Spin 2010 (jmf) 19

Lecture 3: Spinor representations

Yes now I’ve met me another spinor...
— Suzanne Vega (with apologies)

It was Élie Cartan, in his study of representations of simple Lie algebras, who came across repres-
entations of the orthogonal Lie algebra which were not tensorial; that is, not contained in any tensor
product of the fundamental (vector) representation. These are the so-called spinorial representations.
His description [Car38] of the spinorial representations was quite complicated (“fantastic” according
to Dieudonné’s review of Chevalley’s book below) and it was Brauer and Weyl [BW35] who in 1935 de-
scribed these representations in terms of Clifford algebras. This point of view was further explored in
Chevalley’s book [Che54] which is close to the modern treatment. This lecture is devoted to the Pin and
Spin groups and to a discussion of their (s)pinorial representations.

3.1 The orthogonal group and its Lie algebra

Throughout this lecture we will let (V, Q) be a real finite-dimensional quadratic vector space with Q
nondegenerate. We will drop explicit mention of Q, whence the Clifford algebra shall be denoted C�(V)
and similarly for other objects which depend on Q. We will let B denote the bilinear form defining Q.
We start by defining the group O(V) of orthogonal transformations of V:

(51) O(V) = {a : V → V|Q(av) = Q(v) ∀v ∈ V} .

We write O(Rs,t ) = O(s, t ) and O(n) for O(n, 0). If a ∈ O(V), then det a = ±1. Those a ∈ O(V) with det a =
1 define the special orthogonal group SO(V). If V is either positive- or negative-definite then SO(V)
is connected: otherwise it has two connected components. This can be inferred by the fact that the
connectedness of a Lie group is controlled by that of its maximal compact subgroup, which in the case
of SO(s, t ), for s, t > 0, is

S (O(s) × O(t )) = {(a, b) ∈ O(s) × O(t )|det a = det b} ,

which has two connected components. The Lie algebra so(V) of SO(V) is defined by

(52) so(V) = {X : V → V|B(Xu, v) = −B(u, Xv) ∀u, v ∈ V} .

As a vector space, so(V) ∼

= Λ2 V, where the skewsymmetric endomorphism u � v ∈ so(V) corresponding
2
to u ∧ v ∈ Λ V is defined by

(53) (u � v)(x) = B(u, x)v − B(v, x)u .

It is easy to check that u � v ∈ so(V) as it is to compute the commutator

(54) [u � v, x � y] = B(u, x)v � y − B(u, y)v � x − B(v, x)u � y + B(v, y)u � x .

The Clifford algebra C�(V) being associative, becomes a Lie algebra under the commutator and contains
so(V) as a Lie subalgebra via the embedding

(55) ρ : so(V) → C�(V) where ρ(u � v) = 14 (uv − vu) .

Indeed, it is a simple calculation using the Clifford relation uv = −vu − 2B(u, v)1 to show that

(56) [ρ(u � v), x] = B(u, x)v − B(v, x)u = (u � v)(x) ,

and hence that

(57) [ρ(u � v), ρ(x � y)] = B(u, x)ρ(v � y) − B(u, y)ρ(v � x) − B(v, x)ρ(u � y) + B(v, y)ρ(u � x) ,

whence ρ is an injective Lie algebra homomorphism.

Spin 2010 (jmf) 20

Exponentiating so(V) in End(V) generates the identity component SO0 (V) of SO(V), whereas expo-
nentiating ρ(so(V)) in C�(V) generates a covering group of SO0 (V). We will see this in full generality
below, but let us motivate this with an example. Suppose that V contains a positive-definite plane with
orthonormal basis e 1 , e 2 . Then relative to this basis, the restriction to this plane of e 1 � e 2 ∈ so(V) has
matrix
� �
0 −1
(58)
1 0

whose exponential is
� �
cos θ − sin θ
(59) a(θ) = exp (θ(e 1 � e 2 )) = ,
sin θ cos θ

whence, in particular, a(2π) is the identity matrix. On the other hand, exponentiating the image of the
same Lie algebra element ρ(e 1 � e 2 ) = 21 e 1 e 2 in C�(V) we obtain

(60) b(θ) = exp( 12 θe 1 e 2 ) = cos( 12 θ)1 + sin( 12 θ)e 1 e 2 ,

using that (e 1 e 2 )2 = −1. In particular we see that b(2π) = −1, so that the periodicity of b(θ) is 4π. In
other words, it suggests that the Lie group generated by exponentiating so(V) in C�(V) is a double cover
of SO0 (V). We will see that this is indeed the case.

