CARNEGIE MELLONDepartment of Electrical and Computer Engineering

Design and Constructionof a Scanning

Tunneling Microscope

Roland Schaefer

1989

\ MellOn

Design and Construction

of a

Scanning Tunneling Microscope

Roland Schaefer

Department of Electrical and Computer Engineering

Carnegie Mellon University

Pittsburgh, Pennsylvania 15213

September 12, 1989

Submitted in partial fulfillment of the requirements for

the degree of Master of Science in Electrical Engineering.

(~)Roland H. Schaefer; M1 rights reserved

Abstract

This report describes the design and construction of a Scanning Tunneling Mi-

croscope (STM). A Scanning Tunneling Microscope is a relatively new instrument

in the field of surface science and has the capability of atomic resolution of surfaces.

This report presents some background theory pertinent to the understanding of

the STM. Theories of electron tunnefing, STM operation, and control theory are all

briefly covered.

This report also presents the mechanical and electrical designs used in the con-

struction of the STM. Some historical developments are illustrated to justify design

choices. Both the mechanical and electrical designs are described in detail.

Also presented in this report are some results obtained during the design and

construction of the STM. The imaging capabilities of the STM is presented with the

images of different surfaces: a machined surface, an epitax_ially grown monocrystal

of gold~ and graphite.

Acknowledgments

There are many people who gave me a great deal of help and support during the

course of this project. Without their help, I’d still be struggling today.

I’d Like to thank Michael Reed for his understanding and patience when things

got rough. Without his help this project would never have been possible. Michael

sense of humour always made the meetings worthwhile; always had the appropriate

joke for the situation. I’d also Like to thank the two committee members for t:heir

time involved in this project; Ed Schlesinger for his observations and comments on

other operating STMs, and Dan Stancil for taking time off during his holidays to

read this report.

I am also greatly indebted to many other students: David Wong for his eternal

pessimism to which I should have paid more attention; Rob Sturgill for his help in

the laser lab; Jack Kenney for his comments and assistance in my circuit designs;

the members of the Semiconductors Group - Tony, Kim, Nik, and Michele.

I’d also Like to thank Jim Schubert for all his time and assistance in the machine

shop. His trust in my use of the machines allowed me to save a lot of time in the

construction of the STM.

Last of a~l I’d Like to thanks all my friends (especially Andi) who gave me the

moral support when I needed it, put up with my frustrations and took me out for

a beer when I was discouraged. Thanks for putting up with me!!

ii

Contents

1 Introduction 1

1.1 History ................................... 1

1.2 Project Outline ............................... 2

1.3 Outline .................................. 2

1.4 Other Documentation ........................... 2

2 Background 3

2.1 Electron Tunneling ............................ 3

2.2 STM Theory ............................... 5

2.3 Resolution ................................. 6

2.4 Tips .................................... 6

2.5 Image Interpretation .. ................... " ....... 7

2.6 Control Theory .............................. 7

Mechanical Designs 10

3.1 Tunnehng Environment ......................... 10

3.2 Vibration Isolation ............................ 10

3.3 Tip Movement .............................. 12

3.4 Mechanical Approach .......................... 14

4 Electronic Designs 15

4.1 Preamplifier ................................ 15

4.2 Feedback Circuit ............................. 18

4.3 LateralMotion .............................. 21

4.3.1 Ramp Circuit ........................... 21

4.3.2 Offset Circuit ........................... 21

4.4 Circuit Performance ........................... 23

111

Results 25

5.1 Calibration of the Tube Scanner .................... 25

5.2 Exponential Dependency ......................... 25

5.3 Sample Images ............................... 26

5.3.1 Machined Surface ......................... 27

5.3.2 Gold Monocrystal ........................ 28

5,.3.3 Graphite ............................. 28

6 Conclusions and Future Work 34

6.1 Future Work ............................... 34

A Computer Interfacing

iv

List of Figures

2.1 Energy Diagrams for Electron Tunneling ................ 4

2.2 Simplified STM .............................. 6

2.3 Frequency response curves for a modeled STM ............. 9

2.4 Proportional plus Integral Controller .................. 9

3.1 Vibration Isolation (a) Springs. (b) Heavy mass with damping 11

3.2 Piezoelectric materials- (a) Orthogonal bars. (b) Tube ........ 13

3.3 Leverage mechalfism for sample approach ................ 14

4.1 Circuit Interconnections .......................... 16

4.2 Current Detection Method ........................ 17

4.3 Graph of Equation 4.2 ........................... 17

4.4 Preamp circuit ............................... 18

4.5 Feedback Circuit .............................. 20

4.6 Ramp Circuit ................................ 22

4.7 High Voltage Regulator Circuit ...................... 23

4.8 Frequency Response of the Feedback Circuit .............. 24

5.1

5.2

5.3

5.4

5.5

5.6

5.7

5.8

5.9

5.10

Laser Interferometry Setup ........................ 26

Experimental setup ............................ 27

Current vs. Gap spacing ......................... 28

Step in a Machined Copper Surface ................... 29

Two Images Showing Scans in Opposite Directions ........... 29

Image of a Gold Monocrystal ....................... 30

Graphite Structure ............................. 31

HOPG Surface Image ........................... 32

Brightness Modulation of a HOPG Sample Image ........... 32

HOPG Lattice Image ........................... 33

V

Chapter 1

Introduction

1.1 History

The resolution limit of optical microscopes was reached decades ago. Since then

many other microscopes have been developed in an attempt to resolve even smaller

features; the Scanning Electron Microscope (SEM), the Transmission Electron Mi-

croscope (TEM), etc.. Now, with the advent of the Scanning Tunneling Microscope

(STM), the resolving power of scientific instruments has been advanced yet agaJ.n.

The STM was developed by researchers at IBM Zurich Research Laboratories

as a tool to "observe structures and compositions on an atomic scale"Ill. Binrdg,

Rohrer and Gerber first published their results on tunneling through a vacuum gap

in 198112]. In 1983, this same group published the first atomic resolution images

of the Si(111) surface[3]. The development of the STM earned the group a Nobel

Prize in Physics in 1986.

Since the first images in 1983, the STM has been modified and developed by

many research groups around the world. Its atomic resolution makes it an excel-

lent tool not only for surface imaging, but also for spectroscopy measurements and

determination of surface electronic structure[4]. The STM has also been used for

lithography[5][6] and has many possible future applications in semiconductor mate-

rials and device processing[7]. Recently, the STM has been used by researchers at

Lawrence Livermore National Labs to image DNA[8]

A number of good review articles have been written about the STM[9][10][ll];

the reader is referred to these articles for a more detailed history of the STM.

