Notes for Calculus A (2024/25) 87

13. Complex numbers and polar coordinates

It is not secret that some of you already know what complex numbers are. Those
look nothing else but a fancy way to represent elements of a two-dimensional vector
space over the real numbers: one just writes x + iy ∈ C for a vector (x, y) ∈ R2
saying that i is the ‘imaginary’ unit. This characterisation of complex numbers
straightforwardly leads to their depiction in the Cartesian coordinates of the plane.
Well, there is a bit more in x+iy than just a traditional mathematics symbolism. . .
n
There is no good (= natural) notion for a product of two vectors
√ in R , but complex
numbers can be multiplied naturally by identifying i with −1, a (formal) root of
the quadratic equation t2 + 1 = 0. How it works? Just do it:
(x1 + iy1 )(x2 + iy2 ) = x1 x2 + ix1 y2 + iy1 x2 + i2 y1 y2 = x1 x2 + ix1 y2 + iy1 x2 − y1 y2
= (x1 x2 − y1 y2 ) + i(x1 y2 + y1 x2 ),
which is a new complex number corresponding to (x1 x2 − y1 y2 , x1 y2 + y1 x2 ) ∈ R2 in
the coordinate notation.
The coordinates x and y of z = x + iy ∈ C are called real and imaginary parts
of z, the notation Re z = x and Im z = y; the conjugate of z is z = x − yi. The
modulus (or absolute value) of z coincides with what it is in R2 :
√ p
|z| = z z = x2 + y 2 .
One more notion related to the geometric visualisation of complex numbers is the
argument of z = x + iy ∈ C. It is only defined for z ̸= 0 as an angle θ ∈ R such
that x = |z| cos θ and y = |z| sin θ. (The argument arg z = θ is an example of a
multi-valued function, because the angle is only defined modulo 2πZ; for practical
purposes, one chooses −π < θ ≤ π or 0 ≤ θ < 2π. The Book uses notation Arg z
for the argument in the range −π < θ ≤ π and calls it the principal argument.)
p
Remark 13.1. The characteristics r = x2 + y 2 ≥ 0 and angle θ for a point (x, y)
are also known as the polar coordinates of the point. A standard geometric argument
then gives the formulae x = r cos θ, y = r sin θ. The polar coordinates are useful
instruments for visualisation of several important graphs of equations and their
scaling/rotation (you may recall two different equations x2 − y 2 = 2 and xy = 1
of the hyperbola after a 90◦ turn from Lecture 3). The polar coordinates make an
important appearance in two-dimensional mathematics and physics problems; we
will see applications in Calculus B.
Passing to the polar coordinates often allows to get a simple parametrisation of
somewhat involved implicit equations. For example, the folium of Descartes from
Exercise 9.10(h) is given by the equation x3 + y 3 = 3axy, where a > 0 is a fixed
parameter. After taking x = r cos θ and y = r sin θ, the equation becomes
r3 (cos3 θ + sin3 θ) = 3ar2 sin θ cos θ,
which for r ̸= 0 translates into
3a sin θ cos θ
r=
cos3 θ + sin3 θ
88 WADIM ZUDILIN

(which can be further simplified using sin θ cos θ = 12 sin 2θ and u3 +v 3 = (u+v)(u2 −
uv + v 2 )). In the final expression r = 0 is not lost, because it corresponds to θ = 0.
But then
3a sin θ · cos2 θ 3a sin2 θ · cos θ
x = r cos θ = , y = r sin θ =
cos3 θ + sin3 θ cos3 θ + sin3 θ
give a parametric description of the folium in terms of parameter θ (that you can
rename to be t, say). Equations of many classical curves can be given directly
in the polar coordinates. These include the lemniscate r2 = a2 cos 2θ, cardioid
a
r = a(1 + cos θ) and parabola r = , where in all the cases a is a positive
1 − cos θ
parameter. You are encouraged to sketch the graphs of the curves and also to check
that the parabola is indeed a parabola.
Complex numbers lose the order structure of reals: we cannot say that one com-
plex number is larger (or smaller) than another. But they win at solving polynomial
equations! All polynomials with complex coefficients can be entirely solved over
complex numbers; a polynomial of degree n has exactly n roots when counted with
multiplicity.
Remark 13.2. One useful way of thinking of complex numbers z = x + iy is through
identifying them with the real two-dimensional vector space of 2 × 2 matrices
 
