TRENDS IN TUNGSTEN COIL ATOMIC SPECTROMETRY GEORGE L. DONATI. A Dissertation Submitted to the Graduate Faculty of

Transcription

1 TRENDS IN TUNGSTEN COIL ATOMIC SPECTROMETRY BY GEORGE L. DONATI A Dissertation Submitted to the Graduate Faculty of WAKE FOREST UNIVERSITY GRADUATE SCHOOL OF ARTS AND SCIENCES In Partial Fulfillment of the Requirements for the Degree of DOCTOR OF PHILOSOPHY in the Department of Chemistry May 2010 Winston Salem, North Carolina Approved By: Bradley T. Jones, Ph.D., Advisor Examining Committee: Martin Guthold, Ph.D, Chairman Christa L. Colyer, Ph.D. Patricia C. Dos Santos, Ph.D. Willie L. Hinze, Ph.D.

2 ACKNOWLEDGEMENTS Every important accomplishment is the result of several well intentioned people working for the common good. This dissertation would never be possible without the support I have received over the years. I would like to thank everyone who direct or indirectly contributed for this work s completion. Nothing would be possible without God s endless love, providing me with faith and blessing me with countless opportunities. I would like to thank my mom, Sara, for being the light guiding my steps and the reason I have become a scientist. Thanks to my dad, Eduardo, for showing me the path of honesty, discipline, humbleness and above all, teaching me how to appreciate the simple things in life. Thanks to my wife, Eunice, for being my unlimited source of support, love and strength, and for being my greatest, most patient fan. Thanks for giving me the most wonderful of all gifts, my daughter Carolina. I would like to express my gratitude to all my former teachers and professors for the patience and support. This work has a touch of every single one of you. I would like to especially thank my advisor, professor and mentor, Dr. Bradley T. Jones. You have always been a source of inspiration. If the best way to learn is by example, you have taught me professionalism, simplicity, patience, passion for teaching, and of course, a great deal of spectroscopy too. I would also like to thank my friend, partner and former advisor, Dr. Joaquim A. Nóbrega. Thanks for believing in me, for all the patience and guidance, for helping me to find one of my passions, and above all, for being my friend. I also owe a great deal of gratitude to my committee, Dr. Martin Guthold, Dr. Christa L. Colyer, Dr. Patricia C. dos Santos and Dr. Willie L. Hinze, for their time, assistance and suggestions in completing this work. Thanks to the Department of Chemistry, including faculty, staff, past and present undergraduate and graduate students. I have been blessed by being surrounded with such kind people and so many good friends. Thank you. ii

3 TABLE OF CONTENTS LIST OF TABLES AND FIGURES vi ABSTRACT xiii CHAPTER I. INTRODUCTION 1 II. DETERMINATION OF Cd IN URINE BY CLOUD POINT EXTRACTION-TUNGSTEN COIL ATOMIC ABSORPTION SPECTROMETRY 17 Published in Talanta, 2008 III. EVALUATION OF EXPERIMENT DESIGNING AND MULTIVARIATE CALIBRATION AS STRATEGIES TO REDUCE MATRIX INTERFERENCES IN TUNGSTEN COIL ATOMIC ABSORPTION SPECTROMETRY 39 IV. SIMULTANEOUS DETERMINATION OF THE LANTHANIDES BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY 59 Published in the Journal of Analytical Atomic Spectrometry, 2008 iii

4 V. MULTI-WAVELENGTH DETERMINATION OF COBALT BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY 88 Published in Analytical Letters, 2010 VI. SIMULTANEOUS DETERMINATION OF Cr, Ga, In AND V IN SOIL AND WATER SAMPLES BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY 116 Published in Spectrochimica Acta Part B, 2009 VII. DOUBLE TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY: SIGNAL ENHANCEMENT AND A NEW GAS PHASE TEMPERATURE PROBE 142 Published in the Journal of Analytical Atomic Spectrometry, 2009 VIII. INDIRECT DETERMINATION OF IODIDE BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY 170 Published in the Microchemical Journal, 2009 IX. A NEW ATOMIZATION MICRO-CELL FOR TRACE METAL DETERMINATIONS BY TUNGSTEN COIL ATOMIC SPECTROMETRY 199 iv

5 X. CONCLUSIONS 234 SCHOLASTIC VITA 238 v

6 LIST OF TABLES AND FIGURES TABLES PAGE CHAPTER II 1. Previously reported WCAAS limits of 33 detection for cadmium. 2. Tungsten coil heating cycle for CPE urine samples. 34 CHAPTER III 1. Operating parameters for ICP OES and tungsten 52 coil heating cycle. 2. Factorial experiment Determination of Cu in water samples by WCAAS 55 with matrix interference reduction by experimental designing and multiple linear regression modeling. CHAPTER IV 1. Tungsten coil heating programs WCAES analytical figures of merit. 80 CHAPTER V 1. Tungsten coil heating cycle Atomic emission lines for cobalt Analytical figures of merit for cobalt determination 115 by WCAES. vi

7 CHAPTER VI 1. Tungsten coil heating cycle WCAES analytical figures of merit WCAES accuracy for the determination of Cr, Ga, 135 In, and V in soil and water samples. CHAPTER VII 1. Tungsten coil heating cycle Gas phase temperature measurements DWCAES analytical figures of merit DWCAES determination of Ti and V in spiked 163 water samples. CHAPTER VIII 1. Tungsten coil heating cycle. 195 CHAPTER IX 1. Tungsten coil heating cycle WCAAS and WCAES analytical figures of merit WCAAS and WCAES accuracy. 233 vii

8 FIGURES PAGE CHAPTER II 1. Absorption profiles collected during the WCAAS 35 analysis of a urine sample. The background profile (circles) is simply the absorbance measured at nm. The analytical profile (squares) is the absorbance measured at nm minus the absorbance measured at nm. 2. Cadmium levels determined in seven urine samples 36 by WCAAS with near-line background correction (white bars), WCAAS with temporal background correction (cross-hatched bars), and by ICP-MS (gray bars). The error bars represent ± one standard deviation (n=3). CHAPTER III 1. Schematic diagram of WCAAS instrumentation. 56 Inset shows an alternate view of the atomization cell. CHAPTER IV 1. Schematic diagram of WCAES instrument. Inserts 82 show alternate views of the atomization cell and the coil image on the adjustable aperture. viii

9 2. Simultaneous multi-element WCAES determination 83 of Yb (1 mg l -1 ), Dy and Tm (2 mg l -1 ), Ho and Er (5 mg l -1 ). 3. Simultaneous multi-element WCAES determination 84 of Sm and Eu (1 mg l -1 ), Lu (5 mg l -1 ), Ce (10 mg l -1 ), Nd and Gd (20 mg l -1 ). CHAPTER V 1. Photographs of WCAES atomization cell (A), and 108 close-up showing the viewing position (black dot) relative to the W-coil (B). 2. Background-corrected emission spectrum for a mg/l Co solution. The five emission lines used for quantification are designated A-E. 3. Contour plot showing the effects of purge gas flow 111 rate and atomization current on the relative Co emission signal at nm. The maximum value (100 %) occurs at 10 A and 0.9 L/min. CHAPTER VI 1. Schematic design of the WCAES instrument: (A) 122 Block diagram; (B) Alternate view of the Atomization cell; (C) Coil image position relative to the spectrograph s slit. ix

10 2. Background corrected spectra for simultaneous 132 multi-element WCAES determination of Cr (1.0 mg l -1 ), Ga (5.0 mg l -1 ), In (1.0 mg l -1 ), and V (20 mg l -1 ): (A) Spectrum collected from s after the beginning of the atomization step; (B) Spectrum collected from s after the beginning of the atomization step. CHAPTER VII 1. DWCAES arrangement: schematic diagram (A), 165 close-up of the atomization cell (B), and coil images projected on the spectrograph entrance slit (C). 2. DWCAES emission signals for some elements 166 with transition energies higher than 350 kj mol -1 : Ag, Cu, Sn = 1000 mg l -1 ; Na = 100 mg l Emission signals for a 2.0 mg l -1 V solution using 167 one (lower trace) and two (upper trace) tungsten coils. CHAPTER VIII 1. Schematic diagram of the tungsten coil 190 instrumentation. The inset shows a close-up view of the monochromator entrance slit with the coil image positioned for emission measurements. x

11 2. Molecular absorption profile observed for InI at nm using a V hollow cathode lamp as the radiation source. The vaporization step begins at time 0. Solution concentrations were 1000 mg l -1 In and 500 mg l -1 I Atomic emission profiles for 10 mg l -1 In at nm in the absence of I - (squares) and in the presence of 5 mg l -1 I - (diamonds). The vaporization step begins at time 0. The maximum iodide sensitivity is observed at 1.0 s after the onset of atomization, and the longest linear dynamic range is achieved at 1.4 s. 4. Analytical calibration curves for the indirect 193 determination of iodide by tungsten coil atomic emission spectrometry. The Indium atomic emission signal was measured at nm in all cases. The highest Iodide sensitivity is observed with a 10 mg l -1 In solution and a measurement time of 1.0 s after the onset of atomization (A). The limit of detection is depicted by the triangle. Longer linear dynamic ranges are observed at a measurement time of 1.4 s with either 10 of 50 mg l -1 In (B). xi

12 CHAPTER IX 1. Schematic diagram of WCAAS and WCAES 225 instrumentation. A- System arrangement overview. B- WCAAS and WCAES atomization cell caps. C- Atomizer projection on the spectrograph entrance slit for WCAES. 2. WCAAS Cd absorption spectrum: 2 µg L -1 solution 226 monitored at nm, using the nm line for background correction. 3. WCAES atomization cell. A- Cell base showing the 227 atomizer at 0 A. B- Cell with the cap on and the atomizer at 3.5 A. 4. WCAES multielement simultaneous determination. 228 Concentrations: Cr and Eu 200 µg L -1 ; Sr 20 µg L -1. Calcium is always present as concomitant in the blank. 5. Effect of the protecting gas mixture on WCAES 229 analytical signal. A- Effect of H 2 volume percentage in Ar. B- Comparison between Ar, N 2 and He mixtures at the best conditions for each element. Concentrations: Cr and Eu 200 µg L -1 ; Sr 20 µg L -1. Calcium is always present as concomitant in the blank. xii

13 ABSTRACT Donati, George L. TRENDS IN TUNGSTEN COIL ATOMIC SPECTROMETRY Dissertation under the direction of Bradley T. Jones, Ph.D., Professor of Chemistry Renewed interest in electrothermal atomic spectrometric methods based on tungsten coil atomizers is a consequence of a world wide increasing demand for fast, inexpensive, sensitive, and portable analytical methods for trace analysis. In this work, tungsten coil atomic absorption spectrometry (WCAAS) and tungsten coil atomic emission spectrometry (WCAES) are used to determine several different metals and even a non-metal at low levels in different samples. Improvements in instrumentation and new strategies to reduce matrix effects and background signals are presented. Investigation of the main factors affecting both WCAAS and WCAES analytical signals points to the importance of a reducing, high temperature gas phase in the processes leading to atomic cloud generation. Some more refractory elements such as V and Ti were determined for the first time by double tungsten coil atomic emission spectrometry (DWCAES). The higher temperatures provided by two atomizers in DWCAES also allowed the detection of Ag, Cu and Sn emission signals for the first time. Simultaneous determination of several elements by WCAES in relatively complex sample matrices was possible after a simple acid extraction. The results show xiii

14 the potential of this method as an alternative to more traditional, expensive methods for fast, more effective analyses and applications in the field. The development of a new metallic atomization cell is also presented. Lower limits of detection in both WCAAS and WCAES determinations were obtained due to factors such as better control of background signal, smaller, more isothermal system, with atomic cloud concentration at the optical path for a longer period of time. Tungsten coil-based methods are especially well suited to applications requiring low sample volume, low cost, sensitivity and portability. Both WCAAS and WCAES have great commercial potential in fields as diverse as archeology and industrial quality control. They are simple, inexpensive, effective methods for trace metal determinations in several different samples, representing an important asset in today s analytical chemistry. xiv

