Creation and Deployment of the Nereus Autonomous Underwater Chemical Analyzer and Kemonaut, an Odyssey Class Submarine

MIT-Y-03-002 C2

CREATION AND DEPLOYMENT OF THE NEREUS AUTONOMOUS UNDERWATER CHEMICAL ANALYZER AND KEMONAUT, AN ODYSSEY CLASS SUBMARINE

Richard Canulli

MITSG 03-32TH

~ ~ 0 ~

~ ~ ~ CREATION AND DEPLOYMENT OF THE NEREUS AUTONOMOUS UNDERWATER CHEMICAL ANALYZER AND KEMONAUT, AN ODYSSEY CLASS SUBMARINE

Richard Camilli

MITSG 03-32TH

SeaGrant CollegeProgram MassachusettsInstitute of Technology Cambridge,Massachusetts 02139

+SF Grant No.: EAR-0330272 Creation and Deployment of the NEREUS AutonomousUnderwater ChemicalAnalyzer and Kemonaut, an OdysseyClass Submarine

Richard Camilli

B,A,, Biology 996! Cheyney University

M.S,, Civil and Environmental Engineering 000! MassachusettsInstitute of Technology

Submittedto the Departmentof Civil andEnvironmental Engineering In PartialFulfillment of the Requirementsfor the Degreeof Doctor of Philosophyin Civil andEnvironmental Engineering

at the

MassachusettsInstitute of Technology

June 2003

Signatureof Author Departmentof Civil andEnvironmental Engineering November, 2002

Certified by Harold F. Hemond Professorof Civil andEnvironmental Engineering Thesis Supervisor

Accepted by Oral Buyukozturk Chairman, Departmental Committee on Graduate Studies Creationand Deployment of the NEREUS Autonomous Underwater ChemicalAnalyzer and Kemonaut, anOdyssey Class Submarine

Richard Camilli

Submittedtothe Department ofCivil and Environmental Engineering onMarch 27, 2003 in partial fulfillment ofthe requirements for theDoctor ofPhilosophy inCivil and Environmental Engineering

ABSTAIN CT: NEREUSisa compactself-contained low-power submersible membrane-inlet mass spectrometer,designed to measure dissolved volatile gasses in the water column. It is capableof intelligent data collection, analysis and state-based mission control while operatingasa stand-aloneinstrument oronboard theKemonaut autonomous underwater vehicle AUV!. Kemonaut isan Odyssey classAUV with increased payload carrying capacityand dynamic stability, and is intendedfor &eshwaterand coastal marine applicationstodepths of300 meters. The NEREUS-Kemonaut systemcharacteristics allowfor greatly improved dissolved gasdata collection rates,accuracy andmapping resolutionoverpresently available technologies, Applications particularly wellsuited for theNEREUS-Kemonaut systeminclude identification andmapping ofpollution sources suchas chemical spills, investigation ofenigmatic freshwater andmarine ecosystems, assessmentofsubsurface natural resources andestimation ofmarine-related greenhouse gascycling.

A giftispure when it is given from the heart tothe right person atthe right time andat the right place, and when we expect nothing inreturn,

The BhagavadGita

To my mother TMsiskind of hefty, soI maCka special rcfrigcrator magnetto go with it. Scc section 2 2.i!

TABLE OF CONTENTS

Chapter 1 INTRODUCTION.

1.1 Autonomous machines. .16 1.2 Autonomous underwater vehicles. .17 1.3Mass spectrometry . .... 1 9 1.4Autonomous In-situ mass spectrometry .. 20 1.5 Applications . 24 1.6 Systemperformance....... ,...... ...... ,...... ,....,...,...... 26 Chapter2 MECHANICAL AND ELECTRONICHARDWARE. 27 2.1 Vacuumsystem. .,28 2.1.1 Vacuumenvelope...... ,...... ,...... ...... ... 28 2.1.2Inlet apparatus........... ,...... ,...... ,,29 2.1.3Ion pump 32 2.2 Analyzer. ..34 2.2.1Analyzer magnet ,... 35 2.2.2Cycloid analyzer ....37 2.3 Electronic Control ...,41 2.3.1Embedded computer 42 2.3.2 Communication,.

2.3.3Data acquisitionand instrumentcontrol ....43 2.3.4 Mass Selector. ...44 2.3.5Emission regulator . ...47 2.3.6Electrometer ...... ,...... ,...... ,...... ,...... 49 2.4 Powersystem ....60 2.5Component housings, frame, and mounting apparatus. ...,62 2.5.1 Frame .....63 2.5.2Mounting plate, .....64 2.5.3 Instrument cover .....65 2.5.4 Attachment disk .....66 2.5.5Electronics housings ...... 67 2.5.6Hardhat protector .....68 2.6Instrument startup and sphere sealing...... 70 Chapter3 SOFTWARE...... ,...... 73 3.1 EmbeddedCode ...... ,...... .......... ,.....,...... 73 3.1.1Embedded display and user interface ...... ,...... 74 3.1.2 Instrument Control. , 76 3.1.3Calibration...... ,...... ,......... ,....... 77 3.1.4Operational modes . 80 3.1.5Signal Averaging 81 3.1.6Data handling..... 82 3.1.7Data storage. .. 82 3.2 Supervisorycontrol...... 83 3.2.1Communications protocol. ...84 3.3 RemoteGraphical User interface...,...... ...... ,...... 86 3.3,1 Main window ., 3.3.2Command initialization,..... 87 3.3,3Data parsing., .....88 ... 89 3.3.4 Data visualization.. 3.3.5Data storage and retrieval, ..... ,...... ,...... ...... ...... , .....90 3.4 Quantitativepeak evaluation...... 93 Chapter4 AUTONOMOUS UNDERWATER VEHICLE DESIGN 94 4.1 OdysseyII designoverview ...97 4.2 OdysseyIII design ..98 4.3Kemonaut development ... 101 4.3.1Design and construction.. ... 101 4.3.2Control and navigation, ... 103 4.3.3Assembly ... 109 4.3.4Stability and ballasting...... ,...... ,...... ,...... ...... ...... 109 Chapter5 NEREUSBENCH TESTING. .... 112 5.1Dissolved gas gradient sampling . . 115 5.2Computerized operation and noise analysis. 118 5.3Instrument sensitivity. ... 120 5.4 Ion fragmentation ,. 124 5.5Temperature associated peak height variation. ... 125 5.5.1Temperature dependent detection limits ... 126 5.6Variability in ionpump efficiency.....,...... ...... ,...... 128 5.7Instrument response time. ...... ...... ...... ,...... ,...... ,...... 130 5.8 Submergedtesting. ... 132 Chapter6 DEPLOYMENTAND SEATRIALS . 137 6.1Tethered deployments. . 139 6.2 Toweddeployment.. 139 6.3 AUV deployment ... 150 Chapter7 CONCLUSIONSAND FUTURE WORK .. 157 7.1 Towardroutine operation. ... 167 7.2 Systemoptimization.. ,. 168 7.3Sensor integration andreal-tiine data interpretation. ... 169 7.4Future prototypes ...... 170 REFERENCES ... 171 NEREUSCIRCUIT DIAGRAM, ... 173 NEREUSEMBEDDED CODE...... 181 THETISSPECTRAL VIEWER CODE . .. 183 NEREUSDATA INTERPRETERCODE, 203 ... 253 TABLE OF FIGURES

Figure1-1: Rendering of theKemonaut AUV Figure1-2;NEREUS instrument pictured beside laboratory microscope 19 Figure2-1: The NEREUS instrument withupper glass hemisphere removed ..... 24 Figure2-2: Vacuum Envelope cutaway view....,...... ,....,...... 27 Figure2-3:Membrane compression andsagging resulting fromhydrostatic pressure.... 2930 Figure2-4: Sample inlet apparatus Figure2-5: Ion pump, cycloid and vacuum envelope assembly. .... 3 1 Figure2-6: Machined ionpump housing.. ...33 Figure2-7: Asymmetrically shaped analyzer magnet ....34 Figure2-8: Ion trajectory ofcycloidal mass analyzer ...... 36 Figure2-9: NEREUS cycloid tube constructed froma modifiedCEC21-620 .... ,. 37 Figure2-10: Conflat flange with dual BNC feed through 39 Figure2-11: Scancontroller detail .41 Figure2-12: Software controlled emission regulator on/off relay...... 46 Figure2-13: Teledyne parametric op-amp electrometer designby Hemond ...... 48 Figure2-14: Optimal feedback resistor value...... 50 Figure2-15: NEREUS electrometer circuit, 51 Figure2-16: Electrometer sensitivity vs.sampling frequency ..53 Figure2-17: NEREUS electrometer ...... ,...... 54 Figure2-18: Electrometer shielding andmounting 55 Figure2-19: Power spectral density ofelectrometer withand without compensator...... 5756 Figure2-20: Cross correlation ofelectrometer noisewith and without compensation ....58 Figure2-21:Power spectrum densitycomparison ofNEREUS andHemond-Teledyne electrometers...... ,...... Figure2-22: Layout ofNEREUS components withinpressure sphere 59 Figure2-23: NEREUS instrument kame...... ,...... ,...... ,...... 6463 Figure2-24: Mounting plateplacement ofanalyzer, detector andion pump assemblies. 65 Figure2-25: Composite instrument cover...... ...... 66 Figure2-26; Embedded computer with memory card. Figure2-27: The NEREUS instrument, sealedwithin itspressure housing...., 68 Figure3-1: Remote calibration GUI for NEREUS instrument 69 Figure3-2: Valid command string example 78 Figure3-3:Thetis Spectral Viewer login window...., ...... ,...... 8785 Figure3-4;Thetis Spectral Viewer main window....,...... 88 Figure3-5:Thetis GUIcommand radiobutton fields andtool bar ...... ,...... ,.....89 Figure3-6:Thetis GUI status bar ...... ,...... ,...... ,...... ,...... 89 Figure3-7: Thetis GUI histogram fieldwith red temporal window Figure3 8.Thetis GUI spectrum field...... ,....,..... ,...... 91 Figure3-9: Thetis Spectral Viewer main window with data table visible... 92 Figure3-10: NEREUS Peak Interpreter GUI 93 Figure4-1: Odyssey I vehicle courtesy ofMIT Sea Grant! 94 Figure4-2:Odyssey IIdXanthos AUV ...... ,...... ,...... ,...... 10098 Figure4-3:MIT Sea Grant Odyssey IIICaribou AUV...... ,...... ,...... ,...... 101 Figure4-4:Glass pressure sphere placement within the Keinonaut AUV ...... ..... 104 Figure4-5: Cut-away view of Kemonaut,exposing upper floatation foam and lift harness Figure4-6: Exploded diagram ofKemonaut vehicle components ...... ...... ,..... 107108 Figure4-7: Positioning of the NEREUS instrument within the Kemonaut forward payloadbay...... 1 0I Figure4-8: Partially assembled Kemonaut, showing the NEREUS instrument and compositegirder "" -""" ....".....",...... ,...... 111 Figure4-9: Structural integrity test of fully laden Kemonaut vehicle... 111 Figure4-10: Kemonaut ballasting inUpper Mystic Lake,,...... ,..... 113 Figure5-1: NEREUS setup using Keithley electrometer and10-turn potentiometer ....115 Figure5-2: First NEREUS scan completed onJanuary 8,2002, showing ambient air spectrumcontaining water, nitrogen, oxygen, and argon peaks...... ...... , 116 Figure5-3: NEREUS ambient air spectrum using on board electrometer with 10' D feedbackresistor, ...,...... ,...... ...... ...... 117 Figure5-4: Measurements ofdissolved gases ina synthesizedgradientusing Mishawum Lakewater.... . 119 Figure5-5; Conceptual diagram ofNEREUS control and data acquisition design...... 121 Figure5-6: Ambient airspectruin, performed using computer control and data acquisition atmass step interval = 0.1M/Z. This spectrum reveals the presence ofnitrogen, oxygen,argon, water vapor, and carbon dioxide...... ,...... ,.. 122 Figure5-7: Comparison oftypical noise characteristics ofthe NEREUS instrument when operatedusing differing signal averaging regimes...... ,..... 123 Figure5-8: RMS noise versus time required forcompletion ofa highresolution, 1385 datapoint, scan using a 100MHzembedded computer ....,...... ,...... 123 Figure5-9: A high-resolutionspectrum ofambient airdetailing nitrogen gasand an apparent' N-' N isotopicpeak..,...... ...... ,...... ,...... ,...... ,...... 124 Figure5-10: 'Qon peak height of dissolved gases inatmospheric equilibrium asa function oftemperature.,...... ,...... ...... ,...... ...... ,...... ,..... 127 Figure5-11: Model ofNEREUS ionpeak height sensitivity asa functionoftemperature. Datapoints are normalized toNEREUS spectrum data from filtered water at 22'C, Figure5-12: NEREUS ionsignal variability, recorded while submerged ina flume. .... 128 Spectrumscans executed using a peakjumping algorithm...... ,...... ,... 131 Figure5-13: Calculated overall NEREUS time to steady state measurement forvarious gases ...... 1 34 Figure5-14: NEREUS time response tocarbon dioxide ...... ,.... 135 Figure5-15: Time response ofHemond backpack portable mass spectrometer tonitrogen gas .. ...,...... ,...... 1 36 Figure5-16: NEREUS time response tobenzene...... ...... ,...... ... 137 Figure6-1: Tethered NEREUS configuration withHydrolab ...,...... 140 Figure6-2: NEREUS spectrum Rom at 2 metersdepth inthe Halls Brook Holding Area ,...... ,...... ,...... ,...... 1 4 I Figure6-3: NEREUS depth profile of ion peaks inthe Halls Brook Holding Area...... 142 Figure6-4: Hydrolab depth profile of dissolved oxygen, salinity and temperature inthe HallsBrook Holding Area ....,.....,...... .... 142

10 Figure6-5: Location oftethered NEREUS deployment inBoston Harbor ...... ...... , 143 Figure6-6:Microscopic comparison ofa newmembrane left!,and after heat forming anddeployment to 12 meters depth right! . Figure6-7: Boston Harbor spectrum at1.5 meters depth, 144 Figure6-8: Boston Harbor spectrum at3 metersdepth. 145 Figure6-9: Boston Harbor spectrum at4.5 meters depth...... 145 Figure6-10: Boston Harbor spectrum at6 meters depth ...... , ...... 146 Figure6-11: Boston Harbor spectrum at7.5 meters depth ... 146 Figure6-12: Boston Harbor spectrum at9 metersdepth. 147 Figure6-13: Boston Harbor spectrum at10.5 meters depth, 147 Figure6-14: Boston Harbor spectrum at12 meters depth 148 Figure6-15: Depth profile of BostonHarbor 148 Figure6-16: Oxygen signal correlation ofNEREUS andHydrolab datafrom Boston 149 Harbor Figure6-17: 3-D map oftowed NEREUS deployments inUpper Mystic Lake ........ Figure6-18; Oxygen signal correlation ofNEREUS andHydrolab datafrom Upper Mystic Lake, Dec. 13..... Figure6-19: Oxygen signal correlation ofNEREUS andHydrolab data&om Upper 153 Mystic Lake, Dec. 15 Figure6-20: Chemical profile ofUpper Mystic Lake, December 15,2002 ...... ...... 154 Figure6-21: The Kemonaut AUVprior to deployment inBoston Harbor...... ,..... 156 Figure6-22;Nautical ChartofBoston Harbor showing GIBplacement andoverall area157 of operation . Figure6-23: Kemonaut pressure transducer record ... ,...... ,...... ,...... 159 Figure6-24:2-D GIBhydrophone surface track log of Kemonaut divemissions 160 Figure6-25: 3-dimensional GIBhydrophone tracklog of Kemonaut divemissions.... 163162 Figure6-26: A typicalNEREUS spectrum thatwas recorded while onboard the KemonautAUV, showing dissolved nitrogen, oxygen andargon...... ,...... ,...... 165 Figure6-27: NEREUS recordof dissolved gasion signal during Kemonaut deployinent inBoston Harbor. 165

Table1-1: Comparison ofsubmersible massspectrometer specifications,... Table2-1: Cycloid pin out, ....23 Table2-2: Waterproof cable pin assignments ...... ,...... ,...... ...... ...... 40 Table2-3: Cycloid electric field potentials ...43 Table2-4: Power system wiring harness connector 47 Table2-5: Current requirements andfuse ratings ofelectronic circuits....,...... 60 Table2-6: 10 base-T power connector wiring...... ,.....,...... 61 Table3-1: Embedded computer BIOS settings 62 Table3-2 Parallel port registers and functions ..;...... 75 Table3-3 Command string functional elements and values ...... 77 Table5-1: Approximate scan input potentials...... 85 ...... 118 Table5-2: Ion 6agmentation ratios...... ,...... ,...... ,...... ...... ...... ,..... 125

11 Equation2-1:Fundamental thermalnoise offeedback resistor . ,, ...50 Equation3-1:Mass calibration equation,, , . ,....,, .79 Equation3-2:Peak-finding algorithm ...... ...... ,...... ...... ,...... ,...... 95 Equation5-1:Membrane permeability asa function oftemperature,...... ,. 127 Equation5-2:Inlet gas flux as a functionofmembrane geometry, permeability, andgas concentration. ...129 Equation5-3:Fick's second lawof diffusion.....,...... ,...... ,...... ,.... 133 Equation5-4:Analyte gasinflux,...... ,...... ,...... ...... ,...... ...... ,...... 133 Equation5-5:Membrane response timelag ...... ,...... ,...... ,.. 133

12 13 16century painting ofAlexander TheGreat descending ina glass diving bell

14 Chapter 1

INTRODUCTION

Theaquatic environment isinterwoven throughout thehistory ofmankind, serving as a transportationconduit as well as a sourceofsustenance, religion, wealth, beauty and death.Approximately 37/0of the world's population lives within 100 km of an ocean [1], yethumanity's understanding ofthe hydrosphere hasremained extremely limited. At present,our understanding of the dark side of themoon exceeds what is knownabout the bottomof the Eath's oceans. Like the surface of themoon, the submarine environment is inhospitabletohumans, requiring technological devices for exploration, and since ancient timesmankind hassought toobserve thesubmarine environment in-situ. Writings as earlyas Aristotle's Physika Problemata chronicle theuse of submergence systems that, "enablethe divers to respire equally well by letting down a cauldron;forthis does not fill withwater, but retains the air, for it isforced down straight into the water; since, if it inclinesatall from an upright position, the water flows in" [2]. Many legends also surroundAlexander TheGreat's use of a divingbell [3], known asthe Colimpha, during hissiege of Tyreto observedemolition divers removing underwater obstacles from the harbor.The advantages ofin-situ observation have persisted through modern times, leadingtothe use of manned submersibles suchas the Trieste, Archimede, Alvin, and NR1for scientific survey ofthe deep [4]. Although extremely useful, the danger and operationalcostsassociated withthese manned submersibles [5]have prompted the developmentofsmaller robotic survey platforms, including buoyed sensors, drifters, remotelyoperated vehicles ROVs!, and autonomous underwater vehicles AUVs!. AUVs,which are the newest ofthese technologies, aredoubtlessly themost challenging ofthe unmanned systems todevelop because oftheir reliance onembedded expert systemsfornavigation control and operations decision-making, butthey also appear to holdthe greatest promise.

15 1.1 Autonomous machines

Waterclocks, arguably. the first autonomousmachines, have existed since at leastthe reignof AmenhotepI in AncientEgypt, some 3500 years ago. Over the millennia technologicaladvancement has expanded the domainautonomous machines to include the steamengine built by Hero of Alexandriain 100AD andthe mechanicalautomata of Generalde Gennes, Jacques de Vaucanson and Henri Maliardet in the17 through19'" centuries[6], but it wasnot until thelatter half of the 19 centurythat the fundamental valueof closed-loopfeedback in controlsystems was recognized [7]. Shortlythereafter, theadvent of long-rangeweapons systems and telephone switching electronics initiated therapid development of highly accurate inertial guidance systems for warshipsand aircraft.By WWIIcontrol and guidance systems had advanced to the point of allowing unmannedweapons such as the Vengeance I buzz bomb, an early form of cruisemissile, to beused against Britain, a technologicalaccomplishment that cost the lives of thousandsof people. Ironically, the M9 gun director and SCR-584 radar system, a then radicallyadvanced type of semi-autonomousanti-aircraft weapon capable of automatic targettracking and firing, was used to fendoff thebuzz bombs, downing 57'/0 of the missilesthat they engaged[7]. Duringthis samewartime period less sinister autonomous systems were under developmentin England by anAmerican neurological researcher, W. GreyWalter. His creations,a light and impact-sensitive mobile robot named Machina specularrix and a later,more advanced, mobile robot further equipped with memory and sound sensitivity, Machina docilis, wereused to deinonstratethat machinescould exhibit traits of spontaneity,autonomy, Pavlovian conditioning, and self-regulation, While studying these machines,Walter identified several emergent behaviors that were sufficiently sophisticatedto be described in anthropomorphized terms such as "speculative mode", "learning","reflex" and "&ee will" [8,9]. In thedecades since Walter's work a great dealof researchhas been dedicated to developingautonomous agents which rely on embeddedintelligence to operatewithin complexnatural environments in orderto fulfill predeterminedmission objectives. The resulting multitude of highlyspecialized systems includes,among others, walking robots such as Genghis [10], space probes such as the

16 DSl[11], and autonomous unmanned submersibles includingthe MIT Odyssey Class AUVs [12].

1.2Autonomous underwater vehicles AlthoughAUV use for oceanographic researchcomplements thestrengths ofmanned submersiblesandtethered robotic vehicles, thechallenges andbenefits ofAUV operationsarefundamentally differentfrom those ofexisting platforms [13!.As a result, AUVtechnologies areevolving ina mchewhere they will serve asa low-costmechanism for"spatio-temporal sampling;using multiple vehicles incombination formapping out rapidlyevolving phenomen[a]; andtransiting longdistances toa site formaking observationsasa partof a firstresponse teamat the sight ofinterest" [14].Unlike other unmannedplatforms suchas moorings, towfish, and remotely operated vehicles, an AUVdoes notrequire a tether, thusallowing a spatial survey tobe conducted faster, followbathymetry betterand operate inhigh sea states; translating intoincreased data collectionperunit time orcost, especially whena priori knowledge ofthe survey areais limited.Presently, thereare several dozentypes ofAUVs inoperation aroundthe world, ranginginsize f'rom thediminutive 37kg 1.6 meter long REMUS vehicle [15], tothe massive12meter long ISE Theseus withits one-ton cargo capacity [16].In recent years severalAUVdesigns havetransitioned f'romtheexperimental stagesintocommercially availablevehicles, including theKongsberg Hugin 3000 and the MIT allied Bluefin OdysseyIII. TheOdyssey ClassAUVs, originally developed atMIT Sea Grant, arerelatively small,easily deployable, lowcomponent costsurvey platforms thathave been used in numerousmissions throughout theworld [13]. These vehicles make use of low-cost, mass produced,hollowglass spheres tohouse vehicle electronics, sensorsand power systems ina dry,one-atmosphere environment.This maximizes buoyancyefficiency andpermits theuse ofa freeflooding vehicle fairing, thusminimizing vehicle displacement. However,theminimization ofvehicle displacement makesthetradeoff between range i.e.power storage! andpayload capacity a critical design consideration [17]. Thereisa needforin-situ chemical sensors forAUVs. The creation ofsuch sensors presentssignificant challenges, particularly forsmaller vehicles likethe Odyssey class

17 designs.Issues such assize, power consumption, andinstrument signalstability have, untilrecently, excluded theuse of highly capable analytical instruments suchas mass spectrometersandgas chromatographs onthese platforms. Thus, AUV generated chemicaldatahas usually been limited toa selectfewparameters usingion-sensitive electrodes,mainly: dissolved oxygen, salinity, ORP, pH, nitrate, phosphate andammonia. However,asBales and Levine aptly state, "To achieve widespread acceptance withinthe oceanographiccommunity, AUVs must also deliver, asa minimum,dataof coinparable qualitytothat collected today. Sensing andmeasurement systems designed for deployment&omthe surface e.g.on a samplingrosetteor tow-body! donot always meet thesize and power limits imposed byAUVs" [18]. In general, ion-sensitive electrodes do not meetthis stringentstandard. Thisresearch focuses onan emerging roleof mass spectrometry forsubmerged, autonomous,in-situoperation. Specifically, wedetail the development anddeployment ofthe Kemonaut AUVand the NEREUS autonomous submersible massspectrometer. TheKemonaut isa small, shallow water, AUV capable ofoperating todepths of300 meters.Itpossesses a unique forward payload bayfor carrying scientific instrumentation, suchas NEREUS. TheNEREUS instrument provides a means forrapid and i~-situ quantificationofmany gases which have, until now, required offsite analysis, Although NEREUSisdesigned foroperation aboard theKemonaut AUV Figure 1-1!, it can also beused within a BluefinOdyssey IIIAUV as well as a hostofother platforms. Operationofthe NEREUS instrument onboard anautonomous underwater vehicle enhancestheadvantages ofin-situ analysis ofdissolved volatile chemicals byincreasing therange and overall accessibility of data collection areas. Figure1-1: Rendering of theKemouaut AUV

1.3 Massspectrometry Theprinciples ofmass spectrometric analysis arewell understood. Ionrepulsion and attraction,a central phenomenon governing massspectrometry, wasobserved nearly 150 yearsago by J. Plucker inhis investigations ofmagnetically induced light pattern deformationswithina glowtube. In 1897J.J. Thomson demonstrated thatcathode rays cauldbe deflected byelectrostatic andmagnetic fields [19], followed in1898 by W. Wein'sdiscovery thata typeofray emitted during electrical discharge atlow pressure, knownatthat time as a Goldensteinray,was actually a positively charged ionbeam and couldalso be deflected bya magneticfield [20]. This phenomenon waslater elucidated withThomson's investigations ofthe parabolic trajectories ofhydrocarbons andother polyatomicionsin his book, Rays ofPositive Electricity and Their Application to ChemicalAnalyses [21]; thus establishing thations could be separated based ontheir massto charge ratio when accelerated through a magnetic field. From this work came the massspectrograph, a predecessor of the mass spectrometer, which relies on a photographicplateto record the presence ofion beams. However, thevariable ion sensitivityofthis early iiistrument's photographic plateprohibited precise quantification ofion intensity. Forthis reason, Thomson modified hisparabola mass spectrograph throughreplacement ofthe photographic platewith a Faradaycupand electroscope. By varyingthemagnetic fieldstrength andmeasuring theresultant ioncurrent, hethereby

19 createdthefirst mass spectrometer andproduced a mass spectrum thatquantitatively describedtherelative abundances ofindividual gases present ina mixture[22]. The terms massspectrometer and mass spectrograph would be later coined by F.W.Aston in 1920 [20]. Insubsequent yearsmass spectrometers haveevolved toaHow for precise mass determinations,measurement ofrelative ion abundances, andelectron impact studies. Thishas lead to mass spectrometric analysis becoming a standard procedure ina wide rangeof investigations outside chemistry and physics, including among others, industrial processcontrol, clinical diagnoses, isotope fractionation, radioactive half-life determinations,andenvironmental monitoring. In1942 Consolidated Engineering Corporationbuilt the world's first commercially available mass spectrometer forthe AtlanticRefining Corporation [20].As new applications haveemerged, specialized instrumentshavebeen developed tobetter suit particular constraints, resulting inhighly innovativecomponent variations such as ion source, mass selector, and detector. Technologicaladvancements, particularly inthe areas ofintegrated circuitry and materialsscience such as magnetic materials! have permitted thedevelopinent ofmore compactmass spectrometers. This, in turn,has enabled mass spectrometers for in-situ investigationsrangingfrom terrestrial battlefields [23],to Jupiter's atmosphere [24]. These"field portable" instruments werereduced insize and ruggedized tobe transported byspacecraft [25],aircraft [26], and large land vehicles [27,28] . Morerecently, a generationofminiaturized mass spectrometers havebeen developed which are suflcientlycompact andlightweight tobe carried bya persontoa measurement site[23, 29-31].

'I.4Autonomous In-situ mass spectrometry Massspectrometers arewell suited forautonomous in-situanalysis ofdissolved gases ina watercol~ becausea mass spectrometer canquickly detect multiple dissolved chemicalsatlow concentrations, andcan work without exhaust orconsumable reagents. Byimplementing automated control and data processing such as calibration and fragmentpeak separation! sample components canbe quickly identified andquantified in thepresence ofmixtures. Incontrast, conventional environmental chemical analysis

20 suchasgas chromatography orwet chemistry involves collecting andtransporting samplestoa laboratoryforoff-site analysis, resulting insampling andmeasurement that isgenerally labor intensive andtime-consuming. Additionally, error often arises because ofchemical andphysical changes occurring tothe sample during transport suchas degassingandbiological orphotochemical degradation. Although in-situ devices such as dissolvedoxygen probes avoid the drawbacks encountered withoff-site analysis, theyare commonlylimited todetecting oneor a fewgas species [32], with separate sensors requiredforeach specie andsensitivities oftypically about 1 ppm. Continuous samphng techniques,suchas the Weiss equilibrator, aregenerally limited toshipboard usewith modestsampling depths, andhave equilibration timesthat vaiy with gas species and range&om minutes tohours [33, 34]. The resulting data sparseness oftenleads to temporalaliasing translating intospatial aliasing when sampling froma moving platform!,which can mask important chemical andbiological features inan aquatic environment. Variousapproaches havebeen undertaken toaddress thedesign constraints of submersiblemassspectrometer operation. Instruments suchas the TUHH submersible membraneintroduction gaschromatograph massspectrometer GCMS! usea commerciallyavailable EM640 field portable GCMS housed within a cylindricalmetal pressurevessel [35!. This system, which uses a hollowfiber analyte introduction system, displacesapproximately 400kg and draws several hundred watts, requiring thatit functionaboard a remotely operated vehicle ROY! which ispowered andcontrolled via a tether&om a surfaceship. RecentlyAUVs have been developed withreliability, payload carrying capacity, and missionendurance adequate fordeployment ofa mass spectrometer. Nevertheless, mass spectrometeroperation aboard anAUV still introduces several critical instrument payload designconstraints, including manditory autonomous operation, smallsize, and exceptionallylowpower consumption [14].Instruments suchas the University ofSouth FloridaCenter forOcean Technology's COT!quadrupole andion trap mass spectrometers[36]utilize off-the-shelf butrepackaged Leybold Inficon Transpector and VarianSaturn 2000 analyzers, respectively. Bothuse a membraneinletsystem designed foroperation aboard anAUV. The COT quadrupole instrument hasbeen deployed aboard

21. the 6.4 meter 1,4 ton displacement ARCS AUV [37], but both systemsconsume power at high rates in excessof 100watts!, havedelicate membrane inlet systemsthat limit depth of operationto 30 metersand samplingrates no greaterthan 12 samplesper hour for their operatingtime of a few hours, The core design concept of the NEREUS membrane inlet mass spectrometer,was to developan in-situ dissolvedgas analyzer that is fast, robustand highly optimizedfor operationonboard a smallautonomous underwater vehicle andvarious other submerged platforms,including as a self-containedautonomous moored system [38]. To address these and other constraints, the NEREUS system is completely self contained within a 17-inch-diaineter glass pressuresphere, compatible with the Odyssey-classAUVs. It weighs22 kilograms,and is capableof operationto depthsof 100meters a depththat can be extendedwith minor modification! on a self contained power supply drawing less than 20 watts, while sampling at rates exceeding 240 samplesper hour. The NEREUS systemis alsounique in that it requiresno reagents,has no movingparts, and possesses an autonomouscontrol system capable of adapting its mission directives and sampling regimesto bettermonitor its environment.Much of the NEREUSprototype draws upon prior researchfocused on componentdevelopment [39], which culminatedin the constructionof vital NEREUSelements such as the vacuumenvelope and data acquisitionsystem, and the modehngof systembehaviors such as membranekinetics and vacuummaintenance within the analyzer.A previousgeneration backpack portable mass spectrometer[29], servedas the conceptualmodel from which the preliminaryNEREUS designwas based,Drawing from theseresources, NEREUS development proceeded with completionof a prototype chapter2! andbench testing chapter3!. After benchtesting, it was deployedas a stationary"buoyed" type sensor chapter5!. Upon successful demonstrationas a buoyedtype sensor,the Kemonautvehicle was designedand built chapter 4!. NEREUS was then configured for operation onboard the Kemonaut AUV and deployed chapter5!.