1.2 Project Outline

The objective of this project was to design and construct an STM. The design was

to be as flexible as possible, with any changes that may be necessary in the future

requiring a minimum amount of work. The STM built in this project will be used

for spectroscopy measurements of Si/SiO~ interface traps by other researchers.

Computer integration is necessary to store and retrieve data, and to facilitate

image processing of the results. The STM is to function on its own with a storage

oscilloscope for display, or under computer control.

1.3 Outline

Chapter 2 contains a brief review of the theory and background information neces-

sary in order to understand the principles of the STM. This chapter will summarize

some of the theories and present their results. Further information can be obtained

from the references provided.

Chapter Three will cover one of the most important aspects of the STM: the

mechanical components. It will outline problems and past solutions, as well as the

implementations in this project.

Chapter Four covers the electronics of the system. It will present each of the

major circuits involved and give a brief explanation of their function

Chapter Five includes the major results of this project. Results include proof of

a tunneling current, images of a machined metal surface and graphite.

1.4 Other Documentation

To keep the report concise, additional information will be included into other docu-

ments. First, a detailed description of all the circuits will be written for application

and debugging purposes. This will go into far greater depth than the descriptions

covered in Chapter Four. Also to be written will be a Users Manual for the opera-

tion of the STM. This will cover how to switch the STM on, practical details of tip

preparation, the correct method to move the sample to the tip, etc.

The computer integration was mostly performed by Raj Basudev, and his Senior

Project report details most of the software written and the interface hardware built.

Chapter 2

Background

The theory behind electron tunneling can be fairly involved. This chapter highlights

a few results of electron tunneling theories, pointing out their relevance to the

operation of the STM.

This chapter also briefly covers the basics of STM theory~ while further chapters

go into the details of the STM design.

2.1 Electron Tunneling

Clear discussions of electron tunneling can be found in many texts and papers on

quantum physics[12]-[15].

One-dimensional models of electron tunneling are reasonably simple. Figure 2.1a

shows the band diagram for two similar metals separated by a vacuum. As a result

of this vacuum gap, an energy barrier exists for electrons. However, there is a finite

probability of electrons tunneling through the barrier~ given by:

P o¢ e-~a (2.1)

where d is the barrier width, and c~ is equal to V/~-~(V-/~). For this specific

case~ ra is the mass of an electron m0~ and (V - E) is the barrier seen by an electron

with energy E. In most metals at room temperature~ few electrons are excited very

much above the Fermi level~ so it is possible to let E be equal to the Fermi energy

of the metal; therefore (V - E) is simply the work function ¢ of the material,.

Figure 2.1b shows two dissimilar metals separated by a vacuum. If there is no

applied bias~ their Fermi levels will be constant across the junction~ and a contact

potentiM will be created, ¢c. Equation 2.1 is still true, except now the barrier height

seen by the electrons can be approximated as the average of the two metal work

d

(a)

functions.

(b) (C) ""1""’"

~’~. e Vbias

Figure 2.1: Energy Diagrams for Electron Tunneling.

Figure 2.1c shows two different metals separated by a vacuum barrier but with

a bias voltage V~, applied. If the bias voltage is small, then the barrier height is

hardly changed. Now, the probability of tunneling in one direction is no longer equal

to the probability of tunneling in the other direction; the result is a net tunne]J.ng

current proportional to

where V~, is the applied bias between the two metals, A is a constant related to

a equal to 1.025 ]k-leV-{, ¢ is the average of the work functions of the two metals,

and , is the gap spacing. Equation 2.2 clearly shows the ohmic nature of the

tunneling gap. Typical work functions are of the order of 3-5 eV. Calculations have

shown that typical gap spacings for STM operation are of the order of 4-6/~[10].

Changes in gap spacing of 1 /~ can result in an order of magnitude change in the

tunneling current. This strong dependence of tunneling current on gap spacing: is

what makes the STM feasible. Clearly, having gap spacings on the order of 5 /~

between two materials is a challenging task. If however, an atomically sharp tip is

used as one of the materials, and the saznple is fiat, this becomes a realizable task.

Since one of the materials is now a tip, the tunneling theory of the STM becomes

much more complex. The tip of the STM can no longer be represented by a plane,

and the wave representation of the conduction electrons at the Fern~ level will

probably be more complicated than that of a simple plane wave. Therefore, solutions

to the problem of tunneling of electrons in the STM must take into account more

of the electronic structure of the surface.

The solution shown does give a good qualitative feel for the tunneling process

and the variables involved. However, a completely accurate picture of the process

involves a much more rigorous approach.

Bardeen[16] and Simmons [17] [18] have both tackled this problem by considering

the tunneling phenomenon as an overlap of the wave functions of two independent

systems. The results are complex, and I refer the reader to the papers mentioned

for further informatiom

2.2 STM Theory

The strong dependence shown in Equation 2.2 should make it straightforward to use

a feedback: loop to keep the current constant. If the tunneling current is compared

to a reference, it is possible to ensure a constant current by moving the tip closer to,

or further away from the sample as necessary. As the tip comes upon a perturbation

in the surface, the change in current causes feedback circuitry to move the tip as

requ.ired to return the current to its initial value. This signal used to drive the tip

movement is an indication of surface height. When scanned across the surface of

a sample, this signM can be plotted as a function of position giving a topographic

image of the surface.

It is important to note that the images obtained are not true surface profiles

(resulting from atomic nucleus locations), but rather images of equal electron state

density contours. This will be discussed later.

There are two basic modes in which the STM can operate: Constant Current

(CI) and Constant Height (CH). The last paragraph explained the STM in CI mode;

the current is kept constant at all times. This involves a relatively slow scan (20

30 Hz), and a fast feedback system~ so that it can respond to any perturbation on

the surface.

In CH mode, the average separation of tip and sample is kept constant (at

selected value). As the tip scans over the surface~ it does not follow every contour’ of

the surface, only the general slope of the surface. The current however, changes as

a result of every contour; this information can then be used to represent the surface

5

MovementServos

Tip

-- V bias

Sample

!

le~romcs

Figure 2.2: Simplified STM.

height. CH mode runs at much faster scan rates (500 Hz) using slower feedback.

2.3 Resolution

Several different researchers have done theoretical calculations of the resolution of

the STM[19][20]. The theoretical expression derived by Tersoff and Hamann can be

written as:

Here :k is the lateral resolution (smallest discernible lateral dimension), R is the

radius of curvature of the tip, s is the gap spacing, ¢ is the average barrier height,

and A is a constant equal to 1.025/~-1 eV-½. In order to get lateral resolution of

5/~, with typical gap spacings of 4-6 ]k, a tip radius of 10 _/k is necessary.

