Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Rational development of the solvent loaded, inductively coupled argon plasma Weir, Douglas Glenn John 1994

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
831-ubc_1994-954078.pdf [ 10.55MB ]
Metadata
JSON: 831-1.0061679.json
JSON-LD: 831-1.0061679-ld.json
RDF/XML (Pretty): 831-1.0061679-rdf.xml
RDF/JSON: 831-1.0061679-rdf.json
Turtle: 831-1.0061679-turtle.txt
N-Triples: 831-1.0061679-rdf-ntriples.txt
Original Record: 831-1.0061679-source.json
Full Text
831-1.0061679-fulltext.txt
Citation
831-1.0061679.ris

Full Text

RATIONAL DEVELOPMENT OF THESOLVENT LOADED, INDUCTIVELY COUPLED ARGON PLASMAByDOUGLAS GLENN JOHN WEIRBachelor of Science, University of Alberta, 1986A THESIS SUBMITfED IN PARTIAL FULFILMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDepartment of ChemistryWe accept this thesis as conformingto the required standard__THE UMVERSITY OF BRITISH COLUMBIASeptember, 1993© Douglas Glenn John Weir, 1993In presenting this thesis in partial fulfilment of the requirementsfor an advanceddegree at the University of British Columbia, I agree that theLibrary shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted bythe head of mydepartment or by his or her representatives. It is understoodthat copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature)__________________________Department of___________________________The University of British ColumbiaVancouver, CanadaDate tLDE-6 (2/88)ABSTRACTIn routine trace metal analyis, the inductively coupled argon plasma (ICAP) convertsanalyte into atoms and atomic ions, then excites them to emit intense, characteristic line spectra.But none of the spectrochemical techniques based on the ICAP are free from interference effects.This dissertation focuses on only one source of interference effects: the solvent plasma loadassociated with injecting sample solution into the discharge as an aerosol mist. This dissertationreveals several physical phenomena that are responsible for both the spectral and nonspectralinterference effects associated with solvent plasma loading, physical phenomena that are spatially,temporally and parametrically complex.In order to address that complexity, a strategy was proposed for characterizing anddeveloping electrical gas discharges for spectrochemical analysis. The strategy consisted of threesteps which are not strictly sequential. Step one: systematically survey the parametric, spatial andtemporal complexity of analyte and background emission (Chapters 3-6); Step two: drawingguidelines from step one, characterize the physical properties of the discharge (Chapters 7-9); Stepthree: Improve the design and methodology of the spectrochemical method based on the insightgained in the first two steps. The first two steps proved to be very effective, but the third stepremains untested.(1Table of ContentsPageAbstract.iiTable of Contents iiiListofTablesList of FiguresGlossary of Symbols and Abbreviations xixAcknowledgements xxiiOpenning Quotation xxiiiChapter 1: Thesis Introduction and Summary 11.1 Context and scope 11.2 General objective 21.3 An overview of the solvent load problem, from sample introduction to signal detection.. 41.4 Relevant properties of the sample aerosol 91.5 The interface between the sample aerosol and the discharge 101.6 Relevant properties of the inductively coupled argon plasma 111.7 Thesis objectives and summary 171.8 References 20‘UChapter 2: The Experimental Setup 222.1 Sample introduction 222.2 Torch 292.3 Power Supply 292.4 Ignition procedure 332.5 Light collection optics 332.6 References 36Chapter 3: Spectral Survey and Observations 373.1 Introduction 373.2 Spectral survey 423.3 Photographs 483.4 Detailed observations 513.5 Conclusions 683.6 References 69Chapter 4: The Parametric Complexity of a Chloroform Loaded Inductively Coupled ArgonPlasma 704.1 Introduction 704.2 Experimental section 964.3 Results 98iv4.4 Chapter summary 1244.5 Conclusions 1254.6 References 126Chapter 5: The Spatial Complexity of the Solvent Loaded Inductively Coupled Argon Plasma 1285.1 Introduction 1285.2 Experimental 1385.3 Results and discussion 1425.4 Discussion 1785.5 Conclusions 1795.6 References 180Chapter 6: The Temporal Complexity of the Solvent Loaded Inductively Coupled Argon Plasma1826.1 Introduction 1826.2 Experimental 1946.3 Results 1976.4 Conclusions 2146.5 References 215Chapter 7: A Parametric Survey of Electron Density in the Solvent Loaded Inductively CoupledArgon Plasma 2197.1 Introduction 219V7.2 Guidelines offered by previous chapters for investigating the physical characteristics of thedischarge 2207.3 An accurate physical description of the ICAP: Plasma parameters 2237.4 Determining electron number densities from the absolute intensity of a single Ar I line... 2377.5 Experimental 2497.6 Results 2517.8 Discussion 2647.9 Conclusions 2667.10 References 267Chapter 8: The Response of Electron Density to Solvent Plasma Load and Inner Argon FlowRate in the Induction Region of the Inductively Coupled Argon Plasma 2698.1 Introduction 2698.2 Experimental 2738.3 Results 2828.4 Discussion 2948.5 Conclusions 3008.6 References 301Chapter 9: The Effect of Solvent Load and Inner Argon Flow Rate on the Atomic StateDistribution Function of Iron in an Inductively Coupled Argon Plasma 3059.1 Introduction 3059.2 Experimental 311vi9.3 Results.3199.4 Conclusions 3509.5 Reference 351Chapter 10: Concluding Remarks and Future Directions 354vi’List of TablesPagTable 2.1: Summary of the ranges of solvent plasma load investigated 27Table 2.2: Summary of the experimental setup 29Table 9.2.1: Wavelength, excitation energy, statistical weight of the upper state, atomictransition probabilities, and error in the transition probability for Fe II lines used in thiswork 316Table 9.2.2: Wavelength, excitation energy, statistical weight of the upper state, atomictransition probabilities, and error in the transition probability for Fe I lines used in thiswork 317vu’List of FiguresPageFigure 1.1: Introductory overview of the solvent loaded inductively coupled argonplasma, showing the confinement tube, the argon flow rates, and the major components ofthe discharge 3Figure 1.2: The path taken by sample aerosol from nebulizer to plasma 7Figure 1.3: The flowfield in the inductively coupled argon plasma 14Figure 2.1: A schematic overview of the experiment 23Figure 2.2: (a) Calibration plot of the chloroform plasma load versus the temperature ofthe aerosol desolvating condenser; (b) Schematic flow chart of the continuous weighingmethod used to calibrate the solvent plasma load against condenser temperature 25Figure 2.3: The flow of power in a solvent loaded ICAP 32Figure 2.4: Sample maps of the light collection efficiency for the optical trains used inthis work 35Figure 3.2.1: Spectral survey of the visible emission from the inductively coupled argonplasma loaded with methanol 43Figure 3.2.2: Spectral survey of the visible emission from the inductively coupledplasma loaded with water 44Figure 3.2.3: Spectral survey of the visible emission from the inductively coupled argonplasma loaded with chloroform 45Figure 3.2.4: The relative luminosity curve or relative eye sensitivity of a standardobserver adjusted to bright lighting 46ixFigure 3.3.1: Photographs of a chloroform loaded ICAP 49Figure 3.3.2: Photographs of 50Figure 3.4.1: Visually observed emission structures of a solvent loaded ICAP 52Figure 3.4.2: Representative observations of an ICAP discharge 55Figure 3.4.3: Representative observations for an ICAP loaded with methanol 59Figure 3.4.4: Representative observations for an ICAP loaded with chloroform 62Figure 3.4.5: Observations of how the shape of the inner plume varied as the chloroformplasma load was increased 64Figure 3.4.6: Representative observations for an ICAP loaded with an intermediate levelof chloroform on an inner argon stream at varying flow rates 66Figure 4.1.1: The spatial averaging intrinsic to line of sight measurements 94Figure 4.3.1: Axially resolved profiles of Mg II emission (279.55 nm) from aninductively coupled argon plasma loaded with chloroform 100Figure 4.3.2: Axially resolved profiles of Mg I emission (285.21 nm) from aninductively coupled argon plasma loaded with chloroform 103Figure 4.3.3: Axially resolved profiles of atomic carbon emission (248 nm) from aninductively coupled argon plasma loaded with chloroform aerosol 106Figure 4.3.4: Axially resolved profiles of diatomic carbon emission (516 nm) from aninductively coupled argon plasma loaded with chloroform aerosol 110Figure 4.3.5: Axially resolved profiles of cyanogen radical emission (388 nm) from aninductively coupled argon plasma loaded with chloroform aerosol 113Figure 4.3.6: Axially resolved profiles of the ratio of Mg II emission (279.55 nm) to MgI emission (285.2 mm) from an inductively coupled argon plasma loaded with chloroform .... 121xFigure 5.1.1: Four possible optical configurations for capturing monochromatic imagesof a spectrochemical source 132Figure 5.3.1: The reference frame and contour intervals for the spatially resolvedintensity maps 143Figure 5.3.2: An overview of spatially resolved maps of analyte and backgroundemission from the solvent loaded inductively coupled argon plasma 145Figure 5.3.3: Maessen and Kreuning’s spatially resolved profiles of emission fromsolvent pyrolysis products 149Figure 5.3.4: Maessen and Kreuning’s spatially resolved profiles of emission fromsolvent pyrolysis products 150Figure 5.3.5: The characteristic structure of emission plumes for (a) soft, (b)intermediate, and (c) hard line emission from the inductively coupled argon plasma 151Figure 5.3.6: Isocontour maps of CN (388.34 nm) emission intensity for a chloroformloaded ICAP 154Figure 5.3.7: Isocontour maps of C2 (516.56 nm) emission intensity for a ineta-xyleneloaded ICAP 157Figure 5.3.8: Isocontour maps of C I (247.61 nm) emission intensity for a chloroformloaded ICAP 159Figure 5.3.9: Isocontour maps of C I (247.61 nm) emission intensity within theinduction region 162Figure 5.3.10: The response of a hard line plume (Mg II 279.55 nm) to chloroformplasma load 164Figure 5.3.11: The response of a soft line plume (Mg I 285.21 nm) to chloroformplasma load 167Figure 5.3.12: The response of the magnesium line intensity ratio (Mg II 279.55 nm /Mg I 285.21 nm)) to chloroform plasma load 169xiFigure 6.1.1: The 3D varicose wave that falls off the argon jet into a ring vortex 191Figure 6.1.2: Vortex shedding from the inductively coupled argon plasma 192Figure 6.3.1: The effect of chloroform plasma load on the 0 - 500 Hz noise spectra forthe Mn II line at 257.610 nm 198Figure 6.3.2: The effect of chloroform plasma load on the 0 - 500 Hz noise spectra forCN emission at 388.340 nm 199Figure 6.3.3: The effect of methanol plasma load on the 0 - 500 Hz noise spectra for theMn II line at 257.610 nm 200Figure 6.3.4: The effect of methanol plasma load on the 0 - 500 Hz noise spectra for CNemission at 388.340 nm 201Figure 6.3.5: The effect of solvent plasma load on the vortex shedding frequency 204Figure 6.3.6: The transient signal for the Ca I line at 422.673 nm 209Figure 6.3.7: Noise spectra for the Ca I line at 422.673 nm 211Figure 6.3.8: Low frequency noise spectra for C2 emission at 5 16.520 nm 213Figure 7.3.1 The departure of experimental atomic state distribution functions from theSaha distribution. (a) depicts the case of an overpopulation of the low lying atom levelsbalanced by and underpopulation of the upper ion levels. (b) depicts the case of anunderpopulation of the low lying atom levels balanced by and overpopulation of the upperion levels 231Figure 7.4.1 A schematic illustration of determining the electron density and electrontemperature in the inductively coupled argon plasma from the absolute intensity of a singleargon line 238Figure 7.4.2 The effect of departures from local thermal equilibrium on thedetermination of electron density and electron temperature in the inductively coupled argonplasma. Note that the departure from LTE has been exaggerated for clarity 242xiiFigure 7.4.3 Error in the absolute line intensity method for realistic departures fromlocal thermal equilibrium 244Figure 7.4.4 Variation of four key quantities of the absolute line method over electrontemperatures typically encountered in the inductively coupled argon plasma 246Figure 7.4.5 Electron density versus the absolute intensity of a single argon line(687.127 nm). The solid line was calculated assuming complete LTE. The dashed linesassume partial LTE and set the bounds of realistic departures from LTE. For thedetermination of electron density from the absolute intensity of the Ar I 687.127 nm line,this Figure shows that the error owing to departures from LTE is bounded by ± 30% 248Figure 7.5.1 The emission spectrum of the inductively coupled argon plasma in thevicinity of the 687.129 nm line used in this work 250Figure 7.6.1(a): The response of electron density to r.f. power and methanol load,revealed by isocontour maps of electron density e (1015 cm-3) 253Figure 7.6.1(b): The response of electron density to r.f. power and methanol load,revealed by isocontour maps of electron density e (1015 cm3) 254Figure 7.6.1(c): The response of electron density to r.f. power and methanol load,revealed by isocontour maps of electron density e (1015 cm3) 255Figure 7.6.2(a): The response of electron density to r.f. power and chloroform load,revealed by isocontour maps of electron density e (1015 cm3) 257Figure 7.6.2(b): The response of electron density to r.f. power and chloroform load,revealed by isocontour maps of electron density e (1015 cm3) 258Figure 7.6.2(b): The response of electron density to r.f. power and chloroform load,revealed by isocontour maps of electron density e (1015 cm3) 259Figure 7.6.3(a): The response of electron density to r.f. power and water load, revealedby isocontour maps of electron density e (1015 cm3) 261Figure 7.6.3(b): The response of electron density to r.f. power and water load, revealedby isocontour maps of electron density e (1015 cm3) 262xlii0.61 11mm; (b) 0.3 mg/s and 0.81 Llmin; (c) 10.0 mg/s and 0.61 11mm; (d) 10.0 mg/s and0.81 1/mm. The r.f. power was 1.25kW.290Figure 8.3.5 The response of radially resolved electron density profiles to inner argonflow rate and water plasma load. The r.f. power was 1.25 kW 292Figure 8.3.6 The response of radially resolved electron density profiles to inner argonflow rate, chloroform plasma load and xylene plasma load. The r.f. power was 1.25 kW 293Figure 8Figure 7.6.3(c): The response of electron density to r.f. power and waterload, revealed by isocontour maps of electron density e (1015 cm-3) 263Figure 8.2.1 Electron density as a function of electron temperature in an argon plasma,according to the Saha equation. The solid curve assumes local thermal equilibrium. Forthe dashed curves, the argon ground state was assumed to be overpopulated by a factor of10 or underpopulated by a factor of 0.1 with respect to local thermal equilibrium 274Figure 8.2.2 Cubic spline interpolated, theoretical profiles for the Hp line for a range ofelectron densities 275Figure 8.2.3 Electron density versus the theoretical full width at half maximum of theH line. The widths were determined from cubic spline interpolated, theoretical profiles 277Figure 8.2.4 Log plot of electron density versus the theoretical full width at halfmaximum of the Hp line. The widths were determined from cubic spline interpolated,theoretical profiles 278Figure 8.2.5 Comparison of a normalized, cubic spline interpolated, theoretical Hp lineprofile with a closely matching(by visual inspection) experimental proffle. Evident in theFigure is spectral interference that could corrupt line width determinations 279Figure 8.2.6 The effect of aperture size on radially resolved profiles of electron density.... 281Figure 8.3.1 The response of radially resolved electron density profiles to inner argonflow rate for an inductively coupled argon plasma loaded with 0.3 mg/s methanol. The r.f.power was 1.25 kW 284xivFigure 8.3.2 The response of radially resolved electron density profiles to inner argonflow rate for an inductively coupled argon plasma loaded with 1.0 mg/s methanol. The r.f.power was 1.25 kW 286Figure 8.3.3 A summary of the response of the plasma region of the discharge tomethanol plasma load. The isocontours in this figure represent electron densityqualitatively; (a) low methanol load, (b) high methanol load 288Figure 8.3.4 The decay of electron density profiles with axial distance in a dischargeloaded with 0.3 or 10.0 mg/s methanol at an inner argon flow rate of 0.61 or 0.81 11mm:(a) 0.3 mg/s and.3.7 A summary of the response of the plasma region of the discharge tochloroform or xylene plasma load. The isocontours in this figure represent electron densityqualitatively; (a) low solvent load, (b) high solvent load 294Figure 9.1.1 The LTE framework. Within this framework, the departure ofexperimentally determined atomic state distribution functions (ASDFs) from local thermalequilirium (LTE)may be assessed. The solid line represents the theoretical ASDF at LTE,or the Saha distribution. The dashed curves represent the predictions fa collisional-radiative (CR) rate model. The open circles represent the experimental ASDF 309Figure 9.2.1 The map of light collection efficiency in the XZ plane) for the optical trainused to collect iron line emission (see Figure 2.1—the XY plane is the horizontal planedefined by the lateral direction across the discharge and the viewing direction or line ofsight. In this map, the collection efficiency has been integrated over a distance of 2.0 mmin the Y direction (vertical, parallel to the discharge axis) 312Figure 9.2.2 (a) foreground spectrum, (b) background spectrum, (c) detector responsefunction, and (d) the corrected spectrum (foreground - background) of iron lie emission forthe linear photodiode array detector window at 370 nm, at an observation height of 15 mmabove the induction coil, for an ICAP loaded with 6 mg/s of chloroform, and an r.f. powerof 1.25 kW 314Figure 9.2.2 (a) foreground spectrum, (b) background spectrum, (c) detector responsefunction, and (d) the corrected spectrum (foreground - background) of iron lie emission forthe linear photodiode array detector window at 265 nm, at an observation height of 15 mmabove the induction coil, for an ICAP loaded with 6 mg/s of chloroform, and an r.f. powerof 1.25 kW 315xvFigure 9.3.1 Radial profiles of normalized Fe I emission intensity for lines covering arange of excitation energies 321Figure 9.3.2 Radial profiles of Fe I emission (373.7 13 nm) at 15 mm above theinduction coil, for a range of distances between the tip of the inner boundary and theobservation zone 323Figure 9.3.3 Curvature in line if sight and radially resolved Fe I and Fe II Boltzmannplots. Values of ln(nlg) represented by the open circles were taken from a radial position (inthe radially resolved plots) of r = 0.0 mm or a lateral position (in the line of sight plots) ofx 0.00 mm; the crosses, r = 0.60 mm or x = 0.60 mm; the filled circles, r 1.8 mm or x= 1.8 mm 325Figure 9.3.4 Radial and lateral profiles of (a) Fe I excitation temperature, (b) curvatureof the Boltzmann plot, and (c) temperature uncertainty for a water loaded ICAP, 15 mmabove the induction coil; water loads: A = 0.10, B = 0.15, C = 0.20 mgls; inner argonflow rate = 0.81 1/mm; r.f. power = 1.25 kW 327Figure 9.3.5 Fe I excitation temperature at 15 mm above the induction coil plottedagainst the axial displacement f the dissociation front from upstream from that viewingheight 330Figure 9.3.6 Radial profiles of Fe II emission (273.955 nm),normalized to unt intensityat 15 mm aboe the induction coil, from an ICAP loaded with (a) chloroform, (b) methanol,(c) water, for ranges of solvent load; inner argon flow rates: 0.61, 0.81, 1.01 1/mm; r.f.power = 1.25 kW. except 1.00 kW for the dotted curve in frame (c) 332Figure 9.3.7 Radially resolved profiles of tangential excitation temperatures andBoltzmann plot curvature. The inductively coupled argon plasma was loaded with 7.4mg/s of chloroform, the r.f. power was 1.25 kW and the inner argon flow rate was 0.811/mm. The Fe I tangential temperatures were evaluated at excitation energies of 3.2, 4.7and 6.2 eV with respect to the atomic ground state. The Fell tangential temperatures wereevaluated at excitation energies of 12.7, 14.4 and 16.0 eV with respect to the atomicground state 336Figure 9.3.8 Radially resolved profiles of tangential excitation temperatures andBoltzmann plot curvature. The inductively coupled argon plasma was loaded with 1.0mg/s of methanol, the r.f. power was 1.25 kW and the inner argon flow rate was 0.81xvi1/mm. The Fe I tangential temperatures were evaluated at excitation energies of 3.2, 4.7and 6.2 eV with respect to the atomic ground state. The Fe II tangential temperatures wereevaluated at excitation energies of 12.7, 14.4 and 16.0 eV with respect to the atomicground state 337Figure 9.3.9 Profiles of argon ionization temperature at z = 15 mm above the inductioncoil, an r.f. power = 1.25 kW and for loading by various solvents, at different solventplasma loads and inner argon flow rates. If one assumes that the argon plasma is close tolocal thermal equilibrium (or that the non equilibrium parameter for the argon atomicground state b1 in equatuation 9.2 is close to 1.0), then these profiles may be regarded asproffles of electron temperature 340Figure 9.3.10 The error intrinsic to electron temperatures obtained from equation 9.2and accurate electron densities. The error results from the argon ASDF departing fromlocal thermal equilibrium. Several realistic values of b1 were inserted into equation 9.2.The axes extend over the range of electron temperatures and electron densities likely to beincountered in the inductively coupled argon plasma. For an example of how to guage theerror in the electron temperature, the b1 = 0.1 curve (partial LTE) lies below the solid b1 =1.0 curve. Consequently, the electron temperature would be overestimated by 800 K if oneassumed that the plasma was in LTE 342Figure 9.3.11 Experimentally determined state densities for Fe I and Fe II (theexperimental ASDF) within the LTE framework. The bold, solid line defines the Sahadistribution—for an electron temperature calculated from an accurate electron density andequation 9.2. The other solid lines assume that the electron temperature has beenoverestimated (b1 < 1.0) because the argon plasma gas is recombining. Also shown arethe predictions of a collisional radiative (CR) model (dashed curve) 344Figure 9.3.12(a) Experimentally determined state densities for Fe I and Fe II (theexperimental ASDF) within the LTE framework an r.f. power of 1.25 kW and chloroformload 346Figure 9.3.12(b) Experimentally determined state densities for Fe I and Fe II (theexperimental ASDF) within the LTE framework for an r.f. power of 1.25 kW and waterload 347xviiFigure 9.3.12(c) Experimentally determined state densities for Fe I and Fe II (theexperimental ASDF) within the LTE framework for an r.f. power of 1.25 kW and highmethanol load 348Figure 9.3.12(d) Experimentally determined state densities for Fe I and Fe II (theexperimental ASDF) within the LTE framework for an r.f. power of 1.25 kW and lowmethanol load. 349xviiiGLOSSARY OF SYMBOLS AND ABBREVIATIONSPhysical QuantitiesA, atomic transition probability of q —> p optical transition, 10 8 s1b nonequilibrium parameter; ratio of the actual, or experimentalatomic state density to the atomic state density for local thermalequilibrium121 nonequilibrium parameter for the atomic ground stateb nonequilibrium parameter for the ith excited statee electron charge, 1.602 19 X ‘9 CE, ionization potentialE, Eq excitation energy of state p, of state qg gain of photomultiplier;degeneracy or statistical weight of atomic statestatistical weight of atomic ground state, excited state q,excited state p, and the atomic ion ground state (singly ionized atom)h Planck constant, 6.626 18 x j s‘qp absolute intensity of q — p optical transition, W cm-31q energy difference between atomic state p and the ionization limit,energy difference between atomic state q and the ionization limitk Boltzmann constant, 1.38066 X 10 jme electron mass, 9.10953 x 10 31 kgn atomic state density;number of photoelectrons per photon for a photomultiplierfli , , flq , n state density for the atomic ground state, atomic excited state p,atomic excited state q, and the ion ground state.xix‘7 . state density per statistical weight.hi, T1p 17q’ 17jf, i state density per statistical weight for the atomic ground state,atomic excited state p, atomic excited state q, the ion ground state,and ionization limit.p5(p) j(fl ,1B() LSE value of 17 for atomic state p. 17 for the atomic ion groundstate, LBE value of i for atomic state p.1e free electron density of the plasmavqp frequency of light for the optical transition q —* pP atmospheric pressure, iO PaS shot noise signalTe, Tg ,T107 Texc electron temperature, atomic gas temperature,Saha ionization temperature and Boltzmann excitation temperature2AAS full width at half maximum of the Doppler and Stark broadenedline profileA2Lexpt ,A2 experimental and instrumental full width at half maximumelectrical conductivity, Ohms1cm-1Qi coolisional cross section of electrons with plasma species i, cm2AbbreviationsASDF atomic state distribution functionCR collisional radiativeICAP inductively coupled argon plasmaICAP-AES inductively coupled argon plasma atomic emission spectrometryLBE local Boltzmann equilibriumxxLIE.local isothermal equilibriumLSE local Saha equilibriumLTE local thermal equilibriumMg ratio ratio of the intensity of the Mg 11(279.55 nm)ion line to the intensity of the Mg 1(285.21 nm) atomic linepLTE partial local thermal equilibriumPMT photo multiplier tubeRSD relative standard deviationRSDB relative standard deviation of the background signalSBR signal to background ratioSNR signal to noise ratioxxiACKNOWLEDGEMENTSI FIRST WISH TO EXPRESS MY GRATITUDE AND SINCERE LOVE TO MY FATHER,DR. DONALD ROBERT WEIR, A HYDROMETALLURGIST AN]) A PRACTICAL MANOF SCIENCE, TO MY MOTHER, MRS. ANN MARIE WEIR, A DEDICATED HIGHSCHOOL TEACHER, AND TO MY GRANDMOTHER, MRS. BARBARA WEIR, ACHAMPION CURLING SKIPPER. IF IT WERE NOT FOR THEIR LOVE, SUPPORT ANDKThJI) WISDOM, I WOULD NEVER HAVE COMPLETED THIS ADVENTURE.NEXT, I WOULD LIKE TO THANK TWO CLOSE FRIENDS WHO SHARED MYAPPRECIATION FOR THE LIFE AND ACCOMPLISHMENTS OF RICHARD P.FEYNMAN—B. OLAV ANDERSON AND CHARLES W. LEBLANC.SPECIAL THANKS ARE DUE TO THE MAN WHO FAiTHFULLY PROOF READ MUCH OFTHIS WORK, WHO YOU WOULD DEFINITELY WANT TO HAVE IN THE TRENCHESWITH YOU—GUY KENNETH CLENTSMITH.THANKS ALSO TO MYLUAGANAM, CHANDRAKUMAR, IVAN, TIM, ADRIAN,ALEXIS, LYLE, PETE, EDGAR, BERNARD, MARTIN, NARLY, MAHEEN, BEN,MARK, ROB, LAURELLE, CHRISTOPHER, KEN, DAMON, LEE, CHRISTIAN,CAROLINE, CHRISTIANE SCHACHT VON LEVERKUSEN, ROCIO, DEBBIE,SYLVIA, SUSAN, AND ALL THE CHEMISTRY DEPARTMENT SECRETARIES,MACHINE SHOP AND ELECTROMCS SHOP STAFF, LARGELY FOR KEEPINGME SANE.MANY THANKS TO MY RESEARCH SUPERVISOR, DR. MICHEAL W. BLADES, FORKEEPING THINGS RUNNING SMOOTHLY, KEEPING THE BUREAUCRATS OFF MYBACK AND FOR LENDING ADVICE, YET ALLOWING ME TO WORK INDEPENDENTLY.FOR SIMILAR REASONS, I WOULD LIKE TO THANK DR ADRIAN P. WADE.FINALLY, I AM INDEBTED TO THE PATIENCE AND UNDERSTANDING OF MY SONAND BEST FRIEND, CAMERON KIMBALL LAING, AND HIS MOM, ELIZABETH.THIS DISSERTATION IS DEDICATED TO THE MEMORY OF MYGRANDPARENTS,ANNA MARIA GASZLER, VINCENT GASZLER AND DONALD HOWARDWEIR.xxiiOPENNING QUOTATIONWhereas I believed myself born for the common good, and reckoned the care of thecommon weal to be among those duties that are of public right, open to all alike, even asthe waters and the air, I therefore asked myself what could most advantage mankind, andfor the performance of what tasks I seemed to be shaped by nature. But when I searched, Ifound no work so meritorious as the discovery and the development of the arts andinventions that tend to civilize the life of man. . . Above all, if any man could succeed—not merely in bringing to light one particular invention, however useful—but kindling innature a luminary which would, at its first rising, shed some light on the present limits andborders of human discoveries, and afterwards, as it rose still higher, would reveal andbring into clear view every nook and cranny of darkness, it seemed to me that such adiscoverer would deserve to be called the true Extender of the Kingdom of Man over theuniverse, the Champion of human liberty, and the Exterminator of the necessities that nowkeep men in bondage. Moreover, I found in my own nature a special adaptation for thecontemplation of truth. For I had a mind at once versatile for that most important object—Imean the recognition of similitudes—and at the same time sufficiently steady andconcentrated for the observation of subtle shades of difference. I possessed a passion forresearch, a power of suspending judgement with patience, of meditating with pleasure, ofassenting with caution, of correcting false impressions with readiness, and of arranging mythoughts with scrupulous pains. I had no hankering after novelty, no blind admiration forantiquity. Imposture in every shape I utterly detested. For all these reasons I consideredthat my nature and disposition had, as it were, a kind of kinship and connection with truth.FRANCIS BACONxxiiiChapter 1Thesis Introduction and Summary1.1. Context and ScopeThis thesis contributes to a particularly active field of research in analytical chemistry—the development of spectrochemical methods for trace metal analysis. In this field, workersgenerally have two principal aims: 1., to invent new spectrochemical methods for trace metalanalysis and 2., to develop existing methods rationally. In order to invent new methods, theymay begin by devising new ways to convert a sample into a volume of atoms or atomic ions.This is usually accomplished with a spectrochemical source, such as a chemical flame or anelectrical discharge. Next, the volume of atoms in the spectrochemical source is probed foratomic emission, fluorescence, absorbance or atomic ion density. The resulting spectrometricsignal must finally be decoded in order to obtain the trace metal composition of the originalsample. In order to do that, a new methodology may have to be devised. Of course, when anew spectrochemical method is devised this way, the prototype usually performs poorly.Nevertheless its analytical performance, precision, accuracy and detection limits, can usually beimproved by tailoring the spectrometer to suit the spectrochemical source, or the source to suitthe sample. Indeed, even established methods have room for improvement. A good example isatomic emission speetrometry with an inductively coupled argon plasma as a spectrochemicalsource. In summary, the task of workers in this field is not only to devise new spectrochemicalmethods, but to rationally develop spectrochemical methods that are already established as well.Boumans argued that in order to develop a spectrochemical method rationally, theexperimentalist must follow four steps [1]. First, the parametric response of the method must beestablished. That reveals where the optimum performance lies in the operating parameter space.Moreover, the performance at non-optimal parameters reveals how the method may sufferIntroduction 2interferences. Second, the physical properties of the method must be characterized. In order todo that one must resort to spectroscopy in order to characterize the physical properties of thespectrochemical source—but not always. Sometimes the interface between the source and thesample introduction system must be examined. Other times, the interface between the sourceand the spectrometer requires scrutiny. Ideally, the experimentalist should characterize everystep in the method, from sample introduction to signal detection. Having characterized themethod satisfactorily, the third step is to explain the analytical performance and the parametricresponse in terms of the physical properties. That gives one a rational basis for the fourthstep—to develop, refine and improve the method. One can do that by tailoring the source,optics and spectrometer, or prescribing a better protocol for routine analysis, all with the insightinto how the physical properties affect the analytical performance.1.2. General ObjectiveThe general objective of this thesis was to rationally develop a relatively mature, yet byno means flawless spectrochemical method: Inductively Coupled Argon Plasma - AtomicEmission Spectrometry (ICAP-AES).At the heart of this spectrochemical method lies an electrodeless discharge, theinductively coupled argon plasma (ICAP). This source is illustrated in Figure 1.1. It is moreaccurate to call it a discharge rather than a plasma because it consists of both plasma andboundary regions. At the core of the discharge lies a toroidal plasma region, aptly described asa low temperature (< 10000 K), weakly ionized (< 2 %) thermal plasma at atmosphericpressure. Its outside diameter is roughly 15 mm, its inside diameter 4 mm, and its length 15mm. Because the plasma gas retains much of its energy as it flows out of the toroidal inductionregion into the room air, a conical region of plasma extends beyond the exit of the torch. Thisregion may be referred to as the plasma decay region. Here, steep thermal gradients betweenthe induction region and the axial channel decay as the plasma gas flows downstream. Thegradients decay by transport processes such as heat conduction, escape of radiation andambipolar diffusion. On the outside of the jet, however, cold room air entrained into the plasmaIntroduction 3000Outer, coolantargon flow,10.0 to 12.0 1/mmIcmFigure 1.1. Introductory overview of the solventloaded inductively coupled argon plasma, showing theconfinement tube, the argon flow rates, and the majorcomponents of the discharge.downstreamboundary regionanalyte plumepisrna tail coneor decay region0Oinductioncoil0torch wallinner argon flow,0.6 to 1.0 I 1mminjector tubeintermediateargon flow,0 to 1.0 11mmIntroduction 4jet extinguishes the plasma before it has a chance to decay. In fact, air entrainment gives theplasma its characteristic conical shape, a shape reminiscent of the potential core of a jet flowinginto quiescent room air. In this case, one may say that thermal gradients actually grow steeperat the outer boundaries the decay region as a result of heat conduction and radiation, and thatthese gradients cause transport by diffusion. (In fact, this development of thermal gradientsrather than decay was emphasized by the external examiner of this dissertation, P.B.Farnsworth, so strictly speaking, decay region is a misnomer for what is really a conical tailregion.) In any case, the conical decay region generally has an apex that is 20 to 25 mmdownstream from a 15 mm diameter base. The entire plasma, cone and toroid together, issurrounded by a boundary region consisting of two distinct components, one upstream and onedownstream. The upstream boundary region (shown in solid black) resides within the torch andwraps around the base of the plasma. This component is often made clearly visible by intenseemission from diatomic carbon, a solvent pyrolysis product. In contrast, the downstreamboundary region (shown in gray) differs considerably from the upstream region. It may bedescribed as a tail flame.Specific regard was paid to the effects that solvent plasma load had on both thedischarge and the analytical performance of ICAP- AES. Solvent plasma load may be definedsimply as the amount of solvent delivered to the discharge per unit time [2],amount of solvent deliveredSolvent Plasma Load, QSPL. .umt timeIt may be conveniently expressed in units of miffigrams per second, mg/s, or micro moles persecond, jimolls. In spite of its simple definition, solvent plasma load complicates trace metalanalysis by ICAP-AES quite considerably. Nevertheless, solvent load is intrinsic toconventional sample introduction practices because analysts rely on the favorable samplingstatistics of homogeneous solutions. It is clearly beneficial to take a sample from a solventextract or an acid digest, and feed it directly into a spectrochemical instrument.Introduction 5The general thesis objective was to rationally develop the ICAP for spectrochemicalapplications that cannot avoid solvent plasma load. In order to define that objective moreclearly, we will now examine the problem of solvent plasma load in greater detail.1.3. An Overview of the Solvent Load Problem, from Sample Introduction to Signal DetectionBriefly, a nebulizer (shown in Figure 1.2.) converts sample solution into an fine mist ofaerosol droplets [3], [4]. At the tip of the nebulizer, the mist is known as the primary aerosol.After nebulization, the primary aerosol is swept through a spray chamber on an argon stream.As the aerosol travels through the spray chamber, its small droplets partially evaporate while thelarge ones collide with the spray chamber walls. In this way, the spray chamber acts as a cutofffilter for droplet sizes. It removes most of the droplets larger than a limiting diameter, usuallylOj.tm to 25 JIm, and allows the rest of the droplets to partially evaporate. The resultingsecondary aerosol leaving the spray chamber may then be swept through a desolvating devicesuch as the heater and condenser assembly shown in Figure 1.2. This device can remove excesssolvent so that the aerosol stream does not overload the ICAP with solvent mass later on.Beyond the desolvator, more argon or molecular gases may be added through a device similar tothe one shown. A number of other technologies may be incorporated to further modify theaerosol stream before it finally reaches the torch, the general idea being to tailor the aerosolsuch that it neither destabilizes the ICAP, creates excessive noise, nor interferes with thespectrochemical analysis. Nevertheless, many of the problems resulting from solvent plasmaload can be traced back to the aerosol modification step. Certainly there is much room forimprovement here.Introduction 6Once the aerosol finally reaches the torch, it is known as the tertiary aerosol. Thetertiary aerosol generally consists of argon, solvent vapor, a large number of aerosol dropletsless than 25im in diameter, and perhaps even some desolvated particles of sample material.The tertiary aerosol travels into the injector tube of the torch, and at the end of the injector tube,a nozzle constricts the gas flow and injects the aerosol stream into the ICAP, as shown in Figure1.2. The aerosol stream next travels into the hollow base of the toroidal induction region, at theupstream end of the discharge, and then follows the axial channel through the centre of thetoroidal induction region. In one capacity, the toroidal induction region acts as a cylindricaloven. Ideally, it rapidly heats the aerosol stream flowing through it, desolvates the aerosoldroplets and then vaporizes the desolvated particles of sample material. Further downstream,several processes transport mass, energy and electrical charge from the outer toroidal plasmainto the axial channel, thus imparting energy to the sample material. Eventually, the aerosolstream coalesces with the plasma somewhere downstream. Beyond that point, energetic speciesin the plasma collide inelastically with the analyte atoms, thereby ionizing and exciting them.Introduction 7Sample‘Uptake—0IPoint whereCoalesces with Plasma,SIInduction44 RegionIIIIIa.pr‘ICAPTorch—c.TertiaryAerosolInjectorTubeCondenserNebulizerSecondaryAerosolFigure 1.2 The path taken by the sample aerosol from nebulizer to plasma.Introduction 8Electronically excited analyte atoms and atomic ions then emit characteristic lineemission amenable to atomic emission spectrometry. Alternatively, the plasma may serve as anabsorption volume for atomic absorption spectrometry (ICP-AAS) or as an excitation volumefor atomic fluorescence spectrometry (ICP-AFS). Perhaps the most powerful method involvesskimming atomic analyte ions into a vacuum chamber for mass spectrometry (ICP-MS).Whichever method is used, the analytical information must be extracted from the small regiondownstream from where the aerosol has coalesced with the plasma and upstream from wherethe analyte ultimately flows out of the plasma and mixes with room air.From this description, one might expect the ICAP to perform satisfactorily as aspeetrochemical source. Presumably, the toroidal induction region imparts enough energy intothe aerosol stream to completely atomize, excite and ionize the analyte. Moreover, one wouldexpect the flicker noise in the background signal, which determines detection limits for ICAPAES, to be low for a discharge with a presumably stable flow field. One might even expectgood precision for replicate determinations, because the ICAP can be operated steadily andcontinuously, in contrast to d.c. arcs and graphite furnaces, which have short burn times andfinite heating cycles. One might further expect the chemical interferences to be minimalbecause the plasma presumably atomizes everything injected into it. For the same reason, onemight expect spectral interference from molecular band emission to be negligible. In short, onemight have inordinately high expectations for simultaneous, multielement, trace metal analysiswhen one uses the ICAP as a spectrochemical source.But in most practical situations, the samples one injects into the ICAP complicatematters considerably. When the sample aerosols injected into the discharge are suspended inslurries, dissolved in volatile solvents, have complex or unknown matrices, or contain largeconcentrations of concomitant solutes, then interference effects—both spectral and nonspectral—are encountered. For further details, Olesik has provided a comprehensive overviewof the interference effects afflicting both ICP-AES and ICP-MS [5j.Introduction 9In particular, solvent plasma loading may cause spectral and non spectral interferencesand further degrade the analytical performance by introducing noise. This becomes clear whenwe examine the analyte and background signals. First consider the background signal. Solventplasma loading may drastically increase the intensity of both atomic and molecular backgroundemission [6], [7], [2]. Solvent pyrolysis products, principally C2 and CN in the boundaryregions of the discharge and atomic carbon in the plasma region, are the sources of backgroundemission. Their complex spectra may overlap with analyte lines, and thus interfere withbackground subtraction. Moreover, intense background signals can degrade the detection limitsfor ICAP-AES.On the other hand, non spectral interferences may result from the effect solvent plasmaload has on the analyte signal. Obviously, solvent plasma load may lower the amount of energyavailable to the analyte. It is likely that the power required to desolvate aerosol droplets,atomize the solvent molecules and then excite the solvent pyrolysis products is supplied at theexpense of the power available to vaporize, atomize and excite the analyte. 10 to 100 W aretypically required to dissociate solvent molecules, while the other processes require far less (seeFigure 2.3). True, these powers are insignificant in comparison to the total power dissipated inthe discharge (500 to 1750 W), but they are quite significant in comparison to the smallpercentage of total power available to the sample (<10%).Alternatively, solvent plasma load may cause non spectral interferences by increasingthe amount of power available to the analyte. One way it can do this is by altering the geometryof the discharge. In general, any molecular material entrained from the aerosol channel into thecoolant gas will drastically alter the temperature profile over the toroidal induction region. Forexample, the plasma may shrink and grow hotter if solvent contaminates the coolant argonflowing into the plasma, because the molecular material increases the thermal conductivity ofthe plasma gas. Essentially, the gas of higher thermal conductivity cools the boundary regions.This lowers the electrical conductivity of the plasma boundary. Consequently, the volumeoccupied by the plasma shrinks while its power density increases in order to maintain theIntroduction 10overall power balance. As a result, the plasma power may actually concentrate towards the axisof the discharge. Then solvent loading would actually increase the power available to theanalyte. Alternatively, solvent material could simply increase the thermal conductivity of theplasma gas, irrespective of the discharge geometry, and hence increase the rate of transportingenergy to the analyte. Of course, all of these arguments are purely speculative at the moment,but they show that the non spectral interference effects associated with solvent plasma load arecomplex, and possibly unpredictable.Apart from causing spectral and non spectral interferences, solvent plasma loading mayalso introduce noise. One can expect signal noise when the physical characteristics of theplasma fluctuate in response to the vaporization of incompletely desolvated droplets, or whenthe overall solvent plasma load drifts. In short, solvent plasma loading introduces noise to boththe analyte and background components of the analytical signal.Summary of the Solvent Load ProblemThe problem of solvent loading is complex, poorly understood and warrantsinvestigation because it adds noise and leads to both spectral and non spectral interferenceeffects, thus degrading the accuracy, precision and detection limits of any spectrochemicalmethod that uses the ICAP for a spectrochemical source.1.4. Relevant Properties of the Sample AerosolFour physical properties of the sample aerosol have received a great deal of attention: 1.the solvent plasma load, 2. the distribution of solvent mass between the vapour and dropletphases of the aerosol, 3. the size distribution of the aerosol droplets, and 4. the physicalproperties and chemical composition of the solvent. Maessen et al. [2] described reliablemethods for determining both the total solvent plasma load and the distribution of solventbetween vapour and droplets phases. In fact, this thesis relies on much of their pioneering work.Canals et al. [8] provided insight into the droplet size distributions of solvent aerosols. Theirfindings indicate that the influence of desolvating droplets in the discharge may not beIntroduction 11significant when the ICAP is loaded with organic solvents. On the other hand, Farnsworth [9]and Olesik [10] have demonstrated that droplet effects are significant in ICAPs loaded withaqueous aerosols. Olesilc et aL [10], [11] demonstrated that although the overwhelming majorityof droplets in an aqueous aerosol are very small, a few large droplets—statistically few yetsignificant in mass—may survive the traverse through the toroidal induction region and on up tothe analytical viewing zone of the discharge. That brings us to the interface between the plasmagas and the sample aerosol.1.5. The Interface between the Sample Aerosol and the DischargeSeveral workers have looked beyond the properties of the sample aerosol and into theinterface between the sample aerosol and the plasma gas. In particular, [12], [111, [13], [10],[14] have studied the interface between the plasma gas and incompletely desolvated waterdroplets. It was found that desolvating droplets not only survive in the plasma, they createregions of localized cooling about 1.5 mm in diameter as they travel downstream. This was astartling revelation indeed, because it implies that the temperature profile of the ICAP fluctuatesas droplet disturbances flow by. It also calls into question the conclusions of many previousinvestigations which assumed that the solvent loaded ICAP had a steady temperature profile.There is another perspective largely ignored by the droplet investigators. They assumedthat the water loading was confined to the aerosol channel, a reasonable assumption if waterenters the ICAP predominantly as droplets rather than vapour. Droplets will follow the argonstream along the axial channel. But for many organic solvents the vapour phase predominates.In contrast to droplets, vapour mass can diffuse across streamlines and away from the axialchannel. Consequently, the distribution of solvent over the argon stream is an important solventload parameter, at least for volatile organic solvents. Browner [7] and Maessen [15] recognizedthis parameter in their work with organic solvents. Browner investigated it experimentally byvarying the auxiliary argon flow rate [7]. The effects of the distribution of solvent vapour overthe argon stream turn out to be critical to the solvent loading process. Further details may befound in subsequent chapters.Introduction 12Whereas the interface between aerosol and plasma gas has received concerted attentionfrom experimentalists only recently, the plasma gas has received experimental scrutiny fordecades.1.6. Relevant Properties of the Inductively Coupled Argon PlasmaIn characterizing the ICAP, the physical properties of interest in this work are the flowdynamics, the thermal state of the plasma gas, the transport processes in the discharge, and thecollisional radiative processes in the discharge.flow dynamicsThe flow dynamics in the ICAP and similar discharges have been investigated with highspeed photography [16], [17], anemometer probes (pitot tubes) [18], [19], [20], particle tracking[18], and analysis of temporally resolved emission [9]. However, most of the flow field remainsinaccessible to experiment for both fundamental and non fundamental reasons. Invasive probesdisrupt the flow stream, high temperatures of the plasma melt the probes and vaporize trackingparticles, and the intense emission complicates laser Doppler anemometry. Fortunately,computer simulations offer complete access to the flow field, as Patankar points out [211. Thesimulations results relevant to solvent loading are the flow structure they predict within theconfinement tube, where solvent distributes over the argon stream. Beyond the torch exit,however, the complex flow field has so far defied simulation. Nevertheless, insight into theflow field beyond the confinement tube may be found in the literature on axisymmetric jets andflames [22], [23]. Both of these flow systems resemble the tail flame of the ICAP in manyrespects. That insight, other experimental results, and the revelations of computer simulationare summarized in Figure 1.3.The recirculation eddy at the base of the discharge can mix solvent vapour from theinner aerosol stream into the outer coolant stream. Note that this eddy is not necessarily aturbulent phenomenon. Downstream from the eddy, the flow field develops a relatively flatvelocity profile, except for a central maximum. Particle tracking (of aerosol droplets) revealsIntroduction 13that the central flow velocity is approximately 25 mIs. Moreover, Reynolds numbers << 2000validate the assumption that the flow field here is laminar rather than turbulent. The Reynoldsnumber is defined as Re Lpv0/u,where L is the characteristic length of the structureconfining the flow, v0 is the centerline velocity, p is the density of the fluid, and u , itsviscosity. In the confinement tube of the ICAP, approximate values for these are 0.Olm, 10mlsec, 0.2 kg/rn3 and 2 x 10kg/msec, respectively [24], so 100 Re <<2000.The laminar flow field has both an axial and tangential component, the latter impartedby the tangential gas inlet for the coolant flow. The tangential component, or swirl, helps tostabilize the discharge. Simulations reveal that the swirl also concentrates the power densitytowards the axis [24]. This happens because the centrifugal moment of the swirl holds the bulkof the coolant stream against the confinement tube. As a result the outer boundary of theinduction region is kept cool, and both the electrical conductivity and power dissipation are keptlow beyond a certain radius, so the plasma is confined to a smaller radius than if swirl wereabsent. All of these flow field characteristics help us understand how solvent material can betransported through the discharge by convection.Beyond the exit of the torch, the flow field becomes far more complex. When theplasma jet flows out of the torch into the quiescent room air, the flow field is no longer boundedby the torch wall, but extends beyond into the argon stream. Where the flow field crosses fromthe argon jet into the air, there is