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 001478770
007 cr mnu|||uuuuu
008 040811s2004 flua sbm s000|0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0000412
Manoosingh, Lane Leslie.
Design of a chemical agent detector based on polymer coated surface acoustic wave (SAW) resonator technology
h [electronic resource] /
by Lane Leslie Manoosingh.
[Tampa, Fla.] :
University of South Florida,
Thesis (Ph.D.)--University of South Florida, 2004.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 113 pages.
ABSTRACT: This dissertation presents the design of a unique prototype chemical agent detector which utilizes an array of polymer coated SAW resonators as the sensor elements. The design's particular embodiment is that of a testing platform for evaluating the utility of constructing a portable chemical agent detector, utilizing commercially available SAW resonators. It involves the consolidation of the sub-systems comprising a large laboratory development system, into a portable enclosure. A combination of design techniques, utilized to achieve an overall balance between the physical dimensions of the system and its detection performance, comprises the unique nature of the overall design of this detection system. Such techniques include; sensor power cycling, individually phase-tunable sensor oscillators, single step down conversion and the locality of the sensor's driving circuitry and sensing chamber. A frequency shift model is developed to characterize the device's response to target analytes. Reported here are the results of the preliminary tests of the detector system and the verifications of the device's operation as per the design requirements. Further, an assay of the system noise is undertaken, and the detector's limit of detection (LOD) is reported. The analytes used in this investigation were simulants of nerve and mustard gas as well as the interferent compound diesel. Among others, the following conclusions are reported: 1) that a mass loading model can adequately describe the frequency shifts of the SAW resonators utilized for sorption sensing; 2) that the quality factor of a polymer coated SAW resonator ultimately determines the noise performance of the driving oscillator; 3) that the lowest usable quality factor for the designed oscillator is 2500; 4) that the implementation of individual phase-tuning networks for each sensor in the sensor array can adequately compensate for phase variations among these sensors, and 5) that commercially available SAW resonators coated with chemo-selective polymers provide a reasonably inexpensive and reliable solution to the detection of chemical warfare agents when incorporated into a miniaturized sensing platform.
Adviser: Wiley, Paris H.
x Electrical Engineering
t USF Electronic Theses and Dissertations.
Design of a Chemical Agent Detector Based on Polymer Coated Surface Acoustic Wave (SAW) Resonator Technology by Lane Leslie Manoosingh A dissertation submitted in partial fulfillment of the requirements of the degree of Doctor of Philosophy Department of Electrical Engineering College of Engineering University of South Florida Major Professor: Pa ris H. Wiley, Ph.D. Kenneth A. Buckle, Ph.D. Christos Ferekides, Ph.D. O. Geoffrey Okogbaa, Ph.D. Pritish Mukherjee, Ph.D. Date of Approval: June 18, 2004 Keywords: sensor, oscillator, warfare, nerve, portable Copyright 2004, Lane Leslie Manoosingh
DEDICATION This dissertation is dedicated to my pa rents, Jasmin and Ralph Manoosingh and to my sister Celine Manoosingh. Their love, suppor t and patience have always been the foundation of my inspiration.
ACKNOWLEDGEMENTS The author wishes to thank his adviso r, Dr. Paris H. Wiley for his guidance, support and caring throughout the co urse of the research and pr eparation of this thesis. The author would also like to express his sincere appreciation for his employer and his coworkers Mark Tesone, Brian Harm on, Richard Sorrels and Chris Traynor for their friendship and guidance. Finally, the author wishes to thank Dr Tim Postelwaithe, Dr. William Swartz and Sal Barone for their patience, faith and trust.
i TABLE OF CONTENTS LIST OF TABLES............................................................................................................ iv LIST OF FIGURES.............................................................................................................v ABSTRACT......................................................................................................................v ii CHAPTER 1: INTRODUCTION........................................................................................1 CHAPTER 2: REVIEW OF LITERATURE.......................................................................5 2.1 Selection of Background Research................................................................5 2.2 Review of Theoretical Work..........................................................................6 2.3 Review of Experimental Work....................................................................15 CHAPTER 3: CHEMISTRY OF DETECTION................................................................20 3.1 Chemistry of Polymer Adsorption...............................................................20 3.2 Chemistry and Definition of Target Analytes and Simulants......................22 CHAPTER 4: SAW RESONA TOR CHARACTERISTICS.............................................25 4.1 General Description of the SAW Resonator................................................25 4.2 Electrical and Mech anical Characteristics ..................................................27 4.3 Polymer Coated SAW Resonator for Vapor Sensing..................................29
ii CHAPTER 5: FREQUENCY SHIFT MODEL.................................................................31 5.1 Frequency Shift via Mass Loading and Polymer Modulus Change ...........31 5.2 Frequency Shift Model................................................................................35 CHAPTER 6: THE CHEMICAL WARFARE AGENT DETECTOR..............................37 6.1 The Bench-Top Sensing System..................................................................37 6.2 Design Specifications for the Portable Detection System...........................40 6.3 Elements of the Portable Design .................................................................42 CHAPTER 7: THE DE TECTOR OSCILLATOR.............................................................45 7.1 SAW Sensor Oscillator Design....................................................................45 7.2 Sensor Oscillator Tuning.............................................................................50 7.3 Physical Oscillator Implementation.............................................................53 CHAPTER 8: THE MULTIPLE-SE NSOR DETECTOR DESIGN.................................58 8.1 The Multiple Sensor Module.......................................................................58 8.2 Sensor Power-Cycling and Signal Multiplexing..........................................61 8.3 Signal Down Conversion.............................................................................67 8.4 Power Supply Noise.....................................................................................68
iii CHAPTER 9: EXPERIME NTAL VERIFICATION.........................................................70 9.1 Simulant Vapor Testing...............................................................................70 9.2 Experimental Data Â– Vapor Dosing Results................................................71 CHAPTER 10: RESULTS AND CONCLUSIONS..........................................................77 10.1 Results and Discussion .............................................................................77 10.2 Conclusions................................................................................................79 10.3 Future Investigations..................................................................................82 CHAPTER 11: SUMMARY..............................................................................................85 REFERENCES..................................................................................................................87 APPENDICES...................................................................................................................91 Appendix A: DMMP Vapor Dosing Conversions and Characteristics...............92 Appendix B: Chesler-Cram Model Fit.................................................................93 Appendix C: Summary of SAW Sensor and Polymer Parameters......................94 Appendix D: Real Time Analyte Dosing Responses...........................................95 Appendix E: Theoretical vs. Measured Frequency Responses............................98 ABOUT THE AUTHOR.......................................................................................End Page
iv LIST OF TABLES Table 1 SAW Sensor Responses for Various Analyte Concentrations........................18 Table 2 Selectivity Template for SAW Sensor Array Sp,a...........................................36 Table 3 General Specifications for the Portable Detection System..............................40 Table 4 Measured S-Para meters for Driving Amplifier...............................................46 Table 5 Summary of Interact ion Data for Multiple Sensor Module............................61 Table 6 Interaction Data for th e Multiple Sensor Module using Power-Cycling........65 Table 7 Average Maximum GB Sensor Response to DMMP.....................................72 Table 8 Average Maximu m GD Sensor Response to DMMP.....................................72 Table 9 Response for HY Sensor on Unit A, B, C and D for Diesel (50 mg/m3)........73 Table 10 Binary Dosing: Average Maximu m Signal Shifts for Unit A, B, C and D....73 Table 11 Measured Selectivity Pa rameters for the SAW Sensor Array........................74 Table 12 Design Criteria Check List.............................................................................80 Table 13 DMMP Va por Delivery Parameters...............................................................92 Table 14 Vapor Pressure and Volati lity Data for Chemical Warfare Agents................92 Table 15 Summary of SAW Sensor and Polymer Calculation Parameters...................94
v LIST OF FIGURES Figure 1 General SA W Detector Architecture..............................................................10 Figure 2 Weak Hydrogen B onding between Analyte and Polymer..............................22 Figure 3 Functional Molecular Structur e of Soman Nerve Agent and Blister Agent...23 Figure 4 Frequency Shift vs Loading Mass on SAW Sensor (n=1, fo= 311.5-MHz)..33 Figure 5 Bench-Top Chemical Sensing System...........................................................38 Figure 6 Block Diagram of th e Portable Chemical Agent Detector.............................44 Figure 7 SAW Sensor Oscillator Architecture..............................................................47 Figure 8 Oscillator Phase Noise vs. Sensor Ql..............................................................49 Figure 9 SAW Sensor Lumped Element Model...........................................................50 Figure 10 Resonance Modes for a De-tuned 312-Mhz SAW Sensor Module................51 Figure 11 Resonance Mode of Phas e-Tuned 312-Mhz SAW Sensor Module...............52 Figure 12 Measured Phase Noise for Sensor 1 with Coat ed SAW Resonator................56 Figure 13 SAW Se nsor Oscillator...................................................................................57 Figure 14 Multiple SAW Sensor Module.......................................................................60 Figure 15 Phase Noise for Sens or Oscillator 1 (SAW Sensor Ql = 3200)......................62
vi Figure 16 Oscillator Start-up Characteristics.................................................................64 Figure 17 Sensor Oscillator-1 Phas e Noise for Power Cycled Sensor Module.............65 Figure 18 Transition Charac teristic for Sensor Switching.............................................66 Figure 19 Chesler-Cram Model Fit for Sensor GB to DMMP at 1.93 mg/m3...............93 Figure 20 Device A: DMMP Concentrations of .97, 1.93 and 3.86 mg/m3 ..................95 Figure 21 Device B: Diesel at 50 mg/m3.......................................................................95 Figure 22 Device C: 3.86 mg/m3 DMMP + 50 mg/m3 diesel........................................96 Figure 23 Device D: DMMP at 1.93 mg/m3..................................................................96 Figure 24 Device D: DMMP Concentrations of .97, 1.93 and 3.86 mg/m3...................97 Figure 25 Device B: DMMP 1.93 mg/m3 + 50 mg/m3 diesel........................................97 Figure 26 Average Theoretical vs. Measured Frequency Response..............................98 Figure 27 Theoretical vs. Measured Response for Low Dose Concentrations..............99 Figure 28 Theoretical vs. Measured Response for High Dose Concentrations...........100 Figure 29 Measured Frequenc y Response for Detector Units d1, d2, d3 and d4...........101
vii DESIGN OF A CHEMICAL AGENT DETECTOR BASED ON POLYMER COATED SURFACE ACOUSTIC WAVE (SAW) RESONATOR TECHNOLOGY Lane Leslie Manoosingh ABSTRACT This dissertation presents the design of a unique prototype chemical agent detector which utilizes an array of poly mer coated SAW resonators as the sensor elements. The designÂ’s particular embodiment is that of a testing platform for evaluating the utility of constructing a portable chemical agent detector, utilizing commercially available SAW resonators. It involves the c onsolidation of the s ub-systems comprising a large laboratory development system into a portable enclosure. A combination of design techniques, util ized to achieve an overall balance between the physical dimensions of the system and its detection performance, comprises the unique nature of the overall de sign of this detection system. Such techniques include; sensor power cy cling, individually phase-tunable sensor oscillators, single step down conversion and th e locality of the sensorÂ’s driving circuitry and sensing chamber. A frequency shift model is developed to characterize the deviceÂ’s response to target analytes. Reported here are the result s of the preliminary tests of the detector system and the verifications of the device Â’s operation as per the design requirements. Further, an assay of the system noise is unde rtaken, and the detecto rÂ’s limit of detection (LOD) is reported. The analytes used in this investigation were simulants of nerve and mustard gas as well as the interferent compound diesel.
viii Among others, the following conclusions are reported: 1) that a mass loading model can adequately describe the frequency sh ifts of the SAW resonators utilized for sorption sensing; 2) that the quality factor of a polymer coated SAW resonator ultimately determines the noise performance of the drivi ng oscillator; 3) that the lowest usable quality factor for the designed oscillator is 2500; 4) that the implementation of individual phase-tuning networks for each sensor in the sensor array can adequately compensate for phase variations among these sensors, a nd 5) that commercia lly available SAW resonators coated with chemo-selective pol ymers provide a reasona bly inexpensive and reliable solution to the detection of chemical warfare agents when incorporated into a miniaturized sensing platform.
1 CHAPTER 1 INTRODUCTION The detection of chemical agents, pa rticularly those harmful to human physiology, has interested sensor investigator s for decades. Heightened awareness of the necessity for chemical detectors in the embodime nt of first responder or first alert devices has been effected by recent terrorist activities throughout the world. Specifically, sensor technologies exhibiting sub-second, low dosage sensitivity to common chemical warfare agents, such as nerve gas and blister agents are of key interest. The utilization of chemical agents as weapons within the real m of military conflict and attacks directed toward civilian populations is not new. Nerv e gas and blister agent weapons have been utilized since the nineteen forties. However, chemical detection systems combining recent advancements in micro-fabrication t echniques, molecular polymer chemistry and miniaturized electronics conti nue to push the limits and me thodologies for detecting these threatening substances. Surface Acoustic Wave (SAW) devices have been utilized as sensors for measuring gases for many years. The prope rties of SAW devices, which make them attractive to researchers of sensor technol ogies, are numerous. These include high quality factors at resonance, comparably low inse rtion losses and ease of manufacturibility, reproducibility and characteriza tion. When coupled with ch emo-selective polymer based coatings, these devices have exhibited their usefulness as highly se nsitive gas sensors. Recent efforts, directed toward the utilizat ion of polymer coated SAW devices as the primary sensor of gaseous chemical warfare agents have met with great success.
2 The laboratory bench-top testing systems utilized in the assay of these sensors have provided a reasonable level of confidence in the detection concept. The usual test configuration is comprised of bench-top power supplies, freq uency counters, spectrum or network analyzers and several interconnected modular test devices for resonating the SAW sensors. The migration of this general test conf iguration to a useful portable detector platform with the potential for mass produc tion has only recently become realizable. Response variations from unit to unit s hould be significantly reduced with newly developed and well characterized, highly selective polymers. The availability of recently developed, low cost commercial SAW de vices exhibiting largely consistent characteristics from device to device is al so key. Internationally, government agencies have expressed keen interest in the potential for such dete ctor devices and variations thereof for placement in office buildings, trai n or subway stations and schools. Other potential applications of interest would s ee these devices installed in unmanned aerial vehicles (UAV) as a diagnostic sensor to alert operators of possible aircraft contamination. Finally, their possible use with in the embodiment of a personal wearable chemical alert device for soldiers in the field has also been discussed. This dissertation describes the design of a unique prototype handheld chemical agent detector which utilizes polymer coated, SA W resonators as the sensor element. It is through this perspective that th is material conveys the existi ng state of the broader field of investigation encompassing polymer coated SAW sensing. Also, this dissertation was in part a pr eliminary evaluation of the possibility of designing a portable chemical agent dete ctor around commercially available SAW sensors. Further, the polymers utilized in this investigation were representative of a class of selective polymers with suitable density, selectivity and sorption characteristics. The mechanism of detection upon which th e design was based was the resonant frequency shift exhibited by a SAW device when mass loaded. This mass loading is defined by a differential mass change of the SAW polymer coating upon sorbtion-desorption of target analyte molecules during gas probing.
3 This dissertation also de scribes another possible m echanism effecting the SAW sensor frequency shift. This mechanism is based on the modification of the dynamic modulus of the polymer upon its absorption of analyte molecules. The use of the Sauerbrey equation , scaled with a modulus amplification factor n is presented to describe the frequency shift upon analyte de tection. This formulation is based on mass loading and n is generally considered to be 1. In prior work, this equation has successfully described the response of such de tectors. A characteri zing equation is also developed here which is the Sauerbrey e quation multiplied by a device constant multiplied by a modulus amplification factor n This dissertation describes the prelimin ary testing of the detector device and presents the data which verifies its operati on as per the design requirements. The analytes used in this investigation were simulant s of nerve gas and th e interferent compound diesel fuel. While the detector array incorporated a sensor for the detectio n of mustard and blister agent, this sensor dosed only w ith large amounts of simulant due to concerns of contamination of the vapor delivery system Mixtures of nerve simulant and interferent were used to verify the selectivity of the array. A relative humidity level of 50% was used in all the dosing tests to verify the usability of the sensors in the presence of normal humidity environments. This dissertation encompasses an investiga tion of the following questions: 1) is a SAW sensor based, handheld chemical agent de tector realizable within the parameters prescribed for handheld chemical weapons identification devices? 2) Does a mass loading model sufficiently describe the rela tionship between the fr equency shift of a particular SAW sensor and the absorption ch aracteristics of a chemo-selective polymer with which it is coated? 3) What reasonable hardware design path to the realization of such a handheld device can be implemented based on the model prediction? There were also several key questions con cerning the general characteris tics of such devices in the embodiment of small, reproducible, low power applications. This dissertation first presen ts an overview of past investigations of SAW devices used as chemical agent sensors.
