Scanning Tunneling Microscopy

(STM)
 The Scanning Tunneling Microscope (STM)

HEINRICH ROHRER GERD BINNIG

STM invented in 1981 by Greg Binnig & Heinrich Rohrer at IBM and were
awarded with a Nobel Prize in 1986
* Ernst Ruska was the other winner

“IBM” spelled in Xenon atoms

 Classical vs. Quantum
Classically, if one threw a ball at a brick wall (and didn’t miss horribly) they would
always find the ball on their side of the wall, because unless you are Superman there is
no way that you are throwing the ball hard enough to break through the wall.

Some
later
time
Quantum mechanically, however, the ball has some finite probability of “tunneling”
through the wall. The ball need not be thrown incredibly hard to achieve this,
although the more energy the ball has the more likely it will be to tunnel.

Some
later
time
 Mathematical Approach
Standard 1-D wave equation where U(z) is the potential of the barrier, and
E is the particle’s energy

 2  2 n ( z )
− + U ( z ) n ( z ) = E n ( z )
2m z 2

Which has the normal solution of a traveling wave:

 n ( z ) =  n (0)e  ikz where

2m( E − U )
k=
2
This is the general solution but in the event that U > E we can factor out an
i and we get a solution that has a completely different meaning.

 n ( z ) =  n (0)e  kz

which for our scenario, where z > 0, we get an exponential decay as the
electron sees if it can tunnel through the barrier.
 Visualization of tunneling
Higher energy particle has a possibility of
passingLow
through
energytheparticle
wall andiscontinuing with less
confined to
energy
stay in itsthan before.
potential well.
 Practical application for probing

A
Current

If two metals are connected to opposite ends of a battery but are

separated from each other than they logically should not conduct
electricity. However, as we bring the metals extremely close to each other
the wavefunctions of the electrons are able to tunnel through the gap and
we will detect a current on our meter.
 Problems with probabilities
After a little bit of math we can come up with the probability that an
electron can tunnel through our barrier,

P   n ( 0) e − 2 kW
2
where W is the barrier width.

We can look at the above probability as the chance of an electron on one

metal being found on the other, but likewise this works the opposite way as
well. This brings rise to another problem with the above equation, even if an
electron could tunnel the gap it has to have a home on the other metal.

First, let’s look at biasing one movement over the gap. This is achieved just
as shown previously by connecting both metals to different ends of a
voltage source. This voltage is very important though, we don’t want to arc
the material (high voltage) but we do want to give the electrons a natural
tendency to flow one way. As such, the applied voltage is normally on the
order of the work function of the material (a few eV, 4-5eV).
 So many electrons to count
Since we are applying a voltage nearly equivalent to the work function of
the material, we are allowing any electron at, or near, the Fermi energy the
possibility of being free and tunneling our gap.

We are now dealing with Fermi energies and as such our discussion must
now change from individual electrons but to the masses. We now have to
look at all the possible electrons that we could see tunneling our gap.
Luckily, we have some bounds on these electrons, so we need not look at all
of them.
To be more exact we care only about a
small band of electrons that are very
close to the Fermi energy of the material.
Ef
Ef-eV If we were to measure the probability as
the metals are brought close we could only
possibly detect those electrons move, and
this is only a half truth since we still
require they have a free state to tunnel
into.
 Currents, a more usable quantity

We need some other measure of electrons moving across our gap. Electrons
moving leads instantly into current so we need a way to quantify our
probabilities in terms of currents.

Without getting into too many mathematical complications, a relatively

simple equation of the current we can expect to detect:

I  V (W , E f )e−2kW

where (W,Ef) is the density of states of one metal through the gap,
W, at the other metal.
 Some engineering concerns

With some physics in the bag let’s look at some engineering challenges. We
need an extremely accurate way to bring a probing material to our sample,
that is stable and is capable of very fine adjustments. To give you a feel for
if one cranked out some numbers for the above equations we need to be
able to bring our probe to within Angstroms of the surface and then be
able to move in Angstrom increments.

Sample Probe

Angstroms
 Some engineering concerns (cont.)

A second problem is that we cannot simply move a block of metal towards a

sample, if we did we would be probing the entire surface at once and that is
only if our surface is perfectly flat!

Sample Probe

We’ll look at this one first.

 A perfect probe

Hopefully, it is clear that we should have a probe that is extremely small, or

ideally on the atomic scale. If we were to look at the math again (which I’m
not) we would see that if we had a probe that had exactly one atom on the
end then the subsequent atoms behind it would contribute almost nothing to
our current.
Current

~e-2(6)k

~2e-2(9)k
6Å
Sample
3Å
As it turns out atomic probes are easily
made with either mechanical or chemical
processes.
 Basic Principles of STM

Bias voltage:
d~6Å mV – V range

Electrons tunnel between the tip and sample, a small current I is

generated (10 pA to 1 nA).