a surface of discontinuity, or sudden jump between flowingargon and the air at rest (For clarity, we will ignore the fact that the surrounding air is actuallydrawn into the argon jet). Varicose instabilities form at this cylindrical surface of discontinuitybetween the argon jet and the quiescent room air. As these instabilities develop, theyIntroduction 14• EntrainmentDevelopment into aUnidirectional Flowfielcl,with both Axial andTangential VelocityComponentsRecirculation Eddy at theBase of the DischargeTangentialArgonInlet PortShedding RingVortexFigure 1.3 The flowfield in the inductively coupled argon plasma.Introduction 15modulate the diameter of the argon jet. Instabilities of this sort are familiar to anyone who hasseen the jumping orange flame of a Bunsen burner. As the varicose pulsations propagatedownstream, they roll up into ring vortices. Winge et al. provide high speed movies of thesestructures. Experimental evidence indicates that the varicose pulsations penetrate to the veryaxis of the discharge. Chapter 6 provides further details.The thermal state of the Plasma GasWe can largely understand the analytical performance of the ICAP in terms of a physicaldescription if we know the thermal state of the plasma and how it is modulated by dynamicprocess such as vortex shedding and aerosol vaporization. We have already discussed thesedynamic processes, and more will be said about them in Chapter 6. As for the thermal sate, itturns out that the thermal state of the ICAP approaches local thermal equilibrium. In localthermal equilibrium, the kinetics of the plasma electrons may be described by the Maxwellvelocity distribution. Moreover, the atomic state distribution functions for bound electrons maybe described by the Saha and Boltzmann distributions. All of this is discussed in greater detailin the introductory sections of Chapters 7 and 9. The reader is referred to those sections forfurther discussion of the thermal state of the plasma, and how it departs from local thermalequilibrium.Two classes of processes determine the thermal state of the plasma—collisionalradiative processes and transport processes. We now turn to them.Introduction 16collisional radiative balancesthe Saha BalanceThe thermal state of the plasma is dominated by the kinetics of the plasma electrons.Through inelastic collisions, free electrons rapidly ionize atoms and recombine with atomicions. In doing so, they rapidly redistribute the atomic species over successive ionization stages.Consider, for example, the balance between collisional ionization and the reverse process ofthree body recombination, between an excited atom and ground state ion. This balance betweentwo collisional processes may be writtenX+e+(I) S )X+e+e, (1.1)where X, is an atom of element X in excited state p, e an electron, X an ion in the groundstate, and I, the energy required to ionize state p. If this balance prevails, then it can bring theion ground state towards local Saha equilibrium (LSE) with the excited states of the atom. InLSE, the atomic state distribution function may be described by the Saha distribution. Moreprecisely, the density of state p in LSE or 1S(p), is determined by the electron density e theelectron temperature Te , and the ion ground state density jjfn( h2 I,‘i (p)=ij(l)—-j I exp—, (1.2)2 22rmekTe j kTewhere the state density i7 is defined as the level density per statistical weight, ij and thegother symbols retain their usual significance.Introduction 17the Boltzmann balanceNow consider balances involving a single ionization stage such as the neutral atom of anatomic species in the plasma. When free electrons collide with atoms and atomic ions, theyrapidly excite and de-excite the bound states through collisional excitation and de-excitation:Xq +e+(Epq)—B-->Xp+e, (1.3)Here, X is an atom in state q, which lies at a lower energy than state p’ and Epq is thedifference in excitation energy between the two states ( E = E — Eq ). If free electronsredistribute the bound states rapidly enough, then the atomic state distribution function willapproach local Boltzmann equilibrium (LBE). In this thermal state, the excitation temperatureTex equals the electron kinetic temperature Te, and the atomic state distribution function maybe described by the Boltzmann distribution,-EjB(p)=j(q)exp pq (1.4)ewhere B (p) is the density of state p in LBE, and ij(q) is the density of state q.The Saha and Boltzmann balances are proper balances in which the forward collisionalprocess is balanced by the reverse process and both processes are dominated by the kinetics ofthe plasma electrons. These balances tend to bring the atomic state distribution functiontowards local thermal equilibrium by bringing the ASDF into LTE. In contrast, improperbalances are not dominated by electron kinetics. In fact they are controlled by energetic atoms,ions and photons. Moreover, the forward processes are not balanced by the correspondingreverse process and they cause the atomic state distribution function to depart from localthermal equilibrium. Perhaps the most important improper balance is radiative de-excitation.introduction 18radiative de-excitationx RDX+hq (1.5)X ‘ X3+hvHere, atoms and ions in upper states p and r spontaneously emit a photon. As a result, they arede-excited from upper states p and r to lower states q and s, respectively. It has been suggestedthat this improper balance provides the dominant pathway for atomic species to depart fromLBE.transport processesTransport processes can upset the Saha and Boltzmann balances outlined above, andprovide a path for further departures from local thermal equilibrium. Chapters 7 and 8 discusstransport processes in greater detail. Fey provides a lucid summary of the transport processeswhich predominate in the ICAP—energy input, particle transport, escape of radiation and heatconduction [25]. Cambel provides a more general introduction to transport phenomena inplasmas in Chapter 7 of his text on plasma dynamics [26]. Dresvin’s text is also worthconsulting for lucid descriptions of plasma phenomena [181.Two transport properties of the plasma gas are particularly relevant to this work—electrical conductivity and thermal conductivity. They appear to govern how the inductionregion of the plasma responds to solvent loading. The reader is referred to Chapter 8 for furtherdetails.Introduction 191.7. Thesis Objectives and SummaryWhen all of the relevant properties of the plasma are put together—the flow dynamics,the thermal state, sundry dynamic processes, collisional-radiative processes and the transportprocesses—one is left with an incomprehensibly complex physical system. Indeed, Feynmanremarked that only a few simple physical processes are required to create an incomprehensiblycomplex system. In fact, the challenge facing experimenters today is not to discover thefundamental mechanisms, because they are already known for the most part, but to understandthe complexity of their interaction. The same may be said of the solvent loaded ICAP. Wealready know the fundamental mechanisms at work in the discharge, but we do not quiteunderstand the complexity with which they interact. We cannot even be sure which onespredominate, at least not for a wide range of operating conditions. So the basic challenge facingus is to understand the complexity of the inductively coupled argon plasma, the complexity ofits temporal behavior, its spatial structure, and in this thesis work, the complexity with which itresponds to loading by various solvents.In order to face that challenge, a systematic strategy was eventually devised forinvestigating the solvent loaded, inductively coupled plasma. The first step of the strategy wasto survey the complexity in physical space, parameter space and time. The idea was that thepreliminary survey of the complexity would direct and facilitate investigations of the physicalproperties of the discharge. Once the physical properties had been characterized, at least theones relevant to the problem of solvent loading, then we would have the most rationalperspective from which to improve the design and methodology of analysis techniques whichuse the solvent loaded ICAP as a spectrochemical source. The objective of this thesis wassimply to apply this strategy to a very specific ask—to develop the inductively coupled argonplasma for applications in which solvent loading imposes an obstacle to trace metal analysis.In fact, the next eight chapters follow the strategy step by step. Chapter 3 reportsdetailed observations of how the ICAP responds to loading by water, chloroform and methanol.In this way, it provides a preliminary indication of the complexity of the discharge, as far as theIntroduction 20eye can discern. It even anticipates a few of the physical properties that will be revealed in thesucceeding chapters. Chapter four then surveys the complexity with which the analytical signalresponds to solvent plasma load. Chapter 5 and 6 explore the spatial and temporal complexityof the discharge, and together with Chapters 3 and 4, offer valuable guidelines for exploring itsphysical properties. Chapters 7 through 9 then explore the physical properties. Chapter 7discusses the rationale behind exploring the electron densities of the plasma gas, then reports theresults of an extensive electron density survey. The electron densities in this survey weredetermined from the absolute intensity of a highly excited argon line, and provide much insightinto how the physical properties of the discharge responds to solvent plasma load. Chapter 8presents more accurate electron densities determined from the Stark broadening of a hydrogenline. The spatially resolved profiles of electron density presented in this chapter reveal how theinduction region responds to solvent loading, and sort out many of the contradictory resultsreported in the literature. But Chapter 8 still leaves ignorant of how the bound states of theanalyte atoms are excited. That question is addressed in Chapter 9, which rigorously examineshow the atomic state distribution function for iron responds to solvent plasma load. Finally,Chapter 10 summarizes all of the conclusions, assesses the effectiveness of the strategyoriginally proposed, and suggests directions for future work in this area.1.8 References1. Boumans, P.W.J.M., SpectrochimicaActa, Part B, 1991. 46B(617): p. 711-739.2. Maessen, F.J.M.J., G. Kreuning, and I. Balke, Spectrochimica Acta, Part B, 1986. 41B:p.3.3. Sharp, B.L., Journal ofAnalytical Atomic Spectrometry, 1988. 3: p. 613-652.4. Browner, R.F. and A.W. Boom, Analytical Chemistry, 1984. 56(7): p. 787A-798A.5. Olesik, J.W., Analytical Chemistry, 1991. 63(1): p. 12A-21A.6. Pan, C., G. Zhu, and R.F. Browner, Journal ofAnalytical Atomic Spectrometry, 1990. 5:p. 537.7. Pan, G., G. Zhu, and R.F. Browner, Journal ofAnalyticalAtoinic Spectrometry, 1992. 7:p. 1231-1237.8. Canals, A. and V. Hernandis, Journal ofAnalytical Atomic Spectrometry, 1990. 5: p. 61-Introduction 219. Cicerone, M.T. and P.B. Farnsworth, Spectrochimica Acta, Part B, 1989. 44B: p. 897.10. Olesik, J.W., L.J. Smith, and E.J. Williamsen, 1 sheet. Anal. Chem., 1989. 61: p. 2002.11. Olesik, J.W. and J.C. Fister, SpectrochimicaActaPartB, 1991. 46B(6/7): p.851-868.12. Olesik, J.W. and J.C.F. Ill, SpectrochimicaActa, Part B, 1991. 46B: p. 869-883.13. Olesik, J.W. and S.-J. Den, SpectrochimicaActa, PartB, 1990. 45B(7): p.731-752.14. Horlick, G. and F. Qin. . in Federation ofAnalytical Chemistry and SpectroscopySocieties Meeting XVII. 1990. Cleveland, OH:15. Maessen, F.J.M.J. and G. Kreuning, Spectrochimica Acta, Part B, 1989. 44B(4): p. 387-384.16. Winge, R.K., et aL, Journal ofAnalytical Atomic Spectrometry, 1988. 3: p. 849-855.17. Winge, R.K., J.S. Cram, and R.S. Houk, Journal ofAnalytical Atomic Spectrometry,1991. 6: p. 601 - 604.18. Donskoi, A.V., V.M. Goldfarb, and V.S. Klubnikin, Physics and Technology ofLow-Temperature Plasmas. English Edition ed. 1977, The Iowa State University Press. 471.19. Barnes, R.M. and R.G. Schleicher, Spectrochimica Acta, Part B, 1981. 36B: p. 81-101.20. Barnes, R.M. and J.L. Genna, Spectrochimica Acta, Part B, 1981. 36B: p. 299-323.21. Patankar, S.V., Numerical Heat Transfer and Fluid Flow. 1980, New York: McGraw-Hill.22. Becker, H.A. and T.A. Massaro, Journal ofFluid Mechanics, 1968. 31: p. 435-448.23. Dahm, W.J.A., C.E. Frieller, and G. Tryggvason, Journal ofFluid Mechanics, 1992. 241:p. 37 1-402.24. Benoy, D.A., Modelling of ThermalArgon Plasmas. Ph.D. Dissertation 1993, TechnischeUniversiteit Eindhoven:25. Fey, F.H.A.G., Excitation Balances and Transport in an inductively coupled Plasma.Ph.D. Dissertation 1993, Technische Universiteit Eindhoven:26. Cambel, A.B., Plasma Physics and Magnetofluidmechanics. McGraw Hill Series inMissile Technology, ed. H.S. Seifert. 1963, New York: McGraw Hill Book Company, Inc. 304.Chapter 2The Experimental SetupFigure 2.1 depicts the overall experimental setup used to study the effects of solventplasma load and other parameters of the inductively coupled argon plasma. The setup essentiallyconsisted of a sample introduction system (a-k), inductively coupled argon plasma (1), lightcollection optics (o,p), and a grating spectrometer (q—x).2.1 Sample IntroductionBriefly, a peristaltic pump (b) fed the nebulizer (e) with test solution (a) through line (d).The nebulizer generated an aerosol stream on argon which flowed through the spray chamber (g).The aerosol stream was then directed through tube (i) into a thermoelectrically cooled, desolvatingcondenser (j). The peristaltic pump drained the condenser through line (h) and the spray chamberthrough line (f) into the waste flask (c). Narrow bore, Intramedic ® tubing was used for all ofthese lines in order to minimize the overall dead volume. Because other avenues of solvent losswere insignificant (e.g. leakage and evaporation), the weight difference between the test solutionand the waste solution determined the amount of solvent mass delivered to the plasma—thesolvent plasma load. Moreover, the solvent plasma load could be calibrated accurately andreproducibly against the condenser temperature. The aerosol gas flows were similarly accurate andreproducible. In most experiments, the primary aerosol gas flow was set at 0.61 11mm by the highpressure, cross flow nebulizer, while any extra gas fed through adapter (k) was controlledaccurately and precisely by a mass flow controller. The extra gas, or adgas, was argon in mostexperiments, but gases such as oxygen, nitrogen, and hydrogen could also be added. In fact, asafe, convenient way to add oxygen to a combustible aerosol stream was to add oxygen diluted inargon through the adgas adaptor (k). In short, the solvent plasma load, inner gas flow rate andinner gas composition could be controlled reproducibly and accurately and the major sources ofirreproducibifity in the experimental setup were eliminated.22(p)Figure2.1.Aschematicoverviewoftheexperiment.Shownarethesampleintroductionsystem(a-k),theinductivelycoupledplasmatorch(1),andthegratingspectrometer(q-x).ThetorchisshowninsideitsFaradayshieldortorchbox.Theshorterboxmountedalongsidehousestheimpedencematchingnetwork.Bothboxesaremountedonatranslationstage.Thestagewasdrivenbyacomputercontrolledsteppermotor(z),andwascapableof translatingtheentiretorch,matchingnetworkandsuspendedcondenser(j)intheXdirection,oracrosstheopticalaxis(intheZdirection).Ther.f.powersupply,detectorhardware,data-acquisitionhardwareandcomputerarenotshown.Seethetextforfurtherdetails.(q)xThe Experimental Setup 24nebulizersA MAK (Meddings, Anderson and Kaiser) nebulizer and MAK spray chamber were usedfor most experiments. The MAK high pressure, cross flow nebulizer [1] not only proved robustto rough handling, but delivered extremely stable and reproducible argon streams. For the MAKnebulizer used in this work, the inner argon flow rate was 0.61 ± 0.01 11mm, where the standarderror was principally due to the technique used to calibrate the flow rates (flow rates could bereliably calibrated by timing the passage of soap bubbles through a burette, downstream from thenebulizer).The experimental work also incorporated Meinhard concentric nebulizers with a Scott,double pass spray chamber. Work with these nebulizers was restricted to water loading studies.solvent plasma load: calibration and controlFor all of the organic solvents investigated, the solvent plasma load was calibrated againstthe condenser temperature using the method depicted schematically in Figure 2.2.b., thecontinuous weighing method devised by Maessen et al. [2] In this work, the method wasmodified slightly by equipping the sample and waste flasks with rubber septa and hypodemiicneedles. Essentially, the solvent plasma load was determined by monitoring the decrease insolvent mass. Maessen et al. [2] provide further details, including the appropriate intervals forsampling the decrease in solvent mass, and the determination of the error in the solvent plasmaload. Figure 2.2.a depicts a sample calibration plot for chloroform plasma load, showing a cubicpolynomial fit through the data by least squares regression. In general, the dependence of solventload on condenser temperature could be satisfactorily fit with a cubic polynomial. A sigmoidalfunction would not be suitable because the solvent plasma load actually began to increase at verylow condenser temperatures. Maessen et aL [21 explained this phenomenon by pointing out that atvery low temperatures the solvent vapour condenses on the aerosol droplets rather than thecondenser wall. Regardless of the true functional form of the solvent load curves, the experimentalwork did not rely on least squares interpolations. Instead, the solvent load was calibrated for a setExperimental25E0-Ja:0LI..0a:0-JIC.)condenserdrainFigure 2.2. (a) Calibration plot of chloroform plasma load versus the temperature of the aerosoldesolvating condenser. (b) Schematic flow chart of the continuous weighing method used tocalibrate the solvent plasma load against condenser temperature.-20 -10 0 10 20CONDENSER TEMPERATURE, CThe Experimental Setup 26of condenser temperatures, and those condenser temperatures were used to set the solvent plasmaload.The continuous weighing method was not suitable for measuring water plasma loadbecause large water droplets tended to cling to the condenser and spray chamber walls. Ratherthan adding surfactant to make the water drain as freely as the organic solvents, the water plasmaload was determined by trapping the aerosol at the exit of the condenser. A cold trap condensedthe vapour while glass wool trapped the droplets.Table 2.1 summarizes the maximum and minimum solvent plasma loads used in theexperimental work. Note that the table expresses solvent load in units of mass, moles of solventmolecules, moles of dissociation product atoms, and total bond dissociation energy, all per unittime. These units are somewhat helpful in interpreting the experimental results. The error in thetabulated values should be taken as ± 5%. While the continuous weighing method was capable ofdetermining the solvent plasma load within a standard error of± 2%, a conservative value of± 5%allows for long term drift.further aerosol modification: adgasIn order to vary the total inner argon flow rate, extra argon was added downstream fromthe condenser through an adgas adapter. In this way, the total inner argon flow rate could bevaried without perturbing the solvent load or the analyte transport efficiency through the nebulizer.The extra gas surrounded the primary aerosol stream as an annular sheath. Observations of laserlight (red helium neon) scattered off the aerosol confirmed that the annular sheath did not mix withthe central aerosol stream, in agreement with the low Reynolds number for the flow stream. As aresult, the aerosol test species were probably concentrated towards the center of the aerosol stream,while freely diffusing solvent vapour was probably distributed more evenly.In some experiments, a sheath of oxygen diluted in argon was added to the aerosol stream.It turned out that adding oxygen this way was both convenient and safe, in contrast to experimentsconducted by earlier workers. Interestingly, when oxygen diluted in argon was added in equimolarproportion27Table 2.1. Summary of the ranges of solvent plasma load investigated.BondMass Molar Carbon Hydrogen Oxygen Chlorine DissociationSolvent Load, Load Load, Load, Load, Load, Load,Q SPL Q SPL Q CPL Q HPL Q OPL Q C1PL Q dissocmg/s pmo1/s jimol/s jimol/s Jimolls !ImolJs Wattswater 0.10 5.6 — 11.2 5.6— 5.20.30 16.7 33.4 16.7 15.5m-xylene 0.20 1.9 15.2 19—— 14.80.45 4.2 33.6 42 32.8i-propanol 0.30 5.0 15.0 40.0 5.0— 22.01.0 16.7 50.1 134.0 16.7 73.5methanol 0.20 6.3 6.3 25.2 6.3 — 13.01.3 40.6 40.6 162.4 40.6 83.6carbon 1.3 8.4 8.4—— 33.6 11.0tetrachloride 6.3 40.9 40.9 164.0 53.2chloroform 3.0 25.0 25.0 25.0— 100.0 35.010.0 83.3 83.3 83.3 250.0 117.0The Experimental Setup 28to the carbon plasma load of xylene or chloroform—solvents that do not contain oxygen—thisresulted in a discharge similar to one loaded by methanol. For brevity, the results of those oxygenaddition experiments are not presented in this document.observations of the aerosol stream: the distribution of solvent plasma load between aerosoldroplets and vapourThe size distributions of the aerosol droplets and the transport efficiency of the test analyteswere not determined in this work. However, observations of the aerosol stream indicated that thetransport efficiency through the condenser was close to 100%, except at very low condensertemperatures, and that large droplets were not abundant in desolvated aerosols. Observations ofthe aerosol stream within the condenser indicated a high transport efficiency through thecondenser: The aerosol stream was clearly stratified and separated from the condenser wall. Itappeared as though the aerosol had imparted a static charge on the condenser tube. One couldsuppose that charged droplets had initially collided with the wall and stuck onto it. Then thecharged wall repelled all the following droplets of like charge. Hence, electrostatic repulsion keptthe droplets from colliding with the condenser wall, so the transport efficiency through thecondenser was kept high.Another set of observations gave some indication of the overall trends of the droplet sizedistribution. When the aerosol was sufficiently desolvated, the visibility through the aerosol at theexit of the condenser increased, presumably because the droplets had vaporized. In this case, themass of the vapour component of solvent plasma load undoubtedly predominated over dropletcomponent. Scattered laser light (from a red helium neon laser) helped in these observations,except when the fluorescence from red test analytes (such as iron acetylacetonate) swamped out thescattered light. In general, the observations indicated that the vapour component predominatedover droplets for most desolvated aerosols. Even so, a rigorous analysis of the droplet sizesshould be conducted, but is beyond the scope of this thesis.29Table 2.2. Summary of Experimental SetupICP Unit Plasma Therm 1CP2500Power supply Plasma Therm HFP2500FImpedance matcher Plasma Therm AMN-PS-1Incident power 1.00 to 1.50 kWReflected power 0 to 50WInduction Coil 3 turn coil,1 in. ID, 1/16.in. spacing between turns,1/8 in. OD copper tubing, 14C to 15C cooling waterTorch U.B.C. Low Flow, copied from MAK designArgon Flow RatesOuter 10.0 1/mm (needle valve control,rotameter reading)Intermediate 0.5 1/mm (needle valve control,rotameter reading)Inner 0.61 to 1.01 1/mm (0.61 1/mm nebulizer flow rateplus 0.00 to 0.40 1/mm adgas controlled by mass flowcontroller)Translation stage stepping motor driven (Superior Electric Slo-Syntype M062-TDO3), variable scan distanceN*0.0125mm lateral stepsSample Introduction System—suspended from translation stageNebulizer high pressure, cross-flow(MAK)Spray Chamber double pass (MAK)Desolvator variable temperature condenser(U-shaped, 1.3 m pyrex tube, 11 mm ID, cooled by 6thermoelectric coolers (Melcor model CP1.4-127-06L)backed by water cooled platesAdgas fitting inner tube (aerosol) 4mm ID,outer tube (adgas)llmmIDSample Uptake rate 1.0 mL/min (controlled by Gilson Minipuls 2peristaltic pump equiped with 2mm ID Isoversinic pumptubing—Mandel Scientific)Solvent load caffibration Organic solvents: continuous weighing methodWater: glass wool filter followed by cold trap30Table 2.2. (continued)Imaging lens Oriel model 41775 fused silica, piano-convex;centre thickness = 6.9 mm, edge thickness = 2.0 mm;diameter = 50.8 mm; radius of curvature= 68±7 mm;focal length (589nm)=l5Omm;back focal length 145.2 mmSpectrometerMonochromator im Czerny-Turner (Schoeffel-McPherson 2061)120 x 140mm, 1200 g/mm holographic grating,0.833 nm/mm reciprocal linear dispersion(Schoeffel-McPherson Model AH-3264)Detectors Photodiode Array DetectorsSpatial Domain: Reticon RL4096/20; 4096 pixels,7pmwide on 15 tm centres;Spectral Domain: Reticon RS 2048; 2048 pixels, 12.tmwide on 25 tm centresboth cooled to -15 C by thermoelectric coolers (MelcorCP14-71-1OL) backed by water cooled plates; drynitrogen purge prevented frostingPhotomultiplier TubeHamamatsu R955; Kiethley model 427 currentamplifier; Kepco ABC 1500 high voltage dc powersupplyfrradiance Standard General Electric tungsten iodine QL-10 lampData Acquisition Board RC Electronics ISC-16 board, sampling rate up to1 MHz for single channel operationSolvents BDH analytical grade chloroform, carbon tetrachloride,iso-propanol, meta-xylene, methanolTest Analytes Organic solvent solutions: 2,4 pentanedionates of iron,magnesium, calcium and manganese, ThiokolChemicals; note that the magnesium and calciumpentanedionates were only slightly soluable inchloroform, xylene and carbon tetrachloride.Aqueous solvent solutions: ammonium ferrous sulfate;(the water was boiled, then sparged with nitrogen todeoxygenate it and prevent oxidation of Fe2 to Fe3jTest Solution Concentrations lion solutions for Fe I and Fell measurements(Chapter 9) were made up to 1000 .tg/ml (0.0182 M).Other solutions were 5pg/ml (all concentrations refer tothe analyte metal).The Experimental Setup 31flush timeSeveral characteristics of the sample introduction system make it unsuitable for routinetrace metal analysis. Perhaps the most important drawback is the flush time, or the time requiredto completely flush all traces of a sample from the spray chamber and condenser. This timeexceeded two minutes, while flush times must be less than one minute for economical routineanalysis. It would not be difficult, however, to tailor the system for practical applications.However, no attempt to do so was made in this work.2.2 TorchThe aerosol stream was fmally directed into the ICP torch. This work relied on torches ofthe MAK, high efficiency design. Dimensions for this design are given elsewhere [3].2.3 Power SupplyDetails of the r.f. power supply, impedance matching network and induction coil aresummarized in Table 2.2. Briefly, the r.f. power at 27.12 MHz was operated between 0.75 and1.75 kW, with most experiments conducted at 1.00, 1.25 and 1.50 kW. Figure 2.3 reveals theflow of power to the ICAP discharge. The reader is referred to the discussion by Ripson and deGalan [4] for further details. Note that much of the power dissipated in the plasma, say 550 Watts,is spent heating the outer coolant stream. As a result, much of the plasma energy is carried awayby convection. This may be understood if one considers that the power dissipated in the inductionregion of the discharge (see Figs. 1.1 and 1.2) is rapidly transported as heat, outward to the outersheath of the coolant stream and through the torch wall by thermal conduction. Additionally, asmaller fraction of the total power, say 100 Watts, is lost from the induction as ultraviolet, visibleand infrared radiation. Ultimately, if one balances all of the avenues of power consumption asRipson and de Galan did, one finds that at an r.f. power of 1.25 kW, only 300 W is available toheat, vaporize, dissociate, and excite the sample material, including the analyte and the solvent.Note that the carrier gas must also be heated, excited and ionized to some extent. Of that 300 Wavailable, then, perhaps 100 W is required to heat the carrier argon, and another 100 W may be lostExperimental32<1 W anaMe line emissionL100 W air entrainment, emission from C2?100 ‘A’ solvent dissociationC100 W inner argonC300Wavailableto sample100W radiation from argon plasma550 W coolant argon and heatconduction through torch wall1250 W generator power300 W power loss to coiland matching networkFigure 2.3. The flow of power in a solvent loaded ICAP (adapted from reference [6]). Most ofthe power dissipated in the discharge is carried away by convection with the coolant flow orconducted out through the torch wall. A minor fraction is left to dissociate the solvent. Thesolvent bond dissociation load is appreciable in comparison with the power available to thesample.The Experimental Setup 33radiatively by the brightly emitting solvent pyrolysis products. That leaves 100 W to dissociate thesolvent material. Looking back at Table 2.1, one finds that the bond dissociation load (by far themost significant component of the solvent load, in terms of energy consumption) is comparable tothe power available. In that case, it is conceivable that the sample could be incompletely pyrolyzed.Clearly, Table 2.1 shows some values of bond dissociation load exceed the power available to thesample. It is useful to bear this in mind in the following chapters.2.4 Ignition ProcedureMaessen et aL [2] described an elaborate ignition procedure for ICAPs loaded with organicsolvents. Their procedure was designed to prevent carbon soot from forming on the walls of theirtorch during ignition. The plasma is ignited by first applying r.f. power to the three turn inductioncoil around outer tube of the torch. Next, the high voltage output from a Tesla coil is used to createan initial discharge in the argon stream. This initiates the ICAP by allowing the induction coil tocouple power into a conductive gas. Finally, the ICAP is sustained by the r.f. induction current,independently form the Tesla coil. However, problems arise during the initial ignition because theinitially unstable ICAP often touches the torch wall. If solvent material were present in the argonstream, then pyrolysis products could form soot on the torch wall during ignition. Any smallamount of soot acts as an anchor or nucleation site for further soot build, which eventually blocksthe argon flow. The answer is to prevent any soot buildup at all during plasma ignition.Maessen’s procedure involved cleaning out all of the solvent material from the sample introductionsystem. In the work presented in this dissertation, it was found that the ICAP could be easilyignited by simply turning up the adgas to extremely high flow rates, thereby diluting the aerosolstream with argon, and preventing any solvent material from interacting with the plasma duringignition.The Experimental Setup 342.5 Light Collection OpticsSpecifications of the lens used in this work are provided in Table 2.2. In general, it wasfitted with a vertical, 30 mm x 5 mm aperture to limit the acceptance angle in the lateral directionyet pressure the light throughput in the vertical direction. Figure 2.4(a—d) depicts some typicallight collection maps for the optical train. These maps were calculated with an exact ray tracingprocedure developed by Farnsworth et al. [51. In this work, the light collection efficiency wascalculated for the entire optical train, or from source to detector. Figure 2.4(a) depicts a map forunit magnification and no aperture fitted to the lens. Clearly, the light collection envelope issmeared out and would be unsuitable for collecting intensity profiles destined for Abel inversion.Figure 2.4(a) depicts a map for unit magnification and a 30 mm x 5 mm aperture, a configurationclearly suitable for Abel inversion. The same applies to Figure 2.4(c)—a map for 0.5Xmagnification (2 to 1 imaging) with a 30 mm x 10 mm aperture. In maps (a) through (c), the lightcollection efficiency has been integrated over a 2 mm vertical distance. Figure 2.4(d) shows thelight collection efficiency integrated along the line of sight. This map reveals that the verticalresolution is limited to ±0.3 mm, which was the vertical dimension of a diode batch on the verticalarray detector (w) in Figure 2.1. Note that these maps do not account for reflectance loss off thelens surfaces. Also note that chromatic aberration was corrected by adjusting the image and objectdistances according to the Lens Maker’s equation (see Chapter 5).Further experimental details are either summarized in Table 2.2 or provided later in thisthesis. The interested reader is urged to consult the introduction to Chapter 5, where severalspectroscopic imaging configurations are discussed.rCCD,-: r‘.-Cz•.CDCDC-.c,3Q,c;rUiCCCD —*CD0I—.CDCDCD4-04—’.-,.UiCD4-’.-4C4_-_zE. 4-’.--4CDoo0CDCDI.—CCDCDCt-4.—,—‘9ABSOLUTEEFFiCIENCY,iO<‘ABSOLUTEEFFICIENCY,io4.524667010EFFiCIENCY,io0jviuauuadx’The Experimental Setup362.6 References1. Anderson, H., H. Kaiser, and B. Meddings. Proceedings of the Winter Conference onPlasma Chemistry. 1980. San Juan: Heydon, London 1981.2. Maessen, F.J.M.J., G. Kreuning, and J. Bailce, Spectrochimica Acta, Part B, 1986. 41:p.3.3. Rezaaiyaan, R., etal.,AppliedSpectroscopy, 1982.36: p. 627-631.4. Ripson, P.A.M. and L.de Galan, Spectrochimica Acta, Part B, 1983. 38: p. 707-726.5. Farnsworth, P.B., B.W. Smith, and N. Omenetto, Spectrochimica Acta, Part B, 1990.45(10): p. 1151-1166.Chapter 3Spectral Survey and Observations3.1 INTRODUCTION3.1.1 ObjectivesEver since Greenfield [1] first used the ICP for spectrochemical analysis, a growingnumber of research groups have contributed their work to an ever expanding body of ICP-AESand ICP-MS literature. However, out of all of the spectroscopic and mass spectrometric studies,few authors have reported what their ICP discharge actually looked like when viewed throughwelder’s glass. Fewer still have described how their discharge altered in appearance when theyvaried its operating parameters. This general failure to report visual observations strikes one asabsurd when one considers the excellent case for reporting them in detail. That case is stated inthis chapter in an attempt to convince the reader that visual observations are an essential part ofdeveloping and characterizing any spectrochemical plasma.In our preliminary investigations of the solvent loaded ICAP, visual observations heldseveral advantages over spectroscopic measurements. Not only did observations yieldinformation more rapidly, they opened access to a far greater range of the operational parameterspace. For example, they opened access into regions of the parameter space where the dischargebecame unstable, far too unstable to measure ‘steady state’ spectroscopic quantities with theavailable hardware, yet stable enough to view its behavior through welder’s glass. Moreover, theentire discharge could be observed at once, including regions within the load coil (observed bylooking down into the torch). By contrast, these regions\ would have been hidden by the load coilObservations38when viewed side on, say for spectroscopic measurements. (In order to take advantage of thecylindrical symmetry of the discharge for calculating radially resolved intensities, spectroscopicmeasurements must be made side on.) Moreover, the three dimensional structure of severalvisible emission features could be perceived by viewing them from any angle. In short, visualobservations opened access to a greater range of the parameter space than that available tospectroscopic measurements, a greater range which could be surveyed more rapidly.In addition, by looking at the discharge through welder’s glass, one could judge thetemporal stability of the discharge or discern spatiotemporal behavior with time contants greaterthan lOOms. One could also check that the discharge was in focus for the spectroscopicmeasurements by checking that the spatial resolution of the spectroscopic measurements wereconsistant with the appearance of the discharge. Lastly, observations could often confirm that thespectroscopic measurements were realistic.Although there were several limitations intrinsic to visual observations, including thelimited spectral response of the human retina, observations revealed much about the physicalproperties of the discharge. For example, one of the most conspicuous emission features of anICAP loaded with organic solvent fell well within visible range of wavelengths--emission fromdiatomic carbon. The parametric behavior of this emission feature proved informative indeed. Itrevealed where sample atomization approached completion as well as indicating where theboundary of the atomic plasma resided. This and all of the advantages of visual observation citedabove made it indispensible to our investigation.One fmal reward of noting and reporting obsevations is worth mention: Carefully reportedobservations provide researchers with a spatial reference frame for comparing their ICPs withthose from other laboratories. Koirtyohann demonstrated that a visually observable, definiteemission zone within the structure of the discharge could be reliably used as a spatial referencepoint for interlaboratory comparisons [2], [3].The objective of this chapter was to take full advantage of visual observations in order toextract as much information on the behavior and physical characteristics of the solvent loadedICAP as possible. In order to achieve these objectives, literature reports of ICAP observationsObservations39were surveyed. Also surveyed were the backround emission spectra over the visible region inorder to identify the spectral components under observation (as well as to present a survey ofspectral features, to uncover the ones which may be useful and most amenable for spectroscopicmeasurements). Finally, the detailed observations made during this thesis work were compiled.The reports of these observations have been carefully illustrated by photographs and detaileddrawings.3.1.2 Literature SurveyBoumans et al. [4] described general appearance of an ICAP loaded with methyl isobutylketone. Their sketches roughly portray the spatial relation between the boundary regions of thedischarge where molecular emission predominates, and the plasma region where atomic emissionpredominates. Moreover, the response of these structures to the variation in operating parametersis described. Overall, the observations reported by these authors are consistent with thosereported in this chapter.Mixed Gas and Molecular Gas ICPsMuch literature reports visual observations of mixed gas ICPs, and ICPs sustained bymolecular gases such as nitrogen and oxygen. This literature is pertinent to the analogous solventload problem; It offers physical explanations for the response of the ICP to variation of gasscomposition and flow rate. In special cases, these physical explanations may also explain theresponse to solvent loading. Lastly, the literature on mixed gas ICPs is pertinent to solvent loadedICAPs because it supplies a terminology for describing what the ICP discharge looks like.Truit and Robinson introduced a terminology for describing mixed gas ICPs [5], whichindeed proves convenient for discussing solvent loaded ICAPs. They described their mixed gasICP as consisting of three zones: the brilliant blue white, opaque core; the bright white,transparent secondary or transition region (weaker in intensity than the core); and the faint blue,transparent tailfiame. Their core (also known by other investigators as the energy loading region)resided within the load coils and assumed the shape of an annular cylinder or toroid. This regionObservations40appeared opaque because its emission was so intense that only a very bright object could bepercieved behind it. The transitition region appeared dimmer than the core, hence appearedtransparent—objects behind it could be readily percieved. (This region is also known by otherauthors as the decay region, under the assumption that energy dissipation from the plasmaoutweighs energy loading there.) Finally, the tailfiame capped the secondary region, and was notstrictly part of the plasma at all. It was actually a boundary region where air was entrained intothe plasma gas resulting in molecular emission and weak atomic emission. The tailfiame wasfaint, transparent violet.Truit and Robinson also conducted a spectroscopic study of an ICP into which they hadintroduced organic compounds [6]. They supplied a typical emission spectrum from such anICP, but they did not go into extensive detail in reporting their observations beyond describing theoverall structure of the discharge.Montaser, Fassel and Zalewski supplied a photographic record of their observations of anargon - nitrogen ICP supplied with various proportions of the two gases, sustained at severaldifferent powers [7]. They compared the effect of introducing nitrogen as a coolant (outer flow)with the effect of introducing it through the aerosol injector tube (inner flow). When nitrogen wasintroduced through the injector tube, the central channel increased in diameter, while the peripheraldiameter remained constant, so that the plasma formed a thinner annulus of constant outerdiameter. Introducing yttrium revealed the analyte distribution within such a discharge (blueemission from ions, red from the diatomic oxide): The analyte distribution did not expand indiameter along with the central channel. This implied that nitrogen introduced through the injectortube had degraded the interaction between the sample and the plasma. On the other hand, whennitrogen was introduced as the outer flow, the discharge assumed the shape of an inverted cone:its diameter was constricted within the load coil, but expanded radially further downstream. Theyexplained this observation in terms of a thermal pinch, originally suggested by Thorpe [8] andexplained in detail by Greenfield et al. [9]. Apparently, this thermal pinch enhanced theinteraction between the analyte and the plasma.Observations41Briefly, the thermal pinch effect depends on the thermal and electrical conductivity of theplasma gas and their effects on energy loading into and energy dissipation out of the plasma.When a plasma gas containing molecular species (including solvents and diatomic gases) reachesa sufficient temperature (at a given pressure) the bulk of molecular constituents dissociate, and theenthalpy of their dissociation not only cools the plasma gas, but increases its thermal conductivity(the thermal conductivity of nitrogen increases from approximately 10 times that of argon to 36times that of argon when the temperature increases from 5000K to 7000K). This increase inthermal conductivity accellerates heat conduction away from the plasma, especially across thesteep thermal gradients at its boundary. This accellerated heat loss rapidly cools the peripheralregions of the plasma volume. As the peripheral regions cool down, they lose their electricalconductivity, causing the plasma volume to shrink But the overall power loading into the plasmatends to remain constant. In order to keep its total power loading constant as it shrinks in volume,the plasma reacts by increasing its power density. This process eventually results in a morecompact, hotter plasma, able to maintain a stable balance between energy dissapation and energyloading. The contraction is known as the thermal pinch effect.As mentioned above, the thermal pinch effect manifested itself in two ways in mixed gasICPs. In one way the plasma cantracted into an inverted cone when the molecular gas wasintroduced via the coolant or outer flow. In the other way the central channel expanded, resultingin a thinner plasma annulus when the molecular gas was introduced via the carrier stream or theinner flow.As a digression, it is interesting to note that Montaser et al. ‘s photograph of an argon ICPloaded with nitrogen through the injector tube appeared very similar to an argon ICP (or ICAP)loaded with methanol, or very similar to an ICAP loaded with other organic solvents to whichoxygen had been added to the carrier gas in stoichiometric proportions to form carbon monoxide(see chapter 10).Greenfield et al. have provided the most careful analysis of visual observations for mixedgas ICPs [1], [9]. Here, we will only note that they projected an image of the discharge ontograph paper in order to trace the boundaries of plasma region. From such geometric records, theyObservations42produced a table of plasma dimensions for different flow rates of molecular gas added to thedischarge.3.2 Spectral SurveyThe line of sight emission from a solvent loaded ICAP was surveyed over visiblewavelengths, for several observation heights, and for loading by three different solvents: water,methanol and chloroform. The results of this survey are presented in Figures 3.2.1, 3.2.2, and3.2.3. In each figure, all of the spectra (each corresponding to a different height above the loadcoil) were scanned simultaneously by a linear photodiode array mounted vertically, to sampleemission from a range of observation heights. The resulting emission survey made it possible toidentify the conspicuous emission features visible to the observer’s eye, and provided a survey ofhow these emission features depended on the observation height. (The emission survey alsorevealed which emission signals would be most amenable to spectrocopic measurement byrevealing their relative intensity, their signal to background ratio, and their freedom from spectralinterference.)ObservationsO.4_____ _______0.4j___EL”____400 440 480 520 560 600 640 680 720Wavelength, nanometersFigure 3.2.1.Spectral survey of the visible emission from theinductively coupled argon plasma loaded with methanol.The spectra were recorded simultaneously at the different heights,6 mm, 9 mm, 12 mm, 15 mm, 18 mm, and 21 mmabove the load coil, as described in the experimental section.44Observations0.6-o.4*Ji0.2-0.0- I I I IIlBmmIJ I0.6-0.4-0.2-0.0-0.6-0.4-0.2-0.0-0.6-0.40.2-0.0-0.6-0.4-0.2-0.0-0.6-0.4-0.2-0.0-JLmJJ400 440 480 520 560 600 640 680 720Wavelength, nanometersFigure 3.2.2. Spectral survey of the visible emission from theinductively coupled argon plasma loaded with water.Observations 450.6-0.4-0.2-0.0-0.6-0.4-0.2-0.0-0.6—0.4-0.2-0.6-0.4-0.2-0.0-0.6-0.4-0.2-0.0-0.6-0.4-0.2-0.0-Wavelength, nanometersFigure 3.2.3 Spectral survey of the visible emission from theinductively coupled argon plasma loaded with chloroform.121mm1 L. I . I . . I IILLtrt— I I I I I I I • I I I IILU JhzILLJJI I • I • I • I • I • I • I • IjFjj400 440 480 520 560 600 640 680 720—.