Notes on Cliord Algebras and

Spin Groups
Thomas Hjortgaard Danielsen
Contents

1 Cliord Algebras 5
1.1 Elementary Properties . . . . . . . . . . . . . . . . . . . . . . . . 5
1.2 Classication of Cliord Algebras . . . . . . . . . . . . . . . . . . 11
1.3 Representation Theory . . . . . . . . . . . . . . . . . . . . . . . . 15

2 Spin Groups 19
2.1 The Cliord Group . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Pin and Spin Groups . . . . . . . . . . . . . . . . . . . . . . . . . 22
2.3 Double Coverings . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.4 Spin Group Representations . . . . . . . . . . . . . . . . . . . . . 28
2.5 Spin Structures . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31
2.6 The Dirac Operator . . . . . . . . . . . . . . . . . . . . . . . . . 38

3
4
Chapter 1

Cliord Algebras
1.1 Elementary Properties
In this chapter we will introduce the Cliord algebra and discuss some of its
elementary properties. The setting is the following:
Let V be a nite-dimensional vector space over the eld (predominantly R
K
or C) and ϕ : V × V −→ K V . ϕ is said to be
a symmetric bilinear form on
positive (resp. negative ) denite if for all 0 6= v ∈ V we have ϕ(v, v) > 0 (resp.
ϕ(v, v) < 0). ϕ is called non-degenerate if ϕ(v, w) = 0 for all v ∈ V implies
w = 0. From a bilinear form we construct a quadratic form Φ : V −→ K given
by Φ(v) := ϕ(v, v). We can recover the original bilinear form by the polarization
identity :
Φ(u + v) = ϕ(u + v, u + v) = Φ(u) + Φ(v) + 2ϕ(u, v),

hence
1
ϕ(u, v) = (Φ(u + v) − Φ(u) − Φ(v)). (1.1)
2
Thus we have a 1-1 correspondence between symmetric bilinear forms and
quadratic forms. Thus a quadratic form Φ is called positive denite /negative
denite /non-degenerate if ϕ is.

Denition 1.1. Let (V, Φ) be a vector space with a quadratic form Φ. The
associated Cliord algebra Cl(V, Φ) (abbreviated Cl(Φ)) is an associative, unital
algebra over K : V −→ Cl(Φ) obeying the relation i(v)2 =
with a linear map iΦ
Φ(v) · 1 (1 is the unit element of Cl(Φ)). Furthermore (Cl(Φ), iΦ ) should have
the property that for every unital algebra A, and every linear map f : V −→ A
2
satisfying f (v) = Φ(v) · 1 there exists a unique algebra homomorphism fb :

Cl(Φ) −→ A, such that f = fb ◦ iΦ .

Two questions immediately arise: given (V, Φ), do such objects exist and if
they do, are they unique? Fortunately, the answer to both questions is yes:

Proposition 1.2. For any vector space V with a quadratic form Φ let = be the
two-sided ideal in the tensor algebra T (V ) spanned by all elements of the form
a ⊗ (v ⊗ v − Φ(v) · 1) ⊗ b (for a, b ∈ T (V ) and v ∈ V ). Then T (V )/=, with the
map iΦ : V −→ T (V )/= being π ◦ ι where ι : V −→ T (V ) is the injection of V
into T (V ), and π : T (V ) −→ T (V )/= is the quotient map, is a Cliord algebra,
and any other Cliord algebra over (V, Φ) is isomorphic to this one.
Proof. Uniqueness. Assume that Cl1 (Φ) and Cl2 (Φ) with linear maps i1 :
V −→ Cl1 (Φ) and i2 : V −→ Cl2 (Φ) are Cliord algebras. Since Cl1 (Φ) is

5
6 Chapter 1  Cliord Algebras

a Cliord algebra, and i2 i2 (v)2 = Φ(v) · 1, it induces

is linear and satises
an algebra homomorphism bi2 : Cl1 (Φ) −→ Cl2 (Φ); likewise i1 induces an al-
i1 : Cl2 (Φ) −→ Cl1 (Φ) such that the following diagram
gebra homomorphism b
commutes:

x FFF
V
xxx FFi2
x i1
FF
xxx FF
x| "
i2
/
b
Cl1 (Φ) o Cl2 (Φ)
i1
b

We see that
i2 = bi2 ◦ i1 = bi2 ◦ bi1 ◦ i2
and since bi2 ◦ bi1 is the unique map satisfying this, it must be idCl (Φ) . Likewise
2
bi1 ◦ bi2 = idCl (Φ) which means that the two Cliord algebras are isomorphic.
1
This proves uniqueness.
Existence. We now show that Cl(Φ) := T (V )/= is indeed a Cliord algebra. iΦ
2
is easily seen to satisfy iΦ (v) = Φ(v) · 1, where 1 ∈ Cl(Φ) is the coset containing
the unit element of T (V ). Now let f : V −→ A be linear with f (v)2 = Φ(v)·1. By
the universal property of the tensor algebra this map factorizes uniquely through
T (V ) to an algebra homomorphism f 0 : T (V ) −→ A, such that f = f 0 ◦ ι. f 0
0 2
inherits the property (f (v)) = Φ(v) · 1, and consequently
f 0 (v ⊗ v − Φ(v) · 1) = (f 0 (v))2 − Φ(v)f 0 (1) = (f 0 (v))2 − Φ(v) · 1

so f 0 vanishes on =. Therefore it factorizes uniquely through Cl(Φ) to fb :

Cl(Φ) −→ A, such that f = fb ◦ π ◦ ι = fb ◦ iΦ . Thus Cl(Φ) = T (V )/= is a
Cliord algebra.

We immediately see that if the quadratic form Φ on V is identically 0, the

Cliord algebra Cl(Φ) is nothing but the well-known exterior algebra Λ∗ (V ).
From now on we will write i instead of iΦ where no confusion is possible. A
simple calculation reveals that for all u, v ∈ V : i(u + v)2 = i(u)2 + i(v)2 + i(u) ·
i(v) + i(v) · i(u). A comparison of this with the polarization identity using that
Φ(u + v) · 1 = i(u + v)2 , Φ(u) · 1 = i(u)2 and Φ(v) · 1 = i(v)2 yields the following
useful formula:
i(u) · i(v) + i(v) · i(u) = 2ϕ(u, v) · 1. (1.2)

It can be used to prove the following:

Proposition 1.3. Assume Φ to be non-degenerate and let {e1 , . . . , en } be an

orthogonal basis for V . Then the set consisting of 1 and all products of the form
i(ej1 ) · · · i(ejk ), where j1 < · · · < jk and 1 ≤ k ≤ n is a basis for Cl(Φ). In
particular i : V −→ Cl(Φ) is injective, and the dimension of Cl(Φ) is 2n .
Proof. First of all we will show that Cl(Φ) is isomorphic to a subalgebra of
EndK (Λ∗ V ). For each v∈V dene endomorphisms ε(v), ι(v) and c(v) on Λ∗ V
by

ε(v) : v1 ∧ · · · ∧ vk 7−→ v ∧ v1 ∧ · · · ∧ vk .
k
X
ι(v) : v1 ∧ · · · ∧ vk 7−→ (−1)j+1 ϕ(v, vj )v1 ∧ · · · ∧ vbj ∧ · · · ∧ vk .
j=1

c(v) = ε(v) + ι(v).

It's a matter of calculations to show that ε(v)2 = ι(v)2 = 0 and that we have

2
c(v) = Φ(v) · 1
c(u)c(v) + c(v)c(u) = 2ϕ(u, v) · 1.
1.1 Elementary Properties 7

The map c : V −→ EndK (Λ∗ V ) is injective for if c(v) = 0 then c(v)2 = Φ(v)·1 =
0 i.e. v = 0 by non-degeneracy of Φ. By the universal property of Cliord alge-
bras there exists an algebra homomorphism b c : Cl(Φ) −→ EndK (Λ∗ V ) extending
c. By injectivity of c this map is an algebra isomorphism on the subalgebra of
EndK (Λ∗ V ) generated by the set {c(v) | c ∈ V }.
∗
Now we consider the subalgebra of EndK (Λ V ). Any element of the form
c(v1 ) · · · c(vm ) can, by linearity of c and by the relations above be written as a
linear combination of elements of the form c(ei1 ) · · · c(eik ) where i1 < · · · < ik .
Let B denote the set of these element. Hence, these elements span the subalge-
bra. To see that they are linearly independent, observe that c(ei1 ) · · · c(eik ) maps
1 to ei1 ∧ · · · ∧ eik . This evaluation map is linear and since the set {ei1 ∧ · · · ∧ eik }
∗
is a basis for Λ V , B is a linearly independent set, hence a basis. By the iso-
morphism b c the set
{i(ei1 ) · · · i(eik ) | i1 < · · · < ik }
is a basis for Cl(Φ).

Since i is injective we can imagine V as sitting as a subspace of Cl(Φ). Thus,

henceforth we won't bother to write i(v) but will write v instead.
The evaluation map gives, by composing it with the isomorphism mentioned
in the proof above, a linear map σ : Cl(Φ) −→ Λ∗ V called the symbol map . If
{e1 , . . . , en } is an orthogonal basis, then it maps ei1 · · · eik to ei1 ∧ · · · ∧ eik , and
is thus a linear isomorphism (although not necessarily an algebra isomorphism!)
The inverse map Q : Λ∗ V −→ Cl(Φ), mapping ei1 ∧ · · · ∧ eik to ei1 · · · eik is
called the quantization map .
If (V, ϕ) and (W, ψ) are two vector spaces with bilinear forms, a linear map
f : V −→ W is called orthogonal w.r.t. the bilinear forms if ψ(f (v), f (w)) =
ϕ(v, w) (or equivalently, if Ψ(f (v)) = Φ(v)). We say that (V, ϕ) and (W, ψ) are
isomorphic if there exists an orthogonal linear map f : V −→ W which is also a
vector space isomorphism and such a map we call an orthogonal isomorphism . If
ϕ is non-degenerate, one can show that an orthogonal endomorphism f : V −→
V is automatically an isomorphism. Thus the set of orthogonal endomorphisms
of V is a group O(V, Φ) (or just O(Φ)), called the orthogonal group . This is a
closed subgroup of the Lie group GL(V ), and thus itself a Lie group. Picking
only those orthogonal endomorphisms having determinant 1 gives the special
orthogonal group SO(V, Φ) or just SO(Φ). This is again a Lie group, being a
closed subgroup of O(Φ).
A curious fact is that there exists an isomorphism among the orthogonal
groups. To see this, let ϕ : (Rp+q , Φp,q ) −→ (Rp+q , Φq,p ) be an anti-orthogonal
linear map, i.e. satisfying

Φq,p (ϕ(x)) = −Φp,q (x)

and consider the map Ψ : O(p, q) −→ O(q, p) given by Φ(A) = ϕ ◦ A ◦ ϕ−1 . It is

easily checked that this is a Lie group isomorphism. Restricting Ψ to SO(p, q),
it maps into SO(q, p) since the determinant is unaltered by a conjugation. Thus
∼
we also have an isomorphism SO(p, q) −−→ SO(q, p).
The next proposition shows that Cl is a covariant functor from the category
of vector spaces with bilinear forms and orthogonal linear maps to the category
of associated unital algebras and algebra homomorphisms.

Proposition 1.4. An orthogonal linear map f : V −→ W between (V, Φ) and

(W, Ψ) induces a unique algebra homomorphism f : Cl(Φ) −→ Cl(Ψ) satisfying
f (iΦ (v)) = iΨ (f (v)). If f is an orthogonal isomorphism, then f is an algebra
isomorphism.
8 Chapter 1  Cliord Algebras

If, furthermore, we have a vector space U with quadratic form Θ and linear
maps f : V −→ U and g : U −→ W satisfying Θ(f (v)) = Φ(v) and Ψ(g(u)) =
Θ(u), then g ◦ f = g ◦ f .
Exercise 1. Prove the preceding proposition.

The most interesting examples of Cliord algebras show up when the bilin-
ear form is non-degenerate. For the time being we consider only real vector
spaces and Cliord algebras. A prominent example of a vector space with non-
degenerate bilinear form is (Rp+q , ϕp,q ) where ϕp,q (ei , ej ) = 0 if i 6= j and
(
1, i≤p
ϕp,q (ei , ei ) =
−1, i > p

(e1 , . . . , ep+q is the standard basis of Rp+q ). The associated quadratic form is
denoted Φp,q and the corresponding Cliord algebra is denoted Clp,q . In this
case O(Φp,q ) and SO(Φp,q ) are the well-known orthogonal groups O(p, q) and
SO(p, q).
Example 1.5. 1) Let us consider the vector space R with the single basis
element e1 := 1 and the quadratic formΦ0,1 (x1 e1 ) = −x21 . By Proposition 1.3
{1, e1 } (where 1 is now the unit element of Cl0,1 ) is a basis for Cl0,1 . The fact
2
that e1 = Φ0,1 (e1 ) · 1 = −1 shows that the linear map Cl0,1 3 1 7−→ 1 ∈ C,
∼
e1 7−→ i ∈ C denes an algebra isomorphism Cl0,1 −−→ C and that the injection
R ,−→ Cl0,1 is given by x 7−→ ix, i.e. R sits inside Cl0,1 ∼= C as the imaginary
part. Thus, Cl0,1 is just the eld of complex numbers.

Exercise 2. Consider R2 with the standard basis {e1 , e2 } and the quadratic
2 2
form Φ0,2 (x1 e1 +x2 e2 ) = −x1 −x2 . Show that the corresponding Cliord algebra
can be identied with the algebra of quaternions H. Because of this, Cliord
algebras are sometimes called generalized quaternions.
Now, let (V, ϕ) be any real vector space with non-degenerate bilinear form ϕ,
and let {e1 , . . . , en } be a basis for V . We will consider the matrix (ϕij ) where
ϕij = ϕ(ei , ej ). Since ϕ is symmetric the matrix (ϕij ) is symmetric as well, i.e.
it can be diagonalized. Let λ1 , . . . , λn be the eigenvalues and f1 , . . . , fn a basis
of diagonalizing eigenvectors. This means that

ϕ(fi , fj ) = 0 if i 6= j and ϕ(fi , fi ) = λi .

Observe, that none of the eigenvalues are 0 (a 0 eigenvalue would violate non-
degeneracy of ϕ). Arrange the eigenvalues so that λ1 , . . . , λk are all strictly
positive and λk+1 , . . . , λn are all strictly negative. Dene
1
fei = p fi
|λi |
then we see that
(
1, i≤k
ϕ(fei , fej ) = 0 if i 6= j and ϕ(fei , fei ) = .
−1, i > k

A basis satisfying this is called a (real) orthonormal basis w.r.t. ϕ. Thus we have
proven

Theorem 1.6 (Classication of Real Bilinear Forms). Let (V, ϕ) be a real

vector space with non-degenerate bilinear form. Then there exists an orthonor-
mal basis for V and the map sending this basis to the standard basis for Rn is
an orthogonal isomorphism (V, ϕ) −→ (Rn , ϕk,n−k ).
1.1 Elementary Properties 9

In particular any real Cliord algebra originating from a non-degenerate

quadratic form is isomorphic to Clp,q for a certain p and q (Proposition 1.4).
The complex case is a bit dierent. On Cn we have bilinear forms ϕn given by
ϕn (ei , ej ) = δij . The corresponding quadratic form is denoted Φn . For an arbi-
trary complex vector space (V, ϕ), chose a basis {e1 , . . . , en } and write ϕ in this
basis (ϕij ). Again, since it is symmetric, it can be diagonalized. Let λ1 , . . . , λ n
be the eigenvalues and {f1 , . . . , fn } a diagonalizing basis of eigenvectors. None
of the eigenvalues are 0 and hence we can dene

1
fei = √ fi .
λi

This basis satises ϕ(fei , fej ) = δij , and is thus called a (complex) orthonormal
basis . Thus we have shown
Theorem 1.7 (Classication of Complex Bilinear Forms). Let (V, ϕ) be
a complex vector space with non-degenerate bilinear form. Then there exists an
orthonormal basis for V and the map sending this basis to the standard basis
for Rn is an orthogonal isomorphism (V, ϕ) −→ (Cn , ϕn ).
Appealing to Proposition 1.4 we see that a complex Cliord algebra is iso-
morphic to Cl(Φn ) for some n.
Again we consider the general situation of Cliord algebras over K. Now we
want to equip the Cliord algebra with two involutions t and α which we will
need later in the construction of various subgroups of Cl(Φ).
Proposition 1.8. Each Cliord algebra Cl(Φ) admits a canonical anti-auto-
morphism, i.e. a linear map t : Cl(Φ) −→ Cl(Φ) that for all x, y ∈ Cl(Φ)
satises
t(x · y) = t(y) · t(x) , t ◦ t = idCl(Φ) , t|V = idV .

Proof. Consider the involution J of the tensor algebra given by v1 ⊗· · ·⊗vk 7−→
J

vk ⊗ · · · ⊗ v1 (and extended by linearity). Now π ◦ J : T (V ) −→ Cl(Φ) is an

anti-homomorphism that vanishes on = since

J(a ⊗ v ⊗ v ⊗ b − a ⊗ (Φ(v) · 1) ⊗ b)
= J(b) ⊗ v ⊗ v ⊗ J(a) − J(b) ⊗ (Φ(v) · 1) ⊗ J(a)
= J(b) ⊗ (v ⊗ v − Φ(v) · 1) ⊗ J(a) ∈ =.

But then π ◦ J induces a unique anti-homomorphism t : Cl(Φ) −→ Cl(Φ) deter-

mined by t[x] = π ◦ J(x) (where [x] ∈ Cl(Φ) is the coset containing x ∈ T (V )).
It is easy to see that J(x · y) = J(y) · J(x), so we also have t(x · y) = t(y) · t(x).
J is clearly an involution, i.e. J ◦ J = idT (V ) . On one hand idT (V ) induces the
map idCl(Φ) , and on the other hand J ◦ J induces the map t ◦ t. By uniqueness
we conclude t ◦ t = idCl(Φ) . The last property, t|V = idV follows from the fact
that J(v) = v for v ∈ V .

Now for the construction of the second involution:

Proposition 1.9. Each Cliord algebra Cl(Φ) admits a canonical automor-

phism α : Cl(Φ) −→ Cl(Φ) which satises α ◦ α = idCl(Φ) and α|V = − idV .
Furthermore we have α(1) = 1 and α(ei1 · · · eik ) = (−1)k ei1 · · · eik .
Proof. Consider the linear bijection e : V −→ V given by v 7−→ −v . By the
α
functorial property of Cl it induces an automorphism α : Cl(Φ) −→ Cl(Φ), and
we see that
α◦α=α
e◦α
e = idV = idCl(Φ) .
10 Chapter 1  Cliord Algebras

The second property of α is obtained from the identity α◦i = i◦α

e which is
seen from the commutative diagram in the proof of Proposition 1.4. This gives
α(i(v)) = i(−v) = −i(v), and by considering v an element of Cl(Φ) (by virtue
of the injectivity of i) we have proven the claim.