22 NEREUS COT Quadru ole COT lon Tra TUHH GC-MS" Power w 20 95 150 500 Wei ht k 21 75 65 Size m' 0.014 0.035 -0.6 0.11B Anat zer mass ran e AMU 2-150 2-200 unknown-500 2-640 Maximum sam lin rate /hr 600 12 12 Detection limit b 1-10 1-10 10-100 Instrument endurance Weeks Da s Da s Unknown De thratin m 100 30 30 300 Vehicle latform AUV AUV Intended for AUV RQV

Table1-1: Comparison of submersible mass spectrometer specifications

Whenused in conjunctionwith the Kemonaut or OdysseyIII vehicles,NEREUS can rapidlygenerate high-resolution data for mappingtemporal and spatial distribution of dissolvedgasses on local to basin scales. The relatively light 200 kg displacement and 300meter depth rating of theKemonaut vehicle enable routine single day operations, withminimal support equipment e,g. a smalldinghy and transponder array!, throughout theentire water column of manyfreshwater bodies such as inland lakes and rivers, as wellas shallow marine coastal areas like Chesapeake Bayand Cape Cod Bay. The larger OdysseyIII,which is capable to6000 meters depth and possesses andendurance ofup to 7 days[41] can, with minor modifications, bedeployed with the NEREUS system to generatehigh resolution maps of volatilechemicals, covering a lineardistance of over 1,000kilometers, containing approximately 7,000 high resolution spectra of multiple chemicalspecies. In contrast,current technologies. require orders of magnitude longer time &amesto generateequivalent data sets.

' Doesnot include submersible inlet apparatus EM640specifications from40, Einfetd,W., et al., Environmental Technology Verificarion ReporI, l 997,US Environmental Protection Agency, Office of Research andDevelopment; Washington DCp. l- 78.

23 Figure1-2:NEREUS instrument pictured beside laboratory microscope for size reference

1.5 AppIIcalions

Dissolvedgases in marineand freshwater environments canvary significantly both spatiallyand temporally [42-45j, Potential applications for theNEREUS-Kemonaut systemare numerous, including geochemical, ecological, hydrological, and chemical fate analyses. Geochemicalapplications arewide ranging including marine mapping of dissolved gases&om hydrothermal vents, active volcanic areas and low temperature ocean floor seeps.Investigations are valuable for offshoreoil exploration[46j and global climatologicalresearch [47, 48j. Methaneis often observedin increasedconcentrations in areasof freshwater inputs [49, 50j, anaerobic zones of bottom sediments [S1 j,oil andgas deposits,destruction of crystallohydrates,andemission along fractures in theearth's crust [52j. In-situmass spectrometric analysis of terrestrialvolcanic ermssions is usedas an indicatorof volcanicactivity, particularly with regardto Carboru'Sulfurratios and Helium concentration[53, 54j. Active underseavolcanic areasare alsoknown to emit volatile chemicalsinto the water column [55-58], Thus, analysis of gasesdissolved in marine watersmay be useful for predicting undersea eruption events. For example, Kick'em Je'nny,aii undersea volcano just north of theisland of Grenadalast erupted in December of 2001,Micro-beam survey of thevolcano has revealed a massive horseshoe shaped faultscar which most likely resulted Rom a majorslope failure of anearlier cone, suggeststhat the volcano may be unstable and a futurehazard for the region [59], Survey of dissolvedchemicals such as SOq, CO~, and He in watersoverlaying the cone crater, whichlies about 190 meters below the sea surface, may be useful in predicting the activityof thevolcano. Advanced warning of impendingeruptions are particularly importantinvolcanic regions such as this to allow adequate time for evacuation ofnearby islandsto preventlarge scale loss of life [60,61]. Deep-seahydrothermal vents have also been subjects of increasedinterest for global climatologicaland biological research. For example, sources such as anhydrite chimneys of theNorth Fiji Basinhave been shown to emit gases in concentrationsexceeding 14,5 partsper thousand of carbon dioxide, 49 ppm of methane, and approximately 937ppb of helium[62].' In some cases, thereduced gases from these vents are known togive rise to entirechemosynthetic ecosysteins [63, 64] which oxidize methane orsulfide to produce energy.Detection of gasesat theseconcentrations is possible with theNERUS instrument. Hydrologicresearch often utilizes dissolved gas data to determine source and age of waters.NEREUS is a potentiallyconvenient tool for studyingmixing of watersin marineand fresh water bodies. Concentrations of oxygen, nitrogen, carbon dioxide, and methanecan provide valuable clues as to originsand inflow rates of groundand surface watersinto lakes, rivers, and coastal areas [65, 66]. Theability of theNEREUS system to identifyand quantify the distribution of thesebiogenic gases in a watercolumn is also usef'ulfor many aspects of ecological research. High spatial and temporal resolution data is valuablefor monitoringphenomena such as algal blooms, seasonal stratification & turnover,productivity, and eutrophication [43]. In addition,the NEREUS system could beused to unobtrusively monitor water quality, especially urban runoff shipping lanes, andpoint sources of pollutionsuch as sewage outfalls [18].

For consistency,concentrations are expressed as a massfraction

25 Mappingof marineoil spillsfor impactassessment and cleanup is yetanother potentialNEREUS application. Oil spillsurveillance relies almost exclusively on remote sensingsuch as infrared, ultraviolet, and radar imaging from satellites and planes [67, 68].These approaches are only suitablefor detectingsurface slicks, and are unable to detectdissolved petroleum fractions at depth.Oil spill assessmentbased on surface slick as well as sub-surfacedata is importantwhen DNAPL or water-soluble&actions are presentand when wave induced mixing is a factor,For instance, approximately 77/0 of the825,000 gallons of petroleumfrom the North Cape oil spillwas dispersed into the watercolumn [69], Estimates of thisspill's size,based upon oil-slick area, underestimatedthemagnitude ofthe disaster, thus contributing toan ill-prepared cleanup response.Although NEREUS would be unable to detectmany of theheavier hydrocarbonswith low volatility or mass to chargeratios greater than 150, such as poly- aromatichydrocarbons in an occurrence such as this, NEREUS sensitivity models predict detectlimits of manypetroleum fractions, including benzene, toluene, butane, and methanein concentrationsdown to thetens of parts-per-billionlevel. Furthermore, the petroleum"fingerprinting" may be possiblebased on the relativedistribution of these hydrocarbonconstituents [70].

1.6 System performance

Freshwaterand marinedeployments have been conducted with theNEREUS instrumentand the Kemonaut AUV. Tethered and towed operations of the NEREUS instrumenthave established theability to map dissolved volatile chemical species in the watercolumn. Sea trials of theKemonaut AUV demonstratedtheunique ability of this Odysseyclass AUV to carrythe NEREUS instrument as payload. Analysis of vehicle designand performance indicates that the Kemonaut-NEREUScombination is well suitedfor scientificsurvey of dissolvedgases in &eshwaterand marine environments to depthsof 100meters. NEREUS data collected during these operations confirm the ability of theinstrument toefficiently collect meaningful chemical data in a varietyofplatform configurations.

26 Chapter 2

MECHANICAL AND ELECTRONIC

HARDWARE

TheNEREUS instrument is fully self-containedwithin a 17"Benthos glass pressure sphere Figure 2-1!. This housing was chosen based on itscompatibility with Odyssey classAUVs andits ability towithstand the high compressive loads experienced at abystnaldepths. Design and construction of theprototype for operationonboard the AUV, whilewithin the pressure sphere, required careful analysis and optimization. Foremostamong design concerns were: instrument miniaturization, response time minimization,chemical sensitivity maximization, power consumption minimization to

Figure2-1: The NEREUS instrument with upper glass hemisphere removed

27 Mappingof marineoil spillsfor impactassessment and cleanup is yetanother potentialNEREUS application. Oil spill surveillancerelies almost exclusively on remote ' sensingsuch as infrared, ultraviolet, and radar imaging from satellites and planes [67, 68].These approaches are only suitablefor detectingsurface slicks, and are unable to detectdissolved petroleum fractions at depth.Oil spill assessmentbased on surfaceslick as well as sub-surfacedata is importantwhen DNAPL or water-solublefractions are presentand when wave induced mixing is a factor,For instance, approximately 77'10 of the825,000 gallons of petroleum&om the North Cape oil spillwas dispersed into the watercolumn [69]. Estimates of thisspill's size,based upon oil-slick area, underestimatedthemagnitude of thedisaster, thus contributing to anill-prepared cleanup response.Although NEREUS would be unable to detectmany of theheavier hydrocarbonswith low volatility or massto chargeratios greater than 150, such as poly- aromatichydrocarbons in an occurrence such as this, NEREUS sensitivity models predict detectlimits of manypetroleum fractions, including benzene, toluene, butane, and methanein concentrationsdown to thetens of parts-per-billionlevel. Furthermore, the petroleum"fingerprinting" may be possiblebased on the relativedistribution of these hydrocarbonconstituents [70].

1.6 System performance

Freshwaterand marinedeployments have been conducted with the NEREUS instrumentand the Kemonaut AUV. Tetheredand towed operations of theNEREUS instrumenthave established the ability to mapdissolved volatile chemical species in the watercolumn. Sea trials of theKemonaut AUV demonstratedthe unique ability of this Odysseyclass AUV to carrythe NEREUS instrument as payload. Analysis of vehicle designand performance indicates that the Kemonaut-NEREUScombination is well suitedfor scientificsurvey of dissolvedgases in &eshwaterand marine environments to depthsof 100meters. NEREUS data collected during these operations confirm the ability of theinstrument to efficientlycollect meaningful chemical data in a varietyof platform configurations.

26 lessthan 25 watts, ultra-high vacuum maintenance within the vacuum system, autonomousoperation and data interpretation, resistance to vibration,impact, and rollover.NEREUS can be ostensibly divided into five component ~.oups: vacuum system, analyzer,electronic hardware, software, and mounting apparatus,

2.1 Vacuum system

Two paramountchallenges to submergeddeployment of marinesensors are to survive thehydrostatic pressures encountered at depthand the corrosive effects of seawater.In additionto thoseconsiderations, in-situ mass spectrometer design presents the unique requirementsof ultra-high vacuum and magnetic field maintenance. Therefore, the vacuumsystem, which consists of themembrane inlet apparatus,inlet tube, vacuum envelope,and ion pump were fabricated using ¹304 stainless steel. Although stainless steelis generallymore porous to gasesthan other metals such as aluminum, 304 stainless waschosen for constructingthe sample inlet apparatus, sphere penetration, inlet line, and vacuumenvelope, based on weldability, low magnetic permeability, chemical inertness, andmechanical strength. Like the pressure sphere, the sa.mple inlet apparatus, which servesas the interface between NEREUS' high vacuum analyzer and its aquatic environment,is required to withstandambient hydrostatic pressure. In addition to externalpressure considerations, theinlet system must also allow for adequate permeationof analyte gases into the vacuum chamber while maintaining an internal vacuum of no greater than 10 Torr;

2.1.1 Vacuum envelope Thevacuum envelope is roughly the size of thesoft drink can Figure2-2! and was designedto besmall enough to fit withinthe pressure sphere, yet permit satisfactory conductanceof excessgases into theion pump[39]. Volume minimization also serves to decreasetheoverall vacuum system's internal volume, thereby minimizing pump down requirementsand increasing conductance, thus decreasing the residence time of analyte gasesin theregion of thecycloid. The cycloid maintains a highvacuum seal with the vacuumenvelope by wayof eightbolts that firmly compressesa 2.112" diameter PTFE 0-ringbetween the 3 3/8"diameter envelope flange and the cycloid flange.

28 Figure 2-2: Vacuum Envelope cutaway view

2.1.2 Inlet apparatus Preferentialselectivity of dissolvedgases over water and water vapor is accomplished by meansof a thinsemi-permeable membrane that is passivelyexposed to thewater column,thereby allowing analyte gases to diffuseinto the vacuum system. The membraneinlet is positioned outside the pressure sphere and connected tothe analyzer with a smalldiameter approximately 2mm! flexible stainless steel inlet tube. The inlet tubetraverses the lower hemisphere through a .6inch diameter penetration point and 1.5" diametermounting flat. Although differing in dimensionand backing plate material, this inletdesign is similarto the membrane inlet described byHoch and Kok to permit continuoussampling of gasesfrom photosynthetic organisms, that utilizes a thin polyethylenesheet which is heldin placeover a porousplate by a threadedannular crew

29 cap and TeQonwasher [71]. Whereas,the NEREUSinlet apparatusconsists of a 1mil thicknesspolyethylene membrane placed over a stainlesssteel micro-etched backing plateand secured by a Teflonwasher in serieswith anannular stainless steel cap that is threadedonto the stainlesssteel inlet body [72]. Polyethylene is superiorto other membranematerials such as PDMS for submergedoperation because its relativelylow permeabilityto water[73]. In addition,the small .01" diameter holes of theporous backingplate alongwith the low compressibilityof this material,relative to other polymers,help avoid excessive compression and sagging encountered at depthwith other inlet designs i.e. sihconerubber tubing!, thereby minimizing unwanted permeability variation as a functionof depth.

1.E+02

- - 1.E+00 gE1,E02 o

1.E-04 Cl1.E-06

1.E-080 10 20 30 40 50 60 70 80 90 100 depth m!

Figure2-3: Membranecompression and saggingresulting from hydrostatic pressure affecting membranepermeabmty for the NEREUSinlet systemusing 0.001" thick LDPE materialversus 0.018"thick silicone tubing inlet system. Permeability change calculated using Young's modtslus to determinemembrane thickness at a givenpressure, svhich isthen applied to calculate change in gas fluxrate as a functionof pressuredependent membrane thickness refer to Equation5-4!. V'alues expressedas percent change normalized to valueat 0 depthor 1 atmosphere.

TheNEREUS membrane inlet threads onto a VCRfitting on the inlet feedthrough, whichfunctions to transportanalyte gases to theanalyzer residing within theglass

30 pressuresphere, while preventing water from entering the glass pressure sphere. The feedthroughtraverscs the lowerhemisphere wall usingan integrated 1 V~inch diameter flangewith a recessedrubber 0-ring that contacts a groundglass flat surroundinga 0,6 inchdiameter feed-through hole on the lower hemisphere's external surface, providing a water-tightseal Figure 2-4!. The feedthrough is secured onto the pressure sphere using a rubberwasher, to preventdamage to theglass, in serieswith a 3"stainless steel washer, followed by a stainlesssteel backing nut.

Figure2-4: Sampleinlet apparatus

Portionsof inlet tubingresiding on theinterior of thepressure sphere are constructed of thin walledcorrugated tubing because these sections are only requiredto v ithstand approximatelyone atmosphere of pressure, but must be flexible to accommodatepartial disassemblyof the instrument during maintenance procedures. A flexible corrugated inlet lineis connectedto the internal junction of theinlet tube penetration to carry sample perineant to a '/~inch diameter manifold, located at the rear of the instrument. This manifoldbifurcates and attaches to two parallelhigh vacuum bellows valves model SS- BNV51!,linking to separate,bent rigid, lines that terminate at a quick-flangetype fitting andat thevacuum envelope inlet. Theseparallel lines allow, respectively, the vacuum systemto berough pumped down to a pointat whichthe ion pumpcan begin useful

31 operation,and the vacuum envelope and ion pump to besealed off fromthe inlet during storage.All vacuumsystem connections use VCR type couplings, Theinlet apparatus is the most likely component to fail underhigh hydrostatic pressure.This is dueto thefact that the inlet membrane resides outside of thepressure sphereand is exposedto ambientwater pressure while inaintaining a near-perfectvacuum onits internalsurface. Possible failure modes include backing plate collapse caused by excessivehydrostatic pressure, membrane rupture caused by excessivehydrostatic pressure,as well as excessivewater vapor influx acrossan intact membranedue to a thermallyor chemicallyinduced increase in membranepermeability. In additionto the problemsassociated with a catastrophicfailure of theinlet, significant water vapor input cancause a reductionin signalresponse through the collision of watermolecules with analyteions. Based on calculationsprevious calculations [39], inlet membrane and backingplate strength was determined to allow for anabsolute maximum deployment depthof approximately100 meters, with the backing plate being the limiting factor. Destructivepressure testing of theinlet system design has shown the inlet to successfully withstandpressures greater than 100 psi, corresponding to a waterdepth of approximately75meters [72]. To date the NEREUS system has been successfully deployedto a depthof 25 meters,

2.1.3 lon pump Thevacuum system must first be rough-pumped down to a pressureof no greater than 10 Torrbefore the instrument's onboard ion pinnpcan be used to createand maintain freemolecular low conditions within the vacuuin system [39]. This 8 Vsdiode type ion pumpis bolted directly to thevacuum envelope at a 2'/~ inch conflat flange fitting locatednear the center of thevacuum envelope to maximizeconductance of excess gas away&om the cycloid. The ion pump is poweredby a highgain DC-DC converter which acceptsthe +12 volt input from the battery and generates +3000 volts at its output. When operatingat undernormal gas loads at steady state conditions, the ion pump typically drawless than 200 pA, or 0.6watts. The ion pump and power supply, combined, require lessthan 2 wattsof power.

32 Figure2-5: Ion pump with un-machined housing!, cycloid and vacuum envelope assembly

Owingto the ion pump's weightiness, thepump housing has been radically machined to helpreduce overall instrument weight to a pointof neutralbuoyancy. This entailed machiningaway the pump housing's mild steelsidewalls to a totalcross sectional area of 2 in, therebystill permitting sufficient return path for the ion pump's magnetic flux densityof 1,270Gauss. Additionally, concave lenticular surfaces were machined into the externalfaces of thepump housing. These surfaces were produced bymachining a series of concentriccircles of parabolicallydecreasing depth as a functionof distancefrom the center,intended to accommodate theincreasing magnetic flux density developed in relationto distance from the center, The resulting housing weighs 2.8 kg, a 33lo reductionfrom its initialweight of 4.2kg, yet maintains a fluxdensity of 1,270Gauss across its interior.

33 Figure2-6: Machinedion pumphousing

2.2 Analyzer

The analyzercomponent group, which is mostly containedwithin the vacuum envelope,includes the ion source,cycloidal massselector, Faraday cup detector,and elecrometer.A permanentmagnet, which is positionedoutside of thevacuum envelope providesa homogenousB-field for the cycloid. During operation,sample gases first diffusefrom the water column in acrossthe membrane inlet, and then migrate through the inlettube, toward the vacuumenvelope. The limited volumeof the inlet system, particularlywithin the vacuumenvelope, has the addedadvantage of shortresidence time. Oncegas molecules enter the vacuumenvelope, they areionized and then acceleratedthrough the massselector, finally impactingthe Faradaycup detector. Electricfield for thecycloid is producedby a highvoltage operational amplifier which suppliesa variablepotential to acceleratorplates within the analyzer.Total instrumentpower consumption is keptbelow 20 watts,in partthrough a highfrequency,

34 duty cycle modulated,emission regulator [38], To furtherreduce power consumption,a permanentmagnet is usedto generatethe requiredB field acrossthe analyzerregion.

2.2.1 Analyzer magnet

The analyzermagnet is of an asymmetricU-shaped geometry with an air gap of 1 inch andpole piecediameters of 3.5 inches.The pole piecesand yoke areconstructed of low carbonsteel to allow for maximummagnetic permeability, while the coerciveforce of themagnet is generatedby two cylindricalNdFeB magnetic members, each residing betweena polepieces and the yoke. Polepieces are shaped to minimize Gingingeffects within the air gapregion, whereasthe yoke is shapedasymmetrically to minimize space requirementsand mass. The analyzermagnet of the Hemondbackpack instrument served as an initial benchmarkfor the NEREUSmagnet, with a weightgoal of lessthan 11 kg and a flux densityof at least3,500 Gauss for the NEREUSanalyzer magnet. Preliminarymagnet designs were prepared using SoldWorks CAD softwarefor shape optimization[39] basedon the maximummagnetic permeability B ! of low carbon steeland a minimumcalculated cross sectional area of 3.85 in . This type of steelwas chosenbecause of its higher B~ relativeto standardmild steel.The yoke andpole piecesof the magnetbody werethen machined &om this metal,using a venerable Monarchlathe and Bridgeport end mills at theParsons Lab machine shop. The magnet's two cylindrical NdFeBelements were constructed by GroupArnold at low costby virtue of pie shapedmagnets that havebeen compressed and sintered together into a cylinder. Thesecylindrical magnets are nickel plated to protectthe brittle magnetic material from crackingand to minimizeexposure to damaginghydrogen gas. Once initial machining and fabricationwas complete,four bolt holeswere drilled and tappedinto the rear of the magnetbody to securethe yoke pieces together. Then, using extreme care, the magnetic elementswith theirpole pieces were positioned on the yoke and are held securely in placeusing only the magnetic field itself.It is worthwhilenoting that serious dangers are posedto fingersand hands by metal objects such as wrenches, which can be caught within themagnet's field andrapidly accelerated to highvelocities.

35 Flux measurementsof this initial designrevealed a flux densityof 4,200Gauss. To further decreasespace requirements and weight, the magnetwas carefttlly disassembled and a subsequentmachining of thc magnetyoke wasundertaken to aggressivelytrim away areasof low magneticflux. Areasof reductionwere selectedpresuming curved lines of magnetic flux through the magnetic return circuit and that cross sectional area requirementsdirectly abovethe magneticelements range from zeroat the distal end of the body, to maximum at the proximal end. Therefore, the external intersections of the magnetbody wherethe bolts areattached! were rounded using a 1.125in radiusof curvatureand a 16' bevel wasmachined on oneof the side surfaces the surfacefacing the electrometerand ion pumpwhen fully assembled!and on the surfacesdirectly above andbelow the magneticelements. The resultingmagnet weighs approximately 8 kg and generatesa flux densityof 4,270Gauss across the air gap,along its centralaxis.

Figure 2-7: Asymmetricallyshaped analyzer magnet

36 2.2.2 Cycloid analyzer NEREUSemploys a modified CEC 21-620 cycloidal type analyzer, asdescribed by Robinsonand Hall [74j, which uses orthogonally crossed fixed homogenous magnetic andvariable homogenous electric fields to impart trochoidal trajectories tosample ions Figure2-9!. The cycloidal ion trajectory isgiven by Bleakney and Hippie [75j. This geometryhasthe inherent property ofperfect direction and velocity focusing, making the analyzerless sensitive tomisalignment andvibration, and because theion trajectories loopin on themselves a relatively compact flight path, and therefore size, is achievable, Thesegeometry characteristics allows for a massrange of approximately2-150 atomic massunits, and a massresolving power of 100I20j, permitting detection of dissolved biogenicgases, atmospheric gases, light hydrocarbons, andthe differentiation ofmany isotopes.

Figure2-8: lon trajectoryof cycloidalmass analyzer

Modificationstothe cycloid include the addition of a cycloid-mountedinlettube, whichdirects incoming analyte gases into the ionization chamber, described indetail by Camilli[39j. Also, the cycloid used within the NEREUS prototype lacked an electron trap.To correct this shortcoming, anelectron trap was fabricated from a stainlesssteel double-buttedbicycle spoke that was cut and filed into an L-shape with a flattenedsquare tip.This electron trap is held in position by a stainlesssteel armature, fabricated from a discardeddesktop PC motherboard shield, which is bolted to plate 0 and,inturn, bonded

37 to the electrontrap with non-conductiveepoxy. In addition,an aluminuinshim hasbeen insertedbetween the cycloid's mounting arch and plate 0, therebyproviding an offset to properly align the cycloid within the vacuumenvelope, The shim measures0.865" x 0.105"x 0.275"with two 0.15"diameter holes drilled with a 0.6"spacing to accommodatetwo cycloidmounting screws. The relatively close tolerances of thecycloid within thevacuum envelope necessitate that the cycloid be aligned to within 1 millimeter of its requiredposition to preventshort-circuiting of theelectric field plates to thewalls of the vacuuinenvelope. Samplesare first ionizedby theion sourcehoused within the ionizing chamber portionof thecycloid, located immediately in front of theion pumpinlet. These ions are producedusing a heatedtungsten filament to thermionicallyemit electronsthat are then acceleratedby a 70voltpotential, causing these electrons to impactand ionize analyte moleculesthat have entered the ionization chamber, An electrontrap located to the exteriorof theionizing chamber collects &ee electrons and negatively charged ions. Positivelycharged ions are then repelled out fromthe ionizing chamber by a repeller plate heldat a positivepotential relative to theion chainber!and accelerated into the cycloidmass selector by theinjector, which carries a positivecharge smaller than that of therepeller. Once within the mass selector region of thecycloid, the analyte ions are actedupon by thefixed magnetic field developed by the analyzer magnet and orthogonal electricfield createdby five electricplates, causing the ions to assumecurved trajectories towardone of two Faradaycup detectors. By varyingthe potential of theelectric field in unisonwith therepeller and injector potentials, ion beamsof varyingm/z can be made to comeinto focuson oneof thetwo Faradaycups. Ions with mass-to-chargeratios m/z! of 12or lesscan be madeto assumetrajectories in thedirection of the 180' sectorFaraday cup,while ions with m/z of 12or morecan be focused on the cycloidal path Faraday cup. All controlsignals are sent to thecycloid through its Amphenoltype 17 pin connector whichis securedto the3 1/8"flange opening of thevacuum envelope. Connection pins and functionsare listed in Table2-1: Cycloid pin out. Figure 2-9: NEREUS cycloid tube constructed from a modified CKC 21-620

39 cloid in Function Mass selector

A Filament

B ShielcUground 142

Plate heater

Plate heater Plate 0/ ground Electron trap

Plate 2 Plate heater omitted! Plateheater omitted!

Filament heater

Plate 5

N Plate 1

Plate 3 Q Injector

R Repeller 10

Filament

Plate 4

Table 2-1: Cycloidpin out

Theion beamsgenerate an ion currentwhich is detectedby theFaraday cup and sent to theelectrorneter via oneof thetwo BNCtype feed-through ports located on the vacuumenvelope's 2 /i inchconflat flange Figure 2-10!. The conflat flange contains two machinedrecesses that prevent the Faraday cup pickups &om contacting the conflat flangeand producing a shortcircuit. Ion current is sentfrom the Faraday cup pickups to curvedmusic box-like tines that extend perpendicularly from the internal feed through rods.These contacts are constructed from the same lightweight stainless steel material as theelectron trap mounting bracket and are tin solderedto thefeed through rods. Using thesefeedthrough contacts, the electrometersenses this ion currentand transforms it to an

40 amplifiedvoltage signal, This electrometer output voltage is thenconverted to a digital signalby a dataacquisition DAC! board and transmitted to anembedded computer, which interprets and stores this data.