-C CCDCD‘—.C)CD C rCD—-“CDCD•C,, CD Cl)CD CD D CD DCD C 3 CDCD CD CD C 0 CD CDRelaliveSensitivityCCCCCCCCC—•4•L•ooccICC CD——CObservations47A comparison of the spectral survey, in Figures 4.2.1 through 4.2.3, with the photopic(bright light adapted) sensitivity of the eye, shown in Figure 4.2.4, reveals which emissionfeatures are accessible to observation. The light adapted eye is most sensitive at 555-nm, in theyellow-green region. Its sensitivity drops below 1% of this maximum value towards the violetwavelength of 430-nm and towards the red wavelength of 690-nm [10], [11]. If we let 1% ofmaximum sensitivity define the limits of the visible region, then comparison of Figures 3.2.1through 3.2.3 with 3.2.4 reveals that green emission, from diatomic carbon, and white emission,from atomic argon line and continuum emission, should be visible. However, the photopicsensitivity only drops off asymptotically at the violet and red extremes, down to 0.01% at 370-nmand 760-nm [12]. Consequently, very intense emission, say from the CN bandhead at 388 nm,is observable within these 0.01% limits.In short, the most conspicuous visible emission features are 1) band emission, fromdiatomic carbon and cyanyl radicals in boundary regions of the discharge, 2) line emission fromatomic carbon and argon and 3) the ubiquitous continuum emission from the atomic plasma. Alsoconspicuous is incandescent emission from soot particles, but this is not evident in the spectralsurvey. Of the three solvents surveyed, chloroform loading resulted in the most intense diatomiccarbon emission, while both chloroform and methanol loading resulted in intense cyanyl radicalemission. By contrast, diatomic carbon emission from the methanol loaded ICAP was weak,presumably because carbon monoxide formation competed with diatomic carbon formation. Inthis case, it is likely that the concentration of carbon monoxide predominated over diatomic carbonbecause of the higher bond energy and greater stability of the carbon monoxide molecule. Ofcourse, niether cyanyl radical, diatomic carbon or atomic carbon emission were observed forwater loading.In general, these emission features displayed four distinct trends for the dependence ofintensity on observation height: Atomic line emission from carbon and argon and the plasmacontinuum emission decreased monotonically with observation height. This makes sense becausetheir emission originated from the atomic plasma volume. Their decrease in intensityObservations48with height is consistent with the decay and extinction of the plasma as it flowed out of theinduction region into cold air downstream. In contrast, emission from the cyanogen radicalincreased with observation height, consistent with its formation by air entrainment into the plasmagas (molecular nitrogen from the air combining with atomic carbon from the plasma). Theemission from diatomic carbon displayed more complex behavior. It began at a low intensity, atlow observation heights, then proceded through a maximum further downstream, ultimately todissappear almost completely beyond a certain height. Because diatomic carbon emissionoriginated from the interface between the aerosol channel and the atomic plasma, its axial trendmay be reconciled with the structure of the upstream boundary region. This region was a hollowcylinder capped by a bullet of intense green emission. Once the solvent material had flowed pastthe top of the green bullet and then into the atomic plasma, the intensity of diatomic carbonemission dropped to insignificant levels. Finally, the analyte emission began weak at lowpositions, then increased as energy is transfered in towards the axis from the toroidal plasma.When the analyte ultimately flowed out through the top of the plasma volume, the analyteemission decayed to zero because the analyte was no longer supplied with sufficient excitationenergy.3.3 PhotographsFigure 3.3.1 compares an ICAP loaded with two different settings of chloroform load.The image on the left depicts the lower load while the one on the right depicts a higher load. Notethat the inner plume of diatomic carbon emission extends farther downstream with increasingload. In contrast, the methanol loaded ICAP depicted in Figure 3.3.2 appears to rise up withincreasing load. It will be useful to refer back to both of these figures while reading the followingchapters.Figure3.3.1Photographsof achloroformloadedinductivelycoupledargonplasma.Thechloroformloadincreasesfromlefttoright.Figure3.3.2Photographsof amethanol loadedinductivelycoupledargonplasma.Themethanolloadincreases fromlefttoright.Observations513.4 Detailed ObservationsFigure 3.4.1 depicts a medium power ICAP operating under conditions of moderatechloroform load. This figure illustrates all of the components of the ICAP emission structure onewould likely encounter when observing an ICAP loaded with any of the solvents investigated inthis work. This and all of the following figures in this section depict the spatial structure of thevisible emission from the discharge as it would appear in cross section, so that the actual plasmamay be regarded as a body of revolution about the cylindrical axis shown.The black area in each figure represents the region of intense emission from diatomiccarbon (which appears brilliant green through the observation port). In some figures, horizontalbars instead of solid black represent the observed diatomic carbon emission, indicating that it wasconspicuously weak or diffuse rather than intense. Whether intense, weak, or diffuse, thediatomic carbon emission usually occupied the spatial structure similar to the one shown inFigure 3.3.1, which may be conveniently regarded as an outer cup joined concentrically at itsupstream end (or base) with a hollow, inner plume, which culminates in a bullet shaped tip asshown.Enveloped within the region of diatomic carbon emission sits the region of intensecontinuum emission and argon line emission. This region will be hereafter refered to as theatomic plasma, for convenience of discussion. The atomic plasma consists of three discernablecomponents: the plasma core, the bright secondary plasma, and the dim secondary plasma, fromall of which one observes continuum emission and atomic argon line emission to the exclusion ofmolecular emission. These regions typically form an annulus or toroid within the load coilswhich appears to coalesce into a cone further downstream, as shown in the figure. Althoughthese three regions of the atomic plasma are not strictly distinct, as they may all blend into onegradual transition, it is useful to invoke them for the convenience of discussion.52000LEGENDATOMIC PLASMAPLASMA CORE-j BRIGHT SECONDARY PLASMADIM SECONDARY PLASMABOUNDARY REGIONENTRAINMENT REGION,%1 OR TAILFLAMEDIATOM IC CARBONEMISSION REGIONINCANDESCENT RADIATIONObservationsFigure 3.4.1. Visually Observed Emission Structures of a Solvent Loaded ICAPObservations53A cross section through the plasma core is represented by the two oval regions of lightestgrey sitting side by side within the confmement tube of the torch. Enveloping the plasma core isthe bright secondary plasma, which in turn is enveloped or bounded by the dim secondaryplasma. Distortions of these two secondary emission regions of the atomic plasma, in responseto variation of the operating parameters and the solvent loading, are perhaps the most importantobservations to note in this chapter; They qualitatively indicate responses in the physicalproperties of the ICAP, responses which determine how energy is transferred to the analyte, andhence responses critical to the analytical performance of ICAP-AES. One further feature of theatomic plasma worth introducing here is the channel along the axis, through the centre of thetoroid. This will be refered to as the axial channel, and does not necessarily coincide with theaerosol channel or distribution of analyte injected into the discharge. Once again, the term, axialchannel, has been introduced for convenience of discussion.The region enveloping the downstream cone of the atomic plasma represents theentrainment region or tailfiame of the discharge. Weak violet emission from the air entrainmentregion, from species such as CN, was clearly visible from this region.One additional emission structure may be observed under certain operating conditions: ahollow cone of incandescent emission, possibly consisting of glowing carbonaceous sootparticles, is often found nested within the hollow axial plume of diatomic carbon emission. Thesolid curve nested underneath the discharge represents the hollow cone of incandescent radiation.This emission feature was usually observed for high levels of loading by solvents with excesscarbon relative to their oxygen content. Significantly, the bright orange cone was never observedfor methanol loading or for ethanol/water mixtures, nor was buildup of carbonaceous soot on thetorch wall ever a problem for these solvents. In both cases, carbon and oxygen were present instoichiometric proportions. On the other hand, build up of carbonaceous soot accompanied theappearance of the hollow incandescent cone for relative high loading of solvents such as xylene,chloroform and hexane.The following figures illustrate how the emission structure depicted in Figure 3.4.1 variedwith solvent, solvent load, inner argon flow rate and forward power. The behavior depicted inObservations54these figures includes the response of the plasma volume (owing to the thermal pinch effect),vertical translation and vertical contraction of the plasma, the spatial structure of the diatomiccarbon emission plume (including nested cone structures) and the behavior of the normalanalytical zone (at the apex of the plasma decay region).Solvent ComparisonIn order to describe the behavior and emission structure of an ICAP loaded with any ofthe solvents investigated in this thesis work, it is sufficient to consider the distinctive behavior andemission structure that resulted from loading by only three of them: water, methanol andchloroform. Representative observations for ICAPs loaded by these three solvents are depicted inFigures 3.3.2(b—d); Figure 3.3.2(a) depicts representative observations for a pure argon ICAPflowing into air as a basis of comparison.In all four discharges, the atomic plasma assumed the geometry of a torus whichcoellesced downstream into a cone. Within this general geometry, one may distinguish theatomic plasma regions of the four figures by the extent that their central channel has dilated, howfar up through the load coil the plasma has translated, and how much the downstream portion ofthe secondary plasma has bloomed open.tn IntFigure3.4.2Representativeobservationsofaninductivelycoupledargonplasma(ICAP)dischargea.withoutsolventload;andloadedwithb.water, c.methanol,andd.chloroform.C Ia.c.UiObservations56In response to loading by water, the atomic plasma region did not appear to have dilatedor to have translated downstream to any great extent when compared with the pure argon ICAP(without solvent loading). However, upon close inspection, one may note that the central channelbecame less difuse and that the atomic plasma appeared to translated upwards approximately0.5 mm to 1.5 mm. The same minor changes with respect to the pure argon ICAP may be notedin response to chlorform loading. In contrast, the response of the plasma region to methanolloading was more pronounced. In this case, the atomic plasma clearly translated downstream, farenough, in fact (under the moderate conditions listed in table 3.3.1), for its base (or upstreamedge) to reside the middle and upper turns of the load coil. Also in response methanol loading,the central channel of the ICAP appeared to dilate, strongly indicating that methanol loadingresulted in a thermal pinch effect.Beyond the characteristics of the atomic plasma region, no further characteristics werefound to distinguish a pure argon ICAP from a water loaded ICAP. In contrast, both themethanol and chloroform loaded ICAPs exhibited brilliant plasma boundary regions owing toemission from solvent pyrolysis products. The chloroform loaded ICAP displayed brilliant greenemission from a sharply defined spatial structure described previously as an outer annulus joinedat the upstream end to a hollow, inner plume. For an indication of how sharply defined thisemission structure was, the cylindrical wall of the inner plume was often less than 0.5mm thick,while the boundaries appeared perfectly sharp. (Incidently, the small barbs of diatomic carbonemission on the downstream end of the outer cup are not artifacts of the figure, but were observedreproducibly, and were clearly visible. They probably indicate the presence of a back eddy in theouter gas stream, just beyond the exit of the torch. In contrast, the methanol loaded ICAPdisplayed relatively dim green emission from a diffusely defined spatial structure. Theboundaries of this diffuse structure gradually faded over a distance of approximately onemillimeter. However, under operating conditions of high methanol loading or low power, themethanol loaded ICAP exhibited a brilliant green, sharply defined structure similar to that of thechloroform loaded ICAP. This sharply defmed structure invariably nested within the diffuseObservations57structure. The methanol and chloroform loaded ICAPs also differed from the pure argon andwater loaded ICAPs by displaying weak violet emission from their tailfiames.In general, the appearance of the discharge depended on the relative amounts of oxygenand carbon in the aerosol stream. For example, xylene, propanol, and hexane loaded ICAPs weresimilar in appearance to a chloroform loaded ICAP. On the other hand, an ICAP loaded with anethanol water mixture was similar in appearance to an ICAP loaded with methanol. Interestingly,an ICAP loaded with xylene, but with oxygen added to the aerosol stream in equimolarproportions to the solvent carbon,.was also similar in appearance to an ICAP loaded withmethanol. Because of this general dependence on the relative amounts of carbon and oxygenload, the following discussion will be confined to chloroform, methanol, and water loadedICAPs, and their response to solvent load, power, and gas flow rates. The appearence of anICAP loaded with any other solvent, solvent mixture, or combination of solvent loading andoxygen addition, may be regarded as a hybrid or intermediate of the methanol and chloroformloaded ICAPs.Response to Water Plasma LoadIn comparison to its response to other solvents, the ICAP appeared to respond veryslightly to water loading (For this reason and because no molecular emission revealing thelocation of the plasma boundary region was visible either, the response of the ICAP to waterloading has not been illustrated here). In fact, the atomic plasma region appeared to translatedownstream through the load coil by only 1mm when the water load was increased from itsminimum attainable level (0.15mg/si) to its maximum attainable level (0.3Omgis). Whether ornot this was a downstream translation or a contraction of the atomic plasma along the direction offlow owing to the thermal pinch effect was not clear from observations alone. Perhaps theplasma was swept downstream or lifted off the tulip of the torch by expanding water vapor andvaporizing droplets near the base of the plasma[13]. Whatever the reason, the central channelbecame perceptibly darker and more clearly defined in response to water loading. On the wholethe ICAP appeared to be relatively insensitive to water loading, an insensitivity which may beObservations58explained by the characteristically low mass loading of water compared with other solvents whennebulized by conventional pneumatic nebulizers. Typically, the maximum water load thatpneumatic nebulizers are capable of delivering to the ICAP is less than 0.35mg/s. One shouldnote that ultrasonic nebulizers are usually capable of delivering much greater water loads to anICAP than pneumatic nebulizers. However, they must be fitted with some sort of desolvationdevice such as a condenser in order to desolvate the aerosol and reduce the water load before theaerosol stream reaches the plasma. Otherwise, one might anticipate that the plasma could beswept too far through the load coil and become unstable. The extreme water load would thenlikely extinguish the discharge.Response to Methanol Solvent LoadA far greater range of methanol loading was accessible to observation. How the ICAPresponded to methanol loading is illustrated in Figures 3.4.3(a) 3.4.3(b) and 3.4.3(c). They depicttypical observations for an ICAP loaded with the minimum obtainable, intermediate andmaximum levels of methanol loading. The most obvious responses were the way the diatomiccarbon plume extended along the central channel as the methanol load increased, and the way theatomic plasma apparently translated. In addition to its downstream translation, the atomic plasmacontracted in the direction of flow with increasing methanol load, a contraction which may haveresulted from the thermal pinch effect.Figure3.4.3RepresentativeobservationsforanICAPloadedwithmethanolata.minimumobtainable,b.intermediate,andc.maximumtolerablesolventload.0 I0 0 001Hr 0a.C.UiObservations60The obvious translation and contraction are accompanied by a more subtle response: Thesecondary plasma appears to bloom open in response to an increase in the methanol load. InFigure 3.4.3(a), the bright secondary plasma (light grey) retains the characteristic shape of a toroidcapped by cone. Then as the methanol load increases from the lowest attainable to theintermediate level, the cone of the bright secondary plasma is almost completely penetrated,leaving only a thin arch near its apex, as shown in Figure 3.4.3(b). In response to highermethanol loading, the apex of the bright secondary plasma blooms open. The dim secondaryplasma appears to bloom open in a similar manner, but a step behind the bright secondaryplasma.This blooming holds profound implications for the analytical performance of ICAP-AES:It reveals that methanol loading drastically alters how much energy flows into the central channelfrom the energy dissapation region or plasma core, energy required to desolvate, vapourize,atomize, ionize and excite the analyte. At maximum methanol load, the discharge may beregarded as having had bloomed open. It effectively retracted from the analyte. As a result, theplasma interacted incompletely with the analyte, if at all. At the other extreme of minimummethanol loading, one would expect the plasma to have interacted or have supplied energy toanalyte quite effectively.Directly linked with the blooming of the secondary atomic plasma is the behavior of thediatomic carbon emission: As the secondary plasma bloomed open, the green, C2 plumeextended downstream. In general, the cup and plume extended further downstream withincreasing solvent load.Observations61Response to chloroform solvent loadFigures 3.4.4(a), 3.4.4(b), and 3.4.4(c) illustrate typical observations of how the ICAPresponded to the variation of chloroform loading. In a manner similar to methanol loading,chloroform loading also caused the the diatomic carbon emission to extend further downstreambut several other characteristics distinguished the response of the ICAP to chloroform loading.In contrast to methanol loading, chloroform loading did not cause the atomic plasma to translatevery far downstream, and no obvious thermal pinch effect was observed. Moreover, the diatomiccarbon emission structure was always appeared sharp. Nested within the sharply defined centralplume, a bright orange, hollow cone of incandescent soot particles formed at high chloroformloads, something never observed for methanol loading. This hollow orange cone became brighterwith greater chloroform loading and appeared to nest closely within the central plume of diatomiccarbon emission, as shown by the continuous solid curve in Figure 3.4.4(c)Because the spatial structure of diatomic carbon emission was so sharply defined in achloroform loaded ICAP relative to the same structure in a methanol loaded ICAP, several subtlespatial responses to chloroform loading may be noted, and are depicted in Figure 3.4.5. The mostremarkable response is the way the central plume changed shape as it extended up through thecentral channel of the discharge. At minimum chloroform loading, the hollow inner plume canbe regarded as a cone. At intermediate solvent loads, the base of the plume remainedapproximately conical, but the tip of the plume extended downstream to form a cylindricalannulus capped by a bullet shaped region. At maximum chloroform loading, the top of thecylinder dilated to give the plume a bulbous end. Accompanying this extension and expansionwere changes in the thickness of the wall of the hollow plume. As the plume extendeddownstream, the wall of its leadind edge appeared to grow thicker, i.e. thicker in the apparentdirection of gas flow, whereas the thickness of the walls in the radial direction appeared to remainconstant.0 0 0Figure3.4.4RepresentativeobservationsforanICAPloadedwithchloroformata.minimumobtainable,b.intermediate,andc.maximumtolerablesolventload.r 0 0I0 0HIWIa.b.HIC.Observations63It is likely that the steepness of thermal gradients across the wall of the plume determinedits observable thickness; Diatomic carbon is probably only stable over a relatively narrowtemperature range, or a small distance along a steep thermal gradient. At temperatures above itsrange of stability, diatomic carbon tends to dissociate, or become unstable relative to atomiccarbon, while at temperatures below its range of stability, it tends to associate into polyatomiccarbon containing species. It follows that the cross section through the diatomic carbon plumemay be regarded as an isothermal contour in space, thinner when the thermal gradient crossing itis steeper. This explains why the wall of the plume varied in thickness—the side walls werethinner than the tip because the temperature gradient was steeper there.In Figure 3.4.5(a), the triangular cross section of the plume may be understood as adissociation front receding radially towards the discharge axis. The aspect ratio of the cone isdetermined by the gas flow velocity and the competitive rates of heat consumption by theenthalpy of dissociation and radial heat conduction towards the axis from the toroidal core. InFigure 3.4.5(a), heat conduction overtakes heat consumption before the gas flows out of the torch.In Figure, 3.4.5(b), they are nearly balanced. Figure 3.4.5(c) is more difficult to explain: Itappears as though heat consumption exceeds heat conduction, so that the solvent material is notcompletely pyrolyzed until very far downstream.Figure3.4.5Observationsofhjowtheshapeoftheinner plumevariedasthechloroformplasmaloadwasincreased.(a)Theshapeoftheinnerplumeatthelowestchloroformload,(b)atanintermediateload,and(c)atthehighestload.0 I(a)4......(b)(c)Observations65The outer or peripheral cup of diatomic carbon emission also displayed behaviordetermined by thermal gradients, gas flow patterns and heat conduction: It was thickest aroundthe base of the discharge and thinnest between the plasma torus and the torch wall, presumablyfor reasons similar to those determining the thickness of the walls of the central plume. However,it is questionable whether the outer cup can be regarded as a dissociation front similar to the innerplume. If it were, then solvent material would have to be swept around the base of the plasma,and enter through the outer periphery. Another, more realistic possibility, is that solvent materialwas folded into the outer argon stream by a recirculation eddy at the upstream edge of thedischarge (in the wake of the intermediate tube). In that case, solvent material could have beeneither swept around the base of the discharge or folded into the plasma, then transported to theperiphery by diffusion, (by either path reaching the region of the outer cup and forming diatomiccarbon, via dissociation of solvent molecules or via association of carbon atoms diffusing out ofthe atomic plasma.) In short, the behavior of the outer cup indicated how solvent material hadbeen distributed within the discharge:The observed response of a chloroform loaded ICAP to variation of the inner argon flowrate provided further insight into how solvent material might be distributed in the discharge.These observations are illustrated by Figures 3.4.6(a—c). All three frames in this figure depict anICAP loaded with an intermediate amount of chloroform, but at low, moderate and high innerargon flow rates (the inner argon flow rate was adjusted indedendently of the solvent load byadding adgas to the aerosol stream immediately before it was sent into the torch—see Chapter 2).As shown in the figure, the length that the plume extended downstream was inversely related tothe length that the outer cup extended downstream.0 0 0Figure3.4.6RepresentativeobservationsforanICAPloadedwithanintermediatelevelofchloroformonaninnerargonstreamataflowrateof a.0.6,b.0.8,andc.1.0litersIminute.0 I.r0ie HUa.r 0b.Uc.C’C’Observations67Effect ofInner Argon Flow Rate and PowerThe atomic plasma also responded to variation of the inner argon flow rate in aconspicuous manner. As the inner argon flow rate was increased, the atomic plasma appeared tomove upstream, or sit down within the intermediate tube, presumably because higher inner argonflow rates prevented the recirculation eddy at the base of the discharge from folding solventmaterial into the outer argon steam, thus preventing the downstream translation and thermal pincheffect.The observed response of the solvent loaded ICAP to variation of power may be statedquite simply: At lower powers, the discharge responded to all of the other parameters asdescribed above, only more sensitively. By contrast, the discharge was more robust at higherpowers; At powers approaching 2.0 kiloWatts, the discharge became insensitive to variation ofany other parameter, including solvent plasma load.Extention ofvisually observed responses to other solventsIt was stated earlier that the observed response of the ICAP to loading by any solventinvestigated in this work could be conveniently described as similar to a chloroform loadedICAP, similar to a methanol loaded ICAP, or resembling a hybrid of the two, depending on therelative content of carbon and oxygen in the solvent. If the carbon to oxygen content of thesolvent approached 1:1, then its appearance was similar to that of a methanol loaded ICAP. If thecarbon content greatly exceded the oxygen content, then its appearance was similar to achloroform loaded ICAP. This generalization may be extended further, to solvent mixtures andto oxygen addition to the aerosol stream. For example, the appearance of an ethanol loaded ICAPmay be described as a hybrid of chloroform and methanol loaded ICAPs, but loading by anequimolar mixture of ethanol with water results in a discharge which is virtually indistinguishablefrom a methanol loaded ICAP. The same is true for a ICAP loaded with xylene when oxygenhas been added to the aerosol stream in equimolar proportion to the amount of carbon.Observations683.5 ConclusionsThe detailed observations reported here reveal that any further investigation must takedistortions and translations of the macroscopic structure of the discharge into consideration. Theyalso reveal that the appearance of the discharge depends on the relative proportions of oxygen andcarbon in the aerosol stream, irrespective of the chemical form that the oxygen and carbon areintroduced. Beyond that, a number of physical phenomena are evident in the observations. Thesinclude a thermal pinch and convective distribution of solvent material over the argon stream.Moreover, incandescent radiation was observed from a conical shell nested within a dissociationfront, indicating that solvent pyrolysis proceeded via macroscopic soot particles. Overall, theobservations reported in this chapter provide a valuable survey of the parametric behavior of thesolvent loaded ICAP.Observations693.6. References1. S. Greenfield and H.M. McGeachin, Analytica ChimicaActa 74: p. 225-245 (1975).2. J.S. Koirtyohann, Jones, C.P. Jester and D.A. Yates, Spectrochimica Part B 36: p. 49-59(1981).3. Koirtyohann, S.R., C. Baber, and M. Franklin, Spectrochimica Acta B 31: p. 589-587(1976).4. Boumans, P.W.J.M. and M.C. Lux-Steiner, SpectrochimicaActa B 37(2): p. 97-126(1982).5. Truitt, D. and J.W. Robinson, Analytica Chimica Acta 49: p. 401 (1970).6. Truitt, D. and J.W. Robinson, Analytica ChimicaActa 51: p. 6 1-67 (1970).7. Montaser, A., V.A. Fassel, and J. Zalewski, Applied Spectroscopy 35: p. 292 (1981).8. Thorpe, NASA Contractor Report 1143, 1968..9. Greenfield, S. and H.M. McGeachin, Aiwlytica Chimica Acta 100: p. 101-119 (1978).10. Halliday, D. and R. Resnick, Fundamentals ofPhysics. 2 ed. 1981, New York: JohnWiley and Sons, Inc. 947.11. Sears, F.W., Optics. 3 ed. 1958, Reading: Addison-Wesley. 386.12. Kaufman, L., Chapter 3: Sensitivity to Light, in Sight and Mind:An Introduction to VisualPerception. 1974, Oxford University Press: New York.13. B.L. Caughlin and M.W. Blades, Spectrochimica Acta, Part B 42(1/2): p. 363 (1987).Chapter 4The Parametric Complexityof a Chloroform LoadedInductively Coupled Argon Plasma4.1. TNURODUCTIONACCORDING TO FEYNMAN [1], the elements of the scientific method are observation,reasoning and experiment. This is perhaps the most accurate definition because of its generality.Using more specific terms, Box and Hunter [2] have described the scientific method as aniterative learning process of first testing a hypothesis against experimental results, followed bymodifying the hypothesis (be it a conjecture, model, or theory), then retesting the hypothesis, andso on, until the problem is understood. Boumans [3] clearly applied the scientific method fromboth perspectives when he formulated a rational approach for developing spectrochemicalsources. The first step of his approach prescribes careful observation in order to establish theparametric response of the method. His second step prescribes experimental investigation, inorder to characterize physical properties of the spectrochemical source. His third step prescribesreasoning and hypothesis formulation in order to explain the analytical performance of the source.His final step is the rational development of the spectrochemical method. In other words, thefmal step is to refine the method after gaining an understanding of the physical characteristics ofthe source and how the physical characteristics determine the analytical performance. FollowingFeynman’s, Boumans’, and Box and Hunters’ lead, the rational strategy proposed by this thesisadds something more. It surveys the complexity of the discharge, a complexity that otherwiseThe Parametric Complexity71defies the formulation of clear hypotheses. Rather than simply establishing the parametricresponse in the first step, this thesis prescribes an extensive survey of the parametric, spatial andtemporal complexity. Then the rational approach can proceed with physical characterization andso forth, as per Boumans’, Feynman and Box and Hunter.This chapter surveys the parametric complexity of an ICAP loaded with chloroform only.The next two chapters will explore the spatial and temporal complexity and extend theinvestigation to other solvents. Then Chapters 7 through 9 will explore the physicalcharacteristics. And although the rational strategy proposed here provides the glue for binding allthe thesis together, Feynman’s description of the scientific method—observation, reasoning andexperiment— should be kept in mind throughout.The survey of the parametric complexity presented here confirms and extends what isalready known about the analytical performance of ICAP-AES, particularly in the area of aerosoldesolvation; it confirms the findings of Maessen et al. [4], that the analytical performance ofICAP-AES can be substantially improved by desolvating the sample aerosol before it reaches theplasma. The results further reveal that relatively low forward powers (750 W lower than thelowest power used by Maessen et al.) can be used to both sustain a stable plasma and achieveacceptable analytical performance, provided the aerosol has been sufficiently desolvated beforereaching the discharge.The survey also reveals important physical relations amongst emitting species in thedischarge. For example, they reveal a relationship between excitation of the analyte and solventdissociation in the boundary region of the discharge. Briefly, the boundary between cold aerosoland hot plasma gas is revealed by intense diatomic carbon emission. This boundary residesupstream from the onset of analyte emission, so the solvent molecules dissociate before theanalyte atoms are excited, assuming that upstream events precede downstream events.Consequently, species such as diatomic carbon play a minimal role in analyte excitation, becausethey are separated from analyte excitation in both space and time.The Parametric Complexity72Another insight revealed in this work is the connection between air entrainment and theaxial maximum of analyte ion line intensity. This connection has not been explicitly identifiedbefore, and it may explain several anomalous findings reported in the literature.4.1.1 Scope and ObjectivesThere were two objectives for the work presented in this chapter. The first one was tosurvey the parametric complexity of the analytical method, ICAP-AES. The second was tosurvey the parametric complexity of the discharge alone. The overall survey should provide arational basis for investigating the physical characteristics of the ICAP by revealing which regionsof the parameter space are worth investigating. The parametric survey should also provide a cleartrail back to the analytical performance from the physical characteristics are known. Such a trailwould aid immensely in explaining the analytical performance in terms of the newly discoveredphysical properties.Ideally, both a survey of the parametric complexity and a survey of the physicalcharacteristics would extend over the same parameter space. This would allow one to span thegap between the insight held by the analyst (who measures net line intensities and interferencecoefficients), and the insight held by the experimentalist (who measures physical properties).Boumans pointed out that this gap remains an obstacle in ICAP-AES [3]. Moreover, it is themost frustrating barrier that blocks the path to refining and developing the method any further.Of course, one may argue that such extensive surveys would be superfluous because onlythe conditions of optimum analytical performance warrant interest [5]. One could find thoseoptimal conditions and then confine further investigations to them. Such an argument would bereasonable if there were only one unambiguous set of optimal operating conditions. In practice,however, several different optima may be established for spectrometric methods based on theICAP. The different optima depend upon the optimization criteria and the method of detection.For example, optimum conditions for low detection limits are often inconsistent with those forminimizing matrix interferences. Optimum conditions for the detection of ion line signals oftenThe Parametric Complexity73differ from those for the detection of atom line signals. Furthermore, optimum conditions formass spectrometry are almost always inconsistent with those for atomic emission spectrometry.Clearly, the optimum conditions for all of the above examples could only be contained within anextensive survey. Moreover, the experimentalist can study interference effects more easily fornon optimal operating conditions because interference effects are exaggerated away from theoptimum.4.1.2 Literature Review of the Parametric Response ofICAP AESA comprehensive review of the ICAP-AES literature is beyond the scope of this thesis.This literature review is limited to the available parametric insight into the solvent plasma loadproblem.Boumans has reviewed the literature concerning the parametric response ICAP-AES upto 1984. His review may be found in Ref. [6]. Important parametric investigations into theanalytical performance of ICAP-AES are covered, from 1969 to 1984. Boumans also lists theparametric trends for ICAP-AES, and proposes a procedure for finding the best compromise ofparameter settings for multielement analysis (a compromise because not all of the analytical lineshave the same parametric optimum). Overall, he clearly spells out how the analyticalperformance for ICAP-AES may be optimized. From the outset, however, Boumans lamentsthat “...data are scattered in the literature...”Here, the details of Boumans’ review relevant to solvent plasma loading will berecapitulated, then brought up to date.According to Boumans, one must exploreforwardpower; observation height; the sampleuptake rate of the nebulizer; and the outer, intermediate and inner argon flow rates in order tofind optimal compromise conditions for multielement analysis. In addition to these parameters,Maessen et al. found that solvent plasma load must also be explored [4], [7], [8].The Parametric Complexity74Out of all these parameters, the analytical performance is not critically dependent upon theouter argon flow rate. Consequently, this may be set to the lowest, most economic levelnecessary to sustain a stable plasma for long periods. For ICAPs loaded with aqueous solventaerosols, or desolvated organic solvent aerosols, only 10.0 1/mm is required (for a high efficiencytorch, such as the one used in this work). In contrast, Boumans and Lux-Steiner found in earlystudies that ICAPs loaded with organic solvents require 15.0 to 20.0 1/mm , presumably toprevent carbonaceous soot form building up on the inner surface of the torch [5]. But in this earlywork, effective measures to desolvate the aerosol were not taken. In fact, solvent plasma load didnot receive much attention in early work at all.Largely ignoring solvent plasma load, early investigations (1966 to 1978) focused on onlyfour parameters: forward power, viewing height, nebulizer gas flow rate, and the excitation andionization energies of the analyte. For example, Boulos et al. [9] investigated how the net lineintensity of several prominent emission lines responded to variation of the nebulizer flow rate.Their investigation covered a range of viewing heights. They found that the emission lines couldbe classified according to excitation and ionization potentials. They also attributed parametricbehavior to macroscopic phenomena, by suggesting that the analyte followed different pathsthrough the plasma toroid for different nebulizer flow rates. Overall, they concluded that the innerargon flow rate is a critical operating parameter.Horlick et al.. later surveyed a grand total of four parameters —line energetics, viewingheight, inner argon flow rate, and r.f. power [10]. They facilitated their emission measurementsby mounting an array detector vertically in the exit focal plane of their grating spectrometer, anarrangement that allowed them to sample an entire range of viewing heights (an entire axialprofile) simultaneously. That neatly took care of one parametric dimension and allowed them toexplore the others more extensively, which they did. They demonstrated that emission linesfrom the ICAP could be unambiguously classified as hard or soft lines according to theirexcitation energy and parametric response. The classification depends on how the axial profile of)the emission line responds to inner flow rate and power. Briefly, they classified the emission lineThe Parametric Complexity75as hard if the maximum intensity of the axial profile did not move higher or lower above theinduction coil in response to variation of the operating parameters. On the other hand, theyclassified the line as soft if the axial profile was sensitive to variation of the operating parameters.In general, they found that ionic lines and atom lines with relatively large excitation potentials, saygreater than 6 eV, are hard. On the other hand, atom lines with excitation potentials significantlylower than 6 eV were soft. Atomic lines with intermediate excitation potentials displayedintermediate behavior.This early work established the parametric response for conventional ICAP-AES inwhich the sample is introduced as an aqueous aerosol. But solvent plasma load was not takeninto consideration, because it was found that water plasma load does not vary appreciably fromabout 0.3 mg/s for most nebulizers. More recently, Maessen et at. found that solvent plasma loadwas indeed interesting for other solvents [4], [7], [8].Maessen et a!. began their investigations by devising the continuous weighing method[4],a method for quantitatively determining and controlling solvent plasma load. Soon after devisingthis methodology, they discovered how solvent plasma load was determined by two keynebulization parameters—the uptake rate of the sample solution to the nebulizer and the nebulizergas flow rate [7]. This discovery allowed them to unravel the ambiguous interdependenciesbetween the analyte transport rate, nebulizer gas flow rate, sample uptake rate and solvent plasmaload. All these parameters had previously displayed ambiguous interdependencies and had notbeen previously separated experimentally. Maessen et al. showed that it made little sense toinvestigate solvent load effects by varying the sample uptake rate or the nebulizer gas flow rate.They found that controlling solvent load directly was a key step in rationally investigating theeffects of solvent load.Equipped with a method for controlling solvent plasma load as a parameter, theyinvestigated the effects of chloroform loading on the analytical performance of ICAP-AES. Theyreported the parametric response of the signal to background ratio (SBR), the relative standarddeviation of the background (RSDB), the background intensity, and the net analyte line intensity.The Parametric Complexity76The net intensity reveals the sensitivity of ICAP-AES. But knowing the sensitivity of asignal is seldom useful unless we also know its precision. Regarding ICAP-AES, we areinterested in the net analyte signal when its intensity is very small—very small compared to thebackground intensity. In this situation, the noise in the background predominates. As a result,the standard deviation of the background determines the detection limits. Often, however, onlythe background and net analyte intensity are surveyed to the neglect of the standard deviation ofthe background signal. It is then assumed that the relative standard deviation of the background isconstant, so that the standard deviation of the background scales up with the background intensity,as one would expect for source limited noise. It may then be easily demonstrated that the signalto background ratio (SBR) gives a good indication of how the detection limits respond (the higherthe SBR, the standard deviation of the background relative to the signal, and the lower thedetection limit). Consequently, a parametric survey of the SBR is also worthwhile.With this in mind, Maessen et al. surveyed the net intensity, the background intensity, thesignal to background ratio and the relative standard deviation of the background. For the netintensity, they found that emission lines previously classified as hard retained their hardness whenthe ICAP was loaded with a constant amount of chloroform. But their behavior became morecomplex when the chloroform loading was varied. The axial maximum for hard lines weresensitive to desolvation. They shifted towards lower axial positions, or to locations furtherupstream upon desolvation. This shift was accompanied by an overall increase in intensity of theentire axial profile. Such a response is similar to how hard lines respond to wide variations ofinner argon flow rate. However, after a certain extent of desolvation, the overall intensity of hardline emission decreased, presumably because the analyte transport fell off at extremely lowcondenser temperatures. On the other hand, the response of soft lines to desolvation wasconsistent for typical soft line behavior—it was sensitivity to the variation of any parameter.When they surveyed the response of the background emission intensity, the survey wasextended over several wavelength channels in order to probe the response of different spectralbackground features. Included in the survey were emission from diatomic carbon, CN, the lineThe Parametric Complexity77wings of atomic carbon, and the background continuum. For comparison, they also surveyed thebackground emission from an ICAP loaded with 0.5 mg/s of water. Significantly, thechloroform loaded ICAP was operated at three relative high forward powers, 1.5, 1.75, and 1.9kW, while the water loaded one was operated at 1.0 kW. Not surprisingly, they found that thehigher power, chloroform loaded ICAP emitted a continuum background ten times more intensethan the low power ICAP loaded with water. Presumably, the higher continuum emissionobserved for the higher power, chloroform loaded ICAP was due to the increase in plasmaelectron density. This is exactly what one would expect for higher applied r.f. power. (Perhapslower powers could have been investigated for the chloroform loaded ICAP if the aerosol hadbeen further desolvated.) In addition to more intense continuum emission, the backgroundemission displayed a conspicuous axial structure at the 500-nm channel. At this wavelength, anaxial maximum retracted into the torch and decreased in intensity with desolvation. This featurecorresponded to the tip of the plume of green diatomic carbon emission commonly observed fororganic solvent loaded ICAPs. Evidently, desolvation could significantly improve the analyticalperformance of solvent loaded ICAPs by reducing the molecular background emission.They surveyed the parametric behavior of the signal to background ratio (SBR) forrepresentative hard and soft lines over forward power, observation height, and chloroform plasmaload. Enhancements in the signal to background ratio were found on going from the highestchloroform load to moderate desolvation, especially for the soft line at low observation heights.Beyond a certain extent of desolvation, however, the signal to background ratio becameinsensitive to chloroform loading power and observation height. This led them to conclude thatthere was no reason to desolvate the chloroform aerosol beyond the point where the signal tobackground ratio became insensitive to chloroform loading, if the signal to background ratio werethe only criterion for optimizing the analytical performance.But, in order to use the SBR as a criterion for optimizing analytical performance, onemust also know the RSDB. Knowing both gives the standard deviation of the background, andhow large the signal is in comparison, hence indicates how low the detection limit lies. WhenThe Parametric Complexity78Maessen and Kreuning surveyed the RSDB they found that it was independent of solvent loadexcept a extremely high solvent loading [4]. They also found that when the chloroform aerosolwas sufficiently desolvated, that the RSDB fell to values comparable with values for ICAPsloaded with aqueous aerosols. They emphasized, however, the importance of preventingcarbonaceous soot from forming on the torch in order to maintain acceptable values of RSDB,and suggested procedures for doing so.The results of their parametric survey of the chloroform loaded ICAP inspired them toinvestigate the physical properties of the ICAP, under operating conditions involving solventplasma load by a variety of solvents [8]. Their new objective was to examine the physicalproperties of the ICAP loaded with known quantities of solvent, and also to examine the effectsof plasma loading by several solvents representative of the different classes of solvent used inICAP-AES applications. The applications of interest included trace metal analysis of solventextracts, of liquid chromatography eluates, and of samples dissolved in organic solvents. Theyselected ethanol and methanol to represent the alcohols, n-heptane to represent aliphatic solvents,toluene to represent the aromatic solvents and chloroform to represent the chlorinated solvents.Water was included for comparison.For each of these solvents, only two values of solvent loading were investigated: a solventplasma load of 15 tmo1Js, and a carbon plasma load of 50 jimolls. They chose 15 pmol/sbecause 15 pmo1Is conspicuously resulted in the optimal analytical performance for ICAPoperation involving both chloroform and water loading. On the other hand, Browner et al. [11]observed that plasma excitation conditions were determined by the amount of carbon loading.This inspired Maessen and Kreuning to compare loading by different solvents, all with theircarbon load set to 50 lImo]Js.The choice of these two levels (carbon load 50 jimolJs and solvent load = 15 pmo1Is)gave them a somewhat curious parameter space to investigate: for some solvents, the massloading of 15 iimolls resulted in a carbon load greater than 50 jimolls. For others, the massloading of 15 pmo1Js resulted in a carbon load less than 50 .tmol/s. At any rate, a two levelThe Parametric Complexity79investigation in solvent load was conducted as far as possible (the load values, carbon load = 50pmo1Is and the solvent load = 15 tmo1Js, exceeded the maximum tolerable load or fell short ofthe minimum obtainable load for a few solvents). The forward power and inner argon flow rateswere held constant at the optimal compromise values for multielement analysis. Axial profileswere measured over this parameter space, extending from 5 mm to 30 mm above the load coil in3 mm increments. They measured axial profiles for several emission signals. The signalsincluded Fe II excitation temperatures, C2, CN and C I background emission profiles and analyteemission profiles.When Kreuning and Maessen examined the effect of solvent load on axial profiles of FeII excitation temperatures, they found that the shape of the profile depended on both the solventplasma load and the nature of the solvent. Essentially, the behavior of the excitation temperatureprofiles was not simply solvent specific, but depended on solvent load as welL In general, theexcitation temperature decreased when the solvent load was increased. This is just what onemight anticipate if one regarded solvent load as a power consuming quantity. The maximumexcitation temperature was always found between 10 mm and 20 mm above the load coil, anddropped off rapidly in the tail flame region (above 25 mm). There were few exceptions to thisotherwise unremarkable behavior.An interesting revelation became apparent when they compared profiles from the ICAPloaded with different solvents and with the carbon load held at 50 pmo1Is. The spread inexcitation temperatures was narrowest for this experiment. The notable outlier was thechloroform profile, which displayed significantly lower temperatures. The authors attributed thisto the momentum of the chloroform vapour and the high mass loading of chloroform relative tothe other solvents. They suggested that these properties confined the chloroform plasma load tothe aerosol channel more than other solvents. In contrast, other solvents may have beendistributed about the base of the discharge. As a result, the load on the central channel wouldhave been alleviated. The essential argument was that the nature of the solvent determined theThe Parametric Complexity80characteristics of the aerosol stream, which in turn determined how the solvent was distributedabout the entrance paths to the plasma.They revealed something even more intriguing when they compared the axially resolvedexcitation temperature profiles for different solvents, but with the solvent load held at 15 imo1Jsthis time. They found that the excitation temperature profiles for methanol and ethanol loadingwere hotter than the profile for 33 tmo1Is of water loading—at all observation heights. Evidently,not all organic solvents cool the axial region of the ICAP with respect to aqueous solvents. Onceagain, it was demonstrated that both the absolute solvent load and the nature of solvent werecritical in determining excitation conditions within the plasma. The authors went on to suggestthat the distribution of solvent over liquid droplet and vapour phases determines how the solventplasma load is distributed about the entrance paths to the plasma. While the droplet phasepredominates for water, the vapour phase predominates for methanol. Moreover, methanolvapour may readily diffuse away from the injector gas stream, beyond the exit of the injectortube. Consequently, the methanol can disperse over the argon stream and relieve the load on theaerosol channel. In contrast, small water droplets tend to follow the flow stream of the injectorgas, concentrating the water loading on the aerosol channel, resulting in cooler conditions in thechannel than for methanol loading. In summary, several parameters determine the spatialdistribution of solvent loading within the discharge. They include the forward power, the innerargon flow rate, the distribution of aerosol between droplets and vapour, and the flow patterns andvelocities of argon within the torch—not just the absolute solvent plasma load or the nature of thesolvent.Revelations of further interest were made when the excitation temperature profiles forwater and the alcohols, methanol and ethanol, were compared with the excitation temperatureprofiles for the solvents that did not contain oxygen, toluene, n-heptane, and chloroform, all at asolvent plasma load of 15 p.molls. Incidentally, this solvent plasma load corresponds to a seventimes greater carbon load for n-heptane and toluene than for methanol or chloroform. As onemight anticipate, the excitation temperature profiles for toluene and n-heptane loading wereThe Parametric Complexity81significantly cooler than the profile for methanol. But they were not significantly cooler than theprofile for chloroform plasma loading, even though their carbon load was seven times greater.Evidently, the Fell excitation temperature profile for chloroform was anomalous.As for the 50 p.molls of carbon case, Kreuning and Maessen explained the chloroformanomaly by considering the paths the solvent material follows as it enters the base of thedischarge. If at one extreme, the solvent was minimally distributed about the base of thedischarge, but confined to the central aerosol channel, then the solvent loading effects would beconcentrated in the central, aerosol channeL This could happen when the solvent has been injectedprimarily as droplets, or when the aerosol stream has a high momentum along the axis of thedischarge. On the other hand, the recirculation eddy residing in the base of the discharge couldmix the solvent vapour of the aerosol stream quite efficiently with the plasma argon. One mightexpect this opposite situation for an aerosol stream of low momentum, with the mass of solventlargely in the vapour phase.Kreuning and Maessen offered an alternative explanation to account for the differencebetween the excitation temperature profiles for methanol and chloroform. They suggested that thetemperature difference could be attributed to the different pyrolysis products of the twosolvents—CC1 and C2 resulting from chloroform loading, and CO from methanol loading.This idea prompted them to measure both axially and radially resolved profiles of C2, CNand C I emission, in order to learn about the distribution of solvent pyrolysis products in thedischarge. Both the radially and axially resolved profiles revealed interesting spatial relationshipsamongst the pyrolysis products. In general, for all solvents, the radially resolved C2 emissionwas confined to within 3 mm of the discharge axis, low along the discharge axis (<5 mm abovethe load coil), and displayed an axial maximum (this contrasts with the observations reported inChapter 3 and the radial results reported in Chapter 5, in which the C2 plume was found to be ahollow tube capped by a bullet shaped region). This axially confined plume of C2 emission wassurrounded by an annulus of C I emission; the radially resolved C I emission profile exhibited anoff axis, or radial maximum at r = 3.5 mm and a local minimum at the axis (r = 0 mm).The Parametric Complexity82Significantly, the axial minimum of C I intensity was close to zero for solvents exhibiting arelative high intensity of C2 emission at the axis, solvents such as chloroform, toluene and nheptane. On the other hand, for solvents exhibiting a relatively weak intensity of C2 emission atthe axis, the radially resolved C I intensity was quite appreciable at the axis, although thesesolvents still exhibited an axial minimum of C I emission with respect to the off axis maximumat r = 3.5 mm. Further up in the discharge, where both the C2 and C I intensities fell off, CNemission formed a second annulus which surrounded the C I emission; its radially resolvedemission profile exhibited an axial minimum and a radial maximum at r = 5 mm. This result forCN emission was the one Kreuning and Maessen anticipated; CN formed from nitrogenentrainment into the carbon containing plasma (or at least entrainment into its boundary region)from the atmosphere.In summary, Maessen et al. have contributed immensely to our knowledge of theparametric behavior of ICAP-AES for applications in which the effects of solvent load areimportant. They established that solvent plasma load was a parameter critical to analyticalperformance: indeed, they found that for certain organic solvents, desolvation (or a reduction inthe solvent plasma load) could bring the analytical performance up to par with that of an ICAPloaded with aqueous solvents. They further revealed that emission lines from an ICAP retainedtheir hardness in terms of parametric behavior when loaded with organic solvents; the parametricbehavior of emission lines from a ICAP loaded with organic solvents could be classified in thesame way as it had been classified previously for ICAPs loaded with aqueous solvents. Inaddition to this valuable contribution, they discovered several trends and correlations which set thestage for explaining the parametric response of solvent loaded ICAPs in terms of the physicalproperties of the discharge. The most notable of these was the correlation between the shape ofthe axial profile of diatomic carbon emission, the shape of the axial profile of line of sightexcitation temperatures, and the position of the axial maximum for hard line emission.In spite of their valuable contributions, their work has important shortcomings, at leastwithin the context of establishing the parametric response. First, they did not investigate lowThe Parametric Complexity83powers. Second, they did not sample the solvent plasma load for more than two levels, exceptfor chloroform. And third, they did not investigate the effect of the inner argon flow rate,although they acknowledged its significance as a parameter and proposed to investigate it. Theseshortcomings by no means degrade the quality of their work, considering their original objectives,but they must be addressed by anyone attempting to establish the parametric response of theICAP loaded with organic solvents.Beyond the contributions and shortcomings, their work raised many questions anduncovered the complexity of the solvent load problem. What role do pyrolysis products play?How important are pyrolysis pathways? Do droplets play a significant role in organic solventloading, and if so, for which solvents? How is solvent material distributed about the entrancepathways of the discharge? Perhaps some of these questions may be answered by extending theinvestigation over a more extensive parameter space. Perhaps the only way to unravel thecomplexity of the problem will be to extend the investigation from time averaged, line of sightmeasurements to spatial and temporal resolution.After completing the work reviewed above, the Maessen group apparently abandonedtheir investigation into solvent load effects, possibly because of the enormous difficulty posed byconducting spatially resolved measurements of plasma properties and the distribution of plasmaspecies over an adequate range of operating parameters. Even so, several investigators continuedto recognize the immediacy of the solvent load problem and continued to investigate phenomenaassociated with injecting sample solutions into the ICAP as aerosol droplets. Notable amongstthe research groups continuing to report their investigations into aerosol plasma interfacephenomena are Browner et at. [12], Olesilc et at. [13], Mermet et al. [14], and Huang et al.