Quaternions—Algebraic View 1

Quaternions—An Algebraic View

Note. You likely first encounter the quaternions in Introduction to Modern Al-
gebra. John Fraleigh’s A First Course in Abstract Algebra, 7th edition (Addison
Wesley, 2003), defines the quaternions in Part IV (Rings and Fields), Section 24
(Noncommutative Examples—see pages 224 and 225). However, this is an “op-
tional” section for Introduction to Modern Algebra 1 (MATH 4127/5127). In Mod-
ern Algebra 1 (MATH 5410) Thomas Hungerford’s Algebra (Springer-Verlag, 1974)
defines them in his Chapter III (Rings), Section III.1. Rings and Homomorphisms).
We now initially follow definitions of Hungerford.

Definition (Hungerford’s III.1.1). A ring is a nonempty set R together with

two binary operations (denoted + and multiplication) such that:

(i) (R, +) is an abelian group.

(ii) (ab)c = a(bc) for all a, b, c ∈ R (i.e., multiplication is associative).

(iii) a(b + c) = ab + ac and (a + b)c = ac + bc (left and right distribution of

multiplication over +).

If in addition,

(iv) ab = ba for all a, b ∈ R,

then R is a commutative ring. If R contains an element 1R such that

(v) 1R a = a1R = a for all a ∈ R,

then R is a ring with identity (or unity).

Quaternions—Algebraic View 2

Note. An obvious “shortcoming” of rings is the possible absence of inverses under

multiplication. We adopt the standard notation from (R, +). We denote the +
identity as 0 and for n ∈ Z and a ∈ R, na denotes the obvious repeated addition.

Definition (Hungerford’s III.1.3). A nonzero element a in the ring R is a left

(respectively, right) zero divisor if there exists a nonzero b ∈ R such that ab = 0
(respectively, ba = 0). A zero divisor is an element of R which is both a left and
right zero divisor.

Definition (Hungerford’s page 117). Let S = {1, i, j, k}. Let H be the ad-
ditive abelian group R ⊕ R ⊕ R ⊕ R and write the elements of H as formal sums
(a0 , a1 , a2 , a3 ) = a0 1 + a1 i + a2 j + a3 k. We often drop the “1” in “a0 1” and replace
it with just a0 . Addition in H is as expected:

(a0 +a1 i+a2 j+a3 k)+(b0 +b1 i+b2 j+b3 k) = (a0 +b0 )+(a1 +b1 )i+(a2 +b2 )j+(a3 +b3 )k.

We turn H into a ring by defining multiplication as

(a0 + a1 i + a2 j + a3 k)(b0 + b1 i + b2 j + b3 k) = (a0 b0 − a1 b1 − a2 b2 − a3 b3 )

+(a0 b1 +a1 b0 +a2 b3 −a3 b2 )i+(a0 b2 +a2 b0 +a3 b1 −a1 b3 )j +(a0 b3 +a3 b0 +a1 b2 −a2 b1 )k.

This product can be interpreted by considering:

(i) multiplication in the formal sum is associative,

(ii) ri = ir, rj = jr, rk = kr for all r ∈ R,

(iii) i2 = j 2 = k 2 = ijk = −1, ij = −ji = k, jk = −kj = i, ki = −ik = j.

This ring is called the real quaternions and is denoted H in commemoration of Sir
William Rowan Hamilton (1805–1865) who discovered them in 1843.
Quaternions—Algebraic View 3

Definition (Hungerford’s III.1.5). A commutative ring R with (multiplicative)

identity 1R and no zero divisors is an integral domain. A ring D with identity 1D 6= 0
in which every nonzero element is a unit is a division ring. A field is a commutative
division ring.

Note. First, it is straightforward to show that 1 = (1, 0, 0, 0) is the identity in

H. However, since ij = −ji 6= ji, then H is not commutative and so H is not an
integral domain nor a field.

Theorem A. The quaternions form a noncommutative division ring.

Note. Since every nonzero element of H is a unit, the H contains no left zero
divisors: If q1 q2 = 0 and q1 6= 0, then q2 = q1−1 0 = 0. Similarly, H has no right zero
divisors.