The the calculation of t and α on our model algebras C and H we refer to

Example 2.6 in the next chapter.
Due to involutivity of α, we can split the the Cliord algebra in a direct sum
of two subspaces:
Cl(Φ) = Cl0 (Φ) ⊕ Cl1 (Φ) (1.3)
i
where Cl (Φ) = { x ∈ Cl(Φ) | α(x) = (−1)i x } for i = 0, 1. It is easily seen that
Cl (Φ) · Clj (Φ) ⊆ Cli+j (mod 2) (Φ) which says that the Cliord algebra is a Z2 -
i

graded algebra or a super algebra. Cl0 (Φ) is called the bosonic subalgebra (note
that it is actually a subalgebra), and Cl (Φ) is called the fermionic subspace .
1

We see that a product of the form v1 · · · vk for vi ∈ V is bosonic if k is even

0 1
and fermionic if k is odd. Elements of Cl (Φ) ∪ Cl (Φ) are called homogenous
elements and the degree of a homogenous element x is denoted by |x|.
If A and B are two Z2 -graded algebras we could form the tensor product
A ⊗ B with the usual product (a ⊗ b)(a0 ⊗ b0 ) = (aa0 ) ⊗ (bb0 ). However, this
is usually not of great interest since the resulting algebra is not Z2 -graded (at
least not non-trivially). Therefore we dene the so-called graded tensor product
or super tensor product A⊗B
b in the following way: As a vector space it is just
the ordinary tensor product A⊗B but the product is given on homogenous
elements a, a0 ∈ A and b, b0 ∈ B by
0
(a ⊗ b)(a0 ⊗ b0 ) = (−1)|a ||b| (aa0 ) ⊗ (bb0 ).

This gives A⊗B

b a natural grading by dening

b 0 = (A0 ⊗ B 0 ) ⊕ (A1 ⊗ B 1 )
(A⊗B)
b 1 = (A0 ⊗ B 1 ) ⊕ (A1 ⊗ B 0 ).
(A⊗B)

With this at hand we can accomplish out nal task of this section, showing how
Cl reacts to a direct sum of vector spaces. By an orthogonal decomposition of
(V, Φ), we understand a decomposition V = V1 ⊕ V2 such that if v = v1 + v2 we
have Φ(v) = Φ1 (v1 ) + Φ2 (v2 ) (equivalently if ϕ(V1 , V2 ) = 0).

Proposition 1.10. Assume we have an orthogonal decomposition V = V1 ⊕ V2

of (V, Φ), then Cl(V, Φ) ∼ b Cl(V2 , Φ2 ) as algebras.
= Cl(V1 , Φ1 )⊗
Proof. We use the universal property of Cliord algebras to cook up a map.
First, dene g : V −→ Cl(V, Φ) ∼
= Cl(V1 , Φ1 )⊗
b Cl(V2 , Φ2 ) by

g(v) = v1 ⊗ 1 + 1 ⊗ v2 .

A quick calculation shows that g(v)2 = Φ(v)(1 ⊗ 1) and thus by the univer-
sal property of Cl there exists an algebra homomorphism gb : Cl(V, Φ) −→
Cl(V1 , Φ1 )⊗
b Cl(V2 , Φ2 ) extending g . To see that this is indeed an isomorphism
choose a basis {e1 , . . . , em } for V1 and a basis {f1 , . . . , fn } for V2 and put them
together to a basis {e1 , . . . , fn } for V . A basis for Cl(V, Φ) consists of elements of
the form ei1 · · · eik fj1 · · · fjl where i1 < · · · < ik , k ≤ m and j1 < · · · < jl , l ≤ n.
Similarly a basis for Cl(V1 , Φ1 )⊗ b Cl(V2 , Φ2 ) is given by ei1 · · · eik ⊗ fj1 · · · fjl
(with the same restrictions on the indices as above). One can verify that

gb(ei1 · · · eik fj1 · · · fjl ) = ei1 · · · eik ⊗ fj1 · · · fjl

i.e. it maps basis to basis. Thus it is an algebra isomorphism.

1.2 Classication of Cliord Algebras 11

1.2 Classication of Cliord Algebras

In this section we set out to classify Cliord algebras originating from non-
degenerate bilinear forms. At rst we concentrate on real Cliord algebras. We
have taken the rst step in the classication in that we have shown in the
previous section, that it is enough to concentrate our treatment of Cliord
algebras to the case where the vector space is Rp+q equipped with the quadratic
form

Φp,q (x1 e1 + · · · + xp+q ep+q ) := x21 + · · · + x2p − (x2p+1 + · · · + x2p+q )

(where {e1 , . . . , ep+q } denotes the usual standard basis for Rp+q ). The associated
Cliord algebra was denoted Clp,q . As we saw in Example 1.5, we have

Cl0,1 ∼
=C and Cl0,2 ∼
= H. (1.4)

Likewise, one can show that the following isomorphisms hold

Cl1,0 ∼
=R⊕R and Cl2,0 ∼
= Cl1,1 ∼
= R(2) (1.5)

(K(n) is the algebra of n × n matrices over K, where K can be either R, C or

H). As is apparent, all four Cliord algebras are isomorphic to either an algebra
of matrices over R, C or H, or to a direct sum of two such algebras. The goal of
this section is to show that this is no coincidence. Actually, it is a consequence
of the Cartan-Bott Periodicity Theorem  which we will prove at the end of this
section  that this holds for every Cliord algebra Clp,q .
Before proving it, we need two lemmas:

Lemma 1.11. We have the following algebra isomorphisms :

R(m) ⊗ R(n) ∼
= R(mn) C ⊗R C ∼
=C⊕C C ⊗R H ∼
= C(2) H ⊗R H ∼
= R(4).

If K denotes either C or H, then

R(n) ⊗R K ∼
= K(n).

This is well-known so we won't prove it here.

1 Instead we show the next
lemma which really does most of the work:

Lemma 1.12. We have the following three algebra isomorphisms :

Cl0,n+2 ∼
= Cln,0 ⊗ Cl0,2 , Cln+2,0 ∼
= Cl0,n ⊗ Cl2,0 and
Clp+1,q+1 ∼
= Clp,q ⊗ Cl1,1 .
for all p, q, n ∈ N ∪ {0}.
Proof. To prove the rst isomorphism, the strategy is the following: we will
construct a linear map f : Rn+2 −→ Cln,0 ⊗ Cl0,2 , show that f factorizes
through Cl0,n+2 , and that the induced map fb : Cl0,n+2 −→ Cln,0 ⊗ Cl0,2 is
n+2
an isomorphism. Letting {e1 , . . . , en+2 } denote the standard basis for R ,
0 0 n 00 00 2
{e1 , . . . , en } the usual basis for R , and {e1 , e2 } the usual basis for R , and
thereby generators for the Cliord algebras Cl0,n+2 , Cln,0 and Cl0,2 respectively,
we dene f by
(
e0i ⊗ e001 e002 if 1 ≤ i ≤ n
f (ei ) = 00
1 ⊗ ei−n if n + 1 ≤ i ≤ n + 2

1 For a proof consult for instance [Lawson and Michelson], pp 26-27, Proposition 4.2.
12 Chapter 1  Cliord Algebras

For 1 ≤ i, j ≤ n we compute, using the rules for multiplication in a Cliord

algebra, that

f (ei ) · f (ej ) + f (ej ) · f (ei ) = (e0i ⊗ e001 e002 ) · (e0j ⊗ e001 e002 ) + (e0j ⊗ e001 e002 ) · (e0i ⊗ e001 e002 )
= (e0i e0j ) ⊗ (e001 e002 e001 e002 ) + (e0j e0i ) ⊗ (e001 e002 e001 e002 )
= −(e0i e0j + e0j e0i ) ⊗ (e001 e001 e002 e002 ) = −2δij 1 ⊗ 1

because e0i e0j + e0j e0i = 2δij · 1, as {e01 , . . . , e0n } is basis for Rn orthonormal w.r.t.
Φn,0 . For n + 1 ≤ i, j ≤ n + 2 we have

f (ei ) · f (ej ) + f (ej ) · f (ei ) = (1 ⊗ e00i−n ) · (1 ⊗ e00j−n ) + (1 ⊗ e00j−n ) · (1 ⊗ e00i−n )

= 1 ⊗ e00i−n e00j−n + 1 ⊗ e00j−n e00i−n
= 1 ⊗ (e00i−n e00j−n + e00j−n e00i−n ) = −2δij 1 ⊗ 1

where the last minus is due to {e001 , e002 } R2 orthonormal w.r.t.

being a basis for
Φ0,2 . A similar computation shows that f (ei )f (ej ) + f (ej )f (ei ) = 0 if 1 ≤ i ≤ n
and n + 1 ≤ j ≤ n + 2. But then for x = x1 e1 + · · · + xn+2 en+2 we have by
linearity of f that

f (x)2 = −(x21 + · · · + x2n+2 )1 ⊗ 1 = Φ0,n+2 (x) · 1 ⊗ 1.

Therefore f factorizes uniquely through Cl0,n+2 to an algebra homomorphism

fb : Cl0,n+2 −→ Cln,0 ⊗ Cl0,2 . f maps to a set of generators for Cln,0 ⊗ Cl0,2 ;
thus fb maps to a set of generators for Cln,0 ⊗ Cl0,2 . Since fb is an algebra ho-

momorphism, fb must then be surjective. Since

dim Cl0,n+2 = 2n+2 = 2n · 22 = (dim Cln,0 )(dim Cl0,2 ) = dim(Cln,0 ⊗ Cl0,2 ),

the Dimension Theorem from linear algebra tells us that fb is also injective. Thus
fb is the desired isomorphism.
The second isomorphism is proved in exactly the same way, and we avoid
repeating ourselves.
The proof of the third isomorphism is essentially the same as the two rst. Let
{e1 , . . . , ep+1 , ε1 , . . . , εq+1 } be an orthogonal basis for Rp+q+2 (i.e. ϕ(v, w) = 0
when v 6= w ) with the quadratic form Φp+1,q+1 , such that Φp+1,q+1 (ei ) = 1,
Φp+1,q+1 (εi ) = −1, and let {e01 , . . . , e0p , ε01 , . . . , ε0q } and {e001 , ε001 } be similar bases
p+q 1+1
for R and R (and thereby generators for the Cliord algebras Clp+1,q+1 ,
Clp,q and Cl1,1 respectively). We now dene a linear map f : Rp+q+2 −→
Clp,q ⊗ Cl1,1 by
( (
e0i ⊗ e001 ε001 if 1 ≤ i ≤ p ε0j ⊗ e001 ε001 if 1 ≤ j ≤ q
f (ei ) = , f (ε j ) = .
1 ⊗ e001 if i = p + 1 1 ⊗ ε001 if j = q + 1

Just like before it can be shown that

f (x)2 = Φp+1,q+1 (x) · 1 ⊗ 1,

∼
and thus f induces an isomorphism fb : Clp+1,q+1 −−→ Clp,q ⊗ Cl1,1 .

Now we are ready to state and prove the Cartan-Bott Theorem:

Theorem 1.13 (Cartan-Bott I). We have the following isomorphisms :

Cl0,n+8 ∼
= Cl0,n ⊗R(16) and Cln+8,0 ∼
= Cln,0 ⊗R(16).
1.2 Classication of Cliord Algebras 13

Proof. Using the two rst isomorphisms from Lemma 1.12 a couple of times
yields

Cl0,n+8 ∼
= Cln+6,0 ⊗ Cl0,2 ∼
= ··· ∼= Cl0,n ⊗ Cl2,0 ⊗ Cl0,2 ⊗ Cl2,0 ⊗ Cl0,2
∼
= Cl0,n ⊗ Cl2,0 ⊗ Cl2,0 ⊗ Cl0,2 ⊗ Cl0,2
where the last isomorphism follows from the fact that for arbitrary real algebras
A and B we have A ⊗ B ∼ = B ⊗ A. From (1.4) we have Cl0,2 ∼
= H, and from (1.5)
that Cl2,0 = R(2). Thus, using H ⊗R H ∼
∼ = R(4) from Lemma 1.11, we get
Cl2,0 ⊗ Cl2,0 ⊗ Cl0,2 ⊗ Cl0,2 ∼
= R(2) ⊗ R(2) ⊗ (H ⊗R H) ∼
= R(4) ⊗ R(4)
∼
= R(16).
This completes the proof of the rst isomorphism. The proof of the second
isomorphism is identical to this.

Now it's evident that once we know the Cliord algebras Cl0,0 , Cl0,1 , . . . , Cl0,7
and Cl0,0 , Cl1,0 , . . . , Cl7,0 we know all of them: Consider Clp,q and assume that
p ≥ q . Then by the third isomorphism in Lemma 1.12 we have that Clp,q is iso-
⊗q ∼
morphic to Clp−q,0 ⊗(Cl1,1 ) = Clp−q,0 ⊗R(2q ), and Clp−q,0 can be expressed
as a tensor product of Clk,0 (with 0 ≤ k ≤ 7) and some copies of R(16). We can
do the same if q ≥ p. Using the isomorphisms from Lemma 1.11 one obtains
the table of Cliord algebras in Appendix A. From this table we see that any
Cliord algebra is either a matrix algebra or a sum of two such algebras, as we
pointed out in the beginning of this section.

Example 1.14. As an example, let us show that Cl2,11 ∼

= C(64):
Cl2,11 ∼
= Cl0,9 ⊗ Cl1,1 ⊗ Cl1,1 ∼
= Cl0,1 ⊗R(16) ⊗ R(2) ⊗ R(2)
∼ C ⊗R R(64) ∼
= = C(64) .
Exercise 3. Show that Cl3,2 ∼
= R(4) ⊕ R(4).
By repeating the arguments of Example 1.14 in a more general setting we
obtain:

Corollary 1.15. Clp,q is a direct sum of two matrix algebras exactly when
p−q ≡1 (mod 4) .
We conclude this treatment by proving

Proposition 1.16. We have the following algebra isomorphism Clp,q ∼

= Cl0p,q+1 .
Proof. Let {e1 , . . . , ep+q+1 } denote an orthogonal basis for Rp+q+1 which sat-
ises Φ(ei ) = 1 for i = 1, . . . , p and Φ(ej ) = −1 for j = p + 1, . . . , p + q + 1.
p+q
Assume the basis has been chosen so that {e1 , . . . , ep+q } is a basis for R .
p+q 0
Now dene a linear map f : R −→ Clp,q+1 by
f (ei ) = ep+q+1 ei
fori ≤ p+q . Like in the proof of Lemma 1.12 one checks that f satises f (x)2 =
Φ(x) · 1 and thus factorizes to an algebra homomorphism fb : Clp,q −→ Cl0p,q+1 .
By inspection, this is the desired isomorphism.

For the rest of this section we will consider complex Cliord algebras. It turns
out that complex Cliord algebras behave even nicer than their real counter-
parts. As we have already seen a complex Cliord algebra associated with a
non-degenerate bilinear form is isomorphic to Cl(Φn ). The fact that there is
only one index on Φ and not two as in the real case, indicates some sort of
simplication.
But rst we introduce a way of turning real vector spaces/algebras into com-
plex ones:
14 Chapter 1  Cliord Algebras

Denition 1.17. By the complexication of a real vector space V we mean the

real tensor product V C := V ⊗R C. V carries a bilinear
If If the vector space
form φ, then the complexication of the bilinear form is ϕ (v ⊗ λ, v ⊗ λ ) :=
C 0 0

λλ ϕ(v, v ). The complexication of the corresponding quadratic form Φ is then

0 0

ΦC (v ⊗ λ) = λ2 Φ(v).
In the same way we dene the complexication of a real algebra A by A :=
C

A ⊗R C carrying the product

(a ⊗ λ)(a0 ⊗ λ0 ) = (aa0 ) ⊗ (λλ0 ).
Assume (V, ϕ) to be a real vector space with ϕ a non-degenerate bilinear form.
Then ϕC is a non-degenerate bilinear form on V C for assume v0 ⊗ λ0 to satisfy
ϕC (v0 ⊗ λ0 , v ⊗ λ) = λ0 λϕ(v0 , v) 6= 0 for all v ⊗ λ. Then λ0 6= 0 and ϕ(v0 , v) 6= 0
for all v and by non-degeneracy of ϕ this implies that v0 6= 0. Thus v0 ⊗ λ0 6= 0.
C C
In particular, if p + q = n we have that ϕp,q is equivalent to ϕn and thus Φp,q
is equivalent to Φn .
Now, one can pose the question: is the complexication of a real Cliord
algebra a complex Cliord algebra? By the following lemma the answer is yes.

Lemma 1.18. For p + q = n we have Cl(ΦCp,q ) ∼

= Cl(Φn ) ∼ 0,n .
= ClC
Proof. The rst isomorphism is due to the fact that the complexication of
Φp,q is equivalent to Φn and thus the corresponding Cliord algebras are iso-
morphic.
ϕ : Cn −→ ClC
To verify the second isomorphism we construct a linear map 0,n
∼
b : Cl(Φn ) −−→ ClC
and show that it factorizes to an isomorphism ϕ 0,n . At rst,
we remark that C ∼
n
= Rn ⊗R C. We then dene ϕ by ϕ(v ⊗ z) = i(v) ⊗ z where
n
i : R −→ Cl0,n denotes the usual embedding. Since
ϕ(v ⊗ z)2 = (i(v) ⊗ z)2 = i(v)2 ⊗ z 2
= Φ0,n (v)z 2 · 1 ⊗ 1 = ΦC
n (v ⊗ z) · 1 ⊗ 1,

ϕ factorizes uniquely to an algebra homomorphism ϕb : Cl(Φn ) −→ ClC 0,n . Both

n
algebras have complex dimension 2 so it's enough to show that ϕ b is surjective.
But ϕ b is surjective since ϕb is an algebra homomorphism and ϕ maps onto a
C
set of generators of Cl0,n . Namely, the set of elements of the form i(v) generate
Cl0,n , and 1 ∈ C generates C.