[15].Where Maessen’s group left off investigating the effects of solvent load, Browner et a!.resumed their investigations into analyte transport efficiencies and solvent evaporation rates.Moreover, they recognized that the droplet size distribution of the sample aerosol was critical inthe solvent loading process. Briefly, they set out to investigate what they called theThe Parametric Complexity84solventlplasma interface, by examining both the solvent vapour, aerosol droplet and desolvatedaerosol particle components of the overall solvent load.They began by characterizing aerosol droplet size distributions for several solventsnebulized into argon [16]. Their results were quite astounding—they revealed that the longstanding empirical formula for predicting droplet size distributions, the Nukiyama Tanasawaequation [17], grossly overestimated the Sauter mean droplet diameter (a measure of the ratio ofthe total volume to the total surface area of the aerosol) for both aqueous and organic solventaerosols nebulized under typical ICAP-AES operating conditions. Moreover, they revealed thatthe mean droplet sizes for several organic solvents were always smaller than those of aqueoussolvents - contrary to the Nukiyama Tanasawa predictions. They also confirmed earlier reportsthat solvent evaporation rates are critical in determining analyte transport efficiencies.With this insight into the nature of the aerosol stream flowing into the ICAP, they went onto compare the effects that water loading and carbon tetrachioride loading had on ICAP-AES.Unfortunately, they investigated only one inner gas flow rate, just as Maessen et al. had done. Incontrast to the work done by Maessen et al., only one forward power and two solvents wereinvestigated, whereas the response measurements were somewhat more elaborate. They reportedbackground and analyte spectra along with axial profiles of diatomic carbon emission, Fe Iexcitation temperatures and the intensity ratio of Mg Ito Mg II emission lines. In addition, a newquantity, the emission magnitude, was introduced to report emission from background features.The emission magnitudes were simply the integrated axial profiles of C2, CN and C Iemission and were plotted as a function of the temperature of the condenser used to desolvate theaerosoL These plots revealed that the three carbon containing emitters each displayed a distinctresponse to solvent loading: when the condenser temperature was raised from -10 C to +20 C, theC2 emission magnitude increased by an order of magnitude, the CN emission magnitudedoubled, and the atomic carbon emission magnitude remained relatively constant. Overall, theemission from C2 displayed the most sensitive response to desolvation. This lead Browner et al.The Parametric Complexity85to suggest that C2 emission was a good candidate to use as a working diagnostic for monitoringsolvent plasma load [18].Browner et al. also demonstrated the significance of solvent loading as a source ofspectral interference. For example, upon desolvation, the overall intensity of the C2 axial profiledecreased as the position of its axial maximum receded into the torch. This behavior, alsoobserved by Maessen et al. [8], corresponds to the retraction of the C2 plume with desolvation.In contrast, the Ba II axial profile increased in intensity upon desolvation, while its axialmaximum moved upwards with respect to the induction coil. (This fmding was consistent withthe behavior observed by Maessen et al. [4]for hard line emission from a chloroform loadedICAP without desolvation, the C2 Swan bands with bandheads at 516 nm and 512 nm swampedthe Ba II line at 493.4 nm.) When desolvation was employed, the C2 Swan band disappearedfrom the spectral background, and the Ba II line stood out clearly, free from spectral interference.Browner et al. hypothesized that the Ball behavior could be explained by the change inthe ambient plasma power available to excite the analyte caused by a change in solvent loading.In order to test this hypothesis, they investigated the plasma excitation conditions by measuringan Fe I excitation temperature, and the ionization conditions by measuring the ratio of the intensityof Mg II emission to the intensity of Mg I emission, hereafter referred to as the Mg ratio (Mgratios were favored because of the absence of molecular background emission features in the MgII and Mg I wavelength region).When the Mg I to Mg II intensity ratio was determined for an observation height of 20mm above the load coil, and when the condenser temperature was varied from -10°C to + 20°C,the ratio remained close to 25 for water loading, but decreased from approximately 17 down to 2for carbon tetrachloride loading. This behavior was only representative for the range ofobservation heights from 15 to 20 mm. Higher up (>25 mm), the authors explained that thevariation was less extreme because the maximum possible temperature was much cooler,presumably as a result of the distance from the power loading region. This, however, wasdangerous to assume, given that the degree of spatial inhomogeneity and radial stratification ofThe Parametric Complexity86temperatures at 25 mm are apt to make line of sight Mg intensity ratios inaccurate. Nevertheless,the response of the Mg intensity ratio lead the authors to conclude that the decrease in the ratiocould only have resulted from a decrease in the excitation temperature.They also determined Fe I excitation temperatures for the solvent loaded ICAP from lineof sight Fe I line intensities, making the usual assumption that the electronic excitation of Fe Tinthe plasma followed a Boltzmann distribution. They then compared axial profiles of excitationtemperature measured under conditions of water loading to those measured under conditions ofcarbon tetrachloride loading. These profiles revealed that carbon tetrachloride loading resulted in alower excitation temperature in the plasma than water loading, at all observation heights, whenthe aerosol condenser was set to 18°C for both solvents. Moreover, with carbon tetrachlorideloading, the axial maximum of excitation temperature resided further downstream, or higherabove the load coils, than for water loading. These results suggested that the much greater massloading by carbon tetrachioride consumed the available excitation power in the plasma.Significantly, the position of the axial maximum for the excitation temperature corresponded tothe position of the axial maximum for hard line emission from a water loaded ICAP but not forthe carbon tetrachloride loaded ICAP. In this case the hard line maximum resided significantlyfurther downstream (in disagreement with Kreuning et al. ‘s results). Could this be explained bythe spatial averaging, and off axis bias intrinsic to line of sight measurements?More recently, Browner et al. [12]. have investigated the role of the auxiliary argon flowrate in the solvent loading process. These results conclusively reveal that the distribution ofsolvent about the entrance paths of the discharge—partly determined by the auxiliary argon flowrate—are critical to the solvent plasma loading process.In summary, the work conducted by Browner et al. and Maessen et al. firmly establishedthat the nature of the solvent, the absolute quantity of solvent loading, and the physicalcharacteristics of the aerosol stream (including the distribution of solvent between droplet andvapour phases, the momentum of the aerosol stream and the inner argon flow rate) are critical indetermining how the ICAP responds to solvent loading. Unfortunately, their investigationsThe Parametric Complexity87covered only limited regions of the ICAP operational parameter space. Had their investigationsextended over a more comprehensive parameter space, and had their determination of theresponse over this parameter space not been under sampled, then further trends may have beenrevealed, as postulated above. From an analytical perspective, Ebdon et aL [19] have establishedthat the desolvation of organic solvent aerosols can be employed to optimize the analyticalperformance of ICAP- AES, although it is not clearly understood why. In short, although manyprovocative insights have been gained into the solvent load problem, the task remains to fullyestablish the parametric response of ICAP-AES in the special cases of solvent loading anddesolvation.A further task remained to investigate the actual interaction between the aerosol and theplasma—the solvent/plasma interface had to be investigated. This task was initially taken up byFarnsworth et al. [201, then comprehensively pursued by Olesik et al. [21], [22], [23], [13],with some work done by Horlick et al. [24]. These investigators looked at the effect of aerosoldroplets which survived the traverse through the toroidal region of the discharge, and resulted incomplex spatiotemporal effects further downstream. The reader is cautioned against extendingtheir findings to solvent loading in general, because their work was confined to the investigationof loading by aqueous solvents only. Their work is summarized here because it includes aparametric investigation of desolvating droplet effects, upon which other parametric resultsreviewed here may hinge.The general scenario of desolvating droplets in the ICAP is as follows: the desolvatingdroplets represent a minute fraction of the total number of droplets which leave the spray chamberand enter the torch. However, they are the largest droplets in the aerosol, and hence make up asignificant proportion of the total volume of the liquid phase of the aerosol—they contain asignificant proportion of the analyte transported to the ICAP. When these droplets desolvate asthey travel along the axis of the ICAP, they create small regions of local cooling within theplasma. Also within the flowing plasma resides analyte which has been previously atomized.This analyte originated from droplets that were originally much smaller than the ones continuingThe Parametric Complexity88to desolvate. As the droplets (and attendant regions of localized cooling) fly through theobservation zone, their passage perturbs the otherwise steady state signal from the previouslyatomized analyte (the precise nature of these perturbations will be described shortly. Briefly,small regions of plasma with radically cooler thermal conditions traverse the observation zone,resulting in spikes in the atom line emission and depressions in the ion line emission). It isimportant to note that the passing droplets perturb the emission from the analyte that has beenpreviously atomized. By comparison, analyte within the droplets is dark matter. This darkanalyte may eventually begin to emit, but further downstream, after the droplet has completelydesolvated. Once there, the dark material exists in the form of vaporizing particles, left behind bythe progenitive droplets. These vaporizing particles may also perturb the otherwise steady state ofthe analyte emission, but in a different manner than the desolvating droplets: they give rise to localconcentrations of analyte near their surfaces, resulting in spikes in the analyte emission.An in depth overview of the investigations into the droplet effects in the ICAP has beenprovided by Olesik and Fister [13], [13]. Their work represents the most current in this area.On the sub millisecond time scale, spikes may observed in the time resolved signal forsoft line emission while dips may observed in the time resolved signal for hard line emission[21]. Three experiments provided solid evidence that droplet phenomena may be responsible forsuch fluctuations. These three experiments proved that incompletely desolvated droplets canindeed exist in what was formally thought to be a region of atomic plasma, when ICAPs areloaded with aqueous solvent aerosols generated by conventional pneumatic nebulizers. In a twochannel experiment conducted by Olesilc et al. [21], spikes in the sub millisecond, time resolvedwave forms for Ba I and Ca I correlated with each other, even when Ba and Ca were introducedthrough separate nebulizers. The spikes could not have been a result of vaporizing particles orlocal concentrations of either species, otherwise the spikes from Ba and Ca would not correlate;the only rational explanation was that incompletely desolvated droplets created regions of localcooling, which shifted the ionization balance from ions towards atoms for previously atomizedanalyte. In that way, transient peaks in the atomic emission for Ba would correlate with those forThe Parametric Complexity89Ca. In a second experiment, Olesik et al. [21] found that the emission spikes correlated with thelaser light scattering signals. The optical parameters were such that only relatively large aerosolparticle—incompletely desolvated droplets rather than desolvated particles—could have resultedin a scattering signaL In yet a third experiment conducted by Horlick et al. [24], the characteristicfluctuations in the analyte signal were found to disappear when the nebulizer spray chamber washeated (thus desolvating all the droplets, so that only solvent vapour and desolvated particles weretransported to the ICAP). But when the heated, analyte carrying aerosol stream was combinedwith an aerosol steam of nebulized water, the fluctuations in the analyte signal returned. All threeexperiments left little doubt that droplets can and do exist in conventional ICAPs used in routineanalytical practice.In order to remove any remaining doubt about the existence of incompletely desolvateddroplets within the ICAP, and to probe further into droplet related phenomena, Olesik and Fister[13]conducted an experiment in which they simultaneously recorded two line of sight emissionchannels and laser light scattering. All signals were resolved on the sub millisecond time scale.This allowed them to search for correlations between time resolved atom line emission, ion lineemission, and the scattering signal. They also revealed correlations between the time resolvedfluctuations and the form of time averaged, axial emission profiles [23].This brings us up to date on the literature reports of the parametric response of the ICAPassociated with solvent loading.It is difficult to surmise from literature reports what the comprehensive parametricresponse for ICAP-AES actually is. This difficulty arises primarily because: 1., the investigatedparameter space covers an inadequate range; 2., the parameter space has been under sampled; 3.,the reference frame used for the spatial coordinates of the discharge varies amongst researchgroups; and 4., the response signal chosen to investigate the parameter space often provides onlylimited information. In many cases, only line of sight, net emission intensities of a few analytelines have been reported—no indication of their noise statistics or the noise statistics of theirspectral background components have been reported. Moreover, very little information about theThe Parametric Complexity 90spatial structure of the discharge has been provided, and visual observations of the discharge haveseldom been reported.In spite of these problems, the literature provides valuable guidelines for the choice of therelevant parameters to study, and for choice of the emission signals that enable one to gauge mosteffectively the response surface over the entire parameter space. Moreover, the literature reportsthe results of many case studies and parametric investigations, which may not take solvent loadinto consideration, but still provide a basis of comparison for the results reported here. Inaddition, many of the literature reports offer physical interpretations for the parametric response,which in spite of being speculative, provide valuable insight.4.1.3 Surveying the Parametric Response with Axial ProfilesIn order to establish the parametric response of the analytical performance, 1. theparameter space must be defined and 2. appropriate response signals must be chosen.The parameter space in ICAP-AES is defined by several parameters, including thesolvent, the solvent load, the forward power setting, at least three argon flow rates, severalnebulizer parameters, the design and dimensions of the nebulizer, the design and dimensions ofthe torch, the load coil configuration, and the concentration of concomitant solutes in the samplesolution. Several of these have more or less been standardized, and only a few are critical indetermining the analytical performance and physical characteristics of the discharge. In thisstudy, all of the ICAP-AES operating conditions are held constant except the inner argon flowrate, the forward power, and the solvent load. These three parameters defme the parameter space.Only the results for a chloroform loaded ICAP are presented here—as a case study.The response signal measured over such a parameter space would ideally assess theanalytical performance, indicate physical characteristics of the discharge, and be readilydetermined. More specifically, the criteria for selecting the response signal are the following: 1.,the signal must be representative of the important analytical lines used in multielement analysis;2., the background must be well represented; 3., reliable estimates of the standard deviation of theThe Parametric Complexity91net signal and background signal must be obtainable; 4., spectral interference with the signal mustbe minimal; and 5., rapid measurement of the signal must be possible. It should also be possibleto extract physical information from the signal. Unfortunately, no single ICAP emission signalmeets all of these criteria; it turns out that several response signals must be measured in order tomeet all of these criteria.To this end, the emission intensity of two analyte lines, Mg 11279 nm and Mg I 286 nm,and their background contributions were measured (4 signals). A sufficient number of replicatemeasurements were made to reliably estimate the standard deviations of the both the net signaland the background. The ion line was representative of hard line behavior, whereas the atom linewas representative of soft line behavior. These lines are also relatively free from spectralinterference from molecular bandheads. Unfortunately, this means that their backgroundcontributions are not representative of background emission from solvent dissociation products,which are the most important source of background emission for an ICAP loaded withchloroform. In order to represent emission from solvent dissociation species, emission from theCN bandhead at 388 nm, the C2 bandhead at 516 nm and the C I line at 248 nm were measured.In addition to these three background signals, the ratio of magnesium ion to magnesium atom lineintensity was determined in order to indicate the thermal conditions. In total, eight responsesignals were determined over a three dimensional parameter space.With eight responses to determine over four parameters, it is important to conduct theparametric survey with expediency: 8 x n4 intensity determinations must be made to cover theparameter space, where n is the number of samples for each parameter (typically 2 or more). Anexceptionally expedient method is to measure axial emission profiles using a linear array detector.This enables one to measure intensities simultaneously over a complete range of observationheights. It permits extensive parameter spaces to be sampled statistically and in short order,rapidly providing the investigator with a valuable preliminary survey—a survey which informshim of the nature of the problem—which he can then probe with acuity, while safe from beingmisled by the conflicting results of other researchers who may have only considered a limitedThe Parametric Complexity 92range of parameters. So the expediency of measuring the response signals as axial profiles isworth consideration, if not indispensable, at least in the initial stages of an investigation.4.1.4 Axial ProfilesDescription ofan Axial Emission ProfileAnalysts and investigators of the ICAP alike usually align the axis of their light collectionoptics such that they view the discharge side on, and such that the optical axis intersects thedischarge axis. When the light collection optics are aligned to collect emission from the dischargein this fashion, the intensity measurements thus obtained are known as line of sight intensities.This distinguishes them from spatially resolved intensities obtained using radial inversion ortomography. (It would be even more precise to call them side on, line of sight intensities. Butside on is tacitly understood because the side on geometry is the conventional geometry forviewing the discharge. However, one should note that investigators have obtained promisingresults by viewing the discharge axially, with the light collection optics aimed down its axis.) To afirst approximation, this region of light collection for such an optical configuration may beregarded as a line integral, extending through the discharge, from one side to the other. However,it is more realistic to regard the region of light collection as an envelope, a light collectionenvelope, over which the efficiency of collecting light varies.When measuring line of sight intensities from the discharge, one may vary the position ofthe light collection envelope along the discharge axis; equivalently, one may vary the observationheight (the reference point for the observation height is usually set to the top of the load coil).When one plots intensity (line of sight) versus the observation height (above the load coil), oneobtains an axial profile of the line of sight emission intensity, or an axial emission profile.Axial profiles can be measured in two general ways—laboriously or expediently.Laboriously, the viewing height may be scanned by translating the discharge with respect to thelight collection optics. Expediently, an array detector may be mounted vertically in the exit focalplane of a grating spectrometer. Then the detector elements sample different observation heightsThe Parametric Complexity93and a complete axial profile can be measured at once. In either case, it is important to note thatspatial region in the discharge sampled by the detector( or by each detector element) is not simplya line integral through the discharge, but is defined by a light collection envelope.Spatial averaging results from the overlap of this light collection envelope with theemission volume. Moreover, temporal averaging is also intrinsic to axial profiles: in order toobtain acceptably low noise levels, the emission intensity must be integrated for times muchlonger than the time scale of important dynamic discharge phenomena.The Parametric Complexity 94hollow cone ofincandescanti radiationwallFigure 4.1.1. The spatial averaging intrinsic to line of sight measurements. The light collectionenvelopes intersecting the discharge collect light over enormous gradients of intensity,temperature and density. Several envelopes stacked vertically collect light for an axial profile.discharge axisCN Emissioncollection— atomic plasma_—C2 plumeinjectortubeThe Parametric Complexity 95Spatial and Temporal Integration ofAxial ProfilesFigure 4.1.1. shows the plume of analyte emission overlapping with two light collectionenvelopes. Each light collection envelope corresponds to a different observation height, or adiscrete detector element in the exit focal plane of the spectrometer. At each point within theregion of overlap between the envelope and the plume, the contribution to the line of sightintensity is simply the product of the collection efficiency and the local intensity. Consequently,the line of sight intensity is spatially integrated over the region of overlap. This spatial integrationproves critical in setting the limits for interpreting axial profiles, primarily because it can bias theintensity measurements towards locations away from the discharge axis. This situation becomesevident when one inspects the lower end of the analyte plume in the figure. Clearly, one cannotalways assume that line of sight measurements represent conditions along the discharge axisbecause of the bias towards off axis conditions. It is also clear from Figures 4.1.1 that line ofsight intensity information spatially integrates emission over regions in the discharge that spanenormous gradients in composition and thermodynamic conditions, so that the informationcontent about the fundamental properties—notably temperature and electron density—is severelylimited; the spatial integration intrinsic to axial profiles significantly limits their interpretation.One must also consider temporal integration. Long integration times are routinely used toobtain acceptable signal to noise ratios in both analytical and diagnostic measurements of ICAPemission. These long integration times are far greater than the time scale of importantmacroscopic phenomena at work in the discharge, masking the vaporization of incompletelydesolvated aerosol droplets and the vortex shedding in the tail flame region of the discharge. Theexperimentalist must realize that he may not study these phenomena directly by measuring axialprofiles, at least not in the manner described here.In spite of such severe limitations imposed on the interpretation of axial profiles by spatialand temporal integration, a wealth of information can still be extracted from them, as thefollowing sections attest.The Parametric Complexity964.2 EXPERIMENTAL SECTIONThe instrumentation was described in Chapter 2. Several experimental details specific torecording axially resolved profiles are presented here, details involving the imaging optics, thebackground subtraction procedure, data management and the data collection procedure. Forexperimental details of very similar experiments, the reader is referred to Horlick [25] ,Maessen[4], Weir [26] and Browner [18].Briefly, light from the solvent loaded ICAP was focused onto the entrance slit of a onemeter Czerny -Turner monochromator with a 2 inch diameter, fused silica, piano convex lenswith a focal length of 150 mm at 594 nm. The image to object ratio was 1:2, or 225 mm to 450mm at 594 nm, so that the magnification of the ICAP’s image on the entrance slit was 0.5. Inorder to bring spherical aberration of the optical train down to acceptable levels, the flat face of thelens was directed towards the entrance slit and the lens was equipped with a 15 mm diameteraperture. Adjusting the image and object distances, according to the lens maker’s formula for athick lens and the manufacturer’s specifications for the lens, compensated for the chromaticaberration. The 15 mm aperture also limited the acceptance angleThe exit focal plane and the plane of the monochromator was nearly stigmatic with theplane of the entrance slit This meant rays forming a monochromatic image on the plane of theentrance slit would be refocused by the collimator and camera mirrors of the monochromator toform an inverted, monochromatic image on the exit focal plane. Similarly, rays passing throughthe entrance slit were refocused to form an inverted image of the entrance slit on the exit focalplane. A linear photodiode array placed in the exit focal plane with its pixels aligned along thevertical direction would thus collect an axially resolved profile of plasma emission. Its spectralresolution would be determined by the linear dispersion of the monochromator and the pixeldimension in the horizontal direction, its axial resolution would be determined by the pixeldimension in the vertical direction.The Parametric Complexity97There were two further imaging considerations: the spatial response function of the opticsand spectrometer (in the vertical, or Y direction) was measured and found to be acceptably flat.Moreover, the entrance slit was carefully aligned with the 2x reduced image of the ICAP so thatthe image of plasma’s axis of cylindrical symmetry fell precisely on the slit.Using similar optics, Horlick et al.[25] placed a 512 pixel, self scanning, linear photodiode array in the exit focal plane of their monochromator, orienting it vertically so that it wouldact as an exit slit while giving axial resolution of 512 points over a 24 mm axial distance along theplasma. This gave them 50 p.m axial resolution. However, in this study 1000 pixels of a 4096pixel LPDA could be illuminated, covering an axial distance of 30 mm along the axis of theplasma, potentially giving an axial resolution of 30 p.m. But such fine spatial resolution wasfound to over sample the axial profile excessively, when the most complex profile displayedstructural detail that required only sub millimeter resolution. For this reason, the diodes werebinned into groups of 20, which essentially meant that the output from diodes within successivebatches of 20 along the array was summed together. Binning reduced the effective number ofdiscrete detectors from 1000 to 50, resulting in 0.6 mm axial resolution, which was far moreappropriate for this investigation, and resulted in reduced read noise (which was averaged out bybinning) and helped improve the data handling capacity—important for rapid and comprehensivestatistical determinations of axially resolved emission behavior.Twenty replicates were recorded for each profile, but the statistical results will not bepresented here. The averaged profiles were finally smoothed with a digital 5 point smoothingfilter [27], [28], [29]. This filter length in no way degraded the vertical resolution of thesmoothly varying axial profiles.The Parametric Complexity984.3 RESULTSThe axially resolved results presented in this chapter include emission from analytespecies and from atomic and molecular products of solvent dissociation and combustion. For theanalyte species, Mg I, Mg II will be considered. How their net emission signals and backgroundcontributions vary over the parameter space will demonstrate that desolvation is indeed relevant toseeking optimal analytical performance. In order to complement the analyte emission results, theparametric response of emission from the solvent dissociation products, C I, CN, and C2 will beconsidered. These results will demonstrate a.) the diagnostic utility of C2 and CN emission foroptimizing the analytical performance of a chloroform loaded plasma, b.) how the C2 and CNsignals characterize the plasma boundary region whereas C I characterizes the atomic plasma, andc.) how axially resolved emission profiles of these dissociation products can be used incombination with surveys of the background emission spectra to predict spectral interferenceowing to solvent loading. The results also reveal the nature of non spectral interference effectsowing to solvent plasma load.Format ofResultsIn general, results of the parametric study are presented as shown in Figure 4.3.1: thefigure consists of 12 frames, each frame displaying 6 axially resolved profiles. From left to right,the frames are labeled I, II, and III, corresponding to increasing forward powers, 1.00, 1.25, and1.50 kW. Descending from top to bottom, the frames are labeled A, B, C, and D, correspondingto increasing chloroform loads, 3.2, 4.2, 6.2 and 7.4 mg/s. Within each frame, the six axiallyresolved profiles are labeled 1 through 6, corresponding to inner argon flow rates which increasefrom 0.65 1/mm to 0.90 11mm in increments of 0.05 11mm. This format is retained for all of the12 frame figures in this chapter, and the horizontal axis always represents height relative to theload coil, although the quantities represented by the vertical axes vary.The Parametric Complexity994.3.1. The Parametric Response ofAnalyte Net IntensityMg II 279.55Figure 4.3.1. shows the response of the Net Mg II 279.55 nm signal to variation ofviewing height, power, solvent load, and total inner argon flow rate. This figure contains a greatdeal of information (as do all of the 12-frame figures in this chapter). But the general trends arereadily perceived if one notes that all the axial profiles of this figure are similar curves. For all ofthe Mg II profiles, the net intensity appears to increase from zero at a point within or just abovethe load coil, then proceed through a maximum somewhere between 5 and 15 mm above the loadcoil, ultimately to decay back to zero before reaching the axial position of 30 mm above the loadcoil. Upon closer inspection, the only significant responses of the profile over the entireparameter space appear to be 1. axial translation of the position of maximum intensity, 2. changein the slope of the onset in intensity, and 3. change in the intensity of the maximum. In fact,these three responses appear to be related; in general, if the onset in intensity begins higher abovethe load coil, then the slope of the onset decreases and the intensity of the maximum decreases. Afinal noteworthy feature common to all of the profiles is their apparent convergence to a commondecay point between about 25 mm and 30 mm above the load coil.100The Parametric Complexity30- -25- IA - hA lilA\‘j’ 25- lB IIB IIIB/120- 127/2425- lID hID20- ID11020’30 120’30Height Above Load Coil (mm)Figure 4.3.1. Axially resolved profiles of Mg II emission (279.55 nm)from an inductively coupled argon plasma loaded with chloroform;Forward Powers: I = 1.00, II = 1.25, 111= 1.50 kiloWatts;Chloroform Loads: A = 3.2, B = 4.2, C = 6.2, D 7.4 milligrams per second;Nebulizer Argon + Sheathgas Flow Rates:1 = 0.65, 2 = 0.70, 3 = 0.75, 4 = 0.80, 5 = 0.85, 6 = 0.90 litres per minute;The concentration of magnesium in the chloroform was 5 micrograms per millilitre.The Parametric Complexity101Several provocative trends are revealed when individual profiles from among the 12frames of Figure 4.3.1. are compared. For example, when one compares the profiles labeled 6(0.90 1/mm inner argon )from frames IA, TB, IC, and ID, (1.0 kW forward power, 1.5, 3.0, 4.5and 6.0 mg/s chloroform, respectively) one notes that the axial profiles are extremely sensitive tochloroform load at the relatively low power of 1.0 kW and at the high inner argon flow rate of0.90 1/mm—the intensity of the profile maximum decreases from 6.0 to 1.5 relative units and theonset moves from 5 mm to 12 mm above the load coil, which is accompanied by a shift in theposition of the maximum from 12 mm to 17 mm. On the other hand, when one compares theprofiles labeled 1 from frames lIlA, IIIB, IIIC, and IIID (1.50 kW forward power, 1.5, 3.0, 4.5and 6.0 mg/s chloroform, respectively), the profiles of highest power and lowest inner argon flowrate, one notes that the profiles are relatively insensitive to solvent load. Although the onset ofintensity is obscured for the profiles in these frames, the position of the maximum shifts less than2 mm and the intensities of the maxima appear to remain constant, except for a slight decrease inthe case of the lowest chloroform load. In the high power, low flow rate case, the profiles displayhard line behavior with respect to solvent load—the positions of their maxima are insensitive to ahuge increase in solvent loading—whereas in the low power, high flow rate case the profilesdisplay soft line behavior.The Parametric Complexity102Mg 1(285.21 nm)Figure 4.3.2 depicts the response of Mg I 285.21 nm net line intensity to viewing height,chloroform load, r.f. power and inner argon flow rate. There are two conspicuous differencesbetween the Mg II and the response of this soft Mg I line (with an excitation potential = 4.35eV). One immediately notes that the intensity appears to increase with increasing solvent load, incontrast to the ion line behavior. The second discrepancy becomes evident when one comparesthe three columns. The change in intensity with solvent load becomes more pronounced at higherforward powers. A third discrepancy is also evident. The maxima of the Mg I profilesconsistently appear at lower axial positions than the corresponding maxima for the Mg II profiles.It appears as though the cooler conditions promote Mg I emission, while hotter conditionspromote Mg II emission at the expense of Mg I emission. One might infer that Mg II and Mg Iare coupled by a balance which depends on conditions within the discharge. Whether or not thisbalance is dynamic or determined by some sort of equilibrium is not clear. For the moment, itcan only be stated that the balance leans towards Mg I emission (and possibly Mg I density) atrelatively cool locations along the discharge axis (averaged space and time), while promoting MgII emission at relatively hot locations.However, it is clear that the sensitivity of analyte line emission depends on the forwardpower, solvent load and inner argon flow rate. This is indeed the behavior one might intuitivelyexpect for the solvent loaded inductively coupled argon plasma if one were to interpret itsbehavior in terms of residence time and power consumption. Indeed, it is tempting to speculateupon the physical processes involving energy and mass transport, the excitation of analyte atomswithin the plasma, and upon the thermodynamic conditions within the plasma. But one mustproceed with extreme caution when one interprets the behavior of relative, line of sight emissionintensities such as these from analyte species within the plasma, firstly because the intensities aretemporally and spatially integrated and secondly because they are a function of both thetemperature and the concentration of emitting analyte.103The Parametric Complexity3.0-2.5- IA hA lilA2.0-1 .5 -lB IIB TuB!...2.5- Ic iuc HIC•1___1111ID hID hID2.5-0itii’0’2’0’30 0’2’0’30Height Above Load Coil (mm)Figure 4.3.. Axially resolved profiles of Mg I emission (285.21 nm)from an inductively coupled argon plasma loaded with chloroform;Foreward Powers: I = 1.00, II = 1.25, III = 1.50 kiloWatts;Chloroform Loads: A = 3.2, B = 4.2, C = 6.2, D = 7.4 milligrams per second;Nebulizer Argon + Sheathgas Flow Rates:1 = 0.65, 2 = 0.70, 3 0.75, 4 = 0.80, 5 = 0.85, 6 = 0.90 litres per minute;The concentration of magnesium in the chloroform was 5 micrograms per millilitre.The Parametric Complexity1044.2.2 The Parametric Response ofBackground IntensityThe spectral survey of ICAP background emission presented in chapter 3, which spannedfrom the near infrared to the near ultraviolet, showed that the most conspicuous backgroundfeatures in common to chloroform and methanol loaded ICAPs were atomic line emission fromcarbon and argon, and molecular band emission from diatomic carbon and the cyanyl radicaLRelatively weak OH and NO band emission could be detected from ICAPs loaded with water andalcoholic solvents, but emission from other molecular species, including CH, CO. Nil, N2+, N2and CC1, and from singly ionized argon could not be verified.The parametric behavior of the Ar I, C I, C2 and CN emission signals are interesting fortwo reasons: 1., they indicate the level of spectral interference and background intensity; and 2.,they reveal where the solvent dissociation front resides with respect to the analytical viewingzone, indicating how robust the plasma is (in a more robust plasma, the onset of solventdissociation should reside at lower heights).For both reasons, these background signals may serve as working diagnostics ofinterference effects in ICAP-AES, just as Olesik [30] suggested that OH emission, and Browner[18] suggested that C2 emission may serve as working diagnostics.Here, axial profiles of C I, C2 and CN are presented for an ICAP loaded with chloroformover the same parameter space of solvent load, inner argon flow rate and forward power asbefore. The wavelengths for sampling their emission were chosen simply by the criterion that theintensity maximum selected was freest from interference from other emission features: the 247.86nm line for C I, The 516.52 nm bandhead for C2 and the 388.34 nm bandhead for CN. Theobject of these measurements was simply to obtain a survey of the parametric response ofemission from solvent dissociation products. (No supposition pertaining to the relativecontributions of different bandheads or lines of a particular species were made).The Parametric Complexity105The Parametric Response ofEmission from Atomic Carbon: The C 1247.86 nm LineFigure 4.3.3 shows the parametric response of the line of sight, axially resolved intensityof the atomic carbon line at 247.86 nm from a chloroform loaded ICAP. The C I profiles arepresented in the same format as the manganese and magnesium lines, except that the results foronly two solvent loads are shown, 3.0 mg/s and 6.0 mg/s.Interpreting these axial profiles is more difficult because the C I signal arises from theatomic plasma, which extends radially well beyond the aerosol channel where the analyte isconfmed.At a relatively high power of 1.5 kW, the C I intensity shown in frames HIB and BIDincreases with chloroform solvent load for all of the profiles at all viewing heights along theprofile, with the most pronounced increase at 12 mm above the load coil This is what onewould intuitively expect for a robust, relatively high power plasma: an increase in the amount ofcarbon introduced into a robust plasma results in an increase in the C I intensity, especially athigher viewing heights where the maximum amount of energy has transferred from the r.f.power dissipation region into the aerosol channel In contrast, the presence of the aerosol channelbegins to manifest itself at the lowest viewing heights, where the correlation between chloroformload and C I intensity becomes ambiguous, and where a minimum in C I emission maycorrespond to the relatively cool region of the aerosol channel, where atomization is incomplete.The Parametric Complexity 10630-25-20-15- I -4.0.._______ _______25- ID20-15-I__10’2’0’30Height Above Load Coil (mm)Figure 4.3.3 Axially resolved profiles of atomic carbon emission (248 nm)from an inductively coupled argon plasma loaded with chloroform aerosol;Forward Powers: I = 1.00, II = 1.25, III = 1.50 kiloWatts;Chloroform Plasma Load: B = 4.2, D = 7.4 milligrams per second;Nebulizer + Sheath gas Flow Rates:1 = 0.65, 2 = 0.70, 3 = 0.75, 4 0.80, 5 = 0.85, 6 = 0.90 liters per minute;Pure chloroform was nebulized for these profiles.The Parametric Complexity107Further insight may be gained by noting the conspicuous correlation between the maximaof the profiles in frames 1118 and I11D with the corresponding Mg II profiles (Figure 4.3.1). Thiscorrelation corroborates the notion that hard lines emit most intensely where the maximumamount of energy has transferred into the axial channel. In other words, hard line emission ismost intense in the hottest part of the aerosol channel. This assumes that energy continues totransfer into the aerosol channel as the sample material travels downstream until the plasma isextinguished by air entrainment. It also assumes that there is no region along the aerosol channelwhere the plasma begins to cool down because of energy loss exceeding energy input. In otherwords, energy is simply supplied to the aerosol channel up to a cut off point, where the plasma isextinguished by air entrainment But this correlation begins to fail for profiles 4,5 and 6 of framehID. Here one must carefully interpret the spatially averaged results, bearing in mind theextended radial structure of the atomic plasma with respect to the analyte plume. Perhaps theconical structure of the atomic plasma combined with the penetration of the cool region of theaerosol channel masks an otherwise perfect correlation? In order to answer this question, radialresolution is required.In contrast to the behavior of the robust 1.5 kW plasma, frames lB and ID reveal that C Iemission from a 1.0 kW plasma exhibits the opposite response to chloroform loading; at 1.0 kW,all of the C I profiles decrease in intensity when the solvent load, hence amount of carbon, isincreased, except at heights lower than 10 mm. Presumably the plasma is not robust enough atthis power to contain the cool region of the aerosol channel below 5 mm, or within the load coilregion. In fact, observations reported in chapter three confirm that the cool region completelypenetrates the conical plasma up along its axis, under conditions identical to those of frame ID.Apparently the aerosol material simply blasts straight through the plasma.One further feature of Figure 4.3.3 worth noting is the remarkable similarity betweencertain profiles in frames lB. 118 and IIIB: the lowest power, low inner argon flow rate profileslabeled 1 and 2 in frame lB closely match the intermediate power, intermediate flow rate profileslabeled 3 and 4 in frame IIB, which in turn match the highest power, high flow rate profilesThe Parametric Complexity108labeled 5 and 6 of frame TuB. This similarity reveals that inner argon flow rate and the forwardpower have competing effects; an increase in forward power increases the C I intensity while anincrease in the inner flow rate decreases the C I intensity. Another parametric balance appears toinvolve power and solvent load, which becomes evident when profiles are compared betweendiagonally situated frames, from the upper left to the lower right; an increase in forward powerincreases the C I intensity while an increase in the solvent load decreases the C I intensity. It isinteresting to note that the same parametric balances were revealed by the Mg II response. Onemay surmise from these parametric balances that solvent load and inner argon flow rate arepower consuming quantities, but that they do not act independently—the inner argon flow ratedetermines how the solvent load consumes power and visa versa.In summary, the way C I emission responds to chloroform load, inner argon flow rateand forward power may be understood in the light of three perspectives: 1., how the robustnessof the atomic plasma changes over the parameter space; 2., how aerosol material is distributedwithin the argon flow system; and 3., how far the cool region of the aerosol channel penetratesalong the axis of the atomic plasma. The plasma is clearly robust to solvent load at high power—an increase in solvent load increases the carbon signal, indicating fairly constant plasmaconditions. But at high power, the plasma remains sensitive to the inner argon flow rate,indicating that the inner argon flow rate may determine how the aerosol material is distributedwithin the argon flow system and therefore how solvent dissociation consumes power. At lowpower, the plasma is considerably less robust—an increase in solvent load depresses the C I• intensity, indicating that the plasma conditions have become cooler. At any rate, it is difficult toisolate the effects of power consuming quantities, mass distribution effects and distortion orfolding of the atomic plasma from these spatially and temporally averaged, line of sight profilesof C I emission.The Parametric Complexity109The Parametric Response ofEmission from Diatomic Carbon: The C2 Bandhead at 516 nmIn contrast to the C I response, the axially resolved profiles of diatomic carbon emissionshown in Figure 4.3.4. reveal a remarkably simple parametric response: it appears that the generalshape of the diatomic carbon emission profile remains constant over the parameter space and thatonly its intensity and axial displacement change.Profile 6 of frame ID reveals the general shape of the axial C2 emission profile mostclearly: it has a single, clearly defined maximum at its upper end which abruptly decays to zero athigher axial positions, but gradually decays to a finite value at lower axial positions. This profileshape may be best understood in light of the observations reported in Chapter three. Torecapitulate those observations, the characteristically intense green emission from diatomic carbonformed a cup around the periphery of a plasma loaded with chloroform. This cup wrappedaround the plasma’s base to join up with a hollow cylinder of diatomic carbon emission runningup along the axis of the plasma toroid. The hollow, axial cylinder eventually coalesced into abullet shaped tip further downstream, which terminated at axial positions ranging from within theload coils up to 20 mm above them, depending on the operating parameters. This bullet shape tipcorresponds to the maxima of the diatomic carbon emission profiles. The hollow, axial cylinderis evident in plot 6 of frame ID; the non zero minimum at 5 mm is the integrated intensity acrossthe hollow plume. The integrated intensity increases with height as the hollow cylinder collapsesinto a bullet shaped tip, then abruptly decays to zero above the tip. Contributions from theperipheral cup are not apparent because the viewing range is too high—the peripheral cup isconfined to the region within the torch because peripheral air entrainment just above the torch rimconverts carbon species to CN.The Parametric Complexity250-200-1 00-IA hA lilA110I I I-I I I I I- IIBI I I4 8 12 16 4 8 12 16Height Above Load Coil (mm)I • I • I • I4 8 12 16Figure 4.2.4 Axially resolved profiles of diatomic carbon emission (516nm)from an inductively coupled argon plasma loaded with chloroform aerosol;Forward Powers: I = 1.00, II = 1.25, III = 1.50 kiloWatts;Chloroform Plasma Loads: A 3.2, B = 4.2, C = 6.2, D = 7.4 milligrams per second;Nebulizer ÷ Sheath gas Flow Rates:1 = 0.65, 2 = 0.70, 3 = 0.75, 4 = 0.80, 5 = 0.85, 6 0.90 liters per minute;Pure chloroform was nebulized for these profiles.1 50-I I I IlB50-0-250—200-1 50-1 00-50-0-250-200-1 50—1 00—50—0—250-200-1 50 -1 00-50-0-II I I I I IIIIB:I I I I I‘ I I I I I‘Ii6“IC34hID34I • IThe Parametric Complexity111The diatomic carbon profiles in Figure 4.3.4 display remarkably simple behavior. Theirresponse to inner argon flow rate and forward power may be described as a linear axial translationwhile their response to solvent load appears to be a linear scaling of intensity and axial translationwith the amount of carbon introduced.Inspection of plots 1 through 6 of frames ID, lID and hID reveals that the maximumintensity of the profiles is insensitive to both the inner argon flow rate and the forward power.Moreover, comparison of the plots in frames IA, TB, IC and ID reveal that the maximumintensity is directly proportional to solvent load. This is surprising because it suggests that theintensity of diatomic carbon emission is independent of plasma conditions, which one wouldexpect to vary considerably over the parameter space, just as the parametric response of Mn II,Mg II and C I indicated.The key to explaining this apparently anomalous behavior of diatomic carbon emissionmight be that it arises from the boundary region between the aerosol channel and the atomicplasma proper, a point that the observations reported in Chapter 3 suggested. As carboncontaining material is transported by diffusion or convection from the aerosol channel, across theboundary region and into the atomic plasma, it must traverse an enormous temperature gradient.Dissociation products ranging from undissociated solvent molecules to completely dissociatedatoms may plausibly occupy progressive regions of stability corresponding to progressive rangesof temperature across the temperature gradient, forming a spectrum ranging from molecules toatoms. It is plausible that the diatomic carbon molecule is only stable over a very narrowtemperature zone along the boundary gradient, ensuring that the plasma conditions are similarwherever diatomic carbon emission is observed. This would certainly explain the apparentindependence of the diatomic carbon emission intensity from plasma conditions and its simpledependence on plasma composition and amount of carbon.The Parametric Complexity112The Parametric Response ofEmissionfrom The Cyanyl Radical: The CN Bandhead at 388.34 nmWhen CN emission from a ICAP loaded with organic solvent is measured with axialresolution, it becomes clear that the CN emission structure consists of two components. Thesetwo component were first reported by Maessen and Kreuning [8], and were further investigatedby Browner [18], [12]; their existence is evident in the results presented here. Comparison ofthe CN profiles in frame ID of Figure 4.3.5 with profiles over the rest of the parameter spacereveals the two components quite clearly. The minor component of CN emission is found at lowaxial positions, and displays a parametric response similar to the C2 emission profiles presentedearlier. On the other hand, the major component of CN emission makes an overwhelmingcontribution at higher axial positions, and displays a very simple parametric response. It alwaysdisplays a maximum at approximately 22 mm above the load coil and appears to increase inintensity linearly with solvent load. The minor component originates from the inner boundaryregion, or the aerosol channel of the discharge, whereas the major component arises from theouter boundary region. The source of nitrogen for the major component is obviously atmosphericnitrogen. Browner established that the source of nitrogen for the minor component is mostcertainly a trace impurity of nitrogen in the argon. Those fmdings are corroborated by the resultspresented here.In contrast to the minor CN component, evidence presented in chapter 5 reveals thatemission from the major component originates from downstream positions and from beyond theouter periphery of the atomic plasma, or from what may be called the plasma tail flame. It istherefore clear that the major component results from entrainment of atmospheric nitrogen into theplasma and its reaction with carbon containing dissociation products.113TBlilA. I— —.IIIB,—The Parametric Complexity40-IA30-20-100 r403020-100 r40303020100/ICN.