4 1 Introduction

1.1 Introduction to Scanning Tunneling Microscopy

Today the scanning probe microscope is a very important tool in nanoscience. The
principle of scanning probe microscopes is to move a sharp tip close to a surface in
order to measure various properties with a spatial resolution on the nanometer or even
atomic scale. The first kind of scanning probe microscope, the scanning tunneling
microscope, (STM) was invented in 1981/1982 by Binnig and Rohrer who received
the Nobel prize in physics 1986 for this invention. The most striking property of this
kind of microscope is that it provides resolution down to the atomic scale in real
space (Fig. 1.3b).
Here is an analogy which shows the precision of an STM working with atomic
resolution. Such instruments are about 10 cm in size and can image with a resolution
of about 1 Å, corresponding to a precision of about 10−9 of its size. Scaling this
precision of 10−9 up to macrosize dimensions would correspond to using a pencil
1,000 km in length to write letters from Cologne (Germany) in a notebook in Rome
(Italy) with 1 mm resolution!
A schematic of an STM, with fine metal tip used as a probe, is shown in Fig. 1.3a. A
voltage is applied between the tip and the (conducting) sample. The tip is approached
toward the sample surface until a current flows. A current (the tunneling current) can
be detected shortly before tip and sample come into direct contact. This happens
at distances between tip and sample of the order of 0.5–1 nm. The tunneling cur-
rent increases monotonously with decreasing tip-sample distance. Thus a certain
measured tunneling current corresponds to a specific tip-sample distance. Since the
tunneling current varies strongly (exponentially) with the tip-sample distance this

Feedback
(a) (b)
x, y, z
Scanning unit
I
z
x
y
V

Fig. 1.3 a Schematic of a scanning tunneling microscope (STM). b STM image of the Si(111)
surface. Individual atoms are observed as yellow dots. The rhombic unit cell is indicated by white
lines. Besides the periodic arrangement of the atoms also defects such as single missing atoms can
be observed
1.1 Introduction to Scanning Tunneling Microscopy 5

quantity can be used to measure (and control) the tip-sample distance very precisely.
We will see later that a 20 % change in the tunneling current corresponds to a change
in the tip-sample distance of only 0.1 Å. The tip is positioned with such high accu-
racy using piezoelectric actuator elements. The mechanical extension of this actuator
elements is proportional to the voltage applied to their electrodes. In this way, the tip
can be moved in x, y and z directions with sub-ångström resolution.
While the tip is scanned along the surface in x and y directions, a feedback
mechanism constantly adjusts the tip height by approaching or retracting the tip to
a tip-sample distance at which the tunneling current remains constant. If there is an
atomic step at the surface, as shown in Fig. 1.3a, and the tip approaches this step
edge laterally during scanning, the tunneling current will rise due to the smaller
distance between tip and sample. As a reaction to this the feedback circuit will
retract the tip in order to maintain a constant tunneling current, i.e. a constant tip-
sample distance. Recording the feedback signal (tip height) as a function of the lateral
position results in a map (or image) of the tip height, which often corresponds to the
surface topography of the sample surface.
The interpretation of the tip height for constant tunneling current as the topography
of the surface is a first approximation. So-called electronic effects can change this
interpretation. A simplified example of this are atoms on a surface which have the
same height (of their nuclei) but their electronic properties are different in the sense
that one atom has a “higher electrical conductivity” than the other. The atom with the
“higher conductivity” will appear higher (same tunneling current at larger tip-sample
distances) while for the case of the “less conducting atom” the tip has to approach
closer to maintain the same tunneling current.
Figure 1.3b shows an atomically resolved image of a Si(111) surface. Single sili-
con atoms are observed as yellow dots. The operation of an STM can be visualized
experimentally by combining a scanning electron microscope (SEM) with an STM.
The SEM can be used to image the motion of the STM tip during scanning. A
movie of a scanning STM imaged during operation with an SEM can be accessed at
http://www.fz-juelich.de/pgi/pgi-3/microscope.
The tunneling junction (sample-gap-tip) can be treated in different approxima-
tions. Here in the introduction, we consider a simple one-dimensional approximation
for one electron tunneling in order to grasp the very important exponential depen-
dence of the tunneling current on the tip-sample distance. Later we will look more
deeply into the theory of STM.
In quantum mechanics, electrons in a solid are described by a wave function ψ(r).
In the free electron approximation the wave function of an electron of energy E is
an oscillating function. The one-dimensional Schrödinger equation is solved by the
(not normalized) wave function
!
±ikz 2m e E
ψ(z) ∝ e , k= . (1.1)
!2

When drawing such a wave function, it should be always remembered that the quan-
tum mechanical wave function is genuinely a complex function, which is difficult to
6 1 Introduction

Solid Vacuum Solid Vacuum Solid

(a) (b)
Classically forbidden Classically forbidden
Evac region Evac region
E
Eparticle Eparticle

0
0 z 0 z
Incoming particle Incoming particle d
wave function wave function
Particle wave function Particle wave function
in vacuum past the barrier