4 Within this discussion is an introducti on to the general wo rkings of a SAW device. The material presented in this background study focused on two significant aspects of SAW sensor utility. The two aspect s are: 1) the mechanisms through which the SAW device exhibits a frequency shift upon in teraction with analyte vapors, and 2) existing and past experiments related to the assay of polymer coated SAW resonators as used for chemical agent sensing. The dissertation next proceeds to define a basic model for characterizing the response of the detector device when in the presence of target vapors. Proceeding to the design process and hard ware implementation, the text makes a pointed comparison between the new handheld detector and the usual laboratory assay system. Emphasis is placed on conveying the details of the design challenges encountered and the solutions devised for this investig ation, hopefully presenting relevant knowledge for the benefit of fu ture investigators. Apart from the techniques for miniaturiza tion specific to the design challenge at hand, other unique aspects of the design include d: 1) the phase tunable oscillator/sensor module; 2) the methodology of power cycled sensors for reduced power consumption and minimization of sensor to se nsor interference, and 3) the embodiment of sensors, oscillator circuitry, frequenc y tuning and down conversion w ithin a handheld enclosure. The integrated solution to the above questions within the parameters of the design of a SAW sensor based handheld chemical agent de tector thus comprises the unique scientific contribution of the following material.
5 CHAPTER 2 REVIEW OF LITERATURE 2.1 Selection of Background Research The review of literature related to the de sign and characterization of this chemical detection device focused on the following areas: 1) the design of SAW oscillators, and 2) the chemical kinetics of polymer coat ed SAW devices. The re levant theoretical contributions from each category are presente d in the next section of this chapter. Similarly, relevant experimental information is given in section 2.3. These materials were selected for their relevance to both the design of the dete ction system as well as its characterization. Within the area of hardware implementation, the design of stable, low noise SAW oscillators was of specific interest. The characterization of the detector syst em was crucial even in the preliminary design stages of the device since ballpark calculations were necessary for defining parameters such as sensor bandwidth, re solution and sampling tim e. Once designed, the predictability of the deviceÂ’s opera tion was in itself of great value. For the selection of background material for characterization purposes, emphasis was placed on defining the basic relationshi p between analyte concentration and the associated frequency shift of the natural SAW resonance. Thus, the most relevant review material included formulations of frequency sh ift with respect to the physical properties of both the coating polymer and SAW device. With respect to the physical chemistry of the coating polymer, experts such as Grat e , ,  an d, [22Â…26], McGill ,  and, Ballantine   presented the most releva nt and insightful material. Their research is the most cited and contributed gr eatly to the success of the detector design.
6 2.2 Review of Theoretical Work Salmon reviewed the basic criteria fo r SAW oscillator design in his 1979 publication, Â“Practical Aspects of Surface-Ac oustic-Wave OscillatorsÂ”. He presented the usual describing function for the frequency of oscillation (equation 1), and defined the relationship between the freque ncy deviations with respect to the oscill ator-loop delay line length (equation 2). He al so gave an important formulation of the phase noise sideband-spectral-density for offset frequenc ies near the natural resonance of a SAW resonator (equation 3). He used this approach to show that such oscillators could be tuned with an external phase shifter and th at the achievable tuni ng range was inversely proportional to the delay line length. However, he also mentioned that a SAW oscillatorÂ’s noise output could be minimized with the use of a high-power drivi ng amplifier rated for low noise output when properly matched to a 50-ohm delay line . 2 n2f 1 c (1) f2l (2) S f () 10log10Gk 2Tof Po dBc Hz (3)
7 Where f is the frequency of oscillation l is the SAW delay line length Po is the saturated output power of the amplifier To is ambient temperature is the wave velocity Gk is the power gain of the amplifier c is the remaining loop phase shift n is an integer dBc is dB relative to Po McGill  presented an alternate method fo r characterizing the phase noise of a SAW oscillator in terms of its loaded quality factor QL, shown as equation 4. McGill explained that QL is a critical parameter affecting the driving oscillator noise performance. Spf () NRFc 4fc2Ql 2NE fc 3 2NRQlfc 32GFK T Po fc2Ql2 f2 NEf 2GFK T Po (4) Where fc is the resonance frequency of the resonator Ql is the loaded quality factor of the SAW device NR is the resonator flicker noise coefficient NE is amplifier flicker noise coefficient G is the compressed power gain of the amplifier F is the noise factor of the amplifier Po is the power level of the amplifier output K is BoltzmannÂ’s constant T is temperature in degrees Kelvin
8 Klemer, in his minimization procedure for the design of low phase-noise oscillators also illustrates the dependence of the phase noise on Ql. His formulation is shown as equation 5. He applied the techniqu e to a voltage controlled phase-shift SAW oscillator. Performing an iterative minimization, he identified the gr eatest contributors of noise to be the resonator and amplifier elements within the oscillator loop. Like McGill, Klemer reported that the loaded quality f actor determines which of these components presents the greatest source of noise. Also, hi s phase noise prediction illustrated that the contributions of the resonator and the amplifier are maximum for carrier offset frequencies between 10-Hz to 10-kHz . SfmRfo 4fm 3 E2g2fm 3 2R Qlfo 3 fm 2 2GFKT Po 2g2fm 2 Efm 2 GFKT Po (5) Where fm is the offset frequency fo is the natural center frequency E is the flicker noise cons tant of the loop amplifier R is the flicker noise constant of the resonator and G,F,K, T and Po are the same as for equation 4. KlemerÂ’s variable substitutions reveal the dependence of SAW oscillator phase noise on the SAW resonators group delay g and the flicker noise of the driving amplifier. Finally, Klemer reported that the resu lts of his optimization produced Â‘unreasonableÂ’ values for the variable g. Since g is proportional to the loaded qua lity factor of the resonator, Klemer concluded that for quality factors above 2000, the effect of phase noise on the SAW oscillator is minimal.
9 The choice of SAW resonator and the asso ciated circuit topology showed wide variability among the designs illustrated in the background material. Typically, either a single or dual port SAW device was chosen , . With respect to the driving or amplification stage, several implementations utilized a single transistor. Less common was the use of newer monolithic amplifiers. Me ier, however, mentioned the utilization of monolithic amplifiers as being a general advancement in the design of robust gain stages for SAW oscillators . Another common design feature was the use of variable capacitor elements to provide phase tuning of the SAW resonant frequency. Solie stat ed that the use of a single varactor diode within the oscillator feedback loop can implement a phase shift of ~40o . Apart from the sensor oscillator, a SA W detector system generally encompasses sub modules to interpret the state of the sensor and provide an indicati on of the presence of target analytes. Typically a down conversion of the sensor signal is performed through mixing stages . For sensor frequenc ies above a few hundred MHz, several down conversion stages are usually utilized. Schi metta gives an excellent illustration of a complex, multiple down conversion architec ture for a SAW based pressure sensor platform . Another common practice was the use of a reference SAW oscillator to produce a difference signal. This refe rence resonator was implemen ted in order to provide temperature tracking of the SAW sensor thus mi nimizing deviations of the baseline signal due to ambient temperature changes. Fo r SAW devices the relationship between temperature and associated resonant frequenc y deviation is linear. Lao presented an indepth discussion of this relati onship and stated that the dr iving amplifierÂ’s dependence on temperature is not compensated for in th is reference SAW scheme . Avramov, however, suggested that the major baseline de viations due to temp erature effects are associated with the quartz material of the SAW device. He suggested that minimization of temperature deviation can best be achie ved through the proper selection of the SAW device. He stated that the quartz substrat e cut angle for the SAW-mode is the least temperature sensitive implementation among acoustic wave devices .
10 Further, he mentioned that for chemical se nsing applications the use of the SAW-mode resonator is an excellent c hoice. Figure 1 shows the gene ral method of implementing the SAW sensor hardware. Figure 1 Â– General SAW Detector Architecture As a final note on SAW oscillators and down conversion, mention should be made of the usual techniques for coupling the sensor signal out of the oscillator loop. Specifically, coupling issues are generally due to excessive power loss within the coupler and an increase in the noise power of the os cillator from mismatched sensor-mixer path impedances. Among the reviewed signal coupling techni ques, the most intriguing was perhaps the pure, reactively matched splitter presented by Tanski . He presented a third order T-network power splitter used as a coupler. Th is T-network utilized two series inductors and a single shunt capacitor. Although he does not report on the possibility of phase detuning the sensor osci llator he mentions that a purel y resistive coupler is perhaps a better implementation. Leaving the overview of the electronic implementations, the following material now focuses on the chemical kinetics of polymer coated SAW devices. In the past it has been widely held that the frequency shift of a polymer coated SAW device upon sorption of analyte molecules was solely due to mass loading. The widely referenced Sauerbrey Equation (6) defines the frequency shift, f, with respect to the concentration C of a sorbed analyte into a polymer coating.
11 This equation is derived from HookeÂ’s Law fo r the displacement of a spring and gives the frequency shift f as: fKfo 2m A (6) Where K is a device dependent constant fo is natural SAW resonant frequency A is area of the SAW membrane m is the mass change at the surface of the membrane The constant K was also utilized in the backgroun d material to represent a parameter known as the partition coefficient and this coeffi cient is particular to the chemistry of the sensing mechanism. In order to preserve th e clarity of the following content of this investigation, K was used for both instances. When used, the parameter is clearly specified as being either a device dependant constant or a partition coefficient. The reader is thus cautioned as to its use. The partition coefficient, the second definition of K is an expression of the solubility of a gas into a pe rmeable surface . The ratio is given as equation 7. For the present application it describes the solubi lity of an analyte into the SAW polymer coating. K CsCv (7) For equation 7, Cs is the concentration of va por in the sorbent phase and Cv is the concentration of vapor in the vapor phase. Th e Sauerbrey equation defined in equation 6 is often represented in terms of the partition coefficient. Equation 8 gives the relationship between K and the associated frequency shift of a SAW sensor as effected by the mechanism of mass loading , .
12 fvfsCvKs (8) For equation 8, fv is the frequency shift for the mechanism of mass loading; fs is the amount of sorbent phase on the sensor surface (as a frequency shift); and s is the density of the sorbent phase of the analyte. For ch emical sensing applications, the description given by equation 8 is more meaningf ul than that given by equation 6. Evident in recent literature is a clear controversy over the accuracy of the mass loading model used to describe the SA W resonators frequency shift upon analyte sorption. This debate is particular to its utility for describing a polymer coated SAW sensor used as a vapor detect or. It is now suggested that since equation 8 describes the shift of the SAW sensor signal as a sole function of mass loading, it neglects the contributions of the change in the dynami c modulus of the polymer upon absorption of analyte molecules. It is worth noting that the original Sauerbrey equation (equation 6) includes a modulus am plification factor n. Several research groups have suggested that the total frequency shift of the device should be characterized as the sum of the ma ss loading effect and modulus change effect , , and . The disa greement is specifically associated with SAW sensors coated with thin rubbery polymers, such as those used in this investigation. Grate, one of the leading proponents of th is theory suggests a formulation for this total frequency shift through modification of equation 8 wh ich is expressed in terms of the polymerÂ’s fractional free volume or plastic ization. His interpretation is that the polymerÂ’s modulus change is strictly the eff ect of volume increase. Equation 9, gives the formulation .
13fvnfsCvKs n1flsl ASAW (9) Where fv is the total frequency shift fl is the fractional free volume of the analyte as a liquid s is the density of the sorbent phase l is the density of vapor as a liquid ASAW is the kilohertz change in frequency due to a 1oC change in temperature per kHz of coating is the polymerÂ’s coefficient of thermal expansion fs is the amount of sorbent phase on the sensor surface, expressed as a frequency shift Ballantine also reported on an estimation of the contribution of th e modulus modification to the SAW sensor frequency shift. Howeve r, his interpretation was not based on the polymer volume change theory which Grate prop oses. Alternatively Ballantine stated that a rubbery polymer, typica lly characterized as havi ng a shear modulus of ~108 dyne/cm2, glass transitions when excited by high freque ncies . The glassy stateÂ’s typical shear modulus is ~1010 dyne/cm2. In an application of this idea to polymer coated SAW devices he presented a different form ulation for frequency shift f. His interpretation is given as equation 10.
14fk1k2hfo 2k2hfo 24Vr 2 2 (10) Where k1 and k2 are material constants for the quartz substrate h is the polymer film thickness fo is the fundamental frequency of the device is the shear modulus Vr is the Rayleigh wave velocity is the Lame constant Steindl and colleagues have used rubbery spray coated polymers and do not directly address the issue of modulus change. However th ey state that there is virtually no change in the resonators quality factor Q when it absorbs analyte molecules . This would directly contradict the interpretations of Grate and Ballantine, whose modulus models evidence modified Ql values, unless the modulus c ontributions are small. For general insight into the mechanics of SAW devices, Panasik  and Fildes  provided useful explanations of the time domain and harmonic analysis of SAW devices. The information was helpful in calculating oscillator startup times as well as for quantifying the associated harmonics. Also, Pe veriniÂ’s reduced order model  based on Greens function provided the necessary basic review of the usual rigorous mathematical techniques for descri bing general SAW device behavior.
15 2.3 Review of Experimental Work Chung and colleagues have measured the re sidual phase noise of several sets of polymer coated SAW resonators via their gr oup delay characteristics shown on a network analyzer . They constructed an oscillat or loop to drive the SA W devices and through the use of equation 4 calculated the phase noi se of this oscillator. They utilized an amplifier with a fractional frequency stability of 1x10-10 obtaining average oscillator phase noise levels of ~ -160 dBc/Hz. Thei r findings also suggested that when the frequency of the phase noise exceeded the bandwidth of the resonator, the 1/f 2 noise term was maximized. In a follow up investigation McGill incorporated similar polymers into a sensor platform with two sensor arrays. His goal was to gather data on the applicability of such multiple SAW sensor arrays to portable applications. McGillÂ’s parameters for the platform design were based on present specif ications for portable chemical sensors as defined by military officials. He defined some general criteria for such a device as: 1) weighing less than 10 lb; 2) able to operate on either battery power or usual line power, and 3) incorporates the abili ty to perform near real-time reporting and analysis of detection data within its enclosure. McG illÂ’s sensor array used different polymers sensitive to different analytes The two sensor arrays c onsisted of four and six SAW sensors respectively . Grate and colleagues presented the results of their experiments using a foursensor polymer coated SAW array in a bench to p test configuration . They utilized a standard vapor delivery generator system but added a pre-concentrator between the vapor outlet and the SAW test chamber. They re ported the frequency shifts for various concentrations of the nerve agent simula nt DMMP (dimethyl methylphosphonate). For a DMMP concentration of 11 mg/m3 a frequency shift of 8.5-kHz was observed and, for a concentration of 1.1 mg/m3, they recorded a shift of 900-Hz. Finally, they reported that their detectable concentration limit was .12 mg/m3.