3.2 Pin and Spin

Definition 3.1. The Pin group Pin(V) of (V, Q) is the subgroup of (the group of units of ) C�(V) generated
by v ∈ V with Q(v) = ±1. In other words, every element of Pin(V) is of the form u 1 · · · u r where u i ∈ V and
Q(u i ) = ±1. We will write Pin(s, t ) for Pin(Rs,t ) and Pin(n) for Pin(n, 0).

Let v ∈ V ⊂ C�(V) and let Q(v) �= 0. Then v is invertible in C�(V) and v −1 = −v/Q(v). We define, by
analogy with the case of a Lie group, the adjoint action Adv : V → V, by

−1 −1 B(x, v)
(61) Adv (x) = v xv −1 = v xv = (−xv − 2B(x, v)1) v = −x + 2 v = −R v x ,
Q(v) Q(v) Q(v)

where R v stands for the reflection on the hyperplane perpendicular to v and x ∈ V. We can extend
this to a group homomorphism from the Pin group: Adv 1 ···v p = Adv 1 ◦ · · · ◦ Adv p . Since we would prefer
not to see the sign on the right-hand side of Adv (x), we define the twisted adjoint action by Ad � v (x) =
−1 � −1
(−v)xv = R v x or more generally Ada = axa
� for a an element of the Pin group and a �→ a� the grading
automorphism of C�(V), which is induced by the orthogonal transformation v �→ −v. Let a = u 1 · · · u r ∈
� a = Ru ◦ · · · ◦ Ru . Since reflections are orthogonal transformations, Ad
Pin(V), then Ad � defines a group
1 r
� �
homomorphism Ad : Pin(V) → O(V). It follows from the following classic result that Ad is surjective.

Theorem 3.2 (Cartan–Dieudonné). Every g ∈ O(V) is the product of a finite number of reflections g =
Ru1 ◦ · · · ◦ Rur along non-null lines (Q(u i ) �= 0) and moreover r ≤ dim V.

We will now determine the kernel of Ad. � Let a ∈ Pin(V) be in the kernel of Ad. � This means that
av
� = v a for all v ∈ V. Let us break up a = a 0 + a 1 with a 0 ∈ C�(V)0 and a 1 ∈ C�(V)1 , whence a� = a 0 − a 1 .
� if and only if the following pair of equations are satisfied for all v ∈ V:
Therefore a ∈ ker Ad

(62) a0 v = v a0 and a 1 v = −v a 1 .

Suppose that v ∈ V with Q(v) �= 0 and consider a 0 = α + vβ, where α and β do not involve v. Since
α ∈ C�(V)0 and does not involve v, then vα = αv, whereas since β ∈ C�(V)1 and does not involve v, then
vβ = −βv. The first equation in (62) says that β = 0, whence a 0 = α does not involve v. Repeating this
argument for all the elements of an orthonormal basis (e i ) for V, we see that a 0 does not involve any of
the e i and hence must be a multiple of the identity: a 0 = α1 for some α ∈ R. Similarly, write a 1 = γ + vδ,
where γ, δ do not involve v. Now we have that γv = −vγ, whereas δv = vδ. The second equation in (62)
Spin 2010 (jmf) 21

says that δ = 0, whence a 1 = γ does not involve v. Repeating this argument for the basis (e i ), we see that
a 1 does not involve any of the e i and hence must be a multiple of the identity, but a 1 ∈ C�(V)1 whereas
1 ∈ C�(V)0 , whence a 1 = 0. Hence all elements of Pin(V) in the kernel of Ad � are multiples of the identity.
Now let u 1 · · · u p = α1 for Q(u i ) = ±1. Let us compute the norm of this element using the Clifford inner
product (41), to arrive at

(63) (α1, α1) = α2 (1, 1) = (u 1 · · · u p , u 1 · · · u p ) = (1, (−u p ) . . . (−u 1 )u 1 · u p ) = Q(u 1 ) · · · Q(u p )(1, 1) .

Since (1, 1) �= 0 and Q(u i ) = ±1, it follows that α2 = ±1. Since α ∈ R the only solutions to this equation are
α = ±1 and hence ker Ad � = {±1}. In summary we have proved

Proposition 3.3. The following sequence is exact:

�
Ad
1 � {±1} � Pin(V) � O(V) �1.

