Quaternion Equations and Quaternion Polynomial Matrices
Uploaded by
pablo rojoQuaternion Equations and Quaternion Polynomial Matrices
Uploaded by
pablo rojoQuaternion Equations and Quaternion Polynomial Matrices
by
Liji Huang
A Thesis submitted to the Faculty of Graduate Studies of
The University of Manitoba
in partial fulfilment of the requirements of the degree of
MASTER OF SCIENCE
Department of Mathematics
University of Manitoba
Winnipeg
Copyright © 2012 by Liji Huang
Contents
Abstract i
Acknowledgments iii
1. Introduction 1
1.1. History of Quaternions . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2. Background for Quaternion Equations . . . . . . . . . . . . . . . . . 5
1.3. Background for Quaternion Polynomial Matrices . . . . . . . . . . . . 8
2. Quaternion Equations 9
2.1. Quaternions and Equivalence Classes . . . . . . . . . . . . . . . . . . 9
2.2. General Quaternion Quadratic Equations . . . . . . . . . . . . . . . . 15
2.3. Special Quaternion Equations and problems . . . . . . . . . . . . . . 20
2.3.1. A quadratic equation . . . . . . . . . . . . . . . . . . . . . . . 20
2.3.2. Another quadratic equation . . . . . . . . . . . . . . . . . . . 36
2.3.3. A homogeneous quadratic equation . . . . . . . . . . . . . . . 40
2.3.4. A pair of quaternion equations . . . . . . . . . . . . . . . . . . 48
2.3.5. Another pair of quaternion equations . . . . . . . . . . . . . . 52
2.3.6. A least norm problem . . . . . . . . . . . . . . . . . . . . . . 55
2.3.7. Another least norm problem . . . . . . . . . . . . . . . . . . . 58
3. Quaternion Polynomial Matrices 61
3.1. Generalized Inverse . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
3.2. Leverrier-Faddeev Method . . . . . . . . . . . . . . . . . . . . . . . . 74
3.3. Generalized Inversion by Interpolation . . . . . . . . . . . . . . . . . 82
A. Maple Codes 89
A.1. Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
A.2. The Codes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90
Bibliography 119
Abstract
In this thesis, we mainly consider finding formula solutions to quaternion equations
and calculating the generalized inverses of quaternion polynomial matrices.
In the first chapter, we give a condensed review of the history of quaternions.
The topic of the second chapter is solving quadratic and other quaternion equations.
Many mathematicians have worked on solving various quaternion equations. How-
ever, due to the complex nature of quaternions, the current "best" non-numerical
result is a special case of quadratic quaternion equation presented in (Au-Yeung,
2003), which pales in comparison to what is known about equations in complex
numbers. In this chapter, we present a new way to solve not only almost all of the
quaternion equations that have been solved so far, including Au-Yeung’s result, but
also several more complicated ones. This method can also be easily implemented
into computer algebra systems such as Maple. Examples will be included.
The topic of the third chapter is calculating the generalized inverses of quaternion
polynomial matrices. It has been shown that a real (Karampetakis, 1997) or complex
(Stanimirović et al., 2007) polynomial matrix has the generalized inverse that can
be calculated using the Leverrier-Faddeev algorithm (Faddeeva, 1959). It is natural
to ask whether a quaternion polynomial matrix has a generalized inverse, and if yes,
how it can be calculated. In this chapter, we first give conditions that a quaternion
polynomial matrix must satisfy in order to have the generalized inverse. We then
i
Contents
present an interpolation method that can be used to calculate the generalized inverse
of a given quaternion polynomial matrix.
In the appendix, we include our quaternion polynomial matrix Maple module as
well as some examples.
ii
Acknowledgments
First and foremost, I would like to express my most sincere gratitude to my thesis
advisor, Dr. Zhang Yang. I greatly appreciate his encouragement and dedicated
support. Countless hours of discussion with Dr. Zhang helped me overcome many
obstacles during the writing up of this thesis. Without his wisdom and motivation
this thesis would not have been possible.
I greatly appreciate the faculty and staff in the Department of Mathematics for
their teaching and support during my two years of being a graduate student in
mathematics.
I greatly thank my thesis committee members, Dr. Guenter Krause, Dr. Xikui
Wang and Dr. Joe Williams, for reading my thesis and giving me feedback.
On behalf of myself and my wife, I would like to thank Dr. Zhang, the Faculty of
Graduate Studies, the Faculty of Science, and the Department of Mathematics, for
financially supporting me through my graduate studies.
I would like to thank the University of Manitoba library system for providing a rich
supply of online and offline publications. I would also like to thank mapleprimes.com
and latex-community.org for all of the excellent technical support I received from
iii
Acknowledgments
them, and their continued voluntary service to current and future graduate students
and academics all over the world.
Finally, I would like to express my appreciation to my family and friends, for all of
their support and love.
iv
Dedicated to my wife Zifang.
1. Introduction
1.1. History of Quaternions
Quaternions were first discovered by Sir William Rowan Hamilton of Ireland on
October 16th , 1843, in a "flash of genius", on the Brougham Bridge. It is uncommon
that both the date and location of a major mathematical discovery are known, but
we precisely know these because at the end of Hamilton’s life, in a letter to his
second son, Archibald Henry, he wrote:
Every morning in the early part of [October 1843], on my coming down to breakfast,
your (then) 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 the same month—which happened to be Monday, and a
Council day of the Royal Irish Academy—I was walking in to attend and preside,
and your mother was walking with me along the Royal Canal, to which she had
perhaps driven; and although she talked with me now and then, yet an undercurrent
of thought was going on in my mind which gave at last a result, whereof it is not
too much to say that I felt at once the importance. An electric circuit seemed to
close; and a spark flashed forth the herald (as I foresaw immediately) of many long
years to come of definitely directed thought and work by myself, if spared, and, at all
1
1.1 History of Quaternions
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, as we passed
it, the fundamental formula with the symbols i, j, k:
i2 = j2 = k2 = ijk = −1
which contains the Solution of the Problem, but, of course, the inscription has long
since mouldered away (Derbyshire, 2006).
It was a breakthrough discovery in the sense that after the step from one-dimensional
real numbers to two-dimensional complex numbers, the next step was not the natural
one to three-dimensional super-complex numbers but to four-dimensional ones, i.e.,
quaternions. Hamilton embarked on an eight-year journey to develop an algebra for
bracketed triplets but ended up discovering quaternions in a moment of euphoria.
The Dutch mathematician Van der Waerden called Hamilton’s inspiration "the leap
into the fourth dimension" (van der Waerden, 1985). The Scottish mathematical
physicist Peter Guthrie Tait, the leading quaternions exponent of his time, claimed
"...no figure, nor even model, can be more expressive or intelligible than a quaternion
equation" (Tait, 1890). However, Tait’s lifetime friend, William Thomson, 1st Baron
Kelvin, disliked quaternions and would have nothing to do with them (Smith and
Wise, 1989). Tait’s other great friend, the Scottish mathematical physicist James
Clerk Maxwell, remained unconvinced of quaternion’s value in actual calculation, as
he put it "...quaternions...is a mathematical method, but it is a method of thinking,
and not, at least for the present generation, a method of saving thought" (Maxwell,
1873).
Furthermore, the American scientist Josiah Willard Gibbs did not find quaternions
suited to his needs and rejected the quaternionic analysis in favor of his own in-
2
1.1 History of Quaternions
vention: vector analysis (Hastings, 2010; Gibbs et al., 1902). The English electrical
engineer Oliver Heaviside, who independently formulated vector analysis similar to
Gibbs’s system, wrote "...it is impossible to think in quaternions, you can only pre-
tend to do it" (Nahin, 2002). The British mathematician Arthur Cayley gave a
poetic illustration on the matter "...I compare a quaternion formula to a pocket-
map, a capital thing to put in one’s pocket, but which for use must be unfolded; the
formula, to be understood, must be translated into coordinates" (Cayley, 1894). Even
the English mathematician Alexander McAulay, a quaternion advocate, stated that
"...not much advance in physics has been made by the aid of quaternions" (McAulay,
1892). Indeed, in the century that has passed since McAulay wrote, it can be argued
that quaternions faded away to a minor role in mathematics and physics while the
Gibbs-Heaviside vector algebra became the everyday work tool of all modern engi-
neers and scientists. In 1892, Heaviside, hinting subtly that quaternions are what
he called "a positive evil of no inconsiderable magnitude", summarized the matter
in this way, and probably the best way "...the invention of quaternions must be re-
garded as a most remarkable feat of human ingenuity...but to find out quaternions
required a genius" (Nahin).
Today, quaternions still play a role in quantum physics (Adler, 1995; Finkelstein
et al., 1962) and other branches of physics (Churchill, 1990). Quaternions are now
used throughout the aerospace industry for attitude control of aircraft and space-
craft (Kuipers, 2002). Some people claim that quaternions are an important aspect
in modern computer graphics for calculations involving three-dimensional rotations
because quaternions use less memory, compose faster, and are naturally suited for ef-
ficient interpolation (Hanson, 2005). Although, like in the academic circle of physics
over a century ago, there are controversies regarding the advantages of quaternion
orientation representations in today’s computer graphics and modeling community
3
1.1 History of Quaternions
(Hanson, 2005). Quaternions are also attracting more and more attention in math-
ematics. There has been a new surge of interest in quaternions in the mathemat-
ics world—a quick search in the American Mathematical Society’s database shows
that more than half of the 3232 publications containing the word “quaternion” or
“quaternionic” in the title have been published after the year of 1996.
4
1.2 Background for Quaternion Equations
1.2. Background for Quaternion Equations
This section attempts to give a general and brief review on what has been done
on quaternion equations, §2.1 offers a more detailed introduction to quaternions
themselves.
Let x be a quaternion indeterminate and let ai be quaternions. In 1941, Niven
looked into the equation:
xm + a1 xm−1 + a2 xm−2 + · · · + am = 0, (1.1)
where am 6= 0. He showed that 1.1 has at least one quaternion root.
In the following year, Niven completely solved the equation
xm − am = 0.
In 1944, Eilenberg and Niven showed that even the most general equation with a
unique highest term,
a1 xa2 x · · · xam + φ (x) = 0,
where φ (x) is a sum of finite number of monomials of the form b1 xb2 x · · · xbk , k < m,
has at least one quaternion root. This is also known as the "fundamental theorem
of algebra" for quaternions.
In 1944, Johnson gave necessary and sufficient conditions for the equation
xa1 + a2 x + a3 = 0
to have a quaternion solution. His result was extended by Tian to include
xa1 + a2 x + a3 = 0
5
1.2 Background for Quaternion Equations
and
xa1 x + a2 = 0,
where x denotes the quaternion conjugate of x.
Very little was done to obtain formula solutions of quaternion equations after John-
son’s result, until, in 2002, Huang and So solved the special quadratic equation:
x2 + a1 x + a2 = 0. (1.2)
Then in the following year, Au-Yeung solved another special quadratic equation,
which included both Johnson’s and Huang and So’s results as special cases:
x2 + a1 x + xa2 + a3 = 0. (1.3)
Au-Yeung’s paper was the latest non-numerical attempt to find the exact roots of a
non-linear quaternion equation.
In 2010, Tian studied and solved two pairs of linear quaternion equations
xa1 y = a2 ,
(1.4)
ya2 x
= a1 ,
and
xa1 y = a2 ,
(1.5)
ya2 x
= a1 ,
where y is also a quaternion indeterminate. In §2.3, we present a new method to
finding the solutions to 1.3, 1.4 and 1.5. In addition, we give solutions to several
previously unsolved equations and least norm problems, including, for example, the
"two-sided" quadratic homogeneous equation:
x2 + a1 xa2 = 0,
6
1.2 Background for Quaternion Equations
and the least norm problem:
min |a1 x − xa2 − a3 | .
x∈H
In §A.2, we present some maple codes to solve these quaternion equations and
problems.
7
1.3 Background for Quaternion Polynomial Matrices
1.3. Background for Quaternion Polynomial Matrices
In 1955, Penrose introduced the term generalized inverse that exists for any matrix
with complex elements. In 1965, Decell gave a method for computing the general-
ized inverse of a constant complex matrix based on Leverrier-Faddeev’s algorithm
(Faddeeva, 1959), which recursively computes coefficients of the characteristic poly-
nomial of the matrix. This method was later extended by Karampetakis to real
polynomial matrices. In 2006, Stanimirović and Petković applied an interpolation
method to the result of Karampetakis. Then in the following year, Stanimirović
et al. extended Stanimirović and Petković’s result to include complex polynomial
matrices with interpolations at real number data points.
In Chapter 3, we first give conditions that a quaternion polynomial matrix must
satisfy in order to have a generalized inverse. Then we present an interpolation
method that can be used to calculate the generalized inverse of a given quaternion
polynomial matrix.
In §A.2, we include our Maple codes for calculating the generalized inverse of a given
quaternion matrix or quaternion polynomial matrix. We also include our Maple
codes that can be used to do Newton and Lagrange quaternionic interpolation, which
are closely related to the interpolation of the generalized inverse of a quaternion
polynomial matrix.
8
2. Quaternion Equations
2.1. Quaternions and Equivalence Classes
In this section, we will outline some definitions and properties of quaternions. First,
recall the definition of quaternions:
Definition 2.1.1. Let R denote the field of the real numbers. Let H be a four-
dimensional vector space over R with basis 1, i, j and k. A quaternion is a vector
x = x0 · 1 + x1 i + x2 j + x3 k ∈ H
with real coefficients x0 , x1 , x2 and x3 . Throughout this thesis, for convenience, if
h ∈ H, then we will assume that h = h0 + h1 i + h2 j + h3 k where hi ∈ R, 0 ≤ i ≤ 3.
We denote the real part, h0 , by Re h and the imaginary part, h1 i + h2 j + h3 k, by
Im h.
Addition and multiplication in H can be defined as follows:
Definition 2.1.2. Let x, y ∈ H. Then quaternion addition is defined as:
x + y = (x0 + y0 ) + (x1 + y1 ) i + (x2 + y2 ) j + (x3 + y3 ) k.
9
2.1 Quaternions and Equivalence Classes
Quaternion multiplication is defined as:
xy =x0 y0 − x1 y1 − x2 y2 − x3 y3
+ (x0 y1 + x1 y0 + x2 y3 − x3 y2 ) i
+ (x0 y2 − x1 y3 + x2 y0 + x3 y1 ) j
+ (x0 y3 + x1 y2 − x2 y1 + x3 y0 ) k.
Remark 2.1.3. By direct calculation, we obtain:
ii = jj = kk = −1,
ij = −ji = k, jk = −kj = i, ki = −ik = j.
It is also worth noting that with the above equations, the multiplication defined in
Definition 2.1.2 follows.
Lemma 2.1.4. Quaternion multiplication is not commutative in general. H is a
division ring with center R.
Proof. By direct calculation.
Next, conjugate and norm for quaternions are defined as follows:
Definition 2.1.5. Let h ∈ H. Then h denotes the conjugate of h, h = h0 − h1 i −
√ q
h2 j − h3 k, and |h| denotes the norm of h, |h| = hh = h20 + h21 + h22 + h23 .
Lemma 2.1.6. Let x, y ∈ H. Then |x| is non-negative and |·| is a norm on H, i.e.,
|x| = 0 ⇔ x = 0,
|x + y| ≤ |x| + |y| ,
|xy| = |yx| = |x| |y| .
10
2.1 Quaternions and Equivalence Classes
Proof. We will only show that |x + y| ≤ |x| + |y|. The rest of the lemma is true by
direct calculation. Since
|x + y|2 = |(x0 + y0 ) + i (x1 + y1 ) + j (x2 + y2 ) + k (x3 + y3 )|2
q 2
= (x0 + y0 )2 + (x1 + y1 )2 + (x2 + y2 )2 + (x3 + y3 )2
= (x0 + y0 )2 + (x1 + y1 )2 + (x2 + y2 )2 + (x3 + y3 )2
=x20 + x21 + x22 + x23 + y02 + y12 + y22 + y32 + 2x0 y0 + 2x1 y1 + 2x2 y2 + 2x3 y3
= |x|2 + |y|2 + 2x0 y0 + 2x1 y1 + 2x2 y2 + 2x3 y3 ,
it suffices to show that
|2x0 y0 + 2x1 y1 + 2x2 y2 + 2x3 y3 | ≤ 2 |x| |y|
|x0 y0 + x1 y1 + x2 y2 + x3 y3 | ≤ |x| |y|
(x0 y0 + x1 y1 + x2 y2 + x3 y3 )2 ≤ |x|2 |y|2
⇔
(x0 y0 + x1 y1 + x2 y2 + x3 y3 )2 ≤ x20 + x21 + x22 + x23 y02 + y12 + y22 + y32 ,
which is true by the Cauchy-Schwartz inequality.
Next, inverse and conjugacy class for quaternions are defined as follows:
Lemma 2.1.7. For 0 6= h ∈ H, h−1 , the inverse of h, is uniquely determined by
h−1 = h
|h|2
. Furthermore, we have |h−1 | = 1
|h|
.
Proof. By direct calculation.
Definition 2.1.8. Two quaternions x and y are said to be similar, or in the same
11
2.1 Quaternions and Equivalence Classes
conjugacy class, if these exists a quaternion v 6= 0 such that v −1 xv = y, and we
write x v y.
Lemma 2.1.9. x v y implies |x| = |y|.
Proof. By Lemma 2.1.7,
1
x v y ⇒ v −1 xv = y ⇒ |y| = v −1 xv = |x| |v| = |x| .
|v|
Next, we state and prove an important result regarding similar quaternions.
Theorem 2.1.10. x v y if and only if Re x = Re y and |Im x| = |Im y|.
Proof. We only prove the forward direction of the theorem. The backward direction
is straightforward.
If x = 0 or y = 0, then the claim is true trivially. Assuming x 6= 0 and y 6= 0, we
will first show that x0 = y0 . Let xv = vy where v 6= 0. That is,
(x0 + x1 i + x2 j + x3 k) (v0 + v1 i + v2 j + v3 k) (2.1)
= (v0 + v1 i + v2 j + v3 k) (y0 + y1 i + y2 j + y3 k) .
After expanding 2.1 and comparing coefficients, we get:
x0 v 0 − x1 v1 − x2 v2 − x3 v3 = y0 v0 − y1 v1 − y2 v2 − y3 v3 = A,
x0 v 1 + x1 v0 + x2 v3 − x3 v2 = y1 v0 + y0 v1 + y3 v2 − y2 v3 = B,
(2.2)
+ x0 v2 + x3 v1 − x1 v3 = y2 v0 + y0 v2 + y1 v3 − y3 v1 = C,
x2 v 0
x v + x3 v0 + x1 v2 − x2 v1 = y0 v3 + y3 v0 + y2 v1 − y1 v2 = D,
0 3
12
2.1 Quaternions and Equivalence Classes
which is equivalent to:
v0 (x0 − y0 ) = v1 (x1 − y1 ) + v2 (x2 − y2 ) + v3 (x3 − y3 ) ,
v0 (x1 − y1 ) = −v1 (x0 − y0 ) + v2 (x3 + y3 ) − v3 (x2 + y2 ) ,
(2.3)
− y2 ) = −v1 (x3 + y3 ) − v2 (x0 − y0 ) + v3 (x1 + y1 ) ,
v0 (x2
v (x − y3 ) = v1 (x2 + y2 ) − v2 (x1 + y1 ) − v3 (x0 − y0 ) .
0 3
Multiplying the four equations of 2.3 by (x0 + y0 ), (x1 + y1 ), (x2 + y2 ) and (x3 + y3 )
respectively, then adding them, we get
x0 (v1 y1 + v2 y2 + v3 y3 ) = y0 (v1 x1 + v2 x2 + v3 x3 )
x0 (y0 v0 − A) = y0 (x0 v0 − A)
Ax0 = Ay0 .
If x0 = y0 , we are done. Assume x0 6= y0 . Then A = 0.
Next, multiplying the four equations of 2.3 by (x1 + y1 ), − (x0 + y0 ), − (x3 − y3 )
and (x2 − y2 ) respectively, then adding them, we get
x0 (y1 v0 − y2 v3 + y3 v2 ) = y0 (x2 v3 − x3 v2 + x1 v0 )
x0 (B − y0 v1 ) = y0 (B − x0 v1 )
Bx0 = By0 .
Since x0 6= y0 , B = 0.
Next, multiplying the four equations of 2.3 by (x2 + y2 ), (x3 − y3 ), − (x0 + y0 ) and
13
2.1 Quaternions and Equivalence Classes
− (x1 − y1 ) respectively, then adding them, we get
x0 (y2 v0 + y1 v3 − y3 v1 ) = y0 (x2 v0 + x3 v1 − x1 v3 )
x0 (C − y0 v2 ) = y0 (C − x0 v2 )
Cx0 = Cy0 .
Since x0 6= y0 , C = 0.
Lastly, multiplying the four equations of 2.3 by (x3 + y3 ), − (x2 − y2 ), (x1 − y1 ) and
− (x0 + y0 ) respectively, then adding them, we get
x0 (y3 v0 + y2 v1 − y1 v2 ) = y0 (x3 v0 + x1 v2 − x2 v1 )
x0 (D − y0 v3 ) = y0 (D − x0 v3 )
Dx0 = Dy0 .
Since x0 6= y0 , D = 0. That is, A = B = C = D = 0 ⇒ v = 0 or x = y = 0.
Contradiction. Therefore, x0 = y0 . Then by Lemma 2.1.9, x21 +x22 +x23 = y12 +y22 +y32 ,
that is, |Im x| = |Im y|.
