USF Libraries
USF Digital Collections

Oganic vapor sensing using high frequency thickness shear mode resonators

MISSING IMAGE

Material Information

Title:
Oganic vapor sensing using high frequency thickness shear mode resonators
Physical Description:
Book
Language:
English
Creator:
Williams, Randolph
Publisher:
University of South Florida
Place of Publication:
Tampa, Fla.
Publication Date:

Subjects

Subjects / Keywords:
Thickness shear mode resonators
Vapor sensing
QCM
QCR
Dissertations, Academic -- Chemical Engineering -- Masters -- USF   ( lcsh )
Genre:
government publication (state, provincial, terriorial, dependent)   ( marcgt )
bibliography   ( marcgt )
theses   ( marcgt )
non-fiction   ( marcgt )

Notes

Summary:
ABSTRACT: Thickness shear mode (TSM) sensors, also known as quartz crystal micro-balances (QCM) are a class of acoustic wave sensors that have been used for gas/vapor sensing. Fast and sensitive chemical vapor sensing, specifically of hydrocarbon vapors is an important application for these vapor sensors. The TSM sensors typically used have a lower sensitivity compared with other acoustic wave sensors. This thesis describes the development of high sensitivity organic vapor sensors using thin polymer film coatings on TSM devices. Commercially available AT-quartz TSM devices were milled leaving a thin quartz membrane surrounded by a thicker outer ring. This resulted in an increased frequency and a consequent increase in sensitivity, as described by the Sauerbrey equation. The TSM sensors were then coated with thin sensing films of rubbery polymers. Isothermal experiments at room temperature were conducted.A fully instrumented and automated test bed consisting of a temperature-controlled organic vapor dilution system, a precision impedance analyzer, and computer based data acquisition was developed and used to evaluate the performance of the coated TSM devices. The TSM devices compared in this study were AT cut with fundamental resonant frequencies of 10, 20, and 96 MHz. The results of tests conducted are presented to demonstrate increase in sensitivity for higher fundamental frequency TSM devices. 96 MHz TSM resonators were found to be 8 to 27 times more sensitive than 10 MHz resonators. Sensitivity was limited by the difficulty in coating sensing layers and damping of the resonator. Additionally, each sensor was evaluated and compared in terms of detection limit and noise level. 96 MHz resonators had higher noise levels than 10 MHz or 20 MHz resonators; as a result, 96 MHz resonators did not show significant improvements in LOD.Also, response times for 96 MHz resonators were quicker than 10 MHz or 20 MHz resonators and response times generally decreased with analyte concentration. Several rubbery polymer films as well as copolymers were investigated to determine which sensing film would have the optimal performance in terms of response time, recovery, reproducibility, repeatability, frequency noise, and baseline drift. The organic vapors studied were benzene, toluene, hexane, cyclohexane, heptane, dichloroethane, and chloroform at levels ranging from 0.2 to over 13.7 volume percentage in nitrogen gas. The Butterworth-VanDyke (BVD) equivalent circuit model was used to model both the perturbed and unperturbed TSM resonator. Monitoring the sensor response through the equivalent circuit model allowed for discriminating between the organic vapors. Vapor discrimination, in turn, depended upon the changes in the resistance parameter.
Thesis:
Thesis (M.S.)--University of South Florida, 2005.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: World Wide Web browser and PDF reader.
System Details:
Mode of access: World Wide Web.
Statement of Responsibility:
by Randolph Williams.
General Note:
Title from PDF of title page.
General Note:
Document formatted into pages; contains 114 pages.

Record Information

Source Institution:
University of South Florida Library
Holding Location:
University of South Florida
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
aleph - 001670375
oclc - 62311528
usfldc doi - E14-SFE0001243
usfldc handle - e14.1243
System ID:
SFS0025564:00001


This item is only available as the following downloads:


Full Text
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001670375
003 fts
005 20051216093331.0
006 m||||e|||d||||||||
007 cr mnu|||uuuuu
008 051121s2005 flu sbm s000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0001243
035
(OCoLC)62311528
040
FHM
c FHM
1 100
Williams, Randolph.
0 245
Oganic vapor sensing using high frequency thickness shear mode resonators
h [electronic resource] /
by Randolph Williams.
260
[Tampa, Fla.] :
b University of South Florida,
2005.
502
Thesis (M.S.)--University of South Florida, 2005.
504
Includes bibliographical references.
516
Text (Electronic thesis) in PDF format.
538
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
500
Title from PDF of title page.
Document formatted into pages; contains 114 pages.
520
ABSTRACT: Thickness shear mode (TSM) sensors, also known as quartz crystal micro-balances (QCM) are a class of acoustic wave sensors that have been used for gas/vapor sensing. Fast and sensitive chemical vapor sensing, specifically of hydrocarbon vapors is an important application for these vapor sensors. The TSM sensors typically used have a lower sensitivity compared with other acoustic wave sensors. This thesis describes the development of high sensitivity organic vapor sensors using thin polymer film coatings on TSM devices. Commercially available AT-quartz TSM devices were milled leaving a thin quartz membrane surrounded by a thicker outer ring. This resulted in an increased frequency and a consequent increase in sensitivity, as described by the Sauerbrey equation. The TSM sensors were then coated with thin sensing films of rubbery polymers. Isothermal experiments at room temperature were conducted.A fully instrumented and automated test bed consisting of a temperature-controlled organic vapor dilution system, a precision impedance analyzer, and computer based data acquisition was developed and used to evaluate the performance of the coated TSM devices. The TSM devices compared in this study were AT cut with fundamental resonant frequencies of 10, 20, and 96 MHz. The results of tests conducted are presented to demonstrate increase in sensitivity for higher fundamental frequency TSM devices. 96 MHz TSM resonators were found to be 8 to 27 times more sensitive than 10 MHz resonators. Sensitivity was limited by the difficulty in coating sensing layers and damping of the resonator. Additionally, each sensor was evaluated and compared in terms of detection limit and noise level. 96 MHz resonators had higher noise levels than 10 MHz or 20 MHz resonators; as a result, 96 MHz resonators did not show significant improvements in LOD.Also, response times for 96 MHz resonators were quicker than 10 MHz or 20 MHz resonators and response times generally decreased with analyte concentration. Several rubbery polymer films as well as copolymers were investigated to determine which sensing film would have the optimal performance in terms of response time, recovery, reproducibility, repeatability, frequency noise, and baseline drift. The organic vapors studied were benzene, toluene, hexane, cyclohexane, heptane, dichloroethane, and chloroform at levels ranging from 0.2 to over 13.7 volume percentage in nitrogen gas. The Butterworth-VanDyke (BVD) equivalent circuit model was used to model both the perturbed and unperturbed TSM resonator. Monitoring the sensor response through the equivalent circuit model allowed for discriminating between the organic vapors. Vapor discrimination, in turn, depended upon the changes in the resistance parameter.
590
Adviser: Venkat R. Bhethanabotla.
653
Thickness shear mode resonators.
Vapor sensing.
QCM.
QCR.
690
Dissertations, Academic
z USF
x Chemical Engineering
Masters.
049
FHMM
090
TP145 (Online)
773
t USF Electronic Theses and Dissertations.
4 856
u http://digital.lib.usf.edu/?e14.1243



PAGE 1

Organic Vapor Sensing Usi ng High Frequency Thickness Shear Mode Resonators By Randolph Williams A thesis submitted in partial fulfillment of the requirements for the degree of Masters of Science in Chemical Engineering Department of Chemical Engineering College of Engineering University of South Florida Major Professor: Venkat Bhethanabotla, Ph.D. Scott Campbell, Ph.D. Babu Joseph, Ph.D. Date of Approval July 11, 2005 Keywords : TSM sensor, QCM, gas sensing Copyright 2005, Randolph Williams

PAGE 2

DEDICATION This dissertation is dedicated to my parents, Jagdat and Hansranie Williams

PAGE 3

ACKNOWLEDGEMENTS I would like to express my gratitude to my major professor Dr. Venkat Bhethanabotla for his support and guidance throughout my res earch experience and graduate studies. I would like to especially thank Stefan Cular, Krishnan Srinivisan and Shibendra Pobi for their technical help. Thanks to Tom Payne and MTronPTI Inc. for supplying inverse mesa resonators. Finally, thanks to all my friends.

PAGE 4

TABLE OF CONTENTS LIST OF TABLES iii LIST OF FIGURES iv ABSTRACT vii CHAPTER 1 INTRODUCTION 1 1.1 Motivation 1 1.2 Acoustic Wave Sensors 2 1.3 TSM Sensors 3 1.4 Thesis Organization 4 CHAPTER 2 TSM CHEMICAL SENSORS: FABRICATION AND OPERATIONAL THEORY 5 2.1 Introduction 5 2.2 Temperature Stability and Pressure Effects 6 2.3 Fabrication Process 8 2.4 TSM Sensory Theory 9 2.5 Electrical Circuit Model 13 2.6 Polymer Film 17 2.7 Performance Criteria 20 2.7.1 Selectivity 20 2.7.2 Reversibility 21 2.7.3 Dynamic Range 21 2.7.4 Stability, Repeatability, and Reproducibility, and Response Time 22 CHAPTER 3 EXPERIMENTAL APPARATUS 24 3.1 Introduction 24 3.2 Static and Dynamic Gas Generation Methods 25 3.2.1 Major Component/Zero Gas/Dilutant/Carrier Gas 26 i

PAGE 5

3.2.2 Dynamic Method Advantages 27 3.2.3 Flow Control 27 3.3 Experimental Apparatus Design 30 3.4 Vapor Dilution System Components 32 3.4.1 Solvent Cell 33 3.4.2 Sensor Cell Designs 34 3.4.3 Automation 36 CHAPTER 4 RESULTS AND DISCUSSION 43 4.1 Introduction 43 4.2 Polymer Film Selection 43 4.3 Experimental Results with Polyisobutylene 47 4.3.1 Coating and Chemicals 51 4.3.2 Sensor Response, Repeatability, and Sensor Parameters 52 4.3.3 Sensitivity 58 4.4 Vapor Discrimination 64 4.5 Temperature Correction 72 CHAPTER 5 CONCLUSION 76 REFERENCES 79 APPENDICIES 83 Appendix A Solvent Concentrations 84 Appendix B Printed Circuit Boards, Lid Designs and Wiring Diagram 87 Appendix C Frequency Responses and Calibration Curves 91 ii

PAGE 6

LIST OF TABLES Table 2.1 Coating Methods 19 Table 3.1 Values for Molecular Correction Factor 29 Table 4.1 Glass Transition Temperature of Polymers 46 Table 4.2 Design Parameters of the TSM Devices 47 Table 4.3 Coating Guide 48 Table 4.4 Film Thickness 50 Table 4.5 Comparison of Experimenal Sensor Response Parameters 56 Table 4.6 Frequency Noise During Benzene Exposure 57 Table 4.7 Sensivity Ratio and Sesnitivity Dependence on Experimental Power 61 Table 5.1 Dynamic Range 77 Table A.1 Solvent Concentrations: Dichloroethane, Heptane, and Chloroform 84 Table A.2 Solvent Concentrations: Benzene and Toluene 85 Table A.3 Solvent Concentrations: Cyclohexane and Hexane 86 Table A.4 Wagner Equation Constants 86 Table C.1.1 Hexane Experimental Sensitivity Power 92 Table C.1.2 Hexane Calibration Coefficients 93 Table C.2.1 Dichloroethane Experimental Sensitivity Power 94 Table C.2.2 Dichloroethane Calibration Coefficients 95 Table C.3.1 Cyclohexane Experimental Sensitivity Power 96 Table C.3.2 Cyclohexane Calibration Coefficients 97 Table C.4.1 Heptane Experimental Sensitivity Power 98 Table C.4.2 Heptane Calibration Coefficients 99 Table C.5.1 Chloroform Experimental Sensitivity Power 100 Table C.5.2 Chloroform Calibration Coefficients 101 Table C.6.1 Toluene Experimental Sensitivity Power 102 Table C.6.2. Toluene Calibration Coefficients 103 iii

PAGE 7

LIST OF FIGURES Figure 2.1 AT Cut Quartz 6 Figure 2.2 AT Cut Frequency vs. Temperature 7 Figure 2.3 Milled TSM Device 13 Figure 2.4 Equivalent Circuit Model of Uncoated TSM Resonator 15 Figure 2.5 Equivalent Circuit Model of Coated TSM Resonator 16 Figure 3.1 Production of Standard Gas Mixtures 25 Figure 3.2 Conceptual Apparatus Design 30 Figure 3.3 Experimental Apparatus 33 Figure 3.4 Bubbler Unit, Fritted Tube, and Fittings 34 Figure 3.5 Sensor Cell Design 1 35 Figure 3.6 Sensor Cell Design 2 36 Figure 3.7 Overall Automation Programs 38 Figure 3.8 Initiation of All Valves 39 Figure 3.9 First Nitrogen Purge Cycle 40 Figure 3.10 Purge Cycle Control Loop 40 Figure 3.11 Cycle Control Loop 41 Figure 4.1 20 MHz Response to Benzene, PVA Film 44 Figure 4.2 10 MHz Response to Benzene, PVP Film 45 Figure 4.3 10 MHz, 20 MHz, and 96 MHz Devices 47 Figure 4.4 Quartz Thickness 49 Figure 4.5 Resonator Sensitivity 50 Figure 4.6 10 MHz Device Response to Benzene Vapors 53 Figure 4.7 20 MHz Device Response to Benzene Vapors 54 Figure 4.8 96 MHz Device Response to Benzene Vapors 54 Figure 4.9 Comparison of TSM Resonator Responses 55 Figure 4.10 Experimental Mass Sensitivity of 10 and 20 MHz TSM Resonators 59 iv

PAGE 8

Figure 4.11 Experimental Mass Sensitivity of 96 MHz TSM Resonator 60 Figure 4.12 Theoretical and Experimental Comparison of Sensitivity 61 Figure 4.13 Device Sensitivity Comparison 63 Figure 4.14 L.O.D. Determination 64 Figure 4.15 Resistance Changes of a PIB Coated 96 MHz Resonator 65 Figure 4.16 Resistance Changes due to Hexane, Cyclohexane and Heptane 67 Figure 4.17 Resistance Changes due to Benene, Hexane and Dichloroethane 67 Figure 4.18 Resistance Changes due to Benzene and Toluene 68 Figure 4.19 Resistance Changes due to Dichloroethane and Chloroform 69 Figure 4.20 Resistance and Frequency Changes due to Sorption of Heptane, Chloroform, Toluene and Cyclohexane 69 Figure 4.21 Resistance and Frequency Changes due to Sorption of Benzene, Hexane and Dichloroethane 70 Figure 4.22 Vapor Discrimination Curve for Benzene, Hexane and Dichloroethane 71 Figure 4.23 Vapor Discrimination Curve for Heptane, Chloroform, Toluene and Cyclohexane 71 Figure 4.24 Activity of Benzene in PIB 75 Figure B.1 PCB Design 88 Figure B.2 High Frequency Resonator Lid 1 89 Figure B.3 High Frequency Resonator Lid 2 89 Figure B.4 Solenoid Valves, Relays, MFCs Wiring Diagram 90 Figure C.1.1 Hexane Calibration Curve 91 Figure C.1.2 Hexane LOD 92 Figure C.2.1 Dichloroethane Calibration Curve 94 Figure C.1.2 Dichloroethane LOD 95 v