16 Grate and NelsonÂ’s experiments with SAW chemical transducer s provided data on their reaction to toluene vapor. Toluene, like the diesel fuel utilized for this investigation is a hydrocarbon interferent that is widely used for testing chemical detectors of the type embodied here. Grate and Nelson reported frequency shifts of 900 to 1000-Hz for 1mg/m3 concentrations of this hydrocarbon compound . A sim ilar investigation performed by Hsieh and Zellers produced SAW sensor sensitivity values between 3 to 5 Hz/ppm for nerve simulant and between .8 and 2 Hz/ppm for hydrocarbon compounds . Using similar flow rates and dilution c oncentrations, their data closely corresponded with those of Grate and Nelson. In another report, Grate presented data fr om his experimental verification of an Inverse Least Squares method of modeli ng polymer coated SAW resonators. The polymers PIB (polyisobutylene), SXFA (organopolysiloxane) and SIL (cyanopropylsiloxane) were gas probed with dimethylfo rmamide for concentrations between .0001 and 2000 mg/m3. He explained that dimethylformamid eÂ’s molecular structure has reactivity similar to both nerve agent and hydrocarbons The PIB and SXFA coatings have strong attractions for hydrocarbons and nerve simula nt vapors, respectiv ely, while the SIL coating is attractive to both. Grate observ ed that PIB and SXFA exhibited nonlinear frequency shift curves for various concentr ations of dimethylformamide. Further he reported that the dipolar SIL coating gave a highly linear frequency shift response. Finally, his data showed that the SXFA coa tingÂ’s maximum frequency shift was 3.5-kHz, whereas SIL and SXFA had maximum frequenc y responses of about 14-kHz  and, . Ballantine provided an experi mental verification of his earlier theoretical work on modulus effects . He reported on the frequency shift characteristic of a 158-MHz SAW chemical transducer. In his experiment he used a .1 micron thick polymer coating of density 1000 Kg/m3. For large analyte doses he reported a total frequency shift for the SAW resonance frequency of 392-kHz. The a ssay was time independent and the sensors limit of detection was noted to be 1g of an alyte. While the mass loading characterization predicted ~ 335-kHz shift, he repor ted a total shift of 392-kHz.
17 Further Ballantine stated that through height profiling he measured the volume increase of the polymer with sorbed analyte. He re ported that of the 392 kHz shift, 55 kHz was likely due to the modulus volume increase. Pehrsson, in a more diverse investiga tion, evaluated several common polymer coatings applied to SAW sensors of re sonant frequencies 112-MHz and 158-MHz. Among the polymers assayed, of specific inte rest were PIB, PEI (polyethylenimine), PECH (polyepichlorohydrin), FPOL (fluor opolyol), PBOH (polybutadiene-hydroxylate) and PEM (polyethylene-maleate). The chemical transducers were gas probed with nerve simulant (DMMP), hydrocarbon compounds (tol uene and isopropelene) and water vapor . The motivation behind this investigation was to acquire a sufficiently large data set to describe the selectivityÂ’s of various polymers to a group of analytes. This data was then used to minimize the sensor count by identifying redundant sens ors. Initially, the investigation incorporated a 12 sensor array. Through their optimizations they concluded that as few as four sensors would gi ve sufficient discrimination among nerve, hydrocarbon interferent and wate r vapor. A tabulated subset of this data is shown in Table 1.
18 Table 1 SAW Sensor Responses fo r Various Analyte Concentrations Vapor Concentration (g/L) FPOL PEM PEI PECH PIB PBOH DMMP 2230 257.62 90.66 2.64 44.84 8.46 39.53 1100 207.48 56.64 2.25 24.07 8.24 21.42 559 160.8 31.99 1.88 11.48 7.97 10.32 372 142.23 24.28 2.01 8.01 8.09 6.73 137 128.81 20.5 1.45 3.9 5.95 2.95 84 102.79 13.44 1.15 2.32 4.9 1.63 52 83.89 8.51 0.84 1.53 4.27 0.93 29 58.7 4.54 0.62 0.8 3.18 0.44 WATER 7180 9.71 8.23 35.23 1.95 2.2 1.35 3550 4.86 5.41 12.85 0.28 1.46 0.72 1820 2.51 3.63 4.91 0.01 1.2 0.52 1210 1.89 2.93 2.86 0.13 1.49 0.24 ISO 129000 13.43 0.54 0.82 21.25 78.95 25.22 64300 5.9 0.06 0.61 10.78 33.49 11.63 33100 2.52 0.04 0.45 5.1 13.84 4.9 22000 1.77 0.06 0.13 3.57 9.19 3.42 TOL 57300 22.07 11.73 3.22 69.59 77.31 77.83 28400 14.53 7.41 1.28 33.95 37.1 40.68 14500 7.68 3.91 0.56 16.52 15.66 17.89 96800 5.11 2.68 0.17 11.58 9.36 11.15 Of the usual methods for applying polymers to the surface of SAW devices, spray coating and pulsed laser deposition are th e most common. Tepper , however, described a unique coating proce ss utilizing supercritical fluid as the polymer solvent. As he explained, this method allows the polym er to be deposited onto the surface of the SAW device as a uniform coating. The inert super critical solvent quickly evaporates after the film application is complete. Through film characterization and gas probing experiments, Tepper claimed that this coating process facilitates faster response times and greater sensitivity of vapor detection than evidenced by sensors coated using other methods. He attributed this to the uniform ity of polymer coverage on the SAW quartz substrate. Frequency-shift in Hz/kHz
19 Interestingly, Cifra and Bleha , in thei r characterizations of thin film polymer coatings, indicated that the partition coefficient K for such polymers tends to decrease with increasing solvent or decreasing solute molecular size. This supports TepperÂ’s claims of the superiority of the supercritical fluid process since this method, unlike others such as spray coating, is not associated with macro-molecule Â“bubblesÂ” at the polymerSAW substrate interface.
20 CHAPTER 3 CHEMISTRY OF DETECTION 3.1 Chemistry of Polymer Adsorption In this investigation SAW resonators co ated with chemo-selective polymers were used as sensors. These polymers are tailored to selectively adsorb a specific target analyte. The mechanism of inte raction between the polymer and target analyte is that of weak hydrogen bonding and const itutes the adsorption of the analyte into the polymer . The functional polymer group is comprise d of a hydroxyl (OH) unit and a fluorinated carbinol (CFn) group. The polar nature of the hydroxyl unit attracts a single oxygen atom within the target anal yte molecule and a weak hydrogen bond is thus formed. The ability of the polymer to form weak hydrogen bonds with target analyte molecules is a most basic functionality prescribing the polymers utility in the design of a useful chemical warfare agent detector. A useful chemical warfare agent detector should embody the following characteristics: 1) fast res ponse and restore time and, 2) sele ctivity of detection and 3) reusability. The weak hydr ogen bond between analyte and polymer is such that its force of attraction is strong enough to eff ect adsorption of the analyte at ambient pressure. This bond, however, is also weak enough to allow desorption of a sorbed analyte within a relatively short time . Desorption of the analyte constitutes the return of the SAW device to its base resonant frequency. The sensing mechanism described above is generally known as Reversible Vapor Sorption (RVS) and defi nes the sensors reusable nature .
21 The inclusion of CFn groups within the polymers mol ecular structure leads to the polymers hydrophobic qualities. It is important to consider th at the polymer retains the ability to repulse H20 molecules while remaining attract ive to analyte molecules when both molecules present an accepting oxygen atom with unpaired electrons. The polymers molecular geometry holds the answer. The de pendence of the repulsion or attraction of the analyte or water molecules on the molecula r geometry of the polymer is based on an effect known as Steric Hindran ce. Specifically, Steric Hind rance is the repulsion of a molecule by another based on the geometry of the orbits of th e atoms comprising each molecule . The fluorinated carbinol (CFn) structures of the polymer are situated to present a maximal number of hydroxyl dipoles normal to the surface of the SAW device. Each dipole segment constitutes a possible sorption s ite due to its magnetic dipole attraction of the oxygen atoms of analytes such as nerve gas. The CFn unit is non-polar and provides the polymers hindrance to the water mo lecule. Effectively, highly polarized H2O molecules have no affinity for the non polar CFn group which masks the hydroxyl group due to their close proximities. The immense size of the H2O molecule further restricts its reaction with the hydroxyl group. Figure 2 illust rates the interaction between the polymer surface and nerve agent. In contrast, the analyte molecules have alternating oxygen atoms with one being double bonded to the central un it and the other singly bonded. Also, the molecule has a fluorine atom orthogonal to the double bonded oxygen atom. The overall effect is the alignment of the entire molecule thro ugh the attraction of the double bonded oxygen atom to the polymers hydroxyl unit and the null effect of the non polar CFn group.
22 Figure 2 Weak Hydrogen Bonding Between Analyte and Polymer To summarize, the key functionalities of th e polymers are: 1) the reversibility of the hydrogen bonding interaction between the pol ymer and analyte molecules; 2) the selectivity of the polymer to its target an alyte, and 3) the hydrophobic component of the polymer. 3.2 Chemistry and Definition of Target Analytes and Simulants The chemical warfare compounds of intere st are nerve and mustard gas. Due to the lethality of such compounds only a few national laboratories are allowed their possession. As such, less deadly compounds de fined as uniquely similar in molecular structure, viscosity and reactivity, to these agen ts are utilized in usual laboratory assays. These compounds are referred to as simulants. Simulant testing was the method used in this investigation for the assay of the ch emical sensor device. A third compound of interest is the category known as hydrocarbon interferent. Thes e interferents have similar chemical structures to those of nerve agen ts. Since both analytes have similar vapor pressures, the presence of these compounds illicits false nerve agent alarms. Diesel, gasoline and jet-fuel are examples of usual hydrocarbon interferents.
23 The three analytes investigated are nerve agent (Sarin or Soman), mustard gas and a hydrocarbon interferent. A brief descri ption of each is given below. Nerve agent interrupts nerve impulse s by inhibiting the enzyme acetyl cholinesterase. The simulant, dimethylme thylphosphonate (DMMP) was used here to simulate Sarin and Soman nerve agents. The reaction type is weak hydrogen bonding Blister agent, when in contact with th e body, causes severe destruction of tissue and rapidly impairs vision and respiratory tract functions. Mu stard gas is the particular blister agent used here. The mustard gas f unctional molecule group is Cl-S-Cl and the bond formed with the polymers hydroxyl group is HCL. This bond is generally of the strong hydrogen type but is destabilized by the polymers CFn group. The net effect is the weakening of the hydrogen bond thereby facili tating the reversibil ity of the entire interaction. Hydrocarbon interferents form weak hydrogen bonds with the polymers. The interferent used here was diesel fuel. Figure 3 below gives a simple illustration of the functional molecular groups for the nerve ag ent, Soman, and the blister agent, mustard gas. Blister Agent Nerve Agent Mustard Soman Figure 3 Â– Functional Molecular Structures of Soman Nerve Agent and Blister Agent S Cl Cl
24 There are two variations of the nerve polymer, type-1 and type-2. Both have affinities for hydrocarbon interferents as well as nerve agent. Type-1 is more selective to the nerve agent Soman while type-2 is selective to the nerve agent Sarin. The simulant DMMP is used to verify the responses from both type-1 and type-2 coated sensors. The chemo-selective polymers utilized to sense these nerve agents are called GB and GD, respectively. The pol ymer HY (poly-isobutylene) is the chemoselective coating used for sensing interferen ts. Finally the polymer VX (poly-epichloride) is used to detect mustard gas. These f our polymers comprised the sensor array upon which the detector design was based.
25 CHAPTER 4 SAW RESONATOR CHARACTERISTICS 4.1 General Description of the SAW Resonator The operation of Surface Acoustic Wave resonators is based on the piezoelectric properties of its quartz substrate. In genera l piezoelectric materials are mechanically stressed when they are subjected to electri c signals and, conversely, exhibit an electric potential when mechanically stressed. Acous tic wave devices such as SAW devices are designed to be dual transducers which take a dvantage of this electromechanical coupling. The device input transducer c onverts an oscillating input signal into a mechanical wave as its quartz substrate is stre ssed. The propagating mechanical wave is directed toward its output transducer where its energy is reconve rted into an output electrical signal. Effectively, the device serves to delay the pha se and velocity of the input signal. The input and output transducers of a SAW devi ce are implemented as interdigitated metal fingers in an interlocking comb-like structure. These structures are commonly referred to as interdigitated transducer s or IDTÂ’s. The IDT dimensi ons, as well as its physical implementation on a piezoelectric substrate, define the characteristics of the acoustic wave propagated between the device transducer s. When appropriate structure geometry is used a surface-normal wave propagation mode (SAW mode) is formed on the substrate. These surface normal waves are known as Ra yleigh waves after thei r discoverer, Lord Rayleigh . The physical dimensions of SAW devices can be modified such that multiple reflections of the surface-normal wave between the two transducers can be obtained thereby forming standing waves on th e substrate surface. This is the basic principle of the constructi on of a SAW resonator device.
26 Typical quality factors for SAW resona tors range from 2000 to 11000 with center frequencies ranging from 20 MHz up to 1GHz. Among acoustic wave resonators the SAW re sonator is perhaps the most ideal for this investigationÂ’s chemical sensing application. This is due to the fact that wave energy for the SAW mode is propagated orthogonal to the surface of quartz substrate. This allows for a high degree of coupling between the SAW surface and the cross-sectional thickness of the covering polymer. Thus, the SA W device is very sensitive to changes in any medium in contact w ith its substrate surface. The Raleigh wave equation describes the velocity of propaga tion of the acoustic wave as it travels across the substrate surface . ft () fs () 1 a2 tet22 2 (11) Where a is the substrate area is the velocity of wave propagation fs is the natural resonant frequency Equation 11 is actually a modi fied form of the Raleigh wave equation where the natural resonant frequency is replaced with the reso nant frequency of the device after being coated with a thin film substance such as a low density polymer.
27 4.2 Electrical and Mechan ical Characteristics A basic yet useful description of th e SAW device can be obtained by modeling the resonator as a mechanical spring with spring constant, k and an applied damping coefficient b Further, through its second-order el ectrical equivalence to a parallel resonant capacitor-inductor circ uit with series resistor (R LC circuit), two resonance quality factors are obtained . Here, the dc gain is (k)-1 which is equal to 1 for a lossless electrical system. ok m o1 LC Qom b mk b2 Q R2C L Mechanical system Electrical system These equivalences assume that an external force F is applied on the functional area of the SAW device due to the effect of mass loading. However, in light of the intended application, it was assumed that addi tional visco-elastic e ffects might contribute to the sensors response. In an extension of the spring analogy th e effect of mass loading on the resonant frequency of a SAW device is described by the Sauerbrey equation , , and . The Sauerbrey equation, from equa tion 6 is restated on the fo llowing page. The membrane area, A for the SAW devices used is 3mm2 and its shifted resonance frequency fo is 311.5 MHz.
28fKfo 2m A Where K is a device dependant constant fo is the SAW resonant frequency after polymer coating m is the mass change at the surface of membrane, in grams A is the area of membrane (1mm x 3mm = 3mm2) This equation shows that the frequency change for a mass loaded SAW device is proportional to the square of the fundamental frequency. In order to be usable, the value of K must be related to the properties of the specific SAW device used in the investigation. For a laboratory system utilizing a very stable frequency counter and an RF shielded, high stability driving oscillator, the f resolution is about 15~20 Hz. Using this resolution, the theoretical limit of detectable mass is between 2 and 3 pico-grams. Drafts, in his report on Â“Acoustic Wave Technology Sensors,Â” reported on the use of a 200-MHz SAW device of similar type to that used he re as a mass scale. The scaleÂ’s limit of detection was reported to be about 3 pico-grams . The specific SAW resonator utilized for this investigation had a typical resonant frequency of 312 MHz. The SAW delay line produces a 180o phase shift from the signal input to output port. The typica l Insertion Loss for this resonator family is 8.1 dB with an unloaded Q (Q unloaded) of 14,000. A loaded Q (Q 50 loaded) of 8,500, is specified for a load impedance of 50 The resonator is further specifi ed as having a frequency aging characteristic of 10 parts per million per year . Apart from SAW sensors, other modern methods utilized for detecting such chemical agents include gas chromatographs (GCÂ’s), mass spectrometry (MS) and diode laser or passive infrared (IR) systems. Gene rally speaking, these methods of detection are categorized into one of three detection type s. They are: 1) ioni zation; 2) electronic excitation, and 3) sorption.
