FUNDAMENTAL INVESTIGATION OF NONFLAME AND FLAME ATOMIZATION WITH COMPUTER-CONTROLLED SPECTROMETRIC SYSTEMS Dissertation for the Degree of Ph. D. MICHIGAN STATE UNIVERSITY AKBAR MONTASER 1 9 7 4 - Qn".~”fi LIBRARY Michigan 5 rate This is to certify that the thesis entitled FUNDAMENTAL INVESTIGATION OF NONFLAME AND FLAME ATOMIZATION WITH COMPUTER-CONTROLLED SPECTROMETRI C SYSTEMS - presented by AKBAR MONTASER has been accepted towards fulfillment of the requirements for 1’: ,. / I. F [I h 9 degree in C Lit/)0" >/ I; T //-”’ - ' ,7. ’ ffl/l fiéffl c/( WV / Major professor Date {/25/77] 04639 a? amSmo av V "IMF 8: SUNS' 300K BIMIERY INC. Llamnv amozns mmmr memes: o if t u lull-u... ,; 9.. I I! " WSW m ’j ‘1‘ ”WU/‘8‘ ‘ a It The use 0*“ -, fir, ‘. f 'i (w w 603“"? Hm 1 3L“ 7‘ Mme In W ,H.‘ ‘ . ‘ 1-3%? 3‘,- filtration“: at rm. n. ; I ' W91 ilt‘vcst‘Iqat (,5 .’ .-'-s aim. 53', .. ea -..- “.‘fiy' ‘fmactertu- I’VE Dhyiltut tin-'1 -.“H"‘.L ‘- "AVG—fifty: ': ti.) kl» ' m ‘fitef‘dctlfl! u-f {sevnfeL-sv'.‘ 5 595390 was flaw-£28! :v new first Nineteamr analyst» toffi’q‘m-se‘; is a . I— ,‘i ‘31 this tnvesttwtten. three as- nem".. - .ignmfl,_ Mr ‘—————————‘ ABSTRACT FUNDAMENTAL INVESTIGATION OF NONFLAME AND FLAME ATOMIZATION WITH COMPUTER—CONTROLLED SPECTROMETRIC SYSTEMS By Akbar Montaser The use of nonflame atomic vapor cells for atomic absorption (AA) and atomic fluorescence (AF) spectrometry has become increasingly widespread in recent years. The major use of these vapor cells is for analysis of nanograms of elements in small samples, such as in microgram or microliter samples. In order to take the full advantage of nonflame atomization, computer-controlled AA/AF spectrometric systems, which utilize electrically heated nonflame atomizers, were designed and evaluated. A fundamental investigation of nonflame atomization was performed in order to characterize the physical and chemical processes, to isolate and minimize the interaction of parameters, to design new atomizers and to introduce new fast multielement analysis techniques. As a direct result of this investigation, three new nonflame atomizers, four new heating techniques for nonflame atomizers and a new multielement technique, which can employ flame and nonflame devices, were proposed, fully evaluated and characterized. FUNDAMENTAL INVESTIGATION OF NONFLAME AND FLAME ATOMIZATION WITH COMPUTER-CONTROLLED SPECTROMETRIC SYSTEMS By Akbar Montaser A DISSERTATION Submitted to Michigan State University in partial fulfillment of the requirements for the degree of DOCTOR OF PHILOSOPHY Department of Chemistry l974 4 HIM. and arms-'1 .4‘~ ' ‘- pt humid iii-i: . c‘fifitwuw 2m _ S“ i ‘ To Eugene ,3? My Hife, Father. and Mother, LA éh’n “no, they made all the sacrifices. 2': 'Lgv':_‘;;‘ ."5 '7'1M‘1y. UTE 3‘1 ACKNOWLEDGMENTS The author gratefully acknowledges the guidance, encouragement and friendship of Professor S. R. Crouch during the course of this investigation. Thanks are given to Professor C. G. Enke for serving as his second reader, and providing helpful comments. He would like to express his gratitude to Michigan State University for granting him teaching and research assistantships. To Eugene Palermo, Dave Rothman, Phil Notz, Eugene Pals, Brian Hahn and Dave-Baxter thanks also go for their friendship and advice. Finally, the Author wishes to thank his parents for their unfalter- ing support, and, of course, his wife, Shirin, for making everything worthwhile, for making him smile, and for keeping him out of trouble. iii ‘___-'II!II-II----IIIIIIIIIIIIIIIlIII------------------------------.'lllll TABLE OF CONTENTS Chapter Page LIST OF TABLES ..................... xi LIST OF FIGURES ..................... xiii I INTRODUCTION .................... 1 II HISTORICAL ..................... 5 A. Atomic Absorption and Atomic Fluorescence Spectrometry .................. 5 B. Flame atomizers ................ 7 1. Advantages of Flame ------------ 8 2. Fundamental Disadvantages of Flame Atomizers ................. l0 3. Practical Disadvantages of Flame Atomizers ................. 1] C. Nonflame Atomizer ............... 12 l. Furnaces .................. 13 a. The L'vov Furnace ----------- l3 b. The Woodriff Furnace ---------- l5 c. The Massmann Furnace ---------- 16 d. Induction Furnaces ----------- 17 e. Furnaces with Continuous Sample Introduction .............. l8 f. Other Furnaces ------------- l9 2. Filament Atomizers ------------- 21 a. Gold, Silver and Copper Wire or Foil Atomizers ............. 2] b. Tungsten and Platinum Atomizers - - . - 22 Tantalum Atomizers ........... 23 d. The AA Microsampling System Involving Sampling Boats and the Delves CUP - - . 25 e. Carbon and Graphite Filament Atomizers ............... 26 ‘— Chapter Page 3. Cold Vapor Techniques .......... 29 4. Nonflame Atomizers Using Electrical Discharges ................ 29 a. Cooled Hollow Cathode ........ 29 b. Hot Hollow Cathodes ......... 30 c. Arc and Spark ............ 3l d. Plasmas ............... 32 e. Radiation Heating .......... 33 5. Controlled Explosion Atomizer ...... 35 D. Radiation Sources for Atomic Absorption and Atomic Fluorescence Spectrometry ....... 36 l. Radiation Source Requirements in AA and AF Spectrometry ........... 36 2. Laser Sources .............. 38 3. Pulsed Hollow Cathode Lamps ....... 41 E. Multielement Analysis ............ 44 III THEORETICAL DESCRIPTION OF NONFLAME ATOMIZATION AND ANALYTICAL SIGNAL .............. 48 A. Distribution Laws .............. 48 l. Planck's Radiation Law .......... 49 2. The Boltzmann Law ............ 49 3. The Maxwell Law ............. 50 4. The Saha Law ............... 50 5. The Mass Action Law ........... 50 B. Radiance Expressions ............. 51 l. Atomic Absorption Expressions ...... 51 2. Atomic Fluorescence Expressions ..... 52 3. Relationship Between n and Solution Concentration .............. 53 C. Atomization Processes in Nonflame Atomizers . 54 l. Atomization Rate ............. 54 2. Atom Populations in Furnace-type Atomizers ................ 57 3. Atom Populations With Filament-type Atomizers ................ 61 D. Forced and Natural Convection ........ 61 l. Forced Convection ............ 62 v Chapter 2. Dimensional Analysis for Forced Flow in a Furnace . . . . ........ 3. Buckingham IFTheorem ........... 4. Natural Convection in Nonflame Atomizers ................ 5. Forced and Natural Convection Combined in Nonflame Atomizers .......... E. Analytical Signal in Nonflame AA and AF Spectrometry ................. l. Temperature-Time Dependence . ._ ..... 2. Time Dependence of Atom Population. . . . IV INSTRUMENTATION ................. A. Introduction ................. B. Nonflame Atomic Absorption/Fluorescence Spectrometer ................. l. General Description ........... 2. Atomization System ............ a. Nonflame Atomizer Assembly ...... b. Gas Sheath Designs .......... c. Wire Loop Atomizers ......... 3. The Optical System ............ C. Multistep Current Programming Systems for Electrical Heating of Nonflame Atomizers. . . I. Introduction ............... 2. Hardware Programming System for Electrical Heating of Nonflame Atomizers ...... a. Principles ofOperation ........ b. Sequencing System .......... c. D-to-A Converter ........... d. Current Regulator .......... e. Procedure .............. f. Linearity and Reproducibility of the Hardware Programmer ......... 3. Computer-Controlled Heating of Nonflame Atomizers ................ D. Automatic Samplers for Nonflame Atomizers - - I. Introduction ............... vi .1...’ L. Page 64 65 69 7O 78 79 82 85 85 87 87 89 89 9] 92 93 94 94 95 95 97 104 104 105 105 108 Ill lll —————————‘ Chapter Page 2. Automatic Sampler - Design I ........ 113 a. Principles of Operation and Design Consideration .......... 114 b. Sequencing System ........... 117 c. Computer Interface ........... 119 d. Precision and Accuracy ......... 120 3. Automatic Sampler - Design II ....... 120 a. Principles of Operation ........ 120 b. Sample Turntable ............ 122 c. Stepping Motor Driven Micrometer Syringe ........... 122 d. The Purge Function ........... 124 e. The Fill Function ........... 127 f. The Deposit Function .......... 127 9. Computer Interface ........... 130 E. Integration Methods .............. 135 1. Variable-Time Analog Integration ...... 136 2. Fixed—Time Digital Integration ....... 136 3. Fixed-Variable Time Digital Integration ................ 138 F. General Description of the Program ....... 139 V FUNDAMENTAL INVESTIGATION OF NONFLAME ATOMIZATION FROM Pt and W WIRE LOOP .............. 145 A. Introduction .................. 145 B. Effect of Sheath Gas Flow Rate ......... 146 l. Atomization Efficiency ........... 149 2. Atom Residence Time ............ 149 3. Quenching Effect .............. 152 C. Optimization of Experimental Parameter ..... 158 1. Introduction ................ 158 2. Atomization Processes Without Pro— grammed Heating .............. 160 3. Optimization of Experimental Parameter for Programmed-Heated Wire Loop Atomizers ................. 162 vii ——_———‘ Chapter Page a. Introduction .............. 152 b. Preliminary Optimization ........ 153 c. Optimization Procedure ......... 153 d. Inner-Outer Flow Rate ......... 154 e. Inner Flow Rate — Atomization Temperature .............. 166 f. Desolvation Time - Desolvation Current ................ 156 9. Optimized Parameters For Cd, Zn and Hg ................... 165 D. Analytical Results ............... I69 1. System Stability .............. I70 2. Analytical Working Curves ......... I70 3. Detection Limits .............. 173 a. Results ................ 173 b. Discussion ............... 173 E. The Physiochemical Properties of W and Pt Atomizers ................... 179 1. Evaporation of the Atomizer ........ I80 2. Chemical Reactions ............. 180 VI THE GRAPHITE BRAID, A NEW NONFLAME ATOMIZER , , , , 183 A. Introduction .................. 133 B. Characteristics of the GBA ........... 185 1. Physical and Chemical Properties ...... 185 2. Braid Temperature ............. 187 3. Braid Emission Spectra ........... 190 4. AA and AF Signal Characteristics ...... 193 5. Mechanism of Atom Formation ........ 194 6. Braid Lifetime ............... 196 a. Sheath Gas Flow - Sheath Gas Type . . . 196 b. Sample Size - Sample Matrix ...... 198 c. Influences of the Heating Stages. . . . 198 C. New Methods for Programmed Heating of Non- flame Atomizers ................ 199 viii Chapter 1. Introduction ................ 2. Current and Voltage Programming ...... 3. Power Programming ............. 4. Radiation Programming ........... 5. Temperature Programming .......... 6. Influences of Heating Methods and Flow Rates on Atomizer Temperature ....... D. AA and AF Analytical Applications of the GBA. . Analytical Curves and Detection Limits. . . 2. Matrix Effect Studies with the GBA ..... 3. Application of the GBA to Real Chemical System .............. E. Conclusions .................. VII TIME-DIVISION MULTIPLEXED ATOMIC FLUORESCENCE SPECTROMETRY .................... A. Introduction .................. B. Time-Division Multiplexed Atomic Fluorescence Spectrometry .................. 1. Principles ofTime-Division Multiplexing ................ 2. Multichannel Atomic Fluorescence Spectrometry ................ 3. Instrumentation .............. a. Atomization System ........... b. Radiation Sources and Computer Interface ............... c. Detection System and Computer Interface ............... d. General Software Description ...... Analytical Results ............... Burner Parameters ............. N—J Pulsed Source Characteristics ....... Flame Analytical Results .......... 4500 ix Nonflame Analytical Results ........ ' Page 199 200 204 210 218 221 223 224 237 239 242 245 245 247 247 249 251 251 261 263 263 267 270 272 T""""""""""""""""""""""""""""""""""""'"""""""""'lllll Chapter Page 5. Applications ofFiber Optics in Atomic 275 Absorption and Atomic Fluorescence Spectrometry ................ 275 VIII PROSPECTIVES .................... 277 A. Improvements in Instrumentation ........ 277 B. Future Application of the GBA ......... 278 REFERENCES ........................ 280 x " Table LIST OF TABLES Page Physical Quantities and the Dimen- sional Equations for a Heated Tube ....... 66 Influences of Atomization Temperature and Sheath Gas Type on Natural Con- vection ..................... 73 Influences of Atomization Temperature and Sheath Gas Type on Forced Convection . . . . 74 Influences of Atomization Temperature and Sheath Gas Type on the Relative Magnitude of Natural and Forced Convection ................... 74 Initialization and Optimization Questions .................... 140 Experimental Conditions in the Investi- gation of Nonflame Atomization ......... _ 147 Optimum Platinum Loop Atomizer Parameters Using Programmed Heating ............ 169 Comparison of AF Detection Limits (ng/ml) with and Without Programmed Heating of Platinum Loop .................. 173 Atomic Absorption Detection Limits (pg/m1) With Programmed Heated Platinum and Tungsten Loop Atomizers ............. 175 Physical Properties of the Graphite Braid Atomizer ................. 186 A Typical List of Major Elements in the Graphite Braid ............... 186 Experimental Conditions for the Investiga— tion of Various Heating Techniques with Experimental Paremters in the Study of Graphite Braid Atomizer in Atomic Fluores- cence Spectrometry ............... 225 xi Table 20 21 22 23 Experimental Parameters in the Study of Graphite Braid Atomizer in Atomic Absorption Spectrometry ............ GBA Detection Limits in Atomic Fluores- cence ..................... GBA Detection Limits in Atomic Absorption- - - - Percentage Depression Caused by Ten-Fold Amount of Matrix Ion on Analytical Signal .................... Experimental Parameters for Analysis of Iron and Copper with GBA ---------- Experimental Results of Analysis of Iron and Copper with GBA ........... Relative Lamp Intensity as a Function of 0N Time .................. Detection Limits Obtained with Computer Controlled Multielement Flame AFS System .................... Fluorescent Signal Obtained with Multi- element and Single Element Solutions ..... Detection Limits Obtained with Platinum Loop and Graphite Braid Atomizers ....... Page 226 227 228 238 243 . .243 269 271 273 274 """-IIIIIIIIIII--------------------------------—————_____.................lll‘lll LIST OF FIGURES Figure Page 1 Correlation of data for free convection heat transfer from horizontal cylinders in gases and liquids (271) 71 2 Influence of atomizer temperature and sheath gas type on natural convection 75 3 Influence of atomizer temperature and sheath gas type on forced convection 76 4 Influence of atomizer temperature and sheath gas type on the relative magni- tude of natural and forced convection 77 5. Block diagram of the computer-controlled AA/AF spectrometer 88 6 Schematic diagram of the atomization chamber with a wire loop atomizer located on top of an adjustable gas sheath 9O 7. Block diagram of the multistep current programmer 96 8 General block diagram of the current programmer sequencing system (C and C are clock input and P1 is the informaiion from switch register) 99 9 Circuit diagram of the initial time delay (ITD) step 100 10 Circuit diagram of the desolvation time unit (DTU) 102 11 Circuit diagram of the atomization time unit (ATU) and "reset and clear" systems 103 12 The effects of clock rate on the magni— tude of the SDAC output and its repro- ducibility 107 13 Computer interface for electrical heating of nonflame atomizers 110 14 Schematic diagram of solution pumping device utilized in sampler design I 115 xiii Figure 15 17-19 20 21 22 23 24 25 26 27 28 29 3O 31 32 Operational cycle of the sampler design I as a function of position of the cam and pistons Circuit diagram of sampler design I sequencing system Various positions of automatic sampler design II Block diagram of the connection of the stepping motor and the driver circuit Cycle time of the Purge function in automatic sampler design II Cycle time of the Fill function in auto— matic sampler design II Cycle time of the Deposit function in automatic sampler design II Circuit diagram for the isolation of the computer interfaces and the line voltage Computer interface for the control of atomizer temperature and stepping motor forward and reverse movement Computer interface for the control of automatic sampler design II Circuit diagram for analog integration General flow chart of the computer program Operating cycle of one experiment in the computer-controlled system Influence of inner and outer flow rate on the cadmium fluorescence (arbitrary unit) signal Influences of flow rate on the integrated AF signal (arbitrary unit) and atomization temperature Influence of atom residence time on the cadmium AF (arbitrary unit) signal xiv Page 116 118 121 125 128 129 131 132 133 134 137 143 144 148 150 151 Figure 33 34 35 36 37 38 39 4O 41 42 43 44 45 46 Influences of flow rate and atomization current on the Zn AF (arbitrary unit) signal Influences of the quenching effects and flow rate on the integrated AF (arbitrary unit) signal Influences of the quenching effects and flow rate on the integrated AF (arbitrary unit) signal Two atomic fluorescence signals of 10 ppm Hg under non-optimized conditions Influence of flow rate on the signal-to- noise ratio Influences of atomization temperature and inner flow rate on the AF signal and signal—to-noise ratio Influences of desolvation time and current on the AF signal and signal—to-noise ratio Long term stability of the computer-controlled AF spectrometer AF analytical working curves for Cd, Zn and Hg obtained under optimized conditions Typical boiling curve for a wire in a pool of water at atmospheric pressure (heat flux has arbitrary unit) Influence of applied power and applied current on the GBA temperature Influence of applied power and voltage across the atomizer on the temperature of a 0.5 mm diameter GBA Spectral emittance (arbitrary unit) of GBA as a function of atomization tem- perature and wavelength Comparison of the spectral emittance (arbitrary unit) of GBA with two photo- tubes XV Page 153 155 156 161 165 167 168 171 172 176 188 189 191 192 Figure Page 47 Influence of the atomization temperature on the atomizer evaporation rate 197 48 Variation of the emission intensity (arbitrary unit) of GBA as a function of atomization current and number of experi- ments with current programmed heating method 202 49 The circuit diagram of the instrumentation for power programmed heating of nonflame atomizers 205 50 Variation of GBA emission intensity (arbitrary unit) as a function of atomiza- tion power and number of experiment with power programmed heating method 207 51 Influence of different heating methods and atomization parameters on the lifetime of the GBA and the GBA emission intensity variation 209 52 The circuit diagram of the instrumentation for the combined power and radiation programmed heating of nonflame atomizers 212 53 Intensity of GBA as a function of time and as a function of three programmed heating methods 215 54 Duration of the steady state period as a function of the atomization parameters for three programmed heating methods 216 55 Duration of the pre—steady state period as a function of the atomization parameters for three programmed heating methods 217 56 Influence of flow rate and heating methods on the atomizer temperature 222 57 Cadmium working curve (AF signal has arbitrary unit) 229 58 1 Gold working curve (AA signal has arbitrary unit) 230 59 Iron working curve (AA signal has ar- bitrary unit) 231 60 Nickel working curve (AA signal has arbitrary unit) 232 xvi __ AA.’ Figure 61 62 63 64 65 66 67 68 69 7O 71 72 73 74 Silver working curve (AA signal has arbitrary unit) Zinc working curve (AA signal has ar- bitrary unit) Cobalt working curve (AA signal has arbitrary unit) Manganese working curve (AA signal has arbitrary unit Standard Addition analytical curve for Cu added to ten—fold diluted serum 4-channe1 TOM system a) general diagram, b) waveforms in the transmission path for the pulse amplitude modulation Block diagram of computer-controlled multi- channel AF spectrometer Circuit for pulsing four hollow cathode lamps in TDM mode Computer interface for pulsing four HCL's in TDM mode Cycle time of four pulsed HCL's in the TDM mode Computer interface for the synchronous integrator Cycle time of the operation of the pulsed HCL's, synchronous integrator and the ADC Magnesium and nickel fluorescence intensity (arbitrary unit) as a function of argon flow rate Cobalt and iron fluorescence intensity (arbitrary unit) as a function of argon flow rate xvii Page 233 234 235 236 240 248 250 254 255 257 259 260 265 266 7T"""""""""""""""""""""""""""""""""""""""""""""""Illlll I. INTRODUCTION Computer automation of routine analytical measurements is of great practical importance because measurement throughput can be increased significantly and measurement reliability can be greatly enhanced. The laboratory minicomputer is making possible entirely new instrumental measurements which were inconceivable merely a few years ago. The area in which computer—aided experimentation promises to bring about the most profound changes, however, is in the automation of fundamental research studies. Less emphasis has been placed on the automation of fundamental studies because such studies involve widely varying experimental conditions, decision-making based upon what is known about the system, decision-making based upon prior experiments, and decision-making concerning the goals of the experiment, which may change as new information is acquired. This type of complete automation would greatly facilitate funda- mental chemical studies. The influence of all experimental variables could be investigated in a logical manner using the prOper statistical design scheme. In addition experimental variables could be optimized for different goals either in real time or by an iterative procedure. Finally, the complete automation of a chemical experiment could make the investigation of effects which are very difficult to study manually a routine matter. In many types of chemica1 experiments the obstruction to com- p1ete automation comes in the steps of sample preparation and sample hand1ing which occur prior to the instrumental measurements. The automation of these steps along with the more conventional steps of 1 3’ data acquisition, data processing and real-time control over instru- mental variables would then allow computer feedback in a "closed-loop” type system. The study of matrix interferences in atomic spectrometry or the study of ionic strength effects in chemical kinetics are examples of studies which are tedious to perform manually because of the time- consuming sample preparation and handling steps. Consequently many of these studies never get performed or get carried out in a rather haphazard manner. Complete computer automation should greatly ease the burden in performing such experiments. In this thesis the development of computer-controlled nonflame atomic absorption (AA) and atomic fluorescence (AF) spectrometric systems, which feature a high degree of automation, is described. Although a chemical combustion flame can also be utilized with all of these systems, the systems are designed to operate with electric- ally heated nonflame atomizers, such as filaments and furnaces. Des- pite the advantages of flames, namely convenience and reliability of operation, variety of available flames, low costs, and adequate sen- sitivity and precision, the flame atomizers exhibit some severe dis- advantages which limit the attainable sensitivity and convenience in their use for trace-metal analysis. First, in sample-limited situa- tions, such as the analysis of biological samples, the volumes of available samples may frequently be less than that required for analysis in the flame. Flames are also seldom able to atomize solids directly. Second, the transport of the sample solution to the flame and the production of a vapor of the analyte in the flame are highly inefficient. Third, the presence of flame gas radicals results in formation of stable species such as monoxides and monohydroxides, AT"""""""""""""""""""""""""""""""""""'"""""""""'Illll 3 which consequently decreases the free atom fractions. Therefore, precise control over the chemical environment of the analyte and concomitant atoms in flame cells is not possible. Fourth, the achiev- able analyte atom concentration in flames is limited by the dilution effect and the expansion of flame gases. Fifth, in atomic fluores- cence the presence of molecular quenchers in analytical flames de- creases the fluorescence quantum efficiency and thus the fluorescence signal. Furthermore, in both AA and AF, spectral interferences are caused by emission from the flame and the analyte. This can be mini- mized, however, by modulating the radiation source(s) and using a phase-locked detection system. When a nonflame atomizer is utilized, the transport efficiency and the subsequent desolvation of the sample are nearly 100% complete. It is also possible to have precise control over the chemical and spectral environment of the system as well as electrical control over the temperature. The environment of the atomizer is simple to adjust for maximum production of free analyte atoms and for the highest quantum yield in atomic fluorescence. The major use of nonflame atomic vapor cells, however, is for the analysis of nanograms of elements in small samples, such as in microgram or microliter samples. Full utilization of all of these advantages for both routine analysis and fundamental investigations is not possible without a fully computer- controlled system. Therefore, in all of the systems described here, the entire system operation, data acquisition, data treatment and optimization are performed under computer control with no operator intervention. A fundamental investigation of nonflame atomization was undertaken —_——‘ 4 in an attempt to better understand the complicated physical and chemical processes, to separate the interdependence of parameters, to design new atomizers, and to introduce fast multielement analysis methods. As a direct result of these fundamental investigations, three new nonflame atomizers, four new methods of programmed heating and a new multi— element analysis technique were proposed and fully investigated. The new filament atomizers made from graphite braids, threads, or tape incorporate all the fundamental and practical advantages of other nonflame atomic vapor cells as well as several advantages which are not present in other furnace and filament atomizers. The widespread application of one of these vapor cells, the graphite braid atomizer, was demonstrated by the determination of 15 elements by atomic absorp- tion and atomic fluorescence and by its application to biological samples. The various heating techniques introduced are applicable to any electrically heated nonflame atomizer. Application of these techniques to a platinum loop and a graphite braid atomizer showed that the method of programmed heating influences the atomizer lifetime, the time re- quired for the atomizer to reach a steady state temperature and the separation and optimization of atomization parameters. A new multi- element analysis technique which operates in the time-division multi- plexed mode has performed a multielement analysis for 4—8 elements in 1ess than 3 seconds. The application of this method to flame and nonflame atomizers is also presented. II. HISTORICAL The contents of Chapter II are divided into five parts. In the first part, the history of atomic absorption and atomic fluores- cence is reviewed along with the fundamental principles behind these techniques. In Part two, the advantages and disadvantages of flame atomizer as well as the previous work on the separated flames are discussed. The various nonflame atomization techniques are reviewed in the third part, and the advantages and drawbacks of each method are presented along with the practical application of various non- flame atomizers. The fourth part reviews the radiation sources for AA and AF. The previous work on pulsed lasers and pulsed hollow cathodes is discussed in more detail. Finally multielement analysis by AA and AF is briefly reviewed in the fifth part. A. Atomic Absorption and Atomic Fluorescence Spectrometry In their analytical context, atomic absorption (AA) and atomic f1uorescence(AF)spectrometryaremethods for determining the concen- tration of an element in a sample. In AA, an atomic vapor is produced from the sample by an atomizer. The vapor is then excited by means of an external light source, and the fraction of the radiation ab- sorbed as a result of radiational excitation is monitored at a wave— length that is Specific and characteristic of the element under in- vestigation. Atomic absorption was observed as early as 1802, when Wollaston observed dark lines in the sun's spectrum, but the exact 5 explanation of the phenomenon was presented in 1859 by Kirchhoff and Bunsen who laid the foundations of spectrochemical analysis. It was not until 1955, however, that the analytical potentialities of atomic absorption were proposed by Walsh (1), and independently, but some- what later, by Alkemade and Milatz (2,3). Since that time several publications have dealt with the application of AA to the solution of practical problems, and the method has proven to be an indispens- able analytical tool. Several books (4-11) have been written on the subject and atomic absorption spectrometers are presently available from more than twenty instrument manufacturers throughout the world. Atomic fluorescence spectrometry differs from AA only in the means of measurement of the characteristic signal. The analyte atoms are radiationally excited as in AA, but the measured parameter is a portion of the atomic fluorescence radiation resulting when a frac- tion of the excited atoms undergo radiational de-excitation. The phenomenon of fluorescence by free atoms was first noted in sodium vapor above molten sodium by Wood (12) in 1905, and atomic fluores- .cence in flames was observed during 1924-1961 for about ten elements (13—16). It was Alkemade (17) who suggested the analytical utility of the phenomenon in 1962 and for the first time, Winefordner. Vickers and Staab (18-19) demonstrated AF as an analytical technique in 1964. The pioneering work of Winefordner and coworkers was closely followed by work done by West and coworkers (20-21). A number of excellent review articles have been published (24-28), and numerous chapters have dealt with the subject in atomic spectroscopy books (10,11, 29-32). It is obvious that since AF was introduced nine years after the develop— ment of AA, there have been a smaller number of publications, and the method is still relatively new and untried in practical applications. In contrast to AA, commercial atomic fluorescence spectrometers are not available. However, prototype multichannel systems have been demonstrated (23,257,258). Although the unavailability of commercial instruments can be partly responsible for the lack of widespread acceptance of AF as an analytical tool, it is the author's opinion that the origin of the problem has been closely tied to the lack of suitable atomizers and radiation sources for AF. The instrumentation involved in AA and AF mainly depends on the desire to perform single or multielement analysis, but in general, it consists of a radiation source(s), an atomizer, a wavelength isola- tion device(s), a detector(s) and a readout. Both techniques require entrance optics to focus an image of the source upon the vapor cell. Electrically or mechanically modulated light sources and suitable electronics are needed in order to minimize the response to thermal emission from flames and from some of the nonflame atomizers for analysis of elements with analytical lines above 3000 A. Although a greater signal is obtained with a dc electronic system and an unmodulated source, the effect of both white noise and l/f noise can be more severe. Since similar instrumentation is used for both AA and AF, a review of some of the basic components will be given in later sections. 8. Flame Atomizers In atomic absorption and atomic fluorescence spectrometry the achievable sensitivity and precision are limited by the techniques with which atoms of the analyte element are produced from the sample and by the characteristics of the external radiation source. The methods of atom production are discussed in this section, and a review of the radiation sources for both AA and AF appears in a later sec- tion. Flame and nonflame atomizers have been the subject of many review articles (24,25,27,33,34) and chapters (9, 31, 35). The pur- pose of an atomizer is to provide a stable, noise-free system to con- vert the sample into free atoms. At the present time, all commercial instruments are equipped with flame atomization systems, though non- flame attachments are also available. 1. Advantages of Flame; The popularity of flame atomizers originates from the following advantages: a) The attainable signal—to-background and signal—to-noise ratios are sufficiently high to provide adequate sensitivity and precision for the wavelength range of 2000-3000 A. b) They are convenient to use, reliable and relatively free from memory effects. c) Most burner systems are inexpensive, small and stable. d) The wide variety of flames in terms of composition and tempera- ture allows the selection of proper optimum conditions for many different analytical purposes. Despite their advantages, the flame atomization systems exhibit some disadvantages which limit the attainable sensitivity and con- venience in their use for analysis. A discussion of these drawbacks appears later in this section. Since flame atomization systems have been mostly used in conjunction with AF in our investigation, this discussion will be limited to those combustion type flames which are the most suitable for AF. For such flames it is desirable to have a minimal concentra— tion of quenchers, a long residence time of the atomic vapor in front of the observation window and a low background emission. Since the radiance of the external source in AA is many times greater than the radiance of the flame, the effect of flame background is not as critical as in AF. In other words, since the atomic fluorescence and flame radiances are comparable, the detection power of any AF system may be limited by the atomizer background. In a conventional flame, the background emission comes from both the primary and secondary reaction zones. Virtually all the hydro— carbons are broken down in the primary cone and produce among others, unburned products such as carbon monoxide and hydrogen. As a result of the diffusion of aunospheric oxygen towards the center of the flame (37) and the combustion of unburned products in the secondary zone, emission bands result, the most intense of which comes from OH bands in the region of 3100 A. If the flame is sheathed (37) with an inert gas, the secondary combustion is prevented and atomic fluorescence measurements may be made in the secondary zone where background emis- sion is extremely low. It should be noted, however, that the tem- perature of the primary cone decreases with increasing sheath gas flow rate. The extent of the cooling effect depends on the sheath gas, flame composition and the height of observation (38). Despite the disadvantages of a lower temperature, separated flames of this type have been used in studies of fluorescence yields (39) and studies IIIIIiE' ‘illl... of atomization efficiencies (40). They have proven to be useful in the determination of metals forming refractory oxides (22) and have been evaluated for atomic emission, atomic absorption and atomic fluorescence (41-46). A modified version (47) of the sheathed burner designed by Larkins(48)was used throughout this investigation. 2. Fundamental Disadvantages of Flame Atomizers a. Precise control over the chemical environment of the analyte and concomitant atoms in a flame cell is not possible. The degree of control of chemical composition that can be obtained by variation of the fuel—to-oxidant ratio causes the simultaneous change of flame temperature as well as flame spectral emission and absorption charac- teristics. Since all analytical flames possess a considerable con— centration of flame gas radicals such as O, OH, etc., which combine with an appreciable fraction of many elements to produce stable mon- oxides, monohydroxides, etc., the free atom fraction, 8, is less than unity (52). It has been shown (53) for copper that 108 atoms should produce 1 percent absorption whereas in practice the best reported sensitivity limit correspondsto 1015 atoms. In other words, the ef- ficiency of atomization is 10'5%. b. The aspiration efficiency, 6, is less than unity. Because of the short residence time of solute particles within the flame gases (54), solutions with a high concentration of analyte and/or other components in the sample matrix may undergo incomplete solute vapor- ization. Scattering of exciting radiation by solute particles can be considerable. The total consumption nubulizer - burner unit is inefficient in the sense that large droplets are produced so that, even in many instances, solvent evaporation is remarkably incomplete. Pneumatic nebulizers seldom provide an aspiration efficiency of more than 15% with aqueous solutions. The remaining test solution is merely wasted. c. The achievable analyte atom concentration in flames is limited by the dilution effect of a relatively high flow rate of unburnt gas used to transport minute volumes of sample solution to the flame. As can be seen from Equation 10 in Chapter III, the flame gas expan- sion also limits the atom concentration. d. All analytical flames contain an appreciable concentration of molecular quenchers such as C0, C02, and N2, which decreases the fluorescence quantum efficiency,Y, in AF (55,57). The decrease in the parameter Y results in a decreased fluorescence signal and signal- to-noise ratio. e. Continuous variation and control of flame temperature is not possible. 3. Practical Disadvantages of Flame Atomizers a. There is considerable absorption background with flames employing carbonaceous fuels. At wavelengths below 2000 A, an ap- preciable reduction in intensity results owing to absorption of radia- tion by flame gas products. Furthermore, because of chemilumines- cence reactions, such as the oxidation of CO with 02, flames exhibit considerable amounts of emission from their primary and secondary zones. Flames also produce thermal excitation of elements having resonance lines at wavelengths greater than about 3000 A. Since the emission from these elements is constant with time, its contribution can be minimized by modulating the AA or AF source and using a phase-sensitive amplifier locked to the modulating frequency. b. The volume of sample solutions may frequently be less than that required for analysis in flames. This is especially true in clinical chemistry. c. Explosion hazards are always present with flames of high burning velocity. In closed automated systems where no operator is present, it may not be desirable to employ flames as the atom cells. Efficient exhaust systems are required to remove the released toxic products. d. Flames usually produce audible noise, although they can be made to operate fairly quietly. e. Flame atomizers are seldom able to atomize samples directly. C. Nonflame Atomizers The first application of nonflame atomizers dates back to 1908 when King (49,51) suggested the use of a graphite furnace for the investigation of atomic spectra and the measurement of oscillator strengths. In the last two decades, there has been considerable research into new nonflame atomizers and their application to the solution of analytical problems. In fact, the disadvantages of the flame mentioned in the previous section have directed many workers to devise new techniques for atomization of samples. In a nonflame atomizer, it is possible to have precise control over the chemical and spectral environment of the system as well as electrical control over the temperature. The environment of an atomizer is simple to adjust for maximum 68, and the quantum yield is maximized by use of an inert gas. A nonflame atomizer, particularly the filament-type atomizer, should have lower background and back- ground flicker than a flame. The major use of nonflame atomic vapor cells would seem to be for the analysis of nanograms of elements in small sample sizes, such as micrograms or microliters. It is because of these advantages that the use of nonflame atomizers has gathered considerable momentum in recent years. The major nonflame atomizers are furnaces and filaments. These atomizers are reviewed in this section along with the cold vapor cell technique, the electrical dis- charge and other less popular nonflame methods. 