − 2XȦḂ ḂḊ . (2.74)
From (2.73) or (2.74) we obtain X AB AB = X ȦḂ ȦḂ , and therefore (2.65) implies that
X AB AB = X ȦḂ ȦḂ = 14 R. (2.75)
Thus, from (2.73) and (2.74) we obtain
XA B BD = 18 RεAD , XȦ Ḃ ḂḊ = 18 RεȦḊ , (2.76)
and substituting these expressions into (2.72), it reduces to
CȦḊAD = CADȦḊ . (2.77)
As a consequence of (2.76), XABCD and XȦḂĊḊ are symmetric not only on the first
and second pairs of indices, but also under the exchange of the first and second pairs
of indices. Indeed, from (1.27) and (2.76) we see that
2.2 Curvature 81
XABCD − XCDAB = XABCD − XACBD + XACBD − XCDAB

= εBC XA M MD + εAD XC M MB
= 18 R(εBC εAD + εAD εCB )
= 0,
i.e.,
XABCD = XCDAB , (2.78)
and in a similar manner one concludes that
XȦḂĊḊ = XĊḊȦḂ . (2.79)
From (2.62) and (2.77)–(2.79) it follows that RAȦBḂCĊDḊ = RCĊDḊAȦBḂ , which is

equivalent to the well-known relation Rabcd = Rcdab .
Making use of (2.78) and (2.76), one finds that the totally symmetric part of
XABCD , which will be denoted by CABCD , is
CABCD = X(ABCD)
= 13 (XABCD + XACBD + XADBC )
= XABCD + 13 (XACBD − XABCD + XADBC − XABCD)
= XABCD + 13 (εCB XA M MD + εDB XA M MC )
= XABCD − 24
1
R(εAC εBD + εAD εBC ). (2.80)
Similarly, denoting by CȦḂĊḊ the totally symmetric part of XȦḂĊḊ , one obtains
CȦḂĊḊ = XȦḂĊḊ − 24
1
R(εȦĊ εḂḊ + εȦḊ εḂĊ ). (2.81)
Thus, from (2.67), (2.77), (2.80), and (2.81) we see that the curvature two-forms can
be expressed as
RAB = CABCD SCD + 12

1
R SAB + CABĊḊ SĊḊ ,
(2.82)
RȦḂ = CCDȦḂ SCD + 12
1
R SȦḂ + CȦḂĊḊ SĊḊ ,
and from (2.62), the spinor equivalent of the curvature tensor is given by
RAȦBḂCĊDḊ = CABCD εȦḂ εĊḊ + CȦḂĊḊ εAB εCD

+ CABĊḊ εȦḂ εCD + CCDȦḂ εAB εĊḊ
+ 24
1
R(εAC εBD εȦḂ εĊḊ + εAD εBC εȦḂ εĊḊ
+ εABεCD εȦĊ εḂḊ + εAB εCD εȦḊ εḂĊ ). (2.83)
Making use of the relation
εAB εCD + εAC εDB + εAD εBC = 0, (2.84)

which is equivalent to (1.27), and its dotted version, one can rewrite in various ways
the expression between parentheses in the last two lines of (2.83), e.g.,
2(εAB εCD εȦĊ εḂḊ + εAD εBC εȦḂ εĊḊ ),
2(εAB εCD εȦḊ εḂĊ + εAC εBD εȦḂ εĊḊ ),

and
2(εAC εȦĊ εBD εḂḊ − εAD εȦḊ εBC εḂĊ );
according to (1.21), the last expression is the spinor equivalent of 2(gac gbd −
gad gbc ).
Substituting (2.76) and (2.77) into (2.64), one finds that the spinor equivalent of
the Ricci tensor is related to CABȦḂ by RBḂDḊ = 2CBDḂḊ − 14 RεBD εḂḊ , i.e.,
CABȦḂ = 12 (RAȦBḂ + 14 RεAB εȦḂ ). (2.85)
Thus, recalling that −εAB εȦḂ is the spinor equivalent of gab , (2.85) shows that CABȦḂ
is the spinor equivalent of 12 (Rab − 14 Rgab ); that is, apart from a factor of 12 , CABȦḂ
is the spinor equivalent of the traceless part of the Ricci tensor.
Making use of (1.27), it can readily be seen that
CABĊḊ εȦḂ εCD + CCDȦḂ εAB εĊḊ = CADȦḊ εBC εḂĊ − CACȦĊ εBD εḂḊ
+ CBCḂĊ εAD εȦḊ − CBDḂḊ εAC εȦĊ ,
which means that the second line of (2.83) is the spinor equivalent of the tensor
2 (Rad − 4 Rgad )(−gbc ) − 2 (Rac − 4 Rgac )(−gbd ) + 2 (Rbc − 4 Rgbc )(−gad ) − 2 (Rbd
1 1 1 1 1 1 1
− 14 Rgbd )(−gac ) = − 12 (Rad gbc − Rac gbd + Rbc gad − Rbd gac ) + 14 (gad gbc − gac gbd );
hence, it follows from (2.83) that CABCD εȦḂ εĊḊ + CȦḂĊḊ εAB εCD is the spinor
equivalent of
Cabcd ≡ Rabcd + 12 (Rad gbc − Rac gbd + Rbc gad − Rbd gac ) + 16 R(gac gbd − gad gbc ).
(2.86)
The tensor field Cabcd is known as the Weyl tensor, and CABCD and CȦḂĊḊ are called
Weyl spinors. Owing to its behavior under conformal rescalings, Cabcd is also called
the conformal curvature tensor (see Section 2.3).
The tensor field Cabcd possesses all the symmetries of the curvature tensor Rabcd
(i.e., Cabcd = −Cbacd = −Cabdc , Cabcd + Cacdb + Cadbc = 0, Cabcd = Ccdab ), and in
addition, Ca bad = 0, as can readily be seen using the total symmetry of CABCD and
CȦḂĊḊ [see (1.29)].
Since the curvature tensor is real, making use of (2.83), (1.224), and (1.221) one
finds that when the signature is (+ + + +) or (+ + − −), the curvature spinors
must satisfy
CABCD = CABCD , CȦḂĊḊ = CȦḂĊḊ , CABĊḊ = CABĊḊ , (2.87)
while when the signature is (+ + + −) one has

2.2 Curvature 83
CABCD = CȦḂĊḊ , CABĊḊ = CCDȦḂ (2.88)
[see (1.47)].
Combining (1.41) and (2.62), we see that
XABCD εȦḂ εĊḊ + CCDȦḂ εAB εĊḊ + CABĊḊ εȦḂ εCD + XȦḂĊḊ εAB εCD
1
= σ a AȦ σ b BḂ σ cCĊ σ d DḊ Rabcd ;
4
hence [see (1.69)],
1 ab
XABCD = S Scd Rabcd ,
16 AB CD
1 ab
CABĊḊ = S Scd Rabcd , (2.89)
16 AB ĊḊ
1 ab
XȦḂĊḊ = S Scd Rabcd
16 ȦḂ ĊḊ
[cf. (2.85)]. Since CABCD , CABĊḊ , and R are independent of each other, from (2.89)
we have, for instance,
1 ab
CABCD = S Scd Cabcd .
16 AB CD
The spinor components of the curvature can be expressed in terms of the spin co-
efficients by substituting (2.16) into (2.68), making use of (2.15) and (2.55). Since
the differential of a scalar function is expressed locally as d f = (∂a f ) θ a , equiva-
lently we have
d f = −(∂AȦ f ) θ AȦ . (2.90)
In this way we obtain the explicit expressions
XABCD = ∂(C Ḋ Γ|AB|D)Ḋ − ΓAB(C|ĊΓ ĊṘ Ṙ|D) + ΓABRĊΓ R (CD)Ċ + ΓAR (C Ḋ Γ|BR|D)Ḋ (2.91)
and
CABĊḊ = ∂ R (Ċ Γ|ABR|Ḋ) + ΓABCṘΓ Ṙ (ĊḊ)C − ΓABC(ĊΓ CR |R|Ḋ) + ΓA RD (Ċ Γ|BRD|Ḋ) , (2.92)
together with
CCDȦḂ = ∂(C ṘΓ|ȦḂṘ|D) + ΓȦḂĊRΓ R (CD)Ċ − ΓȦḂĊ(C Γ ĊṘ |Ṙ|D) + ΓȦ ṘḊ (C Γ|ḂṘḊ|D) (2.93)
and
XȦḂĊḊ = ∂ D (Ċ Γ|ȦḂ|Ḋ)D − ΓȦḂ(Ċ|C Γ CR R|Ḋ) + ΓȦḂṘC Γ Ṙ (ĊḊ)C + ΓȦ Ṙ (Ċ DΓ|ḂṘ|Ḋ)D . (2.94)
As pointed out above, the spin coefficients can be obtained by means of the com-
mutators (2.23) or from (2.60) and (2.61). Substitution of (2.16) and (2.71) into
(2.60) and (2.61) gives
dSAB = −Γ (AC B) Ḋ θ̆ CḊ − Γ (AC |C| Ḋ θ̆ B)Ḋ (2.95)
and
dSȦḂ = Γ (ȦĊ Ḃ) D θ̆ DĊ + Γ (ȦĊ |Ċ D θ̆ D|Ḃ) . (2.96)
The combinations of the spin coefficients appearing in (2.95) and (2.96) allow
us to find the spin coefficients individually. Writing, say, dSAB = K ABCḊ θ̆ CḊ ,
with KABCḊ = K(AB)CḊ , and comparing with (2.95) one has KABCḊ = −ΓC(AB)Ḋ −
Γ(A|R R Ḋ| εB)C . Then, from this last relation one readily obtains
ΓABCḊ = −KC(AB)Ḋ (2.97)
[cf. (2.12)] (see the examples below). In a similar way one finds that
ΓȦḂĊD = KĊ(ȦḂ)D (2.98)
if the functions KȦḂĊD are defined by dSȦḂ = K ȦḂĊD θ̆ DĊ (cf. Israel 1979 and the
references cited therein).
Frequently, the metric tensor of M is given by a local expression in terms of some
coordinate system, and in order to find a set of one-forms θ AȦ , it is convenient to
notice that from (2.1) we obtain
g = −εAB εȦḂ θ AȦ ⊗ θ BḂ

= −θ 11̇ ⊗ θ 22̇ − θ 22̇ ⊗ θ 11̇ + θ 12̇ ⊗ θ 21̇ + θ 21̇ ⊗ θ 12̇
= −2θ 11̇ θ 22̇ + 2θ 12̇θ 21̇ , (2.99)
where the juxtaposition of one-forms means symmetrized tensor product, and

⎧
⎪ −θAḂ if the σ a AḂ are given by (1.33),
⎪
⎨ θ BȦ if the σ a AḂ are given by (1.34),
θ AḂ = (2.100)
⎪ θ AḂ
⎩ AC ḂḊ
η η θCḊ if the σ a AḂ are given by (1.37).
Thus, instead of looking for an orthonormal or null tetrad, a suitable set of one-forms
θ AȦ can be obtained directly by expressing the metric tensor in the form (2.99) in
terms of one-forms satisfying the appropriate condition (2.100) (see the examples
below).
The behavior of the differential forms SAB, SȦḂ , and θ̆ AḂ under complex conju-
gation can be derived from (2.100) and the relations (2.57) and (2.71).
Example 2.1. The standard metric of S4 .

The metric induced on the sphere S4 ≡ {(x1 , x2 , x3 , x4 , x5 ) ∈ R5 | (x1 )2 + (x2 )2 +
(x )2 + (x4 )2 + (x5 )2 = 1} by the standard Euclidean metric of R5 is locally given by
3
g = dψ 2 + sin2 ψ d χ 2 + sin2 χ (dθ 2 + sin2 θ dϕ 2 ) , (2.101)

2.2 Curvature 85
in terms of spherical coordinates ψ , χ , θ , and ϕ . We can equivalently write

1 1
g = −2 − √ (dψ − i sin ψ dχ ) √ (dψ + i sin ψ dχ )
2 2

1 1
+2 − √ sin ψ sin χ (dθ − i sin θ dϕ ) − √ sin ψ sin χ (dθ + i sin θ dϕ ) ,
2 2
and by comparing with (2.99) one finds that the one-forms θ AȦ satisfying the con-
dition θ AḂ = −θAḂ (i.e., θ 11̇ = −θ 22̇ and θ 12̇ = θ 21̇ ), appropriate for a metric with
Euclidean signature, can be taken as
1 1
θ 11̇ = − √ (dψ − i sin ψ dχ ), θ 22̇ = √ (dψ + i sin ψ dχ ),
2 2
sin ψ sin χ sin ψ sin χ
θ 12̇ = − √ (dθ − i sin θ dϕ ), θ 21̇ = − √ (dθ + i sin θ dϕ ).
2 2
(2.102)
(Alternatively, from (2.101) we see that
θ 1 = sin ψ sin χ dθ , θ 2 = sin ψ sin χ sin θ dϕ ,

θ 3 = dψ , θ 4 = sin ψ dχ ,
form the dual basis of an orthonormal tetrad. The one-forms (2.102) are then ob-
tained with the aid of (2.14) and (1.33).)
The two-forms SAB are then given by [see (2.58)]
S11 = 1
2 sin ψ sin χ (dψ − i sin ψ dχ ) ∧ (dθ − i sin θ dϕ ),
(2.103)
S12 = − 12 i(sin ψ dψ ∧ d χ + sin2 ψ sin2 χ sin θ dθ ∧ dϕ ),
with S22 = S11 . The two-forms SȦḂ can be readily obtained by noticing that the
one-forms (2.102) have the property that under the substitution of ϕ by −ϕ , θ AḂ is
mapped into θ BȦ , and therefore SAB is mapped into SȦḂ [see (2.58)]. A straightfor-
ward computation, making use of (2.71), gives

cot θ 1 cot χ
dS = − √
11
θ̆ − √ 2 cot ψ − i
11̇
θ̆ 12̇ ,
2 sin ψ sin χ 2 sin ψ

1 cot χ 1 cot χ
dS12 = − √ cot ψ + i θ̆ 11̇ − √ cot ψ − i θ̆ 22̇ .
2 sin ψ 2 sin ψ
Hence, the only nonvanishing spin coefficients ΓABCḊ are determined by [see (2.97)
and (2.20)]

1 cot χ 1
Γ1121̇ = − √ cot ψ + i , Γ1211̇ = Γ1222̇ = √ cot ψ ,
2 sin ψ 2 2
1 cot θ
Γ1212̇ = −Γ1221̇ = √ ,
2 2 sin ψ sin χ
and the connection one-forms IΓAB for this tetrad are
1
IΓ11 = − (cos ψ sin χ + i cos χ )(dθ + i sin θ dϕ ),
2
i
IΓ12 = − (cos ψ dχ + cos θ dϕ ), (2.104)
2
1
IΓ22 = − (cos ψ sin χ − i cos χ )(dθ − i sin θ dϕ ).
2
Substituting into (2.68), one finds that the curvature two-forms RAB are given ex-
plicitly by
R11 = dIΓ11 − 2IΓ12 ∧ IΓ11 = S22 ,

R12 = dIΓ12 + IΓ11 ∧ IΓ22 = −S12 , (2.105)
R22 = dIΓ22 + 2IΓ12 ∧ IΓ22 = S ,11
and hence CABȦḂ = 0, CABCD = 0, and R = 12 [see (2.82)]. In a similar way one finds
that CȦḂĊḊ = 0.
Example 2.2. The Schwarzschild metric.

The Schwarzschild metric, which, according to Einstein’s theory of gravitation,
represents the exterior gravitational field of a spherically symmetric distribution of
matter, is usually expressed in the form
rg 2 2 dr2
g = − 1− c dt + + r2 dθ 2 + r2 sin2 θ dϕ 2 , (2.106)
r 1 − rg/r
where rg is a constant (called the gravitational radius or Schwarzschild radius,

rg = 2GM/c2 , where G is Newton’s constant of gravitation, M corresponds to the
Newtonian mass of the source, and c is the speed of light in vacuum), and t, r, θ , and
ϕ are real coordinates (see, e.g., Rindler 1977, Wald 1984). When rg = 0, (2.106)
is the metric of Minkowski space-time in spherical coordinates. By comparing the
metric (2.106), written as

1 rg 1 rg −1
g = −2 1− cdt − dr cdt + 1 − dr
2 r 2 r

r r
+ 2 − √ (dθ − i sin θ dϕ ) − √ (dθ + i sin θ dϕ ) ,
2 2
with (2.99) one finds that the one-forms θ AḂ satisfying the condition θ AḂ = θ BȦ ,
appropriate for Lorentzian signature, can be taken as
2.2 Curvature 87
1 rg 1 rg −1
θ 11̇ = 1− cdt − dr, θ 22̇ = cdt + 1 − dr,
2 r 2 r
(2.107)
r r
θ 12̇ = − √ (dθ − i sin θ dϕ ), θ 21̇ = − √ (dθ + i sin θ dϕ ),
2 2
and hence,
rg −1 1 ∂ ∂ 1 ∂ 1 rg ∂
∂11̇ = − 1 − + , ∂22̇ = − − 1− ,
r c ∂t ∂r 2c ∂ t 2 r ∂r
(2.108)
1 ∂ i ∂ 1 ∂ i ∂
∂12̇ = √ + , ∂21̇ = √ − .
2r ∂ θ sin θ ∂ ϕ 2r ∂ θ sin θ ∂ ϕ
(The coefficients of ∂ /∂ t are chosen negative, so that vAȦ = l A l Ȧ is the spinor equiv-
alent of a future-pointing null vector, for any one-index spinor lA different from
zero.)
The commutators of the vector fields (2.108) can be readily computed, and mak-
ing use of the notation (2.28), we obtain
1 1 rg
[D, δ ] = − δ , [Δ , δ ] = 1− δ,
r 2r r
rg cot θ
[D, Δ ] = − 2 D, [δ , δ ] = √ (δ − δ ),
2r 2r
which, compared with (2.27), implies that the only nonvanishing spin coefficients
are
1 cot θ cot θ rg 1 rg
ρ =− , β = √ , α =− √ , γ = 2, μ =− 1− ;
r 2 2r 2 2r 4r 2r r
(2.109)
therefore, the connection one-forms are given by
1
IΓ11 = − √ (dθ + i sin θ dϕ ),
2

i rg rg −1
IΓ12 = − cos θ dϕ − 2 cdt + 1 − dr , (2.110)
2 4r r
1 rg
IΓ22 = − √ 1 − (dθ − i sin θ dϕ ),
2 2 r
and, substituting into (2.68), one finds that the curvature two-forms RAB are given
explicitly by
rg 22
R11 = dIΓ11 − 2IΓ12 ∧ IΓ11 = − S ,
2r3
rg 12
R12 = dIΓ12 + IΓ11 ∧ IΓ22 = − S , (2.111)
r3
rg
R22 = dIΓ22 + 2IΓ12 ∧ IΓ22 = − 3 S11 ,
2r
which implies that CABȦḂ = 0, R = 0 (that is, Rab = 0, thus showing that the Einstein
vacuum field equations are satisfied), and the only nonvanishing components of the
Weyl spinor are determined by
rg
C1122 = − 3 (2.112)
2r
[see (2.82)].
Example 2.3. The Euclidean Schwarzschild metric.
By reversing the sign of the first term in (2.106) we obtain a metric with
Euclidean signature for r > rg
rg 2 2 dr2
g = 1− c dt + + r2 dθ 2 + r2 sin2 θ dϕ 2 , (2.113)
r 1 − rg/r
which can be expressed in the form −2θ 11̇ θ 22̇ + 2θ 12̇ θ 21̇ , with θ 11̇ = −θ 22̇ and
θ 12̇ = θ 21̇ , choosing

i rg 1 dr
θ = √
11̇
1 − cdt − √ ,
2 r 2 1 − rg/r

i rg 1 dr
θ = √
22̇
1 − cdt + √ ,
2 r 2 1 − rg/r
r r
θ 12̇ = − √ (dθ − i sin θ dϕ ), θ 21̇ = − √ (dθ + i sin θ dϕ ).
2 2
A straightforward computation gives

r rg dr
S =−
11
i 1 − cdt − ∧ (dθ − i sin θ dϕ ),
2 r 1 − rg/r
i
S12 = cdt ∧ dr − r2 sin θ dθ ∧ dϕ ,
2
with S22 = S11 , and

cot θ 11̇ 1 rg rg /r
dS11 = √ θ̆ − √ 1− + θ̆ 12̇ ,
2r 2r r 2 1 − rg/r

1 rg
dS12 = −√ 1 − (θ̆ 11̇ + θ̆ 22̇ ).
2r r
Then, using the fact that when the signature is Euclidean, KABCḊ = −K ABCḊ , the
foregoing relations and (2.97) give

1 rg
IΓ11 = − 1 − (dθ + i sin θ dϕ ),
2 r
i irg
IΓ12 = − cos θ dϕ − 2 cdt,
2 4r
2.2 Curvature 89
with IΓ22 = IΓ11 [cf. (2.110)]. The curvature two-forms are

rg 22 rg 12
R11 = − S , R12 = − S ,
2r3 r3
with R22 = R11 , showing that CABĊḊ = 0 = R and the only nonvanishing component
of the Weyl spinor CABCD is given by C1122 = −rg /(2r3 ) [real, as follows from
(2.87)].
Further examples are given in Sections 2.3 and 4.2.
Bianchi and Ricci Identities
The Bianchi identities in spinor form can be derived from (2.54) or directly by
applying the exterior derivative operator to both sides of the Cartan second struc-
ture equations (2.68), using again these equations in order to eliminate the exterior
derivative of IΓAB or IΓȦḂ . Thus
dRAB = 2 IΓC(A ∧ RC B) ,
(2.114)
dRȦḂ = 2 IΓĊ(Ȧ ∧ RĊ Ḃ) .
Substituting (2.67) and (2.16) into (2.114), with the aid of (2.71), (2.77), (2.60), and
(2.61) one obtains
∇C Ṙ XABCD − ∇DĊCABĊṘ = 0,
(2.115)
∇RĊ XȦḂĊḊ − ∇C ḊCCRȦḂ = 0,
and hence, employing the decompositions (2.80) and (2.81), one finds that these last
equations are equivalent to
∇C ṘCCABD − ∇(AĊCBD)ĊṘ = 0,
∇RĊCĊȦḂḊ − ∇C (ȦC|CR|ḂḊ) = 0, (2.116)
8 ∇DṘ R + ∇ CBDĊṘ = 0.
1 BĊ
Since CABĊḊ is the spinor equivalent of 12 (Rab − 14 Rgab ), the last of these relations
amounts to the so-called contracted Bianchi identities.
Applying the decomposition (1.52) to the commutator of covariant derivatives,
we have
∇AȦ ∇BḂ − ∇BḂ∇AȦ = εȦḂ AB + εAB ȦḂ , (2.117)
where
AB ≡ ∇(A Ṙ ∇B)Ṙ , ȦḂ ≡ ∇R (Ȧ ∇|R|Ḃ) , (2.118)
and the spinor equivalent of the Ricci identity (∇a ∇b − ∇b ∇a )tc = −Rd cabtd , which
follows from (2.48), can be written in the form
(εȦḂ AB + εAB ȦḂ )tCĊ = RDḊCĊAȦBḂtDḊ . (2.119)
Then, making use of (2.83) one finds that (2.119) is equivalent to
ABtCĊ = CDCABtDĊ + CABḊĊ tCḊ + 24

1
R(εCBtAĊ + εCAtBĊ ) (2.120)
and
ȦḂ tCĊ = CḊĊȦḂtCḊ + CDCȦḂtDĊ + 24
1
R(εĊḂtCȦ + εĊȦtCḂ ). (2.121)
Equations (2.120) and (2.121) also follow directly from the definitions (2.118) and
the explicit expressions (2.91)–(2.94).
The commutator of covariant derivatives of a scalar function vanishes; in fact,
(∇a ∇b − ∇b ∇a ) f = ∇a ∂b f − ∇b ∂a f = ∂a ∂b f − Γ c ba ∂c f − ∂b ∂a f + Γ c ab ∂c f , which
is equal to zero if and only if (2.11) holds. Thus, AB f = 0 and ȦḂ f = 0.
In the case of one-index spinor fields, making use of (2.40), (2.80), (2.81), and
(2.91)–(2.94), one finds that
AB ψC = CDCAB ψD + 24 R(εCB ψA + εCA ψB ),

1
(2.122)
AB ψĊ = CAB ḊĊ ψḊ , (2.123)
and
ȦḂ ψC = CDCȦḂ ψD , (2.124)
ȦḂ ψĊ = CḊĊȦḂ ψḊ + 24 R(εĊḂ ψȦ + εĊȦ ψḂ ).

1
(2.125)
These identities imply the identities (2.120) and (2.121) by virtue of the formulas
AB (ψC...Ḋ... φE...Ḟ... ) = ψC...Ḋ... AB φE...Ḟ... + φE...Ḟ... AB ψC...Ḋ...
and
ȦḂ (ψC...Ḋ... φE...Ḟ... ) = ψC...Ḋ... ȦḂ φE...Ḟ... + φE...Ḟ... ȦḂ ψC...Ḋ... ,
which follow from the fact that the covariant derivative satisfies the Leibniz rule. In
this manner one finds that
AB ψCD...Ė Ḟ... = CRCAB ψRD...Ė Ḟ... + CRDAB ψCR...Ė Ḟ... + · · ·

+ 12
1
R εC(A ψB)D...Ė Ḟ... + εD(A ψ|C|B)...Ė Ḟ... + · · ·
+ CABṘ Ė ψCD...ṘḞ... + CABṘ Ḟ ψCD...Ė Ṙ... + · · · (2.126)
and
ȦḂ ψCD...Ė Ḟ... = CṘ Ė ȦḂ ψCD...ṘḞ... + CṘḞ ȦḂ ψCD...Ė Ṙ... + · · ·

+ 12 1
R εĖ(Ȧ ψ|CD...|Ḃ)Ḟ... + εḞ(Ȧ ψ|CD...Ė|Ḃ)... + · · ·
+ CRCȦḂ ψRD...Ė Ḟ... + CR DȦḂ ψCR...Ė Ḟ... + · · · . (2.127)

2.2 Curvature 91
2.2.2 Algebraic Classification of the Conformal Curvature
Since the conformal curvature spinors CABCD and CȦḂĊḊ are totally symmetric,
they can be classified according to the multiplicities of their principal spinors (see
Section 1.7).
If the signature is (+ + + +), the reality conditions (2.87) imply that the only
possible nontrivial cases are

B βC βD) with α A βA = 0,
α(A α
CABCD = (2.128)
±α(A αB αC αD)
[see (1.231)], with analogous independent expressions for CȦḂĊḊ . For instance,
the spinor CABCD of the Euclidean Schwarzschild metric is of the form CABCD =
−α(A αB αC α
D) .
Similarly, if the metric has signature (+ + − −), the only nontrivial possible
cases are
⎧
⎪
⎪ α(A βB γC δD) A = αA , βA = βA , γA = γA , δA = δA ,
with α
⎪
⎪
⎪
⎪ α(A βB γC γD) A = αA , βA = βA ,
with α
⎪
⎪
⎪
⎪
⎪
⎪
⎪ ±α(A αB βC βD) ,
⎪
⎪
⎪
⎪ A = αA , βA = βA , γA = γA ,
⎨ α(A αB βC γD) with α
CABCD = ±α α β β with α A = αA , (2.129)
⎪
⎪ (A B C D)
⎪
⎪
⎪
⎪ ±α(A αB βC βD) with α A = αA , βA = βA ,
⎪
⎪
⎪
⎪ ±α(A αB α C αD) ,
⎪
⎪
⎪
⎪
⎪ α(A αB αC βD) with α
⎪ A = αA , βA = βA ,
⎪
⎩
±αA αB αC αD with α A = αA ,
with analogous independent expressions for CȦḂĊḊ . In particular, there exist

Riemannian manifolds with ultrahyperbolic signature with CABCD = 0 and CȦḂĊḊ =
0 (or vice versa), which turn out to possess interesting applications (see, e.g.,
Dunajski and West 2008). (The algebraic types (2.128) and (2.129) coincide with
the possible algebraic types of a traceless two-index symmetric tensor in a manifold
of dimension three; Torres del Castillo 2003, Sec. 5.3.)
It should be remarked that when the signature of the metric is (+ + + +) or
(+ + − −), the algebraic type of the Weyl spinors is not completely determined by
the multiplicities of their principal spinors, but in some cases there is an overall sign
that has to be specified (cf. Goldblatt 1994a,b and the references cited therein).
In the remaining case, in which the metric has signature (+ + + −), the reality
conditions (2.88) do not impose any restrictions on the principal spinors of CABCD ,
and therefore, if the conformal curvature spinor is different from zero, it can be of
the form
⎧
⎪ α β γ δ (type G or {1, 1, 1, 1}),
⎪ (A B C D)
⎪
⎪
⎨ α(A αB βC γD) (type II or {2, 1, 1}),
CABCD = α(A αB βC βD) (type D or {2, 2}), (2.130)
⎪
⎪
⎪ α(A αB αC βD)
⎪
⎩
(type III or {3, 1}),
αA αB αC αD (type N or {4}),
where it is understood that all the contractions α A βA , α A γA , α A δA , β A γA , β A δA ,
and γ A δA are different from zero. This constitutes the Petrov–Penrose classifica-
tion of the conformal curvature (Penrose 1960, Pirani 1965). Only if the metric has
Lorentzian signature are CABCD and CȦḂĊḊ not independent of each other, being re-
lated by CȦḂĊḊ = CABCD [see (2.88)], and therefore it suffices to consider CABCD
only. Furthermore, only in this case is there associated with each one-index spinor
αA another one-index spinor α Ȧ = αA , so that αA is a principal spinor of CABCD if
and only if α Ȧ is a principal spinor of CȦḂĊḊ . If αA is a principal spinor of CABCD ,
then ±αA α Ȧ is the spinor equivalent of a real null vector referred to as a Debever–
Penrose (DP) vector, and its direction is a principal null direction of the conformal
curvature.
The DP vectors can be defined directly using the conformal curvature tensor. If
la is a real null vector, then the spinor equivalent of l cCabcd is

± εȦḂ α ḊCABCD α C + εABαDCȦḂĊḊ α Ċ ,
assuming that ±α A α Ȧ is the spinor equivalent of l a ; hence

l cCabcd = 0 ⇔ CABCD α C = 0,
which means that αA is a quadruple principal spinor of CABCD . Similarly, the spinor
equivalent of l cCabc[d le] is [see (1.52)]
− 12 (εDE εȦḂ α Ḋ α Ė CABCR α C α R + εABεḊĖ αD αE CȦḂĊṘ α Ċ α Ṙ ),
and hence
l cCabc[d le] = 0 ⇔ CABCR α C α R = 0,
which means that αA is at least a triple principal spinor of CABCD . It can readily be
seen that the spinor equivalent of l b l cCabc[d le] is

± 12 εDE α Ȧ α Ḋ α Ė CABCR α B α C α R + εḊĖ αA αD αE CȦḂĊṘ α Ḃ α Ċ α Ṙ ,
and therefore
l b l cCabc[d le] = 0 ⇔ CABCR α B α C α R = 0,
which means that αA is at least a double principal spinor of CABCD . Finally, the
spinor equivalent of l b l c l[ f Ca]bc[d le] is

1
4 εDE εFA α Ḟ α Ȧ α Ḋ α Ė CSBCR α S α B α C α R + εḊĖ εḞ Ȧ αF αA αD αE CṠḂĊṘ α Ṡ α Ḃ α Ċ α Ṙ ,
2.3 Conformal Rescalings 93
and therefore
l b l c l[ f Ca]bc[d le] = 0 ⇔ CSBCR α S α B α C α R = 0, (2.131)
which means that αA is a principal spinor of CABCD .