29 Diode laser applications are examples of ionization techniques while GC and MS methods are categorized as el ectronic excitation detection. Finally, SAW sensors such as used here, utilize sorbtion as the detection method. Although chromatography and mass spectrome try instruments provide excellent selectivity they require several minutes to complete the detection process and are thus more applicable to laboratory assays rather than portable solutions. Diode laser sensing systems are hampered by the high cost of the particular diode utilized in the device as well as their non-commercial availab ility. Further, their operating wavelengths fluctuate with temperature creat ing problems in controlling the absolute sensing wavelength. Corrective measures for these issues often require the use of additional, expensive filtering optics which te nd to dramatically increase the overall size of the entire sensing device. With respect to the methods presente d above, the advantages of using SAW sensors include: 1) low cost of manufacture (solid state); 2) commercial availability; 3) provision of non ionic detec tion, creating or utilize no radi oactive ioniza tion sources; 4) have small physical size based on the miniaturi zed associated electron ics; 5) have fast response and recovery times, and 6) when constructed using chem o-selective polymers, SAW sensors evidence appropriate selectivit y to target analytes such as nerve and mustard gas. 4.3 Polymer Coated SAW Resonator for Vapor Sensing The individual polymers were applied to form a coating layer of 70nm.The application of the coating perm anently decreases the resonant frequency of the device. These resonant frequency down shifts ( fo-coating) were recorded to be ~400-KHz which is typical for such coating thicknesses. The pol ymer densities are on the order of .50 g/cm3 correlating to the down shift frequency of between 400 to 500-KHz.
30 The application of the polymer also perm anently changes both the quality factor and the insertion loss of the resonator. Th e severe degradation of these parameters, upon application of the polymer coating, is clos ely linked to the choice of acoustic wave sensor. As previously mentioned, the SAW mode is particularly useful for vapor sensing applications due to its surface-normal wave pr opagation. However, since this propagation mode radiates most of its energy into th e polymer, this intimate coupling between the quartz substrate and the polymer causes wave damping to occur . This damping translates into a degradation of the resonators quality factor and increased insertion loss. On the other hand, this facilitates the SAWÂ’s sensitivity to polymer changes. It was necessary to define the relations hip between the analyt e interaction with the polymer and the degraded quality factor of the polymer coated SAW oscillator. This information was necessary in order to estimat e the noise contribution to the output of a generic driving oscillator u tilizing this SAW resonator in such a sensor capacity. Measurements of the noise contribution from this degraded quality factor are given in Chapter 7. The development of the chemical ag ent detector was based largely on the parameters specified and calculated through the theoretical predictions given in Chapters 3 and 4 of this dissertation. The summary of this information defined the formulation of a hypothesis used as both a design goal and a performance prediction.
31 CHAPTER 5 FREQUENCY SHIFT MODEL 5.1 Frequency Shift via Mass Loading and Polymer Modulus Change The concept of detection for this applicat ion was the assumption that the shift of a SAW sensorÂ’s resonant frequency is a func tion of the concentration of the absorbed analyte vapor. There are two principle mechan isms which define this relationship. These mechanisms are mass loading and the eff ects of polymer modulus change . Equation 6 gave the characterization of the effects from mass loading while equations 9 and 10 described the contribu tion of changes in the polymerÂ’s dynamic modulus. Equation 6 can be rewritten as e quation 12 which gives a formulation of the Sauerbrey equation which is more app licable to this situation . fmfo 2n Aqq1 2 (12) Where f is the observed frequency change fo is the fundamental frequency of oscillation = 311.5 MHz n is the harmonic of the fundamental (n=1 here) A is the active area of the SAW device = .03 cm2 q is the shear modulus of quartz = 2.947 x 1011 gcm-1s-1 q is the density of quartz = 2.648 g cm-3
32 The evaluation of equation 12 for these appl ication-specific parameters is given in equation 13. It was assumed that other factor s besides mass loading would have an effect on the frequency shift of the resonato r upon analyte absorption. Although equation 12 defines n as a harmonic frequency, through comp arison to equations 6 and 9, it was evident that n was in fact a modulus contribution scaling factor. As such, n was designated the descriptor of the modulu s effect in equation 13. The constant specified here as the mass loading coefficient, wa s extracted without th e inclusion of n. fnm (13) 311.500 MHz ()2.030 cm22.9471011 gcm1 2.648 g cm3 1 2 3.6611012 Hz gram Equation 13 illustrates that a change in ma ss of 1-pg caused a 3.661-Hz shift in the resonant frequency of the SAW sensor ( n = 1). Accordingly, a 400-KHz frequency shift is produced by an added mass of 109.26-ng. Since 400-KHz was the measured frequency down shift for a polymer coating of 70-nm on the .030 cm2 active area of the SAW sensor, the density of the polymer s was calculated to be .520 g/cm3. Nerve simulant was selected to define the baseline response for the sensor array since it had been well characterized in past investigations. Referring to the reviewed literature, the SAW sensor was expected to exhibit a frequency shift of between 30 to 15000-Hz when gas probed with low concentrat ions of nerve simulant . Through the use of the equation 13, the theoretical fre quency shift versus mass loading was plotted and is shown in figure 4.
33 As previously mentioned, equations 9 and 10 suggested the characteristics of certain polymer types for which modulus ch ange effects should be accounted. Past investigators were in common agreement on the f act that a thin rubbery polymer, such as utilized here, with a ra w dynamic modulus of ~109 N / m2, exhibits modulus modifications through swelling from sorption of molecules and glass-transitions when excited by high frequencies. Their investigatio ns were conducted specif ically for the case where the polymer acted as a chemo-selective coating for sorption sensing where the SAW sensor was also excited at high fre quencies (> 200-MHz) [ 13],  and, . 00.511.522.53 5000 11041.5104 mass ( ng )Frequency Shift in Hz Figure 4 Â– Frequency Shift vs. Loading Mass on SAW Sensor ( n = 1, fo = 311.5-MHz) Although there is much disagreement with in the research community as to the exact contribution of this modulus effect on th e total frequency shift, the basic accuracy of the Sauerbrey equation has not been disputed. In the following section an extension of this mass loading model is discussed. This model extension describes the possi ble relationship betw een the analyte dose concentration and mass absorbed by the polymer.
34 Although equation 13 is reasonably sound, seve ral assumptions had to be made in the formulation of equation 14. This was due to the lack of information for several critical parameters associated with both the polyme rs and the final assembled detector units. These parameters were: 1) the non device-d ependent partition coefficients of the polymers; 2) the exact air flow rates and a ssociated pressure di fferential within the sensing chamber; 3) the added contributi on to the inconsiste ncy of the sorption characteristics of the polymers by the spray coating method used to apply them to the resonatorÂ’s surface, and 4) the fact that onl y four units, of the design illustrated here, were built and tested at the tim e of this paperÂ’s composition. Finally, equation 13 is the proper starting point for associated measurements and calculations of the exact contri butions of modulus changes to the total frequency shift. Due to the nature of the dynamic modulus of the utilized polymers, a modulus factor was included in equation 14. In light of the inform ation presented above it must be made clear that, at present, the uncertainty incurred by the absence of these critical parameters hindered a proper categorization of the exact components of the sensor responses except that of mass loading. As such, no effort was undertaken to supply veri fication of the exact value of n in reporting the analysis of the final data. Nonetheless, it is predicted that, through further investigation, the true na ture of the modulus contribution will be uncovered. As such, its inclusion in the fi nal model was deemed prudent. Based on the information reported in the reviewed literature, for preliminary analysis, n was assumed to be 1.25. The rubbery nature of the polym ers as well as the high frequency of the SAW resonator used here further jus tifies the use of this value for n
35 5.2 Frequency Shift Model Within the sensing manifold, the sensing cavity volume enclosing each of the four SAW sensors is 1 mm3. A typical concentration dose of 1 mg/m3 of DMMP was used as a base line to determine the final frequency shift formulati on of equation 14. For th is concentration of DMMP, the measured absorbed mass was 1 ng of analyte (1 atmosphere, 25 Co) giving a maximum frequency shift of 4.576 KHz. Major , determined the adsorption rate for a similar thin film polymer-analyte in teraction to be about 25 ng per cm2 per second. Such rapid kinetics supports the assumption of co mplete absorption since the polymer is theoretically satu rated with analyte in ~1.5 seconds. Based on these measurements, the device partition coefficient k was calculated. Finally, with a modulus am plification factor of n = 1.25, the general model of frequency shift with respect to the external dose concentration, using the nerve simulant baseline, is given as equation 14. fc () nkc (14) Where is 3.661 109 Hz/mg, n = 1.25 k is 1 10-6 m3 (device partition coefficient) c is analyte dilution concentration in mg/m3 f is the frequency shift in KHz Considering the possible selectivity of the array, a selectivity table was defined to guide the data analysis process. This template is given as Table 2. The linear relationships defining the data analysis procedure given in Chapter 9 attempted to define the various values of the parameter Sp,a.
36 The selectivity parameter was calculate d using the maximum frequency response of the GB sensor, to a DMMP concentration of 1 mg/m3, as a reference or baseline. Table 2 Â– Selectivity Templa te for SAW Sensor Array Sp,a Analytes: a (Sp,a) Polymer: p DMMP (a=1) (reference) Mustard Gas (a=2) Diesel (a=3) GB (p=1) S11 S12 S13 GD (p=2) S21 S22 S23 HY (p=3) S31 S32 S33 VX (p=4) S41 S42 S43
37 CHAPTER 6 THE CHEMICAL WARFAR E AGENT DETECTOR 6.1 The Bench-Top Sensing System The usual bench-top laboratory system consists of two oscillator loops. These oscillator loops comprise the SAW sensor driving circuit and the reference SAW local oscillator circuit. The oscillator signals are connected to a mixer and the intermediate (IF) or difference frequency from the mixer is connected to and disp layed by a frequency counter. The reference oscillat or, utilizing a SAW device of the same type as the SAW sensor, resonates at a higher frequency than the SAW sensor oscillator circuit. This is so due to the frequency down-shift of the SAW resonator when mass lo aded with a polymer coating . The test systemÂ’s architecture is simp ly an implementation of a superheterodyne conversion. Here the IF signal is between 300 and 700-KHz. This difference frequency directly represents the polymer coating th ickness. The associated driving amplifiers, phase-shifter, filters and couplers of the osci llator are usually implemented using coaxial connected test modules. Fifty ohm coaxial connector cables of the SMA type are usually prescribed although fifty ohm BNC connections can be used for the mixer output since the IF frequency is relatively low. For initial testing, such a bench-top sy stem was constructed. Preliminary data from this system was used to aid the porta ble design effort. Fifty sensors from each of four polymer types were assayed. The sensor Â’s resonant frequencies and respective loop phase-shift values were recorded.
38 Also, the stable drive-power levels, load ed quality factors and harmonics for each sensor, were recorded. From these initial as says it was observed that the SAW sensors required an input power level between 5 a nd 13 dBm for maximum stability and overall sensitivity. It was observed that the local os cillator with the uncoated SAW resonator exhibited proper resonance with a fixed l oop phase-shift network. However, it was determined that a phase-shifter with up to 180o of variability was needed within the sensor oscillator loop to facilita te proper resonance of the sensors. This was attributed to variations among the sensor polymer coatings . The final design incorporated a fixed phase-shift local oscillator loop. The sensor oscillator wa s designed with a variable phase-shifter controlled by a DC voltage between 0 and 12 volts. This voltage range corresponded to a loop phase shift of 0o to 180o at the sensor resonance frequency. Figure 5 illustrates the system util ized for initial testing. 1) Reaction Chamber 2) Amplifier 3) Voltage Controller Phase-Shifter 4) DC Power Supply 5) Frequency Counter 6) Phase-Shift Power Cable 7) Band-Pass Filter 8) Mixer 9) Coupler 10) Amplifier 11) Reference SAW 12) Fixed Phase-Shift 13) Amplifier Power Cable 14) DC Power Supply 15) Amplifier 16) SAW Sensor 17) Coupler 18) Band-Pass Filter Figure 5 Â– Bench-Top Chemical Sensing System
39 The polymer coated SAW device was situat ed within the reac tion chamber shown in figure 5. This chamber was designed with an analyte inlet and outlet port. The analyte inlet port was connected to the vapor deliver y system which controlled the humidity, temperature and concentration of the anal yte introduced into th e reaction chamber. The outlet port was connected to an e xhaust pump which evacuated the analyte into a fume hood. Upon gas probing the sensor frequency shifts were observed on the frequency counter and recorded. The counter utilized here was an HP-5335 Universal Counter. Before being placed in the reaction chamber, the sensors were screened to obtain characterizing data. A network anal yzer was utilized to obtain impedance matching parameters (S parameters) for the SAW devices. This test instrument also provided data on the center frequency, Q and filtering characteristics for the system. The network analyzer used in this investiga tion was the HP-8753E which has a frequency range of 30 KHz to 6 GHz. After installation of the test sensor into the reaction chamber a spectrum analyzer was connected to the sensor oscillator output at the signal coupler. The amplifiers were powered up and the phase-shifter voltage wa s varied until the na tural SAW resonance was observed on the spectrum analyzer. This procedure was performed to properly tune the SAW sensor to resonance as well as to prevent damage to the mixer module by the application of an inappropriately high i nput signal. An Agilent E4405B-ESA-E spectrum analyzer was utilized in this investig ation. A Fluke 6060B Synthesized RF signal generator was used as a test signal source to characterize the systemÂ’s filtering networks and mixer sub system. A Tektronix TDS 3232B 320-MHz oscilloscope was utilized for time domain measurements. The precision of th e data gathered from the test system, and the high confidence in its applicability as a pr oper baseline, was due to the use of test modules and bench-top power supply and meas urement devices. The test modules, being designed for accurate laboratory testing, ev idenced consistent operation and excellent signal to noise levels. Also, their isolated metal enclosures and shielded coaxial interconnections further ensured proper impe dance matching with lo w susceptibility to external electromagnetic interference.
40 Unfortunately, many of these advantages are lost in the migration to a miniaturized system. Nonetheless this baseline performance data proved invaluable. 6.2 Design Specifications for th e Portable Detection System The general specifications to which the ch emical agent detector was designed are shown in Table 3. Table 3 General Specifications fo r the Portable Detection System target analytes: nerve, mustard and hydrocarbon interferent sensitivity: nerve < 1mg/m3 Mustard 50 mg/m3 Interferent 20mg/m3 number of sensors 4 sensor type 2 nerve agent sensors 1 mustard gas sensor 1 hydrocarbon interferent sensor data interface to computer RS232 and computer terminal physical system dimensions 1 ft3 or less weight 5lbs or less power requirement < 8 Watts replenishment of consumables in excess of one month base-line restore time two minutes or less operation lifetime > 3000 hrs
41 While portability encompasses many parameters, the focus here was the minimization of the intended deviceÂ’s size, we ight and power while carefully considering the performance goals presented in Table 3. In a review of the data derived from the bench-top system and those presented in the reviewed literature, it was clear that the SAW sensor concept would meet the majority of the other sensing parameters. Particularly, the ultimate design goals were the conversion of the large oscillator loops, bench-top power supplie s and frequency counter into a portable prototype unit. The optimization of the physical robustne ss of the design was left to future improvements. However, certain design precautions utilized in this investigation lend an inherent yet limited robust nature to the instrument. The interested reader is referred to documents such as MIL-STD-810D and EMI -MIL-STD-461 which gives specifications for device operation under vibration, s hock and electromagnetic interference. As mentioned in the introduc tion, a majority of the suggested uses for such a device are military in nature. Examples of im plementation are: 1) ha nd-held monitors for soldiers; 2) a wheeled military vehicle sens or platform; 3) unmanned aircraft mounting, and 4) placement in government and military offices. The portability requirements shown in Table 3 are a compilation of key parameters given by numerous military documents. In order to maintain the applicability of this por table detection concept to an ultimately field deployable device, other specification sources were also considered. Specifically, Table 3 includes certain parameters exhibited by other SAW sensor systems currently under development. These include the JCAD by British Aerospace and the MINICAD from SAWTECH Inc. Having undergone intensive te sting, the US military reported several shortcomings of these devices and are presen tly seeking refinements of the technology. These shortcomings are particularly associat ed with the devices sensitivity to toxins, inconsistencies in discriminating among analytes and slow response time. The overall intent of this effort was to design a prototype SAW detection system whose concept of operation could adva nce chemical detection technology.
