GENERAL ARTICLE

Algebra and Geometry of Hamiltons Quaternions

Well, Papa, Can You Multiply Triplets?

Govind S Krishnaswami and Sonakshi Sachdev

Inspired by the relation between the algebra of

complex numbers and plane geometry, William
Rowan Hamilton sought an algebra of triples for
application to three-dimensional geometry. Un-
able to multiply and divide triples, he invented
a non-commutative division algebra of quadru-
(left) Govind Krishnaswami
ples, in what he considered his most significant is on the faculty of the
work, generalizing the real and complex number Chennai Mathematical
systems. We give a motivated introduction to Institute. His research in
quaternions and discuss how they are related to theoretical physics spans
various topics
Pauli matrices, rotations in three dimensions, the
from quantum field theory
three sphere, the group SU(2) and the celebrated and particle physics to fluid,
Hopf fibrations. plasma and non-linear
dynamics.
1. Introduction
(right) Sonakshi Sachdev is
Every morning in the early part of the above-cited month a PhD student at the
[October 1843], on my coming down to breakfast, your Chennai Mathematical
(then) little brother William Edwin, and yourself, used Institute. She has worked on
the dynamics of rigid bodies,
to ask me, Well, Papa, can you multiply triplets?
fluids and plasmas.
Whereto I was always obliged to reply, with a sad shake
of the head: No, I can only add and subtract them.
W R Hamilton in a letter dated August 5, 1865 to his
son A H Hamilton [1].
Imaginary and complex numbers arose in looking for
impossible solutions to polynomial equations such as
x2 + 1 = 0. Even a sceptic might be convinced of their
utility upon seeing the remarkable FerroTartagliaCar-
danoBombelli formula for a real root of cubic equa-
tions such as x3 = 20x + 25, expressed as sums of cube
roots of complex numbers! No such general expression in Keywords
terms of real radicals is possible. Some of the mystery Quaternions, division algebra,
surrounding complex numbers was removed once they rotations, spheres, Hopf map.

RESONANCE June 2016 529

 GENERAL ARTICLE

could be visualized as vectors in the plane, following the

work of Euler, Wessel, Argand and Gauss. Hamilton
entered the scene in 1830 by dening complex numbers
z = x + yi as ordered pairs (or couples) (x, y) of real
numbers. He abstracted rules for adding, multiplying
and dividing ordered pairs obeying the commutative,
1
Commutative means z1 z2 = z2 associative and distributive laws of arithmetic1. Thus
z1, associative means z1 (z2 z3) =
the complex numbers C constitute a eld just like the
(z1 z2) z3 while z1 (z2 + z3) = z1 z2
+ z1 z3 is distributivity.
reals R. The manner in which the geometry of the plane
was encoded in the algebra of complex numbers2 made
2
The parallelogram law for add- a deep impression on Hamilton. The problem of nding
ing vectors is just the sum of an algebra of triples (, , ) to describe the geometry
complex numbers, a rotation cor-
of vectors in three-dimensional (3D) space haunted him
responds to multiplication by a
complex number of unit magni-
for at least fteen years. While he knew how to add
tude, the cosine of the angle and subtract triples, the problem of multiplying them
between vectors is given in terms and dividing by non-zero triples seemed insurmountable.
of the scalar product of the cor- What is more, by analogy with the complex numbers,
responding complex numbers,
he wanted the Euclidean length (square-root of the sum
etc.
of squares of the components) of the product of a pair
3
The cross product of vectors of triples to equal the product of their lengths3. In cur-
had not yet been discovered.
rent language, Hamilton was looking for a real, three-
Besides being non-associative
a (b c) (a b) c, it does
dimensional, normed, associative, division algebra; we
not satisfy the condition | a b| now know that such an algebra does not exist.
= |a||b|. Cross products were
invented by Grassmann within a In 1843, Hamilton found an ingenious way around this
year of Hamilton's discovery of problem, combining two major innovations. The rst
quaternions. was to drop the commutative law for multiplication. He
wrote a triple as t = +i+j and assumed by analogy
with the complex numbers, that i2 = j 2 = 1. It follows
that
t2 = 2 2 2 + 2i + 2j + (ij + ji). (1)
Moreover, he wanted the law of moduli |t|2 = |t2 | to
hold. Now |t|2 = 2 + 2 + 2 would equal |t2| if ij +ji =
4
The quotations in this para-
0, since
graph are from a letter dated
October 17, 1843 from Hamilton (2 2 2 )2 + (2)2 + (2)2 = (2 + 2 + 2 )2. (2)
to J T Graves, which was re-
printed in Vol.xxv, pp.489495 Hamilton was tempted for a moment to fancy that
of [2]. ij = ji = 04 . But this seemed uncomfortable and

