SCANNING PROBE MICROSCOPY OF PROTEIN NANOWIRES
By
Kathleen Ann Walsh

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
Physics – Doctor of Philosophy
2013

ABSTRACT
SCANNING PROBE MICROSCOPY OF PROTEIN NANOWIRES
By
Kathleen Ann Walsh
The bacterium Geobacter sulfurreducens grows electrically-conductive pili, which act as
protein nanowires, in order to transfer electrons from the cell to electron acceptors in its
environment when direct charge transfer through the cell membrane is not feasible. Understanding the electronic structure of the pili can provide insight into fundamental processes
of electron transfer in biological systems.
This study investigated the electronic structure of these protein nanowires using the
toolbox of scanning probe microscopy, specifically scanning tunneling microscopy and point
tunneling spectroscopy. These measurements were performed at 77 K and at room temperature. The measured data are compared to theoretical calculations.
Density of states measurements using tunneling spectroscopy show that these pili act as
narrow-gap biological semiconductors at 77 K. The onset of nonzero density of states remains
within the metabolically-relevant voltage range. At room temperature, spectroscopy of the
pili retains a gap-like structure, but this pseudogap is raised to a nonzero density of states
at even the smallest applied voltages. These pilus nanowires also exhibit a distinct spatial
dependence of the density of states across the breadth of the pili.

Copyright by
KATHLEEN ANN WALSH
2013

To Mom and Dad—I wouldn’t be here without you.

iv

ACKNOWLEDGMENTS

My heartfelt gratitude goes out to Mrs. Kilgore, the sine qua non of my scientific career.
She stayed after school two evenings every week, offering help to any of her Middle Math II
students who wanted to take her up on it and tutoring such students through topics again
and again until we could succeed. Without her patient assistance in seventh grade, my life
today would be radically different.
I heartily appreciate my friends and colleagues who have made my graduate school experience more cheerful: Jenna, Megan, Sean, Emily, Emi, Carla, Brent, et al. Most timely
thanks goes to my “thesis buddy” Carla for helping me stay on track lo, these many months.
Thanks for the good times, and sorry about the rants.
On the colleague side, Sanela Lampa-Pastirk, in addition to purifying sample for me,
was always willing to help me navigate interdisciplinary research and to encourage me not
to give up on the ideal of rigor. Josh Veazey was a fountain of insight and perspective. Josh
and Sanela laid the groundwork for everything I’ve done with Geobacter. Sean Wagner in
the UHV-STM lab next door (and the desk next to mine in the office) taught me STM and
has been a great colleague and friend ever since. Nathaniel Leonard generously let me test
GSH mode on his samples and worked with me on them. Thanks also go to Scott MacLaren
for GSH Mode collaboration (including the name for it) and other wonderful things.
I’ve been fortunate enough to have several great research mentors. Dr. Elizabeth Behrman
was a superlative undergraduate advisor. She convinced me that “words people” can succeed in physics and showed me how to do research. My REU with Norman Birge made it
abundantly clear that I wanted to be an experimentalist. Alexandra Gade told me, “You
should do something that makes you happy” and encouraged me to go do it.
v

Stuart Tessmer has been the best graduate advisor I could hope for. Seriously, it’s like
working for Mister Rogers. I’ve learned from him not just how to be a better scientist but
how to be a better person. Stuart’s example proves that being nice all the time is a viable
leadership style, one I hope to emulate.
Most CMP experimentalist grad students here thank Reza Loloee, and for good reason.
He’s always willing to help, is tremendously clever and useful, and genuinely cares about the
students. Reza makes sure the basement stays collegial and doesn’t get too grim.
Lastly and mostly, I am immensely fortunate to have parents who understand me and
are also my friends. Thanks, Mom and Dad.

vi

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

x

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xi

Chapter 1

Overture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

1

Chapter 2 Scanning Probe Microscopy . . . . . . . . . . . . . . .
2.1 Key Aspects of Scanning Probe Microscopy . . . . . . . . . . .
2.1.1 Overview . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1.2 STM and AFM Tips . . . . . . . . . . . . . . . . . . . .
Tip Properties . . . . . . . . . . . . . . . . . . . . . . . .
Tip Artifacts . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Scanning Tunneling Microscopy . . . . . . . . . . . . . . . . . .
2.2.1 Review of Tunneling Through an Energy Barrier . . . . .
2.2.2 Topography and Local Density of States Probed by STM
2.3 Point Tunneling Spectroscopy . . . . . . . . . . . . . . . . . . .
2.3.1 Density of States . . . . . . . . . . . . . . . . . . . . . .
2.3.2 Local Density of States Probed by Point Spectroscopy .
2.3.3 Interpretation of dI/dV Curves . . . . . . . . . . . . . .
2.4 Atomic Force Microscopy . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

9
10
10
11
11
13
13
14
18
20
21
23
24
26

Chapter 3 Details of Experimental Apparatus
3.1 STM Hardware Configuration . . . . . . . . .
3.1.1 Besocke Design . . . . . . . . . . . . .
3.1.2 Tip Specifics . . . . . . . . . . . . . . .
3.1.3 Vibration Isolation . . . . . . . . . . .
3.2 STM Calibration . . . . . . . . . . . . . . . .
3.3 Cryogenic Temperature . . . . . . . . . . . . .
3.4 Cypher Scanning Probe Microscope . . . . . .
3.4.1 AFM . . . . . . . . . . . . . . . . . . .
3.4.2 STM . . . . . . . . . . . . . . . . . . .
3.5 Overview of Sample Preparation . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

Chapter 4 Biological and Theoretical Context . . . . .
4.1 Pili Expressed by Geobacter sulfurreducens . . . . . . .
4.1.1 Pilus Structure . . . . . . . . . . . . . . . . . .
4.1.2 Implications of Pilus Structure for Experimental
4.2 Charge Transport Mechanisms in Proteins . . . . . . .
4.2.1 Tunneling . . . . . . . . . . . . . . . . . . . . .
4.2.2 Temperature-Dependent Charge Transport . . .

vii

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.

29
29
29
32
33
34
36
37
37
38
39

. . . .
. . . .
. . . .
Design
. . . .
. . . .
. . . .

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

.
.
.
.
.
.
.

41
42
42
45
46
48
48

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

50
52
54
56

Chapter 5 Results and Interpretation . . . . . . . . . . . . . . . .
5.1 Topography of G. sulfurreducens Pili . . . . . . . . . . . . . . . .
5.1.1 Topographic Characteristics of Pili . . . . . . . . . . . . .
5.1.2 Pilus Topography at 77 K . . . . . . . . . . . . . . . . . .
5.1.3 Structural Longevity of Pilus Fibers . . . . . . . . . . . . .
5.2 Room Temperature Spectroscopy . . . . . . . . . . . . . . . . . .
5.2.1 Positional Dependence of Room Temperature Spectroscopy
5.2.2 Reproducibility of Room Temperature Spectroscopy . . . .
5.3 Spectroscopy at 77 K . . . . . . . . . . . . . . . . . . . . . . . . .
5.3.1 Salient Features of Cryogenic Spectroscopy . . . . . . . . .
5.3.2 Positional Dependence of Cryogenic Spectroscopy . . . . .
5.3.3 Reproducibility of Cryogenic Spectroscopy . . . . . . . . .
5.3.4 Comparison to Room Temperature Spectroscopy . . . . . .
5.3.5 Effect of Thermal Broadening . . . . . . . . . . . . . . . .
5.4 Results in Context . . . . . . . . . . . . . . . . . . . . . . . . . .
5.4.1 Comparison to Theoretical Calculations . . . . . . . . . .
Calculation Parameters . . . . . . . . . . . . . . . . . . . .
Calculated Density of States . . . . . . . . . . . . . . . . .
Discussion of Theoretical Calculations . . . . . . . . . . .
5.4.2 Biological Relevance . . . . . . . . . . . . . . . . . . . . .
5.4.3 Potential Further Investigations . . . . . . . . . . . . . . .
Role of Tyrosines . . . . . . . . . . . . . . . . . . . . . . .
Higher-Resolution Imaging . . . . . . . . . . . . . . . . . .
Temperature Dependence . . . . . . . . . . . . . . . . . .
Electrode Fabrication . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

59
60
60
63
68
69
71
72
73
74
76
78
78
79
82
82
82
83
84
87
88
89
89
91
91

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Appendix A Operation of an Atomic Force Microscopy User Facility
A.1 Installation and Characterization . . . . . . . . . . . . . . . . . . . .
A.2 Technique Development . . . . . . . . . . . . . . . . . . . . . . . . .
A.3 User Training and Troubleshooting . . . . . . . . . . . . . . . . . . .
A.3.1 User Training . . . . . . . . . . . . . . . . . . . . . . . . . . .
A.3.2 Troubleshooting and Maintenance . . . . . . . . . . . . . . . .
Appendix B Artifacts That Almost Thwarted Me . . . . . . . . . . . .
B.1 Step Edges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B.2 Vertebral Shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
B.3 Tendriloid Ripples of the Graphite Surface . . . . . . . . . . . . . . .
B.4 Other Topographic Snares . . . . . . . . . . . . . . . . . . . . . . . .
B.4.1 Graphite Superlattice . . . . . . . . . . . . . . . . . . . . . . .
B.4.2 Multiple-Tip Effects . . . . . . . . . . . . . . . . . . . . . . .

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.

93
94
94
95
102
102
103
106
107
109
110
112
112
114

4.3
4.4

4.2.3 Marcus Theory . . . . . .
4.2.4 Experimental Signatures .
Molecular Dynamics Calculations
Density Functional Theory . . . .

.
.
.
.

.
.
.
.

viii

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

.
.
.
.

B.5 Artifacts in Spectroscopy . . . . . . . . . . . . . . . . . .
B.6 Multi-Characteristic Artifact . . . . . . . . . . . . . . . .
Appendix C Sample Preparation . . . . . . . . . . . . . . . .
C.1 Sample Substrate Mounting . . . . . . . . . . . . . . . .
C.2 Sample Deposition . . . . . . . . . . . . . . . . . . . . .
C.2.1 Deposition Protocol Tuning . . . . . . . . . . . .
C.2.2 Spatial Density of Deposited Pili . . . . . . . . .
C.2.3 Reproducibility of Deposited Samples . . . . . . .
C.2.4 Contact Between Pili and Substrate . . . . . . . .
Appendix D Further Details . . . . . . . . . . . . . . . . . . .
D.1 Feedback . . . . . . . . . . . . . . . . . . . . . . . . . . .
D.2 Piezoelectric Rastering . . . . . . . . . . . . . . . . . . .
D.3 Image Processing . . . . . . . . . . . . . . . . . . . . . .
D.4 Direct and Numerical dI/dV Determination . . . . . . .
D.5 Tip Characterization . . . . . . . . . . . . . . . . . . . .
D.6 AFM Phase Imaging . . . . . . . . . . . . . . . . . . . .
D.7 Conductive AFM . . . . . . . . . . . . . . . . . . . . . .
D.8 STM Topography of Pilus Ends . . . . . . . . . . . . . .
Appendix E Microfabrication of Electrodes . . . . . . . . .
E.1 Goal and Starting Point . . . . . . . . . . . . . . . . . .
E.2 Photolithography . . . . . . . . . . . . . . . . . . . . . .
E.2.1 Materials . . . . . . . . . . . . . . . . . . . . . .
E.2.2 Procedure . . . . . . . . . . . . . . . . . . . . . .
Chip Cleaning . . . . . . . . . . . . . . . . . . . .
Ultraviolet Exposure and Photoresist Developing
Thermal Evaporation . . . . . . . . . . . . . . . .
Liftoff . . . . . . . . . . . . . . . . . . . . . . . .
REFERENCES . . . . . . . . . . . . . . .

ix

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.
.

115
118
121
121
123
124
131
133
135
138
138
139
140
144
146
148
149
150
152
152
153
153
155
155
157
158
158

. . . . . . . . . . . . . . . . . . 162

LIST OF TABLES

Table A.1

GSH optical focus depth measurements of catalyst + Nafion film
thicknesses, using the change in position of the optical microscope
focus to determine the film heights. . . . . . . . . . . . . . . . . . . 105

x

LIST OF FIGURES

Figure 1.1

Figure 1.2

Figure 1.3

Figure 2.1

Each seahorse-shaped subunit of a pilus is an individual pilin. Four
or five pilins make up one turn of the helical pilus [6]. The plural
of pilus is pili. For simplicity, these Geobacter pili are sketched with
no terminating proteins (coded for by pilC in some bacterial pili
[6]). Pilin structure is displayed from Swiss-PdbViewer [7] using the
coordinates in Reference [8]. Pili were drawn using a space-filling
model representation from RasMol [9] as a cookie cutter to form the
shapes. For interpretation of the references to color in this and all
other figures, the reader is referred to the electronic version of this
dissertation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

The pilus nanowires of Geobacter sulfurreducens, observed using different magnifications and different microscopies. Left: Transmission
electron micrograph of G. sulfurreducens cells and pili from Reference [3]. Adapted by permission from Macmillan Publishers Ltd:
NATURE [3], copyright 2005. Center: Atomic force micrograph of
purified Geobacter pili. Right: Scanning tunneling micrograph of
a purified Geobacter pilus (the “echo” above the pilus is due to a
multiple-tip effect). Scale bars are 200 nm. . . . . . . . . . . . . . .

3

Some of the applications of Geobacter. In this diagram, brown rods
represent cells, and black filaments represent pili. All examples here
show biofilms of Geobacter in water-containing sediments. Upper left:
Bioremediation of chemically toxic waste. Other bacteria break up
complex organic waste into fermentation products [15], and Geobacter breaks it down still further. Upper right: Microbial fuel cell. The
example pictured is based on the “BUG” design for seafloor power
generation, as in Reference [15]. Bottom: Bioremediation of nuclear
waste. Soluble nuclear waste (e.g., U(VI) oxide) is chemically reduced and thereby rendered insoluble, at which point it precipitates
out rather than perpetuating in groundwater. . . . . . . . . . . . . .

8

Schematic of an STM tip and sample at progressive layers of magnification moving from left to right. Large circles in the right image
represent atoms surrounded by mobile electrons. The radius of curvature of a typical tip is several tens of nanometers. The apex of a
good tip is essentially a single atom. . . . . . . . . . . . . . . . . . .

12

xi

Figure 2.2

Figure 2.3

Figure 2.4

Figure 2.5

Left: Diagram of a typical multiple tip. The feature on the sample
is measured first with the rightmost tip, then with the left tip, and
the two values are summed together to give the appearance of two
features. The bottom trace is that attributable to a single tip; the
dashed and dotted traces are the superimposed images due to both
the tips, and the top trace is the result. The inset image is how the
feature would appear to the experimenter. Right: Pili and graphite
flakes imaged by a multiple tip; scale bar is 200 nm. Imaged with
applied bias voltage + 0.5 V and tunneling current setpoint 0.1 nA.
(+ 0.5 V, 0.1 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14

Diagram of an STM experiment. An atomically sharp, electrically
conductive tip is rastered very near to an electrically non-insulating
sample, which is biased at a voltage value chosen by the operator.
Electrons tunneling between the tip and the sample are measured by
sensitive charge-detection circuitry, and the z position of the tip is
adjusted to keep the detected tunneling current, proportional to the
number of tunneled electrons per second, constant. . . . . . . . . . .

15

Diagram of one-dimensional tunneling for the STM tip–sample system. Electron wavefunctions are shown superimposed on gray-colored
rectangles representing the electronic states available within the materials. Electron wavefunction in each region shown in fuchsia. Fermi
energy of each material shown in cyan. The difference between Fermi
levels of the materials is shown by a cyan slab (this can be controlled
by applying a bias voltage), and the difference between work functions
of the materials is shown by a brown slab. . . . . . . . . . . . . . . .

16

Diagram of the effect of bias voltage on tunneling. Left: Both tip
and sample are grounded. Electrons can tunnel between tip and
sample with equal probability, here represented by the thickness of
the curves. Right: Positive bias applied to the sample. Electrons
continue to tunnel in both directions, but the tip-to-sample tunneling
probability is enhanced by the applied positive bias. Since this is a
tunneling process, these lines do not represent travel paths; rather,
the detected position of the electron changes discontinuously from tip
to sample or vice versa. . . . . . . . . . . . . . . . . . . . . . . . . .

17

xii

Figure 2.6

Figure 2.7

Figure 2.8

Figure 2.9

Figure 2.10

Diagram of the effect of bias voltage, showing the electronic states
of the tip and the sample. Shaded areas represent filled electronic
states. Arrows denote electron net tunneling direction. The tip has
a smooth density of states. The sample density of states here has
discernible features. Left: Sample is positively biased with respect
to the tip. Electrons from the tip tunnel into sample states with
energies between the Fermi energy and the energy set by the applied
bias voltage. Right: Sample is negatively biased. Electrons from the
sample tunnel into tip states between the Fermi energy and the bias
energy. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

Left: 250 nm × 250 nm STM topograph of a graphite step edge.
The red line marks the location of the height profile to the right.
Right: Line section across the edge. The blue points on the section
are separated by the height of one atomic layer of graphite. (+ 0.5
V, 0.2 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

Diagram of density of states in point tunneling spectroscopy measurements. Left: The number and width of the dashes at a given energy
corresponds to the number of available electronic states at that energy. The outline reflects that the observable which is proportional
to the density of states appears to be a continuous quantity, due to
thermal and experimental smearing of energy levels. Center: Same
as left, with axes switched and dashes removed. Right: Same as left
and center, as it would appear experimentally. . . . . . . . . . . . .

22

Point tunneling spectroscopy on graphite. Black horizontal lines are
at zero. Left: Five I–V curves taken immediately in succession.
Right: Numerically-differentiated dI/dV of the average of those five
curves, with 47-point Savitzky-Golay smoothing prior to differentiation to stabilize the derivative. The extent of the voltage scale is
reduced in order to avoid artifacts of the derivative at the endpoints
of the I–V curve. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

24

Diagrammatic I–V and dI/dV plots for the idealized semiconductor
pictured on the left. As the voltage is swept from V1 to V3 , the
I–V curve behaves linearly in the valence band of the semiconductor
(dI/dV is constant), the I–V curve is zero for the range of voltages
inside the bandgap (dI/dV is zero), and the I–V curve is linear again
(dI/dV is constant) when the applied voltage reaches the conduction
band of the semiconductor. . . . . . . . . . . . . . . . . . . . . . . .

25

xiii

Figure 2.11

Figure 3.1

Figure 3.2

Figure 3.3

Figure 3.4

Figure 3.5

Diagram of an atomic force microscope. The tip is at the end of a
cantilever; deflection of a laser reflected off the back of the cantilever
is proportional to the cantilever bending. Left: Schematic of AFM
hardware. Right Top: When the tip is free of the surface, the deflection of the reflected laser beam is determined by the angle at which
the cantilever is mounted and, if operated in amplitude modulation
mode, the angle of the cantilever at that point in its oscillation. Right
Bottom: When the tip is interacting with the surface, an additional
deflection due to forces from tip–sample interactions is added. . . . .

27

Besocke-design STM with ramps. Left: Schematic of the design.
Right: Photograph of the microscope with no ramps in place. The
caps atop all three “walking” piezotubes can be seen, as well as the
“scanning” piezotube in the center on which a tip is mounted. . . .

30

Diagram of the inertial translation process, shown in two dimensions
and applied to a lateral translation for simplicity of demonstration.
Motion is exaggerated for visibility. A voltage is applied to the quadrants of the walking piezotubes, causing them to bend while they
carry the ramps on their tops. The voltage is then suddenly switched
off, causing the piezotubes to straighten up again while the ramps
and sample retain the offset. . . . . . . . . . . . . . . . . . . . . . .

31

An STM tip in place on the microscope. Left: Photograph of a
mounted tip. Center: Optical microscope photograph of Elizabeth,
the tip used to acquire the clearest low-temperature data. The radius
of curvature of the apex of a good STM tip is only a few to a few
tens of nanometers, so the apex is not visible via optical microscopy.
Right: Labeled components of a mounted tip. . . . . . . . . . . . . .

33

Left: Graphite lattice imaged using the tabletop EasyScan STM at
an outreach event. Scan size is 1.95 nm × 1.95 nm. (+ 0.1 V, 1
nA) Right: Photograph of the tabletop STM in situ in a high school
classroom. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

34

Left: Most of a platinum pit, imaged primarily to calibrate the scanning piezotube extension in the Z direction for a given voltage. The
effect of a multiple tip can be seen by the inverse ziggurat shape of
the pit. Streaky specks are due to particulate debris on the sample
surface. Right: Hexagonal graphite lattice imaged with atomic resolution to obtain the XY calibration for the scanning piezotube. (+
0.1 V, 1 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

35

xiv

Figure 3.6

Figure 3.7

Figure 4.1

Figure 4.2

Figure 4.3

Left: Data acquisition setup for STM and spectroscopy. The microscope is at the bottom of the long stick and is submerged in the dewar
(partially submerged, in this example). The dewar extends below the
bottom edge of the photograph. Right: Lower part of the dewar,
showing the shims and adhesive used for stability. The bright green
tape is a safety measure to warn passersby against bumping into the
apparatus. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

37

AFM topographs obtained using the Cypher. Fabrication of these
samples is discussed in Appendix E. Left: 10 µm × 10 µm scan size.
Center: 4.5 µm × 4.5 µm scan. Right: 5 µm × 5 µm scan. All scans
shown here have a z range grayscale extent of 55 nm. . . . . . . . .

38

Artist’s conception of pilus assembly and extension at the cell membrane, inspired by and simplified from [6]. For simplicity, proteins are
represented schematically. Pilin subunits dispersed throughout the
inner membrane area are assembled by another protein into a helical
structure, then pushed through a pore in the outer cell membrane.
The hydrophobic center of the assembled pilus is a tube formed by
the long α-helices of the pilins, leaving a hollow core to the pilus, and
the heads of the pilins remain on the outside of the pilus to interact
with its environment. In Geobacter, the globular head which other
pilins have is replaced by a short, barb-like shape [8, 52, 53]. . . . .

43

Structure of the G. sulfurreducens pilin subunit based on a homology
model. Coordinates were obtained from Reference [8] and displayed
using DeepView (Swiss-PdbViewer) version 4.1.0 [7]. The α-helix
tail of the pilin is buried inside the assembled pilus, while the small
barb-like head remains on the outside of the pilus. . . . . . . . . . .

44

Structure of the G. sulfurreducens pilin subunit based on solution
NMR [52]. Coordinates were obtained from Reference [53] and displayed using Swiss-PdbViewer [7]. Left: The NMR measurements
yielded several slightly different conformations for some parts of the
pilin, suggesting flexibility in these regions [52]. Right: Same pilin,
viewed from a different angle. Electrostatic potentials due to charged
amino acids are superimposed on the pilin structure. Electrostatic potentials were calculated in DeepView using the Coulomb interaction
and assuming a dielectric constant of 80 for the surrounding medium
and 4 for the pilin, as in Reference [25]. Tyrosine amino acids are
marked with stars. . . . . . . . . . . . . . . . . . . . . . . . . . . . .

45

xv

Figure 4.4

Figure 4.5

Fibrous bundles, likely predominantly composed of aggregated pili
and perhaps also incorporating trapped fluid or contaminants. Left:
A large “haystack” of bundles on gold, with a small region of silicon
dioxide surface in the lower right corner of the image. Right: Expanded view of the image shown in Figure 1.2 showing both clumping and entwining types of bundling. Both scans are atomic force
micrographs. Scale bars are 1 µm. . . . . . . . . . . . . . . . . . . .

46

Diagram of an STM experiment on a pilus. The total tunneling resistance is the sum of the resistances of the tunneling gap and of the
pilus itself. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

53

Figure 4.6

Snapshot from a molecular dynamics calculation by Prof. Dr. G. Feliciano of a Geobacter pilin subunit, beginning from an NMR-determined
structure [53], immersed in water at 77 K. Displayed using the program VMD [66]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55

Figure 4.7

The relationship between functions and functionals. Functions have
different outputs for different variable inputs describing the immediate circumstances where they are applied. Functional s have different
outputs for different functions used to describe the paradigm of the
underlying scenario. . . . . . . . . . . . . . . . . . . . . . . . . . . .

57

Left: Room temperature STM topograph. Scale bar is 50 nm. Right:
Line section through the area of the scan marked with the black bar.
(+ 0.5 V, 0.08 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . .

61

Portion of the filament also seen in Figure 5.1. Scale bar is 20 nm.
Left: Highlighting the longitudinal pitch. Three of the nodule sets
which comprise this pitch are marked with arrows. Spacing between
arrows is ∼ 6 nm. Right: Highlighting the transverse ridge–scallop
structure. The three nodules which comprise this periodic structure
are marked with arrows. Spacing between arrows is ∼ 5–8 nm. (+
0.5 V, 0.08 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

62

Left: Portion of a pilus scanned at 77 K. Right: Portion of a different
pilus scanned at 300 K. A multiple tip is responsible for the apparent
faint “shadow” to the upper left of the actual pilus. Both scan sizes
are 50 nm × 50 nm. (+ 0.5 V, 0.08 nA) . . . . . . . . . . . . . . . .

64

Figure 5.1

Figure 5.2

Figure 5.3

xvi

Figure 5.4

Tip-induced displacement of material. Left: Scan of the area with tip
moving in the downward direction. Center: Acquired near-simultaneously
with the left image but with tip moving upward. Right: Scan of area
again with tip moving downward. The motion of the loosely-adsorbed
material during scans is not due to drift, as can be seen by observing
that the brightest portion of the scan stays in place. Scans performed
at 77 K. Scale bars 50 nm. (+ 0.5 V, 0.5 nA) . . . . . . . . . . . . . 65

Figure 5.5

Four scans of a filamental fragment at 77 K, done immediately successively in the same position. Verification that the disappearance of
segments of the object was not due to drift can be observed via the
conserved position of the patch in the upper right of the image. Scale
bars 200 nm. (+ 0.5 V, 0.1 nA) . . . . . . . . . . . . . . . . . . . .

65

Four scans of a filamental fragment at 300 K, done immediately successively in the same position. Portions of the pilus move back and
forth between scans. All scans shown were in the downward direction.
Scale bars 50 nm. (+ 0.5 V, 0.08 A) . . . . . . . . . . . . . . . . . .

66

Left: AFM topograph of a freshly-deposited sample of pili on SiO2 .
Right: Same sample rescanned after 22.5 months, showing no degradation of the sample. . . . . . . . . . . . . . . . . . . . . . . . . . .

68

350 nm × 350 nm AFM topograph of the 22.5 month old sample
showing intact individual pili and pilus bundles. . . . . . . . . . . .

69

dI/dV on a pilus at room temperature. Left: Graphite control. Center: Edge of pilus. Right: Graphite control. Because of higher levels
of scatter in the data, the pilus dI/dV was smoothed again after differentiation using a 3-point binomial smoothing, in addition to the
procedure described in Appendix D. . . . . . . . . . . . . . . . . . .

71

dI/dV on a pilus at room temperature. Left: Graphite control.
Center: dI/dV obtained in several locations transversally across the
breadth of the pilus. Right: Graphite control. All abscissas are from
- 0.4 V to + 0.4 V. Horizontal lines are dI/dV = 0. . . . . . . . . .

72

Spectroscopy performed on two separate pili, using two different tips
and two separate sample depositions. The left edge of the gap is
similar, and both spectra encounter dI/dV = 0 inside the gap. . . .

73

Figure 5.6

Figure 5.7

Figure 5.8

Figure 5.9

Figure 5.10

Figure 5.11

xvii

Figure 5.12

Figure 5.13

Figure 5.14

Figure 5.15

Figure 5.16

Figure 5.17

Figure 5.18

Figure 5.19

Immediately-sequential point tunneling spectroscopy performed on a
pilus and nearby graphite. Left: Graphite spectroscopy before pilus
spectroscopy. Center: Pilus spectroscopy. Right: Graphite spectroscopy after pilus spectroscopy. Horizontal line is at dI/dV = 0. .

75

Detail of pilus spectroscopy from Figure 5.12. Gap width is 0.2 eV.
Bottom of gap is at dI/dV = 0. Center of gap is slightly more negative
than 0 V. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

75

Immediately-sequential point tunneling spectroscopy performed on
a different part of the pilus from Figure 5.12. Left: Graphite spectroscopy before pilus spectroscopy. Center: Pilus spectroscopy. Right:
Graphite spectroscopy after pilus spectroscopy. . . . . . . . . . . . .

76

On-pilus spectroscopy, labeled by the estimated location across the
diameter of the pilus at which the spectroscopy was taken. Left:
dI/dV on the more diffuse side. Center: dI/dV towards the middle.
Right: dI/dV near the more separate area of the pilus. Bottoms of
gaps are at dI/dV = 0. . . . . . . . . . . . . . . . . . . . . . . . . .

77

Immediately-sequential point tunneling spectroscopy performed on
the middle part of the pilus from Figure 5.12. Left: Graphite spectroscopy before pilus spectroscopy. Center: Pilus spectroscopy. Right:
Graphite spectroscopy after pilus spectroscopy. . . . . . . . . . . . .

77

Comparison between cryogenic and room temperature dI/dV . Left:
77 K spectroscopy. Right: 300 K spectroscopy. Dashed line is a guide
to the eye indicating pseudogap bottom. . . . . . . . . . . . . . . . .

79

Effect of thermal broadening, given by Equation 5.1, on a hypothetical semiconductor with sharp bandgap edges. Left: Density
of states. Right: Thermally-broadened DOS which would be measured as dI/dV of this semiconductor at 300 K, overlaid on the DOS.
dI/dV is plotted as a dashed line as a guide to the eye; actual point
spacing for both plots’ computations is 50 mV. . . . . . . . . . . . .

81

Effect of thermal broadening, given by Equation 5.1, on actual pilus
data obtained at 77 K. Left: dI/dV at 77 K. For the purposes of this
calculation, this is assumed to be equal to the density of states, thus
the effect of thermal broadening will be slightly smaller than shown
here. Horizontal line is dI/dV = 0. Right: Data thermally broadened
by 300 K, overlain in red. Feature edges and small oscillations in the
dI/dV appear smoothed. . . . . . . . . . . . . . . . . . . . . . . . .

81

xviii

Figure 5.20

Figure 5.21

Figure 5.22

Figure 5.23

Figure A.1

Figure A.2

Calculated density of states for a Geobacter pilin [65]. Eight snapshots of pilin and solvent conformations, shown individually to display
the variation in density of states among configurations. All calculations plotted here are displayed over twice the voltage scale of the
experimental data to show more of the structure of the density of
states. Molecular dynamics and density functional theory calculations were performed by G. Feliciano. . . . . . . . . . . . . . . . . .

85

Calculated density of states for a Geobacter pilin [65]. Five snapshots
of pilin and solvent conformations are averaged together. Molecular
dynamics and density functional theory calculations were performed
by G. Feliciano. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

86

Periodic clusters of tyrosine amino acids in the assembled pilus model
adapted from Reference [52]. This research was originally published in
The Journal of Biological Chemistry. P.N. Reardon and K.T. Mueller.
Structure of the Type IVa Major Pilin from the Electrically Conductive Bacterial Nanowires of Geobacter sulfurreducens. 2013; in press.
Copyright the American Society for Biochemistry and Molecular Biology. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

90

Suggested new electrode scheme for pilus longitudinal conductivity
experiment [25]. The graphene flakes provide a more pilus-friendly
interface to the gold electrodes. Diagram is not to scale. . . . . . . .

92

Few-minute subset of a noise test. Top: Amplitude of the noise in
picometers. Noise amplitude decreased when nearby equipment was
turned off. Bottom: Frequency distribution. Mechanical vibration is
evident in the broad band of noise around 20 Hz. . . . . . . . . . . .

96

Left: Diagram of optical-only GSH. The objective lens of the optical
microscope is focused on the top and bottom of the sample, and the
motor position of the objective is recorded at the point at which each
comes into focus. Right: GSH with a tip in place. The focus of the
objective lens of the optical microscope remains set on the tip, while
the tip + objective is moved as a unit until the tip comes in contact with the surface of the sample. The motor position of the tip +
objective is recorded. Because the tip–sample contact involves extension or retraction of the z piezo, the z piezo extension or retraction
can be an important correction factor on the order of a micron, but
involving the tip can give more precise measurements of few-micron
heights. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

97

xix

Figure A.3

Demonstration of the GSH method on a glass slide. Left: Focus on
top of slide. Right: Focus on sample disk underneath slide. Difference
between focus positions is 1.2 mm. . . . . . . . . . . . . . . . . . . .

98

Figure A.4

Focus position determination on catalyst + Nafion films loaded onto
glassy carbon rotating-ring electrodes. Left: Optical micrograph indicating the catalyst layer “cat” and underlying glassy carbon “gc”
electrode. Scale bar is 50 µm. Center: Optical micrograph focusing
on the top of the catalyst, digitally zoomed to twice the magnification of the left and right images for ease of discerning features during
focusing. Scale bar is 25 µm. Right: Optical micrograph focusing on
the underlying electrode. Scale bar is 50 µm. Brightness differences
among these images are due to the light levels used to aid in focusing. 100

Figure A.5

Height measurements of catalyst + Nafion films loaded onto glassy
carbon electrodes using both optical focus depth GSH and tip-enhanced
GSH. A linear fit was made to the data acquired using Electrode 1,
showing a linear dependence of film thickness on sample loading. The
alternate electrode, Electrode 2, appears to have a slightly greater
slope, but there were not enough sample loadings using that electrode to determine this definitively. Within statistical uncertainty,
the optical and tip-enhanced methods are equivalent. . . . . . . . . 101

Figure A.6

I–V curves demonstrating the “ears,” adapted from a post to the
Asylum Research Forum [85]. Left: ± 0.2 V bias sweep. Right: ±
0.5 V bias sweep. Both: The I–V curve on top is as seen by the
10 µA preamp, and the I–V curve on bottom is as seen by the 10
nA preamp, separated by a dashed line for clarity. Horizontal lines
in the lower I–V curves occur when the 10 nA preamp is saturated.
The onset of this saturation corresponds to the “ears” in the upper
I–V curves. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104

Figure B.1

Typical artifacts at graphite step edges on two different parts of a
sample. These are not pili lying against the step edges but rather are
graphite formations. Scale bars are 200 nm. Left scan (+ 0.5 V, 0.5
nA), right scan (+ 0.5 V, 0.4 nA) . . . . . . . . . . . . . . . . . . . 108

Figure B.2

Three scans of a step edge, possessed of a deceptively lumpy structure
along the diagonal from upper left to lower right. The level of visible
detail, obtained by decreasing the scan size, increases from left to
right. Scale bars 10 nm. (+ 0.5 V, 0.2 nA) . . . . . . . . . . . . . . 109

xx

Figure B.3

Artifact comprised of periodic shapes reminiscent of vertebrae in a
spinal column. Left: Close-up of comma-shaped “vertebrae.” Scale
bar is 10 nm. Center: Larger scan demonstrating the abruptness of
the dislocations. Scale bar is 50 nm. Right: Relation of the vertebral
shape to a nearby step edge. The vertebral shape apparently merges
into the step edge near the bottom of the scan. Scale bar is 100 nm.
(+ 0.5 V, 0.08 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . 110

Figure B.4

Left: Tendriloid graphite ripple of approximately the apparent diameter and one half to one third the apparent height expected for a pilus
on graphite. Center: These ripples can be provoked by disruptions
in the graphite surface. Right: Several tendriloid artifacts, showing
a variety of the shapes they can assume. Scale bars 100 nm. (+ 0.5
V, 0.08 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

Figure B.5

Left: Hexagonal pattern on graphite due to a Moir´ pattern from ine
terference between overlapping graphene sheets. A line has been subtracted from each line of the image, and the displayed height shading
range has been truncated to emphasize the hexagonal pattern. Scan
size for left image is 244 nm × 298 nm. (+ 1 V, 0.08 nA) Center:
Diagram of a Moir´ interference pattern. Light and dark lines repree
sent lines of atoms in two overlapping layers of single-atom thickness.
Circles at the intersections indicate the locations where the enhanced
electronic signal from the overlap would be detected as a pattern.
Right: Diagram of a Moir´ pattern for a hexagonal lattice such as
e
graphite. Lattices have been rotated by 30◦ for visibility; actual rotation for the pattern observed here would have been much smaller.