42 In so doing, the final design addressed problems presently being encountered by existing systems. The evaluation and characte rization data given in Chapters 9 and 10 confirm the technology. 6.3 Elements of the Portable Design The design path was chosen with consid eration for advanced miniaturization. Component selection and implementation was based on technologies similar to those employed in highly miniaturized devices su ch as cell phones. It was clear that the migration of the cumbersome test system architecture to a portable design would necessitate the use of miniat ure surface mount and thin film monolithic devices. This further prescribed the use of a printed circuit board (PCB) layout for component integration. This formulation, however necessa ry, was a major design challenge faced in the encompassing investigation. The design of the 4 sensor oscillator module, a particular ly crucial element in the overall system performance, was perhaps the mo st challenging aspect in the pursuit of the portable solution. Its final form and functionality is unique and is disc ussed at length in chapter 8. It suffices here to mention a few key aspects associated with improper sensor oscillator design which have previously acc ounted for performance limitations evidenced by other systems. Such aspects include de-stabi lization of the oscillat or through strip-line cross talk as well as nonlinear frequency sh ift responses due to a lack of phase tuning ability for proper sensor matching. Solutions presented here successfully addressed these as well as other classic issues of cross c oupling among individual se nsor oscillators and problems associated with the restrictions on the locality of the sampling vessel, SAW sensors and their driving oscillators. The elements of the portable solution are: 1) 4 sensors with independent miniaturized phase tunable oscillator drivers; 2) an air sampling manifold;
43 3) down conversion stage; 4) phase tuner, vol tage regulation and os cillator power cycler, and 5) microcontroller and frequency counter. The four SAW sensors are located radially around a teflon manifold at the center of which is a cavity known as the sensing ch amber. Each sensor is the same distance away from the center of the sensing cavit y. This ensures that all sensors have synchronized sampling opportunity. An inlet port to the central cavity provides the only access to the surrounding environment and a singl e outlet port, perpendi cular to the inlet opening, provides for air sample exhaust. To control the sampling period an electro nically actuated valve was positioned normal to the inlet port. When open, surrounding air is allowed to en ter the cavity and, upon its closing, the sample is exhausted. The se nsors within the manifold perform assays of the air sample contained in the sample cavity for the entire duration of this sampling period. A small air pump connected to the exha ust port provides the pressure differential for proper airflow direction through suction. The sensor oscillator was designed on a printed circuit board and each sensor was mounted to its own oscillator circuit. Each a ssembly of sensor and oscillator is hence forth referred to as a sensor module and it wa s designed such that four modules could be attached to a single sampling manifold. The four sensor modules, when mounted on the sampling manifold, form a single integrated un it of four resonating se nsors and single air flow path. A block diagram of the final detection system is shown in figure 6. The following chapters detail the necessi ty, function and design of the individual system components.
44 Figure 6 Â– Block Diagram of the Portable Chemical Agent Detector
45 CHAPTER 7 THE DETECTOR OSCILLATOR 7.1 SAW Sensor Oscillator Design The SAW sensor oscillator was designed to provide the following functions: 1) a SAW resonator signal of +7 dBm; 2) high stability and linearity near resonance; 3) to account for sensor coating variations through variable phase shifting and, 4) to exhibit -100 dBc/Hz phase noise power at offset frequencies greater than or equal to 10-kHz The necessity of having four sensors on a single sampling manifold further complicated the design since this required the oscillatorÂ’s physical wi dth to be less than 15 mm. To achieve the necessary sensor drive power, stability and oscillator bandwidth, wide-band monolithic amplifie rs were utilized. Although singl e transistor designs were considered, high frequency monolithic amplifiers were implemented since they are better characterized and exhibit less variation from device to device. Further, these amplifiers are comprised of several compensating tran sistor networks designed to maximize the flatness of their pass-band response. Finally, such amplifiers require reasonably low bias current and voltage while offering high revers e signal isolation. The monolithic amplifier selected for this application was of the 50 type with a frequency operating range of DC to 2 GHz. Also, the device was specified as a high linearity, low noise amplifier, requiring a typical bias current of ~16 mA at 3.5 volts dc The noise figure of this amplifier was 3 dB and had a th ird order intercept of +14.5 dB.
46 Finally the amplifier was specified as havi ng an input and output voltage standing wave ratio (VSWR) of 1.7 relative to 50 impedance. The amplifierÂ’s scattering parameters were measured using the network analyzer and the data was used to aid oscillator impedance matching. These parameters were me asured for a range of frequencies within the specified pass-band of the device as a linear ity cross check. This data is presented in Table 4. Table 4 Measured S-Parameters for Driving Amplifier Frequency S11 S21 S12 S22 Mhz dB Mag Ang dB Ang dB Mag Ang dB Mag Ang 100 27.9600 0.0400 171 20.10 171.0 22. 51 0.075 5 27.96 0.04 -30 150 276743 0.0414 -11 20.00 -7.9 22.32 0.073 11 26.30 0.04 -62 200 27.3886 0.0428 -22 19.70 -21.0 22.23 0.078 12 26.10 0.05 -71 250 27.1029 0.0443 -37 19.50 -42.0 22.06 0.079 17 -2.00 0.06 -79 300 26.8172 0.0457 -55 19.42 -60.0 22.01 0.082 19 18.00 0.06 -81 350 26.5315 0.0471 -79 19.00 -89.0 21.81 0.082 20 15.92 0.08 -92 400 26.2458 0.0485 -88 18.90 101.0 21.66 0.083 21 16.90 0.09 -100 500 25.9601 0.0499 -105 18.67 138.0 21.53 0.083 21 20.00 0.10 -104 For the purpose of deriving the sensor signa l from the oscillator a resistor bridge was implemented as the oscillat or output coupler. This c oupling technique was utilized since it does not introduce a phase shift within the oscillator loop. This simplifies open loop phase calculations as well as facilitates 50 impedance matching through proper selection of the networkÂ’s resistor values. The disadvantages of this coupling met hod are its relatively high insertion loss (as much as 6 dB) and its non-directional nature. These disadvantages were deemed reasonable trade-offs since the directionality issue was corrected by placing a directional attenuator, also referred to as an isolator, immediately after the coupling network.
47 Also, the power level of the oscillator signal was sufficiently high such that the couplerÂ’s 6 dB insertion loss did not dramatically affect the signal to noise ratio (referenced to a constant -80 dB noise floor). The oscillation criteria was established via -720o phase-shift around the entire loop. This multiple of the basic -360o phase-shift requirement was chosen based on general SAW oscillator stability data given in the background literature , . This total phase-shift was also chosen to facilitate th e necessity of phase-tuning different sensors within the oscillator loop. Figure 7 Â– SAW Sensor Oscillator Architecture As illustrated in figure 7, the SAW resonator provides -180o phase-shift from input to output, as do the tw o amplifier elements. A fixed phase-shift network provides a 120o phase lag at 312 MHz. The total fixed phase shift is -660o. The final 60o shift is obtained with the tuning of the variable pha se-shifter. It is worth noting that the frequency stability of the oscillator is hi ghly dependent on the constancy of the phase shifter supply voltage. The effects and minimization of supply voltage noise for the detector device are discussed in Chapter 8 of this text. Variable Phase Shifter (0-180 deg) (-2 dBm) SAW Sensor (180 deg) ( -7 dBm ) Amplifier (180 deg) ( +17 dBm ) Coupler (0 deg) (-6 dBm) Low Pass Filter Amplifier (180 deg) (+17 dBm) 120 deg (-12 dBm)
48 The oscillator bandwidth was an importa nt factor in the design since it was necessary to keep its linearity of operation within the expe cted band of frequency shift with analyte detection. For the preliminary design efforts, a 10-KHz sensing bandwidth was selected. This was based on estimations of the sensorsÂ’ frequency shifts associated with the detection of analytes at concen trations near the maximum specified dosing requirement , . Another factor in the desi gn of the oscillator was th e minimization of the phase noise associated with the de gradation of the quality f actor of the polymer-coated resonator. Whereas uncoated SAW resonators exhibit 50 loaded quality factors of ~8000, average quality factors fo r the coated sensor are ~300 0. This significant variation in the quality factors of the coated and uncoated SAW resonator prompted an investigation into the relationship between the oscillator phase noise and resonator loaded quality factor. The motive for doing this was to identify the lowest usable SAW sensor quality factor. The design specification for th e oscillator phase noise at 10-KHz offset from the main signal was the determining parameter. From figure 8, it is clear that the lowe st usable SAW resonator Q was about 2500. Through measurements of the oscillator quality factor, the overall Q of the circuit was measured to be roughly 10 % lower than the in stalled resonator. As such a lower limit of ~ 2800 was selected as the lowest usable res onator quality factor due to the phase noise requirement of -100 dBc/Hz at 10 KHz offset from the main signal. The data confirmed that the oscillator loaded quality factor was largely defined by th e resonatorÂ’s quality factor as reported by several sources in the reviewed literature , , and . The theoretical phase noise data was obt ained through the use of equation 5
49 -155 -145 -135 -125 -115 -105 -95 -85 -75 -65 100020003000400050006000 SAW Sensor Loaded QPhase Noise dBc / Hz 1 MHz 100 KHz 10 KHz 1 Khz Figure 8 Â– Oscillator Ph ase Noise vs. Sensor Ql Intuitively, the phase-noise performance of the sensor oscillator has some impact on the detection deviceÂ’s sensitivity. Howeve r, the actual noise level of the entire detection system was a function of total cont ribution of each subsystem to the total limit of detection (LOD) of the device.
50 7.2 Sensor Oscillator Tuning The characteristics of the uncoated SA W device were utilized as the design parameters for the oscillator synthesis. Figur e 9 shows the lumped element model utilized for impedance matching calculations for desi gning the sensor oscillator. As per the device manufacturer, the typical values of Co, Rm, Lm and Cm are 2.2 pF, 84 758 uH and .336 fF respectively . Al so, the transformer coupling factor is specified as .9. 2. Lm Cm 3 Co Co 1 Rm Figure 9 Â– SAW Sensor Lumped Element Model As previously mentioned, a major parameter affecting the SAW oscillator design was the variation in the loaded Q for the indi vidual SAW resonators after the application of their polymer coating. This is associated with the non-uniformity of the spray coating technique utilized to co at the resonators . Through network analyzer measurements, a 170o phase difference was observed for several sets of sensors having similar loaded quality factors and frequency down-shifts. This prescribed an oscillator design incorporating a variable phase shifter within the oscillat ion loop. A phase shifter with 180o phase variability near the typical center frequency of the sensors was implemented. Figure 10 shows a typical output for a sensor module without phase tuning.
51 Figure 10 Resonance Modes for a De -tuned 312-MHz SAW Sensor Module The coated SAW resonator center fre quency was expected to be ~311.5-MHz and the component values for the fixed phase-s hift network were chosen such that an offset frequency of up to 1-MHz would still fall within a linear phase response region. This 1-MHz dynamic range was chosen based upon the frequency down-shift data from the sensor characterizations. Th is data evidenced a maximum coating down shift of 600-KHz of the SAW device from th e SAW natural resonance. It was assumed that the phase tuning adjustabil ity given by the voltage contro lled phase-shifter would be sufficient for the circuit to meet the condi tions for oscillation. Also, the maximum down shift frequency was not deemed large enough to cause severe insertion loss of the SAW device through a resulting im pedance mismatch of the various fixed phase networks. Upon application of the appropriate volta ge, and consequent phase delay within the oscillator loop, the SAW sensor exhibite d its natural resonance mode, down shifted by an appropriate ~ 400-KHz. Figure 11 shows the tuned sensor module output. -52 -42 -32 -22 -12 -2 0100200300400500600Frequency in MHzPower in dB
52 Figure 11 Â– Resonance Mode of Phas e-Tuned 312-MHz SAW Sensor Module The relationship between the tuning voltage and the corresponding phase-shift is 15o/Volt. Thus a phase-shifter tuning voltage of 4 volts was predicted to establish resonance. Indeed, when conducting measurem ents of the required tuning voltage for uncoated SAW resonators, tuning voltages be tween 3.9 and 4.13 volts were required for oscillation. Further, for uncoated resonators this 60o phase shift range corresponded to a tuning frequency span of 200 KHz. However, when coated SAW resonators were placed in the oscillator, proper tuning voltages ra nged between 0 and 12 volts and the tuning frequency span was ~100 KHz. The phase difference between the tuning vol tage of an uncoated resonator versus a coated resonator was due to an impedance mismatch incurred with in the oscillator loop through the degradation of the coated resonators loaded Q. This conclusion was based on the observed difference between the group dela ys of coated and uncoated sensors. An uncoated SAW resonatorÂ’s group delay, g, is proportional to the loaded Q of the resonator, as given by equation 15. Qlfog (15) 311.533 MHz -53 -43 -33 -23 -13 -3 7 0100200300400500600Frequency in MHzPower in dB
53 A characterization of the scattering parameters of 30 coated sensors evidenced group delay values which deviated from the theoretical values pr escribed by equation 15. It was thus concluded that the change in th e resonatorsÂ’ group dela ys, affected by their polymer coating, was the cause of the phase mismatch. This discrepancy between the group delay of polymer coated sensors and that of non coated SAW resonators is further supported by equation 5 7.3 Physical Oscillator Implementation In order to match the impedance of the SAW sensor signal por ts, trace widths of .5 mm were used to obtain a 50 characteristic impedance. Efforts to maintain impedance consistency along the feedback pa th were aided by the use of discrete components designed for 50 port impedances. Impedance uniformity was also maintained through the calculation and us e of coupling capacitor values having an impedance of 1 at 312 MHz. The physical oscillator implementation also incorporated extensive use of signal filtering. A low-pass filter with a break fre quency of 400-MHz was placed at the coupler output to block the first and second harmonic signals of the fundamental frequency. Also, bypass capacitors were placed on both the amp lifier bias networks as well as on the phase-shift tuning voltage power line. Follo wing proper filtering pr actice, these shunt capacitors were placed as close as possi ble to the bias por ts of each device. Proper filtering techniques such as th ose outlined above we re of particular importance for this design because the appli cation required low system noise. Here, the noise effects are further exa ggerated by having four SAW se nsors in close proximity on the sampling manifold. Further, the individual sensor sign als are tightly grouped within the frequency spectrum.
54 Indeed, a significant challenge was presen ted by the necessity of integrating the entire oscillator loop on a pr inted circuit board width of only 15 mm. The requirements for the final detector system dictated the co mpulsory use of four sensors. Further, the sampling manifold width was a fixed paramete r dictated by the radial positioning of the sensors with respect to the sample inlet port As such, the length of the PC board was increased to obtain the necessary ar ea for placement of the components. The lengthening of the oscillator PC board was not a trivial solution to the lack of space prescribed by the sampling manifold. In fact, the decision was not made lightly due to concerns of cross-coupli ng. This cross-coupling interact ion occurs between the two main signal lines of the oscillator. These signa l lines have opposite signal flow directions. Periodic interactions between these paths are affected by the fluctuation of their individual signal phases caused by variances in the background noise floor. Obviously, this coupling is greatest where these signal lines are located in parallel and in close proximity on the printed circuit board. The rest rictive width of the oscillator circuit board often dictated both. Thus, only board length increases which were absolutely necessary were implemented. It was calculated that ev en the minimum required length increase would have too adverse a consequence on th e oscillator performance. A compensating design feature was then sought. The solution to the problem specified above was a technique known as via filtering. To prevent interac tion between the signal lines, multiple vias were placed around the perimeter of the oscillator loop. When these vias are located near signal lines, multiple shortened ground paths for return currents are produced. This filtering technique diminishes the possibility of signal cro ss-coupling by offering direct ground access to stray signals which would have previously ut ilized the surrounding air and PCB substrate as paths to lower potentials. To test the effectiveness of this via filtering technique, th e noise level of 12 oscillator test circuits was measured. Six of these test circuits were patterned with filtering vias. The other six circuits were identical in construction but had no filtering vias.
55 For these test circuits the low-pass filt er on the output port of the coupler was omitted so that the first and second harmonic power levels could be observed. These tests verified the utility of via filtering since th e filtered oscillators exhibited a -20 dB reduction in the noise floor level referenced to that of the non-filtered oscillators. Further, the via-filtered design evidenced a -7 dB re duction of the most di sruptive harmonic signal (~ 620-MHz). The sensor oscillator also incorporated directional attenuator s of between 3 and 6 dB which were utilized to increase the directionality of the signal path. Added directionality decreases the power level of reflected signals caused by impedance mismatches. These reflections can drastically change the oscillator center frequency and in some cases may altogether prevent oscillat ion. This effect was observed in preliminary implementations of the reference SAW local oscillator. As a final word on the coupler design, it is worth mentioning that for this applicationÂ’s operating frequency of ~312-MHz, the resistors forming the bridge do not exhibit the major parasitic reactances and ot her nonlinear characteri stics that might be exhibited at signal freque ncies approaching 800-MHz. Generally, 0805 package styles were used for the passive components comprising the oscillator. Their small area (2mm x 1.2 mm ) allowed the proper filtering capacitors, bias resistors and RF chokes to be implemented within the strict c onfines of the circuit board. Also, these surface mount devices we re implemented on both sides of the PCB, further minimizing the required circuit board length. Figure 12 shows the measured phase noise of the oscillator with Sensor 1 installed as the resonator. Although the Ql for this coated resonator was measured on the Network Analyzer to be 3300, the properl y phase tuned oscillator outpu t evidenced an overall Q of 3005.