Figure2-10: Conflat flange with dual BXC feedthrough

2.3 Electronic Control

Autonomousoperation requires that the MIMS instrumentbe able to performall necessaryfunctions via a pre-programmedembedded computer system i.e. a computer systemresiding within the instrumenthousing!. Minimum functionality includes poweringinstrument on & off at predeterminedintervals, mass step calibration, acceleratorpotential control, emission regulator control, data collection and handling, and systemsdiagnostics. To fulfill theseoperation needs as well as power and space limitations,NEREUS uses embedded PC-104 computer coupled with mass selector controller,data acquisition system, electrometer, and an emission regulator,

41 2.3.1 Embedded computer Althoughhigher clock frequency Pentiurn type systems offer the highest performance,it comes at anenergy use premium. Given the power constraint of approximately5 watts for the computer and DAQ board, an Ampro 386 SX core module operatingat 25MHz was initially utilized in combinationwith an Ampro VFP-II VGA flatpanel controller. The IDE bus of the386SX module does not permit use of a flashablesolid-state PC card! disk for non-volatile memory storage, therefore, the 1 MB EEPROMwas partitionedallowing it to function asa harddrive. After successfulshort- termdeployments, the core module was replaced with a 486DXmodule operating at 100 MHz;thereby permitting use of a removablesolid-state flash disk and thus extending availablemission time. In additionto theincreased storage space, the faster clock speed andonboard math co-processor of the substitutedcore module accelerates command line execution,resulting in anincreased sampling speed of upto approximately1500 samples per secondwhile adheringto a powerbudget of lessthan 5 watts. Eachdata point requires information concerning scan voltage, m/z, and time, imposinga memoryrequirement of up to a maximumof 248 data bits per sample. To addressthe data storage needs, NEREUS uses a conunerciallyavailable 128MB SanDisk CompactFlash disk commonlyused in digitalcameras!, which has the advantages of extremelysmall form factor, low power consumption, high tolerance to vibration,impact, andpressure fluctuation, low cost,and modular data transfer from the NEREUS computerto a laptop i.e. by physicallyexchanging cards!, Presently the embedded computeruses this solid-state disk as a bootabledisk, which contains the operating systemand instrument code compiled as an executable file. The operating system and instrumentcode together require approximately 500KB, allowing the remaining 127MB to bereserved for datastorage, Because NEREUS employs signal averaging, memory requirementsare decreased to less than 2 MBper hour of operation,thus permitting at least2 'li daysworth of datastorage when the instrument is in continuousoperation,

2.3.2 Communication In additionto onboarddata storage, the embedded computer is capable of communicatingto send data or receivecommands from either of twoRS232 serial ports operatingeither a nullmodem toAUV control computers ora halfhandshake protocol to a radiotransceiver viaa tetheredsurface float. The radio transceiver receives power from theinternal NEREUS power system, with a totalpower requirement ofapproximately 150mW, and operates using a 418MHz FM frequency withan effective broadcast range ofapproximately 150meters when the companion laptop transceiver iswithin clear line- of-sightand broadcasts toshorter distances when obstructions arepresent. Communicationswitheither radio transmitter orstandard wired serial link require transmissionat9,600 bits per second through a 50 ft. long 9-pin Impulse MIL-9-MP waterproofmarine cable pin arrangementis described in Table2-2.

m ulse cable in color Function DB-9

Black GND

%hite Power

TXD

Green

Orange RTS

Blue CTS

Purple DSR

Yellow CD

Brown DTR

Table2-2: Waterproof cable pin assignments

2.3.3Data acquisitionand instrumentcontroi Thecontroller/data acquisition system, orDAQ board, is a revisedversion of the circuitpreviously described byCamilli I 39], which functions asa compact2cm x 9 . cm!,low power .5 watt!,16-bit resolution controller andsensor. It is compatible with computersusing 8X86 through Pentium microprocessors, interfacing viaa printerport conformingtoa PS-2/bi-directionalprotocol IEEE1284!. The IEEE 1284 standard permitsa maximumdata width of 8-bits;consequently, theDAQ board makes use of a rnultiplexorarray, which allows for a "twopass" 16-bit transfer of both controller and

43 sensordata. The DAQ boardrelies on anAnalog Devices7884 ADC to sendcommands to themass selector and an Analog Devices 569 DAC to quantify electrometer voltage; bothconverters possess sample-and-hold capabilities, permitting simultaneous operation of the sensor and controller. Thisgeneration DAQ designincorporates several improvements over the previous generationdescribed by Camilli[39]. Included is anoptional 5V inputoffset circuit for electrometerinput jumperselected!, thereby enabling use of thefull 5V to+5V scaleof theAD7884 analog-to-digital converter, Additionally, all 5 statusregister bits of the parallelport havebeen made available for possibleuse as indicatorsof instrumentstates suchas excessive internal pressure, ion pump failure, fail-safe activation, and low battery condition.Bit 3 ofthe control line operatesa relaycontrolling the emission regulator's on/offstate, thus permitting a coinputerselectable low power sleep mode. Additionally, digital-to-analogoutput is nowrouted to anon-board isolation operational amplifier, whichuses the scan board's "floating" power supply to generatescan board input voltages,Finally, the new DAQ designmakes use of AnalogDevices' BQ andAN line of integratedcircuits, which offer improved noise characteristics over AQ models. Errorcorrection algorithms have been incorporated into theDAQ boardexecution codeto increaseaccuracy. The controller correction is simplya computed4mV offset, whilethe sensor correction is a slope-offsetalgorithm designed to accommodatea 2's complementarchitecture, To further decrease sampling error, a signal-averagingregime hasbeen included into the sensor code. Previous DAQ board performance testing demonstratedcapabilities of sense and control functions with sample speed independent 16-bit accuracy,and operational speed dependent on the commandline executionrate of its host computer[39].

2.3.4 Mass Sefector Electricalpotential is deliveredto therepeller, injector, and accelerator plates by an isolatedhigh-gain circuit. This circuit is adaptedRom the Hemond scan board [29]; a designthat is, in turn,based on a CECcontroller developed in the early 195Gs [74] which relies uponvacuum tubes common to that era. TheNEREUS scan control input voltage, generated bythe embedded computer and DAQ,is relayed via an onboard BurrBrown model 124 isolation operational amplifier. Theoutput of the iso opamp isrouted to the negative input of a heat-sinkedoperational amplifier Apex model PA88! acting as an inverting amplifier. This high voltage opamp providesa gain of approximately 60X relativetothe input scan control voltage. It is necessaryto isolate this high voltage circuit so that the repeller, injector, and accelerator platescan make use of a "floating"ground reference, thereby allowing accelerator plates ¹1, ¹3, and¹5 focus!to develop substantially negative potentials without requiring a dedicatednegative power supply. Theoperational amplifier makes use of a "floatingground" power supply created by a %15volt DC-DC converter PICO model 12GR150!, with the +15 volt output transformedinto+300 volts by a secondset of threeserially connected 100 volt DC-DC conveiters PICO model 12A100!. The +300 volt outputis thenrouted through a 1000 ohmresistor to limit current,A 10pFaccumulator capacitor is shuntedto thecommon terminalof thecircuit to stabilizethe 300 volt output, while two 150volt zener diodes in seriesprotect the circuit from reverse polarity damage. Finally, the +300 volt and 15 volt suppliesare routed to thesupply pins of thePA88. Theoutput of the PA88 is connected toa resistorladder, which acts as a voltage dividerfor the accelerator plates, These plates are arranged so that the even numbered plates and4! developrespectively increasing positive potential, and the odd lumbered plates,3, and5! developincreasing negative potential, respectively. Plate 0, which also servesas the cycloid mounting spar is held at ground potential relative to the negative electrodeofthe battery pack and vacuum envelope. Resistor ladder arrangement is detailedin Figure 2-11. Thescan board was initially unable to generatethe requisite plate voltages for focusingions with m/z less than 17. Major revisions toa secondresistor ladder operating inparallel with the accelerator plate ladder tosource voltage tothe injector and repeller plates,were made to reduce the overall current demand placed on the PASS opamp and 100volt DC-DC converters. It is likelythat this scanning limit is a resultof increased platevoltage needed tomatch the 20% increase in field strength of thescan magnet over previousmagnet designs. Toaccomplish current reduction and thereby enable higher platevoltages without requiring increased power, variable voltage to theinjector and repellerplates is providedby way of two parallel 100K 0 1 watt potentiometers connectedto the PA88 outputand shunted to the commonterminal via a 100KQ resistor tail. Throughthis modificationthe overallresistance of the circuit was increasedfrom 45.8K 0 to 55K Q, resultingin a 20'/0decrease in powerdemand and thereby extending the scanrange to m/z of lessthan 12.AAer completingthese modifications, accelerator plate voltageswere tunedto maximizeion peakheight while maintainingbaseline linearity. Greatestperformance was achievedby tuning the potentiometersto producethe voltagesdescribed in Table2-3, whereall potentialsare referenced to the vacuum envelope,and a voltageof -1.174volts is appliedat the input of the scancontroller.

100k PLATE4

PLATE 2

PLATE 0 S VACUUM ENVELOPE

PLATE 1

PLATE 3

PLATE 5 FOCUS!

Figure 2-11: Scan controller detail

46 cloid E-field C cloid inEd e connect in Potential volts

Repeller 10 34.5

Injector 24.36

Plate 0 0,0

Plate 1 -11.16

Plate 2 11.20

Plate 3 -22.12

Plate 4 22.09

Plate 5 -36.02

Table2-3i Cycloidelectric field potentials

2.3.5 Emission regulator Ionizationof sampleanalyte isaccoinplished bymeans of electronimpact, wherein electricityis supplied tothe filament from an electronic circuit, causing electrons tobe ejected&om the heated tungsten filament. This process accounts for the majority of the instrument'spower consumption; therefore, to improve energy efficiency over that of a simpleresistive element regulator, theNEREUS ionization circuit, developed byHeniond [38],relies upon a TL494integrated circuit, operating asa squarewave generator to drive twofield effect power transistors, which are connected in a pushpull arrangement. The currentdriven through these transistors isthen rectified and sent to a DC-DCconverter, whereits voltage is decreasedfrom 12 to approximately2 volts and current increased to approximately3 Amperes before acting upon the filament. By varying the duty cycle of thissquare wave instead of circuitresistance, losses are minimized. Regulation ofthe squarewave duty cycle is maintainedthrough a negativefeedback loop that senses electrontrap current and adjusts the duty cycle to match potentiometer specified current. Electriccurrent is deliveredto thefilament via two12 gauge wires which are affixedto square pads on the circuit board. Typically, the emission regulator draws 9 wattsof power while operating at8 kHz.A Simpsonmicro ammeter that electrically connectsthe electron trap to the emission regulator's sensory circuitry, and is physically

47 mounteddirectly abovethe emissionregulator, can be usedto visually monitortrap currentdeveloped by theionization chamber. Trap current is typicallyaround 7 Ij,A during instrumentoperation. Power and sensorylines including trap currentand on/off control logic! areinput to the emissionregulator trough a DB-25connector. Size of the emissionregulator was reduced to a NEREUScompatible footprint throughthe use of miniaturizedcomponents such as DC-DC converters and ceramic capacitors and the circuit beingetched onto a double-sidedboard measuring 3,5" x 3.5". Carewas taken in circuit boardlayout to placehigh noisecomponents such as switchingpower transistors andconverters into areasaway from sensory components and to minimizetrace lengths. The einissionregulator is switchedon andoff via software,using the computer's parallelport controlbit-3 line to direct a bipolartransistor which drivesa reedtype relay switchconnecting the TIA94 on/offline to ground.In this way,the computer can turn off theemission regulator during periods of inactivity,decreasing power consumption from approximately18 to 9 watts,This control circuit is integralto theDAQ circuitryand, due toshielding its positionforsatisfactoryabove the ion relay pumpactuation, housing and +5analyzer VFFmagnet, requires 4 B-field

CTRL 3

Figure 2-12: Software controlled emissionregulator on/off relay 2.3.6 Electrometer Theextremely low ion currents. generated by theFaraday cup requires that the electrometerproduce a largegain while keeping noise at minimallevels, Given the 16bit resolutionand 5to +5volt inputsignal range of theDAQ's analog to digitalconverter, theelectrometer should be ableto amplifythe smallest ion currentof interestto at least 153p. volts, However,the need for highgain requires the use of a feedbackresistor valuedgreater than 10 ohms,which dominates all otherelectrometer voltage noise sources.Additionally, the response time of theelectrometer is determined by theresistor- capacitortime constant of thefeedback resistor and the. stray capacitance across the resistor.This time constant determines the maximum bandwidth of theelectrometer and, consequently,can be the limiting factor for overallinstrument sampling speed i.e. when multiplemass peaks are required in a spectralscan!. Therefore, adequate instrument performancerequires an optimization of theelectrometer's amplification, noise tolerance, andbandwidth parameters, Initially,an electrometer previously developed for theHemond mass spectrometer Hemond1991! was tested for useon the NEREUS system. This circuit uses a circuit basedon a Teledyne1702 operational amplifier connected asa current-to-voltage converter,followed by a unitygain inverting buffer Figure2-13!. The 1702 op-amp generatesanextremely low bias current, which is achievedwith a parametricinput stage. Thelow noise characteristics of this electrometer permit quantization of ion current signalsassmall as 10 ' amps.However a 10 "0 feedbackresistor and 2pF of capacitance,yield a responsetime of approximatelyone half second, making electrometertime response a majorlimiting factor because successive scans across the spectralrange are required. For example, if 1,000mass steps including molecular ions and&agments! areneeded toaccurately represent mass peaks across a singlespectruxn, andmeasurement of ion currentat eachmass step requires 0.5 second, the time needed to completethe scan exceeds 8 minutes. Therefore, a new electrometer was developed that possessesthesame low noise characteristics asthe Hemond design, while increasing its effective bandwidth.

49 10

FARAOAYC tNPUT

OUTPUT

-18 +18

Figure2-13: Teledyne parametric op-amp eiectrometer design by Hemond Fordesign purposes, minimum required signal amplification can be calculated assuminga 153 @volt minimum DAQ detectable voltage 0 voltsof dynamicrange/16 bits!.Given an assumed minimum detectable ion current of 10' amperes,overall electrometergain must be at least 1.53x10 where1 ampere= 1 volt/ampere!. Based on this gain value,an appropriatefeedback resistor can be calculated.Feedback resistor selectionis limitedbecause large value resistors are produced in onlylimited sizes. Presumingthat the Johnsonnoise of the feedbackresistor will be the dominantnoise source,a noisevalue for thefeedback resistor can be easily computed as a functionof temperature the NEREUS instrument package is not intendedto operatein. water bodies thatexceed 300'K! andfrequency; where i is currentin amps,K is theBoltzmann constantof 1.38x10 Joules/Kelvin, T is temperaturein Kelvin, B is bandwidthin Hertz, and R is resistance in ohms. e' =4KTBR

Equation 2-1: Fundamental thermal noiseof feedback resistor UsingOhm's law to computethe output signal voltage for a givenresistive value, the outputsignal given a 10 Ampereinput! can then be compared against its output thermalnoise voltage to finda signal-to-noiseratio. For example, a 1010 Q resistor producesa voltagesignal of 100IjV, butgenerates Johnson noise of equalvalue S/N=l!

50 whenbandwidth is 60.4Hz andtemperature is 300 Kelvin; while a straycapacitance of lpF in parallelwith 10' Q yieldsa timeconstant of 100Hz.In thiscase the circuit would befrequency limited by theJohnson noise. Assuming that a signal-to-noiseratio of at least1 isrequired for detection of the ion current, a maximumtemperature of 300 Kelvin, a minimumion current of 10-l 4 Amperes,and a straycapacitance of at least 2pF exists in parallelwith the feedback resistor, a feedbackresistor value with optimum signal-to- noiseand frequency characteristics can be found. Figure 2-14 graphically illustrates how circuitswith feedback resistor values of lessthan approximately 5 x10 are frequency limitedby Johnsonnoise and circuits with feedback resistors greater than 1.5 x10' and capacitanceof at least2 pFare frequency limited by theirRC time constant.

N ~ S/N=i ~ S/N=2 0> C=2pF 10 ~ C=3pF CT I LL

1.00E+09 1.00&10 1.00E+11 1.00Etl2 ReSiStiVevailua Oh1TS!

Figure2-14: Optimal feedback resistor value as a functionof RCtime constant and S/N, assuminga minimum input signal of 10 Amperes,temperature of 300K, and Johnson noisedominating.

51 Havingidentified a feedbackresistor of approximately10' Q asoptimal, a glasstype 1.GSx10' 0 Victoreenresistor was chosen, based on cost and availability. Then, using thesecalculations, the activecomponents of the electrometerwere selected. Beginning with theelectrometer input, a NationalSemiconductor LMC6001 field effectoperational amplifier was chosenfor useas the first stagegain. The LMC 6001was selectedbecause of its low inputcurrent noise characteristics .13fA/Hz" !, aiidbecause its field-effect architectureminimizes input leakage.Like the previouselectrometer design, the LMC6001 is arrangedas an inverting amplifier, with its negativeinput connectedto the Faradaycup lead. This arrangement amplifies the 10 ' amperesto 105pV atthe output of the LMC6001. Previouscalculations and experimentation [39j indicatedthat the NEREUS instrumentrequires a dynamicrange of approximately2000 to properly quantify dissolvedchemicals of interestat commonambient environmental concentrations i.e. &om 10parts per billion to 2Gparts per million!. Therefore,to provide adequate amplificationto makefull useof the 10volt digital-to-analogconverter range, the output of the LMC6001 is connectedto the input of a secondstage amplifier which amplifiesthe input with a gainof -50. This secondstage employs a LinearTechnology LTC1151 operationalamplifier configuredas an inverting amplifier! that can convertand amplify a 200mVinput to a 10volt output.The LMC6001 advertiseszero drift andan input offset of 5p.V,minimizing baseline clipping and drift. Finally,the output is routedthrough a resistorcapacitor series low passfilter with its 3dB roll-off centeredat 150Hz. The 150Hzvalue was choseiiso that signalattenuation would be minimizedin the frequency region of interest <50 Hz!. A schematicof the circuit is shownin Figure 2-15. FARADAYCUP INPUT

OUTPUT

Figure 2-15: NEREUS eiectrometer circuit

An overallexpected noise budget as a functionof bandwidthwas determined for the ...electrometer, using the noise figures for the feedback resistor, the various other resistors, andfor theactive components which are listed in thedata sheets of theLMC6001 and . LTC1151.Based on thisnoise budget, a minimumion detectablecurrent as a functionof - bandwidthcan be calculated, assiuning S/N=l and T= 300'K Figure2-16!. Tominimize size, the circuit was constructed on anetched single plane PC-board, using1/8 watt resistors, DIP packages forthe LTCI 151 the LMC6001 Figure 2-17!. The Victoreenfeedback resistor was connected tothe input of theLMC6001 using an "air wiring"technique, This approach takes advantages ofthe insulative and low stray capacitiveproperties of air to minimizethe effects of strayelectronic noise, which althoughsmall, still permeatesthe etchedboard material.

53 A battery power supply consisting of two 9 volt alkaline batteries in series with a centertap, providing +9 volts, in parallel with four centertapped 1.5 volt AA alkaline batteriesproviding +3 volts to the LMC6001 suppliesto the op-ampsand voltage reference.

1.00E-14

H. vJ EX < IL

S M

1.00E-150 5 10 15 20 25 30 35 40 45 Frequency Hz!

Figure2-16: Theoretical electrometer sensitivity vs. sampling frequency assuming circuitis not limited by feedbackresistor RC time constant!

To matchthe 5 to +5 rangeof the DAQ, the 0 to 10 volt outputof, the LTC1151 outputwas offset using an Analog Devices REF195 +5 [change]volt precisionvoltage referencethat is connectedto the outputof the LTC1151via secondLTC1151 operating asa unitygain inverter. This voltage regulator addition, however, generated electronic interferencein excessof theresolution gain attained by centeringthe output range about zero.Therefore, this offset circuit wasre-designed using a highprecision AD588BQ voltagereference that resideson the DAQ board,thus preventing voltage reference inducedload fluctuations from significantly decreasing the electrometer's signal to noise ratio. A 3.7 cm length of coaxialcable is usedin parallel with the Victoreenresistor as a low value - 3 pF! capacitor.This lengthof cableis necessaryto provide feedback

54 amplifierfrequency compensation, thereby avoiding stability problems resulting in a rail- to-railoscillation at 26Hz higherharmonics also! on the electrometer's output. Through trial and error, a minimumcapacitive value was found that preventedthe oscillation.The exactcause of this instabilityis still unclear,but is mostlikely dueto theextremely high gainof theVictoreen resistor in combinationwith strayeffects near the LMC6001 leads. The resultingoverall gain andresponse time of the elcctrometersuggest that the coaxial capacitor may also function as a resistive element to the extent that it decreasesthe feedbackeffective resitance possiblya factorof 5 or more!.

Figure 2-17: NEREUS electrometer

In additionto carefulcircuit design,several other stepswere taken to minimize the effectsof externalnoise sources, A batterypack, sourcing+3 and+9 volts to the electrometerwas built to minimize 60Hz interference,which wasencountered even when usinghigh-quality regulated AC powersupplies. A secondline of defenseagainst externalnoise was to placethe electrometer in a Faradaycage and position it asclose to theFaraday cup aspossible, To accomplishthis, the electrometerwas mountedinside a smallrectangular aluminum BUD box,and positioned so that the vacuum envelope feed- throughactually penetrates the aluminum box, allowing the Faraday cup output to connectdirectly to theinput of theelectrometer's LMC6001 op-amp. Finally, the vacuum

55 envelopeand electrorneterbox were connectedto the batterypack's commonlead, to makethe electrometer resistant to damagefrom electrostatic discharge Figure 2-18!.

Figure 2-18: Klectrometershielding and mounting

A HewlettPackard digital storage oscilloscope was used to collectdata to analyzethe responsecharacteristics of this electrometer design. Analysis consisted of samplinga timeseries of approximately32seconds duration, at 1,000samples per second, resulting in a 32,768 point data series. These data serieswere then saved to disk and converted to Matlabreadable data files. The time series data sets then were transformed into frequency spectra.Once the frequencyspectrum was obtained,this datawas then analyzedfor powerspectral density. Based on time series data taken with thescope probe shorted to ground,the probe introduces approximately 1.19mV, of error to the measurement.To estimatethe noise inherent to theelectrometer circuit, its inputwas tied to ground.Two separatetime seriestaken with the input grounded,indicated that the electrometer generatesfrom 3.42mV, to 8.36mV, of noise. No significantcorrelation is evident betweenscope probe noise and electrorneter noise. However, the electrometer appears to beparticularly sensitive to low frequencymechanical vibrations. Initial spectralanalysis

56 of theelectrometer with its inputconnected to the Faraday cup revealed the 26Hz rail-to- rail oscillation as well as higher harmonics!. After the oscillatory response was attenuatedwith a properlytrimmed coaxial cable acting as a capacitivecompensator!, theelectrometer in coinbination with the Faraday cup generated approximately 9.13rnV of noise Figure 2-19!. Based on cross correlation analysis, the oscillation problemappears to eitherhave been eliminated, or itsmagnitude attenuated to somewherebelow the noise floor of theelectrometer Figure 2-20!, Close inspection of thepower spectral analysis for the compensated electrometer reveals a 60Hzpeak with a magnitudeapproximately anorder of magnitudeabove baseline. This peak is likely to havebeen caused by external power lines and fluorescent lights in thelaboratory room wherethe measurements were made. Other higher frequency peaks, between 400 and 500 Hz,are well in excessof thecritical threshold of the output's low-pass filter, suggesting thatthese frequencies arelikely to have been generated and/or experienced subsequent to the electrometeroutput.

'f0

10 I, 10 ICompensated,' eiectro eter 100 50 100150 200 250 300 350 400 450 500 Frequency Hz!

Figure 2-19: Power spectraldensity of electrometerwith and withoutcoaxial capacitive compensator

57 0,03

0.02

0.01

-0.02 0 1 2 3 4

offset X 10

Figure 2-20: Time series cross correlation of electrometer noise data with and without capacitive compensation

The 9.13mV of measuredelectrometer noise is roughly twice the estimatedvalue, thuselevating the minimum detectable ion currentto 1.75x10 ' amps.Although a minimumdetection limit of 2 x10' ampsand maximum sampling &equency of 25Hz assuming4pF of capacitance!is acceptable,post-processing techniques may allow further improvement.Detailed examination of the frequencyspectra of the electrometer indicatesthat the noisemeasured at the outputof the electrometeris generally"white", or distributeduniformly acrossthe frequencyspectrum. The randomnessof this noiseis exploitedby employingsignal averaging to reducethe magnitudeof noisewithout attenuatingthe signal;thereby increasing the signal-to-noiseratio. Applying this techniqueto the electrometerdata collected by the oscilloscope,we seea 90%error reductionwhen using 100Xsignal averaging, an 86% error reductionwith 50X signal averaging,an 82% error reductionwith 25X signalaveraging, and a 59%error reduction with 5X signalaveraging. These values appear to behavevery nearly to thatof n' /n

58 statistics,corroborating thehypothesis ofnoise randomness. If a 10X signal averaging routinewere to be implemented, a 68'/onoise reduction could be expected. Theanalog- to-digitalconverter of theDAQ system is capableof conversionrates in excessof 150,000samples per second, allowing for many samples in quick succession. Consequently,if the electrometer bandwidth is 25 Hz, the DAQ can be made to do successiveconversions, witha Imstemporal spacing, after an initial time delay of 25ms toallow electrometer output settling!. This would have an overall effect of decreasing thebandwidth toless than 25Hz, but increasing theion current sensitivity. Incomparison tothe Teledyne electrometer, theNEREUS electrometer appears tobe lessaffected byelectromagnetic disturbances below 60 Hz Figure2-21!. Given that the NEREUSelectroineter isdesigned tobe used only within a frequencybandup to approximately25Hz, and incorporates a post-amplification lowpass filter intended to attenuatefrequencies above 50Hz, this electrometer in fact possesses lower noise characteristicsthanthe Hemond design, yet requires a shorter response time, These performanceadvantages aremost likely due to the two-stage amplification process and theextremely low bias current of the6001 op-amp.

10 150 200 250 300 350 400 450 500

Figure2-21: Powerspectrum density comparison of NERKVSand Hemond- Teledyne electrometers.

59 2.4 Power system

NEREUSoperates using a +12Volt DC power supply, Electrical power can be suppliedto theinstrument either from its two internal12 volt 7 ampere-hoursealed lead- acidbatteries. connected in parallel! or via itsexternal 9-pin Impulse cable common returnat pin 1 and+12volts at pin 2!. Thecable uses a sharedpower bus with the batteries,permitting the internal batteries to berecharged when power is suppliedto the instrumentby way of thecable connection, The 12 volt power bus supplies power to 4 electricalcomponent groups: the ion pump, scan controller, enussion regulator, and digitalcircuitry. These components are connected to the+12 volt terminalby wayof a fusedmain toggle switch that then branches to individuallyfused toggle switches for eachof thefour component groups, The five toggle switches are positioned immediately forwardof thescan magnet on theframe's composite mounting plate, and electrical wires to theswitches are routed through a wiringharness with a square12 pin Amphenol connector.This connector can be detached to permiteasier access to componentshoused withinthe lower hemisphere. Connector pin arrangementis detailed in Table2-4.

unction Pin numbe Computer/DAQ 1,5 Emission regulator 2,6 Scan controller 3,7 Ion pump 4 8 Main 9, 10 No connect 11, 12

Table 2-4: Powersystem wiring harness connector

Fuseratings and approximate current draw for individualcomponent groups during operationare specified in Table2-5. Return paths for eachof thepreviously mentioned electroniccomponents are individually routed to a shortlength of 12gauge "common bus"wire which connects to thenegative battery terminals and pin 2 ofthe Impulse cable.

60 Switch Fuserating Amperes! Currentdraw Amperes! Main 1.7 Embeddedcomputer & DAQ 0.5 Emission regulator 0,7 Scan controller 0.2 Ion Pump 0.3

Table2-5: Current requirements and fuse ratings of electroniccircuits

The embeddedcomputer and DAQ rely on a seriesof dedicatedDC-DC converters andvoltage regulators to supplynecessary positive and negative voltages. A DC-DC converter Power One model DFA6U12S5! transforms the battery's unregulated+12 voltsinto+5 volt regulated output of upto 6 watts;the embedded 386 computer and DAQrequire typically require a totalof 900mAduring operation. A parallelsecond DC- DCconverter Power One model DFC10U24D12! uses the unregulated +12 volts to generateregulated 212 volts at 24 watts.A voltageregulator Film Microelectronics modelFM905S7! is connectedin seriesto the 12volt outputof thisDC-DC converter to generatea regulated 5volt source capable of sourcingup to 7.5watts at peak. Both the 212volt converter and the 5volt regulator are capable of generatingpower in excessof thedigital circuitry's respective %50mA and 20mA requirement, but were chosen based ontheir higher efficiency over other types of invertersand regulators e.g. zenier diode, bandgap, etc.!. Outputs from these converters, theregulator, and a commonreturn path arerouted to theembedded computer and DAQ system on a braidedfive-wire power harnessand connector that bifurcates into two 10base-T Ethernet jacks mounted to the instrumentkame's composite plate. Wiring at the Ethernet jacks is describedin Table 2-6.

61 10 base-T Wire ¹ Color Function Orange & white GND Orange +5 volts Green & white NC Blue +12 volts Blue & white -5 volts Green -12 volts Brown & white GND Brown +12 volts

Table 2-6: 10 base-I powerconnector wiring

Electricalpower to theelectrometer is suppliedusing a separateinternal battery pack consistingof four 1,5volt AA andtwo 9 voltalkaline primary batteries. This independent electrometerpower supply was selected to mirumizeinterference &om switching loads generatedby the other electroniccomponents.

2.5 Component housings, frame, and mounting apparatus

TheNEREUS prototype fits withina 17-inchglass Benthos pressure sphere. The upperhemisphere of thepressure sphere houses the analyzer and ancillary components includingembedded computer, DAQ board,electrometer, emission regulator board, ion pump,magnet, and high voltagepower supply. Initial designlayout of the instrument I citeCamilli 2000] was prepared using Ray Dream CAD softwareto projectcomponent shapeswithin the allotted pressure sphere volume. Shape, size, and positioning of the instrumentcomponents were then adapted based on theRay Dream renderings, enabling anapproximately 50% decrease in instrumentsize beyond initial proposalestimates whichcalled for theinstrument to occupyan entire sphere volume. Figure2-22: Layout of NEREVS components within pressure sphere

The volumemade available through this additionalminiaturization of the instrument hasallowed for spacewithin the lower hemisphere to reserved for placementof an internalbattery pack, which can be used to powerthe instrument. This battery pack is composedof two 12 volt 7 Amphour seal lead acid batteries and can easily be modified to accommodateother types batteries, including rechargeable secondary varieties such as lithiumion, nickel cadmium, and silver zinc, and non-rechargeable primary batteries, suchas the alkaline D-cell packs commonly used to power in-situ oceanographic instruments.

2.5.1 Frame NEREUScomponents are held in placeusing a compositeframe, The composite materialhas the advantages of being lightweight, non-corroding, electrically nonconductive,and non-magnetic. The supporting frame uses an inverted quadripartite geometry,similar in designto thoseseen in buildingand bridge vaults. This geometry is resistantto shockloads, permitting batteries to behoused in thelower hemisphere while securelyholding the instrument components in the upper hemisphere. Frame surfaces are constructedfrom 6 oz.medium density 90' wovenglass fiber impregnated with epoxy resin.The upper mounting ring utilizes an internal high-density styrene foam core, with eightradially embedded brass hard-points that are threaded to acceptpan head hex bolts, Theseeight hard-points serve to fastenthe instrument mounting plate to theframe. The frame,in turn,is fastenedto theglass sphere via a compositedisk that is bonded to the base.The frame is secured with stainless steel screws at six hard points in its base.

Figure 2-23: NEREUS instrument frame

2.5.2 Mounting plate Thecomposite plate is shapedwith a seriesof recessesand openings to securely positionthe analyzer magnet, ion pump housing, vacuum system valves, and 10 base-T typepower connector sockets. The ion pump, analyzer magnet, and vacuum envelope are ziptied in place to prevent dislodgment in case of roll over. Construction of the plate involvedfirst vacuum forming 10 layers of epoxyimpregnated 6 oz. medium density 90' wovenglass fiber with weave pattern offset of 45' betweenadjacent layers. A rigid

64 polyurethanecast was then made of theanalyzer magnet-vacuum system assembly, which,in turn,was used to produce a malemold. The male mold's perimeter was then tracedonto the composite plate and the internal area then cut away from the plate. Next, themale mold was placed within the cutaway region and an additional 5 layers of similar epoxyimpregnated glass fiber with 4S' offset was hot vacuum formed onto the existing 10layers. Hot vacuum forming was accomplished byplacing the composite plate within a Reynoldsoven roasting bag commonly used to roast turkeys!. The bag was then sealed andconnected toa vacuumpump to evacuate gas voids from the composite while it cured withina dryingoven set to 85'C. Finally an epoxy top and bottom coat was applied to thecomposite plate, and holes were drilled to accominodate thezip ties, thru-bolts, vacuum valves, and power connectors.