PAGE 9

Figure C.3.1 Cyclohexane Calibration Curve 96 Figure C.3.2 Cyclohexane LOD 97 Figure C.4.1 Heptane Calibration Curve 98 Figure C.4.2 Heptane LOD 99 Figure C.5.1 Chloroform Calibration Curve 100 Figure C.5.2 Chloroform LOD 101 Figure C.6.1 Toluene Calibration Curve 102 Figure C.6.2 Toluene LOD 103 vi

PAGE 10

ORGANIC VAPOR SENSING USING HIGH FREQUENCY THICKNESS SHEAR MODE RESONATORS Randolph D. Williams ABSTRACT Thickness shear mode (TSM) sensors, also known as quartz crystal micro-balances (QCM) are a class of acoustic wave sensors that have been used for gas/vapor sensing. Fast and sensitive chemical vapor sensing, specifically of hydrocarbon vapors is an important application for these vapor sensors. The TSM sensors typically used have a lower sensitivity compared with other acoustic wave sensors. This thesis describes the development of high sensitivity organic vapor sensors using thin polymer film coatings on TSM devices. Commercially available AT-quartz TSM devices were milled leaving a thin quartz membrane surrounded by a thicker outer ring. This resulted in an increased frequency and a consequent increase in sensitivity, as described by the Sauerbrey equation. The TSM sensors were then coated with thin sensing films of rubbery polymers. Isothermal experiments at room temperature were conducted. A fully instrumented and automated test bed consisting of a temperature-controlled organic vapor dilution system, a precision impedance analyzer, and computer based data acquisition was developed and used to evaluate the performance of the coated TSM devices. The TSM devices compared in this study were AT cut with fundamental resonant frequencies of 10, 20, and 96 MHz. The results of tests conducted are presented to demonstrate increase in sensitivity for higher fundamental frequency TSM devices. 96 MHz TSM resonators were found to be 8 to 27 times more sensitive than 10 MHz resonators. Sensitivity was limited by the difficulty in coating sensing layers and damping of the vii

PAGE 11

resonator. Additionally, each sensor was evaluated and compared in terms of detection limit and noise level. 96 MHz resonators had higher noise levels than 10 MHz or 20 MHz resonators; as a result, 96 MHz resonators did not show significant improvements in LOD. Also, response times for 96 MHz resonators were quicker than 10 MHz or 20 MHz resonators and response times generally decreased with analyte concentration. Several rubbery polymer films as well as copolymers were investigated to determine which sensing film would have the optimal performance in terms of response time, recovery, reproducibility, repeatability, frequency noise, and baseline drift. The organic vapors studied were benzene, toluene, hexane, cyclohexane, heptane, dichloroethane, and chloroform at levels ranging from 0.2 to over 13.7 volume percentage in nitrogen gas. The Butterworth-VanDyke (BVD) equivalent circuit model was used to model both the perturbed and unperturbed TSM resonator. Monitoring the sensor response through the equivalent circuit model allowed for discriminating between the organic vapors. Vapor discrimination, in turn, depended upon the changes in the resistance parameter. Finally, the vapor liquid equilibrium at the polymer solvent interface was utilized to correct for perturbations, due to temperature changes, in the sensor response. viii

PAGE 12

CHAPTER 1 INTRODUCTION 1.1 Motivation The scientific and industrial sectors require precise measurement tools for the detection as well as measurement of chemical species. Modern analytical methods such a gas chromatography (GC), x-ray fluorescence (XRF), and surface plasmon resonance require sample preparation, long processing times, restrictive sample environment (vacuum), and excessive associated costs. Sensors that can overcome the limitations of these existing analytical methods would have enormous advantage and appeal to the industrial and scientific community. Such sensors should be capable of real time measurements, have high accuracy and lower processing times and can be taken to the sample environment, rather than the sample taken to the analytical tool. Acoustic wave sensors have these capabilities. In general, a sensor is a device which produces an output signal in response to an input quantity. In this case, the input quantity will be the chemical which is being sensed. The sensor can function as a detector or it can be used to determine the quantity of chemical present. Detectors may find potential applications in areas where fault detection is required, such as detecting the explosive limits of hydrocarbon vapors. Sensors used to quantify chemical amount may find potential applications as process flow 1

PAGE 13

monitors. Additionally, the sensor may be used to identify and distinguish between different chemical species and characterize the species by determining its properties. In summary, chemical sensors have the potential to be employed for the detection of chemical species, the measurement of chemical species, and the characterization of material properties. In many cases all three functions are performed simultaneously by acoustic wave chemical sensors, adding to the advantage of these sensors over typical analytical techniques. Currently there are many classes of sensors, such as electrochemical, acoustic, optical, and thermal sensors. In this work we consider an acoustic wave sensor for the detection and quantification of chemical species. 1.2 Acoustic Wave Sensors Since acoustic wave sensors have a detection mechanism that is mechanical, or acoustic, they are named acoustic wave sensor. Any changes to the characteristics of the propagation path affect the velocity or amplitude of the wave as the wave propagates through the surface of the sensor material. Changes in the resonant frequency of the sensor can then be correlated to the corresponding physical quantity, such as mass, being measured. Almost all acoustic wave devices and sensors use a piezoelectric material to generate the acoustic wave. Acoustic wave devices are also described by the mode of wave propagation through a piezoelectric substrate. One such device, which will be improved in this thesis work, is the thickness shear mode resonator (TSM). Thickness shear mode resonators are bulk acoustic wave devices (the wave propagates through the substrate). The TSM device is also referred to as a quartz crystal microbalance (QCM) or quart crystal resonator (QCR). It is the best-known, oldest, and simplest acoustic wave 2

PAGE 14

device. TSM devices are also known as resonators because the crystal resonates as electromechanical standing waves are created. TSM devices are also very sensitive to surface interactions. The TSM resonator was originally used to measure metal deposition rates in vacuum systems. In the late 1960s, the TSM resonator was shown to operate as a vapor sensor. In this thesis work, the TSM resonator will be used as an organic vapor sensor. 1.3 TSM Sensors Since there are many acoustic wave devices capable of sensing for organic vapors, the choice of the TSM device must be justified. Sensors, in general, must meet certain performance criterion. The most important criterion are the sensitivity of the device to the chemical, the ability of the sensor to distinguish between different chemicals (selectivity), the resolution of the measurement, the reliability of the measurement, and the robustness of the device. In terms of robustness, the TSM device features simplicity of manufacture, ability to withstand harsh environments, temperature stability, and good sensitivity [1]. However, TSM devices have the lowest sensitivity of the acoustic wave sensors. Typical TSM resonators operate between 5 and 30 MHz. TSM devices can operate at higher frequencies, but these sensors are fragile devices that are difficult to manufacture and handle. Recent work has been done to manufacture high-frequency TSM resonators using piezoelectric films and micromachining techniques [2]. In this thesis work, the TSM device operates at frequencies close to 99 MHz, making these TSM sensors a lucrative alternative to other acoustic wave devices because of the increased sensitivity to organic vapors. These sensors were still robust and capable of 3

PAGE 15

producing repeatable results. Seven organic solvents of investigated for their detection, quantification, and distinction. 1.4 Thesis Outline This thesis contains five chapters. Chapter 2 presents the background information needed for understanding how TSM sensors work and how to exploit their parameters for organic vapor sensing and vapor discrimination. Chapter 3 presents the methods of generating standard gas mixtures, and the design of the experimental apparatus based upon these techniques. The experimental results and discussion are presented in chapter 4, and chapter 5 contains a summary of the experimental results and suggestions for improvement and future work. 4

PAGE 16

CHAPTER 2 TSM CHEMICAL SENSORS: FABRICATION AND OPERATIONAL THEORY 2.1 Introduction Thickness shear mode devices are a class of acoustic wave sensors based on piezoelectric crystals. The term piezoelectricity was derived from the Greek word piezein, which means to press. A piezoelectric quartz crystal device will produce an electrical output when a mechanical pressure is applied. This piezoelectric effect was discovered by Pieere and Jacques Currie in 1880 [3]. Piezoelectric materials, like quartz, when acted upon by some outside force will shift electrical charges out of their static position and create an electric potential difference [4]. This potential difference will exist as long as the magnitude of the force is changing. When the motion stops, electricity is no longer produced. If a force is applied in the opposite direction, the electrical polarity will be reversed and output from the crystal will be an alternating voltage which changes at the same rate as the applied force changes direction. TSM devices are also known as quartz crystal microbalances (QCMs); however, the term TSM is more applicable due to fact that it describes the mode of operation as being the thickness shear mode. A TSM device becomes a chemical sensor by the addition of a foreign film or sensing layer onto the surface of the TSM device. The following discussion will introduce the parameters 5

PAGE 17

which affect the design of TSM sensors. These parameters directly impact the operation of a chemical sensor. Next the fundamental operational theory behind TSM chemical sensors will be presented. 2.2 Temperature Stability and Pressure Effects For a sensor application, the range of temperatures where the sensor is stable is important. The temperature stability determines the temperature range over which the TSM device will be capable of maintaining a stable frequency. This temperature stability is dependent upon the direction in which the quartz material (wafer) is cut. The determination of the axis of the quartz bar results in crystal cuts such as AT, CT, BT. The cut angle determines the temperature coefficient, which is a way of expressing how the change in frequency will be affected by a change in temperature. The TSM devices used in this study were AT cut. The orientation of an AT cut quartz crystal is shown in Figure 2.1 [5]. Figure 2.1. AT Cut Quartz [6] 6

PAGE 18

The effect of temperature on the resonant frequency for several cut angles of an AT crystal is shown in Figure 2.2 [6]. Each curve represents a different cut angle. Note that there is a zero temperature coefficient at room temperature. With a reference temperature of 26 C, a predictable range of curves may be developed for different angles of AT cut crystals for specific frequency ranges. Figure 2.2 AT Cut Frequency vs. Temperature [6] The frequency change due to pressure changes has been shown by Stcockbridge to be )10)(015.035.18(10Tffr per torr (2.1) Where T is in K, is the measured frequency shift and f f r is the resonant frequency of the QCM [7]. Since pressure changes in this work were small, changes in frequency due to pressure will not be considered. 7

PAGE 19

2.3 Fabrication Process First the crystal is cut along the axis (determined by x-ray) into an AT cut. The cuts, or wafers, are cut into blanks, which are disks of quartz. Blanks are processed by grinding or lapping. The frequency that the blank will oscillate will be determined by the thickness of the blank. The lapping process is carried out by planetary machines comprised of plates with a high degree of flatness. Blanks are placed in carriers and rotated between the plates, in a slurry of abrasives, which serve to reduce the thickness of the blanks. This is all carried out at controlled rates to achieve the desired blank thickness. If plates are not flat the blanks will not be flat and the temperature coefficient will not be predictable. The thickness of the blank is determined by equation 2.2: rfThickness2.66 (2.2) Where the thickness is in inches and the frequency is in kHz. Blanks are processed with polish finishes until they are transparent, corresponding to 1 or 2 microns. Crystals are then etched in a solution to remove carryover abrasives and slight surface discontinuities. A common solution used is ammonia bi-fluoride or hydrofluoric acid; these etchants also serves to improve the surface finish [8]. Etching procedures are described elsewhere in detail [9]. After etching, the crystal blank is in its final geometric form and has a higher frequency than desired. This is because the crystal must be plated to final frequency in a controlled plating process referred to as plate-back. Plate-back is the process of attaching electrodes to the face of each side of the blank. The electrodes are needed to provide an electrical connection. This is required so 8

PAGE 20

that the crystal can be placed in an electric circuit. The metal is added to the surfaces in equal amounts and thicknesses. The deposit slows down the mechanical vibrations and the process is similar to adding weight to a pendulum to slow down its motion. Plating is usually carried out by vacuum deposition. Common metals used for plating are silver, aluminum, and gold. Chromium, added prior to the gold, silver or aluminum, is used as a binding layer for these metals. The crystal is held in its holder by mounts that make electrical contact with the metallic film on the face of the crystal. After mounting, a special conductive epoxy is applied to the contact point and is cured. After the mounting operation is complete, the crystal unit is finished to the final frequency by adding additional metal to the crystal face in small amounts by an electrical disposition method by a finisher. There are several methods, such as resistance welding, that are available for finishing. 2.4 TSM Sensor Theory To construct gas or vapor sensors, chemically sorbent films are commonly coated onto TSM resonators [10]. Chemical sensitivity and selectivity is imparted by attaching a thin film to the acoustically active region of the TSM device. Devices employed in this work were AT cut quartz crystals with circular gold electrodes on both sides. Because of the piezoelectric properties of the quartz material, application of a voltage between the two electrodes at the surface results in a shear deformation of the crystal. The quartz crystal then vibrates via the piezoelectric effect and this vibrational motion results in the generation of a transverse acoustic wave that propagates across the thickness of the 9

PAGE 21

quartz crystal. The resonant frequency of the TSM device decreases with the crystal thickness when a standing wave condition is met, as 2qqt (2.3) 2/1qqv (2.4) Here is the velocity of sound in the quartz, is the quartz thickness, and v qt q is the acoustic wavelength. Equation 2.4 gives the velocity, where q is the density (2.648 g cm -3 ) and q is the shear modulus of the quartz (2.947 x 10 11 g cm -1 s -2 ). Since the quartz thickness is much larger than the electrode thickness, the electrodes are neglected when determining the resonant frequency. Equations have been developed to yield expressions for the dependence of the resonant frequency on the mass changes occurring within films coated onto the TSM sensor. The Sauerbrey equation is one such relationship, valid for small mass loadings, such as in vapor sensing applications: 2/12)(2qqsrff (2.5) where is the measured frequency shift and f s is the film density (which can be related to the film mass) [11]. Changes in the film mass will cause frequency shifts; these frequency shifts are dependent upon the film selectivity and the device sensitivity. In Sauerbreys model, the sensitivity is given by: fc qqrfvfc22 (2. 6) 10

PAGE 22

Consequently, a commonly used 10 MHz AT cut quartz crystal (with the previously mentioned physical properties) will have a mass sensitivity of 2.26 x 10 8 Hz cm 2 g -1 The addition of material with an areal density of 4.42 ng cm -2 will cause a 1 Hz shift on this resonator, which is easily measurable using common electronics. The sensitivity predicted by equation 2.6 is a theoretical sensitivity, which does not account for the increase in sensitivity due to the addition of a sensing film. Therefore an equation for the experimental sensitivity is needed. This equation is derived from the Sauerbreys equation. First, the frequency shift is defined to account for the additional mass of the polymer sensing film. spfff (2.7) where f p is the frequency shift due to the polymer and f s is the frequency shift due to the solvent (analyte). Similarly for a change in mass, spmmm (2.8) where pphm is the areal mass (mass per area of the electrode). h is the polymer thickness and p is the polymer density. We can define weight fraction w s in terms of the change in mass, spssmmmw (2.9) Rearrange for sm sspswwmm1 (2.10) Substitute in the expression for total mass change, equation 2.8, 11