530 RESONANCE June 2016

 GENERAL ARTICLE

he settled on the less harsh condition ij = ji = k,

reserving to himself to inquire whether k was 0 or not.
The second innovation, was to admit in some sense a
fourth dimension of space for the purpose of calculating
with triples5. In other words, he assumed that k was 5
In taking this step, Hamilton
linearly independent of 1, i and j. Applying the asso- was partly motivated by specu-
ciative law, he inferred that k 2 = ijij = 1, ki = j lations on how time and space
may be girdled together, as
and jk = i. Hamilton vividly describes his elation upon
well as some vague notions
discovering6 these multiplication rules: They started (based on Kantian philosophy)
into life, or light, full grown, on the 16th of October, of geometry and arithmetic be-
1843, as I was walking with Lady Hamilton to Dublin, ing the sciences of space and
and came up to Brougham Bridge, which my boys have time respectively [3].

since called the Quaternion Bridge. That is to say, I 6

The article by N Mukunda in
then and there felt the galvanic circuit of thought close; this issue of Resonance places
Hamilton's discovery of quater-
and the sparks which fell from it were the fundamental
nions in the wider context of his
equations in i, j, k exactly such as I have used them ever life and work. See also the ar-
since. Extract from a letter dated October 15, 1858 ticle by R Nityananda devoted to
from Hamilton to P G Tait [1]. Hamilton's work in optics.

In more recent language [4, 5], quaternions are quadru-

ples of real numbers q = (, , , ) forming a four-
dimensional real vector space H (named after Hamilton).
In terms of the canonical basis:

Figure 1. Plaque on Broome

Bridge, Dublin, commemorat-
ing Hamiltons discovery of
quaternions.
Photograph by Brian Dolan, In-
genious Ireland
(http://ingeniousireland.ie), re-
printed with permission

RESONANCE June 2016 531

 GENERAL ARTICLE

e = (1, 0, 0, 0), i = (0, 1, 0, 0), j = (0, 0, 1, 0)

and k = (0, 0, 0, 1), (3)

q = e + i + j + k. It is natural to write q = e + u,
where e is called the real or scalar part of q. The
other part u = i + j + k may be regarded as a vector
in three-dimensional space and is called the imaginary
or vectorial part of q. In fact, the word vector rst
appears in Hamiltons writings of 1845.

Quaternions derive Quaternions derive their power from the fact that they
their power from the can be multiplied with each other, so that they form a
fact that real algebra (a vector space where vectors can be mul-
they can be multiplied tiplied to get other vectors). Since any quaternion is a
with each other, so
linear combination of the basis quaternions e, i, j and k,
that they form a real
it suces to specify their products. First, e is taken to
be the multiplicative identity (sometimes denoted 1),
algebra (a vector
so eq = qe = q for any quaternion q. In addition,
space where vectors
i2 = j 2 = k 2 = e, while
can be multiplied to
get other ij = ji = k, jk = kj = i and ki = ik = j. (4)
vectors).
Notice the analogy with cross products of unit vectors
i j = k in 3D Euclidean space.

The formula for the product of two quaternions q =

(, u = i + j + k) and q  = ( , u = i + j + k)
can be computed using the quaternion multiplication
table (Table 1) and written compactly in terms of the
dot and cross products of their vectorial parts:

 e i j k
e e i j k
i i e k j
j j k e i
Table 1. Hamiltons quater-
nion relations.
k k j i e

532 RESONANCE June 2016

 GENERAL ARTICLE

qq  = ( + i2 +  j 2 + k 2 , Two quaternions

( + )i + ( +  )j + ( + )k commute if and
+ (ij) + (ik) + (ji) only if their vector
+ (jk) + (ki) + (kj)) parts are collinear.
= ( u u , u +  u + (  )k
+ (  )j + (  )i)
= ( u u , u +  u + u u ) . (5)