Figure B.6

113

Moir´ pattern and graphite lattice comparison. Left: Zoom in on
e
hexagonal pattern from Figure B.5. Image size 123 nm × 147 nm. (+
1 V, 0.08 nA) Right: Atomic resolution scan of the graphite lattice.
Image size 2.1 nm × 2.6 nm. (+ 0.1 V, 1 nA) The left image is
nearly an order of magnitude larger than the right image along each
direction, yet the patterns appear nearly the same size, indicating the
artifactuality of the left image. . . . . . . . . . . . . . . . . . . . . . 114

xxi

Figure B.7

A particularly dramatic example of a multiple tip. Left: Scan showing multiple repetitions of features. From this scan, it appears there
are at least three tips close in height to each other and aligned approximately in the horizontal direction, slightly angled towards the
upper left of the image, and several additional tips approximately
along the vertical direction. Scan size 500 nm × 500 nm. Center:
150 nm × 150 nm scan near the middle of the scan on the left. The
approximately-horizontal multiple tips are even greater in number,
as another nearby tip is causing an overlapping echo on the leftmost
of the three main feature repetitions, as well as causing two fainter
echoes to the left of the three. (+ 0.5 V, 0.2 nA) Right: Artist’s
conception of a tip which could produce the horizontal multiplicity
shown in these images; multiple tips along the other direction are not
considered in this sketch. Only the protrusions near the very end of
the tip are relevant—in this sketch, there are six such—due to the
exponential dependence of tunneling current on tip–sample separation.115

Figure B.8

dI/dV reminiscent of the Coulomb blockade effect convolved with
proper graphite spectroscopy. Repeated peaks at symmetric spacings
along the voltage axis are the chief diagnostic symptom, indicated
with arrows. The offset of the bottom of the spectrum from 0 V
could be due to Fermi level pinning due to contaminants on the surface.117

Figure B.9

Example of spectroscopy with an unstable tip. This dataset sums
together two spectroscopy points on graphite, which should result
in a sigmoidal shape for the I–V curves in the left image and a
slightly rounded V-shape for the dI/dV curves in the right image.
The thicker, black line is the average of the spectra. Left: I–V curves.
Right: Corresponding dI/dV curves, obtained directly using a lock-in
amplifier. The curves are not reproducible. . . . . . . . . . . . . . . 118

Figure B.10 STM of a periodic feature. Bright spots coincident with longitudinal
zig-zags are approximately 100 nm apart. Sub-periodicity is approximately 2 nm. Left: Scan size 500 nm × 500 nm. Center: 125 nm
× 125 nm. Right: 50 nm × 50 nm. (+ 0.5 V, 0.2 nA) . . . . . . . . 119
Figure B.11 STM topographs of the object before and after point spectroscopy.
Indicated features in left and center scans are the same in both images, used to identify the extent of lateral drift between scans. Right:
Same as center, with locations of on-object and graphite control spectroscopy marked by dots. The newly-acquired dimming of a spot on
the lower filamental shape is reflected by dimming on the upper filament, marked by arrows, indicating that the filaments are a single
object. Scan sizes 20 nm × 20 nm. (- 0.5 V, - 0.2 nA) . . . . . . . . 120
xxii

Figure C.1

HOPG mounted on a stainless steel sample disk. Left: Graphite
crystal mounted on a stainless steel sample disk. The conductive
adhesive used to mount the graphite on the disk can be seen extending
beyond the graphite footprint. Photograph taken through an optical
microscope at 20x magnification. Center: Graphite on sample disk,
showing the alignment marks used to center the graphite. Right:
Graphite mounted on sample disk and inserted into ramps. At this
stage, the graphite sample is ready either to have pili deposited onto
it or to be mounted on the STM directly. . . . . . . . . . . . . . . . 123

Figure C.2

AFM topographs of pili on SiO2 . Left: Pili on a circular area of
SiO2 surrounded by gold. The pili on the left of the circle follow
tree-like branching patterns of bundling. The pili on the right of
the circle have perhaps trapped fluid between the filaments, resulting
in a “frayed blanket” look. Center: Another “frayed blanket” pilus
formation. Right: Typical “neuron” structure of pilus bundles on
SiO2 . All scans are 5.7 µm in the vertical direction. Left scan width
9 µm, z grayscale 40 nm; center scan width 7 µm, z grayscale 40 nm;
right scan width 5.7 µm, z grayscale 15 nm. . . . . . . . . . . . . . . 125

Figure C.3

AFM topography scans from the deposition duration investigation,
selected to highlight features from buffer residue. All samples shown
here were deposited from CHES buffer (pH 9.5). Left: 30 minute
deposition time, rinsed, RCA 1 chip pre-cleaning. Center: 2 hour
deposition time, solvent chip pre-cleaning. Right: 12 hour deposition
time, rinsed, RCA 1 chip pre-cleaning. All image sizes are 1 µm × 1
µm. Left z grayscale 6 nm, center z grayscale 20 nm, right z grayscale
34 nm. Lozenge-shaped features are not bacteria, which would be
much larger and which were removed by the purification protocol. . 127

Figure C.4

Highlights from a deposition test conducted with the droplet of pili
in pH 9.5 CHES buffer remaining on the thermally oxidized silicon
surface for 2 hours, then rinsed twice with doubly deionized water.
Upper Left: Focus on several individual pili interacting at a nodule.
Scan size is 150 nm × 150 nm. Upper Right and Bottom Row: Pilus
network topography examples from other locations on the same sample. Scale bars are 1 µm. The z height shading scale is the same for
all of these AFM images except the upper left scan. . . . . . . . . . 129

Figure C.5

Scans showing “film” between pili, from the same deposition test as
Figure C.4. Left and Center: Height and phase images on the same
feature area. Scale bars are 1 µm. Right: Phase image of an 800 nm
× 800 nm portion of the “neural network” shown at left. . . . . . . 130

xxiii

Figure C.6

Unexpected cell observed on a sample using AFM. The lower substrate surface is SiO2 , and the upper surface is a 25 nm-tall evaporated gold electrode. . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

Figure C.7

Unexpected cell observed on a sample using AFM, seen in its entirety
in Figure C.6. Left: Cell and flagellum on gold. Right: Flagellum on
SiO2 with G. sulfurreducens pili forming mats in response to it. . . . 131

Figure C.8

Schematic of the search pattern typically followed within a particular
“region.” Each individual scan area of the nine scan areas comprising
the region was 40 V × 40 V in size, where this voltage is the voltage
applied to the scanning piezotube’s quadrants. At room temperature,
each scan area was approximately 0.8 µm × 0.8 µm, while at 77 K,
the same applied voltages resulted in a scan area of approximately
0.3 µm × 0.3 µm, to one significant figure. Each of the nine boxes
within the region is a complete scan. . . . . . . . . . . . . . . . . . . 133

Figure C.9

A single STM topography scan. The tip moved down and then up
along the same line of the image before proceeding to the next line.
Tip-induced sample motion can be seen most clearly in the center of
the scan, where nanowire is perturbed in the direction of scanning.
Scan size is 763 nm vertically and 749 nm horizontally. The apparent
breadth of the nanowire is likely due to a combination of the convolution of the nanowire topography with the tip’s radius of curvature,
ambiguous resolution of the edges of the nanowire, and motion of the
nanowire along the scan direction. (+ 0.75 V, 0.2 nA) . . . . . . . . 136

Figure C.10 STM scans of one filament. Arrows indicate a feature useful as a
landmark. An unstably imaged tendril is visible below the arrow.
Left: Scan of a pilus unstably affixed to the surface. Center: Nearsimultaneous scan but with the tip moving in the opposite direction.
Right: Smaller-scale scan. Left and center scans are 400 nm × 400
nm; right scan is 100 nm × 100 nm. (+ 0.5 V, 0.2 nA) . . . . . . . 137
Figure C.11 Series of STM scans zooming in on the croissant-shaped mandibles—
two individual pili—protruding from an aggregate bundle of pili. Pili
found near bundles or graphite flakes were generally stable on the
surface over hours of scanning, perhaps because of being anchored to
the surface by the larger objects. Scale bars 20 nm. (+ 0.5 V, 0.1 nA) 137

xxiv

Figure D.1

Example of “piezo creep.” Scan began from bottom of image and
proceeded upward. The apparent bending of the image near the start
of the scan is due to a delay in the full response of the scanning
piezotube to the voltage applied to it. This scan of a silicon dioxide
surface was obtained using a Dimension 3100 AFM with Nanoscope
IIIa controller. Scan area is 1 µm × 1 µm. . . . . . . . . . . . . . . 141

Figure D.2

Changing the colorscale used to display the data. Left: Raw image, as
displayed by the Asylum Research data acquisition software. Right:
After “auto” color range and offset applied. Both images show an
overall tilt whereby the lower left of the image appears higher than
the upper right of the image. (+ 0.5 V, 0.2 nA) . . . . . . . . . . . 142

Figure D.3

Subtracting a background plane from the data. Top: Colorscaled
data as seen in Figure D.2. Bottom: Data after a first-order plane
subtraction was applied in the XY direction. Right: Line sections
through the images on the left. The apparent height interval between
the blue markers changes significantly with processing. (+ 0.5 V, 0.2
nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

Figure D.4

Applying an offset along the X direction to the data from Figure D.3.
Left: Image after subtracting a plane in the XY direction, subtracting the average offset of each line in the slow scan direction, and
choosing the range and offset for the grayscale to match that in the
processed image in Figure D.3. Right: Data as originally displayed
in Figure D.2. Image processing done in Igor Pro using the Asylum
Research software. (+ 0.5 V, 0.2 nA) . . . . . . . . . . . . . . . . . 144

Figure D.5

Five immediately-sequential voltage sweeps above graphite. Black
curves are averages of these curves. Left: I–V curves. Center: Minimally smoothed numerical derivatives calculated by the RHK data
acquisition software. dI/dV = 0 is the bottom of the plot; parts
of these curves also extend, unphysically, below that level. Noise
dominates this plot. Right: Direct dI/dV via the lock-in technique.
dI/dV = 0 is the bottom of the plot. . . . . . . . . . . . . . . . . . 145

Figure D.6

Further analysis of averaged spectroscopy from Figure D.5. Left:
Numerical derivative of averaged I–V curves after smoothing with a
47-point Savitzky–Golay filter using a 2nd-order polynomial to calm
down some of the scatter in the data. Center: Same as left but
only plotted between ± 0.4 V. Right: Direct dI/dV via the lock-in,
averaged and replotted on same scales as center plot. Black lines at
bottoms of plots are dI/dV = 0. . . . . . . . . . . . . . . . . . . . . 146

xxv

Figure D.7

Left: Standard and 3D view of a graphite step, several monolayers
in height, obtained using the cryogenic-capable STM. Scan is 342 nm
in the X direction and 418 nm in the Y direction. (+ 0.5 V, 1 nA)
Center: 2 µm × 2 µm scan of overlapping graphite steps, obtained
using the Cypher STM. Right: 3D display of overlapping steps. (+
0.5 V, 0.2 nA) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

Figure D.8

STM topographs of the “ends” of two separate pili showing the characteristic notch. Both images are 75 nm × 75 nm. Both scans were
performed at an applied bias of + 0.5 V. Left image tunneling current
setpoint 0.2 nA, right 0.08 nA. The ghostly “echoes” in both images
are due to a multiple tip effect. . . . . . . . . . . . . . . . . . . . . . 151

Figure E.1

AFM topograph of a microfabricated electrode and pili. Scan is 3
µm x 3 µm. The brighter region on the left is a gold electrode; the
lower, darker-shaded region is the insulating substrate, exposed as a
Swiss-cheese-like circular cut-out in the electrode. The tendril lying
across the electrode and substrate is a small bundle, likely a pair, of
pili. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154

Figure E.2

Diagram of the photolithography procedure, not to scale. Left to
right: substrate with photoresist deposited on top, photoresist exposed and developed to leave a pattern, gold deposited on the pattern
and photoresist, photoresist “lifted off” to leave solely the pattern. . 155

Figure E.3

Left: Photograph of a chip with electrodes patterned into photoresist
and a layer of gold evaporated over the whole chip, before liftoff. The
pattern of the electrodes can be seen as darker areas on the shiny
surface. Right: Photograph of a chip after liftoff. The electrodes are
the rectangular patterns. Chips are approximately 0.5 inches wide. . 159

Figure E.4

Photograph of a sample used for the CP-AFM experiment. For scale,
the disk on which the chip is mounted is 15 mm in diameter. The
cylinder at the top of the image is the magnet to which the bias wire
from the microscope was attached. Conductive silver paint connects
this to a solid gold pad which is then connected via gold wires or by
direct contact to electrodes on the chip. Residue of sample deposition
can be seen on one of the electrodes at left in the image. . . . . . . . 161

xxvi

Chapter 1
Overture
The search is on for electronic materials with interesting, tunable properties. Complementary to the strategy of devising such materials by doping well-studied crystals to modify
their properties in the desired direction is the approach of using and modifying structures
already present in natural materials [1]. Among the vast range of suggested such materials
are, e.g., nanoscale fibers made from viruses [2]. It has already been established that electrically conductive protein filaments [3] are produced by several different species of bacteria,
including the Geobacter sulfurreducens species studied in this thesis [4, 5]. Endeavoring to
elucidate further the electronic properties of a specific type of protein filament harvested from
G. sulfurreducens in order to exploit them optimally, as well as to understand how they work
on a fundamental level, is the subject of this study—why reinvent the nanowire? The protein
nanowires studied in this work are the pili (singular: pilus) expressed by G. sulfurreducens,
shown in Figures 1.1 and 1.2.
In the course of its metabolic/respiratory process—breathing and eating are much the
same to a single-celled organism—Geobacter sulfurreducens gains energy by breaking apart
organic molecules, which leads to an excess of electrons inside the cell. To discharge itself,
Geobacter donates these electrons to various electron acceptors in its environment, thereby
chemically reducing them. These electron acceptors perform the same role for Geobacter, an
anaerobic bacterium, that oxygen does for aerobic-respiring creatures.
Geobacter sulfurreducens is often provided with the electron acceptor iron (III) oxide to
1

pilin

pilus

pili

Figure 1.1: Each seahorse-shaped subunit of a pilus is an individual pilin. Four or five
pilins make up one turn of the helical pilus [6]. The plural of pilus is pili. For simplicity,
these Geobacter pili are sketched with no terminating proteins (coded for by pilC in some
bacterial pili [6]). Pilin structure is displayed from Swiss-PdbViewer [7] using the coordinates
in Reference [8]. Pili were drawn using a space-filling model representation from RasMol [9]
as a cookie cutter to form the shapes. For interpretation of the references to color in this
and all other figures, the reader is referred to the electronic version of this dissertation.
reduce when grown in the lab because Fe(III) oxide is the classic electron acceptor of this
type of metabolic process [10]. Alternatively, Geobacter can take advantage of other electron
acceptors in the environment, including uranium (VI) oxide [11, 12]. This makes G. sulfurreducens an exciting candidate for bioremediation schemes [13], since reducing uranium (VI)
oxide to uranium (IV) oxide—which Geobacter species can do—decreases its solubility and
precipitates the uranium out of groundwater, which can therefore decrease the spread of
uranium contamination [11]. Such schemes are currently limited by scalability issues due
in part to the large areas involved and the small size of the bacteria, as well as frequentlysignificant differences between laboratory conditions and field-scale circumstances [14]; hence
much further study is needed to aid in making the process viable. Furthermore, this ability
of G. sulfurreducens to transfer negative charge to an acceptor material enables it to be
used in microbial fuel cells, as in Figure 1.3. Replacing the standard environmental electron
acceptors such as Fe(III) with the anode electrode of a battery to be charged or a device
2

Figure 1.2: The pilus nanowires of Geobacter sulfurreducens, observed using different magnifications and different microscopies. Left: Transmission electron micrograph of G. sulfurreducens cells and pili from Reference [3]. Adapted by permission from Macmillan Publishers
Ltd: NATURE [3], copyright 2005. Center: Atomic force micrograph of purified Geobacter pili. Right: Scanning tunneling micrograph of a purified Geobacter pilus (the “echo”
above the pilus is due to a multiple-tip effect). Scale bars are 200 nm.
to be powered, and inserting hydrocarbons such as acetate for Geobacter to break up into
carbon dioxide, enables Geobacter ’s metabolic process to be used for electricity generation
applications [15]. The continued efficiency of these microbial fuel cells as Geobacter cells
grow into stacks and form a biofilm can be attributed to the pili [16] which are the subject
of this research. Pili allow cells in the biofilm whose membranes are not directly touching
the electron acceptor still to be in electrical contact with it [16].
For the cell to reduce the electron acceptor, it must by some means transfer its excess
electrons to that electron acceptor. Soluble electron acceptors can permeate directly into
the cell. Alternatively, the cell membrane can be in direct contact with an insoluble electron
acceptor, for example when insoluble electron acceptors are diffusing through the cell’s environment and occasionally encountering the cell or when the cell is sitting directly atop an
insoluble electrode in a microbial fuel cell [16]. In this case, the cell can transfer the electrons
to the electron acceptor through, e.g., metal-containing cytochromes in the outer membrane
of the cell [17]. If, however, the outer cell membrane is not in direct contact with electron
3

acceptors, for example if it is “land-locked” in the middle of a biofilm, the cell must find an
alternate means of discharging itself. One method is to use intermediary electron acceptors,
called electron shuttles, which diffuse between the cell and the terminal electron acceptor
elsewhere in the environment; this is essentially creating soluble electron acceptors to move
between the cell and an insoluble electron acceptor elsewhere. These electron shuttles diffusively move to the cell membrane, at which point the cell discharges itself onto the electron
shuttle; the newly-reduced electron shuttle then diffuses away and eventually relieves itself
of that excess charge against another electron acceptor elsewhere, freeing it to begin again
[17]. This approach is so favored by some bacteria that they actually express proteins to
manufacture these electron shuttles and release them into their environment; this is done
by the Shewanella species [18]. Although Geobacter species do not manufacture electron
shuttles [19], if electron shuttles are present in the environment, Geobacter will discharge
itself onto them as it would with other soluble electron acceptors [18]. In biofilms, the stacks
of bacteria are surrounded by a medium filled with all manner of cellular byproducts and
assistive molecules, so isolation of the components of the biofilm is necessary to determine
the critical component for electron transfer. The research described here builds on previous
studies, e.g., Reference [3], which have determined that the pili, protein nanowires extended
from the cell membrane, are responsible for electrical transport from G. sulfurreducens to
spatially distant elements of its environment.
Geobacter sulfurreducens, and also certain other bacteria, express electrically conductive
type IV pili to perform electron transport to insoluble electron acceptors at some distance
from the cell [3]. Type IV pili are grouped together based on similarities in the amino acid
sequences of their pilins [20]. The type IV pilus structure is common among bacteria and
used for different purposes by different species; some bacteria use them for locomotion, but
4

this does not appear to be the case for Geobacter sulfurreducens [10], and some bacteria
use them to help hold bacterial biofilms together structurally, which does appear to be one
function of Geobacter ’s pili [21]. Geobacter ’s pili are electrically conductive, but not all
type IV pili are electrically conductive; the pili of Pseudomonas aeruginosa strain K, for
example, are insulators [22]. Type IV pili are composed of subunits, called pilins, which
are expressed inside the cell, assembled inside the cell’s outer membrane and pushed outside
the cell through a pore in the outer membrane [6]. Each pilus has a repeating helical
structure comprised of pilin subunits and can grow to be at least several micrometers long
(see, e.g., Figure 1.2 or Reference [3]).
Veazey et al. [23] established that the protein matrix of the pili, rather than associated
cytochromes or metal atoms, are responsible for the electronic conductivity of the pili. Such
purified protein nanowires [12] were investigated in this research. Pili in bacterial biofilms
are accompanied by various other extracellular productions of the bacteria, including metalcontaining molecules. For this study, these molecules were removed [12], leaving only the
purified protein matrix.
While several studies have addressed axial charge transport along the length of nanowires
of various species (e.g., [24, 25]), transverse properties of the nanowire diameters are also a
subject of consideration, as pili are permitted to intertwine with one another [26], and as
pili are electrically conductive in the transverse direction [3].
The aim of this research is to study the electronic properties of the protein nanowires
grown by Geobacter sulfurreducens. To this end, the pili are investigated by physical techniques as other kinds of nanowires, e.g., carbon nanotubes, would be [27]. Indeed, point
tunneling spectroscopy, a technique used on the pili for this research, was able to categorize carbon nanotubes as having semiconducting or metallic electronic structures, depending
5

on their physical structures [28, 29]. The density of states of the pili contains detailed information about the electronic structure of these protein nanowires, as discussed further
in Chapter 2. Density of states measurements can be obtained with high spatial precision
using the spectroscopic capabilities of the scanning tunneling microscope (STM). Scanning
tunneling microscopy and point tunneling spectroscopy employ a repositionable, sharp, electrically conductive tip in close proximity to the sample to achieve nanometer-scale resolution
of the sample topography. Similarly, nanoscale positionability of the localities for the spectroscopy measurements is obtained. Using scanning tunneling microscopy and spectroscopy,
the density of states has been measured for wild-type G. sulfurreducens pili at 77 K and at
room temperature. Temperatures are generally referred to here using two significant figures,
i.e., 300 K rather than 297 K.
Throughout this document, both the words “pilus” and “nanowire” are used interchangeably to refer to the pili grown by G. sulfurreducens which act as a biological type of nanowire;
the electrically conductive pili behave like wires of nanometer-scale diameter. Often, the
terms “bias” and “voltage” are used synonymously, as well. Because of the interdisciplinary
tone of this research, various terms in this document may reflect different usages in different
fields of study. Such terms include “substrate,” which is always used here in its physical
sense as an underlying surface upon which the subject of primary study is present, rather
than in its biochemical sense. Another source of verbal subtlety lies in the names for different components of the protein nanowires under study here, as but a single letter makes the
difference between discussion of a pilin (subunit of a pilus) and of pili (more than one pilus).
For clarity, a terminology guide is presented in Figure 1.1.
Beginning with a discussion of the scanning probe microscopy techniques (Chapter 2)
and specific microscopes (Chapter 3) used for this research, this document proceeds through
6

an overview of the samples studied and the theoretical underpinnings of the interpretation
framework (Chapter 4). The results are then presented, together with conclusions drawn
from them and suggested directions for future investigations (Chapter 5). The primary
Appendix describes other work done supervising, maintaining, and using an atomic force
microscope facility instrument (Appendix A). The remaining appendices present cautionary
tales of pili-like artifacts (Appendix B) and further details regarding the sample preparation
(Appendix C), techniques, and equipment used in this (Appendix D) and closely-related
research (Appendix E) on the protein nanowires of Geobacter sulfurreducens.

7

Bioremediation of Toxic Waste

hydrocarbons

Electricity Generation

carbon dioxide
and water

Nuclear Waste Containment
Before (soluble nuclear waste)

After (insoluble nuclear waste)

Figure 1.3: Some of the applications of Geobacter. In this diagram, brown rods represent cells, and black filaments represent pili. All examples here show biofilms of Geobacter in water-containing sediments. Upper left: Bioremediation of chemically toxic waste.
Other bacteria break up complex organic waste into fermentation products [15], and Geobacter breaks it down still further. Upper right: Microbial fuel cell. The example pictured is
based on the “BUG” design for seafloor power generation, as in Reference [15]. Bottom:
Bioremediation of nuclear waste. Soluble nuclear waste (e.g., U(VI) oxide) is chemically
reduced and thereby rendered insoluble, at which point it precipitates out rather than perpetuating in groundwater.

8

Chapter 2
Scanning Probe Microscopy
The field of scanning probe microscopy covers a wide range of tools and techniques designed
to investigate the properties of surfaces at the nanoscale [30, 31]. Among the myriad techniques falling under this general heading are the original scanning tunneling microscopy and
other tip-based techniques such as atomic force microscopy. Unlike optical microscopy, the
tip-based scanning probe microscopies are not limited by considerations of the wavelength
of the photons which probe the sample but rather by the sharpness of the tips, making high
spatial resolution on the angstrom or nanometer scale much easier to achieve.
While the primary tools of investigation for this research were scanning tunneling microscopy (STM) and, particularly, its close relative point tunneling spectroscopy, atomic
force microscopy (AFM) was also used to obtain some of the topographs presented here.
For this reason, a brief introduction to AFM is included in the description of the scanning
probe microscopy methods used. The AFM scans discussed here were performed solely at
room temperature, but the STM and point tunneling spectroscopy studies were done both
at room temperature and at liquid nitrogen temperature (77 K) in order to investigate the
temperature-dependence of the measured properties, in particular the density of states of
the pilus nanowires produced by Geobacter sulfurreducens.
This Chapter discusses the underlying principles beneath the scanning probe microscopy
measurements performed in the course of this research. Specific details of the microscopes
used in this study are discussed in Chapter 3. Further details regarding some of the back9

ground for the scanning probe microscopy techniques used in this research are given in
Appendix D.

2.1

Key Aspects of Scanning Probe Microscopy

In its most general sense, scanning probe microscopy uses some type of probe which gives
resolution often on the (sub-)nanometer or micron scale and rasters it along or near the
sample to be scanned. Since the focus of this study was on electronic properties in a sample
as reflected in the density of states at the sample’s surface, the science of interest in this
study was amenable to the techniques of scanning tunneling microscopy, its near cousin point
tunneling spectroscopy, and atomic force microscopy.

2.1.1

Overview

For experiments such as those discussed here, the aim is to gain information on the nanoscale
properties of the system. While properties of bulk amounts of the sample can be measured
with mesoscale electrodes and their properties averaged (for the case of Geobacter pili, see,
e.g., Reference [32]), the optimal way to investigate the properties of a small sample such
as a nanowire is to probe one instance of the sample itself rather than an assembly of such.
For samples which are small and not easily positionable, it behooves the experimenter to
position the measurement device instead: if the sample won’t come to the probe, the probe
will come to the sample. This ability to reposition the probe is a chief strength of scanned
probe microscopies.
The positioning of the tip and sample with respect to each other is conventionally achieved
by mounting either the tip or the sample on a scanner made from a piezoelectric material.

10

(While some microscopes scan the sample with respect to a stationary tip, a larger number
of the microscopes used for this study scan the tip with respect to a stationary sample, so for
simplicity, the tip-scanning scenario will be used for the following description.) Since piezoelectric materials expand or contract in response to applied voltages, computer-controlled
electronic signals can be used to direct the motion of the tip by controlling the extension or
compression of elements of the piezoelectric supports. The tip is rastered along the sample
by being moved back and forth with some speed along one direction while the tip is slowly
moved in the orthogonal direction. Proper tip–sample separation is maintained in the z
direction by a feedback loop which, in the case of the standard “constant current” STM
mode, constantly compares the detected tunneling current to a predetermined setpoint. If
the tunneling current is insufficient or is too great, the tip is extended or retracted until the
detected tunneling current reaches the setpoint. Further discussion of this raster scanning
motion and the feedback is reserved for Appendix D.

2.1.2

STM and AFM Tips

Not all scanning probe microscopies use physical tips to obtain information about the sample,
but tip characteristics are critical for the techniques used in this study. The details of the
specific tips used for this research will be discussed in Chapter 3, but a few general remarks
will be mentioned here as to the properties of and common modes of confusion resulting
from STM and AFM tips.

Tip Properties
In order to achieve the nanometer-scale resolution used for these measurements, the tips must
be extremely sharp. In general, STM and some AFM techniques routinely achieve atomic
11

resolution; however, this was not necessary for this study. The STM tips used here were
mechanically cut from Pt:Ir wire (80% platinum, 20% iridium). Well-formed mechanically
cut tips have a radius of curvature at the tip apex of approximately several tens of nm [30],
although only the few tip atoms nearest the sample are relevant for STM.
During an STM experiment, the tip and sample are brought to a separation of approximately one nanometer, as shown schematically in Figure 2.1. This is close enough
that quantum tunneling of electrons between the tip and the sample, with the dominating directionality set by a voltage between the tip and the sample, can reach the relatively
easily-detectable nanoampere regime. While other portions of the tip may come within a
few nanometers of the surface as well, tunneling will be predominantly through the atom or
few atoms at the apex of the tip.

~ 1 nm

Tip
~ 1 nm

few tens
of nm

Sample
Figure 2.1: Schematic of an STM tip and sample at progressive layers of magnification
moving from left to right. Large circles in the right image represent atoms surrounded by
mobile electrons. The radius of curvature of a typical tip is several tens of nanometers. The
apex of a good tip is essentially a single atom.

12

Tip Artifacts
Because the tunneling current detected by STM depends exponentially on tip–sample separation, only the last few atoms of the STM tip are relevant. For small, flat scan areas, this
can even be only the single atom at the very apex of the tip [33]. When the scan incorporates
features of different heights, such as pili or graphite step edges, multiple “minitips” at the
apex of the tip can be as close as a few angstroms together [33] or several nanometers apart
from one another. These multiple tips are generally at different distances away from the
sample and, hence, contribute to the net image with different strengths. The configuration
of the multiple tips can often be deduced from the scan image itself. The number of multiple
minitips involved can be deduced from the number of times each feature in the image is
repeated. An example of a multiple tip causing repeated images is given in Figure 2.2, and
the topic is discussed further, with an even more egregious example of a multiple tip, in
Appendix B.

2.2

Scanning Tunneling Microscopy

Scanning tunneling microscopy [34] rasters an electrically conductive tip in close proximity
to (but not touching) a surface which is not a strong insulator. Electrons naturally quantum
mechanically tunnel between the tip and the sample, and by applying a bias voltage between
the tip and the sample, the preferential direction of tunneling electrons and the probability
of tunneling can be controlled. This is schematically shown in Figure 2.3.
When looking at feature sizes on the order of 1 nm, as done in this study, the shape of
the molecular orbital at the end of the tip can be ignored, and it can be assumed that the
wavefunction of the apex of the tip is a (spherical) s-orbital [30], meaning that the subatomic
13

Figure 2.2: Left: Diagram of a typical multiple tip. The feature on the sample is measured
first with the rightmost tip, then with the left tip, and the two values are summed together
to give the appearance of two features. The bottom trace is that attributable to a single tip;
the dashed and dotted traces are the superimposed images due to both the tips, and the top
trace is the result. The inset image is how the feature would appear to the experimenter.
Right: Pili and graphite flakes imaged by a multiple tip; scale bar is 200 nm. Imaged with
applied bias voltage + 0.5 V and tunneling current setpoint 0.1 nA. (+ 0.5 V, 0.1 nA)
shape of the tip does not contribute to the resolution of the image.

2.2.1

Review of Tunneling Through an Energy Barrier

This brief review of tunneling through a finite energy barrier is not only relevant to the
basic principle of STM operation but also to the schematic model for tunneling between
available energy states within the pili, as discussed in Chapter 4. Both for tunneling between
the tip and sample in STM and for tunneling from site to site along the nanowire, it is
sufficient to consider the one-dimensional case. Since a well-formed STM tip is extremely
sharp and has a sufficiently high aspect ratio, only the few atoms at the very apex of the tip
contribute to the signal, as will be discussed below, and for site-to-site tunneling through
nanowires, each piecewise step along the tunneling pathway can be considered to be along
one dimension. This discussion follows the simplest case described in Reference [30]; similar
14

Z
Y

charge-sensing
circuitry

X

Qtip

tip
sample surface
DC bias
voltage

Figure 2.3: Diagram of an STM experiment. An atomically sharp, electrically conductive
tip is rastered very near to an electrically non-insulating sample, which is biased at a voltage
value chosen by the operator. Electrons tunneling between the tip and the sample are
measured by sensitive charge-detection circuitry, and the z position of the tip is adjusted to
keep the detected tunneling current, proportional to the number of tunneled electrons per
second, constant.
and more involved discussions can be found in essentially any STM book, as well.
Start from the Schr¨dinger equation for one electron tunneling between the tip and sample
o
in the z direction,
d2
ψ(z) + U (z)ψ(z) = Eψ(z)
2m dz 2
2

−

(2.1)

with U (z) used here to denote the potential to avoid confusion with the voltage V. For
simplicity in the immediate discussion, consider the case where both tip and sample are
metals so the conduction electrons can move freely. Electrons are superpositions of plane
√
2m(E−U )
. The probability
waves inside the metals, ψ(z) = ψ(0) e±ikz , with k given by k =
of an electron being found inside this barrier decreases rapidly the further along z into the

15

barrier the electron is sought. The electron wave function decays exponentially with barrier
√
2m(U −E)
depth: ψ(z) = ψ(0) e−κz , where the decay constant κ =
. The tunneling current
can be described as the number of electrons per unit time which tunnel through the barrier
into the second metal, where they’re detected by the electronics in the rest of the apparatus.
Thus the tunneling current is exponentially dependent upon distance. The work function φ
of a material is the energy required to remove an electron from the material into the vacuum
(this is larger than the energy due to the applied bias in the STM system, e × V [30]), as
indicated in Figure 2.4. Using typical values for work functions of materials used in STM,
the tunneling current decays an order of magnitude for every angstrom of separation between
the materials [30]. This is a significant and frequently-invoked idea, as it is responsible for
the excellent lateral resolution of STM, the relevance of only the nearest few atoms to the
sample in STM tips, and the limited spatial ranges throughout biological samples in which
tunneling conductivity can be significant.

Energy

Sample

Vacuum

Tip

EF

z
Figure 2.4: Diagram of one-dimensional tunneling for the STM tip–sample system. Electron
wavefunctions are shown superimposed on gray-colored rectangles representing the electronic
states available within the materials. Electron wavefunction in each region shown in fuchsia.
Fermi energy of each material shown in cyan. The difference between Fermi levels of the
materials is shown by a cyan slab (this can be controlled by applying a bias voltage), and
the difference between work functions of the materials is shown by a brown slab.

Applying a bias voltage tilts the energy barrier such that one side of the barrier has
16

available empty states at the same energy as there are filled states on the other side of the
barrier. A tilted energy barrier also occurs due to tunneling between materials with different
work functions. This tilt is shown as the difference in heights of the gray backgrounds of
the different materials in Figure 2.4. In the case of a positive bias applied to the sample,
this means energy states in the sample are opened up for electrons from the tip to tunnel
into. Most scans performed during the course of this research were done with a positive
bias applied to the sample, which investigates the emptied states near the Fermi level of the
sample. A sketch of tunneling between the tip and the sample is given in Figure 2.5. An
applied bias voltage enhances the tunneling probability in a particular direction.

tunneling direction
e

tunneling direction
+ bias
voltage

e

e

e

tunneling direction
Tip
Sample

tunneling direction
Sample
Tip

Figure 2.5: Diagram of the effect of bias voltage on tunneling. Left: Both tip and sample
are grounded. Electrons can tunnel between tip and sample with equal probability, here
represented by the thickness of the curves. Right: Positive bias applied to the sample.
Electrons continue to tunnel in both directions, but the tip-to-sample tunneling probability
is enhanced by the applied positive bias. Since this is a tunneling process, these lines do not
represent travel paths; rather, the detected position of the electron changes discontinuously
from tip to sample or vice versa.