Fig. 1.4 a The top graph shows the potential diagram with a barrier of height Φ and the energy
of an electron E particle = E F . The lower graph shows the real part of the electron wave function
with an exponential decay of the wave function in the vacuum region. b The top graph shows the
potential for a solid-vacuum-solid configuration. The lower graph shows the electron wave function
oscillating in front of the barrier, exponentially decaying inside the barrier and again oscillating
past the barrier

draw. Therefore, usually only the real or imaginary part is drawn, as in Fig. 1.4. The
sinusoidal appearance of the real or imaginary part of the wave function should not
make us forget that the absolute value |ψ(z)|2 of such a wave function eikz has the
constant value of one for all z.
In the following, we consider the electrons in a solid with the highest energy
(at the Fermi level E F ) and call this energy the particle energy E = E particle . The
energy of these electrons at the Fermi level is lower than the energy of free electrons
(the vacuum energy). This energy difference is roughly the bonding energy of the
electrons inside the solid. If the Fermi energy were larger than the vacuum energy,
the electrons would leak out of the solid toward the vacuum. The minimum energy
needed to remove an electron from a solid is called the work function Φ, which is
shown graphically in Fig. 1.4a.
Thus at a surface there is a barrier (work function) preventing the electrons from
leaving the solid to the vacuum level E vac . In classical mechanics, particles cannot
penetrate into a barrier which is higher than their energy. In quantum mechanics,
particles can penetrate into a region with a barrier higher than their energy. An ansatz
with an exponentially decaying wave function inside the barrier (in the vacuum) as
ψ(z) = ψ(0)e−κz leads to a solution of the Schrödinger equation inside this potential
barrier (Fig. 1.4a). The probability of a particle being at a position z inside the barrier
is approximately proportional to |ψ(z)|2
!
2m e Φ
|ψ(z)|2 = |ψ(0)|2 e−2κz , κ = . (1.2)
!2
1.1 Introduction to Scanning Tunneling Microscopy 7

If after some distance d the vacuum is replaced by another solid this configuration is
already a one-dimensional model of the tunneling junction (electrode-gap-electrode).
A potential diagram for such a tunneling barrier is shown in Fig. 1.4b. Since inside
the solid the vacuum barrier is not present, the solution for the wave function is an
oscillating wave, which is again a solution inside the second solid. This means that in
quantum mechanics the electron has a finite probability in both metals. In the square
barrier model a barrier, of height Φ = E vac − E F and width d is considered. In the
course of the solution of the square barrier problem, the transmission coefficient for
the wave function behind the barrier can be calculated. (This is usually done in the
quantum mechanics course. We will come to this in a later chapter.) The probability
of an electron being observed on the right side of the barrier is proportional to the
absolute square of the wave function at the end of the barrier |ψ(d)|2 . A transmission
coefficient T can be defined as

|ψ(d)|2
T = ≈ e−2κd . (1.3)
|ψ(0)|2

The main characteristics are: the transmission coefficient decays exponentially with
the tip-sample distance d and decreases exponentially with the square root of the
work function. If we use the right electrode as the tip, the tip probes the probability
density of the electron states at distance d from the surface. Later we will see that
the tunneling current is proportional to the transmission coefficient.
Evaluating (1.2) using the free electron mass for m e and a typical value for the
work function of a metal (Φ ≈ 4.5 eV), 2κ is about 20 nm−1 . Thus a variation of
the barrier thickness of 0.1 nm results in a difference in the transmission factor of
an order of magnitude (∼7.4). Hence the tunneling current increases by about an
order of magnitude if the tip approaches by one Å to the sample. This sensitivity
in the tip-sample distance is the reason for the extremely high vertical resolution of
the STM which can reach the picometer regime. Atoms on the tip which protrude
only 2.5 Å (∼one atomic distance) less toward the sample carry only a factor of 150
less current. This means that the majority of the tunneling current is carried by the
“last atom”, which also explains the very high (ultimately atomic) lateral resolution
of the STM.

1.2 Introduction to Atomic Force Microscopy

One disadvantage of STM is that it can be used only for conducting samples since
the tunneling current is the measured quantity. An atomic force microscope can also
be used on insulating samples. The atomic force microscope (AFM) is alternatively
known as the scanning force microscope (SFM). However, here we will use the more
common name atomic force microscope. Instead of the tunneling current, which is
the measured quantity in STM, in an atomic force microscope force microscopy the
force between the tip and sample is measured. In Fig. 1.5, a qualitative sketch of the
force between tip and sample is given. Three different regimes can be distinguished.
8 1 Introduction

Fig. 1.5 Qualitative

behavior of the force
between tip and sample as
Repulsive force
function of tip-sample
distance

Force
Tip far from
the surface:
No force
0

1 2 Attractive force

Tip-sample distance

(a) If the tip is far away from the surface the force between tip and sample is
negligible. (b) For closer distances an attractive (negative) force between tip and
sample occurs. (c) For very small distances a strong repulsive force between tip and
sample occurs. One problem with this behavior is that the tip-sample force which is
used as measured signal depends non-monotonously on the tip-sample distance, i.e.
for one value of the measured force in the attractive regime there are two tip-sample
distances, point 1 and point 2 on the force distance curve in Fig. 1.5. Care has to
be taken to work only on one of the branches left or right of the minimum in the
force-distance curve on which a monotonous force distance relation holds.
The force between tip and sample can be measured in a static mode using the
deflection of the cantilever on which a tip is mounted. The cantilever acts as a
spring and its deflection is proportional to the tip-sample force. If the stiffness of
the cantilever spring k (spring constant) is known, the force between tip and sample
can be determined by measuring the bending of the cantilever. Hooke’s law gives
F = −kz, where F is the force and z is the distance the cantilever spring is bent
relative to its equilibrium position without the sample present. Figure 1.6 shows a
typical silicon cantilever used as a force sensor in atomic force microscopy with a
sharp tip (probe) at its end. The deflection of the lever is measured for instance using
a laser beam reflected from the back of the cantilever into a split photodiode as shown
in Fig. 1.7.
In the static mode of operation, the surface contour is mapped while scanning by
changing the z-position of the tip in such a way that the tip-sample force and, corre-
spondingly, the tip-sample distance are kept constant. The tip position maintaining
a constant tip-sample distance is recorded as topography signal. In other words: the
feedback loop maintains a constant force between the tip and the sample i.e. con-
stant bending of the cantilever, as shown in Fig. 1.7. The corresponding changes in
the z-position required to maintain a constant tip-sample distance (i.e. constant force)
correspond to the topography of the sample. If the measurements are performed in
1.2 Introduction to Atomic Force Microscopy 9