Thus, henceforth we will stick to the notation ClC

n for the complex Cliord
n
algebra over C equipped with any non-degenerate quadratic form, since the
preceding lemma guarantees that they are all isomorphic.
This result, in combination with the classication results for real Cliord
algebras, we get the complex version of the Cartan-Bott Theorem:

Theorem 1.19 (Cartan-Bott II). We have the following 2- periodicity : ClCn+2 ∼
=
n ⊗C Cl2 ,
ClC and furthermore that ClC2 ∼
= C(2).
C

Proof. Invoking Lemma 1.12 and Lemma 1.18 we obtain the following chain
of isomorphisms:

ClC ∼ ∼
n+2 = Cl0,n+2 ⊗R C = Cln,0 ⊗R C ⊗R Cl0,2
∼
= Cln,0 ⊗R (C ⊗C C) ⊗R Cl0,2 ∼
= (Cln,0 ⊗R C) ⊗C (C ⊗R Cl0,2 )
∼
= ClC ⊗C ClC .
n 2

For the second isomorphism, just recall that Cl0,2 ∼

=H and H ⊗R C ∼
= C(2).
Remembering that Cl1,0 ∼
=R⊕R so that ClC ∼
1 = C ⊕ C, we obtain:

Corollary 1.20. If n = 2k, then ClC

n = C(2 ).
∼ k
If n = 2k + 1, then ClCn ∼
=
C(2 ) ⊕ C(2 ).
k k
1.3 Representation Theory 15

1.3 Representation Theory

In this section we will turn our attention to the representation theory of Cliord
algebras. For the following denition, let us denote by K either R, C or H:

Denition 1.21. Let A be an algebra over K and V a vector space over K.

A K-representation of A is an algebra homomorphism ρ : A −→ EndK (V ). A
subspace U of V is called invariant under ρ if ρ(x)U ⊆ U for all x ∈ A. The
representation ρ is called irreducible if the only invariant subspaces are {0} and
V.

By an intertwiner of two representations ρ and ρ0 of A on V and V0 we

0
understand a linear map f : V −→ V satisfying

ρ0 (x) ◦ f = f ◦ ρ(x)

for all x ∈ A. Two representations are called equivalent if there exists an inter-
twiner between them which is also an isomorphism of vector spaces.
Just as we had complexication of a real algebra, we can complexify complex
representations: If ρ : A −→ End(V ) is a representation of a real algebra on a
complex vector space V , we dene the complexication ρC : AC −→ End(V ) by
ρC (x⊗λ) = λρ(v). The notions of invariant subspaces and irreducibility of a rep-
resentation and its complexied are closely related as the following proposition
shows

Proposition 1.22. Let A be a real unital algebra, and ρ an algebra representa-

tions on the complex V . Let AC and ρC denote the associated complexications.
Then the following hold :
1) A subspace W ⊆ V is ρ-invariant if and only if it is ρC -invariant.
2) ρ is irreducible if and only if ρC is irreducible.
Proof. 1) If W is ρ-invariant then

ρC (x ⊗ λ)W = λρ(x)W ⊆ W.

Conversely, if W is ρC -invariant then for x ∈ A we have ρ(x)W = ρC (x ⊗ 1)W ⊆

W. This proves 1).
2) Follows immediately from 1).

A representation ρ of A on V gives V the structure of a left A-module simply

by dening a · v := ρ(a)v . This is compatible with addition in V.
The next proposition actually contains all the information we need to de-
termine all the irreducible representations (up to equivalence) of the Cliord
algebras:

Proposition 1.23. The matrix algebra K(n) has only one irreducible K-repre-
sentation, namely the dening representation i.e. the natural isomorphism πn :
K(n) −−→ EndK (Kn ). The algebra K(n) ⊕ K(n) has exactly 2 inequivalent irre-
∼

ducible K-representations, namely :

πn0 (x1 , x2 ) := πn (x1 ) and πn1 (x1 , x2 ) := πn (x2 ). (1.6)

We saw in the previous section that Clp,q is of the form K(n) ⊕ K(n) i
p−q ≡1 (mod 4) and a matrix algebra otherwise. This observation along side
with the preceding proposition yields the number of irreducible representations
of the real Cliord algebras.
16 Chapter 1  Cliord Algebras

But there is a slight problem here, in that the real Cliord algebra is not
always a real matrix algebra or a sum of real matrix algebras. Thus the irre-
ducible representations of Proposition 1.23 need not be real! For instance Cl1,4 ∼
=
H(2)⊕H(2), and Proposition 1.23 gives us two irreducible H-representations over
H2 ! But fortunately we can always turn a complex or quaternionic representa-
tion into a real representation if we just remember to adjust the dimension.
2
Without going further into details with keeping track of the dimensions we have
shown:

Theorem 1.24. Consider the real Cliord algebra Clp,q . If p − q ≡ 1 (mod

4) has up to equivalence two real irreducible representations and up to
Clp,q
equivalence exactly one real irreducible representation otherwise.
n ≡ 1 (mod 4) Cl0,n has two irreducible representations, ρ0n
In particular if
1 0
and ρn , and otherwise only one ρn . If n ≡ 1 (mod 4) we dene ρn := ρn such that
to each real Cliord algebra Cl0,n we associate a real irreducible representation
ρn called the real spin representation . The elements of the corresponding vector
spaces are called spinors .
The similar situation for complex Cliord algebras is simpler due to the fact
that each complex Cliord algebra decomposes into complex matrix algebras
(cf. Corollary 1.20). Thus all irreducible representations are complex. This will
make it a lot easier to keep track of the dimensions.

Theorem 1.25. Consider the complex Cliord algebra ClCn . If n = 2k we have

up to equivalence ) exactly one irreducible complex representation κn on C2 ,
k
(

namely the isomorphism ClC2k −− ∼

→ End(C2 ).
k

If n = 2k + 1 we have (up to equivalence ) exactly two irreducible representa-

tions κ0n and κ1n on C2 .
k

k
For n = 2k or n = 2k + 1 like above, dene ∆n := C2 . The elements of ∆n
are called Dirac spinors or complex n-spinors . The irreducible representations
of ClC
n are representations on ∆n .
In the case where n is odd we want to single out κ0n and dene κn := κ0n
which is just the composition
∼ π
κn = κ0n : ClC
n −−→ EndC (∆n ) ⊕ EndC (∆n ) −
−−1
→ EndC (∆n )
of the isomorphism with the projection π1 onto the rst component. Hence, for
each n we have an irreducible complex representation on ∆n called the complex
spin representation .
Finally, let's try to break up the action of the Cliord algebra into smaller
pieces and see how they act on the spinors. This requires introduction of the
so-called volume element. It is well-known that, has a unique volume
Λ∗ (Rn )
element Ω given unambiguously, in any orthonormal basis {e1 , . . . , en }, by Ω =
e1 ∧ · · · ∧ en . Applying the quantization map to this yields an element ω :=
Q(Ω) ∈ Cl0,n , also called the volume element , given by ω = e1 · · · en . For the
C
complex Cliord algebra Cln we dene the volume element by
n+1
ωC := ib 2 c ω.
2 k
In the case n = 2k we note that ω = (−1) and that ω commutes with every
element of Cl00,2k , while ω anti-commutes Cl10,2k , for instance we see that

ωe1 = (e1 · · · e2k )e1 = (−1)2k−1 e21 e2 · · · e2k = −e1 ω.

2 Cn is naturally isomorphic to R2n via the isomorphism ϕ : Cn −→ R2n , (λ , . . . , λ ) 7−→
1 n
(Re λ1 , Im λ1 , . . . , Re λn , Im λn ). If π is a complex representation of an algebra A, then we
just dene a real representation on R2n by πe(x)(v) = ϕ(π(x)ϕ−1 (v)). It's easy to see that
e is irreducible i π is. Likewise with a quaternionic representation we exploit the natural
π
isomorphism Hn ∼ = R4n .
1.3 Representation Theory 17

If n = 2k + 1 then ω commutes with everything.

We are only interested in the even case, so assume n = 2k , dene ωC := ik
and consider the map

f := ik κ2k (ω) : ∆2k −→ ∆2k .

From this we see that f commutes with κ2k (ξ):

f (κ2k (ξ)ψ) = ik κ2k (ω)(κ2k (ξ)ψ) = ik κ2k (ωξ)ψ = ik κ2k (ξω)ψ

= κ2k (ξ)ik κ2k (ω)ψ = κ2k (ξ)f (ψ)

for ξ ∈ Cl00,2k and ψ ∈ ∆2k . Furthermore f is an involution:

f ◦ f = i2k κ2k (ω 2 ) = (−1)k κ2k ((−1)k ) = (−1)2k id∆2k = id∆2k .

Then it is well-known that f has the eigenvalues ±1 and corresponding eigenspaces

∆±
2k (of equal dimension) such that ∆2k ∼
= ∆+ − ±
2k ⊕ ∆2k . The elements of ∆2k are
called positive and negative Weyl spinors or even and odd chiral spinors, respec-
tively.
We can use the map f to induce another splitting of the complexied Clif-
ford algebra (in even dimension!), for left multiplication by ωC on ClC
2k is an
involution, hence the algebra splits into eigenspaces

ClC C C
2k = (Cl2k )+ ⊕ (Cl2k )−

where, in fact
1
(ClC C
2k )± = 2 (1 ± ωC ) Cl2k .

We can combine this splitting with the splitting given by the involution α to
obtain the spaces

0 1 C 0 1 1 C 1
(ClC
2k )± := 2 (1 ± ωC )(Cl2k ) and (ClC
2k )± := 2 (1 ± ωC )(Cl2k ) .

End(∆+
For the next proposition we will identify 2k ) as the subspace of End(∆2k )
+ −
consisting of maps ∆2k −→ ∆2k which map ∆2k to itself and which map ∆2k to
− ± ∓
0, and in a similar way we identify End(∆2k ) and Hom(∆2k , ∆2k ) as subspaces
of End(∆2k ).

Proposition 1.26. The spin representation κ2k : ClC2k −−

∼
→ End(∆2k ) restricts
to the following isomorphisms
0 ∼ + 0 ∼ −
(ClC
2k )+ = End(∆2k ), (ClC
2k )− = End(∆2k )

(ClC )1 =∼ Hom(∆− , ∆+ ), (ClC )1 =∼ Hom(∆+ , ∆− ).

2k + 2k 2k 2k − 2k 2k

Proof. First assume ψ ∈ ∆+

2k , i.e. f (ψ) = ψ , then for ξ ∈ (ClC 0
2k )+ :

f (κ2k (ξ)ψ) = κ2k (ωC ξ)ψ = κ2k (ξωC )ψ

= κ2k (ξ)f (ψ) = κ2k (ξ)ψ,

i.e.κ2k (ξ)ψ ∈ ∆+ C 0
2k . For this result we only used that (ξ ∈ Cl2k ) , but to show
− 1 −
that κ2k (ξ) is 0 on ∆2k we need that ξ = (1 + ωC )ξ . Let ψ ∈ ∆2k , i.e. κ2k (1 +
2
ωC )ψ = 0, then

κ2k (ξ)ψ = 12 κ2k ((1 + ωC )ξ)ψ = 12 κ2k (ξ(1 + ωC )) = 0.

This shows that κ2k maps (ClC 0

2k )+ into End(∆+
2k ). The reasoning is the same in
the other 4 cases. But then, since κ2k is an isomorphism, the restricted maps
must be isomorphisms as well, and this proves the proposition.
18 Chapter 1  Cliord Algebras
Chapter 2

Spin Groups
2.1 The Cliord Group
Having introduced the Cliord algebra Cl(Φ), we proceed to dene its Cliord
group Γ(Φ). The point of doing this is that the Cliord group has two particu-
larly interesting subgroups, the pin and spin groups.
Let V be a nite-dimensional real vector space, and let Φ be a quadratic form
on V.

Denition 2.1. Let Cl∗ (Φ) denote the multiplicative group of invertible ele-
ments of Cl(Φ). The Cliord group (by some also called the Lipschitz group ) of
Cl(Φ) is the group

Γ(Φ) := {x ∈ Cl∗ (Φ) | α(x)vx−1 ∈ V for all v ∈ V }.

One mechanically veries that Γ(Φ) is truly a group. The group Cl∗ (Φ) is
an open subgroup of Cl(Φ), just as Aut(V ) is an open subgroup of End(V ) (at
least when V is nite-dimensional). In the latter case the Lie algebra of Aut(V )
is just End(V ) with the commutator bracket. In the same way the Lie algebra
cl∗ (Φ) of the group Cl∗ (Φ) is just Cl(Φ) with the commutator bracket.
It is very conspicuous from the denition that we are interested in a particular
representation Λ : Γ(Φ) −→ Aut(V ), namely Γ(Φ) 3 x 7−→ Λx where Λx : V −→
V is given by Λx (v) = α(x)vx−1 and called the twisted adjoint representation
(indeed, the form of Λx is reminiscent of the adjoint representation of a Lie
group).
One reason for considering Λx (v) = α(x)vx−1 instead of Adx (v) = xvx−1
is that the twisted adjoint representation keeps track of an otherwise annoying
sign. W.r.t. the bilinear form ϕ we dene, for x ∈ V with Φ(x) 6= 0 the reection
sx through the hyperplane orthogonal to x by

ϕ(v, x)
sx (v) := v − 2 x.
Φ(x)

We then have the following geometric interpretation of Λx which in addition to

being a pretty fact, is crucial in the proof of Lemma 2.10.

Proposition 2.2. For any x ∈ V with Φ(x) 6= 0 we have x ∈ Γ(Φ), and the
map Λx : V −→ V given by Λx (v) = α(x)vx−1 is the reection through the
hyperplane orthogonal to x.
Proof. First, since x2 = Φ(x) · 1 6= 0, x is invertible in Cl(Φ) with inverse

19
20 Chapter 2  Spin Groups

x−1 = 1
Φ(x) x ∈V. Using this and Eq. (1.2) we see that

Φ(x)α(x)vx−1 = −xv(Φ(x)x−1 ) = xvx

= −2ϕ(v, x)x + Φ(x)v ∈ V

i.e. we have x ∈ Γ(Φ) and

ϕ(v, x)
Λx v = v − 2 x = sx (v) ∈ V.
Φ(x)

The following proposition is extremely useful:

Proposition 2.3. Let ϕ be a non-degenerate bilinear form on V. Then for the

twisted adjoint representation ker Λ = R∗ · 1 (where R∗ = R \ {0}).
Proof. ϕ is non-degenerate, we can choose an orthonormal basis e1 , . . . , en
Since
for V Φ(ei ) = ±1 and ϕ(ei , ej ) = 0 when i 6= j. Let x ∈ ker Λ; this
such that
means α(x)v = vx for all v ∈ V. Because of the Z2 -grading of Cl(Φ) we can
0 1
write x = x0 + x1 where x0 and x1 belong to Cl (Φ) and Cl (Φ), respectively.
This gives us the equations

vx0 = x0 v (2.1)

−vx1 = x1 v. (2.2)

The terms x0 and x1 e1 , . . . , en . By successively

can be written as polynomials in
applying the identity ei ej + ej ei = 2ϕ(ei , ej ) we can express x0 in the form
x0 = a0 + e1 a1 where a0 and a1 are both polynomial expressions in e2 , . . . , en .
0 1
Applying α to this equality shows that a0 ∈ Cl (Φ) and a1 ∈ Cl (Φ). Setting
v = e1 in Eq. (2.1) we get

e1 a0 + e21 a1 = a0 e1 + e1 a1 e1 = e1 a0 − e21 a1

where the last equality follows from a0 ∈ Cl0 (Φ) and a1 ∈ Cl1 (Φ). We deduce
2
0 = e1 a1 = Φ(e1 )a1 ; since Φ(e1 ) 6= 0, we have a1 = 0. So, the polynomial
expression for x0 does not contain e1 . Proceeding inductively, we realize that x0
does not contain any of the terms e1 , . . . , en and so must have the form x0 = t · 1
where t ∈ R.
We can apply an analogous argument to x1 to conclude that neither does
the polynomial expression for x1 contain any of the terms e1 , . . . , en . However,
x1 ∈ Cl1 (Φ), so x1 = 0.
Thus, x = x0 + x1 = t · 1. Since x 6= 0, we must have t ∈ R∗ . This shows
ker Λ ⊆ R∗ · 1; the reverse inclusion is obvious.
The assumption that Φ is non-degenerate is not redundant. Consider a real
vector space V dim V ≥ 2. If Φ ≡ 0, then Cl(V, Φ) = Λ∗ V, the exterior
with
algebra of V. Consider the element x = 1 + e1 e2 . Clearly, x−1 = 1 − e1 e2 , and
we have

α(1 + e1 e2 ) v (1 + e1 e2 )−1 = (1 + e1 e2 ) v (1 − e1 e2 ) = v,

i.e. 1 + e1 e2 ∈ ker Λ, yet 1 + e1 e2 is not a scalar multiple of 1. Thus, since

the following propositions use Proposition 2.3 in their proofs we will, from this
point on, always assume Φ to be non-degenerate. This is not a severe restriction
since practically all interesting Cliord algebras originate from non-degenerate
bilinear forms.

We now introduce the important notions of conjugation and norm.

2.1 The Cliord Group 21

Denition 2.4. For any x ∈ Cl(Φ), the conjugate of x is dened as x := t(α(x)).

Moreover, the norm of x is dened as N (x) := xx.

Note that t ◦ α = α ◦ t (it clearly holds on ei1 · · · eir , and by linearity on

any element of Cl(Φ)). Also note that x = x. The term norm is justied in the
following lemma, displaying some elementary properties of the norm

Lemma 2.5. When Φ is non-degenerate, the norm possesses the following prop-
erties :
1) If v ∈ V then N (v) = Φ(v) · 1, i.e. N is an extension of Φ to the algebra.

2) If x ∈ Γ(Φ), then N (x) ∈ R∗ · 1.

When restricted to Γ(Φ), the norm N : Γ(Φ) −→ R∗ ·1 is a homomorphism.
3)
Moreover, N (α(x)) = N (x).
Proof. 1) This is just a simple calculation
N (v) = v(t ◦ α(v)) = −v 2 = −Φ(v) · 1.
2) According to Proposition 2.3, it's enough to show that N (x) ∈ ker Λ. By
denition of the Cliord group, x ∈ Γ(Φ) implies

α(x)vx−1 ∈ V for all v ∈ V.

As t|V = idV , we thus have

α(x)vx−1 = t(α(x)vx−1 ) = t(x−1 )vt(α(x)).