1. Furnaces a. The L'vov Furnace. Several satisfactory long path length graphite furnaces of the type used by King have been designed and tested for analytical use in both AA and AF. The sensitivity of the King furnace, and open end furnaces in general, is determined by the amount of substance existing in the vapor phase. Thus, the increase in the amount of sample would be limited by the fact that the volume of vapor becomes comparable with the internal volume of the furnace. Because of vapor loss by diffusion from both ends, error occurs. For a 10 cm long tube, the loss of Hg in an atmosphere of N2 is reported to be about 4 percent in a time duration of 100 msec at a temperature of 2000 °K (58). Various models of the original King graphite furnace have been used by L'vov and coworkers in extensive AA studies, and the results have been reviewed (9,59). Open ended graphite tubes 30 to 100 mm in length and 2.5 to 5.0 mm in i.d., depending on the analytical requirements, have been employed. In the original L'vov furnace (58), the sample is volatilized into a heated tube with a carbon arc. A tantalum or tungsten foil, placed inside the tube, diminished the loss of sample vapor by diffusion through the walls of the graphite tube. The samples to be analyzed are placed on the electrode either in solution form or as a powder. Solution samples are placed on the ends of a carbon electrode and evaporated to dryness. The tip of the electrode is first coated with a solution of polystyrene in benzene in order to prevent the sample solution from soaking into the carbon. The electrode carrying the sample is introduced into a conical opening in the middle of the furnace. The tube is heated to 2000-3000 °C by an ac current supplied by a 10 kW transformer, and the electrode is simultaneously heated from outside by means of a dc arc. The furnace is placed in an argon chamber which contains two quartz windows to permit radiation from the primary source to pass through the furnace. In a revised version of the graphite cell by L'vov and Lebedev (9), the tantalum or the tungsten foil was discarded, and the tubes were fabricated entirely from pyrolytic graphite or else they were lined both from inside and outside with a layer of pyrographite. The pyrolytic graphite has the advantage of having low permeability, high heat conductivity and high resistance to oxidation. This diminishes diffusion of vapor through the walls, provides uniform heating of the I tube and ensures a longer lifetime of the atomizer. A more efficient method of heating was also employed. A two channel atomic absorption spectrometer was used to correct for non-selective molecular absorption and scattering of source radiation. The spectrometer also allows 15 the simultaneous recording of the absorption of two elements, one of which can be an internal standard. In addition to the extensive use of the graphite furnace for trace element analysis, L'vov and co- workers investigated (9) the Lorentz width of resonance lines, the determination of absolute values of oscillator strengths and estima- tion of atomic diffusion coefficients. The furnace has been used for the determination of 40 elements by AA thus far. b. The Woodriff Furnace. Woodriff and coworkers developed a graphite tube furnace for AA in which a tube is heated continuously by an arc-welder or by a power supply that can provide 200 A at 24 V. The furnace body is made of double-walled stainless steel, is in- sulated with graphite felt and is water cooled. The sample is intro- duced into the tube, 150 mm in length and 7 mm in i.d., through a side arm which is located in the center of the tube furnace. Although the sample can be nebulized into the side arm, it was found (60,61) that water solutions eroded the graphite rapidly while ethanol solutions caused the tube to get thicker. Furthermore, no nebulization occurs at high temperatures. In order to overcome the nebulization problem, a syringe pump was used to introduce solution into pneumatic and ultrasonic nebulizers. The sample can also be introduced into the tube with a carbon cup (61,63). Normally, 20-50 ul of solution are pipetted into the cup and evaporated to dryness under an infrared lamp. The cup is twisted onto a small graphite rod and is inserted into the furnace through the sample port. The side arm narrows just before it opens into the optical path, and the rim of the sample cup makes a tight contact with this construction. Thus, the inside of the cup becomes an integral part of the tube furnace. The sample 16 evaporates quickly and expands in both directions along the optical path. Occasionally a small flake of carbon will fall through the light path as the cup strikes against the retainer ring. With the background correction technique (63,64), this problem can be corrected. The furnace has been used for the determination of seventeen elements (62,63), and no significant matrix effects were observed from the presence of relatively large amounts of Al, Cr, Cu, Fe, Ni, Mn, Zn, Mg on the absorption signal obtained for 10"9 g of silver. The furnace has been employed for the determination of trace level of Pb in the atmosphere (65) and in fish tissue (66) and for the determina— tion of sub-nanogram quantities of silver in snow (67). The life- time of the most improved version of the furnace has been reported to be 10 months (67). The furnace is commercially available from Woodriff Frontier Products (Box 1305, Boyeman, Montana). c. The Massmann Furnace. A somewhat simpler version of the L'vov furnace was designed by Massmann for both AA and AF (68,69). In con- trast to the L'vov design, samples are inserted directly into a tube through an orifice 2 mm in diameter, and evaporation of samples is accomplished by resistance heating of the tube walls instead of an arc. The absorption tube is 55 mm in length, of 6.5 mm i.d., and the wall thickness is 1.5 mm. The tube, mounted in a water cooled chamber, is purged with argon, but the optical path through the tube is open to the atmosphere. For AF work, a cup-shape graphite cuvette is employed. Both the tube and cuvette can be heated in a few seconds to 2600 °C by a current of up to 500 A. If it is assumed that the sample is completely evaporated during the heating period, the cell is ready for the next sample after a cooling time of 20-30 sec. 17 The lifetime of the graphite tube depends on the maximum required temperature and ranges from 200-300 analyses. Sample solution volumes of 5-50 uL were used for AF work and 5-200 ut for AA studies. For solid samples larger than 1 mg, unacceptable background absorption levels were encountered which were minimized by simultaneously measur— ing the intensity of a nearby nonabsorbing line (70). Detection limits for 16 elements by AA and 9 elements by AF have been obtained by Mass- mann. The relative standard deviations of the measurements were matrix and concentration dependent and ranged between 4 and 12 percent. A graphite furnace similar to that described by Massmann is com- mercially available for atomic absorption spectrometers (71) and was first used for the determination of copper and strontium in milk. It is claimed that the furnace can be used for the determination of 60 elements (72). Massmann type furnaces have been used for the direct determination of trace metals in sea water (73) and high purity water (74), for the determination of lead in the atmosphere (75) and in blood (76) and milk (77), for the analysis of chromiwm in urine (78) and for the measurement if iron and copper in high purity silica (79). Interferences were found to occur with most elements when present in large amounts (73,80) and solvent extraction prior to analysis is recommended. Other practical applications of this atomizer have been discussed elsewhere (81). d. Induction Furnaces. From an analytical point of view, two types of induction furnaces (82,85) have been employed for both atomic absorption and atomic emission spectrometry. In these furnaces, the vaporization and atomization of the sample is accomplished by induc- tive heating of a graphite cuvette or tube. Morrison and Talmi (82-84) 18 have described the construction of an induction furnace for AE and AA. The inductively heated graphite cuvette, located in an helium environ- ment, serves as a thermal means of vaporization and atomization, while the excitation of the atomized sample is achieved mainly by the helium plasma formed by an RF field. The furnace consumes 4.5 KW, operates at 3 MHz, and is capable of reaching a maximum temperature of 2500 °C. A maximum of 300 samples can be analyzed before the properties of the furnace are changed significantly (83). Samples can be introduced every 30-60 seconds. For absorption measurements, the light from a hollow cathode lamp passes through the plasma plume located over the mouth of a graphite crucible and is absorbed by the sample vapor. The attenuation of the primary radiation is monitored by a conventional AA spectrometer. The system has been used for the determination of 19 elements by AE and 6 elements by AA, but the absolute sensitivities are inferior (82,83) by one to four orders of magnitude to those ob- tained with some of the nonflame atomizers. Accuracies on the order of 5 percent have been obtained with microsamples of geological, metallurgical and biological materials (83). Headrige and Smith (85) have reported the construction of an induction furnace for the analysis of volatile elements in solution and volatile matrices by AA. Except for the heating method, the general arrangement of the furnace is similar to that described by L'vov, and it can achieve a maximum temperature of 1900 °C. The furnace has been used for the determination of cadmium in zinc-based alloys. e. Furnaces With Continuous Sample Introduction. Some problems associated with nonflame atomizers for both AA and AF spectrometry have included the complexity of design, poor reproducibility of measurements and poor accuracy due to appreciable matrix interferences. The poor reproducibility of measurement mainly originates from dis- crete sampling. A more reliable and convenient method, used with both AA and AF, involves the continuous nebulization of sample solu— tions into heated tubes. Mislan (86) has described the construction and performance of an AA spectrometer that incorporates a 36 cm tube of2-5 cm i.d. heated to a maximum temperature of 1250 °C by a wire wound resistance furnace. Sample solutions were transferred to the absorption tube through a conventional indirect-spray chamber assembly. The device was used for the determination of Cd, and excellent detec- tion limits were obtained. Black et a1. (87) used a heated chamber— condenser sample introduction system (88) to nebulize aqueous samples continuously through a platinum furnace. The furnace was electrically heated and could achieve a maximum temperature of 1600 °C. The atomic vapor was excited by means of an electrodeless discharge lamp which was focused just above the top of the furnace. Two furnace designs were evaluated and compared with respect to limits of detection, ranges of linearity of analytical curves and interferences. The platinum fur- naces have been used for the analysis of Cd, Zn, Cu, Hg and Fe in atomic fluorescence. The relative standard deviations were reported to vary between 1.2 to 3.9 percent. A graphite tube, located in a water cooled assembly, has been described by Murphy et a1. (89) for the determination of Zn, Cu and Bi by atomic fluorescence. The tube was heated by the power from a 3 KW transformer, and the sample aerosol and argon were introduced into the chamber tangentially, exiting up through the tube. The atomic vapor, emerging from the top of the furnace was excited by 20 primary sources such as hollow cathodes, Xe arc lamps, and electrode- less discharge lamps. The resulting atomic fluorescence signals were compared for a photon counting and a lock-in amplifier system. The best detection limits obtained for Zn, Cu, and Bi were 2.7 x 10'7, 3.2 x 10's, and l x 10'3 ug/ml, respectively. f. Other Furnaces. A variety of furnaces have been used for atomic absorption studies (90-94). Hudson reported the use of a stainless-steel absorption cell heated by a resistance wire for the determination of the atomic absorption cross section of Na vapor (90). Choong and Loong-Sing employed a fused silica absorption tube for the investigation of the absorption spectrum of silver vapor (91). Re- cently, an automated atomic absorption spectrophotometer has been described for the acquisition of thermodynamic data (92). The instru- ment can provide automatic measurements of elemental vapor pressure changes with temperature. Using this system, Pemsler and Rapperport (93) have demonstrated that concentrations of extremely small impurities in well characterized metals and alloys can be determined by AA in a sealed cell. The method was illustrated by the determination of 80 ppb of Zn in 99.9999% copper. Robinson and coworkers employed a furnace design which was bas- ically similar to that described by Woodriff. Here the sidearm of the furnace was filled with carbon chunks or rods, and the furnace was heated by an RF coil to about 1400 °C. The sample gas or liquid, I was introduced into the sidearm and was reduced to metallic elements. The atomic vapor then entered a quartz absorption tube heated by a nichrome resistance heater. The system has been used for the deter— mination of Pb, Cd, and Hg in air (95,97) and for the determination 21 of Cd in sea water and urine (98). Background correction was used to correct for molecular absorption. The effect of chemical interferences for 11 anions on the Cd signal was studied and most of the ions produced no interference (98). 2. Filament Atomizers The history of filament-type nonflame atomizers originates from the work of Bunsen in flame emission spectrometry in 1859. A wide variety of these atomizers, in which a rod, a loop, or a cup carries the sample, have been pr0posed in recent years. These atomizers, fabricated from graphite or metallic filaments and strips, are nor- mally electrically heated, and the resulting atomic vapor passes into an unconfined volume in the absorption or fluorescence light path. As a result, with filament-type atomizers, transient analytical sig- nals are obtained. Because of the unconfined nature of the analytical cell volume, there will be less memory effect as well as a considerably greater dilution of the analyte than with the furnace atomizers. Also, 88 should be greater with the furnace atomizers. Furthermore, if a large temperature gradient exists between the hot filament and the cooler volume in which AA or AF measurements are made, it may be dif- ficUlt to achieve freedom from interference effects with the filament atomizers. These interferences should decrease if a flame surrounds the atomizer. a. Gold, Silver and Copper Wire or Foil Atomizers. Ulfuarson (99) employed a gold foil for the determination of Hg in small quantities of biological materials. After the digestion of the biological samples and extraction with dithizone in chloroform, the sample is transferred — 22 to an ignition tube, and the chloroform is evaporated. The mercury- dithizone complex is then destroyed by heating and Hg forms an amalgam with a gold filter. The fail is then heated to vaporize the mercury into the optical path of an AA spectrometer. Biological samples con- taining as low as 10 ng of Hg per 9 can be analyzed with this method. A similar system using tin (II) chloride as a reducing agent has been used for the determination of Hg in geological materials. The mercury is collected on a silver (100) or gold wire (101) and is sub— sequently released by heating the wire in a closed furnace. The detec— tion limit was reported to be 1.0 ng of mercury. Brandenberger and Bader (102) used a copper wire for the quanti- tative amalgamation of Hg. The wire was electrically heated and AA measurements of as low as 0.2 ng of mercury were performed. Stephens (102) described the use of a copper loop which is heated by a capacitor bank initially charged from a conventional 0-450 V variable power supply. The system can provide heating pulses of a few microseconds and was used for the determination and identification of inorganic and organically-bound mercury by AA spectrometry. b. Tungsten and Platinum Atomizers. The amalgamation or the electrolysis technique described above can also be used for the deter- mination of elements other than mercury. Using an electrically heated platinum or tungsten wire, Brandenberger (104) reported the determina- tion of Cd, Zn, Pb, Tl, Cu, Ag, Au and Pt in AA spectrometry. Bratzel et a1. (105,106) have described the application of a filament tech- nique, similar to that described by Brandenberger, for AF spectrometry. After the analyte solution is placed on an electrically heated platinum or tungsten wire loop, the sample is vaporized and swept into the [71 l A hm» ,‘ 23 fluorescence light path where atomic fluorescence is excited by an electrodeless discharge lamp. The device was used for the determina- tion of 11 elements. Williams and Piepmeier (107) evaluated a rigid spiral-wound tungsten filament for the analysis of Ca, Cr, Cu, Fe, Mg, Mn, and Sn by AA and presented a method of compensating for background tungsten emission. A tungsten-rhemium alloy wire loop (108,109) has also been used for the determination of Cd, Pb and Ni by AA. Cantel and West (110) have described the use of a tungsten fila- ment atom reservoir (TFAR), 60 mm long and 2.2 mm in diameter, for atomic absorption spectrometry. The sample is delivered to a 1 mm deep notch (N2 mm wide) at the center of the filament, and the TFAR is heated by dissipating electrical energy. The TFAR atomizer has been used for the determination of traces of Zn, Pb, Cu and Ag, and its performance was compared to the carbon filament atom reservoir for the analysis of lead. It was reported that the lifetime of a TFAR was almost indefinite and that greater sensitivity and fewer matrix effects were observed with the TFAR system. c. Tantalum Atomizer. Donega and Burgess (111) have assembled a filament device which incorporates a sample boat, 50 mm long and 6 mm in width, cut from a graphite sheet, or from tantalum or tungsten foil. The device can be used with volumes of liquid samples as large as 50-100 vi. The sample boat is heated electrically with a current of 30 to 50 A at 12 V, which is sufficient to heat the boat to about 2200 °C in less than 100 msec. The atomizer is mounted in a quartz chamber which can be purged with inert gas in the pressure range of -u-1 24 1 to 760 torr. Although operation at low pressures provides low back- ground interference, the residence time of the atoms in the effective absorption volume decreases and dictates that the detector system be of fast response. The method was used originally for the deter- mination of 10 elements by AA. Using a tantalum strip, Hwang et al. (112) have recently reported the determination of AA sensitivities and detection limits for 37 elements. Tankeuchi et al. (113) compared nickel, tungsten, tantalum and platinum as strip materials. Only small signals were obtained with platinum, and neither nickel nor platinum could be used at very high temperatures. Tantalum and tungsten showed similar results, but the tantalum strip was easier to fabricate and was chosen as the strip material. Samples as small as 0.5 Hi can be delivered to the atomizer. The strip requires 80 W of power to reach a temperature of 2400 °C (114). The system was used for the determination of traces of Al, Cr, Cu, Fe, Mg and Mn and for interference effect studies (113-115). With the exception of the effect of phosphoric acid on the copper absorption signal, the interferences of hydrochloric,perchloric and nitric acid were minimized at high applied powers (115). Similar results were obtained for the interference effects of 12 foreign ions on the chromium absorption signal (114). The tantalum strip has been used for the determination of Mg in control serum and in aluminum alloys (113), for the analysis of Cr in steel (114) and for the microdetermination of Pb in blood (116). From all of the atomizers made from metals, only the tantalum strip is commercially available (Instrumentation Laboratories, Lexington, Mass.). Compared to atomizers made from carbon or graphite the 25 tantalum strip atomizer has the advantage of avoiding carbide forma- tion and the disadvantage of lower temperature. Other practical ap- plications of this atomizer have been discussed elsewhere (117). d. The AA Microsampling System Involving Sampling Boats and the Delves Cup. The AA microsampling system is often described as a “semi-nonflame" atomizer. After placing the sample in a cup or boat-shaped container and drying the sample, the vessel is introduced into a flame. The heat from the flame is sufficient to atomize volatile elements. The boat has a sample capacity ten times that of a cup which for some elements leads to a better detection limit. Both systems give much better detection limits than flames, but only for ten easily atomized elements. The origin of this work started when several attempts had been made to improve the atomization efficiency of total consumption nebilizer burners by using them with long-path absorption tubes of various materials and internal reflective coatings (118-120). The atomization efficiency of a premixed burner-nebulizer system has been improved by means of sample cups, boats and platinum loops. Kahn et al. (121) used a tantalum boat technique for the analysis of lead in urine. White (122) employed a platinum wire loop to vaporize blood samples into a nickel absorption tube for lead determinations. By combining the two previous methods, Delves (123) vaporized blood samples from nickel cups into an absorption tube placed in a long- path length flame. Lead can be determined in a sample volume of 10 pt. As many as 30-50 samples can be analyzed per hour. In all of the above techniques, a minimum of two peaks are seen in each determination, the first of which is from unashed organic F4- a I 26 material and is variable in size, while the second one is the lead signal. A time delay circuit could be used to eliminate the recording of non-specific absorption signals (124). Although the method is easy to use, it has been shown that, it is prone to both positive and negative interferences for the determination of mercury (125) and tellurium (126). Because of the effects of the sample matrix and the cup quality on peak absorbance, integration of the signals has been recommended (127). These atomizers are commercially available (72). e. Carbon and Graphite Filament Atomizers. Among all the filament- type cells, the atomizers made from carbon or graphite appear to be highly applicable because of the high operating temperatures (2500- 3000 °C) which can be achieved. West and Williams (128) have reported the construction and use of an atomic vapor cell made from a graphite filament 1-2 mm in diameter and about 20 mm in length, for AA and AF. The filament was supported by water cooled stainless-steel electrodes and heated to 2000-2500 °C within 5 seconds by applying a current of about 100 A at 5 V. Sample volumes of 1-5 ut are placed one depression it) the filament, and the assembly is housed within a chamber which can be purged with argon. The original device was used for the detection 01’ Ag and Mg by AA and AF, but a modified design did not incorporate a Cl osed chamber assembly (129). Using these systems, West and co- workers have determined Cd (129), Mg, Ag (128, 130), Pb, Zn, Bi, Tl, ‘53 (130), Au (131), so, Cu, Co, Ni, Hg (132) and Mn (132,133) by atomic 1:I‘J'Drescence spectrometry, while Ag, Mg (128,132), Au (131), Pb, Cu (132,134), Ni, A1 (134), Mn (133,135), Ni (136), Cr (137), Fe (133). 2“: M0 (139), Cd (140), V (141) have been measured by atomic absorp- “On. 27 Since, in contrast to furnaces, no further energy is available in the space above the filament-type atomizer, condensation of the atomic vapor occurs and results in a shorter lifetime of free atoms. In order to minimize interelement effects, it is necessary to view the atomic vapor close to the atomizer. It is possible to remove or reduce the severity of interferences by carrying out AF and AA measure- ments in such a way that an area extending only 0.5 mm above the fila— ment is illuminated (131,132,134-136). This method of measurement is referred to as a "limited field of view" in the literature and can also be employed for the study of the decay of atomic populations both by AA and AF (131,132,134,138—140,l42,143). The rate of decrease of the atom population above the atomizer is considerably less if the argon atmosphere is replaced by hydrogen. In operation, the glowing filament ignites the hydrogen, and a diffusion flame results. Using a modified form (144) of West's device and a hydrogen-diffusion flame, lhnos et al. (146) observed considerably less serious spectral and chemi- cal interferences from other ions. Winefordner and coworkers (143- 145,147) have used similar techniques and noted the improvement of free atom lifetimes above the filament. In comparing the West-type atomizer with furnaces, it should be noted that while the filament-type atomizer does not exhibit the same freedom from interferences and can be used for the analysis of a sl'haller number of elements, it has the advantages of compactness, 1ower power consumption, and ease of operation. In an effort to com- tyi'ita the virtues of the West filament and the Massmann furnace, a tr‘ansverse hole of 1.5 Illll in diameter was drilled in a 5 11111 diameter "3S (to produce a miniaturized furnace (148). Other versions, which 28 are basically similar to the miniature furnace, involve the use of a carbon cup or tube (149). The West-type atomizer, the carbon cup and the carbon tube are all commercially available from Varian Techtron (Walnut Creek, Cal.). The various versions of graphite or carbon filament atomizers have had widespread clinical, environmental and industrial applications. They have been used for the determination of Mg (150), Zn (150), Cd (151), Cr (152), Pb (146,150), Fe (150,153), and Cu (150,154,155) in blood (146,150-152), plasma (150), serum (153-154) and tissue (155), for the determination of Cd in food and air (151) and for the analysis of Cd in sea-water, sugar and glycerine (152). Mini-Massmann type carbon rods have been employed for the measurement of Au and Ag (156) as well as Sb (157) in geological and metallurgical samples. Alder and West made direct determinations of Ag and Cu in lubri- cating oil (158) and Hall et al. (159) evaluated the mini-Massmann carbon rod and the West-type filament for routine determinations of A9. Cu, Fe, Ni and Fe in lubricating and crude oil. Winefordner and coworkers (160) have determined Ag and Cu in jet engine oil by atomic f’ltaorescence and have analyzed for Ag, Cr, Cu, Fe, Ni, Pb, and Sn in used jet engine and reciprocating engine oil by atomic absorption (75] ). Finally, West and coworkers have determined Ni (162) and V ”53) in crude and residual fuel oil, while Bratzel and Chakrabarti (1‘5‘3-) and Robbins (165) have reported the determination of Pb in petV‘oleum and petroleum products. T‘ .. 29 3. Cold Vapor Technique A widely used method for the measurement of traces of Hg, As, Se and Sb is the cold vapor technique utilized for both AA and AF. For mercury compounds, the sample is introduced into a reduction cell, and the resulting mercury vapor is swept into a long path length absorption cell for measurement. For As, Se and Sb, the resulting gases are either introduced into an electrically heated absorption tube (166) or into an argon-hydrogen flame. The technique has been used for the determination of mercury in air samples (167), in fish, sediment, and wheat flour (168-170), in urine and plasma (171), in geological materials (169,172,173) and in clothing materials (170). 4. Nonflame Atomizers Using Electrical Discharges a. Cooled Hollow Cathodes. In 1959 Russell and Walsh (174) suggested that cathodic sputtering might be a means to obtain an atomic vapor from solid samples for use in AA. Gatehouse and Walsh (175) were the first to use a hollow cathode tube as an absorption cell, and they determined Ag in metallic copper in the range of 0.005- 0.05%. It must be noted that the determination of Ag in Cu is one of the most favorable since the sputtering yields are high. Later, Walsh ( 7 76-177) reported that Sullivan, who used water-cooled cathodes and a. frlow through gas system, had analyzed phosphorus and silver in copper, 31“! silicon in aluminum and steel. With a similar system, Gobels and Brody (179) evaporated sample V01 umes onto the inner wall of an aluminum hollow cathode and detected 1 149 of Na, Ca, Mg, Si, and Be, although considerable inter-elements eT‘I‘f’ects were observed. Gobel (180,181) has also developed a technique I A r U ." sz Cl. 5. 30 for the determination of isotopic composition by using this device in atomic absorption spectrometry. Recently, Walsh and coworkers (177, 178) determined C, Si, P, Mn, Ni, Cr, Mo, V and Cu in iron-base alloys with a non-dispersive atomic fluorescence system which incorporates a sputtering chamber. The standard deviation at higher concentra- tions was reported to be %l%, and the analysis time was 3-4 minutes. Harrison and Daughtrey (182) have also used a demountable hollow-cathode lamp with a steel cathode for trace analysis of gold by atomic emis— sion. b. Hot Hollow Cathodes. Atomic absorption methods with hot hollow cathodes were first described by Ivanov et al. (183). The cathode cylinder was held in the optical path by a thin molybdenum foil. This configuration causes the conductive heat losses from the cathode to be smaller than radiation losses, and the cathode becomes red-hot during the discharge period. The sample, placed in the cathode as a solid (Jr as a solution, is entirely evaporated in the discharge. A similar type of absorption cell incorporating a graphite cylinder was employed 419' Massmann (184). The discharge was operated at an argon pressure in the range of 1-10 torr and a power of up to 1 kW. It was noticed that the atomic vapor was produced by boiling the sample in the hot hollow cathode and only to a lesser extent by cathodic sputtering. This is a definite advantage of the hot hollow cathode over the cooled version. If the hot hollow cathode is operated with a half-wave rectified cu"went and the primary radiation source is activated only when the absitzarption cell is dark, then AA measurements can be made without s“’EEctral interference. Since the graphite cathode, heated by an e\ ectrical discharge at reduced pressure, cannot reach temperatures .vl' a" O”, 1". 'h I I '5; CV 31 higher than 2000 °C. only volatile elements such as Zn, Pb and A1 can be analyzed. Since the residence time of the atoms in the absorp- tion tube is short, the method is not suitable to detect small amounts if sample size limitations exist. With this device, Massmann obtained detection limits for Ag, Sb, Zn, Cu, Cd, Mg, Mn and Cr in volatile matrices. c. Arc and Spark. The use of an ac or dc arc for sample atomiza- tion in AA has been reported by several workers. Kantor and Erdey (185) employed an electronically controlled ac arc and made a time- resolved study of Cd, Na, Al, and Pb using Osram or tungsten filament lamps as primary radiation sources. It was noticed that for Na the amount of emission was negligible 2.4 msec after the discharge, and for Al and Pb absorption occurred only when the arc was operated in an argon atmosphere. Belyaev et al. (186) utilized an arc atomizer for the determination of silver in rocks and reported a sensitivity caf 1.5 x 10'7%. The standard deviation at the sensitivity threshold was 30%. Marinkovic and Vickers (187) described a long-path stabilized dc arc for the determination of Al, 8, Mg, V and W. The sample was in- troduced into the arc as a solution aerosol. The result showed that "either the analytical curves nor the sensitivities were appreciably different from those obtained with chemical combustion flames. The lack of improvement was due to the fact that the optimum region of t““* arc for AA measurements was far removed from the high temperature mm at region. Jones et al. (188) combined a direct current capillary arc operat- i“SI in argon at atmospheric pressure and a device which produced fine 32 particles from the surface of a solid metal for use in atomic emission. The aerosol generator was a device in which the sample acts as the cathode of a dc arc discharge. A flowing argon gas stream carried the aerosol particles to the capillary arc discharge. The method was applied to the determination of 10 elements in stainless steel alloys. Winge et al. (189) used a similar aerosol generator, but employed a flame as the atomization device for the determination of Cr, Mn, and Ni in steel. Finally, Robinson (190) used a conventional nebulizer and a spark for AA measurements of aluminum. d. Plasmas. Several types of electrically generated flames have been introduced for atomic emission and atomic absorption in recent years. The plasmas can generally be classified in terms of the fre- quency of the exciting field. The induction-coupled plasma and plasma torch were first employed for atomic emission analysis (191,194). The plasma generator usually had a frequency of 4-36 MHz and a power level of 500-5000 W. Wendt and Fassel (195) demonstrated the use of ‘the induction-coupled plasma for the determination of nine elements txy'AA. Friend and Diefenderfer (196) also determined elements forming refractory oxides by AA. With a similar plasma but a different sample introduction system, Vei llon and Margoshes (197) reported pronounced inter-element effects. Dinkinson and Fassel (198) reported the detection limits of 26 elements 1" AE and stated that the conclusions of Veillon and Margoshes were has ed on observations made with particular experimental conditions and thiitztheir deduction does not appear to be generally applicable. Boumans a'lti DeBoer (199) evaluated induction-coupled plasmas for simultaneous mUltielement analyses and determined 33 elements by AE. In many '.n'.' I?” ‘ .-.|31 33 instances detection limits lower than published previously (198) were established. Induction-coupled plasmas have been used for the deter- mination of trace metals in microliter samples in oils, organic com- pounds and biological fluids (200,201). Microwave plasmas, operated at a frequency of 2450 MHz, have also been used for spectrochemical analysis. Most of the work in this area is related to atomic emission. Mavrodineau and Hughes (202) excited the spectra of 75 elements, and Murayama et al. (203) reported the detec- tion limits of 25 elements. Similar systems have been used for the detection of Cd, Ga, In, Hg, and Zn (204). West and coworkers utilized a 40 W microwave-induced plasma in conjunction with a sample introduc- tion system such as a platinum or tungsten loop for the analysis of 12 elements in AE (205). Recently Busch and Vickers discussed the fundamental prOperties of a low-pressure, microwave-induced plasma as an excitation source for analytical spectroscopy (206). e. Radiation Heating. Pulsed discharge lamps and pulsed lasers are two different types of radiation sources which have been employed F“ ‘t() evaporate samples for atomic absorption by heating. In both tech- niques the solid samples can be evaporated by a pulse of short dura- tion. For laser heating, the evaporation is done by a focused laser ~. beam, while for the pulsed discharge lamp, no optical arrangement is L used to focus the radiation onto the sample. The pulse operation of a capacitor discharge flash lamp allows prociuction of large amounts of light with relatively modest equipment. Kue bler and Nelson (207) estimated that the irradiance of such a lamp 2 “as as high as 60 KWcm' . The same authors employed this system to °btain spectra of gaseous species formed at flash-heated surfaces (208). 34 Later absorption spectra of Au, Al, Ag, 8, Ca, Cu, Dy, Eu, Fe, Mg, Pb, W, and Zn were produced directly from solid elements, chloride deposited on graphite strips, or from impurities in W (209). In these experiments, the duration of a flash was about 3 msec. A second flash lamp was used as the primary source. The source flash persisted for 20 usec and could be ignited with variable delays of up to 6 msec. after the start of the heating pulse. A photographic emulsion was used as the detection system, but no quantitative data were reported. In some cases loss of sensitivity for an element may occur due to vaporization as a molecule. Intense absorption of A101 was an indica- tion of this effect. Since the invention of a pulsed, high—power laser in 1960, there have been numerous publications regarding the utility of a laser for atomic spectroscopy (210). In order to achieve direct atomization of certain phosphors and photocathode materials for atomic absorption .analysis, Atwill (211) vaporized analytical samples with the con- centrated output of a gigawatt pulsed laser. This could provide pulses crf 10'I0 chm'2 at the surface of the substance under study and was capable of vaporizing a sizeable portion of any known material. Because of the initial cost and wear on optical elements, Mossotti Git fill. (212) used a laser pulse of medium power as a means of atomizing 0-"--10 ug of analytical material for atomic absorption. The atomic va Dor produced by the Q—switched laser was excited by a Xenon flash ‘iinnla, and the atomic absorption due to the transient vapor cloud was dis‘blcalayed as a function of time on the screen of an oscilloscope. 5‘ “ce the standard deviations of measurements were high, the concen- tV‘ation limits for Ca, Ag and Cu were reported at 10 percent absorption 35 and were 0.0025, 0.0040 and 0.0035 percent respectively. It was also noticed that the laser atomization efficiency was critically dependent on the nature of the target material. Piepmeier and Malmstadt (213) used a Q-switched laser and studied energy absorption in the plasma of an aluminum alloy. It was stated that matrix effects depend upon the atmosphere as well as the sample material and should be significant. A similar system was used in conjunction with pulsed hollow cathode lamps as primary radiation sources for AA measurements (214). Calibration curves were obtained for Cu (0.002-0.15%) in an aluminum alloy and Mn (0.013-0.124%) in graphite pellets. 5. Controlled Explosion Atomizer A new, direct application of chemical energy for atomic absorption has been engineered by Venghiattis (215). The method consisted of controlled burning of a blend of a solid propellant (oxidizer and fuel mixture) and the solid material under study. The generated atomic vapor was introduced into the absorption light path of a conventional atomic absorption spectrometer. The technique has been applied to the determination of trace elements such as Au, Ag, Cu, Pb, and Zn in ores. Although some sample dilution was necessary, the method was sensitive to ppm concentrations of gold in ore. However, it may be difficult to achieve better than 5 percent precision because the method requires considerable sample preparation, especially for materials that are not easily pulverised. The technique may find considerable application in field analysis if transportation of gas cylinders, solutions and chemicals is inconvenient. 36 0. Radiation Sources for Atomic Absorption and Atomic Fluorescence Spectrometry One of the most important components of atomic absorption and atomic fluorescence instrumentation is the radiation source. At least part of the remarkable success of atomic absorption as an analytical technique must be attributed to the successful manufacture of hollow cathode lamps and their commercial availability. 0n the other hand, the lack of suitable radiation sources for atomic fluorescence is a practical disadvantage at the present time. Various sources such as metal vapor discharge lamps, hollow cathode lamps (HCL's), high in- tensity HCL's, demountable and water-cool HCL's, modulated and pulsed HCL's, electrodeless discharge lamps (EDL's), modulated EDL's, tempera- ture-controlled ELD's, continuous sources, induction-coupled plasmas, flames, arcs, sparks and lasers have been used for AA and AF spectrom- etry. A description of almost all of these sources has been given in literature (216,217) and reviewed recently (218). Since lasers and pulsed HCL's have not been reviewed in great detail, a review of the previous work on these sources is presented here. 1. Radiation Source Requirement in AA and AF Spectrometry One of the most important characteristics of a radiation source for both AA and AF is the shape of spectral line emitted by the source. The six factors which govern the shape of spectral line are Natural, Doppler, Lorentz, Holtzmark, pm Resonance), Stark, and Self-absorption' broadening. Lorentz and Holtzmark, broadening together are often called pressure broadening. In the conventional HCL used in atomic absorption, broadening effects that reduce sensitivity most are 37 Doppler and Self-absorption broadening. This situation may change, however, if the lamp is operated in the intermittent high current mode. Before getting to the discussion of pulsed HCL's, the general require- ments of radiation sources are summarized. Because of the differences in the mechanisms of the AA and AF phenomena, there are some differences in the source requirements. For Atomic Absorption: a. The source should radiate light of the element of interest without interference from other spectral lines originating from other components of the source. b. The resonance spectral line must be sharp and bright against a very low background. c. Radiation intensity should be high for low lamp currents. d. The source should have short and long term stability (low noise and drift). e. Ignition and burning voltage should be low to permit pulsed or modulated operation. f. The source should be commercially available, easy to operate, of low cost, small in size, capable of being operated continuously and safely, and should have a long shelf-life. For atomic fluorescence the source requirements are similar to those for AA except that: a. Since the atomic vapor produces essentially monochromatic fluorescence emission, the purity of the source spectrum is not as important as it is for AA. Scattering of light in the vapor cell, if present, may result in erroneous measurements of the fluorescence signal, however. 