Are you sure?
This action might not be possible to undo. Are you sure you want to continue?
Analysis and preliminary characterization
of a MEMS cantilevertype
chemical sensor
by
Daniel Arecco
NEST – NanoEngineering, Science, and Technology
CHSLT – Center for Holographic Studies and Laser micromechaTronics
Mechanical Engineering Department
Worcester Polytechnic Institute
Worcester, MA 016092280
17 December 2003
3
Copyright © 2003
by
NEST – NanoEngineering, Science, and Technology
CHSLT – Center for Holographic Studies and Laser micromechaTronics
Mechanical Engineering Department
Worcester Polytechnic Institute
Worcester, MA 016092280
All rights reserved
4
SUMMARY
This Thesis relates to the continually advancing field of microelectromechanical
systems (MEMS). With MEMS technology, there are many different areas of
concentration available for research. This Thesis addresses analysis and preliminary
characterization of a cantilevertype MEMS chemical sensor for detection of chemicals
and organic components operating at room temperature (20˚C and sea level pressure of 1
atm). Such sensors can be useful in a wide variety of applications.
There currently exist several different types of MEMS chemical sensors. Each is
based on a different detection method, e.g., capacitive, thermal, resistive, etc., and is used
for specific tasks. Out of all currently available detection methods, the most common is
the gravimetric method. The gravimetric sensor works by absorbing the chemical in a
special material, usually a polymer, which alters the overall mass of the sensing element
that can then be measured, or detected, to identify the chemical absorbed.
One of the more exciting developments in the field of gravimetric chemical
MEMS has been with the advancement of cantilevertype sensors. These cantilevers are
small and usually on the order of only about 300 µm in length. In order to utilize the
gravimetric method, a cantilever is coated with a polymer that allows an analyte to bond
to it and change its mass, which in turn changes the resonant frequency of the cantilever.
The change in frequency can then be measured and analyzed and from it, the amount of
absorbed mass can be calculated. Current research in the cantilevertype resonating
sensors for the detection of hydrogen is developing measurement capabilities of 1 ppm
(partpermillion).
5
In this Thesis number of sample cantilevers were qualitatively assessed and their
dimensional geometry measured. Based on these measurements, frequency data were
obtained. In addition, the overall uncertainty in the resonant frequency results was
calculated and the contributing factors to this uncertainty were investigated.
Experimental methods that include laser vibrometry, optoelectronic laser interferometric
microscopy (OELIM), and atomic force microscopy (AFM), were utilized to measure the
frequency responses of the samples. The analytically predicted natural frequencies were
compared to the experimental data to determine correlation subject to the uncertainty
analysis.
Parametric analyses involving chemical absorption processes were also
conducted. Such analyses considered different parameters, e.g., damping and stiffness as
well as changes in their values, to determine contributions they make to the quality of the
frequency data and the effect they have on sensitivity of the MEMS cantilevertype
chemical sensors. Once these parametric analyses were completed, it was possible to
estimate the sensitivity of the cantilever, or the ability for the cantilever to detect
frequency shifts due to absorption of the target chemical. Results of the parametric
analyses of the fundamental resonant frequency were then correlated with the sensitivity
results based on the chemical absorption.
This Thesis correlates many results and ideas and probes problems revolving
around the analysis and characterization of a MEMS cantilevertype chemical sensor.
6
ACKNOWLEDGEMENTS
I would first like to thank my parents, Carlos Arecco M.D. and Vera Arecco
Ph.D. for everything that they have provided for me in my life. Without their undying
love and support, I would not be where I am today.
First off, I have to thank Professor Ryszard J. Pryputniewicz for the wonderful
opportunity he has given me to work in a research field that has challenged me to new
limits. He has taught me things beyond the call of a typical advisor. Thank you.
Secondly, I have to send a special thank you to Mr. (soon to be Dr.) Shivananda
Pai Mizar for all the invaluable support that he showed whenever I was in need and for
his affirming to my getting done. May your teaching career be rewarding. Another
special thank must go to Mr. Peter Hefti from Switzerland, who lent me his wealth of
technical knowledge and support for completing my experimental work and also for his
stories that would always boost morale. May they never catch you. I would also like to
thank Professor Cosme Furlong for his support in my computational work and all the
random questions I had. I would also like to thank everyone else at the Center for
Holographic Studies and Laser micromechaTroncis (CHSLT): Kevin Bruff, Wei Han,
Adam Klempner, Ronald Kok, Ryan Marinis, and Vicky Steward. I would also like to
acknowledge all the people that helped me along the way in there own ways: Barbara
Edilberti, Janice Dresser, and Todd Billings. Finally, I would like to thank my brother,
Andres, because I can.
This M.S. Thesis will stay with me for the rest of my life and I am glad to have
accomplished it with all the great people of Worcester Polytechnic Institute.
7
TABLE OF CONTENTS
Copyright 3
Summary 4
Acknowledgements 6
Table of contents 7
List of figures 11
List of tables 16
Nomenclature 18
1. Objectives 21
2. Introduction 22
2.1. Micromachine beginnings 22
2.2. MEMS Foreword 24
2.3. Sensors 27
2.3.1. Mechanical detection 29
2.3.2. Optical detection 31
2.3.3. Electrical detection 32
2.3.4. Chemical detection 32
2.4. MEMS cantilever chemical sensor 35
2.4.1. Material 35
2.4.2. Fabrication 37
2.4.3. Packaging 39
2.4.4. Sensing types and methods 40
2.4.5. Resonance frequency 42
2.4.6. Cantilever actuation methods 43
2.4.7. Cantilever I/O 44
2.4.8. Polymer thickness 46
2.4.10. Fick’s law of diffusion 47
2.4.11. Reference cantilevers 47
2.4.12. Quality factor 48
2.4.13. Electrical aspects 49
2.4.13.1. Wheatstone bridge 50
2.4.13.2. Feed back Loop 50
2.4.14. Inherent Problems 51
2.4.14.1. Residual Stresses 51
2.4.14.2. Stiction 53
8
2.4.15. Effects of temperature 54
2.4.16. Effects of pressure 54
3. MEMS samples 56
3.1. MikroMasch USA cantilevers 56
3.2. Chemicals used 58
3.2.1. Palladium 59
3.2.2. Hydrogen 60
3.2.3. Titanium 60
3.3. Coating process 61
3.3.1. Complete versus partial coating 62
3.3.2. Effects of humidity 63
3.3.3. Time limits 63
4. Methodology 65
4.1. ACES methodology 65
4.2. Analytical investigations 67
4.2.1. Sensitivity 68
4.2.2. Equivalent variables 71
4.2.3. Gas concentration 71
4.2.4. Absorption 72
4.2.4.1. Diffusion rate 75
4.2.4.2. Volume increase 78
4.2.4.3. Change in modulus of elasticity 79
4.2.5. Resonant frequency and frequency shift 81
4.2.6. Spring constant consideration 83
4.2.7. Damping coefficient consideration 85
4.2.8. Free vibration system 87
4.2.9. Changing stiffness in a free vibration system 89
4.2.10. Harmonically excited system 89
4.2.11. Uncertainty analysis 92
4.3. Computational analysis 93
4.4. Experimental solutions 94
4.4.1. Instrument calibration 95
4.4.2. Geometrical characterization of the cantilevers 95
4.4.3. SEM – scanning election microscope 96
4.4.4. Determination of quality factor and damping coefficient 97
4.4.5. Optoelectronic methodology 98
4.4.6. Determination of resonant frequency by AFM 104
4.4.7. Laser vibrometry 105
4.4.8. Test chamber 107
4.5. Data analysis 109
9
5. Results 111
5.1. Microscope and SEM data 111
5.2. Analytical results 116
5.3. Optoelectronic measurements 121
5.4. AFM results 123
5.5. Computational results 124
5.6. Comparison of results 127
5.7. Parameters of the coated cantilevers 129
5.8. Sensitivity solutions 130
5.9. Absorption 133
5.10. Determination of the damping coefficient 136
5.11. Complete cantilever analysis 139
5.11.1. Case 1: changing mass 140
5.11.2. Case 2: changing mass and stiffness 141
5.11.3. Case 3: changing mass with damping 142
5.11.4. Case 4: changing mass and stiffness with damping 142
5.11.5. Design equation selection 143
5.11.6. Lower limit frequency shift 144
5.11.7. Optimization of the cantilever response 146
6. Conclusions 148
7. References 151
Appendix A. Material properties 161
Appendix B. Analytical calculations 164
B.1. Fundamental resonance frequency 164
B.2. Spring constant 165
B.3. Dynamic mass 168
B.3.1. Dynamic mass constant 170
B.4. Sensitivity calculations of the cantilever 171
B.4.1. Sensitivity of an endloaded cantilever 171
B.4.2. Sensitivity of a completely coated cantilever 173
B.5. Frequency shift for a free, undamped vibration 174
B.6. Frequency shift for a free, undamped vibration with a
changing stiffness 176
B.7. Free vibration with viscous damping 177
A.7.1. Frequency shift for a free, damped vibration 178
A.7.2. Frequency shift for a free, damped vibration
with a changing stiffness 180
A.8. Harmonically excited system with viscous damping 181
10
Appendix C. Uncertainty analysis 183
C.1. Uncertainty in fundamental resonant frequency 183
C.1.1. Uncertainty calculations for the 350 µm long
cantilever 183
C.1.2. Uncertainty calculations for the 300 µm long
cantilever 187
C.1.3. Uncertainty calculations for the 250 µm long
cantilever 189
C.1.4. Average overall percentage uncertainty in frequency 190
C.2. Frequency uncertainty in a palladium coated cantilever 191
C.2.1. Uncertainty in frequency for a free vibrations system 191
C.2.2. Uncertainty in frequency for a free vibrations system
with a changing stiffness 199
C.2.3. Uncertainty in frequency for a free vibrations system
with damping 205
C.2.4. Uncertainty in frequency for a damped, changing
stiffness vibrations system 210
Appendix D. Test chamber 215
Appendix E. Analytical frequency data 218
11
LIST OF FIGURES
Fig. 2.1. A dust mite dwarfs a MEMS device (SNL, 2001) 23
Fig. 2.2. A human is compared to a MEMS device (SNL, 2001) 23
Fig. 2.3. A microchain device (SNL, 2001) 25
Fig. 2.4. Details of a DMD chip (TI, 2003) 26
Fig. 2.5. A schematic illustration of MEMS components and their
interdependence (Madou, 2002) 28
Fig. 2.6. A view of an entire dualaxis accelerometer where the proof mass can
be seen with all the perforations in the middle and a quarterview close
up on one of its corners (AD, 2003) 30
Fig. 2.7. Polymer based FabryPerot sensor (Yoo et al., 1997) 33
Fig. 2.8. Micron sized cantilevers developed at IBM for research purposes
(IBM, 2002) 34
Fig. 2.9. Initial steps in fabrication of MEMS and CMOS chips (MEMC,
Missouri, 2003) 36
Fig. 2.10. Examples of SUMMiT™V process (SNL, 2002) 38
Fig. 2.11. MEMS chip and its different levels of packaging (SNL, 2001) 40
Fig. 2.12. Operation of different gravimetric sensors (Ivanov, 2000) 42
Fig. 2.13. Comparison of different sized and shaped PZTs available (PI, 2003) 44
Fig. 2.14. Graphical illustration of the operational principle for an AFM where
different bend angles reflect the laser beam to different locations on
the PSD 45
Fig. 2.15. Cantilever with integrated piezoresistors (Gotszalk et al., 2000) 46
Fig. 2.16. Wheatstone bridge where changes, e.g., in R
1
, alter the output voltage
with a high accuracy and precision 50
Fig. 2.17. Effects of residual stresses (Dartmouth, 2002) 52
12
Fig. 2.18. The collapse of a thin cantilever due to stiction 53
Fig. 2.19. A femtocalorimeter utilizing thermal bending effects with
piezoresistors (Gotszalk et al., 2000) 55
Fig. 3.1. MikroMasch AFM chip (MM, 2003) 57
Fig. 3.2. Photographs of the AFM chips used in this Thesis 58
Fig. 3.3. Bending moment comparison 62
Fig. 4.1. Graphical representation of ACES methodology and the
interdependence of each system towards determining a final solution 66
Fig. 4.2. Idealization and modeling of a cantilever 69
Fig. 4.3. Concentration and pressure relationships based on experimental data
of many research studies conducted throughout the years, where the
concentration, AR, is in units of atomic ratio (H/Pd) (Lewis, 1967) 74
Fig. 4.4. Permeation rates of hydrogen with circulating or static flows through a
membrane of Pd with a high pressure on one side (Lewis, 1967) 76
Fig. 4.5. Experimental data fitted with a line to approximate the amount of
volumetric change in palladium due to hydrogen absorption (G.
Alefeld and J. Völkl, 1978) 79
Fig. 4.6. Estimated results from experimental data illustrating the change that
occurs in the modulus of elasticity during absorption of hydrogen into
palladium (Lewis, 1967) 80
Fig. 4.7. Representation of cantilever with geometric dimensions and a cross
section view through the middle with each of the layers of the
cantilever labeled 82
Fig. 4.8. Vibration amplitude as a function of frequency ratio for different
values of damping ratio 91
Fig. 4.9. SEM used for measuring thickness of the cantilevers. The samples are
inserted into the chamber on the left (botton of the black tube) and the
microscope is controlled with the console on the right 96
Fig. 4.10. A comparison of how different resonant frequency response curves
affect the width of the bandwidth, which in turn will affect Eq. 4.42 98
13
Fig. 4.11. Singleillumination and singleobservation geometry of a fiber optic
based OEH system (Furlong and Pryputniewicz, 2000) 100
Fig. 4.12. Optical configuration of the OELIM system (Furlong and
Pryputniewicz, 2000) 104
Fig. 4.13. AFM setup utilized 105
Fig. 4.14. A comparison of the PZT used with a MEMS chip attached to the top
and a dime 106
Fig. 4.15. The custom made test chamber containing the PZT transducer, which
is to be used for testing functional operation of the cantileversensing
element in a gaseous environment. 108
Fig. 4.16. Laboratory setup for testing functional operation of MEMS chemical
sensors; the microscope and test chamber will be contained within a
vented hood 108
Fig. 5.1. The microscope setup used in determining the dimensions of the
cantilevers 112
Fig. 5.2. Cantilever details and labelling 113
Fig. 5.3. Cantilevers packed in a GelPack as obtained from the manufacturer 114
Fig. 5.4. SEM photographs of the cantilevers, with a progressive close up on the
tip 115
Fig. 5.5. SEM photographs of E length cantilevers near the tip from different
chips 115
Fig. 5.6. Fundamental resonant frequency of a 1µm thick cantilever as a
function of active length 118
Fig. 5.7. The percentage overall uncertainty in the cantilever is plotted against
the length, while holding the thicknes parameter to 1 µm to show the
decreasing trend 119
Fig. 5.8. The percentage overall uncertainty in the cantilever is plotted against
the thickness, while holding the length parameter to 300 µm to show
the decreasing trend 119
14
Fig. 5.9. Percent contributions by uncertainty of each independent parameter to
the overall uncertainty in resonance frequency of the cantilever, as a
function of length 120
Fig. 5.10. Percent contributions by uncertainty by each independent parameter to
the overall uncertainty in resonance frequency of the cantilever, as a
function of thickness 120
Fig. 5.11. Overall uncertainty in the frequency of the cantilever as a function of
the uncertainty in the thickness 121
Fig. 5.12. OELIM images of cantilever E32 vibrating under resonant conditions 122
Fig. 5.13. OELIM images of the cantilevers (D, E, and F) of chip 22 vibrating at
their fundamental mode of vibration 122
Fig. 5.14. A screen capture of the AFM frequency scan for the cantilever F12,
where a resonant frequency peak is identified as 22,160 Hz with the
accuracy of 10 Hz 123
Fig. 5.15. Pro/Engineer model and Pro/Mechanica results of the three cantilevers
considered 126
Fig. 5.16. Effective mass versus percentage of coating coverage 131
Fig. 5.17. Concentration as a function of time for a cantilever coated with 0.1 µm
thick layer of palladium 136
Fig. 5.18. A digitally enhanced image of a screen capture from the HP Spectrum
analyzer taken while measuring the frequencies of a cantilever 138
Fig. 5.19. The frequency shift in the resonance of a cantilever as a function of
concentration of the hydrogen 146
Fig. A.1. A photograph of palladium samples (EC, 2003) 162
Fig. A.2. A sample of titanium, a popular material due to its corrosion resistance
(EC, 2003) 163
Fig. A.3. A sample of untreated, unprocessed, uncut silicon (EC, 2003) 163
Fig. B.1. Geometric parameters of a cantilever 166
Fig. D.1. Specifications of the main body of the test chamber 215
15
Fig. D.2. Specifications of the lid for the main body of the test chamber 216
Fig. D.3. Fully disassembled view of the chamber showing all parts 216
Fig. D.4. Fully assembled view of chamber with PZT inside 217
Fig. D.5. Isometric view of the assembled chamber 217
Fig. E.1. Schematic of the cantilever and the dimensions used for the data
analysis 218
16
LIST OF TABLES
Table 3.1. Characteristic dimensions, as supplied by the manufacturer,
MikroMasch USA, and the corresponding material properties as listed
by Madou (2002) 57
Table 4.1. Concentrations of hydrogen in Pd based on experimental data of Fig.
4.3 at 20ºC with conversion information (Lewis, 1967) 75
Table 5.1. Measured dimensional data and equivalent lengths for all cantilever
samples of the D type 115
Table 5.2. Summary of measured dimensions for the cantilevers of the D, E, and
F types 116
Table 5.3. Values of material properties of silicon (Madou, 2002) 117
Table 5.4. The uncertainties in each parameter considered 117
Table 5.5. Comparison of the analytical and computational resonance frequency
results 125
Table 5.6. Data comparison of analytical, OELIM, and AFM generated
frequencies for each cantilever 128
Table 5.7. Properties of materials used 130
Table 5.8. Frequency shift per Hertzkilogram for distributed and tip loads 133
Table 5.9. Summary of analytically calculated and experimentally determined
damping coefficients 139
Table 5.10. The results of the four different cases investigated are compared at
1% hydrogen concentration 144
Table A.1. Physical properties of the materials used 162
Table B.1. Material properties and geometric dimensions used in a sample case to
determine sensitivity of the cantilever as a chemical sensor 172
Table C.1. Values of parameters used in uncertainty analysis 184
Table C.2. The uncertainties in each parameter considered 185
17
Table C.3. Material properties and thickness of the composite cantilever 191
Table C.4. The uncertainty values used based on current industry practice 197
Table C.5. Values of the two additional independent parameters in the damped
case 207
Table E.1. Material properties of silicon used for the analytical calculations 218
Table E.2. Geometrical and resonant frequency data of the cantilever utilized 219
18
NOMENCLATURE
b width of the cantilever
b
1
width of the end of cantilever
b
2
width of point of cantilever
b
t
total width of cantilever
c concentration of gas, damping coefficient
c
c
critical damping coefficient
c
o
initial concentration of gas
c
gas
concentration of gas at the surface
erf Gaussian error function
f resonant frequency
f
i
initial resonant frequency
f
1
initial resonant frequency
f
2
final resonant frequency
f frequency
h thickness of the cantilever
h
t
total thickness of the cantilever
h
Ti
thickness of the titanium layer
h
Pd
thickness of the palladium layer
h
∆
change of thickness constant
k spring constant
m effective, or dynamic, mass
n effective mass constant
n
r
effective mass constant for a cantilever coverage at value r
n(x) effective mass constant in terms or x
p pressure
ppm partspermillion
r frequency ratio, percent coverage of cantilever with polymer
s Sievert’s constant, solution to differential equation
s
1,2
solution to differential equation with two roots
t time
v velocity
v
max
maximum velocity of a body
w
o
distributed load
x position along the cantilever
y displacement
y(x) displacement in terms of x
y
max
maximum displacement
z position across the cantilever
A amplitude
A area of shaped tip of cantilevers
C
1,2,3,4
constants
19
D diffusion rate
E modulus of elasticity
E
eq
equivalent modulus of elasticity
E
Ti
modulus of elasticity of titanium
E
Pd
modulus of elasticity of palladium
E
Si
modulus of elasticity of silicon
F
xmax
maximum force acting on a body
F force
F(y,t) time dependent forcing function
FEM finite element method
H thickness of damping fluid
I moment of inertia
K sensitivity vector
K
1
direction of illumination vector
K
2
direction of observation vector
KE kinetic energy of a system
L displacements vector
L length of the cantilever
L
t
total length of the cantilever
L
1
length of shaped end of cantilever
L
2
equivalent length of the shaped end of cantilever
L
3
length of side point of cantilever
M mass
M
x
bending moment along the cantilever
MEMS microelectromechanical systems
PZT piezoresistive transducer
S
e
sensitivity for an end loaded cantilever
S
d
sensitivity for a distributed load cantilever
S
m
mass sensitivity
T temperature
V
Pd
volume of palladium coating
V
Si
volume of silicon cantilever
V
Ti
volume of titanium coating
V
t
total volume of cantilever with coatings
X amplitude
X(x) general solution to a differential equation
ζ damping ratio
θ phase angle
µ viscosity
ν poisson’s ratio
ξ percentage area coverage by the polymer layer
ρ density
ρ
eq
equivalent density
ρ
Ti
density of titanium
20
ρ
Pd
density of palladium
ρ
Si
density of silicon
ω angular frequency
ω
n
natural angular frequency
ω
d
damped angular frequency
Ω fringelocus function
(˜) matrix
( )
T
transpose
δ( ) uncertainty
δm uncertainty in mass
δk uncertainty in stiffness
δf uncertainty in frequency
df infinitesimal change in frequency
dm infinitesimal change in mass
∆f finite change in frequency
∆m finite change in mass
∆k finite change in stiffness
21
1. OBJECTIVES
The objectives of this thesis were to design, analyze, and perform preliminary
characterization of MEMS cantilevertype chemicals sensors for operation at 20˚C and 1
atm pressure, using ACES methodology.
22
2. INTRODUCTION
To understand the methodology and processes used in this Thesis it is important
to have background information on the MEMS cantilevertype chemical sensor. Many
different aspects will be covered, some just for completeness, in this chapter. Each
section will cover some facet of the multidisciplinary technology of these microsensors,
which can be considered to be a part of a broad category of micromachines.
2.1. Micromachine beginnings
Ignoring science fiction roots, the impetus towards micromachines was inspired
by Prof. Richard P. Feynman of the California Institute of Technology in Pasadena,
California in 1959 (Feynman, 1992). In the late 1950’s, Prof. Feynman initiated a new
area of research that is currently expanding at unprecedented rates. Technology has gone
from making macrosized machines to millimeter sized to more recently, within less than
two decades, micrometer sized devices as is shown in Fig. 2.1 (Pryputniewicz, 2002a).
Figure 2.2 compares a micro gear with a human hair for a visual understanding of their
relative scale. While nanotechnology is still largely not of everyday application at the
nanometer size, the size of the new microdevices continues to decrease (Pryputniewicz,
2002b).
23
Fig. 2.1. A dust mite dwarfs a MEMS device (SNL, 2001).
Fig. 2.2. A human hair is compared to a MEMS device (SNL, 2001).
The most researched area of micromachines is with microelectromechanical
systems (MEMS). These systems are a combination of mechanical and electrical
components built into incredibly small devices that are fabricated using sophisticated
integrated circuit (IC) batch processing technologies (Pryputniewicz, 2002a). MEMS
began in the mid 1980’s and some of the first products were accelerometers. MEMS are
intricate devices that can have several different moving parts and coupled together with
24
other MEMS can sense, analyze and perform complex operations in addition to being
able to control and actuate motion on the microscale (Hilbert et al., 2000).
MEMS have been labeled one of the most promising and relevant technologies of
the 21
st
century (Hilbert et al., 2000). Revolutionizing industrial and consumer products
and processes, their steady infiltration into everyday life has begun to dramatically
improve and change the way we live (Madou, 2002).
2.2. MEMS foreword
The acronym MEMS is used today to define both the fabrication processes and
the devices resulting from these processes (Pryputniewicz and Furlong, 2003). The
processes are a result of merging of advanced micromechanical and integrated circuit
(IC) technologies. The methods used for making MEMS are similar to those used in the
silicon wafer/chip market (Baltes et al., 2002). They are in fact mostly fabricated using
silicon wafers or some variation of them. The packaging of MEMS devices also still is,
for the most part, based on how chips are made (Pryputniewicz, 2003a). Packaging is a
technological barrier that must be worked out, at this time, for each specific application
of MEMS.
Adaptation of chip fabrication processes allowed development of bulk and surface
micromachining and highaspect ratio micromachining (HARM) (Gormley et al., 2000).
Advancements in these fabrication processes allowed construction of threedimensional
devices in the micrometer scale, Fig. 2.3.
25
Fig. 2.3. A microchain device: (a) overall view, the white bar at the top
left represents 500 µm, (b) a close up on the microchain, the white bar at
the top left represents 200 µm (SNL, 2001).
The first true MEMS using current fabrication methods appeared around 1988 as
gear trains and tongs (Hsu, 2002). These relatively complex devices were the product of
the fabrication techniques such as lithography and etching. Even though the technology
for lithography was around for about 200 years, this was the first time it was applied at a
micrometer scale. With the continued advancement of the fabrication techniques, more
and more complex devices could be made. Today’s gears are capable of rotating at one
million revolutions per second (Pryputniewicz et al., 2000), a giant leap from the early
versions. Beginning with accelerometers and pressure sensors, development of humidity,
temperature, and chemical sensors soon followed. In addition to sensors, complex
devices such as micro turbines, motors, steam engines, RF switches, optical devices, and
actuators have been since made, to name a few (Hilbert et al., 2000). Most of these
devices have gone from research to actual commercial products. An example of this
occurred in 2002. After about a decade of research, one of the most complex MEMS
devices became a commercial product known as the DMDDLP™ (Digital Micromirror
Device  Digital Light Processing) (TI, 2003).
(a) (b)
26
The DMD chips are used, e.g., in digital projectors capable of displaying high
resolution images by utilizing and rapidly moving some 1.3 million tiny mirrors
characterized by 20 µm diagonals, each representing a single pixel, comprising a single
MEMS DMD. This MEMS device utilizes a multidisciplinary approach (mechanical,
electrical, optical engineering, etc.) simultaneously to produce digital images that are
among the best in the world (TI, 2003). Details of construction of the DMD are shown in
Fig. 2.4.
Fig. 2.4. Details of a DMD chip (TI, 2003): (a) graphic illustration of the construction of
the DMD, (b) close up on the DMD array, (c) close up with one mirror removed from the
surface, (d) all the mirrors removed from the DMD, (e) close up of a single element of
the DMD without mirrors.
While it may seem that the only advantage for MEMS are their small size, there
are indeed many additional benefits (Madou, 2002). Small sizes imply that less material
(b) (c)
(a) (d) (e)
27
is used and less energy is consumed. Their small size allows for the construction of
arrays of hundreds of them on a single chip. Many sensors are also breaking records of
sensitivity with some chemical sensors having the capabilities to detect presence of
individual molecules and atoms (Britton, Jr., et al., 2000).
Perhaps the most prominent advantage to MEMS is the financial factor. By being
able to produce thousands of devices on each individual silicon wafer, the cost per unit
can be driven down to affordable prices. This can also allow development of disposable
devices, which opens entirely new product markets.
MEMS devices are rapidly making their way into every aspect of modern life.
The future is getting smaller, more accurate, and quicker, and MEMS technology is
aiding in the development of NEMS (Nanoelectromechanical Systems) technology
(Feynman, 1992; Pryputniewicz, 2002b). NEMS, true nanotechnology, is similar to
MEMS only that it deals with devices three orders of magnitude smaller in dimensions,
or in the nanometer scale. At this scale, individual molecules can be moved to make
devices only a few atoms in dimension.
2.3. Sensors
MEMS typically contain the following components: mechanical microstructures,
microsensors, microactuators, and microelectronics, Fig. 2.5. These are the general
components that make up MEMS as we know them today.
28
One of the most common applications of MEMS is by utilizing them as sensors.
In fact, the original use for MEMS was as a sensor (Madou, 2002). They have become
varied in their applications and can be found almost everywhere in everyday life. The
popularity of these sensors is mostly due to the great advantages that they posses. In
addition to their small size, MEMS sensors consume very little power and are capable of
delivering accurate measurements, which are unparalleled with macrosized sensors.
Fig. 2.5. A schematic illustration of MEMS
components and their interdependence (Madou, 2002).
Perhaps the most appealing factor to MEMS sensors is that they also are very
inexpensive to make. With each wafer producing thousands of sensors affordably, these
sensors can now be utilized in areas that were cost prohibitive before with other means.
Regardless of application, all MEMS sensors work on the principle of measuring
some form of a change, just as any macro sized sensor (Hilbert et al., 2000). Some utilize
29
common sensor methods while others take advantage of the benefits of the small scale.
While methods may alter, MEMS sensors are able to detect a multitude of different
measurable changes. These include, but are not limited to, mechanical, thermal,
chemical, radiant, magnetic, and electrical changes.
Methods of operation of MEMS sensors are different depending on the use.
While there are many methods along with many types of sensors there are a few
mechanisms that are common among them. All sensors measure a change and MEMS
devices do it with either one or a combination of the following four methods: mechanical,
optical, electrical, and/or chemical. These methods are generalizations for the basic
system in which a MEMS device gathers information from the surrounding environment.
While measurements by MEMS sensors are not limited to these four methods, they are
among the most commonly used. It should be noted that each of the methods also
depends on electrical interconnections between different components of the MEMS
sensor in order for it to function. Sections 2.3.1 through 2.3.4 describe these methods in
more detail.
2.3.1. Mechanical detection
Certain MEMS sensors measure changes based on mechanical displacement or
movement. These are some of the most appealing and intricate of the sensors. This
motion is caused, or accented, by relevant external force(s). It can then be detected either
externally with optics or internally with electrical electrodes. One of the most common is
30
the MEMS accelerometer (Bernstein et al., 1999; Hsu, 2002). A sample accelerometer is
shown in Fig. 2.6.
In the MEMS accelerometer, Fig. 2.6, there is a proof mass that is suspended on
springs and allowed to move in one or two directions, but because of operational
requirements of the device, outofplane motion is limited. External forces act upon the
proof mass causing it to displace. This motion, in turn, changes relative positions
between stationary and moving parts of capacitive pickups in the microaccelerometer
(represented by long fingerlike protrusions in Fig. 2.6), which can be measured
electronically as changes in capacitance. These changes in capacitance are interpreted
into usable data.
Fig. 2.6. A view of the “sensor” part of a dualaxis accelerometer where the
proof mass can be seen with all the perforations in the middle and a quarter
view close up on one of its corners (AD, 2003).
Another common type of mechanically actuated sensor is the pressure sensor,
which also is one of the oldest type of MEMS sensors. There are several different
31
designs for MEMS pressure sensors, but they all can be classified as either one of the
following three types: absolute, gage, and differential pressure (Pryputniewicz et al.
2002a; Johari, 2003). Typically, they are constructed as a cavity covered with a
diaphragm. When an external pressure is applied to the diaphragm, it deforms, due to
stresses produced by applied loads. These stresses are picked up by electrical
piezoresistors on the diaphragm, which in turn produce an electrical signal. By
measuring this signal and interpreting it, the pressure is determined.
2.3.2. Optical detection
While most of the applications for MOEMS (microoptoelectromechanical
systems) are in communications there are a number of uses in other areas (Madou, 2002).
Sensing differences in incoming light compared to the outputted light is the basis for
optical detection outside of communication purposes. Light is usually passed through,
reflected off, or altered by some space containing a medium in question. Some of these
sensors are for the detection of chemicals. This shows the multidisciplinary aspect of
these devices. Infrared spectrometry is an example of a method that can be used for
optical detection (Wang et al., 1999; Wuttig et al., 2002). Another example is the Fabry
Perot optical sensor (Han et al., 1996). Simply said, this sensor works by introducing a
gas into a chamber and then passing light through the chamber. Due to changes in the
chemical composition the light passing through will be different than without the gas.
32
This light is analyzed and from the results a determination can be made about the type of
gas being tested. This sensor works similarly to a test tube experiment.
2.3.3. Electrical detection
The bulk of all MEMS sensors rely on electrical connections for power and as
means of transferring data from the device into the macro world where they can be
accessed and analyzed.
Piezoresistors are used for the production of electric currents. Piezoresistors are
components that convert mechanical energy into electrical and vice versa (MS, 1998).
By attaching a piezoresistor to a mechanism that moves, an electrical signal can be
produced and measured. From the signal measurements, parameters defining the motion
of the body can be calculated.
2.3.4. Chemical detection
The detection of chemicals using other chemicals is the basis for the chemical
detection method. It is based on calculated and known reactions between certain
chemicals. When specific chemicals interact a reaction occurs. Effects of these reactions
are known and can be used to determine what chemical caused the reaction. These
effects are often exploited by, but not limited to, the use of polymers. In the broad sense,
33
a polymer is a laboratory made chemical that can be used for a variety of purposes
including controlled, calculated reactions (Lange et al., 2001; Thundat et al., 2000).
Polymers for MEMS can be engineered to be sensitive only to specific types of
chemicals. When these chemicals are present near the polymer, they get absorbed (or
adsorbed) onto it and change one or a few of the properties (e.g., illuminescence, color,
opacity, conductivity, resistivity, etc.) of the polymers. This effect is particularly useful
and is utilized in several different ways.
One way of utilizing this selective polymer sensitivity is with the artificial tongue
sensor, or FabryPerot sensor (Yoo et al., 1997). Micro storage wells are filled with tiny
beads of a polymer designed to attract a specific chemical, as shown in Fig. 2.7, and are
illuminated from behind.
Fig. 2.7. Polymer based FabryPerot sensor: (a) graphic illustration of
many wells with different polymers, (b) close up of a well, (c) close up
photograph of a FabryPerot well (Yoo et al., 1997).
The analyte is released over the wells and the polymer beads absorb the selected
chemical. Once the polymer absorbs the suspect analyte, there is a chemical reaction that
causes a change in color or luminescence in the bead. Because of the illumination from
behind, it is possible to detect these shifts in color. This change is detected with a CCD
(a) (b) (c)
34
(Charge Coupled Device) array, and means that the sensor identified the suspect
chemical. Since each well is very small, it is possible to make many wells, each with a
different polymer designed for detecting a different chemical. By having many wells
filled with different polymers it is possible to measure complex analytes quickly as well
as being able to detect multiple analytes simultaneously.
A far more robust method of using polymers, however, is with cantilever sensing
elements. Cantilever elements also rely on polymer absorption, but utilize it differently.
A micron sized cantilever, Fig. 2.8, is coated with the polymer of choice that is then
exposed to the suspect analyte (Ilic et al., 2000, 2001).
Fig. 2.8. Micron sized cantilevers developed at IBM
for research purposes (IBM, 2002).
When the polymer absorbs the desired chemical, it changes the mass of the
cantilever, which in turn changes the fundamental frequency of the cantilever. With the
cantilever attached to piezoresistors, this change can then be measured and the analyte
can be identified.
35
2.4. MEMS cantilever chemical sensor
Detection of chemicals is required for many industries. MEMS sensors bring a
novel solution for that requirement in small devices with high sensitivity. One aspect of
these sensors is that they be customized for most applications. While they seem like the
perfect solution, they are not without flaws.
There are many aspects to chemical sensors. They can be simple or highly
complex. In most cases, they are very accurate. It is possible, for example, for a
cantilever sensor to detect mass changes in picograms (Madou, 2002). Many different
technologies must come together in order to allow development of MEMS
(Pryputniewicz, 2002b). Sections 2.4.1 through 2.4.16 discuss some of the aspects of
MEMS cantilever chemical sensor technology and while not all of them will pertain
directly to this Thesis they were inserted for completeness and future work.
2.4.1. Material
The most commonly used material for fabricating MEMS devices is silicon and
silicon compounds (Madou, 2002). Although there are many other materials that can also
be used, the overall properties of silicon are very good at small scales (Petersen, 1982;
Pryputniewicz, 2002b). Combination of the yield strength and modulus of elasticity
(among other properties) of silicon is especially useful for the life and durability of small
devices. Some other materials, such as silicon carbide or diamond, offer better properties
36
but they also cost many times more than silicon making them uneconomical except for
the most demanding and specific purposes (Thaysen et al., 2002).
For the fabrication of MEMS and CMOS (complementary metaloxide
semiconductor) chips alike, silicon is first processed into ingots of about 4 feet in length.
Ingots are bars of silicon (or of any other chip materials) with 99.99% purity. Purity is
essential to making consistent flawless devices. The ingots are specially made with the
crystallographic orientation of the silicon arranged into either [100] or [111] as these are
the most commonly used (Madou, 2002). Silicon ingots and wafers are shown in Fig. 2.9
during fabrication and after slicing.
Fig. 2.9. Initial steps in fabrication of MEMS and CMOS
chips: (a) preparation of a silicon for slicing, (b) visual
inspection of silicon wafers (MEMC, 2003).
This crystallographic orientation of silicon allows MEMS to have well defined
and sharp edges and shapes with perpendicular angles resulting from etching as described
(a)
(b)
37
in Section 2.4.2. The ingots are then sliced into wafers usually ranging from 100 mm to
300 mm in diameter and about 1 mm in thickness. These wafers act as the substrate or
foundation for all MEMS devices.
2.4.2. Fabrication
The fabrication of MEMS is an entire field of research on its own (Pryputniewicz,
2002a). It is extremely diverse and still expanding (Madou, 2002). With the growing
number of companies that fabricate MEMS, the technology is continually improving.
With each breakthrough, MEMS technology becomes more affordable, better, and easier
to fabricate in ever increasing quantities. The technology is also pushing the size
envelope and continually working towards the development of smaller devices such as
NEMS (Madou, 2002).
There are several different processes for producing MEMS; most prominent of
these, however, is lithography. Lithography currently produces the smallest and most
exact features available in MEMS (Pryputniewicz, 2003a). It is comparable to dry
etching, which has many different submethods. Surface micromachining and dry
etching have an advantage over other techniques such as bulk micromachining and LIGA
(German acronym for xray lithography, electrodeposition, and molding) in the size of
the features made.
MEMS devices are fabricated in a manner similar to making a cake. Layers are
deposited on one another, patterned, and the components are released to form 3D
38
structures. Depending on the process, the number of layers, and the complexity of the
design, the fabrication process can take up to a few months.
MEMS devices are fabricated on wafers made usually of silicon, as described in
Section 2.4.1. Using one of the processes available, a MEMS is built up one layer at a
time. In most cases, the layer is grown in a vapor filled environment. This growth can be
time consuming, but there have been improvements in it. These improvements include
automation of the steps. This automation has allowed the price of MEMS to be reduced
while improving quality of the devices that are produced. One place of such
improvement is at Sandia National Laboratories (SNL) in Albuquerque, New Mexico
(SNL, 2001). SNL has developed an automated process known as SUMMiT™V
(Sandia’s Ultraplanar MEMS Multilevel Technology), which is capable of fabricating
MEMS devices out of up to 5 structural layers in complexity, Fig. 2.10 (Pryputniewicz,
2002a). SNL is the only place in the world, at this time, capable of fabricating a 5 layer
MEMS using a sacrificial surface micromachining process. The rest of the world is
limited to 3 layer fabrication processes.
Fig. 2.10. Examples of SUMMiT™V process: (a) a rachet gear, (b) a clutch
activated transmission, (c) a thermally actuated motor, (d) example of the 5
structural layers of deposition (SNL, 2001).
(a) (b) (c)
39
While it may seem simple, the fabrication of each layer is quite complex. Using
lithography, for example, a series of steps are required for each layer (Hsu, 2002; Madou,
2002; Pryputniewicz, 2002a). Starting with the deposition of the first structural
polysilicon layer onto the wafer, the polysilicon is exposed to masks, light, and chemicals
to etch holes and/or dimples. The next layer is sacrificial and it is also exposed to masks,
light, and chemicals to continue with the fabrication. Masks somewhat resemble cookie
cutters cutting out shapes in a layer of cookie dough. The light used is to harden, or
soften, certain parts of a given layer so that when they are exposed to specific chemicals,
they will etch away leaving the desired shapes. The deposition and etching steps
alternate between the structural polysilicon layers and the sacrificial SiO
2
layers making
shapes in each layer and continuing until the desired device is fabricated.
2.4.3. Packaging
One of the biggest hurdles in MEMS technology is with packaging, Fig. 2.11.
The development of MEMS devices has come a long way and new boundaries have been
reached (Pryputniewicz, 1986; Pryputniewicz et.al., 2001a; Hsu, 2002; Pryputniewicz,
2002b; 2003). A problem still remains however in being able to communicate efficiently
with these devices. While MEMS are small, the packages that contain them are still
considerable in size to facilitate their handling and assembly. Packaging of MEMS is
application specific. It is too vast for the scope of this Thesis and will not be addressed
herein.
40
Fig. 2.11. MEMS chip and its different levels of packaging (SNL,
2001): (a) MEMS component, (b) MEMS component contained
sealed chip, (c) Macro sized chip containing the MEMS component
2.4.4. Sensing types and methods
As discussed in Sections 2.3.1 through 2.3.4, MEMS sensors are used for a
variety of purposes. Most of the different types of chemical sensors available are
targeted for the detection of specific chemicals. These can either be regular, simple
composition chemicals or complex biological agents. Chemical sensors have even been
used for the detection of DNA strands with different base sequences (Fritz et al., 2000a)
and are also capable of measuring pH values in analytes (Bashir et al., 2002). With so
many different types of sensors, it is necessary to categorize the chemical detection
methods to better understand MEMS.
Of the different types of sensors, there are two basic methods of operation to
detect chemicals. The two methods could be called contact and noncontact methods.
(c) (a)
(b)
41
Contact methods include any system in which the analyte is physically touched or altered.
In noncontact mode, the analyte is observed from a distance and its behavior and/or
characteristics are measured and studied.
The contact method uses polymers deposited on cantilever sensing elements to
absorb the specific chemicals and thus produce a change in mass, stress, electrical, or
thermal properties of the element. For a change in mass, an increase in mass can be
detected by measuring the change in frequency of the resonating cantilever, which has
mass as one of its parameters, Fig. 2.12. The second way is by measuring change in
stress in the element, which occurs when mass is absorbed into the polymer. This stress
(or surface stress) change can then be measured by piezoresistors on the surface of the
sensing element. The other way to detect the suspect chemical is to use a thermocouple
to measure the change in temperature due to heat produced from the absorption of the
analyte by the polymer. All of these methods have been successfully incorporated into
chemical sensors (Koll et al., 1999; Lang et al., 1999; Jensenius et al., 2000; Kerness et
al., 2000; Thaysen et al., 2001; Subramanian et al., 2002; Baselt et al., 2003).
Two other methods of interest are SAW (Surface Acoustic Wave) and BAW
(Bulk Acoustic Wave) sensors, Fig. 2.12 (Ivanov, 2000). They function when the analyte
comes in contact with the polymer “containing” either a surface or a bulk wave that
travels through the material. A change in these waves can be measured with sensors.
The downside of these methods is that they have a much lower accuracy when compared
with the methods described earlier in this section.
42
Fig. 2.12. Operation of different gravimetric sensors: (a) graphic illustration of
a cantilever bending under the load of analytes attaching to the polymer coating,
(b) illustration behind the principles of BAW and SAW (Ivanov, 2000).
2.4.5. Resonance frequency
The main principle behind cantilever sensors is with the measuring of change in
the resonance frequency (Betts et al., 2000). Resonance frequency is the frequency that
produces the largest amplitude that a vibrating body can achieve. This frequency is
dependent on the spring constant of the body and its dynamic mass. Should one or the
other of these parameters change, the resonant frequency will also change. This is the
basis behind using change in resonance frequency as means of detection. Alteration in
the mass of a vibrating cantilever changes the total mass, which in turn shifts the resonant
frequency to some other frequency. By measuring this shift, one can determine the
amount of mass that was added to the cantilever assuming the spring constant remained
constant. If the cantilever is selective to the kind of analyte it absorbs, then in addition to
detecting the chemical, if present, the sensor can also detect the concentration of it. This
method of chemical detection is highly accurate. Some experiments have successfully
(a) (b)
43
demonstrated the ability to detect a change in mass of as little as 0.7 picograms (Madou,
2002).
2.4.6. Cantilever actuation methods
In order for the MEMS sensor to function in dynamic, or frequency, response
mode, it is necessary for the cantilever to be vibrating (Arecco and Pryputniewicz, 2003).
There are several ways of actuating a cantilever into its resonance frequency, which can
be classified as direct and indirect actuation methods (Cho and Ahn, 2002; Li et al., 2002;
Stephan et al., 2002). Indirect methods include using acoustic vibration to shake the
sensor much like a piece of paper in front of a speaker vibrates from the compression
waves in the air. Another indirect method has a magnetic coat deposited on the cantilever
and an inductance coil underneath it to generate magnetic fields that attract and repel the
cantilever into vibration.
For direct methods, the most common is by the use of a PZT (PbZrTi transducer
or, as it is sometimes called, piezoelectric zaxis actuating transducer), which is capable
of vibrating at different frequencies very accurately depending on the driving voltage
(Harley, 2002; Mehta et al., 2001; Turner and Zhang, 2001). Another method that is less
common, but of great interest, is by fabricating piezoresistive electrodes directly on the
cantilever and bending it rapidly into vibration by bimorph actuation. Figure 2.13 shows
some sample PZTs available commercially.
44
Fig. 2.13. Comparison of different sizes and shapes of PZTs available:
(a) OpenLoop LVPZT Translators, (b) Preloaded Open & Closed
Loop HighLoad HVPZT Translators, (c) PICMA™ Chip Monolithic
Multilayer Piezo Actuators (PI, 2003).
It is of interest to combine and use direct and indirect methods when conducting
research. It is even possible to combine both direct methods into one package that allows
for highspeed functionality (Kim et al., 2003).
2.4.7. Cantilever I/O
In order for the sensor to function it is necessary to have some means of
communicating out the data it collects. There are several ways of accomplishing this and
the most relevant to this Thesis will be discussed.
(b)
(a) (c)
45
One of the most common ways of detecting frequency shifts is by optically
inspecting the cantilever. Using a technique similar to those in AFMs (atomic force
microscopes), a laser is reflected off the cantilever tip and into a position sensitive diode
(PSD) (Raiteri et al., 2001; Kim et al., 2001a; Battiston et al., 2001). The reflected angle
changes due to the bending of the cantilever. This change is detected by the PSD and it
can be measured and analyzed, Fig. 2.14. While the technique is proven and quite
reliable, it is too bulky for portable purposes.
Fig. 2.14. Graphical illustration of the operational principle
for an AFM where different bend angles reflect the laser
beam to different locations on the PSD.
Another method that is also quite popular is by using piezoresistive elements on
the cantilever itself (Porter et al., 2003; Lange et al., 2002). Any changes in the surface
stress due to bending will cause an electrical output from the piezoresistive electrodes.
46
This output can be analyzed and the frequency can be calculated. Figure 2.15 shows how
a typical piezoresistor is used on a cantilever.
Fig. 2.15. Cantilever with integrated piezoresistors: (a) view of piezoresistive cantilever
used for scanning probe microscopy (SPM) applications, (b) close up of the Wheatstone
bridge piezoresistive elements (Gotszalk et al., 2000).
2.4.8. Polymer thickness
In fabricating cantilever sensors with special coatings it is important to take into
account the thickness of this layer. The thickness affects several aspects of the sensor.
Mostly, it affects the rate at which the polymer becomes saturated with the suspect
analyte (Britton et al., 2000). The thicker the polymer, the longer it takes for the polymer
to reach saturation. This is not really a problem unless time is a critical factor in the
function of the sensor. It is generally preferred to make the coating as thin as possible,
however, to minimize other effects. One of the most prominent of these effects is that of
residual stresses which will be discussed in Section 2.4.14.1. A thicker layer of polymer
(a) (b)
47
is more reactive to effects of different coefficients of thermal expansion and also it is
more sensitive to pressure effects, which are discussed in Sections 2.4.16.
2.4.10. Fick’s law of diffusion
For any particle that is absorbed into another there is a rate at which the diffusion
takes place (Hughes and Bastasz, 1988; Streeter et al., 1998; Hu, 2001). During
diffusion, or mass transport, a gas (or liquid) with a higher concentration of a sample will
move into an area of lesser concentration. Diffusion is usually an interstitial motion in
that the sample gas travels to the voids in the lattice structure of a material. The diffusion
can be modeled with Ficks’ laws. These laws address two specific conditions, steady
state and nonlinear (nonsteady) diffusion. During steady state diffusion (first law), the
rate is constant, as it would be in an ideal condition with infinite volumes of gas passing
through a membrane. The nonlinear state (second law) is a more accurate model for
finite volume cases where the diffusion rate will change as the concentration gradient
equalizes.
2.4.11. Reference cantilevers
In order to guarantee sensor functionality and reliability, the data quality must be
high. Because of signal noise from a variety of different sources, such as thermal effects,
48
electrical noise, and others, it is important to include a parallel system to improve the
signal. By incorporating a reference cantilever it is possible to minimize noise and
improve the signal quality. Reference cantilevers function by having them react to only
the noise producing effects, which means that they are not coated with the analyte
attracting polymer (Thaysen et al., 2000). The signals of the reference and coated
cantilevers can then be compared and external forces (noise) can be removed
computationally thus improving quality of data.
2.4.12. Quality factor
Most vibrating mechanical structures have a variable known as the quality factor,
Q (or Qvalue) (Yasumura et al., 2000). This value is a nondimensional parameter for
quantitatively assessing the ‘quality’ of a structure. In other words, the Qvalue is a
measure of how much energy is dissipated in a vibrating system. The higher the Qvalue,
the better the structure is capable to conserve energy. If the structure can conserve
energy efficiently, the structure can operate longer and produce more data. It should be
noted that Q has a great dependence on damping and thus the environment that the
structure operates in. Structures in air usually operate much better than those in liquids
and the Q will be the highest in a vacuum environment (Tamayo et al., 2001).
The quality factor is reflected in the shape of a resonance frequency peak. The
narrower the peak of the frequency response curve, the higher the Q value is while the
opposite is also true. The width of the peak is also sometimes referred to as the
49
bandwidth of the resonance frequency curve. This bandwidth is measured at half power
points, though, for consistency and convenience.
2.4.13. Electrical aspects
Practically all MEMS devices collect measurements electronically. This means
that there must be a mechanism for producing a signal to the macro world. The signal
produced must be directly linked to the motion, or action, that the MEMS performs in
order to determine quantitative results. This mechanical to electrical effect is
accomplished with a phenomenon usually referred to as a piezoresistive effect (MS,
1998). The piezoresistive effect is the change in electrical resistivity that occurs with the
application of mechanical stress, which allows conversion of stress into proportional,
measurable electrical signals. The piezoresistive effect is observed in piezoresistive
elements and materials.
One of the most common gauges used is a piezoresistive strain gauge (Chow et
al., 2002; Thaysen et al., 2002). They are used for detecting the action, or motion, that is
incurred in a sensor. The gauge is usually attached to, or built into, the device as that will
produce the most accurate results. The operation is based upon measuring increasing
stress/strain in an element of the MEMS. When a strain is produced in the gauge, an
electric signal is generated. This signal can then be amplified and measured.
50
2.4.13.1. Wheatstone bridge
The basic and most commonly used method of gathering and measuring the
electric signal generated from gauges is with a Wheatstone bridge circuit, Fig. 2.16. The
Wheatstone bridge is a classic and common configuration for determining voltage
differences in electrical circuits. Four resistors are connected together with a middle
resistor, R
1
, (usually the strain gauge or other measuring gauge) acting as a gate switch.
Any variation in current in the middle resistor will cause a difference in voltage output
from the circuit (Pauw, 1958). For MEMS applications, the middle resistor would be a
piezoresistive element, which can change its resistance with mechanical changes.
Fig. 2.16. Wheatstone bridge where changes, e.g., in R
1
, alter
the output voltage with a high accuracy and precision.
2.4.13.2. Feed back loop
Because the method of extracting data is usually electrical, there is typically noise
present in the results. This noise and errors in the data can be reduced, however, with the
51
use of feed back loops (Abadal et al., 2001; Humphris et al., 2000; Sulcheck et al., 2000).
Feed back loops is a post processing technique used to increase accuracy of the data in
which the data are refined several times. It is utilized with computers and special circuit
setups. Active feed back loops continuously correct the data being processed so that
error is minimized. These can be very complex systems and require a lot of knowledge
in electrical engineering. Because of time, equipment, and complexity concerns, feed
back loops will not be incorporated into the experiments planned for this Thesis.
2.4.14. Inherent problems
In designing a MEMS device, there are a multitude of variables that must be
accounted for. With so many variables, there are inherent problems that arise (Handel,
2001; Manias et al., 2001; Kassegne et al., 2002). Tweaking and adjusting of techniques
and design can overcome most, however. There are two problems, though, that are
reoccurring and of some concern: residual fabrication stresses and stiction.
2.4.14.1. Residual stresses
During the fabrication of a multilayered MEMS device two or more dissimilar
layers are deposited on top of one another depending on the process and design. Because
of differences in coefficients of thermal expansion (CTE) of the materials, residual
52
stresses will be produced in these layers due to the fabrication processes (Fritz et al.,
2000b; Lu et al., 2001; Pryputniewicz et al., 2002c, 2003b). A lot of the fabrication
techniques involve depositing layers at high temperatures (600˚C +). When the
fabrication is over and temperatures return to room temperature (~20˚C), different
materials will shrink (or expand) by different amounts leading to stress gradients
(Pryputniewicz, 2003a, 2003b). These gradients will manifest themselves as warping of
structures, cracks, or other failure modes; Fig. 2.17 shows the effects of residual stresses
on a MEMS component. This is a constant design problem that must be taken into
account. One method of reducing and/or eliminating residual stresses is by annealing the
device after the deposition process. Anneling is a process of heating the device at a
certain temperature for a specific time in order to allow for any stress to dissipate. This is
a proven process, but requires a lot of experimental verification and takes up valuable
production time.
Fig. 2.17. Effects of residual stresses: (a) a MEMS device is bent after its release, (b) a
close up on the device showing how extreme the bending is (Dartmouth, 2002).
(a) (b)
53
2.4.14.2. Stiction
A very common problem that plagues devices with small thickness (1 to 4 µm) is
a phenomenon known as stiction. Stiction is probably the most serious of problems that
occur in micromachining (Buks and Roukes, 2001). Stiction occurs when two flat
surfaces are near each other, Fig. 2.18. This close proximity along with forces such as
van der Waals cause the components to tend to stick to one another.
Fig. 2.18. Collapse of a thin cantilever due to stiction.
In a cantilever type sensor, stiction can cripple a device making it useless. It is
possible to fix stiction, but during that process excessive forces are required which can
damage or destroy the device. Another major problem with stiction is that it can occur at
any time and not just during fabrication. There are solutions for this problem. New
organic modifiers (coatings) have been developed against stiction and have been tested
and shown to lessen the effect (Kim et al., 2001b). With some limited solutions already
in place and continued research, stiction will hopefully become a problem of the past in
the near future.
54
2.4.15. Effects of temperature
Temperature has a critical effect on the functionality of the MEMS sensors
(Hanson et al., 2001; Pryputniewicz, 2003a). Because the cantilever sensor is based on a
bimaterial ( e.g., substrate and polymer coating) there will be two materials and thus two
coefficients of thermal expansion. Any temperature change in the environment will
affect the beam (Gotszalk et al., 2000). Increase or decrease in the temperature will
produce a surface stress on the cantilever due to different CTEs and thus create a
pronounced bending as shown by Stoney (1909) and Pryputniewicz et al. (2003a). This
bending can alter results of any detection system setup. If any temperature change will
be expected it is necessary to include reference beams to allow for the compensation of
this phenomenon (Hanson et al., 2001). For producing better results and data it would be
desirable to reduce the temperature induced bending, as this would lessen the amount of
compensation required from the reference beams. This means that the temperature of the
sample should ideally remain as constant as possible. The effects of temperature have
been, however, capitalized to create calorimeters with femtorange accuracy, Fig. 2.19
(Kerness et al., 2000).
2.4.16. Effects of pressure
The pressure of the environment in which the sensor is in can affect the results
produced. If the pressure of the environment increases, the dynamic viscosity of the
damping fluid increases. This increase in damping reduces the Qvalue and can also
55
reduce the deflection of the cantilever which both, in turn, decreases the data quality and
reliability. In addition to changing the damping, an increase in pressure will also alter the
concentration of the gas in the environment, which can affect the absorption rate, and
equilibrium giving shifted results.
Fig. 2.19. A femtocalorimeter utilizing thermal bending
effects with piezoresistors (Gotszalk et al., 2000).
56
3. MEMS SAMPLES
This chapter describes the MEMS samples used in this Thesis and their
characteristic parameters and includes information for the proposed MEMS chemical
sensor.
3.1. MikroMasch USA cantilevers
Sample cantilevers used in this Thesis were purchased from MikroMasch USA
(MM, 2003). The company is headquartered in Estonia, Spain, and their main product
are atomic force microscope (AFM) chips. AFM chips are siliconbased cantilevers that
have a high aspect ratio tip at the end. They are used for topographic surveying on the
nanometer scale of the samples investigated. The samples used in this Thesis are actual
AFM contact silicon cantilevers that are uncoated and also are without any tips, model
number CSC12 (MM, 2003). They are micromachined cantilevers with the fixed ends
being integral parts of a chip. It is important that the cantilevers be uncoated so that a
special coating can be deposited on them without any other underlying layer. Figure 3.1
shows an AFM chip with cantilevers with tips from MikroMasch, while Fig. 3.2 shows
the special samples purchased for this Thesis. The nomenclature illustration in Fig. 3.1
applies to the samples used in this Thesis. In fact, the labeling of the cantilevers (AF)
shall be used from here on to refer to each cantilever, as only the three longest cantilevers
(DF) shall be investigated in this Thesis. Table 3.1 lists the manufacturer’s
specifications of the samples purchased. As can be seen from the photographs, the chips
57
contain 6 cantilevers of varying lengths, all listed in Table 3.1. The thickness and width
of each of the cantilevers should be the same according to manufacturers specifications.
The cantilevers and the chips are all made of singlecrystal silicon with the dimensions
and material properties listed in Table 3.1.
Fig. 3.1. MikroMasch AFM chip: (a) SEM photograph of cantilevers
(A,B,C) with tips, (b) dimensions of the chip and cantilevers as well as a
standardized nomenclature (MM, 2003).
Table 3.1. Characteristic dimensions, as supplied by the manufacturer, MikroMasch
USA, and the corresponding material properties as listed by Madou (2002).
A B C
Characteristics Min Typical Max Min Typical Max Min Typical Max
Length, L ± 5 µm 110 90 130
Width, b ± 3 µm 35 35 35
Thickness, h ± 0.3 µm 0.7 1 1.3 0.7 1 1.3 0.7 1 1.3
Resonance freq., f kHz 65 105 150 95 155 230 50 75 105
Force constant, k N/m
0.25 0.95 2.5 0.45 1.75 5 0.15 0.6 1.5
D E F
Characteristics Min Typical Max Min Typical Max Min Typical Max
Length, L ± 5 µm 300 350 250
Width, b ± 3 µm 35 35 35
Thickness, h ± 0.3 µm 0.7 1 1.3 0.7 1 1.3 0.7 1 1.3
Resonance freq., f kHz 9.5 14 19 7 10 14 14 20 28
Force constant, k N/m
0.01 0.05 0.1 0.01 0.03 0.08 0.02 0.08 0.2
190 Modulus of elasticity, E ± 3 GPa Density, ρ ± 0.05 g/cm
3
2.33
(a) (b)
58
Fig. 3.2. Photographs of the AFM chips used in this Thesis: (a) a view of the entire chip
from above with all the cantilevers visible. The cantilevers of interest are the ones on the
right, (b) a top view of the primary cantilevers used in this Thesis (D,E,F), (b) a
photograph of the underside of the cantilevers (D,E,F).
3.2. Chemicals used
This Thesis is based around a sensor that is capable of detecting a specific
chemical. Thus it is important to select a chemical that is easy to detect and useful for
research purposes. While it is possible to detect just about any analyte with the right
polymer, some polymers are easier to make and are more accessible than others. Due to
the focus of this Thesis and time/cost issues, availability played an important role in
determining the polymer/analyte combination to use. In the process of coating, there are
(a)
(b) (c)
59
many different types of materials that are fairly standard and can be found and used
easily and quickly. Most are not used for a gravimetric chemical sensor, but there are a
few that are. One of these materials is palladium (Pd). Palladium will, at room
temperature, absorb 800900 times its own volume of hydrogen (Lewis, 1967). This is a
significant amount for a very reasonable condition. The process is also reversible. The
palladiumhydrogen system is very well documented and very easy to conduct
experiments with (Lewis, 1967). While there are other absorption systems, they are
much more complex and can be very difficult to set up and maintain. Most systems are
not reversible in absorption without some kind of treatment to the absorbing material.
3.2.1. Palladium
The element palladium (Pd) was selected as the absorbent layer for the cantilever
because of its sensitivity to hydrogen (Darling, 1958). As mentioned in Section 3.2, it
has an ability to absorb (not adsorb) great quantities of hydrogen. It is in fact a very
unique ability in that there are few other materials that naturally behave as palladium
does with hydrogen. Because of this phenomenon, palladium is often used as a filter of
hydrogen as when it is heated; only hydrogen will diffuse through it. A true benefit of
this interaction for the Thesis is that the absorption will occur at room temperature. The
hydrogen then desorbs out of the palladium when the hydrogen source is removed. These
are highly desired features that simplify the experiments. Properties of palladium are
found in Appendix A.
60
3.2.2. Hydrogen
The element hydrogen (H
2
) is the lightest and most abundant in the universe. It is
a very versatile element and a very promising fuel. When it burns the only byproduct is
water. The one fact about hydrogen that limits its use is its explosive nature. Hydrogen
is explosive from 4%  40% concentration. This means that if the concentration is below
or above that range, hydrogen will not burn. Because of this range, great care must be
taken when setting up any experiment dealing in hydrogen to prevent a concentration
greater than 4%. Premixing a nitrogenhydrogen gas with 1% and less hydrogen
concentration reduces the risk factor. This would be acceptable since this Thesis will
investigate a sensor for detecting the lower concentration limits (<<1% hydrogen).
Nitrogen would be the inert gas carrier for this experiment as it is very stable and helps
reduce humidity from the cantilevers. In the setup, special care will be used to ensure
safety during the experiments. Properties of hydrogen can be found in Appendix A.
3.2.3. Titanium
There will be a small amount of titanium (Ti) used in the coating of the
cantilevers. During the deposition process, a thin layer of Ti will be placed before
coating the cantilever with palladium. The reason behind this thin layer is to help prevent
delamination, or separation, of the palladium from the singlecrystal silicon cantilever. It
has been seen that palladium will delaminate in the presence of hydrogen if it is directly
coated onto silicon. For this reason, a very thin layer (approximately 10 Å of Ti) will be
61
used as an adhesive layer between the silicon and the palladium. Even though the layer is
very small, it will be accounted for in calculations. The properties of titanium can be
found in Appendix A.
3.3. Coating process
Once all the materials have been selected for coating the cantilever, it is necessary
to use a technique that will allow for the deposition of layers with accurate thickness. For
this task, resources such as those at Case Western Reserve University (CWRU) in
Cleveland, Ohio can be used. There, a sputtering machine can be used to coat layers of
Ti and different, controlled thickness of Pd on the cantilevers. The thickness for
palladium, as will be discussed in Chapter 4, only affects the length of time for the
absorption process to complete. To verify this, it would be recommended to get two
different Pd thicknesses: 1000 Å and 2000 Å.
For the calculations in this Thesis, it was assumed that the deposition process
would be ideal in that there is a perfect bonding between the layers and there are no
imperfections. This is a fairly safe assumption to make given the nature of the deposition
process. It was also assumed that the properties of the layers (including the silicon
substrate) are isotropic in nature.
62
3.3.1. Complete versus partial coating
Coating the cantilever with palladium can alter the results, quality, and sensitivity
of the sensor. Therefore, it is important to consider the placement of the palladium
coating. Calculations describing sensitivity and placement of the palladium on the
cantilever are discussed in Chapter 4. As it will be shown in Section 5.8, coating the
entire length of the cantilever is beneficial to the sensitivity. In addition to coating the
entire length, all five sides (top, bottom, left and right sides, and tip) shall be coated. By
doing this, bending effects due to any differences in the values of parameters
characterizing coatings can be minimized, as there will be (ideally) equal and opposite
bending moments and stress gradients. Figure 3.3 illustrates the differences between
bending moments of a fully coated on both sides (i.e., top and bottom) and a oneside
(i.e., top only) cantilever. It is assumed that the deposition of any titanium and palladium
is uniform throughout and that this will also effectively minimize, or even eliminate, any
bending effects.
(a) (b)
Fig. 3.3. Bending moment comparison: (a) the bending moments of a fully
coated cantilever where the moments cancel each other, (b) the moments of a
oneside coated cantilever, where moments add to each another.
63
3.3.2. Effects of humidity
Once the cantilever is coated with palladium, it will be sensitive to humidity in
the air (Baselt et al., 2003). Palladium will naturally start to form palladium oxide in the
presence of water or oxygen, which will slow or prevent hydrogen absorption. This will
cause the palladium to expand as well. As the palladium oxide is exposed to hydrogen,
the oxide will slowly recede and the palladium will become pure again allowing for the
hydrogen to absorb normally. For this reason, care should be taken to ensure that the
cantilever experiments will be conducted in an oxygen free environment with as little
humidity as possible. With the hydrogen being delivered in nitrogen–hydrogen mix, the
relative humidity will be reduced, and potentially eliminated, as oxygen will be removed
from the environment around the cantilevers. This would resolve any humidity problems
that might arise.
3.3.3. Time limits
For the palladium to reach 100% saturation of hydrogen, the time required is
exponentially high in that it will take an infinite amount of time for the hydrogen to reach
complete equilibrium in the palladium. For this reason, the experiments should be run till
the saturation level is only 90%. This diffusion process is described in more detail in
Section 4.2.4.1. For desorption, the time is just as long, if not longer, taking up to days
depending on the varying conditions of the system, which is unacceptably long for
practical operation of the sensor. Because it is very hard to fully remove 100% of all the
64
hydrogen once it has been absorbed into the palladium, there would be some residual left
over. This residual hydrogen will keep the palladium lattice permanently changed. Over
the course of several runs, the absorption/desorption process will deteriorate the structure
eventually leading the palladium to fail via cracks or delamination. Because of this
limitation, the first few runs are the most important for data collection and the total
number of experiments conducted on each chip should be minimized while increasing the
total number of chips experimented on to help improve the quality of the information
gathered.
65
4. METHODOLOGY
This Thesis conducts a study to optimize performance of a cantilever chemical
sensor, which is to operate at the conditions of room temperature (~20˚C) and sea level
pressure (1 atm). Because of its nature, this study will be based on the analytical,
computational, and experimental solutions (ACES) methodology (Pryputniewicz, 1997;
Pryputniewicz et al., 2001a, 2003a).
4.1. ACES methodology
The methodology used in this Thesis is known as ACES (analytical,
computational, experimental solutions), Fig. 4.1. This methodology uses analytical,
computational, and experimental solutions to obtain results which otherwise might be
difficult to obtain.
The analytical part of the solution is based on exact, close form solutions. These
solutions are applicable when the boundary, loading, and initial conditions can be
specified, or in other word, for the equations to be comprehensive to all the aspects of the
problem. The information generated greatly facilitates any computational and
experimental work that needs to be performed.
The computational solution is based on using finite element method (FEM)
software to model the problem and to produce approximations of the solution. The
66
solutions are based on discretization of the governing partial differential equations
(PDEs).
Fig. 4.1. Graphical representation of ACES methodology and the interdependence of
each solutions category towards determining a final result.
Because of continued efforts to advance experimental capabilities, a number of
different methods have been developed. Practicality of these methods, however, depends
on the specific application. For example, in this Thesis, optoelectronic laser
interferometric microscopy (OELIM), laser vibrometry, atomic force microscopy (AFM),
and scanning electron microscopy (SEM) methodologies were used. Each method carries
with it its own strength and weakness. However, comprehensive testing using several
methods can improve the overall quality of the data. This is one reason for comparing
the analytical and computational solutions to the experimental.
67
4.2. Analytical investigations
In order to commence calculations for the analytical investigation of the
cantilever, it is necessary to understand what must be evaluated. There are a number of
publications (Battiston et al., 2001; Bashir et al., 2002; Baselt, 1993; Britton et al., 1999;
Brown and Pryputniewicz, 1992, 1995, 1996, 2000; Pryputniewicz, 1985a, 1988, 1991a,
1991b, 1995, 1996, 1998, 2003b, Pryputniewicz et al., 2001b, 2001c, 2002a, 2002b,
Pryputniewicz and Stetson, 1990) listing different properties and characteristics of
cantilevers. Some of the publications address issues that must be considered. For this
Thesis, a sensor will be investigated that operates at room temperature and 1 atm
pressure. Sections 4.2.1 through 4.2.11 describe some of the necessary issues that need
to be considered in the calculations for this cantilever based chemical sensor.
Several different cases will be considered while analyzing a vibrating cantilever.
They will take into account different combinations of parameters characterizing a
vibrating body. These parameters describe damping, stiffness, external forces, and other
characteristics of the cantilevertype MEMS sensor. Depending on the specific set of
parameters used, the results might change. Once the different cases or scenarios are
taken into account, the results of each will be compared. This will help in determining
how each parameter affects the overall results.
68
4.2.1. Sensitivity
A way of measuring how well a sensor functions is by determining its sensitivity.
It is important to determine the sensitivity of the cantilever, as there are different ways of
coating the cantilever with palladium. By different ways, it is meant that by coating the
entire cantilever versus just coating a small portion at the tip with palladium the
sensitivity will change. It is also important to clarify the condition of the sensitivity. The
cantilever will be exposed to hydrogen until the coating of palladium reaches 90% of
saturation. This will allow mass of hydrogen to accumulate on the cantilever leading to a
shift in frequency. The sensitivity can then be defined as the ability for the cantilever to
detect shifts in the frequency. The higher the sensitivity of the cantilever, the better it
will function.
Sensitivity, S
m
, of this sensor can be defined as a function of the frequency shift
and the absorbed analyte mass, (Madou, 2002) i.e.,
0
1 1
lim ,
m
m
i i
f f
S
f m f m
∆ →
∆ ∂
= =
∆ ∂
(4.1)
where f is the initial frequency, ∆f and ∂f are finite and infinitesimal, respectively,
changes in the frequency, and ∆m and ∂m are the changes mass (see Appendix B).
Additional Figure 4.2 illustrates the boundary conditions and equivalent system of the
cantilever.
Calculating the sensitivity is important in determing the best location for the
polymer coating. Changing the coverage of the cantilever by the polymer can alter the
sensitivity of the sensor.
69
Fig. 4.2. Idealization and modeling of a cantilever.
If we cover the cantilever only at the tip or with just a small percentage compared
to the entire cantilever we get sensitivity for an end loaded cantilever, S
e
, as (see
Appendix B for derivation)
( )
( )
2
1
,
2
e
knr
S
k k m nr m
m m nr m
−
=
+ ∆
+ ∆
(4.2)
where k is the spring constant of the cantilever, n is the effective mass constant, r is the
percent area coverage by the polymer layer of the cantilever, m is the original effective
mass (see Appendix B.3), and ∆m is the change in mass of the cantilever. For
simplification of the problem, it will be assumed that there is no change in the spring
constant.
The percentage area coverage, r, used was estimated from Fig. 5.16 of the
dynamic mass effect on the cantilever. Derivation of the dynamic mass utilizing the
boundary conditions of a cantilever, Fig. 4.2, uses the kinetic energy, KE, equation
2
1
,
2
KE mv = (4.3)
70
where m is the mass of the moving body and v is its velocity. If the equation is solved
subject to the cantilever boundary conditions of a sensing element, the final solution is
2
max
1
( ) ,
2
v
KE C n x = (4.4)
where v
max
is the maximum velocity experienced by the body in motion, C
1
is a constant
that is comprised of structural properties of the cantilever, and n(x) is a constant that
varies with the position along the cantilever that describes the effective mass. From Eqs
4.2 to 4.4, the correct amount of coverage can be calculated to analyze the cantilever
sensitivity.
If we now look at the case where the polymer covers the entire cantilever the
sensitivity for a distributed load, S
d
, will be
( )
( )
2
1
,
33
2
140
33 33
140 140
d
k
S
k k
M m
M M m
−
=
+ ∆
+ ∆
(4.5)
where M is the mass of the cantilever. It is now possible to compare the two equations
and determine which will produce the largest frequency shift. Increasing the sensitivity
increases the detectable shift in frequency of the cantilever. In determining the sensitivity
of the cantilever sensor it is important to maximize this response, as it will directly
determine performance of the sensor (Hu et al., 2001; Lange et al., 2001; Hsu, 2002; Yu
et al., 2002). High sensitivity can be defined as the ability to detect a very small
frequency response.
71
4.2.2. Equivalent variables
In order to model the cantilever more accurately, it is necessary to account for all
the layers and coatings on it. Each layer of the cantilever will have a different material,
which in turn means that it has a different modulus of elasticity and density. When
calculating the frequency shift it is necessary to use an equivalent value for both
parameters. The equation for the equivalent modulus of elasticity is (Jones, 1975;
Berthelot, 1999)
( )
,
Si Ti Pd t
eq
Ti Pd Si Si Pd Ti Si Ti Pd
E E E V
E
E E V E E V E E V
=
+ +
(4.6)
where E
Si,Ti,Pd
are the moduli of silicon, titanium, and palladium, respectively, V
Si,Ti,Pd
are
the volume fractions for each element, and V
t
is the total volume. Equation 4.6 is used
along with the following equation for the equivalent density (Jones, 1975; Berthelot,
1999):
,
Si Si Ti Ti Pd Pd
eq
Si Ti Pd
V V V
V V V
ρ ρ ρ
ρ
+ +
=
+ +
(4.7)
where ρ
Si,Ti,Pd
are the densities of silicon, titanium, and palladium, and h
Si,Ti,Pd
are the
thicknesses of each of the material layers, respectively.
4.2.3. Gas concentration
Since the hydrogen that is to be sensed would be in a gas form it is necessary to
understand how to deal with any gas calculations. The two most common ways of
72
referring to gas concentrations are with percent concentration (%) and with partsper
million (ppm). The relationship between the two forms is
0.0001% 1 , ppm = (4.8)
where the concentration of the volume of the analyte gas to the carrier gas is equated. To
use concentration in equations, it is necessary to convert it into a more useful form such
as density, i.e.,
( )
3 3
_ 273
1 ,
0.0224
ppm molecular weight p
mg
m
m
T
mol
( ⋅ ⋅ ⋅
¸ ¸
=
(
⋅
(
¸ ¸
(4.9)
where ppm is the gas concentration in partspermillion, molecular_weight is the
molecular weight of the sample analyte, and p and T are the partial pressure (in atm) and
temperature (in °K) of the gas, respectively. Using Eq. 4.9 it is possible to estimate the
amount of mass that will be absorbed for a given volume of palladium. It is important to
note though that the partial pressure is of the gas to be detected from a sample of a
gaseous mixture, in this case hydrogen, and will play a vital role in the results.
4.2.4. Absorption
The main operating principle of the cantilever sensor is its ability to absorb an
analyte and thus alter resonant frequency of the sensing element. To determine the
maximum amount of hydrogen that can be absorbed into the palladium it is necessary to
look at the steady state situation. The amount of hydrogen that is absorbed varies with
73
different pressures and temperatures. According to experimental results, reported in
literature throughout the years, absorption of hydrogen is nonlinear at lower
temperatures, Fig. 4.3. This figure shows the relationship between the partial pressure of
hydrogen and the atomic ratio (at equilibrium) at different temperatures. The atomic ratio
(H/Pd) is the volumetric, or mass ratio, between two elements in a given volume. In this
case, the atomic ratio compares the hydrogen volume in palladium. While the figure is
accurate, it does not list the temperature, ~20ºC, or room temperature, that is of interest.
Since the hydrogen that is to be investigated will be in a mixture with nitrogen,
we must look at the partial pressures of the hydrogen. The total pressure for the mixture
will be at 1 atm (101kPa), so for a concentration of 1% hydrogen the partial pressure will
be 0.01 atm (1.01kPa) and the figure indicates that the atomic ratio, or absorption of
hydrogen, will be around 0.621 at 20˚C. While this is not the specified temperature, it is
fairly close and the atomic ratio listed for this temperature (20˚C) will be used.
However, Fig. 4.3 does not have much information about very small pressures,
which for this Thesis equals the small concentrations (<0.1% or <1000 ppm) that are of
interest. There is very little information about the atomic ratio at such small
concentrations available. Because of this, the atomic ratio at 1 ppm had to be
approximated from the available information. Since the concentration is proportional to
the partial pressure, which in turn is proportional to the atomic ratio, a square root
relationship was used to relate the small concentrations with the atomic ratio. The square
root proportionality comes from Sievert’s Law that states (Lewis, 1967)
, c s p = (4.10)
74
where c is the hydrogen concentration in equilibrium in the palladium, s is the Sievert’s
constant parameter, and p is the partial pressure of the hydrogen. Using Eq. 4.10, the
atomic ratio was approximated at small partial pressures for equivalent concentrations. It
is important to note that Fig. 4.3 is assumed as completely correct, as the uncertainty in
the curves in the figure will be zero for isotropic conditions of palladium. Table 4.1 lists
several steady state concentrations at different pressures along with several conversions
of the atomic ratio.
Using information summarized in Table 4.1, it is possible to estimate the amount
of mass that would accumulate on the cantilever and thus cause a shift in frequency.
Fig. 4.3. Concentration and pressure relationships based on experimental
data of many research studies conducted throughout the years, where the
concentration, AR, is in units of atomic ratio (H/Pd) (Lewis, 1967).
75
Table 4.1. Concentrations of hydrogen in Pd based on experimental
data of Fig. 4.3 at 20ºC with conversion information (Lewis, 1967).
4.2.4.1. Diffusion rate
It is of importance to determine how long it will take for the hydrogen
concentration to reach equilibrium in the palladium. For this information, it is necessary
to have the diffusion rate, D. This rate has also been experimentally found for steady
state permeation rates and is 5.4×10
7
cm
2
sec
1
at a temperature of 25ºC for diffusion of
hydrogen into palladium (Alefeld and Völkl, 1978a, 1978b). A graphical representation
was developed by Lewis (1967) to illustrate the permeation rates and importance of flow
at higher temperatures, Fig. 4.4.
Since the temperature at which the experiment should be performed at is around
20ºC and with a static flow of hydrogen, the permeation rate will be fairly high according
to Fig. 4.4. This is what is desired as it also simplifies the experiment by not having to
set up mass flow controllers.
Concentration Pd (H/Pd)
10% 0.667
1% 0.621
0.10% 0.006
0.0001% 6.21E05
1 (H/Pd) = ~1.25·10
3
(ccH
2
/ccPd)
Atomic ratio conversions
1 (H/Pd) = 1.06·10
3
(mgH
2
/100g)
1 (H/Pd) = ~9.5·10
3
(ccH
2
/gPd)
76
Fig. 4.4. Permeation rates of hydrogen with
circulating or static flows through a membrane of
Pd with a high pressure on one side (Lewis, 1967).
The diffusion rate of a sample concentration can be analytically found using
Fick’s second law of diffusion (Streeter et al., 1998; Hu et al., 2001)
2
2
,
c c
D
t x
∂ ∂
=
∂ ∂
(4.11)
where c is the concentration and D is the diffusion rate. Fick’s second law applies to
systems that are not operating in an infinite volume condition, as is the case for this
Thesis. If the boundary conditions of the experimental setup are implemented, including
Sievert’s Law of square root proportionality (Hughes and Bastasz, 1988; Hu et al., 2001),
the equation has a solution of the form
77
,
2
gas
gas o
c c
x
erf
c c Dt
−
 