Isolating the v on the right-hand-side we get, using t ◦ α = α ◦ t,
v = t(x)α(x)v(t(α(x))x)−1 = α(xx)v(xx)−1 ,
i.e. xx ∈ ker Λ. But then xx = x x ∈ ker Λ.
3) Two simple calculations:

N (xy) = xyy x = xN (y)x = xxN (y) = N (x)N (y)

and, since α(x) = α(x), (following from t ◦ α = α ◦ t)
N (α(x)) = α(x)α(x) = α(xx) = α(N (x)) = N (x).
Example 2.6. Let's calculate the conjugate and the norm on the two model
Cliord algebras. First Cl0,1 ∼
= C. Recall thatR sits inside C as the imaginary
line and that α is the involution satisfying α(1) = 1 and α(λ) = −λ for λ ∈ R.
Thus α is just conjugation: α(z) = z . Since t is just the identity on Cl0,1 it
follows that conjugation in the Cliord sense is just usual conjugation

t(α(z)) = z.
Therefore the norm becomes N (z) = zz = |z|2 , i.e. the square of the usual norm.
Exercise 4. Carry out a similar calculation for Cl0,2 ∼ = H.
For the next proposition recall the denition of the orthogonal group O(Φ)
(when Φ is non-degenerate!) as the endomorphisms f : V −→ V satisfying
Φ(f (v)) = Φ(v).
Proposition 2.7. For any x ∈ Γ(Φ), the map Λx is an orthogonal transforma-
tion of V. That is, Λ(Γ(Φ)) ⊆ O(Φ).
Proof. Let x ∈ Γ(Φ) and use the fact that N is a homomorphism:

N (Λx v) = N (α(x)vx−1 ) = N (α(x))N (v)N (x−1 ) = N (x)N (v)N (x)−1 = N (v),

Since N (v) = −Φ(v) · 1, this shows thatΛx is Φ-preserving. Along with the
linearity, this exactly shows that Λx ∈ O(Φ).
22 Chapter 2  Spin Groups

2.2 Pin and Spin Groups

Denition 2.8. The pin group is dened as

Pin(Φ) := {x ∈ Γ(Φ) | N (x) = ±1}.

The spin group consists of those elements of Pin(Φ) that are linear combinations
of even-degree elements:

Spin(Φ) := Pin(Φ) ∩ Cl0 (Φ).

It's not too dicult to verify that Pin(Φ) and Spin(Φ) really are groups. We
will write Pin(p, q) Spin(p, q) for the pin and spin groups associated with
and
Clp,q and likewise Pin(n) := Pin(0, n) and Spin(n) := Spin(0, n) (not to be
confused with the complex pin and spin groups Pin(Φn ) and Spin(Φn ) sitting
C
inside Cln !). Recall that the algebra structure in Cl0,n is that generated by the
2
relations v · v = −kvk · 1.
Since Φ is assumed to be non-degenerate, any real Cliord algebra is (up to
isomorphism) of the form Clp,q and thus we can in fact always assume the real
pin and spin groups to be of the form Pin(p, q) and Spin(p, q).
In addition to the complex pin and spin groups there are also complexica-
tions of the real pin/spin groups, namely let(V, Φ) be a real quadratic vector
space and dene Pinc (Φ) ⊆ Cl(Φ)C = Cl(Φ)⊗C to be the subgroup of invertible
c
elements in Cl(Φ) ⊗ C generated by Pin(Φ) ⊗ 1 and 1 ⊗ U(1). Similarly, Spin (Φ)
is dened as the subgroup generated by Spin(Φ) ⊗ 1 and 1 ⊗ U(1). These are
not to be confused with Spin(ΦC )!
Proposition 2.9. There is a Lie group isomorphism
∼
Spinc (Φ) −−→ Spin(Φ) ×±1 U(1)

i.e. the Spinc (Φ) is quotient of Spin(Φ) × U(1) where we identify (1, 1) and
(−1, −1). In particular Spinc (Φ) is connected if Spin(Φ) is connected.

Proof. We consider the smooth map Spin(Φ) × U(1) −→ Spinc (Φ) given by
(g, z) 7−→ gz . This is a surjective Lie group homomorphism. The kernel consists
−1
of elements (g, z) such that gz = 1 i.e. g = z ∈ U(1). But these elements are
rare, there are only ±1. Thus the kernel equals {±(1, 1)}, and the map above
descends to an isomorphism.

The group Spinc (Φ) is important in the study of complex manifolds.

We have the following important lemma, to be used in the proof of Theorem
2.16.

Lemma 2.10. The map Λ : Pin(Φ) −→ O(Φ) is a surjective homomorphism

with kernel {−1, 1}. Similarly, the map Λ : Spin(Φ) −→ SO(Φ) is a surjective
homomorphism with kernel {−1, 1}.
Proof. First, Proposition 2.7 guarantees Λ(Pin(Φ)) ⊆ O(Φ). It is trivial to
verify that Λ is a homomorphism. To prove that Λ is surjective we use the
1
Cartan-Dieudonne Theorem which states that any element T of O(Φ) can be
written as the composition T = s1 ◦· · ·◦sp of reections where sj is the reection
through the hyperplane orthogonal to some vector uj ∈ V But according to
uj
Proposition 2.2 sj = Λuj . Since N (uj ) = −Φ(uj ) · 1, replacing uj by √
|Φ(uj )|
(this doesn't change the reection) leaves N (uj ) = −1 and thus N (u1 · · · up ) =
1 See for instance [Gallier] ([5]), Theorem 7.2.1.
2.2 Pin and Spin Groups 23

±1, i.e. u1 · · · up ∈ Pin(Φ). Hence T = Λu1 · · · Λup = Λu1 ···up which proves that
Λ|Pin(Φ) is surjective. The kernel is easily calculated:

ker Λ|Pin(Φ) = ker Λ ∩ Pin(Φ) = {t ∈ R∗ · 1 | N (t) = ±1} = {−1, 1}.

To prove the analogous statement for Spin(Φ) we need rst to show that Λ maps
Spin(Φ) SO(Φ). Assume, for contradiction, that this is not the case, i.e. that
to
an element f ∈ O(Φ) \ SO(Φ) exists such that Λx = f for some x ∈ Spin(Φ). By
the Cartan-Dieudonne Theorem f can be written as an odd number of reections
f = s1 ◦ · · · ◦ s2k+1 , and to each such reection sj corresponds a vector uj so
that sj = Λuj . In other words we have

Λx = Λu1 ···u2k+1 or id = Λx−1 u1 ···u2k+1 .

By Proposition 2.3 x−1 u1 · · · u2k+1 = λ·1 for some λ ∈ R∗ , i.e. x = λ−1 u1 · · · u2k+1
1
and this is a contradiction, since u1 · · · u2k+1 ∈ Cl (Φ). Thus Λ maps Spin(Φ)
to SO(Φ) and since {±1} ⊆ Spin(Φ) the kernel is still Z2 .

As a consequence of the proof of this lemma we note, that the set {v ∈

V | Φ(v) = 1} of unit vectors generate Pin(Φ) as a subgroup of Γ(Φ), i.e. any
element of Pin(Φ) can be written as a product of unit vectors in V . Similarly
elements of Spin(Φ) can be written as a product of an even number of unit
vectors in V .
Let's warm up by doing a few simple examples

Example 2.11. Let's calculate Pin(1) and Spin(1). They are subgroups of
Cl0,1 ∼
= C and the vector space, from which the Cliord algebra originates, is
R which sits inside C as the imaginary line (cf. Example 1.5). In R we have
just two unit vectors, namely ±1. They sit in C as ±i and they generate Pin(1)
which are thus seen to be isomorphic to Z4 (the fourth roots of unity). Spin(1)
is then generated by products of two such unit vectors, i.e. Spin(1) = Z2 .

Exercise 5. Calculate Pin(2) and Spin(2).

At this point, it is probably not very clear why there should be anything
particularly interesting about pin and spin groups. The explanation is that they
are double coverings of O(Φ) and SO(Φ) (Theorem 2.16). When n ≥ 3 then
Spin(n) is even the universal double covering of SO(n) (Corollary 2.20). The
following more tedious example can be thought of as a special case of this fact:
it shows that Spin(3) is isomorphic to SU(2) which is known to be the universal
double covering of SO(3).
Example 2.12. Calculation of Spin(3). Choose an orthonormal basis e1 , e2 , e3
of R3 . By Proposition 1.3, the list of elements

1, e1 , e2 , e3 , e1 e2 , e1 e3 , e2 e3 , e1 e2 e3

forms a basis for the Cliord algebra Cl0,3 . The group Spin(3) can be written

Spin(3) = {x ∈ Cl00,3 | ∀v ∈ V : α(x)vx−1 ∈ V and N (x) = 1},

as N (x) = 1 x ∈ Cl∗0,3 (explicitly, x−1 = x). The rst thing we want

implies
0
to show is that x ∈ Spin(3) if and only if x ∈ Cl0,3 and N (x) = 1. The only
0
non-trivial statement to be proven is that the conditions x ∈ Cl0,3 and N (x) = 1
−1
imply that α(x)vx ∈ V for any v ∈ V.
0 1
Since x ∈ Cl0,3 and v ∈ Cl0,3 , we have xvx
−1
∈ Cl10,3 . Thereby,

xvx−1 = u + λe1 e2 e3
24 Chapter 2  Spin Groups

with u ∈ V and λ ∈ R. Moreover, observe that v = −v for all v ∈ V, and

x = x−1 since N (x) = 1, so

xvx−1 = x−1 v x = −xvx−1 .

But e1 e2 e3 = −e3 e2 e1 = e1 e2 e3 , so

−u + λe1 e2 e3 = xvx−1 = −xvx−1 = −u − λe1 e2 e3 ;

hence λ = 0, xvx−1 ∈ V. This is equivalent to α(x)vx−1 ∈ V, as desired.

and so
0
Knowing that x ∈ Spin(3) i x ∈ Cl0,3 and N (x) = 1, we can characterize
0
the elements x of Spin(3) in a very handy way. Namely, since x ∈ Cl3,0 , x must
have the form
x = a1 + be1 e2 + ce1 e3 + de2 e3
where a2 + b2 + c2 + d2 = N (x) = 1. Thus, Spin(3) consists of all elements x of
2 2 2 2
the form x = a1 + be1 e2 + ce1 e3 + de2 e3 with a + b + c + d = 1. This allows
us to establish an isomorphism Spin(3) ∼ = SU(2) like follows:
     
i 0 0 −1 0 −i
e1 e2 7−→ , e1 e3 7−→ , e2 e3 7−→ ,
0 −i 1 0 −i 0
 
a + ib −c − id
i.e. x 7−→ . Thus, Spin(3) is isomorphic to SU(2). One
c − id a − ib
can use similar arguments to prove Spin(4) ∼ = SU(2) × SU(2) and Spin(3, 1)0 ∼ =
SL(2, C) (cf. [Lawson and Michelson], p. 56, Theorem 8.4).
Example 2.13. Let's calculate some of the spin
c
groups also. This is rather
easy, given Proposition 2.9 and the calculations above.
First we show that Spinc (3) ∼
= U(2). Dene the map

Spin(3) × U(1) ∼
= SU(2) × U(1) −→ U(2)

by (A, z) 7−→ zA. It is well-known that this map is a surjective Lie group
homomorphism, and it is easily seen that the kernel is (I2 , 1) (I2 is just the 2×2
identity matrix). Thus a Lie group isomorphism is induced on the quotient, i.e.
∼
an isomorphism Spinc (3) −−→ U(2).
We can argue almost similarly to calculate Spinc (4). For brevity, put

nA 0
 o
1
S(U(2) × U(2)) := | A1 , A2 ∈ U(2), det A1 det A2 = 1
0 A2

Dene a map

Spin(4) × U(1) ∼
= SU(2) × SU(2) × U(1) −→ S(U(2) × U(2))
 
zA1 0
(A1 , A2 , z) 7−→ .
0 zA2
This map is again a surjective Lie group homomorphism, as above, and the
∼
kernel is ±(I2 , I2 , 1), and thus it induces an isomorphism Spinc (4) −−→ S(U(2)×
U(2)).
Note that Pin(Φ) and Spin(Φ) are Lie groups. This is because the multiplica-
tive group Cl∗ (Φ) of invertible elements is an open subset of Cl(Φ) (this is a
general result for algebras) which is a nite-dimensional linear space, hence a
manifold. Thus, Cl∗ (Φ) is a manifold, and since multiplication and inversion are
∗
smooth maps, it is a Lie group. As Pin(Φ) is a closed subgroup of Cl (Φ) (since
N is continuous) and Spin(Φ) is a closed subgroup of Pin(Φ) (since Cl0 (Φ) is a
closed subspace of Cl(Φ)), they are Lie groups.
2.3 Double Coverings 25

2.3 Double Coverings

In this section we will prove that Pin(Φ) and Spin(Φ) are double coverings of
O(Φ) and SO(Φ), respectively. This will allow us to prove furthermore that
Spin(n) is the universal double covering of SO(n) which is our main result. We
rst recall the notion of a covering space in the general setting of two topological
spaces:

Denition 2.14. Let X be topological spaces. A covering map is a

Y and
continuous and surjective map p : Y −→ X with the property that for any
x ∈ X there is an open neighborhood U of x so that p−1 (U ) can be written
as the disjoint union of open subsets Vα (called the sheets ) with the restriction
p|Vα : Vα −→ U a homeomorphism. We say that U is evenly covered by p and
call Y a covering space of X. When all the bers p
−1
(x) have the same nite
cardinality n, we call p an n-covering. If Y is simply connected, the covering is
called a universal covering.
When X is pathwise connected, all bers p−1 (x) will have the same cardinal-
ity. If the covering is the universal covering this is precisely the cardinality of
the fundamental group π1 (X).
Quite often, covering maps between groups arise from group actions G×Y −→
Y. We now introduce the notion of an even action orcovering action as it
allows an elegant proof of Theorem 2.16. A group G is said to act evenly on
the topological space Y if each point y ∈ Y has a neighborhood U such that
g · U ∩ h · U = ∅ if g 6= h. As usual, Y /G denotes the orbit space under the action
of G, equipped with the quotient topology.

Lemma 2.15. Let G be a nite group acting evenly on a topological space Y .

Then the canonical map p : Y −→ Y /G is a |G|-covering.
Proof. p is obviously continuous and surjective. Let [y] = p(y) ∈ Y /G for
some y ∈ Y . We shall produce a neighborhood of [y] which is evenly covered
by p. As G is acting evenly there exists a neighborhood V ⊆ Y of y such that
g · V ∩ h · V = ∅ if g 6= h; dene U := p(V ). U is open, for we have
[
p−1 (U ) = g · V,
g∈G

where g · V are all open (as the map Y 3 x 7−→ g · x is a homeomorphism

for all g ∈ G). Consequently, p−1 (U ) is open and by denition of the quotient
topology, U is open. Thus, we have that U is a neighborhood of [y] and that
p−1 (U ) is a disjoint union of sets homeomorphic to V , and that the number of
sheets is equal to the order of the group G. The only thing we need to show is
that p|V : V −→ U is a homeomorphism. The map is obviously surjective. To
show injectivity assume that p(x) = p(y) (for x, y ∈ V ) that is there exists a
g ∈ G such that y = g · x. But then y ∈ g · V . From the fact that V ∩ g · V = ∅
if g 6= e, we deduce g = e and consequently x = y .
Continuity of p|V is obvious, since it's a restriction of a continuous map. We
now only need to show that p|V is a open map. But p is itself an open map, for
let O ⊆ Y be any open subset. By denition of the quotient topology, p(O) is
−1
open if and only if p (p(O)) is open. But
[
p−1 (p(O)) = g·O
g∈G

is open, being the union of open sets g·O. Being a restriction of an open map p to
an open set V , p|V is an open map, and p|V is therefore a homeomorphism.
26 Chapter 2  Spin Groups

Theorem 2.16. The map Λ : Pin(Φ) −→ O(Φ) is a double covering. Moreover,

Λ : Spin(Φ) −→ SO(Φ) is a double covering.
Proof. By Lemma 2.10, the homomorphism Λ : Pin(Φ) −→ O(Φ) is sur-
jective and has the kernel {1, −1} = Z2 . By standard results we have that
O(Φ) and Pin(Φ)/Z2 are isomorphic as Lie groups. We let Z2 act on Pin(Φ)
by multiplication which is obviously an even action. By the preceding Lemma
2.15, the quotient map Pin(Φ) −→ Pin(Φ)/Z2 is a double covering. Since
Λ : Pin(Φ) −→ O(Φ) can be identied with this map via the above isomorphism
of Lie groups, Λ is a double covering. The proof for Λ : Spin(Φ) −→ SO(Φ) is
completely analogous.

Corollary 2.17. The groups Pin(n) and Spin(n) are compact groups.
Proof. Since O(n) and SO(n) are compact and Pin(n) and Spin(n) are nite
coverings, the result follows from standard covering space theory.

This does not hold in general, for instance Spin(3, 1)0 = SL(2, C) which is
denitely not compact. If the identity component in non-compact, the entire
group must be non-compact.
∼
In Chapter 1 we constructed explicit isomorphisms Ψ : O(p, q) −−→ O(q, p)
∼
and Ψ : SO(p, q) −−→ SO(q, p). A natural question would be if these isomor-
phisms lift to isomorphisms on the level of pin and spin groups? For the pin
groups the answer is no, in general. But for the spin groups the answer is ar-
mative:

Proposition 2.18. There exists a Lie group isomorphism Ψ : Spin(p, q) −−

∼
→
Spin(q, p) such that the following diagram commutes
Spin(p, q)
∼ / Spin(q, p)
Ψ
Λ Λ
 
SO(p, q)
∼ / SO(q, p)
Ψ

Proof. Let ϕ be an anti-orthogonal linear map as mentioned in Chapter 1. By

the same sort of argument as in the proof of Proposition 1.4, it is seen that ϕ
extends to a map on the corresponding Cliord algebras. This is also denoted
ϕ. Dene Ψ : Spin(p, q) −→ Spin(q, p) by (recall that Spin(p, q) is generated by
an even number of unit vectors)

Ψ(v1 · · · v2k ) = (−1)k ϕ(v1 ) · · · ϕ(v2k )

where v1 , . . . , v2k ∈ Rp+q . This is easily seen to be a Lie group isomorphism.

Thus we only need to check commutativity of the diagram. First we observe

Ψ(Λvi )v = ϕ ◦ Λvi ◦ ϕ−1 (v) = ϕ(−vi ϕ−1 (v)vi−1 )

= −ϕ(vi )vϕ(vi−1 ) = −ϕ(vi )vϕ(vi )−1
= Λϕ(vi ) v.