Fig. 1.6 SEM image of a

silicon cantilever used in
atomic force microscopy
with a length of 450 µm

Fig. 1.7 Schematics of Split photodiode

atomic force microscopy Feedback
operation electronics
Laser

z-signal
Cantilever
Sample
z

x,y scanner

the repulsive regime of the force-distance curve the operating mode is called contact
mode. The last atoms of the tip are in direct contact with the surface atoms.
The atomic force microscope can also be operated in the so called dynamic mode
with an oscillating cantilever. This dynamic mode is often operated in the attractive
part of the tip-sample interaction. This mode of operation is called the non-contact
mode. This is important when imaging soft samples (for instance polymers or bio-
logical samples), which would be destroyed by a strong tip sample interaction. In
the dynamic mode, the cantilever is excited to vibrate close to its free resonance
frequency. When the atomic force microscope tip approaches the surface, the inter-
action between tip and sample changes the resonance frequency of the cantilever.
The tip-sample force can be represented by a second spring acting in addition to the
cantilever spring. This additional spring leads to a change of the resonance frequency
of the cantilever and correspondingly to a change of the cantilever amplitude. This
change in amplitude can be used as a scheme of force detection and can serve as
the feedback signal for regulating the tip-sample distance. The distance regulation
10 1 Introduction

will be such that a constant amplitude and therefore a constant force (actually force
gradient, as we will see later) is provided.
The idea of scanning probe methods can be considered more generally. A local
probe is scanned over the surface which can detect physical or chemical properties
with high spatial resolution. These techniques are often called SXM techniques where
“X” stands for some specific interaction between tip and sample.

1.3 A Short History of Scanning Probe Microscopy

It is a strange fact in the history of science that the scanning tunneling microscopy
was invented so late. Nobody was brave enough to dare to think so simple: Use
the blindman’s stick principle all the way down to the atomic scale! The principle
is so simple that there are several projects in which already pupils have built an
STM. All the technical ingredients for an STM were invented long before 1981. The
piezoelectric effect was discovered at the end of the 19th century. The electronics for
the STM is also simple; just a function generator to scan and a feedback controller.
From 1930 on it would have been possible to build an STM as the scanning electron
microscope was invented around this time. But no one dared to do so. This may
be also an encouragement for your scientific carrier: be brave and visionary! Some
important and nevertheless simple things may not have been discovered yet.
Here is a short history of scanning probe microscopy:
• 1972 Development of the Topografiner (precursor of the STM).
• 1981 Construction of the first STM by Binnig, Rohrer, Weibel and Gerber.
• 1982 First image of the atomic structure of the Si(111)-(7 × 7) surface by Binnig,
Rohrer, Weibel and Gerber.
• 1985 Invention of the atomic force microscope (AFM) by Binnig, Quate and
Gerber.
• 1986 Nobel prize in physics for the invention of the STM awarded to Binnig and
Rohrer.
• 1987 Element-sensitive imaging of GaAs by Feenstra.
• 1990 Optical beam deflection method introduced by Meyer and Amer.
• 1990 First positioning of single atoms on a surface with a low temperature STM
by Eigler.
• 1993 Tapping mode introduced by Zhong, Inniss, Kjoller, and Elings.
• 1995 First atomic resolution with an AFM by Giessibl.
• 1998 First vibrational spectroscopy with the STM by Stipe and Ho.
Today scanning probe microscopes are standard tools in materials science,
physics, chemistry, biology and engineering. Many thousands of these microscopes
are in operation worldwide, and they are as common and as popular as the scanning
electron microscopes.
1.4 Summary 11

1.4 Summary

• In scanning probe microscopy (SPM) a sharp probe tip is scanned over a surface
and properties of the surface are sensed at the nano- or atomic scale.
• Different kinds of microscopes are used for nanoscale imaging (scanning and trans-
mission electron microscopes as well as scanning probe microscopes) and all have
their advantages and disadvantages in terms of resolution, working environment,
contrast mechanisms, time to obtain an image, and price.
• The atomic resolution in scanning tunneling microscopy (STM) results from the
exponential dependence of the tunneling current on the tip-sample distance.
• In STM, during scanning the height of the tip is adjusted by a feedback loop (and
recorded as the topography signal) such that the tunneling current and correspond-
ingly the tip-sample distance is kept constant.
• Atomic force microscopy can be also applied to insulating samples. The deflection
of a small cantilever senses the force between tip and sample.
• In the dynamic operation mode, the cantilever oscillates and the resonance fre-
quency and subsequently the amplitude change due to the force between tip and
sample.
Chapter 2
Harmonic Oscillator

In scanning probe microscopy, vibrations play a central role in several areas.

If, for instance, a scanning tunneling microscope is rests on a table you might won-
der what this has to do with vibrations. However, floor vibrations with amplitudes
of roughly one tenth of a micrometer (100 nm) have to be compared to an amplitude
stability of less than 0.01 nm which is necessary for atomically resolved imaging
in STM. Thus the vibrational noise amplitude is about 10,000 times larger than the
signal to be measured. This means that knowledge about vibrations and vibration
isolation is essential for scanning probe methods. Another area where oscillations
are an important topic is atomic force microscopy. In the dynamical mode of atomic
force microscopy, a cantilever vibrating close to (or at) its resonance frequency is
used as a force detector. The simplest way to study vibrations is to study the harmonic
oscillator. In this chapter we will study the mechanical harmonic oscillator.