=

−
\ .
(4.12)
where c
gas
is the concentration of the gas at the surface of the absorbing layer (in this case
palladium), c
o
is the initial concentration within the layer, c is the concentration at a
certain depth, x, and t is time required for the sample at depth x to reach a concentration
of c. The term erf stands for the error function, which is a Gaussian error function that is
tabulated in mathematical tables (Spiegel and Liu, 1999). If the initial concentration of
the palladium is zero then Eq. 4.12 simplifies to
1 ,
2
gas
x
c c erf
Dt
(
 
= −
( 
\ .
¸ ¸
(4.13)
where the error function is always a constant and thus the argument within the error
function is constant. Therefore, Eq. 4.13 can be solved to show that the depth variable, x,
is proportional to the square of the time, i.e.,
( )
. . 2 .
2
x
Const x Const D t x t
Dt
 
= → = ⋅ → ∝

\ .
(4.14)
Equation 4.14 shows that the thickness of the polymer layer (palladium) will only
affect the length of time required for the layer to reach equilibrium. As was stated in
Section 3.3, the desired thickness for conducting experiments shall be 1000 Ǻ and 2000
Ǻ. For the analytical calculations in Sections 4.2.5, a palladium thickness of 1000 Ǻ will
be used.
78
4.2.4.2. Volume increase
Because the hydrogen will be absorbing into the palladium interstitially, there will
be an increase in the lattice structure size. Palladium metal has a lattice constant of 3.889
Å and can increase up to 3.893 Å with hydrogen absorption. This increase varies with
the concentration. There currently are limited data on this lattice increase, but a graph
was compiled over the years that displays several experimental data points from many
different previous experiments and fits a line accordingly, Fig. 4.5 (Alefeld and Völkl,
1978a, 1978b).
From Fig. 4.5 it is possible to estimate how much the volume will be affected at
any concentration of hydrogen. It is interesting to note, though, that the slope of the line
is fairly steep at 1/5 so that when the concentration is at 1% or the atomic ratio is 0.621
H/Pd, the volume change will be approximately 12.4%. It was assumed that the data of
Fig. 4.5 were accurate in that there is no uncertainty in the values generated from it. In
order to account for this volume change, which can affect the modulus of elasticity, it
should be included in the analytical calculations where possible. However, because the
spring constant is dependent on individual geometric values (length, width, height)
instead of volume and it cannot be simplified, this change in volume is assumed to affect
only the height (thickness) of the cantilever. This will lead to results that will vary
slightly from the true value.
79
Fig. 4.5. Experimental data fitted with a line to approximate
the amount of volumetric change in palladium due to hydrogen
absorption (Alefeld and Völkl, 1978a, 1978b).
4.2.4.3. Change in modulus of elasticity
While there are changes in the lattice volume, the absorption of the hydrogen in
palladium will also affect the stiffness (and resonant frequency) because of changes in the
modulus of elasticity. It has been documented and shown that the modulus of elasticity
decreases as hydrogen is absorbed into the palladium (Lewis, 1967). This decrease is
shown in Fig. 4.6, which was estimated from experimentally determined data by Krüger
80
and Jungnitz (1936) at ambient room temperature (~20ºC), which is the temperature
condition for this Thesis.
Fig. 4.6. Estimated results from experimental data illustrating
the change that occurs in the modulus of elasticity during
absorption of hydrogen into palladium (Lewis, 1967).
From Fig. 4.6 it is clear that it will be necessary to include a changing modulus of
elasticity in the analytical calculations in order to maintain accuracy. When the atomic
ratio (H/Pd) is at around 0.621, the modulus of elasticity will decrease by about 10%,
which is a considerable amount. Once again, the data from Fig. 4.6 were assumed as
accurate and when conducting any calculations, no uncertainty was considered for it.
81
However, this change in modulus was only implemented in the analytical calculations
when the spring constant was considered a variable. The condition for incorporating this
changing stiffness will be described in more detail in Section 4.2.5.
4.2.5. Resonant frequency and frequency shift
In order to determine the amount of mass collected, the fundamental frequency
must first be determined. Starting with the general equation for a vibrating body (Rao,
1995) we get
2
2
( )
( , ) ,
d y t dy
F y t m c ky
dt dt
= + + (4.15)
where if damping, c, and external forces, F(y,t), are not included we get a function in
terms of dynamic, or effective, mass, m, and stiffness, k, as
2
2
( )
0 .
d y t
m ky
dt
= + (4.16)
The homogeneous differential equation, represented by Eq. 4.16, can be solved
for the fundamental frequency of vibration to obtain (Pryputniewicz, 2002b)
1
,
2
k
f
m π
= (4.17)
where k is the spring constant for the cantilever and m is its effective or dynamic mass.
The effective mass was calculated for a cantilever and is 33/140, or about 0.24, of the
total (i.e., static) mass for the cantilever. As Eq. 4.17 shows, any change in mass will
82
change the frequency of the cantilever, subject to modifications of the spring constant
(Furlong and Pryputniewicz, 1995). This equation was used for determining frequency of
the cantilever for the simplified case, discussed later in this section, which will be
compared against computational and experimental data. Uncertainty calculations will
also be conducted on this frequency, which will be described in more detail in Section
4.2.11. It should be noted that the calculation of the dynamic mass and all other
equations can be found in Appendix B.
The dynamic mass written in its explicit form is
33
,
140
m bhL ρ = (4.18)
where ρ is the density of the cantilever, while b, h, and L are the geometric dimensions of
the cantilever, as shown in Fig. 4.7.
Fig. 4.7. Representation of cantilever with geometric dimensions and a cross section
view through the middle with each of the layers of the cantilever labeled.
Several different cases of the resonant frequency shall be looked at in the
calculations. The first case will look at the simplest scenario where the only variable that
83
changes during the absorption process is the mass of palladium, as it increases. If the
absorption of hydrogen does not change the spring constant of the palladium, the
equation for calculating the frequency shift due to absorbed mass can be written as
2 2
1 2
2
2
,
f f m
f m
− ∆
= (4.19)
where f
i=1,2
are the initial and final frequencies of vibration, respectively, m the dynamic
mass of the cantilever, and ∆m the absorbed mass accumulated causing f
1
to change to f
2
.
Equation 4.19 can be rewritten in terms of the final frequency, f
2
. Once that is
determined, it is possible to calculate the shift in the frequency due to the absorption of
hydrogen at any given concentration. In this Thesis, 1% hydrogen (in a 1 atm pressure
nitrogenhydrogen mix) will be used as the concentration of hydrogen for the
calculations.
In this Thesis, in order to determine frequency shift, the gas concentration will be
1% hydrogen in a 99% nitrogen mix at a total of 1 atm pressure and 20˚C and the
palladium coating will be 1000 Å thick.
4.2.6. Spring constant consideration
Every oscillating, or vibrating, body has a parameter known as the spring
constant, or stiffness, k. The spring constant is the value that relates to the elastic energy
that can be stored and released in a body for a given distance. It is defined as a ratio of
force and the corresponding displacement, i.e., (Young and Freedman, 2000)
84
,
F
k
x
= (4.20)
where k is the spring constant, F is the force acting on the spring, and x is the
displacement of the spring. It is possible to simplify the spring constant into a set of
variables that pertain to the characteristics of the cantilever. Using the cantilever
boundary conditions of the sensing element, the spring constant can be shown to be
(Pryputniewicz, 2002b)
3
3
,
4
Ebh
k
L
= (4.21)
where E is the modulus of elasticity, while b, h, and L are the width, thickness, and length
of the cantilever, respectively, Fig. 4.7. Equation 4.21 shows how important the
structural dimensions of the cantilever are to the spring constant. Also, since the
thickness and length are cubed, they significantly affect the value of k.
A varying spring constant is one variable that is usually not taken into
consideration in studies dealing with cantilever type MEMS. Most publications simply
calculate the static spring constant of the polymercoated cantilever. However, it is
important to note that as the palladium absorbs the hydrogen, the spring constant of the
cantilever changes (Cherian and Thundat, 2002; Jericho and Jericho, 2002; Gibson et al.,
2001). This dynamically changing spring constant will be included in the calculations.
In Eq. 4.19, a change in frequency is determined based on a change in mass; this equation
will have to be modified, however, if the absorbed mass changes the spring constant. The
new equation where the spring constant changes as the mass changes is derived in
Appendix B, i.e.,
85
2
2 1
1
,
1
k
k
f f
m
m
∆  
+

\ .
=
∆  
+

\ .
(4.22)
where ∆k and ∆m are the changes in the spring constant and effective mass, respectively,
and k and m are the initial spring constant and effective mass, respectively. Substituting
Eqs 4.18 and 4.21 into Eq. 4.17, the initial frequency f
1
can also be expressed as
1
2
1 35
.
2 33
h E
f
L π ρ
= (4.23)
If the spring constant value alters, it will be by a change of the parameters of Eq.
4.21 that, in turn, will affect Eq. 4.23. A change in stiffness will come from either a
change of volume, a change of the modulus of elasticity, or a combination of both. Since
it was shown in Sections 4.2.4.2 and 4.2.4.3, that the modulus of elasticity and the
volume change with the absorption of hydrogen it is reasonable to state that there will be
changes in the spring constant, which depend on these parameters. Stating again,
however, the increase in the palladium volume will be considered to affect the thickness
and, in turn, stiffness. The results of Section 5.11.5 will show that this is a reasonable
argument.
4.2.7. Damping coefficient consideration
Damping is the phenomenon that reduces/absorbs energy of a vibrating system
over a period of time. There is damping everywhere in the real world. It is present in all
86
applications with an exception of systems vibrating in vacuum environment. Damping
helps bring everything to an equilibrium state and maintain it in a static condition.
When doing measurements and calculations, damping plays an important part in
order to get accurate results. It is necessary to understand when and how damping will
affect a given system. Even though the cantilever sensor is only on the order of 300 µm
long, damping will alter results if it is not noticed and considered (Rast et al., 2000).
Only if the cantilever is in vacuum there will be no damping due to the surrounding
medium. However, when the analyte gas is released, it itself will damp the cantilever.
The most common place that damping occurs is in oscillating or vibrating bodies.
Damping itself is a force dependent process that can be expressed as (Young and
Freedman, 2000)
,
F
c
v
= − (4.24)
where F is the force acting upon the body, v is the velocity of the body, and c is the
damping coefficient, which for a specific set of parameters defining a vibrating body and
its environment is a constant that describes the “strength” of the damping force. The
negative sign in Eq. 4.24 indicates that the damping force acts against the body and
attempts to reduce the total energy acting upon it.
The damping coefficient can also be written as a function of several parameters
defining the damping fluid and the volume that it occupies, i.e.,
2
,
Lb
c
H
µ
= (4.25)
87
where µ is the dynamic viscosity of the damping fluid, L and b are the length and width
of the cantilever, respectively, and H is the nominal thickness of the fluid film that the
object moves through. There are two items of importance to note when examining Eq.
4.25. One is that the thickness of the cantilever is not a variable in the relationship
defining the damping coefficient. The other is the obvious relation of how the viscosity
of a fluid alters the damping. This will play a role in the quality factor, which will be
discussed in Section 4.2.8. For this Thesis, the damping coefficient will be calculated
analytically and then it will be experimentally verified in order to ensure that the
frequency shift calculations are being properly represented. The experimental
determination of the damping coefficient will be discussed in Section 4.4.4.
4.2.8. Free vibration system
If we look at Eq. 4.15 defining motion of a free vibration system, based on
discussion in Section 4.2.5, i.e.,
2
2
( ) ( )
0 ,
d y t dy t
m c ky
dt dt
= + + (4.26)
where there are no external forces, it demonstrates that damping and the spring constant
need to be considered in any calculations for the free vibration system. If this
homogeneous equation is solved for an underdamped system characterized by where
(Young and Freedman, 2000)
88
2
2
1 ,
2
c
m
ω
 
− <

\ .
(4.27)
with ω being the angular frequency, the solution will be
( )
( )
/ 2
( ) cos ,
c m t
d
y t Ae t ω θ
−
= + (4.28)
where A is the amplitude of the vibration, t is time, and θ is the phase. The ω
d
is the
damped angular frequency defined by the equation
2
2
,
4
d
k c
m m
ω = − (4.29)
which clearly shows that damping decreases the angular frequency; derivation of Eq.
4.29 can be found in Appendix B. While damping affects the frequency, adding mass
affects the damping. Angular frequency is related to cyclic frequency by the following
equality:
.
2
f
ω
π
= (4.30)
Inserting Eq. 4.29 into Eq. 4.30 and simplifying, a relationship for detecting
frequency shifts, similar to Eq. 4.19, but include damping can be showed to be
(derivation found in Appendix B)
( )
( )
2
2
2 1
2
4
,
4
m m mk mc
f f
k m m
δ
δ
+ −
=
+
(4.31)
where f
1
is the initial frequency and other parameters are as previously defined. Using
Eq. 4.31, the frequency shift in the presence of damping was calculated analytically.
89
4.2.9. Changing stiffness in a free vibration system
Sections 4.2.6 and 4.2.8 described effects that a changing stiffness and damping,
respectively, have on the frequency shift calculations. To further improve the analytical
model, both parameters will be considered together in the same equation. Starting with
Eq. 4.29, the equation takes the form of
( )
( )( )
2
2
2 1 2
4
,
4
m m mk mc
f f
k k m m
δ
δ δ
+ −
=
+ +
(4.32)
which is very similar to Eq. 4.31. Equation 4.32 was used to determine the frequency
shift that occurs when both the stiffness and damping changes are considered
simultaneously.
4.2.10. Harmonically excited system
To model the cantilever system completely, it will be required to take into
account, in addition to the parameters mentioned in Section 4.2.9, the excitation force
needed to get it into its resonance. Using the general equation for a body in motion, Eq.
4.15, it is possible to calculate the maximum amplitude. The steady state solution for the
secondorder nonhomogeneous differential equation is the particular solution, or (Young
and Freedman, 2000)
( ) cos( ) , y t A t ω θ = − (4.33)
90
where ω is the angular frequency of excitation, A is the vibration amplitude and has the
solution of (Young and Freedman, 2000)
( ) ( )
max
2
2
2
1 2
F
A
k r r ζ
=
− +
(4.34)
with F
max
being the maximum force applied to the vibrating body, and ζ being the
damping ratio. The parameter r is defined as the frequency ratio
,
n
r
ω
ω
 
=

\ .
(4.35)
where ω
n
is the natural frequency of the body and the solution for the phase, θ, in Eq.
4.33 is (Young and Freedman, 2000)
1
2
2
tan .
1
r
r
ζ
θ
−
 
=

−
\ .
(4.36)
Looking at the parameters of Eq. 4.35 we see that the driving force will vibrate at
the natural frequency and thus the entire body should be vibrating at this frequency.
When this frequency is the same as the resonance, or natural, frequency, the frequency
ratio will be equal to one and the amplitude will be at its maximum, Fig. 4.8, which is
exactly what is desired when attempting to detect frequencies. The amplitude ratio
shown in Fig. 4.8 is equal to Ak/F
max
so that the displacement is unitless. The damping
ratio,
,
2
c n
c c
c m
ζ
ω
= = (4.37)
is the relationship between the damping, c, and the critical damping, c
c
. The critical
damping coefficient is the magnitude of damping required to make vibrations decrease
91
the quickest, or in other words, reduce the amount of energy in the system and also
prevent any subsequent amplitudes.
The equipment (vibrometer and interferometric methods) used in detecting the
frequency shift, however, will be vibrating the cantilevers over a range of frequencies in
order to see the resonant frequency curve peak. Because the frequency ratio will always
be equal to one (once the resonance frequency is reached), the system can be modeled
with the equation for frequency shift of a free vibration system.
Fig. 4.8. Vibration amplitude as a function of frequency ratio for
different values of damping ratio.
92
4.2.11. Uncertainty analysis
For each case considered in this Thesis, there was an associated overall
uncertainty determined. These uncertainties indicate how “good” the results are and what
contribution each parameter has to the specific overall uncertainty. The derivation for
each case is located in Appendix B, but for illustration purposes, only one case shall be
looked at in this section.
For the uncertainty analysis conducted in this Thesis, the RSS (rootsumsquares)
method shall be used (Pryputniewicz, 1993). The first part of the RSS method is to write
the equation that is going to be analyzed in its phenomenological form, or in an equation
that lists all independent parameters of Eq. 4.23 as
( )
1
, , , , f h L E ρ = (4.38)
where it can be seen that the independent parameters of the equation for frequency are
thickness, h, length, L, modulus of elasticity, E, and density, ρ. Based on Eq. 4.38, the
overall uncertainty in frequency, δf
1
, is
1
2 2 2 2 2
1 1 1 1
1
,
f f f f
f h L E
h L E
δ δ δ δ δρ
ρ
(
  ∂ ∂ ∂ ∂      
= + + + (
   
∂ ∂ ∂ ∂
\ . \ . \ .
( \ .
¸
¸
(4.39)
where δh, δL, δE, and δρ are the uncertainties in the specified parameters. Using Eqs
4.38 and 4.39 it is possible to calculate the percentage overall uncertainty in the
resonance frequency, %δf
1
, as
1
1
1
% 100 .
f
f
f
δ
δ
 
=

\ .
(4.40)
93
It is also possible to calculate the percent contributions of the uncertainties in the
independent parameters to the overall uncertainty by using the following form:
2
1
1
2
1
% 100 ,
f
h
h
f h
f
δ
δ δ
δ
∂  

∂
\ .
= ⋅ (4.41)
where %δfδh is the percent contribution of the uncertainty in thickness to the overall
uncertainty in the frequency. Equation 4.41 can be applied to each parameter to
determine the contributions of each of them to the overall uncertainty. This is important
in order to better understand how reliable the frequency data are and to determine where
any sources of errors come from. Results of uncertainty analysis for each case presented
in Sections 4.2.5 through 4.2.10 are summarized in Appendix C.
4.3. Computational analysis
The ACES methodology requires thorough work with computational analysis.
The computational analysis helped visualize the effects of vibration on the cantilever.
Using finite element method (FEM) software, a model of the cantilever was developed.
For this Thesis, Pro/Engineer was used to model the cantilevers using data gathered from
measurements of the cantilever (PTC, 2003a). The software Pro/Mechanica, a FEM
package, was used for the modal analysis (PTC, 2003b). Using Pro/Mechanica, the
resonance frequencies of a cantilever were deteremined based on the model developed in
Pro/Engineer. For the computational solutions, two cases were considered. The first
94
utilized ideal dimensional values, such as those specified by the manufacturer, and the
second used the averaged values based on measurements of the cantilever. Once the
initial results from Pro/Mechanica were obtained, they were checked to assure their
validity, subject to the input information. This check looks at the convergence of the
solution. The solution should converge to within at least of 1% to be acceptable. The
convergence percent is the amount of difference between two most recent solution points.
When the difference between the two calculated values is 1% or less, the solution is said
to have converged.
4.4. Experimental solutions
ACES methodology requires gathering experiment data. This step was conducted
with the cantilevers that were procured from MikroMasch USA as described in Chapter
3. For the analytical calculations, 6 chips (each with 3 cantilevers) were fully measured
using SEM, vibrometer, AFM, and OELIM systems. Increasing the number of methods
used decreases errors in results. For this Thesis, 6 chips were taken as the sample set.
The experimental results are compared in Section 5.6.
95
4.4.1. Instrument calibration
To properly conduct experiments, it is essential that all of the testing equipment
was functioning and calibrated correctly. All areas of sources of errors must be reduced
to a minimum to ensure quality data. Calibration of instruments is crucial to collecting
valid data. While seeming redundant, it is also important to make sure that each piece of
equipment is working properly.
4.4.2. Geometrical characterization of the cantilevers
To be able to characterize dynamics of the cantilevers used in this Thesis, it was
important to quantitatively determine their dimensions (Pryputniewicz, 1993; Yang et al.,
2000; Pryputniewicz et al., 2002b). Using an optical microscope, geometry of the
cantilevers was measured. This verified quality of fabrication and aided in the analytical
calculations and the computational modeling.
Measurements of dimensions of the cantilevers were made using a measurescope
equipped microscope capable of providing dimensions with the least count of 0.5 µm
(Pryputniewicz, 1993). Each cantilever was measured and all the results were correlated.
Since the free ends of the cantilevers were not square, it was necessary to calculate an
equivalent length that compensated for the irregularly shaped tip based on the
measurements that were taken. With the equivalent length for each cantilever the
analytical calculations were performed.
96
4.4.3. SEM – scanning electron microscope
To inspect and verify the thickness of the cantilevers a scanning electron
microscope (SEM) was used, Fig. 4.9. An optical microscope can easily view in the xy
plane, but because of the geometric configuration of the chips used the thickness or zaxis
of the cantilevers was not easily measurable and thus had to be determined using the
SEM.
Fig. 4.9. SEM used for measuring thickness of the cantilevers. The samples
were inserted into the chamber on the left (bottom of the black tube) and the
microscope was controlled with the console on the right.
97
4.4.4. Determination of quality factor and damping coefficient
The quality factor (Qvalue, or Q) is an important parameter of a vibrating body.
It describes the “quality” or the width of the resonance frequency peak. The narrower the
resonance frequency peak, the easier and more distinct the peak is to detect. The quality
factor is dependent on the damping and, since the damping plays an important role in the
calculations, it is necessary to measure it. In the opinion of this author, it is appropriate
to discuss a procedure for the determination of Q, at this point in the chapter addressing
methodology, because Q is a function of thickness, a method for determination of which
was outlined in Section 4.4.3. The general relationship for the quality factor can be
written as
,
n n
m
Q
c bandwidth
ω ω
= = (4.42)
where ω
n
is the angular frequency, m is the dynamic mass of body, and c is the damping
coefficient. The bandwidth, appearing in Eq. 4.42, is the width of the frequency curve,
Fig. 4.10, at half power point. The half power point is defined to be
2
A
(4.43)
of the peak value of the resonance frequency curve, where A is the maximum amplitude
of the vibration at a specific frequency (Rao, 1995). Once the bandwidth is known, it is
possible to calculate what the damping coefficient is using the identity
.
c
bandwidth
m
= (4.44)
98
The result can then be compared to the analytically determined damping
coefficient in Eq. 4.25.
Fig. 4.10. A comparison of how different resonant frequency response curves
affect the width of the bandwidth, which in turn will affect Eq. 4.42.
The method described above also allows for determination of the H value, or the
nominal thickness of the fluid film that the object moves through, as used in Eq. 4.25.
The value of H is basically the maximum displacement of the cantilever at resonant
frequency. For detecting the resonant frequency curve a laser vibrometer was used, as
discussed in Section 4.4.7.
4.4.5. Optoelectronic methodology
As mentioned in the beginning of Section 4.4, an interferometric setup was used
for fullfield of view identification of the resonant frequencies of the cantilever samples.
99
Specifically, an optoelectronic laser interferometric microscope (OELIM) was used. The
OELIM measured frequency values were then compared to the analytically and
computationally calculated values to determine the degree of correlation between them.
Optoelectronic methodology, as presented in this paper, is based on the principles
of optoelectronic holography (OEH) (Brown, 1999; Furlong, 1999; Furlong and
Pryputniewicz, 2002; Pryputniewicz et al., 2000). Basic configuration of the OEH
system is shown in Fig. 4.11. In this configuration, laser light is launched into a single
mode optical fiber by means of a microscope objective (MO). Then, the single mode
fiber is coupled into two fibers by means of a fiber optic directional coupler (DC). One
of the optical fibers comprising the DC is used to illuminate the object, while the output
from the other fiber provides reference against which the signals from the object are
recorded. Both, the object and reference beams are combined by the interferometer (IT)
and recorded by the system camera (CCD).
Images recorded by the CCD are processed by the imageprocessing computer
(IP) to determine the fringelocus function, Ω, constant values of which define fringe loci
on the surface of object under investigation. The values of Ω relate to the system
geometry and the unknown vector L, defining deformations, via the relationship
(Pryputniewicz, 1995a)
( )
2 1
Ω = − • = • K K L K L , (4.45)
where K is the sensitivity vector defined in terms of vectors K
1
and K
2
identifying
directions of illumination and observation, respectively, in the OEH system, Fig. 4.11.
100
Fig. 4.11. Singleillumination and singleobservation geometry of a fiber optic based
OEH system: LDD is the laser diode driver, LD is the laser diode, OI is the optical
isolator, MO is the microscope objective, DC is the fiber optic directional coupler, PZT
1
and PZT
2
are the piezoelectric fiber optic modulators, IP is the imageprocessing
computer, IT is the interferometer, OL is the objective lens, CCD is the camera, while K
1
and K
2
are the directions of illumination and observation, respectively (Furlong and
Pryputniewicz, 2000).
Quantitative determination of structural deformations due to the applied loads can
be obtained, by solving a system of equations similar to Eq. 4.45, to yield
(Pryputniewicz, 1995a)
  ( ) ,
~ ~ ~
T
1
T
Ω =
−
K K K L
(4.46)
where
T
~
K represents the transpose of the matrix of the sensitivity vectors K.
Equation 4.45 indicates that deformations determined from interferograms are
functions of K and Ω, which have spatial, i.e., (x,y,z), distributions over the field of
interest on the object being investigated. Equation 4.45 can be represented by a
phenomenological equation (Pryputniewicz, 1993)
101
( ) , , Ω = K L L
(4.47)
based on which the RSStype (where RSS represents the square root of the sum of the
squares) overall uncertainty in L, i.e., δL, can be determined to be
,
2 / 1
2 2
(
(
¸
(
¸

.

\

Ω
Ω ∂
∂
+ 
.

\

∂
∂
= δ δ δ
L
K
K
L
L
(4.48)
where K L ∂ ∂ / and Ω ∂ ∂ / L represent partial derivatives of L with respect to K and Ω,
respectively, while δK and δΩ represent the uncertainties in K and Ω, respectively. It
should be remembered that K, L, and Ω are functions of spatial coordinates (x,y,z), i.e.,
K= K(x,y,z), L = L(x,y,z), and Ω = Ω(x,y,z), respectively, when performing partial
differentiations. After evaluating, Eq. 4.48 indicates that δL is proportional to the
product of the local value of L with the RSS value of the ratios of the uncertainties in K
and Ω to their corresponding local values, i.e.,
.
2 / 1
2 2
(
(
¸
(
¸

.

\

Ω
Ω
+ 
.

\

∝
δ δ
δ
K
K
L L
(4.49)
For typical geometries of the OEH systems used in recording of interferograms,
the values of δK/K are less than 0.01. However, for small deformations, the typical
values of δΩ/Ω are about one order of magnitude greater than the values for δK/K.
Therefore, the accuracy with which the fringe orders are determined influences the
accuracy in the overall determination of deformations (Pryputniewicz, 1981). To
minimize this influence, a number of algorithms for determination of Ω were developed.
Some of these algorithms require multiple recordings of each of the two states, in the case
102
of doubleexposure method, of the object being investigated with introduction of a
discrete phase step between the recordings (Furlong, 1999; Furlong and Pryputniewicz,
2000; Pryputniewicz, 1995b).
For example, the intensity patterns of the first and the second exposures, I
n
(x,y)
and I΄
n
(x,y), respectively, in the doubleexposure sequence can be represented by the
following equations:
( ) ( ) ( ) ( ) ( )
{ }
( ) ( )
{ }
1/ 2
, , , 2 , ,
cos , ,
n o r o r
o r n
I x y I x y I x y I x y I x y
x y x y ϕ ϕ θ
( ( = + + •
¸ ¸ ¸ ¸
( − +
¸ ¸
(4.50)
( ) ( ) ( ) ( ) ( )
{ }
( ) ( ) ( )
{ }
1/ 2
'
, , , 2 , ,
cos , , , ,
n o r o r
o r n
I x y I x y I x y I x y I x y
x y x y x y ϕ ϕ θ
( ( = + + •
¸ ¸ ¸ ¸
( − + + Ω
¸ ¸
(4.51)
where I
o
and I
r
denote the object and reference beam irradiances, respectively, with (x,y)
denoting spatial coordinates, φ
o
denotes random phase of the light reflected from the
object, φ
r
denotes the phase of the reference beam, θ
n
denotes the applied nth phase step,
and Ω is the fringelocus function relating to the deformations the object incurred
between the first and the second exposures; Ω is what we need to determine. When Ω is
known, it is used in Eq. 4.46 to find L.
In the case of 5phasesteps algorithm with θ
n
= 0, π/2, π, 3π /2, and 2π the
distribution of the values of Ω can be determined using
( )
( ) ( )
( ) ( ) ( )
2 4
1
3 1 5
2 , ,
, tan .
2 , , ,
I x y I x y
x y
I x y I x y I x y
−
¦ ¹ ( −
¦ ¦
¸ ¸
Ω =
´ `
− −
¦ ¦
¹ )
(4.52)
103
Results produced by Eq. 4.52 depend on the capabilities of the illumination,
imaging, and processing subsystems of the OEH system. Developments in laser, fiber
optic, CCD camera, and computer technologies have led to advances in the OEH
metrology.
In response to the needs of the emerging MEMS technology, an optoelectronic
laser interferometric microscope (OELIM) methodology for studies of objects with
micron size features was developed
(Brown, 1999; Brown and Pryputniewicz, 1998,
2000; Furlong and Pryputniewicz, 2000; Pryputniewicz, 1985b, 1989, 1991). In the
OELIM system, Fig. 4.12, the light beam produced by a coherent light source is directed
through a collimating optics. The resulting light field is then divided into reference and
object beams by the splitter. The reference beam is directed towards a mirror and back to
the beam splitter. The object beam is directed towards the PZT actuated MEMS under
study by a long working distance microscope objective and is reflected back to the beam
splitter. The two beams recombine at the beam splitter and are imaged onto the sensing
element of the CCD camera, which records the resulting interference patterns. These
patterns are transferred to the system computer for subsequent quantitative processing.
Using the systems shown in Fig. 4.12, issues relating to the sensitivity, accuracy,
and precision, associated with application of the algorithm defined by Eq. 4.51, were
studied while evaluating the effects that the use of highspatial and highdigital resolution
cameras would have on the results produced
(Furlong et al., 2002). Mode shape
characteristics of vibrating MEMS cantilevers were determined employing previously
104
developed procedures (Brown and Pryputniewicz, 1992, 1995; Oh and Pryputniewicz,
1990; Pryputniewicz, 1988, 1990, 1991).
Fig. 4.12. Optical configuration of the OELIM
system (Furlong and Pryputniewicz, 2000).
4.4.6. Determination of resonant frequency by AFM
In addition to using the OELIM methodology, resonance frequencies of vibrating
cantilevers were also determined using an atomic force microscope (AFM) in order to
have two independent means of quantifying resonance characteristics of the MEMS
cantilevers used in this Thesis. The AFM, shown in Fig. 4.13, is capable of automatically
measuring the resonance frequencies of a cantilever by scanning a large frequency range
105
and detecting where the largest amplitudes occur. It provides quick and accurate results.
However, to facilitate determination of frequencies by AFM, it is necessary to know
approximately where the frequencies of intent are to avoid detection of harmonics.
Fig. 4.13. AFM setup utilized: (a) overall view of the setup showing measurement and
control subsystems, (b) close up of the measurement subsystem with protective lid open.
In an AFM system, a laser beam is focused and reflected off a cantilever and into
a position sensitive diode (PSD) as illustrated in Fig. 2.14. As the cantilever vibrates, the
angle of the deflection changes and position of the laser beam on the PSD moves. This
motion is measured and used to determine a displacement curve. Wherever there is a
peak in the curve, a resonant frequency is located there.
4.4.7. Laser vibrometry
In this Thesis, laser vibrometry was used to measure damping coefficients of the
vibrating MEMS cantilevers. The vibrometer functions by way of using a laser and a
(a) (b)
106
MachZehnder interferometric setup built within it as well as its capability to measure
Doppler frequency shifts (Polytec, 2003). It functions by examining a signal from an
object on a pointbypoint basis to measure velocity and displacement as a function of
position on the object. Once those two parameters are measured, it is possible to
calculate frequency of the vibrating object with an accuracy of 0.001 Hz. This
vibrometer would be used to measure small shifts in frequency of a vibrating cantilever in
the MEMS chemical sensor.
The cantilevers were placed on a custom made PZT (PbZnTi) transducer, Fig.
4.14, which is capable of excitation frequencies in the MHz range containing the
frequencies of intent in this Thesis.
Fig. 4.14. A comparison of the PZT used with a MEMS
chip attached to the top and a dime.
MEMS
chip
107
4.4.8. Test chamber
The PZT transducer would be contained within a custom made test chamber, Fig.
4.15. The design of the chamber can be found in Appendix D. The chamber will allow
the cantilevers to be in a gaseous environment that can be controlled. This chamber
would have the gas sample flow into it and out via inputs and outputs. A glass window
on the top lid would allow the vibrometer to make unobstructed measurements of
frequencies of the cantilevers. For simplification of the setup, the gas would not flow
through the chamber, but would be pumped into one atmosphere pressure and then
sealed. Because of the volume within the chamber compared to the amount of palladium
present, any effects of reduced partial pressure of hydrogen due to absorption into the
palladium will be assumed negligible. Any changes would be recorded and then the gas
would be released and the neutral carrier gas, nitrogen, would purge the chamber. To
regulate concentration of gas, mass flow controllers would be used to adjust the gas
mixture ratio, Fig. 4.15.
The entire chamber along with an aligning microscope would be contained within
a hood of the vented space. This would ensure that any gas that escapes or is purged
would be safely carried away. The supply gas tanks would not be required to be within
the hood as they are sealed systems in themselves, but can be.
Because of space limitations, the entire setup would have to be as compact as
possible. Figure 4.16 illustrates the setup that would allow for the recording of shifts in
the frequency of vibrating cantilevers in a gaseous mixture containing H
2
. The aligning
microscope would be equipped with a CCD camera so that the cantilevers can be viewed
108
outside of the hood. Because of the remote measurement capability of the vibrometer, it
will be possible to place it outside of the hood.
Fig. 4.15. The custom made test chamber containing the PZT
transducer, which is to be used for testing functional operation
of the cantileversensing element in a gaseous environment.
Fig. 4.16. Laboratory setup for testing functional operation of MEMS chemical
sensors; the microscope and test chamber will be contained within a vented hood.
109
4.5. Data analysis
Once all of the results of the ACES methodology were collected, they were
analyzed in order to determine a set of parameters that would optimize performance
characteristics of the MEMS cantilevers. In the analysis, several correlations between
different types of data were made.
The first calculations were made of the analytical frequencies for each cantilever
based on the measured dimensional data. These results were tabulated, averaged, and are
presented in Section 5.1. The averaged data was then used to calculate the contributions
that uncertainties in each independent parameter have on the overall percentage
uncertainty of the dependent parameter, as discussed in Section 4.2.11.
Two sets of computational results were then compared to the analytically
equivalent cases, as is discussed in Section 5.5. The percentage difference was calculated
and correlation was determined.
Once all the analytical frequencies for each cantilever were calculated, it was
possible to correlate these values with the experimentally determined results. Each
analytically determined cantilever frequency was compared to the AFM and
interferometric data individually. All the percent differences were then averaged and
compared to the overall percentage uncertainty for the analytical case, as is discussed in
Section 5.6.
For analytical calculations dealing with the coated cantilevers, several results
were compared. Each case described in the Sections 4.2.5 through 4.2.10 discussing the
different approaches to the problem was analytically calculated for the longest cantilever.
110
The overall percentage uncertainty was calculated for each case and compared in Table
5.10. The results were analyzed and conclusions were drawn as discussed in Sections
5.11.6 and 5.11.7.
Once all the data were analyzed and conclusions drawn, comments and
recommendations were written for further investigations into cantilever sensors as
discussed in Chapter 6.
111
5. RESULTS
All of the analytical, computation, and experimental results generated in this
Thesis are presented in this chapter. The results are shown herein unless it is specified
that they are located in a specific appendix.
5.1. Microscope and SEM data
In the beginning of the analysis of the cantilevers, quantitative measurements
were made describing all of the geometric dimensions. The sample size was 6 chips,
each with 3 cantilevers apiece. The physical dimensions of the cantilevers were
measured using an optical microscope and a SEM. The microscope used was a Nikon
Measurescope MM11, Fig. 5.1, which has the least count of 0.5 µm. A Nikon CoolPix
5000 with a 3.2 megapixel sensor was connected to the microscope for taking pictures of
each cantilever. Representative pictures of the cantilevers are shown in Fig. 3.2 and also
in Figs 5.2 to 5.5, where it can be seen that the free ends are not square. It was
discovered that the reason for this was due to the way the cantilevers were made. The
MicroMasch cantilevers with tips were fabricated with the ends as shown to facilitate
their use for the AFM applications. During the fabrication of tipless versions the same
masks were used in the process, for cost reasons, so the ends are shaped.
For the purposes of this Thesis, three of the longest cantilevers, i.e., D, E, and F,
per chip were measured for each of the six chips considered. All original measurements
for cantilever D are listed in Table 5.1. Based on the measurements, the shaped areas of
112
the cantilever tips were calculated and an equivalent length was found for a
corresponding “square” tip. This simplified the calculations. The average equivalent
total length of the cantilever D, Fig. 5.2, including its minimum and maximum values,
based on the original measurements listed in Table 5.1, are shown in Table 5.2, which
also shows other pertinent dimensions for the cantilever D as well as for the cantilevers E
and F.
Fig. 5.1. The microscope setup used for determining dimensions of the cantilevers. A
cantilever tip can be seen on the monitor in the middle while the monitor on the right
displays the measured values. The Nikon Measurescope MM11 is on the left.
It was found that the geometric dimensions of the cantilevers were generally
about 1% different (which for length amounted to about 2 to 3 µm) from the
manufacturer’s specifications. Because the cantilevers were made for atomic force
microscope, which is calibrated, the 1% difference does not affect the performance for
113
their originally intended use. However, the differences measured exceed the current
industry practice of 1 µm tolerance on length. The geometric dimensions are summarized
in Table 5.1 with a schematic guide shown in Fig. 5.2. The geometric parameters in Fig.
5.2 include the quantitative values measured and the equivalent calculated values for the
corresponding “square” tip cantilever. A complete list of the dimensions of each
cantilever can be found in Appendix E.
Fig. 5.2. Cantilever details and labelling: a) photograph of the entire cantilever, b) close
up on a tip of a cantilever, c) guide for the labels used for each geometric parameter of
the cantilever describing the data in Tables 5.1 and 5.2.
It should be noted that the sample cantilevers all had a specific designation
number according to the chip that they belonged too. This designation consists of a letter
and two digits. The letter, from A through F, designates the cantilever used on each chip,
differentiable by length with A designating the shortest and F designating the longest
(a)
(b) (c)
114
cantilever. The first digit designates the row number and the second number designated
the column starting with ‘11’ at the upper left corner of the package in which the
cantilevers were shipped and stored in, Fig. 5.3.
Fig. 5.3. Cantilevers packed in a GelPack as obtained from the manufacturer.
In addition to the traditional optical microscope, a scanning electron microscope
(SEM) was used. The SEM was specifically used to measure the thickness of each
cantilever because of its ability to see an object in “threedimensions” as in shown in Fig.
5.4. A few representative images of different samples of cantilevers are shown in Fig.
5.5. The values of the thickness from the SEM that were measured all tended to be
approximately 5% off from the manufacturer’s specifications of 1 µm. This 5% was
considered the uncertainty analysis.
Since all the cantilevers (D, E, and F) are each on the same chip, only the
thickness of the longest cantilever was determined assuming that the thickness will be the
same between them all. The thickness data collected were averaged and are displayed in
Table 5.2.
115
Table 5.1. Measured dimensional data and equivalent
lengths for all cantilever samples of the D type.
Fig. 5.4. SEM photographs of the cantilevers, with a progressive close up on the tip:
a) an overall view of three of the longer cantilevers (DF), b) closeup of cantilever
E, c) closeup of the tip of cantilever E showing the thickness and details of the tip.
Fig. 5.5. SEM photographs of E length cantilevers near the tip from different chips:
a) chip E11, b) chip E22, c) chip E32.
(a) (b) (c)
(a) (b) (c)
Measured D11 D12 D21 D22 D31 D32
characteristics Typical Typical Typical Typical Typical Typical
Length, L mm 296.5 295 293.5 296 295 295
Width, b mm 36 36.5 36.5 35 35 36
Thickness, h mm 1 1.007 1 0.96 0.989 0.956
End width, L1 mm 14 15 15 15 15 15
Eqv. length, L2 mm 8.938 9.106 8.118 7.7 7.757 8.573
Eqv. total length, mm 305.44 304.11 301.62 303.70 302.76 303.57
116
Table 5.2. Summary of measured dimensions for the cantilevers of the D, E, and F types.
5.2. Analytical results
Using Eq. 4.23 and measured dimensions of the cantilevers, the frequency for
each was calculated for the ideal case where they are uncoated, untreated, and assumed
composed of isotropic singlecrystal silicon. For example, taking the measurements
gathered from the microscope and SEM for the cantilever D11 and values of modulus of
elasticity and density from Table 5.3, we find that the analytical nominal frequency is
( )
( )
( )
( )
1 2
9
2
3
1 35
2 33
190 10
1
1 35
2 33 2.33 / 305.44
15, 865 .
h E
f
L
Pa
m
g cm m
Hz
π ρ
µ
π
µ
=
×
=
=
(5.1)
Following the procedure used to evaluate Eq. 5.1 the results for the remaining
cantilevers were determined and are summarized in Section 5.6. The idealized frequency
range that the uncoated cantilevers operate in is shown in Fig. 5.6.
Measured D cantilevers E cantilevers F cantilevers
parameters Min Average Max Min Average Max Min Average Max
Length, L µm 293.5 295.17 296.5 342.5 345.00 346 239.5 245.00 248
Width, b µm 35 35.83 36.5 34 35.00 36 33 35.25 37
Thickness, h µm 0.956 0.985 1.007 0.956 0.985 1.007 0.956 0.985 1.007
End length, L1 µm 9.00 10.08 11.00 9.00 9.83 11.50 8.00 9.58 11.00
Eqv. length, L2 µm 7.700 8.37 9.106 7.236 8.20 9.882 6.721 7.99 9.597
Ave. total length, µm
303.5 305.25 307.5 352 354.83 356.5 249.5 254.58 258
Ave. eq. tot. tength, µm
301.6 303.53 305.4 350.6 353.20 355.7 247.8 252.99 256.1
117
Table 5.3. Values of material properties of silicon (Madou, 2002).
To determine the uncertainty in the frequency results of the simplified case
(uncoated and untreated) the RSS method was used, Eqs 4.39 and 4.40. Using the
averaged equivalent values from Table 5.2 the percent overall uncertainty in resonance
frequency of each of the three cantilevers of different lengths was found, subject to the
uncertainty value for each parameter used is listed in Table 5.4. The percent overall
uncertainty was then averaged from these three values. This percent overall uncertainty
can now be compared to the computational and experimental data. The percent overall
uncertainties were determined based on values shown in, Table 5.4, (Pryputniewicz,
1993).
Table 5.4. The uncertainties in each parameter considered.
The averaged geometric data from Table 5.2 were used in determining the percent
overall uncertainty in the frequency of the cantilevers utilizing Eqs 4.39 and 4.40. These
results can be found in Appendix E. The percent overall uncertainty in frequency was
plotted in terms of length and thickness of the cantilevers and is shown in Figs 5.7 and
5.8. To determine which parameters (height, length, modulus of elasticity, and/or
δ h =0.05
µ
m δ L =5
µ
m δ E =3 GPa δρ =0.005 g/cm
3
Parametric uncertainty values
190
2.33
Modulus of elasticity, E, GPa
Values of material properties used
Density,
ρ
, g/cm
3
118
density) contributed to this uncertainty, each one was plotted as a function of cantilever
length and thickness to determine their percent contribution to the percent overall
uncertainty in the frequency, Figs 5.9 and 5.10 and from them; one can clearly see that in
both figures, the uncertainty in the thickness has the highest contribution to the percent
overall uncertainty in the frequency with 74.9% contribution in the 350 µm long
cantilever as an example. This is an important statement in that the thickness is indeed
the one parameter that has the highest relative uncertainty value of all of the independent
parameters in the frequency equation. If the plot of the percentage overall uncertainty in
frequency with respect to the uncertainty in thickness is generated, Fig. 5.11, the extent
that δh = ± 0.5 µm makes can be seen.
Fig. 5.6. Fundamental resonant frequency of a 1µm thick
cantilever as a function of active length.
119
Fig. 5.7. The percentage overall uncertainty in the cantilever as a function of against the
length while holding the thickness parameter at 1 µm to show the decreasing trend.
Fig. 5.8. The percentage overall uncertainty in the cantilever as a function of the
thickness while holding the length parameter as 350 µm to show the decreasing trend.
120
Fig. 5.9. Percent contributions by uncertainty of each independent parameter to the
overall uncertainty in resonance frequency of the cantilever, as a function of length.
Fig. 5.10. Percent contributions by uncertainty by each independent parameter to the
overall uncertainty in resonance frequency of the cantilever, as a function of thickness.
121
Fig. 5.11. Overall uncertainty in the frequency of the cantilever as a function of the
uncertainty in the thickness, for different lengths of the microcantilevers.
5.3. Optoelectronic measurements
The optoelectronic methodology described in Section 4.4.5, was used to produce
images of the first few bending modes of the cantilevers (up to the 5
th
bending was
found), Fig. 5.12. The first bending mode was very important, however, for it
corresponded to the natural resonant frequency that was analytically calculated. With the
setup used, it was possible to determine this resonant frequency with the accuracy of 10
Hz or better than 0.1% of the experimental determined frequencies as shown in Section
5.6. To determine the resonant frequencies of each cantilever, a range of frequencies
from 1 Hz to 1 MHz was scanned through and the resonances were determined. The first
122
bending resonant frequency is shown in Fig. 5.13 for each cantilever (D, E, and F) on
chip 22. The results from the OELIM measurements are summarized in Section 5.6.
Fig. 5.12. OELIM images of cantilever E32 vibrating under resonant conditions: a) the
1
st
bending mode (resonant frequency) of the F length cantilever where the fringes
indicate the magnitude of deformation and the mode shape at 12, 650 Hz, b) 2
nd
bending
mode at 81,400 Hz, c) 3
rd
bending mode with torsional bending at 115,890 Hz, d) 4
th
bending mode at 456,150 Hz, e) 5
th
bending mode at 754,640 Hz.
Fig. 5.13. OELIM images of the cantilevers (D, E, and F) of chip 22 vibrating at
their fundamental mode of vibration: a) the D length cantilever, resonating at
14,356 Hz, b) the E length cantilever, resonating at 10,602 Hz, c) the F length
cantilever, resonating at 20,716 Hz
(a) (b) (c)
(d) (e)
(a) (b) (c)
123
5.4. AFM results
The cantilever chips were loaded on to a chip carrier and then into the AFM.
Because the chips were originally designed for a use with an AFM, they were easy to set
up and get results. The AFM then scanned the frequency ranges and the resonant
frequency could be detected with an accuracy of 10 Hz. Figure 5.14 shows an example
of the AFM frequency scan with the resonant peak clearly defined and marked for the
cantilever F12. This was how the values were found for each cantilever. The results of
AFM measurements are summarized in Section 5.6.
Fig. 5.14. A screen capture of the AFM frequency scan for
the cantilever F12, where a resonant frequency peak is
identified as 22,160 Hz with the accuracy of 10 Hz.
124
5.5. Computational results
For the computational analysis, Pro/Engineer was used to build the model and
Pro/Mechanica was used for the modal analysis. The model used is shown in Fig. 5.15
where all three cantilevers were analyzed at the same time. The model shown in Fig.
5.15 faithfully represents the chip used in this Thesis.
For the software to generate accurate and valid results it is necessary for the
solution to converge. Current practice is to have the convergence percentage at 0.01%,
but for the software used here, the lowest value attainable was 1% (PTCb, 2003), based
on “fixed solution. Pro/Mechanica uses ptype (polynomial) elements for converging on
the solution quickly and more accurately than with other available methods. The ptype
elements utilize higher order polynomials to obtain the solution instead of increasing the
number of elements as in ntype element software packages. Thus ptype elements
allows for quicker, discreet solutions that are attainable with just a few steps in the
calculation.
The results of the computational simulation were compared to the analytically
determined values to determine degree of correlation. Ideally, the computational results
would be compared to the experimental results. However, to be more efficient, since
each cantilever is of different dimensions and would require a new computational
solution, the computational results were compared to the analytical results for two
different conditions for correlation. This way the analytical results could be directly
compared with the experimental results with confidence that the computational results
would be similar to the analytical results. The first case used ideal structural dimension
125
values (as those specified by the manufacturer in Table 3.1) for the cantilevers while the
second used the averaged values as listed in Table 5.2. For example, the percent
difference between the analytical and FEM results, i.e., Resn. freq. % diff, was calculated
as
FEM freq. Analytical freq.
Resn. freq. % diff 100
Analytical freq.
15314 15326
100
15326
0.078 % .
  −
= ⋅