Figure2-24: Mounting plate placement ofanalyzer, detector and ion pump assemblies

2.5.3 Instrument cover Tofurther secure the analyzer components andprevent electrical short circuiting of theion pump's high voltage power supply, a speciallyshaped composite cover fits over theion pump, power supply, analyzer magnet, vacuum envelope, and electrometer, This coveris constructedof thesame epoxy impregnated glass cloth as the composite plate, utilizing a thicknessof 5 layersin most areas,and up to 12 layersin obvioushigh stress areas.Close fitting of thecover was achieved by sculptinga claymold of thepreviously mentionedcomponent assembly along with requiredhollow areas. After themockup driedto a satisfactoryhardness, it wasthen coated with Vaseline serving as a mold release!and then the composite cloth was applied and hot vacuum formed using the previously mentioned process.

Figure 2-25: Compositeinstrument cover

2.5.4 Attachment disk Theframe is attachedto thesphere via six boltsthat are fastened into to a rigidsemi- sphericalsupport disk. This support disk is constructedof epoxycoated two-part polyurethanefoam with internalcopper mesh and stainless steel wire reinforcement connectedto six threadedstainless steel hard-points. Polyurethane was chosen as foundationmaterial because of its highstrength to weightratio and its castingease.

66 Constructionof the supportdisk requiredthat the frame andinternal surface of the lower pressuresphere be covered with a moldrelease, To minimize cost, common food storage plasticwrap was used. After the release was in place,the six stainless steel hard-points constructedof a stainlesssteel nut welded to two stainlesssteel washers! were bolted to theframe. After securing the hard-points, 1 /2 inch width copper mesh was woven throughthe washers in a six-pointedstar pattern, with overlappingmesh intersections. Internalpoints of thestar with meshoverlapping between hard-point mounts were then tiedto directlyopposing mesh angles, using 16-gauge stainless steel wire. These mesh andwire reinforcement elements were then formed into a bowlshaped contour, similar to theinternal surface of thepressure sphere. Once these steps were complete, the two part polyurethanefoam was mixed in a 1to 1 ratio,to anunexpanded volume equal to 1/10of thevolume between the bottom of theSame and the pressure sphere was poured into the sphere.The frame was then mounted to a squareplywood sheet and placed concentrica11y withinthe lower pressure sphere, ensuring uniform sphere compression space and that the . topof theframe traversed the sphere equator. Once adequately cured, the support disk wasremoved and painted with a thinlayer of epoxyto provideadditional strength and . preventcatastrophic failure of thefoam. The support disk is bondedto theglass pressure spherewith RTV silicone. RTV material was selected as the bonding agent because it providesgood adhesion and is flexible,thereby allowing for spherecompression atdepth andhelping to absorbshock loads created by mechanicalimpacts.

2.5.5 Electronics housings Electronicsboards, including the emission regulator, scan controller, DAQ, and embeddedcomputer are housed within small composite cases to preventshort circuiting, protectcircuitry from water leaks, attenuate electromagnetic interference, and to organize placementof the boards within the pressure sphere. These housings are constructed of the same6 oz.woven cloth usedfor fabricationof the instrumentframe and cover. Thesecases were constructed by applyingepoxy resin impregnated glass cloth over polyurethanefoam molds sizedto maintainadequate internal space for circuitboards andconnectors!. After attainingadequate wall thickness,surfaces of the emission regulator,scan controller, and DAQ were finishedand then coveredin metal foil to

67 minimizeelectromagnetic interference. To minimizeadverse magnetic field permeation affectson an internalreed relay, the DAQ casealso utilizes two .045" thicknessmild steelplates in the areadirectly beneaththe relay. Housings are bolted together using stainless steel screws and nuts. Electronics housingscontaining the PC-104 computer, DAQ system,emission regulator, and scan controllerare fastened to theexternal surface of theinstrument cover using Velcro strips. Althoughunconventional, Velcro was chosen over other fasteners based on its light weight,magnetic inertness, and immunityto vibration.

Figure2-26: Embeddedcomputer with memorycard

2.5.6 Hardhat protector A high-densitypolyethylene "hardhat" protector is usedto protectthe NEREUS instrumentfrom impactsduring transportand tethered deployment. The lower hemisphereof the hard hat has been modified by two separate circular holes that permit theinlet system to protrudeand allow access to thesphere's vacuum port. A seriesof squarecutouts have also been made to theupper hemisphere of thehardhat, which permit viewing of the ion pump andelectron trap currentgauges while the instrumentis covered.

68 Approximateoverall neutralbuoyancy is achievedfor the instrumentand hardhat assemblythrough the use of a rigidpolyurethane foam spider that has been cast into the ribsof theupper hemisphere hardhat, along with a secondpolyurethane foam disk that is attachedto thetop of thehardhat, The placements of these floatation pieces were chosen to maximizethe separation between the system's centers of gravityand buoyancy, allowingthe instrument to remainupright when deployed on a tether.The hardhat is fastenedclosed using four eyebolts. A 'A" diameter rope harness isthen threaded through the bolt eye holes and tied to the tether,

Figure2-27: The NEREUS instrument, shown sealed within its pressure housing prior to deployment

69 2.6 !nstrument startup and sphere sealing

Oncethe componentsof the NEREUSinstrument have been properly wired and assembled,the startup process can be initialized. This involves: 1. Connectroughing pump to NEREUSvacuum system Quickflange. 2. OpenNEREUS vacuum system valve for externalpump down. 3. Vacuum pump systemto 10 Torr, 4, If vacuumenvelope operation pressure has not beenmaintained, then open vacuumenvelope valve and continuepump down process, 5. After propervacuum system pressure has been attained, turn on "main" and "ion pump" power switches. 6. If the ion pump currentmeter is stableand drawsless than 200@A, then turn on "computer","scan board" and"emission regulator" power switches. 7. Verify that the computerboots and that cycloid trap currentmeter displays between5 and10 pA. S. Verify successfulmass calibration using eitheran externalcomputer connected via radio link, or monitor connectedto the embeddedcomputer's VGA output. 9. Verify thatscan routines can be executed as directed &om remote computer. 10.Close roughing pump valve anddisconnect roughing pump 11. Verify that ion pump current remains stable.

12. Disconnect all external wires.

After the NEREUSinstrument initialization hasbeen successfully performed, the instrumentcan then be sealedthrough the following process:

1. Cleanhemisphere mating surfaces with alcoholor similar solvent. 2. Carefullyplace upper hemisphere on lower hemisphere,taking carenot to chip glass surfaces. 3. Rotateupper hemisphere so that alignmentmarks match on hemispheres 4. Remove vacuum port screw and evacuateto 0.8 ATM. 5. Quickly replacevacuum port screwand verify that the sphereholds vacuum. 6. Apply "monkey dung" asa continuousbead around the matingsurfaces.

70 7. Firmlypress and smooth monkey dung to glasssurface. S. Removemonkey dung backing tape, 9. Wrapwide plastic electrical type tape around. the equatorial region three times, beginningslightly below the equator, then transitioning to slightly above the equatoron the second turn, and completing the third wrap on the hemisphere. 10.Secure two metal band hose type! clamps around the sphere. Position clamps in a longitudinalarrangement, intersecting each other to forman "X" atthe top and bottom polesof the sphere. 11.Place sphere in KemonautAUV, or protectivepolyethylene hardhat. 12.Attach dummy plug to MII,-9 cable if externalradio transceiver isnot being used e.g. during AUV deployments!.

71 72 Chapter 3

SOFTWARE

Autonomousoperation requires that the NEREUS instrument be able to performall necessaryfunctions via its embeddedcomputer system. Minimum functionality includes poweringinstrument on and off atpredetermined intervals, mass step calibration, acceleratorpotential control, emission regulator control, data collection and handling, systemsdiagnostics, and communications with auxiliarysystems such as radio transceivers,remote viewing terminals, or vehiclenavigation computers. To satisfy these operationalneeds as well aspower andspace limitations, NEREUS uses an embedded PC-104computer possessing a 486 core module operating at 100MHZwith a standard MS-DOS6.2 operating system. The NEREUS operational code is a compiledstand-alone executablefile thatautomatically loads during the computer booting sequence. Data files andthe executable are stored on a 128MB flashdisk configured as a fastIDE drive; data filesare specified by their *.NDF nereusdata file! extension.Functionality of the instrumentis enhancedthrough a supervisorycontrol architecture that allows for human or machineinput to directNEREUS operation via one of its twoserial ports.

3.1 Embedded Code

Theembedded NEREUS software was written and compiled using Microsoft Quick Basic4.5. This language was chosen based on the Hemond research group's computer languagefamiliarity and its ability to berun both as interpreted and compiled code, a usefulfeature for developmentand debugging. Although Quick Basic is relativelylimited in its functioning,subroutines were developed to overcome these shortcomings, including utilizationof directmemory addressing for accessingparallel port registers, and iterative for-loopstructures to introducetiming delays where precision to lessthan one second is required.

73 3.1.1 Embedded display and user interface

A displayis of minimal valuebecause the embeddedcomputer system is intendedfor unmannedoperation of theNEREUS instrument while it is sealedwithin its protective housingand submerged, Therefore, only a rudimentarydisplay was developed, and is intendedto functionexclusively for computerfile managementand bench top troubleshooting.This displaygenerates a textbased VGA outputand can be connectedto a standardCRT monitor via a VGAjack locatedin theaft of thePC-104 housing. ASCII textis outputto thedisplay in a columnarformat consisting of: massscan voltage, apparentM/Z, ion signal and time. The embeddedcomputer accepts an AT type keyboard,which canbe usedto access theBIOS as well asfiles withinthe operating system, Pressing the FS key during the bootingsequence circumvents execution of theautoexec.bat and embedded code files, therebypermitting file accessfrom a DOSprompt. Alteration to extendedBIOS options is accomplishedby pressingCTRL-ALT-ESC during the boot sequence, which causes a configurablePOST screen to bedisplayed, whereupon various BIOS settings can be modifiedusing the keyboard. Table 3-1 describes the proper extended BIOS settings for NEREUS'embedded coinputer. In additionto theflash disk, the core module's floppy interfacecan be connected to a floppydrive for datatransfer when the upper pressure housingis removed.

74 onfi uration 0 tion Settin System POST Express Core Module extended BIOS Enabled AdvancedPower Mgmt BIOS Enabled Serial Port 1 Enabled Serial Port 2 Enabled Parallel Port SPP/BPP Floppy Interface Enabled IDE Interface Enabled Mono/ColorJumper Color Onboard DIP Socket Disabled Local Bus Video Display CRT& FP Flat PanelDisplay Type 8 Video State Enabled Blank POST Test Disabled Serial Boot Loader Disabled Watchdog Timer Enabled Hot Key Setup Enabled SCSI BIOS service Enabled SCSI Initiator ID SCSI Disk Retries 7 10 SCSI Disk 1 Id 0, Lun0 SCSI Disk 2 Not Active SCSI Disk 3 Not Active SCSI Disk 4 Not Active SCSI Disk 5 Not Active SCSI Disk 6 Not Active SCSI Disk 7 Not Active 1" Hard Disk Auto 2' Hard Disk Not Active 3 Hard Disk Not Active 4 Hard Disk Not Active 5 Hard Disk Not Active 6 Hard Disk Not Active 7 Hard Disk Not Active 8 Hard Disk Not Active ConsoleOutput Device Video ConsoleInput Device Keyboard

Table3-1: Embedded computer BIOS settings

75 3.1.2 tnstrument Control Theembedded computer uses a bi-directionalparallel port connection to the DAQ to maintaininstrument control. The parallel port is accessedthrough direct memory addressing DMA! to four registers: configuration, data, status and control, residing at hexadecimalmemory addresses of 77A, 378, 379 and 37A respectively. The configurationregister serves to specifydata transfer mode; in the case of NEREUS, a valueof 32 writtento the77A address sets the port to PS/2data transfer mode. Bi- directionaldata exchange takes place via the data register using a signahngpattern previouslydescribed byCamilli [cite]. Under this regime data is transmitted across an 8- bit databus using a "twopass" architecture in which binary output data is sentto the 16- bit digital-to-analogconverter asan 8-bit high-byte packet followed by an 8-bit low-byte packet.Binary 2's compliment datais read as input from the 16-bit analog-to-digital converterin a similarfashion, wherein an 8-bit high-byte is readfirst, followedby an 8- bit low-byte.To synchronize thedata transfer process and thereby avoid data packet collisions,the 8 dataregister bits are connected to a double-tieredbank of rnultiplexers onthe DAQ. Bits 0,1 and 2 ofthe parallel port's control register operate the multiplexer gates,while bit 5 dictatesdirectional control for the bi-directional dataregister and bit 3 actuatesa transistor-drivenreed relay functioning as an on/off logic switchfor the emissionregulator. Status register bits 3 through7 arereserved to detectthe logic states of variousfailsafe indicators. Bit assignmentsfor thedata, control and status registers are

detailed in Table 3-2.

76 Registerbit Function DB-25 pin Data 0 Data in/out

Data 1 Data in/out

Data 2 Data in/out

Data 3 Data in/out

Data 4 Data in/out

Data 5 Data in/out

Data 6 Data in/out

Data 7 Data in/out

Status 3 Ion pump failure 15 Status 4 Emergencyvalve 13

Status 5 High pressureindicator 12

Status 6 Low Battery 10

Status 7 Unused

Control 0 Converter select

Control 1 Byte select 14

Control 2 AD569 chipselect 16

Control 3 Emissionregulator on/off 17

Control 5 Datainput/output NC

Table 3-2 Parallelport registersand functions

3.1.3 Calibration Calibrationofmass-to-charge m/z!focus relative tothe scan controller input potentialgenerated bythe DAQ is required toaccurately scan for and identify ion peaks withinthe cycloid's spectral range. Initially a softwareapplication wasdeveloped that scannedthree separate peaks and generated a linear slope-offset equation based on &e threepeak centers Figure 3-1!. This application required human operator input from a remotelaptop computer linked by radio transceiver toselect the calibration gases and to conclude%e slope offset calculation. Extensive testing of this calibration system revealed thatnon-ideal effects, most likely a consequenceofthe scan controller's high gain

77 circuitry,causing the cycloid's electric field potentials to correspondnon-linearly with scancontroller input. Additional analysesindicated that this non-linearitydid not affect themonotonicity of therelationship but did introducea degreeof distortionlarge enough to precludeaccurate tuning of ion peakswith an m/z in excessof 30 AMU. Moreover, radio transmissioninterference occasionally would producetransmission errors that causedthe embeddedcomputer to generatecycloid plate voltagesconsiderably different from theirintended values. Although this error could be overcome either by redundant commandtransmission, or throughpost processing of signalsfor error correction,both methodscould only be implementedwith substantialtime penalties.Therefore, an alternatesystem of calibrationwas developed.

1 Li 'x

O [~gee

,',9~3,

" Nitrogen I

44- = OxygenIk

Figure 3-1: Remote calibration GUI for NEREUS instrument

78 To overcomethe drawbacksof the first generationcalibration system, all functions requiringcommunication with anexternal agent, human or computer,were eliminated in favorof a completelyautonomous system. In addition,an algorithm was developed that iterativelycompares the residuals of likelyvariable values to producea least-squares-fit exponentialequation. This second-generation calibration systein also makes use of an errorchecking procedure, which was developed to verify thatnewly generated calibration variablesare within acceptableerror margins. Failing a successfulcalibration, the instrumentundertakes a limited number of retriesbefore discontinuing the process and revertingback to themost recent valid variables stored in memory. Calibrationis initiated at eachcomputer boot-up and at hourly intervalsthereafter. Thecalibration routine involves scanning for threepeaks at mass-to-chargeratios of 16, 20 and40. These peaks were selected based on theirpresumed universal expression under all marineenvironmental conditions, expecting that an oxygen peak at m/z=16 will persisteither as an atomic fragment of wateror asdoubly charged molecular oxygen. Similarly,persistent peaks appear at m/z = 40and 20 as a resultof singlyand doubly chargedargon. Other peaks occurring at m/z= 17,18 and 28 water&agment, water and nitrogen! arealso universally present, but can generateion currentintensities that exceed the.full scaleof thedetector and in doingso obscure the peak's exact center. Signal saturationis problematic because precise center determination is crucial for generating thetwo calibration variables i.e. slope coefficient and m/z exponent! with precisions resolvedto atleast four significant digits in orderto maintainsufficient focus accuracy in thehigh m/z range of thespectrum. To ensureaccuracy, the calibration curve's RMS erroris verifiedto beless than 10 mV before the new variables are stored in memory. Oncethe calibration process is complete,the variables are used to generateappropriate scancontroller input voltages by raising the desired m/z to thepower of theexponent variable n! andthen multiplying the result with the slope variable m!, yielding the proper scan controller input voltage,

ScanVoltage = m M/Z!"

Equation3-1: Masscalibration equation

79 Thecompiled code relies on a 24-hourclock using a dualprecision floating point valuefor task scheduling. Therefore, to preventthe instrument from generating an error whenthe most recent calibration occurred before midnight and a calibrationis scheduled for aftermidnight, a timechecking algorithm was developed which accounts for this offset andadjusts scheduling accordingly,

3.1.4 Operational rhodes A suiteof embeddedbehaviors allows the NEREUS instrument to respondto input from variousexternal agents and environmentalconditions. These behaviors minimize thepower and time required to gatherrelevant data at an appropriate resolution by controllingthe sleep/wake state, type of scanundertaken, and amount of signalaveraging thatis performed.Execution times vary from seconds to tensof minutes,depending on the scanmode and signalaveraging combination that is executed. A totalof six scanmodes are available to theinstrument, four of whichare spectral sweepsranging from either 12 to 48 or 12to 150with stepintervals of 0.5or 0.1AMU. Theremaining two scanmodes make use of a peakjumping strategy, which allows the instrumentto rapidlymonitor 15 peaks by evaluatingonly the peaks regions and thereby omitting m/z spectralregions of minimal interest. Caadienttracking, one of thepeak jumping modes, permits the instrument to adjust thetime interval between successive scans according to absolutevalue of amplitude changesin the m/z peaks that are being monitored. This behavior relies on an algorithm thatcompares the peak heights of themost recent scan with the corresponding peaks of theprevious scan. If anyof thepeaks have changed amplitude by 100'loor greater,the instrumentimmediately commences a new scan. If themaximum amplitude change is greaterthan 40 lo andless than 100'lo, the scan interval is decreasedby a factorequal to thepeak height change, multiplied by thescan interval, If thelargest peak height change is between30'lo and 40'lo, the scan interval remains unchanged. If the greatest peak heightchange is lessthan 30'lo and the scan interval is lessthan three minutes, the scan intervalis increasedby a quantityof secondsequal to 2 dividedby thepeak height change,up to a totaldelay interval of 3 minutes.When the delay interval is greaterthan 30seconds, the instrument immediately cycles into a standbymode, using the computer

80 to turnoff theemission regulator for a durationten seconds less than the delay interval. Theemission regulator is thenpowered up tenseconds before the standby interval ends to allow sufficient time for electron beam stabilization,

3.1.5 Signal Averaging Whilein operationthe instrument can assume seven different signal-averaging intervals,two of whichinvolve variable signal averaging. The five fixed signal-averaging routinesare SGO, 1G00, 2000, 5000 and 10000X, thereby decreasing statistical error to 4.5%,3.2%, 2.2%, 1.4% aiid 1.0%respectively. The variable signal averaging conducts repeatedsampling until theabsolute value of theintegrated electrometer voltage relative to baselineexceeds either 50 volts or 100volts, and to avoidbaseline clipping, relies on a baselinedetermination which is set equalto 75 mV lessthan the ion currentsensed at m/z = 11.5AMV. Thisdependeiicy on the number of samplesaveraged to theintegrated voltagecauses the same relative magnitude of statisticalerror to begenerated for each datapoint i.e.signal to noiseratio!; thus avoiding excessive sampling of largesignals, or undersampling of smallersignals. To minimizethe time required for calibration,a processrequiring at least300 data points, a limited25X signal averaging routine is used yielding a statisticalerror of 20%! becausejust relative maximaare needed. Thesignal averaging process entails that the appropriate mass first betuned, and then a 20-milliseconddelay be instantiated to allow sufficient time for theelectrometer output to stabilize.After this delayperiod has ended, the DAQ's analog-to-digitalconverter repeatedlysamples the electrometer output in uninterruptedsuccession until the signa1 averagingparameters have been satisfied. Execution of eachsample cycle requires approximately30 microseconds, resulting in respectivetimes of 3S,50, 80, 170and 320 millisecondsfor the 500, 1000, 2000, 5000 and 10000X signal averaging routines. Time requirementsforthe variable signal averaging regimes cannot be predicted with complete certaintyin advanceof theirexecution because sampling times are a functionof the magnitudeof theelectrometer output voltage and various noise sources, causing higher. electrometeroutput signals to requireless sampling time and smaller signals to demand longersignal averaging. In additionto thesignal averaging time requirements, fiuther timeis allotted for data handling and response delays associated withcapacitive coupling of the cycloid E-field plates.

3.1.6 Data handling A measurementof electrometer current is madeof M/Z=11.5at thestart of eachscan. Thisvalue, which is expectedtobe representative of the sensor baseline voltage, isused tonormalize the digital-to-analog converter signal to a baselinevalue. The converter, whichis uses a +5to 5volt scale and 2s complement code to express electrometer voltage,electrometer noise generally causes the electrometer baseline value to fluctuate withina+70 to 70mV range. Therefore, tominimize data character lengths and thereby savedata storage space and speed data transfer, the electrometer voltage signal at M/Z = 11is subtractedfrom the subsequent values. Finally, a valueof 70is addedto theion signalin orderto preventthis normalized signal from dropping below zero.

3.1.7 Data storage Oncethe requisite sainpling and normalization hasbeen fulfilled, the signal data is averagedto produce a singledata point. The data point is thenstored within a datafile usingan 8-character hh-mm-ss.ndf ASCII string time-stamp name. The contents of these datafiles are structured using comma-separated variable formatting. A file header is containedwithin the first fourlines of eachdata file. Thefirst rowrecords the date of the scanusing a month-day-yearformat followed by the descriptor "NEREUS embedded data",denoting that the file was created by the embedded computer. It is worthwhile notingthat the BIOS of theinitially usedPC-104 386 core module was afflicted with the Y2Kproblem, making it necessary todescribe allembedded system dates using a single decadeoffset e.g. a datewith the year 2002 is stored as 1992!. The date problem was correctedby replacingthe 386 core module with a 486core module. Row two of the headerprovides descriptors foreach of thefour data columns and arranged in order of scanvoltage, m/z, ion signal, and time. The scan voltage recorded in additionto them'z andion signal because the scan voltage can be used in conjunctionwith the m/z and ion signaldata to retroactively identify scan controller faults. A timestamp is associatedwith eachdata point to allow the various spectra ion peaks to be temporally referenced with data&om other sensors e.g. 3-D position, conductivity, temperature, etc.!. Rows three

82 andfour of theheader are left blank.Data values are stored beginning at row five and continuing on to the end of the file. To minimize the probability of dataloss, a newdata file is createdat the start of each scan,after saving and closing the currently open data file. In addition,data files are saved andclosed when the instrument enters its sleep mode, While the instrument is in sleep mode,the embeddedcomputer continuously polls the calibrationtimer to seeif a recalibrationis scheduled for immediateaction. If a recalibrationis required, the instrumentawakens from its sleepmode to recalibrateand oncethe recalibrationhas ended,the instrument opens a newdata file usingthe current time stamp for name, then conductsa singlescan using its peak-jumpingmode. After the scan is complete,data from this scanis then savedand the datafile closedbefore the instrumentresumes its sleepstate. This process cycles continuously until thesystem is instructedto commencea scanningmission. To ensure that the instrument responds to anincoming command, a subroutinecontinuously polls the communication system to checkif aninput message has beenreceived. Upon receiving a validnew mission scan command, the system reawakens fromsleep mode the computer generates a new data file usingthe current time for namingand recommences the data collection process.

3.2 Supervisory control

In additionto autonomousoperation, NEREUS is capableof receivingand acting uponmission updates from an external agent usiiig the communications protocol describedin the previous sections. This agent may be human or anexternal computer and inputscan be routed through either of thePC-104's two RS-232C serial ports. This systemof supervisorycontrol enables the external agent to subsumptivelycommand [76] theembedded system to altermission directives in realtime by modifyingthe scan mode, signalaveraging, or enteringinto a sleepmode. TheNEREUS instrument control system can be regarded as an augmented finite state machinein thatit is a situatedrobotic sensor and its possiblemodes of operationare fmitein numberand are executed based on input signals collected by itsdata acquisition systemand f'rom external agents such as its radio transceiver or serialline input. Every degreeof instrumentfreedom isnot controlled by the embedded computer, causing the

83 control systemto functionas a non-holonomicsystem [77]; therefore,a robust communicationssystem was developed to preventthe embedded control system from entering into an indeterminate state,

3.2.1 Communications protocol To reduceinstruction set length, thereby minimizing transmission time requirements andincreasing the likelihood of successfulmessage transmission, commands are sent to theembedded control system as an encoded two letter and one numeral ASCII string Table3-3!. The string consists of a scanmode letter, followed by a signalaverage letter, thenby a singledigit number.Each letter uniquely specifies an option and a correspondinginteger value. The purpose of thenumeric third character is to preventthe instrument&om enteringinto an indeterminatestate as a resultof an erroneouscommand. Thisencoded character is usedto preventthe control system from entering a indeterminatestate by providinga validationthat the two alphabeticcharacters are accurateusing an abbreviated version of a cyclicalredundancy check, which is generated by multiplyingthe associated values of thetwo variables,followed by a modulo7 divisionof theproduct Figure 3-2 provides an example of a validcommand message string!.When a commandisreceived by theembedded system, correct framing of the characterpositions is accomplishedusing a rangeof A through6 for thefirst character,T throughZ forthe second position, and a valueof 0 through7 forthe third position character.The most recently received command satisfying this criterion undergoes the previouslydescribed modulo 7 computationand the resulting value is thencompared with the third position numeric character of the encodedcommand, If the values are identical,the message is determined valid and a confirinationresponse is sent to the remote system, followed by execution of the command.

84 Character Function Position Value 12-48AMU range,0,1 AMU step 12-48AMU range,0.5 AMU step 12-150AMU range,0.1 AMU step

D 12-150AMU range,0.5 AMU step 1st PeakJump 1st Sleep

Gradient Track 1st

Low countvariable signal averaging 2ttd 20

U High countvariable signal averaging 2nd 21

V 500 X sigil averaging 2ttd 22

1,000 X signal averaging 2nd 23 2 tld 2,000 X signal averaging 24 2ttd 5,000 X signal averaging 25

Z 10,000X signalaveraging 2ttd 26

Table 3-3 Command string functional elementsand values

CV3

12-150AMU range,0.1 AMU step 500X signalaveraging x 22! mod 7 = 3

Figure 3-2: Valid commandstring example

85 Messagestransmitted out &om the embedded computer are ASCII strings of variable lengthand can be of twotypes: data, or text, Therefore, each message class uses a unique setof startand stop characters. Data messages are composed of a "$"character start bit, followedby a stringrepresentation of the comma separated scan controller input voltage, mass-to-chargeratio, and electrometer voltage data values, followed by a "k" stop character.Text messages make use of a "¹" startcharacter, followed by the text string message,and concluded with a "." stopcharacter. Data messagesare sent to the externalagent to enablereal time data analysis and redundant data storage. To avoid transmissiontime associated delays, particularly when using the 9600 bit persecond radiotransceiver, only 0,5 m/z interval data is sent.Text messages inform the external agentof instrumentcondition states, These text messages include information regarding statesof calibration,emission regulator, mission changing, sleep mode, and time until next scansequence, and excessivepeak height warnings.

3.3 Remote Graphical User interface

A remotegraphical user interface GUI! hasbeen developed to enablea humanuser to monitorthe instrument performance, modify instrument execution procedures and interpretchemical data in realtime. This capability can significantly enhance the scientificdiscovery process by minimizingtime delays common to iterativesampling and therebyreveal data relationships while still in thefield I 78,79], TheGUI was written andcompiled into an executable file usingthe Microsoft Visual Basic 6.0 language, and it is designedto operatewith a 32bit Windowsbased PC operating systems such as MicrosoftWindows 95, 98, 2000 and XP. The software relies on a RS232serial port to asynchronouslycommunicate with theembedded NEREUS computer. TheGUI, named Thetis Spectral Viewer consists of twowindows, a pop-upwindow thatappears at startuptime, and a mainwindow. The pop-up window Figure3-3! providesthe computer user with informationabout the software and presents two fields for theuser to enterhis or herlogin name and password. The login name and password werewritten into this softwareto preventcommands from being accidentally sent to the NEREUSinstrument or unauthorizedoperation of theinstrument. The pop-up window is

86 withdrawnand main window is presentedonce the user has successfully logged in. A seriesof communicationstartup procedures are initiated when the user logs in, including, identificationof availableserial ports, designation of port to beused and designation of bit transfer rate.

Figure3-3: ThetisSpectral Viewer loginwindow

3.3.1 Main window Themain window consists of severalindependently functioning graphical elements. Acrossthe top of thewindow is a standardupper tool bar.Directly below this is a rowof' checkboxesto allow up to eightgas peaks to besimultaneously monitored and three communicationcommand buttons. Below the checkboxesarc two setsof radio buttons for selectingsignal averaging and scan type, and to theirright is a histogram.Below the histogramis a sliderbar. To theleft of thehistogram and below the radio buttons is a spectrumdisplay. At thebottom of thewindow is a statusbar indicating communication

87 status,scan type, datastate, and degreeof magnification, To the right of this statusbar is a setof checkboxes,which enable scrol.lable viewing of thehistogram using the slider bars and viewing of the data in tabular format.