Note. I use the 8-element multiplicative group {±1, ±i, ±j, ±k} (called by Hunger-
ford the “quaternion group”; see his Exercise I.2.3) to illustrate Cayley digraphs
of groups in Introduction to Modern Algebra (MATH 4127/5127); see my online
notes for that class on Section I.7. Generating Sets and Cayley Digraphs. Consider
the Cayley digraph given below. This is the Cayley digraph for a multiplicative
group of order 8, denoted Q8 . The dotted arrow represents multiplication on the
right by i and the solid arrow represents multiplication on the right by j. The
problem is to use this diagram to create a multiplication table for Q8 .
Quaternions—Algebraic View 4

Solution. For multiplication on the right by i we have: 1 · i = i, i · i = −1,

−1 · i = −i, −i · i = 1, j · i = −k, −k · i = −j, −j · i = k, and k · i = j. For
multiplication on the right by j we have: 1 · j = j, j · j = −1, −1 · j = −j,
−j · j = 1, i · j = k, k · j = −i, −i · j = −k, and −k · j = i. This gives us 16
of the entries in the multiplication table for Q8 . Since 1 is the identity, we get
another 15 entries. All entries can be found from this information. For example,
k · k = k · (i · j) = (k · i) · j = (j) · j = −1. The multiplication table is:

· 1 i j k −1 −i −j −k
1 1 i j k −1 −i −j −k
i i −1 k −j −i 1 −k j
j j −k −1 i −j k 1 −i
k k j −i −1 −k −j i 1
−1 −1 −i −j −k 1 i j k
−i −i 1 −k j i −1 k −j
−j −j k 1 −i j −k −1 i
−k −k −j i 1 k j −i −1
Quaternions—Algebraic View 5

This group is addressed in Hungerford in Exercises I.2.3, I.4.14, and III.1.9(a).

Notice that each of i, j, and k are square roots of −1. So the quaternions are, in
a sense, a generalization of the complex numbers C. The Galois group AutQ Q(α),
q √ √
where α = (2 + 2)(3 + 3), is isomorphic to Q8 (see page 584 of D. Dummit
and R. Foote’s Abstract Algebra, 3rd edition, John Wiley and Sons (2004)).

Note. The quaternions may also be interpreted as a subring of the ring of all 2 × 2
matrices over C. This is Exercise III.1.8 of Hungerford (see page  120): “Let
 R be
z w
the set of all 2 × 2 matrices over the complex field C of the form  , where
−w z
z, w are the complex conjugates of a and w, respectively. Prove that R is a division
ring and that R is isomorphic to the division ring of real quaternions.” In fact,
the quaternion
  group, Q8 ,can bethought of as the group of order 8 generated by
0 1 0 i
A=  and B =  , under matrix multiplication (see Hungerford’s
−1 0 i 0
Exercise I.2.3).

Note. In fact, the complex numbers

 can
 be similarly represented as the field of
a −b
all 2 × 2 matrices of the form   where a, b ∈ R (see Exercise I.3.33 of
b a
Fraleigh).
Quaternions—Algebraic View 6

Note. The complex numbers can be defined as ordered pairs of real numbers,
C = {(a, b) | a, b ∈ R}, with addition defined as (a, b) = (c, d) = (a + c, b +
d) and multiplication defined as (a, b)(c, d) = (ac − bd, bc + ad). We then have
that C is a field with additive identity (0, 0) and multiplicative identity (1, 0).
The additive inverse of (a, b) is (−a, −b) and the multiplicative inverse of (a, b) 6=
(0, 0) is (a/(a2 + b2 ), −b/(a2 + b2 )). We commonly denote (a, b) as “a + ib” so
that i = (0, 1) and we notice that i2 = −1. In fact, this is the definition of the
complex field in our graduate level Complex Analysis 1 (MATH 5510); see my online
notes for that class on Section I.2. The Field of Complex Numbers. The complex
numbers are visualized as the “complex plane” where a + ib ∈ C is associated with
(a, b) ∈ R2 . During the early decades of the 19th century, the complex numbers
became an accepted part of mathematics (in large part due to the development of
complex function theory by Augustin Cauchy). Since the complex numbers have
an interpretation as a sort of “two dimensional” number system, a natural question
to ask is: “Is there a three (or higher) dimensional number system?”