2.4 Tips

As seen in the last section, in order to get atomic resolution in the STM it is

necessary to have a tip with a radius of curvature less than 10 ~. The first STM

images were taken with a conventionally ground Tungsten wire. It seems difficult to

conceive that such a ground wire would have a radius of curvature less than .~0 ~.

It is theorized that what actually happens is that small asperities (or ’tiplets’) form

through which the tunnehng current flows. These ’tiplets’ are generally unstable,

but usually last long enough to get an image. Over the last 7 years, the process

of tip preparation has slowly progressed from an art to a science. There are many

approaches to tip preparation, some of which include[10]:

¯ crashing the tip into the surface and dragging it along the surface away from

the scanning area, retracting the tip and scanning over the desired area

¯ applying a high voltage to the tip to get a low field emmision current

¯ simply letting the system tunnel for an extended period of time before scan-

ning

Recently, electrochemical etching of tips has become popular[21]. Tips produced

in this manner, which appear sharp under a high power optical microscope, tend to

give atomic resolution almost every time[10].

2.5 Image Interpretation

Several papers have been written on STM image interpretation[10][22][23][24]. The

interpretation of the image is tied in very closely with the complex tunneling equa-

tions resulting from overlapping wave functions of two separate systems. The results

show that the tunneling conductance is proportional to the surface local density of

s~ates at the Fermi level. Since the STM operation involves the maintainance of a

constant current, the tip follows contours of constant local density of states (at the

Fermi level). STM images are therefore images of constant local density o/sgate.,; at

the Fermi level. Calculations of such contours have been performed on a number

of materials: Gold - Au(110)[23], Highly Oriented Pyrolytic Graphite - HOPG[24],

and GaAs(110)[193.

2.6 Control Theory

Since the STM is a system with feedback, the stability of the system is of great

concern. In control theory, one talks about a plant and a feedback controller. If we

consider the STM as the plant, which includes the mechanism for tip movement,

the tunneling junction, and the current detection circuit, the control requirements

7

of the system can be better understood (see Figure 2.3a)[25]. As will be seen in the

next chapter, the STM exhibits resonance at some frequencies; these are of great

importance in the model. The STM plemt can be modeled to have a frequency

response as shown in Figure 2.3b [25]; the plant is assumed linear with only the

lowest order resonance of interest. It should be noted that the resonance is a result

of complex poles, and has a 180° phase shift associated with it . The controller

must have:

¯ large gain for frequencies approaching DC;

¯ sufficient attenuation at the resonant frequency;

¯ a unity gain crossover at as high a frequency as possible.

These requirements ensure that there is little error in the system, there is no

oscillation in the system, and that the system responds reasonably fast. Figure 2.3c

shows the open loop response of the plant plus controller system for a non-oscillatory

system (solid line) and an oscillatory system (dashed line).

A proportional plus integral controller satisfies these requirements. The integral

controller has very high DC and low frequency gain, and the proportional controller

(which in our case includes one dominant pole) satisfies the other two requirements.

Figure 2.4a shows the block diagram of the proportional plus integral controller.

Figure 2.4b shows the amplitude response of the controller. The controller has the

transfer function:

where

C(s) Ax(1 + sK) (2.4)s(1 + spl)

Ap(2.5)K =Pl + A--;

By changing parameters A~,, Ax and pl one can change the shape of the frequency

response curve of the controller; if Av << Ax, we have a response which looks more

like an integral controller (right dotted line in Figure 2.4b). For A~ >> At

have a controller which looks more like a proportional controller (left dotted hne in

Figure 2.4b).

Perturbation

Signal

Plant

t Controller

(a)

Controller

S,,~~"~

~Plant

(b)Freq.

Gain

Unit~

(c)

Figure 2.3: Frequency response curves for a modeled STM.

¯

A~

l+s~

Amp

~ ~ Ap<< A I

A~ << Ap

(a) K (b) freq

Figure 2.4: ProportionM plus IntegrM Controller.

9

Chapter 3

Mechanical Designs

A well designed mechanical system is the basis around which the STM is built. T’he

mechanical system acts as the ’stage’ of the microscope. It contains most of the

non-electronic components of the system, and is used to hold both the sample and

tip. The mechanical system includes the mechanism for tip motion, isolation of

vibrations and coarse approach of the sample.

3.1 Tunneling Environment

The original STM designs were. all incorporated into an Ultra High Vacuum (UHV)

system. Images have since been reported in air[26], water[27], and nitrogen[28].

This ability to image in mediums other than in a vacuum is important. It is no

longer necessary to build an UHV system with all its associated problems, and it

will be easier to image biological molecttles in an aqueous solution.

The STM in this project was designed to be operated at room temperature in

air. This made the mechanical design much easier without the constraints of an

UHV system. However, the absence of an UHV system will make it impossible to

image samples which oxidize very quickly. Furthermore, there will be an added

problem of surface contaminants complicating the tip/sample interaction.

3.2 Vibration Isolation

Clearly, with resolutions of interest in the Angstrom range, any vibrations colrld

cause a great number of problems. In fact, the problem of vibrations led earIy :re-

10

searchers to believe that the STM might be technologically unfeasable[29]. In t~he

beginning, STMs were big and bulky with vibration isolation being a definite prob-

lem. To conquer this, researchers used elaborate systems of springs and magnetic

(eddy current) damping[I]. This was a very extensive setup, making it cumbersome

and difficult to fit into a vacuum system.

High frequency vibrations can be removed with a simple mass/spring setup.

The mass/spring system itself has a resonance which is related to the mass of the

STM and the characteristics of the springs. In most STM designs this occurs in

the 2 to 5 Hz range[25]. This is a resonant frequency at which the entire STM

moves up and down. The stiffness of the STM ’stage’ itself controls how much of

this vibration will result in tip to sample vibration. With a small stiff design, the

resonant frequencies tend to be much higher (in the 10 kHz range for STMs :made

from steel with dimensions the order of centimeters)j25].

Frame

MetalPlates Tygon

Tubing

(a) (b)Figure 3.1: Vibration Isolation (a) Springs. (b) Heavy mass with damping.

Our original system is shown in Figure 3.1a. Four springs held a heavy plate

on which the STM ’stage’ sat, separated by a ring of tygon tubing. This system,

however, introduced resonances at 12 Hz and 41 Hz due to overly soft and long

springs.. The system was therefore changed to the system in Figure 3.1b where mul-

tiple heavy plates are again separated by rings of tygon tubing[30]. No resonances

have been found yet with this damping system. The mechanical design was kept

small (approximately 10 cm by 10 cm) so that the resonant frequencies of the ’stage’

11

itself would remain high.