38 b. The source need not radiate such narrow spectral lines. c. The fluorescence radiation observed in an atomic fluorescence measurement is directly proportional to the radiance of the excitation source as is shown by Equation (9) in Chapter III. Thus, the main requirement is a high intensity of the resonance line at the absorption wavelength peak. In fact, if a radiation source such as a laser should provide a highly excited system which is near saturation, then the fluorescence emission shows relatively little dependence upon changes in laser intensity or changes in collisional quenching (219, 220). It should be noted, however, that under this condition, the fluores- cence emission still retains a linear dependence (assuming dilute vapor) upon analyte atomic density in the atomizer. 2. Laser Sources Most of the work done with lasers as excitation sources has been concerned with atomic fluorescence. However, a tunable dye laser has been used for sodium analysis (221) by atomic absorption. The limit of detection was reported to be 2 ng for sodium, and the calibration curve was linear from 2 to 500 ng/ml. The first application of a tunable dye laser for atomic fluores- cence was described by Denton and Malmstadt (222), who reported the analysis of barium. A stable, pulsed, tunable dye laser pumped with a N2 laser has been employed in conjunction with a fast response multiplier phototube and a boxcar integrator by Winefordner and co- workers (223-225) to excite the atomic fluorescence of 27 elements in flames. The dye laser system has a peak power of greater than 10 kW from 360-650 nm, a pulse repetition rate of about 1-25 Hz, a pulse 39 half-width of about 2-8 msec, and a spectral half-width of about 0.1 - 1 nm. The calibration curves were linear over 2-4 orders of magnitude, and detection limits were improved for some of the elements. Since strong ionic fluorescence has been observed for some of the rare earth elements, ionic as well as atomic fluorescence detection limits were reported. For the majority of the rare earths, atomic fluorescence detection limits were essentially the same as the best ones reported by flame atomic absorption, but inferior to flame atomic emission. It should be noted, however, that laser excitation has several advantages over conventional continuously operated line or continuous sources. First, only one source is needed. Second, it is possible to eliminate noise due to scattering of exciting radiation from particles within the vapor cell if nonresonance fluorescence is used. Non- resonance fluorescence with conventional sources is not useful because of low quantum yields. However, the high spectral irradiance of the laser allows these lines to be readily observed. This has been shown for A1 and some of the other elements (224). Third, in addition to resonance fluorescence, other types of fluorescence have been shown to be analytically useful. Fourth, since the saturation condition is approached with laser excitation, the proportional dependence of the fluorescence signal upon the quantum yield, Y, observed at low irradi- ances, is removed. This has been also shown theoretically and experi- mentally both for monochromatic and broad-band laser excitation (219, 220,226,227). In other words, assuming that the atomization efficiency does not change, the saturated fluorescence signal obtained with vapor- cells of different quantum yields would be the same. For example, one could take full advantage of the greater reduction of chemical 4O interferences and/or the greater atomization efficiency of the Csz/NZO flame over the 02/Ar/H2 flame (226). Fifth, when a laser or another high irradiance source is used as an excitation source, the dynamic range is increased. For example, analytical curves for indium were linear from 50 to 10,000 ppm and for strontium from 100 to 50,000 at high irradiances (227). The calibration curve bent at lower concentra- tions when the laser irradiance was decreased. Laser excitation has its disadvantages, however. First, for con- centrations above the detection limit, one of the major sources of noise is laser pulse-to-pulse variation. One convenient way to mini- mize this effect is to assume that the long term average source radiance is constant and then integrate the output signal for a period of time. Furthermore, it has been shown that under saturation condi- tions the fluorescence signal is not greatly influenced by source stability (219,226). Second, the scattered light intensity continues to increase in direct proportion to the laser spectral output, while the excited atom population begins to level off as saturation is ap- proached. It is possible, of course, to use a non-resonance line. Third, at the present time, the enormous consumption of electrical energy by a laser necessitates the focusing of the laser beam to a small region of the atomizer to achieve saturation. Furthermore, typical laser beams may not have uniform cross-section and a careful consideration of the entire optical system is required. Fourth, the lower wavelength range of the present dye laser is approximately 2500 A, although it may be possible to extend the wavelength range down to 2000 A, by frequency doubling. Fifth, the very high cost of the dye laser system and its maintenance is prohibitive at the present 41 time. 3. Pulsed Hollow Cathode Lamps Because of the above disadvantages of the laser, the spectral source which appears to be highly suitable for atomic fluorescence with respect to cost, availability, wavelength coverage, selectivity and intensity is the hollow cathode lamp operated in a pulsed mode. The first analytical application of pulsed hollow cathode lamps was discussed by Dawson and Ellis (228). By passing large currents (300-600 mA) of short duration (15-40 usec) repetitively through conventional hollow cathode lamps, such as Ca, Co, Cu, Mg, Mn and Sr lamps, an increase in intensity from fifty to several hundredfold was achieved. The enhancement obtained was greatest with the metals of lower melting points, and there was no corresponding increase in the intensity of the spectrum of the filler gas. The detection limits and calibration curves obtained were comparable to those reported in the literature with other sources. From these observations and from the fact that the mean current was about 2 mA, the authors concluded that the resonance lines were not significantly broadened in the pulsed mode. Using a similar system in conjunction with a resonance monochromator, Willis (229) reported that the fluorescence signal from the resonance monochromator, excited by a pulsed copper hollow cathode lamp, was only about five times that obtained by operating the lamp in the dc mode at the same average current. In other words, the large increase in emission intensity was achieved at the expense of an increase in broadening of the emitted resonance lines. Since the resonance 42 monochromator has a spectral bandpass of about 0.01 A, this experi- ment showed that the increase in intensity at the center of the line, obtained with pulsed operation, is not as large as found by Dawson and Ellis for the integrated intensity over the whole line. The increase in broadening can be explained in terms of the excita— tion mechanism (216,230) in a hollow cathode lamp Operated at high currents. Initially, the discharge filler gas is excited or ionized by electron impact. The discharge ions then collide with the cathode walls where they are neutralized or release a neutral atom of the cathode material. Then, neutral atoms are brought into the plasma where they can be excited by electron impact or by collisions of the second kind. At higher currents, a larger amount of the discharge gas is ionized and, hence, more neutral atoms are formed. These neutral atoms can either form a deposit at the cathode or leave the discharge by diffusion. The plasma inside the hollow cathode is emitting and partly absorbing the resonance emission. The slab (231,232) outside the electrode contains atomic vapor diffusing from the discharge and is only absorbing. If the concentration of these absorbing atoms between the emitting atoms and the lamp window is significant, additional broadening will be experienced by the resonance lines (233). Therefore, it is not unreasonable that high current pulsed hollow cathode lamps exhibit effects of broadening or self-reversal. Despite the increase in emission line width, pulsed hollow cathode sources are still very useful for atomic fluorescence. Weide and Parsons (234), who use pulsed hollow cathodes in conjunction with a box-car integrator, reported that the limit of detection for zinc (0.003 ppm) with the pulsed system was nearly three orders of 43 magnitude lower than with a conventional system. Mitchell and Johans- son (235-236) used the pulse-modulated technique for an elaborate multi- element atomic fluorescence system. However, they gave no indication of any gain in detection limit using their own pulsing system. Massmann (69) employed HCL's of his own design and manufacture to excite fluorescence for nine elements. Sufficient intensity was obtained by manually pulsing the HCL at a current of several hundred mA for periods of about 15 seconds during which fluorescence measure- ments were made. No information was given on the profiles of the resonance lines emitted by this source, but the author commented on the comparatively low intensity of such sources. Nevertheless, the detec- tion limits obtained by AF for Zn and Cd were superior to those ob- tained by AA. For seven other elements investigated by both tech- niques, absorption was found to give superior results. Several authors have used pulsed hollow cathode lamps for atomic absorption and atomic fluorescence measurements, and results obtained indicate little additional line broadening due to pulsing (214,237- 241). Gains in light intensity due to pulsing were about one to two orders of magnitude (214,238-241). The exact intensity increase de- pends on the element and the filler gas and may vary from lamp to lamp. The lamp lifetime has been reported to be longer in the pulsed operation than in the dc mode (228,240). It has been shown that the ON time, the pulse width/period ratio, the number of periods, and the wait time between each series of ON-OFF periods were all important for providing reliable operation (240). Various versions of pulsed hollow cathode systems have been used for the selective modulation of resonance lines, for the construction 44 of atomic absorption spectrometers with no monochromator, etc. (242- 245). Pulsed hollow cathode sources were also used by KielkOpf (246) who operated them at peak currents between 30 and lSOO A. With this pulsed hollow cathode lamp, excitation of up to the third spectrum of iron and aluminum and parts of the fourth spectrum of the rare earths could be observed. The plasma within this source showed an ionization temperature of about l2,000 °K, electron densities of about l015 cm'3, and electron temperatures of about l5,000 °K. This type of source has proven very useful in classifying lines in ionized rare earth spectra and in analyzing the processes which go on in pulsed discharges. E. Multielement Analysis The important role of minute amounts of elements in physical, chemical and biological systems has emerged as methods of analysis have increased in sensitivity. A wide variety of various instrumental techniques such as X-ray fluorescence spectrometry, spark and arc emis- sion spectrometry, spark source mass spectrometry and thermal neutron activation, involving widely different principles and capabilities, have been introduced for simultaneous multielement analysis. Each of these techniques has its own advantages and disadvantages. While the detection power increases in the order the techniques have been arranged, there is a corresponding increase in the cost and complexity of the instrumentation. In general, it is desirable for any analytical system for simultaneous multielement analysis to meet certain general requirements. First, the system must be capable of handling small 45 sample sizes. Second, the techniques must be capable of determining accurately and precisely many elements at the 0.02-l0 ng/ml concentra- tion level. Third, absolutely no sample preparation should be needed. Fourth, a broad dynamic range is required. Fifth, the technique should be rapid, easy and safe to operate and inexpensive. In a recent review article, Busch and Morrison (247) compared flame spectrometry and other multielement techniques with respect to some of the above requirements. It is an obvious fact that, at the present time, there is no analytical technique which can fulfill all of the above general requirements. However, flame and nonflame spectrometry can fill the gap among the more expensive techniques and are adequate and satisfactory alternatives. There is no question that compared to AA and AF, flame emission is more adaptable to multielement analysis. The development of semi— conductor photo-electronic detection devices (248) will provide a detection system other than the multiple slit-multiple detector system used in expensive emission spectrometric system. Experimental results for the nitrous oxide-acetylene flame (249) indicate that flame emis- sion may cover some 40 elements with detection powers equivalent or superior to flame absorption. Thus, for complete coverage of the common group of some 70 "spectroscopic" elements, flame atomic absorp- tion and/or flame atomic fluorescence must be used complementarily to flame emission (l99). If an induction-coupled plasma similar to that described by Fassel and coworkers (l98,201) is employed, the detection power of many elements would be in the range of 0.02-l0 ng/ml and complementation by atomic absorption and/or atomic fluorescence is not required (l99). In chasing a plasma system, however, the following 46 facts should be considered. First, the high excitation temperature necessitates the requirement of a medium or high resolution spectrometer. Second, matrix effects (l99) do occur and must be studied systematically. Third, the economic factors, such as cost of equipment and large amounts of argon consumption must be carefully considered. Another interesting technique of performing simultaneous multi- element analysis is a completely nondispersive method. In contrast to energy dispersive X-ray fluorescence where the achievable detection limits are in the range of l0-50 ppm, the working range for a non- dispersive atomic fluorescence system is ppb to ppm. The previous work on multielement atomic spectrometry has been reviewed (47,247), and it is the author's intention to briefly review the nondispersive systems. The first nondispersive system for atomic fluorescence was pro- posed by Jenkins (250), who noted that a filter could replace the mono- chromator to isolate the resonance line of the element of interest. Larkins et al.(25l) suggested the application of solar-blind photo- multiplier detectors for atomic fluorescence flame measurements. Since most of the noise emitted by the flame appears at wavelengths longer than 3000 A, it is possible to allow the fluorescence radiation to fall directly on the detector and still avoid flame background noise. Vickers and coworker used this system (252) and a similar system incorporating a chlorine cutoff filter, which responded to radiation of wavelengths shorter than 2800 A (253). Larkins (48) used a solar-blind photomultiplier in conjunction with a sheathed flame and examined the fluorescence of 24 elements. He found that his system gave better detection limits for nine elements compared to atomic ab- sorption flame spectrometry. Similar nondispersive systems have been 47 used by other investigators (254-256). Nondispersive atomic fluorescence methods for multielement analysis were first described by Michell and Johansson (257,258). Their instru- ment was used for the determination of Ag, Cu, Fe and Mn using an air- hydrogen flame, and their results were quite comparable with those ob- tained by others who used a conventional monochromator system. A six- channel atomic fluorescence spectrometer based on this system was manufactured by Technicon Corp., Tarrytown, NY. The AFS-6 (now taken off the market) instrument has been used for the determination of metals in soil extracts (259) in aluminum alloys (260), and with preconcentra- tion by an automated solvent extraction procedure (261). Using the same principle, Cordos and Malmstadt (24l) described a similar dispersive system which incorporated a programmable monochromator for high Speed wavelength isolation. Their instrument may be used to determine up to l2 elements sequentially. III. THEORETICAL DESCRIPTION OF NONFLAME ATOMIZATION AND ANALYTICAL SIGNAL The first section of this chapter discusses the various distribu- tion laws for flame and nonflame atomizers and the influence of various processes on the atomizer temperatures. Section 8 presents the radiance expressions for atomic absorption and fluorescence spectrometry along with the expressions providing the peak atomic concentration in flame and nonflame devices. The atomization processes in nonflame atomizers are discussed in the third section, and the influence of natural and forced convection on nonflame atomization are discussed in the fourth section. In the final section, the time behavior of analytical signals with nonflame devices is described. A. Distribution Laws In deriving the radiance expression relating the concentration of an element in a sample to the readout, it is assumed, in both AA and AF, that the atomizer is in a state of "local" thermodynamic equilibrium. In thermodynamic equilibrium, the spectral distribu— tion of radiation density and the distribution of energy over the various internal and translational degrees of freedom as well as the distribution of ionization and dissociation products are determined by T, the temperature of the system. If it is assumed that singly charged ions and free electrons are the only ionization products in a flame atomizer, there are five statistical distribution laws (35) for the various forms of energy which are summarized below: 48 49 l. Planck's Radiation Law b Planck's radiation law relates the spectral radiance 8A of a black body to the temperature T and the wavelength A by Equation l b _ ZhCZA-S (1) BA ‘ expfhc/AkTI-l where h is Planck's constant, c is the speed of light and k is the Boltzmann constant. The absorption factor a(A) is defined as the ratio of the absorbed radiant power to the incident power and, by definition is equal to unity for a black body. The spectral radiance BA of an arbitrary radiating body with a(x)ri period. The processes of natural and forced convection are dis- Cussed in the following section. 0. Natural and Forced Convection The process of natural of free convection heat transfer happens FINE—1P Whenever the atomizer is located in a fluid medium at a lower or higher t.elnperature than that of the atomizer. Because of the temperature difference, heat flows between the atomizer and the gaseous environ— ment and leads to a change in the density of the gas in the vicinity 0f the atomizer surface. Thus there is an upward flow of the less dense gas and a downward flow of the higher density gas. The only 62 force responsible for natural convection currents in the heated filament case is gravitational attraction. When the motion of particles is caused by some external agent, such as an inert gas flow, we have forced convection. For both natural and forced convection, the motion of the gaseous environment may be laminar or turbulent. In laminar flow, the particles move in an orderly sequence without passing one another, while for the turbulent flow the particles move in a zig zag manner with irregular paths. The statistical averages of the motion of a large number of individual particles is regular and predictable, however. Irrespective of whether the flow is laminar or turbulent, the concept of the boundary layer can be applied to describe the trans- formation of various processes. This concept, which was introduced first by Prandtl, divides the flow field around a body into two domains; a thin layer covering the surface of the body where the viscous forces are large and the velocity gradient is great, and a region outside this layer where the effects of viscosity are negligible and the velocity is nearly the same as the free-stream value. 1. Forced Convection Here we consider the flow of inert gas around the filament atomizer. At the leading edge of the atomizer, only the gaseous molecules in immediate contact with the surface are slowed down, while the remaining particles continue at the undisturbed free-stream velocity. As the inert gas proceeds along the atomizer, the thickness of the boundary layer increases. The velocity profiles near the leading edge are representative of a laminar boundary flow. This laminar flow within 63 the boundary layer changes to a turbulent flow after a certain distance from the leading edge. This distance is related to a dimensionless quantity called the local critical Reynolds number. In general the point of transition depends on the surface contour, the surface rough- ness, the disturbance level and even the heat transfer. If the surface is rough, or if disturbances are introduced into the flow, as for ex— ample by a careless design of a gas sheath or atomizer holder, the flow may become turbulent at lower Reynolds numbers than when the flow is calm. The boundary layer concept is of great importance to an under- standing of convective heat transfer. The rate of convective heat transfer between the atomizer surface and gaseous environment is given by q = AfiC(TA - Tm) (26) where q is the rate of heat flow, A is the surface area of the atomizer, EC is the average convective heat transfer coefficient, TA is the tem- perature of the atomizer and T00 is the temperature of the inert gas away from the atomizer. The convective heat transfer coefficient is a complicated function of the gas flow, thermal properties of the fluid medium, and the geometry of the system. There are four (269) general methods available for the evaluation of he with each method having its limitation which restricts its scope of application. These are: l. Dimensional analysis combined with experiments. 2. Exact mathematical solution of the boundary-layer equations. 3. Approximate analysis of the boundary layer by integral methods. 64 4. The analogy between heat, mass, and momentum transfer. It should be noted that no single method can solve all the problems. The limitation and the scope of each technique have been described (269), and we are only concerned here with the first method, which has found the widest range of application. 2. Dimensional Analysis for Forced Flow in a Furnace Dimensional analysis is mathematically simple and facilitates the interpretation of the experimental data by correlating the data in terms of dimensionless groups such as the Reynolds number. The prin- ciple disadvantage of the technique is that it contributes little to our understanding of the transfer process and at least a preliminary theory is required before dimensional analysis can be performed. The algebraic theory of dimensional analysis will not be developed here. A rigorous treatment of the mathematical background has been given by Langhaar (269). The procedure in dimensional analysis is simply as follows. First, an arbitrary system of primary dimensions is chosen so that all variables can be expressed in terms of them. We shall employ the primary dimensions of length L, time 9, temperature T, and mass M. The dimensional formulae of velocity and mass flow rates would be L/B and M/B respectively. Second, the Buckingham H-theorem (270) is used to determine the number of independent dimensionless groups needed to express the relation describing a phenomenon. Third, ex- perimental data are correlated by plotting the dimensionless numbers against each other. Fourth, the Principle of Similarity is employed to describe the behavior of two systems if they have the same 65 dimensionless number(s) and vice versa. 3 . Buckingham II-Theorem The required number of independent dimensionless groups, Idiml is given by Idimt = Pq - Pd (27) where Pq is the total number of physical quantities (such as density, specific heat, heat transfer coefficient) and Pd is the number of primary dimensions required to express the dimensional formulae of the Pq physical quantities. If these dimensionless groups are desig- nated as n], H2, etc., the equation describing the relationship among the variables has a solution of the form F(II],112,II3,...) = o (28) The above theorem can be applied in correlating the experimental convective heat transfer data for an inert gas flow through a heated tube or a filament atomizer. Although we shall consider the heated tube here, the treatment of other systems such as wire loops and other filament atomizers is quite similar. The physical quantities for a heated tube are listed in Table l. Since there are seven physical quantities and four primary dimen- sions, we need three dimensionless groups to correlate the data. In order to find these groups, we write H as a product of the variables, 66 Table l. Physical Quantities and the Dimensional Equations for a Heated Tube Dimensional Physical Quantities Symbol Equation Heat transfer coefficient hC M/63T Tube diameter D L Density of the inert gas 6 M/L3 Viscosity of the inert gas u M/Le Velocity of the inert gas V L/e Thermal conductivity of the inert gas k ML/B3T Specific heat at constant pressure CP LZ/BZT each raised to an unknown power n = DakacpdueCfpficg (29) Substituting the dimensional formulae into Equation (29), we get u =[L]a[ML/B3T]b[L/e]c[M/L3]d[M/Le]e[L2/62T]f[M/63T]g (30) The exponents of each primary dimension must separately add up to zero if H is to be dimensionless. Since there are four equations containing seven unknowns, we choose values for three of the exponents in such a way that each exponent is independent of the others. By solving the four equations simultaneously, the value of the exponents are obtained and we get the following groups 5 D ”1 "‘E‘ (3‘) 112 = V—EE (32) Cpu “3 ‘ T (33) H] is called the Nusselt number, Nu’ and can be physically inter- preted as the ratio of the temperature gradient in the fluid im- mediately in contact with the atomizer surface to a reference tem- perature gradient. Once its value is known, the convective heat trans- fer coefficient fic can be calculated _ is It can be seen that for a given Nusselt number, He increases as the thermal conductivity of the inert gas increases. It is also obvious that an increase in the significant length 0, causes EC to decrease. In general the parameter 0 specifies the geometry of the object from which heat flows. For the heated tube, 0 represents the diameter. Thus, as far as the convective heat transfer is concerned, it is desirable to minimize the diameter of the atomizer. The second group, H2, is called the Reynolds number, Re' Its magnitude represents the extent of turbulence in a system. The higher the number, the more turbulence is found in the system. It can be seen from Equation (32) that the Reynolds number increases with the inert gas velocity and with the significant length of the atomizer. Since the atmosphere above a filament atomizer is cold, it is desirable to have minimum mixing of the generated atomic vapor with the inert gas. This would decrease the number of collisions of the analyte atoms with the cold inert gas molecules and minimize condensation of 68 the vapor or chemical reactions with impurities in the inert gas en- vi ronment. In atomic fluorescence, the number of quenching collisions would also be lower in a laminar flow than a turbulent flow. The dimensionless group H3 is known as the Prandtl number, Pr’ and relates the temperature distribution to the velocity distribu- tion. It is the ratio of two molecular transport properties, the k“! nematic viscosity v=u/p, which affects the velocity distribution, and the thermal diffusivity k/pCP which affects the temperature profile. It indicates the steepness of the temperature gradients in the flow field. For monoatomic and diatomic molecules the value of Pr ranges from 0.67 to 0.74. It can be seen that the seven original variables involved in the Problem of the heated tube are combined into three dimensionless groups, and the experimental data can now be correlated in terms of three l"ather than seven variables. Using Equation (28), we can write Nu = f(Re’Pr) (35) It 'i 5 now possible to apply the result of only one set of experiments to a variety of other problems. This can be performed by plotting the N'--'$‘-selt number ECO/k (or some product of Nu and Pr) versus the Reynolds nut"her. This correlation of the data permits the evaluation of the heat transfer coefficient for a system using any size filament-type atc>I‘i‘iizer as long as the Reynolds number of the system falls within the range covered in the first set of experiments. This is called “‘3 Principal of Similarity. According to this principle, the behavior of two systems will 69 be similar if the ratios of their linear dimensions, forces, velocities, etc. are the same. Therefore, in geometrically similar systems having the same Pr and Re numbers, the temperature distribution will be similar: These considerations are generally true for furnaces and fi lament-type atomizers where there is a forced flow of inert gas in the system. The followingsection considers the effects of natural co nvection . 4- Natural Convection in Nonflame Atomizers For a static system with no forced flow, the Reynolds number is superfluous and another dimensionless group is required to describe This parameter, the Grashof number, Gr’ is It is analogous to the Reynolds the natural convection. a measure of induced flow in the system. number for forced flow and is a criterion for the transition from 1ear! nar to turbulent processes in natural convection. For low Grashof numbers, the flow is very weak and thick, while for high Gr values, the flow is thin and vigorous, which leads to laminar instability. The latter is more or less characteristic of the production of the ‘31 ume of atoms in a heated filament. The Grashof number is given by 9293(TA'TQID3 or = 2 (36) Ll where g is the gravitational acceleration and the rest of the terms haVe been defined previously. It is interesting to compare Equations (32) and (36). For natural convection, the significant length of the 3“Runner is introduced as 03. Thus, the effect of the dimensions of the atomizer is much more important than the corresponding effect in 70 forced convection. It can also be observed, as expected, that as the temperature of the atomizer is increased, more turbulence is intro- duced in the system. For similarity of the temperature field in forced convection, it was stated that the Prandtl numbers should be equal. This applies also for free convection. Thus, when geometrically similar bodies are cooled or heated by free convection, both the temperature and velocity fields are similar if the Gr and the Pr numbers are equal at corresponding points. Similarly, when the Gr and Pr numbers are equal, the Nusselt numbers for the bodies are the same. The product of GrPr = Ra’ is called the Rayleigh number. The experimental data for natural convection are correlated by plotting the Nusselt number against the Rayleigh number. Nu = f(GrPr) (37) F1Qure 1 shows the correlation (27l) of the experimental data from Van-10.“ sources for free convection from horizontal wires and tubes "'11 ch have been obtained by employing an equation similar to Equation 37 - It can be seen that data for fluids as different as air, glycerin, and water are correlated for Gr numbers from W”5 to 107 for cylinders ranging from small wires to large tubes. 5 ° Forced and Natural Convection Combined in Nonflame Atomizers In the two preceeding sections the effects of forced and natural convection were treated individually. In most of the present nonflame altomizers, the two effects exist simultaneously and a combined or "mixed" convection process exists. In the region where both natural and forced Rd 2‘ I‘v AU I. 7T 20) - I ~ I I ‘ , ,r— -—I-——-[._-..._. . u l w Saw-u Mud 0.l-0000 ”on "and (flu 'oc '. (an MD. '0” In o lo..." Au OOI‘OOI I"? n C. ‘05 C I long-aw A» OW 00’ ‘ ,7 C ‘02, C 0 Ali... 0 'W Au 0”) 0035 | IO M L.» a In. A» i) ll) 0ll-l '1'!” I? l” o 'uu-ol Au 0 H 077-)“ '0 CM . Mud Owl-0°. 0 H 0'7-"4 I. no 0 'm-n (O) 0 ll ")3 I. ... a 00m Ami-n. 00!: I u "-04 H " a o... (a. con: I u "-03 § Dom Gav- m... 00!) I I! u-n . 9..., Volvo" out: l U s 70‘“ . no.» A. 70-“ l too-loo ,‘2 . Adm Worm i l ”-30 34-9? C 00“ M l 440 l ”-1.0 . Womb! h 7 05‘. 9 i ii.” 55-140 - We Magoo 0 ll ”0 vooo : LOW I Cvnhmhhfl.‘" O (m Mm! by ling. W l o 0 Dr T ‘ -- - l ’ - tn" ' / o 06 -___, — _.- .- , .. . Comm" ol I heme-vied Cum ! . ' ~qu ”on "m 0.4 T - __ 7 ‘ ° (N A ‘M -0490 )0") l _ ' ‘0550 l0” : -0.66l IO" 02 g ( ,- ~03" 101' -, -- - . 1.00 0 ‘ l.5l '02 ' 2.” l0 0 ~ 3 I 3.15 10’ ~ ~ - ' 5.37 '0‘ 9.33 10' TO.) I '4” I * no no’ “"""‘ F ‘ 51.3 W L 03.: IO’ 4.. L l —-5 —l -J —7 —l 0 I [I I) 4 3 6 V II 9 L091st I?) Figure l. Correlation of data for free convection heat transfer from horizontal cylinders in gases and liquids (27l). 72 convection effects are of the same order of magnitude, heat transfer is increased if natural convection effects act in the direction of flow and decreased if natural convection acts in the opposite direc- tion. The question that arises is under what circumstances can either natural or forced convection be ignored and under what conditions are the two effects of comparable magnitude. An indication of the relative magnitudes of natural and forced convection may be obtained from the Novier-Stokes boundary - layer differential equation, which describes the flow process. Careful examination of the various terms in the Novier-Stokes equation (272) indicates that the ratio Gr/Rez can provide a good qualitative indication of the buoyancy on forced convection. If the Grashof number is of the same order of magnitude or larger than Re2 , natural convection effects cannot be ignored compared to forced convection. Small values of Re/JE; result in small forced convection effects. Several investigators have reported empirical equations to predict the extent of either forced or natural convection in mixed processes. Most of the results, however, are reported for low tem- peratures where the effects of radiation and conduction losses are minimal. Even at low temperatures, these criteria depend on the geom- etry of the system as well as on a variety of the other factors (272). Collis and Williams have studied long horizontal cylinders 10 air (273). It was suggested that for long fine wires the condition for negligible natural convection in forced flow was Re39(Gr1/3)' The condition for the two effects to be of an equal order of magnitude is reported as Re=0(Gr1/2)° It should be noted that since filament 73 atomizers have a finite length, both the length 2.and diameter d are significant lengths so that the ratio 2/d must be incorporated in the dimensionless group. It has been shown (270) that for three dimensional shapes such as short cylinders the characteristic length is given by T1)-=D—+-D1: (39) where 0V is the height and 0h the average horizontal dimension. From Equations 32 and 36, the Gr' Re and Gr/Re2 numbers were calculated for a filament, 3 cm in length and 3 mm in diameter, at three different temperatures and for four different gases. All the physical properties were calculated at Tm = TA+1»/23 where the tempera- ture of the free-stream gas Ton was assumed to be 28 °C. and the diameter of the atomizer was used as the significant length. No significant change in the significant length results if both length and diameter are introduced in Equation 39. The flow rate was assumed to be l l/min and was distributed over an average area of l inz. The atomizer was located in a chamber with a cross-sectional area of 2x2 inz. The results are shown in Tables 2-4 and plotted in Figures 2-4. Table 2. Influences of Atomization Temperature and Sheath Gas Type on Natural Convection Atomi zer M r N2 He H2 Temp °C Gr Grpr Gr Grpr Gr PrPr Gr Grpr T000 43.5 31 45.4 31.3 0.81 0.56 0.969 0.68 1500 18 13.3 20 13.6 0.321 0.225 0.4 0.264 2250 10.5 7.91 0.218 0.157 74 ”I‘ll. ...¢Q¢..»tild‘ 0.1:? ¢P+o_xmm.P ¢P+o_xm¢.P omNN ep+opx-.e ¢P+opxmw._ ¢F+o_xm~._ ¢P+o_x~._ oom_ ep+opxe_.F ¢F+o~xwp., e_+o.xp_.F ep+opxmp.~ ooo_ m L i . N m a mmm\gw Nmm\gu Mwm\ga on new» . - Lm~weou< N: m: Nz LP< cowuum>cou emote; wen Putnam: mo muauvcmmz m>vumme meg co max» mac gunmen ecu mtaumtmaeoh cowumNPeou< mo mwucmapwcfi .e open» m-opxmm.~ m-o_xmm.m N.Sxm N-opxmm.m ommm m-opxu.m m-opxm.m m-opxm.m m-opxm ~-opxn.N N.2: “-0wae.~ ~-o_xuw.m oom— m-o_xm¢.m m-opxm.m m-o_xmm.m m-o_xm.m ~-opx_e.¢ ~-opx¢.m ~-o_x¢.¢ n-opx~.m coo, game we game we game we game me o. .agop N: a: N2 c_< to~eeoo< cowauw>=ou umugou co waxy moo sumogm ucm agapmgogemh cowum~eeou< to «meson—eca .m m_amb 75 (:1 AIR 0 N 50_ 2 . 0 He x10 EXPANSION A H2 x10 EXPANSION o... P, 10- {g::::f::::r:=1gr__ ‘ A 560 10““‘00 150‘0—"—“'—2000 ATOMIZER TEMPERATURE, °C Figure 2. Influence of atomizer temperature and sheath gas type on natural convection. 76 C] G) (5- O A ‘5' "o 4‘: 0) o: 2 _ 300 1600 1560 2600‘" ATOM IZER TEMPERATURE, °C Figure 3. Influence of atomizer temperature and sheath gas type on forced convection. thfNI N I.- 2 Gr /Re -10 77 1.3% r 1.1 Figure 4. [I] AIR 0 N2 <> ITE! n . H2 :/ TM ATOM IZER TEMPERATURE, 0(3 Influence of atomizer temperature and sheath gas type on the relative magnitude of natural and forced convection. 78 The following conclusion can be drawn. First, for both forced and natural flow, gases with smaller atomic or molecular weights produce more laminar than turbulent flow. Second, the flow becomes more laminar with an increase in temperature. Third, the ratio of Gr/Rez increases, as expected, with temperature, which means that at higher temperatures natural convection becomes more dominant. Further- more, since this ratio is very high, the effect of forced flow is neg- ligible. It should be noted that these calculations are only true for steady-state conditions and not for a transient type situation which actually is present with most nonflame atomizers. Transient convec- tion and conduction have been treated for bodies of different geometry at relatively low temperatures (272). Although investigations in this area can be very helpful in understanding the various heat transfer processes, they will not be treated here. Although the above calculations can provide an indication of the general behavior of the system, they do not specifically describe the behavior of the generated atomic vapor. Since the problem is of a very complicated nature, only the effects of the various heat transfer processes on the time dependence of the atomic vapor are discussed here. E. Analytical Signal in Nonflame AA and AF Spectrometry The analytical signal may be defined as the response of the mea- surement device to the presence of the analyte. This indicates that the information provided by the measurement device is a result of a ‘ . K." MK-L“ 79 convolution of its response characteristics and the physical property being measured. Time is the important factor in considerations of the analytical signal. Steady-state signals are produced when the analyte is continuously introduced in the system, while transient signals are encountered when discrete sampling is employed. Peak height and peak area measurement techniques are normally utilized in recording the analytical signal. The merits of each tech- nique are influenced by the time dependence of the analytical signal which itself is a function of atomizer temperature. In this section the relationship between the heating rate and various heat losses and gains will be developed for electrically heated nonflame atomizers. The temperature-time dependence will serve to indicate the time dependence of atom formation as well as the advantages and disadvan- tages of peak height and peak area as measurement techniques. I. Temperature-Time Dependence The temperature-time dependence is a function of a variety of factors such as the heating techniques, the input power, the heat losses by convection, conduction, and radiation. The influence of the heating technique is discussed in Chapter VI. For the following treatment it is assumed that the atomizer is heated by a constant current i and that the physiochemical pr0perties of the atomizer are not influenced‘ from one experiment to another. Furthermore, it is also assumed that the work done on the system by pressure-volume variations is negligible compared to the magnitude of the internal energy and various heat losses and gains. In general, we have U=Q where Q is the net heat transfer and U 80 is the internal energy, or qinternal energy = qelectrical + thomson heating ' qradiation ' - - (40) qconduction qconvection qsample For a filament-type atomizer of perimeter P, diameter D, and surface area 5, Equation (40) can be expanded into the following equation 3T _ 2 if T . §I_ 4_ 4 pcps %? - Ro(l+aT) + 736% ax - Poe(T To) - k0(l+BT)s %;%-- P(T-T0)h(T) - PmAH (41) where p and CP are the density and specific heat of the atomizer respectively, RO is the atomizer electrical resistance at room tem- perature, a is the temperature coefficient of resistance, a is the Stefan-Boltzmann constant, e is the total emissivity of the atomizer, To is the sheath gas temperature away from the atomizer, k0 is the atomizer thermal conductivity, 8 is the coefficient of thermal conductivity, x is the distance from the center of the atomizer, h is the combined convective heat transfer coefficient due to forced and natural con- vection, m is the amount of analyte deposited on the atomizer, AH is the amount of energy required to atomize the compound, f is the Thomson coefficient and may be of either sign. The heat due to the Thomson effect disappears if ac heating is used. Rearranging Equation (4l), we get 81 .2 _§_T__ P084 + ( Ro‘ll _Ph(T))T Eps’ mp5 PCP S 2 4 (1 R0 + Ph(T)TO - PM” + PoeTo _) + RPS pCPs pCPs pCPs if T , aT 3:T l+3T)7 (42) Ps ax pCoP( It should also be noted that parameters such as f and h are functions of temperature. An exact solution of the above equation is certainly difficult. Numerical integration was attempted for the platinum loop and graphite braid atomizers, but these calculations are not conclusive thus far due to a lack of experimental data over the entire range of temperature. For an approximate solution of the above equation, certain gross assumptions can be made. For a nonflame atomizer such as the graphite braid, which is discussed in Chapter VI, the temperature is uniform 31 and 3:; are practically zero over the entire length of the atomizer. However, a steep tempera- throughout the atomizer. The parameters ture gradient may exist at the contact points of the atomizer and its holder, and the heat lost by conduction and by the Thomson effect can- not be neglected. The same is also true for the carbon and graphite filament atomizers used normally for nonflame atomization. However, since the sample is deposited at the center of the atomizer, the infbrmation desired is the rate of temperature variation in the center and not at the atomizer contact points with its holder. The magnitude of the error introduced by this assumption depends on the length-to- diameter ratio of the atomizer. This ratio is about l0 for the graphite braid atomizer and about 100 for the platinum loop atomizer. 