PAGE 23

ssppsspspwwhhwwhmm111 (2.11) The sensitivity becomes, spspfewhffc11 (2.12) Equation 2.12 represents the measured experimental mass sensitivity of a polymer coated resonator. This equation accounts for the polymer thickness effects on the resonator sensitivity to the analyte. Although a 10 MHz TSM sensor already has a high degree of theoretical sensitivity, it is obvious from Equation 2.6 that this sensitivity can be considerably increased by increasing the resonant frequency. For a 56 MHz device, experimental sensitivity increases proportional to the 2.88 exponent of the resonant frequency have been reported for liquid sensing with the signal to noise ratio improving by a factor of 6.5 [12]. This 2.88 power dependence is greater than predicted by Sauerbreys model. Even with sensitivity increases predicted by the Sauerbrey model, which is more likely valid for gas phase operation, a 100 MHz device will be 100 times more sensitive than a 10 MHz device. Achieving higher frequencies requires thinner quartz crystals; this is difficult to accomplish using conventional fabrication techniques. However, higher frequency devices can be produced by operating at overtones of the fundamental resonant frequency or by using milled devices [13]. Equation 2.13 shows that the quartz thickness determines the wavelength of the fundamental ( qt 1 n ) and harmonic (3, 5, 7 ) resonances n 12

PAGE 24

qntnvf2 (2.13) where is the new fundamental frequency for various thicknesses or at different harmonics. Milled crystals are made thinner only in the center so that a thin quartz membrane is fabricated with a thick outer ring allowing for mechanical stability. These higher frequency devices are long known to have improved mechanical stability and frequency to noise ratios that are better than other acoustic wave devices [14, 15]. Chemically milled devices specifically designed for our experiments, with fundamental frequencies of approximately 99 MHz having 20 mils electrode diameter were fabricated at MTronPTI, Orlando, FL. A diagram of a milled TSM device is shown in Figure 2.3 below. nf Thick Outer Ring Milled Membrane Electrode Figure 2.3 Milled TSM Device 2.5 Electrical Circuit Model An equivalent circuit model, such as the Butterworth-Van Dyke (BVD) model, can be used to describe the electrical characteristics of a QCM coated with a viscoelastic film such as rubbery polymers [16, 17]. Prior to modeling the polymer coated QCM, the electrical characteristics of the uncoated device must also be modeled. The BVD model uses simple lumped element 13

PAGE 25

parameters to simulate the uncoated devices electrical characteristics over a range of frequencies near resonance. Actual physical properties of the QCM as well as the surface loading are related to the lumped elements in the circuit by equations 2.14 to 2.17. sohAC22 (2.14) 221)(8NCKCo (2.15) 1211CLs (2.16) 11CRqq (2.17) where is the static capacitance and ,, are the motional contributions. is electrode area, is the series resonant frequency and oC 1C 1L 1R sh sw 2 K 22 q q and q are the square of the quartz electromechanical coupling coefficient, dielectric permittivity, shear stiffness, mass density, and effective viscosity respectively [18]. The circuit has two branches. The motional branch, which contains the L 1 R 1 and C 1 is the branch that is modified by mass and viscous loading of the crystal. The static branch, which contains the lone C* o element, represents the static capacitance of the crystal electrodes and any cable and fixture capacitance. Physically, R 1 represents the energy loss arising from the viscous effects and internal friction, L 1 the initial mass/motional inertia of the system, and C 1 the mechanical elasticity of the quartz. 14

PAGE 26

C*oL1C1R1 Figure 2.4 Equivalent Circuit Model of Uncoated TSM Resonator Similarly lumped elements can be used to model the electrical characteristics of a coated device. Film loading causes additional contributions to the motional elements and this increase is denoted as eZ qsoseZZCwKNZ24 (2.18) Where Z q depends on the properties of quartz, ,) 2/1)(qqqZ 2(rsfw and Z s is the shear mechanical impedance at the device surface [19]. Admittance data can be obtained from an impedance analyzer or network analyzer and then the circuit model can be fitted to the data to extract the physical properties of the surface perturbations [20]. The polymers storage modulus G and loss modulus G values can be extracted from this data provided that the film thickness and density are known. G (G=G'+jG") is the complex shear modulus of the film. The real part of the shear modulus is called the 15

PAGE 27

storage modulus (G') representing energy storage during oscillation. The imaginary part is the loss modulus (G") which gives rise to power dissipation in the film. By monitoring the changes in the resistance of the film to the presence of different analytes, the identity of the analyte can be determined [21]. C*oL1C1R1 Ze C*oL1C1R1 L2R2 Ze Figure 2.5 Equivalent Circuit model of Coated TSM Resonator The electrical characteristics of the coated QCM can be related to properties of the perturbing polymer coating. Figure 2.5 is an electrical model that approximates the coated QCM [22, 23]. Z e is given by Equation 2.18. In Equation 2.18, Z s is a complex quantity. The motional contributions, R 2 and L 2 of the film can be modeled in terms of Z s as shown in Equations 2.19 and 2.20. 16

PAGE 28

qsosZZCwKNR)Re(422 (2.19) qsosZZCwKNL)Im(4222 (2.20) It should be noted that the frequency can also be determined indirectly by measuring the inductance [19]. Measuring the inductance to determine the frequency allows one to verify whether or not the frequency measurement is accurate. 2.6 Polymer Film As mentioned before, a TSM device is only a resonator if there is no sensing film. The sensing film imparts selectivity and sensitivity towards a particular chemical, thus, making a resonator into a sensor. The sensing films used in this thesis were polymer coatings. The coating should physically or chemically bond to the surface of the TSM resonator and should behave as an ideal mass layer. An ideal mass layer should be infinitesimally thick when compared to the thickness of the quartz material (ignoring the thickness of the gold electrodes). The polymer coatings in this thesis work were applied to both sides of the electrodes, but the total film thickness was calculated as the sum of thickness on both sides of the resonator [24]. In effect the actual film thickness is the half of the total film thickness. Since different resonators were compared in this study, the physical characteristics of the resonators were different, with each sensor having a different electrode area. Since only the coated electrode area is region of the sensor 17

PAGE 29

surface where the coating analyte interaction (physisorbtion) occurs, comparing the sensitivity of devices with different electrode areas is inaccurate. The model developed equation 12 accounts for these differences, however, the polymer film coating f p must be similar for all devices. Since each device has a different sensitivity, the frequency shift due to the polymer coating will be different, but the thickness can be constant. Another complication to having similar film thickness, however, is introduced because each device has a different quartz blank thickness. Consequently, the ratio of the thickness of the quartz blank to the polymer film must be kept constant. This ratio was approximately 0.5 %. The uniformity of the film has little effect on the detection of the chemical vapors, however the film should be adherent and stable in the presence of the organic vapors [19, 25, 26]. Several methods are available for coating TSM resonators and are summarized in Table 1. The performance of several polymer film coatings of poly-vinyl acetate, poly-vinyl pyrrolidone, polystyrene-butadiene, and poly-isobutylene were studied in this thesis. The results based on several performance criterion were used to select the appropriate film and will be presented in Chapter 4. Since, we were sensing for organic vapors, rubbery polymer films known to physisorb with organic vapor, such as the ones mentioned previously, were selected [25, 27-29]. Nevertheless, not all films are ideal for a given application and polymer properties, such as glass transition temperatures, must be taken into account when choosing the sensing film. 18

PAGE 30

Table 2.1 Coating Methods [30] Coating Method Description Spin Coating Quartz crystal is rotating on a horizontal plane. Solution is dropped on the rotating crystal and centrifuge force spreads the solution while solvent is vaporizing. The rotating rate, solution viscosity, and solution density, directly affect the film thickness and uniformity. Spray Coating An airbrush containing a solvent reservoir is drawn upon by a mild flow of carrier gas, usually nitrogen. The solvent is then sprayed through a nozzle onto the quartz crystal surface. The distance between the nozzle and the target quartz crystal needs to be adjusted depending on the solution. The nozzle needs to be moved across the quartz crystal surface for better uniformity. Continual applications are then repeated until the desired thickness is achieved. Operator skill is a key factor for good coating. Solution with more volatile solvent is easier to coat. Drop Coating Solution is dropped on quartz crystal surface. Let solvent evaporate by itself or with the help of gentle heating. This is a simple method of coating. Repeating drops is necessary for desired thickness. This method is highly dependent on solution viscosity and density. Surface tension plays an important role. To aid in the spreading of the solution, the gold surface of quartz crystal may need to be treated to make it hydrophilic or hydrophobic depending on the substance. Oscillating Capillary Nebulizer Solution is injected into a fused silica capillary with 50 ID. This capillary is friction mounted into a fused silica capillary of 250 ID Helium is then introduced into the large capillary thus causing the inner capillary to oscillate. This oscillation caused the solution to be nebulized at the capillary tip. 19

PAGE 31

2.7 Performance Criteria There are many factors which affect the performance of chemical sensors. One of the most important factors is the sensors sensitivity. Experimental and theoretical sensitivity was addressed in detail in Section 2.4 and will be excluded from this discussion. Other important factors which affect a sensors performance are selectivity, reversibility, response time, dynamic range, stability and environmental effects (such as changes in temperature and pressure). Temperature stability and pressure effects were discussed in Section 2.2. 2.7.1 Selectivity Selectivity is the capability of the sensor to distinguish between different analytes and interferences. Selectivity is the most difficult sensor performance parameter to achieve. In most sensor applications selectivity is achieved by the sensing film. In our case, polymer films were used to sense for organic vapors, however, the sensing film was not specific to a particular vapor. Thus, the sensor was not selective towards a particular organic vapor. Nevertheless, we were able to distinguish between different organic vapors though the equivalent circuit parameters. The resistance of a polymer film can be modeled thorough the Butterworth-VanDyke equivalent circuit model and monitored by an impedance analyzer. The mechanical impedance of the polymer film changes when the film is in the presence of the organic vapor. The mechanical impedance is related to the resistance parameter [19]. Literature reports of distinct changes in the resistance 20

PAGE 32

parameter to particular organic vapors show that it is possible to distinguish between different vapors [21, 31]. Selectivity, in this thesis work, refers to the ability to sense for any organic vapor present and to be able to distinguish between these vapors; whereas, most sensors aim to sense for only a single target analyte. 2.7.2 Reversibility The resonant frequency of the TSM device will be the primary parameter monitored in this thesis work. Reversibility refers to the ability of the sensor to return to its baseline frequency after exposure to an organic vapor. Reversibility of an analyte/coating interaction depends upon the strength of the interaction. Since the actual sensing mechanism is physisorbtion at the polymer, reversibility should not be difficult. Complete removal of the analyte from the polymer coatings was accomplished by flooding the sensor with pure nitrogen gas for a sufficient time to allow for the analyte to be removed completely. 2.7.3 Dynamic Range The dynamic range is defined as the concentration interval over which a sensor provides a continuously changing response [19]. The dynamic range has an upper and a lower bound. The lower bound is called the limit of detection (LOD) and the upper end represents the saturation point. The LOD represents the point where the signal is greater than noise levels. The LOD also depends upon the sensitivity of the device. LODs are commonly defined in terms of signal to noise ratios (S/N) of two or three. This corresponds to situations where 21

PAGE 33

the signal exceeds the noise at statistical confidence levels of 95% (for a factor of 2) and 99% (for a factor of 3) [32]. The LOD can be accurately defined by equation 2.21 with units of concentration or ppm. fCNoiseLOD3 (2.21) Physically, the LOD represents the lowest level of a chemical that can be detected accurately by a sensor. The value of the LOD also depends upon the kinetics and thermodynamics of the coating analyte interaction and the thickness and/or surface area of the coating [19]. The saturation limit is dependent upon the limitations of the electronic circuitry and physical and chemical limits. The saturation point is reached when a certain quantity or concentration level of the analyte exceeds the sorptive capacity of the coating. In this thesis work viscoelastic polymer films were used. These films are plasticized at high concentration levels of the analyte, which increases the acoustic wave attenuation and leads to a dampening of the sensor signal. In general, it is desirable to have a wide dynamic range. 2.7.4 Stability, Repeatability, and Reproducibility, and Response Time Stability refers to the noise levels, which are positive and negative deviations from the baseline signal. Stability also refers to the drift in the baseline response of the sensor. Baseline drifts are the result of degradation of the sensor film or of the sensors 22

PAGE 34

circuitry. Both baseline drifts and noise levels should be minimized. It is apparent that noise levels significantly affect the LOD. Repeatability is the extent to which a sensor shows the same signal for the same analyte concentration under the same operating conditions. This is strongly dependent upon stability. Reproducibility refers to the extent to which two sensors fabricated by the same procedures, coated with the same film, in similar operating conditions and tested with the same analtye produce the same response. A sensor response time is the time taken for the sensor response to change to accurately represent the total amount of analyte present. The response time depends upon the coating/analyte behavior. For example, the response time of organic vapors in polymer films depends upon the properties of the polymer film. Polymers below their glass transition temperatures generally have longer response time, whereas amorphous polymers or polymer above the glass transition temperature have shorter response times. The response time, in this thesis work, depends on the rate of diffusion into the polymer, which is governed by temperature and the thickness of the film. 23

PAGE 35

CHAPTER 3 EXPERIMENTAL APPARATUS 3.1 Introduction To accurately test the response of TSM devices, it is imperative that the method of generating and calibrating test samples is also accurate. The type of system that is required for sample generation will depend on the nature of the sample and required measurement. In general the system will be composed of a carrier gas stream, a source of sample gas or vapor, mixing and dilution stages, and provision for both temperature and pressure control, and a method of recording the sensor response [33]. The sensors developed in this thesis were chemical vapor sensors, specifically for organic vapor sensing; hence, they required the generation of calibrated organic vapor samples. There are several methods available for accomplishing this task. The following discussion will describe the principles behind the preparation of standard gas mixtures as it applies to generating organic vapor samples. These principles will be used to design the experimental apparatus. 24

PAGE 36

3.2 Static and Dynamic Gas Generation Methods Vapor mixtures can be generated by a variety of methods. Some of the gas generation methods are presented in the following flow chart (reproduced from [34]). PRODUCTION OF STANDARDGAS MIXTURES STATIC DYNAMIC PRESSURIZED ATMOSPHERIC GRAVIMETRIC SINGLE RIGIDCHAMBER PARTIALPRESSURE MULTIPLE RIGIDCHAMBER VOLUMETRIC FLEXIBLECHAMBER GAS STREAMMIXING INJECTION PERMEATION DIFFUSION ELECTROLYTIC CHEMICALREACTION EXPONENTIALDILUTION Figure 3.1 Production of Standard Gas Mixtures These methods can be classified as static or dynamic methods of vapor generation. The gas or vapor mixture generated by either method should be accurate, reproducible and 25

