Y. D.

Chong (2021) MH2801: Complex Methods for the Sciences

5. Complex Oscillations
In physics and the other quantitative sciences, complex numbers are widely used for
analyzing oscillations and waves. We begin our study of this topic with an elementary
model called the damped harmonic oscillator.

5.1 The damped harmonic oscillator

A particle of mass m moves along one dimension, with x(t) denoting its displacement at
time t. It is subject to two forces: a spring force and a damping force. The spring constant is
k = mω02 , and the damping coefficient is 2mγ. The parameters m, γ, and ω0 are all positive
real numbers.
Spring force

Damping force

According to Newton’s second law,

d2 x dx
m 2
= F (x, t) = −2mγ − mω02 x(t). (5.1)
dt dt
Dividing by the common factor of m, and bringing everything to one side, gives
d2 x dx
2
+ 2γ + ω02 x(t) = 0. (5.2)
dt dt
This is called the damped harmonic oscillator equation. It is a second-order ordinary
differential equation (ODE), so its general solution must contain two free parameters. These
parameters are usually (but not necessarily) specified by the initial displacement x(0) and
initial velocity ẋ(0).

Note—Sometimes, we write the damped harmonic oscillator equation as:

 2 
d d 2
+ 2γ + ω0 x(t) = 0. (5.3)
dt2 dt

The quantity in square brackets is a linear differential operator acting on x(t). The
three terms in the operator correspond to the three ingredients of the damped harmonic
oscillator model: (i) a second derivative term stemming from Newton’s second law, (ii)
a first derivative term representing damping, and (iii) a constant term representing the
spring force.
Writing the differential equation this way emphasizes its linearity, a property that
is important for finding the solutions, as discussed below.

5.1.1 Simple harmonic oscillator limit

For γ = 0 (zero damping), the system reduces to the simple harmonic oscillator. From
previous physics courses, we know the general solution is
x(t) = A cos(ω0 t + ϕ), (5.4)

31
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

where A and ϕ are free parameters. This is a sinusoidal oscillation with amplitude A, phase
ϕ, and frequency ω0 .
The parameter ω0 comes from the spring constant k = mω02 . (In fact, the spring constant
was parameterized this way so that the solution ends up with this nice form.) We call ω0
the natural frequency, meaning the frequency of the oscillator in the absence of damping
or other disturbances.

Note—Some authors call ω0 the “angular frequency”, reserving the term “frequency”
for the quantity f0 = ω0 /2π. But since we will always deal with ω0 rather than f0 , we
will refer to ω0 as simply “frequency”.

5.1.2 Damped oscillations

For γ > 0, there is now a damping force opposing the motion of the oscillator. What
form will the solutions take? Before launching into the mathematics, let’s use our physical
intuition to make a guess.
The damping force does work against the particle (since its sign is always opposite to
the particle’s velocity). If the damping force is very weak, the solution should not be too
different from the simple harmonic oscillator solution—the particle should oscillate around
the equilibrium point x = 0 with a frequency of around ω0 . But the damping force will cause
it to lose a bit of energy every oscillatory cycle, resulting in an oscillation whose amplitude
diminishes slowly over time. In the t → ∞ limit, all the energy is lost, and x (as well as ẋ)
should go asymptotically to zero.
So we would guess something like this:

Let us see if the mathematical analysis agrees with this guess.

5.2 The complex damped harmonic oscillator equation

The variable x(t) is the displacement of the particle, so it ought to be real. However, a
good way to solve the damped harmonic oscillator equation is to generalize x(t) to complex
values. In other words, we convert the harmonic oscillator equation into a complex ODE:

d2 z dz
2
+ 2γ + ω02 z(t) = 0, z(t) ∈ C. (5.5)
dt dt
The parameter-counting rule for real ODEs (see Section 1.3) generalizes to complex ODEs,
except that the free parameters should be complex numbers. In this case, the complex
damped harmonic oscillator equation is a second-order ODE, so its general solution must
have two complex free parameters.
If we can find a solution z(t) for the complex damped harmonic oscillator equation, then
its real part x(t) = Re[z(t)] would be a solution to the real damped harmonic oscillator

32
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

equation, since

d2 x dx 2 d2 Re[z] dRe[z]
+ 2γ + ω0 x(t) = + 2γ + ω02 Re[z(t)] (5.6)
dt2 dt dt2
 2 dt 
d z dz 2
= Re + 2γ + ω 0 z(t) (5.7)
dt2 dt
= 0. (5.8)

Here, we have used the fact that the Re[· · · ] operation can be freely shuffled in or out of
derivatives and sums with real coefficients (see Section 3.1).