\ .
−  
= ⋅

\ .
=
(5.2)
These values are displayed in Table 5.5 along with the analytically calculated values.
As can be seen in Table 5.5, the computational results correlate well with the
analytically determined values. For the ideal case, using values from Table 3.1, the
frequency difference is only about 0.54% and for the averaged case it is about 0.47%.
Table 5.5. Comparison of the analytical and computational
resonance frequency results.
Measured D E F D E F
parameters Typical Typical Typical Typical Typical Typical
Length, L
µ
m 300 350 250 303.5 353.2 253
Width, b
µ
m 35 35 35 35.83 35 35.25
Thickness, h
µ
m 1 1 1 0.985 0.985 0.985
Analytical freq., Hz 15326 11260 22069 14751 10894 21234
FEM freq., Hz 15314 11167 21907 14633 10870 21315
Resn. freq. % diff. 0.0783 0.8259 0.7341 0.8 0.22 0.381
Based on nomicnal
dimensions from the
manufacturers
Based on measured
dimensions, reference
Table 5.2
126
Fig. 5.15. Pro/Engineer model and Pro/Mechanica results of the three cantilevers
considered: a) cantilever D vibrating at its resonant frequency of 14,679 Hz, b) cantilever
E vibrating at its resonant frequency of 10,819 Hz, c) cantilever F vibrating at its
resonant frequency of 21,270 Hz.
(a)
(b)
(c)
127
5.6. Comparison of results
In this section, all the results obtained in this Thesis are compared for determining
quantitative correlation. The data from the OELIM and AFM are to be correlated to the
simplified analytically determined percent overall uncertainty in frequency, as discussed
in Section 5.2. To do this, the difference between the determined AFM frequencies of
each cantilever is compared to the analytically determined frequency. This was also done
for the OELIM determined frequencies and then percent differences were calculated for
both methods. This difference can be seen as the uncertainty in the OELIM and AFM
data if the analytical results are considered correct. These differences were then averaged
so that they could be directly compared to the analytically determined uncertainties. For
example, the percent difference between the analytical and AFM results for cantilever
D11, i.e., Anal./AFM freq. % diff, was calculated as
AFM freq. Analytical freq.
Anal./AFM freq. % diff 100
Analytical freq.
15940 15865
100
15865
0.47 % .
  −
= ⋅

\ .
−  
= ⋅

\ .
=
(5.3)
Relationships similar to Eq. 5.3 were used to determine Anal./OELIM freq. % diff
for the remaining cases and the results are summarized in Table 5.6.
128
Table 5.6. Data comparison of analytical, OELIM, and
AFM generated frequencies for each cantilever.
As can be seen from Table 5.6, the values for the OELIM and AFM methods
where compared to the analytically calculated values. The overall uncertainties, listed
under ‘Uncertainty’ in the table, were calculated in Section 5.2. The optoelectronic
measurements, listed as ‘OELIM’ and the AFM, listed as ‘AFM’ in the table, data results
were discussed in Sections 5.3 and 5.4, respectively. One final average was calculated
D11 E11 F11 D12 E12 F12
Analytical freq., Hz 15865.4 11812.3 22772.1 16116.7 11801.4 23175.9
AFM resn. freq., Hz 15940 11580 22160 16010 11850 22160
OELIM freq., Hz 15700 11440 22100 15940 broke 22040
Anal./AFM freq. % diff. 0.47014 1.96619 2.68777 0.6618 0.41153 4.38341
Anal./OELIM freq. % diff. 1.04259 3.1514 2.95125 1.09613  4.90119
D21 E21 F21 D22 E22 F22
Analytical freq., Hz 16269.8 12044.9 24105.4 15405.6 11377 22230.6
AFM resn. freq., Hz 15849 10690 22450 15670 11700 22580
OELIM freq., Hz 15750 11070 22380 15420 11490 22590
Anal./AFM freq. % diff. 2.58613 11.2484 6.86728 1.71648 2.83881 1.57183
Anal./OELIM freq. % diff. 0.62465 9.11931 7.15767 0.09369 0.99298 1.61682
D31 E31 F31 D32 E32 F32
Analytical freq., Hz 15969.9 11731.8 22919.7 15354.2 11378.5 21576.3
AFM resn. freq., Hz 13230 11560 19630 16990 12720 24280
OELIM freq., Hz 19630 11560 25650 17160 12650 24360
Anal./AFM freq. % diff. 17.1569 1.46404 14.353 10.6536 11.7901 12.5311
Anal./OELIM freq. % diff. 22.9184 1.46404 11.9127 11.7608 11.1749 12.9019
Ave. % diff. of D cantil.
Ave. % diff. of E cantil.
Ave. % diff. of F cantil.
Overall ave. % uncert. 6.150
OELIM
6.256
5.181
6.907
6.114
Uncertainty
6.102
5.865
6.482
5.853
AFM
5.541
4.953
7.066
129
for the overall average percent difference between all the cantilevers for each method.
From the collected data, it is clear that the percent differences in the experimental results
(AFM=5.85% and OELIM=6.11%) were similar to the analytically calculated uncertainty
(Uncertainty=6.15%) results. This correlation of the data validates the parameters
contributing to the overall uncertainty as analytically calculated in Section 5.2.
5.7. Parameters of the coated cantilever
Because of the idea to coat the cantilevers with palladium for hydrogen detection
at room temperature and 1 atm pressure it was necessary to prove the functionality of this
theory. To accurately model this Pd/H
2
system, it was necessary to include all the
parameters that define the coated cantilevers like the different moduli of elasticity and
densities, which would be present. These moduli and densities are listed in Table 5.7.
Using Eqs 4.6 and 4.7 we find that the equivalent moduli of elasticity and density for a
350 µm long cantilever coated with 1000Ǻ of Pd and 10Ǻ of Ti are
( )
( )( )
( ) ( ) ( )
3
3 3 3
190 116 121 14,813.76
116 121 12, 250 190 121 25.24 190 116 2538.52
172.9 ,
Si Ti Pd t
eq
Ti Pd Si Si Pd Ti Si Ti Pd
E E E V
E
E E V E E V E E V
m
m m m
GPa
µ
µ µ µ
⋅ ⋅
=
+ +
⋅ ⋅
=
⋅ + ⋅ + ⋅
=
(5.4)
( ) ( ) ( )
3 3 3
3 3 3
3
2.33 12, 250 4.507 25.24 12.023 2538.52
12, 250 25.24 2538.52
3.995 / .
Si Si Ti Ti Pd Pd
eq
Si Ti Pd
V V V
V V V
m m m
m m m
g cm
ρ ρ ρ
ρ
µ µ µ
µ µ µ
+ +
=
+ +
+ +
=
+ +
=
(5.5)
130
Table 5.7. Properties of materials used.
5.8. Sensitivity solutions
In order to determine the correct coverage of the cantilever with palladium the
sensitivity was calculated for two cases: a completely covered cantilever and a partially
covered, i.e. tip only, cantilever.
It was first necessary to determine the amount of coverage to use in a partially
coated cantilever case. Using Eqs 4.2 and 4.5 it was found that the “sectional” or partial
mass added to the cantilever at the tip affects the overall mass more than if it were added
at the base. As the partial mass that is loaded is moved closer to the base, its effect on the
resonance frequency of the cantilever is reduced. It was found that when 25% of the
cantilever is covered with 0.1 µm palladium at the tip it accounts for about 72% of the
effective mass of the complete cantilever. In Fig. 5.16 the distribution, or percent
coverage, of the mass on the cantilever is related to how much the effective mass value is
affected. As a uniform amount of mass is added to the cantilever, starting from the base
and depositing towards the tip, the effective mass will increase exponentially. As a
coating of mass is added in the opposite direction, starting from the tip and depositing
Density
ρ ± 0.005 g/cm
3
Modulus of elasticity
E ± 3 GPa
Silicon, Si 2.330 190
Palladium, Pd 12.023 121
Titanium, Ti 4.507 116
Equivalent values 3.995 172.9
131
towards the base, the effective mass slope decreases in an opposite manner to the
previous loading method.
Fig. 5.16. Effective mass versus percentage of coating coverage: the dotted line
illustrates the effective mass contribution as a function of the position along the
cantilever moving from the base to the tip as mass is added; the solid line illustrates the
same contribution as the position moves from the tip to the base as mass is added. Both
cases illustrate how the position of where the mass is added, as it gets closer to the tip of
the cantilever, will affect the effective mass.
Since the greatest contribution to the effective mass due to addition of mass
occurs in the end/tip 25% (75% through 100% of the cantilever position), it was decided
132
that this was a reasonable area to use for calculations for the amount of coverage for a
partial covered cantilever.
The second method of calculating the sensitivity was assuming the entire
cantilever was uniformly coated. This case will be slightly different from the partially
coated case, as the differentiation of the same initial equation will produce different
results due to the different formulation of the derivative (see Appendix B for the
derivations).
Using Eqs 4.2 and 4.5 the sensitivity was calculated for the partially and
completely covered cantilever (see Appendix B.4). Using the 350 µm long cantilever
with a stiffness of 0.039 N/m, a mass of 2.854x10
8
g, ∆m = 10
11
g of mass to “add”, an
effective mass constant of n
r
= 71/420 for a coverage of r = 0.25, the sensitivity for a
partial coated cantilever is
( )
( )
2
2
2 2
8
8 11
2
8 11
9
1
2
71
0.039 0.25
420
0.039 0.034
2
33
33 71
2.854 10 2.854 10 0.25 10
140
140 420
1
71
2.854 10 0.25 10
420
3.141 10 ,
r
e
r
r
kn r
S
k k m n r m
m m n r m
kg
s
kg kg
s s
g g g
g g
Hz
Hz kg
−
− −
− −
−
=
+ ∆
+ ∆
 
− ⋅ ⋅

\ .
= •
 
⋅ ⋅ ⋅ ⋅ + ⋅ ⋅

\ .
 
⋅ + ⋅ ⋅

\ .
= − ⋅
−
(5.6)
while the sensitivity for a completely covered cantilever is
133
( )
( )
( )
( )
2
2
2 2
8 8 11
2
8 11
10
1
33
2
140 33 33
140 140
0.039
0.039 0.039
2
33 33
2.854 10 2.854 10 10
140 140
1
33
2.854 10 10
140
1.751 10 .
d
k
S
k k
M m
M M m
kg
s
kg kg
s s
g g g
g g
Hz
Hz kg
− − −
− −
−
=
+ ∆
+ ∆
 
−

\ .
= •
   
 
\ . \ .
⋅ ⋅ +
⋅ +
= − ⋅
−
(5.7)
As can be seen from the results of Eqs 5.6 and 5.7 for detecting the largest
overall shift in resonant frequency at a steady state condition a fully coated cantilever was
preferred, Table 5.8. This coincides with the conclusions of other researchers in the area
of chemical sensors (Betts et al., 2000; Lange et al., 2001).
Table 5.8. Frequency shift per Hertzkilogram for
distributed and tip loads.
5.9. Absorption
The main ability in the hydrogen sensor is for palladium to absorb H
2
. Because of
this, the absorption reaction must be well known. This reaction can be described with
Distributed load Tip loaded
1.751E10 Hz/Hzkg 3.141E9 Hz/Hzkg
Sensitivity
at 1% H
2
134
two main points: the concentration that is absorbed and the amount of time required for
that absorption.
The first part, the amount absorbed, is measured by the atomic ratio, or the
volumetric amount of one element compared to the other at the atomic scale. This
absorption is directly affected by the pressure and temperature of the gas. As the partial
pressure of hydrogen decreases the absorption decreases. This is an important
characteristic because of the conditions of this Thesis to have the sensor work at 1 atm, or
approximately sea level pressure, and with a concentrations equal and less than 1%, the
partial pressure will only be <0.01 atm. This is different than with the temperature,
which seems to show an increase in the atomic ratio as the temperature decreases. These
changes can be seen in Fig. 4.3. This figure, and most data currently available, about
hydrogen absorption into palladium are from experimental data collected throughout the
years (Lewis, 1967). It is important to note that the experimental data available for
determining the atomic ratio of the absorption of hydrogen in palladium at low
concentrations are limited. For this reason, the atomic ratio absorption at lower
concentrations (<1%) is approximated using a square root proportionality decrease from
Sievert’s law (Lewis, 1967).
Since measuring the lower concentrations (<1%) is what is of interest, the
equations used had 1% hydrogen concentration as the upper limit. In addition to
assuming Sievert’s law for absorption amount, another formula (located in Appendix
C.2.2 as Eq. C.80) was generated, simulating Fig. 4.6, to estimate the change in modulus
of palladium with hydrogen concentration being the variable. Equation C.80 was
135
integrated into the equivalent modulus of elasticity relationship so that it could
compensate for the change in the modulus of elasticity of the palladium during
absorption. The final equation then that represented the equivalent amount of mass of
hydrogen accumulated on the cantilever by absorption at different concentrations was
( )( )( )
2
_ _ _ , m absorption multiplier H density Pd volume ∆ = (5.8)
where absorption_multiplier is the conversion of the atomic ratio into the equivalent
volume as described in Table 4.1, H
2
_density is determined with Eq. 4.9, and Pd_volume
is the palladium coating layer volume, which although changes with the absorption
process it does not increase in the overall volume mass (number of palladium molecules).
It is important to note though the importance of the partial pressure of the hydrogen, as it
will affect the amount of absorption greatly and thus the results.
The other important parameter in the absorption process is the length of time
required to reach certain concentration of hydrogen within the palladium, Fig. 5.17.
Using Fick’s second law of diffusion in conjunction with Sievert’s law, it was
possible to estimate the length of time required to reach any particular percentage amount
of equilibrium concentration in the palladium, Fig. 5.17, which was prepared for a
cantilever with a 0.1 µm coating of palladium in a hydrogen concentration of 1%, using
Eq. 4.12. From Fig. 5.17, is it clear that the diffusion occurs very quickly and that within
less than a second the concentration is greater than 90% of the equilibrium. This is very
useful and enabling for a sensor as speed is important in detecting agents.
136
Fig. 5.17. Concentration as a function of time for a cantilever coated with
0.1 µm thick layer of palladium. The time to reach 90% of an equilibrium
concentration of 1%, or 10000 ppm, is about 1 second.
5.10. Determination of the damping coefficient
Unless the cantilever is vibrating in a vacuum, there will be some sort of damping
acting on the system. To accurately model the system, it was necessary to estimate the
damping coefficient, which in this case would be for the damping fluid air. This was
accomplished by utilizing the quality factor equations and the vibrometer. Using a
microscope, the Polytec Instruments vibrometer laser beam was focused into a spot of
about 20 µm in diameter. This spot was smaller than the width of the cantilevers that
were examined, which was required to ensure that the vibrometer was only reading the
cantilever and not the underlying substrate. Once the vibrometer was set up, a digital
137
signal generator (Wavetek 2MHz Variable Phase Synthesizer model 650) was used to
scan a frequency range that would cover the first bending mode resonant frequency.
Using the spectrum analyzer (HP Spectrum Analyzer 3588A) a resonant frequency peak
curve was produced as shown in Fig. 5.18. The maximum peak was found and then two
halfpower points (described in Section 4.4.4) were found at 1/ 2 (~70.8%) of the
amplitude value. These two points were used for measuring the bandwidth, which is
directly related to the mass and damping as was shown in Eq. 4.44. For the cantilever
D22, for which Fig. 5.18 represents the data, the damping was found to be
( )
8
8
21,880 21,155 2.04 10
1.48 10 .
c bandwidth m
Hz Hz g
kg
s
−
−
= ⋅
= − ⋅ ⋅
= ⋅
(5.9)
In addition, using the result from Eq. 5.9, since the damping is defined by Eq.
4.25 and the parameters L and b are known values and µ, which is the viscosity of
nitrogen at 1.77·10
11
kg/µms, it is possible to determine H as
( )( )
11
8
2
2 1.77 10 252.818 33
1.48 10
21.8 .
Lb
H
c
kg
m m
m s
H
kg
s
H m
µ
µ µ
µ
µ
−
−
=
 
⋅

−
\ .
=
⋅
=
(5.10)
Table 5.9 lists the calculated values for Eq. 4.25 and the bandwidth data measured
for a sample chip (D22) and the corresponding calculated damping. It should be noted
that during the frequency sweep, some harmonic resonance frequencies were found, as
138
can be seen in Fig. 5.18. The bandwidth was measured for these harmonics using the
formulation from Eq. 5.9 and it was found that the results are similar to the resonant
frequency bandwidths, Table 5.9. All the damping coefficient values are presented in
Table 5.9. As can be seen from the table, the values for the damping are all of the same
order of magnitude and fairly close to one another. This damping coefficient data were
then used to analytically model the absorption under different cases as will be discussed
in Sections 5.11.1 and 5.11.7.
Fig. 5.18. A digitally enhanced image of a screen capture from the HP Spectrum
analyzer taken while measuring the frequencies of a cantilever. The resonant frequency
can be seen as the distinctive spike (with a diamond marker at the tip) at the frequency of
21,590 Hz. The other peaks are the harmonic frequencies.
139
Table 5.9. Summary of analytically calculated and experimentally
determined damping coefficients.
5.11. Complete cantilever analysis
In order to fully investigate absorption of hydrogen into the palladium coated
cantilever the system had to be analytically investigated. This investigation considered
different cases of varying complexity characterized by different parameters. Four cases
were considered and each one included an increase of mass by absorption of hydrogen.
They are: a simple case with only mass changing, a case with mass and stiffness
changing, a case with changing mass and damping present, and a case with changing
mass and stiffness in the presence of damping. For calculations, the 350 µm long E
cantilever with ideal dimensions and coated with 1000Å of palldium on top of 10Å of
titanium was used as the sensing element. All four cases and results obtained are
Viscosity of nitrogen, kg/
µ
ms 1.86x10
11
Density, ρ g/cm
3
2.33
Measured D22 E22 F22
Characteristics Typical Typical Typical
Length, L
µ
m 303.7 353.4 252.82
Width, b
µ
m 35 34.5 33
Thickness, h
µ
m 0.96 0.96 0.96
Height, H
µ
m 21.8 21.8 21.8
Mass, m kg 2.377x10
11
2.727x10
11
1.866x10
11
Damping calculated, c kg/s 1.725x10
11
1.979x10
11
1.354x10
11
Bandwidth of resonance,
∆
Hz 653 1088 725
Damping experiment, kg/s 1.553x10
11
2.967x10
11
1.353x10
11
Bandwidth of harmonics,
∆
Hz 724 507 
Damping experiment, kg/s 1.721x10
11
1.383x10
11

140
presented in Sections 5.11.1 through 5.11.7 and the derivations for each equation within
these sections are located in Appendices B and C.
5.11.1. Case 1: changing mass
The simplest of the cases considers the situation where there is no damping and
the stiffness does not change. This situation is an idealized case. Using the equivalent
density and modulus of elasticity listed in Table 5.7, the equation used can be rewritten
from Eq. 4.19 into the following form:
2
2
1 35
2 33
,
1
eq
eq
E
h
L
f
m
m
π ρ
=
∆
+
(5.11)
where the change in mass due to hydrogen absorption is ∆m and other parameters are as
previously defined. It was calculated, Section C.2.1, that the percent overall uncertainty
in the frequency shift is 6.926% when the concentration of hydrogen is at 1%. These
results are listed in Table 5.8 where it can be seen that for case 1, at 1% concentration of
hydrogen, the frequency will shift by –1019.801 Hz, where the negative sign indicates
that the frequency will decrease. With an uncertainty of 539.106 Hz in the shifted
frequency of 7783.268 Hz, it is clear that the uncertainty is just about half of the value of
the frequency shift making the accuracy of determining the shift challenging. All of the
results in this section are derived in Section C.2.1.
141
5.11.2. Case 2: changing mass and stiffness
The next case is more complicated in that it will look at both a change in the mass
as well as a change in the stiffness. Because of the experimental results of the palladium
hydrogen system conducted in the past (Lewis, 1967; Alefeld and Völkl, 1978a, 1978b),
there is a known contribution to the change in the stiffness from the modulus of elasticity
and the volume as both can alter during absorption, Figs 4.5 and 4.6. Because of the
limited data for absorption of hydrogen at small concentrations (<1000ppm) the change
in the modulus of elasticity and volume due to hydrogen absorption was assumed correct
without uncertainty. Equation 4.22 used for this determination is repeated here for
continuity, i.e.,
2
2
2
1
1
1 35
,
2 33
1
eq
eq
k
E
h k
f
m L
m
π ρ
∆  
+

 
\ .
 =

∆  
\ .
+

\ .
(5.12)
where the subscript 1 indicates the initial frequency. It was then calculated from Eq. 5.12
that the shift in frequency at 1% concentration is –908.546 Hz for a shifted frequency of
7894.523 Hz with a percent overall uncertainty in this shift of 6.917% equaling 546.06
Hz. As it can be seen from the values of the percent overall uncertainty of this and the
previous case (Section 5.11.1) that the change is minimal due to the addition of a
changing spring constant. This alludes to the conclusion that a changing stiffness will not
play a major role in determining the uncertainty. The derivations of the results in this
section are located in Section C.2.2.
142
5.11.3. Case 3: changing mass with damping
The third case investigates the frequency shift that occurs due to a change in mass
with damping present, which for this case was calculated for the cantilever in air. The
damping values used in this case were experimentally verified in Section 5.10. The
results from this investigation are similar to those without any damping even though there
is a significant change in the formulation. Using Eq. 4.31,
( )
( )
2
2
2
2 2
1
4
1 35
,
2 33
4
eq
eq
m k m m c
E
h
f
L
k m m
π ρ
  ( + ∆ −
¸ ¸
 =

+ ∆
\ .
(5.13)
the contribution that damping has is accounted by the parameter c. When the appropriate
values are substituted into Eq. 5.13 at a hydrogen concentration of 1% the percent overall
uncertainty in the frequency 6.927% with an uncertainty of 539.113 Hz. The results are
very similar to case 1, which had no damping, suggesting that the damping coefficient of
air plays a small to indifferent role in improving the frequency data quality. The data are
presented in Table 5.10.
5.11.4. Case 4: changing mass and stiffness with damping
The final scenario looked at incorporates all the variables: mass, damping, and
stiffness. This case represents the results the most accurately. The formula modeling this
system is Eq. 4.32, repeated here for continuity, i.e.,
( )
( )( )
2
2
2
2 2
1
4
1 35
,
2 33
4
eq
eq
m k m m c
E
h
f
L
k k m m
π ρ
  ( + ∆ −
¸ ¸
 =

+ ∆ + ∆
\ .
(5.14)
143
which is very similar to the damped case described in Eq. 5.13. It has been calculated that
the change in the frequency due to the changing stiffness and damping is 532.137 Hz
while the final percent overall uncertainty in frequency is 6.696%. The results are similar
to those of the case 3 with the changing mass and stiffness.
5.11.5. Design equation selection
In designing cantilever sensors, it is necessary to determine the level of accuracy
in modeling needed in order to maximize data quality. Using the results of Sections
5.11.1 through 5.11.4, it is possible to argue what are necessary variables and
considerations in designing cantilever sensors. From the data collected from each of the
four cases described it was possible to determine which parameter (stiffness and
damping), and in turn equation, were the most and least necessary for accurate results.
All the data are compiled in the Table 5.10.
It can be seen from the analytical results in Table 5.10 that by being more
accurate and including such parameters as damping and a changing stiffness the overall
uncertainty in the frequency does not alter much. This means that it is not really
necessary to include these other parameters to get accurate results under the conditions
used in this Thesis. In other words, to model a vibration system that is damped by air, it
is only necessary to account for changes in mass, as changes in any other parameter will
not lead to significantly improved results. This simplifies the equations and parameters
needed and also the experimental setups as less factors need to be taken into account.
144
This also increases the usability of the sensor in a wider variety of environments, as
effects produced by these environments will not affect the results and reliability of it.
Table 5.10. The results of the four different cases investigated are
compared at 1% hydrogen concentration.
5.11.6. Lower limit frequency shift
As determined in Section 5.11.5, it is not necessary to include variables such as
damping and a changing stiffness as the quality of the results will not improve by much.
With this knowledge, it was possible to estimate the frequency shift amount that would
occur at lower concentrations of hydrogen, such as 1 ppm, absorbing into palladium. Of
course, as mentioned in Section 4.2.4, the exact amount of absorption at such low levels
is not very well known, but can be approximated using diffusion laws. With these
approximations in Eq. 5.11 the frequency shift at 1 ppm was calculated using the same
process as described in Section C.2.1, at room temperature and at 1 atm, as
Frequency data at 1% H
2
Case 1:
(m)
Case 2:
(m,k)
Case 3:
(m,c)
Case 4:
(m,k,c)
Shifted freq. 7783.268 7894.523 7783.313 7673.619
Freq. shift, Hz 1019.801 908.546 1019.756 1129.450
Overall freq. uncert,, Hz 539.106 546.06 539.113 532.137
Overall % freq. uncert. 6.926% 6.917% 6.927% 6.696%
145
( )
( )
2
2
2
3
2
22
11
2
35
1
2 33
1
35 172.9
1 1
2
350
33 3.995
1.652 10
1
5.918 10
8803.069 Hz ,
eq
eq
E
h
L
f
m
m
GPa
m
g
m
cm
f
kg
kg
f
π ρ
µ
π
µ
−
−
=
∆
+
 

\ .
=
⋅
+
⋅
=
(5.15)
where the mass added is very low (10
22
order of magnitude) and therefore the shift at this
low concentration level is very small (<0.001 Hz). Figure 5.19 illustrates how the
frequency shift increases as the concentration of hydrogen is increased.
It is shown that at around 10 ppm the frequency shift is almost unnoticeable and
would be very difficult to detect. There are a few possible reasons for this. The first is
that the data of absorption of hydrogen required to accurately graph the curve of Fig. 4.3
is not readily available at small concentrations, which can allow for misrepresentation of
the behavior of the cantilever. If the palladium does in fact absorb more hydrogen than
what is being used in the equations, then the shift will increase. The second reason is that
the partial pressure of the hydrogen is too small to activate absorption into the palladium.
Since the hydrogen will be mixed down with nitrogen, the partial pressure will be low.
And the other reason is that the mass (quantity) being attempted to detect is too small in
comparison to the mass of the cantilever used, which will lead to very small frequency
shifts and will require highly sensitive frequency shift detecting methods and equipment.
146
Fig. 5.19. The frequency shift in the resonance of a cantilever as a
function of concentration of the hydrogen.
5.11.7. Optimization of the cantilever response
In order for the cantilever to detect a shift in the frequency that is greater that 10
9
Hz, there are several factors that must be considered. Addressing the issues brought up in
Section 5.11.6, parameters of the frequency equation will be altered and will be
documented on their effect on the results. First, the overall length of the cantilevers was
reduced by two orders of magnitude (for example: L=3.5 µm). Next the partial pressure,
and thus in turn the overall pressure of the gas, was increased by two orders in
magnitude. The frequency shift was increased by a total of 4 orders in magnitude with
these modifications. Upon close inspection it was found that for each order of magnitude
147
increase in the partial pressure, the frequency shift increased by one order of magnitude.
For the geometric dimensions, the results of decreasing the length and increasing the
thickness by orders of magnitude increased the resonant frequency of the cantilever that
directly affected the frequency shift. This indicates that higher frequency cantilevers are
favorable for sensors as they are more sensitive to frequency shifts. To attain these
cantilevers with higher resonant frequencies the overall dimensions such as length and
thickness must be reduced.
This has some serious effects on the sensors capabilities due to the uncertainty in
the thickness discovered. For example, if the thickness is desired to be 0.1 µm (or less)
for the sensor, the current uncertainties will basically prohibit the fabrication of these
cantilevers at these scales, as the uncertainties will be extremely high. If a cantilever is
fabricated 5% (or greater) off from the planned structural dimensions the high uncertainty
will undermine sensitivity and function of the sensors. At such small concentrations, 1
ppm or less, slight differences in the geometry will change the sensitivity and thus,
regardless of calibration, the sensor will be out of specifications. For this reason, having
a low uncertainty in the thickness parameter is important for the future development of
cantilever sensors.
148
6. CONCLUSIONS
In this Thesis, MEMS cantilevertype chemical sensors were investigated. This
investigation involved, several chips. Each chip had 6 cantilevers of different lengths and
for this Thesis the three longest cantilevers, i.e., 250 µm, 300 µm, 350 µm, were used.
Analysis and preliminary characterization of vibrations of these cantilevers were
investigated, in this Thesis, using analytical, computational, and experimental solutions
(ACES) methodology.
To characterize the cantilevers, analytical calculations were performed. The
primary investigated aspect involved determination of a nominal frequency and overall
uncertainty in the frequency of the cantilevers. Typical values for the resonance
frequency and the percent overall uncertainty were found to be around 16 kHz and
6.15%, respectively. It was also found that out of all independent parameters defining
resonance frequency, uncertainty in thickness had the greatest contribution to the overall
uncertainty in the frequency. More specifically, δh =0.05 µm contributed 74.9 % to the
overall uncertainty in the resonance frequency of the 350 µm long cantilever. The
resonant frequency for each cantilever was determined for comparison with experimental
results.
Vibrometer, OELIM, and AFM methodologies were used in gathering
experimental results. An AFM was used to determine the frequency response curves
based on which the resonant frequencies of cantilevers were calculated. The OELIM
method was particularly capable of determining mode shapes associated with specific
149
resonance frequencies for a number of vibration modes for each cantilever sample. The
laser vibrometer was used to determine the damping coefficients.
It was found that the analytically calculated percent overall uncertainty was
similar to the percent difference between the analytically calculated frequency and the
frequencies measured by the AFM and the OELIM methodologies. The difference, on
the order of about 0.3%, indicated that the analytically calculated uncertainty was
accurate. The percent overall uncertainty contribution calculations indicated that the
thickness was the greatest source of uncertainty and that the analytical and computational
results correlated well with the experimental results.
Investigations into the cantilever as a chemical sensor for detecting hydrogen with
the palladium/hydrogen system were also conducted. Using equations that included such
variables as damping and a changing spring constant several cases were considered. It
was found from the analytical results of this Thesis that, although damping due to the
nitrogen environment and a changing stiffness do alter the frequency, the percent overall
uncertainty in the frequency, or the quality of the results, did not change much, ~0.3%
change.
Since the damping and a changing stiffness do not affect the frequency
significantly, the sensitivity of the cantilever to the palladium/hydrogen system was
determined without considering these parameters. The sensitivity was calculated to
determine how a sensor would function the best under the conditions of room
temperature (20˚C) and atmosphere pressure (1 atm). It was concluded that increasing
the resonant frequency is the most direct way of improving the sensitivity of the sensor.
150
This can be readily accomplished by changing the overall dimensions of the cantilever,
i.e., decreasing length and increasing the thickness. However, if the length of the
cantilevers is decreased and the thickness is increased, the overall percent uncertainty in
the frequency will increase due to the uncertainty in the thickness. This will cause
problems with sensors designed for detecting 1 ppm, or less, regardless of calibration
because this uncertainty will produce variation in the sensitivity and may put them out of
the specifications.
This Thesis investigated a MEMS cantilevertype chemical sensor and its design.
It was found that to improve the sensitivity of these sensors, it will be necessary to
improve the uncertainty in the thickness. More specifically, the fabrication processes that
make these cantilevers would have to be improved to provide an order of magnitude
decrease in the uncertainty in thickness. As fabrication processes improve, the ability to
make smaller devices more accurately will facilitate development of the cantilevertype
MEMS chemical sensors.
151
7. REFERENCES
G. Abadal, Z. J. Davis, B. Helbo, X. Borrisé, R. Ruiz, A. Boisen, F. Campabadal, J.
Esteve, E. Figueras, F. PérezMurano, and N. Barniol, 2001, "Electromechanical model
of a resonating nanocantileverbased sensor for highresolution and highsensitivity
mass detection", Nanotechnology, 12:100104.
G. Alefeld and J. Völkl, 1978a, Hydrogen in metals, Vol. 1, SpringerVerlag, New York,
NY.
G. Alefeld and J. Völkl, 1978b, Hydrogen in metals, Vol. 2, SpringerVerlag, New York,
NY, pp. 145.
AD, 2003, “iMEMS® accelerometers”, http://www.analogdevices.com, Analog Devices,
Last visited: November 3, 2003.
D. K. Arecco and R. J. Pryputniewicz, 2003, “Design and analysis of a MEMS chemical
sensor,” Proc. 4
th
Internat. Symp. on MEMS and Nanotechnology (4
th
ISMAN), Charlotte,
NC, pp. 211217.
H. Baltes, O. Brand, A. Hierlemann, D. Lange, and C. Hagleitner, 2002, "CMOS MEMS
 present and future", Proc. Symp. on MEMS 2002, Las Vegas, NV, pp. 459466.
F. M. Battiston, J. P. Ramseyer, H. P. Lang, M. K. Baller, Ch. Gerber, J. K. Gimzewski,
E. Meyer, and H. J. Günthherodt, 2001, “A chemical sensor based on a microfabricated
cantilever array with simultaneous resonancefrequency and bending readout”, Sensors
and Actuators B, 77:122131.
D. R. Baselt, 1993, The tipsample interaction in atomic force microscopy and its
implications for biological applications, Ph.D. Dissertation, California Institute of
Technology, Pasadena, CA.
D. R. Baselt, B. Fruhberger, E. Klaassen, S. Cemalovic, C. L. Britton Jr., S. V. Patel, T.
E. Mlsna, D. McCorkle, and B. Warmack, 2003, “Design and performance of a
microcantileverbased hydrogen sensor”, Sensors and Actuators B, 88:120131.
R. Bashir, J. Z. Hilt, A. Gupta, O. Elibol, and N. A. Peppas, 2002, “Micromechanical
cantilever as an ultrasensitive pH microsensor”, Appl. Phys. Lett., 81(16):30913093.
J. M. Berthelot, 1999, Composite materials: mechanical behavior and structural
analysis, SpringerVerlag, New York, NY.
152
J. Bernstein, R. Miller, W. Kelley, and P. Ward, 1999, “Lownoise MEMS vibration
sensor for geophysical applications”, J. MEMS, 8(4):433438.
T. A. Betts, C. A. Tipple, M. J. Sepaniak, and P. G. Datskos, 2000, “Selectivity of
chemical sensors based on microcantilevers coated with thin polymer films”, Analytica
Chimica Acta, 422:8999.
C. L. Britton, Jr., R. L. Jones, P. I. Oden, Z. Hu, R. J. Warmack, S. F. Smith, W. L.
Bryan, and J. M. Rochelle, 2000, "Multipleinput microcantilever sensors,"
Ultramicroscopy, 82:1721.
C. L. Britton, Jr., R. J. Warmack, S. F. Smith, P. I. Oden, R. L. Jones, T. Thundat, G. M.
Brown, W. L. Bryan, J. C. Depriest, M. N. Ericson, M. S. Emery, M. R. More, G. W.
Turner, A. L. Wintenberg, T. D. Threatt, Z. Hu, L. G. Clonts, and J. M. Rochelle, 1999,
"Batterypowered, wireless MEMS sensors for highsensitivity chemical and biological
sensing," Proc. Symp. on Advanced Research in VLSI, Atlanta, GA, pp. 359368.
G. C. Brown, 1999, Laser interferometric methodologies for characterizing static and
dynamic behavior of MEMS, Ph.D. Dissertation, Worcester Polytechnic Institute,
Worcester, MA.
C. Brown and R. J. Pryputniewicz, 2000, “New test methodology for static and dynamic
shape measurements of microelectromechanical systems,” invited paper, Opt. Eng.,
39:127136.
C. Brown and R. J. Pryputniewicz, 1998, “Holographic microscope for measuring
displacements of vibrating microbeams using timeaverage electrooptic holography,”
invited paper, Opt. Eng., 37:13981405.
C. Brown and R. J. Pryputniewicz, 1996, “Holographic microscope for measuring
displacements of vibrating microbeams using timeaverage holography,” SPIE,
2860:111121.
C. Brown and R. J. Pryputniewicz, 1995, “Experimental and computational determination
of dynamic characteristics of microbeam sensors,” SPIE, 2545:108119.
C. Brown and R. J. Pryputniewicz, 1992, “Investigation of submillimeter cantilever
beams of variables cross section by computational and hologram interferometry
methods,” Industrial applications of optical inspection, metrology and sensing, SPIE,
1821:7281.
E. Buks and M. L. Roukes, 2001, "Stiction, adhesion energy, and the Casimir effect in
micromechanical systems", Phys. Rev. B, 63(3):34023405.
153
S. Cherian and T. G. Thundat, 2002, "Determination of adsorptioninduced variation in
the spring constant of a microcantilever," Appl. Phys. Lett., 80(12):2219–2221.
H. J. Cho and C. H. Ahn, 2002, “A bidirectional magnetic microactuator using
electroplated permanent magnet arrays”, J. MEMS, 11(1):7884.
E. M. Chow, G. Yaralioglu, C. F. Quate, and T. W. Kenny, 2002, “Characterization of a
twodimensional cantilever array with throughwafer electrical interconnects”, Appl.
Phys. Lett., 80:664.
A. S. Darling, 1958, Platin. Metals Rev., (2):16.
Dartmouth, 2002, “Residual stress examples”, Microelectronics Laboratory Dartmouth
College, Hanover, NH.
EC, 2003, “Material pictures”, http://elementcollection.com, The Red Green & Blue
Company Ltd., London, UK, Last visited: December 3, 2003.
R. P. Feynman, 1992, “There’s plenty of room at the bottom”, J. MEMS, 1(1):6066.
J. Fritz, M. K. Baller, H. P. Lang, H. Rothuizen, P. Vettiger, E. Meyer, H.J. Güntherodt,
C. Gerber, and J. K. Gimzewski, 2000, “Translating biomolecular recognition into
nanomechamics”, Science, 288:316318.
J. Fritz, M. K. Baller, H. P. Lang, T. Strunz, E. Meyer, H.J. Güntherodt, E. Delamarche,
C. Gerber, and J. K. Gimzewski, 2000, “Stress at the solidliquid interface of self
assembled monolayers on gold investigated with a nanomechanical sensor”, Langmuir
Lett., 16:96949696.
C. Furlong, 1999, Hybrid, experimental and computational, approach for the efficient
study and optimization of mechanical and electromechanical components, Ph.D.
Dissertation, Worcester Polytechnic Institute, Worcester, MA.
C. Furlong and R. J. Pryputniewicz, 2002, “Characterization of shape and deformation of
MEMS by quantitative optoelectronic metrology techniques,” Proc. SPIE, 4778:110.
C. Furlong and R. J. Pryputniewicz, 2000, “Absolute shape measurements using high
resolution optoelectronic holography methods,” Opt. Eng., 39:216223.
C. Furlong and R. J. Pryputniewicz, 1995, “Optomechanical study and optimization of a
cantilever plate dynamics”, SPIE, 2545:192203.
154
C. Furlong, J. S. Yokum, and R. J. Pryputniewicz, 2002, “Sensitivity, accuracy, and
precision issues in optoelectronic holography based on fiber optics and highspatial and
highdigital resolution cameras,” Proc. SPIE, 4778:216223.
C. T. Gibson, B. L. Weeks, J. R. I. Lee, C. Abell, and T. Rayment, 2001, “A
nondestructive technique for determining the spring constant of atomic force microscope
cantilevers”, Rev. Sci. Instr., 72(5):23402343.
C. Gormley, A. Boyle, and S. Blackstone, 2000, HARM processing techniques for MEMS
and MOEMS devices using bonded SOI substrates and DRIE, BCO Technologies (NI)
Ltd., Belfast, Northern Ireland, BCO America, Durham, NH, pp. 41744208.
T. Gotszalk, P. Grabiec, and I. W. Rangelow, 2000, “Piezoresistive sensors for scanning
probe microscopy”, Ultramicroscopy, 82:3948.
J. Han, D. P. Neikirk, M. Clevenger, and J. T. McDevitt, 1996, "Fabrication and
characterization of a FabryPerot based chemical sensor," Proc. Conf. on Microelectronic
Structures and MEMS for Optical Processing II, M. E. Motamedi and W. Bailey, editors,
Austin, TX, SPIE2881:171178.
P. H. Handel, 2001, “Quantum 1/f effects in piezoelectric sensors”, Proc. 55th Internatl.
Frequency Control Symp., Seattle, WA, pp. 430437.
D. S. Hanson, T. F. Marinis, C. Furlong, and R. J. Pryputniewicz, 2001, “Advances in
optimization of MEMS inertial sensor packaging,” Proc. Internat. Congress on
Experimental and Applied Mechanics in Emerging Technologies, Portland, OR, pp. 821
825.
J. A. Harley, 2002, Advances in piezoresistive probes for atomic force microscopy, Ph.D
Dissertation, Stanford University, Stanford, CA.
C. Hilbert, H. Curtis, C. Anagnostopoulos, R. Finnila, G. Hawkins, L. Hones, A.
Jacobson, E. Leonard, F. Matta, J. Melendez, and J. Ojha, 2000, MEMS and
microsystems in Europe, International Technology Research Institute.
T.R. Hsu, 2002, MEMS and microsystems: design and manufacture, McGrawHill, New
York, NY.
Z. Hu, T. Thundat, and R.J. Warmack, 2001, “Investigation of adsorption and absorption
induced stresses using microcantilever sensors,” J. Appl. Phys., 90(1):427431.
R. C. Hughes and R. Bastasz, 1988, “Lowenergy proton detection by Pd metalinsulator
semiconductor diodes,” J. Appl. Phys., 64(12):68396844.
155
A. D. L. Humphris, J. Tamayo, and M. J. Miles, 2000, “Active quality factor control in
liquids for force spectroscopy”, Langmuir Lett., 16:78917894.
IBM, 2002, “Cantilever sensors”,
http://www.zurich.ibm.com/st/nanoscience/cantilever.html, IBM Zurich Research
Laboratory, Zurich, Switzerland, Last visited: November 3, 2003.
B. Ilic, D. Czaplewski, M. Zalalutdinov, and H. G. Craighead, 2001, “Single cell
detection with micromechanical oscillators”, J. Vac. Sci. Tech. B., 19(6):28252828.
B. Ilic, D. Czaplewski, and H. G. Craighead, 2000, “Mechanical resonant
immunospecific biological detector”, Appl. Phys. Lett., 77(3):450452.
D. V. Ivanov, 2000, “Advanced sensors for multifunctional applications”, J. Metals,
52(10), http://www.tms.org/pubs/journals/JOM/0010/Ivanov/Ivanov0010.html.
H. Jensenius, J. Thaysen, A. A. Rasmussen, L. H. Veje, O. Hansen, and A. Boisen, 2000,
“A microcantileverbased alcohol vapor sensorapplication and response model”, Appl.
Phys. Lett., 76(18):26152617.
S. K. Jericho and M. H. Jericho, 2002, “Device for the determination of spring constants
of atomic force microscope cantilevers and micromachined springs”, Rev. Sci. Instr.,
73(6):24832485.
H. Johari, 2003, Development of MEMS sensors for measurements of pressure, relative
humidity, and temperature, MS Thesis, Worcester Polytechnic Institute, Worcester, MA.
R. M. Jones, 1975, Mechanics of composite materials, Hemisphere Publishing Corp.,
New York, NY.
S. K. Kassegne, M. J. Madou, R. Whitten, J. V. Zoval, B. Maher, and K. Sarkar, 2002,
“Design issues in SOIbased highsensitivity piezoresistive cantilever devices”, Proc.
Conf. on Smart Structures and Materials, San Diego, CA, SPIE4693:588597.
S. G. Kelly, 1996, Schaum’s Outline of Theory and Problems of Mechanical Vibrations,
McGraw Hill, New York, NY.
N. Kerness, A. Koll, A. Schaufelbühl, C. Hagleitner, A. Hierlemann, O. Brand, and H.
Baltes, 2000, “Nwell based CMOS calorimetric chemical sensors”, IEEE MEMS 2000,
Myazaki, Japan, pp. 96101.
B. H. Kim, F. E. Prins, D. P. Kern, S. Raible, and U. Weimer, 2001a, “Multicomponent
analysis and prediction with a cantilever array based gas sensor”, Sensors and Actuators
B, 78:1218.
156
B. H. Kim, T. D. Chung, C. H. Oh, and K. Chun, 2001b, “A new organic modifier for
antistiction”, J. MEMS, 10(1):3340.
Y.S. Kim, H.J. Nam, S.M. Cho, J.W. Hong, D.C. Kim, and J. U. Bu, 2003, “PZT
cantilever array integrated with piezoresistors sensor for high speed parallel operation of
AFM”, Sensors and Actuators A, 103:122129.
A. Koll, A. Schaufelbühl, N. Schneeberger, U. Münch, O. Brand, and H. Baltes, 1999,
“Micromachined CMOS calorimetric chemical sensor with onchip low noise amplifier,"
Proc. IEEE MEMS 1999, Orlando, FL, pp. 547551.
F. Krüger and H. Jungnitz, 1936, Z. Tech. Phys., (17):302.
H. P. Lang, M. K. Baller, F. M. Battiston, J. Fritz, R. Berger, J.P. Ramseyer, P. Fornaro,
E. Meyer, H.J. Güntherodt, J. Brugger, U. Drechsler, H. Rothuizen, M. Despont, P.
Vettiger, C. Gerber, and J. K. Gimzewski, 1999, “The Nanomechanical NOSE”, Proc.
12
th
IEEE Internatl. Conf. on MEMS, Orlando, FL, 913.
D. Lange, C. Hagleitner, A. Hierlemann, O. Brand, and H. Baltes, 2002,
“Complementary metal oxide semiconductor cantilever arrays on a single chip: mass
sensitive detection of volatile organic compounds”, Anal. Chem., 74:30843095.
D. Lange, C. Hagleitner, O. Brand, and H. Baltes, 2001, “CMOS resonant beam gas
sensing system with onchip self excitation”, Proc. 14
th
IEEE Internat. Conf. on MEMS,
Interlaken, Switzerland, pp. 547552.
F. A. Lewis, 1967, The Palladium hydrogen system, Academic Press, New York, NY.
H. Li, A. Lai, J. Blanchard, and D. Henderson, 2002, “Selfreciprocating radioisotope
powered cantilever”, J. Appl. Phys., 92(2):11221127.
P. Lu, F. Shen, S. J. O’Shea, K. H. Lee, and T. Y. Ng, 2001, “Analysis of surface effects
on mechanical properties of microcantilevers”, Mater. Phys. Mech., 4:5155.
M. J. Madou, 2002, Fundamentals of microfabrication: the science of miniaturization, 2
nd
ed., CRC Press, New York, NY.
E. Manias, J. Chen, N. Fang, and X. Zhang, 2001, “Polymeric micromechanical
components with tunable stiffness”, Appl. Phys. Lett., 79(11):17001702.
A. Mehta, S. Cherian, D. Hedden, and T. Thundat, 2001, “Manipulation and controlled
amplification of brownian motion of microcantilever sensors", Appl. Phys. Lett.,
78(11):16371640.
157
MEMC, 2003, “Silicon wafers”, http://www.memc.com/, Monsanto Electronic Materials
Company, Last visited: November 3, 2003.
MM, 2003, “AFM chips”, http://www.mikromasch.com, MikroMasch USA, Portland,
OR, Last visited: November 24, 2003.
MS, 1998, “Piezo film sensors technical manual: internet version”, Measurements
Specialties, Inc, Valley Forge, PA.
K. Y. Oh and R. J. Pryputniewicz, 1990, “Application of electrooptic holography in the
study of cantilever plate vibration with concentrated masses,” in Hologram
interferometry and speckle metrology, Soc. For Exp. Mech., Bethel, CT, pp. 245253.
L. J. van der Pauw, 1958, "A Method of Measuring Specific Resistivity and Hall Effects
of Discs of Arbitrary Shape", Philips Research Reports, 13:19.
K. E. Petersen, 1982, “Silicon as a mechanical material”, Proc. IEEE, 70(5):420457.
Polytec, 2003, “Polytec Vibrometer”, http://www.polytec.com, Polytec GmbH, Last
visited: December 1, 2003.
T. L. Porter, M. P. Eastman, C. Macomber, W. G. Delinger, and R. Zhine, 2003, “An
embedded polymer piezoresistive microcantilever sensor”, Ultramicroscopy, 97(1
4):365369.
PI, 2003, “Piezo Tech”, http://www.physikinstrumente.de/, Physik Instrumente GmbH &
Co. KG, Last visited: November 3, 2003.
D. R. Pryputniewicz, 1997, ACES approach to the development of microcomponents, MS
Thesis, Worcester Polytechnic Institute, Worcester, MA.
R. J. Pryputniewicz, 2003a, MEMS Packaging, Worcester Polytechnic Institute,
Worcester, MA.
R. J. Pryputniewicz, 2003b, Thermodynamics of microsystems, Worcester Polytechnic
Institute, Worcester, MA.
R. J. Pryputniewicz, 2002a, MEMS SUMMiT™V Technology, Worcester Polytechnic
Institute, Worcester, MA.
R. J. Pryputniewicz, 2002b, MEMS and nanotechnology, Worcester Polytechnic Institute,
Worcester, MA
158
R. J. Pryputniewicz, 1995a, "Quantitative determination of displacements and strains
from holograms," Ch. 3 in Holographic interferometry, Vol. 68 of Springer Series in
Sciences, SpringerVerlag, Berlin, pp. 3372.
R. J. Pryputniewicz, 1995b, “Hologram interferometry from silver halide to silicon and
beyond,” Proc. SPIE, 2545:405427.
R. J. Pryputniewicz, 1993, Engineering experimentation, Worcester Polytechnic Institute,
Worcester, MA.
R. J. Pryputniewicz, 1991a, “Applications of electrooptic holography in vibration
studies,” Proc. of the 1991 SEM Conf. on Experimental Mechanics, Milwaukee, WI, pp.
912919.
R. J. Pryputniewicz, 1991b, “Static and dynamic measurements using electrooptic
holography,” Photomechanics and speckle metrology, F. P. Chen, ed., Bellingham, WA,
pp. 790798.
R. J. Pryputniewicz, 1989, “Measurements of vibration patterns using electrooptic
holography,” Quantitative analysis of holograms, San Diego, CA, 1162:456467.
R. J. Pryputniewicz, 1988, “Vibration studies using heterodyne hologram
interferometry,” Industrial laser interfermetry, SPIE, 955:154161.
R. J. Pryputniewicz, 1986, “Quantitative holographic analysis of small components,”
Proc. SPIE, 604:7185.
R. J. Pryputniewicz, 1985a, “Timeaverage holography in vibration analysis,” Opt. Eng.,
24:843848.
R. J. Pryputniewicz, 1985b, “Holographic and finite element studies of vibrating beams,”
Optics in Engineering Measurements, SPIE, 599:5462.
R. J. Pryputniewicz, 1981, “High precision hologrammetry,“ Internat. Arch.
Photogramm., 24:377386.
R. J. Pryputniewicz, M. P. de Boer, and G. C. Brown, 2000, “Advances in optical
methodology for studies of dynamic characteristics of MEMS microengines rotating at
high speeds,” Proc. IX Internat. Congress on Exp. Mech., SEM, Bethel, CT, pp. 1009
1012.
R. J. Pryputniewicz, C. Furlong, and E. J. Pryputniewicz, 2003a, “Novel optoelectronics
methodology for testing of MOEMS,” Proc. Internat. Symp. on MOEMS and Miniature
Systems, SPIE, 4983:1125.
159
R. J. Pryputniewicz, C. Furlong, and E. J. Pryputniewicz, 2002a, “Design by analysis of a
MEMS pressure sensor,” Proc. 35
th
Internat. Symp. on Microelectronics, Denver, CO,
pp. 8186.
R. J. Pryputniewicz, C. Furlong, and E. J. Pryputniewicz, 2002b, “Optimization of
contact dynamics for an RF MEMS switch,” Paper No. IMECE200239504, Am. Soc.
Mech. Eng., New York.
R. J. Pryputniewicz, C. Furlong, G. C. Brown, E. J. Pryputniewicz, and M. E. Seta,
2001a, “ACES methodology for studies of MEMS,” Proc. 28
th
Annual Symp. and
Exhibition of IMAPSNE, Boxboro, MA, pp. 208217.
R. J. Pryputniewicz, C. Furlong, G. C. Brown, E. J. Pryputniewicz, and M. E. Seta,
2001b, “Optical methodology for static and dynamic measurements of
nanodisplacements,” Proc. Internat. Symp. On MEMS: Mechanics and Measurements,
Portland, OR, pp. 95100.
R. J. Pryputniewicz, T. F. Marinis, J. W. Soucy, P. Hefti, A. R. Klempner, and C.
Furlong, 2003b, “Novel membrane methodology for characterization of holography for
MEMS inertial sensors,” Proc. Internat. Symp. On Microelectronics, Boston, MA, pp.
392397.
R. J. Pryputniewicz, T. F. Marinis, D. S. Hanson, and C. Furlong, 2001c, “New approach
to development of packaging for MEMS inertial sensors,” Paper No.
IMECE2001/MEMS23906, Am. Soc. Mech. Eng., New York, NY.
R. J. Pryputniewicz and K. A. Stetson, 1990, “Measurement of vibration patterns using
electrooptic holography,” SPIE, 1162:456497.
R. J. Pryputniewicz, P. W. Wilkerson, A. J. Przekwas, and C. Furlong, 2002c, “RF
MEMS: modeling and simulation of switch dynamics,” Proc. 35
th
Internat. Symp. on
Microelectronics, Denver, CO, pp. 267272.
PTC, 2003a, “Pro/Engineer Wildfire”, http://www.ptc.com, PTC, Last visited: November
24, 2003.
PTC, 2003b, “Pro/Mechanica”, http://www.ptc.com, PTC, Last visited: November 24,
2003.
S. S. Rao, 1995, Mechanical vibration, 3
rd
ed., AddisonWesley, New York, NY.
R. Raiteri, M. Grattarola, H.J. Butt, and P. Skladal, 2001, “Micromechanical cantilever
based biosensors”, Sensors and Actuators B, 79:115126.
160
S. Rast, C. Wattinger, U. Gysin, and E. Meyer, 2000, “Dynamics of damped cantilevers”,
Rev. Sci. Instr., 71(7):2772.
SNL, 2001, “Micromachines”, http://mems.sandia.gov/scripts/index.asp, Sandia National
Laboratories, Albuquerque, NM Last visited: November 3, 2003.
M. R. Spiegel and J. Liu, 1999, Schaum’s outline series: Mathematical handbook of
formulas and tables, 2
nd
ed., McGrawHill, New York, NY.
A. C. Stephan, T. Gaulden, A.D. Brown, M. Smith, L. F. Miller, and T. Thundat, 2002,
“Microcantilever chargedparticle flux detector”, Rev. Sci. Instr., 73(1):3641.
G. G. Stoney, 1909, Proc. R. Soc. London Ser. A, 82:172177.
V. L. Streeter, E. B. Wylie, and K. W. Bedford, 1998, Fluid Mechanics, 9
th
ed., McGraw
Hill, New York, NY.
A. Subramanian, P. I. Oden, S. J. Kennel, K. B. Jacobson, R. J. Warmack, T. Thundat,
and M. J. Doktycz, 2002, “Glucose biosensing using an enzymecoated microcantilever”,
Appl. Phys. Lett., 81(2)385387.
T. Sulchek, R. Hsieh, J. D. Adams, G. G. Yaralioglu, S. C. Minne, C. F. Quate, J. P.
Cleveland, A. Atalar, and D. M. Adderton, 2000, “Highspeed tapping mode imaging
with active Q control for atomic force microscopy”, Appl. Phys. Lett., 76(11):14731475.
J. Tamayo, A. D. L. Humphris, A. M. Malloy, and M. J. Miles, 2001, “Chemical sensors
and biosensors in liquid environment based on microcantilevers with amplified quality
factor”, Ultramicroscopy, 86:167173.
TI, 2003, “DLP”, http://www.dlp.com/, Texas Instruments. Last visited: November 3,
2003.
J. Thaysen, A. Boisen, O. Hansen, and S. Bouwstra, 2000, "Atomic force microscopy
probe with piezoresistive readout and a highly symmetrical Wheatstone bridge
arrangement", Sensors and Actuators A, 83(13):4753.
J. Thaysen, R. Marie, and A. Boisen, 2001, "Cantileverbased biochemical sensor
integrated in a microliquid handling system", Proc. Symp. of MEMS, Interlaken,
Switzerland, pp. 401404.
J. Thaysen, A. D. Yalcinkaya, R. K. Vestergaard, S. Jensen, M. W. Mortensen, P.
Vettiger, and A. Menon, 2002, “SU8 based piezoresistive mechanical sensor”, Proc. 15
th
IEEE Internat. Conf. on MEMS, Las Vegas, NV, pp. 320323.
161
T. Thundat, E. Finot, Z. Hu, R. H. Richie, G. Wu, and A. Majumda, 2000, “Chemical
sensing in fourier space”, Appl. Phys. Lett., 77(24):40614063.
K. L. Turner and W. Zhang, 2001, “Design and analysis of a dynamic MEM chemical
sensor”, Proc. Am. Control Conf., Arlington, VA, 2:12141218.
M.J. Walker, Comparison of Bosch and cryogenic processes for patterning highaspect
ratio features in silicon, Oxford Instruments plc, UK, pp. 440712.
X.Q. Wang, A. Desai, Y.C. Tai, and T. D. Lee, 1999, “Polymerbased electrospray chips
for mass spectrometry”, IEEE MEMS 1999, Orlando, FL, pp. 523528.
A. Wuttig, R. Riesenberg, G. Nitzsche, 2002, “Integral field and multiobject
spectrometry with MEMS”, Imaging Spectrometry VII, Proc., SPIE4480:367376.
J. Yang, T. Ono, and M. Esashi, 2000, “Surface effects and high quality factor in ultra
thin singlecrystal silicon cantilevers”, Appl. Phys. Lett., 77(23):38603862.
K. Y. Yasumura, T. K. Stowe, E. M. Chow, T. Pfafman, T. W. Kenny, B. C. Stipe, and
D. Rugar, 2000, “Quality factors in micron and submicronthick cantilevers”, J. MEMS,
9(1):117125.
S.J. Yoo, J. Lavigne, S. Savoy, J. B. McDoniel, E. V. Anslyn, J. T. McDevitt, D. P.
Neikirk, and J. B. Shear, 1997, "Micromachined storage wells for chemical sensing beads
in an 'artificial tongue'," Proc. Symp. Micromachining and Microfabrication:
Micromachined Devices and Components III, K. Chau and P. J. French, editors, Austin,
TX, SPIE, 3224:2930.
H. D. Young and R. A. Freedman, 2000, University physics with modern physics, 10
th
ed., AddisonWesley, New York, NY.
X. Yu, J. Thaysen, O. Hansen, and A. Boisen, 2002, “Optimization of sensitivity and
noise in piezoresistive cantilevers”, J. Appl. Phys., 92(10):62966301.
162
APPENDIX A. PROPERTIES OF MATERIALS USED
For the materials used in this Thesis, the properties of each are tabulated below,
Table A.1. The values listed in the table are the ones used in this Thesis unless otherwise
specified in the text. In addition, pictures of what the materials look like are shown in
Figures A.1 to A.3. These pictures are of small (no scale) samples from The Red Blue &
Green Company (EC, 2003) for the purpose of a visual association to each element.
Table A.1. Physical properties of the materials used.
Fig. A.1. A photograph of palladium samples (EC, 2003).
Material name Palladium Hydrogen Titanium Silicon
Atomic number 46 1 22 14
Atomic weight, g 106.42 1.00794 47.867 28.0855
Crystal structure fccp hcp hcp diamond
Density, kg/m
3
12023 no data 4507 2330
Melting point, °K 1828.05 14.01 1941 1687
Boiling point, °K 3236 20.28 3560 3173
Modulus of elasticity, GPa 121 no data 116 190
Poisson's ratio 0.39 no data 0.32 0.22
Reflectivity 72% no data no data 28%
163
Fig. A.2. A sample of titanium, a popular material due to
its corrosion resistance (EC, 2003).
Fig. A.3. A sample of untreated, unprocessed, uncut silicon (EC, 2003).
164
APPENDIX B. ANALYTICAL CALCULATIONS
In this Appendix, almost very equation used in the main text, Chapters 4 and 5, is
derived stepbystep (Kelly, 1996). This is for validation of work and completeness.
B.1. Fundamental resonance frequency
In determining the fundamental resonance frequency of a cantilever we start with
the general equation for a body in motion, Fig. 4.2,
2
2
( , )
d x dx
m c kx F x t
dt dt
+ + = (B.1)
where m is the dynamic mass of the cantilever, c is the damping coefficient, k is the
spring constant, and F(x,t) is the time dependent forcing function acting on it. If we
assume that there is no damping and no force acting on the cantilever, Eq. B.1 simplifies
to
2
2
0
d x
m kx
dt
+ = (B.2)
which is a simple homogeneous equation. The general solution for Eq. B.2 is
1 2
( ) cos( ) sin( ) X t C t C t ω ω = + (B.3)
 