14
2.2 General Quaternion Quadratic Equations
2.2. General Quaternion Quadratic Equations
Let ai , bi , c ∈ H where i ∈ N. We write ai = ai0 + ai1 i + ai2 j + ai3 k and bi =
bi0 + bi1 i + bi2 j + bi3 k for easier notation. In this section, we focus on the general
quadratic equation:
k
x2 +
X
ai xbi + c = 0, (2.4)
i=1
where k ∈ N.
2.4 is equivalent to the real non-linear system:
(2x0 + m11 ) x1 + m12 x2 + m13 x3 = −c1 + x0 n1 , (1)
m21 x1 + (2x0 + m22 ) x2 + m23 x3 = −c2 + x0 n2 , (2)
(2.5)
+ m33 ) x3 = −c3 + x0 n3 ,
m31 x1 + m32 x2 + (2x0 (3)
x2 − x2 − x2 − x2 + d 1 x0 + d2 x1 + d3 x2 + d4 x3 + c0 = 0, (4)
0 1 2 3
where mij , ni and di are given by:
M = (mij )3×3
k ai0 bi0 − ai1 bi1 + ai2 bi2 + ai3 bi3 ai0 bi3 − ai1 bi2 − ai2 bi1 − ai3 bi0 −ai0 bi2 − ai1 bi3 + ai2 bi0 − ai3 bi1
X
= −ai0 bi3 − ai1 bi2 − ai2 bi1 + ai3 bi0 ai0 bi0 + ai1 bi1 − ai2 bi2 + ai3 bi3 ai0 bi1 − ai1 bi0 − ai2 bi3 − ai3 bi2 ,
i=1
ai0 bi2 − ai1 bi3 − ai2 bi0 − ai3 bi1 −ai0 bi1 + ai1 bi0 − ai2 bi3 − ai3 bi2 ai0 bi0 + ai1 bi1 + ai2 bi2 − ai3 bi3
X k
(−ai0 bi1 − ai1 bi0 − ai2 bi3 + ai3 bi2 )
i=1
k
X
N = (ni )3×1 =
(−ai0 bi2 + ai1 bi3 − ai2 bi0 − ai3 bi1 )
,
i=1
k
X
(−ai0 bi3 − ai1 bi2 + ai2 bi1 − ai3 bi0 )
i=1
15
2.2 General Quaternion Quadratic Equations
and
k
X
(ai0 bi0 − ai1 bi1 − ai2 bi2 − ai3 bi3 )
i=1
X k
(−ai0 bi1 − ai1 bi0 + ai2 bi3 − ai3 bi2 )
i=1
D = (di )4×1 = .
k
X
(−ai0 bi2 − ai1 bi3 − ai2 bi0 + ai3 bi1 )
i=1
X k
(−ai0 bi3 + ai1 bi2 − ai2 bi1 − ai3 bi0 )
i=1
Treating x0 as a constant, let M̌ be the coefficient matrix of (1), (2) and (3) of 2.5:
2x0 + m11 m12 m13
m21 2x0 + m22 m23 .
m31 m32 2x0 + m33
Direct calculation yields:
σ (x0 ) = det M̌
=8x30 + 4(m11 + m22 + m33 )x20
+ 2 (m11 m22 + m11 m33 − m12 m21 − m13 m31 − m23 m32 + m22 m33 ) x0
+ m11 m22 m33 − m11 m23 m32 − m12 m21 m33
+ m13 m21 m32 + m12 m23 m31 − m13 m22 m31 .
We continue the discussion by cases:
Case 1. We assume that det M̌ 6= 0.
Applying Cramer’s rule to (1), (2) and (3) of 2.5, we obtain x1 , x2 and
x3 as rational polynomials of x0 as follows:
p1 (x0 )
x1 = , (2.6)
σ (x0 )
p2 (x0 )
x2 = (2.7)
σ (x0 )
16
2.2 General Quaternion Quadratic Equations
and
p3 (x0 )
x3 = , (2.8)
σ (x0 )
where
p1 (x0 ) =4n1 x30 + 2 (n1 (m22 + m33 ) − n2 m12 − n3 m13 − 2c1 ) x20
+ 2 (m12 c2 + m13 c3 − c1 (m22 + m33 )) x0
+ n1 (m22 m33 − m23 m32 ) x0 + n2 (m13 m32 − m12 m33 ) x0
+ n3 (m12 m23 − m13 m22 ) x0 + c1 (m23 m32 − m33 m22 )
+ c2 (m12 m33 − m13 m32 ) + c3 (m13 m22 − m12 m23 ) ,
p2 (x0 ) =4n2 x30 + 2 (n2 (m11 + m33 ) − n1 m21 − n3 m23 − 2c2 ) x20
+ 2 (m21 c1 + m23 c3 − c2 (m11 + m33 )) x0
+ n1 (m23 m31 − m21 m33 ) x0 + n2 (m11 m33 − m13 m31 ) x0
+ n3 (m13 m21 − m11 m23 ) x0 + c1 (m21 m33 − m23 m31 )
+ c2 (m13 m31 − m11 m33 ) + c3 (m11 m23 − m13 m21 )
and
p3 (x0 ) =4n3 x30 + 2 (n3 (m11 + m22 ) − n1 m31 − n2 m32 − 2c3 ) x20
+ 2 (m31 c1 + m32 c2 − c3 (m11 + m22 )) x0
+ n1 (m21 m32 − m22 m31 ) x0 + n2 (m12 m31 − m11 m32 ) x0
+ n3 (m11 m22 − m12 m21 ) x0 + c1 (m22 m31 − m21 m32 )
+ c2 (m11 m32 − m12 m31 ) + c3 (m12 m21 − m11 m22 ) .
17
2.2 General Quaternion Quadratic Equations
Substituting 2.6, 2.7 and 2.8 back into (4) of 2.5, we obtain:
p1 (x0 )2 + p2 (x0 )2 + p3 (x0 )2
0 =x20 + d1 x0 + c0 −
σ (x0 )2
d2 p1 (x0 ) + d3 p2 (x0 ) + d4 p3 (x0 )
+ .
σ (x0 )
Multiplication of the above equation by σ (x0 )2 yields:
0 =σ (x0 )2 x20 + d1 x0 + c0 − p1 (x0 )2 − p2 (x0 )2 − p3 (x0 )2 (2.9)
+ σ (x0 ) (d2 p1 (x0 ) + d3 p2 (x0 ) + d4 p3 (x0 )) .
2.9 is an 8th order equation of x0 . After solving 2.9 for a real solution
x0 , we substitute its value back into 2.6, 2.7 and 2.8 to get x1 , x2 and
x3 respectively to obtain a solution to 2.4. The prime difficulty here is
that, by Galois theory, there is no formula solution for general eighth
degree equations (Morandi, 1996). Furthermore, there does not seem to
be a general formula to factor 2.9. However, as we will show in the next
section, under certain conditions, 2.9 can be solved.
Case 2. We assume that det M̌ = 0.
Then x0 is known: it is equal to a (fixed) real root (which always exists)
of the cubic polynomial σ (x0 ). Let Ň be the 3 × 1 vector:
−c1 + x0 n1
.
−c + x0 n2
2
−c3 + x0 n3
It is clear that 2.5 has a solution only if the last row of the reduced row
echelon form of the augmented matrix M̌ | Ň is a zero row. Under
these conditions, (1), (2) and (3) of 2.5, now viewed as a linear system
18
2.2 General Quaternion Quadratic Equations
with unknowns x1 , x2 and x3 , has infinitely many solutions of the form:
x = s1 y + t1 z + w1 ,
1
x2 = s2 y + t2 z + w2 , (2.10)
x3 = s3 y + t3 z + w3 ,
where si , ti and wi ∈ R, 1 ≤ i ≤ 3, and y, z ∈ R are parameters. We
plug 2.10 into (4) of 2.5 and obtain:
0 = − (s1 y + t1 z + w1 )2 − (s2 y + t2 z + w2 )2 − (s3 y + t3 z + w3 )2
+ d2 (s1 y + t1 z + w1 ) + d3 (s2 y + t2 z + w2 ) + d4 (s3 y + t3 z + w3 ) (2.11)
+ x20 + d1 x0 + c0 .
The parameters y and z must satisfy 2.11, therefore 2.11 can be seen as
a quadratic equation of, WLOG, y:
y 2 + f (z) y + g (z) = 0, (2.12)
where f (z) is a linear polynomial of z and g (z) is a quadratic polynomial
of z. We can take any z ∈ R such that the discriminant of 2.12, f (z)2 −
4g (z), a quadratic polynomial of z, is non-negative. We then solve for
√
−f (z)± f (z)2 −4g(z)
y= 2
. Finally, we substitute the value of y and z back
into to 2.10 to get x1 , x2 and x3 to obtain a solution to 2.4.
19
2.3 Special Quaternion Equations and problems
2.3. Special Quaternion Equations and problems
In this section, we give solutions to some specific quaternion equations and least
norm problems.
2.3.1. A quadratic equation
In this subsection, we consider the quadratic equation that was studied and solved
in (Au-Yeung, 2003):
t2 + αt + tβ + γ = 0, (2.13)
where α, β and γ ∈ H and t is a quaternion indeterminate. Here I present a
somewhat simpler solution.
α0 +β0
Let x = t + 2
. Then 2.13 becomes:
!2 ! !
α0 + β0 α 0 + β0 α0 + β0
0 = x− +α x− + x− β+γ
2 2 2
= x2 + ax + xb + c, (2.14)
α0 +β0 2
where a = Im α, b = Im β and c = 2
− (α + β) α0 +β
2
0
+ γ. The advantage
of 2.14 over 2.13 is that the left and right coefficients of the indeterminate are both
pure imaginary.
2.14 is equivalent to the real non-linear system:
2x0 x1 + (−a3 + b3 ) x2 + (a2 − b2 ) x3 = (−a1 − b1 ) x0 − c1 , (1)
(a3 − b3 ) x1 + 2x0 x2 + (−a1 + b1 ) x3 = (−a2 − b2 ) x0 − c2 , (2)
(2.15)
(−a2 + b2 ) x1 + (a1 − b1 ) x2 + 2x0 x3 = (−a3 − b3 ) x0 − c3 ,
(3)
x2 + x2 + x2 + (a + b ) x + (a + b ) x + (a + b ) x = x2 + c0 . (4)
1 2 3 1 1 1 2 2 2 3 3 3 0
20
2.3 Special Quaternion Equations and problems
We first show how to solve 2.15 if a = b. When a = b, 2.15 becomes:
2x0 (x1 + a1 ) = −c1 , (1)
2x0 (x2 + a2 ) = −c2 , (2)
(2.16)
2x0 (x3 + a3 ) = −c3 ,
(3)
x2 + x2 + x2 + 2a x + 2a2 x2 + 2a3 x3 = x20 + c0 . (4)
1 2 3 1 1
x1 = −a1 ,
Case 1. If c = c0 ∈ R, then 2.16 can have solutions only if x0 = 0 or x2 = −a2 ,
x3 =
−a3 ,
is part of the solution.
1. When x0 = 0 is part of the solution, (4) of 2.16 becomes:
x21 + x22 + x23 + 2a1 x1 + 2a2 x2 + 2a3 x3 = c
⇔
3 3
(xi + ai )2 = c + a2i = c − a2 .
X X
(2.17)
i=1 i=1
2.17 can only hold if c − a2 ≥ 0. In which case, x1 , x2 and x3 can
be any real numbers that satisfies 2.17.
x = −a1 ,
1
2. When x2 = −a2 , is part of the solution, (4) of 2.16 becomes:
x3 = −a3 ,
−a21 − a22 − a23 = x20 + c
a2 − c = x20 .
21
2.3 Special Quaternion Equations and problems
The above equation can only hold if and a2 − c ≥ 0. In which case,
√
x0 = ± a2 − c.
Case 2. If c 6= R, then x0 = 0 can never be part of the solution. We obtain, using
the method demonstrated in §2.2, a quartic equation for x0 :
1 2
x40 + c0 + a21 + a22 + a23 x20 − c1 + c22 + c23 = 0.
4
Replacing x20 by y, we obtain a quadratic equation for y:
1 2
f (y) = y 2 + c0 + a21 + a22 + a23 y − c1 + c22 + c23 = 0.
4
The solutions of f are:
q
2
± (c0 + a21 + a22 + a23 ) + (c21 + c22 + c23 ) − (c0 + a21 + a22 + a23 )
y= .
2
Since c 6= R, c21 + c22 + c23 > 0, and therefore
q
2
(c0 + a21 + a22 + a23 ) + (c21 + c22 + c23 ) − (c0 + a21 + a22 + a23 )
>0
2
and
q
2
− (c0 + a21 + a22 + a23 ) + (c21 + c22 + c23 ) − (c0 + a21 + a22 + a23 )
< 0.
2
So
vq
u 2
c0 + a21 + a22 + a23 + c21 + c22 + c23 − c0 + a21 + a22 + a23
u
t
x0 = ±
2
vq
u
2 2
t (c0 − a2 ) − (Im c) − (c0 − a2 )
u
= ,
2
22
2.3 Special Quaternion Equations and problems
and
−c1
x1 = − a1 ,
2x0
−c2
x2 = 2x0
− a2 ,
−c3
x3 = − a3 .
2x0
From now on, we can and we will assume that a 6= b. The deter-
minant
of the coefficient matrix
of (1), (2) and (3) of 2.15, M̌ =
2x0 −a3 + b3 a2 − b2
a3 − b3 −a1 + b1 , is:
2x0
−a2 + b2 a1 − b1 2x0
σ (x0 ) = det M̌
=8x30 + 2 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 x0
=2x0 4x20 + (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 .
2.15 can have infinitely many solutions only if σ (x0 ) = 0, this would
mean that x0 = 0.
First, when x0 = 0, (4) of 2.16 becomes
x21 + x22 + x23 + 2a1 x1 + 2a2 x2 + 2a3 x3 = c
⇔
3 3
2
a2i = c − a2 .
X X
(xi + ai ) = c +
i=1 i=1
The above equation can only hold if c − a2 ≥ 0.
Next, when x0 = 0, the augmented matrix of (1), (2) and (3) of 2.15 is
0 −a3 + b3 a2 − b 2 −c1
A=
a3 − b 3 0 −a1 + b1 .
−c2
−a2 + b2 a1 − b 1 0 −c3
23
2.3 Special Quaternion Equations and problems
We will show that A is the augmented matrix of a consistent linear system if and
only if
3
X
ci (ai − bi ) = 0.
i=1
We will show the forward direction first.
Case 1. If ai − bi 6= 0 for 1 ≤ i ≤ 3. We apply Gaussian elimination to A and
.
obtain (= means row equivalent) :
a3 − b3 0 −a1 + b1 −c2
.
A=
0 −a3 + b3 a2 − b 2 −c1 .
P3
ci (ai −bi )
0 0 0 − i=1
a3 −b3
Since A is the augmented matrix of a consistent linear system,
P3
i=1 ci (ai − bi ) = 0.
Case 2. If a1 − b1 = 0, ai − bi 6= 0 for i = 2, 3, then
0 −a3 + b3 a2 − b2 −c1
A=
a3 − b 3 0 0 .
−c2
−a2 + b2 0 0 −c3
Since A is the augmented matrix of a consistent linear system, the last
two rows give:
(a3 − b3 ) x 1 = −c2
+ b2 ) x1 = −c3
(−a2
(a2 − b2 ) c2 + (a3 − b3 ) c3 = 0
⇒
3
X
ci (ai − bi ) = 0.
i=1
24
2.3 Special Quaternion Equations and problems
Similarly, we can show the same result when a2 − b2 = 0, ai − bi 6= 0 for
i = 1, 3 and a3 − b3 = 0, ai − bi 6= 0 for i = 1, 2.
Case 3. If a1 − b1 6= 0, ai − bi = 0 for i = 2, 3, then
0 0 0 −c1
A= 0
0 −a1 + b1 .
−c2
0 a1 − b 1 0 −c3
Since A is the augmented matrix of a consistent linear system, the first
P3
row gives c1 = 0, and therefore i=1 ci (ai − bi ) = 0. Similarly, we can
show the same result when a2 − b2 6= 0, ai − bi = 0 for i = 1, 3 and
a3 − b3 6= 0, ai − bi = 0 for i = 1, 2.
P3
For the converse, we assume that i=1 ci (ai − bi ) = 0.
Case 1. If a3 − b3 6= 0, then we apply Gaussian elimination to A to obtain:
a3 − b3 0 −a1 + b1 −c2
.
A = 0 −a3 + b3 a2 − b 2 −c1
P3
ci (ai −bi )
0 0 0 − i=1
a3 −b3
a3 − b3 0 −a1 + b1 −c2
=
0 −a3 + b3 a2 − b 2 .
−c1
0 0 0 0
A is always the augmented matrix of a consistent linear system as it has
the solution set
−c2 +(a1 −b1 )s
x1 = ,
a3 −b3
c1 +(a2 −b2 )s
x2 = a3 −b3
,
x3 = s,
where s is a parameter. In order for 2.15 to have a solution, we must
25
2.3 Special Quaternion Equations and problems
have c − a2 ≥ 0 and s must be a solution of
−c2 + (a1 − b1 ) s 2 c1 + (a2 − b2 ) s 2
0= + + s2 − c0
a3 − b3 a3 − b3
−c2 + (a1 − b1 ) s c1 + (a2 − b2 ) s
+ (a1 + b1 ) + (a2 + b2 ) + (a3 + b3 ) s
a3 − b3 a3 − b3
⇔
3
!
2
X
0= (ai − bi ) s2 + O (s) , (2.18)
i=1
where O (s) is a complicated linear polynomial of s. The right hand side
of 2.18 is not identically equal to zero as the coefficient of s2 is nonzero,
and thus it can have at most 2 solutions and therefore 2.15 can have at
most 2 solutions.
Case 2. If a3 − b3 = 0 and a2 − b2 = 0, then a1 − b1 6= 0 because a 6= b.
P3
i=1 ci (ai − bi ) = 0 implies that c1 = 0 and we have:
0 0 0 0
A= 0
0 −a1 + b1 .
−c2
0 a1 − b 1 0 −c3
This system is always consistent as it has the solution set
x = s,
1
c2
x2 = a1 −b1
,
−c3
x3 = ,
a1 −b1
where s is a parameter. In order for 2.15 to have a solution, we must
26
2.3 Special Quaternion Equations and problems
have c − a2 ≥ 0 and s must be a solution of
2 2
c2 −c3
0 =s2 + + − c0
a1 − b 1 a1 − b 1
c2 −c3
+ (a1 + b1 ) s + (a2 + b2 ) + (a3 + b3 )
a1 − b 1 a1 − b 1
⇔
0 = (a1 − b1 )2 s2 + O (s) , (2.19)
where O (s) is a complicated linear polynomial of s. The right hand side
of 2.19 is not identically equal to zero as the coefficient of s2 is nonzero,
and therefore it can have at most 2 solutions and therefore 2.15 can have
at most 2 solutions.
Case 3. If a3 − b3 = 0 and a2 − b2 6= 0, then
0 0 a2 − b 2 −c1
A=
0 0 −a1 + b1 .
−c2
−a2 + b2 a1 − b1 0 −c3
P3
i=1 ci (ai − bi ) = 0 implies that c2 (a2 − b2 ) = −c1 (a1 − b1 ) and there-
fore the first two rows are consistent. Furthermore, this system is always
consistent as it has the solution set
c3 +(a1 −b1 )s
x1 = ,
a2 −b2
x2 = s,
−c1
x3 = ,
a2 −b2
where s is a parameter. In order for 2.15 to have a solution, we must
27
2.3 Special Quaternion Equations and problems
have c − a2 ≥ 0 and s must be a solution of
c3 + (a1 − b1 ) s 2 −c1 2
0= + s2 + − c0
a2 − b2 a2 − b2
c3 + (a1 − b1 ) s −c1
+ (a1 + b1 ) + (a2 + b2 ) s + (a3 + b3 )
a2 − b2 a2 − b2
⇔
0 = (a1 − b1 )2 + (a2 − b2 )2 s2 + O (s) , (2.20)
where O (s) is a complicated linear polynomial of s. The right hand side
of 2.20 is not identically equal to zero as the coefficient of s2 is nonzero,
and therefore it can have at most 2 solutions and therefore 2.15 can have
at most 2 solutions.
This concludes our discussion for possible scenarios where 2.15 can have infinitely
many solutions.
When x0 6= 0, we have σ (x0 ) 6= 0. Applying Cramer’s rule to (1), (2) and (3) of
2.15, we obtain x1 , x2 and x3 as rational polynomials of x0 as follows:
p1 (x0 )
x1 = , (2.21)
σ (x0 )
p2 (x0 )
x2 = (2.22)
σ (x0 )
and
p3 (x0 )
x3 = , (2.23)
σ (x0 )
28
2.3 Special Quaternion Equations and problems
where
p1 (x0 ) = − 4 (a1 + b1 ) x30 + 4 (a2 b3 − a3 b2 − c1 ) x20
+ (b1 − a1 ) a21 + a22 + a23 − b21 − b22 − b23 x0
+ 2 (c2 (b3 − a3 ) + c3 (a2 − b2 )) x0
+ (b1 − a1 ) (c1 (a1 − b1 ) + c2 (a2 − b2 ) + c3 (a3 − b3 )) ,
p2 (x0 ) = − 4 (a2 + b2 ) x30 + 4 (a3 b1 − a1 b3 − c2 ) x20
+ (b2 − a2 ) a21 + a22 + a23 − b21 − b22 − b23 x0
+ 2 (c1 (a3 − b3 ) + c3 (b1 − a1 )) x0
+ (b2 − a2 ) (c1 (a1 − b1 ) + c2 (a2 − b2 ) + c3 (a3 − b3 ))
and
p3 (x0 ) = − 4 (a3 + b3 ) x30 + 4 (a1 b2 − a2 b1 − c3 ) x20
+ (b3 − a3 ) a21 + a22 + a23 − b21 − b22 − b23 x0
+ 2 (c1 (b2 − a2 ) + c2 (a1 − b1 )) x0
+ (b3 − a3 ) (c1 (a1 − b1 ) + c2 (a2 − b2 ) + c3 (a3 − b3 )) .