2.1 Free Harmonic Oscillator

The simplest example of a harmonic oscillator is a mass on a spring (Fig. 2.1). The
position to which gravity extends the spring in equilibrium is chosen as the point of
zero extension. The displacement relative to this point is called z. The force exerted
by the spring on the mass m during the oscillation is given by Hooke’s law as

F = −kz, (2.1)

with k being the spring constant. If the spring deflection has negative values (z < 0,
longer spring extension), the direction of the force is positive and vice versa. Thus
the minus sign in (2.1) appears because the force exerted by the spring has a direction
opposite to the deflection z. Newton’s second law tells us that the equation of motion
for the mass m is
d2z
ma = m 2 = m z̈ = F = −kz. (2.2)
dt
© Springer-Verlag Berlin Heidelberg 2015 15
B. Voigtländer, Scanning Probe Microscopy,
NanoScience and Technology, DOI 10.1007/978-3-662-45240-0_2
16 2 Harmonic Oscillator

Fig. 2.1 The simplest

example of a harmonic
oscillator: a mass on a spring

k
k
z
m 0
z <0
m F= -k z

An ansatz for the solution of the equation of motion (2.2) is z = cos(ω0 t) with
ω0 being a parameter which has to be determined.1 We verify that this is a correct
solution by differentiating z two times:

dz d2z
= −ω0 sin(ω0 t); = −ω02 cos(ω0 t). (2.3)
dt dt 2
!
Formally (2.2) is solved if ω0 = mk . But what is the physical significance of ω0 ?
We know that the cosine function repeats itself if the argument is larger than 2π.
Therefore, the mass makes one complete cycle of oscillation if ω0 t = 2π. This time,
we call the period of the oscillation T , and ω0 is given by

ω0 = 2π/T. (2.4)

The angular frequency ω0 is the number of radians through which the oscillation
proceeds per time, while the frequency f 0 is the number of oscillations per time
(ω0 = 2π f 0 ). If the mass is larger it takes a longer time for one oscillation and if the
spring constant is stronger the mass will move more quickly. Note that the period
of oscillation (and also ω0 ) does not depend on how far we stretch the spring at the
beginning. Any solution multiplied by a constant factor is still a solution of (2.2).
We have found a solution to the equation of motion. But is this the only one or
are there more solutions? Also the sine function provides a valid solution. The most
general solution is a linear combination of a sine and a cosine function

z = A cos(ω0 t) + B sin(ω0 t). (2.5)

There is a more intuitive way to find the general solution. When we used the cosine
function as solution, the oscillation started with the maximum extension at time zero.
However, alternatively also any other time during the oscillation could be chosen as
the start of the oscillation. This shift of the time corresponds to a shift of the phase
of the oscillation (the argument of the cosine function is called phase) by a constant

1 The argument of the cosine is named the phase. The phase increases linearly with time if ω0 is
constant.
2.1 Free Harmonic Oscillator 17

phase shift φ. Thus all solutions are captured if the solution is shifted by a constant
(but arbitrary) phase shift φ, and the general solution results as

z = a cos(ω0 t + φ). (2.6)

The two solutions given in (2.5) and (2.6) are in fact equivalent. Using the mathe-
matical identity
cos(α + β) = sin α cos β − cos α sin β, (2.7)

the following relations between A, B in (2.5) and a, φ in (2.6) are obtained

B = a cos φ, A = a sin φ. (2.8)

Moreover, the solutions given in (2.5) and (2.6) are the general solution to the equation
of motion. There are no other solutions.
In the general solution of the equation of motion, we introduced two more
constants: A and B, or a and φ, respectively. How are these constants determined?
They are determined by the initial conditions of the motion. For instance if we start the
motion from a static extension z 0 , B and φ are zero. Now we determine these constants
for the most general initial condition: z 0 , v0 . The acceleration a(t) cannot be speci-
fied as an initial condition. It is given by the spring constant, mass and z(t) according
to (2.2). We use the form for the general solution given in (2.5) and its derivative

v(t) = −ω0 A sin(ω0 t) + ω0 B cos(ω0 t). (2.9)

These equations are valid for all times, but we know z and v at time t = 0. If we
insert t = 0 we obtain

z 0 = A + B · 0 = A v0 = −ω0 A · 0 + ω0 B = ω0 B. (2.10)

We therefore find that the constants A and B can be determined by the initial condi-
tions as
A = z 0 and B = v0 /ω0 . (2.11)

2.2 Driven Harmonic Oscillator

In dynamic atomic force microscopy, we will consider a cantilever which is exited,

driven or moved with a sinusoidal external excitation amplitude. The simplest model
for this is a harmonic oscillator in which the upper fixing point of the spring (cf.
Fig. 2.1) is oscillated (excited) sinusoidally with z drive (t) = Adrive cos(ωdrive t). The
resulting force on the mass m is then F = −k(z − z drive ). The equation of motion
results as
ma = m z̈ = −k(z − z drive ). (2.12)
18 2 Harmonic Oscillator

The driving frequency ωdrive can be different from the natural frequency of the
oscillator ω0 . The question arises at which frequency the driven harmonic oscil-
lator will oscillate. At its natural frequency ω0 , at the driving frequency ωdrive , or at
some value in between? It turns out that the driven harmonic oscillator will oscil-
late in the steady-state at the driving frequency ωdrive . One special solution for the
equation of motion is
z(t) = A cos(ωdrive t). (2.13)

Inserting this ansatz into the equation of motion (2.12) results in

2
− mωdrive A cos(ωdrive t) = −mω02 A cos(ωdrive t) + k Adrive cos(ωdrive t). (2.14)

We find that z = A cos(ωdrive t) is a solution of the equation of motion if

kAdrive
A= . (2.15)
m(ω02 − ωdrive
2 )

