Professional Documents
Culture Documents
Complex Numbers Geometry and Algebra of
Complex Numbers Geometry and Algebra of
Complex Numbers Geometry and Algebra of
Rastko Vukovic∗
Abstract
It is considered the elements of complex numbers. In particular, rotation in
standard complex plane, the real product (dot product), with some applications in
geometry. The generalizations to complex matrices and quaternions are included.
You will notice that the known definition of dot product in standard theory is not
valid more, so I suggested to call it “compound sum”.
1 Introduction
The complex number
√ can be expressed in the form z = x + iy, where x and y are real
numbers and i = −1 is the imaginary unit. The set of complex numbers is C.
Each complex number z = x + iy ∈ C has its real and imaginary part, respectively
<(z) = x and =(z) = y, which can be represented as two orthogonal axes of the Cartesian
coordinate system Oxy, on the figure 1. The distance Oz = |z| is called modulus of the
complex number z, and the angle Arg(z) = φ is argument.
We may use the complex numbers when quantities are not one-dimensional scalar,
but alternate in direction and amplitude. If they are not just a “math oddity”, or the
necessity in quantum mechanics, they can be used for instance, in electrical engineering
∗
Math prof in Gimnazija Banja Luka.
1
Rastko Vukovic: Complex Numbers
to calculate voltage, current and resistance (impedance in AC), or frequency and phase
shift at one time. In the study of electricity and electronics, j is used to represent
imaginary numbers so that there is no confusion with i, which in electronics represents
current.
A special complex number is an imaginary unit, for now marked by i with modulus
|i| = 1 and argument arg(i) = π/2. By changing the sign of the imaginary unit gives the
number −i, which is called conjugated with i. By changing the sign of i in a complex
number z = x + iy the number z becomes conjugated z̄ = x − iy.
Conjugation distributes over the arithmetic operations:
z1 ◦ z 2 = z 1 ◦ z2 , (1)
where “◦” is one of the basic arithmetic operations: addition, subtraction, multiplication
or division. It can be easily proved by substitutions zk = xk + iyk , k = 1, 2. The first
consequence is the identity
p(z) = p(z̄), (2)
for any polynomial p(z) = a0 + a1 z + a2 z 2 + · · · + an z n where a0 , a1 , a2 , . . . , an ∈ C.
Imaginary unit i and it’s conjugated −i are respectively the points (0, 1) and (0, −1)
in C plane. Generally, the reflection of the point z = x + iy about the real p axes gives
the conjugate point x̄ = x − iy. The both p modulus are the same, |z̄| = |z| = x2 + y 2 ,
but Arg(z̄) = − Arg(z). However z z̄ = x2 + y 2 ∈ R.
This is the enter into the geometry of complex numbers C or points in complex plane
C. For example, the inequality |z| < 1 represents the points inside the circle with center
in origin of the coordinates and unit radius. The second example, mark the distance
ρ = Oz on the figure 1. Than we have the transformations:
ρ = x2 + y 2 , φ = arctan xy ;
p
(3)
x = ρ cos φ, y = ρ sin φ.
On the other side, modulus and argument of the complex number z = x + iy can be
calculated by comparing z = ρ(cos φ + i sin φ) and:
!
p x y
z = x + iy = x2 + y 2 p + ip ,
x2 + y 2 x2 + y 2
which gives:
x y
cos φ = p , sin φ = p . (4)
x + y2
2 x + y2
2
2 Rotation
Complex numbers are added by coordinates, real with real, imaginary with imaginary,
that is (x1 , y1 ) + (x2 , y2 ) = (x1 + x2 , y1 + y2 ), or
Proof. On figure 3, the rectangular triangles Oz1 x1 and Oxz are congruent, so ∠Oz1 x1 =
∠xOz = φ, and ∠zOz1 = π2 , so Oz1 ⊥ Oz.
Theorem 1. Multiplying complex numbers z and w, their lengths get multiplied and
their arguments get added.
Proof. Let us take the arbitrary complex numbers z = x + iy and w = u + iv and find
z1 = zw. Then z1 = zw = (x + iy)w = xw + iyw, as we can see on figure 4. There is
Arg(w) = ϕ and we are looking for angle θ = ∠wOz1 .