Two quaternions commute qq  = q q i their vector

parts are collinear: u u = 0. Moreover, the two-
sided multiplicative inverse (reciprocal) of q is q 1 =
(2 + u2 )1 (, u). Quaternion multiplication allows
us to dene polynomials in a quaternion variable. Inter-
estingly, a polynomial of the nth degree may have more
than n quaternion roots, see Box 1 for an example.
Conjugation and norm: The conjugate of a complex
number z = x + yi is z = x yi. It allows us to dene
the scalar product of a pair of complex numbers w, z =
(w z ) = ux + vy, where w = u + vi and  denotes the
real part. This leads to the familiar squared-length of
a complex number |z|2 = z, z = (z z) = x2 + y 2.
The reciprocal (left and right multiplicative inverse) of
z is then z/|z|2 . By analogy with the eld of complex
numbers C, the conjugate of a quaternion q = e + i +
j + k is dened as q = e i j k. One checks
that the conjugate of qq  is qq. The scalar product of q
with another quaternion q  = ( , u),

Box 1. Quaternion Roots of Polynomial Equations in One Variable

The number of real roots of a real polynomial of the nth degree (say x2 + 1, x3 + 1 or
x2 1) can be anywhere between 0 and n. The fundamental theorem of algebra guarantees
that a polynomial of degree n with complex coecients has precisely n complex roots
(counting multiplicity). On the other hand, a polynomial of degree n could have more
than n (even innitely many!) quaternion roots. For instance, i, j and k are obvious
roots of q 2 + 1. In fact, since q 2 = (e + u)2 = 2 u u + 2u, every purely imaginary
quaternion q = (0, u) of unit magnitude |q|2 = u u = 1 is a root, and solutions may be
identied with points on the two-dimensional sphere S 2 .

RESONANCE June 2016 533

 GENERAL ARTICLE

7
Though Hamilton speculated q, q  = (q q) =  +uu =  +  +  + , (6)
on how time and space may be
combined, he did not live to see is dened so as to recover the usual Euclidean7 squared-
their synthesis in the Minkowski norm |q|2 = (q q) = q, q = 2 + 2 + 2 + 2.
space-time of special relativity.
Notice however, that the bilinear Law of moduli and Eulers four-square identity:
form (qq) = u . u on the As noted before, Hamilton was led to his quaternion
space of quaternions is the
product by requiring that the norm of the product of two
Lorentzian inner product of spe-
cial relativity.
quaternions is the product of their norms, as is the case
for complex numbers. For a pair of complex numbers,
|zw| = |z||w| follows from the two-square identity:
(x2 + y 2)(u2 + v 2) = (xu yv)2 + (xv + yu)2, (7)
which goes back to Diophantus and Brahmagupta. The
analogous statement |qq | = |q||q | for two quaternions
requires the following identity:
(2 + 2 + 2 + 2)(2 + 2 + 2 + 2) =
(   )2 + (  +  +   )2
+ (  +  +  )2
+ (  +  +  )2 . (8)
Remarkably, Euler had discovered this identity in 1749,
while trying to prove Fermats conjecture (1659) that
every natural number is a sum of four squares. The
identity also appears in an unpublished note of Gauss
from 1819.
2. Quaternions Represented as 2 2 Matrices

Hamilton was led to

The basis quaternions i, j and k anti-commute (ij =
his quaternion product
ji, etc.) and square to e. For those who have met
them, this is reminiscent of the Pauli matrices arising in
by requiring that
the quantum mechanics of spin:
the norm of the      
product of two 0 1 0 i 1 0
1 = , 2 = and 3 = . (9)
quaternions is the 1 0 i 0 0 1
product of their Indeed, the Pauli matrices anti-commute (12 = 2 1
norms. = i3 and cyclic permutations) and square to the iden-
tity matrix I. If we can get the signs right, it seems