\IIAIIBIICp74.ID‘-1rIjicL:zI76/.-,f5 10 15 20 25 30 5 10 15 20 25 30 5 10 15 20 25 30Height Above Load Coil (mm)Figure 4.3.5 Axially resolved profiles of cyanogen radical emission (388nm)from an inductively coupled argon plasma loaded with chloroform aerosol;Forward Powers: I = 1.00, II = 1.25, III = 1.50 kiloWatts;Chloroform Plasma Load: A = 3.2, B = 4.2, C = 6.2, D = 7.4 milligrams per second;Nebulizer + Sheath gas Flow Rates:1 = 0.65, 2 0.70, 3 = 0.75, 4 = 0.80, 5 = 0.85, 6 = 0.90 liters per minute;Pure chloroform was nebulized for these profiles.The Parametric Complexity114A similar explanation may be offered for the parametric response of the major componentas for the minor component, an explanation which also assumes that the CN emission radiatesfrom a boundary region where the thermal conditions are relatively independent of those withinthe plasma. But in contrast to the steady flow system proposed for the minor component, wherematerial rapidly traverses an enormous temperature gradient, convincing evidence presented inChapters 5 and 6 and reported in numerous references supports the existence of vortex sheddingin the plasma tail flame, resulting in a vortex entrainment mechanism which rapidly extinguishesthe plasma by efficiently mixing the plasma gas with air. The net result of this vortex entrainmentmechanism is that CN emission depends almost linearly on solvent load, which determines theamount of carbon available for CN formation, and that a maximum in the axially resolved, line ofsight, CN emission profile is found at approximately 22 mm above the load coil, at a location justbeyond the conical tip of the of the atomic plasma, and close to the point of maximum mixing ofplasma carbon with atmospheric nitrogen. Further downstream, the CN intensity decreases asmore air is folded into the tail flame by the vortex train and quenches the CN emission.4.3.3 Magnesium Ion to Atom Ratio ProfilesThe axially resolved emission profiles presented thus far provide useful insight into theatomization, excitation and ionization processes in the inductively coupled argon plasma loadedwith chloroform. But emission intensities depend upon not only on excitation and ionizationprocesses, but also the spatial distribution or density of the analyte species within the plasma.This means that both the physical processes and the spatial distribution of the analyte speciesmust be taken into account, making the interpretation of emission intensities complex, if notfraught with sources of error. However, interpreting intensity data can be significantly simplifiedby removing its dependence on the spatial distribution of the emitting species. This can beconveniently accomplished by taking the ratio of emission intensity from singly ionized atomicspecies to the emission intensity from corresponding neutral species. The ion to atom emissionintensity ratio thus obtained is more or less independent of the spatial density of the sum of theThe Parametric Complexity115ionic and atomic components, and under certain conditions may be regarded as depending ononly the excitation and ionization conditions within the plasma.Of course, this reasoning requires that the emitting species exists only as free neutralatoms or free ionized atoms. In other words, none of it can be bound in molecules or trapped inaerosol particles. It also requires that the excitation conditions within the plasma are sufficientlyclose to local thermal equilibrium (see Chapter 7) for the Saha-Boltzmann distribution toapproximate the ionization-excitation distribution of the atomic species in question. Essentially,the analyte must exist as a free atomic gas within the argon plasma and some sort ofexcitation/ionization equilibrium must exist between the neutral and ionized states of the freeatomic analyte. Also, the temperature must be constant along the line of sight.It should be stated parenthetically that experimental evidence shows that this is not true forimportant regions of the parameter space, primarily because incompletely vaporized droplets trapmuch of the analyte and create enormous, transient temperature gradients, which drastically pullthe plasma away from LTE, and which can only be effectively probed by time resolvedmeasurements.We will assume that ion to atom line intensity ratios are independent of the analytedistribution, and faithfully indicate excitation and ionization conditions in the plasma, over anappropriate range of parameters. It is further assumed that if conditions should departsignificantly from the situation where a balance exists only between the atom state and the singlyionized atom, then it will be readily apparent in the results.Selection ofIon and Atom LinesThe question remains of what species to select for ion to atom emission intensity ratiomeasurements. Several compelling reasons favor the measurement of magnesium line intensityratios. In particular, the selection of the Mg I line at 285.21 nm and either one of the Mg II linesat 280.27 nm or 279.55 nm is favored by several practical advantages and physical attributes.Mermet [31] summarizes many of these: the excitation and ionization mechanisms of Mg inThe Parametric Complexity116argon plasma are quite well understood; transition probabilities have been determined accuratelyfor the lines at 285.21, 280.27 and 279.55 nm so that an experimental value of the ion to atomline intensity ratio may be compared confidently with theoretical ones predicted by the Saha andBoltzmann equations (see equation 4.1 below); the excitation potential of the Mg I 285.21 nmline is close to that of the Mg II 280.27 nm and Mg II 279.55 nm lines, so that the effect of thesecond exponential term in equation 4.1 is negligible, simplifying the dependence of the intensityratio on the electron density and the ionization temperature; the two Mg II lines have been provento be sensitive to ICAP excitation conditions, so that a Magnesium ratio incorporating one ofthem will also be a sensitive measure; the atom and ion lines are spectrally close together,obviating spectral response corrections; many literature reports report magnesium atom to ion lineintensity ratios using these lines, forming an established basis of comparison. In addition, for achloroform loaded plasma, the Mg II and Mg I lines are found in a spectral region relatively farremoved from interfering molecular band emission and atomic/ionic line overlap. Finally, thesecond ionization potential for magnesium (eV) precludes the formation of Mg2 (Mg ifi lines).Interpreting and Evaluating ofMg Ion to Atom Emission Intense RatiosOnce the profiles of the magnesium ion to atom emission intensity ratio have beendetermined experimentally, they may be interpreted along several avenues.One avenue of interpretation regards the plasma as existing in a thermodynamic stateclose to local thermal equilibrium. The experimentally determined ion to atom line intensityratios are compared to theoretical values predicted for the same electron density as theexperimental plasma, but at local thermal equilibrium.The theoretical ratios can be calculated using the Saha equation, experimentally determinedelectron densities, Dalton’s law, the assumption that the plasma pressure is one atmosphere, andfrom an argon ionization temperature. The theoretical values thus obtained form the basis ofcomparison for the experimentally determined ones. The further the experimental values deviatefrom the theoretical values predicted using assumptions based the existence of LTE, the furtherThe Parametric Complexity117the plasma conditions can be regarded to depart from LTE conditions. The departure from LTEcan then be accounted for by various collisional and radiative processes.It must be remembered that this avenue of interpretation neglects the effects of transportprocesses, flow dynamics and sample transformation (such as aerosol droplet vaporization), andrequires that the analyte exists only as free atoms and singly or multiply ionized atoms whichinteract thermally with the neutral argon atoms, the singly ionized argon atoms and the unboundelectrons of the plasma. Ample experimental evidence supports the prevalence of these sorts ofconditions and the existence of LTE, or at least only a partial departure from LTE in analyticalzone of the inductively coupled argon plasma, provided very little or no sample aerosol is beingintroduced. On the other hand, when sufficient amounts of sample aerosol are introduced for theICAP to be spectrochemically useful, ample experimental evidence supports the existence ofvaporization processes involving aerosol droplets and aerosol particles which drastically affect theemission characteristics of the analyte. Olesik has recently established that ion to atom lineemission intensity ratios must be interpreted as the time averaged result of a myriad of dropletinduced fluctuations on the sub millisecond time scale. This evidence suggests that ion to atomline intensity ratios ought to be interpreted along an alternative avenue to the one based on partiallocal thermal equilibrium principles—the ratios ought to be interpreted in terms of dynamicprocesses.The Parametric Complexity118Mermet has offered a third avenue for interpreting ion to atom line intensity ratios.Essentially, he proposed that experimentally measured ratios of the Mg II and Mg I 285 nm linesmust approach 20, the theoretical local thermal equilibrium value, in order for the plasma to berobust. In effect, he regarded the value of the experimental ratio as a measure of the robustness ofthe plasma. He used the following expression for the theoretical ratio derived from the Sahaequation and the Boltzmann function [31]:= (4.83 x 10y)(iA//)TYzexp(°nJex[E)](4.1)where g is the statistical weight, A the transition probability, A. the wavelength, Eexc the excitationenergy, E,, the ionization energy, Te the electron temperature, Te the excitation temperature, ethe electron density and k the Boltzmann constant. The subscripts i and a refer to the values forthe ion and atom lines, respectively.Mermet suggested that the ICAP must be sufficiently robust to completely vaporize all ofthe aerosol droplets and particles before they enter the analytical viewing zone in order for localthermal equilibrium conditions to prevail. This requires sufficient energy transfer from theplasma to the aerosol sample material, so the forward power must be sufficiently high and theresidence time of sample material in the plasma toroid must be sufficiently long. Otherwise, thematerial may simply flow straight through the plasma toroid and experience minimal vaporizationand dissociation.Digressing, it is interesting to note that the optimal analytical performance of the ICAPmust strike a compromise between low detection limits and robustness to interference effectscaused by constituents of the sample matrix. Boumans and Lux-Steiner have described thisThe Parametric Complexity119compromise: they reported that the lowest practical operating power in combination withrelatively high inner argon flow rates yield the best detection limits for ICAP-AES, primarilybecause its combination of operating conditions results in a spatial separation of the continuumbackground emission from the analyte line emission: the most intense continuum backgroundemission arises from the plasma toroid, whereas the most intense analyte emission comes froman axial plume which extends further above the load coil than the toroid. This spatial separation isevident in Figures 4.1.1. Conversely, the ICAP is more robust to matrix interference effects athigh powers and low inner argon flow rates—at the expense of low, background noise dependentdetection limits—because the plasma responds to these conditions by extending further above theload coil where the noise of its background continuum emission can make a greater contributionto the analytical signal Perhaps, then, the LTE interpretation works for ICAPs optimized to dealwith matrix effects, whereas interpretations within the context of dynamic processes would workbest for ICAPs optimized for the best detection limits. Clearly, some sort of method fordetermining the plasma’s state of robustness, or it’s macroscopic state is required incomprehensive experimental reports.It is by all three avenues of interpretation that the following results of Mg ratios areexamined.The Parametric Complexity120Magnesium Ratio ResultsFigure 4.3.6 shows the parametric response of the line of sight Mg ion to atom lineintensity ratio for an ICAP loaded with chloroform. The shape of the axial profiles of this ratio ismost clearly revealed in frame TIC. Each profile has a sharp onset in the ion to atom emissionratio, which begins at low axial positions and levels off or decreases in slope at higher positions toform a plateau. Between 20 mm and 25 mm above the coil, the profiles all merge together, thenappear to converge with the horizontal axis just above 30 mm. Upon first inspection, the onlyfour differences between all of the profiles over the twelve frames appear to be 1. the axialdisplacement of the initial onset, 2. the residual slope of the plateau region, 3. the axial positionof their maxima, and 4. the height of their maxima. Moreover, these four profile variationsappear to be related; when the initial onset of the intensity ratio shifts to higher axial positions,then the slope of the plateau region becomes steeper, up to the extreme situation where the plateauregion cannot be discerned from the initial onset region. Similarly, when the position of the initialonset shifts to higher axial positions, the maximum of the intensity ratio decreases.Before attempting to interpret this behavior, it is worthwhile to examine the absolutevalues of the atom to ion emission ratio. Without exception, the ratio profiles display absolutevalues less than 20. Moreover, an absolute value of 20 is closely approached only under robustplasma conditions: low solvent load, high power, low inner argon flow rate and at viewingheights between 7 mm and 10 mm above the load coil. This is evidenced by the profiles labeled1 through 3 in frames hA, lilA, and IIIB.The Parametric Complexity 12120-IA15-/10IA IA//__S__20-lB IIB-c 15-p10-20-IC lic Ilic15-110-..... r0-20-ID lID hID15-I10-5-__0- I I I I ‘ I I • I • I10 20 30 10 20 30 10 20 30Height Above Load Coil (mm)Figure 4.3.6. Axially resolved profiles of the Ratioof Mg II emission (279.55 nm) to Mg I emission (285.21 nm)from an inductively coupled argon plasma loaded with chloroform;Forward Powers: I = 1.00, II = 1.25, III = 1.50 kiloWatts;Chloroform Loads: A = 3.2, B = 4.2, C 6.2, D = 7.4 milligrams per second;Nebulizer Argon + Sheathgas Flow Rates:1 = 0.65, 2 = 0.70, 3 = 0.75, 4 = 0.80, 5 = 0.85, 6 = 0.90 litres per minute;The concentration of magnesium in the chloroform was 5 micrograms per millilitre.The Parametric Complexity122If one accepts Mermet’s proposal that the Mg ratio approaches 20 for a robust plasmaclose to state of local thermal equilibrium, then the profiles in figure 4.3.6 indicate that LTE isonly approached for low solvent loads, high power, low inner argon flow rate and sufficientheights above the load coil. One might further conclude that because the plasma appears toapproach local thermal equilibrium under these conditions, vaporization and atomizationprocesses must be complete and that non spectral, matrix interferences with the analyticalperformance should be negligible because the plasma is robust and energy has been efficientlytransported from the plasma toroid into the aerosol channeL But it remains to be understood whathappens in the plasma away from these conditions, where the ratio is much less than 20. Whathappens during the initial onset? What happens during the decay at higher heights? What state isthe plasma in under conditions of high solvent, low power and high inner argon flow rate.The ratio values much less than 20 in the initial onset may indicate a plasma state farremoved from LTE, conditions under which the vaporization and atomization of the aerosolmaterial is just reaching completion, and under which a dynamic model of the plasma state,which accounts for vaporization and atomization may be more successful. This avenue ofinterpretation is supported when the onset of the Mg ratio in Figure 4.3.6 is compared with thedecay of diatomic carbon emission in Figures 4.3.4—the onset of the ratio overlaps with thedecay of C2 emission. Or it is possible that the decay of C2 emission and the onset of the Mgratio do not overlap at all. It is quite possible that the onset of the line ofsight Mg ratio is biasedto off axis positions in the plasma, where the Mg emission may be more intense than on axis, andthat the tip of the C2 emission plume lies nested within the hollow base of the Mg I and Mg IIemission plumes. Only radially resolved information requiring further spatial resolution than theaxially resolved, line of sight results presented here can confirm this supposition. Radially andaxially resolved emission profiles presented in Chapter 5 appear to confirm this, although they donot cover quite as extensive a parameter space as the axial resolved results presented here. But theobservations reported in Chapter 3 support the nested plume supposition.The Parametric Complexity123Summary of Mg Ratio ResultsIn summary, the magnesium ion to atom emission intensity ratio is an informativediagnostic in three capacities: 1., it locates the axial position where the aerosol vaporization andatomization processes have effectively reached completion; 2., it locates the downstreamboundary of the plasma; and 3., it indicates how far the conditions within the plasma depart fromlocal thermal equilibrium. It was shown that the onset of the ratio profile correlates with thedecay of the diatomic carbon emission signal, presumably because the point of overlap betweenthe two profiles roughly locates the axial position where diatomic carbon has completelydisassociated and where the magnesium has been completely vaporized. An almost steplikedecay (complete decay to zero over a short axial distance) correlates well with the position wherethe hard line emission profiles decay to zero—this is apparently the downstream boundary of theplasma where vortex entrainment of air extinguishes the plasma. The Mg ratios for a robustICAP (low solvent load, high power and low inner argon flow rate) are remarkably close to thetheoretical local thermal equilibrium values calculated using realistic electron densities.The Parametric Complexity1244.6 CHAPTER SUMMARYThe axial profiles reported in this chapter provide a valuable survey of the parameter spaceimportant to ICAP operation; valuable because it reveals how physical processes such asvaporization and air entrainment vary or remain constant over the entire parameter space, andunder what parametric conditions partial local thermal equilibrium conditions prevail—hencewhere pLTE models should be realistic, how the analytical performance varies over the parameterspace, and which emission signals are useful as diagnostic checks for indirectly monitoring theanalytical performance. A valuable survey indeed for the analyst and researcher alike. Torecapitulate, this in depth survey of the parametric response of several emission signals providedthe following insight: the physical processes of vaporization and dissociation were clearlyrevealed by the diatomic carbon profiles and the profiles of the Mg ion to atom emission intensityratio. The decay of the C2 profiles correlated with the onset of the ratio profiles—their cross overpoint likely indicates where atomization of the aerosol material has reached completion. Theresponse of the crossover point clearly revealed that vaporization and dissociation processespersist well into the analytical zone under a wide range of parametric conditions. This survey alsorevealed the significance of how far the diatomic carbon plume, or the dissociation frontpenetrates up through the atomic plasma: if it penetrates too far, say beyond 12 mm, then theconditions within the viewing zone deviated significantly from pLTE, presumably because notenough energy is transported into the aerosol channeL This is an important revelation to theanalyst and researcher alike, because under plasma conditions far from LTE the dynamicprocesses associated with non spectral interference effects can be expected to be important,making interpretations in terms of collisional radiative balances difficult or unrealistic.Beyond the influence of the dissociation front, and past the analytical zone, a sharp cutoffin the Mg ratio profiles correlated with the decay of the hard line profiles, the decay of the C Iprofiles, and with the downstream maxima of the CN profiles—strongly suggesting that all ofthese features are tied together by the air entrainment mechanism which extinguishes or bringsabout an abrupt quenching of the plasma downstream.The Parametric Complexity1254.7. CoNcLusioNsSpatial and temporal averaging imposes severe limitations on the interpretation of theaxially resolved, line of sight emission profiles presented in this chapter. Further limitations arisefrom the complex excitation environment and the spatial distribution of the emitting matter.However, several conclusions may be drawn from the axially resolved profiles of emissionintensity and emission intensity ratios reported in this chapter—conclusions regarding analyticalutility and physical processes. The measurement of axial profiles are an efficient and effectivemeans for determining the parametric response in ICAP AES. Profiles of C2 emission and ofthe Mg ratio are excellent diagnostics of the dissociation processes in ICAP loaded withchloroform. It is evident that the further the C2 dissociation front penetrates the atomic plasma,the less efficient the energy transfer into the aerosol channel. Moreover, C2 and CN emissionintensities are largely independent of the plasma conditions. They probably indicate conditionswithin the boundary region rather than in the plasma. In other words, C2 emission indicates theconditions at the solvent plasma interface while CN emission indicates the conditions in the tailflame. As a result, the intensity of emission from both diatomic species is proportional to thecarbon load, in contrast to C I emission, which depends on both plasma conditions and the carbonload. Nevertheless, radially and temporally resolved measurements are required to further explainthe parametric response.The Parametric Complexity1264.8. REFERENCES1. R.P. Feynman, R.B. Leighton and M. Sands, The Feynman Lectures on Physics. Vol.Volume I. , Reading, Mass.: Addison-Wesley Publishing Company, Inc. (1968).2. G.E.P.Bo , W.G. Hunter and J.S. Hunter, Statistics for Experimenters. Wiley Series inProbability and Mathematical Statistics, New York: John Wiley and Sons (1978).3. P.W.J.M. Boumans, SpectrochimicaActa, PartB 46(6/7): P. 711-739 (1991).4. F.J.M.J. Maessen, G. Kreuning and J. Balke, Spectrochimica Acta B, 41: p. 3 (1986).5. P.W.J.M. Boumans and M.C. Lux-Steiner, Spectrochimica Acta B, 37(2): p. 97-126(1982).6. Boumans, P.W.J.M., Part I; Section 3.2.2. ICP Configuration and Appearance, inInductively Coupled Plasma Emission Spectroscopy, P.W.J.M. Boumans, Editor. 1987,John Wiley and Sons: New York. p. 584.7. F.J.M.J. Maessen and G. Kreuning, Spectrochimica Acta B, 42: p. 677 (1987).8. F.J.M.J. Maessen and G. Kreuning, Spectrochimica Acta B, 44(4): p. 387-384 (1989).9. G. Dube and M.I. Boulos, Canadian Journal ofSpectroscopy, 22(3): p. 68-76 (1977).10. M.W. Blades and G. Horlick, SpectrochimicaActa B, 36: p. 861-880 (1981).11. A.W. Boom, M.S. Cresser and R.F. Browner, Spectrochimica Acta, Part B, 35: p. 823(1980).12. G. Pan, G. Zhu, and R.F. Browner, Journal ofAnalytical Atomic Spectrometry, 7: p.123 1-1237 (1992).13. J.W. Olesilc and J.C. Fister, Spectrochemica Acta Part B, 46(6/7): p. 85 1-868 (1991).14. M. Marichy, M. Mermet, and J.M. Mermet, Spectrochimica Acta, Part B, 45(11):p. 1195-1201 (1990).15. Huang, B., et al., Spectrochimica Acta, Part B, 46(3): p. 407-416 (1991).16. Canals, A. and V. Hernandis, Journal ofAnalytical Atomic Spectrometry 5: p. 6 1-66(1990).17. Nukiyama, S. and Y. Tanasawa, Trans. Soc. Mech. Eng. Jpn.5: p. 68 (1939).18. Pan, C., G. Zhu, and R.F. Browner, Journal ofAnalytical Atomic Spectrometry 5: p. 537(1990).The Parametric Complexity12719. L. Ebdon, E.H. Evans, and N.W. Barnett, Journal ofAnalytical Atomic Spectrometry 4:p. 505-508 (1989).20. M.T. Cicerone and P.B. Farnsworth, Spectrochimica Acta, Part B, 44: p. 897 (1989).21. J.W. Olesik, L.J. Smith, and E.J. Williamsen, Analytical Chemistry 61: p. 2002 (1989).22. J.W. Olesik and E.J. Williamsen, Applied Spectroscopy 43: p. 933 (1989).23. J.W. Olesik and J.C.F. ifi, Spectrochimica Acta, Part B, 46: p. 869-883 (1991).24. G. Horlick and F. Qin. Proceedings of the Federation ofAnalytical Chemistry andSpectroscopy Societies Meeting XVII. 1990. Cleveland, OH:25. G. Horlick and N. Furuta, Spectrochimica Acta, Part B, 37(11): p. 999-1008 (1982).26. D.G.J. Weir and M.W. Blades, SpectrochimicaActa, Part B, 45(6): p. 615-618 (1990).27. A. Savitzky and M.J.E. Golay, Analytical Chemistry 36(8): p. 1627-1639 (1964).28. H.H. Madden, Analytical Chemistry, 50(9): p. 1383-1386 (1978).29. W.H. and S.A. Teukolsky, Computers in Physics : p. 669-672 (1990).30. J.W. Olesik, Analytical Chemistry, 63(1): p. 12A-21A (1991).31. J.M. Mermet, Analytica Chimica Acta 250: p. 85-94 (1991).Chapter 5The Spatial Complexityof the Solvent LoadedInductively Coupled Argon Plasma1. INTRODUCTION5.1.1. Objectives and ScopeThe observations in Chapter 3 revealed that the structure and behavior of the solventloaded ICAP are extremely complex. Chapter 4 then revealed that the parametric behavior of theanalyte and background emission is correspondingly complex. Although both chapters providedvaluable insight into the workings of the discharge, the full complexity of the discharge remainshidden by spatial and temporal averaging. Moreover, both chapters failed to reveal the physicalproperties of the discharge. As a result, chapters 3 and 4 leave us with several tasks: We mustcontinue to assess the spatial and temporal complexity of the discharge and then unravel itsphysical properties.Some might suggest that we proceed directly from here and probe the physical properties.If we were fortunate, we could identify all of the physical mechanisms and processes at work,and assess their significance in determining the analytical behavior of the discharge. Then wecould formulate a model that would take all of the significant mechanisms and processes intoaccount. Next, we could use this model to predict the analytical behavior of the discharge. If themodel stood up to the experimental tests, we could finally say that we understood how thedischarge worked, until further experiments proved us wrong.The Spatial Complexity... 1 29But such a strategy ignores the complex behavior of the discharge. Even if all of themechanisms and processes at work in the discharge were known, it is likely that the dischargewould still be too complex (spatially, temporally and parametrically) for us to devise a predictivemodel that was suitably realistic. Indeed, Feynman points out that it does not take very manysimple processes to create a system of incomprehensible complexity [1]. This is especially trueif the system happens to be inhomogenous, flowing and out of equilibrium—a system like thesolvent loaded ICAP. In reality, then, our ultimate understanding of the discharge may be limitedto simply knowing the fundamental mechanisms and processes at work in the discharge. Wecannot discover how those processes balance, compete and interact until we have a grasp of thecomplexity of the discharge. Until then, we can only devise approximate models.Clearly, characterizing, developing and understanding the solvent loaded ICAP is not asimple matter of formulating predictive models and then testing them experimentally. Theexperimentalist must resort to the experimental strategy outlined in Chapter 1. According to thatstrategy, it is irrational to probe the physical properties of the discharge without some grasp of itscomplexity. And one way we can come closer to grasping its full complexity is by surveying thespatial and temporal complexity of the analyte and background emission. That is exactly whatthis chapter and the next one aim to do.This chapter considers time averaged, spatially resolved maps of analyte and backgroundemission intensity, from both the scientific literature and our own work. The next chapterexplores the temporal behavior of analyte and background emission intensity. No attempt wasmade to measure intensity maps resolved in both space and time.5.1.2. Literature Review ofSpatially Resolved Intensity ProfilesIn order to determine the most efficient and informative experimental approach towardssurveying the spatial complexity, we will now critically review how investigators have alreadymeasured spatially resolved intensity profiles of emission from the ICAP We will examine howthey sampled their spatial profiles, what procedures they used to reconstruct spatially resolvedintensity maps, and how comprehensively they surveyed the operational parameter space. In theThe Spatial Complexity... 130results section, we will also examine what their results revealed regarding the spatial structure andbehavior of analyte and background emission features, particularly the emission from solventpyrolysis products.The earliest ICAP literature reported vertical and horizontal emission profiles obtained byscanning the image of the discharge over the entrance aperture of a monochromator. Obviously,such a point by point approach would be impractical for an extensive survey of the spatialcomplexity. Example of this sort of experiment are provided by Boulos et aL [2] and Maessen eta!. [3]. In their work, the surveys were severely limited. For example, Maessen et al. (furtherdetails may be found in the literature review of Chapter 4) only surveyed one inner argon flowrate [3]. Moreover, they only sampled the radial coordinate at three viewing heights. Finersampling of the axial coordinate was restricted to axial profiles (line of sight intensity resolvedover a range of viewing heights above the induction coil). We can only speculate that theseshortfalls resulted from the difficulty in measuring intensities at one spatial location at a time.In order to facilitate and extend their surveys, ICAP investigators have taken advantage ofthe roughly stigmatic imaging characteristics of Czerny Turner monochromators. They equippedtheir monochromators with one dimensional array detectors mounted vertically in the exit focalplane, as shown in Figure 5.1(c). (One dimensional array detectors are simply a row of smalldetector elements, say 128 to 4096 in a row. They are available as charge coupled devices orphoto diodes.) This configuration allowed them to rapidly record either vertical or lateral profilesof ICAP emission for a single wavelength channel. For example, Franklin et al. [4] recordedboth laterally and axially resolved profiles of calcium emission from the ICAP and studied theeffects concomitant potassium. Blades and Horlick made extensive use of linear photo diodearrays to classify emission lines according to the behavior of their axial profiles [5] and to studythe effects of concomitant easily ionizable elements on analyte emission [6]. Furuta and Horlickextended this work by obtaining lateral profiles with a linear array detector, and then radiallyinverting them [7]. Their work sorted out many of the spatial complexities of analyte emission,The Spatial Complexity... 1 3 1but their spatial sampling intervals were not optimal—more could have been revealed with lessdata.In all of this work, it was expedient to sample the lateral coordinate with the small intervaldetermined by the size of the pixels on the one dimensional array detector. Such a small sampleinterval was appropriate for the sharp intensity gradients of the lateral profiles. However, axialprofiles from both the literature and Chapter 4 reveal that a sampling interval of 0.5 mm—twentytimes the size of a typical array pixel—is more appropriate for sampling the axial coordinate.Evidently, methods employing array detectors could improved by batching or binning detectorpixels together in order to sample the emission profiles over larger intervals. The extra datahandling capacity could then be used to estimate statistics such as the relative standard deviation.Other ICAP investigators used both photographic emulsions and two dimensional arraydetectors in combination with monochromatic imaging spectrometers. This configurationallowed them to record complete monochromatic images of the spectrochemical source. Briefly,a monochromatic imaging spectrometer is essentially a slitless monochromators similar to theone depicted in Figure 5.1(a) or Figure 5.1(b). With a wide entrance aperture rather than a slit, themonochromator accepts the entire image. Rather than focusing a wavelength dispersed image ofthe entrance slit on the exit focal plane, it focuses an entire spectrally resolved image of thespectrochemical source on the exit focal plane, where it can be captured on film or by a chargecoupled device (CCD)—a two dimensional array detector. The image thus obtained is known asa image spectrogram.The Spatial Complexity(a)13217 Collimator(b) rang(c)CameraCollimating LensceCamera Lens 2D ArrayDetectorLens withInterference FilterFigure 5.1.1 Four possible optical configurations for capturing monochromatic images of aspectrochemical source.SourceImageSpectrogramn OpticsSourceLinear ArrayDetectorSource FocalPlaneThe Spatial Complexity... 1 33Of course, not all emission signals can be imaged this way. The emission signal must be verynarrow in the wavelength domain as well as being spectrally removed (not just resolved) fromother features. Molecular bandheads will smear out the image while images of closely spacedlines will overlap with each other. The best emission signals for image spectrograms are narrowemission lines spectrally removed from any other emission feature. The spectral surveypresented in Chapter 3 reveals that many emission signals from the ICAP meet this requirement.Good examples of image spectrograms have been presented by Horlick and Furuta [8]. Theirphotographic records qualitatively reveal the spatial distinction between the hard and softparametric behavior for emission lines from an ICAP.All the imaging problems resulting from spectral dispersion may be avoided by recordingmonochromatic images with camera equipped with an interference filter. Of course, a separatefilter would be required for each spectral feature. This configuration is depicted in Figure 5.1(d).It is interesting to ponder how an imaging interferometer could be devised.Dittrich and Niebergall made extensive use of imaging spectrometers and camerasequipped with interference filters to record monochromatic images of high voltage sparks, d.carcs, d.c. plasma jets and laser induced plasma plumes, all in addition to inductively coupledargon plasmas [9]. Their work should be consulted, if not to gain a wealth of insight intospectrochemical sources, then to appreciate the aesthetic beauty of their images. Essentially, theycaptured the monochromatic images on photographic film. The film was either mounted in theexit focal plane of an imaging spectrometer, or held in a miniature camera equipped with aninterference filter. They analyzed the images by converting the blackening density of thephotographic images into equidensigrams. In other word, they obtained contour maps of theimage intensity. This allowed them to subtract the spectral background, and in some cases,radially invert the images. They also demonstrated that imaging spectrometers can capture timeresolved images when used with high speed emulsions. But they admit that the photographictechniques involved in equidensitometry can be very laborious.The Spatial Complexity... 134An alternative configuration for an imaging spectrometer has been described by Hieftjeand is depicted in Figure 5.1(b) [10]. Unlike the configuration used by Horlick and Furuta orDittrich and Niebergall, this one does not focus an image of the spectrochemical source onto theentrance plane of the monochromator. On the contrary, it sends collimated light through themonochromator. The collimated beam may be stopped down at the entrance aperture so that avery narrow spectral bandwidth may be selected at the exit aperture. As a result, the spectralbandpass of the image is determined not by the size of the image, but by the dimensions of theentrance and exit apertures. Moreover, spectral dispersion does not smear out the image in thisconfiguration. After the collimated, spectrally resolved beam leaves the exit aperture, the spatialinformation is then retrieved by focusing the beam onto a two dimensional detector such as aCCD array. Better spectral and spatial resolution may be thus obtained at the expense of lightthroughput and the number of illuminated rulings on the diffraction grating. Admittedly, fewerilluminated rulings widen the spectral bandpass, but not nearly to the extent of a complete imagein the exit focal plane. Moreover, decreasing the light throughput does not present a problem forcapturing monochromatic images from a light source as bright as the ICAP, especially with CCDdetectors available today. Clearly, the advantages of this configuration warrant its considerationover the one depicted in Figure 5.1.(a).Figure 5.1(c). depicts a third configuration in which a thin, spectrally resolved, verticalslice of the image is accepted by the entrance slit. The spectrally resolved slice is then focusedonto the exit focal plane. This solves the problem of spectrally smeared or overlapping imageswithout sacrificing the light throughput or the number of illuminated rulings on the grating. Butthe entire image must be acquired slice by slice by scanning over its lateral coordinate.5.1.3. Obtaining Spatially Resolved Maps from Line of Sight ImagesAll of the optical configurations depicted in Figure 5.1 enable the experimentalist toacquire monochromatic images of a spectrochemical source. Such images are informative, yetbecause they are averaged over the line of sight, they are only resolved over two spatialdimensions rather than three. This leaves the structure of the spectrochemical source obscured.The Spatial Complexity... 1 35Fortunately, two properties of emission from the ICAP allow three dimensionally resolvedinformation to be retrieved from line of sight images. First, many of the interesting emissionfeatures from the ICAP are optically thin, so line of sight images may be regarded as lineintegrals of emission across the source. In other words, a lateral profile of emission from theICAP is really a set of line integrals of emission over parallel chords through the discharge.Second, the ICAP is almost cylindrically symmetric. For a cylindrically symmetric body, thevalue of a function along the radius can be determined from the line integrals across a set ofparallel chords. This may be accomplished by radial or Abel inversion [10]. Hieftje hascompared Abel inversion to peeling off successive layers of an onion [10]. The outermost lineof sight intensity allows the contribution from the outermost annulus to be subtracted from eachlateral location towards the centre. This is akin to peeling off the outer layer of the onion. Oncethe outermost layer has been peeled off, then successive layers can be stripped away in the sameway until the intensity at each annulus has been determined down to the centre. The intensity ateach annulus down to the centre, or the radially resolved intensity, provides complete spatialresolution for a cylindrically symmetric source.But experimental evidence shows that the ICAP is not cylindrically symmetric. On thecontrary, it has some degree of bilateral asymmetry. In order to account for this, experimentalistshave resorted to two other numerical methods for retrieving three dimensional information fromline of sight images: asymmetric Abel inversion [11] and tomographic reconstruction [10].Asymmetric Abel inversion first calculates asymmetry factors across the lateral profile. Thesefactors are simply the ratio of intensity between points on the left and right of the centre of thelateral profile. A set of asymmetry factors is calculated for pairs of lateral positions from thecentre out to the edge of the profile. Next, the average of the of the left and right sides of thelateral profile is radially inverted. Finally, multiplying or dividing the average radial profile by theset of asymmetry factors gives the asymmetric left and right sides of the radial profile. As onemight expect, the effectiveness of this technique is limited to small degrees of bilateralasymmetry. In order to account for further asymmetry, the experimentalist may resort totomographic reconstruction. This method assumes no axes of symmetry at all. It requires,The Spatial Complexity... 1 36however, much more intensity information than Abel inversion. Line of sight images must becaptured from several different viewing angles in order to apply tomographic reconstruction.Fortunately, Hieftje et al. have already applied this technique to the ICAP, and have noted thatwhile the ICAP is far from being cylindrically symmetric, the results of asymmetric Abelinversion compare favorably with those of tomographic reconstruction [10]. Apparently,asymmetric Abel inversion is adequate for retrieving spatial information from line of sightimages of the ICAP.5.1.4. Temporal ResolutionAll of the techniques discussed so far yield spatially resolved information, but generallyfail to provide temporal resolution. Temporal resolution requires short exposure times, such asthose available to gated, intensified detectors, or high speed film. An example of such technologyapplied to the ICAP has been provided by Olesik et al. [121. They synchronized their gated,intensified array detector to events owing to droplet vaporization. It may also be informative tophase average the experiment, as suggested by Winge et al. [13], and synchronize a shutter oroptical chopper to periodic fluctuations of discharge intensity. Such experiments would yield tophase averaged rather than temporally averaged intensities.5.1.5. The Imaging Technique Used in This WorkThe practical objective of this work was to survey temporally averaged, yet spatiallyresolved maps of analyte and background intensity. We could not do this using theconfigurations depicted in Figures 5.1.(a). or 5.l.(b). because we did not have a two dimensionalarray detector. Alternatively, the techniques required to radially invert photographic images wereimpractically complex and laborious. Moreover, two of the spectral features we planned toinvestigate, the CN bandhead at 388 nm and the C2 bandhead at 516 nm, displayed very broadspectral structures which would have corrupted the spectrally resolved image. Perhaps theconfiguration depicted in Figure 5.1.(c). could have been used. This avenue was not pursued.We rejected the point by point method out of hand for obvious reasons. We also rejected theThe Spatial Complexity... 137option of acquiring vertical and horizontal profiles by recording the array output directly: In theexamples from the literature, the array oversampled vertical profiles excessively and did not havethe flexibility to sample horizontal profiles of different widths (one would have had to change themagnification of the imaging optics or stretch the array). It also turned Out to be quite laborious tosample the entire discharge in this manner. We sought a method which could circumvent all ofthe above shortcomings and survey spatial intensity profiles of the entire discharge with optimalefficiency. Ultimately, we batched the output from an array mounted vertically, so as not tooversample in the vertical direction, and scanned the image of the discharge across the entranceslit in order to appropriately sample the lateral coordinate. In this manner, we could recordprofiles of at least half of the discharge quite rapidly (in about three minutes). Moreover, theentire images were of a manageable size and the individual lateral profiles at each observationheight could be radially inverted quite readily using an asymmetric Abel inversion procedure.5.1.6. Chapter SummaryThis chapter surveys the spatial complexity of the solvent loaded ICAP. It presentsspatially resolved maps of both analyte and background emission. Several physical phenomenaare evident in the maps, notably air entrainment into the argon jet by vortex shedding, entrainmentof solvent material into the outer argon stream by a recirculation eddy at the base of the discharge,and shrinking of the induction region by a thermal pinch effect. Although these phenomena areevident in the results, they are obscured by time averaging and the complexity of the discharge.Nevertheless, the results provide several guidelines for further research. They also suggest atleast three ways to improve the design and methodology of ICAP-AES, 1. by using backgroundsignals to monitor the solvent plasma load, 2. by eliminating air entrainment, and 3. by using thewealth of information in spatially resolved emission maps to optimize the operating parameters.The Spatial Complexity... 1 3 85.2. EXPERIMENTALThe instrumentation and procedure for igniting the plasma, generating the solvent aerosol,preparing solutions, controlling the operational parameters, and measuring spectroscopicquantities were presented in Chapter 2. Only the modifications and procedures specific tosurveying the spatially resolved parametric response of the solvent loaded ICAP are presentedhere.In order to adequately sample the emission profile and to meet the requirements of thenumerical Abel inversion procedure, up to 200 intensity samples were recorded along the lateralcoordinate of the discharge at each observation height. So the spatial maps required up to twohundred times more samples of emission intensity than the axial profiles presented in theprevious chapter. Further complicating the task, the optical train used to collect light from thedischarge had to be stopped down in order to meet Abel inversion requirements. With thisdecrease in light throughput, the detector required longer integration times to provide signal tonoise ratios that were limited by source noise rather than detector noise. With such large volumesof data and long measurement times, it was unrealistic to explore the sampling statistics andparametric response of the spatial emission maps to the extent that they were surveyed in Chapter4. For practical reasons, then, the parameter space of the survey was less extensive than the onein Chapter 4, and only enough replicate measurements were made to ensure reproducibility andrepeatability—not enough to reliably estimate the sampling statistics.5.2.1. Procedure for Recording a Monochromatic ImageThe discharge was translated laterally across the axis of the light collection optics, so thatthe image of the discharge was translated laterally across the entrance slit of the monochromator.Because the monochromator was equipped with a vertical array detector, several lateral profiles—each at different observation heights—could be scanned simultaneously. As a result, an entiremonochromatic image of the discharge could be recorded in a single lateral scan. The bandpassof the monochromatic image was determined by the width of the array detector and the reciprocalThe Spatial Complexity... 139linear dispersion of the monochromator. For example, the width of the detector of 0.506 mmmultiplied by the reciprocal linear dispersion of the monochromator at 516 nm (0.76 nmlmm)gave a spectral bandpass of 0.38 nm. Note that the nominal reciprocal linear dispersion for themonochromator is 0.83 nm/mm, but this value is defined by convention for a wavelength of0 nm. With increasing wavelength, the linear dispersion increases according to a cosine function.At 516 nm, one may calculate a value of 0.76 nmlmm for the reciprocal linear dispersion. Thiswas verified experimentally. The bandpass of 0.38 nm could be reduced by placing a slit maskon top of the detector, but this was found to be unnecessary. The sampling interval along thedetector corresponded to 0.6 mm intervals along the axis of the discharge. The total axial rangewas 25 mm. This range extended from either 5 mm to 30 mm above the top of the load coil (thetail cone), or -15 mm to 10 mm (the induction region). In the lateral direction, the samplinginterval was variable, but was generally set at 0.045 mm or 0.09 mm. In all cases, 200 lateralpositions were sampled, giving a lateral range of ±4.5 mm or ±9.0 mm. This oversampled thelateral profile, but not excessively. In fact, it reduced the numerical accumulation of noisetowards the centre of the radial profiles. In summary, monochromatic images of either the tailcone or induction region of the discharge were recorded with a bandpass of approximately 0.4nm, a lateral resolution of 0.09 mm and a vertical resolution of 0.6 mm. Moreover, images of theforeground, background, and dark signal could be recorded for background subtraction.5.2.2. Speciflcations of the Optical TrainIn order to generate radially resolved maps from line of sight images, the light collectionenvelope for each point in the image had to approximate a line integral. The optical train met thisrequirement as follows: A plano convex lens (see Table 2.2) was fitted with one of three apertures(5 mm diameter, 10 mm diameter or 5 mm wide x 30 mm high), and focused a 0.5X image ofthe discharge onto the entrance slit of a 1 meter focal length Czerny-Turner monochromator. At516. nm, the object distance was 450 mm (from the discharge axis to the front principal point ofthe lens), while the image distance was 225 mm (from the back principal point of the lens to theentrance slit). The Lens Maker’s Equation for thick lenses was used to calculate image and objectThe Spatial Complexity... 