Quaternion Algebraic Geometry

Dominic Widdows
June 5, 2006
Abstract
QUATERNION ALGEBRAIC GEOMETRY
DOMINIC WIDDOWS
St Anne’s College, Oxford
Thesis submitted Hilary Term, 2000, in support
of application to supplicate for the degree of D.Phil.
This thesis is a collection of results about hypercomplex and quaternionic manifolds,
focussing on two main areas. These are exterior forms and double complexes, and the
‘algebraic geometry’ of hypercomplex manifolds. The latter area is strongly influenced
by techniques from quaternionic algebra.
A new double complex on quaternionic manifolds is presented, a quaternionic version
of the Dolbeault complex on a complex manifold. It arises from the decomposition of
real-valued exterior forms on a quaternionic manifold M into irreducible representations
of Sp(1). This decomposition gives a double complex of differential forms and operators
as a result of the Clebsch-Gordon formula V
r
⊗V
1

= V
r+1
⊕V
r−1
for Sp(1)-representations.
The properties of the double complex are investigated, and it is established that it is
elliptic in most places.
Joyce has created a new theory of quaternionic algebra [J1] by defining a quaternionic
tensor product for AH-modules (H-modules equipped with a special real subspace). The
theory can be described using sheaves over CP
1
, an interpretation due to Quillen [Q].
AH-modules and their quaternionic tensor products are classified. Stable AH-modules
are described using Sp(1)-representations.
This theory is especially useful for describing hypercomplex manifolds and forming
close analogies with complex geometry. Joyce has defined and investigated q-holomorphic
functions on hypercomplex manifolds. There is also a q-holomorphic cotangent space
which again arises as a result of the Clebsch-Gordon formula. AH-module bundles are
defined and their q-holomorphic sections explored.
Quaternion-valued differential forms on hypercomplex manifolds are of special inter-
est. Their decomposition is finer than that of real forms, giving a second double complex
with special advantages. The cohomology of these complexes leads to new invariants of
compact quaternionic and hypercomplex manifolds.
Quaternion-valued vector fields are also studied, and lead to the definition of quater-
nionic Lie algebras. The investigation of finite-dimensional quaternionic Lie algebras
allows the calculation of some simple quaternionic cohomology groups.
Acknowledgements
I would like to express gratitude, appreciation and respect for my supervisor Do-
minic Joyce. Without his inspiration as a mathematician this thesis would not have
been conceived; without his patience and friendship it would certainly not have been
completed.
Dedication
To the memory of Dr Peter Rowe, 1938-1998, who taught me relativity as an
undergraduate in Durham. His determination to teach concepts before examination
techniques introduced me to differential geometry.
i
Contents
Introduction 1
1 The Quaternions and the Group Sp(1) 4
1.1 The Quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 The Lie Group Sp(1) and its Representations . . . . . . . . . . . . . . . 9
1.3 Difficulties with the Quaternions . . . . . . . . . . . . . . . . . . . . . . . 12
2 Quaternionic Differential Geometry 17
2.1 Complex, Hypercomplex and Quaternionic Manifolds . . . . . . . . . . . 17
2.2 Differential Forms on Complex Manifolds . . . . . . . . . . . . . . . . . . 20
2.3 Differential Forms on Quaternionic Manifolds . . . . . . . . . . . . . . . 23
3 A Double Complex on Quaternionic Manifolds 25
3.1 Real forms on Complex Manifolds . . . . . . . . . . . . . . . . . . . . . . 25
3.2 Construction of the Double Complex . . . . . . . . . . . . . . . . . . . . 28
3.3 Ellipticity and the Double Complex . . . . . . . . . . . . . . . . . . . . . 32
3.4 Quaternion-valued forms on Hypercomplex Manifolds . . . . . . . . . . . 40
4 Developments in Quaternionic Algebra 42
4.1 The Quaternionic Algebra of Joyce . . . . . . . . . . . . . . . . . . . . . 42
4.2 Duality in Quaternionic Algebra . . . . . . . . . . . . . . . . . . . . . . . 48
4.3 Real Subspaces of Complex Vector Spaces . . . . . . . . . . . . . . . . . 52
4.4 K-modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.5 The Sheaf-Theoretic approach of Quillen . . . . . . . . . . . . . . . . . . 56
5 Quaternionic Algebra and Sp(1)-representations 61
5.1 Stable AH-modules and Sp(1)-representations . . . . . . . . . . . . . . . 61
5.2 Sp(1)-Representations and the Quaternionic Tensor Product . . . . . . . 68
5.3 Semistable AH-modules and Sp(1)-representations . . . . . . . . . . . . . 74
5.4 Examples and Summary of AH-modules . . . . . . . . . . . . . . . . . . 77
6 Hypercomplex Manifolds 80
6.1 Q-holomorphic Functions and H-algebras . . . . . . . . . . . . . . . . . . 81
6.2 The Quaternionic Cotangent Space . . . . . . . . . . . . . . . . . . . . . 86
6.3 AH-bundles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
6.4 Q-holomorphic AH-bundles . . . . . . . . . . . . . . . . . . . . . . . . . 93
ii
7 Quaternion Valued Forms and Vector Fields 100
7.1 The Quaternion-valued Double Complex . . . . . . . . . . . . . . . . . . 101
7.2 Vector Fields and Quaternionic Lie Algebras . . . . . . . . . . . . . . . . 109
7.3 Hypercomplex Lie groups . . . . . . . . . . . . . . . . . . . . . . . . . . 112
References 117
iii
Introduction
This aim of this thesis is to describe and develop various aspects of quaternionic algebra
and geometry. The approach is based upon two pillars, namely the differential geometry
of quaternionic manifolds and Joyce’s recent theory of quaternionic algebra. Contribu-
tions are made to both fields of study, enabling these strands to be woven together in
describing the algebraic geometry of hypercomplex manifolds.
A recurrent theme throughout will be representations of the group Sp(1) of unit
quaternions. Our contribution to the theory of quaternionic manifolds relies on decom-
posing the Sp(1)-action on exterior forms, whilst the main new insight in quaternionic
algebra is that the most important building blocks of Joyce’s theory are best described
and manipulated as Sp(1)-representations. The importance of Sp(1)-representations to
both areas is chiefly responsible for the successful synthesis of methods in the work on
hypercomplex manifolds.
Another frequent source of motivation is the behaviour of the complex numbers.
Many situations in complex algebra and geometry have interesting quaternionic ana-
logues. Aspects of complex geometry can often be described using the group U(1) of unit
complex numbers; replacing this with the group Sp(1) can lead directly to quaternionic
versions. The decomposition of exterior forms on quaternionic manifolds is precisely
such an example, as is all the work on q-holomorphic functions and forms on hypercom-
plex manifolds. On the other hand, Joyce’s quaternionic algebra is such a rich theory
precisely because real subspaces of quaternionic vector spaces behave so differently from
real subspaces of complex vector spaces.
Much of the original work presented is enticingly simple — indeed, I have often felt
both surprised and privileged that it has not been carried out before. One of the main
explanations for this is the relative unpopularity suffered by the quaternions in the 20
th
century. This situation has left various aspects of quaternionic behaviour unexplored.
To help understand the reasons for this omission and the consequent opportunities for
development, the first chapter is devoted to a survey of the history of the quaternions
and their applications. Background material also includes an introduction to the group
Sp(1) and its representations. The irreducible representations on complex vector spaces
and their tensor products are described, as are real and quaternionic representations.
Chapter 2 is about quaternionic structures in differential geometry. The approach
is based on the work of Salamon [S3]. Taking complex manifolds as a model, hyper-
complex manifolds (those possessing a torsion-free GL(n, H)-structure) are defined, fol-
lowed by the broader class of quaternionic manifolds (those possessing a torsion-free
Sp(1)GL(n, H)-structure). After reviewing the Dolbeault complex, we consider the
decomposition of differential forms on quaternionic manifolds, including an important
elliptic complex discovered by Salamon upon which the integrability of the quaternionic
1
structure depends.
This complex is in fact the top row of a hitherto undiscovered double complex on
quaternionic manifolds, which is the subject of Chapter 3. As an Sp(1)-representation,
the cotangent space of a quaternionic manifold M
4n
takes the form T

M

= 2nV
1
, where
V
1
is the basic representation of Sp(1) on C
2
. Decomposition of the induced Sp(1)-
representation on Λ
k
T

M is a simple process achieved by considering weights. That this
decomposition gives rise to a double complex results from the Clebsch-Gordon formula
V
r
⊗T

M

= V
r
⊗2nV
1

= 2n(V
r+1
⊕V
r−1
). The new double complex is shown to be elliptic
everywhere except along its bottom row, consisting of the basic representations V
1
and
the trivial representations V
0
. This double complex presents us with new quaternionic
cohomology groups.
In the fourth chapter (which is partly a summary of the work of Joyce [J1] and
Quillen [Q]) we move to our other major area of interest, the theory of quaternionic
algebra. The building blocks of this theory are H-modules equipped with a special real
subspace. Such an object is called an AH-module. Joyce has discovered a canonical
tensor product operation for AH-modules which is both associative and commutative.
Using ideas from Quillen’s work, we classify AH-modules up to isomorphism. Dual
AH-modules are defined and shown to have interesting properties. Particularly well-
behaved is the category of stable AH-modules. In Chapter 5 it is shown that all stable
AH-modules and their duals are conveniently described using Sp(1)-representations.
The resulting theory is ideally adapted for describing hypercomplex geometry, a pro-
cess begun in Chapter 6. A hypercomplex manifold M has a triple of global anticom-
muting complex structures which can be identified with the imaginary quaternions. This
identification enables tensors on hypercomplex manifolds to be treated using the tech-
niques of quaternionic algebra. Joyce has already used such an approach to define and
investigate q-holomorphic functions on hypercomplex manifolds, which are seen as the
quaternionic analogue of holomorphic functions. There is a natural product map on
the AH-module of q-holomorphic functions, which gives the q-holomorphic functions an
algebraic structure which Joyce calls an H-algebra.
Using the Sp(1)-version of quaternionic algebra, we define a natural splitting of the
quaternionic cotangent space H ⊗ T

M

= A ⊕ B, and show that q-holomorphic func-
tions are precisely those whose differentials take values in A ⊂ H ⊗ T

M. The bundle
A is hence defined to be the q-holomorphic cotangent space of M. These spaces are
examples of AH-module bundles or AH-bundles, which we discuss. Several parallels
with complex geometry arise. There are q-holomorphic AH-bundles with q-holomorphic
sections. Q-holomorphic sections are described using the quaternionic tensor product
and the q-holomorphic cotangent space, and seen to form an H-algebra module over the
q-holomorphic functions.
In the final chapter, such methods are applied to quaternion-valued tensors on hy-
percomplex manifolds. The double complex of Chapter 3 is revisited and adapted to
quaternion-valued differential forms. The global complex structures give an extra decom-
position which generalises the splitting H⊗T

M

= A ⊕B, further refining the double
complex. The quaternion-valued double complex has advantages over the real-valued
version, being elliptic in more places. The top row of the quaternion-valued double com-
plex is particularly well-adapted to quaternionic algebra, which presents close parallels
with the Dolbeault complex and motivates the definition of q-holomorphic k-forms.
2
Quaternion-valued vector fields are also interesting. The quaternionic tangent space
splits as H⊗TM

=
´
A⊕
´
B in the same way as the cotangent space. Vector fields taking
values in
´
A are closed under the quaternionic tensor product and Lie bracket, a result
which depends upon the integrability of the hypercomplex structure. This is the quater-
nionic analogue of the statement that on a complex manifold, the (1, 0) vector fields are
closed under the Lie bracket. The vector fields in question therefore form a quaternionic
Lie algebra, a new concept which we introduce. Interesting finite-dimensional quater-
nionic Lie algebras are used to calculate some quaternionic cohomology groups on Lie
groups with left-invariant hypercomplex structures.
3
Chapter 1
The Quaternions and the Group
Sp(1)
1.1 The Quaternions
The quaternions Hare a four-dimensional real algebra generated by the identity element 1
and the symbols i
1
, i
2
and i
3
, so H = ¦r
0
+r
1
i
1
+r
2
i
2
+r
3
i
3
: r
0
, . . . , r
3
∈ R¦. Quaternions
are added together component by component, and quaternion multiplication is given by
the quaternion relations
i
1
i
2
= −i
2
i
1
= i
3
, i
2
i
3
= −i
3
i
2
= i
1
, i
3
i
1
= −i
1
i
3
= i
2
, i
2
1
= i
2
2
= i
2
3
= −1 (1.1)
and the distributive law. The quaternion algebra is not commutative, though it does
obey the associative law. The quaternions are a division algebra (an algebra with the
property that ab = 0 implies that a = 0 or b = 0 ).
• Define the imaginary quaternions I = ¸i
1
, i
2
, i
3
). The symbol I is not standard,
but we will use it throughout.
• Define the conjugate ¯ q of q = q
0
+q
1
i
1
+q
2
i
2
+q
3
i
3
by ¯ q = q
0
−q
1
i
1
−q
2
i
2
−q
3
i
3
.
Then (pq) = ¯ q¯ p for all p, q ∈ H.
• Define the real and imaginary parts of q by Re(q) = q
0
∈ R and Im(q) = q
1
i
1
+
q
2
i
2
+ q
3
i
3
∈ I. As with complex numbers, ¯ q = Re(q) −Im(q).
• We regard the real numbers R as a subfield of H, and the quaternions as a direct
sum H

= R ⊕I.
• Let q = q
1
i
1
+ q
2
i
2
+ q
3
i
3
∈ I. Then q
2
= −1 if and only if q
2
1
+ q
2
2
+ q
2
3
= 1, so
the set of ‘quaternionic square-roots of minus-one’ is naturally isomorphic to the
2-sphere S
2
. We shall often identify these sets, writing ‘q ∈ S
2
’ as a shorthand for
‘q ∈ H : q
2
= −1’.
• If q ∈ S
2
then ¸1, q) is a subfield of H isomorphic to C. We shall call this subfield
C
q
.
4
1.1.1 A History of the Quaternions
The quaternions were discovered by the Irish mathematician and physicist, William
Rowan Hamilton (1805-1865),
1
whose contributions to mechanics are well-known and
widely used. By 1835 Hamilton had helped to win acceptance for the system of complex
numbers by showing that calculations with complex numbers are equivalent to calcula-
tions with ordered pairs of real numbers, governed by certain rules. At the time, complex
numbers were being applied very effectively to problems in the plane R
2
. To Hamilton,
the next logical step was to seek a similar 3-dimensional number system which would
revolutionise calculations in R
3
. For years, he struggled with this problem. In a touching
letter to his son [H1],
2
dated shortly before his death in 1865, Hamilton writes:
Every morning, on my coming down to breakfast, your brother and yourself
used to ask me: “Well, Papa, can you multiply triplets?” Whereto I was
always obliged to reply, with a sad shake of the head, “No, I can only add
and subtract them”.
For several years, Hamilton tried to manipulate the three symbols 1, i and j into an
algebra. He finally realised that the secret was to introduce a fourth dimension. On 16th
October, 1843, whilst walking with his wife, he had a flash of inspiration. In the same
letter, he writes:
An electric circuit seemed to close, and a spark flashed forth, the herald
(as I foresaw immediately) of many long years to come of definitely directed
thought and work ... I pulled out on the spot a pocket-book, which still
exists, and made an entry there and then. Nor could I resist the impulse —
unphilosophical as it may have been — to cut with a knife on the stone of
Brougham Bridge, as we passed it, the fundamental formula with the symbols
i, j, k:
i
2
= j
2
= k
2
= ijk = −1,
which contains the solution of the problem, but of course, as an inscription,
has long since mouldered away.
Substituting i
1
, i
2
, i
3
for i, j, k, this gives the quaternion relations (1.1).
Rarely do we possess such a clear account of the genesis of a piece of mathematics.
Most mathematical theories are invented gradually, and only after years of development
can they be presented in a lecture course as a definitive set of axioms and results. The
quaternions, on the other hand, “started into life, or light, full grown, on the 16th of
Ocober, 1843...”
3
“Less than an hour elapsed” before Hamilton obtained leave of the
Council of the Royal Irish Academy to read a paper on quaternions. The next day,
Hamilton wrote a detailed letter to his friend and fellow mathematician John T. Graves
1
Letters suggest that both Euler and Gauss were aware of the quaternion relations (1.1), though
neither of them published the disovery [EKR, p. 192].
2
Copies of Hamilton’s most significant letters and papers concerning quaternions are currently avail-
able on the internet at www.maths.tcd.ie/pub/HistMath/People/Hamilton
3
Letter to Professor P.G.Tait, an excerpt of which can be found on the same website as [H1].
5
[H2], giving us a clear account of the train of research which led him to his breakthrough.
The discovery was published within a month on the 13th of November [H3].
The timing of the discovery amplified its impact upon Hamilton and his followers.
The only other algebras known in 1843 were the real and complex numbers, both of
which can be regarded as subalgebras of the quaternions. (It was not until 1858 that
Cayley introduced matrices, and showed that the quaternion algebra could be realised
as a subalgebra of the complex-valued 2 2 matrices.) As a result, Hamilton became
the figurehead of a school of ‘quaternionists’, whose fervour for the new numbers far
exceeded their usefulness. Hamilton believed his discovery to be of similar importance
to that of the infinitesimal calculus, and devoted the rest of his career exclusively to its
study. Echos of this zeal could still be heard this century; for example, while Eamon
de Valera was President of Ireland (from 1959 to 1973), he would attend mathematical
meetings whenever their title contained the word ‘quaternions’ !
Such excesses were bound to provoke a reaction, especially as it became clear that
the quaternions are just one example of a number of possible algebras. Lord Kelvin, the
famous Scottish physicist, once remarked that “Quaternions came from Hamilton after
all his really good work had been done; and though beautifully ingenious, have been an
unmixed evil to those who have touched them in any way” [EKR, p.193]. A belief that
quaternions are somehow obsolete is often tacitly accepted to this day.
This is far from the case. The quaternions remain the simplest algebra after the
real and complex numbers. Indeed, the real numbers R, the complex numbers C and
the quaternions H are the only associative division algebras, as was proved by Georg
Frobenius in 1878: and amongst these the quaternions are the most general. The dis-
covery of the quaternions provided enormous stimulation to algebraic research and it is
thought that the term ‘associative’ was coined by Hamilton himself [H3, p.5] to describe
quaternionic behaviour. Investigation into the nature of and constraints imposed by
algebraic properties such as associativity and commutativity was greatly accelerated by
the discovery of the quaternions.
The quaternions themselves have been used in various areas of mathematics. Most
recently, quaternions have enjoyed prominence in computer science, because they are the
simplest algebraic tool for describing rotations in three and four dimensions. Certainly,
the numbers have fallen short of the early expectations of the quaternionists. However,
quaternions do shine a light on certain areas of mathematics, and those who become
familiar with them soon come to appreciate an intricacy and beauty which is all their
own.
1.1.2 Quaternions and Matrices
In this section we will make use of the older notation i = i
1
, j = i
2
, k = i
3
. This makes
it easy to interpret i as the standard complex ‘square root of −1’ and j as a ‘structure
map’ on the complex vector space C
2
.
It is well-known that the quaternions can be written as real or complex matrices,
because there are isomorphisms from H into subalgebras of Mat(4, R) and Mat(2, C).
The former of these is given by the mapping
6
q
0
+ q
1
i + q
2
j + q
3
k →
_
_
_
_
q
0
q
1
q
2
q
3
−q
1
q
0
−q
3
q
2
−q
2
q
3
q
0
−q
1
−q
3
−q
2
q
1
q
0
_
_
_
_
.
More commonly used is the mapping into Mat(2, C). We can write every quaternion
as a pair of complex numbers, using the equation
q
0
+ q
1
i + q
2
j + q
3
k = (q
0
+ q
1
i) + (q
2
+ q
3
i)j. (1.2)
In this way we obtain the expression q = α+βj ∈ H

= C
2
. The map j : α+βj →−
¯
β+¯ αj
is a conjugate-linear involution of C
2
with j
2
= −1. This identification H

= C
2
is not
uniquely determined: each q ∈ S
2
determines a similar isomorphism.
Having written this down, it is easy to form the map
ι : H →H ⊂ Mat(2, C) α + βj →
_
α β

¯
β ¯ α
_
. (1.3)
The quaternion algebra can thus be realised as a real subalgebra of Mat(2, C), using the
identifications
1 =
_
1 0
0 1
_
i
1
=
_
i 0
0 −i
_
i
2
=
_
0 1
−1 0
_
i
3
=
_
0 i
i 0
_
. (1.4)
Note that the squared norm q¯ q of a quaternion q is the same as the determinant
of the matrix ι(q) ∈ H. The isomorphism ι gives an easy way to deduce that H is an
associative division algebra; the inverse of any nonzero matrix A ∈ H is also in H, and
the only matrix in H whose determinant is zero is the zero matrix.
1.1.3 Simple Applications of the Quaternions
There are a number of ways in which quaternions can be used to express mathematical
ideas. In many cases, a quaternionic description prefigures more modern descriptions.
We will outline two main areas – vector analysis and Euclidean geometry. An excellent
and readable account of most of the following can be found in Chapter 7 of [EKR].
Every quaternion can be uniquely written as the sum of its real and imaginary parts.
If we identify the imaginary quaternions I with the real vector space R
3
, we can consider
each quaternion q = q
0
+q
1
i
1
+q
2
i
2
+q
3
i
3
as the sum of a scalar part q
0
and a vectorial part
(q
1
, q
2
, q
3
) ∈ R
3
(indeed, it is in this context that the term ‘vector’ first appears [H4]).
If we multiply together two imaginary quaternions p, q ∈ I, we obtain a quaternionic
version of the scalar product and vector product on R
3
, as follows:
pq = −p q + p ∧ q ∈ R ⊕I

= H. (1.5)
Surprising as it seems nowadays, it was not until the 1880’s that Josiah Willard Gibbs
(1839-1903), a professor at Yale University, argued that the scalar and vector products
7
had their own meaning, independent from their quaternionic origins. It was he who
introduced the familiar notation p q and p ∧ q — before his time these were written
‘−Spq’ and ‘V pq’, indicating the ‘scalar’ and ‘vector’ parts of the quaternionic product
pq ∈ H.
Another crucial part of vector analysis which originated with Hamilton and the
quaternions is the ‘nabla’ operator ∇ (so-called by Hamilton because the symbol ∇
is similar in shape to the Hebrew musical instrument of that name). The familiar gra-
dient operator acting on a real differentiable function f(x
1
, x
2
, x
3
) : R
3
→ R was first
written as
∇f :=
∂f
∂x
1
i
1
+
∂f
∂x
2
i
2
+
∂f
∂x
3
i
3
. (1.6)
Hamilton went on to consider applying the operator ∇ to a ‘differentiable quaternion
field’ F(x
1
, x
2
, x
3
) = f
1
(x
1
, x
2
, x
3
)i
1
+ f
2
(x
1
, x
2
, x
3
)i
2
+ f
3
(x
1
, x
2
, x
3
)i
3
, obtaining the
equation
∇F = −
_
∂f
1
∂x
1
+
∂f
2
∂x
2
+
∂f
3
∂x
3
_
+
_
∂f
3
∂x
2

∂f
2
∂x
3
_
i
1
+
_
∂f
1
∂x
3

∂f
3
∂x
1
_
i
2
+
_
∂f
2
∂x
1

∂f
1
∂x
2
_
i
3
,
which we recognise in modern terminology as
∇F = −div F + curl F.
Applying the ∇ operator to Equation (1.6) leads to the well-known Laplacian operator
on R
3
:

2
f = −
_

2
f
∂x
2
1
+

2
f
∂x
2
2
+

2
f
∂x
2
3
_
=: ∆f.
Having obtained the standard scalar product on R
3
, we can obtain the Euclidean
metric on R
4
in a similar fashion by relaxing the restriction in Equation (1.5) that p
and q should be imaginary. For any p, q ∈ H, we have
Re(p¯ q) = p
0
q
0
+ p
1
q
1
+ p
2
q
2
+ p
3
q
3
∈ R,
and so we define the canonical scalar product on H by
¸p, q) = Re(p¯ q). (1.7)
We define the norm of a quaternion q ∈ H in the obvious way, putting
[q[ =

q¯ q. (1.8)
The norm function is multiplicative, i.e. [pq[ = [p[[q[ for p, q ∈ H. As with complex
numbers, we can combine the norm function with conjugation to obtain the inverse of
q ∈ H ¸ ¦0¦ (it is easily verified that q
−1
= ¯ q/[q[
2
is the unique quaternion such that
qq
−1
= q
−1
q = 1).
Another quaternionic formula, similar to Equation (1.7), is the identity
Re(pq) = p
0
q
0
−p
1
q
1
−p
2
q
2
−p
3
q
3
∈ R. (1.9)
8
If we regard p and q as four-vectors in spacetime with the ‘time-axis’ identified with
R ⊂ H and the ‘spatial part’ identified with I, this is identical to the Lorentz metric of
special relativity.
Let a and b be quaternions of unit norm, and consider the involution f
a,b
: H →H
given by f
a,b
(q) = aqb. Then [f
a,b
(q)[ = [q[ and the function f
a,b
: R
4
→R
4
is a rotation.
Clearly, f
−a,−b
= f
a,b
, so each of these choices for the pair a, b gives the same rotation.
In 1855, Cayley showed that all rotations of R
4
can be written in this fashion and that
the two possibilities given above are the only two which give the same rotation. Also,
all reflections can be obtained by the involutions
¯
f
a,b
(q) = a¯ qb.
One of the beauties of this system is that having obtained all rotations of R
4
, we
obtain the rotations of R
3
simply by putting a = b
−1
, giving the inner automorphism
of H, q → aqa
−1
. (By Cayley’s theorem, all real-algebra automorphisms of H take this
form.) This fixes the real line R and rotates the imaginary quaternions I. Identifying I
with R
3
, the map q → aqa
−1
is a rotation of R
3
. According to Cayley, within a year
of the discovery of the quaternions Hamilton was aware that all rotations of R
3
can be
expressed in this fashion.
These discoveries provided much insight into the classical groups SO(3) and SO(4),
and helped to develop our knowledge of transformation groups in general. There are
interesting questions which arise. Why do the unit quaternions turn out to be so im-
portant? In view of the quaternionic version of the Lorentz metric in Equation (1.9),
can we use quaternions to write Lorentz tranformations as elegantly? Are there other
spaces which might lend themselves to quaternionic treatment? These questions are best
addressed using the theory of Lie groups, which the pioneering work of Hamilton and
Cayley helped to develop.
1.2 The Lie Group Sp(1) and its Representations
The unit quaternions form a subgroup of H under multiplication, which we call Sp(1).
Its importance to the quaternions is equivalent to that of the circle group U(1) of unit
complex numbers in complex analysis. As a manifold Sp(1) is the 3-sphere S
3
. The
multiplication and inverse maps are smooth, so Sp(1) is a compact Lie group. The
isomorphism ι : H →H ⊂ Mat(2, C) of Equation (1.4) maps Sp(1) isomorphically onto
SU(2).
The Lie algebra sp(1) of Sp(1) is generated by the elements I, J and K, the Lie
bracket being given by the relations [I, J] = 2K, [J, K] = 2I and [K, I] = 2J. We can
write these using the matrices of Equation (1.4):
I =
_
i 0
0 −i
_
J =
_
0 1
−1 0
_
K =
_
0 i
i 0
_
.
The complexification of sp(1) is the Lie algebra sl(2, C). This is generated over C by
the elements
H =
_
1 0
0 −1
_
X =
_
0 1
0 0
_
Y =
_
0 0
1 0
_
, (1.10)
and the relations
[H, X] = 2X, [H, Y ] = −2Y and [X, Y ] = H. (1.11)
9
Hence the equations
I = iH, J = X −Y and K = i(X + Y ) (1.12)
give one possible identification sp(1) ⊗
R
C

= sl(2, C).
For any Q ∈ sp(1), the normaliser N(Q) of Q is defined to be
N(Q) = ¦P ∈ sp(1) : [P, Q] = 0¦ = ¸Q),
which is a Cartan subalgebra of sp(1). Identifying a unit vector Q = aI+bJ+cK ∈ sp(1)
with the corresponding imaginary quaternion q = ai
2
+ bi
2
+ ci
3
∈ S
2
, the exponential
map exp : sp(1) →Sp(1) maps ¸Q) to the unit circle in C
q
, which we will call U(1)
q
.
In the previous section it was shown that rotations in three and four dimensions can
be written in terms of unit quaternions. This is because there is a commutative diagram
of Lie group homomorphisms
Sp(1)

= Spin(3) → Sp(1) Sp(1)

= Spin(4)
↓ ↓
SO(3) → SO(4).
(1.13)
The horizontal arrows are inclusions, and the vertical arrows are 2 : 1 coverings with
kernels ¦1, −1¦ and ¦(1, 1), (−1, −1)¦ respectively. The applications of these homomor-
phisms in Riemannian geometry are described by Salamon in [S1].
From this, we can see clearly why three- and four-dimensional Euclidean geometry
fit so well in a quaternionic framework. We can also see why the same is not true for
Lorentzian geometry. Here the important group is the Lorentz group SO(3,1). Whilst
there is a double cover SL(2, C) →SO
0
(3, 1), this does not restrict to a suitably interest-
ing real homomorphism Sp(1)→SO
0
(3,1). It is possible to write Lorentz transformations
using quaternions (for a modern example see [dL]), but the author has found no way
which is simple enough to be really effective. There have been attempts to use the bi-
quaternions ¦p + iq : p + q ∈ H¦

= H ⊗
R
C to apply quaternions to special relativity
[Sy], but the mental gymnastics involved in using four ‘square roots of −1’ are cum-
bersome: the equivalent description of ‘spin transformations’ using the matrices of the
group SL(2,C) are a more familiar and fertile ground.
1.2.1 The Representations of Sp(1)
A representation of a Lie group G on a vector space V is a Lie group homomorphism
ρ : G →GL(V ). We will sometimes refer to V itself as a representation where the map
ρ is understood. The differential dρ is a Lie algebra representation dρ : g →End(V ).
The representations of Sp(1)

= SU(2) will play an important part in this thesis.
Their theory is well-known and used in many situations. Good introductions to Sp(1)-
representations include [BD, ¸2.5] (which describes the action of the group SU(2)) and
[FH, Lecture 11] (which provides a description in terms of the action of the Lie algebra
sl(2, C)). We recall those points which will be of particular importance.
Because it is a compact group, every representation of Sp(1) on a complex vector
space V can be written as a direct sum of irreducible representations. The multiplicity
10
of each irreducible in such a decomposition is uniquely determined. Moreover, there is a
unique irreducible representation on C
n
for every n > 0. This makes the representations
of Sp(1) particularly easy to describe. Let V
1
be the basic representation of SL(2,C)
on C
2
given by left-action of matrices upon column vectors. This coincides with the
basic representation of Sp(1) by left-multiplication on H

= C
2
. The unique irreducible
representation on C
n+1
is then given by the n
th
symmetric power of V
1
, so we define
V
n
= S
n
(V
1
).
The representation V
n
is irreducible [BD, Proposition 5.1], and every irreducible rep-
resentation of Sp(1) is of the form V
n
for some nonnegative n ∈ Z [BD, Proposition
5.3].
Let x and y be a basis for C
2
, so that V
1
= ¸x, y). Then
V
n
= S
n
(V
1
) = ¸x
n
, x
n−1
y, x
n−2
y
2
, . . . , x
2
y
n−2
, xy
n−1
, y
n
).
The action of SL(2,C) on V
n
is given by the induced action on the space of homogeneous
polynomials of degree n in the variables x and y.
Each of the Lie group representations V
n
is a representation of the Lie algebras sl(2, C)
and sp(1). Another very important way to describe the structure of these representations
is obtained by decomposing them into weight spaces (eigenspaces for the action of a
Cartan subalgebra). In terms of H, X and Y , the sl(2, C)-action on V
1
is given by
H(x) = x X(x) = 0 Y (x) = y
H(y) = −y X(y) = x Y (y) = 0.
(1.14)
To obtain the induced action of sl(2, C) on V
n
we use the Leibniz rule
4
A(a b) =
A(a) b + a A(b). This gives
H(x
n−k
y
k
) = (n −2k)(x
n−k
y
k
)
X(x
n−k
y
k
) = k(x
n−k+1
y
k−1
) (1.15)
Y (x
n−k
y
k
) = (n −k)(x
n−k−1
y
k+1
).
Each subspace ¸x
n−k
y
k
) ⊂ V
n
is therefore a weight space of the representation V
n
, and
the weights are the integers
¦n, n −2, . . . , n −2k, . . . , 2 −n, −n¦.
Thus V
n
is also characterised by being the unique irreducible representation of sl(2, C)
with highest weight n. We can compute the action of sp(1) on V
n
by substituting I, J
and K for H, X and Y using the relations of (1.12).
Another important operator which acts on an sp(1)-representation is the Casimir
operator C = I
2
+ J
2
+ K
2
= −(H
2
+ 2XY + 2Y X). It is easy to show that for any
x
n−k
y
k
∈ V
n
,
C(x
n−k
y
k
) = −n(n + 2)x
n−k
y
k
. (1.16)
4
This can be found in [FH, p. 110], which describes the action of Lie groups and Lie algebras on
tensor products.
11
Each irreducible representation V
n
is thus an eigenspace of the Casimir operator with
eigenvalue −n(n + 2).
Let V
m
and V
n
be two Sp(1)-representations. Then their tensor product V
m
⊗ V
n
is naturally an Sp(1) Sp(1)-representation,
5
and also a representation of the diagonal
Sp(1)-subgroup, the action of which is given by
g(u ⊗v) = g(u) ⊗g(v).
The irreducible decomposition of the diagonal Sp(1)-representation on V
m
⊗V
n
is given
by the famous Clebsch-Gordon formula,
V
m
⊗V
n

= V
m+n
⊕V
m+n−2
⊕ ⊕V
m−n+2
⊕V
m−n
for m ≥ n. (1.17)
This can be proved using characters [BD, Proposition 5.5] or weights [FH, Exercise
11.11].
Real and quaternionic representations
It is standard practice to work primarily with representations on complex vector spaces.
Representations on real (and quaternionic) vector spaces are obtained using antilinear
structure maps. A thorough guide to this process is in [BD, ¸2.6]. In the case of Sp(1)-
representations, we define the structure map σ
1
: V
1
→V
1
by
σ
1
(z
1
x + z
2
y) = −¯ z
2
x + ¯ z
1
y z
1
, z
2
∈ C.
Then σ
2
1
= −1 , and σ
1
coincides with the map j of Section 1.1.2. Let σ
n
be the map
which σ
1
induces on V
n
, i.e.
σ
n
(z
1
x
n−k
y
k
) = (−1)
k
¯ z
1
x
k
y
n−k
. (1.18)
If n = 2m is even then σ
2
2m
= 1 and σ
2m
is a real structure on V
2m
. Let V
σ
2m
be the set
of fixed-points of σ
2m
. Then V
σ
2m

= R
2m+1
is preserved by the action of Sp(1), and
V
σ
2m

R
C

= V
2m
.
Thus V
σ
2m
is a representation of Sp(1) on the real vector space R
2m+1
.
If on the other hand n = 2m− 1 is odd, σ
2
2m−1
= −1. Then Sp(1) acts on the un-
derlying real vector space R
4m
. This real vector space comes equipped with the complex
structure i and the structure map σ
2m−1
, in such a way that the subspace
¸v, iv, σ
2m−1
(v), iσ
2m−1
(v))

= R
4
is isomorphic to the quaternions; thus V
2m−1

= H
m
. This is why a complex antilinear
map σ on a complex vector space V such that σ
2
= −1 is called a quaternionic structure.
5
Since Sp(1) Sp(1)

= Spin(4), we can construct all Spin(4) and hence all SO(4) representations in
this fashion — see [S1, ¸3].
12
1.3 Difficulties with the Quaternions
Quaternions are far less predictable than their lower-dimensional cousins, due to the great
complicating factor of non-commutativity. Many of the ideas which work beautifully for
real or complex numbers are not suited to quaternions, and attempts to use them often
result in lengthy and cumbersome mathematics — most of which dates from the 19
th
century and is now almost forgotten. However, with modesty and care, quaternions can
be used to recreate many of the structures over the real and complex numbers with which
we are familiar. The purpose of this section is to outline some of the difficulties with a
few examples, which highlight the need for caution: but also, it is hoped, point the way
to some of the successes we will encounter.
1.3.1 The Fundamental Theorem of Algebra for Quaternions
The single biggest reason for using the complex numbers in preference to any other num-
ber field is the so-called ‘Fundamental Theorem of Algebra’ — every complex polynomial
of degree n has precisely n zeros, counted with multiplicities. The real numbers are not
so well-behaved: a real polynomial of degree n can have fewer than n real roots.
Quaternion behaviour is many degrees freer and less predictable. If we multiply
together two ‘linear factors’, we obtain the following expression:
(a
1
X + b
1
)(a
2
X + b
2
) = a
1
Xa
2
X + a
1
Xb
2
+ b
1
a
2
X + b
1
b
2
.
It quickly becomes obvious that non-commutativity is going make any attempt to fac-
torise a general polynomial extremely troublesome.
Moreover, there are many polynomials which display extreme behaviour. For ex-
ample, the cubic X
2
i
1
Xi
1
+ i
1
X
2
i
1
X − i
1
Xi
1
X
2
− Xi
1
X
2
i
1
takes the value zero for
all X ∈ H. At the other extreme, since i
1
X − Xi
1
∈ I for all X ∈ H, the equation
i
1
X −Xi
1
+ 1 = 0 has no solutions at all!
In order to arrive at any kind of ‘fundamental theorem of algebra’, we need to restrict
our attention considerably. We define a monomial of degree n to be an expression of the
form
a
0
Xa
1
Xa
2
a
n−1
Xa
n
, a
i
∈ H¸ ¦0¦.
Then there is the following ‘fundamental theorem of algebra for quaternions’:
Theorem 1.3.1 [EKR, p. 205] Let f be a polynomial over H of degree n > 0 of the
form m + g, where m is a monomial of degree n and g is a polynomial of degree < n.
Then the mapping f : H →H is surjective, and in particular f has zeros in H.
This is typical of the difficulties we encounter with quaternions. There is a quater-
nionic analogue of the theorems for real and complex numbers, but because of non-
commutativity the quaternionic version is more complicated, less general and because of
this less useful.
13
1.3.2 Calculus with Quaternions
The monumental successes of complex analysis make it natural to look for a similar
theory for quaternions. Complex analysis can be described as the study of holomorphic
functions. A function f : C → C is holomorphic if it has a well-defined complex
derivative. One of the fundamental results in complex analysis is that every holomorphic
function is analytic i.e. can be written as a convergent power series.
Sadly, neither of these definitions proves interesting when applied to quaternions —
the former is too restrictive and the latter too general. For a function f : H → H to
have a well-defined derivative
df
dq
= lim
h→0
(f(q + h) −f(q))h
−1
,
it can be shown [Su, ¸3, Theorem 1] that f must take the form f(q) = a +bq for some
a, b ∈ H, so the only functions which are quaternion-differentiable in this sense are affine
linear.
In contrast to the complex case, the components of a quaternion can be written as
quaternionic polynomials, i.e. for q = q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
,
q
0
=
1
4
(q −i
1
qi
1
−i
2
qi
2
−i
3
qi
3
) q
1
=
1
4i
1
(q −i
1
qi
1
+ i
2
qi
2
+ i
3
qi
3
) etc. (1.19)
This takes us to the other extreme: the study of quaternionic power series is the same
as the theory of real-analytic functions on R
4
. Hamilton and his followers were aware
of this — it was in Hamilton’s work on quaternions that some of the modern ideas in
the theory of functions of several real variables first appeared. Despite the benefits of
this work to mathematics as a whole, Hamilton never developed a successful ‘quaternion
calculus’.
It was not until the work of R. Fueter, in 1935, that a suitable definition of a ‘regular
quaternionic function’ was found, using a quaternionic analogue of the Cauchy-Riemann
equations. A regular function on H is defined to be a real-differentiable function f :
H →H which satisfies the Cauchy-Riemann-Fueter equations:
∂f
∂q
0
+
∂f
∂q
1
i
1
+
∂f
∂q
2
i
2
+
∂f
∂q
3
i
3
= 0. (1.20)
The theory of quaternionic analysis which results is modestly successful, though it is
little-known. The work of Fueter is described and extended in the papers of Deavours
[D] and Sudbery [Su], which include quaternionic versions of Morera’s theorem, Cauchy’s
theorem and the Cauchy integral formula.
As usual, non-commutativity causes immediate algebraic difficulties. If f and g are
regular functions, it is easy to see that their sum f + g must also be regular, as must
the left scalar multiple qf for q ∈ H; but their product fg, the composition f ◦ g and
the right scalar multiple fq need not be. Hence regular functions form a left H-module,
but it is difficult to see how to describe any further algebraic structure.
1.3.3 Quaternion Linear Algebra
In the same way as we define real or complex vector spaces, we can define quaternionic
vector spaces or H-modules — a real vector space with an H-action, which we shall call
14
scalar multiplication. There is the added complication that we need to say whether this
multiplication is on the left or the right. We will work with left H-modules — this choice
is arbitrary and has only notational effects on the resulting theory. A left H-module is
thus a real vector space U with an action of H on the left, which we write (q, u) →q u
or just qu, such that p(qu) = (pq)u for all p, q ∈ H and u ∈ U. For example, H
n
is an
H-module with the obvious left-multiplication.
Several of the familiar ideas which work for vector spaces over a commutative field
work just as well for H-modules. For example, we can define quaternion linear maps
between H-modules in the obvious way, and so we can define a dual H-module U
×
of
quaternion linear maps α : U →H.
However, if we try to define quaternion bilinear maps we run into trouble. If A, B and
C are vector spaces over the commutative field F, then an F-bilinear map µ : AB →C
satisfies µ(f
1
a, b) = f
1
µ(a, b) and µ(a, f
2
b) = f
2
µ(a, b) for all a ∈ A, b ∈ B, f
1
, f
2
∈ F.
This is equivalent to having µ(f
1
a, f
2
b) = f
1
f
2
µ(a, b).
Now suppose that U, V and W are (left) H-modules. Let q
1
, q
2
∈ H and let u ∈ U,
v ∈ V . If we try to define a bilinear map µ : U V →W as above, then we need both
µ(q
1
u, q
2
v) = q
1
µ(u, q
2
v) = q
2
q
1
µ(u, v) and µ(q
1
u, q
2
v) = q
2
µ(q
1
u, v) = q
1
q
2
µ(u, v).
Since q
1
q
2
,= q
2
q
1
in general, this cannot work.
Similar difficulties arise if we try to define a tensor product over the quaternions. If
A and B are vector spaces over the commutative field F, their tensor product over F is
defined by
A ⊗
F
B =
F(A, B)
R(A, B)
, (1.21)
where F(A, B) is the vector space freely generated (over F) by all elements (a, b) ∈ AB
and R(A, B) is the subspace of F(A, B) generated by all elements of the form
(a
1
+ a
2
, b) −(a
1
, b) −(a
2
, b) (a, b
1
+ b
2
) −(a, b
1
) −(a, b
2
)
f(a, b) −(f(a), b) and f(a, b) −(a, f(b)),
for a, a
j
∈ A, b, b
j
∈ B and f ∈ F. Not surprisingly, this process does not work for
quaternions. Let U and V be left H-modules. Then if we define the ideal R(U, V ) in
the same way as above, we discover that R(U, V ) is equal to the whole of F(U, V ), so
U ⊗
F
V = ¦0¦.
One way around both of these difficulties is to demand that our H-modules should
also have an H-action on the right. We can now define an H-bilinear map to be one which
satisfies µ(q
1
u, vq
2
) = q
1
µ(u, v)q
2
, or alternatively µ(q
1
u, q
2
v) = q
1
µ(u, v)¯ q
2
. Similarly,
for our tensor product we can define a generator uq ⊗v −u ⊗qv (replacing f(u) ⊗v −
u ⊗f(v)) for R. In this (more restricted) case we do obtain a well-defined ‘quaternionic
tensor product’ U⊗
H
V = F(U, V )/R(U, V ) which inherits a left H-action from U and a
right H-action from V .
The drawback with this system is that it does not really provide any new insights.
If we insist on having a left and a right H-action, we restrict ourselves to talking about
H-modules of the form H
n
= H ⊗
R
R
n
. In this case, we do get the useful relation
H
m

H
H
n

= H
mn
, but this is only saying that H⊗
R
R
m

R
R
n

= H⊗
R
R
mn
. In this context,
our ‘quaternionic tensor product’ is merely a real tensor product in a quaternionic setting.
15
1.3.4 Summary
By now we have become familiar with some of the more elementary ups and downs of the
quaternions. In the 20
th
century they have often been viewed as a sort of mathematical
Cinderella, more recent techniques being used to describe phenomena which were first
thought to be profoundly quaternionic. However, we have seen that it is possible to
produce quaternionic analogues of some of the most basic algebraic and analytic ideas of
real and complex numbers, often with interesting and useful results. In the next chapter
we will continue to explore this process, turning our attention to quaternionic structures
in differential geometry.
16
Chapter 2
Quaternionic Differential Geometry
In this chapter we review some of the ways in which quaternions are used to define
geometric stuctures on differentiable manifolds. In the same way that real and complex
manifolds are modelled locally by the vector spaces R
n
and C
n
respectively, there are
manifolds which can be modelled locally by the H-module H
n
. These models work by
defining tensors whose action on the tangent spaces to a manifold is the same as the
action of the quaternions on an H-module. There are two important classes of manifolds
which we shall consider: those which are called ‘quaternionic manifolds’, and a more
restricted class called ‘hypercomplex manifolds’.
In this chapter we describe these important geometric structures. We also review
the decomposition of exterior forms on complex manifolds, and examine some of the
parallels of this theory which have already been found in quaternionic geometry. Many
of the algebraic and geometric foundations of the material in this chapter are collected
in (or can be inferred from) Fujiki’s comprehensive article [F].
2.1 Complex, Hypercomplex and Quaternionic Man-
ifolds
Complex Manifolds
A complex manifold is a 2n-dimensional real manifold M which admits an atlas of
complex charts, all of whose transition functions are holomorphic maps from C
n
to
itself. As we saw in Section 1.3, the simplest notions of a ‘quaternion-differentiable map
from H to itself’ are either very restrictive or too general, and the ‘regular functions’ of
Fueter and Sudbery are not necessarily closed under composition. This makes the notion
of ‘quaternion-differentiable transition functions’ less interesting that one might hope.
An equivalent way to define a complex manifold is by the existence of a special
tensor called a complex structure. A complex structure on a real vector space V is an
automorphism I : V → V such that I
2
= −id
V
. (It follows that dimV is even.) The
complex structure I gives an isomorphism V

= C
n
, since the operation ‘multiplication
by i’ defines a standard complex structure on C
n
. An almost complex structure on a
2n-dimensional real manifold M is a smooth tensor I ∈ C

(End(TM)) such that I is a
complex structure on each of the fibres T
x
M.
17
Now, if M is a complex manifold, each tangent space T
x
M is isomorphic to C
n
, so
taking I to be the standard complex structure on each T
x
M defines an almost complex
structure on M. An almost complex stucture I which arises in this way is called a
complex structure on M, in which case I is said to be integrable. The famous Newlander-
Nirenberg theorem states that an almost complex structure I is integrable if and only if
the Nijenhuis tensor of I
N
I
(X, Y ) = [X, Y ] + I[IX, Y ] + I[X, IY ] −[IX, IY ]
vanishes for all X, Y ∈ C

(TM), for all x ∈ M. The Nijenhuis tensor N
I
measures the
(0, 1)-component of the Lie bracket of two vector fields of type (1, 0).
1
On a complex
manifold the Lie bracket of two (1, 0)-vector fields must also be of type (1, 0).
Thus if I is an almost complex structure on M and N
I
≡ 0, then M is a complex
manifold in the sense of the first definition given above. We can talk about the complex
manifold (M, I) if we wish to make the extra geometric structure more explicit —
especially as the manifold M might admit more than one complex structure.
Hypercomplex Manifolds
This way of defining a complex manifold adapts itself well to the quaternions. A hyper-
complex structure on a real vector space V is a triple (I, J, K) of complex structures on
V satisfying the equation IJ = K. (It follows that dimV is divisible by 4.) If we iden-
tify I, J and K with left-multiplication by i
1
, i
2
and i
3
, a hypercomplex structure gives
an isomorphism V

= H
n
. Equivalently, a hypercomplex structure is defined by a pair of
complex structures I and J with IJ = −JI. It is easy to see that if (I, J, K) is a hyper-
complex structure on V , then each element of the set ¦aI+bJ+cK : a
2
+b
2
+c
2
= 1¦

= S
2
is also a complex structure. We arrive at the following quaternionic version of a complex
manifold:
Definition 2.1.1 An almost hypercomplex structure on a 4n-dimensional manifold M is
a triple (I, J, K) of almost complex structures on M which satisfy the relation IJ = K.
If all of the complex structures are integrable then (I, J, K) is called a hypercomplex
structure on M, and M is a hypercomplex manifold.
A hypercomplex structure on M defines an isomorphism T
x
M

= H
n
at each point
x ∈ M. As on a vector space, a hypercomplex structure on a manifold M defines a
2-sphere S
2
of complex structures on M.
Some choices are inherent in this standard definition. A hypercomplex structure as
defined above gives TM the structure of a left H-module, since IJ = K. This induces
a right H-module structure on T

M, using the standard definition ¸I(ξ), X) = ¸ξ, I(X))
etc., for all ξ ∈ T

M, X ∈ TM, since on T

M we now have
¸ξ, IJ(X)) = ¸I(ξ), J(X)) = ¸JI(ξ), X).
In this thesis we will make more use of the hypercomplex structure on T

M than that on
TM. Because of this we will usually define our hypercomplex structures so that IJ = K
on T

M rather than TM. This has only notational effect on the theory, but it does pay to
1
Tensors of type (p, q) are defined in the next section.
18
be aware of how it affects other standard notations. For example, for us the hypercomplex
structure acts trivially on the anti-self-dual 2-forms ω

1
= dx
0
∧ dx
1
− dx
2
∧ dx
3
etc.
This encourages us to think of a connection whose curvature is acted upon trivially by
the hypercomplex structure as anti-self-dual rather than self-dual, whereas some authors
use the opposite convention. Such things are largely a matter of taste — we are choosing
to follow the notation of Joyce [J1] for whom regular functions form a left H-module,
whereas Sudbery’s regular functions form a right H-module.
One important difference between complex and hypercomplex geometry is the exis-
tence of a special connection. A complex manifold generally admits many torsion-free
connections which preserve the complex structure. By contrast, on a hypercomplex
manifold there is a unique torsion-free connection ∇ such that
∇I = ∇J = ∇K = 0.
This was proved by Obata in 1956, and the connection ∇ is called the Obata connection.
Complex and hypercomplex manifolds can be decribed succinctly in terms of G-
structures on manifolds. Let P be the principal frame bundle of M, i.e. the GL(n, R)-
bundle whose fibre over x ∈ M is the group of isomorphisms T
x
M

= R
4n
. Let G be a
Lie subgroup of GL(n, R). A G-structure Q on M is a principal subbundle of P with
structure group G.
Suppose M
2n
has an almost complex structure. The group of automorphisms of T
x
M
preserving such a structure is isomorphic to GL(n, C). Thus an almost complex structure
I and a GL(n, C)-structure Q on M contain the same information. The bundle Q admits
a torsion-free connection if and only if there is a torsion-free linear connection ∇ on M
with ∇I = 0, in which case it is easy to show that I is integrable. Thus a complex
manifold is precisely a real manifold M
2n
with a GL(n, C)-structure Q admitting a
torsion-free connection (in which case Q itself is said to be ‘integrable’).
If M
4n
has an almost hypercomplex structure then the group of automorphisms
preserving this structure is isomorphic to GL(n, H). Following the same line of argument,
a hypercomplex manifold is seen to be a real manifold M with an integrable GL(n, H)-
structure Q. The uniqueness of any torsion-free connection on Q follows from analysing
the Lie algebra gl(n, H). This process is described in [S3, ¸6].
Quaternionic Manifolds and the Structure Group GL(1, H)GL(n, H)
Not all of the manifolds which we wish to describe as ‘quaternionic’ admit hypercomplex
structures. For example, the quaternionic projective line HP
1
is diffeomorphic to the
4-sphere S
4
. It is well-known that S
4
does not even admit a global almost complex
structure; so HP
1
can certainly not be hypercomplex, despite behaving extremely like
the quaternions locally.
The reason (and the solution) for this difficulty is that GL(n, H) is not the largest
subgroup of GL(4n, R) preserving a quaternionic structure. If we think of GL(n, H)
as acting on H
n
by right-multiplication by n n quaternionic matrices, then the action
of GL(n, H) commutes with that of the left H-action of the group GL(1, H). Thus
the group of symmetries of H
n
is the product GL(1, H)
R

GL(n, H) , which we write
GL(1, H)GL(n, H). We can multiply on the right by any real multiple of the identity
19
(since GL(1, H) and GL(n, H) share the same centre R

), so without loss of gener-
ality we can reduce the first factor to Sp(1). Thus GL(1, H)GL(n, H) is the same as
Sp(1)GL(n, H) = Sp(1)
Z
2
GL(n, H).
Definition 2.1.2 [S3, 1.1] A quaternionic manifold is a 4n-dimensional real manifold M
(n ≥ 2) with an Sp(1)GL(n, H)-structure Q admitting a torsion-free connection.
When n = 1 the situation is different, since Sp(1)Sp(1)

= SO(4). In four dimensions
we make the special definition that a quaternionic manifold is a self-dual conformal
manifold.
In terms of tensors, quaternionic manifolds are a generalisation of hypercomplex
manifolds in the following way. Each tangent space T
x
M still admits a hypercomplex
structure giving an isomorphism T
x
M

= H
n
, but this isomorphism does not necessarily
arise from globally defined complex structures on M. There is still an S
2
-bundle of
local almost-complex structures which satisfy IJ = K, but it is free to ‘rotate’. For a
comprehensive study of quaternionic manifolds see [S3] and Chapter 9 of [S4].
Riemannian Manifolds in Complex and Quaternionic Geometry
Suppose the GL(n, C)-structure Q on a complex manifold M admits a further reduc-
tion to an integrable U(n)-structure Q

. Then M admits a Riemannian metric g with
g(IX, IY ) = g(X, Y ) for all X, Y ∈ T
x
M for all x ∈ M. We also define the differentiable
2-form ω(X, Y ) = g(IX, Y ). If such a metric arises from an integrable U(n)-structure
Q

then ω will be a closed 2-form, and M is a symplectic manifold — so M has com-
patible complex and symplectic structures. In this case M is called a K¨ahler manifold;
an integrable U(n)-structure is called a K¨ahler structure; the metric g is called a K¨ahler
metric and the symplectic form ω is called a K¨ahler form.
The quaternionic analogue of the compact group U(n) is the group Sp(n). A hyper-
complex manifold whose GL(n, H)-structure Q reduces to an integrable Sp(n)-structure
Q

admits a metric g which is simultaneously K¨ahler for each of the complex structures
I, J and K. Such a manifold is called hyperk¨ahler . Using each of the complex structures,
we define three independent symplectic forms ω
I
, ω
J
and ω
K
. Then the complex 2-form
ω
J
+ iω
K
is holomorphic with respect to the complex structure I, and a hyperk¨ahler
manifold has compatible hypercomplex and complex-symplectic structures. Hyperk¨ahler
manifolds are studied in [HKLR], which gives a quotient construction for hyperk¨ ahler
manifolds.
Similarly, if a quaternionic manifold has a metric compatible with the torsion-free
Sp(1)GL(n, H)-structure, then the Sp(1)GL(n, H)-structure Q reduces to an Sp(1)Sp(n)-
structure Q

and M is said to be quaternionic K¨ahler. The group Sp(1)Sp(n) is a
maximal proper subgroup of SO(4n) except when n = 1, where as we know Sp(1)Sp(1) =
SO(4). In four dimensions a manifold is said to be quaternionic K¨ahler if and only if it
is self-dual and Einstein. Quaternionic K¨ahler manifolds are the subject of [S2].
2.2 Differential Forms on Complex Manifolds
This section consists of background material in complex geometry, especially ideas which
encourage the development of interesting quaternionic versions. More information on this
20
and other aspects of complex geometry can be found in [W] and [GH, ¸0.2].
Let (M, I) be a complex manifold. Then I gives TM and T

M the structure of a
U(1)-representation and the complexification C⊗
R
TM splits into two weight spaces with
weights ±i. The same is true of C ⊗
R
T

M. There are various ways of writing these
weight spaces; fairly standard is the notation C ⊗
R
T

M = T

1,0
M ⊕T

0,1
M, where these
summands are the +i and −i eigenspaces of I respectively. However, for our purposes
it will be more useful to adopt the notation of [S3], writing
C ⊗
R
T

M = Λ
1,0
M ⊕Λ
0,1
M, (2.1)
so Λ
1,0
M = T

1,0
M and Λ
0,1
M = T

0,1
M. A holomorphic function f ∈ C

(M, C) is
a smooth function whose derivative takes values only in Λ
1,0
M for all m ∈ M, i.e. f
is holomorphic if and only if df ∈ C

(M, Λ
1,0
M). A closely linked statement is that
Λ
1,0
M is a holomorphic vector bundle. Thus Λ
1,0
M is called the holomorphic cotangent
space and Λ
0,1
M is called the antiholomorphic cotangent space of M. If we reverse the
sense of the complex structure (i.e. if we swap I for −I) then we reverse the roles of the
holomorphic and antiholomorphic spaces, which is why a function which is holomorphic
with respect to I is antiholomorphic with respect to −I.
From standard multilinear algebra, the decomposition (2.1) gives rise to a decompo-
sition of exterior forms of all powers
C ⊗
R
Λ
k
T

M =
k

p=0
Λ
p
(T

1,0
M) ⊗Λ
k−p
(T

0,1
M).
Define the bundle
Λ
p,q
M = Λ
p
T

1,0
M ⊗Λ
q
T

0,1
M. (2.2)
With this notation, Equation (2.1) is an example of the more general decomposition into
types
C ⊗
R
Λ
k
T

M =

p+q=k
Λ
p,q
M. (2.3)
A smooth section of the bundle Λ
p,q
M is called a differential form of type (p, q) or just
a (p, q)-form. We write Ω
p,q
(M) for the set of (p, q)-forms on M, so

p,q
(M) = C

(M, Λ
p,q
M) and Ω
k
(M) =

p+q=k

p,q
(M).
Define two first-order differential operators,
∂ : Ω
p,q
(M) →Ω
p+1,q
(M)
∂ = π
p+1,q
◦ d
and
∂ : Ω
p,q
(M) →Ω
p,q+1
(M)
∂ = π
p,q+1
◦ d,
(2.4)
where π
p,q
denotes the natural projection from C ⊗ Λ
k
M onto Λ
p,q
M. The operator ∂
is called the Dolbeault operator.
These definitions rely only on the fact that I is an almost complex structure. A
crucial fact is that if I is integrable, these are the only two components represented in
21
Figure 2.1: The Dolbeault Complex

0,0
= C

(M, C)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj



1,0
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj



0,1
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj



1,1
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj



2,0
r
r
r
r
r
rj ∂
. . . . . . etc.

0,2
¨
¨
¨
¨
¨
¨B

. . . . . . etc.
the exterior differential d, i.e. d = ∂ +∂ [WW, Proposition 2.2.2, p.105]. An immediate
consequence of this is that on a complex manifold M,

2
= ∂∂ + ∂∂ = ∂
2
= 0. (2.5)
This gives rise to the Dolbeault complex. Writing ∂
p,q
for the particular map ∂ :

p,q
→Ω
p,q+1
, we define the Dolbeault cohomology groups
H
p,q

=
Ker(∂
p,q
)
Im(∂
p,q−1
)
.
A function f ∈ C

(M, C) is holomorphic if and only if ∂f = 0 and for this reason ∂ is
sometimes called the Cauchy-Riemann operator. Similarly, a (p, 0)-form α is said to be
holomorphic if and only if ∂α = 0.
A useful way to think of the Dolbeault complex is as a decomposition of C⊗Λ
k
T

M
into types of U(1)-representation. The representations of U(1) on complex vector spaces
are particularly easy to understand. Since U(1) is abelian, the irreducible representations
all are one-dimensional. They are parametrised by the integers, taking the form

n
: U(1) →GL(1, C) = C


n
: e

→e
niθ
for some n ∈ Z. Representations of the Lie algebra u(1) are then of the form d
n
: z →
nz. The Dolbeault complex begins with the representation of the complex structure
¸I)

= u(1) on C ⊗ T

M. This induces a representation on C ⊗ Λ

T

M. It is easy
to see that if ω ∈ Λ
p,q
M, I(ω) = i(p − q)ω. In other words, Λ
p,q
M is the bundle of
U(1)-representations of the type
p−q
.
22
2.3 Differential Forms on Quaternionic Manifolds
Any G-structure on a manifold M induces a representation of G on the exterior algebra
of M. Fujiki’s account of this [F, ¸2] explains many quaternionic analogues of complex
and K¨ahler geometry.
The decomposition of differential forms on quaternionic K¨ahler manifolds began by
considering the fundamental 4-form
Ω = ω
I
∧ ω
I
+ ω
J
∧ ω
J
+ ω
K
∧ ω
K
,
where ω
I
, ω
J
and ω
K
are the local K¨ahler forms associated to local almost complex
structures I, J and K with IJ = K. The fundamental 4-form is globally defined and
invariant under the induced action of Sp(1)Sp(n) on Λ
4
T

M. Kraines [Kr] and Bonan
[Bon] used the fundamental 4-form to decompose the space Λ
k
T

M in a similar way to
the Lefschetz decomposition of differential forms on a K¨ahler manifold [GH, p. 122]. A
differential k-form µ is said to be effective if Ω∧∗µ = 0, where ∗ : Λ
k
T

M →Λ
4n−k
T

M
is the Hodge star. This leads to the following theorem:
Theorem 2.3.1 [Kr, Theorem 3.5][Bon, Theorem 2]
For k ≤ 2n + 2, every every k-form φ admits a unique decomposition
φ =

0≤j≤k/4

j
∧ µ
k−4j
,
where the µ
k−4j
are effective (k −4j)-forms.
Bonan further refines this decomposition for quaternion-valued forms, using exterior
multiplication by the globally defined quaternionic 2-form Ψ = i
1
ω
I
+i
2
ω
J
+i
3
ω
K
. Note
that Ψ∧ Ψ = −2Ω.
Another way to consider the decomposition of forms on a quaternionic manifold is as
representations of the group Sp(1)GL(n, H). We express the Sp(1)GL(n, H)-representation
on H
n
by writing
H
n

= V
1
⊗E, (2.6)
where V
1
is the basic representation of Sp(1) on C
2
and E is the basic representation
of GL(n, H) on C
2n
. (This uses the standard convention of working with complex
representations, which in the presence of suitable structure maps can be thought of as
complexified real representations. In this case, the structure map is the tensor product
of the quaternionic structures on V
1
and E.)
This representation also describes the (co)tangent bundle of a quaternionic manifold
in the following way. Let P be a principal G-bundle over the differentiable manifold M
and let W be a G-module. We define the associated bundle
W = P
G
W =
P W
G
,
where g ∈ G acts on (p, w) ∈ P W by (p, w) g = (f g, g
−1
w). Then W is a vector
bundle over M with fibre W. We will usually just write W for W, relying on context
23
to distinguish between the bundle and the representation. Following Salamon [S3, ¸1],
if M
4n
is a quaternionic manifold with Sp(1)GL(n, H)-structure Q, then the cotangent
bundle is a vector bundle associated to the principal bundle Q and the representation
V
1
⊗E, so that we write
(T

M)
C

= V
1
⊗E (2.7)
(though we will usually omit the complexification sign). This induces an Sp(1)GL(n, H)-
action on the bundle of exterior k-forms Λ
k
T

M,
Λ
k
T

M

= Λ
k
(V
1
⊗E)

=
[k/2]

j=0
S
k−2j
(V
1
) ⊗L
k
j

=
[k/2]

j=0
V
k−2j
⊗L
k
j
, (2.8)
where L
k
j
is an irreducible representation of GL(n, H). This decomposition is given
by Salamon [S3, ¸4], along with more details concerning the nature of the GL(n, H)
representations L
k
j
. If M is quaternion K¨ahler, Λ
k
T

M can be further decomposed
into representations of the compact group Sp(1)Sp(n). This refinement is performed in
detail by Swann [Sw], and used to demonstrate significant results.
If we symmetrise completely on V
1
in Equation (2.8) to obtain V
k
, we must antisym-
metrise completely on E. Salamon therefore defines the irreducible subspace
A
k

= V
k
⊗Λ
k
E. (2.9)
The bundle A
k
can be described using the decomposition into types for the local almost
complex structures on M as follows [S3, Proposition 4.2]:
2
A
k
=

I∈S
2
Λ
k,0
I
M. (2.10)
Letting p denote the natural projection p : Λ
k
T

M → A
k
and setting D = p ◦ d, we
define a sequence of differential operators
0 −→C

(A
0
)
D=d
−→C

(A
1
= T

M)
D
−→C

(A
2
)
D
−→. . .
D
−→C

(A
2n
) −→0. (2.11)
This is accomplished using only the fact that M has an Sp(1)GL(n, H)-structure; such
a manifold is called ‘almost quaternionic’. The following theorem of Salamon relates the
integrability of such a structure with the sequence of operators in (2.11):
Theorem 2.3.2 [S3, Theorem 4.1] An almost quaternionic manifold is quaternionic if
and only if (2.11) is a complex.
This theorem is analogous to the familiar result in complex geometry that an almost
complex structure on a manifold is integrable if and only if ∂
2
= 0.
2
This is because every Sp(1)-representation V
n
is generated by its highest weight spaces taken with
respect to all the different linear combinations of I, J and K. We will use such descriptions in detail in
later chapters, and find that they play a prominent role in quaternionic algebra.
24
Chapter 3
A Double Complex on Quaternionic
Manifolds
Until now we have been discussing known material. In this chapter we present a discovery
which as far as the author can tell is new — a double complex of exterior forms on
quaternionic manifolds. We argue that this is the best quaternionic analogue of the
Dolbeault complex. The ‘top row’ of this double complex is exactly the complex (2.11)
discovered by Simon Salamon, which plays a similar role to that of the (k, 0)-forms on a
complex manifold.
The new double complex is obtained by simplifying the Bonan decomposition of
Equation (2.8). Instead of using the more complicated structure of Λ
k
T

M as an
Sp(1)GL(n, H)-module, we consider only the action of the Sp(1)-factor and decompose
Λ
k
T

M into irreducible Sp(1)-representations — a fairly easy process achieved by con-
sidering weights. The resulting decomposition gives rise to a double complex through
the Clebsch-Gordon formula, in particular the isomorphism V
r
⊗V
1

= V
r+1
⊕V
r−1
. This
encourages us to define two ‘quaternionic Dolbeault’ operators T and T, and leads to
new cohomology groups on quaternionic manifolds.
The main geometric difference between this double complex and the Dolbeault com-
plex is that whilst the Dolbeault complex is a diamond, our double complex forms an
isosceles triangle, as if the diamond is ‘folded in half’. This is more like the decomposition
of real-valued forms on complex manifolds.
Determining where our double complex is elliptic has been far more difficult than for
the de Rham or Dolbeault complexes. The ellipticity properties of our double complex are
more similar to those of a real-valued version of the Dolbeault complex. Because of this
similarity, we shall begin with a discussion of real-valued forms on complex manifolds.
3.1 Real forms on Complex Manifolds
Let M be a complex manifold and let ω ∈ Λ
p,q
= Λ
p,q
M. Then ω ∈ Λ
q,p
, and so ω + ω
is a real-valued exterior form in (Λ
p,q
⊕ Λ
q,p
)
R
, where the subscript R denotes the fact
that we are talking about real forms. The space (Λ
p,q
⊕ Λ
q,p
)
R
is a real vector bundle
associated to the principal GL(n, C)-bundle defined by the complex structure. This gives
25
a decomposition of real-valued exterior forms,
1
Λ
k
R
T

M =

p+q=k
p>q

p,q
⊕Λ
q,p
)
R
⊕Λ
k
2
,
k
2
R
. (3.1)
The condition p > q ensures that we have no repetition. The bundle Λ
k
2
,
k
2
R
only appears
when k is even. It is its own conjugate and so naturally a real vector bundle associated
to the trivial representation of U(1).
Let ω + ω ∈ (Ω
p,q
⊕ Ω
q,p
)
R
. Then ∂ω + ∂ω ∈ (Ω
p+1,q
⊕ Ω
q,p+1
)
R
. Call this operator
∂ ⊕∂. Then
d(ω + ω) = (∂ ⊕∂)(ω + ω) + (∂ ⊕∂)(ω + ω)
∈ (Ω
p+1,q
⊕Ω
q,p+1
)
R
⊕(Ω
p,q+1
⊕Ω
q+1,p
)
R
.
When p = q, ω = ω, so ∂ ⊕ ∂ = ∂ ⊕ ∂ = d, and there is only one differential operator
acting on Ω
p,p
. This gives the following double complex (where the real subscripts are
omitted for convenience).
Figure 3.1: The Real Dolbeault Complex

0,0
= C

(M)
¨
¨
¨
¨
¨
¨B
d

1,0
⊕Ω
0,1
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
∂ ⊕∂
∂ ⊕∂

1,1
¨
¨
¨
¨
¨
¨B
d

2,0
⊕Ω
0,2
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
∂ ⊕∂
∂ ⊕∂

2,1
⊕Ω
1,2
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
∂ ⊕∂
∂ ⊕∂
. . . . . . etc.

3,0
⊕Ω
0,3
r
r
r
r
r
rj
∂ ⊕∂
. . . . . . etc.
Thus there is a double complex of real forms on a complex manifold, obtained by
decomposing Λ
k
x
T

M into subrepresentations of the action of u(1) = ¸I), induced from
the action on T

x
M. This double complex is less well-behaved than its complex-valued
counterpart; in particular, it is not elliptic everywhere. We shall now show what this
means, and why it is significant.
Ellipticity
Whether or not a differential complex on a manifold is ‘elliptic’ is an important ques-
tion, with striking topological, algebraic and physical consequences. Examples of elliptic
1
This decomposition is also given by Reyes-Cari´on [R, ¸3.1], who calls the bundle (Λ
p,q
⊕ Λ
q,p
)
R
[[Λ
p,q
]].
26
complexes include the de Rham and Dolbeault complexes. A thorough description of
elliptic operators and elliptic complexes can be found in [W, Chapter 5].
Let E and F be vector bundles over M, and let Φ : C

(E) →C

(F) be a differential
operator. (We will be working with first-order operators, and so will only describe this
case.) At every point x ∈ M and for every nonzero ξ ∈ T

x
M, we define a linear map
σ
Φ
(x, ξ) : E
x
→ F
x
called the principal symbol of Φ, as follows. Let ∈ C

(E) with
(x) = e, and let f ∈ C

(M) with f(x) = 0, df(x) = ξ. Then
σ
Φ
(x, ξ)e = Φ(f)[
x
.
In coordinates, σ is often found by replacing the operator

∂x
j
with exterior multiplication
by a cotangent vector ξ
j
dual to

∂x
j
. For example, for the exterior differential d :

k
(M) →Ω
k+1
(M) we have σ
d
(x, ξ)ω = ω ∧ ξ. The operator Φ is said to be elliptic at
x ∈ M if the symbol σ
Φ
(x, ξ) : E
x
→F
x
is an isomorphism for all nonzero ξ ∈ T

x
M.
A complex 0
Φ
0
−→ C

(E
0
)
Φ
1
−→ C

(E
1
)
Φ
2
−→ C

(E
2
)
Φ
3
−→ . . .
Φ
n
−→ C

(E
n
)
Φ
n+1
−→ 0 is
said to be elliptic at E
i
if the symbol sequence E
i−1
σ
Φ
i
−→ E
i
σ
Φ
i+1
−→ E
i+1
is exact for all
ξ ∈ T

x
M and for all x ∈ M. The link between these two forms of ellipticity is as follows.
If we have a metric on each E
i
then we can define a formal adjoint Φ

i
: E
i
→ E
i−1
.
Linear algebra reveals that the complex is elliptic at E
i
if and only if the Laplacian
Φ

i
Φ
i
+ Φ
i−1
Φ

i−1
is an elliptic operator.
One important implication of this is that an elliptic complex on a compact manifold
has finite-dimensional cohomology groups [W, Theorem 5.2, p. 147]. Whether a complex
yields interesting cohomological information is in this way directly related to whether or
not the complex is elliptic. The following Proposition answers this question for the real
Dolbeault complex.
Proposition 3.1.1 For p > 0, the upward complex
0 −→Ω
p,p
−→Ω
p+1,p
⊕Ω
p,p+1
−→Ω
p+2,p
⊕Ω
p,p+2
−→. . .
is elliptic everywhere except at the first two spaces Ω
p,p
and Ω
p+1,p
⊕Ω
p,p+1
.
For p = 0, the ‘leading edge’ complex
0 −→Ω
0,0
−→Ω
1,0
⊕Ω
0,1
−→Ω
2,0
⊕Ω
0,2
−→. . .
is elliptic everywhere except at Ω
1,0
⊕Ω
0,1
= Ω
1
(M).
Proof. When p > q + 1, we have short sequences of the form

p−1,q

−→ Ω
p,q

−→ Ω
p+1,q


q,p−1

−→ Ω
q,p

−→ Ω
q,p+1
.
(3.2)
Each such sequence is (a real subspace of) the direct sum of two elliptic sequences, and
so is elliptic. Thus we have ellipticity at Ω
p,q
⊕Ω
q,p
whenever p ≥ q + 2.
27
This leaves us to consider the case when p = q, and the sequence
0 −→ Ω
p,p

¸
¸


p+1,p

−→ Ω
p+2,2
−→. . . etc.


p,p+1

−→ Ω
2,p+2
−→. . . etc.
(3.3)
This fails to be elliptic. An easy and instructive way to see this is to consider the simplest
4-dimensional example M = C
2
.
Let e
0
, e
1
= I(e
0
), e
2
and e
3
= I(e
2
) form a basis for T

x
C
2

= C
2
, and let e
ab...
denote e
a
∧ e
b
∧ . . . etc. Then I(e
01
) = e
00
−e
11
= 0, so e
01
∈ Λ
1,1
. The map from Λ
1,1
to Λ
2,1
⊕ Λ
1,2
is just the exterior differential d. Since σ
d
(x, e
0
)(e
01
) = e
01
∧ e
0
= 0 the
symbol map σ
d
: Λ
1,1
→Λ
2,1
⊕Λ
1,2
is not injective, so the symbol sequence is not exact
at Λ
1,1
.
Consider also e
123
∈ Λ
2,1
⊕Λ
1,2
. Then σ
∂⊕∂
(x, e
0
)(e
123
) = 0, since there is no bundle
Λ
3,1
⊕Λ
1,3
. But e
123
has no e
0
-factor, so is not the image under σ
d
(x, e
0
) of any form
α ∈ Λ
1,1
. Thus the symbol sequence fails to be exact at Λ
2,1
⊕Λ
1,2
.
It is a simple matter to extend these counterexamples to higher dimensions and higher
exterior powers. For k = 0, the situation is different. It is easy to show that the complex
0 −→C

(M)
d
−→Ω
1,0
⊕Ω
0,1
∂⊕∂
−→Ω
2,0
⊕Ω
0,2
−→. . . etc.
is elliptic everywhere except at (Ω
1,0
⊕Ω
0,1
).
This last sequence is given particular attention by Reyes-Cari´on [R, Lemma 2]. He
shows that, when M is K¨ahler, ellipticity can be regained by adding the space ¸ω) to
the bundle Λ
2,0
⊕Λ
0,2
, where ω is the real K¨ahler (1, 1)-form.
The Real Dolbeault complex is thus elliptic except at the bottom of the isosceles
triangle of spaces. Here the projection from d(Ω
p,p
) to Ω
p+1,p
⊕ Ω
p,p+1
is the identity,
and arguments based upon non-trivial projection maps no longer apply. We shall see
that this situation is closely akin to that of differential forms on quaternionic manifolds,
and that techniques motivated by this example yield similar results.
3.2 Construction of the Double Complex
Let M
4n
be a quaternionic manifold. Then T

x
M

= V
1
⊗ E as an Sp(1)GL(n, H)-
representation for all x ∈ M. Suppose we consider just the action of the Sp(1)-factor.
Then the (complexified) cotangent space effectively takes the form V
1
⊗ C
2n

= 2nV
1
.
Thus the Sp(1)-action on Λ
k
T

M is given by the representation Λ
k
(2nV
1
).
To work out the irreducible decomposition of this representation we compute the
weight space decomposition of Λ
k
(2nV
1
) from that of 2nV
1
.
2
With respect to the
action of a particular subgroup U(1)
q
⊂ Sp(1), the representation 2nV
1
has weights
+1 and −1, each occuring with multiplicity 2n. The weights of Λ
k
(2nV
1
) are the k-
wise distinct sums of these. Each weight r in Λ
k
(2nV
1
) must therefore be a sum of p
occurences of the weight ‘+1’ and p−r occurences of the weight ‘−1’, where 2p−r = k
2
This process for calculating the weights of tensor, symmetric and exterior powers is a standard
technique in representation theory — see for example [FH, ¸11.2].
28
and 0 ≤ p ≤ k (from which it follows immediately that −k ≤ r ≤ k and r ≡ k mod 2).
The number of ways to choose the p ‘+1’ weights is
_
2n
p
_
, and the number of ways
to choose the (p − r) ‘−1’ weights is
_
2n
p−r
_
, so the multiplicity of the weight r in the
representation Λ
k
(2nV
1
) is
Mult(r) =
_
2n
k+r
2
__
2n
k−r
2
_
.
For r ≥ 0, consider the difference Mult(r)−Mult(r+2). This is the number of weight
spaces of weight r which do not have any corresponding weight space of weight r + 2.
Each such weight space must therefore be the highest weight space in an irreducible
subrepresentation V
r
⊆ Λ
k
T

M, from which it follows that the number of irreducibles
V
r
in Λ
k
(2nV
1
) is equal to Mult(r)−Mult(r + 2). This demonstrates the following
Proposition:
Proposition 3.2.1 Let M
4n
be a hypercomplex manifold. The decomposition into irre-
ducibles of the induced representation of Sp(1) on Λ
k
T

M is
Λ
k
T

M

=
k

r=0
__
2n
k+r
2
__
2n
k−r
2
_

_
2n
k+r+2
2
__
2n
k−r−2
2
__
V
r
,
where r ≡ k mod 2.
We will not always write the condition r ≡ k mod 2, assuming that
_
p
q
_
= 0 if q ,∈ Z.
Definition 3.2.2 Let M
4n
be a quaternionic manifold. Define the coefficient

n
k,r
=
_
2n
k+r
2
__
2n
k−r
2
_

_
2n
k+r+2
2
__
2n
k−r−2
2
_
,
and let E
k,r
be the vector bundle associated to the Sp(1)-representation
n
k,r
V
r
. With
this notation Proposition 3.2.1 is written
Λ
k
T

M

=
k

r=0

n
k,r
V
r
=
k

r=0
E
k,r
.
Our most important result is that this decomposition gives rise to a double complex
of differential forms and operators on a quaternionic manifold.
Theorem 3.2.3 The exterior derivative d maps C

(M, E
k,r
) to C

(M, E
k+1,r+1

E
k+1,r−1
).
Proof. Let ∇ be a torsion-free linear connection on M preserving the quaternionic
structure. Then ∇ : C

(M, E
k,r
) → C

(M, E
k,r
⊗ T

M). As Sp(1)-representations,
this is
∇ : C

(M,
n
k,r
V
r
) →C

(M,
n
k,r
V
r
⊗2nV
1
).
29
Using the Clebsch-Gordon formula we have
n
k,r
V
r
⊗2nV
1

= 2n
n
k,r
(V
r+1
⊕V
r−1
). Thus the
image of E
k,r
under ∇ is contained in the V
r+1
and V
r−1
summands of Λ
k
T

M ⊗T

M.
Since ∇ is torsion-free, d = ∧ ◦ ∇, so d maps (sections of) E
k,r
to the V
r+1
and V
r−1
summands of Λ
k+1
T

M. Thus d : C

(M, E
k,r
) →C

(M, E
k+1,r+1
⊕E
k+1,r−1
).
This allows us to split the exterior differential d into two ‘quaternionic Dolbeault
operators’.
Definition 3.2.4 Let π
k,r
be the natural projection from Λ
k
T

M onto E
k,r
. Define the
operators
T : C

(E
k,r
) →C

(E
k+1,r+1
)
T = π
k+1,r+1
◦ d
and
T : C

(E
k,r
) →C

(E
k+1,r−1
)
T = π
k+1,r−1
◦ d.
(3.4)
Theorem 3.2.3 is equivalent to the following:
Proposition 3.2.5 On a quaternionic manifold M, we have d = T +T, and so
T
2
= TT +TT = T
2
= 0.
Proof. The first equation is equivalent to Theorem 3.2.3. The rest follows immediately
from decomposing the equation d
2
= 0.
Figure 3.2: The Double Complex
C

(E
0,0
= R)
¨
¨
¨
¨
¨
¨B
T
C

(E
1,1
= T

M)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
T
T
C

(E
2,0
)
¨
¨
¨
¨
¨
¨B
T
C

(E
2,2
)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
T
T
C

(E
3,1
)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
T
T
. . . . . . etc.
C

(E
3,3
)
r
r
r
r
r
rj
T
. . . . . . etc.
Here is our quaternionic analogue of the Dolbeault complex. There are strong sim-
ilarities between this and the Real Dolbeault complex (Figure 3.1). Again, instead of
a diamond as in the Dolbeault complex, the quaternionic version only extends upwards
to form an isosceles triangle. This is essentially because there is one irreducible U(1)-
representation for each integer, whereas there is one irreducible Sp(1)-representation only
for each nonnegative integer. Note that this is a decomposition of real as well as complex
differential forms; the operators T and T map real forms to real forms.
30
By definition, the bundle E
k,k
is the bundle A
k
of (2.9) — they are both the subbundle
of Λ
k
T

M which includes all Sp(1)-representations of the form V
k
. Thus the leading edge
of the double complex
0 −→C

(E
0,0
)
D
−→C

(E
1,1
)
D
−→C

(E
2,2
)
D
−→. . .
D
−→C

(E
2n,2n
)
D
−→0
is precisely the complex (2.11) discovered by Salamon.
Example 3.2.6 Four Dimensions
This double complex is already very well-known and understood in four dimensions.
Here there is a splitting only in the middle dimension, Λ
2
T

M

= V
2
⊕ 3V
0
. Let I, J
and K be local almost complex structures at x ∈ M, and let e
0
∈ T

x
M. Let e
1
= I(e
0
),
e
2
= J(e
0
) and e
3
= K(e
0
). In this way we obtain a basis ¦e
0
, . . . , e
3
¦ for T

x
M

= H.
Using the notation e
ijk...
= e
i
∧ e
j
∧ e
k
∧ . . .etc., define the 2-forms
ω
±
1
= e
01
±e
23
, ω
±
2
= e
02
±e
31
, ω
±
3
= e
03
±e
12
. (3.5)
Then I, J and K all act trivially
3
on the ω

j
, so E
2,0
= ¸ω

1
, ω

2
, ω

3
). The action of
sp(1) on the ω
+
j
is given by the multiplication table
I(ω
+
1
) = 0
J(ω
+
1
) = −2ω
+
3
K(ω
+
1
) = 2ω
+
2
I(ω
+
2
) = 2ω
+
3
J(ω
+
2
) = 0
K(ω
+
2
) = −2ω
+
1
I(ω
+
3
) = −2ω
+
3
J(ω
+
3
) = 2ω
+
1
K(ω
+
3
) = 0.
(3.6)
These are the relations of the irreducible sp(1)-representation V
2
, and we see that E
2,2
=
¸ω
+
1
, ω
+
2
, ω
+
3
).
These bundles will be familiar to most readers: E
2,2
is the bundle of self-dual 2-
forms Λ
2
+
and E
2,0
is the bundle of anti-self-dual 2-forms Λ
2

. The celebrated splitting
Λ
2
T

M

= Λ
2
+
⊕Λ
2

is an invariant of the conformal class of any Riemannian 4-manifold,
and I
2
+ J
2
+ K
2
= −4(∗ + 1), where ∗ : Λ
k
T