The details of the tunneling are perhaps clearer in terms of the local density of states
(LDOS) of the tip and the sample in an energy level picture. The term “density of states”
refers to the number of available quantum states for particles (considered here to be electrons
for simplicity) to inhabit as a function of energy. STM, by applying a bias voltage between
the sample and the tip, provides an energy equal to the charge of the electron multiplied
by the applied voltage. Electrons will fill states between the Fermi energy (zero applied
17

voltage, hence zero applied energy) and the energy corresponding to the bias voltage, hence
the STM signal integrates over the energies between the Fermi energy and the bias energy.
Assuming the density of states of the tip is constant throughout the scanned image and
only the LDOS of the sample can vary, and neglecting thermal effects on the Fermi-Dirac
distribution describing the electrons, the tunneling current signal
eV

I(V ) ∝
EF

∗
ψsample ( ) ψsample ( ) d =

eV
EF

LDOSsample ( ) d

(2.2)

where EF is the Fermi energy and eV is the energy corresponding to the applied bias voltage
V [30]. The number of electrons which tunnel at a given voltage in an STM experiment is a
function of the density of states in that energy range. When there are more states available
for the electrons to fill, the detected tunneling current will be greater in magnitude. The
tip is made of a material which has an innocuous density of states in the energy regimes
relevant to these experiments; consequently, any features seen in the observed tunneling
current are due to the properties of the sample. A sketch of the density of states of the tip
and the sample during an STM measurement is given in Figure 2.6. Further discussion of
the density of states is reserved for the description of point tunneling spectroscopy below.

2.2.2

Topography and Local Density of States Probed by STM

STM topographs are a convolution of the physical height of features on the sample and the
tip and of the local density of states of the sample and the tip. Ordinarily, the tip density
of states can be ignored unless the tip becomes contaminated by extraneous material. If the
tip is uncontaminated and is neither a multiple tip nor particularly blunt, its effect on the
resulting STM image can be ignored, and features in the STM topographs can be attributed
18

+0.25 V
Tip

-0.25 V

Sample

Tip

Sample

Figure 2.6: Diagram of the effect of bias voltage, showing the electronic states of the tip
and the sample. Shaded areas represent filled electronic states. Arrows denote electron net
tunneling direction. The tip has a smooth density of states. The sample density of states
here has discernible features. Left: Sample is positively biased with respect to the tip.
Electrons from the tip tunnel into sample states with energies between the Fermi energy and
the energy set by the applied bias voltage. Right: Sample is negatively biased. Electrons
from the sample tunnel into tip states between the Fermi energy and the bias energy.
to a convolution of the sample geometry and local density of states.
An example of STM topography is shown in Figure 2.7, along with a height profile along
part of the image. The high spatial resolution of STM permits topography of very small
features, including single atomic layers of graphite as in Figure 2.7. STM topographs are
generally labeled by the bias voltage applied to the sample and the tunneling current setpoint
used during the particular scan, thus: (+ 0.5 V, 0.1 nA).
250

nm

Height [nm]

0.3
0.2
0.1
0
0

0

nm

3A
one
atomic
layer
20

250

40
Length [nm]

60

Figure 2.7: Left: 250 nm × 250 nm STM topograph of a graphite step edge. The red line
marks the location of the height profile to the right. Right: Line section across the edge.
The blue points on the section are separated by the height of one atomic layer of graphite.
(+ 0.5 V, 0.2 nA)

19

The word “topograph” is used somewhat euphemistically, since electronic effects often
dominate the details of the topographs, as STM is sensitive to the entire wavefunction of the
sample, not just to the spatial positions of various elements of the sample. For this reason,
pronouncements about the true conformation of sample features must take into account
the local density of states information about the sample, either by acquiring topographs at
various bias voltages or by performing spectroscopy measurements [35]. Since the samples
studied in this work were not perfectly flat, the geometric properties of the sample terrain
also contributed greatly to the resulting topographs.

2.3

Point Tunneling Spectroscopy

A primary motivation for the development of STM was the quest for the ability to obtain
tunneling spectroscopy measurements with high spatial resolution [36, 37]. Point tunneling
spectroscopy, hereafter often referred to for brevity as “spectroscopy,” is a process whereby
a scanning tunneling microscope probe is parked in position at one point of a sample, the
bias voltage is ramped over a range, and the resultant tunneling current is monitored. (This
is the “tunneling” in “point tunneling spectroscopy.”)
Tunneling spectroscopy allows a nearly-direct view of the electronic levels in the sample,
since it measures tunneling current at a range of voltages instead of at a single bias voltage. The “nearly” qualifier is because the signal will be slightly obscured by convolution
with instrument properties and with statistical effects due to nonzero temperature, discussed in Chapter 5. Similar information could be obtained by acquiring STM whole-image
topographs at a range of bias voltages (voltage dependent imaging), but the spectroscopy
measurement approach allows the range of bias voltages to be investigated at each individual

20

selected point on the sample. (This is the “point” in “point tunneling spectroscopy.”) An
extended variant of this technique, scanning tunneling spectroscopy (STS), performs spectroscopic measurements at every point in the scan, as does the current imaging tunneling
spectroscopy (CITS) technique [38]. In this research, point tunneling spectroscopy was used
to probe the local density of states. Point tunneling spectroscopy is a subset of scanning
tunneling spectroscopy wherein the voltage is ramped and the local density of states is investigated at one location within an STM scan, rather than at every point. The primary
reason for choosing to perform point tunneling spectroscopy rather than scanning tunneling spectroscopy measurements with the apparatus used for this research is that, since the
measurement is done once per scan rather than at every pixel of the scan image, the scans
complete far more quickly, thus decreasing the impact of mechanical or thermal drift, which
could otherwise cause the target features to shift within the image. Since spatial dependence
of the spectroscopic results across the breadth of the pili was one of the research questions
of interest, minimizing drift was a strong incentive.

2.3.1

Density of States

STM and its associated spectroscopic techniques probe the density of electronic states in a
localized region of the sample, in addition to surface geometry in the case of STM topographic
imaging. The detected tunneling current signal is proportional to the number of quantum
states available for electrons to inhabit at a particular energy or range of energies. The
“density” in the term refers to a number density of electronic states available as a function
of energy. This provides valuable information about the electronic structure of the sample.
The density of states of a sample is probed slightly differently by point tunneling spectroscopy
and by STM. These methods integrate over the energy range between the Fermi energy and
21

whichever voltage is applied at the time; since this voltage is swept through a range during
spectroscopy, more energies can be more easily probed by spectroscopy.
In point spectroscopic density of states measurements, the number of states at each
energy is obtained, where the applied voltage selects the energies probed. Point tunneling
spectroscopy measurements like the ones discussed here differ from transport measurements
in that the tip is spatially separated from the sample, meaning the electrons must tunnel
across the gap between the tip and the sample rather than move through an electrical
contact with some contact resistance. Because of the exponential dependence of tunneling
current on distance, these measurements are localized to a small area at the surface of the
sample. The resultant measurement is the density of states, convolved with effects from the
nonzero temperature at which the measurement was performed, as well as by settings on the
experimental apparatus that affect the measurement’s energy resolution. Such smearings
cause the discrete, closely-spaced energy levels to appear experimentally as a continuum, as

Number of states

dI/dV
density of states

Energy

Number of states

sketched in Figure 2.8.

Energy [eV]

Applied voltage [V]

Figure 2.8: Diagram of density of states in point tunneling spectroscopy measurements.
Left: The number and width of the dashes at a given energy corresponds to the number of
available electronic states at that energy. The outline reflects that the observable which is
proportional to the density of states appears to be a continuous quantity, due to thermal
and experimental smearing of energy levels. Center: Same as left, with axes switched and
dashes removed. Right: Same as left and center, as it would appear experimentally.

22

2.3.2

Local Density of States Probed by Point Spectroscopy

The local density of states of the sample and tip are probed by point tunneling spectroscopy.
For sample feature sizes ≥ 1 nm, such as the pili studied here, the tip can be considered
to be cylindrically symmetric and its angular momentum orbital shapes neglected, thus
spectroscopy probes the electronic states of the sample for features of this size [30].
The “dynamic conductance” or “differential conductance” dI/dV is proportional to the
local density of states of the sample, as can be observed by differentiating Equation 2.2,
albeit features in the spectrum are smeared by thermal effects and equipment settings. This
can either be obtained by performing numerical derivatives of the I–V curves or by direct
measurement using a lock-in amplifier, discussed further in Appendix D. The relationship
of features in the dI/dV curves to one another is the useful information, not the absolute
magnitude of the differential conductance, save that dI/dV = 0 is an important landmark in
some dI/dV spectra. Because of this and of the subtleties involved in calibrating an absolute
dI/dV measurement, it is conventional to present these results in arbitrary units, plotted
against the swept bias voltage [30]. An alternate convention for displaying this information
is the “logarithmic density of states,” dI/dV / (I/V ) = dln(I)/dln(V ). The wary may note
that this quantity increases vigorously near zero voltage. This can be somewhat alleviated by
averaging over nearby values of the current [30] or by selecting one particular nonzero voltage
to calculate the (I/V ) for normalization [39], but rather than introduce more layers into the
data analysis, this work presents dI/dV data rather than the logarithmic density of states.
Furthermore, normalization issues in the differential conductance become predominantly
relevant for larger voltages than those used in this work.
An example of point tunneling spectroscopy on graphite is shown in Figure 2.9. The

23

amount of scatter in the data is normal and is partially influenced by the electronics used to
measure such small currents. This scatter adds substantial noise to the numerical derivatives
used to obtain dI/dV curves, so after the five immediately-sequential I–V curves are obtained
and averaged together, they are smoothed using a standard Savitzky-Golay algorithm before

d /dV [arbitrary units]

differentiation, as discussed in more detail in Appendix D.

+0.2
0

I [nA]

I

-0.2

0
-0.4

-0.2

0

+0.2 +0.4

-0.4

Applied Bias [V]

-0.2

0

+0.2

+0.4

Applied Bias [V]

Figure 2.9: Point tunneling spectroscopy on graphite. Black horizontal lines are at zero.
Left: Five I–V curves taken immediately in succession. Right: Numerically-differentiated
dI/dV of the average of those five curves, with 47-point Savitzky-Golay smoothing prior to
differentiation to stabilize the derivative. The extent of the voltage scale is reduced in order
to avoid artifacts of the derivative at the endpoints of the I–V curve.

2.3.3

Interpretation of dI/dV Curves

As mentioned above, the ordinate of the dI/dV curves is conventionally plotted using arbitrary units, as it is the relative heights of features in the dI/dV curves with respect to
one another that are meaningful. Zero is the only special value on the ordinate axis, as a
value of zero dI/dV for a particular voltage means there are no electronic states available

24

at that energy. If dI/dV is zero over a range of voltages, as in the example sketch shown
in Figure 2.10, this means in turn that I–V is constant over that voltage range, so adding
electrons to the system does not increase the observed current; this is characteristic behavior
of a semiconductor inside its bandgap or of an insulator.

Applied voltage

V3

I

dI/dV

V2

V

V1

V

Sample states
Figure 2.10: Diagrammatic I–V and dI/dV plots for the idealized semiconductor pictured
on the left. As the voltage is swept from V1 to V3 , the I–V curve behaves linearly in the
valence band of the semiconductor (dI/dV is constant), the I–V curve is zero for the range
of voltages inside the bandgap (dI/dV is zero), and the I–V curve is linear again (dI/dV is
constant) when the applied voltage reaches the conduction band of the semiconductor.

In analyzing dI/dV curves where gap-like features are observed in the density of states,
as will be discussed in Chapter 5, the edges of the gap are conventionally determined by the
voltage at which the dI/dV curve rises steeply above the flat gap bottom. This is remarked
upon here because often width determinations in science are done using the full width at half
maximum, and gap edge location is not conventionally determined that way except when
the logarithmic density of states, dI/dV / (I/V ) = dln(I)/dln(V ), is used instead [40].
Additionally, as is common in tip-based scanning probe techniques, the tip would not
infrequently change its characteristics semi-spontaneously. Since the actual measurement is
a convolution of the properties of the sample and of the tip, it was important to characterize
the tip frequently to ensure that artifacts were not causing spurious results such as those
25

highlighted in Appendix B. For example, if during a dI/dV measurement, a fragment
of the sample was attracted to the tip and remained affixed there afterwards, a control
measurement on graphite would then exhibit density of states characteristics of the sample.
For this reason, repeated control measurements on a portion of the sample with a wellknown density of states, such as the graphite substrate near the pili, are advisable and were
frequently performed in this work.

2.4

Atomic Force Microscopy

While the heart of the research discussed here lies in scanning tunneling microscopy and
in spectroscopy, some of the topographic scans presented here were obtained using atomic
force microscopy (AFM), so it will be briefly discussed here. Additionally, the technique is
intimately connected with the discussion in Appendix E. Only the amplitude modulation
mode of AFM scanning will be discussed in this Chapter, with discussion of conductive
AFM reserved for Appendix D. Originally developed by the team that invented the STM
[41], the atomic force microscope has progressed over the past nearly three decades into
a standard surface inspection tool. Just as standard STM topographs maintain a constant
current between the tip and the sample, AFM topographs maintain a constant force between
the tip and the sample.
Standard atomic force microscopes detect the interaction between tip and sample based
on the deflection of a laser beam reflected from the back of a cantilever on which the AFM
tip is mounted, as in Figure 2.11. The reflection of the laser beam obeys Snell’s law—the
angle of incidence is equal to the angle of reflection when the index of refraction is the same
throughout the light path—thus the angle at which the reflected beam hits the detector

26

depends on the angle of the cantilever back surface with respect to the incident laser beam.
The force between the tip and the sample can be calibrated using the spring constant of
the cantilever, determined by observing the resonance frequency of the cantilever when it is
just subject to Brownian motion, and taking a “force curve,” watching the deflection of the
cantilever as the tip is brought from far away from the surface to the point where it firmly
touches the surface and then is retracted away [42].

Tip-shaking
motor

Free

Laser
Laser
detector source

Interacting

Tip
Sample
Z motion
XY motion

Figure 2.11: Diagram of an atomic force microscope. The tip is at the end of a cantilever;
deflection of a laser reflected off the back of the cantilever is proportional to the cantilever
bending. Left: Schematic of AFM hardware. Right Top: When the tip is free of the surface,
the deflection of the reflected laser beam is determined by the angle at which the cantilever
is mounted and, if operated in amplitude modulation mode, the angle of the cantilever at
that point in its oscillation. Right Bottom: When the tip is interacting with the surface, an
additional deflection due to forces from tip–sample interactions is added.

Tapping mode atomic force microscopy, also called “AC mode” or “amplitude modulation” or “intermittent-contact” AFM, is the most commonly-used AFM mode today. The
cantilever is driven at or near its resonance frequency. The feedback acts on the amplitude
of the cantilever’s oscillation, as determined by the deflection of the position of the laser
beam reflected off the back of the cantilever, as detected in a postion-sensitive photodetec-

27

tor. When the tip interacts with the sample, this dampens the oscillation of the cantilever,
thus the atomic forces from the sample affect the amplitude of the cantilever’s oscillation,
which provides the data-giving signal. Contact mode, where the tip moves along the sample
surface in almost the exact same fashion as a record player needle moves along the record,
was not used for AFM imaging in this experiment because the tip was prone to pushing the
pili laterally around on the surface when operated in this mode.
Amplitude modulation AFM can be likened to dribbling a basketball. The feedback
mechanism for the basketball player is the amount of force needed to maintain the ball
at the same height when it bounces back. Similarly, the feedback mechanism for amplitude
modulation AFM is the amplitude of the sinusoidal voltage signal applied to the piezoelectric
element controlling the cantilever and tip motion in order to make it oscillate. The damping
of the amplitude, the need to apply more driving force to make it reach the desired height
or “setpoint” amplitude at the top of its oscillation, gives information about the properties
of the surface with which it is interacting. If one dribbles a basketball down a set of stairs,
one finds that one must apply more force to the ball to make it rebound to the same height.
Similarly, the AFM signal gives information about the height of features on the sample.
Dribbling a basketball in the mud instead of the driveway requires more force and delays
the basketball in its return to its setpoint height; similarly, the AFM “phase” signal gives
information about the relative softness or the adhesion of materials. In this work, the phase
signal was useful for ascertaining the location of the biological material on the hard substrate.
With these scanning probe microscopy techniques in hand, specific implementations of
these microscopes (discussed in Chapter 3) can be used on the pilus nanowires expressed by
Geobacter sulfurreducens (discussed in Chapter 4) to gain information about the density of
states of that system at ambient and cryogenic temperatures (discussed in Chapter 5).
28

Chapter 3
Details of Experimental Apparatus
Moving from the general principles in Chapter 2 to specific instrumentation, this Chapter
presents an overview of the experimental equipment used in this research. The cryogeniccapable scanning tunneling microscope used for this study was built at Michigan State
University by Sergei Urazhdin and Aleksandra Tomic and operated using an RHK SPM
100 controller. While some of the research discussed here was done at room temperature,
the majority of the data came from measurements taken with the microscope and sample
immersed in liquid nitrogen. Additional room temperature topography was obtained using
a Cypher scanning probe microscope in STM and AFM modes.

3.1

STM Hardware Configuration

A plethora of strategies and philosophies are used in the design of scanning tunneling microscopes. The cryogenic-capable STM used here followed the design developed by Besocke
et al. [43, 44] for tip–sample coarse approach, chosen for its thermal compensation and
noise-resistant qualities, since the microscope was designed for cryogenic experiments.

3.1.1

Besocke Design

The Besocke, or beetle, STM design [43] was selected for the microscopes in this laboratory
because of its superior thermal compensation. Since all the piezotubes (discussed further
29

in Appendix D) in the microscope are the same size and are mounted on the same structure, they expand or contract identically with temperature changes. Because the piezotubes
supporting the sample and, separately, the tip, are all mounted on the same rigid base,
microscope vibrations affect them all equally, which makes the design vibration-resistant as
well [30]. The “business end” of the cryogenic-capable STM used here is shown in Figure 3.1.

Figure 3.1: Besocke-design STM with ramps. Left: Schematic of the design. Right: Photograph of the microscope with no ramps in place. The caps atop all three “walking” piezotubes
can be seen, as well as the “scanning” piezotube in the center on which a tip is mounted.

The coarse approach between the tip and the sample in this design uses inertial translation
to advance or retract the sample, which is mounted on a set of ramps [44]. Essentially, the
overall shape of the sample motion is screw-like. The process of inertial translation is shown
in Figure 3.2. A voltage is applied to the “walking” piezotubes, causing them to bend
like an elephant’s trunk; the ramps, supported by the stainless steel balls atop the walking
piezotubes, are rotated gently by this motion. The voltage on the walking piezotubes is then
abruptly returned to zero, causing the walking piezotubes to snap back to their equilibrium
position. This occurs so quickly that the ramps are no longer in contact with the walking
piezotube caps and don’t move, as in the magician’s tablecloth trick. This means that a

30

different part of the ramps then sits atop the stainless steel balls summiting the walking
piezotubes, which corresponds to the ramps advancing or retreating a small amount.

Sample

Walking

Before

Ramps
Tip

Piezotubes

After

+V
Tip

Tip

-V
Net sample motion
with respect to tip

Figure 3.2: Diagram of the inertial translation process, shown in two dimensions and applied
to a lateral translation for simplicity of demonstration. Motion is exaggerated for visibility.
A voltage is applied to the quadrants of the walking piezotubes, causing them to bend while
they carry the ramps on their tops. The voltage is then suddenly switched off, causing the
piezotubes to straighten up again while the ramps and sample retain the offset.

Micrometer-scale sample translation in the lateral (X,Y) directions was also performed
via the inertial translation, or stick–slip method, as shown in Figure 3.2. By this process,
several square millimeters of the sample could be probed, although experiments generally
ended due to irrecoverable tip damage before this much area could be scanned. This method
does not necessarily result in reproducible translations over many-micron distances, since
there is some variance in the motion due to the interplay between sample lateral and z
motion and to slight irregularities on the ramps. Nanometer-scale offsetting of the scanning

31

region was done by applying lateral offset voltages to the scanning piezotube quadrants.

3.1.2

Tip Specifics

STM tips must be made of conductive material that does not form an insulating oxide layer
on any timescale during which the experiment would be performed and which preferably has
a simple, or at least known, density of states so that states in the tip do not obscure states
in the sample. The tips used for this experiment were an alloy of platinum and iridium in an
80:20 Pt:Ir ratio. This is a conventional material, frequently used for STM tips, possessed
of these desired characteristics.
To obtain good resolution, it is important to have tips with apexes which are sharp on
the atomic scale. This can be achieved a not inconsiderable fraction of the time by cutting
the tips mechanically with a pair of diagonal cutters. For optimal cutting, the wire should
not be entirely severed by the cutters, but rather the cutters serve to weaken the wire while
a pulling motion applied to the cutters stretches the wire until it breaks. The procedure
used to cut most of the tips used in this experiment is demonstrated in Reference [45].
After mechanically cutting the STM tip wire, tips were affixed to mounting chips before
being loaded onto the cryogenic-capable STM. A mounted tip is shown in Figure 3.3. The
mounting chips were made from standard silicon wafers cleaved to be a few millimeters on
a side; silicon chips were used to prevent electrical shorts between the tip and the cap of
the scanning piezotube, upon which the tip chip was to be mounted. A small amount of
cryogenic-compatible conductive epoxy was dabbed onto the chip. A piece of thin gold wire,
to be used for connecting the tip to the coaxial tunneling current wire at the microscope,
was cut and lain into the epoxy, then the tip was lain into the resulting epoxy cushion. After
curing the epoxy, the tip chip was affixed to the cap of the scanning piezotube and the gold
32

wire was soldered to the coaxial tunneling current wire using indium solder. In this way, the
tip and tunneling current wire were electrically connected.

Tip Au wire

Ag epoxy

In solder

Chip
Cap of
scanning
piezotube

Tunneling
current
wire

Figure 3.3: An STM tip in place on the microscope. Left: Photograph of a mounted tip.
Center: Optical microscope photograph of Elizabeth, the tip used to acquire the clearest
low-temperature data. The radius of curvature of the apex of a good STM tip is only a few
to a few tens of nanometers, so the apex is not visible via optical microscopy. Right: Labeled
components of a mounted tip.

3.1.3

Vibration Isolation

It is traditional in scanning probe microscopy measurements to suspend or support a microscope using a heavy mass and a soft spring, such that the resonant frequency of the
microscope and support is far below that of typical excitations due to building vibrations.
Extensive vibration isolation systems are not always necessary. For example, atomic resolution of the graphite lattice was routinely obtained using the portable EasyScan STM when it
was placed on a sturdy table, including in a high school classroom during an outreach event,
as shown in Figure 3.4. The commercial EasyScan STM is mounted on a heavy, marble slab
with vibration-damping soft feet beneath. For the research done with the cryogenic-capable
STM, the microscope was mounted on a helium storage dewar, discussed in more detail
below in the context of cryogenic operations. This dewar was filled with liquid nitrogen for
the 77 K experiments, for which experiments the microscope was lowered into the dewar.
33

For room-temperature measurements, the microscope was mounted atop the floor-secured
dewar for mechanical stability purposes. The wheels of this portable storage dewar were
secured with cardboard shims wedged beneath them and duct tape adhering them to the
floor in order to stabilize the dewar. Sufficient vibration isolation was attained with this
arrangement to enable atomic resolution of graphite at 77 K.

Figure 3.4: Left: Graphite lattice imaged using the tabletop EasyScan STM at an outreach
event. Scan size is 1.95 nm × 1.95 nm. (+ 0.1 V, 1 nA) Right: Photograph of the tabletop
STM in situ in a high school classroom.

3.2

STM Calibration

Due to time or applied voltage, the piezoelectric material used in the piezotubes will depolarize to some extent, that is, extend or contract less for a given applied voltage [30]. For
this reason, the cryogenic-capable STM used in this study was recalibrated.
Calibration of the correspondence between voltage applied to the scanning piezotube’s
quadrants and actual distance in nanometers along the sample traversed by the tip in the X
and Y dimensions was done on a sample of graphite at room temperature. The microscope
was mounted on a dewar suspended off the ground using bungee cords to decouple it best from
floor vibrations. Knowing the lattice parameters of graphite and measuring the apparent
34

distances between atoms in an atomic-resolution scan of graphite such as that in Figure 3.5,
the conversion factors of 17.15 nm/V in the X direction and 20.88 nm/V in the Y direction
were obtained. Atomic resolution STM scans of graphite only show half of the atoms on the
top surface of the graphite; the bright spots in the atomic resolution image in Figure 3.5 are
those atoms on the surface without another graphite atom immediately beneath them. The
layers which comprise graphite are stacked on top of one another in such a way that every
other atom in a given sheet is atop an atom in the sheet below it, while the other atoms do
not have a Z-direction neighbor until two layers below them. Care should be taken when
performing the XY calibration via atomic resolution on graphite, lest an artifactual hexagonal
pattern be mistaken for the actual lattice structure (see Appendix B and Reference [46]).

Figure 3.5: Left: Most of a platinum pit, imaged primarily to calibrate the scanning piezotube extension in the Z direction for a given voltage. The effect of a multiple tip can be seen
by the inverse ziggurat shape of the pit. Streaky specks are due to particulate debris on the
sample surface. Right: Hexagonal graphite lattice imaged with atomic resolution to obtain
the XY calibration for the scanning piezotube. (+ 0.1 V, 1 nA)

The Z calibration was determined by scanning a grid of pits 180 nm deep etched out

35

of a silicon substrate and coated with platinum (see Figure 3.5), giving a Z calibration of
4.60 nm/V. To check the Z calibration, single steps of graphite were observed and checked
against the literature values for graphite layer spacing [47].

3.3

Cryogenic Temperature

STM and, particularly, spectroscopy experiments were done primarily at 77 K, liquid nitrogen
temperature. These measurements were done with the microscope submerged in a dewar
capable of containing liquid helium but used in this instance with liquid nitrogen. The
dewar was on wheels with taped cardboard shims for balance and stability, as seen in the
apparatus photographs in Figure 3.6. After initially pumping out the microscope and stick
area to a vacuum on the order of a few to tens of microtorr, the microscope and stick area were
backfilled with nitrogen gas to act as an exchange gas for temperature equilibration. The
microscope was slowly lowered into the dewar while monitoring the microscope temperature
with an Allen-Bradley carbon resistor thermometer [48, 49] to ensure that thermal changes
were gradual enough to prevent microscope damage.
When cooled to 77 K, the range of the piezoelectric elements in each direction decreases
by a factor of approximately 3 [30]. In order to determine this factor with more precision,
atomic resolution images of graphite were obtained at 77 K, and the XY calibration was
redetermined at this temperature. Using the known spacing between points on the graphite
lattice of 0.246 nm [47], conversion between apparent and actual atomic spacings gave a
thermal contraction conversion factor

actual distance at 77 K = 0.356 × apparent distance

36

(3.1)

Microscope
stick

Dewar

Figure 3.6: Left: Data acquisition setup for STM and spectroscopy. The microscope is at
the bottom of the long stick and is submerged in the dewar (partially submerged, in this
example). The dewar extends below the bottom edge of the photograph. Right: Lower part
of the dewar, showing the shims and adhesive used for stability. The bright green tape is a
safety measure to warn passersby against bumping into the apparatus.
under the fair assumption that thermal contraction affects all directions of the piezotube equally.

3.4

Cypher Scanning Probe Microscope

Further room temperature measurements for verification of the STM and spectroscopy results, as well as AFM investigations of topography and deposition characteristics, were performed using an Asylum Research Cypher scanning probe microscope. While the Cypher
is predominantly used as an atomic force microscope, it is capable of a variety of scanning
probe measurement techniques, including STM and basic point tunneling spectroscopy.

3.4.1

AFM

Observations were made of the pili deposited on different materials such as gold and thermallyoxidized silicon using amplitude modulation atomic force microscopy. This is referred to by

37

Asylum Research for their equipment as “AC Mode.” This mode can work in either the noncontact (soft tapping) or intermittent-contact (hard tapping) regimes for interacting with
the surface. Most of the AFM topographic images, such as those seen in Figure 3.7, were
obtained using intermittent-contact scanning which, while interacting more strongly with
the sample surface, gives higher resolution of the sample topography and more accurate determinations of feature widths. The AFM tips used for this imaging had cantilever spring
constants of nominally either 2 N/m or a few tens of N/m.

SiO2

Au
Pili

Au

SiO2 Au

SiO2

Figure 3.7: AFM topographs obtained using the Cypher. Fabrication of these samples is
discussed in Appendix E. Left: 10 µm × 10 µm scan size. Center: 4.5 µm × 4.5 µm scan.
Right: 5 µm × 5 µm scan. All scans shown here have a z range grayscale extent of 55 nm.

3.4.2

STM

The implementation of STM on the Cypher, as of software version 111111 which was the
stable version at the time this research was performed, did not allow some of the most
useful conventional methods of STM tip recovery but had a far faster tip change procedure
compared to the cryogenic-capable STM. The tip holder could be removed, the tip could
be recut or replaced, and the tip holder could be repositioned all within two minutes. By
contrast, tip replacement on the cryogenic-capable STM took longer, generally on the order

38

of fifteen minutes, predominantly because of soldering time and logistical considerations.
This is not solely a matter of convenience; it is also desirable to minimize the amount of
time the sample spends outside of the inert nitrogen atmosphere.
There are a few main advantages of the cryogenic-capable STM over the Cypher STM.
The Cypher enclosure, while it can be purged with nitrogen, cannot be pumped out to
achieve vacuum. This means the nitrogen atmosphere is achieved through mixing and displacement of the ambient air inside the enclosure, which is slower than pumping out the air
and replacing it with nitrogen gas, as was done with the cryogenic-capable STM. Similarly,
it cannot be cooled to cryogenic temperatures. This means that this complete temperaturedependence study could not have been performed using the Cypher. Furthermore, the numerical dI/dV curves were often prohibitively noisy. Consequently, the room temperature
dI/dV curves obtained using the cryogenic-capable STM were more useful than those obtained using the Cypher STM. Partially counterbalancing these disadvantages, the Cypher
STM tip change procedure is fast, as previously mentioned, and requires neither epoxy nor
soldering. Additionally, the sample can be moved more reproducibly over longer distances,
leading to improved sample positioning.

3.5

Overview of Sample Preparation

A brief description of sample preparation for the experiments presented in Chapter 5 is given
here. For a more detailed discussion of the sample preparation, including refinement of the
deposition protocol, see Appendix C.
Geobacter pili were purified by Dr. Sanela Lampa-Pastirk of the MSU Department of
Microbiology and Molecular Genetics. 10 microliters of a suspension of purified pili in doubly

39

deionized water were deposited on a freshly-cleaved graphite substrate. The substrate was
mounted on a stainless steel disk attached to stainless steel ramps, discussed above in the
context of sample translation, in order to deliver bias voltage to the sample. After leaving
the droplet to physisorb onto the surface for ten minutes, the droplet was wicked dry. Two
sequential rinses of 10 µL of doubly deionized water were deposited onto the surface, left
for five minutes, then wicked dry. The sample was then dried under a stream of nitrogen
gas. Despite this, the sample remained slightly hydrated due to the persistence of water on
surfaces [50]. Although the experiments were not done in solution, the samples were most
likely covered with a few-nanometer layer of water [50], resulting in a hydrated environment
relevant to the natural state of the protein.

40

Chapter 4
Biological and Theoretical Context
To provide context for the measurements presented in Chapter 5, it is important to have a
basic understanding of the pilus structure of Geobacter sulfurreducens. The structure of the
pilin subunit plays an important role in determining the density of states of the entire pilus.
To this end, density functional theory calculations performed by Professor Gustavo Feliciano
on the Geobacter pilin are introduced for comparison to the measured dI/dV results.
Geobacter sulfurreducens is an anaerobic bacterium initially discovered in hydrocarboncontaminated near-surface sediments in Oklahoma in 1994 [10]. The bacterial cell itself is
two or three micrometers long and approximately 0.5 µm wide [10]. As part of its respiration,
it breaks up substances such as acetate and develops an excess of electrons internal to the
cell. If the cell membrane is in direct contact with an electron acceptor in the environment,
this excess charge can be transferred through the cell membrane into the electron acceptor.
If the cell is not in direct contact with an electron acceptor, for example if it is a cell stacked
in the middle of other cells in a biofilm or if the only available electron acceptors are insoluble and far away, it must find other methods of ridding itself of these excess electrons.
If electron shuttles, molecules which serve as intermediaries by diffusing through the environment, becoming reduced by bacteria, and in turn reducing other electron acceptors in
the environment [17], are present in the environment, Geobacter will take advantage of them
[18], but Geobacter does not produce its own electron shuttles [19]. Instead, Geobacter grows
electrically conductive type IVa-like pili to contact electron acceptors in its environment.
41

The aim of this overall research program is to investigate the electronic properties of these
pili both from a materials science perspective and to address biological questions about the
mechanisms of long-range electron transfer through conductive proteins. This line of inquiry
is pursued in this work through investigation of the electronic structure of these pili.

4.1

Pili Expressed by Geobacter sulfurreducens

Like other Gram-negative bacteria, bacteria which have a membrane outside of their cell wall,
G. sulfurreducens grows type IV pili [26]. Unlike many such pili found in other organisms,
but not uniquely, the pili that G. sulfurreducens grows are electrically conductive, both along
their lengths [25] and transversally across the diameter of the pilus [3], which is the subject
of the research described here. In other bacteria, pili with similar structure are used for
propulsion or anchoring or for gene transfer between bacteria [20]. Geobacter appears to
use its pili primarily for electron transfer, to attach the cells to electron acceptors [3], and
to link cells together to stack in sturdy biofilms [21]. Additionally, Geobacter and similar
species [4, 5] use these pili for electron transfer to spatially-distant electron acceptors. These
pili grow up to many microns in length and are formed from pilin subunits polymerized by
a molecular motor in the cell membrane and assembled into a helical pilus.

4.1.1

Pilus Structure

Pili are composed of subunits, called pilins, which interlock with one another in a helical
structure, generally with four pilins per coil of the helix; each coil repeats every approximately
4 nm [51]. The pilins are assembled by the bacterium in its inner cell membrane and pushed
through a pore in the outer cell membrane, through which they can then be retracted and

42

disassembled again to recycle the pilins, if so desired by the bacterium, at considerable
speeds—for the structurally similar, but electrically insulating, Pseudomonas aeruginosa pili,
at rates of approximately 0.5 µm of pilus extension or retraction per second [6]. Figure 4.1
shows an artist’s rendition of individual pilins being inserted onto an assembled pilus. Each
bacterium can produce multiple pili, particularly when grown under conditions with a large
fuel source and limited access to electron acceptors in the environment. A single pilus is
sufficient to discharge a Geobacter bacterium [24, 25] if it can contact an electron acceptor.

pore

outer
membrane

one turn
assembly

inner
membrane

four pilins
assembled

hydrophobic end
(pilus interior)
head region
(pilus exterior)

pilins

Figure 4.1: Artist’s conception of pilus assembly and extension at the cell membrane, inspired
by and simplified from [6]. For simplicity, proteins are represented schematically. Pilin
subunits dispersed throughout the inner membrane area are assembled by another protein
into a helical structure, then pushed through a pore in the outer cell membrane. The
hydrophobic center of the assembled pilus is a tube formed by the long α-helices of the
pilins, leaving a hollow core to the pilus, and the heads of the pilins remain on the outside
of the pilus to interact with its environment. In Geobacter, the globular head which other
pilins have is replaced by a short, barb-like shape [8, 52, 53].