15 CHAPTER I INTRODUCTION The scientific and technological revolution observed in the last century has changed the face of humanity and provided virtually endless possibilities to modern society. Although responsible for significant improvement in people s lives, the development of different new technologies came with some important challenges. Environmental and health issues are only a few examples of those new challenges. 1, 2 The remarkable evolution of analytical methods in the last few decades is a consequence of the necessity to identify and quantify several different chemical compounds, at increasingly lower levels, in the most varied types of samples. Rising global demand for minerals and metals, as well as recent environmental regulations have been fueling the development of more sensitive, efficient, and economical analytical techniques. Due to characteristics such as robustness, selectivity, sensitivity, and fast sample output, atomic spectrometry represents one of the most important analytical methods 3 with a world market estimated at $2.5 billion. 4 Quantitative analysis by atomic spectrometry may be roughly divided into three main steps: sample introduction, atomization, and detection. Sample introduction can be either continuous as in flame and plasma methods, or 1

16 discrete as in electrothermal methods. There are several different methods of detection, but the most common are related to atomic absorption, emission, or fluorescence. Most compounds are decomposed at high temperatures yielding free atoms. In atomic spectrometric methods, the energy provided by the atomizer, at temperatures varying from 2000 to K, is enough to promote atomic transitions. Therefore, electromagnetic radiation either absorbed or emitted by the analyte atoms at some specific wavelength may be used to determine their concentration in the sample. 3 Based on the type of atomizer employed, atomic spectrometric methods may be divided in three main categories. 5 Flame-based methods are the simplest and most economical. However, problems related to poor sample introduction efficiency, dilution of the atomic cloud by the gases in the flame, and short residence time of the analyte atoms in the optical path may lead to reduced sensitivity. Typical limits of detection (LOD) for flame-based methods rarely reach the μg L -1 level. 6 A second category is based on plasma atomizers. Due to the possibility of simultaneous determinations of several different analytes, methods such as inductively coupled plasma optical emission spectrometry (ICP OES) and inductively coupled plasma mass spectrometry (ICP-MS) have become important analytical tools in different fields such as environmental, food quality control, clinical analysis, and others. These techniques allow high sample throughput, high sensitivity, and straightforward coupling to efficient separation methods. 4, 7 A disadvantage associated with these methods is the cost for 2

17 maintenance and supplies. In addition, the equipment is typically bulky, making it difficult for applications in the field. A third category of atomic spectrometric methods is based on electrothermal atomizers. The most common of them is graphite furnace atomic absorption spectrometry (GFAAS). Introduced in 1961 by L vov, 8 GFAAS is one of the most widely used analytical methods for trace and ultra-trace metal determinations. 9 The first arrangement proposed by L vov consisted of a graphite tube lined with tantalum foil to eliminate analyte loss by diffusion though the graphite tube walls. This atomizer was kept in a sealed chamber and purged with argon at 1 atm. Sample introduction was performed with a graphite electrode at the side of the furnace. After heating the furnace to approximately 2500 K, a small aliquot of the sample solution (about 100 μl) was deposited onto the graphite electrode. Then, a supplementary graphite electrode vaporized and atomized the sample with a DC arc. Later improvements in the arrangement were made and both graphite furnace and electrode were rapidly heated by passing an AC current through them, with no need for a supplementary electrode. In 1968, a much simpler arrangement was proposed by Massmann. 10 The graphite furnace was electrically heated from its ends, but unlike previous arrangements, it was not preheated. The sample aliquot was introduced with a small pipette through an orifice in the side of the tube. The furnace was then resistively heated in different stages to dry, vaporize and atomize the sample. To prevent the graphite tube from oxidation, an inert gas flowed continuously during the heating cycles. Unlike flame-based atomizers, the graphite furnace exhibited 3

18 practically no losses during the sample introduction, and the atomic cloud produced during the atomization was concentrated in a small space for a longer time. Hence, improvements in sensitivity and LOD as large as 4-5, and 2-3 orders of magnitude, respectively, were achieved. 9 Due to its simplicity, the Massmann arrangement became very popular and it was the first GFAAS device to be commercially produced, by Perkin Elmer, in The arrangement was not without limitations however. Unlike L vov s cuvette, Massmann s furnace had to be heated from ambient temperature in the presence of a liquid sample aliquot. The tube walls were not uniformly heated during the different atomization stages. Frequently the energy necessary to dissociate and atomize the analyte varied, and problems related to recombination and molecular absorption became significant. Spectral and chemical interferences were severe and precision was poor. Some researchers noted a gradient in temperature along the length of the furnace. The water-cooled electrodes caused the cooling of the tube ends, and condensation of analytes and matrix species on those sites often caused an increase in spectral and nonspectral interferences. To solve those problems, L vov suggested the use of a graphite platform inserted inside the Massmann furnace for sample deposition and vaporization. 11 The platform had little contact with the tube walls and reduced interference effects were observed. The heating of the sample was slightly delayed, allowing the temperature inside the furnace to reach a constant value at the time of the analyte atomization. Slavin et al. 12 showed that the use of L vov s platform alone would not be enough to reduce more severe interferences 4

19 and provide accurate results. These authors proposed the stabilized temperature platform furnace method (STPF), which aggregated several additional actions: fast heating and digital electronics to accurately capture the transient absorption signal; integrated area rather than peak height absorbance measurement; pyrolytic graphite coated tubes; solid pyrolytic graphite platforms; Zeeman-effect background correction; and the use of chemical modifiers. Since then, the STPF method has become the standard approach for GFAAS analyses. This method has been applied successfully for determination of most elements in the periodic table. 6, 9 However, GFAAS is a relatively expensive method with some drawbacks. For example, some metals cannot be determined due to the formation of stable, refractory metallic carbides at high temperatures. Metallic atomizers are a less expensive alternative to graphite furnaces. Although not as widely used as graphite devices, metal tubes, wires, loops, platforms and coils have been used in different applications as atomizers and 5, 13 vaporizers to determine several different elements. Some advantages of using this type of atomizer when compared to graphite furnaces are: lower thermal gradient, lower power input requirements, carbide-free species, higher heating rates, no cooling system required, long atomizer lifetime, and lower background signal at the UV region. 14 One of the first works on metallic atomizers used a platinum loop to determine Cd, Ga, and Hg by atomic fluorescence. 15 In 1970, Donega and Burges 16 used a tungsten boat to determine several elements by eletrothermal atomic absorption spectrometry (ETAAS). In 1972, Williams and Piepmeier 17 were the first to propose the use of 5

20 a tungsten filament extracted from a light bulb as the atomizer for atomic spectrometry. The filament from a 24 W commercial light bulb was coupled to an AA spectrometer and several elements were determined with LOD s as low as 10 pg for Cu, and 20 pg for Ca. Despite its advantages, AAS with metallic atomizers has never reached the popularity of graphite furnace devices. Fast electronics are essential to capture rapid transient signals produced with such small mass, fast heating rate atomizers. In the 1970 s, fast electronics were not available and methods using metallic atomizers were rarely reported. Detector response times (500 ms or longer) 18 were incompatible with the fast transient signals produced. The evolution of electronics in the 1980 s enabled the capture of these transient signals. In 1988, Berndt and Schaldach 19 used tungsten coil atomic absorption spectrometry (WCAAS) to determine 13 elements in synthetic urine samples. The authors used a filament extracted from a commercially available 12 V, 150 W projector light bulb enclosed in a protective atmosphere of 90 % Ar and 10 % H 2. Sensitivities for Cd, Co, Cr, Eu, Mn, Ni, Pb, Sn and V were comparable to those reported for commercial GFAAS. Several metals have been used as atomizers in electrothermal atomic 5, 20, 21 spectrometry, but tungsten became the material of choice due to its particular properties. 13 Platinum, for example, has a relatively low melting point (2042 K) and therefore it is not suitable for the determination of refractory elements. Molybdenum and tantalum have high melting points (2895 and 3269 K, respectively) but they become brittle after several heating cycles, and the 6

21 atomizer lifetime is reduced. 5 On the other hand, tungsten is an almost perfect material for resistively-heated atomizers. It has the highest melting point of all elements (3680 K), the lowest vapor pressure at temperatures above 2000 K (2.62 x Pa at 2000 K, and 1.59 x 10-6 Pa at 2400 K), and a very low specific heat (0.132 J g -1 K -1 ). 22 Thus, using a simple, low-voltage power supply, it is possible to reach temperatures as high as 3673 K, at high heating rates, without melting the atomizer. 23 A great advantage of using tungsten is that filaments of this material are mass produced with strict specifications for projector light bulbs (Osram Xenophot 64633, Pullach, Germany). Therefore, little variation from coil to coil is observed, which enables more precise measurements. These filaments are inexpensive, fabricated with high purity tungsten, and have lifetimes reaching about 300 heating cycles. 24 In addition, tungsten is very stable: low reactivity with high resistance to concentrated acids. Since the pioneer works in metallic atomizers, several applications have been developed, especially for tungsten coils. 25, 26 Most predominantly, works in atomic absorption (WCAAS) have been applied to many different samples to determine several metals at the μg L -1 and sub-μg L -1 levels One limitation of tungsten-coil-based methods is the significant temperature gradient observed around the atomizer. 19 Values ºC lower than at the atomizer surface were observed 31 by measuring the gas phase temperature up to 3 mm above the coil. This lower gas phase temperature may increase matrix interferences and reduce both precision and accuracy. 32, 33 An important drawback associated with WCAAS is its single-element nature. 7

22 Although some works on multiple elements have been reported, 34, 35 a different approach is necessary to make tungsten coil atomic spectrometry faster, simpler and portable. An alternative is to detect atomic emission rather than absorption using a high resolution spectrograph and a charged coupled device (CCD) detector. The advantages of such approach are that several elements can be detected simultaneously and a simpler, less bulky instrument may be used since 26, 36 no radiation source is required. Recently, Rust et al. used tungsten coil atomic emission spectrometry (WCAES) to simultaneously determine several elements using a small solid-state, constant current power source, a high resolution Czerny-Turner spectrograph, and a CCD detector. The method is promising, with LOD s at the low μg L -1 level for several elements, and the potential for applications on the field. The instrumental design is simple and the use of an inexpensive, commercially available tungsten coil atomizer extracted from a projector light bulb turn the method into an interesting alternative to more traditional spectrometric methods. Modern analytical chemistry challenges are responsible for a renewed interest in metallic atomizers for atomic spectrometry. An increasing demand for inexpensive, fast, sensitive and portable methods has contributed to the development of new electrothermal atomic spectrometry approaches. Tungstencoil-based atomic spectrometric methods may be an interesting alternative to traditional, more expensive analytical methods. WCAAS and WCAES present sensitivities comparable to ICP OES and GFAAS for several elements with much lower costs. Another great advantage is the possibility of applications in the 8

23 field. 37, 38 The development of a sensitive, portable method could improve the quality of analyses and contribute to fast, more effective actions based on the results. The following chapters describe methods developed to improve and expand tungsten-coil-based atomic spectrometry applications by advancing instrumentation, acquiring basic knowledge on both WCAAS and WCAES atomization processes, and setting the first steps toward portability. New approaches are presented to improve WCAAS sensitivity, precision and accuracy by reducing the effects of sample matrix on the analytical signal. In addition, new WCAES instrumentation advances are presented and the method is used to simultaneously determine several elements at low levels in relatively complex samples after simple sample preparation procedures. The strategies explored include power output enhancement by using two atomizers, indirect determinations by monitoring atomic emission attenuation, and atomic cloud concentration in a newly developed atomization micro-cell. As a result, some difficult elements are detected for the first time by WCAES, including Ti, V and I. WCAES is both a sensitive, portable method for field applications and a less expensive alternative for atomic spectrometric analyses. 9