By virtue of the equivalence (2.131), it follows that there exist four (not necessar-
ily distinct) real null directions represented by null vectors l a satisfying the condition
l b l c l[ f Ca]bc[d le] = 0. If the four principal null directions of the conformal curvature
are all distinct [which corresponds to the type G or {1,1,1,1} in (2.130)], the cur-
vature is said to be algebraically general. The conformal curvature is algebraically
special or degenerate if there are fewer than four distinct principal null directions
(which corresponds to the existence of a repeated principal spinor of CABCD ).
2.3 Conformal Rescalings
Two metric tensors g and g on M are conformally equivalent if there exists a positive
function φ such that g = φ 2 g . Given a null tetrad ∂AȦ for the metric g,
∂ AȦ = φ ∂AȦ (2.132)
is a null tetrad for g . Note that since the values of φ are real, the transformation
(2.132) preserves conditions (2.100). Each metric tensor defines a Riemannian con-
nection, and in order to find the spin coefficients corresponding to the Riemannian
connection defined by g with respect to the null tetrad ∂ AȦ , we note that the two-
forms SAB and their analogues SAB are then related by SAB = φ −2 SAB; therefore,
writing dSAB = K ABCḊ θ̆ CḊ as in Section 2.2, and similarly, dSAB = K ABCḊ θ̆ CḊ ,
we have
dSAB = φ −2 dSAB − 2φ −3dφ ∧ SAB

= φ −2 K ABCḊ θ̆ CḊ + 2φ −3 (∂RḊ φ ) θ RḊ ∧ SAB
= (φ K ABCḊ + 2∂ (A Ḋ φ ε B)C ) θ̆ CḊ
[see (2.71) and (2.90)], i.e., K ABCḊ = φ KABCḊ + εAC ∂BḊ φ + εBC ∂AḊ φ . Thus, accord-
ing to (2.97), Γ ABCḊ = −K C(AB)Ḋ , that is,
Γ ABCḊ = φ ΓABCḊ − εC(A ∂B)Ḋ φ . (2.133)
In a similar manner one finds that
Γ ȦḂĊD = φ ΓȦḂĊD − εĊ(Ȧ ∂|D|Ḃ) φ . (2.134)
Equations (2.133) and (2.134) give the spin coefficients of the Riemannian connec-
tion corresponding to g with respect to the null tetrad ∂ AȦ .
The spinor components of the curvature of the rescaled metric g with respect
to the tetrad ∂ AȦ can be obtained by substituting (2.132), (2.133), and (2.134) into
(2.91)–(2.94); in this way, from (2.91) one obtains
X ABCD = φ 2 XABCD + 14 (εAC εBD + εAD εBC ) φ 3 ∇RṘ ∇RṘ φ −1 ,
which is equivalent to the relations
CABCD = φ 2CABCD , R = φ 2 R + 6 φ 3 ∇RṘ ∇RṘ φ −1 , (2.135)
whereas (2.92) yields
CABĊḊ = φ 2CABĊḊ + φ ∇A(Ċ ∇|B|Ḋ) φ . (2.136)
Similarly, from (2.93) we obtain
CABĊḊ = φ 2CABĊḊ + φ ∇(A|Ċ| ∇B)Ḋ φ , (2.137)
which is equivalent to (2.136), while from (2.94) we obtain
CȦḂĊḊ = φ 2CȦḂĊḊ . (2.138)
It should be stressed that equations (2.135)–(2.138) give the relation between

the components of the curvature spinors of g and g with respect to two different
null tetrads related by (2.132). With the aid of (2.135)–(2.138), one finds that the
components of the curvature tensors of g and g (with respect to the basis induced
by a coordinate system) are related by
Rμν = Rμν + 2φ −1 ∇μ ∇ν φ + φ −1 gμν ∇ρ ∇ρ φ − 3φ −2 gμν (∇ρ φ )(∇ρ φ )
and

Cμνρσ = φ −2Cμνρσ ,
and therefore, C μ νρσ = C μ νρσ (see Appendix A).
Only CABCD and CȦḂĊḊ transform homogeneously under conformal rescalings,
and therefore, CABCD and CȦḂĊḊ vanish for a metric conformally equivalent to a flat
one. It can be shown that conversely, if CABCD and CȦḂĊḊ are equal to zero, then the
metric tensor g is locally conformally equivalent to a flat metric.
For instance, the curvature of the standard metric of the sphere S4 satisfies
CABCD = 0 = CȦḂĊḊ , and we can see explicitly that this metric is indeed locally
conformally flat. In fact, letting r ≡ cot 12 ψ , from (2.101) we obtain

dψ 2 + sin2 ψ dχ 2 + sin2 χ (dθ 2 + sin2 θ dϕ 2 )

2
2 2

= dr + r2 d χ 2 + sin2 χ (dθ 2 + sin2 θ dϕ 2 ) .
1+r 2
2.4 Killing Vectors. Lie Derivative of Spinors 95
The expression within braces can be recognized as the standard metric of R4 , which
is flat, written in spherical coordinates. (Note that the conformal factor φ = 2/(1 +
r2 ) = 2 sin2 21 ψ vanishes at ψ = 0.)
Example 2.4. The hyperbolic space.
The hyperbolic space H4 ≡ {(x1 , x2 , x3 , x4 ) ∈ R4 | x4 > 0} possesses the metric

g = (x4 )−2 (dx1 )2 + (dx2)2 + (dx3 )2 + (dx4 )2
conformally equivalent to the flat metric g = (dx1 )2 + (dx2 )2 + (dx3 )2 + (dx4 )2 .

Thus, in the notation employed throughout this section, φ = x4 . The Laplacian of
φ −1 = (x4 )−1 , relative to the flat metric g, can be readily computed by making use of
Cartesian coordinates xa , and one obtains ∇a ∇a φ −1 = −∇a (φ −2 δa4 ) = 2φ −3 . Sim-
ilarly, ∇a ∇b φ = 0, which amounts to ∇AḂ ∇CḊ φ = 0. Then, from (2.135)–(2.138)
we see that the only nonvanishing component of the curvature of g is the scalar
curvature
R = −12.
2.4 Killing Vectors. Lie Derivative of Spinors
The infinitesimal generator K of a one-parameter group of isometries satisfies the

condition £K g = 0, where £K denotes the Lie derivative with respect to the vector
field K (see, e.g., Kobayashi and Nomizu 1963, Petersen 1998), or equivalently, in
terms of the Levi-Civita connection
∇a Kb + ∇b Ka = 0. (2.139)
These equations are referred to as the Killing equations, and their solutions are
called Killing vector fields, or simply Killing vectors. In a four-dimensional mani-
fold, the Killing equations constitute a set of ten independent linear partial differ-
ential equations, and the set of Killing vector fields is a real Lie algebra, with the
commutator or Lie bracket [ , ], of dimension less than or equal to ten. Thus, a vector
field K is a Killing vector field if and only if ∇a Kb is antisymmetric and therefore
its spinor equivalent is of the form (1.50),
∇AȦ KBḂ = LAB εȦḂ + LȦḂ εAB , (2.140)
with LAB and LȦḂ being symmetric spinor fields. From (2.140) we obtain
LAB = 12 ∇A Ṙ KBṘ , LȦḂ = 12 ∇R Ȧ KRḂ . (2.141)
The spinor components LAB and LȦḂ can be conveniently calculated by express-
ing the differential of the one-form Ka θ a in terms of the two-forms SAB and SȦḂ ,
since
d(Ka θ a ) = ∇(A Ṙ KB)Ṙ SAB + ∇R (Ȧ K|R|Ḃ) SȦḂ (2.142)
(see Exercise 2.5). In this way, the spin coefficients need not be known in order to
find LAB and LȦḂ .
The Lie derivative £K X of an arbitrary vector field X = X a ∂a = −X AȦ ∂AȦ with
respect to a Killing vector K = K a ∂a = −K AȦ ∂AȦ coincides with their commuta-
tor, or Lie bracket, [K, X]; therefore, if the torsion of the connection vanishes, then
£K X = [K, X] = ∇K X − ∇X K [see (2.5)], and making use of (2.140), we find that the
spinor equivalent of £K X is determined by
£K X = (K AȦ ∇AȦ X BḂ − X AȦ∇AȦ K BḂ ) ∂BḂ

= (K AȦ ∇AȦ X BḂ + LB A X AḂ + LḂ Ȧ X BȦ) ∂BḂ
≡ −(£K X BḂ ) ∂BḂ .
This result suggests the following definition for the Lie derivative of an arbitrary
spinor field with respect to a Killing vector:
£K ψC...Ḋ... = −K ∇RṠ ψC...Ḋ... − L S ψC...Ḋ... − · · · − L Ṡ ψC...Ḋ...

A...Ḃ... RṠ A...Ḃ... A S...Ḃ... Ḃ A...Ṡ...
− · · · + LSC ψS...Ḋ... + · · · + L Ḋ ψC...Ṡ... + · · · .

A...Ḃ... Ṡ A...Ḃ...
(2.143)
Then, for instance, £K εAB = 0 and £K εȦḂ = 0.
The Killing equations (2.140) have some integrability conditions. From (2.141)
and (2.117) we obtain
∇AȦ LBC = 12 ∇AȦ ∇B Ṙ KCṘ

= 12 (∇AȦ ∇B Ṙ − ∇B Ṙ ∇AȦ + ∇BṘ ∇AȦ )KCṘ
= 12 (εAB Ȧ + εȦ
Ṙ Ṙ
AB + ∇B ∇AȦ )KCṘ
Ṙ
= 12 εAB Ȧ KCṘ − 2 AB KCȦ − 2 ∇B ∇CṘ KAȦ ,

Ṙ 1 1 Ṙ
where we have made use of the fact that ∇AȦ KCṘ = −∇CṘ KAȦ [see (2.139)]. Since
the left-hand side of the last equation is symmetric on the indices BC, we have
∇AȦ LBC = 12 εA(B |Ȧ|

Ṙ
KC)Ṙ − 12 A(B KC)Ȧ − 2
1
BC KAȦ
= −C ABC KDȦ − CBC Ȧ KAḊ − 12 RεA(B KC)Ȧ ,

D Ḋ 1
(2.144)
by virtue of the Ricci identities (2.120) and (2.121). Analogously, we obtain
∇AȦ LḂĊ = −CḊ ȦḂĊ KAḊ − CD AḂĊ KDȦ − 12

1
RεȦ(Ḃ K|A|Ċ) . (2.145)
Equations (2.144) together with (2.145) are equivalent to the well-known relation
∇a ∇b Kc = Rd abc Kd , which follows from the Killing equations. From (2.144) and
(2.145) one finds that the spinor fields LAB and LȦḂ are covariantly constant along
the Killing vector K,
K AȦ ∇AȦ LBC = 0, K AȦ ∇AȦ LḂĊ = 0,

or equivalently [see (2.143)],
£K LBC = 0, £K LḂĊ = 0. (2.146)
Combining (2.118) and (2.144), one can compute AD LBC and ȦḊ LBC ; then,
making use of (2.140) and the Ricci and Bianchi identities, one finds that
K RṠ ∇RṠCABCD − 4LR(ACBCD)R = 0,

K RṠ ∇RṠ R = 0, (2.147)
K RṠ ∇RṠCABĊḊ − 2LR (ACB)RĊḊ − 2LṘ(ĊC|AB|Ḋ)Ṙ = 0,
which can also be written as £K CABCD = 0, £K R = 0, and £K CABĊḊ = 0, respectively.

In a similar way one finds that £K CȦḂĊḊ = 0.
It is also possible to define the Lie derivative of spinor fields with respect to a
conformal Killing vector. A vector field K is a conformal Killing vector if ∇a Kb +
∇b Ka = 2χ gab, where χ is a function different from zero (if χ is a nonzero constant,
K is a homothetic Killing vector); hence, the spinor equivalent of a conformal Killing
vector satisfies
∇AȦ KBḂ = LAB εȦḂ + LȦḂ εAB − χεABεȦḂ ,
with LAB and LȦḂ symmetric. Proceeding as in the case of a Killing vector, one
obtains the formula
£K ψC...Ḋ... = −K ∇RṠ ψC...Ḋ... − (L S + 2 χδS )ψC...Ḋ... − · · ·

A...Ḃ... RṠ A...Ḃ... A 1 A S...Ḃ...
− (LḂ Ṡ + 12 χδṠḂ )ψC...

A...Ṡ...
Ḋ...
− · · · + (LSC + 12 χδCS )ψS...
A...Ḃ...
Ḋ...
+ · · · + (LṠ Ḋ + 12 χδḊṠ )ψC...
A...Ḃ...
Ṡ...
+ ··· , (2.148)
which reduces to (2.143) when χ = 0. Now we have, for instance, £K ε AB = −χε AB

and £K εAB = χεAB .
Example 2.5. Total angular momentum.

A flat four-dimensional Riemannian manifold possesses ten linearly indepen-
dent Killing vectors. In fact, from (2.144) and (2.145) we obtain ∇AȦ LBC = 0,
∇AȦ LḂĊ = 0. On the other hand, if xa denote Cartesian coordinates, the vector fields
∂a = ∂ /∂ xa form an orthonormal tetrad with Γabc = 0; therefore, ∇AȦ LBC = 0 and
∇AȦ LḂĊ = 0 reduce to ∂AȦ LBC = 0 and ∂AȦ LḂĊ = 0, which means
√ that the com-
ponents LBC and LḂĊ with respect to the null tetrad ∂AȦ = (1/ 2) σ a AȦ ∂ /∂ xa are
√
constant. Letting xAȦ ≡ (1/ 2) σa AȦ xa , we have
1 ∂
∂AȦ xBḂ = σ a AȦ a σb BḂ xb = −δAB δȦḂ ,
2 ∂x
which means that
∂
∂AȦ = − . (2.149)
∂ xAȦ
Then, from (2.140) we obtain
∂ KBḂ
− = LAB εȦḂ + LȦḂ εAB ,
∂ xAȦ
and the solution is
KBḂ = LBS xS Ḃ + LḂṠ xB Ṡ + bBḂ,
where the bAȦ are constants (which correspond to rigid translations). The Killing
vectors of the form
KAȦ = LAB xB Ȧ + LȦḂ xA Ḃ = (LAB εȦḂ + LȦḂ εAB )xBḂ (2.150)
generate a group of isometries isomorphic to SO(p, q).

In the specific case of the Minkowski space-time (Lorentzian signature), the
Killing vector given by (2.150) with LAB = −iα(A βB) , where αA and βA are con-
stant with α A βA = 1, generates rotations in the plane spanned by the orthogonal
spacelike unit vectors given by
1 i
vAȦ = √ (αA β Ȧ + βAα Ȧ ), wAȦ = − √ (αA β Ȧ − βAα Ȧ ).
2 2
Then v(A Ṙ wB)Ṙ = LAB , and from (1.52) and (1.69) we have
vAȦ wBḂ − vBḂwAȦ = LAB εȦḂ + LȦḂ εAB
and
va wb SabAB = va σa(A Ṙ wb σ|b|B)Ṙ = 2v(AṘ wB)Ṙ = 2LAB .
Therefore, for instance, the Lie derivative of a bispinor, defined as the bispinor
formed by the Lie derivatives of the two-component spinors, is [see (2.143), (2.150),
and (1.216)]
C
£ K ψA ψ A L A ψ C
= −K BḂ ∇BḂ +
£K φ Ȧ φ Ȧ −LȦĊ φ Ċ

ψA 1 a b SabA ψC
C
= −(v w − v w ) xCĊ ∇BḂ
BḂ CĊ CĊ BḂ
+ v w
φ Ȧ 2 −SabȦĊ φ Ċ

ψA 1 ψA
= −(vb wc − vc wb ) xc ∂b + va wb [γa , γb ]
φ Ȧ 4 φ Ȧ

ψ
∂ ∂ A 1 a b ψA
= v w xa b − xb a
a b
+ v w [γa , γb ] .
∂x ∂x φ Ȧ 4 φ Ȧ
Apart from a factor −ih̄, this Lie derivative represents the component of the total
angular momentum along the spatial direction orthogonal to the plane spanned by
va and wa .
The fact that the components of the metric tensor (2.106) do not depend on the
coordinate ϕ implies that ∂ /∂ ϕ is a Killing vector of the Schwarzschild metric.
The nonvanishing components of the vector field K = ∂ /∂ ϕ with respect to the
√ √
null tetrad (2.108) are K 12̇ = ir sin θ / 2 and K 21̇ = −ir sin θ / 2; thus, according
to (2.141) and (2.109) [or using (2.142) and (2.107)], the components of the corre-
sponding field LAB are
i i i rg
L11 = √ sin θ , L12 = cos θ , L22 = − √ 1 − sin θ .
2 2 2 2 r
Then, from (2.143) and (2.109) one obtains the remarkably simple expression
∂ A
£ ∂ /∂ ϕ ψ A = ψ .
∂ϕ
The Schwarzschild metric possesses the two additional spacelike Killing vectors
∂ ∂ ∂ ∂
K(1) ≡ − sin ϕ − cot θ cos ϕ , K(2) ≡ cos ϕ − cot θ sin ϕ ,
∂θ ∂ϕ ∂θ ∂ϕ
which, together with K(3) ≡ ∂ /∂ ϕ , generate a group of isometries isomorphic to
SO(3) (one can verify that [K(i) , K( j) ] = −εi jk K(k) ). Following the steps outlined
above, one finds that
∂ ∂ cos ϕ A
£K(1) ψ A = − sin ϕ − cot θ cos ϕ − is ψ ,
∂θ ∂ϕ sin θ
∂ ∂ sin ϕ A
£K(2) ψ A = cos ϕ − cot θ sin ϕ − is ψ ,
∂θ ∂ϕ sin θ
where s is the spin weight of ψ A (s = −1/2 for ψ 1 , s = 1/2 for ψ 2 ) (see also Torres
del Castillo 2003, Chap. 3). It should be noted that all these expressions do not
involve the parameter rg , and therefore, they are also valid when rg = 0, in which
case the Schwarzschild metric (2.106) reduces to the metric of Minkowski space-
time in spherical coordinates.
Exercises
2.1. Show that dθ̆ AȦ = IΓA B ∧ θ̆ BȦ + IΓȦ Ḃ ∧ θ̆ AḂ .
2.2. Show that θ AȦ ∧ θ̆ BḂ = ε AB ε ȦḂ (θ 11̇ ∧ θ 12̇ ∧ θ 21̇ ∧ θ 22̇ ).
2.3. Show that
SAB ∧ SCD = 12 (ε AC ε BD + ε AD ε BC ) θ 11̇ ∧ θ 12̇ ∧ θ 21̇ ∧ θ 22̇ ,

SAB ∧ SĊḊ = 0,
SȦḂ ∧ SĊḊ = − 12 (ε ȦĊ ε ḂḊ + ε ȦḊ ε ḂĊ ) θ 11̇ ∧ θ 12̇ ∧ θ 21̇ ∧ θ 22̇ .
(The second of these relations amounts to the fact that with respect to the inner
product induced on the two-forms, the self-dual two-forms are orthogonal to the
anti-self-dual two-forms [cf. (1.71)].)
2.4. Show that
1
θ̆ AḂ = √ σ̌abc AḂ θ a ∧ θ b ∧ θ c ,
6 2
which, together with (1.82), means that the three-form θ̆ AȦ is proportional to the
dual of the one-form θ AȦ .
2.5. Show that
d(−KAḂ θ AḂ ) = (∇AĊ KBĊ )SAB + (∇C Ȧ KCḂ )SȦḂ ,
d(φAB SAB ) = −(∇C Ḃ φCA )θ̆ AḂ ,
d(φȦḂ SȦḂ ) = (∇AĊ φĊḂ )θ̆ AḂ .
2.6. Find the curvature of the metric (2.113) in the region r < rg , where it has ultra-
hyperbolic signature. Find the algebraic type of the Weyl spinors.
2.7. Show that if K is a Killing vector, then
∇AȦ £K f = £K ∇AȦ f ,
∇AȦ £K ψB = £K ∇AȦ ψB ,
∇AȦ £K ψḂ = £K ∇AȦ ψḂ ,
where f is a differentiable function, and ψB and ψḂ are one-index spinors.

2.8. Show that if K1 and K2 are two Killing vectors, then [£K1 , £K2 ]ψA = £[K1 ,K2 ] ψA
and [£K1 , £K2 ]ψȦ = £[K1 ,K2 ] ψȦ .
2.9. A two-index Killing tensor is a symmetric tensor field Kab such that ∇(a Kbc) = 0
[cf. (2.139)]. Show that this equation is equivalent to
∇(A (Ḋ PBC) Ė Ḟ) = 0
and
∇CḊ PACḂḊ = 34 ∂AḂ K,
where PABĊḊ = P(AB)(ĊḊ) is the spinor equivalent of the traceless part of Kab and
K = Kaa.
2.10. The three-index tensor field Habc is a Lanczos potential if
Habc = −Hbac , Hab b = 0, Habc + Hbca + Hcab = 0,

and the conformal curvature tensor can be expressed in the form
Cabcd = ∇d Habc − ∇c Habd + ∇b Hcda − ∇a Hcdb

+ H(bc) gad − H(ac) gbd + H(ad)gbc − H(bd)gac ,
where Hab ≡ ∇c Hacb (Lanczos 1962). The spinor equivalent of Habc is of the form
HAȦBḂCĊ = hABCĊ εȦḂ + hȦḂĊC εAB , with hABCĊ = h(ABC)Ċ and hȦḂĊC = h(ȦḂĊ)C (see
Exercise 1.3). Show that the Weyl spinors are given by
CABCD = −2∇(A Ṙ hBCD)Ṙ , CȦḂĊḊ = −2∇R (Ȧ hḂĊḊ)R .
(Cf. also Maher and Zund 1968, Zund 1975, Bampi and Caviglia 1983, Ares de
Parga et al. 1989, O’Donnell 2003.)
Chapter 3
Applications to General Relativity
In this chapter some applications of the spinor formalism to special and general
relativity are considered. Some of the examples given here are related to tensor fields
(such as the electromagnetic and the gravitational fields), which can be studied using
the traditional tensor formalism; in these cases one can compare and appreciate the
advantages of the use of the spinor formalism. Other examples involve spinor fields
in an essential manner (such as those of the Dirac field and the Weyl neutrino field)
and also serve to show the usefulness of the two-component spinor formalism.
Even though one can always construct the spinor equivalent of any vector or
tensor, in some cases it is more convenient to use the original objects than their
spinor equivalents. (For instance, the four-momentum of a particle with zero rest
mass, being a null vector, can be represented by a single one-index spinor, but the
four-momentum of a particle with nonzero rest mass, which is a timelike vector,
does not have a particularly simple spinor equivalent.) The examples considered
here have been selected in such a way that the advantages of the spinor formalism
can be more clearly exhibited. These examples include the Maxwell equations, the
self-dual electromagnetic fields, the energy–momentum tensor, and the principal
null directions of the electromagnetic field. The Dirac equation is expressed in terms
of two-component spinors, as well as the zero-rest-mass field equations for any spin.
This chapter contains proofs of the Mariot–Robinson and the Goldberg–Sachs
theorems. In these theorems the shear-free congruences of null geodesics, which
are represented by one-index spinor fields, play a central role. These congruences
also appear in most of the examples considered in the rest of the book, such as the
Killing spinors. The Ernst equations for solutions of the Einstein equations or the
Einstein–Maxwell equations that admit a Killing vector are also derived by means
of the spinor formalism.
Throughout this chapter it is assumed that M is a possibly curved space-time
manifold with Lorentzian signature and that the Infeld–van der Waerden symbols
satisfy the condition σaAḂ = σaBȦ .

Physics 59, DOI 10.1007/978-0-8176-4984-5_3,
104 3 Applications to General Relativity
3.1 Maxwell’s Equations
In the framework of special or general relativity, the electromagnetic field is rep-

resented by an antisymmetric two-index tensor field Fab (or differential two-form)
on the space-time manifold. It is assumed that the electromagnetic field in vacuum
satisfies the Maxwell equations
4π a
∇b F ab = J , ∇b ∗ F ab = 0, (3.1)
c
where the Fab are the components of the electromagnetic field tensor with respect to
an orthonormal tetrad ∂a , J a are the components of the four-current density, and c is
the velocity of light in vacuum (see, e.g., Wald 1984). (The effect of the electromag-
netic field on a charged particle is given by dU a /dτ = (q/m0 c)F a bU b , where τ , q,
m0 , and U a are the proper time, electric charge, rest mass, and four-velocity of the
particle, respectively; if the signature of the metric is chosen as (− − − +), instead
of the first equation in (3.1) we have ∇b F ba = (4π /c)J a.)
According to (1.50), the spinor equivalent of Fab is of the form
FAȦBḂ = fAB εȦḂ + fȦḂ εAB , (3.2)
with fAB = 12 FA Ṙ BṘ and fȦḂ = 12 F R ȦRḂ being symmetric spinor fields, and since Fab
is real, FAḂCḊ = FBȦDĊ [see (1.47)]. Hence
fȦḂ = fAB . (3.3)
Thus, the spinor equivalent of the Maxwell equations (3.1) is given by

4π AȦ
−∇BḂ ( f AB ε ȦḂ + f ȦḂ ε AB ) = J
c
and
i∇BḂ ( f AB ε ȦḂ − f ȦḂ ε AB ) = 0
[see (1.65)]. By combining these equations one obtains

2π AȦ 2π AȦ
∇B Ȧ f AB = J , ∇A Ḃ f ȦḂ = J . (3.4)
c c
Owing to (3.3) and to the fact that J AḂ is the spinor equivalent of a real vector field,
the second equation is the complex conjugate of the first one.
Making use of (1.72) and the second line of (1.77) we find an explicit relation be-
tween the components of the electromagnetic spinor fAB and the tetrad components
Fab :
( f A B ) = 14 (F ab Sab A B )

= 12 (−F 14 + iF 23 ) σ1 + (F 24 − iF 31 ) σ2 + (−F 34 + iF 12 ) σ3 ;
3.1 Maxwell’s Equations 105
hence,

1 −F 14 + iF 23 + i(F 24 − iF 31 ) F 34 − iF 12
( fAB ) =
2 F 34 − iF 12
F 14 − iF 23 + i(F 24 − iF 31 )

1 Ex + iBx − i(Ey + iBy ) −(Ez + iBz )
= , (3.5)
2 −(Ez + iBz ) −(Ex + iBx ) − i(Ey + iBy )
where (Ex , Ey , Ez ) and (Bx , By , Bz ) are the Cartesian components of the electric and
magnetic field, respectively, with respect to the orthonormal tetrad ∂a .
The Maxwell equations ∇b ∗ F ab = 0, which can also be written as ∇a Fbc +
∇b Fca + ∇c Fab = 0, are locally equivalent to the existence of a vector field Aa (the
four-potential) such that
Fab = ∇a Ab − ∇b Aa . (3.6)
Thus, if ABḂ is the spinor equivalent of Ab , then
fBC = ∇(B Ḃ AC)Ḃ , fḂĊ = ∇B (Ḃ A|B|Ċ) . (3.7)
For a given electromagnetic field, the four-potential Aa is defined by (3.6) up to a

gauge transformation
Aa → Aa + ∂a ξ , (3.8)
where ξ is an arbitrary differentiable function.
In a source-free region, ∇b F ab = 0; therefore, by analogy with (3.6), there exists
locally a second (real) vector field Ǎa such that
∗
Fab = ∇a Ǎb − ∇b Ǎa ,
or equivalently [see (1.65)],
fBC = i∇(B Ḃ ǍC)Ḃ , fḂĊ = −i∇B (Ḃ Ǎ|B|Ċ) , (3.9)
where ǍBḂ is the spinor equivalent of Ǎb . Then, the complex vector field
Φa ≡ 12 (Aa − iǍa )
is a potential for the self-dual field 12 (Fab − i∗ Fab ) [i.e., ∗ (Fab − i∗ Fab ) = i(Fab −
i∗ Fab )]; in fact, from (3.7) and (3.9) we obtain
∇(B Ḃ ΦC)Ḃ = 12 (∇(B Ḃ AC)Ḃ − i∇(B Ḃ ǍC)Ḃ ) = 0,
that is,
∇(B Ḃ ΦC)Ḃ = 0. (3.10)
Thus, we have proved the following result.
Proposition 3.1. If Fab is a (real) solution of the source-free Maxwell equations,

then there exists locally a complex vector field Φa satisfying the self-duality condi-
tion (3.10) such that Fab = ∇a Ab − ∇b Aa , where Aa = 2 Re Φa .
The self-duality condition (3.10) has the advantage of being a set of first-order
partial differential equations for Φa , in contrast to the original source-free Maxwell
equations ∇b F ab = 0, which, expressed in terms of the four-potential, are second-
order equations. Making use of (3.7) and the complex conjugate of (3.10), one
finds that the self-dual part of the real electromagnetic field generated by the four-
potential Aa = 2 Re Φa is given by
fȦḂ = ∇B (Ȧ Φ|B|Ḃ) . (3.11)
In terms of the Newman–Penrose notation, the self-duality condition (3.10) is

equivalent to the explicit expressions
(D − ε + ε − ρ )Φ12̇ − (δ − β − α + π )Φ11̇ + κΦ22̇ − σ Φ21̇ = 0,