The structure of the individual pilins has very recently been solved using NMR [52, 53],
43

but it is primarily known through theoretical studies [8, 54]. In contrast to other type
IVa pilins, which are a close analogue to Geobacter ’s pilin structure, the pilin subunit of
G. sulfurreducens pili does not have a large globular head but instead has only a small barb
protruding at an angle from the α-helix comprising most of the body of the pilin, as seen
in Figure 4.2, which is based on the published coordinates of the theoretically calculated
structure from Reference [8]. Because it is not obstructed by a large globular head, the
pilin’s native dipole moment remains discernible (the positive part of the dipole is inside
the assembled pilus), although that is modified somewhat by the surrounding water in its
environment [8].

exterior of pilus
interior of pilus

Geobacter sulfurreducens pilin subunit
Figure 4.2: Structure of the G. sulfurreducens pilin subunit based on a homology model. Coordinates were obtained from Reference [8] and displayed using DeepView (Swiss-PdbViewer)
version 4.1.0 [7]. The α-helix tail of the pilin is buried inside the assembled pilus, while the
small barb-like head remains on the outside of the pilus.

A recent study which used NMR to determine the Geobacter pilin structure at room temperature in solution [52] determined several relevant properties of the pilin which may even-

44

tually shed considerable light on the specific pathway of charge transport through Geobacter pili. Two parts of the pilin in particular were found to be rather flexible—NMR measurements are averages over ensembles of many molecules, so they explore the physical
configuration space of the molecules, and this measurement observed a variety of relative
angles between the barb and the α-helix.

red:
negative
blue:
positive

particularly
flexible
areas

star:
tyrosine

Figure 4.3: Structure of the G. sulfurreducens pilin subunit based on solution NMR [52].
Coordinates were obtained from Reference [53] and displayed using Swiss-PdbViewer [7].
Left: The NMR measurements yielded several slightly different conformations for some parts
of the pilin, suggesting flexibility in these regions [52]. Right: Same pilin, viewed from a
different angle. Electrostatic potentials due to charged amino acids are superimposed on
the pilin structure. Electrostatic potentials were calculated in DeepView using the Coulomb
interaction and assuming a dielectric constant of 80 for the surrounding medium and 4 for
the pilin, as in Reference [25]. Tyrosine amino acids are marked with stars.

4.1.2

Implications of Pilus Structure for Experimental Design

Pilins have a net charge of -2e, where e is the magnitude of the electron charge, at pH 7, the
nominal pH of water [8]. Type IV pili in general are known to form aggregates, or bundles,
due to interactions between patches of static charge distribution on the surfaces of nearby
45

pili [51]. Commonly on the samples used throughout this research, pili would tend to form
clumps, out of which only a few individual tendrils would extend, or to entwine with another
in rope-like structures. In addition to the STM and AFM observations of large aggregates
as in Figure 4.4, optical microscopy also showed the presence of even larger conglomerations
of sample substance. For the experiments described here, then, it was necessary to take
steps to break up such large aggregates. Suspensions of pili in doubly deionized water were
stirred by pipetting the solution up and down multiple times before deposition. A general
discussion of sample preparation is given in Appendix C.

100 nm

clump

large
bundles

50 nm

entwined pili
single pili

0 nm

0 nm

single pili

clump

Figure 4.4: Fibrous bundles, likely predominantly composed of aggregated pili and perhaps
also incorporating trapped fluid or contaminants. Left: A large “haystack” of bundles on
gold, with a small region of silicon dioxide surface in the lower right corner of the image.
Right: Expanded view of the image shown in Figure 1.2 showing both clumping and entwining types of bundling. Both scans are atomic force micrographs. Scale bars are 1 µm.

4.2

Charge Transport Mechanisms in Proteins

The motion of charged particles through proteins can follow different paradigms in different
materials. It is not yet concretely established what the charge carrier is in G. sulfurre-

46

ducens pili, although one experiment showed increased electrical conductivity of biofilms
when pH was lowered—more positively ionized hydrogen ions were added to the environment, hence attracting away electrons from the pili—suggesting that holes, rather than
electrons, are the charge carriers in these nanowires [32]. Additionally, proton conduction
is also a common concept in the biological and molecular chemistry electron transfer community; since positively-ionized hydrogen atoms are the smallest non-electron component
of proteins, if any part of the protein structure must detach itself to move, these will be
the ions to do so. In another form of charge transport seen in proteins, particularly in proteins in which tyrosine amino acids are important contributors to the charge transport [55],
positively-ionized hydrogen atom (proton) motion can be associated with electron motion in
charge transport in the process known as proton-coupled electron transport, in which either
both an electron and a proton move at the same time or the transfer of one lowers the potential energy barrier for the transfer of the other [56]. For the purposes of this discussion, the
charge carriers in pili will not be positively identified as electrons, holes, or other, but simply
considered in a generic sense. Extensive speculation will not be ventured as to the nature
of charge transport through the pili, since the work presented here focused most strongly on
observing the electronic structure.
In the end, no matter what the underlying charge transport mechanism is in the pili, STM
and tunneling spectroscopy are sensitive to the density of states at the tipwards surface of the
protein, so many of the transport-related details will not impact the measurements discussed
here. The chief observable result of different charge transport mechanisms in proteins in these
measurements is due to the temperature-dependence of the study; at 77 K, the temperaturedependent transport mechanisms will contribute much less to the available energy levels at
the protein’s surface than they would at 300 K. Density of states experiments like those
47

presented here give information about the available energy levels in a sample; transport
measurements would address how the charge carriers would reach those energy levels.

4.2.1

Tunneling

One way in which charge carriers can move from site to site through a protein is by means
of quantum tunneling through the potential barriers between sites. This process does not
depend on the temperature of the system, provided that the barrier is high enough or the
temperature is low enough that the thermal energy does not allow the charge carriers to
travel “over” the barrier in the classical sense and they must instead tunnel “through.”
This does not include superexchange, or thermally-assisted tunneling, which is in part a
temperature-dependent process. Since the probability of tunneling depends exponentially
on distance, this process is limited to short distances on the order of the diameter of the
pilus. Single-step tunneling across 2 nm distances is not overly improbable but instead is
a biologically-relevant process [57]. Since the diameter of the pili is on the order of this
value, it is not beyond belief that a pure tunneling pathway could exist between “top” and
“bottom” of the pilus breadth.

4.2.2

Temperature-Dependent Charge Transport

Multiple mechanisms for charge transport through proteins exist which can be “helped
along,” or even made at all feasible, with the addition of thermal energy from the environment. This can either provide the charge carriers with enough energy to “hop” over
potential barriers between sites on the protein (charge hopping) or, by providing virtual
intermediate states, extend the range of distances over which tunneling processes can occur

48

(superexchange, or thermally-assisted tunneling). In many proteins upon which temperature
dependence of conductivity studies have been performed at room and cryogenic temperatures, charge transport rates decrease upon cooling of the protein to cryogenic temperatures
[58]. By contrast, increased conductivity of a different protein electron transfer system upon
decrease in temperature to 170 K was attributed by the authors of that study to the protein
compactifying when it was cooled [57]. This serves as a reminder that many factors, not
least including protein conformation changes with temperature, can contribute to the conductivity through these systems. Since proteins are generally less stable in their structure
than most traditional metallic or semiconducting nanowires—in ambient conditions, protein
structures often fluctuate slightly—the transport properties depend in part on the protein
structure at a given temperature.
Recent work from Malvankar et al. proposes a model by which G. sulfurreducens nanowires
are conductive in the same way as a metal is [32], where the charge carriers are delocalized,
in contrast to the charge transfer mechanisms discussed above, wherein charges move from
locale to locale through the nanowire. This is based on observations of increased conductivity through G. sulfurreducens biofilms and mats of purified pili similar to those shown in
Figure C.2 as the temperature was lowered to 200 K, which the authors interpret as evidence
that charge carriers were able to move for greater distances without scattering off of phonons
from room temperature heating of the pili [32].
Since these temperature-dependent charge transport mechanisms are multitudinous in
form and are significantly decreased in relevance at the cryogenic temperature of most interest to the research presented here, the following discussion generally considers them generically as temperature-dependent processes rather than delving into the details of the specific
temperature-dependent charge transport mechanism(s) involved in pilus conductivity.
49

4.2.3

Marcus Theory

Charge transport through pili can be, at least to a first approximation, described by Marcus theory, which is a typical starting point for considerations of charge transport in proteins. This describes charge transport as a product of the contributions from tunneling and
temperature-dependent processes. The famous “Marcus equation” is given in Equation 4.1,
in terms of the electron transfer rate kET , which is proportional to the current.

kET =

2π

|HDA |2

1
(λ + ∆G0 )2
exp −
4λkB T
4πλkB T

(4.1)

In this equation, |HDA | describes the coupling of the donor and acceptor, which, in
this case, would be different components of the pilus. The tunneling part of the Marcus
equation, 2π |HDA |2 , is independent of temperature, neglecting any conformational changes
to the pilus, thus it will underlie both the 77 K and the room temperature conductivity in
the system studied here. In contrast, the latter part of the Marcus equation, all that is not
tunneling, depends strongly and explicitly on temperature.
At room temperature, akin to the natural thermal environment of the pili when attached
to living bacteria, both tunneling and temperature-dependent processes can contribute to the
conductivity of the pili. In order to determine whether tunneling is a meaningful contributor
to the conductivity, a substantial change in temperature is necessary. Cooling the pili reduces
the contribution of temperature-dependent processes to the overall conductivity but does
not affect tunneling. This simplified description neglects conformational changes which may
change the distances over which tunneling would occur.
Besides temperature, the adjustable parameters in the Marcus equation are the reorganization energy λ describing the energy needed for the nearby charges in the protein and

50

in its surrounding water to move to accommodate the new electrostatic scenario once the
charge has transferred [56] and the driving energy of the reaction ∆G0 . For the purpose of
this temperature dependence estimate to investigate implications for the research program
discussed here, the reorganization energy λ is taken to be a constant of temperature. If
the reorganization energy increases upon cooling, the effect of cooling on the significance
of temperature-dependent mechanisms will become even greater. The value of 0.2 eV for λ
used for this estimate is obtained from a study on Shewanella oneidensis, a similar organism
which produces more or less similar protein nanowires [59]. The free energy ∆G0 is taken to
be zero, since it is small compared to λ in the discussion of Shewanella nanowires [59]. Note
that kB T is 25 meV at room temperature and 6.6 meV at 77 K. It has long been known
and shown that freezing proteins to cryogenic temperatures changes the reorganization energy and the free energy driving the charge transport reaction [60]. Thus in addition to the
explicit temperature dependence on kB T , both λ and ∆G0 are affected by the freezing of
the protein and the thin layer of water surrounding the pilus, and the free energy changes
because the entropy decreases with temperature [60].
The current through the pilus is proportional to the electron transfer rate, I ∝ kET .
Therefore, with reference to Equation 4.1,

I77K
k
= 77K = 0.007
I295K
k295K

(4.2)

Thus I77K = 0.7% I295K —under the aforementioned assumptions, the predicted tunneling current will be lower by two orders of magnitude at 77 K than at room temperature.
Temperature dependence studies of materials using STM and tunneling spectroscopy
tend to focus particularly on systems which have some form of transition and to make small

51

temperature steps around that transition temperature, whether it is a transition between
normal metal and superconductor [39], metal and insulator (or better-conducting semiconductor and poorer-conducting semiconductor) [40], charge (or spin) density wave behavior
change [61, 62], and so forth. It is not expected that there is a discrete transition temperature
at which temperature-dependent conductivity mechanisms in proteins turn off completely.
Marcus theory, for example, describes a smooth decrease in temperature-dependent conductivity as the sample is cooled. Consequently, such fine-detailed temperature gradations
around a transition temperature as are common in the aforementioned type of studies are not
necessary, and two widely-spaced temperatures, 77 K and room temperature, are sufficent
for sampling the electronic structure of Geobacter pili.

4.2.4

Experimental Signatures

The change in current through the pilus with temperature, estimated above using Marcus
theory, could potentially affect the ability of the pilus to be conductive enough at 77 K to be
imaged stably using STM. The upper limit on the resistance of an object to be imaged is set
by Ohm’s law, with the resistance of the sample in series with the effective resistance of the
tunneling gap, as in Equation 4.3. The total resistance between the tip and the sample must
be sufficiently small that the applied bias voltage will induce a tunneling current detectable
by the electronics, as sketched in Figure 4.5, for values of “small” in the tens of gigaohms.
Early STM studies on protein-coated DNA [63] substantiate the claim that the tunneling
resistance in this type of measurement is dominated by the tip–sample gap resistance and
that charge was conducted through the bulk of the protein-coated DNA, analogous to the
pili in this research, rather than around the outside [63]. This suggests that pili would still
be imageable using STM at 77 K, and the experiments described here would be feasible.
52

Vbias
Itunneling

= Rtunneling = Rpilus + Rgap

(4.3)

Itunneling A
Rgap

Rtunneling

Rpilus
Vbias

Figure 4.5: Diagram of an STM experiment on a pilus. The total tunneling resistance is the
sum of the resistances of the tunneling gap and of the pilus itself.
If the temperature-dependent conductivity mechanisms are incoherent, which is likely,
and thus distributed across all relevant energies set by the bias voltages in the dI/dV curves,
they will contribute equally to the conductivity. This will serve to offset the dI/dV measurements uniformly up the ordinate axis at room temperature as compared to cryogenic
spectroscopy, rather than to change the shape of the dI/dV curves. Since the scaling along
the ordinate axis is arbitrary, with the exception of zero dI/dV , this will be essentially undetectable using this technique unless it serves to raise a cryogenic value of zero dI/dV to a
value of greater than zero dI/dV . For this reason, more detailed theoretical investigations
are valuable to aid in interpretation of the data.
It has long been recognized that energy levels in proteins can be treated analogously to
those in more traditional condensed matter sytems [64]. At a basic level, the pilus nanowires
can be treated as unified objects. Further insight into the underlying scientific processes
53

driving the results requires a more finely-grained treatment of pilus structure. In order to
interpret the results on the sub-pilus scale, theoretical assistance was obtained from Professor
Gustavo Feliciano of S˜o Paulo State University. Dr. Feliciano performed density functional
a
theory calculations to determine the electronic structure of a pilin subunit in order to provide
insight into density of state characteristics of the pili.

4.3

Molecular Dynamics Calculations

Because of the prohibitively large number of atoms involved in the structure of Geobacter pilins (over 940 [8, 65] for the pilin alone, not even including surrounding water), calculations were made by Dr. Feliciano on a single pilin subunit rather than on an entire pilus.
A sense of the complexity of the calculations can be gained from observing Figure 4.6. This
is a snapshot of Dr. Feliciano’s simulation of a Geobacter pilin at 77 K. The initial pilin
structure coordinates were obtained from one of the experimentally-determined structures
[52, 53], and it was allowed to evolve through time with periodic boundary conditions. The
water molecules are deliberately faded in color in Figure 4.6 to gain visibility for the pilin,
which is surrounded by water in three dimensions.
In order to calculate the density of states of the pilin under as near to the experimental
conditions as feasible, Dr. Feliciano first determined the most likely conformation of the
pilin, surrounded by simulated water with a small number of ions to neutralize the pilus
[8]. This was done by means of molecular dynamics calculations using the GROMACS code
[67]. At this point in the calculations, the temperature dependence was included. The
effect of temperature on the calculated properties of the pilin was in the pilin and water
conformations at the level of the molecular dynamics simulations; temperature dependence

54

water
pilin

Figure 4.6: Snapshot from a molecular dynamics calculation by Prof. Dr. G. Feliciano of
a Geobacter pilin subunit, beginning from an NMR-determined structure [53], immersed in
water at 77 K. Displayed using the program VMD [66].
did not enter into the subsequent density functional theory calculations. At 300 K, the pilin
structure fluctuated [8]. As expected due to the low temperature, however, the pilin did not
significantly fluctuate in conformation at 77 K. Because the conformation at 77 K was similar
to the multitude of conformations seen at 300 K, and because the temperature dependence
of the density of states calculations depends solely on the structure as determined by the
molecular dynamics simulations, this means the calculated density of states using the 300 K
pilin conformation can be used also to describe the low-temperature results as well, as is done
here. This suggests that further physical input to the model besides the physical structure
of the pilin and its surrounding waters may need to be incorporated in order to explain the
differences between the 77 K and the 300 K dI/dV measurements shown in Chapter 5.
Once the physical conformation of the pilin and its surrounding water was simulated, the
electronic structure could be calculated using density functional theory.

55

4.4

Density Functional Theory

Density functional theory (DFT) is one of the most popular quantum mechanical calculation
frameworks today, particularly popular for calculations of band structures and binding energies [68]. The name thoroughly describes the approach—everything about the ground state
of a system (energy, wavefunction, and backdrop potential) is a functional of the ground
state spatial number density. For simplicity, only the density of electrons will be discussed
here, although the theory is more general. Occupied and unoccupied electronic states can
be determined, as can excited states of the system [68]. The “density” in “density functional
theory” refers to the number density of electrons in a given area. This reflects the overall
wavefunction of the system, including the underlying potential describing the non-electron
parts of the system which provide the context or backdrop for the electrons’ experience with
one another [68]. This particle number density is the “functional” in the theory’s name.
The functionals themselves are properties such as the ground-state energy of the system,
determined by the function chosen to describe the density of electrons in the system. For
clarity of terminology, a descriptive sketch showing the density functional describing the
number of electrons in a system is shown in Figure 4.7. Choosing the correct function to
describe the charge density of a system, represented in the figure by “n,” permits discernment
of functionals of “n” which are the properties of interest.
While DFT itself is an exact theory, tractable calculations and descriptions of all but
the very simplest systems require some type of approximation, in no small part because the
correlations between the electrons in the calculation are not easy to pin down to a definite
description. One of the most convenient formulations of these is the generalized gradient
approximation. The “gradient” part of the name comes because this approximation at-

56

Case 2
n(x)

Case 1

here

# of electrons

# of electrons

n(x)

there

x

here

... etc.

x

there

there

N [n] =
functional

n(x) dx
here

function (could be n or n or ...)

Figure 4.7: The relationship between functions and functionals. Functions have different
outputs for different variable inputs describing the immediate circumstances where they
are applied. Functional s have different outputs for different functions used to describe the
paradigm of the underlying scenario.
tempts to deal with spatial variations in the electron density [68]. Density functional theory,
particularly using the generalized gradient approximation (GGA-DFT), is well-known to underestimate band gaps due to difficulties treating exchange correlations [8], so calculations
of the Geobacter pilin density of states would be expected to predict smaller gap widths
than observed in tunneling spectroscopy. However, since the measured density of states of
the pilus is a combination of the influences of many pilins within the vicinity of the tip,
the actual measured dI/dV may rightly have a smaller gap than that predicted by theoretical calculations. This may come about because the detected tunneling current which may
vanish due to a gap in the density of states of one pilin may be supplemented and hence
nonzero by “leakage” from nearby pilins. The density of states is not constant over the whole
pilin length [65], so the available energy levels from the “head” part of a nearby pilin may

57

be available as well as the energy levels from the “tail” part of the target pilin. For this
reason, consideration of the density of states of the entire assembled pilus, calculated as a
whole, is both very important and computationally demanding, thus must wait for further
computational advances.
The particular implementation of DFT used for the quantum mechanical calculations
on the Geobacter pilin [8, 54, 65] is the SIESTA code [69], which uses GGA-DFT. For
this application, the PBE version of GGA-DFT, which is more physically inspired and less
empirical than some others [70], is used. Since computation time using SIESTA scales linearly
with the number of particles considered, rather than to a power of that quantity, SIESTA is
well-suited to simulations containing a large number of particles, such as the Geobacter pilin
in water [54]. The outcome of these theoretical calculations by Prof. Dr. Feliciano of the
density of states of a Geobacter pilin is compared to experimental data in Chapter 5.

58

Chapter 5
Results and Interpretation
Geobacter sulfurreducens pili were investigated at room temperature and at 77 K using
scanning tunneling microscopy (STM) and point tunneling spectroscopy (“spectroscopy”).
Measurements were performed on multiple samples in order to ensure the reproducibility
of results, and the room temperature measurements served as controls for the 77 K experiments. This work expanded upon that of Veazey et al., Reference [23], which investigated
the topography and spectroscopy of Geobacter pili at room temperature using a control of
the non-conductive Pseudomonas aeruginosa strain K pili [22]. Results obtained at room
temperature and at 77 K are presented here and their salient features discussed, along with
a comparison to theoretical calculations and suggestions for further investigations.
In brief, G. sulfurreducens pili were found to remain conductive at 77 K via acquisition
of stable STM topographs. At 77 K, the pilus density of states displayed a distinct gap,
as observed by point tunneling spectroscopy. Nonzero density of states began at energies
easily reached by metabolically-relevant potentials, however. At room temperature, the
pseudogaps in the density of states were shallower and less abrupt than were the gaps in the
77 K dI/dV spectra. This indicates that additional electronic states are made available at
higher temperatures. Additionally, the density of states characteristics varied with position
across the pilus breadth.

59

5.1

Topography of G. sulfurreducens Pili

Building on previous investigations of Geobacter pilus topography by J. Veazey [22, 23], the
topographic characteristics of pili were examined at room temperature. The twin aims of this
portion of the experiment were to determine whether the use of buffer in the previous work’s
sample deposition suspension had any effect on the electronic properties or topographic characteristics of the pili and to determine whether pili were amenable to STM measurements at
77 K. While it was not anticipated that any residual buffer would introduce artifactual effects,
it was important to verify this assumption in order to compare experimental results using
these different sample preparations, as was done in Reference [25]. Topographic similarity
of the pili deposited from suspension in doubly deionized H2 O, as described in Appendix C,
to the results obtained by Veazey for pili deposited from suspension in PBS buffer [22] was
verified, and further topographic studies were performed. Pilus topography at 77 K was
found to be similar to room temperature topography. The topographic investigation was
extended to the ends of pili or pilus fragments using room temperature STM, as discussed
in Appendix D, and the long-term structural persistence of deposited pilus networks was
observed using AFM.

5.1.1

Topographic Characteristics of Pili

In STM topographs, the pili had an apparent height of approximately 1 to 5 nm and an
apparent cross-sectional diameter on the order of 10 to 20 nm at positive bias voltages,
consistent with the tip-broadening effects discussed in Reference [22]. Recall that, for STM
topography, the apparent topographic values for the image are a convolution of height and
electrical conductivity. A line section across a typical topograph is shown in Figure 5.1.

60

4.5
Height [nm]

8.5
nm

0

0

0

Width [nm]

100

Figure 5.1: Left: Room temperature STM topograph. Scale bar is 50 nm. Right: Line
section through the area of the scan marked with the black bar. (+ 0.5 V, 0.08 nA)
The pili do not appear perfectly cylindrical, as can be seen in the topograph and crosssection in Figure 5.1. Leaving aside the convolution between sample and tip shapes which
makes the pili appear so much wider than they are tall, the slanted character of the crosssections displays pilus substructure. At certain values of positive bias voltages (see Reference [22] for voltage-dependent imaging of Geobacter pili), including the + 0.5 V which was
the typical scanning bias in this work, and when the tip sharpness and sample–substrate adhesion were sufficient to obtain sub-pilus resolution, cross-sections of pili displayed a distinct
“ridge–scallop” structure. This consisted of a sharper, cliff-like ridge on one side of the pilus
breadth and a less-distinct other side which trailed off to meet the substrate at a smaller
angle. The trailing edge appeared scalloped because of a longitudinal periodicity.
Early STM scans of protein-coated DNA, which had dimensions similar to the pili studied
here, observed periodic structure [63]. Similarly, a periodic structure was resolved on the
pili, as demonstrated in Figure 5.2. This longitudinal pattern was comprised of periodic
nodules of increased brightness in the image, generally separated by ∼ 6 nm. The increased
brightness could be attributed to either increased height or increased local density of states,
which are convolved with one another due to the nature of STM topography.
61

g
rid

e

g
ed

e

Figure 5.2: Portion of the filament also seen in Figure 5.1. Scale bar is 20 nm. Left:
Highlighting the longitudinal pitch. Three of the nodule sets which comprise this pitch are
marked with arrows. Spacing between arrows is ∼ 6 nm. Right: Highlighting the transverse
ridge–scallop structure. The three nodules which comprise this periodic structure are marked
with arrows. Spacing between arrows is ∼ 5–8 nm. (+ 0.5 V, 0.08 nA)
In addition to the longitudinal periodicities, a transverse periodicity is also observable
across the breadth of the pilus. Either two or three distinct repetitions of the nodules can be
observed in many STM topographs of filaments. The brightest line of these nodules forms the
ridge structure, and the two less-distinct, possibly sometimes smeared into the appearance
of one, nodules form the scallop structure. In Figure 5.2, the spacing between the ridge and
the center nodule is approximately 6–8 nm, and the spacing between the middle nodule and
the scallop-side nodule is approximately 5 nm. Because the periodicity was more apparent
in the image at the top of the scan in Figure 5.2, that image was used for this determination.
The scalloping of one edge is due to the longitudinal periodicity of the transverse pattern.
These periodic structures were also observed in the work by Veazey [22], showing that any
residual buffer salts in the previous research did not attach to the pili in sufficient quantity to
obscure topographic variations in the pilus. Therefore, choice of water or PBS as the material
in which to suspend the pili is equivalent with respect to the small-scale topography of the

62

pili, so the earlier results using samples deposited from PBS are extendable to the case of
pili on bacteria in a subsurface environment rich with groundwater, applications in which
pili not associated with bacteria are in an aqueous environment, or other work done with
samples deposited from water [25].
Because STM is a convolution of geometric and local density of states information, the
nodule-like bright spots in the STM topographs could be due to structural periodicity in the
helical pitch of the pilins assembled to make the pilus, or they could be due to enhancement of
the local density of states caused by tyrosine amino acids in close proximity to one another
[52], as discussed further below. The longitudinal spacing between nodules is the same
order of magnitude as either of these effects [52], but it is not close enough to either to
make attribution definite, and measurements with higher spatial resolution are needed to
investigate this further.

5.1.2

Pilus Topography at 77 K

Well below the temperatures in the neighborhood of 300 K which are relevant to the bacteria’s lives, the relative importance of different mechanisms of pilus conductivity can alter
significantly. Temperature-dependent charge transport mechanisms such as those relying on
motion of parts of the protein will be abated when the pilus is at 77 K and its surrounding
water is frozen. If the pili were to become electrically insulating at 77 K, it would be difficult
to establish a stable tunneling current, and the pili would not be easily imaged by STM at
this temperature. Before spectroscopy was attempted on the pili at 77 K, then, it was important to establish that the pili were amenable to STM measurements at this temperature. As
seen in Figure 5.3, pili could be imaged by STM at 77 K. The bias voltage commonly used to
get reliably stable STM images with a tunneling current setpoint of approximately 80 pA at
63

room temperature, + 0.5 V, was also sufficient for stable imaging at 77 K, as in the example
in Figure 5.3. This is likely far greater than the minimum voltage required for the pilus to
conduct well enough to be imaged by STM; it was chosen as a compromise between a bias
voltage so low that the pili might not be sufficiently conductive and a bias voltage so high
that it might perturb the loosely-bound pili on the surface. The value of the pilus resistance
could have been determined by pressing the tip into the top of the pilus and thereby using
the STM tip as a contact in a two-terminal transport measurement, but the resulting damage
to the tip would make subsequent imaging and spectroscopic measurements infeasible.

Figure 5.3: Left: Portion of a pilus scanned at 77 K. Right: Portion of a different pilus
scanned at 300 K. A multiple tip is responsible for the apparent faint “shadow” to the upper
left of the actual pilus. Both scan sizes are 50 nm × 50 nm. (+ 0.5 V, 0.08 nA)

Another example of topography at 77 K is given in Figure 5.4. The speckles in the
image are indicative of locally unstable imaging, but since a nearby graphite step edge also
showed speckles, this is likely due to an instability on the tip rather than on the sample. The
smearing-type alteration between scans in Figure 5.4 can be attributed to loose coupling of
biological material to the graphite substrate; the feature in the lower left of the scans appears
to have been displaced by the tip.
64

Down

Up

Down

filament
loose
material

not
drift

Figure 5.4: Tip-induced displacement of material. Left: Scan of the area with tip moving in
the downward direction. Center: Acquired near-simultaneously with the left image but with
tip moving upward. Right: Scan of area again with tip moving downward. The motion of
the loosely-adsorbed material during scans is not due to drift, as can be seen by observing
that the brightest portion of the scan stays in place. Scans performed at 77 K. Scale bars
50 nm. (+ 0.5 V, 0.5 nA)
Both at 77 K and at room temperature, acquisition of stable pilus topography was complicated by the lack of rigorous coupling between the pili and the graphite substrate in many
cases, as evidenced by the apparent appearing and disappearing of pili mid-scan and between
scans seen in Figures 5.5 and 5.6. When a feature which was previously imaged disappears
partway through the scan, it can either have physically moved during scanning, or it can
have lost its electrical connection to the bias voltage source.

Figure 5.5: Four scans of a filamental fragment at 77 K, done immediately successively in
the same position. Verification that the disappearance of segments of the object was not
due to drift can be observed via the conserved position of the patch in the upper right of
the image. Scale bars 200 nm. (+ 0.5 V, 0.1 nA)

65

Figure 5.6: Four scans of a filamental fragment at 300 K, done immediately successively in
the same position. Portions of the pilus move back and forth between scans. All scans shown
were in the downward direction. Scale bars 50 nm. (+ 0.5 V, 0.08 A)
A layer of water likely existed between the pili and the graphite. The ambient air to
which the graphite was exposed between cleaving to expose a fresh surface and depositing
the sample is known to create an adsorbed water layer on graphite [50]. More tellingly,
the sample was deposited from a suspension of pili in doubly deionized water. Since the
pili were suspended in water, it is entirely plausible that a thin layer of water could have
been trapped between the adsorbed pili and the graphite surface, which then froze when the
temperature was lowered. The water layer surrounding a protein can contribute significantly
to the charge transport characteristics of said protein [71]. Indeed, the water layer adsorbed
onto the sample surface in a moderately humid environment has enabled STM studies, using
specialized equipment, of biological materials on insulating substrates, where the bias was
delivered to the sample through the water layer [72]. Making measurements when the sample
had been reduced from atmospheric conditions to a vacuum of a few or a few tens of µtorr
is not sufficient to remove all the water, including the last adsorbed water monolayer, from
the sample [50]. Consequently, the pili may be considered to be in a hydrated state for the
measurements at room temperature. At 77 K, all water which may remain on the sample
will exist in a frozen state. Since pili were not necessarily well-coupled to the graphite

66

substrate at room temperature as well, as shown in Figure 5.6, ice formation alone is not
responsible for sample–substrate connectivity issues. Water is not the only substance which
can cause lessened electrical contact between the sample and the graphite substrate; some
studies suggest hydrocarbons present in the air can cause changes to the work function of the
graphite surface over time, even in vacuum [73]. Several studies have shown “patchy” areas in
work function mapping [73] and mechanical properties [74] of graphite surfaces investigated
either in high vacuum or in ambient conditions, suggesting that areas of the graphite surface
may have different surface properties which could affect where the pili tend to adsorb well.
Since the pili investigated here were deposited from a suspension in doubly deionized water and the pili previously investigated at room temperature were deposted from a suspension in phosphate buffered saline [22] (PBS, or sometimes colloquially, if redundantly,
“PBS buffer”), additional measurements were performed at room temperature to provide a
control better reflecting the conditions under which the cryogenic results were obtained,
rather than simply using the previous results as the control. Conventional imaging of
biomolecules, e.g., DNA by AFM, often takes place in a buffer solution to neutralize the
charged biomolecules so that they will more easily lie on a charge-neutral surface [75].
Many of the STM topographs and endpoints of spectroscopy voltage sweeps in the previous work on Geobacter pili deposited from PBS [22] were obtained at 1 V bias. It was
empirically determined during this work that such a voltage tended to induce more instability
in the imaging and spectroscopy. In order to obtain more stable, hence more reproducible,
results which applied less electrostatic force to the sample, most topographs and spectroscopy
voltage sweeps here were obtained at a bias magnitude of 0.5 V. The general decrease in
imaging stability at 1 V compared to the previous work [22] could speculatively be attributed
to pilus–substrate contact differences due to the suspension fluids used in the experiments.
67

5.1.3

Structural Longevity of Pilus Fibers

If pili are to be used in device applications separate from their originating bacteria, they must
maintain their characteristics over time. Proteins are conventionally preserved by chemically
fixing them with substances such as glutaraldehyde, which causes topographic changes in
the pili as observed by STM [22] and may also change the conductive properties of the pili.
In the interests of determining the longevity of the unfixed pilus nanowires studied here, a
sample was prepared of pili deposited on SiO2 with gold electrodes also present, as part of
the longitudinal conductivity study discussed in Reference [25]. The sample was then stored,
spending much of that time under a gentle overpressure of nitrogen gas. This sample was
scanned using the same AFM immediately after deposition and again 22.5 months later;
example scans from this comparison are shown in Figure 5.7.

4

um

10
nm

0

0

um

4 0

um

4

0

Figure 5.7: Left: AFM topograph of a freshly-deposited sample of pili on SiO2 . Right: Same
sample rescanned after 22.5 months, showing no degradation of the sample.

The pilus fibers remained structurally intact over this duration, as can be seen in Figure 5.7, where the distribution of pili follows a conventional pattern in both images. Addi68

tionally, the slightly bulbous appearance of the bundled pili, whether due to windings of pili
around one another or to trapped fluid between pili, was retained over this length of time, as
seen in the smaller scan-range topograph in Figure 5.8. Conductivity properties of the pili
were not investigated over time, nor was the lifetime of pili on conductive substrates such as
graphite probed over such a duration. Both would be useful further studies. The observation that the pili remained structurally intact, at least to 1 nm resolution, on an insulating
substrate over a minimum span of nearly two years is a promising sign for applications of
pili in devices with realistic lifespans.

350

nm

4
nm

0

0

nm

0
350

Figure 5.8: 350 nm × 350 nm AFM topograph of the 22.5 month old sample showing intact
individual pili and pilus bundles.

5.2

Room Temperature Spectroscopy

Point tunneling spectroscopy was performed at room temperature to act as a control for
two different experiments. First, and primarily, it was to act as a control for the cryogenic

69

spectroscopy that was the prime focus of this work. Second, it was to verify that no changes
were introduced to the electronic properties of the pili due to the purification procedure
and deposition in doubly deionized water rather than PBS buffer. While room temperature
spectroscopy had been obtained by Veazey [22, 23, 65], the sample purification procedure
and sample deposition method had evolved since those experiments. The room temperature
results of Veazey [22, 23, 65] were verified using both the cryogenic-capable STM and the
Cypher scanning probe microscope in STM mode, so there are no microscope-dependent
effects on the results. Because the tip was prone to acquiring contaminants between spectroscopy measurements, a control measurement was generally performed on graphite before
and after on-pilus measurements.
An example of room temperature dI/dV on a pilus is shown in Figure 5.9. Spectroscopy
was performed on the lower edge of the brightest filament shown in Figure 5.1 with graphite
control spectroscopy before and after. The process of obtaining the numerically-differentiated
dI/dV curves is discussed in Appendix D.
In the main, dI/dV curves obtained on the surfaces of pili at room temperature showed
pseudogaps in the vicinity of zero applied bias. Unlike a standard bandgap in the delocalized
charge carriers paradigm or HOMO–LUMO gap in the molecular orbitals paradigm, the
bottoms of these pseudogaps in the density of states did not entirely reach zero dI/dV .
For this reason, the pili could be described as pseudo-semiconductors at room temperature.
The other criterion for semiconducting behavior, that of being susceptible to doping, can be
achieved by creating mutant strains with different amino acids [25] or by changing the pH of
the surroundings [32]. The finite dI/dV at room temperature for all voltage values probed
highlights the conductivity of the pili at even very small applied voltages.

70

dI/dV [arb. units]

Graphite Control
Before

Graphite Control
After

gaplike

0

-0.4

Pilus

nonzero

0
0

+0.4 -0.4

0

0
+0.4
-0.4
Bias [V]

0

+0.4

Figure 5.9: dI/dV on a pilus at room temperature. Left: Graphite control. Center: Edge
of pilus. Right: Graphite control. Because of higher levels of scatter in the data, the
pilus dI/dV was smoothed again after differentiation using a 3-point binomial smoothing,
in addition to the procedure described in Appendix D.