Substituting 2.21, 2.23 and 2.23 back into (4) of 2.15, we obtain:
p1 (x0 )2 + p2 (x0 )2 + p3 (x0 )2
x20
0 = − − c0 +
σ (x0 )2
(a1 + b1 ) p1 (x0 ) + (a2 + b2 ) p2 (x0 ) + (a3 + b3 ) p3 (x0 )
+ .
σ (x0 )
29
2.3 Special Quaternion Equations and problems
Multiplication of the above equation by σ (x0 )2 yields a 6th degree equation of x0 :
0 =16x60 + 8 a21 + a22 + a23 + b21 + b22 + b23 + 2c0 x40
+ 4c0 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 x20
2
+ a21 + a22 + a23 − b21 − b22 − b23 x20
+ (8a2 b3 − 8a3 b2 − 4c1 ) c1 x20
+ (8a3 b1 − 8a1 b3 − 4c2 ) c2 x20
+ (8a1 b2 − 8a2 b1 − 4c3 ) c3 x20
3
!2
X
− ci (ai − bi ) .
i=1
Replacing x20 by y, we obtain a cubic equation of y:
0 =f (y)
=16y 3 + 8 a21 + a22 + a23 + b21 + b22 + b23 + 2c0 y 2
+ 4c0 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 y
2
+ a21 + a22 + a23 − b21 − b22 − b23 y
+ (8a2 b3 − 8a3 b2 ) c1 − 4c21 y
+ (8a3 b1 − 8a1 b3 ) c2 − 4c22 y
+ (8a1 b2 − 8a2 b1 ) c3 − 4c23 y
3
!2
X
− ci (ai − bi )
i=1
=16y 3 + D2 y 2 + D1 y + D0 . (2.24)
We next show that f has exactly one strictly positive solution.
P 2
3
Since x0 6= 0, D0 = − i=1 ci (ai − bi ) < 0. By Descartes’ rule of signs, f has
at least one strictly positive solution. Furthermore, f can have multiple positive
solutions only if D2 < 0 and D1 > 0. We show that this is impossible.
30
2.3 Special Quaternion Equations and problems
a21 +a22 +a23 +b21 +b22 +b23
If D2 < 0, then c0 < − 2
. Therefore,
2
4c0 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 + a21 + a22 + a23 − b21 − b22 − b23
+ (2a2 b3 − 2a3 b2 )2 + (2a3 b1 − 2a1 b3 )2 + (2a1 b2 − 2a2 b1 )2
< − 2 a21 + a22 + a23 + b21 + b22 + b23 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2
2
+ a21 + a22 + a23 − b21 − b22 − b23
+ (2a2 b3 − 2a3 b2 )2 + (2a3 b1 − 2a1 b3 )2 + (2a1 b2 − 2a2 b1 )2
2
= − (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 < 0.
This gives:
2
4c0 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 + a21 + a22 + a23 − b21 − b22 − b23
< − (2a2 b3 − 2a3 b2 )2 − (2a3 b1 − 2a1 b3 )2 − (2a1 b2 − 2a2 b1 )2 .
Then D1 can be simplified as follows:
2
D1 =4c0 (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2 + a21 + a22 + a23 − b21 − b22 − b23
+ (8a2 b3 − 8a3 b2 ) c1 − 4c21 + (8a3 b1 − 8a1 b3 ) c2 − 4c22 + (8a1 b2 − 8a2 b1 ) c3 − 4c23
< − (2a2 b3 − 2a3 b2 )2 − (2a3 b1 − 2a1 b3 )2 − (2a1 b2 − 2a2 b1 )2
+ (8a2 b3 − 8a3 b2 ) c1 − 4c21 + (8a3 b1 − 8a1 b3 ) c2 − 4c22 + (8a1 b2 − 8a2 b1 ) c3 − 4c23
= − (2a2 b3 − 2a3 b2 − 2c1 )2 − (2a3 b1 − 2a1 b3 − 2c2 )2 − (2a1 b2 − 2a2 b1 − 2c3 )2
≤0.
√
Therefore, f has exactly one strictly positive solution z. We can plug x0 = ± z
into 2.21, 2.23 and 2.23 to get x1 , x2 and x3 respectively.
In all of the above discussion, solutions to 2.13 can be obtained by calculating
31
2.3 Special Quaternion Equations and problems
α0 +β0
t0 = x0 − 2 ,
t1 = x1 ,
t = x − α0 +β
2
0
= x0 − α0 +β0
2
+ x1 i + x2 j + x3 k, that is, for each
t2 = x2 ,
t3 = x3 ,
x.
We are now able to give a theorem that sums up our result so far in this subsection.
Theorem 2.3.1. The solutions of the quadratic equation
t2 + αt + tβ + γ = 0, (2.25)
where α, β and γ ∈ H and t is a quaternion indeterminate, can be obtained by
α0 +β0 2
formulas according to the following cases, let a = Im α, b = Im β and c = 2
−
(α + β) α0 +β
2
0
+ γ. In the following, we call
√ α0 +β0 √ α0 +β0
t = s− , t = − s− ,
0 0
2 2
√ √
p 1 ( s) p1 (− s)
= =
t1 √ ,
t1 √ ,
σ ( s) σ (− s)
√ and √
p 2 ( s) p2 (− s)
t2 = √ ,
t2 = √ ,
σ ( s) σ (− s)
√ √
p 3 ( s) p3 (− s)
t = √ , t3 = √ ,
3 σ ( s) σ (− s)
where s is the unique positive root of 2.24, the standard solutions.
1. a = b.
a) c ∈ R.
i. c − a2 > 0. t0 = − α0 +β
2
0
and (t1 , t2 , t3 ) can be any real number triple
(ti + ai )2 = c − a2 . This is the only
P3
that satisfies the equation i=1
case where 2.25 has infinitely many solutions.
32
2.3 Special Quaternion Equations and problems
ii. a2 − c ≥ 0.
√ α0 +β0
t0 = ± a2 − c − ,
2
t1 = −a1 ,
= −a2 ,
t2
t = −a3 .
3
b) c ∈
/ R.
α0 +β0
t0 = s− ,
2
c1
t1 = −a1 − ,
2s
c2
−a2 −
t2 = 2s
,
c2
t = −a2 − ,
3 2s
or
α0 +β0
t0 = −s − ,
2
c1
t1 = −a1 + ,
2s
c2
= −a2 +
t2 2s
,
c2
t = −a2 + ,
3 2s
r√
(c0 −a2 )2 −(Im c)2 −(c0 −a2 )
where = 2
.
2. a 6= b. Standard solutions always exist. Furthermore,
P3
a) i=1 ci (ai − bi ) = 0 and c − a2 ≥ 0.
33
2.3 Special Quaternion Equations and problems
i. a3 6= b3 . Possible solutions must be of the form
t0 = − α0 +β 0
,
2
−c2 +(a1 −b1 )s
t1 = ,
a3 −b3
c1 +(a2 −b2 )s
t2 = a3 −b3
,
t = s,
3
where s is a real number that satisfies the quadratic equation 2.18
ii. a3 = b3 and a2 = b2 . Possible solutions must be of the form
t0 = − α0 +β 0
,
2
t1 = s,
c2
t2 = a1 −b1
,
−c3
t = ,
3 a1 −b1
where s is a real number that satisfies the quadratic equation 2.19
iii. a3 = b3 and a2 6= b2 . Possible solutions must be of the form
t0 = − α0 +β 0
,
2
c3 +(a1 −b1 )s
t1 = ,
a2 −b2
t2 = s,
−c1
t = ,
3 a2 −b2
where s is a real number that satisfies the quadratic equation 2.20
b) Otherwise. No additional solutions exist.
Example 2.3.2. Find all the solutions of
t2 + (1 − 2i − 2j − 2k) t + t (1 + i − 3j + k) + (−1 − 3i + j − 2k) = 0. (2.26)
34
2.3 Special Quaternion Equations and problems
The following steps have been implemented into a single procedure in Maple. For
details, see the “AY” procedure of §A.2.
Let x = t + 1. Then 2.26 becomes
x2 + (−2i − 2j − 2k) x + x (i − 3j + k) + (−2 − 2i + 6j − k) = 0. (2.27)
We have
3
X
ci (ai − bi ) = (−2) · (−2 − 1) + 6 · (−2 + 3) + (−1) · (−2 − 1) = 15 6= 0.
i=1
By Theorem 2.3.1, there are only standard solutions and we obtain a cubic poly-
nomial of y = x20 :
f (y) = 16y 3 + 152y 2 − 251y − 225.
f (y) has one positive root
13 √ 1 18 √ 1 19
y= 13 sin arctan 10091901 + π − = s.
6 3 85634 6 6
The two solutions to 2.26 are
√
√ √ √
p1 ( s) p2 ( s) p3 ( s)
s−1 + √ i+ √ j+ √ k ≈ 0.41 + 0.8i + 1.14k
σ ( s) σ ( s) σ ( s)
and
√ √ √
√ p1 (− s) p2 (− s) p3 (− s)
− s−1 + √ i+ √ j+ √ k ≈ −2.41 + 0.87i + 1.65j − 1.93k
σ (− s) σ (− s) σ (− s)
These answers are verified to be correct by direct claculation.
35
2.3 Special Quaternion Equations and problems
2.3.2. Another quadratic equation
In this subsection, we consider another quadratic equation:
x2 + axb + c = 0, (2.28)
where a and b ∈ H with a0 = b0 = 0, c ∈ R and x is a quaternion indeterminate.
We assume that ab 6= 0. Otherwise, 2.28 is reduced to x2 + c = 0, which was studied
and solved in (Niven, 1942). We also assume that c 6= 0. Otherwise, 2.28 is reduced
to x2 + axb = 0, which we will discuss in §2.3.3.
2.28 is equivalent to the real non-linear system:
(2x0 + m11 ) x1 + m12 x2 + m13 x3 = (−a2 b3 + a3 b2 ) x0 , (1)
m21 x1 + (2x0 + m22 ) x2 + m23 x3 = (a1 b3 − a3 b1 ) x0 , (2)
(2.29)
m31 x1 + m32 x2 + (2x0 + m33 ) x3 = (−a1 b2 + a2 b1 ) x0 , (3)
x2 − x2 − x2 − x2 + d + d2 x1 + d3 x2 + d4 x3 + c = 0, (4)
1 x0
0 1 2 3
where mij and di are given by:
M = (mij )3×3
−a1 b1 + a2 b2 + a3 b3 −a1 b2 − a2 b1 −a1 b3 − a3 b1
=
−a1 b2 − a2 b1 a1 b 1 − a2 b 2 + a3 b 3 −a2 b3 − a3 b2 ,
−a1 b3 − a3 b1 −a2 b3 − a3 b2 a1 b1 + a2 b2 − a3 b3
and
−a1 b1 − a2 b2 − a3 b3
a2 b 3 − a3 b 2
D = (di )4×1 =
.
−a1 b3 + a3 b1
a1 b 2 − a2 b 1
We already discussed in §2.2 how to solve 2.28 if the determinant of the coefficient
36
2.3 Special Quaternion Equations and problems
2x0 + m11 m12 m13
matrix, M̌ =
m21 2x0 + m22 m23 ,
is zero. We can and will assume
m31 m32 2x0 + m33
that det M̌ 6= 0.
σ (x0 ) = det M̌
= (b1 a1 + a2 b2 + a3 b3 + 2x0 ) 4x20 − a21 + a22 + a23 b21 + b22 + b23 6= 0,
therefore, 4x20 − (a21 + a22 + a23 ) (b21 + b22 + b23 ) 6= 0. Applying Cramer’s rule to (1), (2)
and (3) of 2.29, we obtain x1 as a rational polynomial of x0 ,
[x0 (a3 b2 − a2 b3 )] (4x20 − (a21 + a22 + a23 ) (b21 + b22 + b23 ))
x1 =
(b1 a1 + a2 b2 + a3 b3 + 2x0 ) (4x20 − (a21 + a22 + a23 ) (b21 + b22 + b23 ))
x0 (a3 b2 − a2 b3 )
= = p1 (x0 ) . (2.30)
b1 a1 + a2 b2 + a3 b3 + 2x0
Similarly,
x0 (a1 b3 − a3 b1 )
x2 = = p2 (x0 ) (2.31)
b1 a1 + a2 b2 + a3 b3 + 2x0
and
x0 (a2 b1 − a1 b2 )
x3 = = p3 (x0 ) . (2.32)
b1 a1 + a2 b2 + a3 b3 + 2x0
Substituting 2.30, 2.31 and 2.32 back into (4) of 2.29, then multiplying through by
(b1 a1 + a2 b2 + a3 b3 + 2x0 )2 , we obtain a quartic equation of x0 ,
3 1 1
f (x0 ) = x40 + c − α x20 + β (4c − α) x0 + cβ 2 = 0, (2.33)
4 4 4
where
α = a21 + a22 + a23 b21 + b22 + b23
and
β = a1 b 1 + a2 b 2 + a3 b 3 .
37
2.3 Special Quaternion Equations and problems
The discriminant of f is 4f = 1
256
β2 (α − β 2 ) (α + 4c) (3α − 4c)3 − 1024c3 β 2 .
Since α − β 2 ≥ 0 by the Cauchy–Schwartz inequality, 4f < 0 if an only if
(α + 4c) (3α − 4c)3 − 1024c3 β 2 < 0.
Once we solve f for x0 , we substitute its value back into 2.30, 2.31 and 2.32 to get
x1 , x2 and x3 respectively to obtain a solution to 2.28.
Example 2.3.3. Find all the solutions of
x2 + (−i − 3j + 2k) x (−2i + j − k) + 21 = 0. (2.34)
The following steps have been implemented into a single procedure in Maple. For
details, see the “AXBc” procedure of §A.2.
2.34 is equivalent to
(2x0 − 7) x1 − 5x2 + 3x3 = −x0 , (1)
−5x1 + (2x0 + 3) x2 − 5x3 = 5x0 , (2)
(2.35)
3x1 − 5x2 + (2x0
+ 1) x3 = 7x0 , (3)
x2 + x2 + x2 − x + 5x2 + 7x3 = x20 + 21. (4)
1 2 3 1
2x0 − 7 −5 3
When the determinant of the coefficient matrix, −5 −5 , is 0,
2x0 + 3
3 −5 2x0 + 1
that is, when σ (x0 ) = 8x30 − 12x20 − 168x0 + 252 =
0, 2.35 has no solutions because
2x0 − 7 −5 3 −x0
the last row of the reduced row echelon form of −5
2x0 + 3 −5 5x0
3 −5 2x0 + 1 7x0
is not a zero row. Otherwise, by 2.33, we obtain a quartic polynomial of x0 ,
189
f (x0 ) = x40 − 42x20 +
4
q √
f (x0 ) has four positive roots x0 = ± 21 84 ± 30 7. Therefore, by 2.30, 2.31 and
38
2.3 Special Quaternion Equations and problems
2.32, the four solutions to 2.34 are
q √ q √ q √ q √
84 + 30 7 84 + 30 7 84 + 30 7 84 + 30 7
+ p1 i + p2 j + p3 k
2 2 2 2
≈6.39 − 0.65i + 3.27j + 4.57k,
q √ q √ q √ q √
84 − 30 7 84 − 30 7 84 − 30 7 84 − 30 7
+ p1 i + p2 j + p3 k
2 2 2 2
≈1.08 + 1.27i − 6.34j − 8.87k,
q √ q √ q √ q √
− 84 + 30 7 − 84 + 30 7 − 84 + 30 7 − 84 + 30 7
+ p1 i + p2 j + p3 k
2 2 2 2
≈ − 6.39 − 0.40i + 2.02j + 2.83k,
and
q √ q √ q √ q √
− 84 − 30 7 − 84 − 30 7 − 84 − 30 7 − 84 − 30 7
+ p1 i + p2 j + p3 k
2 2 2 2
≈ − 1.08 − 0.21i + 1.04j + 1.46k
These answers are verified to be correct by direct claculation.
39
2.3 Special Quaternion Equations and problems
2.3.3. A homogeneous quadratic equation
In this subsection, we consider the currently unsolved homogeneous quadratic equa-
tion:
x2 + axb = 0, (2.36)
where a and b ∈ H and x is the quaternion unknown. We assume that a ∈
/ R and
b∈
/ R. Otherwise, 2.36 is reduced to x (x + ab) = 0 or (x + ab) x = 0.
2.36 is equivalent to the real non-linear system:
(2x0 + m11 ) x1 + m12 x2 + m13 x3 = x0 n1 , (1)
m21 x1 + (2x0 + m22 ) x2 + m23 x3 = x0 n2 , (2)
(2.37)
m31 x1 + m32 x2 + (2x0 + m33 ) x3 = x0 n3 , (3)
x2 − x2 − x2 − x2 + d 1 x0 + d2 x1 + d3 x2 + d4 x3 + c0 = 0, (4)
0 1 2 3
where mij , ni and di are given by:
M = (mij )3×3
a0 b 0 − a1 b 1 + a2 b 2 + a3 b 3 a0 b 3 − a1 b 2 − a2 b 1 − a3 b 0 −a0 b2 − a1 b3 + a2 b0 − a3 b1
=
−a0 b3 − a1 b2 − a2 b1 + a3 b0 a0 b0 + a1 b1 − a2 b2 + a3 b3 ,
a0 b 1 − a1 b 0 − a2 b 3 − a3 b 2
a0 b 2 − a1 b 3 − a2 b 0 − a3 b 1 −a0 b1 + a1 b0 − a2 b3 − a3 b2 a0 b 0 + a1 b 1 + a2 b 2 − a3 b 3
−a0 b1 − a1 b 0 − a2 b 3 + a3 b 2
N = (ni )3×1 = −a b
0 2 ,
+ a1 b 3 − a2 b 0 − a3 b 1
−a0 b3 − a1 b2 + a2 b1 − a3 b0
40
2.3 Special Quaternion Equations and problems
and
a0 b 0 − a1 b 1 − a2 b 2 − a3 b 3
−a b − a b + a b − a b
0 1 1 0 2 3 3 2
D = (di )4×1 =
.
−a b − a b − a b + a b
0 2 1 3 2 0 3 1
−a0 b3 + a1 b2 − a2 b1 − a3 b0
We already showed in §2.2 how
to solve such an equation when the
determinant of
2x0 + m11 m12 m13
the coefficient matrix, M̌ =
m21 2x0 + m22 m23 ,
is zero. We can
m31 m32 2x0 + m33
and will assume that det M̌ 6= 0. Direct calculation yields:
σ (x0 ) = det M̌
=8x30 + 4(m11 + m22 + m33 )x20
+ 2 (m11 m22 + m11 m33 − m12 m21 − m13 m31 − m23 m32 + m22 m33 ) x0
+ m11 m22 m33 − m11 m23 m32 − m12 m21 m33
+ m13 m21 m32 + m12 m23 m31 − m13 m22 m31 .
Applying Cramer’s rule to (1), (2) and (3) of 2.37, we obtain x1 , x2 and x3 as rational
polynomials of x0 as follows:
p1 (x0 )
x1 = , (2.38)
σ (x0 )
p2 (x0 )
x2 = (2.39)
σ (x0 )
and
p3 (x0 )
x3 = , (2.40)
σ (x0 )
41
2.3 Special Quaternion Equations and problems
where
p1 (x0 ) =4n1 x30 + 2 (n1 (m22 + m33 ) − n2 m12 − n3 m13 ) x20
+ n1 (m22 m33 − m23 m32 ) x0
+ n2 (m13 m32 − m12 m33 ) x0
+ n3 (m12 m23 − m13 m22 ) x0 ,
p2 (x0 ) =4n2 x30 + 2 (n2 (m11 + m33 ) − n1 m21 − n3 m23 ) x20
+ n1 (m23 m31 − m21 m33 ) x0
+ n2 (m11 m33 − m13 m31 ) x0
+ n3 (m13 m21 − m11 m23 ) x0
and
p3 (x0 ) =4n3 x30 + 2 (n3 (m11 + m22 ) − n1 m31 − n2 m32 ) x20
+ n1 (m21 m32 − m22 m31 ) x0
+ n2 (m12 m31 − m11 m32 ) x0
+ n3 (m11 m22 − m12 m21 ) x0 .
Substituting 2.38, 2.39 and 2.40 back into (4) of 2.37, we obtain:
0 =x20 + d1 x0
p1 (x0 )2 + p2 (x0 )2 + p3 (x0 )2 d2 p1 (x0 ) + d3 p2 (x0 ) + d4 p3 (x0 )
− + ,
σ (x0 )2 σ (x0 )
Multiplication of the above equation by σ (x0 )2 yields an 8th degree polynomial of
42
2.3 Special Quaternion Equations and problems
x0 ,
0 =f (x0 ) (2.41)
=σ (x0 )2 x20 + d1 x0 − d21 x20 − p2 (x0 )2 − p3 (x0 )2
+ σ (x0 ) (d2 p1 (x0 ) + d3 p2 (x0 ) + d4 p3 (x0 )) .