24 REFERENCES 1. Feldmann, J.; Cullen, W. R. Occurrence of volatile transition metal compounds in landfill gas: synthesis of molybdenum and tungsten carbonyls in the environment. Environ. sci. Technol. 1997, 31(7), Merian, E.; Anke, M.; Ihnat, M.; Stoeppler, M. Metals and Their Compounds. 2nd ed.; Wiley-VCH: Weinheim, 2004; Vol. 2, p Harris, D. C. Solutions Manual for Quantitative Chemical Analysis 5th ed.; W. H. Freeman: New York, 1998; p Schimid, L. S. Spectroscopy: a new technology for all seasons. Spectrosc. 2008, 23(3), Zhong, G.; Luo, H.; Zhou, Z.; Hou, X. Molybdenum, platinum, and tantalum metal atomizers or vaporizers in analytical atomic spectrometry. Appl. Spectrosc. Rev. 2004, 39(4), Welz, B.; Sperling, M. Atomic Absorption Spectrometry 3rd ed.; Wiley-VCH: Weinheim, 1999; p

25 7. Caruso, J. A.; Sutton, K. L.; Ackley, K. L. Elemental Speciation. 1st ed.; Elsevier Science: New York, 2000; Vol. 33, p L'vov, B. V. The analytical use of atomic absorption spectra. Spectrochim. Acta, Part B 1961, 17(7), Jackson, K. W. Electrothermal Atomization for Analytical Atomic Spectrometry. 1st ed.; John Wiley & Sons: Chichester, 1999; p Massmann, H. Vergleich von atomabsorption und atomfluoreszenz in der graphitküvette. Spectrochim. Acta, Part B 1968, 23(4), L'vov, B. V.; Pelieva, L. A.; Sharnopolsky, A. I. Zh. Prikl. Spektrosk. 1977, 27, Slavin, W.; Manning, D. C.; Carnrick, G. R. The stabilized temperature platform furnace. At. Spectrosc. 1981, 2, Nóbrega, J. A.; Silva, M. M.; Oliveira, P. V.; Krug, F. J.; Baccan, N. Espectrometria atômica com atomização eletrotérmica em superficies metálicas. Quím. Nova 1995, 18(6),

26 14. Suzuki, M.; Ohta, K. Electrothermal atomic absorption spectrometry with metal atomizers. Prog. Analyt. Atom. Spectrosc. 1983, 6, Bratzel, Jr. M. P.; Dagnall, R. M.; Winefordner, J. D. A New, simple atom reservoir for atomic fluorescence spectrometry. Anal. Chim. Acta 1969, 48(2), Donega, H. M.; Burgess, T. E. Atomic absorption analysis by flameless atomization in a controlled atmosphere. Anal. Chem. 1970, 42(13), Williams, M.; Piepmeier, E. H. Commercial tungsten filament atomizer for analytical atomic spectrometry. Anal. Chem. 1972, 44(7), Dipierro, S.; Tessari, G. Determination of nanogram amounts of nickel by flameless atomic-absorption spectroscopy. Talanta 1971, 18(7), Berndt, H.; Schaldach, G. Simple low-cost tungsten-coil atomiser for electrothermal atomic absorption spectrometry. J. Anal. At. Spectrom. 1988, 3(5), Hwang, J. Y.; Ullucci, P. A.; Smith, Jr. S. B.; Malenfant, A. L. Microdetermination of lead in blood by flameless atomic absorption spectrometry. Anal. Chem. 1971, 43(10),

27 21. Ohta, K.; Suzuki, M. Trace metal analysis of rocks by flameless atomicabsorption spectrometry with a metal micro-tube atomizer. Talanta 1975, 22(4-5), Lide, D. R., Ed., CRC Handbook of Chemistry and Physics. 88th ed.; CRC: Boca Raton, 2008; p Levine, K.; Wagner, K. A.; Jones, B. T. Low-cost, modular electrothermal vaporization system for inductively coupled plasma atomic emission spectrometry. Appl. Spectrosc. 1998, 52(9), Salido, A.; Sanford, C. L.; Jones, B. T. Determination of lead in blood by chelation with ammonium pyrrolidine dithio-carbamate followed by tungsten-coil atomic absorption spectrometry. Spectrochim. Acta, Part B 1999, 54(8), Hou, X.; Levine, K. E.; Salido, A.; Jones, B. T.; Ezer, M.; Elwood, S.; Simeonsson, J. B. Tungsten coil devices in atomic spectrometry: absorption, fluorescence, and emission. Anal. Sci. 2001, 17(1), Rust, J. A.; Nóbrega, J. A.; Calloway, Jr. C. P.; Jones, B. T. Advances with tungsten coil atomizers: continuum source atomic absorption and emission spectrometry. Spectrochim. Acta, Part B 2005, 60(5),

28 27. Lopes, G. S.; Nogueira, A. R. A.; Oliveira P. V.; Nóbrega, J. A. Determination of cobalt in animal feces by tungsten coil atomic absorption spectrometry. Anal. Sci. 1999, 15(2), Wagner, K. A.; Levine, K. E.; Jones, B. T. A simple, low cost, multielement atomic absorption spectrometer with a tungsten coil atomizer. Spectrochim. Acta, Part B 1998, 53(11), Silva, J. C. J.; Garcia, E. E.; Nogueira, A. R. A.; Nóbrega, J. A. Determination of dysprosium and europium in sheep faeces by graphite furnace and tungsten coil electrothermal atomic absorption spectrometry. Talanta 2001, 55(4), Ribeiro, A. S.; Arruda, M. A. Z.; Cadore, S. Espectrometria de absorção atômica com atomização eletrotérmica em filamento de tungstênio: uma re-visão crítica. Quím. Nova 2002, 25(3), Bruhn, C. G.; Ambiado, F. E.; Cid, H. J.; Woerner, R.; Tapia, J.; Garcia, R. Analytical evaluation of a tungsten coil atomizer for cadmium, lead, chromium, manganese, nickel and cobalt determination by electrothermal atomic absorption spectrometry. Anal. Chim. Acta 1995, 206(2-3),

29 32. Oliveira, P. V.; Krug, F. J.; Silva, M. M.; Nóbrega, J. A.; Queiroz, Z. F.; Rocha, F. R. P. Influence of Na, K, Ca and Mg on lead atomization by tungsten coil atomic absorption spectrometry. J. Braz. Chem. Soc. 2000, 11(2), Queiroz, Z. F.; Krug, F. J.; Oliveira, P. V.; Silva, M. M.; Nóbrega, J. A. Electrothermal behavior of sodium, potassium, calcium and magnesium in a tungsten coil atomizer and review of interfering effects. Spectrochim. Acta, Part B 2002, 57(1), Salido, A.; Jones, B. T. Simultaneous determination of Cu, Cd and Pb in drinking-water using W-Coil AAS. Talanta 1999, 50(3), Rust, J. A.; Nóbrega, J. A.; Calloway, Jr. C. P.; Jones, B. T. Analytical characteristics of a continuum-source tungsten coil atomic absorption spectrometer. Anal. Sci. 2005, 21(8), Rust, J. A.; Nóbrega, J. A.; Calloway, Jr. C. P.; Jones, B. T. Tungsten coil atomic emission spectrometry. Spectrochim. Acta, Part B 2006, 61(2), Batchelor, J. D.; Thomas, S. E.; Jones, B. T. Determination of cadmium with a portable, battery-powered tungsten coil atomic absorption spectrometer. Appl. Spectrosc. 1998, 52(8),

30 38. Sanford, C. L.; Thomas, S. E.; Jones, B. T. Portable, battery-powered, tungsten coil atomic absorption spectrometer for lead determinations. Appl. Spectrosc. 1996, 50(2),

31 CHAPTER II DETERMINATION OF Cd IN URINE BY CLOUD POINT EXTRACTION-TUNGSTEN COIL ATOMIC ABSORPTION SPECTROMETRY George L. Donati, Kathryn E. Pharr, Clifton P. Calloway, Jr., Joaquim A. Nóbrega and Bradley T. Jones The following manuscript was published in Talanta, volume 76, pages , 2008 (DOI /j.talanta ), and is reprinted with permission. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati and Kathryn E. Pharr. The manuscript was prepared by George L. Donati and edited by Bradley T. Jones. 17

32 ABSTRACT Cadmium concentrations in human urine are typically at or below the 1 μg L -1 level, so only a handful of techniques may be appropriate for this application. These include sophisticated methods such as graphite furnace atomic absorption spectrometry and inductively coupled plasma mass spectrometry. While tungsten coil atomic absorption spectrometry is a simpler and less expensive technique, its practical detection limits often prohibit the detection of Cd in normal urine samples. In addition, the nature of the urine matrix often necessitates accurate background correction techniques, which would add expense and complexity to the tungsten coil instrument. This manuscript describes a cloud point extraction method that reduces matrix interference while preconcentrating Cd by a factor of 15. Ammonium pyrrolidinedithiocarbamate and Triton X-114 are used as complexing agent and surfactant, respectively, in the extraction procedure. Triton X-114 forms an extractant coacervate surfactant-rich phase that is denser than water, so the aqueous supernatant is easily removed leaving the metalcontaining surfactant layer intact. A 25 µl aliquot of this preconcentrated sample is placed directly onto the tungsten coil for analysis. The cloud point extraction procedure allows for simple background correction based either on the measurement of absorption at a nearby wavelength, or measurement of absorption at a time in the atomization step immediately prior to the onset of the Cd signal. Seven human urine samples are analyzed by this technique and the results are compared to those found by the inductively coupled plasma mass 18

33 spectrometry analysis of the same samples performed at a different institution. The limit of detection for Cd in urine is 5 ng L -1 for cloud point extraction tungsten coil atomic absorption spectrometry. The accuracy of the method is determined with a standard reference material (toxic metals in freeze-dried urine) and the determined values agree with the reported levels at the 95 % confidence level. Keywords: Tungsten coil; Atomic absorption; Cloud point extraction; Cadmium; Urine 19

34 1. INTRODUCTION Urinary cadmium concentration is considered the best biomarker for total body burden, past exposure, and renal accumulation [ 1,2]. As a result, this particular sample type has been a common target for analytical method development [ 3 ]. However, severe matrix interference and very low analyte concentration contribute to the challenging nature of this problem. The most sensitive and accurate urine-cd analyses are therefore performed by state of the art techniques such as graphite furnace atomic absorption spectrometry (GFAAS) or inductively coupled plasma mass spectrometry (ICP-MS). Simpler, less expensive techniques such as tungsten coil atomic absorption spectrometry (WCAAS) often lack the sophistication necessary to account for the complicated background presented by the urine matrix, and this problem is exacerbated by limits of detection (LODs) that approach normal urine-cd levels even in the absence of a complicated matrix. Previously reported Cd LODs for WCAAS devices vary from 0.04 to 3 μg L -1 [4,5,6,7,8], in relatively simple sample matrices (Table 1). Preconcentration techniques employing fullerene C-60 [9] and Chelex-100 [10] provide LODs that are lower still: and μg L -1 respectively when relatively simple sample matrices are analyzed. While not yet applied to WCAAS, another powerful preconcentration technique is cloud point extraction (CPE). This method has been employed for the determination of Cd in human hair by flame atomic absorption spectrometry with an LOD of 3 μg L -1 [11], and for the determination of Cd in blood by GFAAS with an LOD of 0.02 μg L -1 20

35 [ 12]. The work described below is the first application of a combined CPE- WCAAS method. The CPE step reduces the WCAAS LOD to the point where normal urine-cd levels may be detected. At the same time, the urine matrix is reduced so that simple background correction techniques may be employed. 21

36 2. EXPERIMENTAL 2.1. Instrumental The WCAAS system was designed and assembled in the laboratory, and is similar to those reported elsewhere [4,5,13]. The atomizer was the 10-turn tungsten filament extracted from a 150 W, 15 V commercially-available slide projector light bulb (Osram Xenophot HXL, Pullach, Germany). The fused silica bulb envelope was removed, leaving the filament and bulb base intact. The bulb base was mounted in a standard ceramic two-pronged power socket. Positioning of the filament from coil to coil was highly reproducible since the bulbs were mass produced to strict optical specifications. The filament was housed inside a glass atomization cell with fused silica windows. A purge gas composed of 10 % v/v H 2 /Ar, at a 1.0 L min -1 flow rate, prevented coil oxidation and cooled the filament after atomization. Power was provided by a programmable, constant-current, solid-state DC power source (BatMod, Vicor, Andover, MA, USA). The output from the radiation source (10 W Cd EDL, Perkin- Elmer, Norwalk, CT, USA) was focused through the W-coil atomizer using a 25 mm diameter, 75 mm focal length fused silica lens. The unabsorbed radiation exiting the opposite end of the atomization cell was collected with a second identical fused silica lens, and imaged onto the entrance slit of a crossed Czerny- Turner monochromator (MonSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). The monochromator was equipped with a