(D + ε + ε + ρ − ρ )Φ22̇ − (δ + β − α + τ + π )Φ21̇
+ (δ − α + β − π − τ )Φ12̇ − (Δ − γ − γ − μ + μ )Φ11̇ = 0, (3.12)
(δ + α + β − τ )Φ22̇ − (Δ + γ − γ + μ )Φ21̇ + νΦ11̇ − λ Φ12̇ = 0.
As an application of the last proposition, we can find the solution of the source-
free Maxwell equations assuming that the metric of the space-time is the Schwarz-
schild metric. Making use of (2.108) and (2.109), after some rearrangement, one
finds that equations (3.12) take the form

1 rg −1 1 ∂ ∂ 1 ∂ i ∂
− 1− + rΦ12̇ − √ + Φ11̇ = 0,
r r c ∂t ∂r 2 r ∂ θ sin θ ∂ ϕ

rg −1 1 ∂ ∂ 1 ∂ i ∂
− 1− + Φ22̇ − √ + + cot θ Φ21̇
r c ∂t ∂r 2 r ∂ θ sin θ ∂ ϕ

1 ∂ i ∂
+√ − + cot θ Φ12̇
2 r ∂ θ sin θ ∂ ϕ

1 rg −1 1 ∂ ∂ rg
+ 1− + 1− Φ11̇ = 0,
2 r c ∂t ∂r r

1 ∂ i ∂ 1 1∂ rg ∂
√ − Φ22̇ + + 1− rΦ21̇ = 0.
2 r ∂ θ sin θ ∂ ϕ 2r c ∂ t r ∂r
These equations admit separable solutions of the form

rg −1
Φ11̇ = 2 1 − F(r)Y jm (θ , ϕ ) e−iω t ,
r
√
Φ12̇ = 2 G(r) 1Y jm (θ , ϕ ) e−iω t ,
(3.13)
√
Φ21̇ = 2 f (r) −1Y jm (θ , ϕ ) e−iω t ,
Φ22̇ = g(r)Y jm (θ , ϕ ) e−iω t ,
where the sY jm are spin-weighted spherical harmonics (Newman and Penrose 1966,
Torres del Castillo 2003), j and m are integers, j = 1, 2, . . ., − j m j, and ω
is a constant. The factor 2(1 − rg/r)−1 is introduced for convenience and corrects
the asymmetry introduced in the definition of the null tetrad (2.108). Following the
conventions of Newman and Penrose (1966), we have

∂ i ∂
+ − s cot θ sY jm = − j( j + 1) − s(s + 1) s+1Y jm ,
∂ θ sin θ ∂ ϕ

(3.14)
∂ i ∂
− + s cot θ sY jm = j( j + 1) − s(s − 1) s−1Y jm ,
∂ θ sin θ ∂ ϕ
and the 0Y jm are the ordinary spherical harmonics Y jm . Thus, the one-variable func-
tions F, G, f , and g must obey the system of ordinary differential equations

iω rg d
+ 1− rG + j( j + 1) F = 0,
c r dr

iω rg d j( j + 1) rg
+ 1− g+ 1− ( f + G)
c r dr r r

iω rg d
+ − + 1− F = 0, (3.15)
c r dr

iω rg d
j( j + 1) g + − + 1 − r f = 0.
c r dr
These equations imply that f + G must satisfy the decoupled equation

2
rg f +G ω rg d rg d
1− j( j + 1) = 2 + 1− 1− r( f + G), (3.16)
r r c r dr r dr
leaving f − G unspecified, which is related to the gauge freedom (3.8) [see equation
(3.17) below].
Then, letting
r( f − G) f +G
ξ≡ Y jm e−iω t , χ≡ Y jm e−iω t ,
j( j + 1) j( j + 1)
from (2.108) and (3.13)–(3.16) one obtains
Φ11̇ = ∂11̇ ξ + ∂11̇ (r χ ), Φ22̇ = ∂22̇ ξ − ∂22̇ (r χ ),

(3.17)
Φ12̇ = ∂12̇ ξ + ∂12̇ (r χ ), Φ21̇ = ∂21̇ ξ − ∂21̇ (r χ ),
where ξ is an arbitrary function [cf. (3.8)] and χ is a solution of the second-order

linear partial differential equation

rg ∂ rg ∂ 1 ∂ ∂ 1 ∂2 1 ∂2
1− 1− + sin θ + − 2 2 r χ = 0.
r ∂r r ∂ r sin θ ∂ θ ∂ θ sin2 θ ∂ ϕ 2 c ∂t
(3.18)
(When rg = 0, this equation reduces to the usual scalar wave equation for χ .) Note
that (3.17) and (3.18) do not contain the separation constants j, m, and ω . By choos-
ing the function ξ equal to r χ or to −r χ , one makes Φ2Ȧ = 0 or Φ1Ȧ = 0, respec-
tively (see (3.76) below).
A similar result holds in other space-times; the solution of the self-duality condi-
tion (3.10) can be expressed locally in terms of a single complex function if there ex-
ists a repeated principal spinor of the conformal curvature lA such that l A l B ∇AȦ lB = 0
(see Proposition 3.6 below).
When the Ricci tensor vanishes, each Killing vector K is the four-potential of a
(possibly trivial) solution of the source-free Maxwell equations. From (2.144) and
(2.85) we have ∇A Ȧ LBC = CACḊȦ K AḊ − 18 R KCȦ = 12 RCȦAḊ K AḊ , which is equal to
zero if Rab = 0 or under the weaker condition Rab K b = 0. In any of these cases, fAB =
LAB satisfies the source-free Maxwell equations (3.4). Furthermore, by virtue of
(2.146), the Lie derivative with respect to K of this electromagnetic field is equal to
zero. In the case of Minkowski space-time one obtains only uniform electromagnetic
fields in this manner (see Example 2.5). However, the timelike Killing vector ∂ /∂ t
of the Schwarzschild metric gives rise to a nontrivial spherically symmetric field.
Energy–Momentum Tensor of the Electromagnetic Field
In special relativity, the four-vector 1c F a b J b is the Lorentz force density on the

charge and current distribution represented by J a , in the presence of the electro-
magnetic field Fab (see, e.g., Jackson 1975, Sec. 12.10). The spinor equivalent of
1 a b
c F b J , expressed in terms of the electromagnetic field alone, is
1 1
− F AȦ BḂ J BḂ = − ( f A B ε Ȧ Ḃ + f Ȧ Ḃ ε A B )J BḂ
c c
1 A BȦ
= − ( f B J + f Ȧ Ḃ J AḂ )
c
1
= − ( f A B ∇BĊ f ȦĊ + f Ȧ Ḃ ∇C Ḃ f AC )
2π
1
= ∇ ( f AB f ȦḂ ),
2π BḂ
where we have made use of (3.2) and (3.4). The spinor field
1
TABȦḂ ≡
f f (3.19)
2π AB ȦḂ
thus obtained is the spinor equivalent of the energy–momentum tensor of the elec-
tromagnetic field Tab . As a consequence of the symmetries TABȦḂ = T(AB)(ȦḂ) , Tab is
symmetric and traceless.
In order to find the expression for Tab in terms of the electromagnetic field tensor
Fab , we rewrite (3.19) in the form
4π TABȦḂ = −( fAC εȦĊ + fȦĊ εAC )( fBC εḂĊ + fḂĊ εBC )

− fAC fBC εȦḂ − fȦĊ fḂĊ εAB .
Owing to (1.23), fAC fBC = − fAC fBC = − fBC fAC ; hence, by virtue of (1.26), fAC fBC
= − 12 fRC f RC εAB , and therefore, by complex conjugation, fȦĊ fḂĊ = − 12 fṘĊ f ṘĊ εȦḂ .
Thus
4π TABȦḂ = −( fAC εȦĊ + fȦĊ εAC )( fBC εḂĊ + fḂĊ εBC )
+ 12 ( fRC f RC + fṘĊ f ṘĊ ) εAB εȦḂ
= −( fAC εȦĊ + fȦĊ εAC )( fBC εḂĊ + fḂĊ εBC )
+ 14 ( fRC εṘĊ + fṘĊ εRC )( f RC ε ṘĊ + f ṘĊ ε RC ) εAB εȦḂ ,
and this amounts to
4π Tab = Fac Fb c − 14 Fcd F cd gab . (3.20)
Equivalently, we can start from
4π TABȦḂ = ( fAC εȦĊ − fȦĊ εAC )( fBC εḂĊ − fḂĊ εBC )

+ fAC fBC εȦḂ + fȦĊ fḂĊ εAB ,
which leads to the alternative form [see (1.65)]
4π Tab = ∗ Fac ∗ Fb c + 14 Fcd F cd gab .
By combining the two expressions found for Tab we also have
4π Tab = 12 (Fac Fb c + ∗ Fac ∗ Fb c ). (3.21)
In addition to the algebraic properties Tab = T(ab) and Ta a = 0, the energy–

momentum tensor of the electromagnetic field satisfies the so-called dominant en-
ergy condition: for every timelike vector t a , Tabt at b 0 and Tabt b is a nonspacelike
vector, i.e., (Tabt b )(T a ct c ) 0. Physically, this means that for any observer, the
energy density of the electromagnetic field is always positive and the energy flux
vector is nonspacelike. We can readily prove the validity of this condition by mak-
ing use of the fact that fAB can be expressed as the symmetrized tensor product of
its principal spinors (see Section 1.7).
If the principal spinors αA , βA of fAB are not proportional to each other, the
electromagnetic field is called algebraically general; then, writing fAB = α(A βB) ,
we have
Tabt at b = TABȦḂt AȦt BḂ

1
= α β α β t AȦt BḂ
2π (A B) (Ȧ Ḃ)
1 AȦ
= (t αA α Ȧt BḂ βB β Ḃ + t AȦβA α Ȧt BḂαB β Ḃ ). (3.22)
4π
On the other hand,
t ata = −εAB εȦḂt AȦt BḂ

1
=− (αA βB − αB βA )(α Ȧ β Ḃ − α Ḃ β Ȧ )t AȦt BḂ
α R βR α Ṡ β Ṡ
2
= − R 2 (t AȦ αA α Ȧt BḂ βB β Ḃ − t AȦβA α Ȧt BḂ αB β Ḃ );
|α βR |
hence, t ata < 0 if and only if t AȦ αA α Ȧ t BḂ βB β Ḃ > |t AȦ αA β Ȧ |2 , and from (3.22) we
then obtain Tabt at b > 21π |t AȦ αA β Ȧ |2 0.
When the two principal spinors of the electromagnetic spinor fAB are propor-
tional to each other, the electromagnetic field is called algebraically special or null,
and by absorbing the proportionality factor into αA , we can write fAB = αA αB ; then
Tabt at b = 21π (t AȦ αA α Ȧ )2 0. As a matter of fact, for tat a < 0, t AȦ αA α Ȧ vanishes
only if αA = 0 (see Exercise 3.1).
In order to prove that for every timelike vector t a , Tabt b is a nonspacelike vector,
we first note that the spinor equivalent of Tab T a c is given by
1 1
− f f f A f Ȧ = − f f AR εBC fȦṠ f ȦṠ εḂĊ
(2π )2 AB ȦḂ C Ċ 16π 2 AR
1 2
=− fAR f AR εBC εḂĊ ,
16π 2
which means that Tab T a c is proportional to gbc . In terms of the electromagnetic field
tensor and its dual, f AB fAB can be expressed as [see (1.65)]
f AB fAB = 2f ε fAB εȦḂ

1 AB ȦḂ
= 2(f ε + f ȦḂ ε AB ) fAB εȦḂ

1 AB ȦḂ
∗
= 2 (FABȦḂ + i FABȦḂ )
1 ABȦḂ 1
2F
∗ ab
= 4 (F Fab + i F Fab );
1 ab
(3.23)
hence,
1 rs
Tab T a c = (F Frs )2 + (∗ F rs Frs )2 gbc , (3.24)
(16π )2
and from this last expression we readily see that (Tabt b )(T a c t c ) 0 for every time-
like vector t a . As a matter of fact, Tabt a T a ct c = 0 for any timelike vector t a , that is,
the energy flux vector is null for any observer if and only if f AB fAB = 0.
The proportionality of Tab T a c to gbc is a characteristic of the energy–momentum
tensor of the electromagnetic field. More precisely, if Tab = T(ab) , T a a = 0, and
Tab T a c is proportional to gbc , then there exists Fab = −Fba such that 4π Tab =
± (Fac Fb c − 14 Fcd F cd gab ). In order to prove this assertion we note that for an ob-
ject μABCD such that μABCD = −μCDAB , the identity
μABCD = 12 (μABCD − μCDAB)

= 12 (μABCD − μCBAD + μCBAD − μCDAB )
= 12 (εAC μ R BRD + εBD μC R AR )
= 12 εAC (μ R (B|R|D) + 12 εBD μ RS RS )
+ 12 εBD (μ(C R A)R + 12 εCA μ SR SR )
= 12 (εAC μ R (B|R|D) + εBD μ(C R A)R ) (3.25)
holds. Since we are assuming that Tab is symmetric and traceless, it follows that
TABȦḂ = T(AB)(ȦḂ) , and from (3.25) we have
T AB ȦḂ T CDĊḊ − T CD ȦḂ T ABĊḊ = ε AC T R(B ȦḂ TR D)ĊḊ + ε BD T R(C ȦḂ TR A)ĊḊ .
Now T R(B ȦḂ TR D)ĊḊ = −T R(BĊḊ TR D) ȦḂ , and applying the analogue of (3.25) for
objects with dotted indices, we find that
T AB ȦḂ T CDĊḊ − T CD ȦḂ T ABĊḊ

= 12 ε AC (εȦĊ T R(B|Ṙ| (Ḃ T|R D) Ṙ|Ḋ) + εḂḊ T R(B|Ṙ| (Ȧ T|R D) Ṙ|Ċ) )
+ 12 ε BD (εȦĊ T R(A|Ṙ| (Ḃ T|RC) Ṙ|Ḋ) + εḂḊ T R(A|Ṙ| (Ȧ T|RC) Ṙ|Ċ) ),
which is equal to zero, since by hypothesis, TABȦḂ T AC ȦĊ is proportional to εBC εḂĊ .
Hence,
TABȦḂ TCDĊḊ = TCDȦḂ TABĊḊ , (3.26)
and contracting both sides of this last equation with α C α Ċ α D α Ḋ , assuming that
TCDĊḊ α C α Ċ α D α Ḋ is different from zero, we obtain
TABĊḊ α Ċ α Ḋ TCDȦḂ α C α D
TABȦḂ = ,
TCDĊḊ α C α Ċ α D α Ḋ
thus showing that TABȦḂ is of the form
1
TABȦḂ = ± f f . (3.27)
2π AB ȦḂ
The spinor fAB appearing in this last expression is defined by TABȦḂ up to a duality
rotation fAB → eiχ fAB , where χ is an arbitrary real-valued function. According to the
preceding discussion, the sign on the right-hand side must be positive if Tabt at b 0
for every timelike vector t a .
Moreover, it may be noticed that by contracting both sides of (3.26) with
φ CD φ ĊḊ , where φ AB is a symmetric spinor, we obtain
TABĊḊ φ ĊḊ TCDȦḂ φ CD

TABȦḂ = ,
TCDĊḊ φ CD φ ĊḊ
which is again of the form (3.27), with fAB proportional to TABĊḊ φ ĊḊ and fȦḂ
proportional to TCDȦḂ φ CD , provided that TCDĊḊ φ CD φ ĊḊ is different from zero.
Then, using the fact that TABṘṠ φ ṘṠ εȦḂ + TRSȦḂ φ RS εAB is the spinor equivalent of
Tac φ c b − Tbc φ c a , where φab is the tensor equivalent of φAB εȦḂ + φ ȦḂ εAB , one con-
cludes that up to duality rotations, the electromagnetic field tensor corresponding to
a given energy–momentum tensor Tab is given by
√ Tac φ c b − Tbc φ c a
Fab = 4π , (3.28)
Tcd φ c s φ ds
where φab is any (real) antisymmetric tensor field such that Tcd φ c s φ ds is different
from zero.
Principal Null Directions of the Electromagnetic Field
If the electromagnetic field is different from zero, the electromagnetic spinor has
one of the forms
α(A βB) with α A βA = 0,
fAB =
αA αB .
In the first case (where the electromagnetic field is algebraically general) f AB fAB =
− 12 (α A βA )2 = 0, and in the second case (where the electromagnetic field is alge-
braically special) f AB fAB = 0. Since the two well-known invariants of the electro-
magnetic field F ab Fab and ∗ F ab Fab are both real, from (3.23) we see that the elec-
tromagnetic field is algebraically special if and only if F ab Fab and ∗ F ab Fab are both
equal to zero. Furthermore, when the electromagnetic field is algebraically special,
from (3.19) we see that
1
Tab = la lb ,
2π
where la is the vector equivalent of αA α Ȧ , which is a real null vector. Conversely, if
Tab is proportional to la lb , where la is a real vector, la must be null (since T a a = 0),
and from (3.19) or (3.24) it follows that the electromagnetic field is algebraically
special.
Each principal spinor αA of an algebraically general electromagnetic field fAB

√
gives rise to a (real) null vector l a = −(1/ 2) σ a AȦ α A α Ȧ . Since the principal
spinors of fAB are defined up to a complex factor, l a is defined up to a (positive)
real factor. Hence, at each point of the space-time, an algebraically general electro-
magnetic field defines two real null directions, called principal null directions, of
the electromagnetic field. When the electromagnetic field is algebraically special,
these two principal null directions coincide. (Note that if fAB = αA αB , then αA is
defined up to sign.)
The principal null directions of the electromagnetic field can be characterized as
those null directions l a along which Tab l a l b attains its minimum value (zero). Indeed,
the spinor equivalent of a real null vector l a is of the form l AȦ = ±γ A γ Ȧ . Then
Tab l a l b = 21π |γ A αA |2 |γ B βB |2 , which is equal to zero if and only if γA is proportional
to αA or to βA .
The principal null directions of the electromagnetic field are the only real null
eigenvectors of the electromagnetic field tensor Fab . If la is a real null eigenvector of
Fab , then its spinor equivalent is of the form ±γA γ Ȧ , and the condition Fab l b = λ la
is equivalent to
( fAB εȦḂ + fȦḂ εAB )γ B γ Ḃ = −λ γA γ Ȧ .
Hence, fAB γ B γ Ȧ + fȦḂ γA γ Ḃ = −λ γA γ Ȧ . Contracting this last equation with γ A , we
obtain fAB γ A γ B = 0, and therefore γA is a principal spinor of fAB [see (1.230)].
Conversely, if γA is a principal spinor of fAB , then fAB γ A = μγB , for some μ , and
( fAB εȦḂ + fȦḂ εAB )γ B γ Ḃ = (μ + μ )γA γ Ȧ , which means that γA γ Ȧ is the spinor equiva-
lent of a real null eigenvector of Fab . The condition Fab l b = λ la is usually expressed
in the equivalent form
l a Fa[b lc] = 0
[cf. (2.131)].
In the case of an algebraically general electromagnetic field, from (3.2) and (1.27)
we obtain
FAȦBḂ = fAB εȦḂ + fȦḂ εAB

1 1
= (αA βB + αB βA ) (α Ȧ β Ḃ − α Ḃ β Ȧ ) + c.c.
2 α β
Ṡ
Ṡ
1
= (αA α Ȧ βB β Ḃ − αB α Ḃ βA β Ȧ + αB α Ȧ βA β Ḃ − αA α Ḃ βB β Ȧ ) + c.c.,
2α Ṡ β Ṡ
where c.c. denotes complex conjugate. If la and na are the vector equivalents of
αA α Ȧ and βA β Ȧ , respectively (that is, la and na point along the two principal null
directions of the electromagnetic field), the spinor equivalent of εabcd l c nd is [see
(1.64)]

i(εAC εBD εȦḊ εḂĊ − εAD εBC εȦĊ εḂḊ ) α C α Ċ β D β Ḋ = i αA α Ḃ βB β Ȧ − αB α Ȧ βA β Ḃ ,
and hence
Fab = (Re F ) (la nb − lb na ) + (ImF ) εabcd l c nd , (3.29)
with F ≡ 1/(α S β
S ) [cf. (1.157)]. The function F is related to the complex invariant
f AB fAB through f AB fAB = − 12 F −2 . Directly from (3.29) we have
∗
Fab = −(Im F ) (la nb − lb na ) + (Re F ) εabcd l c nd .
The energy–momentum tensor can also be expressed in terms of the null vectors
la , na ; the result is
1
Tab = (la nb + nalb − 12 l c nc gab ),
4π
which shows that the subspace spanned by la , na is formed by eigenvectors of T a b ,
with eigenvalue l c nc /8π , and the vectors orthogonal to la and na are eigenvectors of
T a b , with eigenvalue −l c nc /8π .
When the electromagnetic field is algebraically special one can find γA such that
α A γA = 1; then
FAȦBḂ = fAB εȦḂ + fȦḂ εAB

= αA αB (α Ȧ γ Ḃ − α Ḃ γ Ȧ ) + α Ȧ α Ḃ (αA γB − αB γA )
= αA α Ȧ (αB γ Ḃ + γB α Ḃ ) − αB α Ḃ (αA γ Ȧ + γA α Ȧ ).
Thus, if sa is the vector equivalent of αA γ Ȧ + γA α Ȧ , we have
Fab = la sb − lb sa , (3.30)
and l a sa = 0, sa sa = 2. Furthermore,
∗
FAȦBḂ = −iαA αB (α Ȧ γ Ḃ − α Ḃ γ Ȧ ) + iα Ȧ α Ḃ (αA γB − αB γA )
= αA α Ȧ (iγB α Ḃ − iαB γ Ḃ ) − αB α Ḃ (iγA α Ȧ − iαA γ Ȧ ),
that is,
∗
Fab = la rb − lb ra ,
where ra is the vector equivalent of i(γA α Ȧ − αA γ Ȧ ), which is also orthogonal to
la . Hence, we conclude that when the electromagnetic field is algebraically special,
Fab and ∗ Fab are simple and Fab l b = 0 = ∗ Fab l b , where la is a null vector that points
along the repeated principal null direction of the electromagnetic field.
For example, in Minkowski space-time, the electric and magnetic fields with
Cartesian components

E = A cos(kz − ω t), ± sin(kz − ω t), 0 ,

B = A ∓ sin(kz − ω t), cos(kz − ω t), 0 ,
where A is a positive real constant, correspond to a monochromatic, circularly po-

larized, electromagnetic plane wave propagating along the z-axis, with angular fre-
3.2 Dirac’s Equation 115
quency ω and wave number k = ω /c (the upper or the lower sign determines the
sense of rotation of the fields). According to (3.5), the components of the electro-
magnetic spinor are simply
f11 = Ae±i(kz−ω t) , f12 = 0 = f22 . (3.31)
Therefore f AB fAB = 0, which means that this field is algebraically special. By in-
spection, one finds that fAB = αA αB with

α1 1/2 ±i(kz−ω t)/2 1
= ±A e .
α2 0
The only principal null direction of this field is proportional to (0, 0, 1, 1).
A simple example of an algebraically general electromagnetic field is provided
by the field of a static electric point charge E = q(x, y, z)/r3 , B = 0, with r = (x2 +
y2 + z2 )1/2 . Making use of (3.5) one obtains the spinor components

q x − iy −z q e−iϕ sin θ − cos θ
( fAB ) = 3 = 2 ,
2r −z −x − iy 2r − cos θ −eiϕ sin θ
where we have replaced the Cartesian coordinates by spherical coordinates. The

ratios α 1 /α 2 and β 1 /β 2 of the components of the principal spinors of fAB are the
roots of the polynomial [see (1.229)]
q −iϕ
f11 z2 + 2 f12z + f22 = (e sin θ z2 − 2 cos θ z − eiϕ sin θ ),
2r2
which are −eiϕ tan 12 θ and eiϕ cot 12 θ . Hence, the principal spinors of fAB can be
chosen as

α1 iϕ /2 sin 1 θ β1 iϕ /2 cos 1 θ
q −e 2 1 e 2
= , = ,
α2 r e−iϕ /2 cos 1 θ β2 r e−iϕ /2 sin 1 θ
2 2
and F −1 = α S βS = q/r2 . The first of these spinors corresponds to a null vector

whose spatial part points radially away from the origin [see (1.163)], and the second
spinor corresponds to a null vector with a spatial part pointing radially toward the
origin.
3.2 Dirac’s Equation
Another illustrative example of the applications of the spinor formalism is given

by the Dirac equation. As a matter of fact, in the study of this equation, Infeld and
van der Waerden (1933) developed the two-component spinor formalism for curved
space-times.
The Dirac equation in flat space-time, expressed in terms of Cartesian coordi-

nates (x1 , x2 , x3 , x4 ) = (x, y, z, ct), has the form
∂Ψ m0 c
γa = Ψ, (3.32)
∂ xa h̄
where m0 is the rest mass of the particle and the γ a must satisfy the relations (1.194),
that is,
γ a γ b + γ b γ a = 2g ab
(see, e.g., Messiah 1962, Davydov 1988, Merzbacher 1998). As shown in Section
1.5, we can find a linear representation of the γ a in a complex vector space of di-
mension four and with respect to a suitable basis [see (1.211)]

ψA σaAḂ φ Ḃ
γa = ,
φ Ȧ −σa BȦ ψB
where the σaAḂ are Infeld–van der Waerden symbols. Hence, the Dirac equation
(3.32) is equivalent to
m0 c m0 c
∂AḂ φ Ḃ = √ ψA , ∂BȦ ψ B = √ φȦ , (3.33)
2 h̄ 2 h̄
with
1 ∂
∂AḂ = √ σ a AḂ a .
2 ∂x
Since the spin coefficients for this null tetrad are all equal to zero, equations (3.33)
are equivalent to
m0 c m0 c
∇AḂ φ Ḃ = √ ψA , ∇BȦ ψ B = √ φȦ . (3.34)
2 h̄ 2 h̄
This set of equations is form-invariant under null tetrad transformations, and it is
assumed that it also holds in a curved space-time.
The combination of the Dirac equations (3.34) and their complex conjugates
yields the continuity equation
m0 c
∇AȦ (φ A φ Ȧ + ψ A ψ Ȧ ) = √ (ψ Ȧ φ Ȧ + φ A ψA + ψ A φ A + ψ Ȧ φȦ ) = 0.
2 h̄
The fact that φ A φ Ȧ and ψ A ψ Ȧ are the spinor equivalents of two future-pointing
null vectors implies that J AȦ ≡ φ A φ Ȧ + ψ A ψ Ȧ is the spinor equivalent of a future-
pointing timelike current density J a ; hence, J 4 is nonnegative and can be interpreted
as a probability density.
Making use of the definitions (2.25), (2.26), and (2.28), the Dirac equation (3.34)
takes the form
m0 c
(δ − α + π )φ1̇ − (D + ε − ρ )φ2̇ = √ ψ1 ,
2 h̄
m0 c
(Δ − γ + μ )φ1̇ − (δ + β − τ )φ2̇ = √ ψ2 ,
2 h̄
(3.35)
m0 c
(δ − α + π )ψ1 − (D + ε − ρ )ψ2 = √ φ1̇ ,
2 h̄
m0 c
(Δ − γ + μ )ψ1 − (δ + β − τ )ψ2 = √ φ2̇ .
2 h̄
As in the case of the Maxwell equations, the Dirac equation is integrable in flat
space-time or in a curved space-time without restriction on the curvature.
Example 3.1. Solution of the Dirac equation in the Schwarzschild background.

The Dirac equation can be solved by taking as the background metric that of the
Schwarzschild space-time, assuming that the energy–momentum tensor of the Dirac
field is small enough and that the space-time metric is not affected by the presence
of the Dirac field. Substituting (2.108) and (2.109) into (3.35), we obtain

m0 c 1 ∂ i ∂ cot θ
√ ψ1 = √ + + φ1̇
2 h̄ 2 r ∂ θ sin θ ∂ ϕ 2

1 rg −1 1 ∂ ∂
− − 1− + rφ2̇ ,
r r c ∂t ∂r

m c 1 1 ∂ rg ∂ rg
√ 0 ψ2 = − + 1− + 2 rφ1̇
2 h̄ 2r c ∂ t r ∂ r 2r

1 ∂ i ∂ cot θ
−√ − + φ2̇ ,
2 r ∂ θ sin θ ∂ ϕ 2

m c 1 ∂ i ∂ cot θ
√ 0 φ1̇ = √ − + ψ1
2 h̄ 2r ∂ θ sin θ ∂ ϕ 2

1 rg −1 1 ∂ ∂
− − 1− + r ψ2 ,
r r c ∂t ∂r

m c 1 1 ∂ rg ∂ rg
√ 0 φ2̇ = − + 1− + 2 r ψ1
2 h̄ 2r c ∂ t r ∂ r 2r

1 ∂ i ∂ cot θ
−√ + + ψ2 .
2 r ∂ θ sin θ ∂ ϕ 2
(When rg = 0, the background space-time is flat.) These equations admit separable

solutions of the form
√ rg −1/2
ψ1 = 2 1− F(r) 1 Y jm (θ , ϕ ) e−iEt/h̄ ,
r 2
ψ2 = G(r) − 1 Y jm (θ , ϕ ) e −iEt/h̄ ,
2
(3.36)
√ rg −1/2
φ1̇ = 2 1 − f (r) − 1 Y jm (θ , ϕ ) e−iEt/h̄ ,
r 2
φ2̇ = g(r) 1 Y jm (θ , ϕ ) e −iEt/h̄ ,

2
where the sY jm are spin-weighted spherical harmonics (Newman and Penrose 1966,
Torres del Castillo 2003),√j and m are half-integers with − j m j, and E is a
real constant. The factor 2 (1 − rg /r)−1/2 is introduced for later convenience and
corrects the asymmetry in the definition of the null tetrad (2.108) [cf. (3.13)]. (The
spin weight of the functions on the right-hand side of (3.36) is determined by the
type of the corresponding component appearing on the left-hand side as defined in
Section 2.1.) According to (3.14) we have

∂ i ∂ 1
± + cot θ ∓ 1 Y jm = ∓( j + 12 ) ± 1 Y jm , (3.37)
∂ θ sin θ ∂ ϕ 2 2 2
and therefore the one-variable functions F, G, f , and g must obey the system of
ordinary differential equations

m0 c 1 1 rg 1/2 iE rg −1 d
F = − (j + 2)f −
1
1− 1− + rg,
h̄ r r r h̄c r dr

m0 c 1 rg 1/2 iE rg −1 d 1
G = − 1− − 1− + r f − ( j + 12 )g,
h̄ r r h̄c r dr r

m0 c 1 1 rg 1/2 iE rg −1 d
f = ( j + 12 )F − 1− 1− + rG,
h̄ r r r h̄c r dr

m0 c 1 rg 1/2 iE rg −1 d 1
g = − 1− − 1− + rF + ( j + 12 )G.
h̄ r r h̄c r dr r
These equations can be partially decoupled by expressing them in the equivalent
form

m0 c j + 12 i rg 1/2 d iE rg −1
A=i A− 1− + 1− rB,
h̄ r r r dr h̄c r

m0 c j + 12 i rg 1/2 d iE rg −1
B = −i B+ 1− − 1− rA,
h̄ r r r dr h̄c r

m0 c j + 12 i rg 1/2 d iE rg −1
C = −i C+ 1− + 1− rD,
h̄ r r r dr h̄c r

m0 c j + 12 i rg 1/2 d iE rg −1
D=i D− 1− − 1− rC,
h̄ r r r dr h̄c r
with A ≡ F + i f , B ≡ G − ig, C ≡ F − i f , and D ≡ G + ig.