The special solution (2.13) means that m oscillates at the driving frequency with an
amplitude which depends on the driving frequency and also on the natural frequency
of the oscillator. If ωdrive < ω0 then displacement and driving excitation are in
the same direction. If ωdrive > ω0 then A becomes negative. This is equivalent to
a positive amplitude and a phase shift of −180◦ of the oscillation z(t) relative to
the driving excitation. The amplitude and phase for an undamped driven harmonic
oscillator are shown in (Fig. 2.2). If ωdrive # ω0 the amplitude A approaches the

Fig. 2.2 Amplitude and 10

phase of an undamped driven
harmonic oscillator as a
function of ωdrive showing a 8
resonance at ω0
6
A/A drive

2
0

0
0 1 2
drive / 0

0
0

0
-90

0
-180
0

0 1 2
drive / 0
2.2 Driven Harmonic Oscillator 19

excitation amplitude Adrive . If ωdrive $ ω0 the amplitude approaches zero because

the mass can no longer follow the high frequency of the driving excitation.
As can be seen in Fig. 2.2 the amplitude A approaches infinity if ωdrive approaches
ω0 . We will see in the next section that damping of the harmonic oscillator prevents
this “resonance catastrophe”.

2.3 Driven Harmonic Oscillator with Damping

Including damping to the driven harmonic oscillator is a more realistic case which
we consider in the following. An additional friction term has to be included to the
equation of motion (2.12). We consider this term as proportional to the speed at
which the oscillating mass moves Ffrict = mγ ż. Also here we assume an external
exciting amplitude z drive (t) = Adrive cos(ωt). Here and in the following we replaced
ωdrive ≡ ω, in order to have a simpler notation. The spring force acting on the
oscillating mass is again proportional to the difference between the position of the
mass z and the excitation amplitude z drive as F = −k(z − z drive ). With this the
equation of motion reads

m z̈ = −mγ ż − k(z − z drive ). (2.16)

Replacing ω02 = k/m results in

z̈ + γ ż + ω02 z = ω02 z drive . (2.17)

Solving this equation would be quite difficult without the use of complex numbers.
The trick here is to consider z and z drive as complex numbers (z̃ and z̃ drive ) and find
the complex solution for the differential equation. Since the physical quantities are
real and the differential equation is linear, at the end only the real part of z̃ is our
solution. The amplitudes are regarded as complex numbers as

z̃ = Aei(ωt+φ) = Aeiφ eiωt = ẑeiωt and z̃ drive = Adrive eiωt . (2.18)

Without loss of generality we set the phase shift of the excitation amplitude z drive
to zero, i.e. Adrive is real, while ẑ is regarded as a complex number with a (real)
phase shift φ and (real) oscillation amplitude A as, ẑ = Aeiφ . The real part of z̃
will later be the real solution for the motion of the mass m. The nice thing about
the complex notation is that differentiation of z̃ is now just multiplication with iω
( ddtz̃ = ẑiωeiωt = iω z̃). This means differentiation in (2.17) (with z → z̃) can
be easily executed and this differential equation converts to the simple algebraic
equation " #
(iω)2 ẑ + γ(iω)ẑ + ω02 ẑ eiωt = ω02 Adrive eiωt . (2.19)
20 2 Harmonic Oscillator

After dividing both sides by eiωt , we obtain the complex solution

ω02 Adrive
ẑ = . (2.20)
ω02 − ω 2 + iγω

Now the real z is the real part of the complex quantity z̃ as

z = Re(z̃) = Re(ẑeiωt ) = Re(Aei(ωt+φ) ). (2.21)

Since A and φ are real, the resulting real position z reads

z = A cos(ωt + φ), (2.22)

with the amplitude A and phase shift φ between excitation amplitude and oscillation
amplitude.
In order to calculate A we recall that ẑ = Aeiφ . Therefore, ẑ ẑ ∗ = A2 and A2 can
be written as

ω04 A2drive ω04 A2drive

A2 = $ 2 %$ 2 %=$ %2 . (2.23)
ω0 − ω 2 + iγω ω0 − ω 2 − iγω ω 2 − ω0 2 + γ 2 ω 2

Now we introduce as a convenient abbreviation the quality factor Q = ω0 /γ. The

physical significance of the quality factor will be elucidated later. This replacement
results in

ω04 A2drive
A2 = $ %2 . (2.24)
ω02 ω 2
ω 2 − ω02 + Q2

Furthermore, the oscillation amplitude A can be written as a function of the normal-

ized frequency ω/ω0 as

A2drive
A2 = & '2 . (2.25)
2 1 ω2
1 − ωω2 + Q 2 ω02
0

The phase φ of the oscillation relative to the excitation can be obtained as follows.
In general the phase ϕ of a complex number x = r eiϕ can be obtained from the
relation tan ϕ = IRe(x)
m(x)
. In order to calculate the phase φ, we recall that ẑ = Aeiφ .
However, according to (2.20) the real and imaginary parts of 1/ẑ are much easier to
find. Therefore, we write

1 1 1 −iφ 1 ( )
2 2
= = e = ω 0 − ω + iγω . (2.26)
ẑ Aeiφ A ω02 Adrive
2.3 Driven Harmonic Oscillator with Damping 21

Using the fact that tan(−φ) = − tan φ, we see that

−γω −ω0 ω
tan φ = = $ 2 %. (2.27)
ω02−ω 2 Q ω0 − ω 2

Also the phase φ can be written as function of the normalized frequency ω/ω0 as

− ωω0
tan φ = * ( )2 + . (2.28)
Q 1 − ωω0

With these results, the amplitude (2.25) and phase (2.28) in the solution (2.22) are
calculated as a function of given variables. The resonance curve in Fig. 2.3 shows the
amplitude and the phase of a driven damped harmonic oscillator. For small driving
frequencies ω # ω0 , the motion of the oscillator mass just follows the outer excitation
with a phase approaching zero; i.e. the oscillation is in phase with the excitation. On