5.2.1 Complex ansatz

We now aim to derive the general solution for the complex damped harmonic oscillator
equation.
First, note that the equation is linear. This means that for any two solutions z1 (t) and
z2 (t), a linear superposition

z(t) = a1 z1 (t) + a2 z2 (t), where a1 , a2 ∈ C (5.9)

is also a solution. This can be verified by direct substitution into the ODE.
Due to linearity, a good strategy for finding the general solution is to identify two different
specific solutions, z1 (t) and z2 (t). Then we can construct the above linear superposition, with
the two complex coefficients a1 and a2 serving as free parameters. Any solution containing
two free parameters is automatically the general solution.
So now we have to find some specific solutions. Let us make a guess, or ansatz:

z(t) = e−iωt . (5.10)

Here, ω is a constant to be determined. The first and second derivatives are:

dz
= −iω e−iωt , (5.11)
dt
d2 z
= −ω 2 e−iωt . (5.12)
dt2
Substituting these into the damped harmonic oscillator equation gives

−ω 2 − 2iγω + ω02 e−iωt = 0.

(5.13)

This equation holds for all t if and only if the complex second-order polynomial on the
left-hand side is zero:

−ω 2 − 2iγω + ω02 = 0. (5.14)

The solutions to the polynomial can be obtained from the quadratic formula:
q
ω = −iγ ± ω02 − γ 2 . (5.15)

Hence, we have found the specific solutions

q
z± (t) = exp (−iω± t) , where ω± = −iγ ± ω02 − γ 2 . (5.16)

Either ω+ or ω− gives a valid specific solution to the damped harmonic oscillator equation.
Note that the solution is specific and contains no free parameters, since ω± is entirely deter-
mined by γ and ω0 (which are fixed parameters appearing in the ODE, not free parameters).

33
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

5.2.2 Complex frequencies

The specific solutions we found have the form e−iωt , where ω is a complex “frequency”.
There are two choices of complex frequency,
q
ω± = −iγ ± ω02 − γ 2 . (5.17)

The values depend on the oscillator parameters γ and ω0 . The plot below shows how ω±
move in the complex plane as γ and ω0 are varied:

Note the following features:

1. For γ = 0, the frequencies are both real, with values ±ω0 .

2. If we increase γ from zero with ω0 fixed, both ω+ and ω− move downwards in the
complex plane, along a circular arc.
3. At γ = ω0 , the frequencies meet along the imaginary axis.
4. For γ > ω0 , the two frequencies move apart along the imaginary axis.

To understand the implications of these complex frequencies, let uswrite the real and
imaginary parts of ω as ωR + iωI . Then

z(t) = e−iωt = eωI t e−iωR t . (5.18)

If both ωR and ωI are non-zero, this describes a spiral trajectory in the complex plane (see
Section 3.6) whose magnitude either increases or decreases with time, depending on the sign
of ωI . To see this explicitly, we can write

z(t) = eωI t e−iωR t . (5.19)

Taking the real part,

Re [z(t)] = eωI t cos [ωR t] . (5.20)

Assuming that γ > 0, the complex frequencies given by Eq. (5.17) always have ωI < 0.
Referring to Eq. (5.20), this means the solutions are always damped. Moreover, if ωR ̸= 0,
then z(t) executes a clockwise (for ω+ ) or counterclockwise (for ω− ) inward spiral; in either
case, Re[z(t)] describes an oscillation with diminishing amplitude, consistent with our guess
from Section 5.1.2. On the other hand, if ωR = 0, then the solution is a pure exponential de-
cay with no oscillation. In the next few sections, we will undertake a systematic examination
of these two distinct behaviors.

34
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

5.3 General solution for the damped harmonic oscillator

For now, let us suppose that ω0 ̸= γ. Then we have two distinct specific solutions,
q
z± (t) = e−iω± t , where ω± = −iγ ± ω02 − γ 2 . (5.21)

By taking a linear superposition of these specific solutions, we obtain the general solution
for the complex damped harmonic oscillator equation:

z(t) = a+ e−iω+ t + a− e−iω− t , (5.22)

where a+ and a− are independent complex free parameters.

(Note that if ω0 = γ, then z+ (t) = z− (t), so a+ and a− would be coefficients multiplying
the same function, and we would not be allowed to treat them as two independent free
parameters. We will discuss how to handle this case in Section 5.3.3.)
To find solutions to the real damped harmonic oscillator equation, we take x(t) =
Re[z(t)]. The resulting expression will depend on whether ω0 > γ or ω0 < γ. These
two cases lead to under-damped solutions and over-damped solutions, respectively,
as discussed in the next two sections.