2
2
1 2
2
( ) cos( ) sin( ) ,
d
X t C t C t
dt
ω ω ω = − + (B.4)
where C
1,2
are arbitrary constants. Substituting Eqs B.3 and B.4 into Eq. B.2, it is
obtained
165
( )  
2
1 2
cos( ) sin( ) 0 m k C t C t ω ω ω
(
− + + =
¸ ¸
(B.5)
which produces two results. One is trivial and the other is equal to
,
k
m
ω = (B.6)
which can be simplified into the following form
1
2
k
f
m π
= (B.7)
by using the equality of
,
2
f
ω
π
= (B.8)
where f is the frequency of the body in motion and ω is the angular frequency. Equation
B.8 can be further simplified by substituting Eqs B.24 and B.33 to obtain
3
3
1
4
33
2
140
Ebh
L
f
bhL
π
ρ
= (B.9)
which becomes
2
1 35
.
2 33
h E
f
L π ρ
= (B.10)
B.2. Spring constant
The general equation for the cyclic frequency of a vibrating body, given by Eq.
B.7, is
166
1
,
2
k
f
m π
= (B.11)
where m is the dynamic mass of the structure and k, the spring constant, is
,
F
k
y
= (B.12)
in which F is the force applied to the structure. If we assume a simple cantilevertype
structure, Fig. B.1, with an applied force, the general equation of the elastic curve is
2
2
,
x
M d y
EI dx
= (B.13)
where M
x
is the moment acting along the cantilever, E is the modulus of the elasticity,
and I is the moment of inertia.
Fig. B.1. Geometric parameters of a cantilever.
However, M
x
can be represented as
( ) .
x
M F L x = − (B.14)
167
Inserting Eq. B.14 into Eq. B.13 and then integrating we get
2
1
2
dy F x
Lx C
dx EI
 
= − +

\ .
(B.15)
and by intergrating Eq. B.15 we obtain
2 3
1 2
.
2 6
F x x
y L C x C
EI
 
= − + +

\ .
(B.16)
In Eq. B.16, we have two unknowns. We can determine them by utilizing boundary
conditions for a cantilever. The boundary conditions are
0 0
0 0 .
dy
at x
dx
y at x
= =
= =
(B.17)
Solving Eqs B.16 and B.15 with the boundary conditions of Eq. B.17 we find that
1 2
0 and 0 . C C = = (B.18)
Substituting Eq. B.18 into Eq. B.15 we find that
( )
2
2 .
2
dy F
Lx x
dx EI
= − (B.19)
Similarly, Eq. B.16 becomes
( )
2 3
3 .
6
F
y Lx x
EI
= − (B.20)
If we now take Eq. B.20 and solve it at x=L, we get the following equation that represents
the maximum displacement
( )
3
3 3
max
3 .
6 3
F FL
y y L L
EI EI
= = − = (B.21)
Inserting the result of Eq. B.21 into Eq. B.12 we get
168
3
max
3
,
F EI
k
y L
= = (B.22)
where I is the moment of inertia for the cantilever, i.e.,
3
.
12
bh
I = (B.23)
When Eq. B.23 is inserted into Eq. B.22 we finally get
3
3
,
4
Ebh
k
L
= (B.24)
which is the stiffness for a cantilever shown in Fig. B.1.
B.3. Dynamic mass
To determine the dynamic mass of a vibrating cantilever we must consider the
kinetic energies involved. A vibrating mass will affect these energies depending on the
boundary conditions. With this in mind we start with the basic equation for a kinetic
energy
, KE dKE =
∫
(B.25)
where the kinetic energy differential dKE can be substituted by
2 2
,
2 2
y y
dKE dm bhdx ρ = =
& &
(B.26)
where ( y& )
2
is the velocity of the vibrating body with displacement y
max
based on the
boundary conditions. Equation B.26 is assuming that the beam has a constant cross
section in that the variables b and h are not changing as a function of position along the
169
cantilever. Assuming proportionality, we start by dividing Eq. B.21 into Eq. B.20 and
getting the following form
2 3
3
max
3
,
2
y Lx x
y L
−
= (B.27)
which translates into
2 3
max
3
3
.
2
Lx x
y y
L
  −
=

\ .
(B.28)
After taking the first derivative with respect to time, Eq. B.28 becomes
2 3
max 3
3
.
2
Lx x
y y
L
  −
=

\ .
& & (B.29)
Using Eqs B.25, B.26, and B.29 and applying limits of integration we get
2
2 2 3
max
3
0
3
.
2 2
L
y Lx x
KE bhdx
L
ρ
  −
=

\ .
∫
&
(B.30)
Solving Eq. B.30 we get
2
max
33
.
140 2
y
KE bhL ρ
 
=

\ .
&
(B.31)
Since the general equation for kinetic energy is
2
1
2
KE mv = (B.32)
we find that the quantity in parenthesis in Eq. B.31 is equivalent to
33 33
140 140
bhL M m ρ = = (B.33)
where M is the total, i.e., static mass of the beam and m is the dynamic mass of the
cantilever.
170
B.3.1. Dynamic mass constant
If we look at the result of Eq. B.33 we see that the mass has a constant of 33/140
in front of it. This constant describes the effect that mass will have on the cantilever. To
determine the effect, or the constant, at an arbitrary point along the cantilever we start
with Eq. B.30, but modify it slightly to
2
2 2 3
max
3
0
3
,
2 2
r
y Lx x
KE bhdx
L
ρ
  −
=

\ .
∫
&
(B.34)
where the upper limit of integration will be replaced with the parameter r to signify the
location along the cantilever that we are interested in. If we now solve Eq. B.34, it
becomes
( ) ( )
2 5
2 2 max
6
1
5 35 63 ,
2 140
y r
KE bh r Lr L
L
ρ
(
= ⋅ − +
(
¸ ¸
&
(B.35)
and we can equate
( )
( )
1
5
2 2
6
1
5 35 63 ( ) ,
140
bh C
r
r Lr L n r
L
ρ =
(
− + =
(
¸ ¸
(B.36)
where C
1
is just a constant of structural properties and n(x) is a function that will produce
a the effective, or dynamic, mass constant along the length of the cantilever. We can then
rewrite Eq. B.35 as
2
max
1
( ) .
2
y
KE C n x =
&
(B.37)
171
B.4. Sensitivity calculations of the cantilever
In this section, the equations for the sensitivity of the cantilever will be derived
for two cases: a cantilever partially and a completely covered with palladium. The
explicit quantitative solutions are in Section 5.8.
B.4.1. Sensitivity of an endloaded cantilever
In determining sensitivity of a frequencybased sensor we can use Eq. 4.1,
repeated here for continuity, i.e.,
0
1 1
lim ,
m
m
i i
f f
S
f m f m
∆ →
∆ ∂
= =
∆ ∂
(B.38)
where sensitivity is a measure of normalized frequency over mass with f
i
equaling the
initial resonant frequency. In order to solve Eq. B.38 for an endloaded case to determine
S
e
, there are a few substitutions that must be made. First, we make the appropriate
substitutions for frequency, mass, and change in mass into the equation so that it becomes
( )
1 1
,
2
1
2
e
r
df k
S
d m m n r m
k
m
π
π
 
= 

∆ + ∆
\ .
(B.39)
where m is the dynamic mass of the cantilever defined in Eq. B.33, and ∆m is the added
mass. The parameter r is the percentage coverage of the cantilever with palladium so that
it can be directly compared with a fully coated cantilever later and the parameter n is the
effective mass constant, or boundary condition mass factor for a cantilever, (i.e.,
n=33/140 for r = 100%) of the added mass that is dependent on the value of r. By
172
assigning the variable r a value of 0.25, equaling 25% coverage of the cantilever, and
using Fig. 5.13 to determine the effective mass at this value we find the effective mass
equal to n
r
= 71/420 (~0.169). Equation B.39 can then be solved, i.e. derivative taken
and solution simplified, and the final equation takes the form of
2
1
.
33
2
33
33
140
140 140
e
r
r
knr
S
k k
M n r m
M M n r m
δ
δ
−
=
 
+

 
\ .
+

\ .
(B.40)
If we assign nominal values to the variables in Eq. B.40, we can solve for a
sample case and determine the sensitivity for it in order to compare the sensitivity of the
partially and completely covered cantilever. Using the values in Table B.1, we can solve
for the sensitivity of a 350 µm long cantilever.
Table B.1. Material properties and geometric
dimensions used in a sample case to determine
sensitivity of the cantilever as a chemical sensor.
Using the values from Table B.1 we find
( )
3
8
2.33 / 35 1 350
2.854 10 ,
M bhL
g cm m m m
g
ρ
µ µ µ
−
=
= ⋅ ⋅
= ⋅
(B.41)
190
2.33
1
350
35
Length, L µm
Width, b µm
Parametric values used
Modulus of elasticity, E , GPa
Density, ρ , g/cm
3
Thickness, h µm
173
which is the mass of the cantilever. To set a value for the change in mass, i.e., added
mass, an arbitrary number shall be picked. The ∆m, or change in mass, will be assumed
to be 3 orders of magnitude less than the cantilever mass itself or
11
1 10 , m g
−
∆ = ⋅ (B.42)
which will be held as a constant, or in other words the addition of mass will be equal to
this value and will not change. For the value of the spring constant, k, Eq. B.24 yields
( )
( )
3
3
3
3
4
190 35 1
4 350
0.039 .
Ebh
k
L
GPa m m
m
N
m
µ µ
µ
=
⋅ ⋅
=
=
(B.43)
Then using the values of n
r
= 71/420 and r = 0.25 we can determine the sensitivity, as
was solved in Section 5.8, of an end loaded cantilever with Eq. B.40 to be
9
3.141 10 ,
e
Hz
S
Hz kg
= − ⋅
−
(B.44)
which is in units of changing frequency per unit frequency – mass.
B.4.2. Sensitivity of a completely coated cantilever
In determining the sensitivity of a uniformly distributed polymer over a cantilever
we start with Eq. B.38. The equation for a uniformly distributed load, with the same
174
substitutions for frequency, mass, and changing mass as done in Eq. B.39, will take the
form of
( )
1 1
,
33
2
1
140
33
2
140
d
df k
S
d m
k
m m
m
π
π
 

= 
∆

+ ∆

\ .
(B.45)
which contains parameters that were all defined previously. If Eq. B.45 is solved, i.e., the
derivative and resultant equation are simplified to take the final form
( )
( )
2
1
,
33
2
140
33 33
140 140
d
k
S
k k
m m
m m m
−
=
+ ∆
+ ∆
(B.46)
which is similar in formulation to Eq. B.40.
Using the values of Table B.1 and from Eqs B.41 through B.43, Eq. B.46, yields
10
1.751 10 ,
d
Hz
S
Hz kg
= − ⋅
−
(B.47)
which is in units of changing frequency per unit frequency – mass. As can be seen, the
sensitivity of the distributed load is about one order of magnitude greater than for an end,
or tip, loaded cantilever.
B.5. Frequency shift for a free, undamped vibration
To determine the frequency shift for an ideal, i.e. free and undamped, system the
following identity is used
175
2 2 2 2
1 2 1 2
2 2
2 2
,
f f f f
f f
− −
= (B.48)
which can be rewritten, by substituting Eq. B.7 and realizing that frequency relation
incorporates a change in mass as,
( )
( )
2
2
2 2
1
1 2 2
2 2
2
2
1 1
2 2
.
1
2
k k
m m m
f f
f
k
m m
π π
π
 
 
− 


+ ∆
− \ .
\ .
=
 


+ ∆
\ .
(B.49)
Equation B.49 can then be simplified into
( )
( )
( )
2 2
1 2
2
2
1
k k
m m m
f f m m m m m m
k
m m m m f
m m
−
+ ∆
− + ∆ − + ∆ ∆
= = ⋅ =
+ ∆
+ ∆
(B.50)
and further to
2
1 2
2
2 2
f f m
f m f
δ
= + (B.51)
and finally reformulated, by substituting Eq. B.10 in for f
1
, as
2
2
1 35
2 33
.
1
eq
eq
E
h
L
f
m
m
π ρ
δ
=
+
(B.52)
Equation B.52 can be used to determine the frequency shift due to the absorption of
hydrogen.
176
B.6. Frequency shift for a free, undamped vibration with a changing stiffness
To determine the frequency shift for a free, undamped system that has a spring
constant that changes with absorption, a different formulation will have to be used than
the one used in Section B.5. For this case, Eq. B.48, can be written, with the appropriate
frequency substitutions that include both a changing mass and stiffness, as
( )
( )
( )
( )
2
2
2 2
1
1 2 2
2 2
2
2
1 1
2 2
,
1
2
k k
k
m m m
f f
f
k k
m m
δ
π π δ
δ
π δ
 
  +
 −


 +
\ .
−
\ .
=
 
+

 +
\ .
(B.53)
which can then be simplified into
( )
( )
( )
( )
( )
( )
( ) ( )
2 2
1 2
2
2
k k
k
m m m m m
f f k m m k k m m k
k k m m m k k m k k f
m m
+ ∆
−
+ ∆ + ∆
− ∆ − ∆ ∆ − ∆
= = ⋅ =
+ ∆ + ∆ + ∆ + ∆
+ ∆
(B.54)
and further to
( )
2 2
2 2 1 1
2 1
1
f f mk m k
f f
k m m k k m mk
k m mk
m k k mk m k
+ ∆
= = =
∆ − ∆ ∆ +
∆ +
+
+ ∆ + ∆
(B.55)
and reformulated, by substituting Eq. B.10 in for f
1
, as
2
2
2
1
1
1 35
.
2 33
1
eq
eq
k
E
h k
f
m L
m
δ
δ π ρ
 
+

 
\ .
 =

 
\ .
+

\ .
(B.56)
177
Equation B.56 can be used to determine the frequency shift due to the absorption of
hydrogen.
B.7. Free vibration with viscous damping
In determining the resonance frequency of a cantilever with damping present we
start with the general equation for a body in motion, i.e.,
2
2
( , ) .
d x dx
m c kx F x t
dt dt
+ + = (B.57)
If we assume that there is damping and no force acting on the cantilever, Eq. B.57
simplifies to
2
2
0 .
d x dx
m c kx
dt dt
+ + = (B.58)
The general solution for Eq. B.58 is assumed as
( ) .
st
X t Ae = (B.59)
If we substitute Eq. B.59 into Eq. B.58 we get
2
0 . ms cs k + + = (B.60)
Solving Eq. B.60 for its roots produces a quadratic solution, which when simplified takes
the form of
2
1,2
,
2 2
c c k
s
m m m
 
= − ± −

\ .
(B.61)
which can then be substituted back into Eq. B.57 giving a solution with two parts
178
1 2
1 2
( ) .
s t s t
X t Ae A e = + (B.62)
In this Thesis, the cantilever system will be operating in air and thus it will be assumed
that it resembles an under damped system (Rao, 1995). For an under damped system, the
radical of Eq. B.61 is
2
1
2
c k
m m
 
− <

\ .
(B.63)
and thus the radical will be negative. Solving Eq. B.62 for the condition given by Eq.
B.63 we simplify the equation with trigonometric identities to
( )
( )
/ 2
( ) cos ,
c m t
d
X t Ae t ω θ
−
= + (B.64)
where A is the amplitude, t is time, θ is the phase shift, and ω
d
is the damped angular
frequency
2
2
,
4
d
k c
m m
ω = − (B.65)
which indicates that damping alters the resonant frequency by changing the damped
angular frequency.
B.7.1. Frequency shift for a free, damped vibration
To determine the frequency shift for a free, damped system the same form as Eq.
B.48 is used. The difference is that in substituting in the functions, Eq. B.65 will be used
in the shifted frequency so that the formulation becomes
179
( )
( )
( )
( )
2
2
2
2
2 2
1
1 2 2
2 2
2 2
2
2
1 1
2 2
4
.
1
2
4
k k c
m m m
m m
f f
f
k c
m m
m m
π π δ
δ
π δ
δ
 
 

− −


 +
+
\ .
−
\ .
=
 

−
 +
+
\ .
(B.66)
Equation B.66 does not simplify as cleanly as Eq. 56, but after some operation, it can be
put into the following form:
( )
( )
( )
( )
( )
( )
2
2
2
2 2
1 2
2 2
2
2
2
4 4
4
4
k k c
m m m
m m k m m m mc
f f
f k c
m k m m c
m m
m m
− −
+ ∆
+ ∆ + ∆ ∆ −
−
= =
( ⋅ + ∆ −
¸ ¸
−
+ ∆
+ ∆
(B.67)
and further to
( )
( )
( )
( )
2
2
2 2 1
2 1
2 2
2
4
4
4
1
4
m k m m c
f
f f
k m m m mc
k m m
mk m m c
δ
δ
( + −
¸ ¸
= =
+ ∆ ∆ −
+
+
+ ∆ −
(B.68)
and finally reformulated, by substituting Eq. B.10 in for f
1
, as
( )
( )
2
2
2
2 2
1
4
1 35
.
2 33
4
eq
eq
m k m m c
E
h
f
L
k m m
δ
π ρ
δ
  ( + −
¸ ¸
 =

+
\ .
(B.69)
Equation B.69 can be used to determine the frequency shift due to the absorption of
hydrogen.
180
B.7.2. Frequency shift for a free, damped vibration with a changing stiffness
To determine the frequency shift for a free vibration system that has damping and
a changing spring constant the same form as Eq. B.48 is used. The difference is that
there will also be the altering stiffness so the equation, incorporating a rewritten Eq. B.65,
becomes
( )
( )
( )
( )
( )
( )
2
2
2
2
2 2
1
1 2 2
2 2
2 2
2
2
1 1
2 2
4
.
1
2
4
k k
k c
m m m
m m
f f
f
k k
c
m m
m m
π π
π
 
  + ∆

− −


 + ∆
+ ∆
\ .
−
\ .
=
 
+ ∆

−
 + ∆
+ ∆
\ .
(B.70)
This equation simplifies much like Eq. 69, so it can be put into the following form:
( )
( )
( )
( )
( )
( )
( )( )
( )( )
2
2
2
2 2
1 2
2
2 2
2
2
4 4
4
4
k k
k c
m m m
m m k m m k m m mc
f f
k k f m k k m m c c
m m
m m
δ
δ
δ
δ
δ
δ
+
− −
+
+ ∆ − ∆ + ∆ −
−
= =
+ ( + ∆ + ∆ −
¸ ¸
−
+
+
(B.71)
and further to
( )( )
( )( )
( )
( )( )
2
2
2 2 1
2 1
2 2
2
4
4
4
1
4
m k m m c
f
f f
k m m k m m mc
k k m m
m k k m m c
δ
δ δ
( + −
¸ ¸
= =
∆ − ∆ + ∆ −
+ +
+
( + ∆ + ∆ −
¸ ¸
(B.72)
and finally reformulated, by substituting Eq. B.10 in for f
1
, as
( )
( )( )
2
2
2
2 2
1
4
1 35
.
2 33
4
eq
eq
m k m m c
E
h
f
L
k k m m
δ
π ρ
δ δ
  ( + −
¸ ¸
 =

+ +
\ .
(B.73)
Equation B.73 can be used to determine the frequency shift due to the absorption of
hydrogen, as discussed in Section 3.2.
181
B.8. Harmonically excited system with viscous damping
In determining the resonance frequency of a cantilever under harmonic excitation
we start with the general equation for a body in motion
2
2
cos( ) .
d x dx
m c kx F t
dt dt
ω + + = (B.74)
If we look at the particular solution for this case, we can assume a solution of the form
( ) cos( ) . X t A t ω θ = − (B.75)
The homogeneous solution to Eq. B.75 will not play a part, as it will be zero under
equilibrium conditions. By substituting Eq. B.75 into Eq. B.74 we get
( )
2
cos( ) sin( ) cos( ) . X k m t c t F t ω ω θ ω ω θ ω
(
− − − − =
¸ ¸
(B.76)
Using trigonometric relations
cos( ) cos( ) cos( ) sin( ) sin( )
sin( ) sin( ) cos( ) cos( ) sin( ) ,
t t t
t t t
ω θ ω θ ω θ
ω θ ω θ ω θ
− = +
− = −
(B.77)
we can simplify Eq. B.76 into two forms
( )
( )
2
2
cos( ) sin( )
sin( ) cos( ) 0 .
X k m c F
X k m c
ω θ ω θ
ω θ ω θ
(
− + =
¸ ¸
(
− − =
¸ ¸
(B.78)
By solving Eq. B.78 simultaneously, it is obtained (Hsu, 2002)
( )
2
2 2 2
F
X
k k m c ω ω
=
− +
(B.79)
where the solution for the phase is
1
2
tan .
c
k m
ω
θ
ω
−
 
=

−
\ .
(B.80)
182
If the following equations:
, ,
2
n
n n
k c
and r
m m
ω
ω ζ
ω ω
= = = (B.81)
are substituted into Eqs B.79 and B.80, they can be rewritten as
( ) ( )
max
2
2
2
1 2
F
X
k r r ζ
=
− +
(B.82)
where X is the amplitude and the phase is
1
2
2
tan ,
1
r
r
ζ
θ
−
 
=

−
\ .
(B.83)
which allows Eqs B.82 and B.83 to be functions of the frequency ratio r.
183
APPENDIX C. UNCERTAINTY ANALYSIS
In this Appendix the uncertainty calculations used in this Thesis are performed.
C.1. Uncertainty in fundamental resonant frequency
In this Section the calculations of the uncertainty in frequency for the ideal,
uncoated cantilever are shown.
C.1.1. Uncertainty calculations for the 350 µm long cantilever
To determine the overall percentage uncertainty in the resonant frequency for the
averaged, uncoated case, we start with the general relationship for fundamental
frequency, based on Eq. 4.23, i.e.,
2
1 35
,
2 33
h E
f
L π ρ
= (C.1)
where h is the thickness of the cantilever, L is the length, E is the modulus of elasticity,
and ρ is the density. The values used for this calculation are listed in Table C.1 and
include the averaged geometric dimensions measured for these cantilevers.
184
Table C.1. Values of parameters used in uncertainty
analysis.
If the values listed in Table C.1 are inserted into Eq. C.1 we get
( )
( )
2
2
3
1 35
2 33
190 1 0.985 35
2 33
353.2
2.33
11690.232 Hz .
h E
f
L
GPa m
g
m
cm
π ρ
µ
π
µ
=
=
 

\ .
=
(C.2)
To determine the overall percentage uncertainty in the resonant frequency, the RSS (root
sumsquares) method is used. The phenomenological equation, corresponding to Eq. C.1,
is
( , , , ) . f h L E ρ = (C.3)
Using Eq. C.3 we write the equation for the overall percentage uncertainty as
1
2 2 2 2 2
,
df df df df
f h L E
dh dL dE d
δ δ δ δ δρ
ρ
(
 
     
= + + + (
   
\ . \ . \ .
\ . (
¸ ¸
(C.4)
where δh, δL, δE, and δρ are the uncertainties in each independent parameter defining the
resonance frequency of the cantilever. The values used for each of these parameters are
listed in Table 5.4, and are also repeated in Table C.2, for completeness of this Section.
190
2.33
0.9853
353.2 Length, L µm
Parametric values used
Modulus of elasticity, E , GPa
Density, ρ , g/cm
3
Thickness, h µm
185
Table C.2. The uncertainties in each parameter
considered.
Contributions by the uncertainties of each independent parameter to the overall
uncertainty in f are
( )
( )
1 35
2
2 33
1 1 35
2
2 33
190
1 1 35
0.05
2
2 33
353.2
2.33
3
593.232
df df h E
h h
dh dh
L
E
h
L
GPa
m
g
m
cm
δ δ
π ρ
δ
π ρ
µ
π
µ
(
 
 
= ⋅
( 

\ .
(
\ .
¸ ¸
 
= ⋅

\ .
 


= ⋅

 



\ .
\ .
=
(C.5)
330.981
df
L
dL
δ
 
=

\ .
(C.6)
92.291
df
E
dE
δ
 
=

\ .
(C.7)
12.543 .
df
d
δρ
ρ
 
=

\ .
(C.8)
Substituting Eqs C.5 to C.8 into Eq. C.4, the overall uncertainty in f is calculated to be
δ h =0.05
µ
m δ L =5
µ
m δ E =3 GPa δρ =0.005 g/cm
3
Uncertainty in parameters defining f
186
( ) ( ) ( ) ( )
1
2 2 2 2 2
1
2 2 2 2 2
593.232 330.981 92.291 12.543
685.673 .
df df df df
f h L E
dh dL dE d
Hz
δ δ δ δ δρ
ρ
(
 
     
= + + + (
   
\ . \ . \ .
\ . (
¸ ¸
(
= + + +
¸ ¸
=
(C.9)
Once the overall uncertainty in the frequency is found, it is possible to determine the
percentage uncertainty in f as
% 100
685.673
100
11690.232
5.865% ,
f
f
f
δ
δ = ⋅
= ⋅
=
(C.10)
which is fairly high.
To discover the source of the uncertainty in Eq. C.10, it is possible to calculate the
percent contribution by each parameter listed in Eq. C.3, as
( )
( )
( )
2
2
2
2
% 100
593.232
100
685.673
74.854%
df
h
dh
f h
f
δ
δ δ
δ
 

\ .
= ⋅
= ⋅
=
(C.11)
% 23.301% f L δ δ = (C.12)
% 1.812% f E δ δ = (C.13)
% 0.033% , f δ δρ = (C.14)
187
where it can be seen that the thickness, h, contributes the most with 74.8% to the overall
percentage uncertainty in the frequency. It should be noted that all the percent
contributions, Eqs C.11 thru C.14, add up to 100%, as they should. Therefore, the most
effective way to reduce the overall uncertainty in f is to improve fabrication processes
and decrease the δh.
C.1.2. Uncertainty calculations for the 300 µm long cantilever
The determination method of the overall percent uncertainty in the resonance
frequency for this shorter cantilever is the same as with the 350 µm long one. Thus, the
values from Tables C.1 and C.2 are used with the exception of the length value, which is
now shorter at L=303.53 µm long and the explicit solutions are the same as those in
Section C.1.1. Using Eq. C.4, the contributions due to uncertainties of each individual
parameter to the overall uncertainty in f are
803.272
df
h
dh
δ
 
=

\ .
(C.15)
521.506
df
L
dL
δ
 
=

\ .
(C.16)
124.968
df
E
dE
δ
 
=

\ .
(C.17)
16.984
df
d
δρ
ρ
 
=

\ .
(C.18)
188
Substituting Eqs C.15 to C.18 into Eq. C.4, the overall uncertainty in f is calculated,
following the same formulation as C.9, to be
965.982 f Hz δ = (C.19)
which is a higher value than that of the 350 µm long cantilever. Once the overall
uncertainty in the frequency is found to 965.982 Hz, the percent overall uncertainty in f is
determined to be
% 6.102% , f δ = (C.20)
which is higher than with the 350 µm long cantilever.
To discover the source of the uncertainty in Eq. C.20, it is possible to calculate the
percent contribution of each parameter listed in Eq. C.3 following the same formulation
as Eq. C.11, as
% 69.149% f h δ δ = (C.21)
% 29.146% f L δ δ = (C.22)
% 1.674% f E δ δ = (C.23)
% 0.031% , f δ δρ = (C.24)
where it can be seen that the thickness, h, once again contributes the most with 69.1% ,
although less than for the 350 µm long cantilever, to the overall percentage uncertainty in
the frequency. The total sum of the percentages adds up to 100%, as they should.
189
C.1.3. Uncertainty calculations for the 250 µm long cantilever
The final determination of the overall percent uncertainty in the resonance
frequency for the 250 µm long cantilever will be the same as in Sections C.1.1 and C.1.2.
Thus, the values from Tables C.1 and C.2 are used with the exception of the length value,
which is now even shorter at L=252.99 µm long and the explicit solutions are the same as
those in Section C.1.1. Using Eq. C.4, the contributions due to uncertainties of each
individual parameter to the overall uncertainty in f are
1156.27
df
h
dh
δ
 
=

\ .
(C.25)
900.647
df
L
dL
δ
 
=

\ .
(C.26)
179.885
df
E
dE
δ
 
=

\ .
(C.27)
24.448 .
df
d
δρ
ρ
 
=

\ .
(C.28)
Substituting Eqs C.25 to C.28 into Eq. C.4, the overall uncertainty in f is calculated,
following the same formulation as C.9, to be
1476.848 , f Hz δ = (C.29)
which is higher than that of both the 300 and 350 µm long cantilevers. This is because of
the shorter length of cantilever D and higher effect the uncertainty in length has. Once
the overall uncertainty in the frequency is found to be 1474.848 Hz, the percent overall
uncertainty in f is determined to be
% 6.482% , f δ = (C.30)
190
which is higher than both the 300 and 350 µm long cantilever.
To discover the source of the uncertainty in Eq. C.10, it is possible to calculate the
percent contribution of each parameter listed in Eq. C.3 following the same formulation
as Eq. C.11, as
% 61.298% f h δ δ = (C.31)
% 37.191% f L δ δ = (C.32)
% 1.484% f E δ δ = (C.33)
% 0.027% , f δ δρ = (C.34)
where it can be seen that the thickness, h, for a third time contributes the most with 61%
to the overall percentage uncertainty in the frequency. The total sum of the percentages
adds up to 100%, as they should.
C.1.4. Average overall percentage uncertainty in frequency
Once all the overall percentage uncertainties were found for each cantilever of
different length, they were averaged for an overall percentage. Taking the results of Eqs
C.10, C.20, and C.30 we obtain
5.865 6.102 6.482
6.15%
3
+ +
 
=

\ .
(C.35)
and thus the overall percentage uncertainty in the resonant frequency for these cantilevers
is 6.15%.
191
C.2. Frequency uncertainty in a palladium coated cantilever
Sections C.2.1 through C.2.4 discuss the uncertainty in the frequency for several
different cases for a palladiumcoated cantilever absorbing a 1% hydrogen concentration
at 20˚C and 1 atm pressure. The concentration will not be considered for anything above
1%. For each case as described in Sections 5.11.1 through 5.11.4 the overall uncertainty,
percent uncertainty, and contributing variables uncertainties will be calculated for an
ideal cantilever of L=350 µm long, h=1 µm thick, and b=35 µm wide. Table C.3 presents
the values used for this analysis.
Table C.3. Material properties and thickness of the composite cantilever.
C.2.1. Uncertainty in frequency for a free vibrations system
The uncertainty in frequency for a free, undamped vibrating system will be
determined in the equation for the frequency shift. Equation 5.11, which relates to the
frequency shift due to hydrogen absorption, is
Density
r (g/cm
3
)
Modulus of elasticity
E (GPa)
Thickness
h (mm)
Silicon, Si 2.330 190 1
Palladium, Pd 12.023 121 0.1
Titanium, Ti 4.507 116 0.001
Equivalent values 3.995 172.9 1.101
192
2
2
35
1
2 33
1
eq
eq
E
h
L
f
m
m
π ρ
=
∆
+
(C.36)
where ∆m is the increase in mass, m is the initial dynamic mass, E
eq
is the equivalent
modulus as defined in Eq. 4.6, and ρ
eq
is the equivalent density as defined in Eq. 4.7. The
values for the equivalent modulus and density can be found in Table C.3. The increase in
mass, ∆m, can be written as
( )
( )( )
2
_ _ ,
Pd
m absorption multiplier H density V ∆ = (C.37)
where it is desirable to write each component in terms of the concentration in units of
ppm. Thus, based on Table 4.1,
2
3
2
3
_ 1.25 10 0.621
10000
10000
_ 1.25 10 0.621
10000
_ 776.25
ppm
absorption multiplier
absorption multiplier
absorption multiplier
 
= ⋅

\ .
 
= ⋅

\ .
=
(C.38)
converts the atomic ratio into the equivalent multiplier factor of the volume of hydrogen
in the palladium in terms of the concentration. Using Eq. 4.9, density of hydrogen at 1%
concentration, 20˚C, and 1 atm (0.01 atm partial pressure) is written as
193
( ) ( )
( )
2
3
2
3
2
3
_ 273
_
0.0224 1
10000 2 1.00794 273 0.01
_
0.0224 293 1
_ 8.385 ,
ppm molecular weight p
H density
m
T atm
mol
g
ppm K atm
mol
H density
m
K atm
mol
kg
H density
m
 

⋅ ⋅ ⋅
 =

⋅ ⋅

\ .
   
⋅ ⋅ ⋅ ⋅
 
\ .

=

⋅ ⋅

\ .
=
(C.39)
where ppm is the gas concentration, molecular_weight is the molecular weight of the
diatomic hydrogen, and p and T are the partial pressure (in atm) and temperature (in °K)
of the gas, respectively. Since the concentration is directly related to the partial pressure,
p, it can be written in terms of concentration as
6
1 ,
10
ppm
p atm = ⋅ (C.40)
when the total pressure of the gas mixture, in this case H
2
and N
2
, is equal to 1 atm. The
last part of Eq. C.37, V
Pd
, is the volume of palladium deposited on the cantilever and it
can be written as a function of just the cantilever dimensions and the layer thickness of
Pd and Ti as
3
35 1 350
12, 250 ,
V bhL
m m m
m
µ µ µ
µ
=
= ⋅ ⋅
=
(C.41)
( )( )( )
( ) ( ) ( )
3
3
2 2
35 2 0.001 1 2 0.001 350 2 0.001 12, 250
25.236 ,
Ti
Ti Ti Ti
V b h h h L h V
m m m m
m
µ µ µ µ
µ
(
= + + + −
(
¸ ¸
( = + ⋅ ⋅ + ⋅ ⋅ + ⋅ −
¸ ¸
=
(C.42)
194
( ) ( ) ( ) ( )( )
( ) ( ) ( ) ( )
( )
2 2
35 2 0.001 0.1 1 2 0.001 0.1
350 0.001 0.1
14, 813.755 ,
t Ti Pd Ti Pd Ti Pd
V b h h h h h L h h
m m
m
m
µ µ
µ
µ
= + + + + + +
= + + ⋅ + + •
+⋅ +⋅
=
(C.43)
and
( )
( )
3 3 3
3
14, 813.755 25.236 12,250
2538.519 ,
Pd
Ti
V V V V
t
m m m
m
µ µ µ
µ
= − +
= − +
=
(C.44)
with h
Ti
and h
Pd
being the thickness of the titanium and palladium layer, respectively. As
can be seen the increase in mass, ∆m, will be comprised of Eqs C.38, C.39, and C.44 and
will have five uncertainty contributing variables aside from being in terms of ppm. Thus
we will write ∆m as
( )
, , , ,
Ti Pd
m m b h L h h ∆ = ∆ (C.45)
to show which parameters ∆m depends on. Using the values listed in Table C.3, a
concentration of 1% hydrogen can be found to be equal to
( )
( )( )
( ) ( )
2
3
3
11
_ _ ,
776.25 8.385 2538.519 ,
1.652 10 ,
Pd
m absorption multiplier H density V
kg
m
m
kg
µ
∆ =
 
=

\ .
= ⋅
(C.46)
which is quite large compared to
195
3 3
3 3
3
3
11
2.33 12, 250 4.507 25.236
12.023 2538.519
5.918 10 ,
Pd Pd Ti Ti
m V V V
g g
m m
cm cm
g
m
cm
kg
ρ ρ ρ
µ µ
µ
= + +
= ⋅ + ⋅ +
⋅
= ⋅
(C.47)
which is the mass for the entire cantilever.
The uncertainty analysis of Eq. C.36 will be very similar to that of Section C.1
except Eq. C.36 will contain many more parameters. The mass and equivalent modulus
and density are functions of other variables themselves. They can be written as
( )
, , , , ,
Ti Pd
m b h L h h m = (C.48)
( )
, , , , , , , ,
eq eq
Ti Pd Si Ti Pd
E E b h L h h E E E = (C.49)
and
( )
, , , , , , , .
eq eq
Ti Pd Si Ti Pd
b h L h h ρ ρ ρ ρ ρ = (C.50)
Thus, based on Eqs C.48 to C.50, Eq. C.36 can be written in terms of all the variables that
comprise it as
( )
( )
( )
( )
2
2
35 , , , , , , ,
1
2
33 , , , , , , ,
, , , ,
1
, , , ,
eq Ti Pd Si Ti Pd
eq Ti Pd Si Ti Pd
Ti Pd
Ti Pd
E b h L h h E E E
h
L
b h L h h
f
m b h L h h
m b h L h h
π
ρ ρ ρ ρ
=
∆
+
(C.51)
based on which the frequency shift equation in its phenomenological form is
196
( ) 2 2
, , , , , , , , , , .
Ti Pd Si Ti Pd Si Ti Pd
f f b h L h h E E E ρ ρ ρ = (C.52)
If the values listed in Table C.3 are inserted into Eq. C.51 we get a resonant frequency of
( )
( )
2
2
2
3
2
11
11
2
35
1
2 33
1
35 172.9
1 1
2
350
33 3.995
1.652 10
1
5.918 10
7783.268 Hz
eq
eq
E
h
L
f
m
m
GPa
m
g
m
cm
f
kg
kg
f
π ρ
µ
π
µ
−
−
=
∆
+
 

\ .
=
⋅
+
⋅
=
(C.53)
for this case. Using Eq. C.51, it is now possible to write the overall uncertainty equation
2
2 2 2
2 2 2 2
2
2 2 2
2 2 2
2 2 2
2 2 2
2
,
Ti
Ti
Pd Si Ti
Pd Si Ti
Pd Si Ti
Pd Si Ti
Pd
Pd
df df df df
f b h L h
db dh dL dh
df df df
h E E
dh dE dE
df df df
E
dE d d
df
d
δ δ δ δ δ
δ δ δ
δ δρ δρ
ρ ρ
δρ
ρ
 
     
= + + + +
   
\ . \ . \ .
\ .
¸
     
+ + +
  
\ . \ . \ .
     
+ + +
  
\ . \ . \ .

\
1
2 2
(

(

(
.
¸
(C.54)
where each parameter will individually be taken into account to determine its
contribution to the overall uncertainty. The values used for each of the parameters
appearing in Eq. C.54 are listed in Table C.4.
197
Table C.4. The uncertainty values used based on current industry practice.
Contributions due to uncertainties of each individual parameter to the overall uncertainty
in f, using the values in Table C.4 and the steps shown in Eq. C.5, are
2
5.806
df
b
db
δ
 
 =

\ .
(C.55)
2
488.218
df
h
dh
δ
 
 =

\ .
(C.56)
2
222.33
df
L
dL
δ
 
 =

\ .
(C.57)
2
1.353
Ti
Ti
df
h
dh
δ
 

=

\ .
(C.58)
2
10.248
Pd
Pd
df
h
dh
δ
 

=

\ .
(C.59)
2
46.243
Si
Si
df
E
dE
δ
 

=

\ .
(C.60)
2
0.256
Ti
Ti
df
E
dE
δ
 

=

\ .
(C.61)
δ h =0.05
µ
m δ L =5
µ
m
δ E
Si
=3 GPa δρ
Si
=0.005 g/cm
3
δ h
Ti
=0.0005
µ
m δ b =3
µ
m
δ E
Ti
=3 GPa δρ
Ti
=0.005 g/cm
3
δ h
Pd
=0.0005
µ
m  δ E
Pd
=3 GPa δρ
Pd
=0.005 g/cm
3
Uncertainty in parameters defining f
198
2
23.628
Pd
Pd
df
E
dE
δ
 

=

\ .
(C.62)
2
3.149
Si
Si
df
d
δρ
ρ
 

=

\ .
(C.63)
2
0.006
Ti
Ti
df
d
δρ
ρ
 

=

\ .
(C.64)
2
0.653 .
Pd
Pd
df
d
δρ
ρ
 

=

\ .
(C.65)
Substituting Eqs C.55 to C.65 into Eq. C.54, the overall uncertainty in f is calculated,
following similar steps as in Eq. C.9, as
2
539.106 , f Hz δ = (C.66)
which is the deviation possible in the resonance frequency for a system absorbing
hydrogen, based on the values of the parameters used in the uncertainty analysis. Thus
the overall percentage uncertainty in the frequency, following similar steps as in Eq.
C.10, for this case is
2
% 6.926% , f δ = (C.67)
which is similar to the simplified case in Section C.1.
To discover the contributing factors to this percent uncertainty, each variable
listed in Eq. C.54 is evaluated and the percent contributions are determined. The percent
contribution by uncertainties of each independent parameter, following similar steps as
Eq. C.11, is
199
% 0.012% f b δ δ = (C.68)
% 82.012% f h δ δ = (C.69)
% 17.008% f L δ δ = (C.70)
% 0.001%
Ti
f h δ δ = (C.71)
% 1.05%
Pd
f h δ δ = (C.72)
% 0.736%
Si
f E δ δ = (C.73)
5
% 2.248 10 %
Ti
f E δ δ = ⋅ (C.74)
% 0.192%
Pd
f E δ δ = (C.75)
% 0.003%
Si
f δ δρ = (C.76)
8
% 1.448 10 %
Ti
f δ δρ
−
= ⋅ (C.77)
4
% 1.465 10 % ,
Pd
f δ δρ = ⋅ (C.78)
where it can be seen that the thickness, h, contributes the most with 82% to the overall
percentage uncertainty in the frequency. The total sum of the percentages adds up to
100%, as they should.
C.2.2. Uncertainty in frequency for a free vibrations system with changing stiffness
For a free, undamped vibrating system that has a spring constant that changes the
equation for the frequency shift will be slightly different that that one of Eq. C.36.
Equation 5.12 relates the frequency shift due to hydrogen absorption as
200
2
2
2
1
1
1 35
2 33
1
eq
eq
k
E
h k
f
m L
m
π ρ
∆  
+

 
\ .
 =

∆  
\ .
+

\ .
(C.79)
where ∆k is the change in the damping coefficient. The change in the stiffness will come
from the ability for the modulus of elasticity of palladium to change as it absorbs
hydrogen. Using Fig. 4.6, a function was created to fit along the curve. This function
takes a complex form and for a concentration of 1% becomes
2
2
0.57 5
=1+ cos 0.621 0.05
100 10000 0.621
5 0.57
cos 0.05
100 0.621
0.57 5
=1+ cos 0.621 0.05 0.049
100 10000 0.621
= 0.915
incr
ppm
E
ppm
π
π
π
 
(
−   (
 ∆ ⋅ − −
(

(

\ . ¸ ¸
(
¸ ¸
\ .
  − (
−

(
¸ ¸
\ .
 
(
−   (
 ⋅ − −
(

(

\ . ¸ ¸
(
¸ ¸ \ .
(C.80)
and fits into the equivalent modulus equation, Eq. 4.6. If Eq. C.80 is substituted into Eq.
5.4 and solved for the case of 1%, or 10000 ppm, hydrogen it becomes
( )
( )( )
( ) ( ) ( )
190 116 0.915 121 14, 813.76
116 0.915 121 12, 250 190 0.915 121 25.24 190 116 2538.52
167.3 ,
Si Ti incr Pd t
eq
Ti incr Pd Si Si incr Pd Ti Si Ti Pd
E E E E V
E
E E E V E E E V E E V
GPa
∆
⋅ ⋅ ∆
=
∆ + ∆ +
⋅ ⋅ ⋅
=
⋅ ⋅ + ⋅ ⋅ + ⋅
=
(C.81)
where the influence of the change in the modulus is put into concentration terms, ppm.
To determine the stiffness for this case, Eq. B.43 is modified so that
201
( )( )
( )
( ) ( )
( )
3
3
3
3
3
3
4
2 2 2 2
4
172.9 35 2 0.001 2 0.1 1 2 0.001 2 0.1
4 350 0.001 0.1
0.062 ,
eq t t
t
eq Ti Pd Ti Pd
Ti Pd
E b h
k
L
E b h h h h h
L h h
GPa m m
m
N
m
µ µ
µ
=
+ + + +
=
+ +
( ( + ⋅ + ⋅ + ⋅ + ⋅
¸ ¸ ¸ ¸
=
( + +
¸ ¸
=
(C.82)
where b
t
, h
t
, and L
t
are the total width, thickness, and length, respectively, following Fig.
4.7. To determine the change in the stiffness, ∆k, it will be necessary to include the
changing volume, limited only to the thickness, as a function of concentration
2
2
0.621
1
10000 5
10000 0.621
1
10000 5
1.124 ,
ppm
h
∆
 
= +

\ .
 
= +

\ .
=
(C.83)
where the increase in thickness due to 1% hydrogen is accounted for. It is then possible
to calculate the shifted modulus of elasticity with a modified Eq. C.82 as
( )
( )
( )
( )
( )
( )
3
3
3
3
3
3
4
2 2
2 2
4
167.3 35 2 0.001 2 0.1
4 350 0.001 0.1
1 2 0.001 1.124 2 0.1
0.063 .
eq t t
t
eq Ti Pd
Ti Pd
Ti Pd
E b h
k
L
E b h h
h h h h
L h h
GPa m
m
m
N
m
µ
µ
µ
∆
∆
∆
∆
=
+ +
= ⋅ + +
+ +
( + ⋅ + ⋅
¸ ¸
= •
( + +
¸ ¸
( + ⋅ + ⋅ ⋅
¸ ¸
=
(C.84)
202
To determine the change in stiffness the difference between Eqs C.84 and 82 is calculated
as
0.063 0.062
0.001 .
k k k
N N
m m
N
m
∆
∆ = −
= −
=
(C.85)
If we use the same values listed in Table C.3 and including Eqs C.82 and C.85 we find
that the frequency for Eq. C.79 is equal to
( )
( )
2
2
2
1
2
2
2
11
3
11
1
2
1
1 35
2 33
1
0.001
1
0.062
35 172.9
1 1
2
1.652 10 350
33 3.995
1
5.918 10
7894.523Hz .
eq
eq
k
E
h k
f
m L
m
N
m
N
GPa
m
m
f
g
kg m
cm
kg
f
π ρ
µ
π
µ
−
−
∆  
+

 
\ .
 =

∆  
\ .
+

\ .
 

+

 

 
\ .

=

    ⋅
+
  
\ . ⋅
\ .
\ .
=
(C.86)
The phenomenological relation for Eq. C.86 is the same as in Section C.2.1, i.e.,
( ) 2 2
, , , , , , , , , ,
Ti Pd Si Ti Pd Si Ti Pd
f f b h L h h E E E ρ ρ ρ = (C.87)
where the change in the stiffness does not add any additional variables. Thus, we can use
the same overall uncertainty given by Eq. C.54 and determine the contributions due to
uncertainties of each individual parameter to the overall uncertainty in f, using the values
in Table C.4 and the steps as in Eq. C.5, as
203
2
6.172
df
b
db
δ
 
 =

\ .
(C.88)
2
492.682
df
h
dh
δ
 
 =

\ .
(C.89)
2
229.036
df
L
dL
δ
 
 =

\ .
(C.90)
2
1.556
Ti
Ti
df
h
dh
δ
 

=

\ .
(C.91)
2
9.54
Pd
Pd
df
h
dh
δ
 

=

\ .
(C.92)
2
46.584
Si
Si
df
E
dE
δ
 

=

\ .
(C.93)
2
0.257
Ti
Ti
df
E
dE
δ
 

=

\ .
(C.94)
2
26.002
Pd
Pd
df
E
dE
δ
 

=

\ .
(C.95)
2
3.244
Si
Si
df
d
δρ
ρ
 

=

\ .
(C.96)
2
0.007
Ti
Ti
df
d
δρ
ρ
 

=

\ .
(C.97)
204
2
0.672 .
Pd
Pd
df
d
δρ
ρ
 

=

\ .
(C.98)
Substituting Eqs C.88 to C.98 into Eq. C.54, the overall uncertainty in f is calculated as
2
546.06 , f Hz δ = (C.99)
which is the deviation possible in the resonance frequency for a system absorbing
hydrogen. Thus the overall percentage uncertainty in the frequency for this case is
% 6.917% ,
2
f δ = (C.100)
which is very similar to the result of Eq. C.67.
To discover the contributing factors of the percent uncertainty given by Eq.
C.100, each variable listed in Eq. C.54 is evaluated and the percent contribution is
determined. The percent contribution of each parameter is
% 0.013% f b δ δ = (C.101)
% 81.405% f h δ δ = (C.102)
% 17.593% f L δ δ = (C.103)
4
% 8.116 10 %
Ti
f h δ δ = ⋅ (C.104)
% 0.91%
Pd
f h δ δ = (C.105)
% 0.728%
Si
f E δ δ = (C.106)
5
% 2.223 10 %
Ti
f E δ δ = ⋅ (C.107)
% 0.227%
Pd
f E δ δ = (C.108)
% 0.004%
Si
f δ δρ = (C.109)
205
8
% 1.498 10 %
Ti
f δ δρ = ⋅ (C.110)
4
% 1.515 10 % ,
Pd
f δ δρ = ⋅ (C.111)
where it can be seen that the thickness, h, contributes the most with 81.4% to the overall
percentage uncertainty in the frequency. The total sum of the percentages adds up to
100%, as they should.
C.2.3. Uncertainty in frequency for a free vibrations system with damping
For a free, damped vibrating system the equation for the frequency shift will be
different as the angular frequency is now different, see Eq. B.69. The new equation that
relates the frequency shift due to hydrogen absorption, is
( )
( )
2
2
2
2 2
1
4
1 35
2 33
4
eq
eq
m k m m c
E
h
f
L
k m m
π ρ
  ( + ∆ −
¸ ¸
 =

+ ∆
\ .
(C.112)
where c is the damping coefficient. The damping coefficient, Eq. 4.25, is equal to
( )( )
11
8
2
2 1.77 10 350 35
21.8
1.988 10
Lb
c
H
kg
m m
m s
m
N s
m
µ
µ µ
µ
µ
−
−
=
 
⋅

−
\ .
=
−
= ⋅
(C.113)
which shows that it is a function of the variables µ, L, b, and H and we can write it in its
phenomenological form as
206
( ) , , , c c L b H µ = (C.114)
for simplification. This indicates that there are now two additional variables to the
frequency and thus the phenomenological relation based on Eq. C.112 becomes
( ) 2 2
, , , , , , , , , , , , ,
Ti Pd Si Ti Pd Si Ti Pd
f f b h L h h E E E H ρ ρ ρ µ = (C.115)
where the viscosity, µ, and the damping fluid thickness, H, are listed. If we use the
values listed in Table C.3 and additionally in Table C.5 we find that the frequency for Eq.
C.112 is equal to
( )
( )
( )
( )
( )
( )
( )
2
2
2
2 2
1
2
2
3
1
11
11 11
2
11 8
1 35
2 33
4
35 172.9
1 1
2
350
33 3.995
5.918 10
5.918 10 1.652 10
5.918 10 1.988 10
4 0.06
eq
eq
E
h m mc
f
m m L
k m m
GPa
m
g
m
cm
kg
kg kg
N s
kg
m
π ρ
µ
π
µ
−
− −
− −
   
  = −
 
+ ∆
+ ∆
\ . \ .
 


= •

 
 
\ .
\ .
¸
¦
⋅
¦
−
´
⋅ + ⋅
¦
¹
−  
⋅ ⋅

\ .
( )
1
2
2
11 11
2 5.918 10 1.652 10
7783.313Hz .
N
kg kg
m
− −
( ¹
(
¦
¦
(
`
(
 
¦
⋅ + ⋅
(

¦
\ .
(
)
¸
=
(C.116)
Because of the two additional variables, the overall uncertainty equation is going to
change to take them into account and thus the equation takes the form of
207
2
2
2 2
2 2 2 2
2
2 2 2
2 2 2
2 2 2
2 2 2
2
,
Ti
Ti
Pd Si Ti
Pd Si Ti
Pd Si Ti
Pd Si Ti
Pd
Pd
df
df df df
f b h L h
db dh dL dh
df df df
h E E
dh dE dE
df df df
E
dE d d
df
d
δ δ δ δ δ
δ δ δ
δ δρ δρ
ρ ρ
δρ
ρ
 
 
   
 = + + + +
  

\ . \ .
\ .
\ .
¸
     
+ + +
  
\ . \ . \ .
     
+ + +
  
\ . \ . \ .
1
2 2
2 2
2 2
df df
H
d dH
δµ δ
µ
(
   
 
+ +
(
  
\ .
\ . ( \ .
¸
(C.117)
where each variable will individually be taken into account as before with the viscosity
and damping fluid thickness included.
Table C.5. Values of the two additional independent
parameters in the damped case.
Contributions due to uncertainties of each individual parameter to the overall
uncertainty in f, using the values in Table C.4 and the steps as in Eq. C.5, are
2
5.806
df
b
db
δ
 
 =

\ .
(C.118)
2
488.225
df
h
dh
δ
 
 =

\ .
(C.119)
2
222.333
df
L
dL
δ
 
 =

\ .
(C.120)
Viscosity of nitrogen, µ kg/
µ
ms 1.769×10
11
δ m =5×10
14
Damping fluid thickness, H
µ
m 21.8 δ H=0.05
Parameter values
208
2
1.353
Ti
Ti
df
h
dh
δ
 

=

\ .
(C.121)
2
10.247
Pd
Pd
df
h
dh
δ
 

=

\ .
(C.122)
2
46.244
Si
Si
df
E
dE
δ
 

=

\ .
(C.123)
2
0.256
Ti
Ti
df
E
dE
δ
 

=

\ .
(C.124)
2
23.628
Pd
Pd
df
E
dE
δ
 

=

\ .
(C.125)
2
3.149
Si
Si
df
d
δρ
ρ
 

=

\ .
(C.126)
3
2
6.487 10
Ti
Ti
df
d
δρ
ρ
 

= ⋅

\ .
(C.127)
2
0.652
Pd
Pd
df
d
δρ
ρ
 

=

\ .
(C.128)
4
2
4.664 10
df
d
δµ
µ
 
 = ⋅

\ .
(C.129)
4
2
3.785 10 .
df
H
dH
δ
 
 = ⋅

\ .
(C.130)
209
Substituting Eqs C.118 to C.130 into Eq. C.117, the overall uncertainty in f is calculated,
following similar steps as in Eq. C.9, as
2
539.113 , f Hz δ = (C.131)
which is the deviation possible in the resonance frequency for a system absorbing
hydrogen. Thus the overall percentage uncertainty in the frequency for this case is
2
% 6.927% , f δ = (C.132)
which is very similar to the result of Eq. C.100.
To discover the contributing factors of this percent uncertainty, each variable
listed in Eq. C.54 is evaluated and the percent contribution is determined. The percent
contribution of each parameter is
% 0.012% f b δ δ = (C.133)
% 82.012% f h δ δ = (C.134)
% 17.008% f L δ δ = (C.135)
% 0.001%
Ti
f h δ δ = (C.136)
% 1.05%
Pd
f h δ δ = (C.137)
% 0.736%
Si
f E δ δ = (C.138)
5
% 2.248 10 %
Ti
f E δ δ = ⋅ (C.139)
% 0.192%
Pd
f E δ δ = (C.140)
% 0.003%
Si
f δ δρ = (C.141)
8
% 1.448 10 %
Ti
f δ δρ = ⋅ (C.142)
210
4
% 1.465 10 %
Pd
f δ δρ = ⋅ (C.143)
11
% 7.484 10 % f δ δµ = ⋅ (C.144)
11
% 4.93 10 % , f H δ δ = ⋅ (C.145)
where it can be seen that the thickness, h, contributes the most with 82% to the overall
percentage uncertainty in the frequency. The total sum of the percentages adds up to
100%, as they should.
C.2.4. Uncertainty in frequency for a damped, changing stiffness vibrations system
For a free, damped vibrating system that has a spring constant that changes the
equation for the frequency shift will be slightly different that that one of Eq. C.112. The
final modified equation that relates the frequency shift due to hydrogen absorption, is
( )
( )( )
2
2
2 2 2
1
4
1 35
2 33
4
eq
eq
m k m m c E
h
f
L
k k m m
π ρ
 
+ ∆ − (
¸ ¸
= 

+ ∆ + ∆
\ .
(C.146)
where ∆k, ∆m, and c are all included. If we use the same values listed in Tables C.3 and
C.5 we find that the frequency for Eq. C.146 is equal to
211
( )
( )( )
( )
( )
( ) ( )
2
2
2
2 2
1
2
2
3
1
11 11 11
4
1 35
2 33
4
35 172.9
1 1
2
350
33 3.995
5.918 10 4 0.062 5.918 10 1.652 10
4 0.062 0.001 5.9
eq
eq
m k m m c
E
h
f
L
k k m m
GPa
m
g
m
cm
N
kg kg kg
m
N N
m m
π ρ
µ
π
µ
− − −
  ( + ∆ −
¸ ¸
 =

+ ∆ + ∆
\ .
 