140distances at other wavelengths [14]. Of course, the planar surface of the lens faced towards theentrance slit in order to minimize spherical aberration. Finally, the entrance slit was lined up withthe cylindrical axis of the discharge. In practice, the cylindrical axis of the discharge veered off toone side with increasing observation heights, but this veering could be removed from the imageprior to radial inversion.The spatial response of the monochromator along the entrance slit was determinedexperimentally. It was found to be slight over the 25 mm imaging range (± 6.25 mm along theentrance slit) and was not taken into account. Maps of the absolute light collection efficiency ofthe optical system were calculated using an exact ray tracing algorithm developed by Farnsworthet al. [15]. Beyond providing the light collecting efficiency of the optical train, these mapsconfirmed that the light collection optics met the requirements for radial inversion. The spectralresponse function of the detector was also determined using an irradiance standard lamp [16].Knowing both the detector response and the light collection efficiency made it possible tocalculate radially resolved maps of absolute emission intensity. In this chapter, however, onlynormalized intensities are presented.5.2.3. Image ProcessingThe monochromatic images were corrected for background or dark signal contributionsby subtracting the background or dark signal images. Next, the images were smoothed with adigital filter in both the axial and lateral directions [17], [18], [19]. Because the axis of thedischarge reproducibly veered to one side, the veering could be removed by shifting the lateralprofile at each observation height by a predetermined number of sampling intervals (<6). Oncebackground corrected, smoothed and straightened out, the images were radially inverted.Although oversampling the lateral coordinate improved the numerical radial inversion, 200points in the radial coordinate were found to be excessive. So after the profiles were inverted,they were condensed from 200 x 40 intensities to 50 x 40 intensities by taking the average of 4points in the radial direction. This rendered the images much more manageable in later analysis,The Spatial Complexity... 141with little or no loss of information. The corrected, smoothed, radially inverted and condensedimages were then converted to topographical maps.5.2.4. Summary of the Spectral ChannelsIn this way, radially resolved maps were recorded for more than ten spectral channels,principally 516.56 nm (C2), 388.34 nm (CN), 247.86 nm (C 1), 257.61 nm (Mn ll), 279.55 nm(Mg Ii), 285.21 nm (Mg I), several Fell lines, An (687.13 nm), and H 1(486.16 nm). Some ofthese were measured for three solvents, methanol, water, and chloroform, over a range of innerargon flow rates, and a range of forward powers. The most extensively surveyed feature was theAr I. Absolute intensities of Ar I emission are considered in Chapter 7. Here, only aninformative sample of background features (CN, C2 and C I) and analyte emission lines (Mn II,Mg II and Mg I) will be presented.The Spatial Complexity... 1425.3. RESULTS AND DISCUSSION5.3.1. Format ofPresentationThe spatially resolved intensity maps presented in the following sections may be regardedas emission from the planar slice through the discharge shown in Figure 5.3.1(a) The plane ofeach map corresponds to the plane in the figure, with most maps bounded by z =5 mm to z =30mm and r = -7 mm to r = +7 mm, and a few bounded by z = -15 mm to z = +10 mm and r = -7mm to r = +7 mm. Obvious shadows of the induction coil reveal which maps extend into theload coils (that is. those bounded by z = -15 mm to z = +10 mm). There is some error in theradial intensities for the induction region because the turns of the load coil were not horizontal—they were helical.Intensity from the planar slice is represented by topographical isocontours as shown inFigure 5.3.1(b) The isocontours always depict the normalized intensity ranging from 0.1 to 1.0in increments of 0.1. These isocontours were either normalized to the maximum intensity forindividual maps or the overall maximum for a set of maps.The Spatial ComplexitySSa)C)CCaCd)CCa0.oooRadial Distance, mm5.0143Figure 5.3.1 The reference frame and contour intervals for the spatially resolved intensity maps.(a) How the image plane corresponds to a vertical slice through the discharge. (b) Typicalcontour map of the image intensity with a contour interval of 0. 1X the maximum intensity.SS0)15.010.0The Spatial Complexity... 1445.3.2. An Overview ofSpatially Resolved Analyte and Background EmissionFigure 5.3.2 depicts representative intensity maps of all the analyte and backgroundspecies considered in this chapter. The maps were measured for a ICAP operated at 1.25 kW,with an inner argon flow rate of 0.81 11mm. and a chloroform load of 4.5 mg/s.From left to right, the first map depicts emission from the CN bandhead at 388.34 nmand second map depicts the C2 bandhead at 516.56 nm. They reveal the boundary regions of theplasma. The CN map reveals the downstream boundary region known as the plasma mantle.This is where air entrains into the plasma jet. In effect, the CN map reveals the interface betweenthe hot atomic plasma and the cold room air. Within the induction coil, the C2 map partiallyreveals the upstream boundary region of the plasma. This region surrounds the base of theplasma and lines the inside of the aerosol channel. The C2 emission may be regarded as adissociation front between the hot atomic plasma and the relatively cold mixture of argon andundissociated solvent vapour.It is interesting to note the relative thickness of the C2 and CN boundaries. In general, theCN boundary was significantly wider than the C2 boundary. Only the cap of the C2 region was ofcomparable thickness to the CN boundary. This may indicate that the temperature gradients weresteeper across the C2 region except at its cap. Steeper temperature gradients would result in anarrower region over which the C2 molecule was stable. Alternatively, the C2 cap and the CNboundary may have been broadened by time averaging. Varicose instabilities resulting fromvortex shedding may have blurred or broadened the CN boundary while droplet vaporization mayhave blurred the cap of the C2 map. Further evidence for time averaging is provided by the extentof overlap between the molecular boundary region and the atomic plasma region. The overlapbetween the molecular boundary region and the atomic plasma region is greatest where thethickness of the boundary regions is greatest. It turns out that shedding ring vortices, or varicoseinstabilities in the flowfield, are the most plausible explanation for this broadening and overlap.Varicose instability will be discussed in later sections.30 25E S20a 0 C1Cu,.,0 010 5.C.RadialDistance, mmFigure5.3.2Anoverviewofspatiallyresolvedmapsof analyteandbackgroundemissionfromthesolventloadedICAP.CNM9ilCIMnIIMgMgMgIC2SoE C)-5C Cl)-lOö Cu-15cq00RadialDistance, mmLMThe Spatial Complexity... 146The next three maps in Figure 5.3.2 depict Ar I and C I emission. The C I and Ar Iemission maps define the volume of the atomic plasma (where atomic emission predominates)and reveal how it nests within the molecular boundary region (where molecular emissionpredominates). Two views of the atomic plasma are provided, within the induction coil andabove the torch rim. Within the induction coil, the C I isocontour reveals the toroidal inductionregion of the plasma. Above the torch rim, both Ar I and C I maps reveal that the toroidalinduction region coalesces into a tail cone. As mentioned earlier, the isocontours of atomicemission overlap with the boundary layer emission to a greater extent downstream from the torchrim than within the induction coil. It is possible that the temperature gradient downstream ismore gradual, but more likely that the downstream boundary is time averaged. At any rate, thedimensions of the atomic region, the overlap between the atomic and boundary region, and howthese spatial features influence analyte emission all deserve special attention.The three narrow maps on the right side of Figure 5.3.2 reveal the intensity distribution ofanalyte emission. The first two maps reveal the plumes of emission typically displayed by hardlines (atomic ion lines or atom lines with excitation potentials > 6 eV). Shown are the intensitydistributions of Mn 11(257.6 nm) and Mg II (279.5 nm). The upstream base of the hard lineplume begins at approximately 5 mm to 10 mm above the top of the induction coil. From thislocation, the plume extends 15 to 20 mm downstream, up to 25 mm above the induction coil.Typically, the hard line plume varied in length, width and intensity when one varied the operatingparameters. In spite of this, the plume retained its distinctive oval shape with no obvious taperingor constrictions upstream or downstream. The oval plume always displayed an unambiguousaxial maximum residing well downstream at 10 to 15 mm above the induction coil. The plumeextended radially to approximately 3 mm. In general, this maximum radius was found atapproximately 15 mm above the induction coil.In contrast, the third narrow map reveals the structure displayed by soft lines (atom lineswith excitation potentials <<6 eV). This map depicts the spatial distribution of emission from theMg I (285.2 nm) line, an atom line with an excitation energy of 4.35 eV. The most conspicuousThe Spatial Complexity... 147differences between this soft line structure and the hard line plumes to its left are the relativepositions of their maxima and their characteristic shapes. The maxima for the soft line structurelies significantly further upstream, or at a lower height above the induction coil than the hard linemaxima which reside above 10 mm above the induction coil. The bases of the hard line plumesalso lie well downstream so that the plumes display a characteristic oval shape. In contrast, thebase of the soft line plume appears to stretch out into a narrow haft which extends into the torch,giving it the characteristic club shape of spatially resolved soft line emission.The last, rightmost frame of Figure 5.3.2 shows a map of the ratio of Mg 11(279.55 nm)and Mg 1(285.21 nm) line emission. It reveals the characteristic trough in the thermal conditionsalong the axis of the discharge. The trough is most pronounced upstream and disappears furtherdownstream. This reflects the decay of thermal gradients in the plasma downstream from wherethe solvent material has completely dissociated. Chapter 4 discusses how this ratio reveals howthermal conditions in the plasma approach local thermal equilibrium. Discussion of how the ratioindicates the robustness of the plasma may also be found in Chapter 4. The spatial mapspresented in this chapter extend our understanding of how robust the plasma is, and how itapproaches local thermal equilibrium.Incidentally, Figure 5.3.2 clearly shows the value of recording complete spatial maps ofintensity from the analyte plume, the atomic plasma and the plasma boundary region. Readerscan immediately see both the boundaries and overlap between the different regions of thedischarge. If all researchers reported maps like this, then comparing results from differentlaboratories would be greatly facilitated.The Spatial Complexity... 1485.3.3 Comparison with Emission Maps Reported in the LiteratureFigure 5.3.3. presents Maessen and Kreuning’s diagram of the general emissiondistribution from pyrolysis products in the ICAP loaded with organic solvent [3]. Clearly, theresults presented in Figure 5.3.2 are consistent with those of Maessen and Kreuning, who arrivedat their diagram after recording more than 100 radial profiles of CN, C I and C2 emission over arange of observation heights. The structure of their CN, C I and C2 regions are all consistent withthose in Figure 5.3.2., except that much of their spatial detail appears to have been obscured.Figure 5.3.4. reproduces Dittrich, Brauer and Niebergall’s diagram for a water loadedICAP derived from equidensigrams of Ar I (360.6 nm), N2+ (391.4 nm) and Cu I (324.8 nm)emission [9]. Significantly, these were equidensigrams of line of sight intensity, not radiallyinverted intensity. In fact, Dittrich et al. radially inverted only a limited sample of their intensitydata. Even so, Figure 5.3.4. extends our view to ICAPs loaded with aqueous solvent, revealingmany similarities between the effects of organic and aqueous solvent loading. It also provides areliable basis for interlaboratory comparison, because it reveals the location of boundary regions,atomic plasma regions as well as analyte plumes.Spatial Complexity... 149(a) Chloroform, Solvent Load = 15 pmol/s10- 10 10-‘ CN,.. S I‘‘ C,5mm ; ‘20mmC2, 5 mm‘ above aboveabove SI coilcoil coil -5I_____44_____- 0-- I 0• •_‘I 0• - I0 10 0 10 0 10•0. Toluene, Carbon Load =50 iimol/s10 10 10-. C,5mm‘.2OmmIabove : above ‘ bove‘ -a\,5mm, CN,coil I‘ coil- coiIII-I..II‘4.0- I 0• s—I 0. —0 10 0 10 0 10Radial Distance (mm)—s 30I(b)0 c CCN20.4’I IillI aill‘Ii• I‘ci• f•o 10 i £-0‘I’I I1 I I-10 0 10Radial Distance (mm)Figure 5.3.3 Maessen and Kreuning’s spatially resolved profiles of emission fromsolvent pyrolysis products. (a) A sample of their radial profiles of C2, C I and CNemission. (b) Their general spatial distribution based on more than one hundred profilessimilar to those shown in (a). These figures are roughly adapted from the figures in Ref.[3].Spatial Complexity...10(a)(b)Figure 5.3.4 A summary of Dittrich, Niebergall and Brauer’s equidensitometry results.(a) equidensigrams of Ar I, Cu I and N2+ emission. (b) An overlay of the lowest intensityequidensites from (a).1503020-10 0 10 -10 0 10 10Distance from Axis (mm)The Spatial Complexity 151(a) (c)Figure 5.3.5 - The characteristic structure of emission plumes for (a) soft, (b)intermediate, and (c) hard line emission from the inductively coupled argonplasma. The bold line indicates the outer boundary of the plasma.000000(b)The Spatial Complexity... 152The spatial structure of analyte emission plumes revealed Dittrich et aL [9], Horlick et at.[8], [6]and Franklin et al. [4] is summarized in Figure 5.3.5. Soft lines, or atomic lines withexcitation potentials much less than 6 eV, display a conical plume low in the discharge (Figure5.3.5.(a).). This cone resides close to the inner boundary between the aerosol channel and theplasma. The inner plume may be surrounded by an outer halo, similar to those revealed byDitthch et at. Evidence for this outer halo has also been provided by Franklin et at. [4] and byprofiles of Fe I emission in Chapter 9. The outer halo appears as an outer wing on lateralemission profiles. In contrast, the image spectrograms and lateral profiles reported by Horlick etat. reveal that outer halos may not always be present [8]. This indicates that the outer halodepends on the individual experimental setup. Dittrich et at. have suggested that air entrainment,which depends on the torch geometry and ventilation of the room air, may be responsible forsuch large discrepancies between experimental results [9]. In contrast to the soft line plumes, thehard line plumes depicted in Figure 5.3.5.(c). occupy the hot apex of the tail cone. Intermediateline display intermediate plumes (Figure 5.3.5.(b).). The literature is replete with much morespatial information regarding the emission from the ICAP, although no reports are ascomprehensive or detailed as this report, or those by Dittrich et at. [9], Horlick et at. [8] andMaessen et al. [3].The Spatial Complexity... 1535.3.4. The Parametric Response of Emission MapsThe Parametric Response of Diatomic Background Emission: The Response of Emission fromthe Plasma Boundary RegionsThe eighteen spatially resolved maps of emission from the CN bandhead at 388.34 nm inFigure 5.3.6 reveal how CN emission responds to chloroform load and forward power. Note thetwo distinctive structural features of CN emission, the central plume and the conical outer mantle:The central plume is absent under conditions of high power and low solvent load (in the lower lefthand plots) but increases in height and extends upward to meet the outer mantle with increasingload and decreasing power (in the upper right hand plots). On the other hand, the outer mantle(which caps the atomic plasma or resides at the boundary between the plasma and thesurrounding air) appears to collapse inward and downward with decreasing power and increasingsolvent load. Also note the intensity response of the CN emission: From left to right, or withincreasing chloroform load, the CN intensity of the mantle increases almost linearly withchloroform load at all locations. The response of CN intensity to power is more ambiguous(power increases from the top row plots to the bottom row plots). Although the CN intensityincreases with forward power at specific locations (for example, z 10 mm and r = 6 mm), alinear increase at all locations is not evident. Indeed, at some locations the CN intensity appears todecrease with forward power.3.4mg/s4.2mg/s6.2mg/s7.4mg/s8.6mg/s10.0mg/s25.020.015.010.0•Figure5.3.6Isocontourmapsof CN(388.34nm) emissionintensityforachloroformloadedICAP.Theinner argonflowratewas0.8111mm.Thecontourintervalis0.1Xthemaximumintensityfor theentiresetof eighteenmaps.Evidentinthemapsareanouter,conicalmantleandaninner plume.1.001.25kW1.50kW30.05.0otcORadial Distance, mmThe Spatial Complexity... 155In order to reveal the response of CN intensity to chloroform loading and forward power,one may first remove the spatial confounds by integrating the intensity over all space. This isessentially what Browner et a!. have done [20]. Alternatively, one may choose a spatial locationwhere the structure of the plasma remains fairly constant even when the forward power andchloroform load are varied. On the axis and beyond the tip of the plasma(z = 30.0 mm, r = 0.0mm), the intensity response appears to be largely determined by how far the atomic plasmaextends downstream. Here the effects of power and solvent load are severely confounded by thespatial response. Similarly, at a position on the axis and at an intermediate observation height (z =15.0 mm, r = 0.0 mm), the intensity is also determined by the spatial response. Here it isdetermined by how far the inner plume extends downstream, and whether or not the plumeextends past z = 15.0 mm. In contrast, the structure of CN emission near z = 15.0 mm and r-5.2 mm appears to be relatively independent of power and solvent load.Because the structure here appears to be constant, this is a good location to examine theresponse of CN intensity to forward power and solvent plasma load. At 1.5 kW, the intensityincreases almost linearly with solvent load, indicating that excitation conditions in the CN mantleare constant and that the amount of solvent material determines the CN intensity. At lowerpowers, the CN response departs from direct proportionality and displays a maximum atintermediate loading. Evidently, high levels of solvent plasma load sap enough energy from thedischarge to lower the CN emission intensity. Moreover, it is likely that vortex shedding entrainsair into the argon stream, effectively mixing N2 with the argon and solvent material, so that thesolvent carbon combines effectively with the atmospheric nitrogen, and the thermal conditions inthe boundary region are intermediate between the plasma and the air—according to the respectivegas temperatures and heat capacities. Further discussion of the mechanism behind this responseis reserved for later sections. Note, however, that one can optimize the CN signal as a workingdiagnostic by selecting the right viewing location. z= 15.0 mm and r = -5.2 mm appears to be thebest place to monitoring solvent plasma during routine analysis. A similar analysis can beapplied to C I and C2 emission maps.The Spatial Complexity... 156In contrast to CN emission maps, C2 emission maps reveal the upstream boundary of thedischarge. This boundary is characterized by solvent pyrolysis and a steady recirculation eddyrather than air entrainment and vortex shedding. Both processes are evident in the C2 emissionmaps presented in Figure 5.3.7. This figure depicts C2 emission within the torch for an ICAPloaded with meta-xylene. For these maps, the meta-xylene load was 0.2 mg/s. the forward powerwas set at 1.25 kW, and the inner argon flow rate was varied from 0.6 to 1.1 Jlmin in 0.1 1/mmincrements. The inner argon flow rate increases from left to right. At low inner argon flow rates,the C2 emission wraps around the base of the plasma while the central plume of C2 emission onlyextends a short distance along the axis. It is unlikely that diffusion could account for the C2emission around the base of the discharge. On the contrary, a recirculation eddy near the base ofthe discharge, predicted by computer simulations, could account for convective mixing of solventmaterial with the outer argon stream. (This recirculation eddy is illustrated in Figure 1.3.)Evidently, such a recirculation eddy entrains solvent material quite effectively into the outer argonstream at low inner argon flow rates, thus reducing the load on the axial channel. However, whenthe inner argon flow rate is increased, the eddy is less effective at entraining solvent material.Interestingly, computer simulations predict that extremely high inner argon flow rates, the innerargon stream may actually sweep the recirculation eddy away. Consequently, the load on theaxial channel increases, and the central plume extends farther downstream, while the outer C2emission intensity decreases. Moreover, the entire profile settles down into the torch. Thisindicates that the plasma translates axially when the outer argon flow is loaded with solventmaterial. C I emission maps provide further insight into this apparent translation.L::Ip1-IDII-loItRadial Distance,mmFigure5.3.7Isocontourmapsof C2(516.56nm)emissionintensityforameta-xyleneloadedICAP.Theinnerargonflowratewasvariedfrom0.6111mmto1.1111mmin0.101/mmincrements.Thecontourintervalis0.1Xthemaximumintensityfortheentiresetofsixmaps.Evident inthissetofmapsisthedistributionof solventmaterialovertheargonstream.0.61/mm0.71/mm0.81/mm0.91/mm1.01/mm1.1I/rimn7.02.0C)-o.vC) C-8.0-13.0-18.00.0.000L1The Spatial Complexity... 158Maps of C2 emission from an ICAP loaded with other solvents at different rates ofsolvent plasma load are all consistent with the behavior depicted in Figure 5.3.7. In response tomethanol loading, the peripheral component of C2 emission was more intense than the centralplume, and the central plume extended a shorter distance downstream. Evidently, the inner argonstream laden with methanol vapour could not penetrate the recirculation eddy as effectively as aninner stream laden with heavier solvent molecules, because the methanol laden stream wouldhave had less momentum. Indeed, chloroform loading displayed the opposite response.In summary, two features of the flowfield are evident in the parametric response of C2and CN emission maps, vortex shedding associated with air entrainment, and a recirculation eddyat the base of the discharge which entrains solvent material into the outer argon stream. Theextent to which solvent material is entrained depends on the flow properties of the inner stream.In addition, the C2 and CN emission maps provide guidelines for using C2 and CN signals ascontrol diagnostics during routine anaiysis.5.3.4 The Parametric Response ofAtomic Background EmissionContrasting quite sharply with the CN results is the response of spatially resolvedemission from atomic carbon (from the C I line at 247.9 nm). The fifteen maps shown in Figure5.3.8 reveal how C I emission responds. Once again, the top row corresponds to the lowestpower of 1.00 kW, the middle row to the intermediate power of 1.25 kW, and the bottom row tothe highest power investigated of 1.50 kW. From left to right the chloroform load increases fromX mg/s to X mgls, so that the rightmost 15 maps of CN emission correspond precisely withthose shown here.I1.00kW1.25kW30.025.020.01.50kW15.010.05.0qp000ODd’oRadialDistance, mmFigure5.3.8IsocontourmapsofCI (247.61nm) emissionintensityfor achloroformloadedICAP.Theinnerargonflowratewas0.811/mm.Thecontourintervalis0.1Xthemaximumintensityfortheentiresetof+jfteenmaps.Evidentineachmapisthetoroidalremnant oftheinductionregionthatcoalescesintoacone.The Spatial Complexity... 160In contrast to the outer mantle and inner plume of CN emission, the spatial distribution ofC I emission fills in the conical void underneath the CN mantle and around the inner plume: Itassumes the shape of a toroid just above the rim of the torch and coalesces downstream into acone. This shape characteristic of the plasma region where atomic line emission predominatesover molecular emission, indicating that the C I emission emanates from hot, atomic plasma andnot from the molecular, flame like conditions of the boundary regions. Nevertheless, it isimportant to note that the spatial distribution of C I emission overlaps with the spatial distributionof CN emission. In other words, it does not fit perfectly into the conical void. Again, one mayturn to the complex dynamic processes at work in the tail flame to explain this overlap. It is clearthat a steady state explanation would be inadequate. Of course, there could be gradual, steadystate transitions between molecular and dissociated species, with the transition described byequilibrium concentrations, but the dynamic processes discussed above most not be ignored.The gradients in these C I maps qualitatively reveal how energy is transported from theouter regions of the plasma into the axial channel. It is not appropriate to describe these timeaveraged gradients in detail, except to note that the gradients decay downstream from the pointwhere the solvent material has completely dissociated. Comparing the inner plume of CNemission with the C I emission nicely illustrates this point.The Spatial Complexity... 1 6 1Figure 5.3.9 depicts C I emission from the induction region of the ICAP. The four mapson the left reveal the response of C I emission to inner argon flow rate and methanol load, whilethe four on the right reveal the response to inner argon flow rate and chloroform load. In each setof four maps, the inner argon flow rate increases from top to bottom, while the solvent plasmaload increases from left to right, and the isocontours are normalized to the overall maximumintensity. The maps reveal that the volume containing the C I emission shrinks in the axialdirection in response to an increase in solvent load or decrease in inner argon flow rate. In fact,the volume shrinks in the axial direction by 10 mm when the methanol load is increased and theinner argon flow rate is decreased. Observations reported in Chapter 3 indicate that the C Iemission volume corresponds closely to the plasma volume, so the redistribution of carboncannot account for the response. Moreover, Figure 5.3.8 reveals that vertical translation cannotaccount for this response. Electron density measurements presented in Chapter 8 reveal that anenormous increase in electron number density accompanies the axial shrinking. For methanolloading, the electron density between the top and middle turns of the induction coil increases from8.0 x 1015 cm3 to 1.3 x 1016 cm3 . Evidently, the maps in Figure 5.3.9 reveal a thermalpinch effect. The effect is less conspicuous for chloroform loading because chloroform is lesseffectively entrained into outer argon stream, for reasons discussed earlier. Alternatively, thebond dissociation enthalpy of the C—O bond in methanol may contribute much moresignificantly to the thermal pinch effect than C—Cl or C—H bonds.In summary, plasma decay, time averaging of the plasma boundary and the thermal pincheffect are all evident in the parametric response of C I emission maps.0.3mg/smethanol1.0mg/smethanol3.2mg/s CHCI3Ii”cOOQRadialDistance, mmFigure53.9Isocontourmapsof CI(247.61nm)emissionintensitywithintheinductionregion.Thefourmapsontheleftdepicttheresponsetomethanolloadandinner argonflowrate.Thefourmapsontherightdepicttheresponsetochloroformloadandinnerargonflowrate.Theisocontoursineachsetoffourhavebeennormalizedtotheirrespectivemaxima—thecontourintervalis0.lxthemaximumforeachset.Evidentinthesemapsareaxialandradialshrinkingoftheinductionregion.LJ(OQ010.05.0E E: -10.0 -15.0The Spatial Complexity... 1635.3.5 The Parametric Response ofAnalyte Emission MapsIn general, two distinctive responses were observed for maps of analyte emission. Oneresponse was observed for hard lines, or atom lines with excitation potentials > 6 eV and allatomic ion lines. This response reduces to the hard line behavior discussed in chapter four whenspatially averaged by line of sight optics. The other response was observed for soft lines, or atomlines with excitation potentials <<6 eV.The response ofhard line emissionFigure 5.3.10 reveals how hard line emission responds spatially to solvent load(Mg II 279.55 nm). At the lowest attainable solvent load depicted in the left hand frame, the hardline plume displays the lowest intensity, presumably because the low condenser temperaturerequired to trap the solvent vapour has also lead to significant sample loss. The entire plume andits intensity maximum also sit the furthest upstream at the lowest solvent load. With an increasein solvent load, shown in the second frame, the overall intensity increases significantly,presumably because the condenser no longer traps a significant amount of analyte. The intensitymaximum also moves marginally upstream, whereas the plume lengthens significantly so that itstip resides at 25 mm above the top of the induction coil. A further increase in solvent load resultsin the plume shown in the third frame. With this increase in solvent load, the top of the plumeextends no further than previously, and the intensity maximum only moves marginally upstreamby perhaps 1.5 mm. On the other hand, the base of the plume moves downstream by 3 mm andclear of the torch, while The overall intensity of the plume decreases significantly. In addition tothese major changes, the 0.1 isocontour has become narrower, possibly indicating that the analyteis confined closer to the axis. (plot all four with individual normalization) Further increasing thesolvent load to the maximum tolerable load (in this case, for chloroform) results in the plumeshown in the right hand frame. With this increase in solvent load, the plume continues todecrease in intensity and become narrower, while the downstream tip of the plume once againextends no further than 25 mm above the top of the induction coil. It is interesting to note thatthis downstream limit for the tip of the plume overlaps the downstream limit for the atomicRadial Distance,mm8.6mg/sCHCI30•aq30.025.020.00 0 a Cu 4-15.010.05.04.2mg/sCHCI37.4mg/sCHCI3Figure5.3.10Theresponseof ahardlineplume(MgII279.55nm) tochloroformplasmaload.The Spatial Complexity... 1 65plasma, as revealed by the C I maps and by the 0.1 isocontour for the CN maps. This suggeststhat analyte plumes are time averaged as well.In order to interpret the response of the hard line emission plume to solvent load, severalphysical processes must be taken into consideration. These physical processes determine thelocal concentration of atomized or ionized analyte, and the energy available to excite the analyte sothat it emits: Relatively far upstream from the plume, the analyte is essentially confined to theaerosol stream because undesolvated particles and droplets must follow the gas stream owing totheir minute inertial moments compared to their immense viscous drag. However, once theanalyte begins to desolvate and vaporize, it becomes free to diffuse across the stream lines of theplasma flow. It can disperse radially as it flows across the boundary region of the aerosol channeland into the plasma. Once in the plasma, the analyte continues to diffuse across the streamlines.As the analyte is transported upward and out from the axis by convection and diffusion, energy istransported upward and in toward the axis from the toroidal energy loading region by convection,radiation and heat conduction across enormous thermal gradients. As a result, the time averageddensity of analyte and the time averaged density of energy available to excite the analyte varyenormously throughout the plasma.To a first approximation, only two properties determine the local emission intensity ofhard line species. These are the time averaged density of analyte and the time averaged density ofenergy available to excite the analyte. Consequently, the maximum in hard line emission resideswhere the maximum amount of energy is available to excite the hard line species, and where thehard line species has not dispersed appreciably by any mass transport process. By similarreasoning, the limits of the hard line plume reside where the energy available to excite the hardline is cut off, or where the local concentration of hard line species is very low. Accordingly, theupper boundary of the hard line plume coincides with the plasma boundary, the radial limits ofthe plume are determined by the radial transport of analyte, and the base of the plume isdetermined by vaporization, atomization, and ionization processes that convert the analyte intohard line species.The Spatial Complexity... 1 66The response ofsoft line emissionThe contrasting response of a moderately soft line, Mg I (285.21 nm) is depicted inFigure 5.3.11. In contrast to hard line plumes, the soft line plumes depicted here do not extendpast the plasma boundary. On the contrary, they appear to be nested within the hard line plumes.In fact, it appears as though the soft line emission occupies a cooler temperature band in theplasma than the hard line plumes, a band that surrounds the hollow, inner boundary. Thissuggests that the geometry of soft line plumes can be explained in terms of norm temperatures,as Blades et al. have pointed out previously [5].The norm temperature for an optical transition is simply the temperature at which theemission intensity for that transition displays a maximum. In a thermal plasma, one generallyencounters a single maximum for line intensity with increasing plasma temperature because oftwo competing processes. First, with increasing temperature, electron collisions increasinglypopulate an excited bound state, for a particular line, according to the Boltzmann function. As aresult, the emission intensity for that line increases with temperature. Second, at sufficiently hightemperatures, the atomic species begins to ionize into the next ionization stage, thus depopulatingthe excited state. Alternatively, emitting molecules dissociate with increasing temperature, thusdepopulating molecular excited states. Overall, atomic, ionic and diatomic emission from theICAP can be roughly characterized by a norm temperature (if one ignores dynamic processes).4.2m!sCHCI36.2mg/sCHCI3RadialDistance,mm20.030.025.0E E a, 0 C15.0210.05.000oFigure5.3.11Theresponseof asoftlineplume(MgI285.21nm) tochloroformplasmaload.The Spatial Complexity... 168As it happens, the electron kinetic temperature in the plasma, the temperature whichlargely governs electronic excitation of plasma bound states, ranges from approximately 6000 Kto 9000 K. In general, the norm temperatures for molecular species fall below this range byapproximately 1000 K whilst the norm temperatures for soft lines fall within this range, and thenorm temperatures for hard lines exceed this range by approximately 1000 K (see, for example,Dittrich or Olesik). Consequently, molecular emission generally occupies the plasma boundary,soft line emission occupies diffuse temperature bands within the plasma (time averaged bands, ofcourse), and the most intense hard line emission may generally be found where the most plasmaenergy is available for electronic excitation.Magnesium Line Intensity RatiosMaps of the ratio of ion line to atom line intensity (Mg II 279.55 nm andMg I 285.21 nm) are presented in Figure 5.3.12. The reader may verify that they agree with theline of sight, axial profiles presented in Chapter 4. It is apparent in these maps that the coolregions lie close to the inner boundary, and that the excitation environment grows hotter towardsthe downstream limit, where the maps become discontinuous.4.2mg/sCHCI36.2mg/sCHCI37.4mg/sCHCI38.6mg/sCHCI3Figure5.3.1ZTheresponseofthemagnesiumlineintensityratio(MgII 279.55nmIMgIchloroformplasmaload.285.21nm) to20.0E E 0’C U) ci15.0•30.025.0RadialDistance, mm10.05.000 00The Spatial Complexity... 1705.4. DISCUSSION5.4.1 Physical Properties Revealed by Emission Intensity MapsPhysical processes including diffusion, heat conduction, radiation, and convection are allevident in the maps presented in the preceding sections, but they are all obscured by their complexinteractions with each other and by temporal averaging. For example, convectional processes andhuge thermal gradients obscure the diffusion of neutral species and the ambipolar diffusion ofelectron ion pairs, if diffusion constants or rates of diffusion are to defined at constanttemperature. Vortex shedding obscures diffusion beyond the exit of the torch while recirculationof the argon obscures diffusion at the upstream end of the discharge. Moreover, the huge thermalgradients between the induction region and the axial channel obscures diffusion near the axis ofthe discharge. True, the recirculation eddy and vortex entrainment is not immediately evident inthe time averaged maps, but they can be inferred from the results of computer simulations andexperimental work reported elsewhere. Indeed, all of the maps were consistent with the presenceof a recirculation eddy and vortex shedding, so these convectional features cannot be dismissed.If one accepts the recirculation eddy and the vortex shedding mechanism for airentrainment, one will note that they have profound effects on the physical characteristics of thesolvent loaded ICAP. For example, vortex shedding determines the downstream boundary of theplasma. It cuts the plasma off by folding cold room air into the argon stream and thusextinguishing the plasma. As a result, the tail cone of the plasma cannot be regarded as a regionwhere the plasma decays gradually and steadily owing to microscopic processes such as threebody recombination or radiative loss. On the contrary, one must consider the possibility of anabrupt, fluctuating, discontinuous limit at the downstream boundary of the plasma, more akin tothe bounds of the potential core in a round jet.The Spatial Complexity... 17 1The recirculation eddy has similarly profound effects on the discharge. It determines howthe solvent load is distributed over the argon stream. Hence, it determines whether the axialchannel or the induction region will be heavily loaded with solvent material. The balance betweenthese two extremes determines the temperature and density profile of the plasma gas downstreamfrom the torch. Hence it determines how energy is transported to the analyte. Clearly, both therecirculation eddy and vortex shedding must be taken into consideration because they are centralto the solvent load problem.Beyond mass transport by convection and diffusion, the effects of heat conduction andheat capacity are evident in the maps of C I emission from the induction region. Inspection ofthese maps reveals that the plasma volume shrinks in the axial direction in response to solventloading. This indicates that the discharge responds to solvent loading with a thermal pinch in theaxial direction rather than a simple translation downstream. Although such an effect may beobscured in the C I maps by the mass transport of carbon in the argon stream, the thermal pincheffect is corroborated by observations, comparison with similar effects reported in the literature,and by electron density measurements presented in Chapter 8.Phenomena related to heat conduction are also evident beyond the torch rim. Forexample, maps of the Mg II to Mg I intensity ratio show that solvent dissociation consumes theenergy transported from the outer regions into the axial channel. Much of the energy is thusconsumed at the very boundary, so the very center is kept relatively cool. Only once all thesolvent material has dissociated does an appreciable amount of energy flow from the outer regioninto the very center of the axial channel. However, the mechanisms for energy transfer from theouter regions into the channel are not clear, nor obvious. Even so, the maps provide no reason toinvoke anything beyond thermal transport or heat conduction. Any further argument aboutradiation trapping, the transport of metastable species or other nonthermal channels of energytransport must be regarded as speculative—after all, the results presented here are obscured bytemporal averaging.The Spatial Complexity... 172Beyond heat conduction within the plasma, the intensity of CN, C I and C2 maps indicatethat those emission features may provide effective avenues for radiative loss of energy from thedischarge. Although all the emission maps and absolute intensity calibrations provide all theinformation required to assess the degree of radiative loss by solvent dissociation products, theproblem was not pursued in this work.In summary, several physical processes and phenomena are evident in the emission mapspresented in this chapter. Moreover, the emission maps begin to unfold the complexity withwhich these processes interact. But the quantitative physical characteristics of the solvent loadedICAP remain obscure. Several experiments for removing the ambiguity and obscurity suggestthemselves, further demonstrating the value of surveying the spatial complexity of thespectrochemical source. Those experiments will be discussed in the following chapters.The Spatial Complexity... 1735.4.2. The Air Entrainment MechanismVortex SheddingVortex shedding has been observed for flame-jets as well as for round jets of cold gas,and is most readily observed when the jets are excited acoustically. Becker and Massaro havereviewed the literature on acoustically excited jets up to 1967 [21]. Their review brought themback to 1858, when Leconte[22j reported that a coal gas flame-jet jumped in response to certainnotes from a violoncello and suggested that, “We must look upon all jets as musically inclined.”The jumping flame was just a manifestation of vortex shedding. Soon afterwards, it was foundthat combustion was inessential for acoustic sensitivity in a jet—round jets of cold gas madevisible by smoke particles behaved in a similar manner. Later on, Lord Rayleigh analyzed theinstability problem [23], and employed stroboscopic illumination to study it [24]. He foundthat both varicose and sinuous instabilities could be acoustically excited in round jets. Thevaricose instabilities took the form of symmetric swelling and constriction of the jets diametersynchronous with the exciting tone while sinuous instabilities took the form of rhythmicundulation or twisting of the jet (vortex evolution is associated with the varicose instability).Varicose instabilities predominate in jets with a flat, or top hat velocity profile with a thinboundary region. Becker and Massaro’s literature review also reveals that vortex shedding andthe acoustic sensitivity of jets has been extensively studied in more recent times and that theassociated theory has also been developed extensively. But that is beyond the scope of this thesis.Becker and Massaro themselves have presented photographic records and detailedobservations of vortex evolution in a round jet [211. Their study focused on an axisymmetric jetof cold gas from a nozzle with a flat flow velocity profile at the nozzle mouth except for a thinboundary layer of laminar flow near the nozzle wall. Their study is very informative, because itspans a wide range of Reynolds numbers for the axisymmetric jet, including the range typicallyencountered for ICAPs. They divided the complete range of Re = 600 to 20,000 into eight flowThe Spatial Complexity... 174regimes. (Incidentally, the ICAP (Re = 100 to 600 ) resides in the first flow regime of 600 < Re<1450 [251, [26]. Indeed, displays behavior remarkably similar to Becker and Massaro’s coldjet in the first regime.) They found that successive vortices shedded off the jet according to ageneral frequency law (or wave velocity law). In general,-Econstant, (5.1)U0where f is the vortex shedding frequency, 2 is the wavelength of the varicose disturbance, andU0may be regarded as the centre line velocity of the jet. The constant is approximately 0.5, 50vortices shed off the jet at approximately half the velocity of the jet stream. Vortex sheddingfollowed this frequency law both in the presence and absence of acoustic excitation. Interestingly,they found that when the vortex shedding was excited by pure acoustic tones, discrete frequencyjumps were observed which turned out to be related to the resonant properties of the nozzle tube.(Would similar resonances in the ICAP be dependent upon torch dimensions?) Among otherthings, they found that varicose instabilities were prevalent for thin boundary layers, whereassinuous instabilities were prevalent for fully developed laminar flows. One further point worthmentioning is that they observed they transition from vortex shedding to turbulent flow (note thatvortices need not be turbulent). In certain flow regimes, successive vortices would collide, thenbreak up into turbulent eddies. In other flow regimes, the onset of turbulent flow residedupstream from the tip of the potential core. Which flow regime the ICAP is found, if it indeeddisplays similar phenomena, is not known.The ICAP displays many of the characteristics of an axisymmetric round jet in which onewould expect vortex shedding. Before the hot argon jet flows out of the confinement tube into therelatively stagnant air of the torch box, it’s flow is essentially laminar rather than turbulent. Itsflow profile also appears to have a thin boundary layer and a top hat velocity profile, except forthe axial channel [27]. Moreover, the argon exists as unionized gas—as a coolant stream—closeto the torch wall. So at least the outer flow bears similarities to a round jet prone to varicoseinstabilities. On the other hand, it may be dangerous to assume that the flow dynamics of hotplasma bear any similarity to those of cold argon. It should also be noted that the ICAP has bothThe Spatial Complexity... 175axial and tangential velocity components of flow. It is not understood how tangential componentsaffect varicose instabilities. In spite of these departures from a cold, axisymmetric jet with a flatflow profile, we can develop the following conjecture:if the ICAP indeed behaves as a round jet, the argon flow remains laminar just beyond theexit of the torch, but varicose instabilities arise in the cylindrical surface of discontinuity in theflowfield. As a result, the surface of discontinuity eventually folds into a toroidal vortex (asshown in Figures 6.1.1 and 6.1.2—see next chapter, pages 191 and 192) much like a breakingwave. As this vortex moves downstream, it entrains air and radically convolutes the cylindricalinterface between the air and the argon. As it continues on its course downstream, it grows inthickness, folds in on itself and continues to entrain more air into the argon jet at its interior; Asthe toroidal vortex moves upward, its inner edge makes contact with the hot plasma and likelyfolds cool material in with the plasma gas, likely quenching the plasma. In other words, theplasma is probably confined to the potential core of the jet, where the potential core is simply theregion of the flowfield unperturbed by air entrainment. In short, the plasma boundary is likelydefmed by a modulated limit or cutoff determined by vortex entrainment of cold air rather than bythermal or radiative dissipation of energy.The time averaged picture of this modulated plasma boundary is revealed by the0.1 isocontours of CN intensity in Figure 5.3.6. In the outer boundary region, the CN emissionmaps and the tail flame observed above the plasma are probably time averaged pictures of thevortex shedding. Within this region, the interface between air and plasma—over which masstransport by diffusion takes place—becomes radically convoluted, even though the flow may stillbe laminar (experiments with cold gas jets and diffusion flames reveal that the flow eventuallybecomes turbulent as the vortices collide with each other and disintegrate, at downstreamdistances of more than two jet diameters from the nozzle). The net result of the vortex sheddingis complex, pulsating mixing mechanism, with a frequency corresponding to the sheddingfrequency of the vortex rings, which modulates the outer diameter of the plasma. It seemsreasonable to conclude, then, that the overlap of the contours for CN and C I maps above theThe Spatial Complexity... 176torch rim is a consequence of temporal averaging, while the minimal overlap between C I and C2isocontours within the torch is characteristic of the steady plasma boundary there.To our knowledge, detailed pictures of this vortex shedding process are only available forcold jets and diffusion flames and not for plasma jets. But reliable experimental evidence showsthat the ICAP also displays vortex shedding. This evidence includes high speed movie frames inwhich vortex structures are plainly visible and noise power spectra in which bands correspondingto vortex shedding frequencies are unmistakable [13]. This and other evidence is discussed infurther detail in Chapter 6. Lastly, one of the most unforgettable observations of the ICAP is thatat high power and outer argon flow rates(2.25 kW, 15.0 1/mm.), the discharge begins to sing,emitting a very clear acoustic tone.The Spatial Complexity... 1775.4.3. Implicationsfor the Analytical PeiformancePhysical explanations and conjecture aside, the structure and behavior of the analyteplumes are critical to the analytical performance of the ICAP. Of particularly relevance are thespatial relationships between the analyte plumes and background emission, because the ratiobetween analyte signal and background emission is something the analyst would like tomaximize. Note that one may predict the signal to background ratio from emission maps. Oneneed only multiply the light collection efficiency for each spatial location by the intensity ofanalyte or background emission, where the light collection efficiency can be calculated by exactray tracing. After integrating over all space, one obtains analyte and background signals. In thisway, one could calculate the signal to background ratio for a variety of light collectionconfigurations, and optimize the signal to background ratio numerically.Also relevant to the analytical performance is the influence of vortex shedding and airentrainment Air entrainment influences the analytical signal in several ways. It introduces flamelike conditions to the periphery of the analyte plumes, rendering the analyte signal susceptible toall of the matrix interference effects normally encountered with flames. It also introduces noise bymodulating the analyte plume. It may even corrupt the analytical blank by entraining dust or otherpollutants. Finally, vortex shedding and air entrainment are sensitive to acoustic excitation andchanges in the flow dynamics of the room air. Consequently, environmental sound and changesin the flowfield of the room air may corrupt the analytical signal. Evidently, it would bebeneficial to eliminate air entrainment altogether.The Spatial Complexity... 