Figure 3-4: Thetis Spectral Vie>vermain window

3.3.2 Command initialization Commandinitialization is carriedout by selectingthe degree of signalaveraging and type of scanand then clicking the greenScan button in the upperright handcorner of the screen.This redundantly sends the appropriate 3 characterASCII command refer to section3.2.1! fifty times for outputto the serial line. The statusbar at the bottom of the GUI windowis thenupdated to indicateif theserial port is open,which serial port it is,

subroutinecollects all subsequentcharacters into an input string until it encountersa stop characterof @ ifthe header character was ¹! or & if theheader character was $!, If the characteris ¹ anda matching@ characteris found, then the string is consideredto containa validtext message and the string is trimmedof its startand stop characters beforeappearing in the centraltext box within the statusbar at the bottomof the window. If a $header character and matching & stopcharacter is identified,then the internal cd'aeter stringis consideredto bea NEREUSdata string. If a secondstart character is encounteredbefore a matchingstop character, the input string characters are cleared and a newstring is started.If thetext message received &om NEREUS is "Awaitingnew mission",the GUI automaticallyreplies by sendingthe latest command that was selected. Whena NEREUSdata string is found,the data string undergoes a secondary parsing andconfirmation process before being stored as valid data and being incorporated into thedata display. This secondary parsing is executedby trimmingthe start and stop charactersfrom the string and then dividing the string into 4 discretestrings scan voltage,mass/charge, ionsignal, time stamp! based on comma separators within the string.The 4 stringsare then checked to confirmthat each is withinits respective valid rangeand that these strings are not identical to theprevious data string received i.e. redundantdata!. If aHof theseparameters are satisfied, the data strings are considered valid and stored in the data matrix.

3.3.4 Oata visualization TheGUI visually presents the gathered data to theuser in theforms of a histogram,a spectrumand an optional data table. The histogram, which is locatedon the right side of the window, describespeak data as a functionof time, whereinthe time domainis expressedalong the X-axis and the ion signal intensity is conveyedalong the Y-axis. The histogramfield is dividedevenly into a numberof horizontalrows equaling the total numberof gasesselected in the "GasesMonitored" checkbox field andare labeled accordingto thegas ion being monitored in thatparticular row, Vertical scaling of each rowis derived6om the maximum peak height of all monitoredion peaks in thedata range.Temporal scaling is determinedby thetab positions on theslider bar when the "scrollenabled" checkbox is selected.If this checkbox is not selected, the histogram only

90 graphsthe 100 most recent data points. Histogram graphing is accomplishedby constructingindividual arrays of M/Z datacorresponding to the gas being monitored withinthe temporal domain specified by theslider bar tabs, The search process is computationallydemanding, and therefore is automaticallyrecommenced only after the serialinput register has received 30 charactersor if theslider bar tabs are inoved or the "Gases Monitored" checkbox selections are modified.

Pigure3-7: Thetis GUI histogramfield with red temporalwindow and accompanying slider bar

TheSpectrum appearing in the lower left of thewindow is the last complete spectrum withinthe time domain specified by the slider bar tabs. The red box outline appearing withinthe histogram field corresponds tothe temporal domain of the spectrum. This time

91 domainwindow allows the user to placea particularspectrum within the contextual historyof previousdata. The spectrum makes use of anauto scaling function that scales thespectrum both vertically, as a functionof themaximum clectrometer signal observed withinthe spectrum range, and horizontally to thewidth of thedisplay field. ion signalintensity in boththe histogram and the spectrum can be re-scaled by selectinga magnificationfactor of 1 to9 withthe "magnification" combo box located in thestatus bar at thebottom of thewindow. This causes the spectrum to displaythe verticalfraction of dataequal to thelower 1/n of theion signal range. Meanwhile, the histogramdisplays the lower 1/n signal range fraction of theions monitored, and the solid greyline is replacedby a dottedred line if a particularion is off scalefor thattime period.

Figure3-8:Thetis GUI spectrumfield

92 In additionto thehistogram and spectrum displays, an optional data table is availablc for viewing.This data table appears as an overlay of thehistogram field andis selected usingthe checkbox on the right hand side of thestatus bar, The data table is continuously updatedwith new incoming data and is scrollableusing the vertical scroll bar on its right.

X 7 557' 07.3177 022620' 2936 0232.32 ' OZ35

Figure 3-9: Thetis Spectral Viewer main window with data table visible

3.3»5 Oata storage and retrieval Datastorage and retrieval is availabledirectly from the software, These tasks are accoinplishedby selectingeither the "Open" or "Save" commandswithin the "File" tnenuin the menubar. Whendata transmitted from the NEREUSinstrument to the GUI is saved,it is savedin a formatidentical to *.ndffiles generated by theNFREUS embeddedcomputer. File naming is leftup to theuser. The file headergenerated by the

93 GUI differs from the embeddedcomputer in that its first line does not include the descriptor "NEREUS, embedded,data". The OUI is capable of opening all files with the *.ndf designation. Any currently open file is automatically closed when a new file is opened.

I~INIW~QJPII}:}I5~67 ~

M "16} LWW".'" I'~»w'4"'.,'-I ~~~4 -; 7~.-,+-;*".3'6»446 6'ddl . ' ' 1~076»~ -7»64 -kd ~~LL td2~" tdi};,:.:,4i;,:%~."-:;."-8}:2, 5...." ';.7}V:-"-'@::» L ~d, 60 ..":*;.,.-76} 4. I~, '7~ .;QLL 'QSI' t ~'.~'61}3@@;.~6 L,14 66 1L@IL +~'I3}t}74 .d»' +%4Q»"+W: 756

< 062 115 75 1 4725 4 027 116 123 174025 -3992 117 115 174925 116 127 174025 -3525 119 133 17dt}25 *- 12 134 174025 396: 121 'I29 174025 122 125 1740 25 3 797 '!23 120 174025 3 767 124 111 174025 '12.5 115 1741725 3 7QI7 126 1I7 174625

Figure 3-IG: NEREUS Peak Interpreter GUI

3.4 Quantitative peak evaluation

As the NEREUS instrument has evolved, its data collection rates have significantly increasedto the point of making unaided analysis and quantification by a human impractical. During three brief NEREUS instrument deployinents in the Upper Mystic Lake and Boston Harbor over 400 discrete spectrawere generated,with each spectra composedof 1385 data points. Therefore, to automatethe process of quantifying ion peak heights and thereby significantly reduce the required time for analysis, an additional piece of software was developed. This program, named NEREUS Data Interpreter,

94 facilitatesthe assessment, compression and merging of NEREUSspectrum data, The sofbvarewas specifically developed to identify and quantify ion peaks at M/Z = 14,15, 16,17, 18, 28, 32, 40, 44, 78and 93, corresponding to nitrogen,methane, oxygen, water, argon, carbon dioxide, benzene, and toluene ions. A simplepeak apex detection algorithm [80] searchesfor majorpeaks within+0.5 M/Zof thesespecified ranges by examiningdata points adjacent to M/Z<~so that:

M/2 »< M/Z,»< M/Z,~!

Equation3-2: Peak-findingalgorithm

If a peakis identified,the program then calculates the baseline using three data points on eitherside of thepeak apex. When the peak M/Z is lessthan 20, the upper and lower base lineaverages are calculated using data points in theM/Z regions&om +0.3to +0.5M/Z, and-0.3 to -0.5M/Z. If thepeak M/Z is greaterthan 21, the upper and lower base line averagesare calculated using data points further away to compensatefor decreasedM/Z resolutionin thehigher mass ranges, in theM/Z regionsfrom +1.4to +1.6M/Z, and-1.4 to -1.6M/Z. Theseupper and lower baseline values are then averaged create an ensemble baselineaverage and is subtracted&om the peak apex to findthe net peak height, RMS noisevalues are calculated independently for theupper and lower baselines and then combinedto createa compositeRMS baseline noise value. If thenet peak height exceeds thecomposite RMS baseline noise value by at least a factorof 3,the peak is considered to bestatistically distinct &om the baseline noise and is recordedas an actual ion peak. TheGUI presents the user with a displayof thespectrum, along with red lines through eachof theidentified peaks. To the left andright below this spectrum display are two scrollabledata tables containing the raw ~.ndf data and the calculated peak height data, respectively.The calculated peak height data is organizedin a columnarformat consistingof: iontype label, apparent M/Z, ion peak height, time stamp. In addition,this datatable also calculates ion peak height ratios when available of nitrogen-to-argon, nitrogen-to-oxygen,oxygen-to-argon and methane-to-carbon dioxide. These ratios are also displayedin largetext in the areabetween the two datatables. Below the text

95 displayof the calculatedratios are two commandbuttons, "OPEN" and"SAVE". The "OPEN" button allows the userto browseand select stored *.ndf files, andthe "SAVE" commandsaves the calculated peak heights and ratios to a *.npi NEREUSPeak Interpreted!file. If a newfile nameis notspecified, the GUI will automaticallysave the *.npifile withthe name of *.ndffile thatwas used to generatethe peak height data. All *.npi filesare stored in a columnarformat of commaseparated variables, similar to the ~.ndffiles and include a fileheader specifying the *.ndf file thatwas used to generatethe data.

96 Chapter 4

AUTONOMOUS UNDERWATER

VEHICLE DESIGN

TheNEREUS instrument is designedto functionaboard an Odyssey Class autonomousunderwater vehicle, having beenoriginally proposedas payload within an SeaGrant Odyssey II or BlueflinRobotics Odyssey III typevehicle. Both the OdysseyII andIII arederived &om. the Odyssey I predecessorwhich was developed by theMIT Sea GrantAUV Laboratoryand first underwentfield trials off of New Englandin 1992.The Odysseyconcept focuses on AUV s thatare small enoughto be easilydeployed and are of low componentcost. A keyfactor among these design parameters is the use of low-cost massproduced glass flotation spheres,which commonlyfunction as floats to full ocean depthof 6,000meters. These glass spheres house vehicle electronics, sensors and power systemsin a dryone-atmosphere environment, permitting the use of a &eeflooding vehiclehull, therebyminimizing vehicle volume and displacement requirements, The OdysseyI vehicledesign which made use of 3 glasspressure spheres and a &eeflooding Qberglasshull displaced195 kg, with an overall lengthof 2.1 metersand maximum diameter of 0.59 meters [81].

97 Figure4-1: OdysseyI vehic1e courtesy of MIT SeaGrant!

4.1 Odyssey II design overview

In 1994the Odyssey II generation began deployments in the Beaufort Sea to explore mechanicsof Arctic sea ice. Like the Odyssey I, this second generation does not require anyspecial handling equipment for launchand recovery, and to functionat full ocean depthusing spherical glass instrument housings, Unlike the highly experimental Odyssey I, the OdysseyII is a smallerdisplacement vehicle, which utilizes a moldeddouble hull designcomposed of high-density polyethylene, allowing for inexpensivemass production aswell asincreased impact resistance [17]. Other significant differences are decreased vehicledisplacement, the core vehicle displacing approximately 120 kg andan increased emphasison modularity, permitting the Odyssey II vehicle reconfiguration according to missionrequirements such as substitution of sensorpayloads and rapid updating mission profile libraries storedin the vehiclesnavigation computer. These design elements assist in optimization for simultaneoususe of multiple vehicles,leading to an overall increase in synopticcoverage efficiency, energymanagement and robustness to componentfailure [82]. Over the pastdecade, this designphilosophy has proven both functionaland robust, resulting in severalhundred successful dives andthe OdysseyIIb, IIc and IId subclasses, Thesesubclasses, which were developedto satisfyspecific mission goals, retain the saine hull designand external dimensions of theOdyssey II, at anoverall length of 2.2meters and0.58 meters diameter. The first of theseadaptations, the Odyssey IIb configuration, wasequipped with updatednavigation control systems that relied upon a simplified distributedintelligence working acrossRS232 serial communication lines coupledto onboardsensors and actuators. The IIb designwas first deployedin a month-long experimentin Juneof 1996to study the dynamicsof &ontalmixing in the Haro Strait nearVancouver Island, in which five vehicleswere arrayed with variouswater quality sensors,side-scan sonar, and acoustic Doppler current profilers [83]. Followingthe successfulHara Strait experiments, the IIb designwas auginented with varioustypes of camerasand deployedin numerousexpeditions, Among others,missions to New Zealand in searchof giantsquid in theKaikoura Trench and to theLabrador Sea to study thermohalinecycling of oceanwaters [13]. In thefirst half of 2000two OdysseyIIb vehicles,named Xanthos and Borealis, were again updated with newnavigation control andsensor systems and dubbed IIc. Navigationcontrol updates included replacing the onboardMotorola 68040based computer with a newPC-104 based main vehicle computer MVC!, a highprecision long-baseline navigation system using Kalman filtering, a fluxgateinertial navigationsensor system, GPS synchronized microsecond timing,and acoustic telemetry, During the Generic Oceanographic Array Technology experiment GOATS 2000! the XanthosIIc vehicle was equippedwith an 8-element sonararray mountedin a "swordfish"configuration on the vehicle's nose,along with a Roxann bottom classification echo sounder, while the Borealis IIc vehicle carried a sub- bottomprofiler [84].Collectively, these modifications sought to permitclosed loop navigationcontrol and multi-vehicle operations in whichthe AUVs workedtogether to achievea commonsurvey mission to transecta pre-specifiedarea and measure the

99 acousticfield producedby theinsonified targets and accurately navigate by acoustic means [84]. Followingthese experiments, the Borialis vehicle was cannibalized for spareparts andthe remaining Xanthos vehicle was reconfigured by theAUV Laboratoryin Juneof 2001for Archaeologicalvideo survey work in theMediterranean Sca [85]. This new surveyconfiguration came to be known asthe OdysseyIId, and containedall of the core vehiclecomponents included in theIIc configurationas well asa ROSNavigator Video camerapositioned just forward of thevehicle thruster along with a downwardlooking illuminationsource in thenose of thevehicle cite SGwebsite!. The Odyssey IId Xanthos wasupdated again by MIT SeaGrant in 2002with a WHOImicro modem, an alkaline primarybattery replacing the silver-zinc secondary battery power supply, and installation of a navigationsoftware named Mission Oriented Operating System MOOS! [86, 87].

HOPE Quter Fainn

Cont

uster

.",:: '-' Camera Illumination Source

Figure 4-2: OdysseyIId Xanthos AUV

100 4.2 Odyssey III design

TheOdyssey III, whichwas developed by Bluefin Robotics, an MIT startup company,is basedon the OdysseyII design.In 2001 the MIT SeaGrant AUV Laboratoryacquired an OdysseyIII vehicle,named Caribou, from Bluefin Roboticsfor sonarsurvey research. Like the previous Odyssey designs, it uses internal glass pressure spheresto houseelectronics and instrumentation, and a free-floodingouter hull. In 2002 the Cariboucontrol systemwas adaptedby MIT SeaGrant to usethc MOOS architecture [87],Unlike the Odyssey I and II, theOdyssey III usesa modularhull, designed to accommodatevarious sonar, camera, and oceanographic systems in modularcenter payloadsections and relics on a singlevectored thruster, rather than a combinationof fixedaxis propeller with elevator and rudder fins. In additionto these differences, the OdysseyIII power supply is a lithiuinpolymer battery housed in a cube-shapedpressure compensatedepoxy-resin case instead of a glasspressure sphere. The Caribou is significantlylarger than previous Odyssey Generations, displacing approximately 400kg, witha lengthof 2.6 to 3,4 meters with and without central payload module, respectively!.

Figure4-3: MIT SeaGrant OdysseyIII CaribouAUV

4.3 Kemonaut development

In Augustof 2002,after having undergone successful limnologic and marine trials, theNEREUS instrument was ready for integrationinto an Odyssey class AUV. At this timeMIT SeaGrant's single remaining Odyssey II Xanthosvehicle used both of its internalpressure spheres to housevehicle power, control, and communication systems, makingit physically impossibleto accommodatethe NEREUSinstrument without

101 consolidationof the components into a singlesphere. The process of Xanthossphere consolidationwas estimated to takeseveral months, at a costof severalthousand dollars, and was not entirelyclear to work satisfactorily,On the other hand,NEREUS deploymentaboard MIT SeaGrant's Caribou vehicle was also impractical due to an insufficientunderstanding of vehicle dynamics using a centerpayload section with the vectoredthruster and MOOS architecture, and because a centerpayload section was unavailablefor NEREUS,Purchase of a payloadsection from Bluefin Roboticswas estimatedto costseveral tens of thousandsof dollars, with an estimated completion time of a few months,again making deployment aboard the Caribouvehicle out of the question. After lengthyanalysis it wasdecided that a newtype of Odysseyvehicle could be built specifically for NEREUS missionsat an economicand time cost lower thanthat of the Cariboucenter section or the Xanthosreconfiguration. Design of this new vehicle, namedKemonaut, began in lateOctober 2002 with the goal of producingan operational vehiclebefore winter weather delayed NEREUS deployment until thefollowing spring. To speedthe development process, it wasdetermined that the Kemonaut design should usean external hull, tail cone,control and power system spheres identical to theOdyssey IId [84].However, because the Kemonaut vehicle required space for threeglass pressure spheres two for standardOdyssey IId powerand controland one for the NEREUS instrument!while retaining the same external dimensions of an Odyssey II hull, a number of newdesign factors had to beaddressed, including:

increasedvehicle displacement

decreased hull stiffness decreasedvolume for corevehicle components decreased volume for floatation foam unknownvehicle center of buoyancy,center of mass,and dynamics unknownnavigation and communication system interference from the NEREUS instrument

Tofurther speed the design process, Kemonaut design plans were developed on a computerusing Solid Works CAD sofbvare,allowing analysis work to beundertaken

102 beforevehicle componentswere built and therebyminimizing the excessivetime and economiccosts of iterativephysical fabrication and construction cycles.

4.3.1 Design and construction Thedesign process began by first obtainingan Odyssey II outerhull fairingFrom MIT SeaGrant's AUV Laboratory.The fairing's diameterwas thenmeasured at 30 centimetertransect intervals along its longitudinalaxis. Thesedimensions were then usedto loft a computermodel of the externalhull shape,which servedas relative boundaryvolume for vehiclecomponents. Three 17" diameter spheres were then describedwithin this internalvolume, yielding a residualvolume for all other vehicle components.To further increaseclearance between the 3 spheres,as well as overall vehiclestability loweringthe vehicle center of mass!the glass pressure spheres were asymmetricallypositioned relative to thevehicle centerline Figure 4-4!. In this arrangementthe heaviest sphere containing the vehicle battery supply was located in the lowercentral position, the NEREUS sphere was located in theforward position, and the vehiclecontrol sphere was positioned in theaft. Thisarrangement provided several benefits,including:

Additional volume in the uppersection of the hull is madeavailable for floatation foam,~er improving stability characteristics. Spacingbetween the vehicle'snavigation and communicationsystems and the NEREUS instrument is increased,thereby minimizing electro-magnetic interferencebetween vehicle and. NEREUS systems. Distancebetween the two vehiclespheres and the tail coneis minimized,thereby minimizing requiredcable length. NEREUSsampling occurs before water is perturbedby the vehicle'smotion.

Despitethe advantages of this arrangement,the layout revealed that large sections of the OdysseyII structuralinner fairing would needto be removedin orderto accommodatethe third NEREUS sphere, dramatically compromising the strength of the innerhull. Furthermore,the additional 48 kg NEREUSsphere would require greater hull strengththan a standardtwin sphereOdyssey II layout.

103 Figure4-4iGlass pressure sphere placement within the KemonautAUV

Toprevent hogging or moresevere structural failure leading to possiblepressure spheredamage during launch and recovery, an additional hull girderwas essential for improvedhull rigidity. However, the third sphere and an additional girder significantly reduceavailable volume for floatationfoam, meanwhile requiring supplementary buoyancyto offsetthe increased displacement produced by thesenew components. Similarstrength and mass constraints had previously been confronted during developmentof the NEREUS instrument, leading to theusc of compositelaminates and foamsin its structuralelements. These materials functioned v ell in thatthey: reduced componentweight by approximately80/o, enduredhigh shockloads without failure duringinstrument deployments. In addition, these materials are electrically and magneticallynon-conductive a consideration when using fluxgate sensors!, are generally immuneto corrosivesea v ater,and can be fabricated into coinplex shapes relatively quicklyand at low cost. Based on the positive results met by thecomposite NEREUS components,a computer design of a compositelongitudinal girder was developed to fulfill the structuralrequirements of the Kemonautvehicle. The contoursof this surfboard-shapedcomponent allow it to occupythe normally empty four-inch gap, runningthe length of thehull, between the upper and lower Odyssey II innerfairings. To furtherreduce the vehicle's floatation foam requirement, this girder design utilizes marinestructural foam p = 288kg/m ! sheathedin a carbonfiber cloth p = 1,570 kg/m'!.The structural foam itself possesses flexural strength rise parallel tobeam thickness!of 4.8MPa, while the autoclave cured carbon material possesses a tensile strengthof 448 MPa. Because the NEREUS instrument isintended to operate to depthsof

104 approximately100in, the structural foam was chosen asa compromisebetween the high buoyancybut depth limited low-density foams and the enormous depth capability of less buoyantsyntactic foam. The carbon fiber is an autoclave cured bi-directional '/90 ! 33 msi2x2 twill, layered to 1.5 mm thickness, which is impregnated withan epoxy binding matrix,providing a measure ofarmor plating for the foam as well as a largeproportion of the girder's rigidity. Constructionentailed creation of a plywoodtemplate to patternthe contours onto the carboncloth and structural foam. These materials were then cut to shape using a high- speedspiral saw. After trimming,the upper and lower carbon fiber sectionswere bonded tothe central foam section and vacuum cured using a smallrotary pump and simple vacuumbag constructed of trash bin liners.Autoclave curing of thecarbon/foam structurewas not attempted, asit wouldhave been deleterious to the strength of thefoam, Externalsides of thefoam piece were also sheathed in carboncloth, but were bonded usingan epoxy containing glass microfibers to decrease brittleness and thereby minimize thelikelihood of delamination.Curing was accomplished using the previously mentioned . vacuumbag technique. Cutouts for vehicle components were then made to the girder aftercuring was completed, The resulting composite structure has a massof approximately9 kg and is calculatedto withstanddepths to 300meters without significantcompression taking place and, due to the density ofthe foam, provides approximately16kg of floatation. AAerfabricating the girder, the lower inner hull fairing was modified to accept the thirdNEREUS sphere byexcising two sphere cradles from a surplusOdyssey II inner hulland repositioning these further apart within a secondOddysey II inner hull that was trimmedto accept the transferred cradles. The third, central, sphere cradle was constructedofa non-ribbedpolyethylene Benthos "hardhat", Accurate positioning of the spherecradles was necessary because ofthe limited clearance between the spheres. Therefore,to insure proper spacing, glass hemispheres were placed in thecradles and alignedusing the composite girder before the sphere cradles were riveted into place with stainlesssteel pop rivets. A double-pointlifling harness was designed instead ofthe single-point system that OdysseyII vehiclesuse for multiplereasons. First, adequate clearance was not available

105 for a singlecentral lift pointbecause of thelocation of thecenter sphere. Second, two relatively wide-spacedlift pointswould causethe vehicle's centerof massto reside betweenthe two points, allowing the vehicle to belifted without the vehicle experiencing theuncontrolled seesawing common to thesingle-point harness. Third, by distributingthe loadingbetween two attachment points, failure was less likely to occur.After finalizing a computer design, the lift frame was fabricated of I/8"x 1" cross section annealed0316 stainlesssteel bar stockand 1" diameterstainless steel eyebolts. Placementsof thosecomponents residing outside of thespheres were decided based on orientationrequirements for properoperation and computer generated clearance estimates,As such,the sonaraltimeter was positioned vertically betweenthe forward and middlesphere on the vehicle's starboard side and a cutawaywas made in theunderside of thelower hull to exposethe altimeter's acoustic head. To helpequalize weight distribution,placement of the depth sensor mirrored the sonar altimeter along the central axis,residing on the portside between the forward and middle sphere, Careful considerationled to verticalplacement of theGPS Intelligent Buoy GIB! transponderin thenose of thevehicle, permitting the transponder head to sit proudof thehull for clearer signaltransmission. The batteryhousing supplying power to the strobebeacon was placedimmediately forward and starboard of thetail coneassembly in anotherattempt to aid balanceby offsettingthe asymmetric girder cutaway required for thetail cone assembly.The strobelight beacon,GPS antenna, and radio antennawere mounted immediatelyaft of thelift harness'rear eyebolt on top of a 3.5inch diameter cylindrical long-base-line LBL! transducerthat had been permanently affixed to theupper hemisphereof the Xanthoscontrol sphere. A modularthree-piece floatation system was designed to mountwithin the upper half of thevehicle between the middle sphere and outer hull to increasevehicle payload carryingcapacity and further improve stability. Six pieces of 3 A" thicknesspressure resistantmarine foam identicalto materialused in thecomposite girder! were cut to track the hull shapeusing dimensionsderived from clearanceestimates between the outer hull fairing,lift harness,and pressure spheres. The four forward-mostpieces were then bondedtogether and coated with anepoxy to improvedurability, whereas the remaining two foampieces were coated with epoxyand cured separately. The large 4-piece section

106 providesapproximately 1S kg of buoyancywith an additional2 kg of buoyancyavailable throughthe inclusion of eachof the small single sections,thereby permitting buoyancy adjustmentaccording to vehicle-payloadconfiguration. This modularsystem, when combinedwith the compositegirder, provides between 31 and35 kg of grossvehicle buoyancy when deployed in freshwater.

Figure4-5: Cut-awayview of Kemonaut,exposing upper floatation foam and lift harness

107 ~ Upper outerhull

Lift harness

Floatation foam

Beacon battery housing

Upper inner hull

Composite girder

Strobe beacon

Vehicle control sphere Vehicle power sphere

NEREUS instrument sphere

GiB transponder

Pressure transducer

Sonar Altimeter

Lower inner hull

Tail cone assembly Lead weight

Lower outer hull NEREUS inlet guard

Figure 4-6: Explodeddiagram of Kemonautvehicle components

108 4.3.2 Control and navigation

Vehicle control was madepossible using the OdysseyIId Xanthostail coneassembly and control sphere,Both componentsare powered using the centralbattery spherewhich housesa batterypack capableof generating800 watt-hoursof energy.The tail cone assemblyincludes a propulsionunit poweredby a 100watt PittmanELCOM series5100 brushlessthruster enclosed within a cylindricalmetal Benthospressure housing in concertwith an elevatorand rudder subassembly composed of two oil compensated Pittman3100 actuatorsindividually directedby JR Kerr motor controllersusing Hall effect feedback sensors[88]. Navigationis carriedout througha simpleopen-loop system, using a global positioningsystem and CrossBowAHRS altitude, headingand reference system composedof a 3-axisaccelerometer and magnetometer! contained within the control sphere,along with a DataSonics PSA-916 sonar altimeter and Paroscientific Digiquartz pressuresensor[88]. Data from thesesensors are suppliedto the main vehicle computer MVC!, a 300 MHz PC-104 stack, which uses MIT SeaGrant's Mission Oriented OperatingSystem MOOS! andis locatedwithin the aft control sphere.A FreeWave spreadspectrum radio modemhoused in the aft spheresupplies communication between the MVC andbase station. This basicconfiguration allows the vehicleto obtainposition fixes and missionupdates by way of the radio Cansmitterand onboardGPS when the vehicleis surfaced.While submerged, the MVC executesmission directives using dead reckoningestnnates via the AHRS, altimeterand pressure transducer. The MVC automaticallylogs time stampedMOOS dataduring both surfaceand submerged operation,permitting post-mission analysis of vehicle operationsand mme synchronized integrationof vehicle track log datawith NEREUSchemical data.

4.3.3 Assembly

After fabricationwas complete, the new 3-sphereinner structuralhull was fastenedto the outerhull using stainlesssteel screws. The tail coneassembly was then fitted to the externalhull, using a laserto align the thrusteraxis with the vehicle centerline,and held in placewith radially boredstainless steel screws. Next the sonaraltimeter and pressure transducerwere mountedin the lower hull with hoseclamps banding their housingsto

109 stainlesssteel L-brackets that are in turn boltedto the inner hull. Desiccantpacks were addedto the pressurespheres before sealing to preventtemperature-induced condensation on internalelectronic surfaces, then evacuated to approximately0.8 atmospheres and placedwithin their appropriateinner hull cradles.A small hull cutawaywas made in starboardbow to accommodatethe NEREUSinstrument's sampling inlet anda low-cost enginecoolant water intake guard was thenbolted in placeover the cutawayto act as a guardfor thesampling inlet. Electricalwiring residing outside of thepressure spheres was connectedusing waterproof wet pluggableMIL-9 andcoaxial cables. The composite girderwas then positioned on top of thelower inner hull, followedby theupper inner hull, floatationfoam and lift harness,in respectiveorder. These components were then securedin placeby eight stainlesssteel bolts, placedvertically alongthe perimeter,with thefotu centralbolts attaching to thelift harnesslegs for loadtransfer, buttressed by two pairsof boltsfore and aft to preventhull separation.The top outerhull waspositioned andsecured along its edgeusing Zeus fasteners once all componentswere in place. Finally,several small holes were drilled in thehull tutderbellyto permitadequate water drainageduring vehiclerecovery.

Figure 4-7: Positioning of the NEREUS instrument within the Kemonaut forward payload bay

110 Figure4-8: Partiallyassembled Kemouaut, showing the NKRKUSinstrument and composite girder

Figure4-9: Structural integrity testof fully ladenKemonaut vehicle

111 4.3.4 Stability and ballasting Predictionof Kemonaut'sdisplacement and ballasting requirements were made prior to constructionusing computer-generatedmodels of the assembledvehicle, The overall vehiclewas calculated to havea massof nearly185 kg anddisplace some 0.196 m' volume,thus yielding a netbuoyancy of about15 kg in freshwaterby assigning appropriatematerial density characteristics to individual Kemonaut components. Modelingand calculations also indicated that proper trim couldbe achieved by placing 1Gkg of leadweight in the aft portion of inner hull immediatelyin front of the tail cone assembly,along with approximately4 kgof leadbelow the forward sphere. Thus, producinga net1 kgpositive buoyancy in freshwaterwith a centerof masslocated approximately6 cmbelow the centerof buoyancy. Initial ballastingwas carried out on December10, 2002 in the UpperMystic Lake aAer the vehicle had been assembled. At this time the lake surface had frozen to a thicknessof severalcentimeters, requiring that a holebe madein theice for ballasting. Usingthe Mystic Boat Club's facilities, the vehicle was lowered by electrichoist into the waterto observestatic buoyancy. Lead weight was then added iteratively until thevehicle establishedapproximately 1 kgpositive buoyancy and relatively neutral trim. A total of 15.4kg of lead weightwas addedto the vehicle,with 8.8 kg two 10 lb lead dive weights!placed immediately forward of the tail cone,4.4 kg one 1Glb dive weight! betweenthe forwardand middle inner hull cradles,and two 0.9kg lb diveweights! on theport and starboard beam each. Based on thisballasting, the vehicle was calculated to needbetween 3.6 and 5.4 kg of additionalballast weigh for marinedeployment. Kemonautwas ballast tested prior to marinedeployment in BostonHarbor from a two-ton electrichoist. Basedon resultsfrom earlier&eshwater ballasting, an additional 3.8kg of weightwas pre-fitted to thevehicle while undergoing assembly in thelab. This additionalweight was distributed in two equallyweighted 1.8 kg bagsof leadshot. The first bagwas zip-tie wrappedto the 10 lb dive weight in the noseof the vehicle,and the secondweight was wired tightly to theexternal aft surfaceof thevehicle, directly below the two 10 lb dive weights,using 4 corrosivelinks. Thesecorrosive links causethe ballastweight to bejettisoned after a periodof 12hours, thereby improving the prospects of successfulrecovery in theevent of a vehiclesystem failure. Although theoretically

112 buoyantfor seawater,the vehicle proved to be correctlyballasted without any additional weight.Salinity tests confirmed the harbor water to beslightly brackish at 26-28 parts per thousand,thus helping to accountfor thedensity disparity. A small-300 cc piece of floatation foam was added inside the bow of the vehicle to achieve neutral trim.