Note. Sir William Rowan Hamilton (1805–1865) spent the years 1835 to 1843
trying to develop a three dimensional number system based on triples of real num-
bers. He never succeeded. However, he did succeed in developing a four dimensional
number system, now called the quaternions and denoted “H” in his honor. In a
letter he wrote late in his life to his son Archibald Henry, Hamilton tells the story
of his discovery:
Quaternions—Algebraic View 7

“Every morning in the early part of [October 1843], on my coming down

to breakfast, your little brother, William Edwin, and yourself, used to
ask me, ‘Well, papa, can you multiply triplets?’ Whereto I was always
obliged to reply, with a sad shake of the head: ‘No, I can only add and
subtract them.’ But on the 16th day of that some month. . . An electric
circuit seemed to close; and a spark flashed forth the herald (as I foresaw
immediately) of many long years to come of definitely directed through
and work by myself, is spared, and, at all events, on the part of others if
I should even be allowed to live long enough distinctly to communicate
the discovery. Nor could I resist the impulse—unphilosophical as it
may have been—to cut with a knife on a stone of Brougham Bridge
[in Dublin, Ireland; now called “Broom Bridge”], as we passed it, the
fundamental formula with the symbols i, j, k:

i2 = j 2 = k 2 = ijk = −1

which contains the Solution of the Problem, but, of course, the inscrip-
tion has long wince mouldered away.”

Images from the MacTutor biography webpage for Hamilton and the Tripadvisor
website (both accessed 1/5/2024)
Quaternions—Algebraic View 8

So the exact date of the birth of the quaternions is October 16, 1843. [This note
is based on Unknown Quantity: A Real and Imaginary History of Algebra by John
Derbyshire, John Henry Press (2006).]

Note. You are probably familiar with the Factor Theorem which relates roots of a
polynomial to linear factor of the polynomial. You might not recall that it requires
commutivity, though:
The Factor Theorem. (Hungerford’s Theorem III.6.6). Let R be a com-
mutative ring with identity and f ∈ R[x]. Then c ∈ R is a root of f if and only if
x − c divides f .

Note. The Factor Theorem is used to prove the following, which might remind
you of the Fundamental Theorem of Algebra:
Hungerford’s Theorem III.6.7. If D is an integral domain contained in an
integral domain E and f ∈ D[x] has degree n, then f has at most n distinct roots
in E.
So in an integral domain, an n degree polynomial at most n roots. Surprisingly in
a division ring, this can be violated.

Note. It is easy to see that the polynomial q 2 + 1 ∈ H[q] has more than two roots.
Along with ±i are the roots ±j and ±k. In fact, the polynomial has an infinite
Quaternions—Algebraic View 9

number of roots in H! Let x1 , x2 , x3 ∈ R with x21 + x22 + x23 = 1. Then

(x1 i + x2 j + x3 k)2 = x21 i2 + x1 x2 ij + x1 x3 ik + x2 x1 ji + x22 j 2 + x2 x3 jk

+x3 x1 ki + x3 x2 kj + x23 k 2 by the definition of multiplication

= −x21 − x22 − x23 since ij = −ji, ik = −ki, jk = −kj

= −1 since x21 + x22 + x23 = 1.

Note. We now turn our attention to polynomials in H[x]. We are particularly

interested in roots of such polynomials, a version of the Factor Theorem, and the
concept of algebraic closure. Much of this material is of fairly recent origins. The
remainder of the supplement is mostly based on the following references:

1. T. Y. Lam, A First Course in Noncommutative Rings, Graduate Texts in Math-

ematics #131, Springer-Verlag (1991).

2. G. Gentili and D. C. Struppa, A New Theory of Regular Functions of a Quater-

nionic Variable, Advances in Mathematics 216 (2007), 279–301.