2.41 can be factored into the product of x0 , a cubic and a quartic polynomial of x0 ,
f (x0 ) = x0 h (x0 ) g (x0 ) ,
where the cubic polynomial is
h (x0 ) =x30 + 2a0 b0 x20 (2.42)
4 2 2 1 2
+ a0 b 0 − a0 − 3a21 − 3a22 − 3a23 b20 − 3b21 − 3b22 − 3b23 x0
3 12
1
+ B (a0 b0 − a1 b1 − a2 b2 − a3 b3 ) ,
4
and the quartic polynomial is
1 1 1
g (x0 ) = x40 + 2a0 b0 x30 + Ax20 + a0 b0 Bx0 + B 2 , (2.43)
2 2 16
with
A = 4a20 b20 − a21 + a22 + a23 − a20 b21 + b22 + b23 − b20
and
B = a20 + a21 + a22 + a23 b20 + b21 + b22 + b23 .
Furthermore, g (x0 ) can be factored in R as
g (x0 ) = g1 (x0 ) g2 (x0 ) ,
43
2.3 Special Quaternion Equations and problems
where
1
g1 (x0 ) = x20 + sx0 + B
4
and
1
g2 (x0 ) = x20 + tx0 + B,
4
with
q
s = a0 b 0 + (b21 + b22 + b23 ) (a21 + a22 + a23 )
and
q
t = a0 b 0 − (b21 + b22 + b23 ) (a21 + a22 + a23 ).
The discriminants of g1 and g2 are
4g1 = − b23 a20 − a21 b20 − a22 b20 − a23 b20 − b22 a20 − a20 b21
q
+ 2b0 a0 (b21 + b22 + b23 ) (a21 + a22 + a23 )
q q 2
= − a0 (b21 + b22 + b23 ) − b0 (a21 + a22 + a23 ) ≤0
and
4g2 = − b23 a20 − a21 b20 − a22 b20 − a23 b20 − b22 a20 − a20 b21
q
− 2b0 a0 (b21 + b22 + b23 ) (a21 + a22 + a23 )
q q 2
= − a0 (b21 + b22 + b23 ) + b0 (a21 + a22 + a23 ) ≤ 0.
We continue the discussion by cases:
Case 1. When a0 = b0 = 0.
44
2.3 Special Quaternion Equations and problems
Then h (x0 ), g (x0 ) and σ (x0 ) can be simplified to:
1 2
h (x0 ) =x30 − 3a1 + 3a22 + 3a23 3b21 + 3b22 + 3b23 x0 (2.44)
12
1 2
− a1 + a22 + a23 b21 + b22 + b23 (a1 b1 + a2 b2 + a3 b3 ) ,
4
2
1 2
g (x0 ) = x20 − a1 + a22 + a23 b21 + b22 + b23 (2.45)
4
and
1 2
σ (x0 ) = 4 (2x0 + a1 b1 + a2 b2 + a3 b3 ) x20 − a1 + a22 + a23 b21 + b22 + b23 .
4
(2.46)
Note that we already assumed that σ (x0 ) 6= 0, and therefore 2.45 does
not contribute any roots. We next show that 2.44 always has three real
solutions. The discriminant of h is
27 2 2 2
4h = a1 + a22 + a23 b21 + b22 + b23
16
× (a1 b2 − a2 b1 )2 + (a1 b3 − a3 b1 )2 + (a2 b3 − a3 b2 )2
≥0.
So the solution to f (x0 ) are 0 and the 3 real roots of 2.44.
q
a0
(a21 +a22 +a23 )
Case 2. Otherwise. 4g1 < 0 unless b0
= q , that is, a0 = cb0 and
(b21 +b22 +b23 )
a21 + a22 + a23 = c2 (b21 + b22 + b23 ) for some 0 < c. Under these conditions,
4g2 < 0 and therefore g2 does not contribute any solutions. However, g1
can be simplified to
1 2
x20 + c b20 + b21 + b22 + b23 x0 + c2 b20 + b21 + b22 + b23 . (2.47)
4
q
a0
(a21 +a22 +a23 )
Similarly, 4g2 < 0 unless = −q
b0
= c for some c < 0. Under
(b21 +b22 +b23 )
these conditions, 4g1 < 0 and therefore g1 does not contribute any solu-
45
2.3 Special Quaternion Equations and problems
tions. However g2 can also be simplified to 2.47. So the roots of f (x0 )
a0 (b20 +b21 +b22 +b23 )
are 0, the real root(s) of 2.42 and − 2b0
which has multiplicity
2 and is contributed by 2.47.
Once we obtain x0 , we substitute its value back into 2.38, 2.39 and 2.40 to get x1 ,
x2 and x3 respectively to obtain a solution to 2.36.
Example 2.3.4. Find all the solutions of
x2 + (−19 + 15i − j + 11k) x (9 + 10i − 4j + 10k) = 0. (2.48)
The following steps have been implemented into a single procedure in Maple. For
details, see the “AXB” procedure of §A.2.
2.48 is equivalent to
(2x0 − 207) x1 − 219x2 − 345x3 = 21x0 , (1)
359x1 + (2x0 + 85) x2 − 271x3 = −27x0 , (2)
(2.49)
−175x1 + 379x2 + (2x0 − 127) x3 =
141x0 , (3)
x2 + x2 + x2 − 89x + 107x − 41x = x20 − 435x0 . (4)
1 2 3 1 2 3
When the determinant of the coefficient matrix,
2x0 − 207 −219 −345
359 2x0 + 85 −271
−175 379 2x0 − 127
is 0, that is, when σ (x0 ) = 8x30 − 996x20 + 237708x0 − 91470060 = 0, 2.49 has no
solutions because the last row of the reduced row echelon form of
2x0 − 207 −219 −345 21x0
359 2x0 + 85 −271 −27x0
−175 379 2x0 − 127 141x0
46
2.3 Special Quaternion Equations and problems
is not a zero row. Otherwise, by 2.42, we obtain a cubic polynomial of x0 ,
f (x0 ) = x3 − 342x2 + 6858x − 22867515
f (x0 ) has one real root:
3 √ 1 7140
x0 = 3710932 + 4 853947883289 3 + √ 1 + 114
2
3710932 + 4 853947883289 3
=δ.
Therefore, by 2.38, 2.39 and 2.40, the solutions to 2.48 are 0 and
p1 (δ) p2 (δ) p3 (δ)
δ+ i+ j+ k
σ (δ) σ (δ) σ (δ)
≈443.03 + 62.3i − 7.35j + 100.33k.
This answer is verified to be correct by direct claculation.
47
2.3 Special Quaternion Equations and problems
2.3.4. A pair of quaternion equations
Two nonzero quaternions a and b are said to be semisimilar if the following system
has a solution:
yax = b,
(2.50)
xby
= a,
where x and y are nonzero quaternion indeterminates. 2.50 was studied and solved
in (Tian, 2010). In this subsection, we present a somewhat simpler solution.
2.50 can be written as
= bx−1 a−1 ,
y
= b−1 x−1 a,
y
this gives:
bx−1 a−1 = b−1 x−1 a (2.51)
|b| a−1 = b−1 |a| ⇒ |a| = |b| .
Therefore, a and b must satisfy the following equation for 2.55 to have solutions,
a20 + a21 + a22 + a23 = b20 + b21 + b22 + b23 . (2.52)
2.51 is equivalent to xb2 = a2 x, which is equivalent to the real linear homogeneous
system:
m11 x1 + m12 x2 + m13 x3 + m14 x0 = 0,
m21 x1 + m22 x2 + m23 x3 + m24 x0 = 0,
(2.53)
m31 x1
+ m32 x2 + m33 x3 + m34 x0 = 0,
m x + m42 x2 + m43 x3 + m44 x0 = 0.
41 1
48
2.3 Special Quaternion Equations and problems
where mij is given by the 4 × 4 skew-symmetric matrix:
M = (mij )4×4
d −2 (a0 a3 + b0 b3 ) 2 (a0 a2 + b0 b2 ) 2 (a0 a1 − b0 b1 )
2 (a a + b b )
0 3 0 3 d −2 (a0 a1 + b0 b1 ) 2 (a0 a2 − b0 b2 )
,
−2 (a a + b b ) 2 (a0 a1 + b0 b1 ) d 2 (a0 a3 − b0 b3 )
0 2 0 2
−2 (a0 a1 − b0 b1 ) −2 (a0 a2 − b0 b2 ) −2 (a0 a3 − b0 b3 ) d
where the diagonal entries are d = a20 − a21 − a22 − a23 − b20 + b21 + b22 + b23 . By 2.52,
d = a20 − b20 + a20 − b20 = 2a20 − 2b20 .
Therefore, M becomes:
a20
− b20 − (a0 a3 + b0 b3 ) a0 a2 + b 0 b 2 a0 a1 − b 0 b 1
a a +b b
0 3 0 3 a20 − b20 − (a0 a1 + b0 b1 ) a0 a2 − b 0 b 2
.
− (a a + b b )
0 2 0 2 a0 a1 + b 0 b 1 a20 − b20 a0 a3 − b 0 b 3
− (a0 a1 − b0 b1 ) − (a0 a2 − b0 b2 ) − (a0 a3 − b0 b3 ) a20 − b20
We continue the discussion by cases.
Case 1. If a0 = b0 or a0 = −b0 , then det M = 0. Therefore, the homoge-
neous system 2.53 has a solution set with an infinite number of solutions
(x0 , x1 , x2 , x3 ). Each such x gives a solution to 2.50,
x = x,
= bx−1 a−1 .
y
Case 2. Otherwise, we show that 2.50 has no solutions. Let A = a20 + a21 + a22 + a23
and B = b20 + b21 + b22 + b23 . Direct calculation yields:
det M =a20 b20 (A − B) A − B − 4a20 + 4b20
+ (a0 + b0 ) (a0 − b0 ) (a0 A − b0 B) (a0 A + b0 B) .
49
2.3 Special Quaternion Equations and problems
A = B by 2.52 and therefore,
det M =0 + (a0 + b0 ) (a0 − b0 ) A (a0 − b0 ) A (a0 + b0 )
= (a0 + b0 )2 (a0 − b0 )2 A2 .
Since a 6= 0, A 6= 0 and therefore det M 6= 0. This implies that the
homogeneous system 2.53 has only trivial solutions (x0 , x1 , x2 , x3 ) =
(0, 0, 0, 0). A contradiction. Therefore, 2.50 has no solutions.
Example 2.3.5. Find all (x, y) such that the following system holds
y (2 + 5i + 12j − 5k) x = 2 + 4i − 3j + 13k,
(2.54)
x (2 + 4i − 3j + 13k) y = 2 + 5i + 12j − 5k.
The following steps have been implemented into a single procedure in Maple. For
details, see the “SemiSim” procedure of §A.2.
Since Re (2 + 4i − 3j + 13k) = 2 = Re (2 + 5i + 12j − 5k) and
|2 + 4i − 3j + 13k| =22 + 42 + (−3)2 + 132 = 198
=22 + 52 + 122 + (−5)2 = |2 + 5i + 12j − 5k| ,
2.54 has infinitely many solutions. First we solve the linear system
−16x2 + 18x3 − 2x0 = 0,
16x1 − 18x3 − 30x0 = 0,
18x1 − 18x2 − 36x0 = 0,
2x + 30x − 36x = 0.
1 2 3
50
2.3 Special Quaternion Equations and problems
Direct calculation yields:
x0 = −8t2 + 9t3 ,
x1 = −15t2 + 18t3 ,
x2 = t2 ,
x = t3 ,
3
where t2 and t3 are arbitrary real numbers such that x = x0 + x1 i + x2 j + x3 k 6= 0.
Then we have a solution set with an infinite number of solutions of the form:
x = x,
= (2 + 5i + 12j − 5k) x−1 (2 + 4i − 3j + 13k)−1 .
y
For example, when t2 = 2 and t3 = 3, we have:
x = −11 + 24i + 2j + 3k,
11 12 1 3
= − 710 − − −
y
i j k.
355 355 710
Direct calculation shows that (x, y) satisfies 2.54.
51
2.3 Special Quaternion Equations and problems
2.3.5. Another pair of quaternion equations
Two nonzero quaternions a and b are said to be consemisimilar if the following
system has a solution:
yax = b,
(2.55)
xby
= a,
where x and y are nonzero quaternion indeterminates. 2.55 was studied and solved
in (Tian, 2010). In this subsection, we present a somewhat simpler solution.
Since st = ts and s−1 = s−1 for all s, t ∈ H, 2.55 can be written as
= a−1 x−1 b,
y
= b−1 x−1 a,
y
this gives:
a−1 x−1 b = b−1 x−1 a (2.56)
a−1 b = b−1 |a| ⇒ |a| = |b| .
Therefore a and b must satisfy the following equation in order for 2.55 to have
solutions,
a20 + a21 + a22 + a23 = b20 + b21 + b22 + b23 . (2.57)
2.56 is equivalent to x−1 ba−1 = ab−1 x−1 . On the other hand, we have:
a0 − a1 i − a2 j − a3 k a
a−1 = 2 2 2 2
= 2
a0 + a1 + a2 + a3 a0 + a1 + a22 + a23
2
and
b0 − b 1 i − b2 j − b3 k b
b−1 = 2 2 2 2
= 2 .
b0 + b1 + b2 + b3 b0 + b1 + b22 + b23
2
52
2.3 Special Quaternion Equations and problems
By 2.57, 2.56 is equivalent to bax = xab ⇔ xab = bax, which is equivalent to the
following system:
m11 x1 + m12 x2 + m13 x3 + m14 x0 = 0,
m21 x1 + m22 x2 + m23 x3 + m24 x0 = 0,
(2.58)
m31 x1 + m32 x2 + m33 x3 + m34 x0 = 0,
m x + m42 x2 + m43 x3 + m44 x0 = 0,
41 1
where mij is given by the 4 × 4 skew-symmetric matrix:
M = (mij )4×4
0 a0 b 3 + a3 b 0 − (a0 b2 + a2 b0 ) a2 b 3 − a3 b 2
− (a b + a b ) 0 a0 b1 + a1 b0 − (a1 b3 − a3 b1 )
0 3 3 0
=
.
a b +a b
0 2 2 0 − (a0 b1 + a1 b0 ) 0 a1 b 2 − a2 b 1
− (a2 b3 − a3 b2 ) a1 b 3 − a3 b 1 − (a1 b2 − a2 b1 ) 0
Direct calculation yields det M = 0. Therefore, the homogeneous system 2.58 has a
solution set with an infinite number of solutions (x0 , x1 , x2 , x3 ). Each such x gives
a solution to 2.55:
x = x,
b−1 x−1 a.
y
=
Example 2.3.6. Find all (x, y) such that the following system holds:
y (1 − 25i + 8j + 15k) x = 17 + i − 24j − 7k,
(2.59)
x (17 + i − 24j − 7k) y = 1 − 25i + 8j + 15k.
The following steps have been implemented into a single procedure in Maple. For
details, see the “CSSim” procedure of §A.2.
53
2.3 Special Quaternion Equations and problems
Since
|17 + i − 24j − 7k| =172 + 12 + (−24)2 + (−7)2 = 915
=12 + (−25)2 + 82 + 152 = |1 − 25i + 8j + 15k| ,
2.59 has infinitely many solutions. First we solve the linear system:
248x2 − 112x3 − 304x0 = 0,
−248x1 − 424x3 + 160x0 = 0,
+ 424x2 − 592x0 = 0,
112x1
304x − 160x2 + 592x3 = 0,
1
to obtain:
31 53
x0 = − 20 t1 − t,
20 3
x1 = t1 ,
19 37
x2 = t
10 1
+ t,
10 3
x = t3 ,
3
where t1 and t3 are arbitrary real numbers such that x = x0 + x1 i + x2 j + x3 k 6= 0.
Then we have a solution set with an infinite number of solutions of the form:
x = x,
(17 + i − 24j − 7k)−1 x−1 (1 − 25i + 8j + 15k) .
y
=
For example, when t1 = 1 and t3 = 3, we have:
= − 19
x + i + 13j + 3k,
2
2 56 4 32
− −
y
= 1077
+ 1077
i 359
j 1077
k.
Direct calculation shows that (x, y) satisfies 2.59.
54
2.3 Special Quaternion Equations and problems
2.3.6. A least norm problem
The following least norm problem was proposed in (Tian, 2010),
min |ax − xb − c| ,
x∈H
where a, b and c ∈ H and x is a quaternion indeterminate that minimizes the value
of |ax − xb − c|. In this subsection, we present a solution.
We consider:
ax − xb = c. (2.60)
2.60 is equivalent to the real linear homogeneous system:
m11 x1 + m12 x2 + m13 x3 + m14 x0 = c1 ,
m21 x1 + m22 x2 + m23 x3 + m24 x0 = c2 ,
(2.61)
m31 x1
+ m32 x2 + m33 x3 + m34 x0 = c3 ,
m x + m42 x2 + m43 x3 + m44 x0 = c0 ,
41 1
where mij is given by:
a0 − b 0 −a3 − b3 a2 + b 2 a1 − b 1
a +b
3 3 a0 − b0 −a1 − b1 a2 − b 2
M = (mij )4×4 =
.
−a − b a1 + b 1 a0 − b 0 a3 − b 3
2 2
−a1 + b1 −a2 + b2 −a3 + b3 a0 − b0
Direct calculation yields:
2
det M = a21 + a22 + a23 − b21 − b22 − b23
+ (a0 − b0 )2 (a0 − b0 )2 + 2a21 + 2a22 + 2a23 + 2b21 + 2b22 + 2b23 .
We continue the discussion by cases:
Case 1. a ∼ b. Then det M = 0. Therefore, we can find the least norm solution
55
2.3 Special Quaternion Equations and problems
y1
y2
of 2.61 by calculating
= M † c0 where M † is the Moore-Penrose
y3
y0
c1
c2
pseudoinverse of M and c0 = (Bjõrck, 1996). This means that
c3
c0
x = y0 + y1 i + y2 j + y3 k satisfies |ax − xb − c| ≤ |az − zb − c| for all
z ∈ H.
Case 2. Otherwise. Then det M 6= 0. Therefore,
we can solve the linear system
y1
y2
2.61 to obtain a unique solution
such that 2.61 holds. This means
y3
y0
that x = y0 + y1 i + y2 j + y3 k satisfies |ax − xb − c| = 0.
Example 2.3.7. Solve the following least norm problem:
min |(5 − 10i − 5j + 2k) x − x (3 − 4i − 4j − 8k) − (−9 − 2i + 10j − 2k)| . (2.62)
x∈H
The following steps have been implemented into a single procedure in Maple. For
details, see the “LN1” procedure of §A.2.
Since 5 − 10i − 5j + 2k 3 − 4i − 4j − 8k, there exists an exact solution. We solve
56
2.3 Special Quaternion Equations and problems
the linear system:
2x1 + 6x2 − 9x3 − 6x0 = −2,
−6x1 + 2x2 + 14x3 − x0 = 10,
− 14x2 + 2x3 + 10x0 = −2,
9x1
6x + x2 − 10x3 + 2x0 = −9,
1
to obtain:
x0 = − 3364 ,
2905
128
x1 = ,
415
= − 1073
x2 2905
,
2372
x = .
3 2905
Direct calculation shows that x = − 3364
2905
+ 128
415
i − 1073
2905
j + 2372
2905
k satisfies 2.62.
57
2.3 Special Quaternion Equations and problems
2.3.7. Another least norm problem
The following least norm problem was proposed in (Tian, 2010):
min |ax − x̄b − c| , (2.63)
x∈H
where a, b and c ∈ H and x is a quaternion indeterminate that minimizes the value
of |ax − x̄b − c| . In this subsection, we present a solution.
We consider:
ax − x̄b − c = 0. (2.64)
2.64 is equivalent to the real linear homogeneous system:
m11 x1 + m12 x2 + m13 x3 + m14 x0 = c1 ,
m21 x1 + m22 x2 + m23 x3 + m24 x0 = c2 ,
(2.65)
m31 x1
+ m32 x2 + m33 x3 + m34 x0 = c3 ,
m x + m42 x2 + m43 x3 + m44 x0 = c0 ,
41 1
where mij is given by:
a0 + b 0 −a3 + b3 a2 − b 2 a1 − b 1
a −b a0 + b 0 −a1 + b1 a2 − b 2
3 3
M = (mij )4×4 =
.
−a + b a1 − b 1 a0 + b 0 a3 − b 3
2 2
−a1 − b1 −a2 − b2 −a3 − b3 a0 − b0
Direct calculation yields:
det M = (a0 + b0 )2 + (a1 − b1 )2 + (a2 − b2 )2 + (a3 − b3 )2
× a20 + a21 + a22 + a23 − b20 − b21 − b22 − b23 .
We continue the discussion by cases:
Case 1. a ∼ b. Then det M = 0. Therefore, we can find the least norm solution
58
2.3 Special Quaternion Equations and problems
y1
y2
of 2.65 by calculating
= M † c0 where M † is the Moore-Penrose
y3
y0
c1
c2
pseudoinverse of M and c0 = (Bjõrck, 1996). This means that
c3
c0
x = y0 + y1 i + y2 j + y3 k satisfies |ax − x̄b − c| ≤ |az − zb − c| for all
z ∈ H.
Case 2. Otherwise. Then det M 6= 0. Therefore,
we can solve the linear system
y1
y2
2.65 to obtain a unique solution
such that 2.65 holds. This means
y3
y0
that x = y0 + y1 i + y2 j + y3 k satisfies |ax − x̄b − c| = 0.
Example 2.3.8. Solve the following least norm problem:
min |(6 − 8i + j + 5k) x − x (6 + i + 5j − 8k) − (−3 + i + j − 5k)| . (2.66)
x∈H
The following steps have been implemented into a single procedure in Maple. For
details, see the “LN2” procedure of §A.2.
Since 6 − 8i + j + 5k ∼ 6 + i + 5j − 8k, we can only find a least square
solution. We calculate M † c0 where M † is the Moore-Penrose pseudoinverse of
59
2.3 Special Quaternion Equations and problems
12 −13 −4 −9 1
13 12 9 −4 1
M=
and c0 =
to obtain:
4 −9 12 13 −5
7 −6 3 0 −3
39
x0 = − 205 ,
119
x1 = − 7380 ,
1133
x2 = 8610
,
2519
x = − 17220 .