Since iw ⊥ w, the line segment from O to iyw is perpendicular to the segment from
O to xw. Therefore O, xw, z1 forms a right triangle. The length of the opposite side to
angle θ is y|w|, the length of the adjacent side is y|w|. The ratio of these two lengths is
y : x = tan θ. Hence θ = Arg(z).
The angle θ on figure 5 is the same as angle φ on figure 3. The length of the
hypotenuse of the right triangle O, xw, z1 is |w| times hypotenuse of triangle O, x, z,
so |zw| = |z||w|. Because the argument of zw is the angle θ + ϕ, and θ = Arg(z),
ϕ = Arg(w), we get the two formulas:
|zw| = |z||w|, Arg(zw) = Arg(z) + Arg(w), (7)
and this is what was to be proved.
z1 z2 = ρ1 ρ2 [(cos φ1 cos φ2 − sin φ1 sin φ2 ) + i(sin φ1 cos φ2 + cos φ1 sin φ2 )]. (9)
Directly, by using the trigonometry addition formulas for cosine and sine:
cos(φ1 + φ2 ) = cos φ1 cos φ2 − sin φ1 sin φ2 ,
(10)
sin(φ1 + φ2 ) = sin φ1 cos φ2 + cos φ1 sin φ2 ,
we find
z1 z2 = ρ1 ρ2 [cos(φ1 + φ2 ) + i sin(φ1 + φ2 )], (11)
which is equivalent to (7). Reversely, using the result (7), from (9) we prove the trigonom-
etry formulas (10). So, (7) is equivalent to the trigonometric addition formulas (10).
As we know from the trigonometry, by dividing formulas (10) stems:
The arguments are Arg(w) = π/4, Arg(z1 ) = φ1 , Arg(z) = φ, and mark the angle
φ1 − φ = θ. Then:
tan φ1 − tan φ
tan θ = tan(φ1 − φ) = =
1 + tan φ1 tan φ
x+y y
x−y − x x(x + y) − y(x − y)
= x+y y = = 1,
1+ x−y x
x(x − y) + y(x + y)
that is θ = π/4. But Arg(w) = π/4 too.
The rotation of unit number z = cis(φ), which applies |z| = 1, are especially inter-
esting. Each unit complex number can be written as:
For example:
which means that for any n = 1, 2, 3, . . . exists the angle φ = 2π/n for which is nφ = 2π.
These numbers z for which is z n = 1 are called the (n-th) roots of unity or de Moivre
nubers. So, numbers (14) are (4-th) roots of unit.
√
Example 2. Find all the roots of the equation z = 3 2.
√
Solution. The solutions are zk = 3 2, for k = 0, ±1, ±2, . . . , precisely:
p
zk = 3 2(cos 2kπ + i sin 2kπ), k ∈ Z,
√
3 2kπ 2kπ
zk = 2(cos
+ i sin ),
3 3
but 2kπ/3 = 0, ±2π/3, ±4π/3, ±6π/3, . . . , and we have only 3 different solutions,
let’s say for k = 0, 1, 2. So, the asked roots are:
( √
z0 = 3 2,
√ √
z12 = 3 2(− 21 ± i 23 ).
Check them in z 3 = 2.
Multiplication by cis(φ) causes the rotation of multiplied number for angle θ without
changing its modulus. It is convenient to define the rotation in the plane C as isometric
transformation. From z1 cis(θ) = z2 we get:
x2 = x1 cos θ − y1 sin θ,
(16)
y2 = x1 sin θ + y1 cos θ,
or
x2 cos θ − sin θ x1
= , (17)
y2 sin θ cos θ y1
for z1 = x1 + iy1 before, and z2 = x2 + iy2 after the rotation. In this way, complex
numbers reveal its vector and the matrix structure.