37 grooves mm -1 grating (52 x 40 mm), resulting in a reciprocal linear dispersion of 2.4 nm mm -1 at 230 nm. The detector was a thermoelectrically-cooled charge coupled device (CCD, Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA) with a two-dimensional array of 1340 x 100 pixels. Each pixel was 20 x 20 μm in size, making the image area on the CCD camera 26.8 mm x 2 mm. Based on a slit width approximately the size of a single detector pixel (25 μm), the theoretical spectral bandpass of the system was 0.05 nm. In practice, the measured full width at half maximum (FWHM) for an emission line from the lamp was 2-3 pixels, so the practical bandpass was 0.15 nm. The spectral window covered by the CCD detector was slightly less than 65 nm, and the central wavelength was set by adjusting the monochromator wavelength dial. User selectable CCD integration times as low as 1 ms were possible, and the maximum signal to noise ratio (S/N) was observed with an integration time of 20 ms. One hundred successive spectra were recorded during the atomization step. Calibration was performed in two modes: peak height (absorbance signal measured for the single spectrum collected 500 ms after the onset of atomization) and peak area (absorbance signals summed for the 12 spectra corresponding to the period of greatest absorption, ms after the onset of atomization) Standards and reagents Reference solutions were prepared by diluting a Cd stock solution (1000 mg L -1, SPEX CertPrep, Metuchen, NJ, USA) with distilled-deionized water (Milli- 23

38 Q, Millipore, Bedford, MA, USA). A standard reference material from the National Institute of Standards and Technology (Toxic metals in freeze-dried urine, normal level, SRM 2670, NIST, Gaithersburg, MD, USA) was used to check the method accuracy. Trace metal grade HNO 3 14 mol L -1 (Fisher Scientific, Pittsburg, PA, USA) and H 2 O 2 30 % v/v (Acros, Morris Plains, NJ, USA) were used in the microwave-assisted sample digestions. Ammonium pyrrolidinedithiocarbamate (APDC, Sigma-Aldrich, St. Louis, MO, USA) and octylphenoxypolyethoxyethanol (Triton X-114, Sigma-Aldrich) were used as complexing agent and surfactant, respectively, in the cloud point extraction. Solutions of APDC (0.5 % m/v) and TX-114 (5 % v/v) were prepared by diluting appropriate amounts of the reagents with distilled-deionized water (Millipore). All ph adjustments before the cloud point extraction were performed with trace metal grade ammonium hydroxide (Fisher) Sample preparation Sample aliquots of 3.0 ml urine were transferred to PTFE microwave digestion vessels. Volumes of 1.0 ml concentrated HNO 3 and 0.50 ml of H 2 O 2 were added to the samples, and the solutions were submitted to a five-step heating program in the microwave oven (630 W CEM MDS 2000 closed-vessel microwave oven, CEM Corp. Matthews, NC, USA): 25 % power for 2 min, 0 % for 2 min, 35 % for 3 min, 50 % for 4.5 min, and 60 % for 6 min. After cooling, the solutions were transferred to 15-mL centrifuge tubes and neutralized with NH 4 OH 24

39 to ph 6.0. Then, the solutions were diluted to 10 ml with distilled-deionized water, and subjected to cloud point extraction. The CPE procedure was based on the optimal conditions suggested in the literature [14,15]. A 0.5 ml aliquot of APDC (0.5 % m/v) was added to the 10 ml diluted/digested sample volume. After 20 min, 0.5 ml of TX-114 (5 % v/v) was added, and the solution was heated in a water bath for 20 min at 45 o C. To facilitate phase separation, the solution was centrifuged at 2500 rpm (660 x g) for 10 min. Then, the solution was chilled in a freezer for 20 min (at 10 o C) to increase the viscosity of the lower TX-114 surfactant-rich phase, which in turn facilitated the removal of the supernatant (aqueous surfactant-poor phase) with a micropipette. Chilling for longer than 20 min risked loss of Cd, since the solution would eventually reform a one phase homogeneous mixture at temperatures below the cloud point. The final volume of the organic phase was then increased to 0.20 ml with the addition of ethanol, which served to reduce the viscosity of the surfactant phase. This step facilitated the transfer of the preconcentrated solution to the W-coil. Standard Cd solutions were treated in the same fashion. The CPE method resulted in a concentration factor of 15 (3.0 ml of urine reduced to 0.20 ml of extract) WCAAS Analysis A 25 μl volume of the CPE preconcentrated sample (or standard) was transferred to the W-coil using a micropipette. The atomization cycle consisted of 25

40 a 7-step heating program (Table 2). The first two steps accounted for gradual solvent volatilization to prevent potential sample loss. The coil reached dryness at the end of step 2, as evidenced by an increase in the applied voltage necessary to maintain a constant current during the last few seconds of this stage. Since a wet coil had a lower resistance, a dry coil reached a higher temperature at a given constant current [16]. Therefore, the ashing stage of the cycle, steps 3 and 4, employed lower currents. Using progressively lower currents prevented the W-coil from glowing red prior to atomization, thus reducing a potential loss of analyte. A cooling period (step 5) ensured a reproducible high temperature atomization step, as the beginning temperature was always the same (near room temperature). Finally, the high current (10 A) atomization step generated the atomic cloud. This current was the maximum available from the power supply, and therefore produced the greatest possible peak heights. These peak heights also allowed for shorter detector integration times, which resulted in reduced background blackbody emission measurements from the coil. The detector was triggered at the beginning of the atomization step (step 6), and one hundred spectra were collected automatically. The final cooling step readied the W-coil for the deposition of the next sample aliquot ICP-MS Analysis For comparison purposes, the seven human urine samples analyzed by WCAAS were also transported (frozen) to a different laboratory for ICP-MS 26

41 analysis: Research Triangle Institute, Research Triangle Park, NC. For the ICP- MS method, the urine samples were thawed and then vortexed for 15 s. Sample aliquots of 2.0 ml were transferred to digestion tubes. Volumes of 1.0 ml of concentrated HNO 3 and 0.25 ml of H 2 O 2 were added to each tube. The tubes were then placed in a DigiPreP MS heating block (SCP Science, Champlain, NY, USA) and heated in the following stages: 60 min at 60 o C (ramp time 30 min), 60 min at 80 o C (ramp time 10 min), and 120 min at 110 o C (ramp time 20 min). After cooling, the sample solutions were diluted with distilled-deionized water to 10.0 ml and vortexed for 15 s. Finally, a 50.0 μl aliquot of a 89 Y internal standard solution (1 mg L Y solution in 2 % v/v HNO 3 ) was added to a 5.00 ml aliquot of the digested samples. The resulting solution was analyzed with a quadrupole ICP-MS instrument (Thermo X Series II X0637, Waltham, MA, USA). 27

42 3. RESULTS AND DISCUSSION 3.1. Background correction Figure 1 depicts the absorption profiles collected during the high temperature atomization step for the WCAAS analysis of a cloud point extracted urine sample. The 10 amp atomization step began at time zero, and spectra were collected at 20 ms intervals. Two emission lines from the Cd EDL were monitored simultaneously: the nm Cd absorption line which is sensitive to both Cd and background absorption, and the nm line which is not sensitive to Cd and thus reflects only the background absorption. The profile represented by circles is the background absorbance measured at the nm wavelength. The profile depicted in squares represents the absorbance measured at the nm line minus the absorbance measured at the nm line. This profile, therefore, represents the background corrected Cd signal, assuming that the background absorbance is the same at the two wavelengths. Notice that the background profile contains two features, a peak with maximum occurring at 480 ms and a second one with maximum at 580 ms. The first peak appears only for samples containing Triton X-114, so it may be attributed to the CPE process. The second peak occurs only for extracted urine samples, so it may be attributed to those components, other than Cd in urine, that are also concentrated by the CPE technique. Pure aqueous samples produce neither background peak, and CPE extracts of aqueous solutions 28

43 produce only the first peak. These peaks are generally the same size and shape at nm, with the Cd peak positioned in between, so that the single Cd profile in the figure results from their subtraction. This background-corrected urine-cd signal is almost identical in time and shape to that observed for the WCAAS analysis of purely aqueous Cd standard solution. This off-line, or near-line, background correction technique has been reported previously for WCAAS determinations of Cd [5,6], and in fact tables with suggested close lines for background correction for many elements have been published [17]. While near-line correction has provided reasonable analytical results for simple Cd-containing samples, the complex urine matrix may affect the background spectrum such that the background absorbance measured at nm differs from the background that might be present at nm. While the background for the relatively clean cloud point extracts is expected to vary very little across a 2.3 nm region, the difference could become significant at low Cd concentrations. High resolution continuum source atomic absorption spectra for urine samples show that overall background absorbance levels may differ by orders of magnitude, even for samples with similar Cd levels [18]. For the high background samples reported in that work, the signal variation over a 2.3 nm range is clearly measurable. In an effort to correct for this potential error, a second background correction approach was investigated. The surfactant background peak (Figure 1) was very reproducible, regardless of sample type or Cd level. Notice in the profile that the background absorbance measured at 460 ms, prior to the appearance of the Cd signal, closely approximates the 29

44 background signal measured at 500 ms (black diamonds). This is also the case for absorption profiles measured at nm for CPE samples containing no Cd. Therefore, the background absorbance level measured at 460 ms for the Cd nm absorption line closely approximates the background level present at the time of maximum Cd signal (500 ms). Thus a temporal background correction may be applied using only the nm line, by subtracting the absorbance values measured at these two times. While neither near-line nor temporal correction may be expected to match the accuracy observed for conventional background correction techniques (using the Zeeman effect or a continuum source), both can be performed without adding complexity to the WCAAS system. Continuum source atomic absorption spectrometry has also identified structured background spectra in the vicinity of the Cd nm line for very complex matrices like those present in solid coal samples [ 19 ]. While the molecular species giving rise to these spectra is not identified, one might speculate that similar structure in a urine spectrum may arise from the classic phosphate interference [20]. To test for the possibility of phosphate interference with the CPE-WCAAS procedure, a 5 μg ml -1 phosphate solution was subjected to the CPE procedure described above. ICP emission analysis of the aqueous supernatant revealed that at most 20 % of the phosphate was extracted into the surfactant layer. The phosphate present in a urine sample, therefore, had the potential to produce an interference. However, a 10 μg ml -1 solution of phosphate produced no measurable absorbance at nm by WCAAS. 30

45 3.2. Urine Analyses The CPE-WCAAS method resulted in a urine-cd LOD (3 σ) of μg L -1, which is in line with the other WCAAS LODs using preconcentration techniques (Table 1), and roughly a factor of 100 lower than normal urine-cd levels. The CPE-WCAAS analysis of NIST standard reference material #2670 (toxic metals in freeze-dried urine normal level) resulted in 0.38 ± 0.03 μg L -1 (mean ± standard deviation, n=3) using the near-line correction method, and 0.36 ± 0.05 μg L -1 using the temporal correction method. The NIST reported Cd level for SRM #2670 is 0.40 μg L -1. Figure 2 shows the urine-cd levels measured for the seven human urine samples. The amount of Cd was determined in each sample by applying the near line correction technique (white bars) and the temporal background correction technique (cross-hatched bars). These two values were determined for exactly the same runs (n=3). The ICP-MS results (gray bars) were collected at a different time using a different sample preparation technique as described above. The error bars represent the precision for each method (± 1σ, n=3). Clearly the ICP- MS technique provides more precise results. The average standard deviation observed for the seven samples was 0.03 μg L -1 by ICP-MS, compared to 0.12 and 0.15 μg L -1 by CPE-WCAAS using near-line and temporal background correction respectively. Compared with ICP-MS results, the CPE-WCAAS technique with near-line correction under-reported the urine-cd levels for the seven samples by an average of μg L -1. In one case (sample #2) a 31