56 -155 -145 -135 -125 -115 -105 -95 1101001000Frequency Offset in KHzPhase Noise in dBc / Hz Sensor 1 Phase Noise Figure 12 Â– Measured Phase Noise for Se nsor 1 with Coated SAW Resonator In the final design, the SAW sensor, am plifiers, bias and matching networks, filtering capacitors, phase-shift network, isol ators, low-pass filter and mounting holes were all neatly packaged into a single low noise oscillator module whose basic form is illustrated in figure 13.
57 Figure 13 Â– SAW Sens or Oscillator
58 CHAPTER 8 THE MULTIPLE-SENSOR DETECTOR DESIGN 8.1 The Multiple Sensor Module The multiple-sensor module consists of 4 individual SAW sensor oscillator modules mounted to a single sampling manifol d. As mentioned in the previous chapter, the width of the individual oscillator modules was determ ined by the width of a single side of the sampling chamber housing. This os cillator layout allowed a sensor to be placed on each face of the four-sided manifold. The sensor oscillator was designed for the signal and ground pins of the coated SAW device to be directly soldered to their respective driving circuits. In the preliminary phases of this inves tigation, several methods for connecting the SAW sensors to their respective driving circ uits were considered. One method proposed the full separation of the SAW resonators from the oscillator circuitry. In such a configuration the sensors alone would be m ounted to the sides of the manifold with coaxial cables connecting the sensors to the dr iving circuitry. The dr iving circuitry would then be implemented on a single, separate circuit board. Several variations of this implementation were initially considered. Su ch configurations were entertained only because they avoided the extreme difficulty of implementing the ideal solution. This ideal solution required the SAW sensors to be dire ctly mated to their own driving circuit. Initially, this solution seemed unfeasible base d on the anticipated i ssues of incorporating the entire oscillator loop w ithin the restrictive width of the manifold. After much consideration it was determined that such an undertaking was warranted.
59 The decision to design the oscillator to allo w the sensor to be directly mounted to it and then having the entire assembly mounted to the manifold was based on the low noise, high stability and multiple sensor count required for this applic ation. Basically this avoided the impedance discont inuities and electromagnetic noise associated with soldering coaxial cabling between the signal por ts of the resonator and oscillator. It was also anticipated that these interference problems would be further exaggerated by the susceptibility of high quality factor resonato rs to harmonic generati on. The possibility of being able to mount 4 sensors and their associ ated oscillators on a single manifold in such a tightly packaged, modular fashion was a ve ry attractive propositi on. The final design was based on this architecture and an illustra tion of this multiple-s ensor module is shown in figure 14.
60 Figure 14 Â– Multiple SAW Sensor Module 1) Spacer 2) Low Pass Filter 3) Phase Shifter 4) Oscillator Module 5) Inlet Port 6) Sampling Chamber 7) O-ring 8) SAW sensor 9) Sensor Socket 10) Isolator 11) Exhaust Tube
61 8.2 Sensor Power-Cycling and Signal Multiplexing This design utilized the method of po wer-cycling the 4 individual SAW sensors. Basically, this involves applying power to the amplifiers of one sensor oscillator at a time. As such, only one sensor is operating at any given time. The necessity for power cycling is discussed below. Initial tests of the multiple sensor m odule were conducted using the spectrum analyzer to view the output of each sensor oscillator. A bench top power supply was used to power the drive circuits and another power supply was used to vary the phase-shifter tuning voltage of each sensor circuit. The sensors were powered on, one at a time, and tuned. Their quality factor, signal power, noi se floor level and resonant frequency were recorded. Two sensor modules were th en simultaneously powered and these measurements were again taken. The procedur e was repeated for three and finally four sensors simultaneously operating. For those ca ses where no resonance at the expected ~311.5-MHz was visible, the bandwidth was cha nged to the appropriate scale. Otherwise, the observation bandwidth was set to 1MHz. Table 5 summarizes the data. Table 5 Â– Summary of Interaction Data for Multiple Sensor Module Powered Powered Resonance Tuning Voltage Quality Noise Floor Sensors Sensor(s) Frequencies (Volts) Factor (dB) 1 Sensor 1 311.5-MHz 8.1 3005 -75 2 Sensor 1 311.5-MHz 8.1 1000 -51 Sensor 2 311.5-MHz 8.3 1200 -53 Sensor 1 311.5-MHz 8.1 100 -35 3 Sensor 2 -31 Sensor 3 311.6-MHz 9.1 324 -30 Sensor 1 311.5-MHz -21 Sensor 2 170-MHz -19 4 Sensor 3 155-MHz -26 Sensor 4 311.5-MHz 0 400 -27 Note: All values are in generic units unless otherwise specified.
62 In addition to this data, the oscillatorsÂ’ phase noises were measured. Recall that figure 12 illustrated the phase noise for the Sensor 1 oscillator before being mounted on the manifold assembly in close proximity to th e other three sensors. For comparison, figure 15 illustrates four separate measurements of the phase noise for the Sensor 1 oscillator when mounted on the sensing manifold. -155 -145 -135 -125 -115 -105 -95 1101001000Frequency Offset in KHzPhase Noise in dBc / Hz 1 Powered Sensor 2 Powered Sensors 3 Powered Sensors 4 Powered Sensors Figure 15 Â– Phase Noise for Sensor Oscillator 1 (SAW Sensor Ql = 3200)
63 Figure 15 illustrates the severity of th e cross-coupling which occurred between the individual sensor oscillators when power ed simultaneously. It was clear that this configuration was particular ly noisy and would not meet the desired phase noise requirement of -100 dBc/Hz at 10-KHz offset from the main signal. From the data presented in Table 5, it was concluded that usable data could only be obtained from a sensor when it alone wa s being powered. Otherwise, the sensor to sensor interference would likely render the entire sensor module useless. The powercycling scheme was implemented as a method to avoid having more than one sensor operating at any given time. This eliminated unwanted interactions between simultaneously resonating oscillators. During the period in which a sensor was allowed to resonate, a sampling device logged its resonant frequency. In this case the sampling device was a basic microcontroller. Through an onchip comparator, the controller was programmed to count the number of rising signal edges within a defined sampling time. The sampling time was calculated as tsampling = (tsensor-on-time) Â– (.1 tsensor-on-time). Thus, for a sensor power-on time of 100 milliseconds, the sampling time was 90 milliseconds. The 10 millisecond sampling delay was implemented to account fo r the oscillator star tup time. The actual startup time for 10 sensor modules was meas ured. These oscillators, having typical resonator loaded quality factors between 2500 to 4000, re quired 500 microseconds to stabilize after power was app lied to their driving amplifie rs. Figure 16, shows the typical startup characteristic of the sensor oscillator.
64 -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.006260.007260.008260.009260.010260.01126Time (ms)Ampliture (Volts ) Figure 16 Â– Oscillator St art-up Characteristics The power-cycling scheme added switching no ise to the oscillator output. This narrow band switching noise was observed to have a maximum power level of Â–12-dBc. This switching noise was decreased to near noi se floor levels through the use of isolation attenuators. However, as evidenced by figure 17, the switching noise ultimately contributed to a marginal increase in the phase noise of the individual oscillators. Table 6 gives a summary of the sensorsÂ’ performan ce within the sensor module with power cycling implemented. Here, a decrease in the overall sensor 1 oscillator Ql from 3005 to 2800 (with four sensors being power cycled consecutively), explains the measurements illustrated in figure 17.
65 Phase Noise For Sensor 1-155 -145 -135 -125 -115 -105 -95 1101001000 Frequency Offset in KHzPhase Noise in dBc / Hz 4 Power Cycled Sensors Figure 17 Â– Sensor Oscillator-1 Phase Noise for Power Cycled Sensor Module Table 6 Interaction Data for the Multiple Sensor Module using Power-Cycling Powered Powered Resonance Tuning Voltage Quality Noise Floor Sensors Sensor(s) Frequencies (Volts) Factor (dB) 1 Sensor 1 311.5-MHz 8.1 3005 -66 2 Sensor 1 311.5-MHz 8.1 3000 -60 Sensor 2 311.6-MHz 8.3 2980 -60 Sensor 1 311.5-MHz 8.1 2960 -55 3 Sensor 2 311.6-MHz 8.3 2980 -52 Sensor 3 311.6-MHz 9.1 3100 -55 Sensor 1 311.5-MHz 8.1 2800 -55 Sensor 2 311.6-MHz 8.3 2980 -55 4 Sensor 3 311.6-MHz 9.1 3120 -55 Sensor 4 311.5-MHz 0 3900 -54 Note: All values are in generic units unless otherwise specified.
66 In conjunction with the use of power cycling, signal multiplexing was also implemented. A high frequency signal multiplexer was utilized to apply the signal of the operating sensor to the comparator input. Th is 4 to 1 multiplexer was synchronized with the power cycling timing and the entire proc ess of sampling each of the 4 sensors was completed in 60ms. The looping of this sens ing pattern defines the devices real-time sensing characteristic. Within the 60ms period, each sensor was powered for 15 milliseconds and sampled between the second and fourteenth millisecond. The multiplexer path was gated open for 14.5 millisec onds. Figure 18 illustrates the output of the multiplexer and shows the transition characteristic when switching from one sensor to the next. The implementation of signal multiplexing and power cycling allowed the use of a single mixing circuit since th e four sensor outputs were converted into a single timevarying signal. The derivation of the se nsor signal through the process of down conversion is discussed next. -0.2 -0.15 -0.1 -0.05 0 0.05 0.1 0.15 0.2 0.00090.00140.00190.00240.0029 Figure 18 Â– Transition Characte ristic for Sensor Switching Time (ms) Amplitude (Volts)
67 8.3 Signal Down Conversion As mentioned, the four sensor signals were multiplexed and applied to a single down-conversion module. A high-isolation mixer was implemented and the down conversion was performed by one local osci llator and one mixer. A double-balanced, monolithic mixer was used here. The SAW se nsor signal (~311.6 MHz) was mixed with a 312 MHz SAW stabilized local oscillator signal to produce a base-band signal of approximately 400 kHz. While the single down conversion minimized the overall device count, size and power consumption of the unit, it also added complexity to the design. Whereas usual methods prescribe the impl ementation of at le ast two stages of down conversion, portability requirements prompted the utilization of a single superheterodyne topology. Multiple down convers ion stages are generally preferred since the mixer devices usually exhibit leakage betw een their signal and local oscillator ports. The mixers ability to isolate these signals is critical. Such internal device leakage can severely hinder the proper operation of any ci rcuitry associated with these signal ports. This is particularly true for instances where the local oscillat or frequency is within a few MHz of the signal frequency and where th e operating frequencies are several hundred MHz. This was the problem faced in this ap plication for the implementation of a single down conversion stage. In the preliminary local oscillator desi gns the circuit displayed an intermittent tendency to stop resonating upon connection to the appropriate mixer port. It was observed that this problem occurred only wh en both the sensor signals and the local oscillator signal were connected to the mixe r. Since the frequencies of the SAW sensors were very similar to that of the local osci llator frequency, it was surmised that leakage signals from the sensor port of the mixer to the local oscillator port were impeding proper oscillation. Again, the implementation of is olation attenuators allowed for sufficient decoupling of these input ports and facilitate d continuous, stable oscillation of the local oscillator circuit.
68 The local oscillator utilized a SAW resonator element of the same type as the sensor resonator. This uncoated SAW re sonator exhibits a f undamental resonant frequency of 312-MHz. This oscillator was de signed without a variable phase-shifter since it utilized an uncoated SAW resonator exhibiting little vari ation from device to device. This is due to the high toleran ce standards to which these devices are manufactured. The proper re sonance frequency of 312-MH z was achieved with any combination of local oscillator circui t board and uncoated resonator device. A local oscillator topology, stabilized with the same SAW resonator (uncoated) as the coated SAW resonator, was chosen for two reasons. First, the SAW device has inherent stability of oscillation and a hi gh resonance quality f actor. Second, this implementation provides temper ature-tracking capability. The previously mentioned temperature tracking is the result of the similar frequency shift of both the local oscillator SAW resonator and the SAW sensor when the ambient temperature of the devices changes. 8.4 Power Supply Noise The power-cycler design mentioned above was included in the design of a power supply distribution module. The design of the power supply module is, in itself, a unique solution and is particular to this applica tion. To achieve the low noise power levels necessary for high sensitivit y, extensive use was made of new miniature surface-mount power regulators. Further, th ese regulators are low power devices and are relatively inexpensive. These low power, low dropout regu lators supply steady biasing voltages to SAW driver amplifiers and phase-shifters. Mu lti-turn potentiometers were incorporated on the power distributio n module and were used to set the tuning voltage for the phase shifters. While this tuning method was a si mple, cost-effective solution, the metal coils
69 comprising the potentiometer were considered to be a potential noi se-coupling source. Therefore, the generic ceramic capacitors on th e phase-shifter pin we re replaced with low equivalent series resistance (ESR) tantalum capacitors de signed for wide bandwidth filtering. Also, these tantalum capacitors were 22f, a factor of 10 increase from the previous 2.2f ceramic capacitors. The amplifier devices were much less se nsitive to power fluctuations but good design practice prescribed that they also be supplied with properly regulated voltage. The amplifiers were filtered with generic .1 f ceramic capacitors.
70 CHAPTER 9 EXPERIMENTAL VERIFICATION 9.1 Simulant Vapor Testing Laboratory testing of the device was pe rformed with a vapor delivery system. The carrier gas used was nitrogen. Dry nitr ogen was bubbled through simulant containers and the resulting mixture was routed through thermo-stated metal tubing into a mass flow controller. This computerized mass flow c ontroller was used to verify the analyte concentration before the vapors were applie d to the sampling port of the detector under test. Teflon tubing was utilized to construc t the short flow path between the flow controller and the outpu t dosing port. The choice of mate rial comprising the outlet path was important since the low concentrations of analyte being delivered to the device under test could adhere to or be absorbed by the walls of this tubing. Teflon is commonly utilized for such an application due to its highly passive chemical properties. The vapor flow generator setup, more commonly known as a dosing system, incorporated a debounce compensated on/o ff valve on the output. This valve was electronically manipulated duri ng dosing tests. The generator system was thus designed to provide pulses of analyte vapor to the devi ce under test. The devices were tested at 50 % relative humidity to verify each polymer coatingÂ’s ability to repel water molecules. The vapor delivery system was constructed with two delivery paths terminating in a single outlet port. These delivery paths were de vised to allow one of two target analytes to be delivered to the detector device. The analytes used in the dual flow path dosing system were DMMP and diesel.
71 Dosing was performed with nerve simu lant concentrations (DMMP) of .97, 1.93, 3.86 and 6.75 mg/m3 and a diesel interferen t concentration of 50 mg/m3. Due to the cost of implementing additional flow paths, mustard simulants were not utilized in this dosing scheme. The mustard gas SAW sensor reaction was gauged by alternative means, i.e. its response to a high concentration of mustar d gas simulant, di-c hloroethyl ether (>50 mg/m3), as well as its non-responsiveness to both DMMP and diesel vapors. Appendix A gives a tabulated summary of the chemical characteristics of the relevant nerve agents. Two vapor-mixture tests we re performed. Mixture 1 combined 50 mg/m3 of diesel with 3.86 mg/m3 of DMMP and mixture 2 combined 50 mg/m3 of diesel with 1.93 mg/m3 of DMMP. These combined vapor tests were conducted to verify the selectivity of the sensor array. All dosing tests on the device were performed at 25 Co. 9.2 Experimental Data Â– Vapor Dosing Results The data presented here were obtained from four detector test units. The device under test was connected to a data logging interface computer and a graphics display application allowed the individu al sensor frequencies to be monitored in real time. The data was obtained from ten dosing experime nts at each of the four concentrations of DMMP. These tests were repeated for each of the four test devices. Table 7 gives the frequency shift for the GB sensor averaged fo r each of the four test units and Table 8 gives the frequency shift for the GD interferent sensor averaged for each of the four test units. Similarly, Table 9 gives the frequenc y shift of each of the four HY sensors, averaged across the four test units, for th e fixed diesel dose concentration of 50 mg/m3. For the binary analyte mixtures of DMMP a nd diesel, Table 9 gives the average response for each sensor. For the response times repor ted in these data tables, the Time to Detection Threshold (TDT) was a scaled paramete r referenced to a signal to noise ratio of 4 for the given sensor sampling time.