The fact that the pair of functions A, B is decoupled from the pair C, D implies
that one can consider particular solutions of the Dirac equation such that either
A = 0 = B or C = 0 = D; that is, there exist solutions (3.36) such that f = ±iF
and g = ∓iG, which contain only two radial functions instead of four. As we shall
show below (Section 3.4), this possibility is related to the existence of a symmetry
operator for the Dirac equation in any type-D vacuum space-time [see (3.128)]. (The
solution of the Dirac equation in the Kerr background given in Chandrasekhar 1983,
§104, belongs to this class of particular solutions that involve two radial functions
only.) When rg = 0 the radial functions are linear combinations of spherical Bessel
functions of order j ± 12 (recall that j is a half-integer).
The Dirac equation in a possibly curved space-time (3.34) can be written in terms
of the Dirac matrices γa and bispinors. Indeed, from (3.34), (2.39), (1.211), (2.18),
and (1.216), we obtain

m 0 c ψA √ ∂AḂ φ Ḃ − Γ Ḃ ṠḂA φ Ṡ
= 2
h̄ φ Ȧ −∂ BȦ ψB − Γ xS B BȦ ψS

ψA 1 SabḂṠ σ c AḂ φ Ṡ
= γ ∂a
a
− Γabc
φ Ȧ 4 SabS B σ cBȦ ψS

ψA 1 ψA
= γ ∂a
a
+ Γabc γ [γ , γ ]
c a b
,
φ Ȧ 8 φ Ȧ
or equivalently,
m0 c
γ a ∇aΨ = Ψ (3.38)
h̄
[cf. (3.32)], where
1
∇aΨ ≡ ∂aΨ + Γbca [γ b , γ c ]Ψ (3.39)
8
is the covariant derivative of a bispinor (see also Lichnerowicz 1964).
Letting m0 = 0 in (3.34), one obtains the decoupled pair of equations
∇AȦ φ Ȧ = 0, ∇AȦ ψ A = 0. (3.40)
Each of these equations is known as the Weyl equation.
Zero-Rest-Mass Fields
The Weyl equation and the source-free Maxwell equations [see (3.4)] are two par-
ticular cases of the so-called spin-s zero-rest-mass field equations
∇AȦ Φ AB...L = 0 (3.41)

or
∇AȦ Φ ȦḂ...L̇ = 0, (3.42)
where Φ AB...L and Φ ȦḂ...L̇ are totally symmetric 2s-index spinor fields. (See the
discussion about spin at the end of Chapter 1.)
In flat space-time the zero-rest-mass field equations admit plane-wave solutions.
The Ricci rotation coefficients for the orthonormal basis ∂a = ∂ /∂ xa , induced by a
set of Cartesian coordinates xa , are equal to zero; therefore ΓABCḊ and ΓȦḂĊD are also
equal to zero, and substituting, for instance, Φ AB...L = M AB...L exp(−ikRṘ xRṘ ) into
(3.41), where M AB...L = M (AB...L) and kAȦ are constants, and xAȦ denotes the spinor
equivalent of xa [cf. (3.31)], we obtain
ikAȦ M AB...L exp(−ikRṘ xRṘ ) = 0.
Hence, kAȦ M AB...L = 0, and by contracting this last equation with kSȦ it follows that
kAȦ kAȦ M SB...L = 0, which implies that in order to have a nontrivial solution, kAȦ must
be null and therefore kAȦ = ±αA α Ȧ , for some (constant) one-index spinor αA . Thus
we have αA M AB...L = 0, which implies that M AB...L is proportional to α A α B · · · α L
[see (1.230)]. Thus
Φ AB...L = Cα A α B · · · α L exp(∓iαR α Ṙ xRṘ ), (3.43)
where C is some constant [cf. (3.31)].

Similarly, one finds that the equations (3.42) admit plane wave solutions of the
form
Φ ȦḂ...L̇ = Cα Ȧ α Ḃ · · · α L̇ exp(∓iαR α Ṙ xRṘ ), (3.44)
where C is a complex constant and αA is a constant one-index spinor.
As in the case of the circularly polarized electromagnetic wave (3.31), the fields
(3.43) and (3.44) possess a definite helicity. Let βA be a constant one-index spinor
such that α A βA = 1. The vector equivalents of αA β Ȧ + βA α Ȧ and i(αA β Ȧ − βA α Ȧ )
span a two-dimensional spacelike plane orthogonal to the wave four-vector ka ,
and conversely, any two-dimensional spacelike plane orthogonal to ka can be con-
structed in this manner, with a suitable choice of βA (see Exercise 1.10). The
SL(2, C) matrix with entries K A B = eiθ /2 α A βB − e−iθ /2 β A αB corresponds to a ro-
tation through an angle θ in this plane (see Exercise 1.8); under this rotation kAȦ is
invariant and Φ AB...L , given by (3.43), is mapped into eisθ Φ AB...L . At a point with
fixed spatial coordinates (x1 , x2 , x3 ), the effect of this rotation is the same as that pro-
duced in the course of the time t = ∓sθ /ω , where ω is the angular frequency of the
wave (k4 = ω /c). Thus, at a point with fixed spatial coordinates, the field (3.43) ro-
tates with angular velocity ω /s in any two-dimensional spacelike plane orthogonal
to the wave four-vector ka (the sense of the rotation depends on the sign of k4 ).
In an analogous way, under the rotation corresponding to K A B = eiθ /2 α A βB −
e−iθ /2 β A αB , Φ ȦḂ...L̇ is mapped into e−isθ Φ ȦḂ...L̇ , and therefore, for a given sign
of ω , at a point with fixed spatial coordinates the fields (3.43) and (3.44) rotate in
opposite senses.
The zero-rest-mass field equations (3.41) and (3.42) maintain their form under
conformal rescalings of the metric. For example, making use of (2.40), (2.132), and
(2.133), one finds that if Φ AB...L obeys the zero-rest-mass field equations (3.41), then
0 = ∂AȦ Φ ABC...L − Γ A SAȦ Φ SBC...L − (2s − 1)Γ (B SAȦ Φ C...L)SA

= φ −1 ∂ AȦ Φ ABC...L − (φ −1Γ A SAȦ + 32 ∂ SȦ φ −1 )Φ SBC...L
(B
− (2s − 1)(φ −1Γ (B SAȦ + 12 δA ∂ SȦ φ −1 )Φ C...L)SA
= φ −s−2 ∇ AȦ (φ s+1 Φ ABC...L ),
which shows that the zero-rest-mass field equations are form-invariant under con-
formal rescalings if we take Φ AB...L = φ s+1 Φ AB...L , provided that the null tetrads
are related as in (2.132).
A conformally invariant equation for a spin-0 zero-rest-mass field is given by
∇AȦ ∇AȦ Φ + 16 RΦ = 0. (3.45)
In effect, making use of (2.40), (2.132), and (2.133)–(2.135), one finds that
∇AȦ ∇AȦ Φ = ∂AȦ ∂ AȦ Φ − Γ A SAȦ ∂ SȦ Φ − Γ Ȧ ṠȦA ∂ AṠ Φ

= φ −1 ∂ AȦ (φ −1 ∂ AȦ Φ ) − (φ −1Γ A SAȦ + 32 ∂ SȦ φ −1 )φ −1 ∂ SȦ Φ
− (φ −1Γ Ȧ ṠȦA + 32 ∂ AṠ φ −1 )φ −1 ∂ AṠ Φ
= φ −2 ∇ AȦ ∇ AȦ Φ + 2φ −3 (∂ AȦ φ )∂ AȦ Φ
= φ −3 ∇ AȦ ∇ AȦ (φ Φ ) − φ −3 (∇ AȦ ∇ AȦ φ )Φ
= φ −3 ∇ AȦ ∇ AȦ (φ Φ ) − 16 RΦ + 16 R φ −3 Φ ,
that is,
∇AȦ ∇AȦ Φ + 16 RΦ = φ −3 ∇ AȦ ∇ AȦ (φ Φ ) + 16 R (φ Φ ) .
The zero-rest-mass field equations (3.41) and (3.42) are not satisfactory for spins
greater than 1 if the conformal curvature of the space-time is different from zero
(Buchdahl 1958, Plebański 1965). Making use of the Ricci identities (2.126), one
obtains the integrability conditions for (3.41)
0 = ∇B Ȧ ∇AȦ Φ ABCD...L
= BA Φ
ABCD...L
= −(2s − 2)CABS(C Φ D...L)ABS . (3.46)
Note that by virtue of the Bianchi identities (2.116), when the traceless part of the
Ricci tensor vanishes, the Weyl spinor satisfies the zero-rest-mass field equations,
but the right-hand side of (3.46) vanishes identically when ΦABCD = CABCD , and
therefore no algebraic constraints on the Weyl spinor arise in this way.
3.3 Einstein’s Equations
With the Ricci tensor defined as in (2.63), the Einstein field equations are given by
8π G
Rab − 12 Rgab = Tab , (3.47)
c4
where G is Newton’s constant of gravitation and Tab denotes the energy–momentum
tensor of the matter present (see, e.g., Rindler 1977, Wald 1984). Equations (3.47)
yield R = −(8π G/c4)T a a ; therefore, in the case of vacuum (Tab = 0), the Einstein
equations reduce to
Rab = 0, (3.48)
while in the case that the source of the gravitational field is the electromagnetic field
[see (3.20)],
2G
Rab = 4 (Fac Fb c − 14 Fcd F cd gab ),
c
or equivalently, making use of (2.85) and (3.19),
2G
CABȦḂ = f f . (3.49)
c4 AB ȦḂ
The tensor equivalent of CABCDCȦḂĊḊ is a real, traceless, totally symmetric four-
index tensor field Tabcd , known as the Bel–Robinson tensor [cf. (3.19)]. If the trace-
less part of the Ricci tensor vanishes, as a consequence of the Bianchi identities
(2.116), we have ∇AȦ (CABCDCȦḂĊḊ ) = 0, that is, ∇a Tabcd = 0.
Since CABCD εȦḂ εĊḊ and CȦḂĊḊ εAB εCD are the spinor equivalents of 12 (Cabcd +
i Cabcd ) and 12 (Cabcd − i∗Cabcd ), respectively, where
∗
∗
Cabcd ≡ 12 εabrsCrs cd
and
CABCDCȦḂĊḊ = CARBS εȦṘ εḂṠCĊ Ṙ Ḋ Ṡ εC R εD S ,
it follows that
Tabcd = 14 (Carbs + i∗Carbs )(Cc r d s − i∗Cc r d s ).
Using the fact that Cabcd , ∗Cabcd , and Tabcd are real, we conclude that
Tabcd = 14 (CarbsCc r d s + ∗Carbs ∗Cc r d s ) (3.50)
[cf. (3.21)] and

Carbs ∗Cc r d s − ∗CarbsCc r d s = 0.
The Bel–Robinson tensor is analogous to the energy–momentum tensor of the elec-
tromagnetic field in various senses; however, it is not possible to give a local defini-
tion of energy or momentum for the gravitational field (see, e.g., Wald 1984).
3.3 Einstein’s Equations 123
A totally symmetric, traceless, four-index tensor Tabcd possesses a spinor equiva-

lent of the form CABCDCȦḂĊḊ if and only if its spinor equivalent satisfies the relation
TABCDȦḂĊḊ TEFGH Ė Ḟ ĠḢ = TABCDĖ Ḟ ĠḢ TEFGH ȦḂĊḊ (3.51)

[cf. (3.26)]. However, the tensor equivalent of this condition has a complicated form
(see Bergqvist and Lankinen 2004). If TABCDȦḂĊḊ = T(ABCD)(ȦḂĊḊ) satisfies the con-
dition (3.51), then the corresponding spinor CABCD is uniquely defined up to a phase
factor (see Exercise 3.7).
3.3.1 The Goldberg–Sachs Theorem
The following two propositions imply that the integral curves of a repeated principal
null direction of a zero-rest-mass field have several special geometric properties.
Proposition 3.2. If fAB is an algebraically special solution of the source-free Maxwell

equations and lA is a (repeated) principal spinor of fAB , then
l A l B ∇AȦ lB = 0. (3.52)
Proof. By hypothesis, fAB = α lA lB , for some function α , and ∇AȦ fAB = 0; hence
0 = l B ∇AȦ (α lA lB ) = l B [α lA ∇AȦ lB + lB ∇AȦ (α lA )] = α l B lA ∇AȦ lB .

It may be noticed that condition (3.52) is invariant under rescalings of the spinor
field lA . With respect to a spinor frame such that lA ∝ δA2 , (3.52) is equivalent to
Γ111Ȧ = 0 (that is, κ = 0 = σ , in the Newman–Penrose notation); then, making use
of (2.91), one finds that C1111 = 0, which means that any spinor field lA satisfying
(3.52) is a principal spinor of CABCD .
The foregoing proposition is a special case of the following more general result.
Proposition 3.3. If lA is a repeated principal spinor of a totally symmetric spinor

field φAB...L that satisfies the zero-rest-mass field equations
∇AȦ φAB...L = 0, (3.53)

then lA satisfies (3.52).
Proof. Denoting the components φAB...L by φ( j) , where j = 0, 1, . . . , 2s is the number

of indices taking the value 2, i.e., φ(0) = φ11...1 , φ(1) = φ21...1 , . . . , φ(2s) = φ22...2 ,
equations (3.53) are explicitly given by
0 = [∂1Ȧ − 2(s − 1 − j)Γ121Ȧ − (2s − j)Γ112Ȧ]φ( j+1)

− [∂2Ȧ − 2(s − j)Γ122Ȧ + ( j + 1)Γ221Ȧ]φ( j)
+ jΓ222Ȧφ( j−1) + (2s − 1 − j)Γ111Ȧφ( j+2) , (3.54)
j = 0, 1, . . . , 2s − 1. In a spin frame such that lA ∝ δA2 , the fact that lA is a repeated

principal spinor of φAB...L amounts to φ(0) = 0 = φ(1) , and from (3.54), assuming
that φAB...L is different from zero, one obtains Γ111Ȧ = 0, which is equivalent to the
condition (3.52) when lA ∝ δA2 .

There is a converse of Proposition 3.2. Given a one-index spinor lA that satisfies

l A l B ∇AȦ lB = 0, there exist (locally) algebraically special solutions fAB of the source-
free Maxwell equations such that lA is the repeated principal spinor of fAB . Indeed,
looking for solutions of the source-free Maxwell equations (equations (3.54) with
s = 1) such that φ(0) = 0 = φ(1) with respect to a spin frame for which lA ∝ δA2 , we
obtain
0 = Γ111Ȧφ(2) , 0 = (∂1Ȧ + 2Γ121Ȧ − Γ112Ȧ)φ(2) .
The first of these equations is satisfied by virtue of the hypothesis, while the second
one can be written in the equivalent form
∂1Ȧ ln φ(2) = Γ112Ȧ − 2Γ121Ȧ. (3.55)
The integrability condition for these partial differential equations for φ(2) is ob-
tained by applying ∂1 Ȧ to both sides of (3.55):
∂1 Ȧ ∂1Ȧ ln φ(2) = ∂1 Ȧ (Γ112Ȧ − 2Γ121Ȧ). (3.56)
The left-hand side of this last equation contains the commutator
∂1 Ȧ ∂1Ȧ = ∂11̇ ∂12̇ − ∂12̇ ∂11̇ = −Γ S 11 Ȧ ∂SȦ − Γ Ṡ Ȧ Ȧ 1 ∂1Ṡ (3.57)
[see (2.23)], which, since Γ111Ȧ = 0, reduces to
∂1 Ȧ ∂1Ȧ = Γ121 Ȧ ∂1Ȧ − Γ Ȧ Ṡ Ṡ 1 ∂1Ȧ . (3.58)
On the other hand, from (2.91) we obtain
C1211 = (∂1 Ȧ − Γ ȦṘ Ṙ1 )Γ121Ȧ (3.59)

and
2C1112 = (∂1 Ȧ − Γ121Ȧ − Γ ȦṘ Ṙ1 )Γ112Ȧ . (3.60)
Substituting (3.58), (3.55), (3.59), and (3.60) into (3.56), one obtains an identity,
which means that the equations (3.55) are locally integrable. This result, together
with Proposition 3.2, is known as the Mariot–Robinson theorem (Mariot 1954,
Robinson 1961).
If the one-index spinor field lA satisfies the condition (3.52), then the integral
curves of the (real) null vector field v = −l A l Ȧ ∂AȦ form a shear-free family (or
congruence) of null geodesics. From (3.52) we obtain
l A l B l Ȧ ∇AȦ lB = 0, (3.61)
which means that the integral curves of v are geodesics; in fact, (3.61) is equivalent
to
l A l Ȧ ∇AȦ lB = ε lB , (3.62)
for some complex-valued function ε , and therefore,
l A l Ȧ ∇AȦ (lB l Ḃ ) = l A l Ȧ (l Ḃ ∇AȦ lB + lB∇AȦ l Ḃ ) = (ε + ε ) lB l Ḃ ,
that is, ∇v v ∝ v, which is the definition of a geodesic. (Only when ε + ε = 0 will

these geodesics be affinely parameterized.)
By replacing the spinor field lA by l A ≡ f lA , where f is a complex-valued func-
tion, from (3.62) we have
l A l Ȧ ∇AȦ l B = | f |2 ( f ε lB + lB l A l Ȧ ∂AȦ f ).
Hence, choosing f as a solution of the equation l A l Ȧ ∂AȦ ln f = −ε , we obtain

l A l Ȧ ∇AȦ l B = 0; that is, l A is translated parallel to itself along the congruence of
geodesics and ∇v v = 0, with v = −l A l Ȧ ∂AȦ (the integral curves of v are geodesics
affinely parameterized). From l A l Ȧ ∇AȦ l B = 0 we have
l A l B l Ȧ ∇AȦ l B = 0,
which implies that

l A l B ∇AȦ l B = σ l Ȧ , (3.63)
for some function σ .
In order to exhibit the geometric meaning of σ , we consider the behavior of
vectors joining each of the geodesics with its neighbors. More precisely, we deter-
mine how a vector ξ rotates and is stretched as it is dragged along the congruence
of the affinely parameterized geodesics, that is, £v ξ = 0. Since £v ξ = [v , ξ ] =
∇v ξ − ∇ξ v and ∇v v = 0, the directional derivative of the inner product g(v , ξ )
along v vanishes,
v [g(v , ξ )] = g(∇v v , ξ ) + g(v , ∇v ξ ) = g(v , ∇ξ v ) = 12 ξ [g(v , v )] = 0,
where v [ f ] denotes the directional derivative of f along v . Therefore, if ξ is orthog-

onal to v at some point, it will remain orthogonal to v along the geodesic passing
through that point. Hence, we can assume that ξ is orthogonal to v everywhere,
which means that the spinor equivalent of ξ is of the form
ξ AȦ = zmA l Ȧ + zl A mȦ + bl A l Ȧ , (3.64)
where mA is a spinor field such that mA l A = 1 and we can impose the condi-
tion l A l Ȧ ∇AȦ mB = 0 (that is, l A and mA are covariantly constant along v ), z is
a complex-valued function, and b is a real-valued function. Substituting (3.64) into
the spinor equivalent of the relation ∇v ξ − ∇ξ v = 0, we obtain
0 = −l A l Ȧ ∇AȦ ξBḂ + ξ AȦ∇AȦ (l B l Ḃ )

= v [z] mB l Ḃ + v [z] l B mḂ + v [b] l B l Ḃ
+ zmA l Ȧ ∇AȦ (l B l Ḃ ) + zl A mȦ ∇AȦ (l B l Ḃ ).
The contraction of this last equation with l B yields
0 = −v [z] + ρ z + σ z, (3.65)
where
ρ ≡ mA l Ȧ l B ∇AȦ l B (3.66)
and we have made use√ of (3.63). Thus, if s is an affine parameter of the geodesics and
we write z = (x + iy)/ 2, ρ = θ + iω , and σ = |σ | eiχ , equation (3.65) is equivalent
to a pair of real ordinary differential equations that can be expressed in the matrix
form
d x θ + |σ | cos χ −ω + |σ | sin χ x
= . (3.67)
ds y ω + |σ | sin χ θ − |σ | cos χ y
Since
mA l Ȧ + l A mȦ im l − il m
ξAȦ = x √ + y A Ȧ√ A Ȧ + bl A l Ȧ
2 2
[see (3.64)], the real-valued functions x and y are the components, with respect to
an orthonormal basis, of the projection of ξ on a two-dimensional plane orthogonal
to v (see Exercise 1.10). At each point of M where the real-valued function χ has
some specific value χ0 , say, by means of a rotation in this plane through an angle
χ0 /2, which amounts to replacing z by eiχ0 /2 z, one obtains an equation of the form
(3.67) with χ = 0, at that point only [see (3.65)], and then one can see that θ rep-
resents the rate of expansion of the congruence, ω is the angular velocity, or twist,
with which the congruence rotates, and |σ | corresponds to a distortion, or shear, of
the congruence whose principal axes form an angle χ0 /2 with respect to the basis
vectors employed in (3.67).
Since l A = f lA , from (3.52) it follows that l A l B ∇AȦ l B = 0, which, according to
the definition (3.63), means that σ = 0; thus, the integral curves of the vector field
v = −l A l Ȧ ∂AȦ form a shear-free congruence of null geodesics, as stated above.
As pointed out at the end of Section 3.2, when the traceless part of the Ricci
tensor vanishes, the Bianchi identities imply that the Weyl spinor satisfies the zero-
rest-mass field equations; therefore, according to the last proposition, if CABĊḊ = 0
(as in the case in which the Einstein vacuum field equations hold), each repeated DP
spinor defines a shear-free congruence of null geodesics.
There exists a converse of this result, which does not require the vanishing of the
traceless part of the Ricci tensor.
Proposition 3.4. Let lA be a one-index spinor field such that
l A l B ∇AȦ lB = 0 and l A l BCABȦḂ = 0. (3.68)

Then lA is a repeated principal spinor of CABCD , i.e., l A l B lCCABCD = 0.
Proof. In a spin frame such that lA ∝ δA2 , conditions (3.68) are equivalent to Γ111Ȧ =
0 and C11ȦḂ = 0, respectively. Then, from (2.91) we obtain C1111 = 0, together with
C1211 = (∂1 Ȧ − Γ ȦṘ Ṙ1 )Γ121Ȧ (3.69)

and
2C1112 = (∂1 Ȧ − Γ121Ȧ − Γ ȦṘ Ṙ1 )Γ112Ȧ . (3.70)
On the other hand, with C11ȦḂ = 0 and Γ111Ȧ = 0, we have ∇1 ṠC11ṘṠ = ∂1 ṠC11ṘṠ +
2Γ A 11 ṠCA1ṘṠ + Γ Ȧ Ṙ Ṡ 1C11ȦṠ + Γ Ȧ Ṡ Ṡ 1C11ṘȦ = 0, and taking into account that C1111 =
0, the Bianchi identities (2.116) yield
0 = ∇A ṘC111A
= ∂ A ṘC111A + 3Γ S 1 A ṘC11SA + Γ S A A ṘC111S
= (∂1Ṙ − 2Γ121Ṙ − 4Γ112Ṙ)C1112 ,
i.e.,
∂1ṘC1112 = (2Γ121Ṙ + 4Γ112Ṙ)C1112 . (3.71)
Then, making use of (3.71), (3.70), and (3.69), expressed in the equivalent form
C1211 = (∂1 Ȧ − Γ121Ȧ − Γ ȦṘ Ṙ1 )Γ121Ȧ , (3.72)
we obtain
∂1 Ṙ ∂1ṘC1112 = (2Γ121Ṙ + 4Γ112Ṙ)∂1 ṘC1112 + C1112 ∂1 Ṙ (2Γ121Ṙ + 4Γ112Ṙ)

= C1112 ∂1 Ṙ (2Γ121Ṙ + 4Γ112Ṙ)

= 2[C1112 + (Γ121Ȧ + Γ ȦṘ Ṙ1 )Γ121Ȧ ]

+ 4[2C1112 + (Γ121 Ȧ + Γ ȦṘ Ṙ1 )Γ112Ȧ ] C1112

= 10C1112 + (Γ121 Ȧ + Γ ȦṘ Ṙ1 )(2Γ121Ȧ + 4Γ112Ȧ) C1112 .
According to (3.58) and (3.71), the left-hand side of this last equation is also equal
to
(Γ121 Ȧ + Γ ȦṘ Ṙ1 )∂1ȦC1112 = (Γ121 Ȧ + Γ ȦṘ Ṙ1 )(2Γ121Ȧ + 4Γ112Ȧ)C1112 .
Thus, we conclude that C1112 = 0, which, together with C1111 = 0, is equivalent to
C111A = 0 and to the covariant expression l A l B lCCABCD = 0 in a frame such that
lA ∝ δA2 .

Thus, if the Einstein vacuum field equations hold, the conformal curvature is
algebraically special if and only if the space-time admits a shear-free congruence of
null geodesics. This result is known as the Goldberg–Sachs theorem (Goldberg and
Sachs 1962).
In the case of a solution of the Einstein–Maxwell equations with an algebraically

special electromagnetic field, from (3.49) we have l A l BCABȦḂ = 0 if lA is the repeated
principal spinor of fAB , and from (3.52) and (3.68) it follows that lA is also a repeated
principal spinor of CABCD , which is therefore algebraically special.
The space-times that admit a shear-free congruence of null geodesics defined by
a multiple DP spinor have several remarkable properties.
Proposition 3.5. If lA is a repeated principal spinor of the Weyl spinor CABCD

(CABCD l A l B lC = 0) such that l A l B ∇AȦ lB = 0, then there exists locally a complex-
valued function φ such that
l B ∇AȦ lB = lA l B ∂BȦ ln φ . (3.73)
Proof. In a spinor frame such that lA = δA2 , equation (3.73) is equivalent to
Γ111Ȧ = 0, Γ112Ȧ = ∂1Ȧ ln φ . (3.74)
The integrability condition for the function φ is obtained by applying the operator
∂1 Ȧ to both sides of the second equation in (3.74), which, making use of (3.58) and
(3.74), is equivalent to the condition
∂1 ȦΓ112Ȧ = Γ211 ȦΓ112Ȧ − Γ Ṡ Ȧ Ȧ 1Γ112Ṡ ,
which is satisfied, since by hypothesis, C1112 = 0 [see (3.70)].

In a similar manner one finds that (3.72) with C1112 = 0 implies the local exis-
tence of a function ζ such that
Γ121Ȧ = ∂1Ȧ ln ζ . (3.75)
As a first application of the existence of the functions φ and ζ we can now establish
the following proposition.
Proposition 3.6. In a space-time that admits a shear-free congruence of null geodesics

defined by a multiple DP spinor lA , up to gauge transformations the most general
solution of the self-duality condition (3.10) can be expressed locally as
Φ1Ḃ = 0, Φ2Ḃ = (∂1Ḃ + 2Γ121Ḃ + Γ112Ḃ)χ , (3.76)
with respect to a spin frame such that lA ∝ δA2 , where χ is a complex potential
satisfying the equation
(∂2 Ḃ + Γ122Ḃ − Γ ḂṠ Ṡ2 )(∂1Ḃ + 2Γ121Ḃ + Γ112Ḃ)χ = 0.
Proof. The self-duality condition ∇(B Ḃ ΦC)Ḃ = 0 is explicitly given by the equations
0 = ∂(B Ḃ ΦC)Ḃ + Γ S (BC) Ḃ ΦSḂ + Γ Ṡ Ḃ Ḃ (B ΦC)Ṡ , (3.77)

which are equivalent to (3.12). Taking B = 1 = C in (3.77), we have
0 = (∂1 Ḃ − Γ121Ḃ − Γ ḂṠ Ṡ1 )Φ1Ḃ + Γ111Ḃ Φ2Ḃ . (3.78)
Making use of the fact that we have Γ111 Ḃ = 0 with respect to a spin frame such
that the multiple DP spinor that defines a shear-free geodesic congruence has the
form lA ∝ δA2 , from (3.78) it follows that there exists locally a function ξ such that
Φ1Ḃ = −∂1Ḃ ξ . Indeed, the integrability condition for ξ is given by
∂1 Ȧ ∂1Ȧ ξ = Γ121 Ȧ ∂1Ȧ ξ − Γ Ȧ Ṡ Ṡ 1 ∂1Ȧ ξ
[see (3.58)], which amounts to
−∂1 Ȧ Φ1Ȧ = −Γ121 Ȧ Φ1Ȧ + Γ Ȧ Ṡ Ṡ 1 Φ1Ȧ
and coincides with (3.78). Then, the gauge transformation ΦBḂ → ΦBḂ + ∂BḂ ξ
yields Φ1Ḃ = 0.
Taking B = 1, C = 2 in (3.77) with Φ1Ḃ = 0, we obtain the condition
0 = (∂1 Ḃ + Γ121Ḃ + Γ112Ḃ − Γ ḂṠ Ṡ1 )Φ2Ḃ . (3.79)
Making use of (3.74) and (3.75), this last condition can also be written as
0 = (∂1 Ḃ − Γ121Ḃ − Γ ḂṠ Ṡ1 )(ζ 2 φ Φ2Ḃ )
[cf. (3.78)], which implies the local existence of a function η such that ζ 2 φ Φ2Ḃ =
∂1Ḃ η . Hence, letting χ ≡ ζ −2 φ −1 η , we have
Φ2Ḃ = ζ −2 φ −1 ∂1Ḃ (ζ 2 φ χ ) = (∂1Ḃ + 2Γ121Ḃ + Γ112Ḃ)χ , (3.80)
where we have made use again of (3.74) and (3.75).