I proportional to e-2κd, I decreases by a factor of 10 when d is

increased by 1 Å.
Electron Tunneling:
In scanning tunneling
microscopy a small bias
voltage V is applied so
that due to the electric
field the tunneling of
electrons results in a
tunneling current I. The
height of the barrier can
roughly be approximated
by the average
workfunction of sample
and tip.
Ideally a STM probe tip is very pointed (1-2 atoms
at the end) and has a relatively low work function.
Etched tungsten crystals are ideal and are nearly
identical to field emitters.
The probe is scanned
over the surface in a
raster pattern similar
to that of a SEM or
Confocal. Each
coordinate (X,Y, & Z)
is recorded by a
computer.
 Accurate distance control

Angstrom level distance control is required which would be impossible for

any mechanical gear workings.

Enter the world of piezoelectrics

A piezoelectric is a crystal that creates potential differences (voltages)

when mechanical stresses are imposed on it.

Voltage

0
 Piezoelectric motors

The mechanical to electrical effect is completely reversible and by applying

voltages to these materials we force mechanical effects out of them.

In fact this effect is so common place that we can find it in everyday

objects like speakers to printers, and they can be as sophisticated as
extremely accurate mirror control in laser laboratories.

One can reliably place our sampling tip within 10’s of Angstroms from a
surface without using piezoelectric motors but beyond that the accurate
probing is handled by the piezoelectric.

We’ll even use these motors in a feedback system to adjust the height of
our probe.
 Probing a surface

With all of the elements in place we can start probing our metallic surface.

Or
Wewecould
could
either
vary probe
the height
at a constant
in search
height and
for look
constant
at current
current.
variations
as we pass.

PROBE
Current
Height

Distance
 Two Modes of Scanning

Constant
Height
Mode

Constant
Current
Mode
Usually, constant current mode is superior.
Instrumental Design: Controlling the Tip

Precise tip control is achieved with

Piezoelectrics

Displacement accurate to ± .05 Å

Raster scanning
 Probing a surface

Although more complex, one normally modulates the height of the probe.
We can see this as a major advantage in the sense that we are not
contingent on the surface being very uniform, a crag or cliff of atoms does
not pose a collision hazard as well as at a constant current we can suppress
effects that may come with higher currents (heat, breakdown of tip, etc.).

The biggest thing we must keep in mind is that an STM cannot “see” atoms.

An STM probes the density of states of a surface.

As such, holes are a common artifact of STM, a hole does not mean that no
atom exists it merely means that the probe cannot reach below the first
layer of atoms. Likewise, peaks in an STM do not necessarily mean there is
a mountain of atoms, the atoms may be a different element from the
surroundings and may possess a higher density of states resulting in a
higher current/probe height.
 Images of an STM

The Nobel prize winning first STM.

A more current example.

STM produced images

STM image, 7 nm x 7 nm, of a

single zig-zag chain of Cs atoms
(red) on the GaAs(110) surface
(blue). *

*National Institute of Standards and Technology

 Surface of platinum.
IBM, Almaden Research Facility

Surface of nickel.
IBM, Almaden Research Facility

Surface of copper.
IBM, Almaden Research Facility
Ring of iron atoms on a copper blanket.
IBM, Almaden Research Facility
 Startling aside...

In 1989, at the Almaden IBM Research Facility scientists found

that an STM could be used to lift atoms off a surface of metal and
placed back in a different location, at low temperatures that is.
 Advantages Disadvantages
No damage to the sample Samples limited to conductors and
semiconductors
Vertical resolution superior to SEM

Limited Biological Applications:

Spectroscopy of individual atoms AFM

Generally a difficult technique

Relatively Low Cost to perform

Figures of Merit
Maximum Vertical Maximum Lateral
Resolution: .1 Å Resolution: 1 Å

Maximum Field of View: 100 μm

 Applications of STM
Surface Structure: Compare to bulk structure

Semiconductor surface structure, Nanotechnology,Superconductors,

etc.

Metal-catalyzed reactions

Spectroscopy of single atoms

Limited biological applications: Atomic Force Microscopy

Future Developments: Improve understanding of how electronic

structure affects tunneling current, continue to develop STM
offshoots
Scanning Tunneling Spectroscopy
Scanning tunneling spectroscopy (STS) is a technique that allows the
study of the electronic structure of surfaces with atomic resolution.

Transmission of the electric current between the tip and the sample
depends on two factors:

(1) Geometry of the sample (influenced by presence/absence of

atom/molecule etc.)
(2) Electron density : Function of both position and energy described
as Local Density of States (LDOS).