M →Λ
4−k
T

M is the Hodge star map.
This also serves to illustrate why in four dimensions we make the definition that
a quaternionic manifold is a self-dual conformal manifold. The relationship between
quaternionic, almost complex and Riemannian structures in four dimensions is described
in more detail in [S4, Chapter 7], a classic reference being [AHS].
Because there is no suitable quaternionic version of holomorphic coordinates, there is
no ‘nice’ co-ordinate expression for a typical section of C

(E
k,r
). In order to determine
the decomposition of a differential form, the simplest way the author has found is to use
the Casimir operator ( = I
2
+ J
2
+ K
2
. Consider a k-form α. Then α ∈ E
k,r
if and
only if (I
2
+ J
2
+ K
2
)(α) = −r(r + 2)α. This mechanism also allows us to work out
expressions for T and T acting on α.
Lemma 3.2.7 Let α ∈ C

(E
k,r
). Then
Tα = −
1
4
_
(r −1) +
1
r + 1
(I
2
+ J
2
+ K
2
)
_

3
We are assuming throughout that uppercase operators like I, J and K are acting as elements of a
Lie algebra, not a Lie group. This makes no difference on T

M but is important on the exterior powers
Λ
k
T

M. In particular, for k ,= 1 we no longer expect I
2
= J
2
= K
2
= −1, and by ‘act trivially’ we
mean ‘annihilate’.
31
and
Tα =
1
4
_
(r + 3) +
1
r + 1
(I
2
+ J
2
+ K
2
)
_
dα.
Proof. We have dα = Tα +Tα, where Tα ∈ E
k+1,r+1
and Tα ∈ E
k+1,r−1
. Applying the
Casimir operator gives
(I
2
+ J
2
+ K
2
)(dα) = −(r + 1)(r + 3)Tα −(r + 1)(r −1)Tα.
Rearranging these equations gives Tα and Tα.
Writing T
k,r
for the particular map T : C

(E
k,r
) → C

(E
k+1,r+1
), we define the
quaternionic cohomology groups
H
k,r
D
(M) =
Ker(T
k,r
)
Im(T
k−1,r−1
)
. (3.7)
3.3 Ellipticity and the Double Complex
In this section we shall determine where our double complex is elliptic and where it is
not. It turns out that we have ellipticity everywhere except on the bottom two rows of
the complex. This is exactly like the real Dolbeault complex of Figure 3.1, and though
it is more difficult to prove for the quaternionic version, the guiding principles which
determine where the two double complexes are elliptic are very much the same in both
cases.
Here is the main result of this section:
Theorem 3.3.1 For 2k ≥ 4, the complex
0 →E
2k,0
D
→E
2k+1,1
D
→E
2k+2,2
D
→. . .
D
→E
2n+k,2n−k
D
→0
is elliptic everywhere except at E
2k,0
and E
2k+1,1
, where it is not elliptic.
For k = 1 the complex is elliptic everywhere except at E
3,1
, where it is not elliptic.
For k = 0 the complex is elliptic everywhere.
The rest of this section provides a proof of this Theorem. Note the strong similarity
between this Theorem and Proposition 3.1.1, the analogous result for the Real Dolbeault
complex. Again, it is the isosceles triangle as opposed to diamond shape which causes
ellipticity to fail for the bottom row, because d = T on E
2k,0
and the projection from
d(C

(E
2k,0
)) to C

(E
2k+1,1
) is the identity.
On a complex manifold M
2n
with holomorphic coordinates z
j
, the exterior forms
dz
a
1
∧ . . . ∧ dz
a
p
∧ d¯ z
b
1
∧ . . . ∧ d¯ z
b
q
span Λ
p,q
. This allows us to decompose any form
ω ∈ Λ
p,q
, making it much easier to write down the kernels and images of maps which
involve exterior multiplication. On a quaternionic manifold M
4n
there is unfortunately no
easy way to write down a local frame for the bundle E
k,r
, because there is no quaternionic
version of ‘holomorphic coordinates’. However, we can decompose E
k,r
just enough to
enable us to prove Theorem 3.3.1.
32
A principal observation is that since ellipticity is a local property, we can work on H
n
without loss of generality. Secondly, since GL(n, H) acts transitively on H
n
¸ ¦0¦, if the
symbol sequence . . .
σ
e
0
−→E
k,r
σ
e
0
−→E
k+1,r+1
σ
e
0
−→. . . is exact for any nonzero e
0
∈ T

H
n
then it is exact for all nonzero ξ ∈ T

H
n
. To prove Theorem 3.3.1, we choose one such
e
0
and analyse the spaces E
k,r
accordingly.
3.3.1 Decomposition of the Spaces E
k,r
Let e
0
∈ T

x
H
n

= H
n
and let (I, J, K) be the standard hypercomplex structure on H
n
. As
in Example 3.2.6, define e
1
= I(e
0
), e
2
= J(e
0
) and e
3
= K(e
0
), so that ¸e
0
, . . . , e
3
)

= H.
In this way we single out a particular copy of Hwhich we call H
0
, obtaining a (nonnatural)
splitting T

x
H
n

= H
n−1
⊕H
0
which is preserved by action of the hypercomplex structure.
This induces the decomposition Λ
k
H
n

=

4
l=0
Λ
k−l
H
n−1
⊗Λ
l
H
0
, which decomposes each
E
k,r
⊂ Λ
k
H
n
according to how many differentials in the H
0
-direction are present.
Definition 3.3.2 Define the space E
l
k,r
to be the subspace of E
k,r
consisting of exterior
forms with precisely l differentials in the H
0
-direction, i.e.
E
l
k,r
≡ E
k,r
∩ (Λ
k−l
H
n−1
⊗Λ
l
H
0
).
Then E
l
k,r
is preserved by the induced action of the hypercomplex structure on Λ
k
H
n
.
Thus we obtain an invariant decomposition E
k,r
= E
0
k,r
⊕E
1
k,r
⊕E
2
k,r
⊕E
3
k,r
⊕E
4
k,r
. Note
that we can identify E
0
k,r
on H
n
with E
k,r
on H
n−1
.
(Throughout the rest of this section, juxtaposition of exterior forms will denote ex-
terior multiplication, for example αe
ij
means α ∧ e
ij
.)
We can decompose these summands still further. Consider, for example, the bun-
dle E
1
k,r
. An exterior form α ∈ E
1
k,r
is of the form α
0
e
0
+ α
1
e
1
+ α
2
e
2
+ α
3
e
3
, where
α
j
∈ Λ
k−1
H
n−1
. Thus α is an element of Λ
k−1
H
n−1
⊗ 2V
1
, since H
0

= 2V
1
as an sp(1)-
representation. Since α is in a copy of the representation V
r
, it follows from the isomor-
phism V
r
⊗ V
1

= V
r+1
⊕ V
r−1
that the α
j
must be in a combination of V
r+1
and V
r−1
representations, i.e. α
j
∈ E
0
k−1,r+1
⊕E
0
k−1,r−1
. We write
α = α
+
+ α

= (α
+
0
+ α

0
)e
0
+ (α
+
1
+ α

1
)e
1
+ (α
+
2
+ α

2
)e
2
+ (α
+
3
+ α

3
)e
3
,
where α
+
j
∈ E
0
k−1,r+1
and α

j
∈ E
0
k−1,r−1
.
The following Lemma allows us to consider α
+
and α

separately.
Lemma 3.3.3 If α = α
+
+ α

∈ E
1
k,r
then both α
+
and α

must be in E
1
k,r
.
Proof. In terms of representations, the situation is of the form
(pV
r+1
⊕qV
r−1
) ⊗2V
1

= 2p(V
r+2
⊕V
r
) ⊕2q(V
r
⊕V
r−2
),
where α
+
∈ pV
r+1
and α

∈ qV
r−1
. For α to be in the representation 2(p + q)V
r
,
its components in the representations 2pV
r+2
and 2qV
r−2
must both vanish separately.
The component in 2pV
r+2
comes entirely from α
+
, so for this to vanish we must have
33
α
+
∈ 2(p +q)V
r
independently of α

. Similarly, for the component in 2qV
r−2
to vanish,
we must have α

∈ 2(p + q)V
r
.
Thus we decompose the space E
1
k,r
into two summands, one coming from E
0
k−1,r−1

2V
1
and one from E
0
k−1,r+1
⊗2V
1
. We extend this decomposition to the cases l = 0, 2, 3, 4,
defining the following notation.
Definition 3.3.4 Define the bundle E
l,m
k,r
to be the subbundle of E
l
k,r
arising from
V
m
-type representations in Λ
k−l
H
n−1
. In other words,
E
l,m
k,r
≡ (E
0
k−l,m
⊗Λ
l
H
0
) ∩ E
k,r
.
To recapitulate: for the space E
l,m
k,r
, the bottom left index k refers to the exterior
power of the form α ∈ Λ
k
H
n
; the bottom right index r refers to the irreducible Sp(1)-
representation in which α lies; the top left index l refers to the number of differentials in
the H
0
-direction and the top right index m refers to the irreducible Sp(1)-representation
of the contributions from Λ
k−a
H
n−1
before wedging with forms in the H
0
-direction. This
may appear slightly fiddly: it becomes rather simpler when we consider the specific
splittings which Definition 3.3.4 allows us to write down.
Lemma 3.3.5 Let E
l,m
k,r
be as above. We have the following decompositions:
E
0
k,r
= E
0,r
k,r
E
1
k,r
= E
1,r+1
k,r
⊕E
1,r−1
k,r
E
2
k,r
= E
2,r+2
k,r
⊕E
2,r
k,r
⊕E
2,r−2
k,r
E
3
k,r
= E
3,r+1
k,r
⊕E
3,r−1
k,r
and E
4
k,r
= E
4,r
k,r
.
Proof. The first isomorphism is trivial, as is the last (since the hypercomplex structure
acts trivially on Λ
4
H
0
). The second isomorphism is Lemma 3.3.3, and the fourth follows
in exactly the same way since Λ
3
H
0

= 2V
1
also. The middle isomorphism follows a
similar argument.
Recall the self-dual forms and anti-self-dual forms in Example 3.2.6. The bundle E
2,r
k,r
splits according to whether its contribution from Λ
2
H
0
is self-dual or anti-self-dual. We
will call these summands E
2,r+
k,r
and E
2,r−
k,r
respectively, so E
2,r
k,r
= E
2,r+
k,r
⊕E
2,r−
k,r
.
3.3.2 Lie in conditions
We have analysed the bundle E
k,r
into a number of different subbundles. We now
determine when a particular exterior form lies in one of these subbundles. Consider
a form α = α
1
e
s
1
...s
a
+ α
2
e
t
1
...t
a
+ . . . etc. where α
j
∈ E
0
k−a,b
. For α to lie in one of the
spaces E
a,b
k,r
the α
j
will usually have to satisfy some simultaneous equations. Since these
are the conditions for a form to lie in a particular Lie algebra representation, we will
refer to such equations as ‘Lie in conditions’.
To begin with, we mention three trivial Lie in conditions. Let α ∈ E
0
k,r
. That α ∈ E
0,r
k,r
is obvious, as is αe
0123
∈ E
4,r
k,r
, since wedging with e
0123
has no effect on the sp(1)-action.
34
Likewise, the sp(1)-action on the anti-self-dual 2-forms ω

1
= e
01
− e
23
, ω

2
= e
02
− e
31
and ω

3
= e
03
−e
12
is trivial, so αω

j
∈ E
2,r−
k,r
for all j = 1, 2, 3.
This leaves the following three situations: those arising from taking exterior products
with 1-forms, 3-forms and the self-dual 2-forms ω
+
j
. As usual when we want to know
which representation an exterior form is in, we apply the Casimir operator.
The cases l = 1 and l = 3
Let α
j
∈ E
0
k,r
. Then α = α
0
e
0
+ α
1
e
1
+ α
2
e
2
+ α
3
e
3
∈ E
1,r
k+1,r+1
⊕ E
1,r
k+1,r−1
, and α is
entirely in E
1,r
k+1,r+1
if and only if (I
2
+ J
2
+ K
2
)α = −(r + 1)(r + 3)α.
By the usual (Leibniz) rule for a Lie algebra action on a tensor product, we have that
I
2

j
e
j
) = I
2

j
)e
j
+ 2I(α
j
)I(e
j
) + α
j
I
2
(e
j
), etc. Thus
(I
2
+ J
2
+ K
2
)α =
3

j=0
_
(I
2
+ J
2
+ K
2
)(α
j
)e
j
+ α
j
(I
2
+ J
2
+ K
2
)(e
j
) +
+ 2
_
I(α
j
)I(e
j
) + J(α
j
)J(e
j
) + K(α
j
)K(e
j
)
__
= −r(r + 2)α −3α + 2
3

j=0
_
I(α
j
)I(e
j
) + J(α
j
)J(e
j
) + K(α
j
)K(e
j
)
_
= (−r
2
−2r −3)α + 2
_
I(α
0
)e
1
−I(α
1
)e
0
+ I(α
2
)e
3
−I(α
3
)e
2
+
+J(α
0
)e
2
−J(α
1
)e
3
−J(α
2
)e
0
+ J(α
3
)e
1
+
+K(α
0
)e
3
+ K(α
1
)e
2
−K(α
2
)e
1
−K(α
3
)e
0
_
.
(3.8)
For α ∈ E
1,r
k+1,r+1
we need this to be equal to −(r + 1)(r + 3)α, which is the case if and
only if
−rα = I(α
0
)e
1
−I(α
1
)e
0
+ I(α
2
)e
3
−I(α
3
)e
2
+ J(α
0
)e
2
−J(α
1
)e
3
−J(α
2
)e
0
+ J(α
3
)e
1
+
+ K(α
0
)e
3
+ K(α
1
)e
2
−K(α
2
)e
1
−K(α
3
)e
0
.
Since the α
j
have no e
j
-factors and the action of I, J and K preserves this property,
this equation can only be satisfied if it holds for each of the e
j
-components separately.
We conclude that α ∈ E
1,r
k+1,r+1
if and only if α
0
, α
1
, α
2
and α
3
satisfy the following Lie
in conditions:
4

0
−I(α
1
) −J(α
2
) −K(α
3
) = 0

1
+ I(α
0
) + J(α
3
) −K(α
2
) = 0

2
−I(α
3
) + J(α
0
) + K(α
1
) = 0

3
+ I(α
2
) −J(α
1
) + K(α
0
) = 0.
(3.9)
Suppose instead that α ∈ E
1,r
k+1,r−1
. Then (I
2
+J
2
+K
2
)α = −(r−1)(r+1)α. Putting
this alternative into Equation (3.8) gives the result that α ∈ E
1,r
k+1,r−1
if and only if
4
Our interest in these conditions arises from a consideration of exterior forms, but the equations
describe sp(1)-representations in general: they are the conditions that α ∈ V
r
⊗V
1
must satisfy to be in
the V
r+1
subspace of V
r+1
⊕V
r−1

= V
r
⊗V
1
. The other Lie in conditions have similar interpretations.
35
(r + 2)α
0
+ I(α
1
) + J(α
2
) + K(α
3
) = 0
(r + 2)α
1
−I(α
0
) −J(α
3
) + K(α
2
) = 0
(r + 2)α
2
+ I(α
3
) −J(α
0
) −K(α
1
) = 0
(r + 2)α
3
−I(α
2
) + J(α
1
) −K(α
0
) = 0.
(3.10)
Consider now α = α
0
e
123
+ α
1
e
032
+ α
2
e
013
+ α
3
e
021
∈ E
3,r
k+3,r+1
⊕ E
3,r
k+3,r−1
. Since
Λ
3
H
0

= H
0
, the Lie in conditions are exactly the same: for α to be in E
3,r
k+3,r+1
we need
the α
j
to satisfy Equations (3.9), and for α to be in E
3,r
k+3,r−1
we need the α
j
to satisfy
Equations (3.10).
The case l = 2
We have already noted that wedging a form β ∈ E
0
k,r
with an anti-self-dual 2-form ω

j
has no effect on the sp(1)-action, so βω

j
∈ E
2,r−
k+2,r
. Thus we only have to consider the
effect of wedging with the self-dual 2-forms ¸ω
+
1
, ω
+
2
, ω
+
3
)

= V
2
⊂ Λ
2
H
0
. By the Clebsch-
Gordon formula, the decomposition takes the form V
r
⊗ V
2

= V
r+2
⊕ V
r
⊕ V
r−2
. Thus
for β = β
1
ω
+
1
+ β
2
ω
+
2
+ β
3
ω
+
3
we want to establish the Lie in conditions for β to be in
E
2,r
k+2,r+2
, E
2,r+
k+2,r
and E
2,r
k+2,r−2
.
We calculate these Lie in conditions in a similar fashion to the previous cases, by con-
sidering the action of the Casimir operator I
2
+J
2
+K
2
on β and using the multiplication
table (3.6). The following Lie in conditions are then easy to deduce:
β ∈ E
2,r
k+2,r+2
⇐⇒
_
_
_
(r + 4)β
1
= J(β
3
) −K(β
2
)
(r + 4)β
2
= K(β
1
) −I(β
3
)
(r + 4)β
3
= I(β
2
) −J(β
1
).
(3.11)
β ∈ E
2,r+
k+2,r
⇐⇒
_
_
_

1
= J(β
3
) −K(β
2
)

2
= K(β
1
) −I(β
3
)

3
= I(β
2
) −J(β
1
).
(3.12)
β ∈ E
2,r
k+2,r−2
⇐⇒
_
_
_
(2 −r)β
1
= J(β
3
) −K(β
2
)
(2 −r)β
2
= K(β
1
) −I(β
3
)
(2 −r)β
3
= I(β
2
) −J(β
1
).
(3.13)
Equation 3.12 is particularly interesting. Since this equation singles out the
V
r
-representation in the direct sum V
r+2
⊕ V
r
⊕ V
r+2
, it must have dimV
r
= r + 1
linearly independent solutions. Let β
0
∈ V
r
and let β
1
= I(β
0
), β
2
= J(β
0
), β
3
= K(β
0
).
Using the Lie algebra relations 2I = [J, K] = JK − KJ, it is easy to see that β
1
, β
2
and β
3
satisfy Equation 3.12. Moreover, there are r + 1 linearly independent solutions
of this form (for r ,= 0). We conclude that all the solutions of Equation (3.12) take the
form β
1
= I(β
0
), β
2
= J(β
0
), β
3
= K(β
0
).
3.3.3 The Symbol Sequence and Proof of Theorem 3.3.1
We now describe the principal symbol of T, and examine its behaviour in the context
of the decompositions of Definition 3.3.2 and Lemma 3.3.5. This leads to a proof of
Theorem 3.3.1. First we obtain the principal symbol from the formula for T in Lemma
3.2.7 by replacing dα with αe
0
.
36
Proposition 3.3.6 Let x ∈ H
n
, e
0
∈ T

x
H
n
and α ∈ E
k,r
. The principal symbol
mapping σ
D
(x, e
0
) : E
k,r
→E
k+1,r+1
is given by
σ
D
(x, e
0
)(α) =
1
2(r + 1)
_
(r + 2)αe
0
−I(α)e
1
−J(α)e
2
−K(α)e
3
_
.
Proof. Replacing dα with αe
0
in the formula for T obtained in Lemma 3.2.7, we have
σ
D
(x, e
0
)(α) = −
1
4
_
(r −1) +
1
r + 1
(I
2
+ J
2
+ K
2
)
_
αe
0
=
−1
4(r + 1)
_
( (r −1)(r + 1) −r(r + 2) −3 ) αe
0
+2
_
I(α)e
1
+ J(α)e
2
+ K(α)e
3
_
_
=
1
2(r + 1)
_
(r + 2)αe
0
−I(α)e
1
−J(α)e
2
−K(α)e
3
_
,
as required.
Corollary 3.3.7 The principal symbol σ
D
(x, e
0
) maps the space E
l,m
k,r
to the space
E
l+1,m
k+1,r+1
.
Proof. We already know that σ
D
: E
k,r
→ E
k+1,r+1
, by definition. Using Lemma 3.3.6,
we see that σ
D
(x, e
0
) increases the number of differentials in the H
0
-direction by one, so
the index l increases by one. The only action in the other directions is the sp(1)-action,
which preserves the irreducible decomposition of the contribution from Λ
k−a
H
n−1
, so the
index m remains the same.
(To save space we shall use σ as an abbreviation for σ
D
(x, e
0
) for the rest of this
section.)
The point of all this work on decomposition now becomes apparent. Since σ : E
l
k,r

E
l+1
k+1,r+1
, we can reduce the (somewhat indefinite) symbol sequence
. . .
σ
−→E
k−1,r−1
σ
−→E
k,r
σ
−→E
k+1,r+1
σ
−→. . . etc.
to the 5-space sequence
0
σ
−→E
0
k−2,r−2
σ
−→E
1
k−1,r−1
σ
−→E
2
k,r
σ
−→E
3
k+1,r+1
σ
−→E
4
k+2,r+2
σ
−→0. (3.14)
Using Lemma 3.3.5 as well, we can analyse this sequence still further according to the
different (top right) m-indices, obtaining three short sequences (for k ≥ 2, k ≡ r mod 2)
0 → E
2,r+2
k,r
→ E
3,r+2
k+1,r+1
→ E
4,r+2
k+2,r+2
→ 0
⊕ ⊕
0 → E
1,r
k−1,r−1
→ E
2,r
k,r
→ E
3,r
k+1,r+1
→ 0
⊕ ⊕
0 → E
0,r−2
k−2,r−2
→ E
1,r−2
k−1,r−1
→ E
2,r−2
k,r
→ 0.
(3.15)
37
This reduces the problem of determining where the operator T is elliptic to the problem
of determining when these three sequences are exact.
For a sequence 0 → A → B → C → 0 to be exact, it is necessary that dimA −
dimB + dimC = 0. Given this condition, if the sequence is exact at any two out of A,
B and C it is exact at the third. We shall show that for r ,= 0 this dimension sum does
equal zero.
Lemma 3.3.8 For r > 0, each of the sequences in (3.15) satisfies the dimension condi-
tion above, i.e. the alternating sum of the dimensions vanishes.
Proof. Let r > 0. We calculate the dimensions of the spaces E
l,m
k,r
for l = 0, . . . , 4. Recall
the notation E
k,r
=
n
k,r
V
r
from Definition 3.2.2. It is clear that dimE
0
k,r
= (r + 1)
n−1
k,r
,
since E
0
k,r
on H
n
is simply E
k,r
on H
n−1
. Thus dimE
0,r−2
k−2,r−2
= (r − 1)
n−1
k−2,r−2
and
dimE
4,r+2
k+2,r+2
= (r + 3)
n−1
k−2,r+2
.
The cases a = 1 and a = 3 are easy to work out since they are of the form E
0
k,r
⊗2V
1
.
For a = 1, we have dimE
1,r−2
k−1,r−1
= 2r
n−1
k−2,r−2
and dimE
1,r
k−1,r−1
= 2r
n−1
k−2,r
. For a = 3,
dimE
3,r
k+1,r+1
= 2(r + 2)
n−1
k−2,r
and dimE
3,r+2
k+1,r+1
= 2(r + 2)
n−1
k−2,r+2
.
The case a = 2 is slightly more complicated, as we have to take into account exterior
products with the self-dual 2-forms V
2
and anti-self-dual 2-forms 3V
0
in Λ
2
H
0
. The spaces
E
2,r+2
k,r
and E
2,r−2
k,r
receive contributions only from the self-dual part V
2
, from which we
infer that dimE
2,r+2
k,r
= (r + 1)
n−1
k−2,r+2
and dimE
2,r−2
k,r
= (r + 1)
n−1
k−2,r−2
. Finally, the
space E
2,r+
k,r
has dimension (r+1)
n−1
k−2,r
and the space E
2,r−
k,r
has dimension 3(r+1)
n−1
k−2,r
,
giving E
2,r
k,r
a total dimension of 4(r + 1)
n−1
k−2,r
.
It is now a simple matter to verify that for the top sequence of (3.15)

n−1
k−2,r+2
(r + 1 −2(r + 2) + r + 3) = 0,
for the middle sequence

n−1
k−2,r
(2r −4(r + 1) + 2(r + 2)) = 0,
and for the bottom sequence

n−1
k−2,r−2
(r −1 −2r + r + 1) = 0.
The case r = 0 is different. Here the bottom sequence of (3.15) disappears altogether,
the top sequence still being exact. Exactness is lost in the middle sequence. Since the
isomorphism
n−1
k−2,0
V
0
⊗V
2

=
n−1
k−2,0
V
2
gives no trivial V
0
-representations, there is no space
E
2,0+
k,0
. Thus E
2,0
k,0
is ‘too small’ — we are left with a sequence
0 −→3
n−1
k−2,0
V
0
−→2
n−1
k−2,0
V
1
−→0,
which cannot be exact. (As there is no space E
2
0,0
, this problem does not arise for the
leading edge 0 →E
0,0
→E
1,1
→. . . etc.)
We are finally in a position to prove Theorem 3.3.1, which now follows from:
38
Proposition 3.3.9 When r ,= 0, the three sequences of (3.15) are exact.
Proof. Consider first the top sequence 0
σ
−→ E
2,r+2
k,r
σ
−→ E
3,r+2
k+1,r+1
σ
−→ E
4,r+2
k+2,r+2
−→ 0.
The Clebsch-Gordon formula shows that there are no spaces E
3,r+2
k+1,r−1
or E
4,r+2
k+2,r
. Thus
T = 0 on E
2,r+2
k,r
and E
3,r+2
k+1,r+1
, so T = d for the top sequence. It is easy to check using
the relevant Lie in conditions that the map ∧e
0
: E
2,r+2
k,r
→E
3,r+2
k+1,r+1
is injective and the
map ∧e
0
: E
3,r+2
k+1,r+1
→E
4,r+2
k+2,r+2
is surjective.
To show exactness at E
1
k−1,r−1
, consider α = α
0
e
0

1
e
1

2
e
2

3
e
3
∈ E
1
k−1,r−1
. A
calculation using Proposition 3.3.6 shows that
σ(α) =
1
2r
_
(rα
1
+ I(α
0
))e
10
+ (rα
2
+ J(α
0
))e
20
+ (rα
3
+ K(α
0
))e
30
+ (3.16)
+ (2α
1
−J(α
3
) + K(α
2
))e
32
+ (2α
2
−K(α
1
) + I(α
3
))e
13
+ (2α
3
−I(α
2
) + K(α
1
))e
21
_
.
Since the α
i
have no e
j
-components, σ(α) = 0 if and only if all these components
vanish. This occurs if and only if α
1
= −
1
r
I(α
0
), α
2
= −
1
r
J(α
0
), α
3
= −
1
r
K(α
0
) (since
as remarked in Section 3.3.1 these equations also guarantee that 2α
1
−J(α
3
)+K(α
2
) = 0
etc.), in which case it is clear that
σ(α) = 0 ⇐⇒α = σ
_
2(r+1)
r
α
0
_
.
This shows that the sequence E
0
k−2,r−2
→ E
1
k−1,r−1
→ E
2
k,r
is exact. Restricting to
E
1,r
k−1,r−1
and E
1,r−2
k−1,r−1
, we see that exactness holds at these spaces in the middle and
bottom sequences respectively of (3.15).
Consider α ∈ E
0
k−2,r−2
. Then
σ(α) =
1
2(r −1)
_
rαe
0
−I(α)e
1
−J(α)e
2
−K(α)e
3
_
.
Since these are linearly independent, σ(α) = 0 if and only if α = 0, and σ : E
0,r−2
k−2,r−2

E
1,r−2
k−1,r−1
is injective. Hence the bottom sequence 0 −→ E
0,r−2
k−2,r−2
σ
−→ E
1,r−2
k−1,r−1
σ
−→
E
2,r−2
k,r
−→0 is exact.
Finally, we show that the middle sequence 0 −→E
1,r
k−1,r−1
σ
−→E
2,r
k,r
σ
−→E
3,r
k+1,r+1
−→
0 is exact at E
2,r
k,r
, which is now sufficient to show that the sequence is exact.
Let β = β
1
ω
+
1
+ β
2
ω
+
2
+ β
3
ω
+
3
∈ E
2,r+
k,r
. Recall the Lie in condition (3.12) that β
must take the form β =
1
r
_
I(β
0

+
1
+ J(β
0

+
2
+ K(β
0

+
3
_
for some β
0
∈ E
0
k−2,r
. (The
1
r
-factor makes no difference here and is useful for cancellations.) Thus a general element
of E
2,r
k,r
is of the form
β + γ =
1
r
_
I(β
0

+
1
+ J(β
0

+
2
+ K(β
0

+
3
_
+ γ
1
ω

1
+ γ
2
ω

2
+ γ
3
ω

3
,
for β
0
, γ
j
∈ E
0,r
k−2,r
. A similar calculation to that of (3.16) shows that
σ(β + γ) = 0 ⇐⇒
_
¸
¸
_
¸
¸
_
(r + 2)β
0
+ I(γ
1
) + J(γ
2
) + K(γ
3
) = 0
(r + 2)γ
1
−I(β
0
) −J(γ
3
) + K(γ
2
) = 0
(r + 2)γ
2
+ I(γ
3
) −J(β
0
) −K(γ
1
) = 0
(r + 2)γ
3
−I(γ
2
) + J(γ
1
) −K(β
0
) = 0.
39
But this is exactly the Lie in condition (3.10) which we need for β
0
e
0

1
e
1

2
e
2

3
e
3
to be in E
1,r
k−1,r−1
, in which case we have
β + γ = σ
_
2(β
0
e
0
+ γ
1
e
1
+ γ
2
e
2
+ γ
3
e
3
)
_
.
This demonstrates exactness at E
2,r
k,r
and so the middle sequence is exact.
As a counterexample for the case r = 0 and k ≥ 4, consider α ∈ E
0
k−4,0
. Then
αe
0123
∈ E
4,0
k,0
and σ(αe
0123
) = 0, so σ : E
k,0
→ E
k+1,1
is not injective, which is exactly
the same as saying that the symbol sequence is not exact at E
k,0
. It is easy to see
that this counterexample does not arise when k = 0 or 2, and to show that the maps
σ : E
0,0
→E
1,1
and σ : E
2,0
→E
3,1
are injective.
As a counterexample for the case r = 1 and k ≥ 2, consider α ∈ E
0
k−2,0
. Then
αe
123
∈ E
3,0
k+1,1
and αe
123
∧ e
0
∈ E
4
k+2,0
. Thus σ(αe
123
) = 0. Since αe
123
has no e
0
-
components at all it is clear that αe
123
,= σ(β) for any β ∈ E
2
k,0
. Thus the symbol
sequence fails to be exact at E
k+1,1
. Again, it is easy to see that this counterexample
does not arise when k = 0, and to show that the sequence E
0,0
σ
−→ E
1,1
σ
−→ E
2,2
is
exact at E
1,1
.
This concludes our proof of Theorem 3.3.1.
3.4 Quaternion-valued forms on Hypercomplex Man-
ifolds
Let M be a hypercomplex manifold. Then M has a triple (I, J, K) of complex structures
which we can identify globally with the imaginary quaternions. Thus we have globally
defined operators which generate the sp(1)-action on Λ
k
T

M.
Consider also the quaternions themselves. Equation (2.6) describes the
Sp(1)GL(n, H)-representation on H
n
as V
1
⊗ E. In the case n = 1 this reduces to
the representation
H

= V
1
⊗V
1
, (3.17)
where we can interpret the left-hand copy of V
1
as the left-action (p, q) → pq, and the
right-hand copy of V
1
as the right-action (p, q) →qp
−1
, for q ∈ H and p ∈ Sp(1).
We can now use our globally defined hypercomplex structure to combine the Sp(1)-
actions on H and Λ
k
T

M. This motivates a thorough investigation of quaternion-valued
forms on hypercomplex manifolds. Consider, for example, quaternion-valued exterior
forms in the bundle E
k,r
=
n
k,r
V
r
. The Sp(1)-action on these forms is described by the
representation
H⊗E
k,r

= V
1
⊗V
1

n
k,r
V
r
.
Leaving the left H-action untouched, we consider the effect of the right H-action and the
hypercomplex structure simultaneously. This amounts to applying the operators
1 : α →I(α) −αi
1
, ¸ : α →J(α) −αi
2
and / : α →K(α) −αi
3
to α ∈ H⊗E
k,r
. Under this diagonal action the tensor product V
1

n
k,r
V
r
splits, giving
the representation
H⊗E
k,r

= V
1

n
k,r
(V
r+1
⊕V
r−1
). (3.18)
40
Each of these summands inherits the structure of a left H-module from the left-action
V
1
, which is not affected by our splitting.
This situation mirrors our discussion of real and complex forms on complex mani-
folds. There is a decomposition of real-valued forms, which is taken further when we
consider complex-valued forms. In the same way, considering quaternion-valued forms
on a hypercomplex manifold allows us to take our decomposition further.
This point of view turns out to be very fruitful. It will, over the next few chapters,
lead to quaternionic analogues of holomorphic functions and k-forms, the holomorphic
tangent and cotangent spaces, and complex Lie groups and Lie algebras.
The algebraic foundation for this geometry lies in considering objects like our left
H-modules in Equation (3.18). Each of the summands V
1

n
k,r
V
r±1
is a left H-module
which arises as a submodule of H ⊗ E
k,r

= (r + 1)
n
k,r
H
n
. Thus each summand is an
H-linear submodule of H
n
. In the next chapter we will introduce a new algebraic theory
which is based upon such objects.
41
Chapter 4
Developments in Quaternionic
Algebra
This chapter describes an algebraic theory which will be central to our description of
hypercomplex geometry. The theory is that of my supervisor, Dominic Joyce, and is
presented in [J1]. The basic objects of study are H-submodules U of H ⊗ R
n
. Joyce
shows that the inclusion ι
U
: U → H
n
is determined up to isomorphism by the H-
module structure of U and the choice of a real vector subspace U

⊂ U satisfying a
certain condition. The pair (U, U

) is an augmented H-module, or AH-module.
The most important discovery in [J1] is a canonical tensor product for AH-modules
with interesting properties. (Recall from Section 1.3 that the most obvious definitions of
a tensor product over the quaternions are not especially fruitful.) For two AH-modules
U ⊂ H
m
and V ⊂ H
n
, we can define a unique AH-module U⊗
H
V ⊂ H
mn
. The
operation ‘ ⊗
H
’ will be called the quaternionic tensor product. It has similar properties
to the tensor product over a commuting field; for example it is both associative and
commutative. This allows us to develop the algebra of AH-modules as a parallel to that
of vector spaces over R or C. This analogy is particularly strong for certain well-behaved
AH-modules which will be called stable AH-modules.
There are other algebraic operations which are equivalent to Joyce’s quaternionic ten-
sor product. A sheaf-theoretic point of view is presented by Quillen [Q], in which he dis-
covers a contravariant equivalence of tensor categories between AH-modules and regular
sheaves on a real form of CP
1
. This allows us to classify all AH-modules and determine
their tensor products. In the next chapter, we will see that the most important classes of
AH-modules are conveniently described and manipulated using Sp(1)-representations.
4.1 The Quaternionic Algebra of Joyce
The following is a summary of parts of Joyce’s theory of quaternionic algebra. The
interested reader should consult [J1] for more details and proofs.
4.1.1 AH-Modules
We begin by defining H-modules and their dual spaces. A (left) H-module is a real vector
space U with an action of H on the left which we write as (q, u) → q u or qu, such
that p(q(u)) = (pq)(u) for p, q ∈ H and u ∈ U. For our purposes, all H-modules will
42
be left H-modules. By dimU we will always mean the dimension of U as a real vector
space, even if U is an H-module.
We write U

for the dual vector space of U. If U is an H-module we also define the
dual H-module U
×
of linear maps α : U → H that satisfy α(qu) = qα(u) for all q ∈ H
and u ∈ U. If q ∈ H and α ∈ U
×
, define q α by (q α)(u) = α(u)q for u ∈ U. Then
q α ∈ U
×
, and U
×
is a (left) H-module. Dual H-modules behave just like dual vector
spaces.
Definition 4.1.1 [J1, Definition 2.2] Let U be an H-module. Let U

be a real vector
subspace of U. Define a real vector subspace U

of U
×
by
U

= ¦α ∈ U
×
: α(u) ∈ I for all u ∈ U

¦. (4.1)
Conversely, U

determines U

(at least for finite-dimensional U) by
U

= ¦u ∈ U : α(u) ∈ I for all α ∈ U

¦. (4.2)
An augmented H-module, or AH-module, is a pair (U, U

) such that if u ∈ U and α(u) = 0
for all α ∈ U

, then u = 0. We consider H to be an AH-module, with H

= I. AH-
modules should be thought of as the quaternionic analogues of real vector spaces.
Usually we will refer to U itself as an AH-module, assuming that U

is also given. If we
consider only the real part Re(α(u)) (for u ∈ U and α ∈ U
×
), we can interpret U
×
as
the dual of U as a real vector space, and then U

is the annihilator of U

. Thus if U is
finite-dimensional, dimU

+ dimU

= dimU = dimU
×
and an isomorphism U

= U
×
determines an isomorphism U/U


= U

.
Let U be an AH-module and let u, v ∈ U such that α(u) = α(v) for all α ∈ U

.
Then since α is a linear map we have α(u − v) = 0 for all α ∈ U

and it follows from
Definition 4.1.1 that u = v. Thus U is an AH-module if and only if each u ∈ U is
uniquely determined by the values of α(u) for α ∈ U

. In effect, the definition of an AH-
module demands that U

should not be too large. Definition 4.1.1 demands for any u ∈ U,
its H-linear span H u should not be entirely contained in U

, so dim(U

∩ H u) ≤ 3.
If V is an AH-module, we say that U is an AH-submodule of V if U is an H-submodule
of V and U

= U ∩V

. As U

is the restriction of V

to U, if α(u) = 0 for all α ∈ U

then
u = 0, so U is an AH-module. Define W to be the quotient H-module V/U and define
W

to be the real subspace (V

+ U)/U of W. We would like to define (W, W

) to be
the quotient AH-module V/U. However, there is a catch: W may not be an AH-module,
as the condition in Definition 4.1.1 may not be satisfied.
Example 4.1.2 [J1, Definition 6.1] Let Y ⊂ H
3
be the set Y = ¦(q
1
, q
2
, q
3
) : q
1
i
1
+
q
2
i
2
+ q
3
i
3
= 0¦. Then Y

= H
2
is a left H-module. Define a real subspace Y

= Y ∩ I
3
;
so Y

= ¦(q
1
, q
2
, q
3
) : q
j
∈ I and q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦. Then dimY = 8 and
dimY

= 5.
Let ν : Y → H, ν(q
1
, q
2
, q
3
) = i
1
q
1
+ i
2
q
2
+ i
3
q
3
. Then im(ν) = I and ker(ν) = Y

.
Since Y/Y


= (Y

)

, ν induces an isomorphism (Y

)


= I

= V
2
.
Here is the natural concept of linear map between AH-modules:
Definition 4.1.3 Let U, V be AH-modules and let φ : U → V be H-linear. We say
that φ is an AH-morphism if φ(U

) ⊂ V

. If φ is also an isomorphism of H-modules we
say φ is an AH-isomorphism.
43
One obvious question is whether we can classify AH-modules up to AH-isomorphism.
Real and complex vector spaces are classified by dimension, but clearly this is not true
for AH-modules, as there are several choices of U

for each H-module U

= H
n
. We will
return to this question in some detail later.
We note the following points:
• If φ : U → V and ψ : V → W are AH-morphisms, then ψ ◦ φ : U → W is an
AH-morphism.
• Let U, V be AH-modules and φ : U → V an AH-morphism. Define an H-linear
map φ
×
: V
×
→ U
×
by φ
×
(β)(u) = β(φ(u)) for β ∈ V
×
and u ∈ U. Then
φ(U

) ⊂ V

implies that φ
×
(V

) ⊂ U

.
• Let U be an AH-module. Then H⊗(U

)

is an H-module, with H-action p(q⊗x) =
(pq) ⊗ x. Define a map ι
U
: U → H ⊗ (U

)

by ι
U
(u) α = α(u), for u ∈ U and
α ∈ U

. Then ι
U
is H-linear, so that ι
U
(U) is an H-submodule of H⊗(U

)

.
Suppose u ∈ ker ι
U
. Then α(u) = 0 for all α ∈ U

, so that u = 0 as U is an
AH-module. Thus ι
U
is injective, and ι
U
(U)

= U.
• From Equation (4.2), it follows that ι
U
(U

) = ι
U
(U) ∩ (I ⊗ (U

)

). Thus the
AH-module (U, U

) is determined by the H-submodule ι
U
(U).
This shows as promised that every AH-module is isomorphic to a (left) submodule
of (H ⊗ R
n
, I ⊗ R
n
) for n = dim(U

)

. Example 4.1.2 shows how ths works for the
AH-module Y . As an abstract AH-module Y is isomorphic to H
2
and (Y

)


= V
2
. One
of the easiest and most symmetrical ways to obtain Y is as an 8-dimensional subspace
of H
3
= H⊗V
2
. We will find this version of events very useful in many situations, as we
shall see immediately.
4.1.2 The Quaternionic Tensor Product
Let U and V be AH-modules. Then they can be regarded as subspaces of H⊗(U

)

and
H ⊗ (V

)

respectively. Since the H-action on both of these is the same, we can paste
these AH-modules together to get a product AH-module. Here is the key idea of the
theory:
Definition 4.1.4 [J1, Definition 4.2] Let U, V be AH-modules. Then H⊗(U

)

⊗(V

)

is an H-module, with H-action p (q ⊗x ⊗y) = (pq) ⊗x ⊗y. Exchanging the factors of
H and (U

)

, we may regard (U

)

⊗ ι
V
(V ) as a subspace of H ⊗ (U

)

⊗ (V

)

. Thus
ι
U
(U) ⊗(V

)

and (U

)

⊗ι
V
(V ) are AH-submodules of H⊗(U

)

⊗(V

)

. We define
their intersection to be the quaternionic tensor product of U and V ,
U⊗
H
V = (ι
U
(U) ⊗(V

)

) ∩ ((U

)

⊗ι
V
(V )) ⊂ H⊗(U

)

⊗(V

)

. (4.3)
The vector subspace (U⊗
H
V )

is then given by (U⊗
H
V )

= (U⊗
H
V )∩(I⊗(U

)

⊗(V

)

)
and with this definition U⊗
H
V is an AH-module. The operation ⊗
H
will be called the
quaternionic tensor product.
44
A few words of explanation may be useful at this point. At the end of Chapter 1 we
saw that it is possible to define a sort of tensor product H
m

H
H
n

= H
mn
, but that this is
not really any different from taking tensor products over R. The theory of AH-modules
and the quaternionic tensor product is a way of taking the quaternionic behaviour into
account. How U

behaves in relation to the H-action determines a particular subspace
ι
U
(U) of H ⊗ R
n
. The quaternionic tensor product is the natural way of combining
these choices for AH-modules U ⊆ H ⊗ R
m
and V ⊆ H ⊗ R
n
into an AH-module
U⊗
H
V ⊆ H⊗R
mn
.
We also define the tensor product of two AH-morphisms:
Definition 4.1.5 Let U, V, W, X be AH-modules, and let φ : U → W and ψ : V → X
be AH-morphisms. Then φ
×
(W

) ⊂ U

and ψ
×
(X

) ⊂ V

. Taking the duals gives
maps (φ
×
)

: (U

)

→ (W

)

and (ψ
×
)

: (V

)

→ (X

)

. Combining these, we have a
map
id ⊗(φ
×
)

⊗(ψ
×
)

: H⊗(U

)

⊗(V

)

→H⊗(W

)

⊗(X

)

. (4.4)
Define φ⊗
H
ψ : U⊗
H
V →W⊗
H
X to be the restriction of id ⊗(φ
×
)

⊗(ψ
×
)

to U⊗
H
V .
Then φ⊗
H
ψ is an AH-morphism from U⊗
H
V to W⊗
H
X . This is the quaternionic
tensor product of φ and ψ.
It can be proved [J1, Lemma 4.3] that there are canonical AH-isomorphisms
H⊗
H
U

= U, U⊗
H
V

= V ⊗
H
U and (U⊗
H
V )⊗
H
W

= U⊗
H
(V ⊗
H
W). (4.5)
This tells us that ⊗
H
is commutative and associative, and that the AH-module H acts
as an identity element for ⊗
H
. Since ⊗
H
is commutative and associative we can define
symmetric and antisymmetric products of AH-modules:
Definition 4.1.6 [J1, 4.4] Let U be an AH-module. Write

k
H
U for the product
U⊗
H

H
U of k copies of U, with

0
H
U = H. Then the k
th
symmetric group S
k
acts
on

k
H
U by permutation of the U factors in the obvious way. Define S
k
H
U and Λ
k
H
U
to be the AH-submodules of

k
H
U which are symmetric and antisymmetric respectively
under the action of S
k
.
Much of the algebra that works over R or C can be adapted to work over H, using
AH-modules and the quaternionic tensor product instead of vector spaces and the real
or complex tensor product. There are, however, many situations where the quaternionic
tensor product behaves differently from the standard real or complex tensor product.
For example, the dimension of U⊗
H
V can behave strangely. It can vary discontinuously
under smooth variations of U

or V

, and it is possible to have U⊗
H
V = ¦0¦ when
both U and V are non-zero. If φ and ψ are both injective AH-morphisms, it is possible
to prove [J1, Lemma 7.4] that φ⊗
H
ψ is also injective. However, if φ and ψ are both
surjective then φ⊗
H
ψ is not necessarily surjective.
Given u ∈ U and v ∈ V it is not possible in general to define an element u⊗
H
v ∈
U⊗
H
V . However, we do have the following special case:
Lemma 4.1.7 [J1, 4.6] Let U, V be AH-modules, and let u ∈ U and v ∈ V be nonzero.
Suppose that α(u)β(v) = β(v)α(u) ∈ H for every α ∈ U

and β ∈ V

. Define an
45
element u⊗
H
v of H⊗(U

)

⊗(V

)

by (u⊗
H
v) (α ⊗β) = α(u)β(v) ∈ H. Then u⊗
H
v
is a nonzero element of U⊗
H
V .
It is easy to visualise how this Lemma ‘works’. If α(u)β(v) = β(v)α(u) ∈ H for every
α ∈ U

and β ∈ V

, then α(u) and β(v) must both be in some commutative subfield
C
q
⊂ H. This is the same as saying that
ι
U
(u) ∈ C
q
⊗(U

)

and ι
V
(v) ∈ C
q
⊗(V

)

,
and the element u⊗
H
v is just the complex tensor product ι
U
(u) ⊗
C
q
ι
V
(v) ∈ C
q
⊗(U

)


(V

)

. So Lemma 4.1.7 tells us that on complex subfields of H, the quaternionic tensor
product is the same as the complex tensor product.
4.1.3 Stable and Semistable AH-Modules
In this section we define two special sorts of AH-modules, which we shall call semistable
and stable. These AH-modules behave particularly well, and we can exploit their ‘nice’
properties to cement further the analogy between real and quaternionic algebra.
Definition 4.1.8 [J1, ¸8] Let U be a finite-dimensional AH-module. We say that U is
semistable if it is generated over H by the subspaces U

∩ qU

for q ∈ S
2
.
We can describe semistable AH-modules by the following property:
Lemma 4.1.9 Suppose that U is semistable, with dimU = 4j and dimU

= 2j + r,
for integers j, r. Then U

+ qU

= U for generic q ∈ S
2
. Thus r ≥ 0.
The next logical step is to require this property for all q ∈ S
2
, motivating the
following definition:
Definition 4.1.10 Let U be a finite-dimensional AH-module. We say that U is stable
if U = U

+ qU

for all q ∈ S
2
.
In effect, our definitions of stable and semistable AH-modules act as a balance to
Definition 4.1.1 by demanding that U

should not be too small. Many of the properties
of semistable and stable AH-modules can be characterised by exploring the properties
of a particularly important type of AH-module:
Definition 4.1.11 Let q ∈ I ¸ ¦0¦. Define an AH-module X
q
by X
q
= H, X

q
= ¦p ∈
H : pq = −qp¦. In other words, X

q
is the subspace of H which is perpendicular to C
q
with respect to the standard scalar product.
We quote the following results, mainly taken from [J1, ¸8]:
• X
q
is semistable, but not stable.
• X
q
= X
λq
for all λ ∈ R ¸ ¦0¦, but for p ,= λq, X
p
and X
q
are not AH-isomorphic
to one another. There is thus a distinct AH-module X
q
given by each pair of
antipodal points ¦q, −q¦ for q ∈ S
2
.
46
• There is a canonical AH-isomorphismX
q

H
X
q

= X
q
, but if p ,= λq then X
p

H
X
q
=
¦0¦.
• Let χ
q
: X
q
→ H be the identity map on H. Then χ
q
and id ⊗
H
χ
q
: U⊗
H
X
q

U⊗
H
H

= U are injective AH-morphisms.
• There is an isomorphism (U⊗
H
X
q
)


= U

∩ qU


= C
n
q
. It follows that U⊗
H
X
q

=
nX
q
.
• Therefore if q ∈ S
2
and U is an AH-module with dimU = 4j and dimU

= 2j +r,
then U⊗
H
X
q

= nX
q
with n ≥ r .
• If U is semistable then U⊗
H
X
q

= rX
q
for generic q ∈ S
2
, by Lemma 4.1.9.
• An AH-module U is stable if and only if U⊗
H
X
q

= rX
q
for all q ∈ S
2
.
• It follows that if U and V are stable AH-modules then U⊗
H
V ⊗
H
X
q

= U⊗
H
(sX
q
)

=
rsX
q
for all q ∈ S
2
, so using the associativity of the quaternionic tensor we infer
that U⊗
H
V is a stable AH-module.
The AH-module X
q
is an important bridge from quaternionic to complex algebra. In
effect, X
q
= C
q
⊕C

q
, and the operation ‘⊗
H
X
q
’ converts an AH-module U into copies
of X
q
, so it turns the AH-module structure on U which is quaternionic information into
a set of what are effectively complex vector spaces.
It is clear that all stable AH-modules are semistable. There is a sense in which the
X
q
’s are the ‘only’ class of AH-modules which are semistable but not stable, due to the
following Proposition:
Proposition 4.1.12 [J1, 8.8]: Let V be a finite-dimensional AH-module. Then V is
semistable if and only if V

= U ⊕(

l
i=1
X
q
i
), where U is stable and q
i
∈ S
2
.
Joyce also shows that generic AH-modules with appropriate dimensions are stable or
semistable:
Lemma 4.1.13 [J1, 8.9] Let j, r be integers with 0 ≤ r ≤ j. Let U = H
j
and let U

be a real vector subspace of U with dimU

= 2j + r. For generic subspaces U

, (U, U

)
is a semistable AH-module. If r > 0 then for generic subspaces U

, (U, U

) is a stable
AH-module.
The benefits of working with stable and semistable AH-modules become increasingly
apparent as one becomes more familiar with the theory. For now, we will quote the
following theorems:
Theorem 4.1.14 [J1, 9.1] Let U and V be stable AH-modules with
dimU = 4j, dimU

= 2j + r, dimV = 4k and dimV

= 2k + s. (4.6)
Then U⊗
H
V is a stable AH-module with dim(U⊗
H
V ) = 4l and dim(U⊗
H
V )

= 2l +t,
where l = js + rk −rs and t = rs .
47
Proof.(Sketch of formula for total dimension) The proof works along the following lines.
If dimU = 4j and dimU

= 2j +r, then dim(U

)

= 2j −r and similarly dim(V

)

=
2k −s. So dim(H⊗(U

)

⊗(V

)

) = 4(2j −r)(2k −s).
Let A = ι
U
(U) ⊗ (V

)

and B = (U

)

⊗ ι
V
(V ), so that U⊗
H
V = A ∩ B. Then
dimA = 4j(2k − s) and dimB = 4k(2j − r), and dim(A + B) = dimA + dimB −
dim(A ∩ B). Now, if r, s ≥ 0 then dimA + dimB ≥ dim(H⊗(U

)

⊗(V

)

), and so if
the subspaces A and B are suitably transverse in H ⊗ (U

)

⊗ (V

)

we expect that
A + B = H⊗(U

)

⊗(V

)

, in which case
dim(U⊗
H
V ) = dimA + dimB −dim(H⊗(U

)

⊗(V

)

)
= 4(js + rk −rs).
The rest of Joyce’s proof consists of showing that if U and V are stable then this
intersection is transverse and finding (U⊗
H
V )

, from which it is easy to see that U⊗
H
V
is stable.
In fact, these dimension formulae still hold if V is only semistable. Theorem 4.1.14
and Proposition 4.1.12 combine to give the following:
Corollary 4.1.15 [J1, 9.3] Let U, V be semistable AH-modules. Then U⊗
H
V is
semistable.
Thus both stable and semistable AH-modules form subcategories of the tensor cat-
egory of AH-modules, closed under direct and tensor products. Let U be a stable
AH-module, with dimU = 4j and dimU

= 2j + r. We define r to be the virtual
dimension of U. Then Proposition 4.1.12 shows that the virtual dimension of U⊗
H
V is
the product of the virtual dimensions of U and V . We end this section by quoting the
following result:
Proposition 4.1.16 [J1, 9.6] Let U be a stable AH-module, with dimU = 4j and
dimU

= 2j + r. Let n be a positive integer. Then S
n
H
U and Λ
n
H
U are stable AH-
modules, with dim(S
n
H
U) = 4k, dim(S
n
H
U)

= 2k+s, dim(Λ
n
H
U) = 4l and dim(Λ
n
H
U)

=
2l + t, where
k = (j−r)
_
r+n−1
n −1
_
+
_
r+n−1
n
_
, s =
_
r+n−1
n
_
, l = (j−r)
_
r−1
n−1
_
+
_
r
n
_
, t =
_
r
n
_
.
4.2 Duality in Quaternionic Algebra
The objects which are naturally dual to AH-modules are called SH-modules. Under
certain circumstances an SH-module can also be regarded as an AH-module. In this
case we obtain interesting algebraic results which use dual AH-modules to tell us about
AH-morphisms between AH-modules.
48
4.2.1 SH-modules
In this section we will describe the class of objects which are dual to AH-modules. These
will be called strengthened H-modules, or SH-modules. SH-modules are introduced by
Quillen in [Q], as a link between sheaves and AH-modules. Being very much part of
quaternionic algebra rather than sheaf theory, we discuss them in this section.
Let (U, U

) be an AH-module. Then u ∈ U is completely determined by the values
of α(u) for α ∈ U

, and if u is determined then so is β(u) for all β ∈ U
×
. So for all
β ∈ U, β(u) is determined by the action of U

on u. Since the only other structure
present is the H-action, each β ∈ U
×
must be an H-linear combination of elements
of U

, so U
×
is generated over H by U

. The converse is clearly true as well — if
every β ∈ U
×
is of the form q α for some α ∈ U

, then (U, U

) is an AH-module by
Definition 4.1.1. The natural dual to an AH-module is thus an H-module equipped with
a generating real subspace.
Definition 4.2.1 Let Q be a left H-module and Q

a real linear subspace of Q. We
say that the pair (Q, Q

) is a strengthened H-module or SH-module if Q is generated
over H by Q

.
If U = (U, U

) is an AH-module then (U
×
, U

) is an SH-module, the SH-module
corresponding to U. Just as we sometimes write U for the AH-module (U, U

), we will
often write U
×
for the SH-module (U
×
, U

).
A further link between these two ideas is provided by choosing a (hyperhermitian)
metric on the AH-module (U, U

), giving an H-module isomorphism U

= U
×
. This
in turn identifies U

with (U

)

, and so realises (U

)

as a subspace of U which is
perpendicular to U

. We see that choosing a metric gives us a decomposition U

=
U

⊕(U

)

, where (U, (U

)

) is an SH-module. Every AH-module can thus be regarded
as an SH-module — the point of view depends on whether we think of U

or (U

)

as the
‘special’ subspace. The simplest example is that of the quaternions themselves: we can
regard them as the AH-module (H, I) or the SH-module (H, R), and these definitions
are exactly equivalent.
We define morphisms and quaternionic tensor products for SH-modules, by taking
the definitions from their corresponding AH-modules: the whole theory works in exactly
the same way. For example, if (U, U

) and (V, V

) are AH-modules and φ : U →V is an
AH-morphism, then (U
×
, U

) and (V
×
, V

) are SH-modules and the dual H-morphism
φ
×
: V
×
→ U
×
satisfies φ
×
(V

) ⊆ U

, which makes φ
×
an SH-morphism. We write
U
×

H
V
×
for the quaternionic tensor product of two SH-modules; in other words we
define
U
×

H
V
×
= (U⊗
H
V )
×
. (4.7)
Another useful example is given by stable and semistable AH-modules. An AH-
module is stable (respectively semistable) if and only if it satisfies the identity U =
U

+ qU

for all (respectively for generic) q ∈ I. But
U

+ qU

= U ⇐⇒(U

)

∩ q(U

)

= ¦0¦,
where (U

)

is the subspace orthogonal to U

with respect to a (hyperhermitian) metric
on U. Since there is an SH-isomorphism (U, (U

)

)

= (U
×
, U

), we see that
U

+ qU

= U ⇐⇒U

∩ qU

= ¦0¦.
49
Definition 4.2.2 An SH-module is called stable (respectively semistable) if and only if
it has the property that U

∩ qU

= ¦0¦ for all (respectively for generic) q ∈ I.
This is sometimes easier to demonstrate than the property for the corresponding AH-
modules. We shall work happily with either AH-modules or SH-modules according to
the needs of each situation, since their theories are interchangeable.
4.2.2 Dual AH-modules
In this section we take a new step and ask what happens if we consider (U
×
, U

) as an
AH-module — the dual AH-module of (U, U

). There are immediate attractions to this
approach. If V is a vector space over the commutative field F then we define the dual
space V

to be the space of F-linear maps φ : V → F. In the same way, if U is an
H-module we define U
×
to be the space of H-linear maps φ : U → H. Once we also
have a real subspace U

⊂ U we define U

to be the set of maps
U

= ¦α ∈ U
×
: α(u) ∈ I for all u ∈ U

¦.
But an H-linear map α : U → H such that α(U

) ⊆ I is precisely an AH-morphism
from U into H. Thus the space (U
×
, U

) consists of two sets of maps φ : U → H,
namely the H-linear maps and AH-morphisms respectively. This suggests that defining
(U
×
, U

) to be the dual AH-module of (U, U

) could be a good quaternionic analogue
of the concept of a dual vector space in real or complex algebra.
There is an obvious possible catch: (U
×
, U

) might not even be an AH-module! Thus
if we are to talk about dual AH-modules, we need to discern which AH-modules have
well-defined duals. We have in fact already done this in Section 4.2.1.
Lemma 4.2.3 The AH-module (U, U

) has a well defined dual AH-module (U
×
, U

) if
and only if (U, U

) is also an SH-module.
Proof. In Section 4.2.1 we showed that (U, U

) is an AH-module if and only if (U
×
, U

)
is an SH-module. We simply reverse this argument: (U
×
, U

) is an AH-module if and
only if the AH-module (U, U

) is also an SH-module.
Definition 4.2.4 Let U be an H-module and U

a real subspace of U . We say that
the pair (U, U

) is a strengthened augmented H-module, or SAH-module, if (U, U

) is
both an AH-module and an SH-module.
A comprehensive way to sum this up is to say that (U, U

) is an AH-module if
and only if it has no submodule isomorphic to (H, H) , an SH-module if and only if it
has no submodule isomorphic to (H, ¦0¦), and an SAH-module if and only if it has no
submodule isomorphic to either (H, H) or (H, ¦0¦). Unless otherwise stated, when we
refer to properties of an SAH-module such as stability, we mean this in terms of the
structure of U as an AH-module.
Example 4.2.5 Any semistable AH-module U has the property that U = U

+ qU

for generic q ∈ I, so U is also an SH-module and so an SAH-module.
50
Definition 4.2.6 Let U = (U, U

) be an AH-module which is also an SAH-module.
Then we define U
×
= (U
×
, U

) to be the dual AH-module of U.
With this definition, it is clear that U
×
is also an SAH-module. For finite dimensional
U, there are canonical isomorphisms U

= (U
×
)
×
and (U
×
)


= U

.
Definition 4.2.7 Let (U, U

) be an AH-module. Then U is called antistable if and
only if U

∩ qU

= ¦0¦ for all q ∈ I.
It is clear that antistable AH-modules are almost always dual to stable AH-modules.
The only irreducible exception is the AH-module (H, ¦0¦), which we regard as an anti-
stable AH-module in spite of the fact that its dual (H, H) is not an AH-module.
4.2.3 Spaces of AH-morphisms and Duality
The space U

is, as we have remarked, the space of AH-morphisms φ : U → H. This
fact is an example of a more general result which makes the theory of dual spaces
in quaternionic algebra particularly useful. Let A and B be (free, finite dimensional)
modules over the commutative ring R, and let Hom
R
(A, B) denote the space of R-linear
maps φ : A →B. It is well-known that there is a canonical isomorphism Hom
R
(A, B)

=
A


R
B.
There is an analogous result in quaternionic algebra. We start with the following
definition:
Definition 4.2.8 Let U and V be AH-modules. Then Hom
AH
(U, V ) denotes the
space of AH-morphisms from U into V .
Here is the main result of this section:
Theorem 4.2.9 Let U be an SAH-module and V be an AH-module. Then there is a
canonical isomorphism
Hom
AH
(U, V )

= (U
×

H
V )

.
Proof. Let φ ∈ (U
×

H
V )

. Then
φ ∈ (ι
U
×(U
×
) ⊗(V

)

) ∩ ((U

)

⊗ι
V
(V )) ∩ (I ⊗(U

)

⊗(V

)

),
or equivalently
φ ∈ (ι
U
×(U

) ⊗(V

)

) ∩ ((U

)

⊗ι
V
(V

)).
Consider φ ∈ ι
U
×(U

) ⊗ (V

)

. The mapping ι
U
× identifies U

with ι
U
×(U

). Using
this and the canonical isomorphism of real vector spaces Hom(A, B)

= A

⊗ B we see
that φ is exactly equivalent to an real linear map
Φ : (U

)

→(V

)

,
which in turn is equivalent to an H-linear map
Φ
H
: H⊗(U

)

→H⊗(V

)

,
51
which by definition is an AH-morphism.
Every AH-morphism ψ : U →V is equivalent to an AH-morphism Ψ : H⊗(U

)


H ⊗ (V

)

with the property that Ψ : ι
U
(U) → ι
V
(V ). Using the fact that φ ∈
(U

)

⊗ι
V
(V

) and the natural identification U


= ι
U
(U

) we see that
Φ
H
: ι
U
(U

) →ι
V
(V

).
Since U is an SAH-module, ι
U
(U) is generated over H by ι
U
(U

), from which it follows
that Φ
H
: ι
U
(U) → ι
V
(V ). Thus Φ
H
is equivalent to an AH-morphism from U to
V . Reversing these steps, we can construct an element of (U
×

H
V )

from each AH-
morphism from U to V .
The quaternionic tensor product can be used in this way to tell us about spaces of
AH-morphisms, which is another piece of evidence suggesting that Joyce’s definition of
the quaternionic tensor product is the right one for AH-modules.
4.3 Real Subspaces of Complex Vector Spaces
We have seen that choosing different real subspaces U

of an H-module U

= H
n
gives
rise to different algebraic properties. As always with the quaternions, it is useful to
compare this situation with that of the complex numbers. As an instructive example
(and a bit of light relief!) we shall give a classification of real linear subspaces of complex
vector spaces up to complex linear isomorphism. In other words, we classify the possible
orbits of a subspace R
k
of C
n
under the action of GL(n, C). This is not difficult, though
as far as the author can tell both the problem and its solution are original.
Theorem 4.3.1 Let U


= R
k
be a real linear subspace of C
n
. Then we can choose a
complex basis ¦e
j
: j = 1, . . . , n¦ of C
n
such that
U

= ¸e
1
, . . . , e
p
, ie
1
, . . . , ie
q
)
R
,
where q ≤ p ≤ n, p + q = k.
Proof. Both U

+ iU

and U

∩ iU

are complex subspaces of C
n
. Let U

+ iU


= C
p
and let U

∩ iU


= C
q
. Then dim
R
(U

) = p + q.
Choose a complex basis ¦e
1
, . . . , e
q
¦ for U

∩ iU

. Then ¦e
1
, ie
1
, . . . , e
q
, ie
q
¦ is a
real basis for U

∩ iU

. Extend this to a real basis ¦e
1
, ie
1
, . . . , e
q
, ie
q
, e
q+1
, . . . , e
p
¦ for
U

. Then ¦e
1
, . . . , e
p
¦ spans U

+ iU


= C
p
(over C), and so ¦e
1
, . . . , e
p
¦ is linearly
independent over C. The result follows.
It is easy to see from this theorem that choosing a real subspace of C
n
is always
compatible with the complex structure, in the sense that each basis vector of the real
subspace U

can be chosen to lie in one and only one copy of C. The pair (U, U

)

=
(C
n
, R
k
) can always be completely reduced to a direct sum of copies of C, each of which
contains 0, 1 or 2 basis vectors for U

.
The situation is very different for H-modules. For example, consider the AH-module
(U, U

) with
U = H
2
and U

= ¸(1, 0)(0, 1), (i
1
, i
2
)).
52
There is no way to decompose U into two separate copies of H, the first of which contains
2 basis vectors for U

and the second of which contains the remaining basis vector.
This motivates the following definition:
Definition 4.3.2 An AH-module (U, U

) is irreducible if and only if it cannot be
written as a direct sum of two non-trivial AH-modules, i.e. there are no two non-trivial
AH-modules (U
1
, U

1
), (U
2
, U

2
) such that (U, U

) = (U
1
⊕U
2
, U

1
⊕U

2
).
The classification of irreducible AH-modules up to AH-isomorphism is a much more
difficult problem than its analogue for complex vector spaces. It will be addressed in the
next section using a class of algebraic objects called K-modules.
4.4 K-modules
A K-module is an algebraic object based on a pair of complex vector spaces. Real and
quaternionic vector spaces, as so often, occur as complex vector spaces with particular
structure maps.
Definition 4.4.1 [Q, 4.1] A K-module is a pair (W, V ) of (finite dimensional) complex
vector spaces together with a linear map e : W → H ⊗ V , where H

= C
2
is the basic
representation of GL(2, C).
The reason why K-modules are important to quaternionic algebra is that in the
presence of suitable structure maps, some K-modules are equivalent to AH-modules.
This allows us to use the classification of irreducible K-modules, a problem with a known
solution, to write down all irreducible AH-modules very explicitly. This link between
K-modules and AH-modules was discovered by Quillen [Q]. Quillen is more interested in
an interpretation of quaternionic algebra in terms of sheaves over the Riemann sphere, a
powerful theory which we will review in the next section. We follow a slightly different
approach from that in Quillen’s paper to obtain a more immediate link between K-
modules and AH-modules.
A K-module e : W →H ⊗V is called indecomposable if it cannot be written as the
direct sum of two non-trivial K-modules. A K-module morphism is a map φ : (W
1
e
1

H ⊗ V
1
) −→ (W
2
e
2
→ H ⊗ V
2
) which respects the K-module structure. A K-module
can equivalently be defined as a pair of linear maps e
1
, e
2
: W → V . In this guise, a
K-module is a representation of the Kronecker quiver. We recover the first definition by
setting e(w) = h
1
⊗e
1
(w) +h
2
⊗e
2
(w) where ¦h
1
, h
2
¦ is a basis for H. Representations
of the Kronecker quiver are discussed in Benson’s book [Ben, ¸4.3]. The important result
is the following classification theorem of Kronecker (which we have summarised slightly),
which shows that every indecomposable K-module is isomorphic to one of three basic
types.
Theorem 4.4.2 [Ben, p. 101] Let e
1
, e
2
: W →V be a pair of linear maps constituting
an indecomposable K-module. Then one of the following holds:
(i) The vector spaces W and V have the same dimension. In this case, if det e
1
,= 0
53
then e
1
and e
2
can be written in the form
e
1
→id e
2

_
_
_
_
_
α 0
1 α
.
.
.
.
.
.
0 1 α
_
_
_
_
_
.
If det e
1
= 0 a modification is necessary which in some sense corresponds to the ‘rational
canonical form at infinity’.
(ii) The dimension of W is one larger than the dimension of V , and bases may be chosen
so that e
1
and e
2
are represented by the matrices
e
1
=
_
_
_
1 0 0
.
.
.
.
.
.
0 1 0
_
_
_
e
2
=
_
_
_
0 1 0
.
.
.
.
.
.
0 0 1
_
_
_
.
(iii) The dimension of W is one smaller than the dimension of V , and bases may be
chosen so that e
1
and e
2
are represented by the transposes of the above matrices.
Definition 4.4.3 Define A
n,α
1
to be the indecomposable K-module of type (i) with
dimV = n and α on the leading diagonal of e
2
. (We write A
n,∞
1
for the case det e
1
=
0 ). Define A
n
2
to be the irreducible K-module of type (ii) with dimV = n. Define A
n
3
to be the irreducible K-module of type (iii) with dimV = n.
Let σ
H
be the standard quaternionic structure map on H defined by σ
H
(z
1
h
1
+
z
2
h
2
) = −¯ z
2
h
1
+¯ z
1
h
2
. This gives H⊗V the structure of a complex H-module. Our aim
is for the K-module e : W →H ⊗V to define a real subspace of a real H-module. This
is accomplished by compatible structure maps on W and V .
Definition 4.4.4 [Q, 11.2] An SK-module is a K-module e : W →H⊗V equipped with
antilinear operators σ
W
and σ
V
of squares 1 and −1 respectively, such that e σ
W
=

H
⊗σ
V
) e.
Suppose W →H ⊗V is an SK-module. Then σ
W
is a real structure on W, and its
set of fixed points is the real vector space W
σ
. The map σ
H
⊗σ
V
is also a real structure,
and in the same way we define the real vector space (H⊗V )
σ
. Then (H⊗V )
σ
is a real
H-module, whose H-module structure is inherited from that on H, and e(W
σ
) is a real
subspace of this H-module. In many circumstances an SK-module is therefore equivalent
to an AH-module.
Example 4.4.5 Consider the K-module A
2
2
which has dimW = 3 and dimV = 2.
Let ¦w
1
, w
2
, w
3
¦ be a basis for W and let ¦v
1
, v
2
¦ be a basis for V . We have
e(w
1
) = h
1
⊗v
1
, e(w
2
) = h
1
⊗v
2
+ h
2
⊗v
1
, e(w
3
) = h
2
⊗v
2
.
Let σ
V
be the standard quaternionic structure on V , so σ
V
(v
1
) = v
2
and σ
V
(v
2
) = −v
1
.
There is a compatible real structure σ
W
on W given by σ
W
(w
1
) = w
3
, σ(w
3
)
W
= w
1
54
and σ
W
(w
2
) = −w
2
, so that with these structure maps A
2
2
is an SK-module. We have
real vector spaces
W
σ
= ¸w
1
+ w
3
, iw
2
, i(w
1
−w
3
))
and
(H ⊗V )
σ
=
_
h
1
⊗v
2
−h
2
⊗v
1
i(h
1
⊗v
2
+ h
2
⊗v
1
)
h
1
⊗v
1
+ h
2
⊗v
2
i(h
1
⊗v
1
−h
2
⊗v
2
)
_
.
An H-module isomorphism (H⊗V )
σ

= H is obtained by mapping these basis vectors to
1, i
1
, i
2
and i
3
respectively. Under this isomorphism the real subspace e(W
σ
) is mapped
to the imaginary quaternions I. This demonstrates explicitly that the SK-module A
2
2
is
equivalent to the AH-module (H, I ) = H.
Every real subspace U

of a quaternionic vector space U can be obtained in this
fashion, so a classification of SK-modules gives a classification of AH-modules.
Corollary 4.4.6 Every indecomposable SK-module is isomorphic to one of the following:
(A) The direct sum of a pair of K-modules of type (i) of the form A
n,α
1
⊕A
n,−¯ α
−1
1
.
(B) An indecomposable K-module of type (ii) or (iii) with dimV even; in other words
A
2m
2
or A
2m
3
.
(C) An indecomposable K-module of type (ii) or (iii) with dimV odd, tensored with the
basic representation H equipped with its standard structure map σ
H
; in other words
A
2m+1
2
⊗H or A
2m+1
3
⊗H.
Proof. This follows from Theorem 4.4.2 and explicit calculations using standard structure
maps on the vector spaces V .
It remains to check which pairs (U, U

) arising in this fashion are AH-modules. As
noted earlier, a pair (U, U

) fails to be an AH-module if and only if it has a subspace
of the form (H, H). This pair is given by the SK-module H ⊗ A
1
3
. The ‘annihilating
K-module’ A
0
3
also fails to give an AH-module. Any SK-module containing neither of
these indecomposables is equivalent to an AH-module.
Here are some important facts about irreducible AH-modules which can now be
deduced:
• SK-modules of type (ii) correspond to stable AH-modules and SK-modules of type
(iii) correspond to antistable AH-modules.
• Indecomposable SK-modules of type (i) of the form A
1,α
1
⊕A
1,−¯ α
−1
1
correspond to
the semistable AH-modules X
q
.
• The irreducible stable AH-module corresponding to A
2m
2
has dimU = 4m and
dimU

= 2m + 1. Thus U

= H
m
and the virtual dimension of U is 1.
• The irreducible stable AH-module of the form H⊗A
2m+1
2
has dimU = 4(2m+1)
and dimU

= 4(m + 1). Thus U

= H
2m+1
and the virtual dimension of U is 2.
• This shows that there is an irreducible stable AH-module with virtual dimension
1 in every dimension and an irreducible stable AH-module with virtual dimension
2 in every odd dimension.
55
• The isomorphism class of an irreducible stable AH-module is thus uniquely deter-
mined by the dimension and virtual dimension of U.
4.5 The Sheaf-Theoretic approach of Quillen
Much of Joyce’s quaternionic algebra can be described using (coherent) sheaves over
the complex projective line CP
1
. This interpretation is due to Daniel Quillen [Q].
Quillen’s paper works by recognising that certain exact sequences of sheaf cohomology
groups are K-modules. Thus in the presence of certain structure maps, we obtain SH-
modules. (Quillen deals primarily with SH-modules rather than AH-modules.) Quillen
uses slightly different structure maps from those we used in the previous section to obtain
SK-modules, but the resulting theory is exactly the same.
The most interesting new result in this section is that the equivalence between sheaves
and SH-modules respects tensor products, enabling us to calculate the quaternionic
tensor product of two SH-modules from knowing the tensor products of the corresponding
sheaves. Thus by the end of this section we will have succeeded in classifying all AH-
modules and their tensor products.
4.5.1 Sheaves on the Riemann Sphere
We describe the algebraic geometry of (coherent) sheaves over CP
1
.
1
Quillen demon-
strates that every coherent sheaf over CP
1
is the direct sum of a holomorphic vector
bundle and a torsion sheaf (one whose support is finite).
2
These summands factorise
very neatly — every torsion sheaf is the sum of indecomposable sheaves supported at a
single point, and every holomorphic vector bundle is a sum of holomorphic line bundles.
We will describe the vector bundles first.
Every holomorphic line bundle over CP
n
is a tensor power of the hyperplane section
bundle L [GH, p. 145]. In the case n = 1, we use the open cover of CP
1
= C ∪ ¦∞¦
consisting of the two open sets U
0
= CP
1
¸¦∞¦ and U
1
= CP
1
¸¦0¦. A holomorphic line
bundle over CP
1
is determined by a holomorphic transition function ψ : U
0
∩U
1
→C

,
so ψ : C

→C

. Two transition functions ψ and ψ

determine the same line bundle if
and only if there exist non-vanishing holomorphic functions f, g : C

→C

such that
ψ

=
f
g
ψ.
Two functions are equivalent under this relation if and only if they have the same winding
number, so each line bundle on CP
1
is given by one of the transition functions g(z) = z
n
,
n ∈ Z. The line bundle given by the transition function z
n
is in fact

n
L. Following
standard notation, we write O(n) for the sheaf of its holomorphic sections. Thus O =
1
Background material can be found in [GH], which introduces sheaves and their cohomology [pp.
34-49], coherent sheaves [pp. 678-704], holomorphic vector bundles [pp. 66-71] and holomorphic line
bundles [pp. 132-139]. A more thorough exposition of the differential geometry of holomorphic vector
bundles, including many of the properties of sheaves used in Quillen’s paper, can be found in Kobayashi’s
book [K].
2
This uses the convention of identifying a holomorphic vector bundle with its sheaf of holomorphic
sections.
56
O(0) is the structure sheaf of CP
1
. It is easy to see (by multiplying the transition
functions together) that L
n
⊗L
m

= L
n+m
, or in sheaf-theoretic terms O(n) ⊗
O
O(m)

=
O(n + m).
Every holomorphic vector bundle over CP
1
can be written as a direct sum of these
line bundles, the summands being unique up to order.
3
Thus every holomorphic vector
bundle E is a sum of irreducible line bundles, and can be written E =


−∞
a
n
L
n
,
where the multiplicities a
n
are unique (though the decomposition itself may not be).
This leaves us to consider sheaves which are supported at a finite set of points, which
are called torsion sheaves. Quillen [Q, ¸2] demonstrates that every coherent sheaf over
CP
1
is the sum of a vector bundle and a torsion sheaf. Torsion sheaves themselves
split into sheaves supported at one point only. Let z ∈ CP
1
be such a point, and let
O
z
be the ring of germs of holomorphic functions at z. Define m
z
to be the unique
maximal ideal of O
z
consisting of germs of functions whose first derivative vanishes at
z. Every torsion sheaf splits into sheaves of the form O
z
/(m
z
)
n
, which we write O/m
n
z
by extending m
z
by O on the complement of z. We have the following Theorem:
Theorem 4.5.1 [Q, 2.3] Any coherent sheaf over CP
1
splits with unique multiplicities
into indecomposable sheaves of the form O(n) for n ∈ Z and O/m
n
z
for n ≥ 1 and
z ∈ CP
1
.
Cohomology Groups and Exact Sequences
There are various ways to calculate the cohomology groups of these sheaves.
4
The
method Quillen outlines uses the properties of exact sequences of sheaves. It is from the
maps in these sequences that we obtain K-modules and thence SH-modules.
Let H

= C
2
be the basic representation of GL(2, C). Then CP
1
can be identified
with the set of quotient lines of H and there is a basic exact sequence
0 →Λ
2
H ⊗O(−1) →H ⊗O →O(1) →0. (4.8)
Tensoring (over O) with the sheaf F and choosing an identification Λ
2
H

= C yields
the exact sequence
0 →F(−1) →H ⊗F →F(1) →0, (4.9)
where F(n) denotes the sheaf F ⊗
O
O(n).
Example 4.5.2 If we put F = O(n), n ≥ 0, we have the exact sequence
0 →O(n −1) →H ⊗O(n) →O(n + 1) →0. (4.10)
We know that H
0
(O)

= C (global holomorphic functions on CP(1) ) and that
H
0
(O(−1)) = 0. By the exact sequences of (4.10) and induction, it follows that
H
0
(O(n))

= S
n
(H) for n ≥ 0 and zero otherwise, and there is an exact sequence
of cohomology groups given by
0 →S
n−1
H →H ⊗S
n
H →S
n+1
H →0. (4.11)
3
This follows from the Harder-Narasimhan filtration of a holomorphic vector bundle E over any
Riemann surface M [K, p. 137]
4
For example [W, p.11], where H
0
(O(n)) is shown to be isomorphic to the (complex) vector space
of homogeneous polynomials of degree n in 2 variables.
57
Torsion sheaves can be dealt with in a similar fashion, using a resolution involving
the sheaf cohomology groups H
0
(F). It is easy to see that H
0
(O/m
n
z
)

= C
n
. We call
the sheaves O(n) where n ≥ 0, torsion sheaves, and sums thereof regular sheaves. For
all regular sheaves F, the first cohomology group H
1
(F) is zero.
This leaves the sheaves O(n) where n < 0 and sums thereof, which we call negative
vector bundles. These must be treated slightly differently, using the first cohomology
groups H
1
(F). Since H
0
(O(n)) = 0 for n < 0, we obtain an exact sequence
0 →H
1
(O(n −1)) →H ⊗H
1
(O(n)) →H
1
(O(n + 1)) →0, (4.12)
from which it follows that H
1
(O(n))

= S
−n−2
H for n ≤ −2 and zero otherwise.
Sheaves and K-modules
Consider the regular sheaf O(n) for n ≥ 0. The exact sequence (4.10) gives rise to
an injection of cohomology groups H
0
(O(n − 1)) → H
0
(H ⊗ O(n)) which takes the
form e : S
n−1
H → H ⊗ S
n
H. This is clearly a K-module, the irreducible K-module
A
n+1
3
. Similarly for the torsion sheaves O/m
n
x
there is an injection H
0
(O/m
n
x
(−1)) →
H
0
(H ⊗O/m
n
x
). This gives a K-module of type (i) and dimW = n. Thus we obtain a
K-module from each regular sheaf F which we call ξ
+
F.
Let O(n), n < 0 be a negative vector bundle. The exact sequence (4.12) gives a map
H
1
(O(n − 1)) → H ⊗ H
1
(O(n)) which is equivalent to the indecomposable K-module
A
−n−1
3
. Thus for any negative vector bundle G we obtain a K-module which we call
ξ

G.
Comparing the classifications of indecomposable sheaves and K-modules (Theorems
4.4.2 and 4.5.1), it is clear that these categories are equivalent. Quillen proves this in
detail [Q, ¸¸4,5] and uses the tensor product of sheaves F⊗
O
G to construct an equivalent
tensor product operation for K-modules [Q, ¸6].
The rest of the programme begins to take shape. Some sheaves will correspond to SK-
modules, from which we obtain SH-modules. Quillen formulates this slightly differently
from our treatment in Section 4.4. Let W → H ⊗ V be a K-module. Instead of a
real structure on W and a quaternionic structure on V , Quillen uses K-modules with
a quaternionic structure σ
W
on W and a real structure σ
V
on V , such that the map
e intertwines σ
W
and σ
H
⊗ σ
V
. In this situation, W and H ⊗ V are H-modules and
e is an H-linear map. He calls this structure a σK-module. A σK-module is not itself
an SH-module, but the inclusion of the real subspace V
σ
in the cokernel (H ⊗ V )/W
is an SH-module if the K-module has no submodule for which the map e is surjective,
in which case the K-module is called reduced.
There is a parallel description in terms of sheaves. Let σ : z → ¯ z
−1
be the antipodal
map on the Riemann sphere. This induces a map of sheaves σ

: F → σ

(F) which
we call the σ-transform, and allows us to define a ‘ σ-invariant sheaf’ or just ‘ σ-sheaf’.
For example, a torsion sheaf F is a σ-sheaf if it is supported at a finite set of points
which is preserved by the antipodal map σ — so it must consist of sheaves of the
form O/m
n
z
⊕O/m
n
σ(z)
, where σ

interchanges the two summands. Quillen investigates
σ-sheaves thoroughly, and discovers that:
58
Proposition 4.5.3 [Q, 10.7] Any σ-sheaf splits with unique multiplicities into the fol-
lowing irreducible σ-sheaves:
(1) O/(m
z
m
σ(z)
)
n
for any pair ¦z, σ(z)¦ of antipodal points and n ≥ 1.
(2) O(2m) for m ∈ Z.
(3) O(2m + 1) ⊗H for m ∈ Z.
The formal similarity between this result and Corollary 4.4.6 is clear. He also proves
that:
Proposition 4.5.4 [Q, 12.6] The categories of reduced σK-modules and SH-modules are
equivalent.
We can also use σ-sheaves to obtain SK-modules which lead directly to SH-modules.
Example 4.5.5 Let F be a regular σ-sheaf. Then we have the exact sequence
0 →H
0
(F) →H ⊗H
0
(F(1)) →H
0
(F(2)) →0, (4.13)
and H
0
(F) →H ⊗H
0
(F(1)) is an SK-module e : W →H ⊗V . Taking real subspaces
gives an SH-module which we call η
+
(F).
Example 4.5.6 Let G be a negative σ-vector bundle with no summand H ⊗ O(−1)
or O(−2). Then we have the exact sequence
0 →H
1
(G) →H ⊗H
1
(G(1)) →H
1
(G(2)) →0, (4.14)
Just as in the previous example, this sequence gives an SK-module
H
1
(G) →H ⊗H
1
(G(1)). (4.15)
We call this SH-module η

(G). We also define η

(O(−2)) = η

(H ⊗O(−1)) = 0.
If we have a σ-sheaf A = F+G, with F and G as above, then H
0
(G) = H
0
(G(1)) =
0 and H
1
(F) = H
1
(F(1)) = 0 ; so η
+
(G) = η