PAGE 37

easy to prepare. Generation of gas mixtures are much less accurate than liquid mixtures because liquids can be weighed easily, volumes may change during handling and temperature and pressure effects must be considered [34]. Static and dynamic methods are reasonably accurate for most application purposes. Static methods are limited in the volume of the sample generated since they involve preparing a sample gas and storing it in a closed vessel. Static methods are preferred when a large concentration is required. Also, since the sample is stored in a container, there are significant losses of componenets of the mixture to the walls of the containers [34]. Dynamic methods, as the name suggests, generate a continuous flow of a mixture, produce large sample volumes, and have less losses to surfaces due to an equilibrium between the walls of the system and the flowing gas stream. 3.2.1 Major Component/Zero Gas/Dilutant/Carrier Gas An important consideration in the design of a vapor generation apparatus, in both static and dynamic methods, is the major component of the gas mixture. This is frequently air or some inert gas, such as nitrogen. This is especially critical for dynamic systems, since the volume of the major component must be large. The term zero gas is applied to the major component. Since the sensors that will be tested have high sensitivity, very low concentrations and high purity mixtures are required. Gases generated in this thesis work will use nitrogen as the major component or zero gas. The nitrogen used will be of ultra high purity grade. It should be noted that even though nitrogen gas has a certified purity of 99.99 %, it may still contain up to 100 p.p.m. of impurity; therefore, there is an obvious lower limit of test sample concentration which 26

PAGE 38

can be accurately tested [35]. Depending on the requirements for the concentrations desired for testing the sensor response, methods of generating higher purity zero gases are available [36]. Based upon the above discussion, the method chosen for vapor generation was the dynamic method. The following discussion relates to the design principles behind dynamic vapor generation methods. 3.2.2 Dynamic Method Advantages From Figure 3.1, it can be seen that the first advantage of the dynamic methods is the variety of generation methods available. Dynamic methods require an uninterrupted blending of component parts for some specified time period. In comparison to static methods, this has an advantage when dealing with reactive mixtures. The undesired side products can be swept away, discarded, and replace by the pure un-reacted test sample. Since the samples considered in this thesis were not reactive, this will not be an issue. Another advantage of the dynamic methods is that the range of sample concentrations varies from p.p.m. to p.p.b. levels. Also, wall absorption, as previously mentioned, will not be a problem. 3.2.3 Flow Control Gas flow control is important due to the kinetic nature of the sensor response [33]. The most widely used method of mixing gases is to dilute the gases with one another [37-45]. Turbulence of the components in the streams of gases are usually enough to promote homogenous mixing in the lines [46]. Precise flow control is also a desired design requirement for obvious reasons such as accuracy of the generated sample concentration. 27

PAGE 39

The method of flow control, utilized in this work, was mass flow controllers. Mass flow controllers allow for control of gases from 1 mL per minute to 5000 mL per minute. Mass flow meters depend on the rate of cooling or heat transfer. The heat transfer depends on the amount of heat added to the gas, the number of molecules passing the heat source and the heat capacity of each molecule. The output is dependent on only the mass flow of gas, not the temperature or pressure variations. The bulk gas goes through a laminar-flow by pass while a portion goes through a sensor section. The sensor section is composed of two resistance temperature detector coils around the sensor tube, which directs a constant quantity of heat into the gas stream [47]. When no gas is flowing, the heat at each sensing coil is the same; however, when the gas is flowing, heat is carried from the upstream sensor to the downstream sensor. The temperature difference is proportional to the flow of gas and is converted to a 0 to 5 V output signal which is viewed on a meter readout. The flow controllers utilized in this study were MKS flow controllers. When using mass flow controllers, corrections must be made for different gasses. The MKS flow controllers allow for various gases to be used, however, nitrogen gas was the zero gas used in this study. The gas conversion factors are dimensionless ratios which relate the sensitivity of the mass flow meter to two different gasses [48]. Conveniently, the conversion factor for nitrogen was approximately zero, and other gases are referenced to nitrogen. 2121KKQQ (3.1) Where Q i is the volumetric flow rate at standard conditions at 0 C and 1 atm and K i is the conversion factor. Subscript 1 is the unknown gas and subscript 2 is the reference 28

PAGE 40

gas, nitrogen. The conversion factor can be obtained from the manufacturer of the mass flow controller (MKS), TCNHQpm (3.2) Where Q m is the mass flow rate (g/min), N is the correction factor for the molecular structure (See Table 3.1), H is the constant amount of heat applied to the sensor tube (cal) and C p is the specific heat (cal/g), and T is the temperature difference between upstream and downstream sensor coils. The mass flow rate can be written as, QQm (3.3) Where is the gas density at standard conditions (g/L). The temperature difference is proportional to aET (3.4) Where a is a constant and E is the output voltage. Table 3.1 Values for Molecular Correction Factor Gas Value of N Monatomic, i.e. argon, helium, xenon 1.01 Diatomic, i.e. nitrogen, oxygen, nitric oxide 1.00 Triatomic, i.e. carbon dioxide, nitrous oxide 0.94 Polyatomic, i.e. ammonia, arsine, diborane 0.88 Combining equations and making the necessary substitution, Q becomes pCbNQ (3.5) 29

PAGE 41

Where aEHb at a constant output voltage. The required ratio, however, is the ratio of the actual gas to the ratio of the reference gas. Combining equations and eliminating b leads to 1122212121ppCNCNKKQQ (3.6) This expression is used to accurately generate flow rates of the zero or carrier gas, nitrogen. 3.3 Experimental Apparatus Design The experimental apparatus was designed based on the design principles mentioned before. A visual schematic of the vapor dilution system is given in the Figure 3.2. This schematic does not contain the details of the design, and its purpose is to familiarize the reader with the general operation of the vapor generation apparatus. Impedance Solenoid Valve Solenoid Valve Computer NI-PCIData AquisitionCardAnalyzerTest Cell EthyleneGlycolCirculator TemperatureController VaporGeneratorMass FlowControllers Impedance Solenoid Valve Solenoid Valve Computer NI-PCIData AquisitionCardAnalyzerTest Cell EthyleneGlycolCirculator TemperatureController VaporGeneratorMass FlowControllers N2N2N2 To Exhaust Figure 3.2 Conceptual Apparatus Design 30

PAGE 42

Organic liquids were contained in four bubbler units housed in a temperature bath. MKS mass flow controllers were used to regulate nitrogen carrier gas through the bubbler unit. The bubbler unit consisted of a temperature controlled flask in which the organic liquid is contained. Multiple mass flow controllers (100 sccm, and 200 sccm) allowed for a variation of test sample concentrations. In the present design, there are three streams: The carrier stream passing through the bubbler, and two dilutant streams. The vapor pressure of the liquid at the bubbler temperature is calculated using an accurate vapor pressure correlation (Wagners equation) [49]. xVPDxVPCxVPBxVPAxPPcvp****(*)1()ln(35.11 (3.7) Where, cTTx1 Pvp is the vapor pressure of the gas, P c is the critical pressure, T c is the critical temperature, and T is the solvent temperature. Constants VPA, VPB and VPC for all chemicals studied are presented in Appendix A, Table A.4. Multiple mass flow controllers allowed for a variation of test sample concentrations. In the present design, there were three streams: The carrier stream (Qcs) passing through the bubbler, and the dilutant streams (Q1 and Q2). From Wagners equation, the vapor pressure of the liquid at a given temperature is calculated. The mole fraction is approximated from the volume of the vapor generated (Qs). The fraction of vapor generated is further diluted and a new diluted mole fraction is calculated. This mole fraction is equated to the volume fraction to determine the volume of gas generated and then the concentration is approximated with the ideal gas law [33]. Equation 3.8 gives the resulting expression used to determine the ppm level generated,. vvPPM 31

PAGE 43

21)10(6QQQcsQsPPMvv (3.8) The above equations were used to calculate theoretical concentrations for various organic vapors. The concentration levels of the solvents generated are given in Appendix A, Table A.1, A.2 and A.3. The temperature of the liquid bath was controlled by circulating antifreeze coolant with a Lauda water circulator. The temperature of the bath was logged and found to be accurate to within 0.5 C. An Agilent 4294 A Precision impedance analyzer was used to monitor the resonant frequency and equivalent circuit parameters of the TSM device. A custom made stainless steel test cell was designed for housing the sensor and was kept under temperature control using Thermolyne heating tape and an Omega PID controller. The PID controller was accurate to within 0.1 C. The TSM device was attached to a printed circuit board using a commercially available socket holder, obtained from International Crystal Manufacturing (ICM), Oklahoma City, OK, for 10 MHz and 20 MHz resonators and a transistor socket for higher fundamental frequency resonators (96 MHz). The sockets facilitated easy removal of the sensor without disturbing electrical connections. The 96 MHz transistor socket and the printed circuit boards (PCB) were custom made. See Appendix B for the details on the PCB fabrication process and the 96 MHz socket. 3.4 Vapor Dilution System Components The apparatus is shown in Figure and consists of four sections 1. Solvent cell 2. Sensor cell 3. Vapor generation and dilution 32

PAGE 44

4. Automation The actual experimental apparatus is given in the Figure 3.3. Mass Flow Controllers B C D F-1 F-2 F-3 F-4 G-1 G-4 G-3 G-2 E I AFHJKLMA Nitrogen CylinderG-1,G-2,G-3,G-4 Vapor Generator UnitsB MKS Mass Flow Controller (Carrier Gas)H Omega ThermometerC MKS Mass Flow Controller (Dilutant 1)I Stainless Steel CellD MKS Mass Flow Controller (Dilutant 2)J Heating TapeE Lauda Heater/Chiller Ethylene Glycol CirculatorK PID Temperature ControllerF Temperature BathL Personal ComputerF-1, F-2,F-3,F-4 Solenoid ValvesM Agilent Impedance AnalyzerTo FumeHood F-1 F-2 F-3 F-4 Figure 3.3 E xperim e ntal Apparatus 3.4.1 Solvent Cell The sensor cell is essen tia lly a f l ask w h ich contains the org a nic solv ent in liquid phase. The solvent flask, or bubbler, was a 250 m l round botto m flask with an inlet and outlet port and is shown in the Figure 3.4. The inlet port allowed for the carrier nitrogen gas stre am t o enter in to the bubble r The inle t p o rt was fitte d with a fritted tube, wh ich served as a sparger that allowed for the n itrogen gas to bubble through the solvent liquid. This allowed for a longer residence tim e in the bubbler, better disper sion of the nitrogen gas, and it also aided in m a ss tran sfer fr om the liquid phase to the vapor phase. The bubbles coalesce when they arrive at the surf ace of the liquid, to release the entrapped gas. This gas then leaves the bubbler unit th rough the top outlet port. All glass fittings 33

PAGE 45

were wrapped with Teflon sleeves to prevent leaks. In addition, multiple o rings were used to prevent leaks. Figure 3.4. Bubbler Unit, Fritted Tube, and Fittings 3.4.2 Sensor Cell Designs The design of the sensor cell was important in determining the sensor response. Figure 3.5 shows the first sensor cell design. There are some notable features of this design which should be avoided if the correct sensor responses are to be determined. Some of the sensor parameters which are affected by the cell design are the response time, frequency noise, and sensor recoverability. Response times were up to 5 minutes with all resonators; the frequency noise was over 100 Hz, depending upon the resonator studied, and the recoverability was slow. These effects were mainly due to the large cell volume [19]. Tapered Fittings Inlet Outlet 34

PAGE 46

SMAThermocoupleQCM TeflonSheetPCB Gas Inlet/OutletSS Tubing Cross Section 0.5 "1.75"0.5 Welded Top Screw 0.5 2.75 1.5 0.5 0.5"0.25 "0.25 "1.5" PCBGroove Screw 2/16 0.25 "Open 2.0" Screws Figure 3.5 Sensor Cell Design 1 1). No streamlined flow across the QCM because the inlets were located at the top of the cell. 2). Long equilibrium times which resulted in long sensor response times 3). Channeling of N 2 which resulted in unstable baselines and high frequency noise levels. to reduce the errors in the sensor parameters and responses a cell was designed to minimize the volume and to allow a streamline flow of the sample gas to flow past the sensor. The design of this new cell is shown in detail in Figure 3.6. 35

PAGE 47

Figure 3.6 Sensor Cell Design 2 The volume of the cell was minimized and the resonators were placed in parallel with the direction of the gas flow stream. Several lids were fabricated to allow for multiple resonators with different dimensions to be easily installed in the same sensor cell. Diagrams for these lids are given in Appendix B. 3.4.3 Automation Labview 7.0 was used to control all instruments. The purpose of the automation was to accurately log the sensor parameters and to reliably generate gas samples with minimal time being spent by the experimenter. Experimental running times, with a single 36

PAGE 48

solvent, lasted for over 3 hours; hence, automation was necessary. The automation consisted of two parts: 1. Logging the sensor response from the impedance analyzer. 2. Controlling all experimental conditions by controlling all instruments (solenoid valves and flow controllers). The Labview flow diagram for automation and data logging is shown in Figure 3.7. The wiring diagram for the solenoid valves and flow controllers is given in Appendix B. An outermost while loop was used to execute the program for the desired number of times. In effect, this served to repeat an experimental run to check the repeatability of the sensor responses to a particular solvent. It also allowed for the flexibility of selecting the bubblers filled with four different solvents from one through four. The user can choose to run one solvent for several times, or four solvents one at a time. The program consisted of many while loops and cases which are described in detail in the next few pages. Sensor parameters, such as the resonant frequency and equivalent circuit parameters, were logged from the impedance analyzer by using a GPIB controller to send GPIB commands to the analyzer. All experimental values were logged to a text file or excel sheet for processing and analysis. Figure 3.8 is the first sequence of the program. Relays were used to activate solenoid valves. These relays must be deactivated if the program is to shut done properly. Relays remained activated after the experimental run was completed which was due to an issue in the programming. To compensate any kind of possible erroneous behavior of the nine relays it is desired to set all the relays to OFF position. The relays were paired together, and are represented by the green T/F boxes in Figure 5.8. 37

PAGE 49

Figure 3.7 Overall Automation Programs 38

PAGE 50

Figure 3.8 Initialization of All Valves There were 5 T/F cases (one pair for each bubbler, and one for the purge line). Initializing the program with the false case (F) sends a false value to all the relays which shuts off all relays and valves prior to running any other programs. A time delay of 2 seconds was set prior to activating any other programs to ensure that all relays were deactivated. A DAQ Assistant for the PCI 6025E (200 kS/s, 12-bit 16 analog input multifunction DAQ) connected with SCB100 (I/O connector block for 100-pin digital devices), was responsible for sending the digital outputs to all the relays and analog outputs to all mass flow controllers. Figure 3.9 represents the code for the second sequence in the Labview program. Sequence two sets the purge solenoid valve to a true value, and sends a setpoint of 100 sccm nitrogen to the flow controllers. This causes a purge of pure nitrogen to clean the entire vapor dilution apparatus and to establish the baseline response of the sensor. The purge time has a time delay of 10 minutes; this sets the purge duration to 10 minutes. 39

PAGE 51

Figure 3.9 First Nitrogen Purge Cycle Figure 3.10 Purge Cycle Control Loop 40

PAGE 52

Figure 3.10 shows the third sequence of the program. This sequence is divided in two cases, one true and false. Loop iteration was used to determine the action of the case. The iteration number is divided by two and the quotient is checked for zero or greater than zero values. When the quotient is zero, a true value is sent to both the cases; and when it is greater than zero, a false is sent to both the cases. Both cases are responsible for setting constant flow through the sorption cell. When the sequence is case 1, one turns off all the relays except the purge valve and case 2 sets the purge flow to 100 sccm. Figure 3.11 Cycle Control Loop 41