Then we immediately get

Λ(Ψ(v1 · · · v2k )) = Λ((−1)k ϕ(v1 ) · · · ϕ(v2k ))

= Λϕ(v1 ) · · · Λϕ(v2k ) = Ψ(Λv1 ) · · · Ψ(Λv2k )
= Ψ(Λv1 · · · Λv2k ) = Ψ(Λ(v1 · · · v2k ))

Restricting attention to Spin(n) we can prove another delicious fact concern-

ing its topology:
2.3 Double Coverings 27

Theorem 2.19. Spin(n) is path-connected when n ≥ 2 and simply connected

when n ≥ 3.
Proof. We rst remark that Spin(n) is pathwise connected. Consider the el-
ement v1 · · · v2k vj ∈ Rn and Φ0,n (vj ) = 1 and note that each vj can
where
n−1
be connected to v1 by a continuous path running on the unit sphere S =
n n
{v ∈ R | Φ0,n (v) = 1} in R . Thus there is a continuous path from v1 · · · v2k to
v1 · · · v1 = 1 and thereby Spin(n) is path-connected.
Next we need to show, that the fundamental group at each point of Spin(n)
is the trivial group. Since Spin(n) is path-connected it suces to show this for
just a single point. So let x0 ∈ Spin(n). By standard covering space theory ρ∗ :
π1 (Spin(n), x0 ) −→ π1 (SO(n), ρ(x0 )) is injective, and the index of the subgroup
ρ∗ (π1 (Spin(n), x0 )) in π1 (SO(n), ρ(x0 )) is equal to the number of sheets of the
covering ρ, which we have just showed was 2. For n ≥ 3 the fundamental group
π1 (SO(n), ρ(x0 )) is Z2 , so as ρ∗ : π1 (Spin(n), x0 ) has index 2, it must be the
trivial subgroup. Since ρ∗ was injective also π1 (Spin(n), x0 ) is trivial, proving
that Spin(n) is simply connected.

Putting p=0 in Theorem 2.16 and combining with Theorem 2.19 we get the
main result:

Corollary 2.20. For n ≥ 3, the group Spin(n) is the universal (double ) cover-
ing of SO(n).
G
It's a classical fact from dierential geometry that for connected Lie groups
and H, F : G −→ H is a covering map, then the induced map F∗ : g −→ h is
if
2
an isomorphism. Spin(n) is connected and simply connected (Theorem 2.19),
and SO(n) is connected. By Theorem 2.16, the homomorphism Λ : Spin(n) −→
SO(n) is a covering map, and thus we have an isomorphism
∼
Λ∗ : spin(n) −−→ so(n),
n(n−1)
in particular, dim spin(n) = dim so(n) = 2 .
Let's investigate this map a little further. Recall that the Lie algebra of Cl∗0,n
is just the Cliord algebra Cl0,n itself with the commutator bracket. Spin(n) is
a Lie subgroup of Cl∗0,n and hence the Lie algebra spin(n) is a Lie subalgebra of
Cl0,n .
Proposition 2.21. Let {e1 , . . . , en } be an orthonormal basis for Rn ,
then spin(n) ⊆ Cl0,n is spanned by elements of the form ei ej where 1 ≤ i <
j ≤ n. Furthermore Λ∗ maps ei ej to the matrix 2Bij ∈ so(n) where Bij is the
n × n-matrix which is −1 in its ij 'th entry and 1 in its ji'th entry.
Proof. Consider the curve

t 7−→ (ei cos t + ej sin t)(−ei cos t + ej sin t) = cos(2t) + sin(2t)ei ej .

Spin(n) since it is the product of two unit vectors, and its value at
It is a curve in
t=0 1. Upon dierentiating at t = 0 we get 2ei ej , which
is the neutral element
is then an element of T1 Spin(n) ∼ = spin(n). They are all linearly independent i
Cl0,n hence also in spin(n), and there are exactly n(n−1)
2 of them, i.e. they span
spin(n).
Now, Λ is the restriction of the twisted adjoint representation to Spin(n), and
0 −1 n
since Spin(n) ⊆ Cl0,n we get Λ(g)v = gvg for g ∈ Spin(n) and v ∈ R . As for
the usual adjoint representation one can calculate

(Λ∗ X)v = Xv − vX (2.3)

2 See for instance [Warner], Proposition 3.26.

28 Chapter 2  Spin Groups

in particular we get


 0,
 6 i, j
k=
Λ∗ (ei ej )ek = ei ej ek − ek ei ej = 2ej , k=i

−2ei , k=j


We see that Λ∗ (ei ej ) acts in the same way on Rn as the matrix 2Bij , thus we
may identify Λ∗ (ei ej ) = 2Bij .

We can rephrase the rst part of this proposition by saying that under the
∼
symbol map σ : Cl0,n −−→ Λ∗ Rn , the Lie algebra spin(n) gets mapped to Λ2 Rn .
We end the section by a short description of some covering properties of
Spinc (Φ).

Proposition 2.22. The map Λc : Spinc (Φ) −→ SO(Φ)×U(1) given by [g, z] 7−→
(Λ(g), z 2 ) is a double covering.
Proof. Λ(−g) = Λ(g) and (−z)2 =
It is easy to see that it is well-dened (since
2
z ). It is a covering map because it is the quotient map of the even Z2 -action on
Spinc (Φ) given by (−1, [g, z]) 7−→ [−g, z] = [g, −z]. Thus it follows from Lemma
2.15.

SO(n) × U(1) is compact, Spinc (n) is also compact. Furthermore, for

Since
n ≥ 2 Spinc (n) is connected (cf. Proposition 2.9 and connectivity of Spin(n)).
c
Thus π1 (Spin (n)) can be identied with a subgroup of π1 (SO(n) × U(1)) =
Z2 × Z of index 2, i.e. π1 (Spinc (n)) ∼
= Z.

2.4 Spin Group Representations

In this section we will treat the basics of the representation theory of spin
groups. In this section we will restrict our attention to a particular complex
representation of Spin(n), the spinor representation, dened as follows:

Denition 2.23. complex spinor representation κn of Spin(n) we under-

By the
stand the restrictionκn |Spin(n) −→ AutC (∆n ) of the complex spin representation
κn of ClC
n to Spin(n).
Similarly, we dene the spin -representation κn of Spin (n) by restricting the
c c c
c C
spin representation to Spin (n) ⊆ Cln .

We stress that we use the term spin representation for the irreducible Cliord
algebra representations and spinor representation for the associated spin group
representations.
Of course we can in a similar way dene the real spinor representation of
Spin(n) by restricting ρn to Spin(n), but we will not consider them here.

Theorem 2.24. For each n the complex spinor representation κn is a faithful

representation of Spin(n).
Proof
∼
. If n = 2k is even, then κn is a restriction of the isomorphism κn :
ClC
n −−→ EndC (∆2k ) and therefore injective.
So let's assume that n = 2k + 1. By denition we have ∆2k = ∆2k+1 and
consequently Aut(∆2k ) = Aut(∆2k+1 ). We can think of Spin(2k) as sitting
2.4 Spin Group Representations 29

inside Spin(2k + 1).3 Denoting the injection ι the following diagram commutes:

Spin(2k)
κ2k
/ Aut(∆2k )

ι id
 
Spin(2k + 1) / Aut(∆2k+1 )
κ2k+1

Now, put H := ker κ2k+1 ⊆ Spin(2k + 1). The goal is to verify H = {1}, but
rst we show that H ∩ Spin(2k) = {1}. The inclusion ⊇ follows since 1 clearly
sits in ker κ2k+1 . Now, assume that h ∈ H ∩ Spin(2k). In particular, h ∈ H and

h = ι(e h) for some eh ∈ Spin(2k). Since h sits in H , κ2k+1 (h) = id∆2k+1 . From
the commutativity of the diagram it follows that κ2k (e h) = id∆2k+1 . But since
κ2k is injective, eh must be 1, and so must h. This shows H ∩ Spin(2k) ⊆ {1}.
Identifying elements A ∈ SO(n) with elements in SO(2k + 1) of the form
diag(A, 1), we obtain SO(2k) ⊆ SO(2k + 1) like the spin groups. Recall that
Λ : Spin(2k + 1) −→ SO(2k + 1) is a surjective homomorphism. Thus Λ(H) is a
normal subgroup of SO(2k + 1), since H as a kernel is normal in Spin(2k + 1).
Now we claim
Λ(H ∩ Spin(2k)) = Λ(H) ∩ SO(2k).
The inclusion  ⊆ is obvious, and  ⊇ follows from the surjectivity ofΛ. Hence
we have Λ(H) ∩ SO(2k) = {I} (here, I denotes the identity matrix). We want
to show that Λ(H) = {I}, so let A ∈ Λ(H) ⊆ SO(2k + 1). Its characteristic
polynomial is of odd degree, and it thus has a real root. As A ∈ SO(2k + 1),
all eigenvalues have modulus 1. Moreover, A preserves orientation, so this root
must be 1. Denote the corresponding eigenvector by v0 and choose an ordered,
2k+1
positively oriented orthonormal basis for R containing v0 as the last vector.
If B denotes the change-of-basis matrix, then we have the block diagonal matrix
BAB −1 = diag(A, e 1), where Ae ∈ SO(2k) and 1 is the unit of R. We can now
−1 −1
identify A with BAB
e . Hence, BAB ∈ SO(2k), and since Λ(H) was normal,
−1
we also have BAB ∈ Λ(H). All together we have BAB −1 ∈ Λ(H) ∩ SO(2k) =
{I} and so A = I .
Now we have Λ(H) = {I}. We have two possibilities: H = {1} or H = {±1}.
But −1 cannot be in the kernel of the spinor representation (because it's not in
the kernel of the spin representation, from which it came). Therefore, H = {1},
and κ2k+1 is injective.

This theorem is not as innocent as it might look. It actually tells us that the
spinor representations do not arise as lifts of SO(n)-representations, since a lift
of an SO(n)-representation necessarily contains {±1} in its kernel.
We now want to decompose the spinor representations into irreducible repre-
sentations. To this end we need:

Lemma 2.25. For any complex vector space V the endomorphism algebra
End(V ) is a simple algebra, i.e. the only ideals are the trivial ones. In par-
ticular if dim W < dim V , then any homomorphism ϕ : End(V ) −→ End(W ) is
trivial : ϕ ≡ 0.
Proof. Let n be the dimension of V V . Then we can think
and x a basis for
of End(V ) as the algebra of complex n × n-matrices. Now let I ⊆ End(V )
be any non-zero ideal, and let 6 a ∈ I . Then a has an eigenvalue λ 6= 0
0 =
(because C is algebraically closed). By a suitable basis transformation, given by
3 If Cl
0,2k is generated by {e1 , . . . , e2k } and Cl0,2k+1 is generated by {e1 , . . . , e2k+1 } then
0 0

we have a linear injection ι : Cl0,2k ,−→ Cl0,2k+1 by dening ι(ej ) = e0j . This restricts to an
injection ι : Spin(2k) ,−→ Spin(2k + 1).
30 Chapter 2  Spin Groups

a change-of-basis-matrix b, and subsequently multiplying diag(1/λ, 0, . . . , 0) on

b−1 ab from the left we obtain the matrix diag(1, 0 . . . , 0). It has been constructed
from a just by multiplication, so it's in I . By a similar argument we obtain
diag(0, . . . , 0, 1, 0 . . . , 0) ∈ I . The sum of all these is just the identity matrix,
which therefore is also in I . Thus, I = End(V ).
If ϕ : End(V ) −→ End(W ) is a homomorphism, ker ϕ is an ideal in End(V ),
thus ker ϕ = {0} or ker ϕ = End(V ). But since dim W < dim V injectivity of ϕ
is impossible. Therefore ker ϕ = End(V ) and ϕ ≡ 0.

Decomposing κ2k+1 into irreducibles is easy:

Theorem 2.26. The spinor representation κ2k+1 of Spin(2k + 1) is irreducible.

Proof. Let's assume that {0} =
6 W ( ∆2k+1 Spin(2k + 1)-invariant sub-
is a
space, i.e. for each g ∈
Spin(2k + 1)
κ2k+1 (g)W ⊆ W . Consider an element of the form ei ej , i < j ({e1 , . . . , e2k+1 }
2k+1
is an orthonormal basis for the vector space R underlying Cl0,2k+1 ). It is an
element of Spin(2k + 1) and therefore κ2k+1 (ei ej )W ⊆ W . I.e. κ2k+1 is actually
dened on all elements {ei1 · · · eim } where i1 < · · · < im and m is even. On the
C 0
other hand these elements constitute a complex basis for (Cl2k+1 ) . Hence we
C 0
get an algebra representation ϕ : (Cl2k+1 ) −→ End(W ) by extending κ2k+1

linearly. But recall that (Cl2k+1 ) ∼ ∼

C 0
= ClC2k = End(∆2k ) (Proposition 1.16) so
that we get an algebra homomorphism

ϕ : End(∆2k ) −→ End(W ).

W was a proper subspace of ∆2k+1 = ∆2k , so dim W < dim ∆2k . Lemma 2.25
now guarantees that ϕ ≡ 0. Since ϕ is an extension of κ2k+1 , this should also be
zero. As κ2k+1 is injective by Theorem 2.24 this is a contradiction, so W cannot
be invariant.

Example 2.27. Again we consider our favorite spin group Spin(3). What is
the spinor representation of Spin(3)? Recall that Spin(3) ∼
= SU(2) and that for
each n SU(2) has exactly one irreducible representation πn of dimension n + 1
on the space of homogenous n-degree polynomials in two variables. The spinor
representation κ3 is a 2-dimensional irreducible representation, thus it must be
equivalent to π1 .

κ2k is not an irreducible representation, but it can quite easily be decom-

posed into such. To do this recall the volume element , the unique element ω in
Cl0,2k e1 · · · e2k . It commutes with everything in the even part of Cl0,2k
given by
k
and anti-commutes with the odd part. The map f = i κ2k (ω), which was an
involution, gave rise to a splitting ∆2k ∼
+ −
∆
= 2k ⊕ ∆ 2k .

2k and ∆2k are κ2k -invariant subspaces. Thus, κ2k induces

Lemma 2.28. ∆+ −

representations κ±
2k on ∆2k such that κ2k = κ2k ⊕ κ2k .
± + −

Proof. We want to show κ2k (g)∆± ±

2k ⊆ ∆2k for any g ∈ Spin(2k), so let ψ be a
positive Weyl spinor. Then

f (κ2k (g)ψ) = κ2k (g)f (ψ) = κ2k (g)ψ

so κ2k (g)ψ ∈ ∆+
2k . Likewise if ψ is a negative Weyl spinor.

2k are irreducible representations of Spin(2k).

Theorem 2.29. κ±
2.5 Spin Structures 31

Proof. Like in the proof of Theorem 2.26 a κ+

2k -invariant subspace {0} =
6 W (
∆+
2k gives rise to a representation

0
ϕ : (ClC
2k ) −→ End(W ).

Again, by Proposition 1.16 and Corollary 1.20:

0 ∼ ∼
(ClC C
2k ) = Cl2k−1 = End(∆2k−1 ) ⊕ End(∆2k−1 ).

We get homomorphisms ϕ1 , ϕ2 : End(∆2k−1 ) −→ End(W ) simply by

ϕ1 (x) = ϕ(x, 1) and ϕ2 (y) = ϕ(1, y).

By assumption dim W < dim ∆+ 2k = dim ∆2k−1 and so by Lemma 2.25 ϕ1 , ϕ2 ≡

0. This means ϕ(x, y) = ϕ((x, 1)(1, y)) = ϕ1 (x)ϕ2 (y) = 0, hence ϕ ≡ 0 and thus
κ+
2k ≡ 0 which is a contradiction.

The covering space results in the previous section yields the following result

Corollary 2.30. Let κ : Spin(n) −→ Aut(V ) be a nite-dimensional repre-

sentation of Spin(n) which is the restriction of an algebra representation ρ :
Cl0,n −→ End(V ) (e.g. the spinor representation κn ). Then for the induced
representation κ∗ : spin(n) −→ End(V ) it holds that
κ∗ (X)v = ρ(X)v

for X ∈ spin(n) ⊆ Cl0,n and v ∈ V .

Proof. Note that ρ is a linear map, hence the induced representation of the
restriction ρ|Cl∗0,n is just ρ itself (where we have identied cl∗0,n ∼
= Cl0,n ). The
induced representation of κ = ρ|Spin(n) is just the restriction of ρ|Cl∗ , hence
0,n
the formula follows.

A close examination of the proofs above will reveal that nothing is used which
does not hold for Spinc (n) as well. We may therefore summarize the results above
c
in the following statement about the spin -representations:

Theorem 2.31. For each n the spinc -representation κcn is faithful. If n is odd,
the representation is irreducible and if n is even it splits in a direct sum of two
irreducible representations κcn = (κcn )+ ⊕ (κcn )− where (κcn )± are representations
on the space ∆± n.

2.5 Spin Structures

c
In this section let us recall/introduce the notions of spin and spin structures
and x some notation.