46 completely erroneous and negative urine-cd result was reported (-0.16 μg L -1 ). Perhaps this trend can be explained by the background absorbance value at nm being consistently higher than that measured at nm. Such a result would not be surprising, since background absorption tends to increase with decreasing wavelength. The CPE-WCAAS technique with temporal background correction did not show a consistent systematic error. Compared to the ICP-MS results, the temporal-corrected values differed by just μg L -1 on average. ACKNOWLEDGEMENTS This material is based upon work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET The authors would also like to thank Keith E. Levine of the Research Triangle Institute, Research Triangle Park, North Carolina, for providing access to the ICP-MS system. 32

47 Table 1. Previously Reported WCAAS Limits of Detection for Cadmium. Matrix Limit of Detection μg L -1 Reference Water 3 [4] Water 0.2 [5] Water 0.1 [6] 0.2 % HNO [7] Mussel Acid Digests 0.04 [8] Fullerene C-60 Extract from Water Chelex-100 Extract from Beverages [9] 0.01 [10] 33

48 Table 2. Tungsten coil heating cycle for CPE urine samples. Step Current (A) Temperature a (ºC) Time (s) b b c a. Temperature estimated by: T = 309 A + 52 (Reference [16]). b. In steps 1 and 2, the coil is wet, so while the temperature must be below 100, it cannot be calculated using the above equation. c. Data was collected during this step. 34

49 Figure 1. Absorption profiles collected during the WCAAS analysis of a urine sample. The background profile (circles) is simply the absorbance measured at nm. The analytical profile (squares) is the absorbance measured at nm minus the absorbance measured at nm Cadmium Urine Matrix Background Absorbance Surfactant Background Atomization Time (s) 35

50 Figure 2. Cadmium levels determined in seven urine samples by WCAAS with near-line background correction (white bars), WCAAS with temporal background correction (cross-hatched bars), and by ICP-MS (gray bars). The error bars represent ± one standard deviation (n=3) Cd Found (μg L -1 ) Sample Number 36

51 REFERENCES [1] J. Godt, F. Scheidig, C. Grosse-Siestrup, V. Esche, P. Brandenburg, A. Reich and D. A. Groneberg, J. Occup. Med. Toxicol., 1 (2006) 1-6. [2] H. Choudhury, T. Harvey, W. C. Thayer, T. F. Lockwook, W. M. Stiteler, P. E. Goodrum, J. M. Hassett and G. L. Diamond, J. Toxicol. Environ. Health, Part A 63 (2001) [3] A. C. Davis, P. Wu, X. Zhang, X. Hou and B. T. Jones, Appl. Spectrosc. Rev., 41 (2006) [4] J. D. Batchelor, S. E. Thomas and B. T. Jones, Appl. Spectrosc., 52(8) (1998) [5] A. Salido and B. T. Jones, Talanta, 50 (1999) [6] K.A. Wagner, K.E. Levine, B.T. Jones, Spectrochim. Acta Part B, 53 (1998) [7] C.G. Bruhn, F.E. Ambiado, H.J. Cid, R. Woerner, J. Tapia, R. Garcia, Anal. Chim. Acta, 306 (1995) [8] C.G. Bruhn, N.A. San Francisco, J.Y. Neira, J.A. Nobrega, Talanta, 50 (1999) [9] M.M. Silva, M.A.Z. Arruda, F.J. Krug, P.V. Oliveira, Z.F. Queiroz, M. Gallego, M. Valcarcel, Anal. Chim. Acta, 368 (1998) [10] C.G. Bruhn, F.E. Ambiado, H.J. Cid, R. Woerner, J. Tapia, R. Garcia, Quim. Anal., 15 (1996) [11] J. Manzoori, A. Bavili-Tabrizi, Anal. Chim. Acta., 470 (2002)

52 [12] D.L.G. Boges, M.A.M.S. da Veiga, V.L.A. Frescura, B. Welz, A.J. Curtius, J. Anal. Atom. Spectrom., 18 (2003) [13] C.L. Sanford, S.E. Thomas, B.T. Jones, Appl. Spectrosc. 50 (1996) [14] P. Wu, Y. Zhang, Y. Lv and X. Hou, Spectrochim. Acta Part B, 61 (2006) [15] G. L. Donati, C. C. Nascentes, A. R. A. Nogueira, M. A. Z. Arruda and J. A. Nóbrega, Microchem. J., 82 (2006) [16] K. E. Levine, K. A. Wagner and B. T. Jones, Appl. Spectrosc., 52 (1998) [17] J. Sneddon, Spectroscopy, 2 (1987) 38. [18] R. Fernando, F.K. Ennever, B.T. Jones, Appl. Spectrosc. 47 (1993) [19] A.F. da Silva, D.L.G. Borges, F.G. Lepri, B. Welz, A.J. Curtius, U. Heitmann, Anal. Bioanal. Chem., 382 (2005) [20] K. Saeed, Y. Thomassen, Anal. Chim. Acta, 130 (1981)

53 CHAPTER III EVALUATION OF EXPERIMENTAL DESIGN AND MULTIVARIATE CALIBRATION AS STRATEGIES TO REDUCE MATRIX INTERFERENCES IN TUNGSTEN COIL ATOMIC ABSORPTION SPECTROMETRY George L. Donati, Joaquim A. Nóbrega and Bradley T. Jones The following manuscript was prepared for submission to Analytical Sciences. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati. The manuscript was prepared by George L. Donati and edited by Bradley T. Jones. 39

54 ABSTRACT Experimental design and multiple linear regression are used to reduce matrix interferences in WCAAS. A factorial experiment is applied to evaluate the effects of some common concomitants in water samples such as Ca, K, and Na on Cu signal. Multiple linear regression is used to determine a mathematical model relating experimental Cu concentrations and actual values. The accuracy of the method is verified by determining Cu in a standard reference material (SRM 1643e) and in a spiked water sample. No statistically significant difference is observed between certified and determined values at a 95% confidence level. A 97% recovery is obtained for the spiked sample. By applying the calibration method of acid-based standards to the same samples a concentration 82% smaller than the certified value and a recovery of 27% were obtained. The method proposed is applied to 7 water samples with different concentrations of analyte and concomitants. 40

55 INTRODUCTION The use of tungsten coil atomizers in atomic spectrometry represents an interesting alternative to other traditional electrothermal methods due to characteristics such as simplicity, high sensitivity and flexibility to be applied to different methods. 1-3 Tungsten filaments can be extracted from mass-produced, commercially available 150 W, 15 V light bulbs, which ensures reproducibility and low cost. The low mass and low specific heat of the tungsten coil enable for fast heating and apparatus simplification since no cooling system is necessary. The possibility of using a CCD detector and a small power supply has allowed a portable arrangement and the application of tungsten coil atomic absorption spectrometry (WCAAS) in the field. 4 Nevertheless, some inherent characteristics of tungsten coil atomizers can play an important role in WCAAS limitations. The non-isothermal atmosphere around the coil and the relatively low temperature of the gas phase contribute to severe matrix interferences. 5 Generation of oxides resulting in a smaller population of analyte atoms in the atomic cloud and non-specific molecular absorption are some probable causes of such interferences. On the other hand, condensed-phase formation of stable analyte-concomitant compounds also has a negative effect on both WCAAS sensitivity and accuracy, especially considering samples with high concomitant concentrations. 1,5-6 These effects can be minimized by using a purge gas composed of a low percentage of H 2 dissolved in Ar. However, matrix matched procedures are still required to correct for matrix 41

56 interferences. 7 Considering the complexity and variability in the composition of real samples, the application of a multicomponent matrix matched method could represent a tedious and time-consuming task. Additionally, analyzing a great number of constitutionally different samples by such a method could represent economical and ecological disadvantages. A different approach to reduce matrix interferences and improve accuracy considers experimental desig and multivariate calibration. Andrade et al. 8 present an extensive review on such an approach applied to atomic spectrometry techniques For AAS the most common method used is based on multiple linear regression (MLR). Results obtained for 5 elements determined by FAAS with reduction of matrix interferences by MLR calibration were compared to those from ICP OES and XRF in tungsten carbide samples. No statistically significant difference was observed between the results. 15 In another work, Grotti et al. 16 applied multiple linear regression modeling to reduce interferences in Fe, Mn and Cd determination by GFAAS. Results obtained for the MLR method were not statistically different from the ones obtained by other traditional calibration methods such as acid-based standards, standard addition, and matrix matched standards. The present work evaluates the possibility of using experimental desig combined to MLR modeling as an alternative strategy to reduce matrix interferences in WCAAS. The method proposed is applied for Cu determination in water samples. Interferences on Cu absorption signal caused by common concomitants such as Ca, K and Na are studied. 42

57 EXPERIMENTAL Apparatus A tungsten coil filament extracted from a 150 W, 15 V commercially available slide projector light bulb (Osram Xenophot HXL, Pullach, Germany) was used as atomizer in the WCAAS set-up. The bulbs are mass produced to strict optical specifications, so the positioning of the filament from bulb-to-bulb is highly reproducible. The fused silica envelope protecting the filament was removed, leaving coil and bulb base intact. The bulb base was mounted in a standard ceramic two-pronged power socket and the filament was housed inside a glass atomization cell with fused silica windows. Power was provided by a programmable, constant-current, solid-state DC power source (BatMod, Vicor, Andover, MA, USA). A schematic diagram depicting the WCAAS system is presented in Figure 1. The atomization cell and tungsten coil atomizer have been described in detail elsewhere. 3,17 A purge gas composed of 10% v/v H 2 /Ar, flowing at a 1.0 L min -1 rate, was used to prevent coil oxidation and to cool the atomizer. A Cu hollow cathode lamp (HCL, Fisher Scientific, Loughborough, Leicestershire, UK) was used as radiation source. The output from the HCL lamp was focused through the tungsten coil atomizer using a 25 mm diameter, 75 mm focal length fused silica lens. The non-absorbed radiation exiting the opposite end of the atomization cell was collected with a second identical fused silica lens, and imaged onto the 43

58 entrance slit of a crossed Czerny-Turner monochromator (MonSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). The monochromator is equipped with a 2400 gr/mm grating (110 x 110 mm), resulting in a reciprocal linear dispersion of 2 nm/mm at 400 nm. Based on a slit width approximately the size of a single detector pixel (20 μm), the theoretical spectral bandpass of the system is 0.04 nm. The detector is a thermoelectrically-cooled charge coupled device (CCD, Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA). The CCD detector consists of a two-dimensional array of 1340 x 100 pixels. Each pixel is 20 x 20 μm in size, so the image area on the CCD camera is 26.8 x 2 mm. In order to have the greatest signal to noise ratio (S/N) a 20 ms integration time was used for all Cu determinations. One hundred successive spectra were recorded and each two consecutive spectra were averaged to further improve the S/N. The best results were obtained by integrating the absorption signals between ms. A 25 μl volume was adopted as the solution aliquot for all WCAAS measurements. A micropipette (Ependorf μl, Brinkmann Instruments Inc., Westbury, NY, USA) was used to place the aliquot directly onto the tungsten coil. A heating cycle with 7 gradually decreasing current steps 2,18 was used in all Cu determinations. An inductively coupled plasma optical emission spectrometer with an axial view configuration, a cyclonic spray chamber, and a concentric nebulizer (Prodigy/Prism high dispersion ICP, Teledyne Leeman Labs, Hudson, NH, USA) was used to determine the concentrations of concomitants in the water 44