= •

 
 
\ .
\ .
¸
(
 
⋅ ⋅ + ⋅
 (
\ . ¸ ¸
 
+

\ .
( )
( )
( )
2
11 11
1
2 2
11 8
2
11 11
18 10 1.652 10
5.918 10 1.988 10
4 0.062 0.001 5.918 10 1.652 10
7673.619 Hz .
kg kg
N s
kg
m
N N
kg kg
m m
− −
− −
− −

−
⋅ + ⋅
\
( 
−  
⋅ ⋅ ( 

\ .
( 
( 
 
+ ⋅ + ⋅
( 


\ .
(
.¸
=
(C.147)
The phenomenological form for Eq. C.147 is the same as the damped case, expressed by
Eq. C.112, so the phenomenological relationship takes the form of
( ) 2 2
, , , , , , , , , , , , ,
Ti Pd Si Ti Pd Si Ti Pd
f f b h L h h E E E H ρ ρ ρ µ = (C.148)
where all the variables are included. Thus, we can use the same overall uncertainty as
Eq. C.117 and determine the contributions due to uncertainties of each individual
parameter to the overall uncertainty in f, using the values in Table C.4 and the steps as in
Eq. C.5, are
2
5.457
df
b
db
δ
 
 =

\ .
(C.149)
212
2
483.599
df
h
dh
δ
 
 =

\ .
(C.150)
2
215.823
df
L
dL
δ
 
 =

\ .
(C.151)
2
1.162
Ti
Ti
df
h
dh
δ
 

=

\ .
(C.152)
2
10.905
Pd
Pd
df
h
dh
δ
 

=

\ .
(C.153)
2
45.883
Si
Si
df
E
dE
δ
 

=

\ .
(C.154)
2
0.254
Ti
Ti
df
E
dE
δ
 

=

\ .
(C.155)
2
21.372
Pd
Pd
df
E
dE
δ
 

=

\ .
(C.156)
2
3.056
Si
Si
df
d
δρ
ρ
 

=

\ .
(C.157)
2
0.006
Ti
Ti
df
d
δρ
ρ
 

=

\ .
(C.158)
2
0.633
Pd
Pd
df
d
δρ
ρ
 

=

\ .
(C.159)
213
4
2
4.527 10
df
d
δµ
µ
 
 = ⋅

\ .
(C.160)
4
2
3.674 10 .
df
H
dH
δ
 
 = ⋅

\ .
(C.161)
Substituting Eqs C.149 to C.161 into Eq. C.117, the overall uncertainty in f is calculated,
following similar steps as in Eq. C.9, as
2
532.137 , f Hz δ = (C.162)
which is the deviation possible in the resonance frequency for a system absorbing
hydrogen. Thus the overall percentage uncertainty in the frequency for this case is
2
% 6.696% , f δ = (C.163)
which is very similar to the result of Eq. C.100.
To discover the contributing factors of this percent uncertainty, each variable
listed in Eq. C.54 is evaluated and the percent contribution is determined. The percent
contribution of each parameter is
% 0.011% f b δ δ = (C.164)
% 82.589% f h δ δ = (C.165)
% 16.449% f L δ δ = (C.166)
4
% 4.764 10 %
Ti
f h δ δ = ⋅ (C.167)
% 1.189%
Pd
f h δ δ = (C.168)
% 0.743%
Si
f E δ δ = (C.169)
5
% 2.271 10 %
Ti
f E δ δ = ⋅ (C.170)
214
% 0.161%
Pd
f E δ δ = (C.171)
% 0.003%
Si
f δ δρ = (C.172)
8
% 1.4 10 %
Ti
f δ δρ = ⋅ (C.173)
4
% 1.417 10 %
Pd
f δ δρ = ⋅ (C.174)
11
% 7.238 10 % f δ δµ = ⋅ (C.175)
11
% 4.768 10 % , f H δ δ = ⋅ (C.176)
where it can be seen that the thickness, h, contributes the most with 82.3% to the overall
percentage uncertainty in the frequency. The total sum of the percentages adds up to
100%, as they should.
215
APPENDIX D. TEST CHAMBER
In this Appendix the blueprints for the test chamber are shown along with some
pictures of the finished component.
Fig. D.1. Specifications of the main body of the test chamber.
216
Fig. D.2. Specifications of the lid for the main body of the test chamber.
Fig. D.3. Fully disassembled view of the chamber showing all parts.
217
Fig. D.4. Fully assembled view of chamber with PZT inside.
Fig. D.5. Isometric view of the assembled chamber.
218
APPENDIX E. ANALYTICAL FREQUENCY DATA
Using measured dimensions of the cantilevers, it was possible to calculate the
theoretical resonant frequency for each cantilever based on Eq. 4.23. Figure E.1
illustrates the guide for the measurements taken and labels used while Table E.1 has the
values for the modulus of elasticity and the density used. The parameters L1, L3, b1, and
b2 are used to determine the area, A=L2 x b, represent by the shaded part in Fig. E.1.
Table E.1. Material properties of silicon used for the analytical calculations.
Fig. E.1. Schematic of the cantilever and the dimensions
used for the data analysis.
190
2.33 Density, ρ , g/cm
3
Modulus of elasticity, E, GPa
Parameter values used
219
Taking cantilever D12 as an example, the resonance frequency is
( )
( )
1
2
1
2
3
1
1 35
2 33
190
1 1.007 35
2 33
304.106
2.33
16,116.660 ,
h E
f
L
GPa
m
f
g
m
cm
f Hz
π ρ
µ
π
µ
=
=
 