82 In order to neglect the conduction effect the 2/0 ratio should be greater than 20,000 for a temperature of about 300 °C (274). This indicates the grossness of the above assumption. For the following equation it is further assumed that the convective heat transfer and the Thomson coefficients do not change significantly with temperature. We can then write dt--A_ B. .37 - T4 + T + C (43) where -1 =: Poe A pCPs 2 3"]:iRoa- Ph pCPs pCPS 2 4 C'1 = 1 R0 + PhTO PmAH PoeTo pCPs pCPs - pCPs + pCps Integration of Equation (43) yields t=§-T‘3+B-TogT+c-T+D (44) where D is the integration constant. It can be seen that temperature is a complicated function of time. The contribution of each term de- pends on the magnitude of the constants in Equation 44. 2. Time Dependence of the Atom Population The rate of introduction of atoms by an electrically heated atom- izer follows the rate of temperature growth in the system. The first treatment of transient signals in nonflame atomizers was initiated by L'vov (9) who assumed that the rate of introduction of atoms into a 83 cell was a linear function of the evaporation temperature which can be approximated to be a linear function of time. L'vov, however, provided no justification for his assumption. The Justification can be seen from Equation 44 when either coefficients A and B or A and C are set to zero. Using this assumption, we can then write the equation for the balance between the atoms entering and leaving the cell as follows (9) 31:- = nut) - nzm (45) For a nonflame atomizer "1(t) = Et (46) where E is a proportionallity constant. Since the analyte atoms, No, in the sample are transferred into the cell in T] seconds, then Tl Jf n1(t)dt = NO (47) Therefore n (t) = 2N tr'z (48) l o l Since the analyte atom spends an average time, 12, in the cell, we have n (t) = -”- (49) 2 12 and ‘T = T ' 'T— ‘50) T] 2 D. 84 If Equation (50) is solved for N(t), we obtain 2 N0212 t N(t) = Tet— - 1 4' exp{-t/12} T1 2 for t.§ T] (513) and N 2 2 0 T2 T1 T1 N(t) = 73—472 4 + exp{- EDeXM-(t-Tfl/Iz} for t.: T] (51b) If it is assumed that the measurement device has an impulse response of unity, the ratio fil-is a function of 11/12. The smaller the 11/12 ratio, the larger th: N/No ratio. The peak height measure- ment can provide its best results if atomization occurs instantaneously. The sensitivity is decreased if T] is greater than zero. Integration of N(t) provides information which is independent of the atomization time as shown by Equation (52) ‘l’ l .. I n(t) + IN“) = N012 (52) O T] Integration of the signal thus provides a more linear working curve and better precision. IV. INSTRUMENTATION A. Introduction The instrumentation used in nonflame atomic absorption and atomic fluorescence spectrometry is similar to that used in conventional flame spectrometry except for the atomization and the detection systems. The instrumentation employed throughout this work varied from a com- pletely non-automated system to a system in which the operational functions were performed by a logic sequencing system and finally to a system in which the entire operation, data treatment, and optimiza- tion were performed under computer control with no operator interven- tion. This transition was dictated by the rigorous analysis of ex- perimental results during the various stages of this investigation and not by a simple desire to have an automated system. Since the experimental data obtained with the non-automated system have all been repeated with the computer-controlled system, only the data obtained with the latter are presented. Furthermore, no detailed descriptions of either the non-automated or the sequencer-controlled system is given here. The latter has been published (275), and the interested reader can refer to that article. However, the experimental results obtained with the non-automated system provided valuable information which helped us in the design of both the sequencer-controlled and the computer-controlled system and some of these conclusions will be summarized here. 85 86 First, in order to employ nonflame atomizers for routine analysis, the sample delivery step should be automated. In all nonflame atomizers used so far, this has not been considered, and samples are usually transported to the atomizer manually by means of a microsyringe. Since the sample volume required by many of the atomizers is ex- tremely small (l-lO pl), sampling imprecision can be a limiting factor in the overall measurement. Second, there are many interdependent parameters which can influence the free atom concentration in nonflame atomizers. For example, the temperature of the atomizer is a compli- cated function of the power applied to the atomizer, the sheath gas flow rate, and several other parameters. The sheath gas can similarly influence the residence time of the atomic vapor in front of the observation window and, in the case of atomic fluorescence, can also influence the quenching of the excited atoms. Thus, optimization of experimental parameters for high signal-to-noise ratios can be a critical factor. Although this can be performed by a non-automated system, it is time consuming and tedious, and the treatment of the experimental data is subject to Operator malfunction or at least bias. It is, therefore, essential to have a computer-controlled system. Third, because of the interdependence of the experimental parameters, it is desirable to simplify the atomization process either by decreasing the number of variables and/or decreasing the dependence of these parameters on each other. For instance, it was shown during our primary investigation, that it was practically impossible to prevent the simultaneous vaporization of solvent and solute from the atomizer when some of the gas sheath designs were employed to transport 87 the atomic vapor. Furthermore, if the sample is present in a compli- cated matrix, it is desirable to reduce matrix effects by removing some of the interfering species before the atomization step. This requires a multistep programming system for the electrical heating of nonflame atomizers. In this chapter, the design of the various components of the computer-controlled spectrometer is described. In the first section, a general description of the apparatus is given along with a descrip- tion of the atomization system and the optical system. The design and operation of a hardware and a software multistep programming system for electrical heating of nonflame atomizers are discussed in the second section. Two different automatic samplers are described in the third section. With one of the samplers it is possible to have computer control over the size and type of the sample. Finally, the various integration methods employed for the measurement of transient signals are described and the various functions of the computer- controlled spectrometer are discussed. B. Nonflame Atomic Absorption/Fluorescence Spectrometer 1. General Description Figure 5 shows a block diagram of the AF spectrometer. The system functions as an AA spectrometer if the positions of the sampler and the light source are interchanged. The AA/AF spectrometer is controlled by a minicomputer, which directs the sampler, the hardware temperature programmer or the current regulator, the optimization, the data processing and the readout. Under computer control, the 88 53850583 “:32 uwZoficooLouanEou of a8 .532“. xuoE .m 959.... «womooum mo\6 uaoomoqflomo \H'l'l lllllllllll // 2C3”: \ \\I|ll lllllllllll \ \\\.l.l||||||.l||.l|...l| \ x _ HUM" 3&8 hzmmmao _ musazoezi Iillv mm _ 332352: _ _ 5:592 5&3 mozzomzom amaze: 5:23 .3205 _ d #2.". -0202 m" m2<4nTzoz — I -+ \II II 75-3 533%. moto— _ mozaom _ maoomojamo \ t8... .IIIIIIIL, 89 sampler delivers a desired amount of a desired solution onto the atomizer. The computer then commands the hardware temperature pro- grammer or the current regulator to provide the atomizer with appro- priate currents and time delays for the desolvation, ashing, and atomization steps. The atomic vapor, which is transported by an in- ert gas, is excited by the light source, and the transmitted/emitted light is received by the monochromator and photomultiplier transducer. The resulting signal is amplified, digitized and sent to the computer. After the digitized signal is stored by the computer, the atomizer is shut-off, and the integrated signal is printed out on a teletype. The remaining operations depend on operator decisions and are des- cribed in a later section. 2. Atomization System a. Nonflame Atomizer Assembly. Figure 6 shows the atomization chamber with a wire loop atomizer located on top of an adjustable gas sheath. This arrangement was used for AF investigations. The whole system is fixed on a 9.5 x 5 in. plate and can be mounted in optical alignment with the monochromator in a matter of a few seconds. This type of arrangement was not subject to air drafts, and a change in the position of one component did not affect the other components. The wire loop holder was mounted on the plexiglass walls of the chamber. The two L-shaped holders were made from either brass or stainless steel and had two slots at the lower ends. The wire loop was inserted in the slots and was held in place by a spring clip arrangement as can be seen in Figure 6. The horizontal and vertical positions of the atomizer could be changed by means of a set of screws. 90 Figure 6. Schematic Diagram of the atomization chamber with a wire loop atomizer located on top of an adjustable gas sheath, 91 A lens holder was mounted on one of the plexiglass walls. The lens holder had an angle of 90° with respect to the monochromator entrance slit. The lens was used to form a cylinder of primary radia- tion at the top of the atomizer. No lens was used to focus the fluor- esced radiation on the monochromator. For AA work a similar lens holder was employed at an angle of 180° with respect to the monochromator entrance slit. The plexiglass walls were supported by a stainless steel base. One of the plexiglass walls can be removed to allow replacement of the atomizer. In a: later modification, the atomizer was held by a holder from below rather than from the top. This decreases light scattering by the atomizer holder. Light scattering was also reduced by introducing suitable light traps and painting all components with optical black paint. For multielement analysis, the plexiglass walls were removed from the assembly. b. Gas Sheath Designs. The atomizer was positioned over an adjustable inert gas sheath which transports the atomic vapor plume to the optical axis, minimizes oxide formation, and reduces quench- ing of the fluorescence radiation in AF. Numerous of gas sheaths were designed and constructed from stainless steel. One such sheath can be seen in Figures 6 and 19. The gas sheath could be supplied with an inert gas through suitable hose nipples. In all designs, the height of the sheath was variable over a two inch range starting from just below the entrance slit of the monochromator. The design of only two different gas sheaths are described here. The first design consisted of inner and outer clusters of holes drilled on a circular piece of stainless steel. The distance between 92 the inner and outer cluster was 0.32 in. The holes in the inner cluster had a diameter of 0.018 in. each and were located on four rings of 0.04, 0.08, 0.12, and 0.16 in. radii. The central ring had nine holes, but the rest of the rings had seventeen holes each. The outer cluster had four rings of holes located on 0.62, 0.66, 0.7 and 0.74 in. radii. The holes, which were staggered to improve the sheathing efficiency, had diameters of 0.032 in. each. The sheathing gas was supplied through the inner and outer hose nipples. The second design consisted of a square piece of 2 x 2 in. stain- less steel with inner, middle and outer clusters of holes. The dis- tances between the inner-middle and middle-outer cluster were 0.125 and 0.25 in., respectively. The middle and outer cluster each con- sisted of three square matrices of holes. The holes in these two clusters had diameters of 0.032 in. and were 0.048 in. apart. The inner cluster consisted of a square matrix of holes with five parallel rows of holes, which had diameters of 0.018 in. and were 0.032 in. apart. In contrast to the first design, the holes were not staggered in the second design. c. The Hire LoopgAtomizers. A variety of wire loop atomizers. such as wires from gold and tungsten, wires from alloys of nickel and chromium, and wires from alloys of platinum were investigated as atomization devices. In the nickel-chromium alloys, the Cr content makes the alloys resistant under oxidizing conditions, and the Ni content enables them to retain resistance under reducing conditions and in strongly alkaline solutions. The best alloy from an electrical standpoint was the 80% Ni - 20% Cr alloy. However, because of its chemical reactions with various analytes, this atomizer was discarded. 93 Although gold wire was very resistant to chemical reactions and to oxidation, the melting point of this element is about 1060 °C, and gold cannot be used for the atomization of high boiling point elements. Platinum has a melting point of about 1770 °C and is only attacked by aqua regia and somewhat slowly by hydrochloric acid and other oxidizing agents. Alloying with iridium and ruthenium reduces the rate of corrosion. Iridium and rhodium have a boiling point of about 2400 and l970 °C respectively, but they undergo oxidation to a greater extent than platinum, and they are moderately hard. In general, Pt-Rh alloys were found most useful, and an alloy of 90% Pt and 10% Rh was chosen. This particular alloy was selected because of its stiffness and overall strength, higher operating temperatures than pure platinum and low susceptibility to oxidation. Although fatigue is commonly exhibited in pure metals under repeated heating and cooling, the Pt-Rh alloy exhibited this effect to a lesser extent than pure platinum. For high atomization temperatures, tungsten wire loops were also used. The melting point of this element is about 3390 °C. but it undergoes oxidation. Thus, the metal wire loop atomizers used throughout this investi- gation were made of Pt and N metals, and 90% Pt - l0% Rh. The diameters of wires used ranged from 0 032 to 0.008 in. for Pt and its alloys and from 0.01 to 0.02 in. for tungsten. The physiochemical pr0perties of these atomizers are discussed in Chapter V. 3. Optical System The optical system used was similar to that used in conventional AA/AF spectrometers. The radiation from a metal vapor discharge lamp 94 (George N. Gates & Co., Inc., Franklin Square, NY) or a hollow cathode lamp (Fisher Scientific, Naltham, MA)was focused above, and slightly beyond the loop atomizer by a 2 inch focal length quartz lens (f/l.3), which was only employed for AF work. The rest of the spectrometer consisted of a grating monochromator (EU-700, f/6.8, Heath Co., Benton Harbor, MI), a photomultiplier module (EU-70l-30, Heath Co.) a current-to-voltage converter (Model 427, Keithley Instruments, Inc., Cleveland, ON), a digital voltmeter (EU-805, Heath Co.), an oscilloscope (Type 564, Tektronix Inc., Portland, Oregon), a recorder (EU-ZOS-ll, Heath Co.), and a PDP Lab 8/e mini- computer (Digital Equipment Corp., Maynard, MA). C. Multistep Current Programming Systems for Electrical Heating of Nonflame Atomizers 1. Introduction Nonflame atomizers for atomic absorption (AA) and atomic fluores- cence (AF) spectrometry have been shown to be of considerable value for the detection and analysis of trace amounts of metals in a variety of matrices. In order to reduce matrix effects and to allow some con- trol over the atomization process, most nonflame atomizers utilize a two or three stage electrical current program for heating the atomiza- tion element. In the first step, a small current is applied to the atomizer to vaporize the solvent. If organic material is present, a second larger current step is applied to ash the material, and in the final step, a high current (often > 100 A) is applied to atomize the sample. 95 For most efficient use of such a current programming system for nonflame atomizers, it is desirable to have precise control over the current applied during each stage, the length of time for which each current step is applied, and the times for various other operations such as data acquisition and sample delivery. In this section two different programming methods are described for multistage electrical heating of nonflame atomizers, such as carbon filaments, tantalum strips and hot-wire loops. The operation of the first system is con- trolled by a logic system, while the second programmer is controlled entirely by a minicomputer. 2. Hardware Programming System for Electrical Heatingof Nonflame Atomizers A digital programming system is described for multistage elec- trical heating of nonflame atomizers. The integral parts of the programmer consist of a logic system, a serial D-to-A converter, and a current-regulated power supply. The design and characteristics of the system are presented and discussed. a. Principles of Operation. A block diagram of the entire multi- step current programmer is shown in Figure 7. Before the experiment begins, the operator presets the digital sequencer by entering binary numbers from switch registers in order to select the following: 1. An initial time delay for the sample delivery to the atomizer. 2. The desolvation time and current. 3. The ashing time and current if an ashing step is necessary. 4. The atomization time and current. The sequence begins when the sequencer receives a logic 1+0 transition 96 .emsseemoea pcmeezu amumwppze we» do segmepu xuopm .u weaned A0200 (ID Eth>m 02.02w30wm ._<._._0_D Nb. 1‘ «e ph. _ gem ZO_._.Ow._.mO A< um u< 305325: L 5&5... GUN-gb‘ IIII “mmsSRUQ .hflfivl ~mw>>hvm >._n_QDw 092.5me thmmDO 97 either from a manual switch or from an automatic sampler. This start pulse opens a gate and allows clock pulses to enter preset binary counters, which are the heart of the programmer. After the initial time delay, a count-serial digital signal (a number of standard sized pulses) is sent to a countrserial digital-to-analog converter (SDAC) where an analog output voltage is produced corresponding to the desired desolvation current. The SDAC output is applied to a current-regulated programmable DC power supply which controls the current in the atomizer. When the preselected desolvation time is over, the ashing current is applied, if an ashing step is desired. When the ashing step is complete, the atomization current is applied for a preselected time. During the atomization step, the sequencer signals the data acquisition system to start the data collection process, and the AA or AF data are then acquired. In the final step the current through the atomizer is shut off, and the entire logic system is cleared and prepared for the next start pulse. Details of each of the functional blocks shown in Figure 7 are given below. b. Sequencing,System. The operating cycle of the programmer is under the complete control of the sequencing system which provides the appropriate time delays and pulses for the different steps. Since a binary counter stores the maximum possible amount of information, the system was constructed from 4-bit binary counters. The logic system is described for a two step program of desolvation and atomiza- tion.: An ashing step can be implemented by the addition of a third similar circuit. All components were mounted on circuit cards and inserted into an Analog-Digital Designer (Model EU-801A, Heath 00., Benton Harbor, MI), which supplied the necessary power to operate the .f’ ' r’ rvv-j V 98 sequencing system. Figure 8 shows a general block diagram of the entire sequencing system along with its connections to the SDAC. The initial time delay (ITD), the desolvation time unit (DTU), the desolvation current unit (DCU), the atomization current unit (ACU), and the atomization time unit (ATU) are preset independently by means of parallel switch registers. Figure 8 also shows that two different clock rates are utilized for the time controlling and current controlling units. The clock rates were“ normally 0.l-1 Hz and 1 kHz for the time and current controlling units, respectively. The different clock rates were obtained from a 1 MHz crystal oscillator and frequency dividers.which are not shown in the figure. Figure 9 shows the circuit diagram of the initial time delay (ITD) step. The four Exclusive-NOR gates and NAND gate 1 form a digital comparator. The comparator output, known as the final count detector (F.C.D. O), and the reset signal operate the clock gate. The gated clock output is sent to the binary counter input. Signals A], B], C], and D] are switch outputs which can provide logic 0 or l levels to the Exclusive-NOR gate inputs. These switches are used to preset the desired number of counts. After the desired number of counts is selected, the rest of the operation is as follows. A logic 1 at "reset" results in clearing all flip-flop outputs to zero, while a logic 0 allows the clock pulses to enter the counter. The comparator matches the desired count with the counter output and undergoes a l+0 transition as soon as the balance is reached. This transition prevents NAND gate 2 (gated clock) from passing more clock pulses and starts the operation of the 99 .2333.» 533.0. 23 so: 532285 «5 2 E use 335. goo—u wee Po use UV Emumxm aerosozamm Lossecmoen ucmcesu on» we Euemmvv xuorn Poconos .w meamwm _n_ 0 _o_emee 8:8 co seemeee useeeeu .oP meagre © IFIL _ e _ E E. k. . mm... .6. 3!. ml... 4m. ...m. ....- z «flee. E" «a... .sz. meO xUOaU xUOdU 0wh<0 N .00.“. DU< O... .msmamxm eeempu new «mmwee use A2F 4-10000-r CLOCK RATE. Hz Figure 12. The effects of clock rate on the magnitude of the SDAC output and its reproducibility. 108 open at any time during any one pulse. Thus, even though the pulse is counted, the SDAC does not "see" the entire pulse, which results in a small loss in the output. The same is also true about the VOFTOUS‘ time delays in the system. In order to decrease the relative error due to this effect, one pulse should contribute a very small amount to the value Of any parameter (time, current). Therefore, a large number of counts should be accumulated during each current and time step. With the present system sufficient pulses were counted to achieve errors Of less than 1%. In order to reduce the error even further, higher resolution counters and comparators are required. However, since the SDAC is a serial converter, the conversion time will increase and this may result in a gradual evaporation of the sample in the atomization process. This problem can be solved if a parallel DAC replaces the serial DAC. For our purposes, the effect Of this error was very small and could be neglected. Although the above programmer has only desolvation and atomiza- tion steps, the system can be expanded to have an ashing step. The electronic circuit is similar to the other steps and is not discussed here. The analytical results Obtained with this programmer compare favorably with results Obtained with most nonflame AA and AF techniques and are discussed in Chapter V and VI. 3. Computer-Controlled Heatingyof Nonflame Atomizers The hardware programming system described previously has a number Of disadvantages which can be best taken care of through software control over the electrical parameters which govern the atomizer temperature. First, iterative Optimization Of parameters such as desolvation, ashing, and atomization time or current, if not practically impossible, is extremely tedious. This prevents a fundamental investi- gation Of the chemical problems in a routine manner. Second, multi- element analysis and samples of complicated matrices require additional heating steps which necessitates the expansion Of the sequencing system. Third, the logic system is subject to various transients and should be protected by suitable isolation devices. The computer-controlled heating system employs the same current regulator described in the previous section. The serial DAC and the sequencing system are replaced by a parallel DAC and a minicomputer, respectively. The minicomputer was a PDP Lab 8/e system (Digital Equipment Corp., Maynard, MA). For all the interfacing designs, a minicomputer interface system (EU-801E, Heath Co., Benton Harbor, MI) with appropriate circuit cards was used. Figure 13 shows the inter- face which was designed for the computer-controlled system. The parameters which control the temperature Of the atomizer and the various time delays are stored in the computer through an initial dialog between the Operator and the teletype. The stored digital information relating to the amount of current for each heating step, is transferred to the computer interface buffer (CIB) under computer control. The Data Bus card directs the digital signal to a 10-bit DAC card, which produces a corresponding analog voltage between D to -10 V when it is properly addressed. The DAC output is sent through a unity gain inverter to provide appropriate polarity for the current regulator circuit. This output voltage is compared at the 0A input terminal (0A2) with the voltage drop across a current sensing resistor in series with the atomizer. 110 .mgm~reoae msepmcoc mo we'vem; Feueguuopm so; «acetone? couzasou .mp oezmwm A1 @ @ G a) 0 b 2.5 55m>20u in © M o 1“ 2:6 m2. <53 @ 4 11le rue 9:8 $5083 250 no “If n. Tm QILIIMJTIIO 0.1 N I a _. 0 : 53mm e 3.30 we 3:. .n..l.. 39:» .i 29 e O) & 9 b Of a T 1.. Oi b 11 J8 oemtmorll rill \ .7 a. :23 3230: 5:3 “3:32. thamau cmhaus‘OU am»:m<<0U.Z_<< 111 In order to be addressed unambiguously, the DAC also needs a Strobe and a Device Select signal from the computer. The Data Bus card provides these different Input/Output signals (IOPl, 2, and 4), and IOPl was used as the Strobe signal. The Octal Decoder card was also employed to provide the Device Select Code (11). In operation the system functions as follows. Under program con- trol a small current from the programmable current supply is applied to the atomizer to desolvate the sample fOr a preselected length of time. After the desolvation step is complete, a larger current is applied to ash the sample if the operator has indicated in the initial dialog that an ashing step is necessary. This is done by commanding the DAC to provide a larger voltage at the 0A2 input terminals. Again, the ashing current and ashing time are under computer control. Finally, a third high-current step is applied to atomize the sample. During the atomization time, the computer enters a data acquisition routine to Obtain quantitative absorption or fluorescence information. The data acquisition system is described in a later section. 0. Automatic Samplers for Nonflame Atomizers 1. Introduction In many types Of chemical experiments, the "hang up" to complete automation comes in the steps of sample preparation and sample handling which occur prior to the instrumental measurement. The automation of these steps along with the more conventional steps of data acquisition, data processing and real-time control over instrumental variables would then allow computer feedback in a "closedeloop" type system. 112 The effects of the sample delivery process are more pronounced in nonflame spectrometry compared to most other chemical experimentation. This is due to the fact that the required sample volumes range between 1 - 10 ul, and it is essential tO deliver reproducible volumes Of sample solution to the same location on the nonflame atomizer. In those filament-type atomizers, such as a wire loop or West-type carbon fila- ment, where the heat is confined to the center Of the atomizer and there is a larger temperature gradient throughout the atomizer, the sample must be delivered to exactly the same point every time to provide reproducible free atom populations. The first automatic sampler for a wire loop nonflame atomizer was reported by S. R. GOOde (276) in these laboratories. The sampling system consisted Of a pump, similar to a peristaltic pump, and a de- livery capillary. A motor-driven "wiper" compresses tygon tubing lying ‘between an adjustable stage and the wiper, and this compression forces a certain amount of sample through the tubing. The sample is delivered to the wire loop by a quartz capillary tube, which is positioned directly below the wire loop. The droplet attaches itself to the loop in preference to the quartz. This type of automatic sampler has a certain number of disadvantages, however. First, the sampler is only suitable to the loop type atomizer and it is not possible to employ the sampler with furnace and filament atomizers. Second, since the quartz capillary tube is close to atomizer at all times, it might clog or break because Of solvent evaporation and high temperatures. The solvent evaporation introduces a systematic error in the overall measurement. Third, the presence of the quartz capillary tube disturbs the inert gas flow in the atomization chamber. 113 Fourth, the solution level in the solution container should be closely monitored to prevent erroneous results. Fifth, the pump-capillary system suffers from the siphoning effect. Sixth, there is sample adsorption onto the tygon tubing and this makes trace analysis im- possible. Finally, with this type Of arrangement, computer control over the size and type of the sample cannot be exercised. In sample-limited situations such as in the analysis of trace elements in biological fluids it is desirable to use only a few micro- liters Of the sample. This is not possible if the sampler has a large dead volume. Furthermore, although it is possible to employ this sampler as a component of an automated spectrometer for fundamental chemical studies, its capabilities are practically and fundamentally limited. In order to eliminate all Of the above drawbacks, the author designed and evaluated a variety of sampling systems. In this section the designs, characteristics, and the computer interfaces for two sampling systems are discussed. In both cases the sampler is not pres- ent in the atomizer environment during the atomization process, and sample delivery is made from above the atomizer. In one design, it is possible to have software control over six different sample types with a software-selected volume ranging between 0.05 p1 to 200 pl. The operation of each system is discussed separately. 2. Automatic Sampler - Desigpyl With this design, the desired amount Of sample is selected by the operator. The amount of sample can be continuously varied from O to 5.6 pl. The principles of operation, sequencing system and 114 analytical results are reported in this section. a. Principles of Operation and Design Consideration. The sampler consisted basically of two parts; a pump and a moveable syringe. A schematic diagram of the pump is shown in Figure 14. Three spring loaded stainless steel pistons (0.5 in. diameter) sequentially press a piece Of tygon tubing. The movement of the pistons is governed by a cam attached to a small motor. The amount of delivered solution, which is continuously variable and covers an overall range from O to 5.6 pl, can be set by a circular plate (c) attached to a set screw (a). The amount Of solution delivered on each Operation cycle depends on the pressure applied on the tubing by the set screw (a). A locking device (a') is provided to maintain each setting during operation. The Operational cycle of the sampler as a function of the position Of the cam and piston is shown in Figure 15. At the start Of the Opera— tion piston 3 is closed and pistons l and 2 are Open. This situation prevents the flow Of solution. As the motor starts to rotate the cam, piston 2 and then piston 1 press the tygon tubing, and the solution is pushed out. After a 120° rotation of the cam, piston 3 and then piston 2 open, and the solution comes through. Finally, the pistons return to their initial positions. The second part Of the automatic sampler consisted of a moveable handle connected to a miniature air cylinder. The piston of the cylinder was actuated by a 4-way, solenoid-operated valve. The other side Of the moveable handle holds a 5 in. long stainless steel tube, which was sharpened on one end. The tube is electrically insulated from the rest of the system. The tygon tubing was clamped to the handle and attached tO the stainless steel syringe. The X, Y, Z .H cmvmmu LquEem c? um~vpvaa muw>mu mcpqsaa covuapom mo Enemewu Owuusocum .ep meamwm Z. ZQhBOm u j 115 L. 50 20:38 «05.2 .. -. . My -q—u-q-d-fi-a 5:3. T“ C CM , p . f a s . 5 me a: 116 0' 60° 120° 180° 240° 300° 360° PISTON No.1 8. 21 ARE 240° DOWN a 120°ur PISTON No.2 180° UP a. 180° DOWN Figure 15. Operational cycle of the sampler design I as a function Of position of the cam and pistons. 117 positions of the syringe could be changed by three screws on the moveable handle. In Operation the delivery syringe is initially in a position out of the Optical path Of the spectrometer. On command Of the computer or operator the solenoid - actuated air piston moves the syringe into position over the atomizer, a preselected amount Of solution is delivered to the atomizer by the pump, and the syringe is moved back out of the light path by the release of air pressure. b. Sequencing System. The circuit diagram of the sequencing system is shown in Figure 16. The entire Operating cycle Of the sampler was controlled by a simple electronic circuit which consisted Of three relay switches R51, R52 and RS3 and two microswitches M51, and M52. The MSl switch was mounted below the cam, and its moveable contact was normally held closed by the pump cam. The sampler starts its cycle when the manual switch is pushed momentarily by an operator or when a +5V dc signal is applied to the R53 coil by another instrument. This closes switch RSl, actuates the solenoid and moves the syringe to the atomization chamber. As soon as the syringe "sits" on the tOp of the atomizer, the moveable contact of switch M52 is pressed by the moveable handle, switch M52 closes, and the motor is activated. When the motor turns, the cam rotates and releases the M51 contact, which results in closing switch R52. Because of the new contacts, the solenoid remains "actuated", and the motor continues its operation to deliver the sample. When the sample is left on the atomizer, the cam presses the contact of M51 again, which deactivates switch R52 and the solenoid. The solenoid valve is then opened to release the air pressure and, as a result, the syringe leaves the atomization chamber. .503? 9.3533 H $33 .3353 .3 523:8 3:25 .8. menace Cb...— n. :923m >445“. 8 mm Cha— mv rub—3m Duo-2 a 92 118 10.526 mom Du . meaeqs. _a 1 .8 23-.., . coEmoa Eco E3 5 go ago 956 3.5a *0 Dome—O an >0 ; . oz 1 coco 33.9 enxxai “:eT. ego: mm: 02.. 02 .m: /. ’lululalllllnlllnl 02 .mm .:o i O o.‘ 10.2261— Xx 44:24.21; _ 10:26 nmm. mmgom 1:00. >0 +I_L1 , 119 The system has the capability Of purging itself. This is done when the purge switch is left at the "2" position. It can be seen from the circuit diagram, that the solenoid is deactivated in this case, and the sampler runs continuously to remove the solution from the tygon tubing. The time required for cleaning the sampler depends on a variety Of parameters such as the size Of tubing, sample type and concentration. This sample delivery system has the drawback of possible adsorption onto the tygon tubing. This problem was eliminated by the automatic sampler described in the next section. c. Comppter Interface. The function of the computer interface was only to command the sampler to move to the atomization chamber. The remaining operations were performed by the sampler itself. In order to perform this function, the computer was programmed to provide a +5V signal for an Operator selected length of time. This can be done by a variety Of software and hardware combinations, and one pos- sible method is as follows. The computer was programmed to provide a variable time delay between two input-output instructions, 6411 and 6421. The device select signal 41 was connected to the Latch card input, and the corresponding Latch card output was sent to the sampler. In order to transfer the signal from the computer to the sampler, the Latch card should be addressed with a minimum Of two gating signals. Device select signals 41 and 42 were taken from the Octal Decoder card and connected to an OR gate. The OR gate output and IOPl were sent to the Select and Strobe inputs of the Latch card, respectively. With the first instruction, 6411, the Latch card is gated and transfers a +5V signal to the sampler. After a preselected time delay, the computer performs the 6421 instruction, which results in the 120 elimination of the +5V signal. d. Precision and Accuracy. The sampling system delivers samples with a precision of 2 to 3% relative standard deviation. The precision was found to be a function of the sample volume delivered. The pre- cision increases initially with sample volume and passes through a maximum for a sample volume of l to 2 p1. The accuracy of the sampler was determined for only a sample volume of 2 p1 and was found tO be :3 percent. The precision and accuracy were determined by weighing droplets and also by measuring the absorbances of solutions prOduced when the sampling system injected a dye into a spectrophotometer cell. Even though the precision in dispensing a sample was better for a manual syringe (l to 2% relative standard deviation) the overall precision of absorption and fluorescence measurements was increased with the automatic sampling system because of the excellent repro- ducibility in positioning samples on the atomizer. 