1785.4.4 Suggested Improvements for the Design and Methodology ofICAP-AESThe fmal step in the rational strategy, proposed in Chapter 1, called for improving thedesign and methodology of the analytical technique after gaining insight into the complexity andphysical properties of the discharge. Here, we have only surveyed the parametric and spatialcomplexity, yet two improvements immediately suggest themselves: 1. Eliminate vortexshedding by tailoring the flow dynamics of the discharge, and perhaps eliminate molecularbackground. For example, an outer sheath of argon and oxygen could suppress vortex shedding,C2 formation and CN formation. It may also be possible to manipulate the plasma shape usingflow dynamics and improve the analytical performance. 2. Optimize the light collection envelopewith ray tracing calculations and maps of analyte and background emission. For example, thesignal to background ratio could be optimized.5.4.5. Diagnostic Utility ofSpatially Resolved Intensity MapsThe diagnostic utility of radially resolved, monochromatic images of emission intensitymaps, for both physical diagnostics and control diagnostics, is evident in this chapter. The mapsreveal that emission from C2, CN and C I may all be used as control diagnostics for solventplasma load, provided they are viewed at the appropriate location. For example, the C2 intensityis proportional to solvent load when viewed down the axis, while CN and C I intensity areproportional to solvent load when viewed off axis. On the other hand, for physical diagnosticwork, the C I maps within the induction region reveal that the plasma shrinks both axially andradially in response to solvent load. C2 emission within this region indicates how solventmaterial is distributed over the argon flowstream. Chapter 7 underscores the utility of radiallyresolved, monochromatic images by presenting maps of electron density calculated from maps ofthe absolute intensity of a single argon line.The Spatial Complexity... 1795.4. CoNcLusioNsFrom a physical perspective, a downstream, conical limit is evident in all maps ofemission intensity at locations above the torch rim. This limit, and the overlap betweenisocontours of different emitting species, may be attributed to time averaged, varicose instabilitiesin the plasma jet. In contrast, the flowstream within the induction region is apparently steady andunperturbed by fluctuations. In maps of this region, it is evident that the distribution of solventover the argon stream depends on the inner argon flow rate and the properties of the solvent.Briefly, less solvent material is distributed over the outer argon stream when the momentum(flow rate x density) of the inner stream increases. This response may be attributed to arecirculation eddy at the base of the discharge. It also evident in these maps that the plasmaresponds to solvent plasma load by shrinking both axially and radially. More about this shrinkingwill be revealed in Chapter 8.From an analytical perspective, the emission maps reveal the analytical utility of emissionfrom argon and solvent pyrolysis products. In particular, the three dimensional informationreveals the observation zones where the emission intensity of these species is proportional tosolvent load and plasma excitation conditions.In terms of the rational strategy proposed for this thesis work, the spatial maps presentedhere indeed offer many useful guidelines for pursuing physical diagnostic investigations. Thoseguidelines are reviewed in Chapter 7.The Spatial Complexity... 1 801. R.P. Feynman, R.B. Leighton, and M. Sands, The Feynman Lectures on Physics. Vol.Volume I. 1965, Reading, Mass.: Addison-Wesley Publishing Company, Inc (1965).2. G. Dube and M.I. Boulos, Canadian Journal ofSpectroscopy 22(3): p. 68-76 (1977).3. F.J.M. J. Maessen and G. Kreuning, Spectrochimica Acta, Part B, 44(4): p. 387-384(1989).4. S.R. Koirtyohann, C. Baber, and M. Franklin, Spectrochimica Acta, Part B, 31: p. 589-587 (1976).5. M.W. Blades and G. Horlick, Spectrochimica Acta, Part B, 36: p. 86 1-880 (1981).6. M.W. Blades, and G. Horlick, Spectochimica Acta, Part B, 36(9): p. 881-900 (1981).7. N. Furuta and G. Horlick, Spectrochimica Acta, Part B, 37(1): p. 53-64 (1982).8. 0. Horlick and N. Furuta, Spectrochimica Acta, Part B, 37(11): p. 999-1008 (1982).9. K. Dittrich and K. Niebergall, Progress inAnalytical Atomic Spectroscopy 7: p. 315-372 (1984).10. G.M. Hieftje, Spectrochimica Acta, Part B, 47(1): p. 3-25 (1991).11. Y. Yasutomo, et al., IEEE Trans. Plasma Sci. PS-9: p. 18 (1981).12. J.W. Olesik, J.W. and J.C.Fister III, Spectrochimica Acta, Part B, 46: p. 869-883 (1991).13. R.K. Winge et al., Journal ofAnalytical Atomic Spectrometry 3: p. 849-855 (1988).14. Hecht, E. and A. Zajac, Optics. 1 ed., Reading: Addison-Wesley. 565 (1979).15. P.B. Farnsworth, B.W. Smith, and N. Omenetto, Spectrochimica Acta, Part B, 45(10): p.1151-1166 (1990).16. R.S. Stair, W.E. Schneider, and J.K. Jackson, Applied Optics 2(11): p. 1151-1154.17. A. Savitzky and M.J.E. Golay, Analytical Chemistry 36(8): p. 1627-1639 (1964).18. Madden, H.H., Analytical Chemistry 50(9): p. 1383-1386 (1978).19. Press, W.H. and S.A. Teukolsky, Computers in Physics: p. 669-672 (1990).20. Pan, C., 0. Zhu, and R.F. Browner, Journal ofAnalytical Atomic Spectrometry 5: p. 537(1990).21. Becker, H.A. and T.A. Massaro, Journal of Fluid Mechanics, 31: p. 435-448 (1968).The Spatial Complexity... 1 8 122. J. Leconte, Phil. Mag. 15: p. 235 (1858).23. Lord Rayleigh, Phil. Mag. 15: p. 235 (1879).24. Lord Rayleigh, Phil. Mag. 17: p. 188 (1884).25. Benoy, D.A., Modelling of ThermalArgon Plasmas. Ph.D Dissertation, TechnischeUniversiteit Eindhoven (1993).26. J. Davies and R.D. Snook, Journal of Analytical Atomic Spectrometry 1: p. 195-201(1986).27. P. Yang and R.M. Barnes, SpectrochimicaActa, Part B, 13(4): p. 275-309 (1990).Chapter 6The Temporal Complexityof the Solvent LoadedInductively Coupled Argon Plasma6.1 INTRODUCTION6.1.1 ObjectivesThe previous chapters surveyed both the parametric and structural complexity of thedischarge. In doing so, they revealed where solvent load effects occur in parameter space andphysical space. They also set the stage for exploring the physical properties of the discharge—except for one thing—they did not survey temporal complexity. In fact, the results of theprevious chapters mask the temporal complexity because their results are averaged over timescales ranging from 1 second up to 10 minutes.True, time averaged measurements would suffice if the discharge had both a steady flowfield and a steady temperature field. Then one could assume that time averaged results wereequivalent to time resolved results. One might even assume that downstream events weretemporally related to upstream events. This would allow one to regard gradients across thestream lines as evolving in time as one translated the frame of reference downstream. Moreover,one could assume that local thermodynamic properties were constant in time. Then one couldmeasure them simply by increasing the spatial resolution until thermal gradients becameinsignificant. Indeed, many such assumptions—which could only be validated by a steadyflowfield—have been relied upon in the past.But fluctuations are known to disrupt the flowfield of the discharge. Perhaps the twoThe Temporal Complexity... 1 83most important fluctuations result from vortex shedding beyond the exit of the torch, and dropletvaporization within the axial channel.First consider vaporizing droplets. Relatively large droplets (approximately 25 im indiameter) of an aqueous aerosol may actually survive in the plasma all the way up to the analyticalviewing zone (z = 15 mm above the induction coil). Although these large droplets are statisticallyfew in number, they are significant in volume, so they contain a significant amount of analyte andsolvent material compared with the rest of the aerosol. When such large droplets are swept alongby the argon stream, their solvent evaporates and cools small regions of the surrounding plasma.When such regions of local cooling pass through the viewing zone, they cause the emissionsignal to fluctuate aperiodically on the sub millisecond time scale. Similar fluctuations resultwhen relatively large particles left behind by the large droplets fly through the viewing zonebecause they are surrounded by local concentrations of analyte vapor. In general, the fluctuationsresulting from aerosol vaporization are aperiodic and drastically alter plasma conditions on thesub millisecond time scale.In contrast to vaporization events, vortices set up periodic fluctuations when they shed offthe plasma jet. In fact, they modulate the emission signal at approximately 200 Hz.Experimental evidence suggests that vortices modulate both the outer boundary of the dischargeand the flowfield within the plasma. In fact, it appears as if vortices peristaltically pinch the flowof the axial channel as they shed off the plasma jet. At any rate, it is clear that both vortexshedding and aerosol vaporization disrupt the flow field and temperature field of the discharge.Consequently, time averaged measurements of plasma properties must be regarded as biased.This bias may be avoided by completely resolving measurements over both space andtime. But such an approach has not met much success to date. Alternatively, one could surveythe complexity of the temporal behavior to determine whether other approaches, say phaseaveraging, would suffice. The objective of this chapter is to survey the temporal complexity ofthe discharge. This survey, in combination with the spatial and parametric surveys of previouschapters, provides a rational basis for probing the physical properties of the discharge.The Temporal Complexity... 1 84Surveying the temporal complexity of the discharge essentially equates to identifying, thenexplaining the significant fluctuations in the flow field and temperature field. A number ofexperiments have been pursued to do just that. The following sections review these experiments,experiments including the measurement of noise spectra, the study of vaporizing droplets anddesolvated particles in the discharge, and flow diagnostics of the discharge.6.1.2 Noise SpectraAmong the first to explore the temporal behavior of the ICAP, Walden [1], and Horlickand Belchamber [2] studied the noise in the analyte signal by examining its noise powerspectrum. A noise power spectrum may be obtained by sampling the emission signal at anappropriate rate, and then taking the Fourier transform of the time varying part. Next, differentfrequency components of the noise, such as 1/f noise, white noise, whistle noise, and power linenoise, can all be identified and then traced back to their respective sources, such as a defectivepower supply, the gas flow dynamics of the discharge or nebulizer drift. Incidentally, Talmi et al.[3] had already applied this approach to other spectrochemical sources. They found that it was apowerful method for isolating sources of noise in spectrochemical methods.Following Talmi’s lead, Horlick and Belchamber obtained noise power spectra of theanalyte emission signal from an ICAP and identified several distinct components of the noise.For example, they identified a low frequency component which they attributed to sample flicker,or 1/f fluctuations of the analyte transport efficiency. They also identified an acoustic componentwhich they concluded was due to plasma rotation. They were led to this conclusion by theresults of an interesting experiment. Essentially, they used two monochromators and detectors tosimultaneously monitor signal fluctuations from an ICAP via two optical channels. One opticalchannel viewed the discharge at 900 with respect to the other. Remarkably, the acousticfluctuations for the two channels were 90° out of phase. Evidently, the plasma was rotating, or soBeichamber and Horlick concluded.Winge et al. later demonstrated that the acoustic frequency noise at 200 Hz resultedThe Temporal Complexity... 1 85from vortex shedding [4]. This conclusion has been corroborated by the noise power spectrareported by Houk et aL [5], Hieftje et al. [6], [7], Furuta et al. [8], Snook et al. [9] and LoosVollebregt et al. [10]. Of all this work, Winge et al. [4] provide the most convincing evidencefor the vortex shedding phenomenon. They found that the acoustic noise band at 200 Hzdisappeared when the outer tube of the ICAP torch was extended by 6 mm. The extensionprevented air from entraining into the argon stream and hence suppressed vortex shedding.Consequently, the acoustic noise band at 200 Hz disappeared from their noise power spectra forthe extended torch. They went further by capturing images of shedding vortices with a highspeed movie camera. The shedding frequency revealed by the movie frames matched a peak intheir noise power spectra. Although this finding proved that vortex shedding was responsible forthe acoustic bands discovered earlier by Walden [1], and Horlick et al. [2], one must note thatvortex shedding and plasma rotation are not mutually exclusive. For example, if the vorticeswere asymmetric, then the rotation imparted to them by the tangential argon flow would beevident as a phase shift. Hence, the shedding of asymmetric vortices is consistent with the resultsof Horlick and Belchamber?s dual optical channel experiment [2].Aside from the vortex shedding phenomenon, noise power spectra may reveal the effectsof droplet vaporization. In fact, noise spectra reported by Antanavicius et a!. [11] demonstratethat broad band noise (0 to 2.5 kHz) in ion line signals may be attributed to aerosol vaporization.In typical noise power spectra, however, this broad band noise is buried beneath the baseline ofwhite noise. But Antanavicius et aL lowered the white noise level in their spectra by suppressingthe shot noise of their photomultiplier [11]. In order to do that, they used high concentrations oftest analyte. These concentrations yielded large radiation fluxes for the analyte signal and allowedthem to suppress the photomultiplier shot noise by turning down the photomultiplier gain. Aftersuppressing the white noise, they compared noise spectra for test analyte introduced as vapor withthose for test analyte introduced in an aqueous aerosol. They found that for the frequency rangebetween 0 to 2.5 kHz, the noise for aerosol introduction exceeded the noise for vapor introductionby an order of magnitude. In short, they demonstrated that aerosol vaporization may beinvestigated with noise power spectra, provided large concentrations of test analyte are used.The Temporal Complexity... 1 86They suggested that the time scale of the vaporization event (500 ps) corresponded to theinverse of the noise bandwidth (2.5 kHz), and offered two explanations. A time scale of 500 msmay correspond to the pulse relaxation time of the analyte signal. In that case, the wave form ofthe pulse would be determined by the vaporization of analyte particles and the pulse durationwould follow Poisson statistics. On the other hand, 500 ms could be the observation time of theflowing aerosol (after Eckert [12]). In that case, one must assume that emitting particles survivethe entire distance over the observation window. Interestingly, one obtains a particle velocity of20 m/s when one divides their observation distance (slit height) of 10 mm by the time scale of500 ps. This velocity agrees quite well with typical centerline velocities in the ICAP. In short,some doubt surrounds the correct assignment of the broad band noise. Does it correspond toparticle lifetimes or observation times? Recall that similar doubt surrounded the assignment ofthe acoustic band at 200 Hz—did the acoustic band correspond to plasma rotation or vortexshedding?In summary, noise power spectra can reveal the presence or absence of temporalfluctuations owing to aerosol vaporization and vortex shedding, but beyond that, noise powerspectra do not reveal the processes or mechanisms behind the fluctuations.6.1.3 Studies of Vaporizing Droplets and ParticlesFarnsworth et al. [131 and Olesik et al. [14], [15], [16], [17]looked beyond noisepower spectra and examined the form of the transient signal itself. They did this for Mg I, Ca Iand Ba I emission lines, and discovered conspicuous emission spikes on the sub millisecond timescale. Their discovery led them to investigate the relationship between aerosol droplets and thetime resolved emission of low energy atom lines and atomic ion lines. By correlating both dipsin the ion signal and spikes in the atom signal with laser light scattering off incompletelydesolvated droplets, they proved that incompletely desolvated droplets can exist in a water loadedICAP, at least when the aerosol stream has not been modified by heating or desolvation.Moreover, both droplets and desolvated particles induce temporal fluctuations in the plasma.The Temporal Complexity... 1 87Olesik et al. pursued these experiments much further. They surveyed the parametricresponse of droplet induced fluctuations [16]. They also established relationships betweendroplet phenomena, emission profiles resolved in time and space, and time averaged axial profilessimilar to those presented in Chapter 4 [17]. These experiments allowed them to demonstratethat the parametric behavior of analyte emission could be explained in terms of dynamicphenomena rather than the popularly unquestioned thermal characteristics of an ICAP in steadystate. Horlick et al. followed a different tack [18]. They examined the auto-correlation ofdroplet modulated signals. They did this for different entrance slit geometries on theirmonochromator. This allowed them to determine the size of droplet induced disturbances in theplasma. The disturbances turned out to be about 1.5 mm across. In short, several experimentalmethodologies for exploring the dynamic behavior of the ICAP have evolved from the originalexperiment of examining the wave form of the analyte emission signal.6.1.4 Flow DiagnosticsFurther experimental techniques for exploring the dynamic behavior of the ICAP havebeen borrowed from experimental fluid mechanics. Techniques such as high speed photography,particle tracking and anenometry have been applied to the ICAP, while flow visualizationtechniques using Schlieren photography or Mie scattering have not been applied as far as weknow. Nevertheless, good examples of high speed photography have been reported by Winge eta!. [4], who have examined vortices shedding off the ICAP with a high speed movie camera.Their high speed movie frames clearly reveal vortex structures in the tail section of the ICAP.The wavelength of the vortex structure, the position where the varicose waves roll over intovortices, and periodic constriction of the analyte plume are all evident in the movie frames.Moreover, the vortex shedding is consistent with noise power spectra of the same discharge. Thefrequency of a prominent peak in the noise power spectrum corresponds to the frequency ofvortex shedding. (All of these results compelled Winge et al. to suggest that diagnosticmeasurements of the ICAP should be phase averaged rather than time averaged.) The findings ofWinge et al. were corroborated by high speed videos of the ICAP recorded by Furuta [19] andThe Temporal Complexity... 1 8 8high speed photography of the ICAP by Houk eta!. [5] and Grey [20]. In addition to vortexphenomena, the photographs reported by Houk et al. [5] capture images of a single aerosolparticle vaporizing.In addition to high speed photography, anenometry and particle tracking have providedfurther insight into the temporal fluctuations and flow dynamics in the ICAP. Particle trackingexperiments have revealed that the flow velocity along the centerline of the discharge isapproximately 25 mIs. Cicerone and Farnsworth measured velocities between 20 m/s and 40 m/swithin 1 mm of the axis. Interestingly, they found that the velocities increased with power andviewing height. Olesik et al. found that this centerline velocity was independent of the innerargon flow rate, a finding that suggests that the axial channel dilates in response to an increase inflow rate, in order to keep the axial flow velocity constant. More about the dilation of the axialchannel will be said in Chapter 8. Beyond particle tracking, Barnes et al. [21], [22] haveinserted anenometry probes into the jet to measure its velocity profile. For an ICAP confined byan extended torch, they found that the velocity profile was essentially flat except for a thinboundary layer and an axial maxima corresponding to the axial flow. These results indicate that anunconfined ICAP would be prone to vortex shedding because vortex shedding typically arises injets with flat velocity profiles. On the other hand, jets with fully developed flows (the boundarythickness approaches the jet radius), are prone to sinuous instabilities [23].6.1.5 Axisymmetric JetsThe work reviewed in the previous sections provides insight into the velocity profile of theargon jet, the effects of desolvating droplets and vaporizing particles, the acoustic noise resultingfrom vortex shedding, and the gross structures of shedding vortices in the tail flame of thedischarge. Moreover, these experimental findings have been extended by numerical simulationsof the flowfiekl and temperature fields [24]. But neither simulations nor experimental resultsavailable to date provide insight into how the flowfield of the discharge may be perturbed byvortex shedding. Consequently, the temporal complexity of the discharge remains unclear.Fortunately, a wealth of insight into vortex shedding has been provided by experimental andThe Temporal Complexity... 1 89theoretical work with axisymmetric jets and diffusion flames. Several reports of this work areworth noting. While Prandtl [25] provides a good introduction to fluid mechanics, Becker andMassaro’s [23] paper on vortex evolution in a round (axisymmetric) jet provides a good startingpoint. They illuminated an axisymmetric jet through a slit, in order to obtain images of the flowstructure for a narrow slice through the jet. Because the jet was laden with oil condensationsmoke, and because the light was stroboscopically synchronized with the vortex shedding, theywere able to obtain sharp, phase averaged photographs of the sagittal cross section of the jet.These photos clearly reveal how the vortex structures evolve as they travel downstream.Moreover, they reveal how vortex shedding varies as the Reynolds number is increased from 600to 20000, and how the structured, laminar flow of the vortices break down into turbulence (theICAP occupies a flow regime of Re = 200— 600). In fact, the phase averaged photos of thesmoke laden jet are far clearer than high speed photos of the ICAP, because the exposure is moreeven (neither under nor overexposed) and the boundaries of the flow structures are sharp.Complementary photos are provided by Subbarao and Cantwell [26], Longmire andEaton [27], Roquemore et at. [28] , and Dahm et al. [29]. Dahm et at. [29] supply laserinduced fluorescence images of coaxial jets. These images reveal how vortex formation may beattenuated by surrounding the jet with an outer sheath flow. Roquemore et at. [28] supply 2Dphotos of a diffusion flame illuminated by a sheet of laser light. The laser light was scattered offTi02 particles formed by TiCl4 in the fuel and air reacting with combustion products. Theirimages reveal flow structures both at the outer boundary of the jet (visualized by Ti02 particles)and close to the axis (visualized by combustion). Interestingly, the inner combustion region andouter vortex structures of a diffusion flame resemble the pinched analyte plume and sheddingvortices of the ICAP. Further insight into the inner flow field is provided by Longmire andEaton’s study of a particle laden jet [27]. Although the flow regimes they studied correspondedto Reynolds numbers much higher than in the ICAP, the velocity and density information theyobtained from their laser Doppler anenometry experiment reveal much about how the axial flowof a jet interacts with ring vortices shedding off the periphery. Finally, Subbarao and Cantwellprovide stroboscopic Schlieren photographs of jets with Reynolds numbers in the immediateThe Temporal Complexity... 190vicinity of the ICAP [26]. Although their photos only reveal the flow at the interface betweentheir helium jet (the jet was buoyant) and coflowing air, the sharp Schlieren images reveal thevortex structure clearly.The experimental work cited above is complemented by the numerical simulationsconducted by Martin and Meiburg [30]. Not only is their report lucidly written, it suppliesbeautiful and informative 3D graphics of their simulation results. Also complementingexperimentally recorded images are power spectra of the acoustic fluctuations in axisymmetricjets. These have been reported by Bridges and Hussain [31], and Kyle and Sreenivasan [32],and bear many similarities with the noise spectra of emission from the ICAP.Although the theory developed on the structure and evolution of vortices is beyond thescope of this thesis, the reader is urged to consult the references cited above because they providevaluable insight into the vortex shedding phenomenon discussed in this chapter. Moreover, it ispossible to construct a conceptual model of vortex shedding in the solvent loaded ICAP once thework on axisymmetric jets as been consulted. Such a model is illustrated in Figures 6.1.1 and6.1.2. Figure 6.1.1 depicts a three dimensional wave that forms at the surface of discontinuity ofthe flow field. As the wave crests, analyte material is drawn into the crest and pinched away fromthe trough, in much the same way that particles behaved in Longmire and Eaton’s experiment.Figure 6.1.2 reveals how this process takes place in the ICAP. The six frames in the figurecorrespond to different phases of the vortex train. The drawing were based on the photographsreported by Winge et al. [4], and compare the typical time averaged view of the discharge withthe phase averaged view. Evident in the figure is the peristaltic pinching of the analyte plume, aprocess which probably accounts for much of the overlap between the time averaged imagesreported in Chapter 5 of analyte emission and molecular emission from the boundary regions.The Temporal ComplexityFigure 6.1.1 The 3D varicose wave that rolls off the argon jet into a ring vortex. Theperistaltic bunching of the axial stream has been exaggerated to illustrate how the varicosedisturbance interacts with the axial stream.191YAFigure6.1.2Vortexsheddingfromtheinductivelycoupledargonplasma.Thephaseaveragedviewisshownontheleftsideofeachframe.Therightsideofeachframeshowsthetimeaveragedview.Theanalyteplumeisshadedgrey,theplasmaiswhite,andthedownstreamboundaryregionishatched.I0 0 0The Temporal Complexity... 1936.1.6 Chapter SummaiyOut of all the powerful experimental techniques designed to explore the temporalcomplexity of the discharge, the measurement of noise power spectra is the most appropriate forthis work. Rather than being specifically tailored to droplet vaporization or vortex shedding, noisespectra can test for the presence of several phenomena. Quantitative information may also beextracted from noise power spectra (for example, frequencies and noise levels), and when thespectral density is scaled to the right units, results can be compared with a growing body of noisepower spectra reported in the literature. Moreover, the spectra contain information relevant toboth the analytical performance and physical characteristics of the discharge. Finally, noise powerspectra can be easily measured over a wide range of parameters, so one may rapidly survey thetemporal behavior of the discharge over a wide range of experimental conditions. Such a survey,in combination with a parametric survey and spatial survey, can provide one with a solid basisfrom which to investigate the physical characteristics of the discharge.This chapter presents the salient features of noise spectra measured over three frequencyranges: 0- 50 Hz, 0- 500 Hz, and 0- 100 kHz. For the 0 - 500 Hz and 0- 100 kHz ranges, threeparameters were explored: solvent, solvent load, and observation height. Only the peristalticpump rate was studied for the 0 - 50 Hz range. Noise spectra were measured for low energyatom lines and ion lines of test analytes, and for molecular bandhead emission of diatomic speciesin the boundary region. Droplet disturbances are evident in both the noise spectra and transientwaveforms for the low energy atom lines. Evidence for vortex shedding is evident in all spectrathat span acoustic frequencies above 100 Hz.The Temporal Complexity... 1946.2. EXPERIMENTAL6.2.1 Noise SpectraAll transients were detected with a Hammamatsu R955 photomultiplier fitted to the 1meter Czerny Turner monochromator described in Chapter 2. The voltage applied to thephotomultiplier was set to 600 V (yielding a photomultiplier gain — 10), and the output wasamplified by a Kiethley 428 current amplifier. The amplifier gain was generally set to 106, butgains of i05 and i07 were also used. The built in low pass filter of the Kiethley 428 currentamplifier prevented high frequency noise from being aliased into the noise spectra. The cut offfrequency was taken as the reciprocal of the filter rise time, and the signal was sampled at theNyquist frequency, or twice the cut off frequency. For example, when the 0 to 500 Hz frequencyrange was investigated, the filter rise time was set to 1 msec, giving a cutoff frequency of 1 kHz.Consequently, the signal was sampled at 2 kHz. For these settings, the -3 dB point of the filterresponse was at 350 Hz, and the filter response rolled off at -60 dB per decade (-18 dB peroctave). Note that much sharper cut off filters are available, but the built in filter was adequate.Data were sampled using the RC Electronics Compuscope software and ISC-16 data-acquisitionboard. The DADiSP worksheet ( DSP development corporation) was used to calculate the noisepower spectra in batch mode with a command file.In general, transients 16K long were sampled, then divided into 16 records of 1024samples each. For each 1K record, the d.c. component was first removed by subtracting themean signal. Then the time varying component was apodized with a Hanning window. A fastFourier transform was then taken, and the sum of the squares of the real and imaginary parts gavethe noise power spectrum. An average of the 16 resulting noise power spectra gave a goodestimate of the noise power spectrum for the original 16K transient. Finally, the noise power, orThe Temporal Complexity... 195noise spectral density S, was expressed in a variety of units so the average noise spectra could becompared with literature results. Beyond expressing the noise power S in arbitrary units, thenoise amplitude A could be obtained by taking the square root of S, and then expressed inphotocurrent units (A s). It was also useful to nonnalize the spectral density to the intensity of thed.c. component in order to correct for differences in signal strength. A variety of unit scales wereavailable to do this. The results could be expressed in decibels, where the decibel scale representsthe ratio of the noise amplitude A to the d.c. photocurrent A0, dB 2Olog(A/). This facilitatedcomparison with ref. [4]. An alternative decibel scale was proposed by Houk et at. [5] andrepresents the ratio of the noise amplitude A to the amplitude of noise approached asymptoticallyat high frequencies (essentially the photomultiplier shot noise level). For comparison with theresults in ref. [11], the spectral density was expressed as noise power S, divided by the d.c.photocurrent A0, while for comparison with the results in ref. [7], the spectral density wasexpressed in reduced units of noise amplitude A divided by the d.c. photocurrent A0. In total,noise spectra in the literature are reported according to more than six different unit scales.6.2.2 imagingThe discharge was imaged onto the entrance slit with an image to object ratio of 2:1, andthe slit height was 1.5 mm, so the projected aperture of the slit height was 3 mm. Winge et al.[4] discuss how this dimension limits the spatial sampling frequency of periodic plasmafluctuations traveling past the observation zone. They point out that if the wavelength of aperiodic plasma fluctuation were equal to the projected slit height, then the fluctuation would betotally attenuated to the d.c. level. As it happens, a projected slit height of 3 mm yields aneffective spatial sampling frequency of approximately 3000 Hz, when the wave velocity of typicalperiodic plasma fluctuations (10 m/s) are taken into account. Ref. [4], should be consulted forfurther discussion of how both the spatial filtering of the slit-height aperture and the time domainfiltering determine the overall frequency response of the measuring system.The Temporal Complexity... 1 96The observation height was varied by translating the 1.5 mm slit-height aperture over theentrance slit. Neither the spatial response of the photomultiplier, nor the spherical aberration ofthe lens were taken into account.6.2.3 Emission ChannelsEmission signals of the Mn II (257.61 nm) line and the CN (388.34 nm ) bandhead weresampled at 2000 Hz to investigate the 0 to 500 Hz frequency range where vortex fluctuations canbe expected. Emission signals of the Ca I (422.70 nm) line and the Mg I (285.22 nm) line weresampled at 200 kHz to investigate both the form of sub millisecond transients and the 0 to 100kHz frequency range where droplet vaporization effects can be expected. Finally, the emissionsignal from the C2 (516.56 nm) bandhead was sampled at 100 Hz to investigate low frequencypulsations owing to the peristaltic pump.6.2.4 Test Solutions and Solvent LoadLoading by three solvents was studied: water, methanol and chloroform. The chloroformload was varied from 3.2 to 10 mg/s. the methanol load from 0.3 to 0.7 mg/s while desolvatedaqueous aerosol (0.15 mgls) from a MAK nebulizer was compared with undesolvated aerosolfrom a Meinhard concentric nebulizer (0.32 mg/s). The metal concentrations were 20 pg/ml (20ppm) in chloroform and methanol (as 2,4 pentanedionates), and 50 ig/m1 (50 ppm) in aqueoussolvent (calcium carbonate in dilute HC1, magnesium chloride and manganese sulfate).The Temporal Complexity... 1976.3 RESULTSNoise spectra covering three frequency ranges are presented here. First the 0 - 500 Hzrange reveals the acoustic features owing to vortex shedding. Then the 0 - 100 kHz range revealsthe broad band noise resulting from vaporizing droplets and particles. Finally, the 0 - 50 Hzrange reveals the low frequency noise associated with sample introduction.6.3.1 The 0 to 500 Hz Frequency Range: Vortex SheddingThe following four figures depict noise spectra for Mn 11(257.6 nm) line emission andCN (388.3 nm) bandhead emission from a solvent loaded ICAP. Noise spectra are shown atthree viewing heights above the induction coil: z = 6 mm, z = 12 mm and z = 18 mm. Figure6.3.1 depicts the response of Mn II (257.6 nm) noise spectra to chloroform load; Figure 6.3.2, theresponse of CN (388.3 nm) noise spectra to chloroform load; Figure 6.3.3, the response of Mn II(257.6 nm) noise spectra to methanol load; and Figure 6.3.4, the response of CN (388.3 nm) tomethanol load.Clearly, all of the noise spectra evolve in the same way with viewing height. In general,an acoustic noise band at 200 Hz dominates the spectra at all viewing heights. A subharmonicand two harmonics of the 200 Hz band emerge at 12 mm, then become comparable in magnitudeto the 200 band Hz at z = 18 mm. This evolution of acoustic noise with height is typical of thevortex shedding phenomenon for ICAPs sustained in short torches.The Temporal Complexity(a)I Iz.•V .iOC.20I I•30_______.40•50 I IZ 0 100 200 300400(b)I’Z 0 100 200 300 400C”IIII/C’,c::.t—CZ•Z 0 100 200 300 400Frequency (Hz)Figure 6.3.1 The effect of chloroform plasma loadon the 0- 500 Hz noise spectra for the Mn II line at 257.6 10 nm:(a) z 18 mm, (b) z = 12 mm, and (c) z = 6 mm above the induction coil.The Temporal Complexity 199(a)0 100 200 300 400(b)Z 0 100 200 300 4001:Z 0 100 200 300 400Frequency (liz)Figure 6.3.2 The effect of chloroform plasma loadon the 0 - 500 Hz noise spectra for CN emission at 388.340 nm:(a) z = 18 mm, (b) z = 12 mm, and (c) z = 6 mm above the induction coil.200The Temporal Complexity(a)Z 0 100 200 300 400(b)I_______________________________________0 100 200 300 400I I IZ 0 100 200 300 400Frequency (Hz)Figure 6.3.3 The effect of methanol plasma loadon the 0 - 500 Hz noise spectra for the Mn II line at 257.6 10 nm:(a) z = 18 mm, (b) z = 12 mm, and (c) z = 6 mm above the induction coil.Figure 6.3.4 The effect of methanol plasma loadon the 0- 500 Hz noise spectra for CN emission at 388.340 nm:(a) z = 18 mm, (b) z 12 mm, and (c) z = 6 mm above the induction coil.201The Temporal Complexity(a)0.o .io.20.30•.4o.I I0 100 200 300 4000•I EL__0 100 200 300 400(c)Iz C 000 200 300 400Frequency (Hz)‘V &The Temporal Complexity... 202The f/2 subharmonic of the 200 Hz band first emerges in the z = 6 mm spectra. Itbecomes more intense at z = 12 mm, and is joined by 3/2 f and 2 f harmonics at 18 mm. Anexplanation for the f/2 subharmonic was found by Winge et al. [4]. Their high speed, black andwhite movies revealed that vortex shedding alternates between strong and weak fluctuations. Onthe other had, the 3/2f and 2 f harmonics emerge when the varicose wave begins to crest. Thishappens immediately before the wave rolls over into a vortex. Note that additional harmonics arerequired to describe the wave as it crests, because its waveform clearly departs from a simplesinusoidal function. Above z 18 mm, the cresting wave rolls over into a vortex. When thishappens, Winge et al. [4] have found that 5/2 f and 3 f harmonics emerge. Undoubtedly, theseadditional frequency components are required to describe the evolving wave form as it departseven further from a simple sinusoidal function.All of the 0- 500 Hz noise spectra evolved this way. The only difference between the CNand Mn II spectra was a flicker component below 100 Hz displayed by the Mn II spectra at z =6mm. This 1/f noise probably resulted from sample flicker (drift and low frequency fluctuationsin the aerosol transport efficiency), but not from droplet vaporization. The next section revealsthat droplet vaporization results in broad band noise that spans a much greater frequency rangethan the 1/f components displayed here. In any case, the 1/f components in these spectra are farweaker than those reported in the literature. Interestingly, the noise spectra reported in theliterature are for sample introduction without desolvation. Evidently, the low levels of 1/f noise inFigures 6.3.1 - 6.3.4 corroborate what Maessen et al. [33], [34] observed—that the desolvatingcondenser filters out the sample flicker noise. Of course, one must also remember that sampleflicker depends quite sensitively on the design of the pneumatic nebulizer. Perhaps the MAKnebulizer is inherently quiet [35].Apart from the 1/f component, the similarity between the noise spectra for Mn IIemission and CN emission was unanticipated. While the CN noise reveals fluctuations in theboundary region where one would expect vortex shedding to prevail, one would expect the Mn IIThe Temporal Complexily... 203emission from the analyte plume to be removed from processes ii the boundary region. On thecontrary, vortex shedding modulates both the analyte plume and the boundary region to the sameextent. Evidently, the varicose instability is not confined to the boundary regions, but penetratesright down to the axial channel. For insight into how this happens, one must go beyond noisespectra.The high speed photographs reported by Winge et al. [4] reveal that shedding vorticespinch the analyte plume of the ICAP. A clearer picture of how shedding vortices affect the flowfield is provided by the large body of work on axisymmetric jets. This work reveals that ringvortices constrict the flow along the axis of the jet. Between successive vortices, the axial flow isforced radially outwards. In fact, experimental results suggest that shedding vortices bunch thematerial flowing along axis into packets. One experiment revealed that particles in anaxisymmetric, particle laden jet are bunched together between successive vortices [27].Although the vortex shedding frequency is independent of viewing height in Figures 6.3.1through 6.3.4, Figure 6.3.5 reveals that it decreases with solvent plasma load. Moreover, thedecrease is far steeper for methanol load than for chloroform load. In order to explain thisresponse, and reconcile it with reports that the frequency increases with r.f. power, outer argonflow rate, and a short extension (1 cm) of the outer tube, we propose the following model. Thewave velocity of the varicose disturbance is roughly half the velocity of the coolant stream.Consequently, any variation in the operating parameters that increases the velocity of the coolantstream also increases the shedding frequency. An increase in the outer stream velocity can beexplained in terms of two properties: 1., how much argon is carried by the outer stream; and 2.,how effectively the hot plasma transfers heat to the argon in the outer stream and therebyaccelerates the outer stream. Increasing the r.f. power increases the shedding frequency byaccelerating the coolant stream—more heat is transferred to the coolant stream. Increasing theouter flow increases the shedding frequency at first, because a greater volume of coolant argonmust flow through an outer annulus of constant area.5.06.07.0SolventPlasmaLoad,mg/sFigure6.3.5Theeffectofsolventplasmaloadonthevortexsheddingfrequency.Theerrorinthesheddingfrequencyof±3Hzwastakenfromthepeakwidthathalfmaximumfortheacousticband.Theerrorinthesolventplasmaloadof±5%wasdeterminedbytheprecisionofthecontinuousweighingmethoddescribedinChapter2.I220216212208204200196192 1881 84180176 172(3 CD ci) CL) 0:chlciroformloadmethanolload.— 0.01.02.03.04.08.09.010.011.012.0The Temporal Complexity... 205But with further increase, the outer flow ultimately levels off. It may even reverse as moreenergy is convectively lost from the discharge, and the induction region can no longer acceleratethe outer flow as effectively by transferring heat to it. Similarly, an increase in solvent plasmaload consumes plasma energy and decreases the plasma’s ability to accelerate the outer stream.This effect is more pronounced for methanol load because the plasma lifts and contracts axially(see Chapter 5). As a result, the axial distance over which the coolant stream comes into contactwith the induction region decreases with methanol load. Far less heat can be transferred to theouter stream over shorter distances. The converse argument applies to extending the outer tubeby 1 cm (Winge et al. observed that this increases the shedding frequency [4]). A 1 cmextension of the outer tube confines the outer flow next to the hot plasma for a longer distance, sothe outer flow accelerates more than in a shorter torch.Alternatively, the frequency dependence could be explained in terms of the area of theunobstructed annulus through which the outer stream must escape past the plasma. This areawould vary with the dimensions of the plasma. In reality, the true mechanism by which the outerstream velocity and shedding frequency vary is probably a combination of both the heat transferefficiency and the geometry of the discharge.6.3.2 The 0 to 100 kHz Frequency Range: Droplet VaporizationIn contrast to vortex shedding, it is not quite as simple to observe fluctuations owing tovaporizing aerosol, for two reasons. First, Olesik et al. [14], [15], [16], [17] have pointed Outthat droplet disturbances may be so numerous that they actually overlap and swamp one anotherout and yield an apparent d.c. signal. Second, Antanavicius et al. [11] pointed out that thewaveform resulting from particle vaporization is typically buried beneath the photomultiplier shotnoise. However, appropriate measures may be taken to expose the effects of aerosol vaporizationin the noise power spectra. Here, we consider how Antanavicius brought the broad band noiseowing to aerosol vaporization above the baseline of white noise in their noise spectra. Theirstrategy was essentially to use large concentrations of test analyte in order to increase the radiationflux of the analytical line. This allowed them to turn down the voltage applied to theirThe Temporal Complexity... 206photomultiplier, thus lower the photomultiplier gain and lower photomultiplier shot noise—asignificant component of the total white noise. Consequently, using large concentrations of testanalyte allowed them to lower the white noise in their final signal and see the broad band noiseattributable to aerosol vaporization.Essentially, the photomultiplier shot noise S, contributes significantly to the total whitenoise, but scales down with the square of the gain, g, so the total white noise may be reducedquite considerably by decreasing the photomultiplier gain. More precisely,=2engf’, (6.1)where e is the electron charge, n is the number photoelectrons ejected from the photocathode,and F’ is a factor that accounts for the fluctuations in the gain owing to statistical nature ofamplification (1.1 < F’ <1.4) [11]. Antanavicius et al. were interested in bringing the signalfluctuations owing to aerosol vaporization above the level of photomultiplier shot noise.Consequently, they wished to know how the ratio of the shot noise to signal (anodic current)scaled with the photomultiplier gain. Noting that the signal, or the anodic current, is given by thecathodic current multiplied by the gain, orI=10g neg, (6.2)where I is the anodic current and I is the cathodic current, one obtains the ratio of shot noiseSSh(, to the anodic current by dividing equation 6.1 by equation 6.2.= 2egF’. (6.3)Evidently, the ratio of shot noise to the anodic current scales proportionally with thephotomultiplier gain. Typically, values for SShO,’ exceed iW’3 A s when the photomultipliergain exceeds 106. It turns Out that such high levels of mask the broad band noise owingto aerosol vaporization. But by using high concentrations of analyte which resulted in highradiation fluxes, Antanavicius et al. were able to decrease the photomultiplier gain to 1 Bydoing so, they brought the white noise down below 5 x 1014 A s and unmasked the broad bandThe Temporal Complexity... 207noise in the signal for ion line emission (Call ).While the noise in the ion line signal that results from aerosol vaporization may lie buriedin the photomultiplier shot noise, the noise in low energy atom line signal that results fromdroplet vaporization is far more conspicuous. Consequently, the arguments applied byAntanavicius to particle vaporization should also apply to droplet vaporization. The noise powerspectra of low energy atom line emission should reveal the broad band noise attributable toaerosol vaporization. This can be ensured by using large concentrations and low photomultipliergains.On the other hand, it is possible for droplet disturbances to overlap and yield an apparentd.c. signal. The way around this problem is to extend the survey over a range where dropletevents are separated or absent. Then the conditions where droplet events coalesce should beevident, if it occurs.In this way, it was established that incompletely desolvated droplets are not present in theICAP when the aerosol has been desolvated. Moreover, the broadband noise attributable toaerosol vaporization was only encountered at low observation heights for an ICAP loaded withwater, without desolvation, as the following paragraphs describe.The Temporal Complexity... 208Figure 6.3.6 depicts noise resulting from droplet vaporization in a temporally resolvedsignal for emission from a low energy atom line (Ca I 422.70 nm). The signal is characterizedby intense spikes superimposed on a weak d.c. component. The d.c. component results fromatom line emission from the unperturbed plasma. In the unperturbed plasma, such atom lineemission is usually weak for elements with low ionization potentials because the thermalconditions are such that the analyte is > 99 % ionized. However, when a droplet vaporizes tocreate a region of local cooling, the analyte’s degree of ionization within the region of local coolingdecreases from > 99% down to 50% or even 0 %. Consequently, the density of neutral atoms ismultiplied, say by 10 to 50 times, and an intense emission spike in the atom line signal isobserved as the region of local cooling is swept through the observation zone. Of course, thisswitch-like, ionization effect is most conspicuous when the atom line has a low excitationpotential and the neutral analyte has a low ionization potential. Because they meet these criteria,atom line signals from the alkaline-earth elements are the most sensitive indicators of dropletperturbations in the plasma.It is recognized that the number of droplet disturbances traversing the observation zoneper unit time may be so high that the emission spikes could coalesce into an apparent d.c. signal.In our studies, however, no spikes were observed under the typically robust operating conditionsof our 1CM?. Spikes only began to emerge when the power and viewing height were decreased,or when the water load and inner argon flow rate were increased. In other words, emissionspikes began to appear as the plasma grew less robust. In fact, we could follow the increase inthe frequency of emission spikes, and their density never reached the point of coalescing into anapparent d.c. component.2000-D >%1000-LIA-______1112131411618192Otime,msFigure6.3.6Thetransient signalfortheCaIlineat422.673nm.Theviewingheightwasz=6mmabovetheinductioncoil,andthedischargewasloadedwithwaterat0.32mg/s(without desolvation).CThe Temporal Complexity... 21 0In general, most observed emission spikes have a peak width on the order of 100 JIs. Inview of the work done by Cicerone and Farnsworth [13] and Olesik et al. [14], [15], [16],[17], and considering that the projected slit height was only 3 mm, this time scale is undoubtedlythe observation time of the droplet disturbance rather than the life time of a vaporizing droplet. Inother words, the droplet disturbances survive for long distances compared to 3 mm, and flythrough the observation zone with a velocity on the order of 30m/s, a typical centerline velocityfor the ICAP, so the observation time 3 mm / 30 mIs 100 ps. Moreover, if a noise powerspectra were taken of a transient with emission spikes, then one would expect broad band noisefrom 0 to approximately 10 kHz — the reciprocal of 100 ts.Figure 6.3.7 depicts exactly that sort of broad band noise in an ICAP loaded with 0.32mg/s water, without a desolvating condenser interposed between the spray chamber and the torch.Significantly, the broad band noise rapidly subsides with an increase in viewing height, and wasonly observed for water loading without desolvation. Evidently, of the three solvents consideredin this work, the effects of desolvating droplets are confined to high levels of water loading.Further discussion of aerosol vaporization effects may be found in Chapter 4. Beyond the effectsof desolvating droplets, further conclusions cannot be drawn from the noise spectra measured inthis study. But the approach taken by Antanavicius et al. [11] offers promise for investigatingthe effects of vaporizing particles.ci) :3 C.) 0 0•1..Cl) CL) h, C.) Li, ci) Cl) 06mm1 o101 1 o1210-1310-14Frequency,HzFigure6.3.7NoisespectrafortheCaIlineat422.673nm.Thedischargewasloadedwithwaterat0.32mg/s(withoutdesolvation).05101520x103The Temporal Complexity... 21 26.3.3 The 0 to 50 Hz Range: Sample IntroductionFigure 6.3.8 reveals that fluctuations in sample delivery are evident in emission signals.Specifically, the pulsations of the peristaltic pump are evident in C2 emission. Moreover, the peakcorresponding to the pulse frequency is the dominant component of low frequency noise, and itsamplitude decreases with an increase in the pumping rate. Similar effects were observed foranalyte emission by Loos-Vollebregt and Goudzwaard. They point out several ways of dealingwith noise, including standard addition [Myers and Tracy] and modulated sample introduction[Steele and Hieftje]. Here, we only point out that the C2 emission signal can be used to monitorlow frequency noise.NoisePower,arbitraryunits0 0Co--‘r\)I\)001001001333xlOxlOxlOC 0 0 01 0 0—.rc.01oo000000000o 0 0 (ii b 0010(7100001\)I)C.)0010ooo.ci-..C/I 0C-C-.I1C/IL1 II IIP 0 P 01 0 0 01 9) 0C.)P 0 P CJ1 I\.) 