(F) = 0. In theory, we could combine
the functors η
+
and η

into a single functor η = η
+


, since η(A) = η
+
(F) +η

(G)
as required.
4.5.2 Sheaves and the Quaternionic Tensor Product
We have seen how the tensor product of sheaves encourages us to define a ‘reduced
tensor product’ operation for K-modules. It turns out that this tensor product agrees
remarkably with the quaternionic tensor product for SH-modules. This is a considerable
bonus from Quillen’s theory — the correspondences between σ-sheaves, SK-modules and
SH-modules allow us to compute tensor products in each category. The main theorem
is as follows:
Theorem 4.5.7 [Q, 7.1]
5
5
Quillen proves this theorem for the tensor product of K-modules — the version given here is obtained
by performing the simple translation into SH-modules.
59
Let F
i
be regular σ-sheaves and G
i
be negative σ-vector bundles. Then
η
+
F
1

H
η
+
F
2
= η
+
(F
1

O
F
2
),
η
+
F
1

H
η

G
1
= η

(F
1

O
G
1
)
and
η

G
1

H
η

G
2
= ¦0¦.
Since we will usually work with AH-modules, we will often find ourselves using this
theorem for the corresponding AH-modules, which of course take the form (η
+
F)
×
and


G)
×
.
Thus if F is a torsion σ-sheaf and G is a negative σ vector bundle, (η
+
F)
×

H


G)
×
= ¦0¦. If F = O(m) and G = O(n) (possibly tensored with H if m or n is
odd) with m ≥ 0 and n < −2 then

+
F)
×

H


G)
×
=
_
0 m + n ≥ −2
η

(F ⊗
O
G)
×
m + n ≤ −3
since η

(O(k)) = ¦0¦ for k ≥ −2.
The sheaf-theoretic approach is thus a very powerful tool for describing quaternionic
algebra. For example, it is possible to obtain the dimension theorems of Section 4.1.3
by translating known results about the degree and rank of the tensor product of two
sheaves.
60
Chapter 5
Quaternionic Algebra and
Sp(1)-representations
The representations of the group Sp(1) occur in so many different situations, from K¨ ahler
geometry to particle physics, that they are by far the most ubiquitous Lie group rep-
resentations in modern mathematical literature. Given this versatility, it is no surprise
that these representations are a powerful tool in quaternionic algebra, especially since
Sp(1) is just the group of unit quaternions and the Lie algebra sp(1) can be identified
with I.
In this chapter, we will see how stable and antistable AH-modules can be handled
using Sp(1)-representations. We encountered the germ of this idea in Section 3.4, where
we came across the space H ⊗ E
k,r
and its splitting into two H-submodules. Because
the Riemann sphere CP
1
can be described as the homogeneous space Sp(1)/U(1), the
holomorphic sections of line bundles over CP
1
are naturally Sp(1)-representations. The
structure maps necessary to define an SK-module e : W → H ⊗ V of type (ii) or (iii)
arise from Sp(1)-representations on W, H and V . Not only do Sp(1)-representations
underlie all of these phenomena — they also make the theory of quaternionic algebra
very easy to predict and manipulate.
This point of view turns out to have fruitful applications in hypercomplex geometry,
towards which our exposition is deliberately geared. We use representations to explain
the structure of stable AH-modules and their tensor products. We also investigate the
role of semistable AH-modules and their interaction with the Sp(1)-representation struc-
ture of stable AH-modules.
5.1 Stable AH-modules and Sp(1)-representations
5.1.1 Sp(1)-representations on the quaternions
As a motivating example, we review the case of the quaternions themselves, viewed as the
stable AH-module (H, I). Recall the description of the quaternions as a tensor product
of two Sp(1)-representations H

= V
1
⊗V
1
, given in Equation (3.17), where the left hand
61
copy of V
1
gives the left H-action, and the right-hand copy gives the right H-action.
1
In other words, we think of H as an Sp(1) Sp(1)-representation by defining
(p, q) : r →prq
−1
r ∈ H, p, q ∈ Sp(1) ⊂ H.
Consider now the action of the diagonal Sp(1)-subgroup ¦(q, q) : q ∈ Sp(1)¦ ⊂
Sp(1) Sp(1) on V
1
⊗ V
1
. The Clebsch-Gordon formula gives the splitting V
1
⊗ V
1

=
V
2
⊕V
0
(equivalent to the standard isomorphism V ⊗V

= S
2
V ⊕Λ
2
V ). Each of these
summands inherits a real structure from the real structure on V
1
⊗V
1
so we obtain the
splitting
V
1
⊗V
1

= V
2
⊕V
0
(5.1)
into real subspaces of dimensions three and one respectively, just as we would expect.
This is the same as taking the action by conjugation r → qrq
−1
, which as we know
preserves the splitting H

= I ⊕R. For the quaternions, the AH-module structure H


= I
and (H

)


= R is a concept which arises naturally when we take both the Sp(1) actions
into account. It is this account of the AH-module H which we will generalise to all stable
AH-modules.
This description of the quaternions is very similar to that of Example 4.4.5, where
H

= A
2
2
. In terms of Sp(1)-actions, the basic vector space H we used so much in the
previous chapter is simply a copy of the basic representation V
1
.
5.1.2 Notation for Several Sp(1)-representations
It will be a sound investment at this point to introduce some notation to help us keep
track of the structure of representations when we have several copies of Sp(1) acting on
a vector space. We have already encountered the action of Sp(1) Sp(1) on V
1
⊗ V
1
.
Here we have two copies of Sp(1) acting, so there is already the possibility of ambiguity
concerning which Sp(1) is acting on which V
1
. We remove this ambiguity by writing
upper-case superscripts with the groups and the representations, to make it clear which
group is acting on which vector space.
For left H-modules there will always be a left H-action to consider. We will denote
this by V
L
1
, and the copy of Sp(1) which acts on this factor by Sp(1)
L
. Other copies of
Sp(1) and other representations will be labelled with the letters M, N etc. So we would
write the above example as
Sp(1)
L
Sp(1)
M
acting on V
L
1
⊗V
M
1
.
When we decompose such a representation using the Clebsch-Gordon formula, we
are decomposing the action of the diagonal subgroup ¦(q, q)¦ ⊂ Sp(1)
L
Sp(1)
M
. We
will call this subgroup Sp(1)
LM
, thus stating explicitly of which two groups this is the
diagonal subgroup. Similarly, we can combine superscripts for the representations to
write
V
L
1
⊗V
M
1

= V
LM
2
⊕V
LM
0
.
1
We are talking about the representation V
1
⊗ V
1
as a real representation on R
4
, implicitly using
the induced map σ
1
⊗σ
1
as a real structure on V
1
⊗V
1
. For more details, refer to Section 1.2.1.
62
This book-keeping comes into its own when we come to consider tensor products of
many Sp(1)-representations. For example, if we have three copies of Sp(1) acting, we
write this as
Sp(1)
L
Sp(1)
M
Sp(1)
N
acting on V
L
1
⊗V
M
j
⊗V
N
k
.
In this situation there are various diagonal actions we could be interested in, and we
can join the superscripts as above to indicate exactly which one we are considering. For
example, supposing we want to restrict to the diagonal subgroup in the first two copies of
Sp(1), i.e. ¦(q, q)¦ Sp(1)
N
. We denote this subgroup Sp(1)
LM
Sp(1)
N
. We combine
superscripts for the representations in the same way, so that we now have
Sp(1)
LM
Sp(1)
N
acting on (V
LM
j+1
⊕V
LM
j−1
) ⊗V
N
k
.
If, however, we considered the diagonal subgroup of the first and last copies of Sp(1), we
would write this as
Sp(1)
LN
Sp(1)
M
acting on (V
LN
k+1
⊕V
LN
k−1
) ⊗V
M
j
.
This provides an unambiguous and (it is hoped) easy way to understand tensor prod-
ucts of several representations and their decompositions into irreducibles under different
diagonal actions.
5.1.3 Irreducible stable AH-modules
We have described the quaternions as an Sp(1)
L
Sp(1)
M
-representation. This allows
us to interpret the primed part H


= I as a representation of the diagonal subgroup
of Sp(1)
LM
⊂ Sp(1)
L
Sp(1)
M
. In this section we will demonstrate how this idea can
be adapted to describe more general stable AH-modules. The basic idea is exactly the
same — a stable AH-module is a (real) Sp(1)
L
Sp(1)
M
-representation V
L
1
⊗ W
M
,
where W
M
=

k
1
a
j
V
M
j
. The left H-action is given by the action of the left subgroup
Sp(1)
L
⊂ Sp(1)
L
Sp(1)
M
. The primed part is then a representation of the diagonal
subgroup Sp(1)
LM
⊂ Sp(1)
L
Sp(1)
M
.
Let (U, U

) be an irreducible stable (or antistable) AH-module. Let M be the group
of AH-automorphisms of U whose (real) determinant is equal to 1. Using Theorem
4.2.9, we see that M = ±1 if the virtual dimension U is 1, and M

= Sp(1) if the
virtual dimension of U is 2.
Consider now the more general group G of real linear isomorphisms φ : U →U such
that:
• φ(U

) = U

,
• The (real) determinant of φ is 1,
• There exists some q ∈ Sp(1) such that φ(pu) = (qpq
−1
)φ(u) for all p ∈ H, u ∈ U.
Then G is a compact Lie group of which M is a normal subgroup. Because of this the
Lie algebra of G splits into two orthogonal ideals
g = sp(1) ⊕m
63
and the exponential map determines a homomorphism ρ : Sp(1) →G. Since the elements
of G map U

to itself, ρ is a representation of Sp(1) on U

.
Since every stable AH-module is a sum of such irreducibles, there is an action of
Sp(1) on U

for all stable AH-modules U, and the irreducible decomposition of U as
an AH-module determines the irreducible decomposition of the Sp(1)-action on U

, and
vice versa. As we shall see, there is a unique irreducible stable AH-module for each
irreducible Sp(1)-representation.
Let U be a stable AH-module and suppose that U

is preserved by some diagonal
action of Sp(1). This diagonal action will be the result of the left H-action on V
1
and
some other Sp(1)-action on U. Thus our AH-modules will follow the basic form
U = V
L
1
⊗V
M
m
(5.2)
as a representation of the group Sp(1)
L
Sp(1)
M
. The primed part of U is then the
V
LM
n+1
summand in the decomposition
V
L
1
⊗V
M
n

= V
LM
n+1
⊕V
LM
n−1
.
For example, recall the splitting
H⊗E
k,r

= V
1
⊗ε
n
k,r
(V
r+1
⊕V
r−1
)
from Section 3.4. We see that each of these summands is an AH-submodule of H⊗E
k,r
.
The larger (left-hand) submodule is stable; the smaller one is antistable. Both stable
and antistable AH-modules arise as Sp(1) Sp(1)-representations.
This is not necessarily the case for AH-modules which are neither stable nor anti-
stable. For example, there is no representation of Sp(1) on X
q
which couples with the
left H-action to give a representation of Sp(1) on X

q
.
AH-modules of the form V
1
⊗V
2m−1
Consider an even-dimensional irreducible Sp(1)-representation V
2m−1
. Because σ =
σ
1
⊗ σ
2m−1
is a real structure, there is a real representation V
L
1
⊗ V
M
2m−1

= R
4m
with
a left H-action defined by q : a ⊗ b → (qa) ⊗ b. Under the diagonal Sp(1)
LM
-action
q : a ⊗b →(qa) ⊗(qb), we have the splitting
V
L
1
⊗V
M
2m−1

= V
LM
2m
⊕V
LM
2m−2
, (5.3)
which is a splitting of real vector spaces (technically we could write (V
L
1
⊗ V
M
2m−1
)
σ

=
(V
LM
2m
)
σ
⊕(V
LM
2m−2
)
σ
).
Proposition 5.1.1 The pair (V
L
1
⊗ V
M
2m−1
, V
LM
2m
) forms a stable AH-module (U, U

)
with U

= H
m
and U


= R
2m+1
.
Proof. Consider the maximal stable submodule W of U. Since W is an H-submodule
it must be invariant under the left H-action. Also, W must depend solely on the
Sp(1) Sp(1)-representation structure: in particular W

must be preserved by the
diagonal action. So W

must be an Sp(1)
LM
-invariant subspace of U

= V
2m
and
by Schur’s Lemma [FH, p.7] W

= V
2m
or W

= ¦0¦. Since dimU

>
1
2
dimU we must
have W ,= ¦0¦. Hence (W, W

) = (U, U

) and thus U is stable.
64
Lemma 5.1.2 The AH-module U = V
1
⊗ V
2m−1
is the irreducible stable AH-module
corresponding to the SK-module A
2m
2
.
Proof. This follows from Proposition 5.1.1 and the remarks in Section 4.4. The virtual
dimension of U is 1, from which it follows that U must be irreducible. The isomorphism
with A
2m
2
follows because irreducible stable AH-modules are uniquely determined by
their dimensions.
It is instructive to describe the splitting of V
1
⊗V
2m−1
thoroughly in terms of basis
vectors. To make statements less cumbersome, we let n = 2m−1 throughout, bearing
in mind that n is odd.
Let V
1
= ¸x, y) and V
n
= ¸a
n
, a
n−1
b, . . . , ab
n−1
, b
n
) be Sp(1)-representations. The
actions of sl(2, C) on V
1
and V
n
are given by Equations (1.14) and (1.16) of Section
(1.2.1). We want to understand the actions on the tensor product
V
1
⊗V
n
=
_
x ⊗a
n
, x ⊗a
n−1
b, . . . , x ⊗ab
n−1
, x ⊗b
n
y ⊗a
n
, y ⊗a
n−1
b, . . . , y ⊗ab
n−1
, y ⊗b
n
_
.
In particular, we would like to find out how the left H-action interacts with the splitting
V
1
⊗V
n

= V
n+1
⊕V
n−1
.
Consider the structure of A
2m
2
as a K-module e : W →H ⊗V , where H

= V
1
and
V

= V
n
. Using these isomorphisms and Theorem 4.4.2, the image e(W) is spanned by
the vectors
¦x⊗a
n
, x⊗a
n−1
b+y⊗a
n
, . . . , x⊗a
n−k
b
k
+y⊗a
n−k+1
b
k−1
, . . . , x⊗b
n
+y⊗ab
n−1
, y⊗b
n
¦.
It is easier to observe the sp(1) action on the complementary subspace which we can
identify as the U

-part of an SH-module. Given a suitable choice of metric, the perpen-
dicular subspace to e(W) is spanned by vectors of the form x⊗a
n−k
b
k
−y⊗a
n−k+1
b
k−1
.
Calculating the action of the Casimir operator C = H
2
+ 2XY + 2Y X reveals that
C(x ⊗a
n−k
b
k
−y ⊗a
n−k+1
b
k−1
) = (n + 1)(n −1)(x ⊗a
n−k
b
k
−y ⊗a
n−k+1
b
k−1
).
This shows that x ⊗a
n−k
b
k
−y ⊗a
n−k+1
b
k−1
∈ V
n−1
, giving the result that
V
n−1
= Span¦x ⊗a
n−k
b
k
−y ⊗a
n−k+1
b
k−1
: 1 ≤ k ≤ n¦. (5.4)
The subspaces e(W) and e(W)

of the SK-module H ⊗ V are thus equivalent to the
subspaces V
n+1
and V
n−1
respectively in the splitting V
1
⊗V
n

= V
n+1
⊕V
n−1
.
The SK-module structure maps σ
W
and σ
V
are exactly the standard real structure
σ
n−1
and the quaternionic structure σ
n
introduced in Section 1.2.1. The H-action is as
usual determined by the action of sp(1) on x and y using the correspondence
1 ←→x i
1
←→ix i
2
←→y i
3
←→−iy.
This formalism enables us to write out the structure of V
1
⊗V
n
as an H-module in the
same fashion as in Example 4.4.5.
65
AH-modules of the form V
1
⊗V
2m
We can also obtain a stable AH-module from an odd-dimensional irreducible
Sp(1)-representation V
2m

= C
2m+1
. If we take the tensor product V
L
1
⊗ V
M
2m

= C
4m+2
we obtain the splitting V
L
1
⊗V
M
2m

= V
LM
2m+1
⊕V
LM
2m−1
and a left H-action in the same way
as above. However this does not restrict to an H-action on any suitable real vector space
U such that U ⊗
R
C = V
1
⊗V
2m
(this is obviously impossible since R
4m+2
,

= H
k
for any
k ). The reason for this is that the structure map σ
1
⊗σ
2m
has square −1 instead of 1,
and so V
L
1
⊗V
M
2m
is a quaternionic rather than a real representation of Sp(1)
L
Sp(1)
M
.
There are two ways round this difficulty. Firstly, we could simply take the underlying
real vector space R
8m+4

= V
1
⊗ V
2m
to be an H-module. Secondly, we can tensor with
H

= C
2
equipped with its standard structure map. The vector space H is unaffected
by the Sp(1)
L
Sp(1)
M
-action; thus we can think of V
L
1
⊗V
M
2m
⊗H as a direct sum of
two copies of V
L
1
⊗V
M
2m
, which we write 2V
L
1
⊗V
M
2m
. This space comes equipped with a
real structure σ = σ
1
⊗σ
2m
⊗σ
H
, and so we have a stable AH-module
((2V
L
1
⊗V
M
2m
)
σ
, (2V
LM
2m+1
)
σ
). (5.5)
(As usual, once the correct structure maps have been specified, we will not usually
mention the σ-superscript.) This approach is the equivalent of dealing with SK-modules
of the form A
2m+1
k=2,3
⊗H and the σ-sheaves O(2m + 1) ⊗H.
Both these approaches give exactly the same AH-module; both effectively leave the
Sp(1) Sp(1)-representation V
1
⊗V
2m
untouched, whilst doubling the dimension of the
real vector space we are considering so that it is divisible by four.
5.1.4 General Stable and Antistable AH-modules
Definition 5.1.3 Let U
2n
denote the AH-module (V
L
1
⊗V
M
2n+1
, V
LM
2n+2
).
Let U
2n−1
denote the AH-module (2V
L
1
⊗V
M
2n
, 2V
LM
2n+1
).
The AH-module U
2n
corresponds to the SK-module A
2n+2
2
and the σ-sheaf O(2n).
The AH-module U
2n−1
corresponds to the SK-module A
2n+1
2
⊗H and the σ-sheaf
O(2n −1) ⊗H.
By analogy with the quaternions themselves, we will refer to the action of the ‘left’
subgroup Sp(1)
L
on V
1
as the left H-action, and the action of the ‘right’ subgroup
Sp(1)
M
on V
n
as the Sp(1)
M
-action. (This could also be taken to signify ‘module’
action.) We will often omit the superscripts L and M from expressions like V
L
1
⊗V
M
n
if the context leaves no ambiguity as to which group acts on what. Thus we write
U
n
= aV
1
⊗V
n+1
,
where a = 1 if n is even and a = 2 if n is odd.
This formulation allows us to see the relationship between stable AH- and SH-modules
very explicitly. The AH-module U
n
= aV
1
⊗V
n+1
splits as a(V
n+2
⊕V
n
). We can choose
to regard this as a stable AH-module by thinking of aV
n+2
as the ‘primed part’, or as a
stable SH-module by regarding aV
n
as the ‘generating real subspace’.
The classification results of Sections 4.4 and 4.5 allow us to state the following theo-
rem:
66
Theorem 5.1.4 Every stable AH-module can be written as a direct sum of the irre-
ducibles U
n
with unique multiplicities.
Consider the direct sum U =

n
j=0
a
j
U
j
. We can write U more explicitly in terms
of Sp(1)
L
Sp(1)
M
-representations, as the sum
U = V
L
1

_
m

j=1
V
M
2j+1
⊕2
n

k=1
V
M
2k
_
. (5.6)
The left H-action on V
L
1
is common to all the irreducibles. The Sp(1)
M
-action can
be much more complicated. However, we know that since Sp(1) is a compact group, any
such representation can be written as a sum of irreducibles with unique multiplicities.
Having done this, it is then easy to separate these representations to form separate AH-
modules, provided that each odd-dimensional representation V
M
2k
appears with even
multiplicity. Thus the following is equivalent to Theorem 5.1.4:
Theorem 5.1.5 To every stable AH-module (U, U

) can be attached an Sp(1)
M
-action
which intertwines with the left H-action in such a way that the diagonal Sp(1)
LM
-action
preserves U

.
In the decomposition of Equation (5.6), each irreducible subrepresentation of the
Sp(1)
M
-action contributes 1 to the virtual dimension of U. Thus the virtual dimension
of

n
j=0
c
j
U
j
is equal to

j even
c
j
+ 2

j odd
c
j
.
Antistable AH-modules
Let U
n
= aV
1
⊗ V
n+1
be a stable AH-module. Then its dual AH-module U
×
n
is an
antistable AH-module. Just like stable SH-modules, antistable AH-modules are formed
by taking the smaller summand in the splitting aV
1
⊗V
n+1
= a(V
n+2
⊕V
n
). The following
Lemma then follows immediately from Definition 5.1.3.
Lemma 5.1.6 The antistable AH-module U
×
2n
takes the form (V
L
1
⊗V
M
2n+1
, V
LM
2n
).
The antistable AH-module U
×
2n−1
takes the form (2V
L
1
⊗V
M
2n
, 2V
LM
2n−1
).
The AH-module U
×
2n
corresponds to the SK-module A
2n+2
3
and the σ-sheaf O(−2n−
4). The AH-module U
×
2n−1
corresponds to the SK-module A
2n+1
3
⊗H and the σ-sheaf
O(−2n − 3) ⊗ H. There is a ‘unique factorisation theorem’ for antistable AH-modules
which is exactly dual to Theorem 5.1.4.
5.1.5 Line Bundles over CP
1
and Sp(1)-representations
There is naturally a link between Sp(1)-representations and the cohomology groups of
vector bundles over CP
1
. In Section 4.5.1 we demonstrated that H
0
(O(n))

= S
n
(H),
where H

= C
2
is the basic representation of GL(2,C). From the inclusion SL(2,C) ⊂
GL(2,C), H is also the basic representation V
1
of SL(2,C) and therefore Sp(1). The
67
induced action of Sp(1) on S
n
(H) is by definition the irreducible representation V
n
.
Thus the cohomology groups of line bundles over CP
1
are Sp(1)-representations; we
have
H
0
(O(n))

= V
n
and H
1
(O(−n))

= V
n−2
. (5.7)
The exact sequence (4.11) of Section 4.5.1 is thus the same as the exact sequence
0 −→V
n−1
−→V
1
⊗V
n

= V
n−1
⊕V
n+1
−→V
n+1
−→0. (5.8)
The fact that the cohomology groups of line bundles over CP
1
have the structure
of irreducible Sp(1)-representations is already known in the context of the theory of
homogeneous spaces. Let G be a compact Lie group and let T be a maximal toral
subgroup. Then the homogeneous space G/T has a homogeneous complex structure.
(This famous result is due to Borel.) The right action of T on G gives G the structure
of a principal T-bundle over G/T. Let t be the Lie algebra of T, so that t is a
Cartan subalgebra of g. For each dominant weight λ ∈ t

there is a one-dimensional
representation C
λ
of T. The holomorphic line bundle associated to the principal bundle
G and the representation λ is then
L
λ
= G
T
C
λ
= (GC
λ
)/¦(g, v) ∼ (gt, t
−1
v), t ∈ T¦.
Since G acts on L
λ
, the cohomology groups of L
λ
are naturally representations of G.
For more information see [FH, p. 382-393].
In the case of the group Sp(1), each maximal torus is isomorphic to U(1), and the
homogeneous space Sp(1)/U(1)

= CP
1
is the Hopf fibration S
1
→ S
3
→ S
2
. The line
bundle Sp(1)
U(1)
C
λ
is then L
−λ
, where L is the hyperplane section bundle of CP
1
.
5.2 Sp(1)-Representations and the Quaternionic Ten-
sor Product
This section describes the quaternionic algebra of stable and antistable AH-modules
using the ideas of the previous section.
5.2.1 The inclusion map ι
U
(U)
We begin by discussing the map ι
U
and its image. Let U
n
be an irreducible stable
AH-module. Then
U
n
= a(V
L
1
⊗V
M
n+1
), U

n
= aV
LM
n+2
and U

n

= (U

n
)

= aV
LM
n
.
There is an injective map ι
U
n
: U
n
→H⊗(U

n
)

. This map has a natural interpretation
in terms of the Sp(1)-representations involved. Writing the quaternions as the stable
AH-module V
L
1
⊗V
R
1
, we have
H⊗(U

n
)


= V
L
1
⊗V
R
1
⊗aV
N
n
.
68
This is exactly like the motivating example of H ⊗ E
k,r
in Section 3.4. Leaving the
left-action untouched and taking the diagonal Sp(1)
RN
-action gives the isomorphism
H⊗(U

n
)


= V
L
1
⊗a(V
RN
n+1
⊕V
RN
n−1
) (5.9)
as an Sp(1) Sp(1)-representation. The AH-submodule ι
U
n
(U
n
) is clearly the V
L
1

aV
RN
n+1
subrepresentation of H⊗(U

n
)

.
Antistable AH-modules behave in a similar fashion. Consider the AH-module U
×
n
,
so that
U
×
n
= a(V
L
1
⊗V
M
n+1
), (U
×
n
)

= aV
LM
n
and (U
×
n
)


= ((U
×
n
)

)

= aV
LM
n+2
.
There is a similar splitting
H⊗((U
×
n
)

)


= V
L
1
⊗V
R
1
⊗aV
N
n+2

= V
L
1
⊗a(V
RN
n+3
⊕V
RN
n+1
).
This time, the AH-submodule ι
U
×
n
(U
×
n
) is the smaller AH-submodule V
L
1
⊗aV
RN
n+1
. Thus
the splitting
H⊗aV
N
n

= aV
L
1
⊗(V
RN
n+1
⊕V
RN
n−1
) (5.10)
splits H ⊗ aV
n
into the direct sum of a stable AH-module isomorphic to U
n
and an
antistable AH-module isomorphic to U
×
n−2
.
The subspaces ι
U
n
(U

n
) and ι
U
×
n
((U
×
n
)

) have a similar interpretation. Treating the
imaginary quaternions I as a copy of V
LR
2
, we have
I ⊗(U

n
)


= V
LR
2
⊗aV
N
n

= a(V
LRN
n+2
⊕V
LRN
n
⊕V
LRN
n−2
).
and ι
U
n
(U

n
) is the aV
n+2
subrepresentation of I ⊗(U

n
)

. In exactly the same way, for
U
×
n
we have
I ⊗((U
×
n
)

)


= V
LR
2
⊗aV
N
n+2

= a(V
LRN
n+4
⊕V
LRN
n+2
⊕V
LRN
n
).
In this case, ι
U
×
n
((U
×
n
)

) is the smallest subrepresentation aV
LRN
n
. This also shows why
we would not expect U

to be closed under the left H-action — the group Sp(1)
L
does
not act upon it, since we do not have an intact copy of V
L
1
.
It is worth noting that so far we have been able consistently to interpret stable AH-
modules and their subspaces as representations of highest weight in tensor products of
Sp(1)-representations, and antistable AH-modules and their subspaces as representations
of lowest weight.
5.2.2 Tensor products of stable AH-modules
We shall now see how to use our description of stable AH-modules to form the quater-
nionic tensor product. The results in this section can be obtained through Quillen’s
sheaf-theoretic version of AH-modules by using Theorem 4.5.7. However, the author
hopes that including a little more description will help the reader to get more of a feel
for what is going on.
Let U
m
= aV
1
⊗V
m+1
, U
n
= bV
1
⊗V
n+1
be stable AH-modules. By Definition 4.1.4,
U
m

H
U
n
= (ι
U
m
(U
m
) ⊗(U

n
)

) ∩ ((U

m
)

⊗ι
U
n
(U
n
)) ⊂ H⊗(U

m
)

⊗(U

n
)

.
69
In terms of Sp(1)-representations,
H⊗(U

m
)

⊗(U

n
)


= V
L
1
⊗V
R
1
⊗aV
P
m
⊗bV
Q
n
. (5.11)
Using Equation (5.9), we write ι
U
m
(U
m
)

= aV
L
1
⊗ V
RP
m+1
⊂ V
L
1
⊗ a(V
RP
m+1
⊕ V
RP
m−1
)

=
H⊗(U

m
)

. Tensoring this expression with (U

n
)


= bV
Q
n
gives
ι
U
m
(U
m
) ⊗(U

n
)


= a(V
L
1
⊗V
RP
m+1
) ⊗bV
Q
n
. (5.12)
In the same way, we form the isomorphism
(U

m
)

⊗ι
U
n
(U
n
)

= aV
P
m
⊗b(V
L
1
⊗V
RQ
n+1
). (5.13)
A rearrangement of the factors leaves us considering the spaces abV
L
1
⊗ V
RP
m+1
⊗ V
Q
n
and abV
L
1
⊗ V
P
m
⊗ V
RQ
n+1
. We now have an Sp(1)
L
Sp(1)
RP
Sp(1)
Q
-representation
and an Sp(1)
L
Sp(1)
P
Sp(1)
RQ
-representation. From these we want to obtain a
single Sp(1) Sp(1)-representation which leaves the left H-action intact. The way to
proceed is to leave the V
L
1
-factor in each of these expressions alone and consider the
representations of the diagonal subgroup Sp(1)
RPQ
. We examine the factors
V
RP
m+1
⊗ V
Q
n
and V
P
m
⊗ V
RQ
n+1
. To obtain a stable AH-module, we want to reduce these
two Sp(1) Sp(1)-representations to a single Sp(1)-representation. In so doing, we hope
to find the intersection of these two spaces.
This is summed up in the following diagram:
d
d
d
d
d
d
d
ds

d
d
d
d
d
d
d
ds
U
m

H
U
n

= abV
L
1
⊗(

V
RPQ
?
)
ι
U
m
(U
m
) ⊗(U

n
)


= V
L
1
⊗aV
RP
m+1
⊗bV
Q
n
(U

m
)

⊗ι
U
n
(U
n
)

= aV
P
m
⊗V
L
1
⊗bV
RQ
n−1
H⊗(U

m
)

⊗(U

n
)


= V
L
1
⊗V
R
1
⊗aV
P
m
⊗bV
Q
n
c
U
m
⊗(U

n
)


=
c
(U

m
)

⊗U
n

=
The upward arrows here are inclusion maps. The argument goes in the opposite direction,
as we restrict our attention to particular subspaces. If we go down the left hand side, we
consider the diagonal action of the subgroup Sp(1)
RP
, and restrict to the higher weight
subspace ι
U
m
(U
m
) ⊗ (U

n
)

. We then consider the action of Sp(1)
RPQ
on this. If on
the other hand we go down the right hand side, we consider the diagonal action of the
70
subgroup Sp(1)
RQ
, restrict to the higher weight subspace (U

m
)

⊗ ι
U
n
(U
n
), and then
consider the action of Sp(1)
RPQ
on this. At each ‘half-way stage’ we are considering
representations of diagonal subgroups of different pairs of groups, and in both cases we
take the higher weight representation in a sum V
k+1
⊕V
k−1
and discard the V
k−1
part.
Since we do this for different diagonal subgroups we expect to be left with different
subspaces.
Using the Clebsch-Gordon formula, we obtain the two decompositions
V
RP
m+1
⊗V
Q
n

=
min{m+1,n}

j=0
V
RPQ
m+1+n−2j
and V
P
m
⊗V
RQ
n+1

=
min{m,n+1}

j=0
V
PRQ
m+n+1−2j
.
These decompositions contain fairly similar summands, with differences arising as the
index j approaches the region of min¦m, n¦. However, just because we have two
Sp(1)
RPQ
-representations of the same weight, we cannot say that they are automati-
cally the same subspace of H ⊗ (U

)

⊗ (V

)

. We want to know which parts end up
contributing to the final Sp(1)
RPQ
-representation whichever path we take. This will
identify the subspace U
m

H
U
n
⊆ H⊗(U

m
)

⊗(U

n
)

.
One thing that we can guarantee for any m, n > 0 is that the representation with
highest weight will be the same in both cases — both expressions have leading summand
V
m+n+1
. We conjecture that this is the summand which we find in U
m

H
U
n
. This would
fit well with the observation that stable AH-modules arise as representations of highest
weight in decompositions of tensor products of Sp(1)-representations.
We will show that this is in fact the case, using Joyce’s dimension formulae for stable
AH-modules. Here is the main result of this section:
Theorem 5.2.1 Let U
m
, U
n
be irreducible stable AH-modules. If m or n is even then
U
m

H
U
n

= U
m+n
.
If m and n are both odd then
U
m

H
U
n

= 4U
m+n
.
Proof. We have already noted that each irreducible representation of the Sp(1)
M
-action
on a stable AH-module U contributes 1 to the virtual dimension of U. Thus any
stable AH-module of virtual dimension k must be a sum of at least k/2 and at most k
irreducibles, depending on whether the irreducibles are odd or even.
We will deal with the three possible cases in turn.
Case 1( m and n both even): Let m = 2p, n = 2q. Then
dimU
m
= 4(p + 1) dimU

m
= 2p + 3 dimU
n
= 4(q + 1) and dimU

n
= 2q + 3.
Using Theorem 4.1.14 we find that dimU
m

H
U
n
= 4(p + q + 1) and that the virtual
dimension of U
m

H
U
n
is equal to 1. But any stable AH-module whose virtual dimension
is equal to 1 must be irreducible. The irreducible stable AH-module whose dimension
71
is 4(p + q + 1) and whose virtual dimension is 1 is V
1
⊗ V
2(p+q)+1
= U
m+n
. Hence
U
m

H
U
n
= U
m+n
.
Case 2( m even and n odd): Let m = 2p, n = 2q −1. Then
dimU
m
= 4(p + 1) dimU

m
= 2p + 3 dimU
n
= 4(2q + 1) and dimU

n
= 4(q + 1).
Using Theorem 4.1.14 we find that dimU
m

H
U
n
= 4(2p + 2q + 1) and that the virtual
dimension of U
m

H
U
n
is equal to 2. Thus U
m

H
U
n
must be either an even irreducible
or a sum of two odd irreducibles.
Consider the space ι
U
m
(U
m
) ⊗(U

n
)


= a(V
L
1
⊗V
RP
m+1
) ⊗bV
Q
n
of Equation (5.12). For
m = 2p and n = 2q −1 this becomes
2V
L
1
⊗V
RP
2p+1
⊗V
Q
2q−1

= V
L
1
⊗2(V
RPQ
2p+2q
⊕V
RPQ
2p+2q−2
⊕. . .) (5.14)
The virtual dimension of the tensor product U
m

H
U
n
must be equal to 2, so we cannot
have more than 2 of the irreducibles of the Sp(1)
RPQ
-action. We also need a total
dimension of 4(2p + 2q + 1). Examining Equation (5.14) we see that the only way this
can occur is if U
m

H
U
n

= V
1
⊗ 2V
2p+2q
, as all the other irreducibles of the Sp(1)
RPQ
-
action have smaller dimension. Hence U
m

H
U
n

= 2V
1
⊗V
2p+2q
= U
m+n
.
Case 3 ( m and n both odd): The argument is very similar to that of Case 2.
Let m = 2p −1, n = 2q −1. Then
dimU
m
= 4(2p+1) dimU

m
= 4(p+1) dimU
n
= 4(2q +1) and dimU

n
= 4(q +1).
Using Theorem 4.1.14 we find that dimU
m

H
U
n
= 16(p + q) and that the virtual
dimension of U
m

H
U
n
is equal to 4.
Consider the space ι
U
m
(U
m
) ⊗(U

n
)


= a(V
L
1
⊗V
RP
m+1
) ⊗bV
Q
n
of Equation (5.12). For
m = 2p −1 and n = 2q −1 this becomes
4V
L
1
⊗V
RP
2p
⊗V
Q
2q−1

= V
L
1
⊗4(V
RPQ
2p+2q−1
⊕V
RPQ
2p+2q−3
+ . . .). (5.15)
The only way U
m

H
U
n
can have a virtual dimension of four and a total dimension of
16(p + q) is if U
m

H
U
n

= 4V
1
⊗V
2p+2q−1
= 4U
m+n
.
Quaternionic tensor products of more general stable AH-modules can be computed
from this result by splitting into irreducibles and using the fact that the quaternionic
tensor product is distributive for direct sums.
This result is parallel to Theorem 4.5.7 applied to non-negative vector bundles. For
the canonical sheaves O(n) over CP
1
, H
0
(O(n))

= V
n
. The isomorphism O(n) ⊗
O
O(m)

= O(n + m) induces a map of cohomology groups H
0
(O(m)) ⊗ H
0
(O(n)) →
H
0
(O(m + n)). In terms of Sp(1)-representations, this is a map
V
P
m
⊗V
Q
n

= V
PQ
m+n
⊕V
PQ
m+n−2
⊕. . . →V
m+n
.
The map in question is projection onto the irreducible of highest weight V
n+m
. This is
really what this whole section has been about — the idea that the behaviour of stable
AH-modules can be thoroughly and flexibly described by taking subrepresentations of
highest weight in tensor products of Sp(1)-representations.
72
5.2.3 Tensor Products of Antistable AH-modules
It is not difficult to extend Joyce’s results for tensor products of stable AH-modules to
irreducible antistable AH-modules — we can follow the same argument as in the proof
of Theorem 4.1.14, since the generic properties of sums and intersections guaranteed by
stability also hold if one or both of the AH-modules is irreducible and antistable.
Let U and V be antistable AH-modules with dimU = 4j, dimU

= 2j−r, dimV =
4k and dimV

= 2k − s. Let A = ι
U
(U) ⊗ (V

)

and let B = (U

)

⊗ ι
V
(V ). Then
dimA = 4j(2k + s) and dimB = 4k(2j + r), so dim(H ⊗ (U

)

⊗ (V

)

) = 4(2j +
r)(2k + s) > dimA + dimB. Thus in generic situations we would expect dimA ∩ B =
dimU⊗
H
V = 0.
Suppose instead that V is stable, so now dimV

= 2k + s. A similar calculation
shows that dimA + dimB ≥ dimH ⊗ (U

)

⊗ (V

)

if and only if s(j + r) ≥ kr. In
this case we might expect dimU⊗
H
V = dimA + dimB −dim(H⊗(U

)

⊗(V

)

). For
example, let U = U
×
m
= (aV
1
⊗ V
m+1
)
×
and V = U
n
= bV
1
⊗ V
n+1
. Then we would
expect that dim(U
×
m

H
U
n
) = 2ab(m−n + 2).
As in Section 5.2, we can describe what is going on in terms of diagonal actions
on tensor products of Sp(1)-representations. We will illustrate the case U
m

H
U
×
n
. Let
U
m
= aV
1
⊗V
m+1
and U
×
n
= (bV
1
⊗V
n+1
)
×
, so (U

m
)


= V
m
and ((U
×
n
)

)


= V
n+2
. This
gives rise to the standard descriptions
H⊗
R
(U

m
)


= aV
L
1
⊗V
R
1
⊗V
P
m

= aV
L
1
⊗(V
RP
m+1
⊕V
RP
m−1
)
and
H⊗
R
((U
×
n
)

)


= bV
L
1
⊗V
R
1
⊗V
Q
n+2

= bV
L
1
⊗(V
RQ
n+3
⊕V
RQ
n+1
).
The only difference between this and equation (5.9) is that we have an antistable AH-
module involved, and thus to obtain ι
U
×
m
(U
×
m
), we take the smaller summand of V
n+3

V
n+1
. We use the Clebsch-Gordon formula to describe
ι
U
m
(U
m
) ⊗
R
((U
×
n
)

)


= abV
L
1
⊗V
RP
m+1
⊗V
Q
n+2

= abV
L
1

_
_
min{m+1,n+2}

j=0
V
RPQ
m+n+3−2j
_
_
and
(U

m
)


R
ι
U
×
n
(U
×
n
)

= abV
L
1
⊗V
P
m
⊗V
RQ
n+1

= abV
L
1

_
_
min{m,n+1}

j=0
V
RPQ
m+n+1−2j
_
_
.
As with stable AH-modules, our task is to find which of these summands is in the
intersection U
m

H
U
×
n
= ι
U
m
(U
m
) ⊗
R
((U
×
n
)

)

∩ (U

m
)


R
ι
U
×
n
(U
×
n
). This time since
min¦m + 1, n + 2¦ = min¦m, n + 1¦, we can guarantee that the summand of smallest
weight will appear in both expressions. From our dimensional arguments, we only expect
a non-zero intersection if m < n + 2, in which case min¦m, n + 1¦ = m, and we would
predict that U
m

H
U
×
n
contains the summand V
n−m+1
. Because its virtual dimension is
not positive, U
m

H
U
×
n
cannot be stable. This suggests that
U
m

H
U
×
n

= (abV
1
⊗V
n−m+1
)
×
=
_
_
_
¦0¦ if m ≥ n + 2
U
×
n−m
n or m even and m < n + 2
4U
×
n−m
n and m both odd and m < n + 2.
(5.16)
73
Rather than try to emulate Joyce’s (difficult) proof of Theorem 4.1.14, we will confirm
these conjectures by appealing to Quillen’s powerful results.
Proposition 5.2.2 Let U
m
be an irreducible stable AH-module and let U
×
n
be an irre-
ducible antistable AH-module.
If m ≥ n + 2 then U
m

H
U
×
n
= ¦0¦.
If m < n + 2 and m or n is even then U
m

H
U
×
n

= U
×
n−m
.
If m < n + 2 and m and n are both odd then U
m

H
U
×
n

= 4U
×
n−m
.
Let U
×
m
and U
×
n
be antistable irreducible AH-modules. Then U
×
m

H
U
×
n
= ¦0¦.
Proof. This follows from Theorem 4.5.7 (due to Quillen), using the correspondences
U
m
= η
+
(aO(m)) and U
×
n
= η

(aO(−n −4)), where as usual a = 1 or 2 depending on
whether m, n are even or odd.
We can use this result about tensor products of antistable AH-modules to tell us
about AH-morphisms between stable AH-modules, using the isomorphism
Hom
AH
(U
m
, U
n
)

= (U
×
m

H
U
n
)

of Theorem 4.2.9.
Proposition 5.2.3 Let U
m
and U
n
be stable AH-modules. Then
Hom
AH
(U
m
, U
n
)

=
_
(aU
×
m−n
)

= aV
m−n
n ≤ m
¦0¦ n > m
where a = 4 if m and n are both odd and a = 1 otherwise.
Proof. This follows immediately by combining Theorems 4.2.9 and 5.2.2.
In particular, since U
0
= H, we see that there are always AH-morphisms from U
n
into
H (and indeed, this is a defining property for AH-modules), but never AH-morphisms
from H into U
n
unless n = 0.
Similarly, we can now see that there are always AH-morphisms from antistable AH-
modules into stable AH-modules, but never AH-morphisms from stable AH-modules into
antistable AH-modules.
5.3 Semistable AH-modules and Sp(1)-representations
In this section we shall consider how the AH-module X
q
fits into the picture of stable
AH-modules and Sp(1)-representations. Recall from Section 4.1.3 that for q ∈ S
2
,
X
q
= H and X

q
= ¦p ∈ H : pq = −qp¦ so that X

q

= (X

q
)


= C
q
.
If we consider the dual AH-module X
×
q

= (H, C
q
) we see that the left-multiplication
L
q
: H → H defined by L
q
(p) = q p gives an AH-isomorphism X
q

= X
×
q
. This
suggests that Theorem 4.2.9 might be particularly interesting in the case of X
q
. For any
AH-module U, there is a canonical isomorphism
Hom
AH
(X
q
, U)

= (X
×
q

H
U)

74
and so
Hom
AH
(X
×
q
, U)

= Hom
AH
(X
q
, U)

= (X
q

H
U)


= (X
×
q

H
U)

.
Let φ : X
×
q
→ U be an AH-morphism. Then we need φ(1) = u ∈ U

and φ(q) ∈ U

.
Since φ is H-linear, φ(q) = qu, and so u ∈ U

∩ qU

. It follows that
Hom
AH
(X
q
, U)

= (X
q

H
U)


= U

∩ qU

. (5.17)
As noted by Joyce (see the summary in Section 4.1.3), the second of these isomorphisms
is given by the map (id
U

H
χ
q
) : (U⊗
H
X
q
)

→(U⊗
H
H)


= U

.
Though X

q
is not itself an Sp(1)-representation, the subspaces C
q
and X

q
= C

q
are acted on by the Cartan subgroup U(1)
q
⊂ Sp(1). As we shall see, taking the tensor
product of a stable AH-module with the AH-module X
q
serves to restrict attention
from information about Sp(1)-representations to information concerning representations
of the group U(1)
q
and its Lie algebra u(1)
q
.
Let Q = aI
1
+ bI
2
+ cI
3
∈ sp(1) with a
2
+ b
2
+ c
2
= 1 and let q = ai
1
+ bi
2
+ ci
3

S
2
. Any irreducible representation of sp(1) is also a representation of the subalgebra
u(1)
q
= ¸Q). The analysis of V
n
as a U(1)
q
-representation is already familiar: it is the
decomposition of V
n
into weight spaces of the operator Q : V
n
→ V
n
. If we choose an
identification sp(1) ⊗
R
C = sl(2, C) such that Q = iH, this is exactly the same as the
decomposition of V
n
into eigenspaces of H with weights ¦−n, −n + 2, . . . , n − 2, n¦.
This decomposition gives important information about the action of q on the AH-module
U
n
= aV
1
⊗V
n+1
.
This is exactly what we have done in the explicit calculations of Section 5.1.3 for the
case q = i
1
. We define a basis ¸x, y) for the space V
1
in such a way that Q(x) = ix
and Q(y) = −iy. This also gives the left action of q ∈ Sp(1) on V
1
, since the actions of
q and Q coincide for this representation (see Section 1.2.1). This gives the left action
of q ∈ H on U
n
= aV
1
⊗V
n+1
.
The goal of this discussion is to describe the AH-module U
n

H
X
q
. We do this with
the aid of the following lemma. For ease of notation we work with the AH-module U
n−1
.
Lemma 5.3.1 The subspace U

n−1
∩ qU

n−1
is given by the sum of the weight spaces of
Q with highest and lowest possible weights.
Proof. Recall from Section 5.1.3 that
V
1
⊗V
n
=
_
x ⊗a
n
, x ⊗a
n−1
b, . . . , x ⊗ab
n−1
, x ⊗b
n
y ⊗a
n
, y ⊗a
n−1
b, . . . , y ⊗ab
n−1
, y ⊗b
n
_
,
and that the space (U
n−1
)


= V
n+1
is spanned by the vectors
¦x⊗a
n
, x⊗a
n−1
b+y⊗a
n
, . . . , x⊗a
n−k
b
k
+y⊗a
n−k+1
b
k−1
, . . . , x⊗b
n
+y⊗ab
n−1
, y⊗b
n
¦.
Define the basis vectors w
k
≡ x⊗a
n−k
b
k
+y ⊗a
n−k+1
b
k−1
, including w
0
= x⊗a
n
and
w
n+1
= y ⊗b
n
.
The action of H ∈ sl(2, C) is given by H(x ⊗ a
n−k
b
k
) = (n − 2k + 1)x ⊗ a
n−k
b
k
and H(y ⊗ a
n−k
b
k
) = (n − 2k − 1)y ⊗ a
n−k
b
k
. Thus each basis vector w
k
is a weight
vector with weight (n −2k + 1). The two extreme vectors x ⊗a
n
and y ⊗b
n
are also
75
weight vectors with weights n +1 and −n −1 respectively. Note that all these weights
are different.
The left H-action of q on V
1
⊗V
n
is given by
q(x ⊗a
n−k
b
k
) = ix ⊗a
n−k
b
k
q(y ⊗a
n−k
b
k
) = −iy ⊗a
n−k
b
k
.
Left multiplication by q therefore preserves the weight space decomposition with respect
to Q of a vector w ∈ V
1
⊗V
n
.
On the basis vectors w
k
= x ⊗a
n−k
b
k
+y ⊗a
n−k+1
b
k−1
we have
q(x ⊗a
n−k
b
k
+y ⊗a
n−k+1
b
k−1
) = ix ⊗a
n−k
b
k
−iy ⊗a
n−k+1
b
k−1
.
Since each vector w
k
has a different weight from all the others we have q(w
k
) ∈ V
n+1
if
and only if q(w
k
) ∈ ¸w
k
), and
q(

λ
k
w
k
) ∈ V
n+1
⇐⇒λ
k
= 0 or q(w
k
) ∈ V
n+1
for all k.
It is evident that
q(x ⊗a
n
) = ix ⊗a
n
∈ V
n+1
and q(y ⊗a
n
) = −iy ⊗a
n
∈ V
n+1
,
but for all the other basis vectors w
k
,
q(w
k
) ,∈ ¸w
k
) and so q(w
k
) ,∈ V
n+1
.
Hence
V
n+1
∩ qV
n+1
= ¸x ⊗a
n
, y ⊗b
n
),
the weight spaces with highest and lowest weight. Taking the σ-invariant subspace
¸x ⊗ a
n
− y ⊗ b
n
) if U
n−1
is an even AH-module yields the desired result for the AH-
module U
n−1
.
Since the AH-submodule (id ⊗
H
χ
q
)(U
n−1

H
X
q
) ⊂ U
n−1
is the subspace generated
over H by U

∩ qU

, this demonstrates the main result of this section which describes
U
n

H
X
q
as follows:
Theorem 5.3.2 Let U
n
= aV
1
⊗V
n+1
be an irreducible stable AH-module, whose eigenspace
decomposition with respect to Q ∈ sp(1) takes the form
V
1
⊗V
n+1
=
_
x ⊗a
n+1
, x ⊗a
n
b, . . . , x ⊗ab
n
, x ⊗b
n+1
y ⊗a
n+1
, y ⊗a
n
b, . . . , y ⊗ab
n
, y ⊗b
n+1
_
.
If U
n
is an even AH-module then U
n

H
X
q

= X
q
and
(id ⊗
H
χ
q
)(U
n

H
X
q
) = H ¸x ⊗a
n+1
−y ⊗b
n+1
)
R
,
where the H-action is induced by the H-action on V
1
.
If U
n
is an odd AH-module then U
n

H
X
q

= 2X
q
and
(id ⊗
H
χ
q
)(U
n

H
X
q
) = H ¸x ⊗a
n+1
, y ⊗b
n+1
)
R
= V
1
⊗¸a
n+1
, b
n+1
).
Thus the quaternionic tensor product U
n

H
X
q
picks out the representations of ex-
treme weight in the decomposition of U
n
into weight spaces of Q.
76
5.4 Examples and Summary of AH-modules
By now, the reader should be familiar with the ideas of quaternionic algebra. In this
section we shall briefly sum up this information, giving explicit constructions of the AH-
modules which will occur most frequently in the following chapter, in the forms in which
they occur most naturally.
Example 5.4.1 Recall the AH-module Y = ¦(q
1
, q
2
, q
3
) : q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦ of
Example 4.1.2. Since Y is stable and has virtual dimension 1, it follows that Y is
irreducible and is isomorphic to U
2
= V
1
⊗V
3
. A calculation shows that (Y

)


= V
2
and
that the equation q
1
i
1
+q
2
i
2
+q
3
i
3
= 0 is precisely the condition for (q
1
, q
2
, q
3
) to lie in
the subspace V
L
1
⊗V
RM
3
of H⊗V
M
2

= V
L
1
⊗V
R
1
⊗V
M
2

= V
L
1
⊗(V
RM
3
⊕V
RM
1
).
Consider the AH-modules

k
H
Y , S
k
H
Y and Λ
k
H
Y . Using the dimension formulae of
Theorem 4.1.14 and Proposition 4.1.16, we discover that Λ
n
H
Y = ¦0¦ and that

n
H
Y =
S
n
H
Y with dim(S
n
H
Y ) = 4(n + 1) and dim(S
n
H
Y )

= 2n + 3. From this we deduce that
S
n
H
Y

= U
2n
, and that all the even irreducible stable AH-modules can be realised as tensor
powers of the AH-module Y .
Example 5.4.2 [J1, Example 10.1] Let Z ⊂ H⊗R
4
be the set
Z = ¦(q
0
, q
1
, q
2
, q
3
) : q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦.
Then Z

= H
3
is a left H-module. Define a real subspace Z

= ¦(q
0
, q
1
, q
2
, q
3
) : q
j

I and q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦. Then dimZ = 12, dimZ

= 8 and Z is a stable
AH-module.
In fact, Z is isomorphic to the first odd irreducible AH-module U
1

= 2V
1
⊗V
2
. We
have (Z

)


= R
4

= 2V
1
. The equation q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0 is the condition for
(q
0
, q
1
, q
2
, q
3
) to lie in the subspace 2V
L
1
⊗ V
RM
2
of H ⊗ 2V
M
1

= V
L
1
⊗ V
R
1
⊗ 2V
M
1

=
2V
L
1
⊗(V
RM
2
⊕V
RM
0
).
Example 5.4.3 We can obtain the rest of the odd AH-modules as tensor products of
those above. From Theorem 5.2.1 we know that
U
2n+1
= U
2n

H
U
1
.
This combined with the previous examples shows that
U
2n+1

= S
n
H
Y ⊗
H
Z.
Thus we have obtained all the irreducible stable AH-modules in terms of previously
known AH-modules. We can also use these constructions to write down formulae giving
all those irreducible antistable AH-modules which are dual to stable AH-modules.
Example 5.4.4 Consider the AH-module (U, U

) where
U = H
2
and U

= ¸(1, 0), (i
1
, i
2
), (0, 1), (0, i
1
)).
Then ¦(0, q) : q ∈ H¦ is an AH-submodule of U isomorphic to X
i
1
. However, there is no
complementary AH-submodule V such that U

= X
i
1
⊕V , and indeed U is irreducible.
77
It is also clear that U is not semistable — nor is it the dual AH-module of any semistable
AH-module. We call such an AH-module irregular.
Irregular AH-modules correspond to type (i) SK-modules of the form A
n,α
1
⊕A
n,−¯ α
−1
1
and torsion sheaves of the form O/m
n
x
, where n ≥ 2. In the present case which singles
out the subfield C
i
1
, we have
U

= η
+
(O/m
2
{0,∞}
).
For a general formula linking pairs of antipodal points ¦z, σ(z)¦ in CP
1
and complex
subfields of H see [Q, ¸14].
These are the only AH-modules which we do not describe in detail, for two reasons.
Firstly, they are badly behaved compared to semistable and antistable AH-modules. Sec-
ondly, as far as the author is aware, they do not arise naturally in geometrical situations
in the way that the other AH-modules do.
Example 5.4.5 There is one remaining irreducible to consider — the AH-module
(H, ¦0¦). This trivially satisfies Definition 4.1.1, and so is an AH-module. Any AH-
module whose primed part is zero is a direct sum of copies of (H, ¦0¦). It is easy to see
that for any irreducible AH-module U, U⊗
H
(H, ¦0¦) = ¦0¦ unless U = H, in which
case we have H⊗
H
(H, ¦0¦) = (H, ¦0¦).
Though badly behaved, the AH-module (H, ¦0¦) does arise naturally in quaternionic
algebra — for example, Z⊗
H
H
×
= (H, ¦0¦). This suggests the notation
(H, ¦0¦) = U
×
−1
,
in which case this result agrees with Theorem 5.2.2. Such notation is consistent with the
sheaf description of antistable AH-modules, since we have
U
×
−1
= η

(2O(−3))
as expected. Thus we interpret (H, ¦0¦) as the ‘antistable part’ of 2V
1
⊗V
0
. The dual
space (H, H) is of course not an AH-module, so there is no stable AH-module U
−1
which
is dual to U
×
−1
. In spite of this we still regard U
×
−1
as ‘antistable’, because treating it as
an exception every time would be cumbersome.
We end this chapter with a diagram (overleaf) summarising much of our theory.
78
Figure 5.1: Irreducible AH-modules and the ratio of their dimensions.
0
1
4
1
2
3
4
1 dimU

/ dimU
AH-morphisms
-
Quaternionic tensor products

H
Z

= U
1
c
Y

= U
2
c
Y
×

= U
×
2
c
Z
×

= U
1
c
H
×
X
q
and irregular
AH-modules
c
U
×
−1
= (H, ¦0¦)
. . . . . .
U
×
n
(3 ≤ n < ∞)

. . . . . .
U
n
(3 ≤ n < ∞)
d
d
ds
This describes the AH-modules we have met so far — all the finite-dimensional irre-
ducible AH-modules. The quaternionic tensor product of two AH-modules is always to
the left of both of them in Figure 5.1. On the other hand, there are AH-morphisms from
an AH-module into itself and any AH-modules to its right — and never from an AH-
module to any AH-module to its left. These statements are closely linked by Theorem
4.2.9. If U and V are irreducible stable or antistable AH-modules, this demonstrates
that that there will always be AH-morphisms from U⊗
H
V into U and V , but never
any AH-morphisms from U or V into U⊗
H
V , unless U or V is equal to H.
79
Chapter 6
Hypercomplex Manifolds
This chapter uses the quaternionic algebra developed in Chapters 4 and 5 to describe
hypercomplex manifolds. It is in three parts (the first of which is a summary of Joyce’s
work, the other two being original). The first part (Section 6.1) summarises Joyce’s
theory of q-holomorphic functions. A q-holomorphic function on a hypercomplex man-
ifold M is a smooth function f : M → H which satisfies a quaternionic version of
the Cauchy-Riemann equations. We let T
M
denote the AH-module of q-holomorphic
functions on M. The q-holomorphic functions on the hypercomplex manifold H are
precisely the regular functions of Fueter and Sudbery. Because of the noncommuta-
tivity of the quaternions, the product of two q-holomorphic functions is not in general
q-holomorphic. Nonetheless, q-holomorphic functions possess a rich algebraic structure.
Joyce attempts to capture this using the concept of an H-algebra, a quaternionic version
of a commutative algebra over the real or complex numbers.
In the second part (Section 6.2), we use the Sp(1)-representation T

M

= 2nV
1
defined by the hypercomplex structure to obtain a natural splitting of the quaternionic
cotangent space of a hypercomplex manifold M, which we write H ⊗ T

M

= A ⊕ B.
This is precisely a version of the splitting H ⊗ aV
k

= aV
1
⊗ (V
k+1
⊕ V
k−1
) used in the
previous chapter to construct and describe all stable and antistable AH-modules, and
the Sp(1)-version of quaternionic algebra gives a complete description of the geometric
situation. A function f is q-holomorphic if and only if its differential df takes values in
the subspace A ⊂ H⊗T

M. This is very similar to the situation in complex geometry
where the differential of a holomorphic function takes values in the holomorphic cotangent
space Λ
1,0
. It follows that the subbundle A should be thought of as the q-holomorphic
cotangent space of M, its complement B being the q-antiholomorphic cotangent space.
The third part (Sections 6.3 and 6.4) is about AH-bundles, which are smooth vector
bundles whose fibres are AH-modules. They can be defined over any smooth manifold
M, but are most interesting when M is hypercomplex. An AH-bundle E is said to
be q-holomorphic if it is simultaneously holomorphic with respect to the whole 2-sphere
of complex structures on M. This condition is met if and only if E carries an anti-
self-dual connection which is compatible with the structure of E as an AH-bundle. We
generalise the theory of q-holomorphic functions to that of q-holomorphic sections, so
that a q-holomorphic function is precisely a q-holomorphic section of the trivial bundle
MH. We investigate the algebraic structure of q-holomorphic sections in some detail,
showing that the q-holomorphic sections of a q-holomorphic vector bundle E form an
80
H-algebra module over T
M
.
6.1 Q-holomorphic Functions and H-algebras
6.1.1 Quaternion-valued functions
This section is mainly a summary of [J1, ¸¸3,5], to which the reader is referred for
more detail and proofs of the important results. Let M be a smooth manifold and
let C

(M, H) be the vector space of smooth quaternion-valued functions on M. An
H-action on C

(M, H) is defined by setting (q f)(m) = q(f(m)) for all m ∈ M,
f ∈ C

(M, H). Thus C

(M, H) is a left H-module. What is less immediately obvious
is that C

(M, H) is an AH-module.
Lemma 6.1.1 Define a linear subspace
C

(M, H)

= ¦f ∈ C

(M, H) : f(m) ∈ I for all m ∈ M¦ = C

(M, I).
With these definitions, (C

(M, H), C

(M, H)

) is an AH-module.
Proof. We use the fact that M can be embedded in C

(M, H)

in the following way.
For each m ∈ M, define the ‘evaluation map’
θ
m
: C

(M, H) →H by θ
m
(f) = f(m).
Then θ
m
(q f) = q θ
m
(f), so θ
m
∈ C

(M, H)
×
. Also, if f ∈ C

(M, I) then θ
m
(f) ∈ I
for all m ∈ M, so θ
m
∈ C

(M, H)

.
Suppose that f ∈ C

(M, H), and α(f) = 0 for all α ∈ C

(M, H)

. Since θ
m

C

(M, H)

, f(m) = 0 for all m ∈ M ; so f ≡ 0. Thus C

(M, H) is an AH-module,
by Definition 4.1.1.
This technique of linking a point m ∈ M to an element of C

(M, H)

is extremely
useful. Joyce has used this process to reconstruct hypercomplex manifolds from their
H-algebras. Note that we have assumed no geometric structure on M other than that
of a smooth manifold.
Complex and quaternionic functions
For each q ∈ S
2
let ι
q
: C → H be the inclusion obtained by identifying i ∈ C with
q ∈ S
2
, so that ι
q
(a + ib) = a + qb.
1
For any real vector space E, the inclusion ι
q
extends to a map ι
q
: C⊗E →H⊗E given by ι
q
(e
0
+ie
1
) = e
0
+qe
1
(for e
0
, e
1
∈ E ).
Note that the images of ι
q
and ι
−q
are the same, but the two maps are not identical:
since ι
q
(a + ib) = a + qb and ι
−q
(a + ib) = a −qb, we see that ι
q
(z) = ι
−q
(¯ z).
Let f = a + ib be a complex-valued function on M. Then for every q ∈ S
2
,
the function ι
q
(f) = a + qb is a quaternion-valued function on M. In this way we
obtain quaternion-valued functions from complex ones, and we shall soon see that this
construction can be used to obtain q-holomorphic functions from holomorphic ones.
1
The map ι
q
will be distinguished from the inclusion map ι
U
of AH-modules by the use of lower-case
rather than upper-case subscripts.
81
6.1.2 Q-holomorphic functions
In this section we will define q-holomorphic functions, the hypercomplex version of holo-
morphic functions, and familiarise ourselves with some of their basic properties. Consider
the complex manifold (M, I). A complex-valued function f = f
0
+ idf
1
∈ C

(M, C) is
holomorphic (with respect to I ) if and only if f satisfies the Cauchy-Riemann equations
df
0
+ Idf
1
= 0. (6.1)
Let (M, I
1
, I
2
, I
3
) be a hypercomplex manifold. Here is the definition of a q-holomorphic
function on M.
Definition 6.1.2 Let f ∈ C

(M, H) be a smooth H-valued function on M. Then
f = f
0
+f
1
i
1
+f
2
i
2
+f
3
i
3
with f
j
∈ C

(M). The function f is said to be q-holomorphic
if and only if it satisfies the Cauchy-Riemann-Fueter equations
Df ≡ df
0
+ I
1
df
1
+ I
2
df
2
+ I
3
df
3
= 0.
The set of all q-holomorphic functions on M is called T
M
.
The term q-holomorphic is short for quaternion-holomorphic, and it is intended to
indicate that a q-holomorphic function on a hypercomplex manifold is the appropriate
quaternionic analogue of a holomorphic function on a complex manifold. If a function f
is q-holomorphic then the function q f is also q-holomorphic for all q ∈ H. Thus the
set of q-holomorphic functions T
M
forms a left H-submodule of C

(M, H). We adopt
the obvious definition for T

M
, namely
T

M
= ¦f ∈ T
M
: f(m) ∈ I for all m ∈ M¦.
So T

M
= T
M
∩ C

(M, H)

, and T
M
is an AH-submodule of C

(M, H). Thus the
q-holomorphic functions T
M
on a hypercomplex manifold M form an AH-module.
The product of two q-holomorphic functions is not in general q-holomorphic — we can
observe this simply by noting that all constant functions are trivially q-holomorphic, but
T
M
is not even closed under right-multiplication by quaternions. Thus q-holomorphic
functions do not form an algebra in the same sense that the holomorphic functions on a
complex manifold do.
We show that there are many interesting q-holomorphic functions on a hypercomplex
manifold M, by observing that every complex-valued function on M which is holo-
morphic with respect to any complex structure Q ∈ S
2
gives rise to a q-holomorphic
function.
Lemma 6.1.3 Let q = a
1
i
1
+ a
2
i
2
+ a
3
i
3
∈ S
2
and let Q = a
1
I
1
+ a
2
I
2
+ a
3
I
3
be the
corresponding complex structure on M. Let f = x + iy ∈ C

(M, C) be holomorphic
with respect to Q. Then ι
q
(f) = x + qy is a q-holomorphic function.
Proof. The proof is a simple substitution. If f = x + iy is holomorphic with respect to
Q then it satisfies the Cauchy-Riemann equations dx + Q(dy) = 0, in which case
0 = dx + (a
1
I
1
+ a
2
I
2
+ a
3
I
3
)(dy) = D(x + a
1
yi
1
+ a
2
yi
2
+ a
3
yi
3
) = D(x + qy).
82
So the H-valued function ι
q
(f) = x + qy is q-holomorphic.
If f = x + iy ∈ C

(M, C) is holomorphic with respect to Q, then its conjugate
¯
f = x − iy is holomorphic with respect to −Q. Does this mean that both x + qy and
x − qy are q-holomorphic functions? Closer inspection shows that this is not the case
— we see that
ι
q
(x + iy) = ι
−q
(x −iy) = x + qy.
So the holomorphic functions with respect to Q and −Q are mapped to the same
functions when we consider their images under ι
±q
in C

(M, H). We shall see that this
is a consequence of (indeed is equivalent to) the fact that the AH-modules X
q
and X
−q
are the same.
Now, if the functions f, g ∈ C

(M, C
q
) are both holomorphic with respect to Q,
then so is fg ∈ C

(M, C
q
). By Lemma 6.1.3, ι
q
(f), ι
q
(g), and ι
q
(fg) = ι
q
(f)ι
q
(g)
are all q-holomorphic. In this special case where the q-holomorphic functions f and
g both take values in a commuting subfield of H, their product is also q-holomorphic.
This is reminiscent of the situation described in Lemma 4.1.7, where two elements u ∈
U and v ∈ V such that ι
U
(u) ∈ C
q
⊗ (U

)

and ι
V
(v) ∈ C
q
⊗ (V

)

define an
element u⊗
H
v ∈ U⊗
H
V . In this case we have two well-defined ‘products’ of f and
g — their quaternionic tensor product f⊗
H
g ∈ T
M

H
T
M
and their product as C
q
-
valued functions fg = gf ∈ T
M
. This algebraic situation is described by the theory of
H-algebras.
6.1.3 H-algebras and Q-holomorphic functions
An H-algebra is a quaternionic version of an algebra over a commutative field. Let F
be a commutative field (usually the real or complex numbers). An F-algebra is a vector
space A over F, equipped with an F-bilinear multiplication map µ : AA →A which
has certain algebraic properties. For example if µ(a, b) = µ(b, a) for all a, b ∈ A then µ
is said to be commutative. As we have already seen in Section 1.3, this formulation is of
no great use to us when seeking a quaternionic analogue because the non-commutativity
of the quaternions makes the notion of an H-bilinear map untenable.
The axioms for an algebra over F can alternatively be written in terms of tensor
products. The main feature of tensor algebra is that a bilinear map on the cartesian
product AB translates into a linear map on the tensor product A⊗
F
B. So our bilinear
multiplication map µ : A A → A becomes an F-linear map µ : A ⊗
F
A → A. The
commutative axiom µ(a, b) = µ(b, a) becomes µ(a⊗b) = µ(b ⊗a), so µ(a⊗b −b ⊗a) =
µ(a ∧ b) = 0. Hence we obtain a ‘tensor algebra version’ of the commutative axiom,
saying that µ : A ⊗
F
A →A is commutative if and only if Λ
2
A ⊆ ker µ.
Once we have reformulated our axioms in terms of tensor products and linear maps we
can translate them into quaternionic algebra, replacing ‘vector space’, ‘linear map’ and
‘tensor product’ with ‘AH-module’, ‘AH-morphism’ and ‘quaternionic tensor product’.
This is precisely what Joyce does, producing the following definition [J1, ¸5]:
Axiom H. (i) T is an AH-module.
(ii) There is an AH-morphism µ
P
: T⊗
H
T →T, called the multiplication
map.
83
(iii) Λ
2
H
T ⊂ ker µ
P
. Thus µ
P
is commutative.
(iv) The AH-morphisms µ
P
: T⊗
H
T → T and id : T → T com-
bine to give AH-morphisms µ
P

H
id and id ⊗
H
µ
P
: T⊗
H
T⊗
H
T →
T⊗
H
T. Composing with µ
P
gives AH-morphisms µ
P
◦ (µ
P

H
id)
and µ
P
◦ (id ⊗
H
µ
P
) : T⊗
H
T⊗
H
T → T. Then µ
P
◦ (µ
P

H
id) =
µ
P
◦ (id ⊗
H
µ
P
). This is associativity of multiplication.
(v) An element 1 ∈ A called the identity is given, with 1 / ∈ T

and I1 ⊆
T

.
(vi) Part (v) implies that if α ∈ T

then α(1) ∈ R. Thus for each a ∈ T,
1⊗
H
a and a⊗
H
1 ∈ T⊗
H
T by Lemma 4.1.7. Then µ
P
(1⊗
H
a) =
µ
P
(a⊗
H
1) = a for each a ∈ T. Thus 1 is a multiplicative identity.
Definition 6.1.4 T is an H-algebra if T satisfies Axiom H.
Here H-algebra stands for Hamilton algebra. We also define morphisms of H-algebras:
Definition 6.1.5 Let T, Q be H-algebras, and let φ : T → Q be an AH-morphism.
Write 1
P
, 1
Q
for the identities in T, Q respectively. We say that φ is an H-algebra
morphism if φ(1
P
) = 1
Q
and µ
Q
◦ (φ⊗
H
φ) = φ ◦ µ
P
as AH-morphisms T⊗
H
T →Q.
Let M be a hypercomplex manifold. Joyce has constructed a multiplication map
µ
M
with respect to which the q-holomorphic functions T
M
form an H-algebra, and such
that for f, g ∈ C

(M, C
q
) holomorphic with respect to Q we have µ(f⊗
H
g) = fg.
In this way the the H-algebra T
M
includes the algebras of holomorphic functions with
respect to all the different complex structures on M.
Let M, N and (therefore) M N be hypercomplex manifolds. We define an AH-
morphism φ : T
M

H
T
N
→ T
M×N
. Let U and V be AH-modules. Let x ∈ U

⊗ V

and y ∈ U⊗
H
V . Then y x ∈ H, where ‘ ’ contracts together the factors of U

⊗ V

with those of (U

)

⊗(V

)

in U⊗
H
V ⊂ H⊗(U

)

⊗(V

)

. Define a linear map
λ
UV
: U

⊗V

→(U⊗
H
V )

by setting λ
UV
(x)(y) = y x, (6.2)
for all x ∈ U

⊗V

and y ∈ U⊗
H
V .
2
Consider again the maps θ
m
∈ T

M
of Lemma 6.1.1, which allow us to interpret
points of m ∈ M as AH-morphisms θ
m
: T
M
→ H. In the same way, we can associate
to each point (m, n) ∈ M N the map θ
m
⊗θ
n
∈ T

M
⊗T

N
. Then λ
P
M
,P
N

m
⊗θ
n
) ∈
(T
M

H
T
N
)

.
Definition 6.1.6 We define a map φ : T
M

H
T
N
→ C

(M N, H) as follows. Let
f ∈ T
M

H
T
N
. Then λ
P
M
,P
N

m
⊗ θ
n
) f ∈ H and we define φ(f) : M N → H by
setting
φ(f)(m, n) = λ
P
M
,P
N

m
⊗θ
n
) f.
For infinite-dimensional vector spaces U and V we define U ⊗ V to consist of
finite sums of elements u ⊗ v. Thus φ(f) is a sum of finitely many smooth func-
tions, and so is smooth. It is easy to see that φ(f) is q-holomorphic, since D
M×N
=
2
See [J1, Definition 4.2], where Joyce defines this map and uses it to prove that U⊗
H
V is an AH-
module.
84
D
M
(φ(f)) + D
N
(φ(f)) = 0, i.e. we can evaluate the derivatives in the M and N
directions separately. It is clear that φ is H-linear and that φ(T

M

H
T

N
) ⊆ T

M×N
.
Thus we have a canonical AH-morphism φ : T
M

H
T
N
→T
M×N
.
The second (and easier) step is to consider the case M = N, so φ : T
M

H
T
M

T
M×M
. The diagonal submanifold M
diag
= ¦(m, m) : m ∈ M¦ ⊂ M M is a submani-
fold of M M isomorphic to M as a hypercomplex manifold, and each q-holomorphic
function on M M restricts to a q-holomorphic function on M
diag
. Let ρ be the re-
striction map ρ : T
M×M
→ T
M
diag
; then ρ is an AH-morphism. Thus we can define an
AH-morphism
µ
M
= ρ ◦ φ : T
M

H
T
M
→T
M
. (6.3)
Here is the key theorem of this section:
Theorem 6.1.7 [J1, 5.5]. Let M be a hypercomplex manifold, so that T
M
is an AH-
module. Let 1 ∈ T
M
be the constant function on M with value 1, and let µ
M
:
T
M

H
T
M
→ T
M
be the AH-morphism µ
M
= ρ ◦ φ of Equation (6.3). With these
definitions, T
M
is an H-algebra.
This statement is also true for C

(M, H). An AH-morphism φ : C

(M, H)⊗
H
C

(N, H) → C

(M N, H) can be defined just as in Definition 6.1.6, and ρ is
still the obvious restriction. The q-holomorphic functions T
M
form an H-subalgebra
of C

(M, H), and the inclusion map id : T
M
→C

(M, H) is an H-algebra morphism.
Example 6.1.8 Q-holomorphic functions on H [J1, ¸10].
Consider the manifold H with coordinates (x
0
, x
1
, x
2
, x
3
) representing the quaternion
x
0
+ x
1
i
1
+ x
2
i
2
+ x
3
i
3
. Then H is naturally a complex manifold with hypercomplex
structure (I
1
, I
2
, I
3
) given by
I
1
dx
2
= dx
3
, I
2
dx
3
= dx
1
, I
3
dx
1
= dx
2
and I
j
dx
0
= dx
j
, j = 1, 2, 3. (6.4)
Consider the linear H-valued functions on H. Using the Cauchy-Riemann-Fueter
Equation (1.20) with the hypercomplex structure in (6.4), we find that f = q
0
x
0
+
q
1
x
1
+ q
2
x
2
+ q
3
x
3
is q-holomorphic if and only if q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0. Thus
the linear q-holomorphic functions on H form an AH-submodule of the set of all linear
functions, and this submodule is isomorphic to the AH-module Z

= U
1
of Example
5.4.2.
Consider the AH-module U
(k)
of homogeneous q-holomorphic polynomials of degree
k. The spaces U
(k)
are important in quaternionic analysis and are studied by Sudbery
[Su, ¸6]. Joyce uses the dimension formulae of Proposition 4.1.16 to prove that S
k
H
Z

=
U
(k)
. We can easily show that
S
k
H
Z

= (k + 1)V
1
⊗V
k+1
=
_
(k + 1)U
k
k even
1
2
(k + 1)U
k
k odd.
This constructs not only the spaces U
(k)
, but also their structure as Sp(1)-representations,
which is crucial to Sudbery’s approach.
The space of all q-holomorphic polynomials on H is therefore given by the sum


k=0
S
k
H
Z, which is naturally an H-algebra. Joyce [J1, Example 5.1] calls this F
Z
, the
free H-algebra generated by Z , and this idea enables us to write down the multiplication
map µ
H
. The full H-algebra T
H
is constructed by adding in convergent power series.
85
6.2 The Quaternionic Cotangent Space
Let M be a complex manifold and recall the splitting C ⊗
R
T

M = Λ
1,0
M ⊕ Λ
0,1
M
of Equation (2.1). A function f ∈ C

(M, C) is holomorphic if df ∈ Λ
1,0
M. The
Cauchy-Riemann operator is defined by ∂ = π
0,1
◦ d, and f is holomorphic if and only
if ∂f = 0.
This section presents the quaternionic analogue of this description. Let M be a
hypercomplex manifold. We show that there is a natural splitting of the quaternionic
cotangent space H⊗T

M

= A⊕B. The bundle A is then the q-holomorphic cotangent
space of M. A function f ∈ C

(M, H) is q-holomorphic if and only if df ∈ A, and the
Cauchy-Riemann-Fueter operator can be written
¯
δ = π
B
◦ d, where π
B
is the natural
projection to the subspace B ⊂ H ⊗ T

M. The ease with which quaternionic algebra
presents such close parallels with complex geometry could scarcely be more rewarding.
Theorem 6.2.1 Let M
4n
be a hypercomplex manifold and let H⊗T

M be the quater-
nionic cotangent bundle of M. The hypercomplex structure determines a natural splitting
H⊗T

M

= A ⊕B,
where dimA = 12n and dimB = 4n.
Proof. Recall from Section 3.2 that T

M

= 2nV
1
as an Sp(1)-representation. Call this
representation 2nV
G
1
(since it is the action defined by the geometric structure of M ).
Following the standard arguments of Chapter 5, we have a splitting
H⊗T

M

= V
L
1
⊗V
R
1
⊗2nV
G
1

= 2nV
L
1
⊗(V
RG
2
⊕V
RG
0
)

= A ⊕B,
where A = 2nV
L
1
⊗V
RG
2
and B = 2nV
L
1
⊗V
RG
0
.
This is a situation with which we are by now very familiar. Using the theory of
Chapter 5, we immediately deduce that
• The (fibres of the) subspaces A and B are AH-modules with A

= nU
1
and
B

= nU
×
−1
.
• Dim A = 12n and dimA

= 8n, where A

= A ∩ I ⊗ T

M. A is a stable AH-
module.
• Dim B = 4n and dimB

= 0, where B

= B ∩ I ⊗ T

M. B is an antistable
AH-module.
• The mapping A ⊕ B → H ⊗ T

M is an injective AH-morphism which is an
isomorphism of the total spaces. It is not an AH-isomorphism because dim(A

⊕B

)
is smaller than dim(I ⊗T

M).
86
The importance of this splitting lies partly in the following result:
Theorem 6.2.2 A function f ∈ C

(M, H) is q-holomorphic if and only if df ∈
C

(A).
The proof of this theorem will be in two parts. Firstly, we investigate the projection
operators π
A
and π
B
from H⊗T

M onto A and B. We then show that the Cauchy-
Riemann-Fueter operator can be written as π
B
◦ d.
To work out projection maps to A and B we use the Casimir operator for the
diagonal Sp(1)
RG
-action, obtained by coupling right-multiplication on H with the action
of the hypercomplex structure. The diagonal Lie algebra action is given by
1(ω) = I
1
(ω) −ωi
1
¸(ω) = I
2
(ω) −ωi
2
/(ω) = I
3
(ω) −ωi
3
. (6.5)
Then C(ω) = (1
2

2
+/
2
)(ω) = −6ω−2χ(ω), where χ(ω) = I
1
(ω)i
1
+I
2
(ω)i
2
+I
3
(ω)i
3
.
The operator χ satisfies the equation χ
2
= 3 − 2χ, so has eigenvalues +1 and −3. If
ω is in the −3 eigenspace then we have C(ω) = 0, so ω ∈ V
RG
0
. If ω is in the +1
eigenspace then we have C(ω) = −8, so ω ∈ V
RG
2
. Thus the subspaces A and B are
the eigenspaces of the operator χ. This enables us to write down expressions for π
A
and π
B
such that π
A
+ π
B
= 1.
Lemma 6.2.3 Projection maps π
A
: H⊗T

M →A and π
B
: H⊗T

M →B are given
by
π
A
(ω) =
1
4
(3ω+I
1
(ω)i
1
+I
2
(ω)i
2
+I
3
(ω)i
3
) π
B
(ω) =
1
4
(ω−I
1
(ω)i
1
−I
2
(ω)i
2
−I
3
(ω)i
3
).
These maps depend only on the structure of M as a hypercomplex manifold.
Just as we defined the Dolbeault operators in Equation (2.4), we define new differ-
ential operators δ and
¯
δ:
δ : C

(M, H) →C

(M, A)
δ = π
A
◦ d
and
¯
δ : C

(M, H) →C

(M, B)
¯
δ = π
B
◦ d.
(6.6)
Then for a function f ∈ C

(M, H), df = δf +
¯
δf.
Proposition 6.2.4 A function f ∈ C

(M, H) is q-holomorphic if and only if
¯
δf = 0.
Proof. Let f = f
0
+ f
1
i
1
+ f
2
i
2
+ f
3
i
3
be a q-holomorphic function on M, so Df =
df
0
+ I
1
df
1
+ I
2
df
2
+ I
3
df
3
= 0. This is exactly the real part of the equation
4
¯
δf ≡ df −I
1
dfi
1
−I
2
dfi
2
−I
3
dfi
3
= 0. (6.7)
Moreover, the three imaginary parts are the equations I
j
(Df) = 0, which are satisfied
if and only if Df = 0. Thus the real equation Df = 0 is exactly the same as the
quaternionic equation
¯
δf = 0.
87
It now follows that a function f is q-holomorphic if and only if df = δf ∈ C

(M, A).
This proves Theorem 6.2.2 and motivates the following definition:
Definition 6.2.5 Let M be a hypercomplex manifold and let H⊗T

M

= A ⊕B, as
in Theorem 6.2.1. Then A is the q-holomorphic cotangent space of M and B is the
q-antiholomorphic cotangent space of M.
Here are some practical methods for writing the spaces A and B. Let ω = 1 ⊗ω
0
+
i
1
⊗ω
1
+ i
2
⊗ω
2
+ i
3
⊗ω
3
. It is straightforward to verify that
ω ∈ A ⇐⇒ω
0
+ I
1

1
) + I
2

2
) + I
3

3
) = 0,
and
ω ∈ B ⇐⇒ω
0
= I
1

1
) = I
2

2
) = I
3

3
).
Let M
4n
be a hypercomplex manifold and let ¦e
a
: a = 0, . . . , 4n − 1¦ be a basis
for T

x
M such that I
b
(e
4a
) = e
4a+b
. In other words, we choose a particular isomorphism
T

x
M

= H
n
such that the subspace ¸e
4a
, e
4a+1
, e
4a+2
, e
4a+3
) is a copy of H with its
standard hypercomplex structure. With respect to this basis the fibres of A and B at
the point x are given by
A
x
=
_

q
a
e
a
: q
4b
+ q
4b+1
i
1
+ q
4b+2
i
2
+ q
4b+3
i
3
= 0 for all b = 0, . . . , n −1
_
(6.8)
and
B
x
=
_

q
a
e
a
: q
4b
= −q
4b+1
i
1
= −q
4b+2
i
2
= −q
4b+3
i
3
for all b = 0, . . . , n −1
_
.
(6.9)
Equation (6.8) is a higher dimensional version of the equation q
0
+q
1
i
1
+q
2
i
2
+q
3
i
3
= 0
which gives the q-holomorphic cotangent space on H (Example 6.1.8). This gives an
AH-isomorphism between A
x
and the AH-module nZ

= nU
1
.
Let us pause for a moment to reflect on what has happened. On a complex manifold,
we have a splitting of C⊗T

M into two spaces of equal dimension, which are conjugate
to one another. On a hypercomplex manifold, we have a splitting into two subspaces,
but they are not equal in size. Instead, one of them has three times the dimension of the
other, because the subspaces are defined by the equation V
1
⊗V
1

= V
2
⊕V
0
. This type
of splitting into spaces of dimension 3n and n is typical of the quaternions, echoing the
fundamental isomorphism H

= I ⊕R.
6.2.1 The Holomorphic and Q-holomorphic cotangent spaces
Let M
4n
be a hypercomplex manifold with hypercomplex structure (I
1
, I
2
, I
3
). Let
q = a
1
i
1
+a
2
i
2
+a
3
i
3
∈ S
2
and let Q = a
1
I
1
+a
2
I
2
+a
3
I
3
be the corresponding complex
structure on M. We have already seen (Lemma 6.1.3) that if f is a complex-valued
function on M which is holomorphic with respect to the complex structure Q, then
ι
q
(f) is a q-holomorphic function. This is a correspondence not only of holomorphic
and q-holomorphic functions, but also of the holomorphic and q-holomorphic cotangent
spaces.
88
Each complex structure Q on M defines a holomorphic cotangent space Λ
1,0
Q
. Using
the embedding ι
q
we can consider C⊗T