Denition 2.32 (Spin Structure). Let E −→ M be a real oriented Rieman-

nian vector bundle of rank n≥3
π : PSO (E) −→ M its oriented orthonor-
and
mal frame bundle. A spin structure on E is a lift of PSO (E) to a principal
Spin(n)-bundle. More precisely, a spin structure is a pair (PSpin (E), Φ) of a prin-
e : PSpin (E) −→ M and a bundle map Φ : PSpin (E) −→
cipal Spin(n)-bundle π
PSO (E) such that
Φ(p · g) = Φ(p) · Λ(g)
where Λ : Spin(n) −→ SO(n) is the double covering.
32 Chapter 2  Spin Groups

Often, the vector bundle in question will be the tangent bundle of some man-
ifold M (provided of course that this manifold is oriented and has been given
a metric). Then we will write PSO (M ) and PSpin (M ) for the oriented frame
bundle resp. the spin bundle. If T M has a spin structure we say that M is a
spin manifold .
Let's try and see this from a local perspective. It is well-known that a principal
G-bundle over a manifold M can be (uniquely) described by the following data:
a cover (Uα )α∈A of open sets and for each pair (α, β) ∈ A × A a smooth map
(the transition function ) gαβ : Uαβ −→ G (where Uαβ := Uα ∩ Uβ ) satisfying
that gαα (x) = 1 for all x ∈ Uα and satisfying the cocycle condition ,

gαβ (x)gβγ (x)gγα (x) = 1

for all triples (α, β, γ) ∈ A × A × A and for all x ∈ Uαβγ . This collection of
data is called a gluing cocycle . SO(n)-bundle
So if we consider our principal
PSO (E), then there exists a cover (Uα ) and transition functions gαβ : Uαβ −→
SO(n) satisfying the cocycle condition. The existence of a spin structure is then
eαβ : Uαβ −→ Spin(n) over Λ satisfying the
equivalent to the existence of lifts g
cocycle condition. It is well-known that such a set of lifts exists if and only if the
second Stiefel-Whitney class is zero. In fact this follows more or less by denition
of the second Stiefel-Whitney class (in the setting of Cech cohomology). In the
armative case the possible spin structures on M are parametrized by elements
of the rst Cech cohomology group Ȟ 1 (M ; Z2 ) which is, of course, isomorphic to
1
the singular cohomology group H (M ; Z2 ) with coecients in Z2 . For instance
S for n ≥ 3 admits a spin structure, since H 2 (S n ; Z2 ) = 0, so the second Stiefel-
n
1 n
Whitney class can be nothing but 0. Since H (S ; Z2 ) = 0 the spin structure
must be unique.
The spin group Spin(n) has a distinguished complex representation, called the
spinor representation κn : Spin(n) −→ Aut(∆n ) where ∆n = C
2k
and where
k = b n2 c, n
the integer part of
2 . This is a faithful representation (hence does
not descend to a representation of SO(n)) and when n is odd it is irreducible.
For n even, it decomposes into a direct sum of two irreducible representations
−
κn = κ+ +
n ⊕ κn and the corresponding representation spaces are denoted ∆n and
−
∆n respectively.
Given a principal Spin(n)-bundle π : Q −→ M (originating from a spin
structure, say) we can form the associated complex vector bundle w.r.t. the
spinor representation, namely S := Q ×κn ∆n which is the quotient of the direct
product Q × ∆n under the equivalence relation (p, v) = (p · g, κn (g −1 )v) and
with projection q : S −→ M given by q([p, v]) = π(p). This is called the spinor
bundle and sections of this bundle are called spinors or Dirac spinors . If n is
even this bundle splits, in the same way as the representation, into two bundles
S = S+ ⊕ S− where in fact S± is the associated bundle Q ×κ± ∆±
n. Sections
Weyl spinors
n
of these vector bundles are called positive resp. negative or even
resp. odd chiral spinors . We will discuss these vector bundles and some of their
properties in more detail in the next section when we dene the Dirac operator.
Next, recall how the Lie group Spinc (n) is dened: It is the group inside
Cl0,n ⊗ C generated by Spin(n) ⊗ 1 and 1 ⊗ U(1). Equivalently, Spinc (n) =
Spin(n) ×±1 U(1), the quotient where we collapse the subgroup {±(1, 1)}. Thus,
in Spin(n) × U(1) we identify (g, z) with (−g, −z). The equivalence class con-
c
taining (g, z) will be denoted [g, z]. This is usually how we will view Spin (n).
c c
We can dene a Lie group homomorphism ρ : Spin (n) −→ SO(n) by
[g, z] 7−→ Λ(g) (again, Λ : Spin(n) −→ SO(n) is the double covering). This
is well-dened, since Λ(−g) = Λ(g), however, contrary to Λ, this is no longer a
covering map, since its bers are not discrete. Instead we may view this map as
an n-dimensional representation of Spinc (n).
2.5 Spin Structures 33

Similarly, we dene a Lie group homomorphism λ : Spinc (n) −→ U(1) by

[g, z] 7−→ z . Again, this is well-dened since (−z) = z 2 . We may view λ as a
2 2
c
1-dimensional unitary representation of Spin (n).
Finally, Spin (n) is a covering space, not of SO(n) as we saw above, but of
c

SO(n) × U(1). We simply dene the homomorphism Λc : Spinc (n) −→ SO(n) ×

U(1) by Λc ([g, z]) = (Λ(g), z 2 ). One can then check that this is a smooth double
covering of SO(n) × U(1).

Denition 2.33 (Spinc -structure). Let E −→ M be a real oriented Rie-

mannian vector bundle of rank n ≥ 3 with oriented orthonormal frame bun-
dle π : PSO (E) −→ M . A spinc -structure on E is a principal Spinc (n)-bundle
c
π
e : PSpin (E) −→ M and a bundle map Φc : PSpin c
(E) −→ PSO (E) such that
c c c c c
Φ (p · g) = Φ (p) · ρ (g) where ρ : Spin (n) −→ SO(n) is the map dened
above.
c
Two spin -structures
c
(PSpin (E)1 , Φc1 ) c
(PSpin (E)2 , Φc2 ) are said to be iso-
and
morphic if there exists a bundle
c 1 c 2
isomorphism PSpin (E) −→ PSpin (E) making
the following diagram commutative:

c
PSpin (E)1 / PSpin
c
(E)2
LLL rr
LLL r
L rrrc
Φc1 LLL rr
% yrr Φ 2

PSO (E)

c
The set of isomorphism classes of spin -structures on E is denoted Spinc (E). In
the case where E happens to be the tangent bundle of an oriented Riemannian
c
manifold a manifold equipped with a spin -structure is called a spinc -manifold .
c
The set of isomorphism classes of spin -structures on M is denoted Spinc (M ).

Assume that E is an oriented Riemannian vector bundle with a spin structure

π : PSpin (E) −→ M and bundle map Φ : PSpin (E) −→ PSO (E). Then E has a
c
canonical spin -structure given in the following way: Dene

c
PSpin (E) := PSpin (E) ×±1 U(1)

more precisely we take the product PSpin (E) × U(1) and mod out by the equiv-
alence relation ∼ given by (p, z) ∼ (p0 , z 0 ) i π(p) = π(p0 ) and (p0 , z 0 ) = ±(p, z).
The space
c
PSpin (E) can then be equipped with a right Spinc (n) action

[p, z] · [g, z 0 ] = [p · g, zz 0 ]
c
which in combination with the projection map π
e : PSpin (E) −→ M given by
c c
π
e([p, z]) = π(p) turns PSpin (E) into a principal Spin (n)-bundle. Finally, dene
Φc : PSpin
c
(E) −→ PSO (E) by Φc ([p, z]) = Φ(p). We see that

Φc ([p, z] · [g, z 0 ]) = Φc ([p · g, zz 0 ]) = Φ(p · g) = Φ(p) · Λ(g)

= Φc ([p, z]) · ρc ([g, z 0 ])
c
and thus that (PSpin (E), Φc ) c
is a spin -structure on E.
c
This shows that the concept of a spin -structure is more general than that of a
c
spin structure. I will not go into a topological discussion of when spin -structures
exist, except mentioning the following result

Theorem 2.34 (Hirzebruch-Hopf ). Any oriented Riemannian 4-manifold is

a spinc -manifold.
A proof of this statement can be found in [ ], Lemma 3.1.2. ?
34 Chapter 2  Spin Groups

Denition 2.35 (Determinant Line Bundle). Given a principal Spinc (n)-

bundle Q −→ M we can form the complex line bundle L := Q ×λ C associated
to the 1-dimensional representation λ dened above. This bundle is called the
determinant line bundle , det(Q) for the Spinc (n)-bundle Q. If the Spinc (n)-
c
bundle originates from a spin -structure σ , we will often write det(σ) for the
determinant line bundle.

The determinant line bundle can be given a hermitian ber metric by dening
h[p, w], [p, w0 ]i := hw, w0 iC = ww0 , we simply transfer the usual inner product
on C to L. This is well-dened since

h[p · g, λ(g)w], [p · g, λ(g)w0 ]i = hλ(g)w, λ(g)w0 iC = hw, w0 iC

sinceλ(g) ∈ U(1) and we may therefore form the unitary frame bundle L0 :=
PU (L), a principal U(1)-bundle over M .

Lemma 2.36. There is a bundle isomorphism L0 ∼ c

= PSpin (E) ×λ U(1).

Proof. We dene the bundle map

c
PSpin (E) ×λ U(1) −→ L0 in the following
c
way: the element [p, z] ∈ PSpin (E) ×λ U(1) should be mapped to the frame (i.e.
the isometric isomorphism C −→ Lπ e(p) ) given by w 7−→ [p, zw]. Since z ∈ U(1),
this is an isometric isomorphism, i.e. a frame at π
e(p). Since any frame at π
e(p)
must be of the form w 7−→ w[p, z] = [p, zw] for a unit vector [p, z] ∈ Lπe(p) , i.e.
for z ∈ U(1), we see that the bundle map is surjective. Its not hard to see that
it is injective also and hence a bundle isomorphism.

Let's put a local perspective on this as we did for the spin structures. Our
starting point is again a principal SO(n)-bundle which is given in terms of a cover
(Uα ) and transition functions gαβ : Uαβ −→ SO(n). Picking a spinc -structure
eαβ : Uαβ −→ Spinc (n) along ρc
(if it exists) is then equivalent to picking lifts g
such that the cocycle condition is satised. We see that g eαβ must be of the form
[hαβ , zαβ ] where hαβ : Uαβ −→ Spin(n) and zαβ : Uαβ −→ U(1) are such that
Λ ◦ hαβ = gαβ and such that the pair

(hαβ (x)hβγ (x)hγα (x), zαβ (x)zβγ (x)zγα (x))

is either (−1, −1) or (1, 1), thus (hαβ ) and (zαβ ) need only satisfy the cocycle
condition up to a sign. However if the bundle admits a spin structure, we may
pickhαβ such that it does satisfy the cocycle condition, and then we can pick
c
zαβ = 1, this is the canonical spin -structure of a spin structure.
2
Given the families of maps (hαβ ) and (zαβ ) we can dene λαβ := zαβ which
maps Uαβ U(1). This family of maps satises the cocycle condition and thus
into
represents a principal U(1)-bundle. This bundle is nothing but the unitary frame
bundle of the determinant line bundle. By the remarks above we then conclude
c
that the determinant line bundle of the canonical spin -structure induced by a
spin structure has trivial determinant line bundle.
c c
Let M be a spin -manifold. How many dierent spin -structures does this
manifold have? To shed some light on this question, let Pic∞ (M ) denote the set
of complex Riemannian line bundles (i.e. bundles carrying a sesquilinear, conju-
gate symmetric, positive denit 2-form). This is a group under tensor product,
known as the Picard group . This group is in 1-1 correspondence with the set of
principal U(1)-bundles over M - the map from Pic∞ (M ) to the set of principal
U(1)-bundles is simply given by forming the unitary frame bundle. Recall that
the rst Chern class is a bijection

c1 : Pic∞ (M ) −→ H 2 (M ; Z). (2.4)

2.5 Spin Structures 35

We can let Pic∞ (M ) act on Spinc (M ) σ ∈ Spinc (M ) is

in the following way: If
c ∞
a spin -structure given by the gluing cocycle [hαβ , zαβ ] and if L ∈ Pic (M ) is
c
given by the gluing cocycle (ζαβ ) then we dene the spin -structure σ ⊗ L by
2 2
the gluing cocycle [hαβ , zαβ ζαβ ]. Since (zαβ ζαβ ) = λαβ ζαβ we see that

det(σ ⊗ L) = det(σ) ⊗ L⊗2 .

Proposition 2.37. The action of Pic∞ (M ) on Spinc (M ) is free and transi-

tive. Thus for a xed spinc -structure σ0 the map L 7−→ σ0 ⊗ L is a bijection
Pic∞ (M ) −→ Spinc (M ). Composing with (2.4) we obtain a bijection
∼
Spinc (M ) −−→ H 2 (M ; Z).

Moreover, at most nitely many spinc -structures have the same determinant line
bundle.
Proof. First, the action is free: if σ ⊗ L = σ , i.e. if [hαβ , zαβ ζαβ ] = [hαβ , zαβ ],
then we must have zαβ ζαβ = zαβ , i.e. ζαβ = 1 and since this is the gluing cocycle
for L, this bundle must be trivial.
c
The action is transitive: Assume we have two spin -structures σ1 and σ2
(i) (i) (1) (2)
given by gluing cocycles [hαβ , zαβ ]. Since Λ(hαβ ) = Λ(hαβ ) = gαβ we must have
(1) (2) (1) (2)
hαβ = ±hαβ . By a change of sign if necessary we can thus assume hαβ = hαβ .
(2) (1)
Now put ζαβ := zαβ /zαβ . Clearly ζαβ maps into U(1) and we see that

(2) i
(1) (1) zαβ
h
(2) (2) (1) (1)
[hαβ , zαβ ] = hαβ , zαβ (1) = [hαβ , zαβ ζαβ ]
zαβ

and hence that σ2 = σ1 ⊗ L.

c
At last, assume σ1 and σ2 are two spin -structures having the same determi-
nant line bundle. By the rst part of the proof, there exists a unique line bundle
L such thatσ2 = σ1 ⊗ L. If L is given by the gluing cocycle (ζαβ ), then the
2
requirement det(σ1 ) = det(σ2 ) implies that ζαβ = 1, i.e. ζαβ maps into Z2 . Thus
(ζαβ ) determines an element of the Cech cohomology group Ȟ 1 (M ; Z2 ) which
1
is isomorphic to the singular cohomology group H (M ; Z2 ). Since this is nite,
L belongs to a nite set, hence the conclusion.
In the same way we dened the spinor bundles associated to a principal
Spin(n)-bundle, we can form spinor bundles associated to a Spinc (n)-bundle.
Spinc (n) sits inside ClC
n and the fundamental representation of this algebra
c
on the space ∆n of Dirac spinors restricts to a group representation κn :
c
Spin (n) −→ Aut(∆n ). Thus if E is a real vector bundle carrying a spinc -
structure PSpin (E) we dene the complex spinor bundle :
c

S c (E) := PSpin
c
(E) ×κcn ∆n .

If the principal Spinc (n)-bundle

is given by the gluing cocycle [hαβ , zαβ ] then
S c (E) is given by the gluing cocycle κcn ([hαβ , zαβ ]). If we change spinc -structure
from σ to σ ⊗ L where L is a line bundle given by the gluing cocycle (ζαβ ) then
the spinor bundle is given by the gluing cocycle

κcn ([hαβ , zαβ ζαβ ] = κcn ([hαβ , zαβ ])ζαβ

and hence the new spinor bundle is just S c (E) ⊗ L.

As observed above k := dimC ∆n = rankC S c (E) is an even number. In the
representation theory of spin groups it is shown that the two representations
κ∧k
n = κn ∧ · · · ∧ κn and λ⊗k/2 are equivalent. This means that the associated
36 Chapter 2  Spin Groups

vector bundles are isomorphic, giving us the following relations between the
spinor bundles and determinant line bundle L:

Λk S c (E) ∼
= L⊗k/2 . (2.5)

The same result holds for S c (E)± :

Λk/2 (S c (E)± ) ∼
= L⊗k/4 . (2.6)

The formula (2.5) also explains the name determinant line bundle, since the top
exterior product of a vector space or a vector bundle traditionally is called the
determinant.
We also need to discuss the notion of connections on vector bundles and prin-
cipal bundles. First we recall the denitions

Denition 2.38 (Connection on a Vector Bundle). Let E be a smooth

K-vector bundle (K being either R or C) over M and let Ω1 (M, E) denote the E -
valued 1-forms, i.e. sections of T M ⊗K E . By a connection on E we understand
∗

a K-linear map
∇ : Γ(E) −→ Ω1 (M, E)
satisfying the Leibniz rule

∇(f s) = df ⊗ s + f ∇s (2.7)

for f ∈ C ∞ (M ) and s ∈ Γ(E). ∇s is called the covariant derivative of s.

If E is a Riemannian vector bundle, we say that a connection ∇ is compatible
with the metric (or just metric ) if
Xhs, s0 i = h∇X s, s0 i + hs, ∇X s0 i

for all vector elds X and all sections s, s0 ∈ Γ(E).

One can show that any vector bundle can be equipped with a connection, and
that the space of connections is an ane space modeled on Ω1 (M, End(E)), i.e.
1
any two connections dier by an element in Ω (M, End(E)).
Given a connection on TM we dene its torsion by

τ∇ (X, Y ) := ∇X Y − ∇Y X − [X, Y ].

This is an anti-symmetric 2-tensor. The Fundamental Theorem of Riemannian

Geometry states that on a Riemannian manifold, there exists a unique metric
connection whose torsion tensor vanishes identically. This is called the Levi-
Civita connection on M.
π
Let G ,−→ P − −→ M be a smooth principal G-bundle over M . For each p ∈ P
we have the so-called vertical subspace Vp P ⊆ Tp P , namely the kernel of the
dierential dπp : Tp P −→ Tπ(p) M . A connection on P is then loosely speaking
a smooth choice of an algebraic complement over each point. Formally

Denition 2.39 (Connection on a Principal Bundle). For the principal

G-bundle G ,−→ P −→ M a connection is a smooth tangent distribution HP
on P such that for each p ∈ P we have Tp P = Hp P ⊕ Vp P (Hp P is called the
horizontal subspace ) and such that Hp·g P = (dσg )p Hp P where σg : P −→ P is
the map p 7−→ p · g .

There are several equivalent denitions of a connection on a principal G-

bundle. One of them is given in terms of a connection 1-form: If g denotes the Lie
algebra of G, then a connection 1-form 1
is a smooth g-valued 1-form ω ∈ Ω (P, g)
2.5 Spin Structures 37

satisfying the following two axioms: (σg )∗ ω = Adg−1 ◦ω and ωp (A] (p)) = A for
allA ∈ g where A ]
is the fundamental vector eld on P determined by A.
Letting σp be the map G −→ P , g 7−→ p · g , then A] is given by

d
A] (p) = dσp (A) = (p · exp(tA)
dt t=0

so the second axiom could be phrased as ωp ◦ dσp = id.

A connection 1-form induces a connection on the principal bundle, simply by
Hp P = ker ωp , and vice versa.
Yet a third denition of a connection is via local gauge potentials satisfying
a compatibility conditions on the overlap of their local domains.