534 RESONANCE June 2016

 GENERAL ARTICLE

plausible that the quaternion algebra can be represented

in terms of 2 2 complex matrices.
Let us rst see how this is done for complex numbers,
which may be represented by certain anti-symmetric 2
2 real matrices:
 
x y
z = x + yi  A(z) = = xI y(i2). (10)
y x

The set of matrices of this sort is an algebra: it is closed

under matrix addition and multiplication. In fact, the
map A is an isomorphism of real algebras: A(zz ) =
A(z)A(z ), A(z + z  ) = A(z) + A(z ) and A(z) = A(z)
for R. Remarkably, though multiplication of matri-
ces is generally non-commutative, matrices of the above
sort commute with each other (as (10) involves only one
of the Pauli matrices 2 ), simulating the commutative
multiplication of complex numbers. z = 1 + 0i is rep-
resented by the identity matrix and |z|2 = x2 + y 2 is
det A(z). The reciprocal of a non-zero complex number
is mapped to the inverse matrix. The map A takes the
set of unit complex numbers z = ei comprising the 1-
sphere S 1 : x2 + y 2 = 1, tothe space SO(2)  of 2 2
cos sin 8
A matrix is orthogonal if its
orthogonal8 matrices R = with unit
sin cos rows (or columns) are orthonor-
determinant. mal: R t R = R R t = I, where t
denotes transposition. The set
Similarly, quaternions may be represented as 2 2 com- of n n orthogonal matrices of
plex matrices by exploiting the Pauli matrix algebra. unit determinant O(n) implement
rotations in n-dimensional Eu-
Hamiltons relations (Table 1) are satised if we repre- clidean space.
sent the basis quaternions e, i, j and k by the following
matrices (this choice is not unique!)
   
1 0 i 0
F (e) = , F (i) = = i3 ,
0 1 0 i
   
0 1 0 i
F (j) = = i2, F (k) = = i1 . (11)
1 0 i 0

This map extends by linearity to a 1-1 homomorphism of

RESONANCE June 2016 535

 GENERAL ARTICLE

Unit quaternions H into the associative algebra of 2 2 complex matrices:

comprise the
3-sphere embedded q = e + i + j + k  F (q), where
in four-dimensional    
+ i + i z w
Euclidean space. F (q) = = . (12)
+ i i w z

In this representation, the non-commutativity of quater-

nion multiplication is reected in that of the matrices.
As before, the reciprocal of a non-zero quaternion is
mapped to the inverse matrix and |q|2 = 2 + 2 + 2 + 2
is detF (q). Under this map, the image of the set of unit
quaternions (|q| = 1) is the space SU(2) of 2 2 uni-
tary matrices (F F = F F = I, where denotes com-
plex conjugate transpose) with unit determinant. Notice
moreover that unit quaternions comprise the 3-sphere
S 3 : 2 + 2 + 2 + 2 = 1 embedded in R4 .
3. Quaternions and Rotations
Rotations are an important symmetry of many prob-
lems in geometry, physics and astronomy and it is use-
ful to have explicit formulae for how rotations trans-
form vectors. Hamilton developed quaternions in part
to describe rotations in 3D, inspired by how complex
numbers describe rotations in the plane. Multiplication
of w = u + vi by a complex number z = ei of unit
magnitude represents a counter-clockwise rotation R
of w = (u, v) by angle . These rotations may be com-
posed R R = R+ , and undone R1 = R . Thus,
unit complex numbers ei form the circle group U(1) of
1 1 unitary matrices. The map A of equation (10)
Hamilton developed restricted to |z| = 1 is an isomorphism from U(1) to the
quaternions in part to group SO(2) of rotations of vectors in a plane.
describe rotations in
three-dimensions, Quaternions give a convenient way of expressing rota-
inspired by how tions in three dimensions. First, we identify vectors in
complex numbers
3D space with imaginary quaternions H = R3. By
analogy with the unit complex numbers acting as rota-
describe rotations
tions of the plane, the unit quaternions (|q| = 1) act as
in the plane.

536 RESONANCE June 2016

 GENERAL ARTICLE

rotations of H. However, the action is not by multipli-

cation, but by conjugation v  v  = qvq 1, which takes
H to H. Indeed, one checks using (5) that if v is
purely imaginary, then so is v . Moreover, by the law of
moduli |v | = |qvq 1| = |v| so that lengths of vectors are
preserved under conjugation. Explicitly, if v = (0, v),
q = (, u) and q 1 = q = (, u) with 2 + |u|2 = 1,
then using (5) we get

qvq 1 = (u v, v + u v) (, u)
2
= (0, v + 2u v + (u v)u + u (u v))
= 0, ( |u|2 )v + 2(u v) + 2(u v)u
2

(0, R v). (13)