M

= C
q
⊗T

M as a subspace of H⊗T

M. The
relationship between these subspaces of H⊗T

M is clarified in the following Lemma:
Lemma 6.2.6 Let A ⊂ H ⊗ T

M be the q-holomorphic cotangent space of a hyper-
complex manifold M and let Λ
1,0
Q
⊂ C⊗T

M be the holomorphic cotangent space with
respect to the complex structure Q. Then
H ι
q

1,0
Q
) = (id
A

H
χ
q
)(A⊗
H
X
q
).
Proof. We illustrate the case dimM = 4, the general result holding in exactly the
same way but involving more coordinates. For any point x ∈ M we choose a basis
¦e
0
, e
1
, e
2
, e
3
¦ for T

x
M such that the hypercomplex structure at x is given by the
standard relations (6.4). Without loss of generality take q = i
1
and consider the complex
manifold (M, I
1
). Then
Λ
1,0
I
1
= ¸e
0
+ ie
1
, e
2
+ ie
3
)
C
.
Since ι
i
1
(a + ib) = a + i
1
b, we see that
H ι
q

1,0
I
1
) = H ¸e
0
+ i
1
e
1
, e
2
+ i
1
e
3
). (6.10)
Recall from Section 5.3 that (id
U

H
χ
q
)(U⊗
H
X
q
)

= U

∩ qU

for any AH-module
U. Using Equation (6.8), we have A = ¦(q
0
e
0
+ . . . + q
3
e
3
) : q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦
and A

= I ⊗T

M. It follows that
A

∩ i
1
A

= ¦q
0
e
0
+ . . . + q
3
e
3
∈ A : q
j
∈ ¸i
2
, i
3

= ¸i
2
e
0
−i
3
e
1
, i
3
e
0
+ i
2
e
1
, i
2
e
2
−i
3
e
3
, i
3
e
2
+ i
2
e
3
), (6.11)
and the image (id
A

H
χ
i
1
)(A⊗
H
X
i
1
)

= 2X
i
1
is generated over H by this subspace.
Observe that −i
2
(i
2
e
0
− i
3
e
1
) = e
0
+ i
1
e
1
and that −i
2
(i
2
e
2
− i
3
e
3
) = e
2
+ i
1
e
3
.
Therefore H (A

∩ i
1
A

) is exactly the same as the subspace H ι
q

1,0
I
1
) of Equation
(6.10).
This gives us a description of the q-holomorphic and q-antiholomorphic cotangent
spaces of M in terms of complex geometry:
Corollary 6.2.7 Let M
4n
be a hypercomplex manifold, with q-holomorphic cotangent
space A and q-antiholomorphic cotangent space B. Then
A =

Q∈S
2
H ι
q

1,0
Q
)
and
B =

Q∈S
2
H ι
q

0,1
Q
).
Thus the q-holomorphic cotangent space is generated over H by the different holomorphic
cotangent spaces.
Proof. Use the fact that A is a stable AH-module; thus A is generated over H by the
subspaces A

∩ qA

for q ∈ I. The proof now follows immediately from Lemma 6.2.6.
89
6.3 AH-bundles
An AH-bundle is the natural quaternionic analogue of a real or complex vector bundle:
a set of AH-modules E parametrised smoothly by a base manifold M.
Definition 6.3.1 Let M be a differentiable manifold and (W, W

) a fixed AH-
module. A smooth AH-module bundle with fibre W or simply AH-bundle is a family
¦(E
x
, E

x

x∈M
of AH-modules parametrised by M, together with a differentiable man-
ifold structure on E =

x∈M
E
x
such that
• The projection map π : E →M, π : E
x
→x is C

.
• For every x
0
∈ M, there exists an open set U ⊆ M containing x
0
and a diffeo-
morphism
φ
U
: π
−1
(U) →U W
such that φ(E
x
) →¦x¦ W is an AH-isomorphism for each x ∈ U.
A section of an AH-bundle is a smooth map α : M → E such that α(x) ∈ E
x
for
all x ∈ M, and the left H-module of smooth sections of E is denoted by C

(M, E).
Define C

(M, E)

= C

(M, E

). With this definition, it is easy to adapt Lemma 6.1.1
to show that if E is an AH-bundle then C

(M, E) is an AH-module. We can multiply
sections on the left by quaternion-valued functions, and C

(M, E) is a left module over
the ring C

(M, H).
The standard real or complex algebraic operations which give new bundles from old
can be transfered en masse to quaternionic algebra and AH-bundles. Let (E, π
1
, M) and
(F, π
2
, N) be AH-bundles. We define a bundle (E⊗
H
F, MN) by setting (E⊗
H
F)
(m,n)
=
E
m

H
F
n
. We can define a bundle E ⊕F in the same way.
In particular, let E and F be AH-bundles over M. By restricting to the diago-
nal submanifold of M M we define a bundle (E⊗
H
F, M) by setting (E⊗
H
F)
m
=
E
m

H
F
m
. In the same way (E ⊕ F, M), Λ
k
H
E and S
k
H
(E) are AH-bundles, as is E
×
if the fibres of E are SAH-modules. We will always make clear whether an AH-bundle
‘ E⊗
H
F ’ refers to (E⊗
H
F, M M) or the diagonal restriction (E⊗
H
F, M).
The q-holomorphic and q-antiholomorphic cotangent bundles A and B are both
AH-bundles, and are AH-subbundles of H ⊗ T

M. In fact, every AH-bundle can be
regarded as an AH-subbundle of some bundle H⊗V , as is easy to demonstrate:
Lemma 6.3.2 Let E = (E, π, M) be an AH-bundle. Then E is isomorphic to an
AH-subbundle of H⊗V , for some real vector bundle (V, π
1
, M).
Proof. This is an extension of the fact that any AH-module W is isomorphic to ι
W
(W) ⊆
H ⊗ (W

)

. Consider the map ι
E
x
(E
x
) : E
x
→ H ⊗ (E

x
)

for x ∈ M. Let V
x
= (E

x
)

.
Then V =

x∈M
V
x
is a real vector bundle, and H ⊗ V is an AH-bundle. Thus E is
isomorphic to the AH-subbundle
_
x∈M
ι
E
x
(E
x
) ⊆ H⊗V,
and the lemma is proved.
90
Abusing notation slightly, if E is an AH-bundle we will often write ι
E
: E →
H⊗(E

)

for the set of inclusion maps

x∈M

E
x
: E
x
→H⊗(E

x
)

).
We could broaden our definition of an AH-bundle, demanding only that E should be
a real vector bundle possessing a left H-action and a real subbundle E

such that (E
x
, E

x
)
is an AH-module for all x ∈ M. The following example shows why this definition would
be unwieldy:
Example 6.3.3 Let M = R, and define a real vector bundle E = RH. Define a real
subspace E

⊂ E by
E

x
= ¸i
1
, i
2
+ xi
3
).
It is easy to check that (E
x
, E

x
) = X
i
3
−xi
2
as an AH-module. Thus E is not an AH-
bundle, since the fibres E
x
and E
y
are not AH-isomorphic for x ,= y. There is good
reason for excluding E from being an AH-bundle. For example,
(E⊗
H
X
i
3
)
x
=
_
X
i
3
if x = 0
0 otherwise.
So despite the fact that E is a smooth vector bundle, E⊗
H
X
q
is not; if we considered
this broader class of objects to be ‘AH-bundles’ then they would not be closed under the
tensor product.
Let V be a real vector bundle over the manifold M with fibre R
k
, and let P
V
be the bundle of frames associated with V . Then P
V
is a principal bundle with fibre
GL(k, R), and V = P
V

GL(k,R)
R
k
. In the same way, let E be an AH-bundle with fibre
W and let G = Aut
AH
(W). We can define a principal G-bundle P
E
associated with
E so that E = P
E

G
W. As we know, G will be a subgroup of GL(k, H). In most
cases, G will be significantly smaller than GL(k, H). For example, if W = U
2j
for some
j then by Theorem 4.2.9 we have Hom
AH
(U
2j
, U
2j
)

= R, so Aut
AH
(U
2j
) = R

= R¸ ¦0¦
and P
E
is just a principal R

-bundle. Principal bundles associated with AH-bundles
tend to be much smaller than those associated with real or complex vector bundles.
6.3.1 Connections on AH-bundles
Definition 6.3.4 Let E be an AH-bundle over M. An AH-connection on E is an
AH-morphism

E
: C

(M, E) −→C

(M, E ⊗T

M) = Ω
1
(M, E)
which satisfies the rule

E
(f α) = df ⊗α + f ∇
E
α
for all α ∈ C

(M, E), f ∈ C

(M, H).
We will sometimes omit the subscript and just write ‘ ∇’ when the AH-bundle E
is clearly implied. Let ∇ be an AH-connection on E. Then the map id ⊗(∇

E
)

is
an AH-connection on H⊗(E

)

which preserves ι
E
(E). Thus every AH-connection ∇
on E is equivalent to a connection (∇

)

on the real vector bundle (E

)

. We will not
distinguish between ∇ and (∇

)

, but will write ∇ for both. This equivalence allows us
to construct AH-connections on tensor products, in very much the same way that Joyce
constructs the AH-morphism φ⊗
H
ψ from AH-morphisms φ and ψ [J1, Definition 4.4].
91
Lemma 6.3.5 Let (E, M) and (F, N) be AH-bundles and let ∇
E
and ∇
F
be AH-
connections on E and F. Then ∇
E
and ∇
F
induce an AH-connection ∇
E⊗
H
F
on the
AH-bundle (E⊗
H
F, M N).
Proof. We regard ∇
E
as a connection on (E

)

. The bundle ((E

)

, M) extends
trivially to a bundle ((E

)

, M N) over M N. Using the natural identification
T

(m,n)
(M N)

= T

m
M ⊕T

n
N, define a differential operator ∇
E,M
on ((E

)

, M N),
where ∇
E,M
differentiates only in the M-directions of M N. In the same way define
a connection ∇
F,N
on ((F

)

, M N) which differentiates only in the N-directions.
Adding these gives a differential operator

E,M
⊗id +id ⊗∇
F,N
: C

((E

)

⊗(F

)

) −→C

((E

)

⊗(F

)

⊗T

(M N)).
This operator is a connection on the real vector bundle (E

)

⊗ (F

)

, so the operator
id ⊗(∇
E,M
⊗ id +id ⊗∇
F,N
) on the bundle (H ⊗ (E

)

⊗ (F

)

, M N) is an AH-
connection.
Since ∇
E
is an AH-connection, the operator id ⊗∇
E,M
⊗id maps sections of ι
E
(E)⊗
(F

)

to sections of ι
E
(E) ⊗(F

)

⊗T

(MN), and acts trivially on the (F

)

factor;
thus id ⊗∇
E,M
⊗id maps sections of E⊗
H
F to sections of E⊗
H
F ⊗T

(M N). The
same is true for id ⊗id ⊗∇
F,N
, and so the AH-connection id ⊗(∇
E,M
⊗ id +id ⊗∇
F,N
)
preserves the AH-subbundle (E⊗
H
F, MN), on which it is therefore an AH-connection.
We call this connection ∇
E⊗
H
F
.
If E and F are AH-bundles over the same manifold M we can use Lemma 6.3.5 to
define an AH-connection on the diagonal bundle (M, E⊗
H
F).
This process can be described in terms of principal bundles.
3
Suppose that φ ∈
Aut
AH
(W) for some AH-module W. Then φ induces a real linear isomorphism (φ
×
)

:
(W

)

→(W

)

. In this way the action of G = Aut
AH
(W) on W gives rise to an action
on (W

)

. Let E be an AH-bundle with fibre W, and let P be the associated principal
bundle with fibre G, so that E = P
G
W. An AH-connection ∇
E
on E gives rise
to a connection D
P
on the principal bundle P, and vice versa. Since connections on
principal bundles always exist, we can use this to guarantee that every AH-bundle has
an AH-connection.
Let F be another AH-bundle with fibre V , let H = Aut
AH
(V ) and let Q be the
associated principal bundle with fibre H, so that F = Q
H
V . An AH-connection ∇
F
on F gives rise to a connection D
Q
on the principal bundle Q. Consider the principal
bundle P Q, a bundle over M N with fibre G H. Since G acts on (W

)

and
H acts on (V

)

, there is an induced action of G H on W⊗
H
V . The AH-bundle
(E⊗
H
F, MN) is the associated bundle (P Q)
G×H
W⊗
H
V . Since the connections
D
P
and D
Q
define a unique connection D
P
⊕D
Q
on the product bundle P Q, this
gives rise to a connection on the associated bundle E⊗
H
F. Since G H acts as AH-
automorphisms on W⊗
H
V , this induced connection is an AH-connection, which we can
identify with the AH-connection ∇
E⊗
H
F
of Lemma 6.3.5.
3
A similar discussion for connections in complex vector bundles can be found in [K, ¸1.5].
92
6.4 Q-holomorphic AH-bundles
Using complex geometry as a model, we define a q-holomorphic AH-bundle over a
hypercomplex manifold M to be one which is holomorphic with respect to each complex
structure Q ∈ S
2
. An AH-bundle is q-holomorphic if and only if it admits a connection
whose curvature takes values only in E
2,0
= Λ
2

. We define the q-holomorphic sections of
a q-holomorphic vector bundle using a version of the Cauchy-Riemann-Fueter equations.
Q-holomorphic sections have interesting algebraic properties, which generalise those of
Joyce’s q-holomorphic functions. We give an interpretation of q-holomorphic sections in
terms of quaternionic algebra and the q-holomorphic cotangent space, which makes the
step-by-step correspondence with complex geometry extremely clear.
Here is the main definition of this section:
Definition 6.4.1 Let E be an AH-bundle over the hypercomplex manifold M, equipped
with an AH-connection ∇. Suppose that ∇ gives E the structure of a holomorphic
vector bundle over the complex manifold (M, Q) for all Q ∈ S
2
. Then (E, ∇) is a
q-holomorphic AH-bundle.
If we want to be really specific we can write (E, π, M, ∇) for our q-holomorphic
bundle; on the other hand, if a connection is already specified, we can simply refer to
the total space E as a q-holomorphic AH-bundle. The existence of such a connection
ensures that the different holomorphic structures are compatible. A lot can be learned
about q-holomorphic AH-bundles by studying the connection ∇.
6.4.1 Anti-self-dual connections and q-holomorphic AH-bundles
Let E be a complex vector bundle over the complex manifold (M, I) and let ∇ be
a connection on E. Just as we split the exterior differential d = ∂ + ∂, we can define

E
= π
1,0
◦∇ : C

(E) →C

(E⊗Λ
1,0
M) and ∂
E
= π
0,1
◦∇ : C

(E) →C

(E⊗Λ
0,1
M).
Proposition 6.4.2 The connection ∇ gives E the structure of a holomorphic vector
bundle if and only if ∂
E
◦ ∂
E
= 0, i.e. the (0, 2)-component of the curvature R of
∇ vanishes. Conversely, if E is a holomorphic vector bundle then E admits such a
connection.
Proof. This is Propositions 3.5 and 3.7 of [K, p. 9].
Let E be a q-holomorphic AH-bundle over the hypercomplex manifold M, so E has
the structure of a holomorphic vector bundle over (M, Q) for every complex structure
Q ∈ S
2
. Let ∇ be a connection on E and let R be the curvature of ∇. Each
Q ∈ S
2
defines a splitting ∇ = ∂
E
Q
+∂
E
Q
, and so a decomposition of the curvature tensor
R = R
2,0
Q
+ R
1,1
Q
+ R
0,2
Q
. By Proposition 6.4.2, ∇ defines a holomorphic structure on
E with respect to Q if and only if R
0,2
Q
≡ 0. If we reverse the sense of Q then we
reverse the decomposition of R, so R
0,2
−Q
= R
2,0
Q
. Thus ∇ gives E the structure of a
holomorphic vector bundle with respect to every Q ∈ S
2
if and only if R
2,0
Q
= R
0,2
Q
= 0
for all Q ∈ S
2
, i.e. the curvature of ∇ is of type (1, 1) with respect to all Q ∈ S
2
.
93
A 2-form ω is of type (1, 1) with respect to every Q ∈ S
2
if and only if it is
annihilated by the action of sp(1) on Λ
2
T

M, so ω ∈ E
2,0
⊂ Λ
2
T

M. By analogy
with the 4-dimensional theory (see Example 3.2.6), we call E
2,2
the space of self-dual
2-forms Λ
2
+
and E
2,0
the space of anti-self-dual 2-forms Λ
2

. A connection ∇ whose
curvature R takes values only in C

(End(E)⊗E
2,0
) is therefore called an anti-self-dual
connection.
Several authors have considered self-dual and anti-self-dual connections, particulary
on quaternionic K¨ahler manifolds: for example Galicki and Poon [GP], Nitta [N] and
Mamone Capria and Salamon [MS]. This is partly because such connections give minima
of a Yang-Mills funtional on M. It is important to note that Mamone Capria and
Salamon refer to the bundle E
2,0
as self-dual rather than anti-self-dual, and so refer
to connections taking values in E
2,0
as self-dual connections. There is no unanimous
convention in the literature: our choice is made because we are using the conventions
that I
j
(e
0
) = e
j
not −e
j
, and that the volume form e
0123
gives a positive orientation of
H.
We have established the following result:
Theorem 6.4.3 Let E be an AH-bundle over the hypercomplex manifold M equipped
with an anti-self-dual AH-connection nabla. Then (E, ∇) is a q-holomorphic AH-
bundle.
Conversely, an AH-bundle E equipped with an AH-connection ∇ is q-holomorphic
if and only if ∇ is anti-self-dual.
This is similar to the idea of hyperholomorphic bundles described by Verbitsky [V, ¸2].
Verbitsky considers the case where B is a Hermitian vector bundle over a hyperk¨ahler
manifold, so the fibres of such a bundle are not normally H-modules. There is no reason
why his definition cannot be extended to hypercomplex manifolds, since as we have seen,
the Hermitian inner product is not necessary to force the connection to be anti-self-
dual if it is to be compatible with the 2-sphere of holomorphic structures. If E is a
q-holomorphic AH-bundle then it is easy to see that H ⊗ (E

)

is also q-holomorphic,
and (E

)

⊗ C is a hyperholomorphic bundle. Conversely, if B is a hyperholomorphic
bundle with a real structure σ such that B = B
σ
⊗C, then H⊗B
σ
is a q-holomorphic
AH-bundle. Any AH-subbundle E of H⊗B
σ
which is preserved by the anti-self-dual
connection ∇ will also be q-holomorphic.
The following Proposition gives further insight into the analogy between holomorphic
and q-holomorphic bundles:
Proposition 6.4.4 Let (E, π, M, ∇) be a q-holomorphic AH-bundle over the hypercom-
plex manifold M. Then E is itself a hypercomplex manifold.
Proof. Consider the splitting of TE into horizontal and vertical subbundles TE

=
E ⊕TM, where the vertical subbundle E is naturally defined by the structure of E as
an AH-bundle and the horizontal subbundle TM is defined by the connection.
Define a hypercomplex structure (I
1
, I
2
, I
3
) on TE as follows. On the vertical
subbundle isomorphic to E, the action of I
1
, I
2
and I
3
is given by the fixed left H-action
of i
1
, i
2
and i
3
on the fibres. On the horizontal subbundle isomorphic to TM, define
94
the hypercomplex structure (I
1
, I
2
, I
3
) to be the horizontal lift of the hypercomplex
structure on M. This defines an almost-hypercomplex structure on E. The integrability
of this structure is guaranteed by the fact that (E, I
j
) is a holomorphic vector bundle
for j = 1, 2, 3.
Example 6.4.5 Let M be a hypercomplex manifold. Then the Obata connection
∇ on T

M is anti-self-dual and defines an anti-self-dual AH-connection on H⊗T

M,
which is thus a q-holomorphic AH-bundle. Let A ⊂ H ⊗ T

M be the q-holomorphic
cotangent space of M,
A = ¦ω
0
+ i
1
⊗ω
1
+ i
2
⊗ω
2
+ i
3
⊗ω
3
: ω
0
+ I
1
ω
1
+ I
2
ω
2
+ I
3
ω
3
= 0, ω
j
∈ T

M¦.
Let a = a
0
+i
1
a
1
+i
2
a
2
+i
3
a
3
∈ C

(M, A). Then ∇a
0
+I
1
(∇a
1
)+I
2
(∇a
2
)+I
3
(∇a
3
) = 0
(where I
1
, I
2
and I
3
act on the A factor of A⊗T

M ), since ∇I
j
= 0 for j = 1, 2, 3. It
follows that ∇a ∈ C

(A⊗T

M). Hence the Obata connection defines an AH-connection
on A. Since the Obata connection is anti-self-dual, A is a q-holomorphic AH-bundle.
Example 6.4.6 Let E and F be q-holomorphic AH-bundles. Then so are the var-
ious associated bundles E⊗
H
F, E ⊕ F, Λ
k
H
E and so on, all with their induced AH-
connections.
6.4.2 Q-holomorphic sections
Q-holomorphic sections of AH-bundles are defined using a version of the Cauchy-Riemann-
Fueter equations. We cannot automatically rewrite Equation (6.7) to define a Cauchy-
Riemann-Fueter operator on a general AH-bundle E, because the fibres of E will not
in general have a well-defined right H-action. Instead, we use the inclusion map ι
E
to
manipulate sections of e in a more manageable form.
Definition 6.4.7 Let E be an AH-bundle over the hypercomplex manifold M equipped
with an AH-connection ∇. Let e ∈ C

(M, E) so that ι
E
(e) = e
0
+ i
1
e
1
+ i
2
e
2
+ i
3
e
3
,
e
j
∈ C

(M, (E

))

. Then e is a q-holomorphic section of E if and only if
∇e
0
+ I
1
(∇e
1
) + I
2
(∇e
2
) + I
3
(∇e
3
) = 0, (6.12)
where I
1
, I
2
and I
3
act on the T

M factor of E ⊗T

M.
Let T(M, E) = T(E) be the space of q-holomorphic sections of E. Let T(E)

=
T(E) ∩ C

(M, E

) so that T(E) is an AH-submodule of C

(M, E).
Q-holomorphic sections are the natural generalisation of q-holomorphic functions,
which are precisely the q-holomorphic sections of the trivial bundle M H equipped
with the flat connection. So T
M
= T(M H).
A general AH-bundle might have no q-holomorphic sections. However, if the AH-
bundle (E, ∇) is q-holomorphic then E is holomorphic with respect to the complex
structure Q, and admits holomorphic sections. It is easy to adapt Lemma 6.1.3 to
show that the holomorphic sections of this holomorphic vector bundle give rise to q-
holomorphic sections of E.
95
Not all q-holomorphic AH-bundles have interesting q-holomorphic sections. For ex-
ample, the q-antiholomorphic cotangent space B ⊂ H⊗T

M is closed under the action
of the Obata connection, and so is technically a q-holomorphic AH-bundle. Recall that
B = ¦ω −i
1
⊗I
1
(ω) −i
2
⊗I
2
(ω) −i
3
⊗I
3
(ω) : ω ∈ T

M¦.
Let b = b
0
−i
1
I
1
b
0
−i
2
I
2
b
0
−i
3
I
3
b
0
∈ C

(M, B), so b is q-holomorphic if and only if
∇b
0
−I
1
∇I
1
b
0
−I
2
∇I
2
b
0
−I
3
∇I
3
b
0
= 0.
Since ∇I
j
= 0 and I
2
j
= −1, this equation is satisfied if and only if ∇b = 0, and the
AH-bundle B admits no non-trivial q-holomorphic sections. In fact, this follows from
the fact that B

= 0, and such behaviour is predicted and described by the algebra of
the quaternionic cotangent space.
Q-holomorphic sections and the quaternionic cotangent space
Let E be a q-holomorphic AH-bundle over the hypercomplex manifold M. We can
describe the q-holomorphic sections of E using the q-holomorphic cotangent space, just
as we did for q-holomorphic functions in the previous chapter. The inclusion map ι
E
and the AH-connection ∇ (regarded as a connection on (E

)

) define an AH-morphism
∇ ◦ ι
E
: C

(M, E) → C

(M, H ⊗ (E

)

⊗ T

M). After swapping the H and (E

)

factors, we have a copy of the quaternionic cotangent space H ⊗ T

M, which we can
split into the q-holomorphic and q-antiholomorphic spaces A and B. Thus we have a
map ∇◦ ι
E
: C

(M, E) → C

(M, (E

)

⊗ (A ⊕ B)). We can now use the projections
π
A
and π
B
of the previous chapter to split the action of ∇◦ ι
E
into two operators.
Definition 6.4.8 Let (E, ∇) be a q-holomorphic AH-bundle. We define the pair of
operators δ
E
: C

(E) →C

((E

)

⊗A) and
¯
δ
E
: C

(E) →C

((E

)

⊗B) by
δ
E
= (id
(E

)
∗ ⊗π
A
) ◦ ∇◦ ι
E
and
¯
δ
E
= (id
(E

)
∗ ⊗π
B
) ◦ ∇◦ ι
E
.
Hence we regard E as a subspace of H ⊗ (E

)

, and then the action of ∇ on E
splits into a q-holomorphic part δ
E
and a q-antiholomorphic part
¯
δ
E
so that in effect
∇ = δ
E
+
¯
δ
E
. This is an exact analogue of the complex case where a connection ∇
splits as ∇ = π
1,0
◦ ∇ + π
0,1
◦ ∇. Let V be a holomorphic vector bundle with a
connection ∇ compatible with the holomorphic structure, so that ∂
E
= ∂. A section
v ∈ C

(V ) is holomorphic if and only if π
0,1
◦ ∇(s) = 0. In just the same way, the
operator e →∇e
0
+I
1
(∇e
1
) +I
2
(∇e
2
) +I
3
(∇e
3
) of Definition 6.4.7 is precisely the real
part of
¯
δ
E
, and a section e ∈ C

(E) is q-holomorphic if and only if
¯
δ
E
(e) = 0.
This analogy goes further. In complex geometry, a section v ∈ C

(V ) is holomorphic
if and only if ∂v = 0, which means that ∇v ∈ C

(V ⊗
C
Λ
1,0
M). Here is the quaternionic
analogue of this statement:
96
Proposition 6.4.9 Let (E, π, M, ∇) be a q-holomorphic AH-bundle, and let e ∈ C

(E)
be a q-holomorphic section. Then
∇◦ ι
E
(e) ∈ C

(E⊗
H
A),
where A is the q-holomorphic cotangent space of M.
Proof. Let e be q-holomorphic, so
¯
δ
E
(e) = 0 and ∇◦ ι
E
(e) = δ
E
(e) ∈ C

(H⊗(E

)


T

M). Using the identification (A

)


= T

M we can regard δ
E
(e) as a section of
H⊗(E

)

⊗(A

)

.
The induced connection ∇ on (E

)

preserves ι
E
(E), so δ
E
(e) ∈ C


E
(E)⊗(A

)

).
Clearly δ
E
(e) ∈ C

((E

)

⊗ι
A
(A)), since δ
E
is defined by projection to this subspace.
Thus ∇◦ ι
E
(e) ∈ C

(E⊗
H
A), by Definition 4.1.4.
In complex geometry, a holomorphic section v of a holomorphic vector bundle V is
one whose covariant derivative ∇v takes values in the complex tensor product of V with
the holomorphic cotangent space. In hypercomplex geometry, a q-holomorphic section e
of a q-holomorphic vector bundle E is one whose covariant derivative ∇e takes values
in the quaternionic tensor product of E with the q-holomorphic cotangent space.
6.4.3 Sections of Tensor Products
For real and complex vector bundles, there is a natural inclusion C

(M, E) ⊗C

(N, F)
→C

(MN, E⊗F) given by (e⊗f)(m, n) = e(m)⊗f(n). The quaternionic analogue
of this map is more delicate. Let E and F be AH-bundles. We want to define an AH-
morphism φ : C

(M, E)⊗
H
C

(N, F) → C

(M N, E⊗
H
F). The obvious difficulty
is that for sections e and f of E and F respectively, there will not in general be a
section e⊗
H
f of the AH-bundle (E⊗
H
F, M N). Instead we use a generalisation of
Joyce’s map φ : T
M

H
T
N
→T
M×N
(Definition 6.1.6).
Consider first the fibres E
m
and F
n
for m ∈ M and n ∈ N. Let α ∈ E

m
and
β ∈ F

n
. Define an AH-morphism α
m
: C

(M, E) → H by α
m
(e) = α(e(m)) for all
e ∈ C

(M, E). Then α
m
∈ C

(M, E)

. Similarly, define the ‘evaluation at n ∈ N ’,
β
n
(f) = β(f(n)). Then β
n
∈ C

(N, F)

. The operators α
m
and β
m
are generalisations
of Joyce’s θ
m
∈ C

(M, H)

, introduced in Lemma 6.1.1. With H-valued functions
(sections of M H), the evaluation map θ
m
generates the whole of H

m

= R. For
more general AH-bundles, we need to consider the combination of a point m at which
to evaluate sections, and an AH-morphism α ∈ E

m
, since we can no longer assume that
the map α = id generates the whole of E

m
.
For AH-modules U and V , recall the linear map λ
UV
: U

⊗ V

→ (U⊗
H
V )

of
Equation (6.2). Thus there are linear maps λ
E
m
,F
n
: E

m
⊗ F

n
→ (E
m

H
F
n
)

and
λ
C

(M,E),C

(N,F)
: C

(M, E)

⊗C

(N, F)

→(C

(M, E)⊗
H
C

(N, F))

.
Definition 6.4.10 Let ∈ C

(M, E)⊗
H
C

(N, F). Define φ
E,F
()(m, n) ∈ E
m

H
F
n
by the equation
λ
E
m
,F
n
(α ⊗β) φ
E,F
()(m, n) = λ
C

(M,E),C

(N,F)

m
⊗β
n
)
for all m ∈ M, α ∈ E

m
and for all n ∈ N, β ∈ F

n
.
97
Thus φ
E,F
() defines a section of the AH-bundle (E⊗
H
F, M N). By the same
arguments as in Definition 6.1.6, φ
E,F
() is smooth because it is a finite sum of smooth
sections, and so we have a linear map
φ
E,F
: C

(M, E)⊗
H
C

(M, F) →C

(M N, E⊗
H
F).
It is easy to see that φ
E,F
is an injective AH-morphism. We shall call φ
E,F
the section
product map for AH-bundles.
If e ∈ C

(M, E) and f ∈ C

(N, F) satisfy the conditions of Lemma 4.1.7 (so if
ι
E
(e) and ι
F
(f) both take values in C
q
for some q ∈ S
2
) then the section φ
E,F
(e⊗
H
f) is
the complex product of the sections, so φ(e⊗
H
f)(m, n) = e(m)⊗
H
f(n)

= e(m) ⊗
C
q
f(n).
The section product map is the natural tool for relating tensor products of sections with
sections of tensor products and allows us to treat sections of more complicated AH-
bundles using similar techniques to those used by Joyce for q-holomorphic functions.
It is well known that sums and tensor products of holomorphic sections are themselves
holomorphic. The same is true for sums of q-holomorphic sections, and there is an
analogous description for q-holomorphic sections of quaternionic tensor products, using
the section product map φ
E,F
. This is a generalisation of the fact (Definition 6.1.6) that
φ : T
M

H
T
N
→T
M×N
.
Theorem 6.4.11 Let (E, π
1
, M, ∇
E
) and (F, π
2
, N, ∇
F
) be q-holomorphic AH-bundles,
and let T(M, E) ⊂ C

(M, E) and T(N, F) ⊂ C

(N, F) be the AH-modules of their
q-holomorphic sections. Then
φ
E,F
: T(M, E)⊗
H
T(N, F) −→T(M N, E⊗
H
F),
where E⊗
H
F is equipped with the connection ∇
E⊗
H
F
, so the section product map takes
q-holomorphic sections to q-holomorphic sections.
Proof. Consider the anti-self-dual connection ∇
E⊗
H
F
= id ⊗(∇
E,M
⊗ id +id ⊗∇
F,N
).
Then ∇
E⊗
H
F
= δ
E⊗
H
F
+
¯
δ
E⊗
H
F
. Using the natural splitting T

(MN)

= T

M⊕T

N,
we see that
¯
δ
E⊗
H
F
=
¯
δ
E,M
⊗id +id ⊗
¯
δ
F,N
, where
¯
δ
E,M
differentiates sections of ι
E
(E) in
the M directions and then projects to the q-antiholomorphic cotangent space of MN,
and similarly for
¯
δ
F,N
.
Let ∈ T(E)⊗
H
T(F), and let φ
E,F
() ∈ C

(MN, E⊗
H
F), where φ
E,F
is the sec-
tion product map of Definition 6.4.10. Since ∈ ι
P(E)
(T(E)) ⊗(T(F)

)

, it follows that
(
¯
δ
E,M
⊗id)(φ
E,F
()) = 0. Similarly, (id ⊗
¯
δ
F,N
)(φ
E,F
()) = 0. Hence
¯
δ
E⊗
H
F

E,F
()) = 0,
and φ
E,F
() is q-holomorphic.
Just as for q-holomorphic functions, when M = N we can restrict to the diagonal
bundle (E⊗
H
F, M). The restriction map ρ clearly preserves q-holomorphic sections,
and we obtain a natural product ρ ◦ φ
E,F
: T(M, E)⊗
H
T(M, F) → T(M, E⊗
H
F). In
particular, let E be a q-holomorphic AH-bundle over M, let T(E) be the AH-module
of q-holomorphic sections of E and let T
M
= T(H) be the H-algebra of q-holomorphic
functions on M. Then by Theorem 6.4.11, we have a natural AH-morphism
ρ ◦ φ
H,E
: T
M

H
T(E) −→T(H⊗
H
E)

= T(E). (6.13)
We can descibe such an algebraic situation by saying that T(E) is an H-algebra module
over T
M
. Here is the definition of an H-algebra module:
98
Axiom M. (i) Q is an AH-module and (T, µ
P
) is an H-algebra.
(ii) There is an AH-morphism µ
Q
: T⊗
H
Q → Q, called the module
multiplication map.
(iii) The maps µ
P
and µ
Q
combine to give AH-morphisms µ
P

H
id and
id ⊗
H
µ
Q
: T⊗
H
T⊗
H
Q → T⊗
H
Q. Composing with µ
Q
gives AH-
morphisms µ
Q
◦ (µ
P

H
id) and µ
Q
◦ (id ⊗
H
µ
Q
) : T⊗
H
T⊗
H
Q →Q.
Then µ
Q
◦(µ
P

H
id) = µ
Q
◦(id ⊗
H
µ
Q
). This is associativity of module
multiplication.
(iv) For u ∈ Q, 1⊗
H
u ∈ T⊗
H
Q by Lemma 4.1.7. Then µ
Q
(1⊗
H
u) = u
for all u ∈ Q. Thus 1 acts as an identity on Q.
Definition 6.4.12 Q is an H-algebra module if Q satisfies Axiom M.
The idea of an H-algebra module was suggested by Joyce, and the axioms follow a
very similar pattern to those for an H-algebra (Definition 6.1.4). It is relatively easy
to see that the q-holomorphic sections T(E) of a q-holomorphic AH-bundle form an
H-algebra module over the H-algebra T
M
, the module multiplication map being the
section product map φ
H,E
of Equation (6.13). The proof of this statement is obtained
by adapting Joyce’s proof that the q-holomorphic functions T
M
form an H-algebra [J1,
Theorem 5.5].
One possible application of this idea is to study anti-self-dual connections (instantons)
on H. Different connections will give rise to different H-algebra modules of q-holomorphic
sections. This suggests that the theory of H-algebra modules over T
H
might lead to an
algebraic description of instantons on H.
99
Chapter 7
Quaternion Valued Forms and
Vector Fields
This final chapter uses the algebra and geometry developed so far to describe quaternion-
valued tensors on hypercomplex manifolds. On a hypercomplex manifold we have global
complex structures, which allows us to adapt the real-valued double complex of Chapter
3 to decompose quaternion-valued forms. The splitting H ⊗ T

M

= A ⊕ B of the
previous chapter is the first example of this type of decomposition. More generally, for
all k and r we obtain a splitting of H⊗E
k,r
into two AH-bundles. This decomposition
gives rise to a double complex of quaternion-valued forms on hypercomplex manifolds,
which has some advantages over the real-valued double complex. Not only the top row
but the top two rows of the quaternion-valued double complex are elliptic, and in four
dimensions the whole complex is elliptic. The top row of the quaternion-valued double
complex is particularly well-behaved, and can be constructed using quaternionic algebra.
This allows us to define q-holomorphic k-forms, and to describe their algebraic structure
using ideas from the previous chapter.
A similar approach to quaternion-valued vector fields is also fruitful. Just as with
the quaternionic cotangent space, there is a splitting of the quaternionic tangent space
H⊗TM using which we define a hypercomplex version of the ‘(1, 0) vector fields’ on a
complex manifold. These vector fields are closed under the Lie bracket (suitably adapted
to the quaternionic situation). This encourages us to adapt the axioms for a Lie algebra
to form a quaternionic version, in a similar way to that in which Joyce arrived at H-
algebras. The Lie bracket on hypercomplex manifolds defines a natural operation on
vector fields which satisfies these axioms.
In recent times, Spindel et al. [SSTP] and Joyce [J3] demonstrated the existence of in-
variant hypercomplex structures on certain compact Lie groups and their homogeneous
spaces. This discovery presents us with lots of examples of finite-dimensional quater-
nionic Lie algebras. We use this idea to calculate the quaternionic cohomology groups of
the group U(2), and suggest how these methods may be extended to higher-dimensional
hypercomplex Lie groups.
100
7.1 The Quaternion-valued Double Complex
In Chapter 3, we saw how the real-valued exterior forms Λ
k
T

M on a quaternionic
manifold M are acted upon by the principal Sp(1)-bundle Q of local almost complex
structures on M. Recall that the subbundle of Λ
k
T

M consisting of V
r
-type represen-
tations is denoted E
k,r
= ε
n
k,r
V
r
.
Suppose in addition that M is hypercomplex, and so possesses global complex struc-
tures I
1
, I
2
and I
3
= I
1
I
2
(in other words, Q is the trivial bundle MSp(1) ). Instead
of just real or complex forms (which we can think of as taking values in the trivial repre-
sentation V
0
), consider forms taking values in some Sp(1)-representation W, i.e. sections
of the bundle W ⊗Λ
k
T

M. Since the Sp(1)-action on Λ
k
T

M is now defined by global
complex structures, we can take the diagonal action under which the subspace W ⊗E
k,r
splits according to the Clebsch-Gordon formula.
The situation in which we are particularly interested is that of forms taking values
in the quaternions H = V
L
1
⊗ V
R
1
. As suggested in Section 3.4, we obtain splittings by
coupling the right Sp(1)-action on H with the Sp(1)-action on Λ
k
T

M. In symbols, this
takes the form
H⊗E
k,r

= V
L
1
⊗V
R
1
⊗ε
n
k,r
V
G
r

= ε
n
k,r
V
L
1
⊗(V
RG
r+1
⊕V
RG
r−1
). (7.1)
Proposition 7.1.1 Let M
4n
be a hypercomplex manifold. The AH-bundle H⊗Λ
k
T

M
decomposes as
H⊗Λ
k
T

M

= V
L
1

_
k+1

r=0

n
k,r+1
+ ε
n
k,r−1
)V
RG
r
_
,
where r ≡ k + 1 mod 2.
Proof. By Proposition 3.2.1, we have
H⊗Λ
k
T

M

= V
L
1
⊗V
R
1

_
k

r=0
ε
n
k,r
V
G
r
_
,
where the ‘Sp (1)
G
-action’ is the action of Sp(1) on Λ
k
T

M induced by the hypercomplex
structure. Taking the diagonal Sp(1)
RG
-action using the Clebsch-Gordon formula, this
becomes
H⊗Λ
k
T

M

= V
L
1

_
k

r=0
ε
n
k,r
(V
RG
r+1
⊕V
RG
r−1
)
_
. (7.2)
Collecting together the V
r
representations yields the formula in the Proposition.
Definition 7.1.2 Define F
k,r
to be the subspace (ε
n
k,r+2

n
k,r
)V
1
⊗V
r+1
⊆ H⊗Λ
k
T

M.
The primed part of the space F
k,r
is obtained using the theory of Chapter 5. As in
Equation 5.10, Equation 7.1 is a decomposition of H ⊗ E
k,r
into stable and antistable
AH-modules. (This is a generalisation of the splitting H ⊗ T

M

= A ⊕ B.) Defining
F

k,r
= F
k,r
∩ (I ⊗ Λ
k
T

M), the space F
k,r
is an AH-subbundle of H ⊗ (F

k,r
)

= H ⊗
101
(E
k,r+2
⊕ E
k,r
), and each (fibre of the) AH-bundle F
k,r
is the direct sum of stable and
antistable components. The splitting
H⊗E
k,r

= ε
n
k,r
V
L
1
⊗(V
RG
r+1
⊕V
RG
r−1
)
gives an H-module isomorphism
H⊗E
k,r

=
_
_
_
ε
n
k,r
U
r
⊕ε
n
k,r
U
×
r−2
for r even
1
2
ε
n
k,r
U
r

1
2
ε
n
k,r
U
×
r−2
for r odd.
(7.3)
As with the decomposition H⊗T

M

= A⊕B, these are not AH-isomorphisms because
some of the primed part is lost in the splitting. We give names to these spaces as follows
(where as usual a = 1 if n is even and a = 2 if n is odd):
Definition 7.1.3 Define F

k,r
to be the AH-bundle
1
a
ε
n
k,r
U
r
⊆ H ⊗ E
k,r
. Define F

k,r
to be the AH-bundle
1
a
ε
n
k,r+2
U
×
r
⊆ H ⊗ E
k,r+2
. Thus F

k,r
is the U
r
-type subspace of
H⊗Λ
k
T

M and F

k,r
is the U
×
r
-type subspace of H⊗Λ
k
T

M, so that
F
k,r
= F

k,r
⊕F

k,r
.
The q-holomorphic cotangent space A is F
1,1
= F

1,1
and the q-antiholomorphic cotan-
gent space B is F
1,−1
= F

1,−1
.
With these definitions, we have (F

k,r

)

= (F

k,r−2

)

= E
k,r
, and H ⊗ E
k,r

= F

k,r

F

k,r−2
. Just as with the splitting H⊗T

M

= A⊕B, there is an injective AH-morphism
F

k,r
⊕F

k,r−2
→H⊗E
k,r
which is an H-linear isomorphism of the total spaces but is not injective on the primed
parts. A short calculation shows that
dimF

k,r
= 2(r + 2)ε
n
k,r
dimF

k,r

= (r + 3)ε
n
k,r
and
dimF

k,r
= 2(r + 2)ε
n
k,r+2
dimF

k,r

= (r + 1)ε
n
k,r+2
.
Since we can consider the bundles F

k,r
and F

k,r
separately, it may appear strange
to mix up the stable and antistable fibres in the single bundle F
k,r
. However, it soon
becomes clear that exterior differentiation does not necessarily map stable fibres to stable
fibres or antistable fibres to antistable fibres, so in order to obtain a double complex it
is necessary to amalgamate the stable and antistable contributions.
The definition of the operators δ and
¯
δ of Section 6.2 can now be generalised to
cover the whole of H ⊗ Λ

T

M, giving rise to a double complex of quaternion-valued
forms.
Definition 7.1.4 Let π
k,r
be the natural projection map π
k,r
: H ⊗ Λ
k
T

M → F
k,r
.
Define the differential operators δ,
¯
δ : Ω
k
(M, H) →Ω
k+1
(M, H) by
δ : C

(F
k,r
) →C

(F
k+1,r+1
)
δ = π
k+1,r+1
◦ d
and
¯
δ : C

(F
k,r
) →C

(F
k+1,r−1
)
¯
δ = π
k+1,r−1
◦ d .
102
Theorem 7.1.5 The exterior derivative d maps C

(M, F
k,r
) to C

(M, F
k+1,r+1

F
k+1,r−1
), so
d = δ +
¯
δ.
It follows that
δ
2
= δ
¯
δ +
¯
δδ =
¯
δ
2
= 0.
and the decomposition H⊗Λ
k
T

M =

k+1
r=0
F
k,r
gives rise to a double complex.
Proof. The proof works in exactly the same way as that of Theorem 3.2.3. Let ∇ be
the Obata connection on M, so ∇ : C

(M, F
k,r
) →C

(M, F
k,r
⊗T

M). Since
F
k,r
⊗T

M = (ε
n
k,r+2
+ ε
n
k,r
)V
1
⊗V
r+1
⊗2nV
1
,
it follows (from the Clebsch-Gordon splitting V
r+1
⊗2nV
1

= 2n(V
r+2
⊕V
r
), followed by
the antisymmetrisation d = ∧◦∇) that d : C

(M, F
k,r
) →C

(M, F
k+1,r+1
⊕F
k+1,r−1
).
The rest of the theorem follows automatically.
Figure 7.1: The Quaternion-Valued Double Complex
C

(M, F
0,0
= H)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
C

(M, F
1,1
= A)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
C

(M, F
1,−1
= B)
¨
¨
¨
¨
¨
¨B
δ
C

(M, F
2,0
)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
C

(M, F
2,2
)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
C

(M, F
3,1
)
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
C

(M, F
3,3
)
r
r
r
r
r
rj
¯
δ
. . . . . . etc.
C

(M, F
3,−1
)
¨
¨
¨
¨
¨
¨B
δ
. . . . . . etc.
As already hinted, the operators δ and
¯
δ do not preserve stable or antistable sub-
spaces. For example, though we have
¯
δ : F
k,r
→ F
k+1,r−1
, it is not the case that
¯
δ : F

k,r
→F

k+1,r−1
. This can be observed in the simplest of cases — a quaternion-valued
function f is a section of F
0,0
= F

0,0
, and unless f is q-holomorphic
¯
δf is a nonzero
section of F

1,−1
.
Much of the theory from the real-valued double complex of Chapter 3 can be adapted
to describe the quaternion-valued version as well. For example, the operators δ and
¯
δ
can be expressed in a similar fashion to T and T, using the Casimir element technique
of Lemma 3.2.7. The Casimir operator in question is that of the diagonal Lie algebra
action given by the operators
1(ω) = I
1
(ω) −ωi
1
¸(ω) = I
2
(ω) −ωi
2
/(ω) = I
3
(ω) −ωi
3
.
103
of Equation (6.5). This leads to the following adaptation of Lemma 3.2.7:
Lemma 7.1.6 Let α ∈ C

(F
k,r
). Then
δα = −
1
4
_
r +
1
r + 2
(1
2

2
+/
2
)
_

and
¯
δα =
1
4
_
(r + 4) +
1
r + 2
(1
2

2
+/
2
)
_
dα.
The results on ellipticity in Section 3.3 can also be adapted to the new situation. We
can infer that the operator δ is elliptic except at the bottom spaces F
2k−1,−1
and F
2k,0
.
Again, the operator δ is elliptic at some of these lowest-weight spaces for low exterior
powers; in particular the leading edge of spaces F
k,k
forms a complex which is elliptic
throughout. Closer examination also reveals that the ‘second row’ of spaces F
k,k−2
is
also an elliptic complex with respect to δ.
Lemma 7.1.7 The complex
0 −→C

(F
1,−1
)
δ
−→C

(F
2,0
)
δ
−→. . .
δ
−→C

(F
2n+1,2n−1
)
δ
−→0
is elliptic.
Proof. Ellipticity at C

(F
k,k−2
) for all k ≥ 3 follows from a generalisation of the
techniques used in the proof of Theorem 3.3.1. We need to show that the complex is
elliptic at C

(F
1,−1
) = C

(B) and C

(F
2,0
). As in Section 3.3, it is enough to choose
some e
0
∈ T

M and show that the symbol sequence
0 −→F
1,−1
σ
−→F
2,0
σ
−→F
3,1
σ
−→. . . etc.
is exact, where σ(ω) = σ
δ
(ω, e
0
) = π
k+1,r+1
(ωe
0
) for ω ∈ F
k,r
. (As usual ωe
0
means
ω ∧ e
0
.)
Since δ = d on F
1,−1
= B, we have σ(β) = βe
0
for all β ∈ B. It follows from the
expression for B in Equation (6.9) that σ : B → F
2,0
is injective, so the complex is
elliptic at B.
Let ω ∈ F
2,0
. Then σ(ω) = 0 if and only if ωe
0
∈ F
3,−1
, which is the case if and
only if 1(ωe
0
) = ¸(ωe
0
) = /(ωe
0
) = 0. The first of these equations is
I
1
(ωe
0
) −ωe
0
i
1
= I
1
(ω)e
0
+ ωe
1
−ωe
0
i
1
= 0.
It follows by taking exterior product with e
0
that ωe
10
= 0. The same arguments for
¸ and / show that
σ(ω) = 0 =⇒ωe
10
= ωe
20
= ωe
30
= 0,
and so ωe
0
must be equal to zero and ω = γe
0
for some γ ∈ H⊗T

M.
104
It remains to show that we can choose β ∈ B such that ω = βe
0
. Suppose instead
that ω = αe
0
, with α ∈ A. Since the complex is elliptic at A it follows that αe
0
∈ F
2,0
if and only if α = σ(q) =
1
4
q(3e
0
+ e
1
i
1
+ e
2
i
2
+ e
3
i
3
) for some q ∈ H. But then
αe
0
= βe
0
, where β =
1
4
q(−e
0
+e
1
i
1
+e
2
i
2
+e
3
i
3
) ∈ B. Hence we can find β ∈ B such
that σ(ω) = 0 implies that ω = σ(β) for all ω ∈ F
2,0
, so the complex is exact at F
2,0
.
This completes the proof.
This is an unexpected bonus — on hypercomplex manifolds, the double complex of
quaternion-valued forms has not just one but two rows which with respect to the operator
δ are elliptic throughout, namely the top row F
k,k
and the second row F
k,k−2
.
Example 7.1.8 Quaternion-valued forms in four dimensions
Figure 7.2: The Quaternion-Valued Double Complex in Four Dimensions
F
0,0
= H
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
A

= Z
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
B

= U
×
−1
¨
¨
¨
¨
¨
¨B
δ
F
2,0

= 3H⊕H
×
¨
¨
¨
¨
¨
¨B
r
r
r
r
r
rj
δ
¯
δ
Λ
2
H
A

= Y
r
r
r
r
r
rj
¯
δ
F
3,1

= Z
r
r
r
r
r
rj
¯
δ
F
3,−1

= U
×
−1
¨
¨
¨
¨
¨
¨B
δ
F
4,0

= H
In four dimensions the situation is particularly friendly towards quaternion-valued
forms, because the whole double complex is elliptic. To show this, choose a standard
basis of 1-forms ¦e
0
, . . . , e
3
¦ for T

M so that as usual I
j
(e
0
) = e
j
. Explicitly, we have
H⊗T

M

= F
1,1
⊕F
1,−1
, where
F
1,1
= A = ¦q
0
e
0
+ q
1
e
1
+ q
2
e
2
+ q
3
e
3
: q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦,
F
1,−1
= B = ¦q
0
e
0
+ q
1
e
1
+ q
2
e
2
+ q
3
e
3
: q
0
= q
1
i
1
= q
2
i
2
= q
3
i
3
¦.
Next, H⊗Λ
2
T

M

= F
2,2
⊕F
2,0
, where
F
2,2
= Λ
2
H
A = ¦q
1
e
01+23
+ q
2
e
02+31
+ q
3
e
03+12
: q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦,
F
2,0
= F

2,0
⊕F

2,0
= ¸e
01−23
, e
02−31
, e
03−12
)
H
⊕¦q
1
e
01+23
+ q
2
e
02+31
+ q
3
e
03+12
: q
1
i
1
= q
2
i
2
= q
3
i
3
¦.
Lastly, H⊗Λ
3
T

M

= F
3,1
⊕F
3,−1
, where
F
3,1
= ¦q
0
e
123
+ q
1
e
032
+ q
2
e
013
+ q
3
e
021
: q
0
+ q
1
i
1
+ q
2
i
2
+ q
3
i
3
= 0¦,
F
3,−1
= ¦q
0
e
123
+ q
1
e
032
+ q
2
e
013
+ q
3
e
021
: q
0
= q
1
i
1
= q
2
i
2
= q
3
i
3
¦.
It follows that the symbol map σ
δ
is an isomorphism between F
3,−1
and F
4,0
, and so
in four dimensions the entire quaternion-valued double complex is elliptic.
105
7.1.1 The Top Row Λ
k
H
A and Q-holomorphic k-forms
In contrast with the the role of the Cauchy-Riemann operator ∂ in the Dolbeault com-
plex, the Cauchy-Riemann-Fueter operator
¯
δ in the quaternion-valued double complex
(Figure 7.1) does not begin a longer elliptic complex.
1
On the other hand, the operator
δ on functions does extend to give the elliptic complex
0 −→C

(F
0,0
)
δ
−→C

(F
1,1
)
δ
−→. . .
δ
−→C

(F
2n,2n
)
δ
−→0.
This section discusses these top row spaces F
k,k
which are of particular interest.
In complex geometry the Hodge decomposition of forms results immediately from the
isomorphism C⊗T

M

= Λ
1,0
⊕Λ
0,1
and the isomorphism in exterior algebra Λ
k
(U⊕V ) =

p+q=k
Λ
p
U ⊗ Λ
q
V . With hypercomplex geometry we are not quite so lucky, since the
splitting A ⊕ B

= H ⊗ T

M is not an AH-isomorphism but rather an injective AH-
morphism which is not surjective on the primed parts. Despite the fact that we can
identify H⊗Λ
k
T

M with Λ
k
H
(H⊗T

M), the induced AH-morphism
ι
k
: Λ
k
H
(A ⊕B) →H⊗Λ
k
T

M
is therefore not an isomorphism for k > 1. This is why we resort to our more complicated
analysis of the Sp(1)-representation on H⊗Λ
k
T

M to discover the quaternionic version
of the Dolbeault complex. In spite of this, the inclusion map ι
k
is still of special interest,
because it describes explicitly the top row of the double complex.
Proposition 7.1.9 Let A be the q-holomorphic cotangent space of a hypercomplex man-
ifold M. The inclusion map ι
k
: Λ
k
H
(A⊕B) →H⊗T

M identifies Λ
k
H
A with the highest
space F
k,k
⊆ H⊗Λ
k
T

M.
Proof. There is a natural identification Λ
k
H
(A ⊕B) =

p+q=k

p
H
A⊗
H
Λ
q
H
B), and so
ι
k
:

p+q=k
Λ
p
H
A⊗
H
Λ
q
H
B →H⊗Λ
k
T

M.
Since B

= ¦0¦, B⊗
H
B = ¦0¦ and so Λ
k
H
B = ¦0¦ for k > 1 ; also A⊗
H
B = ¦0¦. Thus
Λ
k
H
(A ⊕ B) = Λ
k
H
A for k > 1, and the map ι
k
: Λ
k
H
(A ⊕ B) → H ⊗ Λ
k
T

M is none
other than the normal inclusion map ι : Λ
k
H
A → H ⊗ (Λ
k
H
A

)

. It follows that there is
an AH-submodule of H⊗Λ
k
T

M which is isomorphic to Λ
k
H
A.
Now, A

= nU
1
, so by Theorem 5.2.1, Λ
k
H
A ⊆

k
H
A

= mU
k
for some m. (Recall
that U
k
= aV
1
⊗V
n+1
where a = 1 for k even and a = 2 for k odd.) By Proposition
4.1.16, dimΛ
k
H
(nU
1
) = 2(k + 2)
_
2n
k
_
. It follows that
Λ
k
H
A

=
1
a
_
2n
k
_
U
k
=
_
2n
k
_
V
1
⊗V
k+1
. (7.4)
Thus ι
k

k
H
A) must be a subspace of H⊗Λ
k
T

M of this form.
1
In 1991, Baston [Bas] succeeded in extending the Cauchy-Riemann-Fueter operator to a locally
exact complex with a different construction involving second order operators.
106
From Definition 7.1.2, we see that F
k,k
=
_
2n
k
_
V
1
⊗ V
k+1
as well. Since there are no
other representations of this weight in H⊗Λ
k
T

M, it follows that
ι
k

k
H
A) = F
k,k
,
so there is a natural isomorphism ι
k
: Λ
k
H
A

= F
k,k
.
Just as we define the q-holomorphic cotangent space A as being a particular sub-
module of H ⊗ T

M, so that (A

)

= T

M, we identify its exterior powers with the
corresponding AH-submodules of H ⊗ Λ
k
T

M. Thus we will omit to write the ‘ ι
k
’,
writing Λ
k
H
A = F
k,k
and (Λ
k
H
A

)

= E
k,k
.
In Section 6.2.1 we showed that the q-holomorphic cotangent space A is generated
over H by the various holomorphic cotangent spaces (Corollary 6.2.7). We generalise
this result to higher exterior powers as follows:
Theorem 7.1.10
2
Let A be the q-holomorphic cotangent space of a hypercomplex man-
ifold M
4n
. Then
(id ⊗
H
χ
q
)(Λ
k
H
A⊗
H
X
q
) = H ι
q

k,0
Q
).
It follows that
Λ
k
H
A =

Q∈S
2
H ι
q

k,0
Q
),
i.e. the space Λ
k
H
A is generated over H by the spaces of (k, 0)-forms Λ
k,0
Q
.
Proof. Since Λ
k
H
A

=
_
2n
k
_
V
1
⊗ V
k+1
, it follows that Λ
k
H
A⊗
H
X
q

=
_
2n
k
_
X
q
. This is AH-
isomorphic to the submodule (id ⊗
H
χ
q
)(Λ
k
H
A⊗
H
X
q
) ⊂ Λ
k
H
A, which by Theorem 5.3.2 is
generated over H by the weight-spaces of Q with extremal weight.
Consider the weights of the action of Q ∈ S
2
⊂ sp(1). The highest-weight vectors in
C ⊗Λ
k
T

M are the (k, 0)-forms Λ
k,0
Q
, which are mapped to C
q
⊗Λ
k
T

M by the map
ι
q
. Thus H ι
q

k,0
Q
) ⊂ (id ⊗
H
χ
q
)(Λ
k
H
A⊗
H
X
q
), and since dim
C
Λ
k,0
Q
=
_
2n
k
_
we see that
H ι
q

k,0
Q
) = (id ⊗
H
χ
q
)(Λ
k
H
A⊗
H
X
q
),
proving the first part of the Theorem.
Since Λ
k
H
A is stable, it is generated by these subspaces. The result follows.
Just as the q-holomorphic cotangent space A is our quaternionic analogue of the
holomorphic cotangent space T

1,0
M, the AH-bundle Λ
k
H
A is the quaternionic analogue
of the bundle of (k, 0)-forms Λ
k,0
= Λ
k
C
T

1,0
M. Both are formed in the same way using
exterior algebra over their respective fields, and both form the ‘top row’ of their double
complexes. This suggests a natural definition of a ‘q-holomorphic k-form’:
Definition 7.1.11 Let M be a hypercomplex manifold. A quaternion-valued k-form
ω ∈ Ω
k
(M, H) is q-holomorphic if and only if ω ∈ C

(M, Λ
k
H
A) and
¯
δω = 0. The
AH-module of q-holomorphic k-forms on M will be written T
k
M
, so the H-algebra of
q-holomorphic functions on M is T
0
M
= T
M
.
2
This theorem is really a quaternionic version of (Salamon’s) Equation 2.10, which is essentially the
same result for complexified k-forms.
107
This gives rise to what we may call the q-holomorphic de Rham complex
0 −→T
M
d=δ
−→T
1
M
d=δ
−→T
2
M
d=δ
−→. . .
d=δ
−→T
2n−1
M
d=δ
−→T
2n
M
d=δ
−→0. (7.5)
Just as (on a complex manifold) Λ
k,0
is a holomorphic vector bundle, it is easy to
see that (Λ
k
H
A, ∇) is a q-holomorphic AH-bundle where ∇ is (the connection induced
by) the Obata connection on M. The q-holomorphic k-forms T
k
M
are precisely the q-
holomorphic sections T(Λ
k
H
A) of the AH-bundle (Λ
k
H
A, ∇) as introduced in Definition
6.4.7, as the following Proposition demonstrates:
Proposition 7.1.12 A quaternion-valued k-form ω ∈ C

(M, Λ
k
H
A) satisfies the
equation
¯
δω = 0 if and only if ∇ω ∈ C


k
H
A⊗
H
A), i.e. ω is a q-holomorphic section
of (Λ
k
H
A, ∇).
Proof. The ‘if’ part is automatic, since ∇ω ∈ C


k
H
A⊗
H
A) implies that dω = ∧◦∇ω ∈
C


k+1
H
A) and so
¯
δω = 0.
The reverse implication is nontrivial and depends upon analysing the Sp(1)-
representation on Λ
k
H
A ⊗T

M. Using Equation (7.4), we have
Λ
k
H
A ⊗T

M

=
_
2n
k
_
V
L
1
⊗V
M
k+1
⊗2nV
G
1

= 2n
_
2n
k
_
V
L
1
⊗(V
MG
k+2
⊕V
MG
k
).
The bundle Λ
k
H
A⊗
H
A is precisely the higher-weight subspace 2n
_
2n
k
_
V
L
1
⊗ V
GH
k+2
. The
complementary subspace 2n
_
2n
k
_
V
L
1
⊗ V
GH
k
is revealed by Proposition 7.1.1 to be none
other than the bundle F
k+1,k−1
⊂ H⊗Λ
k+1
T

M. The component of ∇ω taking values
in F
k+1,k−1
is of course
¯
δω, the vanishing of which therefore guarantees that ∇ω ∈
C


k
H
A⊗
H
A).
The q-holomorphic de Rham complex therefore inherits a a rich and interesting alge-
braic structure. We have already noted that d : T
k
M
→T
k+1
M
. It follows from the theory
of q-holomorphic sections that q-holomorphic forms are closed under the tensor product.
Explicitly, let
φ
Λ
k
H
A,Λ
l
H
A
: C

(M, Λ
k
H
A)⊗
H
C

(M, Λ
l
H
A) −→C

(M M, Λ
k
H
A⊗
H
Λ
l
H
A)
be the section product map (Definition 6.4.10). Define φ
k,l
to be the restriction to
C

(M, Λ
k+l
H
A), brought about by the restriction ρ : M M → M to the diagonal
submanifold M
diag
⊂ MM followed by the skewing map ∧ : C

(M, Λ
k
H
A⊗
H
Λ
l
H
A) →
C

(M, Λ
k+l
H
A). It follows (from Theorem 6.4.11 and Proposition 7.1.12) that
φ
k,l
: T
k
M

H
T
l
M
−→T
k+l
M
.
Thus the H-algebra structure on the q-holomorphic functions T
M
extends to one
on the q-holomorphic k-forms T

M
, and we say that T

M
forms a differential graded
H-algebra. It is well known that on a real or complex manifold M one can use exterior
products over R or C to give an algebraic structure to de Rham or Dolbeault cohomology,
which is often called the cohomology algebra of M. It may be that the differential graded
H-algebra structure on T

M
can be used to give a similar description of the quaternionic
cohomology of a hypercomplex manifold.
108
7.2 Vector Fields and Quaternionic Lie Algebras
Quaternionic algebra can also be used to describe vector fields on a hypercomplex man-
ifold M. Using similar ideas to those of Section 6.2, we define a splitting of the quater-
nionic tangent space H ⊗ TM

=
´
A ⊕
´
B. Quaternion-valued vector fields which take
values in the subspace
´
A ⊂ H ⊗ TM turn out to be a good hypercomplex analogue of
the ‘ (1, 0) vector fields’ in complex geometry. In particular, an almost hypercomplex
structure is integrable if and only if these vector fields are closed under a quaternionic
version of the Lie bracket operator. This encourages us to treat these vector fields as a
quaternionic Lie algebra, a new concept which we proceed to define and explore.
7.2.1 Vector Fields on Hypercomplex Manifolds
As so often, we take our inspiration from complex geometry. An almost complex structure
I on a manifold M defines a splitting C ⊗ TM

= T
1,0
M ⊕ T
0,1
M, where T
1,0
M is
the holomorphic and T
0,1
M the antiholomorphic tangent space of M. Let 1 be the
set of smooth vector fields on M. The Lie bracket is a bilinear map from 1 1 to
1. Let 1
1,0
= C

(M, T
1,0
M) be the vector fields of type (1, 0) on M. The almost
complex structure I is integrable if and only if the Lie bracket preserves (1, 0) vector
fields, which is expressed by the inclusion
[1
1,0
, 1
1,0
] ⊆ 1
1,0
. (7.6)
The obstruction to this equation is the Nijenhuis tensor N
I
which measures the (0, 1)
component of the Lie bracket of two (1, 0) vector fields.
We present a similar theorem for quaternionic vector fields on hypercomplex mani-
folds. Let M
4n
be a hypercomplex manifold. We define a splitting of the quaternionic
tangent space H ⊗ TM, which is roughly dual to the splitting H ⊗ T

M

= A ⊕ B of
Section 6.2.
Definition 7.2.1 Let M
4n
be a hypercomplex manifold so that TM

= 2nV
1
as an
Sp(1)-representation. The quaternionic tangent space H ⊗ TM splits according to the
equation
H⊗TM

= V
L
1
⊗V
R
1
⊗2nV
G
1

= 2nV
L
1
⊗(V
RG
2
⊗V
RG
0
).
Define the AH-subbundles
´
A = 2nV
L
1
⊗V
RG
2
and
´
B = 2nV
L
1
⊗V
RG
0
, using the natural
definitions
´
A

=
´
A ∩ (I ⊗ TM) and
´
B

=
´
B ∩ (I ⊗ TM) = ¦0¦. Then
´
A is the
q-holomorphic tangent bundle and
´
B is the q-antiholomorphic tangent bundle of M.
This definition is perfectly natural, though not quite ideal from the point of view of
quaternionic algebra. We would like
´
A and
´
B to be dual to the cotangent spaces A
and B. However, whilst there are H-linear bundle isomorphisms
´
A

= A
×
and
´
B

= B
×
,
these spaces are not isomorphic as AH-bundles, nor is there any fruitful way to alter the
definitions to make them so.
The Lie bracket of vector fields is a bilinear map [ , ] : 1 1 →1, and so defines
a natural linear map λ : 1 ⊗1 →1. As with H-algebras, we will find it much more
meaningful to talk about linear maps on tensor products, and we can use this formulation
to obtain a quaternionic analogue of Equation (7.6).
109
Definition 7.2.2 Let M be a hypercomplex manifold with q-holomorphic tangent
space
´
A ⊂ H ⊗ TM. Let 1 = C

(M, TM) be the vector space of smooth real vector
fields on M.
Define 1
H
= H⊗1 to be the AH-module of smooth quaternion-valued vector fields
on M, so 1
H
= C

(M, H⊗TM). Define 1
A
= C

(M,
´
A) to be the AH-submodule of
H-valued vector fields on M taking values in
´
A ⊂ H ⊗ TM. We will refer to elements
of 1
A
as A-type vector fields.
The relationship between these vector fields and the map λ : 1 ⊗ 1 → 1 is par-
ticularly interesting. Using the canonical isomorphism (H ⊗ TM)⊗
H
(H ⊗ TM)

=
H ⊗ TM ⊗ TM, the Lie bracket defines an AH-morphism (which we shall also call
λ)
λ : 1
H

H
1
H
→1
H
,
which is effectively a Lie bracket operation on quaternionic vector fields.
Here is the quaternionic version of Equation (7.6). The formulation and proof is
similar in spirit to Theorem 6.4.11.
Theorem 7.2.3 Let M be a hypercomplex manifold with q-holomorphic tangent space
´
A, and let 1
A
denote the space of A-type vector fields on M. Then the Lie bracket λ
on quaternionic vector fields preserves 1
A
, so that
λ : 1
A

H
1
A
→1
A
.
Proof. To begin with, we use the section product map φ

A,

A
(Definition 6.4.10) to regard
elements of 1
A

H
1
A
as sections in C

(
´
A⊗
H
´
A) ⊂ C

(H⊗TM ⊗TM).
Let v
j
, w
j
∈ 1 be vector fields such that

j
q
j
⊗v
j
⊗w
j
∈ C

(
´
A⊗
H
´
A), which means
that

q
j
⊗v
j
,

q
j
⊗w
j
∈ 1
A
. We want to find an expression for λ(

q
j
⊗v
j
⊗w
j
) =

q
j
⊗[v
j
, w
j
].
The Obata connection ∇ is an AH-connection on
´
A (and in fact is anti-self-dual, so
that
´
A is a q-holomorphic AH-bundle). Thus ∇(

q
j
⊗w
j
) is an element of C

(
´
A ⊗
T

M). Contracting the T

M-factor with v
j
∈ TM, it follows that

q
j
⊗∇
v
j
w
j
∈ 1
A
,
and similarly

q
j
⊗∇
w
j
v
j
∈ 1
A
.
Since ∇ is torsion-free, the Lie bracket [v
j
, w
j
] is given by the difference ∇
v
j
w
j
−∇
w
j
v
j
.
It follows immediately that
λ
_

q
j
⊗v
j
⊗w
j
_
=

q
j
⊗(∇
v
j
w
j
−∇
w
j
v
j
) ∈ 1
A
,
proving the theorem.
110
The reason why the hypercomplex structure must be integrable to obtain this result
is that the integrability of I, J, and K ensures that the connection ∇ with ∇I =
∇J = ∇K = 0 is torsion-free, and otherwise we would not have [v, w] = ∇
v
w −∇
w
v.
Another way to understand this result is in terms of the decomposition of tensors
with respect to different complex structures. Just as in Theorem 7.1.10, the bundle
´
A
is generated over H by the tensors of type (1, 0) with respect to the different complex
structures, and the bundle
´
A⊗
H
´
A is generated by the tensors of type (2, 0). In other
words, we have
´
A =

Q∈S
2
H ι
q
(T
1,0
Q
M) and
´
A⊗
H
´
A =

Q∈S
2
H ι
q
(T
1,0
Q
M ⊗T
1,0
Q
M).
If every complex structure Q ∈ S
2
is integrable, we have
[1
1,0
Q
, 1
1,0
Q
] ⊆ 1
1,0
Q
for all Q ∈ S
2
, where 1
1,0
Q
denotes the vector fields which are of type (1, 0) with
respect to the complex structure Q. Since the fibres of
´
A are stable, it follows that the
Lie bracket must map sections of
´
A⊗
H
´
A to sections of
´
A.
7.2.2 Quaternionic Lie Algebras
The result of the previous section encourages us to think of the vector fields 1
A
as a
quaternionic Lie algebra with respect to the Lie bracket map λ. We describe this idea as
an abstract algebraic structure, and see how it fits in with some of Joyce’s other algebraic
structures over the quaternions.
As always, we do not talk about bilinear maps, but rather about linear maps on
tensor products. A quaternionic Lie algebra will be an AH-module A together with an
AH-morphism λ : A⊗
H
A →A, whose properties reflect those of a Lie bracket on a real
or complex vector space: namely antisymmetry and the Jacobi identity. Here are the
axioms for a quaternionic Lie algebra:
Axiom QL. (i) A is an AH-module and there is an AH-morphism λ = λ
A
: A⊗
H
A →
A called the Lie bracket.
(ii) S
2
H
A ⊂ ker λ. Thus λ is antisymmetric.
(iii) The composition λ ◦ (id ⊗
H
λ) defines an AH-morphism from
A⊗
H
A⊗
H
A to A such that Λ
3
H
A ⊂ ker(λ ◦ (id ⊗
H
λ)). This is the
Jacobi identity for λ.
Definition 7.2.4 The pair (A, λ) is a quaternionic Lie algebra if it satisfies Axiom QL.
We will often refer to A itself as a quaternionic Lie algebra when the map λ is
understood. Axiom QL(iii) is probably the least familiar-looking of these axioms. By
way of explanation, let (V, λ) be a real or complex Lie algebra. Then in terms of tensor
products the Jacobi identity is
λ ◦ (id ⊗λ)(x ⊗y ⊗z + y ⊗z ⊗x + z ⊗x ⊗y) = 0.
111
Since we can identify x⊗y−y⊗x with x∧y, we see that the Jacobi identity is equivalent
to the condition that λ ◦ (id ⊗λ)(x ∧ y ∧ z) = 0, so that Λ
3
V ⊂ ker(λ ◦ (id ⊗λ)).
Example 7.2.5 Let M be a hypercomplex manifold. The space of quaternion-valued
vector fields 1
H
is a quaternionic Lie algebra with respect to the Lie bracket operator
λ : 1
H

H
1
H
→1
H
. Axiom QL follows from the corresponding identities satisfied by the
Lie bracket on real vector fields.
Theorem 7.2.3 now shows that the A-type vector fields 1
A
form a quaternionic Lie
subalgebra of 1
H
. This is the quaternionic version of the well-known fact that on a
complex manifold, the (1, 0) vector fields form a complex Lie subalgebra of the complex
vector fields.
Joyce has already considered the notion of a Lie algebra over the quaternions, but in a
different fashion. In [J1, ¸6], he writes down axioms called Axiom L and Axiom P, which
define quaternionic analogues of Lie algebras and Poisson algebras: but instead of a Lie
bracket λ : A⊗
H
A →A, Joyce’s quaternionic Lie bracket is a map ξ : A⊗
H
A →A⊗
H
Y ,
where Y

= U
2
is the AH-module of Example 4.1.2.
The reason for this is that Joyce’s main purpose is to describe the algebraic structure
of q-holomorphic functions on hyperk¨ahler manifolds. Since a hyperk¨ahler manifold M
has three independent symplectic forms, two functions f and g have three different
Poisson brackets and the H-algebra T
M
of q-holomorphic functions on M has three
independent Poisson structures. Using the fact that I

= R
3
, Joyce describes the three
Poisson brackets using a single map ξ : T
M

H
T
M
→ T
M
⊗ I. Recalling that (Y

)


=
V
2
= I, we have a map ξ : T
M

H
T
M
→ ι
P
M
(T
M
) ⊗ (Y

)

which is in fact an AH-
morphism whose image is contained in T
M

H
Y . This is why, for Joyce, quaternionic
Lie algebras and Poisson algebras are defined by an AH-morphism ξ : A⊗
H
A →A⊗
H
Y .
We can relate these two algebraic ideas very simply by choosing an AH-morphism
η : Y → H. The space of such AH-morphisms is of course Y


= V
2
. Then for every
AH-module A there is a map id ⊗
H
η : A⊗
H
Y → A⊗
H
H

= A. Suppose that ξ :
A⊗
H
A → A⊗
H
Y satisfies Joyce’s Axiom L. Define a Lie bracket λ : A⊗
H
A → A by
setting λ = (id ⊗
H
η) ◦ ξ. It follows immediately that the pair (A, λ) is a quaternionic
Lie algebra in the sense of Axiom QL.
To go in the opposite direction, suppose that (A, λ) is a quaternionic Lie algebra,
and consider the AH-module A⊗
H
Y . In order to define an HL-algebra we need to form
a map from (A⊗
H
Y )⊗
H
(A⊗
H
Y ) to (A⊗
H
Y )⊗
H
Y . Let τ be the natural isomorphism
τ : A⊗
H
Y ⊗
H
A⊗
H
Y →A⊗
H
A⊗
H
Y ⊗
H
Y which interchanges the second and third factors.
Define an AH-morphism
ξ = (λ⊗
H
id ⊗
H
id) ◦ τ : (A⊗
H
Y )⊗
H
(A⊗
H
Y ) →(A⊗
H
Y )⊗
H
Y.
Then ξ satisfies Joyce’s Axiom L, and the pair (A⊗
H
Y, ξ) forms an HL-algebra. Pre-
cisely which HL-algebras may be obtained from quaternionic Lie algebras and vice versa
using these constructions remains open to question.
7.3 Hypercomplex Lie groups
As a final example, we consider hypercomplex structures on compact Lie groups, and
show how these give rise to finite-dimensional quaternionic Lie algebras. Since the early
112
1950s, the mathematical world has been aware that every compact Lie group of even
dimension is a homogeneous complex manifold.
3
This was first announced by Samelson,
whose proof is an extension of Borel’s celebrated result that the quotient of a compact
Lie group by its maximal torus is a homogeneous complex manifold. In 1988 and 1992
respectively, Spindel et al. [SSTP] and Joyce [J3] demonstrated independently that these
results extend to hypercomplex geometry. Joyce’s approach also gives hypercomplex
structures on more general homogeneous spaces. It relies on being able to decompose
the Lie algebra of a compact Lie group as follows:
Lemma 7.3.1 [J3, Lemma 4.1] Let G be a compact Lie group, with Lie algebra g.
Then g can be decomposed as
g = b ⊕
n

k=1
d
k

n

k=1
f
k
, (7.7)
where b is abelian, d
k
is a subalgebra of g isomorphic to su(2), b+

k
d
k
contains the
Lie algebra of a maximal torus of G, and f
1
, . . . , f
n
are (possibly empty) vector subspaces
of g, such that for each k = 1, 2, . . . , n, f
k
satisfies the following two conditions:
(i) [d
l
, f
k
] = ¦0¦ whenever l < k, and
(ii) f
k
is closed under the Lie bracket with d
k
, and the Lie bracket action of d
k
on
f
k
is isomorphic to the sum of m copies of the basic representation V
1
of su(2) on C
2
,
for some integer m.
By adding an additional p (with 0 ≤ p ≤ Max(3, rk G) ) copies of the abelian Lie
algebra u(1) to b if necessary, this decomposition allows us to define a hypercomplex
structure on pu(1) ⊕g as follows. For a 1-dimensional subspace b
k
of pu(1) ⊕b there
is an isomorphism b
k
⊕d
k

= R ⊕I = H. The action of d
k
on f
k
gives an isomorphism
f
k

= H
m
. This gives an isomorphism pu(1) ⊕ g

= H
l
, in other words a hypercomplex
structure. Normally there are many choices to be made in such an isomorphism, which
give rise to distinct hypercomplex structures. That such a hypercomplex structure on
the vector space pu(1) ⊕g defines an integrable hypercomplex structure on the manifold
U(1)
p
G follows from Samelson’s original work on homogeneous complex manifolds.
This leads to the following result:
Theorem 7.3.2 [J3, Theorem 4.2] Let G be a compact Lie group. Then there exists
an integer p with 0 ≤ p ≤ Max(3, rk G) such that U(1)
p
G admits a left-invariant
homogeneous hypercomplex structure.
There is a strong link between these hypercomplex structures and the quaternionic
Lie algebras of the previous section. Suppose that g is any real Lie algebra. Then
g
H
≡ H ⊗ g is a quaternionic Lie algebra because Axiom QL is obviously satisfied.
The quaternionic Lie algebra structure of g is much more interesting when g admits a
hypercomplex structure as described above. In this case we define the subspace
g
A
= ¦v
0
+ i
1
v
1
+ i
2
v
2
+ i
3
v
3
: v
0
+ I
1
v
1
+ I
2
v
2
+ I
3
v
3
= 0¦ ⊂ g
H
which we shall call the space of A-type elements of g
H
. Lie groups with integrable left-
invariant hypercomplex structures then give rise to interesting quaternionic Lie algebras.
3
Note that not all ‘Lie groups possessing a complex structure’ are complex Lie groups, because their
multiplication and inverse maps might not be holomorphic.
113
Corollary 7.3.3 Let G be a hypercomplex Lie group with hypercomplex structure
(I
1
, I
2
, I
3
), and let g
A
⊂ g
H
be the subspace of A-type elements of g
H
. Then g
A
is
a quaternionic Lie subalgebra of g
H
.
Proof. By Theorem 7.2.3, the set of A-type vector fields on G is closed under the Lie
bracket operator λ. Since λ also preserves the left-invariant vector fields g, it preserves
g
A
.
In this way, Joyce’s hypercomplex structures on compact Lie groups give rise to
many interesting finite-dimensional quaternionic Lie algebras. We can also begin to
calculate the quaternionic cohomology groups of these manifolds. On a Lie group G the
exterior differential d on G-invariant k-forms can be expressed as a formal differential
d : Λ
k
g

→ Λ
k+1
g

. The map d is induced by the dual of the Lie bracket, which is a
linear map λ : g ⊗g →g. Since λ is antisymmetric it is effectively a map λ : Λ
2
g →g,
so its dual is the map d = λ

: g

→Λ
2
g

. In practice, this means that
dω(u, v) = ω([u, v]) ω ∈ g

v, w ∈ g.
The map d extends to a unique antiderivation on Λ

g

in the usual fashion. Questions
about the cohomology of G as a real, complex or hypercomplex manifold can then be
rephrased in terms of the cohomology of the complex (Λ

g

, d). This is true at least for
G-invariant forms, which accounts for all the de Rham cohomology since it is known that
every de Rham cohomology class has a G-invariant representative. This is less obvious
for Dolbeault and quaternionic cohomology groups, whose theory is possibly more subtle
for this reason.
Example 7.3.4 Let M = U(2) be the unitary group in two dimensions, i.e. the
subgroup of GL(2, C) preserving the standard hermitian metric on C
2
. It is well-known
that U(2) is isomorphic to U(1)
Z
2
SU(2), which is diffeomorphic to the Hopf surface
S
1
S
3
.
The Lie algebra u(2) appears naturally in the form of Equation (7.7), thanks to the
decomposition u(2) = u(1) ⊕su(2). Let u(1) = ¸e
0
) and su(2) = ¸e
1
, e
2
, e
3
), so that
[e
0
, e
j
] = 0 and [e
i
, e
j
] = ε
ijk
e
k
where i, j, k ⊂ ¦1, 2, 3¦. Let ¦e
α
¦ be the dual basis for g

. It follows that de
0
=
0 and de
i
= e
j
∧ e
k
, where ¦i, j, k¦ is an even permutation of ¦1, 2, 3¦. Thus e
0
and e
123
both generate de Rham cohomology classes, and we have b
0
(M) = b
1
(M) =
b
3
(M) = b
4
(M) = 1, b
2
(M) = 0. This complex is best described using the structure
of u(2) as a representation τ of su(2) = ¸e
1
, e
2
, e
3
). This subgroup acts on u(2) via
the adjoint representation, so that τ
v
(w) = [v, w] for v ∈ su(2), w ∈ u(2). Since
[v, e
0
] = 0, e
0
generates a copy of the trivial representation V
0
and ¸e
1
, e
2
, e
3
) is just
the adjoint representation of su(2), which is V
2
. Thus T

M

= V
0
⊕V
2
, and this induces
a representation of su(2) on Λ

T

M by the usual Leibniz rule τ
v
(w
1
∧ w
2
) = [v, w
1
] ∧
w
2
+ w
1
∧ [v, w
2
]. This gives the following decompositions:
T

M = ¸e
0
) ⊕¸e
1
, e
2
, e
3
)

= V
0
⊕V
2
Λ
2
T

M = ¸e
23
, e
31
, e
12
) ⊕¸e
01
, e
02
, e
03
)

= V
2
⊕V
2
Λ
3
T

M = ¸e
032
, e
013
, e
021
) ⊕¸e
123
)

= V
2
⊕V
0
Λ
4
T

M = ¸e
0123
)

= V
0
.
(7.8)
114
The benefit of this approach is that the map d is su(2)-equivariant. Thus once we
have shown that de
1
= e
23
, it follows immediately that d gives an su(2)-equivariant
isomorphism
d : ¸e
1
, e
2
, e
3
) →¸e
23
, e
31
, e
12
).
This allows us to read off cohomological information, including the quaternionic co-
homology groups H
k,r
D
(M) and H
k,r
δ
(M). To do this we compare the decomposition of
Equation (7.8) with the decomposition of Λ
k
T

M induced by the hypercomplex struc-
ture. In other words, we have two representations of sp(1) on T

M. The first is the
representation T

M

= 2V
1
defined by the hypercomplex structure, and the second
is the representation T

M

= V
0
⊕ V
2
given by the adjoint action of the subalgebra
su(2) ⊂ u(2). The quaternionic cohomology of M is defined using the first action, but
the d operator is best described using the second. By collating the two we obtain a
complete picture of the situation.
In four dimensions, E
2,2
is the space of self-dual 2-forms ¸e
01+23
, e
02+31
, e
03+12
), where
e
ab+cd
= e
ab
+e
cd
. The image under d of T

M is ¸e
01
, e
02
, e
03
), and the projection π
2,2
is an su(2)-equivariant surjection onto E
2,2
. The sequence
0 −→E
0,0
−→E
1,1
−→E
2,2
−→0
is therefore exact at E
2,2
, and we obtain the standard self-dual cohomology of S
1
S
3
,
H
0,0
D
(M) = R H
1,1
D
(M) = ¸e
0
)

= R H
2,2
D
(M) = 0. (7.9)
Moving to quaternion-valued forms, it is easy to show that non-zero δ-cohomology
occurs only at F
0,0
, F
1,1
, F
3,−1
and F
4,0
, the sequence 0 →B →F
2,0
→F
3,1
→0 being
exact. Explicitly, we have
H
0,0
δ
(M) = H H
1,1
δ
(M) = ¸3e
0
+ i
1
e
1
+ i
2
e
2
+ i
3
e
3
)
H

= U
×
−1
H
4,4
δ
(M) = ¸e
0123
)
H

= H H
3,−1
δ
(M) = ¸e
123
−i
1
e
032
−i
2
e
013
−i
3
e
021
)
H

= U
×
−1
.
(7.10)
It follows from the property of ellipticity that the cohomology sequences should be
exact, as indeed they are. However, they are clearly not AH-exact, because they are
not exact on the primed parts. The quaternionic algebra of such phenonomena might be
interesting and merit closer study.
It would be desirable to extend this work to higher-dimensional groups and homo-
geneous spaces. The group SU(3) provides an interesting case. A similar analysis to
that above may provide the correct results, though because of the additional four dimen-
sions this option is difficult and complicated. The principle would nonetheless be very
much the same, and essentially involves comparing different sp(1)-actions, one defined
by Joyce’s hypercomplex structure and the other arising from the adjoint representation
of a particular subalgebra su(2) ⊂ su(3). For example, it is easy to demonstrate that
H
1,1
δ
(SU(3)) ,= 0. Quaternionic cohomology is in this case certainly not a subset of de
Rham cohomology, since b
1
(SU(3)) = 0.
A logical precursor to developing the quaternionic cohomology theory of hypercom-
plex Lie groups would be to understand thoroughly the Dolbeault cohomology of ho-
mogeneous complex manifolds. Surprisingly, this theory seems to be lacking or at best
115
extremely obscure. The theory for complex groups and homogeneous spaces is docu-
mented in Bott’s important paper [Bot, ¸2], but the link between theory and results in
this paper is quite opaque and certainly contains some mistakes. Pittie’s paper [P] uses
a simpler bicomplex than Bott’s, conjecturing that the cohomology of these complexes
is the same. Certain hypercomplex nilmanifolds discussed by Dotti and Fino [DF] might
also provide fruitful rseults. This would be an interesting area for future research.
116
Bibliography
[AHS] M.F. Atiyah, N.J. Hitchin, I.M. Singer: Self-duality in four-dimensional
Riemannian geometry, Proc R. Soc. Lond. A. 362 (1978), pp. 425-61.
[Bas] R. J. Baston: Quaternionic complexes, J. Geom. Phys. 8 (1992), pp. 29-52.
[Ben] D. Benson: Representations and cohomology I: Basic representation the-
ory of finite groups and associative algebras, Cambridge University Press,
Cambridge (1991).
[Bes] A.L. Besse: Einstein Manifolds, Springer-Verlag (1987).
[Bon] E. Bonan: Sur l’alg`ebre ext´erieure d’une vari´et´e presque hermitienne
quaternionique, C. R. Acad. Sc. Paris 295 (1982), pp. 115-118 and 296
(1983), pp. 601-602.
[Bot] R. Bott: Homogeneous Vector Bundles, Ann. of Math. 66 no. 2 (1957),
pp. 203-248.
[BD] T. Br¨ocker, T. tom Dieck: Representations of Compact Lie Groups,
Springer-Verlag, Graduate Texts in Math. 98 (1985).
[D] C.A. Deavours: The Quaternion Calculus, Amer. Math. Monthly 80
(1973), pp. 995-1008.
[DF] I. G. Dotti, A. Fino: Abelian Hypercomplex 8-dimensional nilmanifolds,
Ann. Global Anal. Geom. 18 (2000), pp. 47-59.
[EKR] Ebbinghaus et al: Numbers, Springer-Verlag, Graduate Texts in Math.
123 (1990) (in particular M. Koecher, R. Remmert: Chapter 7, Hamilton’s
Quaternions).
[F] A. Fujiki: On the de Rham cohomology group of a compact K¨ahler sym-
plectic manifold, Adv. Stud. Pure Math. 10, 1987 (Algebraic Geometry,
Sendai, 1985), pp. 105-165.
[FH] W. Fulton, J. Harris: Representation Theory, Springer-Verlag, Graduate
Texts in Math. 129 (1991).
117
[GH] P. Griffiths, J. Harris: Principles of Algebraic Geometry, John Wiley and
Sons (1978).
[GP] K. Galicki, Y. S. Poon: Duality and Yang-Mills fields on quaternionic
K¨ahler manifolds, J. Math. Phys. 32 (1991), pp. 1263-1268.
[H1] Sir William Rowan Hamilton: Letter to the Rev. Archibald Hamilton,
August 5th 1865. Hamilton’s Math. Papers III p. 15.
[H2] Sir William Rowan Hamilton: Letter to John T.Graves, October 17th,
1843. London, Edinburgh and Dublin Philosophical Magazine and Journal
of Science 25 (1844), pp. 189-95.
[H3] Sir William Rowan Hamilton: On a new Species of Imaginary Quantities
connected with a theory of Quaternions, Proc. Royal Irish Acad. 2 (1843),
pp. 424-434.
[H4] Sir William Rowan Hamilton: On Quaternions, Proc. Royal Irish Acad.
3 (1847), pp. 1-16.
[HKLR] N.J. Hitchin, A. Karlhede, U. Lindstr¨om, M. Roˇcek: Hyperk¨ahler metrics
and supersymmetry, Commun. Math. Phys. 108 (1987), pp. 535-589.
[J1] D. Joyce: Hypercomplex Algebraic Geometry, Quart. J. Math. Oxford, 49
(1998), pp. 129-162.
[J2] D. Joyce: The Hypercomplex quotient and the Quaternionic quotient,
Math. Ann. 290 (1991), pp. 323-340.
[J3] D. Joyce: Compact Hypercomplex and Quaternionic manifolds, J. Diff.
Geom. 35 (1992), pp. 743-761.
[K] S. Kobayashi: Differential Geometry of Complex Vector Bundles, Prince-
ton University Press (1987).
[Kr] V. Y. Kraines: Topology of Quaternionic Manifolds, Trans. Amer. Math.
Soc. 122 (1966) pp. 357-367.
[MS] M. Mamone Capria, S.M. Salamon: Yang-Mills fields on quaternionic
spaces, Nonlinearity 1 (1988), pp. 517-530.
[N] T. Nitta: Vector bundles over Quaternionic K¨ahler manifolds, Tohoku
Math. J. 40 (1988), pp. 425-440.
[P] H. V. Pittie: The nondegeneration of the Hodge–de Rham spectral se-
quence, Bull. AMS 10 (1989), pp. 19-22.
[Q] D. Quillen: Quaternionic Algebra and Sheaves on the Riemann Sphere,
Quart. J. Math. Oxford 49 (1998), pp. 163-198.
118
[R] R. Reyes-Carri´on: A generalisation of the notion of instanton, Diff. Geom.
and its Apps. 8 (1998), pp. 1-20.
[S1] S. Salamon: Topics in four-dimensional Riemannian geometry, Geome-
try seminar “Luigi Bianchi”, Springer-Verlag lecture notes in math. 1022
(1982).
[S2] S. Salamon: Quaternionic K¨ahler Manifolds, Invent. Math. 67 (1982), pp.
143-171.
[S3] S. Salamon: Differential Geometry of Quaternionic Manifolds, Ann. Sci-
ent.
´
Ec. Norm. Sup. Paris. series 4 no. 19 (1986), pp. 31-55.
[S4] S. Salamon: Riemannian geometry and holonomy groups, Pitman Res.
Notes in Math. 201, Longman, Harlow (1989).
[SSTP] Ph. Spindel, A. Sevrin, W. Troost, A. van Proeyen: Extended super-
symmetric σ-models on group manifolds, Nuclear Phys. B 308 (1988), pp.
662-698.
[dL] S. de Leo: Quaternions and Special Relativity, J. Math. Phys. 37.3 (1996),
pp. 2955-2968.
[Su] A. Sudbery: Quaternionic Analysis, Math. Proc. Camb. Phil. Soc. 85
(1979), pp. 199-225.
[Sw] A. Swann: Quaternion K¨ahler Geometry and the Fundamental 4-form,
Proc. Curvature Geom. workshop, University of Lancaster (1989), pp.
165-173.
[Sy] J.L. Synge: Quaternions, Lorentz Transformations and the Conway-
Dirac-Eddington Matrices, Comm. Dublin Inst. for advanced studies, se-
ries A no. 21 (1972).
[V] M. Verbitsky: Hyperholomorphic bundles over a Hyperk¨ahler manifold, J.
Alg. Geom. 5 (1996), pp. 633-669.
[v.d.W] B.L. van der Waerden: A History of Algebra, Springer-Verlag (1985).
[WW] R.S. Ward, R.O. Wells: Twistor Geometry and Field Theory, Cambridge
Monographs on Math. Phys. (1990).
[W] R.O Wells: Differential Analysis on Complex Manifolds, Springer-Verlag,
Graduate Texts in Math. 65 (1980).
119

Abstract

QUATERNION ALGEBRAIC GEOMETRY DOMINIC WIDDOWS
St Anne’s College, Oxford Thesis submitted Hilary Term, 2000, in support of application to supplicate for the degree of D.Phil.