3
39 119 1133 2519
Direct calculation shows that x = − 205 − 7380
i + 8610
j − 17220
k satisfies 2.66.
60
3. Quaternion Polynomial Matrices
3.1. Generalized Inverse
Definition 3.1.1. H [x] denotes the set of quaternion polynomials
( n )
i
X
ai x | ai ∈ H, an 6= 0, n ∈ N ,
i=0
where x commutes element-wise with H. H [x]m×n denotes the space of m × n
matrices with entries from H [x].
Remark 3.1.2. If no ambiguity arises, the indeterminate x will be omitted and we
write p ∈ H [x] or A ∈ H [x]m×n and so on. Im denotes the m × m identity matrix.
0m×n denotes the m × n zero matrix. When the dimension is unspecified, a matrix
is of appropriate dimension.
Definition 3.1.3. Let A ∈ H [x]m×n and
Pn
i=0 pi xi = p0 +p1 x+· · ·+pn xn = p ∈ H [x]
where pi ∈ H. We give a list of some notations as follows:
Pn
i
• p= i=0 pi x , A denotes the conjugate of A, A = (Aij ).
ij
!
m×n P Q
• If A = P + Qj where P , Q ∈ C [x] , then χA = ∈ C [x]2m×2n
−Q P
denotes the complex adjoint of A.
• AT ∈ H [x]n×m denotes the transpose of A, AT = Aji .
ij
61
3.1 Generalized Inverse
• A∗ ∈ H [x]n×m denotes the conjugate transpose of A, (A∗ )ij = (Aji ). In
particular, for all p ∈ H [x], p∗ = p.
• A† ∈ H [x]n×m , a generalized inverse of A (Penrose, 1955), if it exists, denotes
the solution of the following system of equations,
AXA = A (3.1)
XAX = X (3.2)
(AX)∗ = AX (3.3)
(XA)∗ = XA (3.4)
We first show some elementary properties that will be used often throughout this
chapter.
Lemma 3.1.4. Let A ∈ H [x]m1 ×n and B ∈ H [x]n×m2 . Then (AB)∗ = B ∗ A∗ .
Proof. By direct calculation.
Corollary 3.1.5. For A ∈ H [x]m×n , let B = AA∗ . Then B = B ∗ .
Proof. B ∗ = (AA∗ )∗ = (A∗ )∗ A∗ = AA∗ = B.
Lemma 3.1.6. Let A ∈ H [x]m×n have a generalized inverse A† . Then (A∗ )† =
∗
A† .
∗
Proof. We show that A† satisfies the defining relations of a generalized inverse
stated in Definition 3.1.3,
∗ ∗
A∗ A† A∗ = AA† A = A∗ ,
∗ ∗ ∗ ∗
A† A∗ A† = A† AA† = A† ,
h ∗ i ∗ h ∗ i ∗ ∗ ∗
A∗ A† = A† A = A† A = A† A = A∗ A† ,
h ∗ i∗ h ∗ i ∗ ∗ ∗
A† A∗ = AA† = AA† = AA† = A† A∗ .
62
3.1 Generalized Inverse
∗
Therefore, (A∗ )† = A† .
Lemma 3.1.7. (Adaption of (Penrose, 1955)) Let A ∈ H [x]m×n have a generalized
∗ ∗
inverse A† . Then relations satisfied by A† include: A† A† A∗ = A† = A∗ A† A†
and A† AA∗ = A∗ = A∗ AA† .
Proof. First,
∗ ∗
A† A† A∗ =A† AA† = A† AA† = A†
∗ ∗
=A† AA† = A† A A† = A∗ A† A† .
Next,
∗ ∗
A† AA∗ = A† A A∗ = AA† A = A∗
∗ ∗ ∗
= AA† A = A∗ A† A∗ = A∗ AA† = A∗ AA† .
We next show that A† is uniquely defined for quaternion polynomial matrices.
Lemma 3.1.8. (Adaption of (Penrose, 1955)) Let A ∈ H [x]m×n have a generalized
inverse A† . Then A† is unique.
Proof. We first show that 3.2 and 3.3 are equivalent to
XX ∗ A∗ = X. (3.5)
Substituting 3.3 into 3.2, we obtain X = XAX = X (AX)∗ = XX ∗ A∗ .
Conversely, 3.5 implies that:
AXX ∗ A∗ = AX ⇒ (AXX ∗ A∗ )∗ = (AX)∗ ⇒ AXX ∗ A∗ = (AX)∗ .
63
3.1 Generalized Inverse
So AX = (AX)∗ and therefore X = XX ∗ A∗ = X (AX)∗ = XAX. Similarly 3.1
and 3.4 can be replaced by the equation:
XAA∗ = A∗ . (3.6)
Therefore, 3.5 and 3.6 are equivalent to 3.1, 3.2, 3.3 and 3.4.
Likewise the following two equations are also equivalent to 3.1, 3.2, 3.3 and 3.4,
X ∗ A∗ A = A (3.7)
A∗ X ∗ X = X. (3.8)
Suppose A† satisfies 3.5 and 3.6, B satisfies 3.7 and 3.8. Then
∗ ∗
A† =A† A† A∗ = A† A† (B ∗ A∗ A)∗
∗
=A† A† A∗ AB = A† AB = A† AA∗ B ∗ B = A∗ B ∗ B = B.
Therefore, if it exists, A† is uniquely defined.
Due to the non-commutativity of quaternions, there are two types of eigenvalues:
right eigenvalues and left eigenvalues. Right eigenvalues have been studied exten-
sively (Brenner, 1951; Lee, 1949; Baker, 1999). We will work with right eigenvalues
towards our main result. For convenience, we will just use the term “eigenvalue”
from now on.
Definition 3.1.9. A ∈ Hm×m is unitary if AA∗ = A∗ A = I.
Lemma 3.1.10. (Zhang, 1997) A ∈ Hm×m is hermitian, that is, A
= A∗ , if and
d1
..
m×m ∗
only if there exists a unitary matrix U ∈ H such that U AU = . ,
dm
where di are the eigenvalues of A.
Proof. Corollary 6.2 of (Zhang, 1997).
64
3.1 Generalized Inverse
Adapting ideas from (Puystjens and de Smet, 1980), we will give conditions that
quaternion polynomial matrices must satisfy to have generalized inverses. We need
to prove some lemmas first.
Lemma 3.1.11. (Adaption of (Puystjens and de Smet, 1980)) Let A ∈ H [x]m×n
have the generalized inverse A† . If U ∈ Hm×m is a unitary matrix, then (U A)† =
A† U ∗ .
Proof. We show that A† U ∗ satisfies the defining relations of the generalized inverse
stated in Definition 3.1.3,
U A A† U ∗ U A = U AA† A = U A,
A† U ∗ (U A) A† U ∗ = A† AA† U ∗ = A† U ∗ ,
∗ ∗
U AA† U ∗ = U AA† U ∗ = U AA† U ∗ ,
∗ ∗
A† U ∗ U A = A† A = A† A = A† U ∗ U A.
Therefore, (U A)† = A† U ∗ .
Lemma 3.1.12. A ∈ Hm×n induces a homomorphism H [x]n×1 to H [x]m×1 , that is,
for all P , Q ∈ H [x]n×1 , A (P + Q) = AP + AQ ∈ H [x]m×1 .
Proof. By direct calculation.
Lemma 3.1.13. (Adaption of (Puystjens and de Smet, 1980)) Let A ∈ H [x]m×n
have the generalized inverse A† . Consider A as a homomorphism from H [x]n×1 to
H [x]m×1 . Then ImageA = ImageAA∗ = ImageAA† and ImageA∗ = ImageA∗ A =
ImageA† A.
Proof. Let P ∈ H [x]n×1 . Then
∗ h ∗ i
AP = AA† AP = A A† A P = AA∗ A† P ,
65
3.1 Generalized Inverse
∗
where AP ∈ H [x]m×1 and A† P ∈ H [x]m×1 . Therefore, ImageA ⊆ ImageAA∗
and ImageA ⊆ ImageAA† .
For any Q ∈ H [x]m×1 , it is clear that AA∗ Q = A (A∗ Q) where A∗ Q ∈ H [x]n×1 and
that AA† Q = A A† Q where A† Q ∈ H [x]n×1 . Thus, ImageAA∗ ⊆ ImageA and
ImageAA† ⊆ ImageA.
Therefore, ImageA = ImageAA∗ and ImageA = ImageAA† , i.e., ImageA =
ImageAA∗ = ImageAA† .
Similarly we can show that ImageA∗ = ImageA∗ A = ImageA† A.
Lemma 3.1.14. (Adaption of (Puystjens and de Smet, 1980)) If E ∈ H [x]m×m is
a symmetric projection, that is, E = E 2 = E ∗ , then E ∈ Hm×m .
Proof. If f1 , · · · , fm are the entries in the first row of E, with f1 = f1 , then
m m
fi fi = f1 ⇒ f12 +
X X
f1 f1 + fi fi = f1 .
i=2 i=2
Since f1 = f1 , the leading coefficient of f12 is a positive real number. Note that the
Pm
leading coefficient of i=2 fi fi is also a positive real number. Thus,
m m
! ( !)
f12 f12 f12
X X
deg ≥ deg f1 = deg + fi fi = max deg , deg fi fi
i=2 i=2
≥ deg f12 .
P
m
This shows that f1 ∈ H. Furthermore, 0 = deg f1 = deg i=2 fi fi and therefore
fi ∈ H for all 1 ≤ i ≤ m.
The same can be done for the other rows of E. Therefore, E ∈ Hm×m .
Now we are ready to give conditions that quaternion polynomial matrices must
satisfy in order to have generalized inverses.
66
3.1 Generalized Inverse
Theorem 3.1.15. (Adaption of (Puystjens and de Smet, 1980))
Let A ∈ H [x]m×n .
A1 A2
Then A has the generalized inverse A† if and only if A = U with U ∈
0 0
r×r
Hm×m
unitary and A1 A∗1 + A2 A∗2 a unit in H [x] with r ≤ min {m, n}. Moreover,
∗
A1 (A1 A∗1
+ 0 ∗ A2 A∗2 )−1
A† = U .
∗ −1
∗ ∗
A2 (A1 A1 + A2 A2 ) 0
Proof. If A has the generalized inverse A† , then
2 ∗
AA† = AA† AA† = AA† = AA† .
By Lemma 3.1.14, AA† ∈ Hm×m . AA† is hermitian and hence, by Lemma 3.1.10,
there exists a unitary matrix U ∈ Hm×m such that U ∗ AA† U = D where D is
diagonal. Since
D2 = U ∗ AA† U U ∗ AA† U = U ∗ AA† AA† U = U ∗ AA† U = D,
the diagonal entries
ofD are either 1 or 0. Therefore, we can rearrange the rows of
Ir 0
U so that D =
with r ≤ min {m, n}.
0 0
Set A0 =
U ∗ A. By Lemma
3.1.11,
A0 has its own generalized inverse A0† and
Ir 0 A1 A2
A0 A0† = . Set A0 = , for arbitrary quaternion polynomial matrices
0 0 A3 A4
A1 ∈ H [x]r×r , A2∈ H [x]
r×(n−r)
3 ∈ H [x]
, A
(m−r)×r
and A4 ∈ H [x](m−r)×(n−r)
. Since
Ir 0 A1 A2 A1 A2 A1 A2
A0 = A0 A0† A0 =
= , we must have A0 =
0 0 A3 A4 0 0 0 0
∗ ∗
A1 A1 + A2 A2 0 B1 0
and therefore A0 A0∗ = . Similarly, A0† = for some B1
0 0 B2 0
and B2 .
67
3.1 Generalized Inverse
By Lemma 3.1.13,
Ir 0
ImageA0 A0∗ = ImageA0 = ImageA0 A0† = Image
.
0 0
This implies the surjectivity of A1 A∗1 + A2 A∗2 on H [x]r×1 . Therefore, A1 A∗1 + A2 A∗2
is a unit in H [x]r×r and
A1 A2
A = U A0 = U
.
0 0
Next, we have that:
∗
† †
A0† =A0† A0† A0∗ = A0† (A0∗ ) A0∗ = A0∗ (A0 A0∗ )
∗
A1 0 (A1 A∗1 + A2 A∗2 )−1 0
=
A∗2 0 0 0
∗
A1 (A1 A1
∗
+ A2 A∗2 )−1 0
= ,
A2 A∗2 )−1
A∗2 (A1 A∗1 + 0
which gives:
∗
A1 + (A1 A∗1 0 ∗ A2 A∗2 )−1
A† = U .
∗ −1
∗ ∗
A2 (A1 A1 + A2 A2 ) 0
We will show the converse by direct computation.
A1 A2
Let A = U be in H [x]m×n with U ∈ Hm×m unitary and A1 A∗1 + A2 A∗2 a
0 0
r×r
unit in H [x] with r ≤ min {m, n}. Let
∗
A1+ (A1 A∗1 0 ∗ A2 A∗2 )−1
B= U .
∗ −1
∗ ∗
A2 (A1 A1 + A2 A2 ) 0
We show that A† U ∗ satisfies the defining relations of the generalized inverse stated
in Definition 3.1.3,
68
3.1 Generalized Inverse
• the first condition:
ABA
A1 A2 A∗1 (A1 A∗1 + A2 A∗2 )−1 0 A1 A2
=U U ∗U
A2 A∗2 )−1
0 0 A∗2 (A1 A∗1 + 0 0 0
A1 A2 A∗1 (A1 A∗1 + A2 A∗2 )−1 0 A1 A2
=U
A2 A∗2 )−1
0 0 A∗2 (A1 A∗1 + 0 0 0
A1 A2
=U
= A,
0 0
• the second condition:
BAB
∗
A1 (A1 A∗1
+ A2 A∗2 )−1
0 ∗ A1 A2 A∗1 (A1 A∗1 + A2 A∗2 )−1 0
= U U U∗
∗ −1
A2 A∗2 )−1
∗ ∗
A2 (A1 A1 + A2 A2 ) 0 0 0 A∗2 (A1 A∗1 + 0
∗
A1 (A1 A1
∗
+ A2 A∗2 )−1 0 A1 A2 A∗1 (A1 A∗1 + A2 A∗2 )−1 0
= U∗
A2 A∗2 )−1 A2 A∗2 )−1
A∗2 (A1 A∗1 + 0 0 0 A∗2 (A1 A∗1 + 0
∗
A1 (A1 A1
∗
+ A2 A∗2 )−1 0
= U∗ = B,
A2 A∗2 )−1
A∗2 (A1 A∗1 + 0
• the third condition:
∗
A1 A2 A∗1 (A1 A∗1 + A2 A∗2 )−1 0
(AB)∗ = U ∗
U
A∗2 (A1 A∗1 + A2 A∗2 )−1 0
0 0
∗
Ir 0 Ir 0
=
=
= AB,
0 0 0 0
69
3.1 Generalized Inverse
• the last condition:
∗
∗
A1 (A1 A∗1
+ A2 A∗2 )−1
0 ∗ A1 A2
(BA)∗ = U U
∗ −1
∗ ∗
A2 (A1 A1 + A2 A2 ) 0 0 0
∗
Ir 0 Ir 0
=
=
= BA.
0 0 0 0
∗ ∗ ∗ −1 0
A1 (A1 A1 + A2 A2 )
Therefore, B = U∗ = A† . The uniqueness of A† is the
A∗2 (A1 A∗1 + A2 A∗2 )−1 0
result of Lemma 3.1.8.
Lemma 3.1.16. Let A ∈ Hm×m be hermitian. Then the eigenvalues of A are real.
Let λ ∈ H be an eigenvalue of A with corresponding eigenvector 0 6= X =
Proof.
x1
..
. such that AX = Xλ. Then X ∗ AX = X ∗ Xλ. So X ∗ AX = λ∗ X ∗ X and
xm
therefore
X ∗ Xλ = λ∗ X ∗ X = (X ∗ Xλ)∗
⇒
x1
∗
. X X
(x1 , · · · xm ) .
. λ = xi xi λ = xi xi λ
xm
X ∗ X
= λ∗ xi xi = λ∗ x i xi .
Note that 0 6= xi xi ∈ R [x]. Division of the above equation by
P P
xi xi gives
λ = λ∗ ⇒ λ ∈ R.
Theorem 3.1.17. (Zhang, 1997)(Cayley-Hamilton theorem for Quaternion Matri-
ces) Let A ∈ Hm×m . fA (λ) = det (λI2m − χA ) is called the characteristic polyno-
70
3.1 Generalized Inverse
mial of A, where λ is a complex indeterminate. Then fA (A) = 0. Furthermore,
fA (λ0 ) = 0 if and only if λ0 is an eigenvalue of A.
Proof. Theorem 8.1 of (Zhang, 1997).
Definition 3.1.18. Let A ∈ H [x]m×n have the generalized inverse A† . For any
x ∈ R, set B = AA∗ . Then fB (λ) = det (λI2m − χB ) is called the characteristic
polynomial of B. By Lemma 3.1.16, λ can be assumed to be an indeterminate that
enjoys the following: λ = λ and λ commutes element-wise with H [x].
Lemma 3.1.19. Let A ∈ H [x]m×n have the generalized inverse A† . For any x ∈ R,
set B = AA∗ . Then fB (λ) = g (λ)2 where g (λ) ∈ (R [x]) [λ].
Proof. We first show that fB (λ) ∈ (R [x]) [λ]. Since
det (λI2m − χB )T = det (λI2m − χB ) = det ((λI2m − χB )∗ )
det (λI2m − χB ) = det (λI2m − χB ) = det (λI2m − χB ),
we have that:
det (λI2m − χB ) = fB (λ) ∈ (R [x]) [λ] . (3.9)
Next, we show that fB (λ) = g (λ)2 where g (λ) ∈ (C [x]) [λ]. Let B = P + Qi. For
any fixed 1 ≤ i, j ≤ m, we have Bij = a + bi + cj + dk where a, b, c and d ∈ R [x].
Since B is hermitian, Bji = a − bi − cj − dk and therefore Pij = a + bi, Pji = a − bi
and Qij = c + di, Qji = −c − di. So P T = P and Q = −QT . Therefore,
P Q P Q
χB =
=
T
−Q P −Q P
71
3.1 Generalized Inverse
λIm −P Q
λI2m − χB =
.
−Q λIm − P T
Next, we have:
Im −Im Im 0 Im −Im λIm − P Q
0 Im Im Im 0 Im −Q λIm − P T
T
Q P − λIm
=
λIm − P Q
and
Im −Im Im 0
det
= det
= 1.
0 Im Im Im
Therefore,
T
λIm −P Q Q P − λIm
fB (λ) = det
= det
.
T
−Q λIm − P λIm − P Q
Note that:
T
T T
Q P − λIm Q P − λIm
−
=
,
λIm − P Q λIm − P Q
Q P T − λIm
which implies that is skew-symmetric. By (Muir, 2003), the
λIm − P Q
Q PT − λIm
determinant of , also called its Pfaffian, can be written as
λIm − P Q
the square of a polynomial in its entries. Therefore, fB (λ) = g (λ)2 where g (λ) ∈
(C [x]) [λ].
Finally we show that g (λ) ∈ (R [x]) [λ]. Suppose otherwise. Then g (λ) =
a (λ) + b (λ) i where a and b ∈ (R [x]) [λ] with b (λ) 6= 0. By 3.9, g (λ)2 ∈
(R [x]) [λ]. So a (λ) = 0 and therefore fB (λ) = g (λ)2 = (b (λ) i)2 = −b (λ)2
72
3.1 Generalized Inverse
where b (λ) ∈ (R [x]) [λ]. For a fixed x ∈ R, let λ0 ∈ R be large enough such that
λ0 I2m − χB ∈ C2m×2m is diagonally dominant with non-negative diagonal entries
and that (b (x)) (λ0 ) 6= 0. Since λ0 I2m − χB is also hermitian, λ0 I2m − χB is positive
definite (Horn and Johnson, 1990). But det (λ0 I2m − χB ) = − ((b (x)) (λ0 ))2 < 0.
Contradiction. So, b = 0 and therefore fB (λ) = g (λ)2 where g (λ) ∈ (R [x]) [λ].
Corollary 3.1.20. Let A ∈ H [x]m×n have the generalized inverse A† . For any
x ∈ R, set B = AA∗ and fB (λ) = g (λ)2 . Then g (B) = 0. We will call g (λ) the
generalized characteristic polynomial of A.
Proof. g (λ) ∈ (R [α]) [λ] by Lemma 3.1.19, so χg(B) = g (χB ). Next, fB (χB ) = 0 by
the Cayley-Hamilton theorem for complex polynomial matrices (Horn and Johnson,
1990), so g (χB ) = 0. Therefore 0 = g (χB ) = χg(B) , that is, g (B) = 0.
73
3.2 Leverrier-Faddeev Method
3.2. Leverrier-Faddeev Method
Lemma 3.2.1. Let A ∈ H [x]m×n have the generalized inverse A† . Set B = AA∗ .
Then B † = (A∗ )† A† and B † B = BB † .
Proof. First we show that (A∗ )† A† satisfies the defining relations of the generalized
inverse stated in Definition 3.1.3,
• the first condition:
h i
B (A∗ )† A† B =AA∗ (A∗ )† A† AA∗
=AA∗ (A∗ )† A∗ = AA∗ = B,
• the second condition:
h i h i
(A∗ )† A† B (A∗ )† A† = (A∗ )† A† AA∗ (A∗ )† A†
= (A∗ )† A∗ (A∗ )† A† = (A∗ )† A† ,
• the third condition:
h i∗ ∗ ∗
B (A∗ )† A† = AA∗ (A∗ )† A† = AA†
h i
=AA† = AA∗ (A∗ )† A† = B (A∗ )† A† ,
• the last condition:
h i ∗ ∗ ∗
(A∗ )† A† B = (A∗ )† A† AA∗ = (A∗ )† A∗
h i
= (A∗ )† A∗ = (A∗ )† A† AA∗ = (A∗ )† A† B.