3. Automatic Sampler - Design 11 a. Principle of Operation. The sampler described previously could be used for optimization studies involving a fixed volume of sample. For studies which involved variation Of the sample size or automatic selection of the sample, the sample handling system pictured in Figures 17 through 19 was designed. This sampling system consists of a sample turntable, a stepping motor driven micrometer syringe, and a pneumatic system for moving the syringe. The moveable arm to which the syringe is attached has 3 positions. In the position shown in Figure 17, the syringe is tilted backward so that it is out Of the Optical path. Figure 18 shows the syringe in the second position, 122 where it is moved to the sample turntable for sample pickup. In the third position shown in Figure 19, the syringe is positioned directly over the atomization element for sample delivery. All Operations of the sampler are controlled by means of five small motors and two miniature air cylinders. The sample turntable employs one motor, and the micrometer syringe is driven by a stepping motor. The remaining three motors perform three basic Operations called the Purge, the Fill, and the Deposit functions. The miniature air cylinders are responsible for transporting the moveable arm to the atomization chamber or the sample turntable. A brief descrip- tion Of the various functions and the computer interfaces is given here. b. Sample Turntable. The sample turntable has provisions for six sample cups and suitable height control adjustment with respect to the micrometer syringe. The turntable is moved by a small motor, and any of the six sample cups can be moved into position under the syringe for sample pickup. The sample handling system has provision for flushing. One cup is normally kept filled with deionized water and a second one, the waste cup, is kept empty for the sampler to dump its content if so commanded. The movement of the turntable can be controlled by the computer or a sample selector switch mounted on the front panel. c. Stepping Motor Driven Micrometer Syripge. The basic component of this system consisted of a stepping motor, a micrometer syringe and a stepping motor driver circuit. The stepping motor was con- nected to the micrometer syringe by means of flexible mechanical link- ages, and the combination was attached to the moveable arm. The X, Y 123 and Z positions Of the micrometer syringe - stepping motor combination could be changed by means of a set of screws. Two small microswitches were mounted on this combination to indicate when the micrometer syringe had reached its upper and lower limits. As soon as the syringe is filled with a solution the upper microswitch Opens and stops the stepping motor. The same function is performed at the lower limit when the syringe is empty. Since in most of our investigations the computer keeps track of the amount Of solution in the syringe, these micro- switches were not employed as limit indicators. They are utilized, however, if the computer is replaced by a hardware sequencer. The stepping motor (Model 20-2215 0 200-Fl.5, Sigma Instruments, Inc., Braintree, Mass.) was an inductor-type motor with 200 steps per revolution. The angular accuracy reported by the manufacturer was :3 percent and was non-cumulative. The accuracy of the 200 p1 micro- meter syringe (Catalog NO. 7840, Cole-Farmer, Chicago, IL) was 0.5 percent. The luer joint of the syringe allowed the attachment of a hypodermic needle. Each division in the sleeve corresponds to 0.2 pl, and each revolution of the micrometer head dispenses 10 pl of the sample. Thus each step Of the stepping motor delivers 0.05 pl of sample solution. Delivery is accomplished at a maximum rate Of 10 pl/sec. The stepping motor driver circuit (Type A202, Sigma Instrument), rated at 28V dc and 2A, was used to convert pulses at either forward or reverse input terminals into four current patterns for proper bi- directional rotation Of the stepping motor. The required pulse was a negative six volt transition with a minimum pulse width Of 20 psec. By decoding in this manner, it was possible to use only two bits Of information to drive the stepping motor. Setting up the proper current 124 sequence within the computer was discarded on the grounds that this required four bits Of information, more programming, and was more expensive. The stepping motor driver was powered from a 28V dc, 8A power supply. For the maximum possible speed, the stepping motor was con- nected to the driver card by the arrangement shown in Figure 20. The resistors are required because the nominal motor voltage based on the rated coil resistance and current, is less than the power supply voltage. A large difference between these voltages results in a high stepping rate. For an inductor-type stepping motor, the rate Of rise Of cur- rent is determined by the motor winding time constant L/R, where L is the coil inductance and R is the total circuit resistance. It is, therefore, desirable to have low values for the ratio, L/R. The forward or reverse pulses for the drive circuit were Obtained either from a hardware sequencing circuit or from the computer. The detailed description of the hardware system design will not be given here. It basically consisted Of a counter for sample size selection. The same counter in conjunction with the two microswitch limit indi- cators could also be utilized for filling or emptying the syringe. The description of the computer interface will be given later in this section. d. The Purge Function. In this function, the moveable arm to which the micrometer syringe and the stepping motor is attached, is Carried to the sample turntable by means Of two miniature air cylinders. The pistons of the air cylinders were actuated by two 4-way single solenoid-Operated valves. The first cylinder transports the arm from its initial position towards the sample turntable. This process 125 .uwauewu em>ecn me» new gopoe mcvaamum we» we cowuumccou exp we Between xuopm .oN weaned H No 94.0) n 5.2. m u».:3\30...:> 33:» 3.5.. a male 33>: .L «~“a . v +. $23 2.1338253 : 323.0 m «Opci 2:2. 3. 02:85 35.. o 01422.0“. ixianqgfiz, m 300. C2 3.13:3: 2:. mm — o— 1? <+ae U0>mu+ 126 changes the position of the arm in only the horizontal direction. The second air cylinder then transports the arm in the vertical direc- tion and positions it on top Of the sample turntable with the tip Of the syringe in the waste sample cup. A pulse train is then delivered at the forward input terminal Of the stepping motor system which causes the syringe to dump its contents into the waste cup. The air pressure on the second air cylinder is then released, and the syringe leaves the sample turntable. The sample turntable is automatically actuated and moves to the deionized water cup. The syringe then moves to the sample turntable, is filled with deionized water, and leaves the sample turntable. The sample turntable moves to the waste cup position, and the syringe moves to the sample turntable and dumps its contents into the waste cup. The syringe leaves the sample turntable by releasing the air pres- sure on the second piston and finally moves to its initial position as soon as the air pressure on the first piston is also released. The computer responsibility for this operation is to provide an initial pulse to actuate the Purge function as well as to control the stepping motor driven micrometer syringe. The rest Of the Operations, such as the transport of the moveable arm to the sample turntable and the selection Of deionized and waste cups are performed automatically by a small motor and a number Of cams attached to the motor shaft. Each cam is in a form of circular plate with the motor shaft passing through the center of the plate. Thus all cams are parallel, and the movement of the motor shaft rotates all the cams simultaneously. The circular movement of these cams results in the opening or closing Of a number Of microswitches, which are in contact with the circumferences 127 of the circular plates. This provides a programming system for the various Operations of the Purge function. The cycle time for the opera- tion of this function is shown in Figure 21. e. Fill Function. The Operation of this function is similar to the Purge function described in theprevious section. The cycle time Of the Fill function is shown in Figure 22. The sample type is either selected by means Of a computer command signal or manually by the sample selector switch mounted on the front panel. The syringe moves to the sample turntable, and a pulse train is then applied tO the reverse input terminal of the stepping motor system by the computer, which causes the syringe to be filled with the desired solution. The moveable arm then leaves the sample turntable and returns to its initial position. The computer has the responsibility for providing an initial pulse to actuate the Fill function, a second pulse to supply the proper sample cup for the syringe, and a pulse train for the stepping motor driven micrometer syringe. The remaining operations are performed by a motor-cam system similar to that described in the previous section. f. Deposit Function. In this function, the moveable arm trans- ports the stepping motor driven micrometer syringe to the atomizer by means Of only one air cylinder. After the syringe is positioned at the top Of the atomizer, a software-selected amount Of sample is delivered to the atomizer, and the syringe returns to its initial position. The computer is only responsible for sample size selection and delivery and for providing an initial pulse to activate the Deposit function. The rest Of the operations are performed by a motor-cam mechanism similar to that described for the Purge function. The 128 1 2 —FL 3 ———f '— 4 r—L 5 J l a l ' 7 I 1__ 8 "_L- l T .T 1 l I 1 l I 1 O 20 40 60 80 100 120 140 160 180 TIME,sec. l PURGE ACTIVATION SIGNAL. 2 SYRINGE MOVES TO SAMPLE TURNTABLE, FROM ITS INITIAL POSITION. 3 SYRINGE DUMPS OUT SOLUTION, INTO THE WASTE CUP. 4 SYRINGE LEAVES THE SAMPLE TURNTABLE. 5 SYRINGE RETURNS TO SAMPLE TURNTABLE, AND IS FILLED WITH THE SOLVENT. 6 SYRINGE LEAVES THE SAMPLE TURNTABLE. 7 SYRINGE DUMPS OUT SOLVENT INTO THE WASTE CUP. 8 SYRINGE RETURN TO ITS INITIAL POSITION. Figure 21. Cycle time of the Purge function in automatic sampler design 11. 129 ,r—L 2_J-—_I 3 J j 4 4 L - 1 I 1 1 1 1 1 O 10 2O 30 4O 50 60 7O TlME,sec. 1 FILL ACTIVATION SIGNAL 2 SYRINGE MOVES TO SAMPLE TURNTABLE FROM ITS INITIAL POSITION. 3 SYRINGE IS FILLED WITH THE DESIRED SOLUTION 4 SYRINGE RETURNS TO ITS INITIAL POSITION. Figure 22. Cycle time Of the Fill function in automatic sampler design II. 130 cycle time of the Deposit function is shown in Figure 23. At the conclusion of the Purge, Fill, and Deposit functions, the sampler can provide a pulse to indicate that its Operation has con- cluded. These pulses are only utilized when a hardware sequencer controls the automated spectrometer. When a computer-controlled system is used, the proper time delay patterns are generated by the computer, and the above signals are not employed. Figures 21 and 23 show that the duration Of the Purge, Fill, and Deposit functions are 185, 65, and 35 seconds, respectively. The time durations of these functions can be decreased by a factor Of three if the stepping motor provides 50 steps per revolution instead of 200 steps, if the three motors employed for the above functions are three times faster and have three times the torque of the present motors. It should also be noted that in order to Operate the sampler independent of a computer, front panel switches were available. 9. Computer Interface. In order to Operate the sampler, the sample turntable, the Purge, the Fill and the Deposit functions, llOV ac were required. Since there were six sample cups and three functions, the computer should switch ON and OFF 110 V ac for nine different devices. A combination Of optoelectronic switching devices and triacs was employed in order to isolate the logic system from the line voltage. The circuit diagram for remote activation Of one device is shown in Figure 24. This circuit is repeated nine times for the above nine devices. All circuits, however, are powered by the same 9 V dc power supply. I In order to control the atomizer and the various components Of the sampler, a minimum of 12 different devices should be under computer 131 2—I'_'l 3__I 7— 4 I—'L_ F T I 1 l 0 10 20 30 40 TIME,sec. 1 DEPOSIT ACTIVATION SIGNAL. 2 SYRINGE IS POSITIONED ON THE TOP OF ATOMIZER. 3 SYRINGE DEPOSITS THE SAMPLE. 4 SYRINGE RETURNS TO INITIAL POSITION. Figure 23. Cycle time of the Deposit function in automatic sampler design II. 132 . .mmmppo> mc*_ on» new mmumwgmpcp empaasou on» Co cowumpomw on» 20$ Eeemmwv upzucwo .em mg:m_m m.p<< nnnnZn ”U452AVP_ATIIIIIL kudoe. 3+.ch 3. 3. >ama3m F my ~:u>>nvm mw U0>o 133 Logos mcvaqoum use meauagmaeou LONrsope mo Poppeou we» com ouaweoucw Luuaaeou .mm mega?“ .ucmeo>os omco>og new uguzgow — OOH _ _ _ N mhm¢ «OFOE 021mmhm cmZOO m3 (\o (.55 33>: «052 02:85 O M m :e M 1 a be - - N :e I 235.0“. «052 02:85 O a .1; u .r n E. w wO OOs [11: 3:1 .5“. w Of 4 Ilmfllrll.) a I” : a o¢>¢Om umN_¢wn 134 mOaam O... :mOmmo O» :E O... u-«nvno .OZ .5 U m4m<<mo 135 control. A variety of software and hardware combinations was employed to perform these functions. One such interface is shown in Figures 25 and 26. The computer provides +5V logic level signals to the components through the computer interface. The Device Select Signals in both figures are provided by the same Octal Decoder card. Figure 25 shows that the same DAC is employed for the Operation of the atomizer and for supplying pulse trains to the forward and reverse input terminals of the stepping motor driver circuit. The DAC output is, however, 'gated by means of the Latch card and FET switches tO prevent simul- taneous activation Of different devices. Figure 26 shows the inter- face for the sampler. It should be noted that the same Data Latch card is used in both circuits shown in Figures 25 and 26. The +5V output signals from this card are connected to the Optoelectronic switching circuit shown in Figure 24. The computer does not provide a signal for the Fill function. However, since the sample turntable should be activated before the filling process, the same signals which command the sample cups are employed to control the F111 function simultaneously. E. Integration Methods Early in this work a system for the integration of the peak- shaped photocurrent signals was found necessary. Analytical curves constructed using peak height data were nonlinear at high concentra- tion. Both fixed and variable time analog and digital integration have been employed throughout this work. Analog integration generally suffered from long term drifts, and almost all data have been Obtained with digital integration. 136 1. Variable-Time AnalogyIntegration Figure 27 shows the circuit diagram for analog integration. The circuit consisted Of an OA (Model PP25 AU, Philbrick/Nexus Research A Teledyne Co., Dedham, Mass.), connected as an integrator, a comparator (Comparator/V-F card, EU-BOO-HB, Heath CO.) and a digital voltmeter. A buffer was introduced between the i-v converter and comparator to prevent loading Of the signal by the comparator. The photomultiplier output was converted to a proportional voltage by a current amplifier whose transfer function could be varied from 103 to 1011 V/A. The resulting signal was connected to both the comparator and the integrator input. As soon as the voltage crosses a preselected level, the comparator output changes its state and closes the switch 51. Integration is then performed if switch 52 is Open. In fact, switch 51 provides a variable time window which is the function of signal. The integrated signal is then displayed on the digital voltmeter. The integrator output voltage eo is related to the input voltage ein by t -eo =lef ein dt where t is the integration time and R and C are the values Of the resistor and capacitor in Figure 27. The integrator is reset by closing switch 52, which discharges the capacitor. 2. Fixed-Time Digital Integration In this method a computer functions as a digital integrator and the readout system. During the atomization period, the analog signal is continuously digitized at regular intervals by an analog-to-digital 137 mmhmiha0> ion—.05 .cowuecmmucw mopmce co; seemewv upaugvu .NN meamvm mO.—<¢ 1_ , ¢0h<¢me 138 converter (ADC) and stored in the computer core memory. The integra- tiOn is then performed by summing all the stored digital information. The result Of the integration is printed on a teletype and is also stored in the computer for statistical treatment. The stored digital signal is also available for visual Observation by activating a display routine. Integration Of the baseline is performed prior to a fluores- cence or absorption measurement and is automatically subtracted from each integrated signal value. After a preselected number Of experi- ments have been performed, the mean, the standard deviation, and the percent relative standard deviation are reported on the teletype. 3. Fixed-Variable Time Digital Integration This method is basically a combination of the two previously described integration techniques. A computer was used as a digital comparator, integrator and readout system. The analog signal is con- tinuously digitized and stored by the computer during the entire atomization time, which is normally a fixed period. After the data acquisition is over and the atomizer is shut Off, the computer functions as a digital comparator by utilizing a peak recognition routine. The computer first searches through the stored digital information for only the largest stored number, the peak maximum, and also finds the average of the first five A-to-D conversions. The leading edge Of a peak is defined as those points that have a value equal tO or greater than twice this average and less than the peak maximum, with the condition that the values of successive points are continuously increasing. The same condition is also applied for the trailing edge except that the values Of successive points should continuously decrease. After the leading and trailing edges are recognized, the computer functions 139 as an integrator by summing all the data points between the start of the leading edge and the end of the trailing edge. The rest of the operation is similar tO that described for the fixed-time integration technique. The programs for both integration techniques have been written in assembly language. The fixed-variable time technique should only be employed when the behavior Of the signal can be predicted. If a double peak is present in the system, erroneous results are obtained. This situation does not normally exist in nonflame atomizers, however. Furthermore, double or multiple peaks can also be handled by additional software. F. General Description of the Program Since two types of automatic samplers have been used in this work, two different types of programs were written using assembly languages. Only a description Of the program employing sampler Design 11 is given here. A simpler version Of this program was employed when the fixed-volume sampler was used. Before the experiment begins, the operator communicates through the teletype with the computer and provides responses to seventeen questions listed in Table 5. Except for questions C, D, and N, the meaning of the rest of the questions is obvious. For the sample size in question C, a number 20 times larger than that required should be typed. For example, the number 20 means 1 pl while 40 means 2 pl, etc. This number actually represents the number Of pulses which are applied to the forward input terminal of the stepping motor driver card. In question 0, the sample type should be specified by a number 140 Table 5. Initialization and Optimization Questions DOW) Isa-urn H O OZZF'KL No. Of Experiments? Amplifier Gain? Sample Size = ? Sample N0. = ? II 0‘) First Time Delay II 0‘) Desolvation Time ? Desolvation Power Ashing Time = 7 ? Ashing Power Atomization Time = ? Atomization Power = ? Cooling Time = 7 Type "1" for AF and "2" for AA mode. Type "1" for optimization submode and "2" for normal submode. Type “1" to generate & "2" to delete paper tape copy of the data. Type "1" to display & "2" to delete the display of data on the scope. Type ”R“ if Ready to start the experiment. 141 between 1 to 6. These numbers represent the different sample cups. The program can provide two submodes Of operation in both AA and AF. The Operator has to select these submodes by answering question N. In the Optimization submode, after each experiment or sets of experiments has been executed, the operator is asked if variations in the parameters or experimental functions are desired. The operator can ask for variation of any parameter or function by typing the charac- ters A through 0, which refer to the questions listed in Table 5. After the parameter is changed, the computer prints the message "what else", which asks if any other variable or function is to be changed. This process continues until the Operator types "O", which indicates that optimization is completed. In the Normal submode, it is assumed that Optimization has been performed, and the computer provides a closed-loop Operation. After each experiment, there will be a preselected time delay for cooling the atomizer, and the computer starts its new cycle. When a preselected- number of experiments has been performed, the mean, the standard devia- tion and the percent relative standard deviation are printed on the teletype. In this mode of operation the computer can automatically vary atomization parameters to a limited extent after a preselected number of experiments has been completed. Programming this type of operation for all experimental variables requires mass storage or a computer memory Of more than 8K. Iterative Optimization in real time is, however, possible if the Operator provides proper experi- mental parameters in a logical manner. After initialization, the computer asks if the operator is willing to start the experiment. The Operator can respond by typing "R" and 142 the computer starts its control operation. A general flow chart of the program is shown in Figure 28. Under computer command, the auto- matic sampler syringe is first rinsed and then filled with the desired sample solution. The sampler moves to the top of the atomizer and provides a preselected sample volume. Since the syringe volume is 200 pl, the sampler has enough samples for 100 experiments if a 2 pl sample is employed in each experiment. Thus, the Purge and Fill Operations are not required for every experiment. The operating cycle of one experiment can be best understood with the aid Of Figure 29. When the sample delivery is completed, the sampler moves out Of the Optical path of the spectrometer. A fixed time delay ensures that the sampler has moved out of position. The computer then begins to control electrical heating of the atomizer. A two or three step current program is normally employed to desolvate, ash, and atomize the sample. During the atomization step, a stream Of inert gas carries the atomic vapor to the level of the monochromator entrance slit where AA or AF occurs. The fluorescence or absorption signal is converted to a proportional voltage by the photomultiplier tube and the i-v converter. The analog signal is continuously digitized at a specified rate by the computer's lO-bit ADC, and the result is stored in core memory. The atomizer is then shut Off, and the signal is integrated and printed on the teletype. The atomizer is allowed to cool for a preselected amount Of time. The program has the capability of providing a punched paper tape of the stored integrated signals and a display of the signals on an Oscilloscope if these routines are requested in the initial dialog. 143 "Ll IN! SYIINOE NOV! '0 AIOMIXII APO DIPOSII TN! SAMPLE ill!" I 510.! INTIOIAIID . IIOVIDID ONIV 1' '° ' “ ‘ ' ' ' " " OPIIATOI HAS IIOUISTID DURING INITIALIZAIION NO IIOVIOI 'APEI "INCH OF IN! GIAYE' DATA "OW ID NLV I -‘- O’IIAIOI "AS REQUESTED DURING INITIAUIAHON UITIMIIAHON 0 PROVIDED (Iva II NOIHAL ---~-'-‘---- OPIIAIOI MAS IIOUISVID SUIMODI WIINO lNIIlAlIlAIION Figure 28. General flow chart of the computer program. 144 .Eoumxm umppoeucounempaneou use :2 acmewemaxo ago we mpuxu mcpuoeoao .mm mesm_d 326 262 IL OZZOOU amN_<<._mm_o ZO_._._m_DOU< ._Omwo ><4mo Chm—m FLI V. FUNDAMENTAL INVESTIGATION OF NONFLAME ATOMIZATION FROM Pt and W WIRE LOOPS A. Introduction The computer-controlled nonflame atomic spectrometer described in Chapter IV is capable Of handling any electrically heated nonflame atomizer such as a carbon rod, a tantalum strip, and a platinum or tungsten wire loop. Since we were interested in the investigation of the parameters influencing nonflame atomization, all the experimental results reported in this chapter have been Obtained with the platinum or tungsten wire loop atomizer. Although a wire loop made of platinum or its alloys suffers from a low melting point (1700 °C) and can only be utilized for atomization of low boiling point elements, it has several advantages when used for fundamental investigations. First, the atomizer has a long lifetime. Approximately 5000 determinations can be performed with one platinum loop. Second, in contrast tO carbon rods or furnaces, there is practically no gas evolution from the atomizer. Third, it requires less than 10 H.0f electrical power to provide an atomization temperature Of about 1600 °C. Fourth, it is possible to monitor preatomization processes such as desolvation Of the sample solution on the atomizer. This is not possible when a porous atomizer such as some versions Of carbon rods or furnaces are employed as atomization devices. In the first section of this chapter, the influences Of the sheath gas flow rate on atomic populations, atomic fluorescence signals and 145 146 the precision of measurements are discussed. The second section dis- cusses the complication Of the Optimization of experimental parameters in programmed and non-programmed heating systems and presents the Optimized parameters Obtained with the former system. The analytical data Obtained with Pt and W loops are presented in the third section. The last section describes the physiochemical properties of W and Pt atomizers. B. Effect Of Sheath Gas Flow Rates The experimental results reported in this section were all Ob- tained under the conditions shown in Table 6. All measurements were performed in the atomic fluorescence mode with a two step current program to desolvate and to atomize the aqueous sample. Thus, before atomization occurs, the sample is present on the atomizer as a solid. Figure 30 shows the effect of inner and outer sheath gas flow rates on the cadmium fluorescence signal. Argon was employed for both the inner and the outer sheaths. The inner flow rate (IFR) was kept at 1 l/min, when the outer flow rate (OFR) was varied between 0 and 6 l/min Of argon. The same is true for IFR variations. It can be seen in Figure 30 that the AF signal decreases with an increase in the flow rates above 1 l/min. The signal-to-noise ratio (l/relative standard deviation) normally follows the AF signal variation with both the IFR and OFR. The change in AF signal is small between flow rates Of l to 2 l/min but becomes larger as the flow rate increases. The de- crease in integrated AF signal with increasing flow rate can be attributed tO the smaller residence time Of free atoms in front of 147 Table 6. Experimental Conditions in the Investigation of Nonflame Atomization Elements Cd Zn Hg 1. Atomization Parameters Type of Atomizer Pt Pt W Diameter of the wire, in. 0.0126 0.0126 0.01 Diameter of the lOOp, in. 0.0625 0.0625 0.0625 Sample Size, pl 3 3 2 Sample concentration, ppm .1 .2 10 Desolvation current, A 2 2 2 Desolvation time, sec. 33 33 45 Atomization Temperature, °C 1200 1200 900 Atomization time, sec. 4 4 4 Atomizer Location Bottom of slit Gas Sheath Design II Distance between gas Sheath and Atomizer, in. l l l 2. Instrumental Parameters Metal Vapor Lamp current, A 1.2 1.2 1.2 Monochromator slit width, mm 1.5 1.5 1.5 Analysis wavelength, nm 228.8 213.86 253.65 Photomultiplier supply voltage, V 800 800 800 i-v converter rise time, msec l l 1 148 ‘01 O OUTER FLOW RATE 9" I INNER FLOW RATE (INTEGRATED AF SIGNAL)X10 3 0 . 1 4 8 now RATE I/min..Ar Figure 30. Influence Of inner and outer flow rate on the cadmium fluorescence (arbitrary unit) signal. 149 the observation window, to a lower atomization efficiency, or to both. It is also possible that increased quenching effects occur at high flow rates. 1. Atomization Efficiency In order to examine the effect of flow rate on the atomization efficiency, the atomizer temperature as a function of flow rate was monitored. The results, shown in Figure 31, demonstrate that the integrated AF signal follows the atomization temperature very closely. Furthermore, an increase in the inner sheath gas flow rate causes the atomizer temperature and consequently the atomization efficiency to decrease. On the other hand, the outer sheath gas flow rate has, within a 95% confidence internal, no influence on the wire lOOp tempera- ture. However, the integrated AF signal decreases with an increase in the outer sheath gas flow rate. This indicates that although the atomization efficiency decreases with flow rate, the influence of atom residence time is also quite significant. 2. Atom Residence Time The effect Of atom residence time on the AF signal was confirmed by two sets of AF experiments with Cd as the test element. The results are shown in Figure 32. In the first set, similar to the previous experiments, the electrical current through the wire loop was kept constant and therefore, the atomizer temperature varied with the flow rate. In the second case, the computer was programmed to provide different currents at different flow rates so that the loop temperature was kept constant. It can be seen that for both sets, whether the temperature is constant or not, the integrated AF signal decreases 150 7~ -1200 \ OAF vsIFR 6‘ IT vsIFR ,1050 "2 5- *1000 X 3 2 Q 4. "-950 U) ‘2 O “4 3- ~900 2 M 8 .— .Z. 2~ ~850 1- ~800 o . . . . . 750 o. 0 l 2 3 4 5 INNER FLOW RATE, I / min 3' aflanUSdWSL USZIWOIV Figure 31. Influences of flow rate on the integrated AF signal (arbitrary unit) and atomization temperature. 151 12 .. H. O CURRENT CONSTANT g ,0. TVARYING x - CURRENT VARYING 3 94 \ T CONSTANT < z 9 8i W "<- 74 O a bi § 51 .— é 4" Si 2‘ I. O I I I I T I O 1 2 3 4 5 6 INNER now RATE, I/min.,Ar Figure 32. Influence of atom residence time on the cadmium AF (ar- bitrary unit) signal. 152 with increasing flow rate, but in all cases the signal is greater in the experiments with constant temperature. The same phenomenon was Observed for other elements such as Hg and Zn. The results for Zn are illustrated in Figure 33. Again, the OFR was kept at 1 l/min when the inner flow rate was varied and vice versa. The integrated AF signal decreases with both the IFR and OFR. The system was then programmed to change the atomization current between 3.5 to 4.3 A when both the inner and outer flow rates were kept constant at 5 and 1 l/min respectively. Currents higher than those shown resulted in loop melting. It can be seen that although the integrated Signal has increased by a factor of more than fifty, the signal is about twenty times smaller than when both the IFR and OFR were at 1 l/min. The effect of OFR on temperature for Zn was similar to that observed with cadmium. 3. OuenchingEffect In atomic fluorescence, collisions Of the radiationally excited atoms with free electrons, or with atomic and molecular particles present in the vapor cell may deactivate the excited atoms with no radiation being produced. The quenching effect in flames has been reviewed in detail by Alkemade and Zeegers (35). From a classical point of view, the number of binary collisions per second and per unit volume, ZA,B is given by 2A,B = [A][B]n(rA + rB)2 are]. (55) where srel is the average relative velocity Of the colliding particles, and rA and r8 are the radii of two particles A and B as hard spheres, 153 ATOMIZATION CURRENT, A 35 34.6 3.7 3.8 3.9 410 4g 42 4A3 15 15 14- I AF SIGNAL vs CURRENT 14 E A AF SIGNAL vs OFR r411 13.. 0 AF SIGNAL vs IFR 13 9, )> 12" 12 a 1"“ .1] z 92 23'H)‘ 'lCl g2 >< 68* -8 a; a) it 7 .7 2. D - 63 _ 1 in 0 5‘ 1.5 _ ‘*‘-‘ E '2' 4‘ P4 .3 V 9° 3" __3 SA . P 1? 2 2 ‘<:; §. 1 - _] P T I T I L 1 2 3 4 5 a INNER now RATE, I /an., A Figure 33. Influences Of flow rate and atomization current on the Zn AF (arbitrary unit) signal. 154 and [A] and [B] are the concentrations of A and B, respectively. The number of quenching collisions then is proportional to the concentra- tion of the colliding species, their cross sections and their relative velocity. The variation of AF signal with the flow rate should also be influenced by the quenching effect. The quenching cross section for noble gases such as He, Ne, and Ar, is small compared to molecular species like N2, 02, and 002. However, because atmospheric gases can become entrained in the Ar flow, the quenching effect is present when argon is employed as the sheath gas and should vary with Ar flow rate. In order to demonstrate the quenching effect, four sets Of experiments were performed. The computer was programmed to provide nearly the same temperature at various flow rates and for two different sheath gases. Thus, the variation of the AF signal with the flow rate can be accounted for in terms of the residence time Of the atoms and the quenching effect. In the first two sets, nitrogen and argon were employed for the inner and outer sheath gases, respectively. For the second two sets, argon was used for the inner and nitrogen as the outer sheath gas. In all four cases, the flow for one gas sheath was kept constant when the second one was varied. The variation of the cadmium AF signal with flow rate is shown in Figures 34 and 35. Comparison of these results with those given in Figure 30 to 32 indicates that the presence of N2 in the system decreases the AF signal. Furthermore, N2 has also caused a shift in the Optimum argon flow rate from about 2 to 3 l/min. The influence of the flow rate on the quenching effect cannot be evaluated from these results, however. This is due to the fact that the loop temperature 155 II Il‘R , Aw OOFR,N2 “é: >< 3 (10.. 2! £2 9- (I) 2 8—- 8 7.. .— <( 8‘ 6- 2‘. z 5— v 4— 3.. ‘4 21— \O\ l \ T \ at...___ l l 1 .‘fl 1 O l 2 3 4 5 6 FLOW RATE, I/min. Figure 34. Influences of the quenching effects and flow rate on the integrated AF (arbitrary unit) signal. 156 3 I IER.N2 o OFR,Ar 2.5— 2... —l U! I (INTEGRATED AF SIGNAL) XIO3 l l l 1 I) 1 2 3 4 FLOW RATE, I/min. Figure 35. Influences of the quenching effects and flow rate on the integrated AF (arbitrary unit) Signal. 157 was measured with an optical pyrometer and had an uncertainty of :50 °K. Moreover, although the number of quenching collisions is proportional to the relative velocity of the colliding particles, the effective quenching cross section decreases as the velocity increases. From these experimental results, the following conclusions may be reached. First, an increase in sheath gas flow rate results in a decrease of the atomization efficiency and of the residence time Of the atom and influences quenching. Second, when atomization occurs, the atomic vapor will be uniformly distributed around the atomizer if there is no forced flow in the system. If the atomizer is located at the bottom of the observation window, as was the case in these experi- ments, the atomic plume covers the entire area of the slit. Because of natural convection, however, the atomic vapor expands and gradually leaves the Observation window. If the production of atomic vapor re- sults in a laminar plume, the integrated AF signal would be expected to be at its maximum and would be very reproducible. On the other hand, a turbulent plume would result in a lower signal-to-noise ratio (higher relative standard deviation). The inner gas sheath confines the atomic vapor and causes an improvement in the signal-tO-noise ratio. However, as the inner flow rate increases the volume of the atomic plume in front of the slit shrinks and becomes more cylindrical in shape. In the extreme situation, for large flow rates, the cross-sectional area Of the cylinder becomes smaller or comparable to the area Of the slit. Because of turbulence, the atomic vapor does not pass in front of the observation window in a reproducible manner and the signal-tO-noise ratio deteriorates. Third, the decrease Of AF signal with an increase in the OFR can also be explained similarly. The atomic 158 vapor should occupy a volume with a cross-sectional area greater than the Size of the inner gas sheath. Increasing the outer flow results in a shorter residence time, but the decrease in the signal-to-noise ratio should be less than that expected for increases in the inner flow rate. Thus, even though the AF signal decreases because of a decrease in atomization efficiency and a decrease in residence time, the OFR would not be able to reduce the atomic vapor volume to the same extent as the IFR. It was experimentally Observed that the rela- tive standard deviation changes by more than one order Of magnitude for variations of the IFR, but the same variation in the OFR produces a less pronounced effect. When helium gas was employed as the sheath gas instead Of argon in the study of Hg atomic fluorescence, it was Observed that although the AF signal was decreased, the signal-tO-noise ratio was improved by at least a factor of two to three. The improvement can be explained in terms of the thermal conductivity of He, which is about 8 times that of Ar. Helium also has a smaller quenching cross-section than argon. Furthermore, our calculations in Chapter II Show that low atomic or molecular weight gases like N2 can provide a more laminar flow than the high atomic or molecular weight gases like N2. C. Optimization Of Experimental Parameters 1. Introduction In order to Obtain the maximum signal and minimum relative stan- dard deviation, a large number Of measurements was performed to find the Optimum experimental parameters. Critical optimization was only performed for the analysis of Cd, Zn and Hg in the atomic fluorescence 159 mode. No rigorous optimization was done for the detection Of other elements. Although a set of optimum parameters in AF are not generally optimum for AA measurements, the same set Of optimum parameters was employed for AA measurements. Furthermore, it was found that, except for the atomization period, the Optimum conditions were approximately the same when programmed or non-programmed heating of the atomizer was employed. The detection limits and the signal-tO-noise ratios were drastically different, however. Therefore, the optimum parameters for programmed heating are reported here. Under optimum conditions, the AA and AF signals were approximately Gaussian in shape and had a time duration which varied from 100 msec at low concentrations to 4 sec at high concentrations when aqueous samples were used. Prior to optimization of the atomizer parameters, all the optical and electrical parameters of the Spectrometer were adjusted for the highest signal-tO-noise ratio. These adjustments included the position and electrical parameters of the excitation sources, the position of the nonflame cell with respect to the source, the monochromator spectral handpass, and the photomultiplier anode- to-cathode supply voltage. In the optimization of the atomizer param- eters, the atomization temperature was kept constant for the variation of the flow rate as well as other parameters. The optimum parameters were found to be independent of the type of the wire lOOp atomizer provided that the wire diameter and the loop size were identical. Unless otherwise stated, all optimum parameters have been Obtained with gas sheath design II. The results of the optimization study are summarized later in this section. 160 2. Atomization Processes Without PrograImIed Heatipg The experimental results presented in the previous section were all Obtained with a two step current program to desolvate and to atomize the sample. In other words, it was assumed that the sample is present as a solid just prior to atomization. ‘Because the analysis time can be reduced and sample loss can be prevented during the desolvation and ashing steps, it is desirable to atomize the sample in one step. This method Of atomization, however, may complicate the Optimization Of experimental parameters as well as the atom formation processes and the detection processes. A discussion Of some Of these complications is presented here. As was previously stated, the limits of detections and the relative standard deviations Obtained under optimum conditions, with a multistep programming system were superior to those Obtained when atomization was performed in a single step. Another important charac- teristic of the non-programmed system is that the signal behavior may vary significantly when the system is not Operated under Optimum con- ditions. Figure 36 shows two atomic fluorescence signals for 10 ppm Hg when the atomizer was not optimized and gas sheath design I was employed. The long leading edge Of the signal is due to sample explosion on the wire loop. In other words, as a result Of a change in the mechanism of heat transfer, the solvent and solute are not separated in time and there is some light scattering by the sample. Thus, the fluorescence signal is seen on top Of the scattering signal. When gas sheath design II was tried, it was realized that the optimum parameters Obtained with the previous gas sheath could not be duplicated. A variety of parameters such as the inner and outer flow 161 .mcoruvveou uw~22ruaoucoz gone: a: sun op mo m—ecmvm oucmummgoapu Oveoue oz» .03 .922. HOVI'T OA .om mg:o_m 162 rates, the distance between the atomizer and the gas sheath, and the atomizer temperature were varied to find the Optimum parameters. In contrast to design I, it was practically impossible to prevent the simultaneous vaporization of the solvent and solute. The detection limits and relative standard deviations increased drastically. In all cases, the signal—tO-blank ratio was maximum at a flow rate of 2 l/min., however. It is obvious that the changes in the behavior Of the system are due to the different designs and constructions of the gas sheaths. Thus, optimization, if not impossible, can be extremely tedious. This originates from the fact that most of the parameters are dependent variables. In order to simplify the system, the number of parameters and their interdependence should be decreased. Programmed heating of the atomizer not only separates the vaporization Of the solvent and solute in time, it makes the function of the carrier gas simpler, the effect of the gas sheath design less critical and the sources of light scattering and quenching (AF) more evident. 