Here R = R(q) is the 3 3 matrix R = (2 |u|2 )I

2
 u + 2uut with u the matrix whose entries are uab =
9
9
Here abc is the Levi-Civita sym-
c abc uc . Thus, conjugation is an origin and length- bol, anti-symmetric under ex-
preserving linear transformation of H. Such transfor- change of any pair of indices
mations R must either be proper rotations (det R = 1, and with 123 = 1.
comprising the group SO(3)) or rotations composed with
a reection (det R = 1). It is possible to show that Figure 2. The infinitesimal
the space SU(2) of unit quaternions is path connected, change v v in a
so that the continuous function det R(q) must always vector v due to a counter-
clockwise rotation by a small
equal the value 1 that it takes at the identity quater-
angle about the axis .
nion q = e. Thus, conjugation by unit quaternions act
Notice that | v| |v| sin
as proper rotations. Now, Euler showed in 1776 that
and that v points along v.
any rotation of 3D space may be regarded as a counter-
clockwise rotation by angle about an axis specied by
a unit vector n . It remains to relate the parameters
and u of the rotation R(q) to and n . It turns out that
= cos(/2) and n is the unit vector along u.
To see this, we rst consider an innitesimal rotation,
which should correspond to conjugation by a unit quater-
nion q = (, u) that diers innitesimally from the iden-
tity q = e = (1, 0) (i.e., 2 1 and |u|2 0). In this
case, from (13) we nd

Rv = v = v + v v + 2u v. (14)

RESONANCE June 2016 537

 GENERAL ARTICLE

10
Recall Euler's formula ei = A delightful piece of trigonometry shows that this rep-
cos + i sin for a unit complex resents a counter-clockwise rotation of v about the axis
number as an exponential of an
imaginary number. Similarly, unit = u/|u| by a small angle with || = 2|u| 2|u|
n
quaternions are expressible as since 2 = 1 |u|2 and |u|2 0 (see Figure 2). More
exponentials of purely imaginary
quaternions e(/2) n = cos (/2)e + generally, we may always express the unit quaternion
sin(/2) n . To show this, notice q = e + u as10
that the unit imaginary quater-
nionn = nxi + nyj + nz k satisfies q = cos(/2) e + sin(/2) n
, (15)
n = e, n 3 = n, n 4 = e, etc.,
2
where n is the unit vector along u. The division by
allowing one to sum the expo-
nential series. two ensures that for small |u|, may be interpreted as
the angle of rotation, since |2u| = |2 sin(/2)n| ||.
Plugging = cos(/2) and u = sin(/2) n into R in
(13) we arrive at
 
Rv = cos2 (/2) sin2(/2) v
+ 2 cos(/2) sin(/2) n v + 2 sin2 (/2)(
n v) n

= cos v + sin n v + (1 cos)(n v) n
. (16)
Interestingly, this formula for the result of rotating a
vector v by angle about an axis n was obtained by B
O Rodrigues (1840) shortly before Hamiltons discovery
of quaternions, using Eulers 4-square identity. For small
, (16) reduces to (14). In general, the rotated vector
v is expressed as a linear combination of (the generally
non-orthogonal vectors) v, n v and n . For planar
rotations (v perpendicular to the axis n ), these vectors
become orthogonal and the third term is absent.
The ability to multiply real
The set of unit quaternions {|q| = 1} is closed under
numbers gives the unit reals
S0 = {1,1} the structure of
multiplication and inverses and forms a group isomor-
the cyclic group of order two: phic to SU(2) via the map F of equation (12). The map
/2 . The multiplication of R(q) denes a homomorphism from unit quaternions (or
complex numbers gives the SU(2)) onto the rotation group SO(3). The homomor-
unit circle S1 in the complex
phism is 2 : 1 since conjugation by both q = e and
plane the structure of the
group U(1). Similarly, the ex-
q = e correspond to the identity rotation (the kernel
istence of the quaternions en- is {e}). The appearance of the half-angle /2 in (15)
dows the unit quaternions is symptomatic of this. We say that SU(2) is a double
(i.e., the 3-sphere S3) with cover of SO(3). This fact has many interesting conse-
the structure of the group
quences, such as the possibility for particles such as the
SU(2).
electron to have spin one-half [6, 7].