x y
.
−y x
You may (attempt to) show that this define a bijection (one-to-one correspondence)
preserving arithmetic operations between the field of complex numbers C and the
set of matrices   
x y
: x, y ∈ R .
−y x
Quite remarkably all standard manipulations with real numbers transform to
complex ones; even the power series
∞
X
an z n
n=0

become well defined for z = x + iy thought to be complex (and their coefficients an

can be complex numbers as well!). The formulae for the radius of convergence of
such power series,
|an | 1
R = lim or R = lim p
n→∞ |an+1 | n→∞ n
|an |
(when the limits exist!), remain the same; but now they characterise disks |z| < R
on the complex plane rather than intervals (−R, R) on the real line. (At least, we
see that there is a good reason to call R the radius of convergence.) Of course, when
 Notes for Calculus A (2024/25) 89

R = ∞, and this is for example the case of the power series

∞
X zn
,
n=0
n!
then the convergence happens for all z ∈ C. We can now officially introduce the
exponential function of complex variable z as the series above:
∞
z
X zn
e = exp(z) = ,
n=0
n!
as well as trigonometric and hyperbolic functions:
∞ ∞
X (−1)n z 2n+1 X (−1)n z 2n
sin z = , cos z = ,
n=0
(2n + 1)! n=0
(2n)!
∞ ∞
X z 2n+1 X z 2n
sinh z = , cosh z = .
n=0
(2n + 1)! n=0
(2n)!
All these series converge at all z ∈ C (they define analytic functions of complex
variable z — something that will be clarified and discussed at a greater depth in
your Complex Analysis course). Then it comes as no suprise that
ez = cosh z + sinh z and eiz = cos z + i sin z,
the latter formula is due to Euler. It also implies that
eiz + e−iz eiz − e−iz
cos z = and sin z = ;
2 2i
in other words, these trigonometric functions are intimately related to the exponen-
tial function when viewed as complex-variable functions.
De Moivre’s formula
(cos y + i sin y)m = cos my + i sin my for m = 0, 1, 2, . . .
iy
is a precursor to Euler’s formula e = cos y + i sin y. The former follows from the
latter by noticing that (eiy )m = eimy by the properties of the exponential function.
Multiplying two complex numbers z1 = r1 eiθ1 and z2 = r2 eiθ2 in their polar form is
particularly nice: z1 z2 = r1 r2 ei(θ1 +θ2 ) . This reflects the properties of the exponential
function: ez1 ez2 = ez1 +z2 .
Exercise 13.1. Compute
√
(a) (−i)79 ,
(b) (1 + i)100 , (c) (1 − i)999 , (d) (1 + i 3)1111 ,
√ √ √
(e) (1 + 5 − i 2(5 − 5))1000 .
√ √ √ √
Hints. (d) 1 + i 3 = −2e4πi/3 ; (e) 1 + 5 − i 2(5 − 5) = −4e4πi/5 . □
90 WADIM ZUDILIN

Linear differential equations

We now return to the theme we ended with at Lecture 10, namely, to the class of
linear differential equations that is particularly well structured — to homogeneous
linear differential equations. Why do we return to it with such a delay? Because
we now have additional tools to treat it: complex numbers but also the connection
between the exponential of complex variable and trigonometric functions.
We recall a general version of the linear equation of order n:
y (n) + fn−1 (x)y (n−1) + · · · + f1 (x)y ′ + f0 (x)y = 0,
where f0 (x), f1 (x), . . . , fn−1 (x) are given functions of x. Then there are (generically)
n solutions y1 (x), . . . , yn (x) of the equation, forming a basis of solutions, which are
linearly independent over R, such any solution y(x) belongs to a linear space spanned
by them:
y(x) = c1 y1 (x) + · · · + cn yn (x), where c1 , . . . , cn ∈ R.
Thus, to solve our equation means to construct a basis of its solutions. Though this
problem cannot be solved in general, as solutions depend critically on the coefficients
fn−1 (x), . . . , f0 (x) of the differential equation, one special situation stands out, when
these coefficients do not depend on x. Examples of the linear equations y ′′ + y = 0
and y ′′ − 2y = 0 of second order we discussed in Lecture 10, but we can really do
things in a much-much greater generality.
From now on, we will assume that coefficients fk (x) in the equation are constant,
and so we will write them as fk (x) = ak ∈ R (we will see though that we can also
deal with ak complex):
y (n) + an−1 y (n−1) + · · · + a1 y ′ + a0 y = 0.
To such an equation we assign its characteristic equation
tn + a1 tn−1 + · · · + an−1 t + an = 0,
and we look for its complex roots t1 , t2 , . . . , tr with multiplicities m1 , m2 , . . . , mr
respectively. The latter simply means that we factorise the polynomial into linear
factors:
tn + a1 tn−1 + · · · + an−1 t + an = (t − t1 )m1 (t − t2 )m2 · · · (t − tr )mr , (13.1)
which is always possible over the complex numbers. Notice that the sum of all
multiplicities m1 + · · · + mr = n, the order of differential equation.
Now with each root t = tj of multiplicity m = mj we can associate m independent
solutions of the linear differential equation, namely,
etx , xetx , x2 etx , . . . , xm−1 etx .
The proof of this fact is simply based on verification that each of this functions
satisfy our differential equation. When we combine all such solutions for different
roots t = tj of the characteristic equation, we get precisely m1 + · · · + mr = n
independent solutions y1 (x), y2 (x), . . . , yn (x) of the equation.
Well, it is clear how things work for a real root t. But what happens when
t = a + bi with b ̸= 0? Because the characteristic polynomial has all coefficients real,
 Notes for Calculus A (2024/25) 91