3. Optimization of parameters for Programmed Heated Wire Loop Atomizers 3. Introduction. Among the parameters unique to electrically heated filament atomizers are the diameter and size Of the filament, the sample size, the sample type, the gas sheath designs, the inner and outer sheath gas flow rate, the distance between the atomizer and the gas Sheath, the vertical distance between the Observation win- dow and the atomizer, the desolvation, ashing, and atomization times and temperatures. All the Optimized parameters reported here have been obtained with aqueous solutions. Therefore, a two step current 163 programming technique was employed. The Optimized parameters for the atomization step are reported in terms of temperature instead Of cur- rent or power. The latter are a function of loop dimensions and other parameters in the system. The temperature was measured by an optical pyrometer and was reproduced for different flow rates by making relative emission measurements at 6500 A. For the desolvation period, where there is no visible emission from the atomizer, the electriCal current through the atomizer was optimized instead of temperature. b. Preliminary Optimization. Wire loops with diameters larger than the maximum monochromator slit width were not employed in the work. The best loop size had a diameter Of 2 mm. The best wire diameter was found tO be 0.0126 in. Larger diameter wires could not normally provide a uniform temperature at the loop center. Wires with diameters less than 0.0126 in. were hard to fabricate. With this type Of a loop, a sample size of 3 pl was found to provide the best precision. Since the desolvation current for a Hg sample should be very small to prevent the covaporization of Hg, a sample size of 2 p1 was employed to prevent long desolvation times. All optimized parameters have been obtained with gas sheath design II. The distance between the atomizer and the gas sheath was selected as l in. The atomizer was located slightly below the mono- chromator slit with the primary source focused 1 mm above the wire loop. This arrangement allows AF measurement before condensation Of the atomic vapor occurs. c. Optimization Procedure. After the preliminary optimization, the following procedure for Optimization of other parameters was 164 normally followed. First, the desolvation time and current were ad- justed so that the water vaporizes smoothly and leaves the solute on the atomizer. Second, an atomization temperature well above the boiling point of the element and an atomization period about twice the actual atomization time were also selected as initial estimates. Third, using these parameters, the inner and outer sheath gas flow rates were Optimized for maximum signal-to-noise ratios. Fourth, since the outer flow rate did not change the atomizer temperature ap- preciably, the variation of inner flow rate with atomization tempera- ture was monitored at the Optimum outer flow rate. Both the AF signal and the signal-tO-noise ratio were measured. Fifth, the desolvation current and time were optimized at the Optimum atomization temperature and optimum inner and outer flow rates. d. Inner-Outer Sheath Gas Flow Rates. Using the above procedure, the parameters were optimized for Cd, Zn, and Hg. The desolvation time and current for both Cd and Zn were chosen to be 33 sec and 2 A respectively. The atomization period was started immediately after water vaporization was visibly complete. The desolvation parameters for H9 were 45 sec and 0.6 A, and atomization was initiated before complete vaporization of water. This prevents possible vaporization Of Hg during desolvation. In all cases, an atomization temperature and period of 1200 °C and 4 sec were employed as initial estimates. Figure 37 shows a diagram Of the influence Of flow rate upon the relative standard deviation (RSD) of Cd AF signals. For the center contour line, the R50 of AF measurements was 5 percent or less, while for the next to the center line, the R50 varied between 5 to 10 per- cent. The Optimum sheath gas flow rates were chosen to be inside 165 .ovuee mmvocIouIchmpm on» so one; sop» mo mucmzfimcH .mm weaned .e_E\_.memo 0¢_h<._m~_ wZ... mDOPZOU ._Pommu mo mmucm:_m:~ m2... «DOPZOU ZO.._.<_>mo 0¢_._.<._w~_ m2: «DOhZOU .220.» m< .l .85. m2: 20:45.5me mm «m mm mm % 1 <3 00 ,_'. l ‘1 cu J ‘t c: N N V ‘lNSHUUO NOILVA'IOSHCI .mm «Lammm 169 Table 7. Optimum Platinum Loop Atomizer Parameters Using Programmed Heating IFR OFR Desolvation time Desolvation Atomization Element 1/min 1/min time, sec. Current, A Temperature, °C Hg 1 l 40 0.8 1000 Cd 2 2 34 2 1250 Zn 2 2 34 2 1300 parameters for these elements. These parameters were employed to obtain the analytical results presented in the next section for these elements. As can be seen from Figures 37 to 39, the optimum parameters for any system can vary within a certain range. For example, Hg can be analyzed between 1000 to 1300 °C with practically identical results. The optimum parameters for Cd are almost the same as for Zn because these elements have almost the same boiling points. 0. Analytical Results After optimum parameters were chosen for the analysis of Cd, Zn and Hg, a study of the long term stability of the AF system was made, and analytical data were obtained. A comparison of the detection limits obtained with a programmed and a non-programmed heated platinum loop was also performed. The optimum parameters obtained by AF were also employed for the determination of Cd, Zn, Hg and other elements by AA. 170 1. System Stability In all cases, the overall stability of the system with multi- step programmed heating was found to be superior to cases when a one step atomization was used. Figure 40 demonstrates the long term stability of the system for 100 consecutive samples each of which contained 150 pg of cadmium. A 10 sec. cooling period was allowed between successive analyses. The system was operated in the AF mode with a programmed heated Pt loop and with the fixed-volume sampler. No operator adjustments were made during the total analysis time of 100 min. The largest deviation from the mean was approximately 15%, and the relative standard deviation of 100 samples was 6.5 percent. In a similar study for the AF analysis of Zn, and Hg, the relative standard deviations of 100 samples were 6.3 and 7 percent, respec- tively. When a one step atomization was employed the relative standard deviation ranged between 8 and 10 percent under the same operating conditions. Both the short and long term reproducibility of the system are influenced by, among other factors, the physiochemical properties of the atomizer. This is discussed in a later section. 2. Analytical Working Curves AF calibration curves under optimum conditions were obtained for Cd, Zn and Hg with programmed heating. All the working curves showed a sigmoid shape. In all cases the monochromator was flushed with argon and triply distilled deionized water was used in making all solutions. The analytical curves, Figure 41, for Cd, and Zn have a linear range of 5 orders of magnitude, while the linear range for 171 901 o ' O SAMPLE NUMBER -2o -10 o 10 20 DEVIATION FROM MEAN,% Figure 40. Long term stability of the computer-controlled AF spectrometer. l cidd NOIlVUlNBDNOD 01 [:0 [0'0 [000 01 172 INTEGRATED AF SIGNAL(ARBITRARY UNIT) ‘ —l 9N T ‘Q'E: v 90 Q N 3 1 W ‘ E ., 3.7. o 8 l L l l Tl Figure 41. AF analytical working Curves for Cd, Zn, and Hg obtained under optimized conditions. 173 Hg is about 3 orders of magnitude. The vertical markings around each experimental point indicate the relative standard deviation of 10 samples. All working curves show a slope of unity in the linear region. For extremely low concentrations, the slope differs from unity, although the precision of measurements is about 15 percent relative standard deviation. In this area, the blank concentration of the elements becomes comparable to the concentration of the analyte in solution. 3. Detection Limits a. Results. The lowest concentration in the unity slope calibration curve was defined as the detection limit. It should be noted that this definition gives conservative detection limits compared to the most common definition (S/N=2). Table 8 shows a comparison of the Table 8. Comparison of AF Detection Limits (ng/ml) With and Without Programmed Heating of Platinum Loop Elements Programmed Heating Non-Programmed Heating Hg 50 1000 Cd 0.001 10 Zn 0.04 50 detection limits for Cd, Zn, and Hg with and without the programmed heating of the Pt loop. The same optimum parameters were also employed to obtain the detection limits of the same elements by the AA technique. 174 Detection limits were also obtained for Tl, Mg, Mn, Ni and Pb in atomic absorption with a W loop atomizer. Table 9 summarizes the detection limits of the elements obtained in AA. b. Discussion. There are many factors responsible for the lower detection limits with the programmed heated system. Some of these factors are evident if one refers to some of the advantages of nonflame atomizers over flame atomization, which are described in Chapter II. In general there is: l) more precise control over the spectral and chemical environment; 2) no flame gas expansion on com- bustion; 3) higher efficiency of free atom production; 4) less scatter- ing of light and far less quenching of excited atoms (AF). In other words, in any nonflame method the increase in S/N is not only due to increases in the signal, but also to a decrease in noise. If the nonflame system is not programmed heated, the signal decreases because points 1, 2, and 3 are not present, and the noise increases due to the opposite effect of points 1, 2, and 4. Even in the simplest case of an aqueous solvent, the non-programmed system is complicated by the nature of the boiling process, which is not yet fully understood. In order to describe the different processes which occur during solvent evaporation, we first simplify the system by assuming that a hypothetical high melting point wire loop is heated electrically in a pool of deionized water at atmospheric pressure and at the saturation temperature of 100 °C. We further assume that the wire is heated uniformly and define the difference between the wire and the solvent temperature as ATE or the excess temperature. A plot of heat flux (fléggc h = heat transfer coefficient, A = surface area) against excess temperature (277) indicates, that there are five distinct rates of 175 Table 9. Atomic Absorption Detection Limits (pg/ml) With Programmed Heated Platinum and Tungsten Loop Atomizers Analysis Wavelength Detection Elements A Limit Atomizer Hg 2536.5 0.2 Pt Cd 2288 0.0005 Pt Zn 2138.6 0.0005 Pt Tl 2767.9 0.06 Pt, W Mg 2852.1 0.02 W Mn 2794.8 0.08 W Ni 2320 0.7 W Pb 2170 0.08 W The following parameters were kept constant for all elements. PMT voltage, V 800 Monochromator Slit Width, mm 0.1 Sample size, 01 3 Flow Rate, filmin 2 For W Atomizer Desolvation time, sec 25 Desolvation current, A 3 Atomizer temperature, °C 2000 HEAT FLUX 176 BCDHJPMCB PURE NUCLEATE “ng LTABIE FILM ' BOILING BOILING 199NVECT'9'#. REGIME REGIME 5 [TRANSITION I if- 15C) l L l l l 1.0 10 100 1000 10,000 ATJF EXCESS TEMPERATURE Figure 42. Typical boiling curve for a wire in a pool of water at atmospheric pressure (heat flux has arbitrary unit). 177 heat flux variation. These five regions are shown in Figure 42 and are as follows. In the first region (0ATE<100 °F), and the heat flux changes rapidly with increasing wire temperature. Visual observations indicate that in the second region, individual bubbles form at certain places on the wire. They grow and push hot water from the proximity of the wire into the colder bulk of water. In the third region, one sees many rising streams of bubbles as ATE increases. These columns cause a considerable stirring action in the water and the heat transfer co- efficient increases. It is obvious that the transfer rate is greater from wire to water than from wire to vapor. Thus. as ATE increases more, a greater percentage of the wire surface is blanketed by bubbles. Eventually, this increased blanketing begins to counteract the tendency of higher ATE values to produce higher flux, and the two effect balance at some point. This maximum heat flux occurs at a temperature called the "burnout" point. Beyond this point an increase in the heater temperature results in a decrease in the heat flux. This region is called the transition boiling region. Finally, in the stable film boiling regions, a vapor film covers the entire wire surface. An event similar to this process can be observed when droplets are sprayed on a red hotplate. Because of the formation of a vapor film between the plate and water, the droplets dance on the plate and do not evaporate immediately. Moreover, the effect of radiation from the wire in this region becomes quite high, and this results in an 178 increase in the heat flux with respect to ATE. In the transition boiling region, stable film and nucleate boiling take place alternatively, and this results in a very violent process. However, in the case of an electrically heated wire, transition boiling is not usually observed. The reason is that if at the maximum heat flux, the rate of energy generation is increased, the wire temperature rise provides a boiling process which cannot remove heat as rapidly as at the maximum heat flux. The difference between the generation and the dissipation rate produces a higher wire temperature with a conse- quent jump to the stable film boiling region. We believe that the mechanism of solvent evaporation in the non- programmed system is similar to the boiling processes described above. The problem is, however, more complicated because of the presence of the solute and increased surface roughness, which tend to increase the heat transfer coefficient. Moreover, the sample is not at the saturation temperature initially, the instantaneous amount of both solvent and solute is a strong function of wire temperature, and the free and the forced convection which exist because of gas flow in the system complicate the processes. Due to the limited amount of the sample, we believe the boiling process from a heated wire filament involves only the first three regions. When the current is turned on, the wire temperature increases gradually (not instantaneously). Even if the bulk temperature is below the saturation temperature, the boundary layer is superheated and bubbles form next to the wire. However, because of the temperature gradient the process would be more violent and much higher heat fluxes can be obtained in this case. The extreme agitation is responsible for 179 partial evaporation of the solute. Some of the solute is also lost when the cold sample splashes against the hot wire. Furthermore, as the amount of solvent is decreased by evaporation, the solute would be left behind on the loop. Since the final part of solvent evaporation is usually located on only one part of the atomizer (276), some solute evaporation results in this manner. A combination of these effects results in light scattering. Moreover the presence of water as well as other molecular species result in the quenching of excited atoms in AFS. In a programmed heated system, on the other hand, the excess temperature can be kept low (<10 °F), which results in having a low heat transfer coefficient. The process is a conventional free convec- tion. Although it is true that in a programmed heated system, the amount of heat for each of the steps and the time duration for each step should be optimized, the estimation of these quantities, because of the simpler heat transfer process, provides less of a problem. E. The Physiochemical Properties of W and Pt Atomizers The % RSD for the optimum parameters ranges between 2.7 to 7% for both Pt and W wire loops. The long term reproducibility of the W atomizer was not as good as for the Pt atomizer. This can be ac- counted for by the oxidation of the W loop, by its evaporation, and by chemical reactions between tungsten and the solute. These effects are discussed below. 180 l. Evaporation of the Atomizer It has been shown that ac heating results in a higher evaporation rate than the dc method (278). The rate of evaporation of tungsten in a vacuum is given by (279) Log m = 7.5 - £99 (57) where m is in g/cmz. sec and T is °K. It can be seen that if the temperature is increased from 2000 °C to 3000 °C the evaporation rate is increased by a factor of 2 x 105. Although the evaporation rate is reduced in an inert atmosphere, it cannot be neglected. This effect is more or less present in the case of any nonflame atomizer and results in an increase in the electrical resistivity of the atomizer. 2. Chemical Reactions A study of the chemical reactions of tungsten shows that although the metal is very inert, it is dissolved by strong oxidizing agents, and it is slightly attacked by phosphoric, sulfuric and nitric acids between 100 - 110 °C. Moreover, even for simple sample matrices, it can react with oxygen, water vapor, hydrocarbons, nitrogen, and other species between 500 - 1500 °C. According to our experience, oxidation of tungsten was one of the major factors responsible for the poor long term reproducibility. It was possible to differentiate different tungsten oxides by their colors. At 500 °C tungsten oxidizes and forms a black protective film "18049 or W02 (280). The gain in weight of tungsten follows a parabolic law (281). G = D.k.t (58) 181 where GW is weight gain, 0 is a diffusion constant, k is a constant and t is time. With additional oxidation, the black oxidized layer changes to a green (at 600 - 950 °C) or a yellow (at 1000 - 1100 °C) layer which is no longer protective. This new trioxide layer changes the parabolic law into a linear law (281). W = k' . t (59) The yellow trioxide layer, which is also an electrical insulator, vaporizes rapidly above 1000 °C, and this opposes the gain in weight. Although the above laws depend on the amount of oxygen in the system, we had no success in preventing the loop oxidation by the use of outer gas sheaths. In fact, it has been shown that tungsten is oxidized even when the partial pressure of oxygen is 10"13 abms (282). It should be realized that since heat is localized in the center of the atomizer, different forms of oxides exist at different parts of the W loop. The two ends are black and this gradually changes to green and yellow toward the loop center. Finally, the atomizer is usually ruptured at the yellow trioxide layer. Even if it is assumed that evaporation of the W loop is minimal in the presence of argon gas and oxide layers and it can be neglected in the discussion of long term reproducibility, formation of oxides and chemical reactions lead to an increase in the electrical and thermal resistance of the atomizer. Moreover, at temperatures of 2600 - 3200 °C, the loop may deform by shearing along the boundaries of recrystallized grains or creeping in the wire, which results in lowering the electrical efficiency of the wire. This means that the required power to atomize the sample is a function of at least time 182 and the number of experiments. If the power is kept constant, then the long term reproducibility deteriorates. The same explanation also holds true for the platinum loop. The better long term reproducibility of Pt is due to its low chemical affinity. It forms a monoxide at 510 - 560 °C when heated in oxygen, but the oxide decomposes above this temperature. Thus, the loop remains bright. The rate of volatili- zation is smaller in an inert gas than in oxygen or air, yet there is a loss in weight. The gradual evaporation of the nonflame atomizer material is a fundamental disadvantage of these vapor cells. It results in a drift of atomization temperature and determines the lifetime of the atomizers. However, several methods can be employed to compensate for the variation of temperature and to extend the lifetime of the atomizer. These tech— niques will be discussed in Chapter VI. VI. THE GRAPHITE BRAID, A NEW NONFLAME ATOMIZER A. Introduction The Pt loop atomizer, as stated in the previous chapter, is of considerable value in the investigation of the general behavior of filament-type atomizers. The atomizer, however, is of no significant value for the analysis of elements with boiling points greater than 1500 °C. Although a tungsten loop atomizer can provide high Operating temperatures, the metal is subject to oxidation and extreme control of the atomization chamber is necessary. The lifetime of the atomizer at a temperature of 900 °C is about 250 samples, while increasing the temperature to 2000 °C reduces this figure to a maximum of 40-50 samples under the best conditions. Furthermore, since the mechanism of atom formation may be quite different from the mechanism of atomiza— tion from carbon or graphite, higher operating temperatures may be required to provide the same atom population. This increases the background emission from the atomizer and deteriorates the signal-to- noise ratio. For a tantalum loop or strip atomizer, the temperature is limited to a maximum of 2600 °C by the characteristics of tantalum metal. The atomizers made from carbon or graphite appear to be highly applicable for analytical purposes because of the high operating temperatures (2500-3000 °C) which can be achieved. These atomizers, however, exhibit several practical limitations. Graphite or carbon 183 184 tubes and furnaces, for example, may require several kilowatts of electrical power to achieve atomization temperatures in the range of 2500 to 3000 °C (9.67.146). The different versions of graphite and carbon rods and cuvettes provide similar temperatures with a power of 500 - 1500 W, which depends on the size and design of the atomizer (128,146). Since the transfer of electrical energy into thermal energy is not perfectly efficient, a fraction of the electrical energy will be dissipated as electromagnetic radiation. Extensive shielding and/or careful circuit design must be employed to minimize these electromagnetic interferences. The sheer bulk of a system which requires kilowatts of power also prevents its application as a portable field analyzer. The power requirements of carbon rods or filaments can be reduced by narrowing the central part of the atomizer. This intentional con- finement of heat to the central region of the rod type atomizers also results in prolonged filament lifetime and reduction of background radiation (131), but it can cause some loss of sensitivity due to sample diffusion in the rod. The problem is more serious when larger volumes of the sample are used to increase sensitivity. A pyrolitic graphite coating makes the atomizer impervious to most gases and liquids, yet it is desirable to have a porous filament to minimize creeping of the sample, as in the analysis of fuel oil (162). Thus, it would be advantageous to have a uniform temperature throughout the atomizer. Furthermore, since in atomic absorption, absorbance is proportional to the optical path length, a uniform temperature throughout the atomizer can be helpful in increasing the absorbance. In an effort to combine the virtues of Pt and W loop atomizers 185 (compactness, low power consumption, speed of operation) with the West rod (compactness, speed of operation, range of elements) and the L'vov furnace (range of elements, freedom from interference), a new filament- type nonflame atomizer, the graphite braid atomizer (GBA), has been developed and investigated. The electrically heated GBA is capable of providing temperatures similar to other carbon or graphite atom- izers, but with lower applied powers. The porous nature of the atomizer provides a furnace-type environment with a uniform temperature through- out the GBA. The new atomizer has been found to give high atomization efficiencies and low detection limits in the AA and AF determinations of several elements. In contrast to other carbon or graphite fila— ments or furnaces where a cooling system is required for the atomizer holder, no cooling system was needed for the GBA during this work. In the first part of this chapter, the characteristics of the GBA are discussed. The various heating methods of nonflame atomizers are presented in the second section along with the application of these techniques, to the graphite braid and wire loop atomizers. Finally, the AA and AF analytical applications of the GBA are demonstrated in the last section. 8. Characteristics of the GBA 1. Physical and Chemical Properties The graphite braid (Union Carbide Corp., Parma, OH) contains high strength, flexible, continuous filaments of graphite which are woven to form a thread of 1.5-2 mm in diameter. The physical properties of the GBA are summarized in Table 10. The type of gases evolved from 186 Table 10. Physical Properties of the Graphite Braid Atomizer Properties Unit Magnitude Density g/cc 1.5 Melting Point '°F does not melt Sublimation Point °F 6600 Specific Heat at 70 °F BTU/(lb)(°F) 0.17 Mean specific heat between 70 to 2700 °F °F 0.4 Electrical Resistivity at 70 °F ohm - in 0.0013 930 °F 0.0010 2730 °F 0.0006 Nominal Yield Yard/pound 412 Price S/pound 85 Table 11. A typical List of Major Elements in the Graphite Braid Concentration Melting Boiling Element PPM Point °C Point °C Fe 35-50 1535 3000 Ca 20 850 1487 Mg 20 651 1109 Si >5 1427 2290 Ti >5 1800 5100 Cu 5 1083 2595 Al 3 660 2057 Na >l 98 892 187 this material has been determined (283) in the presence of a vacuum and at a temperature of 1500 °C. Although the exact quantities are not known, some small quantities of sulfur, nitrogen, oxygen and hydrogen are evolved. Sulfur has been measured at 0.003 percent. The results of an emission spectroscopic analysis of the braid, shown in Table 11, suggest that the total impurity content does not exceed 150 ppm. Braids used in this work were outgassed by resistive heating to about 2500 °C for a period of 2-3 sec. The cleaning procedure was repeated until the AA or AF baseline was constant. This usually re- quired two or three repetitions of the cleaning cycle. 2. Braid Temperature Figure 43 shows the influence of applied power and applied cur- rent on the atomizer temperature. The temperature of the braid was found to be variable up to 2600 °C. The temperature dependence on power for a GBA 3 cm long, was approximately linear over the range 1000 °C to 2500 °C and had a slope of about 5 °C/W in this range, with only 350 W needed to heat the atomizer to 2500 °C. When the length of the braid is shortened, the required power needed to reach a given 'temperature is considerably less than the power needed for the longer braid as expected. Furthermore, when the diameter of the braid is decreased, the braid resistance increases and allows less current through the GBA. The relationship between the atomizer temperature, the voltage applied across the atomizer, and the applied power for a GBA of about 0.5 mm diameter is shown in Figure 44. The power requirement has decreased by a factor of about 3 and only 125 W are needed to reach a temperature of 2500 °C. The voltage across the C ATOMIZATION TEMPERATURE 188 CURRENT.A 5 10 15 20 25 2600 ‘ I r I a C J 2400 r 0 2200 - O O 2000 _ 1800 b 0 O 1600 -- o o 1400 P 1200 0 0 1000 C, @ TEMPERATURE vs POWER 800 A TEMPERATURE VS CURRENT I I I I l I I 100 200 300 400 500 600 POWER.W Figure 43. Influence of applied power and applied current on the GBA temperature. ATOMIZATION TEMPERATURE 189 VOLTAGE .V 10 20 30 1 j r O TEMPERATURE vs POWER A TEMPERATURE vs VOLTAGE 3200 — 2800 - A 2400 - 2000 I- 1600 _ A 1200 - A / soo - l l l 0 so 100 150 zoo POWER .W Figure 44. Influence of applied power and voltage across the atomizer on the temperature of a 0.5 mm diameter GBA. 190 atomizer, however, increases approximately linearly with temperature and about 28 V are required to reach a temperature of 2500 °C. The time required to reach a given temperature with this diameter GBA is shorter than when larger diameter braids are employed. Moreover, the background emission of the atomizer is smaller, because of the smaller dimensions of the atomizer. This type of braid is not com- mercially available, but because of the lower power requirement, lower background emission, and fast response, it should be investi- gated in detail. Throughout this work, the GBA had a diameter of l~2 mm and was 3 cm long. 3. Braid Emission Spectra Figure 45 shows the spectral emittance of the graphite braid as a function of wavelength and atomization temperature. For these experiments, the GBA was mounted in the optical path of the spectrometer, about 5 mm above the bottom of the monochromator slit, by an arrange- ment similar to that described for the Pt loop atomizer. The spectrom- eter was operated in the emission mode and the spectral emittance was monitored as a function of the wavelength at three different tempera- tures. The maximum spectral emittance moves towards lower wavelengths as the temperature is increased. In order to improve the signal-to- noise ratio, a R166 photomultiplier tube was employed instead of the 1P28 tube. The R166 phototube spectral response extends from 1600 to 3200 °A. Figure 46 compares the spectral emittance of the GBA obtained with the two phototubes. Note that the photocurrent from the R166 tube is amplified by three orders of magnitude. When the GBA is mounted below the optical path of the spectrometer, as is the case in AA and 191 covumNPsoua mo corpucam a ma m3 vcm mazumcmnEmu <. 1525m><>> come coon _ 00¢? comm _ .mw mgsmva ooou .aup S d «4. m ruup “Mu V 1 «a. m _|L rnO— I. V N 0 n4. Av— SPECTRAL EMITTANCE 192 3 .. PHOTOMULTIPLIER lP28 2 >— 1 p ] ..ullll”l”” i” l 2000 3200 4400 5600 6800 8000 3000) PHOTOMULTIPLIER R166 2000 — 1000 — l‘ 1| (Immwflflflfllflfllflflflmmmm l 2000 3200 4400 5600 6800 8000 WAVELENGTH,A Figure 46. Comparison of the spectral emittance (arbitrary unit) of GBA with two phototubés. 193 AF measurements, the braid emission would be negligible below 3200 A. For the determination of high boiling point (b.p. >3000 °C) elements such as platinum, the atomizer is employed at its maximum operating temperature, and the atomic vapor should be observed as close to the braid surface as possible. For these determinations the GBA emission may cause some interference at concentrations close to the detection limits. This problem can be eliminated, however, if a suitable baffle is utilized to limit the braid emission reaching the observa- tion window or if an ac detection system is employed in conjunction with a modulated radiation source instead of a dc system. Such systems were not found necessary throughout this work. 4. AA and AF signal Characteristics During the atomization step, the atomic vapor above the atomizer gives rise to an AA or AF signal with a time duration which depends on a variety of factors such as the sample size, matrix, and atomization temperature. Typical signals were approximately Gaussian in shape with half-widths on the order of 0.5-1 sec. Integration of the absorp- tion and fluorescence peaks was used to improve the accuracy and to' extend the linearity of the wrrking curves. The photometric signal was integrated for the duration of the atomization step. After an aqueous sample is placed on top of the braid, it soaks down between the graphite fibers. Heating the fibers produces a furnace-like environment. The heating processes are very efficient because a high surface area is kept in contact with the sample. The possibility of sample explosion is minimized, again due to the furnace- tyDe environment around the sample. An explosion and a concurrent 194 loss of sample could be observed if the sample were too viscous to soak into the braid and the desolvation temperature were set well above the boiling point of the solvent. The effects of varying the length of time which the sample re- mained on the atomizer prior to desolvation were studied. For soaking times of up to 20 seconds, the mean AF signal for a 1 pg/ml aqueous zinc solution remained constant, within a 95% confidence interval. The standard deviations of replicate trials were also unaffected by the soaking time, probably because the entire length of the braid is heated to a uniform temperature. Soaking of the sample into the braid may be a problem with a different sample matrices, however. When programmed heating was not employed, the mean AF signal increased initially with the soaking time, reached a maximum and finally de- creased. In contrast to the platinum loop, no covaporization of the solvent and solute was observed with the GBA. 5. Mechanisms of Atom Formation 1 The process of atom formation in flames can either be thermal or chemical in nature. For many easily atomized elements one needs to elevate the temperature to a sufficient extent in order to liberate free atoms. For other elements, which form refractory monoxides, ; molecules with dissociation energies on the order of 7 eV or more this should not be the case. The efficiency of free atom formation is higher in fuel-rich flames, although stoichiometric flames can provide higher temperatures. The determination of 8, Si, and W in the nitrous oxide - acetylene flame is one of the best examples of the effect of chemical formation of free atoms. The mechanism of free 195 atom formation in this flame is still unclear, but because of the presence of the reducing environment, the following mechanisms have been suggested. + M0+C+M+CO M0+NHIM+N0+H M0+CNIM+CO+N The mechanism of atom formation in nonflame atomizers depends on the type of nonflame device, the element, the chemical compound being analyzed, and the amount of oxygen in the system. For those atomizers made from metals such as Pt, W, and Ta, the process should be thermal in nature. If a sheath gas such as H2 is employed, a hydrogen diffus- ion flame is produced during the atomization period. This reducing environment should minimize interferences, remove the oxide layer on the atomizer, and extend the range of determinable elements. For atomizers made from carbon or graphite, the hot carbon may assist in the atom formation process as well as the high temperature. The presence of carbon or graphite, however, presents a difficulty in analyzing for elements such as Al, Si, W, and B that form notoriously stable carbides. Although it has been possible to dissociate such carbides by thermal means in the case of Al and Si, such is not the case for W and B with carbon or graphite devices not coated with pyrolitic graphite. The less reactive nature of the pyrolitic graphite also reduces the role of hot carbon in assisting the atom formation process. The nature of the process is questionable, however. Whatever the mechanism 196 of atom formation, thermal or chemical, the GBA should be superior or at least equivalent to the graphite or carbon filament devices in free atom formation. This is due to the porous nature of the atomizer and the large surface area which is in contact with the sample. Whether the atom formation is thermal, chemical or a combination of the two, the higher surface area of the GBA should provide higher free atom populations at the same temperature than other metals, carbon, or graphite filament devices. 6. Braid Lifetime The lifetime of any nonflame atomizer depends on a variety of parameters such as the sheath gas flow, the sheath gas type, the sample size, the matrix, the method of programmed heating of the atomizer, the desolvation, ashing and atomization temperatures and the duration of each heating step. The following discussion is applicable to all electrically heated nonflame atomizers. a. Sheath Gas Flow - Sheath Gas Type, The rate of vaporization of any material in a vacuum depends on its vapor pressure and tempera- ture. The rate of evaporation in an inert gas atmosphere at atmospheric pressure is about 60 times slower than the rate in vacuum. Using Equation 12, the evaporation rate for a carbon atomizer was calculated and plotted as a function of atomization temperatures as shown in Figure 47. It can be seen that when the atomizer temperature is varied from 2100 to 3200 °C, the evaporation rate changes by five orders of magnitude. If the sheath gas or the impurities in the sheath gas react with the atomizer, the evaporation rate is still higher. The evaporation rate is also higher in a dynamic system compared to W T." 197 0 I04 0' o a: In N. (E) 10° \ . OI LL! #- '7 ‘I: 1(1-1 (I: «0 Z 9 I— "_ ‘<: l() {. a: O 3: > 16'" LL] 10 2000 2400 2800 3200 ATOMIZATION TEMPERATURE .C Figure 47. Influence of the atomization temperature on the atomizer evaporation rate. 198 a static system. As the flow rate is increased, the vaporized material is swept out of the atomization chamber and the atomizer consequently produces more vapor. In addition to the above effects, for a GBA of very small diameter, the sheath gas applies a pressure on the braid which can cause the deformation or rupture of the braid fibers at higher flow rates. b. Sample Size - Sample Matrix. The effect of sample size on atomizer lifetime is evident. The larger the sample, the more energy is required to atomize it, and this decreases the atomizer lifetime. This indicates that the duration of each heating step should be extended to completely desolvate, ash, and atomize the compound. The sample matrix not only influences the magnitude and the dura- tion of each heating step, but it can react with the atomizer and alter its physiochemical properties. The types of corrosive agents which both carbon and graphite do not resist are the strong oxidizing agents. Even in some cases of this kind, however, they may be found to be more serviceable than other atomizers. c. Influences of Heating Stages - Methods of Programmed Heating. In order to reduce the matrix effects and to allow some control over the atomization process, most nonflame atomizers utilize a two or three stage electrical current program for heating the atomization elements. These steps are utilized to desolvate, ash, and atomize the sample. The lifetime of the atomizer decreases as the current through the atomizer and the duration of each stage increases. The effect of current during the atomization step on the number of deter- minations with the GBA is shown later in Figure 51. For atomizers made from carbon or graphite, vaporization of the atomization element 199 is the major factor responsible for the decreased lifetime of the atom- izer. Vaporization results in increased electrical resistance. If the current through the atomizer is constant, as is the case with all programming systems reported so far, the atomizer temperature increases gradually and this can influence the long range reproducibility in the system. It is practically impossible to prevent the vaporization of atomization elements. However, other methods of programmed heating can be employed to compensate for the vaporization problem. Since these methods are shown to influence other parameters in the system as well as the atomizer lifetime, they are discussed in the next sec- tion. C. New Methods for Programmed Heating of Nonflame Atomizers 1. Introduction There are actually 5 different methods of programmed heating of nonflame atomizers, depending upon the electrical or physical parameters which are controlled during the heating steps. Most nonflame atomizers utilize a two or three stage electrical current program for heating the atomization elements. However, programmed heating can be accom- plished by controlling the voltage across the atomizer, the power dissipated in the atomizer, the radiation emitted by the atomizer, and the actual atomizer temperature. The instrumentation employed for various heating methods is similar in that all techniques use a sensing device and a feedback system for heating regulation. In this section the instrumentation required for the various techniques is discussed along with the advantages and disadvantages of each method. The method of programmed heating is shown to influence the atomizer 200 lifetime, the time required for the atomizer to reach a steady state temperature and the separation and optimization of atomization parameters. 