Taking A = 2 = C in (3.77) we obtain
0 = (∂2 Ḃ + Γ122Ḃ − Γ ḂṠ Ṡ2 )Φ2Ḃ ,
which, by virtue of (3.80), implies that the potential χ must satisfy the equation
0 = (∂2 Ḃ + Γ122Ḃ − Γ ḂṠ Ṡ2 )(∂1Ḃ + 2Γ121Ḃ + Γ112Ḃ)χ (3.81)
(Torres del Castillo 1999).

It can be readily verified that the covariant expression
ΦBḂ = φ −2 ∇S Ḃ (φ 2 lS lB ψ ) (3.82)
reduces to (3.76), with χ = (l2 )2 ψ , in a spinor frame such that lA ∝ δA2 . If we assume
that φ and ψ are scalar functions, then both sides of (3.82) are spinor fields of the
same type, and therefore the equality (3.82) is valid in any spinor frame.
Hence, according to the results of Section 3.1, the most general real source-free
electromagnetic field is given locally by
fȦḂ = ∇B (Ȧ [φ −2 ∇S Ḃ) (φ 2 lS lB ψ )] (3.83)
(see also Cohen and Kegeles 1974, Wald 1978, Mustafa and Cohen 1987). The
second-order linear partial differential equation (3.81) can be solved by separa-
tion of variables in all the type-D solutions of the Einstein vacuum field equations
(Kamran 1987, Torres del Castillo 1988).
In the case of flat space-time there exist covariantly constant one-index spinors
(which form a complex two-dimensional vector space), and any of them satisfies the
conditions of Proposition 3.5. Then the function φ defined by (3.73) can be taken
equal to 1, and (3.83) reduces to
fȦḂ = ∇A Ȧ ∇B Ḃ (lA lB ψ ),
where lA is a covariantly constant spinor and ψ is a complex-valued function that

has to satisfy the wave equation (see also Penrose 1965).
The solution of the Weyl equation can also be expressed locally in terms of
a complex potential when the space-time admits a shear-free congruence of null
geodesics defined by a multiple DP spinor. The Weyl equation 0 = ∇A Ȧ ηȦ is given
explicitly by
0 = (∂1 Ȧ − Γ ȦṠ Ṡ1 )ηȦ , (3.84)

0 = (∂2 − ΓȦ ȦṠ
Ṡ2 )ηȦ . (3.85)
Making use of (3.75), equation (3.84) can be rewritten as
0 = (∂1 Ȧ − Γ121Ȧ − Γ ȦṠ Ṡ1 )(ζ ηȦ ),
which implies the local existence of a complex-valued function η such that ζ ηȦ =
∂1Ȧ η , and writing η = ζ χ , we have
ηȦ = ζ −1 ∂1Ȧ (ζ χ ) = (∂1Ȧ + Γ121Ȧ)χ . (3.86)
Substituting this expression into (3.85), we obtain
0 = (∂2 Ȧ − Γ ȦṠ Ṡ2 )(∂1Ȧ + Γ121Ȧ)χ . (3.87)
Then, as in the case of (3.82), one finds that with respect to an arbitrary frame, the
solution of the Weyl equation is given by
ηḂ = φ −1 ∇S Ḃ (φ lS ψ ), (3.88)
where ψ is a complex potential that has to satisfy the condition

∇BḂ φ −1 ∇S Ḃ (φ lS ψ ) = 0
(the left-hand side of this last equation is proportional to l B , and therefore this equa-
tion imposes a single condition on ψ ).
In any type-D solution of the Einstein vacuum field equations, equation (3.87)
can be solved by separation of variables.
3.3.2 Space-Times with Symmetries. Ernst Potentials
With each nonnull Killing vector of a solution of the Einstein vacuum field equa-
tions, there is associated a complex function, called the Ernst potential, which can
be used to construct the metric. In the case of a solution of the Einstein–Maxwell
equations, there is a pair of Ernst potentials associated with each nonnull Killing
vector, provided that the Lie derivative of the electromagnetic field with respect to
the Killing vector vanishes. In this subsection the equations for the Ernst poten-
tials are obtained by means of the spinor formalism (alternative derivations, making
use of the tensor formalism or of differential forms, can be found, for instance, in
Chandrasekhar 1983, Heusler 1996, Stephani et al. 2003). Owing to their nonlin-
earity, it is difficult to find explicit solutions of the Ernst equations; fortunately, the
symmetries of the Ernst equations make it possible to generate new solutions from
a given one (see, e.g., Stephani et al. 2003).
An arbitrary vector field K is (locally) hypersurface orthogonal if and only if
ωa ≡ εabcd K b ∇c K d
vanishes (see below). Making use of (1.64), one finds that the spinor equivalent of
ωa is given by
ωAȦ = −i(K BḂ ∇AḂ KBȦ − K BḂ∇BȦ KAḂ ).
According to (2.142) and (2.71), the components ωAȦ can be obtained by means of
the relation
Ka θ a ∧ d(Kb θ b ) = iωAȦ θ̆ AȦ . (3.89)
In particular, if K is a Killing vector, from (2.140) it follows that
ωAȦ = 2i(K B Ȧ LAB − KA Ḃ LȦḂ ). (3.90)
Similarly, if we let
f ≡ K BḂ KBḂ (3.91)
(i.e., f = −K a Ka ), then using again (2.140), we have
∇AȦ f = 2K BḂ ∇AȦ KBḂ = 2(K B Ȧ LAB + KA Ḃ LȦḂ ), (3.92)

[cf. (3.90)]. Hence

ωAȦ = 4iK B Ȧ LAB − i∇AȦ f . (3.93)
We can now prove the assertion made at the beginning of this subsection. When
ωa = 0, from (3.93) we see that ∇AȦ f = 4K B Ȧ LAB ; therefore, if K is nonnull,
K(C Ȧ ∇A)Ȧ f = −2 f LAC , or equivalently, K(C Ȧ ∇A)Ȧ f = f ∇(A|Ȧ| KC) Ȧ [see (2.141)].
This last equation can also be written in the form ∇(A|Ȧ| [ f −1 KC) Ȧ ] = 0, and taking
into account that the spinor equivalent of the curl of ωa is given by
∇AȦ ωBḂ − ∇BḂ ωAȦ = εȦḂ ∇(A Ṙ ωB)Ṙ + εAB ∇R (Ȧ ω|R|Ḃ) ,
it follows that there exists (locally) a real-valued function u such that
KAȦ = f ∇AȦ u, (3.94)
thus showing that K is orthogonal to the hypersurfaces u = constant. Alternatively,

if ωa = 0, from (3.89) and Frobenius’s theorem it follows that the one-form Ka θ a
is locally integrable; that is, locally there exist functions F and G such that Ka θ a =
FdG, which amounts to
KAȦ = F∇AȦ G.
(Note that this last proof is valid also if K is null and even if K is not a Killing
vector; on the other hand, (3.94) explicitly gives an integrating factor for the one-
form Ka θ a .)
By virtue of (2.140) and (2.144), the anti-self-dual part of the curl of ωa is deter-
mined by
∇(C Ȧ ωA)Ȧ = 4iK B Ȧ ∇(C Ȧ LA)B = 4iK BḂCB(A|Ḃ| Ṡ KC)Ṡ . (3.95)
The right-hand side of (3.95) vanishes if K is an eigenvector of the Ricci tensor or
if the traceless part of the Ricci tensor is equal to zero. Thus, if we assume that the
Einstein vacuum field equations are satisfied, from (3.95) it follows that there exists
(at least locally) a real-valued function ω , the twist potential, such that
ωAȦ = −∇AȦ ω , (3.96)
and (3.93) yields

4K B Ȧ LAB = ∇AȦ χ , (3.97)
where
χ ≡ f + iω . (3.98)
Therefore, if K is nonnull (i.e., f = 0), we have
1
LAB = − K Ȧ ∇ χ . (3.99)
2 f B AȦ
Then (3.97), (2.144), with Rab = 0, (2.140), and (3.99), give
∇AȦ ∇AȦ χ = 4K B Ȧ ∇AȦ LAB + 4LAB∇AȦ K B Ȧ

= 8LAB LAB
1
= (∇AȦ χ )(∇AȦ χ ).
f
Thus, χ obeys the nonlinear equation
(Re χ )∇a ∇a χ = (∇a χ )(∇a χ ). (3.100)
This equation is known as the Ernst equation and χ is the Ernst potential (Ernst
1968a). (Note that from (3.97) it follows that K AȦ ∇AȦ χ = 0.)
The Ernst potential is useful in finding exact solutions of the Einstein vacuum
field equations. For instance, the metric of a stationary axisymmetric space-time
can be locally expressed in the form

g = −e2U (dt + Adϕ )2 + e−2U e2γ (dρ 2 + dz2 ) + ρ 2dϕ 2 ,
where t, ρ , ϕ , z form a coordinate system and the functions U, γ , and A depend

on ρ and z only (Papapetrou 1953; see also Chandrasekhar 1983, Stephani et al.
2003). This metric possesses the timelike Killing vector K = ∂ /∂ t, for which f =
K AȦ KAȦ = −K a Ka = e2U ; hence, we can also write

g = − f (dt + Adϕ )2 + f −1 e2γ (dρ 2 + dz2) + ρ 2dϕ 2 .
Then, the Ernst equation (3.100) takes the explicit form

1 ∂ ∂χ ∂ 2χ ∂χ 2 ∂χ 2
(Re χ ) ρ + 2 = + ,
ρ ∂ρ ∂ρ ∂z ∂ρ ∂z
which does not involve the unknown functions f , γ , and A. The real part of a given
solution of the Ernst equation gives f [see (3.98)], while the imaginary part of χ
determines A by means of

f2 ∂A ∂A ∂ω ∂ω
+i = −i . (3.101)
ρ ∂ρ ∂z ∂z ∂ρ
In fact, Ka θ a = − f (dt + Adϕ ), and therefore Ka θ a ∧ d(Kb θ b ) = f 2 dt ∧ dA ∧ dϕ .

Then (3.89) and (3.96) lead to (3.101).
The function γ is also determined by the Ernst potential, as a consequence of the
Einstein equations (see, e.g., Heusler 1996, Sec. 4.1, Stephani et al. 2003, §19.5).
Solutions of the Einstein–Maxwell Equations
When there is an electromagnetic field present, expressions analogous to (3.99) and

(3.100) can be given, provided that the electromagnetic field is invariant under a
one-parameter group of isometries. The electromagnetic field is invariant under the

one-parameter group of isometries generated by a Killing vector K = −K AȦ∂AȦ if
the Lie derivative of the electromagnetic spinor fAB with respect to K vanishes, i.e.,
−KCĊ ∇CĊ fAB + 2LC (A fB)C = 0
[see (2.143)], or equivalently, making use of (1.27) and (2.140),
0 = −KCĊ ∇(A|Ċ fC|B) + K(AĊ ∇S |Ċ fS|B) + 2LC (A fB)C

= −∇(A|Ċ [KCĊ fC|B) ] + fC(A ∇B)Ċ KCĊ + K(AĊ ∇S |Ċ fS|B) + 2LC (A fB)C
= −∇(A|Ċ [KCĊ fC|B) ] + K(AĊ ∇S |Ċ fS|B) .
Thus, if the source-free Maxwell equations are satisfied, letting PBĊ ≡ KC Ċ fBC , we
have ∇(AĊ PB)Ċ = 0. Furthermore,
∇B (Ȧ P|B|Ċ) = ∇B (Ȧ [KCĊ) fBC ] = fBC ∇B (Ȧ KC Ċ) = 0
[see (2.140)]. Hence the curl of the tensor equivalent of PAȦ is equal to zero, and
therefore, there exists, locally, a complex-valued function Φ such that
c2
KCĊ fBC = √ ∇BĊ Φ , (3.102)
2 G
where the constant factor is introduced for later convenience. The symmetry of fBC
implies that K BĊ ∇BĊ Φ = 0. Thus, if Ka is nonnull,
c2
fBC = − √ KC Ċ ∇BĊ Φ . (3.103)
f G
The source-free Maxwell equations imply K BḂ ∇A Ḃ fAB = 0, which, making use
of (3.103) and (2.140), reduces to
f ∇AȦ ∇AȦ Φ = 4K BȦLA B ∇AȦ Φ . (3.104)
(See Exercise 3.9.)

Substituting the Einstein equations (3.49) and (3.102) into (3.95), one finds that
∇(C Ȧ ωA)Ȧ = −2i(∇(C Ȧ Φ )(∇A)Ȧ Φ )

= i∇(C Ȧ Φ ∇A)Ȧ Φ − Φ ∇A)Ȧ Φ ,
which implies (locally) the existence of a real-valued function ω such that ωAȦ =
i(Φ ∇AȦ Φ − Φ ∇AȦ Φ ) − ∇AȦ ω , and therefore (3.93) gives
4K B Ȧ LAB = ∇AȦ χ + 2Φ ∇AȦ Φ , (3.105)

where
χ ≡ f − ΦΦ + iω . (3.106)
From (3.105), making use of (2.140), (2.144), (3.104), the Einstein equations, and
(3.102), we obtain
∇AȦ ∇AȦ χ = 4LAB ∇AȦ K B Ȧ + 4K BȦ ∇AȦ LAB − 2(∇AȦΦ )(∇AȦ Φ )

− 2Φ ∇AȦ ∇AȦ Φ
8G
= 8LAB LAB + 4 K B Ȧ fBR K RṠ fṠ Ȧ − 2(∇AȦΦ )(∇AȦ Φ )
c
−1 BȦ A
− 8 f K L B Φ ∇AȦ Φ
= f −1 (∇AȦ χ + 2Φ ∇AȦ Φ )∇AȦ χ .
Thus, from (3.104)–(3.106) we finally obtain the system of nonlinear equations

(Ernst 1968b)
(Re χ + ΦΦ )∇a ∇a χ = (∇a χ + 2Φ ∇a Φ )∇a χ ,

(3.107)
(Re χ + ΦΦ )∇a ∇a Φ = (∇a χ + 2Φ ∇a Φ )∇a Φ .
The Ernst equations (3.107) can be used to find stationary axisymmetric solutions
of the Einstein–Maxwell equations, and as in the case of the Ernst equation (3.100),
the symmetries of these equations make it possible to generate new solutions from
a given one.
Relations Between Killing Vectors and the Conformal Curvature
The equations derived above are also useful in the derivation of some relations be-
tween the properties of a Killing vector and the conformal curvature (equivalent
results using other formalisms have been given in Debney 1971a,b and Catenacci
et al. 1980). In the rest of this section we shall restrict ourselves to solutions of the
Einstein vacuum field equations.
If the space-time admits a null Killing vector K, we can write
KAḂ = ±αA α Ḃ , (3.108)
where αA is some one-index spinor field, and from the Killing equations (2.140) we
obtain
0 = α A α C (∇AḂ KCḊ + ∇CḊ KAḂ ) = ±(α A α Ḋ α C ∇AḂ αC + α A α Ḃ α C ∇CḊ αA ),
which implies
α A α C ∇AḂ αC = 0,
that is, K is tangent to a shear-free congruence of null geodesics [see (3.52)]. Then,
by virtue of the Goldberg–Sachs theorem (Proposition 3.4), since the Einstein vac-
uum field equations are assumed to hold, αA is a repeated DP spinor and the confor-
mal curvature is algebraically special.
In the present case, in which the function f = K BḂ KBḂ vanishes, from (3.92) and
(3.108) we have
0 = K B Ȧ LAB + KA Ḃ LȦḂ = ±(α B α Ȧ LAB + αA α Ḃ LȦḂ ),
which implies that LAB α B = iBαA , where B is a real-valued function; then (3.90)
and (3.108) give ωAȦ = −4BKAȦ .
The condition ∇(C Ȧ ωA)Ȧ = 0 [see (3.95)] yields
0 = B∇(C Ȧ KA)Ȧ + K(A|Ȧ| ∇C) Ȧ B = 2BLCA ± α(A α |Ȧ| ∇C) Ȧ B.
Contracting this relation with α C , we obtain 2 iB2 αA ± 12 αA α Ȧ α C ∇C Ȧ B = 0, which

implies B = 0.
Thus, LAB α B = 0; that is, αA is a double principal spinor of LAB , and LAB is
algebraically special. Hence, we can write
LAB = ψαA αB (3.109)
for some complex-valued function ψ . Furthermore, ωAȦ = 0; that is, K is hypersur-

face orthogonal.
Substituting (3.108) and (3.109) into (2.144), we see that
∇AȦ (ψαB αC ) = ∓CD ABC αD α Ȧ .
Hence, contracting with α B α Ȧ we obtain (assuming ψ = 0)
α B α Ȧ ∇AȦ αB = 0,
which shows that the twist and the expansion of the congruence of null geodesics
defined by αA are equal to zero [see (3.66)].
On the other hand, if we assume that LAB is algebraically special [as in (3.109)]
but that the corresponding Killing vector K is nonnull, writing LAB = αA αB , from
(2.144), assuming again that the Einstein vacuum field equations hold, we arrive at
0 = α B α C ∇AȦ (αB αC ) = −CD ABC α B α C KDȦ .
Since K is nonnull, it follows that CABCD α B α C = 0, thus showing that the conformal
curvature is of type III or N, or equal to zero.
3.4 Killing Spinors 137
3.4 Killing Spinors
Walker and Penrose (1970) showed that any type-D solution of the Einstein vacuum
field equations admits a quadratic integral of the null geodesic equations (see also
Hughston et al. 1972, Hughston and Sommers 1973). The existence of this first
integral follows from that of a certain two-index spinor field, which has been called
a Killing spinor.
Apart from their relationship with first integrals of the geodesic equations, the
Killing spinors also appear in “symmetry operators” for the Weyl, Dirac, and
Maxwell equations (Carter and McLenaghan 1979, Kamran and McLenaghan 1983,
1984a, Torres del Castillo 1985, 1986, McLenaghan et al. 2000).
A D(k, 0) Killing spinor is a totally symmetric 2k-index spinor field MAB...D such
that
∇(A Ȧ MBC...L) = 0. (3.110)
(Note that the equations for a conformal Killing vector can be expressed in the
form ∇(A (Ȧ KB) Ḃ) = 0.) It can readily be seen that the symmetrized tensor product of
Killing spinors is also a Killing spinor. Equation (3.110) is also known as the twistor
equation.
By virtue of the Ricci identities (2.126),
∇(A|Ȧ| ∇B Ȧ MCD...L) = − (AB MCD...L) = −2kCS (ABC MD...L)S ;
hence, an integrability condition for the set of equations (3.110) is
CS (ABC MD...L)S = 0. (3.111)
As we shall show below, if k = 2, 4, 6, . . ., or if the traceless part of the Ricci tensor

vanishes, these equations imply that the conformal curvature must be of type D
or N, or equal to zero [see (2.130)]. Denoting the components of MAB...D by M( j) ,
where j = 0, 1, . . . , 2k is the number of indices taking the value 2, i.e., M(0) = M11...1 ,
M(1) = M21...1 , . . . , M(2k) = M22...2 , the integrability conditions (3.111) are given
explicitly by the set of 2k + 3 equations
0 = (2k + 2 − j)(2k + 1 − j)(2k − j)C(5)M( j+1)

+ 2(2 j − k)(2k + 2 − j)(2k + 1 − j)C(4)M( j)
+ 6 j(2k + 2 − j)( j − k − 1)C(3)M( j−1)
+ 2 j( j − 1)(2 j − 3k − 4)C(2)M( j−2)
− j( j − 1)( j − 2)C(1)M( j−3) = 0, (3.112)
where j can take the values 0, 1, . . . , 2k + 2, C(5) ≡ C1111 , C(4) ≡ C1112 , C(3) ≡ C1122 ,
C(2) ≡ C1222 , and C(1) ≡ C2222 .
If lA is a principal spinor of MAB...D , then in a spinor frame such that lA ∝ δA2 ,

we have M(0) = 0, and from (3.112) it follows that C(5) = 0, which means that a
principal spinor of MAB...D is also a principal spinor of CABCD . Substituting the values
M(0) = 0 and C(5) = 0 into (3.112), one finds that if k = 2, 4, 6, . . ., then C(4) = 0;
hence, when k is not an even integer, each principal spinor of MAB...D is at least a
double principal spinor of CABCD . Therefore (if CABCD = 0), there are at most two
principal spinors of MAB...D that are not proportional to each other. When MAB...D
has a 2k-fold repeated principal spinor, from (3.111) or (3.112) one concludes that
CABCD is of type N; otherwise, CABCD must be of type D (Torres del Castillo 1984a).
The fact that a principal spinor of a D(k, 0) Killing spinor, MAB...D , is a principal
spinor of CABCD also follows from the fact that each principal spinor of MAB...D sat-
isfies condition (3.52) (Hughston et al. 1972). Indeed, equations (3.110) amount to
∂(A Ȧ MBC...L) + 2kΓ S (AB Ȧ MC...L)S = 0,
or more explicitly,
0 = j [∂2Ȧ − 2(k + 1 − j)Γ122Ȧ − (2k + 1 − j)Γ221Ȧ] M( j−1)

+ (2k + 1 − j)[∂1Ȧ − 2(k − j)Γ121Ȧ + jΓ112Ȧ ] M( j)
− j( j − 1)Γ222Ȧ M( j−2) + (2k + 1 − j)(2k − j)Γ111Ȧ M( j+1) , (3.113)
where j can take the values 0, 1, . . . , 2k + 1. Substituting M(0) = 0 into (3.113), one
finds that Γ111Ȧ = 0, which implies C1111 = 0 [see (2.91)]. Furthermore, as we have
seen, Γ111Ȧ = 0 means that any one-index spinor lA such that lA ∝ δA2 satisfies condi-
tion (3.52); hence, if the traceless part of the Ricci tensor vanishes, Proposition 3.4
implies that each principal spinor of MAB...D is at least a double principal spinor of
CABCD (without restriction on the value of k). Therefore, when the traceless part of
the Ricci tensor vanishes, CABCD must be of type N or D, or equal to zero.
If we do not assume that the traceless part of the Ricci tensor vanishes, when k =
2, for instance, the integrability conditions (3.111) imply that MABCD is proportional
to CABCD , without restriction on the algebraic type of the conformal curvature.
When CABCD is of type D one can find a spin frame such that only C(3) is different
from zero; then (3.112) implies that M(k) is the only nonvanishing component of
MAB...L , and necessarily k must be an integer. In the case that CABCD is of type N,
one can find a spin frame such that only C(1) is different from zero, and (3.112)
implies that M(2k) is the only nonvanishing component of MAB...L .
In the case that CABCD is of type D, assuming that M(k) is the only nonvanishing
component of MAB...D , equations (3.113) reduce to
0 = Γ111Ȧ,
0 = Γ222Ȧ,
0 = (∂1Ȧ + kΓ112Ȧ)M(k) , (3.114)
0 = (∂2Ȧ − kΓ221Ȧ)M(k) .
These equations are integrable if the traceless part of the Ricci tensor vanishes
(Walker and Penrose 1970); in fact, the Bianchi identities (2.116) with CABȦḂ = 0
imply that the Weyl spinor satisfies the zero-rest-mass equations (3.53), and as-
suming that C(3) is the only nonvanishing component of CABCD , equations (3.54)
reduce to
0 = Γ111Ȧ,
0 = Γ222Ȧ,
0 = (∂1Ȧ − 3Γ112Ȧ)C(3) ,
0 = (∂2Ȧ + 3Γ221Ȧ)C(3) .
Therefore, the solution of (3.114) is given by
M(k) = const. [C(3) ]−k/3 . (3.115)
Thus, a type-D solution of the Einstein vacuum field equations (such as the
Schwarzschild or the Kerr metric) admits a D(k, 0) Killing spinor, for k = 1, 2, . . . .
(According to (3.115), up to a constant factor, all these Killing spinors are sym-
metrized products of the D(1, 0) Killing spinor with itself.)
Similarly, one can give the solution of (3.114) if the space-time metric satisfies
the Einstein–Maxwell equations with an algebraically general electromagnetic field
such that each principal spinor of fAB satisfies (3.52) (as in the case of the Reissner–
Nordström and the Kerr–Newman solutions). With these assumptions, the Einstein
equations (3.49) imply that if l A is a principal spinor of fAB , then l A l BCABĊḊ = 0,
and by Proposition 3.4, l A is a repeated DP spinor, which implies that the conformal
curvature is of type D. With respect to a spinor frame such that f12 is the only
nonvanishing component of fAB , we have Γ111Ȧ = 0 = Γ222Ȧ , and the source-free
Maxwell equations [(3.54) with s = 1] give
0 = (∂1Ȧ − 2Γ112Ȧ)φ(1) ,
0 = (∂2Ȧ + 2Γ221Ȧ)φ(1) .
Thus, the solution of (3.114) is
M(k) = const. (φ(1) )−k/2 = const. ( f12 )−k/2 .
When CABCD is of type N, assuming that M(2k) is the only nonvanishing compo-
nent of MAB...D , equations (3.113) give
0 = Γ111Ȧ,
0 = (∂1Ȧ + 2kΓ121Ȧ + 2kΓ112Ȧ)M(2k) , (3.116)
0 = (∂2Ȧ + 2kΓ122Ȧ)M(2k) .
The system of equations (3.116) is integrable for k = 1/2, 1, 3/2, 2, . . . if and only
if it is integrable for k = 1/2; in fact, if we denote by Q the solution of equations
(3.116) for k = 1/2, that is,
(∂1Ȧ + Γ121Ȧ + Γ112Ȧ)Q = 0, (∂2Ȧ + Γ122Ȧ)Q = 0,
then the solution of equations (3.116) for any other value of k is M(2k) = Q2k , which
means that in the present case, any D(k, 0) Killing spinor MAB...L is of the form
MAB...L = MA MB · · · ML , where MA is a D(1/2, 0) Killing spinor.
As pointed out at the beginning of this section, the existence of a D(1, 0) Killing
spinor allows us to construct a quadratic integral of the null geodesic equations.
Indeed, if αA α Ȧ is the spinor equivalent of the tangent vector of an affinely param-
eterized null geodesic, that is, α A α Ȧ ∇AȦ (αB α Ḃ ) = 0, then from (3.110) it follows
that if MAB is a D(1, 0) Killing spinor, the directional derivative of the real-valued
function MAB M ȦḂ α A α Ȧ α B α Ḃ along the geodesic is equal to zero. If MAB satisfies
the additional condition ∇BĊ MAB + ∇A Ḃ MĊḂ = 0, one can find a quadratic constant
of motion for any geodesic (see Exercise 3.14).
According to the definition (3.110), MAB...D is a D(k, 0) Killing spinor if and only
if there exists KAB...DL̇ = K(AB...D)L̇ such that
2k
∇A Ȧ MBC...L = ε K Ȧ
. (3.117)
2k + 1 A(B C...L)
Then
KAB...L Ȧ = ∇SȦ MSAB...L , (3.118)
and making use of (2.117) and (3.117), one finds that
∇(A (Ȧ KB...L) Ḃ) = ∇(A (Ȧ ∇|S|Ḃ) MB...L)S

= ∇S(Ḃ ∇(A Ȧ) MB...L)S − ȦḂ
MAB...L
1
= ∇ (Ḃ K Ȧ)
− ȦḂ
MAB...L ;
2k + 1 (A B...L)
hence,
∇(A (Ȧ KB...L) Ḃ) = −(2k + 1)C(ASȦḂ MB...L)S , (3.119)
where we have made use of the Ricci identities (2.127).
D(1, 0) Killing Spinors. Symmetry Operators
According to the definitions given above, a D(1, 0) Killing spinor MAB is a symmet-
ric two-index spinor field such that
∇A Ȧ MBC = 13 (εAB KC Ȧ + εAC KB Ȧ ), (3.120)
with KA Ȧ = ∇SȦ MSA [see (3.117) and (3.118)]. Then, making use of (2.117) and
(3.120), we obtain
∇BȦ K A Ȧ = ∇BȦ ∇S Ȧ MS A
= ∇S Ȧ ∇BȦ MS A + 2 BS
MS A
= 13 ∇S Ȧ (ε B S K AȦ + ε BA KS Ȧ ) + 2 BS
MS A
= 13 ∇B Ȧ K AȦ − 13 ε BA ∇SȦ KSȦ + 2 BS
MS A .
Thus, by virtue of (2.126),
3 ∇ K Ȧ
4 BȦ A
= − 13 ε BA ∇SȦ KSȦ + 2CBASC MSC − 13 RM BA .
Contracting the last equation with εBA , one finds that
∇AȦ KAȦ = 0, (3.121)
and therefore
3 ∇ K Ȧ = 2CBASC MSC − 13 RM BA .
4 BȦ A
(3.122)
On the other hand, from (3.119) and (3.121) it follows that KAȦ is the spinor equiv-
alent of a (possibly complex) Killing vector if and only if C(A SȦḂ MB)S = 0. Thus, in
particular, KAȦ corresponds to a possibly complex Killing vector if the space-time
metric is a type-D solution of the Einstein vacuum field equations or of the Einstein–
Maxwell equations with an algebraically general electromagnetic field such that
each principal spinor of fAB satisfies (3.52), since, as shown above, in this case MAB
is proportional to fAB .
Proposition 3.7. Let MAB be a D(1, 0) Killing spinor and KAȦ ≡ ∇B Ȧ MBA . Then for
an arbitrary spinor field ΦA ,
∇CȦ (M AB ∇BȦ − 23 K A Ȧ )ΦA = (MCB ∇BȦ − 13 KC Ȧ )∇AȦ ΦA . (3.123)
Proof. We begin by noticing that, making use of (1.27), (2.117), and (2.122),
M AB ∇CȦ ∇BȦ ΦA = MCB ∇AȦ ∇BȦ ΦA + ε AC M SB ∇S Ȧ ∇BȦ ΦA

= MCB ∇BȦ ∇AȦ ΦA + 2MCB B ΦA + M
A SB
SB Φ
C
= MCB ∇BȦ ∇AȦ ΦA − 6 RM ΦB + M C SB ΦR .