538 RESONANCE June 2016

 GENERAL ARTICLE

Box 2. Quaternion Group

The unit quaternions have an interesting discrete subgroup: the basis quaternions and
their negatives e, i, j, k form the quaternion group of order eight. Each element
other than e generates a cyclic subgroup of order 4, e.g. i = {i, i2 = 1, i3 = i, i4 =
e}. The quaternion group acts by left (or right) multiplication on H. For instance,
i(e + i + j + k) = e + i j + k may be interpreted as a simultaneous rotation
by 90 counter-clockwise in the 1 i and j k planes.

Box 3. Quaternions and Rotations in Four Dimensions

Remarkably, quaternions also help us to understand rotations in R4 . As before, one
may show that the group SU(2) of unit quaternions act via left and right multiplication
v  qvq  as length and orientation preserving linear transformations of the 4D space of
quaternions v. This gives a homomorphism from SU(2)SU(2) onto the rotation group
SO(4) with kernel (I, I). Thus the former is a double cover of the latter. The group
SO(4) arises as a symmetry of the Kepler problem as well as the hydrogen atom. The
two copies of SU(2) are related to the conservation of angular momentum and Laplace
RungeLenz vectors.

The abelian group of unit complex numbers U(1) has

discrete subgroups consisting of the nth roots of unity
(the cyclic group Z/nZ for each n = 1, 2, 3, . . .). Simi-
larly, the unit quaternions have discrete subgroups, see
Box 2 for an interesting non-abelian example. Remark-
ably, quaternions also help in understanding rotations
in four dimensions as discussed in Box 3.
4. Hopf Maps for Reals, Complexes and Quater-
nions
We have met the unit circle S 1, the surface of the unit
sphere S 2 and the unit reals {1} denoted S 0 . More
generally, S n = {(x1, x2 , . . . , xn+1 ) Rn+1 | x21 + x22 +
+ x2n+1 = 1} is called the unit n-sphere. The Hopf 11
The octonions are an eight-
map (H Hopf, 1931) and its generalizations are many- dimensional normed division al-
gebra generalizing , and
to-one maps between certain spheres. There are Hopf
. An octonion is a linear com-
maps associated to the reals, complexes, quaternions bination of the basis octonions
and octonions11 (but no more!): e0, ..., e7, which obey a non-
commutative and non-associa-
S 1 S 1, S 3 S 2 , S 7 S 4 and S 15 S 8 . (17) tive product rule [11].

RESONANCE June 2016 539

 GENERAL ARTICLE

Figure 3. A circle bundle

over a patch of the two-
sphere.

The maps go from a total-space sphere upstairs to a

base-space sphere downstairs. In each case, the pre-
image of any point on the base sphere turns out (see be-
low) to be a sphere of dimension one less than the base!
These pre-images of points on the base are called bres
and the total space is called a bre bundle (see Figure
3). The bres in the above four cases are S 0, S 1 , S 3 and
S 7 . This allows the sphere upstairs to be viewed locally
(though not globally) as the Cartesian product of the
base sphere and the bre sphere. Thus we say that S 3 is
a circle (S 1) bundle over S 2 and that S 7 is an S 3 bundle
over S 4. In the rst three cases, the bres have a group
structure: S 0 = Z/2Z, S 1 = U(1) and S 3 = SU(2), so
that the base spheres S , S 2 and S 4 may be viewed as
1

quotients of the total spaces by group actions:

S1
= S 1 /S 0 , S2
= S 3 /S 1 and S 4
= S 7 /S 3 . (18)

To understand the Hopf maps in a little more detail, let

us start with the real case, where it goes from the real
1-sphere S 1 (which, with a view towards generalization,
we denote by SR1 ) to the real projective line P 1 (R).
The latter is the set of lines through the origin in R2.
The sphere upstairs
The line through (x1, x2 ) = (0, 0) is an equivalence class
may be viewed locally
[(x1, x2)] of ordered pairs that dier by rescaling by a
(though not globally)
non-zero real number : (x1, x2) (x1 , x2 ). Any such
as the Cartesian line (other than the vertical line through the origin),
product of the base contains exactly one point with x1 = 1, that is, a point
sphere and the fibre of the form (1, x2). This point can be used to label that
sphere. line. Using the label (0, ) (or (0, )) for the vertical

540 RESONANCE June 2016

 GENERAL ARTICLE

Figure 4. A `cartoon' of the

2:1 real Hopf map S1 S1
with S0 fibres. For easy visu-
alization, the points identified
`upstairs' are not antipodal.