PAGE 53

Figure 3.11 is the false case of the third sequence. When the sequence is in case 1, the relays and solenoid values particular to a specific bubbler are activated. As expected, in this sequence, case 2 turns off the purge. In this sequence, the value of the flow through the bubbler, carrier gas, is stepped up from 10 sccm to 100 sccm in increments of 10 sccm. Although the carrier flow rate is constantly changing to vary the concentration of the analyte, the total flow at the sensor must always be 100 sccm. A total flow rate of 100 sccm must always be maintained to ensure that flow effects do no affect the sensor response, and also to establish the sensors baseline frequency. Consequently, the flow through the remaining mass flow controller was incremented or decremented to maintain the constant 100 sccm flow at the sensor cell. 42

PAGE 54

CHAPTER 4 RESULTS AND DISCUSSION 4.1 Introduction Several polymer films were used as the sensing layer in this work, however, not all sensing films yielded desirable responses. Results of trials with several polymers are presented first to determine which polymer was best suited for more extensive studies. Three sensors were exposed to seven organic vapors with the appropriate sensing films. Experimental results of exposures to benzene are presented in this chapter, with the remaining results of the other six organic vapors presented in Appendix C. Finally, the discrimination of the seven solvents is presented. 4.2 Polymer Film Selection Several polymer films were utilized for sensing organic vapors. These polymers were polyvinyl acetate, polyvinyl pyrrolidone, polyisobutylene, polystyrene, polystyrene-butadiene copolymers. An ideal sensing film should be able to recover after exposure to the analyte, have a stable baseline frequency, and repeatable responses. These features are affected by the properties of the polymer. Mainly the film should be rigid enough to move with the oscillation of the TSM resonator, but the film should also be soft enough to allow for sorption of the analyte. Films that are too rigid have longer equilibrium 43

PAGE 55

times, consequently, these films may not be practical for a sensing application. Equilibrium times of less than one minute are ideal. Polystyrene, polyvinyl acetate, and polyvinyl pyrrolidone films were found to have equilibrium times of more than 20 minutes when exposed to 27,000 mg/m 3 of benzene at room temperature. Figure 4.1 shows a typical response of a 20 MHz TSM resonator coated with 23.25 kHz of polyvinyl acetate (PVA) to benzene vapors (0.823 to 7.66 volume percentages). Notice that the resonator frequency continuously decreases and that an equilibrium is never reached for 0 2000 4000 6000 8000 10000 12000 14000 -4500 -4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 500 Time (seconds) Frequency (Hz)0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 %7.66 % Figure 4.1 20 MHz Response to Benzene, PVA Film 44

PAGE 56

0 2000 4000 6000 8000 10000 12000 14000 9.9796 9.9796 9.9797 9.9797 9.9798 9.9799x 106 Time (seconds) Frequency (Hz)0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 %7.66 % Figure 4.2 10 MHz Response to Benzene, PVP Film all exposure levels. This response is typical of a film that has a long equilibrium time and does not readily desorb the analyte. A similar response was obtained with polyvinyl pyrrolidone (PVP) coated to 20.03 kHz on a 10 MHz resonator (Figure 4.2). Hence, polystyrene would not be an ideal film for sensing benzene. Polymer films of polybutadiene were too soft, resulting in unstable baseline frequencies, large baseline drifts, and poor repeatability. The polymer glass transition temperature was found to be factor which determined whether a film would behave ideally. Glass transition temperatures of the polymers investigated in this thesis work are presented in Table 4.1. The temperature at which sensing occurs should be above the glass transition temperature 45

PAGE 57

Table 4.1 Glass Transition Temperature of Polymers Polymer T g ( C ) Polybutadine -102 Polyisobutylene -76 Polystyrene 60-93 Polyvinylacetate 30 Polyvinylpyrrolidone 160 of the polymer for equilibrium times to be low. This is because the as the temperature increases the thermal energy in the polymer solvent system is sufficient to overcome molecular forces between the polymer, allowing for sorption [19]. However, too low of a glass transition temperature resulted in a polymer film that was difficult to coat onto the TSM resonators. Additionally, as the polymer sorbs the analyte, properties of the polymer change. In particular the glass transition temperature becomes depressed, depending upon the concentration of the analyte [19]. This is because the analyte has a lubricating effect on the polymer and causes the individual chains in the polymers to move more freely. The net effect is a plasticization of the polymer and a depression of the glass transition temperature. Changes in the shear modulus of the polymer also result from sorption. This is because the viscoelastic properties of the polymer changes in response to the sorbed vapor. The shear modulus is directly related to the rigidness of the polymer and the intermolecular forces within the polymer. Consequently, as these forces change due to sorption, the shear modulus also changes. The motional resistance of the TSM resonator increases as the polymer film becomes softened. This change in the motional resistance can be used to determine the shear modulus. Since the extent of change in the viscoleastic properties of the film varies according to the quantity and 46

PAGE 58

identity of the organic vapor, it was possible to distinguish between the organic vapors by monitoring the motional resistance of the TSM resonator. 4.3 Experimental Results with Polyisobutylene Polyisobutylene and copolymers of polystyrene-butadiene produced ideal sensor responses at room temperatures. Here, results for a polyisobutylene sensing film tested with benzene analyte are presented. Sensor responses to the other six organic vapors are presented in the Appendix C. Each of the resonators in Table 4.2 was exposed to various concentrations of benzene. Figure 4.3 shows the actual TSM resonators used. Table 4.2. Design Parameters of the TSM Devices Resonator (MHz) Blank Diameter (cm) Electrode Diameter (cm) Electrode Thickness (A o ) Electrode Material Milled Area (cm 2 ) 9.9823750 19.969062 96.888522 1.376 0.8077 0.508 0.5105 0.3480 0.0508 1000/100 1000 Au / Cr Au / Cr Au / Cr 0.0507 Figure 4.3 10 MHz (left), 20 MHz (middle), and 96 MHz TSM Devices (right) The frequency responses of the sensor were recorded during exposure to compare the devices sensitivities, limit of detection, and frequency noise. The concentration was 47

PAGE 59

increased in values from 27,000 mg/m 3 to 256,000 mg/m 3 with purges of pure nitrogen gas between each concentration to allow the benzene vapor to desorb and the film to recover. A total flow rate of 100 1 sccm was always maintained over the surface of the device. The temperature of the cell was maintained at 22.5 o C and the benzene vapor was generated at 15 o C. Temperature fluctuation of the cell was within 0.1 o C. Frequency measurements taken using Labview and the Agilent impedance analyzer were accurate to within 1 Hz. The polymer film thickness cannot exceed more than 1 % of the thickness of the quartz blank thickness. If the sensing film exceeds this percentage, the film can no longer be considered ideal, and the Sauerbreys equation will Table 4.3. Coating Guide Fundamental Frequency (MHz) Film Mass (g/cm 2 ) Film Thickness (nm) Coating (kHz) 10 1.9706E-04 1670 45 20 9.8530E-05 835 89 30 6.5687E-05 557 134 40 4.9265E-05 418 178 50 3.9412E-05 334 223 60 3.2843E-05 278 267 70 2.8151E-05 239 312 80 2.4633E-05 209 356 90 2.1896E-05 186 401 100 1.9706E-05 167 445 110 1.7915E-05 152 490 be invalid. Using the 1 % as the guideline for polymer film thickness, Table 4.3 was generated to serve as a guide during coating. Also, sensor response dampens as the polymer film thickness is increased, such that a limit of sensing film thickness exists for 48

PAGE 60

each sensor with a given fundamental resonant frequency. Since different resonators have different blank thicknesses, the thickness of the polymer sensing film cannot be constant; however, the ratio of the thickness of the quartz blank to the thickness of the film can be kept at a constant ratio. One of the requirements for a film to be considered ideal is that the film must be infinitimesally thin. TSM resonators used in this study had varying thickness depending upon the operating frequency. The thickness of the TSM devices and their corresponding sensitivities are shown in Figures 4.4 and 4.5. These figures represent the ideal thickness and resonator sensitivity which are calculated according to the Sauerbreys equation. 0 20 40 60 80 100 120 0 20 40 60 80 100 120 140 160 180 Frequency (MHz) Quartz Thickness ( m) Figure 4.4 Quartz Thicknesses 49

PAGE 61

0 20 40 60 80 100 120 0 0.5 1 1.5 2 2.5 3x 1010 Frequency (MHz) Sensitivity(Hz cm 2/g) Figure 4.5 Resonator Sensitivity The actual thickness parameters corresponding to each resonator are given in Table 4.3. Table 4.4 represents the devices which were actually used in this study and the polymer film thicknesses which were coated onto the quartz. Table 4.4. Film Thickness Resonator Frequency (MHz) Film Thickness (nm) Frequency Shift (kHz) Thickness Ratio 9.993250 19.993400 96.888522 961 420 141 20,112 34,440 275,413 0.58 0.50 0.82 50

PAGE 62

Due to its high sensitivity, the 96 MHz resonator was difficult to coat, consequently, the thickness ratio was higher when compared to the other resonators. The frequency shift values in Table 4.4 represent the total frequency shift due to the mass loading of the polymer coating. In the experiments conducted, the responses of each resonator to various concentrations or volume percentages of benzene were recorded. From this data the resonator sensitivity, detection limit, and frequency noise were determined and compared. 4.3.1 Coating and Chemicals TSM devices were coated according to the spray coating method utilizing a spray brush [15]. In the first test of the apparatus, polyisobutylene was used as the sensing film. The polymer had an average molecular weight of 400,000 and was obtained from Acros Organics. A dilute solution of the polymer (0.1 %), dissolved in chloroform (99.9 % HPLC), was aspirated through an atomizing nozzle using compressed nitrogen gas. The atomized droplets impact the device surface, and the volatile solution evaporates to leave the polymer coating. Polymer coatings formed using this method may have irregular texture and coverage; however, thicknesses were reproducible because the device frequency was monitored throughout the coating process. The resonators were soaked in chloroform and cleaned in a Harrick plasma cleaner prior to coating. Polymers were coated equally onto each side of the 10 MHz, 20 MHz and 96 MHz TSM resonators to achieve various frequency shifts corresponding to different film thicknesses. The device was allowed to air dry after each coating application, and was cured to anneal the film. 51

PAGE 63

Unless handled with care, resonators of approximately 96 MHz frequency were fragile and shattered during spray coating. The frequency and equivalent circuit parameters of the uncoated device were recorded before and after application of the polymer. Also, the resonant frequency of the device was monitored with time throughout the sorption and desorption of the vapors in the polymer. The TSM device is cleaned and coated by following the following procedure: 1. Soak QCM in chloroform for 1 hour. 2. Rinse QCM with deionized water. 4. Air dry QCM with nitrogen gas. 5. Acetone wash QCM. 6. Rinse QCM with deionized water and air dry in nitrogen gas again. 7. Heat QCM at 40 C for 15 minutes. 8. Clean QCM in plasma cleaner for 15 minutes. 6. Record baseline resonant frequency and equivalent circuit parameters prior to spray coating. 7. Spray coat one electrode at a time, while monitoring the resonant frequency. Coat each side with an approximately equivalent frequency. 4.3.2 Sensor Response, Repeatability, and Sensor Parameters Figures 4.6 to 4.8 demonstrate that the resonator responses (frequency changes) due to exposure to the analyte are repeatable. Three trials were conducted with the 10 MHz 20 MHz and 96 MHz resonators. The frequency changes observed in Figures 4.6 to 4.8 can be interpreted in terms of the physical phenomena (Sauerbreys equation); 52

PAGE 64

therefore, the frequency shifts are negative. A stream of ultra high purity grade nitrogen gas was passed over the polymer coated resonator for 1200 seconds; next the resonator was exposed to benzene vapors (0.823 % by volume in nitrogen) for 600 seconds, followed by another exposure to pure nitrogen gas for 600 seconds to remove the absorbed benzene. This procedure was repeated again with increasing volume percentages of benzene; consequently the frequency shifts in the following graphs increased with each increasing volume percentage of benzene vapor. The procedure was chosen to demonstrate the viability of the resonator as a sensor in terms of recovery after exposure the test sample. The inherent increase in sensitivity of the 96 MHz resonator, due to the increase in resonant frequency, is demonstrated by a comparison of the sensor response from each resonator, as shown in Figure 4.9. 0 2000 4000 6000 8000 10000 12000 14000 -4500 -4000 -3500 -3000 -2500 -2000 -1500 -1000 -500 0 500 Time (seconds) Frequency Shift (Hz)0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 % Trial 1Trial 2Trial 3 Figure 4.6 10 MHz Device Response to Benzene Vapors 53

PAGE 65

0 2000 4000 6000 8000 10000 12000 -9000 -8000 -7000 -6000 -5000 -4000 -3000 -2000 -1000 0 1000 Time (seconds) Frequency Shift (Hz)0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 % Trial 1Trial 2Trial 3 Figure 4.7 20 MHz Device Response to Benzene Vapors 0 2000 4000 6000 8000 10000 12000 14000 -12 -10 -8 -6 -4 -2 0 2x 104 Time (seconds) Frequency Shift (Hz)0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 % Trial 1Trial 2Trial 3 Figure 4.8 96 MHz Device Response to Benzene Vapors 54

PAGE 66

0 2000 4000 6000 8000 10000 12000 14000 -12 -10 -8 -6 -4 -2 0 2x 104 Time (seconds) Frequency Shift (Hz)0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 % 10 MHz20 MHz96 MHz Figure 4.9 Comparison of TSM Resonator Responses A noticeable drift in the baseline frequency was observed for all resonators. The base line drift over the duration of the entire experiment was 49 Hz for the 10 MHz device. The 20 MHz and 96 MHz resonators had higher baseline drifts corresponding to 260 and 2342 Hz respectively. Over the entire duration of the experiment it was observed that there was an increase in the baseline drift. However, due to the erratic nature of the baseline drift, this drift did not have an accumulating effect with time. The baseline drift was probably the result of de-wetting effects. De-wetting affects the shape of a thin film polymer by reducing the area of the film/surface interface. During exposure to the analyte the polymer film breaks up into beads of isolated droplets. This effect was also observed during the coating procedure and has also been noted by previous studies [16]. The frequency of a polymer coated resonator was observed to slightly increase when left overnight. This was probably due to absorption of water vapor from the air. 55

PAGE 67

Consequently, the initial rise in resonant frequency during first exposure to pure nitrogen was due to de-sorption of the water. Although it appears from Figure 4.7 that the responses were not repeatable at the highest exposure to benzene, the average difference of the weight fraction at each exposure was calculated to show that the responses were repeatable. This average was found to be 2.5 % (between trial 1 and trial 2), 2.9 % (between trial 1 and trial 2), and 5.3 % (between trial 1 and trial 3). The limit of detection (LOD) and the noise levels for each resonator was also determined from the sensor responses in Figures 4.6 to 4.8. Similar to the baseline drift, the frequency noise increased with the fundamental frequency of each resonator. The frequency noise is important since it determines the detection limit for the sensor. The frequency noise was defined as the standard deviation (S.D.) of the mean resonant frequency taken over a 7 minute period in the presence of 100 sccm of pure nitrogen gas. Using this definition, the noise level for the 10 MHz resonator was 0.444 Hz, 0.900 Hz for the 20 MHz resonator and 3.650 Hz for the 96 MHz resonator. The determination of sensitivity is discussed in the next section. A comparison of the LOD, signal noise, and baseline drift for each resonator is shown in Table 4.5. The LOD was determined from the device sensitivity in the next section. Table 4.5. Comparison of Experimental Sensor Response Parameters Resonator (MHz) Total Baseline Drift (Hz/220 min) Noise (Hz, S.D. Mean Signal) L.O.D (mg/m 3 3Noise/Cf) 9.9823750 19.969062 96.888522 49 260 2342 0.444 0.900 3.650 132 267 256 56