Figure4-10: Kemonautballasting in Upper Mystic Lake

113 114 Chapter 5

NEREUS BENCH TESTING

Testingof theNEREUS instrument commenced following fabrication and assembly of theinstrument frame, emission regulator, scan controller, analyzer magnet, vacuum envelope,cycloid and ion pumpcomponents, The electrometer and computer operated scancontroller and data acquisition system were omitted during initial testing,in favorof a simple10-turn potentiometer controller and commercially available Keithley electrometer Figure 5-1!, Although these two componentsare comparatively cumbersome,in that they are large and require hand tuning, they permit straightforward assessmentof the NEREUScycloid and ion systemelements. The NEREUS instrument recordedits firstspectrum on January 8, 2002,using the previously mentioned Keithley electrometerto measureion currentand manually operated potentiometer to scanacross theinstrument's M/Z range. The initial spectrum, which took approximately 5 hours to manuallyscan, clearly indicated the presence of water, nitrogen, oxygen and argon peaks Figure 5-2!; therebydemonstrating operation of the variousfabricated NEREUS componentsand overall systemdesign.

FigureS-I: NEREUSsetup using Keithley electrorneter and l0-turn potentiometer

115 0.31

0.11

0.013.50 3.25 3.00 2.75 2.50 2.25 2.00 1.75 1.50 1.25 1.00 Scan voltage volts!

Figure5-2: First NEREUS scan completed on January 8, 2002,showing ambient air spectrumcontaining water, nitrogen, oxygen, and argonpeaks.

Scanswere carried out by varyingthe potentiometer output from 3.5to 0 volts, developingcycloid plate electrical field potentialsto focusion beams ranging from 12to 150M/Z. Table5-1! Repeated scanning revealed that capacitive coupling of theFaraday cupcaused cycloid plate voltage tuning to producea baselinetransient in theion signal output.The 180' Faraday cup detector, which extends to theconflat flange pick-ups in parallelto thecycloid Faraday cup, was shorted to thevacuum envelope, effectively creatingground plane shield to decreasethis response delay. Following this modification,the Keithley electrometer was also replaced with anelectrometer and shield designedspecifically for the NEREUS instrument see section 2.3.6!. This Faraday cup andelectrometer assembly greatly reduced the required scan time as well asnoise, while generatingadequate amplification to permit signal output to berouted to a paperchart recorder Figure 5-3!, Chartrecordings indicated that the electrometer, which used a 10'

116 Q feedbackresistor, yielded approximately 35 mV of noisewhile generating maximum peakheights in therange of 100-200mV, suggestingthat the electrorneter's coaxial compensatorattenuated its overallgain by approximatelya factorof 10. Therefore,to moreclosely match the+SV dynamic range of thedata acquisition system and to improve signal-to-noisecharacteristics, the10' 0 feedbackresistor was replaced with a 10"0 feedbackresistor, This modification generated anamplification increase of approximately 10X,while slightly improving overall signal-to-noise characteristics and maintaining an electrorneterresponse time of lessthan 30 ms.

Figure5-3: NKRKUSambient air spectrumusing on boardelectrometer with 10 0 feedback resistor

117 M/z scan V rn/z scan V m/z scan V m/z scan V rn/z scan V rn/z scan V 1 47.04 26 1.84 51 0,94 76 0.63 101 0A7 126 0.3 2 23.6'I 27 1.77 52 G.92 77 0.62 102 OA7 127 0.381 3 15.78 28 1,71 . 53 0.90 78 0.61 103 0.46 128 0,37 4 1'1.85 29 1,65 54 0.89 79 0 611 104 0.46 129 0,37 5 9A9 30 1.60 55 0.87 80 0,60 105 0.46 130 0.372 6 792 31 1.54 56 0.86 81 0.59 106 OA5 131 0.37 7 679 32 1.50 57 0.84 82 0.58 107 0,45 132 0.36 8 5,95 33 1.45 58 0.831 83 0.58 108 0,44 133 03 9 5.29 34 1.41 59 0,81 84 0.57 109 0.44 134 0.361 10 4.76 35 1.37 60 0,80 85 0.56 110 0.44 135 0.35 11 4.33 36 1.33 61 0,79 86 0.56 111 GA3 136 0.35 12 3.97 37 1.29 62 0.77 87 0.55 112 0.43 137 0.35 13 367 38 1.26 63 0,76 88 0.54 113 G.42 138 0.351 14 3.41 39 1.23 64 075 89 0.54 114 0,42 139 0.34 15 3.18 40 1.20 65 0,74 90 0.53 115 0.421 140 0.34 16 2.98 41 1.17 66 0,731 91 0.531 116 OA1 141 0.344 17 2.81 42 1.14 67 0.72 92 0.52 117 OA1 142 0.341 18 265 43 1.11 68 G,70 93 0.52 118 0,41 143 0.33 19 251 44 1.09 69 0,69 94 05$ 119 0.40 144 0.33 20 2.39 45 1.06 70 0,68 95 0.50 120 OAO 145 0.3 21 2.281 46 1.04 71 0.67 96 0.50 121 0.40 146 0.332 22 2.17 47 1.02 72 0.67 97 0.49 122 0.39 147 0.330 23 2.08 48 1.00 73 0.661 98 OA9 123 0.39 148 0,327 24 1.99 49 0.98 74 0.65 99 0.48 124 0.39 149 0.32 25 1.91 50 0.96 75 0,64 100 0.48 125 0,38 150 0.323

Table 5-1: Approximatescan input potentials negative voltage! required to focusa givenM/Z. Cycloidalscanning below M/Z=11 is not possiblebecause of physicalion focuslimitations but out-of-rangescan voltageshave beenincluded to illustrate the non-linearrelationship between M/Z focus and scanvoltage.

5.1 Dissolved gas gradient sampling

AAer successfullydemonstrating the ability to generatespectra from air and water samples,a laboratory-synthesizedgradient was created to testthe instrument's capability to accuratelyquantify dissolvedgas concentrations. This simulatedspatial gradient was createdby seriallydiluting an anoxic water sample previously collected in a BODbottle from the bottom of MishewamLake locatedin the AberjonaRiver watershed,north of Boston,Ma.!. Analysiswas performedby first exposingthe membraneinlet to the undilutedanoxic sample,followed by six serialdilutions with aeratedmini-Q filtered

118 waterto yieldprogressively decreasing anoxic water mixtures of: 96%,84%, 68%, 41%, 24%, 10%and 2%, The resulting dataclearly show simultaneousmeasurement of multiplegas species at usefuland representative environmental concentrations Figure 5-4!.The comparatively linear relationship of gasconcentration to mixingratio and correspondenceof slope trends at ionpeaks of M/Z = 15,32, and 44 with expected tendenciesof methane,oxygen, and carbondioxide suggestthat the NEREUSinstrument sensitivityis stableand ion currentincreases monotonically and fairly linearlywith respectto concentration.Although actual gas concentration was not establishedthrough independentgas analyses or meastuementwith known gasstandards, the simulated surfacewater wasnear atmospheric equilibrium, allowing for estimationof dissolved nitrogen,oxygen, argon, and carbon dioxide concentrations at the various mixing ratios.

I 40

~ -" 50 X 0 C 60 ~O 70

100 0 50 100 150 200 250 gas concentration as measured in mV

Figure 5-4: Measurementsof dissolvedgases in a synthesizedgradient using Mishawum Lake water.

119 5.2 Computerized operation and noise analysis

Followingtesting using the chart recorder and potentiometer, the data acquisition systemand computer were integrated into the instrument. The digital-to-analog converter connectionto thescan controller required an intermediate isolation amplifier to permit thecycloid's center electric field plateto maintaina "floating"ground potential. Based onelectrometer performance calculations and results from manually executed scan rate tests,the embedded computer's execution code was adjusted to employa 30ms wait periodto ensureadequate time for theelectrometer stabilization. Digital ground return pathsof theembedded PC-104 computer and DAQ circuitry were isolated from analog electronicground paths to preventhigh speed switching pulses &om affecting analog circuitstability, particularly analog-to-digital conversion, Finally, seven differing signal- averagingroutines were included as a partof thecomputerized data collection system in an effort to increasespectrum signal-to-noise ratio.

120 t.aptop computer

Figure5-5: Conceptual diagram of NEREUS control and data acquisition design Spectragenerated by thiscomputerized control and data collection system exhibited severalcharacteristic improvements over manually generated data, including: flatterbasehne, substantially higher signal-to-noise ratio, better peak shape resolution in higherm/z ranges, and the ability to rapidlytune to a precisem/z value Figure 5-6!.

2650

2150

5' E 1650 C C5 CO 0 1150

650

15010 20 30 40 50 60 70 80 90 100110 120 130 140 150 HI/z

Figure 5-6: Ambientair spectrum,performed using computer control and data acquisitionat niassstep interval = 0.1 M/Z. Thisspectrum reveals the presenceof nitrogen,oxygen, argon, water vapor, and carbondioxide.

Thesignal averaging techniques used by theembedded computer can be broadly grouped intotwo categories: fixed and variable number signal averaging. Fixed number signal averagingis capableof reducingRMS noise by asmuch as a factorof 20,but thiscomes atthe expense of instrumentspeed. The tradeoff of speedversus time requirement is illustratedin Figure5-8 wherein a 10Xincrease in samplepopulation generally results in a 3Xdecrease in RMSnoise, closely following the electroineter model predictions for noise.Variable signal averaging was developed in anattempt to decreaseinstrument samplingtime requirements by varying the number of samplingcycles according to integratedsignal, thereby generating uniform signal-to-noise ratio throughout the entire scanrange and avoiding over-sampling see section 3.1.5!.

122 300

280

240

220 E I 200 O 180

160

120

100 100 105 110 115 120 125 130 135 140 145 150

Figure 5-7: Comparisonof typicalnoise characteristics of the NEREUS instrument whenoperated using differing signal averaging regimes

54 51 48 45 42 39 y 36 E 33 30 O 27 2 24 21 I 18 15 12 93 6

0 00:00,0 00:45.0 01:30.0 02:15.0 03:00.0 03:45.0 04:30.0 05:15.0 06:00.0 06:45.0 Time mm:ss!

Figure 5-$: RMS noiseversus time required for completionof a high resolution, 1385data point,scan using a 100MHz embeddedcomputer.

123 Thesevarious signal averaging routines enable the NEREUS instrument toexchange scanspeed for appropriate levels of accuracy. Based on the results shown in Figure 5-8, a 500Xsignal averaging routine residing inthe knee of the S/N vs. time curve! appears to bemost suitable for quick, moderate resolution, spectrum examinations in that it efficientlyminimizes both the required instrument scan time and signal noise. The variablesignal averaging routines are most likely to be advantageous insituations where extremelyhigh scanning speed isneeded, whereas high count signal averaging can be usefulfor high-precision analyses such as ' N/' N, ' 0/' 0 isotopicdeterminations Figure 5-9!,

100

a 60

c 50

27.5 28 28.5 29.5 30 30.5

Figure5-9: A high-resolutionspectrumof ambient air,taken usin~ 10,000X signal averaging,detailing nitrogen gas at M/Z = 28and an apparent N- 'N isotopicpeak at M/Z=29.

5.3 Instrument sensitivity

Basedon earlier calculationsl 39],the analyzer requires a minimum gas influx rate of approximatelyofapproximately 10' gas molecules persecond, togenerate a detectable

124 Faradaycupsignal of 10 ' Amperes.Using this estimate, along with Henrys law constantsforvarious gases, membrane geometry, and permeability constants ofthese gasesacross the polyethylene membrane, minimum NEREUS detection limits of individualgasspecies were projected torange inthe tens to hundreds ofparts per billion mass&action! for most gases [39]. Although this model does not account for variables suchas temperature, hydrostatic pressure, variability inionization efficiency, or fragmentation,sensitivity estimates based on the. inlet system's physical characteristics correspondwellwith observed detection limits of atmospheric gases such as oxygen, argon, nitrogen, and carbon dioxide.

5.4 ion fragmentation

Fragmentationpeaks at M/Z = 20,17, 16 and 14 can be predicted asfunctions of their relativeintensities, normalized tothe intensity of theunfragmented molecular ion. The M/Z= 17peak ispresumed tobe a waterion fragment lacking a single hydrogen atoin Okf!,while the M/Z = 16ispresumed tobe composed ofhydrogen depleted water &agments'!, and doubly charged molecular oxygen 2 !.Fragmentation peaksat M/Z= 20and 14 are assumed tobe exclusively composed ofdoubly charged argon and nitrogen,respectively. NEREUS fragmentation data Table 5-2!, with the exception of argon,generally appear insimilar ratios to those specified inthe UNMIX algorithm [89].

M/Z ion ratio UNMIX al orithm NEREUS

Argon 20/40 13'/o 27'/0

Nitrogen 14/28 14'/0 10'/0

Water 17/18 21'/0 26e/0

Water 16/18 1 0/o 2'/0

Oxygen 16/32 9'/0 100/0

Table 5-2: Ion fragmentationratios

Oxygenioncontribution tothe M/Z = 16peak iscalculated assuming a NEREUS waterfragmentation ratio that is normalizedto the UNMIX ratio.

125 5.5 Temperature associated peak height variation

Testswere conducted to determinethe effects of membranetemperature on instrumentsensitivity. Prior modeling of inletmembrane kinetics suggested that decreasingmembrane temperature would causea simultaneousdecrease in membrane permeability,-resultingin anoverall decrease in instrument sensitivity. Testing was carriedout by conductingsequential scans while the instrument'smembrane inlet was submergedin a beakerof filteredwater. The water was first chilled to approximately1 C andair bubbleswere flushed from the surface of themembrane before scanning commenced.Scans were thenconducted in direct successionwhile the waterwarmed and thewater's temperature was recorded. The peak height data was then time correlated withtemperature measurements to produce a temperature-ionpeak data series for water, nitrogen,oxygen and argon ions. With the exception of argon,data from all gaseswithin thetemperature range of 5 to22 'C reveala generallyexponential relationship between ionsensitivity and membrane temperature. Data taken at temperatures below 5 'Cappear to divergefrom the exponential trend and appear to plateauto a constantvalue. The cause of thislow temperature trend is unclear,and maybe due to anynumber of mechanisms, including:1! the membrane material undergoing a discontinuous change in its permeability,2! membrane material being at a greatertemperature than the sample water, or otheranalyte influx mechanisms donunating such as: 3! degassingfrom internal vacuuinsystem surfaces, 4! decreasedion pump pumping efficiency, or 5! vacuum system leaks. A nearexact correlation is visible between the slopes of M/Z = 17and 18, signifying thatM/Z = 17is anion fragment OH ! ofwater vapor, In contrast,the ion peaks at M/Z = 16exhibit a slopemore closely following that of M/Z= 32,suggesting that the M/Z = 16peaks are predominantly composed of doubly charged oxygen gas 2 '!, not hydrogendepleted water ion fragments '!. Peakheights at M/Z = 40appear to exhibita trendof increasingamplitude as a functionof increasingtemperature, which is the inverseof all otherM/Z peaks. This irregularity is likelycaused by decreased ion pumpingefficiency at lower temperaturesand may be further intensifieddue to the fact thatargon, a noblegas, is relativelydifficult to sequesterwith thediode-type pump used by NEREUS see section 5.6!.

10000

1000

lO C CI N c 0

101 3 5 7 9 11 13 15 17 19 21 Temperature Cerltlgrede!

Figure5-10: Uon peakheight of dissolvedgases in atmosphericequilibrium as a function of temperature

Theoretically,the membraneperineability of thesegases should change as an exponentialfunction of temperature,where permeability equals the pre-exponential permeabilityconstant multiplied by thenegative exponent of theactivation energy of permeationdivided by theBoltzmann constant multiplied with theabsolute temperature of themembrane Equation 5-1! [90].This relationship can be augmentedwith temperaturedependent Henry's Law gas solubility constants [91] to modelexpected sensitivityvariation acrossa temperaturerange.

Equation5-1: Membrane permeability as a functionof temperature

127 10000.0

1000.0

100.0

10.01 2 3 4 5 6 7 8 9 1011 12 13 'f4 16 16 17 'f8 19 20 21 Temperature centigrade!

Figure5-11: Model of NEREUSion peakheight sensitivity as a functionof temperature i.e. with respect to membraneenergy of activationand Henry's Law solubilitiesin freshwater!. Data points are normalized toNEREUS spectrum data from filtered water at 22'C.

Modeldata derived using these parameters forwater vapor, nitrogen and oxygen demonstrateexponential trends similar to measured datatrends Figure 5-11!, Although, modelcalculations suggest a more precipitous decrease inpermeability asthe membrane'stemperature decreases than the experimental data, the slope ratios of these gases i.e. model slope ratios compared with measured slope ratios! exhibit a closer degreeof congruence.

5.5.1 Temperature dependent detection limits Thetemperature dependent permeability parameters may also be coupled to previouslymodeled expected instrument detection limits [39j to yield dissolved gas detectionlimits of theinstrument atvarying temperatures. This model, which takes into

128 accountthe membrane thickness, surface area, permeability ofthe membrane to a gas, and the equivalentpartial pressureof dissolvedgas to determinea volumetricflux rate of theanalyte gas, assumes that the analyzer requires a minimumgas influx rateof approximately10' grams of argon per second, orapproximately 10'' gas molecules per second,to generatea detectableFaraday cup signal. Using this model, minimum NEREUSdetection limits of individualgas species can be estimated. By applyingthe idealgas equation, the measured molecule flux ratecan be expressedin termsof a volumetricflux rateat standard temperature and pressure. Volumetric flux rate,Q cm' .STP/sec!,can then be used to findthe equivalent partial pressure, pj atm!,of thegas encounteredon the external surface of themembrane Equation 5-2! [92].

Equation5-2: Inlet gasflux as a functionof membranegeometry, permeability, and gas concentration.

Thepermeability coefficient, P [cm@STP em]/[cm sec atmj!, of themembrane to a givengas is calculatedas a functionof temperatureusing Equation 5-1, The membrane surfacearea, A cm'!,and thickness, l cm!, are known, and the internal partial pressure of the gas,pz atm!, is assumedto be negligible.An estimateof minimum detectable dissolvedgas concentration can be calculated from this partial pressure using Henry' s law constants. Temperaturedependent detection limits can also be estimated by simplyconsidering thesignal-to-noise ratios of ionpeaks, while taking into account gas solubility at a given temperaturebecause the conception of a gasis determinedby its Henry'slaw equilibrium with the atmosphere.Thus, minimum detectionlimit estimatesare calculated by dividingthe equilibrated gas concentrations bytheir ion signal, and then multiplying theseterms with theappropriate noise value. Based on overallNEREUS noise characteristics section 5.2!, a minimumnoise value of approximately2.5 mV can be achievedwhen signal averaging 10,000 times. Although the oxygen and mtrogen ion peakscontained in temperaturedata appear to be attenuatedby a factorof two or more

129 dissolvedoxygen and nitrogen peaks at room temperature are typically in thevicinity of 1000and 2000mV, respectively!,order of magnitudeestimates from this dataindicate thatthe NEREUS instrument is capable of detectingdissolved oxygen and nitrogen in concentrationsas low as 0.1 to 0.01 p,l/l.

5.6 Variable ion pump efficiency

Duringbench testing of theNEREUS instrument it became apparent that the pumping efficiencywas variable. Occasionally, sharp increases in ion pump current demand were observed,along with a simultaneousheightening of ion peaks at M/Z = 28,32 and 40, suggestingthat bursts of sequesteredgases &om the pump's walls were being reemitted, According to the Varian productliterature,

Whenan ion pumpis newor hasbeen regenerated, for exampleby baking,the surfacelayer of the cathodeis cleanand the gas re-emission from it is negligible.In this condition,the ion pump is called 'unsaturated'and the pumping effect is dueboth to thegettering effect as well as to ion implantationand diffusion.As the numberof gas mo1eculesimplanted into the cathodeincreases, the re-emissiondue to the ion bombardmentincreases, As a consequence,the net pumping speeddecreases until anequilibrium condition between ion implantation andgas re-emission is reached.In this condition,the ion pumpis 'saturated'and the net pumping speed due only to thegettering action of thematerial sputtered from the cathode! is abouthalf the pumping speed of theunsaturated pump... The main characteristic of noblegases is that theydo notreact with anyother element. Therefore, the film produced by thesputtering of cathodematerial does not providegetter pumping for heliumand argon. Moreover, since these gases do nottend to diffuse intothe cathode, the pumping effect due to theion implantationis not permanent.Nevertheless, all the ion pumpelements have some capacity to removethese gases, Noble gasesare pumpedby beingburied by titanium. Noble gas ions can be neutralized and scatteredfrom the

130 cathode without losing their energy. These neutral atoms maintain enoughenergy to implantor stickon theanode and on the pump walls wherethey will beburied by sputteredtitanium and thus permanently pumped. In the Diode configuration, the neutralization and back scatteringprobability is verysmall, thus the pumping speed for noble gasesis only a small percentageof the nitrogenpumping speed. Moreover,when operating at a relativelyhigh argonpartial pressure i.e., higherthan 10 mbar!,sudden bursts of pressuredue to the re- emissionof temporarilyimplanted argon in the cathodeis observed. Afterthis occurs, a Diodepump is notable to pumpmore argon until its sourceis stopped.This phenomenon is known as 'argon instability'. [93]

Thisexplanation isconsistent with the observed sporadic increases inargon ion peak height,and may also account for the similar behavior encountered with oxygen and nitrogenion peaks. It ishypothesized thata suddenre-emission burst of argon from the pumpmay trigger an avalanche of other gases from internal pump surfaces. The scenario isespecially likely if theion pump has been used for an extended period without having undergoneregenerative baking-out, thus allowing the gas load buried within the pump walls to build up.

1400 5' C 1200 g 1000 C 0 800 C 0 0 600 0 0 O gQ

200 8C 0 lO 0 0 12 17 22 27 32 37 42 47 52 57 62 67 72 77 82 87 92

Figure5-12: NEREUS ion signal variability, recorded while submerged in a Aume, basedon 9 consecutivespectrum scans executed using a peakjumping algorithm.

131 Data from 9 consecutivescans within a 7 minuteperiod during submergedtesting section5.8! of theinstrument in a laboratoryflume filled with stationarywater revealed greatestpeak height variability among M/Z = 28,followed by 32 and40. Although the instrumentwas submergedin water,peak height variability at M/Z = 18 and 17 was considerablyless than 16,and ordersof magnitudebelow that of M/Z = 28, 32 and 40, suggestingthat the deviationwas unlikely to havebeen caused be vacuumsystem leaks or variableionization, Furthermore, there was no appreciablevariability in peakheights at regionsnear M/Z = 12, 15, 78 and 91, indicatingthat electrometerdrift wasnot significant.At the time of this test the ion pumphad operatedfor more than 100hours withoutbake out, while conducting NEREUS air spectrumanalyses. After experiencing similardegassing events during the first tetheredNEREUS deployment, this ion pump was replacedprior to its secondtethered deployment and recorded comparatively negligibleargon and nitrogen signal variation during that deployment. Basedon the observed instrument behavior and ion pump literature, it appearsthat periodichigh-temperature bake out of theion pump aids not only in extendingpump life, but alsominimizes background noble gas enrichment and facilitatesstabilization of instrumentsensitivity. NEREUS ion pump rejuvenation has been successfully completed on severaloccasions by removingthe ion pump housing and heating the pump body to a temperatureof approximately125-150 'C. Oncethe ion pumpbody hasreached its prescribedtemperature, the pump-downinlet valve andvacuum envelope valves are openedand the system is evacuatedusing an oil diffusionpump in serieswith a roughing pump,Suitable results have been achieved when this pump down is performedfor a periodof at least12 hours. This process requires that the ion pump'shigh voltage power supplybe turned off, andcare must be taken not to scorchthe instrument's composite frame,nor to surpassthe nearby analyzer magnet's Curie temperature.

5.T instrument response time

Severalfactors affect NEREUS response time, including electrometer response time, vacuumsystem conductance, membrane diffusivity, and diffusivity of gasesacross the unstirredwater boundary layer residing outside of themembrane. However, prior

132 modelingsuggests that NEREUS response time is primarilygoverned by membrane diffusivity [39]. Thetime required for gasesto crossthe membrane into the vacuum system can be predictedbased on the diffusivity of a givengas within the membraneand on membrane thickness.Provided that the diffusion coefficient of thegas through the polymer is independentof concentration and constant, Fick's second law of diffusion Equation 5-3!

BC O'C ar aX'

Equation 5-3: Fick's secondlaw of diffusion

canbe applied as a onedimensional diffusive transport model to estimatethe rate of gaspermeation across the membrane as a functionof time,as the system approaches equilibrium194]. If thegas concentration onthe external surface is keptconstant and gas concentrationwithin the vacuum system is effective1ymaintained at zero,the cumulative gasflux intothe vacuum system per permeable surface area, Qcan be determined at any point in time Equation5-4!.

Equation5-4: Analytegas influx

Ast approachesoo,a linearrelation between Q, and t develops,yielding a t-axisintercept, oftenreferred to astime lag, L sec![95]. This time lag can be expressed in terms of membranethickness, X cm!,and inembrane diffusivity, D cm/sec! Equation 5-5!. L[ ] XI-]

Equation5-5: Membrane response time lag

133 Steadystate flow is underestimatedby about4% if a periodof approximatelythree times the time lag is used[96], Applying this time constantof 3L as an estimateof membraneresponse time, it is evidentthat membrane thickness strongly affects response time. Given a 1mil membranethickness, and the variousgas permeabilities of polyethylene,it is anticipatedthat the instrument is capableof accuratelyquantifying mostatmospheric gases within 10 to 15seconds. Higher molecularweight gases, includingmany common hydrocarbons, are predicted to requiretime frames on the order of minutesbefore converging upon a steadystate value Figure 5-13!.

Argon

Carbon dioxide

Carbon monoxide

Ethane

Helium

Hydrogen

Methane

Neon

Nitrogen

Oxygen

Propane

Sulfur hexaflouride

1.E+00 1.E+01 1.E+02 1.E+03 time seconds!

Figure 5-13: Calculated overall NEREUS time to steady state measurementfor various gases

Investigationsof instrumentresponse time were conducted to corroboratethe veracity of thesemodels, using carbon dioxide and benzene. These two gaseswere chosen based on theirlow backgroundconcentrations in ambient air andexpected low quantitiesof embeddedmolecules within the ion pump.

134 Carbondioxide testing required that the embedded computer be programmed to monitora singleion peak at M/Z = 44and record electrorneter voltage once per second. Thetesting was conducted by presenting the membrane inlet with an air samplewith a fixed concentrationof carbondioxide. Time was thenallowed for the carbondioxide ion signalto stabilize.After the signal had sufficiently stabilized, the sample was then rapidly evacuatedfrom the inlet using a roughingpump. It is estimatedthat the membraneinlet waspumped down to a smallfraction of 1 atmospherewithin lessthan onesecond. The datarecord shown in Figure5-14 indicates that carbon dioxide ion peak height decreased by 95%within a timeperiod of approximately15 seconds, a valueconsistent with the modelestimate of 17seconds. The data also reveals an unexpected ion peak height increaseduring the initial pump down process,The causeof this transientincrease is uncertain,but hasbeen encountered during nitrogen pump down tests using the Hemond backpackportable mass spectrometer with an identical membrane inlet apparatus Figure 5-15!, Thisphenomenon may be due to themembrane becoming unseated Rom the backingplate, thereby increasing the total permeablecross section of the membraneand causinga transientintensification of gasflux.

9500

9000

8500

g eoOO

7500

c 7000 0 5500 5000-10 10 20 30 40 50 60 70 t/me secands!

Figure 5-14: NEREUStime responseto carbondioxide

135 400

350

300

~ 250 E200 r r -' 150 1000 1020 30 40 50 60 70 '80 90 100 110 time seconds!

Figure5-15: Time responseof Hemondbackpack portable mass spectrometer to nitrogen gas

Althougha specificmodel has not been developed for NEREUStime response to benzene,tests were conducted to assess the overall instrument behavior when monitoring highermolecular weight hydrocarbons. Investigations were carried out by exposingthe inletto a near-instantaneouspulseof toluenerich air and then, monitoring the subsequent decreasein theM/Z = 78ion signal.The resulting curve show in FigureS-16 indicates thatthe instrument's overall time response to benzene,assuming a 95%reduction to baselinestandard, is on the orderof 3 minutes. This value is somewhatfaster than model estimatesof thelower molecular weight hydrocarbons such as propane, propene and ethane.It remainsunclear if thiscomparatively short response time of theNEREUS instrumentto benzeneis generallyshared with otherhydrocarbons, or if it is dueto a preferentialmembrane permeability of aromatichydrocarbons over aliphatic hydrocarbons.

136 6900

5' E 7900 Yl C CO e 06900 59000:000:050:100:150:200:25 0:300;35 0:40 0:45 0:500:55 1:00 'i;05 1:10 Time h:mm!