Fig. 2.3 Amplitude and 15

phase of a damped driven
harmonic oscillator as a Q=30
function of ω ≡ ωdrive , for
different values of damping
10
Q = ω0 /γ
A/Adrive

Q=10

Q=3

0
0 1 2

0
0
Q=30
0
-30

0 Q=10
-60

0
-90

0
-120
Q=3
0
-150

0
-180
0 1 2
22 2 Harmonic Oscillator

the other hand for very large frequencies ω $ ω0 , the amplitude A approaches zero.
In this case the phase approaches −180◦ , i.e. the motion of the oscillator mass is
always in opposite to the excitation.
If we take the limit ω $ ω0 in (2.25) we find that the amplitude is proportional to
1/ω 2 for small damping, i.e. γ # ω0 or Q $ 1. As seen in Fig. 2.3, the smaller the
damping, the higher the maximum amplitude is. For small damping the maximum
of the resonance curve is very close to the resonance frequency of the free harmonic
oscillator ω0 . At any driving frequency the phase is smaller than zero, which means
that the oscillator displacement z always lags behind the driving excitation (Fig. 2.3).
The phase at resonance (ω = ω0 ) is −90◦ , while it approaches −180◦ for large
driving frequencies.
The amplitude at the resonance frequency A(ω0 ) can be obtained using (2.25) as

A(ω0 ) = Q Adrive , (2.29)

i.e. the amplitude at resonance is Q times higher than the excitation amplitude. For
the case of cantilevers in atomic force microscopy this resonance enhancement of
the excitation amplitude can be quite high. Due to damping in air, Q-factors of 500
are usual for cantilevers in air. In vacuum, quality factors higher than 10,000 can
be reached.
For the case that the oscillation frequency is very close to ω0 , i.e. ω ≈ ω0 , the
expression for the resonance curve (2.25) can be approximated as

A2drive A2drive
A2 = "( )( )#2 ≈ ( )2 . (2.30)
1 ω2 1
1 + ωω0 1 − ωω0 + Q 2 ω02
4 1 − ωω0 + Q2

ω2
In order to obtain this we used the approximations 1 + ωω0 ≈ 2 and ω02
≈ 1, which
hold if ω ≈ ω0 .
An important quantity is the width of the resonance curve. Therefore, we calculate
following the frequency ω1/2 at which the amplitude of the oscillation decreases
in the√
to 1/ 2 of its value2 at ω0 . This condition for the amplitudes can be written as

1 1
A1/2 (ω1/2 ) = √ A(ω0 ) = √ Q Adrive . (2.31)
2 2
If we insert ω = ω1/2 in expression (2.30), the following relation results

1 2 A2drive 1 2 2
A (ω1/2 ) ≈ ( )2 ≈ Q Adrive . (2.32)
2 1/2 ω1/2 1 2
4 1− ω0 + Q2

2
√
We use the decrease of the amplitude to 1/ 2 instead of 1/2, because in this case the energy in
the harmonic oscillator, which is proportional to the square of the amplitude, decreases to one half
of its maximum value.
2.3 Driven Harmonic Oscillator with Damping 23

1 ω0
Solving this expression for ω1/2 − ω0 results in ω1/2 − ω0 ≈ 2 Q. Since the full
width of the resonance curve is twice this, we obtain
ω0
!ω1/2 ≈ . (2.33)
Q

This means the larger the Q-factor, the narrower the resonance is.
The maximum of the resonance amplitude, which we determine in the following,
lies at a slightly lower frequency than ω0 . The maximum of the resonance curve occurs
at the frequency at which the denominator in (2.25) becomes minimal. Differentiating
the denominator of (2.25) with respect to ω/ω0 , and equating this derivative to zero
results in the following expression for the frequency ωmax at which the resonance
curve has its maximum
& '
2 1
ωmax = ω02 1 − . (2.34)
2Q 2

The corresponding shift of the resonance curve to lower frequencies results as

, - .
1
δω = ω0 − ωmax = ω0 1 − 1− . (2.35)
2Q 2

For the case of an AFM cantilever considered as a harmonic oscillator we estimate

some values for this frequency shift of the resonance curve due to the damping Q
of the cantilever. For a resonance frequency of ω0 = 300 kHz and quality factors
of Q = 10,000 and Q = 300, a frequency shift of 0.8 mHz and 0.8 Hz results,
respectively. These are very small values and correspondingly in most cases we will
neglect this small shift and consider the maximum of the amplitude to be located at
ω0 , unless the quality factor is very low.

2.4 Transients of Oscillations

The solution for the damped driven harmonic oscillator (2.22) is the so called steady-
state solution after transients due to the initial conditions have died out. An example
for a transient is an oscillation which starts from rest. The amplitude is initially zero,
builds up after the excitation starts, and reaches the steady-state amplitude in the
limit of large times. The steady-state solution (2.22) does not contain such transients
arising from specific initial conditions.
It can be shown that the general solution of the driven damped harmonic oscilla-
tor is the specific solution (2.22) plus a solution of the corresponding homogeneous
problem. The corresponding homogeneous problem is the damped harmonic oscil-
lator without external driving. Here we do not derive the solution for the damped
oscillator without driving but it should be remembered that this is (for small damping)
24 2 Harmonic Oscillator

an exponentially decaying oscillation z hom = G exp(−ω0 /(2Q)t) cos(ωhom t − φ)