72 Test units A, B, C and D were dosed with di-chloroethyl ether (mustard simulant > 50 mg/m3). The four PE sensors responded with frequency shifts in excess of 10 KHz, within 1 second with the three remaining se nsors on each unit exhibiting no reactivity to this analyte. Table 7 Â– Average Maximum GB Sensor Response to DMMP DMMP (mg/m3) Maximum Frequency Shift Time to Detection Threshold (TDT) 6.75 29700 Hz ~ 1 second 3.86 17138 Hz ~ 1.1 second 1.93 8570 Hz ~ 6 second .97 4306 Hz ~ 9 second Table 8 Average Maximum GD Se nsor Response to DMMP DMMP (mg/m3) Maximum Frequency Shift Time to Detection Threshold (TDT) 6.75 3010 Hz ~ 1 second 3.86 1713 Hz ~ 1.1 second 1.93 860 Hz ~ 6 second .97 428 Hz ~ 9 second
73 Table 9 Â– Response for HY Sensor on Unit A, B, C and D for Diesel (50 mg/m3) Diesel (mg/m3) Maximum Frequency Shift Time to Detection Threshold (TDT) Device A 2200 Hz ~ 1.5 second B 2101 Hz ~ 1.7 second C 2102 Hz ~ 1.0 second D 2320 Hz ~ 2 second Table 10 Â– Binary Dosing: Average Maximu m Signal Shifts for Unit A, B, C and D Sensor Binary Dose 1 ( fmax) Binary Dose 2 ( fmax) Time to Detection Threshold (TDT) GB 17.20 KHz 8.55 KHz 1 seconds, 6 second GD 2.15 KHz 1.31 KHz 2 seconds, 9 seconds HY 2 KHz 2 KHz ~ 1.6 seconds For Table 10 above, Binary Dose 1 was the mixture of 3.86 mg/m3 of DMMP and 50 mg/m3 of diesel, and Binary Dose 2 was the mixture of 1.93 mg/m3 of DMMP and 50 mg/m3 of diesel. Appendix D presents several an alyte response curves for the test devices as recorded on the interface computer te rminal during vapor testing. The relative selectivityÂ’s evidenced for the sensor arra y, were constructed ba sed on the baseline formulation of equation 14 for D MMP and are given as Table 11.
74 Table 11 Â– Measured Selectivity Parameters for the SAW Sensor Array Analytes: a (Sp,a) Polymer: p DMMP (a=1) (reference) mustard gas (a=2) Diesel (a=3) GB (p=1) 1 0 0 GD (p=2) .1 0 .1 HY (p=3) 0 0 .5 VX (p=4) 0 1 0 Where, S11 = S42 = 1, S21 = .1, S33 = .5, S23 =.1 S12 = S22 = S13 = S31 = S32 = S41 = S43 = 0 From Table 11 the basic linear relationships describing the detection performance of the entire handheld sensing system were deri ved and are shown on the following page.
75 Single Dosing: DMMP (.97 mg/m3) fGBS11nCDMMPmg m 3 4.43 kHz fGDS21nCDMMPmg m 3 .443 kH z fHYS31nCdiesel0 0kHz fVXS41nCmustard0 0 kHz Single Dosing: diesel fGBS13nCDMMPmg m 3 0 kHz fGDS23nCdiesel1 .443 kHz fHYS33nCdiesel1 2.22 kHz fVXS43nCmustard0 0 kHz Binary Mixture 1: DMMP (3.86 mg/m3) + diesel fGBS11nCDMMPmg m 3 17.1 kHz fGDnS21CDMMPmg m 3 S23Cdiesel1 1.710 kHz .443 kHz 2.153 kHz fHYS33nCdiesel1 2.22 kHz fVXS41nCmustard0 0 kHz Binary Mixture 2: DMMP (1.93 mg/m3) + diesel fGBS11nCDMMPmg m 3 8.55 kHz fGDnS21CDMMPmg m 3 S23Cdiesel1 .855 kHz .443 kHz 1.30 kH z fHYS33nCdiesel1 2.22 kHz fVXS41nCmustard0 0 kHz
76 The final linear relationshi ps are restated below, fGBS11nCDMMPmg m3 (16) fGDnS21CDMMPmg m3 S23Cdiesel01 (17) fHYS33nCdiesel01 (18) fVXS42nCmustard01 (19) where fGB, fGD, fHY and fVX are the frequency shifts of the associated SAW sensors in KHz.
77 CHAPTER 10 RESULTS AND CONCLUSIONS 10.1 Results and Discussion Among the test units, single analyte dosing variation from the calculated values of equations 16 and 19 were w ithin 15 % for GB and GD for the DMMP concentration of ~1 mg/m3. For the DMMP concentration of ~2 mg/m3 the variation decreased to 10 % of the calculated value. For the highest DMMP concentrations the variation was ~ 5%. Figures 26, 27 and 28 of Appendix E show the frequency response for the various dosing concentrations of DMMP and illustrate the nonlinear response characteristic for low dosing concentrati ons. This trend toward nonlinear sensor responses for low DMMP concentrations was co nsistent and similar for each dosing test for all test units. It was concluded that two ma in effects contributed to this trend. These effects were: 1) the lack of uniformity of the polymer coatings and, 2) inaccuracy in calibrating the vapor delivery generator to de liver such small dose c oncentrations. For the unit to unit variation, figures 20 and 24 of A ppendix D illustrate the close similarity of detection responses for the reference sensor GB among the test units. Figure 29 of Appendix E further illustrates th is variation among the units. Variation for the hydrocarbon interf erent sensor HY was within 10 % of the calculated value. Similarly, the variation am ong the test units was also within 10 %. The actual responses for sensors GB, GD and HY, for binary analyte dosing, varied from the linear comb inations given by equations 17 and 18 by ~30%. Surprisingly however, the mustard sensor VX exhibited a frequency shift of several KHz for these binary doses (with no mustard included in the mixture).
78 Figures 22 and 25 of Appendix D illustrate this sensorÂ’s response to the binary mixtures. Figures 21 and 23 of Appendix D display the sensors proper non -reactivity to the individual analyte DMMP and diesel dosing. Finally, referring to Chapter 9, this VX se nsor also exhibited its high selectivity when dosed with mustard gas simulant. Through further investigation of both the ch emical nature of this polymer and the flow characteristics of the va por deliver system, it was c oncluded that the VX polymer was actually responding to elevat ed humidity levels associated with the mixing of both analytes. This theory is al so supported by PehrssonÂ’s inves tigations where he reported on the relatively poor resistance to water vapor exhibited by polyethyl based SAW sensor coatings . Thus, it was also conclude d that an undetermined portion of the 30% variation in the signal respons e among the test units for binary dosing mixtures was probably due to the increased humidity le vels associated with these tests. The false positive shift exhibited by the mustard gas sensor is thus correctable through the proper calibration of the humidity le vels associated with the binary mixtures. The estimated humidity level of the present se tup was estimated to be ~70% (rather that the proper 50% level). It is im portant to mention that even though this sensor exhibited a higher affinity to water vapor (with respect to the other three sens ors), the associated frequency shift is only ~ 2000 Hz. This frequenc y response is minimal with respect to the sensorÂ’s response to harmful concentrations of mustard gas. As such, this false positive is easily neglected by increasing the detection th reshold for this sens or. Therefore the non ideal hydrophobic quality of the VX polymer does not impede the selectivity or the detection limit of the entire sensor array. Within the scope of this experimental investigation, the detector designÂ’s limit of detection could not be estimated for each sens or-analyte pair. This was due to the fact that the prototype units were not dosed with comparably low concentrations of each of the three target vapors. Instead, the detecti on limit was referenced to the system noise floor and defined as the lowest detectable frequency shift assuming a noise figure of 4.
79 The RMS system frequency noise was comparable among the test units and was observed to be ~80 Hz. Thus, the limit of det ection for sensor GB was calculated to be .072 mg/m3. Although the lack of test data prevented the formulat ion of a true detection limit, it would be clearly erroneous to report the detection limit for this sensor as being valid for the coincidence function of the array. Thus, in an effort to maintain the full functionality of the sensing array, a more suitable detecti on limit, based on the sensor GD, is specified here as .72 mg/m3 of the nerve simulant DMMP. Further, the 60 millisecond sampling time for each sensor was verified as being adequate since the devicesÂ’ noise floors exceeded 50 Hz. As such, no advantage would be gained by increasing the sampling window. Rather, futu re efforts will be focused on further reduction of the total system noise. 10.2 Conclusions With respect to the sensor power cycli ng, it was observed that the phase noise response of Sensor 1 with four sensors power cycled was similar to this sensorÂ’s noise level for two continuously powered sensor s. Upon further investigation, it was determined that an error in the timing turn -on sequence had in fact allowed two of the four sensors to be powered (with the other two sensors fo llowing the appropriate poweron sequence). This explained the similarity between Figure 17 and the response for two powered sensors, shown in Figure 15. It is beli eved that with the correction of this timing glitch the method of power cy cling will prove even more valuable for system noise reduction. Table 12 gives a summary of the ove rall design parameters and its successful implementation with respect to the design goals. With respect to the physical parameters of the portable solution, the metal enclosure accounts for as much as 50% of the total device wei ght and does not give a true indication of the minimum necessary volume. As such future efforts will likely
80 implement a rigid plastic enclosure, optimized to provide a smaller, light weight solution. The parameters of the final device are sp ecified below and evidence the successful implementation of the design with respect to the design parameters. Table 12 Â– Desi gn Criteria Check List Parameter Design Goal Actual target analytes: nerve, mustard and hydrocarbon interferent sensitivity: nerve = 1mg/m3 mustard 50 mg/m3 Interferent = 50mg/m3 sensitivity: nerve = 1mg/m3 mustard 50 mg/m3 Interferent = 50mg/m3 # of sensors 4 4 sensor type 2 nerve agent sensors 1 mustard gas sensor 1 interferent sensor 2 nerve agent sensors 1 mustard gas sensor 1 interferent sensor data interface RS232 RS232 dimensions 1 ft3 or less .76 ft3 weight 5lbs or less 3.23 lbs power draw < 8 Watts 3.1 Watts replenishment of consumables in excess of one month Verified base-line restore two minutes or less 20 seconds operation lifetime > 3000 hrs In Process With respect to the linear characterization of the frequency shift versus absorbed concentration, it was concluded that equation 14 should be revised to include a damping factor. This damping term, was included to explain the observed nonlinearity of the frequency shift at low concentrations.
81 Equation 20 gives the general form of the m odified response characteristic and equation 21 gives the specific form of th e final characterizing equation. 2k m b 2m 2 (20) 2f2nkcb2c 2 (21) Where is 3.661 109 Hz/mg, n = 1.25 k is 1 10-6 m3 (device partition coefficient) c is analyte dilution concentration in mg/m3 f is the frequency shift in KHz b is the damping constant The testing of the detector device was lim ited to only a few hundred dosing cycles. Since both the polymer coatings and the detector el ectronics were being tested simultaneously much remains to be completed with respect to the characterization of the entire unit. Although more data is needed in order to fully decouple the components of the sensor responses, the excellent correlati on between the theoreti cal approximations and the actual data gathered from the four test units lend support to the composition of several important conclusions. The following conclusions were formulated as a result of this investigation: 1) that a mass loading model can adequately describe the frequency shifts of the SAW resonators utilized for sorpti on sensing; 2) that the quality factor of a polymer coated SAW resonator ultimately determines the noise performance of the driving oscillator; 3) that the lowest usable quality factor for the designed oscillator is 2500;
82 4) that the implementation of individual pha se-tuning networks for each sensor in the sensor array can adequately compensate fo r phase variations am ong these sensors; 5) that, for high frequency SAW sensor arra ys, the method of power cycling provides a solution to the overall increase in system noise associated with sensors placed in close proximity within the array, and 6) that commercially available SAW resonators coated with chemo-selective polymers can provide a reasonably inexpensive and reliable solution to the detection of chemical wa rfare agents when incorporated into a miniaturized sensing platform. The future utility of the de signed detection system is di scussed in the next section and provides a clear outline of the possible improvements for the system. Apart from the detection of chemical warfare agents, the SAW sensor system designed here is applicable to the detection of multitudes of vapors and light gasses. As an example, the modification of this detector platform for sensing light ga sses such as hydrogen is simply a matter of the manipulation of the SAW transducer. 10.3 Future Investigations Generally speaking, it is the authorÂ’s opi nion that, given the present state of the technology, SAW sensors should play a supplementa l role within detect ion platforms that comprise other sensing devices whos e sensing science is more mature. Several research groups have undertaken th e task of integrating SAW devices into conventional gas chromatography systems [7 ]. Through the sensing characteristics evidenced in this investigati on, it is believed that the utility of such SAW detectors can best be exploited within such systems. As evidenced in Appendix B, the sorption characteristics of the utilized polymer coat ed devices are not dissimilar to the time dependant vapor retention characte ristics of gas chromatographs.
83 Another detector design with great potent ial is one combining diode laser sensors and chemo-selective polymer coated SAW sens ors. Such a system could exhibit greatly increased accuracy and sensitivity when appl ied to the detection of chemical warfare agents. Generally, a highly touted implementati on for modern sensor platforms is the use of pattern recognition or look-up libraries. These methods are perhaps a nice idea for the distant future of polymer coated SAW sensor science. However, much further investigation is necessary to refine the sele ctive nature of the polymer coatings with specific applicability to SAW resonators. At pres ent, variations in the sensor responses to wide varieties of target vapor s or complex mixtures of seve ral vapors illustrate the prematurity of implementing look-up library an alysis. For now, such pattern recognition techniques are predestined to transform these detection devices into unmanageably complex and error prone systems. The next goal for the prototype chemical warfare agent detector designed here is an extensive testing period with actual live -agents. In the coming months, the data gathered from these live-agent tests will be compared to the data gathered in this investigation. As for revision of the present portable de tector design, the next generation of the design is already on the drawi ng board. This next generation detector has been designed with onboard processing capability, audio and visual alarms and adjustable sensitivity. Still under investigation is the calibration of the se nsor array to account for changing environmental conditions. Specifically temperature effects on the SAW sensor frequency are currently being addressed. With the addition of onboard temperature sensors, compensated phase recalibration will be possible. For larg er variations in temperature, a model is being developed whic h incorporates the use of equation 22 . This general relationship of the temp erature effects on the SAW resonant frequency is being implemented as a software compensating routine.
84ffo 1FTCToTc 2 (22) Where f is the nominal SAW frequency fo is the Turn-Over Frequency ~ fc + 2.3 kHz FTC is the Frequency-Temperature coefficient To is Turn-Over Temperature ~39 Co Tc is the case temperature. Although this temperature compensation is useful as a preliminary method of counteracting changes in temperature, the issue is far more complicated. The temperature dependence of the design is largely associated with the SAW resonator element. Specifically, both the i nherent temperature effects on the quartz material comprising the SAW resonator, as well as the absorption properties of the polymer contribute. Although th ese effects have not been specifically addressed here, future attempts to refine the characterization of these sensors will likely have to address these issues. Here, it is at least prudent to outline the major aspects of this temperature dependence. These aspects are: 1) lower te mperatures produce higher quality factors and also reduce the phase noise for the SAW resonator element; 2) lower temperatures cause a higher moisture concentration buildup on the SAW surface, which ultimately degrades the sensorÂ’s performance; 3) the rate of absorption of analyte molecules into the polymer coating increases with increasing temperature but the volatility of the reversible bonds increases, and 4) the polymerÂ’s elastic modulus varies with temperature fluctuations and affects the frequency shift response of the se nsor upon analyte detection. As such the amplification factor n is temperature dependent. Due to the complexity of the SAW sensorÂ’s frequency shift for both analyte dete ction as well as signal drift over extended periods, it is clear that future research into this matter will require the accumulation of significant amounts of test data. Regardless, the embodiment of the future design should be one that supplies a detection solution app licable to mass production at reasonable costs and facilitating the protecti on of the nation through vigi lant wide area monitoring.