Now assume that π : P −→ M is a principal SO(n)-bundle over M and as-

sume it has a connection. Assume furthermore that the bundle lifts to a spin-
bundle e : S(P ) −→ M .
π The map Φ : S(P ) −→ P is in fact a double covering
map: it is not hard to see that we can pick local trivializations of S(P ) and
P on a common neighborhood U ⊆ M such that Φ locally takes the form
(x, g) 7−→ (x, Λ(g)). Since Λ is a double covering map, Φ|U is a double covering
map, and since being a covering map is a local property, Φ itself is a dou-
ble covering map. In particular it is a local dieomorphism and its dierential
dΦp : Tp S(P ) −→ TΦ(p) P is an isomorphism for all p ∈ S(P ). But then we can
lift the connection onP to a connection on S(P ), simply by dening the hori-
−1
zontal subspace Hp S(P ) := (dΦp ) (HΦ(p) P ). Dened in this way from a local
dieomorphism, it is obviously a smooth tangent distribution. The additional
requirement is also satised, as can be seen directly as follows:

Hp·g S(P ) = (dΦp )−1 (HΦ(p·g) P ) = (dΦp )−1 (HΦ(p)·Λ(g) P )

= (dΦp )−1 dσΛ(g) HΦ(p) P = de σg (dΦp )−1 (HΦ(p) P )

= de
σg Hp S(P ),

the fourth identity follows from the requirement Φ ◦ (e

σg ) = σΛ(g) ◦ Φ. Thus, the
connection on P lifts to a connection on S(P ). If P is the oriented orthonormal
frame bundle and the connection is the Levi-Civita connection, the lifted con-
nection is called the spin connection .
c
The situation for spin -structures is somewhat more complicated. Assume again
c
PSO (E) is the frame bundle of a vector bundle E and assume it carries a spin -
c c
structure as well as a connection. Since the map Φ : PSpin (E) −→ PSO (E) is
not c c
a covering map (because ρ : Spin (n) −→ SO(n) is not a covering map),
c
we cannot simply lift a connection from PSO (E) to PSpin (E) as we did before.
c
To x a connection on the Spin (n)-bundle we need not only a connection on
0
the SO(n)-bundle but also a connection A on the frame bundle L of the de-
terminant line bundle L. Let's spend a few moments to describe how this works
out. Consider the product bundle π × π 0 : PSO (E) × L0 −→ M × M . This is
an SO(n) × U(1)-principal bundle over M × M . The connections on PSO (E)
0 0
and L give a natural connection on PSO (E) × L : namely choose in the tan-
0 ∼ 0
gent space T(p,q) (PSO (E) × L ) = Tp PSO (E) × Tq L the horizontal subspace
0
Hp PSO (E) × Hq L . Then we get a decomposition

T(p,q) (PSO (E) × L0 ) = (Hp PSO (E) × Hq L0 ) ⊕ (Vp PSO (E) × Vq L0 ).

It shouldn't be hard to check that this is a connection on the product bundle.

Let ∆ : M −→ M × M be the diagonal map x 7−→ (x, x) and consider the
pullback bundleQ := ∆∗ (PSO (E) × L0 ) (i.e. the restriction to M viewed as the
38 Chapter 2  Spin Groups

diagonal in M × M ). Q is what we will call the bered product or the spliced

bundle of P 0 0
and L . Restrict the connection on PSO (E) × L to this new bundle
(i.e. pull the connection back along ∆). Thus we have a principal SO(n) ×
U(1)-bundle Q −→ M carrying a connection determined by the connections on
PSO (E) and L0 . If π1 : Q −→ PSO (E) and π2 : Q −→ L0 denote the obvious
projection maps, and if ω is the connection 1-form for the connection on P
0
and A is the connection form for the connection on L one can show that the
connection on Q is given by the connection 1-form

ω A := (π1∗ ω) ⊕ (π2∗ A) (2.8)

which takes values in the Lie algebra spinc (n) ∼

= spin(n) ⊕ iR.
c
In order to lift this connection to the Spin (n)-bundle we need a covering of
Q. We know that Λc : Spinc (n) −→ SO(n) × U(1) is a double covering, and
c
so inspired by this we seek a bundle map PSpin (E) −→ Q which locally looks
c c
like Λ . Our candidate: Ξ(p) := (Φ (p), [p, 1]). To see that it locally looks like
Λc , pick trivializations Ψc and Ψ for PSpin
c
(E) resp. PSO (E) over a common
c c −1
domain U ⊆ M such that Ψ ◦ Φ ◦ (Ψ ) (x, [g, z]) = (x, Λ(g)) (remember that
[g, z] ∈ Spinc (n) for g ∈ Spin(n)). The trivialization Ψc for PSpin (E) gives a
0
trivialization Ψ of L over U by (using the isomorphism from Lemma 2.36)

Ψ([s(x), z]) = (x, z)

(where s(x) := (Ψc )−1 (x, e) is the local section of PSpin

c
(E) corresponding to the
c
trivialization Ψ and e ∈ Spin (n) is the neutral element), and further Ψ and Ψ
give a trivialization of Q over U , denoted Ψ × Ψ (with a slight abuse of notation,
since the trivialization is only a berwise product). We want to show that

(Ψ × Ψ) ◦ Ξ ◦ (Ψc )−1 (x, [g, z]) = (x, Λ(g), z 2 ), (2.9)

and to see this put p := (Ψc )−1 (x, [g, z]). This is mapped to (Φc (p), [p, 1]) by Ξ
and Ψ maps the rst component to (x, Λ(g)) as it should. Note that

[p, 1] = [s(x) · [g, z], 1] = [s(x), λ([g, z])1] = [s(x), z 2 ]

which by Ψ is mapped to (x, z 2 ), thus we have veried (2.9).

c
Now we have a double covering map Ξ : PSpin (E) −→ Q and thus we can re-
c
peat what we did before, lifting the connection on Q to a connection on PSpin (E).
c
If E is the tangent bundle for a spin -manifold M , the connection is called the
spin c -connection .

2.6 The Dirac Operator

In this section we let M denote an oriented Riemannian manifold. No compact-
ness condition is imposed unless specied.
Let E be a real oriented Riemannian vector bundle over M of rank n (often
we will take it to be the tangent bundle). Thus we can construct its oriented
frame bundle PSO (E). We want to construct the so-called Cliord bundle over
E, i.e. an algebra bundle over M whose ber at x is isomorphic to the Cliord
algebra Cl(Ex ). The construction is accomplished as an associated bundle in
the following way: Consider Cl0,n , the Cliord algebra over Rn with the usual
negative denit inner product. We have a representation ρ of SO(n) on Cl0,n :
any A ∈ SO(n) Rn −→ Rn preserves the inner prod-
viewed as a linear map
e : Cl0,n −→ Cl0,n , so the
uct, and hence induces an algebra homomorphism A

representation is given as ρ(A) = A.

e
2.6 The Dirac Operator 39

Denition 2.40 (Cliord Bundle). The Cliord bundle of E is the associated

bundle
Cl(E) = PSO (E) ×ρ Cl0,n .

Elements in Cl(E) [p, ξ], where p ∈ PSO (E) and ξ ∈

are equivalence classes
Cl0,n PSO (E) × Cl0,n is (p, ξ) ∼ (p · A, ρ(A−1 )ξ).
and the equivalence relation on
e : Cl(E) −→ M , [p, ξ] 7−→ π(p) where π : PSO (E) −→ M
The projection map is π
is the projection in the frame bundle. The vector space structure on the bers
is given by
a[p, ξ] + b[p, ξ 0 ] = [p, aξ + bξ 0 ]
(note, by transitivity of the right SO(n)-action on each ber in PSO (E) we can
always assume the p's to be equal). Similarly, the algebra structure is given by

[p, ξ] · [p, ξ 0 ] = [p, ξξ 0 ],

and the identity element is [p, 1]. It is easy to check that these operations are
well-dened. Since ξ 2 = −kξk2 · 1, we get

[p, ξ][p, ξ] = [p, ξ · ξ] = [p, −kξk2 · 1] = −kξk2 [p, 1],

thus each ber is indeed a Cliord algebra of type (0, n). Thus Cl(E)x (the
ber in the Cliord bundle) is isomorphic to Cl(Ex ) (the Cliord algebra of the
vector space Ex ).
Observe that we have Rn ⊆ Cl0,n Rn
is a ρ-invariant subspace and
and that
that ρ(A)|Rn = A, so ρ restricted to this invariant subspace is just the dening
n
representation of SO(n) on R . We write it as id. But this means that we have
n ∼
the subbundle PSO (E) ×id R = E sitting inside Cl(E). Elements in E ⊆ Cl(E)
n
are characterized by being of the form [p, v] where v ∈ R . In particular we may
view Γ(E) as sitting inside Γ(Cl(E)).
For the purpose of studying spinor bundles, as we will do later in this section,
we need another description of the Cliord bundle. Assume that the oriented
Riemannian vector bundle E has a spin structure π : PSpin (E) −→ M with dou-
ble covering bundle mapΦ : PSpin (E) −→ PSO (E). Consider the representation
Ad : Spin(n) −→ Aut(Cl0,n ) given by

Ad(g)ξ = gξg −1

(recall that Spin(n) sits inside Cl0,n , so multiplication makes sense), then the
following diagram commutes (simply because Ad(g) is the unique extension of
Λ(g) to Cl0,n ):
Spin(n)
Ad / Aut(Cl0,n )
ll l5
llll
l
lll ρ
Λ
 lll
SO(n)

From the principal Spin(n)-bundle PSpin (E) and the representation Ad, we can
form the associated bundle PSpin (E) ×Ad Cl0,n .
c c c
Analogously, dene Ad : Spin (n) −→ Aut(Cl0,n ) by Ad ([g, z]) = Ad(g).
This is well-dened since Ad(−g) = Ad(g).

Lemma 2.41. The map Ψ : PSpin (E) ×Ad Cl0,n −→ Cl(E) = PSO (E) ×ρ Cl0,n
given by [p, ξ] 7−→ [Φ(p), ξ] is a well-dened smooth algebra bundle isomorphism.
Similarly, the map Ψc : PSpin c
(E) ×Adc Cl0,n −→ Cl(E) given by [p, ξ] 7−→
[Φ (p), ξ] is a well-dened smooth algebra bundle isomorphism.
c
40 Chapter 2  Spin Groups

Proof. It is well-dened, since

Ψ([p · g −1 , Ad(g)ξ]) = [Φ(p · g −1 ), Ad(g)ξ] = [Φ(p) · Λ(g)−1 , ρ(Λ(g))ξ]

= [Φ(p), ξ] = Ψ([p, ξ]),

the third identity is a consequence of equivariance of Ψ and of the commut-

ing diagram just above. Restricted to the ber over x, the map is an algebra
homomorphism (we skip checking linearity):

Ψx ([p, ξ][p, ξ 0 ]) = Ψx ([p, ξ · ξ 0 ]) = [Φ(p), ξ · ξ 0 ] = [Φ(p), ξ] · [Φ(p), ξ 0 ]

= Ψx ([p, ξ])Ψx ([p, ξ 0 ]).

0 = Ψx ([p, ξ]) = [Φ(p), ξ], then ξ must be 0, hence [p, ξ] = 0.

It is injective, for if
Moreover Ψx Φ is. Thus Ψ is an algebra bundle isomorphism.
is surjective, since
c
The verication that Ψ is an algebra bundle isomorphism is completely similar
and so we skip it.

Denition 2.42 (Dirac Bundle). Let E be a real Riemannian vector bundle

over M ∇ a metric connection. A complex vector bundle S over M , is
and
called a Cl(E)-module or a left Cliord module if for each x ∈ M there is a
representation of the algebra Cl(E)x on Sx .
A left Cl(E)-module is called a Dirac bundle over E , provided it is equipped
with a ber metric h , i and a compatible connection ∇ e satisfying the two
additional conditions:

1) Cliord multiplication is skew-adjoint, i.e. for each x∈M and each Vx ∈

Ex and ψ 1 , ψ 2 ∈ Sx :

hVx · ψ1 , ψ2 i + hψ1 , Vx · ψ2 i = 0. (2.10)

2) The connection on S is compatible with the connection on E in the fol-

lowing sense:

∇
e X (V · ψ) = (∇X V ) · ψ + V · (∇
e X ψ) (2.11)

for X ∈ X(M ), V ∈ Γ(E) and ψ ∈ Γ(S).

The single most important example of a Dirac bundle is the spinor bundle
S(E) := PSpin (E) ×κn ∆n

as dened in the previous section. To show that it is a Dirac bundle we rst equip
it with an action of the Cliord bundle Cl(E) = PSpin (E) ×Ad Cl0,n . Consider
on the threefold product PSpin (E) × Cl0,n ×∆n the equivalence relation ∼

(p, ξ, v) ∼ (p · g −1 , Ad(g)ξ, κn (g)v)

for any g ∈ Spin(n). Elements in the quotient space PSpin (E) × Cl0,n ×∆n / ∼
are denoted [p, ξ, v]. As in the proof of Lemma 2.41 one can show that the map

Cl(E) × S(E) −→ PSpin (E) × Cl0,n ×∆n / ∼

given by ([p, ξ], [p, v]) 7−→ [p, ξ, v] is a well-dened bundle isomorphism. This
allows us to dene the Cliord action in the following way: Dene

e : PSpin (E) × Cl0,n ×∆n −→ PSpin (E) × ∆n

µ
2.6 The Dirac Operator 41

by e(q, ξ, v) := (q, ρn (ξ)v) (where ρn : Cl0,n −→ Aut(∆n ) is the spin representa-

µ
tion of the Cliord algebra) and note that the following diagram is commutative

PSpin (E) × Cl0,n ×∆n

µ
e
/ PSpin (E) × ∆n

·g ·g
 
PSpin (E) × Cl0,n ×∆n / PSpin (E) × ∆n
µ
e

where the rst vertical map is (p, ξ, v) 7−→ (p · g −1 , Ad(g)ξ, κn (g)v) and the
−1
second is (p, v) 7−→ (p·g , κn (g)v). Thus µ e induces a map µ : Cl(E)×S(E) −→
S(E) given explicitly by the formula

[p, ξ] · [p, v] := µ([p, ξ], [p, v]) = [p, ρn (ξ)v].

This is the desired Cliord action, turning S(E) into a left Cl(E)-module.
Next we want to give S(E) a metric. Inside Cl0,n we have the nite group

Gn := {ei1 · · · eik | 1 ≤ k ≤ n, 1 ≤ i1 < · · · < ik ≤ n}

(where {e1 , . . . , en } is some orthonormal basis for Rn ). Restricting the spin

representation ρn of Cl0,n to Gn gives a representation of Gn on ∆n , also denoted
ρn . By a well-known result from representation theory, there exists an inner
product h , i∆n on ∆n relative to which ρn is a unitary representation, i.e.

hρn (ei1 · · · eik )v, ρn (ei1 · · · eik )wi∆n = hv, wi∆n .

(To make the notation in the following less cumbersome, we will simply write
Pn
the action of ρn (ξ) on v as ξ · v .) ξ=
i=1 ai ei is a unit vector in
If Rn ⊆ Cl0,n
then ρn (e) is a unitary operator as well: First we observe

hei · v, ej · wi∆n = hej · (ei · v), e2j · wi∆n = −h(ej ei ) · v, wi∆n

= h(ei ej ) · v, wi∆n = h(e2i ej ) · v, ei · wi∆n
= −hej · v, ei · wi∆n ,

and from this we get

n
DX n
X E
hξ · v, ξ · wi∆n = ai ei · v, aj ej · w
∆n
i=1 j=1
n
X X
= a2i hei · v, ei · wi∆n + ai aj hei · v, ej · wi∆n
i=1 i6=j
n
X X
a2i hv, wi∆n +


= ai aj hei · v, ej · wi∆n + hej · v, ei · wi∆n
i=1 i<j

= hv, wi∆n .

Since Spin(n) is generated by unit vectors, we see immediately that κn is a

unitary representation w.r.t. this inner product.
Furthermore, for any ξ ∈ Rn , unit vector or not, ρn (ξ) is a skew-adjoint map:

ξ ξ 1
hξ · v, wi∆n = h kξk · (ξ · v), kξk · wi∆n = hξ 2 · v, ξ · wi∆n = −hv, ξ · wi∆n .
kξk2

Note that this implies that ρn (ξ) is skew-adjoint, when ξ is in Cl10,n (the odd
0
part of Cl0,n ) and that ρn (ξ) is self-adjoint when ξ ∈ Cl0,n (the even part of
Cl0,n ). In particular, κn = ρn |Spin(n) is self-adjoint.
42 Chapter 2  Spin Groups

We can readily extend this inner product to a ber metric on S(E), simply
by dening
h[p, v], [p, w]i := hv, wi∆n .
This is well-dened by unitarity of κn (g):

h[p · g −1 , κn (g)v], [p · g −1 , κn (g)w]i = hκn (g)v, κn (g)wi∆n = hv, wi∆n

= h[p, v], [p, w]i.

Checking condition 1 in the denition of a Dirac bundle is not hard: Let Vx ∈

Ex ⊆ Cl(E)x and ψ1 , ψ2 ∈ Sx (E). We have presentations Vx = [p, v] and ψi =
[p, wi ] where v ∈ Rn ⊆ Cl0,n , p ∈ PSpin (E) and wi ∈ ∆n , and hence:

hVx · ψ1 , ψ2 i = h[p, v] · [p, w1 ], [p, w2 ]i = h[p, v · w1 ], [p, w2 ]i

= hv · w1 , w2 i∆n = −hw1 , v · w2 i∆n = −h[p, w], [p, v] · [p, w2 ]i
= −hψ1 , Vx · ψ2 i.

In the previous section we equipped S(E) with a connection, namely the

lift of some connection on PSO (E). If ω denotes the connection 1-form for the
connection on PSO (E), the connection 1-form of the lifted connection can be
constructed as follows: e := (dΛ)−1 ◦ Φ∗ ω . This is a spin(n)-valued 1-form, and
ω
to see that it is a connection form, we only have to check that the two axioms
are satised. The rst one:

eg∗ ω
σ e = (dΛ)−1 σg∗ Φ∗ ω = (dΛ)−1 Φ∗ σΛ(g)
∗
ω
= (dΛ)−1 Φ∗ Ad(Λ(g −1 )) ◦ ω = (dΛ)−1 ◦ Ad(Λ(g −1 ))(Φ∗ ω)
= Ad(g −1 ) ◦ (dΛ)−1 Φ∗ ω = Ad(g −1 ) ◦ ω
e.

For the second one let p ∈ PSpin (E) and recall σ

ep : Spin(n) −→ PSpin (E) given
by g 7−→ p · g and note that Φ◦σ ep = σΦ(p) ◦ Λ, where σΦ(p) is the similar map
on the bundle PSO (E). Then:

σp )∗ = (dΛ)−1 ◦ (Φ∗ ω)p ◦ de

ep ◦ (e
ω σp = (dΛ)−1 ωΦ(p) ◦ dΦ ◦ de
σp
= (dΛ)−1 ωΦ(p) ◦ dσΦ(p) ◦ dΛ = (dΛ)−1 ◦ dΛ = idspin(n) .