Definition. We denote by S the two dimensional sphere (as a subset of the four
dimensional quaternions H) S = {q = x1 i + x2 j + x3 k | x21 + x22 + x23 = 1}. As
observed above, for any I ∈ S we have I 2 = −1. For x, y ∈ R we let x + yS denote
the two dimensional sphere x + yS = {x + yI | I ∈ S}. (We might think of x + yS
as a two dimensional sphere centered at (x, 0, 0, 0) with radius |y|.)
Quaternions—Algebraic View 10

Note. We take q as the indeterminate in the ring of polynomials H[q]. Since

H is not commutative, we are faced with the case that a monomial of the form
aq n ∈ H[q] is the same as monomial a1 qa1 qa2 q · · · qan ∈ H[q] where a = a0 a1 · · · an ,
but if we evaluate aq n at some element of H, we may get a different value than
if we evaluate aq qa1 q · · · qan at the same element of H. That is, evaluation of an
element of H[q] at r ∈ H is not a homomorphism (recall that Fraleigh deals with
the evaluation homomorphism for field theory in Theorem 22.4). In the remainder
of this supplement, we consider polynomials with the powers of the indeterminate
on the left and the coefficients on the right: p(q) = ni=0 q i ai . We will call p a
P

“quaternionic polynomial.”

Pn i
Definition. For two quaternionic polynomials p1 (q) = i=0 q ai and p2 (q) =
Pm i
i=0 q bi in H[q], define the product
X
(p1 p2 )(q) = q i+j ai bj .
i=0,1,...,n;j=0,1,...,m

Note. We now explore roots of quaternionic polynomials. The following result is

originally due to A. Pogorui and M. V. Shapiro (in “On the Structure of the Set
of Zeros of Quaternionic Polynomials,” Complex Variables 49(6) (2004), 379–389)
but we present an easier proof due to Gentili and Struppa in 2007.

PN n
Theorem B. Let p(q) = n=0 q an be a given quaternionic polynomial. Suppose
that there exist x0 , y0 ∈ R and I, J ∈ S with I 6= J such that p(x0 + y0 I) = 0 and
p(x0 + y0 J) = 0. Then for all L ∈ S we have p(x0 + y0 L) = 0.
Quaternions—Algebraic View 11

Note. In fact, Gentili and Struppa develop a theory of analytic functions of a

quaternionic variable and show that the previous result holds for an analytic func-
tion.

Note. In a ring of polynomials, R[t], each element of R commutes with in-

determinate t (see Hungerford’s Theorem III.5.2(ii)). So in R[t] we have that
f (r) = ni=0 ai ti = ni=0 ti ai . However, for r ∈ R where R is not commutative
P P

we likely have ni=0 ai ri 6= ni=0 ri ai . So in order to evaluate f (r), we must de-

P P

cide on a standard representation of f (t). In this supplement, we use the form

f (t) = ni=0 ti ai ∈ R[t]. Additionally, we may have f (t) = g(t)h(t) in R[t], but we
P

may not have f (r) = g(r)h(r). Consider g(t) = t−a and h(t) = t−b where a, b ∈ R
do not commute (so ab 6= ba). Then we have by the definition of multiplication
that f (t) = g(t)h(t) = (t − a)(t − b) = t2 − t(a + b) + ab. But

f (a) = a2 − a(a + b) + ab = ab − ba 6= 0 = g(a)h(a).

(This “sneaky” behavior results from the term at being expressed as ta in the
representation of g(t)h(t).)

Definition 16.1 of Lam. Let R be a ring and f (t) = ni=0 ti ai ∈ R[t]. An element
P

r ∈ R is a left root of f if f (r) = ni=0 ri ai = 0. If g(t) = ni=0 ai ti ∈ R[t]. An

P P

element r ∈ R is a right root of g if g(r) = ni=0 ai ri = 0.

P

Proposition 16.2 of Lam. (The Factor Theorem in a Ring with Unity). An

element r ∈ R is a left (right) root of a nonzero polynomial f (t) = ni=0 ti ai ∈ R[t]
P

if and only if t − r is a left (right) divisor of f (t) in R[t].

Quaternions—Algebraic View 12

Note. Recall a right ideal of a ring R is a subring I of R such that for all r ∈ R
and x ∈ I we have xr ∈ I (Hungerford’s Definition III.2.1). We see from the Factor
Theorem in a Ring with Unity that the set of polynomials in R[t] having r as a left
root is precisely the right ideal (t − r)R[t] = {(t − r)g(t) | g(t) ∈ R[t]}.

Proposition 16.3 of Lam. Let D be a division ring and let f (t) = h(t)g(t) in
D[t]. Let d ∈ D be such that a = h(d) 6= 0. Then f (d) = h(d)g(a−1 da). In
particular, if d is a left root of f but not of h then the conjugate of d, a−1 da, is a
left root of g.