Therefore, (A∗ )† A† = (AA∗ )† = B † . Next,
(AA∗ )† AA∗ = (A∗ )† A† AA∗ = (A∗ )† A∗
∗ ∗
= (A∗ )† A∗ = AA† = AA∗ A† A† = AA∗ (AA∗ )† ,
74
3.2 Leverrier-Faddeev Method
which is, B † B = BB † .
Lemma 3.2.2. Let A ∈ H [x]m×n have the generalized inverse A† . Set B = AA∗ .
k †
Then B † = Bk where k ∈ N.
k
Proof. We show that B † satisfies the defining relations of the generalized inverse
stated in Definition 3.1.3,
• the first condition:
k k−1
Bk B† B k =B k B † BB † B k−1
k−1
=B k B † B k−1 = · · · = B k ,
• the second condition:
k k k−1 k
B† Bk B† = B† BB † B k−1 B †
k−1 k k
= B† B k−1 B † = · · · = B† ,
• the third condition:
∗
† k
h ik k k
B k
B = (B ∗ )† (B ∗ )k = B † Bk = Bk B† ,
• the last condition:
∗
† k
k k k
B B k
= (B ∗ )k (B ∗ )† = Bk B† = B† Bk.
k †
Therefore, B † = Bk where k ∈ N.
Lemma 3.2.3. (Adaption of (Penrose, 1955)) Let A ∈ H [x]m×n , B ∈ H [x]p×q
and C ∈ H [x]m×q . If A† and B † both exist, then the quaternion polynomial matrix
equation AXB = C has a solution in H [x]n×p if and only if AA† CB † B = C, in
which case the general solution is
X = A† CB † + Y − A† AY BB † ,
75
3.2 Leverrier-Faddeev Method
where Y ∈ H [x]n×p is arbitrary.
Proof. Suppose X satisfies AXB = C, then
C = AXB = AA† AXBB † B = AA† CB † B.
Conversely, if C = AA† CB † B then A† CB † is a particular solution of AXB = C.
For the general solution, we must solve AXB = 0. Any expression of the form
X = Y − A† AY BB † satisfies AXB = 0, and conversely if AXB = 0 then X =
A† CB † + Y − A† AY BB † .
Theorem 3.2.4. (Adaption of (Penrose, 1955)) Let A ∈ H [x]m×n have the general-
ized inverse A† . For any x ∈ R, set B = AA∗ . Suppose the generalized characteristic
polynomial of A is
g (λ) = λm + a1 λm−1 + · · · + ak λm−k + · · · + am−1 λ + am ,
where ai ∈ R [x]. If k is the largest integer such that ak 6= 0, then the generalized
inverse of A is given by
1 ∗ h k−1 i
A† = − A B + a1 B k−2 + · · · + ak−1 I .
ak
If ai = 0 for all 1 ≤ i ≤ m, then A† = 0.
Proof. By Corollary 3.1.20, we have:
0 = B m + a1 B m−1 + · · · + ak B m−k + · · · + am−1 B + am Im .
If k is the largest integer such that ak 6= 0 and define B 0 = I, we may write:
B m−k B k + a1 B k−1 + · · · + ak−1 B + ak I = 0.
This equation guarantees a solution of the matrix equation B m−k X = 0 and hence,
76
3.2 Leverrier-Faddeev Method
by Lemma 3.2.1 and 3.2.3, all solutions are given by:
†
X = Y − B m−k B m−k Y = Y − B † BY,
where Y ∈ Hm×m is arbitrary. In particular, there exists Y1 such that:
B k + a1 B k−1 + · · · + ak−1 B + ak I = Y1 − B † BY1 = Y1 − AA† Y1
Left multiplication of the latter equation by A† gives:
A† B k + a1 A† B k−1 + · · · + ak−1 A† B + ak A† = 0.
By Lemma 3.1.7, this can be simplified to:
A∗ B k−1 + a1 A∗ B k−2 + · · · + ak−1 A∗ = −ak A† ,
which gives:
1 ∗ h k−1 i
A† = − A B + a1 B k−2 + · · · + ak−1 I .
ak
If ai = 0 for all 1 ≤ i ≤ m, then A = 0 and therefore A† = 0.
Lemma 3.2.5. (Adaption of (Kalman, 1978)) Let A ∈ H [x]m×n have the generalized
inverse A† . For any real x ∈ R, set B = AA∗ . Let λ1 , · · · , λm0 , where m0 ≤ m, be
all the non-zero eigenvalues of B. Then for 1 ≤ k ≤ m,
h i
tr B k + a1 B k−1 + · · · + ak−1 B = −kak ,
where the ai arise from the generalized characteristic polynomial of A:
g (λ) = λm + a1 λm−1 + · · · + ak λm−k + · · · + am−1 λ + am .
77
3.2 Leverrier-Faddeev Method
Proof. Let Y = yI where y ∈ R. We can write g (Y ) as:
g (Y ) = (Y − B)
× Y m−1 + (B + a1 I) Y m−2 + B 2 + a1 B + a2 I Y m−3 + · · ·
+ B m−1 + a1 B m−2 + · · · + am I .
As long as y is not an eigenvalue of B, (yI − B) = Y − B is non-singular, so we can
write:
(Y − B)−1 g (Y ) =Y m−1 + (B + a1 I) Y m−2 + B 2 + a1 B + a2 I Y m−3 + · · ·
+ B m−1 + a1 B m−2 + · · · + am I .
Taking the traces gives:
h i
tr (Y − B)−1 g (Y ) =my m−1 + tr [(B + a1 I)] y m−2
h i
+ tr B 2 + a1 B + a2 I y m−3 + · · ·
+ tr B m−1 + a1 B m−2 + · · · + am I .
Let C = (Y − B)−1 g (Y ). Since g (Y ) = g (yI) = g (y) I, C = g (y) (Y − B)−1 .
Therefore,
h i
trC = g (y) tr (Y − B)−1 .
h i
tr (Y − B)−1 is the sum of the eigenvalue of (Y − B)−1 . We will show these eigen-
1 1
values are the fractions y−λ1
, ···, y−λm0
.
Let ζ be an eigenvalue of (Y − B)−1 with corresponding eigenvector Z such that:
(Y − B)−1 Z = Zζ.
78
3.2 Leverrier-Faddeev Method
ζ is real by Lemma 3.1.16, and hence
1
(Y − B) Z = Z
ζ
⇒
!
1
BZ = Z y − .
ζ
Therefore, y − 1
ζ
= λi ⇒ ζ = 1
y−λi
for some 1 ≤ i ≤ m0 .
Since g (y) = (y − λ1 ) (y − λ2 ) · · · (y − λm0 ), we have that g 0 (y) =
1 1
g (y) y−λ1
+ ··· + y−λm0
and
trC = g 0 (y)
The derivative of g is also equal to:
g 0 (y) = my m−1 + a1 (m − 1) y m−2 + · · · + am−1 .
Therefore,
my m−1 + a1 (m − 1) y m−2 + · · · + am−1 =my m−1 + tr [(B + a1 I)] y m−2 + · · ·
+ tr B m−1 + a1 B m−2 + · · · + am I
Comparing the coefficient of y m−k−1 on both sides, we obtain
h i
ak (m − k) = tr B k + a1 B k−1 + · · · + ak−1 B + ak I
⇒
h i
ak (m − k) = tr B k + a1 B k−1 + · · · + ak−1 B + trak I
⇒
h i
−kak = tr B k + a1 B k−1 + · · · + ak−1 B
Lemma 3.2.6. (Adaption of (Decell, 1965; Faddeeva, 1959)) Let A ∈ H [x]m×n have
79
3.2 Leverrier-Faddeev Method
the generalized inverse A† . For any x ∈ R, set B = AA∗ . Suppose the generalized
characteristic polynomial of A is
g (λ) = λm + a1 λm−1 + · · · + ak λm−k + · · · + am−1 λ + am ,
where ai ∈ R [x]. Define a0 = 1. If p is the largest integer such that ap 6= 0 and we
construct the sequence A0 , · · · , Ap as follows:
A0 = 0 −1 = q0 B0 = I
.. .. ..
. . .
trAp−1
Ap−1 = AA∗ Bp−2 = qp−1 Bp−1 = Ap−1 − qp−1 I
p−1
trAp
Ap = AA∗ Bp−1 = qp Bp = Ap − qp I
p
then qi (x) = −ai (x) , i = 0, · · · , p.
Proof. We will show the desired result by mathematical induction.
For p ≥ i ≥ 0, let Pi be the statement:
qi (x) = −ai (x) .
P0 is the statement:
q0 = −a0 ,
which is true by definition.
80
3.2 Leverrier-Faddeev Method
Suppose Pi holds for all 0 ≤ i ≤ k − 1, it remains to show that Pk holds. We have
Ak =BBk−1
=B (Ak−1 − qk−1 I)
=B (B (Ak−2 − qk−2 I) − qk−1 I)
=···
=B k − q1 B k−1 − q2 B k−2 − · · · − qk−1 B
=B k + a1 B k−1 + · · · + ak−1 B.
Therefore,
h i
tr Ak = tr B k + a1 B k−1 + · · · + ak−1 B ,
tr Ak
which by Lemma 3.2.5 is equal to −kak . So qk = k
= −ak .
Thus, Pk holds.
Therefore, by the principle of mathematical induction, Pi holds for all p ≥ i ≥ 0.
That is, qi = −ai , i = 0, · · · , p.
81
3.3 Generalized Inversion by Interpolation
3.3. Generalized Inversion by Interpolation
In this section we present a method to obtain the generalized inverse of a quaternion
polynomial matrix by interpolation at real number data points.
Lemma 3.3.1. (See (Lam, 2001) for more details) An element r ∈ H is a root of a
nonzero polynomial f ∈ H [x] iff x − r is a right divisor of f . The set of polynomials
in H [x] having r as a root is the left ideal H [x] · (x − r).
Proof. By direct calculation.
Lemma 3.3.2. (See (Lam, 2001) for more details) Let f , g and h ∈ H [x], f = gh
and r ∈ H be such that β = h (r) 6= 0. Then
f (r) = g βrβ −1 h (r) .
In particular, if r is a root of f but not of h, then βrβ −1 is a root of g.
Pn Pn Pn
Proof. Let g = i=0 ai xi . Then f = ( i=0 ai x i ) h = i=0 ai hxi . Therefore,
n n
ai h (r) ri = ai βri
X X
f (r) =
i=0 i=0
n n i
ai βri β −1 β = ai βrβ −1 β
X X
=
i=0 i=0
n
!
i
ai βrβ −1 β = g βrβ −1 h (r) .
X
=
i=0
The last conclusion follows since H has no zero-divisors.
Theorem 3.3.3. (Adaption of (Gordon and Motzkin, 1965)) Let f ∈ H [x] be of
degree n. Then the roots of f lie in at most n conjugacy classes of H.
Proof. We will show the desired result by mathematical induction.
For n ≥ 1, let Pn be the statement:
82
3.3 Generalized Inversion by Interpolation
Let f ∈ H [x] be of degree n. Then the roots of f lie in at most n conjugacy classes.
P1 is the statement:
Let f ∈ H [x] be of degree 1. Then the root of f lies in at most 1 conjugacy class.
P1 is trivially true.
Suppose Pn holds for all 1 ≤ n ≤ k − 1, it remains to show that Pk holds. Let
f ∈ H [x] be of degree k with roots c 6= d. By Lemma 3.3.1, we can write
f = g · (x − c) ,
where g has a root that is, by Lemma 3.3.2, conjugate to d. Invoking the inductive
hypothesis, d lies in at most k − 1 conjugacy classes. So, the roots of f lie in at most
k conjugacy classes. Thus, Pk holds. Therefore, by the principle of mathematical
induction, Pn holds for all n ≥ 1.
Theorem 3.3.4. (Adaption of (Bray and Whaples, 1983)) Let c1 , · · · , cn be n
pairwise non-conjugate elements of H. Then there is a unique polynomial gn ∈ H [x],
monic of degree n, such that gn (c1 ) = · · · = gn (cn ) = 0. Moreover, c1 , · · · , cn are
the only roots of gn in H.
Proof. We first show the existence of gn for all n ≥ 1 by mathematical induction.
For n ≥ 1, let Pn be the statement:
Let c1 , · · · , cn ∈ H be pairwise non-conjugate. Then there is a polynomial gn ∈
H [x], monic of degree n, such that gn (c1 ) = · · · = gn (cn ) = 0. Moreover, c1 , · · · , cn
are the only roots of g in H.
P1 is the statement:
Let c1 ∈ H. Then there is a polynomial g1 ∈ H [x], monic of degree 1, such that
g1 (c1 ) = 0. Moreover, c1 is the only root of g in H.
83
3.3 Generalized Inversion by Interpolation
P1 is trivially true as g = x − c1 .
Suppose Pn holds for all 1 ≤ n ≤ k − 1, it remains to show that Pk holds. Let
c1 , · · · , ck ∈ H be pairwise non-conjugate. Invoking the inductive hypothesis, there
exists a monic polynomial gk−1 of degree k − 1 with c2 , · · · , ck as its only roots, that
is, gk−1 (c1 ) 6= 0. Construct gk as follows,
gk = x − gk−1 (c1 ) c1 gk−1 (c1 )−1 · gk−1 .
By Lemma 3.3.2, gk (c1 ) = 0. Thus, Pk holds.
Therefore, by the principle of mathematical induction, for all n ≥ 1, Pn holds.
We next show that gn is unique for all n ≥ 1. For some fixed n, let g 0 6= gn be monic
of degree n, such that g 0 (c1 ) = · · · = g 0 (cn ) = 0 too. Then deg (gn − g 0 ) ≤ n − 1 but
gn − g 0 has roots c1 , · · · , cn which lie in n conjugacy classes of H. This contradicts
Theorem 3.3.3. Therefore, gn is unique for all n ≥ 1.
Lemma 3.3.5. Let c1 , · · · , cn+1 ∈ H be pairwise non-conjugate and let
d1 , · · · , dn+1 ∈ H. Then there is a unique lowest degree polynomial f ∈ H [x],
of degree p ≤ n, such that f (ci ) = di for all 1 ≤ i ≤ n + 1.
Proof. For any 1 ≤ s0 ≤ n + 1, let S = {1, · · · , n + 1} \ {s0 }. By Theorem 3.3.4, for
any s ∈ S, we can find unique hS ∈ H [x], monic of degree n, such that hS (cs ) = 0
and that {cs | s ∈ S} are the only roots of hS in H. So, hS (cs0 ) 6= 0and therefore
0 α ∈ S,
we can construct a polynomial gS of degree n such that gS (cα ) = as
1 0
α=s,
below:
gS = hS (cs0 )−1 hS .
We can then construct the polynomial f , of degree at most n, such that f (ci ) = di
84
3.3 Generalized Inversion by Interpolation
for all 1 ≤ i ≤ n + 1 as below:
n+1
X
f= ds 0 gS .
s0 =1
We next show that f is unique. Suppose we have f 0 ∈ H [x] of degree p0 ≤ p ≤ n,
such that f 0 6= f and that f 0 (ci ) = di for all 1 ≤ i ≤ n + 1 too. Then f − f 0 6= 0 is of
degree at most p ≤ n. But f − f 0 has roots c1 , · · · , cn+1 which lie in n + 1 conjugacy
classes of H. This contradicts Theorem 3.3.3. Therefore, f is unique.
Definition 3.3.6. The degree of a given quaternion polynomial matrix A ∈
H [x]m×n is deg A = max {deg (Aij ) | 1 ≤ i ≤ m, 1 ≤ j ≤ n}.
Lemma 3.3.7. Let A ∈ H [x]m×n have the generalized inverse A† . Then deg A† ≤
(2m − 1) deg A.
Proof. By Theorem 3.2.4, deg A† ≤ deg A∗ (AA∗ )m−2 ≤ deg (A2m−1 ) ≤
(2m − 1) deg A.
Remark 3.3.8. The upper bound mentioned in Lemma 3.3.7 is rarely achieved in
practice, because some elements of AA∗ (and other matrices) do not have maximal
degree.
Lemma 3.3.9. Let c1 , · · · , ck+1 ∈ H be pairwise non-conjugate and let
A1 , · · · , Ak+1 ∈ Hn×m . Then there is a unique lowest degree matrix A ∈ H [x]n×m
of degree p ≤ k, such that A (ci ) = Ai for all 1 ≤ i ≤ k + 1.
Proof. For any 1 ≤ n1 ≤ n and 1 ≤ m1 ≤ m, by Lemma 3.3.5, there is
a lowest degree polynomial An1 m1 determined by the values c1 , · · · , ck+1 and
(A1 )n1 m1 , · · · , (Ak+1 )n1 m1 . In fact, for any 1 ≤ s0 ≤ k + 1, let S = {1, · · · , k + 1} \
{s0 }. Then
k+1
X
An1 m1 = (As0 )n1 m1 gS ,
s0 =1
85
3.3 Generalized Inversion by Interpolation
0 α ∈ S,
where gS (cα ) = . Since n1 and m1 are chosen randomly, a lowest degree
1 0
α=s
matrix A that satisfies A (ci ) = Ai for all 1 ≤ i ≤ k + 1 is determined by A =
Pk+1
s0 =1 As0 gS .
Next we show that A is unique. Suppose A0 6= A of degree p0 ≤ p also satisfies
A0 (ci ) = Ai for all 1 ≤ i ≤ k + 1. Then for some 1 ≤ n2 ≤ n and 1 ≤ m2 ≤ m,
(A − A0 )n2 m2 6= 0. But (A − A0 )n2 m2 , of degree at most p ≤ k, has roots c1 , · · · , ck+1
which lie in k+1 conjugacy classes of H. This contradicts Theorem 3.3.3. Therefore,
A is unique.
Let A ∈ H [x]m×n satisfy conditions of Theorem 3.1.15. For any x ∈ R, set B =
AA∗ . Let p be the largest integer such that ap 6= 0. We construct the sequence
A0 , · · · , Ap as follows:
A0 = 0 −1 = q0 B0 = I
.. .. ..
. . .
trAp−1
Ap−1 = AA∗ Bp−2 = qp−1 Bp−1 = Ap−1 − qp−1 I
p−1
trAp
Ap = AA∗ Bp−1 = qp Bp = Ap − qp I
p
Theorem 3.3.10. In the above setting, let k = (2m − 1) deg A and c1 , · · · , ck+1 ∈
R be k + 1 distinct real numbers such that qp (cs0 ) 6= 0 for any 1 ≤ s0 ≤ k + 1. Let
S = {1, · · · , k + 1} \ {s0 }. Then
k+1
A† = A (cs0 )† gS
X
s0 =1
where
1 h i
A (cs0 )† = A (cs0 )∗ B (cs0 )p−1 − q1 (cs0 ) B (cs0 )p−2 − · · · − qp−1 (cs0 ) I
qp (cs0 )
86
3.3 Generalized Inversion by Interpolation
and
0 α ∈ S,
gS (cα ) =
α = s0 .
1
Proof. By Theorem 3.2.4, Lemma 3.2.6 and Lemma 3.3.9.
Example 3.3.11. Find, by interpolation, the generalized inverse of the quaternion
polynomial matrix
14x + 14 + 76i + 70j + 56k 56 − 28i − 70j + 70k 28j − 56k 14x − 56 − 8i − 14j − 56k
−2x − 2 − 43i − 10j − 8k −8 + 4i + 10j − 10k −4j + 8k −2x + 8 − 31i + 2j + 8k
A=
−3x − 3 + 3i − 15j − 12k −12 + 6i + 15j − 15k −6j + 12k −3x + 12 + 21i + 3j + 12k
−4x − 4 + 4i − 20j − 16k −16 + 8i + 20j − 20k −8j + 16k −4x + 16 + 28i + 4j + 16k
The following steps have been implemented into a single procedure in Maple. For
details, see the “LFI” procedure of §A.2.
As mentioned in Remark 3.3.8, here we only need two data points because direct
calculation shows that deg A† = 2. Let c1 = 0 and c2 = 1. We have:
14 + 76i + 70j + 56k 56 − 28i − 70j + 70k 28j − 56k −56 − 8i − 14j − 56k
−2 − 43i − 10j − 8k −8 + 4i + 10j − 10k −4j + 8k 8 − 31i + 2j + 8k
A (c1 ) =
−3 + 3i − 15j − 12k −12 + 6i + 15j − 15k −6j + 12k 12 + 21i + 3j + 12k
−4 + 4i − 20j − 16k −16 + 8i + 20j − 20k −8j + 16k 16 + 28i + 4j + 16k
and
28 + 76i + 70j + 56k 56 − 28i − 70j + 70k 28j − 56k −42 − 8i − 14j − 56k
−4 − 43i − 10j − 8k −8 + 4i + 10j − 10k −4j + 8k 6 − 31i + 2j + 8k
A (c2 ) = .