2. Current and Voltage Programming In current-programmed heating, the power supply provides the atom- izer with the preselected amount of current during the various heating stages. The current is monitored by a sensing resistor, R s’ and cur- c rent regulation is performed by the feedback circuit described in Chapter IV. In voltage programming, the voltage across the atomizer is preselected and kept constant during each heating stage. The mag- nitude of the voltage, therefore, would be different for the 3 heating stages. The instrumentation for voltage programmed heating is trivial and will not be discussed here. In all programming operations reported here, a dc rather than an ac power supply was utilized for heating the atomizer. The ac evapora- tion rate of the atomizer material has been shown (278) to be greater than the dc evaporation rate under the constraint of either equal efL fective current or voltage. Both current and voltage methods of pro- grammed heating result in atomization temperature drift. This can be easily observed from Equation 60, which has been obtained assuming ohmic heating. .2 _ dT l R(T) - C(T)pS a? k k k T = —Sl(12R) = —s‘—(iv) = —$l(w) (60) where both the atomizer resistance R(T), and specific heat C(T) are func- tions of temperature, and p and s are the atomizer density and cross 201 sectional area. If current programming is employed as the heating method, gradual vaporization results in higher electrical resistance and smaller cross- sectional area for the atomizer. Thus, the atomization temperature will increase from run to run. For voltage - controlled heating, the increase in the resistance allows less current through the atomizer and the temperature decreases. The rate of atomizer evaporation, however, would be smaller than with current programming. These varia- tions can be multiplied if the atomizer or its holder do not have the same temperature at the start of successive experiments. Depending on the magnitude and the sign of the coefficient of electrical resist- ance, different changes may result as shown by Equation 61. R = R0(1+aT) (61) If the atomizer and its holder are made of the same material, a posi- tive value of a results in the intensification of temperature varia- tions, while a negative value may operate in the opposite direction. If the atomizer and the holder are made of materials with different a values, as is normally the case, other temperature variations can be expected. Therefore, the atomizer holder should be water cooled or a cooling period should be allowed between successive analysis. The variation of the GBA temperature as a function of number of the experiments was monitored by measuring the emission intensity of the GBA for atomization currents of 9 to 21 A. An increase in the emission intensity of the GBA indicates higher atomizer temperatures. A two step current program was employed for all the data presented in this section. Figure 48 shows the variation of emission intensity (“n .Ihrzwmm‘do ZO§IF OEmSEExM uO mmmzaz own ova 0mm 00— cup ow ov F _ _ L P _ <2 .HmemDO ZO_.F mUZwamume 3a «wigs? 220.2530 5.2.522 aO~ZZOO OOa thwEEm—QXw ".0 mum—232 00V own com on“ com cap 00— on .om mcsmca >>Om— .mm>>0a ZO;>nwp .mm>>0a ZO_._.< 5 In 400 .. L I 8 e O '7 CD a: 300 - :1 “J P6 _< m -5 2'1 5 > 3 2001 -4 -i Z 5 —-3 ,10()-- \"~ ..2 \.\ ‘— ‘*—-—l bl I 1 l J ATOMIZATION 9 12 15 18 CURRENT,A ATOMIZATION 95 130 185 250 POWER,W ATOMIZATION1485 I650 2100 2250 TEMPERATURE,°C Figure 51. Influence of different heating methods and atomization parameters on the lifetime of the GBA and the GBA emission intensity variation. 210 deterioration begins by forming a hot spot. The hot spot usually forms at a point which is either located between the inner and the outer sheath gas flow or at the contact points of the atomizer and its holder. 4. Radiation Programming According to Stefan‘s Law, the total rate of radiation emitted by a body increases with the 4th power of the Kelvin temperature. The radiation emitted by any electrically heated atomizer can be uti- lized for programmed heating. A radiation transducer such as a photo- transistor can function as a radiation sensing device. In contrast to the current and power methods, where the atomizer temperature is a function of its dimensions as shown by Equation 60, the radiation technique does not suffer from this disadvantage as long as the radia- tion transducer monitors a fraction of the total emitted radiation. Furthermore, the variation of sheath gas flow rate and sheath gas type should not influence the atomizer temperature in the radiation technique. This is an important step in the separation of parameters in nonflame atomization. When the current programming technique is employed for the carbon rod atomizer, it has been shown (284) that the rate of cooling water flow in the carbon rod holder influences the AA signal. Increasing the water flow has been shown to improve the sensitivity. This en- hancement at higher water flow rates has been attributed to an improve- ment in the atomizer contact with its holder. When radiation programming is employed, these physical changes should not affect either the atom- izer temperature or the AA and AF signals. Since the nonflame atomizers emit visible radiation only during 211 the ashing and atomization periods, other programming methods should be utilized for heating the atomizers during the desolvation period where there is no visible radiation. One possible instrumentation system for multistep heating of filament atomizer is shown in Figure 52. The power control technique is used for the desolvation period and the radiation control method is employed during the ashing and atomization periods. A two channel analog switch (06 152 AP, Siliconix Incorporated, Santa Clara, CA) provides the proper heating method during the various heating stages. One analog channel is active at all times. As soon as the desolvation period is complete, the computer supplies a +5V signal to the analog switch logic input for the entire length of the ashing and atomization stages. The analog switch disconnects the power sensing circuit from the control operational amplifier and simultaneously connects the radiation sensing circuit. The photo- transistor (type TIL 64, Texas Instrument, Inc., Dallas, TX) monitors the atomizer radiation and produces a corresponding voltage, which is amplified and applied to the control operational amplifier input. The DA compares this voltage with the reference voltage (radiation) and allows a current through the atomizer such that the radiation circuit output voltage equals the reference output voltage. When a furnace atomizer is used, the desolvation period is not often required because of the confined environment of the furnace. A two stage programmed heating technique is sufficient for the furnace atomizer to ash and atomize the sample. For furnaces, therefore, the radiation method can be employed alone. The required instrumentation is the same as that shown in Figure 25 if the instrumentation amplifier, the analog multiplier, and the analog switch are deleted from the 212 .mgmeEoum mac—Gcoc Go mcvuum; umEEmcmoaa cowpmwuma use smzoq nocwneou as» so» covumpc053gpmcr any mo Enamepc uwzuupu as» .Nm manure h mO 322.93.. . I7 4 In. 1.."I I4I My 33 . . ---n “ pasz_ _ _ :wmosww " 3:32: OOs6 0 P E Q 5‘ ago. an. 5.2 53 L 23532:» :23. $30.. sssssssse 330.. «3.102 fi ‘7 A ‘7 AV 5:: “n “n 5.2 \— Ay Av _ /_ u" ax. “nag. 3:35: . 20:4»2wiauhmz. 213 circuit. In order to test the radiation method with filament atomizers such as the GBA and hot wire loops, the phototransistor was mounted in a holder which could be assembled in the lens holder. The X, Y, and 2 positions of the phototransistor with respect to the atomizer could be changed by means of suitable adjustments. These controls could also function as an analog control over the radiation level (temperature). Because of the intense braid emission, light filters were employed to provide the proper emission intensity for the photo- transistor. The influence of radiation-programmed heating on the reproduc- ibility of atomization temperature and the lifetime of the GBA was investigated using a procedure similar to that described previously for other heating techniques. The ratio of the emission intensities was about 1.3 at 1485 °C, but the ratio was unity at higher temperatures. Compared to other programmed heating methods, it can be argued that when the GBA is heated by the radiation method, excellent temperature reproducibility can be achieved. With the radiation method, the GBA lifetime was found to be shorter than when it is heated by either the current or the power technique. The ratio of the average number of determinations using the power method to the average number of determinations performed with the radiation method varied between 2 to 4. This ratio decays as the atomizer temperature varies between l485° to 2250 °C. The shorter GBA lifetime when the radiation method is employed for heating can be explained in terms of the larger fraction of time that the atomizer spends in the final steady state temperature. 214 Figure 53 demonstrates the rate of radiation (temperature) increase for the three heating techniques. The atomization period is 1.56 sec. and the final GBA temperature is about 1650 °C. It can be noted that the steady state period is considerably longer in the radiation method. It is also interesting to examine the regulation of radiation (tem- perature) in the radiation technique. The duration of the steady state atomization period, ts, as a function of atomization current is shown in Figure 54 for the three heating methods. It can be seen that tS increases with increasing temperature in the current and power tech- niques, while for the radiation method the steady state periods de- creases with increasing temperature. Figure 55 shows the time duration, pre—steady state period, required before the GBA can achieve its final temperature. It can be noted from both Figures 54 and 55 that as far as the vaporization of the atomizer is concerned, the current and the power techniques should cause nearly the same amount of vaporization in a single ex- periment. When radiation programming is utilized, however, the atomizer spends a considerably longer time at its final temperature and a greater amount of vaporization should be expected. The atom- izer lifetime will, therefore, be shorter compared to other tech- niques. It should be noted that the length of the pre-steady state period is a function of maximum possible amount of current that a power sup- ply can deliver. The larger the current, the shorter the pre-steady state period. Furthermore, since in the radiation methods, the atomizer achieves its final temperature at a higher rate, atomization of the analyte should be complete in a shorter atomization period. The 215 POWER PROGAMMING TEMPERATURE °C,I650 CURRENT PROGAMMING TEMPERATURE °C,I650 RADIATION PROGAMMING TEMPERATURE °C, I650 Figure 53, Intensity of GBA as a function of time and as a function of three prograImIed heating methods .1. I <1. "III >A Iq 1“. IJ, 216 1000 - E g 800 F 01 < |_ E a) J 600 o RADIATION PROGRAMMING Ed I CURRENT PROGRAMMING <9 A POWER PROGRAMMING “Jo: "Lu 030. 400 P 200 - ATOMIZATION 9 I2 15 18 21 CURRENT, A ATOMIZATION 95 130 185 250 350 POWER,w ATOMIZATION I485 I650 2100 2250 2500 TEMPERATURE,°C Figure 54. Duration of the steady state period as a function of the atomization parameters for three programmed heating methods. -omwmuE. no.6 .00.N&Muhl MIPD_ E__ o RADIATION PROGRAMMING O a II CURRENT PROGRAMMING <( “ 1000- w d A POWER PROGRAMMING 5.. a, Q - E g g_ 800 .. Q 600 _ 1 l l l L ATOMIZATION 9 12 15 18 21 CURRENT,A ATOMIZATION 95 130 185 250 350 POWER,W ATOMIZATION 1485 1650 2100 2250 2500 TEMPERATURE,°C Figure 55. Duration of the pre-steady state period as a function of the atomization parameters for three programmed heating methods. HI 11 218 lifetime of the atomizer can be improved, if the atomization period is terminated shortly after the sample is completely atomized. 5. Temperature Programming In all the heating methods discussed so far, an electrical or a physical parameter proportional to the atomizer temperature was monitored in the programmed heating. Direct temperature regulation can be performed if a thermocouple is employed as a sensing device. Thermocouples lend themselves conveniently to the measurement of tem- perature profiles in flame and nonflame atomizers, but empirical cor- rection factors must be utilized to allow for radiation and conductance losses, and in any case they can only be used for low temperature flames. Catalytic heating on the wire surface may also cause errors in temperature measurements. For nonflame atomizers, where a violent reaction zone does not exist, a tungsten-rhenium thermocouple system may be used for temperatures from 1600 to 3000 °C. This thermocouple system has not been used in either flame or nonflame atomizers thus far, but it has been evaluated for use in aerospace and nuclear in- dustries (285). The stability of a 0.51 mm diameter wire thermocouple in an induction-heated vacuum furnace has been shown to be :2 percent for 15 hours at 2600 °C (285). When a nonflame atomizer is employed, even if an inert gas is used as a sheath gas, oxidation of the thermocouple can occur, and the thermocouple lifetime is expected to be shorter. Furthermore, attachment of any thermocouple to the atomizer presents a practical problem. Therefore, direct temperature measurement and regulation, despite its fundamental advantages, suffers from major practical 219 limitations. The instrumentation required for direct temperature control is similar to that employed in the current technique except that the thermo- couple output voltage is amplified and applied to the control operational amplifier instead of the voltage drop across the sensing resistor Rcs' However, in temperature measurements, the parameter of importance is the thermal time constant, Tth’ of the measuring device. For a very fine thermocouple, which is immersed in an environment of high thermal conductivity, Tth is about 1 sec or less. For nonflame atomizers, the temperature continuously changes with time and it is necessary to measure the temperature instantaneously. However, when a thermo- couple is subjected to a temperature step, its temperature approaches the steady state value after several time constants. The temperature growth, to a first approximation, is exponential for the thermocouple. The output voltage of the thermocouple V(t) as a function of time t is given by Equation 62 V(t) = Vo(l - e't/Tth) (62) where V0 is the thermocouple output voltage corresponding to the steady state temperature. Differentiating Equation 62 with respect to time, gives Equation 63 dV t 3 -t/T 1th t vo e th (63) which when substituted in Equation 62 provides the following equation = dV(t) Vo V(t) + Tth dt (54) 220 Theoretically, by combining the temperature at any instant V(t) with the rate of change of temperature, g§é£l, the ultimate temperature, V0, can be measured instantaneously. When a nonflame atomizer is heated, it can be assumed that the atomizer is subjected to an infinite number of temperature steps before the final steady state temperature is achieved. The ultimate output of the thermocouple for each step V01 is given by an Equation similar to Equation 64. If the resulting ultimate thermo- couple output for each step can be evaluated instantaneously, a stair 3 case waveform pattern which is proportional to the heat input to the n atomizer can be generated. This stair case function Z V01 can be i=0 utilized as the controlling variable for the input to the control operational amplifier. This also prevents temperature overshot which might occur because of the slow response of the thermocouple. Evaluation of V01 can be either performed by software or by analog circuitry. When a minicomputer is available, the thermocouple output voltage V(t) is amplified and applied to the ADC of the computer, where A-to-D conversions are made at a specified rate. A real-time calculation routine can be employed for evaluation of V01, and Z V01 is subsequently fed to the control operational amplifier. Since the computer should function as a digital integrator for the AA or AF signals, it may be advantageous that evaluation of V01 be executed by an analog circuit. An analog circuit for the determination of V0 has recently been described (286). The limiting factor in the evalua- tion of V0 is the electronic time constant of the circuit. Calculation of ultimate thermocouple output for each temperature step can be executed in 0.1 msec when the thermocouple time constant Tth is about 221 100 msec. 6. Influences of Flow Rates and Heating Methods on Atomizer Temperatures. The influence of the flow rate on the platinum loop under current- programmed heating was discussed in Chapter V. It was noted that the atomizer temperature decreases with the flow rate. The decrease in temperature is due to convective and conductive losses by the atomizer. Since only ohmic heating was assumed in deriving Equation 61, the ef- fect of these losses on atomizer temperature cannot be predicted from the equation. Figure 56 demonstrates the influence of three heating methods and sheath gas flow rates on the atomization temperatures of a graphite braid and a Pt loop atomizer. The vertical markings around each ex- perimental point indicate the standard deviation of 4 successive mea- surements. A ten second cooling period between measurements was used in all cases. The atomizer emission intensity was monitored instead of atomizer temperature in all cases. The GBA and Pt loop temperatures for an inner and outer sheath gas flow rate of 1 l/min were about 1650 and 1200 °C, respectively. When the radiation method is utilized, the atomizer temperature for both the GBA and the Pt loop does not change with the flow rate. Furthermore, the temperature measurements are reproducible. With both the power and the current techniques, the Pt loop temperature decreases with gas flow at nearly the same rate. Aflthough the power technique compensates for resistance changes caused by the flow rate variation, it cannot correct for convective and conductive heat losses in the system. When the current and power techniques are applied to the GBA, 222 I RADIATION PROGRAMING 110 } POWER PROGRAMING I CURRENT PROGRAMING '1'- 100 .. .._____..__ GRAPHITE BRAID ATOMIZER ................ PLATINUM LOOP ATOMIZER 50 I 40 F 30- ATOMIZATION EMISSION INTENSITY 201- 10- ‘5 l 2 3 4 5 6 7 8 9 10 FLOW RATE. l/min Figure 56. Influence of flow rate and heating methods on the atomizer temperature. . ‘3?! IDE 111 (n 223 the atomizer temperature appears to pass through a minimum, but in- creases when the flow rate becomes greater than 6 l/min. The decline in temperature at lower flow rates can be explained in terms of convec- tive and conductive losses by the GBA. The gas flow also causes a decrease in the electrical resistance of the atomizer holder. At higher flow rates, the holder resistance is further decreased and this allows more current through the atomizer with a subsequent increase in temperature. This should also explain the higher standard deviation at higher flow rates. Note that the precision of the atomizer tempera- ture, provided by the power control technique, is superior to the current control method. In both cases, the precision at higher flow rates should be improved when the holder is kept at constant tempera- ture either by water cooling or by allowing cooling periods of greater than 10 seconds. For the Pt loop atomizer, the current control technique seems to provide higher reproducibility, in terms of atomization temperature, compared to the power control technique. No explanation of this varia- tion in precision can be provided at this point. 0. AA and AF Analytical Applications of GBA In order to evaluate the graphite braid atomizer for AA and AF spectrometry, analyses of a variety of elements were performed and analytical curves and detection limits were obtained. A brief matrix effect study was also conducted and the atomizer was used for the analysis of iron and copper in serum to illustrate its potential applica- tion for biological samples. 224 1. Analytical Curves and Detection Limits Analytical data were obtained for 15 elements in aqueous solutions. The experimental conditions employed in the AF and AA analysis are summarized in Tables 13 and 14, respectively. The current programming method for heating the GBA was used in the early part of this work, but later the power control technique was used. Except for the atomiza- tion stage, where a rough optimization was performed, no other param- eters in the system were Optimized in obtaining these data. Analysis of Hg, Cd, Zn, and Pb were performed by atomic fluorescence using pulsed hollow cathode lamps or metal vapor discharge lamps as radiation sources. 1 The description of the pulsed hollow cathode lamps and the computer- controlled system is given in the next chapter. The detection limits for these elements are given in Table 15. All of the analytical curves showed good linearity over 1.5 - 3 orders of magnitude in concentration. A typical calibration curve for Cd is shown in Figure 57. The line parallel to the analytical line is an arbitrary unity slope line. Atomic absorption data were obtained for Cd, Zn, Tl, Pb, Cu, Mn, Mg, Sn, Au, Ag, Fe, Ni, Co, and Pt using dc hollow cathode lamps as the radiation sources. The detection limits for these elements are summarized in Table 16. Analytical curves were obtained for all the above elements except T1 and Pt, and all working curves exhibited good linearity over 1.5 - 3 orders of magnitude. The working curves for Au, Fe, Ni, Ag, Zn, Co, and Mn are shown in Figures 58 to 64. The relative standard deviation of measurements both in AA and AF was usually between 4-7% at concentration of one order of magnitude greater than the detection limits. Since all of the above data were 225 > 83 no: 3 Bug: «3323 a... Low Roux. 330 Zn 5 2.8 «on 83:; 333 E 3v Lava—.2332... 2:3 L28 9. 3 35:28.6 :- 5 .03 Z a... on: 9.3:. 2: A8 9.2 09.23:. .33» p3!- N._>x 3V 8.: 6853 8:2 8:8 36.: E 8.: 85 N w:— n 3: a N N355 05:3 - 8N, S. N .5 EN 8.: N.NNN m. P N mm: 3 N N 88 seats... - SN, 5. N to 3 8 :éN . NE SN 8... 8 N35: 22.3 N oNN dz. N Na 9. 8.: 353 N N. _ 82 o: N Names? 5.8.. s 8. 5.: N _ a: 8.: 8.2N N N; 82 or 2 £4.8ch .26.. N GN. 81.. N N EN 9. N.NNN N p 82 on. 8 N83 Loses N 08 8.: N 9° 3 3:: E E .53: 63 .8: u. .22 3 .538 E 5.82 853. 88 .8: E. 29:3 AN}: 5...: 5 .5282 3:88: one. In 59.03:: 35 c3328; c3338: c2323: 2 £26.. 395.6. 9.53: :0 3.58 3.58 338 33. BE can .35 «3322 5395.523: 9.3.5 533.5. c0333.. «8 535m 5053 «832.9 23?; A5532; 3:32:03: 0:83 5 32.8.3 395 3289a Go in 05 5 «Luau-lung $205,590 .2 03-... 2236 oom N > .oaas_o> N_aa=m Nza E.E\N .ouaa xoaa use gsasgm p.o.su .urpm Lauueocguocot can Lu~rsou< cooxuum mucuumNa —uu.ucu> ”mucosa—o F_m Lo» oeam as» «so: menquacoa we'roppow on» c mm Fo.¢NN N. o.p omNN omm PP m pucm Luzon op mo¢m¢az .zuu_3 prm coNNNNNE6u< coNNostoN< coNNNNNsou< .NENN .LOtba mpasom acpumo: ougaom ovozuau mvmxpoc< Louueoccuocoz ac.»co acwxgo covuo—umm soppoz xcumeocuumam coruaLoma< Orsopq cN LmeEoN< awoam mNNLQNLo Go xuaum on» :N mcuuoeocaa .Nucoewcoaxm .v— «pomp 227 Table 15. GBA Detection Limits in Atomic Fluorescence Element Radiation Source Detection Limit, pg/ml Cd PHCL 0.002 Zn PHCL 0.2 Hg PHCL 0.01 Pb PHCL 0.005 Cd MVL 0.005 Zn MVL . 0.005 PHCL = pulsed hollow cathode lamp. MVL metal vapor discharge lamp. 228 Table 16. GBA Detection Limits in Atomic Absorption Elements Detection Limit,IJg/m1 Cd 0.0005 Zn 0.0005 T1 0.04 Pb 0.01 Cu 0.01 Mn 0.006 Mg 0.002 Sn 0.1 AU 0.02 Ag 0.007 Fe 0.004 Ni 0.02 Co 0.004 Pt 0.9 NOZnu-CUHHQC-I uIIII fluIIhIIIIIII'l TIM E INTEGRATED FLUORESCENCE 229 0 fi 4 U C l 1() - l() ‘ 10 100 1000 CADMIUM CONCENTRATION, PPb Figure 57. Cadmium working curve (AF signal has arbitrary unit). TIME INTEGRATED ABSORBANCE 10’ WITITIII 10‘ IIITIIFI I 1() Figure 58. 230 1 00 1,000 10,000 GOLD CONCENTRATION, PPb Gold working curve (AA signal has arbitrary unit). “IIIVIIIa=u mcwxuoz am>_wm San. .ZO_._.Izm _ _. 3 bp-nb- .NG assmaa EONVQHOSQV OHLVHOBINI 3WI.L 234 .ANN== xgmauwnsm we; Pmcmvm <u=o mcwxaoz uch can .ZO_._.a=o mcwxcoz upmnou .mm mgzmwm SEQ . ZO_._.a=u Pmuwuxpmcm cowpvuv< uamucmum .mo maamwa _E\mi .Enmmm 09:35 07.. z. mummoo mo zo_._.mz an .sacmmpu Feamcwm Am Empmam 2n» PmccmcuIe .oo mazmwa . .. < < o O N m A. 932.6%. mmuanom 20.23.92. 20.25.92. , .I _ - — o R o v . o _ 5:26 205% - - H 0 _ 0 56:52.23 552522 6 .7 0 _ — mu .— .O .01, —. mm — - . . N. I E - _V 1‘. ~1 no nVII # 4‘. _ 249 channels is usually rather easy to minimize with the TDM technique. The only requirement for negligible interchannel crosstalk is that the transmission system must have sufficient bandwidth and linearity in order to prevent the pulse waveforms from overlapping into adjacent time slots. The major disadvantage of the TDM technique is that the information is transmitted sequentially rather than simultaneously as in FDM systems. However, the simplicity of recovering the individual infor- mation channels in a TON system can often make it nearly as fast a transmission method as an FDM system. 2. Multichannel Atomic Fluorescence Spectrometry For atomic fluorescence spectrometry, time-division multiplexing can be achieved in the optical domain by sequentially pulsing several hollow cathode lamps in a low duty cycle mode. Figure 67 shows a general block diagram of a computer-controlled, 4-channel, TDM atomic fluorescence spectrometer. Each hollow cathode lamp is focused on a flame or nonflame atomic vapor cell. The minicomputer controls the circuitry for pulsing each hollow cathode lamp so that the ON and OFF times of all lamps can be controlled by software for optimization purposes. If a nonflame atomizer is utilized, the computer also con- trols the programmed heating of the atomizer. The fluorescent radiation from each element is transmitted through the same optical path, but separated in time. The optical signals thus appear as amplitude modulated, time-division multiplexed pulses of radiant energy. A solar blind photomultiplier tube (PMT) is used to transduce the radiant energy pulses into electrical current pulses. 1‘... .Lmuweoguuwam m< Poccngu*p_:e ump_ocucou-cwu=asou mo sogmawu xqum .um mcamwu «052032. 5:226. 55202 250 3029525 62 m 1'4 mz<:.zoz 4L m0 924: .IIlj . » 0.... [fl m 7 .III mmufiaflz. «Eizou .33 , IL 30120 TII! 30:0: in fianfiw «mp/0a mmkDmIOU J a} mi 251 An analog integrator, synchronized to the time slot of each channel under computer control, integrates and holds the PMT output from each fluorescence pulse. During the hold time of the integrator, an analog- to-digital converter (ADC) produces a digital representation of the output of the integrator. Demultiplexing of the information signals is accomplished by the minicomputer, which directs the ADC outputs for each information channel E into separate locations in core memory. The cycle of multiplexing, transmission, transduction, and demultiplexing is then repeated for a preselected number of times (usually 20) to improve the signal-to- ‘fl; noise ratio in each information channel. With the present system 4 l. elements can readily be determined in less than 1 sec and 8 elements in less than 3 seconds with flame atomization. 3. Instrumentation a. Atomization Systems. Because the system operates without wavelength dispersion, a low background atomizer is necessary. Flame and nonflame atomizers were employed throughout this work. A separated air-hydrogen flame was used with a burner system similar to that des- cribed originally by Larkins (48). The burner was modified (47) by Eugene Palermo in these laboratories, and consisted of two parallel stainless steel plates which could be mounted on a Jarrell Ash pre- mixed burner. The top plate has two circular arrays of holes, and fuel, oxidizer and sample droplets pass through a hole_in the bottom plate and up to the inner circular array. Argon sheath gas enters the bottom plate through tygon tubing and a channel, and passes through the outer circular arrays of holesin the top plate. With this design, 252 the sheath gas surrounds the flame without coming in direct contact with the fuel, oxidizer, or sample droplets. The nonflame atomizers consisted of a 90% - 10% Pt - Rh wire loop and the graphite braid atomizer (GBA) described in Chapter VI. The atomizer was mounted slightly below the optical system as described in Chapter IV. For all the multielement studies, the atomizer was held in place by a holder from the bottom rather than from above to decrease scattered radiation. Gas sheath design 11 was also utilized throughout this work, and the Ar flow rate was kept constant at 2 l/min. A 4 u] sample was placed on the atomizer by a syringe or the automatic sampler. Since only aqueous samples were analyzed, a two-step heating program, using either current regulation or power regulation, was employed for heating the wire loop or the GBA. b. Radiation Source and Computer Interface. For the atomic fluorescence measurements reported in this chapter, hollow cathode lamps operated in the pulsed mode were utilized. The radiation from each lamp was focused on the vapor cell by a one inch diameter plano- convex and biconvex quartz lens (Esco Optics Products, Oak Ridge, NJ). Four lenses were employed to focus the radiation of four HCL's on a common point. In order to perform a simultaneous or nearly simultan- eous multielement analysis, all of the radiation sources must be operated at the same time. If a single channel detection system is to be employed, a mechanism of distinguishing the fluorescence signals of the various elements is required. This may be accomplished in one of two ways. The first method involves simultaneous modulation of radiation sources at different frequencies. The fluorescence inten- sities could be determined by transforming the amplitude-vs-time 253 spectrum to an amplitude-vs-frequency spectrum by the Fourier trans- formation technique. Another possibility is the TDM mode where all the radiation sources are Operated at the same frequency, but out of phase. With a minicomputer, the latter method is more attractive for rapid multielement analysis. The circuitry for pulsing the hollow cathode lamps was similar to that described earlier (47,241) except that the ON and OFF times of the lamps were controlled by the minicomputer. The circuit for operat- ing four lamps in a pulsed mode is shown in Figure 68. An Operational amplifier controls the voltage applied to the base of a driver tran- sistor. A current booster amplifier (EU-QOO-CA, Heath Co.) was used in conjunction with the control 0A when the operational amplifier was incapable of providing the required current. The feedback control for current regulation was obtained by connecting the feedback resistor, Rf, to the emitter of each transistor. The current through each lamp, iL, is given by the expression in Figure 68, and could be individually adjusted for each lamp by varying the corresponding input resistor, R The lamp current was experimentally monitored by measuring the in' voltage drop across the resistor Rs' The FET switches (8-bit analog switch, Model EU-900-JA, Heath, Co.) enable the power supply to be operated in an intermittent mode by switching the reference voltage, er, ON and OFF. The FET switch with the A381C15 control signal was used to ensure complete turn off of the radiation sources when they were not in operation. The pulse width or 0N time of each lamp was usually 2-lO msec, and it was controlled by the computer. This information was trans— ferred through four channels by the computer interface shown in Figure 254 .muos Zoe cw muse, muozumu zo__o; Lace mcwmfiza Lee p_:oewu .mo mesmem C - .m . _m ..._ aw _ Becflufi . . \ V . 98? em . < c38loorsm >u... cm ‘ To €31 [0, - Q- -623: ”I O ‘( UL. em 3 \2an o _ 0+U+m+< 255 .mnos zap c. m..u: esoe m=.m.=a eoe mumegmpcw emuzasou .mo mann.. IU...>>m hm“. .252. «0.7.1032. (“U0 0. 0mwo W T A AI L.) 1.) l3 4) <<<.. wOOI....<.. 92: #53. as :8 _l_ T 92: to v _l_|< n=2<.. m2: m2: 1 I 20 20 258 was also connected to a photomultiplier power supply (Model EU-24A, Heath Co.). The phototube output current was converted to a voltage by a Keithley Model 427 current amplifier. In order to average the AF information present during each time slot, an integrator operated in synchronism with the firing of the hollow cathode lamp was employed. The integrator and its computer interface are shown in Figure 7l. The output of the current-to-voltage r“ converter was connected through a FET switch to the OA voltage integrator. ; A FET switch was also employed to discharge the capacitor in the integrator feedback loOp. The timing pulses from the computer, which cause each lamp to .I‘F' -‘1 ‘....—... turn ON, also control the FET switch at the input of the integrator. The sequence of operation for one cycle is shown in Figure 72. When the first lamp is turned ON, the integrator is also turned ON for a preset time. The lamp is then turned OFF by a command from the com- puter, and the integrator input is opened, which causes the integrated signal to be held for a period twice the lamp ON time. During the integrator hold time, the ADC is enabled, and the resulting digital signal is stored in one location in the core memory. The integration capacitor is then discharged, and a program-controlled delay time equal to the lamp ON time is begun. Then lamp 2 and the-integrator are turned ON, and the digitized signal for the second element is stored in a second location in core memory. The above process is repeated for the remaining lamps, which ends the first data cycle. Thus, for a 4-channel system, the result of one cycle is the storage of 4 digital representations of the AF information in separate core memory locations. The process is then repeated for a preset number of cycles (usually 20), 259 .eopmgmmucw msocoecuc»m one Low mummemuc. emuaaeoo .Pu mesa.“ . 5.50 5:928 mm.) 2 m_u “959, 533 338 IAUh limezau a: 5 K “ I o nxu( nth! a5 /\ nfixv+o+a. v 92 185.53 . 8.. men new ecumemmpc. mnocoggucxm .m..u: www.3a mcp mo cowpmemqo on» we we.“ mpuau .Nn mezmwm = .= = 00$“. 02 .1 LII/l 5330 5255.2. 260 20 10.... >>m w0m<10m5 hm”. . l , , l - o+o+m+< to C C .l_ .8. 261 which represents one experiment. The computer then enters an integra- tion routine and the integrated AF signal for each lamp is printed on the teletype. Under program control the entire process can be repeated for a preset number of experiments, after which average values and standard deviations are calculated. In the case of nonflame atomization, each experiment corresponds to a repetition of the processes of sample deposition on the atomizer, desolvation, atomization and acquisi- é tion of the AF information. d. Sequential, Dispersive AF system. For comparison of the non- a dispersive, TDM AF system to a sequentially-operated dispersive AF system, a monochromator (Model EU-700, Heath Co.) was inserted into 1:« the system, and the software was modified to allow sequential pulsing of the hollow cathode lamps. No change in the pulsing circuitry hard— ware or the integrator was required. One hollow cathode was pulsed for a preset number of cycles, the monochromator was scanned to a new wavelength, the next lamp pulsed for a preset number of cycles, etc. It should be noted that with the TOM system, the total analysis time for N elements is reduced by a factor of ”l/N compared to the sequen- tial system because of the interlacing of pulses from other lamps during the OFF time of any one lamp. The sequential method, however, allows the optimization of parameters between the determination of each element. e. General Software Description. The software was written in assembly language (Macro-8), and it is assumed that the fixed volume sampler is utilized in conjunction with a nonflame atomizer. When. a flame atomizer is employed, the sample delivery, the desolvation 262 and ashing stages are deleted in the initialization step. The lamps were operated only during the atomization step. The length of the atomization period is determined by the lamp ON time, tON’ the duty cycle, dc and the number of data points (cycles) dp requested as shown by Equation 65. .. .__1_, Atomization period — t0N dc dp (65) For example, for a lamp ON time of 5 msec and a duty cycle of l/16, the atomization period would be l.44 sec if 18 data points are obtained for each lamp. When a nonflame atomizer is utilized, because of the transient nature of the signal, the digital data should contain as much of the relevent information in the original signal as practicable. What- ever definition of information is used, if the digital data can be converted back into an analog signal that can be exactly superimposed on the original signal, then the digitized signal is an accurate representation of the original signal. Any difference between the two signals may be called digitization error. For instance, in order to have an error of 0.01% in the peak height measurement, 630 and ll data points are required for representation of a Lorentzian and a Gaussian signal, respectively (290). For single element analysis, the nonflame transient signal is approximately Gaussian in shape. If multielement analyses of mixtures containing low boiling points and high boiling point elements are to be performed, it should be ex- pected that the signal from the low boiling point elements would exhibit more of a Lorentzian than a Gaussian shape. In order to keep the atomization period constant and at the same time minimize 263 the digitization error, t0N can be decreased. For example, for a t0N of 2 msec, 45 points could be obtained to allow a 1.44 sec atomization period. Optimization of the lamp ON time is not required for a flame atomizer, because of the continuous nature of the signal. After the data acquisition is complete, the atomizer is shut off and the integrated signal for each element is printed on the teletype. The program then enters a display routine, and the AF signal correspond- ing to the first hollow cathode lamp is displayed on an oscilloscope for visual observation. Under keyboard control the data for the other three elements can also be displayed. The actual time, relative to the start of atomization, that the atomic vapor for each element is in the observation window varies according to the boiling point of the element. In the analysis of a mixture containing Hg, Cd, Zn, and Pb, the AF signals appeared in time as the elements are arranged. However, there was overlap of the peaks which would rule out the possibility of sequential operation. The display routine can be ended by typing the character "R" and the program starts a new experiment. After the preselected number of experiments is completed, the average, the standard deviation and the percent relative standard deviation are printed on the teletype. The program then enters the Optimization or the Normal submode as described in Chapter IV. C. Analytical Results l. Burner Parameters Among the parameters Optimized in flame studies were the position 254 .muoe zap :. muse. muocpmu zo._o: Lao; mcwm.:q Lo. p.3uewu .mo mezmwe .IIIIL . 4. _ . a . LE 3 xanv o o+u+m+< aw; . c_mm III. A...u.m&. Lmvnwl. _ 3.... are? .m. 962" .m 9.8.702 u em. 2..» L OD 255 IU...>>m kw“. .552. «0.21032. A .muoe 29? c. m..o: eaom mcwmpza Low mumeemucw ewuaqsoo .mo meamwm 970 5:: 2 83mm 533 SE «hm All-ISO. mm Al N w 5 ...Uw.wm mU.>mo M AT o u AF a < Ar .. i .3 .L as}... max—(U 30.... O... (“M 256 69. The computer was programmed to provide input-output instructions 63ll to 634l with a software selected time pattern. The Device Select signals 31 to 34 were connected to the Data Latch card inputs, and the corresponding Data Latch card outputs provided ON signals for the radiation source FET switches A, B, C, And D and the integrator, to be described later. The input-output instruction 6l5l was also utilized to turn the ON lamp OFF. 5" In order to transfer this information, the Latch card required ; two gating signals. The Device Select signals (3l-34) were also con- nected to OR gate l and its output along with Device Select l5 was I .1 "-1 J sent to the OR gate 2. The output of OR gate 2, the IOPl were utilized as the gating signals for the Data Latch card. This arrangement provided great versatility in optimizing the ON and OFF times, which was par- ticularly important when measuring transient signals from nonflame atomizers. For both flame and nonflame atomization a duty cycle of l/lG was utilized. The ON times of the lamps were 2-lO msec for both systems. These times were often varied to account for different atomization rates for different elements with nonflame atomizers. The lamps were operated only during the atomization period. The delay time between pulses for successive lamps was usually three times the ON time. A typical sequencing diagram for a 4-channel system is shown in Figure 70. c. Detection System and Computer Interface. In order to detect the fluorescence radiance, a Rl66 solar blind photomultiplier (Hamamatsu Corp., Lake Success, NY) was utilized as the radiation transducer. The phototube assembly was mounted directly in front of the vapor cell at an angle of 90° with respect to the radiation sources. The PMT .muoe zoe as“ c. m..u= umm_=g Lao. .0 as.» apuao .ou mesm.. .mcw .VMa. : C D as}: , .nev sage F.. Fl. ”nu Au.)alx.. m. w as .38 C C m as...) ms=h .238. as :8 .I. T 92: to v _l_.l < n..).<... m<<=. m.wn. E30 ume>ZOU H um .i 3459 A \ I...2w~.~.DU III/IL ex». mw11, TA 0 u o+u+n+< a. 92m :85 . <55 .FA mann.. 260 chat-5311!. :9. u.4.............r . . .oo< mg» use Louaemmpc. mzocoecucxm .m..u: campaa ms“ mo :o.pmewao mcp mo we.» m_uxu .Nu mg:m.. I: = .= = 00.5.. 09 IfiI/l SE20 mofimoflz C Fm E CE: 20 IO....>>m w0m<10m5 hm“. L c C E ZO 261 which represents one experiment. The computer then enters an integra- tion routine and the integrated AF signal for each lamp is printed on the teletype. Under program control the entire process can be repeated for a preset number of experiments, after which average values and standard deviations are calculated. In the case of nonflame atomization, each experiment corresponds to a repetition of the processes of sample deposition on the atomizer, desolvation, atomization and acquisi- , .