3.3 Tip Movement

Most of the dimensions discussed in the previous chapter have been on the order

of Angstroms. An actuation system is therefore required for precise movement of

distances less than an Angstrom. Piezoelectric materials fulfill this requirement

well. These materials work by coupling mechanical motion to dipole/electric field

interaction. In the piezoelectric material, all dipoles are originally lined up in one

direction, the direction of poling. External electric fields interact with these dipoles

causing stresses which result in mechanical displacement. The Linearity of these

materials is ideal for STM applications. If, however, the electric field applied is

large enough and opposes the direction of poling, it is possible for the dipoles to

be forced out of alignment; this is called depoling. This typically involves fields of

the order of 15 kV / inch[31]. This problem occurred on a number of occasions in

this project. It is possible to repole the piezoelectric material simply by applying

a large electric field in the original direction. This was accomplished by applying

1000V or more across the material. Heating the piezoelectric material also makes

depoling and repoling much easier.

In early STM designs, PZT (Lead Zirconium Titinate) bimorph bars were used.

Three bars connected orthogonally were used for x-, y- and z-direction motion.

However, for large displacements this arrangement becomes troublesome with mo-

tion in one direction coupling to motion in a perpendicular direction. Furthermore,

the resonant frequencies of such setups were as low as 5 kHz (perpendicular to the

sample surface)[32] limiting the bandwidth of the system.

Recently, PZT tubes have become popular for STM designs. In this case, the

piezoelectric material is in the shape of a tube about 1/2 in. long, 1/4 in. diazneter,

and 0.020 in. thick. It is coated with nickel inside and out. The outside contact is

divided into four sections, 90° apart. By applying appropriate voltages to these four

contacts, z-y motion can be performed. Applying a positive potential to one contact

with respect to the center causes the tube to bend in that direction. Applying a

positive potential to the center contact with respect to the outer contacts results in

an elongation of the tube.

The tube is used for several reasons. First, its resonant frequencies are much

higher than those of the bars (40 kHz perpendicular to the sample surface), allowing

12

Y

(a) (b)

Figure 3.2: Piezoelectric materials- (a) Orthogonal bars. (b) Tube.

for a larger bandwidth. Second, there is much less coupling of motion between

orthogonal axes[32]. The sensitivity of the tube is approximately 50/~ per volt in

all three directions.

Initially in this project, the circuits were designed such that an increasing posi-

tive voltage on the inner electrode with respect to the outer electrodes, caused the

tube to elongate (causing the tip to approach the sample). This added the feature

pulling the tip away from the material when the system was switched off. However,

this involved applying a voltage opposing the direction of poling, and the maximum

voltage that could be applied without depoling the tube was approximately 300V

(a throw of about 1.5 microns). It is desirable for the PZT to have as large a range

as possible; However, with this configuration, the larger the range of the PZT, the

more likely depoling will occur. This resulted in depoling the material on a number

of occasions.

To increase the range of movement of the tip, the circuits were redesigned so

that an increasing negative voltage on the inner electrode now causes the tube to

shrink (causing the tip to move further away from the sample). Since the applied

field is now in the direction of poling, it is not possible to depole the tube. The only

limitation is the high voltage power supply limit. However, it is now necessary to

separate the tip and sample mechanica~Lly before switching off the electronics, since

when the voltage on the PZT goes to zero, the tube elongates, possibly crashing

13

the tip into the surface.

3.4 Mechanical Approach

Since the PZT tube used has a range of only about 2/.tm, the mechanical system

also has to handle coarse approach. A mechanism for approaching the sample to

the tip had to be devised. This system also has to include relatively easy sample

removal and replacement.

The mechanism used in this project is based on a design by Andres Bryant[33];

it uses a system of levers to facihtate the approach mechanism. Figure 3.3 shows

the mechanical setup. Vernier X is used to make coarse approach adjustments. Fine

approach adjustments are made with Vernier Y. It is possible to bring the sample

and the tip to within l~m so that they are within the range of the PZT tube and

the electronics.

AI. Tubes

Coarse Fine

Figure 3.3: Leverage mechanism for sample approach.

At this point, the feedback electronics take over, and the mechanical system is

left undisturbed.

14

Chapter 4

Electronic Designs

This chapter describes the body of the STM: the electronics. The electronics en-

compass the feedback system, data acquisition, and tip positioning. There are five

ma~n circuits:

¯ Preamplifier : detects and amplifies the current signal to a level which can be

sent to the feedback circuit;

¯ Feedback circuit : the heart of the system;

¯ Ramp circuit : controls lateral movement of the tip;

¯ z and y offset : controls DC positioning for z and y directions;

¯ Computer interface hardware : facilitates data storage and retrieval.

The circuits are distributed in a series of rack mounted boxes (except the pream-

plifier which is on the STM ’stage’ itself). These boxes are interconnected with

coaxial cables (see Figure 4.1).

4.1 Preamplifier

One of the requirements of the system is to have as large a bandwidth as possible.

Since the impedance of the tunneling gap is high (on the order of 0.1-10 Mf~)[34],

any capacitance due to long lines from the STM stage to the controlling circuit

will result in a large RC time constant and consequent loss in bandwidth. To solve

this problem, it is advantageous to place a preamplification stage as close to the

15

Figure 4.1: Circuit Interconnections.

tunneling gap as possible. Also, since the current is on the order of 1-10 nA, it is

advantageous to raise these signals to enhance the signal to noise ratio.

To do this, a current to voltage conversion is necessary. Since the tunneling

current is an exponential function of spacing, it is advantageous to try to linearize

the measurement for small displacements. Logarithmic amplifiers can accomp]Ssh

this for large displacements, but they are notoriously slow, and have a limited signal

range of operation. Although the method used only linearized the exponential

dependancy for small displacements, it has other advantages.

The tunneling gap can be modeled by a variable resistance R,, where

(4.1)

This equation follows from Equation 2.2. Here Co is a proportionality constant

and c = 1..0254~[ ~-1].

If a voltage divider between Rt and a fixed resistor R,a is set up (see Figure 4,.2),

we can write

1: (4.2)+ ec(’-’o)

where so is the value of s for which Rt is equal to R,et.

An exact value of so is unknown, but given the range of typical tunneling resis-

tances, an Ro,t on the order of 1-2M~ is appropriate.

16

Tip ~ Vb

Rt(s) 1 I TunnelingSampl~

Rset iVm(s)

Gap

Figure 4.2: Current Detection Method.

Plotting this function shows how it is approximately linear for small deviations

around s0 (Figure 4.3). There are several advantages to this method: it is fast (there

is no delay added into the system), it can operate over a large range of currents,

and even if the tip touches the surface, Ro** limits the current.