5.3.1 Under-damped motion

First, consider ω0 > γ. Let us define
q
Ω= ω02 − γ 2 ∈ R, (5.23)

so that ω± = −iγ ± Ω. Plugging this into the complex general solution gives

z(t) = a+ e−γt e−iΩt + a− e−γt eiΩt (5.24)

= e−γt a+ e−iΩt + a− eiΩt .
 
(5.25)

We can use Euler’s formula to simplify the terms in the brackets:

a+ e−iΩt + a− eiΩt = a+ cos(Ωt) − i sin(Ωt) + a− cos(Ωt) + i sin(Ωt)

   
(5.26)



= a+ + a− cos(Ωt) − i a+ − a− sin(Ωt). (5.27)

Hence,

x(t) = Re [z(t)] (5.28)

(
A = Re [a+ + a− ]
= e−γt A cos (Ωt) + B sin (Ωt) , where
 
(5.29)
B = Im [a+ − a− ] .

This is called an under-damped solution. The coefficients A and B are two independent
real parameters, so this serves as a general solution for the real damped harmonic oscillator
equation. Using the trigonometric formulas, the solution can be equivalently written as
( p
−γt

C = A2 + B 2 ,
x(t) = C e cos Ωt + Φ , where (5.30)
Φ = − tan−1 [B/A] .

This shows explicitly that it consists of a sinusoidal oscillation (with frequency Ω) overlaid
on an “envelope” given by the exponentially decreasing function exp(−γt). The graph of
x(t) versus t is plotted below:

35
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

5.3.2 Over-damped motion

p
For ω0 < γ, the square root term in ω± = −iγ ± ω02 − γ 2 is imaginary. Let us define
q
Γ = γ 2 − ω02 ⇒ ω± = i (−γ ± Γ) . (5.31)

Then the complex general solution can be simplified to

z(t) = a+ e−(γ−Γ) t + a− e−(γ+Γ) t , (5.32)

and the real solution is

x(t) = Re [z(t)] (5.33)

−(γ−Γ) t −(γ+Γ) t
= C+ e + C− e , where C± = Re[a± ]. (5.34)

This is called an over-damped solution. It consists of two exponentially decaying terms,

with decay rates (γ − Γ) and (γ + Γ) respectively. Since Γ < γ, both decay rates are positive
real numbers, but note that (γ − Γ) decreases with γ, whereas (γ + Γ) increases with γ, as
shown below:

Since γ + Γ is associated with a faster-decaying exponential, for large t the second term
becomes negligible compared to the first term, and the solution has the limiting form

x(t) ≈ C+ e−(γ−Γ)t (for large t). (5.35)

A plot of x(t) versus t is shown below, with the limiting form plotted as dashes.

36
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

The over-damped solution has an interesting feature: for stronger damping, the decay
rate at long times is slower. This is the opposite of the under-damped oscillator’s behavior!
Mathematically, it happens because the decay rate (γ − Γ), appearing in the limiting form
of x(t) for large t, is a decreasing function of γ.
In the over-damped regime, the motion of the oscillator is dominated by the damping
force rather than the spring force. Therefore, as the oscillator tries to return to its equilib-
rium position x = 0, the damping acts against this motion, and the stronger the damping,
the slower the decay to equilibrium. By contrast, in the under-damped regime, the spring
force is dominant, so stronger damping leads to faster decay via faster dissipation of the
oscillator’s kinetic energy.

5.3.3 Critical damping

Critical damping occurs when ω0 = γ. Under this special condition, the solution given in
Eq. (5.22) reduces to
z(t) = (a+ + a− ) e−γt . (5.36)
This has only one independent complex parameter, i.e. the parameter (a+ +a− ). Therefore,
it cannot be a general solution for the complex damped harmonic oscillator equation.
We will not go into detail here regarding the procedure for finding the general solution for
the critically-damped oscillator, leaving it as an exercise for the interested reader. Basically,
it involves Taylor expanding the solution near the critical point, and then showing that there
is a solution of the form
z(t) = (A + Bt) e−γt , (5.37)
which contains the desired two independent parameters.
The critically-damped solution contains an exponential decay constant of γ, which is
the same as the decay constant for the envelope function in the under-damped regime (Sec-
tion 5.3.1), and larger than the (long-time) decay constants in the over-damped regime
(Section 5.3.2). Hence, we can regard the critically-damped solution as the fastest-decaying
non-oscillatory solution.
This feature of critical damping is employed in many engineering contexts, the most
familiar being automatic door closers. If the damping is too weak or the spring force is too
strong (under-damped), the door slams shut, whereas if the damping is too strong or the
spring force is too weak (under-damping), the door takes unnecessarily long to swing shut.
For best performance, an automatic door closer should be tuned to a “sweet spot” that
corresponds to the critical point of a damped harmonic oscillator.