3 Dot Product
The compound sum (5) in the complex plane C is:
1
z1 ⊕ z2 = (z1 z̄2 + z̄1 z2 ) =
2
1
= [(x1 + iy1 )(x2 − iy2 ) + (x1 − iy2 )(x2 + iy2 )]
2
1 1
= [(x1 x2 + y1 y2 ) − i(x1 y2 − x2 y1 )] + [(x1 x2 + y1 y2 ) + i(x1 y2 − y1 x2 )],
2 2
z1 ⊕ z2 = x 1 x 2 + y1 y2 , (18)
where z1 = x1 + iy1 and z2 = x2 + iy2 , for x1 , x2 , y1 , y2 ∈ R and i2
= −1. Thus,
compound sum for the plane complex numbers is the real number, which resembles to
the scalar product of the vectors. Emphasizing this similarity, we define
z1 · z2 = x1 y1 + x2 y2 , (19)
for z1 = x1 + iy1 and z2 = x2 + iy2 . Expression (19) we call dot product, or real product
of the complex numbers z1 and z2 .
The dot product (19) have the next proprieties, which is easy to prove.
Lemma 2. For dot product (19) is valid:
1. z1 · z2 = 21 (z1 z̄2 + z̄1 z2 );
2. z1 · z2 = z1 · z2 ;
3. z1 · z2 = z2 · z1 ;
4. z · z = |z|2 ;
5. z · (z1 + z2 ) = z · z1 + z · z2 ;
6. λ(z1 · z2 ) = (λz1 ) · z2 = z1 · (λz2 ), for all λ ∈ R;
7. z1 · z2 = 0 ⇐⇒ Oz1 ⊥ Oz2 ;
8. (zz1 ) · (zz2 ) = |z|2 (z1 · z2 );
for all complex numbers z1 = x1 + iy1 , z2 = x2 + iy2 and z = x + iy.
This is the basis of algebra of dot multiplication of complex numbers. The following
statement links this multiplication with scalar multiplication of vectors.
Theorem 3. For a dot product is valid:
z1 · z2 = |z1 ||z2 | cos θ,
where z1 = x1 + iy1 , z2 = x2 + iy2 and θ = ∠z1 Oz2 .
Proof. Using (19) we get z1 · z2 = x1 x2 + y1 y2 =
= x1 (x1 cos θ − y1 sin θ) + y1 (x1 sin θ + y1 cos θ) = (x21 + y12 ) cos θ.
From the same (16) we have x22 + y22 = x21 + y12 , than
q q
z1 · z2 = x21 + y12 x22 + y22 cos θ,
and this is what is supposed to prove.
By comparing arising:
which means, diagonals of the parallelogram are equal (z1 z3 = z2 z4 ), iff2 the adjacent
sides of the parallelogram are perpendicular (z1 z2 ⊥ z2 z3 ).
Task 1. Prove the Thales’ Theorem3 by dot product of the complex numbers:
The diameter of a circle always subtends a right angle to any point on the circle.
Task 2. Prove the Maxwell’s Theorem4 :
Given ∆ABC and a point P , the sides of ∆A0 B 0 C 0 are parallel to the cevians in
∆ABC through P . Prove that the cevians in ∆A0 B 0 C 0 parallel to the sides of ∆ABC
are concurrent.
1
Law of cosines: https://en.wikipedia.org/wiki/Law_of_cosines
2
if and only if
3
Thales’ Theorem: www.mathopenref.com/thalestheorem.html
4
Maxwell’s Theorem: www.cut-the-knot.org/Curriculum/Geometry/Maxwell.shtml
The next is (17), the equation that reveals the matrix structure of the complex
“numbers”. The matrix of rotation
cos θ − sin θ
cis(θ)
c = , (21)
sin θ cos θ
That is equivalent of the complex number z = ρ(cos θ + i sin θ). Given that z = x + iy,
the complex matrix is
x −y 1 0 0 −1
zb = =x +y , (23)
y x 0 1 1 0
or ẑ = xIˆ + y î, where Iˆ is the unit matrix and î matrix of imaginary unit.
It is expected as it is easy to prove by matrix multiplication:
2
2 0 −1 0 −1 0 −1 −1 0 ˆ
î = = = = −I.
1 0 1 0 1 0 0 −1
But the same, raises the question: Is there any another matrix, dimension 2 × 2, whose
square is minus unit?