V

Figure 4.3: Graph of Equation 4.2.

The circuit in Figure 4.4 shows the preamplifier circuit used. The b:iFET

(AD711) op-amp has bias currents a factor of 100 less than the tunneling currents

of interest.. The bias currents are therefore negligible compared with the measured

current.

The instrumentation op-amp (AD524) amplifies the buffered signal to fractions

of a volt level to avoid problems with noise. The gain of this amplifier can be

17

selected by the user.

This is the extent of the preamplifier; it is deliberately kept small since it is

mounted directly on the STM stage.

Sample]

Figure 4.4: Preamp circuit.

4.2 Feedback Circuit

The feedback circuit is made up of four parts:

¯ current comparison;

¯ proportional plus integral controller;

¯ high voltage amplifier;

¯ data output filtering.

The circuit design is based upon a similar design by Sang-il Park[35]. "]?he

redesign of a number of stages for this project was necessary. The most important

change comes as a result of using a PZT tube scanner rather than PZT bimorphs

as in the original design. Since only one contact is available for the z signal, both

the feedback information and the offset signal have to be combined in the circuit.

Also, a high voltage axnplifier for large negative voltages was designed. Another

important change is the inclusion of the integral controller. Since the electrorfics

are designed with flexibility in mind, other options are included; the selection of

CI/CH modes; feedback on/off; external input to V=; computer/manual control.

Figure 4.5 shows the circuit. The preamphfier signal comes in through coax

connector C1. Op-amp A1 buffers this signal and the instrumentation op-amp A2

compares the current signal with the current set point set by potentiometer IP1.

18

The output of A2 passes through the proportional (A3) plus integral (A4)

troller, the two portions being summed at op amp AS. Variable filtering is available

through potentiometer P4 and the switching in of filter capacitors ($7 - S10). The

filtering allows variable bandwidths between 15 Hz and 1 kiiz. The output of the

proportional plus integral controller passes to the high voltage amplifier (A9 and

Q1) with the option of turning feedback off (S1), using an external signal ($2),

having computer control (S3a and $4).

The offset position of the tip is controlled by potentiometer PS. Optional offset

input is available through connector C2. These two signals are stmmaed with the

instrumentation op-amp A6, and are added to the feedback signal at the high voltage

amplifier. The output of the high voltage amplifier then goes to the STM stage (C3).

Selection of CI or CH modes is done by switch $5. The selected signal then

gets buffered (A7), band pass filtered and buffered again with variable gain

instrumentation amplifier AS. The output is C4.

Selection of the bias voltage is controlled by potentiometer P6. Computer control

of the bias voltage is selected with switch S3b, before going to the STM stage (C5).

A more detailed circuit description can be found in the ’Circuit Description’

document..

There are two ways to display the information; direct view and intensity mod-

ulation.

For intensity modulation, the z and ~/scan signals are used as z and ~/inputs to

the oscilloscope. The z data is used as an intensity input to the oscilloscope, which

causes the signal to become brighter or dimmer resulting in different intensities for

different ’heights’ of the sample.

For the direct view, part of the z data signal is added to the !t scan signal going

to the oscilloscope. In this way, the height of the sample is displayed directly on

the oscilloscope screen.

A circuit was designed for this project that adds part of the z signal to the V

signal (to display the features), and adds part of the V signal to the z signal (to show

a more orthogonal view). It is not possible to determine from the images directly

the height of certain features; the z signal must be observed separately to determine

such feature sizes.

19

CI

AIIK

P6

Vblas

Computer

Interface

Integral

P2

IK 100K

Proportional

$60.Integral

off2.2uF

-15~0K A4

P3

5K

10K

5K

luF

$8 ~.F

$9 luF

SIO IuF

P4 .IuF

IK 100K --

A5

Manual

Filter

$2

S3a

$4-400

50K,

200K

Computer

Interface

C310K 2K

HTH2P45

4.3 Lateral Motion

When considering lateral motion, both an offset (DC level) and scan (’ AC’ signal)must be applied. The DC level positions the tip at a certain location, and the ’AC’

signal then scans the area. Using the PZT tube, with four quadrants around ~he

outside, it is a simple matter of using opposite quadrants to connect ’AC’ and DC

signals for a given z or y direction.

4.3.1 Ramp Circuit

The ramp circuits are designed for this project so that a wide range of scan frequen-

cies are available to permit both CI and CH modes of operation. Also, a triangle

waveform was selected in preference to a sawtooth waveform since the feedback may

not be able to respond we]] to surface features with a very fast retrace. Scanning

is only done in one direction, since scans in alternate directions produce slightly

different traces, making the results difficult to interpret.

The circuit used (Figure 4.6) is standard[36]. Op-amp A1 integrates a :~-10V

square wave produced by op-amp A3. Op-amp A2 compares the square wave with

the integral output causing the square wave to change polarity when A1 integrates

to =kl0V. Potentiometer P1 controls the frequency of the scan with switch $1 se-

lecting high or low frequency.scanning ranges. Op-amp A4 in conjunction with

potentiometer P2 controls the amplitude of the output waveform. Connector C1

can be used when the computer is used to control the z and y scans; this is selected

with switch $2. Connector C2 is used for displaying the information on the oscil-

loscope screen. Op-amp A5 buffers the signal which is filtered before going to the

PZT tube.. Switch $3 turns tip scanning on and off.

4.3.2 Offset Circuit

To view as large an area as possible, the offset circuit should have a large range;

this is accomplished using high voltages. Unfortunately, most high voltage power

supplies are designed for a relatively large ,current supply as well. This results

in poor voltage regulation, and large amounts of ripple. With sensitivities of 50

_~/volt on the PZT tube, ripples of 50mV can cause tip movements of 2.5/~ mak:ing

atomic resolution imaging impossible. Therefore, a high voltage regulator is used

to condition the offset voltages. The design (Figure 4.7) is based on a circuit used

21

(..) - Y scanner components

PI.10K~,

5O(

Frequency 6.8n~ / ]N400] +.1.~Low 57K (luF) ., 1.6KFreq. (100K) / ,

~[ ~

1N4OO1

I-(2OOK) IN4001 1.6KlooK

IK (50K) Amplitude -15

20K ~’ S2 5’ Manual 1K 5K S.,3

~~ u~ V (PZT)

- OscilloscopeComputer (~ C1 C?.~In --- .- Out

Figure 4.6: Ramp Circuit.

22

IT1 1N4004 R2 I

+500 V

luF~

50OMeter ~

11. o Out

Figure 4.7: High Voltage Regulator Circuit.

by Sang-il Park in his STM design[35].