Figure 5-16: NEREUS time responseto benzene

5.8 Submerged testing

NEREUSunderwent submerged trial operation in a laboratoryflume prior to deployment.This trial verified that the glass pressure sphere housing could be properly sealedso that its interior remained dry, that the instrument was approximately neutrally buoyantand maintained a more or lessupright position when submerged. Rollover tests alsoconfirmed that the instrument could withstand rapid overturning without damage, Scanningoperations were carried out during these tests without any noticeable problems,

137 138 Chapter 6

DEPLOYMENT AND SEA TRIALS

Following benchtesting of the NEREUSinstrument has underwent a numberof deploymentsasa buoyedsensor, towed sensor, and aboard an AUV. Beginning in August 2002,the NEREUS instrument was deployed as a tetheredbuoy sensor in HallsBrook andBoston Harbor. The instrument was then deployed in theUpper Mystic Lake as a towedbody, followed by anAUV deploymentin Boston Harbor during December of 2002.

6.1 Tethered deployments

Tethereddeployments connnenced onAugust 20, 2002 at the Halls Brook Holding Area HBHA!,a smallbody of contaminatedwater located approximately 30km north of Boston,Massachusetts. This water body, also known as Lake Mishawum, contains high concentrationsseveral types of volatileorganic carbons VOCs! including: benzene, diphenylsulfone, ortho-hydroxybiphenyl, higher hydroxybiphenyls, and alkylated benzenesand are thought to originatefrom contaminated groundwater inflow to the hypolimnion[97], In-situ monitoringof VOCs is desirablebecause substances such as benzeneare known to rapidlyundergo biodegradation when mixed with themetalimnetic and epilimnineticwaters [98]. A verticalprofile consisting of sixspectra from 12 to 48 AMUwas completed during thistrial deployment. Each of thesescans utilized 500X signal averaging. The deploymentwas executed by loweringthe instrument with a Hydrolabsonde and an attacheddepressor weight from a ropethrough the water column to depthsof: 0.5,1, 1,5, 2, 2.5 and2.7 meters. NEREUS ion peaks were displayed in realtime via theradio transceiversand a shore-basedlaptop computer. Data from the Hydrolab sonde was storedand viewed via a cableconnection to its Surveyor4 loggerat thesurface.

139 Figure 6-1: Tethered NEREUS configuration with Hydrolab

Datacollected during from this deploymentreveal the presenceof carbondioxide and methanein thesurface water, and concentration increasing with depth,while oxygen concentrationsexhibited an inversetrend, decreasing as a functionof increaseddepth. Baselineclipping caused by highC ion fragmentpeaks at M/Z=12precludes direct measurementof the ion peakheight abovebaseline the embeddedcode normalized the baselineto M/Z=12values during this deployment!.Some ion peak tops such as methane, oxygenand argon at M/Z= 1S,32 and 40 respectively!were completely concealed by theelevated baseline, while carbon dioxide and nitrogen ion peaks at M/Z=44and 28,

140 respectively!exceed the instrument'sdynamic range at depthsof 2.Sand 2.7 meters. However,quantitative estimates of thespectra peak heights can be reconstructed by scalingthe water vapor peak at M/Z=18to a uniformvalue and then adding the spectrum'sscaling value to eachof thespectrum's peaks. The resulting values suggest factorof three increases in methane and carbon dioxide concentrations at depth. Several peaksappear in M/Z regionsthat do not correspondto dissolvedatmospheric gases, but it is unclearif these are VOC ion fragment peaks or simplyartifact. Hydrolab data reveals thepresence of a thermoclinein thedepth vicinity of 1 to2 meters,coinciding with a similardissolved oxygen trend, suggesting water stratification, Although the NEREUS spectraRMS noiseis lessthan S mV,the oxygenion peaktrend doesnot correlatewell withHydrolab data, indicating that other sources of errordominate i.e. baseline clipping!.

7000

6000

5000

E ~ 4000

CO 3000 0 2000

1000

0 12 14 16 18 20 22 24 26 28 30 32 34 36 38 40 42 44 46 48

Figure6-2: NEREUS spectrum from at 2 metersdepth in theHalls Brook Holding Area

141 0.5

~ 1.5

K CI 2

2,5

1000 2000 3000 4000 5000 6000 7000 8000 ion signai IV!

Figure 6-3: NEREUS depth profile of ion peaks in the Halls Brook Holding Area

0.5

E C15 ClKO 2.53 0

10 15 20 25 30

Figure 6-4: Hydrolab depth profile of dissolved oxygen, salinity and temperature in the Hails Brook Holding Area Followingthis preliminary Halls Brook deployment, the NEREUS instrument baselinenormalization algorithm was modified to use an ion signal at M/Z=11.5 to preventclipping and a newmembrane installed as a safetyprecaution against membrane failure,After installing the membrane andre-establishing therequired vacuum system pressure,the membrane was heated to approximately50'C with a hotair gun to encourageitsconformation tothe shape of thebacking plate Figure 6-6!. TheAugust 24, 2002 NEREUS deployment in Boston's Inner Harbor demonstrated theinstrument's depth capability, ability to operate inmarine environments, andexplored forpossible pollutants. Over the past decade significant cleanup efforts have been made toclean up the harbor, but severalpollution sources still remainsuch as combined sewer overflows CSOs!, which are known to discharge 1arge quantities ofhydrocarbons and traceorganics during wet weather I 99],The deployment area, located at North42 22.234' by%est 71 03,419', experiences a diurnal tidal cycle typically of3.5 meters amplitude andhas a mean low-water depth of approximately9 meters, causing the water to be well mixed.The tether used the same configuratiori asin the previous deployment, and scans wereconducted across a spectralrange of 12to 150AMU using a 2000Xsignal toaveraging1,5,3,4.5,routine 6, 7.5, and9, were 10.5 conductedand 12meters duringwithout high failure.tide at depth intervals««} corresponding

}}Ran 4}}}}} } «f.

}} lP

3b} $7} P+ h P 92}3}

Jd ~ 2 « 36 3I}} Figure6-5: Location of tethered NEREUS deployment in Boston Harbor

143 Figure 6-6: Microscopiccomparison of a new membrane left!, and after heat formingand deploymentto 12 metersdepth right!

Theeight spectra generated during this deployment possess similar ion peak signaturesand amplitudes, which is suggestiveof a well-mixedbody of water.All containnoise of approximately2.5 mV RMS,and have peak amplitudes that are attenuatedby around'l~, compared with laboratorygenerated data. Factors contributing to attenuationmay include diminished permeable cross section due to increasedmembrane adherenceto the backing plate!, and to a lesserextent, decreased gas solubility in seawater,as well as decreasedmembrane permeability because of lower water temperature.Typically stable ion peaks such as water vapor, nitrogen, and argon exhibit trendsof decreasingamplitude with increasingdepth, with anensemble average variation of 30/o,Models of membranepermeability as a functionof temperaturesuggest that the observed water temperaturevariation of 2.3 'C should affect an 18 ' variation in these ion peaks.The remaining variation is thoughtto bedue to pressureeffects on the membrane,particularly through increased pressure causing the membrane to sag and comeinto greater contact with the conically shaped holes of thebacking plate, thereby decreasingthe rnernbrane's total permeable surface area. Permeation may also be effected by molecular-scaledecreases in polymer permeability as a resultof pressureincreases I 100],but it is unlikelythat these molecular effects come to dominateuntil much greater depth.

144 5050

4550

3550

3050

c 2550 CO2050

1050

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

Figure6-7: Boston Harbor spectrum at 1.5meters depth

4550

4050

g 3050

e 2550 Ng 2050

1550

1050

5010 20 30 40 50 60 70 80 90100 110 120 130 140 150 MIZ

Figure6-8: Boston Harbor spectrum at 3 metersdepth

145 4550

4050

3050

2550

g 2050

1550

1050

5010 20 30 40 50 60 70 80 90 100110 120 130 140 150 M/Z

Figure 6-9: BostonHarbor spectrumat 4.5 metersdepth

4050

3550

6 3050

c 2550 PJ S 2050

1050

5010 20 30 40 50 60 70 80 90 100 110 120 130 140 150

Figure 6-10: BostonHarbor spectrumat 6 metersdepth

146 4550

4050

~p3050

a 2550 CO S 2050

1550

50 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150

Figure6-11: Boston Harbor spectrum at 7.5meters depth

3050

c 2550 EA I 2050

1550

10 20 30 40 50 60 70 80 90 100 110 120 130 140 150 IN%

Figure6-12: Boston Harbor spectrum at 9 metersdepth

147 4550

3550

E 3050

c 2550

i 2050

1550

'1050

10 20 30 40' 50 60 70 80 90 100 110 120 130 f40 150

Figure6-13: Boston Harbor spectrum at 10.5meters depth

4550

4050

3550

p 3050

c 2550 CO m 2050

1550

1050

50 10 20 30 40 50 60 70 80 90 100 1 10 120 130 140 150

Figure6-14: Boston Harbor spectrum at 12meters depth

148 Methane,carbon dioxide and oxygenion peakdata exhibit obvioustrends in the 3 to 4.5meter depth range and then again at 12meters depth, Given a meanhigh-water depth of 13meters and the Harbor's soft mud bottom, the increase in thesegases at 12meters is aptto bethe result of theinstrument being near or partially buried within the uppermost portionsof anoxicsediments. Simultaneous decrease in oxygensignals recorded at this depthby boththe Hydrolab and NEREUS support an anoxic bottom sediment hypothesis. Thecausation of themethane and carbon dioxide ion signal increase in theupper depth rangeof 3 to5 metersis lessclear because they do not coincide with proportional decreasesin oxygen or anyobservable deviation in salinityor temperature.A plausible explanationmay be thatthis region of thewater column contains effluent from nearby shipbilges and holding tanks or drainageculverts lining the Hoosac seawall these drainageculverts lie at approximately3 metersdepth during high tide!, but more investigationis neededto clarifythe origin of thischemical feature,

1.5

4.5

X 6 6

7.5

10.5

121.0 10.0 100.0 1000.0 10000.0

Figure 6-15: 9epth profile of BostonHarbor .

149 1200.0

1100.0

~1000.0

c~gooo

800.0 II 700.0

600.0 z 500.0

400.05.0 5.2 5,4 5.6 5.8 6.0 6.2 6.4 6.6 6.8 7,0 HydroLab dissolved oxygen mg/l!

Figure 6-16: Oxygensignal correlation of NEREUS and Hydrolabdata from Boston Harbor

NEREUSoxygen ion peakdata at M/Z=32 exhibit a highdegree of correlationwith Hydrolabdissolved oxygen data, matching all trendslopes and yielding a statisticalR value of 0.93. The minimumNEREUS detection limit for oxygenis estimatedto be approximately4.62 mg/1,by applying a linearleast squares line fit to this dataand assuminga characteristicnoise of 2.5 mV and a signal-tonoise ratio of 3, This detection limit is roughlyan order of magnitudehigher than bench top estimates.However, given the overall peakheight attenuation to approximately1/5 of full scale,the instrument detectionlimits canbe decreasedto lessthan 1 mg/1by simply increasingelectrometer gainor permittingmore analyte influx i,e. allowinggreater permeable membrane surface area!.

6.2 Towed deployment

After successfullycompleting buoyed deployment, the NEREUSinstrument underwenttowed deploymentin the UpperMystic Lake on December13 and 15,2002. The objectivesof thesedeployments were to ascertaininstrument performance when used

150 ina towbody configuration andto develop a detailed three-dimensional chemicalmap of this ice-covered lake. Theupper mystic lake is an eutrophic, dimictic, kettlehole lake with a surfacearea of approximately50hectares, volume of7,000,000 in and average depth of 15m [101]. Thelake sediment iscontaminated withtoxic metals, including: arsenic, cadmium, chromium,copper, and lead from industrial chemical andtannery waste release during theearly 1900s [102]. Bathymetric features include a centralbasin with a maximum depthof25 meters, and a shoalinthe northern third of the lake with a maximumdepth of 9 meters[103]. Prior investigations haveobserved a seasonal thermocline which appears inearly spring, causing ananoxic hypolirnnion todevelop ata depthof8 to10 meters, andthen gradually deepen over the course of fall.Observations indicate that the bottom 9 metersremain stratified into the winter. Oxygen depletion from biological respiration and redoxcycling of metals causes the hypolimnion tobecome anoxic by early summer [104]. Thefirst deployment onDecember 13was conducted by"mowing thelawn" along transectsmostly within the oxic epilimnion, while the second December 15deployment focusedmore on carrying out transects across the epilimnetic/hypolimnetic interface. Bothdeployments madeuse of a 14'aluminum boat equipped with a 5hp outboard motorand the tether apparatus totow the instrument through the water. Boat velocity was keptto approximately 0.5meters per second inorder to minimize oscillatory motion of theinstrument andprevent thetether &om seizing inthe ice layer which ranged in thickness6.om '/4" to 1", A soundinglinerigged with a 25lb, weight was suspended in &ontof and to a depth2 meters below the towed instrument package, with the aim of contactingthelake bottom or submerged debris before the instrument package. The December13deployment used the external radio transceiver andf1oat, whereas the December15deployment required that the radio transceiver beinte~ted inside the pressuresphere and a dummyplug for the MIL-9 electrical feed through so that the instrumentcould be sent to a depthof 25meters the MIL-9 radio transceiver cable is only15 meters long!, Total duration forboth deployments wasapproximately 6 'l~hours. Inthis time period more than 300 spectrum scans were conducted mass range 12-150 AMU,at 0.1 AMU intervals and 500 to 1000Xsignal averaging!. Over 2/3 of these

151 spectramaintain sufficiently high signal-to-noiseratios to containmeaningful data. A major noisesource appears to havebeen caused by microphonicvibration of the electrometerand faraday cup detector when the instrument tether strummed while being towed throughthe water andlarge sheets of brokenice, Nonetheless,the volumeand quality of datacollected during these brief deploymentsis unparalleled.

-71. 148

73.149

,$59 ~

42. 71,151 42.432 42,431

LatitUde

Figure6-17: 3-D mapof towedNEREUS deployments in Upper Mystic Lake

Hydrolab,conductivity and temperature data clearly show the presence two stratified layers,interfacing at a depthof approxiinately13meters. Hydrolab malfunctionin, due to battery failure in the cold winter temperatures,allowed only 54 of the NEREUS spectrato be comparedwith Hydrolabdata. NEREUS and Hydrolab dissolved oxygen datatrends generally parallel each other and correlation comparison reveal statistical R 2 valuesof 0.83 Dec. 13! and0.84 Dec. 15! that are similar to the BostonHarbor tethered deployment.The decreased correlation between oxygen data may be caused by a spatial aliasingof upto 15meters, providing sufficient time for oneinstrument to sampleanoxic waterin thehypolimnion and the other to sampleoxic epilimnion water. This aliasing is duethe boat's velocity of 0.5m/s and asynchronous sampling as much as 30 seconds apart.Hydrolab dissolved oxygen data collected on December13 exhibit relatively high variability at one andfive meterdepth transects Figure 6-20!. It is unknownwhether this variabilityis dueto dissolvedoxygen heterogeneity at thesampled depth transects or

152 Hydrolabsensor drift. However, it is worthwhile noting that Hydrolab dissolved oxygen variabilitywithin the epilimnion isless pronounced within the December 15data Figure 6-21!. Regressionbased extrapolation of December13 data indicates that the NEREUS minimumdetection limit foroxygen is in therange of 3 mg/1;while December 15 data suggestsa detectionlimit of 0 mg/1when ion sign@is below75 mV, Carbondioxide peaks are clearly identified throughout the entire water column, and increasesignificantly asa functionofdepth below the thermocline Figure 6-20, Figure 6-21!.Although a methanesignal at M/Z=15 is absent in theepilimnion, peaks begin to emergeat the thermocline and demonstrate anincreasing trend with depth, similar to that of carbondioxide, In contrast,NEREUS determinations of oxygen concentration remain relativelystable within the epilimnion, and decrease substantially within the hypolimnion.These chemical features are consistent with biologically induced anoxia and methanogenesiswithin the hypolimnion.

300

5' E 200

o 0 cv 150

5 100 XIIl/ 10 12 14 HydrolabdissoNed oxygen nag/I!

Fignre6-18: Oxygen signal correlation ofNKRKUS and Hydrolab data from Upper Mystic Lake, Dec. 13

153 350

300 5' C ~ 250

c 200 0 cv g 150 l%?g 100 12

14 Hydrolsb dissolved oxygen rngll!

Figure 6-19: Oxygensignal correlationof NEREUS and Hydrolab data from Upper Mystic Lake, Dec. 15

The low water temperaturesof 2 to 4' C decreasedthe instrument's inlet membrane permeabilityappreciably, attenuating peak heights by roughly a factorof 3. Furthermore, temperaturevariations are likely to causeincreased permeability and thereforeincrease ion signalstrength! in the highertemperature hypolimnion over that of the epilimnion. Ion peakamplitude differences across the two samplingdays are most pronounced when comparingepilimnetic oxygen concentrations. Possible reasons for this may includeincreased photosynthetic productivity from one day to the next althoughthis is unlikely becauseof the continuouscloud coverand ice coverduring thesedates!, influx of new watersfrom upstreamsources also unlikely dueto a forebayresidence time on the orderof 10days [105]!, or most likely becauseof sensordrift. Sensordrift may be due to any of a numberof factors,including variability in electrometergain, variability in ionizationefficiency, increased baseline signal i.e. throughdecreased pumping efficiency of ion pump and gasbuildup!, The relativestability of carbondioxide, methane,and hypolimnetic oxygen signals suggest that electrometergain and ionization efficiency remainlargely unchangedfrom oneday to the next,

154 CL II

C O +Is c

4I CL

CI I5 IO

O Ef5CL O 6 OI

X IO

Nl R ILl K

Figure6-20I Chemical profile of UpperMystic Lake, December 13, 2002

1/5 Figure6-21: Chemicalprofile of Upper MysticLake, December15, 2002

156 6.3 AUV deployment

On December19 2002the Kemonautvehicle and NEREUSpayload underwent sea trials in BostonHarbor demonstratingthe cargocarrying capacity,maneuverability, and overall seaworthinessof the Kemonautdesign as well as the abi1ityof the NEREUS instrumentto collectmeaningful chemical data while operating onboard a movingAUV platform. A relatively confinedregion of Boston'sInner Harbornear to the CharlesRiver Lockswas selected for operationsbecause of'its unobstructedflat bathymetry,relatively low frequencyof boattraffic duringthe winter months compared to otherregions of the harbor!,and absenceof undergroundtunnels or electricalcables thereby avoiding potential electromagneticinterference to the vehicle navigationsystem,

Figure6-22: The KemonautAUV prior to deploymentin BostonHarbor

157 Operationsbegan with the deploymentof four GIB buoyswith attachedhydrophones aroundthe perimeterof the test area,These buoys were positioned in more or lessa rectangularpattern to maximizethe triangulationdistance while minimizing propagation path obstructionsand interference with shippingtraffic, Buoy rodeswere anchored in placeusing a 5 kg weight attachedto 2 metersof 3/8" steelchain to preventdrift in the channel's +1 kt current and 3.5 meter tidal cycle. Boston Harbor's Pier 4 served as the operations base station for radio communication with Kemonautand crew aboarda skiff that servedas an interceptboat. Operations base stationwas setupin the cargobed of a Rydertruck, parkedat the endof the pier to provide a clear view westward out over the Harbor. Kemonaut was ballast tested for marine deployment from a two-ton electric hoist once GIB transponderswere in position and functioning. Based on results from earlier &eshwaterballasting, an additional3.8 kg of weight hadbeen fitted to the vehiclewhile undergoingassembly in the lab, This additionalweight was distributedin two equally weighted 1.8kg bagsof leadshot, The first bagwas zip-tie wrappedto the internal 10 lb dive weight in the noseof the vehicle, andthe secondweight waswired tightly to the externalaft surfaceof the vehicle,directly below the two 10 lb dive weights,using 4 corrosivelinks. Thesecorrosive links causethe ballastweight to bejettisoned after a period of 12 hours,thereby improving the prospectsof successfulrecovery in the eventof a vehiclesystem failure. Althoughtheoretically buoyant for seawater,the vehicleproved to be correctlyballasted without any additionalweight. Salinity testsconfirmed the harborwater to be slightly brackishat 26-28 partsper thousand,thus helping to account for the density disparity. A small -300 cc piece of floatation foam was added inside the bow of the vehicle to achieve neutral trim. Kemonautand NEREUS clock synchronizationand systems checks were performed subsequentto ballasting.After satisfyingall checks,the AUV wasreleased from its lift hoist and towedinto positionusing a 16ft skiff equippedwith a 35hp outboardengine. Oncein position,mission commands were relayed to Kemonautfrom the basestation using the spreadspectrum radio modem and operationscommenced once the vehicle was in position.

158 I I '2M/ I 41 42 39 3~g~ 41 Oft IV 4 0 R5M 40 40

41 40 43

42 38 37' 364 42

A,~i 39 i 37 36 ~ <~8 36 38 l~ qP 37 40 38 41 c> 3/ 'l 42 42 34 38 42 36 36 i 37 42

Figure 6-23:nauticalChart of BostonHarbor showingGIB placementand overall area of operation

Twoinitial sin face missions were conducted to verify proper vehicle tracking with theGIB hydrophone array, while maintaining visual contact. Following the two surface runs,Kemonaut performed a totalof 6 submergedmissions, ranging in durationfrom 2 to 6 minutesand reachinga maximumdepth of 6.3 meters.Each dive missionwas conductedas level, straightline, run for a specifiedtime durationwith the aim of reducingmission complexity and keeping altitude and heading reference systein AHRS! errorfrom critically affecting the navigation system. The navigation systein was set to abortits missionif the vehicle's sonaraltimeter sensed a distanceof lessthan 3 ineters fromthe bottom to preventthe vehicle from colliding with theharbor bottom, The interceptboat tracked the submerged AUV using GIB hydrophone position fixes of the vehiclerelayed verbally over VHF radio from the basestation. After eachdive the interceptboat was sent to visuallyinspect Kemonaut for obvioussigns of damageand to preventother nearbyvessels from colliding with the somewhatconcealed AUV. In

159 addition to visual inspection after surfacing, NEREUS instrument status was evaluated via a dedicatedradio link from a laptopcomputer located in the interceptboat. As a measureof caution,missions were conducted at progressivelydeeper target depths of 2, 4 and 6 meters.Mission durationwas alsoincreased incrementaHy from 2 minuteson first dive to a maximum of 8 minutes on the sixth dive.

4 CiK4l 6000

Time seconds!

Figure 6-24: Kemonaut pressuretransducer record

After satisfactorycompletion of its first andsecond dive to respectivedepths of 2 and 4 meters,Kemonaut was requestedto dive to a depthof 6 meterson its third mission and reacheda maximumdepth of 6.3 metersbefore resurfacing. At the time it was unclearif the bottom avoidance override had been triggered by an acoustic return &om an obstructionlying on the harborbottom. Although later analysisof sonaraltimeter data revealedthat this had not beenthe case,in the interestof vehicle safety,a decisionwas madeat that time to conductthe next dive at a depthof 4 meters,but with an increased missionduration of 5 minutes,Kemonaut executed this dive missioncleanly by quickly rotatingitself nearly 180' to assumethe properheading and then maintainingits straight

160 coursefor theremainder of themission, logging some 400 meters in distance.The fifth divemission was setup to better ascertain therepeatability ofthe vehicle's navigation and controlsystem by retracingthis fourth dive, Consequently, Kemonaut was towed back to thevicinity of its previous starting position and instructed todive to a depthof 5 meters fora durationof six minutes using the same heading asdive 4. Again, Kemonaut performedcapably, following to within1 meterof itsprevious track line for a distanceof over200 meters. The sixth and final dive mission was intended to furthertest the AUV's abilityto maintaina straightand level track line and to allowthe NEREUS instrument adequatetime to collect multiple chemical spectra while at depth. Mission duration was specifiedfor 8 minutesand a shallowdepth of 3 meterswas chosen in orderto minimize thechances ofbottom collision. The desired heading was increased ina similarattempt to preventthe AUV frombeing run aground on the ruins near the harbor channel's southern seawall and Coast Guard piers. Despite these efforts, Kemonaut was forced to abortthe divemission approximately 4 minutes intothe run because thehuman operator assigned headinghadinadvertently directed thevehicle toswim in a northwesterlytrack,passing underneaththeUSS Cassin Young and a bargebefore swimming into a shallowdry dock area,causing the bottom avoidance override procedure totrigger, thus forcing the vehicle to surface. Basedon recorded MOOS pressure sensor data, Kemonaut has demonstrated a dive ratecapability ofup to 0.18m/s,and is able to maintain a prescribed depth to within +0.3 meters Figure 7!. Despiteuncertainties caused by variabilityin near-surfaceacoustic signalpropagation, GIB hydrophone tracking revealed that Kemonaut was able to maintaina highdegree of lateralaccuracy Figure 6-25, Figure 6-26!. The extent of Kemonaut'sdynamic control isevident when comparing theGIB track logs from dive missionsfour and five, particularly inlight of the fact that these dives were performed whilein thepresence ofa currentwith a flowrate exceeding 1.5 kts.

161 -300

-400

-100 -200 Longitude rreters} Figure 6-25: 2-D GIB hydrophonesurface track log of Kemouautdive missions

l62 ia<~

Figure 6-26: 3-dimensional GIB hy dr rophonetrack logof Kememonant t dived missions

The NEREUS instru'nstrument,nstru, which ' oroperated for the e duration of its time aboardth e K emonaut vehicle , collectedapproximately 90 in' d' ivi 'd uals p ectraove r the courseof these threehours, s, wiwith eaceach spectrums consisting o,of 693 00" 00 discrete measurements. The NEREUUS instrumentappeared too bee lar arge ely unaunaffected by vehiclemotion o ionthrou ou th wateror vibrationfrom the e vehve ic 1 e's ' propulsion ' n and control n o o systemwhile underway. The NERE US instrument conducted spectrals scans of volatile dissolved'ssove gases asesratiratiging ' in molecularweight from 12 to 150Da aaltons p while o eratin ' g onboardo Kemonaut. These ans were pe armedd atata massstep interval of 0,1 Daltoa tonto ensurethat ion peakshape

163 couldbe clearly resolved. Signal averaging was kept to 500samples per data point iii an effortto minimizethe amount of timerequired for a completespectral scan while still maintaimng an adequateS/N ratio, Automated radio communications between the instrumentand its remote computer terminal used for monitoringthe instrument! were scheduledto occurduring the electrometer stabilization period before the instrument took a measurementto further decrease time requirements. NEREUS was able to completean entirescan once every 98 seconds using these techniques. Although spectral scans can be conductedmuch faster by further limiting the ma.ss-to-charge range, mass step interval andsignal averaging, this mode was chosen so that high molecular weight hydrocarbons couldbe clearly identified in theevent of beingencountered by theAUV. Post-deploymentanalysis of NEREUSdata revealed the presence of dissolved atmosphericgases, principally nitrogen, oxygen, and argon Figure6-27!. Due to the cold temperaturesencountered on this day an air temperatureof -5 to -1' C anda water temperatureof 2 to4' C! theinstrument's inlet membrane permeability was significantly decreased,causing all spectrato exhibitattenuated peak heights, Nevertheless, relative ratiosof thesegases remained consistent with previous NEREUS data collected during buoyeddeployment in this area of BostonHarbor. RMS noise was typically about 5 mV, butexhibited sudden dramatic increases, to hundreds of millivolts,appearing to coincide whenthe intercept boat occasionally bumped into the AUV whileit wasbeing tawed. NEREUSspectrum data was processed upon completion of thesix divemissions to quantifyion peaks and then merged with MOOS track log data by wayof timestamp synchronization.The resulting data shows a noticeablevariability in dissolvedgas concentrationnear the water's surface Figure 6-28!. This anomaly is likely to bethe resultof air bubbleentrainment lrom breaking waves generated by the +20kt wind blowingthat day. Further analysis of thedata did not reveal the presence of any hydrocarbon fractions or xenobiotic molecules.

164 2100

E 1600

1100 O

600

1001020 30 40 50 60 70 80 90 100 110 120 130 140 150 M/2

Figure6-27: A typicalNEREUS spectrum that wasrecorded while onboard the KemonautAUV, showing dissolved nitrogen, oxygen and argon.

tP E K3 D

1000 2000 3000 4000 5000 ion Signal mV!

Figure6-28: NEREUS record of dissolvedgas ion signalduring Kemonaut deploymentin BostonHarbor

165 166 Chapter 7

CONCLUSIONS AND FUTURE WORK

Severalcapabilities of theNEREUS mass spectrometer and Kemonaut AUV were demonstratedthrough bench testing and deployments. These deployments rigorously testedthe system's ability to not only survivein real-worldenvironinents but to collect meaningfuldata while operating in harsh environments such as an ice-covered lake, or a busymarine port. NEREUS met or exceeded allof the original design goals, including: size,power, weight, response time, depth capability, sensitivity, endurance and reliability.The Kemonaut vehicle has proven to be a small,easily deployable, low cost, yetrobust platform with adequate stability and cargo carrying capacity todeploy the NEREUSinstrument. Thus far, all tests and operations have been successfully completed withoutinstrument orvehicle failure. Moreover, large quantities ofscientifically valuable datawere collected during these initial deployments. At present, the instrument and vehiclesystem iswell suited for scientific assessment ofdissolved gases in shallow, &eshwaterandcoastal marine environments. Further research and development may greatlyextend system usefulness byvirtue of increaseddepth limit, greater detector sensitivity,elevated scan speed, multiple sensor integration, increased vehicle endurance andclosed-loop navigation. It is worthwhilenoting that the NEREUS instrument was constructedwith a materialscost of approximately$5,000, and $3,000 for theKemonaut vehicle.These budgetary constraints required that all hardware components becompleted withjust a singledesign and build iteration. Nearly all NEREUS components were fabricatedby hand and, in severalcases, component parts were fashioned &om discarded materialsfound around the MIT campus. Kemonaut development was similarly challengingbecause it required in-house fabrication and was completed, &om concept to vehicledeployment, injust six weeks. Despite these budgetary and time constraints, the Kemonaut-NEREUSsystem is a technologicallyadvanced environmental data collection toolthat offers new and unique capabilities forunderstanding thesubsurface aquatic environment.