1 CB SB RC
Hence, with the aid of (3.120), (3.122), and (1.27), we obtain
∇CȦ (M AB ∇BȦ − 23 K A Ȧ )ΦA = M AB ∇CȦ ∇BȦ ΦA + (∇CȦ M AB )∇BȦ ΦA − 23 K A Ȧ ∇CȦ ΦA

− 23 (∇CȦ K A Ȧ )ΦA
= MCB ∇BȦ ∇AȦ ΦA − 16 RMCB ΦB + M SBCRC SB ΦR
+ 13 (ε CA K BȦ + ε CB K AȦ )∇BȦ ΦA − 23 K A Ȧ ∇CȦ ΦA
− CCASB MSB ΦA + 16 RMCA ΦA
= (MCB ∇BȦ − 13 KC Ȧ )∇AȦ ΦA .

Thus, if the spinor field ΦA satisfies the Weyl equation ∇AȦ ΦA = 0 and MAB is a
D(1, 0) Killing spinor, then
χȦ ≡ (M AB ∇BȦ − 23 K A Ȧ )ΦA
satisfies the Weyl equation ∇CȦ χȦ = 0 (see also Kamran and McLenaghan 1984a,b).
A similar result holds in the case of the source-free Maxwell equations; if fAB is a
solution of the source-free Maxwell equations ∇AȦ fAB = 0, then

fṘṠ = ∇A (Ṙ (M|AB MCD| ∇C Ṡ) + 23 M|AB KD|Ṡ) + 13 M|BD KA|Ṡ) ) f BD
satisfies the source-free Maxwell equations ∇RṘ fṘṠ = 0 (Torres del Castillo 1985).
In fact, fṘṠ = ∇A (Ṙ Φ|A|Ṡ) , with

ΦAṠ = MAB MCD ∇C Ṡ + 23 MAB KDṠ + 13 MBD KAṠ f BD ,
and making use of (3.120), (3.122), (2.64), and the result of Exercise 3.2, one finds
that ΦAṠ satisfies the self-duality condition (3.10).
Equation (3.123) can be written in a more symmetric form involving MAB and its
complex conjugate; indeed, making use of (3.123), we obtain
∇CȦ (M AB ∇BȦ − 23 K A Ȧ + MȦ Ḃ ∇A Ḃ )ΦA

= (MCB ∇BȦ − 13 KC Ȧ + MȦ Ḃ ∇C Ḃ )∇AȦ ΦA + (∇CȦ M Ȧ Ḃ )∇A Ḃ ΦA
= (MCB ∇BȦ − 13 KC Ȧ − KC Ȧ + M Ȧ Ḃ ∇C Ḃ )∇AȦ ΦA ,
with M ȦḂ ≡ MAB and K AḂ = ∇A Ṡ M ḂṠ . Hence, if
K AḂ = −KAḂ , (3.124)
we have
∇CȦ (M AB ∇BȦ − 23 K A Ȧ + M Ȧ Ḃ ∇A Ḃ )ΦA

= (MCB ∇BȦ − 23 KC Ȧ + MȦ Ḃ ∇C Ḃ )∇AȦ ΦA . (3.125)
Making use of (3.120) and its complex conjugate, one can verify that condition
(3.124) holds if and only if MAB εȦḂ + MȦḂ εAB is the spinor equivalent of a two-index
Killing–Yano tensor Yab , which is an antisymmetric two-index tensor satisfying
∇aYbc + ∇bYac = 0 (3.126)
(see (3.127) below). Note that if MAB is a D(1, 0) Killing spinor, then for any com-
plex constant λ , λ MAB is also a D(1, 0) Killing spinor, but when KAȦ is different
from zero, MAB and λ MAB satisfy condition (3.124) if and only if λ is real. On the
other hand, if MAB is a D(1, 0) Killing spinor such that KAȦ is equal to zero, then
for any complex constant λ , λ MAB εȦḂ + λ M ȦḂ εAB will be the spinor equivalent of a
two-index Killing–Yano tensor, which implies (taking λ real or pure imaginary) that
the tensor equivalent of MAB εȦḂ + M ȦḂ εAB and its dual are two linearly independent
Killing–Yano tensors. (Cf. Taxiarchis 1985 and the references cited therein.)
By virtue of (3.120), (3.124), (1.27), and (1.64), the spinor equivalent of ∇aYbc is
∇AȦ (MBC εḂĊ + M ḂĊ εBC )

= 13 (εAB KCȦ + εAC KBȦ )εḂĊ − εBC (εȦḂ KAĊ + εȦĊ KAḂ )

= 13 (2εAC KBȦ + εBC εA R KRȦ )εḂĊ − εBC (2εȦĊ KAḂ + εḂĊ εȦ Ṙ KAṘ )
= 23 (εAC εBD εȦḊ εḂĊ − εAD εBC εȦĊ εḂḊ )K DḊ
= − 23 i εAȦBḂCĊDḊ K DḊ ,
that is,
∇aYbc = 23 i εabcd K d , (3.127)
where Ka is the vector equivalent of KAȦ , which implies the validity of (3.126). (Note
that condition (3.124) means that Ka is pure imaginary, and therefore the right-hand
side of (3.127) is real.)
If we now consider a pair of spinor fields ψA , φȦ satisfying the Dirac equation
√ √
2 h̄ ∇AȦ ψ A = m0 c φȦ , 2 h̄ ∇AȦ φ Ȧ = m0 c ψA , the identity (3.125) and its complex
conjugate imply that the spinor fields
Ḃ
A ≡ −(M Ȧ ∇AḂ − 23 K A Ȧ + MA B ∇B Ȧ )φȦ ,
ψ
(3.128)
φȦ ≡ (M AB ∇BȦ − 23 K A Ȧ + MȦ Ḃ ∇A Ḃ )ψA ,
also satisfy the Dirac equation (see also Exercise 3.13).

In the case of a type-D space-time that admits a D(1, 0) Killing spinor (e.g., any
type-D solution of the Einstein vacuum field equations), according to (3.114), with
respect to a frame such that only C1122 is different from zero, the components of the
vector field KAȦ , defined by (3.120), can be explicitly expressed in the form
K11̇ = −3Γ1121̇M12 , K12̇ = −3Γ1122̇M12 ,

(3.129)
K21̇ = −3Γ2211̇M12 , K22̇ = −3Γ2212̇M12 .
Hence, with respect to such a frame, expressions (3.128) can be written explicitly in
the form
ψ
1 = 2M 1̇2̇ (∂12̇ − Γ1̇2̇2̇1 + Γ2̇2̇1̇1 )φ1̇ + (M12 + M1̇2̇ )∇1 Ȧ φȦ ,
ψ
2 = 2M 1̇2̇ (∂21̇ + Γ1̇2̇1̇2 − Γ1̇1̇2̇2 )φ2̇ − (M12 + M1̇2̇ )∇2 Ȧ φȦ ,
(3.130)
φ1̇ = −2M12(∂21̇ − Γ1221̇ + Γ2211̇)ψ1 − (M12 + M1̇2̇ )∇A 1̇ ψA ,
φ2̇ = −2M12(∂12̇ + Γ1212̇ − Γ1122̇)ψ2 + (M12 + M1̇2̇ )∇A 2̇ ψA .
In the case of the Schwarzschild metric, (3.115) and (2.112) imply that M12 is
proportional to r, and making use of (2.109) and (3.129) one finds that (3.124) is
satisfied if we take, for instance,
ir
M12 = √ .
2
Then, making use of (2.108), equations (3.130) reduce to

∂ i ∂ 1
ψ1 = −i + + cot θ φ1̇ ,
∂ θ sin θ ∂ ϕ 2

∂ i ∂ 1
ψ2 = −i − + cot θ φ2̇ ,

(3.131)
∂ i ∂ 1
φ1̇ = −i − + cot θ ψ1 ,

∂ i ∂ 1
φ2̇ = −i + + cot θ ψ2 .
The eigenvalues of the operator that sends ψA , φȦ into ψ A , φȦ , given by (3.131),
are of the form ±( j + 12 ), where j is a half-integer [see (3.37)]. It may be noticed
that this operator does not involve the parameter rg , contained in the Schwarzschild
metric; therefore, also in the Minkowski space-time this operator maps any solution
of the Dirac equation into another solution. In the flat space-time limit, this operator
is essentially that given in the standard notation as β (σ · L + h̄) (see, e.g., Messiah
1962, Davydov 1988).
Exercises
3.1. Show that if tAȦ is the spinor equivalent of a future-pointing timelike vector,
then α , β ≡ t AȦ βA α Ȧ is a positive definite Hermitian inner product for the undot-
ted one-index spinors.
3.2. Show that if fAB satisfies the source-free Maxwell equations, then ∇CĊ ∇CĊ fAB
= 2CABCD f CD − 13 R fAB .
3.3. Show that at each point where the electromagnetic field is algebraically general,
the minimum value of Tabt at b , with the constraint t ata = −1, is
1
[(F ab Fab )2 + (∗ F ab Fab )2 ]1/2 .
16π
(Physically, Tabt at b is the energy density of the electromagnetic field measured by
an observer with four-velocity t a .)
3.4. Let Fab be a bivector. Prove that there exists a nonzero vector l a such that
Fab l b = 0 = ∗ Fab l b if and only if Fab is algebraically special.
3.5. Let Tab be the energy–momentum tensor of the electromagnetic field. Prove that
there exists a nonzero vector l a such that Tab l a = 0 if and only if Tab corresponds to
an algebraically special electromagnetic field Fab and l a points along the repeated
principal null direction of Fab .
3.6. A set of equations for a zero-rest-mass spin-3/2 field is given by
∇AȦ ψ A BḂ = ∇BḂ ψ A AȦ .
Show that these equations can be also expressed in the form HAḂĊ = 0, HABC =
H(ABC) , where
HAḂĊ ≡ ∇D (Ḃ ψ|AD|Ċ) , HABC ≡ ∇(B Ḋ ψ|A|C)Ḋ ,
and that if the Ricci tensor vanishes, HABC satisfies
∇A Ȧ HABC = CABC D ψ A DȦ
[cf. (3.41)]. Show that if the Ricci tensor is equal to zero, there are no algebraic
consistency conditions of the Buchdahl–Plebański type for these equations.
3.7. Using the spinor formalism, show that if the spinor equivalent of a real, trace-
less, totally symmetric four-index tensor Tabcd satisfies the condition (3.51), then
Tabcd can be expressed as Tabcd = 14 (CarbsCc r d s + ∗Carbs ∗Cc r d s ) with
Tacgh φb g d h − Tbcgh φa g d h + Tbdgh φa g c h − Tadgh φb g c h

Cabcd = 2 ,
[T abcd (φarbs φc r d s + ∗ φarbs ∗ φc r d s )]1/2
where φabcd is any real four-index tensor with the symmetries of the Weyl tensor
such that T abcd (φarbs φc r d s + ∗ φarbs ∗ φc r d s ) is different from zero [cf. (3.28)].
3.8. Assuming that the Einstein vacuum field equations are satisfied (i.e., Rab = 0)
and that K is a Killing vector, show that

∇[a 2 f ∇b Kc] + ω εbc]de ∇d K e = 0,
where f = −K a Ka and ω is the twist potential of K; that is, show that

∇AȦ − iχ LAB εȦḂ + iχ LȦḂ εAB = 0,
where χ is the Ernst potential (3.98). This implies the local existence of a vector
field βa such that
2 f ∇[a Kb] + ω εabcd ∇c K d = −∇[a βb] .
Similarly, show that

∇[a εbc]de ∇d K e = 0,
that is, ∇AȦ (LAB εȦḂ + LȦḂ εAB ) = 0, which implies the local existence of a vector
field αa such that
εabcd ∇c K d = 2∇[a αb] .
The vector fields αa and βa can be used to construct a one-parameter family of
solutions of the Einstein vacuum field equations (for details see Geroch 1971).
3.9. Show that the spinor field (3.103) satisfies K(BĊ ∇A |Ċ fA|C) = 0, for any differen-
tiable function Φ .
3.10. Show that if MAB is a D(1, 0) Killing spinor and Qab is the tensor equivalent
of MAB εȦḂ + M ȦḂ εAB , then
∇a Qbc = − 13 εabcd ∇s ∗ Qsd + 13 gab ∇s Qsc − 13 gac ∇s Qsb ,
and therefore ∇(a Qb)c = 13 gab ∇s Qsc − 13 gc(a ∇s Q|s|b) . Show that Qab is a Killing–
Yano tensor if ∇a Qab = 0.
3.11. Show that if ΦAB...L is a totally symmetric spinor field that satisfies the spin-s
zero-rest-mass field equations and MAB...D is a D(k, 0) Killing spinor, with s > k,
then ΦAB...D...L M AB...D also satisfies the zero-rest-mass field equations.
3.12. Show that if MA is a D(1/2, 0) Killing spinor, then
∇AȦ KḂ = − 12
1
R εȦḂ MA − 2CD AȦḂ MD ,
with KḂ ≡ ∇A Ḃ MA . Show that if ΦȦ satisfies the Weyl equation, then
fȦḂ ≡ MC ∇CȦ ΦḂ − Φ(Ȧ ∇C Ḃ) MC
satisfies the source-free Maxwell equations.

3.13. Show that the relation between the bispinors

ψA ψA
Ψ= and
Ψ= ,
φ Ȧ φȦ
given by (3.128), can be expressed in the form
= √1 (Y ab γ5 γa ∇bΨ + 2 K a γaΨ ).
Ψ 3
2
3.14. Show that if Yab is a Killing–Yano tensor and va is the velocity vector of an
affinely parameterized geodesic, then YabYc b va vc is constant along the geodesic (this
amounts to saying that Kac ≡ YabYc b is a two-index Killing tensor, ∇(a Kbc) = 0).
Show that Yab vb is covariantly constant along the geodesic.
3.15. Show that if K is a Killing vector, the Lie derivative of a bispinor Ψ with
respect to K is given by
£KΨ = K a ∇aΨ + 14 (∇a Kb ) γ a γ bΨ
with
ψA
Ψ= .
φ Ȧ
Chapter 4
Further Applications
In this chapter some further examples of the applications of the two-component

spinor formalism are given. In contrast to the preceding chapter, manifolds with Eu-
clidean or ultrahyperbolic signature are also considered here. In Section 4.1 the self-
dual Yang–Mills fields in a possibly curved manifold are considered. In Section 4.2,
a reduced expression for the metric of a Riemannian manifold with one algebraically
special Weyl spinor (CABCD or CȦḂĊḊ ) is derived; this class of manifolds includes
the H spaces, which were originally studied by Newman (see, e.g., Newman 1976,
Hansen et al. 1978) and Penrose (see, e.g., Penrose 1976, Penrose and Ward 1980,
Ward and Wells 1990). Section 4.3 deals with Killing bispinors and eigenbispinors
of the Dirac operator, which can be defined in Riemannian manifolds of any dimen-
sion or signature. The discussion is restricted to four-dimensional manifolds, which
allows one to obtain stronger results than those derived by means of the formal-
ism based on Clifford algebras, applicable to manifolds of arbitrary dimension. The
sections of this chapter are independent of each other.
c
With the only exception of the null tetrad ∂ AḂ introduced in Section 4.2, the null
tetrads employed in this chapter satisfy the condition (2.100) appropriate for the
signature of the metric.
4.1 Self-Dual Yang–Mills Fields
A Yang–Mills field is locally represented by a one-index vector field (or one-form)

Aa with values in the Lie algebra g of some Lie group G. (More precisely, Aa is the
local expression of a connection in a principal fiber bundle with structure group G
(see, e.g., Ward and Wells 1990).) The field strength Fab is then given by
Fab = ∇a Ab − ∇b Aa + [Aa , Ab ] (4.1)
and satisfies the identities [which follow from (4.1)]

Physics 59, DOI 10.1007/978-0-8176-4984-5_4,
150 4 Further Applications
∇a ∗ F ab + [Aa, ∗ F ab ] = 0, (4.2)
where ∗ Fab denotes the dual of Fab (∗ Fab = 12 εabcd F cd ). The Yang–Mills equations
are
∇a F ab + [Aa, F ab ] = 0. (4.3)
The Yang–Mills equations are form-invariant under the gauge transformations
Aa → U −1 AaU + U −1∂aU, (4.4)
where U is a function with values in the (gauge) group G. Under this transformation,
the field strength transforms according to
Fab → U −1 FabU.
The field strength Fab vanishes if and only if there exists locally a function U with
values in G such that Aa = U −1 ∂aU.
By virtue of the identities (4.2), the Yang–Mills equations are trivially satisfied if
∗ F = λ F , for some constant factor λ . Since ∗ (∗ F ) = (−1)q F [see (1.66)], it
ab ab ab ab
follows that λ 2 = (−1)q ; that is, ∗ Fab is proportional to Fab only if the field is self-
dual (∗ Fab = iq Fab ) or anti-self-dual (∗ Fab = −(iq )Fab ). However, when the metric
has Lorentzian signature, a self-dual or anti-self-dual Yang–Mills field would satisfy
∗ F = ±iF , which cannot be fulfilled if A takes values in the (real) Lie algebra of
ab ab a
G; but when the metric of M has Euclidean signature it is possible to have nontrivial
self-dual or anti-self-dual Yang–Mills fields, because in that case, the field strength
would satisfy the condition ∗ Fab = ±Fab.
For instance, in the case of a self-dual Yang–Mills field we have FA Ṙ BṘ = 0,
where FAȦBḂ is the spinor equivalent of Fab [see (1.50) and (1.65)]; hence, making
use of (4.1), one obtains the condition
∇(A Ṙ AB)Ṙ + A(AṘ AB)Ṙ = 0, (4.5)
which involves only first derivatives of ABḂ [cf. (3.10)]. As in the case of the abelian
version of the self-duality conditions (4.5), given by (3.10), equations (4.5) can be
partially integrated, reducing them to a single equation for a potential, if the curva-
ture is equal to zero or if there exists a principal spinor lA of the Weyl spinor CABCD
satisfying the condition (3.52),
l A l B ∇AĊ lB = 0.
As shown in Section 3.3.1, in a manifold with Lorentzian signature, a one-index

spinor satisfying this condition defines a shear-free congruence of null geodesics
(lA l Ȧ is the spinor equivalent of a vector field tangent to the congruence). However,
when the signature of the metric is Euclidean, we cannot define a vector field with lA
alone. Nevertheless, the consequences of the existence of such a spinor field derived
in Section 3.3.1 remain valid (e.g., Propositions 3.2–3.5).
4.1 Self-Dual Yang–Mills Fields 151
The proof is almost identical to that given for Proposition 3.6. With respect to a
null tetrad such that lA = δA2 , we have Γ111Ȧ = 0, and from (4.5), with A = B = 1, we
obtain
0 = (∂1 Ṙ − Γ121Ṙ − Γ ṘṠ Ṡ1 )A1Ṙ + A1 Ṙ A1Ṙ (4.6)
[cf. (3.78)]. This last equation means that the components A1Ṙ of the gauge field can
be eliminated locally by a gauge transformation (4.4). That is, equation (4.6) is the
integrability condition for the local existence of a function U with values in G such
that A1Ṙ = U −1 ∂1ṘU. In fact, applying ∂1 Ṙ to both sides of this last equation, we
have
∂1 Ṙ A1Ṙ = (∂1 ṘU −1 )∂1ṘU + U −1∂1 Ṙ ∂1ṘU

= −U −1 (∂1 ṘU)U −1 ∂1ṘU + U −1(−Γ S 11 Ṙ ∂SṘU − Γ Ṡ Ṙ Ṙ 1 ∂1ṠU)
= −A1 Ṙ A1Ṙ + Γ121Ṙ A1Ṙ + Γ ṠṘ Ṙ1 A1Ṡ ,
which coincides with (4.6).

Thus, with A1Ṙ being equal to zero, taking A = 1, B = 2 in (4.5) we obtain the
condition
0 = (∂1 Ṙ + Γ121Ṙ + Γ112Ṙ − Γ ṘṠ Ṡ1 )A2Ṙ ,
which has the form of (3.79); therefore, there exists locally a matrix-valued function
χ such that
A2Ḃ = (∂1Ḃ + 2Γ121Ḃ + Γ112Ḃ)χ (4.7)
[see (3.80)].
Finally, taking A = B = 2 in (4.5) and substituting the expressions A1Ṙ = 0 and
that for A2Ṙ given by (4.7), one obtains the single nonlinear condition
0 = (∂2 Ṙ + Γ122Ṙ − Γ ṘṠ Ṡ2 )A2Ṙ + A2Ṙ A2Ṙ ,

= (∂2 Ṙ + Γ122Ṙ − Γ ṘṠ Ṡ2 )(∂1Ṙ + 2Γ121Ṙ + Γ112Ṙ)χ

+ (∂1 Ṙ + 2Γ121Ṙ + Γ112Ṙ )χ (∂1Ṙ + 2Γ121Ṙ + Γ112Ṙ)χ . (4.8)
√
When M is flat, we can make use of the null tetrad ∂AȦ = (1/ 2) σ a AȦ ∂ /∂ xa
induced by Cartesian coordinates xa for which all the spin coefficients vanish. Then,
equation (4.8) reduces to

∂ ∂χ ∂ ∂χ ∂χ ∂χ
0= − + , , (4.9)
∂ x21̇ ∂ x12̇ ∂ x22̇ ∂ x11̇ ∂ x11̇ ∂ x12̇
√
where xAȦ ≡ (1/ 2) σa AȦ xa [see (2.149)]. As pointed out above, if the metric of
M has Lorentzian signature, there are no nontrivial self-dual Yang–Mills fields if
Aa takes values in a real Lie algebra. However, one can have nontrivial self-dual
fields by considering, for instance, GL(n, C) as the gauge group G. In that case,
(4.9) corresponds to the equation for Newman’s K-matrix (Newman 1978, Mason
and Woodhouse 1996).
4.2 H and H H Spaces
As shown in Section 3.3.1, in a manifold with Lorentzian signature, a one-index

spinor field lA satisfying the condition
l A l B ∇AĊ lB = 0 (4.10)
defines a shear-free congruence of null geodesics. According to the Frobenius the-

orem, these equations are also the conditions for the complete integrability of the
system of differential equations
lA θ AḂ = 0. (4.11)
That is, equations (4.10) are necessary and sufficient for the local existence of two
(possibly complex-valued) functions q1̇ , q2̇ such that
lA θ AḂ = LḂĊ dqĊ , (4.12)
where the LḂĊ are complex-valued functions whose values at each point form a
nonsingular 2 × 2 matrix (see also Flaherty 1980). In fact, the system (4.11) is com-
pletely integrable if and only if
d(lA θ AḂ ) ∧ lB θ B1̇ ∧ lC θ C2̇ = 0.
Making use of (2.90), the Cartan first structure equations (2.15), (2.55), and Exercise
2.5, one can readily verify that this last condition amounts to (4.10).
For example, in the case of the Schwarzschild metric (2.106), the spin coefficients
for the null tetrad (2.108) satisfy Γ111Ȧ = 0 [see (2.109)], and therefore, the one-
index spinor field lA = δA2 satisfies the condition (4.10). Making use of the explicit
expressions (2.107), we find that the system of differential equations lA θ AḂ = 0 is
equivalent to
dθ + i sin θ dϕ = 0,
rg −1
cdt + 1 − dr = 0.
r
This system is, indeed, completely integrable (actually, in this particular case, each
of these one-forms is integrable); one can readily see that

dθ + i sin θ dϕ = 2e−iϕ cos2 21 θ d eiϕ tan 12 θ ,
rg −1
cdt + 1 − dr = d ct + r + rg ln |r − rg | .
r
Thus, the functions qȦ appearing in (4.12) can be taken as eiϕ tan 12 θ and ct + r +
rg ln |r − rg |, or any pair of functionally independent functions of them (see below).
4.2 H and H H Spaces 153
Assuming that lA is a multiple principal spinor of CABCD , with the aid of the
c
functions qȦ we construct a second null tetrad, θ AḂ , starting with
c
θ 2Ȧ ≡ −φ −2 dqȦ , (4.13)
where φ is defined by
l B ∇AȦ lB = lA l B ∂BȦ ln φ (4.14)
[see (3.73)]. Thus, from (4.12) we obtain
c
θ 2Ȧ = −φ −2 (L−1 )Ȧ Ḃ lA θ AḂ , (4.15)
where (L−1 )Ȧ Ḃ are the entries of the inverse of the matrix (LȦ Ḃ ). Then, the one-
c
forms θ 1Ȧ must be given by
c c
θ 1Ȧ = φ 2 LḂ Ȧ mC θ CḂ + η θ 2Ȧ , (4.16)
where mA is a one-index spinor field such that mA lA = 1 and η is an arbitrary func-

c c
tion, so that −εAB εȦḂ θ AȦ θ BḂ = −εAB εȦḂ θ AȦ θ BḂ [see (2.99)]. (Note that the spinor
field mA is not uniquely defined by the condition mA lA = 1; the function η is related
c
to this ambiguity.) Thus, the vector fields ∂ AḂ are
c
∂ 1Ȧ = −φ −2 (L−1 )Ȧ Ḃ lC ∂CḂ ,
c c
(4.17)
∂ 2Ȧ = −φ 2 LĊ Ȧ mD ∂DĊ − η ∂ 1Ȧ .
A somewhat lengthy computation, making use of (1.106), (2.23), (4.10), the fact
that ddqĊ = d[(L−1 )Ċ Ḃ lA θ AḂ ] = 0, and the Cartan first structure equations (2.15),
yields
c c c c
[∂ 11̇ , ∂ 12̇ ] = ∂ 1 Ṙ ∂ 1Ṙ
−1
= φ −4 det(LṘ Ṡ ) −2lC (∂C Ṙ ln φ )l S ∂SṘ + 32 (∇RḊ lR )l S ∂SḊ

− 32 lR (∇RḊ l S )∂SḊ + 12 l R (∇SḊ lR )∂SḊ
−1 C Ṙ
= −2φ −4 det(LṘ Ṡ ) l ∂C ln φ − mC l B ∇C Ṙ lB l S ∂SṘ ,
which is equal to zero as a consequence of (4.14). We can now introduce a second

pair of (possibly complex-valued) functions, pȦ , defined by
c
∂ 1Ȧ p = δȦ .
Ḃ Ḃ
(4.18)
The integrability conditions for these equations are trivially satisfied since the vector
c
fields ∂ 1Ȧ commute. Then, the differential of pȦ is [see (2.90)]
c c c c c
dpȦ = −(∂ CḂ pȦ ) θ CḂ = − θ 1Ȧ − (∂ 2Ḃ pȦ ) θ 2Ḃ ,
and using (4.13) we obtain

c c
θ 1Ȧ = −dpȦ + φ −2 (∂ 2Ḃ pȦ ) dqḂ . (4.19)
On the other hand, from (4.17) we have

c
∂ 2Ȧ p = −φ L Ȧ m ∂DĊ p − 2η .
Ȧ 2 Ċ D Ȧ
c
Therefore, choosing η = − 12 φ 2 LĊ Ȧ mD ∂DĊ pȦ , we obtain ∂ 2Ȧ pȦ = 0, which means
that
c
QȦḂ ≡ −φ −2 ∂ 2 Ȧ pḂ (4.20)
is symmetric on the indices Ȧ, Ḃ. Then, (4.19) gives
c
θ 1Ȧ = −dpȦ − QȦ Ḃ dqḂ . (4.21)
In this manner, we have proved the validity of the following (Plebański and Robinson
1976, Finley and Plebański 1976, Torres del Castillo 1983) result.
Proposition 4.1. The existence of a repeated principal spinor of the Weyl spinor
CABCD that satisfies (4.10) implies the local existence of (possibly complex) coor-
dinates qȦ , pȦ such that the metric of the manifold can be expressed in the form
g = 2φ −2 dqȦ (dpȦ + QȦḂ dqḂ ). (4.22)

It may be remarked that in the preceding derivation no assumption is made con-
cerning the signature of the metric (see Example 4.1 below). The manifolds de-
scribed in this last proposition are called H H spaces.
In the case of the Schwarzschild metric, already considered at the beginning of
this section, it is convenient to choose the coordinates qȦ in such a way that
rg −1
dq1̇ = csc θ dθ + idϕ , dq2̇ = cdt + 1 − dr. (4.23)
r
Then, by comparing (2.107) with (4.12), taking lA = δA2 as above, we find that
√
(LȦ Ḃ ) = diag (−r sin θ / 2, 1).
The conformal factor φ is determined by Γ112Ȧ = ∂1Ȧ ln φ (i.e., ρ = D ln φ , τ = δ ln φ )

[see (3.74)]; hence, we can take φ = r−1 [see (2.109) and (2.108)]. Then, ac-
c c c
cording to (4.17), the vector fields ∂ 1Ȧ of the new tetrad are ∂ 11̇ = −r2 ∂11̇ , ∂ 12̇ =
√
2 r csc θ ∂12̇ , and one can readily verify that a possible choice for the coordinates
pȦ is
p1̇ = r−1 , p2̇ = − cos θ . (4.24)
c
Taking mA = δ2A , the vector fields ∂ 2Ȧ are given by [see (4.17)]
c sin θ c c 1 c
∂ 21̇ = √ ∂21̇ − η ∂ 11̇ , ∂ 22̇ = − 2
∂22̇ − η ∂ 12̇ .
2r r
Then, making use of the expressions derived above, one finds that η = 0 and
r − rg
Q1̇1̇ = − 12 sin2 θ , Q1̇2̇ = 0, Q2̇2̇ = − . (4.25)
2r3
It may be pointed out that if we replace the relations (4.23) by
rg −1
dq1̇ = csc θ dθ + cdt, dq2̇ = idϕ + 1 − dr,
r
maintaining (4.24), (4.25), and φ = r−1 , then (4.22) gives another type-D solution
of the Einstein vacuum field equations (with Lorentzian signature, assuming that
r, θ , ϕ , and t are real), known as the Ehlers–Kundt B1 metric (Flaherty 1980).
Example 4.1. The standard metric of S4 in complex coordinates.