This thesis is a collection of results about hypercomplex and quaternionic manifolds, focussing on two main areas. These are exterior forms and double complexes, and the ‘algebraic geometry’ of hypercomplex manifolds. The latter area is strongly influenced by techniques from quaternionic algebra. A new double complex on quaternionic manifolds is presented, a quaternionic version of the Dolbeault complex on a complex manifold. It arises from the decomposition of real-valued exterior forms on a quaternionic manifold M into irreducible representations of Sp(1). This decomposition gives a double complex of differential forms and operators as a result of the Clebsch-Gordon formula Vr ⊗V1 ∼ Vr+1 ⊕Vr−1 for Sp(1)-representations. = The properties of the double complex are investigated, and it is established that it is elliptic in most places. Joyce has created a new theory of quaternionic algebra [J1] by defining a quaternionic tensor product for AH-modules (H-modules equipped with a special real subspace). The theory can be described using sheaves over CP 1 , an interpretation due to Quillen [Q]. AH-modules and their quaternionic tensor products are classified. Stable AH-modules are described using Sp(1)-representations. This theory is especially useful for describing hypercomplex manifolds and forming close analogies with complex geometry. Joyce has defined and investigated q-holomorphic functions on hypercomplex manifolds. There is also a q-holomorphic cotangent space which again arises as a result of the Clebsch-Gordon formula. AH-module bundles are defined and their q-holomorphic sections explored. Quaternion-valued differential forms on hypercomplex manifolds are of special interest. Their decomposition is finer than that of real forms, giving a second double complex with special advantages. The cohomology of these complexes leads to new invariants of compact quaternionic and hypercomplex manifolds. Quaternion-valued vector fields are also studied, and lead to the definition of quaternionic Lie algebras. The investigation of finite-dimensional quaternionic Lie algebras allows the calculation of some simple quaternionic cohomology groups.

Acknowledgements
I would like to express gratitude, appreciation and respect for my supervisor Dominic Joyce. Without his inspiration as a mathematician this thesis would not have been conceived; without his patience and friendship it would certainly not have been completed.

Dedication
To the memory of Dr Peter Rowe, 1938-1998, who taught me relativity as an undergraduate in Durham. His determination to teach concepts before examination techniques introduced me to differential geometry.

i

. . . . . . . . . . 6. . . . . . . . . . . . . . .1 Complex. .2 Sp(1)-Representations and the Quaternionic Tensor Product 5. . . . . . . . . .1 Stable AH-modules and Sp(1)-representations . . . . . . . . . . . . . . 3. Hypercomplex and Quaternionic Manifolds . . . . . . . . . . . . . . . 4. . . . 4. . . . . . .3 Ellipticity and the Double Complex .Contents Introduction 1 The 1. .4 Examples and Summary of AH-modules . .2 1. 3 A Double Complex on Quaternionic Manifolds 3. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4. . . . .4 K-modules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Lie Group Sp(1) and its Representations . . . . . . . . . . . . . . . . . . . . . . . . . 3. . . .1 The Quaternionic Algebra of Joyce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .4 Quaternion-valued forms on Hypercomplex Manifolds 4 Developments in Quaternionic Algebra 4.2 Duality in Quaternionic Algebra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5. .1 Real forms on Complex Manifolds . . . . . . . . . . . . . Difficulties with the Quaternions . . . . . . . . . . . . . . . . . . . . 6. . . . . 6 Hypercomplex Manifolds 6. . . .3 Quaternions and the Group Sp(1) The Quaternions . . . . . . . . . . . . . . . . . .2 Differential Forms on Complex Manifolds . . .2 Construction of the Double Complex . . . . . . . . . . . . . . . . . . .3 Differential Forms on Quaternionic Manifolds . . .4 Q-holomorphic AH-bundles . . . . . . . . . . . . . 5 Quaternionic Algebra and Sp(1)-representations 5. .5 The Sheaf-Theoretic approach of Quillen . . . . . . . . . . . . . . . . . . . . . 2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 4 4 9 12 17 17 20 23 25 25 28 32 40 42 42 48 52 53 56 61 61 68 74 77 80 81 86 90 93 2 Quaternionic Differential Geometry 2. . . .3 Real Subspaces of Complex Vector Spaces 4.2 The Quaternionic Cotangent Space . 5. . . . . . . . .3 AH-bundles . . . . . . . . . . . . . . .1 Q-holomorphic Functions and H-algebras 6. . . . . . . . . . ii .3 Semistable AH-modules and Sp(1)-representations . . . . . . . . . . . . . .1 1. . . . . . . . . 3. . . 2. .

7. . . . . . . . . . . . . . . . . . References 100 101 109 112 117 iii . . . .2 Vector Fields and Quaternionic Lie Algebras . . . . . . . . . . . . . . . . . . 7. . . . . . . . . . . . . . .3 Hypercomplex Lie groups .7 Quaternion Valued Forms and Vector Fields 7. . . .1 The Quaternion-valued Double Complex . . . .

Chapter 2 is about quaternionic structures in differential geometry. The approach is based upon two pillars. Our contribution to the theory of quaternionic manifolds relies on decomposing the Sp(1)-action on exterior forms. H)-structure) are defined. as is all the work on q-holomorphic functions and forms on hypercomplex manifolds. The decomposition of exterior forms on quaternionic manifolds is precisely such an example. A recurrent theme throughout will be representations of the group Sp(1) of unit quaternions. Another frequent source of motivation is the behaviour of the complex numbers. The irreducible representations on complex vector spaces and their tensor products are described. Contributions are made to both fields of study. This situation has left various aspects of quaternionic behaviour unexplored. hypercomplex manifolds (those possessing a torsion-free GL(n. replacing this with the group Sp(1) can lead directly to quaternionic versions. The approach is based on the work of Salamon [S3]. as are real and quaternionic representations. H)-structure). the first chapter is devoted to a survey of the history of the quaternions and their applications.Introduction This aim of this thesis is to describe and develop various aspects of quaternionic algebra and geometry. I have often felt both surprised and privileged that it has not been carried out before. Aspects of complex geometry can often be described using the group U(1) of unit complex numbers. The importance of Sp(1)-representations to both areas is chiefly responsible for the successful synthesis of methods in the work on hypercomplex manifolds. After reviewing the Dolbeault complex. followed by the broader class of quaternionic manifolds (those possessing a torsion-free Sp(1)GL(n. Background material also includes an introduction to the group Sp(1) and its representations. Much of the original work presented is enticingly simple — indeed. On the other hand. we consider the decomposition of differential forms on quaternionic manifolds. Taking complex manifolds as a model. whilst the main new insight in quaternionic algebra is that the most important building blocks of Joyce’s theory are best described and manipulated as Sp(1)-representations. including an important elliptic complex discovered by Salamon upon which the integrability of the quaternionic 1 . To help understand the reasons for this omission and the consequent opportunities for development. namely the differential geometry of quaternionic manifolds and Joyce’s recent theory of quaternionic algebra. Many situations in complex algebra and geometry have interesting quaternionic analogues. enabling these strands to be woven together in describing the algebraic geometry of hypercomplex manifolds. Joyce’s quaternionic algebra is such a rich theory precisely because real subspaces of quaternionic vector spaces behave so differently from real subspaces of complex vector spaces. One of the main explanations for this is the relative unpopularity suffered by the quaternions in the 20th century.

These spaces are examples of AH-module bundles or AH-bundles. we define a natural splitting of the quaternionic cotangent space H ⊗ T ∗ M ∼ A ⊕ B. further refining the double = complex. which we discuss. the theory of quaternionic algebra. There is a natural product map on the AH-module of q-holomorphic functions. consisting of the basic representations V1 and the trivial representations V0 . In the fourth chapter (which is partly a summary of the work of Joyce [J1] and Quillen [Q]) we move to our other major area of interest. which are seen as the quaternionic analogue of holomorphic functions. such methods are applied to quaternion-valued tensors on hypercomplex manifolds. The new double complex is shown to be elliptic = = everywhere except along its bottom row. Joyce has already used such an approach to define and investigate q-holomorphic functions on hypercomplex manifolds. The top row of the quaternion-valued double complex is particularly well-adapted to quaternionic algebra. That this decomposition gives rise to a double complex results from the Clebsch-Gordon formula Vr ⊗T ∗ M ∼ Vr ⊗2nV1 ∼ 2n(Vr+1 ⊕Vr−1 ). There are q-holomorphic AH-bundles with q-holomorphic sections. In Chapter 5 it is shown that all stable AH-modules and their duals are conveniently described using Sp(1)-representations. Using the Sp(1)-version of quaternionic algebra. and show that q-holomorphic func= tions are precisely those whose differentials take values in A ⊂ H ⊗ T ∗ M .structure depends. Dual AH-modules are defined and shown to have interesting properties. where = V1 is the basic representation of Sp(1) on C2 . Decomposition of the induced Sp(1)representation on Λk T ∗ M is a simple process achieved by considering weights. The building blocks of this theory are H-modules equipped with a special real subspace. the cotangent space of a quaternionic manifold M 4n takes the form T ∗ M ∼ 2nV1 . Several parallels with complex geometry arise. Such an object is called an AH-module. The quaternion-valued double complex has advantages over the real-valued version. Joyce has discovered a canonical tensor product operation for AH-modules which is both associative and commutative. Using ideas from Quillen’s work. As an Sp(1)-representation. This identification enables tensors on hypercomplex manifolds to be treated using the techniques of quaternionic algebra. This double complex presents us with new quaternionic cohomology groups. which presents close parallels with the Dolbeault complex and motivates the definition of q-holomorphic k-forms. which gives the q-holomorphic functions an algebraic structure which Joyce calls an H-algebra. The resulting theory is ideally adapted for describing hypercomplex geometry. The bundle A is hence defined to be the q-holomorphic cotangent space of M . This complex is in fact the top row of a hitherto undiscovered double complex on quaternionic manifolds. Q-holomorphic sections are described using the quaternionic tensor product and the q-holomorphic cotangent space. A hypercomplex manifold M has a triple of global anticommuting complex structures which can be identified with the imaginary quaternions. 2 . In the final chapter. being elliptic in more places. which is the subject of Chapter 3. and seen to form an H-algebra module over the q-holomorphic functions. we classify AH-modules up to isomorphism. The global complex structures give an extra decomposition which generalises the splitting H ⊗ T ∗ M ∼ A ⊕ B. Particularly wellbehaved is the category of stable AH-modules. a process begun in Chapter 6. The double complex of Chapter 3 is revisited and adapted to quaternion-valued differential forms.

Quaternion-valued vector fields are also interesting. Interesting finite-dimensional quaternionic Lie algebras are used to calculate some quaternionic cohomology groups on Lie groups with left-invariant hypercomplex structures. the (1. The quaternionic tangent space splits as H ⊗ T M ∼ A ⊕ B in the same way as the cotangent space. This is the quaternionic analogue of the statement that on a complex manifold. The vector fields in question therefore form a quaternionic Lie algebra. 0) vector fields are closed under the Lie bracket. Vector fields taking = values in A are closed under the quaternionic tensor product and Lie bracket. a new concept which we introduce. a result which depends upon the integrability of the hypercomplex structure. 3 .

q = Re(q) − Im(q). ¯ • We regard the real numbers R as a subfield of H. • Define the conjugate q of q = q0 + q1 i1 + q2 i2 + q3 i3 by q = q0 − q1 i1 − q2 i2 − q3 i3 . Then q 2 = −1 if and only if q1 + q2 + q3 = 1. The symbol I is not standard. The quaternions are a division algebra (an algebra with the property that ab = 0 implies that a = 0 or b = 0 ). . and the quaternions as a direct sum H ∼ R ⊕ I. As with complex numbers. r3 ∈ R}. so H = {r0 +r1 i1 +r2 i2 +r3 i3 : r0 . so the set of ‘quaternionic square-roots of minus-one’ is naturally isomorphic to the 2-sphere S 2 . We shall call this subfield Cq . i2 = i2 = i2 = −1 1 2 3 (1.1) and the distributive law. 4 . The quaternion algebra is not commutative.Chapter 1 The Quaternions and the Group Sp(1) 1. i2 and i3 . We shall often identify these sets. i2 . • Define the imaginary quaternions I = i1 . • If q ∈ S 2 then 1. i2 i3 = −i3 i2 = i1 . though it does obey the associative law. • Define the real and imaginary parts of q by Re(q) = q0 ∈ R and Im(q) = q1 i1 + q2 i2 + q3 i3 ∈ I. i3 i1 = −i1 i3 = i2 . = 2 2 2 • Let q = q1 i1 + q2 i2 + q3 i3 ∈ I. writing ‘q ∈ S 2 ’ as a shorthand for ‘q ∈ H : q 2 = −1’. Quaternions are added together component by component. but we will use it throughout. q is a subfield of H isomorphic to C. . . and quaternion multiplication is given by the quaternion relations i1 i2 = −i2 i1 = i3 . q ∈ H. ¯ ¯ ¯¯ Then (pq) = q p for all p. .1 The Quaternions The quaternions H are a four-dimensional real algebra generated by the identity element 1 and the symbols i1 . i3 .

or light. which contains the solution of the problem. full grown. 1843.maths. He finally realised that the secret was to introduce a fourth dimension. he struggled with this problem. Substituting i1 . on my coming down to breakfast. the fundamental formula with the symbols i.” 3 “Less than an hour elapsed” before Hamilton obtained leave of the Council of the Royal Irish Academy to read a paper on quaternions. To Hamilton. and a spark flashed forth. At the time. “started into life.G. and only after years of development can they be presented in a lecture course as a definitive set of axioms and results. 2 Copies of Hamilton’s most significant letters and papers concerning quaternions are currently available on the internet at www. your brother and yourself used to ask me: “Well. For years. j. as an inscription.. can you multiply triplets?” Whereto I was always obliged to reply.1.1). on the other hand. The quaternions. “No. 5 . I pulled out on the spot a pocket-book. Rarely do we possess such a clear account of the genesis of a piece of mathematics. i2 . though neither of them published the disovery [EKR..tcd. i3 for i. whilst walking with his wife.ie/pub/HistMath/People/Hamilton 3 Letter to Professor P. I can only add and subtract them”. on the 16th of Ocober. Hamilton tried to manipulate the three symbols 1.. has long since mouldered away. an excerpt of which can be found on the same website as [H1].1. complex numbers were being applied very effectively to problems in the plane R2 . For several years. as we passed it. this gives the quaternion relations (1.1 A History of the Quaternions The quaternions were discovered by the Irish mathematician and physicist. In the same letter. k: i2 = j 2 = k 2 = ijk = −1. The next day. 2 dated shortly before his death in 1865. Hamilton wrote a detailed letter to his friend and fellow mathematician John T. p.. which still exists. Nor could I resist the impulse — unphilosophical as it may have been — to cut with a knife on the stone of Brougham Bridge. with a sad shake of the head. In a touching letter to his son [H1]. Most mathematical theories are invented gradually. k. the next logical step was to seek a similar 3-dimensional number system which would revolutionise calculations in R3 . Hamilton writes: Every morning. William Rowan Hamilton (1805-1865). but of course. 192]. and made an entry there and then.1). governed by certain rules. Graves 1 Letters suggest that both Euler and Gauss were aware of the quaternion relations (1. i and j into an algebra. 1 whose contributions to mechanics are well-known and widely used. he had a flash of inspiration. he writes: An electric circuit seemed to close. Papa. By 1835 Hamilton had helped to win acceptance for the system of complex numbers by showing that calculations with complex numbers are equivalent to calculations with ordered pairs of real numbers. j. the herald (as I foresaw immediately) of many long years to come of definitely directed thought and work . 1843. On 16th October.Tait.

Most recently.) As a result. the real numbers R. However. The timing of the discovery amplified its impact upon Hamilton and his followers. and though beautifully ingenious. once remarked that “Quaternions came from Hamilton after all his really good work had been done. quaternions have enjoyed prominence in computer science. C). while Eamon de Valera was President of Ireland (from 1959 to 1973). especially as it became clear that the quaternions are just one example of a number of possible algebras. the complex numbers C and the quaternions H are the only associative division algebras. 1. Hamilton became the figurehead of a school of ‘quaternionists’. he would attend mathematical meetings whenever their title contained the word ‘quaternions’ ! Such excesses were bound to provoke a reaction. whose fervour for the new numbers far exceeded their usefulness. and those who become familiar with them soon come to appreciate an intricacy and beauty which is all their own.193].2 Quaternions and Matrices In this section we will make use of the older notation i = i1 . and showed that the quaternion algebra could be realised as a subalgebra of the complex-valued 2 × 2 matrices. (It was not until 1858 that Cayley introduced matrices. Investigation into the nature of and constraints imposed by algebraic properties such as associativity and commutativity was greatly accelerated by the discovery of the quaternions. Indeed. It is well-known that the quaternions can be written as real or complex matrices.1. The discovery was published within a month on the 13th of November [H3].[H2]. quaternions do shine a light on certain areas of mathematics. Lord Kelvin. The quaternions themselves have been used in various areas of mathematics. j = i2 .5] to describe quaternionic behaviour. because there are isomorphisms from H into subalgebras of Mat(4. have been an unmixed evil to those who have touched them in any way” [EKR. p. the numbers have fallen short of the early expectations of the quaternionists. The discovery of the quaternions provided enormous stimulation to algebraic research and it is thought that the term ‘associative’ was coined by Hamilton himself [H3. The only other algebras known in 1843 were the real and complex numbers. k = i3 . giving us a clear account of the train of research which led him to his breakthrough. R) and Mat(2. Certainly. The quaternions remain the simplest algebra after the real and complex numbers. and devoted the rest of his career exclusively to its study. because they are the simplest algebraic tool for describing rotations in three and four dimensions. This makes it easy to interpret i as the standard complex ‘square root of −1’ and j as a ‘structure map’ on the complex vector space C2 . the famous Scottish physicist. Hamilton believed his discovery to be of similar importance to that of the infinitesimal calculus. both of which can be regarded as subalgebras of the quaternions. p. as was proved by Georg Frobenius in 1878: and amongst these the quaternions are the most general. This is far from the case. The former of these is given by the mapping 6 . A belief that quaternions are somehow obsolete is often tacitly accepted to this day. for example. Echos of this zeal could still be heard this century.

it is in this context that the term ‘vector’ first appears [H4]). (1. In many cases. C) α + βj → α β ¯ ¯ −β α . q0 q1 q2 q3  −q1 q0 −q3 q2  . The isomorphism ι gives an easy way to deduce that H is an associative division algebra.4) Note that the squared norm q q of a quaternion q is the same as the determinant ¯ of the matrix ι(q) ∈ H. we obtain a quaternionic version of the scalar product and vector product on R3 . This identification H = uniquely determined: each q ∈ S 2 determines a similar isomorphism. a professor at Yale University. C). the inverse of any nonzero matrix A ∈ H is also in H. it was not until the 1880’s that Josiah Willard Gibbs (1839-1903). a quaternionic description prefigures more modern descriptions. (1. = (1. An excellent and readable account of most of the following can be found in Chapter 7 of [EKR]. Having written this down. q0 + q1 i + q2 j + q3 k →   −q2 q3 q0 −q1  −q3 −q2 q1 q0 More commonly used is the mapping into Mat(2. argued that the scalar and vector products 7 . q2 . we can consider each quaternion q = q0 +q1 i1 +q2 i2 +q3 i3 as the sum of a scalar part q0 and a vectorial part (q1 . 1. it is easy to form the map ι : H → H ⊂ Mat(2.1. C). We will outline two main areas – vector analysis and Euclidean geometry.3) The quaternion algebra can thus be realised as a real subalgebra of Mat(2. q ∈ I.5) Surprising as it seems nowadays. q3 ) ∈ R3 (indeed.3 Simple Applications of the Quaternions There are a number of ways in which quaternions can be used to express mathematical ideas. If we multiply together two imaginary quaternions p. using the equation q0 + q1 i + q2 j + q3 k = (q0 + q1 i) + (q2 + q3 i)j. as follows: pq = −p · q + p ∧ q ∈ R ⊕ I ∼ H. If we identify the imaginary quaternions I with the real vector space R3 . and the only matrix in H whose determinant is zero is the zero matrix. using the identifications 1 0 0 1 i 0 0 −i 0 1 −1 0 0 i i 0 1= i1 = i2 = i3 = . The map j : α+βj → −β + αj = 2 2 ∼ C2 is not is a conjugate-linear involution of C with j = −1.2)  ¯ ¯ In this way we obtain the expression q = α+βj ∈ H ∼ C2 . (1. Every quaternion can be uniquely written as the sum of its real and imaginary parts. We can write every quaternion as a pair of complex numbers.

q ∈ H. Another crucial part of vector analysis which originated with Hamilton and the quaternions is the ‘nabla’ operator (so-called by Hamilton because the symbol is similar in shape to the Hebrew musical instrument of that name). q = Re(p¯). x3 ) = f1 (x1 .7). i. we can obtain the Euclidean metric on R4 in a similar fashion by relaxing the restriction in Equation (1.e. which we recognise in modern terminology as F = − div F + curl F. x2 . x3 )i1 + f2 (x1 . putting √ ¯ |q| = q q . Another quaternionic formula. q and so we define the canonical scalar product on H by p. (1.6) f := ∂x1 ∂x2 ∂x3 Hamilton went on to consider applying the operator to a ‘differentiable quaternion field’ F (x1 .5) that p and q should be imaginary. similar to Equation (1. (1. q ∈ H. obtaining the equation F =− ∂f1 ∂f2 ∂f3 + + ∂x1 ∂x2 ∂x3 + ∂f3 ∂f2 − ∂x2 ∂x3 i1 + ∂f1 ∂f3 − ∂x3 ∂x1 i2 + ∂f2 ∂f1 − ∂x1 ∂x2 i3 .7) (1. As with complex numbers. x2 . x3 ) : R3 → R was first written as ∂f ∂f ∂f i1 + i2 + i3 . is the identity Re(pq) = p0 q0 − p1 q1 − p2 q2 − p3 q3 ∈ R. The familiar gradient operator acting on a real differentiable function f (x1 . For any p. It was he who introduced the familiar notation p · q and p ∧ q — before his time these were written ‘−Spq’ and ‘V pq’. |pq| = |p||q| for p. x2 .8) The norm function is multiplicative.6) leads to the well-known Laplacian operator 2 f =− ∂2f ∂2f ∂2f + 2+ 2 ∂x2 ∂x2 ∂x3 1 =: ∆f. x2 . we can combine the norm function with conjugation to obtain the inverse of q ∈ H \ {0} (it is easily verified that q −1 = q /|q|2 is the unique quaternion such that ¯ qq −1 = q −1 q = 1). x2 .9) 8 . x3 )i2 + f3 (x1 . q We define the norm of a quaternion q ∈ H in the obvious way.had their own meaning. x3 )i3 . Having obtained the standard scalar product on R3 . independent from their quaternionic origins. Applying the on R3 : operator to Equation (1. indicating the ‘scalar’ and ‘vector’ parts of the quaternionic product pq ∈ H. (1. we have Re(p¯) = p0 q0 + p1 q1 + p2 q2 + p3 q3 ∈ R.

Cayley showed that all rotations of R4 can be written in this fashion and that the two possibilities given above are the only two which give the same rotation.b (q)| = |q| and the function fa.b : R4 → R4 is a rotation.4): I= i 0 0 −i J= 0 1 −1 0 K= 0 i i 0 . The Lie algebra sp(1) of Sp(1) is generated by the elements I. (1. There are interesting questions which arise.) This fixes the real line R and rotates the imaginary quaternions I. We can write these using the matrices of Equation (1. J] = 2K. C) of Equation (1. X] = 2X. C).4) maps Sp(1) isomorphically onto SU(2).2 The Lie Group Sp(1) and its Representations The unit quaternions form a subgroup of H under multiplication. Why do the unit quaternions turn out to be so important? In view of the quaternionic version of the Lorentz metric in Equation (1. we obtain the rotations of R3 simply by putting a = b−1 . giving the inner automorphism of H. ¯ all reflections can be obtained by the involutions fa.If we regard p and q as four-vectors in spacetime with the ‘time-axis’ identified with R ⊂ H and the ‘spatial part’ identified with I. According to Cayley. and consider the involution fa. Also. b gives the same rotation. q → aqa−1 .11) . which the pioneering work of Hamilton and Cayley helped to develop. As a manifold Sp(1) is the 3-sphere S 3 . Clearly. which we call Sp(1). The isomorphism ι : H → H ⊂ Mat(2. this is identical to the Lorentz metric of special relativity. f−a. Identifying I with R3 .b (q) = aqb.−b = fa. all real-algebra automorphisms of H take this form. [J. the map q → aqa−1 is a rotation of R3 .b (q) = a¯b. q One of the beauties of this system is that having obtained all rotations of R4 . Y ] = H. (By Cayley’s theorem. so Sp(1) is a compact Lie group. In 1855. J and K.9). Its importance to the quaternions is equivalent to that of the circle group U(1) of unit complex numbers in complex analysis. [H. These discoveries provided much insight into the classical groups SO(3) and SO(4). and helped to develop our knowledge of transformation groups in general. Then |fa. The multiplication and inverse maps are smooth. This is generated over C by the elements 0 0 1 0 0 1 H= X= Y = . within a year of the discovery of the quaternions Hamilton was aware that all rotations of R3 can be expressed in this fashion. (1.10) 0 −1 0 0 1 0 and the relations [H.b . 1. K] = 2I and [K. Y ] = −2Y 9 and [X. so each of these choices for the pair a. Let a and b be quaternions of unit norm. the Lie bracket being given by the relations [I.b : H → H given by fa. The complexification of sp(1) is the Lie algebra sl(2. can we use quaternions to write Lorentz tranformations as elegantly? Are there other spaces which might lend themselves to quaternionic treatment? These questions are best addressed using the theory of Lie groups. I] = 2J.

= For any Q ∈ sp(1).13) The horizontal arrows are inclusions. this does not restrict to a suitably interesting real homomorphism Sp(1)→ SO0 (3. and the vertical arrows are 2 : 1 coverings with kernels {1. 1.5] (which describes the action of the group SU(2)) and [FH. This is because there is a commutative diagram of Lie group homomorphisms Sp(1) ∼ Spin(3) → Sp(1) × Sp(1) ∼ Spin(4) = = ↓ SO(3) → ↓ SO(4). The differential dρ is a Lie algebra representation dρ : g → End(V ). The applications of these homomorphisms in Riemannian geometry are described by Salamon in [S1].and four-dimensional Euclidean geometry fit so well in a quaternionic framework. From this. Good introductions to Sp(1)representations include [BD.2. (1. 1). 1).1 The Representations of Sp(1) A representation of a Lie group G on a vector space V is a Lie group homomorphism ρ : G → GL(V ). C) → SO0 (3. every representation of Sp(1) on a complex vector space V can be written as a direct sum of irreducible representations.C) are a more familiar and fertile ground. −1} and {(1. but the mental gymnastics involved in using four ‘square roots of −1’ are cumbersome: the equivalent description of ‘spin transformations’ using the matrices of the group SL(2. We recall those points which will be of particular importance. which is a Cartan subalgebra of sp(1). Lecture 11] (which provides a description in terms of the action of the Lie algebra sl(2. C).12) give one possible identification sp(1) ⊗R C ∼ sl(2. There have been attempts to use the biquaternions {p + iq : p + q ∈ H} ∼ H ⊗R C to apply quaternions to special relativity = [Sy]. Identifying a unit vector Q = aI +bJ +cK ∈ sp(1) with the corresponding imaginary quaternion q = ai2 + bi2 + ci3 ∈ S 2 .1). the normaliser N (Q) of Q is defined to be N (Q) = {P ∈ sp(1) : [P. Because it is a compact group. We can also see why the same is not true for Lorentzian geometry. In the previous section it was shown that rotations in three and four dimensions can be written in terms of unit quaternions. (−1. Whilst there is a double cover SL(2. but the author has found no way which is simple enough to be really effective. C)).Hence the equations I = iH. −1)} respectively. Here the important group is the Lorentz group SO(3. We will sometimes refer to V itself as a representation where the map ρ is understood. It is possible to write Lorentz transformations using quaternions (for a modern example see [dL]). = Their theory is well-known and used in many situations. Q] = 0} = Q . we can see clearly why three. The representations of Sp(1) ∼ SU(2) will play an important part in this thesis. which we will call U(1)q .1). J =X −Y and K = i(X + Y ) (1. The multiplicity 10 . §2. the exponential map exp : sp(1) → Sp(1) maps Q to the unit circle in Cq .

xyn−1 . Let V1 be the basic representation of SL(2.12). It is easy to show that for any xn−k yk ∈ Vn .1]. so we define Vn = S n (V1 ). xn−2 y2 . Let x and y be a basis for C2 . . This coincides with the basic representation of Sp(1) by left-multiplication on H ∼ C2 .14) A(a · b) = To obtain the induced action of sl(2. which describes the action of Lie groups and Lie algebras on tensor products. C(xn−k yk ) = −n(n + 2)xn−k yk .C) on Vn is given by the induced action on the space of homogeneous polynomials of degree n in the variables x and y. C) and sp(1). x2 yn−2 . the sl(2. . The representation Vn is irreducible [BD. and the weights are the integers {n. (1. Moreover. . . y . J and K for H.16) 4 This can be found in [FH. n − 2k. The action of SL(2. −n}.C) on C2 given by left-action of matrices upon column vectors. .of each irreducible in such a decomposition is uniquely determined. yn . Thus Vn is also characterised by being the unique irreducible representation of sl(2.3]. and every irreducible representation of Sp(1) is of the form Vn for some nonnegative n ∈ Z [BD. C) with highest weight n. n − 2. X and Y . C) on Vn we use the Leibniz rule A(a) · b + a · A(b). so that V1 = x. 2 − n. Proposition 5. Another important operator which acts on an sp(1)-representation is the Casimir operator C = I 2 + J 2 + K 2 = −(H 2 + 2XY + 2Y X). .15) Each subspace xn−k yk ⊂ Vn is therefore a weight space of the representation Vn . 110]. . 11 . . . We can compute the action of sp(1) on Vn by substituting I. C)-action on V1 is given by H(x) = x H(y) = −y X(x) = 0 X(y) = x Y (x) = y Y (y) = 0. Proposition 5. . there is a unique irreducible representation on Cn for every n > 0. In terms of H. The unique irreducible = n+1 th representation on C is then given by the n symmetric power of V1 . Then Vn = S n (V1 ) = xn . Another very important way to describe the structure of these representations is obtained by decomposing them into weight spaces (eigenspaces for the action of a Cartan subalgebra). X and Y using the relations of (1. xn−1 y. This makes the representations of Sp(1) particularly easy to describe. This gives H(xn−k yk ) = (n − 2k)(xn−k yk ) X(xn−k yk ) = k(xn−k+1 yk−1 ) Y (xn−k yk ) = (n − k)(xn−k−1 yk+1 ). 4 (1. Each of the Lie group representations Vn is a representation of the Lie algebras sl(2. (1. . p. .

17) This can be proved using characters [BD.11].1. thus V2m−1 ∼ Hm . Exercise 11. Let σn be the map which σ1 induces on Vn . §3]. in such a way that the subspace v. σ2m−1 = −1. This is why a complex antilinear = map σ on a complex vector space V such that σ 2 = −1 is called a quaternionic structure. i. 5 and also a representation of the diagonal Sp(1)-subgroup. Let V2m be the set 2m+1 σ of fixed-points of σ2m . 2 Then σ1 = −1 . Real and quaternionic representations It is standard practice to work primarily with representations on complex vector spaces.18) 2 σ If n = 2m is even then σ2m = 1 and σ2m is a real structure on V2m . Then Sp(1) acts on the underlying real vector space R4m . we define the structure map σ1 : V1 → V1 by σ1 (z1 x + z2 y) = −¯2 x + z1 y z ¯ z1 . the action of which is given by g(u ⊗ v) = g(u) ⊗ g(v). z2 ∈ C.Each irreducible representation Vn is thus an eigenspace of the Casimir operator with eigenvalue −n(n + 2). §2. Then their tensor product Vm ⊗ Vn is naturally an Sp(1) × Sp(1)-representation. = σ Thus V2m is a representation of Sp(1) on the real vector space R2m+1 . 2 If on the other hand n = 2m − 1 is odd. A thorough guide to this process is in [BD. ¯ (1. and σ1 coincides with the map j of Section 1. Proposition 5. 5 12 . Representations on real (and quaternionic) vector spaces are obtained using antilinear structure maps. Let Vm and Vn be two Sp(1)-representations. iσ2m−1 (v) ∼ R4 = is isomorphic to the quaternions.5] or weights [FH. This real vector space comes equipped with the complex structure i and the structure map σ2m−1 . σ2m−1 (v). Then V2m ∼ R is preserved by the action of Sp(1). iv. we can construct all Spin(4) and hence all SO(4) representations in = this fashion — see [S1. Since Sp(1) × Sp(1) ∼ Spin(4). σn (z1 xn−k yk ) = (−1)k z1 xk yn−k . In the case of Sp(1)representations. Vm ⊗ Vn ∼ Vm+n ⊕ Vm+n−2 ⊕ · · · ⊕ Vm−n+2 ⊕ Vm−n = for m ≥ n.6].e. and = σ V2m ⊗R C ∼ V2m . (1. The irreducible decomposition of the diagonal Sp(1)-representation on Vm ⊗ Vn is given by the famous Clebsch-Gordon formula.2.

However. due to the great complicating factor of non-commutativity. Many of the ideas which work beautifully for real or complex numbers are not suited to quaternions. counted with multiplicities. Then the mapping f : H → H is surjective. p. and in particular f has zeros in H. quaternions can be used to recreate many of the structures over the real and complex numbers with which we are familiar.1. 205] Let f be a polynomial over H of degree n > 0 of the form m + g. we obtain the following expression: (a1 X + b1 )(a2 X + b2 ) = a1 Xa2 X + a1 Xb2 + b1 a2 X + b1 b2 . where m is a monomial of degree n and g is a polynomial of degree < n. since i1 X − Xi1 ∈ I for all X ∈ H.3.3 Difficulties with the Quaternions Quaternions are far less predictable than their lower-dimensional cousins. which highlight the need for caution: but also. The purpose of this section is to outline some of the difficulties with a few examples.1 The Fundamental Theorem of Algebra for Quaternions The single biggest reason for using the complex numbers in preference to any other number field is the so-called ‘Fundamental Theorem of Algebra’ — every complex polynomial of degree n has precisely n zeros. It quickly becomes obvious that non-commutativity is going make any attempt to factorise a general polynomial extremely troublesome. There is a quaternionic analogue of the theorems for real and complex numbers. Quaternion behaviour is many degrees freer and less predictable. point the way to some of the successes we will encounter. 13 . we need to restrict our attention considerably. If we multiply together two ‘linear factors’. it is hoped.3. This is typical of the difficulties we encounter with quaternions. The real numbers are not so well-behaved: a real polynomial of degree n can have fewer than n real roots. the cubic X 2 i1 Xi1 + i1 X 2 i1 X − i1 Xi1 X 2 − Xi1 X 2 i1 takes the value zero for all X ∈ H. At the other extreme. Moreover. the equation i1 X − Xi1 + 1 = 0 has no solutions at all! In order to arrive at any kind of ‘fundamental theorem of algebra’. and attempts to use them often result in lengthy and cumbersome mathematics — most of which dates from the 19th century and is now almost forgotten. less general and because of this less useful. Then there is the following ‘fundamental theorem of algebra for quaternions’: Theorem 1. ai ∈ H \ {0}. For example. but because of noncommutativity the quaternionic version is more complicated.1 [EKR. there are many polynomials which display extreme behaviour. 1. We define a monomial of degree n to be an expression of the form a0 Xa1 Xa2 · · · an−1 Xan . with modesty and care.

Hence regular functions form a left H-module. which we shall call 14 . Fueter. dq h→0 it can be shown [Su. A function f : C → C is holomorphic if it has a well-defined complex derivative.3 Quaternion Linear Algebra In the same way as we define real or complex vector spaces. it is easy to see that their sum f + g must also be regular. non-commutativity causes immediate algebraic difficulties. in 1935. so the only functions which are quaternion-differentiable in this sense are affine linear. It was not until the work of R.e. As usual.3. but their product f g. using a quaternionic analogue of the Cauchy-Riemann equations. Complex analysis can be described as the study of holomorphic functions. ∂q0 ∂q1 ∂q2 ∂q3 (1.20) The theory of quaternionic analysis which results is modestly successful. that a suitable definition of a ‘regular quaternionic function’ was found. though it is little-known. can be written as a convergent power series.1. i. the composition f ◦ g and the right scalar multiple f q need not be. Sadly. Theorem 1] that f must take the form f (q) = a + bq for some a.3. neither of these definitions proves interesting when applied to quaternions — the former is too restrictive and the latter too general. Cauchy’s theorem and the Cauchy integral formula. Hamilton never developed a successful ‘quaternion calculus’. we can define quaternionic vector spaces or H-modules — a real vector space with an H-action. §3. The work of Fueter is described and extended in the papers of Deavours [D] and Sudbery [Su].e. In contrast to the complex case. 1. One of the fundamental results in complex analysis is that every holomorphic function is analytic i. 4i1 (1. which include quaternionic versions of Morera’s theorem. Hamilton and his followers were aware of this — it was in Hamilton’s work on quaternions that some of the modern ideas in the theory of functions of several real variables first appeared. the components of a quaternion can be written as quaternionic polynomials. 1 q0 = (q − i1 qi1 − i2 qi2 − i3 qi3 ) 4 q1 = 1 (q − i1 qi1 + i2 qi2 + i3 qi3 ) etc. as must the left scalar multiple qf for q ∈ H. b ∈ H. but it is difficult to see how to describe any further algebraic structure.19) This takes us to the other extreme: the study of quaternionic power series is the same as the theory of real-analytic functions on R4 . for q = q0 + q1 i1 + q2 i2 + q3 i3 . If f and g are regular functions. Despite the benefits of this work to mathematics as a whole. A regular function on H is defined to be a real-differentiable function f : H → H which satisfies the Cauchy-Riemann-Fueter equations: ∂f ∂f ∂f ∂f + i1 + i2 + i3 = 0.2 Calculus with Quaternions The monumental successes of complex analysis make it natural to look for a similar theory for quaternions. For a function f : H → H to have a well-defined derivative df = lim (f (q + h) − f (q))h−1 .

f2 b) = f2 µ(a. b1 ) − (a. q for our tensor product we can define a generator uq ⊗ v − u ⊗ qv (replacing f (u) ⊗ v − u ⊗ f (v)) for R. Then if we define the ideal R(U. B) is the vector space freely generated (over F) by all elements (a. b ∈ B. v)q2 . q2 v) = q2 µ(q1 u. f (b)). There is the added complication that we need to say whether this multiplication is on the left or the right. In this context. V ) is equal to the whole of F (U. Now suppose that U . B) generated by all elements of the form (a1 + a2 . for a. q ∈ H and u ∈ U . V ). v) and µ(q1 u. b). In this (more restricted) case we do obtain a well-defined ‘quaternionic tensor product’ U ⊗H V = F (U.21) where F (A. v). If we try to define a bilinear map µ : U × V → W as above. then an F-bilinear map µ : A×B → C satisfies µ(f1 a. We can now define an H-bilinear map to be one which satisfies µ(q1 u. B) . R(A. f1 . b) − (f (a). v ∈ V . b) and (a. B) is the subspace of F (A. Several of the familiar ideas which work for vector spaces over a commutative field work just as well for H-modules. we can define quaternion linear maps between H-modules in the obvious way. f2 b) = f1 f2 µ(a. This is equivalent to having µ(f1 a. B and C are vector spaces over the commutative field F. and so we can define a dual H-module U × of quaternion linear maps α : U → H. q2 ∈ H and let u ∈ U . We will work with left H-modules — this choice is arbitrary and has only notational effects on the resulting theory. b) − (a2 . Since q1 q2 = q2 q1 in general. Let U and V be left H-modules. this process does not work for quaternions. we restrict ourselves to talking about H-modules of the form Hn = H ⊗R Rn . V )/R(U. or alternatively µ(q1 u. such that p(qu) = (pq)u for all p. q2 v) = q2 q1 µ(u. V ) in the same way as above. However. which we write (q. Not surprisingly. Similar difficulties arise if we try to define a tensor product over the quaternions. but this is only saying that H⊗R Rm ⊗R Rn ∼ H⊗R Rmn .scalar multiplication. if we try to define quaternion bilinear maps we run into trouble. If A. then we need both µ(q1 u. we discover that R(U. = = our ‘quaternionic tensor product’ is merely a real tensor product in a quaternionic setting. we do get the useful relation Hm ⊗H Hn ∼ Hmn . b1 + b2 ) − (a. v) = q1 q2 µ(u. bj ∈ B and f ∈ F. f2 ∈ F. Let q1 . q2 v) = q1 µ(u. V and W are (left) H-modules. b2 ) f (a. b) f (a. u) → q · u or just qu. b) = f1 µ(a. this cannot work. If A and B are vector spaces over the commutative field F. 15 . b. aj ∈ A. A left H-module is thus a real vector space U with an action of H on the left. b) ∈ A×B and R(A. b) − (a1 . V ) which inherits a left H-action from U and a right H-action from V . B) (1. their tensor product over F is defined by A ⊗F B = F (A. For example. Similarly. For example. One way around both of these difficulties is to demand that our H-modules should also have an H-action on the right. In this case. b) for all a ∈ A. b) and µ(a. The drawback with this system is that it does not really provide any new insights. vq2 ) = q1 µ(u. Hn is an H-module with the obvious left-multiplication. q2 v) = q1 µ(u. If we insist on having a left and a right H-action. v)¯2 . b) − (a. so U ⊗F V = {0}.

1.3.4

Summary

By now we have become familiar with some of the more elementary ups and downs of the quaternions. In the 20th century they have often been viewed as a sort of mathematical Cinderella, more recent techniques being used to describe phenomena which were first thought to be profoundly quaternionic. However, we have seen that it is possible to produce quaternionic analogues of some of the most basic algebraic and analytic ideas of real and complex numbers, often with interesting and useful results. In the next chapter we will continue to explore this process, turning our attention to quaternionic structures in differential geometry.

16

Chapter 2 Quaternionic Differential Geometry
In this chapter we review some of the ways in which quaternions are used to define geometric stuctures on differentiable manifolds. In the same way that real and complex manifolds are modelled locally by the vector spaces Rn and Cn respectively, there are manifolds which can be modelled locally by the H-module Hn . These models work by defining tensors whose action on the tangent spaces to a manifold is the same as the action of the quaternions on an H-module. There are two important classes of manifolds which we shall consider: those which are called ‘quaternionic manifolds’, and a more restricted class called ‘hypercomplex manifolds’. In this chapter we describe these important geometric structures. We also review the decomposition of exterior forms on complex manifolds, and examine some of the parallels of this theory which have already been found in quaternionic geometry. Many of the algebraic and geometric foundations of the material in this chapter are collected in (or can be inferred from) Fujiki’s comprehensive article [F].

2.1

Complex, Hypercomplex and Quaternionic Manifolds

Complex Manifolds A complex manifold is a 2n-dimensional real manifold M which admits an atlas of complex charts, all of whose transition functions are holomorphic maps from Cn to itself. As we saw in Section 1.3, the simplest notions of a ‘quaternion-differentiable map from H to itself’ are either very restrictive or too general, and the ‘regular functions’ of Fueter and Sudbery are not necessarily closed under composition. This makes the notion of ‘quaternion-differentiable transition functions’ less interesting that one might hope. An equivalent way to define a complex manifold is by the existence of a special tensor called a complex structure. A complex structure on a real vector space V is an automorphism I : V → V such that I 2 = − idV . (It follows that dim V is even.) The complex structure I gives an isomorphism V ∼ Cn , since the operation ‘multiplication = by i’ defines a standard complex structure on Cn . An almost complex structure on a 2n-dimensional real manifold M is a smooth tensor I ∈ C ∞ (End(T M )) such that I is a complex structure on each of the fibres Tx M .

17

Now, if M is a complex manifold, each tangent space Tx M is isomorphic to Cn , so taking I to be the standard complex structure on each Tx M defines an almost complex structure on M . An almost complex stucture I which arises in this way is called a complex structure on M , in which case I is said to be integrable. The famous NewlanderNirenberg theorem states that an almost complex structure I is integrable if and only if the Nijenhuis tensor of I NI (X, Y ) = [X, Y ] + I[IX, Y ] + I[X, IY ] − [IX, IY ] vanishes for all X, Y ∈ C ∞ (T M ), for all x ∈ M . The Nijenhuis tensor NI measures the (0, 1)-component of the Lie bracket of two vector fields of type (1, 0). 1 On a complex manifold the Lie bracket of two (1, 0)-vector fields must also be of type (1, 0). Thus if I is an almost complex structure on M and NI ≡ 0, then M is a complex manifold in the sense of the first definition given above. We can talk about the complex manifold (M, I) if we wish to make the extra geometric structure more explicit — especially as the manifold M might admit more than one complex structure. Hypercomplex Manifolds This way of defining a complex manifold adapts itself well to the quaternions. A hypercomplex structure on a real vector space V is a triple (I, J, K) of complex structures on V satisfying the equation IJ = K. (It follows that dim V is divisible by 4.) If we identify I, J and K with left-multiplication by i1 , i2 and i3 , a hypercomplex structure gives an isomorphism V ∼ Hn . Equivalently, a hypercomplex structure is defined by a pair of = complex structures I and J with IJ = −JI. It is easy to see that if (I, J, K) is a hypercomplex structure on V , then each element of the set {aI+bJ +cK : a2 +b2 +c2 = 1} ∼ S 2 = is also a complex structure. We arrive at the following quaternionic version of a complex manifold: Definition 2.1.1 An almost hypercomplex structure on a 4n-dimensional manifold M is a triple (I, J, K) of almost complex structures on M which satisfy the relation IJ = K. If all of the complex structures are integrable then (I, J, K) is called a hypercomplex structure on M , and M is a hypercomplex manifold. A hypercomplex structure on M defines an isomorphism Tx M ∼ Hn at each point = x ∈ M . As on a vector space, a hypercomplex structure on a manifold M defines a 2-sphere S 2 of complex structures on M . Some choices are inherent in this standard definition. A hypercomplex structure as defined above gives T M the structure of a left H-module, since IJ = K. This induces a right H-module structure on T ∗ M , using the standard definition I(ξ), X = ξ, I(X) etc., for all ξ ∈ T ∗ M, X ∈ T M , since on T ∗ M we now have ξ, IJ(X) = I(ξ), J(X) = JI(ξ), X . In this thesis we will make more use of the hypercomplex structure on T ∗ M than that on T M . Because of this we will usually define our hypercomplex structures so that IJ = K on T ∗ M rather than T M . This has only notational effect on the theory, but it does pay to
1

Tensors of type (p, q) are defined in the next section.

18

whereas some authors use the opposite convention. The group of automorphisms of Tx M preserving such a structure is isomorphic to GL(n. H). Thus a complex manifold is precisely a real manifold M 2n with a GL(n. C)-structure Q on M contain the same information. It is well-known that S 4 does not even admit a global almost complex structure. A complex manifold generally admits many torsion-free connections which preserve the complex structure. H)structure Q. C)-structure Q admitting a torsion-free connection (in which case Q itself is said to be ‘integrable’). H). Thus the group of symmetries of Hn is the product GL(1. in which case it is easy to show that I is integrable. the quaternionic projective line HP 1 is diffeomorphic to the 4-sphere S 4 . The bundle Q admits a torsion-free connection if and only if there is a torsion-free linear connection on M with I = 0. Quaternionic Manifolds and the Structure Group GL(1. This process is described in [S3. Complex and hypercomplex manifolds can be decribed succinctly in terms of Gstructures on manifolds. R)bundle whose fibre over x ∈ M is the group of isomorphisms Tx M ∼ R4n . the GL(n. H) .e. If M 4n has an almost hypercomplex structure then the group of automorphisms preserving this structure is isomorphic to GL(n. The uniqueness of any torsion-free connection on Q follows from analysing the Lie algebra gl(n. R) preserving a quaternionic structure. A G-structure Q on M is a principal subbundle of P with structure group G. Suppose M 2n has an almost complex structure. H) ×R∗ GL(n. despite behaving extremely like the quaternions locally.be aware of how it affects other standard notations. a hypercomplex manifold is seen to be a real manifold M with an integrable GL(n. and the connection is called the Obata connection. on a hypercomplex manifold there is a unique torsion-free connection such that I= J= K = 0. so HP 1 can certainly not be hypercomplex. H) commutes with that of the left H-action of the group GL(1. Let P be the principal frame bundle of M . This encourages us to think of a connection whose curvature is acted upon trivially by the hypercomplex structure as anti-self-dual rather than self-dual. We can multiply on the right by any real multiple of the identity 19 . The reason (and the solution) for this difficulty is that GL(n. §6]. H)GL(n. H) is not the largest subgroup of GL(4n. H) as acting on Hn by right-multiplication by n × n quaternionic matrices. i. One important difference between complex and hypercomplex geometry is the existence of a special connection. whereas Sudbery’s regular functions form a right H-module. H). Following the same line of argument. for us the hypercomplex − structure acts trivially on the anti-self-dual 2-forms ω1 = dx0 ∧ dx1 − dx2 ∧ dx3 etc. then the action of GL(n. C). For example. H). H)GL(n. H) Not all of the manifolds which we wish to describe as ‘quaternionic’ admit hypercomplex structures. Such things are largely a matter of taste — we are choosing to follow the notation of Joyce [J1] for whom regular functions form a left H-module. For example. which we write GL(1. If we think of GL(n. Thus an almost complex structure I and a GL(n. By contrast. R). This was proved by Obata in 1956. Let G be a = Lie subgroup of GL(n.

In four dimensions a manifold is said to be quaternionic K¨hler if and only if it a is self-dual and Einstein. H)GL(n.2 Differential Forms on Complex Manifolds This section consists of background material in complex geometry. and a hyperk¨hler a manifold has compatible hypercomplex and complex-symplectic structures.2 [S3. H) is the same as Sp(1)GL(n. There is still an S 2 -bundle of local almost-complex structures which satisfy IJ = K. When n = 1 the situation is different. a The quaternionic analogue of the compact group U(n) is the group Sp(n). J and K. Y ) = g(IX. a an integrable U(n)-structure is called a K¨hler structure. Each tangent space Tx M still admits a hypercomplex structure giving an isomorphism Tx M ∼ Hn . Such a manifold is called hyperk¨hler . For a comprehensive study of quaternionic manifolds see [S3] and Chapter 9 of [S4]. Definition 2. so without loss of generality we can reduce the first factor to Sp(1). Y ∈ Tx M for all x ∈ M . Then M admits a Riemannian metric g with g(IX. which gives a quotient construction for hyperk¨hler a manifolds. Using each of the complex structures. H) = Sp(1) ×Z2 GL(n. but this isomorphism does not necessarily = arise from globally defined complex structures on M . a we define three independent symplectic forms ωI . C)-structure Q on a complex manifold M admits a further reduction to an integrable U(n)-structure Q . quaternionic manifolds are a generalisation of hypercomplex manifolds in the following way.1] A quaternionic manifold is a 4n-dimensional real manifold M (n ≥ 2) with an Sp(1)GL(n. We also define the differentiable 2-form ω(X. Quaternionic K¨hler manifolds are the subject of [S2]. Then the complex 2-form ωJ + iωK is holomorphic with respect to the complex structure I. Thus GL(1. H)-structure Q reduces to an Sp(1)Sp(n)structure Q and M is said to be quaternionic K¨hler. H). if a quaternionic manifold has a metric compatible with the torsion-free Sp(1)GL(n. H) share the same centre R∗ ). H)-structure Q admitting a torsion-free connection. and M is a symplectic manifold — so M has compatible complex and symplectic structures. IY ) = g(X. 1. Similarly. a 2. H)-structure Q reduces to an integrable Sp(n)-structure Q admits a metric g which is simultaneously K¨hler for each of the complex structures a I. especially ideas which encourage the development of interesting quaternionic versions. If such a metric arises from an integrable U(n)-structure Q then ω will be a closed 2-form. The group Sp(1)Sp(n) is a a maximal proper subgroup of SO(4n) except when n = 1. H)-structure. Y ). Y ) for all X. ωJ and ωK . A hypercomplex manifold whose GL(n. Riemannian Manifolds in Complex and Quaternionic Geometry Suppose the GL(n. but it is free to ‘rotate’. the metric g is called a K¨hler a a metric and the symplectic form ω is called a K¨hler form. H) and GL(n. In this case M is called a K¨hler manifold.(since GL(1. More information on this 20 . Hyperk¨hler a manifolds are studied in [HKLR]. where as we know Sp(1)Sp(1) = SO(4). In four dimensions = we make the special definition that a quaternionic manifold is a self-dual conformal manifold.1. since Sp(1)×Sp(1) ∼ SO(4). In terms of tensors. then the Sp(1)GL(n.

Let (M.0 M = T1. A crucial fact is that if I is integrable.q M.0 M ⊕ T0.q M ) and Ωk (M ) = p+q=k p. These definitions rely only on the fact that I is an almost complex structure.4) where π p.1) gives rise to a decomposition of exterior forms of all powers k C ⊗R Λk T ∗ M = p=0 ∗ ∗ Λp (T1.1 M ). so Ωp.q+1 ◦ d. Define two first-order differential operators.1) ∗ ∗ so Λ1.and other aspects of complex geometry can be found in [W] and [GH.2) With this notation. which is why a function which is holomorphic with respect to I is antiholomorphic with respect to −I.q+1 (M ) ∂ = π p. A holomorphic function f ∈ C ∞ (M.q (M ) ∂ = π p+1. Thus Λ1.1 M is called the antiholomorphic cotangent space of M . (2. where these summands are the +i and −i eigenspaces of I respectively. A closely linked statement is that Λ1.e.1 M. q) or just a (p. the decomposition (2. Λ1.1 M = T0. (2. I) be a complex manifold. §0. ∂ : Ωp.2].q M = Λp T1.e. i. The same is true of C ⊗R T ∗ M . If we reverse the sense of the complex structure (i. Λp. The operator ∂ is called the Dolbeault operator.q M .1 M .0 M and Λ0.0 M ⊕ Λ0.3) p+q=k A smooth section of the bundle Λ M is called a differential form of type (p.0 M for all m ∈ M . Equation (2.q (M ) → Ωp+1. However.0 M ) ⊗ Λk−p (T0.q ◦ d and ∂ : Ωp.q (M ) → Ωp. (2. if we swap I for −I) then we reverse the roles of the holomorphic and antiholomorphic spaces.0 M is a holomorphic vector bundle. There are various ways of writing these ∗ ∗ weight spaces. fairly standard is the notation C ⊗R T ∗ M = T1.0 M ).0 M is called the holomorphic cotangent space and Λ0. for our purposes it will be more useful to adopt the notation of [S3].1 M . q)-forms on M . q)-form. f is holomorphic if and only if df ∈ C ∞ (M.q denotes the natural projection from C ⊗ Λk M onto Λp.1) is an example of the more general decomposition into types C ⊗R Λk T ∗ M = Λp.q Ωp. C) is a smooth function whose derivative takes values only in Λ1.1 M.q (M ) = C ∞ (M. these are the only two components represented in 21 . writing C ⊗R T ∗ M = Λ1. Then I gives T M and T ∗ M the structure of a U(1)-representation and the complexification C⊗R T M splits into two weight spaces with weights ±i. From standard multilinear algebra. Define the bundle ∗ ∗ Λp.q (M ). We write Ωp.q (M ) for the set of (p.0 M ⊗ Λq T0. (2.

C) is holomorphic if and only if ∂f = 0 and for this reason ∂ is sometimes called the Cauchy-Riemann operator.Figure 2.0 r rr r Ω0.q−1 p. This induces a representation on C ⊗ Λ• T ∗ M . the exterior differential d. a (p. etc. i. . I(ω) = i(p − q)ω. etc.1 ∂ ¨ ¨¨ B ¨ r rr r j ∂ rr B ∂ ¨¨ ¨ Ω0. It is easy = to see that if ω ∈ Λp.2. Since U(1) is abelian. Similarly. .q+1 . The Dolbeault complex begins with the representation of the complex structure I ∼ u(1) on C ⊗ T ∗ M . .2. p. . 0)-form α is said to be holomorphic if and only if ∂α = 0.1: The Dolbeault Complex Ω2. C) r rr r j ∂ rr ¨¨ j ∂ rr ¨¨ ¨ B ∂ ¨¨ ¨ Ω1.0 B ∂ ¨¨ ¨¨ ¨ . p. Proposition 2. They are parametrised by the integers.q 2 (2. The representations of U(1) on complex vector spaces are particularly easy to understand.e. Representations of the Lie algebra u(1) are then of the form d n : z → nz. In other words.q M .q H∂ = Ker(∂ ) Im(∂ p. C) = C∗ n : eiθ → eniθ for some n ∈ Z.q M is the bundle of U(1)-representations of the type p−q .0 = C ∞ (M. .5) This gives rise to the Dolbeault complex. . taking the form n : U(1) → GL(1. . Writing ∂ for the particular map ∂ : p. . ) A function f ∈ C ∞ (M. we define the Dolbeault cohomology groups p. ∂ 2 = ∂∂ + ∂∂ = ∂ = 0. the irreducible representations all are one-dimensional. An immediate consequence of this is that on a complex manifold M . . . d = ∂ + ∂ [WW. Λp. rr rr r ¨ ∂ j r Ω ¨ ¨¨ B ∂ ¨¨ ¨ 1.2 . A useful way to think of the Dolbeault complex is as a decomposition of C ⊗ Λk T ∗ M into types of U(1)-representation. 22 .105].q .q Ω → Ωp.1 rr r rr ∂ j r ¨ ¨¨ Ω0.

H) on C2n . Then W is a vector bundle over M with fibre W . We define the associated bundle W = P ×G W = P ×W . g −1 · w). H).1 [Kr.6) = where V1 is the basic representation of Sp(1) on C2 and E is the basic representation of GL(n. 122]. w) ∈ P × W by (p. Fujiki’s account of this [F. where ∗ : Λk T ∗ M → Λ4n−k T ∗ M is the Hodge star. where ωI . Note that Ψ ∧ Ψ = −2Ω. Let P be a principal G-bundle over the differentiable manifold M and let W be a G-module. We will usually just write W for W.5][Bon. Theorem 3. relying on context 23 .) This representation also describes the (co)tangent bundle of a quaternionic manifold in the following way. Kraines [Kr] and Bonan [Bon] used the fundamental 4-form to decompose the space Λk T ∗ M in a similar way to the Lefschetz decomposition of differential forms on a K¨hler manifold [GH. We express the Sp(1)GL(n. ωJ and ωK are the local K¨hler forms associated to local almost complex a structures I. In this case. J and K with IJ = K.2. Bonan further refines this decomposition for quaternion-valued forms. p. using exterior multiplication by the globally defined quaternionic 2-form Ψ = i1 ωI + i2 ωJ + i3 ωK . The fundamental 4-form is globally defined and invariant under the induced action of Sp(1)Sp(n) on Λ4 T ∗ M .3 Differential Forms on Quaternionic Manifolds Any G-structure on a manifold M induces a representation of G on the exterior algebra of M . where the µk−4j are effective (k − 4j)-forms. (This uses the standard convention of working with complex representations.3. the structure map is the tensor product of the quaternionic structures on V1 and E. Another way to consider the decomposition of forms on a quaternionic manifold is as representations of the group Sp(1)GL(n. This leads to the following theorem: Theorem 2. H)-representation on Hn by writing Hn ∼ V1 ⊗ E. every every k-form φ admits a unique decomposition φ= 0≤j≤k/4 Ωj ∧ µk−4j . A a differential k-form µ is said to be effective if Ω ∧ ∗µ = 0. which in the presence of suitable structure maps can be thought of as complexified real representations. G where g ∈ G acts on (p. w) · g = (f · g. Theorem 2] For k ≤ 2n + 2. a The decomposition of differential forms on quaternionic K¨hler manifolds began by a considering the fundamental 4-form Ω = ωI ∧ ωI + ωJ ∧ ωJ + ωK ∧ ωK . §2] explains many quaternionic analogues of complex and K¨hler geometry. (2.

Theorem 4. Proposition 4. We will use such descriptions in detail in later chapters. H)action on the bundle of exterior k-forms Λk T ∗ M .3. and find that they play a prominent role in quaternionic algebra. we define a sequence of differential operators 0 −→ C ∞ (A0 ) −→ C ∞ (A1 = T ∗ M ) −→ C ∞ (A2 ) −→ . so that we write (2. we must antisymmetrise completely on E.11) is a complex. −→ C ∞ (A2n ) −→ 0. The following theorem of Salamon relates the integrability of such a structure with the sequence of operators in (2. D=d D D D (2. H) a representations Lk .10) Letting p denote the natural projection p : Λk T ∗ M → Ak and setting D = p ◦ d. Λk T ∗ M can be further decomposed j into representations of the compact group Sp(1)Sp(n). This decomposition is given j by Salamon [S3. = (2. This refinement is performed in detail by Swann [Sw].9) The bundle Ak can be described using the decomposition into types for the local almost complex structures on M as follows [S3.8) where Lk is an irreducible representation of GL(n.to distinguish between the bundle and the representation.8) to obtain Vk . H)-structure. . I (2. [k/2] [k/2] Λ T M ∼ Λk (V1 ⊗ E) ∼ = = k ∗ j=0 S k−2j (V1 ) ⊗ Lk j ∼ = j=0 Vk−2j ⊗ Lk .11) This is accomplished using only the fact that M has an Sp(1)GL(n. if M 4n is a quaternionic manifold with Sp(1)GL(n. This induces an Sp(1)GL(n. 2 24 . §4]. This is because every Sp(1)-representation Vn is generated by its highest weight spaces taken with respect to all the different linear combinations of I. If we symmetrise completely on V1 in Equation (2. j (2. then the cotangent bundle is a vector bundle associated to the principal bundle Q and the representation V1 ⊗ E. §1]. H)-structure Q. Following Salamon [S3. H). and used to demonstrate significant results. such a manifold is called ‘almost quaternionic’. J and K. This theorem is analogous to the familiar result in complex geometry that an almost 2 complex structure on a manifold is integrable if and only if ∂ = 0. If M is quaternion K¨hler.2]: 2 Ak = I∈S 2 Λk.2 [S3. .0 M.7) (T ∗ M )C ∼ V1 ⊗ E = (though we will usually omit the complexification sign).11): Theorem 2.1] An almost quaternionic manifold is quaternionic if and only if (2. Salamon therefore defines the irreducible subspace Ak ∼ Vk ⊗ Λk E. along with more details concerning the nature of the GL(n.

q ⊕ Λq. in particular the isomorphism Vr ⊗ V1 ∼ Vr+1 ⊕ Vr−1 . The ellipticity properties of our double complex are more similar to those of a real-valued version of the Dolbeault complex. Then ω ∈ Λq. we consider only the action of the Sp(1)-factor and decompose Λk T ∗ M into irreducible Sp(1)-representations — a fairly easy process achieved by considering weights. This = encourages us to define two ‘quaternionic Dolbeault’ operators D and D.11) discovered by Simon Salamon.q ⊕ Λq. We argue that this is the best quaternionic analogue of the Dolbeault complex. Because of this similarity.Chapter 3 A Double Complex on Quaternionic Manifolds Until now we have been discussing known material. Determining where our double complex is elliptic has been far more difficult than for the de Rham or Dolbeault complexes. C)-bundle defined by the complex structure. and leads to new cohomology groups on quaternionic manifolds. Instead of using the more complicated structure of Λk T ∗ M as an Sp(1)GL(n. 3. The resulting decomposition gives rise to a double complex through the Clebsch-Gordon formula.p . we shall begin with a discussion of real-valued forms on complex manifolds.p )R is a real vector bundle associated to the principal GL(n. H)-module. This gives 25 . This is more like the decomposition of real-valued forms on complex manifolds. where the subscript R denotes the fact that we are talking about real forms.8). which plays a similar role to that of the (k. and so ω + ω is a real-valued exterior form in (Λp. The space (Λp. The new double complex is obtained by simplifying the Bonan decomposition of Equation (2. In this chapter we present a discovery which as far as the author can tell is new — a double complex of exterior forms on quaternionic manifolds.q M . 0)-forms on a complex manifold. our double complex forms an isosceles triangle.q = Λp. as if the diamond is ‘folded in half’.p )R . The ‘top row’ of this double complex is exactly the complex (2. The main geometric difference between this double complex and the Dolbeault complex is that whilst the Dolbeault complex is a diamond.1 Real forms on Complex Manifolds Let M be a complex manifold and let ω ∈ Λp.

0 ⊕ Ω0. (3. ∈ (Ωp+1. and why it is significant.1 Thus there is a double complex of real forms on a complex manifold. It is its own conjugate and so naturally a real vector bundle associated to the trivial representation of U(1).0 ⊕ Ω0.p )R . . so ∂ ⊕ ∂ = ∂ ⊕ ∂ = d. induced from x ∗ the action on Tx M .3 ¨ ¨¨ . . Then d(ω + ω) = (∂ ⊕ ∂)(ω + ω) + (∂ ⊕ ∂)(ω + ω) .0 = C (M ) ∞ ¨ ∂⊕∂ rr rr j r ∂ ⊕ ∂ rr j Ω1.p )R .1 ⊕ Ω1. . This gives the following double complex (where the real subscripts are omitted for convenience).1) k k 2 The condition p > q ensures that we have no repetition. Let ω + ω ∈ (Ωp. p.q ⊕ Λq. it is not elliptic everywhere. Examples of elliptic This decomposition is also given by Reyes-Cari´n [R. The bundle ΛR 2 only appears when k is even.p )R ⊕ ΛR 2 .q+1 ⊕ Ωq+1.p )R o [[Λ ]]. .1]. Call this operator ∂ ⊕ ∂.1: The Real Dolbeault Complex Ω3.q ⊕ Ωq. obtained by decomposing Λk T ∗ M into subrepresentations of the action of u(1) = I . Λk T ∗ M = R p+q=k p>q 1 k k 2 (Λp. . who calls the bundle (Λp.2 ∂ ⊕ ∂¨ ¨ ¨¨ B ¨¨ rr rr ∂ ⊕ ∂ rr j ∂ ⊕ ∂ ¨¨ ¨ ¨¨ B ¨ Ω1. When p = q. §3. .1 ¨¨ Ω2.q ⊕ Λq. and there is only one differential operator acting on Ωp. d ¨¨ B ¨ rr d ¨¨ B ¨ Ω 0.0 ⊕ Ω0. . ω = ω.p . ∂ ⊕ ∂ ¨¨ B ¨ rr r r ∂ ⊕ ∂ rr j Ω2. Ellipticity Whether or not a differential complex on a manifold is ‘elliptic’ is an important question. . . .a decomposition of real-valued exterior forms. etc. We shall now show what this means. etc. Then ∂ω + ∂ω ∈ (Ωp+1. in particular. .p+1 )R .q ⊕ Ωq. with striking topological. Figure 3.q ⊕ Ωq.p+1 )R ⊕ (Ωp. algebraic and physical consequences. This double complex is less well-behaved than its complex-valued counterpart.2 rr r ¨¨ ¨ .q 1 26 .

complexes include the de Rham and Dolbeault complexes. A thorough description of elliptic operators and elliptic complexes can be found in [W, Chapter 5]. Let E and F be vector bundles over M , and let Φ : C ∞ (E) → C ∞ (F ) be a differential operator. (We will be working with first-order operators, and so will only describe this ∗ case.) At every point x ∈ M and for every nonzero ξ ∈ Tx M , we define a linear map σΦ (x, ξ) : Ex → Fx called the principal symbol of Φ, as follows. Let ∈ C ∞ (E) with (x) = e, and let f ∈ C ∞ (M ) with f (x) = 0, df (x) = ξ. Then σΦ (x, ξ)e = Φ(f )|x .
∂ In coordinates, σ is often found by replacing the operator ∂xj with exterior multiplication ∂ by a cotangent vector ξ j dual to ∂xj . For example, for the exterior differential d : Ωk (M ) → Ωk+1 (M ) we have σd (x, ξ)ω = ω ∧ ξ. The operator Φ is said to be elliptic at ∗ x ∈ M if the symbol σΦ (x, ξ) : Ex → Fx is an isomorphism for all nonzero ξ ∈ Tx M .
n 0 1 2 3 A complex 0 −→ C ∞ (E0 ) −→ C ∞ (E1 ) −→ C ∞ (E2 ) −→ . . . −→ C ∞ (En ) −→ 0 is i said to be elliptic at Ei if the symbol sequence Ei−1 −→ Ei −→ Ei+1 is exact for all ∗ ξ ∈ Tx M and for all x ∈ M . The link between these two forms of ellipticity is as follows. If we have a metric on each Ei then we can define a formal adjoint Φ∗ : Ei → Ei−1 . i Linear algebra reveals that the complex is elliptic at Ei if and only if the Laplacian Φ∗ Φi + Φi−1 Φ∗ is an elliptic operator. i−1 i One important implication of this is that an elliptic complex on a compact manifold has finite-dimensional cohomology groups [W, Theorem 5.2, p. 147]. Whether a complex yields interesting cohomological information is in this way directly related to whether or not the complex is elliptic. The following Proposition answers this question for the real Dolbeault complex.

Φ

Φ

Φ

Φ

Φ

Φn+1

σΦ

σΦi+1

Proposition 3.1.1 For p > 0, the upward complex 0 −→ Ωp,p −→ Ωp+1,p ⊕ Ωp,p+1 −→ Ωp+2,p ⊕ Ωp,p+2 −→ . . . is elliptic everywhere except at the first two spaces Ωp,p and Ωp+1,p ⊕ Ωp,p+1 . For p = 0, the ‘leading edge’ complex 0 −→ Ω0,0 −→ Ω1,0 ⊕ Ω0,1 −→ Ω2,0 ⊕ Ω0,2 −→ . . . is elliptic everywhere except at Ω1,0 ⊕ Ω0,1 = Ω1 (M ). Proof. When p > q + 1, we have short sequences of the form Ωp−1,q −→ Ωp,q −→ Ωp+1,q (3.2) Ωq,p−1 −→ Ωq,p −→ Ωq,p+1 . Each such sequence is (a real subspace of) the direct sum of two elliptic sequences, and so is elliptic. Thus we have ellipticity at Ωp,q ⊕ Ωq,p whenever p ≥ q + 2.
∂ ∂ ∂ ∂

27

This leaves us to consider the case when p = q, and the sequence

Ωp+1,p −→ Ωp+2,2 −→ . . . etc. (3.3) Ω
p,p+1

0 −→ Ωp,p

−→ Ω

2,p+2

−→ . . . etc.

This fails to be elliptic. An easy and instructive way to see this is to consider the simplest 4-dimensional example M = C2 . ∗ Let e0 , e1 = I(e0 ), e2 and e3 = I(e2 ) form a basis for Tx C2 ∼ C2 , and let eab... = a b 01 00 11 01 1,1 denote e ∧ e ∧ . . . etc. Then I(e ) = e − e = 0, so e ∈ Λ . The map from Λ1,1 to Λ2,1 ⊕ Λ1,2 is just the exterior differential d. Since σd (x, e0 )(e01 ) = e01 ∧ e0 = 0 the symbol map σd : Λ1,1 → Λ2,1 ⊕ Λ1,2 is not injective, so the symbol sequence is not exact at Λ1,1 . Consider also e123 ∈ Λ2,1 ⊕ Λ1,2 . Then σ∂⊕∂ (x, e0 )(e123 ) = 0, since there is no bundle Λ3,1 ⊕ Λ1,3 . But e123 has no e0 -factor, so is not the image under σd (x, e0 ) of any form α ∈ Λ1,1 . Thus the symbol sequence fails to be exact at Λ2,1 ⊕ Λ1,2 . It is a simple matter to extend these counterexamples to higher dimensions and higher exterior powers. For k = 0, the situation is different. It is easy to show that the complex 0 −→ C ∞ (M ) −→ Ω1,0 ⊕ Ω0,1 −→ Ω2,0 ⊕ Ω0,2 −→ . . . etc. is elliptic everywhere except at (Ω1,0 ⊕ Ω0,1 ). This last sequence is given particular attention by Reyes-Cari´n [R, Lemma 2]. He o shows that, when M is K¨hler, ellipticity can be regained by adding the space ω to a the bundle Λ2,0 ⊕ Λ0,2 , where ω is the real K¨hler (1, 1)-form. a The Real Dolbeault complex is thus elliptic except at the bottom of the isosceles triangle of spaces. Here the projection from d(Ωp,p ) to Ωp+1,p ⊕ Ωp,p+1 is the identity, and arguments based upon non-trivial projection maps no longer apply. We shall see that this situation is closely akin to that of differential forms on quaternionic manifolds, and that techniques motivated by this example yield similar results.
d ∂⊕∂

3.2

Construction of the Double Complex

∗ Let M 4n be a quaternionic manifold. Then Tx M ∼ V1 ⊗ E as an Sp(1)GL(n, H)= representation for all x ∈ M . Suppose we consider just the action of the Sp(1)-factor. Then the (complexified) cotangent space effectively takes the form V1 ⊗ C2n ∼ 2nV1 . = k ∗ k Thus the Sp(1)-action on Λ T M is given by the representation Λ (2nV1 ). To work out the irreducible decomposition of this representation we compute the weight space decomposition of Λk (2nV1 ) from that of 2nV1 . 2 With respect to the action of a particular subgroup U(1)q ⊂ Sp(1), the representation 2nV1 has weights +1 and −1, each occuring with multiplicity 2n. The weights of Λk (2nV1 ) are the kwise distinct sums of these. Each weight r in Λk (2nV1 ) must therefore be a sum of p occurences of the weight ‘+1’ and p − r occurences of the weight ‘−1’, where 2p − r = k

This process for calculating the weights of tensor, symmetric and exterior powers is a standard technique in representation theory — see for example [FH, §11.2].

2

28

and 0 ≤ p ≤ k (from which it follows immediately that −k ≤ r ≤ k and r ≡ k mod 2). The number of ways to choose the p ‘+1’ weights is 2n , and the number of ways p 2n to choose the (p − r) ‘−1’ weights is p−r , so the multiplicity of the weight r in the representation Λk (2nV1 ) is Mult(r) = 2n
k+r 2

2n
k−r 2

.

For r ≥ 0, consider the difference Mult(r)−Mult(r +2). This is the number of weight spaces of weight r which do not have any corresponding weight space of weight r + 2. Each such weight space must therefore be the highest weight space in an irreducible subrepresentation Vr ⊆ Λk T ∗ M , from which it follows that the number of irreducibles Vr in Λk (2nV1 ) is equal to Mult(r)−Mult(r + 2). This demonstrates the following Proposition: Proposition 3.2.1 Let M 4n be a hypercomplex manifold. The decomposition into irreducibles of the induced representation of Sp(1) on Λk T ∗ M is
k

Λk T ∗ M ∼ =
r=0

2n
k+r 2

2n
k−r 2

2n
k+r+2 2

2n
k−r−2 2

Vr ,

where r ≡ k mod 2. We will not always write the condition r ≡ k mod 2, assuming that
p q

= 0 if q ∈ Z.

Definition 3.2.2 Let M 4n be a quaternionic manifold. Define the coefficient
n k,r

=

2n
k+r 2

2n
k−r 2

2n
k+r+2 2

2n
k−r−2 2

,
n k,r Vr .

and let Ek,r be the vector bundle associated to the Sp(1)-representation this notation Proposition 3.2.1 is written
k k n k,r Vr r=0

With

Λ T M∼ =
k ∗

=
r=0

Ek,r .