5.4 Stating the solution in terms of initial conditions

The general solution for the complex damped harmonic oscillator equation contains two
undetermined parameters which are the complex amplitudes of the “clockwise” and “coun-
terclockwise” complex oscillations:
q
z(t) = a+ e−iω+ t + a− e−iω− t , where ω± = −iγ ± ω02 − γ 2 . (5.38)

However, mechanics problems are often expressed in terms of an initial value problem,
specifying the state of the system at some initial time t = 0. In other words, given z(0) ≡ x0
and ż(0) ≡ v0 , what is z(t) in terms of x0 and v0 ?
We can solve the initial-value problem by finding z(0) and ż(0) in terms of the above
general solution for z(t):
z(0) = a+ + a− = x0 (5.39)
ż(0) = −iω+ a+ − iω− a− = v0 . (5.40)

37
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

These two equations can be combined into a 2 × 2 matrix equation:

    
1 1 a+ x
= 0 . (5.41)
−iω+ −iω− a− v0

So long as ω+ ̸= ω− , the matrix is non-singular, and we can invert it to obtain a± :

   
a+ 1 −iω− x0 − v0
= . (5.42)
a− i(ω+ − ω− ) iω+ x0 + v0

We can plug these coefficients back into the general solution. After some algebra, the result
simplifies to
 
γx0 + v0
q
z(t) = e−γt x0 cos(Ωt) + sin(Ωt) , where Ω ≡ ω02 − γ 2 . (5.43)
Ω

For the under-damped case, Ω is real, and this solution is consistent with the one we derived
om Section 5.3.1, except that it is now explicitly expressed in terms our initial conditions
x0 and v0 . As for the over-damped case, we can perform the replacement
q
Ω → iΓ = i γ 2 − ω02 . (5.44)

Then, using the relationships between trigonometric and hyperbolic functions from Sec-
tion 3.5.3, the solution can be re-written as
 
−γt γx0 + v0
z(t) = e x0 cosh(Γt) + i sinh(Γt) (5.45)
iΓ
   
x0 γx0 + v0 x0 γx0 + v0
= + e−(γ−Γ)t + − e−(γ+Γ)t , (5.46)
2 2Γ 2 2Γ

which is consistent with the result found in Section 5.3.2.

In either case, so long as we plug in real values for x0 and v0 , the solution is guaranteed
to be real for all t. That’s to be expected, since the real solution is also one of the specific
solutions for the complex harmonic oscillator equation.

5.5 Exercises
1. In Section 5.2.2, we encountered the complex frequencies
q
ω± = −iγ ± ω02 − γ 2 . (5.47)

For fixed ω0 and ω0 > γ (under-damping), prove that ω± lie along a circular arc in
the complex plane.
2. Derive the general solution for the critically damped harmonic oscillator by following
these steps:
(a) Consider the complex ODE, in the under-damped regime ω0 > γ. We saw in
Section 5.3 that the general solution has the form
 q    q  
z(t) = ψ+ exp −γ − i ω02 − γ 2 t + ψ− exp −γ + i ω02 − γ 2 t (5.48)

for some complex parameters ψ+ and ψ− . Define the positive parameter ε =

p
ω02 − γ 2 . Re-write z(t) in terms of γ and ε (i.e., eliminating ω0 ).

38
Y. D. Chong (2021) MH2801: Complex Methods for the Sciences

(b) The expression for z(t) is presently parameterized by the independent parameters
ψ+ , ψ− , ε, and γ. We are free to re-define the parameters, by taking

α = ψ+ + ψ− (5.49)
β = −iε(ψ+ − ψ− ). (5.50)

Using these equations, express z(t) using a new set of independent complex pa-
rameters, one of which is ε. Explicitly identify the other independent parameters,
and state whether they are real or complex.
(c) Expand the exponentials in z(t) in terms of the parameter ε. Then show that in
the limit ε → 0, z(t) reduces to the critically-damped general solution (5.37).
3. Repeat the above derivation for the critically-damped solution, but starting from the
over-damped regime γ > ω0 .

4. Let z(t) be a complex function of a real input t, which obeys the differential equation

dz
= −i (ω1 − iγ) z(t), (5.51)
dt
where ω1 and γ are real. Find the general solution for z(t), and hence show that z(t)
satisfies the damped oscillator equation
 2 
d d 2
+ 2γ + ω 0 z(t) = 0 (5.52)
dt2 dt

for some ω02 . Finally, show that this harmonic oscillator is always under-damped.
[solution available]

39