“ iii: tion of the AF information. i d. Sequential, Dispersive AF system. For comparison of the non- dispersive, TDM AF system to a sequentially-operated dispersive AF T1 :71 m—-.. system, a monochromator (Model EU-700, Heath Co.) was inserted into l the system, and the software was modified to allow sequential pulsing of the hollow cathode lamps. No change in the pulsing circuitry hard- ware or the integrator was required. One hollow cathode was pulsed for a preset number of cycles, the monochromator was scanned to a new wavelength, the next lamp pulsed for a preset number of cycles, etc. It should be noted that with the TDM system, the total analysis time for N elements is reduced by a factor of ”l/N compared to the sequen- tial system because of the interlacing of pulses from other lamps during the OFF time of any one lamp. The sequential method, however, allows the optimization of parameters between the determination of each element. e. General Software Description. The software was written in assembly language (Macro-8), and it is assumed that the fixed volume sampler is utilized in conjunction with a nonflame atomizer. When. a flame atomizer is employed, the sample delivery, the desolvation 262 and ashing stages are deleted in the initialization step. The lamps were operated only during the atomization step. The length of the atomization period is determined by the lamp 0N time, tON’ the duty cycle, dc and the number of data points (cycles) dp requested as shown by Equation 65. .. ._1. Atomization period - tON HE- dp (65) For example, for a lamp ON time of 5 msec and a duty cycle of 1/16, the atomization period would be 1.44 sec if 18 data points are obtained for each lamp. When a nonflame atomizer is utilized, because of the transient nature of the signal, the digital data should contain as much of the relevent information in the original signal as practicable. What- ever definition of information is used, if the digital data can be converted back into an analog signal that can be exactly superimposed on the original signal, then the digitized signal is an accurate representation of the original signal. Any difference between the two signals may be called digitization error. For instance, in order to have an error of 0.01% in the peak height measurement, 630 and 11 data points are required for representation of a Lorentzian and a Gaussian signal, respectively (290). For single element analysis, the nonflame transient signal is approximately Gaussian in shape. If multielement analyses of mixtures containing low boiling points and high boiling point elements are to be performed, it should be ex- pected that the signal from the low boiling point elements would exhibit more of a Lorentzian than a Gaussian shape. In order to keep the atomization period constant and at the same time minimize 263 the digitization error, t0N can be decreased. For example, for a t0N of 2 msec, 45 points could be obtained to allow a 1.44 sec atomization period. Optimization of the lamp 0N time is not required for a flame atomizer, because of the continuous nature of the signal. After the data acquisition is complete, the atomizer is shut off and the integrated signal for each element is printed on the teletype. The program then enters a display routine, and the AF signal correspond- ing to the first hollow cathode lamp is displayed on an oscilloscope for visual observation. Under keyboard control the data for the other three elements can also be displayed. The actual time, relative to the start of atomization, that the atomic vapor for each element is in the observation window varies according to the boiling point of the element. In the analysis of a mixture containing Hg, Cd, Zn, and Pb, the AF signals appeared in time as the elements are arranged. However, there was overlap of the peaks which would rule out the possibility of sequential operation. The display routine can be ended by typing the character "R" and the program starts a new experiment. After the preselected number of experiments is completed, the average, the standard deviation and the percent relative standard deviation are printed on the teletype. The program then enters the Optimization or the Normal submode as described in Chapter IV. C. Analytical Results l. Burner Parameters Among the parameters Optimized in flame studies were the position ' a 6.1". .‘M Anhx‘m . I. 264 of the burner with respect to the observation window, and the flow rate of the sheath gas in both dispersive and nondispersive systems. Burner position profiles obtained in the dispersive mode dictated that the burner should be placed as close to the observation window as possible for the maximum AF signal. However, in the nondispersive mode, the burner should be positioned as far away from the observation window as possible to reduce flame background. Thus for nondispersive fay operation, a compromise has to be reached to obtain high signal-to- j noise ratios. The burner was positioned at 2 and 5.25 inches below the observation window in the dispersive and nondispersive modes, respectively. ng In order to determine the influence of sheath gas flow rate on the fluorescence intensity and background signals, flow rate profiles were obtained with the optimum burner position. The argon flow rate profiles for Ni, Mg, Co, and Fe are shown in Figure 73 and 74. The signal-to-noise ratio and the fluorescence intensity initially increase as the flow rate increases. This is due to the fact that the sheath gas renders the concentration of atoms more uniform, reduces the quench- ing effects of atmospheric nitrogen, and reduces the flame background by separating the flame into its primary and secondary zones. Further— more, the atomization efficiency should also improve because the sheath gas reduces the diffusion of atmospheric oxygen into the flame and subsequent oxide formation. As the flow rate increases further, the flame temperature decreases and the fluorescence intensity declines. The optimum flow rate was taken at 15 l/min for Mg and 20 l/min for Ni, Co and Fe. Flow rate profiles were also obtained in the nondispersive mode. - w w ——.._. A INTEGRATED AF SIGNAL 265 L 10:- -- A Ni 0 M9 .. 1'0 W 2'0 3‘0 FLOW RATE, I/min Magnesium and nickel fluorescence intensity (arbitrary unit) as a function of argon flow rate. Figure 73. am 266 A Co 50+ 0 Fe .1 ‘1 < 2 S2. 0) 30% UL < O )1.” < 4 o: 0 LL] p.. :2; .u. 10i To 26 3'0 FLOW RATEJ/min Figure 74. Cobalt and Iron fluorescence intensity (arbitrary unit) ‘as a function of argon flow rate. 267 However, measurements could not be made at low flow rates, because the flame was not effectively separated, and hence the flame background was too great. An argon flow rate of 35 l/min was utilized for multi- element analysis. 2. Pulsed Source Characteristic One of the advantages of using pulsed hollow cathode sources is that the increase in current during the ON time is accompanied by a corresponding increase in the emission intensity, while the average current over a full cycle is still maintained at a safe operating level. The exact relationship between the intensity, I, and the current, i, has been given by L'vov (9) I = a 1'" (66) where a and n are specific constants for each combination of cathode material and filler gas, respectively. If self-absorption occurs in a lamp, there will be a deviation in the relationship between I and i, which will appear as a deflection of the slope toward the horizontal axis in a plot of Log I vs Log i. Once n (i.e., the slope of the line) for a lamp has been calculated, the gain in intensity, for the same average lamp current, may be determined from the following equa- tion Ipulsed g (ipulsed)n (67) Idc idc For the experimental data presented, peak currents of normally 200 to 300 mA were employed. Since the duty cycle was 1/16, the corres- ponding mean currents were 12 to 18 mA. Since the atomic fluorescence 268 sensitivity and the detection limit are proportional to the intensity of the source, maximum source current should be used without arriving at the point of self-reversal. The maximum current at which lamps can be operated was determined by monitoring the current through the resistor RS on an oscilloscope. As long as the lamp is not overloaded, the lamp current follows the shape of the square pulse which has been applied from the computer. An overloading of the lamp appears as a distortion of the square shape of the pulse during its last part. The gain in intensity shows a definite dependence on the type of lamp and filler gas. The gain ranged between 12 to more than 200 as reported by Cordos and Malmstadt (240). This is, however, achieved at the expense of the dispersion of radiation as we have noticed in our laboratories. Since the radiation is dispersed, it is not possible to focus the radiation on the vapor cell to take the full advantage of all the available radiation. Dispersion of radiation results in AF detection limits which are not as low as predicted by the gain in intensity. When the lamp ON time is decreased, the problem is minimized, however. With the present source power supply, it was not possible to achieve lamp ON times shorter than 2 msec. The effect of pulse width on lamp intensity was also investigated for Hg, Cd, Zn, Mg, Ni, Co, Pb, and Fe lamps. Pulse widths of l to 5 msec were employed, and results for a Zn lamp are listed in Table 20. Except for a pulse width of l msec, the intensity appears to be independent of the lamp 0N time. Similar results were obtained for other lamps. The stability of the lamp radiances was determined by introducing a metal cylinder between the sources and the detection system to produce a scattered light signal. For a pulse width of 5 269 Table 20. Relative Lamp Intensity as a Function of ON Time Relative Intensity, “ ..__.( \.___.4 .. _—__ ON Time Arbitrary Unitsa l msec 200 2 msec 245 3 msec 265 4 msec 265 5 msec 274 aZn hollow cathode lamp. 270 msec, a duty cycle of 1/16, and 20 total pulses, the relative standard deviations of the scattering signals ranged from 0.67 to 0.85%. 3. Flame Analytical Results In order to test the computer-controlled system, the analysis of mercury and cadmium was performed, and the detection limits obtained with this system were found identical to those obtained with a hard- ware-controlled system designed in these laboratories (47). Detection limits and working curve were then obtained for Mg, Ni, Co and Fe. The calibration curves were linear over l.5 orders of magnitude and all showed a slope of unity. Table 21 summarizes the detection limits i Q':, ' obtained with the flame atomizer in the sequential, dispersive mode. Similar results were obtained for TDM operation. Extreme care should be exercised in focusing the sources on the atomizer in multielement analysis. Inferior results are obtained if the radiance reaching the vapor cell is smaller than the radiance in sequential analysis. The detection limits obtained for Co, Ni and Fe are not unreason- able considering that these are high boiling point elements, 2900 °C, 2732 °C, and 3000 °C respectively. However, the detection limit for Mg appears quite high considering the fact that its boiling point is only 1107 °C. The discrepancy may be explained by the fact that the gain in intensity of the Mg hollow cathode lamp is only 12 when operated at a peak current of 220 mA compared to dc operation as reported by Cordos and Malmstadt (240). This lack of intensity increase is probably due to the ease of self-reversal in the case of Mg. The solutions that were used for the working curves were composed of mixtures of the four elements. To determine if any interelement 27l Table 21. Detection Limits Obtained with Computer-Controlled Multi- element Flame A.F.S. System. Optimum Ar Detection Peak Flow Rate Limit Elements Current (t/min) (ppm) Hg 265 O 2 Cd 200 20 .02 flex Mg 200 15 1 fl Ni 275 20 10 E Co 290 20 2 g P Fe 330 20 20 “ “IL—1.4.. . 272 effects were present, the fluorescent signals obtained for the mixture solutions were compared to fluorescence signals obtained for solutions containing only the single element. All solutions were made from the chloride salts. The data obtained are presented in Table 22. As can be seen only slight interelement effects were obtained for these elements. 4. Nonflame Analytical Results With a 90% Pt - 10% Rh wire loop atomizer wirking curves were determined for four elements, Cd, Hg, Zn, and Pb in the dispersive mode. For these working curves, the atomizer was positioned directly below the entrance slit of the monochromator, and the radiation from the source was focused directly above the atomizer. The detection limits obtained with the loop atomizer are shown in Table 23. For Cd and Hg, the detection limits were lower than those obtained with the flame system. This decrease in detection limit is due to the fact that background noise from the nonflame atomizer is less than the background noise from the flame atomizer. Because the Pt loop temperature cannot exceed 1500 °C, the detection limit for Pb is high. Detection limits were then obtained for Cd, Hg and Zn with the nonflame, TDM, AF system. These detection limits are also shown in Table 23. The detection limits for Cd and Hg are comparable to those obtained in the dispersive mode. However, the Zn detection limit is higher. This increase is due to the fact that the atomizer position had to be lowered relative to the observation window because of the radiation from the atomizer and because of scattering of the primary sources caused by the dispersion of radiation. As the loop is lowered, —-—-E..—-—-_~— 273 Table 22. Fluorescence Signals Obtained with Multielement and Single Element Solutions Average A.F. Signal, Arbitrary Units Concentration Element (ppm) Mixture Single Ni 50 123 130 200 326 287 *""" 1000 662 657 CO 50 311 419 200 600 718 1000 1402 1317 Fe 50 400 ~ 386 200 719 600 1000 1728 1540 Mg 50 550 403 200 1271 1232 1000 2230 2557 _- “—- 274 Table 23. Detection Limits Obtained with Platinum Loop and Graphite Braid Atomizer Sequential TDM Dispersive Nondispersive Element Atomizer pg/ml. pg/ml. Cd Pt loop 0.008 0.005 Hg Pt loop 0.5 0.5 ft Zn Pt loop 0.1 2.0 ! Pb Pt loop 50 ; Cd GBA .002 , Hg GBA .01 lg, O 0 Zn GBA 0.2 O Pb GBA .005 275 the temperature decreases drastically. Hence the atomic population decreases. The nonflame TDM system may be used for multielement analysis. However, the detection limits obtained for certain elements will be inferior to those obtained for the single element dispersive case. Analytical results were also obtained with the GBA in the dispersive mode under the experimental conditions described in Chapter VI. It can be seen from Table 23 that the GBA detection limits are at least equivalent or superior to those obtained with the Pt loop atomizer. 5. Application of Fiber Optics in Atomic Absorption and Atomic Fluores- cence Spectrometer In order to perform simultaneous or near simultaneous analysis, it is necessary that the radiation from various radiation sources be focused to a common point on the vapor cell. As the number of lamps is increased, the spatial problem becomes more and more critical. Even for four HCL's, we encountered extreme difficulty in focusing the radiation. This was particularly true when filament-type nonflame atomizers were utilized as atomization devices. In contrast to the flame where the atomic population is distributed over a larger cross sectional area, the atom population in nonflame filament atomizers is concentrated in a small area. Focusing of radiation sources is there- fore more critical with nonflame devices. Fiber optics may be utilized in reducing the spatial problem. It has been used in atomic emission (291) for simultaneous measurement of two adjacent wavelengths, but no application of fiber optics have been reported for simultaneous analyses, by either AA or AF. We have 1 n-H no ‘2 t-.' i ' 3! —~—hm 276 designed a four branch light guide for multielement analyses by AA and AF. The four branches are combined at one end. The radiation from four different radiation sources is focused on four separate branches, 6.3 mm in diameter each, of the light guide and is directed to a single beam spectrometer. The total length of the light guide is about 40 cm, with the common end having a diameter of 1.27 cm. This ‘ 1 type of arrangement in conjunction with pulsed hollow cathodes allows sequential multielement analysis by AA if a rapid scanning monochromator " I‘-<'V.-o“fl, _ is available. For AF analysis, where the dispersing element can be eliminated, multielement analysis can be performed in the TDM mode. The only dis- .1- advantage of fiber optics is its low radiation transmission. Our initial results (292) indicate that the transmission is about 10 percent of the incident radiation. This system is currently under more extensive investigation in our laboratories. VIII. PROSPECTIVES A. Improvements in Instrumentation Various improvements can be made in the sample handling, atomiza- tion, and excitation systems. The second automatic sampler should be i made faster and more versatile in terms of the number of samples which i it can handle and the capacity of the sample cups. Without major modifications, the sample turntable can be replaced with one that has i 20 sample cups instead of 6. The sample cups would, therefore, have .3' a smaller sample capacity which is valuable for clinical analysis. The sample turntable motor, however, should be replaced with a stepping motor to allow more accurate positioning of the sample cups. For complete automation, the variation of the sheath gas flow rates and the three dimensional positioning of the atomizer should also be automated. One possible method for electronic flow measurement and control is to heat the gas stream uniformly, before it reaches the atomization chamber, by a heater coil and to monitor the gas flow at upstream and downstream positions by two sensors. When there is no gas flow, a balanced bridge circuit is established to provide a zero flow output signal. As the gas flows, a temperature difference is created between the upstream and downstream sensors, which is related to the mass flow rate. The resulting signal can be applied to the input of an instrumentation amplifier, and the amplifier can generate a dc voltage output which is related to the mass flow rate. The output 277 -I—-‘~.w 278 signal may be utilized to drive a motor for flow adjustments. For accurate atom profile studies, automation of the three dimen- sional positioning of the atomizer is desirable. This can be accom- plished by a combination of stepping motors which are operated under computer control. Various improvements can also be made in the radiation sources. For pulsed operation of HCL's, a new power supply should be designed which allows pulsing of the lamps on the usec time scale and which can provide larger output currents. The shorter the lamp 0N time, the easier the focusing of the radiation sources. Furthermore, we predict that operation in usec time intervals should diminish the possible self-absorption problem and should improve the spectral line intensity. As far as the dimensions of the HCL's are concerned, attention should be directed in miniaturizing these sources for multielement analysis. Fiber Optics can also be utilized where spatial problems exist in focusing the radiation sources on the vapor cell. 8. Future Application of the GBA It is obvious from the work presented here that the GBA should prove to be an excellent nonflame atomizer for the analysis of solu- tions. The low power requirements of the GBA and its high atomization efficiency suggest that the GBA should have widespread application in AA and AF spectrometry. It is interesting to note that the cost of each GBA is less than 1 cent. The GBA cost is so small that it is incomparable to other commercially available atomizers. The atomizer should be tested for more elements, and its capabilities in real 279 sample analysis should be further demonstrated. A complete matrix effect study should also be undertaken. The matrix effects, if present, can be minimized by surrounding the atomizer with a H2 diffusion flame. For GBA's of about 0.5 mm in diameter, it was shown in Chapter VI that only lZOliare required to reach atomization temperatures of 2500 °C. The capability of this small diameter graphite filament atomizer (GFA) should also be studied. For direct analysis of solids, the GBA E can be replaced with a graphite tape atomizer (GTA) of a width of about i 1 cm. Solid samples can be directly placed on the center of the GTA. This atomizer should also be valuable for the analysis of solutions. ‘.- The GBA, GFA, and GTA can also be utilized as sample vaporization sources for a miniature spark and an induction - coupled plasma. Because of the porous nature of these atomizers, they have applications in gas and liquid chromatography as well as in electrochemistry. R EF EREN CE S Li 0“ Elk! n—p—“W‘ E 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. REFERENCES A. Walsh, Spectrochim. Acta., 7, 108 (1955). E. T.)J. Alkemade and J. M. W. Milatz, J, Opt, Sop, Am,, 45, 583 1955 . '—- C. T. J. Alkemade and J. M. W. Milatz, Appl. Sci. Res. 43, 289 (1955). “‘ J. Ramirez-Mufioy, "Atomic-Absorption Spectroscopy", Elsevier Publishing Co., Amsterdam, 1968. W. Salvin, "Atomic Absorption Spectroscopy", Interscience Pub- lishers, New York, 1968. I. Rubeska and B. Moldan, "Atomic Absorption Spectrophotometry", Iliffe Books Ltd., London, 1969. R. J. Reynolds and K. Aldous, "Atomic Absorption Spectroscopy, A Practical Guide", C. Griffin Co., Ltd., London, 1970. G. 0. Christian and F. J. Feldman, "Atomic Absorption Spectroscopy. Application in Agriculture, Biology and Medicine", John Wiley, New York, 1970. B. V. L'vov, "Atomic Absorption Spectrochemical Analysis", J. H. Dixon, tran., Adam Hilger, London, 1970. L. DeGalan, "Analytical Spectrometry", Adam Hilger, Ltd., London, 1971. D. R. Browning, Ed. "Spectroscopy", McGraw-Hill, New York, NY, 1970. :0 . W. Wood, Phil, Mag, 12, 513 (1905). E. L. Nichols and H. L. Howes, Phys. 531. g, 472 (1924). R. M. Badger, Z, Phys, 22, 56 (1929). R. Mankopff, yer, Deutsch Phys. Ges., i2! 16 (1933). J. W. Robinson, Anal. Chim. Acta. 2;, 254 (1961). C. T. J. Alkemade, in Proc. X Colloq. Spectrs. Internat., Spartan Books, Washington, 0.C., 963: p. 143. J. D. Winefordner and T. J. Vickers, Anal. Chem. 22, 161 (1964). J. D. Winefordner and R. A. Staab. Ibid., 36, 1367 (1964). 280 20. 21. 22. 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 39. 40. 281 T. S. West, lecture to the Society for Analytical Chemistry, October 6, 1965, see Analyst, 91, 69 (1966). R. M. Dagnall, T. S. West, and P. Young, Talanta, 13, 803 (1966). P. L.)Larkins and J. B. Willis, Spectrochem. Acta., 268, 491 1971 . “" H. V. Malmstadt and E. Cordos, Pittsburgh Conference on Applied SpectrOSCOPY. Cleveland, Ohio, March 7, 1972, papers 132 and 132A. J. O. Winefordner and T. J. Vickers, Anal. Chem., 42, 206R (1970). J. D. Winefordner, V. Svoboda, and L. J. Cline, CRC Crit. Rev. Anal. Chem., 4, 233 (1970). J. D. Winefordner and R. C. Elser, Anal. Chem., 44, 24A (1971). Cu D. Winefordner and T. J. Vickers., Anal. Chem., 44, 150R (1972). .g S. West and M. S. Cresser, Appl. Spectros. Rev., Z, 79 (1973). W. J. Price in "Spectroscopy", D. R. Browning Ed., McGraw-Hill, Ne York, NY, 1969. R. Smith in "Spectrochemical Methods of Analysis", J. D. Winefordner, Ed., John Wiley, New York, NY, 1971. J. D. Winefordner and R. Smith in "Analytical Flame Spectrometry", R. Mavrodineanv, Ed., Macmillan, Ltd., London, 1970. A Syty in "Flame Emission and Atomic Absorption Spectrometry", J. A. Dean and T. C. Rains, Ed., Vol. 2 Marcel Dekker, New York, NY. 1971. T. S. West and M. S. Cresser, Appl. Spectros. Rev., 2, 79 (1973). G. F. Kirkbright, Analyst.,.g§, 609 (1971). C. T. J. Alkemade and P. J. J. Zeegers in ”Spectrochemical Methods of Analysis, J. D. Winefordner, Ed., John Wiley, New York, NY, 1971. G. F. Kirkbright, A. Semb and T. S. West, Talanta, 14, 1011 (1967). G. F. Kirkbright and T. S. West, Appl. Optics., 1, 1305 (1968). G. F. Kirkbright and s. Vetter, Spectrochim. Acta., _2__6_B_. 505 (1971). D. R. Jenkins, Ibid, 258, 47 (1970). P. J. Slevin, V. I. Muscat, and T. J. Vickers, Appl. Spectrosc., 34, 296 (1972). IF‘ m .uwnm ." l‘ ‘ LXI-Mi. V _ Y'NIEKLL: ' . 1%.. 41. 42. 43. 43. 44. 45. 46. 47. 48. 49. 50. 51. 52. 53. 54. 55. 56. 57. 58. 59. 60. 61. 343 (1970). Transfer. , 7,251 (1967). pp, 1607 (1963). 282 D. N. Hingle, G. F. Kirkbright, and T. S. West, Talanta, 1_, 199 (1968). "' G. F. Kirkbright, A. Semb and T. S. West, Ibid, 444_(l968). R. S. Hobbs, G. F. Kirkbright, M. Sargents and T. S. West, Ibi , 987 (1968). D. N. )Hingle, G. F. Kirkbright and T. S. West, Analyst. , 93, 522 1968 R. M. Dagnall, G. F. Kirkbright, T. S. West and R. M. Wood, Anal. Chem., 44,1029 (197OL K. M. Aldous, R. F. Browner, R. M. Dagnall, and T. S. West, Ibi , R. F. Browner and D. C. Manning, Ibid, 44, 843 (1972). . Palermo, Ph. D. Thesis, Michigan State University, 1973. Larkins, Spectrochim. Acta., 444, 477 (1971). . King, AstrOphys. 4,, 44, 318 (1922). MM?“ E P A. A . King and R. B. King, AstrOphys. 4,, 44, 377 (1935). R . B. King and D. C. Stockbarger, Astrophys. 4,, 44, 488 (1940). L. DeGalan and J. D. Winefordner, 4. Quant. Spectry. Radiative J. W. Robinson, "Atomic Absorption Spectroscopy", Marcel Dekker Inc., New York, 1966, P. 62. J. D. Winefordner, C. T. Mansfield and T. J. Vickers, Anal. Chem., H. P. Hooymayers and C. T. J. Alkemade, 4. Quant. Spectry. Radiative Tax—B : 3".y .-. ‘§ ‘9’ Transfer,6 __, 501 (1966). Ibid, 4, 847 (1966). D. R. Jenkins, Spectrochim. Acta., 444, 167 (1967). B. V. L‘vov, Spectrochim. Acta., 1 , 761 (1961). Ibid,_g44, 53 (1969). R. Woodriff and G. Ramelow, Ib i 2 B, 665 (1968). b—-H*- — R. Woodriff and R. W. Stone., Appl . t., 4, 1337 (1968). 62. 63. 64. 65. 66. 67. 68. 69. 70. 71. 72. 73. 74. 75. 76. 77. 78. 79. 80. 81. 82. 83. 84. 85. 283 . Woodriff, R. W. Stone and A. M. Held, Ibid, 44, 408 (1968). . Woodriff, B. R. Culver, K. W. Olson, Ibid, 44, 630 (1970). . Woodriff and D. Shrader, Anal. Chem., 44, 1918 (1971). . Woodriff and J. F. Lech, Ibid, 44, 1323 (1972). 306350wa . Woodriff, B. R. Culver, D. Shrader and A. B. Super, Ibid, 44, 230 (1973). "‘ H. Massmann, Proceedings XIIth Colloquium Spectroscopicum Inter- nationale, Hilger and Watts, London, 1965, pp. 275-278. H. Massmann, Spectrochim. Acta., 444, 215 (1968). H. Massmann, Z, Analyt. Chem., 444, 203 (1967). D. C. Manning and F. Fernandez, Atomic Absorption Newsletter, 4, 65 (1970). '— Perkin-Elmer Corp., Norwalk, CT 06852, ORDER NO. L-332. D. A. Segar and J. D. Gonzalez, Anal. Chem. Acta., 44, 7 (1972). C. J. Pickford and G. Rossi, Analyst, 44, 329 (1973). . Janssens and R. Dams, Anal. Chim. Acta., 44, 41 (1973). . C. Manning, Amer. Lab., 4(8), 37 (1973). M E. Norval and L. R. Butler, Anal. Chim. Acta., 44, 47 (1972). D R . T. Ross, J. G. Gonzalez and D. A. Segar, Anal. Chim. Acta., 44, 205 (1973). G. W. Fuller, Anal. Chim. Acta., 44, 261 (1972). D. Clark, R. M. Dagnall and T. S. West, Anal. Chim. Acta., 44, 11 (1973). J. W. Robinson and P. J. Slevin, 4943, 444,, 458), 10 (1972). G. H. Morrison and Y. Talmi, Anal. Chem., 44, 809 (1970). Y. Talmi and G. H. Morrison, 4414,_44, 1455 (1972). 4414, __44, 1467 (1972). J. B. Headridge and D. R. Smith, Talanta, 18, 247 (1971). K. Pagenkopf, D. R. Neuman, and R. Woodriff, Ibid, 44, 2243(1972). 284 86. J. P. Mislan, Paper presented at the 7th Conference on Analytical Chemistry in Nuclear Technology, Gatlinberg, Tennessee, 1963. 87. M. S. Black, T. H. Glenn, M. P. Bratzel and J. D. Winefordner, Anal. Chem., 44, 1769 (1971). 88. C. Veillon and M. Margoshes, Spectrochim. Acta., 444, 553 (1968). 89. M. K.)Murphy, S. A. Clylurn and C. Veillon, Anal. Chem., 44, 1468 1973 . "' 90. R. D. Hudson, Phyg. Rev., 135, 1212 (1964). 91. S. P. Choong and W. Loong-Seng, Nature, 204, 276 (1964). 92. E. J. Rapperport, J. P. Pemsler, and E. Adler, Rev. Sci. Instr., 44, 1168 (1970). 93. 4. P.)Pemsler and E. J. Rapperport, Anal. Chim. Acta., 44, 15 1972 . 94. F. S. Tomkins and B. Ercoli, Appl. Optics., 4, 1299 (1967). 95. H. P. Loftin, C. M. Christian and J. W. Robinson, §pectrosc., Let. 4, 161 (1970). 96. C. M. Christian and J. W. Robinson, Anal. Chem. Acta., 44, 466 (1971). 97. J. W. Robinson, P: J. Slevin, G. D. Hindman and D. K. Wolcott, Ibid, 4;, 431 (1972). 98. J. W. Robinson, D. K. Wolcott, P. J. Slevin and G. D. Hindman, bid, 44, 13 (1973). 99. U. Ulfuarson, Acta. Chem. Scand., 41, 641 (1967). 100. G. W. Kalb, Atomic Absorption Newsletter, 4, 84 (1970). 101. P. C. Head and R. A. Nicholson, M144, 44, 53 (1973). 102. H. Brandenberger and H. Bader, Helv. Chim. Acta., 44, 1409 (1967). 103. R. Stephens, Anal. Letters., 4, 851 (1972). 104. H. Brandenberger, 441414,.44, 449 (1968). 105. M. P. Bratze1,.R. M. Dagnall and J. D. Winefordner, Anal. Chim. Acta., 44, 197 (1969). 106. Ibid, Appl. §pectros., 44, 518 (1970). 107. M. Williams and E. H. Piepmeier, Anal. Chem., 44, 1342 (1972). 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 285 M. P.)Newton, J. V. Chauvin and D. G. Davis, Anal. Letters, 4, 89 1973 . '— J. V. Chauvin, M. P. Newton and D. G. Davis, Anal. Chem. Acta., 44, 291 (1973). J. E. Cantle and T. S. West, Talanta, 44, 459 (1973). H. M. Donega and T. E. Burgess, Anal. Chem., 44, 1521 (1970). J. Y. Hwang, C. J. Mokeler and P. A. Ullucci,4414, 44, 2018 (1972). T. Takeuchi, M. Yanagisawa and M. Suzuki, 4414444, 44, 465 (1972). T. Murata and T. Takeuchi, Anal. Chim. Acta., 44, 5 (1973). Ibid, 3, 253 (1972). J. Y. Hwang, P. A. Ullucci, S. B. Smith and A. L. Malenfant, Anal. Chem., 44, 1319 (1971). P. A. Ullucci, C. J. Mokeler and J. Y. Hwang, Amer. 444, 4(8). 63 (1972). " K. Puma and B. L. Vallee, Anal. Chem., 35, 942 (1963). A. Ando, K. Fuwa and B. L. Vallee, Ibid, 44, 818 (1970). W. W.)McGee and J. D. Winefordner, Anal. Chim. Acta., 44, 429 1967 . ““- H. L. Kahn, G. E. Peterson and J. E. Shallis, Atomic Absorption Newsletter, 4, 35 (1968). R. A. White, paper presented at the International Atomic Absorp- tion Spectroscopy Conference, Sheffield, 1969. H. T. Delves, Analyst., 44, 431 (1970). H. T. Delves and R. B. Reeson, Ibid, 44, 343 (1973). D. Clark, R. M. Dagnall and T. S. West, Anal. Chim. Acta., 44, 339 (1972). "‘ R. D. Beaty, Anal. Chem., 44, 234 (1973). K. M.)A1dous, D. G. Mitchell and E. J. Ryan, Anal. Chem., 44, 1990 1973 . “' T. S. West and X. K. Williams., Anal. Chim. Acta., 44, 27 (1969). J. F. Alder and T. S. West, Ibid, 4;, 365 (1970). 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144. 145. 146. 147. 148. 149. 150. 151. 152. 286 R. G. Anderson, I. S. Maines and T. S. West, Ibid,_51, 355 (1970). J. Aggett and T. S. West, Ibid, 55, 349 (1971). D. Alger, R. G. Anderson, I. S. Mains and T. S. West, Ibid, 55, 271 (1971). . “' L. Ebdon, G. F. Kirkbright and T. S. West, Talanta, 55, 1301 (1972). R. G. Anderson, H. N. Johnson and T. S. West, Anal. Chim. Acta., 55, 281 (1971). L. Ebdon, G. F. Kirkbright and T. S. West, 5515, 55, 39 (1972). K. W. Jackson and T. S. West, 5515, 55, 187 (1972). 55155515, 363 (1973). L. Ebdon, G. F. Kirkbright and T. S. West, 5515, 51, 15 (1972). D. J. Johnson, T. S. West and R. M. Dagnall, 5515, 55, 171 (1973). J. Aggett and T. S. West, I_bi_d_, 5_7_, 15 (1971). G. L. Everett, T. S. West and R. W. Williams, 5515, 55, 301 (1973). R. D. Reeves, B. M. Patel, C. J. Molnar and J. D. Winefordner, Anal. Chem., 55, 246 (1973). B. M. Patel, R. D. Reeves, R. F. Browner, C. J. Molnar and J. D. Winefordner, Appl. Spectros., 55, 171 (1973). M. o. Amos, Amer. 95., g, 33 (1970). C. J. Molnar, R. D. Reeves, J. D. Winefordner, M. T. Glenn, J. R. Ahlstrom, and J. Savory, Appl. Spectros., 25, 606 (1972). M. D. Amos, P. A. Bennett, K. G. Brodie, P. W. Y. Lung and J. P. Matsusek, Anal. Chem., 55, 41 (1971). M. T. Glenn, J. Savory, L. P. Hart, T. H. Glenn and J. D. Winefordner, Anal. Chem., 55. 246 (1973). J. P. Matousek, Amer. 555,, 516), 45 (1971). M. D. Amos, Ibid, 1(8), 57 (1972). . P. Matousek and B. J. Stevens, Clin. Chem., 15, 363 (1971). . R.)Parker, J. Rowe and D. P. Sandoz, Amer. 555,, 5(8), 53 1973 . '— . Tessari and G. Torsi, Talenta, 19, 1059 (1972). OAOL 153. 154. 155. 156. 157. 158. 159. 160. 161. 162. 163. 164. 165. 166. 167. 168. 169. 170. 171. 172. 173. 174. 287 M. Glenn, J. Savory, S. A. Fein, R. D. Reeves, C. J. Molnar and J. D. Winefordner, Anal. Chem., 55, 203 (1973). M. Glenn, J. Savory, L. Hart, T. Glenn and J. D. Winefordner, Anal. Chem. Acta., 5;, 263 (1971). B. J. Stevens, Clin. Chem., 15, 1379 (1972). M. P. Bratzel, C. L. Chakrabarti, R. E. Sturgeon, M. W. McIntyre and H. Agemian, Anal. Chem., 55, 372 (1972). M. Yanagisawa, T. Takeuchi and M. Suzuki, Anal. Chem. Acta., 55, 381 (1973). J. F. Alder and T. S. West, Ibid, 55, 331 (1972). %1 Ha11, M. P. Bratzel and C. L. Chakrabarti, Talanta, 55, 755 973 . 5. M.)Pate1 and J. D. Winefordner, Anal. Chim. Acta., 55, 135 1973 . R. D. Reeves, C. J. Molnar, M. T. Glenn, J. R. Ahlstrom and J. D. Winefordner, Anal. Chem., 55, 2205 (1972). J. F. Alder and T. S. West, Anal. Chim. Acta., 51, 132 (1972). G. L. Everett and T. S. West, Ibid, 55, 301 (1973). M. P. Bratzel and C. L. Chakrabarti, Ibid, 51, 25 (1972). W. K. Robbins, Anal. Chem. Acta.,:55, 285 (1973). R. C. Chu, G. P. Barron and P. A. W. Baumgarner, Anal. Chem., 55, 1476 (1972). J. Y. Hwang, P. A. Ullucci and A. L. Malenfant, Can. Spectrosc., 55, 100 (1971). F. M. D'Itri, C. S. Annett and A. W. Fast, MTS Journal, 5, 10 (1971). g. I.)Muscat, T. J. Vickers and A. Andren, Anal. Chem., 5;, 218 1972 . T. R. Gillert and D. N. Hume, Anal. Chem. Acta., 5, 461 (1973). N. P.)Kubasik, H. E. Sine and M. T. Volosin, Clin. Chem., 18, 1326 1972 . ‘__ S. R. Aston and J. P. Riley, Ibid, 52, 349 (1972). S. H. Omang and P. E. Paus, Ibid, 55, 393 (1971). B. J. Russell and A. Walsh, Spectrochim. Acta., 15, 883 (1959). 175. 176. 177. 178. 179. 180. 181. 182. 183. 184. 185. 186. 187. 188. 189. 190. 191. 192. 193. 194. 195. 196. 197. 198. 199. 288 B. Gatehouse and A. Walsh, Ibid, 15, 602 (1960). A. Walsh, Spectroscopy, The Institute of Petroleum, London, 1962, pp. 13-26. A. Walsh, Appl. §pectrosc., 55, 335 (1973). D. S. Gou h, P. Hannaford and A. Walsh, Spectrochim. Acta., 555, 197 (1973) J. A. Goleb and J. K. Brody, Anal. Chim. Acta., 55, 457 (1963). J. A. Goleb, Anal. Chem., 55, 1978 (1963). J. A. Goleb, Anal. Chim. Acta., 55, 135 (1966). W. W. Harrison and E. H. Daughtrey, 5515, 55, 35 (1973). N. P. Ivanov, M. N. Gusinski and A. D. Esikov, Zhur. Anal. Khim., Q, 1133 (1965). H. Massmann, Spectrochim. Acta., 555, 393 (1970). T. Kantor and L. Erdey, Ibid, 555, 283 (1969). Yu. I. Belyaev, A. M. Pchelintsev and N. F. Zvereva, Zhurna1 Analiticheskoi Khimii, 55, 1295 (1971). M. Marinkovic and T. J. Vickers, Appl. Spectros.,:;5, 319 (1971). . L. Jones, R. L. Dablguist and R. E. Hoyt, Ibid, 55, 628 (1971). . K. Winge, V. A. Fassel and R. N. Kniseley, Ibid, 555 (1971). . W. Robinson, Anal. Chim. Acta., 5;, 465 (1962). Greenfield, 1. Jones and C. T. Berry, Analyst, 55, 713 (1964). . H. Wendt and V. A. Fassel, Anal. Chem., 5;, 920 (1965). D. West and D. N. Hume, 5515, 55, 412 (1964). . C. Hoare and R. A. Mostyn, 5515, 55, 1153 (1967). H. Wendt and V. A. Fassel, 5515,155, 337 (1966). E. Friend and A. J. Diefenderfer, 5515,155, 1763 (1966). Veillon and M. Margoshes, Spectrochim. Acta., 555, 503 (1968). . W. Dickinson and V. A. Fassel, Anal. Chem., 51, 1021 (1969). . W. J. M. Bowmans and J. F. DeBoer, Spectrochim. Acta., 555, 91 (1972). w'umoxxxnxmaxc. mmnmm,__vfi 91‘!!th l“— ! - 'u j_‘. "I 1‘ ., ~ ‘i I 1 D 200. 201. 202. 203. 204. 205. 206. 207. 208. 209. 210. 211. 212. 213. 214. 215. 216. 217. 218. 219. 220. 221. 289 S. Greenfield and P. B. Smith, Anal. Chim. Acta., 55, 341 (1972). R. N. Kniseley, V. A. Fassel and C. C. Butler, Clin. Chem., 15, 807 (1973). ?. Magrodineanu and R. L. Hughes, Spectrochim. Acta., 55, 1309 1963 . S. Murayama, H. Matsuno and M. Yamamoto, Spectrochim. Acta., 555, 513 (1968). K. Faligatter, V. Svoboda and J. D. Winefordner, Appl. Spectros., 55, 347 (1971). K. M. Aldous, R. M. Dagnall, B. L. Sharp and T. S. West, Anal. Chim. Acta., 54, 233 (1971). K. W. Busch and T. J. Vickers, 5pectrochim. Acta., 555, 85 (1973). N. A. Kuebler and L. S. Nelson, 5, 555, Soc. Am,,:51, 1411 (1961). S. Nelson and N. A. Kuebler, J. Chem. Phys., 55, 47 (1962). S. Nelson and N. A. Kuebler, Spectrochim. Acta., 19, 781 (1963). Margoshes and B. F. Scribner, Anal. Chem., 55, 223R (1968). x: 3: r- r- Atwill, International Electronics, 7, 18 (1964). < . G. Mossotti, K. LaQua and W. D. Hagenah, Spectrochim. Acta., 55:, 197 (1967). E. H. Piepmeier and H. Malmstadt, Anal. Chem., 51, 700 (1969). D. E. Osten and E. H. Piepmeier, Appl, Spectrosc., 55, 165 (1973). A. A. Venghiattis, Atomic Absorption Newsletter, 5, 19 (1967). L. R. P. Butler and J. A. Brink in "Flame Emission and Atomic Absorption Spectrometry", J. A. Dean and T. C. Rains, Ed., Vol. 2 Marcel Dekker, New York, NY 1971. High Frequency Excited Electrodeless Sources in Analytical Chemistry, by P. M. Beckwith, R. F. Browner and J. D. Winefordner, Marcel Dekker, 1974 (in press). T. S. West and M. S. Cresser, Appl. Spectrosc. Rev., 5, 79 (1973). E. H. Piepmeier, Spectrochim. Acta., 555, 431 (1972). I id, 515, 445 (1972). J. Kuhl, G. Marowsky and R. Torge, Anal. Chem., 55, 375 (1972). nJi r I} AI: . I ' 2.1—” — "y 222. 223. 224. 225. 226. 227. 228. 229. 230. 231. 232. 233. 234. 235. 236. 237. 238. 239. 240. 241. 242. 243. 244. 245. 290 lZi. B.)Denton and H. V. Malmstadt, Appl. Pth. Lett. , 1:8, 485 1971 . L. M. Fraser and J. D. Winefordner, Anal. Chem., 55, 1693 (1971). Ibid, 44, 1444 (1972). N. Omenetto, N. N. Hatch, L. M. Fraser and J. D. Winefordner, Ib d, 45,195 (1973). N. Omenetto, P. Benetti, L. P. Hart, J. D. Winefordner and C. T. Alkemade, Spectrochim. Acta., 555, 289 (1973). N. Omenetto, L. P. Hart, P. Benetti and J. D. Winefordner, Ibid, :551.(1973). J. . Dawson and D. J. Ellis, Ibid, 555, 565 (1967). . Willis, Rev. Pure. and Appl, Chem., 11, 111 (1967). B B A. Buger and W. Fink, §pectrochim. Acta., 555, 1359 (1970). . V. Gelder, Ibi , 555, 669 (1970). V F 0 . Tright, Phys. Rev., 151, 97 (1969). LOONVL . Weide and M. L. Parsons, Anal. Letters,:5, 363 (1972). 5. G. )Mitchell and A. Johansson, Spectrochim. Acta. , 25B,175 1970 Ibid, 555, 677 (1971). Y. Yokoyama and S. Ikeda, 5515, 555, 118 (1969). J. V. Sullivan and A. Walsh, ABEL-9E” ,1271 (1968). E. Cordos and H. V. Malmstadt, Anal. Chem., 55, 2277 (1972). 5515, 55, 27 (1973). H. V. Malmstadt and E. Cordos, 5555, 555,,:5(8) 35 (1972). M. Lowe, Spectrochim. Acta., 555, 191 (1969). 0. Lloyd and R. M. Lowe, Ibid,=555, 23 (1972). G. C. Human, Ibid, 555 (1972). . A. Sebestyen, Ibid, 555 (1970). 21‘” . Bruce and P. Hannaford, §pectrochim. Acta., 555, 207 (1971). 246. 247. 248. 249. 250. 251. 252. 253. 254. 255. 256. 257. 258. 259. 260. 261. 262. 263. 264. 265. 266. 267. 291 J. F. Kielkopf, Ibid, 555 (1971). K. W. Busch and G. H. Morrison, Anal. Chem., 55, 712A (1973). P. W. J. M. Boumans and G. B. Brouwer, 5pectrochim. Acta., 555, 247 (1972). G. 0. Christian and F. J. Feldman, Appl. Spectrosc., 55, 660 (1971). D. R. Jenkins, Spectrochim. Acta., 555, 167 (1967). P. L. Larkins, R. M. Lowes, J. V. Sullivan and A. Walsh, Ibid, 555, 187 (1969). T. J. Vickers and R. M. Vaught, Anal. Chem., 55, 1476 (1969). T. J. Vickers, P. J. Sleirn, V. I. Muscot and L. T. Farias, Ibid, 55, 930 (1972). P. D. Warr, Talanta, 55, 543 (1970). ?. C.)Elser and J. D. Winefordner, Appl, Spectrosc., 55, 345 1971 . T. S. West and X. K. Williams, Anal. Chem., 55, 335 (1968). ?. G.)Mitchell and A. Johansson, Spectrochem. Acta., 555, 175 1970 . Ibid, 555, 677 (1971). R. M. Dagnall, G. F. Kirkbright, T. S. West and R. Wood, Ana1. Chem., 55, 1765 (1971). Ibid, Analyst., 5;, 245 (1972). M. Jones, G. F. Kirkbright, L. Ranson and T. S. West, Anal. Chim. Acta., 55, 210 (1973). E. Cordos and H. V. Malmstadt, Ana1. Chem., 55, 425 (1973). C. T. J. Alkemade, Sheffield Atomic Absorption Conference, Sheffield, England, July 1969. H. P. Hooymayers, Spectrochim. Acta., 555, 567 (1968). P. T. J. Zeegers, R. Smith and J. D. Winefordner, Anal. Chem., 55, 26A (1968). I. Langmuir, Phys. 555,, Q, 329 (1913). G. R. Fonda, Ibid, 51, 343 (1923). I. PHI)“. “.1.- ;~'_v.a* 1 _ 4‘ 1. IA .. _L-‘L—fl r Que-'7‘“ ‘n 268. 269. 270. 271 272. 273. 274. 275. 276. 277. 278. 279. 280. 281. 282. 283. 284. 285. 286. 287. 288. 292 Ibid, 51, 260 (1928). H. L. Langhaar, Dimensional Ana1ysis and Theory of Models, New York, John Wiley 8 Sons, Inc., 1951. W. J. King, Mech, Eng., 55, 410 (1932). W. H. McAdams, Heat Transmission, 3rd Ed., New York, McGraw-Hill Book Company, Inc., 1954. B. Gebhart, Heat Transfer, 2nd Ed. ,New York, McGraw-Hill Book Company, Inc. 0. C. Collis and M. J. Williams, 5, Fluid Mech., 5, 357 (1959). Ibid, Aerodynamics, Note 140, Aeronautical Research Laboratories, Ml bourne, Australia, 1954. S. R. Goode, Akbar Montaser and S. R. Crouch, Appl. Spectrosc., £1, 355 (1973). S. R. Goode, Ph. D. Thesis, Michigan State University, 1973. . A. Farber and R. L. Scorah, Trans. ASME,:55, 369,(1948). E A. D. Wilson, Applied Optics, 5, 1055 (1963). C. J. Smithells, Tungsten, Chapman and Hall, London, 1952. E. Kellett and S. Rogers, 5, Electrochem. Soc., 555, 502 (1963). W. Webbe, J. Norton, and C. Wanger, J. Electrochem. Soc. , 555, 107 (1956). R. Speiser and G. St. Pierre, Agard Conference, Oslo, Pergamon Press, Oxford, 1963. D. J. Page, Union Carbide Corp., Parma, Ohio, Personal correspondence. W. J. Findlay, et al., Fourth International Conference on Atomic Spectroscopy, Toronto, Canada, 29 October - 2 November, 1973, Paper Tr—16. R. R. Asamoto and P. E. Novak, Rev. Sci. Instrum.,:55, 1047 (1967). S. H. Praul and L. V. Hmurcik, Rev. Sci. Instrum., 55, 1363 (1973). A. Lorber and L. S. Cutler and C. C. Chang, Arthritis and Rheumatism, 11(1), 65 (1968). P. T. Parter, “Modulation, Noise, and Spectral Ana1ysis", McGraw- Hill, New York, 1965, pp. 548-589. 289. 290. 291. 292. 293 . R. Aaron in "Modern Filter Theory and Design", G. C. Temes and . K. Mitra, Eds., Wiley-Interscience, New York, 1973. pp. 425-431. . C. Kelly and G. Horlick, Anal. Chem., 55, 518 (1973). ‘UMZ C . Haisch, K. Laqua and W. D. Hagenah, Spectrochem. Acta., 555, 651 (1971). E. Pals, Akbar Montaser and S. R. Crouch, Current Research (1974).