➢ In STS, the STM is used to measure the number of electrons

(LDOS) as function of electron energy.
➢ Electron energy is set by electrical potential difference between
the sample and the tip.
➢ Location is set by the position of the tip.
In general, STM involves a very sharp conductive tip that is brought
within tunneling distance (sub-nanometer) of a conductive sample
surface, thereby creating a metal-insulator-metal (MIM)
configuration.

In the representation of one-dimensional tunneling , the tunneling

wave of the sample electrons, s , and the wave of a STM tip
electrons, t , overlap in the insulating gap, allowing a current to flow.
➢Thus, to observe the tunneling current I of electrons through the vacuum gap
between the sample and the tip, a bias voltage, Vbias, is applied, as shown in
Figure .
➢At Vbias= 0, the electrons cannot flow in either direction since the Fermi
level, Ef, of both the tip and the sample is equal, i.e., the gradient is zero.
➢When Vbias > 0 (positive bias), the Fermi level of the sample is raised by Vbias
and the electrons in the occupied state (filled with electrons) of the sample
can tunnel into the unoccupied state of the tip.
➢Similarly, when Vbias < 0 (negative bias), the electrons in the occupied state of
the tip tunnel into the unoccupied state of the sample.
Tunnel Current, Vacuum Gap Size and Density of States

The tunneling current decays exponentially with the distance gap distance z,
and is strongly affected by the density of states (DOS) of the sample at the
Fermi level.

An electronic state describes a specific configuration, an electron can possess.

For instance, it can have either a spin up or spin down, or a particular magnetic
momentum etc. A state is described by a set of quantum mechanical numbers.
Each state can only be filled by one electron.

Density of state (DOS), N, and Fermi energy XF in two classroom settings.

(left) N is constant. (right) N(x).
Many physical properties are affected or depend on the
number of states within an energy range (i.e., the energy
density of states).

While in metals and semi-metals, there is relatively small

variation in the density of states due to the large electron
delocalization, the density of energy levels in
semiconductors varies noticeably.

Thus, knowledge about DOS is of immense importance for

electronic applications involving semiconducting materials,
where the availability of empty valence and conduction
states (states below and above the Fermi level) is crucial
for the transition rates.
Figure illustrates the tunneling mechanism involving a semiconductor.
The filled area (grey) is not uniform, representing the variation in
electron density, and the lines in the unoccupied levels represent the
variation in density of the energy levels that the tunneling electrons
can occupy.
STM and Local Density of States

STM constant current maps provide information about the variations in the
electron density, and do not necessarily correspond to the location of atoms
(nuclei).

Figure illustrates that a location of high tunneling current in a STM image can
be either a compounded affected of two atoms, leading to a current maximum
in between the atoms, or be identical with the location of an atom.
➢This is for instance found for the silicon (001) 2x1 surface. A π
molecular orbital of the silicon-silicon dimers (Si=Si) creates the
highest electron density (probability) at the center of the dimers,
while an antibonding π* molecular orbital has a node (a location where
the probability is zero) at the center of the dimers.

➢Thus, when a negative bias is applied, the electrons in the π-

molecular orbital (occupied state) tunnel and the resulting image,
similar to the case shown in Figure (a), will be obtained.

When a positive bias is applied, the electrons of the tip tunnel into the
anti-bonding π* molecular orbital (unoccupied state), revealing a gap
between the dimers, as in Figure (b).

When the variation in the local DOS (LDOS) of metals is small, the
contour of STM images often can be safely interpreted as the
topography of the atomic lattice.
In the STM imaging mode, the tunneling current I is
continuously recorded at each location (x,y) at a constant
bias voltage Vbias, generating a two-dimensional map of
tunneling conductance I/Vbias.

In contrast, tunneling spectroscopy (I-V curve) focuses on

the tunneling conductance, or commonly, a normalized
differential tunneling conductance (dI/dV)/(I/V) .

Tunneling spectroscopy studies are usually performed

without scanning at a particularly chosen location, based
on an initial STM current or height map. However, it is
also possible to scan while the bias is ramped (scanning
tunneling spectroscopy (STS)).
The general profile of the density of state around the Fermi level, i.e.,
(dI/dV)/(I/V), can be used to classify the material based on its conductivity,
as illustrated in Figure.
For semiconductors and insulators, the conductance around
the Fermi level is zero. The threshold voltage, i.e., band
gap, Eg = |V+bias| + |V-bias|, is relatively small for
semiconductor (< 3eV, used as definition for
semiconductors).

As shown in Figure, semiconductors show a highly bend

DOS, which is flat as for insulators at low voltages, where
the energy gap Eg cannot be bridged. It is well known that
doping semiconductors with impurities or defect sites
affect reduce Eg , and thus, can modify the density of
states at the Fermi level to such a degree that it
resembles nearly a semi-metal.
December 2015 Physical Review B 92(21) 214506

A plot of scanning-tunneling spectroscopy data for superconducting

amorphous tungsten-based nanoscale deposits (Tc=4.15K)