Exact sequences

A sequence of groups and group homomorphisms

i p
1 �A �B �C �1

is said to be exact if the kernel of each homomorphism is the image of the preceding one. In
the above diagram, 1 denotes the one-element group. This is both an initial and final object in
the category of groups, since there is only one homomorphism into it (sending all elements to
the identity) and only one homomorphism out of it (sending the identity to the identity). This
explains why we have not given names to the homomorphisms 1 → A and C → 1. Exactness
at A means that i : A → B is injective, since its kernel is the image of 1 → A, whence consists
only of the identity. Similarly, exactness at C says that p : B → C is surjective, since the kernel
of C → 1 is all of C, and that is precisely the image of p. Finally, exactness at B says that the
kernel of p : B → C is precisely the image of i : A → B. Such an exact sequence says that B is an
extension of C by A.

Finally, let us define the spin group.

Definition 3.4. The spin group of (V, Q) is the intersection

Spin(V) = Pin(V) ∩ C�(V)0 .

Equivalently, it consists of elements u 1 · · · u 2p , where u i ∈ V and Q(u i ) = ±1. We will write Spin(s, t ) for
Spin(Rs,t ) and Spin(n) for Spin(n, 0).
� a = 1 for a ∈ Pin(V) if
Since for a reflection Ru ∈ O(V), we have that det Ru = −1, it follows that det Ad
�
and only if a ∈ Spin(V). Since the kernel of Ad belongs to Spin(V), we immediately have the following

Proposition 3.5. The following sequence is exact:

�
Ad
1 � {±1} � Spin(V) � SO(V) �1.

� : Spin(V) → SO(V) is a nontrivial cov-

For V of signature (s, t ) with at least one of s, t ≥ 2, the map Ad
ering. This can be shown by exhibiting a continuous path between 1 and −1 in Spin(V). Let e 1 , e 2 be an
orthonormal basis for a positive- or negative-definite plane. That such a plane exists is a consequence
of our assumption on the signature of V. Then consider the following continuous (in fact, analytic)
curve in Spin(V):

a(t ) = (e 1 cos t + e 2 sin t )(e 2 sin t − e 1 cos t ) = Q(e 1 ) cos(2t )1 + sin(2t )e 1 e 2 .

Spin 2010 (jmf) 22

We see that a(0) = Q(e 1 )1, whereas a(π/2) = −Q(e 1 )1, whence it joins 1 to −1.
Finally let us remark that for V either positive- or negative-definite, SO(V) and hence Spin(V) is
connected, whereas for indefinite V, Spin(V) has two connected components. Let Spin0 (V) denote the
identity component. In definite or lorentzian signatures, Spin0 (V) → SO0 (V) is a universal covering, but
Spin0 (s, t ) is not simply connected when both s, t ≥ 2. The simplest interesting examples of spin covers
are SU(2) → SO(3) and SL(2, C) → SO0 (3, 1).

3.3 Pinors and spinors

Informally, pinors (resp. spinors) are vectors in an irreducible representation of a Clifford algebra (resp.
its even subalgebra) and, by restriction, of the corresponding Pin (resp. Spin) group. In order to define
them properly we need to introduce some notation.

Definition 3.6. Let A be a real associative algebra and let K = R, C or H. By a K-representation of A we

mean an R-linear homomorphism
ρ : A → EndK (E)
for some K-vector space E. Two K-representations ρ : A → EndK (E) and ρ� : A → EndK (E� ) are equivalent
if there is a K-linear isomorphism f : E → E� such that the triangle

� A ��� �
ρ ��� �� ρ
�� ��
� ��
��� Ad f �
EndK (E) � EndK (E� )

where Ad f : EndK (E) → EndK (E� ) is defined by ϕ �→ f ◦ϕ◦ f −1 . In other words, for all a ∈ A, we have that
f ◦ ρ(a) = ρ� (a) ◦ f .

Quaternionic vector spaces

Because H is not commutative, one must distinguish between left and right quaternionic vec-
tor spaces. This is largely a matter of convention, since quaternionic conjugation relates left
and right vector spaces. Throughout these lectures we shall adopt the convention that Hn is a
right quaternionic vector space. In this way, the matrix algebra H(n) can act H-linearly on Hn
from the left. (The fact that left and right multiplication commute is precisely associativity.)
This defines an isomorphism of real algebras H(n) ∼ = EndH (Hn ). In fact, notice that EndH (E)
for a quaternionic vector space E is only a real algebra! This is because of the nonexistence of
H-bilinears.