An LM317 regulator (R1) is used since it is specifically designed for high voltage

regulation.. The transformer (T1) and 7815 regulator (R2) ensure that the regulator

R1 and power transistor Q1 are always in the forward active state. The LM317

develops a fixed 1.25 volts across its output pins (pins 2 and 3). Since negligfble

current flows into pin 3 of the LM317, potentiometer P1 controls the current flow

into potentiometer P2. The voltage drop across P2 (plus the fixed regulator output

voltage) is used as the offset voltage for the z and y directions.

4.4 Circuit Performance

Due to power supply limits, noise and other electrical anomalies prevalent in analog

circuit designs~ there are certain limitations to the use of these circuits.

As mentioned in Section 4.1, the maximum tunneling current limitation is set by

the bias voltage applied to the tip. Since the STM is typically run at a bias voltage

of 100 mV, the maximum tunneling current is 100 nA. Electrical noise on the power

supply lines also appears on the preamplifier output lines. The maximum noise is

typically around 5 mV which corresponds to a current of 0.05 nA. This current is

equivalent to a 1.46.~ change in gap spacing at a steady state current of I nA. This

sets the lower electronic limit on the resolution perpendicular to the surface of the

STM.

Similarly, the output high voltage amplifiers which drive the PZT also have

23

associated with them a certain amount of noise. 5 mV of noise appears on the

z, ~ and z signal hnes. This corresponds to a movement of 0.25 ~. This sets the

electronic lateral resolution of the STM.

The range of the z, ~, and z motion of the tip is limited by the high voltage

amplifier. The maximum range in the z direction is about 1.6 gin. This is limil~ed

by the breakdown voltage of the p-channel MOSFET used. The z and V (offset)

limits are set by the high voltage supply since the M J10011 breakdown voltage is

well over 1000 V. At present, the range of the z and V (offset) motion is about it v,m.

The scanning bruits are set by the circuit components used. The maxi4.0 scan area

is about 1300 .& by 1300 ]k. The minimum area is 5 ./k (z) by 11 X (V). Scanning

frequencies range from 1.4 to 300 Hz for z scans and 0.07 to 175 Hz for V scans.

Figure 4.8 shows the frequency response of the feedback circuit for various setting

of the gain and integrator time constant.

Controller Frequency Response

Low Frequency roll-off

2030 ~/// of oscilloscope.

-~o

E< -20

-30 ........ , ¯ . .1 10 100 1000 10000

FrequencyFigure 4.8: Frequency Response of the Feedback Circuit.

=0

24

Chapter 5

Results

This chapter reports some of the results obtained during the design of the STM.

Some images of samples are also presented.

5.1 Calibration of the Tube Scanner

Calibration of the piezoelectric tube was done to see what motion the tip would have

for applied voltages. The tip was calibrated using a laser interferometry setup (see

Figure 5.1). A mirror was mounted on one end of the tube, and a 0 - 500 V r~mp

was applied to the outside with respect to the inside. The movement of the mirror

resulted in interference patterns on a detector. The spacing between two maxima

or minima equates to a displacement of one wavelength of the laser used. This

experiment established a calibration of 64 ~/vo]t in the z direction. Calibration of

the z and y directions was not performed since the miror could not be mounted on

the side of the tube.

5.2 Exponential Dependency

It is possible to establish the existence of a tunneling current. This can be done by

showing the exponential dependence of the current on the tip to sample spacing.

Figure 5.2 shows the setup used with a graphite sample. A function generator

was used to obtain a 100 mV peak-to-peak bias voltage with a 100 mV DC level.

The tunneling current was detected using a Lock-In Amplifier. An external tip

position signal was applied to the feedback electronics, which resulted in a controlled

25

Laser

Detector

Intensity

Silvered Mirror ZTMirror

/

TubeScanner

Plotter ~

Voltage

PowerSupply

Figure 5.1: Laser Interferometry Setup.

z displacement of the tip. This tip position signal was plotted against the output

of the Lock-In Amplifier. Data was then read from the curve and replotted on a

logarithmic scale.

From Equation 2.2, the slope of the graph is expected to be equal to

Figure 5.3 shows the results obtained. It clearly shows an exponential dependence

of current on spacing. The slo.pe of the graph is approximately equal to 1.71/~-1.

This corresponds to a ¢ =2.8 eV. The actual average of the work functions of carbon

(5 eV) and tungsten (4.5 eV) is about 4.75 eV[37]. The result is consistent

other researchers[2][34] who found that the measured values of ¢ were typically

to 2 eV lower than the clean metal work functions.

5.3 Sample Images

The STM typically operates at a 1 nA tunneling current and a 100 mV tip bias

voltage (tip positive). All the images in this section were taken at this operating

point. Images are displayed on a storage scope. A circuit designed for this project

that adds part of the z signal to the y signal (to display the features), and adds

part of the y signal to the z signal (to show a more orthogonal view) is used

display the images. It is not possible to determine from the images directbr the

height of certain features; the z signal must be observed separately to determine

26

PZT

Vbias

I Function ~

Generato~

Reference

Tunneling Current

Electronics

Tip Position (z)

Lock-InAmplifier

Output J

Plotter

feature heights.

Figure 5.2: Experimental setup.

5.3.1 Machined Surface

As a first test for the imaging capabilities of the STM, a sample with gross surface

features was used. The san~p, le holders seemed an ideal choice, since they are a

machined copper surface with an abundance of scratches. The sample holder was

cleaned and had a 2200/~ layer of gold deposited on it to prevent the copper from

oxidizing.

Many striking surface features were detected. Figure 5.4 shows a typical image.

The oscilloscope grid represents 50/~ per division for the z and y scales. The step

is approximately 150A high.

To indicate the importance of only displaying a scan in one direction, Figure 5.5

shows images taken one right after the other. The grid in these images is 25 ~ per

division in both directions. The image on the right was taken scanning from the

bottom to the top, and the image on the left was take scanning from the top to the

bottom. It is important to note that all the images scanned in the ~ame di~ction

looked the same, but images scanned in opposite directions may not look the same.

This is probably a result of the feedback electronics reacting differently to a sudden

drop of[ or a sudden wall.

27

lOO

.1

Tunneling Current vs. Gap Spacing

I I I

Gap Spaclng (A)

Figure 5.3: Current vs. Gap spacing.

5.3.2 Gold Monocrystal

A gold monocrystal epitaxially grown on a sodium chloride substrate was also im-

aged. The gold layer (approximately 440/~ thick) floated onto a graphite substrate

and annealed for 30 minutes at.400° C. The graphite substrate was then mounted on

a sample holder with silver paint. This sample had far fewer gross surface features.