with the oscillation frequency ωhom being slightly / lower than the natural frequency
ω0 of the free harmonic oscillator ωhom = ω0 1 − 1/(4Q 2 ) and with G and φ as
coefficients specified by the initial conditions.
If we call the specific solution z in (2.22) z s , the general solution for the driven,
damped harmonic oscillator is given as z general = z hom + z s . It is necessary to
include the solution of the damped harmonic oscillator without external driving z hom
since it can describe the transients which are not described by z s . All aspects of z s are
specified in terms of the driving frequency, the driving amplitude, and the phase shift.
Yet we still need some way to impose the constraints given by the initial conditions
z(0) and v(0) in the general solution. The two coefficients G and φ give the freedom
to match the general solution to z(0) and v(0).
As an example we consider as initial condition that the oscillation starts from rest.
In Fig. 2.4 the general solution for the initial condition: starting from rest, is shown to
be composed of the specific solution of the inhomogeneous system (Fig. 2.4a) plus
the solution for the homogeneous system (transient) z hom (Fig. 2.4b). In Fig. 2.4c the
sum of both is shown for the case that ω = ωhom . The specific solution in Fig. 2.4a
is approached within the decay time for the homogeneous solution Fig. 2.4b. The
fact that the situation is not always simple is shown in Fig. 2.4d. Here the driving
frequency deviates from ωhom , which leads to a beating behavior before a steady-state
solution is reached.

(a) (b)

0
Z

t/(2Q)
cos( hom t+ ) e cos( hom t+ )

(c) (d)

0
Z

hom 1.2 hom

Time Time

Fig. 2.4 The general solution for a damped driven harmonic oscillator is composed of the specific
solution of the inhomogeneous driven system (steady-state solution), shown in (a) plus the solution
of the homogeneous system without driving (transient), shown in (b). The initial conditions are
chosen such that the general solution satisfies the given initial conditions (start from rest in this
example). c and d show two examples of general solutions (for two different driving frequencies)
starting from rest and approaching the steady-state solution for long times
2.4 Transients of Oscillations 25

If the driven damped oscillator is oscillating in steady-state (Fig. 2.4a) and the
driving amplitude is stopped suddenly, the problem is converted to a homogeneous
one and the oscillator will de-excite as shown in Fig. 2.4b. This is a sinusoidal oscil-
lation with the envelope decreasing as exp(−ω0 /(2Q)t). This means that after a time
τ = 2Q/ω0 = TQ/π the amplitude has decreased by 1/e. This characteristic time is
called ring down time and increases with smaller damping. The same time is needed
to build up the steady-state oscillation amplitude after a start from rest.
This time can be expressed in terms of the Q-factor as τ = 2Q/ω0 = TQ/π. This
means that the oscillation builds up (decays) within roughly Q oscillation cycles and
Q can be expressed as

1
Q= τ ω0 . (2.36)
2

2.5 Dissipation and Quality Factor of a Damped

Driven Harmonic Oscillator

When the mass is initially at rest and an external oscillatory excitation is applied,
energy is successively stored in the oscillator with the buildup of the oscillation (tran-
sient). If the oscillator is finally in a steady-state, the energy stored in the oscillator
is constant and all the energy supplied by the external force ends (on average) up
in the dissipative term. The instantaneous power dissipated is Ffrict v = γmv 2 and
varies over one period, as v varies. The mean power consumed by the oscillator in
steady-state can be written as

*P+ = *Ffrict v+ = γm*v 2 +. (2.37)

The brackets indicate an averaging over one oscillation period. Since z = A cos(ωt +
φ), differentiation results in v 2 = ω 2 A2 sin2 (ωt + φ). If sin2 is averaged over one
period a factor of one half results. Therefore, the average power results in

1
*P+ = γm*v 2 + = γmω 2 A2 . (2.38)
2
With this the energy dissipated per cycle is

Energy dissipated per cycle = *P+T = *P+2π/ω = πγmω A2 . (2.39)

Another important quantity is the total energy stored in the oscillator. If we con-
sider driving frequencies close to ω0 , the energy stored in the driven oscillator is
approximately the energy of the free oscillator with the same amplitude A

1 2 1
*E+ ≈ k A = mω02 A2 . (2.40)
2 2
26 2 Harmonic Oscillator

The efficiency of an oscillator is defined by how much energy is stored, compared

with how much work is supplied (dissipated) by the external force per oscillation
cycle. This is called the quality factor of the oscillator and is defined by 2π times the
mean energy stored, divided by the energy dissipated per cycle

Energy stored in the oscillator

Q = 2π × . (2.41)
Energy dissipated per cycle

Close to the resonance frequency (ω ≈ ω0 ), Q can be written using (2.39) and

(2.40) as
ω0
Q≈ . (2.42)
γ

This is consistent with the abbreviation for Q introduced in the previous section.

2.6 Effective Mass of a Harmonic Oscillator

In this chapter, we always considered an idealized system consisting of a mass-less

spring and a mass m at its end. However, in some cases of practical relevance this
approximation is not fulfilled. For instance, in the case of a cantilever-type spring,
often used in atomic force microscopy, the mass (of the cantilever) is distributed
throughout the whole cantilever (Fig. 2.5b). We introduce the concept of the effective
mass for the example of a coil spring (with mass m spring ) and assume that the mass

(a) (b)
z

dz
v(z) L x z(x), v(x)
m spring
z

M L

vmax
Fig. 2.5 a For a spring with mass m spring , the velocity of a volume element depends on the position,
i.e. v = v(z). The effective mass turns out to be 1/3 of the spring mass. b For a cantilever beam the
deflection and the velocity are non-linear as a function of x
2.6 Effective Mass of a Harmonic Oscillator 27

is distributed homogeneously along its length. In the following, we calculate the

maximum kinetic energy (which corresponds to the total energy) of the spring with
a mass and we do not consider a mass M at the end of the spring.
When calculating the (maximum) kinetic energy of the spring, we regard v(z) as
the maximum velocity during one oscillation cycle. The (maximum) kinetic energy
of a length element dz of the spring is given by