Note. A result similar to Proposition 16.3 holds for right roots.

Note. If D is an integral domain and p ∈ D[x] is of degree n, then p has at most

n roots in D (see Hungerford’s Theorem III.6.7, mentioned above). This is not the
case in a division ring as illustrated by p(q) = q 2 + 1 ∈ H[q], as described above.
The following result is analogous to Hungerford’s Theorem III.6.7, but for division
rings. It does not imply at most n roots, but roots from at most n conjugacy
classes.

Note. Quaternion a is a conjugate of quaternion b (in the algebraic sense) if

a = cbc−1 for some quaternion c. Notice that if a = c1 b1 c−1 −1
1 and a = c2 b2 c2 , then

b1 = c−1 −1 −1 −1 −1 −1
1 ac1 so that b1 = c1 (c2 b2 c2 )c1 = (c1 c2 )b2 (c1 c2 ) . So conjugation is an

equivalence relation and the conjugacy classes partition H.

Quaternions—Algebraic View 13

Theorem 16.4 of Lam. (Gordon-Motzkin Theorem.) Let D be a division

ring and let f be a polynomial of degree n in D[t]. Then the left (right) roots of f
lie in at most n conjugacy classes of D. If f (t) = (t − a1 )(t − a2 ) · · · (t − an ) where
a1 , a2 , . . . , an ∈ D, then any left (right) root of f is conjugate to some ai .

Definition. For q = a + bi + cj + dk ∈ H, we define the quaternionic conjugate

q = a − bi − cj − dk.

Note. For q = a + bi + cj + dk ∈ H, we have

qq = (a + bi + cj + dk)(a + (−b)i + (−c)j + (−d)k)

= ((a)(a) − (b)(−b) − (c)(−c) − (d)(−d))

+((a)(−b) + (b)(a) + (c)(−d) − (d)(−c))i

+((a)(−c) + (c)(a) + (d)(−b) − (b)(−d))j

+((a)(−d) + (d)(a) + (b)(−c) − (c)(−b))k

= a2 + b2 + c2 + d2 .
√
We define the modulus of a ∈ H as qq.

Lemma A. For q1 , q2 ∈ H we have q1 q2 = q2 q1 .

Note. Recall that a field is algebraically closed if every nonconstant polynomial

over the field has a root in the field. This is the motivation for the following
definition.
Quaternions—Algebraic View 14

Definition (Lam, page 169). A division ring D is left (right) algebraically closed
if every nonconstant polynomial in D[t] has a left (right) root in D.

Note. By Proposition 16.2, if f ∈ D[t] for left or right algebraically closed division
ring D, then f can by factored into a product of linear factors in D[t] (that is, f
splits in D[t]).

Note. The following is the Fundamental Theorem of Algebra for Quaternions.

The result originally appeared in I. Nivens’ “Equations in Quaternions,” American
Mathematical Monthly, 48 (1941), 654–661.

Theorem 16.14 of Lam. (“Niven-Jacobson” in Lam) Fundamental The-

orem of Algebra for Quaternions. The quaternions, H, are left (and right)
algebraically closed.

Note. Now that we have our Fundamental Theorem of Algebra, we conclude with
a brief exploration of the structure of the set of quaternions for which a polynomial
has a left (right) root. The following result is from A. Pogorui and M. Shapiro’s “On
the Structure of the Set of Zeros of Quaternionic Polynomials,” Complex Variables:
Theory and Applications 49(6) (2004), 379–389.
Quaternions—Algebraic View 15

Theorem (Pogorui and Shapiro). For f a (nonzero) polynomial in H[q]. The

set of left (right) roots of f consists of isolated points or isolated two dimensional
spheres of the form S = x + yS for x, y ∈ R. The number of isolated roots plus
twice the number of isolated spheres is less than or equal to n.

Note. We would hope that the “less than or equal to n” of Pogorui and Shapiro’s
theorem could simply be replaced with “equal to n.” The punch-line is that we
cannot have “the number. . . equal to the degree n,” but with a well thought-out
definition of multiplicity, we can get the multiplicities of the roots to sum to the
degree. This is explored in another supplement to these notes on Quaternions—The
Degree of a Polynomial versus the Number of Zeros.

Revised and Updated: 1/11/2024