In Section 1.4.4 we have already seen one example of an R-representation of C�(V), namely ΛV. This
representation is not irreducible, however.

Definition 3.7. A pinor representation of Pin(V) is the restriction of an irreducible representation of

C�(V). Similarly, a spinor representation of Spin(V) is the restriction of an irreducible representation of
C�(V)0 . (It is not hard to see that both pinor and spinor representations are irreducible.)

The irreducible representations of C�(V) are easy to determine from the classification of real Clifford
algebras in the second lecture. Recall that as a real algebra, C�(s, t ) is isomorphic to either K(2n ) or
K(2n ) ⊕ K(2n ) depending on the signature. The following result can be extracted from [Lan84, § XVII].

Theorem 3.8. 1. Every irreducible R-representation of the real algebra R(n) is isomorphic to Rn , where
the matrix A ∈ R(n) acts via left matrix multiplication.

2. Every irreducible H-representation of the real algebra H(n) is isomorphic to Hn as a right H-vector
space and where A ∈ H(n) acts via left matrix multiplication.
Spin 2010 (jmf) 23

3. Every irreducible C-representation of the real algebra C(n) is isomorphic either to Cn with the nat-
ural action given by left matrix multiplication by A ∈ C(n) or to Cn with the complex conjugate
action given by left matrix multiplication by A ∈ C(n).

This result together with the classification of real Clifford algebras, allows us to determine the pinor
representations easily. First of all, we notice that because of the third isomorphism in (49), the type of
the Clifford algebra does not change, only the dimension does, when we moved diagonally in the table.
This means that the type of the representation of C�(s, t ) only depends on s − t and, moreover, because
of Bott periodicity, only on s − t (mod 8). Thus we need only remember one small part of the Clifford
chessboard to determine the rest:

C(2)
R(2)
R⊕R
R C H H ⊕ H H(2)
−−−−−−−→
s

Notice that if we colour the squares of the chessboard according to whether C�(s, t ) has one or two
inequivalent irreducible representations, then we do indeed end up with a chessboard pattern.

2 1 2 1 2
1 2 1 2 1
2 1 2 1 2
1 2 1 2 1
−−−−−−−→
s

This dichotomy can also be explained by means of the volume element of C�(V). Given an ordered
orthonormal basis (e 1+ , . . . , e s+ , e 1− , . . . , e t− ) for Rs,t , with Q(e i± ) = ±1, there is associated a volume element
of C�(s, t ) defined as the Clifford product ω = e 1+ · · · e s+ e 1− · · · e t− .

Lemma 3.9. The volume element ω ∈ C�(s, t ) satisfies the following properties:

1. ω2 = (−1)s+d (d −1)/2 1, where d = s + t ,

2. ω is central if s + t is odd, and

3. ωv = −vω for all v ∈ V, if s + t is even.

It follows from the first part that the sign of ω2 depends only on s − t (mod 4):

− − + +
�
1, s − t = 0, 3 (mod 4) − + + −
ω2 =
−1, s − t = 1, 2 (mod 4) + + − −
+ − − +
−−−−−−−→
s

Suppose that s + t (equivalently, s − t ) is odd, so that ω is central. Then if ω2 = 1 there are two
pinor representations P± , distinguished by the action of ω: ω = ±1 on P± . If s − t = 3 (mod 8), P± is
quaternionic, whereas if s − t = 7 (mod 8), P± is real. If ω2 = −1, so that s − t = 1, 5 (mod 8), there
are two complex pinor representations P and P, distinguished by the action of ω: ω = ±i on P and P,
respectively.
Spin 2010 (jmf) 24

In summary, the type and dimension of the pinor representations follows from the classification
theorem 2.7. Similarly, the type and dimension of the spinor representations, being representations of
the even subalgebra C�(V)0 , follow from Corollary 2.11. It remains to understand how the pinor and
spinor representations are related, for which we need a brief scholium about representation theory.