5.2.1

Positional Dependence of Room Temperature Spectroscopy

As was previously observed by Veazey [22], the density of states characteristics of pili at
room temperature depended on where across the breadth of the pilus the spectroscopy point
was positioned. Briefly, the center of the pilus had a density of states which was more
graphite-like, while spectroscopy performed on either the ridge or the scallop edges of the
pilus showed more structure.
The varied characters of spectroscopic measurements taken over different parts of the
pilus could possibly be attributed to summation of the densities of states of multiple pilins
in the assembed pilus. Since pili are composed of a spiral pattern of pilins, neighboring pilins
will have different parts of their structure beneath the tip. As the calculated density of states
has a spatial dependence along the pilin [65], pilins can supplement their neighbors’ densities
of states to varying extents depending on where above the pilus the tip is positioned.
The zero-bias point on the dI/dV spectrum was not necessarily in the center of the
71

dI/dV [arb. units]

Graphite
before

Pilus

Pilus

Repeated
at one point

Pilus

Pilus
edge

Graphite
after

0
-0.4 0 +0.4
Bias [V]

Figure 5.10: dI/dV on a pilus at room temperature. Left: Graphite control. Center:
dI/dV obtained in several locations transversally across the breadth of the pilus. Right:
Graphite control. All abscissas are from - 0.4 V to + 0.4 V. Horizontal lines are dI/dV = 0.
pseudogap, as also observed in the cryogenic spectroscopy discussed below. This could
be interpreted as a diode-like asymmetry in the onset of conductivity; in the interests of
circumspection, this assertion is not made conclusively. An experimental source of similar
results could be “Fermi level pinning” due to non-pilus materials present on the sample. For
example, a lesser quality of electrical contact between the sample and the substrate could
act as though it were a dopant decreasing the net overall conductivity, or the substrate itself
could dope the nanowires [28]. This does not speak against the results presented here, since
results were reproduced on more than one pilus, and it is unlikely that identical contaminants
would be uniformly present.

5.2.2

Reproducibility of Room Temperature Spectroscopy

Similar spectroscopy was obtained on several pili at room temperature. Because of the gradualness of many of the pseudogap edges, pseudogap widths were difficult to assign definitively.
Further, the position-dependence of the spectroscopy across the breadth of the pilus introduced scatter in the comparison between samples. Despite these caveats, the pseudogaps at
room temperature were found to have an average width of 0.4 V, with a standard deviation of
72

0.1 V. This value was obtained by averaging pseudogap widths from seven objects on which
spectroscopy could be obtained. The difference between pseudogap widths on different sides
of the pili was not statistically significant, in large part due to the variation in pseudogap
width determinations. With these room temperature spectroscopy results in mind, results
of spectroscopy at 77 K can be put in a firmer context.

5.3

Spectroscopy at 77 K

The primary novel thrust of this research was investigation of the density of states of G. sulfurreducens pili as seen in dI/dV measurements from tunneling spectroscopy at 77 K. Similar
results were obtained on multiple samples, as in Figure 5.11, indicating that the results were
reproducible and not artifactual. The gross features of the spectroscopy were the same across
samples, although the matter is complicated based on the variability of dI/dV curves depending on the exact position along the pilus at which they were obtained, as observed at

0
-0.4

dI/dV [arb. units]

dI/dV [arb. units]

room temperature and even more pronouncedly at cryogenic temperature.

Gap Edge

0
-0.4

0
+0.4
Bias [V]

Gap Edge

0
+0.4
Bias [V]

Figure 5.11: Spectroscopy performed on two separate pili, using two different tips and two
separate sample depositions. The left edge of the gap is similar, and both spectra encounter
dI/dV = 0 inside the gap.

73

For dI/dV curves such as those in Figure 5.11, determination of gap widths was done
from numerically-differentiated I–V curves which had been smoothed by a 47-point Savitzky–
Golay smoothing. Since five voltage sweeps, yielding five I–V curves, were generally obtained
sequentially at a given point on the sample, the gap widths were determined independently
for each of the five curves. The onset of the gap was identified as the voltage at which the
dI/dV curve began to ascend steeply from its baseline value. The chief source of uncertainty
in gap width determinations at a given point on the sample was slight, non-systematic
differences between sequential curves acquired at the same point. For this reason, gap
widths are quoted only to one significant figure.

5.3.1

Salient Features of Cryogenic Spectroscopy

Certain features were reproducibly observed in 77 K spectroscopy of the pilus nanowires.
For simplicity, these features will be discussed with reference to one data point in particular,
and the dependence of spectroscopic features on position across the pilus will be discussed
as a subsequent topic.
Figure 5.12 presents dI/dV spectroscopy on a pilus at 77 K, bookended by control spectroscopy on the nearby graphite surface before and after the spectroscopy performed on
the pilus. The absolute values along the ordinate axis are not relevant; it is the relative
heights of peaks and valleys on the curve that matter. It is accepted practice in this type of
measurement to give the ordinate axis in arbitrary units.
As discussed in Chapter 2, zero is the only meaningful value along the ordinate axis,
as it means there are no electronic states available over that entire voltage range for which
dI/dV is zero. This, in turn, means that adding electrons by increasing the magnitude
of the voltage does not correspondingly alter the current response, which is a sign of a
74

semiconductor or insulator. The gap width of 0.2 eV, as shown in Figure 5.13, is an order
of magnitude smaller than that for proteins which aren’t used for charge transport, such as
Pseudomonas pili [22]. This means the Geobacter pilus has electronic states available for

0

-0.4

0
+0.4
Bias [V]

Pilus

0
-0.4

dI/dV [arb. units]

Graphite

dI/dV [arb. units]

dI/dV [arb. units]

very little energy investment.

Graphite

0

0
+0.4
Bias [V]

-0.4

0
+0.4
Bias [V]

dI/dV [arb. units]

Figure 5.12: Immediately-sequential point tunneling spectroscopy performed on a pilus and
nearby graphite. Left: Graphite spectroscopy before pilus spectroscopy. Center: Pilus
spectroscopy. Right: Graphite spectroscopy after pilus spectroscopy. Horizontal line is at
dI/dV = 0.

gap

dI/dV = 0

0
-0.4

slight offset

0
Bias [V]

+0.4

Figure 5.13: Detail of pilus spectroscopy from Figure 5.12. Gap width is 0.2 eV. Bottom of
gap is at dI/dV = 0. Center of gap is slightly more negative than 0 V.

75

5.3.2

Positional Dependence of Cryogenic Spectroscopy

As in the room temperature spectroscopy results, the shape of the dI/dV curves on pili at 77
K depended on where along the breadth of the pilus the spectroscopy point was positioned.
The spectroscopy in Figure 5.14 was obtained on the same pilus as that in Figure 5.12,
but its gap is significantly wider, 0.4 eV. The repeated before-and-after graphite control
spectroscopy shows that this is not due to tip contamination. The bottom of the gap is still

0

-0.4

0
+0.4
Bias [V]

Pilus

Graphite

0

0

-0.4

dI/dV [arb. units]

Graphite

dI/dV [arb. units]

dI/dV [arb. units]

at dI/dV = 0.

0
+0.4
Bias [V]

-0.4

0
+0.4
Bias [V]

Figure 5.14: Immediately-sequential point tunneling spectroscopy performed on a different
part of the pilus from Figure 5.12. Left: Graphite spectroscopy before pilus spectroscopy.
Center: Pilus spectroscopy. Right: Graphite spectroscopy after pilus spectroscopy.

For the particular pilus in Figures 5.12–5.14, additional dI/dV curves were obtained in
different locations on the pilus and are shown in Figure 5.15. All pilus spectroscopy in
Figure 5.15 was preceded and followed by graphite control spectroscopy. Since this not only
reproduces the room temperature finding, originally observed in Reference [22], that positiondependent spectroscopy is a characteristic of Geobacter pili but enhances the differences
considerably, it suggests that the position-dependent dI/dV arises because of the density of
states of the pili.
The peculiar shape of the dI/dV curve in the center of Figure 5.15 cries out for control
measurements. While the dI/dV above the middle of the pilus is unusually shaped, the
76

0
+0.4
Bias [V]

0

0

-0.4

Ridge
dI/dV [arb. units]

0

-0.4

Middle
dI/dV [arb. units]

dI/dV [arb. units]

Scallop

0
+0.4
Bias [V]

-0.4

0
+0.4
Bias [V]

Figure 5.15: On-pilus spectroscopy, labeled by the estimated location across the diameter
of the pilus at which the spectroscopy was taken. Left: dI/dV on the more diffuse side.
Center: dI/dV towards the middle. Right: dI/dV near the more separate area of the pilus.
Bottoms of gaps are at dI/dV = 0.
control measurements of graphite immediately before and after that data are acceptably
interpretable as graphite, as can be seen in Figure 5.16. Perhaps the shape is due to interplay
between density of states contributions of neighboring pilins. This was only clearly observed
on this particular pilus, so it remains to be convincingly reproduced and is presented here

0

-0.4

0
+0.4
Bias [V]

0
-0.4

Pilus

dI/dV [arb. units]

Graphite

dI/dV [arb. units]

dI/dV [arb. units]

as a matter of interest rather than the subject of substantive claims or interpretations.

Graphite

0

0
+0.4
Bias [V]

-0.4

0
+0.4
Bias [V]

Figure 5.16: Immediately-sequential point tunneling spectroscopy performed on the middle
part of the pilus from Figure 5.12. Left: Graphite spectroscopy before pilus spectroscopy.
Center: Pilus spectroscopy. Right: Graphite spectroscopy after pilus spectroscopy.

77

5.3.3

Reproducibility of Cryogenic Spectroscopy

As in the room temperature spectroscopy, the position-dependence of the spectroscopy at
77 K led to a variation in gap width for spectra obtained at different positions across the
breadth of the pilus, although gaps in the spectroscopy at 77 K were easier to determine due
to more abrupt gap edges. The gaps at 77 K were found to have an average width of 0.3 V,
with a standard deviation of 0.2 V. This value was obtained by averaging gap widths from
six objects on which spectroscopy could be obtained at 77 K. The larger standard deviation
in the 77 K spectroscopy compared to that at room temperature is in part due to the small
widths of gap-like features such as those in the spectroscopy on the middle of the pilus in
Figure 5.16. Gaps and pseudogaps at 77 K and at room temperature were consistent in
width with one another.

5.3.4

Comparison to Room Temperature Spectroscopy

Overall, the dI/dV spectroscopy obtained on Geobacter pilus nanowires at 77 K was comparable to that obtained on pili at 300 K, with two important differences, shown in Figure 5.17.
The cryogenic spectroscopy had more abrupt gap edges. This includes accounting for the
difference in spectroscopy at different locations transversally across the breadth of the pilus,
and even accounting for thermal broadening effects at room temperature, which almost—
but not quite—soften the gap edges enough to make up the difference and are discussed
further below. Secondly, the gaps in the dI/dV at 77 K genuinely reached zero density of
states, with gap widths on the order of 0.2–0.4 eV, which is narrow-gap semiconductor-like
behavior. Pseudogaps in the room temperature spectroscopy did not quite reach down to
zero differential conductance. Room temperature spectroscopy was in general slightly less

78

stable, perhaps because at 77 K, most contaminants were frozen into greater stability. Despite these differences, however, the cryogenic and room temperature dI/dV spectra were
recognizably similar in character to each other, displaying either semiconductor (77 K) or

dI/dV [arb. units]

reminiscent-of-semiconductor (300 K) dI/dV .

0
-0.4

77 K

300 K
pseudogap

gap

0
Bias [V]

dI/dV=0 0
+0.4
-0.4

nonzero
0
Bias [V]

+0.4

Figure 5.17: Comparison between cryogenic and room temperature dI/dV . Left: 77 K
spectroscopy. Right: 300 K spectroscopy. Dashed line is a guide to the eye indicating
pseudogap bottom.

5.3.5

Effect of Thermal Broadening

Thermal broadening of the density of states contributes to the measured dI/dV . At temperatures above 0 K, temperature-induced fluctuations in the occupation of the available states
will smear out the ability of a dI/dV measurement to discern the pure density of states. For
this reason, when comparing 77 K and 300 K results, it is useful to set forth the expected
contribution of purely thermal broadening, statistical effects so this can be qualitatively disentangled from effects on the spectroscopic results due to different temperature-dependent
processes. An example of a case in which such thermal broadening effects can mimic or
obscure the effect of more exotic physical processes on a system’s temperature dependence
is discussed in Reference [76].
Previous room-temperature spectroscopy on G. sulfurreducens pili [22] showed pseudo-

79

gaps in the density of states where the conductivity was consistently minimal, slightly above
zero, for small magnitudes of the applied voltage. From this, it could be predicted that such
gap-like features would widen and deepen as the temperature was decreased. As an example
of this, assume an idealized semiconductor which has perfectly straight bandgap edges when
the density of states is measured, as shown in Figure 5.18. Raising the temperature to 300
K and considering only thermal effects would amount to convolving the function which describes the shape of the density of states with the derivative of the Fermi function, because
the charge carriers in this system obey Fermi-Dirac statistics, at each voltage value. This
is done in Equation 5.1, where the specific voltage value for which the thermal broadening
is evaluated is represented by V0 in units of mV. Additionally, all energy units, conventionally in electron volts, have been converted to mV to match experimental quantities, and
values of V are expressed in units of mV to correspond to the voltage range relevant to this
experiment, recalling that at room temperature, kB T = 25 meV.
+∞

dI/dV 300K (V0 ) =

−∞
+∞

DOS(V ) ∗ f (V − V0 ) dV
DOS(V ) ∗

=
−∞

1
exp[(V − V0 )/25]
dV
25 {exp[(V − V0 )/25] + 1}2

(5.1)

As can be seen in Figure 5.18, the statistical effect due to an increase in temperature
to room temperature is insufficient to erase the bandgap of this hypothetical material; it
only softens the edges somewhat. Thus cooling the samples, especially only down to 77 K,
will not open up tenths-of-eV-scale gaps in the density of states of the pili by decreasing
thermal broadening. Since the effect of thermal broadening is noticeable in the idealized
semiconductor in Figure 5.18, it is valuable to repeat this calculation to see the effect of
thermal broadening on actual dI/dV data. The results of this are shown in Figure 5.19. For
80

these calculations, the dI/dV at 77 K is assumed to be equal to the density of states; since
there is some minor thermal broadening already present in the 77 K data, the actual thermal
broadening will not be quite as pronounced as that seen in Figure 5.19, which nevertheless

DOS

dI/dV [arb. units]

does not remove the gap structure sufficiently to match the measured dI/dV at 300 K.

-400 -200

0
200
bias [mV]

400

-400 -200

0
200
bias [mV]

400

dI/dV [arb. units]

dI/dV [arb. units]

Figure 5.18: Effect of thermal broadening, given by Equation 5.1, on a hypothetical semiconductor with sharp bandgap edges. Left: Density of states. Right: Thermally-broadened
DOS which would be measured as dI/dV of this semiconductor at 300 K, overlaid on the
DOS. dI/dV is plotted as a dashed line as a guide to the eye; actual point spacing for both
plots’ computations is 50 mV.

-400 -200
0
200
bias [mV]

400

-400 -200
0
200
bias [mV]

400

Figure 5.19: Effect of thermal broadening, given by Equation 5.1, on actual pilus data
obtained at 77 K. Left: dI/dV at 77 K. For the purposes of this calculation, this is assumed
to be equal to the density of states, thus the effect of thermal broadening will be slightly
smaller than shown here. Horizontal line is dI/dV = 0. Right: Data thermally broadened by
300 K, overlain in red. Feature edges and small oscillations in the dI/dV appear smoothed.

81

5.4

Results in Context

Using the room temperature spectroscopy measurements as a control, STM and point tunneling spectroscopy on G. sulfurreducens pilus nanowires at 77 K showed that the pili are
transversally conductive and have a semiconductor-like density of states when probed at
cryogenic temperatures. The edges of this gap, energies at which electronic states become
available, are at some few tens or hundreds of mV, which is on the order of biologicallyrelevant potentials involved in Geobacter metabolism [77]. It can be instructive to put these
findings in a larger context, particularly by comparison to theoretical insights.

5.4.1

Comparison to Theoretical Calculations

Theoretical investigations by means of ab initio calculations can provide a physical context
for interpretation of the observed dI/dV of the pili. As discussed in Chapter 4, the temperature dependence of the calculated density of states enters the calculations at the level of
the molecular dynamics simulation, where the physical shape of the pilin is determined at a
given temperature, surrounded by water, since the pili remained hydrated even after the experimental drying process [50]. The density of states itself is temperature-independent; how
many and which of these states are actually filled at a given applied voltage—the independent variable in dI/dV measurements—depends in part on temperature-dependent processes
in the system.

Calculation Parameters
It has long been known for semiconducting proteins that adsorbed water on the protein
greatly affects its conductivity [78]. Proteins deposited from water and wicked dry, even

82

when placed in a high vacuum microscope, will be surrounded by at least a small amount
of water [50]. For this reason, considering some water molecules in conjunction with the
protein itself is important; also, pili in vivo are surrounded by water, and water affects the
conductivity properties of proteins, so it is more realistic to consider pili surrounded by
water. Furthermore, the relationship of the water and dissolved ions to the pilin subunit
accounts for much of the propensities of its calculated electronic structure.
Beginning from the homology model [8] for the approximate configuration of the atoms
in the Geobacter pilin molecule, Prof. Dr. Gustavo Feliciano performed classical molecular
dynamics simulations at 300 K to determine the configurations the pilin and its surrounding
waters would take [8, 65]. These atomic positions at different “snapshots” in time were then
input into the ab initio quantum mechanical SIESTA [54] code to perform the density of
states calculations [8, 65]. The density of states calculations simulated the electronic structure of the pilin in the context of its surroundings, based on atomic charges and atomic
positions determined by the molecular dynamics simulations. Because the pilin at 300 K
sampled such a large variety of conformations in the course of the molecular dynamics simulation [8] and the pilin at 77 K was found to remain essentially in its initial configuration, it
could be assumed that the 300 K calculations could also describe the 77 K density of states.

Calculated Density of States
Simulations of Geobacter pilins at room temperature show a gap in the density of states in
which the differential conductivity goes completely to zero. This gap is much narrower than
the few-eV gap seen in most proteins [8, 65], as expected, since Geobacter pilins are part of
electrically-conductive pili, by contrast to most proteins, which are electrically insulating.
However, this calculated gap is wider by a factor of two to four than that observed in
83

experimental spectroscopy. This is interesting in light of the expectation that GGA-DFT
would be expected to underpredict the gap width [68], although published calculations on a
room temperature pilin also overpredict the gap [8]. For a broader perspective, the calculated
density of states is shown in Figure 5.20 over twice the voltage range used to display most of
the experimental data. This shows the shape of the non-gap density of states; restriction to
the experimentally-probed voltage range would show very little beyond the gap. In general,
the character of the calculated density of states is closely akin to that of the experimentallydetermined dI/dV , although the calculated gap is wider. A slight offset of the center of
the gap from 0 V can be observed in some of the calculated curves, as can be compared
to the experimental data in Figure 5.13, indicating that Fermi level pinning is not the only
mechanism by which experimental curves could be offset in energy. Specific peaks in the
theoretical curves are not easily correlated with peaks in experimental curves.
The density of states presented in Figure 5.21 is formed from an average of five normalized
density of states calculation results. The five component curves, among those shown over a
larger voltage range in Figure 5.20, are density of states calculations using density functional
theory, obtained for different conformations of the pilin, ions, and water at statistically
independent snapshots of the molecular dynamics calculation; density of states calculations
were performed by Dr. Feliciano. Both experiment and theory show a gap which reaches to
zero on the ordinate axis; the calculated gap is wider.

Discussion of Theoretical Calculations
Discrepancies between measured and calculated densities of states in this system could be
attributed to several possible origins. The calculated density of states for each individual simulated spectrum in Figure 5.20 is based on the physical structure of the pilin and
84

DOS
0
+0.8
Bias [V]

DOS

0
-0.8

0
-0.8

0
+0.8
Bias [V]

Figure 5.20: Calculated density of states for a Geobacter pilin [65]. Eight snapshots of pilin
and solvent conformations, shown individually to display the variation in density of states
among configurations. All calculations plotted here are displayed over twice the voltage scale
of the experimental data to show more of the structure of the density of states. Molecular
dynamics and density functional theory calculations were performed by G. Feliciano.
water at one snapshot in the molecular dynamics simulation and does not include the motion of the protein. Proteins reconfigure, at least slightly, in response to their electrostatic
environments—this contributes to the reorganization energy λ in Marcus theory, as discussed in Chapter 4. Considerations of small motions of components of the pilus due to
the electronic interaction with the STM tip are neglected in the experimental interpretation.
Similarly, the theoretical calculations consider the pilin density of states to be determined
by its structure and interactions with the water but do not include interactions with the tip,
which can be important but difficult to quantify [79].
Perhaps more significantly, the calculations are only computationally tractable at the
pilin level at the moment, and the measurements were done on fully-assembled pili. Since
the majority of the pilin structure is buried within the assembled pilus, this can not only

85

DOS

Calculated

0
-1.5

gap width ~ 0.9 V
0
+1.5
Bias [V]

dI/dV=0

Figure 5.21: Calculated density of states for a Geobacter pilin [65]. Five snapshots of pilin
and solvent conformations are averaged together. Molecular dynamics and density functional
theory calculations were performed by G. Feliciano.
have implications for the conformation of the pilins [52] but also for the electrostatic and
hydration environment of the pilins. The model shown in Figure 4.6 is evaluated with
periodic boundary conditions, thus essentially describing a chain of pilins surrounded by
boxes of water rather than assembled with one another. Since the interior of the pilus is
hydrophobic, the α-helices are less exposed to water in the assembled pilus than they are
in the calculated scenario as shown in Figure 4.6. As postulated in Chapter 4, “leakage” of
available electronic states from pilin to pilin within the assembled pilus could act to decrease
the gap width as the electronic structures of adjoining pilins overlap. Further calculations
in progress from Dr. Feliciano begin from the very recently-determined NMR pilin structure
[52] rather than the homology model [53], and efforts are currently underway to simulate a
portion of an assembled pilus.
The difference between calculated and experimental results is not significantly impacted
by thermal broadening to 77 K, at which the experimental data were obtained. This can
be extrapolated by reference to Figure 5.19, which shows a more extreme case of thermal
86

broadening than is present in this particular example. This indicates that further detail in the
model is needed to account for differences between the 300 K and the 77 K experimentallyobserved dI/dV . Freezing of the pilus and water as they are cooled to 77 K decreases the
ability of the pilus to move. Protein motion and reorganization of surroundings are important
factors in the way biological molecules react to charge transfer processes, as discussed in the
context of of Marcus theory in Chapter 4, and it can be speculated that the room temperature
gap is raised to a non-zero density of states in the experiment due to small motions of the pilus
or surrounding water below the spatial resolution obtained during this experiment. Since
parts of the pilin subunit were found to be particularly flexible, as displayed in Figure 4.3,
which the authors of that study suggest may impact the ability of the assembled pilus to
accommodate bending [52], these regions of the pilin could be promising locations to consider
further for this purpose.
In both theory and cryogenic experiment, the density of states is zero for a gap extent on
the order of a few hundred millivolts. The unexpectedly wide gap in theoretical calculations
could potentially be addressed by future calculations of an assembled pilus. The raising
of the room temperature dI/dV above zero in the pseudogap region will likely require the
incorporation of other physical processes into the model. In all, the calculations for the pilin
subunit are not identical to the experimentally-determined dI/dV curves, but they share
their most profound feature, a semiconductor-like gap in the density of states.

5.4.2

Biological Relevance

The pili have been previously proposed to behave as molecular rectifiers [22, 65]. In this
scenario, there is a preferential direction for current to travel along the pilus, which could
help prevent undesirable transport of electrons in the “wrong” direction into rather than
87

away from the bacterium. The offset of the gap center from 0 V indicated in Figure 5.13
demonstrates a smaller positive than negative voltage needed to reach the nonzero density of
states region beyond the gap. Thus an electric potential established in one direction across
the pilus would find the pilus more amenable to conductivity by providing more available
electronic states, while a potential established in the opposite direction would need to be
slightly greater in order to have electronic states available to use for conductivity. On the
opposite edge of the pilus as shown in Figure 5.14, an asymmetry is also present between
the positive and negative biases required to observe nonzero density of states. For that
edge of the pilus, however, the gap is reproducibly slightly wider. This is consistent with
the rectifying picture [22, 65] wherein current travels in a spiral direction along the pilus
length. To address questions of current pathways directly, transport measurements such
as those in Reference [25] are crucial. The pili were found to conduct ohmically in the
longitudinal direction at room temperature [25] within the voltage regime considered in this
work, suggesting that pseudogaps in the local density of states at room temperature do not
seriously diminish pilus conductivity.

5.4.3

Potential Further Investigations

While significant progress has been made over the past several years on the properties of the
pili in Geobacter sulfurreducens, there remain many more details to be dived into. In particular, the questions of what specific parts of the pili are most responsible for the conductivity
can be addressed by futher scanning probe microscopy studies. Directed mutagenesis to
produce pili which lack components which are suspected of being keys to charge transport
in Geobacter pili, combined with transport measurements [25], as well as other sub-pilus
studies can narrow down the most important contributors to conductivity in these pili. In
88

addition to suggestions of further research directions, potential improvements to the STM
and conductive AFM [25] experimental configurations are discussed here.

Role of Tyrosines
A recent study [52] overlaid the experimentally-determined Geobacter pilin structure [53]
onto a known pilus structure from another bacterium to model an assembled Geobacter pilus.
This pilus model demonstrates that tyrosine amino acids from multiple pilins, shown for an
individual pilin in Figure 4.3, are concentrated together periodically throughout the pilus,
as can be seen in Figure 5.22. Since tyrosines are traditional culprits for biological charge
transport [55], this strongly suggests that pili should have alternating spots of higher and
lower density of electronic states. The pilin calculations by Prof. Dr. Feliciano show variation
in the local density of states along the pilin [65]. This may be related to some of the
periodic bright spots in STM images of the pili, and comparison between this model and
data presented both here and in Reference [22] is presently in progress.

Higher-Resolution Imaging
While STM is capable of atomic resolution with relative ease, as shown in Figure 3.5, atomic
resolution of the surface of protein structures such as the pili investigated here remains
temporarily out of reach. In large part, this is due to the difficulties involved in sample
stabilization. The pili in these experiments did not always thoroughly stick to the surface,
sometimes losing contact with the substrate mid-scan, as in Figure 5.5. Adhering the pili
more strongly to the surface could help stabilize them for high-resolution imaging [27]. This
was not done in this work because a substrate for the pili with a well-known density of states
was desired, as well as the minimum possible chemical perturbation to the pili. Often in

89

tyrosine grouping

tyrosine grouping
pilin

tyrosine grouping

Figure 5.22: Periodic clusters of tyrosine amino acids in the assembled pilus model adapted
from Reference [52]. This research was originally published in The Journal of Biological
Chemistry. P.N. Reardon and K.T. Mueller. Structure of the Type IVa Major Pilin from
the Electrically Conductive Bacterial Nanowires of Geobacter sulfurreducens. 2013; in press.
Copyright the American Society for Biochemistry and Molecular Biology.
similar experiments, such substance which binds both to the protein and to the substrate
is incorporated into the protein under study [80], although physisorption as was done here
is also a common deposition strategy [75]. Now that initial work has been done on these
pili such that their density of states in different circumstances is broadly predictable, adding
other molecules to the sample to attach the pili to a substrate for higher-resolution imaging
can be ventured.
Higher spatial resolution for spectroscopy on pili which stay in place on the surface could
help disentangle relative contributions of separate pilins to the spatial dependence of the
pilus density of states. If G. sulfurreducens pili prove to be molecular rectifiers (diodes) as
proposed in References [22] and [65], it would be interesting to investigate whether these pili,
without their originating bacteria to control their motion, tend to counteralign when forming
bundles—in other words, whether the pili themselves have an electric dipole moment which
would cause them to associate opposite to one another, much as magnets counteralign their
90

poles when brought close together—or whether they tend to align such that a bacterium
with several pili intertwined in effect has a bundle of wires in parallel, hence decreasing the
effective resistance of the bundle.

Temperature Dependence
While the present study focused on creating a large contrast between the two temperatures
probed, 77 K and 300 K, in order to exacerbate temperature dependence for visibility of
effects, future work can focus on smaller-scale temperature changes. Of particular interest is
further observation of the freezing process and the effect of frozen versus liquid water on pilus
properties. Since Geobacter bacteria are often found in sediments containing groundwater
which could freeze, the behavior of pili under frozen-water conditions is environmentally relevant. It can be predicted that the density of states of the pili as observed in the 77 K frozen
water scenario and presented in this work will be observed to be the same for all temperatures at which water is frozen, after accounting for thermal broadening effects as discussed
above. If so, the metabolically-surmountable gaps observed in 77 K spectroscopy indicate
that the pili could continue to participate in charge transport in the bacterial environment
even when groundwater is frozen. This prediction can be tested by investigating the pili as
the sample is slowly cooled in the temperature range around the freezing point of water,
perhaps using, e.g., a Peltier thermoelectric cooling mechanism.

Electrode Fabrication
Further tuning can be performed on the pilus longitudinal conductivity study of Reference [25]. In the interests of enhancing pilus–electrode contact, flakes of multilayer graphene
could be used as an interface to the gold. In this vision, schematically sketched in Fig-

91

ure 5.23, flakes of graphene would be deposited onto the standard substrate of silicon with
a 300 nm thermally-grown oxide layer as is commonly done for studies of graphene itself
[81], comprised of different thicknesses and widths and scattered in no particular pattern
on the SiO2 substrate, then larger-scale gold electrodes would be sputtered through a shadowmask onto the substrate and graphene flakes. The sputtered electrodes would provide
electrical conductivity to some of the graphene flakes on the substrate. Sputtering through
a shadowmask, in which the material is laid down on the substrate everywhere except where
blocked by a physical mask, obviates the need for the photolithographic procedure described
in Appendix E. The conductive AFM measurement can then proceed as before [25].

SiO2

SiO2 + graphene SiO2 + graphene SiO2 + graphene
+ Au
+ Au + pili

Figure 5.23: Suggested new electrode scheme for pilus longitudinal conductivity experiment
[25]. The graphene flakes provide a more pilus-friendly interface to the gold electrodes.
Diagram is not to scale.

Several advantages to this design can be envisioned, including the shorter height of fewnanometer graphene compared to 25 nm electrodes, the opportunity for pili to be wedged
under the graphene flakes as a secondary method of achieving electrical contact. The chief
advantage comes from the history of study of G. sulfurreducens pili on graphite surfaces,
e.g., [3, 22], and thus the better-established characterization of the pilus–electrode interface,
as furthered by the research discussed here.

92

APPENDICES

93

Appendix A
Operation of an Atomic Force
Microscopy User Facility
The MSU Center of Research Excellence in Complex Materials (CORE–CM) purchased a
Cypher atomic force microscope/scanning probe microscope from Asylum Research, which
was installed in September 2011. This Cypher AFM/SPM was to be operated as a user
facility under the general supervision of Reza Loloee, but the opportunity existed for another
person to take a strong role in the day-to-day operation of the facility, training users and
providing first-line technical support. This Appendix documents some of the components of
this work, as well as presenting a technique developed to extend the range of measurements
for which the Cypher could be used.

A.1

Installation and Characterization

As a user facility instrument, this Cypher needed to be situated in a freely-accessible space. It
was housed in an available small room until noise characterization measurements established
that a more traditional lab space was necessary, at which point it was moved to its longterm home in B138 Biomedical and Physical Sciences Building. Since noise levels were the
primary reason for moving the instrument, characterization of the noise was important to
quantify the benefits of each location.
94

Since the scanning probe microscopy techniques pertinent to the research presented here
are extraordinarily sensitive to the separation between the tip and the sample, it is extremely
important to minimize vibrations of the system that could affect the tip–sample separation
distance. The natural frequency of building structures, even in basements, is on the order of
a few tens of Hz [30]. The scanning probe microscope should be vibration-isolated in such a
way that this range of frequencies is damped out.
The sensitivity of the scanning probe measurement to ambient vibration depends on the
measurement technique being used. The simultaneous operation of other equipment in the
same lab as the Cypher was observed to disrupt STM measurements but did not significantly
impact amplitude modulation AFM scans. For this reason, noise characterization tests are
done with the AFM tip in mechanical contact with the surface. As an example, a portion of
a noise characterization test is shown in Figure A.1. The Asylum Research control software
for the Cypher monitors the motion of the cantilever due solely to noise while the tip is on
the surface, and it plots the amplitude and the frequency distribution to aid in diagnosing
noise sources. Noise in the frequency range of a few tens of hertz can generally be attributed
to mechanical vibration [30], while electronic noise displays itself in a narrow band at 60 Hz
and its higher harmonics, since this is the frequency of mains current in the United States.

A.2

Technique Development

The range of motion of the z piezo on the Cypher scanner places a limit on the heights
of features that can be measured by this microscope. For the scanner on this particular
Cypher, the z piezo range was measured to be 6.2 µm. A significant proportion of potential
facility users, however, desire to measure heights ranging from several to several tens of µm.

95

Frequency [Hz] Amplitude [pm]

40
30
20

100
10
1
1640

1642
Time of Day

1644

Figure A.1: Few-minute subset of a noise test. Top: Amplitude of the noise in picometers.
Noise amplitude decreased when nearby equipment was turned off. Bottom: Frequency
distribution. Mechanical vibration is evident in the broad band of noise around 20 Hz.
A workaround method to determine large heights was devised in collaboration with Scott
MacLaren of the University of Illinois at Urbana–Champaign and members of the Asylum
Research scientific staff, as worked out on the Asylum Research Forum [82].
The method, dubbed the Giant Step Height tool, involves focusing the optical microscope
inside the Cypher on the top and underneath the sample. This is done most simply at an edge
of a sample or at a trench scratched into a sample so that the sample mounting disk surface or
the substrate underlying the sample can be focused on as the bottom surface. For transparent
samples, the underlying substrate seen through the sample can be used for this measurement,
although the index of refraction of the sample must then be considered. The optical encoder
counts corresponding to the position of the focal point of the Cypher’s internal optical
microscope can be read out from the control software. These encoder counts are accurate to
200 nm [82], although scatter in visual determination of the in-focus locations of the sample
top and bottom was empirically found to be the limiting source of uncertainty. Consequently,

96

the µm accuracy of the display on the Engage Panel within the software is sufficent for this
method. For samples shorter than several µm in height but still uncomfortably tall for AFM,
greater precision can be achieved by mounting a tip in the Cypher tip holder, as sketched in
Figure A.2.

Tip-based GSH

Tipless GSH

Objective lens of
optical microscope

Sample
Mounting disk

Focus position

Focus position

Objective lens of
optical microscope
Tip holder
motor
Tip

Sample
Mounting disk

Z correction

z piezo
(fixed extension)
Scanner base

Fixed
distance

z piezo
Scanner base

Figure A.2: Left: Diagram of optical-only GSH. The objective lens of the optical microscope
is focused on the top and bottom of the sample, and the motor position of the objective
is recorded at the point at which each comes into focus. Right: GSH with a tip in place.
The focus of the objective lens of the optical microscope remains set on the tip, while the
tip + objective is moved as a unit until the tip comes in contact with the surface of the
sample. The motor position of the tip + objective is recorded. Because the tip–sample
contact involves extension or retraction of the z piezo, the z piezo extension or retraction
can be an important correction factor on the order of a micron, but involving the tip can
give more precise measurements of few-micron heights.