59 samples. Table 1 shows ICP OES operating parameters and WCAAS heating cycle. Standards and samples Reference solutions were prepared by diluting Cu, Ca, K, and Na monoelemental stock solutions (1000 mg L -1, SPEX CertPrep, Metuchen, NJ, USA) with distilled-deionized water (Milli-Q, Millipore, Bedford, MA, USA). A standard reference material from the National Institute of Standards and Technology (Trace elements in water SRM 1643e, NIST, Gaithersburg, MD, USA), and a water sample spiked with 15 μg L -1 of Cu were used to check for accuracy. Copper concentrations in 7 water samples from different regions were determined by the method proposed. Experimental design and multivariate calibration A factorial experiment based on a central composite design 19 was used to evaluate the combined interference effects of Ca, K, and Na on Cu signal in WCAAS. Each variable was evaluated at 5 levels, coded as -2, -1, 0, +1, and +2. For Cu the coded values corresponded to 1.00, 5.75, 10.5, 15.3, and 20.0 μg L -1, respectively. Considering the concomitants at higher levels in the samples when compared to Cu, a logarithmic scale was adopted to attribute coded values to Ca, K, and Na. For each of these elements the coded values corresponded to 0.50, 45

60 1.58, 5.00, 15.8, and 50.0 mg L -1, respectively. Table 2 presents the factorial experiment matrix with all experiments performed and the coded values for each variable. The 2 4 factorial with 16 experiments plus 8 experiments at the star points (experiments 17-24) and 6 more measurements at the central point (experiments 25-30) allowed the determination of the main effects, the quadratic effects and the interactions among the variables. Measurements at the central point were used to estimate the experimental variance, which was used to determine the levels of significance of each variable effect. Copper concentrations were determined using a calibration curve with aqueous standard solutions of the element. Each experiment was carried out in replicate and the results were the averaged concentrations. Data collected were used to obtain a second degree equation relating the variables evaluated: Cu(f) = β 0 + β 1 Cu + β 2 Ca + β 3 K + β 4 Na + β 5 Ca 2 + β 6 K 2 + β 7 Na 2 + β 8 CaK + β 9 CaNa + β 10 KNa (1) where Cu(f) is the average concentration of copper determined by WCAAS (dependent variable), β j are the coefficients for each variable, and Cu, Ca, K, Na are the added concentrations of analyte and concomitants, respectively. The near-line background correction method was used in all WCAAS determinations. 7 Copper lines at nm (analytical line) and nm (BG) were monitored consecutively for background correction. 46

61 RESULTS AND DISCUSSION WCAAS analytical performance Some figures of merit for Cu determination by WCAAS were determined in order to establish a working range for the application of MLR method. Aqueous standard solutions of Cu were used to determine WCAAS limit of detection (LOD), linear dynamic range (LDR), and precision. The limit of detection was calculated as 3 times the standard deviation of the blank solution signal (n = 20) divided by the calibration curve slope. The precision was calculated as the percent relative standard deviation (% RSD) of the absorption signals obtained with a 5 μg L -1 Cu solution (n = 10). The values determined for LOD, LDR, and precision were 0.64 μg L -1, μg L -1, and 3.7 %, respectively. Based on those results, Cu solutions in the range of μg L -1 were used to evaluate the effects of some concomitants such as Ca, K, and Na on Cu determination by WCAAS. Matrix interferences Tungsten coil atomizers are prone to severe matrix interferences and typical concomitants present in water samples such as Ca, K, and Na can affect significantly the results in a WCAAS determination. Due to its specific 47

62 characteristics as an open atomizer method, 20 both gas and condensed-phase interferences can reduce WCAAS sensitivity and accuracy. 5,6 A factorial experiment was used to determine the effects of Ca, K, and Na on Cu determination by WCAAS. Thirty experiments (Table 2) were carried out in a totally random fashion and the results were the averaged Cu concentrations determined by WCAAS from 2 replicates. The method of least squares was used to determine the coefficients of each variable in a second degree equation relating found Cu concentration and added Cu concentration. All calculations were performed using Microsoft Excel. Equation 2 represents the mathematical model with terms at the 95% significance level: Cu(f) = Cu Ca K Ca K 2 (2) where Cu(f) represents the averaged Cu concentration determined by WCAAS in μg L -1, Cu represents the added copper concentration, in μg L -1 ; and Ca and K represent the concentrations of calcium and potassium in mg L -1. From Equation 2, it can be seen that only Ca and K present significant effects on the Cu signal and no interaction among the variables can be considered statistically significant at the 95% confidence level. The quadratic terms for both concomitants indicate that, at higher concentrations, an overestimation of the analyte concentration would be observed, and that both Ca and K present negative and positive effects on the Cu signal depending on their concentrations. As pointed out in previous works 5,20-21 competition for H 2 48

63 molecules, which are necessary to both analyte and concomitants atomization, would be the reason for the negative effects observed for both Ca and K. On the other hand, the positive effects observed could be related to non-specific molecular absorption from compounds formed at high concentrations of concomitants. Model validation and method accuracy To verify the statistical quality of the mathematical model presented in Eq. 2, a cross-validation procedure 22 was performed and the explained variance (EV) was calculated. 16 According to this procedure, each of the 30 experiments was separately removed from the factorial matrix and the remaining 29 results were used to recalculate the model. Then, both experimental and predicted values were compared to determine the explained variance. The efficiency of the model to predict Cu actual concentrations based on experimental results was considered satisfactory since an EV of 96% was obtained in this procedure. The accuracy of the MLR-WCAAS method was evaluated by determining Cu concentration in a standard reference material (Trace elements in water, NIST SRM #1643e). Additionally a water sample spiked with 15 μg L -1 of Cu was analyzed to verify the performance of the method proposed in real samples. Results are presented in Table 3. No dilution was required for the SRM analysis and no statistically significant difference was observed between certified and determined Cu values in a 95% confidence level. For the spiked water sample a 49

64 97% recovery was obtained. Concentration corrections provided by the model were significant. By applying the calibration method of the acid-based standards to the same samples, a Cu concentration 82% lower than the certified value and a recovery of only 27% were obtained. Application to real samples The method proposed was applied to 7 water samples from different regions. Results are presented in Table 3. Copper concentrations varied from 3.28 to 109 μg L -1. Samples NCLT and VAW had to be diluted with distilleddeionized water in order to fit the concentration range used in the MLR model. It is interesting to observe the high concentrations obtained for samples not submitted to water treatment. Concentration of Cu for a sample collected from a well (VAW) was almost 6-fold higher than the average of the other samples. This result could be an indication of groundwater contamination by rain leaching of Cu compounds originally present in the soil or products of anthropogenic activities. 50

65 CONCLUSIONS The combination of experimental design and multiple linear regression is an interesting alternative to reduce matrix interferences in WCAAS. Different from other more traditional calibration methods, MLR considers non-linear relationships and allows for a better understanding of the concomitant effects on the analytical signal. Tungsten coil atomic absorption spectrometry is simple, economical, sensitive and potentially portable. The application of the method proposed is suitable for a great number of samples in which the concentration of concomitants is much higher than the analyte. ACKNOWLEDGEMENTS This material is based upon work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET J.A.N. is thankful to Fundação de Amparo à Pesquisa do Estado de São Paulo for travel funds (Process 07/ ). 51

66 Table 1. Operating parameters for ICP OES and tungsten coil heating cycle. ICP OES operating parameters WCAAS heating cycle Element a Wavelength / nm Step Current b / A Time / s Ca II K I Power / kw Plasma gas flow rate / L min Auxiliary gas flow rate / L min c 10 5 Sample uptake rate / ml min a. I, atomic line; II ionic line. b. The temperature (T) in Kelvin of the dry coil surface may be estimated from the current (I) in amps: T 309 I c. Tungsten coil atomic absorption was collected during this step. 52

67 Table 2. Factorial experiment Experiment Cu Ca Na K Coded Coded Coded Coded

68

69 Table 3. Determination of Cu in water samples by WCAAS with matrix interference reduction by experimental design and multiple linear regression modeling Cu content a / μg l -1 NCHT NCLT CT VAT VAW DCT CAT Spiked b SRM 1643e c Recovery / % Certified / μg l -1 Found / μg l ± ± ± ± ± ± ± ± 3.3 a. NCHT = North Carolina drinking water; NCLT = North Carolina tap water; CT = China drinking water; VAT = Virginia drinking water; VAW = Virginia well water; DCT = D.C. drinking water; CAT = California drinking water. Values are the mean ± 1 standard deviation (n = 3). b. Drinking water spiked with 15 μg l -1 of Cu. c. NIST Standard Reference Material 1643e. Found values are the mean ± 1 standard deviation (n = 5). 55

70 Radiation source (HCL) Lens Lens C C D Lamp power supply 10 % H 2 / Ar Constant current power supply D/A Converter Sample introduction Quartz window Quartz window W coil 10 % H 2 / Ar Power supply Fig. 1. Schematic diagram of WCAAS instrumentation. Inset shows an alternate view of the atomization cell. 56

71 REFERENCES 1. J. A. Rust, J. A. Nóbrega, C. P. Calloway Jr., and B. T. Jones, Spectrochim. Acta Part B, 2005, 60, G. L. Donati, J. Gu, J. A. Nóbrega, C. P. Calloway Jr., and B. T. Jones, J. Anal. At. Spectrom., 2008, DOI /b710600a. 3. J. A. Rust, J. A. Nóbrega, C. P. Calloway Jr., and B. T. Jones, Anal. Chem., 2005, 77, J. D. Batchelor, S. E. Thomas, and B. T. Jones, Appl. Spectrosc., 1998, 52, Z. F. Queiroz, F. J. Krug, P. V. Oliveira, M. M. Silva, and J. A. Nóbrega, Spectrochim. Acta part B, 2002, 57, P. V. Oliveira, F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz, and F. R. P. Rocha, J. Braz. Chem. Soc., 2000, 11, K. A. Wagner, K. E. Levine, and B. T. Jones, Spectrochim. Acta Part B, 1998, 53, J. M. Andrade, M. J. Cal-Prieto, M. P. Gomez-Carracedo, A. Carlosena, and D. Prada, J. Anal. At. Spectrom., 2008, 23, G. Henrion, R. Genrion, R. Hebisch, and B. Boeden, Anal. Chim. Acta, 1992, 268, N. Majcen, Fresenius J. Anal. Chem., 1996, 355, M. Grotti, R. Leardi, and R. Fache, Anal. Chim. Acta, 1998, 376, M. Grotti, Ann. Chim., 2004, 94, 1. 57

72 13. M. Rupprecht and T. Probst, Anal. Chim. Acta, 1998, 358, M. Z. Martin, N. Labbe, T. G. Rials, and S. D. Wullschleger, Spectrochim. Acta Part B, 2005, 60, T. Piippanen, J. Jaatinen, A. Pirjeta, and J. Tummavuori, Fresenius J. Anal. Chem., 1997, 358, M. Grotti, M. L. Abelmoschi, F. Soggia, C. Tiberiade, and R. Fache, Spectrochim. Acta Part B, 2000, 55, A. Salido and B. T. Jones, Talanta, 1999, 50, P. V. Oliveira, M. Catanho, J. A. Nóbrega, and P. O. Luccas, Quim. Nova, 2000, 23, W. G. Cochran and G. M. Cox, Experimental Designs, 1950, Wiley, New York. 20. J. V. Chauvin, D. G. Davis, and L. G. Hargis, Anal. Lett., 1992, 25, B. L vov, Spectrochim. Acta Part B, 1997, 52, M. Stone, J. R. Stat. Soc. Series B, 1974, 36,

73 CHAPTER IV SIMULTANEOUS DETERMINATION OF THE LANTHANIDES BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY George L. Donati, Jiyan Gu, Joaquim A. Nóbrega, Clifton P. Calloway, Jr. and Bradley T. Jones The following manuscript was published in the Journal of Analytical Atomic Spectrometry, volume 23, pages , 2008 (DOI /b710600a), and is reprinted with permission. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati and Jiyan Gu. The manuscript was prepared by George L. Donati and edited by Bradley T. Jones. 59