\ .
=
(E.177)
where the values for each parameter and result for each cantilever are listed in Table. E.2.
Table E.2. Geometrical and resonant frequency data of the cantilever utilized.
Measured D12 E12 F12
characteristics Typical Typical Typical
Length, L mm
295 345.5 244
Width, b mm 36.5 36 36
Thickness, h mm 1.007 1.007 1.007
End width, L1 mm 15 18 15
End length, b1 mm 11 11 11
Offset, b2 mm
3 3.5 4
Point width, L3 mm 3 4.5 4
Added area, A mm
2
332.375 355.75 345.5
Eqv. length, L2 mm 9.106 9.882 9.597
Resonance freq., f Hz
16116.660 11801.434 23175.895
Measured D11 E11 F11
characteristics Typical Typical Typical
Length, L mm
296.5 345 246.5
Width, b mm 36 35.5 36.5
Thickness, h mm 1 1 1
End width, L1 mm 14 13 14.5
End length, b1 mm 11 11.5 10.5
Offset, b2 mm
2 2.5 1.5
Point width, L3 mm 4.5 2 4.5
Added area, A mm
2
321.75 318.875 308.25
Eqv. length, L2 mm 8.938 8.982 8.445
Resonance freq., f Hz
15865.410 11812.252 22772.060
220
Measured D21 E21 F21
characteristics Typical Typical Typical
Length, L mm 293.5 342.5 239.5
Width, b mm 36.5 34 34
Thickness, h mm 1 1 1
End width, L1 mm 15 14.5 14.5
End length, b1 mm 10 9.5 10
Offset, b2 mm
2 2.5 2
Point width, L3 mm 3.5 4 4
Added area, A mm
2
296.3125 273.625 282
Eqv. length, L2 mm 8.118 8.048 8.294
Resonance freq., f Hz
16269.757 12044.854 24105.383
Measured D22 E22 F22
characteristics Typical Typical Typical
Length, L mm 296 346 246
Width, b mm 35 34.5 33
Thickness, h mm 0.96 0.96 0.96
End width, L1 mm 15 15 15
End length, b1 mm 9 9 8
Offset, b2 mm
3 2 3
Point width, L3 mm 3.5 3 2
Added area, A mm
2
269.5 255.375 225
Eqv. length, L2 mm 7.700 7.402 6.818
Resonance freq., f Hz
15405.567 11377.028 22230.573
Measured D31 E31 F31
characteristics Typical Typical Typical
Length, L mm 295 346 246
Width, b mm
35 35 35
Thickness, h mm 0.989 0.989 0.989
End width, L1 mm 15 13 13
End length, b1 mm 9.5 9 8
Offset, b2 mm 2 2 3
Point width, L3 mm 3 3.5 3.5
Added area, A mm
2
271.5 253.25 235.25
Eqv. length, L2 mm 7.757 7.236 6.721
Resonance freq., f Hz
15969.949 11731.758 22919.660
221
Measured D32 E32 F32
characteristics Typical Typical Typical
Length, L mm 295 345 248
Width, b mm
36 35 37
Thickness, h mm 0.956 0.956 0.956
End width, L1 mm 15 15 15
End length, b1 mm 10 9 10
Offset, b2 mm 3 2.5 2
Point width, L3 mm 4.5 4 3.5
Added area, A mm
2
308.625 267.5 299.25
Eqv. length, L2 mm 8.573 7.643 8.088
Resonance freq., f Hz
15354.224 11378.466 21576.252
Copyright © 2003 by NEST – NanoEngineering, Science, and Technology CHSLT – Center for Holographic Studies and Laser micromechaTronics Mechanical Engineering Department Worcester Polytechnic Institute Worcester, MA 016092280 All rights reserved
3
SUMMARY This Thesis relates to the continually advancing field of microelectromechanical systems (MEMS). With MEMS technology, there are many different areas of concentration available for research. This Thesis addresses analysis and preliminary characterization of a cantilevertype MEMS chemical sensor for detection of chemicals and organic components operating at room temperature (20˚C and sea level pressure of 1 atm). Such sensors can be useful in a wide variety of applications. There currently exist several different types of MEMS chemical sensors. Each is based on a different detection method, e.g., capacitive, thermal, resistive, etc., and is used for specific tasks. Out of all currently available detection methods, the most common is the gravimetric method. The gravimetric sensor works by absorbing the chemical in a special material, usually a polymer, which alters the overall mass of the sensing element that can then be measured, or detected, to identify the chemical absorbed. One of the more exciting developments in the field of gravimetric chemical MEMS has been with the advancement of cantilevertype sensors. These cantilevers are small and usually on the order of only about 300 µm in length. In order to utilize the gravimetric method, a cantilever is coated with a polymer that allows an analyte to bond to it and change its mass, which in turn changes the resonant frequency of the cantilever. The change in frequency can then be measured and analyzed and from it, the amount of absorbed mass can be calculated. Current research in the cantilevertype resonating sensors for the detection of hydrogen is developing measurement capabilities of 1 ppm (partpermillion). 4
In this Thesis number of sample cantilevers were qualitatively assessed and their dimensional geometry measured. Based on these measurements, frequency data were obtained. In addition, the overall uncertainty in the resonant frequency results was calculated and the contributing factors to this uncertainty were investigated. Experimental methods that include laser vibrometry, optoelectronic laser interferometric microscopy (OELIM), and atomic force microscopy (AFM), were utilized to measure the frequency responses of the samples. The analytically predicted natural frequencies were compared to the experimental data to determine correlation subject to the uncertainty analysis. Parametric analyses involving chemical absorption processes were also conducted. Such analyses considered different parameters, e.g., damping and stiffness as well as changes in their values, to determine contributions they make to the quality of the frequency data and the effect they have on sensitivity of the MEMS cantilevertype chemical sensors. Once these parametric analyses were completed, it was possible to estimate the sensitivity of the cantilever, or the ability for the cantilever to detect frequency shifts due to absorption of the target chemical. Results of the parametric analyses of the fundamental resonant frequency were then correlated with the sensitivity results based on the chemical absorption. This Thesis correlates many results and ideas and probes problems revolving around the analysis and characterization of a MEMS cantilevertype chemical sensor.
5
Thank you. Without their undying love and support. I would also like to thank Professor Cosme Furlong for his support in my computational work and all the random questions I had. Andres. I would not be where I am today. and Vera Arecco Ph.ACKNOWLEDGEMENTS I would first like to thank my parents. May they never catch you. Finally. Thesis will stay with me for the rest of my life and I am glad to have accomplished it with all the great people of Worcester Polytechnic Institute. Janice Dresser. First off. He has taught me things beyond the call of a typical advisor. Ryan Marinis. I would like to thank my brother. Pryputniewicz for the wonderful opportunity he has given me to work in a research field that has challenged me to new limits. I have to thank Professor Ryszard J. I have to send a special thank you to Mr. who lent me his wealth of technical knowledge and support for completing my experimental work and also for his stories that would always boost morale. May your teaching career be rewarding. I would also like to thank everyone else at the Center for Holographic Studies and Laser micromechaTroncis (CHSLT): Kevin Bruff. I would also like to acknowledge all the people that helped me along the way in there own ways: Barbara Edilberti. This M.D. Adam Klempner.D. for everything that they have provided for me in my life. Ronald Kok. Another special thank must go to Mr. Carlos Arecco M. because I can. Secondly. Peter Hefti from Switzerland. Wei Han.S. 6 . and Vicky Steward.) Shivananda Pai Mizar for all the invaluable support that he showed whenever I was in need and for his affirming to my getting done. and Todd Billings. (soon to be Dr.
14.3. Sensing types and methods 2. Cantilever actuation methods 2.4.2.4.3. Objectives 2.1.3.TABLE OF CONTENTS Copyright Summary Acknowledgements Table of contents List of figures List of tables Nomenclature 1.4.13.11. Mechanical detection 2. Quality factor 2. Electrical aspects 2.3. Optical detection 2.13.3. MEMS cantilever chemical sensor 2.12.4. Inherent Problems 2.4.4.4. Stiction 7 3 4 6 7 11 16 18 21 22 22 24 27 29 31 32 32 35 35 37 39 40 42 43 44 46 47 47 48 49 50 50 51 51 53 .4. Fick’s law of diffusion 2.4.2.2.4.6. Reference cantilevers 2.4.5.8. Sensors 2.2.1.4.4. Polymer thickness 2.4. Chemical detection 2.4.4.14. Micromachine beginnings 2.3. Resonance frequency 2.7.3.14. Feed back Loop 2.13.10. Cantilever I/O 2. Material 2. MEMS Foreword 2. Packaging 2.4. Wheatstone bridge 2. Introduction 2.2.4. Residual Stresses 2.4.1.4.1. Electrical detection 2.1. Fabrication 2.
2. Analytical investigations 4.2.3. Methodology 4.3.15.5.1. Uncertainty analysis 4.4.7. Time limits 4. Instrument calibration 4.6.5.5.3.4.2.6. Data analysis 54 54 56 56 58 59 60 60 61 62 63 63 65 65 67 68 71 71 72 75 78 79 81 83 85 87 89 89 92 93 94 95 95 96 97 98 104 105 107 109 8 . Effects of pressure 3. Laser vibrometry 4.2.3.11. SEM – scanning election microscope 4.2.10.4. Palladium 3. Hydrogen 3.2.4.2. Absorption 4.2.4. Gas concentration 4. Test chamber 4.2.1. Computational analysis 4.3. MikroMasch USA cantilevers 3.2.2.4.9.4. Experimental solutions 4.3.2.4. Geometrical characterization of the cantilevers 4. Coating process 3.4.16.1. Changing stiffness in a free vibration system 4.3.4.1. Free vibration system 4. Volume increase 4.2.2.8.4.7.8. Titanium 3.4.3. Effects of humidity 3.2. Equivalent variables 4. Chemicals used 3. Damping coefficient consideration 4. Optoelectronic methodology 4. Harmonically excited system 4.2. MEMS samples 3.2.2.2. Resonant frequency and frequency shift 4.2. Determination of quality factor and damping coefficient 4.2.1.2. Determination of resonant frequency by AFM 4.2.4.2.3.4. Effects of temperature 2.1.1. Spring constant consideration 4.2. ACES methodology 4.3.4. Sensitivity 4.4. Change in modulus of elasticity 4. Diffusion rate 4. Complete versus partial coating 3.
Material properties Appendix B.8.1. Case 3: changing mass with damping 5.2. Results 5. Design equation selection 5.8. Lower limit frequency shift 5. undamped vibration with a changing stiffness B.6.11.1.11.3.11.3.5. Optoelectronic measurements 5.1. Determination of the damping coefficient 5.10. References Appendix A. Microscope and SEM data 5.5.11. Dynamic mass B. Sensitivity calculations of the cantilever B. AFM results 5.6. Absorption 5.2.5. Frequency shift for a free. damped vibration with a changing stiffness A.7.5. Sensitivity solutions 5.3. damped vibration A. Case 1: changing mass 5. Case 2: changing mass and stiffness 5.11.4.1.7. Sensitivity of a completely coated cantilever B.4. Dynamic mass constant B. Spring constant B.2.7. Analytical results 5. Free vibration with viscous damping A.11.3. Computational results 5.11.7.2. Case 4: changing mass and stiffness with damping 5.7.1. Fundamental resonance frequency B.6. Frequency shift for a free.2.9. Parameters of the coated cantilevers 5. Conclusions 7. Frequency shift for a free.4.1. Optimization of the cantilever response 6.4.11. Comparison of results 5. Analytical calculations B. Sensitivity of an endloaded cantilever B. Complete cantilever analysis 5.4. undamped vibration B. Frequency shift for a free. Harmonically excited system with viscous damping 111 111 116 121 123 124 127 129 130 133 136 139 140 141 142 142 143 144 146 148 151 161 164 164 165 168 170 171 171 173 174 176 177 178 180 181 9 .
1.2.2.2.1.1. Test chamber Appendix E. Uncertainty calculations for the 250 µm long cantilever C.Appendix C.1.1. Uncertainty analysis C. changing stiffness vibrations system Appendix D. Analytical frequency data 183 183 183 187 189 190 191 191 199 205 210 215 218 10 .4.1. Uncertainty in frequency for a damped. Uncertainty calculations for the 350 µm long cantilever C. Uncertainty in frequency for a free vibrations system with damping C.2. Frequency uncertainty in a palladium coated cantilever C.4.1.2.3. Uncertainty calculations for the 300 µm long cantilever C. Uncertainty in frequency for a free vibrations system with a changing stiffness C.3. Uncertainty in frequency for a free vibrations system C. Average overall percentage uncertainty in frequency C. Uncertainty in fundamental resonant frequency C.2.2.
15. Effects of residual stresses (Dartmouth. 2001) Details of a DMD chip (TI. MEMS chip and its different levels of packaging (SNL. Examples of SUMMiT™V process (SNL. 2.11.. 2. Cantilever with integrated piezoresistors (Gotszalk et al. 2.. Operation of different gravimetric sensors (Ivanov.9. Fig. 2. 2003) Polymer based FabryPerot sensor (Yoo et al. 1997) Micron sized cantilevers developed at IBM for research purposes (IBM. Comparison of different sized and shaped PZTs available (PI. 2. alter the output voltage with a high accuracy and precision Fig. 2002) Initial steps in fabrication of MEMS and CMOS chips (MEMC.3. in R1.2. 2. 2. Missouri. e. 2002) A view of an entire dualaxis accelerometer where the proof mass can be seen with all the perforations in the middle and a quarterview close up on one of its corners (AD.7. 2. 2. Fig.13. Fig.6. Fig.4. 2001) Fig. 2. A dust mite dwarfs a MEMS device (SNL.8. 2000) Fig. 2003) A schematic illustration of MEMS components and their interdependence (Madou.5. 2001) A human is compared to a MEMS device (SNL. Fig. Fig. Fig. 2003) Fig.g. 2. 2002) Fig. 2. 2001) A microchain device (SNL.1. 2. 2.. 2.10. 2000) Fig. 2002) 11 45 46 50 52 . 2. Graphical illustration of the operational principle for an AFM where different bend angles reflect the laser beam to different locations on the PSD Fig.12.16. 2. Wheatstone bridge where changes. Fig.17. 2003) 23 23 25 26 28 30 33 34 36 38 40 42 44 Fig.14.LIST OF FIGURES Fig.
4.10. 4.6. 79 Fig. 4. where the concentration. 96 98 Fig. The samples are inserted into the chamber on the left (botton of the black tube) and the microscope is controlled with the console on the right 53 55 57 58 62 66 69 74 76 Fig. 3. 2000) Fig.5. 4.3. 4. 4. 1967) Experimental data fitted with a line to approximate the amount of volumetric change in palladium due to hydrogen absorption (G. 4. A femtocalorimeter utilizing thermal bending effects with piezoresistors (Gotszalk et al. 4. 2.3. which in turn will affect Eq.1. 1967) Permeation rates of hydrogen with circulating or static flows through a membrane of Pd with a high pressure on one side (Lewis.1. Fig. The collapse of a thin cantilever due to stiction Fig.18.9. 1978) Estimated results from experimental data illustrating the change that occurs in the modulus of elasticity during absorption of hydrogen into palladium (Lewis.. Völkl. 82 91 Fig.42 12 . 2. Fig. 2003) Photographs of the AFM chips used in this Thesis Bending moment comparison Graphical representation of ACES methodology and the interdependence of each system towards determining a final solution Idealization and modeling of a cantilever Concentration and pressure relationships based on experimental data of many research studies conducted throughout the years.Fig. Fig.19. Fig.7. AR. 4. Fig. 80 Fig. A comparison of how different resonant frequency response curves affect the width of the bandwidth.4. Fig. Alefeld and J. 4. is in units of atomic ratio (H/Pd) (Lewis. 3.2.2.8. Fig. 1967) Representation of cantilever with geometric dimensions and a crosssection view through the middle with each of the layers of the cantilever labeled Vibration amplitude as a function of frequency ratio for different values of damping ratio SEM used for measuring thickness of the cantilevers. 4. 3. MikroMasch AFM chip (MM.
4.6.8.Fig.4. 5. Laboratory setup for testing functional operation of MEMS chemical sensors. 5. 119 13 . 5. Fig. the microscope and test chamber will be contained within a vented hood Fig. while holding the length parameter to 300 µm to show the decreasing trend 100 104 105 106 108 108 112 113 114 115 115 118 119 Fig. 4. 2000) Fig.15.14. Singleillumination and singleobservation geometry of a fiber optic based OEH system (Furlong and Pryputniewicz. 4.7. 4. AFM setup utilized Fig. 5. while holding the thicknes parameter to 1 µm to show the decreasing trend The percentage overall uncertainty in the cantilever is plotted against the thickness. The microscope setup used in determining the dimensions of the cantilevers Cantilever details and labelling Cantilevers packed in a GelPack as obtained from the manufacturer SEM photographs of the cantilevers. Fig. A comparison of the PZT used with a MEMS chip attached to the top and a dime Fig. 5.11. 5. Fig. Optical configuration of the OELIM system (Furlong and Pryputniewicz.13. 4. 4.1. 5. Fig.16. Fig. with a progressive close up on the tip SEM photographs of E length cantilevers near the tip from different chips Fundamental resonant frequency of a 1µm thick cantilever as a function of active length The percentage overall uncertainty in the cantilever is plotted against the length.3.12. The custom made test chamber containing the PZT transducer.5. which is to be used for testing functional operation of the cantileversensing element in a gaseous environment.2. Fig. 2000) Fig. 5. Fig.
A digitally enhanced image of a screen capture from the HP Spectrum analyzer taken while measuring the frequencies of a cantilever Fig. 2003) Fig.160 Hz with the accuracy of 10 Hz Fig. 5. A. B.13. and F) of chip 22 vibrating at their fundamental mode of vibration Fig.1 µm thick layer of palladium Fig. 5. 5.19.11. 5.1. The frequency shift in the resonance of a cantilever as a function of concentration of the hydrogen Fig. 2003) Fig. 2003) Fig. as a function of length 120 Fig. 5. OELIM images of the cantilevers (D. uncut silicon (EC. where a resonant frequency peak is identified as 22.1.1. Pro/Engineer model and Pro/Mechanica results of the three cantilevers considered Fig.Fig. A.18. 5.17.16. Percent contributions by uncertainty by each independent parameter to the overall uncertainty in resonance frequency of the cantilever.9.12. 5. Overall uncertainty in the frequency of the cantilever as a function of the uncertainty in the thickness Fig. unprocessed.2. A. Percent contributions by uncertainty of each independent parameter to the overall uncertainty in resonance frequency of the cantilever. 5. Concentration as a function of time for a cantilever coated with 0.14.3. D. E. 5. a popular material due to its corrosion resistance (EC. 5.10. A sample of titanium. A photograph of palladium samples (EC. as a function of thickness Fig. Specifications of the main body of the test chamber 120 121 122 122 123 126 131 136 138 146 162 163 163 166 215 14 . OELIM images of cantilever E32 vibrating under resonant conditions Fig. A sample of untreated. 5. Geometric parameters of a cantilever Fig. A screen capture of the AFM frequency scan for the cantilever F12. Effective mass versus percentage of coating coverage Fig.15.
5.1. D. Fully disassembled view of the chamber showing all parts Fig. Specifications of the lid for the main body of the test chamber Fig. D. E.3.Fig. D.2. Isometric view of the assembled chamber Fig. D.4. Schematic of the cantilever and the dimensions used for the data analysis 216 216 217 217 218 15 . Fully assembled view of chamber with PZT inside Fig.
MikroMasch USA. 1967) Table 5. Values of material properties of silicon (Madou. and AFM generated frequencies for each cantilever Table 5. E. Summary of analytically calculated and experimentally determined damping coefficients Table 5. Values of parameters used in uncertainty analysis Table C. Comparison of the analytical and computational resonance frequency results Table 5.2.1.1. 2002) Table 5.6.1. Properties of materials used Table 5.1.7. and F types Table 5. Concentrations of hydrogen in Pd based on experimental data of Fig. 4. Data comparison of analytical. Characteristic dimensions.5. Measured dimensional data and equivalent lengths for all cantilever samples of the D type Table 5. Summary of measured dimensions for the cantilevers of the D.10. The results of the four different cases investigated are compared at 1% hydrogen concentration Table A. Material properties and geometric dimensions used in a sample case to determine sensitivity of the cantilever as a chemical sensor Table C. as supplied by the manufacturer.3.2.1.LIST OF TABLES Table 3. and the corresponding material properties as listed by Madou (2002) Table 4.9. Physical properties of the materials used Table B.8. The uncertainties in each parameter considered Table 5. Frequency shift per Hertzkilogram for distributed and tip loads Table 5. OELIM.3 at 20ºC with conversion information (Lewis.1. The uncertainties in each parameter considered 16 57 75 115 116 117 117 125 128 130 133 139 144 162 172 184 185 .4.
Material properties of silicon used for the analytical calculations Table E.2. The uncertainty values used based on current industry practice Table C. Geometrical and resonant frequency data of the cantilever utilized 191 197 207 218 219 17 .Table C.5.1. Material properties and thickness of the composite cantilever Table C.3. Values of the two additional independent parameters in the damped case Table E.4.
mass effective mass constant effective mass constant for a cantilever coverage at value r effective mass constant in terms or x pressure partspermillion frequency ratio.4 width of the cantilever width of the end of cantilever width of point of cantilever total width of cantilever concentration of gas. percent coverage of cantilever with polymer Sievert’s constant.NOMENCLATURE b b1 b2 bt c cc co cgas erf f fi f1 f2 f h ht hTi hPd h∆ k m n nr n(x) p ppm r s s1. solution to differential equation solution to differential equation with two roots time velocity maximum velocity of a body distributed load position along the cantilever displacement displacement in terms of x maximum displacement position across the cantilever amplitude area of shaped tip of cantilevers constants 18 . or dynamic.3. damping coefficient critical damping coefficient initial concentration of gas concentration of gas at the surface Gaussian error function resonant frequency initial resonant frequency initial resonant frequency final resonant frequency frequency thickness of the cantilever total thickness of the cantilever thickness of the titanium layer thickness of the palladium layer change of thickness constant spring constant effective.2 t v vmax wo x y y(x) ymax z A A C1.2.
t) FEM H I K K1 K2 KE L L Lt L1 L2 L3 M Mx MEMS PZT Se Sd Sm T VPd VSi VTi Vt X X(x) ζ θ µ ν ξ ρ ρeq ρTi diffusion rate modulus of elasticity equivalent modulus of elasticity modulus of elasticity of titanium modulus of elasticity of palladium modulus of elasticity of silicon maximum force acting on a body force time dependent forcing function finite element method thickness of damping fluid moment of inertia sensitivity vector direction of illumination vector direction of observation vector kinetic energy of a system displacements vector length of the cantilever total length of the cantilever length of shaped end of cantilever equivalent length of the shaped end of cantilever length of side point of cantilever mass bending moment along the cantilever microelectromechanical systems piezoresistive transducer sensitivity for an end loaded cantilever sensitivity for a distributed load cantilever mass sensitivity temperature volume of palladium coating volume of silicon cantilever volume of titanium coating total volume of cantilever with coatings amplitude general solution to a differential equation damping ratio phase angle viscosity poisson’s ratio percentage area coverage by the polymer layer density equivalent density density of titanium 19 .D E Eeq ETi EPd ESi Fxmax F F(y.
ρPd ρSi ω ωn ωd Ω (˜) ( )T δ( ) δm δk δf df dm ∆f ∆m ∆k density of palladium density of silicon angular frequency natural angular frequency damped angular frequency fringelocus function matrix transpose uncertainty uncertainty in mass uncertainty in stiffness uncertainty in frequency infinitesimal change in frequency infinitesimal change in mass finite change in frequency finite change in mass finite change in stiffness 20 .
1. OBJECTIVES The objectives of this thesis were to design, analyze, and perform preliminary characterization of MEMS cantilevertype chemicals sensors for operation at 20˚C and 1 atm pressure, using ACES methodology.
21
2. INTRODUCTION To understand the methodology and processes used in this Thesis it is important to have background information on the MEMS cantilevertype chemical sensor. Many different aspects will be covered, some just for completeness, in this chapter. Each section will cover some facet of the multidisciplinary technology of these microsensors, which can be considered to be a part of a broad category of micromachines.
2.1. Micromachine beginnings Ignoring science fiction roots, the impetus towards micromachines was inspired by Prof. Richard P. Feynman of the California Institute of Technology in Pasadena, California in 1959 (Feynman, 1992). In the late 1950’s, Prof. Feynman initiated a new area of research that is currently expanding at unprecedented rates. Technology has gone from making macrosized machines to millimeter sized to more recently, within less than two decades, micrometer sized devices as is shown in Fig. 2.1 (Pryputniewicz, 2002a). Figure 2.2 compares a micro gear with a human hair for a visual understanding of their relative scale. While nanotechnology is still largely not of everyday application at the nanometer size, the size of the new microdevices continues to decrease (Pryputniewicz, 2002b).
22
Fig. 2.1. A dust mite dwarfs a MEMS device (SNL, 2001).
Fig. 2.2. A human hair is compared to a MEMS device (SNL, 2001). The most researched area of micromachines is with microelectromechanical systems (MEMS). These systems are a combination of mechanical and electrical components built into incredibly small devices that are fabricated using sophisticated integrated circuit (IC) batch processing technologies (Pryputniewicz, 2002a). MEMS began in the mid 1980’s and some of the first products were accelerometers. MEMS are intricate devices that can have several different moving parts and coupled together with
23
other MEMS can sense, analyze and perform complex operations in addition to being able to control and actuate motion on the microscale (Hilbert et al., 2000). MEMS have been labeled one of the most promising and relevant technologies of the 21st century (Hilbert et al., 2000). Revolutionizing industrial and consumer products and processes, their steady infiltration into everyday life has begun to dramatically improve and change the way we live (Madou, 2002).
2.2. MEMS foreword The acronym MEMS is used today to define both the fabrication processes and the devices resulting from these processes (Pryputniewicz and Furlong, 2003). The processes are a result of merging of advanced micromechanical and integrated circuit (IC) technologies. The methods used for making MEMS are similar to those used in the silicon wafer/chip market (Baltes et al., 2002). They are in fact mostly fabricated using silicon wafers or some variation of them. The packaging of MEMS devices also still is, for the most part, based on how chips are made (Pryputniewicz, 2003a). Packaging is a technological barrier that must be worked out, at this time, for each specific application of MEMS. Adaptation of chip fabrication processes allowed development of bulk and surface micromachining and highaspect ratio micromachining (HARM) (Gormley et al., 2000). Advancements in these fabrication processes allowed construction of threedimensional devices in the micrometer scale, Fig. 2.3. 24
Digital Light Processing) (TI. 2001). this was the first time it was applied at a micrometer scale. 2002). steam engines. more and more complex devices could be made. development of humidity. 2003).3. Most of these devices have gone from research to actual commercial products. a giant leap from the early versions. The first true MEMS using current fabrication methods appeared around 1988 as gear trains and tongs (Hsu. Today’s gears are capable of rotating at one million revolutions per second (Pryputniewicz et al..(a) (b) Fig. the white bar at the top left represents 500 µm. 2000). motors. After about a decade of research. one of the most complex MEMS devices became a commercial product known as the DMDDLP™ (Digital Micromirror Device . Even though the technology for lithography was around for about 200 years. and chemical sensors soon followed.. optical devices. and actuators have been since made. Beginning with accelerometers and pressure sensors. the white bar at the top left represents 200 µm (SNL. to name a few (Hilbert et al. temperature. 25 . With the continued advancement of the fabrication techniques. (b) a close up on the microchain. A microchain device: (a) overall view. RF switches. An example of this occurred in 2002. complex devices such as micro turbines. These relatively complex devices were the product of the fabrication techniques such as lithography and etching. 2. In addition to sensors. 2000).
Details of a DMD chip (TI. 2003): (a) graphic illustration of the construction of the DMD. This MEMS device utilizes a multidisciplinary approach (mechanical. e.. 2003).3 million tiny mirrors characterized by 20 µm diagonals. Details of construction of the DMD are shown in Fig. 2.) simultaneously to produce digital images that are among the best in the world (TI. (c) close up with one mirror removed from the surface. (b) (c) (a) (d) (e) Fig.4.4. (b) close up on the DMD array. electrical. in digital projectors capable of displaying highresolution images by utilizing and rapidly moving some 1. comprising a single MEMS DMD. (e) close up of a single element of the DMD without mirrors. there are indeed many additional benefits (Madou.The DMD chips are used.g. etc. Small sizes imply that less material 26 . 2. optical engineering. 2002). While it may seem that the only advantage for MEMS are their small size. each representing a single pixel. (d) all the mirrors removed from the DMD.
individual molecules can be moved to make devices only a few atoms in dimension.3. 2002b). and microelectronics. microactuators. Jr. Fig. et al. Perhaps the most prominent advantage to MEMS is the financial factor. more accurate. 1992.. 2.is used and less energy is consumed. MEMS devices are rapidly making their way into every aspect of modern life. microsensors. or in the nanometer scale. and MEMS technology is aiding in the development of NEMS (Nanoelectromechanical Systems) technology (Feynman. The future is getting smaller.. which opens entirely new product markets. true nanotechnology. and quicker.5. the cost per unit can be driven down to affordable prices. Many sensors are also breaking records of sensitivity with some chemical sensors having the capabilities to detect presence of individual molecules and atoms (Britton. 2. These are the general components that make up MEMS as we know them today. 27 . This can also allow development of disposable devices. is similar to MEMS only that it deals with devices three orders of magnitude smaller in dimensions. At this scale. Their small size allows for the construction of arrays of hundreds of them on a single chip. 2000). By being able to produce thousands of devices on each individual silicon wafer. NEMS. Pryputniewicz. Sensors MEMS typically contain the following components: mechanical microstructures.
The popularity of these sensors is mostly due to the great advantages that they posses. A schematic illustration of MEMS components and their interdependence (Madou. Fig. They have become varied in their applications and can be found almost everywhere in everyday life. Regardless of application. the original use for MEMS was as a sensor (Madou. In fact. With each wafer producing thousands of sensors affordably. 2002).5. Some utilize 28 . 2002). which are unparalleled with macrosized sensors. 2000). MEMS sensors consume very little power and are capable of delivering accurate measurements.. In addition to their small size. Perhaps the most appealing factor to MEMS sensors is that they also are very inexpensive to make. 2. all MEMS sensors work on the principle of measuring some form of a change. these sensors can now be utilized in areas that were cost prohibitive before with other means.One of the most common applications of MEMS is by utilizing them as sensors. just as any macro sized sensor (Hilbert et al.
While methods may alter. they are among the most commonly used. optical. chemical. magnetic. but are not limited to. and electrical changes.3. or accented. It can then be detected either externally with optics or internally with electrical electrodes. It should be noted that each of the methods also depends on electrical interconnections between different components of the MEMS sensor in order for it to function.4 describe these methods in more detail. All sensors measure a change and MEMS devices do it with either one or a combination of the following four methods: mechanical. These methods are generalizations for the basic system in which a MEMS device gathers information from the surrounding environment. MEMS sensors are able to detect a multitude of different measurable changes. by relevant external force(s). 2. While measurements by MEMS sensors are not limited to these four methods. This motion is caused. Sections 2. One of the most common is 29 . mechanical. thermal. These are some of the most appealing and intricate of the sensors. These include.1 through 2.1. electrical.3.3.common sensor methods while others take advantage of the benefits of the small scale. Methods of operation of MEMS sensors are different depending on the use. and/or chemical. Mechanical detection Certain MEMS sensors measure changes based on mechanical displacement or movement. While there are many methods along with many types of sensors there are a few mechanisms that are common among them. radiant.
These changes in capacitance are interpreted into usable data.6. 2003). 2002). A view of the “sensor” part of a dualaxis accelerometer where the proof mass can be seen with all the perforations in the middle and a quarterview close up on one of its corners (AD. Hsu. Fig. which can be measured electronically as changes in capacitance. there is a proof mass that is suspended on springs and allowed to move in one or two directions. There are several different 30 . 2. which also is one of the oldest type of MEMS sensors. but because of operational requirements of the device. 2.6.6). 2. changes relative positions between stationary and moving parts of capacitive pickups in the microaccelerometer (represented by long fingerlike protrusions in Fig.the MEMS accelerometer (Bernstein et al..6. This motion. External forces act upon the proof mass causing it to displace. in turn. 1999. A sample accelerometer is shown in Fig. 2. In the MEMS accelerometer. Another common type of mechanically actuated sensor is the pressure sensor. Fig. outofplane motion is limited.
which in turn produce an electrical signal. Optical detection While most of the applications for MOEMS (microoptoelectromechanical systems) are in communications there are a number of uses in other areas (Madou. Due to changes in the chemical composition the light passing through will be different than without the gas. Johari. or altered by some space containing a medium in question. These stresses are picked up by electrical piezoresistors on the diaphragm. 2003).designs for MEMS pressure sensors.3.. Sensing differences in incoming light compared to the outputted light is the basis for optical detection outside of communication purposes. this sensor works by introducing a gas into a chamber and then passing light through the chamber. By measuring this signal and interpreting it. Wuttig et al. 31 . Another example is the FabryPerot optical sensor (Han et al. Simply said. This shows the multidisciplinary aspect of these devices. 2. Typically. they are constructed as a cavity covered with a diaphragm. 2002a. the pressure is determined. reflected off.. 1996). gage. 2002).2.. it deforms. 1999. When an external pressure is applied to the diaphragm. but they all can be classified as either one of the following three types: absolute. Some of these sensors are for the detection of chemicals. Light is usually passed through. and differential pressure (Pryputniewicz et al. due to stresses produced by applied loads. 2002). Infrared spectrometry is an example of a method that can be used for optical detection (Wang et al.
This sensor works similarly to a test tube experiment. Piezoresistors are components that convert mechanical energy into electrical and vice versa (MS. 1998).3. 2. Electrical detection The bulk of all MEMS sensors rely on electrical connections for power and as means of transferring data from the device into the macro world where they can be accessed and analyzed. In the broad sense. Effects of these reactions are known and can be used to determine what chemical caused the reaction. parameters defining the motion of the body can be calculated. From the signal measurements. These effects are often exploited by. 2. Piezoresistors are used for the production of electric currents.4. It is based on calculated and known reactions between certain chemicals.3. By attaching a piezoresistor to a mechanism that moves. but not limited to. When specific chemicals interact a reaction occurs. Chemical detection The detection of chemicals using other chemicals is the basis for the chemical detection method.This light is analyzed and from the results a determination can be made about the type of gas being tested. the use of polymers. an electrical signal can be produced and measured.3. 32 .
color. there is a chemical reaction that causes a change in color or luminescence in the bead. Polymers for MEMS can be engineered to be sensitive only to specific types of chemicals. Thundat et al. When these chemicals are present near the polymer. (b) close up of a well. One way of utilizing this selective polymer sensitivity is with the artificial tongue sensor. illuminescence. 2.a polymer is a laboratory made chemical that can be used for a variety of purposes including controlled. 2001. The analyte is released over the wells and the polymer beads absorb the selected chemical. as shown in Fig.. Because of the illumination from behind.g. Once the polymer absorbs the suspect analyte. (a) (b) (c) Fig. (c) close up photograph of a FabryPerot well (Yoo et al. 1997). or FabryPerot sensor (Yoo et al. Micro storage wells are filled with tiny beads of a polymer designed to attract a specific chemical.7. This change is detected with a CCD 33 .) of the polymers. opacity..7. 2000). calculated reactions (Lange et al. it is possible to detect these shifts in color.. This effect is particularly useful and is utilized in several different ways. they get absorbed (or adsorbed) onto it and change one or a few of the properties (e. etc. conductivity. resistivity.. 1997).. Polymer based FabryPerot sensor: (a) graphic illustration of many wells with different polymers. 2. and are illuminated from behind.
2000. Cantilever elements also rely on polymer absorption. 34 . is coated with the polymer of choice that is then exposed to the suspect analyte (Ilic et al. Since each well is very small. 2. but utilize it differently. and means that the sensor identified the suspect chemical. With the cantilever attached to piezoresistors. this change can then be measured and the analyte can be identified. however. A micron sized cantilever. which in turn changes the fundamental frequency of the cantilever. Fig.(Charge Coupled Device) array. 2. it is possible to make many wells. 2002). is with cantilever sensing elements.8. 2001). each with a different polymer designed for detecting a different chemical. it changes the mass of the cantilever. By having many wells filled with different polymers it is possible to measure complex analytes quickly as well as being able to detect multiple analytes simultaneously. Fig.8. A far more robust method of using polymers. When the polymer absorbs the desired chemical.. Micron sized cantilevers developed at IBM for research purposes (IBM.
Although there are many other materials that can also be used. In most cases. MEMS sensors bring a novel solution for that requirement in small devices with high sensitivity.2.16 discuss some of the aspects of MEMS cantilever chemical sensor technology and while not all of them will pertain directly to this Thesis they were inserted for completeness and future work. They can be simple or highly complex. 2002). 1982. 2. offer better properties 35 .4. 2002b).4. 2002b). for a cantilever sensor to detect mass changes in picograms (Madou. 2002). they are very accurate. they are not without flaws. Some other materials.4. There are many aspects to chemical sensors.4. Pryputniewicz. Sections 2. One aspect of these sensors is that they be customized for most applications. Material The most commonly used material for fabricating MEMS devices is silicon and silicon compounds (Madou.1.1 through 2. It is possible. MEMS cantilever chemical sensor Detection of chemicals is required for many industries. Many different technologies must come together in order to allow development of MEMS (Pryputniewicz. the overall properties of silicon are very good at small scales (Petersen. such as silicon carbide or diamond. Combination of the yield strength and modulus of elasticity (among other properties) of silicon is especially useful for the life and durability of small devices. While they seem like the perfect solution. for example.
This crystallographic orientation of silicon allows MEMS to have well defined and sharp edges and shapes with perpendicular angles resulting from etching as described 36 . Ingots are bars of silicon (or of any other chip materials) with 99. 2. (b) visual inspection of silicon wafers (MEMC. The ingots are specially made with the crystallographic orientation of the silicon arranged into either [100] or [111] as these are the most commonly used (Madou. Initial steps in fabrication of MEMS and CMOS chips: (a) preparation of a silicon for slicing. 2002).but they also cost many times more than silicon making them uneconomical except for the most demanding and specific purposes (Thaysen et al.. 2003). For the fabrication of MEMS and CMOS (complementary metaloxide semiconductor) chips alike.99% purity. Purity is essential to making consistent flawless devices. 2.9. Silicon ingots and wafers are shown in Fig. 2002). (a) (b) Fig.9 during fabrication and after slicing. silicon is first processed into ingots of about 4 feet in length.
2002a).4. Fabrication The fabrication of MEMS is an entire field of research on its own (Pryputniewicz. There are several different processes for producing MEMS. is lithography. the technology is continually improving.4. electrodeposition. Lithography currently produces the smallest and most exact features available in MEMS (Pryputniewicz.in Section 2. and the components are released to form 3D 37 . and easier to fabricate in ever increasing quantities. which has many different submethods. however. With the growing number of companies that fabricate MEMS. better. The technology is also pushing the size envelope and continually working towards the development of smaller devices such as NEMS (Madou. patterned. The ingots are then sliced into wafers usually ranging from 100 mm to 300 mm in diameter and about 1 mm in thickness. It is extremely diverse and still expanding (Madou. 2002). It is comparable to dry etching. 2003a). and molding) in the size of the features made.2. 2002). Surface micromachining and dry etching have an advantage over other techniques such as bulk micromachining and LIGA (German acronym for xray lithography. most prominent of these.2. These wafers act as the substrate or foundation for all MEMS devices. Layers are deposited on one another. MEMS technology becomes more affordable. With each breakthrough. MEMS devices are fabricated in a manner similar to making a cake. 2.
2002a). One place of such improvement is at Sandia National Laboratories (SNL) in Albuquerque. capable of fabricating a 5 layer MEMS using a sacrificial surface micromachining process. Fig. Depending on the process. the layer is grown in a vapor filled environment. (d) example of the 5 structural layers of deposition (SNL.structures. (c) a thermally actuated motor. In most cases. (a) (b) (c) Fig. 2.10. The rest of the world is limited to 3 layer fabrication processes. 2001). but there have been improvements in it. These improvements include automation of the steps.1. Examples of SUMMiT™V process: (a) a rachet gear. (b) a clutch activated transmission. SNL has developed an automated process known as SUMMiT™V (Sandia’s Ultraplanar MEMS Multilevel Technology). as described in Section 2. the fabrication process can take up to a few months. 2. Using one of the processes available. SNL is the only place in the world. which is capable of fabricating MEMS devices out of up to 5 structural layers in complexity. New Mexico (SNL. 38 . a MEMS is built up one layer at a time. 2001).10 (Pryputniewicz. the number of layers. MEMS devices are fabricated on wafers made usually of silicon. This automation has allowed the price of MEMS to be reduced while improving quality of the devices that are produced. at this time. and the complexity of the design. This growth can be time consuming.4.
11. the fabrication of each layer is quite complex.4. It is too vast for the scope of this Thesis and will not be addressed herein.al. While MEMS are small. 1986. they will etch away leaving the desired shapes. 2002.3.While it may seem simple. The next layer is sacrificial and it is also exposed to masks. the polysilicon is exposed to masks. Masks somewhat resemble cookie cutters cutting out shapes in a layer of cookie dough. and chemicals to etch holes and/or dimples. 2002. 2001a. Packaging of MEMS is application specific. and chemicals to continue with the fabrication. light. The development of MEMS devices has come a long way and new boundaries have been reached (Pryputniewicz. light. Starting with the deposition of the first structural polysilicon layer onto the wafer. 2. certain parts of a given layer so that when they are exposed to specific chemicals. A problem still remains however in being able to communicate efficiently with these devices. 2002b. 2003). 39 . Fig. 2002a). or soften. Packaging One of the biggest hurdles in MEMS technology is with packaging. Using lithography. Pryputniewicz. Pryputniewicz et. the packages that contain them are still considerable in size to facilitate their handling and assembly. Pryputniewicz. Madou. The light used is to harden. 2. for example. a series of steps are required for each layer (Hsu. 2002. The deposition and etching steps alternate between the structural polysilicon layers and the sacrificial SiO2 layers making shapes in each layer and continuing until the desired device is fabricated. Hsu..
40 .3. Sensing types and methods As discussed in Sections 2. 2000a) and are also capable of measuring pH values in analytes (Bashir et al. Of the different types of sensors.11. MEMS chip and its different levels of packaging (SNL. 2002).1 through 2.3. there are two basic methods of operation to detect chemicals. The two methods could be called contact and noncontact methods.4. simple composition chemicals or complex biological agents.4.4. Most of the different types of chemical sensors available are targeted for the detection of specific chemicals. MEMS sensors are used for a variety of purposes. 2001): (a) MEMS component. 2. (c) Macro sized chip containing the MEMS component 2. (b) MEMS component contained sealed chip.(c) (a) (b) Fig. With so many different types of sensors. it is necessary to categorize the chemical detection methods to better understand MEMS.. Chemical sensors have even been used for the detection of DNA strands with different base sequences (Fritz et al.. These can either be regular.
The second way is by measuring change in stress in the element. Subramanian et al. A change in these waves can be measured with sensors. This stress (or surface stress) change can then be measured by piezoresistors on the surface of the sensing element. Fig. The contact method uses polymers deposited on cantilever sensing elements to absorb the specific chemicals and thus produce a change in mass.. Lang et al. 2000). They function when the analyte comes in contact with the polymer “containing” either a surface or a bulk wave that travels through the material.12.12 (Ivanov. 2002.. 1999. electrical. 2. Two other methods of interest are SAW (Surface Acoustic Wave) and BAW (Bulk Acoustic Wave) sensors. 1999. Kerness et al. Thaysen et al. or thermal properties of the element. 2001. The other way to detect the suspect chemical is to use a thermocouple to measure the change in temperature due to heat produced from the absorption of the analyte by the polymer. which has mass as one of its parameters.. the analyte is observed from a distance and its behavior and/or characteristics are measured and studied. 2000.. The downside of these methods is that they have a much lower accuracy when compared with the methods described earlier in this section. Fig. 2000. 2. stress. 2003).. Jensenius et al. In noncontact mode.. For a change in mass.Contact methods include any system in which the analyte is physically touched or altered. which occurs when mass is absorbed into the polymer. All of these methods have been successfully incorporated into chemical sensors (Koll et al. Baselt et al.. 41 . an increase in mass can be detected by measuring the change in frequency of the resonating cantilever.
Should one or the other of these parameters change. which in turn shifts the resonant frequency to some other frequency.12. then in addition to detecting the chemical. the resonant frequency will also change. Operation of different gravimetric sensors: (a) graphic illustration of a cantilever bending under the load of analytes attaching to the polymer coating. Some experiments have successfully 42 . 2000). one can determine the amount of mass that was added to the cantilever assuming the spring constant remained constant. This method of chemical detection is highly accurate. Alteration in the mass of a vibrating cantilever changes the total mass. If the cantilever is selective to the kind of analyte it absorbs. if present.. This is the basis behind using change in resonance frequency as means of detection. 2. By measuring this shift.5. 2000). Resonance frequency is the frequency that produces the largest amplitude that a vibrating body can achieve.4. Resonance frequency The main principle behind cantilever sensors is with the measuring of change in the resonance frequency (Betts et al. This frequency is dependent on the spring constant of the body and its dynamic mass. the sensor can also detect the concentration of it. (b) illustration behind the principles of BAW and SAW (Ivanov.(a) (b) Fig. 2.
which is capable of vibrating at different frequencies very accurately depending on the driving voltage (Harley. 2001). Figure 2.. Stephan et al. Cantilever actuation methods In order for the MEMS sensor to function in dynamic.. 2002). Another method that is less common. 2002. the most common is by the use of a PZT (PbZrTi transducer or. Indirect methods include using acoustic vibration to shake the sensor much like a piece of paper in front of a speaker vibrates from the compression waves in the air. Another indirect method has a magnetic coat deposited on the cantilever and an inductance coil underneath it to generate magnetic fields that attract and repel the cantilever into vibration.6. 2003). Li et al.4. 2001. it is necessary for the cantilever to be vibrating (Arecco and Pryputniewicz.13 shows some sample PZTs available commercially. Turner and Zhang.demonstrated the ability to detect a change in mass of as little as 0. Mehta et al. but of great interest. There are several ways of actuating a cantilever into its resonance frequency. as it is sometimes called. 43 . response mode. 2002.. For direct methods.7 picograms (Madou. 2. or frequency. piezoelectric zaxis actuating transducer). is by fabricating piezoresistive electrodes directly on the cantilever and bending it rapidly into vibration by bimorph actuation. 2002. which can be classified as direct and indirect actuation methods (Cho and Ahn. 2002).
There are several ways of accomplishing this and the most relevant to this Thesis will be discussed. It is of interest to combine and use direct and indirect methods when conducting research. 2003).13. Comparison of different sizes and shapes of PZTs available: (a) OpenLoop LVPZT Translators.(b) (a) (c) Fig.. It is even possible to combine both direct methods into one package that allows for highspeed functionality (Kim et al. 2003). (c) PICMA™ Chip Monolithic Multilayer Piezo Actuators (PI. 2. (b) Preloaded Open. 44 . 2.& ClosedLoop HighLoad HVPZT Translators. Cantilever I/O In order for the sensor to function it is necessary to have some means of communicating out the data it collects.7.4.
2002). 45 . Graphical illustration of the operational principle for an AFM where different bend angles reflect the laser beam to different locations on the PSD. 2001). Using a technique similar to those in AFMs (atomic force microscopes). 2... a laser is reflected off the cantilever tip and into a position sensitive diode (PSD) (Raiteri et al. Fig.One of the most common ways of detecting frequency shifts is by optically inspecting the cantilever.. Lange et al. 2003. Any changes in the surface stress due to bending will cause an electrical output from the piezoresistive electrodes. While the technique is proven and quite reliable. This change is detected by the PSD and it can be measured and analyzed. The reflected angle changes due to the bending of the cantilever. it is too bulky for portable purposes.14.. 2. Fig. 2001. Kim et al. Another method that is also quite popular is by using piezoresistive elements on the cantilever itself (Porter et al..14. Battiston et al. 2001a.
.1. A thicker layer of polymer 46 .15. the longer it takes for the polymer to reach saturation. The thicker the polymer. (b) close up of the Wheatstone bridge piezoresistive elements (Gotszalk et al.. 2000).4. (a) (b) Fig. 2. Cantilever with integrated piezoresistors: (a) view of piezoresistive cantilever used for scanning probe microscopy (SPM) applications. 2000).14.This output can be analyzed and the frequency can be calculated. The thickness affects several aspects of the sensor. Mostly. however.8. It is generally preferred to make the coating as thin as possible. This is not really a problem unless time is a critical factor in the function of the sensor. it affects the rate at which the polymer becomes saturated with the suspect analyte (Britton et al. Figure 2.15 shows how a typical piezoresistor is used on a cantilever. to minimize other effects. Polymer thickness In fabricating cantilever sensors with special coatings it is important to take into account the thickness of this layer. 2. One of the most prominent of these effects is that of residual stresses which will be discussed in Section 2.4.
11. a gas (or liquid) with a higher concentration of a sample will move into an area of lesser concentration. 1988. 2. such as thermal effects. 47 . 2. as it would be in an ideal condition with infinite volumes of gas passing through a membrane..4. the rate is constant. Because of signal noise from a variety of different sources. Diffusion is usually an interstitial motion in that the sample gas travels to the voids in the lattice structure of a material. Streeter et al. 1998. Reference cantilevers In order to guarantee sensor functionality and reliability. These laws address two specific conditions.4. 2001). which are discussed in Sections 2. Hu. or mass transport. Fick’s law of diffusion For any particle that is absorbed into another there is a rate at which the diffusion takes place (Hughes and Bastasz. the data quality must be high.16. steady state and nonlinear (nonsteady) diffusion.is more reactive to effects of different coefficients of thermal expansion and also it is more sensitive to pressure effects. The nonlinear state (second law) is a more accurate model for finite volume cases where the diffusion rate will change as the concentration gradient equalizes. During diffusion.10.4. The diffusion can be modeled with Ficks’ laws. During steady state diffusion (first law).
. 2. It should be noted that Q has a great dependence on damping and thus the environment that the structure operates in. 2000). In other words. The signals of the reference and coated cantilevers can then be compared and external forces (noise) can be removed computationally thus improving quality of data. which means that they are not coated with the analyteattracting polymer (Thaysen et al. Quality factor Most vibrating mechanical structures have a variable known as the quality factor. Reference cantilevers function by having them react to only the noise producing effects. the higher the Q value is while the opposite is also true.. the structure can operate longer and produce more data.4. it is important to include a parallel system to improve the signal. Q (or Qvalue) (Yasumura et al. the Qvalue is a measure of how much energy is dissipated in a vibrating system. The narrower the peak of the frequency response curve.12. The width of the peak is also sometimes referred to as the 48 . The quality factor is reflected in the shape of a resonance frequency peak. The higher the Qvalue. By incorporating a reference cantilever it is possible to minimize noise and improve the signal quality. Structures in air usually operate much better than those in liquids and the Q will be the highest in a vacuum environment (Tamayo et al. 2001). If the structure can conserve energy efficiently.. and others. the better the structure is capable to conserve energy. This value is a nondimensional parameter for quantitatively assessing the ‘quality’ of a structure. 2000).electrical noise.
13. The signal produced must be directly linked to the motion. the device as that will produce the most accurate results. 49 . or motion. The piezoresistive effect is observed in piezoresistive elements and materials. Electrical aspects Practically all MEMS devices collect measurements electronically..bandwidth of the resonance frequency curve. 2002). When a strain is produced in the gauge. Thaysen et al. They are used for detecting the action. The gauge is usually attached to.4. though. that is incurred in a sensor. This signal can then be amplified and measured. This mechanical to electrical effect is accomplished with a phenomenon usually referred to as a piezoresistive effect (MS. The piezoresistive effect is the change in electrical resistivity that occurs with the application of mechanical stress. which allows conversion of stress into proportional. The operation is based upon measuring increasing stress/strain in an element of the MEMS. measurable electrical signals. an electric signal is generated. This means that there must be a mechanism for producing a signal to the macro world. or action.. or built into. This bandwidth is measured at half power points. that the MEMS performs in order to determine quantitative results. 2002. 2. for consistency and convenience. One of the most common gauges used is a piezoresistive strain gauge (Chow et al. 1998).
(usually the strain gauge or other measuring gauge) acting as a gate switch. Fig. Feed back loop Because the method of extracting data is usually electrical.4. Fig. Wheatstone bridge The basic and most commonly used method of gathering and measuring the electric signal generated from gauges is with a Wheatstone bridge circuit. 1958).16. e. with the 50 . Wheatstone bridge where changes. This noise and errors in the data can be reduced.1.13. which can change its resistance with mechanical changes. The Wheatstone bridge is a classic and common configuration for determining voltage differences in electrical circuits. there is typically noise present in the results.g.2. alter the output voltage with a high accuracy and precision. however. 2. Four resistors are connected together with a middle resistor. in R1.4. R1. For MEMS applications. the middle resistor would be a piezoresistive element.13. Any variation in current in the middle resistor will cause a difference in voltage output from the circuit (Pauw. 2.16..2. 2.
.use of feed back loops (Abadal et al. residual 51 . Tweaking and adjusting of techniques and design can overcome most. There are two problems. and complexity concerns.1. Manias et al. Feed back loops is a post processing technique used to increase accuracy of the data in which the data are refined several times. 2002). Inherent problems In designing a MEMS device....14. Because of differences in coefficients of thermal expansion (CTE) of the materials. It is utilized with computers and special circuit setups. Residual stresses During the fabrication of a multilayered MEMS device two or more dissimilar layers are deposited on top of one another depending on the process and design. there are a multitude of variables that must be accounted for. Kassegne et al. 2000). 2000. 2. however. there are inherent problems that arise (Handel. that are reoccurring and of some concern: residual fabrication stresses and stiction. though. 2001.14.4. equipment.. Humphris et al. With so many variables.4. Sulcheck et al. These can be very complex systems and require a lot of knowledge in electrical engineering. Active feed back loops continuously correct the data being processed so that error is minimized. 2001. Because of time. feed back loops will not be incorporated into the experiments planned for this Thesis. 2001. 2.
Anneling is a process of heating the device at a certain temperature for a specific time in order to allow for any stress to dissipate... 52 . A lot of the fabrication techniques involve depositing layers at high temperatures (600˚C +). cracks. 2003b). Effects of residual stresses: (a) a MEMS device is bent after its release.17 shows the effects of residual stresses on a MEMS component. (a) (b) Fig. 2000b..17. 2. When the fabrication is over and temperatures return to room temperature (~20˚C). 2003a. This is a proven process. 2002). This is a constant design problem that must be taken into account. but requires a lot of experimental verification and takes up valuable production time. Lu et al. 2002c. One method of reducing and/or eliminating residual stresses is by annealing the device after the deposition process. 2. or other failure modes. 2001. Fig. Pryputniewicz et al.stresses will be produced in these layers due to the fabrication processes (Fritz et al. These gradients will manifest themselves as warping of structures. (b) a close up on the device showing how extreme the bending is (Dartmouth. 2003b). different materials will shrink (or expand) by different amounts leading to stress gradients (Pryputniewicz.
18. Another major problem with stiction is that it can occur at any time and not just during fabrication.2. 2001). stiction can cripple a device making it useless. There are solutions for this problem. It is possible to fix stiction.18.4. This close proximity along with forces such as van der Waals cause the components to tend to stick to one another. 53 . 2.2. Collapse of a thin cantilever due to stiction. Stiction occurs when two flat surfaces are near each other. In a cantilever type sensor. stiction will hopefully become a problem of the past in the near future. With some limited solutions already in place and continued research. Fig. but during that process excessive forces are required which can damage or destroy the device. Fig.14. New organic modifiers (coatings) have been developed against stiction and have been tested and shown to lessen the effect (Kim et al. 2001b). 2. Stiction is probably the most serious of problems that occur in micromachining (Buks and Roukes.. Stiction A very common problem that plagues devices with small thickness (1 to 4 µm) is a phenomenon known as stiction.
This increase in damping reduces the Qvalue and can also 54 . 2. 2000).g. as this would lessen the amount of compensation required from the reference beams.. Effects of pressure The pressure of the environment in which the sensor is in can affect the results produced. If the pressure of the environment increases. The effects of temperature have been. Increase or decrease in the temperature will produce a surface stress on the cantilever due to different CTEs and thus create a pronounced bending as shown by Stoney (1909) and Pryputniewicz et al. Any temperature change in the environment will affect the beam (Gotszalk et al. For producing better results and data it would be desirable to reduce the temperature induced bending.. the dynamic viscosity of the damping fluid increases. If any temperature change will be expected it is necessary to include reference beams to allow for the compensation of this phenomenon (Hanson et al. however. 2000). This means that the temperature of the sample should ideally remain as constant as possible.16.4. 2001. Pryputniewicz.19 (Kerness et al. 2001). Fig. Effects of temperature Temperature has a critical effect on the functionality of the MEMS sensors (Hanson et al. This bending can alter results of any detection system setup...2. (2003a).. capitalized to create calorimeters with femtorange accuracy. 2003a).15. Because the cantilever sensor is based on a bimaterial ( e.4. substrate and polymer coating) there will be two materials and thus two coefficients of thermal expansion. 2.
55 . decreases the data quality and reliability. an increase in pressure will also alter the concentration of the gas in the environment. which can affect the absorption rate. 2000).19. Fig.reduce the deflection of the cantilever which both. 2.. in turn. A femtocalorimeter utilizing thermal bending effects with piezoresistors (Gotszalk et al. and equilibrium giving shifted results. In addition to changing the damping.
the labeling of the cantilevers (AF) shall be used from here on to refer to each cantilever. The company is headquartered in Estonia. In fact. AFM chips are siliconbased cantilevers that have a high aspect ratio tip at the end. 3. Spain. As can be seen from the photographs. and their main product are atomic force microscope (AFM) chips.1 applies to the samples used in this Thesis. The nomenclature illustration in Fig. MEMS SAMPLES This chapter describes the MEMS samples used in this Thesis and their characteristic parameters and includes information for the proposed MEMS chemical sensor. Figure 3. the chips 56 .1. model number CSC12 (MM.1 shows an AFM chip with cantilevers with tips from MikroMasch. 2003). 2003). while Fig. Table 3. MikroMasch USA cantilevers Sample cantilevers used in this Thesis were purchased from MikroMasch USA (MM. as only the three longest cantilevers (DF) shall be investigated in this Thesis. They are used for topographic surveying on the nanometer scale of the samples investigated. The samples used in this Thesis are actual AFM contact silicon cantilevers that are uncoated and also are without any tips. They are micromachined cantilevers with the fixed ends being integral parts of a chip. 3. 3.3. It is important that the cantilevers be uncoated so that a special coating can be deposited on them without any other underlying layer.2 shows the special samples purchased for this Thesis.1 lists the manufacturer’s specifications of the samples purchased.
Characteristics Length. and the corresponding material properties as listed by Madou (2002).01 0. L ± 5 µm Width.3 14 20 28 0. f kHz Force constant.6 1.7 1 1.03 0.3 95 155 230 0. b ± 3 µm Thickness. b ± 3 µm Thickness. The cantilevers and the chips are all made of singlecrystal silicon with the dimensions and material properties listed in Table 3.08 0. h ± 0.3 7 10 14 0.5 14 19 0. 3.05 g/cm3 C Min Typical Max 130 35 0.1 190 57 B Min Typical Max 90 35 0.02 0.25 0. Characteristic dimensions.7 1 1. E ± 3 GPa .7 1 1. k N/m Characteristics Length. all listed in Table 3.1.15 0.45 1.. 2003).2 2.3 65 105 150 0. as supplied by the manufacturer.1. k N/m A Min Typical Max 110 35 0. L ± 5 µm Width.1.3 µm Resonance freq. MikroMasch AFM chip: (a) SEM photograph of cantilevers (A.05 0.C) with tips.75 5 E Min Typical Max 350 35 0.01 0. f kHz Force constant.contain 6 cantilevers of varying lengths. The thickness and width of each of the cantilevers should be the same according to manufacturers specifications. (a) (b) Fig.95 2. Table 3.3 50 75 105 0. h ± 0.3 µm Resonance freq.B.7 1 1.08 Density.3 9.7 1 1. ρ ± 0..1.5 D Min Typical Max 300 35 0. MikroMasch USA.7 1 1.33 Modulus of elasticity. (b) dimensions of the chip and cantilevers as well as a standardized nomenclature (MM.5 F Min Typical Max 250 35 0.
2.E.(a) (b) (c) Fig.E. While it is possible to detect just about any analyte with the right polymer. Photographs of the AFM chips used in this Thesis: (a) a view of the entire chip from above with all the cantilevers visible. 3.2. some polymers are easier to make and are more accessible than others. Due to the focus of this Thesis and time/cost issues. Thus it is important to select a chemical that is easy to detect and useful for research purposes. In the process of coating.F).F). (b) a top view of the primary cantilevers used in this Thesis (D. there are 58 . availability played an important role in determining the polymer/analyte combination to use. 3. Chemicals used This Thesis is based around a sensor that is capable of detecting a specific chemical. The cantilevers of interest are the ones on the right. (b) a photograph of the underside of the cantilevers (D.
only hydrogen will diffuse through it. This is a significant amount for a very reasonable condition.2. Palladium The element palladium (Pd) was selected as the absorbent layer for the cantilever because of its sensitivity to hydrogen (Darling. The process is also reversible. it has an ability to absorb (not adsorb) great quantities of hydrogen. A true benefit of this interaction for the Thesis is that the absorption will occur at room temperature. The palladiumhydrogen system is very well documented and very easy to conduct experiments with (Lewis. absorb 800900 times its own volume of hydrogen (Lewis. Most are not used for a gravimetric chemical sensor. Most systems are not reversible in absorption without some kind of treatment to the absorbing material. While there are other absorption systems. Palladium will.1. but there are a few that are. 1967). 59 .many different types of materials that are fairly standard and can be found and used easily and quickly. they are much more complex and can be very difficult to set up and maintain. at room temperature. 1967). palladium is often used as a filter of hydrogen as when it is heated. Properties of palladium are found in Appendix A. As mentioned in Section 3. One of these materials is palladium (Pd).2. It is in fact a very unique ability in that there are few other materials that naturally behave as palladium does with hydrogen. Because of this phenomenon. 1958). 3. These are highly desired features that simplify the experiments. The hydrogen then desorbs out of the palladium when the hydrogen source is removed.
Because of this range. During the deposition process.2. great care must be taken when setting up any experiment dealing in hydrogen to prevent a concentration greater than 4%. Hydrogen is explosive from 4% . When it burns the only byproduct is water.2. a very thin layer (approximately 10 Å of Ti) will be 60 . For this reason. Hydrogen The element hydrogen (H2) is the lightest and most abundant in the universe. special care will be used to ensure safety during the experiments. hydrogen will not burn. a thin layer of Ti will be placed before coating the cantilever with palladium.3. In the setup. Properties of hydrogen can be found in Appendix A. This means that if the concentration is below or above that range. It has been seen that palladium will delaminate in the presence of hydrogen if it is directly coated onto silicon. of the palladium from the singlecrystal silicon cantilever. or separation. This would be acceptable since this Thesis will investigate a sensor for detecting the lower concentration limits (<<1% hydrogen). The one fact about hydrogen that limits its use is its explosive nature. Premixing a nitrogenhydrogen gas with 1% and less hydrogen concentration reduces the risk factor.2. Titanium There will be a small amount of titanium (Ti) used in the coating of the cantilevers. 3.3. Nitrogen would be the inert gas carrier for this experiment as it is very stable and helps reduce humidity from the cantilevers.40% concentration. It is a very versatile element and a very promising fuel. The reason behind this thin layer is to help prevent delamination.
Even though the layer is very small. it is necessary to use a technique that will allow for the deposition of layers with accurate thickness. Ohio can be used. a sputtering machine can be used to coat layers of Ti and different. 3. There. For the calculations in this Thesis. controlled thickness of Pd on the cantilevers. it was assumed that the deposition process would be ideal in that there is a perfect bonding between the layers and there are no imperfections. Coating process Once all the materials have been selected for coating the cantilever. To verify this.used as an adhesive layer between the silicon and the palladium. it would be recommended to get two different Pd thicknesses: 1000 Å and 2000 Å. It was also assumed that the properties of the layers (including the silicon substrate) are isotropic in nature. it will be accounted for in calculations.3. The thickness for palladium. 61 . For this task. as will be discussed in Chapter 4. only affects the length of time for the absorption process to complete. This is a fairly safe assumption to make given the nature of the deposition process. The properties of titanium can be found in Appendix A. resources such as those at Case Western Reserve University (CWRU) in Cleveland.
as there will be (ideally) equal and opposite bending moments and stress gradients. and tip) shall be coated.8. Bending moment comparison: (a) the bending moments of a fully coated cantilever where the moments cancel each other. In addition to coating the entire length. It is assumed that the deposition of any titanium and palladium is uniform throughout and that this will also effectively minimize.1. 3.3. bending effects due to any differences in the values of parameters characterizing coatings can be minimized.3 illustrates the differences between bending moments of a fully coated on both sides (i. Calculations describing sensitivity and placement of the palladium on the cantilever are discussed in Chapter 4.3. left and right sides.. top and bottom) and a oneside (i. and sensitivity of the sensor. or even eliminate..3. all five sides (top. top only) cantilever. By doing this.e. coating the entire length of the cantilever is beneficial to the sensitivity. quality. As it will be shown in Section 5. bottom. Therefore. Figure 3. it is important to consider the placement of the palladium coating. 62 . any bending effects. (b) the moments of a oneside coated cantilever. where moments add to each another. (a) (b) Fig.e. Complete versus partial coating Coating the cantilever with palladium can alter the results.
3. it will be sensitive to humidity in the air (Baselt et al. as oxygen will be removed from the environment around the cantilevers. care should be taken to ensure that the cantilever experiments will be conducted in an oxygen free environment with as little humidity as possible. This would resolve any humidity problems that might arise.4. the time is just as long.3.. For this reason. the time required is exponentially high in that it will take an infinite amount of time for the hydrogen to reach complete equilibrium in the palladium. As the palladium oxide is exposed to hydrogen. Time limits For the palladium to reach 100% saturation of hydrogen. the oxide will slowly recede and the palladium will become pure again allowing for the hydrogen to absorb normally. which will slow or prevent hydrogen absorption. With the hydrogen being delivered in nitrogen–hydrogen mix. This will cause the palladium to expand as well. For desorption. taking up to days depending on the varying conditions of the system. Palladium will naturally start to form palladium oxide in the presence of water or oxygen. 2003). and potentially eliminated. For this reason.1. if not longer. This diffusion process is described in more detail in Section 4. the relative humidity will be reduced. which is unacceptably long for practical operation of the sensor. Effects of humidity Once the cantilever is coated with palladium.2.3. the experiments should be run till the saturation level is only 90%.3. 3.2. Because it is very hard to fully remove 100% of all the 63 .
hydrogen once it has been absorbed into the palladium. 64 . This residual hydrogen will keep the palladium lattice permanently changed. the first few runs are the most important for data collection and the total number of experiments conducted on each chip should be minimized while increasing the total number of chips experimented on to help improve the quality of the information gathered. there would be some residual left over. the absorption/desorption process will deteriorate the structure eventually leading the palladium to fail via cracks or delamination. Over the course of several runs. Because of this limitation.
2001a. 4. experimental solutions). for the equations to be comprehensive to all the aspects of the problem. These solutions are applicable when the boundary. The analytical part of the solution is based on exact. loading. and initial conditions can be specified. computational.1. and experimental solutions to obtain results which otherwise might be difficult to obtain. computational. The 65 . The computational solution is based on using finite element method (FEM) software to model the problem and to produce approximations of the solution. The information generated greatly facilitates any computational and experimental work that needs to be performed. Because of its nature. ACES methodology The methodology used in this Thesis is known as ACES (analytical. computational. Pryputniewicz et al. this study will be based on the analytical. close form solutions. which is to operate at the conditions of room temperature (~20˚C) and sea level pressure (1 atm).. 4. METHODOLOGY This Thesis conducts a study to optimize performance of a cantilever chemical sensor. Fig.4. This methodology uses analytical.1. and experimental solutions (ACES) methodology (Pryputniewicz. 2003a). 1997. or in other word.
66 . depends on the specific application. 4. a number of different methods have been developed. Fig. Because of continued efforts to advance experimental capabilities. and scanning electron microscopy (SEM) methodologies were used. atomic force microscopy (AFM). in this Thesis. Practicality of these methods. comprehensive testing using several methods can improve the overall quality of the data. laser vibrometry. For example. Graphical representation of ACES methodology and the interdependence of each solutions category towards determining a final result. optoelectronic laser interferometric microscopy (OELIM). However. Each method carries with it its own strength and weakness.solutions are based on discretization of the governing partial differential equations (PDEs).1. This is one reason for comparing the analytical and computational solutions to the experimental. however.
1990) listing different properties and characteristics of cantilevers. They will take into account different combinations of parameters characterizing a vibrating body. 2001. the results of each will be compared. Britton et al. 1991a. Sections 4. 1993. These parameters describe damping. 2003b.4. 1999. 2001c. There are a number of publications (Battiston et al. a sensor will be investigated that operates at room temperature and 1 atm pressure.. For this Thesis. 1996. and other characteristics of the cantilevertype MEMS sensor. 1985a. 1995. 1996. 67 . This will help in determining how each parameter affects the overall results.2. 2002a. Analytical investigations In order to commence calculations for the analytical investigation of the cantilever... 1995. 1991b. Depending on the specific set of parameters used..11 describe some of the necessary issues that need to be considered in the calculations for this cantilever based chemical sensor. Pryputniewicz. Some of the publications address issues that must be considered. Pryputniewicz and Stetson. 2002. stiffness. Pryputniewicz et al. the results might change. external forces. Several different cases will be considered while analyzing a vibrating cantilever.1 through 4. 2000. Once the different cases or scenarios are taken into account.2. Brown and Pryputniewicz. Bashir et al. it is necessary to understand what must be evaluated. 2002b. 2001b. 1988. 1998. 1992.2. Baselt.
4. the better it will function. 68 . Calculating the sensitivity is important in determing the best location for the polymer coating. The higher the sensitivity of the cantilever. (Madou. This will allow mass of hydrogen to accumulate on the cantilever leading to a shift in frequency. ∆f and ∂f are finite and infinitesimal.1. Sm.1) where f is the initial frequency. (4. It is also important to clarify the condition of the sensitivity. and ∆m and ∂m are the changes mass (see Appendix B). Changing the coverage of the cantilever by the polymer can alter the sensitivity of the sensor. It is important to determine the sensitivity of the cantilever. S m = lim ∆m → 0 1 ∆f 1 ∂f = fi ∆m fi ∂m . respectively.2 illustrates the boundary conditions and equivalent system of the cantilever. Sensitivity.. Additional Figure 4. By different ways. changes in the frequency. as there are different ways of coating the cantilever with palladium.e. 2002) i. of this sensor can be defined as a function of the frequency shift and the absorbed analyte mass. Sensitivity A way of measuring how well a sensor functions is by determining its sensitivity.2. The cantilever will be exposed to hydrogen until the coating of palladium reaches 90% of saturation. The sensitivity can then be defined as the ability for the cantilever to detect shifts in the frequency. it is meant that by coating the entire cantilever versus just coating a small portion at the tip with palladium the sensitivity will change.
r is the percent area coverage by the polymer layer of the cantilever. (4. 5. 4. r. The percentage area coverage. If we cover the cantilever only at the tip or with just a small percentage compared to the entire cantilever we get sensitivity for an end loaded cantilever. Idealization and modeling of a cantilever.3) 69 . KE. m is the original effective mass (see Appendix B. used was estimated from Fig. it will be assumed that there is no change in the spring constant. 4. Derivation of the dynamic mass utilizing the boundary conditions of a cantilever. and ∆m is the change in mass of the cantilever. For simplification of the problem. (4. n is the effective mass constant.2. equation KE = 1 2 mv 2 . as (see Appendix B for derivation) Se = −1 k 2 m knr 2 k ( m + nr ∆m ) ( m + nr ∆m ) . Se. uses the kinetic energy.16 of the dynamic mass effect on the cantilever.3).Fig. Fig.2) where k is the spring constant of the cantilever.2.