01 9) 0 9) (71 009) 01C), 9’ 0 91 01C.)01 0 (71 91 0 1 (110 (71 0C)10) 0(Ti1 09) 01010) 0The Temporal Complexity... 2146.4. CONCLUSIONSThe physical properties, particularly the thermal state of the ICAP, are modulated by twomacroscopic processes. Aerosol vaporization modulates the temperature field aperiodically whilevortex shedding modulates both the temperature and flow fields periodically. Incompletelyvaporized droplets in the plasma can be eliminated by desolvation. It is not clear what the effectsof desolvation have on the vaporization of desolvated particles. However, it is clear that vortexshedding is always present torches without extension tubes. Moreover, the vortex sheddingfrequency depends on the solvent and the solvent plasma load, and modulates both the plasmaboundary and analyte plume of the discharge to a similar extent. In light of vortex shedding,future diagnostics should be phase averaged as Winge et al. have already suggested [4].Clearly, the survey of the temporal complexity of the discharge presented in this chapter isa valuable contribution to the rational strategy proposed in Chapter 1—it provides an importantguideline for further diagnostic work—that measurements should be phase averaged, or that theirinterpretation should at least account for varicose modulations and aerosol vaporization.Once again, improvements in the analytical technique suggest themselves at an early stagein the strategy. For example, CN emission, a potential source of spectral interference, could beeliminated by modifying the flow dynamics of the discharge to prevent air entrainment and vortexshedding. Perhaps a coflowing, outer sheath of argon and oxygen could be tailored to eliminatethe formation of diatomics that emit in the ultraviolet and visible regions.The Temporal Complexity... 21 56.5. REFERENCES1. G.L. Walden, Spectrochimica Acta, Part B, 35: p. 535-546 (1980).2. R.M. Beichamber and G. Horlick, SpectrochemicaActa Part B, 37(1): p. 17-27 (1982).3. Y. Talmi, R. Crosmun, and N.M. Larson, Analytical Chemistry 48: p. 326 (1976).4. R.K. Winge, D.E. Eckels, E.L. DeKaib and V.A. Fassel, Journal ofAnalytical AtomicSpectrometiy 3: p. 849-855 (1988).5. R.K. Winge, R.K., J.S. Cram, and R.S. Houk, Journal ofAnalytical Atomic Spectrometry6: p. 601 - 604 (1991).6. N. Furuta, C.A. Monnig, P. Yang and G.M. Hieftje, Spectrochemica Acta, Part B, 44(7):p. 649-656 (1989).7. N.N. Sesi, P.J. Galley, and G.M. Hieftje, Journal ofAnalytical Atomic Spectrometry 8: p.65-70 (1993).8. N. Furuta, Journal ofAnalytical Atomic Spectrometry 6: p. 199-203 (1991).9. J. Davies and R.D. Snook, Journal ofAnalytical Atomic Spectrometzy 1: p. 195-201(1986).10. M.P. Goudzwaard and M.T.C. de Loos-Vollebregt, Spectrochemica Acta Part B, 45(8):p. 887-901 (1990).11. R. Antanavicius, P. Serapinas, and P. Sirnkus, Journal ofPhysics D: Applied Physics22: p. 254-257 (1989).12. H.U. Eckert, Spectrochimica Acta, Part B, 40: p. 145 (1985).13. M.T.. Cicerone and P.B. Farnsworth, SpectrochimicaActa, Part B, 44: p. 897 (1989).The Temporal Complexity... 21 614. J.W. Olesik, L.J. Smith, and E.J. Williamsen, Analytical Chemistry 61: p. 2002 (1989).15. J.W. Olesik, and E.J. Williamsen, Applied Spectroscopy 43: p. 933 (1989).16. J.W. Olesik, and J.C. Fister ifi, Spectrochemica Acta Part B, 46(6/7): p. 85 1-868(1991).17. J.W. Olesik, and J.C.Fister III, Spectrochimica Acta,, Part B, 46B: p. 869-883 (1991).18. G. Horlick, G. and F. Qin, Federation ofAnalytical Chemistry and Spectroscopy SocietiesMeeting XVII. Cleveland, OH (1990).19. Furuta, N.J., Journal ofAnalyticalAtomic Spectrometry 6: p. 199 (1991).20. Gray, A.L. Journal ofAtomic Analytical Spectrometry 7: p. 1151-1153 (1992).21. R.M. Barnes and R.G. Schleicher, SpectrochimicaActa, Part B, 36: p. 8 1-101 (1981).22. R.M. Barnes, R.M. and J.L. Genna, Spectrochimica Acta, Part B, 36: p. 299-323 (1981).23. H.A. Becker and T.A. Massaro, Journal of Fluid Mechanics 31: p. 435-448 (1968).24. P. Yang and R.M. Barnes, Spectrochimica Acta Review 13(4): p. 275-309 (1990).25. Ludwig Prandtl, Essentials ofFluid Dynamics. London: Blackie and Son Limited (1952).26. E.R. Subbarao and B.J. Cantwell, Journal ofFluid Mechanics 245: p. 69-90 (1993).27. E.K. Longmire and J.K. Eaton, Journal of Fluid Mechanics 236: p. 2 17-257 (1992).28. L-D. Chen, L.P. Goss, W.F. Lynn and W.M. Roquemore The Structure ofJet DiffisionFlames, in Turbulent Reactive Flows, S.N.B.Murthy, R. Borghi, Editors., Springer-Verlag: Berlin (1987).29. W.J.A. Dahm, C.E. Frieller, and G. Tryggvason, Journal of Fluid Mechanics 241: p.371-402 (1992).The Temporal Complexity... 21 730. J.E. Martin and E. Meiburg, Journal ofFluid Mechanics 230: p. 271-318 (1991).31. J. Bridges, J. and F. Hussain, Journal ofFluid Mechanics 240: p. 469-501 (1992).32. D.M. Kyle and K.R. Sreenivasan, Journal ofFluid Mechanics 249: p. 6 19-664 (1993).33. F.J.M.J. Maessen, G. Kreuning, and J. Balke, Spectrochimica Acta, Part B, 41: p. 3(1986).34. F.J.M.J. Maessen and G. Kreuning, SpectrochimicaActa, Part B, 44(4): p. 387-384(1989).35. H. Anderson, H. Kaiser, and B. Meddings. Proceedings of the Winter Conference onPlasma Chemistry. 1980. San Juan: Heydon, London (1981).The Temporal Complexity... 21 8NotesChapter 7A Parametric Survey of Electron Densityin the Solvent Loaded Inductively Coupled Argon Plasma7.1 IntroductionDrawing on the insight provided by the previous chapters, this one begins to explore thephysical characteristics of the plasma region of the discharge; It surveys the electron numberdensity ne (the density of unbound plasma electrons) in the tail cone of the discharge. The surveycovers a range of inner argon flow rates, solvent load, and forward power. It extends over theentire volume of the plasma beyond the exit of the torch, from z = 5 mm to 25 mm, and r = ± 8mm. Moreover, it explores three different solvents—water, methanol and chloroform—and thencompares them with a pure argon ICAP. In order to facilitate such an extensive survey, e wasdetermined from the absolute intensity of a single argon line. This method required only twoemission channels, one for the line intensity and one to subtract the continuum background.Consequently, the amount of data required to cover a given set of experimental parameters wasfar less than for multichannel diagnostics, such as those based on line profile measurements.This convenience made it practical to explore a very extensive parameter space. The results ofthe survey reveal how e , the thermal conditions, and the transport of energy respond over anextensive, previously unexplored range of experimental parameters.A Parametric Survey ofElectron Density 220Before proceeding to examine how electron density varies over experimental parameters,we will first review the guidelines offered by the previous chapters for exploring the physicalcharacteristics of the discharge. Then we will consider why three particular plasma parameters,the electron density ne, the electron temperature Te and the non equilibrium parameter b, areparticularly informative in describing the physical characteristics of the ICAP. After brieflyreviewing how these plasma parameters have already been investigated for ICAPs loaded withwater, we will consider the theoretical basis for determining the electron density from theabsolute intensity of a single argon line, and discuss why this technique holds promise for theICAP. Finally, we will examine the results of an electron density survey based on the absoluteintensity of a single argon line.7.2 Guidelines Offered by Previous Chapters for Investigating the Physical Characteristics ofthe DischargeThe previous four chapters fulfilled the first step in our rational strategy—they revealedthe parametric, structural, and temporal complexity of the inductively coupled argon plasma. Indoing so, they provided guidelines for investigating its physical characteristics. In fact, theguidelines provided by the previous chapters suggest how one may most effectively explore thetemporal and spatial characteristics of the discharge. Moreover, they reveal which spectralfeatures and which spectral windows would be most useful for spectroscopic diagnostics, andwhat physical processes would be worthwhile investigating.For example, the noise power spectra presented in Chapter 6 reveal periodic fluctuationsin the plasma flowfield suggest. If spectroscopic diagnostic measurements were time averagedover those fluctuations, then the results could well be biased. But if the measurements werephase averaged or phase locked to the fluctuations (i.e.., sampled at discrete intervals along theperiod of the fluctuation), then the bias owing to time averaging could be avoided.A Parametric Survey ofElectron Density 221Beyond guidelines for exploring the temporal characteristics, guidelines for exploring thespatial characteristics were provided by the observations of Chapter 3, and by the radiallyinverted images of Chapter 5. Although the results presented in these chapters are all timeaveraged, they reveal the specific locations in the discharge where the effects of solvent loadingare most extreme, hence where further investigations should be directed. For example, theyreveal that the induction region is extremely sensitive to changes in solvent plasma load.Obviously, diagnostic measurements should directed there. On the other hand, the spatial surveyreveals that the plasma decay region further downstream must also be investigated—it is evidentthat plasma decay downstream from the torch rim follows the response of the induction region ina complex manner. In any case, it is clear that measurements cannot be restricted to a singleviewing height or radial position—the discharge does not respond as a static entity whosethermal conditions can be understood simply in terms of energy consumption, dissipation andmass transport. On the contrary, further investigations should follow the guidelines offered byChapters 3 and 5, and extend the diagnostic investigation over a wide range spatial dimensions,ideally over the entire discharge, in order to account for spatial translation and distortions.Further investigations should also follow the guidelines offered by the spectral survey inChapter 3 , a survey which revealed spectral windows and emission features which hold promisefor physical investigations. This survey extends from ultraviolet to near infrared emission, andreveals where one may find spectral regions free from complex background emission. In thosespectral windows, diagnostic measurements of analyte intensities would be free of spectralinterference from background signals. The spectral survey also reveals several intense atomicargon lines, as well as other spectral features that may yield diagnostic information.Significantly, the spectral survey reveals that many argon lines are free from spectralinterference. On the other hand, the survey reveals that the intensity of the H line, a usefulspectroscopic diagnostic signal for electron number density measurements, is much weaker inICAPs loaded by solvents containing small proportions of hydrogen (e.g., CHC13)than for thoseloaded with water or methanol. Moreover, the H line suffers interference from C2 emission.A Parametric Survey ofElectron Density 222Clearly, diagnostic measurements based on the intensity and spectral lineshapes of atomic argonlines offer advantages over other the H method examined in the next chapter.Lastly, the spatial results of Chapter 5 reveal that molecular emission from C2,Nj andCN are not suitable for probing the physical characteristics of the solvent loaded argon plasma.In fact, chapters 4 and 5 revealed that the most intense molecular emission originates from theboundary region or outside of the atomic plasma ( see also ref. [7] of Chapter 6 and refs. [3] and[9] of Chapter 5). Consequently, molecular intensities contain information from the boundaryregion, but generally do not reveal the physical characteristics within the atomic plasma. This isparticularly true for line of sight measurements. Diagnostic measurements based on molecularemission should therefore be reserved for plasmas supported in molecular gases, or confined tosituations in which the molecular emission is intense enough to supply information from theplasma region.Apart from offering guidelines for investigating the physical characteristics of thedischarge, the emission intensities reported in previous chapters offers qualitative insight into thephysical properties . Even though the information they provide is only qualitative, it gives agood indication of the fluid dynamics, at least when compared with the results of computersimulations reported in the literature. In fact, simulation results and temporally resolvedmeasurements led to the proposal of flowfield model of both steady and unsteady regions.Similarly, some indication of the transport processes in the boundary regions of the dischargewas provided by spatially resolved emission maps, as the C2 emission maps of Chapter 5 willattest. They reveal how solvent material is entrained into the outer argon stream. But suchinsight is only qualitative because emission intensities depend on both the local density of theemitting species and the local excitation conditions, properties which are species specific andunknown in general. Consequently, emission intensities alone leave us far from understandingthe physical characteristics of the discharge quantitatively. In fact, they raise more questionsthan they answer. But at least they bring us to the next step in our rational strategy—investigating the fundamental physical characteristics of the discharge.A Parametric Survey ofElectron Density 223In summary, the previous chapters offer valuable guidelines for further investigation.They even provide qualitative insight into the physical properties of the discharge. Now theproblem is to decide which physical properties or plasma parameters should be explored toobtain an adequate, quantitative, physical description of the discharge.7.3 An Accurate Physical Description of the ICAP: Plasma ParametersA complete physical description of the ICAP must account for enormous thermalgradients, inhomogeneity, complex energy balances and the four states of matter (solid, liquid,gas, plasma) assumed by the sample material. Moreover, it must describe the fluid dynamics, themass and energy transport processes, the collisional-radiative processes, and the thermodynamicstate of the discharge. Such a description would be impractical—if not impossible—to obtain fora system as incomprehensibly complex as the ICAP. Consequently, the most reasonablequestion to ask right now is—what do we need to know? Within the context of this thesis, thisquestion breaks down into the following four questions: 1. What physical properties of theplasma determine the analytical signal? 2. What plasma parameters define those physicalproperties? 3. How are they modulated by dynamic processes? 4. And finally, how do thoseplasma parameters respond to operating parameters, including solvent plasma load?We already know of two dynamic processes that disrupt the steady flow and temperaturefields—aerosol vaporization and vortex shedding. In Chapter 6, the effects of dropletvaporization were found to be limited to ICAPs loaded with undesolvated, aqueous aerosol. Theeffects of particle vaporization ( as distinct from droplet vaporization) have been shown to besignificant by other workers [11, [2]. That leaves vortex shedding to be dealt with, possibly byphase locking or phase averaging.A Parametric Survey ofElectron Density 224Beyond macroscopic disturbances, we already now a great deal about thermal state of thedischarge. (By thermal state, we simply mean the state of how energy quanta are distributedover matter according to thermal, or statistical distribution functions. ) This knowledge is veryuseful because the distribution of energy quanta over matter determines the analyte andbackground signal.It turns out that thermal state of the plasma is dominated by kinetics of the plasmaelectrons. This is true because inelastic collisions between atomic species and plasma electronsredistribute energy quanta over bound states more effectively than any other collisional orradiative process. In order to see exactly how the plasma electrons collisionally redistribute thebound states, we first need a description of the kinetic energies of the plasma electrons.Experiment has confirmed that in the ICAP, the kinetic energies of the plasma electrons follow aMaxwellian distribution [3]. Theoretical work supports this finding [4]. If the kinetic energiesof the plasma electrons indeed follow a Maxwellian distribution, then their kinetic energies maybe described by only one parameter—the electron kinetic temperature Te But how are thepopulations of bound states governed by the kinetics of the plasma electrons?A Parametric Survey ofElectron Density 225When 1. the internal energy states of atoms or molecules in the gas state are rapidlyredistributed by inelastic collisions, 2. the colliding particles follow a Maxwellian velocitydistribution, and 3. the particles do not interact with each other beyond their collisions, then it isa consequence of classical, Maxwell Boltzmann statistics that the internal energies of the atomsand molecules are distributed according to the Maxwell Boltzmann distribution [5], [6].According to this distribution, in the state of maximum thermodynamic probability, the numberof particles n in the ith energy level is given byg u./kT (7.1)where u, is the energy of the ith energy level, Z is the partition function for bound states in thespecies, k is Boltzmann’s constant, T is the temperature, and g1 is the degeneracy, or statisticalweight of the energy level (an integer that specifies the number of different internal states thathave the same energy). The three prerequisites listed above for the Maxwell Boltzmanndistribution are closely met by electrons and atoms in the ICAP. (For example, the electrondensity is far too low, < 1O7cm, and the temperature far too high, > 5000K, for quantuminteractions between electrons—consequently Maxwell Boltzmann statistics apply rather thanFermi Dirac statistics). Moreover, the internal states are limited to the electronically excited,atomic bound states.A Parametric Survey ofElectron Density 226In the ICAP, then, the Boltzmann distribution for bound states within a single ionizationstage may be written/exp,_EJ, (7.2)where n, and flq are the respective populations of the p and q excitation states, E and Eq aretheir respective excitation potentials, g and gq their statistical weights, k the Boltzmannconstant, and Texc the electronic excitation temperature.In the ICAP, the Boltzmann balance is maintained by excitation and deexcitation of thebound states by electron impact:Xq + e < Boltzmann > X, + e (7.3)Eq K.E.(1) E K.E.(2)where X,is an atom in state p , and Xq is the same atom in state q, and e is an electron with anunspecified kinetic energy of K.E.(l) or K.E.(2), where K.E.(1) differs from K.E.(2) by AE =(E — Eq). As a result of these collisions, the excitation temperature Texc which describes theBoltzmann distribution of bound states approaches the electron temperature of the plasmaelectrons.A Parametric Survey ofElectron Density 227If these collisional processes prevail over any other excitation or deexcitation process, and Txcindeed equals Te, then we can say that the plasma is in local Boltzmann equilibrium (LBE). Inthat case we may write=yex[;Eq)] (7.4)Here, Te was written in place of Tec to emphasize that the Boltzmann balance (equation 7.3) isdominated by the kinetics of the plasma electrons.Beyond the Boltzmann balance, the atomic state distribution functions are more generallydescribed by the Saha balance, which determines how energy quanta are distributed over thebound states of successive ionization stages. Similar to the Boltzmann balance, the Saha balanceis maintained by inelastic collisions between plasma electrons and atomic species.Consequently, the ionization temperature I for the Saha distribution approaches the electrontemperature. Unlike the Boltzmann balance, however, the Saha balance is maintained bycollisional ionization and three body recombination rather than collisional excitation anddeexcitation. The Saha distribution for bound states may be writtenfle = 2gitJ(22r1nekTe )expg, (7.5)where n is the ion ground state density, n1 the density of the ith neutral excited state, g1 and gthe respective statistical weights, E the ionization potential, me the electron mass, n thedensity of excited state i for the neutral, E its excitation potential, and h , Planck’s constant.A Parametric Survey ofElectron Density 228Here, Te was written in place of T to emphasize that the Saha balance (equation 7.6) tends tobe dominated by the kinetics of the plasma electrons. In fact, the Saha balance is maintained bycollisional ionization and the corresponding reverse process of three body recombination:X1 + e ( Saha > X + e + e (7.6)E K.E.(l) E K.E.(2)where X is an atom in state i, X is the corresponding ground state ion, and e is one of twoelectrons with an unspecified kinetic energy of K.E.(1), K.E.(2) or K.E.(3), where K.E.(l) differsfrom K.E.(2) + K.E.(3) by AE = (E — E). If this balance prevails, and‘rn indeed equals Te,then we can say that the plasma is in local Saha equilibrium (LSE).Beyond LBE and LSE, another convenient reference state for describing the thermal stateof the plasma is LIE, or local isothermal equilibrium, in which the kinetic temperature of heavyparticles (atomic species) T5 equals the kinetic temperature of the plasma electrons Te. If theICAP were in the theoretical state of local thermal equilibrium, we would have LBE, LSE, andLIE. In other words, Texc, 1ion’ Te , and Tg would all be equal.Departuresfrom LTEIt turns Out that the theoretical state of local thermal equilibrium (LTE) is quite useful indescribing the thermal state of the ICAP, because the ICAP approaches LTE very closely. But inorder obtain an accurate description of the thermal state of the plasma, and consequently anaccurate description of how excited states of analyte and background emitters are populated, weneed to know how the thermal state departs from LTE.A Parametric Survey ofElectron Density 229Departure from LIEIn reality, the kinetic temperature is much cooler for the heavy particles than for theelectrons. This is true because the different particles experience different forces and collisionalinteractions according to their mass and charge.In contrast to the heavy argon atoms and ions, unbound electrons absorb most of theirenergy directly from the electromagnetic field between their collisions with other particles. Theelectromagnetic field can accelerate the electrons to high kinetic energies, depending on themean free path between collisions and the local field strength. Moreover, because the electronscollide with heavy particles and other electrons much more frequently than the electromagneticfield changes direction, it is the collisions which determine how the population of electrons isdistributed over kinetic energy. As a result, the electrons follow a Maxwellian velocitydistribution. In fact, such a distribution has been verified experimentally for electrons in theICAP. This distribution yields a very important plasma parameter indeed, the electron kinetictemperature, also known simply as the electron temperature, Te.The heavy particles also follow a Maxwellian velocity distribution, but because they areheavy and predominantly uncharged, they do not gain a significant amount of energy directlyfrom the electromagnetic field. In contrast to the electrons, the heavy particles receive most oftheir kinetic energy indirectly, through elastic collisions with the more energetic electrons. As aresult, the mean kinetic energy for heavy particles lags below that of the electrons. This isparticularly evident where the field strength is high, say in the induction region. But furtherdownstream in the plasma decay region, the heavy particle temperature, (or gas temperature,Tgas) approaches the electron temperature, Te , very closely. In fact, in the decay region, it isgenerally assumed that O.9Te < Tgas < Te . In order to refine this approximation, Tgas can becalculated from Te by accounting for the discrepancy in kinetic energy between the heavyparticles and the electrons. For this calculation, only the electromagnetic field strength and themean free path are required. Details of such an estimation may be found in reference [6].A Parametric Survey ofElectron Density 230It is important to consider this departure from LIE because Tg and not Te determinesthe local density of atomic species (through Dalton’s Law and the popularly unquestionedassumption that the ICAP exists at atmospheric pressure).Departure from LBE and LSEBecause Dalton’s law determines the total population of atomic species over all excitationstates and ionization stages, a departure from isothermal equilibrium will cause the absolutedensity of plasma species to deviate from n = P/k Te—a more accurate density will be given byfl = P/kTgas, where Tgas typically differs from T by a factor of 0.8 (as a very rough estimate), asa consequence of the electromagnetic field imparting energy to the electrons more effectivelythan to the heavy particles. Accompanying the departure from local isothermal equilibrium aredepartures from local Boltzmann equilibrium and local Saha equilibrium.In general, departures from local Boltzmann equilibrium and local Saha equilibriumbecome clear when one compares the theoretical atomic state distribution function for localthermal equilibrium with the actual, or experimental ASDF. Such a comparison is shownschematically in Figure 7.3.1. In both frames, the solid line depicts the ASDF predicted from theSaha equation, and given values of n , the ion ground state population, and Te (not specified inthis schematic illustration). In the figure, the natural log of state density per statistical weight isplotted against the excitation energy from the ground state of the neutral atom, past the firstionization potential and ground state energy of the ion, up to the second ionization potential. Asa result, the Saha distribution for a given ne , Te and ion ground state population follows astraight line with slope- 1/kTe, an with a discontinuous jump at the first ionization potential.(Note that for a single ionization stage, the Saha distribution reduces to the Boltzmanndistribution. Also bear in mind that energy levels lie at discrete excitation energies—they do notoccupy a continuum of energy states as the curves in the figure suggest.)A Parametric Survey ofElectron Density 231In(n/g)Excitation Energy, e VFigure 7.3.1 The departure of experimental atomic state distribution functions from theSaha distribution. (a) depicts the case of an overpopulation of the low lying atomic levelsbalanced by an underpopulation of the upper ion levels. (b) depicts an underpopulation ofthe of the low lying atom levels balanced by an overpopulation of the upper ion levels.‘pExcitation EnergyIn(n/g)‘pA Parametric Survey ofElectron Density 232In contrast to the solid Saha line, the dashed, experimental ASDFs generally displaycurvature and depart from LTE. In general, their departure is manifest as either a generaloverpopulation or underpopulation of energy levels within a given ionization stage, with respectto the values predicted by the Saha distribution. Moreover, there are two general cases fordeparture from the Saha distribution. Figure 7.3.1(a) depicts the first case of an overpopulationof the low lying atom levels balanced by and underpopulation of the upper ion levels.Figure 7.3.1(b) depicts the second case of an underpopulation of the low lying atom levelsbalanced by and overpopulation of the upper ion levels.Note that in both cases, the departure from the Saha distribution tends to zero for energylevels close to first ionization potential. This happens because the atomic levels close to theionization limit are very closely spaced, so that collisions between plasma electrons and atomicspecies rapidly redistribute the energy quanta over neighboring levels—electron collisions are farmore effective at redistributing energy quanta over closely spaced levels, because many plasmaelectrons have kinetic energies on the order of the energy differences between the closely spacelevels and Rydberg states: collisional excitation and deexcitation bring neighboring energy levelsinto Boltzmann equilibrium with an excitation temperature Texc equal to Te, while collisionalionization and three body recombination bring the upper atomic levels into Saha equilibriumwith ion ground state, with an ionization temperature Tkrn equal to Te. The same cannot be saidof other energy levels, especially the low lying atomic and ionic levels which spaced furtherapart. Moreover, other processes compete with electron collisions to populate or depopulate thelow lying levels.Before considering the physical processes which depopulate or populate low lying atomicand ionic levels and hence cause the ASDFs to depart from LSE and LBE, it is worthwhile todefme a convenient parameter for describing the departure of experimental ASDFs from the Sahadistribution.A Parametric Survey ofElectron Density 233The nonequilibrium parameter b is defined as the ratio of the experimental population of anenergy level (per statistical weight) to the Saha, or theoretical local thermal equilibrium value.b = b• (expt) / b(LTE) (7.7)In general, each energy level has a unique value. Fey [2] discussed an interesting example of thisdependence of b on energy levels—in an ionizing argon plasma, such as the induction region ofthe ICAP, ions and electrons are rapidly lost via transport processes, so the thermal balance shiftssuch that ionization is favored, to sustain the population of charged species in the plasma. Inother words the atom levels are overpopulated with respect to the Saha distribution, whereas theion levels are underpopulated. Moreover, the degree of overpopulation increases towards theground state of the atom, to sustain the ionization flow from one level to the next. As a resultb1>b23... >b (7.8)for the atomic energy levels. On the other hand, for the ionic energy levels,< b < b3 <... <ba, (7.9)by the same reasoning. The opposite applies to a recombining plasma.More generally, one may consider departures from local thermal equilibrium owing totransport processes other than particle transport and to collisional radiative processes orimproper balances which compete with electron collisions. One may also consider apparentdepartures owing to temporal and spatial averaging. In fact, all of these sources of departurecomplicate matters considerably. Consequently, explaining departures from LTE is a dauntingchallenge. In addition to real departures from LTE that result from collisional radiativeprocesses, and transport processes, one can also expect apparent departures that result fromtemporal and spatial averaging. Moreover, different operating parameters take the plasma intoregimes where one or more of these sources of departure may dominate over the others. Beyondthat, the degree and direction of departure (value and sign of b1) depends on the atomic species.A Parametric Survey ofElectron Density 234As a result, there are considerable discrepancies between literature reports of the departure fromLTE.Fey et al. [71, [8] point out that the departure differs markedly between argon, solventdissociation products and analyte metal. The differences arise partly because the first excitedstates for metal atoms lie much closer to the ground state than for argon or hydrogen, but alsobecause the atomic state distribution function for analyte species may be sensitive to the hugegradients surrounding a vaporizing particle [9]. Beyond that, Fey [2] has summarized the mostimportant transport processes responsible for the departure of atomic state distribution functionsfrom local thermal equilibrium. Briefly, 1. energy input decouples the plasma temperatures ( T,> Tgo.s). Consequently, energy input significantly affects the atomic state distribution functionsbecause the Saha and Boltzmann balances are Te dependent; 2. Ion and electron transport out ofthe plasma by ambipolar diffusion leads to the ionizing system that was discussed earlier; 3.Radiation escapes from the plasma, especially via optically thin transitions between highlyexcited states. As a result, the Saha and Boltzmann balances shift towards lower energy states inorder to compensate for the depletion of high energy states. In this case, the lower energy statesare overpopulated by radiative decay and one encounters ASDFs similar to the one depicted inFigure 7.3.1(a); 4. Heat conduction drains large amounts of energy from the plasma, especiallyat locations high in the tail cone and above the toroidal remnants of the induction region. Thei,the atomic state distribution function for argon takes on the recombining character depicted inFigure 7.3.1(b).A Parametric Survey ofElectron Density 235In contrast to argon, the departure from local thermal equilibrium differs completely foranalyte species, as Blades and coworkers have revealed. Briefly, they explored the departure ofanalyte ASDFs by following two experimental strategies. First they compared the experimentalratios of ion to atom state densities with theoretical ratios [10], [11]. The theoretical ratiosassumed local Saha equilibrium, and were calculated from the Saha equation and experimentallydetermined values of electron density and electron temperature. The value for electron densitywas obtained from Stark broadening measurements of the H line ( see Chapter 8 ) henceinvolved no assumptions of local thermal equilibrium. On the other hand, the value for Te wasobtained from the electron density, where it was assumed that the argon ASDF was close to LTE.They found that the ratios of ion to atom state density were within a factor of five of thetheoretical values. Interestingly, they were infrathermal rather than suprathermal with respect thetheoretical values. In other words, the atom states were overpopulated with respect to the ionstates.Following a second strategy, Blades et al. looked beyond ion to atom state density ratiosand at the experimentally determined atomic state distribution functions [12], [131. Once again,they compared the experimental results with theoretical ASDFs calculated from the Sahaequation, experimentally determined electron densities, and electron temperatures determinedfrom the electron density. The comparison to theoretical ASDFs led them to attribute thedepartures from local thermal equilibrium to radiative decay of excited states and charge transferbetween argon ions and analyte atoms. Radiative decay resulted in a departure similar to thatdepicted in Figure 7.3.1(a). A model proposed to account for the departure was based on onlythree processes: collisional excitation by electron impact, collisional deexcitation by electronimpact, and spontaneous radiative decay [13]. In spite of its simplicity, the model predictedrealistic ASDFs. On the other hand, they found that the charge transfer process overpopulatedthe and 2S levels of the magnesium ion, in which case the simple collisional radiative modelfailed to predict accurate ASDFs for magnesium. This charge transfer effect had also beenreported by van der Mullen [141, and verified by Farnsworth et al. [15]. But Farnsworth haveA Parametric Survey ofElectron Density 236provided experimental proof that the charge transfer mechanism itself is balanced, and does notlead to departures from LTE.Beyond collisional radiative processes, Olesik and coworkers demonstrated that underoperating conditions normally used for routine trace metal analysis, the effects of aerosolvaporization cannot be ruled out as an explanation for the departure of experimental ASDFs fromtheoretical predictions [161. In particular, if measurements were temporally averaged overdroplet induced fluctuations, then the emission intensity from regions of local cooling would beaveraged with the intensity from the unperturbed plasma. This would result in an apparentoverpopulation of the energy levels close to the atom ground state because of the intensitycontribution of the emission spikes discussed in Chapter 4 and Chapter 6. Similarly, an apparentdeparture could result from averaging measurements over other fluctuations, such as vortexshedding. Spatial averaging may also result in an apparent departure, such as the excesscurvature in the ASDF for iron reported in Chapter 9. One should account for all of theseapparent departures before drawing conclusions about departures owing to transport processes orcollisional radiative processes. Clearly, the experimental determination of the nonequilibriumparameter b is sensitive to gradients and fluctuations, largely because the measurement of bothhigh and low energy states are required to define the departure of ASDFs from theoreticalequilibrium values.To summarize this section, a physical explanation of the analytical performance demands1. knowledge of the thermal state of the plasma, and 2. knowledge of how macroscopic processesmodulate the thermal state. Apparently, the two most important macroscopic processesresponsible for modulating the thermal state have already been identified—aerosol vaporizationand vortex shedding. As far as defining the thermal state goes, three plasma parameters arerequired—n, Te and b. Because the determination of b is fraught with experimental error, itseems prudent to begin investigations by surveying the electron density.A Parametric Survey ofElectron Density 2377.4 Determining Electron Number Densities from the Absolute Intensity of a Single Ar I LineOver the past decade, workers in the plasma and Atomic Physics group, at EindhovenUniversity, have collaborated with the Analytical and Inorganic Preparative Chemistry group atPhilips Research Laboratories, also at Eindhoven, in an effort to accurately describe the thermalstate of the ICAP. Over the course of their fruitful collaboration, they revealed that two keyplasma parameters, e’ and Te, can be determined quite accurately and reliably from the absoluteintensity of a single, high energy argon line [17], [18], [19], [20], [21].Figure 7.4.1 summarizes the method. First, the density of a high energy, atomic energystate is obtained from the absolute line intensity. This point corresponds to the filled circle justabove 14 eV in the figure. The state density per statistical weight 17q may be obtained from theEinstein transition probability for spontaneous emission A, the statistical weight of the upperstate gq’ the absolute intensity ‘qp’ Planck’s constant h, and the frequency of light Vqp?7 =-= ‘qp (7.10)qgq hv1,gqAqpThe method requires a second state density—the neutral argon ground state density. This can bedetermined from Dalton’s law if one assumes that the plasma gas behaves as an ideal gas,P = + n + n + n)kTgos + nekTe. (7.11)i>1 i>1A Parametric Survey ofElectron Density 2382010- Calculated Points1018 -.- Experimentally Determined16104.I101OCl)10“rI‘CoI I I I I I I0 2 4 6 8 10 12 14Excitation Energy, eVFigure 7.4.1 A schematic illustration of the method used to determine electron densityand electron temperature in the inductively coupled argon plasma from the absolute intensityof a single argon line.A Parametric Survey ofElectron Density 239Further assuming that the plasma gas exists at atmospheric pressure ( 1O Pa) and that thepopulation of all excited states and ion states are insignificant in comparison with the atomground state, one obtainsp (7.12)g1 kTgaswhere the statistical weight for the atom ground state equals one. For the moment, we willassume that Tg may be approximated quite satisfactorily by Te. Then inserting the expressionfor n1 (equation 7.12) into the Boltzmann equation (equation 7.4) yieldsTlq = [__)exp(jiJ. (7.13)Taking the natural logarithm of both sides gives(p”j Elnij =1nI—I—-----, (7.14)q kTe) kTefrom which Te may be obtained graphically, after obtaining 17q from the absolute intensity andequation 8. Then may be obtained from ijq Te and the Saha equation (equation 7.5) afterassuming that ne equals n (charge neutrality of the plasma gas):(7.15)We must consider the error resulting from departures from local isothermal equilibriumor LIE. The principal source of error here is evident in equation 7.13, where we assumed thatTgg may be approximated quite satisfactorily by Te. If we instead let T5 differ from Te, andintroduce a correction factor c = TgO/Te to account for the difference. Then Tg = CTe andA Parametric Survey ofElectron Density 240equation 7.13 may be written as follows after expanding the logarithmic term(P (PmlLkCTe) LkTe)(p Ein ij = lnI— I — + Inc. (7.16)q kTe) kTeExperimental evidence shows that 0.9 < c 1.0 in the ICAP, except in the induction region andat the edges of the plasma, so that the term in c is insignificant compared with the other terms inequation 7.16 [2], [22], [23], [24], [25]. In other words, departures from LIE contributeminimally to the error in e when ‘e is obtained from the absolute intensity of a single argonline.Nowak et al. took a similar route, except that they extrapolated a straight line through theexperimentally determined densities of several excited states to obtain an more accurate value ofi. They obtained the experimental state densities by measuring the absolute intensities ofseveral high energy argon lines.Whether one or many argon lines are used, the determination of ne from absolute Ar Iline intensities assumes that the ICAP is in local thermal equilibrium. We know that the ICAPdeparts from local thermal equilibrium, but as Raaijmakers first suggested in 1982 [17], realisticdepartures should introduce no more than ±10% error to values of ne determined from absoluteAr I line intensities, because under typical operating conditions and viewing locations [19], [181,[20] the ICAP can be assumed to be close to LTE. The following discussion explains how theclose to LTE condition limits the error in e•A Parametric Survey ofElectron Density 241In order to account for errors in the absolute line method that result from departures fromLIE, we can introduce the non equilibrium parameter b1 for the atom ground state. There is noneed, however, to introduce nonequilibrium parameters for the highly excited argon states,because electron collisions keep them in Boltzmann equilibrium with each other, and in Sahaequilibrium with the ion ground state and the continuum electron density. From the definition ofnonequilibrium parameter,nlb1 exp , (7.17)LTEwe obtain the LTE value for the ground state atom population in terms of b1 and theexperimental ground state population,expt PLTE = = . (7.18)V1 1e’1Thus b1 may be introduced into equation 7.14 as -mb1,p Eln (7.19)q kTe kTeFigure 7.4.2 illustrates the effect of the lnb1 term in the determination of e• For clarity, theoverpopulation (b1 = 10000) and the underpopulation(0.0001) of the atom ground state havebeen greatly exaggerated compared to the realistic bounds in the ICAP (0.1 < b <10). For b1>> 1, the atomic state distribution function curves up to meet n1 expt Consequently, thetemperature obtained by the single line method, and by assuming LIE, underestimates the actualelectron temperature and electron density. In contrast, the atomic state distribution functioncurves down to meet expt for b1 << 1, and the LIE assumption results in the opposite error.A Parametric Survey ofElectron Density 24215.76Figure 7.4.2 The effect of departure from local thermal equilibrium on thedetermination of electron density and electron temperature in the inductivelycoupled argon plasma. Note that the departure from LTE has been exaggeratedfor clarity.10180 Excitation Energy, eVA Parametric Survey ofElectron Density 243Although the absolute line method incurs error by assuming LTE, Figure 7.4.2 revealstwo properties of the method that limit the error in the electron temperature and densitydetermination. First, only the natural logarithm of b1 contributes to the error in the slope. As aresult, ln b1, in combination with the wide energy gap between the atomic ground state and thehighly excited state, only introduces an error of 10% in the electron temperature, even when b1 isallowed to range between 0.1 and 10. Second, the error in ij is significantly scaled down fromthe error in b1 LTE because flq is close to ij0,, and far from the atomic ground state.Figure 7.4.3 depicts more realistic departures from LTE. Here, b1 is bounded within anorder of magnitude (0.1 < b1 < 10). It turns out that these limits of departure limit the error inthe electron density determination to ±15% when the Ar I line at 687.129 nm is used. Of course,this is only the maximum error that could result from departures from local thermal equilibrium.Errors in the atomic transition probability (±15%) and in the absolute intensity determinationboth contribute further error.116-cj--b1=1O—A—bl;=O1•11510141013Eflqfromtheabsoluteintensityof the687.129nmArIline/101210111010 1 0 1/ /extrapolationtoionizationlimitassuming—0—1,1=114.815.015.215.415.6EXCITATIONENERGY,eVIIFigure7.4.3Errorintheabsolutelinemethodfor realisticdeparturesfromlocal thermalequilibrium.A Parametric Survey ofElectron Density 245For a better picture of the error intrinsic to the LTE assumption, Figure 7.4.4 reveals howfour key quantities of the absolute line method vary with the electron temperature of the plasmagas. Figure 7.3.4(a) depicts the atomic ground state density nlexpt versus electron temperature.Note that n1ex is fixed within ±15% of 1.05 x 1018 cm-3 over the range of electron temperaturesnormally encountered in the ICAP. As a result, the errors in the estimation of this quantity fromDalton’s Law should contribute minimally to the overall error of the method. Figure 7.4.4(b)depicts, the excited state density nq versus the experimental electron temperature. This quantityranges over two orders of magnitude, but can be determined experimentally, up to the error in theatomic transition probability, from absolute intensity measurements. Figure 7.4.4(c) depicts theelectron temperature calculated from“sq’ ‘1lexpt’ and equation 7.14 (the Boltzmann distribution).Of course, the calculated electron temperature equals the actual electron temperature for the solidLTE line (b1=1.0). Also shown are the partial LTE lines for b1=0.1 and b1=10. Evidently,departures from LTE can lead to ±15% error in the determination of the electron temperaturefrom absolute Ar I line intensities. Finally, 7.4.4(d) depicts the LTE and pLTE values for thenear continuum state density ij, versus the electron temperature. The error in this quantity isrestricted by the scaling property discussed earlier. Its error may be further reduced if several ArI lines are used in its determination [19], [181, [20].A Parametric Survey ofElectron Density 246Figure 7.4.4 Variation of four key quantities of the absolute line method overelectron temperatures typically encountered in the inductively coupled argonplasma.1.20x106 -1.15-1.10-1.05 -1.00-C’)I;I I1010wI 1 0a111x103 —10-I 9-8-7-1 0I,E00z1 087500 8000 8500 9000THEORETiCAL ELECTRON TEMPERATURE, K9500A Parametric Survey ofElectron Density 2477.4.5 depicts our final goal, the theoretical calibration curve of electron density versusabsolute line intensity. In order to generate this curve, a range of excited state densities nq werecalculated for a range of electron temperatures using equation 7.13. Next, a range of absoluteintensities were calculated from the nq values by equation 7.10. Finally, three ranges of electrondensity (for b1 =0.1,1.0, 10) were obtained by inserting the flq values, and three ranges ofelectron temperatures (Figure 7.4.4(c), equation 7.18) into equation 7.15.IIIIIIIIIIIIIIIII8b1=1(LTE)--—“b1=0.1(pLTE)::.:::;E.::.:::z.E:E:::b,=10(pLTE)E:::E8 :—tr-i89Figure7.4.5Electrondensityversusabsoluteintensityof asingleargonline(687.129nm).ThesolidlinewascalculatedassumingLTE.ThedashedlinesassumepartialLTEandsettherealisticboundsfordeparturefromLTE.oc1016III(7 ci i;j1015892,356110ABSOLUTEINTENSITY(687.129nm),mWcm423I-I576V100A Parametric Survey ofElectron Density 2497.5 ExperimentalRadially resolved intensity maps of the Ar I 687.129 nm line were obtained exactly asdescribed in Chapter 5. An intensity map of the continuum background at 690.000 nm wassubtracted from each Ar I 687.129 nm map in order to obtain the net intensity. (See spectrum,Figure7.5.1. Note that the bandpass of 0.4 nm was wider than the emission line profile.) Thenthe absolute line intensity was determined from the absolute response of the detector and thelight collection efficiency of the optical train. The absolute response of the vertical arraydetector was determined using a tungsten iodide lamp as an irradiance standard. Further detailsare provided in ref. [26]. The light collection efficiency of the optical train was determinedseparately with an exact raytracing procedure [27]. This calculation did not account forreflectance loss off the surfaces of the quartz lens. In practice, large determinate errors could bedetected by comparing the absolute intensity results with H line broadening results (presented inthe next chapter).2000-687.129nm1500-687.929nm696.543nm688.7lOnm696.023nm688.817nm695.146nm1000-693.767nmBackground500-692.501nmhA-0—686688690692694696698700WAVELENGTH,nmFigure7.5.1Theemissionspectraoftheinductivelycoupledargonplasmainthevicinityofthe687.129nmlineusedinthiswork.CA Parametric Survey ofElectron Density 2517.6 ResultsThis section presents an extensive survey of electron density in the ICAP above the torchrim for three r.f. powers (1.00, 1.25, and 1.50 kW), three inner argon flow rates (0.61 1/mmnebulizer argon + 0.00, 0.20,. and 0.40 11mm argon through the aerosol sheath gas adapter), andthree solvents (water, chloroform and methanol). For each solvent, the solvent plasma loadcovered a range of up to four settings, from the minimum obtainable up to the maximumtolerable. For each set of experimental conditions, a radially resolved map of electron density ispresented, covering a radial range from r +8 mm to r = -8 mm and an axial range from z= 5mm above the induction coil to r = 25 mm above the induction coil. Maps are missing for sets ofconditions where the plasma would have been unstable. The only exceptions to this rule are themissing maps for maximum water load and 1.00 kW r.f. power. The data for these maps wasinadvertently corrupted. The remaining maps have been arranged on nine pages, three for eachsolvent corresponding to the three inner argon flow rates. On each page, for a single inner argonflow rate, the r.f. power increases from the bottom row to the top while the solvent plasma loadincreases from left hand column to the to right hand column. On all pages, the left most columncontains maps of electron density in a pure argon plasma, for comparison. Because the spatialresponse of electron density in the maps is largely self evident, further description and discussionwill be brief.A Parametric Survey ofElectron Density 252General FeaturesAll of the electron density maps display a similar downstream, conical limit. This limit,presumably owing to vortex shedding and air entrainment, appears to recede further downstreamwith increasing power, advances upstream with increasing inner argon flow rate and bulges intoa cylinder at the bases under certain conditions. Apart from the response of the conical limit, theregion of low electron density along the discharge axis extends to different heights. Accordingly,the isocontours of high electron density for the remnants of the toroidal induction region archover to meet at the axis or bloom open. Note that the general features of the electron densitymaps presented in the following pages agree quite well for the electron density maps reported byHuang et a!. [22].The Response to Methanol Plasma LoadFigures 7.6.l(a—c) depict the response to methanol plasma load, r.f. power and innerargon flow rate. In general, the electron density increases with methanol load, except at lowpower and high flow rates. At low flow rates and high methanol load, the discharge appears torise out of the confinement tube. It appears to settle back into the tube with increasing flow rate.This may be perceived as an axial translation of the plasma region in response to methanol load,but detailed observations, the C I maps in Chapter 5 a