Our most important result is that this decomposition gives rise to a double complex of differential forms and operators on a quaternionic manifold. Theorem 3.2.3 The exterior derivative d maps C ∞ (M, Ek,r ) to C ∞ (M, Ek+1,r+1 ⊕ Ek+1,r−1 ). Proof. Let be a torsion-free linear connection on M preserving the quaternionic structure. Then : C ∞ (M, Ek,r ) → C ∞ (M, Ek,r ⊗ T ∗ M ). As Sp(1)-representations, this is : C ∞ (M, n Vr ) → C ∞ (M, n Vr ⊗ 2nV1 ). k,r k,r 29

. . C ∞ (E3.r+1 ⊕ Ek+1. etc.3 is equivalent to the following: Proposition 3.2. .5 On a quaternionic manifold M . (3. Again.r−1 ◦ d.r−1 ). instead of a diamond as in the Dolbeault complex.2 ) ¨¨ B ¨ D ¨¨ ¨ r rr r ¨ D rr j ¨¨ . and so D2 = DD + DD = D = 0.r ) → C ∞ (Ek+1.3 ) ¨¨ ¨ . Ek.0 = R) r r D rr j C ∞ (E2.1 ) ∞ D rr j D ¨¨ B ¨ C ∞ (E1.r under is contained in the Vr+1 and Vr−1 summands of Λk T ∗ M ⊗ T ∗ M .r be the natural projection from Λk T ∗ M onto Ek. Ek+1.r image of Ek. . 2 Figure 3.1). The first equation is equivalent to Theorem 3.4) Theorem 3. Thus the = k. so d maps (sections of) Ek. Definition 3.2. . the operators D and D map real forms to real forms. The rest follows immediately from decomposing the equation d2 = 0.r . we have d = D + D.r ) → C ∞ (Ek+1.r k. d = ∧ ◦ . . Define the operators D : C ∞ (Ek.0 ) ¨ ¨¨ D ¨¨ B ¨ rr r r D rr j Here is our quaternionic analogue of the Dolbeault complex. etc. whereas there is one irreducible Sp(1)-representation only for each nonnegative integer. 30 . rr r D ¨¨ B ¨ r C (E2.2: The Double Complex C ∞ (E3. . Since is torsion-free.3.r+1 ) D = πk+1. . This allows us to split the exterior differential d into two ‘quaternionic Dolbeault operators’. This is essentially because there is one irreducible U(1)representation for each integer. . .r to the Vr+1 and Vr−1 summands of Λk+1 T ∗ M . Note that this is a decomposition of real as well as complex differential forms.r−1 ) D = πk+1. There are strong similarities between this and the Real Dolbeault complex (Figure 3.Using the Clebsch-Gordon formula we have n Vr ⊗2nV1 ∼ 2n n (Vr+1 ⊕Vr−1 ).4 Let πk.r+1 ◦ d and D : C ∞ (Ek. Proof. Thus d : C ∞ (M.2.1 = T ∗ M ) ¨ ¨¨ D ¨¨ B ¨ rr C ∞ (E0. the quaternionic version only extends upwards to form an isosceles triangle.r ) → C ∞ (M.2.

almost complex and Riemannian structures in four dimensions is described in more detail in [S4. Let e1 = I(e0 ). . .6) These are the relations of the irreducible sp(1)-representation V2 . The relationship between quaternionic. and we see that E2. and let e ∈ Tx M . 31 . for k = 1 we no longer expect I 2 = J 2 = K 2 = −1. and by ‘act trivially’ we mean ‘annihilate’.etc. Because there is no suitable quaternionic version of holomorphic coordinates. In this way we obtain a basis {e . e3 } for Tx M ∼ H. .2.0 ) −→ C ∞ (E1. there is no ‘nice’ co-ordinate expression for a typical section of C ∞ (Ek. This makes no difference on T ∗ M but is important on the exterior powers Λk T ∗ M ..1 ) −→ C ∞ (E2. The celebrated splitting + − Λ2 T ∗ M ∼ Λ2 ⊕ Λ2 is an invariant of the conformal class of any Riemannian 4-manifold. Here there is a splitting only in the middle dimension. J = 0 ∗ and K be local almost complex structures at x ∈ M . Then Dα = − 3 1 4 (r − 1) + 1 (I 2 + J 2 + K 2 ) dα r+1 We are assuming throughout that uppercase operators like I. Thus the leading edge of the double complex 0 −→ C ∞ (E0. In order to determine the decomposition of a differential form. In particular. Let I.11) discovered by Salamon.6 Four Dimensions This double complex is already very well-known and understood in four dimensions.9) — they are both the subbundle of Λk T ∗ M which includes all Sp(1)-representations of the form Vk . define the 2-forms ± ω1 = e01 ± e23 . Λ2 T ∗ M ∼ V2 ⊕ 3V0 . D D D D D (3.r ). ω3 . so E2. J and K are acting as elements of a Lie algebra. ω3 . 2 0 3 0 0 ∗ e = J(e ) and e = K(e ). . a classic reference being [AHS]. not a Lie group.0 is the bundle of anti-self-dual 2-forms Λ2 . This mechanism also allows us to work out expressions for D and D acting on α. −→ C ∞ (E2n. . Lemma 3. Consider a k-form α. J and K all act trivially 3 on the ωj . Example 3. the bundle Ek. Chapter 7]. = ijk.0 = ω1 . ω2 .k is the bundle Ak of (2.5) − − − − Then I. .2. This also serves to illustrate why in four dimensions we make the definition that a quaternionic manifold is a self-dual conformal manifold. ± ω3 = e03 ± e12 . These bundles will be familiar to most readers: E2.2 ) −→ .2 is the bundle of self-dual 2forms Λ2 and E2. i j k Using the notation e = e ∧ e ∧ e ∧ . = + − and I 2 + J 2 + K 2 = −4(∗ + 1).. ± ω2 = e02 ± e31 . Then α ∈ Ek. .. the simplest way the author has found is to use the Casimir operator C = I 2 + J 2 + K 2 .By definition.7 Let α ∈ C ∞ (Ek. .r ).2n ) −→ 0 is precisely the complex (2. The action of + sp(1) on the ωj is given by the multiplication table + I(ω1 ) = 0 + + J(ω1 ) = −2ω3 + + K(ω1 ) = 2ω2 + + I(ω2 ) = 2ω3 + J(ω2 ) = 0 + + K(ω2 ) = −2ω1 + + I(ω3 ) = −2ω3 + + J(ω3 ) = 2ω1 + K(ω3 ) = 0.2 = + + + ω1 . (3.r if and only if (I 2 + J 2 + K 2 )(α) = −r(r + 2)α. where ∗ : Λk T ∗ M → Λ4−k T ∗ M is the Hodge star map. ω2 .

This is exactly like the real Dolbeault complex of Figure 3.1 For 2k ≥ 4. Writing Dk. .r for the particular map D : C ∞ (Ek. it is the isosceles triangle as opposed to diamond shape which causes ellipticity to fail for the bottom row. Again.0 )) to C ∞ (E2k+1.1 . . → E2n+k. . we define the quaternionic cohomology groups k. 32 D D D D D . the exterior forms a1 dz ∧ .2n−k → 0 is elliptic everywhere except at E2k. ∧ d¯bq span Λp. because there is no quaternionic version of ‘holomorphic coordinates’. where Dα ∈ Ek+1.2 → .0 and E2k+1. The rest of this section provides a proof of this Theorem. Rearranging these equations gives Dα and Dα. r+1 Proof.r+1 and Dα ∈ Ek+1. .1.3 Ellipticity and the Double Complex In this section we shall determine where our double complex is elliptic and where it is not. we can decompose Ek. ∧ dz ap ∧ d¯b1 ∧ .r ) . Applying the Casimir operator gives (I 2 + J 2 + K 2 )(dα) = −(r + 1)(r + 3)Dα − (r + 1)(r − 1)Dα. . the complex 0 → E2k.r−1 ) (3. On a quaternionic manifold M 4n there is unfortunately no easy way to write down a local frame for the bundle Ek.r ) → C ∞ (Ek+1. For k = 1 the complex is elliptic everywhere except at E3. making it much easier to write down the kernels and images of maps which involve exterior multiplication. This allows us to decompose any form z z ω ∈ Λp. the guiding principles which determine where the two double complexes are elliptic are very much the same in both cases.7) 3. where it is not elliptic. Im(Dk−1. where it is not elliptic. It turns out that we have ellipticity everywhere except on the bottom two rows of the complex.1.r+1 ).1. Here is the main result of this section: Theorem 3.r .r−1 .1.3. the analogous result for the Real Dolbeault complex. However.q . We have dα = Dα + Dα. . Note the strong similarity between this Theorem and Proposition 3. On a complex manifold M 2n with holomorphic coordinates z j . For k = 0 the complex is elliptic everywhere.1 → E2k+2. and though it is more difficult to prove for the quaternionic version.1 .and Dα = 1 4 (r + 3) + 1 (I 2 + J 2 + K 2 ) dα.1 ) is the identity.3.r HD (M ) = Ker(Dk.0 → E2k+1.0 and the projection from d(C ∞ (E2k.r just enough to enable us to prove Theorem 3. because d = D on E2k.q .

r ⊕ Ek. is exact for any nonzero e0 ∈ T ∗ Hn then it is exact for all nonzero ξ ∈ T ∗ Hn . . .r−1 . Note n n−1 0 that we can identify Ek.r . Consider.r ⊕ Ek.1. define e1 = I(e0 ).r ⊕ Ek. Secondly. the situation is of the form (pVr+1 ⊕ qVr−1 ) ⊗ 2V1 ∼ 2p(Vr+2 ⊕ Vr ) ⊕ 2q(Vr ⊕ Vr−2 ). To prove Theorem 3.r is preserved by the induced action of the hypercomplex structure on Λk Hn . where αj ∈ Λk−1 Hn−1 . since H0 ∼ 2V1 as an sp(1)= representation.r ⊂ Λk Hn according to how many differentials in the H0 -direction are present. 3. We write + − + − + − + − α = α+ + α− = (α0 + α0 )e0 + (α1 + α1 )e1 + (α2 + α2 )e2 + (α3 + α3 )e3 . αj ∈ Ek−1. its components in the representations 2pVr+2 and 2qVr−2 must both vanish separately. K) be the standard hypercomplex structure on Hn .2 Define the space Ek. for example αeij means α ∧ eij . + − 0 0 where αj ∈ Ek−1. Proof. J.r ≡ Ek.r to be the subspace of Ek.A principal observation is that since ellipticity is a local property.r ∗ Let e0 ∈ Tx Hn ∼ Hn and let (I. juxtaposition of exterior forms will denote exterior multiplication.r on H . = where α+ ∈ pVr+1 and α− ∈ qVr−1 .3.r+1 and αj ∈ Ek−1. so for this to vanish we must have 33 . we choose one such e0 and analyse the spaces Ek.3. which decomposes each = l=0 Ek.r+1 ⊕ Ek−1. obtaining a (nonnatural) ∗ splitting Tx Hn ∼ Hn−1 ⊕ H0 which is preserved by action of the hypercomplex structure.r accordingly. e2 = J(e0 ) and e3 = K(e0 ). since GL(n.r .r ∩ (Λk−l Hn−1 ⊗ Λl H0 ).r = Ek. (Throughout the rest of this section.3 If α = α+ + α− ∈ Ek. = In this way we single out a particular copy of H which we call H0 . 0 1 2 3 4 Thus we obtain an invariant decomposition Ek. Thus α is an element of Λk−1 Hn−1 ⊗ 2V1 . i. we can work on Hn without loss of generality.r−1 .2. As = in Example 3. = This induces the decomposition Λk Hn ∼ 4 Λk−l Hn−1 ⊗ Λl H0 .r . . The following Lemma allows us to consider α+ and α− separately. l Definition 3. . l Then Ek.e. e3 ∼ H. . for example. Since α is in a copy of the representation Vr .r consisting of exterior forms with precisely l differentials in the H0 -direction. l Ek.e.3.r ⊕ Ek. An exterior form α ∈ Ek. i. it follows from the isomorphism Vr ⊗ V1 ∼ Vr+1 ⊕ Vr−1 that the αj must be in a combination of Vr+1 and Vr−1 = 0 0 representations. For α to be in the representation 2(p + q)Vr .r is of the form α0 e0 + α1 e1 + α2 e2 + α3 e3 . so that e0 . . H) acts transitively on Hn \ {0}. 1 1 Lemma 3. if the σe0 σe0 σe0 symbol sequence . . the bun1 1 dle Ek.r then both α+ and α− must be in Ek. .) We can decompose these summands still further. The component in 2pVr+2 comes entirely from α+ .6.r −→ Ek+1. −→ Ek.3.r+1 −→ .1 Decomposition of the Spaces Ek. In terms of representations.r on H with Ek.

Since these are the conditions for a form to lie in a particular Lie algebra representation.r 4.r and 4. 2.r 1.r = Ek.r = Ek.r respectively.r the αj will usually have to satisfy some simultaneous equations. l. Let α ∈ Ek. 0.r and Ek.4 Define the bundle Ek...r 2. 3.r 2.r .r 4 Ek.ta + .b .r Recall the self-dual forms and anti-self-dual forms in Example 3. 2.r . 3.2 Lie in conditions We have analysed the bundle Ek. The middle isomorphism follows a = similar argument. The second isomorphism is Lemma 3. 0 1 Thus we decompose the space Ek.r = Ek. We now determine when a particular exterior form lies in one of these subbundles. l.3. l. since wedging with e0123 has no effect on the sp(1)-action.r 2. We 2.r+1 ⊗2V1 .r = Ek.r ≡ (Ek−l.r+1 1.r−1 ⊗ 0 2V1 and one from Ek−1.α+ ∈ 2(p + q)Vr independently of α− . In other words. one coming from Ek−1.r− will call these summands Ek. Consider 0 a form α = α1 es1 . That α ∈ Ek. the bottom right index r refers to the irreducible Sp(1)representation in which α lies. l.3.r+ 2.r ⊕ Ek. 34 .r .. 4.r be as above. we mention three trivial Lie in conditions.4 allows us to write down.2.3.r into a number of different subbundles.r ..r to be the subbundle of Ek.m ⊗ Λl H0 ) ∩ Ek. etc. where αj ∈ Ek−a. Similarly. the bottom left index k refers to the exterior power of the form α ∈ Λk Hn .3.r−1 1 Ek.r− 2.r ⊕ Ek.sa + α2 et1 .m To recapitulate: for the space Ek.r−2 2 Ek. as is the last (since the hypercomplex structure acts trivially on Λ4 H0 ).r ⊕ Ek.r arising from n−1 Vm -type representations in Λk−l H .m 0 Ek. the top left index l refers to the number of differentials in the H0 -direction and the top right index m refers to the irreducible Sp(1)-representation of the contributions from Λk−a Hn−1 before wedging with forms in the H0 -direction. so Ek.r 0 To begin with.r 0 Ek. .r ⊕ Ek.b spaces Ek.r is obvious. and the fourth follows in exactly the same way since Λ3 H0 ∼ 2V1 also. We have the following decompositions: 0.r . The bundle Ek.6.r−1 3 Ek.r . we must have α− ∈ 2(p + q)Vr .5 Let Ek.m l Definition 3.r 3. This may appear slightly fiddly: it becomes rather simpler when we consider the specific splittings which Definition 3.r splits according to whether its contribution from Λ2 H0 is self-dual or anti-self-dual. as is αe0123 ∈ Ek. .r = Ek. we will refer to such equations as ‘Lie in conditions’. Proof.3. defining the following notation.r ⊕ Ek.r+2 2. We extend this decomposition to the cases l = 0.3.r into two summands.r+1 3.r+ 2. For α to lie in one of the a. for the component in 2qVr−2 to vanish.r = Ek. The first isomorphism is trivial.m Lemma 3.

(3. Since the αj have no ej -factors and the action of I. 3.r entirely in Ek+1. so αωj ∈ Ek. = 4 35 .r 1. ω2 = e02 − e31 2.− − Likewise.r−1 . Then (I 2 +J 2 +K 2 )α = −(r −1)(r +1)α.8) gives the result that α ∈ Ek+1. 2.r . Thus 3 (I + J + K )α = j=0 2 2 2 (I 2 + J 2 + K 2 )(αj )ej + αj (I 2 + J 2 + K 2 )(ej ) + + 2 I(αj )I(ej ) + J(αj )J(ej ) + K(αj )K(ej ) 3 = −r(r + 2)α − 3α + 2 j=0 I(αj )I(ej ) + J(αj )J(ej ) + K(αj )K(ej ) I(α0 )e1 − I(α1 )e0 + I(α2 )e3 − I(α3 )e2 + +J(α0 )e2 − J(α1 )e3 − J(α2 )e0 + J(α3 )e1 + +K(α0 )e3 + K(α1 )e2 − K(α2 )e1 − K(α3 )e0 .r this alternative into Equation (3. we have that 2 I (αj ej ) = I 2 (αj )ej + 2I(αj )I(ej ) + αj I 2 (ej ). J and K preserves this property. and α is 1.r We conclude that α ∈ Ek+1.r−1 if and only if Our interest in these conditions arises from a consideration of exterior forms.r 0 Let αj ∈ Ek. 1. but the equations describe sp(1)-representations in general: they are the conditions that α ∈ Vr ⊗ V1 must satisfy to be in the Vr+1 subspace of Vr+1 ⊕ Vr−1 ∼ Vr ⊗ V1 . 3-forms and the self-dual 2-forms ωj .r− − − 03 12 and ω3 = e − e is trivial.r Suppose instead that α ∈ Ek+1.r+1 if and only if (I 2 + J 2 + K 2 )α = −(r + 1)(r + 3)α. Putting 1. etc. This leaves the following three situations: those arising from taking exterior products + with 1-forms.r−1 . (3.r For α ∈ Ek+1. α1 .r+1 ⊕ Ek+1. which is the case if and only if −rα = I(α0 )e1 − I(α1 )e0 + I(α2 )e3 − I(α3 )e2 + J(α0 )e2 − J(α1 )e3 − J(α2 )e0 + J(α3 )e1 + + K(α0 )e3 + K(α1 )e2 − K(α2 )e1 − K(α3 )e0 . this equation can only be satisfied if it holds for each of the ej -components separately.r+1 we need this to be equal to −(r + 1)(r + 3)α. The cases l = 1 and l = 3 1. we apply the Casimir operator. Then α = α0 e0 + α1 e1 + α2 e2 + α3 e3 ∈ Ek+1.r+1 if and only if α0 .8) = (−r2 − 2r − 3)α + 2 1. α2 and α3 satisfy the following Lie in conditions: 4 rα0 − I(α1 ) − J(α2 ) − K(α3 ) rα1 + I(α0 ) + J(α3 ) − K(α2 ) rα2 − I(α3 ) + J(α0 ) + K(α1 ) rα3 + I(α2 ) − J(α1 ) + K(α0 ) = = = = 0 0 0 0. By the usual (Leibniz) rule for a Lie algebra action on a tensor product. the sp(1)-action on the anti-self-dual 2-forms ω1 = e01 − e23 . As usual when we want to know which representation an exterior form is in.r for all j = 1. The other Lie in conditions have similar interpretations.9) 1.

r− − has no effect on the sp(1)-action. ω3 ∼ V2 ⊂ Λ2 H0 .r+2 . β2 and β3 satisfy Equation 3.13)  (2 − r)β3 = I(β2 ) − J(β1 ). We calculate these Lie in conditions in a similar fashion to the previous cases. there are r + 1 linearly independent solutions of this form (for r = 0).r−1 .7 by replacing dα with αe0 . 3.   (2 − r)β1 = J(β3 ) − K(β2 ) 2.r+2 ⇐⇒ (3. (3. K] = JK − KJ.5.r Ek+2.r−2 ⇐⇒ (3. it must have dim Vr = r + 1 linearly independent solutions.12.r+ 2β2 = K(β1 ) − I(β3 ) β ∈ Ek+2. By the Clebsch= Gordon formula. β2 = J(β0 ).r+1 we need = 3.2.r (2 − r)β2 = K(β1 ) − I(β3 ) β ∈ Ek+2. Since 3.r−1 we need the αj to satisfy Equations (3. the Lie in conditions are exactly the same: for α to be in Ek+3.r (r + 4)β2 = K(β1 ) − I(β3 ) β ∈ Ek+2.10) 3.3.3. Using the Lie algebra relations 2I = [J.3.3. Let β0 ∈ Vr and let β1 = I(β0 ). The case l = 2 − 0 We have already noted that wedging a form β ∈ Ek. Thus = + + + for β = β1 ω1 + β2 ω2 + β3 ω3 we want to establish the Lie in conditions for β to be in 2.r Consider now α = α0 e123 + α1 e032 + α2 e013 + α3 e021 ∈ Ek+3.12)  2β3 = I(β2 ) − J(β1 ).   2β1 = J(β3 ) − K(β2 ) 2.r Λ3 H0 ∼ H0 .1 We now describe the principal symbol of D. Moreover.r+1 ⊕ Ek+3.(r + 2)α0 + I(α1 ) + J(α2 ) + K(α3 ) (r + 2)α1 − I(α0 ) − J(α3 ) + K(α2 ) (r + 2)α2 + I(α3 ) − J(α0 ) − K(α1 ) (r + 2)α3 − I(α2 ) + J(α1 ) − K(α0 ) = = = = 0 0 0 0. Ek+2.r 3.r the αj to satisfy Equations (3. Since this equation singles out the Vr -representation in the direct sum Vr+2 ⊕ Vr ⊕ Vr+2 . This leads to a proof of Theorem 3. and examine its behaviour in the context of the decompositions of Definition 3.r . the decomposition takes the form Vr ⊗ V2 ∼ Vr+2 ⊕ Vr ⊕ Vr−2 .r with an anti-self-dual 2-form ωj 2. β3 = K(β0 ).11)  (r + 4)β3 = I(β2 ) − J(β1 ). Thus we only have to consider the + + + effect of wedging with the self-dual 2-forms ω1 . ω2 . and for α to be in Ek+3.10). β3 = K(β0 ).r and Ek+2.12) take the form β1 = I(β0 ).r+ 2.6).9). it is easy to see that β1 .2 and Lemma 3.r−2 . First we obtain the principal symbol from the formula for D in Lemma 3. 36 . by considering the action of the Casimir operator I 2 +J 2 +K 2 on β and using the multiplication table (3.r 2. Equation 3. so βωj ∈ Ek+2.3. The following Lie in conditions are then easy to deduce:   (r + 4)β1 = J(β3 ) − K(β2 ) 2. β2 = J(β0 ).1.r ⇐⇒ (3. We conclude that all the solutions of Equation (3.12 is particularly interesting.3 The Symbol Sequence and Proof of Theorem 3.

so the index m remains the same. σ σ σ σ σ σ σ σ σ σ (3. .r → l+1 Ek+1.r−1 −→ Ek.6 Let x ∈ Hn . e0 ∈ Tx Hn and α ∈ Ek. . e0 )(α) = − 1 1 (r − 1) + (I 2 + J 2 + K 2 ) αe0 4 r+1 −1 = ( (r − 1)(r + 1) − r(r + 2) − 3 ) αe0 4(r + 1) +2 I(α)e1 + J(α)e2 + K(α)e3 = as required. Replacing dα with αe0 in the formula for D obtained in Lemma 3.2.r to the space l+1. 1 (r + 2)αe0 − I(α)e1 − J(α)e2 − K(α)e3 .r+1 → 0 ⊕ 2. The only action in the other directions is the sp(1)-action.r 3.14) Using Lemma 3.r+1 −→ Ek+2.r+1 → Ek+2.) l The point of all this work on decomposition now becomes apparent.r+2 3.r−1 −→ Ek. l. to the 5-space sequence 2 3 4 0 1 0 −→ Ek−2. Using Lemma 3.r → 0. 2(r + 1) Proof.7 The principal symbol σD (x.6.r+1 −→ .3. which preserves the irreducible decomposition of the contribution from Λk−a Hn−1 . .r → Ek.∗ Proposition 3.r−1 2.m Corollary 3. etc. e0 ) maps the space Ek.r+2 → 0 ⊕ ⊕ 2.r+2 −→ 0.r+1 is given by σD (x.r+2 4.r → Ek−1. e0 ) for the rest of this section.3.7. (To save space we shall use σ as an abbreviation for σD (x.r+1 . we see that σD (x.r . e ) : Ek. we can reduce the (somewhat indefinite) symbol sequence .3. We already know that σD : Ek. obtaining three short sequences (for k ≥ 2.r → Ek+1. we can analyse this sequence still further according to the different (top right) m-indices. (3.r+1 .r−2 1.r → Ek+1.r −→ Ek+1.r+2 → Ek.r−2 −→ Ek−1.r−2 → Ek−1.m Ek+1.15) 37 .r → Ek+1. −→ Ek−1. . e0 ) increases the number of differentials in the H0 -direction by one.5 as well. we have σD (x.r−2 0 → Ek−2.r → Ek+1.r−1 ⊕ 1. by definition. Since σ : Ek. The principal symbol 0 mapping σD (x.r −→ Ek+1.r+1 . so the index l increases by one.3. k ≡ r mod 2) 0 0 0. 2(r + 1) Proof. e0 )(α) = 1 (r + 2)αe0 − I(α)e1 − J(α)e2 − K(α)e3 .r−2 → Ek.

2 which cannot be exact.r The case a = 2 is slightly more complicated. Here the bottom sequence of (3.r . It is clear that dim Ek.m Proof. for the middle sequence n−1 k−2.r = (r + 1) k−2. Since the isomorphism n−1 V0 ⊗V2 ∼ n−1 V2 gives no trivial V0 -representations.r 2.r = (r + 1) k−2. Finally. we have dim Ek−1. l.2. .r dim Ek+1.3.r n−1 giving Ek. 3. Lemma 3.r+2 k−2. B and C it is exact at the third.This reduces the problem of determining where the operator D is elliptic to the problem of determining when these three sequences are exact.r has dimension (r +1) n−1 and the space Ek.2.r+2 3.r−2 and 4.r−2 1.r ⊗ 2V1 .r has dimension 3(r +1) n−1 .r−2 (r − 1 − 2r + r + 1) = 0.r on Hn−1 . which now follows from: 38 . The case r = 0 is different. k.r−1 = 2r k−2.0+ Ek.r+2 2.1 → . For a sequence 0 → A → B → C → 0 to be exact.r−2 = (r − 1) k−2.3. Let r > 0. etc.r+1 = 2(r + 2) n−1 and dim Ek+1.r+ 2. k−2. and for the bottom sequence n−1 k−2. . each of the sequences in (3.0 → E1.r n−1 n−1 For a = 1. Recall 0 the notation Ek.15) satisfies the dimension condition above.0 . Thus dim Ek−2. We shall show that for r = 0 this dimension sum does equal zero.0 V0 −→ 2 n−1 k−2.r+2 (r + 1 − 2(r + 2) + r + 3) = 0. if the sequence is exact at any two out of A.r a total dimension of 4(r + 1) k−2.0 is ‘too small’ — we are left with a sequence 0 −→ 3 n−1 k−2.r+2 and dim Ek.r = n Vr from Definition 3. as we have to take into account exterior products with the self-dual 2-forms V2 and anti-self-dual 2-forms 3V0 in Λ2 H0 .e. i. . 4. Thus Ek.15) n−1 k−2.r k−2.r−2 Ek.r on Hn is simply Ek.r−2 .r−1 = 2r k−2.r+2 2. the alternating sum of the dimensions vanishes.r−2 n−1 0 since Ek.0 V1 −→ 0. . k−2. k−2.0 k−2.r− space Ek.8 For r > 0.0 2.r+2 dim Ek+2.r for l = 0.0 2. .15) disappears altogether.r+2 = (r + 3) n−1 . It is now a simple matter to verify that for the top sequence of (3. (As there is no space E0. Given this condition. For a = 3.r+1 = 2(r + 2) n−1 . from which we 2.r .r k. We calculate the dimensions of the spaces Ek. The spaces 2.r−2 and dim Ek−1. 1. it is necessary that dim A − dim B + dim C = 0. .r−2 n−1 n−1 infer that dim Ek. the 2.r receive contributions only from the self-dual part V2 .r = (r + 1) n−1 .r (2r − 4(r + 1) + 2(r + 2)) = 0. there is no space = k−2. the top sequence still being exact.1.r and Ek.r 0. Exactness is lost in the middle sequence.r+2 0 The cases a = 1 and a = 3 are easy to work out since they are of the form Ek. this problem does not arise for the leading edge 0 → E0.) We are finally in a position to prove Theorem 3.0 .

r+2 −→ 0.r −→ Ek+1.r−2 1. we show that the middle sequence 0 −→ Ek−1.9 When r = 0.r+2 3.r+1 −→ Ek+2.r . Then σ(α) = 1 rαe0 − I(α)e1 − J(α)e2 − K(α)e3 . Recall the Lie in condition (3. A calculation using Proposition 3.r−2 Ek−1.r−2 Since these are linearly independent. Since the αi have no ej -components.r−1 or Ek+2.r+1 is injective and the 3. Restricting to 1. 0 Consider α ∈ Ek−2.r−1 . in which case it is clear that σ(α) = 0 ⇐⇒ α = σ 2(r+1) α0 r .r → Ek+1.r .r −→ Ek+1.r of Ek.15).r−1 and Ek−1. σ 2.r−2 → σ σ 1.r−1 .r+2 The Clebsch-Gordon formula shows that there are no spaces Ek+1.r+2 4. It is easy to check using 3.15) are exact.r is of the form β+γ = 1 − + + + − − I(β0 )ω1 + J(β0 )ω2 + K(β0 )ω3 + γ1 ω1 + γ2 ω2 + γ3 ω3 .16) shows that   (r + 2)β0 + I(γ1 ) + J(γ2 ) + K(γ3 ) =   (r + 2)γ1 − I(β0 ) − J(γ3 ) + K(γ2 ) = σ(β + γ) = 0 ⇐⇒  (r + 2)γ2 + I(γ3 ) − J(β0 ) − K(γ1 ) =   (r + 2)γ3 − I(γ2 ) + J(γ1 ) − K(β0 ) = 0 0 0 0.r+2 2. γj ∈ Ek−2.Proposition 3. 2.3. Consider first the top sequence 0 −→ Ek. 1 1 To show exactness at Ek−1.r−2 .r Finally.r+1 .r−2 Ek−1.r−1 −→ Ek. Hence the bottom sequence 0 −→ Ek−2. 0 1 2 This shows that the sequence Ek−2.12) that β + + + 0 must take the form β = 1 I(β0 )ω1 + J(β0 )ω2 + K(β0 )ω3 for some β0 ∈ Ek−2.3. α3 = − 1 K(α0 ) (since r r r as remarked in Section 3.).r 1. and σ : Ek−2.r+2 is surjective.r and Ek+1.r−2 −→ Ek−1.r for β0 .r+1 → Ek+2.r σ 3.r−1 . 2.r . This occurs if and only if α1 = − 1 I(α0 ).3.r is exact.6 shows that σ σ σ σ(α) = 1 (rα1 + I(α0 ))e10 + (rα2 + J(α0 ))e20 + (rα3 + K(α0 ))e30 + (3.r−1 → Ek. the three sequences of (3.r+2 4.r+2 3.r+2 D = 0 on Ek.r−2 Ek.16) 2r + (2α1 − J(α3 ) + K(α2 ))e32 + (2α2 − K(α1 ) + I(α3 ))e13 + (2α3 − I(α2 ) + K(α1 ))e21 .r+2 Proof. so D = d for the top sequence.r+2 3. (The r 1 -factor makes no difference here and is useful for cancellations. α2 = − 1 J(α0 ).r+ + + + Let β = β1 ω1 + β2 ω2 + β3 ω3 ∈ Ek.r−2 → Ek−1. we see that exactness holds at these spaces in the middle and bottom sequences respectively of (3. r 0. 39 .) Thus a general element r 2. consider α = α0 e0 + α1 e1 + α2 e2 + α3 e3 ∈ Ek−1.r .r+1 −→ 2. which is now sufficient to show that the sequence is exact. 4.r+2 map ∧e0 : Ek+1.1 these equations also guarantee that 2α1 −J(α3 )+K(α2 ) = 0 etc. Thus 2. σ(α) = 0 if and only if all these components vanish. σ(α) = 0 if and only if α = 0.r+2 the relevant Lie in conditions that the map ∧e0 : Ek.r −→ 0 is exact. A similar calculation to that of (3.r−1 −→ 2. 2(r − 1) 0.r−2 0.r .r 1.r−1 is injective.r 0 is exact at Ek.

0 → E3. so σ : Ek. consider α ∈ Ek−2. Under this diagonal action the tensor product V1 ⊗ the representation H ⊗ Ek. Then 3. Equation (2.0 −→ E1. This amounts to applying the operators I : α → I(α) − αi1 . This concludes our proof of Theorem 3. and to show that the sequence E0. Consider.0 . for example. q) → qp−1 .0 αe0123 ∈ Ek.3.r−1 .10) which we need for β0 e0 + γ1 e1 + γ2 e2 + γ3 e3 1. We can now use our globally defined hypercomplex structure to combine the Sp(1)actions on H and Λk T ∗ M .0 and σ(αe0123 ) = 0.But this is exactly the Lie in condition (3. and the right-hand copy of V1 as the right-action (p.1 is not injective.r representation H ⊗ Ek.1.1 −→ E2.0 → Ek+1. which is exactly the same as saying that the symbol sequence is not exact at Ek.1 and αe123 ∧ e0 ∈ Ek+2. 0 As a counterexample for the case r = 1 and k ≥ 2. it is easy to see that this counterexample σ σ does not arise when k = 0.17) = where we can interpret the left-hand copy of V1 as the left-action (p.1 are injective. giving (3. in which case we have β + γ = σ 2(β0 e0 + γ1 e1 + γ2 e2 + γ3 e3 ) . (3.2 is exact at E1.r 40 splits.r . Again.r ∼ V1 ⊗ V1 ⊗ n Vr . 2. Thus we have globally defined operators which generate the sp(1)-action on Λk T ∗ M .1 . This motivates a thorough investigation of quaternion-valued forms on hypercomplex manifolds.0 → E1. and to show that the maps σ : E0. Consider also the quaternions themselves.r Leaving the left H-action untouched.r = n Vr . Thus the symbol sequence fails to be exact at Ek+1. It is easy to see that this counterexample does not arise when k = 0 or 2.r to be in Ek−1. Then M has a triple (I. Thus σ(αe123 ) = 0.r and so the middle sequence is exact. K) of complex structures which we can identify globally with the imaginary quaternions. Then 4. The Sp(1)-action on these forms is described by the k.1 . = k.1 and σ : E2. 3.0 . In the case n = 1 this reduces to the representation H ∼ V1 ⊗ V1 . H)-representation on Hn as V1 ⊗ E.4 Quaternion-valued forms on Hypercomplex Manifolds Let M be a hypercomplex manifold. = k. Since αe123 has no e0 2 components at all it is clear that αe123 = σ(β) for any β ∈ Ek.0 .r ∼ V1 ⊗ n (Vr+1 ⊕ Vr−1 ). quaternion-valued exterior forms in the bundle Ek.r Vr to α ∈ H ⊗ Ek. q) → pq. consider α ∈ Ek−4. 0 As a counterexample for the case r = 0 and k ≥ 4.18) .0 . J : α → J(α) − αi2 and K : α → K(α) − αi3 n k. J.r This demonstrates exactness at Ek.6) describes the Sp(1)GL(n. we consider the effect of the right H-action and the hypercomplex structure simultaneously.0 4 αe123 ∈ Ek+1.0 . for q ∈ H and p ∈ Sp(1).

lead to quaternionic analogues of holomorphic functions and k-forms.18). The algebraic foundation for this geometry lies in considering objects like our left H-modules in Equation (3.Each of these summands inherits the structure of a left H-module from the left-action V1 .r H-linear submodule of Hn . Each of the summands V1 ⊗ n Vr±1 is a left H-module k. It will. This point of view turns out to be very fruitful. over the next few chapters. There is a decomposition of real-valued forms. which is taken further when we consider complex-valued forms. the holomorphic tangent and cotangent spaces.r which arises as a submodule of H ⊗ Ek. Thus each summand is an = k. and complex Lie groups and Lie algebras. which is not affected by our splitting. 41 .r ∼ (r + 1) n Hn . This situation mirrors our discussion of real and complex forms on complex manifolds. In the next chapter we will introduce a new algebraic theory which is based upon such objects. In the same way. considering quaternion-valued forms on a hypercomplex manifold allows us to take our decomposition further.

4. Dominic Joyce.) For two AH-modules U ⊂ Hm and V ⊂ Hn . For our purposes. all H-modules will 42 . for example it is both associative and commutative. The pair (U. 4. The interested reader should consult [J1] for more details and proofs. The basic objects of study are H-submodules U of H ⊗ Rn . we will see that the most important classes of AH-modules are conveniently described and manipulated using Sp(1)-representations. or AH-module. There are other algebraic operations which are equivalent to Joyce’s quaternionic tensor product.Chapter 4 Developments in Quaternionic Algebra This chapter describes an algebraic theory which will be central to our description of hypercomplex geometry. The theory is that of my supervisor.1 AH-Modules We begin by defining H-modules and their dual spaces. U ) is an augmented H-module. and is presented in [J1]. In the next chapter. It has similar properties to the tensor product over a commuting field. The operation ‘ ⊗H ’ will be called the quaternionic tensor product. we can define a unique AH-module U ⊗H V ⊂ Hmn . u) → q · u or qu. This allows us to develop the algebra of AH-modules as a parallel to that of vector spaces over R or C.3 that the most obvious definitions of a tensor product over the quaternions are not especially fruitful. A sheaf-theoretic point of view is presented by Quillen [Q]. such that p(q(u)) = (pq)(u) for p.1 The Quaternionic Algebra of Joyce The following is a summary of parts of Joyce’s theory of quaternionic algebra. q ∈ H and u ∈ U . This allows us to classify all AH-modules and determine their tensor products. This analogy is particularly strong for certain well-behaved AH-modules which will be called stable AH-modules. in which he discovers a contravariant equivalence of tensor categories between AH-modules and regular sheaves on a real form of CP 1 . The most important discovery in [J1] is a canonical tensor product for AH-modules with interesting properties. A (left) H-module is a real vector space U with an action of H on the left which we write as (q.1. Joyce shows that the inclusion ιU : U → Hn is determined up to isomorphism by the Hmodule structure of U and the choice of a real vector subspace U ⊂ U satisfying a certain condition. (Recall from Section 1.

If φ is also an isomorphism of H-modules we say φ is an AH-isomorphism. If we consider only the real part Re(α(u)) (for u ∈ U and α ∈ U × ). Let U be a real vector subspace of U . as the condition in Definition 4. U † determines U (at least for finite-dimensional U ) by U = {u ∈ U : α(u) ∈ I for all α ∈ U † }. We would like to define (W.3 Let U. ν induces an isomorphism (Y † )∗ ∼ I ∼ V2 . or AH-module. Since Y /Y ∼ (Y † )∗ . = = = Here is the natural concept of linear map between AH-modules: Definition 4. assuming that U is also given.1 that u = v. Then since α is a linear map we have α(u − v) = 0 for all α ∈ U † and it follows from Definition 4. Definition 2. Dual H-modules behave just like dual vector spaces. We write U ∗ for the dual vector space of U . Define a real vector subspace U † of U × by U † = {α ∈ U × : α(u) ∈ I for all u ∈ U }. with H = I.1. its H-linear span H · u should not be entirely contained in U . so U is an AH-module. Definition 4.1 may not be satisfied. there is a catch: W may not be an AH-module. AHmodules should be thought of as the quaternionic analogues of real vector spaces.be left H-modules. We say that φ is an AH-morphism if φ(U ) ⊂ V . = so Y = {(q1 . the definition of an AHmodule demands that U should not be too large. 43 . Usually we will refer to U itself as an AH-module. Example 4. However. In effect. Definition 4. q3 ) = i1 q1 + i2 q2 + i3 q3 . By dim U we will always mean the dimension of U as a real vector space. V be AH-modules and let φ : U → V be H-linear. Then q · α ∈ U × .1] Let Y ⊂ H3 be the set Y = {(q1 . If U is an H-module we also define the dual H-module U × of linear maps α : U → H that satisfy α(qu) = qα(u) for all q ∈ H and u ∈ U .2) (4. Define W to be the quotient H-module V /U and define W to be the real subspace (V + U )/U of W . We consider H to be an AH-module. = Let U be an AH-module and let u.1) An augmented H-module.1 demands for any u ∈ U . Let ν : Y → H. Conversely. define q · α by (q · α)(u) = α(u)q for u ∈ U .1. Thus if U is finite-dimensional. ν(q1 .2] Let U be an H-module. v ∈ U such that α(u) = α(v) for all α ∈ U † . Then im(ν) = I and ker(ν) = Y . Definition 6. q2 . and U × is a (left) H-module. q2 . we say that U is an AH-submodule of V if U is an H-submodule of V and U = U ∩ V . we can interpret U × as the dual of U as a real vector space. W ) to be the quotient AH-module V /U .2 [J1.1 [J1. if α(u) = 0 for all α ∈ U † then u = 0. and then U † is the annihilator of U . so dim(U ∩ H · u) ≤ 3. As U † is the restriction of V † to U . (4. is a pair (U. Then Y ∼ H2 is a left H-module.1. q3 ) : q1 i1 + q2 i2 + q3 i3 = 0}.1. dim U + dim U † = dim U = dim U × and an isomorphism U ∼ U × = determines an isomorphism U/U ∼ U † . q2 . Thus U is an AH-module if and only if each u ∈ U is uniquely determined by the values of α(u) for α ∈ U † . U ) such that if u ∈ U and α(u) = 0 for all α ∈ U † .1. Then dim Y = 8 and dim Y = 5.1. even if U is an H-module. q3 ) : qj ∈ I and q1 i1 + q2 i2 + q3 i3 = 0}. If q ∈ H and α ∈ U × . If V is an AH-module. Define a real subspace Y = Y ∩ I3 . then u = 0.

so that ιU (U ) is an H-submodule of H ⊗ (U † )∗ . U ) is determined by the H-submodule ιU (U ).2). V be AH-modules. Suppose u ∈ ker ιU . Then ιU is H-linear. then ψ ◦ φ : U → W is an AH-morphism.1. We will = return to this question in some detail later. Thus ιU is injective. with H-action p · (q ⊗ x ⊗ y) = (pq) ⊗ x ⊗ y. V be AH-modules and φ : U → V an AH-morphism. Then they can be regarded as subspaces of H ⊗ (U † )∗ and H ⊗ (V † )∗ respectively. Then H ⊗ (U † )∗ ⊗ (V † )∗ is an H-module. = • From Equation (4. Here is the key idea of the theory: Definition 4. Exchanging the factors of H and (U † )∗ .2 The Quaternionic Tensor Product Let U and V be AH-modules. The operation ⊗H will be called the quaternionic tensor product. Then φ(U ) ⊂ V implies that φ× (V † ) ⊂ U † . we can paste these AH-modules together to get a product AH-module. We define their intersection to be the quaternionic tensor product of U and V . it follows that ιU (U ) = ιU (U ) ∩ (I ⊗ (U † )∗ ).1. I ⊗ Rn ) for n = dim(U † )∗ . Example 4. U ⊗H V = (ιU (U ) ⊗ (V † )∗ ) ∩ ((U † )∗ ⊗ ιV (V )) ⊂ H ⊗ (U † )∗ ⊗ (V † )∗ .2 shows how ths works for the AH-module Y . with H-action p·(q⊗x) = (pq) ⊗ x.4 [J1. Real and complex vector spaces are classified by dimension. Define a map ιU : U → H ⊗ (U † )∗ by ιU (u) · α = α(u). As an abstract AH-module Y is isomorphic to H2 and (Y † )∗ ∼ V2 . as there are several choices of U for each H-module U ∼ Hn . Thus the AH-module (U. and ιU (U ) ∼ U . • Let U. Then H⊗(U † )∗ is an H-module.3) The vector subspace (U ⊗H V ) is then given by (U ⊗H V ) = (U ⊗H V )∩(I⊗(U † )∗ ⊗(V † )∗ ) and with this definition U ⊗H V is an AH-module.2] Let U. • Let U be an AH-module. We note the following points: • If φ : U → V and ψ : V → W are AH-morphisms. we may regard (U † )∗ ⊗ ιV (V ) as a subspace of H ⊗ (U † )∗ ⊗ (V † )∗ . so that u = 0 as U is an AH-module.One obvious question is whether we can classify AH-modules up to AH-isomorphism. Thus ιU (U ) ⊗ (V † )∗ and (U † )∗ ⊗ ιV (V ) are AH-submodules of H ⊗ (U † )∗ ⊗ (V † )∗ . We will find this version of events very useful in many situations. Definition 4. (4. Since the H-action on both of these is the same. but clearly this is not true for AH-modules.1. for u ∈ U and α ∈ U † . Define an H-linear map φ× : V × → U × by φ× (β)(u) = β(φ(u)) for β ∈ V × and u ∈ U . One = of the easiest and most symmetrical ways to obtain Y is as an 8-dimensional subspace of H3 = H ⊗ V2 . 4. 44 . This shows as promised that every AH-module is isomorphic to a (left) submodule of (H ⊗ Rn . Then α(u) = 0 for all α ∈ U † . as we shall see immediately.

but that this is = not really any different from taking tensor products over R.A few words of explanation may be useful at this point. we do have the following special case: Lemma 4.1. Then φ× (W † ) ⊂ U † and ψ × (X † ) ⊂ V † . This is the quaternionic tensor product of φ and ψ. and let φ : U → W and ψ : V → X be AH-morphisms.6] Let U. the dimension of U ⊗H V can behave strangely. Then the k th symmetric group Sk acts H k k U by permutation of the U factors in the obvious way. however.6 [J1. Taking the duals gives maps (φ× )∗ : (U † )∗ → (W † )∗ and (ψ × )∗ : (V † )∗ → (X † )∗ . Define SH U and Λk U on H H to be the AH-submodules of k U which are symmetric and antisymmetric respectively H under the action of Sk . Then φ⊗H ψ is an AH-morphism from U ⊗H V to W ⊗H X . 4. Much of the algebra that works over R or C can be adapted to work over H.5) This tells us that ⊗H is commutative and associative. and let u ∈ U and v ∈ V be nonzero. (4. Suppose that α(u)β(v) = β(v)α(u) ∈ H for every α ∈ U † and β ∈ V † . If φ and ψ are both injective AH-morphisms. For example.3] that there are canonical AH-isomorphisms H⊗H U ∼ U.7 [J1. = U ⊗H V ∼ V ⊗H U = and (U ⊗H V )⊗H W ∼ U ⊗H (V ⊗H W ).4) Define φ⊗H ψ : U ⊗H V → W ⊗H X to be the restriction of id ⊗(φ× )∗ ⊗ (ψ × )∗ to U ⊗H V . Lemma 7. Given u ∈ U and v ∈ V it is not possible in general to define an element u⊗H v ∈ U ⊗H V . many situations where the quaternionic tensor product behaves differently from the standard real or complex tensor product. V be AH-modules.4] Let U be an AH-module. and that the AH-module H acts as an identity element for ⊗H . There are.5 Let U. with 0 U = H.1. At the end of Chapter 1 we saw that it is possible to define a sort of tensor product Hm ⊗H Hn ∼ Hmn . W. How U behaves in relation to the H-action determines a particular subspace ιU (U ) of H ⊗ Rn . Lemma 4. The theory of AH-modules and the quaternionic tensor product is a way of taking the quaternionic behaviour into account. It can be proved [J1.1. Combining these. It can vary discontinuously under smooth variations of U or V . = (4.4] that φ⊗H ψ is also injective. if φ and ψ are both surjective then φ⊗H ψ is not necessarily surjective. and it is possible to have U ⊗H V = {0} when both U and V are non-zero. Since ⊗H is commutative and associative we can define symmetric and antisymmetric products of AH-modules: k Definition 4. The quaternionic tensor product is the natural way of combining these choices for AH-modules U ⊆ H ⊗ Rm and V ⊆ H ⊗ Rn into an AH-module U ⊗H V ⊆ H ⊗ Rmn . we have a map id ⊗(φ× )∗ ⊗ (ψ × )∗ : H ⊗ (U † )∗ ⊗ (V † )∗ → H ⊗ (W † )∗ ⊗ (X † )∗ . Define an 45 . it is possible to prove [J1. However. 4. V. Write H U for the product U ⊗H · · · ⊗H U of k copies of U . using AH-modules and the quaternionic tensor product instead of vector spaces and the real or complex tensor product. However. X be AH-modules. We also define the tensor product of two AH-morphisms: Definition 4.

and the element u⊗H v is just the complex tensor product ιU (u) ⊗Cq ιV (v) ∈ Cq ⊗ (U † )∗ ⊗ (V † )∗ . There is thus a distinct AH-module Xq given by each pair of antipodal points {q. §8] Let U be a finite-dimensional AH-module.1 by demanding that U should not be too small. We say that U is stable if U = U + qU for all q ∈ S 2 .11 Let q ∈ I \ {0}. Xq = {p ∈ H : pq = −qp}. So Lemma 4.1. the quaternionic tensor product is the same as the complex tensor product. We can describe semistable AH-modules by the following property: Lemma 4. for integers j.7 tells us that on complex subfields of H. If α(u)β(v) = β(v)α(u) ∈ H for every α ∈ U † and β ∈ V † . It is easy to visualise how this Lemma ‘works’.8 [J1. r. In effect. Then U + qU = U for generic q ∈ S 2 .1. This is the same as saying that ιU (u) ∈ Cq ⊗ (U † )∗ and ιV (v) ∈ Cq ⊗ (V † )∗ . Definition 4. which we shall call semistable and stable.3 Stable and Semistable AH-Modules In this section we define two special sorts of AH-modules. We quote the following results. 46 .1. Xp and Xq are not AH-isomorphic to one another. mainly taken from [J1.1. but not stable. Then u⊗H v is a nonzero element of U ⊗H V . Define an AH-module Xq by Xq = H. The next logical step is to require this property for all q ∈ S 2 . 4. We say that U is semistable if it is generated over H by the subspaces U ∩ qU for q ∈ S 2 .10 Let U be a finite-dimensional AH-module. then α(u) and β(v) must both be in some commutative subfield Cq ⊂ H.element u⊗H v of H ⊗ (U † )∗ ⊗ (V † )∗ by (u⊗H v) · (α ⊗ β) = α(u)β(v) ∈ H.9 Suppose that U is semistable. but for p = λq. Xq is the subspace of H which is perpendicular to Cq with respect to the standard scalar product.1. motivating the following definition: Definition 4. These AH-modules behave particularly well. our definitions of stable and semistable AH-modules act as a balance to Definition 4. and we can exploit their ‘nice’ properties to cement further the analogy between real and quaternionic algebra. Many of the properties of semistable and stable AH-modules can be characterised by exploring the properties of a particularly important type of AH-module: Definition 4. Thus r ≥ 0. −q} for q ∈ S 2 .1. • Xq = Xλq for all λ ∈ R \ {0}. with dim U = 4j and dim U = 2j + r.1. §8]: • Xq is semistable. In other words.

so using the associativity of the quaternionic tensor we infer that U ⊗H V is a stable AH-module. due to the following Proposition: Proposition 4. = • There is an isomorphism (U ⊗H Xq ) ∼ U ∩ qU ∼ Cn . The AH-module Xq is an important bridge from quaternionic to complex algebra.13 [J1.1. dim U = 2j + r. The benefits of working with stable and semistable AH-modules become increasingly apparent as one becomes more familiar with the theory. 9.12 [J1. but if p = λq then Xp ⊗H Xq = = {0}. (U. Xq = Cq ⊕ C⊥ . so it turns the AH-module structure on U which is quaternionic information into a set of what are effectively complex vector spaces. It is clear that all stable AH-modules are semistable.1. Then χq and id ⊗H χq : U ⊗H Xq → U ⊗H H ∼ U are injective AH-morphisms. where l = js + rk − rs and t = rs .1. (4. It follows that U ⊗H Xq ∼ = = q = nXq . then U ⊗H Xq ∼ nXq with n ≥ r .8]: Let V be a finite-dimensional AH-module. For now.1. 8. dim V = 4k and dim V = 2k + s.14 [J1.1] Let U and V be stable AH-modules with dim U = 4j. we will quote the following theorems: Theorem 4. For generic subspaces U . (U. Let U = Hj and let U be a real vector subspace of U with dim U = 2j + r. = • It follows that if U and V are stable AH-modules then U ⊗H V ⊗H Xq ∼ U ⊗H (sXq ) ∼ = = rsXq for all q ∈ S 2 .9] Let j. In effect. • Let χq : Xq → H be the identity map on H. Then V is semistable if and only if V ∼ U ⊕ ( l Xqi ). 47 . where U is stable and qi ∈ S 2 . = • An AH-module U is stable if and only if U ⊗H Xq ∼ rXq for all q ∈ S 2 . by Lemma 4. 8.• There is a canonical AH-isomorphism Xq ⊗H Xq ∼ Xq . U ) is a semistable AH-module. There is a sense in which the Xq ’s are the ‘only’ class of AH-modules which are semistable but not stable. = • If U is semistable then U ⊗H Xq ∼ rXq for generic q ∈ S 2 . and the operation ‘⊗H Xq ’ converts an AH-module U into copies q of Xq . • Therefore if q ∈ S 2 and U is an AH-module with dim U = 4j and dim U = 2j + r. r be integers with 0 ≤ r ≤ j.6) Then U ⊗H V is a stable AH-module with dim(U ⊗H V ) = 4l and dim(U ⊗H V ) = 2l + t.9. U ) is a stable AH-module. If r > 0 then for generic subspaces U . = i=1 Joyce also shows that generic AH-modules with appropriate dimensions are stable or semistable: Lemma 4.

48 . Theorem 4. The rest of Joyce’s proof consists of showing that if U and V are stable then this intersection is transverse and finding (U ⊗H V ) .12 shows that the virtual dimension of U ⊗H V is the product of the virtual dimensions of U and V . Then Proposition 4. with dim U = 4j and dim U = 2j + r. Then SH U and Λn U are stable AHH n n modules.Proof.2 Duality in Quaternionic Algebra The objects which are naturally dual to AH-modules are called SH-modules. We end this section by quoting the following result: Proposition 4. We define r to be the virtual dimension of U . in which case dim(U ⊗H V ) = dim A + dim B − dim(H ⊗ (U † )∗ ⊗ (V † )∗ ) = 4(js + rk − rs).1. with dim(SH U ) = 4k. If dim U = 4j and dim U = 2j + r. n−1 n t= r .1. n 4. s ≥ 0 then dim A + dim B ≥ dim(H ⊗ (U † )∗ ⊗ (V † )∗ ).14 and Proposition 4.1. Then dim A = 4j(2k − s) and dim B = 4k(2j − r). So dim(H ⊗ (U † )∗ ⊗ (V † )∗ ) = 4(2j − r)(2k − s). Then U ⊗H V is Thus both stable and semistable AH-modules form subcategories of the tensor category of AH-modules.1. Let A = ιU (U ) ⊗ (V † )∗ and B = (U † )∗ ⊗ ιV (V ). then dim(U † )∗ = 2j − r and similarly dim(V † )∗ = 2k − s. and so if the subspaces A and B are suitably transverse in H ⊗ (U † )∗ ⊗ (V † )∗ we expect that A + B = H ⊗ (U † )∗ ⊗ (V † )∗ .(Sketch of formula for total dimension) The proof works along the following lines. if r. dim(SH U ) = 2k + s. dim(Λn U ) = 4l and dim(Λn U ) = H H 2l + t. Under certain circumstances an SH-module can also be regarded as an AH-module.15 [J1.1. Now. closed under direct and tensor products. n l = (j−r) r−1 r + . these dimension formulae still hold if V is only semistable. In fact.16 [J1. 9. 9. so that U ⊗H V = A ∩ B. n−1 n s= r+n−1 . semistable. V be semistable AH-modules. Let U be a stable AH-module. where k = (j−r) r+n−1 r+n−1 + . with dim U = 4j and n dim U = 2j + r. In this case we obtain interesting algebraic results which use dual AH-modules to tell us about AH-morphisms between AH-modules. from which it is easy to see that U ⊗H V is stable.6] Let U be a stable AH-module. Let n be a positive integer. and dim(A + B) = dim A + dim B − dim(A ∩ B).3] Let U.12 combine to give the following: Corollary 4.

Then u ∈ U is completely determined by the values of α(u) for α ∈ U † . U ) is an AH-module then (U × . where (U. each β ∈ U × must be an H-linear combination of elements of U † .2.4. we see that = U + qU = U ⇐⇒ U † ∩ qU † = {0}. 49 . U ) is an AH-module by Definition 4. We define morphisms and quaternionic tensor products for SH-modules. as a link between sheaves and AH-modules. where (U )⊥ is the subspace orthogonal to U with respect to a (hyperhermitian) metric on U . then (U. Q† ) is a strengthened H-module or SH-module if Q is generated over H by Q† .2. V † ) are SH-modules and the dual H-morphism φ× : V × → U × satisfies φ× (V † ) ⊆ U † . R). then (U × . Let (U. and if u is determined then so is β(u) for all β ∈ U × . Just as we sometimes write U for the AH-module (U.1 Let Q be a left H-module and Q† a real linear subspace of Q. Since the only other structure present is the H-action. For example.1. We write U × ⊗H V × for the quaternionic tensor product of two SH-modules. U ). SH-modules are introduced by Quillen in [Q]. I) or the SH-module (H.7) Another useful example is given by stable and semistable AH-modules. Since there is an SH-isomorphism (U.1 SH-modules In this section we will describe the class of objects which are dual to AH-modules. U † ) is an SH-module. V ) are AH-modules and φ : U → V is an AH-morphism. A further link between these two ideas is provided by choosing a (hyperhermitian) metric on the AH-module (U. Being very much part of quaternionic algebra rather than sheaf theory. we will often write U × for the SH-module (U × . so U × is generated over H by U † . U ) be an AH-module. and these definitions are exactly equivalent. if (U. So for all β ∈ U . We see that choosing a metric gives us a decomposition U ∼ = U ⊕ (U † )∗ . by taking the definitions from their corresponding AH-modules: the whole theory works in exactly the same way. U ) and (V. U ). These will be called strengthened H-modules. the SH-module corresponding to U . The converse is clearly true as well — if every β ∈ U × is of the form q · α for some α ∈ U † . But U + qU = U ⇐⇒ (U )⊥ ∩ q(U )⊥ = {0}. The simplest example is that of the quaternions themselves: we can regard them as the AH-module (H. Every AH-module can thus be regarded as an SH-module — the point of view depends on whether we think of U or (U )⊥ as the ‘special’ subspace. This = in turn identifies U † with (U † )∗ . and so realises (U † )∗ as a subspace of U which is perpendicular to U . Definition 4. or SH-modules. An AHmodule is stable (respectively semistable) if and only if it satisfies the identity U = U + qU for all (respectively for generic) q ∈ I. U † ) and (V × . The natural dual to an AH-module is thus an H-module equipped with a generating real subspace. which makes φ× an SH-morphism. (U )⊥ ) ∼ (U × . β(u) is determined by the action of U † on u. in other words we define U × ⊗H V × = (U ⊗H V )× . We say that the pair (Q. we discuss them in this section.1. (U † )∗ ) is an SH-module. U † ). (4. If U = (U. giving an H-module isomorphism U ∼ U × . U † ).

U ) is also an SH-module. But an H-linear map α : U → H such that α(U ) ⊆ I is precisely an AH-morphism from U into H.1 we showed that (U. if (U. If V is a vector space over the commutative field F then we define the dual space V ∗ to be the space of F-linear maps φ : V → F.2 Dual AH-modules In this section we take a new step and ask what happens if we consider (U × . Example 4. U † ) is an AH-module if and only if the AH-module (U. Unless otherwise stated.2 An SH-module is called stable (respectively semistable) if and only if it has the property that U † ∩ qU † = {0} for all (respectively for generic) q ∈ I. U ) is both an AH-module and an SH-module. There are immediate attractions to this approach. Once we also have a real subspace U ⊂ U we define U † to be the set of maps U † = {α ∈ U × : α(u) ∈ I for all u ∈ U }.2. U ) is also an SH-module. Definition 4. Proof. U † ) to be the dual AH-module of (U. H) or (H. namely the H-linear maps and AH-morphisms respectively. This is sometimes easier to demonstrate than the property for the corresponding AHmodules. There is an obvious possible catch: (U × .3 The AH-module (U.4 Let U be an H-module and U a real subspace of U . In the same way. or SAH-module. since their theories are interchangeable. an SH-module if and only if it has no submodule isomorphic to (H. We have in fact already done this in Section 4. U † ) consists of two sets of maps φ : U → H. We shall work happily with either AH-modules or SH-modules according to the needs of each situation. This suggests that defining (U × . {0}). Lemma 4.1. when we refer to properties of an SAH-module such as stability.5 Any semistable AH-module U has the property that U = U + qU for generic q ∈ I. U ) is an AH-module if and only if it has no submodule isomorphic to (H. U ) is a strengthened augmented H-module. We simply reverse this argument: (U × . so U is also an SH-module and so an SAH-module. we need to discern which AH-modules have well-defined duals. 4. H) .2. Thus the space (U × . U ) is an AH-module if and only if (U × . U ) could be a good quaternionic analogue of the concept of a dual vector space in real or complex algebra. In Section 4. A comprehensive way to sum this up is to say that (U. U † ) if and only if (U. we mean this in terms of the structure of U as an AH-module. U ) has a well defined dual AH-module (U × .2.2. We say that the pair (U. 50 . U † ) is an SH-module. if U is an H-module we define U × to be the space of H-linear maps φ : U → H. U † ) might not even be an AH-module! Thus if we are to talk about dual AH-modules.Definition 4. {0}). U ). U † ) as an AH-module — the dual AH-module of (U.2. and an SAH-module if and only if it has no submodule isomorphic to either (H.2.2.

This fact is an example of a more general result which makes the theory of dual spaces in quaternionic algebra particularly useful. there are canonical isomorphisms U ∼ (U × )× and (U × )† ∼ U .7 Let (U.2. Then φ ∈ (ιU × (U × ) ⊗ (V † )∗ ) ∩ ((U )∗ ⊗ ιV (V )) ∩ (I ⊗ (U )∗ ⊗ (V † )∗ ). Then HomAH (U. U † ) to be the dual AH-module of U . the space of AH-morphisms φ : U → H. We start with the following definition: Definition 4. or equivalently φ ∈ (ιU × (U † ) ⊗ (V † )∗ ) ∩ ((U )∗ ⊗ ιV (V )). U ) be an AH-module.2. Then there is a canonical isomorphism HomAH (U.2. Using this and the canonical isomorphism of real vector spaces Hom(A. There is an analogous result in quaternionic algebra. V ) denotes the space of AH-morphisms from U into V . Here is the main result of this section: Theorem 4. and let HomR (A. It is clear that antistable AH-modules are almost always dual to stable AH-modules.2. V ) ∼ (U × ⊗H V ) .2.6 Let U = (U. 4. Then we define U × = (U × . B) ∼ A∗ ⊗ B we see = that φ is exactly equivalent to an real linear map Φ : (U † )∗ → (V † )∗ .8 Let U and V be AH-modules.9 Let U be an SAH-module and V be an AH-module. finite dimensional) modules over the commutative ring R. Let φ ∈ (U × ⊗H V ) . For finite dimensional U .Definition 4.3 Spaces of AH-morphisms and Duality The space U † is. = = Definition 4. 51 . B) denote the space of R-linear maps φ : A → B. Let A and B be (free. as we have remarked. which we regard as an antistable AH-module in spite of the fact that its dual (H. It is well-known that there is a canonical isomorphism HomR (A. H) is not an AH-module. The mapping ιU × identifies U † with ιU × (U † ). it is clear that U × is also an SAH-module. With this definition. which in turn is equivalent to an H-linear map ΦH : H ⊗ (U † )∗ → H ⊗ (V † )∗ . B) ∼ = ∗ A ⊗R B. U ) be an AH-module which is also an SAH-module. The only irreducible exception is the AH-module (H. {0}). Then U is called antistable if and only if U ∩ qU = {0} for all q ∈ I. Consider φ ∈ ιU × (U † ) ⊗ (V † )∗ . = Proof.

1 Let U ∼ Rk be a real linear subspace of Cn . . . . 4. Then we can choose a = j complex basis {e : j = 1. . Proof. ep . ie1 . Using the fact that φ ∈ (U )∗ ⊗ ιV (V ) and the natural identification U ∼ ιU (U ) we see that = ΦH : ιU (U ) → ιV (V ). ieq R . U ) with U = H2 and U = (1. . eq . . p + q = k. ιU (U ) is generated over H by ιU (U ). ieq . . Both U + iU and U ∩ iU are complex subspaces of Cn . . . ep } for U . . ep } is linearly = independent over C. 0)(0. and so {e1 . . . n} of Cn such that U = e1 . . Every AH-morphism ψ : U → V is equivalent to an AH-morphism Ψ : H ⊗ (U † )∗ → H ⊗ (V † )∗ with the property that Ψ : ιU (U ) → ιV (V ). ie1 . . . which is another piece of evidence suggesting that Joyce’s definition of the quaternionic tensor product is the right one for AH-modules. The situation is very different for H-modules. we can construct an element of (U × ⊗H V ) from each AHmorphism from U to V . 1). Theorem 4. In other words. For example. This is not difficult. eq+1 . . . we classify the possible orbits of a subspace Rk of Cn under the action of GL(n. in the sense that each basis vector of the real subspace U can be chosen to lie in one and only one copy of C. U ) ∼ = (Cn . Extend this to a real basis {e1 . though as far as the author can tell both the problem and its solution are original. ieq } is a real basis for U ∩ iU . 52 . . . . . .3. it is useful to compare this situation with that of the complex numbers. i2 ) . Thus ΦH is equivalent to an AH-morphism from U to V . As always with the quaternions. . where q ≤ p ≤ n.which by definition is an AH-morphism.3 Real Subspaces of Complex Vector Spaces We have seen that choosing different real subspaces U of an H-module U ∼ Hn gives = rise to different algebraic properties. and let U ∩ iU = Choose a complex basis {e1 . The quaternionic tensor product can be used in this way to tell us about spaces of AH-morphisms. The pair (U. . eq . . C). As an instructive example (and a bit of light relief!) we shall give a classification of real linear subspaces of complex vector spaces up to complex linear isomorphism. eq } for U ∩ iU . . . Then {e1 . (i1 . . . Then {e1 . consider the AH-module (U. 1 or 2 basis vectors for U . ie1 . ep } spans U + iU ∼ Cp (over C). The result follows. . from which it follows that ΦH : ιU (U ) → ιV (V ). Let U + iU ∼ Cp = ∼ Cq . It is easy to see from this theorem that choosing a real subspace of Cn is always compatible with the complex structure. Rk ) can always be completely reduced to a direct sum of copies of C. each of which contains 0. Since U is an SAH-module. Then dimR (U ) = p + q. . . . Reversing these steps. . . .

occur as complex vector spaces with particular structure maps.2 An AH-module (U. 4.3. In this case. (U2 . 101] Let e1 . i. Theorem 4. which shows that every indecomposable K-module is isomorphic to one of three basic types. some K-modules are equivalent to AH-modules.4.2 [Ben. there are no two non-trivial AH-modules (U1 . §4. The classification of irreducible AH-modules up to AH-isomorphism is a much more difficult problem than its analogue for complex vector spaces. Definition 4.4 K-modules A K-module is an algebraic object based on a pair of complex vector spaces.3]. h2 } is a basis for H. U2 ) such that (U.1 [Q. Representations of the Kronecker quiver are discussed in Benson’s book [Ben. In this guise. We follow a slightly different approach from that in Quillen’s paper to obtain a more immediate link between Kmodules and AH-modules. p. It will be addressed in the next section using a class of algebraic objects called K-modules.1] A K-module is a pair (W. A K-module morphism is a map φ : (W1 → e2 H ⊗ V1 ) −→ (W2 → H ⊗ V2 ) which respects the K-module structure. a K-module is a representation of the Kronecker quiver. We recover the first definition by setting e(w) = h1 ⊗ e1 (w) + h2 ⊗ e2 (w) where {h1 . Quillen is more interested in an interpretation of quaternionic algebra in terms of sheaves over the Riemann sphere. a powerful theory which we will review in the next section. Real and quaternionic vector spaces. where H ∼ C2 is the basic = representation of GL(2. the first of which contains 2 basis vectors for U and the second of which contains the remaining basis vector. A K-module e : W → H ⊗ V is called indecomposable if it cannot be written as the e1 direct sum of two non-trivial K-modules. This allows us to use the classification of irreducible K-modules. U1 ⊕ U2 ). U1 ). 4.e. a problem with a known solution.4. This motivates the following definition: Definition 4. to write down all irreducible AH-modules very explicitly.There is no way to decompose U into two separate copies of H. The reason why K-modules are important to quaternionic algebra is that in the presence of suitable structure maps. This link between K-modules and AH-modules was discovered by Quillen [Q]. U ) is irreducible if and only if it cannot be written as a direct sum of two non-trivial AH-modules. Then one of the following holds: (i) The vector spaces W and V have the same dimension. e2 : W → V be a pair of linear maps constituting an indecomposable K-module. C). The important result is the following classification theorem of Kronecker (which we have summarised slightly). e2 : W → V . U ) = (U1 ⊕ U2 . as so often. V ) of (finite dimensional) complex vector spaces together with a linear map e : W → H ⊗ V . if det e1 = 0 53 . A K-module can equivalently be defined as a pair of linear maps e1 .

Let σV be the standard quaternionic structure on V .∞ dim V = n and α on the leading diagonal of e2 . Then σW is a real structure on W . . e(w3 ) = h2 ⊗ v2 . n. w3 } be a basis for W and let {v1 . 2 Example 4.  . We have e(w1 ) = h1 ⊗ v1 .4. Define X2 to be the irreducible K-module of type (ii) with dim V = n.  . such that e · σW = (σH ⊗ σV ) · e. There is a compatible real structure σW on W given by σW (w1 ) = w3 . (We write X1 for the case det e1 = n n 0 ). w2 .4. and its set of fixed points is the real vector space W σ . .4. 0 1 0 0 0 1 (iii) The dimension of W is one smaller than the dimension of V . Our aim z ¯ is for the K-module e : W → H ⊗ V to define a real subspace of a real H-module..then e1 and e2 can be written in the form  e1 → id α 0  1 α  e2 →  . This gives H ⊗ V the structure of a complex H-module. ... Suppose W → H ⊗ V is an SK-module. The map σH ⊗σV is also a real structure.  If det e1 = 0 a modification is necessary which in some sense corresponds to the ‘rational canonical form at infinity’. so σV (v1 ) = v2 and σV (v2 ) = −v1 . This is accomplished by compatible structure maps on W and V .4 [Q. 11. . 0 1 α    . . Define X3 to be the irreducible K-module of type (iii) with dim V = n. and bases may be chosen so that e1 and e2 are represented by the transposes of the above matrices. and e(W σ ) is a real subspace of this H-module. and in the same way we define the real vector space (H ⊗ V )σ .  .  e1 =  e2 =  . whose H-module structure is inherited from that on H. (ii) The dimension of W is one larger than the dimension of V . Then (H ⊗ V )σ is a real H-module.2] An SK-module is a K-module e : W → H ⊗V equipped with antilinear operators σW and σV of squares 1 and −1 respectively.α Definition 4. . Definition 4. . . e(w2 ) = h1 ⊗ v2 + h2 ⊗ v1 . Let σH be the standard quaternionic structure map on H defined by σH (z1 h1 + z2 h2 ) = −¯2 h1 + z1 h2 .5 Consider the K-module X2 which has dim W = 3 and dim V = 2. v2 } be a basis for V . and bases may be chosen so that e1 and e2 are represented by the matrices     1 0 0 0 1 0   .3 Define X1 to be the indecomposable K-module of type (i) with n. In many circumstances an SK-module is therefore equivalent to an AH-module.. σ(w3 )W = w1 54 . . Let {w1 .

in other words 2m+1 2m+1 X2 ⊗ H or X3 ⊗ H.α 1. The ‘annihilating 0 K-module’ X3 also fails to give an AH-module. It remains to check which pairs (U. i2 and i3 respectively.2 and σW (w2 ) = −w2 . 55 . Here are some important facts about irreducible AH-modules which can now be deduced: • SK-modules of type (ii) correspond to stable AH-modules and SK-modules of type (iii) correspond to antistable AH-modules. in other words 2m 2m X2 or X3 . −1 correspond to 2m • The irreducible stable AH-module corresponding to X2 has dim U = 4m and dim U = 2m + 1. (C) An indecomposable K-module of type (ii) or (iii) with dim V odd. Under this isomorphism the real subspace e(W σ ) is mapped 2 to the imaginary quaternions I. Proof.α (A) The direct sum of a pair of K-modules of type (i) of the form X1 ⊕ X1 α . = 2m+1 • The irreducible stable AH-module of the form H ⊗ X2 has dim U = 4(2m + 1) 2m+1 and dim U = 4(m + 1). I ) = H. = • This shows that there is an irreducible stable AH-module with virtual dimension 1 in every dimension and an irreducible stable AH-module with virtual dimension 2 in every odd dimension. a pair (U.4. This follows from Theorem 4. H). iw2 .− ¯ • Indecomposable SK-modules of type (i) of the form X1 ⊕ X1 α the semistable AH-modules Xq .− ¯ n. This pair is given by the SK-module H ⊗ X3 . −1 (B) An indecomposable K-module of type (ii) or (iii) with dim V even. so that with these structure maps X2 is an SK-module. This demonstrates explicitly that the SK-module X2 is equivalent to the AH-module (H. We have real vector spaces W σ = w1 + w3 . (H ⊗ V )σ = Every real subspace U of a quaternionic vector space U can be obtained in this fashion. As noted earlier. U ) arising in this fashion are AH-modules.2 and explicit calculations using standard structure maps on the vector spaces V . so a classification of SK-modules gives a classification of AH-modules.4. 1. U ) fails to be an AH-module if and only if it has a subspace 1 of the form (H. Any SK-module containing neither of these indecomposables is equivalent to an AH-module. Thus U ∼ Hm and the virtual dimension of U is 1. Thus U ∼ H and the virtual dimension of U is 2. i(w1 − w3 ) and h1 ⊗ v2 − h2 ⊗ v1 i(h1 ⊗ v2 + h2 ⊗ v1 ) . tensored with the basic representation H equipped with its standard structure map σH .6 Every indecomposable SK-module is isomorphic to one of the following: n. Corollary 4. i1 . h1 ⊗ v1 + h2 ⊗ v2 i(h1 ⊗ v1 − h2 ⊗ v2 ) An H-module isomorphism (H ⊗ V )σ ∼ H is obtained by mapping these basis vectors to = 1.

5 The Sheaf-Theoretic approach of Quillen Much of Joyce’s quaternionic algebra can be described using (coherent) sheaves over the complex projective line CP 1 . can be found in Kobayashi’s book [K]. but the resulting theory is exactly the same. so each line bundle on CP 1 is given by one of the transition functions g(z) = z n . The line bundle given by the transition function z n is in fact L. Every holomorphic line bundle over CP n is a tensor power of the hyperplane section bundle L [GH. Thus in the presence of certain structure maps. Quillen’s paper works by recognising that certain exact sequences of sheaf cohomology groups are K-modules. 34-49]. 678-704]. 1 56 . holomorphic vector bundles [pp. we obtain SHmodules. 4. coherent sheaves [pp. Following standard notation. 2 This uses the convention of identifying a holomorphic vector bundle with its sheaf of holomorphic sections. The most interesting new result in this section is that the equivalence between sheaves and SH-modules respects tensor products. 145]. enabling us to calculate the quaternionic tensor product of two SH-modules from knowing the tensor products of the corresponding sheaves. including many of the properties of sheaves used in Quillen’s paper. which introduces sheaves and their cohomology [pp. This interpretation is due to Daniel Quillen [Q]. Thus O = Background material can be found in [GH]. p. n n ∈ Z.• The isomorphism class of an irreducible stable AH-module is thus uniquely determined by the dimension and virtual dimension of U . g : C∗ → C∗ such that ψ = f ψ. and every holomorphic vector bundle is a sum of holomorphic line bundles. 66-71] and holomorphic line bundles [pp. Thus by the end of this section we will have succeeded in classifying all AHmodules and their tensor products. In the case n = 1.5. We will describe the vector bundles first. we use the open cover of CP 1 = C ∪ {∞} consisting of the two open sets U0 = CP 1 \{∞} and U1 = CP 1 \{0}. so ψ : C∗ → C∗ . Two transition functions ψ and ψ determine the same line bundle if and only if there exist non-vanishing holomorphic functions f.) Quillen uses slightly different structure maps from those we used in the previous section to obtain SK-modules. (Quillen deals primarily with SH-modules rather than AH-modules. A more thorough exposition of the differential geometry of holomorphic vector bundles. 2 These summands factorise very neatly — every torsion sheaf is the sum of indecomposable sheaves supported at a single point. we write O(n) for the sheaf of its holomorphic sections. 1 Quillen demonstrates that every coherent sheaf over CP 1 is the direct sum of a holomorphic vector bundle and a torsion sheaf (one whose support is finite). 132-139].1 Sheaves on the Riemann Sphere We describe the algebraic geometry of (coherent) sheaves over CP 1 . 4. A holomorphic line bundle over CP 1 is determined by a holomorphic transition function ψ : U0 ∩ U1 → C∗ . g Two functions are equivalent under this relation if and only if they have the same winding number.

11]. Let H ∼ C2 be the basic representation of GL(2. which we write O/mn z by extending mz by O on the complement of z. (4. which are called torsion sheaves. (4. 137] 4 For example [W.10) C (global holomorphic functions on CP (1) ) and that exact sequences of (4. §2] demonstrates that every coherent sheaf over CP 1 is the sum of a vector bundle and a torsion sheaf.3] Any coherent sheaf over CP 1 splits with unique multiplicities into indecomposable sheaves of the form O(n) for n ∈ Z and O/mn for n ≥ 1 and z z ∈ CP 1 .5. It is from the maps in these sequences that we obtain K-modules and thence SH-modules. Let z ∈ CP 1 be such a point. 3 Thus every holomorphic vector ∞ n bundle E is a sum of irreducible line bundles.9) where F (n) denotes the sheaf F ⊗O O(n). Torsion sheaves themselves split into sheaves supported at one point only.2 If we put F = O(n). Then CP 1 can be identified = with the set of quotient lines of H and there is a basic exact sequence 0 → Λ2 H ⊗ O(−1) → H ⊗ O → O(1) → 0.5. p. and can be written E = −∞ an L . We have the following Theorem: Theorem 4. This leaves us to consider sheaves which are supported at a finite set of points. where H 0 (O(n)) is shown to be isomorphic to the (complex) vector space of homogeneous polynomials of degree n in 2 variables. C). Every holomorphic vector bundle over CP 1 can be written as a direct sum of these line bundles.8) Tensoring (over O ) with the sheaf F and choosing an identification Λ2 H ∼ C yields = the exact sequence 0 → F (−1) → H ⊗ F → F (1) → 0.O(0) is the structure sheaf of CP 1 . p.10) and induction. It is easy to see (by multiplying the transition functions together) that Ln ⊗ Lm ∼ Ln+m .1 [Q. Cohomology Groups and Exact Sequences There are various ways to calculate the cohomology groups of these sheaves. 57 . the summands being unique up to order. and there is an exact sequence by (4. Quillen [Q. By the H 0 (O(n)) ∼ S n (H) for n = of cohomology groups given (4.11) 0 → S n−1 H → H ⊗ S n H → S n+1 H → 0. Define mz to be the unique maximal ideal of Oz consisting of germs of functions whose first derivative vanishes at z. and let Oz be the ring of germs of holomorphic functions at z. we have the exact sequence 0 → O(n − 1) → H ⊗ O(n) → O(n + 1) → 0. Every torsion sheaf splits into sheaves of the form Oz /(mz )n . Example 4. or in sheaf-theoretic terms O(n) ⊗O O(m) ∼ = = O(n + m). We know that H 0 (O) ∼ = 0 H (O(−1)) = 0. it follows that ≥ 0 and zero otherwise. 4 The method Quillen outlines uses the properties of exact sequences of sheaves. n ≥ 0. 3 This follows from the Harder-Narasimhan filtration of a holomorphic vector bundle E over any Riemann surface M [K. where the multiplicities an are unique (though the decomposition itself may not be). 2.

Quillen uses K-modules with a quaternionic structure σW on W and a real structure σV on V . This is clearly a K-module.4.10) gives rise to an injection of cohomology groups H 0 (O(n − 1)) → H 0 (H ⊗ O(n)) which takes the form e : S n−1 H → H ⊗ S n H. For all regular sheaves F . A σK-module is not itself an SH-module. using a resolution involving the sheaf cohomology groups H 0 (F ). Some sheaves will correspond to SKmodules.Torsion sheaves can be dealt with in a similar fashion.2 and 4. but the inclusion of the real subspace V σ in the cokernel (H ⊗ V )/W is an SH-module if the K-module has no submodule for which the map e is surjective. It is easy to see that H 0 (O/mn ) ∼ Cn . Similarly for the torsion sheaves O/mn there is an injection H 0 (O/mn (−1)) → x x H 0 (H ⊗ O/mn ). These must be treated slightly differently.4. such that the map e intertwines σW and σH ⊗ σV . In this situation. torsion sheaves. Quillen investigates z σ(z) σ-sheaves thoroughly. a torsion sheaf F is a σ-sheaf if it is supported at a finite set of points which is preserved by the antipodal map σ — so it must consist of sheaves of the form O/mn ⊕ O/mn . and allows us to define a ‘ σ-invariant sheaf’ or just ‘ σ-sheaf’. Let W → H ⊗ V be a K-module. The rest of the programme begins to take shape. Quillen proves this in detail [Q.12) gives a map H 1 (O(n − 1)) → H ⊗ H 1 (O(n)) which is equivalent to the indecomposable K-module −n−1 X3 . We call z = the sheaves O(n) where n ≥ 0. which we call negative vector bundles. Thus for any negative vector bundle G we obtain a K-module which we call − ξ G. Quillen formulates this slightly differently from our treatment in Section 4. Since H 0 (O(n)) = 0 for n < 0. where σ ∗ interchanges the two summands. the first cohomology group H 1 (F ) is zero. Instead of a real structure on W and a quaternionic structure on V . Thus we obtain a x K-module from each regular sheaf F which we call ξ + F . This leaves the sheaves O(n) where n < 0 and sums thereof. He calls this structure a σK-module.5] and uses the tensor product of sheaves F ⊗O G to construct an equivalent tensor product operation for K-modules [Q. Let σ : z → z −1 be the antipodal ¯ map on the Riemann sphere. For example.12) . There is a parallel description in terms of sheaves. n < 0 be a negative vector bundle. §§4. from which we obtain SH-modules.1). and discovers that: 58 (4. The exact sequence (4.5. using the first cohomology groups H 1 (F ). §6]. and sums thereof regular sheaves. it is clear that these categories are equivalent. This induces a map of sheaves σ ∗ : F → σ ∗ (F ) which we call the σ-transform. This gives a K-module of type (i) and dim W = n. in which case the K-module is called reduced. The exact sequence (4. we obtain an exact sequence 0 → H 1 (O(n − 1)) → H ⊗ H 1 (O(n)) → H 1 (O(n + 1)) → 0. = Sheaves and K-modules Consider the regular sheaf O(n) for n ≥ 0. Comparing the classifications of indecomposable sheaves and K-modules (Theorems 4. the irreducible K-module n+1 X3 . Let O(n). W and H ⊗ V are H-modules and e is an H-linear map. from which it follows that H 1 (O(n)) ∼ S −n−2 H for n ≤ −2 and zero otherwise.