167 7.1 Toward routine operation

Kemonautvehicle characteristics allow rapid, low-cost access to thesubsurface; meanwhile,the NEREUS instrument is well suitedfor identifyingchemical structures withinthis environment. Together, this platform and sensor permits straightforward, automatedmonitoring of dissolvedgases such as nitrogen,carbon dioxide, methane and oxygen.Better understanding of roleand variability of thesegases in biogeochemical cyclesare critical for insightinto ecological, atmospheric, geologic and hydrologic dynamics.Several deployment missions are currently possible for theKemonaut- NEREUSsystem in its currentconfiguration, including detailed survey of smalllakes, synopticassessment of largerwater bodies and calibration of chemicalsensor networks. TheKemonaut-NEREUS system is probablybest suited for monitoringdissolved gasesin lakeswith simple bowl-shaped bathymetry, such as the Upper Mystic Lake. This typeof environmentis lesschallenging to navigationfor a numberof reasons.First, currentsare generally below a few centimetersper secondand the lake volume is relativelysmall, enabling full coverageby a singlevehicle within a shortperiod of time, Second,collision obstacles such as large debris resting proud of thebottom and deep draft vesselsare not commonlypresent. Third it is unusualfor lake bottom debrisand geologyto generatemagnetic signatures of sufficient intensity to causesignificant course deviations.Furthermore, the relatively short vehicle track lines do not permit large spatial deviations.Fourth, acoustic noise and reflections i.e, from corrugated metal seawall pilings!are uncommon in such lakes, minimizing acoustic navigation error. Finally, althoughthe corrosive drop weights are unlikely to functionreliably in anoxicfreshwater environments,thelake's small and closed volume facilitates recovery in theevent of vehiclefailure. Limnologic deployments may involve the vehicle swimming vertical sinusoidsor maintainingfixed depth along a gridpattern to determinevariations in stratificationdepth i.e. whenduring a periodof internalseiching! or localizedchemical concentrationheterogeneity. Morechallenging &eshwater deployments may include deployment in ice-covered Antarcticlakes. Many of theselakes are super-saturated with respect to nitrogen,oxygen, argon,and carbon dioxide, but the contributions of their biological and non-biological sourcesare not well understood I 106]. Nitrogen, oxygen and argon gases together have

168 beenused to determined thefraction of gasproduced bybiology compared tofreezing andalso determine the nature of thefreezing process [107]. However, research indicates thatsignificant variability in gasconcentrations occur throughout the water column and samplingislimited by the time required todrill multiple boreholes inice up to 5 meters thickas well as the abilities of scubadivers to collect subsurface samples [108]. NEREUSdeployments inthe Upper Mystic Lake has demonstrated itsability to operate withina frozenlake, and subsequent Boston Harbor deployment established that the instrumentand AUV both operate satisfactorily in water temperatures lessthan 5 C. Thus,the Kemonaut-NEREUS system should be able to withstand the harsh Antarctic environment,andif equippedwith a LBLnavigation system may prove valuable not onlyin increasingthe data collection rate, but also by eliminating the need for scuba diversto performdangerous sample collection work below the thick ice cover. Yetanother type of near-termdeployment can be monitoring of coastalmarine areas forpollution &om sewage orhydrocarbon spills. The instrument package could potentiallybe used in areassuch as Boston Harbor, where recent cleanup efforts have leadto the creation ofa newoffshore outfall and the closing of sewageoutfalls within the harbor[109]. The instrument andvehicle could easily identify areas ofdecreased oxygen andincreased methane andcarbon dioxide, which may help in identifying remaining pollutionsources tothe Harbor. This type of mission would only require simple straight- linesurveys and therefore could be conducted using Kemonaut's existing navigation system.

7.2 System optirnizatlon

Althoughthe instrument and vehicle met or exceededtheir design goals, a fewareas ofpotential improvement wereidentified, both in the instrument andvehicle, following thedeployment tests. Each improvement iseasily achieved with simple modifications to hardwarecomponents oralgorithms. Improvements toNEREUS may include a modified inletsystem, designed without recesses thattrap gas bubbles within the inlet housing and a membranemounting apparatus that permits easy membrane replacement andincreased operationaldepth, The cycloid tube possesses a 180' sector analyzer and Faraday cup for M/Z= 1to 12 scanning. Thissecondary analyzer, which isnecessary formeasuring

169 hydrogengas and helium, can be utilized if a secondelectrometer and switching circuitry is incorporated.Redesign of thecircuit to incorporatean "auto-zeroing" or "chopper- stabilized" operationalamplifier [110] in its first gain stagemay further increase electrometersignal-to-noise ratio by limiting electrometerbandwidth to a narrowrange around30Hz [1 1 j.1 Thedual electrometer configuration should also include a systemfor scalableelectrometer gain permittinggreater dynamic ion peakmeasurement range! and additionalvibration isolation to dampvehicle or tow bodyvibration!. Tow body vibration canalso be decreasedby enclosingthe NEREUSinstrument within a streamlinedprotective fairing and by usinga shortshock line e.g.bungee cord! connectedbetween the towline and the instrumentto decouplesurface craft motion and towline strummingfrom the instrument.Signal-to-noise characteristics of NEREUS data maybenefit from additional post-processing techniques such as median filtering when the systemis exposedto excessivevibration or impacts.A medianfilter for electrometerdata canpotentially help reduce noise by automaticallyrejecting sample outliers and thereby preventinglarge oscillatory signals from influencing data values i.e. averaging only a subsetof samplemeasurements that arecentered about the median!.Both the NEREUS instrumentand Kemonaut AUV canbenefit from high energy-density storage batteries suchas rechargeablelithium polymer or lithium ion batteries.Further advance of navigationand signal processing software will improveaccuracy and expand system capabilities.This may include the development of a closedloop navigation system augmentedwith navigationsensors such as a longbase line LBL! array,an active hydrophoneand GIB transponderonboard the vehicle,a Dopplervelocity log DVL!, or a high-precisionspread-spectrum acoustic network such as the EXACT system, Finally, incorporationof an acoustictelemetry system will allow both Kemonautand NEREUS to communicatewith externalagents while submerged,permitting its use asa nexusof chemicaldata, linking nearby fixed arrays such as sensor buoys that may be distributed throughout a body of water.

7.3 Sensor integration and reai-time data interpretation

In-situsubmerged deployment presents the challenges of monitoringlarge spatial volumesand highly dynamic environmental states when a prioriknowledge is limited.

170 Thesefactors, when combinedwith the limited communicationbandwidth available to a maneuveringunderwater platform, requirethat autonomousdecision and control processesbe employed.The NEREUS systemuses an autonomousarchitecture to both controlthe instrument and interpret data in realtime. However, data generated by other sensors such as temperatLue, depth, height above bottom, salinity and various chemical parameters!remain external to theNEREUS decision control process. Instrument- platformcontrol and data collection can be greatly enhanced through sensor integration andreal-time data assimilation, thereby improving state-based decision processes and real-timeforecasting of environmentalconditions [79, 112]. For example, uniform temporalor spatialchemical sampling may be inefficientif a chemicalfeature is localizedor intermittent.If a uniformcoarse sampling strategy isused, the phenomenon maybecome aliased or be overlooked completely. Then again, a fine-scalesampling strategymay over sample, leading to monitoringinefficiency. Currently,the NEREUS instrument uses an embedded type of expertsystem intelligencethat allows the instrimMnt to identify individual ion peaks and vary its samplinginterval in real-time,according to therate of changeof ionsignatures. Its capabilitytointelligently monitor multiple chemicals may be broadened through integrationof other sensors onboard the Kernonaut AUV or with a distributedarray of sensorsusing acoustic or radiotelemetry. Such in-situ data distribution is advantageous becauseit cooperatively permits rapid, automated procedures such as in-situ calibration, dataassimilation, forecasting, state-based monitoring, and chemotactic navigation Camilliand Heinond 2002!. Several sensor networks, including the LEO 15 [113] site off NewJersey's coast, and the proposed Neptune array [114] in thePacific Ocean can potentiallyuse the Kemonaut-NEREUS system to rapidlymeasure multiple ions across widespatial domains, providing synoptic coverage of large-scaletransient chemical phenomena,andto supplement data collected by other moored, drifting, ship-borne and airborne instruments.

7.4 Future prototypes

Asnew materials and technologies become available, more capable prototypes can be derived&om the Kemonaut-NEREUS system's design. It is difficult to accurately predict

171 what typesof innovationsare ahead, but thesefuture generationsof autonomous submersiblemass spectrometers and AUVs may include many of thepromising technologiesthat are just nowemerging. These may include new power systems such as fuel cells,which enable month-long, multi-&ousand kilometer mission durations [115, 116].Other AUV advancesmay include advanced vehicle control systems that permit vehicle hovering,and precise navigation even in energeticenvironments such as surf zones[117, 118].Control systemevolution may alsopermit more flexible missions throughfeature-based concurrent mapping, which incorporateenvironmental measurementsin navigationestimation to improvenavigational performance in unmappedenvironments [119]. Theseadaptive control systemsmay assimilatechemical datain real-timefrom an onboardmass spectrometer, making possible chemotactic navigation.In additionto higher-levelinstrument behaviors, future submersiblemass spectrometerswill be much smallerdue to high-energyproduct rare-earth magnets and metalswith increasedB-field permeability [30]. Micro electro-mechanicalsystems MEMS!will, undoubtedly,also cause overall sensor size to shrinkconsiderably, reducingthe vacuum requirements and thereby also decreasing the instrument's power requirement.It is conceivablethat advances in MEMSand integrated circuit technologies will soonlead to thedevelopment of sub-centimeter-scaleion path length analyzers operatingat pressuresof a few Torr, permittingthe useof much smaller,lower cost AUVs,At somepoint these mass spectrometers should diminish in size,power, and cost to a pointthat they can be carriedas payload on miniatureautonomous vehicles which ran besimultaneously deployed as a self-configuringdistributed network by a single personto coveran entirevolume, Such systems will revolutionizethe way we observe theaquatic environment, enabling an in-situ presence to exploreand understand chemical processesat appropriatespatial and temporal frequencies.

172 REFERENCES

Cohen,J.E., et al., Estimates ofcoastal populations. Science, 1997. 278341!: p. 1211-1212. Aristotle,The Complete Works of Aristotle the Revised Oxford Translation, in Problems Physika Problemata!, J.Barnes, Editor. 1984, Princeton University: Princeton.p. 1514, Edgerton,H.E., Diving Deep on the Bathyscaphe, in MIT Institute Archives and SpecialCollections: Edgerton Collections. 1988: Cambridge. Ballard,R.D., The History of WoodsHole's Deep Submergence Program, in 50 YearsofOcean Discovery, O,S.B.N.R. Council, Editor. 2000, National Academy Press:Washington, D.C. p, 67-84. Fricke,J,R., Down to the Sea in Robots, in Technology Review. 1994. p. 46-55. Brooks,R.A., I leshand Machines How Robots Will Change Us. 2002, New York: Pantheon. 260. Mindell,D.A., Between Human and Machine I'eedback, Control, and Computing beforeCybernetics. 2002, Baltimore: Johns Hopkins, 439. Walter,W.G., An Imitation of Life. Scientific American, 1950. 182!; p. 42-45. Walter,W.G., A Machinethat Learns. Scientific American, 1953. 185!: p. 60- 63. Brooks,R.A., A Robotthat Walks; Emergent Behaviors from a CarefullyEvolved Network.1989, Massachusetts Institute ofTechnology Artificial Intelligence Laboratory;Cambridge. p, 12. Bernard,D., et al. Spacecraft autonomy flight experience: TheDSI Remote Agent Experiment.in Proceedingsof theAIAA Conference1999. 1999. Bellingham',J.G., et al. Performance Characteristics ofthe Odyssey AUV. in 8th InternationalSymposium onUnmanned Untethered Submersible Technology. 1993.Portsmouth NH.: Autonomous Undersea Systems Institute. Bellingham,J.G., New oceanographic uses of Autonomous Underwater Vehicles. MarineTechnology Society Journal, 1997. 31!; p. 34-47, Singh,H., D. Yoerger, and A. Bradley. Issues inAUV Design and Deployment for OceanographicResearch. in Proceedingsof theI 997IEEZ International ConferenceonRobotics and Automation, 1997. Albuquerque: IEEE. Purcell,M., et al.New Capabilities of theREMUS Autonomous Underwater Vehicle.in Oceans2000. 2000, Providence, Ri. Butler,B. andM.R. Black. The Theseus A UV, Two Successful Missions. in Proceedingsof the Tenth International Symposium on UnmannedUntethered SubmersibleTechnology. 1997. Lee, New Hampshire. Bellingham,J.G., et al. A SecondGeneration Survey A UV. inProceedings of the IEEESymposium onAutonomous Underwater VehicleTechnology AUV '94!. 1994.Cambridge, Ma.; IEEE. Bales,J.W. and E.R. Levine. Sensors for OceanographicApplications of AutonomousUnderwater Vehicles. inAUVS-94. 1994. Arlington: Association for UnmannedVehicle Systems,

173 19. Thomson,J,J., Cathode Rays. The London,Edinburgh, and Dublin Philosophical Magazine, 1897. 44: p. 293, 20. Kiser, R.W,, Introductionto MassSpectrometry and its Applications.1965, Englewood Cliffs, Nj: Prentice-Hall. 356. 21. Thomson,J,J,, Rays of positive electricityand their application to chemical analyses.1913, London: Longmans, Green and Co. 132. Thomson,J.J., Multiply-charged Atoms. The London,Edinburgh, and Dublin PhilosophicalMagazine, 1912. 24!; p, 668-672. 23. Syage,J.A., et al., Field-portable,high-speed GC/TOFMS, Journal of the AmericanSociety for MassSpectrometry, 2001. 12!: p. 648-655. 24. Niemann,H.B., et al., TheGalileo probe massspectrometer: Composition of' Jupiter's atmosphere.Science, 1996. 272263!; p. 846-849. 25. Palmer,P,T. and T.F. Limero, Massspectrometry in the USspace program: Past, present,and future. Journalof the AmericanSociety for MassSpectrometry, 2001. 12!: p, 656-675. 26. Neuman,J.A., et al., Fast-responseairborne in situ measurementsof HNO3 during the Texas2000 Air Quality Study.Journal of GeophysicalResearch- Atmospheres,2002. 107 D20!:p. art. no,-4436. 27. Richardson,L.B., Thesis:A field-usablemembrane-probe mass spectrometer, in Deptartmentof Civil Engineering.1988, Massachusetts Institute of Technology; Cambridge. p. 62. 28. Baykut, G, andJ. Franzen,Mobile Mass-Spectrometry a Decade of Field Applications.Trac-Trends in AnalyticalChemistry, 1994. 13!: p. 267-275. 29. Hemond,H.F., A Backpack-PortableMass-Spectrometer for Measurementof Volatile Compoundsin the Environment.Review of ScientificInstruments, 1991, 62!: p. 1420-1425. 30. Sinha,M.P. and A,D. Tomassian,Development of a Miniaturized,Lightweight Magnetic-Sectorfor a Field-PortableMass Spectrograph, Review of Scientific Instruments, 1991, 621!: p. 2618-2620, 31. Sinha,M.P. and G. Gutnikov,Development of a MiniaturizedGas- ChromatographMass- Spectrometer with a MicroboreCapillary Column and an Array Detector.Analytical Chemistry,1991. 638!: p, 2012-2016. 32. Takahashi,T., C. Wunsch,and e. al., Applications in AnalyticalChemistry to OceanicCarbon Cycle Studies. Oceanographic Measurements; Technologies for ChemicalMeasurements, 1993, Washington D,C.: National Academy Press. 85. 33. Johnson,J.E., Evaluation of a seawaterequilibrator for shipboardanalysis of dissolvedoceanic trace gases. Analytica Chimica Acta, 1999. 395-2!: p. 119- 132. 34. Park,S., Surface Seawater Sampling Device During ¹vigation. OceanResearch, 1995. 17!: p. 59-63. 35. Matz, G., W. Schroder, and G. Kibelka, Submersible Membrane Introduction- Gaschromatograph- Mass Spectrometer. unpublished, Technische Universitat Hamburg-Harburg. 36. Short,R.T., et al., Underwatermass spectrometers for in situ chemicalanalysis of thehydrosphere. Journal of theAmerican Society for MassSpectrometry, 2001, 12!: p. 676-682.

174 37. Kibelka,G., et al., Underwater Mass Spectrometry for Detection of VOCsand DissolvedGases. unpublished, Center for OceanTechnology Applied MicrosystemsLtd. 38. Hemond,H.F. and R, Camilli.Underwater Mass Spectrometers: Some Critical EngineeringIssues. in Third 8'orkshopon Harsh-Environment Mass Spectrometry.2002. Pasadena,Ca. 39. Camilli,R., Thesis: The Development of Components for an In-situ Mass Spectrometer,inDepartment of Civil and Environmental Engineering. 2000, MassachusettsInstitute of Technology:Cambridge. p. 102. 40. Einfeld,Wet al.,Environmental Technology Verification Report. 1997, US EnvironmentalProtection Agency, Office of Researchand Development: WashingtonDC. p. 1-78. 41. Poel,L.v.d., Product Sur vey on Commercial A UVs, in HydroInternational. 2002. p. 37-39. Conrad,R. and W. Seiler, Methane and Hydrogen in Seawater Atlantic-Ocean!, Deep-SeaResearch Part a-Oceanographic Research Papers, 1988. 352!: p. 1903-1917, 43. Schmidt,U. andR, Conrad,Hydrogen, Carbon-Monoxide, and Methane Dynamicsin Lake Constance. Limnology and Oceanography, 1993. 38!: p. 1214-1226. 44. Scranton,M.I., P. Donaghay,and J.M. Sieburth, Nocturnal Methane Accumulationin the Pycnocline ofan Anoxic Estuarine Basin. Limnology and Oceanography,1995. 40!: p. 666-672. 45 Tilbrook,B.D. and D.M. Karl, Methane Sources, Distributions and SinksPom CaliforniaCoastal 8'aters to theOligotrophic North Pacific Gyre. Marine Chemistry,1995. 49 l!: p. 51-64. Brekke,T,, O. Lonne, and S.E. Ohm, Light hydrocarbon gases in shallow sedimentsin the northern North Sea. Marine Geology, 1997. 137-2!: p. 81-108, 47. Deangelis,M.A. andM.I. Scranton,Fate of Methanein theHudson River and Estuary.Global Biogeochemical Cycles, 1993. 7!: p. 509-523. 48. Chen,G.T. and F.J. Millero, Gradual Increase of OceanicCo2. Nature, 1979. 277693!: p. 205-206. 49. Whiticar,M.J., Diagenetic relationships of methanogenesis, nutrients, acoustic turbidity, pockmarks andPeshwater seepages in Eckernforde Bay. Marine Geology,2002. 182-2!: p. 29-53. 50. Scranton,M.I, andK. McShane,Methane Fluxes in theSouthern North-Sea - the Roleof European Rivers. Continental Shelf Research, 1991. 11!: p. 37-52. 51. Frenzel,P., B. Thebrath,and R. Conrad, Oxidation of Methanein theOxic Surface-Layerofa DeepLake Sediment Lake Constance!. Ferns Microbiology Ecology, 1990.73!: p. 149-158. 52. Dafher,E., A. Obzhirov,and O. Vereshzhagina, Distribution of methanein waters of theOkhotsk and western Bering Seas, and the area of theKuril Islands. Hydrobiologia,1998. 362: p. 93-101. 53. Diaz,J.A., C,F. Giese, and W.R. Gentry, Mass spectrometryfor in-situ volcanic gasmonitoring. Trac-Trends in Analytical Chemistry, 2002. 21 8!:p. 498-514.

175 54. Sutton, A,J, et al., Implications for eruptive processes as indicated by sulfu r dioxide emissionsfrom Ki1auea Volcano, Hawai'i, 1979-1997. Journal of Volcanologyand GeothermalResearch, 2001. 108-4!: p. 283-302. 55. Tsunogai,U., et al., Peculiar Featuresof SuiyoSeamount Hydrothermal Fluids, Izu- Bonin Arc DifferencesPomSubaerial Volcanism.Earth and Planetary ScienceLetters, 1994.126!: p. 289-301. 56. Lilley, M.D,, M,A, Deangelis, and L.I. Gordon, Ch4, H-2, Co and N2o in SubmarineHydrothermal Vent Waters.Nature, 1982.300887!: p. 48-50. 57. Kodera,M,, G. Igarashi,and M. Ozima,Noble-Gases in HydrothermalPlumes of Loihi Seamount.Earth and Planetary Science Letters, 1988. 87!: p. 266-272. 58. Gamo, T., et al,, Methane Anomalies in Seawater above the Loihi Submarine SummitArea, Hawaii. GeochimicaEt CosmochimicaActa, 1987.510!: p. 2857-2864. 59. Sigurdsson,H., Kick'emJenny, R. Camilli, Editor. 2002,personal communication. 60. Robertson, R., et al., The explosive eruption of Soujiere Hills Volcano, Montserrat,West Indies, 17 September,1996. Geophysical Research Letters, 1998. 258!: p. 3429-3432. 61. Robertson,R.E.A., An Assessmentof theRiskPom Future Eruptionsaf the SoufriereVolcano of St-Vincent,West-Indies. Natural Hazards, 1995. 11!: p. 163-191. 62. Ishibashi,J,I., et al., Helium and CarbonGeochemistry of HydrothermalFluids from theNorth Fjii BasinSpreading Ridge SouthwestPacific!. Earth and PlanetaryScience Letters, 1994. 128-4!: p, 183-197. 63. Karl, D.M., C.O. Wirsen,and H,W. Jannasch,Beep-Sea Primary Production at the GalapagosHydrothermal Vents. Science, 1980. 207437!: p. 1345-1347. 64, Alper, J,, Oasesin the oceanicdesert. ASM AmericanSociety for Microbiology News, 1990. 560!: p. 536-538. 65. Tsunogai,U., et al., Fresh waterseepage and pore water recyclingon the seafloor: SagamiTrough subduction zone, Japan. Earth and Planetary Science Letters, 1996. 138-4!: p. 157-168. 66. Bussmann,I. and E. Suess,Groundwater seepage in EckernfordeBay Western Baltic Sea!:Effect on methaneand salinity distribution of the water column. ContinentalShelf Research,1998, 184-15!: p. 1795-1806. 67. Fingas,M. and C, Brown,Remote sensing of oil spills. SeaTechnology, 1997, 38 9!: p. 37-&. 68. Fingas,M, and C. Brown,A reviewof the statusof advancedtechnologies for the detectionof oil in and with ice. Spill Science& TechnologyBulletin, 2000.6- 6!: p. 295-302. 69. Lehr, W., J. Gait, ande. al. Trackingthe North CapeOil Spill. in Proceedingsof the 19thArctic and Marine Oilspill Program. 1996.Calgary. 70. Wang,Z.D,, et al,, Characterizationand sourceidentification of hydrocarbonsin watersamples using multiple analytical techniques.Journal of Chromatography A, 2002. 971-2!: p. 173-184, 71. Hoch,J. andB. Kok, A MassSpectrometer Inlet Systemfor SamplingGases Dissolvedin Liquid Phases,Archives of Biochemistryand Biophysics, 1963. 101: p. 160-170.

176 Allen, G., Final Report: MassSpectrometers. 1996, Massachusetts Institute of Technology:Unpublished; Cambridge. 73. Maden,A. andM. Hayward,Sheet Materials for Useas Membranes in Membrane IntroductionMass Spectrometry Analytical Chemistry, 1996. 68: p. 1805-1811. 74. Robinson,C,F, and L.G. Hall, Review of ScientificInstrumentation, 1956. 27: p, 504. 75 Bleakney,W. and J.A, Hippie, A NewMass Spectrometer with Improved FocusingProperties. Physical Review, 1938. 53. p. 521-528. 76. Brooks,R.A., A.I. Memo 864: A RobustLayered Control System For a Mobile Robot.1985, Massachusetts Institute of Technology:Cambridge. p. 1-25. 77 Russell,S. and P. Norvig, Artificial Intelligence a ModernApproach. 1995, Upper SaddleRiver, Nj.: PrenticeHall. 932. 78 Carver,S,J., et al., Evaluating field-based GIS for environmental characterization,modeling and decision suppor t. InternationalJournal of HeographicalInformation Systems, 1995. 9!: p, 475-486. 79 Vivoni, E.R.and R, Carnilli,Real-time Streaming of EnvironmentalField Data. Computersand Geosciences, 2002. 29!: p. 457-468. 80 Gates,S.C. and J. Becker, Laboratory Automation using the IBMPC. 1989, EnglewoodCliffs Nj.: PrenticeHall. 322. 81 Bellingham,J.G., et al. A smalllong range vehicle for deepocean exploration. in ProceedingsInternational Onshore and Polar Engineering Conference. 1992, San Francisco. Curtin,T.B., et al., Autonomous Ocean Sampling Networks. Oceanography, 1993. 6!: p. 86-94, 83 Schmidt,H., et al. Real-TimeFrontal Mapping with A UVsin a Coastal Environment.in Oceans 96 MTSlIEEE. 1996. Ft, Lauderdale, Florida. 84 Manley,J. andJ. Rieffel,GOATS 2000 September 25-October 12, 2000 A UV OperationsReport. 2000, MIT SeaGrant. p. 13. 85 Chryssostomides,C.,MIT Reports to thePresident, Sea Grant College Program. 2001,Massachusetts Institute of Technology. 86 Newman,P,, Thisisthe MOOS, in htt:lloceanai,rnit.edulnewmanlMOOSl. 2000, MIT Sea Grant. Manley,J., A UVSea Trials March 6 -15, 2002 Key 8'est, Florida Operations Report.2002, MIT SeaGrant AUV Laboratory:Cambridge. p. 15. 88. Damus,R., S, Desset, and J. Morash, Xanthos vehicle control system, R. Camilli, Editor. 2002. 89. Doherty,R.E., An Algorithm for MixtureComponent Determination For UseIn MembraneInlet Mass Specrometry, in Civil Engineering. 1987, Massachusetts Instituteof Technology:Cambridge. p. 272. Brandrup,J., E.H. Immergut, and E.A. Grulke, eds, Polymer Handbook. 4th ed. 1999, Wiley-Interscience: New York. 91. Weiss,R.F., The solubility of nitrogen, oxygen and argon in waterand seawater. Deep-SeaResearch, 1970. 17: p. 721-725. 92. Comyn,J., Polymer Permeability. 1985, New York: Elsevier Applied Science. 93. Varian,Ion Pumps-vacuum productsPom Varian. 2001. 94. Allen,G., Comprehensive Polymer Science, 1989, New York: Pergamon,

177 95. Crank,J., TheMathematics of Diffusion, 1975,Oxford UniversityPress: New York. 96. Jenkins,R., P. Nelson,and e. al., eds,Transactions of the FaradaySociety, 1970. 97. Wick, L. Y, andP.M. Gschwend,By-products of a formerphenol manufacturing site in a smalllake adjacentto a Superfundsite in theAbej rona watershed. EnvironmentalHealth Perspectives, 1998. 106: p. 1069-1074. 98. Wick,L.Y., et al.,Fate of benzenein a StratifiedLake receiving contaminated groundwaterdischarges from a Superfundsite. EnvironmentalScience & Technology,2000. 340!: p. 4354-4362. 99. Eganhouse,R,P. and P.M. Sherblom,Anthropogenic organic contaminantsin the effluent of a combinedsewer overflow: impact on BostonHarbor. Marine EnvironmentalResearch, 2001. 51!: p. 51-74. 100. Futo, I. andH. Degn,Effect of SamplePressure on MembraneInlet Mass- Spectrometry.Analytica ChimicaActa, 1994.294!: p. 177-184. 101. Senn,D,B,, Thesis:Coupled Arsenic, Iron, andNitrogen cyclingin Arsenic- ContaminatedUpper Mystic Lake, in Departmentof Civil and Environmental Engineering,2001, Massachusetts Institute of Technology:Cambridge MA. p, 284, 102. Sphethoff,H.M andH.F. Hemond, History of toxicmetal discharge to surface watersof theAberj ona iWatershed.Environmental Science & Technology,1996. 30!: p. 121-128, 103. Fricker, P,D., Thesis:The Effect of Stratificationand Bathymetryon Internal SeicheDynamics, in Departmentof Civil andEnvironmental Engineering. 2000, MassachusettsInstitute of Technology:Cambridge MA, p, 174. 104. Senn, D,B, and H.F. Hernond, Nitrate controls on iron and arsenic in an urban lake. Science,2002, 296577!: p. 2373-2376. 105. Andradottir,H., Thesis:Circulation and Mixing in the UpperForebay of the Mystic Lake System, iWinchester,Massachusetts, in Civil and Environmental Engineering.1997, Massachusetts Institute of Technology:Cambridge, p. 109. 106. Andersen,D.T., C.P. McKay, and R. Wharton,Dissolved gases in perenniallyice- coveredlakes of theMcMurdo Dry Valleys,Antarctica. Antarctic Science,1998. 10!: p. 124-133. 107. Craig, H., R.A. Wharton,and C.P.McKay, OxygenSupersaturation in Ice- CoveredAntarctic Lakes BiologicalVersus Physical Contributions. Science, 1992. 255042!: p. 318-321. 108. McKay, C.P.,personal communicationon samplingmethods inPozen Antarctic lakes, R, Camilli, Editor. 2002. 109. Signell, R.P.,H.L. Jenter,and A,F. Blumberg,Predicting thephysical effectsof relocatingBoston's sewage outfall. EstuarineCoastal and Shelf Science,2000. 50!: p. 59-71, 110. Horowitz, P. and W. Hill, The Art of Electronics. 2nd ed. 1989, New York. CambridgeUniversity Press.1125. 111. Landhal,C,E. andR.G. Merryman,Modifications to CEC 21-620Mass Spectrometerfor IncreasedSensitivity. 1960, Los Alamos ScientificLaboratory of the University of California:Los AlamosNM. p. 1-20.

178 112. Camilli,R. andH.F. Hemond. Multisensor Data Integration and Adaptive SamplingStrategies for anAutonomous Underwater Mass Spectrometer. in Third 8'orkshopon Harsh-EnvironmentMass Spectrometry. 2002. Pasadena, Ca. 113. Laursen,A.E. andS.P. Seitzinger, The role of denitrificationin nitrogenremoval and carbon mineralizationin Mid-Atlantic Bight sediments.Continental Shelf Research,2002. 22 9!: p, 1397-1416. 114. Delaney,J.R, and A.D. Chave,NEPTUNE: A fiber-optic'telescope'to inner space.Oceanus, 2000, 42!: p. 10-11, 115. Aoki, T. andT. Shimura,Fuel Cellfor Long-RangeAUV. SeaTechnology, 1997. 38 8!: p. 69-73. 116. Rudd,E,J, and D.W. Gibbons, High energydensity aluminum/oxygen cell. Journal of PowerSources, 1994. 47: p. 329-340. 117. Curico,J.A,, F.S. Hover, and J.G, Bellingham, CETUS; Design of a Production AUV. SeaTechnology, 1998, 392!: p. 10-17, 118. Hillenbrand,C., UUVScience 4 TechnologyTrends. Sea Technology, 1997. 382!: p. 10-15. 119. Smith,C.M., et al. Feature-basedconcurrent mapping and localization for autonomousunderwater vehicles, in Proceedingsof IEEE Oceans.1997.

179