ˆ in the square matrix of order 2.
Example 5. Solve the matrix equation ê = −I,
If bc 6= 0, from the second and third equation from the system we get d = −a and
a2 = −1–bc. But a2 ≥ 0 leads to bc ≤ −1. The first solution is (generated from) b = −1,
c = 1, than a = d = 0. That is the matrix î. √
But, the equation bc = −1 has another simple solution: b = c = i = −1, and again
a = d = 0. Let say it is matrix ĵ.
Next, if bc = 0, than a2 = d2 = −1. The basic solution is b = c = 0, d = −a = i.
That is the matrix k̂.
ˆ
with the required feature î2 = ĵ 2 = k̂ 2 = −I.
For now we are working only with one, î, of these three matrices (24), which together
with unit matrix Iˆ expands the 2-dim vector space. Such space is analog to the complex
plane C. Let’s look at it’s appropriate properties once more.
Conjugate numbers z = x + iy are z̄ = x − iy of property z z̄ = x2 + y 2 . Likewise,
the conjugate matrices are:
x −y ∗ x y
ẑ = , ẑ = , (25)
y x −y x
of property: 2
x + y2
∗ 0 ˆ
ẑ ẑ = = (x2 + y 2 )I. (26)
0 x2 + y 2
Consistently, we define compound sum (5) for the complex matrices:
1
ẑ1 ⊕ ẑ2 = (ẑ1 ẑ2∗ + ẑ1∗ ẑ2 ) =
2
1 x1 −y1 x 2 y2 x1 y1 x2 −y2
= +
2 y1 x1 −y2 x2 −y1 x1 y2 x2
1 x1 x2 + y1 y2 x1 y2 − y1 x2 x1 x2 + y1 y2 −x1 y2 + y1 x2
= +
2 y1 x2 − x1 y2 y1 y2 + x1 x2 −y1 x2 + x1 y2 y1 y2 + x1 x2
1 2(x1 x2 + y1 y2 ) 0 1 0
= = (x1 x2 + y1 y2 ) .
2 0 2(x1 x2 + y1 y2 ) 0 1
So:
ẑ1 ⊕ ẑ2 = (x1 x2 + y1 y2 )Iˆ = ẑ1 · ẑ2 , (27)
because this is analog to (19), to the dot product z1 · z2 = x1 x2 + y1 y2 .
Equivalent of the square modulus |z|2 = z z̄ is determinant:
x −y x −y
= x2 + y 2 ,
det(ẑ) = det = (28)
y x y x
cos φ1 − sin φ1 cos φ2 − sin φ2
ẑ1 ẑ2 = ρ1 ρ2 ,
sin φ1 cos φ1 sin φ2 cos φ2
cos(φ1 + φ2 ) − sin(φ1 + φ2 )
ẑ1 ẑ2 = ρ1 ρ2 , (29)
sin(φ1 + φ2 ) cos(φ1 + φ2 )
which is the analog of the theorem 1. Next stands for analog of the theorem 3:
ˆ
p p
ẑ1 · ẑ2 = det(ẑ1 ) det(ẑ2 ) cos θ I, (30)
but the angle θ has no more the same sense.
As the unit matrix Iˆ with î extends plane C, so Iˆ makes with ĵ some 2-dim space
analog to C. However, the matrix k̂ is not forming the similar 2-dim space. Besides,
the same matrices must be used a few times. That’s why these matrices form “only” a
4-dim space. This can be proven by using the features of vector spaces.
The linear combination of these four matrices, for x, y, u, v ∈ R, is:
x − iv −y + iu z̄ − w̄
q̂ = xIˆ + y î + uĵ + v k̂ = = . (31)
y + iu x + iv w z
where z = x + iv, w = y + iu are arbitrary. Than q̂ = 0̂ iff both z = 0 and w = 0, that
is x = y = u = v = 0. Accordingly, these four matrices make exactly 4-dim space.
4 Quaternions
The three matrices (24) with unit Iˆ and the linear combinations (31) we call quaternions.