The formal similarity between this result and Corollary 4.6 Let G be a negative σ-vector bundle with no summand H ⊗ O(−1) or O(−2).5.6 is clear.Proposition 4. 4. 59 .6] The categories of reduced σK-modules and SH-modules are equivalent.5.1] 5 5 Quillen proves this theorem for the tensor product of K-modules — the version given here is obtained by performing the simple translation into SH-modules. then H 0 (G) = H 0 (G(1)) = 0 and H 1 (F ) = H 1 (F (1)) = 0 .5. Taking real subspaces gives an SH-module which we call η + (F ). We can also use σ-sheaves to obtain SK-modules which lead directly to SH-modules. Just as in the previous example.2 Sheaves and the Quaternionic Tensor Product We have seen how the tensor product of sheaves encourages us to define a ‘reduced tensor product’ operation for K-modules. Then we have the exact sequence 0 → H 1 (G) → H ⊗ H 1 (G(1)) → H 1 (G(2)) → 0. we could combine the functors η + and η − into a single functor η = η + + η − . In theory.7 [Q. with F and G as above. (2) O(2m) for m ∈ Z. It turns out that this tensor product agrees remarkably with the quaternionic tensor product for SH-modules. 10. He also proves that: Proposition 4. SK-modules and SH-modules allow us to compute tensor products in each category. The main theorem is as follows: Theorem 4.5.4. Example 4. so η + (G) = η − (F ) = 0. We also define η − (O(−2)) = η − (H ⊗ O(−1)) = 0. since η(A) = η + (F ) + η − (G) as required. 7.5.5 Let F be a regular σ-sheaf. If we have a σ-sheaf A = F +G. (4. This is a considerable bonus from Quillen’s theory — the correspondences between σ-sheaves. σ(z)} of antipodal points and n ≥ 1. (4.3 [Q. this sequence gives an SK-module H 1 (G) → H ⊗ H 1 (G(1)).4 [Q.14) We call this SH-module η − (G).5.13) and H 0 (F ) → H ⊗ H 0 (F (1)) is an SK-module e : W → H ⊗ V . 12. Then we have the exact sequence 0 → H 0 (F ) → H ⊗ H 0 (F (1)) → H 0 (F (2)) → 0. Example 4. (3) O(2m + 1) ⊗ H for m ∈ Z.7] Any σ-sheaf splits with unique multiplicities into the following irreducible σ-sheaves: (1) O/(mz mσ(z) )n for any pair {z.15) (4.

which of course take the form (η + F )× and (η − G)× . For example. we will often find ourselves using this theorem for the corresponding AH-modules. Since we will usually work with AH-modules. (η + F )× ⊗H − (η G)× = {0}. η + F1 ⊗H η − G1 = η − (F1 ⊗O G1 ) and η − G1 ⊗H η − G2 = {0}. Then η + F1 ⊗H η + F2 = η + (F1 ⊗O F2 ).3 by translating known results about the degree and rank of the tensor product of two sheaves. If F = O(m) and G = O(n) (possibly tensored with H if m or n is odd) with m ≥ 0 and n < −2 then (η + F )× ⊗H (η − G)× = 0 m + n ≥ −2 × η (F ⊗O G) m + n ≤ −3 − since η − (O(k)) = {0} for k ≥ −2. 60 . it is possible to obtain the dimension theorems of Section 4. The sheaf-theoretic approach is thus a very powerful tool for describing quaternionic algebra.1. Thus if F is a torsion σ-sheaf and G is a negative σ vector bundle.Let Fi be regular σ-sheaves and Gi be negative σ-vector bundles.

We use representations to explain the structure of stable AH-modules and their tensor products. This point of view turns out to have fruitful applications in hypercomplex geometry.1 Stable AH-modules and Sp(1)-representations Sp(1)-representations on the quaternions As a motivating example. especially since Sp(1) is just the group of unit quaternions and the Lie algebra sp(1) can be identified with I.r and its splitting into two H-submodules. where the left hand = 61 . from K¨hler a geometry to particle physics. we will see how stable and antistable AH-modules can be handled using Sp(1)-representations. Because the Riemann sphere CP 1 can be described as the homogeneous space Sp(1)/U(1). Recall the description of the quaternions as a tensor product of two Sp(1)-representations H ∼ V1 ⊗ V1 . Given this versatility.Chapter 5 Quaternionic Algebra and Sp(1)-representations The representations of the group Sp(1) occur in so many different situations. We also investigate the role of semistable AH-modules and their interaction with the Sp(1)-representation structure of stable AH-modules. the holomorphic sections of line bundles over CP 1 are naturally Sp(1)-representations. Not only do Sp(1)-representations underlie all of these phenomena — they also make the theory of quaternionic algebra very easy to predict and manipulate.4. I). that they are by far the most ubiquitous Lie group representations in modern mathematical literature. viewed as the stable AH-module (H. where we came across the space H ⊗ Ek. The structure maps necessary to define an SK-module e : W → H ⊗ V of type (ii) or (iii) arise from Sp(1)-representations on W . In this chapter. We encountered the germ of this idea in Section 3.1. given in Equation (3. it is no surprise that these representations are a powerful tool in quaternionic algebra. H and V . 5. towards which our exposition is deliberately geared. we review the case of the quaternions themselves.17).1 5.

For the quaternions. to make it clear which group is acting on which vector space. Similarly. q) : q ∈ Sp(1)} ⊂ Sp(1) × Sp(1) on V1 ⊗ V1 . q ∈ Sp(1) ⊂ H.1. It is this account of the AH-module H which we will generalise to all stable AH-modules. where H ∼ X2 .4.1. we think of H as an Sp(1) × Sp(1)-representation by defining (p.copy of V1 gives the left H-action. the basic vector space H we used so much in the = 2 previous chapter is simply a copy of the basic representation V1 . which as we know preserves the splitting H ∼ I ⊕ R. This description of the quaternions is very similar to that of Example 4. the AH-module structure H ∼ I = = † ∗ ∼ and (H ) = R is a concept which arises naturally when we take both the Sp(1) actions into account. We remove this ambiguity by writing upper-case superscripts with the groups and the representations. For left H-modules there will always be a left H-action to consider.1) = into real subspaces of dimensions three and one respectively. Each of these V2 ⊕ V0 (equivalent to the standard isomorphism V ⊗ V = summands inherits a real structure from the real structure on V1 ⊗ V1 so we obtain the splitting V1 ⊗ V1 ∼ V2 ⊕ V0 (5. For more details. The Clebsch-Gordon formula gives the splitting V1 ⊗ V1 ∼ = ∼ S 2 V ⊕ Λ2 V ). We have already encountered the action of Sp(1) × Sp(1) on V1 ⊗ V1 . Here we have two copies of Sp(1) acting. We will call this subgroup Sp(1) LM .5. We will denote this by V1L . q) : r → prq −1 r ∈ H. p. q)} ⊂ Sp(1)L × Sp(1)M . 5.2 Notation for Several Sp(1)-representations It will be a sound investment at this point to introduce some notation to help us keep track of the structure of representations when we have several copies of Sp(1) acting on a vector space. N etc. 1 62 . and the right-hand copy gives the right H-action. implicitly using the induced map σ1 ⊗ σ1 as a real structure on V1 ⊗ V1 . 1 Consider now the action of the diagonal Sp(1)-subgroup {(q. and the copy of Sp(1) which acts on this factor by Sp(1) L . = We are talking about the representation V1 ⊗ V1 as a real representation on R4 . In terms of Sp(1)-actions. so there is already the possibility of ambiguity concerning which Sp(1) is acting on which V1 . So we would write the above example as Sp(1)L × Sp(1)M acting on V1L ⊗ V1M . In other words. thus stating explicitly of which two groups this is the diagonal subgroup. we can combine superscripts for the representations to write V1L ⊗ V1M ∼ V2LM ⊕ V0LM . refer to Section 1. When we decompose such a representation using the Clebsch-Gordon formula.2. just as we would expect. we are decomposing the action of the diagonal subgroup {(q. This is the same as taking the action by conjugation r → qrq −1 . Other copies of Sp(1) and other representations will be labelled with the letters M.

The primed part is then a representation of the diagonal subgroup Sp(1)LM ⊂ Sp(1)L × Sp(1)M . The basic idea is exactly the same — a stable AH-module is a (real) Sp(1)L × Sp(1)M -representation V1L ⊗ W M . and M ∼ Sp(1) if the = virtual dimension of U is 2. i. We denote this subgroup Sp(1)LM × Sp(1)N .9. we considered the diagonal subgroup of the first and last copies of Sp(1). Let (U. 5. U ) be an irreducible stable (or antistable) AH-module. For example. Then G is a compact Lie group of which M is a normal subgroup. Using Theorem 4. q)} × Sp(1)N . {(q. Consider now the more general group G of real linear isomorphisms φ : U → U such that: • φ(U ) = U . however. so that we now have Sp(1)LM × Sp(1)N LM LM acting on (Vj+1 ⊕ Vj−1 ) ⊗ VkN .e. We combine superscripts for the representations in the same way. where W M = k aj VjM . If. we write this as Sp(1)L × Sp(1)M × Sp(1)N acting on V1L ⊗ VjM ⊗ VkN . • The (real) determinant of φ is 1. we see that M = ±1 if the virtual dimension U is 1. we would write this as Sp(1)LN × Sp(1)M LN LN acting on (Vk+1 ⊕ Vk−1 ) ⊗ VjM . Let M be the group of AH-automorphisms of U whose (real) determinant is equal to 1.2. For example. This provides an unambiguous and (it is hoped) easy way to understand tensor products of several representations and their decompositions into irreducibles under different diagonal actions.3 Irreducible stable AH-modules We have described the quaternions as an Sp(1)L × Sp(1)M -representation. and we can join the superscripts as above to indicate exactly which one we are considering. In this section we will demonstrate how this idea can be adapted to describe more general stable AH-modules. if we have three copies of Sp(1) acting. The left H-action is given by the action of the left subgroup 1 Sp(1)L ⊂ Sp(1)L × Sp(1)M .This book-keeping comes into its own when we come to consider tensor products of many Sp(1)-representations.1. This allows us to interpret the primed part H ∼ I as a representation of the diagonal subgroup = LM L M of Sp(1) ⊂ Sp(1) × Sp(1) . supposing we want to restrict to the diagonal subgroup in the first two copies of Sp(1). u ∈ U . Because of this the Lie algebra of G splits into two orthogonal ideals g = sp(1) ⊕ m 63 . • There exists some q ∈ Sp(1) such that φ(pu) = (qpq −1 )φ(u) for all p ∈ H. In this situation there are various diagonal actions we could be interested in.

64 . U ) m 2m+1 with U ∼ H and U ∼ R . U ) and thus U is stable. So W must be an Sp(1)LM -invariant subspace of U = V2m and by Schur’s Lemma [FH.4. we have the splitting M LM V1L ⊗ V2m−1 ∼ V2m ⊕ V2m−2 . there is a unique irreducible stable AH-module for each irreducible Sp(1)-representation. and the irreducible decomposition of U as an AH-module determines the irreducible decomposition of the Sp(1)-action on U . Because σ = M σ1 ⊗ σ2m−1 is a real structure. there is no representation of Sp(1) on Xq which couples with the left H-action to give a representation of Sp(1) on Xq . AH-modules of the form V1 ⊗ V2m−1 Consider an even-dimensional irreducible Sp(1)-representation V2m−1 .7] W = V2m or W = {0}.r from Section 3. there is an action of Sp(1) on U for all stable AH-modules U . = LM For example. and vice versa. M LM Proposition 5. Also. = LM (5. there is a real representation V1L ⊗ V2m−1 ∼ R4m with = a left H-action defined by q : a ⊗ b → (qa) ⊗ b. Since every stable AH-module is a sum of such irreducibles. Since the elements of G map U to itself.r . Both stable and antistable AH-modules arise as Sp(1) × Sp(1)-representations. W ) = (U. the smaller one is antistable. We see that each of these summands is an AH-submodule of H ⊗ Ek. The larger (left-hand) submodule is stable. Since dim U > 1 dim U we must 2 have W = {0}.1. = = Proof. The primed part of U is then the LM Vn+1 summand in the decomposition LM M V1L ⊗ Vn ∼ Vn+1 ⊕ Vn−1 .r ∼ V1 ⊗ εn (Vr+1 ⊕ Vr−1 ) = k.3) M which is a splitting of real vector spaces (technically we could write (V1L ⊗ V2m−1 )σ ∼ = LM LM (V2m )σ ⊕ (V2m−2 )σ ). For example. This diagonal action will be the result of the left H-action on V1 and some other Sp(1)-action on U . Let U be a stable AH-module and suppose that U is preserved by some diagonal action of Sp(1). As we shall see. W must depend solely on the Sp(1) × Sp(1)-representation structure: in particular W must be preserved by the diagonal action.2) as a representation of the group Sp(1)L × Sp(1)M . p. Hence (W.and the exponential map determines a homomorphism ρ : Sp(1) → G. Consider the maximal stable submodule W of U . Thus our AH-modules will follow the basic form M U = V1L ⊗ Vm (5. recall the splitting H ⊗ Ek. Since W is an H-submodule it must be invariant under the left H-action. V2m ) forms a stable AH-module (U. ρ is a representation of Sp(1) on U .1 The pair (V1L ⊗ V2m−1 . Under the diagonal Sp(1)LM -action q : a ⊗ b → (qa) ⊗ (qb). This is not necessarily the case for AH-modules which are neither stable nor antistable.

Proof. the image e(W ) is spanned by = the vectors {x⊗an . we would like to find out how the left H-action interacts with the splitting V1 ⊗ Vn ∼ Vn+1 ⊕ Vn−1 . . This shows that x ⊗ an−k bk − y ⊗ an−k+1 bk−1 ∈ Vn−1 . It is easier to observe the sp(1) action on the complementary subspace which we can identify as the U † -part of an SH-module. This follows from Proposition 5.1. . an−1 b. y ⊗ an−1 b. The H-action is as usual determined by the action of sp(1) on x and y using the correspondence 1 ←→ x i1 ←→ ix i2 ←→ y i3 ←→ −iy. . from which it follows that U must be irreducible. . x ⊗ abn−1 . . y and Vn = an . Using these isomorphisms and Theorem 4. . Let V1 = x. The virtual dimension of U is 1. where H ∼ V1 and = V ∼ Vn .2.5. Calculating the action of the Casimir operator C = H 2 + 2XY + 2Y X reveals that C(x ⊗ an−k bk − y ⊗ an−k+1 bk−1 ) = (n + 1)(n − 1)(x ⊗ an−k bk − y ⊗ an−k+1 bk−1 ). C) on V1 and Vn are given by Equations (1.4) The subspaces e(W ) and e(W )⊥ of the SK-module H ⊗ V are thus equivalent to the subspaces Vn+1 and Vn−1 respectively in the splitting V1 ⊗ Vn ∼ Vn+1 ⊕ Vn−1 . . . = 2m Consider the structure of X2 as a K-module e : W → H ⊗ V . . x ⊗ an−1 b. x⊗an−1 b+y⊗an . The isomorphism 2m with X2 follows because irreducible stable AH-modules are uniquely determined by their dimensions. . .4. To make statements less cumbersome. .4. x ⊗ bn y ⊗ an . y ⊗ abn−1 . y⊗bn }.1 and the remarks in Section 4. 65 . In particular. bearing in mind that n is odd. we let n = 2m − 1 throughout.1). y ⊗ bn . the perpendicular subspace to e(W ) is spanned by vectors of the form x⊗an−k bk −y⊗an−k+1 bk−1 . x⊗bn +y⊗abn−1 . We want to understand the actions on the tensor product V1 ⊗ Vn = x ⊗ an . x⊗an−k bk +y⊗an−k+1 bk−1 . abn−1 . .2. . bn be Sp(1)-representations. .16) of Section (1.2.1.1. = The SK-module structure maps σW and σV are exactly the standard real structure σn−1 and the quaternionic structure σn introduced in Section 1.14) and (1. Given a suitable choice of metric. The actions of sl(2. . .4. This formalism enables us to write out the structure of V1 ⊗ Vn as an H-module in the same fashion as in Example 4. . It is instructive to describe the splitting of V1 ⊗ V2m−1 thoroughly in terms of basis vectors. .Lemma 5. giving the result that Vn−1 = Span{x ⊗ an−k bk − y ⊗ an−k+1 bk−1 : 1 ≤ k ≤ n}.2 The AH-module U = V1 ⊗ V2m−1 is the irreducible stable AH-module 2m corresponding to the SK-module X2 . . (5.

(This could also be taken to signify ‘module’ M action. thus we can think of V1L ⊗ V2m ⊗ H as a direct sum of M M two copies of V1L ⊗ V2m .and SH-modules very explicitly. This space comes equipped with a real structure σ = σ1 ⊗ σ2m ⊗ σH . and so we have a stable AH-module M LM ((2V1L ⊗ V2m )σ . If we take the tensor product V1L ⊗ V2m ∼ C4m+2 = = L M ∼ LM LM we obtain the splitting V1 ⊗ V2m = V2m+1 ⊕ V2m−1 and a left H-action in the same way as above. we can tensor with = ∼ C2 equipped with its standard structure map. We can choose to regard this as a stable AH-module by thinking of aVn+2 as the ‘primed part’. This formulation allows us to see the relationship between stable AH. (2V2m+1 )σ ). 2n+2 The AH-module U2n corresponds to the SK-module X2 and the σ-sheaf O(2n). The reason for this is that the structure map σ1 ⊗ σ2m has square −1 instead of 1. or as a stable SH-module by regarding aVn as the ‘generating real subspace’.1. Firstly. (5. 5.4 General Stable and Antistable AH-modules LM M Definition 5. and the action of the ‘right’ subgroup Sp(1)M on Vn as the Sp(1)M -action.AH-modules of the form V1 ⊗ V2m We can also obtain a stable AH-module from an odd-dimensional irreducible M Sp(1)-representation V2m ∼ C2m+1 .3 Let U2n denote the AH-module (V1L ⊗ V2n+1 . where a = 1 if n is even and a = 2 if n is odd.5) (As usual. once the correct structure maps have been specified. The vector space H is unaffected H = M by the Sp(1)L × Sp(1)M -action. 2n+1 The AH-module U2n−1 corresponds to the SK-module X2 ⊗ H and the σ-sheaf O(2n − 1) ⊗ H.4 and 4. which we write 2V1L ⊗ V2m . M and so V1L ⊗ V2m is a quaternionic rather than a real representation of Sp(1)L × Sp(1)M .) This approach is the equivalent of dealing with SK-modules 2m+1 of the form Xk=2.3 ⊗ H and the σ-sheaves O(2m + 1) ⊗ H.5 allow us to state the following theorem: 66 . Thus we write Un = aV1 ⊗ Vn+1 . The AH-module Un = aV1 ⊗ Vn+1 splits as a(Vn+2 ⊕ Vn ). whilst doubling the dimension of the real vector space we are considering so that it is divisible by four. Both these approaches give exactly the same AH-module.) We will often omit the superscripts L and M from expressions like V1L ⊗ Vn if the context leaves no ambiguity as to which group acts on what. we will refer to the action of the ‘left’ subgroup Sp(1)L on V1 as the left H-action. LM M Let U2n−1 denote the AH-module (2V1L ⊗ V2n .1. However this does not restrict to an H-action on any suitable real vector space U such that U ⊗R C = V1 ⊗ V2m (this is obviously impossible since R4m+2 ∼ Hk for any = k ). By analogy with the quaternions themselves. The classification results of Sections 4. There are two ways round this difficulty. V2n+2 ). 2V2n+1 ). we will not usually mention the σ-superscript. we could simply take the underlying real vector space R8m+4 ∼ V1 ⊗ V2m to be an H-module. Secondly. both effectively leave the Sp(1) × Sp(1)-representation V1 ⊗ V2m untouched.

The Sp(1)M -action can be much more complicated.1.1. we know that since Sp(1) is a compact group.1.5 Line Bundles over CP1 and Sp(1)-representations There is naturally a link between Sp(1)-representations and the cohomology groups of vector bundles over CP 1 .C). The AH-module U2n−1 corresponds to the SK-module X3 O(−2n − 3) ⊗ H.1. The following Lemma then follows immediately from Definition 5. × 2n+2 and the σ-sheaf O(−2n− The AH-module U2n corresponds to the SK-module X3 × 2n+1 ⊗ H and the σ-sheaf 4).6 The antistable AH-module U2n takes the form (V1L ⊗ V2n+1 . 2V2n−1 ).3.4. Consider the direct sum U = n aj Uj . any such representation can be written as a sum of irreducibles with unique multiplicities. × LM M The antistable AH-module U2n−1 takes the form (2V1L ⊗ V2n .1. it is then easy to separate these representations to form separate AHM modules. 5. U ) can be attached an Sp(1)M -action which intertwines with the left H-action in such a way that the diagonal Sp(1)LM -action preserves U . However. provided that each odd-dimensional representation V2k appears with even multiplicity.C) ⊂ where H = GL(2.1.Theorem 5.4 Every stable AH-module can be written as a direct sum of the irreducibles Un with unique multiplicities.5. There is a ‘unique factorisation theorem’ for antistable AH-modules which is exactly dual to Theorem 5. Thus the virtual dimension n of j odd cj .6). V2n ). × M LM Lemma 5.C). In Section 4.6) The left H-action on V1L is common to all the irreducibles.C) and therefore Sp(1).5 To every stable AH-module (U. j even cj + 2 j=0 cj Uj is equal to Antistable AH-modules × Let Un = aV1 ⊗ Vn+1 be a stable AH-module.1. Just like stable SH-modules. as the sum m n M V2j+1 ⊕ 2 j=1 k=1 M V2k U = V1L ⊗ . = ∼ C2 is the basic representation of GL(2. The 67 . (5. each irreducible subrepresentation of the Sp(1)M -action contributes 1 to the virtual dimension of U . Then its dual AH-module Un is an antistable AH-module.1 we demonstrated that H 0 (O(n)) ∼ S n (H). H is also the basic representation V1 of SL(2. antistable AH-modules are formed by taking the smaller summand in the splitting aV1 ⊗Vn+1 = a(Vn+2 ⊕Vn ).4: Theorem 5. From the inclusion SL(2. Thus the following is equivalent to Theorem 5. In the decomposition of Equation (5. We can write U more explicitly in terms j=0 of Sp(1)L × Sp(1)M -representations. Having done this.

Let t be the Lie algebra of T . we have † N H ⊗ (Un )∗ ∼ V1L ⊗ V1R ⊗ aVn . the cohomology groups of Lλ are naturally representations of G.5. 382-393]. The holomorphic line bundle associated to the principal bundle G and the representation λ is then Lλ = G ×T Cλ = (G × Cλ )/{(g. Let G be a compact Lie group and let T be a maximal toral subgroup. where L is the hyperplane section bundle of CP 1 . = 68 . (This famous result is due to Borel.8) The fact that the cohomology groups of line bundles over CP 1 have the structure of irreducible Sp(1)-representations is already known in the context of the theory of homogeneous spaces. For more information see [FH. 5. p. The line = −λ bundle Sp(1) ×U(1) Cλ is then L . we have H 0 (O(n)) ∼ Vn and H 1 (O(−n)) ∼ Vn−2 . For each dominant weight λ ∈ t∗ there is a one-dimensional representation Cλ of T . Since G acts on Lλ .induced action of Sp(1) on S n (H) is by definition the irreducible representation Vn . and the homogeneous space Sp(1)/U(1) ∼ CP 1 is the Hopf fibration S 1 → S 3 → S 2 . LM Un = aVn+2 † LM and Un ∼ (Un )∗ = aVn .1 is thus the same as the exact sequence 0 −→ Vn−1 −→ V1 ⊗ Vn ∼ Vn−1 ⊕ Vn+1 −→ Vn+1 −→ 0. Then the homogeneous space G/T has a homogeneous complex structure. so that t is a Cartan subalgebra of g. 5.11) of Section 4.2. Then M Un = a(V1L ⊗ Vn+1 ). Thus the cohomology groups of line bundles over CP 1 are Sp(1)-representations. (5. each maximal torus is isomorphic to U(1). t ∈ T }. This map has a natural interpretation in terms of the Sp(1)-representations involved.1 The inclusion map ιU (U ) We begin by discussing the map ιU and its image.2 Sp(1)-Representations and the Quaternionic Tensor Product This section describes the quaternionic algebra of stable and antistable AH-modules using the ideas of the previous section. v) ∼ (gt. = (5. t−1 v). Let Un be an irreducible stable AH-module. In the case of the group Sp(1). Writing the quaternions as the stable AH-module V1L ⊗ V1R .7) = = The exact sequence (4.) The right action of T on G gives G the structure of a principal T -bundle over G/T . = † † There is an injective map ιUn : Un → H ⊗ (Un )∗ .

and antistable AH-modules and their subspaces as representations of lowest weight. = = † and ιUn (Un ) is the aVn+2 subrepresentation of I ⊗ (Un )∗ . since we do not have an intact copy of V1L . we have † N LRN LRN LRN I ⊗ (Un )∗ ∼ V2LR ⊗ aVn ∼ a(Vn+2 ⊕ Vn ⊕ Vn−2 ). × LM (Un ) = aVn × × LM and (Un )† ∼ ((Un )† )∗ = aVn+2 . = = × RN × This time. 69 .7. This also shows why we would not expect U to be closed under the left H-action — the group Sp(1)L does not act upon it.10) = splits H ⊗ aVn into the direct sum of a stable AH-module isomorphic to Un and an × antistable AH-module isomorphic to Un−2 .5. In exactly the same way.4. for × Un we have × N LRN LRN LRN I ⊗ ((Un )† )∗ ∼ V2LR ⊗ aVn+2 ∼ a(Vn+4 ⊕ Vn+2 ⊕ Vn ). the author hopes that including a little more description will help the reader to get more of a feel for what is going on. Let Um = aV1 ⊗ Vm+1 . The results in this section can be obtained through Quillen’s sheaf-theoretic version of AH-modules by using Theorem 4. The AH-submodule ιUn (Un ) is clearly the V1L ⊗ RN † aVn+1 subrepresentation of H ⊗ (Un )∗ .2. ιUn ((Un ) ) is the smallest subrepresentation aVn . Treating the imaginary quaternions I as a copy of V2LR .1. Thus the splitting N RN RN H ⊗ aVn ∼ aV1L ⊗ (Vn+1 ⊕ Vn−1 ) (5. It is worth noting that so far we have been able consistently to interpret stable AHmodules and their subspaces as representations of highest weight in tensor products of Sp(1)-representations. = There is a similar splitting × N RN RN H ⊗ ((Un )† )∗ ∼ V1L ⊗ V1R ⊗ aVn+2 ∼ V1L ⊗ a(Vn+3 ⊕ Vn+1 ). By Definition 4.9) as an Sp(1) × Sp(1)-representation. † † † † Um ⊗H Un = (ιUm (Um ) ⊗ (Un )∗ ) ∩ ((Um )∗ ⊗ ιUn (Un )) ⊂ H ⊗ (Um )∗ ⊗ (Un )∗ . Leaving the left-action untouched and taking the diagonal Sp(1)RN -action gives the isomorphism RN RN † H ⊗ (Un )∗ ∼ V1L ⊗ a(Vn+1 ⊕ Vn−1 ) = (5.This is exactly like the motivating example of H ⊗ Ek. However. Un = bV1 ⊗ Vn+1 be stable AH-modules. × Antistable AH-modules behave in a similar fashion. × × The subspaces ιUn (Un ) and ιUn ((Un ) ) have a similar interpretation. so that × M Un = a(V1L ⊗ Vn+1 ).4. Consider the AH-module Un . = = LRN × × In this case. 5. the AH-submodule ιUn (Un ) is the smaller AH-submodule V1L ⊗aVn+1 .r in Section 3.2 Tensor products of stable AH-modules We shall now see how to use our description of stable AH-modules to form the quaternionic tensor product.

The argument goes in the opposite direction. If on the other hand we go down the right hand side.13) RP Q A rearrangement of the factors leaves us considering the spaces abV1L ⊗ Vm+1 ⊗ Vn RQ P and abV1L ⊗ Vm ⊗ Vn+1 . = (5. we write ιUm (Um ) ∼ aV1L ⊗ Vm+1 ⊂ V1L ⊗ a(Vm+1 ⊕ Vm−1 ) ∼ = = † ∗ † ∗ ∼ Q H ⊗ (Um ) . we consider the diagonal action of the subgroup Sp(1)RP . This is summed up in the following diagram: † † P Q H ⊗ (Um )∗ ⊗ (Un )∗ ∼ V1L ⊗ V1R ⊗ aVm ⊗ bVn =      † Um ⊗ (Un )∗ s d d d d d d d d † (Um )∗ ⊗ Un             ∼ = c ∼ = c RP Q † ιUm (Um ) ⊗ (Un )∗ ∼ V1L ⊗ aVm+1 ⊗ bVn = RQ † P (Um )∗ ⊗ ιUn (Un ) ∼ aVm ⊗ V1L ⊗ bVn−1 = s d d d d d d d d                  Um ⊗H Un ∼ abV1L ⊗ ( = V RP Q ) ? The upward arrows here are inclusion maps.11) RP RP RP Using Equation (5. We now have an Sp(1)L × Sp(1)RP × Sp(1)Q -representation L and an Sp(1) × Sp(1)P × Sp(1)RQ -representation. = (5. † † P Q H ⊗ (Um )∗ ⊗ (Un )∗ ∼ V1L ⊗ V1R ⊗ aVm ⊗ bVn . = (5. we hope to find the intersection of these two spaces. We then consider the action of Sp(1)RP Q on this. We examine the factors RQ P Q RP Vm+1 ⊗ Vn and Vm ⊗ Vn+1 . Tensoring this expression with (Un ) = bVn gives † RP Q ιUm (Um ) ⊗ (Un )∗ ∼ a(V1L ⊗ Vm+1 ) ⊗ bVn .12) In the same way. we want to reduce these two Sp(1) × Sp(1)-representations to a single Sp(1)-representation. The way to proceed is to leave the V1L -factor in each of these expressions alone and consider the representations of the diagonal subgroup Sp(1)RP Q . From these we want to obtain a single Sp(1) × Sp(1)-representation which leaves the left H-action intact.In terms of Sp(1)-representations. we consider the diagonal action of the 70 . we form the isomorphism RQ † P (Um )∗ ⊗ ιUn (Un ) ∼ aVm ⊗ b(V1L ⊗ Vn+1 ). In so doing.9). and restrict to the higher weight † subspace ιUm (Um ) ⊗ (Un )∗ . To obtain a stable AH-module. If we go down the left hand side. as we restrict our attention to particular subspaces.

Case 1( m and n both even): Let m = 2p. Since we do this for different diagonal subgroups we expect to be left with different subspaces. We will show that this is in fact the case. we obtain the two decompositions min{m+1. Thus any stable AH-module of virtual dimension k must be a sum of at least k/2 and at most k irreducibles. = Proof. depending on whether the irreducibles are odd or even. At each ‘half-way stage’ we are considering representations of diagonal subgroups of different pairs of groups. and then RP Q consider the action of Sp(1) on this. These decompositions contain fairly similar summands. This would fit well with the observation that stable AH-modules arise as representations of highest weight in decompositions of tensor products of Sp(1)-representations.† subgroup Sp(1)RQ . The irreducible stable AH-module whose dimension 71 . Un be irreducible stable AH-modules. One thing that we can guarantee for any m.14 we find that dim Um ⊗H Un = 4(p + q + 1) and that the virtual dimension of Um ⊗H Un is equal to 1. However. But any stable AH-module whose virtual dimension is equal to 1 must be irreducible. Using the Clebsch-Gordon formula. n = 2q. = If m and n are both odd then Um ⊗H Un ∼ 4Um+n .2. n}. restrict to the higher weight subspace (Um )∗ ⊗ ιUn (Un ). We want to know which parts end up contributing to the final Sp(1)RP Q -representation whichever path we take. n > 0 is that the representation with highest weight will be the same in both cases — both expressions have leading summand Vm+n+1 .n+1} RP Q Vm+1+n−2j j=0 ⊗ Q Vn ∼ = and P Vm ⊗ RQ Vn+1 ∼ = j=0 P RQ Vm+n+1−2j . using Joyce’s dimension formulae for stable AH-modules. just because we have two Sp(1)RP Q -representations of the same weight. We have already noted that each irreducible representation of the Sp(1)M -action on a stable AH-module U contributes 1 to the virtual dimension of U .1 Let Um . We will deal with the three possible cases in turn. If m or n is even then Um ⊗H Un ∼ Um+n . Then dim Um = 4(p + 1) dim Um = 2p + 3 dim Un = 4(q + 1) and dim Un = 2q + 3. We conjecture that this is the summand which we find in Um ⊗H Un . Here is the main result of this section: Theorem 5.n} RP Vm+1 min{m. This will † † identify the subspace Um ⊗H Un ⊆ H ⊗ (Um )∗ ⊗ (Un )∗ . and in both cases we take the higher weight representation in a sum Vk+1 ⊕ Vk−1 and discard the Vk−1 part.1. Using Theorem 4. with differences arising as the index j approaches the region of min{m. we cannot say that they are automatically the same subspace of H ⊗ (U † )∗ ⊗ (V † )∗ .

Let m = 2p − 1.is 4(p + q + 1) and whose virtual dimension is 1 is V1 ⊗ V2(p+q)+1 = Um+n . Hence Um ⊗H Un = Um+n . this is a map PQ PQ P Q Vm ⊗ Vn ∼ Vm+n ⊕ Vm+n−2 ⊕ . Hence Um ⊗H Un ∼ 2V1 ⊗ V2p+2q = Um+n . → Vm+n . n = 2q − 1. Thus Um ⊗H Un must be either an even irreducible or a sum of two odd irreducibles. Then dim Um = 4(2p + 1) dim Um = 4(p + 1) dim Un = 4(2q + 1) and dim Un = 4(q + 1). n = 2q − 1. Using Theorem 4. Case 2( m even and n odd): Let m = 2p. . This is really what this whole section has been about — the idea that the behaviour of stable AH-modules can be thoroughly and flexibly described by taking subrepresentations of highest weight in tensor products of Sp(1)-representations. For = m = 2p − 1 and n = 2q − 1 this becomes Q RP Q RP Q RP 4V1L ⊗ V2p ⊗ V2q−1 ∼ V1L ⊗ 4(V2p+2q−1 ⊕ V2p+2q−3 + . = (5. H 0 (O(n)) ∼ Vn . Then dim Um = 4(p + 1) dim Um = 2p + 3 dim Un = 4(2q + 1) and dim Un = 4(q + 1).7 applied to non-negative vector bundles.12).15) The only way Um ⊗H Un can have a virtual dimension of four and a total dimension of 16(p + q) is if Um ⊗H Un ∼ 4V1 ⊗ V2p+2q−1 = 4Um+n . 72 . as all the other irreducibles of the Sp(1)RP Q = action have smaller dimension. Q RP † Consider the space ιUm (Um ) ⊗ (Un )∗ ∼ a(V1L ⊗ Vm+1 ) ⊗ bVn of Equation (5. = Case 3 ( m and n both odd): The argument is very similar to that of Case 2. .1.1. so we cannot have more than 2 of the irreducibles of the Sp(1)RP Q -action. We also need a total dimension of 4(2p + 2q + 1). The isomorphism O(n) ⊗O = O(m) ∼ O(n + m) induces a map of cohomology groups H 0 (O(m)) ⊗ H 0 (O(n)) → = H 0 (O(m + n)).) = (5.14) The virtual dimension of the tensor product Um ⊗H Un must be equal to 2.5.14) we see that the only way this can occur is if Um ⊗H Un ∼ V1 ⊗ 2V2p+2q . = The map in question is projection onto the irreducible of highest weight Vn+m . Examining Equation (5.14 we find that dim Um ⊗H Un = 16(p + q) and that the virtual dimension of Um ⊗H Un is equal to 4. For the canonical sheaves O(n) over CP 1 . Q RP † Consider the space ιUm (Um ) ⊗ (Un )∗ ∼ a(V1L ⊗ Vm+1 ) ⊗ bVn of Equation (5. .12). .). Using Theorem 4.14 we find that dim Um ⊗H Un = 4(2p + 2q + 1) and that the virtual dimension of Um ⊗H Un is equal to 2. This result is parallel to Theorem 4. For = m = 2p and n = 2q − 1 this becomes Q RP Q RP Q RP 2V1L ⊗ V2p+1 ⊗ V2q−1 ∼ V1L ⊗ 2(V2p+2q ⊕ V2p+2q−2 ⊕ . In terms of Sp(1)-representations. . . = Quaternionic tensor products of more general stable AH-modules can be computed from this result by splitting into irreducibles and using the fact that the quaternionic tensor product is distributive for direct sums.

5. We use the Clebsch-Gordon formula to describe   min{m+1.2. we only expect a non-zero intersection if m < n + 2. we can describe what is going on in terms of diagonal actions × on tensor products of Sp(1)-representations. Let × † × Um = aV1 ⊗ Vm+1 and Un = (bV1 ⊗ Vn+1 )× . This = = gives rise to the standard descriptions H ⊗R (U † )∗ ∼ aV L ⊗ V R ⊗ V P ∼ aV L ⊗ (V RP ⊕ V RP ) = = m 1 1 m 1 m+1 m−1 and RQ RQ Q × H ⊗R ((Un )† )∗ ∼ bV1L ⊗ V1R ⊗ Vn+2 ∼ bV1L ⊗ (Vn+3 ⊕ Vn+1 ). our task is to find which of these summands is in the × × † × × intersection Um ⊗H Un = ιUm (Um ) ⊗R ((Un )† )∗ ∩ (Um )∗ ⊗R ιUn (Un ). In this case we might expect dim U ⊗H V = dim A + dim B − dim(H ⊗ (U † )∗ ⊗ (V † )∗ ). A similar calculation shows that dim A + dim B ≥ dim H ⊗ (U † )∗ ⊗ (V † )∗ if and only if s(j + r) ≥ kr. so now dim V = 2k + s. From our dimensional arguments. n + 2} = min{m. n + 1}. so (Um )∗ ∼ Vm and ((Un )† )∗ ∼ Vn+2 . Let U and V be antistable AH-modules with dim U = 4j. Let A = ιU (U ) ⊗ (V † )∗ and let B = (U † )∗ ⊗ ιV (V ). Then we would × expect that dim(Um ⊗H Un ) = 2ab(m − n + 2).2. This time since min{m + 1.n+1}  RP Q Vm+n+1−2j  . we take the smaller summand of Vn+3 ⊕ Vn+1 .14. = = The only difference between this and equation (5.3 Tensor Products of Antistable AH-modules It is not difficult to extend Joyce’s results for tensor products of stable AH-modules to irreducible antistable AH-modules — we can follow the same argument as in the proof of Theorem 4. This suggests that   {0} if m ≥ n + 2 × U× n or m even and m < n + 2 Um ⊗H Un ∼ (abV1 ⊗ Vn−m+1 )× = =  n−m × 4Un−m n and m both odd and m < n + 2. For × example. and thus to obtain ιUm (Um ).16) 73 .1. so dim(H ⊗ (U † )∗ ⊗ (V † )∗ ) = 4(2j + r)(2k + s) > dim A + dim B. Suppose instead that V is stable. Thus in generic situations we would expect dim A ∩ B = dim U ⊗H V = 0. n + 1} = m. let U = Um = (aV1 ⊗ Vm+1 )× and V = Un = bV1 ⊗ Vn+1 . and we would × predict that Um ⊗H Un contains the summand Vn−m+1 . dim V = 4k and dim V = 2k − s. in which case min{m. We will illustrate the case Um ⊗H Un .9) is that we have an antistable AH× × module involved. we can guarantee that the summand of smallest weight will appear in both expressions. As with stable AH-modules. dim U = 2j −r. Um ⊗H Un cannot be stable. since the generic properties of sums and intersections guaranteed by stability also hold if one or both of the AH-modules is irreducible and antistable. Because its virtual dimension is × not positive.n+2} × ιUm (Um ) ⊗R ((Un )† )∗ ∼ abV L ⊗ V RP ⊗ V Q ∼ abV L ⊗  = n+2 = 1 m+1 1 j=0 RP Q Vm+n+3−2j  and  RQ † × P × (Um )∗ ⊗R ιUn (Un ) ∼ abV1L ⊗ Vm ⊗ Vn+1 ∼ abV1L ⊗  = = j=0 min{m. (5. Then dim A = 4j(2k + s) and dim B = 4k(2j + r). As in Section 5.

Recall from Section 4.3 Semistable AH-modules and Sp(1)-representations In this section we shall consider how the AH-module Xq fits into the picture of stable AH-modules and Sp(1)-representations.2.14. we will confirm these conjectures by appealing to Quillen’s powerful results. U ) ∼ (Xq ⊗H U ) = 74 . = × × × If m < n + 2 and m and n are both odd then Um ⊗H Un ∼ 4Un−m . For any AH-module U . † † Xq = H and Xq = {p ∈ H : pq = −qp} so that Xq ∼ (Xq )∗ ∼ Cq . In particular. = = × ∼ If we consider the dual AH-module Xq = (H. Then HomAH (Um .2 Let Um be an irreducible stable AH-module and let Un be an irreducible antistable AH-module.1. we see that there are always AH-morphisms from Un into H (and indeed. Proof.2. this is a defining property for AH-modules).5.2.9 might be particularly interesting in the case of Xq .9. but never AH-morphisms from stable AH-modules into antistable AH-modules. = × × × × Let Um and Un be antistable irreducible AH-modules.2. using the isomorphism HomAH (Um . × Proposition 5. we can now see that there are always AH-morphisms from antistable AHmodules into stable AH-modules. Proof.3 Let Um and Un be stable AH-modules. × If m ≥ n + 2 then Um ⊗H Un = {0}.Rather than try to emulate Joyce’s (difficult) proof of Theorem 4. Proposition 5. This follows immediately by combining Theorems 4. n are even or odd. there is a canonical isomorphism × HomAH (Xq . We can use this result about tensor products of antistable AH-modules to tell us about AH-morphisms between stable AH-modules. since U0 = H. Then Um ⊗H Un = {0}. Similarly. 5.3 that for q ∈ S 2 .2. This = × suggests that Theorem 4.7 (due to Quillen).2. This follows from Theorem 4. where as usual a = 1 or 2 depending on whether m.2. Un ) ∼ = × (aUm−n ) = aVm−n {0} n≤m n>m where a = 4 if m and n are both odd and a = 1 otherwise. but never AH-morphisms from H into Un unless n = 0. × If m < n + 2 and m or n is even then Um ⊗H Un ∼ Un−m . Un ) ∼ (Um ⊗H Un ) = × of Theorem 4. Cq ) we see that the left-multiplication Lq : H → H defined by Lq (p) = q · p gives an AH-isomorphism Xq ∼ Xq .1. using the correspondences × Um = η + (aO(m)) and Un = η − (aO(−n − 4)).9 and 5.

Then we need φ(1) = u ∈ U and φ(q) ∈ U . . the subspaces Cq and Xq = C⊥ q are acted on by the Cartan subgroup U (1)q ⊂ Sp(1).3. φ(q) = qu. x⊗an−1 b+y⊗an . .1 The subspace Un−1 ∩ qUn−1 is given by the sum of the weight spaces of Q with highest and lowest possible weights. and so u ∈ U ∩ qU . y⊗bn }. This decomposition gives important information about the action of q on the AH-module Un = aV1 ⊗ Vn+1 . . . Recall from Section 5. Thus each basis vector wk is a weight vector with weight (n − 2k + 1). This also gives the left action of q ∈ Sp(1) on V1 . Proof. . It follows that HomAH (Xq .1. n − 2. .1. Since φ is H-linear. U ) ∼ (Xq ⊗H U ) ∼ U ∩ qU . Lemma 5. x ⊗ abn−1 . .2. . y ⊗ abn−1 . Any irreducible representation of sp(1) is also a representation of the subalgebra u(1)q = Q . C) such that Q = iH. . y ⊗ bn . . y for the space V1 in such a way that Q(x) = ix and Q(y) = −iy.and so × × HomAH (Xq . For ease of notation we work with the AH-module Un−1 . the second of these isomorphisms is given by the map (idU ⊗H χq ) : (U ⊗H Xq ) → (U ⊗H H) ∼ U . = Though Xq is not itself an Sp(1)-representation. this is exactly the same as the decomposition of Vn into eigenspaces of H with weights {−n.17) As noted by Joyce (see the summary in Section 4. . The goal of this discussion is to describe the AH-module Un ⊗H Xq . Let Q = aI1 + bI2 + cI3 ∈ sp(1) with a2 + b2 + c2 = 1 and let q = ai1 + bi2 + ci3 ∈ S 2 . We define a basis x. = = (5. and that the space (Un−1 ) ∼ Vn+1 is spanned by the vectors = {x⊗an . x ⊗ bn y ⊗ an . . This is exactly what we have done in the explicit calculations of Section 5.1. C) is given by H(x ⊗ an−k bk ) = (n − 2k + 1)x ⊗ an−k bk and H(y ⊗ an−k bk ) = (n − 2k − 1)y ⊗ an−k bk . As we shall see. Define the basis vectors wk ≡ x ⊗ an−k bk + y ⊗ an−k+1 bk−1 . . . = = = × Let φ : Xq → U be an AH-morphism. . taking the tensor product of a stable AH-module with the AH-module Xq serves to restrict attention from information about Sp(1)-representations to information concerning representations of the group U(1)q and its Lie algebra u(1)q . U ) ∼ (Xq ⊗H U ) ∼ (Xq ⊗H U ) . We do this with the aid of the following lemma. U ) ∼ HomAH (Xq . x⊗an−k bk +y⊗an−k+1 bk−1 . If we choose an identification sp(1) ⊗R C = sl(2.3 that V1 ⊗ Vn = x ⊗ an . The analysis of Vn as a U(1)q -representation is already familiar: it is the decomposition of Vn into weight spaces of the operator Q : Vn → Vn . −n + 2. . since the actions of q and Q coincide for this representation (see Section 1. This gives the left action of q ∈ H on Un = aV1 ⊗ Vn+1 .1). y ⊗ an−1 b. . . x ⊗ an−1 b. The two extreme vectors x ⊗ an and y ⊗ bn are also 75 . including w0 = x ⊗ an and wn+1 = y ⊗ bn . . x⊗bn +y⊗abn−1 . The action of H ∈ sl(2.3). .3 for the case q = i1 . n}.

. Since the AH-submodule (id ⊗H χq )(Un−1 ⊗H Xq ) ⊂ Un−1 is the subspace generated over H by U ∩ qU . Thus the quaternionic tensor product Un ⊗H Xq picks out the representations of extreme weight in the decomposition of Un into weight spaces of Q. and q( It is evident that q(x ⊗ an ) = ix ⊗ an ∈ Vn+1 but for all the other basis vectors wk . . . q(wk ) ∈ wk Hence Vn+1 ∩ qVn+1 = x ⊗ an . bn+1 . x ⊗ an b. Since each vector wk has a different weight from all the others we have q(wk ) ∈ Vn+1 if and only if q(wk ) ∈ wk . λk wk ) ∈ Vn+1 ⇐⇒ λk = 0 or q(wk ) ∈ Vn+1 for all k. On the basis vectors wk = x ⊗ an−k bk + y ⊗ an−k+1 bk−1 we have q(x ⊗ an−k bk + y ⊗ an−k+1 bk−1 ) = ix ⊗ an−k bk − iy ⊗ an−k+1 bk−1 . The left H-action of q on V1 ⊗ Vn is given by q(x ⊗ an−k bk ) = ix ⊗ an−k bk q(y ⊗ an−k bk ) = −iy ⊗ an−k bk . x ⊗ abn .3.2 Let Un = aV1 ⊗Vn+1 be an irreducible stable AH-module. If Un is an even AH-module then Un ⊗H Xq ∼ Xq and = (id ⊗H χq )(Un ⊗H Xq ) = H · x ⊗ an+1 − y ⊗ bn+1 R .weight vectors with weights n + 1 and −n − 1 respectively. 76 . If Un is an odd AH-module then Un ⊗H Xq ∼ 2Xq and = (id ⊗H χq )(Un ⊗H Xq ) = H · x ⊗ an+1 . Taking the σ-invariant subspace x ⊗ an − y ⊗ bn if Un−1 is an even AH-module yields the desired result for the AHmodule Un−1 . . Note that all these weights are different. . y ⊗ bn+1 R = V1 ⊗ an+1 . and q(y ⊗ an ) = −iy ⊗ an ∈ Vn+1 . y ⊗ bn+1 . x ⊗ bn+1 y ⊗ an+1 . . y ⊗ an b. . y ⊗ abn . this demonstrates the main result of this section which describes Un ⊗H Xq as follows: Theorem 5. and so q(wk ) ∈ Vn+1 . the weight spaces with highest and lowest weight. where the H-action is induced by the H-action on V1 . Left multiplication by q therefore preserves the weight space decomposition with respect to Q of a vector w ∈ V1 ⊗ Vn . whose eigenspace decomposition with respect to Q ∈ sp(1) takes the form V1 ⊗ Vn+1 = x ⊗ an+1 . . y ⊗ bn .

5.4

Examples and Summary of AH-modules

By now, the reader should be familiar with the ideas of quaternionic algebra. In this section we shall briefly sum up this information, giving explicit constructions of the AHmodules which will occur most frequently in the following chapter, in the forms in which they occur most naturally. Example 5.4.1 Recall the AH-module Y = {(q1 , q2 , q3 ) : q1 i1 + q2 i2 + q3 i3 = 0} of Example 4.1.2. Since Y is stable and has virtual dimension 1, it follows that Y is irreducible and is isomorphic to U2 = V1 ⊗ V3 . A calculation shows that (Y † )∗ ∼ V2 and = that the equation q1 i1 + q2 i2 + q3 i3 = 0 is precisely the condition for (q1 , q2 , q3 ) to lie in the subspace V1L ⊗ V3RM of H ⊗ V2M ∼ V1L ⊗ V1R ⊗ V2M ∼ V1L ⊗ (V3RM ⊕ V1RM ). = = k k Consider the AH-modules Y , SH Y and Λk Y . Using the dimension formulae of H H n Theorem 4.1.14 and Proposition 4.1.16, we discover that Λn Y = {0} and that H HY = n n n SH Y with dim(SH Y ) = 4(n + 1) and dim(SH Y ) = 2n + 3. From this we deduce that n SH Y ∼ U2n , and that all the even irreducible stable AH-modules can be realised as tensor = powers of the AH-module Y . Example 5.4.2 [J1, Example 10.1] Let Z ⊂ H ⊗ R4 be the set Z = {(q0 , q1 , q2 , q3 ) : q0 + q1 i1 + q2 i2 + q3 i3 = 0}. Then Z ∼ H3 is a left H-module. Define a real subspace Z = {(q0 , q1 , q2 , q3 ) : qj ∈ = I and q0 + q1 i1 + q2 i2 + q3 i3 = 0}. Then dim Z = 12, dim Z = 8 and Z is a stable AH-module. In fact, Z is isomorphic to the first odd irreducible AH-module U1 ∼ 2V1 ⊗ V2 . We = have (Z † )∗ ∼ R4 ∼ 2V1 . The equation q0 + q1 i1 + q2 i2 + q3 i3 = 0 is the condition for = = (q0 , q1 , q2 , q3 ) to lie in the subspace 2V1L ⊗ V2RM of H ⊗ 2V1M ∼ V1L ⊗ V1R ⊗ 2V1M ∼ = = 2V1L ⊗ (V2RM ⊕ V0RM ). Example 5.4.3 We can obtain the rest of the odd AH-modules as tensor products of those above. From Theorem 5.2.1 we know that U2n+1 = U2n ⊗H U1 . This combined with the previous examples shows that U2n+1 ∼ SH Y ⊗H Z. = n Thus we have obtained all the irreducible stable AH-modules in terms of previously known AH-modules. We can also use these constructions to write down formulae giving all those irreducible antistable AH-modules which are dual to stable AH-modules. Example 5.4.4 Consider the AH-module (U, U ) where U = H2 and U = (1, 0), (i1 , i2 ), (0, 1), (0, i1 ) .

Then {(0, q) : q ∈ H} is an AH-submodule of U isomorphic to Xi1 . However, there is no complementary AH-submodule V such that U ∼ Xi1 ⊕ V , and indeed U is irreducible. = 77

It is also clear that U is not semistable — nor is it the dual AH-module of any semistable AH-module. We call such an AH-module irregular. n,α n,− ¯ −1 Irregular AH-modules correspond to type (i) SK-modules of the form X1 ⊕ X1 α and torsion sheaves of the form O/mn , where n ≥ 2. In the present case which singles x out the subfield Ci1 , we have U ∼ η + (O/m2 = {0,∞} ). For a general formula linking pairs of antipodal points {z, σ(z)} in CP 1 and complex subfields of H see [Q, §14]. These are the only AH-modules which we do not describe in detail, for two reasons. Firstly, they are badly behaved compared to semistable and antistable AH-modules. Secondly, as far as the author is aware, they do not arise naturally in geometrical situations in the way that the other AH-modules do. Example 5.4.5 There is one remaining irreducible to consider — the AH-module (H, {0}). This trivially satisfies Definition 4.1.1, and so is an AH-module. Any AHmodule whose primed part is zero is a direct sum of copies of (H, {0}). It is easy to see that for any irreducible AH-module U , U ⊗H (H, {0}) = {0} unless U = H, in which case we have H⊗H (H, {0}) = (H, {0}). Though badly behaved, the AH-module (H, {0}) does arise naturally in quaternionic algebra — for example, Z⊗H H× = (H, {0}). This suggests the notation
× (H, {0}) = U−1 ,

in which case this result agrees with Theorem 5.2.2. Such notation is consistent with the sheaf description of antistable AH-modules, since we have
× U−1 = η − (2O(−3))

as expected. Thus we interpret (H, {0}) as the ‘antistable part’ of 2V1 ⊗ V0 . The dual space (H, H) is of course not an AH-module, so there is no stable AH-module U−1 which × × is dual to U−1 . In spite of this we still regard U−1 as ‘antistable’, because treating it as an exception every time would be cumbersome. We end this chapter with a diagram (overleaf) summarising much of our theory.

78

Figure 5.1: Irreducible AH-modules and the ratio of their dimensions.

AH-morphisms
-

Y × ∼ U2 = ×

Xq and irregular AH-modules

Y ∼ U2 =

Z × ∼ U1 =
× U−1 = (H, {0})

Z ∼ U1 = H
3 4


1 4

c c

c

0

...... 

     

1 2

......
s d

c c

dim U / dim U 1

× Un (3 ≤ n < ∞)

Un (3 ≤ n < ∞)

d d 

Quaternionic tensor products

This describes the AH-modules we have met so far — all the finite-dimensional irreducible AH-modules. The quaternionic tensor product of two AH-modules is always to the left of both of them in Figure 5.1. On the other hand, there are AH-morphisms from an AH-module into itself and any AH-modules to its right — and never from an AHmodule to any AH-module to its left. These statements are closely linked by Theorem 4.2.9. If U and V are irreducible stable or antistable AH-modules, this demonstrates that that there will always be AH-morphisms from U ⊗H V into U and V , but never any AH-morphisms from U or V into U ⊗H V , unless U or V is equal to H.

79

showing that the q-holomorphic sections of a q-holomorphic vector bundle E form an 80 . q-holomorphic functions possess a rich algebraic structure.1) summarises Joyce’s theory of q-holomorphic functions. the other two being original). A function f is q-holomorphic if and only if its differential df takes values in the subspace A ⊂ H ⊗ T ∗ M . This condition is met if and only if E carries an antiself-dual connection which is compatible with the structure of E as an AH-bundle. Because of the noncommutativity of the quaternions. It is in three parts (the first of which is a summary of Joyce’s work. A q-holomorphic function on a hypercomplex manifold M is a smooth function f : M → H which satisfies a quaternionic version of the Cauchy-Riemann equations. but are most interesting when M is hypercomplex. This is very similar to the situation in complex geometry where the differential of a holomorphic function takes values in the holomorphic cotangent space Λ1. We investigate the algebraic structure of q-holomorphic sections in some detail. An AH-bundle E is said to be q-holomorphic if it is simultaneously holomorphic with respect to the whole 2-sphere of complex structures on M . The first part (Section 6. which we write H ⊗ T ∗ M ∼ A ⊕ B. The third part (Sections 6.3 and 6. we use the Sp(1)-representation T ∗ M ∼ 2nV1 = defined by the hypercomplex structure to obtain a natural splitting of the quaternionic cotangent space of a hypercomplex manifold M . We let PM denote the AH-module of q-holomorphic functions on M . so that a q-holomorphic function is precisely a q-holomorphic section of the trivial bundle M × H.0 . Joyce attempts to capture this using the concept of an H-algebra.4) is about AH-bundles. They can be defined over any smooth manifold M .2). its complement B being the q-antiholomorphic cotangent space. the product of two q-holomorphic functions is not in general q-holomorphic. and the Sp(1)-version of quaternionic algebra gives a complete description of the geometric situation. We generalise the theory of q-holomorphic functions to that of q-holomorphic sections. It follows that the subbundle A should be thought of as the q-holomorphic cotangent space of M . In the second part (Section 6. The q-holomorphic functions on the hypercomplex manifold H are precisely the regular functions of Fueter and Sudbery. a quaternionic version of a commutative algebra over the real or complex numbers. = ∼ aV1 ⊗ (Vk+1 ⊕ Vk−1 ) used in the This is precisely a version of the splitting H ⊗ aVk = previous chapter to construct and describe all stable and antistable AH-modules.Chapter 6 Hypercomplex Manifolds This chapter uses the quaternionic algebra developed in Chapters 4 and 5 to describe hypercomplex manifolds. which are smooth vector bundles whose fibres are AH-modules. Nonetheless.

Joyce has used this process to reconstruct hypercomplex manifolds from their H-algebras. we see that ιq (z) = ι−q (¯). H) is an AH-module. Note that we have assumed no geometric structure on M other than that of a smooth manifold. 1 81 . H). I) then θm (f ) ∈ I for all m ∈ M . so θm ∈ C ∞ (M. the function ιq (f ) = a + qb is a quaternion-valued function on M . Suppose that f ∈ C ∞ (M. but the two maps are not identical: since ιq (a + ib) = a + qb and ι−q (a + ib) = a − qb. H) → H by θm (f ) = f (m). to which the reader is referred for more detail and proofs of the important results. Then θm (q · f ) = q · θm (f ). Then for every q ∈ S 2 . §§3. In this way we obtain quaternion-valued functions from complex ones.1. H) ) is an AH-module. H)† . Lemma 6. e1 ∈ E ). Let M be a smooth manifold and let C ∞ (M. Complex and quaternionic functions For each q ∈ S 2 let ιq : C → H be the inclusion obtained by identifying i ∈ C with q ∈ S 2 . 1 For any real vector space E. 6. the inclusion ιq extends to a map ιq : C ⊗ E → H ⊗ E given by ιq (e0 + ie1 ) = e0 + qe1 (for e0 . and we shall soon see that this construction can be used to obtain q-holomorphic functions from holomorphic ones.1 Define a linear subspace C ∞ (M. f ∈ C ∞ (M. Note that the images of ιq and ι−q are the same. For each m ∈ M . C ∞ (M.5]. f (m) = 0 for all m ∈ M . H) is a left H-module. so θm ∈ C ∞ (M. H) be the vector space of smooth quaternion-valued functions on M . Since θm ∈ C ∞ (M. H).1.H-algebra module over PM . An H-action on C ∞ (M. H) = {f ∈ C ∞ (M.1 Q-holomorphic Functions and H-algebras Quaternion-valued functions This section is mainly a summary of [J1.1 6. H)† is extremely useful. Thus C ∞ (M. z Let f = a + ib be a complex-valued function on M . if f ∈ C ∞ (M. H)† . I). H) : f (m) ∈ I for all m ∈ M } = C ∞ (M. The map ιq will be distinguished from the inclusion map ιU of AH-modules by the use of lower-case rather than upper-case subscripts. Proof. H)× . and α(f ) = 0 for all α ∈ C ∞ (M. Also. Thus C ∞ (M. H)† . H) is an AH-module. H) is defined by setting (q · f )(m) = q(f (m)) for all m ∈ M . This technique of linking a point m ∈ M to an element of C ∞ (M. H). With these definitions.1. We use the fact that M can be embedded in C ∞ (M. (C ∞ (M.1. H)† in the following way. What is less immediately obvious is that C ∞ (M. define the ‘evaluation map’ θm : C ∞ (M. by Definition 4. so that ιq (a + ib) = a + qb. so f ≡ 0.

H) . If f = x + iy is holomorphic with respect to Q then it satisfies the Cauchy-Riemann equations dx + Q(dy) = 0. the hypercomplex version of holomorphic functions. Then ιq (f ) = x + qy is a q-holomorphic function. Thus the q-holomorphic functions PM on a hypercomplex manifold M form an AH-module.2 Q-holomorphic functions In this section we will define q-holomorphic functions. in which case 0 = dx + (a1 I1 + a2 I2 + a3 I3 )(dy) = D(x + a1 yi1 + a2 yi2 + a3 yi3 ) = D(x + qy). A complex-valued function f = f0 + idf1 ∈ C ∞ (M. H) be a smooth H-valued function on M . Proof. H). Consider the complex manifold (M.1. The product of two q-holomorphic functions is not in general q-holomorphic — we can observe this simply by noting that all constant functions are trivially q-holomorphic. C) be holomorphic with respect to Q. The function f is said to be q-holomorphic if and only if it satisfies the Cauchy-Riemann-Fueter equations Df ≡ df0 + I1 df1 + I2 df2 + I3 df3 = 0. C) is holomorphic (with respect to I ) if and only if f satisfies the Cauchy-Riemann equations df0 + Idf1 = 0. Here is the definition of a q-holomorphic function on M . Let f = x + iy ∈ C ∞ (M. The term q-holomorphic is short for quaternion-holomorphic. I1 . Definition 6. If a function f is q-holomorphic then the function q · f is also q-holomorphic for all q ∈ H. and it is intended to indicate that a q-holomorphic function on a hypercomplex manifold is the appropriate quaternionic analogue of a holomorphic function on a complex manifold.3 Let q = a1 i1 + a2 i2 + a3 i3 ∈ S 2 and let Q = a1 I1 + a2 I2 + a3 I3 be the corresponding complex structure on M . by observing that every complex-valued function on M which is holomorphic with respect to any complex structure Q ∈ S 2 gives rise to a q-holomorphic function. We show that there are many interesting q-holomorphic functions on a hypercomplex manifold M . (6. Thus q-holomorphic functions do not form an algebra in the same sense that the holomorphic functions on a complex manifold do. I2 . The set of all q-holomorphic functions on M is called PM .2 Let f ∈ C ∞ (M. namely PM = {f ∈ PM : f (m) ∈ I for all m ∈ M }. Lemma 6. H).1.1) Let (M. Then f = f0 +f1 i1 +f2 i2 +f3 i3 with fj ∈ C ∞ (M ). 82 . I). So PM = PM ∩ C ∞ (M.1. and familiarise ourselves with some of their basic properties. I3 ) be a hypercomplex manifold.6. but PM is not even closed under right-multiplication by quaternions. The proof is a simple substitution. We adopt the obvious definition for PM . Thus the set of q-holomorphic functions PM forms a left H-submodule of C ∞ (M. and PM is an AH-submodule of C ∞ (M.

ιq (g). 6. Let F be a commutative field (usually the real or complex numbers). C) is holomorphic with respect to Q.1. producing the following definition [J1. b) = µ(b. saying that µ : A ⊗F A → A is commutative if and only if Λ2 A ⊆ ker µ. We shall see that this is a consequence of (indeed is equivalent to) the fact that the AH-modules Xq and X−q are the same. So our bilinear multiplication map µ : A × A → A becomes an F-linear map µ : A ⊗F A → A. 83 . equipped with an F-bilinear multiplication map µ : A × A → A which has certain algebraic properties. a) for all a.3. §5]: Axiom H. called the multiplication map. This algebraic situation is described by the theory of H-algebras. By Lemma 6. their product is also q-holomorphic. This is reminiscent of the situation described in Lemma 4. The commutative axiom µ(a. In this case we have two well-defined ‘products’ of f and g — their quaternionic tensor product f ⊗H g ∈ PM ⊗H PM and their product as Cq valued functions f g = gf ∈ PM . ‘AH-morphism’ and ‘quaternionic tensor product’.3. ιq (f ). if the functions f. this formulation is of no great use to us when seeking a quaternionic analogue because the non-commutativity of the quaternions makes the notion of an H-bilinear map untenable. a) becomes µ(a ⊗ b) = µ(b ⊗ a). and ιq (f g) = ιq (f )ιq (g) are all q-holomorphic. If f = x + iy ∈ C ∞ (M. then its conjugate ¯ f = x − iy is holomorphic with respect to −Q. so µ(a ⊗ b − b ⊗ a) = µ(a ∧ b) = 0. b) = µ(b.1.1. where two elements u ∈ U and v ∈ V such that ιU (u) ∈ Cq ⊗ (U † )∗ and ιV (v) ∈ Cq ⊗ (V † )∗ define an element u⊗H v ∈ U ⊗H V . In this special case where the q-holomorphic functions f and g both take values in a commuting subfield of H. b ∈ A then µ is said to be commutative. For example if µ(a. Cq ). replacing ‘vector space’. An F-algebra is a vector space A over F. ‘linear map’ and ‘tensor product’ with ‘AH-module’. Cq ) are both holomorphic with respect to Q. H). As we have already seen in Section 1. Does this mean that both x + qy and x − qy are q-holomorphic functions? Closer inspection shows that this is not the case — we see that ιq (x + iy) = ι−q (x − iy) = x + qy. The main feature of tensor algebra is that a bilinear map on the cartesian product A×B translates into a linear map on the tensor product A⊗F B. (ii) There is an AH-morphism µP : P⊗H P → P.7. So the holomorphic functions with respect to Q and −Q are mapped to the same functions when we consider their images under ι±q in C ∞ (M. then so is f g ∈ C ∞ (M. g ∈ C ∞ (M.So the H-valued function ιq (f ) = x + qy is q-holomorphic. Hence we obtain a ‘tensor algebra version’ of the commutative axiom. Now. The axioms for an algebra over F can alternatively be written in terms of tensor products. Once we have reformulated our axioms in terms of tensor products and linear maps we can translate them into quaternionic algebra.3 H-algebras and Q-holomorphic functions An H-algebra is a quaternionic version of an algebra over a commutative field. This is precisely what Joyce does. (i) P is an AH-module.

1Q for the identities in P. Here H-algebra stands for Hamilton algebra.2]. Write 1P . where Joyce defines this map and uses it to prove that U ⊗H V is an AHmodule. For infinite-dimensional vector spaces U and V we define U ⊗ V to consist of finite sums of elements u ⊗ v. Thus 1 is a multiplicative identity.PN (θm ⊗ θn ) · f ∈ H and we define φ(f ) : M × N → H by setting φ(f )(m. Then µP ◦ (µP ⊗H id) = µP ◦ (id ⊗H µP ).4 P is an H-algebra if P satisfies Axiom H. Thus µP is commutative. H (iv) The AH-morphisms µP : P⊗H P → P and id : P → P combine to give AH-morphisms µP ⊗H id and id ⊗H µP : P⊗H P⊗H P → P⊗H P. Let M .1.1.6 We define a map φ : PM ⊗H PN → C ∞ (M × N.(iii) Λ2 P ⊂ ker µP . Thus φ(f ) is a sum of finitely many smooth functions. which allow us to interpret points of m ∈ M as AH-morphisms θm : PM → H. (v) An element 1 ∈ A called the identity is given. In the same way.1. In this way the the H-algebra PM includes the algebras of holomorphic functions with respect to all the different complex structures on M . Then λPM .1.PN (θm ⊗ θn ) ∈ (PM ⊗H PN )† . Then λPM . Definition 6. Q respectively.2) for all x ∈ U † ⊗ V † and y ∈ U ⊗H V . Let M be a hypercomplex manifold. Then y · x ∈ H. Thus for each a ∈ P. n) = λPM . and such that for f. where ‘ · ’ contracts together the factors of U † ⊗ V † with those of (U † )∗ ⊗ (V † )∗ in U ⊗H V ⊂ H ⊗ (U † )∗ ⊗ (V † )∗ . This is associativity of multiplication.5 Let P. n) ∈ M × N the map θm ⊗ θn ∈ PM ⊗ PN .1. Define a linear map λU V : U † ⊗ V † → (U ⊗H V )† by setting λU V (x)(y) = y · x. H) as follows. 1⊗H a and a⊗H 1 ∈ P⊗H P by Lemma 4.7. Composing with µP gives AH-morphisms µP ◦ (µP ⊗H id) and µP ◦ (id ⊗H µP ) : P⊗H P⊗H P → P. Q be H-algebras. Joyce has constructed a multiplication map µM with respect to which the q-holomorphic functions PM form an H-algebra. we can associate † † to each point (m. Let f ∈ PM ⊗H PN . Let U and V be AH-modules. Cq ) holomorphic with respect to Q we have µ(f ⊗H g) = f g. We also define morphisms of H-algebras: Definition 6. 2 † Consider again the maps θm ∈ PM of Lemma 6. with 1 ∈ P and I·1 ⊆ / P. It is easy to see that φ(f ) is q-holomorphic. Let x ∈ U † ⊗ V † and y ∈ U ⊗H V . Definition 6. since DM ×N = See [J1.1. g ∈ C ∞ (M. N and (therefore) M × N be hypercomplex manifolds.PN (θm ⊗ θn ) · f. We define an AHmorphism φ : PM ⊗H PN → PM ×N . (6. 2 84 . (vi) Part (v) implies that if α ∈ P † then α(1) ∈ R. Definition 4. and so is smooth. Then µP (1⊗H a) = µP (a⊗H 1) = a for each a ∈ P. and let φ : P → Q be an AH-morphism. We say that φ is an H-algebra morphism if φ(1P ) = 1Q and µQ ◦ (φ⊗H φ) = φ ◦ µP as AH-morphisms P⊗H P → Q.

(6.3). Let 1 ∈ PM be the constant function on M with value 1. §10]. Then H is naturally a complex manifold with hypercomplex structure (I1 . 5.4. Thus the linear q-holomorphic functions on H form an AH-submodule of the set of all linear functions.DM (φ(f )) + DN (φ(f )) = 0. I2 dx3 = dx1 .7 [J1. H) is an H-algebra morphism. PM is an H-algebra. With these definitions. Thus we can define an AH-morphism µM = ρ ◦ φ : PM ⊗H PM → PM .3) Here is the key theorem of this section: Theorem 6. Thus we have a canonical AH-morphism φ : PM ⊗H PN → PM ×N . and each q-holomorphic function on M × M restricts to a q-holomorphic function on Mdiag .8 Q-holomorphic functions on H [J1. so φ : PM ⊗H PM → PM ×M . This statement is also true for C ∞ (M. An AH-morphism φ : C ∞ (M.20) with the hypercomplex structure in (6. The q-holomorphic functions PM form an H-subalgebra of C ∞ (M. j = 1. We can easily show that k SH Z ∼ (k + 1)V1 ⊗ Vk+1 = = 1 (k 2 (k + 1)Uk k even + 1)Uk k odd. x3 ) representing the quaternion x0 + x1 i1 + x2 i2 + x3 i3 . 2.16 to prove that SH Z ∼ = (k) U . the free H-algebra generated by Z . Consider the AH-module U (k) of homogeneous q-holomorphic polynomials of degree k.5].1. 3. and this idea enables us to write down the multiplication map µH . (6. It is clear that φ is H-linear and that φ(PM ⊗H PN ) ⊆ PM ×N . Let ρ be the restriction map ρ : PM ×M → PMdiag . and the inclusion map id : PM → C∞ (M. This constructs not only the spaces U (k) .1. H) → C ∞ (M × N. so that PM is an AHmodule.e. I3 dx1 = dx2 and Ij dx0 = dxj .1. which is naturally an H-algebra. x1 . Using the Cauchy-Riemann-Fueter Equation (1. H) can be defined just as in Definition 6. Joyce [J1. Consider the manifold H with coordinates (x0 . I2 . §6]. x2 . The spaces U (k) are important in quaternionic analysis and are studied by Sudbery k [Su. H). Example 6. Example 5. m) : m ∈ M } ⊂ M × M is a submanifold of M × M isomorphic to M as a hypercomplex manifold. H).6. I3 ) given by I1 dx2 = dx3 . then ρ is an AH-morphism. 85 . The space of all q-holomorphic polynomials on H is therefore given by the sum ∞ k Z k=0 SH Z. we find that f = q0 x0 + q1 x1 + q2 x2 + q3 x3 is q-holomorphic if and only if q0 + q1 i1 + q2 i2 + q3 i3 = 0. Joyce uses the dimension formulae of Proposition 4.4). and let µM : PM ⊗H PM → PM be the AH-morphism µM = ρ ◦ φ of Equation (6. The second (and easier) step is to consider the case M = N .1.2. i.4) Consider the linear H-valued functions on H. The full H-algebra PH is constructed by adding in convergent power series.1] calls this F . Let M be a hypercomplex manifold. and this submodule is isomorphic to the AH-module Z ∼ U1 of Example = 5. The diagonal submanifold Mdiag = {(m. we can evaluate the derivatives in the M and N directions separately. which is crucial to Sudbery’s approach. and ρ is still the obvious restriction. H)⊗H C ∞ (N. but also their structure as Sp(1)-representations.

The ease with which quaternionic algebra presents such close parallels with complex geometry could scarcely be more rewarding. • The mapping A ⊕ B → H ⊗ T ∗ M is an injective AH-morphism which is an isomorphism of the total spaces. We show that there is a natural splitting of the quaternionic cotangent space H ⊗ T ∗ M ∼ A ⊕ B. • Dim B = 4n and dim B = 0. B= −1 • Dim A = 12n and dim A = 8n. A function f ∈ C ∞ (M. 86 .0 M .1). where B = B ∩ I ⊗ T ∗ M . C) is holomorphic if df ∈ Λ1. and f is holomorphic if and only if ∂f = 0. Theorem 6. Let M be a hypercomplex manifold. The hypercomplex structure determines a natural splitting H ⊗ T ∗ M ∼ A ⊕ B.6. A function f ∈ C ∞ (M. Recall from Section 3. This is a situation with which we are by now very familiar. we have a splitting H ⊗ T ∗ M ∼ V1L ⊗ V1R ⊗ 2nV1G = ∼ 2nV L ⊗ (V RG ⊕ V RG ) = 1 2 0 ∼ A ⊕ B. B is an antistable AH-module. Using the theory of Chapter 5. It is not an AH-isomorphism because dim(A ⊕B ) is smaller than dim(I ⊗ T ∗ M ). where π B is the natural projection to the subspace B ⊂ H ⊗ T ∗ M .1 ◦ d. The bundle A is then the q-holomorphic cotangent = space of M . where A = A ∩ I ⊗ T ∗ M . H) is q-holomorphic if and only if df ∈ A.2 that T ∗ M ∼ 2nV1 as an Sp(1)-representation.1 M of Equation (2. we immediately deduce that • The (fibres of the) subspaces A and B are AH-modules with A ∼ nU1 and = ∼ nU × . This section presents the quaternionic analogue of this description. = where A = 2nV1L ⊗ V2RG and B = 2nV1L ⊗ V0RG . Call this = representation 2nV1G (since it is the action defined by the geometric structure of M ).2. Proof. = where dim A = 12n and dim B = 4n. Following the standard arguments of Chapter 5. The Cauchy-Riemann operator is defined by ∂ = π 0.1 Let M 4n be a hypercomplex manifold and let H ⊗ T ∗ M be the quaternionic cotangent bundle of M . and the ¯ Cauchy-Riemann-Fueter operator can be written δ = π B ◦ d. A is a stable AHmodule.0 M ⊕ Λ0.2 The Quaternionic Cotangent Space Let M be a complex manifold and recall the splitting C ⊗R T ∗ M = Λ1.

H) is q-holomorphic if and only if df ∈ C ∞ (A). We then show that the CauchyRiemann-Fueter operator can be written as π B ◦ d.2. obtained by coupling right-multiplication on H with the action of the hypercomplex structure. H). Firstly.6) ¯ Then for a function f ∈ C ∞ (M. 87 . A) δ = πA ◦ d and ¯ δ : C∞ (M. H) → C ∞ (M. (6.2 A function f ∈ C ∞ (M. If ω is in the −3 eigenspace then we have C(ω) = 0. where χ(ω) = I1 (ω)i1 +I2 (ω)i2 +I3 (ω)i3 .7) Moreover. The operator χ satisfies the equation χ2 = 3 − 2χ. (6.5) Then C(ω) = (I 2 +J 2 +K2 )(ω) = −6ω−2χ(ω). This is exactly the real part of the equation ¯ 4δf ≡ df − I1 df i1 − I2 df i2 − I3 df i3 = 0. Thus the subspaces A and B are the eigenspaces of the operator χ. H) → C ∞ (M. ¯ Proposition 6. df = δf + δf . Let f = f0 + f1 i1 + f2 i2 + f3 i3 be a q-holomorphic function on M . H) is q-holomorphic if and only if δf = 0. the three imaginary parts are the equations Ij (Df ) = 0. so ω ∈ V2RG . so has eigenvalues +1 and −3. 4 These maps depend only on the structure of M as a hypercomplex manifold. which are satisfied if and only if Df = 0. Thus the real equation Df = 0 is exactly the same as the ¯ quaternionic equation δf = 0. The diagonal Lie algebra action is given by I(ω) = I1 (ω) − ωi1 J (ω) = I2 (ω) − ωi2 K(ω) = I3 (ω) − ωi3 .4). B) ¯ δ = π B ◦ d. we define new differ¯ ential operators δ and δ: δ : C∞ (M.The importance of this splitting lies partly in the following result: Theorem 6. Just as we defined the Dolbeault operators in Equation (2.2. we investigate the projection operators π A and π B from H ⊗ T ∗ M onto A and B. so ω ∈ V0RG .2. (6. If ω is in the +1 eigenspace then we have C(ω) = −8.4 A function f ∈ C ∞ (M. Lemma 6. Proof. so Df = df0 + I1 df1 + I2 df2 + I3 df3 = 0.3 Projection maps π A : H ⊗ T ∗ M → A and π B : H ⊗ T ∗ M → B are given by 1 π A (ω) = (3ω+I1 (ω)i1 +I2 (ω)i2 +I3 (ω)i3 ) 4 1 π B (ω) = (ω−I1 (ω)i1 −I2 (ω)i2 −I3 (ω)i3 ). To work out projection maps to A and B we use the Casimir operator for the diagonal Sp(1)RG -action. The proof of this theorem will be in two parts. This enables us to write down expressions for π A and π B such that π A + π B = 1.

then ιq (f ) is a q-holomorphic function. 4n − 1} be a basis ∗ for Tx M such that Ib (e4a ) = e4a+b . On a hypercomplex manifold. Instead. .1. e4a+2 . but also of the holomorphic and q-holomorphic cotangent spaces. . as = in Theorem 6. Let q = a1 i1 + a2 i2 + a3 i3 ∈ S 2 and let Q = a1 I1 + a2 I2 + a3 I3 be the corresponding complex structure on M . . (6.2. . In other words. one of them has three times the dimension of the other. but they are not equal in size. we have a splitting into two subspaces. . echoing the fundamental isomorphism H ∼ I ⊕ R. This type = of splitting into spaces of dimension 3n and n is typical of the quaternions.8). On a complex manifold. . .2 and motivates the following definition: Definition 6. We have already seen (Lemma 6. With respect to this basis the fibres of A and B at the point x are given by qa ea : q4b + q4b+1 i1 + q4b+2 i2 + q4b+3 i3 = 0 for all b = 0. . This proves Theorem 6.8) is a higher dimensional version of the equation q0 + q1 i1 + q2 i2 + q3 i3 = 0 which gives the q-holomorphic cotangent space on H (Example 6. .8) for all b = 0. = Let us pause for a moment to reflect on what has happened. which are conjugate to one another. A).2.1.2. Let M 4n be a hypercomplex manifold and let {ea : a = 0. . . and ω ∈ B ⇐⇒ ω0 = I1 (ω1 ) = I2 (ω2 ) = I3 (ω3 ). = Bx = qa ea : q4b = −q4b+1 i1 = −q4b+2 i2 = −q4b+3 i3 6. It is straightforward to verify that ω ∈ A ⇐⇒ ω0 + I1 (ω1 ) + I2 (ω2 ) + I3 (ω3 ) = 0. e4a+1 . because the subspaces are defined by the equation V1 ⊗ V1 ∼ V2 ⊕ V0 . This is a correspondence not only of holomorphic and q-holomorphic functions.1. we have a splitting of C ⊗ T ∗ M into two spaces of equal dimension. .9) Equation (6. e4a+3 is a copy of H with its = standard hypercomplex structure. n − 1 Ax = and (6. we choose a particular isomorphism ∗ Tx M ∼ Hn such that the subspace e4a . This gives an AH-isomorphism between Ax and the AH-module nZ ∼ nU1 . Here are some practical methods for writing the spaces A and B. I2 . 88 . Then A is the q-holomorphic cotangent space of M and B is the q-antiholomorphic cotangent space of M . I3 ).5 Let M be a hypercomplex manifold and let H ⊗ T ∗ M ∼ A ⊕ B. Let ω = 1 ⊗ ω0 + i1 ⊗ ω1 + i2 ⊗ ω2 + i3 ⊗ ω3 .1 The Holomorphic and Q-holomorphic cotangent spaces Let M 4n be a hypercomplex manifold with hypercomplex structure (I1 .2.3) that if f is a complex-valued function on M which is holomorphic with respect to the complex structure Q. n − 1 .It now follows that a function f is q-holomorphic if and only if df = δf ∈ C ∞ (M.

i3 e2 + i2 e3 . Then A= Q∈S 2 H · ιq (Λ1.1 ).0 . . we have A = {(q0 e0 + . (6. we see that H · ιq (Λ1. . the general result holding in exactly the same way but involving more coordinates. Q Thus the q-holomorphic cotangent space is generated over H by the different holomorphic cotangent spaces. = Observe that −i2 · (i2 e0 − i3 e1 ) = e0 + i1 e1 and that −i2 · (i2 e2 − i3 e3 ) = e2 + i1 e3 .11) and the image (idA ⊗H χi1 )(A⊗H Xi1 ) ∼ 2Xi1 is generated over H by this subspace. This gives us a description of the q-holomorphic and q-antiholomorphic cotangent spaces of M in terms of complex geometry: Corollary 6. i2 e2 − i3 e3 . Using Q the embedding ιq we can consider C⊗T ∗ M ∼ Cq ⊗T ∗ M as a subspace of H⊗T ∗ M . + q3 e3 ) : q0 + q1 i1 + q2 i2 + q3 i3 = 0} and A = I ⊗ T ∗ M . with q-holomorphic cotangent space A and q-antiholomorphic cotangent space B.4). I1 Since ιi1 (a + ib) = a + i1 b.0 = e0 + ie1 . i3 e0 + i2 e1 . .7 Let M 4n be a hypercomplex manifold. e2 + i1 e3 .0 ) = H · e0 + i1 e1 . .0 Therefore H · (A ∩ i1 A ) is exactly the same as the subspace H · ιq (ΛI1 ) of Equation (6.6 Let A ⊂ H ⊗ T ∗ M be the q-holomorphic cotangent space of a hypercomplex manifold M and let Λ1.0 H · ιq (ΛQ ) = (idA ⊗H χq )(A⊗H Xq ). 1. For any point x ∈ M we choose a basis ∗ {e0 . + q3 e3 ∈ A : qj ∈ i2 .6.Each complex structure Q on M defines a holomorphic cotangent space Λ1.2. The = relationship between these subspaces of H ⊗ T ∗ M is clarified in the following Lemma: Lemma 6.2. It follows that A ∩ i1 A = {q0 e0 + . Use the fact that A is a stable AH-module. 89 . thus A is generated over H by the subspaces A ∩ qA for q ∈ I.0 ) Q and B= Q∈S 2 H · ιq (Λ0. We illustrate the case dim M = 4. Using Equation (6.8). i3 } = i2 e0 − i3 e1 .10) Recall from Section 5. e2 . e2 + ie3 C . Then Λ1. Then 1. Without loss of generality take q = i1 and consider the complex manifold (M. e3 } for Tx M such that the hypercomplex structure at x is given by the standard relations (6. Proof.0 ⊂ C ⊗ T ∗ M be the holomorphic cotangent space with Q respect to the complex structure Q. e1 . I1 ). Proof.2. I1 (6. The proof now follows immediately from Lemma 6.10).3 that (idU ⊗H χq )(U ⊗H Xq ) = U ∩ qU for any AH-module U .