1 m spring 2
d E kin = v (z)dz. (2.43)
2 L
According to Fig. 2.5a, the velocity distribution along the spring is linear with z and
can be written as v(z) = vmax z/L, with vmax being the maximum velocity at the end
of the spring, i.e. v(L). Integrating the (maximum) kinetic energy along the spring
results in

0L 0L
1 m spring 2 1 m spring 2 z2
E kin = v (z)dz = vmax dz
2 L 2 L L2
0 0
& '
1 1 2 1 2
= m spring vmax = m eff vmax . (2.44)
2 3 2

Thus a mass-containing spring is equivalent to a massless spring with an effective

mass m eff = 1/3m spring fixed to the end of the spring. If an additional mass M at the
end of a spring is also considered, the effective mass becomes m eff = M +1/3m spring .
While we only considered the expression of the kinetic energy here, the same
effective mass also enters into the equations of motion, and thus also into all following
results. For instance, when calculating the natural frequency of a harmonic oscillator
in which the spring contains mass, the effective mass has to be used instead of the
mass M at the end of a massless spring.
For the situation of a cantilever beam the situation is more complicated, because
the deflection z (in reaction to a force applied at the end of the cantilever) is not linear
along the cantilever beam as shown in Fig. 2.5b. According to [1], the bending has
the form z(x) ∝ −x 3 + 3x 2 L. Since a harmonic oscillation is considered throughout
the beam, the velocity distribution along the beam is proportional to the deflection
v(x) = cz(x). The constant of proportionality is determined by the condition v(L) =
vmax as c = vmax /(2L 3 ). Thus the maximum velocity at position x along the beam
results as
vmax ( 3 2
)
v(x) = −x + 3x L . (2.45)
2L 3
Using this expression for the velocity distribution along the beam, the (maximum)
kinetic energy can be obtained by integration along the beam as
28 2 Harmonic Oscillator

0L 2 ( )2 & '
1 m cant vmax 3 2 1 33 2
E kin = −x + 3x L d x = m spring vmax
2 L 4L 6 2 140
0
1 2
= m eff vmax . (2.46)
2
Thus the effective mass for a cantilever beam turns out to be ∼0.2357, instead of
1/3 for a coil spring with a linear extension.
In the case of a cantilever spring, an effective mass has to be used
√ in the equation of
motion and all subsequently derived expressions such as ω0 = k/m eff . Throughout
this text we use the concept of the harmonic oscillator and denote the mass as m in
order to keep the notation simple. It has to be kept in mind that in fact the appropriate
effective mass haves to be used.

2.7 Linear Differential Equations

At the end of this chapter, we consider some general properties of linear differential
equations with constant coefficients. A homogeneous linear differential equation
up to the second order can be written as

0 = a1 x + a2 ẋ + a3 ẍ. (2.47)

The following propositions hold for the homogeneous equation.

• Homogeneity: If x is a solution of the linear differential equation, C x is also a

solution.
• Superposition: If x1 and x2 are solutions of the linear differential equation, x1 + x2
is also a solution.
• Combining the two, we see that all linear combinations of two solutions are also
solutions.

The corresponding inhomogeneous equations including an external driving force

F(t) can be written as
F(t) = a1 x + a2 ẋ + a3 ẍ. (2.48)

If we have a (special) solution of the inhomogeneous equation x1 , we can add

any solution x2 of the homogenous (free) equation F(t) = 0 and the sum x =
x1 + x2 will be also a solution of the inhomogeneous system as we see if we add the
inhomogeneous equation and the homogeneous equation as

F(t) = a1 (x1 + x2 ) + a2 (ẋ1 + ẋ2 ) + a3 (ẍ1 + ẍ2 ) = a1 x + a2 ẋ + a3 ẍ. (2.49)

Finally, we come to another important property of linear differential equations.

If we have a solution x1 for an external force F1 (t) and a second solution x2 for another
2.7 Linear Differential Equations 29

external force F2 (t), then a solution for the problem with the force F1 (t) + F2 (t) is
x1 +x2 . This superposition principle is remarkable and is the basis for decomposing a
complicated (arbitrary) force into Fourier components and composing the solution of
the problem with a complicated force as a superposition of the solutions obtained for
simple harmonic forces. This is also a late justification for why we only considered
an external excitation (force) of simple harmonic form for the harmonic oscillator.

2.8 Summary
!
• The free harmonic oscillator has the natural frequency of ω0 = mk .
• The driven harmonic oscillator oscillates at the driving frequency ω with an ampli-
tude depending on ω and ω0 .
• If ω = ω0 the amplitude becomes very large (resonance).
• For the damped driven oscillator the amplitude at resonance is damped with
increasing damping force Ffrict = mγ ż.
• The phase between driving excitation and oscillation is zero if ω # ω0 , it is −90◦
if ω = ω0 , and −180◦ if ω $ ω0 .
• The quality factor of the oscillation Q is the ratio between the energy stored in the
oscillator to the energy dissipated per cycle. Q ≈ ωγ0 ≈ !ωω0
≈ A(ω0 )/Adrive , with
!ω being the width of the resonance curve and Adrive the excitation amplitude.
• The build up or the decay of the steady-state amplitude takes about Q oscillations,
i.e. the corresponding time constant is τ = 2Q/ω0 .
• If a spring has a non-negligible mass, the effective mass has to be used in the
equations of the harmonic oscillator.