−6 + 3i − 15j − 12k −12 + 6i + 15j − 15k −6j + 12k 9 + 21i + 3j + 12k
−8 + 4i − 20j − 16k −16 + 8i + 20j − 20k −8j + 16k 12 + 28i + 4j + 16k
By the algorithm stated in Lemma 3.2.6, we calculate and obtain:
1
A (c1 )† = A (0)† = ×
230175
140 − 560i − 228j − 342k 355 + 1730i − 96j + 81k −255 − 870i + 126j + 54k −340 − 1160i + 168j + 72k
276 + 88i + 426j − 382k 282 + 416i − 93j − 149k −252 − 276i − 72j + 204k −336 − 368i − 96j + 272k
32 + 16i − 176j + 292k −176 − 88i + 68j + 194k 96 + 48i + 12j − 204k 128 + 64i + 16j − 272k
−140 − 122i + 228j + 342k −355 + 2021i + 96j − 81k 255 − 1176i − 126j − 54k 340 − 1568i − 168j − 72k
87
3.3 Generalized Inversion by Interpolation
and
1
A (c2 )† = A (1)† = ×
230175
152 − 550i − 244j − 330k 289 + 1675i − 8j + 15k −219 − 840i + 78j + 90k −292 − 1120i + 104j + 120k
268 + 104i + 406j − 402k 326 + 328i + 17j − 39k −276 − 228i − 132j + 144k −368 − 304i − 176j + 192k
.
32 + 16i − 160j + 300k −176 − 88i − 20j + 150k 96 + 48i + 60j − 180k 128 + 64i + 80j − 240k
−152 − 132i + 244j + 330k −289 + 2076i + 8j − 15k 219 − 1206i − 78j − 90k 292 − 1608i − 104j − 120k
By Theorem 3.3.10, we have:
2
X
A† = A (cs0 )† gS = A (0)† (1 − x) + A (1)† x
s0 =1
1
= ×
230175
(12 + 10i − 16j + 12k) x + 140 − 560i − 228j − 342k (−66 − 55i + 88j − 66k) x + 355 + 1730i − 96j + 81k
(−8 + 16i − 20j − 20k) x + 276 + 88i + 426j − 382k (44 − 88i + 110j + 110k) x + 282 + 416i − 93j − 149k
(16j + 8k) x + 32 + 16i − 176j + 292k (−88j − 44k) x − 176 − 88i + 68j + 194k
(−12 − 10i + 16j − 12k) x − 140 − 122i + 228j − 342k (66 + 55i − 88j + 66k) x − 355 + 2021i + 96j − 81k
(36 + 30i − 48j + 36k) x − 255 − 870i + 126j + 54k (48 + 40i − 64j + 48k) x − 340 − 1160i + 168j + 72k
(−24 + 48i − 60j − 60k) x − 252 − 276i − 72j + 204k (−32 + 64i − 80j + 80k) x − 366 − 368i − 96j + 272k
.
(48j + 24k) x + 96 + 48i + 12j − 204k (64j + 32k) x + 128 + 64i + 16j − 272k
(−36 − 30i + 48j − 36k) x + 255 − 1176i − 126j − 54k (−48 − 40i + 64j − 48k) x + 340 − 1568i − 168j − 72k
Direct calculation shows that A† satisfies the defining relations of the generalized
inverse stated in Definition 3.1.3.
88
A. Maple Codes
A.1. Overview
At the early stage of my graduate studies, I realized the need of utilizing Maple
when trying to solve quaternion quadratic equations. There are existing Maple
packages dedicated to quaternions (Carter, 2007; Harder, 2010), but not quaternion
polynomials or quaternion polynomial matrices. Therefore, I started writing my
own Maple module. During my two years of graduate studies, I gradually extended
this module to have over thirty procedures. These procedures enable the user to do
basic quaternion and quaternion polynomial matrix operations such as additions and
multiplications; to do Lagrange and Newton interpolations of a list of quaternion
pairs; to obtain the generalized inverse of a quaternion polynomial matrix; to solve
multiple quaternion equations and quaternion least norm problems; and so on. In
§A2 the codes of the procedures are provided with descriptions and examples.
89
A.2 The Codes
A.2. The Codes
Q := module()
option package;
export M, ‘*‘, ‘^‘, eval, Qdef, Qrand, QPrand, Mrand,
PMrand, Qreal, Qimag, Qconj, Mconj, Qnorm, Qinv, AXXB,
sortcollect, ZeroP, QMFNorm, PRotate, LagrangeP, AXBc,
NewtonP, QAdjointMatrix, LFI, NoSim1, NoSim2, Meval, AXB,
SemiSim, CSSim, LN1, LN2;
local QMultiplyScalar, QMatrixMultiply, QScalarMultiply;
M := proc(a,b)
#M is the multiplication operator for quaternions and quaternion polynomials.
local a1 := Qreal(a), b1 := coeff(a, I),
c1 := coeff(a, J), d1 := coeff(a, K),
a2 := Qreal(b), b2 := coeff(b, I),
c2 := coeff(b, J), d2 := coeff(b, K);
return
sortcollect (a1*a2 - b1*b2 - c1*c2 - d1*d2
+ (a1*b2 + b1*a2 + c1*d2 - d1*c2)*I
+ (a1*c2 - b1*d2 + c1*a2 + d1*b2)*J
+ (a1*d2 + b1*c2 - c1*b2 + d1*a2)*K, x);
end proc;
90
A.2 The Codes
QMatrixMultiply := proc(A::Matrix, B::Matrix)
#Local procedure for quaternion matrix multiplication.
local nrowsA := LinearAlgebra:-RowDimension(A),
ncolsA := LinearAlgebra:-ColumnDimension(A),
ncolsB := LinearAlgebra:-ColumnDimension(B),
AB := Matrix(1..nrowsA, 1..ncolsB), i, j;
for i from 1 to nrowsA do
for j from 1 to ncolsB do
AB(i, j) := simplify(add(M(A(i, k), B(k, j)), k = 1 .. ncolsA));
end do;
end do;
end proc;
QMultiplyScalar := proc(A::Matrix, B)
#Local procedure to multiply a quaternion matrix by a quaternion scalar.
local nrowsA := LinearAlgebra:-RowDimension(A),
ncolsA := LinearAlgebra:-ColumnDimension(A),
AB := Matrix(1..nrowsA, 1..ncolsA), i, j;
for i from 1 to nrowsA do
for j from 1 to ncolsA do
AB(i, j) := simplify(M(A(i, j), B));
end do;
end do;
91
A.2 The Codes
end proc;
QScalarMultiply := proc(A, B::Matrix)
#Local procedure to multiply a quaternion scalar by a quaternion matrix.
local nrowsB := LinearAlgebra:-RowDimension(B),
ncolsB := LinearAlgebra:-ColumnDimension(B),
AB := Matrix(1..nrowsB, 1..ncolsB), i, j;
for i from 1 to nrowsB do
for j from 1 to ncolsB do
AB(i, j) := simplify(M(A, B(i, j)));
end do;
end do;
end proc;
‘*‘ := proc(m, n)
#Newly defined multiplication operator for quaternions related elements.
options remember;
if type(m, Matrix) and type(n, Matrix) then
QMatrixMultiply(m, n);
elif type(m, Matrix) then
QMultiplyScalar(m, n);
elif type(n, Matrix) then
QScalarMultiply(m, n);
else M(m, n);
92
A.2 The Codes
end if;
end proc;
‘^‘ := proc(m, n::integer)
#Newly defined exponential operator to include quaternion raised to positive integer
powers.
options remember;
local i, p := 1;
if type(m, freeof({I, J, K})) = false and n > 1 then
for i from 1 to n do
p := M(p, m);
end do;
return p;
else m^n;
end if;
end proc;
Qdef := proc(a, b, c, d)
#Defines a quaternion a + bi + cj + dk.
a + b*I + c*J + d*K;
end proc:
Qrand := proc(m::integer, n::integer)
#Generates a random quaternion with all integer coordinates between m and n.
Qdef(rand(m..n)(), rand(m..n)(), rand(m..n)(), rand(m..n)());
93
A.2 The Codes
end proc:
QPrand := proc (m,n,k)
#Generates a random quaternion polynomial of degee k, with coefficients
Qrand (m, n).
local j := k;
return convert([seq(Qrand(m, n)*x^(j - i), i = 0 .. k)], ‘+‘)
end proc;
Mrand := proc(m, n, p, q)
#Generates a random p × q Matrix with quaternion entries Qrand (m, n).
local M := Matrix(1 .. p, 1 .. q), i, j;
for i from 1 to p do
for j from 1 to q do
M(i, j) := Qrand(m, n);
end do;
end do;
end proc:
PMrand := proc(m, n, k1, k2, p, q)
#Generates a random p × q Matrix with random quaternion polynomial entries.
local M := Matrix(1 .. p, 1 .. q), i, j;
for i from 1 to p do
for j from 1 to q do
M(i, j) := QPrand(m, n, rand(k1..k2)());
94
A.2 The Codes
end do;
end do;
end proc:
Qreal := proc(a)
#Calculates the real part of a quaternion.
subs(I = 0, J = 0, K = 0, a);
end proc:
Qimag := proc(a)
#Calculates the imaginary part of a quaternion.
a - Qreal(a);
end proc:
Qconj := proc(f)
#Calculates the conjugate of a quaternion or a quaternion polynomial.
sortcollect(Qreal(f) - Qimag(f), x);
end proc:
Mconj := proc(A)
#Calculates the conjugate of a quaternion or quaternion polynomial matrix.
local nrowsA:=LinearAlgebra:-RowDimension(A),
ncolsA:=LinearAlgebra:-ColumnDimension(A),
AC := Matrix(1..nrowsA, 1..ncolsA), i, j;
for i from 1 to nrowsA do
for j from 1 to ncolsA do
95
A.2 The Codes
AC(i, j) := Qconj(A(i, j))
end do;
end do;
end proc:
sortcollect := proc (A, a)
#Returns A as a polynomial in standard form of variable a.
return sort(collect(A, a), a)
end proc;
eval := proc(N, n)
#Calculates N (n) where N (x) is a polynomial of x.
options remember;
local i, f := 1, m := N;
if has(N, x) = false then
expand(N);
elif is(expand(N),‘+‘) = true then
map(eval,expand(N),n);
else for i from 1 to degree(N, x) do
f := M(f, n);
end do;
M(coeff(N, x, degree(N, x)), f);
end if;
end proc;
96
A.2 The Codes
Meval := proc(N, n)
#Calculates N (n) where N (x) is a quaternion polynomial matrix of x.
options remember;
local i, j,
m := Matrix(LinearAlgebra:-RowDimension(N),
LinearAlgebra:-ColumnDimension(N));
for i from 1 to LinearAlgebra:-RowDimension(m) do
for j from 1 to LinearAlgebra:-ColumnDimension(m) do
m(i, j) := eval(N(i, j), n);
end do;
end do;
end proc;
PRotate := proc(m::integer)
#Generates a block matrix.
local M := Matrix(1 .. 2*m, 1 .. 2*m), i;
for i from 1 to m do
M(i, i + m) := -1;
end do;
for i from m + 1 to 2*m do
M(i, i - m) := 1;
end do;
end proc;
97
A.2 The Codes
Qnorm := proc(a)
#Calculates the 2 norm of a quaternion.
(Qreal(a)^2 + coeff(a, I)^2 + coeff(a, J)^2 + coeff(a, K)^2)^(1/2);
end proc:
Qinv := proc(a)
#Calculates the inverse of a quaternion or a quaternion polynomial.
sortcollect(Qreal(a) - Qimag(a), x)/sortcollect
((Qreal(a)^2 + coeff(a,I)^2 + coeff(a, J)^2 + coeff(a, K)^2), x)
end proc:
NoSim1 := proc(A::list)
#Ensures that a list of quaternion is ready for zero polynomial interpolation.
options remember;
local i, j, k, B := A, n := nops(B);
for i from 2 to n do
for j from 1 to i - 1 do
if Qreal(B[j]) = Qreal(B[i]) and Qnorm(B[j]) = Qnorm(B[i]) then
for k from j + 1 to i - 1 do
if Qreal(B[k]) = Qreal(B[i]) and Qnorm(B[k]) = Qnorm(B[i]) then
B := subsop(i = z, B); next;
end if;
end do;
end if;
98
A.2 The Codes
end do;
end do;
return remove(t -> t = z, B);
end proc;
ZeroP := proc(B::list)
#Calculates an zero polynomial for a list of quaternions.
options remember;
local A := [op({op(NoSim1(B))})], i, n := nops(A), f;
f[1] := x - op(1, A);
for i from 2 to n do
f[i] := M((x - M(M(eval((f[i - 1]), op(i, A)), op(i, A)),
Qinv(eval((f[i - 1]), op(i, A))))), f[i - 1]);
end do;
return sortcollect(f[n], x);
end proc;
NoSim2 := proc(A::listlist)
#Ensures that a list of quaternion pairs is ready for Lagrange Interpolation.
options remember;
local i, j, k, B := A, C, n1 := nops(B), n2;
for i from 2 to n1 do
for j from 1 to i - 1 do
if A[j][1] = A[i][1] then
99
A.2 The Codes
B := subsop([i,1] = z, B);
end if;
end do;
end do;
C := remove(t -> t[1] = z, B); n2 := nops(C);
for i from 3 to n2 do
for j from 1 to i-2 do
if Qreal(C[j][1]) = Qreal(C[i][1]) and Qnorm(C[j][1]) = Qnorm(C[i][1]) then
for k from j + 1 to i - 1 do
if Qreal(C[k][1]) = Qreal(C[i][1]) and Qnorm(C[k][1]) = Qnorm(C[i][1]) then
C := subsop([i, 1] = z, C); next;
end if;
end do;
end if;
end do;
end do;
return remove(t -> t[1] = z, C);
end proc;
LagrangeP := proc(B::listlist)
#Calculates the Lagrange polynomial for a list of quaternion pairs.
local A := NoSim2(B), n := nops(A), b, c, i, j, f, g;
options remember;
100
A.2 The Codes
for i from 1 to n do
b[i] := op(1, op(i, A));
c[i] := op(2, op(i, A));
end do;
for i from 1 to n do
f[i] := ZeroP([seq(b[j], j in {seq(1..n)} minus {i})]);
end do;
for i from 1 to n do
g[i] := M(Qinv(eval(f[i], b[i])), f[i]);
end do;
return sortcollect(add(M(c[i], g[i]), i = 1..n), x);
end proc:
QAdjointMatrix := proc(A::Matrix)
#Calculates the adjoint matrix of a quaternion matrix.
local nrowsA := LinearAlgebra:-RowDimension(A),
ncolsA := LinearAlgebra:-ColumnDimension(A),
A1 := Matrix(1..nrowsA, 1..ncolsA), A1C := Matrix(1..nrowsA, 1..ncolsA),
A2 := Matrix(1..nrowsA, 1..ncolsA), A2C := Matrix(1..nrowsA, 1..ncolsA),
AM := Matrix(1..2*nrowsA, 1..2*ncolsA), i, j;
for i from 1 to nrowsA do
for j from 1 to ncolsA do
A1(i, j) := A(i, j) - coeff(A(i, j), J)*J - coeff(A(i, j), K)*K;
101
A.2 The Codes
A1C(i, j) := Qconj(A1(i, j));
A2(i, j) := coeff(A(i, j), J) + coeff(A(i, j), K)*I;
A2C(i,j) := Qconj(A2(i, j));
AM(i, j) := A1(i, j);
end do;
end do;
for i from 1 to nrowsA do
for j from ncolsA + 1 to 2*ncolsA do
AM(i, j) := A2(i, j - ncolsA);
end do;
end do;
for i from nrowsA + 1 to 2*nrowsA do
for j from 1 to ncolsA do
AM(i, j) := -A2C(i - nrowsA, j);
end do;
end do;
for i from nrowsA + 1 to 2*nrowsA do
for j from ncolsA + 1 to 2*ncolsA do
AM(i, j) := A1C(i - nrowsA, j - ncolsA);
end do;
end do;
for i from 1 to 2*nrowsA do
102
A.2 The Codes
for j from 1 to 2*ncolsA do
AM(i, j) := subs(I = ii, AM(i, j));
end do;
end do; return AM;
end proc;
QMFNorm := proc(A::Matrix)
#Calculates the frobinius norm of a quaternion matrix.
local nrowsA := LinearAlgebra:-RowDimension(A),
ncolsA := LinearAlgebra:-ColumnDimension(A), i, j, N := 0;
for i from 1 to nrowsA do
for j from 1 to ncolsA do
N := N + coeff(A[i, j], I)^2 + coeff(A[i, j], J)^2
+ coeff(A[i, j], K)^2 + (Qreal(A[i, j]))^2;
end do;
end do;
return (expand(N))^(1/2);
end proc;
NewtonP := proc(B::listlist)
#Calculates the Newton polynomials and the divided difference for a list of quater-
nion pairs.
local g, i, j, d, b, c, A := NoSim2(B), n := nops(A);
options remember;
103
A.2 The Codes
for i from 1 to n do
b[i] := op(1, op(i, A)); c[i] := op(2, op(i, A));
end do;
d[0] := c[1]; g[0] := 1;
for i from 1 to n - 1 do
g[i] := ZeroP([seq(b[j], j = 1..i)]);
d[i] := M(c[i + 1] - c[1] - add(M(d[j], eval(g[j], b[i + 1])), j = 1..i - 1),
Qinv(eval(g[i], b[i + 1])));
end do;
return [seq(g[i], i = 0..n - 1)], [seq(d[i], i = 0..n - 1)];
end proc;
LFI := proc (A::Matrix)
#Calculates the generalized inverse of a quaternion polynomial matrix.
local nrowsA := LinearAlgebra:-RowDimension(A),
H := LinearAlgebra:-Transpose(Mconj(A)),
P := QMatrixMultiply(A, H),
a, B, AA, i, j, k := 0, LFI; a[0] := 1;
B[0] := LinearAlgebra:-IdentityMatrix(nrowsA);
for i from 1 to nrowsA do
AA[i] := QMatrixMultiply(P, B[i - 1]);
a[i] := -LinearAlgebra:-Trace(AA[i])/i;
B[i] := AA[i] + QScalarMultiply(a[i], B[0]);
104
A.2 The Codes
end do;
for j from 1 to nrowsA do
if a[j] <> 0 then
k := j;
end if;
end do; if k = 0 then
LFI := 0;
else
LFI := [(-1/a[k]), QMatrixMultiply(H, B[k - 1])];
end if;
end proc:
AXXB:= proc (A, B, C)
#Calculates the solutions of x2 + Ax + xB + C = 0.
local a0 := Qreal(A), a1 := coeff(A, I),
a2 := coeff(A, J), a3 := coeff(A, K),
b0 := Qreal(B), b1 := coeff(B, I),
b2 := coeff(B, J), b3 := coeff(B, K),
c := ((a0 + b0)/2)^2 - M((A + B), ((a0 + b0)/2)) + C,
c0 := Qreal(c), c1 := coeff(c, I),
c2 := coeff(c, J),c3 := coeff(c, K),
one := -4*(a1 + b1)*x^3 + 4*(a2*b3 - a3*b2 - c1)*x^2
+ (b1 - a1)*(a1^2 + a2^2 + a3^2 - b1^2 - b2^2 - b3^2)*x
105
A.2 The Codes
+ 2*(c2*(b3 - a3) + c3*(a2 - b2))*x
+ (b1 - a1)*(c1*(a1 - b1) + c2*(a2 - b2) + c3*(a3 - b3)),
two := -4*(a2 + b2)*x^3 + 4*(a3*b1 - a1*b3 - c2)*x^2
+ (b2 - a2)*(a1^2 + a2^2 + a3^2 - b1^2 - b2^2 - b3^2)*x
+ 2*(c1*(a3 - b3) + c3*(b1 - a1))*x
+ (b2 - a2)*(c1*(a1 - b1) + c2*(a2 - b2) + c3*(a3 - b3)),
three := -4*(a3 + b3)*x^3 + 4*(a1*b2 - a2*b1 - c3)*x^2
+ (b3 - a3)*(a1^2 + a2^2 + a3^2 - b1^2 - b2^2 - b3^2)*x
+ 2*(c1*(b2 - a2) + c2*(a1 - b1))*x
+ (b3 - a3)*(c1*(a1 - b1) + c2*(a2 - b2) + c3*(a3 - b3)),
d := 8*x^3 + 2*((a1 - b1)^2 + (a2 - b2)^2 + (a3 - b3)^2)*x,
f := 16*x^3
+ 8*(a1^2 + a2^2 + a3^2 + b1^2 + b2^2 + b3^2 + 2*c0)*x^2
+ 4*c0*((a1 - b1)^2 + (a2 - b2)^2 + (a3 - b3)^2)*x
+ (a1^2 + a2^2 + a3^2 - b1^2 - b2^2 - b3^2)^2*x
+ (8*(a2*b3 - a3*b2)*c1 - 4*c1^2)*x
+ (8*(a3*b1 - a1*b3)*c2 - 4*c2^2)*x
+ (8*(a1*b2 - a2*b1)*c3 - 4*c3^2)*x
- (a1*c1 + a2*c2 + a3*c3 - b1*c1 - b2*c2 - b3*c3)^2,
Z := evalf(op(select(t -> evalf(t) > 0,
map(Re + Im*ii, simplify(evalc([solve(f,x)]))))));
return map(t ->
106
A.2 The Codes
(t + subs(x = t,one/d)*I
+ subs(x = t,two/d)*J + subs(x = t,three/d)*K - (a0 + b0)/2),
{sqrt(Z), -sqrt(Z)});
end proc:
AXBc:= proc (A, B, c)
#Calculates the solutions of x2 + AxB + c = 0.