This is the connection 1-form of the lifted connection, since one can easily check
that ker(e
ω )p = Hp PSpin (E).
There is a standard procedure for transforming connections on principal bun-
dles to connections on the associated bundles. In general let π : P −→ M be
a principal G-bundle, ρ : G −→ Aut(V ) a nite-dimensional representation of
G on V and E := P ×ρ V the associated vector bundle. Recall that there is a
1-1 correspondence between sections of E and functions f : P −→ V satisfying
f (p · g) = ρ(g −1 )f (p) (the so-called equivariant functions ). If ψ ∈ Γ(E) we write
ψb for the associated equivariant function.
The map ∇ : X(M ) × Γ(E) −→ Γ(E) given by (X, ψ) 7−→ ∇X ψ where ∇X ψ

is the section of E corresponding to the equivariant function p 7−→ X p (ψ) b (here

X is the unique lift of X to a horizontal vector eld on P ) denes a connection
on E . In short
∇[ X ψ(p) = X p (ψ).
b (2.12)

We also have a local description of the situation. Given a set of trivializations

(Uα , Φα )α∈I and transition functions gαβ : Uαβ −→ GL(n, K) there is a 1-1
correspondence between smooth sections of E and collections (ψα )α∈I of smooth
functions ψα : Uα −→ Rn satisfying ψα (x) = gαβ (x)ψβ (x) for x ∈ Uαβ : Given a
2.6 The Dirac Operator 43

section ψ ∈ Γ(E), ψα : Uα −→ Rn is the unique function satisfying Φα ◦ ψ(x) =

(x, ψα (x)), i.e.
ψα (x) = pr2 ◦Φα (ψ(x)).
So the question arises: how does the induced connection on E look locally?
Well, given a local section sα : Uα −→ P of the principal bundle, the local
function of ∇X ψ relative to the corresponding trivialization of E is given by
(see L. Claessens Field Theory from a Bundle Point of View, Section 6.2, in
?
[ ] Section 6.2)

(∇X ψ)α (x) = Xx ψα − ρ∗ (Aα (Xx ))ψα (x) (2.13)

where ρ∗ : g −→ End(V ) is the induced Lie algebra representation of ρ and

Aα = s∗α ω is the so-called local gauge potential .
In the case of the Levi-Civita connection on E = T M , the formula reads
X 
(∇X Y )α (x) = Xx Yα − Aα i (Xx )Bi Yα (x) (2.14)
i∈I

where (Bi )i∈I is some basis for the Lie algebra so(n) where m = dim M .
Getting back to the spin bundle case, the connection ω
e onPSpin (E) induces
a connection ∇
e S(E). Let's try to unveil (2.13) in this
on the spinor bundle
particular setting. Lettα : Uα −→ PSpin (E) be a local section of the spin
bundle and put sα := Φ ◦ tα . This is a local section of the frame bundle PSO (E).
Let
eα := t∗α ω
A e and Aα := s∗α ω
denote the local gauge potentials of the connection ω, resp. the connection ω
e.
Then we have

eα = t∗α ω
A e = t∗α ((dΛ)−1 ◦ Φ∗ ω) = (dΛ)−1 ◦ t∗α Φ∗ ω
= (dΛ)−1 ◦ s∗α ω = (dΛ)−1 ◦ Aα ,

so the gauge potentials are related in the nicest possible way.

Putting this into (2.13) we obtain

(∇ eα (Xx ))ψα (x)

e X ψ)α (x) = Xx ψα − (κn )∗ (A
= Xx ψα − ρn ((dΛ)−1 (Aα (Xx )))ψα (x)

for x ∈ Uα (recall that the action of (κn )∗ is just ρn itself ). Now we pick the
usual basis (Bij )i<j for so(n) (where Bij is the n × n-matrix whose ij 'th entry
is −1, the ji'th entry is 1 and all other entries are 0) and write the so(n)-valued
α α α
P
1-form A in terms of this basis: A = i<j Aij Bij . Then (remembering that
dΛ : spin(n) −→ so(n) maps ei ej to 2Bij , where (ei ) is the standard basis for
Rn ) we get
 X 
−1 α
(∇X ψ)α (x) = Xx ψα − ρn (dΛ)
e Aij (Xx )Bij ψα (x)
i<j
1 X 
= X x ψ α − ρn Aα
ij (Xx )ei ej ψα (x)
2 i<j
1X α
= X x ψα − A (Xx )ei ej · ψα (x). (2.15)
2 i<j ij

Now, let us show compatibility of the connection ∇

e with the ber metric. We
will use the local expression above. Locally ψ(x) = [sα (x), ψα (x)] (for some
44 Chapter 2  Spin Groups

section sα : Uα −→ PSpin (E) dened around x) we get by denition of the ber

metric
hψ(x), ψ 0 (x)i = hψα (x), ψα0 (x)i∆n
for x ∈ Uα . Thus

e X ψ(x), ψ 0 (x)i = Xx ψα − 1
D X E
0
h∇ Aα
ij (Xx )ei ej · ψα (x) , ψα (x)
2 i<j ∆n

1X α
= hXx ψ, ψ 0 (x)i∆n − A (Xx )hei ej · ψα (x), ψα0 (x)i∆n .
2 i<j ij

Note that

hei ej · ψα (x), ψα0 (x)i∆n = −hej · ψα (x), ei · ψα0 (x)i∆n = hψα (x), ej ei · ψα0 (x)i∆n
= −hψα (x), ei ej · ψα0 (x)i∆n ,

and therefore

e X ψ(x), ψ 0 (x)i + hψ(x), ∇

h∇ e X ψ 0 (x)i = hXx ψ, ψ 0 (x)i∆ + hψ(x), Xx ψ 0 i∆ .
n n

As mentioned Xx ψα should be interpreted componentwise, i.e. pick a complex

n
basis {v1 , . . . , vN } for ∆n (of course N = 2b 2 c ) and write
N
X
ψα = ψα,i vi
i=1

where ψα,i , the i'th component of ψα , is a complex-valued function on Uα , then

N
X
Xx ψα = (Xx ψα,i )vi ∈ ∆n .
i=1

Since we have a metric in play, it would be wise of us to assume the basis

{v1 , . . . , vN } to be orthonormal. Then

N
X N
X
hXx ψα , ψα0 (x)i∆n + hψα (x), Xx ψα0 i∆n = 0
(Xx ψα,i )ψα,i (x) + 0
ψα,i (x)Xx ψα,i
i=1 i=1
N
X 
0
= Xx ψα,i ψα,i = Xx hψα , ψα0 i∆n .
i=1

Thus we have proved that the spin connection is compatible with the metric.
Finally we need to check condition 2 in the denition of a Dirac bundle, that
is
∇
e X (Y · ψ)(x) = (∇X Y )(x) + Yx · (∇
e X ψ(x)) (2.16)

for each x ∈ M. Again we use the local expressions, i.e. we consider a cover
(Uα ) which are domains of trivializations of both TM and E. Let Φα denote
the trivializations of the tangent bundle (we may assume it to preserve the
∼
metric on M, i.e. Φx : Tx M −−→ Rm is an isometry). Since ∇
eX is C-linear and
satises the Leibniz rule, it is sucient to verify the above condition for Y = Ek
−1
where Ek (x) = Φα (x, ek ) are local orthonormal vector elds.
But rst, recall the following formula for the dierential of the double covering

dΛ(X)v = Xv − vX
2.6 The Dirac Operator 45

X ∈ spin(n) and v ∈ Rn ⊆ Cl0,n and replace in that X by i<j Aα

P
for ij (Xx )ei ej ∈
spin(n) and v by ek to get
X  X  X 
Aαij (X )e e
x i j ek = ek Aα
ij (X )e e
x i j + dΛ Aα
ij (X x i j ek
)e e
i<j i<j i<j
X  X
= ek Aα
ij (Xx )ei ej + 2 Aα
ij (Xx )Bij ek . (2.17)
i<j i<j

Note that concatenation here means multiplication inside the Cliord algebra,
and not the Cliord action. Recall also formula (2.14) for the local form of
the Levi-Civita connection. It will be used in the following calculations (for
explanations see below):

1 X 
(∇
e X (Ek · ψ))α (x) = Xx (ek · ψα ) − Aα
ij (Xx )ei ej ek · ψα (x)
2 i<j
1 X 
= ek · (Xx ψα ) − ek Aα
ij (Xx )ei ej · ψα (x)
2 i<j
X
− (Aα
ij (Xx )Bij ek ) · ψα (x)
i<j

= ek · (∇
e X ψ)α (x) + (∇X Ek )α (x) · ψα (x)

and this is precisely the local form of the right-hand side of (2.16). For the rst
identity we used that (Ek · ψ)α = ek · ψα and in the second we used (2.17) as
well as the fact, that ek · is a linear map, and thus commutes with Xx . This
veries condition 2 and hence we have shown that the spinor bundle S(E) is a
Dirac bundle.
The second most important example of a Dirac bundle is the complex spinor
bundle S c (E). We can proceed in almost the same way we did before and we
begin by giving S c (E) a metric: On ∆n we can, as before, nd an inner product
h , i∆n such that κcn (g) is unitary for each g ∈ Spinc (n) and such that ρn (v) is
n
skew-adjoint for any v ∈ R ⊆ Cl0,n . We transfer this inner product to a ber
c
metric on S (E) in the usual way by dening

h[p, v], [p, w]i := hv, wi∆n .

Unitarity of κcn (g) guarantees that this is well-dened.

Thanks to the isomorphism Cl(E) ∼ c
= PSpin (E) ×Adc Cl0,n we can also dene
c c
a Cliord action Cl(E) × S (E) −→ S (E) by

([p, ξ], [p, v]) 7−→ [p, ρn (ξ)v].

Checking condition 1 in the denition of a Dirac bundle is done as above.

c
In the previous section we gave PSpin (E) a connection depending on the choice
0
of connection A on L . The corresponding connection 1-form is given in the same
way as for the spin bundle

e A = (dΛc )−1 ◦ (Ξ∗ ω A )

ω
c
where Ξ : PSpin (E) −→ Q is the double covering and Q is the bundle is as

dened in the previous section. This connection induces a connection eA

∇ on
c
S (E). We are interested in calculating its local expression from (2.13). Note
rst that dΛc : spinc (n) ∼
= spin(n) ⊕ iR −→ so(n) ⊕ iR (we identify the Lie
algebra of U(1) with iR) is simply given by

(t, iθ) 7−→ (dΛ(t), 2iθ). (2.18)

46 Chapter 2  Spin Groups

c c
Consider trivializations of PSpin (E) and Q over Uα , let tα : Uα −→ PSpin (E)
denote the corresponding local section and put sα := Ξ ◦ tα . If

Aα := s∗α ω A and eα := t∗α ω

A eA

denote the local gauge potentials (local connection 1-forms), then we have eα =
A
c −1 α α
(dΛ ) ◦ A . Note also that since A is an so(n) ⊕ iR-valued 1-form, we may
Aα = Aα
split it:
α α
ω ⊕ AA where Aω is the gauge potential for the connection ω on
PSO (E) and AA is the gauge potential for the connection A on L0 . From (2.18)
α

we get
(dΛc )−1 Aα = (dΛ−1 (Aα 1 α
ω ), 2 AA )

and the induced Lie algebra representation of κcn is just ρn restricted to the
1 α
Lie algebra. κn ( AA (Xx )) is just multiplication with the imaginary number
2
1 α −1 α
2 AA (Xx ) and (dΛ) Aω is already known to us from our discussion above.
Hence we get:

eA 1X α 1
(∇ X ψ)α (x) = Xx ψa − (A )ij (Xx )ei ej · ψα (x) + Aα (Xx )ψα (x). (2.19)
2 i<j ω 2 A

With this local expression at our disposal we can show that the requirements
of Denition 2.42 are satised, and hence that S c (E) is a Dirac bundle. The
arguments are identical to the ones above for the spin bundle and so we skip
them.

Denition 2.43 (Dirac Operator). Let S be a Dirac bundle. The Dirac

operator ∂/ is then dened as the composition
∇ ∼
Γ(S) −−→ Γ(T ∗ M ⊗ S) −−→ Γ(T M ⊗ S) −−→ Γ(S)
e
(2.20)

where the last map is Cliord multiplication X(M ) ⊗ Γ(S) −→ Γ(S).

Next follow three elementary facts about general Dirac operators. Proofs are
not included here, they can be found in any standard treatment of the subject,
?
for instance in [ ] Part II.5.

Lemma 2.44. The Dirac operator is a rst order dierential operator, and
given a local orthonormal frame {E1 , . . . , Em } for T M over U , the Dirac oper-
ator takes the local form
m
X
/
∂ψ|U = Ej · (∇
e E ψ).
j
(2.21)
j=1

As a dierential operator, we can compute its symbol:

Proposition 2.45. For ξx ∈ Tx∗ M , let ξx] · : Sx −→ Sx be Cliord multiplication

with the metric dual ξx] of ξx . Then
/ x ) = iξx] ·
σ(∂)(ξ and 2
σ(∂/ )(ξx ) = kξx k2 .

Thus both ∂/ and ∂/2 (called the Dirac Laplacian) are elliptic.
Proposition 2.46. The Dirac operator is formally self-adjoint, i.e. for ψ1 and
ψ2 in Γc (S) (the set of sections of S with compact support ) we have
/ 1 |ψ2 ) = (ψ1 |∂ψ
(∂ψ / 2 ). (2.22)

We augment this list of elementary properties of the Dirac operator with a

perhaps less renowned result. The proof may be found in [ ] Theorem 8.2: ?
2.6 The Dirac Operator 47

Theorem 2.47 (Unique Continuation Property). If ψ ∈ Γ(S) is in the

kernel of the Dirac operator ∂/ and ψ is 0 on some open set, then ψ is identically
equal to 0.
In the next example we present two of the most important Dirac operators.

Example 2.48. (Spin-Dirac operator). We consider the spinor bundle S(E)

associated to some oriented Riemannian vector bundle E carrying a spin struc-
ture. We saw earlier that this vector bundle is a Dirac bundle and thus it carries
a Dirac operator, /
D called the spin-Dirac operator or just the Dirac opera-
?
tor (in [ ] it is called the Atiyah-Singer operator ). Thanks to Lemma 2.44 and
the calculations done in the previous example, we arrive at the following local
description

m
X  m
X
/ ψ)α (x) =
(D Ek · ∇
eE ψ
k
(x) = ek · (∇
e E ψ)α (x)
k
α
k=1 k=1
m n
X  1X 
= ek · (Ek )x ψα − Aα
ij ((Ek )x )ei ej · ψα (x)
2 i<j
k=1
m
X 1X α
= ek · (Ek )x ψα − A ((Ek )x )ek ei ej · ψα (x) (2.23)
2 i<j ij
k=1

where {E1 , . . . , Em } is a local tangent frame (m = dim M ) such that Ek (x) =

Φ−1 4
α (x, ek ) .
( Geometric Dirac operator). The Dirac operator associated to a complex
spinor bundle of a principal Spinc (n)-bundle is called the geometric Dirac op-
erator. Inserting the local expression (2.19) of the connection into (2.21) we
get

m 
X 1X α
/ A ψ)α (x) =
(D ek · (Ek )x ψα − (A )ij ((Ek )x )ek ei ej · ψα
2 i<j ω
k=1
1 
+ Aα ((E k )x )ek · ψ α (x) . (2.24)
2 A
where A is a choice of a connection on the determinant line bundle, and where
m is the dimension of the base manifold. If we replace the connection A on the
determinant line bundle by A + β for some β ∈ iΩ1 (M ) the local connection
1-forms change to from Aα α
A to AA + β and from the local expression for the
Dirac operator we immediately deduce

1
/ A+β ψ = D
D / A ψ + β · ψ. (2.25)
2
We will make extensively use of this result.

In the Cliord algebra ClC

n we have the volume form given unambiguously by
b n+1
2 c
ωC = i e1 · · · en whenever {e1 , . . . , en } is an orthonormal basis for Rn . In
the same manner we may dene a volume section ωC ) of the com-
(also denoted

plexied Cliord bundle Cl(M )⊗C by the local formula ωC |U = ib 2 c E1 · · · Em

m+1

when (E1 , . . . , Em ) is a local orthonormal tangent frame over U .

k
Assume now that m = 2k , then ωC = i E1 · · · E2k . In this dimension it is well-
C 0
known that the volume form commutes with everything in (Clm ) - the even
part of the Cliord algebra. For a Dirac bundle S over E , ∇ e the connection and
ψ a section of S , we get since Xx (ωC · ψ)α = Xx (ωC · ψα ) = ωC · (Xx ψα ) (in the
4 Note that often (Ek )x ψα is written as ∂ψα
(x) .
∂ek
48 Chapter 2  Spin Groups

rst expression ωC denotes the volume section and in the two last expressions
it denotes the volume form ) that
1X α
[∇
e X (ωC · ψ)]α (x) = Xx (ωC · ψ)α − A (Xx )ei ej · (ωC · ψα (x))
2 i<j ij
1X α
= ωC · (Xx ψα ) − ωC · A (Xx )ei ej · ψα (x)
2 i<j ij

= ωC · (∇
e X ψ)α (x).

If E S(E) is the associated spinor bundle, then we

has a spin structure and
−
have the splittingS = S + ⊕S − induced by the decomposition ∆2k = ∆+ 2k ⊕∆2k .
±
The subspaces ∆2k are the ±1-eigenspaces of the action of ωC on ∆2k . Therefore,
±
sections of S(E) are exactly the spinor elds satisfying ωC · ψ = ±ψ . Therefore
the formula above implies that ∇ e X maps Γ(S ± ) −→ Γ(S ± ). To see how the
Dirac operator reacts to this splitting, note the following:

m
X m
X
/ C · ψ) =
D(ω Ei · ∇
e E (ωC · ψ) =
i
Ei · ω C · ∇
e E (ψ)
i
i=1 i=1
m
X
= −ωC · Ei · ∇ /
e E (ψ) = −ωC · Dψ
i
i=1

(m / maps Γ(S(E)± ) −→ Γ(S(E)∓ ).

= dim M ). From this it is apparent, that D
±
/ denotes the restriction of D
If D / to Γ(S(E)± ) we may write the Dirac operator
relative to the splitting as a matrix

!
−
0 /
D
/=
D + .
/
D 0
c
Exactly the same holds true for the geometric Dirac operator. Also the spin -
c c
representation κ2k : Spin (2k) −→ Aut(∆2k ) decomposes into irreducible rep-
+ −
resentation spaces ∆2k = ∆2k ⊕ ∆2k and hence also the complex spinor bundle
S (E) exhibits a splitting S (E) = S c (E)+ ⊕ S c (E)− , relative to which the
c c

geometric Dirac operator /A

D takes the form

!
−
0 /A
D
/A =
D +
/A
D 0

±
where / A : Γ(S c (E)± ) −→ Γ(S c (E)∓ ).
D