Figure 5.6 shows a typical image. The oscilloscope grid is 200A per division for

both z and y directions. The large left and smaller right steps are approximately

50/~ and 25/~ high respectively.

5.3.3 Graphite

Highly Oriented Pyrolytic Graphite (HOPG) is a popular material for STM imaging.

The material is easily cleaved with a piece of adhesive tape leaving an atomically

flat surface.

The lattice pattern for graphite is shown in Figure 5.7[33]. Typical STM im-

ages show an array of atomic bumps corresponding to locations marked A. These

hexagonal rings of bumps are joined by ’bridges’ passing through locations marked

B. C locations result in a ’pit’ in the surface structure.

28

Figure 5.4: A step in a machined copper surface. Grid is 50/~ per division. Step

height is 150

Figure 5.5: Two images showing scans in ,opposite directions, (1) scanning

bottom to top, (r) scann.ing from top to bottom, Grid is 25

division.

29

Figure 5.6: Image of a gold monocrystal, Grid is 200 ~ per division. ,Left step is

50 ~ and right step is 25 ~.

Getting the atomic resolution was a very painstaking process, the most impor-

tant step of which was the tip preparation. A previously damaged tip was electro-

chemically etched in a 4 % KOH solution at 1 mA for 25 seconds. This resulted !in

a clean tip (though not as sharp as the new tips when viewed through an opticM

microscope). The resulting image is shown in Figure 5.8. Here, the z and V scale

is 1.25 ./k per division. There are distinguishable rows visible with a separation of

about 2.5 ~. CMculations of constant density of state contours indicate that rows

should occur with a spacing of 2.46 ]k. It may be possible to cMibrate the PZT tube

scanner in the z and V directions using these rows, but a better image must first be

obtained, to do this. Figure 5.9a shows the same area with brightness modulation.

This photograph is enlarged to better point out the features. Figure 5.9b shows a

schematic of the hexagonal ring in the lower central portion of the photograph. The

dimensionM ratio a : b is approximately equal to 0.58. This corresponds well to the

expected ratio of dimensions expected as shown in Figure 5.7 which is 0.577.

In Figure 5.10 it is possible to discern the hexagonal pattern. An hexagonal

overlay is provided to point out the hexagonal array. The difficulty in identifying

the structure may possibly be due to a ’dragging effect’ described by Park[35]. This

dragging effect is a result of hysteresis between the actual tunneling point on tlhe

tip and the z motion of the tip. Vertical lines of ~bumps’ are clearly visible. This

30

1.42 I st layer

2nd layer

Figure 5.7: Graphite Structure.

image was taken with a new tip that was etched in 4 % KOH at 0.5 mA for IL0

seconds to remove surface oxide on the tip.

Figure 5.8: HOPG surface image. Grid is 1.25 It per division. Height is approxi-

mately 4

Figure 5.9: Brightness modulation of a HOPG surface. 2.5 i per division.

32

Figure 5.10: HOPG lattice image. Grid is 1.25/~ per division. Overlay shows the

expected hexagonal pattern.

33

Chapter 6

Conclusions and Future Work

This re:port outlined the work done during the design and construction of a Scan-

ning Tunneling Microscope. It briefly covered some background theory necessary

for STM understanding; it covered the mechanical and electronic designs; and. it

presented results of the observation of a tunneling current, as well images of different

surfaces.

During the course of the project, the electrical and mechanical components were

designed, built, tested, modified and retested until the STM operated as required.

Over the course of the project, resonant ~requencies were rejected, noise was elim-

inated, vibrations were damp.ed and results improved. The results of Chapter 5

show the results of the functionality of the STM.

The field of Scamling Tunneling Microscopy is still f~irly new, and more appl~-

cations will undoubtedly be thought of for this tool. As applications come tbrth,

this STM will easily be enhanced by future improvements and additions.

6.1 Future Work

As use of this tool progresses, problems and possible improvements will present

themselves. The STM was designed so that these changes would be as straightfor-

ward as possible.

There are some aspects of the STM and its design that have to be addressed

very soon. This work is important if the STM is to be used as an effect research

tool.

First and foremost, work is required on tip preparation. This is the one area

that is of great importance to the performance of the STM. Very httle time was

34

spent on tip preparation during this project, and it must be mastered to obtain

reliable results.

Also requiting work is the computer interface. The computer interface has :not

been fully tested or used. Some foreseeable problems exist due to the speed at

which the computer interface operates. Some prograzn optimization will have to

be done to speed up the functioning of the interface. Also, some hardware glitdhes

have been noticed that have been temporarily bypassed. These do not affect the

STM when doing a ’Constant Height’ scan, but are important if the interface is to

be used for spectroscopy measurements.

The mechanical design also requires some improvements. At present there is no

provision for moving the sample once it is mounted. The z and V offset signals only

have a :range of about 2 ~m, this means that is a particular feature is to be imaged,

it must be mounted on the sample holder at the correct spot. This is very difl%ult

to do. Some mechanical design to do coarse movement of the sample would be: of

great benefit.

These are major points that wi].l need work. Other minor changes include: build-

ing all the circuits on a printed circuit board; including a switch so that negative

bias voltages can be applied; allowing external control of the output data bandpass

filter. These changes are not essential, but would enhance the flexibility of the STM.

35

Appendix A

Computer Interfacing

For practical use as a research tool, the STM must be able to store and recall data

and images. Computer interfacing will not ouly facilitate this, but also automate the

operation of the STM. For this reason, an HP9836 microcomputer was interfaced

with the STM.

In the automated system, the computer controls most functions usually con-

trolled by the user. It is also desirable to have ~partial automation’, where only

certain operations are controlled by the computer. In particular, the computer

must be able to:

scan the tip in z and y ~rections;

¯ do precise z and y positioning to be able to move to one particular spot;

¯ control the tip bias voltage;

¯ control the z positioning of tip;

¯ read and store data.

Combinations of these options will allow a variety of different functions:

1. obtaining topographic images while scanning (this information can then be

stored and the data image processed at a later time);

2. having complete computer control of the feedback system;

3. ramping the tip bias at a specific spot to get spectroscopic information.

36

In order to achieve this, l~aj Basudev designed digital circuitry to handle the

computer interface handshaldng and signal generation and aquisition. To monitor

the tunneling current/topography information, a 12 bit analog to digital converter

is used. For the bias voltage and z, y ~z z outputs, 10 bit digital to analog converters

are used.

The computer allows the user to select the precision of z and y scans (number

of bits used) as well as the range. For example, when doing a scan over a range of

500 /~, a precision between steps of 0.25 /~ is unnecessary and very slow; whereas

for a scan of 10 ~, 0.25/~ precision steps is quite appropriate.