local a1 := coeff(A, I), a2 := coeff(A, J), a3 := coeff(A, K),
b1 := coeff(B, I), b2 := coeff(B, J), b3 := coeff(B, K),
one := x*(a3*b2 - a2*b3),
two := x*(a1*b3 - a3*b1),
three := x*(a2*b1 - a1*b2),
a := (a1^2 + a2^2 + a3^2)*(b1^2 + b2^2 + b3^2),
b := a1*b1 + a2*b2 + a3*b3,
d := 2*x + b,
f := x^4 + (c - (3/4)*a)*x^2
+ (1/4)*b*(4*c - a)*x + (1/4)*c*b^2,
sol := fsolve(f, x, real);
return map(t -> (t
+ subs(x = t,one/d)*I + subs(x = t,two/d)*J + subs(x = t,three/d)*K),
{sol});
end proc:
AXB:= proc (A, B)
107
A.2 The Codes
#Calculates the solutions of x2 + AxB = 0.
local a0 := Qreal(A), a1 := coeff(A, I),
a2 := coeff(A, J), a3 := coeff(A, K),
b0 := Qreal(B), b1 := coeff(B, I),
b2 := coeff(B, J), b3 := coeff(B, K),
m11 := a0*b0 - a1*b1 + a2*b2 + a3*b3,
m12 := a0*b3 - a1*b2 - a2*b1 - a3*b0,
m13 := -a0*b2 - a1*b3 + a2*b0 - a3*b1,
m21 := -a0*b3 - a1*b2 - a2*b1 + a3*b0,
m22 := a0*b0 + a1*b1 - a2*b2 + a3*b3,
m23 := a0*b1 - a1*b0 - a2*b3 - a3*b2,
m31 := a0*b2 - a1*b3 - a2*b0 - a3*b1,
m32 := -a0*b1 + a1*b0 - a2*b3 - a3*b2,
m33 := a0*b0 + a1*b1 + a2*b2 - a3*b3,
d := 8*x^3 + 4*(m11 + m22 + m33)*x^2
+ 2*(m11*m22 + m11*m33 - m12*m21 - m13*m31 - m23*m32 + m22*m33)*x
+ m11*m22*m33 - m11*m23*m32 - m12*m21*m33
+ m13*m21*m32 + m12*m23*m31 - m13*m22*m31,
n1 := -a0*b1 - a1*b0 - a2*b3 + a3*b2,
n2 := -a0*b2 + a1*b3 - a2*b0 - a3*b1,
n3 := -a0*b3 - a1*b2 + a2*b1 - a3*b0,
one := 4*n1*x^3 + 2*(n1*(m22 + m33) - n2*m12 - n3*m13)*x^2
108
A.2 The Codes
+ n1*(m22*m33 - m23*m32)*x
+ n2*(m13*m32 - m12*m33)*x
+ n3*(m12*m23 - m13*m22)*x,
two := 4*n2*x^3 + 2*(n2*(m11 + m33) - n1*m21 - n3*m23)*x^2
+ n1*(m23*m31 - m21*m33)*x
+ n2*(m11*m33 - m13*m31)*x
+ n3*(m13*m21 - m11*m23)*x,
three := 4*n3*x^3 + 2*(n3*(m11 + m22) - n1*m31 - n2*m32)*x^2
+ n1*(m21*m32 - m22*m31)*x
+ n2*(m12*m31 - m11*m32)*x
+ n3*(m11*m22 - m12*m21)*x,
ab := (a0^2 + a1^2 + a2^2 + a3^2)*(b0^2 + b1^2 + b2^2 + b3^2),
h := x^3 + 2*a0*b0*x^2 + ((4/3)*a0^2*b0^2
- (1/12)*(a0^2 - 3*a1^2 - 3*a2^2 - 3*a3^2)
*(b0^2 - 3*b1^2 - 3*b2^2 - 3*b3^2))*x
+ (1/4)*ab*(a0*b0 - a1*b1 - a2*b2 - a3*b3),
g := x^4 + 2*a0*b0*x^3
+ (1/2)*(4*a0^2*b0^2
- (a1^2 + a2^2 + a3^2 - a0^2)*(b1^2 + b2^2 + b3^2 - b0^2))*x^2
+ (1/2)*a0*b0*ab*x + (1/16)*ab^2,
solg := fsolve(g, x, real),
solh := fsolve(h, x, real);
109
A.2 The Codes
if a0 = 0 and b0 = 0 then
return map(t -> (t
+ subs(x = t,one/d)*I + subs(x = t,two/d)*J + subs(x = t,three/d)*K),
{solh});
else
return map(t -> (t
+ subs(x = t, one/d)*I + subs(x = t, two/d)*J + subs(x = t, three/d)*K),
{solg, solh});
end if;
end proc:
SemiSim:= proc (B, A)
yAx = B
#Calculates (x, y) such that .
xBy
= A
local a0 := Qreal(A), a1 := coeff(A, I), a2 := coeff(A, J),
a3 := coeff(A, K), b0 := Qreal(B), b1 := coeff(B, I),
b2 := coeff(B, J), b3 := coeff(B, K), Z;
if Qnorm(A) <> Qnorm(B) then
return "No Solution"
elif abs(a0) <> abs(b0) then
return "No Solution"
else Z := LinearAlgebra:-LinearSolve(
Matrix([[a0^2 - b0^2, -a0*a3 - b0*b3, a0*a2 + b0*b2, a0*a1 - b0*b1, 0],
110
A.2 The Codes
[a0*a3 + b0*b3, a0^2 - b0^2, -a0*a1 - b0*b1, a0*a2 - b0*b2, 0],
[-a0*a2 - b0*b2, a0*a1 + b0*b1, a0^2 - b0^2, a0*a3 - b0*b3, 0],
[-a0*a1 + b0*b1, -a0*a2 + b0*b2, -a0*a3 + b0*b3, a0^2 - b0^2, 0]]),
free = ’t’);
end if;
return sort(Z[4] + Z[1]*I + Z[2]*J + Z[3]*K, [I, J, K]) and
simplify(M(B, M(Qinv(Z[4] + Z[1]*I + Z[2]*J + Z[3]*K),Qinv(A))));
end proc:
CSSim:= proc (B, A)
yAx = B
#Calculates (x, y) such that .
xBy
= A
#Calculates x,y such that
local a0 := Qreal(A), a1 := coeff(A, I), a2 := coeff(A, J),
a3 := coeff(A, K), b0 := Qreal(B), b1 := coeff(B, I),
b2 := coeff(B, J), b3 := coeff(B, K), Z;
if Qnorm(A) <> Qnorm(B) then
return "No Solution"
else Z := LinearAlgebra:-LinearSolve(
Matrix([[0, a0*b3 + a3*b0, -a0*b2 - b0*a2, a2*b3 - a3*b2, 0],
[-a0*b3 - a3*b0, 0, a0*b1 + a1*b0, -a1*b3 + a3*b1, 0],
[a0*b2 + a2*b0, -a0*b1 - a1*b0, 0, a1*b2 - a2*b1, 0],
[-a2*b3 + a3*b2, a1*b3 - a3*b1, -a1*b2 + a2*b1, 0, 0]]),
111
A.2 The Codes
free = ’t’);
end if;
return simplify(Z[4] + Z[1]*I + Z[2]*J + Z[3]*K) and
simplify(M(M(Qinv(B),Qinv(Qconj(Z[4] + Z[1]*I + Z[2]*J + Z[3]*K))),A))
end proc:
LN1:= proc (A, B, C)
#Calculates the solution of min |ax − xb − c|.
x∈H
local a0 := Qreal(A), a1 := coeff(A, I), a2 := coeff(A, J),
a3 := coeff(A, K), b0 := Qreal(B), b1 := coeff(B, I),
b2 := coeff(B, J), b3 := coeff(B, K), c0 := Qreal(C),
c1 := coeff(C, I), c2 := coeff(C, J), c3 := coeff(C, K), Z;
if Qnorm(A) = Qnorm(B) and Qreal(A) = Qreal(B) then
Z := LinearAlgebra:-MatrixMatrixMultiply(LinearAlgebra:-MatrixInverse(
Matrix([[a0 - b0, -a3 - b3, a2 + b2, a1 - b1],
[a3 + b3, a0 - b0, -a1 - b1, a2 - b2],
[-a2 - b2, a1 + b1, a0 - b0, a3 - b3],
[-a1 + b1, -a2 + b2, -a3 + b3, a0 - b0]]), method = pseudo),
Matrix(4, 1, {(1, 1) = c1, (2, 1) = c2, (3, 1) = c3, (4, 1) = c0}));
else
Z := LinearAlgebra:-LinearSolve(
Matrix([[a0 - b0, -a3 - b3, a2 + b2, a1 - b1, c1],
[a3 + b3, a0 - b0, -a1 - b1, a2 - b2, c2],
112
A.2 The Codes
[-a2 - b2, a1 + b1, a0 - b0, a3 - b3, c3],
[-a1 + b1, -a2 + b2, -a3 + b3, a0 - b0, c0]]), free = ’t’);
end if;
return simplify(Z[4] + Z[1]*I + Z[2]*J + Z[3]*K);
end proc:
LN2:= proc (A,B,C)
#Calculates the solution of min |ax − x̄b − c|.
x∈H
local a0 := Qreal(A), a1 := coeff(A, I), a2 := coeff(A, J),
a3 := coeff(A, K), b0 := Qreal(B), b1 := coeff(B, I),
b2 := coeff(B, J), b3 := coeff(B, K), c0 := Qreal(C),
c1 := coeff(C, I), c2 := coeff(C, J), c3 := coeff(C, K), Z;
if Qnorm(A) = Qnorm(B) and Qreal(A) = Qreal(B) then
Z := LinearAlgebra:-MatrixMatrixMultiply(LinearAlgebra:-MatrixInverse(
Matrix([[a0 + b0, -a3 + b3, a2 - b2, a1 - b1],
[a3 - b3, a0 + b0, -a1 + b1, a2 - b2],
[-a2 + b2, a1 - b1, a0 + b0, a3 - b3],
[-a1 - b1, -a2 - b2, -a3 - b3, a0 - b0]]), method = pseudo),
Matrix(4, 1, {(1, 1) = c1, (2, 1) = c2, (3, 1) = c3, (4, 1) = c0}));
else
Z := LinearAlgebra:-LinearSolve(
Matrix([[a0 + b0, -a3 + b3, a2 - b2, a1 - b1, c1],
[a3 - b3, a0 + b0, -a1 + b1, a2 - b2, c2],
113
A.2 The Codes
[-a2 + b2, a1 - b1, a0 + b0, a3 - b3, c3],
[-a1 - b1, -a2 - b2, -a3 - b3, a0 - b0, c0]]), free = ’t’);
end if;
return simplify(Z[4] + Z[1]*I + Z[2]*J + Z[3]*K);
end proc:
end module:
114
A.2 The Codes
115
A.2 The Codes
116
A.2 The Codes
117
A.2 The Codes
118
Bibliography
Adler, S., 1995. Quaternionic quantum mechanics and quantum fields. International
Series of Monographs on Physics. Oxford University Press, USA.
Au-Yeung, Y., 2003. An explicit solution for the quaternionic equation x2 + bx +
xc + d = 0. Southeast Asian Bull. Math. 26 (5), 717–724.
Baker, A., 1999. Right eigenvalues for quaternionic matrices: a topological approach.
Linear Algebra Appl. 286 (1-3), 303–309.
Bjõrck, A., 1996. Numerical methods for least squares problems. Society for Indus-
trial and Applied Mathematics.
Bray, U., Whaples, G., 1983. Polynomials with coefficients from a division ring.
Canad. J. Math. 35 (3), 509–515.
Brenner, J. L., 1951. Matrices of quaternions. Pacific J. Math. 1, 329–335.
Carter, M., March 2007. Quaternions. Maple package. Maplesoft.
URL http://www.maplesoft.com/applications/view.aspx?SID=4886
Cayley, A., 1894. On coordinates versus quaternions. Proc. Roy. Soc. Edinburgh 20,
271–275.
Churchill, L., 1990. Quaternion analysis in physics. East Carolina University.
Decell, H., 1965. An application of the Cayley-Hamilton theorem to generalized
matrix inversion. SIAM Rev. 7, 526–528.
119
Bibliography
Derbyshire, J., 2006. Unknown quantity: a real and imaginary history of algebra.
Joseph Henry Press.
Eilenberg, S., Niven, I., 1944. The “fundamental theorem of algebra” for quaternions.
Bull. Amer. Math. Soc. 50, 246–248.
Faddeeva, V., 1959. Computational methods of linear algebra. Dover books on ad-
vanced mathematics. Dover Publications.
Finkelstein, D., Jauch, J., Schiminovich, S., Speiser, D., 1962. Foundations of quater-
nion quantum mechanics. Journal of Mathematical Physics 3 (2), 207–220.
Gibbs, J., Bumstead, H., Van Name, R., Longley, W., 1902. The collected works of
J. Willard Gibbs. No. v. 2 in The Collected Works of J. Willard Gibbs. Longman,
Green and Co.
Gordon, B., Motzkin, T. S., 1965. On the zeros of polynomials over division rings.
Trans. Amer. Math. Soc. 116, 218–226.
Hanson, A., 2005. Visualizing quaternions. The Morgan Kaufmann Series In Inter-
active 3D Technology. Morgan Kaufmann.
Harder, D., September 2010. Quaternions. Maple package. Maplesoft.
URL http://www.maplesoft.com/applications/view.aspx?SID=96897
Hastings, C., 2010. Biographical Memoir of Josiah Willard Gibbs, 1839-1903: By
Charles S. Hastings. BiblioBazaar.
Horn, R., Johnson, C., 1990. Matrix analysis. Cambridge University Press.
Huang, L., So, W., 2002. Quadratic formulas for quaternions. Appl. Math. Lett.
15 (5), 533–540.
Johnson, R. E., 1944. On the equation χα = γχ + β over an algebraic division ring.
Bull. Amer. Math. Soc. 50, 202–207.
120
Bibliography
Kalman, D., 1978. An extension of the remainder theorem. Delta 8, 77–80.
Karampetakis, N. P., 1997. Computation of the generalized inverse of a polynomial
matrix and applications. Linear Algebra Appl. 252, 35–60.
Kuipers, J., 2002. Quaternions and rotation sequences: a primer with applications
to orbits, aerospace and virtual reality. Princeton University Press.
Lam, T., 2001. A first course in noncommutative rings. Graduate Texts in Mathe-
matics (131). Springer.
Lee, H. C., 1949. Eigenvalues and canonical forms of matrices with quaternion co-
efficients. Proc. Roy. Irish Acad. Sect. A. 52, 253–260.
Maxwell, J., 1873. Quaternions. Nature 9, 137–138.
McAulay, A., 1892. Quaternions as a practical instrument of physical research. Philo-
sophical Magazine 33, 477–495.
Morandi, P., 1996. Field and Galois theory. Graduate Texts in Mathematics.
Springer.
Muir, T., 2003. A treatise on the theory of determinants. Dover Phoenix Editions.
Dover Publications.
Nahin, P., 2002. Oliver Heaviside: the life, work, and times of an electrical genius
of the Victorian age. Oliver Heaviside. Johns Hopkins University Press.
Niven, I., 1941. Equations in quaternions. Amer. Math. Monthly 48, 654–661.
Niven, I., 1942. The roots of a quaternion. Amer. Math. Monthly 49, 386–388.
Penrose, R., 1955. A generalized inverse for matrices. Proc. Cambridge Philos. Soc.
51, 406–413.
Puystjens, R., de Smet, H., 1980. The Moore-Penrose inverse for matrices over skew
121
Bibliography
polynomial rings. Ring Theory Antwerp, Lecture Notes in Mathematics 825, 94–
103.
Smith, C., Wise, N., 1989. Energy and empire: a biographical study of Lord Kelvin.
Cambridge University Press.
Stanimirović, P., Petković, M., 2006. Computing generalized inverse of polynomial
matrices by interpolation. Appl. Math. Comput. 172 (1), 508–523.
Stanimirović, P., Tasić, M., Krtolica, P., Karampetakis, N., 2007. Generalized in-
version by interpolation. Filomat 21 (1), 67–86.
Tait, P., 1890. Vii. On the importance of quaternions in physics. Philosophical Mag-
azine Series 5 29 (176), 84–97.
Tian, Y., 2004. The equations ax − xb = c, ax − xb = c and xax = b in quaternions.
Southeast Asian Bull. Math. 28 (2), 343–362.
Tian, Y., 2010. Solving two pairs of quadratic equations in quaternions. Adv. Appl.
Clifford Algebr. 20 (1), 185–193.
van der Waerden, B., 1985. A history of algebra: from al-Khwārizmı̄ to Emmy
Noether. Springer.
Zhang, F., 1997. Quaternions and matrices of quaternions. Linear Algebra and its
Applications 251 (0), 21–57.
122
You might also like
- Lectures On Groups of Transformations: J. L. KoszulNo ratings yetLectures On Groups of Transformations: J. L. Koszul85 pages
- Hatcher A. - Algebraic Topology. Errata (Web Draft, Nov. 2004) PDFNo ratings yetHatcher A. - Algebraic Topology. Errata (Web Draft, Nov. 2004) PDF6 pages
- Exactly Solvable Schrödinger Equations With Singularities: A Systematic Approach To Solving Complexified Potentials (Part1)100% (1)Exactly Solvable Schrödinger Equations With Singularities: A Systematic Approach To Solving Complexified Potentials (Part1)36 pages
- Differential Topology in Equilibrium TheoryNo ratings yetDifferential Topology in Equilibrium Theory494 pages
- Mathematical Methodsoftheo-Reticalphysics: Karlsvozil100% (2)Mathematical Methodsoftheo-Reticalphysics: Karlsvozil275 pages
- Patidar Nonstandard Finite Difference Methods PDF100% (1)Patidar Nonstandard Finite Difference Methods PDF34 pages
- Read Online Here: Introduction To Abstract Algebra Solutions 2No ratings yetRead Online Here: Introduction To Abstract Algebra Solutions 25 pages
- Entropy and Perpetual Computing InsightsNo ratings yetEntropy and Perpetual Computing Insights9 pages
- Representations and Cohomology Ii: Cambridge Studies in Advanced Mathematics 31100% (1)Representations and Cohomology Ii: Cambridge Studies in Advanced Mathematics 31291 pages
- 01-S - Antoniou - Lie Symmetries, Differential Equations and Applications100% (1)01-S - Antoniou - Lie Symmetries, Differential Equations and Applications1,451 pages
- Duistermaat J.J., Kolk J.a.C. - Multidimensional Real Analysis II - Integration (2004)No ratings yetDuistermaat J.J., Kolk J.a.C. - Multidimensional Real Analysis II - Integration (2004)396 pages
- Math Symbol Tables: B.1 Hebrew and Greek LettersNo ratings yetMath Symbol Tables: B.1 Hebrew and Greek Letters20 pages
- Eighteen Essays in Non-Euclidean Geometry: Vincent Alberge Athanase Papadopoulos EditorsNo ratings yetEighteen Essays in Non-Euclidean Geometry: Vincent Alberge Athanase Papadopoulos Editors477 pages
- Transform Methods: Partial Differential EquationsNo ratings yetTransform Methods: Partial Differential Equations4 pages
- A Review On Cartans Structure Equations For CertaNo ratings yetA Review On Cartans Structure Equations For Certa7 pages
- A Primer of Infinitesimal Analysis - Portada0% (1)A Primer of Infinitesimal Analysis - Portada4 pages
- A Postquantum Theory of Classical GravityNo ratings yetA Postquantum Theory of Classical Gravity37 pages
- Review of Quaternion Differential EquatiNo ratings yetReview of Quaternion Differential Equati19 pages
- Dickson QuaternionsGeneralizationHistory 1919No ratings yetDickson QuaternionsGeneralizationHistory 191918 pages
- JAMP-Algebraic Techniques For Linear EquationsNo ratings yetJAMP-Algebraic Techniques For Linear Equations14 pages
- Szynal-Liana-Włoch2021 Article GeneralizedCommutativeQuaterniNo ratings yetSzynal-Liana-Włoch2021 Article GeneralizedCommutativeQuaterni9 pages
- MMA-Algebraic Techniques For Least Squares Problem OverNo ratings yetMMA-Algebraic Techniques For Least Squares Problem Over14 pages
- Koleksi Soalan Matematik Tambahan SPM 2004-2009-Differentiation50% (6)Koleksi Soalan Matematik Tambahan SPM 2004-2009-Differentiation10 pages
- IIT-JEE 2004 Mains Questions & Solutions - Maths - Version 2 (The Questions Are Based On Memory)No ratings yetIIT-JEE 2004 Mains Questions & Solutions - Maths - Version 2 (The Questions Are Based On Memory)15 pages
- Theory of Errors Chapter 9 391668376849522No ratings yetTheory of Errors Chapter 9 3916683768495227 pages
- Boyce/Diprima 9 Ed, CH 2.5: Autonomous Equations and Population DynamicsNo ratings yetBoyce/Diprima 9 Ed, CH 2.5: Autonomous Equations and Population Dynamics22 pages
- No Rules Rules 1st Edition Reed Hastings & Erin Meyer (Hastings Download100% (15)No Rules Rules 1st Edition Reed Hastings & Erin Meyer (Hastings Download115 pages
- Civil Engineering Unit 1: Engineering MathematicsNo ratings yetCivil Engineering Unit 1: Engineering Mathematics3 pages
- LO - Unit-3 - Decimals, Percentages & RoundingNo ratings yetLO - Unit-3 - Decimals, Percentages & Rounding7 pages
- 7.9 Age Word Problems - Intermediate AlgebraNo ratings yet7.9 Age Word Problems - Intermediate Algebra7 pages
- Zimbabwe School Examinations Council Pure Mathematics 6042/2100% (4)Zimbabwe School Examinations Council Pure Mathematics 6042/28 pages
- Equation Calculation Guide for Graphing CalculatorsNo ratings yetEquation Calculation Guide for Graphing Calculators9 pages
- Slope and Intercepts in Linear EquationsNo ratings yetSlope and Intercepts in Linear Equations387 pages