2002). C1 is a constant that is comprised of structural properties of the cantilever.. (4.4) where vmax is the maximum velocity experienced by the body in motion. It is now possible to compare the two equations and determine which will produce the largest frequency shift. the correct amount of coverage can be calculated to analyze the cantilever sensitivity. High sensitivity can be defined as the ability to detect a very small frequency response. 2 (4. In determining the sensitivity of the cantilever sensor it is important to maximize this response.5) where M is the mass of the cantilever. will be Sd = −1 k 2 33 M 140 k 33 ( M + ∆m ) 140 k 33 2 ( M + ∆m ) 140 .where m is the mass of the moving body and v is its velocity. Increasing the sensitivity increases the detectable shift in frequency of the cantilever. 2002. the final solution is 2 vmax KE = C1n( x) . Sd.2 to 4. Lange et al. and n(x) is a constant that varies with the position along the cantilever that describes the effective mass. If we now look at the case where the polymer covers the entire cantilever the sensitivity for a distributed load. 70 . If the equation is solved subject to the cantilever boundary conditions of a sensing element. 2001. Hsu.4. 2001. From Eqs 4... Yu et al. as it will directly determine performance of the sensor (Hu et al.
Equivalent variables In order to model the cantilever more accurately. 1999): ρeq = ρ SiVSi + ρTiVTi + ρ PdVPd VSi + VTi + VPd .Ti. 1975. 4.2.6) where ESi.3. it is necessary to account for all the layers and coatings on it. which in turn means that it has a different modulus of elasticity and density. Berthelot. respectively. 1999) Eeq = ( ESi ETi EPd )Vt ETi EPdVSi + ESi EPdVTi + ESi ETiVPd .6 is used along with the following equation for the equivalent density (Jones. and Vt is the total volume.Ti.Pd are the moduli of silicon.2. and palladium.Ti. titanium. and palladium.Pd are the volume fractions for each element.Pd are the densities of silicon. When calculating the frequency shift it is necessary to use an equivalent value for both parameters. VSi.Ti. respectively. Gas concentration Since the hydrogen that is to be sensed would be in a gas form it is necessary to understand how to deal with any gas calculations. The two most common ways of 71 .2. 1975.4. titanium. (4. The equation for the equivalent modulus of elasticity is (Jones.Pd are the thicknesses of each of the material layers. Berthelot.7) where ρSi. and hSi. Equation 4. (4. Each layer of the cantilever will have a different material.
i.0001% = 1 ppm . Using Eq. (4.e.2.8) where the concentration of the volume of the analyte gas to the carrier gas is equated.referring to gas concentrations are with percent concentration (%) and with partspermillion (ppm). 4. molecular_weight is the molecular weight of the sample analyte.9) where ppm is the gas concentration in partspermillion. respectively. (4. it is necessary to convert it into a more useful form such as density.. in this case hydrogen. It is important to note though that the partial pressure is of the gas to be detected from a sample of a gaseous mixture. Absorption The main operating principle of the cantilever sensor is its ability to absorb an analyte and thus alter resonant frequency of the sensing element.9 it is possible to estimate the amount of mass that will be absorbed for a given volume of palladium. To determine the maximum amount of hydrogen that can be absorbed into the palladium it is necessary to look at the steady state situation. To use concentration in equations. The relationship between the two forms is 0. 4. and will play a vital role in the results.4. 1 mg ppm ⋅ ( molecular _ weight ) ⋅ 273 ⋅ p = 3 3 m m ⋅T 0.0224 mol . The amount of hydrogen that is absorbed varies with 72 . and p and T are the partial pressure (in atm) and temperature (in °K) of the gas.
reported in literature throughout the years. or absorption of hydrogen. that is of interest. the atomic ratio at 1 ppm had to be approximated from the available information. absorption of hydrogen is nonlinear at lower temperatures. In this case. it is fairly close and the atomic ratio listed for this temperature (20˚C) will be used. the atomic ratio compares the hydrogen volume in palladium. While the figure is accurate. However. This figure shows the relationship between the partial pressure of hydrogen and the atomic ratio (at equilibrium) at different temperatures. Fig. it does not list the temperature.different pressures and temperatures.10) 73 . While this is not the specified temperature.3. 4. so for a concentration of 1% hydrogen the partial pressure will be 0. which for this Thesis equals the small concentrations (<0. ~20ºC. 1967) c=s p . Because of this. There is very little information about the atomic ratio at such small concentrations available. Fig. we must look at the partial pressures of the hydrogen.1% or <1000 ppm) that are of interest. or room temperature. will be around 0. a square root relationship was used to relate the small concentrations with the atomic ratio.01kPa) and the figure indicates that the atomic ratio. Since the hydrogen that is to be investigated will be in a mixture with nitrogen.3 does not have much information about very small pressures. According to experimental results. The square root proportionality comes from Sievert’s Law that states (Lewis.621 at 20˚C. 4. Since the concentration is proportional to the partial pressure. The atomic ratio (H/Pd) is the volumetric. or mass ratio. between two elements in a given volume.01 atm (1. The total pressure for the mixture will be at 1 atm (101kPa). which in turn is proportional to the atomic ratio. (4.
where the concentration. Using Eq. Fig. and p is the partial pressure of the hydrogen. as the uncertainty in the curves in the figure will be zero for isotropic conditions of palladium. is in units of atomic ratio (H/Pd) (Lewis. the atomic ratio was approximated at small partial pressures for equivalent concentrations. 4.3 is assumed as completely correct.where c is the hydrogen concentration in equilibrium in the palladium. 74 . Using information summarized in Table 4. s is the Sievert’s constant parameter. 1967). Table 4. Concentration and pressure relationships based on experimental data of many research studies conducted throughout the years.3.1.10. AR. 4. It is important to note that Fig.1 lists several steady state concentrations at different pressures along with several conversions of the atomic ratio. it is possible to estimate the amount of mass that would accumulate on the cantilever and thus cause a shift in frequency. 4.
D. 1978b). 4.3 at 20ºC with conversion information (Lewis. A graphical representation was developed by Lewis (1967) to illustrate the permeation rates and importance of flow at higher temperatures. This is what is desired as it also simplifies the experiment by not having to set up mass flow controllers.4.06·103 (mgH2/100g) 1 (H/Pd) = ~9. it is necessary to have the diffusion rate.1.2. 1967). Since the temperature at which the experiment should be performed at is around 20ºC and with a static flow of hydrogen.1.5·103 (ccH2/gPd) 4. 4.006 0.667 1% 0. Concentration Pd (H/Pd) 10% 0. 75 .4×107 cm2sec1 at a temperature of 25ºC for diffusion of hydrogen into palladium (Alefeld and Völkl.25·103 (ccH2/ccPd) 1 (H/Pd) = 1. Concentrations of hydrogen in Pd based on experimental data of Fig. 1978a.Table 4. This rate has also been experimentally found for steady state permeation rates and is 5.4. Diffusion rate It is of importance to determine how long it will take for the hydrogen concentration to reach equilibrium in the palladium. For this information.0001% 6.621 0.10% 0.4. the permeation rate will be fairly high according to Fig. 4.21E05 Atomic ratio conversions 1 (H/Pd) = ~1. Fig.
including Sievert’s Law of square root proportionality (Hughes and Bastasz. 1988..11) where c is the concentration and D is the diffusion rate. the equation has a solution of the form 76 . as is the case for this Thesis.. 2001). (4. Permeation rates of hydrogen with circulating or static flows through a membrane of Pd with a high pressure on one side (Lewis. 1967).Fig. 1998.. 2001) ∂c ∂2c =D 2 ∂t ∂x .4. Fick’s second law applies to systems that are not operating in an infinite volume condition. 4. The diffusion rate of a sample concentration can be analytically found using Fick’s second law of diffusion (Streeter et al. If the boundary conditions of the experimental setup are implemented. Hu et al. Hu et al.
− co 2 Dt (4. For the analytical calculations in Sections 4.3. x. 4. Eq. the desired thickness for conducting experiments shall be 1000 Ǻ and 2000 Ǻ.12) where cgas is the concentration of the gas at the surface of the absorbing layer (in this case palladium).. is proportional to the square of the time.5. 1999). co is the initial concentration within the layer. → x = Const.14 shows that the thickness of the polymer layer (palladium) will only affect the length of time required for the layer to reach equilibrium. If the initial concentration of the palladium is zero then Eq. and t is time required for the sample at depth x to reach a concentration of c. 77 . x = Const.13 can be solved to show that the depth variable. As was stated in Section 3. x. which is a Gaussian error function that is tabulated in mathematical tables (Spiegel and Liu. (4. i.cgas − c cgas x = erf . 4. a palladium thickness of 1000 Ǻ will be used. (4.2. The term erf stands for the error function.12 simplifies to x c = cgas 1 − erf 2 Dt .14) Equation 4. ⋅ 2 D t → x ∝ t 2 Dt ( ) .13) where the error function is always a constant and thus the argument within the error function is constant. c is the concentration at a certain depth.e. Therefore.
4%. This increase varies with the concentration.5 (Alefeld and Völkl. In order to account for this volume change. that the slope of the line is fairly steep at 1/5 so that when the concentration is at 1% or the atomic ratio is 0. 4.2. 4. 1978a. but a graph was compiled over the years that displays several experimental data points from many different previous experiments and fits a line accordingly. it should be included in the analytical calculations where possible. 78 .5 were accurate in that there is no uncertainty in the values generated from it.5 it is possible to estimate how much the volume will be affected at any concentration of hydrogen. width. 4. From Fig.889 Å and can increase up to 3. there will be an increase in the lattice structure size. It was assumed that the data of Fig. which can affect the modulus of elasticity.621 H/Pd. It is interesting to note.2. However.4. this change in volume is assumed to affect only the height (thickness) of the cantilever. Fig. 1978b). height) instead of volume and it cannot be simplified. the volume change will be approximately 12. Volume increase Because the hydrogen will be absorbing into the palladium interstitially. This will lead to results that will vary slightly from the true value. There currently are limited data on this lattice increase.893 Å with hydrogen absorption.4. though. Palladium metal has a lattice constant of 3. because the spring constant is dependent on individual geometric values (length.
This decrease is shown in Fig. Change in modulus of elasticity While there are changes in the lattice volume. It has been documented and shown that the modulus of elasticity decreases as hydrogen is absorbed into the palladium (Lewis.3. 4.4.6. 4. 1978b).2. 1978a. 1967).Fig.5. Experimental data fitted with a line to approximate the amount of volumetric change in palladium due to hydrogen absorption (Alefeld and Völkl. 4. the absorption of the hydrogen in palladium will also affect the stiffness (and resonant frequency) because of changes in the modulus of elasticity. which was estimated from experimentally determined data by Krüger 79 .
the modulus of elasticity will decrease by about 10%.621. 4.6. which is the temperature condition for this Thesis. When the atomic ratio (H/Pd) is at around 0. no uncertainty was considered for it. 80 .and Jungnitz (1936) at ambient room temperature (~20ºC). Fig. 4. 4. which is a considerable amount. Once again. Estimated results from experimental data illustrating the change that occurs in the modulus of elasticity during absorption of hydrogen into palladium (Lewis.6 were assumed as accurate and when conducting any calculations.6 it is clear that it will be necessary to include a changing modulus of elasticity in the analytical calculations in order to maintain accuracy. From Fig. 1967). the data from Fig.
can be solved for the fundamental frequency of vibration to obtain (Pryputniewicz. Starting with the general equation for a vibrating body (Rao. Resonant frequency and frequency shift In order to determine the amount of mass collected. or effective.16) The homogeneous differential equation. of the total (i. mass.17) where k is the spring constant for the cantilever and m is its effective or dynamic mass. dt 2 (4.15) where if damping. F(y. As Eq. m.5. represented by Eq. The condition for incorporating this changing stiffness will be described in more detail in Section 4. this change in modulus was only implemented in the analytical calculations when the spring constant was considered a variable.2. c. 4. 4. 1995) we get d 2 y (t ) dy F ( y. 2 dt dt (4. and stiffness. k. The effective mass was calculated for a cantilever and is 33/140.2.e. t ) = m + c + ky . the fundamental frequency must first be determined.. (4.t).24. or about 0. static) mass for the cantilever. any change in mass will 81 . 4. as 0=m d 2 y (t ) + ky . are not included we get a function in terms of dynamic.16.However.17 shows.5. 2002b) f = 1 2π k m . and external forces.
1995).7. 4. 140 (4. which will be compared against computational and experimental data. 4. The dynamic mass written in its explicit form is m= 33 ρ bhL . Fig. The first case will look at the simplest scenario where the only variable that 82 . Representation of cantilever with geometric dimensions and a cross section view through the middle with each of the layers of the cantilever labeled. It should be noted that the calculation of the dynamic mass and all other equations can be found in Appendix B.18) where ρ is the density of the cantilever. Uncertainty calculations will also be conducted on this frequency. which will be described in more detail in Section 4. h.11. while b. Several different cases of the resonant frequency shall be looked at in the calculations. and L are the geometric dimensions of the cantilever.7.2.change the frequency of the cantilever. as shown in Fig. subject to modifications of the spring constant (Furlong and Pryputniewicz. This equation was used for determining frequency of the cantilever for the simplified case. discussed later in this section.
body has a parameter known as the spring constant. In this Thesis. In this Thesis. the equation for calculating the frequency shift due to absorbed mass can be written as f12 − f 22 ∆m = f 22 m . it is possible to calculate the shift in the frequency due to the absorption of hydrogen at any given concentration. or stiffness. Once that is determined. 4.2 are the initial and final frequencies of vibration. 1% hydrogen (in a 1 atm pressure nitrogenhydrogen mix) will be used as the concentration of hydrogen for the calculations. or vibrating. respectively. (4. The spring constant is the value that relates to the elastic energy that can be stored and released in a body for a given distance. f2. (Young and Freedman.19 can be rewritten in terms of the final frequency. 2000) 83 .19) where fi=1. as it increases. If the absorption of hydrogen does not change the spring constant of the palladium. m the dynamic mass of the cantilever.6. k. i. the gas concentration will be 1% hydrogen in a 99% nitrogen mix at a total of 1 atm pressure and 20˚C and the palladium coating will be 1000 Å thick. Equation 4.. It is defined as a ratio of force and the corresponding displacement. in order to determine frequency shift. Spring constant consideration Every oscillating.changes during the absorption process is the mass of palladium.e.2. and ∆m the absorbed mass accumulated causing f1 to change to f2.
2001). 2002. while b. (4. In Eq.21 shows how important the structural dimensions of the cantilever are to the spring constant. the spring constant can be shown to be (Pryputniewicz. they significantly affect the value of k. respectively. and x is the displacement of the spring. however.21) where E is the modulus of elasticity. It is possible to simplify the spring constant into a set of variables that pertain to the characteristics of the cantilever.20) where k is the spring constant. 4. 2002b) k= Ebh3 4 L3 . Also. Jericho and Jericho. and length of the cantilever. This dynamically changing spring constant will be included in the calculations. 84 .e. A varying spring constant is one variable that is usually not taken into consideration in studies dealing with cantilever type MEMS.7.. this equation will have to be modified. it is important to note that as the palladium absorbs the hydrogen.19. thickness. The new equation where the spring constant changes as the mass changes is derived in Appendix B. 2002. 4. Most publications simply calculate the static spring constant of the polymercoated cantilever. Equation 4. Fig. h. i. if the absorbed mass changes the spring constant. (4.k= F x . and L are the width. Gibson et al. However. a change in frequency is determined based on a change in mass. Using the cantilever boundary conditions of the sensing element. F is the force acting on the spring. the spring constant of the cantilever changes (Cherian and Thundat.. since the thickness and length are cubed.
respectively.4. the increase in the palladium volume will be considered to affect the thickness and.2 and 4. a change of the modulus of elasticity.17. 4. the initial frequency f1 can also be expressed as f1 = 1 h 2π L2 35 E 33 ρ .7. stiffness. in turn. Substituting Eqs 4.11.18 and 4.2.2. (4.2. in turn. It is present in all 85 .4. Since it was shown in Sections 4.f2 = ∆k 1 + k 2 f1 ∆m 1 + m . or a combination of both. Damping coefficient consideration Damping is the phenomenon that reduces/absorbs energy of a vibrating system over a period of time. that the modulus of elasticity and the volume change with the absorption of hydrogen it is reasonable to state that there will be changes in the spring constant.22) where ∆k and ∆m are the changes in the spring constant and effective mass. 4.21 into Eq. A change in stiffness will come from either a change of volume. The results of Section 5. 4. it will be by a change of the parameters of Eq. which depend on these parameters. (4.23. respectively.5 will show that this is a reasonable argument. There is damping everywhere in the real world. Stating again.21 that. however.23) If the spring constant value alters. and k and m are the initial spring constant and effective mass. will affect Eq.3. 4.
(4. Only if the cantilever is in vacuum there will be no damping due to the surrounding medium. 2000). v is the velocity of the body. it itself will damp the cantilever.. It is necessary to understand when and how damping will affect a given system.applications with an exception of systems vibrating in vacuum environment.e. The negative sign in Eq.. When doing measurements and calculations.24 indicates that the damping force acts against the body and attempts to reduce the total energy acting upon it. Even though the cantilever sensor is only on the order of 300 µm long. Damping helps bring everything to an equilibrium state and maintain it in a static condition. However.25) 86 .24) where F is the force acting upon the body. 2000) c=− F v . 4. and c is the damping coefficient. damping plays an important part in order to get accurate results. (4. Damping itself is a force dependent process that can be expressed as (Young and Freedman. i. which for a specific set of parameters defining a vibrating body and its environment is a constant that describes the “strength” of the damping force. damping will alter results if it is not noticed and considered (Rast et al. when the analyte gas is released. c= 2µ Lb H . The most common place that damping occurs is in oscillating or vibrating bodies. The damping coefficient can also be written as a function of several parameters defining the damping fluid and the volume that it occupies.
respectively. i. and H is the nominal thickness of the fluid film that the object moves through.25. it demonstrates that damping and the spring constant need to be considered in any calculations for the free vibration system.e. which will be discussed in Section 4. One is that the thickness of the cantilever is not a variable in the relationship defining the damping coefficient. 4.15 defining motion of a free vibration system.4.8. L and b are the length and width of the cantilever.5. 2 dt dt (4..2.26) where there are no external forces. Free vibration system If we look at Eq.8. the damping coefficient will be calculated analytically and then it will be experimentally verified in order to ensure that the frequency shift calculations are being properly represented. There are two items of importance to note when examining Eq. The experimental determination of the damping coefficient will be discussed in Section 4. based on discussion in Section 4. 2000) 87 . 4.2. This will play a role in the quality factor. If this homogeneous equation is solved for an underdamped system characterized by where (Young and Freedman. The other is the obvious relation of how the viscosity of a fluid alters the damping. 4. For this Thesis.4.where µ is the dynamic viscosity of the damping fluid.2. 0=m d 2 y (t ) dy (t ) +c + ky .
19. a relationship for detecting frequency shifts.29 into Eq. While damping affects the frequency.31. (4.30) Inserting Eq. t is time. 88 . and θ is the phase. 4.28) where A is the amplitude of the vibration. the solution will be y (t ) = Ae −( c / 2 m )t cos (ω d t + θ ) .30 and simplifying. 4. but include damping can be showed to be (derivation found in Appendix B) f2 = f 2 1 4 ( m + δ m ) mk − mc 2 4k ( m + δ m ) 2 .29) which clearly shows that damping decreases the angular frequency. (4. similar to Eq. 2 (4. the frequency shift in the presence of damping was calculated analytically. derivation of Eq. (4. Angular frequency is related to cyclic frequency by the following equality: f = ω 2π . 4. The ωd is the damped angular frequency defined by the equation ωd = k c2 − m 4m 2 . 2m with ω being the angular frequency. 4. adding mass affects the damping.31) where f1 is the initial frequency and other parameters are as previously defined. Using Eq. c 2 −ω <1 .27) (4.29 can be found in Appendix B. 4.
4. the excitation force needed to get it into its resonance.9.2. respectively. the equation takes the form of f2 = f 2 1 4 ( m + δ m ) mk − mc 2 4 ( k + δ k )( m + δ m ) 2 . have on the frequency shift calculations. Starting with Eq. (4.9. 4. The steady state solution for the secondorder nonhomogeneous differential equation is the particular solution.4. Using the general equation for a body in motion. 2000) y (t ) = A cos(ω t − θ ) . Changing stiffness in a free vibration system Sections 4.2. both parameters will be considered together in the same equation.2.15. or (Young and Freedman. Eq. Equation 4.2. (4. it will be required to take into account. Harmonically excited system To model the cantilever system completely.29. in addition to the parameters mentioned in Section 4.10.2.6 and 4. it is possible to calculate the maximum amplitude.32) which is very similar to Eq. To further improve the analytical model. 4.32 was used to determine the frequency shift that occurs when both the stiffness and damping changes are considered simultaneously.31. 4.8 described effects that a changing stiffness and damping.33) 89 .
Fig. The damping ratio. in Eq. When this frequency is the same as the resonance. and the critical damping. the frequency ratio will be equal to one and the amplitude will be at its maximum.33 is (Young and Freedman.8. c.35 we see that the driving force will vibrate at the natural frequency and thus the entire body should be vibrating at this frequency. A is the vibration amplitude and has the solution of (Young and Freedman. The amplitude ratio shown in Fig. 4. frequency. The parameter r is defined as the frequency ratio ω r = . ζ = c c = cc 2mωn . 2000) A= k Fmax (1 − r ) + ( 2ζ r ) 2 2 2 (4.where ω is the angular frequency of excitation. 2000) θ = tan −1 2ζ r .8 is equal to Ak/Fmax so that the displacement is unitless. 4. θ. (4. or natural. ωn (4. 4. The critical damping coefficient is the magnitude of damping required to make vibrations decrease 90 .36) Looking at the parameters of Eq.35) where ωn is the natural frequency of the body and the solution for the phase.37) is the relationship between the damping.34) with Fmax being the maximum force applied to the vibrating body. cc. 4. 2 1− r (4. and ζ being the damping ratio. which is exactly what is desired when attempting to detect frequencies.
The equipment (vibrometer and interferometric methods) used in detecting the frequency shift. the system can be modeled with the equation for frequency shift of a free vibration system. 4. 91 . Vibration amplitude as a function of frequency ratio for different values of damping ratio. or in other words. Because the frequency ratio will always be equal to one (once the resonance frequency is reached). will be vibrating the cantilevers over a range of frequencies in order to see the resonant frequency curve peak.8. reduce the amount of energy in the system and also prevent any subsequent amplitudes.the quickest. however. Fig.
the RSS (rootsumsquares) method shall be used (Pryputniewicz. 1993). length. or in an equation that lists all independent parameters of Eq. as δ f %δ f1 = 1 100 . h. f1 (4. For the uncertainty analysis conducted in this Thesis. These uncertainties indicate how “good” the results are and what contribution each parameter has to the specific overall uncertainty. Using Eqs 4.23 as f1 = ( h. δE. δf1. modulus of elasticity. and density. 4. Uncertainty analysis For each case considered in this Thesis. E . ρ. the overall uncertainty in frequency. Based on Eq.39) where δh. E. is 2 2 2 ∂f1 2 ∂f1 ∂f1 ∂f1 δ f1 = δ h + δ L + δ E + δρ ∂E ∂ρ ∂h ∂L 1 2 . but for illustration purposes.4.11. and δρ are the uncertainties in the specified parameters.38) where it can be seen that the independent parameters of the equation for frequency are thickness. (4. The derivation for each case is located in Appendix B. The first part of the RSS method is to write the equation that is going to be analyzed in its phenomenological form. L. ρ ) . %δf1. (4. δL.2. 4. only one case shall be looked at in this section.38 and 4. there was an associated overall uncertainty determined. L.40) 92 .38.39 it is possible to calculate the percentage overall uncertainty in the resonance frequency.
The software Pro/Mechanica. a model of the cantilever was developed. the resonance frequencies of a cantilever were deteremined based on the model developed in Pro/Engineer. two cases were considered. The computational analysis helped visualize the effects of vibration on the cantilever. Results of uncertainty analysis for each case presented in Sections 4.41 can be applied to each parameter to determine the contributions of each of them to the overall uncertainty. For the computational solutions. Using Pro/Mechanica.2.It is also possible to calculate the percent contributions of the uncertainties in the independent parameters to the overall uncertainty by using the following form: ∂f1 ∂h δ h ⋅ 100 .5 through 4. Computational analysis The ACES methodology requires thorough work with computational analysis. Pro/Engineer was used to model the cantilevers using data gathered from measurements of the cantilever (PTC. This is important in order to better understand how reliable the frequency data are and to determine where any sources of errors come from.2. 2003a). a FEM package. The first 93 . For this Thesis. was used for the modal analysis (PTC. 4.41) where %δfδh is the percent contribution of the uncertainty in thickness to the overall uncertainty in the frequency.3. 2003b). Using finite element method (FEM) software. Equation 4.10 are summarized in Appendix C. %δ f1δ h = 2 δ f1 2 (4.
94 . This step was conducted with the cantilevers that were procured from MikroMasch USA as described in Chapter 3.4.6. For the analytical calculations. 6 chips (each with 3 cantilevers) were fully measured using SEM. Once the initial results from Pro/Mechanica were obtained. the solution is said to have converged. The convergence percent is the amount of difference between two most recent solution points.utilized ideal dimensional values. they were checked to assure their validity. For this Thesis. The solution should converge to within at least of 1% to be acceptable. AFM. 4. vibrometer. This check looks at the convergence of the solution. Experimental solutions ACES methodology requires gathering experiment data. The experimental results are compared in Section 5. 6 chips were taken as the sample set. When the difference between the two calculated values is 1% or less. and the second used the averaged values based on measurements of the cantilever. subject to the input information. such as those specified by the manufacturer. Increasing the number of methods used decreases errors in results. and OELIM systems.
it is also important to make sure that each piece of equipment is working properly.2..4. While seeming redundant. Yang et al. Measurements of dimensions of the cantilevers were made using a measurescope equipped microscope capable of providing dimensions with the least count of 0. 1993. This verified quality of fabrication and aided in the analytical calculations and the computational modeling. Each cantilever was measured and all the results were correlated.1. it was important to quantitatively determine their dimensions (Pryputniewicz. With the equivalent length for each cantilever the analytical calculations were performed. 95 . All areas of sources of errors must be reduced to a minimum to ensure quality data. 2000.. Using an optical microscope. it is essential that all of the testing equipment was functioning and calibrated correctly. Instrument calibration To properly conduct experiments. Calibration of instruments is crucial to collecting valid data. 2002b). Pryputniewicz et al. 4.4. Geometrical characterization of the cantilevers To be able to characterize dynamics of the cantilevers used in this Thesis. geometry of the cantilevers was measured.4. Since the free ends of the cantilevers were not square. 1993). it was necessary to calculate an equivalent length that compensated for the irregularly shaped tip based on the measurements that were taken.5 µm (Pryputniewicz.
9. The samples were inserted into the chamber on the left (bottom of the black tube) and the microscope was controlled with the console on the right. SEM used for measuring thickness of the cantilevers. Fig. SEM – scanning electron microscope To inspect and verify the thickness of the cantilevers a scanning electron microscope (SEM) was used. 4.4. An optical microscope can easily view in the xy plane.4. 96 .9. 4.3. but because of the geometric configuration of the chips used the thickness or zaxis of the cantilevers was not easily measurable and thus had to be determined using the SEM. Fig.
is the width of the frequency curve.43) of the peak value of the resonance frequency curve. Once the bandwidth is known. The bandwidth.4. Determination of quality factor and damping coefficient The quality factor (Qvalue. because Q is a function of thickness. where A is the maximum amplitude of the vibration at a specific frequency (Rao. it is possible to calculate what the damping coefficient is using the identity bandwidth = c .4. a method for determination of which was outlined in Section 4. The general relationship for the quality factor can be written as Q= mω n ωn = c bandwidth . the easier and more distinct the peak is to detect. In the opinion of this author. The half power point is defined to be A 2 (4. m (4. appearing in Eq. it is appropriate to discuss a procedure for the determination of Q. (4. at this point in the chapter addressing methodology.4.3. Fig. It describes the “quality” or the width of the resonance frequency peak.44) 97 .42. or Q) is an important parameter of a vibrating body. The quality factor is dependent on the damping and.10. 1995).42) where ωn is the angular frequency. 4. at half power point. The narrower the resonance frequency peak. m is the dynamic mass of body. and c is the damping coefficient. 4. since the damping plays an important role in the calculations.4. it is necessary to measure it.
25. For detecting the resonant frequency curve a laser vibrometer was used. an interferometric setup was used for fullfield of view identification of the resonant frequencies of the cantilever samples. or the nominal thickness of the fluid film that the object moves through.10. 4. 4.4. 4. which in turn will affect Eq.The result can then be compared to the analytically determined damping coefficient in Eq.7. 98 . The value of H is basically the maximum displacement of the cantilever at resonant frequency. 4. as discussed in Section 4.25. Fig. Optoelectronic methodology As mentioned in the beginning of Section 4. A comparison of how different resonant frequency response curves affect the width of the bandwidth.42. 4.4. as used in Eq.4.5. The method described above also allows for determination of the H value.
Optoelectronic methodology.11. In this configuration. Furlong and Pryputniewicz. Images recorded by the CCD are processed by the imageprocessing computer (IP) to determine the fringelocus function. laser light is launched into a single mode optical fiber by means of a microscope objective (MO). 2000). in the OEH system.11. Basic configuration of the OEH system is shown in Fig. defining deformations. Fig. the object and reference beams are combined by the interferometer (IT) and recorded by the system camera (CCD). Ω. via the relationship (Pryputniewicz.Specifically. respectively. 1999.45) where K is the sensitivity vector defined in terms of vectors K1 and K2 identifying directions of illumination and observation. 99 . constant values of which define fringe loci on the surface of object under investigation. The OELIM measured frequency values were then compared to the analytically and computationally calculated values to determine the degree of correlation between them. the single mode fiber is coupled into two fibers by means of a fiber optic directional coupler (DC). is based on the principles of optoelectronic holography (OEH) (Brown. 1999. Then. 4. 4. (4. while the output from the other fiber provides reference against which the signals from the object are recorded. Both. One of the optical fibers comprising the DC is used to illuminate the object. 2002. Pryputniewicz et al. as presented in this paper. The values of Ω relate to the system geometry and the unknown vector L. an optoelectronic laser interferometric microscope (OELIM) was used.. 1995a) Ω = ( K 2 − K1 ) • L = K • L . Furlong.
z).e. respectively (Furlong and Pryputniewicz. while K1 and K2 are the directions of illumination and observation. i. to yield (Pryputniewicz. 1995a) ~ ~ −1 ~ L = K TK K TΩ [ ]( ) . MO is the microscope objective.45 indicates that deformations determined from interferograms are functions of K and Ω. PZT1 and PZT2 are the piezoelectric fiber optic modulators.Fig.11. CCD is the camera.45. 2000). Quantitative determination of structural deformations due to the applied loads can be obtained. ~ Equation 4. (x. IP is the imageprocessing computer. LD is the laser diode. OI is the optical isolator. Singleillumination and singleobservation geometry of a fiber optic based OEH system: LDD is the laser diode driver.y.. DC is the fiber optic directional coupler. distributions over the field of interest on the object being investigated.46) T where K represents the transpose of the matrix of the sensitivity vectors K. 4. which have spatial. Equation 4. 1993) 100 . (4. IT is the interferometer. by solving a system of equations similar to Eq. OL is the objective lens. 4.45 can be represented by a phenomenological equation (Pryputniewicz.
To minimize this influence. K= K(x. (4.z).e.48) where ∂L / ∂K and ∂L / ∂Ω represent partial derivatives of L with respect to K and Ω. (4.y.z).e. Some of these algorithms require multiple recordings of each of the two states.01. the accuracy with which the fringe orders are determined influences the accuracy in the overall determination of deformations (Pryputniewicz. the values of δK/K are less than 0.Ω ) . 1981). It should be remembered that K. while δK and δΩ represent the uncertainties in K and Ω. δL. Eq. L.e. δK 2 δΩ 2 δL ∝ L + Ω K 1/ 2 . in the case 101 . Therefore. i. i. a number of algorithms for determination of Ω were developed. After evaluating. and Ω = Ω(x. can be determined to be 2 2 ∂L ∂L δL = δK + δΩ ∂Ω ∂K 1/ 2 . (4..z). respectively.y.z).47) based on which the RSStype (where RSS represents the square root of the sum of the squares) overall uncertainty in L.y.48 indicates that δL is proportional to the product of the local value of L with the RSS value of the ratios of the uncertainties in K and Ω to their corresponding local values. when performing partial differentiations. i. for small deformations. respectively.. respectively. 4.L = L(K .49) For typical geometries of the OEH systems used in recording of interferograms. L = L(x. and Ω are functions of spatial coordinates (x..y. the typical values of δΩ/Ω are about one order of magnitude greater than the values for δK/K. However.
1999. 4.46 to find L. and Ω is the fringelocus function relating to the deformations the object incurred between the first and the second exposures.52) 102 . in the doubleexposure sequence can be represented by the following equations: I n ( x.51) .of doubleexposure method. respectively. y ) + I r ( x . y ) I r ( x. y ) = I o ( x. y ) + θ n + Ω ( x. y ) cos ϕ o ( x. 3π /2. y ) cos ϕ o ( x. where Io and Ir denote the object and reference beam irradiances. In the case of 5phasesteps algorithm with θn = 0. y ) (4. with (x.50) ' I n ( x . φo denotes random phase of the light reflected from the object. y ) { { } } 1/ 2 • (4. Ω is what we need to determine. y ) + 2 I o ( x. φr denotes the phase of the reference beam. y ) + θ n { { } } 1/ 2 • (4. the intensity patterns of the first and the second exposures. Pryputniewicz. y ) − I 5 ( x. y ) I r ( x. y ) = I o ( x. y ) − ϕ r ( x. π.y) denoting spatial coordinates. π/2. 2000.y). y ) + 2 I o ( x. When Ω is known. y ) + I r ( x. of the object being investigated with introduction of a discrete phase step between the recordings (Furlong. y ) − ϕ r ( x. Furlong and Pryputniewicz. y ) − I1 ( x.y) and I΄n(x. respectively. 2 I 3 ( x. θn denotes the applied nth phase step. and 2π the distribution of the values of Ω can be determined using 2 I 2 ( x. In(x. For example. y ) = tan −1 . 1995b). it is used in Eq. y ) Ω ( x. y ) − I 4 ( x.
. In the OELIM system. an optoelectronic laser interferometric microscope (OELIM) methodology for studies of objects with micron size features was developed (Brown. and processing subsystems of the OEH system. The two beams recombine at the beam splitter and are imaged onto the sensing element of the CCD camera. 1999. 4. 2000. issues relating to the sensitivity.12. were studied while evaluating the effects that the use of highspatial and highdigital resolution cameras would have on the results produced (Furlong et al. which records the resulting interference patterns. 2002). The object beam is directed towards the PZT actuated MEMS under study by a long working distance microscope objective and is reflected back to the beam splitter. 4. These patterns are transferred to the system computer for subsequent quantitative processing. Pryputniewicz. The resulting light field is then divided into reference and object beams by the splitter. associated with application of the algorithm defined by Eq. Developments in laser. imaging. Furlong and Pryputniewicz. and precision. 1998. Brown and Pryputniewicz. CCD camera. Fig.12. 4. Using the systems shown in Fig. 1991). 1985b. and computer technologies have led to advances in the OEH metrology. The reference beam is directed towards a mirror and back to the beam splitter. the light beam produced by a coherent light source is directed through a collimating optics. fiber optic. Mode shape characteristics of vibrating MEMS cantilevers were determined employing previously 103 .51. 4.52 depend on the capabilities of the illumination. In response to the needs of the emerging MEMS technology.Results produced by Eq. accuracy. 2000. 1989.
1990. resonance frequencies of vibrating cantilevers were also determined using an atomic force microscope (AFM) in order to have two independent means of quantifying resonance characteristics of the MEMS cantilevers used in this Thesis.12. Oh and Pryputniewicz. 4. 1990.4.13.developed procedures (Brown and Pryputniewicz. is capable of automatically measuring the resonance frequencies of a cantilever by scanning a large frequency range 104 . Pryputniewicz. The AFM. 1992. shown in Fig. 4. 2000). Optical configuration of the OELIM system (Furlong and Pryputniewicz. 1995. 1991). 4.6. Determination of resonant frequency by AFM In addition to using the OELIM methodology. 1988. Fig.
It provides quick and accurate results. a resonant frequency is located there. In an AFM system. 4. a laser beam is focused and reflected off a cantilever and into a position sensitive diode (PSD) as illustrated in Fig.14. (a) (b) Fig. This motion is measured and used to determine a displacement curve. The vibrometer functions by way of using a laser and a 105 . it is necessary to know approximately where the frequencies of intent are to avoid detection of harmonics. the angle of the deflection changes and position of the laser beam on the PSD moves. to facilitate determination of frequencies by AFM. As the cantilever vibrates.and detecting where the largest amplitudes occur. (b) close up of the measurement subsystem with protective lid open. laser vibrometry was used to measure damping coefficients of the vibrating MEMS cantilevers. 4. 2. Wherever there is a peak in the curve.7.4.13. However. Laser vibrometry In this Thesis. AFM setup utilized: (a) overall view of the setup showing measurement and control subsystems.
MachZehnder interferometric setup built within it as well as its capability to measure Doppler frequency shifts (Polytec. 106 .14. 4. Once those two parameters are measured. it is possible to calculate frequency of the vibrating object with an accuracy of 0. MEMS chip Fig.14.001 Hz. A comparison of the PZT used with a MEMS chip attached to the top and a dime. This vibrometer would be used to measure small shifts in frequency of a vibrating cantilever in the MEMS chemical sensor. which is capable of excitation frequencies in the MHz range containing the frequencies of intent in this Thesis. It functions by examining a signal from an object on a pointbypoint basis to measure velocity and displacement as a function of position on the object. 2003). Fig. The cantilevers were placed on a custom made PZT (PbZnTi) transducer. 4.
would purge the chamber. The design of the chamber can be found in Appendix D. Any changes would be recorded and then the gas would be released and the neutral carrier gas. The supply gas tanks would not be required to be within the hood as they are sealed systems in themselves.4. 4. This chamber would have the gas sample flow into it and out via inputs and outputs.15. Figure 4. Test chamber The PZT transducer would be contained within a custom made test chamber. The aligning microscope would be equipped with a CCD camera so that the cantilevers can be viewed 107 . Fig. The entire chamber along with an aligning microscope would be contained within a hood of the vented space. The chamber will allow the cantilevers to be in a gaseous environment that can be controlled. Because of space limitations. the entire setup would have to be as compact as possible. To regulate concentration of gas. nitrogen.4. 4. For simplification of the setup.15. but would be pumped into one atmosphere pressure and then sealed. This would ensure that any gas that escapes or is purged would be safely carried away. A glass window on the top lid would allow the vibrometer to make unobstructed measurements of frequencies of the cantilevers. the gas would not flow through the chamber.8. but can be.16 illustrates the setup that would allow for the recording of shifts in the frequency of vibrating cantilevers in a gaseous mixture containing H2. any effects of reduced partial pressure of hydrogen due to absorption into the palladium will be assumed negligible. Fig. mass flow controllers would be used to adjust the gas mixture ratio. Because of the volume within the chamber compared to the amount of palladium present.
Laboratory setup for testing functional operation of MEMS chemical sensors. Because of the remote measurement capability of the vibrometer. 4. the microscope and test chamber will be contained within a vented hood.15. 108 . which is to be used for testing functional operation of the cantileversensing element in a gaseous environment. Fig. Fig. 4.16.outside of the hood. The custom made test chamber containing the PZT transducer. it will be possible to place it outside of the hood.
and are presented in Section 5. For analytical calculations dealing with the coated cantilevers.11. several correlations between different types of data were made.5.2. as is discussed in Section 5.6. The first calculations were made of the analytical frequencies for each cantilever based on the measured dimensional data. In the analysis. Two sets of computational results were then compared to the analytically equivalent cases. Each case described in the Sections 4. several results were compared. they were analyzed in order to determine a set of parameters that would optimize performance characteristics of the MEMS cantilevers. The percentage difference was calculated and correlation was determined. Once all the analytical frequencies for each cantilever were calculated.2.5 through 4. All the percent differences were then averaged and compared to the overall percentage uncertainty for the analytical case.1. These results were tabulated. as is discussed in Section 5.2. Each analytically determined cantilever frequency was compared to the AFM and interferometric data individually. averaged.5. 109 .10 discussing the different approaches to the problem was analytically calculated for the longest cantilever. The averaged data was then used to calculate the contributions that uncertainties in each independent parameter have on the overall percentage uncertainty of the dependent parameter. as discussed in Section 4. it was possible to correlate these values with the experimentally determined results. Data analysis Once all of the results of the ACES methodology were collected.4.
10.11.7. The results were analyzed and conclusions were drawn as discussed in Sections 5. comments and recommendations were written for further investigations into cantilever sensors as discussed in Chapter 6. 110 .The overall percentage uncertainty was calculated for each case and compared in Table 5. Once all the data were analyzed and conclusions drawn.6 and 5.11.
5. E. For the purposes of this Thesis.5 µm. per chip were measured for each of the six chips considered. The MicroMasch cantilevers with tips were fabricated with the ends as shown to facilitate their use for the AFM applications.2 and also in Figs 5. and F.2 to 5. where it can be seen that the free ends are not square. A Nikon CoolPix 5000 with a 3. The physical dimensions of the cantilevers were measured using an optical microscope and a SEM.e. 5. Microscope and SEM data In the beginning of the analysis of the cantilevers. 3. i. All original measurements for cantilever D are listed in Table 5. D. It was discovered that the reason for this was due to the way the cantilevers were made.5. Representative pictures of the cantilevers are shown in Fig. computation. which has the least count of 0.. and experimental results generated in this Thesis are presented in this chapter.2 megapixel sensor was connected to the microscope for taking pictures of each cantilever. During the fabrication of tipless versions the same masks were used in the process. quantitative measurements were made describing all of the geometric dimensions. each with 3 cantilevers apiece.1. Based on the measurements. 5. The sample size was 6 chips. the shaped areas of 111 . so the ends are shaped. three of the longest cantilevers. Fig. The microscope used was a Nikon Measurescope MM11.1.1. RESULTS All of the analytical. for cost reasons. The results are shown herein unless it is specified that they are located in a specific appendix.
5.2. The Nikon Measurescope MM11 is on the left.the cantilever tips were calculated and an equivalent length was found for a corresponding “square” tip. including its minimum and maximum values. 5. the 1% difference does not affect the performance for 112 . Fig. which is calibrated. A cantilever tip can be seen on the monitor in the middle while the monitor on the right displays the measured values. are shown in Table 5.1. The microscope setup used for determining dimensions of the cantilevers. Fig. This simplified the calculations. based on the original measurements listed in Table 5. The average equivalent total length of the cantilever D. It was found that the geometric dimensions of the cantilevers were generally about 1% different (which for length amounted to about 2 to 3 µm) from the manufacturer’s specifications. which also shows other pertinent dimensions for the cantilever D as well as for the cantilevers E and F.1.2. Because the cantilevers were made for atomic force microscope.
This designation consists of a letter and two digits. (a) (b) (c) Fig.2 include the quantitative values measured and the equivalent calculated values for the corresponding “square” tip cantilever. A complete list of the dimensions of each cantilever can be found in Appendix E. from A through F.1 with a schematic guide shown in Fig. The letter. designates the cantilever used on each chip.their originally intended use.1 and 5. 5.2. The geometric dimensions are summarized in Table 5. b) closeup on a tip of a cantilever. Cantilever details and labelling: a) photograph of the entire cantilever. However. It should be noted that the sample cantilevers all had a specific designation number according to the chip that they belonged too. differentiable by length with A designating the shortest and F designating the longest 113 .2. c) guide for the labels used for each geometric parameter of the cantilever describing the data in Tables 5. 5. the differences measured exceed the current industry practice of 1 µm tolerance on length. The geometric parameters in Fig. 5.2.
2.3. The thickness data collected were averaged and are displayed in Table 5. This 5% was considered the uncertainty analysis. a scanning electron microscope (SEM) was used. 5.3. only the thickness of the longest cantilever was determined assuming that the thickness will be the same between them all. and F) are each on the same chip. 5.5. The SEM was specifically used to measure the thickness of each cantilever because of its ability to see an object in “threedimensions” as in shown in Fig. 5. Since all the cantilevers (D. Fig. The first digit designates the row number and the second number designated the column starting with ‘11’ at the upper left corner of the package in which the cantilevers were shipped and stored in. 5. Fig.cantilever. The values of the thickness from the SEM that were measured all tended to be approximately 5% off from the manufacturer’s specifications of 1 µm. In addition to the traditional optical microscope.4. A few representative images of different samples of cantilevers are shown in Fig. Cantilevers packed in a GelPack as obtained from the manufacturer. 114 . E.
938 305.5 36 1 14 8. (a) (b) (c) Fig.573 303. total length.007 15 9.4. Measured dimensional data and equivalent lengths for all cantilever samples of the D type.70 D31 Typical 295 35 0. length.5 1. b mm Thickness.57 (a) (b) (c) Fig. h mm End width. 115 . Measured characteristics Length. L1 mm Eqv.1.Table 5. c) chip E32.989 15 7. L2 mm Eqv.757 302.11 D21 Typical 293.118 301.76 D32 Typical 295 36 0.96 15 7. with a progressive close up on the tip: a) an overall view of three of the longer cantilevers (DF). b) closeup of cantilever E.5 1 15 8.7 303. SEM photographs of E length cantilevers near the tip from different chips: a) chip E11. c) closeup of the tip of cantilever E showing the thickness and details of the tip. SEM photographs of the cantilevers. b) chip E22.44 D12 Typical 295 36. L mm Width.956 15 8.5 36.5. 5.62 D22 Typical 296 35 0.106 304. 5. mm D11 Typical 296.
700 8. The idealized frequency range that the uncoated cantilevers operate in is shown in Fig.00 9. 116 .5 245.00 9. 4.83 356.5 345.597 249. Summary of measured dimensions for the cantilevers of the D.2. total length.20 355.985 1. the frequency for each was calculated for the ideal case where they are uncoated.08 11.5 301.5 254.3.5 305. E.007 8.5 295.1 the results for the remaining cantilevers were determined and are summarized in Section 5.53 305. L µm Width.37 9. Measured parameters Length.2.00 346 34 35.882 352 354. ( ( ) ) (5. we find that the analytical nominal frequency is f1 = 1 h 2π L2 35 E 33 ρ 9 35 190 ×10 Pa 33 2.00 248 33 35.00 10.83 36. L2 µm Ave. and F types.007 9.00 6. tot. tength. length.5 0.25 37 0.17 296.83 11.50 7.23 and measured dimensions of the cantilevers.58 11.956 0. For example.6. 5.007 9.99 9.865 Hz .20 9.00 36 0.236 8.5 350.25 307.1 5.7 F cantilevers Min Average Max 239.985 1.6 303. 5. taking the measurements gathered from the microscope and SEM for the cantilever D11 and values of modulus of elasticity and density from Table 5.58 258 247.956 0.99 256.1) Following the procedure used to evaluate Eq.985 1. µm Ave. L1 µm Eqv.721 7.956 0.33g / cm3 (1µ m ) 1 = 2π ( 305.8 252.106 303. untreated.Table 5.5 35 35. µm D cantilevers Min Average Max 293. eq.6 353.6. and assumed composed of isotropic singlecrystal silicon.00 7. Analytical results Using Eq.44 µ m )2 = 15. h µm End length. b µm Thickness.4 E cantilevers Min Average Max 342.
Values of material properties used Modulus of elasticity.3. The percent overall uncertainty was then averaged from these three values. 1993). length.05 µm δ L =5 µm δ E =3 GPa δρ =0.33 To determine the uncertainty in the frequency results of the simplified case (uncoated and untreated) the RSS method was used. To determine which parameters (height. The percent overall uncertainty in frequency was plotted in terms of length and thickness of the cantilevers and is shown in Figs 5.4. E .2 were used in determining the percent overall uncertainty in the frequency of the cantilevers utilizing Eqs 4.40. subject to the uncertainty value for each parameter used is listed in Table 5. g/cm 3 2. The uncertainties in each parameter considered. These results can be found in Appendix E. modulus of elasticity.4. Table 5. (Pryputniewicz. Parametric uncertainty values 3 δ h =0.8.39 and 4.39 and 4. The percent overall uncertainties were determined based on values shown in. ρ.Table 5.40.4.005 g/cm The averaged geometric data from Table 5. This percent overall uncertainty can now be compared to the computational and experimental data. Eqs 4. Values of material properties of silicon (Madou. Table 5.2 the percent overall uncertainty in resonance frequency of each of the three cantilevers of different lengths was found. Using the averaged equivalent values from Table 5. 2002). GPa 190 Density. and/or 117 .7 and 5.
each one was plotted as a function of cantilever length and thickness to determine their percent contribution to the percent overall uncertainty in the frequency.density) contributed to this uncertainty. If the plot of the percentage overall uncertainty in frequency with respect to the uncertainty in thickness is generated. one can clearly see that in both figures.9% contribution in the 350 µm long cantilever as an example. 5. Figs 5. the extent that δh = ± 0. Fundamental resonant frequency of a 1µm thick cantilever as a function of active length.6.5 µm makes can be seen. the uncertainty in the thickness has the highest contribution to the percent overall uncertainty in the frequency with 74.11. 118 . Fig. Fig.9 and 5. This is an important statement in that the thickness is indeed the one parameter that has the highest relative uncertainty value of all of the independent parameters in the frequency equation. 5.10 and from them.
Fig. The percentage overall uncertainty in the cantilever as a function of against the length while holding the thickness parameter at 1 µm to show the decreasing trend.8.7. The percentage overall uncertainty in the cantilever as a function of the thickness while holding the length parameter as 350 µm to show the decreasing trend. 5.Fig. 119 . 5.
5. 120 .9. Fig. as a function of thickness. as a function of length. Percent contributions by uncertainty of each independent parameter to the overall uncertainty in resonance frequency of the cantilever.Fig. Percent contributions by uncertainty by each independent parameter to the overall uncertainty in resonance frequency of the cantilever. 5.10.
Fig. 5.11. Overall uncertainty in the frequency of the cantilever as a function of the uncertainty in the thickness, for different lengths of the microcantilevers.
5.3. Optoelectronic measurements
The optoelectronic methodology described in Section 4.4.5, was used to produce images of the first few bending modes of the cantilevers (up to the 5th bending was found), Fig. 5.12. The first bending mode was very important, however, for it corresponded to the natural resonant frequency that was analytically calculated. With the setup used, it was possible to determine this resonant frequency with the accuracy of 10 Hz or better than 0.1% of the experimental determined frequencies as shown in Section 5.6. To determine the resonant frequencies of each cantilever, a range of frequencies from 1 Hz to 1 MHz was scanned through and the resonances were determined. The first
121
bending resonant frequency is shown in Fig. 5.13 for each cantilever (D, E, and F) on chip 22. The results from the OELIM measurements are summarized in Section 5.6.
(a)
(b)
(c)
(d)
(e)
Fig. 5.12. OELIM images of cantilever E32 vibrating under resonant conditions: a) the 1st bending mode (resonant frequency) of the F length cantilever where the fringes indicate the magnitude of deformation and the mode shape at 12, 650 Hz, b) 2nd bending mode at 81,400 Hz, c) 3rd bending mode with torsional bending at 115,890 Hz, d) 4th bending mode at 456,150 Hz, e) 5th bending mode at 754,640 Hz.
(a)
(b)
(c)
Fig. 5.13. OELIM images of the cantilevers (D, E, and F) of chip 22 vibrating at their fundamental mode of vibration: a) the D length cantilever, resonating at 14,356 Hz, b) the E length cantilever, resonating at 10,602 Hz, c) the F length cantilever, resonating at 20,716 Hz
122
5.4. AFM results
The cantilever chips were loaded on to a chip carrier and then into the AFM. Because the chips were originally designed for a use with an AFM, they were easy to set up and get results. The AFM then scanned the frequency ranges and the resonant frequency could be detected with an accuracy of 10 Hz. Figure 5.14 shows an example of the AFM frequency scan with the resonant peak clearly defined and marked for the cantilever F12. This was how the values were found for each cantilever. The results of AFM measurements are summarized in Section 5.6.
Fig. 5.14. A screen capture of the AFM frequency scan for the cantilever F12, where a resonant frequency peak is identified as 22,160 Hz with the accuracy of 10 Hz.
123
5.5. Computational results
For the computational analysis, Pro/Engineer was used to build the model and Pro/Mechanica was used for the modal analysis. The model used is shown in Fig. 5.15 where all three cantilevers were analyzed at the same time. The model shown in Fig. 5.15 faithfully represents the chip used in this Thesis. For the software to generate accurate and valid results it is necessary for the solution to converge. Current practice is to have the convergence percentage at 0.01%, but for the software used here, the lowest value attainable was 1% (PTCb, 2003), based on “fixed solution. Pro/Mechanica uses ptype (polynomial) elements for converging on the solution quickly and more accurately than with other available methods. The ptype elements utilize higher order polynomials to obtain the solution instead of increasing the number of elements as in ntype element software packages. Thus ptype elements allows for quicker, discreet solutions that are attainable with just a few steps in the calculation. The results of the computational simulation were compared to the analytically determined values to determine degree of correlation. Ideally, the computational results would be compared to the experimental results. However, to be more efficient, since each cantilever is of different dimensions and would require a new computational solution, the computational results were compared to the analytical results for two different conditions for correlation. This way the analytical results could be directly compared with the experimental results with confidence that the computational results would be similar to the analytical results. The first case used ideal structural dimension
124
the frequency difference is only about 0. freq.2) These values are displayed in Table 5.985 21234 21315 0.2 35 0. 15314 − 15326 = ⋅100 15326 = 0.25 0. For example. was calculated as FEM freq. % diff = ⋅100 Analytical freq. L µm Width. freq.7341 D Typical 303. % diff. Table 5.078 % . For the ideal case.8259 F Typical 250 35 1 22069 21907 0. Comparison of the analytical and computational resonance frequency results.985 10894 10870 0. reference Table 5. i. % diff. freq. b µm Thickness. using values from Table 3. D Typical 300 35 1 15326 15314 0.e.47%.83 0. Based on nomicnal dimensions from the manufacturers Based on measured dimensions. − Analytical freq.0783 E Typical 350 35 1 11260 11167 0. h µm Analytical freq.54% and for the averaged case it is about 0.5 35.2 Measured parameters Length.5.8 E Typical 353.381 125 .22 F Typical 253 35. (5.5.5 along with the analytically calculated values.values (as those specified by the manufacturer in Table 3. Resn. Resn. Hz FEM freq. Hz Resn.985 14751 14633 0.. the percent difference between the analytical and FEM results. the computational results correlate well with the analytically determined values.1) for the cantilevers while the second used the averaged values as listed in Table 5.1... As can be seen in Table 5.2.
(a)
(b)
(c)
Fig. 5.15. Pro/Engineer model and Pro/Mechanica results of the three cantilevers considered: a) cantilever D vibrating at its resonant frequency of 14,679 Hz, b) cantilever E vibrating at its resonant frequency of 10,819 Hz, c) cantilever F vibrating at its resonant frequency of 21,270 Hz.
126
5.6. Comparison of results
In this section, all the results obtained in this Thesis are compared for determining quantitative correlation. The data from the OELIM and AFM are to be correlated to the simplified analytically determined percent overall uncertainty in frequency, as discussed in Section 5.2. To do this, the difference between the determined AFM frequencies of each cantilever is compared to the analytically determined frequency. This was also done for the OELIM determined frequencies and then percent differences were calculated for both methods. This difference can be seen as the uncertainty in the OELIM and AFM data if the analytical results are considered correct. These differences were then averaged so that they could be directly compared to the analytically determined uncertainties. For example, the percent difference between the analytical and AFM results for cantilever D11, i.e., Anal./AFM freq. % diff, was calculated as AFM freq. − Analytical freq. Anal./AFM freq. % diff = ⋅100 Analytical freq. 15940 − 15865 = ⋅100 15865 = 0.47 % .
(5.3)
Relationships similar to Eq. 5.3 were used to determine Anal./OELIM freq. % diff for the remaining cases and the results are summarized in Table 5.6.
127
Table 5.6. Data comparison of analytical, OELIM, and AFM generated frequencies for each cantilever.
Analytical freq., Hz AFM resn. freq., Hz OELIM freq., Hz Anal./AFM freq. % diff. Anal./OELIM freq. % diff. Analytical freq., Hz AFM resn. freq., Hz OELIM freq., Hz Anal./AFM freq. % diff. Anal./OELIM freq. % diff. Analytical freq., Hz AFM resn. freq., Hz OELIM freq., Hz Anal./AFM freq. % diff. Anal./OELIM freq. % diff. Ave. % diff. of D cantil. Ave. % diff. of E cantil. Ave. % diff. of F cantil. Overall ave. % uncert.
D11 15865.4 15940 15700 0.47014 1.04259 D21 16269.8 15849 15750 2.58613 0.62465 D31 15969.9 13230 19630 17.1569 22.9184
E11 11812.3 11580 11440 1.96619 3.1514 E21 12044.9 10690 11070 11.2484 9.11931 E31 11731.8 11560 11560 1.46404 1.46404
F11 22772.1 22160 22100 2.68777 2.95125 F21 24105.4 22450 22380 6.86728 7.15767 F31 22919.7 19630 25650 14.353 11.9127
D12 E12 F12 16116.7 11801.4 23175.9 16010 11850 22160 15940 broke 22040 0.6618 0.41153 4.38341 1.09613 4.90119 D22 15405.6 15670 15420 1.71648 0.09369 D32 15354.2 16990 17160 10.6536 11.7608 E22 11377 11700 11490 2.83881 0.99298 E32 11378.5 12720 12650 11.7901 11.1749 F22 22230.6 22580 22590 1.57183 1.61682 F32 21576.3 24280 24360 12.5311 12.9019
Uncertainty 6.102 5.865 6.482 6.150
AFM 5.541 4.953 7.066 5.853
OELIM 6.256 5.181 6.907 6.114
As can be seen from Table 5.6, the values for the OELIM and AFM methods where compared to the analytically calculated values. The overall uncertainties, listed under ‘Uncertainty’ in the table, were calculated in Section 5.2. The optoelectronic measurements, listed as ‘OELIM’ and the AFM, listed as ‘AFM’ in the table, data results were discussed in Sections 5.3 and 5.4, respectively. One final average was calculated 128
for the overall average percent difference between all the cantilevers for each method. From the collected data, it is clear that the percent differences in the experimental results (AFM=5.85% and OELIM=6.11%) were similar to the analytically calculated uncertainty (Uncertainty=6.15%) results. This correlation of the data validates the parameters contributing to the overall uncertainty as analytically calculated in Section 5.2.
5.7. Parameters of the coated cantilever
Because of the idea to coat the cantilevers with palladium for hydrogen detection at room temperature and 1 atm pressure it was necessary to prove the functionality of this theory. To accurately model this Pd/H2 system, it was necessary to include all the parameters that define the coated cantilevers like the different moduli of elasticity and densities, which would be present. These moduli and densities are listed in Table 5.7. Using Eqs 4.6 and 4.7 we find that the equivalent moduli of elasticity and density for a 350 µm long cantilever coated with 1000Ǻ of Pd and 10Ǻ of Ti are Eeq = =
( ESi ⋅ ETi ⋅ EPd )Vt
ETi EPdVSi + ESi EPdVTi + ESi ETiVPd
116 ⋅121 12, 250 µ m3 + 190 ⋅121 25.24 µ m3 + 190 ⋅116 2538.52 µ m3
(
(190 ⋅116 ⋅121) (14,813.76 µ m3 )
)
(
)
(
)
(5.4)
= 172.9 GPa , ρ V + ρTiVTi + ρ Pd VPd ρeq = Si Si VSi + VTi + VPd
=
2.33 12, 250 µ m3 + 4.507 25.24 µ m3 + 12.023 2538.52 µ m3 12, 250 µ m + 25.24 µ m + 2538.52 µ m
3 3 3
(
)
(
)
(
)
(5.5)
= 3.995 g / cm3 .
129
2 and 4. Sensitivity solutions In order to determine the correct coverage of the cantilever with palladium the sensitivity was calculated for two cases: a completely covered cantilever and a partially covered. It was found that when 25% of the cantilever is covered with 0.16 the distribution. Ti Equivalent values 5. its effect on the resonance frequency of the cantilever is reduced. of the mass on the cantilever is related to how much the effective mass value is affected. the effective mass will increase exponentially.9 Silicon.7. Density 3 ρ ± 0. or percent coverage. Using Eqs 4.005 g/cm 2.e.5 it was found that the “sectional” or partial mass added to the cantilever at the tip affects the overall mass more than if it were added at the base. It was first necessary to determine the amount of coverage to use in a partially coated cantilever case. As a uniform amount of mass is added to the cantilever.8. Properties of materials used. starting from the base and depositing towards the tip.Table 5.507 3. As a coating of mass is added in the opposite direction. cantilever. As the partial mass that is loaded is moved closer to the base. In Fig.1 µm palladium at the tip it accounts for about 72% of the effective mass of the complete cantilever. 5. Pd Titanium.995 Modulus of elasticity E ± 3 GPa 190 121 116 172. i. Si Palladium.330 12. tip only. starting from the tip and depositing 130 .023 4.
towards the base. the solid line illustrates the same contribution as the position moves from the tip to the base as mass is added. Since the greatest contribution to the effective mass due to addition of mass occurs in the end/tip 25% (75% through 100% of the cantilever position). Effective mass versus percentage of coating coverage: the dotted line illustrates the effective mass contribution as a function of the position along the cantilever moving from the base to the tip as mass is added. the effective mass slope decreases in an opposite manner to the previous loading method. will affect the effective mass. Fig. Both cases illustrate how the position of where the mass is added. 5.16. it was decided 131 . as it gets closer to the tip of the cantilever.
854 ⋅10 g + 420 Hz .039 2 ⋅ ⋅ 0.25 ⋅10−11 g 2.034 2 s 2 33 33 71 ⋅ 2.854 ⋅10−8 g + ⋅ 0. = −3.25 ⋅10−11 g 140 420 140 1 2 kg 0.that this was a reasonable area to use for calculations for the amount of coverage for a partial covered cantilever.854 ⋅10−8 g ⋅ 2.039 N/m.854x108 g. as the differentiation of the same initial equation will produce different results due to the different formulation of the derivative (see Appendix B for the derivations).5 the sensitivity was calculated for the partially and completely covered cantilever (see Appendix B. a mass of 2.141⋅109 Hz − kg while the sensitivity for a completely covered cantilever is 132 . This case will be slightly different from the partially coated case.25.4).25 s 420 kg 0. The second method of calculating the sensitivity was assuming the entire cantilever was uniformly coated. the sensitivity for a partial coated cantilever is Se = −1 k 2 m k ( m + nr r ∆m ) ( m + nr r ∆m ) knr r 2 = kg 71 − 0. an effective mass constant of nr = 71/420 for a coverage of r = 0. ∆m = 1011 g of mass to “add”. Using the 350 µm long cantilever with a stiffness of 0.6) 71 −8 ⋅ 0.2 and 4. Using Eqs 4.039 2 s • (5.
2001).141E9 Hz/Hzkg 5. the absorption reaction must be well known.9.. Table 5.751⋅1010 . This coincides with the conclusions of other researchers in the area of chemical sensors (Betts et al.8.039 2 s s 2 33 33 2.854 ⋅10−8 g + 10−11 g 140 Hz = −1. This reaction can be described with 133 ..7 for detecting the largest overall shift in resonant frequency at a steady state condition a fully coated cantilever was preferred. Absorption The main ability in the hydrogen sensor is for palladium to absorb H2. Table 5.854 ⋅10−8 g + 10−11 g 140 140 1 2 33 2. Sensitivity at 1% H2 Distributed load 1.6 and 5. Frequency shift per Hertzkilogram for distributed and tip loads.039 2 0. 2000.039 2 s = • kg kg 0.8. Because of this. Hz − kg (5.Sd = 2 −1 k 33 M 140 k 33 ( M + ∆m ) 140 k 33 2 ( M + ∆m ) 140 kg − 0.854 ⋅10−8 g 2.7) ( ) ( ) As can be seen from the results of Eqs 5. Lange et al.751E10 Hz/Hzkg Tip loaded 3.
or the volumetric amount of one element compared to the other at the atomic scale. and most data currently available. This absorption is directly affected by the pressure and temperature of the gas. or approximately sea level pressure. and with a concentrations equal and less than 1%. These changes can be seen in Fig. simulating Fig. the amount absorbed. the partial pressure will only be <0. 1967). It is important to note that the experimental data available for determining the atomic ratio of the absorption of hydrogen in palladium at low concentrations are limited. another formula (located in Appendix C.2. For this reason. This figure. the equations used had 1% hydrogen concentration as the upper limit. which seems to show an increase in the atomic ratio as the temperature decreases. Since measuring the lower concentrations (<1%) is what is of interest. This is an important characteristic because of the conditions of this Thesis to have the sensor work at 1 atm.two main points: the concentration that is absorbed and the amount of time required for that absorption. the atomic ratio absorption at lower concentrations (<1%) is approximated using a square root proportionality decrease from Sievert’s law (Lewis. 1967).6. 4. 4.3. In addition to assuming Sievert’s law for absorption amount. about hydrogen absorption into palladium are from experimental data collected throughout the years (Lewis. is measured by the atomic ratio. Equation C.80) was generated. The first part.80 was 134 . to estimate the change in modulus of palladium with hydrogen concentration being the variable.2 as Eq. This is different than with the temperature. C.01 atm. As the partial pressure of hydrogen decreases the absorption decreases.
17. as it will affect the amount of absorption greatly and thus the results. which was prepared for a cantilever with a 0.12. 5. using Eq.1 µm coating of palladium in a hydrogen concentration of 1%.17. Fig. From Fig. and Pd_volume is the palladium coating layer volume. 4. Using Fick’s second law of diffusion in conjunction with Sievert’s law. H2_density is determined with Eq. 5.1.9. This is very useful and enabling for a sensor as speed is important in detecting agents.integrated into the equivalent modulus of elasticity relationship so that it could compensate for the change in the modulus of elasticity of the palladium during absorption. it was possible to estimate the length of time required to reach any particular percentage amount of equilibrium concentration in the palladium. It is important to note though the importance of the partial pressure of the hydrogen. which although changes with the absorption process it does not increase in the overall volume mass (number of palladium molecules). The other important parameter in the absorption process is the length of time required to reach certain concentration of hydrogen within the palladium. (5. The final equation then that represented the equivalent amount of mass of hydrogen accumulated on the cantilever by absorption at different concentrations was ∆m = ( absorption _ multiplier )( H 2 _ density )( Pd _ volume ) . Fig.17.8) where absorption_multiplier is the conversion of the atomic ratio into the equivalent volume as described in Table 4. 135 . 5. 4. is it clear that the diffusion occurs very quickly and that within less than a second the concentration is greater than 90% of the equilibrium.
Determination of the damping coefficient Unless the cantilever is vibrating in a vacuum. To accurately model the system.Fig. Using a microscope. which in this case would be for the damping fluid air. the Polytec Instruments vibrometer laser beam was focused into a spot of about 20 µm in diameter. The time to reach 90% of an equilibrium concentration of 1%. 5. Concentration as a function of time for a cantilever coated with 0. or 10000 ppm. which was required to ensure that the vibrometer was only reading the cantilever and not the underlying substrate. Once the vibrometer was set up. This spot was smaller than the width of the cantilevers that were examined.17.10. a digital 136 . This was accomplished by utilizing the quality factor equations and the vibrometer. it was necessary to estimate the damping coefficient. 5.1 µm thick layer of palladium. is about 1 second. there will be some sort of damping acting on the system.
which is directly related to the mass and damping as was shown in Eq. using the result from Eq. 5.18 represents the data. For the cantilever D22. 4. which is the viscosity of nitrogen at 1.4. (5. The maximum peak was found and then two halfpower points (described in Section 4. H= (5.8%) of the amplitude value.10) Table 5.18.155 Hz ) ⋅ 2. These two points were used for measuring the bandwidth. it is possible to determine H as 2 µ Lb c kg 2 1.04 ⋅ 10−8 g = 1.9 lists the calculated values for Eq. Using the spectrum analyzer (HP Spectrum Analyzer 3588A) a resonant frequency peak curve was produced as shown in Fig. as 137 .77·1011 kg/µms.48 ⋅ 10−8 s H = 21. some harmonic resonance frequencies were found. 5.880 Hz − 21.9.signal generator (Wavetek 2MHz Variable Phase Synthesizer model 650) was used to scan a frequency range that would cover the first bending mode resonant frequency.25 and the bandwidth data measured for a sample chip (D22) and the corresponding calculated damping.44.25 and the parameters L and b are known values and µ.77 ⋅ 10−11 ( 252. since the damping is defined by Eq. for which Fig.818µ m )( 33µ m ) µm − s H= kg 1. 4.9) In addition.48 ⋅ 10−8 kg s .4) were found at 1/ 2 (~70. 5. It should be noted that during the frequency sweep. 4. the damping was found to be c = bandwidth ⋅ m = ( 21.8 µ m .
18. The resonant frequency can be seen as the distinctive spike (with a diamond marker at the tip) at the frequency of 21. 5.9 and it was found that the results are similar to the resonant frequency bandwidths. 138 .590 Hz. Fig.can be seen in Fig.11.9. 5.7. As can be seen from the table. 5.11. The bandwidth was measured for these harmonics using the formulation from Eq. A digitally enhanced image of a screen capture from the HP Spectrum analyzer taken while measuring the frequencies of a cantilever.18. This damping coefficient data were then used to analytically model the absorption under different cases as will be discussed in Sections 5.9.1 and 5. The other peaks are the harmonic frequencies. All the damping coefficient values are presented in Table 5. the values for the damping are all of the same order of magnitude and fairly close to one another. Table 5.
the 350 µm long E cantilever with ideal dimensions and coated with 1000Å of palldium on top of 10Å of titanium was used as the sensing element. L µm Width.4 35 34.721x10 11 11 11 3 2. Complete cantilever analysis In order to fully investigate absorption of hydrogen into the palladium coated cantilever the system had to be analytically investigated.8 11 1.8 11 11 2. They are: a simple case with only mass changing. c kg/s Bandwidth of resonance.96 21.7 353.727x10 1. Summary of analytically calculated and experimentally determined damping coefficients.553x10 724 1. a case with changing mass and damping present. m kg Damping calculated.866x10 1.967x10 507 1. ρ g/cm 1. kg/µms Measured Characteristics Length.11. ∆ Hz Damping experiment. Viscosity of nitrogen. b µm Thickness.86x10 D22 E22 Typical Typical 303. kg/s Bandwidth of harmonics.383x10 11 11 5. This investigation considered different cases of varying complexity characterized by different parameters. h µm Height.8 21.377x10 2. H µm Mass.725x10 653 11 1.33 F22 Typical 252. kg/s Density.Table 5.354x10 725 11 1. and a case with changing mass and stiffness in the presence of damping.353x10 11 1.9. ∆ Hz Damping experiment. For calculations.96 0. Four cases were considered and each one included an increase of mass by absorption of hydrogen.5 0.82 33 0.979x10 1088 11 2. All four cases and results obtained are 139 .96 21. a case with mass and stiffness changing.
106 Hz in the shifted frequency of 7783.11) where the change in mass due to hydrogen absorption is ∆m and other parameters are as previously defined.2.2. that the percent overall uncertainty in the frequency shift is 6.11. 140 .7. It was calculated. With an uncertainty of 539.7 and the derivations for each equation within these sections are located in Appendices B and C.926% when the concentration of hydrogen is at 1%. All of the results in this section are derived in Section C. This situation is an idealized case. Case 1: changing mass The simplest of the cases considers the situation where there is no damping and the stiffness does not change. 4.1 through 5.1. (5.11.presented in Sections 5.1.19 into the following form: 1 h 2π L2 35 Eeq 33 ρeq f2 = ∆m +1 m . where the negative sign indicates that the frequency will decrease.1. These results are listed in Table 5. it is clear that the uncertainty is just about half of the value of the frequency shift making the accuracy of determining the shift challenging. 5. Using the equivalent density and modulus of elasticity listed in Table 5.268 Hz.801 Hz. Section C. at 1% concentration of hydrogen. the equation used can be rewritten from Eq. the frequency will shift by –1019.8 where it can be seen that for case 1.11.
Because of the experimental results of the palladium hydrogen system conducted in the past (Lewis.5 and 4. 1 h f2 = 2π L2 35 Eeq 33 ρeq 1 2 ∆k 1 + k ∆m 1 + m . i.06 Hz.5. Case 2: changing mass and stiffness The next case is more complicated in that it will look at both a change in the mass as well as a change in the stiffness. Figs 4.546 Hz for a shifted frequency of 7894.2. Alefeld and Völkl. 1978a. (5..22 used for this determination is repeated here for continuity.12 that the shift in frequency at 1% concentration is –908. It was then calculated from Eq.1) that the change is minimal due to the addition of a changing spring constant. 1978b). This alludes to the conclusion that a changing stiffness will not play a major role in determining the uncertainty. 141 .2. The derivations of the results in this section are located in Section C.2.11.6. 1967. Equation 4. Because of the limited data for absorption of hydrogen at small concentrations (<1000ppm) the change in the modulus of elasticity and volume due to hydrogen absorption was assumed correct without uncertainty.12) where the subscript 1 indicates the initial frequency. there is a known contribution to the change in the stiffness from the modulus of elasticity and the volume as both can alter during absorption. 5.523 Hz with a percent overall uncertainty in this shift of 6.e.11. As it can be seen from the values of the percent overall uncertainty of this and the previous case (Section 5.917% equaling 546.