Initially, a test of the procedure was done by measuring the height of a glass slide,
correctly determining its height of 1.2 mm. Portions of screenshots used for the first GSH
measurement are shown in Figure A.3.
The primary test of this method on a scientifically-relevant sample was done with the
collaboration of Nathaniel Leonard from the MSU Department of Chemical Engineering and
97

Focus Position:
2.235 mm

Focus Position:
0.9808 mm

100 um

100 um

Figure A.3: Demonstration of the GSH method on a glass slide. Left: Focus on top of slide.
Right: Focus on sample disk underneath slide. Difference between focus positions is 1.2 mm.
Materials Science. Mr. Leonard studies catalysts comprised of a pyrolyzed combination of
iron, nitrogen compounds, and a porous carbon support, working towards a replacement for
platinum-based catalysts for oxygen-reduction applications [83, 84]. The sample inks were
created by dispersing catalyst in a Nafion + ethanol solution by sonication, using ethanol to
dilute the solution to have different densities of the catalyst + Nafion system in the ethanol.
After deposition of 5 µL of the solution onto a glassy carbon rotating ring-disk electrode
[83, 84], the ethanol was permitted to evaporate. Before measurement, a razor blade was
used to apply pressure laterally to the film, swiping it away from the glassy carbon electrode
surface and leaving a crisp edge to use for the height measurement.
The samples were prepared by Mr. Leonard, and the determination of “in focus” was made
initially by concurrence between the two experimenters to decrease the effect of personal
perception. The scatter in perceived heights was greater than the 200 nm resolution of the

98

detector encoding the position of the optical microscope focus, yet both experimenters agreed
on each focus position despite choosing slightly different surface features to focus on. Since
the roughness of the sample was often in the neighborhood of a few hundred nanometers,
comparable to the wavelength of the optical light being used to view the sample, this amount
of height variation in focus did not materially affect the results.
Three measurements were obtained at each location on the samples, with locations selected to represent the center and edges of the films. Since the film covered the entire
electrode except where it had been scraped away, the film edges were also at the electrode
edges. It was observed through the optical microscope that the film was cracked in a manner
reminiscent of parched ground, as seen in Figure A.4. This concurred with previous observations by the chemical engineering and materials science researchers, who had noted this as
a characteristic of the films. It is therefore unsurprising that measurements taken towards
an edge of the film and those obtained near the center could vary.
Table A.1 contains the data for GSH measurements of catalyst + Nafion film thicknesses.
Quoted uncertainty is the standard deviation, calculated in Excel and rounded to two significant digits. After inter-rater reliability was established for the first three measurements
based on independent consensus between the experimenters regarding the positions of the
top and undersurface of the films, it was sufficient to have one experimenter continue the
remainder of the measurements, beginning with the second 300 µg/cm2 sample loading in
Table A.1. The 200 µg/cm2 sample loading and the first 100 µg/cm2 sample loading in
Table A.1 were done on a different glassy carbon electrode of the same type as the electrode
used for all the other sample depositions.
Film thicknesses are shown in Figure A.5 as a function of sample loading. The alternate
electrode average film thickness for the 100 µg/cm2 loading was higher than, although still
99

gc
cat

Figure A.4: Focus position determination on catalyst + Nafion films loaded onto glassy
carbon rotating-ring electrodes. Left: Optical micrograph indicating the catalyst layer “cat”
and underlying glassy carbon “gc” electrode. Scale bar is 50 µm. Center: Optical micrograph
focusing on the top of the catalyst, digitally zoomed to twice the magnification of the left
and right images for ease of discerning features during focusing. Scale bar is 25 µm. Right:
Optical micrograph focusing on the underlying electrode. Scale bar is 50 µm. Brightness
differences among these images are due to the light levels used to aid in focusing.
consistent with, those measured on the usual electrode. The film thicknesses for the 300
µg/cm2 loading determined by one experimenter (lower value in the plot in Figure A.5) and
by both experimenters (upper value) using two different depositions of the sample on the
same electrode with the same loading were consistent within uncertainty, implying that there
was a sufficient amount of inter-rater reliability in the height determination perceptions. A
sample series of screenshots of the optical microscope images used for the GSH measurement
is presented in Figure A.4.
The tipless, or optical focus depth, GSH measurements were confirmed for some of the
samples by including the use of an AFM tip to determine the location of the surface, rather
than relying on optical focusing. These methods are compared in Figure A.2. For tipenhanced GSH, the tip was engaged on both the upper and the lower surfaces in turn, with
the optical microscope set to maintain its focus on and move with the tip, and the difference

100

y = 0.041x - 1.702

Figure A.5: Height measurements of catalyst + Nafion films loaded onto glassy carbon
electrodes using both optical focus depth GSH and tip-enhanced GSH. A linear fit was made
to the data acquired using Electrode 1, showing a linear dependence of film thickness on
sample loading. The alternate electrode, Electrode 2, appears to have a slightly greater slope,
but there were not enough sample loadings using that electrode to determine this definitively.
Within statistical uncertainty, the optical and tip-enhanced methods are equivalent.
in optical microscope position was measured via encoder counts [82]. In addition, during
this process, it was determined that the cantilever deflection contributed appreciably to the
measured heights when using this method for film thicknesses of only a few microns, since
the range of the z piezo was 6.2 µm. To calculate the correction to the height using the
tip-enhanced GSH method, use

z piezo f ull travel range
voltage range of z piezo
6.2 µm
= −(Ztop [V] − Zbot [V]) ∗
160 V

Z correction [µm] = −(Ztop − Zbot ) ∗

101

(A.1)

where Ztop and Zbot are the z piezo extension voltages on the sample top and the surface
underlying the sample, as observed during the measurements. The correction is negative
because a positive value of the z piezo extension causes a decrease in the relative objective
lens–sample separation. For shorter films, this more precise tip-enhanced GSH using the optical encoder count measurement was performed and compared to the optical-only method.
Results of the comparison are shown in Figure A.5, showing that tip-enhanced GSH results
fall within the uncertainty determined by multiple measurements using the optical measurement. The uncertainty in the z piezo extension determination was such that only two
significant figures, with some uncertainty, in the correction can be claimed, so no error bars
are shown for the tip-enhanced GSH data in Figure A.5.

A.3

User Training and Troubleshooting

In addition to experimental design and technique development for the aforementioned measurements and staff-assisted sample scanning, this role included responsibility for the first
line of troubleshooting on the apparatus and for interacting with Asylum Research technical
support as needed.

A.3.1

User Training

The Cypher was intended to be operated as a user facility. In order to insure safe and effective operation of the microscope, users were required to undergo training, which necessitated
the development and coordination of training sessions. Over the course of two years, several
users were trained to a level enabling them to operate the AFM independently, not only for
the standard amplitude modulation in air mode but also for more complicated modes of op-

102

eration. In addition, verification and validation of the independent trainings of further users
was performed. An efficient and sustainable method for training needed to be developed. To
this end, several instructional videos were created to aid in training future users. These were
produced both independently and in collaboration with Reza Loloee or Victor Ramirez.
Assisting with experimental design is an important component of proper user training.
Significant input into the experimental design was provided for experiments by users from
a variety of different departments and different colleges within the university. In addition,
staff-assisted scanning was performed on samples from a similar diversity of departments.
This is a facility-provided service whereby researchers who only need a few samples scanned,
or whose projects require advanced techniques, bring their samples to be scanned rather
than becoming trained as AFM users.

A.3.2

Troubleshooting and Maintenance

Part of facility operation consists of maintaining the equipment in proper working condition
and identifying and addressing issues. The Cypher controller developed a malfunction that
required its return for repair. The troubleshooting process surrounding this malfunction
involved disassembling and removing components of the electronics for testing in collaboration with Asylum Research technical support. This procedure provided the very interesting
opportunity to interact with the “inner workings” of the Cypher control electronics.
Conductive AFM I–V measurements used the dual-gain ORCA cantilever holder, which
includes one preamplifier with a sensitivity set to measure currents in the range of ± 10
nanoamperes and another preamplifier for currents in the range of ± 10 µA. It was observed
that when the 10 nA preamp was saturated with current, “ears” would appear in the current
reading of the 10 µA preamp, as seen in Figure A.6. This had also been independently
103

observed by another dual-gain ORCA user and by Asylum Research technical support [85]
and was identified as cross-talk between the preamplifiers.
20
10
0
10
0

Cur2 [nA]

60 Cur2 [nA]
0
10

Cur [nA]

-10
-0.2

-0.1

0
0
Bias [V]

+0.1

+0.2

-10

Cur [nA]

-0.4

-0.2

0
+0.2
Bias [V]

+0.4

Figure A.6: I–V curves demonstrating the “ears,” adapted from a post to the Asylum
Research Forum [85]. Left: ± 0.2 V bias sweep. Right: ± 0.5 V bias sweep. Both: The
I–V curve on top is as seen by the 10 µA preamp, and the I–V curve on bottom is as
seen by the 10 nA preamp, separated by a dashed line for clarity. Horizontal lines in the
lower I–V curves occur when the 10 nA preamp is saturated. The onset of this saturation
corresponds to the “ears” in the upper I–V curves.

104

Loading [µg/cm2 ]

Location on Sample

Layer Height [mm]

± [mm]

500

1

20

0.47

2

24

1.6

3

16

0.94

1

17

1.2

2

14

2.4

3

12

0.0

1

16

2.9

2

7

0.82

1

12

2.5

2

3

0.47

200

1

21

0.47

alternate electrode

2

4

2.5

3

19

0.94

100

1

0

1.4

alternate electrode

2

5

1.4

3

13

1.6

1

5

0.47

2

3

1.9

3

3

1.2

400

300

300

100

Table A.1: GSH optical focus depth measurements of catalyst + Nafion film thicknesses,
using the change in position of the optical microscope focus to determine the film heights.

105

Appendix B
Artifacts That Almost Thwarted Me
Since visual searches of data begin with a query in mind—e.g., “Is there a pilus in this
scan?”—before searching the image to answer the question [86], looking for pili primes the
researcher to focus on pilus-like features in scans, which may be artifactual. For this reason,
this Appendix presents several examples of artifacts, that they may be more easily recognized
as such in future work. Just as the examples of pilus topography shown in Chapter 5 provide
priming for future searches for pili on the surface using STM, these examples of artifacts can
provide references to assist in pattern recognition when checking for artifacts in future work
[87]. Additionally, presenting artifacts and actual pilus topography in the same document can
build connections to bring to mind both genuine pili and common artifacts when observing
nanowire-like filamental features, thus providing an aid to feature identification by priming
the future worker to recognize both pili and common artifacts [88, 89].
While distinguishment of artifacts from genuine pili can be done to some extent based
on feature topography, the proof is in the spectroscopic pudding, so to speak. This is
particularly important when searching for nanowire-like objects on a graphite substrate,
as graphite is known to produce nanowire-like artifacts [89]. No matter how pilus-like a
feature may look, its density of states will betray its identity. Relying on the spectroscopy
for object identification comes with its own requirements for vigilance; contamination or
unstable behavior of the tip or sample result in uncharacteristic spectroscopy. For this reason,
reproducible results are the most important safeguard against being tricked by artifacts.
106

Although graphite is a popular substrate for STM studies of biological materials (the
first STM studies on DNA [90] used a carbon film substrate instead of the highly-oriented
pyrolytic graphite, HOPG, more commonly in use today), it has a tendency to produce
multifarious artifactual formations which can be highly reminiscent of biological features
[89]. Since STM-on-graphite experiments have been performed for roughly three decades,
the database of such formations is well-developed, even if some of the artifacts are still not
fully physically explained [46].

B.1

Step Edges

The accumulation of graphitic debris along graphite step edges can lead to artifacts which
appear at first glance very similar to pili. In particular, they tend to be a few nanometers
in width, comparable to expected pilus widths. In addition, the instability of this debris
leads to spikes in the tunneling current as the feature is scanned over, thereby masking the
true height of the feature. When the step edges are distinct and have sharp edges, as in the
two examples in Figure B.1, these are easier to identify quickly as artifacts. For this reason,
careful nearly-real-time image processing can be a helpful practice, as short step edges may
be difficult to identify as such without subtracting a background tilt of the sample.
A particularly insidious example of artifactual behavior comes with step edges which
are not straight. One of the criteria that helps differentiate elongated carbon nanostructures
from biological nanowires is the tendency of carbon nanostructures to be straight over longer
distances, although this is not a hard and fast rule. When flakes of graphite laying atop
a more uniform graphite substrate have edges which are curved, particularly edges along
which debris has accumulated as described above or edges which show ridge behavior, as

107

Figure B.1: Typical artifacts at graphite step edges on two different parts of a sample. These
are not pili lying against the step edges but rather are graphite formations. Scale bars are
200 nm. Left scan (+ 0.5 V, 0.5 nA), right scan (+ 0.5 V, 0.4 nA)
discussed in Reference [46], the more subtle indication that the flake area is higher than its
surroundings may seem less convincing evidence than the curvedness of the edge. Further, the
aforementioned tunneling current spikes due to the debris can make it particularly difficult
to determine the relative heights of the flake and the underlying graphite, since spikes tend
to dominate the height scale in linecuts across the edges.
In addition, the density of states can be enhanced at the edges of some graphite steps due
to topological effects [91]. Since STM topographs are a convolution of surface height with
the local density of states of the sample, these can make the edges appear “taller” or brighter
in the STM image, solely due to electronic enhancement. This is a potential explanation for
the artifact on the right side of Figure B.1.
Alternatively, step edges can be stably imaged but possessed of deceptively periodic
lumpy protrusions as though the edges were crimped, as is seen in Figure B.2. Despite the
lumpiness, identification of this as a step edge is straightforward based on the surrounding
topography–the upper right half of the images in Figure B.2 is lower than the lower left half,
108

with the boundary marked by the lumpy feature. The periodicity of such step edge can even
mimic a chiral pattern [89], making it particularly important to detect when artifactual so
as not to confuse this with pilus structure.

Figure B.2: Three scans of a step edge, possessed of a deceptively lumpy structure along the
diagonal from upper left to lower right. The level of visible detail, obtained by decreasing
the scan size, increases from left to right. Scale bars 10 nm. (+ 0.5 V, 0.2 nA)

B.2

Vertebral Shape

A more insidious artifact seen in the course of this research at low temperature bore a strong
resemblance to a column of vertebrae and is shown in Figure B.3. A similar structure was
observed by Veazey [22] at room temperature on a similar sample as a zipper shape. The
aspect of this artifact that made it most perilously misleading was that it had a similar
breadth and only a slightly shorter height to that customarily seen for pili and multiple
periodic substructures, as were commonly observed in well-resolved pili. The non-pilus
identity of this object was eventually realized due to its spectroscopy often appearing similar
to spectroscopy on graphite and, more tellingly in this particular instance, based on the
topographic merging of the feature with a nearby graphite step edge, as seen in Figure B.3.
Because atomic resolution was not obtained under these specific conditions, observation of
109

the atomic lattice structure in the vicinity of this artifact was not possible, thus precise
identification of the origin of this artifact remains in the realm of speculation. This feature
may be a graphite grain boundary, where the density of states is electronically enhanced as
the direction of the graphite lattice changes its angle. Grain boundaries in graphite, where
two different orientations of the graphite lattice meet in a way analogous to magnetic domain
walls, produce periodic structures along the boundary as the mechanical deformation due to
stresses causes periodic defects which lead to an enhancement in the local electronic density
of states near the Fermi level, thus causing a periodic “ridge” in the image due to electronic
effects rather than actual physical height [92].

Figure B.3: Artifact comprised of periodic shapes reminiscent of vertebrae in a spinal column.
Left: Close-up of comma-shaped “vertebrae.” Scale bar is 10 nm. Center: Larger scan
demonstrating the abruptness of the dislocations. Scale bar is 50 nm. Right: Relation of
the vertebral shape to a nearby step edge. The vertebral shape apparently merges into the
step edge near the bottom of the scan. Scale bar is 100 nm. (+ 0.5 V, 0.08 nA)

B.3

Tendriloid Ripples of the Graphite Surface

Tendril-shaped ripples in the graphite surface can also mimic the shape and approximate
diameter of pili, although they are far shorter, often much less than a nanometer tall and

110

some fraction of the expected apparent height of the pili. They are not necessarily straight
over long distances, and this vermiform waviness of some of these artifacts can cause additional confusion with the flexible biological nanowires. This was the most common artifact
observed. The weak van der Waals attraction between the two-dimensional graphene layers
that are stacked to form a graphite crystal is not always strong enough to keep the graphene
layers mechanically stable with respect to one another. The perturbation due to the strong
electric field between the tip and the sample, particularly pronounced due to the small area
of the apex of the tip [30], can deform the graphite locally, applying shear forces in the
many-megapascal regime [46]. Indeed, the tip–sample interaction is strong enough that it
can be used deliberately to perturb the graphene layer atop the rest of the bulk graphite,
lifting it up or folding it over [93]. These ripples are particularly prevalent in areas where
the graphite has been disturbed, as is seen in the center of Figure B.4, where a perturbation
in the graphite surface has has a relatively large number of ripples in its vicinity.

Figure B.4: Left: Tendriloid graphite ripple of approximately the apparent diameter and one
half to one third the apparent height expected for a pilus on graphite. Center: These ripples
can be provoked by disruptions in the graphite surface. Right: Several tendriloid artifacts,
showing a variety of the shapes they can assume. Scale bars 100 nm. (+ 0.5 V, 0.08 nA)

111

B.4

Other Topographic Snares

In addition to the nanowire-like corrugations of the graphite surface mentioned above,
graphite can show “superlattice” behavior, in which the STM image shows a pattern of
bright and dark spots in the shape expected for the graphite lattice but with incorrect dimensions, due to issues during the growth or cleavage of the graphite sample [46]. It is useful
to be aware of this effect to avert worries about piezotube calibrations. It is also profitable
to expound further upon multiple-tip effects, as these are quite prevalent artifacts.

B.4.1

Graphite Superlattice

Graphite artifacts are not solely nanowire-like; sometimes, the distortions cause false latticelike patterns, such as the hexagonal pattern in Figure B.5. Because of the hexagonal pattern,
it is instinctive to assume initially that this is the atomic lattice of graphite; however, the
apparent atomic size and spacing is too large. One could speculate that the scanning piezotube had suddenly lost resolution and was only scanning over a far smaller than instructed
distance or that the tip had somehow become mechanically mired in place. The reproduciblity of this image, the sharpness of the step edge as seen in Figure B.5, and the fact that
subsequent scans in different areas displayed the expected nm/V resolution argues strongly
against this. According to the conventional explanation, that’s a Moir´ pattern caused by
e
interactions between the topmost and second of the overlapping sheets of graphene, which
stack atop each other to comprise the graphite crystal. When the topmost sheet is rotated
with respect to the sheet underneath it, the interactions cause the formation of a Moir´
e
interference pattern; also, strain in the graphite lattice can cause buckling that results in a
similar Moir´ pattern [46].
e

112

Figure B.5: Left: Hexagonal pattern on graphite due to a Moir´ pattern from interference
e
between overlapping graphene sheets. A line has been subtracted from each line of the image,
and the displayed height shading range has been truncated to emphasize the hexagonal
pattern. Scan size for left image is 244 nm × 298 nm. (+ 1 V, 0.08 nA) Center: Diagram of
a Moir´ interference pattern. Light and dark lines represent lines of atoms in two overlapping
e
layers of single-atom thickness. Circles at the intersections indicate the locations where the
enhanced electronic signal from the overlap would be detected as a pattern. Right: Diagram
of a Moir´ pattern for a hexagonal lattice such as graphite. Lattices have been rotated by 30◦
e
for visibility; actual rotation for the pattern observed here would have been much smaller.
In this experimental observation of a graphite superlattice, the bright-spot spacing was
approximately 100 times that of the actual atomic spacing of graphite, as shown by the comparison of image sizes showing similar periodicities in Figure B.6. Following Reference [46],
the rotation angle can be determined via

θrotation = 2 arcsin

1 atomic spacing
2 bright spot spacing

= 2 arcsin

1 0.246 nm
2 21.4 nm

= 0.6◦

(B.1)

showing that this observed pattern in Figures B.5 and B.6 would be due to a very small
rotation angle between overlapping sheets. This could either suggest that the top layer of
graphite was generally very well-matched to the underlying layers in terms of rotation or that
the Moir´ pattern was not caused by the relative rotation of layers of graphite but rather by
e
some other process such as lattice strain.

113

Figure B.6: Moir´ pattern and graphite lattice comparison. Left: Zoom in on hexagonal
e
pattern from Figure B.5. Image size 123 nm × 147 nm. (+ 1 V, 0.08 nA) Right: Atomic
resolution scan of the graphite lattice. Image size 2.1 nm × 2.6 nm. (+ 0.1 V, 1 nA) The
left image is nearly an order of magnitude larger than the right image along each direction,
yet the patterns appear nearly the same size, indicating the artifactuality of the left image.

B.4.2

Multiple-Tip Effects

As mentioned in Chapter 2, multiple-tip effects can cause the illusory appearance of more
features than are actually present. A frequently-seen imaging artifact in this research was a
repeated apparent image of pili when found on the surface, caused by more than one nearby
“minitip” at the end of the STM tip. In scenarios such as that in Figure B.7, a useful
metaphor may be trying to play a phonograph record using an upside-down porcupine as
the needle. Generally, due to the exponential drop-off in tunneling current with increased
tip–sample separation, there are only two or perhaps three competing minitips at the end of
a multiple tip. This was seen in studies using the cryogenic-capable STM and the Cypher
STM, as it is one of the most common STM artifacts. A particularly egregious example
of this echo-like repetition of features in a scan is given in Figure B.7. Careful observation
of the bends of the objects in this image show that there are multiple tips along different
114

lateral directions. Confirmation that this is not the actual surface topography comes from
observing the same echoing pattern on separate features in the image. Disentangling the
number of minitips involved in producing this image is a task for the sharp-eyed; an attempt
in one direction only is made in Figure B.7. The customary response to such a scan image
is to endeavor to recover the tip or to replace it entirely.

Figure B.7: A particularly dramatic example of a multiple tip. Left: Scan showing multiple
repetitions of features. From this scan, it appears there are at least three tips close in height
to each other and aligned approximately in the horizontal direction, slightly angled towards
the upper left of the image, and several additional tips approximately along the vertical
direction. Scan size 500 nm × 500 nm. Center: 150 nm × 150 nm scan near the middle of
the scan on the left. The approximately-horizontal multiple tips are even greater in number,
as another nearby tip is causing an overlapping echo on the leftmost of the three main feature
repetitions, as well as causing two fainter echoes to the left of the three. (+ 0.5 V, 0.2 nA)
Right: Artist’s conception of a tip which could produce the horizontal multiplicity shown
in these images; multiple tips along the other direction are not considered in this sketch.
Only the protrusions near the very end of the tip are relevant—in this sketch, there are six
such—due to the exponential dependence of tunneling current on tip–sample separation.

B.5

Artifacts in Spectroscopy

The health of an STM tip for spectroscopy can be evaluated by dI/dV measurements done
on the graphite substrate on which the sample was deposited. A clean graphite surface with a
clean, stable Pt:Ir tip should produce spectra with V-shaped dI/dV curves, like those shown
115

in Chapters 2 and 5. If the tip had acquired a contaminant while scanning the surface, the
spectroscopy could then take on the characteristics of that contaminant.
Alternatively, if the contaminant cluster of particles were small enough, perhaps a few
nanometers in diameter [94], that it acted as a small capacitor, a “Coulomb staircase”
[95] could be observed in the spectroscopy. This could potentially explain the example in
Figure B.8. While this spectrum is not a precise example of high-quality Coulomb staircases
as may be seen in, e.g., Reference [95], this could be because the formation of the Coulomb
staircase is an unfortunate side effect of a mesoscopic particle cluster on the tip rather than
being deliberately designed. Note that the particle cluster does not need to be an alien
substance; it can simply be a slightly-detached cluster of tip atoms with only a very small,
confined pathway leading to the rest of the body of the tip. The periodic peaks in the
dI/dV spectra come about because the mesoscopic structure acts capacitively; it only passes
the tunneling current on to the tip once its own energy levels have filled. The peaks are due
to the resumption of current flow once the structure’s energy levels are filled.
When a few atoms on the surface of the tip near its apex are not thoroughly connected
to the bulk material of the tip and can move somewhat freely along the surface of the tip,
the potential exists for these atoms to move into or out of the tunneling gap, thus changing
the effective tip–sample separation distance, which can drastically change the amount of
current detected. If a group of atoms farther away from the apex of the tip are not firmly
connected to the rest of the tip body, this satellite chunk of atoms can wobble back and
forth, changing the tip characteristics. A primary symptom of this is unstable spectroscopy,
as shown in Figure B.9. The wide variety of slopes in the I–V curves in that figure shows that
the tip–sample effective distance, which scales the tunneling resistance, is changing between
curves. Careful examination of the zero-current line in the I–V curves in Figure B.9 shows
116

dI/dV [arb. units]
0

-0.4

0
Bias [V]

+0.4

Figure B.8: dI/dV reminiscent of the Coulomb blockade effect convolved with proper
graphite spectroscopy. Repeated peaks at symmetric spacings along the voltage axis are
the chief diagnostic symptom, indicated with arrows. The offset of the bottom of the spectrum from 0 V could be due to Fermi level pinning due to contaminants on the surface.
that several of the curves do in fact show consistently zero current as the voltage is swept;
equivalently, these curves hug the bottom of the dI/dV plot. This occurs because the tip–
sample separation is set before each voltage sweep, with the voltage set at the scanning
bias voltage level, and the feedback is turned off during spectroscopy. Although the tip
and sample generally reach an equilibrium separation during this time, the changing voltage
during the voltage sweep changes the electric field between the tip and the sample, which
can tug unstable particles into different locations. This can push a few atoms at the end
of an unstable tip out of the tunneling gap, thus increasing the tip–sample separation such
that no current is detected, or vice versa.
Small portions of the sample can be adsorbed onto the tip, particularly those which are
loosely bound to the surface and attracted to the strong electric field between the tip and
the sample. Changes to the tip state, which can occur even in the middle of an individual
scan, can have significant effects on the position of peaks in the dI/dV measurements, causing

117

dI/dV [arb. units]

Current [nA]

4
0
-4
-0.4

0
Bias [V]

0

+0.4

-0.4

0
Bias [V]

+0.4

Figure B.9: Example of spectroscopy with an unstable tip. This dataset sums together two
spectroscopy points on graphite, which should result in a sigmoidal shape for the I–V curves
in the left image and a slightly rounded V-shape for the dI/dV curves in the right image.
The thicker, black line is the average of the spectra. Left: I–V curves. Right: Corresponding
dI/dV curves, obtained directly using a lock-in amplifier. The curves are not reproducible.
irreproducibility in the measurements [96]. For this reason, it is important to perform control
spectroscopy measurements before and after spectroscopy measurements performed on a
feature of interest in order to verify the pure quality of the tip. Figure 5.12 is convincing in
no small part due to the proper graphite spectroscopy acquired immediately before and after
the spectra acquired on the pilus. In multiple cases throughout this research, the “before”
graphite spectroscopy looked as expected and the spectra on the pilus candidate looked nongraphitic, but the “after” graphite showed signs of looking like the pilus candidate spectra. It
can be hypothesized that portions of the sample had transferred to the tip in those instances.

B.6

Multi-Characteristic Artifact

As a summary exhortation to believe a candidate feature is actually a pilus based upon
dI/dV spectroscopic identity with possible corroboration from other ways of interrogating
the sample, such as topography, rather than the other way around, this cautionary example
is here presented. This object shared most characteristics with what were considered to
118

be authentic pili and was often misidentified as such upon preliminary analysis. While
topographic scanning is an effective way to discover pilus candidates, classification of such
as authentic pili rests on the shoulders and peaks of spectroscopy.
The topographic characteristic most prominently shared between this object and previouslyobserved pili is multiple nested periodicities approximating those observed for protein nanowires.
Such an extensive level of topographic similarity between filamental biological objects and
artifactual propensities of graphite is precedented [89]. Suspiciously, the object shown here
was straight over at least 2 µm, save for dislocations where it zig-zagged as though relieving
stress between mismatched lattice boundaries, as seen in Figure B.10. Enhancement of the
local density of states at grain boundaries on graphite, leading to increased feature brightness in STM scans, as well as producing artifacts with small and large periods as the stress
at the boundaries is relieved, is a well-known graphitic artifact that could be responsible for
such an image [92].

100

nm

100

nm

Figure B.10: STM of a periodic feature. Bright spots coincident with longitudinal zig-zags
are approximately 100 nm apart. Sub-periodicity is approximately 2 nm. Left: Scan size
500 nm × 500 nm. Center: 125 nm × 125 nm. Right: 50 nm × 50 nm. (+ 0.5 V, 0.2 nA)
Adding yet further obfuscation, there is a multiple tip imaging this object. Not only is
there a far-away echo of the object near the center of the left image in Figure B.10, but the
smaller scan size shows that the object is shadowed by multiple image(s) of itself nearby.
119

Based on line sections through the brightest “strand,” the object itself is likely approximately
4 nm in apparent width, with the other image(s) of the feature being multiple-tip echo(es).
This is demonstrated in Figure B.11, where an alteration that was induced in the feature by
the performance of spectroscopic measurements is visible not only in the location in which
the spectroscopy was performed but also in the brighter view of the object above it.

Before

After

Points

Figure B.11: STM topographs of the object before and after point spectroscopy. Indicated
features in left and center scans are the same in both images, used to identify the extent of
lateral drift between scans. Right: Same as center, with locations of on-object and graphite
control spectroscopy marked by dots. The newly-acquired dimming of a spot on the lower
filamental shape is reflected by dimming on the upper filament, marked by arrows, indicating
that the filaments are a single object. Scan sizes 20 nm × 20 nm. (- 0.5 V, - 0.2 nA)

If this object is indeed an artifact, it is one which shows similar topographic characteristics
to objects often identified as pili. This serves as an enthusiastic reminder that beyondtopography measurements such as tunneling spectroscopy are key to discerning the identity
of sample components.

120

Appendix C
Sample Preparation
A primary challenge in scanning probe microscopy measurements of biological samples is
appropriate preparation of said samples such that realistic insight into the native properties of
the samples is obtained. Much of the difficulty lies in ensuring the sample will be sufficiently
stable on the substrate surface to obtain meaningful measurements [79]. This Chapter thus
discusses in detail the sample preparation used in this study.
Sample preparation for scanning probe microscopy studies of biological samples has long
been recognized as key to achieving good results [63]. It is important, particularly for STM
and spectroscopy which require electrical contact between sample and voltage source, that
the sample is sufficiently well-connected to the substrate. Furthermore, proper treatment of
the biological sample itself is important for producing samples which are amenable to the
intended experiments. The purified pili used in this research were not chemically fixed, in
contrast to the conventional tradition for observing biological samples, because the preceding room-temperature study had shown that chemical fixation tended to obscure periodic
features on the pili [23].

C.1

Sample Substrate Mounting

Graphite was chosen as a substrate for the samples in this study for several reasons. First, it
is an electrically conductive material, and STM and spectroscopy require electrical contact

121

to the sample in order to deliver the bias voltage. Next, it is a mostly-renewable resource
due to the now-classic mechanical exfoliation method [81], which exploits the weak bonding
between the stacked graphene layers which make up graphite. Adhesive tape is applied to the
top of the graphite crystal and peeled away. Several layers of graphite will prefer to adhere
to the tape rather than to the main graphite crystal remaining behind, thus exposing a fresh
graphite surface. This makes it a fiscally feasible substrate for experiments requiring frequent
redepositions of new doses of sample, as opposed to gold, which has also historically been
used for STM studies on DNA [97]. One disadvantage of graphite compared to materials
such as gold is that the edges of the layered flakes of graphite on the surface can show
enhanced electronic behavior [91] or curl up topographically, which results in features not
dissimilar in appearance to pili, as described in Appendix B and observed by other workers
[22]. Additionally, the cleaving procedure can result in fragmented particles remaining on the
surface, which can find their way onto the tip. Difficulties arise either when such debris is of
a size to exhibit Coulomb blockade effects in the spectroscopy [95] or when it makes the tip–
sample junction distance unstable, as discussed in more detail in Appendix B. Familiarity
with common graphite artifacts can go some way towards ameliorating this downside.
Samples were deposited on highly ordered pyrolitic graphite (HOPG), grade ZYA. This
was mounted on stainless steel sample disks in order to deliver the bias voltage to the
sample. The graphite was affixed to the stainless steel using silver paint, an adhesive with
silver particles dispersed within it, to provide a continuously conductive path between the
graphite and the sample disk. The crystal was centered on the disk such that the STM tip
could encounter the largest possible area of the graphite. Photographs of a mounted graphite
substrate are shown in Figure C.1.

122

Figure C.1: HOPG mounted on a stainless steel sample disk. Left: Graphite crystal mounted
on a stainless steel sample disk. The conductive adhesive used to mount the graphite on
the disk can be seen extending beyond the graphite footprint. Photograph taken through
an optical microscope at 20x magnification. Center: Graphite on sample disk, showing the
alignment marks used to center the graphite. Right: Graphite mounted on sample disk and
inserted into ramps. At this stage, the graphite sample is ready either to have pili deposited
onto it or to be mounted on the STM directly.

C.2

Sample Deposition

Sanela Lampa-Pastirk from the MSU Department of Microbiology and Molecular Genetics
prepared the biological sample. Pili were sheared from cultures of G. sulfurreducens grown,
prepared, and purified by Dr. Lampa-Pastirk according to the procedure laid out in Reference [25] based on the protocol in Reference [12]. Briefly and schematically, the bacteria
were grown under conditions that encouraged them to express pili, then the bacteria with
their attendent pili were broken apart mechanically and via detergent, and the resulting
amalgamation was separated by filtering away the other cell products through a porous gel
before rinsing and drying the pili which remained. At this point, the pili were resuspended
in a pH 9.5 solution to encourage deaggregation of the proteins so they would be deposited
as single nanowires rather than bundles, additional contaminants were removed, and the
sample was dried and frozen for storage. Before the pili were deposited for this research,
they were thawed and resuspended in doubly deionized water.
For the STM and point tunneling spectroscopy experiments that were the main focus

123

of this work, 10 µL of pili suspended in doubly deionized water, corresponding to a density
of approximately 5 mg/µL [25] were deposited using a micropipette onto a freshly-cleaved
graphite surface mounted on a stainless steel sample disk that fit into the ramps in the
microscope used for this study, as shown in Figure C.1. The sample was left to physisorb
onto the graphite for ten minutes, then the droplet was wicked away using filter paper,
and the sample was rinsed by depositing doubly deionized water of the same volume as the
sample droplet. The rinsate was left to soak for five minutes and wicked away, then the
rinsing process was repeated to make a total of two rinses after sample deposition. The
sample was then dried under a stream of N2 gas for approximately one hour. When the
sample was not being scanned or in the process of being mounted on the microscope, it was
stored in a container with a positive overpressure flow of N2 gas to minimize its exposure
to atmospheric oxygen and other contaminants. After the sample was mounted on the
cryogenic-capable STM, the microscope area was pumped down to a few or a few tens of
µtorr and filled with N2 gas before cooling down to 77 K by immersion in liquid nitrogen.

C.2.1

Deposition Protocol Tuning

While the sample deposition protocol from Veazey’s work [22] provided an excellent starting
point, further optimization of the deposition process was warranted for the experiments
discussed here. Further tuning of this protocol addressed some of the challenges faced in
obtaining reproducible, conductive samples with a sufficient spatial distribution of pili to
aid in timely identification of individual pili. While this protocol tuning was carried out on
silicon, covered with a layer of either native oxide or 300 nm of thermally grown oxide, this
was the origin of the protocol used to deposit the sample on graphite for STM.
Concerns over a possible less-conductive layer left behind on the sample after deposi124

tion stemmed from studies on pili deposited on gold electrodes formed on a silicon dioxide
substrate, discussed in Appendix E. During AFM investigations of these samples, it was
observed that regions of the sample appeared to be blanketed with a relatively soft and relatively non-conductive substance, unlike the bare gold that was expected. Two hypotheses
for the identity of this layer were that portions of the gold were nonconducting due to a
contaminant film on the surface or that the film itself was composed of pili matted together.
This concern was probed in part by varying the amount and type of rinsing and drying of
the samples after deposition, in case contaminants from the air had found their way into the
water in which the pili were suspended during the time allotted for deposition. Figure C.2
shows a few of the scans which seemed to indicate the presence of what is perhaps a dense,
near-monolayer mat of pili with trapped fluid in the interstices.