The base5 vectors of the 4-dim quaternion space are all four matrices:
ˆ 1 0 0 −1 0 i −i 0
I= , î = , ĵ = , k̂ = . (32)
0 1 1 0 i 0 0 i
To form the quaternions we need variables x, y, u, v ∈ R on the corresponding axes, like
in Figure 7.
Especially, depending on the coordinate axes, there are six possible 2-dim quaternion
subspaces :
Oxy : xIˆ + y î, Oxu : xIˆ + uĵ, Oxv : xIˆ + v k̂,
(33)
Oyu : y î + uĵ, Oyv : y î + v k̂, Ouv : uĵ + v k̂.
5
quaternion base can be defined and otherwise, as in [2]
each of them similar to C. Still, the planes Oxv and Oyu (opposite axes) are more alike
than other four (adherent axes). The following statement helps in future calculations.
This is easily proved by direct matrix multiplication. Note that the multiplication
of the successive matrices in the cyclic series î, ĵ, k̂, î, ĵ, etc. in right direction produces
positive, in reverse negative
√ result. Now backward, we could define three “ordinary”
imaginary units i, j, k = −1, in accordance with the lemma.
The general form of quaternions for the above mentioned four coordinates is (31).
Consistently, if we like conjugates of imaginary units i, j, k to be −i, −j, −k, and to hold
distribution laws (1), we need the definitions:
The immediate consequence is: conjugation of matrix is the transposition of the matrix
and conjugation of all its elements. So:
x − iv −y + iu ∗ x + iv y − iu
q̂ = , q̂ = , (35)
y + iu x + iv −y − iu x − iv
or:
z̄ −w̄ ∗ z w̄
q̂ = , q̂ = , (36)
w z −w z̄
where z = x + iv, w = y + iu, i2 = −1.
Example 7. Prove:
1 1
(q̂ + q̂ ∗ ) = xI,
ˆ (q̂ q̂ + q̂ ∗ q̂ ∗ ) = (x2 − y 2 − u2 − v 2 )I,
ˆ
2 2
for the quaternion (36).
Solution. The first is obvious. Check the second, successive:
2
z̄ −w̄ z̄ −w̄ z̄ − w̄w −z̄ w̄ − w̄z
q̂ q̂ = = ,
w z w z wz̄ + zw −ww̄ + z 2
2
∗ ∗ z w̄ z w̄ z − w̄w z w̄ + w̄z̄
q̂ q̂ = = ,
−w z̄ −w z̄ −wz − z̄w −ww̄ + z̄ 2
2
x − y 2 − u2 − v 2
1 ∗ ∗ 0
(q̂ q̂ + q̂ q̂ ) = ,
2 0 x2 − y 2 − u2 − v 2
which is correct.
For lemma 3, the matrices î, ĵ, k̂ are multiplied as unit vectors ~i, ~j, ~k in rectangular
Cartesian coordinate system Oyuv. It gives us the idea to split quaternion matrix into
two matrices, or just to watch the rest:
ˆ
r̂ = q̂ − xI, (37)
that is r̂ = y î + uĵ + v k̂. Analog Cartesian vectors, let us define dot and cross product
for the rest matrix:
î ĵ k̂
r̂1 · r̂2 = y1 y2 + u1 u2 + v1 v2 , r̂1 × r̂2 = y1 u1 v1 , (38)
y2 u2 v2
Notice that the result (41) incorporates many of the previous. When we choose a
2-dim subspace (33), in the cases of Oxy, Oxu and Oxv we have “complex” spaces very
similar to C, but in spaces without x (real) axes, that is not.
For example, dot product stay in the same space if it has the real axes (r̂1 × r̂2 = 0).
However, it transforms points from 2-dim spaces with both imaginary axes into some of
the first cases.
References
[1] Titu Andreescu, Dorin Andrica: : Complex Numbers, from A to ... Z, Second Edition,
ISBN-13: 978-0817684143, 2010.
[2] Rastko Vukovic: Quaternion fields, The six dimensions of space-time, GSJ6 , 2014.
6
Quaternion filds: http://www.gsjournal.net/Science-Journals/Research%
20Papers-Mathematics%20and%20Applied%20Mathematics/Download/5791