The damping values used in this case were experimentally verified in Section 5. 5. Case 3: changing mass with damping The third case investigates the frequency shift that occurs due to a change in mass with damping present.11. 1 h f2 = 2π L2 2 35 Eeq m 4k ( m + ∆m ) − c 2 33 ρeq 4 k ( m + ∆m ) 1 2 . 4. The results are very similar to case 1.. Case 4: changing mass and stiffness with damping The final scenario looked at incorporates all the variables: mass. Using Eq.3.32.5.e.14) 142 . When the appropriate values are substituted into Eq.4.10.31.11. suggesting that the damping coefficient of air plays a small to indifferent role in improving the frequency data quality. The formula modeling this system is Eq. The data are presented in Table 5. This case represents the results the most accurately. 5. i. 1 h f2 = 2π L2 2 35 Eeq m 4k ( m + ∆m ) − c 33 ρeq 4 ( k + ∆k )( m + ∆m )2 1 2 .13) the contribution that damping has is accounted by the parameter c. (5. which had no damping. and stiffness. The results from this investigation are similar to those without any damping even though there is a significant change in the formulation. damping.13 at a hydrogen concentration of 1% the percent overall uncertainty in the frequency 6.10. 4. repeated here for continuity. which for this case was calculated for the cantilever in air.113 Hz. (5.927% with an uncertainty of 539.
it is necessary to determine the level of accuracy in modeling needed in order to maximize data quality.1 through 5. 143 .4.10. 5. All the data are compiled in the Table 5.5. and in turn equation. From the data collected from each of the four cases described it was possible to determine which parameter (stiffness and damping).11. Design equation selection In designing cantilever sensors. In other words. This simplifies the equations and parameters needed and also the experimental setups as less factors need to be taken into account. it is only necessary to account for changes in mass.which is very similar to the damped case described in Eq.11. to model a vibration system that is damped by air.13. as changes in any other parameter will not lead to significantly improved results. It has been calculated that the change in the frequency due to the changing stiffness and damping is 532. It can be seen from the analytical results in Table 5. were the most and least necessary for accurate results.10 that by being more accurate and including such parameters as damping and a changing stiffness the overall uncertainty in the frequency does not alter much. This means that it is not really necessary to include these other parameters to get accurate results under the conditions used in this Thesis. The results are similar to those of the case 3 with the changing mass and stiffness. it is possible to argue what are necessary variables and considerations in designing cantilever sensors.11.137 Hz while the final percent overall uncertainty in frequency is 6. Using the results of Sections 5. 5.696%.
313 1019. shift. The results of the four different cases investigated are compared at 1% hydrogen concentration.801 539.137 6.5.k. Freq. as effects produced by these environments will not affect the results and reliability of it.106 6. Hz Overall % freq. it was possible to estimate the frequency shift amount that would occur at lower concentrations of hydrogen. such as 1 ppm.k) 7894. as mentioned in Section 4.696% 5..11 the frequency shift at 1 ppm was calculated using the same process as described in Section C.546 546. uncert.11.6. With these approximations in Eq.c) 7783.619 1129. but can be approximated using diffusion laws. Frequency data at 1% H2 Shifted freq.1.927% Case 4: (m.917% Case 3: (m. 5.10.523 908. Of course. the exact amount of absorption at such low levels is not very well known.2. Lower limit frequency shift As determined in Section 5. at room temperature and at 1 atm. Case 1: (m) 7783.756 539.450 532. Hz Overall freq. With this knowledge.2.c) 7673. as 144 .11. it is not necessary to include variables such as damping and a changing stiffness as the quality of the results will not improve by much.4.268 1019.06 6.113 6.926% Case 2: (m.This also increases the usability of the sensor in a wider variety of environments. uncert. Table 5. absorbing into palladium.
And the other reason is that the mass (quantity) being attempted to detect is too small in comparison to the mass of the cantilever used. 145 .069 Hz . If the palladium does in fact absorb more hydrogen than what is being used in the equations. where the mass added is very low (1022 order of magnitude) and therefore the shift at this low concentration level is very small (<0.995 3 cm −22 f2 = 1.1 h 2π L2 f2 = 35E 33ρ eq eq ∆m +1 m 1 1µ m 2π ( 350 µ m )2 35(172. which can allow for misrepresentation of the behavior of the cantilever.3 is not readily available at small concentrations. It is shown that at around 10 ppm the frequency shift is almost unnoticeable and would be very difficult to detect. The second reason is that the partial pressure of the hydrogen is too small to activate absorption into the palladium.001 Hz). Figure 5. 4. Since the hydrogen will be mixed down with nitrogen. There are a few possible reasons for this.9GPa ) g 33 3. The first is that the data of absorption of hydrogen required to accurately graph the curve of Fig.652⋅10 kg +1 5. then the shift will increase. the partial pressure will be low.918⋅10−11 kg (5.15) f 2 = 8803.19 illustrates how the frequency shift increases as the concentration of hydrogen is increased. which will lead to very small frequency shifts and will require highly sensitive frequency shift detecting methods and equipment.
19. was increased by two orders in magnitude. The frequency shift in the resonance of a cantilever as a function of concentration of the hydrogen.11. Upon close inspection it was found that for each order of magnitude 146 . the overall length of the cantilevers was reduced by two orders of magnitude (for example: L=3. there are several factors that must be considered. 5. First. Next the partial pressure. and thus in turn the overall pressure of the gas.Fig.5 µm). Optimization of the cantilever response In order for the cantilever to detect a shift in the frequency that is greater that 109 Hz.6. The frequency shift was increased by a total of 4 orders in magnitude with these modifications.7. Addressing the issues brought up in Section 5. parameters of the frequency equation will be altered and will be documented on their effect on the results. 5.11.
the frequency shift increased by one order of magnitude. 147 . the current uncertainties will basically prohibit the fabrication of these cantilevers at these scales. This indicates that higher frequency cantilevers are favorable for sensors as they are more sensitive to frequency shifts. having a low uncertainty in the thickness parameter is important for the future development of cantilever sensors. 1 ppm or less. For example. as the uncertainties will be extremely high. For this reason.increase in the partial pressure. the sensor will be out of specifications. This has some serious effects on the sensors capabilities due to the uncertainty in the thickness discovered.1 µm (or less) for the sensor. For the geometric dimensions. slight differences in the geometry will change the sensitivity and thus. the results of decreasing the length and increasing the thickness by orders of magnitude increased the resonant frequency of the cantilever that directly affected the frequency shift. if the thickness is desired to be 0. At such small concentrations. If a cantilever is fabricated 5% (or greater) off from the planned structural dimensions the high uncertainty will undermine sensitivity and function of the sensors. regardless of calibration. To attain these cantilevers with higher resonant frequencies the overall dimensions such as length and thickness must be reduced.
were used. computational. More specifically. using analytical.15%. Typical values for the resonance frequency and the percent overall uncertainty were found to be around 16 kHz and 6. To characterize the cantilevers. CONCLUSIONS In this Thesis. Analysis and preliminary characterization of vibrations of these cantilevers were investigated. δh =0.05 µm contributed 74. The primary investigated aspect involved determination of a nominal frequency and overall uncertainty in the frequency of the cantilevers. Vibrometer. uncertainty in thickness had the greatest contribution to the overall uncertainty in the frequency. analytical calculations were performed.. MEMS cantilevertype chemical sensors were investigated. It was also found that out of all independent parameters defining resonance frequency. i.9 % to the overall uncertainty in the resonance frequency of the 350 µm long cantilever. several chips. and AFM methodologies were used in gathering experimental results. 250 µm. in this Thesis.e. This investigation involved. respectively. The resonant frequency for each cantilever was determined for comparison with experimental results. 300 µm. and experimental solutions (ACES) methodology. OELIM. Each chip had 6 cantilevers of different lengths and for this Thesis the three longest cantilevers. An AFM was used to determine the frequency response curves based on which the resonant frequencies of cantilevers were calculated. 350 µm. The OELIM method was particularly capable of determining mode shapes associated with specific 148 .6.
the percent overall uncertainty in the frequency. 149 . It was found from the analytical results of this Thesis that. The sensitivity was calculated to determine how a sensor would function the best under the conditions of room temperature (20˚C) and atmosphere pressure (1 atm). did not change much.3%. ~0.3% change. Since the damping and a changing stiffness do not affect the frequency significantly. The laser vibrometer was used to determine the damping coefficients. The percent overall uncertainty contribution calculations indicated that the thickness was the greatest source of uncertainty and that the analytical and computational results correlated well with the experimental results. or the quality of the results. Using equations that included such variables as damping and a changing spring constant several cases were considered. on the order of about 0. indicated that the analytically calculated uncertainty was accurate. although damping due to the nitrogen environment and a changing stiffness do alter the frequency. The difference. Investigations into the cantilever as a chemical sensor for detecting hydrogen with the palladium/hydrogen system were also conducted. the sensitivity of the cantilever to the palladium/hydrogen system was determined without considering these parameters. It was concluded that increasing the resonant frequency is the most direct way of improving the sensitivity of the sensor.resonance frequencies for a number of vibration modes for each cantilever sample. It was found that the analytically calculated percent overall uncertainty was similar to the percent difference between the analytically calculated frequency and the frequencies measured by the AFM and the OELIM methodologies.
. it will be necessary to improve the uncertainty in the thickness. the ability to make smaller devices more accurately will facilitate development of the cantilevertype MEMS chemical sensors. It was found that to improve the sensitivity of these sensors. the overall percent uncertainty in the frequency will increase due to the uncertainty in the thickness. However. i. 150 . This will cause problems with sensors designed for detecting 1 ppm.This can be readily accomplished by changing the overall dimensions of the cantilever. regardless of calibration because this uncertainty will produce variation in the sensitivity and may put them out of the specifications. This Thesis investigated a MEMS cantilevertype chemical sensor and its design. decreasing length and increasing the thickness.e. the fabrication processes that make these cantilevers would have to be improved to provide an order of magnitude decrease in the uncertainty in thickness. As fabrication processes improve. or less. if the length of the cantilevers is decreased and the thickness is increased. More specifically.
D. Baselt. Z. Ramseyer. New York. K. Phys. CA. Battiston. Peppas. NY. D. AD. 77:122131. Ruiz. SpringerVerlag.analogdevices.present and future". R. O. Appl. 1993. NC. E. Baselt. Borrisé. A. and N. 2003. L. Last visited: November 3. 145. Analog Devices. The tipsample interaction in atomic force microscopy and its implications for biological applications. “Design and performance of a microcantileverbased hydrogen sensor”. “Micromechanical cantilever as an ultrasensitive pH microsensor”. NV. 2003. E. REFERENCES G. Nanotechnology. Hydrogen in metals. 151 . Hagleitner. Vol. Brand. B. X. Davis. J. R. G. K.. 4th Internat.com. pp. Klaassen. Pryputniewicz. 2003. and N. J. S. Esteve. O. Baltes. McCorkle. and H. 1999. J. R. Lett. http://www. Hydrogen in metals. Sensors and Actuators B. M. Abadal. pp. D. 1978b. J. 211217. SpringerVerlag. Barniol. Z. D. “iMEMS® accelerometers”. Figueras. J. Sensors and Actuators B. Symp. Ph. H. Hilt. Dissertation. PérezMurano. A. pp. 81(16):30913093. Campabadal. 2. F. 459466. Baller. Boisen. 2001. J. Lang. Hierlemann. Composite materials: mechanical behavior and structural analysis. NY. B. A. 12:100104.” Proc. Elibol. Völkl. New York. K. 88:120131. Berthelot. 1. G. Ch. Arecco and R. Fruhberger. “A chemical sensor based on a microfabricated cantilever array with simultaneous resonancefrequency and bending readout”. P. 2002. A. Bashir. E. "Electromechanical model of a resonating nanocantileverbased sensor for highresolution and highsensitivity mass detection". M. Mlsna. E. 1978a.D. T. Warmack. Helbo. F. Pasadena. Charlotte. New York. Proc. California Institute of Technology. 2003. Gerber. Lange. Britton Jr. R. on MEMS and Nanotechnology (4thISMAN). 2001. Günthherodt. S. 2002. D. "CMOS MEMS . Gimzewski. C. Alefeld and J. on MEMS 2002. SpringerVerlag. Las Vegas. Vol. Völkl. NY. Patel. F. “Design and analysis of a MEMS chemical sensor. V. Gupta. Cemalovic. J. Alefeld and J. and C. Symp. and B. M. P. H. J. Meyer.7..
Jr. 2860:111121. P. R. Sepaniak. Ph. 359368. 2000. Roukes. D. C. Atlanta.. Datskos. MEMS. Bryan. and P.D. Rev. Depriest. "Stiction. L. Laser interferometric methodologies for characterizing static and dynamic behavior of MEMS. Brown.” invited paper. More. Pryputniewicz. L. Brown and R. W. and J. 63(3):34023405.. C. T. Kelley. Pryputniewicz.. 1999. Britton. C. Betts. “Holographic microscope for measuring displacements of vibrating microbeams using timeaverage holography. 37:13981405. C. C. W. on Advanced Research in VLSI. Worcester. G. 1998. Brown and R. adhesion energy. and the Casimir effect in micromechanical systems". I. L. W. Warmack. M. Hu. 8(4):433438. B. 1999. J. J. "Batterypowered. P. L. T. A. Rochelle. Bryan. L. M. Opt. E. Jr. W. A. J. 2000. J. Dissertation. J. 1995. Worcester Polytechnic Institute.” Industrial applications of optical inspection. A.. Pryputniewicz. MA. L. SPIE. F. Pryputniewicz. 82:1721. Tipple. Warmack. M. Buks and M. 1992. 2001. “Lownoise MEMS vibration sensor for geophysical applications”." Ultramicroscopy. 1999. C. J. Smith. F. M. “Investigation of submillimeter cantilever beams of variables cross section by computational and hologram interferometry methods. J. Brown and R. J. Analytica Chimica Acta.” SPIE. “Experimental and computational determination of dynamic characteristics of microbeam sensors. C. Wintenberg. Hu. M. R.J. pp. Bernstein. R. Opt. Smith. L. Brown and R. Turner. Threatt. C. Eng. "Multipleinput microcantilever sensors. G. Oden. J. S. C. 2000. 2545:108119. G. L." Proc. Britton. M. and P. N. S. Jones. 422:8999. wireless MEMS sensors for highsensitivity chemical and biological sensing. “Holographic microscope for measuring displacements of vibrating microbeams using timeaverage electrooptic holography. Emery. Z. Eng. 152 . L. Clonts. 1996. I. “Selectivity of chemical sensors based on microcantilevers coated with thin polymer films”. Ward. Miller. Ericson. M. Symp. GA. Thundat. 1821:7281. Brown. Rochelle. Jones. G. metrology and sensing. and J.” invited paper. R. J. R. Oden. Z. T. “New test methodology for static and dynamic shape measurements of microelectromechanical systems. 39:127136. Phys.” SPIE. R. Pryputniewicz. C. G. S. Brown and R.
Worcester. J. T. Baller. 16:96949696. J. 1995. and J.J. 288:316318. J. SPIE. C. G. Pryputniewicz. C. 80(12):2219–2221. Gimzewski. C. (2):16. C. “Characterization of a twodimensional cantilever array with throughwafer electrical interconnects”. 2003. Dartmouth. 2002. Cherian and T. H. J. “Stress at the solidliquid interface of selfassembled monolayers on gold investigated with a nanomechanical sensor”. 80:664. 2000. P. K. Meyer. K. Appl. Langmuir Lett.. W. UK. Pryputniewicz. C. “Material pictures”. G. 2002. M. MA.D. Dissertation. 2002. K. London. “Translating biomolecular recognition into nanomechamics”.. Lang. Meyer. Fritz. “A bidirectional magnetic microactuator using electroplated permanent magnet arrays”. SPIE. H. J. Ph. Hanover. NH. Ahn. 2000.. MEMS." Appl. S. Science. P. P. Quate. Gerber. E. and J. F. Darling.” Proc. Güntherodt. C. Furlong and R. H.S.” Opt. Lett. Yaralioglu. 2003. J. Phys. Worcester Polytechnic Institute. 2002. Microelectronics Laboratory Dartmouth College. 1(1):6066. MEMS. H. 2545:192203. Lett. Kenny. M. J. Furlong and R. "Determination of adsorptioninduced variation in the spring constant of a microcantilever.com. and T. “There’s plenty of room at the bottom”. “Residual stress examples”. Phys. Cho and C. Gimzewski. 11(1):7884. EC. “Characterization of shape and deformation of MEMS by quantitative optoelectronic metrology techniques. E. E. Hybrid. Furlong. H. “Absolute shape measurements using highresolution optoelectronic holography methods. Furlong and R. Fritz. 153 . 4778:110. H. Last visited: December 3... approach for the efficient study and optimization of mechanical and electromechanical components. M. Vettiger. Feynman.J. Rothuizen. Chow. C. Platin. experimental and computational. Baller. Gerber. R. Delamarche. J. Lang. Strunz. 2000. E. 39:216223. Metals Rev. Eng. 1992. P. H. 1958. “Optomechanical study and optimization of a cantilever plate dynamics”. A. K.. http://elementcollection. The Red Green & Blue Company Ltd. Thundat. 1999. Güntherodt. Pryputniewicz. 2002.
Matta. McDevitt. 2001. E. Phys. Boyle. MEMS and microsystems in Europe.. 2001. Appl. Proc. McGrawHill. Rayment. T. Clevenger. Hu. 2001. on Microelectronic Structures and MEMS for Optical Processing II. C. R. L. OR. Warmack. NH. WA. Congress on Experimental and Applied Mechanics in Emerging Technologies. "Fabrication and characterization of a FabryPerot based chemical sensor. Marinis. 154 ." Proc. NY.R. and R. G. Yokum. Abell.. 2000. “Sensitivity. editors. 90(1):427431. Anagnostopoulos. J. C. 2002. International Technology Research Institute. Portland.D Dissertation. Pryputniewicz. S. Weeks. Harley. L. R. Grabiec. J.. “Lowenergy proton detection by Pd metalinsulatorsemiconductor diodes. Ph. 64(12):68396844. 4778:216223. T. “A nondestructive technique for determining the spring constant of atomic force microscope cantilevers”.. Ojha. J. E. “Advances in optimization of MEMS inertial sensor packaging. BCO America. F. 1988. and J. I. Gormley. Jacobson. and I. J. Pryputniewicz. Internat. C.” J. Lee. A.” Proc. Z. Stanford. Hilbert. S. TX. 2000.. C. Finnila. pp. Northern Ireland. and R. J. Blackstone.C. 2000. MEMS and microsystems: design and manufacture. 55th Internatl. A. Rangelow. 2002. T. D. 1996. Austin. T. Durham. Hughes and R. Leonard. Hones. Advances in piezoresistive probes for atomic force microscopy. W.” Proc. Bastasz. New York. and S. Conf. 430437. Han. “Investigation of adsorption and absorptioninduced stresses using microcantilever sensors. C.” J. and R. Ultramicroscopy. Furlong. Instr. BCO Technologies (NI) Ltd. B. and T. J. Bailey. 2001. Phys. Furlong. 2002. 41744208. accuracy. Belfast. 82:3948. “Piezoresistive sensors for scanning probe microscopy”. Appl. Gibson. Hawkins. Motamedi and W. pp. C. P. Frequency Control Symp. 72(5):23402343. SPIE. P. A. Gotszalk. 821825. Thundat. Hsu. T. T. “Quantum 1/f effects in piezoelectric sensors”. CA. F. H. D.J. P. Hanson. Handel. M. Melendez. and J. R. SPIE2881:171178. C. Stanford University. J. M. Sci. Neikirk. Curtis. and precision issues in optoelectronic holography based on fiber optics and highspatial and highdigital resolution cameras. Rev. H. pp. HARM processing techniques for MEMS and MOEMS devices using bonded SOI substrates and DRIE. Seattle.
and K. 155 . H. Kern. H. “Single cell detection with micromechanical oscillators”. H. and U. and temperature. Myazaki. San Diego. Mechanics of composite materials. 2001a. A.zurich.html. Craighead. CA. Craighead. Sarkar. Appl. Worcester. L. 2000. L. A. Lett. Proc. Prins. “A microcantileverbased alcohol vapor sensorapplication and response model”. C. 2003. J. NY. A. Schaufelbühl. Maher. IEEE MEMS 2000. S.. G. Tamayo. 19(6):28252828. Vac. “Active quality factor control in liquids for force spectroscopy”. M. Sci. Instr. 2003. Phys. http://www. O. 2000. IBM Zurich Research Laboratory. J. 16:78917894. Jericho and M. 77(3):450452. Hemisphere Publishing Corp. 2001. IBM. “Mechanical resonant immunospecific biological detector”. 78:1218. Boisen. S. 2002. B. Rev. N.tms. Veje.A. Thaysen. A. B. R.. pp. P. D. R. Japan. F. Appl. Schaum’s Outline of Theory and Problems of Mechanical Vibrations. D. Weimer. Hansen. 2000. http://www.. New York. S. K. Madou.. 52(10). D.html. Kassegne. Kerness. Ilic. Development of MEMS sensors for measurements of pressure. Kelly. Langmuir Lett. “Design issues in SOIbased highsensitivity piezoresistive cantilever devices”. O. Phys. Sensors and Actuators B. “Advanced sensors for multifunctional applications”. MS Thesis. Switzerland. NY. Ilic.. “Cantilever sensors”. Baltes. 1996. Zoval. 73(6):24832485. J. 76(18):26152617. 2000. MA. and H. 2000. V. “Nwell based CMOS calorimetric chemical sensors”. Metals. M. Jensenius. Zurich. B. Rasmussen. Czaplewski. J. G. J. Koll.ibm. V. Brand. A. J. Zalalutdinov. Last visited: November 3. Johari. Worcester Polytechnic Institute. Jericho. and A. relative humidity. Whitten. Czaplewski. New York. K. Tech. on Smart Structures and Materials. and H. and H. Hagleitner. and M. “Multicomponent analysis and prediction with a cantilever array based gas sensor”. “Device for the determination of spring constants of atomic force microscope cantilevers and micromachined springs”. Jones. S. Lett. 2002.org/pubs/journals/JOM/0010/Ivanov/Ivanov0010.com/st/nanoscience/cantilever. 1975. Sci. SPIE4693:588597. G. Raible. H. Humphris. J. E. Ivanov. D. B. M. Kim. Hierlemann. McGraw Hill. 2002.. Miles. 96101. D. H. Conf. B.
A. J. A. Chung. A. Tech. Schneeberger.P. Hong. M. C. Brugger. Chem. 2002. Baller. Sensors and Actuators A. Appl. P.W. H. 78(11):16371640. 1936. Li. “A new organic modifier for antistiction”.. O. “Complementary metal oxide semiconductor cantilever arrays on a single chip: masssensitive detection of volatile organic compounds”. CRC Press. Y. H. H. IEEE MEMS 1999. D. Oh. Lai. Lett. J. Berger. D. O’Shea. U. 4:5155. Thundat. FL. Zhang. Mater. E. Hedden. “CMOS resonant beam gas sensing system with onchip self excitation”. Meyer. and H. O. Ng. Conf. 547552. Kim. Mech. New York. 103:122129. 2001. Proc. 14th IEEE Internat. H. Lee. Lange. “Analysis of surface effects on mechanical properties of microcantilevers”. 2002. and H.B. J. 10(1):3340. A. Fang. Güntherodt. and D. C. Hierlemann. and J. K. 2nd ed. M. Anal. H. 547551. P. Orlando. K. New York. 2001. C. J. Madou. C. Phys. 2002. Manias. Brand. Baltes. Academic Press. D. Fornaro. 2001. “Polymeric micromechanical components with tunable stiffness”. J. S. (17):302. Phys.S. Phys.. Cherian. O. Shen. Chen. 12th IEEE Internatl. J. Hagleitner. T. Appl.J. Fundamentals of microfabrication: the science of miniaturization. D. MEMS. “Manipulation and controlled amplification of brownian motion of microcantilever sensors". J. Koll. N. 2001. Proc. Krüger and H. 2003. 1967. Bu. K. NY. Battiston. “Selfreciprocating radioisotopepowered cantilever”. NY. Baltes. Appl. The Palladium hydrogen system. P. 2001b. Gerber. Fritz. R. F. J. S. 92(2):11221127. H. 156 .J. U. and J. Phys. Ramseyer. E. U. Mehta. Lewis. Schaufelbühl.M. Gimzewski. Lu. Nam.. S.. “Micromachined CMOS calorimetric chemical sensor with onchip low noise amplifier. Kim. Henderson. Y. FL. Orlando. Lang. F. 1999. H. 79(11):17001702. Lange. on MEMS. Brand. Blanchard. Lett. Kim. Drechsler. Münch. pp. A. Jungnitz. M. Cho. Interlaken. Conf. H. Z.C. F. M. “The Nanomechanical NOSE”. and X. on MEMS.. and T. F.. Switzerland. and H. Rothuizen. pp. A. 913. J. Vettiger. P. Baltes. Phys. N. Despont. “PZT cantilever array integrated with piezoresistors sensor for high speed parallel operation of AFM”.. and T. Chun. 1999. D. 74:30843095." Proc. Hagleitner. and K. J. Brand.
J. Worcester Polytechnic Institute. 1982. MEMS SUMMiT™V Technology. D. L. “Silicon as a mechanical material”. IEEE. http://www. "A Method of Measuring Specific Resistivity and Hall Effects of Discs of Arbitrary Shape". Measurements Specialties. J.MEMC. Last visited: November 24. Y. KG. MA 157 . MEMS Packaging. 2003. R. 2003a. Zhine.de/. 245253. For Exp. Oh and R. 2003. 1998. Valley Forge. MEMS and nanotechnology. 2003. OR. Pryputniewicz. Mech. Worcester. 1958. “Polytec Vibrometer”. Portland. Worcester Polytechnic Institute. J. “Piezo Tech”. 2002a.memc. “Silicon wafers”. 2002b. 2003. MA. R. ACES approach to the development of microcomponents. Worcester. Worcester. Worcester Polytechnic Institute. “AFM chips”. MA.com. MA. MS Thesis.” in Hologram interferometry and speckle metrology.physikinstrumente. 97(14):365369. R. van der Pauw. 2003. 2003. R. Worcester Polytechnic Institute. Inc. MS. 13:19. Thermodynamics of microsystems. W. Pryputniewicz. 1997. Worcester Polytechnic Institute. Pryputniewicz. PI. P. Pryputniewicz. Polytec. and R. “An embedded polymer piezoresistive microcantilever sensor”. Last visited: December 1. http://www. Physik Instrumente GmbH & Co. Bethel. E. Last visited: November 3. PA. Eastman. CT. 70(5):420457. “Piezo film sensors technical manual: internet version”. MA. Petersen. Polytec GmbH. 2003b. J. Pryputniewicz. MM. Worcester. Pryputniewicz. Soc. J. 1990. Delinger. 2003. G. J.polytec. Proc. 2003. R.mikromasch.. MikroMasch USA. Monsanto Electronic Materials Company. 2003. K. http://www.com/. Philips Research Reports. pp. Last visited: November 3. T. Porter. Ultramicroscopy. M. “Application of electrooptic holography in the study of cantilever plate vibration with concentrated masses. L. Macomber. http://www. Worcester.com. C. K.
” Industrial laser interfermetry.” Quantitative analysis of holograms. 912919. WI. “Vibration studies using heterodyne hologram interferometry. 604:7185. "Quantitative determination of displacements and strains from holograms. Symp. Pryputniewicz. SPIE. SPIE. R. IX Internat.“ Internat. “Applications of electrooptic holography in vibration studies. Bellingham.” Photomechanics and speckle metrology. San Diego. Congress on Exp. Photogramm. MA. “Quantitative holographic analysis of small components. pp. SEM. R. Pryputniewicz. CA. R. Chen.” Proc. J. Pryputniewicz. ed. J. Pryputniewicz. Mech. pp. Furlong. 2000. 1162:456467. “Advances in optical methodology for studies of dynamic characteristics of MEMS microengines rotating at high speeds. J. 599:5462. R.” Proc. R. P. pp. 3372. J. Pryputniewicz. Vol. J.” Proc. Pryputniewicz.” Optics in Engineering Measurements. 1986. “Novel optoelectronics methodology for testing of MOEMS. 1989.. M. J. 68 of Springer Series in Sciences. “Static and dynamic measurements using electrooptic holography. “Measurements of vibration patterns using electrooptic holography. C. 158 . R. J. Pryputniewicz. Pryputniewicz. 1981. F. “Hologram interferometry from silver halide to silicon and beyond. 3 in Holographic interferometry. J. Pryputniewicz. 4983:1125. pp. WA. Pryputniewicz. 24:377386. Bethel. Milwaukee. “Timeaverage holography in vibration analysis. SPIE. Worcester. Eng. Pryputniewicz.” Proc. J. P. on MOEMS and Miniature Systems. 1991b.. Arch. 1985a.” Opt. SPIE. Internat. CT. Brown. 1995a. Pryputniewicz. 790798.. R. and E. de Boer. and G. J. Pryputniewicz.R. Berlin. J. C. “High precision hologrammetry. R. of the 1991 SEM Conf. 2003a. 24:843848. R. 1995b. “Holographic and finite element studies of vibrating beams. J. Pryputniewicz. 10091012.. R. Worcester Polytechnic Institute. 1985b. SPIE.” Proc. Engineering experimentation." Ch. SpringerVerlag. R. J. R. 955:154161. J. 1993. 1988. on Experimental Mechanics. 1991a. 2545:405427.
S. and E. 79:115126. and P. Seta. Furlong. “Micromechanical cantileverbased biosensors”. Pryputniewicz.com.” Proc. On Microelectronics. E. New York. Brown. pp. and C. J. 392397. Raiteri. Last visited: November 24. NY. Hefti. C. and M. Soucy. Klempner. 2003a. “Optical methodology for static and dynamic measurements of nanodisplacements. J. and Exhibition of IMAPSNE. 2003b. R. Pryputniewicz. PTC.. MA. F. on Microelectronics. Symp. and C. Butt. W. C. Internat. Marinis. Hanson.com.. IMECE2001/MEMS23906. Pryputniewicz. Furlong. Boston.ptc. Pryputniewicz. Boxboro.” Proc. 159 . 1995. “Pro/Mechanica”. and C. Symp. A. Seta. “ACES methodology for studies of MEMS. J. Pryputniewicz. Am. “RF MEMS: modeling and simulation of switch dynamics. Grattarola. Pryputniewicz. “Measurement of vibration patterns using electrooptic holography. 1990. A. E. 2002b. R. J.J. Soc. J. 267272. Pryputniewicz. G. C. 28th Annual Symp. T. S. Internat. 35th Internat. 2002a. “New approach to development of packaging for MEMS inertial sensors. and M.ptc. Brown. Mech. Mechanical vibration. NY. Wilkerson. 3rd ed. Symp. Eng. P. Eng. 95100. pp. Pryputniewicz. Denver.” Proc. J. 2003. 1162:456497. T. E. J. New York. PTC. Sensors and Actuators B.R. 8186. R. F.. G. R. Rao. “Optimization of contact dynamics for an RF MEMS switch. 2001. J. Furlong. Soc. Pryputniewicz and K. A. MA.” Paper No. J. New York.” Proc. on Microelectronics. Portland. 2002c. R. 2003b. Furlong. E. pp. 208217. CO. pp. Pryputniewicz. “Design by analysis of a MEMS pressure sensor. S. R. 2001b.” Paper No. Furlong. 35th Internat. Pryputniewicz. http://www. Przekwas. On MEMS: Mechanics and Measurements. “Novel membrane methodology for characterization of holography for MEMS inertial sensors. 2001c. J. 2003. Denver. Furlong. Am. pp. CO. R. J. Mech. Skladal. J. and E. “Pro/Engineer Wildfire”. Furlong. PTC. IMECE200239504. Symp. Marinis. C. 2001a. W. H. PTC. C. Last visited: November 24. http://www. J. R.” Proc. Pryputniewicz. M.” SPIE. AddisonWesley. P. OR. Stetson. C. J. D. R.
A. S.. J. and M. A. J. Doktycz. Minne. Warmack. Conf. Switzerland. Last visited: November 3. 320323. TI. F. on MEMS. A. Rev. “DLP”. Appl.gov/scripts/index. G. T. Humphris. J. Miles. 1998. Ultramicroscopy. G.S. R. B. Oden. E. Bedford. “Dynamics of damped cantilevers”. L. 86:167173. Phys. A. J. Vettiger. 2000. Stoney. L. M.. 2001.. and S. Appl. J. 160 . 2001. Proc. Adderton. Menon. Boisen. J. Jensen. Symp. Proc. A.asp. Brown. 15th IEEE Internat. 2003. Thaysen. 1999.dlp. of MEMS. New York. S. D. http://www. 2002. pp. C. Sulchek. NM Last visited: November 3. Smith. R. Marie. J. Vestergaard. Proc. B. L. and K. Atalar. NY. Bouwstra. Lett. Wattinger. Soc. G. Malloy.. Sci. Spiegel and J. Mortensen.sandia. R. Stephan. Rast. and A. Streeter. 73(1):3641. Rev. SNL. Jacobson. P. “SU8 based piezoresistive mechanical sensor”. Meyer. I. A. and E. Texas Instruments. C. T. Hansen. M. Boisen. Las Vegas. R. W.D. Hsieh.. and D. R. Yaralioglu. Quate. K. "Cantileverbased biochemical sensor integrated in a microliquid handling system". P. 1909. Instr. 2001. McGrawHill. C. Subramanian. C. London Ser. Gysin. A. Albuquerque. Cleveland. “Microcantilever chargedparticle flux detector”. Thundat. Gaulden. pp. New York. Sci. Yalcinkaya. J. A. Phys. 2000. S. Lett. Miller. McGrawHill. Thundat. 81(2)385387. W. "Atomic force microscopy probe with piezoresistive readout and a highly symmetrical Wheatstone bridge arrangement". M. Sensors and Actuators A. NV. http://mems. and A. T. 76(11):14731475. 2nd ed. Interlaken. 2000. F.. M. and M. “Micromachines”. Fluid Mechanics. 2002. Thaysen. “Glucose biosensing using an enzymecoated microcantilever”. A. G. Kennel. “Highspeed tapping mode imaging with active Q control for atomic force microscopy”. 2002. P. 2003. M. Liu. NY. 9th ed. R. U. J. K. 83(13):4753. Tamayo. D. Adams. Schaum’s outline series: Mathematical handbook of formulas and tables. 71(7):2772. 401404.com/. “Chemical sensors and biosensors in liquid environment based on microcantilevers with amplified quality factor”. Sandia National Laboratories. 2003. Instr. Wylie. 82:172177. V. D. and T. J. O. Thaysen.
Riesenberg. Am. Stipe.Q. R. 2001. Chow. Wang. Richie. pp. A. pp. Boisen. T. Chau and P. K. “Quality factors in micron. Yasumura. Z. Phys." Proc. Lett. 77(24):40614063. French. editors. 2000. 3224:2930. and M. Austin. 2002. D. G.J. B. E. Tai. 2000. “Polymerbased electrospray chips for mass spectrometry”. Micromachining and Microfabrication: Micromachined Devices and Components III. W. V. “Surface effects and high quality factor in ultra thin singlecrystal silicon cantilevers”. Nitzsche.. “Chemical sensing in fourier space”. and A. Y. 10th ed. Proc. 523528. Finot. Neikirk. “Design and analysis of a dynamic MEM chemical sensor”. 2000. M. Thaysen. A. Lavigne. Kenny. Appl. J. X. S.and submicronthick cantilevers”. Wu. Young and R. Yu. "Micromachined storage wells for chemical sensing beads in an 'artificial tongue'.. Pfafman. H. MEMS. B. New York. D. T. Comparison of Bosch and cryogenic processes for patterning highaspectratio features in silicon. 1999. C. Wuttig. Stowe. 161 . NY. Savoy. S. B. Proc. Control Conf. G. D. K. Shear. University physics with modern physics. J. J. 2000. Hu. 2:12141218. P. SPIE. McDevitt. Yang. IEEE MEMS 1999. Phys. “Integral field and multiobject spectrometry with MEMS”. SPIE4480:367376. A. TX. Desai. 92(10):62966301. Imaging Spectrometry VII. Rugar. E. Lett. Symp. Thundat.. H. T. J. “Optimization of sensitivity and noise in piezoresistive cantilevers”. and D. J. Turner and W.. VA. 440712.J.. L. 9(1):117125. E. Phys. Ono. Appl. and T. O. Yoo. McDoniel. UK. J. AddisonWesley. Lee. K. T. 2002. Esashi. R. K. Oxford Instruments plc. Zhang. Arlington. Majumda. M. FL. Anslyn.. 1997.T. Hansen. X. Y. Orlando.C. Appl. T. J. and A. Freedman. and J. 77(23):38603862. J. Walker.
°K Modulus of elasticity. g Crystal structure Density. The values listed in the table are the ones used in this Thesis unless otherwise specified in the text.32 no data Silicon 14 28. 2003) for the purpose of a visual association to each element. Table A. PROPERTIES OF MATERIALS USED For the materials used in this Thesis. A. Physical properties of the materials used.01 20.1 to A. °K Boiling point. GPa Poisson's ratio Reflectivity Palladium 46 106. A photograph of palladium samples (EC.867 hcp 4507 1941 3560 116 0.1. 2003).APPENDIX A. 162 . the properties of each are tabulated below.3.1. Material name Atomic number Atomic weight. kg/m3 Melting point. These pictures are of small (no scale) samples from The Red Blue & Green Company (EC.28 no data no data no data Titanium 22 47.00794 hcp no data 14.0855 diamond 2330 1687 3173 190 0. pictures of what the materials look like are shown in Figures A. Table A. In addition.1.42 fccp 12023 1828.22 28% Fig.05 3236 121 0.39 72% Hydrogen 1 1.
A.Fig. unprocessed. 2003).2. A. Fig. 163 . A sample of titanium. uncut silicon (EC.3. a popular material due to its corrosion resistance (EC. A sample of untreated. 2003).
2 dt where C1.3 and B. Eq.2.2.4 into Eq. k is the spring constant. Fundamental resonance frequency In determining the fundamental resonance frequency of a cantilever we start with the general equation for a body in motion.1. and F(x. c is the damping coefficient. Fig. m d 2x dx + c + kx = F ( x. Substituting Eqs B. B.t) is the time dependent forcing function acting on it.APPENDIX B.2 are arbitrary constants. ANALYTICAL CALCULATIONS In this Appendix. is derived stepbystep (Kelly.2) which is a simple homogeneous equation.4) 164 .2 is X (t ) = C1 cos(ω t ) + C2 sin(ω t ) d2 X (t ) = −ω 2 [C1 cos(ω t ) + C2 sin(ω t ) ] .1 simplifies to m d 2x + kx = 0 dt 2 (B. B. The general solution for Eq. B.3) (B. If we assume that there is no damping and no force acting on the cantilever. almost very equation used in the main text. 4. 1996). B. t ) 2 dt dt (B.1) where m is the dynamic mass of the cantilever. it is obtained (B. Chapters 4 and 5. This is for validation of work and completeness.
given by Eq. (B. (B. Spring constant The general equation for the cyclic frequency of a vibrating body. is 165 .9) which becomes f = 1 h 2π L2 35E 33ρ .8 can be further simplified by substituting Eqs B. One is trivial and the other is equal to ( ) (B.7. B. Equation B.6) which can be simplified into the following form f = 1 2π k m (B.5) ω= k m . (B.7) by using the equality of f = ω 2π .2.10) B.33 to obtain Ebh3 4 L3 33 ρ bhL 140 f = 1 2π (B.24 and B.8) where f is the frequency of the body in motion and ω is the angular frequency. m −ω 2 + k [C1 cos(ω t ) + C2 sin(ω t ) ] = 0 which produces two results.
If we assume a simple cantilevertype structure.1. E is the modulus of the elasticity. is k= F y .1. the spring constant.11) where m is the dynamic mass of the structure and k. Fig. B. the general equation of the elastic curve is d2 y Mx = EI dx 2 . (B.f = 1 2π k m . (B.12) in which F is the force applied to the structure.14) 166 . with an applied force. and I is the moment of inertia. (B. B. Geometric parameters of a cantilever. Fig. (B.13) where Mx is the moment acting along the cantilever. However. Mx can be represented as M x = F ( L − x) .
18) ( ) . B.Inserting Eq.16 becomes y= F 3Lx 2 − x 3 6 EI ( ) . we get the following equation that represents the maximum displacement y = ymax = F FL3 3L3 − L3 = 6 EI 3EI ( ) . B.15 with the boundary conditions of Eq.15) and by intergrating Eq. B. The boundary conditions are dy =0 dx y=0 at at x=0 x=0 .16) In Eq. B.15 we find that dy F = 2 Lx − x 2 dx 2 EI (B. (B. we have two unknowns. B.17) Solving Eqs B. 6 EI 2 (B. B. (B. B. (B.21 into Eq. We can determine them by utilizing boundary conditions for a cantilever.12 we get 167 . B. B.13 and then integrating we get dy F x2 Lx − + C1 = dx EI 2 (B.20 and solve it at x=L.15 we obtain y= F x 2 x3 L − + C1 x + C2 .20) If we now take Eq. Substituting Eq. (B.16. B.17 we find that C1 = 0 and C2 = 0 . B. Eq.18 into Eq.14 into Eq.19) Similarly.21) Inserting the result of Eq.16 and B.
23 is inserted into Eq. i.23) When Eq.1.3.k= F 3EI = 3 ymax L . (B. where the kinetic energy differential dKE can be substituted by dKE = & & y2 y2 dm = ρ bhdx . A vibrating mass will affect these energies depending on the boundary conditions.. B. bh3 I= 12 . 2 2 (B. Equation B.26) & where ( y )2 is the velocity of the vibrating body with displacement ymax based on the boundary conditions. B.26 is assuming that the beam has a constant cross section in that the variables b and h are not changing as a function of position along the 168 .25) (B.22 we finally get k= Ebh3 4 L3 .22) where I is the moment of inertia for the cantilever. B.24) which is the stiffness for a cantilever shown in Fig.e. Dynamic mass To determine the dynamic mass of a vibrating cantilever we must consider the kinetic energies involved. With this in mind we start with the basic equation for a kinetic energy KE = ∫ dKE . B. (B. (B.
27) which translates into 3Lx 2 − x3 y = ymax . 3 2L (B. B. B. Eq.20 and getting the following form y ymax 3Lx 2 − x 3 = 2 L3 . i.30) Solving Eq. and B.32) we find that the quantity in parenthesis in Eq.28) After taking the first derivative with respect to time.26.21 into Eq.31 is equivalent to 33 33 ρ bhL = M =m 140 140 where M is the total. (B.25.e.cantilever. 3 2L (B.29) Using Eqs B. B. (B. we start by dividing Eq.29 and applying limits of integration we get & y 2 3Lx 2 − x 3 KE = ∫ max ρ bhdx .33) 169 . 2 2 L3 0 L 2 (B. (B. Assuming proportionality.31) Since the general equation for kinetic energy is KE = 1 2 mv 2 (B. B..28 becomes 3Lx 2 − x 3 & & y = ymax . B. B.30 we get &2 33 y ρ bhL max KE = 140 2 . static mass of the beam and m is the dynamic mass of the cantilever.
6 140 L ( ) (B.35) and we can equate ( ρ bh ) = C1 1 r5 5r 2 − 35Lr + 63L2 = n(r ) . B. To determine the effect. but modify it slightly to KE = ∫ 0 r &2 ymax 3Lx 2 − x 3 ρ bhdx .30. mass constant along the length of the cantilever.35 as &2 ymax KE = C1 n( x) . it becomes KE = &2 1 r5 ymax ρ bh ) ⋅ 5r 2 − 35 Lr + 63L2 .3.36) where C1 is just a constant of structural properties and n(x) is a function that will produce a the effective. B.B. 2 (B. ( 6 2 140 L ( ) (B. Dynamic mass constant If we look at the result of Eq. If we now solve Eq. B. We can then rewrite Eq. 2 2 L3 2 (B.33 we see that the mass has a constant of 33/140 in front of it.34.1. at an arbitrary point along the cantilever we start with Eq. or dynamic. B. or the constant.34) where the upper limit of integration will be replaced with the parameter r to signify the location along the cantilever that we are interested in.37) 170 . This constant describes the effect that mass will have on the cantilever.
33. ( m + nr r ∆m ) (B. there are a few substitutions that must be made. or boundary condition mass factor for a cantilever. B. First. and ∆m is the added mass.1.. and change in mass into the equation so that it becomes Se = 1 1 2π df 1 k d ∆m 2π m k . n=33/140 for r = 100%) of the added mass that is dependent on the value of r.38) where sensitivity is a measure of normalized frequency over mass with fi equaling the initial resonant frequency.4. B. The parameter r is the percentage coverage of the cantilever with palladium so that it can be directly compared with a fully coated cantilever later and the parameter n is the effective mass constant. In order to solve Eq. 4. mass.4. B.. we make the appropriate substitutions for frequency. i. Sensitivity of an endloaded cantilever In determining sensitivity of a frequencybased sensor we can use Eq. Sensitivity calculations of the cantilever In this section. the equations for the sensitivity of the cantilever will be derived for two cases: a cantilever partially and a completely covered with palladium. S m = lim 1 ∆f 1 ∂f = fi ∆m f i ∂m . ∆m → 0 (B.38 for an endloaded case to determine Se. repeated here for continuity.e.B.e.1. (i. The explicit quantitative solutions are in Section 5.39) where m is the dynamic mass of the cantilever defined in Eq. By 171 .8.
33 1 350 35 Using the values from Table B.assigning the variable r a value of 0. ρ . and using Fig. Equation B. Parametric values used Modulus of elasticity. and the final equation takes the form of Se = −1 k 2 33 M 140 k knr 2 33 M + nr rδ m 33 140 M + nr rδ m 140 . b µm 190 2.1 we find M = ρ bhL = 2. 5.40. L µm Width. ( ) (B.e.1.1.25. GPa Density. Material properties and geometric dimensions used in a sample case to determine sensitivity of the cantilever as a chemical sensor.33g / cm3 35µ m ⋅ 1µ m ⋅ 350µ m = 2.854 ⋅ 10−8 g . (B.169). B. g/cm 3 Thickness.39 can then be solved. E . we can solve for the sensitivity of a 350 µm long cantilever.40) If we assign nominal values to the variables in Eq.41) 172 . h µm Length. we can solve for a sample case and determine the sensitivity for it in order to compare the sensitivity of the partially and completely covered cantilever. Using the values in Table B.13 to determine the effective mass at this value we find the effective mass equal to nr = 71/420 (~0. i. equaling 25% coverage of the cantilever. Table B. derivative taken and solution simplified.
For the value of the spring constant. or in other words the addition of mass will be equal to this value and will not change. Eq. k. B.24 yields k= = Ebh3 4 L3 190GPa ⋅ 35µ m ⋅ (1µ m ) 4 ( 350 µ m ) N m . The ∆m. B.2.42) which will be held as a constant. an arbitrary number shall be picked. of an end loaded cantilever with Eq. B.25 we can determine the sensitivity.e. will be assumed to be 3 orders of magnitude less than the cantilever mass itself or ∆m = 1 ⋅ 10−11 g . 3 3 (B. B.8. as was solved in Section 5.039 Then using the values of nr = 71/420 and r = 0.40 to be Se = −3.43) = 0. or change in mass. added mass. To set a value for the change in mass. The equation for a uniformly distributed load. (B.4.141 ⋅ 109 Hz Hz − kg . (B.which is the mass of the cantilever.. i.44) which is in units of changing frequency per unit frequency – mass. with the same 173 . Sensitivity of a completely coated cantilever In determining the sensitivity of a uniformly distributed polymer over a cantilever we start with Eq.38.
.41 through B. Eq.substitutions for frequency.e.e. i.45 is solved. B. the derivative and resultant equation are simplified to take the final form Sd = −1 k . will take the form of Sd = 1 1 2π k 33 m 140 df 1 d ∆m 2π . mass.39.5. B.47) which is in units of changing frequency per unit frequency – mass. i. Frequency shift for a free.46) which is similar in formulation to Eq.46. 33 2 k k ( m + ∆m ) 2 140 33 33 m ( m + ∆m ) 140 140 (B. loaded cantilever. free and undamped.1 and from Eqs B.45) which contains parameters that were all defined previously.40. If Eq. B.751 ⋅ 1010 Hz Hz − kg . undamped vibration To determine the frequency shift for an ideal. As can be seen. or tip. Using the values of Table B. system the following identity is used 174 . yields Sd = −1. and changing mass as done in Eq. the sensitivity of the distributed load is about one order of magnitude greater than for an end. 33 ( m + ∆m ) 140 k (B.43. B. (B. B.
by substituting Eq.10 in for f1. B.52 can be used to determine the frequency shift due to the absorption of hydrogen.f12 − f 22 f12 − f 22 = f 22 f 22 .7 and realizing that frequency relation incorporates a change in mass as.48) which can be rewritten. (B. (B.49 can then be simplified into f −f f 22 2 1 2 2 k k − m ( m + ∆m ) m + ∆m − m m + ∆m ∆m = = ⋅ = k 1 m ( m + ∆m ) m ( m + ∆m ) (B.52) +1 Equation B.51) f2 = δm m . B.50) and further to f1 δ m f 22 = + 2 f2 m f2 and finally reformulated. by substituting Eq. as 1 h 2π L2 35 Eeq 33 ρeq (B. (B.49) Equation B. 1 2π k 1 k − m 1 2π ( m + ∆m ) 2 2 1 k 2π ( m + ∆m ) 2 2 2 f12 − f 22 = f 22 . 175 .
56) 176 .5. can be written.6.55) and reformulated. with the appropriate frequency substitutions that include both a changing mass and stiffness. undamped system that has a spring constant that changes with absorption. a different formulation will have to be used than the one used in Section B. Frequency shift for a free.48. For this case. B. as 1 2π 2 k 1 − m 1 2π f12 − f 22 = f 22 (k + δ k ) ( m + δ m ) 2 2 2 1 2π (k + δ k ) ( m + δ m ) 2 .10 in for f1. as 1 h f2 = 2π L2 δk 2 1+ Eeq 35 k 33 ρeq δ m 1 1 + m .53) which can then be simplified into k ( k + ∆k ) − m ( m + ∆m ) k ∆m − m∆k ( m + ∆m ) k ∆m − m∆k ⋅ = m ( m + ∆m ) ( k + ∆k ) m ( k + ∆k ) f −f f 22 2 1 2 2 = ( k + ∆k ) ( m + ∆m ) = (B. Eq. (B. by substituting Eq.B. B. undamped vibration with a changing stiffness To determine the frequency shift for a free.54) and further to f12 f12 mk + m∆k f = = = f12 k ∆m − m∆k k ∆m + mk k ∆m + mk +1 m ( k + ∆k ) mk + m∆k 2 2 (B. (B.
60 for its roots produces a quadratic solution. (B. B. i.61) which can then be substituted back into Eq. (B. 2 dt dt (B. B. m d2x dx + c + kx = F ( x.. Eq. B. which when simplified takes the form of s1.7.59 into Eq.60) Solving Eq.e. B.57 giving a solution with two parts 177 . 2 dt dt (B.56 can be used to determine the frequency shift due to the absorption of hydrogen.58 is assumed as X (t ) = Ae st .58) The general solution for Eq. B.Equation B. t ) .59) If we substitute Eq. Free vibration with viscous damping In determining the resonance frequency of a cantilever with damping present we start with the general equation for a body in motion.2 = − c k c ± − 2m 2m m 2 . B. B.57) If we assume that there is damping and no force acting on the cantilever. (B.57 simplifies to m d 2x dx + c + kx = 0 .58 we get ms 2 + cs + k = 0 .
62) In this Thesis. B.64) where A is the amplitude. B.65 will be used in the shifted frequency so that the formulation becomes 178 . B.X (t ) = A1e s1t + A2e s2 t . 1995).65) which indicates that damping alters the resonant frequency by changing the damped angular frequency. The difference is that in substituting in the functions. B. (B.61 is k c − <1 2m m 2 (B. (B. the radical of Eq. For an under damped system.62 for the condition given by Eq. the cantilever system will be operating in air and thus it will be assumed that it resembles an under damped system (Rao.63) and thus the radical will be negative. Eq.1.63 we simplify the equation with trigonometric identities to X (t ) = Ae −( c / 2 m )t cos (ωd t + θ ) . (B. damped system the same form as Eq.7. and ωd is the damped angular frequency ωd = k c2 − m 4m 2 .48 is used. damped vibration To determine the frequency shift for a free. Solving Eq. B. θ is the phase shift. B. t is time. Frequency shift for a free.
69 can be used to determine the frequency shift due to the absorption of hydrogen. it can be put into the following form: k k c2 − − m ( m + ∆m ) 4 ( m + ∆m )2 k c2 − ( m + ∆m ) 4 ( m + ∆m )2 f12 − f 22 = f 22 = 4k ( m + ∆m ) ∆m − mc 2 m ⋅ 4k ( m + ∆m ) − c 2 (B.69) Equation B. 179 . 56.67) and further to f12 f = 4k ( m + ∆m ) ∆m − mc 2 2 2 = f +1 2 1 m 4k ( m + δ m ) − c 2 4k ( m + δ m ) 2 (B.f12 − f 22 = f 22 1 2π 2 k 1 − m 1 2π k c − ( m + δ m ) 4 ( m + δ m )2 2 2 2 2 1 2π k c2 − ( m + δ m ) 4 ( m + δ m )2 2 .68) 4mk ( m + ∆m ) − c 2 and finally reformulated. as 1 h f2 = 2π L2 2 35 Eeq m 4k ( m + δ m ) − c 2 33 ρeq 4k ( m + δ m ) 1 2 . by substituting Eq. (B. B. but after some operation.66 does not simplify as cleanly as Eq.10 in for f1. (B.66) Equation B.
B.10 in for f1.7. B.65. B. incorporating a rewritten Eq. becomes 1 2π 2 k 1 − m 1 2π f12 − f 22 = f 22 ( k + ∆k ) c2 − ( m + ∆m ) 4 ( m + ∆m )2 2 2 2 1 2π ( k + ∆k ) c − ( m + ∆m ) 4 ( m + ∆m )2 2 2 . by substituting Eq. 180 .73) Equation B.48 is used.73 can be used to determine the frequency shift due to the absorption of hydrogen. as discussed in Section 3.70) This equation simplifies much like Eq.71) and further to f12 f = 4 ( k ∆m − m∆k )( m + ∆m ) − mc 2 2 2 = f +1 2 1 m 4k ( m + δ m ) − c 2 4 ( k + δ k )( m + δ m ) 2 (B. (B. Frequency shift for a free. The difference is that there will also be the altering stiffness so the equation.2.2. (B. so it can be put into the following form: k (k + δ k ) c2 − − m ( m + δ m ) 4 ( m + δ m )2 f12 − f 22 = f 22 (k + δ k ) c2 − ( m + δ m ) 4 ( m + δ m )2 = 4 ( k ∆m − m∆k )( m + ∆m ) − mc 2 m 4 ( k + ∆k )( m + ∆m ) − c 2 (B. as 1 h f2 = 2π L2 2 35 Eeq m 4k ( m + δ m ) − c 4 ( k + δ k )( m + δ m )2 33 ρeq 1 2 . 69. B. damped vibration with a changing stiffness To determine the frequency shift for a free vibration system that has damping and a changing spring constant the same form as Eq.72) m 4 ( k + ∆k )( m + ∆m ) − c 2 and finally reformulated.
2002) X= k F ( k − mω 2 ) 2 + c 2ω 2 (B. 2 k − mω (B.76) (B. B. Harmonically excited system with viscous damping In determining the resonance frequency of a cantilever under harmonic excitation we start with the general equation for a body in motion m d2x dx + c + kx = F cos(ω t ) . B. 2 dt dt (B. Using trigonometric relations cos(ω t − θ ) = cos(ω t ) cos(θ ) + sin(ω t ) sin(θ ) sin(ω t − θ ) = sin(ω t ) cos(θ ) − cos(ω t ) sin(θ ) we can simplify Eq.77) (B. (B.78) By solving Eq. The homogeneous solution to Eq. B. we can assume a solution of the form X (t ) = A cos(ω t − θ ) .75 into Eq.75 will not play a part. By substituting Eq.8. it is obtained (Hsu. B.78 simultaneously. (B.79) where the solution for the phase is θ = tan −1 cω . B.B. as it will be zero under equilibrium conditions.74) If we look at the particular solution for this case.75) ( ) X ( k − mω ) sin(θ ) − cω cos(θ ) = 0 2 X k − mω 2 cos(θ ) + cω sin(θ ) = F .76 into two forms .74 we get X ( k − mω 2 ) cos(ω t − θ ) − cω sin(ω t − θ ) = F cos(ω t ) .80) 181 .
ζ = . they can be rewritten as X= k Fmax (1 − r ) + ( 2ζ r ) 2 2 2 (B. 2 1− r (B.If the following equations: ωn = k c . and 2mω n m r= ω ωn (B.82 and B.83 to be functions of the frequency ratio r.79 and B.80.81) are substituted into Eqs B. 182 .82) where X is the amplitude and the phase is θ = tan −1 2ζ r .83) which allows Eqs B.
UNCERTAINTY ANALYSIS In this Appendix the uncertainty calculations used in this Thesis are performed. Uncertainty calculations for the 350 µm long cantilever To determine the overall percentage uncertainty in the resonant frequency for the averaged. Uncertainty in fundamental resonant frequency In this Section the calculations of the uncertainty in frequency for the ideal.1 and include the averaged geometric dimensions measured for these cantilevers. we start with the general relationship for fundamental frequency. and ρ is the density.1) where h is the thickness of the cantilever. The values used for this calculation are listed in Table C. uncoated case. uncoated cantilever are shown.1.1. 4.1.e. C.APPENDIX C. L is the length. E is the modulus of elasticity. (C.23. C. i. based on Eq. f = 1 h 2π L2 35 E 33 ρ .. 183 .
ρ ) . The values used for each of these parameters are listed in Table 5. E .9853 Thickness. C.2) = 11690. L µm 353.2. 184 .1 we get f = = 1 h 2π L2 35 E 33 ρ 35 (190GPa ) g 33 2.2 If the values listed in Table C.2 µ m )2 (C. L . and are also repeated in Table C.4. h µm Length.3 we write the equation for the overall percentage uncertainty as 2 2 2 2 df df df df δ f = δ h + δ L + δ E + δρ dρ dh dL dE 1 2 .4) where δh. C.1 are inserted into Eq. To determine the overall percentage uncertainty in the resonant frequency.1.232 Hz . is f = ( h. and δρ are the uncertainties in each independent parameter defining the resonance frequency of the cantilever. GPa 190 3 Density. The phenomenological equation. Values of parameters used in uncertainty analysis. (C. corresponding to Eq. (C.985 µ m 2π ( 353. δL. E .33 3 cm 1 0. g/cm 2. ρ .33 0. δE.Table C.3) Using Eq. Parametric values used Modulus of elasticity. C. for completeness of this Section. the RSS (rootsumsquares) method is used.1.
8) Substituting Eqs C.4.981 dL df δ E = 92.33 3 cm = 593.5) df δ L = 330.005 g/cm Contributions by the uncertainties of each independent parameter to the overall uncertainty in f are df df 1 h 35 E ⋅δ h δ h = dh dh 2π L2 33 ρ 1 1 35 E = ⋅δ h 2π L2 33 ρ 1 35 (190GPa ) 1 = ⋅0. the overall uncertainty in f is calculated to be 185 .543 .05 µm δ L =5 µm δ E =3 GPa δ ρ =0.05µ m 2 33 2π ( 353. Uncertainty in parameters defining f 3 δ h =0. The uncertainties in each parameter considered. dρ (C.2.8 into Eq.2µ m ) g 2.6) (C.7) (C.291 dE df δρ = 12.Table C. C.5 to C.232 (C.
2 2 2 2 df df df df δ f = δ h + δ L + δ E + δρ dρ dh dL dE
1
2
2 2 2 2 = ( 593.232 ) + ( 330.981) + ( 92.291) + (12.543)
1
2
(C.9)
= 685.673 Hz .
Once the overall uncertainty in the frequency is found, it is possible to determine the percentage uncertainty in f as %δ f = =
δf
f
⋅ 100 (C.10)
685.673 ⋅ 100 11690.232 = 5.865% , which is fairly high.
To discover the source of the uncertainty in Eq. C.10, it is possible to calculate the percent contribution by each parameter listed in Eq. C.3, as
df δ h dh ⋅ 100 %δ f δ h = 2 (δ f ) =
2
( 593.232 ) 2 ( 685.673)
2
⋅ 100
(C.11)
= 74.854%
%δ f δ L = 23.301% %δ f δ E = 1.812% %δ f δρ = 0.033% ,
(C.12) (C.13) (C.14)
186
where it can be seen that the thickness, h, contributes the most with 74.8% to the overall percentage uncertainty in the frequency. It should be noted that all the percent contributions, Eqs C.11 thru C.14, add up to 100%, as they should. Therefore, the most effective way to reduce the overall uncertainty in f is to improve fabrication processes and decrease the δh.
C.1.2. Uncertainty calculations for the 300 µm long cantilever
The determination method of the overall percent uncertainty in the resonance frequency for this shorter cantilever is the same as with the 350 µm long one. Thus, the values from Tables C.1 and C.2 are used with the exception of the length value, which is now shorter at L=303.53 µm long and the explicit solutions are the same as those in Section C.1.1. Using Eq. C.4, the contributions due to uncertainties of each individual parameter to the overall uncertainty in f are df δ h = 803.272 dh df δ L = 521.506 dL df δ E = 124.968 dE df δρ = 16.984 dρ (C.15)
(C.16)
(C.17)
(C.18)
187
Substituting Eqs C.15 to C.18 into Eq. C.4, the overall uncertainty in f is calculated, following the same formulation as C.9, to be
δ f = 965.982 Hz
which is a higher value than that of the 350 µm long cantilever. Once the overall
(C.19)
uncertainty in the frequency is found to 965.982 Hz, the percent overall uncertainty in f is determined to be %δ f = 6.102% , which is higher than with the 350 µm long cantilever. To discover the source of the uncertainty in Eq. C.20, it is possible to calculate the percent contribution of each parameter listed in Eq. C.3 following the same formulation as Eq. C.11, as %δ f δ h = 69.149% %δ f δ L = 29.146% %δ f δ E = 1.674% %δ f δρ = 0.031% , (C.21) (C.22) (C.23) (C.24) (C.20)
where it can be seen that the thickness, h, once again contributes the most with 69.1% , although less than for the 350 µm long cantilever, to the overall percentage uncertainty in the frequency. The total sum of the percentages adds up to 100%, as they should.
188
C.1.3. Uncertainty calculations for the 250 µm long cantilever
The final determination of the overall percent uncertainty in the resonance frequency for the 250 µm long cantilever will be the same as in Sections C.1.1 and C.1.2. Thus, the values from Tables C.1 and C.2 are used with the exception of the length value, which is now even shorter at L=252.99 µm long and the explicit solutions are the same as those in Section C.1.1. Using Eq. C.4, the contributions due to uncertainties of each individual parameter to the overall uncertainty in f are df δ h = 1156.27 dh df δ L = 900.647 dL df δ E = 179.885 dE df δρ = 24.448 . dρ (C.25)
(C.26)
(C.27)
(C.28)
Substituting Eqs C.25 to C.28 into Eq. C.4, the overall uncertainty in f is calculated, following the same formulation as C.9, to be
δ f = 1476.848 Hz ,
(C.29)
which is higher than that of both the 300 and 350 µm long cantilevers. This is because of the shorter length of cantilever D and higher effect the uncertainty in length has. Once the overall uncertainty in the frequency is found to be 1474.848 Hz, the percent overall uncertainty in f is determined to be %δ f = 6.482% , (C.30)
189
they were averaged for an overall percentage. Taking the results of Eqs C.which is higher than both the 300 and 350 µm long cantilever.15% 3 (C. C. and C.15%. To discover the source of the uncertainty in Eq. 190 .482 = 6.10. h.35) and thus the overall percentage uncertainty in the resonant frequency for these cantilevers is 6. C. for a third time contributes the most with 61% to the overall percentage uncertainty in the frequency. it is possible to calculate the percent contribution of each parameter listed in Eq.102 + 6.3 following the same formulation as Eq.11. The total sum of the percentages adds up to 100%. as they should.298% %δ f δ L = 37. C.20.32) (C. Average overall percentage uncertainty in frequency Once all the overall percentage uncertainties were found for each cantilever of different length.1.865 + 6.34) where it can be seen that the thickness. (C. as %δ f δ h = 61. C.191% %δ f δ E = 1.10.027% . C.484% %δ f δρ = 0.4.33) (C.31) (C.30 we obtain 5.
and contributing variables uncertainties will be calculated for an ideal cantilever of L=350 µm long.1 0.3 presents the values used for this analysis. Uncertainty in frequency for a free vibrations system The uncertainty in frequency for a free.2. Density Silicon.2.001 1.C.3.995 3 Modulus of elasticity E (GPa) 190 121 116 172.1. The concentration will not be considered for anything above 1%.507 3.1 through C.11.023 4.4 discuss the uncertainty in the frequency for several different cases for a palladiumcoated cantilever absorbing a 1% hydrogen concentration at 20˚C and 1 atm pressure.1 through 5.2.9 Thickness h (mm) 1 0. Si Palladium. h=1 µm thick. Material properties and thickness of the composite cantilever.11. undamped vibrating system will be determined in the equation for the frequency shift. Equation 5. Frequency uncertainty in a palladium coated cantilever Sections C.2. which relates to the frequency shift due to hydrogen absorption.11. Table C. Pd Titanium.4 the overall uncertainty. is 191 .101 C. For each case as described in Sections 5. Ti Equivalent values r (g/cm ) 2.330 12. percent uncertainty. Table C. and b=35 µm wide.
Using Eq. 20˚C. ∆m.25 ⋅ 10 0.1. Thus.7. based on Table 4.621 10000 3 2 (C.621 10000 3 2 10000 absorption _ multiplier = 1. and ρeq is the equivalent density as defined in Eq.25 converts the atomic ratio into the equivalent multiplier factor of the volume of hydrogen in the palladium in terms of the concentration. The values for the equivalent modulus and density can be found in Table C. density of hydrogen at 1% concentration. Eeq is the equivalent modulus as defined in Eq. ppm absorption _ multiplier = 1.37) where it is desirable to write each component in terms of the concentration in units of ppm.9. The increase in mass. 4. 4.25 ⋅ 10 0.6. (C. 4.36) where ∆m is the increase in mass. can be written as ∆m = ( absorption _ multiplier ) H _ density V 2 ( )( ) Pd .38) absorption _ multiplier = 776.1 h 2π L2 f2 = 35 E 33 ρ eq eq ∆m +1 m (C. m is the initial dynamic mass.01 atm partial pressure) is written as 192 .3. and 1 atm (0.
and p and T are the partial pressure (in atm) and temperature (in °K) of the gas.001) µ m − 12. ppm⋅molecular _ weight ⋅ 273⋅ p H 2 _ density = m3 ⋅T ⋅1atm 0. respectively. The last part of Eq. Since the concentration is directly related to the partial pressure. (C.001) µ m⋅(1+ 2⋅0. C. it can be written in terms of concentration as p= ppm ⋅ 1atm .236 µ m3 . ( )( )( L + h ) − V Ti = ( 35 + 2⋅0.0224 mol g ⋅( 273K )⋅( 0.41) VTi = b + 2h h + 2h Ti Ti = 25. in this case H2 and N2. is equal to 1 atm.37. 250 µ m3 . 106 (C. VPd.0224 mol H 2 _ density = 8. molecular_weight is the molecular weight of the diatomic hydrogen. (C.001) µ m⋅( 350 + 2⋅0.01atm ) 10000 ppm⋅ 2⋅1. is the volume of palladium deposited on the cantilever and it can be written as a function of just the cantilever dimensions and the layer thickness of Pd and Ti as V = bhL = 35µ m ⋅ 1µ m ⋅ 350 µ m = 12.00794 mol H 2 _ density = 3 m ⋅( 293K )⋅1atm 0.385 kg m3 . p.39) where ppm is the gas concentration.42) 193 .40) when the total pressure of the gas mixture. 250µ m3 (C.
= 14.001 + 0. h. h .001 + ⋅0.1) ) µ m ⋅ (1 + 2 ( 0.1) ) µ m • ( 350 + ⋅0.1) µ m = 14. and VPd = V − V +V t Ti (C. Using the values listed in Table C.3. a concentration of 1% hydrogen can be found to be equal to ∆m = ( absorption _ multiplier ) H _ density V 2 . and C.25 ) 8.44 and will have five uncertainty contributing variables aside from being in terms of ppm. will be comprised of Eqs C. Thus we will write ∆m as ∆m = ∆m b.Vt = ( b + 2 ( hTi + hPd ) ) ( h + 2 ( hTi + hPd ) ) ( L + hTi + hPd ) = ( 35 + 2 ( 0. C.39.755 µ m .43) ( ) ( ) (C.001 + 0. respectively. ∆m.46) kg .38. kg = ( 776. As can be seen the increase in mass.519 µ m3 ) .652 ⋅ 10 11 (C.45) to show which parameters ∆m depends on. h Ti ( Pd ) ( )( ) Pd (C.250µ m3 = 2538.519 µ m3 with hTi and hPd being the thickness of the titanium and palladium layer.44) .385 3 ( 2538.236µ m3 +12. m = 1.813. which is quite large compared to 194 . L.755µ m3 − 25.813.
h. E .507 3 ⋅ 25.51) based on which the frequency shift equation in its phenomenological form is 195 . The mass and equivalent modulus and density are functions of other variables themselves. = 2. L. h . h .48 to C. ρ . h Ti (C. L.49) and ρ eq = ρ eq b. E Ti Si Ti ( Pd ) . h. ρ ) ∆m ( b.36 will contain many more parameters. E .48) Eeq = Eeq b. C. ρ . h . C.47) ( Pd ) Pd . (C. h . h . ρ Ti Pd Si Ti ( Pd ) .36 can be written in terms of all the variables that comprise it as 1 h 2π L2 f2 = 35 E ( b. ρ . h. They can be written as m = m b. h . h. The uncertainty analysis of Eq. h ) eq Ti Pd Si Ti Pd eq Ti Pd Si Ti Pd Ti Pd Ti Pd (C. 250µ m3 + 4.36 will be very similar to that of Section C.236µ m3 + 3 cm cm g 12.m = ρV + ρ Pd VPd + ρTiVTi g g ⋅ 12. (C. (C. ρ . Eq. h .h . E ) 33 ρ ( b.50.33 which is the mass for the entire cantilever. L. L. h .h.h .519 µ m3 cm = 5. E . L. C. h . L . h . h ) +1 m ( b. h. E . L .50) Thus. based on Eqs C.1 except Eq.918 ⋅ 1011 kg .023 3 ⋅ 2538.
C. (C. h. ρ . C.268 Hz for this case. h . E .995 3 cm (C.4. L. C. ρ Ti Pd Si Ti Pd Si Ti ( Pd ) .51.53) f2 = 1. 196 .54 are listed in Table C.652⋅10−11 kg +1 5. it is now possible to write the overall uncertainty equation 2 2 df 2 2 df 2 df 2 df 2 δ f2 = δb + δ h + δ L + δ hTi + dL dhTi db dh 2 df 2 df df δ hPd + 2 δ ESi + 2 δ ETi + dhPd dESi dETi df 2 df df δ EPd + 2 δρ Si + 2 δρTi + dEPd d ρ Si d ρTi 2 df 2 δρ Pd d ρ Pd 1 2 2 2 2 2 2 2 (C.51 we get a resonant frequency of 1 h 2π L2 f2 = 35E 33ρ eq eq ∆m +1 m 1 1µ m 2π ( 350 µ m )2 35(172.f 2 = f 2 b. The values used for each of the parameters appearing in Eq.52) If the values listed in Table C. Using Eq. ρ . where each parameter will individually be taken into account to determine its contribution to the overall uncertainty.54) . h .3 are inserted into Eq. E . E .9GPa ) g 33 3.918⋅10−11 kg f 2 = 7783.
806 db df 2 δ h = 488.56) (C.218 dh df 2 δ L = 222.33 dL df 2 δ h = 1.0005 µm δ b =3 µm δ E Ti =3 GPa δρ Ti =0. C.4.243 Si dE Si df 2 δ E = 0.60) (C.0005 µm  δ E Pd =3 GPa δρ Pd =0.5. Uncertainty in parameters defining f δ h =0. The uncertainty values used based on current industry practice.58) (C.248 (C.61) 197 .005 g/cm 3 Contributions due to uncertainties of each individual parameter to the overall uncertainty in f.005 g/cm3 3 δ h Ti =0.55) (C. are df 2 δ b = 5.59) df 2 δ E = 46.256 Ti dE Ti (C.Table C.05 µm δ L =5 µm δ E Si =3 GPa δρ Si =0.4 and the steps shown in Eq.353 Ti dh Ti df 2 δh Pd dh Pd = 10.005 g/cm δ h Pd =0.57) (C. using the values in Table C.
for this case is %δ f 2 = 6.1. each variable listed in Eq. The percent contribution by uncertainties of each independent parameter. as δ f 2 = 539. following similar steps as in Eq. is (C.653 .106 Hz .926% .62) df 2 δρ = 3. C. df 2 δE Pd dE Pd = 23. C.628 (C. C.149 Si dρ Si df 2 δρ = 0.65) Substituting Eqs C.54 is evaluated and the percent contributions are determined. following similar steps as Eq. C.11.63) (C. which is similar to the simplified case in Section C.10.67) 198 .54.55 to C. (C. the overall uncertainty in f is calculated. following similar steps as in Eq.64) (C. To discover the contributing factors to this percent uncertainty.66) hydrogen.65 into Eq. Thus the overall percentage uncertainty in the frequency. which is the deviation possible in the resonance frequency for a system absorbing (C.006 Ti dρ Ti df 2 δρ Pd dρ Pd = 0. based on the values of the parameters used in the uncertainty analysis. C.9.
003% %δ f δρTi = 1.72) (C.2.36. as they should.73) (C. C. The total sum of the percentages adds up to 100%.465 ⋅ 10 % .448 ⋅ 10−8 % %δ f δρ Pd = 1.74) (C.012% %δ f δ h = 82.001% %δ f δ hPd = 1.70) (C. contributes the most with 82% to the overall percentage uncertainty in the frequency.012% %δ f δ L = 17.736% %δ f δ ETi = 2. Equation 5. 4 where it can be seen that the thickness.12 relates the frequency shift due to hydrogen absorption as 199 .%δ f δ b = 0.248 ⋅ 10 % 5 (C. C. h. Uncertainty in frequency for a free vibrations system with changing stiffness For a free.05% %δ f δ ESi = 0.69) (C.75) (C.77) (C. undamped vibrating system that has a spring constant that changes the equation for the frequency shift will be slightly different that that one of Eq.192% %δ f δρ Si = 0.2.68) (C.008% %δ f δ hTi = 0.76) (C.78) %δ f δ EPd = 0.71) (C.
This function takes a complex form and for a concentration of 1% becomes ∆Eincr ppm 2 π − 0.621 − 0. Eq. If Eq.049 10000 100 0.915 ⋅ 121(12. 5. a function was created to fit along the curve.57 cos −0. 4.915 ⋅ 121)(14. ppm.57 5 = 1+ c os ⋅ 0.24 ) + 190 ⋅ 116 ( 2538. To determine the stiffness for this case.915 ⋅ 121( 25. Using Fig.813.57 5 = 1+ c os ⋅ 0. 250 ) + 190 ⋅ 0.621 5 π − 0.6.05 100 0. C. or 10000 ppm. Eq. B.6.81) 116 ⋅ 0.621 − 0.79) where ∆k is the change in the damping coefficient.80 is substituted into Eq. 4.915 and fits into the equivalent modulus equation. where the influence of the change in the modulus is put into concentration terms. The change in the stiffness will come from the ability for the modulus of elasticity of palladium to change as it absorbs hydrogen.4 and solved for the case of 1%.621 ppm 2 π − 0.80) ( ESi ⋅ ETi ⋅ ∆Eincr EPd )Vt ETi ∆Eincr EPd VSi + ESi ∆Eincr EPd VTi + ESi ETiVPd (190 ⋅ 116 ⋅ 0.621 = 0.05 − 0.43 is modified so that 200 .05 − 10000 100 0.3 GPa . hydrogen it becomes E∆eq = = (C.52 ) = 167. 1 h f2 = 2π L2 35 Eeq 33 ρ eq 1 2 ∆k 1 + k ∆m 1 + m (C.76 ) (C.
1) µ m 4 ( 350 + 0.124 .124 ⋅ 2 ⋅ 0. 4.1) µ m (1 + 2 ⋅ 0.83) where the increase in thickness due to 1% hydrogen is accounted for. and length. as a function of concentration ppm 0.1) µ m N m .9GPa ( 35 + 2 ⋅ 0. To determine the change in the stiffness.82 as k∆ = = = E∆eq bt ht 3 E∆eq ( b + 2hTi + 2hPd ) 4 ( L + hTi + hPd ) 3 4 Lt 3 ⋅ ( h + 2hTi + h∆ 2hPd ) • 3 167. following Fig.001 + 2 ⋅ 0. limited only to the thickness. m 201 .1) µ m 4 ( 350 + 0.1) µ m N = 0.062 where bt.7.001 + 0. ht.621 h∆ = 1 + 5 10000 10000 0. 2 2 (C. respectively.063 .84) (1 + 2 ⋅ 0. thickness. It is then possible to calculate the shifted modulus of elasticity with a modified Eq.k= = Eeq bt ht 3 4 Lt 3 Eeq ( b + 2hTi + 2hPd )( h + 2hTi + 2hPd ) 4 ( L + hTi + hPd ) 3 3 3 = 172.001 + 2 ⋅ 0. and Lt are the total width.82) = 0.001 + 2 ⋅ 0.621 = 1+ 10000 5 = 1.3GPa ( 35 + 2 ⋅ 0. 3 (C. C.001 + 0. it will be necessary to include the changing volume. ∆k.001 + 1.1) µ m 3 3 (C.
= 0. C. E .995 3 cm 1 2 (C.86) The phenomenological relation for Eq.062 m m N . Thus.2.. h .85 we find that the frequency for Eq. L. ρ Ti Pd Si Ti Pd Si Ti ( Pd ) (C. C.1.To determine the change in stiffness the difference between Eqs C.82 and C. ρ . as 202 .063 (C.9GPa ) g 33 3. C. E . C.523Hz .85) If we use the same values listed in Table C. ρ .3 and including Eqs C.4 and the steps as in Eq.001 m = 0. h.062 N m 1. f 2 = f 2 b.e.86 is the same as in Section C.5. i.918 ⋅ 10 kg 1 1µ m f2 = 2π ( 350 µ m )2 f 2 = 7894.79 is equal to 2 1 h f2 = 2π L2 35 Eeq 33 ρeq 1 ∆k 1 + k ∆m 1 + m N 0.84 and 82 is calculated as ∆k = k ∆ − k N N − 0. h .54 and determine the contributions due to uncertainties of each individual parameter to the overall uncertainty in f. we can use the same overall uncertainty given by Eq.652 ⋅ 10−11 kg 1 + −11 5. E . 35 (172. using the values in Table C.87) where the change in the stiffness does not add any additional variables.001 m 1 + 0.
036 dL df 2 δ h = 1.007 Ti dρ Ti (C.556 Ti dh Ti df 2 δh Pd dh Pd = 9.89) (C.95) df 2 δρ = 3.584 Si dE Si df 2 δ E = 0.54 (C.88) (C.172 db df 2 δ h = 492.96) df 2 δρ = 0.682 dh df 2 δ L = 229.90) (C.244 Si dρ Si (C. df 2 δ b = 6.91) (C.94) (C.257 Ti dE Ti df 2 δE Pd dE Pd = 26.002 (C.93) (C.92) df 2 δ E = 46.97) 203 .
106) 5 %δ f δ hPd = 0.100) which is very similar to the result of Eq.100. C.101) (C.728% %δ f δ ETi = 2. C.91% %δ f δ ESi = 0. C.67.54.004% 204 . each variable listed in Eq.105) (C.88 to C.98 into Eq. (C.109) %δ f δ EPd = 0. 2 (C.102) (C.108) (C.103) (C.98) Substituting Eqs C.107) (C. C. Thus the overall percentage uncertainty in the frequency for this case is %δ f = 6.405% %δ f δ L = 17.104) (C.223 ⋅ 10 % (C.227% %δ f δρ Si = 0. The percent contribution of each parameter is %δ f δ b = 0.013% %δ f δ h = 81.54 is evaluated and the percent contribution is determined.06 Hz . which is the deviation possible in the resonance frequency for a system absorbing hydrogen. To discover the contributing factors of the percent uncertainty given by Eq.99) (C.672 .917% . df 2 δρ Pd dρ Pd = 0.116 ⋅ 10 % 4 (C. the overall uncertainty in f is calculated as δ f 2 = 546.593% %δ f δ hTi = 8.
111) %δ f δρ Pd = 1. B.4% to the overall percentage uncertainty in the frequency.515 ⋅ 10 % .498 ⋅ 10 % 8 (C. The damping coefficient.113) which shows that it is a function of the variables µ. h. see Eq. 4 where it can be seen that the thickness. Eq. b. L.110) (C. C. 4.112) where c is the damping coefficient.8µ m N −s = 1. The new equation that relates the frequency shift due to hydrogen absorption.988 ⋅ 10−8 m (C. contributes the most with 81. Uncertainty in frequency for a free vibrations system with damping For a free. is 1 h f2 = 2π L2 2 35 Eeq m 4k ( m + ∆m ) − c 2 33 ρeq 4 k ( m + ∆m ) 1 2 (C.2.77 ⋅ 10−11 ( 350µ m )( 35µ m ) µm − s = 21. and H and we can write it in its phenomenological form as 205 . as they should.3. is equal to c= 2µ Lb H kg 2 1. The total sum of the percentages adds up to 100%.%δ f δρTi = 1.69.25. damped vibrating system the equation for the frequency shift will be different as the angular frequency is now different.
H Ti Pd Si Ti Pd Si Ti Pd ( ) . the overall uncertainty equation is going to change to take them into account and thus the equation takes the form of 206 .652 ⋅ 10−11 kg ( ( ) ) (C. L . h.995 3 cm 1 2 2 1µ m 1 = 2π ( 350 µ m )2 5.115) where the viscosity. b.c = c ( µ . E .918 ⋅ 10−11 kg − 5. h . L . µ .5 we find that the frequency for Eq.918 ⋅ 10 kg + 1. ( ) ( ) Because of the two additional variables. µ. are listed.988 ⋅ 10−8 m 2 N −11 −11 4 0.112 is equal to 1 h f2 = 2π L2 m mc 2 35 Eeq − 33 ρeq ( m + ∆m ) 4k ( m + ∆m )2 1 35 (172. E .918 ⋅ 10−11 kg + 1.062 5. h . E . ρ . ρ . H.652 ⋅ 10 kg m = 7783. C.116) 2 1 2 N −s 5. ρ .313Hz .112 becomes f 2 = f 2 b. This indicates that there are now two additional variables to the frequency and thus the phenomenological relation based on Eq.3 and additionally in Table C.114) for simplification. and the damping fluid thickness.918 ⋅ 10−11 kg 1. (C. H ) (C. C. If we use the values listed in Table C.9GPa ) • g 33 3.
C.05 14 Contributions due to uncertainties of each individual parameter to the overall uncertainty in f. are df 2 δ b = 5.120) 207 .118) (C. H µm 21.5. using the values in Table C.225 dh df 2 δ L = 222.2 2 2 2 df df df 2 df 2 2 2 δ f2 = δ b + δh + δ L + δ hTi + db dh dL dhTi df 2 df df δ hPd + 2 δ ESi + 2 δ ETi + dhPd dESi dETi df 2 df df δ EPd + 2 δρ Si + 2 δρTi + dEPd d ρ Si d ρTi df 2 df df δρ Pd + 2 δµ + 2 δ H d ρ Pd d µ dH 2 2 2 2 2 2 2 2 2 (C.119) (C.117) .4 and the steps as in Eq. µ kg/µms 1. Parameter values Viscosity of nitrogen.333 dL (C.8 11 δ m =5×10 δ H =0.769×10 Damping fluid thickness. Table C.5. Values of the two additional independent parameters in the damped case.806 db df 2 δ h = 488. 1 2 where each variable will individually be taken into account as before with the viscosity and damping fluid thickness included.
487 ⋅ 103 Ti dρ Ti df 2 δρ Pd dρ Pd = 0.124) (C.244 Si dE Si df 2 δ E = 0. (C.247 (C.121) (C.125) df 2 δρ = 3.652 (C.130) 208 .256 Ti dE Ti df 2 δE Pd dE Pd = 23.785 ⋅ 104 dH .128) df 2 δµ = 4.353 Ti dh Ti df 2 δh Pd dh Pd = 10.122) df 2 δ E = 46.149 Si dρ Si df 2 δρ = 6.126) (C.123) (C.664 ⋅ 104 dµ df 2 δ H = 3.127) (C.628 (C. df 2 δ h = 1.129) (C.
117. each variable listed in Eq.003% %δ f δρTi = 1.137) (C.9.142) 209 .001% %δ f δ hPd = 1.134) (C.012% %δ f δ L = 17. To discover the contributing factors of this percent uncertainty.118 to C.927% .100.248 ⋅ 10 % 5 (C.139) (C.012% %δ f δ h = 82.Substituting Eqs C. The percent contribution of each parameter is (C.140) (C. as δ f 2 = 539. C.192% %δ f δρ Si = 0. following similar steps as in Eq. which is the deviation possible in the resonance frequency for a system absorbing (C.136) (C. which is very similar to the result of Eq.05% %δ f δ ESi = 0.008% %δ f δ hTi = 0.141) 8 %δ f δ EPd = 0.135) (C. the overall uncertainty in f is calculated. Thus the overall percentage uncertainty in the frequency for this case is %δ f 2 = 6.131) hydrogen.138) (C.736% %δ f δ ETi = 2. C.448 ⋅ 10 % (C.113 Hz .54 is evaluated and the percent contribution is determined. C.130 into Eq.133) (C. C.132) %δ f δ b = 0.
The total sum of the percentages adds up to 100%. The final modified equation that relates the frequency shift due to hydrogen absorption.93 ⋅ 10 % . C.484 ⋅ 10 % 11 %δ f δ H = 4. C.5 we find that the frequency for Eq.4. as they should.146) where ∆k.112.3 and C. 11 where it can be seen that the thickness. changing stiffness vibrations system For a free. and c are all included. Uncertainty in frequency for a damped. damped vibrating system that has a spring constant that changes the equation for the frequency shift will be slightly different that that one of Eq.2. If we use the same values listed in Tables C.146 is equal to 210 .143) (C. C.465 ⋅ 10 % 4 (C. h.144) (C. ∆m.%δ f δρ Pd = 1. contributes the most with 82% to the overall percentage uncertainty in the frequency.145) %δ f δµ = 7. is 1 h f2 = 2π L2 35 Eeq 33 ρ eq m 4k ( m + ∆m ) − c 2 2 4 ( k + ∆k )( m + ∆m ) 1 2 (C.
918 ⋅ 10−11 kg + 1.147 is the same as the damped case. L .988 ⋅ 10−8 m N N 4 0. H Ti Pd Si Ti Pd Si Ti Pd ( ) . Thus. 1 h f2 = 2π L2 2 35 Eeq m 4k ( m + ∆m ) − c 33 ρeq 4 ( k + ∆k )( m + ∆m )2 1 2 2 35 (172.062 5. (C. expressed by Eq. C.117 and determine the contributions due to uncertainties of each individual parameter to the overall uncertainty in f. h .148) where all the variables are included. ρ . using the values in Table C. h . h.9GPa ) 1µ m 1 • = 2π ( 350µ m )2 g 33 3. are df 2 δ b = 5. E .652 ⋅ 10−11 kg m m ( ) ( ) ( ) 1 N −s 5.918 ⋅ 10 kg 4 0.001 5.062 + 0. we can use the same overall uncertainty as Eq.918 ⋅ 10 kg + 1. E .652 ⋅ 10−11 kg m m ( ) 2 ( ) 2 2 = 7673. ρ .918 ⋅ 10−11 kg + 1.149) 211 .062 + 0.4 and the steps as in Eq.5.147) The phenomenological form for Eq.001 5. so the phenomenological relationship takes the form of f 2 = f 2 b. ρ .112. E . µ .918 ⋅ 10−11 kg 1. (C. C. C.619 Hz .652 ⋅ 10 kg m − 2 N N 4 0. C.457 db (C.995 3 cm 1 N −11 −11 −11 5.
153) df 2 δ E = 45.056 Si dρ Si (C.599 dh df 2 δ L = 215.154) (C.155) (C.152) (C.150) (C.156) df 2 δρ = 3.372 (C.254 Ti dE Ti df 2 δE Pd dE Pd = 21. df 2 δ h = 483.006 Ti dρ Ti df 2 δρ Pd dρ Pd = 0.633 (C.905 (C.159) 212 .162 Ti dh Ti df 2 δh Pd dh Pd = 10.158) (C.157) df 2 δρ = 0.883 Si dE Si df 2 δ E = 0.151) (C.823 dL df 2 δ h = 1.
163) (C.160) (C. Thus the overall percentage uncertainty in the frequency for this case is %δ f 2 = 6. To discover the contributing factors of this percent uncertainty.54 is evaluated and the percent contribution is determined.167) (C.149 to C. C.696% . C.271 ⋅ 10 % (C.170) 213 . as δ f 2 = 532. df 2 δµ = 4.9. which is very similar to the result of Eq. (C. The percent contribution of each parameter is %δ f δ b = 0.449% %δ f δ hTi = 4.137 Hz .743% %δ f δ ETi = 2.589% %δ f δ L = 16.527 ⋅ 104 dµ df 2 δ H = 3.165) (C.764 ⋅ 10 % 4 (C. following similar steps as in Eq.166) (C.674 ⋅ 104 dH .117.162) hydrogen.169) 5 %δ f δ hPd = 1.161) Substituting Eqs C.161 into Eq.164) (C. which is the deviation possible in the resonance frequency for a system absorbing (C. C.189% %δ f δ ESi = 0. each variable listed in Eq.100. C.168) (C.011% %δ f δ h = 82. the overall uncertainty in f is calculated.
176) %δ f δρ Pd = 1.417 ⋅ 10 % 4 %δ f δµ = 7.768 ⋅ 10 % . contributes the most with 82.4 ⋅ 10 % 8 (C. 11 where it can be seen that the thickness.173) (C.161% %δ f δρ Si = 0. as they should.%δ f δ EPd = 0. 214 . h.238 ⋅ 10 % 11 %δ f δ H = 4.003% %δ f δρTi = 1.175) (C.3% to the overall percentage uncertainty in the frequency.172) (C. The total sum of the percentages adds up to 100%.174) (C.171) (C.
D. Fig. Specifications of the main body of the test chamber.1.APPENDIX D. 215 . TEST CHAMBER In this Appendix the blueprints for the test chamber are shown along with some pictures of the finished component.
Fig.2. Fig.3. Fully disassembled view of the chamber showing all parts. 216 . Specifications of the lid for the main body of the test chamber. D. D.
217 .5. Isometric view of the assembled chamber. Fully assembled view of chamber with PZT inside.Fig. D. Fig. D.4.
1. E. GPa 3 Density. 218 .1 has the values for the modulus of elasticity and the density used. Table E. Parameter values used Modulus of elasticity.33 Fig. Schematic of the cantilever and the dimensions used for the data analysis. g/cm 190 2. it was possible to calculate the theoretical resonant frequency for each cantilever based on Eq. Figure E.1 illustrates the guide for the measurements taken and labels used while Table E. b1. 4.APPENDIX E. represent by the shaded part in Fig. E . A=L2 x b.23. ρ .1. and b2 are used to determine the area. ANALYTICAL FREQUENCY DATA Using measured dimensions of the cantilevers. E. The parameters L1.1. L3. Material properties of silicon used for the analytical calculations.
L3 mm Added area. A mm Eqv.007 µ m 2π ( 304.75 8. b2 mm Point width. where the values for each parameter and result for each cantilever are listed in Table. L3 mm Added area.75 9.5 9.5 36 1 14 11 2 4. L2 mm Resonance freq.5 1.5 36 1. L mm Width.106 16116.5 1 13 11.5 1 14.25 8.106 µ m )2 (E. b2 mm Point width.5 321.007 15 11 3 3 332. L2 mm Resonance freq.5 2 318.5 2.116. f Hz Measured characteristics Length.5 10. b mm Thickness.. b mm Thickness. length.2.5 1.5 4.5 36. h mm End width. b1 mm Offset. b1 mm Offset.5 355. Geometrical and resonant frequency data of the cantilever utilized.895 219 .875 8.410 D12 Typical 295 36.882 11801.660 Hz .177) f1 = 16. L mm Width. length. A mm Eqv. L1 mm End length.Taking cantilever D12 as an example.5 4.2. L1 mm End length. Table E.5 308.434 F11 Typical 246.660 E11 Typical 345 35. h mm End width.375 9.33 3 cm 1 1.938 15865.445 22772.252 E12 Typical 345.060 F12 Typical 244 36 1. the resonance frequency is f1 = f1 = 1 h 2π L2 35 E 33 ρ 35 (190GPa ) g 33 2. E. Measured characteristics Length.007 18 11 3.597 23175..007 15 11 4 4 345. f Hz 2 2 D11 Typical 296.982 11812.
989 13 8 3 3.3125 8.989 13 9 2 3.5 4 273. L1 mm End length.5 1 15 10 2 3. A mm2 Eqv. f Hz D31 Typical 295 35 0. L3 mm Added area..700 15405.118 16269.5 296. L mm Width. L3 mm Added area.5 7. length.5 7.5 34 1 14. b mm Thickness.96 15 9 2 3 255. b mm Thickness. b1 mm Offset.Measured characteristics Length.96 15 9 3 3.660 220 .949 E31 Typical 346 35 0.5 34 1 14.. L mm Width. h mm End width. b2 mm Point width. L2 mm Resonance freq. L1 mm End length. b2 mm Point width. h mm End width.294 24105.5 253.5 9. h mm End width.25 6. b1 mm Offset.5 2.5 2 3 271.402 11377. f Hz D22 Typical 296 35 0. L mm Width. A mm2 Eqv. length. A mm2 Eqv. length.854 F21 Typical 239.048 12044.573 Measured characteristics Length.236 11731.721 22919.758 F31 Typical 246 35 0.5 269.25 7. f Hz D21 Typical 293.757 E21 Typical 342.96 15 8 3 2 225 6. L2 mm Resonance freq. L3 mm Added area.5 36..625 8.818 22230.5 0. b2 mm Point width. L1 mm End length.5 10 2 4 282 8.5 235.567 E22 Typical 346 34.989 15 9.375 7.757 15969. b mm Thickness. L2 mm Resonance freq.028 F22 Typical 246 33 0. b1 mm Offset.383 Measured characteristics Length.
length.643 11378. L2 mm Resonance freq.956 15 10 2 3.5 7.625 8. h mm End width.956 15 10 3 4. A mm2 Eqv. L3 mm Added area. f Hz D32 Typical 295 36 0.224 E32 Typical 345 35 0..088 21576.252 221 .956 15 9 2.5 308.Measured characteristics Length. b2 mm Point width. L1 mm End length. b mm Thickness.466 F32 Typical 248 37 0.573 15354.25 8. L mm Width. b1 mm Offset.5 4 267.5 299.
This action might not be possible to undo. Are you sure you want to continue?
arecco will be available on