PAGE 68

Noise levels in Table 4.5 were recorded during exposure of the TSM sensors to pure nitrogen gas. During exposure to benzene, however, noise levels were higher due to the interaction of the polymer with the solvent. The frequency noise level generally increase with increasing concentration of benzene. Noise levels at each exposure for the 10 MHz, 20 MHz, and 96 MHz devices are shown in Table 4.6. The values in Table 4.6 were determined by taking the standard deviation of the frequency response after equilibrium was attained. Consequently the first 100 seconds and the last 25 seconds of exposure to benzene were not included in determining the frequency noise during exposure. Table 4.6 Frequency Noise During Benzene Exposure Benzene Percent 10 MHz 20 MHz 96 MHz 0.82 0.374967 1.669667 79.02567 1.63 0.403433 1.055967 4485.133 2.43 0.531033 4.232567 5.593967 3.21 0.7781 11.87353 7982.367 3.98 1.006067 2.656 442.4933 4.74 1.437533 4.1677 9962.567 5.49 2.010967 4.806433 279.6567 6.23 2.504433 5.626867 5653.333 6.95 3.6527 6.6893 21134 The sensor response to the analyte was expected to follow an exponential decay with time. An exponential function would allow for response time constants to be determined using conventional methods. However, the response did not fit an exponential decay. A value for the mean response was taken as the full response. This value was used to determine the time constant for 63 %, 95 % and 99 % response times and 99 % recovery times. Response and recovery times were similar for all seven 57

PAGE 69

organic vapors studied. 63 % response times were similar for all three sensors at all concentrations of benzene except at the lowest concentration. The 63% response time was between 10 and 20 seconds. The 96 MHz devices responded quicker at 0.82 % levels of benzene. There was a decreasing trend with concentration for the 99 % responses of all devices. This can be explained by the unsteady state diffusion of a solvent through a polymer slab. Recovery times ranged from 100 to 180 seconds for all devices, with the 96 MHz devices recovering quicker 4.3.3 Sensitivity There are several methods available for determining sensitivity. The theoretical sensitivity will be compared to each of these experimental sensitivities. The available methods can be summarized as the device or sensor sensitivity and the measured experimental mass sensitivity. In addition to comparing each device with its corresponding theoretical sensitivity, the devices will also be compared with each other to determine the sensitivity improvement. The theoretical sensitivity can be found from the Sauerbreys equation. The experimental mass sensitivity is defined to account for the polymer film effects on the sensors sensitivity and is expressed by Equation 2.12. The device sensitivity (f s /C) does not account for the film effects and is a function of concentration, C. The frequency shift due to the solvent was taken at the point where the sensor response reaches equilibrium. A plot of the experimental sensitivity for each of the three devices, Figures 4.10 and 4.11, shows that there is an obvious increase in the experimental mass sensitivity of higher fundamental frequency resonators. The dependence of the experimental mass sensitivity on concentration is also obvious, with increasing sensitivity towards higher concentrations of benzene vapor. The 58

PAGE 70

improvement in the experimental mass sensitivity for the 96 MHz resonator in comparison with the 10 MHz resonator was a factor of approximately 111 to 126, which was higher than expected and predicted by the Sauerbreys equation. In contrast, the 20 MHz resonator only showed an improvement by a factor of approximately 4 when compared with the 10 MHz resonator. The improvement factors for each exposure to benzene are shown in Table 4.7. 0 0.5 1 1.5 2 2.5x 105 2 3 4 5 6 7 8 9 10 11 12x 108 Concentration (mg/m3) Sensitivity (Hz cm 2 /g) 10 MHz20 MHz Figure 4.10 Experimental Mass Sensitivity of 10 and 20 MHz TSM Resonators Normalizing the experimental sensitivities with respect to the 10 MHz resonator allows for the calculation of the power. This power value represents the value which determines 59

PAGE 71

0 0.5 1 1.5 2 2.5x 105 2.4 2.6 2.8 3 3.2 3.4 3.6x 1010 Concentration (mg/m3) Sensitivity (Hz cm 2 /g) 96 MHz Figure 4.11. Experimental Sensitivity of 96 MHz TSM Resonator the resonator sensitivity. From Equation 2.6, the theoretical power value corresponds to 2; thus, the theoretical sensitivity varies as the square of the resonant frequency. A comparison of the theoretical power with the experimental power is determined by a double logarithmic plot show in Figure 4.12. A linear regression fit to the data yields the experimental power and allows for comparison of experimental and theoretical sensitivity. From this figure, and Table 4.12, the sensitivity dependence upon concentration is also evident. The experimental power values for each volume percent of benzene were obtained from plots similar to Figure 4.12, however, only the power values for the low (0.823 %) and high levels (6.95 %) of the benzene vapor are represented for clarity. 60

PAGE 72

y = 2x 4.6016y = 2.137x 4.950700.511.522.533.544.5522.533.544.55ln(Resonant Frequency) MHzln(Sensitivity Ratio) Theoretical 0.823 % BNZ 6.95 % BNZ Linear (Theoretical) Linear (6.95 % BNZ) Figure 4.12 Theoretical and Experimental Comparison of Sensitivity Table 4.7. Sensitivity Ratio and Sensitivity Dependence on Experimental Power Volume Percent Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Experimental Power 0 0.823 1.63 2.43 3.21 3.98 4.74 5.49 6.23 6.95 1 1 1 1 1 1 1 1 1 1 4.01 4.01 4.02 4.01 4.02 4.03 4.02 4.04 4.04 4.05 110.78 110.85 111.63 112.57 113.81 115.33 117.24 119.01 120.62 126.19 2.076 2.077 2.080 2.084 2.089 2.095 2.103 2.109 2.116 2.137 61

PAGE 73

Figure 4.13 shows the calibration curves (top) for the three resonators used in this study and the device sensitivity (bottom). As mentioned previously, the device sensitivity was defined as the sensor response to a particular analyte concentration. In this case, this is the change in resonant frequency divided by a change in concentration. The calibration curves show that the sensors exhibited nonlinear responses to the benzene vapors. This nonlinear response was also present in the experimental mass sensitivity in Figures 4.10 and 4.11. Furthermore, the power values in Table 4.7 were not constant but increased with higher concentration levels of benzene. Due to the nonlinear nature of the sensor sensitivity, a linear model was needed to determine the device sensitivity. A non-linear regression second order fit was made to the calibration curves, with the intercept set to zero. The regression coefficient was 0.9974 for the 10 MHz device, 0.9976 for the 20 MHz device, and 0.9921 for the 96 MHz device. The first derivative of the fitted equations represents the sensor response to the analyte, which is the device sensitivity. A comparison of the three resonators device sensitivities are shown in the linear plot of Figure 4.13. Since the slope of the device sensitivities of all devices were not zero, the device sensitivity was dependent upon benzene concentration. The device sensitivity in Figure 4.13 can be utilized for determining whether a specific device meets the sensitivity requirements of a particular process and which sensor would perform better at a particular analyte concentration level. The device sensitivity of a 96 MHz resonator operating as a process monitor for detecting 200,000 mg/m 3 is much higher than the sensitivity of a 10 or 20 MHz device. The device sensitivity was used to determine the LOD of the three devices. The LOD was defined as the lowest exposure level where the response is greater than three 62

PAGE 74

times the standard deviation of the noise and was calculated as the noise divided by the device sensitivity. Here, the sensitivity was determined by taking the limit of the device sensitivity as concentration approaches zero. The LOD and noise level for each resonator is given in Table 4.5. Figure 4.14 shows the limit which exists at the point where the sensitivity crosses the y axes. The noise level for the 96 MHz resonator was considerably higher than the 10 MHz resonator, consequently, the LOD did not improve despite the improvement in the sensitivity of higher fundamental frequency resonators. 0 0.5 1 1.5 2 2.5x 105 0 2 4 6 8 10 12x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure 4.13 Benzene Calibration Curve The calibration curves are represented very well by a quadratic function of the form Where is the frequency shift that results from solvent exposure, and C is the vapor phase concentration of the analyte. Parameters for these regressions are given for each sensor exposed to benzene. bCaCf2 f 63

PAGE 75

10 MHz: xxy)10*439.5()10*390.5(328 20 MHz: xxy)10*008.1()10*839.9(328 96 MHz: xxy)10*280.4()10*661.1(226 0 0.5 1 1.5 2 2.5x 105 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Concentration (mg/m3) Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure 4.14 L.O.D. Determination 4.4 Vapor Discrimination Other factors, in addition to frequency changes, contribute to the response of a TSM sensor. Factors such as coating thickness, polymer viscoelasticity, contact surfaces and temperature all affect the total response of a sensor. Electrical characteristics play an important part in the total response of a sensor. The electrical behavior of the sensor can be determined by monitoring the electrical admittance over a frequency range near resonance (accomplished by using the BVD circuit model) with an impedance analyzer. 64

PAGE 76

Figure 2.5 shows the BVD model of a perturbed and unperturbed TSM resonator. A viscoelastic film contributes the additional inductance () and resistance () to the BVD circuit. Physically the changes in the resistance of the film represent changes in dissipation energy and changes in the films inductance represent changes in mass. Changes in these two circuit parameters allow for vapor discrimination. Figure 4.15 shows the changes in the resistance of a polyisobutylene film coated onto a 96 MHz resonator. From figure 4.15 we see that changes in the resistance parameter are repeatable. Similar changes in the resistance of the polyisobutylene coated 10 MHz and 20 MHz resonators were also observed. Increases in the film resistance were coupled by increases in the inductance (decreases in the resonant frequency). 2L 2R 0 2000 4000 6000 8000 10000 12000 -20 0 20 40 60 80 100 120 Time (seconds) Resistance Change,( )0.823 %1.63 %2.43 %3.21 %3.98 %4.74 %5.49 %6.23 %6.95 % Trial 1Trial 2Trial 3 Figure 4.15 Resistance Changes of a PIB Coated 96 MHz Resonator 65

PAGE 77

Solvents such as benzene and toluene are structurally similar (aromatic), whereas hexane and benzene are structurally dissimilar; and solvents such as dichloroethane are polar. The seven solvents selected for studying vapor discrimination were benzene, toluene, hexane, cyclohexane, heptane, dichloroethane, and chloroform at levels ranging from less than 1 to over 10 volume percentage in nitrogen gas. The aim was to determine whether or not it was possible to distinguish between dissimilar as well as similar solvent. Calibration curves can then be generated depending on the extent of discrimination. These calibration curves can be used in sensor detection applications where the identity of several chemicals are known, but the exact chemical which needs to be detected is unknown. For example, consider a storage room filled with benzene, toluene, hexane, cyclohexane, heptane, dichloroethane, and chloroform bottles. The identities of the chemicals are known such that if one of the bottles breaks, the TSM sensor can identify the solvent which was spilled. It should be pointed out that polyisobutylene coated TSM devices do not selectively sorb any of the 8 solvents, consequently, if two bottles broke simultaneously, the TSM sensor would fail to discriminate between the solvents. Figure 4.16 to 4.19 shows the changes in the resistance of polysiosbutylene due to solvents of benzene, hexane and dichloroethane, heptane, chloroform, toluene, and cyclohexane. From Figure 4.16, it is possible to distinguish between the three solvents. In Figure 4.16 to 4.19 the solvents were plotted separately to illustrate the significance of the resistance changes and also to avoid cluttering. Figure 4.16 shows that two straight chain solvents, heptane and hexane, cause similar changes in resistances. However, cyclohexane caused dampened when exposed to 232,945 mg/m^3 of cyclohexane, consequently the remaining cyclohexane resistance curve in Figure 16 was not obtained. Unfortunatley, 66

PAGE 78

0 2000 4000 6000 8000 10000 12000 14000 -50 0 50 100 150 200 250 300 350 400 Time (seconds) Resistance Change, ( ) HexaneCyclohexaneHeptane Figure 4.16 Resistance Changes due to Hexane, Cyclohexane and Heptane 0 2000 4000 6000 8000 10000 12000 14000 -20 0 20 40 60 80 100 120 140 160 180 Time (seconds) Resistance Change, ( ) BenzeneHexaneDichloroethane Figure 4.17 Resistance Changes due to Benzene, Hexane and Dichloroethane 67

PAGE 79

resistance changes are more distinct at high concentration levels where damping occurs. Figure 4.17 shows resistance changes of benzene, hexane and dichloroethane. These three solvents have different physical properties due to their functional groups and chemical structure. A comparison of the resistance changes for these solvents shows that it is relatively easier to distinguish between these solvents. Comparing Figure 4.18 and 4.19 shows that polarity, rather than structure, may be the cause for the differences in resistance. Here, toluene and benzene cause similar resistance changes at low levels, but 0 2000 4000 6000 8000 10000 12000 14000 -100 0 100 200 300 400 500 600 700 Time (seconds) Resistance Change, ( ) BenzeneToluene Figure 4.18 Resistance Changes due to Beznene and Toluene dichloroethane and chloroform cause different changes in the film resistance. The cause for the resistance changes cannot be completely determined by the nature of the solvent alone. To completely determine the cause of these observed resistance changes would require a study of the viscoelastic changes in the polymer film and a study of the 68

PAGE 80

0 2000 4000 6000 8000 10000 12000 14000 -20 0 20 40 60 80 100 120 Time (seconds) Resistance Change, ( ) DichloroethaneChloroform Figure 4.19 Resistance Changes due to Dichloroethane and Chloroform 0 2 4 6 8 10 12 14 16x 104 0 20 40 60 80 100 120 | f| (Hz)| R| ( ) HeptaneChloroformTolueneCyclohexane Figure 4.20 Resistance and Frequency Changes due to Sorption of Heptane, Chloroform, Toluene, and Cyclohexane 69

PAGE 81

0 2 4 6 8 10 12x 104 0 20 40 60 80 100 120 140 160 180 | f| (Hz)| R| ( ) BenzeneHexaneDichloroethane Figure 4.21 Resistance and Frequency Changes due to Sorption of Benzene, Hexane and Dichloroethane interaction of the solvents with polyisobutylene. Using Figure 4.16 to 4.19 and the frequency response of the TSM resonator to each organic vapor (presented in the appendix), Figures 4.20 and 4.21 were obtained. Figures 4.20 and 4.21 show the corresponding change in resistance for a measurable change in the resonant frequency. These two figures were used to generate the calibration curves shown in Figures 4.22 and 4.23. Since the resistance changes and frequency changes are repeatable for a given solvent concentration, these calibration curves can be used to distinguish between organic vapors. The variation of the resistance responses during the exposure time was calculated as the standard deviation of the mean resistance during exposure. This variation allows one to determine the error in the measured resistances. The variation for benzene was 70