Figure C.2: AFM topographs of pili on SiO2 . Left: Pili on a circular area of SiO2 surrounded
by gold. The pili on the left of the circle follow tree-like branching patterns of bundling. The
pili on the right of the circle have perhaps trapped fluid between the filaments, resulting in
a “frayed blanket” look. Center: Another “frayed blanket” pilus formation. Right: Typical
“neuron” structure of pilus bundles on SiO2 . All scans are 5.7 µm in the vertical direction.
Left scan width 9 µm, z grayscale 40 nm; center scan width 7 µm, z grayscale 40 nm; right
scan width 5.7 µm, z grayscale 15 nm.

The other primary concern in obtaining a satisfactory sample deposition protocol was
optimizing the density of pili on the sample. Ideally, pili would be spread uniformly around
the surface and easily found with the scanning probes without being more than a monolayer
125

thick; however, the density and distribution when the sample was deposited on graphite were
such that this was not the case. In order to shed light on whether an insufficient number
of pili were making it to the surface from the droplet in which they had been suspended,
a series of deposition durations was investigated. Pili were suspended in a droplet of water
or of buffer, then deposited on a clean silicon chip and left for a range of times before the
droplet was wicked away using filter paper, after which the samples were either rinsed or
not, with the rinsate wicked dry using the same method. The resulting samples were then
imaged using amplitude modulation atomic force microscopy.
In the initial test of different deposition conditions, the pili samples were deposited onto
silicon chips on which the silicon was either covered with the few-nanometer native oxide
layer or with a thermally-grown SiO2 layer 300 nm thick. Each of these was cleaned either
according to a standard cleaning protocol similar to that given in Appendix E or using the
“RCA 1” cleaning protocol for more stringent chip cleaning [98]. On these four different
substrate preparations, then, was deposited a sample of pili suspended either in doubly
deionized water or in CHES buffer to maintain a pH of 9.5. Using doubly deionized water
would provide the benefit of not needing to consider possible artifactual effects in measurements due to the presence of buffer salts; using buffer to maintain a stable pH of 9.5 would
provide the benefit of deaggregating the pili [25] to make individual nanowires more accessible for measurement. The amount of time the sample droplet remained on the surface was
also varied in an endeavor to determine whether there would be a denser coverage of pili,
or a larger amount of surface contamination from the droplet, when the droplet was left on
the sample longer. Each sample was sealed using Parafilm (stretchable, wax-covered film)
into a plastic petri dish in company with a piece of lens-cleaning tissue soaked with water
to maintain the humid environment so the sample droplet would not evaporate. After the
126

allotted amount of time—30 minutes, 2 hours, 6 hours, or 12 hours—the droplet was wicked
dry and, in the case of sample deposited from pH 9.5 buffer, rinsed with doubly deionized water, before AFM imaging. The predominant impression from this investigation was
that, whether rinsed or not, and not strongly dependent on deposition duration, the buffered
solution was prone to leaving residue on the surface, as seen in Figure C.3. Consequently,
deposition from a suspension of pili in doubly deionized water was preferable. Note that the
buffer used in previous STM work on Geobacter [22, 23] was a different type, PBS, intended
to stabilize the pH around that of water, and thus this observation of buffer residue does
not reflect unfavorably on the conditions in that work. Note also that the residue in the
samples deposited from CHES buffer was localized; some portions of the sample appeared
as expected for the substrate. No pilus candidates were seen on these samples.

Figure C.3: AFM topography scans from the deposition duration investigation, selected to
highlight features from buffer residue. All samples shown here were deposited from CHES
buffer (pH 9.5). Left: 30 minute deposition time, rinsed, RCA 1 chip pre-cleaning. Center:
2 hour deposition time, solvent chip pre-cleaning. Right: 12 hour deposition time, rinsed,
RCA 1 chip pre-cleaning. All image sizes are 1 µm × 1 µm. Left z grayscale 6 nm, center z
grayscale 20 nm, right z grayscale 34 nm. Lozenge-shaped features are not bacteria, which
would be much larger and which were removed by the purification protocol.

Next, the deposition condition evaluation was repeated for only the 2 hour deposition
duration to reduce the number of variables. The pili, suspended in either doubly deion127

ized water or CHES buffer at pH 9.5, were deposited on silicon chips with either a native
or thick oxide layer, cleaned either using a standard solvent cleaning procedure or using the
RCA 1 process. Several of these preparations resulted in pili on the surface which were
imageable by AFM; a few scans of the most thoroughly piliated sample from this evaluation are shown in Figures C.4 and C.5. Networks of pili, superficially similar in appearance
to neural networks, were frequently seen, and intersections between pili in these networks
resulted in nodules larger than the sum of the pili sizes in the intersection. These nodules
are perhaps due to additional material, such as buffer salts or water, being trapped between
the overlapping pili. Furthermore, AFM phase imaging, which is sensitive to differences in
sample adhesion or softness as discussed in Appendix D, showed the presence of a film-like
material between the pili in these networks, as can be seen in Figure C.5.
The difference in distribution of pili on silicon dioxide compared to that on the graphite
surface used for the primary research project discussed here could potentially be attributed
to the hydrophobicity of graphite versus the hydrophilicity of the oxide-covered silicon—
even the nanometer-scale native oxide layer is sufficient to make silicon chips hydrophilic
[98]—and the consequent interaction with the droplet. While the hydrophilicity of the silicon dioxide surface made it better suited for adsorption of proteins in general [79] than
the graphite surface, its lack of metallic conductivity made it unsuited for the STM-based
experiments discussed here. The silicon dioxide substrate would introduce complications to
the interpretation of spectroscopic data and have a deleterious effect even on the acquisition of topographic images. The deposition conditions chosen for the primary research were
aimed to be as near in vivo pH as possible without adding buffer. Since individual pili were
successfully imaged on the graphite surface when deposited from doubly deionized water,
although not in the neural network configuration seen above, sample deposition was always
128

6 nm

pilus

nodule

0
30 nm

150 nm

0
Figure C.4: Highlights from a deposition test conducted with the droplet of pili in pH 9.5
CHES buffer remaining on the thermally oxidized silicon surface for 2 hours, then rinsed
twice with doubly deionized water. Upper Left: Focus on several individual pili interacting
at a nodule. Scan size is 150 nm × 150 nm. Upper Right and Bottom Row: Pilus network
topography examples from other locations on the same sample. Scale bars are 1 µm. The z
height shading scale is the same for all of these AFM images except the upper left scan.
from water. Additionally, it was found that a deposition time of 10 minutes was sufficient
to permit pili to be deposited with tolerable density on the graphite surface, and the rinsing
time was increased to 5 minutes for each of the two rinses. Thus, despite differences between
the surfaces, the deposition condition study on the silicon dioxide substrate provided a viable
sample deposition protocol for experiments with pili on graphite.
As a final caveat regarding sample preparation, Figure C.6 shows an AFM topograph of a
cell found on an electrode sample discussed in Appendix E. This sample was obtained from
purified pili [12], so this is not a G. sulfurreducens cell. The question remains open as to
from where this cell originated. Possible sources include pre-deposition contamination of the
129

height

phase

phase
800 nm

25 nm

0

film

800 nm

Figure C.5: Scans showing “film” between pili, from the same deposition test as Figure C.4.
Left and Center: Height and phase images on the same feature area. Scale bars are 1 µm.
Right: Phase image of an 800 nm × 800 nm portion of the “neural network” shown at left.
substrate, pre-deposition contamination of the water used to suspend and rinse the pili, or
contamination of the sample during the deposition process. The mats of pili radiating from
what appears to be a flagellum imply that the cell and its flagellum were in place either before
or at the same time as the pili were deposited. If the cell had encountered the pili mats after
the sample was dried, it seems slightly unlikely that it would have “stirred” the pili into such
a configuration. Figure C.7 shows a smaller scan range view of this. Since this was the only
cell observed during the course of this research, it can be considered an exotic contaminant.
While the cell in Figures C.6 and C.7 is almost certainly not a G. sulfurreducens bacterium,
it is useful to include for perspective on the relative sizes of the Geobacter pili of interest
compared to more-familiar objects such as cells and flagella. This cell of unknown origin
is of a typical size for a bacterial cell—approximately 1 µm long, 0.5 µm wide, and 0.1
µm tall. The flagellum is approximately 5 µm long, 0.1 µm wide and 30 nm tall—an order
of magnitude taller than Geobacter pili.

130

5

non-Geobacter
cell
um

non-Geobacter
flagellum

0

Geobacter
pili
um

0

5

Figure C.6: Unexpected cell observed on a sample using AFM. The lower substrate surface
is SiO2 , and the upper surface is a 25 nm-tall evaporated gold electrode.

1.75

1.75

0

45
nm

um

um

190
nm

0
0

um

0

1.75

0
0

um

1.75

Figure C.7: Unexpected cell observed on a sample using AFM, seen in its entirety in Figure C.6. Left: Cell and flagellum on gold. Right: Flagellum on SiO2 with G. sulfurreducens pili forming mats in response to it.

C.2.2

Spatial Density of Deposited Pili

Because of the uncertainty as to how conductive the pili would be at low temperature, thus
how well they would be identifiably imaged by STM, a simple statistical argument was
made to tie the observed density of pili on the sample scanned at room temperature to the
predicted density of pili on the sample when scanning at low temperature. One significant
assumption in this estimate was that the pili were uniformly distributed on the surface.

131

Since this was generally observed not to be the case, this estimate merely provided a rule of
thumb for whether candidate features seen at low temperature were likely to be pili, even if
the spectroscopy proved to be unstable.
In the systematic search for pili, it was convenient to define “regions” of the sample
surface which were scanned in a uniform fashion in order to facilitate comparison. Since
multi-micron–scale lateral (XY) sample translation was based on the stick–slip method, as
discussed in Chapter 3, the regions were generally non-overlapping. It is a fair statement
that the regions are independent samplings of a several mm2 surface. The pili are only a
few nanometers in diameter, so a multi-micron scan size divided into 128 pixels would limit
the apparent pilus size to smaller than one pixel. In order to facilitate visual identification
of pili, the regions were divided into individual scans smaller than 1 µm × 1 µm in area, as
shown by the nine boxes comprising the region in Figure C.8. To ensure systematic coverage
of the entire region and to avoid inadvertent redundant scans, a specific search pattern was
followed of offsets moving to new scan areas within each region, as shown in the same figure.
In the initial room temperature data run, two pilus candidates appeared in the first
six scanned regions, with one “region” systematically defined as in Figure C.8. Using the
estimated scan area size decrease by a factor of 9 at 77 K as compared to its room temperature
area, decreasing by a factor of three in both lateral directions due to thermal effects on
the piezotubes as described in Chapter 3, the number of pilus candidates seen in a given
number of regions should decrease correspondingly by a factor of 9. This was found to
be approximately the case in the first low-temperature data run, as two pilus candidates
were seen in 43 regions. Comparing the number of regions needed to see two candidates,
6
43

≈ 0.14. This has a ∼20% discrepancy with 1 ≈ 0.11, which is not unexpected given
9

the small statistics for the purposes of this calculation. This lent further credence to the
132

first
scan

one region

one scan

40 V ~ 0.8 um at 300 K
~ 0.3 um at 77 K

last
scan
Figure C.8: Schematic of the search pattern typically followed within a particular “region.”
Each individual scan area of the nine scan areas comprising the region was 40 V × 40 V
in size, where this voltage is the voltage applied to the scanning piezotube’s quadrants. At
room temperature, each scan area was approximately 0.8 µm × 0.8 µm, while at 77 K, the
same applied voltages resulted in a scan area of approximately 0.3 µm × 0.3 µm, to one
significant figure. Each of the nine boxes within the region is a complete scan.
determination that the candidate objects seen at 77 K were indeed pili. Later, spectroscopy
was successfully obtained on objects with similiar topography, identifying them as pili.

C.2.3

Reproducibility of Deposited Samples

While substantial efforts were made to ensure that all samples were identical, some sample
depositions did appear to have more pilus candidates than others. In part, this could have
been due to the low statistics—if only a handful of individual pilus candidates were expected
for each sample, based on the rough calculation above, a quantity difference of only a few
pilus candidates between different samples could appear to be substantial while actually
being a normal amount of scatter between samples.
For some sample depositions, the deposited droplet would “wet” the surface of the
graphite differently. Wetting of a surface, or the amount a droplet spreads when deposited
onto the surface, depends on the polarity of the surface and of the fluid. Water deposited on
133

hydrophobic surfaces forms droplets with minimal droplet area in contact with the surface.
The graphite surface is hydrophobic, in contrast to another cleavable substrate, mica [99],
which is also often used in scanning probe microscopy experiments. To cause a difference
in the wetting of the sample surface, either something on the surface was modifying the
hydrophobicity of the graphite, or something in the sample droplet was modifying the polar character of the water. In general, the graphite sample was cleaved immediately prior
to deposition. In a few cases, the graphite was cleaved, proper graphite spectroscopy was
verified using the STM, then the sample was deposited. In each case, since a water layer
rapidly forms on graphite exposed to air, a thin layer of water was present on the graphite
surface before sample deposition, potentially with associated contaminants. Since the sample droplet was exposed to air, albeit in a covered Petri dish, contaminants could also have
entered the water of the sample droplet, although this would not be responsible for a difference in wetting at the moment of deposition. Despite this, the potential contaminants did
not prevent acquisition of correct control spectroscopy on graphite during the experiments.
Some variation among samples could occur because the pili were suspended in the water,
not dissolved, and pili had a tendency to aggregate. Since often only a few microliters of
sample were deposited from the often several µL total volume of water in which the pili
were suspended, the density of pili in the selected droplet could vary. It can be considered,
however, that the anisotropic distribution of the pili on the surface played a greater role in
the difficulty of finding individual pili to study. The droplet size was deliberately chosen
not to cover the entire graphite surface; when water spills over the edges of a graphite slab,
it could intercalate between the layers that comprise the graphite, loosening the connection
between layers and, consequently, encouraging instabilities. In addition, edges and corners
of the graphite slabs tend to possess flakes and outcroppings which can be quite hazardous
134

to STM tips, so the outer edges of the graphite slab were experimentally unusable. This
means that the pili-containing droplet was confined to the center area of the graphite; even
so, the graphite area was at least slightly larger than the droplet. After the droplet had been
repeatedly rinsed and wicked dry, a sample without significant contamination was invisible
by eye, thus positioning the STM tip in a specific location with respect to where the droplet
had been could be a minor challenge. Making matters even trickier, once the sample was
inside the cryogenic STM, it was no longer optically visible, and sample motion choices were
based on a mental model of the overall structure of the sample. Considering that the droplet
size was on the order of a millimeter in diameter and the largest scan size in which pili could
be comfortably identified at room temperature under the scanning conditions used here was
on the order of one square micron, starting in the “wrong” place on the sample for finding
pili could lead to substantial amounts of time spent imaging graphite untouched by the
sample droplet. For this reason, an apparent differing number of pilus candidates per scan
time among samples, as in the ∼20% difference between room temperature and 77 K pilus
densities discussed above, could also be impacted by the amount of time spent searching for
pili on bare graphite where the sample had not been deposited, and hence should not be
taken as an indicator of significant differences between samples.

C.2.4

Contact Between Pili and Substrate

STM imaging and stable spectroscopy rely on a consistent delivery of bias voltage to the
sample. In some cases, it appeared as though the tip was moving the sample around on the
substrate and, consequently, imaging it poorly, as in Figure C.9. In other cases, segments of
the pili would appear and disappear between scans, which was attributed to the tip having
perturbed the pilus out of contact with the surface. Additionally, in Figure C.10, it appears
135

as though the influence of the tip has caused part of the pilus to move back and forth along
the surface. In particular, it appears as though the pilus shown in the left and center images
in Figure C.10 is being moved by the tip. The retrace scan was obtained as the tip moved
backwards into position to begin scanning the next line of the trace scan. It appears that
the lower part of the vertically-oriented filament is in poor contact with the surface, as it
is being imaged sporadically. The scan on the right shows the vertical tendril being imaged

scan direction

scan direction

more stably after it appears to have been wrenched from the upper part of the pilus.

Figure C.9: A single STM topography scan. The tip moved down and then up along the
same line of the image before proceeding to the next line. Tip-induced sample motion can
be seen most clearly in the center of the scan, where nanowire is perturbed in the direction
of scanning. Scan size is 763 nm vertically and 749 nm horizontally. The apparent breadth
of the nanowire is likely due to a combination of the convolution of the nanowire topography
with the tip’s radius of curvature, ambiguous resolution of the edges of the nanowire, and
motion of the nanowire along the scan direction. (+ 0.75 V, 0.2 nA)

In many cases, the most stable spectroscopy was on pili which were “anchored down”
by a graphite flake, as though part of the pilus had become lodged under the flake’s edge
during deposition. These pili could be stably scanned for many hours repeatedly without
degradation, showing that STM imaging under normal conditions did not discernibly alter
the sample. Pili which protruded from aggregate bundles, as in Figure C.11, also seemed to
be anchored down by the other pili in the bundle. These anchored-down pili did not show
136

Trace

Retrace

Slightly later

Figure C.10: STM scans of one filament. Arrows indicate a feature useful as a landmark.
An unstably imaged tendril is visible below the arrow. Left: Scan of a pilus unstably affixed
to the surface. Center: Near-simultaneous scan but with the tip moving in the opposite
direction. Right: Smaller-scale scan. Left and center scans are 400 nm × 400 nm; right scan
is 100 nm × 100 nm. (+ 0.5 V, 0.2 nA)
the sporadic imaging problem shown in Figures C.9 and C.10 and thus were good candidates
for tunneling spectroscopic measurements.

Figure C.11: Series of STM scans zooming in on the croissant-shaped mandibles—two individual pili—protruding from an aggregate bundle of pili. Pili found near bundles or graphite
flakes were generally stable on the surface over hours of scanning, perhaps because of being
anchored to the surface by the larger objects. Scale bars 20 nm. (+ 0.5 V, 0.1 nA)

Once a (relatively-)stable pilus was found on the graphite substrate, STM topography
and point tunneling spectroscopy could be obtained.

137

Appendix D
Further Details
While the discussions and explanations in the main text can generally stand alone, this
Appendix probes a little deeper into some of the details of the techniques and ideas relevant
to this research. Particular emphasis is given to further discussion of the scanning probe
microscopy techniques and apparatus presented in Chapters 2 and 3.

D.1

Feedback

For the scanning probe microscopy modes used in this study, the tip–sample separation is
adjusted dynamically during scanning to reflect changes in the sample topography, convolved
with the sample’s density of states in the case of STM. The feedback signal for amplitude
modulation AFM is the amplitude of the cantilever’s oscillation, but otherwise, the feedback
behaves as it does for STM, where the feedback signal is the tunneling current, so only STM
feedback will be discussed here for simplicity.
In this feedback system, the signal of interest is the tunneling current detected by the
electronics. This detected tunneling current is compared to a user-defined current “setpoint”
value. If the detected current differs from this, the tip–sample separation is increased or
decreased until the detected current matches the setpoint value. This is “negative feedback,”
in that it acts to minimize the difference between the actual and setpoint values. As the
tip is scanned along a surface, it encounters variability in the surface topography or local

138

density of states of the sample. This causes the tunneling current to change, which requires
intervention by the feedback system. Parameters affecting the response time of the feedback
loop can be adjusted by the equipment operator. Both the gain and the time constant of the
feedback loop can be adjusted, and their effects are not independent. In a simple picture,
the time constant can be thought of as describing the reaction time of the feedback loop, and
the gain can be thought of as the enthusiasm of response of the feedback loop. Consider the
case of the STM electronics detecting a too-large tunneling current signal as mapped onto
the case of little Miss Muffet seeing a spider sitting down beside her. The time constant
of the spider-proximity feedback loop would then describe how quickly she was frightened
away, and the gain would be the extremity of her fright impelling her to vacate the premises.

D.2

Piezoelectric Rastering

In order to raster the scanning probe over a sufficiently precise area in a well-controlled
manner, either the tip or the sample has its motion governed by piezoelectric elements.
Piezoelectric materials expand or contract slightly in response to applied voltages, so a tip
or sample supported by piezoelectric elements with the expansion directions appropriately
oriented can be moved in a controlled, nanoscale manner by applied electronic signals.
The piezoelectric elements in the cryogenic-capable STM used for this research are tubeshaped, hence their appelation “piezotubes.” These thin, hollow cylinders of the piezoelectric ceramic material PZT—lead zirconate titanate, a mixture predominantly composed of
PbZrO3 and PbTiO3 [30]—have metallic contacts applied to them to which wires are soldered which carry the voltages from the STM controller. There are four electrical contacts on
the outside of the piezotube and an interior metallization providing the ground for the volt-

139

age applied across each quadrant. When a particular voltage is applied to one quadrant, the
piezoelectric dipoles align such that the material of that quadrant contracts. When a voltage
of the same magnitude and opposite polarity is applied to the quadrant opposite that one
across the diameter of the piezotube, that opposite quadrant expands by the same amount.
Thus the entire tube bends like an elephant’s trunk. For small enough amplitudes of the
bending compared to the diameter of the piezotube, this can be considered as a longitudinal
translation of the top part of the piezotube, as schematically sketched in Figure 3.2.
Because of hysteresis in the piezoelectric material, the piezoelectric elements take a nonnegligible amount of time to reach the elongation or contraction appropriate to the voltage
applied to them. This is colloquially known as “piezo creep” and causes image artifacts
such as apparent bending of straight features near the beginning of a scan, caused by the
piezoelectric elements not yet reaching their full response to the applied voltage when the
probe begins scanning; an example of this is shown in Figure D.1. This can be ameliorated
by rescanning the region, sometimes repeatedly, to allow the piezoelectric elements to reach
their full elongation or contraction before the production scan is acquired. Consequently,
the first few lines of a scan may often be justifiably disregarded.

D.3

Image Processing

While raw scanning probe microscopy data can often be used to view the sample properties
of interest, interpretation of same generally requires a certain amount of image processing
as part of the analysis. This is not generally a case of “making it look prettier,” but rather
of “making it more intelligible,” as artifacts and background effects can mislead both the
experimenter and the experimenter’s data extraction routine. The most basic and most

140

scan
direction
piezo creep
effect
Figure D.1: Example of “piezo creep.” Scan began from bottom of image and proceeded
upward. The apparent bending of the image near the start of the scan is due to a delay in
the full response of the scanning piezotube to the voltage applied to it. This scan of a silicon
dioxide surface was obtained using a Dimension 3100 AFM with Nanoscope IIIa controller.
Scan area is 1 µm × 1 µm.
harmless image processing techniques will be discussed here; many of the scans included
in this thesis have had one or more of these processing techniques used on them. When
measuring feature heights and breadths on a sample, it is particularly useful to process
the data, as background properties of the sample could swamp the small dimensions of the
features. Sometimes, it is sufficient simply to optimize the colorscale in which the scan is
displayed, as in Figure D.2; default display settings may not be appropriate for the scan
currently under consideration.
Most samples are tilted at an angle to the tip, as can be seen in Figure D.2. This
is unsurprising, considering that achieving a sample which is flat to within less than a
nanometer in height over a square micrometer or so of area when the sample is loaded
macroscopically is an occurrence of some statistical unlikeliness. In order to correct for this
near-omnipresent tilt, a background plane can be subtracted from the data. This can be
quite important when measuring heights, as the background tilt can change the apparent
heights considerably. After the plane subtraction, the line section across the graphite step
141

150

150

nm

nm

nm
0

5 nm

1 nm

0

0

150

0

nm

150

Figure D.2: Changing the colorscale used to display the data. Left: Raw image, as displayed
by the Asylum Research data acquisition software. Right: After “auto” color range and
offset applied. Both images show an overall tilt whereby the lower left of the image appears
higher than the upper right of the image. (+ 0.5 V, 0.2 nA)
in Figure D.3 looks much more flat and step-like, providing a better baseline from which to
measure the height of the filamental object.
Separately, or in tandem with a plane subtraction, line-by-line background subtractions
can be made. This accounts for small variations in the tip–sample separation as a function
of which “line” is being scanned at the time. Line offset subtractions are done by finding the
average height of all the data in a scan line along the fast scan direction and subtracting that
average value from all the data on that line; this is repeated for each line of the data. As a
mental model for this terminology, consider a manual typewriter. The fast scan direction is
the direction in which the typist is typing, and the slow scan direction is along the length
of the page, slowly incremented by carriage returns. For example, the faint dark “bands” in
Figure D.3 can be diminished by subtracting the average offset of each line, as is shown in
Figure D.4. Caution should be used when choosing to apply this processing step to scans
with features which align along the fast scan axis, as their heights with respect to the rest
of the image would then be falsely subtracted away to match the rest of the image.
If features on the surface are significantly different from the background and including
142

0.979 nm

5 nm

0

2 nm

Unprocessed
nm

150

0

nm

0

150

50

100
150
Length [nm]

0

0

nm

150

1 nm

4 nm

Planefit
nm

150
1.29 nm

0

50

100
150
Length [nm]

Figure D.3: Subtracting a background plane from the data. Top: Colorscaled data as seen
in Figure D.2. Bottom: Data after a first-order plane subtraction was applied in the XY
direction. Right: Line sections through the images on the left. The apparent height interval
between the blue markers changes significantly with processing. (+ 0.5 V, 0.2 nA)
them in the processing fits would cause further artifacts, they can be “masked” away such
that the processing does not include the properties of that region. This is generally not
necessary for small objects such as the pili discussed here, although it was sometimes necessary when the pili were at unusual angles to the scan direction, which affected the ability
to perform line offset subtraction. As an example of an image that benefited from masking
during image processing, Figures C.6 and C.7, which displayed a large cellular contaminant
on the surface, required masking in order to avoid including the large height of the cell in
the background subtraction.

143

After processing
and grayscale selection
150

150

0

nm

nm

0
150

0

1 nm

nm

5
nm
4 nm

0

Original display

0

nm

150

Figure D.4: Applying an offset along the X direction to the data from Figure D.3. Left:
Image after subtracting a plane in the XY direction, subtracting the average offset of each
line in the slow scan direction, and choosing the range and offset for the grayscale to match
that in the processed image in Figure D.3. Right: Data as originally displayed in Figure D.2.
Image processing done in Igor Pro using the Asylum Research software. (+ 0.5 V, 0.2 nA)

D.4

Direct and Numerical dI/dV Determination

Two ways to obtain dI/dV spectra are prevalent in the field. The first, used in this study, is to
obtain I–V curves and differentiate them numerically. The second is to obtain dI/dV curves
directly by means of a lock-in amplifier. When using the lock-in method, a small modulation
is applied to the bias voltage with a frequency much faster than the voltage sweep, which
causes the current to change correspondingly, yielding ∆I/∆V at each voltage step. This can
give better resolution than the numerical differentiation method, yet it is also susceptible
to simultaneous inadvertent measurements of other properties of the system, which can
sometimes lead to differing results between the direct and numerical dI/dV curves. Since
I–V curves are a more fundamental measure of the sample properties than are dI/dV curves
obtained via the lock-in amplifier, in case of discrepancy, the I–V curves should be hewn to
and numerically differentiated, as was done in this work.

144

Because of inherent scatter in I–V curve points, differentiation without smoothing beforehand resulted in noisy numerical dI/dV curves, as the differentiation interpreted transitions
between scattered points as small steps of steep slope. This can be seen in Figure D.5, where
the numerically-differentiated dI/dV looks impenetrably noisy. The averaged I–V data were
imported into Igor Pro software and smoothed with a 2nd order Savitzky–Golay filter [100]
using a 47-point window. This was found to be a good compromise level of smoothing,
avoiding histrionics in the differentiation while not smoothing over meaningful features in
the spectra. Additionally, the frequent flattening-out of I–V curves near the end of the bias
sweeps led to sudden drops in the derivatives near the endpoints of the voltage sweeps. In
general, the numerical dI/dV presented here have their displayed abscissas pruned to avoid
this; most voltage sweeps in this work were from + 0.5 V to - 0.5 V. The benefit of this can
be seen in Figure D.6, removing the confusing domination of these endpoint effects over the
more physically interesting and relevant features in the spectra at voltages nearer zero bias.

0

- 0.9
- 0.5

0
+ 0.5
Bias [V]

0
- 0.5

dI/dV [arb. units]

dI/dV [arb. units]

Tunneling Current [nA]

+ 0.9

0
+ 0.5
Bias [V]

0
- 0.5

0
+ 0.5
Bias [V]

Figure D.5: Five immediately-sequential voltage sweeps above graphite. Black curves are
averages of these curves. Left: I–V curves. Center: Minimally smoothed numerical derivatives calculated by the RHK data acquisition software. dI/dV = 0 is the bottom of the plot;
parts of these curves also extend, unphysically, below that level. Noise dominates this plot.
Right: Direct dI/dV via the lock-in technique. dI/dV = 0 is the bottom of the plot.

145

0
+ 0.5
Bias [V]

0
- 0.4

dI/dV [arb. units]

dI/dV [arb. units]

dI/dV [arb. units]
0
- 0.5

0
Bias [V]

0
+ 0.4 - 0.4

0
+ 0.4
Bias [V]

Figure D.6: Further analysis of averaged spectroscopy from Figure D.5. Left: Numerical
derivative of averaged I–V curves after smoothing with a 47-point Savitzky–Golay filter
using a 2nd-order polynomial to calm down some of the scatter in the data. Center: Same
as left but only plotted between ± 0.4 V. Right: Direct dI/dV via the lock-in, averaged and
replotted on same scales as center plot. Black lines at bottoms of plots are dI/dV = 0.

D.5

Tip Characterization

To ensure that a sufficiently high-quality tip existed at the start of each experiment, tips
were characterized by scanning bare graphite before sample deposition. In some cases, the
graphite was scanned immediately before deposition to ensure the graphite cleave was of
high quality and did not leave loose graphite dust strewn about the surface; in others, the
graphite was recleaved immediately before deposition. Neither of these procedures appeared
to affect the appearance of the graphite substrate during the experiment. Graphite is a
layered material and hence presents step edges of knowable sharpness against which the
aspect ratio of the tip can be tested. From the apparent sharpness of the graphite step
edges, the approximate dimensions of the pili, when deposited, can be predicted for that
particular tip.
As a “reality check” for the spectroscopic measurements on pili, point spectroscopy mea-

146

3 nm
2 nm
1 nm
0 nm

3 nm
2 nm
1 nm
0 nm

Figure D.7: Left: Standard and 3D view of a graphite step, several monolayers in height,
obtained using the cryogenic-capable STM. Scan is 342 nm in the X direction and 418 nm in
the Y direction. (+ 0.5 V, 1 nA) Center: 2 µm × 2 µm scan of overlapping graphite steps,
obtained using the Cypher STM. Right: 3D display of overlapping steps. (+ 0.5 V, 0.2 nA)
surements were routinely performed on graphite before depositing the sample. This could,
for example, help diagnose difficulties in the electrical connection delivering the bias voltage
to the sample. In addition, control spectra were obtained on graphite before and after spectra were taken on locations on pili, in order to verify that the tip was still giving realistic
spectroscopy as expected for a Pt:Ir tip on a graphite surface and that the tip hadn’t picked
up a particle from the sample, nor that there was a loose piece of tip which was mechanically
unstable. A more thorough discussion of spectroscopic artifacts can be found in Appendix B.
Once an STM tip developed an undesirable characteristic such as a multiple tip or unstable behavior during scanning or spectroscopy, there were several methods by which it could
be reformed into a more convenient shape. If a tip had acquired a contaminant particle,
shaking the tip to dislodge it mechanically could be productive. This could also be useful if
one of the STM “minitips” at the end of the tip was somewhat loose and causing unstable
imaging or spectroscopy. Additionally, strong forces from the electric field between the tip
and the sample in STM could be used to pull off contaminants or unsatisfactory minitips,

147

leaving a different minitip as the apex of the STM tip. AFM tips have more limited recovery
options than STM tips and were customarily replaced rather than recovered.

D.6

AFM Phase Imaging

In amplitude modulation atomic force microscopy, the cantilever is driven in oscillatory motion very near to the sample surface, and the tip’s interactions with the surface affect the
oscillation of the cantilever. The feedback signal for the AFM in this mode, and thus the
source of the height imaging signal, is the change in the tip–sample separation needed to
keep the cantilever’s oscillation amplitude constant. Surface properties can also affect the
phase of the cantilever’s oscillation. In essence, interaction with the sample shifts the resonance frequency of the cantilever from its free space value [101], which shifts the resonance
frequency of the cantilever with respect to the frequency at which the cantilever is driven.
This frequency shift can be described by a phase. Areas of a sample which are stickier can
delay the cantilever’s return to the peak of its amplitude, and areas of a sample which are
softer will absorb more of the cantilever’s vibrational energy [102] and thereby damp the
resonance frequency of the cantilever. Thus the phase contrast in different parts of a scan
can be used to detect surface properties other than height using AFM. Although this is qualitative [102] because it is difficult to disentangle the effects of softness vs. stickiness without
using complementary techniques, it is a tremendously useful technique for detecting different
types of materials which may differ only very slightly from one another in height. Indeed,
for studies of soft pilus nanowires on a hard SiO2 surface, the pili were generally more easily
identifiable in phase than in height images. This also facilitated the identification of patches
in Figure C.5 as being a different substance rather than a layer of well-packed pili.

148

D.7

Conductive AFM

Conductive AFM (often abbreviated C-AFM, or, more commonly in the field of protein
nanowire studies, CP-AFM for “conductive probe” or “conducting probe” AFM) is a natural tool to use for scanning probe electronic measurements of Geobacter pili. While an
overview is given here, results using this technique are discussed in Reference [25]. Conductive AFM tips are generally standard AFM tips coated with gold or platinum to provide
a conducting path from the tip to the measurement electronics. Unlike in standard AFM,
the tip–sample system is biased to induce an electrical current for the measurement. Unlike
STM, this current only flows when the tip is in contact with the sample surface. The tip can
be scanned in contact with the surface to detect this current at all points on the sample as
a function of the applied voltage, with the tip pressing into the sample with a nanonewtonscale force to ensure stable electrical contact. An alternate method of obtaining conductivity
measurements using AFM is by scanning the sample in intermittent-contact mode, then carefully selecting points and performing I–V curves with the tip coming into contact specifically
with these points, a direct-conduction implementation of point spectroscopy.
If these measurements are performed on a conductive substrate, this can give information
about the transverse conductivity of a sample such as pilus nanowires. STM was chosen over
CP-AFM to address the electronic properties of Geobacter pili for the research described
here, for several reasons. First, the spatial resolution of STM is generally higher than that
of intermittent-contact AFM in an ambient environment, which is often used for CP-AFM
studies of proteins’ transverse electronic properties. Second, STM perturbs the structure of
the pili less than CP-AFM does, since it does not need to press the tip into the sample in
order to obtain electronic information. This is an important consideration, both from an

149

in-principle and from a practical standpoint. From an in-principle perspective, touching the
sample with an AFM tip can alter the sample’s structure if too much force is applied; the
tip can press into or even punch through the pilus, complicating interpretation of results,
as it would be difficult to determine exactly how much of the pilus was in contact with the
tip [59]. Practically, the AFM tips tend not to be positioned solely atop pili when doing
these measurements, and much care needs to be taken to avoid applying too much voltage
between the tip and the surface to avoid tip damage. A voltage magnitude which may be
necessary in order to cause a pilus with poor electrical contact to the sample substrate to
conduct may cause a high enough current to melt the conductive coating from the tip when
the tip touches the conductive substrate. Third, the apparatus and expertise were readily
available to perform the temperature-dependence studies using STM.
In addition to this use of CP-AFM to probe the transverse conductivity of the sample, the CP-AFM tip can be used as a movable probe point for a two-terminal electronic
transport measurement along the longitudinal direction, if another immobile terminal is provided for the measurement. Microfabrication of electrodes for a two-terminal conductivity
measurement using CP-AFM is described in Appendix E.