Real, complex and quaternionic representations

Let G be a Lie group, such as Pin(V) or Spin(V). If K = R, C or H, we will denote RepK (G) denote
the (symmetric, monoidal) category of K-representations of G, whose objects are K-vector
spaces E (with the usual caveat about the case K = H) together with group homomorphisms
ρ : G → GLK (E) and where a morphism between ρ : G → GLK (E) and ρ� : G → GLK (E� ) is a
K-linear map f : E → E� such that for all g ∈ G, f ◦ ρ(g ) = ρ� (g ) ◦ f . There are a number of
functors relating these categories, which commute with the direct sum of representations,
which is the categorical coproduct in RepK (G). These functors are neatly summarised in the
following (noncommutative!) diagram, borrowed from [Ada69] via [BtD85]:

(64) RepC (G)

C
eR
����� � ���� H
���r C
��� ���
��� �
RepR (G) c RepH (G)
���� �
��� �� ���
��� �� H
r RC � � ��� e C
RepC (G)

where c takes a complex representation E to its complex conjugate c(E) = E, e RC and e CH are
extension of scalars, taking a real representation E to its complexification e RC (E) = E ⊗R C and a
complex representation E to its quaternionification e CH (E) = E ⊗C H, and where r RC and r CH are
restriction of scalars, so that we simply view a complex representation E as a real represent-
ation r RC (E) and a quaternionic representation E as a complex representation r CH (E). These
functors satisfy a number of identities:

c2 = 1 r RC e RC = 2 cr CH = r CH
(65) ce RC = e RC e RC r RC = 1 + c e CH r CH = 2
e CH c = e CH r RC c = r RC r CH e CH = 1 + c

where 2E = E ⊕ E, (1 + c)E = E ⊕ E, etc.

In the following discussion, d = s + t is the (real) dimension of V. We will let P and S, perhaps with
decorations, denote pinor and spinor representations, respectively; although in order to compare them
we must view them both as representations of Spin(V).
If d is even, then the volume element ω ∈ C�(s, t )0 and commutes with C�(s, t )0 , whence its eigen-
spaces in the pinor representation will correspond to the spinors representations. By contrast, if d is
odd, then ω �∈ C�(s, t )0 and hence C�(s, t ) = C�(s, t )0 ⊕ C�(s, t )0 ω ∼= C�(s, t ) ⊗R R[ω]. This means that we
will be able to induce a pinor representation of C�(s, t ) from a spinor representation S of C�(s, t )0 essen-
tially by tensoring with R[ω]: P = C�(s, t ) ⊗C�(s,t )0 S. If s − t = 1, 5 (mod 8) then ω2 = −1 so that R[ω] ∼= C,
whereas if s − t = 3, 7 (mod 8) then ω2 = 1 so that R[ω] ∼ = R ⊕ R. We will use these facts freely in what
follows.

3.3.1 s − t = 0 (mod 8)
d /2 (d −2)/2
Here P ∼
= R2 and S ± ∼= R2 as vector spaces. The volume element obeys ω2 = 1, whence S ± are the
eigenspaces of ω with eigenvalues ±1 and P = S + ⊕ S − .
Spin 2010 (jmf) 25

3.3.2 s − t = 1 (mod 8)
(d −1)/2 (d −1)/2
Here P ∼ = C2 and S ∼
= R2 as vector spaces. The Clifford algebra C�(s, t ) ∼
= C�(s, t )0 ⊗ C, whence
∼ C ∼
P = e R (S). It follows that P = P as representations of Spin(s, t ).

3.3.3 s − t = 2 (mod 8)
(d −2)/2 (d −2)/2
Here P ∼
= H2 and S, S ∼
= C2 as vector spaces. We have that P ∼
= e CH (S), whence r CH (P) ∼
= S ⊕ S, the
eigenspace decomposition under ω, which obeys ω2 = −1.

3.3.4 s − t = 3 (mod 8)
(d −3)/2 (d −3)/2
Here P± ∼
= H2 and S ∼
= H2 as vector spaces. The Clifford algebra C�(s, t ) ∼
= C�(s, t )0 ⊕ C�(s, t )0
and hence P± ∼
= S.

3.3.5 s − t = 4 (mod 8)
(d −2)/2 (d −4)/2
Here P ∼
= H2 and S ± ∼
= H2 as vector spaces. We have that P ∼
= S + ⊕ S − is the eigenspace decom-
2
position of ω, which obeys ω = 1.

3.3.6 s − t = 5 (mod 8)
(d −1)/2 (d −3)/2
Here P, P ∼
= C2 and S ∼
= H2 as vector spaces. The Clifford algebra C�(s, t ) ∼
= C�(s, t )0 ⊗R C and
∼ H ∼
hence P = r C (S). It follows that P = P as representations of Spin(s, t ).