line through the origin, we see that P 1 (R) may be iden-

tied with the set of real numbers x2 compactied by
the point at to form a circle, so P 1 (R) = S 1 . The
1
Hopf map takes (x1 , x2) SR to the equivalence class
[(x1, x2)]. Under this map all points of the form (x1 , x2)
with || = 1 (i.e., where = 1 is a unit real) are
mapped to the same equivalence class [(x1, x2)]. Thus we
may identify the pre-image {(x1, x2 )/|x|, (x1, x2)/|x|}
of [(x1, x2 )] with S 0 = {1, 1} (here |x|2 = x21 +x22). The
set of unit reals S 0 is the bre over [(x1, x2)]. We say
that the real projective line is the quotient of the real
1-sphere by the group of unit reals. This Hopf map may
also be viewed as the 2 : 1 map ei  e2i of the unit
circle to itself, obtained by identifying antipodal points
(see Figure 4).
Replacing the reals with the complexes or quaternions
leads to corresponding Hopf brations. For instance, we
have the Hopf map from the complex 1-sphere
 
SC1 = (z1, z2 ) C2 | |z1|2 + |z2|2 = 1 = S3 (19)

to the complex projective line P 1 (C) taking (z1, z2) SC1

to the equivalence class [(z1, z2)]. As before, points of the
form (z1 , z2) with = ei , a complex number of unit
magnitude, are mapped to the same point/equivalence
class [(z1, z2)] in P 1 (C). So each bre is a circle S 1 . Fur-
thermore, as for the reals, we may choose the equivalence
class representatives as (1, z2) for z2 C and (0, ),

RESONANCE June 2016 541

 GENERAL ARTICLE

and thereby identify P 1 (C) with the complex plane (or

R2 ) compactied by the point at innity. So, P 1 (C)
=
2 3 2 1
S . Thus the Hopf map S S with S bres allows
us to view S 3 as a circle bundle over the two-sphere (see
Figure 3). Moreover, exploiting the fact that S 3 and
S 1 can be identied with the groups SU(2) and U(1) of
unit quaternions and complexes, it can be shown that
S 2 is the quotient of SU(2) by U(1). This Hopf map
appears in the study of polarized light [8], the Dirac
magnetic monopole [9], single qubit Hilbert spaces [10]
as well as the planar three-body problem. Depend-
ing on the context, the base S 2 is referred to as the
Bloch/Poincare/shape sphere.
Turning to quaternions, we have a map from the quater-
nion 1-sphere

SH1 = {(q1, q2) H2 | |q1 |2 + |q2|2 = 1}

= S7 (20)

to the quaternion projective line P 1 (H)

= S 4 (one point
compactication of H = R4 ) with bres given by the
unit quaternions S 3 . This Hopf bration makes an ap-
pearance in the theory of instantons [9] (Euclidean eld
congurations localized in both space and time) relevant
12
One difference is that unlike
to tunnelling phenomena in particle physics and in the
unit reals, complexes or quater-
nions, the unit octonions (S7) do
study of the geometry of two qubits. The nal Hopf
not form a group, since the bration S 7 S 15 S 8 results from a similar con-
octonion multiplication is non- struction involving the octonions12 . The Hopf maps are
associative. summarized in Box 4.
Box 4. The Hopf Maps for R, C, H and O