the factorisation in (13.1) can only take place when every complex root t = a + bi
appears together with its conjugate t = a − bi as another root as well, with the same
multiplicity. This means that in our collection of independent solutions we will have
both
xk e(a+bi)x = xk eax (cos bx + i sin bx) and xk e(a−bi)x = xk eax (cos bx − i sin bx).
This complex-looking elements of the basis can be then replaced with much simpler
(and absolutely real!) functions
xk eax cos bx and xk eax sin bx.
Indeed, from our linear algebra course we know that changing two vectors of a basis
v 1 , v 2 to λ1 (v 1 + v 2 ), λ2 (v 1 − v 2 ) (where both λ1 and λ2 are nonzero!) will give
another basis of the same vector space.
Let us do some examples.
Exercise 13.2. Solve linear differential equations in y = y(x):
(a) y ′′ − 2y = 0, (b) y ′′ + y ′ − 2y = 0, (c) y ′′ − 4y ′ + 4y = 0,
(d) y ′′′ + 6y ′′ + 9y ′ = 0, (e) y ′′′ − 3y ′′ + 3y ′ − y = 0, (f) y ′′′′ − 2y ′′ + y = 0.
Solution. (d) The characteristic equation reads t3 + 6t2 + 9t = 0, and through an
inspection t3 +6t2 +9t = t(t+3)2 . This means that we have root t1 = 0 of multiplicity
1 and root t2 = −3 of multiplicity 2; this leads to the basis
y1 (x) = e0x = 1, y2 (x) = e−3x , y3 (x) = xe−3x
of solutions, so that the general solution assumes the form y(x) = c1 + c2 e−3x +
c3 xe−3x , where c1 , c2 , c3 run over real numbers. □
Exercise 13.3. Solve linear differential equations in y = y(x):
(a) y ′′ + 2y = 0, (b) y ′′ − 4y ′ + 5y = 0, (c) 4y ′′ + 4y ′ + y = 0,
(d) y ′′′′ − y = 0, (e) y ′′′′ + y = 0, (f) y (5) − 2y (4) − 16y ′ + 32y = 0.
Solution. (f) The characteristic equation is
t5 − 2t4 − 16t + 32 = 0.
Since
t5 − 2t4 − 16t + 32 = (t − 2)(t4 − 16) = (t − 2)2 (t + 2)(t2 + 4),
the roots are 2, −2, 2i, −2i of multiplicity 2, 1, 1, 1 respectively. The above rules
allow us to write the general solution:
y(x) = (c1 + c2 x)e2x + c3 e−2x + c4 cos 2x + c5 sin 2x,
where c1 , . . . , c5 ∈ R. □
Another class of linear differential equations, whose coefficients are not constant,
is represented by
d2 y dy
ax2 2 + bx + cy = 0
dx dx
92 WADIM ZUDILIN

(at least in the second order case). Such an equation is called an Euler equation or
an equidimensional equation. Particular independent solutions of the equation are
searched for in the form y = xr , with r to be determined. Assuming for simplicity
that x > 0 and using
dy d2 y
y = xr , = rxr−1 , = r(r − 1)xr−2 ,
dx dx2
we find out after the substitution in the original equation:
ar(r − 1) + br + c xr = 0.

Thus, the solutions y = xr or, more generally, y = |x|r will correspond to the root of
quadratic polynomial ar(r − 1) + br + c. Such differential equations are not assessed
in your final exam, but you may still consider solving some exercises on them (from
the Book).