3.3.7 s − t = 6 (mod 8)
d /2 (d −2)/2
Here P ∼
= R2 and S, S ∼
= C2 . Then P ∼
= r RC (S), so that e RC (P) ∼
= S ⊕ S is the eigenspace decomposition
2
of ω, which obeys ω = −1, acting on the complexification of P.

3.3.8 s − t = 7 (mod 8)
(d −1)/2 (d −1)/2
Here P± ∼ = R2 and S ∼
= R2 as vector spaces. The Clifford algebra C�(s, t ) ∼
= C�(s, t )0 ⊕ C�(s, t )0
∼
and hence P± = S.
We can summarise and paraphrase these results by saying that in even dimensions the pinor rep-
resentation (or if s − t = 6 (mod 8), its complexification) decomposes into the direct sum of two equi-
dimensional spinor representations, whereas in odd dimensions, we must distinguish several cases: if
s − t = 3, 7 (mod 8) then each of the two pinor representations is isomorphic to the unique spinor rep-
resentation, whilst if s − t = 1, 5 (mod 8) then the two complex pinor representations are isomorphic
either to the complexification of the unique spinor representation, if s − t = 1 (mod 8), or to the restric-
tion of scalars of the unique quaternionic spinor representation, if s − t = 5 (mod 8).

3.4 Inner products for pinors and spinors

The pinor and spinor representations have inner products which are Spin(V) invariant. In fact, the
precise statement, which can be found together with a complete discussion of this topic in [Har90],
requires us to review the Clifford involutions.
Spin 2010 (jmf) 26

Clifford involutions

There are three natural involutions of the Clifford algebra C�(V):

1. the grading automorphism α �→ α � , which extends the orthogonal transformation v �→

p
−v on V, e.g., u�
1 · · · u p = (−1) u 1 · · up ;
·

2. the check involution α → α̌, which is the antiautomorphism of C�(V) defined by revers-
ing the order of the generators in every monomial, e.g., (u 1 . . . u p )ˇ= u p · · · u 1 ; and

3. the hat involution α �→ α̂, obtained by combining the previous two.

If α ∈ C�(V) comes from Λp V under the isomorphism C�(V) ∼

= ΛV, then α̃, α̌ and α̂ will be ±α
according the following signs:

p mod 4 0 1 2 3
˜ + − + −
ˇ + − − +
ˆ + + − −

Notice that on C�(V)0 , the hat and check involutions agree. This is called the canonical in-
volution of C�(V)0 .

The following theorem can be found in [Har90, Chapter 13].

Theorem 3.10. There exists an inner product 〈−, −〉 on every spinor representation S such that

(66) 〈ax, y〉 = 〈x, â y〉 for all a ∈ C�(V)0 and x, y ∈ S.

There exist inner products ε̂ and ε̌ on the pinor representation P (possibly taking the direct sum of the two
irreducible pinor representations when appropriate) such that

(67) ε̌(ax, y) = ε̌(x, ǎ y) and ε̂(ax, y) = ε̂(x, â y).

Moreover all seven types of inner products (real symmetric, real symplectic, complex symmetric, complex
symplectic, complex hermitian, quaternionic hermitian and quaternionic skewhermitian) appear!

The spinor representations (with these Spin(V)-invariant inner products) are behind most of the
isomorphisms between the following low-dimensional Lie groups:

Spin(2) ∼
= U(1) Spin(6) ∼= SU(4) Spin(5, 1)0 ∼
= SL(2, H)
Spin(3) ∼
= Sp(1) Spin(2, 1)0 ∼
= SL(2, R) Spin(2, 2)0 ∼
= SL(2, R) × SL(2, R)
(68)
Spin(4) ∼
= Sp(1) × Sp(1) Spin(3, 1)0 ∼
= SL(2, C) Spin(3, 2)0 ∼
= Sp(4, R)
Spin(5) ∼
= Sp(2) Spin(4, 1)0 ∼
= Sp(1, 1) Spin(4, 2)0 ∼
= SU(2, 2)

In particular, notice the sequence Spin0 (2, 1) ∼

= SL(2, R) , Spin0 (3, 1) ∼
= SL(2, C), Spin0 (5, 1) ∼
= SL(2, H),
which would suggest that Spin0 (9, 1) would be isomorphic to SL(2, O) if the octonions were associative
and such a group could be defined.