An explicit example of a metric with Euclidean signature that satisfies the condi-
tions of Proposition 4.1 and therefore can be written in the form (4.22) is given by
the standard metric of S4 (see Example 2.1, pp. 84–86). As shown in Section 2.2,
the conformal curvature of this metric vanishes, CABCD = 0, CȦḂĊḊ = 0, and the spin
coefficients for the null tetrad (2.102) satisfy Γ111Ȧ = 0; hence, the one-index spinor
lA = δA2 satisfies (4.10) and, trivially, l A l B lCCABCD = 0. The one-forms lA θ AḂ = θ 2Ḃ
must form an integrable system; in fact, θ 2Ȧ = LȦ Ḃ dqḂ , with
dq1̇ = csc θ dθ + idϕ , dq2̇ = csc ψ dψ + id χ ,
and √ √
(LȦ Ḃ ) = diag(− sin ψ sin χ sin θ / 2, sin ψ / 2).
The conformal factor φ is determined by the conditions Γ112Ȧ = ∂1Ȧ ln φ . Hence, one
finds that φ can be taken as
φ = csc ψ csc χ ,
and from (4.17) we obtain

c ∂ i ∂
∂ 11̇ = − sin ψ sin χ + ,
2
∂ ψ sin ψ ∂ χ

c 1 ∂ i ∂
∂ 12̇ = + .
sin θ ∂ θ sin θ ∂ ϕ
By inspection, one readily finds that the coordinates pȦ can be chosen as
p1̇ = −i cot χ , p2̇ = − cos θ .
Then, one finally finds that the functions QȦḂ are given by
Q1̇1̇ = − 12 sin2 θ , Q1̇2̇ = 0, Q2̇2̇ = 12 csc2 χ ,
which are second-degree polynomials in the pȦ .
Going back to the general case, from (4.13) and (4.21) it follows that the null
tetrad ∂ AḂ , in terms of the coordinates qȦ , pȦ , is given by
c ∂
∂ 1Ȧ = ,
∂ p
Ȧ
(4.26)
c ∂ ∂
∂ 2Ȧ = φ2 − QȦ Ḃ .
∂q Ȧ ∂ pḂ
Note that this tetrad need not satisfy one of the conditions (2.100). In fact, equations
c
(4.17) can be written in the form ∂ AḂ = KAC KḂ Ḋ ∂CḊ , with
K1C = −φ −2 [det(LṘ Ṡ )]−1/2 lC ,

K2C = φ 2 [det(LṘ Ṡ )]1/2 mC + ηφ −2 [det(LṘ Ṡ )]−1/2 lC ,
and
KȦ Ḃ = [det(LṘ Ṡ )]1/2 (L−1 )Ȧ Ḃ .
The matrices (K A B ) and (K Ȧ Ḃ ) belong to SL(2, C), but they do not have to satisfy
one of the relations (1.185).
The spin coefficients of the metric (4.22) with respect to the null tetrad (4.26) can
be readily obtained by making use of (2.97) and (2.98). Since
c c c
|2|Ḃ)
S ȦḂ = θ 1(Ȧ
∧θ = φ −2 (dp(Ȧ − QĊ(Ȧ dqĊ ) ∧ dqḂ)
[see (2.57)], we have

c c
d S ȦḂ = −2φ −1 dφ ∧ S ȦḂ − φ −2 dQĊ(Ȧ ∧ dqĊ ∧ dqḂ)
c c c c c
= 2φ −1 (∂ RĊ φ ) θ RĊ ∧ S ȦḂ + φ −2 (∂ RḊ QĊ(Ȧ ) θ |RḊ| ∧ dqĊ ∧ dqḂ)
c c c c c c
= 2φ −1 (∂ RĊ φ ) θ RĊ ∧ S ȦḂ + φ −2 (∂ RḊ QĊ(Ȧ ) θ |RḊ| ∧ εĊ Ḃ) φ 4 S 22 ,
c c c
where we have made use of the fact that S 22 =θ 21̇ ∧ θ 22̇ = φ −4 dq1̇ ∧ dq2̇ . The
exterior products appearing in the last equation are expressed in terms of the three-
c
forms θ̆ AḂ with the aid of (2.71), and we obtain
c c c c
d S ȦḂ = − 2φ −1 (∂ R (Ȧ φ )ε Ḃ) Ḋ − φ 2 (∂ 1Ḋ QȦḂ )ε 2 R θ̆ RḊ . (4.27)
Then, making use of (2.98), we find that

c c c
Γ ȦḂĊD = φ −1 ∂ D(Ȧ φ εḂ)Ċ − φ 2 (∂ 1(Ȧ QḂ)Ċ )ε 2 D , (4.28)
and hence
c
IΓȦḂ = φ −1 ∂(Ȧ φ (dpḂ) + QḂ)Ċ dqĊ ) + φ −1 D(Ȧ φ dqḂ) − ∂(ȦQḂ)Ċ dqĊ , (4.29)
where we have made use of the definitions

∂
∂Ȧ ≡ ,
∂ pȦ
(4.30)
∂ Ḃ ∂
DȦ ≡ − QȦ .
∂ qȦ ∂ pḂ
c c
(Note that ∂ Ȧ = −∂ /∂ pȦ .) In a similar manner, computing separately d S 11 , d S 12 ,
c
and d S 22 , we obtain
c
IΓ11 = φ −3 ∂Ȧ φ dqȦ ,
c
IΓ12 = − 32 φ −1 ∂Ȧ φ (dpȦ + QȦ Ḃ dqḂ ) − 12 (φ −1 DȦ φ + ∂ Ḃ QḂȦ ) dqȦ , (4.31)
c
IΓ22 = −φ DȦ φ (dpȦ + QȦḂ dqḂ ) − φ 2 DḂ QḂȦ dqȦ .
The curvature spinors can now be calculated, e.g., with the aid of (2.91)–(2.94); the
Weyl spinors are
c
CȦḂĊḊ = −φ 2 ∂(Ȧ ∂Ḃ QĊḊ) , (4.32)
and
c
C1111 = 0,
c
C1112 = 0,
c
C1122 = − 16 φ 2 ∂Ȧ ∂Ḃ QȦḂ , (4.33)
c
C1222 = − 12 φ 4 DȦ ∂Ḃ QȦḂ ,
c
C2222 = −φ 6 DȦ DḂ QȦḂ − (DḂ QȦḂ )∂ Ċ QȦĊ ,
the spinor equivalent of the traceless part of the Ricci tensor

c
C11ȦḂ = φ −1 ∂Ȧ ∂Ḃ φ ,
c
C12ȦḂ = − 12 φ 2 ∂ Ċ ∂(Ȧ QḂ)Ċ + φ ∂(Ȧ DḂ) φ , (4.34)
c
C22ȦḂ = −φ 5 ∂(Ȧ φ −1 DĊ QḂ)Ċ + φ 3 (DĊ φ )∂Ċ QȦḂ + φ 3 D(Ȧ DḂ) φ ,
and the scalar curvature
R = 2φ 2 ∂Ȧ ∂Ḃ QȦḂ + 12φ 3 ∂Ȧ (φ −2 DȦ φ ). (4.35)
Integration of Field Equations
Expression (4.22) allows us to partially integrate several sets of equations of ge-

ometric or physical interest. As an example of the reduction that can be achieved
by combining the two-component spinor formalism and the form of the metric
(4.22) (which is itself adapted to the spinor formalism) we shall give an alterna-
tive derivation of two results obtained in Section 3.3.1, showing that locally the
solution of the source-free Maxwell equations, or of the Weyl equation, can be ex-
pressed in terms of a single scalar function that satisfies a second-order linear partial
differential equation.
c c
The source-free Maxwell equations ∇AḂ f ȦḂ = 0 are equivalent to d( f ȦḂ S ȦḂ ) = 0
(see Exercise 2.5). Thus making use of (2.71), (2.90), and (4.27), we have
c c
φ 2 ∂ 1Ḃ (φ −2 f ȦḂ ) = 0 (4.36)
and
c c c c
φ 2 ∂ 2Ḃ (φ −2 f ȦḂ ) + φ 2 f ḂĊ ∂ 1 Ȧ QḂĊ = 0. (4.37)
c
Owing to (4.26), equation (4.36) is equivalent to ∂Ḃ (φ −2 f ȦḂ ) = 0. The solution of
this last equation can be obtained making use of the following proposition.
Proposition 4.2. Let MȦḂ be a set of functions satisfying the conditions MȦḂ = MḂȦ
and ∂Ȧ M ȦḂ = 0. Then there exists locally a function H such that MȦḂ = ∂Ȧ ∂Ḃ H.
Proof. Conditions ∂Ȧ M ȦḂ = 0 are equivalent to
∂ M1̇ Ḃ ∂ M2̇ Ḃ
= ,
∂ p2̇ ∂ p1̇
which imply the local existence of functions F Ḃ such that
∂ F Ḃ ∂ F Ḃ
M1̇ Ḃ = , M2̇ Ḃ = .
∂ p1̇ ∂ p2̇
Since M1̇2̇ = M2̇1̇ , we have ∂ F2̇ /∂ p1̇ = ∂ F1̇ /∂ p2̇ , which in turn implies the local
existence of a function H such that FȦ = ∂ H/∂ pȦ ; hence, MȦḂ = ∂ 2 H/∂ pȦ ∂ pḂ .

Thus, there exists locally a function H such that
c
f ȦḂ = φ 2 ∂Ȧ ∂Ḃ H. (4.38)
Substituting this expression into (4.37), making use of (4.26) and (4.30), we obtain
0 = DḂ ∂ Ȧ ∂ Ḃ H + (∂ Ḃ ∂ Ċ H)∂ Ȧ QḂĊ

∂ Ḃ
=∂ Ȧ
∂ H + QḂĊ ∂ ∂ H
Ḃ Ċ
∂ qḂ

= ∂ Ȧ DḂ ∂ Ḃ H ,
which means that F ≡ DḂ ∂ Ḃ H is a function of qȦ only. The function F can always
be written in the form F = ∂ hȦ /∂ qȦ , where the hȦ are also functions of qȦ only.
Letting H̃ ≡ H − hȦ pȦ , we find that [see (4.30)]
∂ hḂ
DḂ ∂ Ḃ H̃ = DḂ (∂ Ḃ H − hḂ) = F − DḂ hḂ = F − = 0.
∂ qḂ
c
On the other hand, f ȦḂ = φ 2 ∂Ȧ ∂Ḃ H = φ 2 ∂Ȧ ∂Ḃ H̃. Thus (dropping the tilde), we con-
clude that locally, any solution of the source-free Maxwell equations can be written
c
in the form f ȦḂ = φ 2 ∂Ȧ ∂Ḃ H, where H is a function such that DḂ ∂ Ḃ H = 0.
In order to make use of this result it is not necessary to express the metric of
the background space-time in the form (4.22), since with respect to an arbitrary null
tetrad, this solution is given by
fȦḂ = ∇B (Ȧ [φ −2 ∇S Ḃ) (φ 2 lS lB ψ )] (4.39)
(Torres del Castillo 1984b), where ψ is some function [cf. (3.83)].

With the aid of (4.28), one finds that with respect to the null tetrad (4.26), the
Weyl equation ∇AḂ η Ḃ = 0 is given by
c c
φ 3/2 ∂ 1Ḃ (φ −3/2 η Ḃ ) = 0,
c c c c
(4.40)
φ 3/2 ∂ 2Ḃ (φ −3/2 η Ḃ ) + 12 φ 2 (∂ 1 Ċ QĊḂ ) η Ḃ
= 0.
The first of these equations is locally equivalent to the existence of a function H

c
such that ηḂ = φ 3/2 ∂Ḃ H, and substituting into the second equation one finds that the
function H has to satisfy the equation
DḂ ∂ Ḃ H + 12 (∂ Ċ QĊḂ )∂ Ḃ H = 0.
With respect to an arbitrary null tetrad we have
ηḂ = φ −1 ∇C Ḃ (φ lC ψ )
[cf. (3.88)].
Left-Flat Spaces
The so-called H spaces are “half-flat,” which means that the curvature two-forms
RAB (or RȦḂ ) vanish. In the same form as the vanishing of the curvature two-forms
Ωab implies the local existence of a tetrad for which the connection one-forms van-
ish [see (2.51)], the vanishing of the two-forms RAB , for instance, is locally equiv-
alent to the existence of a null tetrad such that IΓAB = 0 [see (2.68)]; therefore,
according to Proposition 4.1, the metric of such a manifold can be expressed in the
form
g = 2dqȦ(dpȦ + QȦḂ dqḂ ) (4.41)
in terms of some local coordinates qȦ , pȦ [since ΓABCḊ = 0, φ can be taken equal
to 1; see (4.14)]. As we shall show below, the coordinates qȦ , pȦ can be chosen
c
in such a way that IΓAB = 0. We start by noticing that ∂Ȧ ∂ Ċ QḂĊ = ∂(Ȧ ∂ Ċ QḂ)Ċ +
c c
2 εȦḂ ∂ ∂ QṘĊ
= 0, as a consequence of C12ȦḂ = 0 and C1122 = 0, taking into account
1 Ṙ Ċ
that φ = 1 [see (4.34) and (4.33)]. Hence, ∂ Ċ QḂĊ are functions of qṘ only.
c
On the other hand, from C1222 = 0, we have ∂Ȧ DḂ QȦḂ = 0, which turns out to
be equivalent to DḂ ∂Ȧ QȦḂ = 0 [see (4.30)], or, since ∂Ȧ QȦḂ depend on the qṘ only,
∂ (∂Ȧ QȦḂ )/∂ qḂ = 0. This last equation is locally equivalent to the existence of a
function Λ (qṘ ) such that
∂ Ȧ QȦḂ = ∂Λ /∂ qḂ (4.42)
(cf. Proposition 4.2).
We can take advantage of the fact that the coordinates qȦ are not uniquely defined
by (4.12); in place of qȦ we can employ any pair of independent functions q Ṙ of the
c
qȦ , which induce a new null tetrad θ AḂ with [see (4.13)]
c
2Ȧ ∂ q Ȧ Ḃ c
θ = −φ −2 dq Ȧ = −φ −2 dq = T Ȧ Ḃ θ 2Ḃ ,
∂q Ḃ
where
∂ q Ȧ
T Ȧ Ḃ ≡ .
∂ qḂ
(There is also some freedom in the choice of the conformal factor φ [see(4.14)].
However, in the present case, we want to maintain the value φ = 1.) Hence,
c c

∂ 1Ȧ = −TȦ Ḃ ∂ 1Ḃ .
c
The new coordinates p Ȧ must satisfy the conditions ∂ 1Ȧ p Ḃ = δȦḂ [see (4.18)], and
therefore,
p Ȧ = −(T −1 )Ḃ Ȧ pḂ + σ Ȧ , (4.43)
where the σ Ȧ are functions of qḂ only.
Since the form of the metric (4.41) must be invariant under this change of coor-
dinates, the functions Q ȦḂ corresponding to the new coordinates q Ȧ , p Ȧ must be
given by
∂ p Ḃ)
Q ȦḂ = (T −1 )Ċ Ȧ (T −1 )Ḋ Ḃ QĊḊ − (T −1 )Ċ(Ȧ . (4.44)
∂ qĊ
c
Using the fact that ∂ 1Ȧ = ∂ /∂ p Ȧ ≡ ∂ Ȧ [see (4.26)] and combining the previous
results, one finds that

Ȧ −1 Ḋ ∂
∂ Q ȦḂ = (T ) Ḃ ∂ QĊḊ − Ḋ ln det(T Ṡ ) .
Ċ Ṙ
∂q
Hence, by means of a change of coordinates such that ln det(T Ṙ Ṡ ) = Λ , we get

∂ Ȧ Q ȦḂ = 0 [see (4.42)].
Now we note that ∂Ȧ DĊ QḂĊ = ∂(Ȧ DĊ QḂ)Ċ + 12 εȦḂ ∂ Ṙ DĊ QṘĊ = 0, as a conse-
c c
quence of C22ȦḂ = 0, C1222 = 0, and φ = 1. Hence, DĊ QḂĊ are also functions of qȦ
only. Assuming that the coordinates q Ȧ , p Ȧ have been chosen in such a way that
c
∂ Ȧ QȦḂ = 0, from the condition C2222 = 0 we have DḂ DĊ QḂĊ = 0, which is locally
equivalent to the existence of a function Ξ (qȦ ) such that
DĊ QḂĊ = ∂ Ξ /∂ qḂ . (4.45)
Under the coordinate transformation q Ȧ = qȦ , p Ȧ = pȦ + σ Ȧ [see (4.43) and

(4.44)], we have
∂
Q ȦḂ = QȦḂ − σḂ) .
∂ q(Ȧ
Then one finds that
∂ 1 ∂ σ Ȧ
DĊ Q ḂĊ = DĊ QḂĊ + .
∂ qḂ 2 ∂ qȦ
Therefore, by means of a change of coordinates q Ȧ = qȦ , p Ȧ = pȦ + σ Ȧ with
∂ σ Ȧ /∂ qȦ = −2Ξ , we get DĊ Q ḂĊ = 0, thus showing that there exist local coor-
dinates qȦ , pȦ such that the metric can be written in the form (4.41) with
∂ Ȧ QȦḂ = 0, DȦ QȦḂ = 0 (4.46)

c
(which amounts to IΓAB = 0 [see (4.31)]). According to Proposition 4.2, ∂ Ȧ QȦḂ = 0
implies the local existence of a function Θ such that
QȦḂ = ∂Ȧ ∂ḂΘ . (4.47)
Substituting (4.47) into the second set of equations in (4.46), we find that [see
(4.26)]

∂ ∂ ∂ Ȧ 1
0= − QȦĊ Q ȦḂ
=∂ Ḃ
∂ Θ− ∂ ∂ Θ ∂Ċ ∂ Θ .
Ċ Ȧ
(4.48)
∂q Ȧ ∂ pĊ ∂ qȦ 2 Ȧ
Equation (4.48) means that the expression between brackets is a function of qȦ only,
which can always be written in the form ∂ hȦ /∂ qȦ , where the hȦ are functions of the
qṘ only. Then, the substitution Θ → Θ + hȦ pȦ , which leaves QȦḂ unchanged [see
(4.47)], gives
∂ Ȧ 1
∂ Θ + ∂Ȧ ∂ḂΘ ∂ Ȧ ∂ ḂΘ = 0. (4.49)
∂ qȦ 2
This is the so-called second heavenly equation (Plebański 1975, Boyer et al. 1980),
and as we have shown, the metric of any left-flat space can be locally expressed in
the form
g = 2dqȦ (dpȦ + ∂Ȧ ∂ḂΘ dqḂ ), (4.50)
where Θ is a solution of (4.49).
According to (4.32), the only components of the curvature that can be different
from zero are given by CȦḂĊḊ = −∂Ȧ ∂Ḃ ∂Ċ ∂ḊΘ .
4.3 Killing Bispinors. The Dirac Operator
Apart from the definition of a Killing spinor given in Section 3.4 for a manifold with
Lorentzian signature, there exists another concept of a Killing spinor in the litera-
ture, which we shall call Killing bispinor (defined for a manifold of any signature).
For a four-dimensional manifold, a pair of one-index spinor fields (ψA , φȦ ) forms a
Killing bispinor with Killing number λ if
∇AȦ ψB = λ εAB φȦ , ∇AȦ φḂ = λ εȦḂ ψA , (4.51)
for some constant λ (cf. Baum 2000, Friedrich 2000). An analogous concept can be
defined in Riemannian manifolds of any dimension (Friedrich 2000, for the three-
dimensional case see also Torres del Castillo 2003). As in the case of the D(k, 0)
Killing spinors, the existence of a nontrivial solution of (4.51) imposes strong con-
ditions on the curvature of the manifold. In fact, combining the equations (4.51),
one obtains
∇C Ȧ ∇AȦ ψB = 2λ 2 εAB ψC ,
(4.52)
∇AĊ ∇AȦ ψB = −λ 2 εĊȦ ψB ,
and hence
CA ψB = λ 2 (εAB ψC + εCB ψA ),
ĊȦ ψB = 0,
4.3 Killing Bispinors. The Dirac Operator 163
and comparing with the Ricci identities (2.122) and (2.124), we see that CABCD = 0,
CABȦḂ = 0, and
R = −24λ 2. (4.53)
In a similar way one finds that CȦḂĊḊ = 0. Equation (4.53) shows that the value of
λ is determined up to sign by the scalar curvature and that λ must be real or pure
imaginary.
Equations (4.51) imply that if (ψA , φȦ ) is a Killing bispinor with Killing num-
ber λ , then (−ψA , φȦ ) is a Killing bispinor with Killing number −λ . Each Killing
bispinor gives rise to a (null, possibly complex) Killing vector KAȦ = ψA φȦ , since
∇AȦ (ψB φḂ ) = λ ψA ψB εȦḂ + λ φȦ φḂ εAB (4.54)
[cf. (2.140)]. Equations (4.52) also yield
∇AȦ ∇AȦ ψB = −2λ 2 ψB . (4.55)
In what follows we shall restrict ourselves to the case that the signature of the
metric is Euclidean. In this case equations (4.51) are equivalent to
B = −λ εAB φȦ ,
∇AȦ ψ ∇AȦ φḂ = −λ εȦḂ ψ
A . (4.56)
Thus, when λ is real (R 0), (ψA , φȦ ) and (−ψ A , φȦ ) are two linearly independent
Killing bispinors with the same Killing number λ , while if λ is pure imaginary
A , φȦ ) are two linearly independent Killing bispinors with
(R 0), (ψA , φȦ ) and (ψ
the same Killing number λ .
From (4.54) and its mate it follows that ψA φȦ − ψ A φȦ and i(ψA φȦ + ψ
A φȦ ) are
the spinor equivalents of two real Killing vectors, and from the relation
∇AȦ (ψB φḂ ± ψ

B φḂ ) = εAB (λ φȦ φḂ ∓ λ φḂ φȦ ) ± εȦḂ (λ ψA ψ
B ∓ λ ψB ψ
A )
it follows that if λ is real, then i(ψA φȦ − ψ

A φȦ ) is the spinor equivalent of a real
Killing vector, while if λ is pure imaginary, ψA φȦ + ψ A φȦ is the spinor equivalent of
a real Killing vector. Thus, a Killing bispinor gives rise to three linearly independent
Killing vectors (cf. Friedrich 2000, Sec. 5.2).
Making use of (4.51) and (4.56) one can verify that if λ is real, then ψ A ψ A −
Ȧ A Ȧ
φ φȦ is a (real) constant, and if λ is pure imaginary, then ψ ψA + φ φȦ is a constant.
Eigenbispinors of the Dirac Operator
We shall say that the pair of nonvanishing one-index spinor fields (ψA , φȦ ) is an
eigenbispinor of the Dirac operator if
∇A Ȧ ψA = λ φȦ , ∇A Ȧ φȦ = λ ψA , (4.57)

for some constant λ [cf. (3.34)]. Note that a Killing bispinor is an eigenbispinor
of the Dirac operator, but the converse is not true. If we further assume that M is
compact, then λ must be pure imaginary, since when the metric of M is positive
definite, we have

λ ψ Aψ
A dv = ψ
A ∇AȦ φȦ dv
M M

= ∇AȦ (ψ
A φȦ ) − φȦ ∇AȦ ψ
A dv
M

= φȦ λ φȦ dv,
M
i.e.,
λ ψ Aψ
A dv = −λ φ Ȧ φȦ dv, (4.58)
M M

and similarly one finds that λ dv = −λ ψ A ψ
Mφ
Ȧ φ A dv, which implies that
Ȧ M

(λ + λ ) A + φ Ȧ φȦ ) dv = 0.
(ψ A ψ
M
Hence, λ = −λ , and if λ does not vanish, (4.58) yields

ψ Aψ
A dv = φ Ȧ φȦ dv. (4.59)
M M
Making use of the Ricci identities, from equations (4.57) and their mates one finds
that if λ is equal to zero then the curvature of M vanishes.
A lower bound for |λ | can be obtained by noting that for any spinor field χ A , the
identity
∇BĊ ∇AĊ χ A = ∇(BĊ ∇A)Ċ χ A + 12 εBA ∇SĊ ∇SĊ χ A

= BA χ + 12 ∇AĊ ∇AĊ χB
A
= 18 RχB + 12 ∇AĊ ∇AĊ χB (4.60)
holds. Therefore, if (ψA , φȦ ) is an eigenbispinor of the Dirac operator, then ∇AĊ ∇AĊ
ψB = −2λ 2 ψB − 14 RψB and

(2λ 2 + 14 R)ψ B ψ
B dv = ψ
B ∇AĊ ∇AĊ ψB dv
M M

= [∇AĊ (ψ
B ∇AĊ ψB ) − (∇AĊ ψ
B )∇AĊ ψB ] dv
M

=− (∇AĊ ψ B )(∇AĊ ψB ) dv. (4.61)
M
4.3 Killing Bispinors. The Dirac Operator 165
The integrand in the last expression is greater than or equal to zero, and therefore

2|λ |2 ψ Bψ
B dv 1
4 Rψ B ψ
B dv 14 R0 ψ Bψ
B dv,
M M M
where R0 denotes the minimum of the scalar curvature on M; thus, |λ |2 18 R0 .

In order to find a smaller lower bound for |λ | we introduce the spinor fields
ξABĊ ≡ ∇AĊ ψB − 12 λ εAB φĊ , ηȦḂC ≡ ∇CȦ φḂ − 12 λ εȦḂ ψC
[cf. (4.51)], so that (4.57) are equivalent to ξABĊ = ξ(AB)Ċ and ηȦḂC = η(ȦḂ)C . Then
(∇AĊ ψ B )(∇AĊ ψB ) = (ξ ABĊ + 12 λ ε AB φ Ċ )(ξABĊ + 12 λ εAB φĊ )
= ξ ABĊ ξABĊ + 12 |λ |2 φ Ċ φĊ ,
and making use of (4.61) and (4.59), we have

0 ξ ABĊ ξABĊ dv
M

= (∇AĊ ψ B )(∇AĊ ψB )− 12 |λ |2 φ Ċ φĊ dv
M
= B − 12 |λ |2 φ Ċ φĊ dv
− (2λ 2 + 14 R)ψ B ψ
M B
2 |λ | − 4 R ψ ψ
= 3 2 1 B dv.
M
Thus |λ |2 16 R0 . Furthermore, if |λ |2 = 16 R0 , then R is constant and ξABĊ = 0, which

means that ∇AĊ ψB = 12 λ εAB φĊ . In a similar manner one finds that ηȦḂC = 0, which
is equivalent to ∇CȦ φḂ = 12 λ εȦḂ ψC , and therefore, (ψA , φȦ ) is a Killing bispinor
with Killing number 12 λ .
Exercises
4.1. Show that the solution of the zero-rest-mass field equations (3.42) with respect
to the null tetrad (4.26) can be expressed in terms of a scalar potential [cf. (4.38)].
4.2. Find the commutators of the three linearly independent Killing vector fields
induced by a Killing bispinor when the signature of the metric is Euclidean.
Appendix A
Bases Induced by Coordinate Systems
In the traditional tensor formalism, the vector or tensor fields and the connections
are given through their components with respect to the bases induced by coordinate
systems, instead of rigid bases as the orthonormal or the null tetrads. Given the
components tμν ...ρ of a tensor field with respect to the basis induced by a coordinate
system xμ , the components of its spinor equivalent are defined by
1 1 1
tAȦBḂ...DḊ ≡ √ σ μ AȦ √ σ ν BḂ · · · √ σ ρ DḊ tμν ...ρ , (A.1)
2 2 2
where the Infeld–van der Waerden symbols σ μ AḂ are complex-valued functions
such that
gμν σ μ AḂ σ ν CḊ = −2εAC εḂḊ (A.2)
and the gμν are the components of the metric tensor with respect to the coordinate
system xμ . (By contrast, the Infeld–van der Waerden symbols associated with a
rigid tetrad are constant.) We use here almost the same notation as in the preceding
chapters for the Infeld–van der Waerden symbols, only with a Greek letter instead
of a Latin letter for the first index.
The Infeld–van der Waerden symbols can be explicitly obtained for a metric
given in terms of a coordinate system, by finding first a set of one-forms θ AḂ such
that the metric tensor is expressed as g = −2θ√ 11̇ θ 22̇ + 2θ 12̇ θ 21̇ [see (2.99)], and
then reading off the coefficients in θ = (1/ 2)σμ AḂ dxμ [cf. (2.14)]. Finally,
A Ḃ
σ ν AḂ = gμν εAC εḂḊ σν CḊ .