PAGE 82

0 0.5 1 1.5 2 2.5 3 3.5x 105 0 20 40 60 80 100 120 140 160 180 Concentration (mg/m3)| R| ( ) BenzeneHexaneDichloroethane Figure 4.22 Vapor Discrimination Curve for Benzene, Hexane and Dichloroethane 0 1 2 3 4 5 6 7x 105 0 20 40 60 80 100 120 Concentration (mg/m3)| R| ( ) HeptaneChloroformTolueneCyclohexane Figure 4.23 Vapor Discrimination Curve for Heptane, Chloroform, Toluene, and Cyclohexane 71

PAGE 83

had an average of 3.43 ohms. Resistance noise was higher at 0.823 % and 6.95 % exposures levels of benzene. Similar resistance variations were noted with the other solvents. 4.5 Temperature Correction Although the TSM resonators are stabilized (according to the cut angle) at different temperatures, the polymer film does not produce the same frequency response at different temperatures. Organic vapors have different solubility at different temperature, being less soluble at higher temperatures. As a result, if the temperature fluctuates the calibration curves, such as figure 4.13 cannot be used to determine the concentration from frequency shift data alone. The actual concentration can only be determined by determining the change in frequency shift due to the change in temperature. A method of determining the effect of temperature change on frequency shift is to model the thermodynamics at the polymer solvent interface. The actual concentration can be determined through the activity of the solvent in the polymer. From vapor liquid equilibrium at the polymer solvent interface, the fugacity of the vapor and the solvent can be equated. solutionvaporff11 (4.1) Where, is the fugacity of the solvent in vapor phase and is the fugacity of the solvent in solution. Accounting for liquid and vapor phase non-ideality introduces the fugacity and activity coefficients into Equation 4.1 vaporf1 solutionf1 lfxPy1111^1 (4.2) Where the partial pressure is given by 72

PAGE 84

11PPy (4.3) The fugacity coefficient for component 1 in the gas phase for a two-component mixture can be determined from the truncated virial equation of state [50] 331113231112expBBByBRTP (4.4) Where R is the gas constant ,are second virial coefficients, ,P is the total system pressure and T is the system temperature. A process measurement of T is required if the temperature correction method is used to determine the actual concentration. The second virial coefficient can be estimated using the corresponding states correlation of Tsonopoulos [51]. 331113,,BBB 11B )()()1()0(11RRccTfTfPRTB (4.5) Here, and ccPT, are the critical temperature, critical pressure and Pitzers accentric factor, respectively. are the reduced temperature and pressure. are given by [51] RRPT, )1()0(,ff 32)0(0121.01358.0330.01445.0)(RRRRTTTTf (4.6) 332)1(0073.0097.05.046.0073.0)(RRRRRTTTTTf (4.7) The fugacity of the liquid is determined by equation 4.8 to 4.10: llfxf1111 (4.8) SatSatlSatSatlPPRTVExpPf11111 (4.9) 73

PAGE 85

SatSatlSatSatlPPRTVExpRTPBExpPf1111111 (4.10) For vapor liquid equilibrium SatSatlSatSatPPRTVExpRTPBExpPxBBByBRTPExpP11111111331113231112 (4.11) Solving for the activity yields SatSatlSatSatPPRTVExpRTPBExpPBBByBRTPExpPax11111133111323111111112 (4.12) Where is the activity coefficient based on weight fraction, is the saturated vapor pressure of the solvent at the system temperature. The saturation pressure is obtained from Wagners equation. Equation 4.12 gives an expression which relates the activity to the temperature. If the activity over a range of temperatures is known for a PIB film, then it is possible to determine the vapor phase concentration. Figure 4.24 shows the activity of benzene in PIB at several temperatures. Using Equations 4.12, one would have to know and activity which means that the temperature and pressure of the system must be measured, and there should be only one known solvent present. The polymer coating on the sensor will be the same, consequently the frequency shift due to the polymer coating,, will also be know. One can calculate the new frequency shift due to the temperature change. With this new information, frequency shift and temperature change, calibration curves such as Figure 4.13 at a new temperature can be used to find the concentration of the analyte. In effect, a second set of calibration curves 1 SatP1 SatT1 1 of 74

PAGE 86

(weight fraction vs. activity) and (concentration vs. frequency at the new temperature) are needed to correct for the changes in frequency due to changes in temperature. In the absence of experimental data, models such as UNIFAC or UNIQUAC may be used to generate approximate corrections (by estimating activity) in the calibration curves due to temperature changes [52]. However, the accuracy of the results may vary depending upon the accuracy of the model. 00.10.20.30.40.50.60.70.80.9100.10.20.30.40.50.6w1a1 T=298 K T=313 K T=338 K T= 283 K Figure 4.24 Activity of Benzene in PIB 75

PAGE 87

CHAPTER 5 CONCLUSIONS In summary it was demonstrated that milled resonators with increased fundamental resonant frequencies can be utilized as vapor sensors despite their fragility. Despite the use of extensive cleaning procedures prior to coating the TSM resonators, baseline drifts due to dewetting effects were observed. Noise levels for higher fundamental frequency resonators were significantly higher when compared to typical TSM devices resulting in an increase of the LOD decreased for higher frequency resonators. Increased noise levels of the 96 MHz resonator contributed to its LOD being higher than expected. Sensor response times were generally quicker for higher frequency devices. It was also determined that the sensor response is dependent upon the concentration of the analyte, being slightly higher at higher concentrations and following a nonlinear trend. There was an increased sensor response for the 96 MHz and 20 MHz resonator when compared to the 10 MHz resonator as seen in the resonant frequency shifts. The experimental analysis of sensitivities showed excellent agreement with theoretical predictions. The lower sensor sensitivity verified that the sensitivity is dependent upon the coating-analyte interaction and the thickness as well as surface area of the polymer. The LOD was directly affected by these effects. For the 20 MHz resonator the experimental mass sensitivity (2.27x 10 10 Hz cm 2 g -1 ) was slightly higher 76

PAGE 88

than the theoretical sensitivity (2.26x 10 10 Hz cm 2 g -1 ). It should be noted that a comparison of the sensitivity of a 96 MHz resonator with a 99 MHz SAW (1.26x 10 10 Hz cm 2 g -1 ) device reveals that the TSM resonator has a higher (2.26x 10 10 Hz cm 2 g -1 ) sensitivity. These higher fundamental frequency TSM resonators offer a new alternative as highly sensitive vapor sensors. In addition to improving upon sensitivity, it was demonstrated the utility of high frequency resonators as vapor sensors for many solvents which span a range of pre-explosive vapor concentration levels. Such sensors may serve as detectors or as process monitors. Table 5 shows the dynamic range of seven organic vapors benzene, hexane, cyclohexane, heptane, dichloroethane, chloroform, and toluene) which were tested with the 96 MHz resonators. Industrial processing streams may contain higher volume Table 5.1 Dynamic Range Dynamic Range (mg/m 3 ) Chemical 10 MHz 20 MHz 96 MHz Benzene 132 231371 267 255094 256 231371 Toluene 78 51,404 134 59840 40 59840 Hexane 334 335614 451 420444 431 335614 Heptane 99 106830 137 136371 49 106830 Cyclohexane 184 232945 161 286494 88 232945 Dichloroethane 118 265623 161 265623 100 265623 Chloroform 461 693076 626 766560 393 693076 77

PAGE 89

percentage levels of these organic vapors which do not fall within the dynamic range. Nevertheless, it is possible to extract a sample from a process stream and dilute the sample to the required volume percentage. This would allow for the detection of an organic vapor outside of the upper limit of the dynamic range. Finally, through the BVD equivalent circuit, it was possible to determine the resistance changes in conjunction with the frequency changes of PIB films. These two pararmeters were used to distinguish between the organic vapors. The changes in resistances is believed to be the result of softening of the PIB film and changes in the viscoelastic properties of the polymer. Future work would include generating more resistance calibration curves for more solvents and to study the viscoelastic behavior of the polymer through the polymer shear modulus. Finally, future work could include using high frequency TSM resonators as liquid sensors. 78

PAGE 90

REFERENCES [1] M. Schweyer, J. Weaver, J.C. Andel, D. McAllister, A Novel Monolithic Piezoelectric Sensor, Proc. Ultrasonics Symposium, 1 (1997) 371-374. [2] S. Martin, Gas Sensing with Acoustic Devices, Proc Ultrasonics Symposium, 1 (1996) 423-434. [3] W. G. Cady, Piezoelectricity, McGraw Hill, New York, 1946. [4] D. S. Ballantine, R. M. White, S. J. Martin, A. J. Ricco, E. T. Zellers, G. C. Frye, H. Wohltjen, Acoustic wave sensors: theory, design, and physico-chemical applications, Academic Press, San Diego, 1997. [5] M. Thompson, D. C. Stone, Surface-launched Acoustic Wave Sensors, Wiley, New York., 1997. [6] PiezoTech, Quartz Crystal Resonators, http://www.piezotech.com/Technical_ Information/techindex.htm, 2004 [7] C. D. Stockbridge, Vacuum Micobalance Techniques, Plenum Press, New York, 1966. [8] M. D. Ward, L. Zuxuan, C. M. Yip, I. S. Joseph, Operation of an ultrasensitive 30 MHz quartz microbalance in liquids, Anal. Chem., 65 (1993) 1546-1551. [9] J. R. Vig, J. Ballato, R. E. Riman, S. Laffey, Etching Quartz Crystals in Anhydrous HF Gas, 1996 IEEE International Frequency Control Symposium (1996) 201-208. [10] S. J. Martin, G. C. Frye, K. Wessendorf, Sensing liquid properties with thickness-shear mode resonators, Sensors and Actuators A, Physical, 44 (1994) 209-218. [11] G. Sauerbrey, The use of quartz oscillators for weighing thin films and for microweighing, Zeitschrift fur Physik, (1959) 206-222. [12] E. Uttenthaler, M. Schraml, J. Mandel, S. Drost, Ultrasensitive quartz crystal microbalance sensors for detection of M13-phages in liquids, Biosensors and Bioelectronics, 16 (2001) 735-743. [13] D. A. Neumeier, Cheical milling of quartz using a solution based on organic solvents and anhydrous hydrofluoric acid, 2002 IEEE International Frequency Control Symposiumand PDA Exhibition (2002) 394-402. 79

PAGE 91

[14] J. R. Hunt, R. C. Smythe, Chemically milled VHF and UHF AT-cut resonators, In Proceedings of the 39 th Annual Frequency Control Symposium (1986) 292-300. [15] G. K. Guttwein, A. D. Ballato, T. J. Lukaszek, VHF-UHF Piezoelectric Resonators. US Patent 3694677, USA, (1972) [16] S. J. Martin. V. E. Granstaff, J. Appl. Phys., 56 (1984) 608. [17] J. F. Rosenbaum, Bulk acoustic wave theory and devices, Artech House, Boston, 1988. [18] A. Ballato, IEEE Trans. Sonics Ultrason., SU-25 (1978) 185-191. [19] D. S. Ballantine, R. M. White, S. J. Martin, A. J. Ricco, E. T. Zellers, G. C. Frye, H. Wohltjen, Acoustic wave sensors: theory, design, and physico-chemical applications, Academic Press, San Diego, 1997. [20] S. J. Martin, G. C. Frye, Polymer Film Characterization Using Quartz Resonators, 1991 Ultrasonics Symposium, 393-398. [21] A. F. Holloway, A. Nabok, M. Thompson, A. K. Ray, D. Crowther, J. Siddiqi, New Method of Vapor Discrimination Using the Thickness Shear Mode (TSM) Resonator, Sensors, 3 (2003) 187-191. [22] V. Mecea, R. V. Bucur, Thin Film Solids, 60 (1979) 73-84. [23] S. J. Martin, V. E. Granstaff, G. C. Frye, Anal. Chem., 63 (1991) 2272-2281. [24] H. Finklea, The Finklea Papers, Pacific Northwest National Laboratory. (1995) [25] J. W. Grate, S. W. Wenzel, R. M. White, Anal. Chem., 63 (1991) 1552. [26] A. W. Snow, W. R. Barger, M. Klusty, H. Wohltjen, N. L. Jarvis, Langmuir, 2 (1986) 513. [27] J. W. Grate, S. W. Wenzel, R. M. White, Anal. Chem., 64 (1992) 413. [28] S. J. Martin, G. C. Frye, S. D. Senturia, Anal. Chem., 66 (1994) 2201. [29] S. J. Martin, G. C. Frye, Appl. Phys. Let, 57 (1990) 1867. [30] J. Tian, Comparative Solubility Study of C60 and C60-Piperazine and Applications on Quartz Crystal Microbalance/Heat Conduction Calorimeter, Ph.D., Chemistry, Drexel University, 2002. 80

PAGE 92

[31] A.F. Holloway, A. Nabok, M. Thompson, A.K. Ray, T. Wilkop, Impedance analysis of the thickness shear mode resonator for organic vapour sensing, Sensors and Actuators B, 99 (2004) 355-360. [32] F. A. Graybill, An Introduction to Linear Statistical Models, McGraw Hill, new York, 1961. [33] M. Thompson, D. C Stone, Surface-Launched Acoustic Wave Sensors: Chemical Sensing and Thin Film Characterization, Wiley and Sons, Toronto, 1997. [34] R. S. Barratt, The Preparation of Standard Gas Mixtures, A Review, The Analyst, 106 (1981) 817-849. [35] J. P. Lodge, Air Pollution, Academic Press, New York, 1968. [36] D. P. Lucero, Calibration in Air Monitoring, 7, Philadelphia, American Socieity for Testing and Materials. (1976) 301. [37] H. L. Kusnetz, B. E. Saltzman, M. E. Lanier, Am. Ind. Hyg. Assoc. J., 21 (1960) 361. [38] P. E. Caplan, Calibration of Air Sampling Instruments, Cincinnati, OH, Air Sampling Instruments for Evaluation of Atmospheric Contaminants (1966) [39] A. F. Smith, Standard Atmospheres, Exeter, England, Heineman Educational Books. 1981, ch 7. [40] H. D. Axelrod, J. P. Lodge, Sampling and Calibration of Gaseous Pollutants, San Francisco, Academic Press. III (1976) [41] J. Namiesnik, J. Chromatog., 300 (1984) 79. [42] W. J. Woodfin, Am. Ind. Hyg. Assoc. J., 45 (1984) 138. [43] S. W. Dixon, J. F. Vasta, L. T. Freeland, D. J. Calvo, R. E. Hemingway, Am. Ind. Hyg. Assoc. J., 45 (1984) 99. [44] M. M. Bownik, J. Namiesnik, L. Tores, Chromatographia, 17 (1983) 503. [45] S. P. Berardinelli, E. S. Moyer, R. C. Hall, Am. Ind. Hyg. Assoc. J., 51 (1990) 595. [46] D. D. Irish, E. M. Adams, Ind. Med. Surg., 1 (1940) 1. [47] F. W. Sunderman, J. F. Kincaid, W. Kooch, E. A. Bermelin, Am. J. Pathol., 26 (1956) 1211. 81