74 ABSTRACT The fourteen lanthanides are determined by tungsten coil atomic emission spectrometry. Twenty-five microliter sample aliquots are placed directly on the coil. A simple constant current power source carefully dries the sample prior to analysis. During this dry step, the voltage is monitored to prevent over heating. This allows for shorter atomization programs while improving sensitivity and coil lifetime. During the 5 s high temperature atomization step, the emission signals for as many as seven lanthanides are determined simultaneously in the same 55 nm spectral window. The analytical figures of merit for all 14 natural lanthanides are reported and compared with nitrous oxide flame atomic emission spectrometry. Tungsten coil atomic emission concentration detection limits are in the range 0.8 (Yb) to 600 (Pr) μg l -1, and are lower than those for the flame in most cases. The absolute detection limits are near or below the ng level: significantly lower than the flame detection limits due to the smaller sample volume required. A three fold improvement in detection limit may be realized by combining the signals for multiple emission lines for a single element. The method is applied to the determination of seven lanthanides in a soil sample acquired from the National Institute of Standards and Technology. After a simple acid extraction, the measured values agree with the reported values with 95 % confidence in all cases but one, Yb. Finally, a conditioning program for new tungsten coils enhances reproducibility and maximizes the emission signal. 60

75 INTRODUCTION Conventional atomic spectrometry using flames, furnaces, or plasmas may be considered a mature technique, and most recent publications report routine developments rather than novel instrumental arrangements. 1 One exception to this generalization is the continued interest in metal speciation, and the search for a better understanding of the role of these species on biological organisms. Accordingly, in situ, low sample volume methods are of considerable interest. 2 Similarly, field methods continue to be developed. 3 Flames, furnaces, and plasmas have limited portability due to their gas, power, and/or cooling requirements. Traditional atomic spectrometers also have considerable bulk. Laser-induced breakdown plasmas have met with some success in field applications, especially for the direct analysis of solids. 4 In addition, open tungsten coil atomizers have been employed in portable battery-powered atomic absorption spectrometers. 5, 6 Atomic absorption devices, however, usually require a separate light source for each element to be determined, so portable applications are limited. More recently, Rust et al. presented a new method with high potential for field applications. 7, 8 Tungsten coil atomic emission spectrometry (WCAES) employs a simple, inexpensive tungsten filament as both atomizer and excitation source for trace metal determinations. The coil is extracted from a massproduced, commercially available 150 W, 15 V light bulb. Power is supplied by a small, solid-state constant current source, and atomic emission lines are 61

76 detected with a high resolution Czerny-Turner monochromator with a charge coupled device (CCD) detector. Since the method is based on atomic emission rather than absorption, simultaneous multi-element determination of eleven metals at the μg L -1 level is reported. Clearly, a portable WCAES spectrometer could be devised by replacing the high resolution detection system with a much smaller one. This work represents the first steps towards a portable WCAES device. A small, 156 mm focal length monochromator replaces the 1.33 m device reported previously. 7,8 While the new device allows for the collection of more of the atomic emission signal, it also passes a higher proportion of the blackbody radiation emanating from the high temperature filament during the atomization step. This interference is reduced by employing a light blocking aperture. The smaller system also provides a broader wavelength window, so more metals may be determined simultaneously. In addition to the change in detection system, improvements in the atomizer are described. A conditioning program for new coils improves the filament-to-filament reproducibility of the atomizer. In addition, a new atomization heating program is employed. The program is composed of sequentially decreasing current steps whose lengths are determined by monitoring the voltage across the coil as the sample dries. Improved reproducibility and increased tungsten coil lifetimes are observed. The system is evaluated for the determination of the fourteen Lanthanide elements (La, Ce, Pr, Nd, Sm, Eu, Gd, Tb, Dy, Ho, Er, Tm Yb, and Lu) in aqueous solution, and seven of the metals are determined in a commercially 62

77 available soil sample following simple acid extraction. The Lanthanides are often present in fertilizers, 9 or used as feeding markers in nutritional and agronomic 10, 11 studies. These elements have also been used as fingerprint markers for cement and concrete, 12 and in geological, 13 or paleontology studies. 14 Clearly, a field instrument could find applications in these areas. While some of the Lanthanides have been determined by tungsten coil atomic absorption 15, 16 spectrometry (WCAAS), the more portable WCAES approach should be easier to deploy. 63

78 EXPERIMENTAL Instrumental A schematic diagram of the WCAES system is presented in Figure 1. The atomizer is the tungsten coil filament produced for a 150 W, 15 V commercially available slide projector light bulb (Osram Xenophot HXL, Pullach, Germany). The fused silica bulb envelope was removed leaving the filament and bulb base intact. A microscope (Model Z45L, Leica Inc., Buffalo, NY, USA) was used to observe the surfaces of both new and used tungsten coils. The bulb base was mounted in a standard ceramic two-pronged power socket. The bulbs are mass produced to strict optical specifications, so the positioning of the filament from bulb to bulb was highly reproducible. Power was provided by a programmable, constant-current, solid state DC power source (BatMod, Vicor, Andover, MA, USA). The filament was housed inside a glass atomization cell with fused silica windows, and a 1.1 l min -1 purge gas composed of 10 % H 2 /Ar served to prevent coil oxidation and to cool the atomizer. The atomization cell and tungsten coil atomizer have been described in detail elsewhere. 17, 18 The atomic emission arising from the atomization cloud at high temperature was collected with a 25 mm diameter 75 mm focal length fused silica lens. The lens formed a 1:1 image of the cloud on a light blocking baffle containing an adjustable aperture set to 2 mm (Figure 1 inset). The aperture was positioned to block the blackbody radiation emitted from the coil surface, with the image of the coil approximately 1 64

79 mm away from the opening. The radiation passing through the aperture was collected with a second identical fused silica lens, and imaged (1:1) onto the entrance slit of a crossed Czerny-Turner monochromator (MonSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). The monochromator was equipped with a 2400 gr/mm grating (52 x 40 mm), resulting in a reciprocal linear dispersion of 2 nm/mm at 400 nm. Based on a slit width approximately the size of a single detector pixel (20 μm), the theoretical spectral bandpass of the system was 0.04 nm. In practice, the observed atomic emission lines had full width at half maximum (FWHM) of approximately 2 detector pixels, so the practical system resolution was 0.08 nm (see the Yb peak at nm in Figure 2). The monochromator had a 156 mm focal length and the collection optics were f/3.8. The detector was a thermoelectrically-cooled charge coupled device (CCD, Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA). The CCD detector consisted of a two-dimensional array of 1340 x 100 pixels. Each pixel was 20 x 20 μm in size, so the image area on the CCD camera was 26.8 x 2 mm. The system provided a spectral window of approximately 55 nm, depending upon the selection of the central wavelength. A 25 μm fixed entrance slit width was employed throughout. The CCD detector had user selectable integration times as low as 1 ms. For the WCAES system, the greatest signal to noise ratio (S/N) for the Lanthanides was observed with an integration time of 500 ms. Ten successive spectra therefore covered the entire 5 s atomization time. All elements reached maximum emission levels during the third spectrum, and all signals returned to 65

80 zero by the seventh spectrum. Therefore, the summation of six successive spectra resulted in a total integration time of 3 s, and ensured that all emission signals were captured. Reference solutions and sample preparation All reference solutions were prepared from dilution of single element stock solutions (1000 mg l -1, SPEX CertPrep, Metuchen, NJ) with 2% v/v HNO 3. The blank solution (2% v/v HNO 3 ) was prepared from dilution of concentrated (14 mol l -1 ) trace metal grade HNO 3 (Fisher Scientific, Pittsburg, PA, USA) with distilleddeionized water (Milli-Q, Millipore Corp., Bedford, MA, USA). A soil sample from the National Institute of Standards and Technology (Montana soil, SRM # 2711) was used to check the method accuracy. The soil sample was submitted to extraction with concentrated HNO 3 (Fisher). Approximately 1 g of soil was weighed accurately into a plastic extraction container, and a 1 ml aliquot of acid was added. The mixture was left to react for 1 min, then 1 ml of distilled-deionized water was added. The plastic container was placed in an aluminum hot block at 100 ºC and the extraction was carried out for 1 h. To prevent the sample from drying, 3 aliquots (1 ml each) of distilled-deionized water were added to the sample at different intervals during the extraction procedure. 66

81 The extracted sample was allowed to cool for 20 min and submitted to filtration with coarse filter paper (Fisherbrand). The filtrate was diluted to a total of 50 ml with distilled-deionized water. This procedure was carried out in triplicate. Safety considerations Material safety data sheets were consulted before using each chemical reagent. Essential safety precautions were taken in each step of the analysis. Aqueous waste was stored in glass containers prior to disposal. Atomization programs A 25 μl volume was adopted as solution aliquot for all WCAES measurements. A micropipette (Ependorf μl, Brinkmann Instruments Inc., Westbury, NY, USA) was used to place the aliquot directly onto the tungsten coil. A 1.1 l min -1 purge gas flow rate was used to protect the coil during the atomization step. Also acting as the atomizer coolant, the 10% H 2 / Ar purge gas provided a reducing atmosphere for generation of the atomic cloud and protection (from oxidation) of the tungsten coil. The atomization program consisted of seven heating steps (Table 1). The two first steps accounted for gradual solvent volatilization. The coil reached dryness at the end of step 2, as evidenced by an increase in the applied voltage necessary to maintain a constant current during the last few seconds of this stage. The dry coil reached higher 67

82 temperature than the wet coil for a given current. Therefore the ashing stage of the program, steps 3 and 4, employed lower currents. Using progressively lower current prevented the coil from glowing red prior to atomization, and thus reduced the potential to lose analyte. A cooling period (step 5) ensured a reproducible high temperature atomization step, since the beginning temperature was always the same (near room temperature). Finally, the high current (10 A) atomization step generated the atomic cloud and provided energy sufficient to excite atomic emission. The detector was triggered at the beginning of the atomization step (step 6) and six successive spectra were collected automatically within the 55 nm spectral window. The final cooling step readied the coil for the deposition of the next sample aliquot. New tungsten coils were subjected to two different heating programs in order to improve the analytical signal. The conditioning program (Table 1) served to break in the new coil, presumably by etching the smooth surface of the tungsten wire. This program was employed outside of the atomization chamber in air, so coil oxidation was promoted. The coil was then placed in the atomization chamber with a reducing atmosphere for a cleaning program (Table 1), with a 1.4 l min -1 purge gas flow rate (10 % H 2 / Ar). This program served to remove the oxidation products adhering to the coil after the air heating. 68

83 RESULTS AND DISCUSSION Background emission signal WCAES measurements are susceptible to interference from the continuum background radiation emanating from the coil surface which is acting as a blackbody emitter. 8 Care must be taken to isolate this bright background signal from the emission arising from the gaseous analyte atoms near the coil surface. To minimize this problem, a radiation-blocking aperture was placed midway between the focusing lenses (Figure 1). After optimizing the coil position, a significant reduction in the background signal permitted the collection of relatively intense analytical signals. The signal to background ratio (S/B) was maximized with a 2 mm aperture and a viewing height 1 mm above the upper surface of the coil. Also, as described before, 8 S/B was higher when the image of the coil was positioned approximately 1 mm to the side of the monochromator entrance slit, so the final viewing position was 1 mm above and 1 mm to the right of the upper surface of the coil. Heating program optimization As described above, the tungsten coil is heated by a constant current power supply. The sample aliquot wets the coil, so during the dry step the resistance is lower than during the ashing and atomization steps. A given drying 69

84 current will begin to evaporate the liquid sample, and as the amount of liquid on the coil is reduced, the resistance rises, producing an increase in the temperature for the resistively heated filament. Even at a relatively low drying current, if the coil is allowed to heat to dryness, the temperature will begin to rise rapidly. Without careful control of the end time for the drying step, the coil may rise in temperature enough to cause significant loss of analyte atoms. Coil temperatures briefly as high as 700 ºC have been observed in this step, resulting in the poor recovery of relatively volatile elements. 19 Furthermore, small changes in analytical sample volume may cause the point of coil dryness to occur at slightly different times from run to run, so a fixed drying time length may result in different ending temperatures and varying degree of sample loss. These sources of error, while potentially small, will clearly affect analytical precision. To counteract these effects, the potential across the coil was monitored during all heating steps. As the coil dried, the resistance rose and higher voltage was necessary to maintain the constant current. Thus, the degree of dryness of the sample was monitored by observing the potential across the coil. The rate of increase in potential with time was low while liquid remained on the coil, and then abruptly increased upon total dryness. This point of abrupt potential change was used to signal the end of the dry step, rather than a fixed time. In a 60 s dry period designed to completely dry the coil in a single step for example, repetitive runs typically reached the potential change in the range 58 ± 2 s. Ending the dry step at the potential change, rather than the fixed time took the guess work out of the equation. 70