Then V = x∈M Vx is a real vector bundle. π2 . M ) by setting (E⊗H F )m = k Em ⊗H Fm . H). The standard real or complex algebraic operations which give new bundles from old can be transfered en masse to quaternionic algebra and AH-bundles. M ) be an AH-bundle. Proof. and C ∞ (M. N ) be AH-bundles. A smooth AH-module bundle with fibre W or simply AH-bundle is a family {(Ex . M ). W ) a fixed AHmodule.1 Let M be a differentiable manifold and (W. In fact. as is E × H if the fibres of E are SAH-modules. The q-holomorphic and q-antiholomorphic cotangent bundles A and B are both AH-bundles. let E and F be AH-bundles over M .1. Ex )}x∈M of AH-modules parametrised by M . We can define a bundle E ⊕ F in the same way. By restricting to the diagonal submanifold of M × M we define a bundle (E⊗H F. and the left H-module of smooth sections of E is denoted by C ∞ (M. x∈M and the lemma is proved. E) is a left module over the ring C ∞ (M. Thus E is isomorphic to the AH-subbundle ιEx (Ex ) ⊆ H ⊗ V.2 Let E = (E. E) = C ∞ (M. π : Ex → x is C ∞ . Λk E and SH (E) are AH-bundles. Then E is isomorphic to an AH-subbundle of H ⊗ V . as is easy to demonstrate: Lemma 6. M × M ) or the diagonal restriction (E⊗H F.6. Define C ∞ (M.3 AH-bundles An AH-bundle is the natural quaternionic analogue of a real or complex vector bundle: a set of AH-modules E parametrised smoothly by a base manifold M . every AH-bundle can be regarded as an AH-subbundle of some bundle H ⊗ V . and H ⊗ V is an AH-bundle. 90 . With this definition. π1 . M ) and (F. Definition 6. M ×N ) by setting (E⊗H F )(m. π1 . • For every x0 ∈ M . We define a bundle (E⊗H F. Consider the map ιEx (Ex ) : Ex → H ⊗ (Ex )∗ for x ∈ M . there exists an open set U ⊆ M containing x0 and a diffeomorphism φU : π −1 (U ) → U × W such that φ(Ex ) → {x} × W is an AH-isomorphism for each x ∈ U . π. Let (E. it is easy to adapt Lemma 6. E ). In particular. We can multiply sections on the left by quaternion-valued functions. Let Vx = (Ex )∗ .n) = Em ⊗H Fn .1 to show that if E is an AH-bundle then C ∞ (M. for some real vector bundle (V. M ). We will always make clear whether an AH-bundle ‘ E⊗H F ’ refers to (E⊗H F. In the same way (E ⊕ F. This is an extension of the fact that any AH-module W is isomorphic to ιW (W ) ⊆ † † H ⊗ (W † )∗ . and are AH-subbundles of H ⊗ T ∗ M . M ).3. E).3. A section of an AH-bundle is a smooth map α : M → E such that α(x) ∈ Ex for all x ∈ M . together with a differentiable manifold structure on E = x∈M Ex such that • The projection map π : E → M . E) is an AH-module.

Ex ) is an AH-module for all x ∈ M . i2 + xi3 . For example. In the same way. H).3.1 Connections on AH-bundles Definition 6. Then P V is a principal bundle with fibre GL (k. R). H). Thus E is not an AHbundle. As we know. The following example shows why this definition would be unwieldy: Example 6.3. It is easy to check that (Ex . For example. G will be significantly smaller than GL (k. and V = P V ×GL(k. Definition 4. Principal bundles associated with AH-bundles tend to be much smaller than those associated with real or complex vector bundles. We will not distinguish between and ( † )∗ . 6.4]. if W = U2j for some j then by Theorem 4. So despite the fact that E is a smooth vector bundle. We will sometimes omit the subscript and just write ‘ ’ when the AH-bundle E is clearly implied. so AutAH (U2j ) = R∗ = R \ {0} = ∗ E and P is just a principal R -bundle.R) Rk . Define a real subspace E ⊂ E by Ex = i1 . if we considered this broader class of objects to be ‘AH-bundles’ then they would not be closed under the tensor product.3. E) −→ C ∞ (M. E). demanding only that E should be a real vector bundle possessing a left H-action and a real subbundle E such that (Ex . (E⊗H Xi3 )x = Xi3 if x = 0 0 otherwise. H). since the fibres Ex and Ey are not AH-isomorphic for x = y. E ⊗ T ∗ M ) = Ω1 (M. In most cases.9 we have HomAH (U2j . We could broaden our definition of an AH-bundle.4 Let E be an AH-bundle over M . let E be an AH-bundle with fibre W and let G = AutAH (W ). Let V be a real vector bundle over the manifold M with fibre Rk . E) E (f which satisfies the rule · α) = df ⊗ α + f · Eα for all α ∈ C ∞ (M. Ex ) = Xi3 −xi2 as an AH-module. Then the map id ⊗( † )∗ is E an AH-connection on H ⊗ (E † )∗ which preserves ιE (E). Let be an AH-connection on E. An AH-connection on E is an AH-morphism E : C ∞ (M. and let P V be the bundle of frames associated with V . There is good reason for excluding E from being an AH-bundle. E⊗H Xq is not. 91 . in very much the same way that Joyce constructs the AH-morphism φ⊗H ψ from AH-morphisms φ and ψ [J1. Thus every AH-connection on E is equivalent to a connection ( † )∗ on the real vector bundle (E † )∗ .2. if E is an AH-bundle we will often write ιE : E → † H ⊗ (E † )∗ for the set of inclusion maps x∈M (ιEx : Ex → H ⊗ (Ex )∗ ). This equivalence allows us to construct AH-connections on tensor products. U2j ) ∼ R. G will be a subgroup of GL (k.3 Let M = R.Abusing notation slightly. and define a real vector bundle E = R × H. We can define a principal G-bundle P E associated with E so that E = P E ×G W . but will write for both. f ∈ C ∞ (M.

define a differential operator E.5 Let (E. This process can be described in terms of principal bundles. be AHE⊗H F on the F Proof.N on ((F † )∗ . thus id ⊗ E.5. so that F = Q ×H V . §1. Adding these gives a differential operator E. there is an induced action of G × H on W ⊗H V .M ⊗id maps sections of ιE (E)⊗ † ∗ (F ) to sections of ιE (E) ⊗ (F † )∗ ⊗ T ∗ (M × N ). and so the AH-connection id ⊗( E. The same is true for id ⊗ id ⊗ F. In this way the action of G = AutAH (W ) on W gives rise to an action on (W † )∗ . M ) and (F. We call this connection E⊗H F . which we can identify with the AH-connection E⊗H F of Lemma 6. we can use this to guarantee that every AH-bundle has an AH-connection. M ) extends trivially to a bundle ((E † )∗ . M × N ) is the associated bundle (P × Q) ×G×H W ⊗H V . M × N ).5 to define an AH-connection on the diagonal bundle (M. the operator id ⊗ E. 3 A similar discussion for connections in complex vector bundles can be found in [K. = ∗ where E. on which it is therefore an AH-connection. M × N ) is an AHconnection. M × N ). Then E and F induce an AH-connection AH-bundle (E⊗H F.M ⊗ id + id ⊗ F. If E and F are AH-bundles over the same manifold M we can use Lemma 6.Lemma 6.M ⊗ id maps sections of E⊗H F to sections of E⊗H F ⊗ T ∗ (M × N ). Let E be an AH-bundle with fibre W .3.M ⊗ id + id ⊗ F. N ) be AH-bundles and let E and connections on E and F .N : C ∞ ((E † )∗ ⊗ (F † )∗ ) −→ C ∞ ((E † )∗ ⊗ (F † )∗ ⊗ T ∗ (M × N )). Let F be another AH-bundle with fibre V . so the operator id ⊗( E. E⊗H F ). This operator is a connection on the real vector bundle (E † )∗ ⊗ (F † )∗ . Since G × H acts as AHautomorphisms on W ⊗H V .3. and let P be the associated principal bundle with fibre G.N . Since the connections DP and DQ define a unique connection DP ⊕ DQ on the product bundle P × Q. In the same way define a connection F. An AH-connection E on E gives rise to a connection DP on the principal bundle P . 92 . An AH-connection F on F gives rise to a connection DQ on the principal bundle Q. and vice versa. Consider the principal bundle P × Q.N ) on the bundle (H ⊗ (E † )∗ ⊗ (F † )∗ . this gives rise to a connection on the associated bundle E⊗H F . Then φ induces a real linear isomorphism (φ× )∗ : (W † )∗ → (W † )∗ . Using the natural identification ∗ ∗ T(m.3. 3 Suppose that φ ∈ AutAH (W ) for some AH-module W .N ) preserves the AH-subbundle (E⊗H F. let H = AutAH (V ) and let Q be the associated principal bundle with fibre H.M differentiates only in the M -directions of M × N .5].M ⊗ id + id ⊗ F. a bundle over M × N with fibre G × H. M ×N ). The AH-bundle (E⊗H F. Since E is an AH-connection. M × N ) which differentiates only in the N -directions. Since G acts on (W † )∗ and H acts on (V † )∗ . We regard E as a connection on (E † )∗ . The bundle ((E † )∗ . this induced connection is an AH-connection. Since connections on principal bundles always exist. and acts trivially on the (F † )∗ factor. so that E = P ×G W . M × N ) over M × N .M on ((E † )∗ .n) (M × N ) ∼ Tm M ⊕ Tn N .

5 and 3. π. i. so E has the structure of a holomorphic vector bundle over (M. Each E 2 E Q ∈ S defines a splitting = ∂Q + ∂ Q . The existence of such a connection ensures that the different holomorphic structures are compatible. we can simply refer to the total space E as a q-holomorphic AH-bundle.1 defines a holomorphic structure on R = RQ + RQ + RQ . on the other hand.2 The connection gives E the structure of a holomorphic vector E E bundle if and only if ∂ ◦ ∂ = 0.e. Thus 2.1 ◦ : C ∞ (E) → C ∞ (E⊗Λ0. if a connection is already specified. 6. which makes the step-by-step correspondence with complex geometry extremely clear. This is Propositions 3. so R−Q = RQ . and so a decomposition of the curvature tensor 0.1 Let E be an AH-bundle over the hypercomplex manifold M .0 0.2 2 holomorphic vector bundle with respect to every Q ∈ S if and only if RQ = RQ = 0 for all Q ∈ S 2 .1 Anti-self-dual connections and q-holomorphic AH-bundles Let E be a complex vector bundle over the complex manifold (M.0 ◦ : C ∞ (E) → C ∞ (E⊗Λ1. Q) for all Q ∈ S 2 . 1) with respect to all Q ∈ S 2 .0 M ) and ∂ = π 0. we can define E ∂ E = π 1. Let E be a q-holomorphic AH-bundle over the hypercomplex manifold M . the (0.2 2. 9].4 Q-holomorphic AH-bundles Using complex geometry as a model.0 = Λ2 . 0. Conversely.4. If we reverse the sense of Q then we 2. We give an interpretation of q-holomorphic sections in terms of quaternionic algebra and the q-holomorphic cotangent space. Here is the main definition of this section: Definition 6. we define a q-holomorphic AH-bundle over a hypercomplex manifold M to be one which is holomorphic with respect to each complex structure Q ∈ S 2 . Then (E.2 gives E the structure of a reverse the decomposition of R. We define the q-holomorphic sections of − a q-holomorphic vector bundle using a version of the Cauchy-Riemann-Fueter equations. p. i. I) and let be a connection on E. which generalise those of Joyce’s q-holomorphic functions. Q) for every complex structure Q ∈ S 2 . By Proposition 6. An AH-bundle is q-holomorphic if and only if it admits a connection whose curvature takes values only in E2.2 E with respect to Q if and only if RQ ≡ 0. 2)-component of the curvature R of vanishes. If we want to be really specific we can write (E. A lot can be learned about q-holomorphic AH-bundles by studying the connection .0 0.4. Let be a connection on E and let R be the curvature of .0 1. ) for our q-holomorphic bundle.7 of [K. Proof. Suppose that gives E the structure of a holomorphic vector bundle over the complex manifold (M.2. M. if E is a holomorphic vector bundle then E admits such a connection. 93 .e. Q-holomorphic sections have interesting algebraic properties.4.6. Proposition 6. equipped with an AH-connection .1 M ). Just as we split the exterior differential d = ∂ + ∂. ) is a q-holomorphic AH-bundle. the curvature of is of type (1.4.

where the vertical subbundle E is naturally defined by the structure of E as an AH-bundle and the horizontal subbundle T M is defined by the connection.0 as self-dual rather than anti-self-dual.6). §2]. and that the volume form e0123 gives a positive orientation of H. On the horizontal subbundle isomorphic to T M . There is no reason why his definition cannot be extended to hypercomplex manifolds. I2 .0 ⊂ Λ2 T ∗ M . I3 ) on T E as follows. Verbitsky considers the case where B is a Hermitian vector bundle over a hyperk¨hler a manifold. define 94 . M. This is similar to the idea of hyperholomorphic bundles described by Verbitsky [V. if B is a hyperholomorphic bundle with a real structure σ such that B = B σ ⊗ C.4 Let (E. we call E2. particulary on quaternionic K¨hler manifolds: for example Galicki and Poon [GP].0 the space of anti-self-dual 2-forms Λ2 . the action of I1 . I2 and I3 is given by the fixed left H-action of i1 .4. Conversely. the Hermitian inner product is not necessary to force the connection to be anti-selfdual if it is to be compatible with the 2-sphere of holomorphic structures. Then E is itself a hypercomplex manifold.2 the space of self-dual 2-forms Λ2 and E2. Proof. Conversely. Any AH-subbundle E of H ⊗ B σ which is preserved by the anti-self-dual connection will also be q-holomorphic. This is partly because such connections give minima of a Yang-Mills funtional on M . π. It is important to note that Mamone Capria and Salamon refer to the bundle E2. and so refer to connections taking values in E2. Then (E.2. A connection whose + − curvature R takes values only in C ∞ (End(E)⊗E2. The following Proposition gives further insight into the analogy between holomorphic and q-holomorphic bundles: Proposition 6. 1) with respect to every Q ∈ S 2 if and only if it is annihilated by the action of sp(1) on Λ2 T ∗ M . We have established the following result: Theorem 6. Nitta [N] and a Mamone Capria and Salamon [MS].0 ) is therefore called an anti-self-dual connection. There is no unanimous convention in the literature: our choice is made because we are using the conventions that Ij (e0 ) = ej not −ej . since as we have seen. If E is a q-holomorphic AH-bundle then it is easy to see that H ⊗ (E † )∗ is also q-holomorphic. On the vertical subbundle isomorphic to E. Consider the splitting of T E into horizontal and vertical subbundles T E ∼ = E ⊕ T M . Define a hypercomplex structure (I1 . so the fibres of such a bundle are not normally H-modules.A 2-form ω is of type (1.3 Let E be an AH-bundle over the hypercomplex manifold M equipped with an anti-self-dual AH-connection nabla. an AH-bundle E equipped with an AH-connection is q-holomorphic if and only if is anti-self-dual. Several authors have considered self-dual and anti-self-dual connections. i2 and i3 on the fibres. then H ⊗ B σ is a q-holomorphic AH-bundle. so ω ∈ E2. By analogy with the 4-dimensional theory (see Example 3.0 as self-dual connections. and (E † )∗ ⊗ C is a hyperholomorphic bundle.4. ) is a q-holomorphic AHbundle. ) be a q-holomorphic AH-bundle over the hypercomplex manifold M .

since Ij = 0 for j = 1. Let A ⊂ H ⊗ T ∗ M be the q-holomorphic cotangent space of M.12) where I1 .4. E) = P(E) be the space of q-holomorphic sections of E. which are precisely the q-holomorphic sections of the trivial bundle M × H equipped with the flat connection. Ij ) is a holomorphic vector bundle for j = 1. all with their induced AHH connections.4. 3. 2. The integrability of this structure is guaranteed by the fact that (E.4. 2. (6. A is a q-holomorphic AH-bundle. E). Example 6. It follows that a ∈ C ∞ (A⊗T ∗ M ).3 to show that the holomorphic sections of this holomorphic vector bundle give rise to qholomorphic sections of E. Then so are the various associated bundles E⊗H F . Let P(M. Instead.the hypercomplex structure (I1 . and admits holomorphic sections. I2 . ) is q-holomorphic then E is holomorphic with respect to the complex structure Q. Λk E and so on. 95 . Definition 6. Then a0 +I1 ( a1 )+I2 ( a2 )+I3 ( a3 ) = 0 (where I1 . This defines an almost-hypercomplex structure on E. Since the Obata connection is anti-self-dual. Then e is a q-holomorphic section of E if and only if e0 + I1 ( e1 ) + I2 ( e2 ) + I3 ( e3 ) = 0. I2 and I3 act on the A factor of A⊗T ∗ M ). So PM = P(M × H). We cannot automatically rewrite Equation (6. we use the inclusion map ιE to manipulate sections of e in a more manageable form. ej ∈ C ∞ (M.4. However. because the fibres of E will not in general have a well-defined right H-action. It is easy to adapt Lemma 6. Q-holomorphic sections are the natural generalisation of q-holomorphic functions. Let e ∈ C ∞ (M.7 Let E be an AH-bundle over the hypercomplex manifold M equipped with an AH-connection . (E † ))∗ . Hence the Obata connection defines an AH-connection on A. I2 and I3 act on the T ∗ M factor of E ⊗ T ∗ M .1. which is thus a q-holomorphic AH-bundle.5 Let M be a hypercomplex manifold. A).6 Let E and F be q-holomorphic AH-bundles. Then the Obata connection on T ∗ M is anti-self-dual and defines an anti-self-dual AH-connection on H ⊗ T ∗ M . ωj ∈ T ∗ M }. E ⊕ F . I3 ) to be the horizontal lift of the hypercomplex structure on M . 6. 3.7) to define a CauchyRiemann-Fueter operator on a general AH-bundle E. A general AH-bundle might have no q-holomorphic sections.2 Q-holomorphic sections Q-holomorphic sections of AH-bundles are defined using a version of the Cauchy-RiemannFueter equations. Let P(E) = P(E) ∩ C ∞ (M. Example 6. if the AHbundle (E. A = {ω0 + i1 ⊗ ω1 + i2 ⊗ ω2 + i3 ⊗ ω3 : ω0 + I1 ω1 + I2 ω2 + I3 ω3 = 0. E) so that ιE (e) = e0 + i1 e1 + i2 e2 + i3 e3 . E ) so that P(E) is an AH-submodule of C ∞ (M. Let a = a0 +i1 a1 +i2 a2 +i3 a3 ∈ C ∞ (M.

Here is the quaternionic analogue of this statement: 96 . In just the same way. Let V be a holomorphic vector bundle with a E connection compatible with the holomorphic structure. Q-holomorphic sections and the quaternionic cotangent space Let E be a q-holomorphic AH-bundle over the hypercomplex manifold M . ◦ ιE Hence we regard E as a subspace of H ⊗ (E † )∗ . For example. and so is technically a q-holomorphic AH-bundle. we have a copy of the quaternionic cotangent space H ⊗ T ∗ M .0 ◦ + π 0. just as we did for q-holomorphic functions in the previous chapter. so b is q-holomorphic if and only if b0 − I1 I1 b0 − I2 I2 b0 − I3 I3 b0 = 0. ∞ This analogy goes further.7 is precisely the real ¯ ¯ part of δ E . which means that v ∈ C ∞ (V ⊗C Λ1. and then the action of on E ¯ splits into a q-holomorphic part δ E and a q-antiholomorphic part δ E so that in effect ¯ = δ E + δ E . so that ∂ = ∂. E) → C ∞ (M. and such behaviour is predicted and described by the algebra of the quaternionic cotangent space. The inclusion map ιE and the AH-connection (regarded as a connection on (E † )∗ ) define an AH-morphism ◦ ιE : C ∞ (M. In complex geometry. 2 Since Ij = 0 and Ij = −1.1 ◦ (s) = 0. (E † )∗ ⊗ (A ⊕ B)).1 ◦ . and the AH-bundle B admits no non-trivial q-holomorphic sections. Let b = b0 − i1 I1 b0 − i2 I2 b0 − i3 I3 b0 ∈ C ∞ (M. Thus we have a map ◦ ιE : C ∞ (M.0 M ).Not all q-holomorphic AH-bundles have interesting q-holomorphic sections. the operator e → e0 + I1 ( e1 ) + I2 ( e2 ) + I3 ( e3 ) of Definition 6. the q-antiholomorphic cotangent space B ⊂ H ⊗ T ∗ M is closed under the action of the Obata connection. this follows from the fact that B = 0. A section v ∈ C ∞ (V ) is holomorphic if and only if π 0. Recall that B = {ω − i1 ⊗ I1 (ω) − i2 ⊗ I2 (ω) − i3 ⊗ I3 (ω) : ω ∈ T ∗ M }. This is an exact analogue of the complex case where a connection splits as = π 1. and a section e ∈ C ∞ (E) is q-holomorphic if and only if δ E (e) = 0. which we can split into the q-holomorphic and q-antiholomorphic spaces A and B. E) → C ∞ (M. Definition 6. a section v ∈ C (V ) is holomorphic if and only if ∂v = 0. After swapping the H and (E † )∗ factors. B). We can describe the q-holomorphic sections of E using the q-holomorphic cotangent space. In fact. We can now use the projections A π and π B of the previous chapter to split the action of ◦ ιE into two operators.4. ) be a q-holomorphic AH-bundle.4.8 Let (E. H ⊗ (E † )∗ ⊗ T ∗ M ). We define the pair of ¯ operators δ E : C ∞ (E) → C ∞ ((E † )∗ ⊗ A) and δ E : C ∞ (E) → C ∞ ((E † )∗ ⊗ B) by δ E = (id(E † )∗ ⊗π A ) ◦ and ¯ δ E = (id(E † )∗ ⊗π B ) ◦ ◦ ιE . this equation is satisfied if and only if b = 0.

Define φE. We want to define an AHmorphism φ : C ∞ (M. F ) → C ∞ (M × N. Using the identification (A† )∗ ∼ T ∗ M we can regard δ E (e) as a section of = H ⊗ (E † )∗ ⊗ (A† )∗ .E). a q-holomorphic section e of a q-holomorphic vector bundle E is one whose covariant derivative e takes values in the quaternionic tensor product of E with the q-holomorphic cotangent space. F ). we need to consider the combination of a point m at which † to evaluate sections. the evaluation map θm generates the whole of H† ∼ R. by Definition 4. M. define the ‘evaluation at n ∈ N ’.Proposition 6. E)† ⊗ C ∞ (N. Then αm ∈ C ∞ (M. In hypercomplex geometry.1. E⊗H F ). and let e ∈ C ∞ (E) ◦ ιE (e) ∈ C ∞ (E⊗H A). so δ E (e) = 0 and ◦ ιE (e) = δ E (e) ∈ C ∞ (H ⊗ (E † )∗ ⊗ ∗ T M ). In complex geometry. since δ E is defined by projection to this subspace. E)⊗H C ∞ (N. β ∈ Fn . Let E and F be AH-bundles.4.6). † Consider first the fibres Em and Fn for m ∈ M and n ∈ N . E)† . there is a natural inclusion C ∞ (M. F ))† . E ∞ † ∗ Clearly δ (e) ∈ C ((E ) ⊗ ιA (A)). so δ E (e) ∈ C ∞ (ιE (E)⊗(A† )∗ ). With H-valued functions (sections of M × H ). Similarly. The operators αm and βm are generalisations of Joyce’s θm ∈ C ∞ (M. n) ∈ Em ⊗H Fn λEm . F )† .9 Let (E.F ( )(m.3 Sections of Tensor Products For real and complex vector bundles. recall the linear map λU V : U † ⊗ V † → (U ⊗H V )† of † † Equation (6. Definition 6.Fn : Em ⊗ Fn → (Em ⊗H Fn )† and λC ∞ (M. M × N ). E) ⊗ C ∞ (N. Thus ◦ ιE (e) ∈ C ∞ (E⊗H A). The obvious difficulty is that for sections e and f of E and F respectively. E). 97 .Fn (α ⊗ β) · φE. n) = λC ∞ (M. The quaternionic analogue of this map is more delicate. E) → H by αm (e) = α(e(m)) for all ∞ e ∈ C (M. E)⊗H C ∞ (N.E). 6.2). Instead we use a generalisation of Joyce’s map φ : PM ⊗H PN → PM ×N (Definition 6. π.1. βn (f ) = β(f (n)). For m = more general AH-bundles.10 Let by the equation ∈ C ∞ (M. E ⊗F ) given by (e⊗f )(m. Then βn ∈ C ∞ (N. The induced connection on (E † )∗ preserves ιE (E).4.1. Define an AH-morphism αm : C ∞ (M. For AH-modules U and V . Let e be q-holomorphic. there will not in general be a section e⊗H f of the AH-bundle (E⊗H F. introduced in Lemma 6. ¯ Proof. a holomorphic section v of a holomorphic vector bundle V is one whose covariant derivative v takes values in the complex tensor product of V with the holomorphic cotangent space. Let α ∈ Em and † β ∈ Fn .4. H)† . F )† → (C ∞ (M.4.F ) (αm ⊗ βn ) · † † for all m ∈ M . since we can no longer assume that † the map α = id generates the whole of Em . F ) → C ∞ (M ×N.F ( )(m.1.C ∞ (N. α ∈ Em and for all n ∈ N . n) = e(m)⊗f (n). E)⊗H C ∞ (N. and an AH-morphism α ∈ Em . be a q-holomorphic section. where A is the q-holomorphic cotangent space of M .C ∞ (N. Then ) be a q-holomorphic AH-bundle.F ) : C ∞ (M. Thus there are linear maps λEm .

F .M differentiates sections of ιE (E) in ¯ ¯ ¯ we see that δ the M directions and then projects to the q-antiholomorphic cotangent space of M ×N .Thus φE. so the section product map takes q-holomorphic sections to q-holomorphic sections. Let ∈ P(E)⊗H P(F ).10.N .1.1.1. In particular. It is easy to see that φE. π2 .F the section product map for AH-bundles. Proof. M × N ). F ) → P(M.6) that φ : PM ⊗H PN → PM ×N . E) and f ∈ C ∞ (N.F ( ) defines a section of the AH-bundle (E⊗H F.F ( )) = 0. Similarly.F : C ∞ (M. using the section product map φE.F is the section product map of Definition 6.M ⊗ id + id ⊗ F. π1 . F ) be q-holomorphic AH-bundles. Using the natural splitting T ∗ (M × N ) ∼ T ∗ M ⊕ T ∗ N . Consider the anti-self-dual connection E⊗H F = id ⊗( E. we have a natural AH-morphism ρ ◦ φH.4. = The section product map is the natural tool for relating tensor products of sections with sections of tensor products and allows us to treat sections of more complicated AHbundles using similar techniques to those used by Joyce for q-holomorphic functions. E⊗H F ). (id ⊗δ F. M. F ) satisfy the conditions of Lemma 4. The same is true for sums of q-holomorphic sections.11. let E be a q-holomorphic AH-bundle over M . N. and so we have a linear map φE.F ( ) is q-holomorphic. M ). so φ(e⊗H f )(m. and there is an analogous description for q-holomorphic sections of quaternionic tensor products. ¯ and similarly for δ F. F ) ⊂ C ∞ (N. It is well known that sums and tensor products of holomorphic sections are themselves holomorphic. E ) and (F. We shall call φE.E : PM ⊗H P(E) −→ P(H⊗H E) ∼ P(E).F (e⊗H f ) is the complex product of the sections.N )(φE. where φE. ¯ Then E⊗H F = δ E⊗H F + δ E⊗H F . This is a generalisation of the fact (Definition 6.F : P(M. If e ∈ C ∞ (M. let P(E) be the AH-module of q-holomorphic sections of E and let PM = P(H) be the H-algebra of q-holomorphic functions on M .6. and we obtain a natural product ρ ◦ φE. E⊗H F ). Hence δ E⊗H F (φE. E)⊗H P(N. where E⊗H F is equipped with the connection E⊗H F . F ) be the AH-modules of their q-holomorphic sections.M ⊗id)(φE. Theorem 6.M ⊗id + id ⊗δ F. and φE. E) and P(N. n) = e(m)⊗H f (n) ∼ e(m) ⊗Cq f (n).4.F ( )) = 0. Here is the definition of an H-algebra module: 98 . and let P(M.4. = (6. By the same arguments as in Definition 6.N ). E⊗H F ).11 Let (E. E)⊗H C ∞ (M. Just as for q-holomorphic functions. when M = N we can restrict to the diagonal bundle (E⊗H F. F ) −→ P(M × N. F ) → C ∞ (M × N.F ( ) ∈ C ∞ (M ×N.13) We can descibe such an algebraic situation by saying that P(E) is an H-algebra module over PM .F : P(M.F ( ) is smooth because it is a finite sum of smooth sections. where δ E. Then φE. and let φE.F is an injective AH-morphism. it follows that ¯ ¯ ¯ (δ E. Since ∈ ιP(E) (P(E)) ⊗ (P(F )† )∗ .7 (so if ιE (e) and ιF (f ) both take values in Cq for some q ∈ S 2 ) then the section φE.F ( )) = 0. E) ⊂ C ∞ (M. The restriction map ρ clearly preserves q-holomorphic sections.N . E⊗H F ). = ¯E⊗H F = δ E. φE. Then by Theorem 6. E)⊗H P(M.

(i) Q is an AH-module and (P. µP ) is an H-algebra.4.4). (ii) There is an AH-morphism µQ : P⊗H Q → Q. (iii) The maps µP and µQ combine to give AH-morphisms µP ⊗H id and id ⊗H µQ : P⊗H P⊗H Q → P⊗H Q. Different connections will give rise to different H-algebra modules of q-holomorphic sections. Then µQ (1⊗H u) = u for all u ∈ Q.7.1. This is associativity of module multiplication. (iv) For u ∈ Q. This suggests that the theory of H-algebra modules over PH might lead to an algebraic description of instantons on H. called the module multiplication map. The proof of this statement is obtained by adapting Joyce’s proof that the q-holomorphic functions PM form an H-algebra [J1. 1⊗H u ∈ P⊗H Q by Lemma 4. the module multiplication map being the section product map φH.12 Q is an H-algebra module if Q satisfies Axiom M.5]. Definition 6. The idea of an H-algebra module was suggested by Joyce.13). Then µQ ◦(µP ⊗H id) = µQ ◦(id ⊗H µQ ). Composing with µQ gives AHmorphisms µQ ◦ (µP ⊗H id) and µQ ◦ (id ⊗H µQ ) : P⊗H P⊗H Q → Q.1. It is relatively easy to see that the q-holomorphic sections P(E) of a q-holomorphic AH-bundle form an H-algebra module over the H-algebra PM . and the axioms follow a very similar pattern to those for an H-algebra (Definition 6.Axiom M. Theorem 5. One possible application of this idea is to study anti-self-dual connections (instantons) on H.E of Equation (6. Thus 1 acts as an identity on Q. 99 .

The Lie bracket on hypercomplex manifolds defines a natural operation on vector fields which satisfies these axioms. These vector fields are closed under the Lie bracket (suitably adapted to the quaternionic situation). there is a splitting of the quaternionic tangent space H ⊗ T M using which we define a hypercomplex version of the ‘(1. In recent times. Not only the top row but the top two rows of the quaternion-valued double complex are elliptic. We use this idea to calculate the quaternionic cohomology groups of the group U(2). This decomposition gives rise to a double complex of quaternion-valued forms on hypercomplex manifolds. which allows us to adapt the real-valued double complex of Chapter 3 to decompose quaternion-valued forms. Spindel et al. This allows us to define q-holomorphic k-forms. The top row of the quaternion-valued double complex is particularly well-behaved.r into two AH-bundles. On a hypercomplex manifold we have global complex structures. This discovery presents us with lots of examples of finite-dimensional quaternionic Lie algebras. A similar approach to quaternion-valued vector fields is also fruitful. and in four dimensions the whole complex is elliptic. [SSTP] and Joyce [J3] demonstrated the existence of invariant hypercomplex structures on certain compact Lie groups and their homogeneous spaces.Chapter 7 Quaternion Valued Forms and Vector Fields This final chapter uses the algebra and geometry developed so far to describe quaternionvalued tensors on hypercomplex manifolds. which has some advantages over the real-valued double complex. The splitting H ⊗ T ∗ M ∼ A ⊕ B of the = previous chapter is the first example of this type of decomposition. 0) vector fields’ on a complex manifold. More generally. and suggest how these methods may be extended to higher-dimensional hypercomplex Lie groups. and to describe their algebraic structure using ideas from the previous chapter. in a similar way to that in which Joyce arrived at Halgebras. Just as with the quaternionic cotangent space. for all k and r we obtain a splitting of H ⊗ Ek. and can be constructed using quaternionic algebra. This encourages us to adapt the axioms for a Lie algebra to form a quaternionic version. 100 .

r k.r splits according to the Clebsch-Gordon formula.r Suppose in addition that M is hypercomplex. Q is the trivial bundle M × Sp(1) ). The AH-bundle H ⊗ Λk T ∗ M decomposes as k+1 H ⊗ Λ T M ∼ V1L ⊗ = k ∗ r=0 n RG (εn k. Recall that the subbundle of Λk T ∗ M consisting of Vr -type representations is denoted Ek. By Proposition 3. k. Instead of just real or complex forms (which we can think of as taking values in the trivial representation V0 ).r )V1 ⊗ Vr+1 ⊆ H ⊗ Λ T M .r (7. this takes the form RG RG H ⊗ Ek. n k ∗ Definition 7.1. i. (This is a generalisation of the splitting H ⊗ T ∗ M ∼ A ⊕ B. and so possesses global complex structures I1 .2 Define Fk.e. consider forms taking values in some Sp(1)-representation W .1 is a decomposition of H ⊗ Ek. Since the Sp(1)-action on Λk T ∗ M is now defined by global complex structures.r ∼ V1L ⊗ V1R ⊗ εn VrG ∼ εn V1L ⊗ (Vr+1 ⊕ Vr−1 ).r )∗ = H ⊗ 101 .7.r−1 )Vr . The situation in which we are particularly interested is that of forms taking values in the quaternions H = V1L ⊗ V1R .r is an AH-subbundle of H ⊗ (Fk.r where the ‘Sp (1)G -action’ is the action of Sp(1) on Λk T ∗ M induced by the hypercomplex structure.r ∩ (I ⊗ Λk T ∗ M ). = = k. In symbols. the space Fk. k.10.r+1 + εk.r is obtained using the theory of Chapter 5.r to be the subspace (εn k.1.r (7. we can take the diagonal action under which the subspace W ⊗ Ek. Taking the diagonal Sp(1)RG -action using the Clebsch-Gordon formula.r = εn Vr .2.r+2 + εk. As in Equation 5.) Defining = † Fk. I2 and I3 = I1 I2 (in other words. Proof. Equation 7.1 The Quaternion-valued Double Complex In Chapter 3. sections of the bundle W ⊗ Λk T ∗ M .1) Proposition 7.4.2) Collecting together the Vr representations yields the formula in the Proposition.r into stable and antistable AH-modules. this becomes k H ⊗ Λ T M ∼ V1L ⊗ = k ∗ r=0 RG RG εn (Vr+1 ⊕ Vr−1 ) .1. where r ≡ k + 1 mod 2.1 Let M 4n be a hypercomplex manifold. we obtain splittings by coupling the right Sp(1)-action on H with the Sp(1)-action on Λk T ∗ M . k. we saw how the real-valued exterior forms Λk T ∗ M on a quaternionic manifold M are acted upon by the principal Sp(1)-bundle Q of local almost complex structures on M . As suggested in Section 3. we have k H ⊗ Λk T ∗ M ∼ V1L ⊗ V1R ⊗ = r=0 εn VrG .r = Fk. The primed part of the space Fk.

r+1 ◦ d and ¯ δ : C ∞ (Fk.r−2 → H ⊗ Ek.r r 2 k. A short calculation shows that ↑ dim Fk.r ⊕ Fk. However. ¯ The definition of the operators δ and δ of Section 6.r ∼ =  1 n × ε U ⊕ 1 εn Ur−2 2 k. Just as with the splitting H ⊗ T ∗ M ∼ A ⊕ B. As with the decomposition H ⊗ T ∗ M ∼ A ⊕ B. The splitting RG RG H ⊗ Ek. ↑ The q-holomorphic cotangent space A is F1.r+2 .r−2 )∗ = Ek.r+2 . ↑ ↓ Since we can consider the bundles Fk.r . these are not AH-isomorphisms because = some of the primed part is lost in the splitting.r+2 ↓ n dim Fk. H) → Ωk+1 (M.r is the Ur -type subspace of H ⊗ Λk T ∗ M . it may appear strange to mix up the stable and antistable fibres in the single bundle Fk.r be the natural projection map πk.r to be the AH-bundle a εn Ur ⊆ H ⊗ Ek.−1 = F1.r = 2(r + 2)εn k.1 and the q-antiholomorphic cotan↓ gent space B is F1. there is an injective AH-morphism = ↑ ↓ Fk.r+2 ⊕ Ek.r and Fk.1.1 = F1.−1 .(Ek.r H ⊗ Ek.r .r+1 ) δ = πk+1. Thus Fk.r Ur ⊕ εn Ur−2 k.r ⊕ Fk. We give names to these spaces as follows (where as usual a = 1 if n is even and a = 2 if n is odd): ↓ ↑ 1 Definition 7. giving rise to a double complex of quaternion-valued forms.r is the direct sum of stable and antistable components.r gives an H-module isomorphism  n ×  εk.1. so in order to obtain a double complex it is necessary to amalgamate the stable and antistable contributions.r k. ¯ Define the differential operators δ.4 Let πk.r ∼ εn V1L ⊗ (Vr+1 ⊕ Vr−1 ) = k.r . so that ↓ ↑ Fk.r )∗ = (Fk.2 can now be generalised to • ∗ cover the whole of H ⊗ Λ T M . Definition 7.r is the Ur -type subspace of k.r : H ⊗ Λk T ∗ M → Fk.r ↑ n dim Fk.r ) → C ∞ (Fk+1.r = (r + 1)εk. and H ⊗ Ek. it soon becomes clear that exterior differentiation does not necessarily map stable fibres to stable fibres or antistable fibres to antistable fibres.r separately.r−2 .r which is an H-linear isomorphism of the total spaces but is not injective on the primed parts. H) by δ : C ∞ (Fk. δ : Ωk (M.r ↑ 1 × to be the AH-bundle a εn Ur ⊆ H ⊗ Ek. and each (fibre of the) AH-bundle Fk.r ).r .3) for r odd.r = (r + 3)εk. Define Fk. 102 .r = Fk.r−1 ) ¯ δ = πk+1.r ∼ Fk.r and ↓ dim Fk.r ⊕ = ↓ Fk.r .r+2 ↓ × H ⊗ Λk T ∗ M and Fk.3 Define Fk.r ) → C ∞ (Fk+1.r−1 ◦ d .r for r even (7.r = 2(r + 2)εn k. we have (Fk. † ↑ † ↓ ↑ With these definitions.

This can be observed in the simplest of cases — a quaternion-valued ↑ ¯ function f is a section of F0.r ⊗ T ∗ M = (εn k. δ ¨¨ B ¨ ¯ As already hinted. .−1 ) . and unless f is q-holomorphic δf is a nonzero ↓ section of F1.1. It follows that ¯ ¯ ¯ δ 2 = δ δ + δδ = δ 2 = 0.r ) → C ∞ (M.2 ) ¨¨ δ ¨¨ B ¨ C ∞ (M.Theorem 7. F1.r gives rise to a double complex. the operators δ and δ do not preserve stable or antistable sub¯ spaces.r → Fk+1.r+2 + εk. The Casimir operator in question is that of the diagonal Lie algebra action given by the operators I(ω) = I1 (ω) − ωi1 J (ω) = I2 (ω) − ωi2 103 K(ω) = I3 (ω) − ωi3 . F3. F2. F3. and the decomposition H ⊗ Λk T ∗ M = k+1 r=0 Fk.r−1 . using the Casimir element technique of Lemma 3. Since n Fk. . etc. . .7.r−1 .1 ) B ¨ δ ¨ ¨¨ ¨ ¨ r rr r rr ¯r δ rr j j r ¨ ¨¨ δ ¨¨ B ¨ C ∞ (M.0 = F0. . .0 ) B ¨ ¨ ¯r δ rr ¨¨ j C ∞ (M. F3. Let the Obata connection on M . though we have δ : Fk. . Fk.r ) to C ∞ (M.1: The Quaternion-Valued Double Complex C ∞ (M. it is not the case that ↑ ↑ ¯ δ : Fk. it follows (from the Clebsch-Gordon splitting Vr+1 ⊗ 2nV1 ∼ 2n(Vr+2 ⊕ Vr ).r → Fk+1.0 = H) r rr C ∞ (M. For example. so : C ∞ (M. For example.−1 . The rest of the theorem follows automatically. Fk.2. be Proof. Much of the theory from the real-valued double complex of Chapter 3 can be adapted ¯ to describe the quaternion-valued version as well.3.r ⊗ T ∗ M ). F0. etc. followed by = ∞ the antisymmetrisation d = ∧ ◦ ) that d : C (M. The proof works in exactly the same way as that of Theorem 3. Fk+1.r−1 ).3 ) ¨ ¨¨ rr . so ¯ d = δ + δ. Fk+1.r ) → C ∞ (M. Fk. the operators δ and δ can be expressed in a similar fashion to D and D.r+1 ⊕ Fk+1.0 . F1.2. . r rr j r δ ¨¨ B ¨ rr C ∞ (M. .r−1 ).r+1 ⊕ Fk+1. Figure 7. F2.r )V1 ⊗ Vr+1 ⊗ 2nV1 .5 The exterior derivative d maps C ∞ (M. . .−1 = B) δ ¨¨ ¨ ¯r δ rr ¨¨ j C ∞ (M.1 = A) B ¨ δ ¨ ¨¨ ¨ ¨ r rr ¯r δ rr j ¨ rr ¯ δ r C ∞ (M. Fk.

We can infer that the operator δ is elliptic except at the bottom spaces F2k−1.r . As in Section 3. It follows by taking exterior product with e0 that ωe10 = 0. etc. Ellipticity at C ∞ (Fk.0 . Then σ(ω) = 0 if and only if ωe0 ∈ F3.6 Let α ∈ C ∞ (Fk. in particular the leading edge of spaces Fk.7 The complex 0 −→ C ∞ (F1.k forms a complex which is elliptic throughout.0 . .0 is injective. the operator δ is elliptic at some of these lowest-weight spaces for low exterior powers. which is the case if and only if I(ωe0 ) = J (ωe0 ) = K(ωe0 ) = 0. it is enough to choose some e0 ∈ T ∗ M and show that the symbol sequence 0 −→ F1.r+1 (ωe0 ) for ω ∈ Fk. −→ C ∞ (F2n+1. Then δα = − and 1 4 r+ 1 (I 2 + J 2 + K2 ) dα r+2 1 (I 2 + J 2 + K2 ) dα. This leads to the following adaptation of Lemma 3.3. Proof.) Since δ = d on F1. we have σ(β) = βe0 for all β ∈ B. The first of these equations is I1 (ωe0 ) − ωe0 i1 = I1 (ω)e0 + ωe1 − ωe0 i1 = 0.k−2 ) for all k ≥ 3 follows from a generalisation of the techniques used in the proof of Theorem 3. is exact. e0 ) = πk+1.−1 ) −→ C ∞ (F2.−1 .0 −→ F3.1. It follows from the expression for B in Equation (6.r ).−1 = B.−1 and F2k. 104 σ σ σ δ δ δ δ . Closer examination also reveals that the ‘second row’ of spaces Fk. so the complex is elliptic at B.7: Lemma 7.3 can also be adapted to the new situation. Again.3.−1 −→ F2.0 ).5). .−1 ) = C ∞ (B) and C ∞ (F2.of Equation (6.k−2 is also an elliptic complex with respect to δ.2n−1 ) −→ 0 is elliptic. where σ(ω) = σδ (ω. (As usual ωe0 means ω ∧ e0 .0 ) −→ . .1. . Let ω ∈ F2.1 −→ . Lemma 7. We need to show that the complex is elliptic at C ∞ (F1. r+2 1 ¯ δα = 4 (r + 4) + The results on ellipticity in Section 3.9) that σ : B → F2. and so ωe0 must be equal to zero and ω = γe0 for some γ ∈ H ⊗ T ∗ M . The same arguments for J and K show that σ(ω) = 0 =⇒ ωe10 = ωe20 = ωe30 = 0.2.1.

−1 ∼ U−1 = × rr j rr r δ ¨¨ B ¨ In four dimensions the situation is particularly friendly towards quaternion-valued forms. H ⊗ Λ2 T ∗ M ∼ F2.2 ⊕ F2.1. where = F1. where = = F3. To show this.1 ⊕ F3.−1 .2: The Quaternion-Valued Double Complex in Four Dimensions Λ2 A ∼ Y = H ¨ r δ¨ ¨¨ rr ¯r B ¨¨ rr rr ¯ δ r A∼Z = ¨¨ ¨ j r F3. F3.0 . H ↑ ↓ F2.1 = A = {q0 e0 + q1 e1 + q2 e2 + q3 e3 : q0 + q1 i1 + q2 i2 + q3 i3 = 0}.0 . where β = 4 q(−e + e1 i1 + e2 i2 + e3 i3 ) ∈ B. . This completes the proof. Hence we can find β ∈ B such that σ(ω) = 0 implies that ω = σ(β) for all ω ∈ F2. with α ∈ A.It remains to show that we can choose β ∈ B such that ω = βe0 . the double complex of quaternion-valued forms has not just one but two rows which with respect to the operator δ are elliptic throughout. because the whole double complex is elliptic. It follows that the symbol map σδ is an isomorphism between F3. This is an unexpected bonus — on hypercomplex manifolds. so the complex is exact at F2.0 . Since the complex is elliptic at A it follows that αe0 ∈ F2. e03−12 H ⊕ {q1 e01+23 + q2 e02+31 + q3 e03+12 : q1 i1 = q2 i2 = q3 i3 }. F1. and so in four dimensions the entire quaternion-valued double complex is elliptic. e02−31 .0 ∼ 3H ⊕ H = ¯r δ rr j ¨ B ∼ U−1 = × B ¨ δ ¨¨ ¨¨ rr j ¨¨ ¨ δ ¨¨ B ¨ δ × F4.0 ⊕ F2.0 = F2.−1 = B = {q0 e0 + q1 e1 + q2 e2 + q3 e3 : q0 = q1 i1 = q2 i2 = q3 i3 }.2 = Λ2 A = {q1 e01+23 + q2 e02+31 + q3 e03+12 : q1 i1 + q2 i2 + q3 i3 = 0}.−1 .1 ∼ Z = r rr ¯r δ ¨¨ B ¨ δ F0. Suppose instead that ω = αe0 . 105 . But then 1 0 0 0 αe = βe . . Explicitly.0 = H r rr F2.k and the second row Fk. Lastly.8 Quaternion-valued forms in four dimensions Figure 7. Example 7.0 ∼ H = ¨ ¯r δ rr ¨¨ j F3.1 ⊕ F1. . Next.−1 = {q0 e123 + q1 e032 + q2 e013 + q3 e021 : q0 = q1 i1 = q2 i2 = q3 i3 }. H ⊗ Λ3 T ∗ M ∼ F3. . e3 } for T ∗ M so that as usual Ij (e0 ) = ej .k−2 .0 e01−23 .−1 and F4.1 = {q0 e123 + q1 e032 + q2 e013 + q3 e021 : q0 + q1 i1 + q2 i2 + q3 i3 = 0}.0 1 if and only if α = σ(q) = 4 q(3e0 + e1 i1 + e2 i2 + e3 i3 ) for some q ∈ H. choose a standard basis of 1-forms {e0 . we have H ⊗ T ∗ M ∼ F1. where = F2. namely the top row Fk.0 .

1. H Now. and the map ιk : Λk (A ⊕ B) → H ⊗ Λk T ∗ M is none H H H other than the normal inclusion map ι : Λk A → H ⊗ (Λk A† )∗ . (Recall = = H H that Uk = aV1 ⊗ Vn+1 where a = 1 for k even and a = 2 for k odd. A ∼ nU1 . so by Theorem 5. In complex geometry the Hodge decomposition of forms results immediately from the isomorphism C⊗T ∗ M ∼ Λ1.1 ) −→ . Proposition 7. −→ C ∞ (F2n.16.9 Let A be the q-holomorphic cotangent space of a hypercomplex manifold M . It follows that there is H H an AH-submodule of H ⊗ Λk T ∗ M which is isomorphic to Λk A. With hypercomplex geometry we are not quite so lucky.2n ) −→ 0. It follows that H k 1 2n Uk = Λk A ∼ = H a k 2n V1 ⊗ Vk+1 . dim Λk (nU1 ) = 2(k + 2) 2n . . the inclusion map ιk is still of special interest. k (7.4) Thus ιk (Λk A) must be a subspace of H ⊗ Λk T ∗ M of this form. the operator δ on functions does extend to give the elliptic complex 0 −→ C ∞ (F0. 1 On the other hand. Proof. B⊗H B = {0} and so Λk B = {0} for k > 1 .k ⊆ H ⊗ Λk T ∗ M .1. 1 106 . Baston [Bas] succeeded in extending the Cauchy-Riemann-Fueter operator to a locally exact complex with a different construction involving second order operators. δ δ δ δ and so Λp A⊗H Λq B → H ⊗ Λk T ∗ M. Thus H Λk (A ⊕ B) = Λk A for k > 1.) By Proposition 4. This section discusses these top row spaces Fk. because it describes explicitly the top row of the double complex. Despite the fact that we can identify H ⊗ Λk T ∗ M with Λk (H ⊗ T ∗ M ).k which are of particular interest.7. In spite of this. since the splitting A ⊕ B ∼ H ⊗ T ∗ M is not an AH-isomorphism but rather an injective AH= morphism which is not surjective on the primed parts.1. also A⊗H B = {0}. The inclusion map ιk : Λk (A⊕B) → H⊗T ∗ M identifies Λk A with the highest H H space Fk. the Cauchy-Riemann-Fueter operator δ in the quaternion-valued double complex (Figure 7.1) does not begin a longer elliptic complex.2. H H Since B = {0}. . H In 1991. Λk A ⊆ k A ∼ mUk for some m.0 ) −→ C ∞ (F1. There is a natural identification Λk (A ⊕ B) = H ιk : p+q=k p q p+q=k (ΛH A⊗H ΛH B).1. the induced AH-morphism H ιk : Λk (A ⊕ B) → H ⊗ Λk T ∗ M H is therefore not an isomorphism for k > 1. This is why we resort to our more complicated analysis of the Sp(1)-representation on H ⊗ Λk T ∗ M to discover the quaternionic version of the Dolbeault complex.0 ⊕Λ0.1 The Top Row Λk A and Q-holomorphic k-forms H In contrast with the the role of the Cauchy-Riemann operator ∂ in the Dolbeault com¯ plex.1 and the isomorphism in exterior algebra Λk (U ⊕V ) = = p q p+q=k Λ U ⊗ Λ V .

The highest-weight vectors in C ⊗ Λk T ∗ M are the (k. we identify its exterior powers with the corresponding AH-submodules of H ⊗ Λk T ∗ M . Q i. We generalise this result to higher exterior powers as follows: Theorem 7.0 .1. so the H-algebra of 0 q-holomorphic functions on M is PM = PM . so that (A† )∗ = T ∗ M .0 M . 0)-forms Λk. H so there is a natural isomorphism ιk : Λk A ∼ Fk. and both form the ‘top row’ of their double complexes. Thus H · ιq (ΛQ ) ⊂ (id ⊗H χq )(ΛH A⊗H Xq ). The H k AH-module of q-holomorphic k-forms on M will be written PM . H) is q-holomorphic if and only if ω ∈ C ∞ (M. Thus we will omit to write the ‘ ιk ’.k . The result follows.0 ) = (id ⊗H χq )(Λk A⊗H Xq ). This suggests a natural definition of a ‘q-holomorphic k-form’: Definition 7. Consider the weights of the action of Q ∈ S 2 ⊂ sp(1). the space Λk A is generated over H by the spaces of (k. which is essentially the same result for complexified k-forms. we see that Fk.0 k ιq .3.0 . which are mapped to Cq ⊗ Λk T ∗ M by the map Q k. H Q H · ιq (Λk.2. H Q Proof. 2 107 . 0)-forms Λk. 0)-forms Λk.10.k .From Definition 7. H Just as the q-holomorphic cotangent space A is our quaternionic analogue of the ∗ holomorphic cotangent space T1. Λk A) and δω = 0.0 M . H H In Section 6. writing Λk A = Fk.k and (Λk A† )∗ = Ek. it follows that Λk A⊗H Xq ∼ 2n Xq . This is AH= k = k H H k k isomorphic to the submodule (id ⊗H χq )(ΛH A⊗H Xq ) ⊂ ΛH A.0 = Λk T1.10 ifold M 4n . This theorem is really a quaternionic version of (Salamon’s) Equation 2.0 ).7).1. it is generated by these subspaces.k = 2n V1 ⊗ Vk+1 as well.k .1.2. Since there are no k other representations of this weight in H ⊗ Λk T ∗ M . Since Λk A ∼ 2n V1 ⊗ Vk+1 .e. Both are formed in the same way using C exterior algebra over their respective fields. Then It follows that Λk A = H Q∈S 2 2 Let A be the q-holomorphic cotangent space of a hypercomplex man(id ⊗H χq )(Λk A⊗H Xq ) = H · ιq (Λk.2 is generated over H by the weight-spaces of Q with extremal weight. it follows that ιk (Λk A) = Fk. and since dimC Λk.11 Let M be a hypercomplex manifold. the AH-bundle Λk A is the quaternionic analogue H ∗ of the bundle of (k. H Q proving the first part of the Theorem.0 = 2n we see that Q k H · ιq (Λk. = H Just as we define the q-holomorphic cotangent space A as being a particular submodule of H ⊗ T ∗ M .1 we showed that the q-holomorphic cotangent space A is generated over H by the various holomorphic cotangent spaces (Corollary 6. A quaternion-valued k-form ¯ ω ∈ Ωk (M. Since Λk A is stable. which by Theorem 5.0 ).2.

the vanishing of which therefore guarantees that ω ∈ k ∞ C (ΛH A⊗H A). The component of ω taking values ¯ in Fk+1. The q-holomorphic de Rham complex therefore inherits a a rich and interesting algek+1 k braic structure. ) as introduced in Definition H H 6. we have H Λk A ⊗ T ∗ M ∼ = H 2n M V1L ⊗ Vk+1 ⊗ 2nV1G k ∼ 2n 2n V L ⊗ (V M G ⊕ V M G ).e. Define φk. It follows from the theory of q-holomorphic sections that q-holomorphic forms are closed under the tensor product. brought about by the restriction ρ : M × M → M to the diagonal H submanifold Mdiag ⊂ M × M followed by the skewing map ∧ : C ∞ (M. let φΛk A. since ω ∈ C ∞ (Λk A⊗H A) implies that dω = ∧◦ ω ∈ H ¯ C ∞ (Λk+1 A) and so δω = 0.l : PM ⊗H PM −→ PM . .12 A quaternion-valued k-form ω ∈ C ∞ (M.1. ) is a q-holomorphic AH-bundle where is (the connection induced H k by) the Obata connection on M .11 and Proposition 7. which is often called the cohomology algebra of M . H Proof. Λk+l A).7. Λl A) −→ C ∞ (M × M.k−1 ⊂ H ⊗ Λk+1 T ∗ M .k−1 is of course δω. = 1 k+2 k k GH The bundle Λk A⊗H A is precisely the higher-weight subspace 2n 2n V1L ⊗ Vk+2 . ω is a q-holomorphic section H of (Λk A. It may be that the differential graded • H-algebra structure on PM can be used to give a similar description of the quaternionic cohomology of a hypercomplex manifold. Λk A⊗H Λl A) H H H H H H be the section product map (Definition 6. Λk A)⊗H C ∞ (M.4. The q-holomorphic k-forms PM are precisely the qholomorphic sections P(Λk A) of the AH-bundle (Λk A. it is easy to see that (Λk A. H The reverse implication is nontrivial and depends upon analysing the Sp(1)representation on Λk A ⊗ T ∗ M .4). We have already noted that d : PM → PM . 108 . Λk A) satisfies the H ¯ equation δω = 0 if and only if ω ∈ C ∞ (Λk A⊗H A). as the following Proposition demonstrates: Proposition 7. i. Thus the H-algebra structure on the q-holomorphic functions PM extends to one • • on the q-holomorphic k-forms PM . Λk+l A). It is well known that on a real or complex manifold M one can use exterior products over R or C to give an algebraic structure to de Rham or Dolbeault cohomology. ).1 to be none other than the bundle Fk+1. Explicitly.This gives rise to what we may call the q-holomorphic de Rham complex 2n−1 1 2 2n 0 −→ PM −→ PM −→ PM −→ .5) Just as (on a complex manifold) Λk.l to be the restriction to C ∞ (M.1.Λl A : C ∞ (M. It follows (from Theorem 6. −→ PM −→ PM −→ 0. Λk A⊗H Λl A) → H H C ∞ (M.12) that H k+l k l φk.0 is a holomorphic vector bundle.1.10).4.4. The H k 2n complementary subspace 2n k V1L ⊗ VkGH is revealed by Proposition 7. Using Equation (7. . and we say that PM forms a differential graded H-algebra. d=δ d=δ d=δ d=δ d=δ d=δ (7. The ‘if’ part is automatic.

Using similar ideas to those of Section 6. 0) vector fields’ in complex geometry.2. though not quite ideal from the point of view of quaternionic algebra.2. and so defines a natural linear map λ : V ⊗ V → V. which is roughly dual to the splitting H ⊗ T ∗ M ∼ A ⊕ B of = Section 6.0 .2 Vector Fields and Quaternionic Lie Algebras Quaternionic algebra can also be used to describe vector fields on a hypercomplex manifold M .2.0 M ⊕ T 0.1 Let M 4n be a hypercomplex manifold so that T M ∼ 2nV1 as an = Sp(1)-representation.6) The obstruction to this equation is the Nijenhuis tensor NI which measures the (0.6). nor is there any fruitful way to alter the definitions to make them so. Let M 4n be a hypercomplex manifold. and we can use this formulation to obtain a quaternionic analogue of Equation (7. 109 .7. Quaternion-valued vector fields which take = values in the subspace A ⊂ H ⊗ T M turn out to be a good hypercomplex analogue of the ‘ (1. The quaternionic tangent space H ⊗ T M splits according to the equation H ⊗ T M ∼ V1L ⊗ V1R ⊗ 2nV1G ∼ 2nV1L ⊗ (V2RG ⊗ V0RG ). which is expressed by the inclusion [V 1.1 M the antiholomorphic tangent space of M . whilst there are H-linear bundle isomorphisms A ∼ A× and B ∼ B × . This definition is perfectly natural. we define a splitting of the quaternionic tangent space H ⊗ T M ∼ A ⊕ B. An almost complex structure I on a manifold M defines a splitting C ⊗ T M ∼ T 1. we take our inspiration from complex geometry. We define a splitting of the quaternionic tangent space H ⊗ T M . we will find it much more meaningful to talk about linear maps on tensor products. V 1. 0) vector fields. This encourages us to treat these vector fields as a quaternionic Lie algebra.0 . Let V be the set of smooth vector fields on M . an almost hypercomplex structure is integrable if and only if these vector fields are closed under a quaternionic version of the Lie bracket operator. T 1. using the natural definitions A = A ∩ (I ⊗ T M ) and B = B ∩ (I ⊗ T M ) = {0}. Definition 7. 7. a new concept which we proceed to define and explore. Let V 1. We would like A and B to be dual to the cotangent spaces A and B.2. where T 1.1 Vector Fields on Hypercomplex Manifolds As so often.0 = C ∞ (M. As with H-algebras. The Lie bracket of vector fields is a bilinear map [ . The Lie bracket is a bilinear map from V × V to V. However. 0) vector fields.1 M . 0) on M . Then A is the q-holomorphic tangent bundle and B is the q-antiholomorphic tangent bundle of M . = = these spaces are not isomorphic as AH-bundles. = = Define the AH-subbundles A = 2nV1L ⊗ V2RG and B = 2nV1L ⊗ V0RG .0 M is = the holomorphic and T 0. The almost complex structure I is integrable if and only if the Lie bracket preserves (1. In particular. We present a similar theorem for quaternionic vector fields on hypercomplex manifolds.0 ] ⊆ V 1. 1) component of the Lie bracket of two (1. (7. ] : V × V → V.0 M ) be the vector fields of type (1.

wj ∈ V be vector fields such that j qj ⊗vj ⊗wj ∈ C ∞ (A⊗H A). ∈ VA . Proof. Since is torsion-free. Using the canonical isomorphism (H ⊗ T M )⊗H (H ⊗ T M ) ∼ = H ⊗ T M ⊗ T M . To begin with. it follows that qj ⊗ and similarly qj ⊗ wj v j vj wj ∈ VA .A (Definition 6. Thus ( qj ⊗ wj ) is an element of C ∞ (A ⊗ T ∗ M ).3 Let M be a hypercomplex manifold with q-holomorphic tangent space A. We will refer to elements of VA as A-type vector fields. We want to find an expression for λ( qj ⊗ vj ⊗ wj ) = qj ⊗ [vj . qj ⊗ wj ∈ VA . Then the Lie bracket λ on quaternionic vector fields preserves VA . qj ⊗ vj ⊗ wj = qj ⊗ ( vj wj − wj vj ) ∈ VA .6). 110 . Let vj . the Lie bracket [vj . Define VH = H ⊗ V to be the AH-module of smooth quaternion-valued vector fields on M . The relationship between these vector fields and the map λ : V ⊗ V → V is particularly interesting. wj ].2. vj wj − wj vj .11.Definition 7. T M ) be the vector space of smooth real vector fields on M . so that A is a q-holomorphic AH-bundle).2 Let M be a hypercomplex manifold with q-holomorphic tangent space A ⊂ H ⊗ T M . A) to be the AH-submodule of H-valued vector fields on M taking values in A ⊂ H ⊗ T M . the Lie bracket defines an AH-morphism (which we shall also call λ) λ : VH ⊗H VH → VH . so that λ : VA ⊗H VA → VA . Define VA = C ∞ (M. so VH = C ∞ (M. Here is the quaternionic version of Equation (7. Contracting the T ∗ M -factor with vj ∈ T M . we use the section product map φA. H ⊗ T M ). The formulation and proof is similar in spirit to Theorem 6. Let V = C ∞ (M. The Obata connection is an AH-connection on A (and in fact is anti-self-dual.4.2. and let VA denote the space of A-type vector fields on M .4. Theorem 7. which is effectively a Lie bracket operation on quaternionic vector fields.10) to regard elements of VA ⊗H VA as sections in C ∞ (A⊗H A) ⊂ C ∞ (H ⊗ T M ⊗ T M ). wj ] is given by the difference It follows immediately that λ proving the theorem. which means that qj ⊗ vj .

J. and K ensures that the connection with I = J = K = 0 is torsion-free. we have 1. and otherwise we would not have [v. 0) with respect to the complex structure Q.1. it follows that the Lie bracket must map sections of A⊗H A to sections of A. the bundle A is generated over H by the tensors of type (1. Thus λ is antisymmetric.2 Quaternionic Lie Algebras The result of the previous section encourages us to think of the vector fields VA as a quaternionic Lie algebra with respect to the Lie bracket map λ. A quaternionic Lie algebra will be an AH-module A together with an AH-morphism λ : A⊗H A → A. Since the fibres of A are stable. and see how it fits in with some of Joyce’s other algebraic structures over the quaternions. If every complex structure Q ∈ S 2 is integrable. whose properties reflect those of a Lie bracket on a real or complex vector space: namely antisymmetry and the Jacobi identity. 0). let (V. Here are the axioms for a quaternionic Lie algebra: Axiom QL.0 1. where VQ denotes the vector fields which are of type (1.0 1. 111 . λ) is a quaternionic Lie algebra if it satisfies Axiom QL. We will often refer to A itself as a quaternionic Lie algebra when the map λ is understood.0 H · ιq (TQ M ) and A⊗H A = Q∈S 2 1. By way of explanation.The reason why the hypercomplex structure must be integrable to obtain this result is that the integrability of I. but rather about linear maps on tensor products.10. In other words. VQ ] ⊆ VQ 1. and the bundle A⊗H A is generated by the tensors of type (2. we have A= Q∈S 2 1. 0) with respect to the different complex structures. As always. Then in terms of tensor products the Jacobi identity is λ ◦ (id ⊗λ)(x ⊗ y ⊗ z + y ⊗ z ⊗ x + z ⊗ x ⊗ y) = 0. Another way to understand this result is in terms of the decomposition of tensors with respect to different complex structures.0 for all Q ∈ S 2 . (iii) The composition λ ◦ (id ⊗H λ) defines an AH-morphism from A⊗H A⊗H A to A such that Λ3 A ⊂ ker(λ ◦ (id ⊗H λ)).2. w] = v w − w v. (i) A is an AH-module and there is an AH-morphism λ = λA : A⊗H A → A called the Lie bracket.0 H · ιq (TQ M ⊗ TQ M ).2. We describe this idea as an abstract algebraic structure. we do not talk about bilinear maps. Just as in Theorem 7. This is the H Jacobi identity for λ. Axiom QL(iii) is probably the least familiar-looking of these axioms. Definition 7.4 The pair (A. λ) be a real or complex Lie algebra. 7.0 [VQ . 2 (ii) SH A ⊂ ker λ.0 1.

Suppose that ξ : = A⊗H A → A⊗H Y satisfies Joyce’s Axiom L. 7. and the pair (A⊗H Y.1.3 Hypercomplex Lie groups As a final example. Theorem 7. This is why.5 Let M be a hypercomplex manifold. λ) is a quaternionic Lie algebra. quaternionic Lie algebras and Poisson algebras are defined by an AH-morphism ξ : A⊗H A → A⊗H Y . The space of quaternion-valued vector fields VH is a quaternionic Lie algebra with respect to the Lie bracket operator λ : VH ⊗H VH → VH . we consider hypercomplex structures on compact Lie groups. but in a different fashion. Then ξ satisfies Joyce’s Axiom L. Define an AH-morphism ξ = (λ⊗H id ⊗H id) ◦ τ : (A⊗H Y )⊗H (A⊗H Y ) → (A⊗H Y )⊗H Y. This is the quaternionic version of the well-known fact that on a complex manifold. we see that the Jacobi identity is equivalent to the condition that λ ◦ (id ⊗λ)(x ∧ y ∧ z) = 0. In order to define an HL-algebra we need to form a map from (A⊗H Y )⊗H (A⊗H Y ) to (A⊗H Y )⊗H Y . which define quaternionic analogues of Lie algebras and Poisson algebras: but instead of a Lie bracket λ : A⊗H A → A. Since a hyperk¨hler manifold M a a has three independent symplectic forms. Then for every = AH-module A there is a map id ⊗H η : A⊗H Y → A⊗H H ∼ A. Let τ be the natural isomorphism τ : A⊗H Y ⊗H A⊗H Y → A⊗H A⊗H Y ⊗H Y which interchanges the second and third factors. To go in the opposite direction. suppose that (A. he writes down axioms called Axiom L and Axiom P.Since we can identify x⊗y −y ⊗x with x∧y. so that Λ3 V ⊂ ker(λ ◦ (id ⊗λ)). Precisely which HL-algebras may be obtained from quaternionic Lie algebras and vice versa using these constructions remains open to question. Since the early 112 . Joyce describes the three = Poisson brackets using a single map ξ : PM ⊗H PM → PM ⊗ I.2.2. λ) is a quaternionic Lie algebra in the sense of Axiom QL. Axiom QL follows from the corresponding identities satisfied by the Lie bracket on real vector fields. Joyce has already considered the notion of a Lie algebra over the quaternions. Using the fact that I ∼ R3 . and show how these give rise to finite-dimensional quaternionic Lie algebras. and consider the AH-module A⊗H Y . ξ) forms an HL-algebra. §6]. 0) vector fields form a complex Lie subalgebra of the complex vector fields.3 now shows that the A-type vector fields VA form a quaternionic Lie subalgebra of VH . for Joyce. Joyce’s quaternionic Lie bracket is a map ξ : A⊗H A → A⊗H Y . In [J1. the (1. It follows immediately that the pair (A. The space of such AH-morphisms is of course Y † ∼ V2 . two functions f and g have three different Poisson brackets and the H-algebra PM of q-holomorphic functions on M has three independent Poisson structures. = The reason for this is that Joyce’s main purpose is to describe the algebraic structure of q-holomorphic functions on hyperk¨hler manifolds. where Y ∼ U2 is the AH-module of Example 4. Example 7. Define a Lie bracket λ : A⊗H A → A by setting λ = (id ⊗H η) ◦ ξ.2. We can relate these two algebraic ideas very simply by choosing an AH-morphism η : Y → H. Recalling that (Y † )∗ ∼ = V2 = I. we have a map ξ : PM ⊗H PM → ιPM (PM ) ⊗ (Y † )∗ which is in fact an AHmorphism whose image is contained in PM ⊗H Y .

In this case we define the subspace gA = {v0 + i1 v1 + i2 v2 + i3 v3 : v0 + I1 v1 + I2 v2 + I3 v3 = 0} ⊂ gH which we shall call the space of A-type elements of gH . 3 113 . Lemma 4. Then g can be decomposed as n n g=b⊕ k=1 dk ⊕ k=1 fk . which give rise to distinct hypercomplex structures. . This gives an isomorphism pu(1) ⊕ g ∼ Hl . . Note that not all ‘Lie groups possessing a complex structure’ are complex Lie groups. and (ii) fk is closed under the Lie bracket with dk . The quaternionic Lie algebra structure of g is much more interesting when g admits a hypercomplex structure as described above. .1] Let G be a compact Lie group. .1 [J3. . Then there exists an integer p with 0 ≤ p ≤ Max(3. . The action of dk on fk gives an isomorphism = ∼ Hm . fk ] = {0} whenever l < k. 2. fn are (possibly empty) vector subspaces of g.7) where b is abelian. such that for each k = 1. fk satisfies the following two conditions: (i) [dl . in other words a hypercomplex fk = = structure. dk is a subalgebra of g isomorphic to su(2). There is a strong link between these hypercomplex structures and the quaternionic Lie algebras of the previous section. Spindel et al. whose proof is an extension of Borel’s celebrated result that the quotient of a compact Lie group by its maximal torus is a homogeneous complex manifold. 3 This was first announced by Samelson. [SSTP] and Joyce [J3] demonstrated independently that these results extend to hypercomplex geometry. rk G) such that U(1)p × G admits a left-invariant homogeneous hypercomplex structure. Normally there are many choices to be made in such an isomorphism. It relies on being able to decompose the Lie algebra of a compact Lie group as follows: Lemma 7. because their multiplication and inverse maps might not be holomorphic.1950s. this decomposition allows us to define a hypercomplex structure on pu(1) ⊕ g as follows. (7. and f1 . the mathematical world has been aware that every compact Lie group of even dimension is a homogeneous complex manifold. In 1988 and 1992 respectively.3.2] Let G be a compact Lie group. for some integer m. Lie groups with integrable leftinvariant hypercomplex structures then give rise to interesting quaternionic Lie algebras. Joyce’s approach also gives hypercomplex structures on more general homogeneous spaces. with Lie algebra g. By adding an additional p (with 0 ≤ p ≤ Max(3. and the Lie bracket action of dk on fk is isomorphic to the sum of m copies of the basic representation V1 of su(2) on C2 . Theorem 4. This leads to the following result: Theorem 7. For a 1-dimensional subspace bk of pu(1) ⊕ b there is an isomorphism bk ⊕ dk ∼ R ⊕ I = H. b + k dk contains the Lie algebra of a maximal torus of G. Then gH ≡ H ⊗ g is a quaternionic Lie algebra because Axiom QL is obviously satisfied. That such a hypercomplex structure on the vector space pu(1) ⊕ g defines an integrable hypercomplex structure on the manifold U(1)p × G follows from Samelson’s original work on homogeneous complex manifolds. n. .3. Suppose that g is any real Lie algebra. rk G) ) copies of the abelian Lie algebra u(1) to b if necessary.2 [J3. .

The map d is induced by the dual of the Lie bracket. and this induces = • ∗ a representation of su(2) on Λ T M by the usual Leibniz rule τv (w1 ∧ w2 ) = [v. thanks to the decomposition u(2) = u(1) ⊕ su(2). The map d extends to a unique antiderivation on Λ• g∗ in the usual fashion. j. = (7. It is well-known that U(2) is isomorphic to U (1) ×Z2 SU(2).3. so that [e0 . which accounts for all the de Rham cohomology since it is known that every de Rham cohomology class has a G-invariant representative. which is a linear map λ : g ⊗ g → g.2.7). the set of A-type vector fields on G is closed under the Lie bracket operator λ. Thus T ∗ M ∼ V0 ⊕ V2 . 3}. so that τv (w) = [v. By Theorem 7. Then gA is a quaternionic Lie subalgebra of gH . w] for v ∈ su(2). It follows that de0 = 0 and dei = ej ∧ ek . e02 . and let gA ⊂ gH be the subspace of A-type elements of gH .3. C) preserving the standard hermitian metric on C2 . 2. The Lie algebra u(2) appears naturally in the form of Equation (7. e021 ⊕ e123 e0123 114 ∼ V0 ⊕ V2 = ∼ V2 ⊕ V2 = ∼ V2 ⊕ V0 = ∼ V0 . e0 generates a copy of the trivial representation V0 and e1 . Thus e0 and e123 both generate de Rham cohomology classes. e3 . e3 is just the adjoint representation of su(2). i. e2 . k ⊂ {1. In this way. this means that dω(u. it preserves gA . ej ] = εijk ek where i. e013 . This subgroup acts on u(2) via the adjoint representation. complex or hypercomplex manifold can then be rephrased in terms of the cohomology of the complex (Λ• g∗ . e2 .3.8) . which is diffeomorphic to the Hopf surface S 1 × S 3. so its dual is the map d = λ∗ : g∗ → Λ2 g∗ . I2 . v) = ω([u. e03 e032 . w2 ]. whose theory is possibly more subtle for this reason. e12 ⊕ e01 .4 Let M = U(2) be the unitary group in two dimensions.Corollary 7. k} is an even permutation of {1. This is less obvious for Dolbeault and quaternionic cohomology groups. 3}.3 Let G be a hypercomplex Lie group with hypercomplex structure (I1 . e3 . This gives the following decompositions: T ∗M Λ2 T ∗ M Λ3 T ∗ M Λ4 T ∗ M = = = = e0 ⊕ e1 . v]) ω ∈ g∗ v. ej ] = 0 and [ei . This is true at least for G-invariant forms. Joyce’s hypercomplex structures on compact Lie groups give rise to many interesting finite-dimensional quaternionic Lie algebras. I3 ). the subgroup of GL(2. Since λ also preserves the left-invariant vector fields g. which is V2 . w1 ] ∧ w2 + w1 ∧ [v. e0 ] = 0. w ∈ u(2). On a Lie group G the exterior differential d on G-invariant k-forms can be expressed as a formal differential d : Λk g∗ → Λk+1 g∗ . In practice. This complex is best described using the structure of u(2) as a representation τ of su(2) = e1 . Since λ is antisymmetric it is effectively a map λ : Λ2 g → g. j.e. Example 7. b2 (M ) = 0. Let {eα } be the dual basis for g∗ . e31 . and we have b0 (M ) = b1 (M ) = b3 (M ) = b4 (M ) = 1. w ∈ g. Let u(1) = e0 and su(2) = e1 . 2. e2 . We can also begin to calculate the quaternionic cohomology groups of these manifolds. Since [v. Proof. e3 e23 . d). e2 . where {i. Questions about the cohomology of G as a real.

where eab+cd = eab + ecd .1 .0 . (7.2 HD (M ) = 0.0 → F3.0 −→ E1. 0.r homology groups HD (M ) and Hδ (M ). In other words. In four dimensions.−1 and F4. E2. The sequence 0 −→ E0. However. as indeed they are. e02 . we have 3e0 + i1 e1 + i2 e2 + i3 e3 H ∼ U−1 = × ∼ e123 − i1 e032 − i2 e013 − i3 e021 H ∼ U−1 . e31 . The group SU(3) provides an interesting case. 0. but the d operator is best described using the second. The quaternionic algebra of such phenonomena might be interesting and merit closer study. since b1 (SU(3)) = 0. For example. e3 → e23 . e03 . This allows us to read off cohomological information. The quaternionic cohomology of M is defined using the first action.1 Hδ (M ) = 3. Surprisingly.2 is the space of self-dual 2-forms e01+23 .9) Moving to quaternion-valued forms.0 Hδ (M ) = H 4. Explicitly. By collating the two we obtain a complete picture of the situation. including the quaternionic cok. The first is the representation T ∗ M ∼ 2V1 defined by the hypercomplex structure. Thus once we have shown that de1 = e23 . F3. The principle would nonetheless be very much the same. e2 .1 Hδ (SU(3)) = 0. though because of the additional four dimensions this option is difficult and complicated. The image under d of T ∗ M is e01 .2 is an su(2)-equivariant surjection onto E2. the sequence 0 → B → F2.1 HD (M ) = e0 ∼ R = 2. A similar analysis to that above may provide the correct results. e02+31 . we have two representations of sp(1) on T ∗ M . e03+12 .0 . this theory seems to be lacking or at best 115 . A logical precursor to developing the quaternionic cohomology theory of hypercomplex Lie groups would be to understand thoroughly the Dolbeault cohomology of homogeneous complex manifolds. e12 . they are clearly not AH-exact.2 . and the projection π2.2 −→ 0 is therefore exact at E2. and essentially involves comparing different sp(1)-actions. one defined by Joyce’s hypercomplex structure and the other arising from the adjoint representation of a particular subalgebra su(2) ⊂ su(3). it follows immediately that d gives an su(2)-equivariant isomorphism d : e1 . it is easy to show that non-zero δ-cohomology occurs only at F0.r k.1 −→ E2.10) It follows from the property of ellipticity that the cohomology sequences should be exact.0 HD (M ) = R 1.1 → 0 being exact. it is easy to demonstrate that 1. To do this we compare the decomposition of Equation (7.8) with the decomposition of Λk T ∗ M induced by the hypercomplex structure.−1 Hδ (M ) = It would be desirable to extend this work to higher-dimensional groups and homogeneous spaces. and the second = is the representation T ∗ M ∼ V0 ⊕ V2 given by the adjoint action of the subalgebra = su(2) ⊂ u(2). because they are not exact on the primed parts.2 .The benefit of this approach is that the map d is su(2)-equivariant.4 Hδ (M ) = e0123 1. = × H = H (7. and we obtain the standard self-dual cohomology of S 1 × S 3 . F1. Quaternionic cohomology is in this case certainly not a subset of de Rham cohomology.

but the link between theory and results in this paper is quite opaque and certainly contains some mistakes. conjecturing that the cohomology of these complexes is the same. §2]. Pittie’s paper [P] uses a simpler bicomplex than Bott’s. The theory for complex groups and homogeneous spaces is documented in Bott’s important paper [Bot. This would be an interesting area for future research. 116 .extremely obscure. Certain hypercomplex nilmanifolds discussed by Dotti and Fino [DF] might also provide fruitful rseults.

362 (1978). Ebbinghaus et al: Numbers. I. Fulton. 995-1008. 601-602. o Springer-Verlag. C. Amer. 18 (2000). 1985). Sendai. Fujiki: On the de Rham cohomology group of a compact K¨hler syma plectic manifold. Besse: Einstein Manifolds. Baston: Quaternionic complexes. Lond. A. Graduate Texts in Math. C. 123 (1990) (in particular M. Global Anal. Ann. J. 115-118 and 296 (1983). R. Bonan: Sur l’alg`bre ext´rieure d’une vari´t´ presque hermitienne e e ee quaternionique. 66 no. 2 (1957). pp. pp. A.J. 47-59. T. Math. of Math. G. Graduate Texts in Math.L. Geom. Hamilton’s Quaternions). Pure Math. 105-165. pp. Springer-Verlag. Harris: Representation Theory.A.Bibliography [AHS] M. Hitchin. Cambridge University Press. Remmert: Chapter 7. R. D. [Bas] [Ben] [Bes] [Bon] [Bot] [BD] [D] [DF] [EKR] [F] [FH] 117 . Fino: Abelian Hypercomplex 8-dimensional nilmanifolds. Geom. Bott: Homogeneous Vector Bundles. pp. Koecher. Graduate Texts in Math. R. tom Dieck: Representations of Compact Lie Groups. Deavours: The Quaternion Calculus. pp. 29-52. Atiyah. Benson: Representations and cohomology I: Basic representation theory of finite groups and associative algebras.F. A. N. Dotti. Springer-Verlag. I. pp. Cambridge (1991). Singer: Self-duality in four-dimensional Riemannian geometry. 8 (1992). Ann. Monthly 80 (1973). E. 10. pp. Paris 295 (1982). pp. Br¨cker. 203-248. Adv. R. J. Sc. Soc. Phys. 98 (1985). Springer-Verlag (1987). Acad. W. 1987 (Algebraic Geometry. Proc R. Stud. A. 425-61. J. T.M. 129 (1991).

Oxford. 15. Y. Math. T. Math. Kraines: Topology of Quaternionic Manifolds. A. V. Tohoku a Math. 189-95.M.[GH] P. Royal Irish Acad. Galicki. August 5th 1865. John Wiley and Sons (1978). D. S. Harris: Principles of Algebraic Geometry. K. Ann. Proc. 424-434. M. 163-198. J. J. M. 357-367. Nonlinearity 1 (1988). Commun. 49 (1998). pp. N. Proc. 129-162. Hitchin. 323-340. Y. pp. Sir William Rowan Hamilton: On a new Species of Imaginary Quantities connected with a theory of Quaternions. a Sir William Rowan Hamilton: Letter to the Rev. Bull. 743-761. 3 (1847). 19-22. D. J. Joyce: Hypercomplex Algebraic Geometry. 1-16. 1263-1268. Quart. 2 (1843). Phys. Griffiths. Hamilton’s Math. pp. Kobayashi: Differential Geometry of Complex Vector Bundles. J. Math. 1843. Math. D. AMS 10 (1989). pp. October 17th. Geom. Edinburgh and Dublin Philosophical Magazine and Journal of Science 25 (1844). J. Archibald Hamilton. Math. Sir William Rowan Hamilton: Letter to John T. Nitta: Vector bundles over Quaternionic K¨hler manifolds. Papers III p. 40 (1988). S. Princeton University Press (1987). Joyce: Compact Hypercomplex and Quaternionic manifolds. pp. S. Mamone Capria. 35 (1992). Diff. Pittie: The nondegeneration of the Hodge–de Rham spectral sequence. 108 (1987). 517-530. pp. pp. Quillen: Quaternionic Algebra and Sheaves on the Riemann Sphere. H. Oxford 49 (1998). Roˇek: Hyperk¨hler metrics o c a and supersymmetry. Phys. Math. J. 425-440. pp. Lindstr¨m. pp.Graves. Trans. Poon: Duality and Yang-Mills fields on quaternionic K¨hler manifolds. 118 [GP] [H1] [H2] [H3] [H4] [HKLR] [J1] [J2] [J3] [K] [Kr] [MS] [N] [P] [Q] . London. Salamon: Yang-Mills fields on quaternionic spaces. pp.J. Joyce: The Hypercomplex quotient and the Quaternionic quotient. 535-589. Soc. pp. 32 (1991). U. Royal Irish Acad. Sir William Rowan Hamilton: On Quaternions. D. Quart. V. pp. Amer. 290 (1991). Karlhede. 122 (1966) pp.

Cambridge Monographs on Math. de Leo: Quaternions and Special Relativity. Math. B. 8 (1998). Troost. A. Ann. a 143-171. 65 (1980). Phys. Springer-Verlag (1985).O. J. Lorentz Transformations and the ConwayDirac-Eddington Matrices. Pitman Res. M. Dublin Inst.[R] R. 31-55. Spindel. S. pp. R. Springer-Verlag lecture notes in math. pp. 21 (1972). 1-20. 662-698. pp.L. B 308 (1988). Sudbery: Quaternionic Analysis. Graduate Texts in Math. Norm. R. Harlow (1989).3 (1996). Salamon: Differential Geometry of Quaternionic Manifolds. 199-225. A. Math. Reyes-Carri´n: A generalisation of the notion of instanton. S. [S1] [S2] [S3] [S4] [SSTP] [dL] [Su] [Sw] [Sy] [V] [v. van der Waerden: A History of Algebra. 37. J. 85 (1979). pp. Phil. Salamon: Topics in four-dimensional Riemannian geometry. S. R. Sup. a Alg. Sci´ ent. Diff. o and its Apps. University of Lancaster (1989). S. Nuclear Phys.O Wells: Differential Analysis on Complex Manifolds. series 4 no. S. Synge: Quaternions. a Proc. Verbitsky: Hyperholomorphic bundles over a Hyperk¨hler manifold. for advanced studies. pp. Soc. van Proeyen: Extended supersymmetric σ-models on group manifolds. Proc. 5 (1996). 67 (1982). pp. 2955-2968. Salamon: Riemannian geometry and holonomy groups. 201. 633-669. Salamon: Quaternionic K¨hler Manifolds. Curvature Geom. Math. (1990). Wells: Twistor Geometry and Field Theory. workshop. Longman. Ward. Ph. J. pp. Phys. Invent.S.L. A. Paris. Springer-Verlag. pp. W. Sevrin.d. Comm. 19 (1986). Geom. 165-173. 1022 (1982). Notes in Math. Geom.W] [WW] [W] 119 . Swann: Quaternion K¨hler Geometry and the Fundamental 4-form. series A no. Geometry seminar “Luigi Bianchi”. Ec. A. Camb.