The components of the covariant derivative of a tensor field involve the Christoffel
symbols, which are given by the well-known formula

μ ∂ gνλ ∂ gρλ ∂ gνρ
Γνρ = 12 gμλ + − . (A.3)
∂ xρ ∂ xν ∂ xλ
For instance, the components of the covariant derivative of a vector field are
∂ tν
∇μ t ν = + Γρνμ t ρ ,
∂ xμ
167
168 A Bases Induced by Coordinate Systems
which are also denoted by t ν ;μ . Therefore, using the fact that
1
t ρ = − √ σ ρ E Ḟ t E Ḟ ,
2
the components of the spinor equivalent of ∇μ t ν are given by
ν
1 μ CḊ ∂ t ν ρ
∇AḂ t = σ AḂ σν
CḊ
+ Γρ μ t
2 ∂ xμ

∂ t E Ḟ ν
1 μ ν E Ḟ ∂ σ E Ḟ ν ρ
= − √ σ AḂ σν CḊ
σ E Ḟ μ + t + Γρ μ σ E Ḟ t E Ḟ
2 2 ∂x ∂ xμ
= ∂AḂ t CḊ − Γ CḊ E ḞAḂ t E Ḟ ,
where the vector fields

1 ∂
∂AḂ ≡ √ σ μ AḂ μ (A.4)
2 ∂x
form a null tetrad and

1 ∂ σ ν E Ḟ
Γ CḊ E ḞAḂ = √ σ μ AḂ σν CḊ + Γ ν ρ
ρμ σ E Ḟ . (A.5)
2 2 ∂ xμ
(An example in which one can identify the functions σ μ AḂ is given by (2.108).)
Since in the present case the components of the metric tensor need not be con-
stant, one has to be careful with the order in which the partial derivatives and
the raising or lowering of indices are applied. Using the fact that ∂ gνλ /∂ xμ =
Γνκμ gκλ + Γλκμ gνκ , which is equivalent to (A.3), from (A.2) we obtain
∂ σ ν E Ḟ ∂ σν CḊ
σν CḊ μ
= −σ ν E Ḟ
∂x ∂ xμ
∂
= −σ ν E Ḟ μ (gνλ σ λ CḊ )
∂x
∂ σ λ CḊ
= −σλ E Ḟ − σ ν E Ḟ σ λ CḊ (Γνκμ gκλ + Γλκμ gνκ ),
∂ xμ
and substituting into (A.5), we find that Γ CḊ E ḞAḂ = −ΓE Ḟ CḊ AḂ , which implies that
ΓCḊE ḞAḂ = ΓCEAḂ εḊḞ + ΓḊḞ ḂA εCE . Hence, the spin coefficients for the null tetrad
(A.4) are related to the Christoffel symbols by
ν
1 ∂ σ E Ḋ
ΓCEAḂ = √ σ μ AḂ σν C Ḋ + Γ ν ρ
ρμ σ E Ḋ ,
4 2 ∂ xμ
ν (A.6)
1 μ ∂ σ CḞ ν ρ
ΓḊḞ ḂA = √ σ AḂ σν Ḋ C
+ Γρ μ σ CḞ
4 2 ∂ xμ
[cf. (2.18)].
A Bases Induced by Coordinate Systems 169
If we consider two conformally related metrics g = φ 2 g as in Section 2.3, for a

given null tetrad ∂AḂ associated with the metric g, we can define a null tetrad ∂ AḂ
for the metric g , by ∂ AḂ = φ ∂AḂ [see (2.132)]. Expressing the vector fields ∂ AḂ in
the form (A.4), in terms of the basis induced by the coordinates xμ , we have
σ μ AḂ = φ σ μ AḂ ,
or equivalently,
σ μ AḂ = φ −1 σμ AḂ .
(The tensor indices of σ μ AḂ are lowered by means of gμν = φ −2 gμν .) With the
aid of the coefficients σ μ AḂ and σ μ AḂ we can find the relation between the tensor
components of objects corresponding to the metrics g and g , with respect to a given
coordinate system. For example, making use of (2.136) and the fact that CABĊḊ is
the spinor equivalent of 12 (Rμν − 14 Rgμν ), we obtain
Rμν − 14 R gμν = Rμν − 14 Rgμν + 2φ −1 (∇μ ∇ν φ − 14 gμν ∇ρ ∇ρ φ ),
where Rμν and R are the components of the Ricci tensor and the scalar curvature,
respectively, of the metric g . Hence,
Rμν = Rμν + 2φ −1∇μ ∇ν φ + φ −1 gμν ∇ρ ∇ρ φ − 3φ −2gμν (∇ρ φ )(∇ρ φ ). (A.7)
In a similar way, making use of the fact that CABCD εȦḂ εĊḊ + CȦḂĊḊ εAB εCD is the
spinor equivalent of the Weyl tensor, from (2.135) and (2.138) we find that

Cμνρσ = φ −2Cμνρσ . (A.8)
References
1. Ares de Parga, G., Chavoya, O., and López-Bonilla, J.L. (1989). Lanczos potential, J. Math.
Phys. 30, 1294.
2. Ashtekar, A. (1987). New Hamiltonian formulation of general relativity, Phys. Rev. D 36,
1587.
3. Ashtekar, A. (1991). Lectures on Non-perturbative Canonical Gravity (World Scientific,
Singapore).
4. Bampi, F. and Caviglia, G. (1983). Third-order tensor potentials for the Riemann and Weyl
tensors, Gen. Rel. Grav. 15, 375.
5. Baum, H. (2000). Twistor and Killing spinors in Lorentzian geometry, Séminaires et Con-
grès, Société Mathématique de France 4, 35.
6. Berestetskii, V.B., Lifshitz, E.M., and Pitaevskii, L.P. (1982). Quantum Electrodynamics, 2nd
ed. (Pergamon, Oxford).
7. Bergqvist, G. and Lankinen, P. (2004). Unique characterization of the Bel–Robinson tensor,
Class. Quantum Grav. 21, 3499.
8. Boyer, C.P., Finley, III, J.D., and Plebański, J.F. (1980). Complex general relativity, H and
HH spaces – a survey of one approach, in General Relativity and Gravitation, Vol. 2, ed. A.
Held (Plenum, New York), pp. 241–281.
9. Buchdahl, H.A. (1958). On the compatibility of relativistic wave equations for particles of
higher spin in the presence of a gravitational field, Nuovo Cim. 10, 96.
10. Cartan, E. (1966). The Theory of Spinors (Hermann, Paris) (Dover, New York, reprinted
1981).
11. Carter, B. and McLenaghan, R.G. (1979). Generalized total angular momentum operator for
the Dirac equation in curved space-time, Phys. Rev. D 19, 1093.
12. Catenacci, R., Marzuoli, A., and Salmistraro F. (1980). A note on Killing vectors in alge-
braically special vacuum space-times, Gen. Rel. Grav. 12, 575.
13. Chandrasekhar, S. (1983). The Mathematical Theory of Black Holes (Clarendon, Oxford).
14. Chevalley, C. (1996). The Algebraic Theory of Spinors and Clifford Algebras (Springer,
Berlin).
15. Cohen, J.M. and Kegeles, L.S. (1974). Electromagnetic fields in curved spaces: A construc-
tive procedure, Phys. Rev. D 10, 1070.
16. Conlon, L. (2001). Differentiable Manifolds, 2nd ed. (Birkhäuser, Boston).
17. Davydov, A.S. (1988). Quantum Mechanics, 2nd ed. (Pergamon, Oxford).
18. Debever, R. (1958). La super-énergie en relativité générale. Bull. Soc. Math. Belg. 10, 112.
19. Debney, G.C. (1971a). Invariant approach to a space-time symmetry, J. Math. Phys. 12, 1088.
20. Debney, G.C. (1971b). On vacuum space-times admitting a null Killing bivector, J. Math.
Phys. 12, 2372.
171
172 References
21. Dunajski, M. and West, S. (2008). Anti-self-dual conformal structures in neutral signature,
in Recent Developments in Pseudo-Riemannian Geometry, ESI Lectures in Mathematics
and Physics, ed. D. Alekseevsky and H. Baum (European Mathematical Society Publishing
House, Zürich), pp. 113–148.
22. Ernst, F.J. (1968a). New formulation of the axially symmetric gravitational field problem,
Phys. Rev. 167, 1175.
23. Ernst, F.J. (1968b). New formulation of the axially symmetric gravitational field problem. II,
Phys. Rev. 168, 1415.
24. Finley III, J.D. and Plebański, J.F. (1976). The intrinsic spinorial structure of hyperheavens,
J. Math. Phys. 17, 2207.
25. Flaherty, Jr., E.J. (1980). Complex variables in relativity, in General Relativity and Gravita-
tion, Vol. 2, ed. A. Held (Plenum, New York), pp. 207–239.
26. Friedrich, Th. (2000). Dirac Operators in Riemannian Geometry (American Mathematical
Society, Providence, Rhode Island).
27. Geroch, R. (1968). Spinor structure of space-times in general relativity. I, J. Math. Phys. 9,
1739.
28. Geroch, R. (1970). Spinor structure of space-times in general relativity. II, J. Math. Phys.
11, 343.
29. Geroch, R.P. (1971). A method for generating solutions of Einstein’s equations, J. Math.
Phys. 12, 918.
30. Geroch, R.P., Held, A., and Penrose, R. (1973). A space-time calculus based on pairs of null
directions, J. Math. Phys. 14, 874.
31. Goldberg, J.N. and Sachs, R.K. (1962). A theorem on Petrov types, Acta Phys. Polon., Suppl.
22, 13. Reprinted in Gen. Rel. Grav. 41, 433 (2009).
32. Goldblatt, E. (1994a). A Newman-Penrose formalism for gravitational instantons, Gen. Rel.
Grav. 26, 979.
33. Goldblatt, E. (1994b). Symmetries of type D+ D− gravitational instantons, J. Math. Phys. 35,
3029.
34. Hall, G.S. (2004). Symmetries and Curvature Structure in General Relativity (World
Scientific, Singapore).
35. Hansen, R.O., Newman, E.T., Penrose, R., and Tod, K.P. (1978). The metric and curvature
properties of H -space, Proc. R. Soc. Lond. A 363, 445.
36. Heusler, M. (1996). Black Hole Uniqueness Theorems (Cambridge University Press,
Cambridge).
37. Hirsch, M. and Smale, S. (1974). Differential Equations, Dynamical Systems, and Linear
Algebra (Academic, New York).
38. Huggett, S.A. and Tod, K.P. (1994). An Introduction to Twistor Theory, 2nd ed. (Cambridge
University Press, Cambridge).
39. Hughston, L.P., Penrose, R., Sommers, P., and Walker, M. (1972). On a quadratic first integral
for the charged particle orbits in the charged Kerr solution, Commun. Math. Phys. 27, 303.
40. Hughston, L.P. and Sommers, P. (1973). The symmetries of Kerr black holes, Commun. Math.
Phys. 33, 129.
41. Infeld, L. and van der Waerden, B.L. (1933). Die Wellengleichung des Elektrons in der all-
gemeinen Relativitätstheorie, Sitzber. preuss. Akad. Wiss. Physik-math. Kl. 9, 380.
42. Israel, W. (1979) Differential Forms in General Relativity, 2nd ed. (Dublin Institute for Ad-
vanced Studies, Dublin).
43. Jackson, J.D. (1975). Classical Electrodynamics, 2nd ed. (Wiley, New York).
44. Jeffryes, B.P. (1984). Space-times with two-index Killing spinors, Proc. R. Soc. Lond. A 392,
323.
45. Kamran, N. (1987). Séparation des variables pour les potentiels de Debye dans toutes les
solutions de type D des équations d’Einstein, C.R. Acad. Sci. Paris, 304, Ser. I, 299.
46. Kamran, N. and McLenaghan, R.G. (1983). Separation of variables and quantum numbers
for Weyl neutrino fields on curved spacetime, Lett. Math. Phys. 7, 381.
47. Kamran, N. and McLenaghan, R.G. (1984a). Symmetry operators for neutrino and Dirac
fields on curved spacetime, Phys. Rev. D 30, 357.
References 173
48. Kamran, N. and McLenaghan, R.G. (1984b). Separation of variables and symmetry operators
for the neutrino and Dirac equations in the space-times admitting a two-parameter abelian
orthogonally transitive isometry group and a pair of shearfree geodesic null congruences,
J. Math. Phys. 25, 1019.
49. Kobayashi, S. and Nomizu, K. (1963). Foundations of Differential Geometry, Vol. 1, (Wiley-
Interscience, New York).
50. Lanczos, C. (1962). The splitting of the Riemann tensor, Rev. Mod. Phys. 34, 379.
51. Lanczos, C. (1970). The Variational Principles of Mechanics, 4th ed. (University of Toronto
Press, Toronto) (Dover, New York, reprinted 1986).
52. Lawson, H.B. and Michelsohn, M.-L. (1989). Spin Geometry (Princeton University Press,
Princeton, N.J.).
53. Lichnerowicz, A. (1964). Propagateurs, commutateurs et anticommutateurs en relativité
générale, in Relativity, Groups and Topology, ed. B.S. DeWitt and C. DeWitt (Gordon and
Breach, New York), pp. 821–861.
54. Lounesto, P. (1997). Clifford Algebras and Spinors (Cambridge University Press, Cam-
bridge).
55. Maher, Jr., W.F. and Zund, J.D. (1968). A spinor approach to the Lanczos spin-tensor, Nuovo
Cim. 57, 638.
56. Mariot, L. (1954). Le champ électromagnétique singulier, C. R. Acad. Sci. Paris 238, 2055.
57. Mason, L.J. and Woodhouse, N.M.J. (1996). Integrability, Self-Duality, and Twistor Theory
(Oxford University Press, Oxford).
58. McLehaghan, R.G., Smith, S.N., and Walker, D.M. (2000). Symmetry operators for spin-1/2
relativistic wave equations on curved space-time, Proc. R. Soc. Lond. A 456, 2629.
59. Merzbacher, E. (1998). Quantum Mechanics, 3rd ed. (Wiley, New York).
60. Messiah, A. (1962). Quantum Mechanics, Vol. II, (North Holland, Amsterdam).
61. Misner, C.W., Thorne, K.S., and Wheeler, J.A. (1973). Gravitation (Freeman, New York),
Chap. 41.
62. Mustafa, E. and Cohen, J.M. (1987). Hertz and Debye potentials and electromagnetic fields
in general relativity, Class. Quantum Grav. 4, 1623.
63. Newman, E.T. (1976). Heaven and its properties, Gen. Rel. Grav. 7, 107.
64. Newman, E.T. (1978). Source-free Yang-Mills theories, Phys. Rev. D 18, 2901.
65. Newman, E.T. and Penrose, R. (1962). An approach to gravitational radiation by a method
of spin coefficients, J. Math. Phys. 3, 566; (1963) 4, 998.
66. Newman, E.T. and Penrose, R. (1966). Note on the Bondi–Metzner–Sachs group, J. Math.
Phys. 7, 863.
67. O’Donnell, P. (2003). Introduction to 2-Spinors in General Relativity (World Scientific,
Singapore).
68. Papapetrou, A. (1953). Eine rotationssymmetrische Lösung in der allgemeinen Relativitäts-
theorie, Ann. Physik 12, 309.
69. Penrose, R. (1960). A spinor approach to general relativity, Ann. Phys. 10, 171.
70. Penrose, R. (1965). Zero rest-mass fields including gravitation: asymptotic behaviour, Proc.
R. Soc. Lond. A 284, 159.
71. Penrose, R. (1976). Nonlinear gravitons and curved twistor theory, Gen. Rel. Grav. 7, 31.
72. Penrose, R. and Rindler, W. (1984). Spinors and Space-Time, Vol. 1, (Cambridge University
Press, Cambridge).
73. Penrose, R. and Ward, R.S. (1980). Twistors for flat and curved space-time, in General Rel-
ativity and Gravitation, Vol. 2, ed. A. Held (Plenum, New York), pp. 283–328.
74. Petersen, P. (1998). Riemannian Geometry (Springer, New York).
75. Pirani, F.A.E. (1965). In Lectures on General Relativity, 1964 Brandeis Summer Institute,
Vol. 1, ed. S. Deser and K.W. Ford (Prentice-Hall, Englewood Cliffs, NJ).
76. Plebański, J.F. (1965). The “vectorial” optics of fields with arbitrary spin, rest-mass zero,
Acta Phys. Polon. 27, 361.
77. Plebański, J.F. (1974). Spinors, Tetrads and Forms, unpublished monograph, Centro de In-
vestigación y de Estudios Avanzados del IPN, México, D.F., México.
174 References
78. Plebański, J.F. (1975). Some solutions of complex Einstein equations, J. Math. Phys. 16,
2395.
79. Plebański, J.F. and Robinson, I. (1976). Left-degenerate vacuum metrics, Phys. Rev. Lett. 37,
493.
80. Porteous, I.R. (1995). Clifford Algebras and the Classical Groups (Cambridge University
Press, Cambridge).
81. Rindler, W. (1977). Essential Relativity, Special, General, and Cosmological, 2nd ed.
(Springer, New York).
82. Robinson, I. (1961). Null electromagnetic fields, J. Math. Phys. 2, 290.
83. Sen, A. (1981). On the existence of neutrino “zero-modes” in vacuum spacetimes, J. Math.
Phys. 22, 1781.
84. Sen, A. (1982). Quantum theory of spin-3/2 field in Einstein spaces, Int. J. Theor. Phys. 21, 1.
85. Shaw, W.T. (1983a). Spinor fields at spacelike infinity, Gen. Rel. Grav. 15, 1163.
86. Shaw, W.T. (1983b). Twistor theory and the energy–momentum and angular momentum of
the gravitational field at spatial infinity, Proc. R. Soc. Lond. A 390, 191.
87. Sommers, P. (1980). Space spinors, J. Math. Phys. 21, 2567.
88. Stephani, H., Kramer, D., MacCallum, M.A.H., Hoenselaers, C., and Herlt, E. (2003). Exact
Solutions to Einstein’s Equations, 2nd ed. (Cambridge University Press, Cambridge).
89. Stewart, J. (1990). Advanced General Relativity (Cambridge University Press, Cambridge).
90. Taxiarchis, P. (1985). Space-Times admitting Penrose–Floyd tensors, Gen. Rel. Grav.
17, 149.
91. Torres del Castillo, G.F. (1983). Null strings and Bianchi identities, J. Math. Phys. 24, 590.
92. Torres del Castillo, G.F. (1984a). A note on D(k, 0) Killing spinors, Commun. Math. Phys.
92, 485.
93. Torres del Castillo, G.F. (1984b). Null strings and Hertz potentials, J. Math. Phys. 25, 342.
94. Torres del Castillo, G.F. (1985). Killing spinors and massless spinor fields, Proc. R. Soc.
Lond. A 400, 119.
95. Torres del Castillo, G.F. (1986). Killing spinors and gravitational perturbations, J. Math.
Phys. 27, 1583.
96. Torres del Castillo, G.F. (1988). The Teukolsky–Starobinsky identities in type D vacuum
backgrounds with cosmological constant, J. Math. Phys. 29, 2078.
97. Torres del Castillo, G.F. (1999). Debye potentials for self-dual fields, Gen. Rel. Grav. 31,
205.
98. Torres del Castillo, G.F. (2003). 3-D Spinors, Spin-Weighted Functions and Their Applica-
tions (Birkhäuser, Boston).
99. Wald, R.M. (1978). Construction of solutions of gravitational, electromagnetic, or other per-
turbation equations from solutions of decoupled equations, Phys. Rev. Lett. 41, 203.
100. Wald, R.M. (1984). General Relativity (University of Chicago Press, Chicago).
101. Walker, M. and Penrose, R. (1970). On quadratic first integrals of the geodesic equations for
type {22} spacetimes, Commun. Math. Phys. 18, 265.
102. Ward, R.S. and Wells, Jr., R.O. (1990). Twistor Geometry and Field Theory (Cambridge
University Press, Cambridge).
103. Zund, J.D. (1975). The theory of the Lanczos spinor, Ann. Mat. Pura Appl., Ser. IV, 104, 239.
Index
A invariance, 121
aberration of light, 37 rescalings, 93, 169
adjoint of a spinor, 40, 44, 56 congruence of null geodesics, 124
algebraic classification conjugate of a spinor, 40, 44, 56
of the conformal curvature, 91 connection, 68
of the electromagnetic field, 112 connection one-forms, 68
of totally symmetric spinors, 60 connection symbols, 6
algebraically general continuity equation, 116
conformal curvature, 93 contracted Bianchi identities, 89
electromagnetic field, 110 covariant derivative
algebraically special of a bispinor, 119
conformal curvature, 93 of a spinor field, 74
electromagnetic field, 110 of a tensor field, 72
angular momentum, 99 of a vector field, 72, 167
anti-self-dual curl, 132
part, 14 curvature
two-forms, 78 tensor, 77
antisymmetrization, 15 two-forms, 77
B D
Bel–Robinson tensor, 122 Debever–Penrose vector, 92
Bianchi identities, 77, 89 differential
bispinors, 53 of a one-form, 69
inner product, 55, 65 of a scalar function, 83
bivector, 14 Dirac adjoint, 55
simple, 14, 61, 62, 64 Dirac equation, 50, 116, 143
boost weight, 76 covariance, 55
Dirac operator, 163
C Dirac spinors, 53
Cartan’s first structure equations, 69 dominant energy condition, 109
Cartan’s second structure equations, 77 Doppler shift, 38
Christoffel symbols, 167 dual of an antisymmetric two-index tensor, 14
Clifford algebra, 50
commutators, 69, 89 E
conformal Einstein’s field equations, 122
curvature tensor, 82 Einstein–Maxwell equations, 128
equivalence, 93, 169 electromagnetic field, 104
175
176 Index
algebraically general, 110 Killing vector, 95, 108, 131, 134, 163
algebraically special, 110 conformal, 97
energy–momentum tensor, 109 homothetic, 97
principal null directions, 113 Killing–Yano tensor, 142, 146
self-dual, 105 Klein–Gordon equation, 50
electromagnetic plane wave, 37, 114 Kleinian signature, 2
electromagnetic spinor, 104
Ernst L
equation, 133 Lanczos potential, 101
potential, 133, 145 Levi-Civita connection, 69
Euclidean Schwarzschild metric, 88 Levi-Civita symbol, 6, 14
Euclidean signature, 2 Lie algebra, 16, 22
Euler angles, 35 Lie derivative, 96
exponential, 26 of a bispinor, 98, 147
exterior derivative, 69 of a spinor field, 96
exterior product, 69 Lie transport, 125
Lorentz boosts, 30, 47
Lorentz transformations, 30
F
improper, 38
four-potential, 105
orthochronous proper, 30, 33
simple, 33
G
Lorentzian signature, 2
gauge transformations, 105, 150
geodesics, 125, 128, 140, 146 M
Geroch–Held–Penrose notation, 75 Mariot–Robinson theorem, 124
Goldberg–Sachs theorem, 127 mate of a spinor, 24, 40, 44, 56
gravitational radius, 86 Maxwell’s equations, 104, 106, 108, 119, 123,
134, 142, 144, 158
H metric tensor, 2, 67
H spaces, 160
heavenly equation, 162 N
helicity, 120 Newman–Penrose notation, 70, 71
H H spaces, 154 null
hyperbolic signature, 2 geodesics, 137, 140
hyperbolic space, 95 rotations, 34
hypersurface orthogonal, 131 tetrad, 5, 52, 168
vectors, 5, 31, 37
I
improper O
O(2, 2) transformations, 41 O(p, q), 2
O(3, 1) transformations, 38 one-index spinors, 10
O(4) transformations, 28 orthogonal
orthogonal transformations, 20 projection, 49
Infeld–van der Waerden symbols, 6, 53, 167 transformations, 2
inner product, 55, 57, 65 orthonormal basis, 2
irreducible representations, 53, 59, 63 orthonormal tetrad, 68, 77
isometries, 95
P
K passive transformations, 65
Kerr metric, 119, 139 Pauli matrices, 8, 16
Kerr–Newman solution, 139 Pauli’s theorem, 55
Killing bispinor, 162 Petrov–Penrose classification, 92
Killing equations, 95 plane waves, 114
Killing spinor, 137, 146 plane-waves, 120
Killing tensor, 100, 146 primed indices, 6
Index 177
principal null direction spin spaces, 10

of the conformal curvature, 92 spin transformations, 44, 45, 54, 59
of the electromagnetic field, 113 irreducible representations, 59
principal spinors, 60 spin weight, 76, 99
proper orthogonal transformations, 20 spin-0 zero-rest-mass field, 121
spin-3/2 field, 64, 145
Q spin-tensors, 15
quaternions, 34 spin-weighted spherical harmonics, 107, 118
R spinor equivalent, 9
R2,2 , 3 of a symmetric two-index tensor, 11
raising and lowering of an antisymmetric four-index tensor, 13
of spinor indices, 6 of an antisymmetric three-index tensor, 12
of tensor indices, 3 of an antisymmetric two-index tensor, 11
rapidity, 30, 32, 33 of the curvature tensor, 79
reflections, 45 of the curvature two-forms, 79
Reissner–Nordström solution, 139 of the electromagnetic field tensor, 104
restricted Lorentz group, 30 of the energy–momentum tensor, 109
Ricci rotation coefficients, 69 of the Maxwell equations, 104
Ricci tensor, 79 of the Ricci identities, 90
Riemannian connection, 69 of the Ricci tensor, 82
of the Weyl curvature tensor, 82
S spinor fields, 73
S4 , 84, 94, 155 spinors, 10
scalar curvature, 79 stationary axisymmetric space-times, 133
Schwarzschild metric, 86, 99, 106, 139, 144, SU(1, 1), 43
152, 154 SU(2), 22, 63
Schwarzschild radius, 86 parameterization, 23, 35
second heavenly equation, 162 symmetrization, 12
self-dual
electromagnetic field, 105 T
part, 14 tensor product, 10
two-forms, 78 tetrad rotations, 73
Yang–Mills fields, 150 torsion, 68
separation of variables, 106, 130 two-forms, 68
shear-free congruence of null geodesics, 124, total angular momentum, 99
126, 128, 130, 136, 150, 152 twist potential, 132, 145
signature, 2 twistor equation, 137
simple two-index Killing tensor, 100, 146
bivector, 14, 28, 61, 62, 64
U
orthochronous proper Lorentz transforma-
ultrahyperbolic signature, 2
tions, 33, 34, 47
SO(4) transformations, 25, 28, 46 W
SL(2, C), 21 weighted operators, 76
SL(2, R), 39 Weyl curvature tensor, 82, 169
SO↑ (3, 1), 29 Weyl equation, 119, 130, 142, 159
SO(2, 2), 39 Weyl spinor, 82
SO(2, 2)0 , 39
SO(3, 1), 29 Y
SO(3, 1)0 , 29 Yang–Mills fields, 149
SO(p, q), 3 gauge transformations, 150
SO(4), 22
spatial inversion, 38, 39 Z
special relativity, 30 zero-rest-mass field equations, 119, 123, 145
spin, 63 conformal invariance, 121
spin coefficients, 70, 168 integrability conditions, 121

(Progress in Mathematical Physics 59) Gerardo F. Torres Del Castillo (Auth.) - Spinors in Four-Dimensional Spaces-BirkhÃ User Basel (2010)

Uploaded by

Copyright:

Available Formats

(Progress in Mathematical Physics 59) Gerardo F. Torres Del Castillo (Auth.) - Spinors in Four-Dimensional Spaces-BirkhÃ User Basel (2010)

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Progress in Mathematical Physics 59) Gerardo F. Torres Del Castillo (Auth.) - Spinors in Four-Dimensional Spaces-BirkhÃ User Basel (2010)

Uploaded by

Copyright:

Available Formats

What is the aim of the book?

What is the aim of the book?

What approach does the book take to presenting spinor formalism?

What approach does the book take to presenting spinor formalism?

Progress in Mathematical Physics

ISBN 978-0-8176-4983-8 e-ISBN 978-0-8176-4984-5

Library of Congress Control Number: 2010930605

c Springer Science+Business Media, LLC 2010

Printed on acid-free paper

Puebla, Puebla, México Gerardo F. Torres del Castillo

2 Connection and Curvature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3 Applications to General Relativity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

4 Further Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149

A Bases Induced by Coordinate Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167

In this chapter the algebraic background of the two-component spinor formalism is

G.F.T. del Castillo, Spinors in Four-Dimensional Spaces, Progress in Mathematical 1

1.1 Orthogonal Groups

Let V be a four-dimensional real vector space with a metric tensor g (i.e., g is a

g(ea , eb ) = g(ea , eb ), (1.2)

signature of g. The matrices satisfying (1.4) with det(La b ) = 1 form a subgroup of

with entries (L−1 )a b , is given by

algebraic conditions detK = 1 = det M, allow us to find orthogonal transformations.

1.2 Null Tetrads and the Spinor Equivalent of a Tensor

(thus, ψ 2 = ψ1 , ψ 1 = −ψ2 ), and similarly for dotted indices. It should be remarked

If (M R S ) is a (real or complex) 2 × 2 matrix, then

In fact, letting ψBD ≡ εAC M A B MC D , it follows that

ψDB = εAC M A D MC B = −εCA MC B M A D = −ψBD ,

A Convenient Choice for the Infeld–van der Waerden Symbols

and if (gab ) = diag (1, 1, −1, −1), then

By contracting both sides of (1.21) with σcCḊ , one obtains

σ a AḂ σ bCḊ σcCḊ gab = −2εAC εḂḊ σcCḊ = −2σcAḂ,

σaAḂ σb AḂ = −2gab. (1.40)

ta sa = −tAȦ sAȦ . (1.46)

Equation (1.45) allows us to consider the complexification of V as the tensor

The Spinor Equivalents of Symmetric or Antisymmetric Tensors

tAȦBḂ = tBḂAȦ , (1.48)

and therefore t R ȦRḂ is proportional to εȦḂ ; thus

tAȦBḂ − tBȦAḂ = εAB 12 t RṠ RṠ εȦḂ = − 12 t s s εAB εȦḂ , (1.49)

τAB = 12 tA Ṙ BṘ = 12 t R ȦRḂ = τȦḂ

2 μAȦBC = −μB Ḃ AC εȦḂ − μBḂ ȦḂ εAC

and hence, making use of (1.27),

3μAȦBC = μ R ȦRC εAB − μBṘ ȦṘ εAC . (1.55)

Contracting now this last equation with ε BC , we obtain

and therefore, defining tBȦ ≡ 23 μ R ȦRB , from (1.55) we obtain

2 μAȦBC = tCȦ εAB + tBȦ εAC . (1.56)

In an entirely similar way, from (1.54) one arrives at

− (tA Ṙ εȦṘ εĊḂ + 2tAḂεȦĊ )εBC ]

and contracting this last equation with ε ȦĊ , we obtain

0 = −φB Ṙ DṘ εḂḊ + φ R ḂRḊ εBD ,

which implies that

φB Ṙ DṘ = 2a εBD , φ R ḂRḊ = 2a εḂḊ , (1.61)

Therefore, the spinor equivalent of the dual of an antisymmetric two-index tensor

Proposition 1.2. If the bivector tab is simple, then

τAB τ AB = τȦḂ τ ȦḂ . (1.67)

Spin-Tensors and Other Derived Objects

tab = 12 σa AȦ σb BḂ (τAB εȦḂ + τȦḂ εAB )

Making use of (1.23), we obtain the alternative expression

Sab AB = 12 (σa AṘ σb B Ṙ + σa BṘ σb A Ṙ )

where the brackets denote antisymmetrization on the indices enclosed. In a similar

then the product of the matrices (

Then, M Ȧ Ḃ γ Ḃ = eiϕ /2 γ Ȧ , M Ȧ Ḃ γḂ = e−iϕ /2 γȦ , and