PAGE 93

[48] G. O. Nelson, K. S. Griggs, Rev. Sci. Instr., 39 (1968) 927. [49] R. C. Reid, J. M. Prausnitz, B. E. Poling, The Properties of Liquids and Gases, McGraw Hill, New York, 1987. [50] J. M. Prausnitz, R. N. Lichtenthaler, E. G. de Azevedo, Molecular Thermodynamics of Fluid Phase Equilibria, Prentice Hall, Englewood Cliff, New Jersey, 1986. [51] C. Tsonopoulos, AIChE J., 20 (1974) 263-272. [52] Wong, H.C., S.W. Campbell, and V.R. Bhethanabotla, Sorption of Bezene, Tetrahydrofuran and 2-butanone by poly(vinyl acetate) at 323.15K Using a Quartz Crystal Balance, Fluid Phase Equilibria, 179 (2000) 181-191. 82

PAGE 94

APPENDICIES 83

PAGE 95

Appendix -A Solvent Concentrations The following concentrations of solvents generated. Percentage levels in nitrogen gas are also listed. Table A1. Solvent Concentrations: Dichloroethane, Heptane, Chloroform Dichloroethane Heptane Chloroform Concentration (mg/m 3 ) Volume (%) Concentration (mg/m 3 ) Volume (%) Concentration (mg/m 3 ) Volume (%) 0 0.00 0 0.00 0 0.00 28171 0.67 15598 0.37 98460 1.95 55966 1.34 31083 0.73 193153 3.83 83391 1.99 46454 1.10 284293 5.63 110455 2.64 61713 1.46 372074 7.37 137164 3.28 76861 1.81 456680 9.04 163525 3.91 91900 2.17 538279 10.66 189545 4.53 106830 2.52 617029 12.22 215230 5.14 121654 2.87 693078 13.73 240588 5.75 136371 3.22 766560 15.18 265623 6.35 150984 3.56 837605 16.59 84

PAGE 96

Appendix A (Continued) Table A2. Solvent Concentrations: Benzene, Toluene Benzene Toluene Concentration (mg/m 3 ) Volume (%) Concentration (mg/m 3 ) Volume (%) 0 0 0 0 27414 0.82 8663 0.22 54378 1.63 17287 0.44 80901 2.43 25873 0.66 106994 3.21 34421 0.88 132669 3.98 42931 1.1 157934 4.74 51404 1.32 182800 5.49 59840 1.54 207276 6.23 68239 1.75 231371 6.95 76601 1.97 255094 7.72 84927 2.18 85

PAGE 97

Appendix A (Continued) Table A3. Solvent Concentrations: Cyclohexane, Hexane Cyclohexane Hexane Concentration (mg/m 3 ) Volume (%) Concentration (mg/m 3 ) Volume (%) 0 0 0 0 30887 0.87 52053 1.43 61242 1.72 102640 2.82 91079 2.56 151822 4.17 120412 3.38 199658 5.48 149253 4.19 246201 6.75 177614 4.99 291504 8 205508 5.77 335614 9.21 232945 6.54 378580 10.39 259937 7.3 420444 11.53 286494 8.05 461248 12.65 Table A4. Wagner Equation Constants Solvent VPA VPB VPC VPD Benzene -6.9827 1.33213 -2.6286 -3.334 Chloroform -6.9555 1.16625 -2.1397 -3.4442 n-Hexane -7.4677 1.44211 -3.2822 -2.5094 Dichloromethane -7.3574 2.17546 -4.0704 3.50701 Dichloroethane -7.3686 1.76727 -3.343 -1.4353 Toluene -7.2861 1.38091 -2.8343 -2.7917 Cyclohexane -6.9601 1.31328 -2.7568 -2.4549 Heptane -7.6747 1.37068 -3.5362 -3.2024 86

PAGE 98

Appendix B Printed Circuit Boards, Lid Designs, and Wiring Diagram PRESENSITIZED BOARD ETCHING The following method was used to make printed circuit boards. 1) Make the layout for obtaining the pattern in AutoCAD or any other suitable designer 2) Obtain (print) the layout on a transparency. NOTE: The PCB has positive photo resist, so exposed areas are removed while developing. 3) Prepare the developer in LAB 120D or any suitable place. Process to prepare developer: Ingredients: NaOH (6-9) gms DI water 1000 ml = 1 litre Beaker Mix the NaOH and DI water and stir it using magnetic stirrer till all the NaOH crystals are dissolved. Note: The higher the amount of NaOH the faster the developing time. PROCESS TO BE DONE IN CLEAN ROOM 4) Then take the PCB and with the cutter cut into suitable dimension 5) Then take the PCB and place the transparency pattern on it 87

PAGE 99

Appendix B (Continued) 6) Place the entire system on the exposing module available in the MEMS LAB-152, with the side to be exposed facing the light through the transparency. 7) Close the box and set the number of lamps to 5 and set the timer to 3. 8) Then after the time elapses open and put the PCB in solution (developer) 9) It takes anywhere from 1-2 minutes for developing which can be clearly seen by the removal of the green photo resist thus exposing the Copper. NaOH solution which is clear also turns green due to this. Note: Do not use old solution of developer as it losses its ability to etch seen previously while doing trial etch. Figure B.1 PCB Design 88

PAGE 100

Appendix B (Continued) Sensor Cell Lids Figures B.2 and B.3 are the lids used to house the high frequency resonators. These lids were needed because each resonator base had a different dimension. All lids were used along with the sensor cell. Figure B.2 Cell Design: High Frequency Resonator Lid 1 Figure B.3 Cell Design: High Frequency Resonator Lid 2 89

PAGE 101

Appendix B (Continued) Vapor Generation Apparatus Wiring Diagram Figure B.4 is the wiring diagram for the vapor generation apparatus. It shows the electrical connectivity for the solenoid valves, relays and mass flow controllers. V 5 V 4 V 3 V 2 V 1 V 6 V 7 V 8 V 9 R 1 R 7 R 6 R 5 R 4 R 3 R 2 R 8 R 9BUSBAR ACACACACAC NNNNN LL ACACACACNNNNACACACACNNNNNACACACACACACAC120 V AC120 V ACNNNNNN PURGE V 5To V 4To V 3To V 2To V 1To V 6To V 7To V 8To V 9 123424 TO 330VAC4 TO 28 V DCCONTROL R 2123424 TO 330VAC4 TO 28 V DCCONTROL R 3123424 TO 330VAC4 TO 28 V DCCONTROL R 4123424 TO 330VAC4 TO 28 V DCCONTROL R 5123424 TO 330VAC4 TO 28 V DCCONTROL R 6123424 TO 330VAC4 TO 28 V DCCONTROL R 7123424 TO 330VAC4 TO 28 V DCCONTROL R 8123424 TO 330VAC4 TO 28 V DCCONTROL R 9123424 TO 330VAC4 TO 28 V DCCONTROL-+--------+++++++ + SCB 100PULLOUT MKS 247 READ OUT MFC 1MFC 2MFC 3 Figure B.4 Solenoid Valves, Relays, MFCs Wiring Diagram 90

PAGE 102

Appendix -C Frequency Responses and Calibration Curves The following pages contain the calibration curves for hexane, dichloroethane, cyclohexane, heptane, chloroform and toluene. The calibration curves were fitted to a second order polynomial equation, which was used to determine the L.O.D. Additionally the power of the experimental sensitivity was determined for comparison to the Sauerbrey equation, where the power has a theoretical sensitivity value of 2.0. C.1 Hexane 0 0.5 1 1.5 2 2.5 3 3.5x 105 0 1 2 3 4 5 6 7 8 9 10x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure C.1.1 Hexane Calibration Curve 91

PAGE 103

Appendix C (Continued) 0 0.5 1 1.5 2 2.5 3 3.5x 105 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Concentration (mg/m3)10 MHz LOD=334 mg/m320 MHz LOD=451 mg/m396 MHz LOD=431 mg/m3 Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure C.1.2 Hexane LOD Table C.1.1 Hexane Experimental Sensitivity Power Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Power 1 4.006092581 110.7922 2.076 1 4.007789769 111.3474 2.0789 1 3.996586747 112.5446 2.0843 1 3.984592074 114.5686 2.0931 1 3.976928971 116.8289 2.1025 1 3.968650066 119.6344 2.114 1 3.957851337 123.1386 2.1279 1 3.958644691 130.613 2.1558 92

PAGE 104

Appendix C (Continued) Table C.1.2 Hexane Calibration Coefficients Sensor Device A B Regression Coefficient 10 MHz 20 MHz 96 MHz 1.296 x 10 -8 2.056 x 10 -8 7.319 x 10 -6 3.993 x 10 -3 5.985 x 10 -3 2.542 x 10 -2 0.9942 0.9997 0.9997 93

PAGE 105

Appendix C (Continued) C.2 Dichloroethane 0 0.5 1 1.5 2 2.5x 105 0 1 2 3 4 5 6 7x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure C.2.1 Dichloroethane Calibration Curve Table C.2.1 Dichloroethane Experimental Sensitivity Power Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Power 1 4.006092581 110.7922 2.0766 1 3.979718786 110.2129 2.0749 1 3.950033565 110.2398 2.0758 1 3.904404187 110.347 2.0775 1 3.887716041 110.975 2.0807 1 3.851802231 110.7106 2.0805 1 3.824031532 110.8948 2.0821 1 3.80627373 111.3268 2.0845 1 3.820158741 112.2926 2.0882 1 3.819833575 113.8713 2.0948 94

PAGE 106

Appendix C (Continued) 0 0.5 1 1.5 2 2.5x 105 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 Concentration (mg/m3)10 MHz LOD=118 mg/m320 MHz LOD=402 mg/m396 MHz LOD=100 mg/m3 Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure C.2.2 Dichloroethane LOD Table C.2.2 Dichloroethane Calibration Coefficients Sensor Device A B Regression Coefficient 10 MHz 20 MHz 96 MHz 2.147 x 10 -8 5.486 x 10 -8 6.217 x 10 -7 1.129 x 10 -2 6.720 x 10 -3 1.096 x 10 -1 0.9982 0.9930 0.9983 95

PAGE 107

Appendix C (Continued) C.3 Cyclohexane 0 0.5 1 1.5 2 2.5x 105 0 2 4 6 8 10 12x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure C.3.1 Cyclohexane Calibration Curve Table C.3.1 Cylcohexane Experimental Sensitivity Power Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Power 1 4.006092581 110.7922 2.0766 1 4.035279345 111.9818 2.0809 1 4.031769594 113.0279 2.0854 1 4.012833552 113.8392 2.0893 1 3.992790212 115.7656 2.0978 1 3.97364845 117.1396 2.1039 1 3.929931317 118.6017 2.1109 1 3.905744949 121.2183 2.1219 1 3.903955669 125.1163 2.137 96

PAGE 108

Appendix C (Continued) 0 0.5 1 1.5 2 2.5x 105 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Concentration (mg/m3)10 MHz LOD=184 mg/m320 MHz LOD=161 mg/m396 MHz LOD=88 mg/m3 Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure C.3.2 Cyclohexane LOD Table C.3.2 Cyclohexane Calibration Coefficients Sensor Device A B Regression Coefficient 10 MHz 20 MHz 96 MHz 6.215 x 10 -8 6.557 x 10 -8 1.528 x 10 -6 7.320 x 10 -3 1.678 x 10 -2 1.245 x 10 -1 0.9992 0.9963 0.9973 97

PAGE 109

Appendix C (Continued) C.4 Heptane 0 2 4 6 8 10 12x 104 0 1 2 3 4 5 6 7x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure C.4.1 Heptane Calibration Curve Table C.4.1 Heptane Experimental Sensitivity Power Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Power 1 4.006092581 110.7922 2.0766 1 4.008156013 111.4187 2.0792 1 3.995956961 112.2108 2.0829 1 3.988169856 113.2419 2.0875 1 3.976770506 114.2345 2.0919 1 3.968535392 115.5249 2.0974 1 3.964048175 116.8384 2.1029 1 3.964899768 120.0723 2.1158 98

PAGE 110

Appendix C (Continued) 0 2 4 6 8 10 12x 104 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 Concentration (mg/m3)10 MHz LOD=99 mg/m320 MHz LOD=137 mg/m396 MHz LOD=49 mg/m3 Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure C.4.2 Heptane LOD Table C.4.2 Heptane Calibration Coefficents Sensor Device A B Regression Coefficient 10 MHz 20 MHz 96 MHz 1.138 x 10 -7 1.903 x 10 -7 3.377 x 10 -6 1.347 x 10 -2 1.973 x 10 -2 2.230 x 10 -1 0.9993 0.9997 0.9978 99

PAGE 111

Appendix C (Continued) C.5 Chloroform 0 1 2 3 4 5 6 7x 105 0 2 4 6 8 10 12 14 16x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure C.5.1 ChloroformCalibration Curve Table C.5.1 Chloroform Experimental Sensitivity Power Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Power 1 4.006092581 110.7922 2.0766 1 4.008667569 109.1832 2.0696 1 4.004377473 108.4111 2.0664 1 4.002752783 110.02 2.0734 1 3.993536464 111.5654 2.0803 1 3.980510677 112.6109 2.085 1 3.977675642 114.0945 2.0913 1 3.979081228 115.552 2.0973 1 4.01426139 119.359 2.1325 100

PAGE 112

Appendix C (Continued) 0 1 2 3 4 5 6 7x 105 0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 Concentration (mg/m3)10 MHz LOD=461 mg/m320 MHz LOD=626 mg/m396 MHz LOD=393 mg/m3 Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure C.5.2 Chloroform LOD Table C.5.2 Chloroform Calibration Coefficients Sensor Device A B Regression Coefficient 10 MHz 20 MHz 96 MHz 8.303 x 10 -9 1.507 x 10 -8 2.581 x 10 -7 2.892 x 10 -3 4.310 x 10 -3 2.789 x 10 -2 0.9997 0.9984 0.9916 101

PAGE 113

Appendix C (Continued) C.6 Toluene 0 1 2 3 4 5 6x 104 0 1 2 3 4 5 6 7 8 9x 104 Concentration (mg/m 3) Frequency Shift (Hz) 10 MHz20 MHz96 MHz10 MHz Fit20 MHz Fit96 MHz Fit Figure C.6.1 Toluene Calibration Curve Table C.6.1 Toluene Experimental Sensitivity Power Cf10/Cf10 Cf20/Cf10 Cf96/Cf10 Power 1 4.006092581 110.7922 2.0766 1 4.004783884 111.2322 2.0785 1 4.005457514 112.4637 2.0837 1 4.001840487 113.8792 2.0897 1 4.10520453 117.9727 2.1037 1 4.115215978 119.7032 2.1103 1 4.166150549 122.5055 2.1199 1 4.21521816 127.4581 2.1374 102

PAGE 114

Appendix C (Continued) 0 1 2 3 4 5 6x 104 0 0.5 1 1.5 2 2.5 3 Concentration (mg/m3)10 MHz LOD=78 mg/m320 MHz LOD=134 mg/m396 MHz LOD=40 mg/m3 Sensitivity (Hz/mg/m 3) 10 MHz20 MHz96 MHz Figure C.6.2 Toluene LOD Table C.6.2 Toluene Calibration Coefficients Sensor Device A B Regression Coefficient 10 MHz 20 MHz 96 MHz 5.112 x 10 -7 1.633 x 10 -6 1.933 x 10 -5 1.701 x 10 -2 2.014 x 10 -2 2.755 x 10 -1 0.9862 0.9992 0.9972 103