{|x| = 1} = S 0 / S1 1
{|z| = 1}
= S1 / S1 3
R = S C = S

 
P 1 (R) = S1 P 1 (C) = S2

{|q| = 1}
= S3 / S1 7
{|o| = 1}
= S7 / S1 15
H = S O = S

 
P 1 (H) = S4 P 1 (O) = S8

542 RESONANCE June 2016

 GENERAL ARTICLE

5. Afterword Hamiltons way of

thinking unlocked the
Hamilton thought that the discovery of quaternions in
freedom available in
the mid-19th century was as important as the discovery
relaxing one or other
of calculus at the end of the 17th century. Quaternions
axiom of arithmetic,
were vigorously researched and taught in parts of Ire-
land, England and the United States (a Quaternion So- and in working with
ciety, the International Association for Promoting the higher dimensional
Study of Quaternions and Allied Systems of Mathemat- algebras, even to
ics was founded at Yale University in 1895). Though the understand problems
importance of quaternions was probably overestimated in ordinary space.
by Hamilton and the quaternionists, their discovery
catalyzed several related developments. Hamiltons way
of thinking unlocked the freedom available in relaxing
one or other axiom of arithmetic, and in working with
higher dimensional algebras, even to understand prob-
lems in ordinary space. Within a few months of Hamil-
tons discovery of quaternions, his friend J T Graves
found another hypercomplex number system by drop-
ping the condition of associativity: the 8-dimensional
normed division algebra of octonions (rediscovered by
Cayley in 1845). The contemporaneous development
of the non-commutative algebra of matrices (especially
beginning with the work of Cayley, recall the Cayley
Hamilton theorem) was arguably as important to the
mathematical sciences and engineering as the discovery
of calculus. Quaternions also fuelled the development
of vector algebra and vector calculus (by Grassmann,
Gibbs, Heaviside and Helmholtz) and the formulation
of the equations of particle and continuum mechanics
and electromagnetism in terms of them.
Though quaternions may be viewed simply as a sub-
algebra of 2 2 complex matrices, they are very special.
Indeed, Frobenius (1877) showed that any nite- dimen-
sional associative division algebra over the reals, is iso-
morphic to one of R, C or H. Relaxing the condition of
associativity allows for the octonions (Hurwitz, 1923).
Quaternions continue to appear in various areas of

RESONANCE June 2016 543

 GENERAL ARTICLE

mathematical research: e.g., quaternionic versions of in-

tegers in number theory, quaternionic representations of
groups and quaternionic analogues of real and complex
manifolds, with many connections to physics. Quater-
nions are also used by computer scientists and engineers!
The parametrization of rotations by unit quaternions is
used by programmers to animate camera movements [12]
in video games (such as Tomb Raider) and to smoothly
interpolate between successive orientations of an air-
plane in ight simulators.
Acknowledgements
We thank B V Rao, R Nityananda and N Mukunda
for carefully reading this article and suggesting improve-
ments.

Suggested Reading

[1] R P Graves, Life of Sir William Rowan Hamilton, Volume II, Reprinted from the collection of the Univ. of
Michigan Library, Chapter xxviii, 1882.
[2] W R Hamilton, On Quaternions; or on a new System of Imaginaries in Algebra, The London, Edinburgh and
Dublin Philosophical Magazine and Journal of Science (3rd Series) Vol.xxv-xxxvi, 18441850; reprinted in The
Mathematical Papers of Sir William Rowan Hamilton, Vol.iii, (Algebra), Edited for the Royal Irish Academy
by H Halberstam and R E Ingram, Cambridge University Press, Cambridge, 1967.
[3] E T Bell, Men of Mathematics, Simon and Schuster, New York, pp.340362, 1986.
[4] L Brand, Vector and Tensor Analysis, John Wiley and Sons, New York, pp.403429, 1947.
[5] M Koecher and R Remmert, Hamilton's Quaternions, Chapter 7 in H-D Ebbinghaus et al, Numbers, Graduate
Texts in Mathematics 123, Springer-Verlag, New York, pp.189220, 1990.
[6] K Gottfried and T M Yan, Quantum Mechanics: Fundamentals, Springer-Verlag, New York, 2nd Edition,
Chapters 3 and 7, 2003.
[7] E C G Sudarshan and N Mukunda, Classical Dynamics: A Modern Perspective, Hindustan Book Agency, New
Delhi, 2015.
[8] R Nityananda, Impossibility of a continuous phase convention for polarised light, Pramana, Vol.12, No.3,
pp.257262, 1979.
[9] M Nakahara, Geometry, Topology and Physics, Taylor and Francis, Boca Raton, 2nd Edition, pp.368369,
2003.
[10] R Mosseri and R Dandoloff, Geometry of entangled states, Bloch spheres and Hopf fibrations, J. Phys. A: Math.
Gen., Vol.34, pp. 1024310252, 2001.
[11] J H Conway and D A Smith, On Quaternions and Octonions: Their Geometry, Arithmetic, and Symmetry, A K
Peters/CRC Press, Massachusetts, 2003.
[12] K Shoemake, Animating rotation with quaternion curves, ACM SIGGRAPH Computer Graphics, Vol.19, No.3,
pp.245254, 1985.

544 RESONANCE June 2016