D.8

STM Topography of Pilus Ends

Complementary to the scans focusing on the middle reaches of pilus lengths, topography
was obtained on the ends of nanowires, as shown in Figure D.8. These ends display a
characteristic notch several nanometers long where one side of the pilus is truncated before
the other side ends. Detailed observation of the left image in Figure D.8 shows that this notch
is longer than the longitudinal separation between nodules. Because the pili were purified, it

150

is not certain whether this “end” is the root of a pilus as would be anchored in the bacterium,
the tip of a pilus farthest away from the bacterium, or a breakage plane along which the
pilus was fragmented. The end could be speculated to be composed of an incomplete turn of
the helical pilus coil, and the tail part which continues past the notch consists of the heads
of pilins forming a partial coil. Alternatively, the end could be speculated to be composed
of exposed α-helices which customarily are buried on the inside of the pili. In the “broken
pilus” case, either of these speculations could apply, or the pilus could be sundered at another
point. Further investigation would be necessary to disentangle these options to identify these
notches. An exquisitely careful deconvolution of tip shape to obtain a more accurate lateral
measurement of the widths of the notches, inspired by the pilin structure [52, 53], could
assist in this determination.

Figure D.8: STM topographs of the “ends” of two separate pili showing the characteristic
notch. Both images are 75 nm × 75 nm. Both scans were performed at an applied bias of
+ 0.5 V. Left image tunneling current setpoint 0.2 nA, right 0.08 nA. The ghostly “echoes”
in both images are due to a multiple tip effect.

151

Appendix E
Microfabrication of Electrodes
While this Appendix does not include experimental data from the study of G. sulfurreducens pilus nanowire longitudinal conductivity (see instead Reference [25]), methods for
photolithographically fabricating the electrodes for that study are included here.

E.1

Goal and Starting Point

The conductive probe atomic force microscopy (CP-AFM) measurement using the electrodes
whose microfabrication is described here aimed to investigate the lateral conductivity along
the pilus nanowires. In brief, the idea was to let a microfabricated gold pattern on an
insulating substrate act as one terminal of a resistance measurement and the conductive
AFM probe act as the other terminal. By making resistance measurements at different tip–
electrode separations along a pilus, resistance as a function of tip–electrode separation would
permit determination of the resistivity of the pilus independent of the contact resistance
between the pilus and the electrode and between the pilus and the tip. Because the tip was
in direct contact with the pilus, contact resistance due to the tip was expected to depend
also on the amount of force applied by the tip.
This project was a continuation of the work of Joshua Veazey and Sanela Lampa-Pastirk
in collaboration with Jiebing Sun and Pengpeng Zhang. This team had obtained one measurement of the pilus longitudinal conductivity [22] which needed reproduction and expan-

152

sion. Further refinements were made to the procedure used for that measurement in an
endeavor to expand that study. Because of the priority of obtaining cryogenic STM and
point tunneling spectroscopy results, this project was continued, and successfully concluded,
by Sanela Lampa-Pastirk [25].
The objective of the microfabrication of these electrodes was to create continuous, yet
thin electrodes over which pili could drape, partially on the electrode and partially on the
nonconductive substrate. An AFM topograph of pili draping over an electrode edge in this
fashion is shown in Figure E.1. The tip would be positioned at various locations along the
pili for the CP-AFM measurement.

E.2

Photolithography

The electrode patterns were fabricated using ultraviolet photolithography via an AB-M Mask
Aligner in the W. M. Keck Microfabrication Facility Class 100 cleanroom at MSU. The gold
electrodes were subsequently thermally evaporated in the adjoining Class 1000 cleanroom
using an Edwards Auto 306 Thermal Evaporator. The resulting electrodes were characterized
using one or both of the following atomic force microscopes: a Digital Instruments Dimension
3100 with Nanoscope IIIa controller for topography and an Asylum Research Cypher for
topography and conductivity.

E.2.1

Materials

Electrodes were microfabricated onto silicon substrates covered by 300 nm thermally-grown
oxide layers, obtained from Silicon Quest International after the optimal substrate supplier
was determined. Because of the thick thermal oxide layer, the substrates were highly resistive

153

Figure E.1: AFM topograph of a microfabricated electrode and pili. Scan is 3 µm x 3 µm.
The brighter region on the left is a gold electrode; the lower, darker-shaded region is the
insulating substrate, exposed as a Swiss-cheese-like circular cut-out in the electrode. The
tendril lying across the electrode and substrate is a small bundle, likely a pair, of pili.
and unlikely to give spurious current. Electrodes were fabricated out of gold of 99.99%
purity atop a titanium “sticking layer” deposited in order to facilitate bonding of gold to
the substrate. For most of the electrodes fabricated for this project, a good compromise
enabling low electrode heights to decrease the steepness of the edges the pili would have to
drape over while maintaining a continuous layer of conducting material was reached when
the electrode consisted of 23 nm of gold atop 2 nm of titanium.

154

E.2.2

Procedure

The following describes the final version of the photolithography procedure using this photomask. Like most photolithographic processes, many iterations were required to devise a
procedure that gave acceptable outcomes. The overall procedure is schematically shown in
Figure E.2; the realities of photolithography process development are more subtle.

Figure E.2: Diagram of the photolithography procedure, not to scale. Left to right: substrate
with photoresist deposited on top, photoresist exposed and developed to leave a pattern, gold
deposited on the pattern and photoresist, photoresist “lifted off” to leave solely the pattern.

Chip Cleaning
The surfaces supporting the electrodes must be extremely clean, lest the electrodes adhere to a removable contaminant layer rather than to the chips themselves. Before prephotolithography cleaning, thermally-oxidized silicon wafers were cut into 1/2 inch × 1/2
inch chips. Dicing the chips rather than hand-cleaving them with a scribe resulted in more
regularly-shaped chips and, therefore, a more reproducible photolithography protocol. A
layer of photoresist was spun onto the wafer to protect the surface during dicing. The chips
were then cleaned according to the following protocol:
Initial removal of protective resist, if chips were diced
• Ultrasonicate for 1 minute in acetone, rinse with acetone
• Ultrasonicate for 1 minute in isopropyl alcohol (IPA), rinse with IPA
155

Cleaning with a detergent, which must then be thoroughly rinsed off
• Ultrasonicate for 3 minutes in a 1% solution of Alconox in deionized water (DI H2 O)
• Rinse for 1 minute with DI H2 O
• Ultrasonicate for 3 minutes in DI H2 O, rinse with DI H2 O
Cleaning with solvents to remove organic contaminants [103]
• Scrub for 1 minute in acetone, scrubbing in only one direction with a cotton swab
• Ultrasonicate for 3 minutes in acetone, rinse with acetone
• Ultrasonicate for 3 minutes in IPA, rinse with IPA
• Ultrasonicate for 3 minutes in DI H2 O, rinse with DI H2 O
• Leave in DI H2 O for transportation to the cleanroom
Cleaning in the cleanroom, immediately before spinning on photoresist
• Ultrasonicate for 2 minutes in acetone, rinse with acetone
• Ultrasonicate for 2 minutes in IPA, rinse with IPA
• Ultrasonicate for 2 minutes in DI H2 O, rinse with DI H2 O
• Dry with N2 gas
• (Optional) Heat for 1 minute on a 100◦ C bakeplate to evaporate additional water

156

Ultraviolet Exposure and Photoresist Developing
The pattern which would form the electrodes was created by shining ultraviolet light through
holes in a photomask in order to create the desired pattern on the substrate. Akin to the
classic photographic developing process, photoresist must also be developed to obtain the
“image” upon which the metal electrodes will be deposited. The Shipley Microposit S1813
photoresist used for this work was a “positive” photoresist [104], meaning that the areas of
the photoresist exposed to the ultraviolet light would dissolve and wash away when exposed
to the developer (here, Shipley 352 developer). The photomask used to define the patterns
for the electrodes was a repurposed photomask from a previous project, which had a veritable
plethora of different patterns on it from which to select. For this project, it was important
for the electrodes to be continuous, since the bias voltage to the electrodes would only be
delivered at one or two locations.
As in most photolithography, this procedure required an appreciable amount of optimization. The exposure and developing steps contribute in a non-independent fashion to
the overall result. Both overexposure and overdeveloping, for example, dissolve “too much”
of the photoresist, resulting in malformed electrodes. Optimization of the procedure does
not imply that this is the best possible way to fabricate these electrodes, but rather that
electrodes made in this way were sufficiently functional. The protocol eventually settled on
involved exposing the photoresist through the mask for 8 seconds, followed by agitating it
in the developer for 30 seconds before rinsing thoroughly with DI H2 O.
The photoresist was initially spin-coated onto the chips, which were then placed on
a bakeplate for a few minutes to ensure the photoresist was sufficiently dry [104] before
ultraviolet exposure of the pattern. Following this, the mask aligner was used to press the

157

photoresist-covered chip against the mask, after which the ultraviolet light source was turned
on and shone through the holes in the photomask, exposing that portion of the photoresist.
The exposed chips were then post-baked on the hotplate. After the photoresist-covered and
exposed chips were swished around in the developer, they were immediately rinsed with
DI H2 O in order to remove residual developer. If they had not been rinsed immediately
thereafter, the remaining developer would have continued to work on the photoresist.

Thermal Evaporation
Thermal evaporation is a commonly-used technique to deposit material on surfaces. The
material, such as gold, to be deposited is placed in solid source form on a metal “boat”
which is resistively heated by having current run through it. This, in turn, heats the gold
until it melts and then evaporates. The evaporated material then sticks to the substrate
surfaces, which are suspended in a vacuum chamber above the evaporation source. Since
gold does not bond well to silicon, it is customary to include an “adhesion layer” of a few
nanometers of titanium between the silicon and the gold [103]. An example of a chip at the
microfabrication stage where gold has been evaporated over the entire surface is shown on
the left side of Figure E.3, contrasted with the post-liftoff outcome.

Liftoff
After evaporation of titanium and gold, the underlying photoresist was removed by immersing, agitating, or rinsing the chip in acetone. Turbulence from agitating the chip in the
acetone was generally sufficient to tear away, or lift off, gold which was not directly attached
to the substrate. Of the liftoff protocols ventured during this experiment, some of the best
results were obtained by soaking the chips in acetone at 90◦ C for ten minutes immediately

158

Figure E.3: Left: Photograph of a chip with electrodes patterned into photoresist and a
layer of gold evaporated over the whole chip, before liftoff. The pattern of the electrodes can
be seen as darker areas on the shiny surface. Right: Photograph of a chip after liftoff. The
electrodes are the rectangular patterns. Chips are approximately 0.5 inches wide.
after evaporation, then leaving them overnight in fresh, room-temperature acetone. The
next morning, the chips were quickly rinsed with acetone from a wash bottle in order to
remove large leaves of gold from between the electrode areas; sonicated in acetone for one
minute in fresh acetone; rinsed with a continuous stream of acetone from a wash bottle for
one minute; and rinsed with acetone, IPA, and DI H2 O before drying with N2 gas.
In principle, this procedure should result in continuous electrodes with uniform edges,
with the gold firmly affixed to the substrate. While this was not always the case, the
highest-quality electrodes thus produced were used for the CP-AFM experiments. For these
experiments, a bias voltage was delivered to the electrodes via an external wire from the
AFM. This wire attached to a magnet on the AFM sample-mounting disk to which was
affixed the completed chip with electrodes and sample. Therefore, there needed to be a continuously conducting path between the disk’s magnet and the electrodes. This was provided
via a continuous path of silver paint, adhesive with a sufficiently large concentration of silver
particles to render the paint conductive, between the magnet and a gold pad on the surface.
159

This gold pad then served as a voltage source for the electrodes. The electrodes were connected to the voltage pad with 1 mil-diameter 99.99% pure gold wire, using a West·Bond
wire bonder to make the connections. The finished product is shown in Figure E.4.

160

Figure E.4: Photograph of a sample used for the CP-AFM experiment. For scale, the disk
on which the chip is mounted is 15 mm in diameter. The cylinder at the top of the image
is the magnet to which the bias wire from the microscope was attached. Conductive silver
paint connects this to a solid gold pad which is then connected via gold wires or by direct
contact to electrodes on the chip. Residue of sample deposition can be seen on one of the
electrodes at left in the image.

161

REFERENCES

162

REFERENCES
[1] F. Qian and Y. Li. Biomaterials: A natural source of nanowires. Nature Nanotechnology, 6:538–539, 2011.
[2] C.-Y. Chiang, C. Mello, J. Gu, E. Silva, K. VanVliet, and A. Belcher. Weaving genetically engineered functionality into mechanically robust virus fibers. Advanced Materials, 19(6):826–832, 2007.
[3] G. Reguera, K. D. McCarthy, T. Mehta, J. S. Nicoll, M. T. Tuominen, and D. R. Lovley.
Extracellular electron transfer via microbial nanowires. Nature, 435:1098–1101, 2005.
[4] Y. A. Gorby, S. Yanina, J. S. McLean, K. M. Rosso, D. Moyles, A. Dohnalkova,
T. J. Beveridge, I. S. Chang, B. H. Kim, K. S. Kim, D. E. Culley, S. B. Reed, M. F.
Romine, D. A. Saffarini, E. A. Hill, L. Shi, D. A. Elias, D. W. Kennedy, G. Pinchuk,
K. Watanabe, S. Ishii, B. Logan, K. H. Nealson, and J. K. Fredrickson. Electrically
conductive bacterial nanowires produced by Shewanella oneidensis strain MR-1 and
other microorganisms. Proceedings of the National Academy of Sciences of the USA,
103(30):11358–11363, 2006.
[5] G. Wanger, Y. Gorby, M. Y. El-Naggar, T. D. Yuzvinsky, C. Schaudinn, A. Gorur,
and P. P. Sedghizadeh. Electrically conductive bacterial nanowires in bisphosphonaterelated osteonecrosis of the jaw biofilms. Oral Surgery, Oral Medicine, Oral Pathology
and Oral Radiology, 115(1):71–78, 2013.
[6] J. S. Mattick. Type IV pili and twitching motility. Annual review of Microbiology,
56:289–314, 2002.
[7] N. Guex and M. C. Peitsch. SWISS-MODEL and the Swiss-Pdb Viewer: An environment for comparative protein modeling. Electrophoresis, 18(15):2714–2723, 1997.
[8] G. T. Feliciano, A. J. R. d. Silva, G. Reguera, and E. Artacho. Molecular and electronic
structure of the peptide subunit of Geobacter sulfurreducens conductive pili from first
principles. The Journal of Physical Chemistry A, 116(30):8023–8030, 2012.
[9] D. S. Goodsell. Representing structural information with RasMol. In Current Protocols
in Bioinformatics. John Wiley & Sons, Inc., 2002.

163

[10] F. Caccavo, D. J. Lonergan, D. R. Lovley, M. Davis, J. F. Stolz, and M. J. McInerney. Geobacter sulfurreducens sp. nov., a hydrogen- and acetate-oxidizing dissimilatory
metal-reducing microorganism. Applied and Environmental Microbiology, 60(10):3752–
3759, 1994.
[11] D. R. Lovley, E. J. Phillips, Y. A. Gorby, and E. R. Landa. Microbial reduction of
uranium. Nature, 350:413–416, 1991.
[12] D. L. Cologgi, S. Lampa-Pastirk, A. M. Speers, S. D. Kelly, and G. Reguera. Extracellular reduction of uranium via Geobacter conductive pili as a protective cellular
mechanism. Proceedings of the National Academy of Sciences of the United States of
America, 108(37):15248–15252, 2011.
[13] D. R. Lovley. Cleaning up with genomics: Applying molecular biology to bioremediation. Nature Reviews Microbiology, 1:35–44, 2003.
[14] P. Sturman, P. Stewart, A. Cunningham, E. Bouwer, and J. Wolfram. Engineering
scale-up of in situ bioremediation processes: a review. Journal of Contaminant Hydrology, 19(3):171–203, 1995.
[15] D. R. Lovley. Bug juice: harvesting electricity with microorganisms. Nature Reviews
Microbiology, 4:497–508, 2006.
[16] G. Reguera, K. P. Nevin, J. S. Nicoll, S. F. Covalla, T. L. Woodard, and D. R. Lovley.
Biofilm and nanowire production leads to increased current in Geobacter sulfurreducens
fuel cells. Applied and Environmental Microbiology, 72(11):7345–7348, 2006.
[17] J. A. Gralnick and D. K. Newman. Extracellular respiration. Molecular Microbiology,
65:1–11, 2007.
[18] D. R. Lovley, J. D. Coates, E. L. Blunt-Harris, E. J. Phillips, and J. C. Woodward.
Humic substances as electron acceptors for microbial respiration. Nature, 382:445–448,
1996.
[19] K. P. Nevin and D. R. Lovley. Lack of production of electron-shuttling compounds
or solubilization of Fe(III) during reduction of insoluble Fe(III) oxide by Geobacter
metallireducens. Applied and Environmental Microbiology, 66:2248–2251, 2000.
[20] M. S. Strom and S. Lory. Structure-function and biogenesis of the type IV pili. Annual
Review of Microbiology, 47(1):565–596, 1993.

164

[21] G. Reguera, R. B. Pollina, J. S. Nicoll, and D. R. Lovley. Possible nonconductive role of
Geobacter sulfurreducens pilus nanowires in biofilm formation. Journal of Bacteriology,
189(5):2125–2127, 2007.
[22] J. P. Veazey. Scanning Probe Studies of the Pilus Nanowires in Geobacter sulfurreducens. PhD thesis, Michigan State University, 2011.
[23] J. P. Veazey, G. Reguera, and S. H. Tessmer. Electronic properties of conductive pili of
the metal-reducing bacterium Geobacter sulfurreducens probed by scanning tunneling
microscopy. Phys. Rev. E, 84:060901, 2011.
[24] M. Y. El-Naggar, G. Wanger, K. M. Leung, T. D. Yuzvinsky, G. Southam, J. Yang,
W. M. Lau, K. H. Nealson, and Y. A. Gorby. Electrical transport along bacterial
nanowires from Shewanella oneidensis MR-1. Proceedings of the National Academy of
Sciences of the United States of America, 107(42):18127–18131, 2010.
[25] S. Lampa-Pastirk, J. P. Veazey, R. Steidl, K. A. Walsh, S. Tessmer, and G. Reguera.
Long-range electron transport via microbial protein nanowires. in preparation, 2013.
[26] L. Craig, M. E. Pique, and J. A. Tainer. Type IV pilus structure and bacterial
pathogenicity. Nature Reviews Microbiology, 2:363–378, 2004.
[27] J. J. Davis. Molecular bioelectronics. Philosophical Transactions of the Royal Society
A, 361:2807–2825, 2003.
[28] J. W. G. Wild¨er, L. C. Venema, A. G. Rinzler, R. E. Smalley, and C. Dekker. Eleco
tronic structure of atomically resolved carbon nanotubes. Nature, 391:59–62, 1998.
[29] T. W. Odom, J.-L. Huang, P. Kim, and C. M. Lieber. Atomic structure and electronic
properties of single-walled carbon nanotubes. Nature, 391:62–64, 1998.
[30] C. J. Chen. Introduction to Scanning Tunneling Microscopy. Monographs on the
Physics and Chemistry of Materials. Oxford University Press, New York, 2nd edition,
2008.
[31] A. F. Raigoza, J. W. Dugger, and L. J. Webb. Review: Recent advances and current
challenges in scanning probe microscopy of biomolecular surfaces and interfaces. ACS
Applied Materials & Interfaces preprint online, 2013.
[32] N. S. Malvankar, M. Vargas, K. P. Nevin, A. E. Franks, C. Leang, B.-C. Kim, K. Inoue,
T. Mester, S. F. Covalla, J. P. Johnson, V. M. Rotello, M. T. Tuominen, and D. R.
165

Lovley. Tunable metallic-like conductivity in microbial nanowire networks. Nature
Nanotechnology, 6:573–579, 2011.
[33] S.-i. Park, J. Nogami, and C. F. Quate. Effect of tip morphology on images obtained
by scanning tunneling microscopy. Phys. Rev. B, 36:2863–2866, 1987.
[34] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel. Surface studies by scanning tunneling
microscopy. Phys. Rev. Lett., 49:57–61, 1982.
[35] R. M. Feenstra. Scanning tunneling spectroscopy. Surface Science, 299–300(0):965–
979, 1994.
[36] G. Binnig, H. Rohrer, C. Gerber, and E. Weibel. Tunneling through a controllable
vacuum gap. Applied Physics Letters, 40(2):178–180, 1982.
[37] G. Binnig and H. Rohrer. Scanning tunneling microscopy–from birth to adolescence.
Rev. Mod. Phys., 59:615–625, 1987.
[38] R. J. Hamers, R. M. Tromp, and J. E. Demuth. Surface electronic structure of Si
(111)-(7×7) resolved in real space. Phys. Rev. Lett., 56:1972–1975, 1986.
[39] T. Arai, K. Ichimura, K. Nomura, S. Takasaki, J. Yamada, S. Nakatsuji, and H. Anzai.
Superconducting and normal-state gaps in κ-(BEDT-TTF)2 Cu(NCS)2 studied by STM
spectroscopy. Solid State Communications, 116(12):679–682, 2000.
[40] K. Jordan, A. Cazacu, G. Manai, S. Ceballos, S. Murphy, and I. Shvets. Scanning
tunneling spectroscopy study of the electronic structure of Fe3 O4 surfaces. Physical
Review B, 74:085416–1–6, 2006.
[41] G. Binnig, C. F. Quate, and C. Gerber. Atomic force microscope. Phys. Rev. Lett.,
56:930–933, 1986.
[42] R. Proksch, T. E. Sch¨ffer, J. P. Cleveland, R. C. Callahan, and M. B. Viani. Fia
nite optical spot size and position corrections in thermal spring constant calibration.
Nanotechnology, 15(9):1344, 2004.
[43] K. Besocke. An easily operable scanning tunneling microscope. Surface Science,
181:145–153, 1987.

166

[44] J. Frohn, J. Wolf, K. Besocke, and M. Teske. Coarse tip distance adjustment and positioner for a scanning tunneling microscope. Review of Scientific Instruments, 60:1200–
1201, 1989.
[45] K. A. Walsh, M. E. Romanowich, M. Gasseller, I. Kuljanishvili, R. Ashoori, and S. Tessmer. Scanning-probe single-electron capacitance spectroscopy. Journal of Visualized
Experiments, 77:e50676, 2013.
[46] W.-T. Pong and C. Durkan. A review and outlook for an anomaly of scanning tunnelling microscopy (STM): superlattices on graphite. Journal of Physics D: Applied
Physics, 38(21):R329, 2005.
[47] Y. Baskin and L. Meyer. Lattice constants of graphite at low temperatures. Physical
Review, 100(2), 1955.
[48] P. Kes, C. v. d. Klein, and D. d. Klerk. A new R–T relation for Allen-Bradley carbon
resistor thermometers. Cryogenics, 14(3):168–169, 1974.
[49] W. N. Lawless, S. K. Hampton, and C. F. Clark. Calibration scheme for Allen–Bradley
resistors, 4.2–100 k. Review of Scientific Instruments, 59(11):2505–2507, 1988.
[50] J. Freund, J. Halbritter, and J. H¨rber. How dry are dried samples? Water adsorption
o
measured by STM. Microscopy Research and Technique, 44(5):327–338, 1999.
[51] L. Craig, N. Volkmann, A. S. Arvai, M. E. Pique, M. Yeager, E. H. Egelman, and J. A.
Tainer. Type {IV} pilus structure by cryo-electron microscopy and crystallography:
Implications for pilus assembly and functions. Molecular Cell, 23(5):651 – 662, 2006.
[52] P. N. Reardon and K. T. Mueller. Structure of the type IVa major pilin from the
electrically conductive bacterial nanowires of Geobacter sulfurreducens. Journal of
Biological Chemistry preprint online, 2013.
[53] P. Reardon and K. Mueller. Structure of the type IVa major pilin from the electrically
conductive bacterial nanowires of Geobacter sulfurreducens.
[54] E. Artacho, E. Anglada, O. Di´guez, J. D. Gale, A. Garc´ J. Junquera, R. M. Martin,
e
ıa,
P. Ordej´n, J. M. Pruneda, D. S´nchez-Portal, and J. M. Soler. The SIESTA method;
o
a
developments and applicability. Journal of Physics: Condensed Matter, 20(6):064208,
2008.
[55] J. M. B. Jr. Electron relay in proteins. Science, 320:1730–1731, 2008.
167

[56] R. I. Cukier and D. G. Nocera. Proton-coupled electron transfer. Annual Review of
Physical Chemistry, 49(1):337–369, 1998.
[57] H. Gray and J. Winkler. Electron flow through proteins. Chemical Physics Letters,
483:1–9, 2009.
[58] L. K. Skov, T. Pascher, J. R. Winkler, and H. B. Gray. Rates of intramolecular electron
transfer in Ru(bpy)2(im)(His83)-modified azurin increase below 220 K. Journal of the
American Chemical Society, 120(5):1102–1103, 1998.
[59] N. F. Polizzi, S. S. Skourtis, and D. N. Beratan. Physical constraints on charge transport through bacterial nanowires. Faraday Discussions, 155:43–62, 2011.
[60] T. Kakitani and N. Mataga. Effects of freezing out protein vibrational modes on
electron transfer kinetics in bacterial reaction centers. Photosynthesis Research, 22:187–
193, 1989.
[61] J.-J. Kim, I. Ekvall, and H. Olin. Temperature-dependent scanning tunneling spectroscopy of 1T -TaS2 . Physical Review B, 54(3):2244–2248, 1996.
[62] K. Ichimura, M. Kobayashi, K. Nomura, T. Hanajiri, J. Yamada, S. Nakatsuji, and
H. Anzai. SDW gap in (TMTSF)2 PF6 : STM spectroscopy. Synthetic Metals, 120(1–
3):897–898, 2001. Proceedings of the International Conference on the Science and
Technology of Synthetic Metals.
[63] G. Travaglini, H. Rohrer, M. Amrein, and H. Gross. Scanning tunneling microscopy
on biological matter. Surface Science, 181(1–2):380–390, 1987.
[64] A. Szent-Gy¨rgyi. Towards a new biochemistry? Science, 93(2426):609–611, 1941.
o
[65] J. Veazey, S. Lampa-Pastirk, G. Feliciano, G. Reguera, E. Artacho, and S. Tessmer.
Evidence of rectification in the local electronic structure of microbial nanowires. to be
submitted, 2013.
[66] W. Humphrey, A. Dalke, and K. Schulten. VMD: Visual molecular dynamics. Journal
of Molecular Graphics, 14(1):33 –38, 1996.
[67] B. Hess, C. Kutzner, D. v. d. Spoel, and E. Lindahl. GROMACS 4: Algorithms for
highly efficient, load-balanced, and scalable molecular simulation. Journal of Chemical
Theory and Computation, 4:435–447, 2008.

168

[68] K. Capelle. A bird’s-eye view of density-functional theory. arXiv:cond-mat/0211443v5
1–35.
[69] J. M. Soler, E. Artacho, J. D. Gale, A. Garc´ J. Junquera, P. Ordej´n, and D. Snchezıa,
o
Portal. The SIESTA method for ab initio order-N materials simulation. Journal of
Physics: Condensed Matter, 14(11):2745–2779, 2002.
[70] J. P. Perdew, K. Burke, and M. Ernzerhof. Generalized gradient approximation made
simple. Physical Review Letters, 77:3865–3868, 1996.
[71] R. K. Clayton. Effects of dehydration on reaction centers from Rhodopseudomonas
sphaeroides. Biochimica et Biophysica Acta (BBA)–Bioenergetics, 504(2):255–264,
1978.
[72] R. Guckenberger, M. Heim, G. Cevc, H. F. Knapp, W. Wiegr¨be, and A. Hillebrand.
a
Scanning tunneling microscopy of insulators and biological specimens based on lateral
conductivity of ultrathin water films. Science, 266:1538–1540, 1994.
[73] D. Martinez-Martin, R. Longuinhos, J. G. Izquierdo, A. Marele, S. S. Alexandre,
M. Jaafar, J. M. G´mez-Rodr´
o
ıguez, L. Ba˜ares, J. M. Soler, and J. Gomez-Herrero.
n
Atmospheric contaminants on graphitic surfaces. Carbon, 61:33–39, 2013.
[74] R. Proksch. Multifrequency, repulsive-mode amplitude-modulated atomic force microscopy. Applied Physics Letters, 89(11):113121, 2006.
[75] D. J. M¨ller, M. Amrein, and A. Engel. Adsorption of biological molecules to a solid
u
support for scanning probe microscopy. Journal of Structural Biology, 119:172–188,
1997.
[76] S.-F. Shi, K. I. Bolotin, F. Kuemmeth, and D. C. Ralph. Temperature dependence
of anisotropic magnetoresistance and atomic rearrangements in ferromagnetic metal
break junctions. Physical Review B, 76:184438, 2007.
[77] R. Mahadevan, D. R. Bond, J. E. Butler, A. Esteve-Nu˜ez, M. V. Coppi, B. O. Palsson,
n
C. H. Schilling, and D. R. Lovley. Characterization of metabolism in the Fe(III)reducing organism Geobacter sulfurreducens by constraint-based modeling. Applied
and Environmental Microbiology, 72(2):1558–1568, 2006.
[78] B. Rosenberg. Electrical conductivity of proteins. II. Semiconduction in crystalline
bovine hemoglobin. The Journal of Chemical Physics, 36(3), 1962.

169

[79] J. J. Davis and H. A. O. Hill. The scanning probe microscopy of metalloproteins and
metalloenzymes. Chemical Communications 393–401, 2002.
[80] N. S. Wigginton, K. M. Rosso, B. H. Lower, L. Shi, and M. F. H. Jr. Electron tunneling properties of outer-membrane decaheme cytochromes from Shewanella oneidensis.
Geochimica et Cosmochimica Acta, 71(3):543–555, 2007.
[81] K. S. Novoselov, A. K. Geim, S. V. Morozov, D. Jiang, Y. Zhang, S. V. Dubonos,
I. V. Grigorieva, and A. A. Firsov. Electric field effect in atomically thin carbon films.
Science, 306(5696):666–669, 2004.
[82] K. Walsh,
S. MacLaren,
et al.
Engage panel tip height.
https://support.asylumresearch.com/forum/showthread.php?1546-Engage-paneltip-height, February 2013.
[83] N. Leonard, V. Nallathambi, and S. Calabrese Barton. Carbon supports for nonprecious metal oxygen reducing catalysts. Journal of the Electrochemical Society,
160(8):F788–F792, 2013.
[84] N. D. Leonard and S. Calabrese Barton. Basicity of non-precious metal catalysts for
oxygen reduction. Meeting Abstracts, MA2012-02(13):1698, 2012.
[85] K.
Walsh
et
al.
Ears
in
dual-gain
ORCA
IV
curves.
https://support.asylumresearch.com/forum/showthread.php?1124-Ears-in-dualgain-ORCA-IV-curves, November 2011.
[86] C. Ware. Chapter 5–Visual salience and finding information. In Information Visualization: Perception for Design. Morgan Kaufmann, 2013.
[87] C. Ware. Chapter 6–Static and moving patterns. In Information Visualization: Perception for Design. Morgan Kaufmann, 2013.
[88] C. Ware. Chapter 11–Visual thinking processes. In Information Visualization: Perception for Design. Morgan Kaufmann, 2013.
[89] C. R. Clemmer and J. Beebe, Thomas P. Graphite: A mimic for DNA and other
biomolecules in scanning tunneling microscope studies. Science, 251(4994):640–642,
1991.
[90] G. Binnig and H. Rohrer. Scanning tunneling microscopy. In J. Janta and
J. Pantofliˇek, editors, Trends in Physics 1984–Proceedings of the 6th General Conferc
ence of the European Physical Society 38–45. European Physical Society, 1984.
170

[91] Y. Niimi, T. Matsui, H. Kambara, K. Tagami, M. Tsukada, and H. Fukuyama. Scanning tunneling microscopy and spectroscopy of the electronic local density of states of
graphite surfaces near monoatomic step edges. Phys. Rev. B, 73:085421, 2006.
[92] P. Simonis, C. Goffaux, P. Thiry, L. Biro, P. Lambin, and V. Meunier. STM study of
a grain boundary in graphite. Surface Science, 511:319–322, 2002.
[93] H. S. Wong, C. Durkan, and N. Chandrasekhar. Tailoring the local interaction between
graphene layers in graphite at the atomic scale and above using scanning tunneling
microscopy. ACS Nano, 3(11):3455–3462, 2009.
[94] R. Wilkins, M. Amman, E. Ben-Jacob, and R. C. Jaklevic. Single-electron and oxideimpurity effects in junctions formed by a cryogenic scanning tunneling microscope.
Phys. Rev. B, 42:8698–8701, 1990.
[95] C. Sch¨nenberger, H. v. Houten, and H. C. Donkersloot. Single-electron tunnelling
o
observed at room temperature by scanning-tunnelling microscopy. EPL (Europhysics
Letters), 20(3):249, 1992.
[96] J. P. Pelz. Tip-related electronic artifacts in scanning tunneling spectroscopy. Phys.
Rev. B, 43:6746–6749, 1991.
[97] S. Lindsay, N. Tao, J. DeRose, P. Oden, L. Yu.L., R. Harrington, and L. Shlyakhtenko.
Potentiostatic deposition of DNA for scanning probe microscopy. Biophysical Journal,
61(6):1570–1584, 1992.
[98] X. Xi, J. Shi, S. Maghsoudy-Louyeh, and B. R. Tittmann. Characterization of silicon wafer surfaces after hydrophilic and hydrophobic treatments by atomic force microscopy. AIP Conference Proceedings, 1211(1):1493–1498, 2010.
[99] F. Zhang, H.-N. Du, Z.-X. Zhang, L.-N. Ji, H.-T. Li, L. Tang, H.-B. Wang, C.-H. Fan,
H.-J. Xu, Y. Zhang, J. Hu, H.-Y. Hu, and J.-H. He. Epitaxial growth of peptide nanofilaments on inorganic surfaces: Effects of interfacial hydrophobicity/hydrophilicity.
Angewandte Chemie International Edition, 45(22):3611–3613, 2006.
[100] A. Savitzky and M. J. E. Golay. Smoothing and differentiation of data by simplified
least squares procedures. Analytical Chemistry, 36(8):1627–1639, 1964.
[101] S. Magonov, V. Elings, and M.-H. Whangbo. Phase imaging and stiffness in tappingmode atomic force microscopy. Surface Science, 375(2–3):L385–L391, 1997.

171

[102] I. Schmitz, M. Schreiner, G. Friedbacher, and M. Grasserbauer. Phase imaging as
an extension to tapping mode AFM for the identification of material properties on
humidity-sensitive surfaces. Applied Surface Science, 115(2):190–198, 1997.
[103] K. R. Williams, K. Gupta, and M. Wasilik. Etch rates for micromachining processing—
Part II. Journal of Microelectromechanical Systems, 12(6):761–778, 2003.
[104] A. M. Collins. Chapter 5—Physical Techniques. In Nanotechnology Cookbook 205–253.
Elsevier, Oxford, 2012.

172