85 CHAPTER 11 SUMMARY A detection device for sensing chemical warfare agents was designed and tested. The device utilized an arra y of four polymer coated surface acoustic wave (SAW) resonant sensors. The designÂ’s particular em bodiment was that of a testing platform for evaluating the utility of constructing a por table chemical agent detector, utilizing commercially available SAW sensors. The combination of sensor power cyc ling, individually phase-tunable sensor oscillators single step down conversion method and the locality of the sensors, driving circuitry and sensing chamber, comprises the unique nature of the overall design of this detection system. Reported here are the resu lts of the preliminary te sts of the detector and verification of the deviceÂ’s operation as per the design requirements. The analytes used in this preliminary investigation were simulant s of nerve and mustard gas as well as the interferent compound diesel. Mixtures of nerv e simulant and interferent were utilized here to verify the selectivity of the detectors sensor array. A frequency shift model based on the Sa uerbrey equation of mass loading was developed to aid in character izing the operation of the detector device. This model included an elastic polymer m odulus compensation factor of 1.25. This linear model was ultimately developed into a set of scaled selectivity relationships which described the selectivity of the sensor arra y. Also, an assay of the system noise was undertaken, and the detectorÂ’s Limit of detection (LOD) was reported. Ultimately, the motivation behind this dissertation was to contribute to the small pool of existing da ta describing the design and general performance of miniaturized construc ts of present laboratory detection systems.
86 While much past work has involved th e use of large labo ratory measurement systems, few progressive efforts have been undertaken to develop this technology for use in portable detection applications. As such, in past investigations lit tle attention has been paid to the physical design parameters of th e laboratory setup per its migration into a useful portable detection plat form. The design requirements of such a portable device are vastly different from those of the present laboratory systems. This dissertation reported on several key design problems associated w ith the efforts to balance the deviceÂ’s physical size, weight, power consumption and cost against the performance parameters prescribed for detection limits, system noise and reusability. Finally, the following conclusions were re ported: 1) that a mass loading model can adequately describe the frequency shifts of the SAW resonators utilized for sorption sensing; 2) that the quality factor of a polymer coated SAW resonator, ultimately determines the noise performance of the drivi ng oscillator; 3) that the lowest usable quality factor for the designed oscillator is 2500; 4) that the implementation of individual phase-tuning networks for each sensor in the sensor array can adequately compensate for phase variations among these sensors; 5) that for high frequency SAW sensor arrays, the method of power cycling provides a solution to the overall increase in system noise associated with sensors placed in close proximity within the array and, 6) that commercially available SAW resonators coated with chemo-selective polymers provide a reasonably inexpensive and reliable solution to the detection of chemical warfare agents, when incorporated into a mi niaturized sensing platform.
87 REFERENCES  Salmon, S. K., Â“Practical Aspects of Surface-Acoustic-Wave Oscillators,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 27, Dec 1979, pp. 1012-1018.  McGill, A. R., et Al., Â“Performance Optimization of Surface Acoustic Wave Chemical Sensors,Â” IEEE Trans. on Ultr asonics, Ferroelectrics, and Frequency Control, Vol. 45, NO. 5, Sept 1998, pp. 1370-1380.  Klemer, D. P., et Al., Â“Phase Noise Characterization of SA W Oscillators Based on Newton Minimization Procedure,Â” I EEE Trans. on Microwave Theory and Techniques, Vol. 39, NO. 5, May 1991, pp. 883-889.  Drafts, Bill., Â“Acoustic Wave Tec hnology Sensors,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 49, NO.4, Apr 2001, pp. 795-802.  Weigel, R., et Al., Â“Microwave Acoustic Materials, Devices and Applications,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 50, NO. 3, Mar 2002, pp. 738-749.  Meier, H., Baier, T., Riha, G., Â“Minia turization and Advanced Functionalities of SAW Devices,Â” IEEE Trans. on Microwav e Theory and Techniques, Vol. 49, April 2001, pp. 743-748.  Solie, L. P., Wohlers, M. D., Â“Use of an SAW Multiplexer in FMCW Radar Systems,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 29, No. 5, May 1981, pp. 419-423.
88  Schimetta, G., Dollinger, F., Weigel, R., Â“A Wireless Pressure-Measurement System Using a Hybrid Sensor,Â” IE EE Trans. on Microwave Theory and Techniques, Vol. 48, No. 12, Dec 2000, pp. 2730-2735.  Lao, B. Y., et Al., Â“SAW Oscillators in UHF Transit Satellite Links,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 29, No. 12, Dec 1981, pp. 13271332.  Avramov, I. D., et Al., Â“Investiga tions on Plasma-Polymer-Coated SAW and STW Resonators for Chemical Gas-Se nsing Applications,Â” IEEE Trans. on Microwave Theory and Technique s, Vol. 49, NO. 4, Apr 2001, pp. 827-837.  Tanski, W. J., et Al., Â“A Radar System Application of an 840-MHz SAW Resonator Stabilized OscillatorÂ” IEEE Trans. on Microwave Theory and Techniques, Vol. 29, No.5, May 1981, pp. 424-428.  Grate, J. W., Â“Acoustic Wave Micros ensor Arrays for Vapor Sensing,Â” American Chemical Society, Vol. 100, Mar 2000, pp. 2627-2648.  Grate, J. W., Martin, S. J., White R. M., Â“Acoustic Wave Microsensors,Â” Analytical Chemistry, Vol. 65, Nov 1993, pp. 987-995.  Grate, J. W., Zellers, E. T., Â“The Fr actional Free Volume of the Sorbed Vapor in Modeling the Viscoelastic Contributi on to Polymer-Coated Surface Acoustic Wave Vapor Sensor Responses,Â” Anal. Chem., Vol. 72, 2000, pp. 2861-2868.  Ballantine, D. S., Wohltjen, H ., Â“USE OF SAW DE VICES TO MONITOR VISCO-ELASTIC PROPERTIES OF MATERIALS,Â” IEEE ULTRASONICS SYMPOSIUM, 88, 1988, pp. 559-562.  Steindl, R., Pohl, A., Seifert, F., Â“I mpedance Loaded SAW Sensors Offer a Wide Range of Measurement OpportunitiesÂ” IEEE Trans. on Microwave Theory and Techniques, Vol. 47, No. 12, Dec 1991, pp. 2625-2629.
89  Panasik, C. M., Hunsinger, B. J., Â“Harmonic Analysis of SAW Transducers,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 26, MO. 6, Jun 1978, pp. 447-450.  Fildes, R. D., Hunsinger, B. J., Â“T ime Domain Analysis of SAW Reflectors,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 27, No. 2, Feb 1979, pp. 194-198.  Peverini, O. A., Orta, R., Tascone, R., Â“A new Reduced-Order Model of SAW Interdigital Transducers,Â” IEEE Trans. on Microwave Theory and Techniques, Vol. 49, No. 10, Oct 2001, pp. 1785-1791.  Chung, R., McGill, A. R., Matthews, P., Â“Phase Noise Characterization of Polymer Coated-SAW Gas Sensors: Imp lications for the Performance of an Oscillator Circuit,Â” IEEE International Frequency Control Symposium., 1997, pp. 169-174.  McGill, A. R., et Al., Â“SAW SE NSOR TECHNOLOGY TESTED FROM THE LABAROTARY TO THE FIELD,Â” Solid -State Sensors and Actuators-8th International Conference, Stockhol m, Sweden, June 25-29, 1995, pp. 761-764.  Grate, J.W., et Al., Â“Smart Sens or System for Trace Organophosphorus and Organosulfur Vapor Detection Employing a Temperature-Controlled Array of Surface Acoustic Wave Sensors, Automated Sample Preconcentration, and Pattern Recognition,Â” Anal. Chem., Vol. 65, 1993, pp. 1868-1881.  Grate, J.W., Nelson, A. A., Â“Sorptiv e Polymeric Materials and Photopatterened Films for Gas Phase Chemical Microsensors,Â” IEEE Proceedings., Vol. 91, NO. 6, June 2003, pp. 881-889.  Hsier, M., Zellers, E.T., Â“Limits of Recognition for Simple Vapor Mixtures Determined with a Microsensor Array,Â” Anal. Chem., Vol. 76, No. 7, Apr 2004, pp. 1885-1895.
90  Grate, J. W., Patrash, S. J., Kaganove S. N., Â“Inverse Least-Squares Modeling of Vapor Descriptors Using Polymer-Coated Surface Acoustic Wave Sensor Array Responses,Â” Anal. Chem., Vol. 73, No. 21, Nov 2001, pp. 5247-5259.  Grate, J. W., McGill, A. R., Abraha m, M. H., Â“Chemically selective Polymer Coatings for Acoustic Vapor Sensor s and Arrays,Â” IEEE ULTRASONICS SYMPOSIUM., 92, 1992, pp. 275-279.  Ballantine Jr., D. S., et Al., Â“Co rrelation of Surface Acoustic Wave Device Coating Responses with Solubility Pr operties and Chemical Structure Using Pattern Recognition,Â” Anal. Chem., Vol. 58, 1986, pp. 3058-3056.  Rose-Pehrsson, S. L., et Al., Â“Detec tion of Hazardous Vapors Including Mixtures Using Pattern Recognition Analysis of Responses from Surface Acoustic Wave Devices,Â” Anal. Chem., Vol. 60, Sept 1988, pp. 2801-2811.  Tepper, G., Levit, N., Â“Polymer De position from Supercritical Solutions for Sensing Applications,Â” American Chemical Society, Vol. 39, 2000, pp. 44454449.  Cifra, P., Bleha, T., Â“Partition Coeffi cients and the Free Energy of Confinement from Simulations of Nonideal Polymer Systems,Â” American Chemical Society, Vol. 34, 2001, pp. 605-613.  RF Monolithics, Inc., Da ta Sheet Publication 1220298, May 2000.  Major, J. S., Blanchard, G. J., Â“A dsorption Behavior of Polymer-Modified Interfaces,Â” American Chemical Society, Vol. 18, 2002, pp. 6548-6553.  Lu Jung, C., et Al., Â“Portable Gas Ch romatograph with Tunable Retention and Sensor Array Detection for Determinati on of Complex Vapor Mixtures,Â” Anal. Chem., Vol. 75, 2003, pp. 1400-1409.
92 Appendix A: DMMP Vapor Dosing C onversions and Characteristics Table 13 DMMP Vapor Delivery Parameters Temp Saturation Vapor Carrier Flow Rate Vapor Flow Rate Dilution Concentration Co mg/m3 ml/min ml/min mg/m3 ppm ppb 0 764 20000 5 0.19 0.04 36.96 5 1178.453 20000 10 0.59 0.11 113.98 -15 183 20000 10 0.09 0.02 17.70 20 3861.325 20000 5 0.97 0.19 186.78 20 3861.325 20000 10 1.93 0.37 373.47 20 3861.325 20000 20 3.86 0.75 746.56 20 3861.325 20000 35 6.75 1.31 1305.50 20 3861.325 20000 56 10.78 2.09 2086.61 Table 14 Â– Vapor Pressure and Volatility Data for Chemical Warfare Agents Chemical Warfare Agent Simulant / Interferent Vapor Pressure (ppm) at 25 Co Relative Volatility GB (Sarin) Simulant: dimethylmethylphosphonate (DMMP) Interferent: 4000 Medium Triethyl Phosphate (TEP) Diesel, Gasoline Jet Fuel 720 Medium GD 550 Medium dichlordiethylene sulphide (Mustard Gas, VX) di-chloroethyl ether 200 Lowest
93 Appendix B: Chesler-Cram Model Fit yx () yoae1 xx1 22w b1.5 () 1tanhxx2k2 e.5k3xx3 xx3 (16) 0 1000 2000 3000 4000 5000 6000 7000 8000 9000 012345678910111213 Time (seconds)Frequency Shift (Hz) Figure 19 Chesler-Cram Model Fit fo r Sensor GB to DMMP at 1.93 mg/m3 ft () yoae1 tx1 22w b1.5 () 1tanhtx2k2 e.5k3tx3 tx3 (17) Where a is the peak height = 63.129 x1 is the time of maximum peak = 60.875 w is the weighting value of the rate of adsorption = 31.75 x2 is the duration of equilibrium state = 66.58 x3 defines the desorption rate = 148.58 k3 is the standard devi ation of the peak = .096 b and k2 defines the maximal flatness of the equilibrium state = 1.93 yo is the offset or baseline level = 1.05
94 Appendix C: Summary of SAW Sensor a nd Polymer Calculation Parameters Table 15 Summary of SAW Sensor a nd Polymer Calculation Parameters Parameter Value Active area .03 cm2 Coating thickness 70 nm (70 10-7 cm) Frequency Down-Shift with coating 400-kHz Polymer Density .520 g /cm3 n 1 to 1.5 Added mass from polymer 109.26 10-9 g (109.26 10-12 kg) Frequency Shift per gram of added mass 3.661 kHz per ng
95 Appendix D: Real Time Analyte Dosing Responses Figure 20 Device A: DMMP Concentrations of .97, 1.93 and 3.86 mg/m3 Figure 21 Device B: Diesel at 50 mg/m3 Sensor GB -200 1800 3800 5800 7800 9800 11800 13800 15800 010002000300040005000SecondsFrequency Shift in Hz HY GD
96 Appendix D (Continued) Figure 22 Device C: 3.86 mg/m3 DMMP+ 50 mg/m3 Diesel Figure 23 Â– Device D: DMMP at 1.93 mg/m3 -1000 1000 3000 5000 7000 9000 11000 13000 0500100015002000250030003500400045005000Seconds GB HY & VX GD -200 1800 3800 5800 7800 9800 11800 13800 15800 17800 050100150200250300SecondsFrequency Shift in Hz GB HY VX GD Fre q uenc y Shift in Hz
97 Appendix D (Continued) Figure 24 Device D: DMMP Concentrations of .97, 1.93 and 3.86 mg/m3 Figure 25 Device B: DMMP 1.93 mg/m3 + 50 mg/m3 Diesel -5000 0 5000 10000 15000 20000 0500100015002000250030003500400045005000 Time in SecondsFrequency Shift ( Hz ) Sensor GB -2000 0 2000 4000 6000 8000 10000 12000 010002000300040005000SecondsFrequency in Hz HY GB VX GD
98 Appendix E: Theoretical vs. M easured Frequency Responses 0 5000 10000 15000 20000 25000 30000 35000 01234567 Concentration in mg per cubic meterFrequency Shift in KH z Theoretical Measured Figure 26 Â– Average Theoretical vs Measured Frequency Response
99 Appendix E (Continued) 1 3 5 7 9 11 13 15 0.511.522.5 Concentration in mg per cubic meterFrequency Shift in KH z Theoretical Measured Figure 27 Theoretical vs. Measured Re sponse for Low Dose Concentrations
100 Appendix E (Continued) 10 15 20 25 30 35 33.544.555.566.57 Concentration in mg per cubic meterFrequency Shift in KH z Theoretical Measured Figure 28 Theoretical vs. Measured Re sponse for High Dose Concentrations
101 Appendix E (Continued) 0.972.984.997 4 10.5 17 23.5 30 Frequency Shift in KHzd1d2d3d4CDMMP Figure 29 Measured Frequency Response for Detector Units d1, d2, d3 and d4
ABOUT THE AUTHOR Lane Manoosingh received his BachelorÂ’s and Masters degree s in Electrical Engineering from The University of South Fl orida. He held a position as an Electrical Engineer at Constellation T echnology Corporation while in the MasterÂ’s program and served as an instructor of Electronics at th e University of South Fl orida. After receiving his Masters degree, he continued on in both positions through the Ph.D. program. While in the Ph.D. program at the Univ ersity of South Flor ida, Mr. Manoosingh was very active in several areas related to national defense and filed a patent application based on the detection of chemical warfare agents.