The computer can function in one of three principle modes:

1. constant current;

2. constant height;

3. tip ramp.

In constant height mode, the computer provides z and y scan signals, and stores

the output, of the feedback circuitry. It plots this information on the terminal in

gray-scale format.

In constant current mode, the computer receives the actual tunneling current

information, and uses it to prqvide a z output signal which drives the PZT. It also

provides z and y scan signals, and plots this along with the z-output signal it sends

to the STM. In this mode, the analog feedback circuitry is essentially disabled.

Tip ramp allows the user to ramp the bias voltage on the tip while monitoring

the tunneling current signal. The user indicates the range of the ramp sigh’d, as

well as the precision of the ramp. This information is plotted on the monitor and

stored. The feedback circuitry is temporarily disabled during this operation.

Further functions on the computer allow calibration of the A/D and D/A con-

verters and some limited image processing

A full description of the hazdware can be found in Raj Basudev’s Senior Project

Report (Spring 1989).

37

Bibliography

[2]

[3]

[4]

[5]

[6]

[7]

[8]

[9]

[10]

G. Binnig and H. Rohrer. Scanning tunneling microscopy. Helv. Phys. Acta,

5~, 726, (1982).

G. Binn.ig,H. Rohrer, Ch. Gerber and E. Weibel. Tunneling through a control-

lable vacuum gap. Appl. Phys. LetS. 40, 178, (1982).

G. Binnig, H. Rohrer, Ch. Gerber and E. Weibel. 7 X 7 reconstruction on

Si(11.1) resolved in real space. Phys. Rev. Lett. 50, 120, (1983).

W. Kaiser and R. Jaklevic. Spectroscopy o] electronic states o] metals with a

scanning tunneling microscope. IBM J. Res. Develop. 30,411, (1986).

M. McCord and R. Pease. JLithography with the scanning tunneling microscope.

J. Vac. Sci. Technol. B4, 86, (1986).

M. Ringger, H. Hidber, R. Schloegl, P. Oelhafen and H. Guentherodt. Nanome-

ter lithography with the scanning tunneling microscope. Appl. Phys. LetS., 46,

832, (1985).

N. Lipari. STM applications for semiconductor materials and devices. Surf.

Sci., 181,285, (1987).

T. Beebe, T. Wilson, F. Ogletree, J. Katz, R. Baldhorn, M. Salmerson and

W. Siekhaus. Direct observation of native DNA stuctures with the tcanning

tunneling microscope. Science, 243,370, (1989).

C. Quate. Vacuum tunneling: a new technique .Cot microscopy. Physics Today,

~9, 26, (1986).

P. Hansma and J. Tersoff. Scanning tunneling microscopy. J. Appl. Phys. 61,

R1, (1987).

38

[11]

[12]

[13]

[14]

[15]

[161

[17]

[18]

[19]

[20]

[21]

[22]

[23]

[24]

G. Binnig and H. Rohrer. Scanning tunneling microscopy. Sci. Am. 253, 50,

(1985).

W. Harrison. Tunneling from an independent-particle point of view. Phys.

Rev. 123, 85, (1961).

C. Turner T. VanDuzer. Principles of Superconductive Devices and Circuits.

Elsevier, New York, 1981.

W. Harrison. Solid State Theory. Dover Publications, Inc., New York, 1979.

L. Schiff. Quantum Mechanics. McGraw-Hill, New York, 1968.

J. Bardeen. Tunneling from a many-particle point of view. Phys. Rev. Lett.. 6,

57, (1961).

:I. Sinmaons. Generalized ]ormula .for the electric tunnel effect between similar

electrodes separated by a thin insulating film. 3. Appl. Phys. 84, 1793, (1963).

:1. Simmons. Electric tunnel effect between dissimilar electrodes separated by a

thin insulating film. J. Appl. Phys. 34, 2581, (1963).

~I. Tersoff and D. Hamaan. Theory of the scanning tunneling microscope. Phys.

Rev. B, 31,805, (1985).

E. Stoll. Resolution of the scanning tunneling microscope. Surf. Sci. 143, L4:ll,

(1984).

P. Bryant, H. Kim, Y. Zheng and R. Yang. Technique for shaping scanning

tunneling microscope tips. Rev. Sci. Instrurn. 58, 1115, (1987).

N. Garcia, C. OcM and F. Flores. Model theory .for scanning tunneling rai.

croscopy: application to Au(110) (lX~). Phys. Rev. Lett. 50, 2002, (1983).

:I. Tersoff and D. Hamann. Theory and application .for the scanning tunneling

microscope. Phys. Rev. Lett. I~0, 1998, (1983).

A. Selloni, P. Carnevali, E. Tosatti and C. Chen. Voltage-dependent scanning.

tunneling microscopy of a crystal surface: graphite. Phys. Rev. B, 31, 2602,

(1985).

39

[25] D. Pohl. Some design criteria in scanning tunneling microscopy. IBM J. Res.

Develop. 30,417, (1986).

[26] S. Park and C. Quate. Tunneling microscopy of graphite in air. Appl. Phys.

Lett. 48, 112, (1986).

[27] R. Sonnenfeld and P. Hansma. Atomic-resolution microscopy in water. Sci-

ence, 232,211, (1986).

[28] R. Sonnenfeld and B. Schardt. Tunneling microscopy in an electrochemical

cell: images of Ag plating. Appl. Phys. Lett. 49, 1172, (1986).

[29] R. Young, J. Ward and F. Scire. The topografiner; an instrument for measuring

surface microtopography. The Rev. Sci. Instrum. 43, 999, (1961).

[30] T. Albrecht. Private Communication, Stanford university, 1989.

[31] R. Petrucci. Private communication, EBL Company, 1988.

[32] D. Smith. New Applications of Scanning Tunneling Microscopy. PhD t:hesis,

Stanford University, 1987.

[33] A. Bryant. High Speed Imaging with the Scanning Tunneling Microscope. PhD

thesis~, Stanford University, 1986.

[34] S. Elrod. Low Temperature Tunneling Microscopy. PhD thesis, Stanford Uni-

versity, 1985.

[35] S. Park. Scanning Tunneling Microscopy for Surf. Sci.. PhD thesis, Stemford

University, 1986.

[36] G. Tobey, J. Graeme and L. Huelsman. Operational Amplifiers. McGraw-Hill

Books Inc., New York, NY, 1971.

[37] R. Weast and M. Astle, Editors. CRC Handbook of Chemistry and Physics.

CRC Press Inc., Boca Raton, FI., 1981.

4O