85 To further improve precision, the drying program was divided into multiple steps having incrementally smaller constant currents (Table 1). This insured that the completely dry coil state was reached gradually, reducing the risk of sample loss and splattering. In addition, the probability of beginning the ashing or atomization stages with a wet coil were greatly diminished. Sending the coil to high current, even for an instant, while liquid water was still present resulted in severe oxidation and coil damage. Previously published works typically employed multi-step programs with gradually increasing currents, 7,8,18, 20 but the optimized program presented here produced higher analytical signals, longer coil lifetimes (~ 200 runs), and lower background signals when compared to a conventional program. Coil surface and sensitivity Using a new, untreated coil, the atomic emission signal observed for a test solution increased slowly during the first few dozen atomization programs before leveling off at a maximum stable level after approximately 100 firings. The analytical signal using a fresh coil was typically 50 % lower, on average, than the signal observed for a seasoned coil. Microscopic comparisons of new and used coils in the past have indicated that the surface of a new filament was smooth when compared to a seasoned one that was pitted. 21 The tungsten coils in the current work were also examined with a microscope. As expected, new coils were clean and smooth, while seasoned ones were rougher on the surface, and 71

86 also contained tiny crystals which resembled tungstate (though no further analysis was employed to verify the composition). One possible explanation for the crystalline material may be found in the literature. 22 Non-stoichiometric compounds with the general formula M x WO 3 were prepared at high temperature under a H 2 atmosphere during the production of tungsten bronzes. These crystals were characterized as metal-like and yellow in color with high conductivity. While detailed studies on the effects of chemical modifiers have been reported for tungsten coil atomic absorption spectrometry, 23, 24 a reasonable first approximation may be that the increased surface area resulting from these crystals may allow for better adhesion of analyte atoms during the heating program. While the investigation of the exact role of the surface effects on emission signal is beyond the scope of this manuscript, a conditioning program for new coils was developed. Initially a current was applied to a new dry coil under exposure to air. During this brief program (Table 1) the outer surface of the coil was oxidized. Then a second program was applied under a 10% H 2 /Ar atmosphere (Table 1) to remove the oxidation products, resulting in a clean, seasoned coil. Reproducible, maximized signals were then obtained throughout the lifetime of the coil. Multi-element WCAES spectra The 55 nm spectral window provided by the detection system facilitated the simultaneous determination of up to 7 lanthanide elements by WCAES. 72

87 Figures 2 and 3 present portions of two different spectral windows. Within the 30 nm region centered around 410 nm (Fig. 2) sharp atomic emission lines for five different lanthanide elements were observed: Tm (2 mg l -1 ), Yb (1 mg l -1 ), Er (5 mg l -1 ), Dy (2 mg l -1 ), and Ho (5 mg l -1 ). In addition, the Ca line at nm was visible. Ca was always present as a low level contaminant in the blank (~1 μg l -1 ). Even though the WCAES spectra were rich, little overlap occurred. An example of the resolving power of the detection system is demonstrated by the near baseline resolution of the Ho and Tm peaks at and nm respectively. The full width at half maximum for the Yb peak at nm is 0.08 nm. Since the resolution is limited by the spectrometer, all of the peaks in Figure 2 have nearly the same 0.08 nm FWHM (2 pixel widths). Figure 3 shows a slightly more complicated region of the spectrum. Signals due to six lanthanides are present: Eu (1 mg l -1 ), Lu (5 mg l -1 ), Ce (10 mg l -1 ), Sm (1 mg l -1 ), Nd (20 mg l -1 ), and Gd (20 mg l -1 ). The spectrum also seems to contain an underlying molecular band in the region 465 to 470 nm. Analytical figures of merit Analytical figures of merit were determined with aqueous solutions of the lanthanides (Table 2). Limits of detection (LOD) were calculated as 3 times the standard deviation in the blank signal (n=20), divided by the slope of the calibration curve. The observed LODs range from 0.8 μg l -1 (Yb) to 600 μg l -1 (Nd and Pr). These values compare quite favorably to traditional flame 73

88 emission detection limits using N 2 O as the oxidant. 25 In fact, these data suggest that nitrous oxide flame emission detection limits could be used to predict WCAES LODs to within approximately an order of magnitude. The WCAES LOD is lower than the flame LOD in all but 3 cases: Er (LODs are equal), Eu (flame LOD lower by a factor of 2), and Pr (flame LOD lower by a factor of 9). In addition, the WCAES system simultaneously monitors a 55 nm spectral window, so multiple lines for a single element falling in the window may be used for calibration purposes. 8 Summing the signals at multiple wavelengths lowers the detection limit in most cases, by roughly a factor of 3. Finally, the sample volume for WCAES is quite small (25 μl) compared to the amounts required for traditional flame analysis (approximately 500 μl), so the absolute detection limits ( ng) compare even more favorably. Reference 25 reports flame LODs for 68 elements, suggesting that WCAES might find wide applications outside of the Lanthanides. Other, more recent publications report similar detection limits in the nitrous oxide acetylene flame. 26 The precision of the method was calculated as the relative standard deviation (RSD, n = 10) for each element (Table 2). Aqueous solutions providing WCAES emission signals in the mid-range of the calibration curve were used for these measurements: 25 μg l -1 (Eu, Yb), 200 μg l -1 (Tm), 500 μg l -1 (Sm, Ho, Lu), 1 mgl -1 (Dy, Er), 5 mg l -1 (Ce, Pr, Gd, Tb), and 10 mg l -1 (La, Nd). The RSD was always below 10% and often below 5 %. The performance of the WCAES method was evaluated using a moderately complex sample. Seven lanthanides were determined in a 74

89 commercially available soil sample (Montana soil, NIST SRM #2711). The sample was prepared with a simple nitric acid extraction technique that could be employed in the field. Table 2 presents the recovery values observed for each of the elements present in the sample above the WCAES detection limit. The Lanthanides were reported by NIST to be present in the solid sample at concentrations between 1 and 40 μg g -1, however NIST does not certify these reported values. The Lanthanides were determined to have concentrations in the range 0.9 to 55 μg g -1 in the sample. While more rigorous sample preparation techniques like microwave digestion might improve some of the individual recovery values, clearly all of these elements could be determined at concentrations within a few μg g -1 of their actual values using a field-friendly sample preparation method. The statistical t-test showed that the found value agreed with the NIST reported value with 95 % confidence in all cases except for Yb. The error in the Yb determination could be due to incomplete extraction, or to inaccuracy due to the relatively low amount of the metal present in the sample. Another potential source of inaccuracy in the determination of lanthanides by WCAES may arise from interference effects caused by concomitant elements in the sample matrix. Indeed, interference effects have been documented for atomic absorption spectrometry using metal atomizers such as the wire loop, 27 and the tungsten coil. 28, 29 To date, no interference studies have been published for WCAES methods, and such a study is beyond the scope of this manuscript. 75

90 CONCLUSIONS WCAES is a potentially portable technique for the simultaneous determination of the Lanthanides. Concentration detection limits are similar to those reported for nitrous oxide flame atomic emission spectrometry (in the μg l -1 range), and WCAES LODs may be predicted using previously reported flame emission data. Absolute detection limits are at the low ng level and below, due to the low volume requirement of the WCAES technique (25 μl). Detection limits may be further improved (by approximately a factor of 3) if multiple emission lines for a single element are monitored simultaneously. The WCAES instrument is very simple, requiring only a 150 W constant current power supply. This power could be provided by an automobile battery, rendering the technique portable. In addition, a simple extraction technique for the analysis of a soil sample provides good results for seven lanthanides present at the μg g -1 level in the solid. Finally, the multi-wavelength capability of the system provides for the determination of single elements at more than one emission line. This characteristic improves the confidence level of the analysis and potentially increases the linear dynamic range of the measurement. All 14 lanthanides may be determined by WCAES. 76

91 ACKNOWLEDGEMENTS This material is based upon work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET

92 TABLE CAPTIONS Table 1. Tungsten Coil Heating Programs. Table 2. WCAES Analytical Figures of Merit. FIGURE CAPTIONS Figure 1. Schematic diagram of WCAES instrument. Inserts show alternate views of the atomization cell and the coil image on the adjustable aperture. Figure 2. Simultaneous multi-element WCAES determination of Yb (1 mg l -1 ), Dy and Tm (2 mg l -1 ), Ho and Er (5 mg l -1 ). Figure 3. Simultaneous multi-element WCAES determination of Sm and Eu (1 mg l -1 ), Lu (5 mg l -1 ), Ce (10 mg l -1 ), Nd and Gd (20 mg l -1 ). 78

93 Table 1. Tungsten coil heating programs. Sample Atomization a New Coil Conditioning b New Coil Cleaning c Step Current d (A) time (s) Current (A) time (s) Current (A) time (s) e a. Atomization program for a 25 μl aliquot of an aqueous solution under a 1.1 l min -1 flow rate of 10 % H 2 in Ar. b. Conditioning program for a new dry tungsten coil in open air. c. Cleaning program for a newly conditioned coil under a flow rate of 1.4 l min -1 of 10 % H 2 in Ar (for oxide removal). d. The approximate temperature (T) in Kelvin of the dry coil surface may be estimated from the current (I) in amps: T 309 I While the current associated with each step is fixed, the actual temperature will depend upon the age of the coil, and the presence of liquid or salt on the coil surface. e. Tungsten coil atomic emission was collected during this step. 79

94 Table 2. WCAES Analytical Figures of Merit. Limit of Detection (μg l -1 ) Montana Soil Content (μg g -1 ) f Emission Wavelengths (nm) WCAES a WCAES b Flame c LDR d Precision e Reported Found Ce g h , Dy h g ± 0.7 Er g Eu g ± 0.12 Gd Ho La h g g h g ± ± 15 Lu g Nd g ± 17 80

95 Pr h g 600 NA i Sm Tb Tm h g h g h g ± Yb g 0.8 NA i ± 0.1 a. Detection limit determined using the signal at the primary wavelength only. b. Detection limit determined using the combined signals at all listed wavelengths within the 55 nm window. c. Flame atomic emission detection limits reported for a nitrous oxide acetylene flame. 25 d. Linear dynamic range in decades, beginning at the detection limit. e. Precision (repeatability) reported as % relative standard deviation for an aqueous solution in the midrange of the calibration curve (n = 10). f. NIST Standard Reference Material NIST does not certify the reported Lanthanide values in SRM Found values are the mean ± 1 standard deviation (n = 3). g. Primary emission wavelength for WCAES. h. Primary emission wavelength for Flame emission (if different from WCAES). i. Not Available. Only one emission line in available spectral window. 81

96 Figure 1 Lens Lens Adjustable Aperture C C D 10 % H 2 / Ar Constant current power supply D/A Converter Sample introduction Atomic cloud Coil image Adjustable Aperture W coil Silica window 10 % H 2 / Ar Power supply 82

97 Figure Yb 80 Relative Emission Signal Tm Er Dy Ho Ho Tm Tm Ho Er Ca Tm Tm Dy Dy Wavelength (nm) 83

98 Figure Eu 80 Relative Emission Signal Lu Ce Lu Sm Sm Sm Sm Sm Nd Gd Ce Wavelength (nm) 84

99 REFERENCES 1. E. H. Evans, J. A. Day, W. J. Price, C. M. M. Smith, K. Sutton and J. F. Tyson, J. Anal. At. Spectrom., 2003, 18, J. Szpunar and R. Lobinski, Hyphenated Techniques in Speciation Analysis, Royal Society of Chemistry, Gateshead, 1st ed., X. Hou and B. T. Jones, Microchem. J., 2000, 66, K. Song, Y. I. Lee, and J. Sneddon, Appl. Spectrosc. Rev., 2002, 37, C. L. Sanford, S. E. Thomas and B. T. Jones, Appl. Spectrosc., 1996, 50, J. D. Batchelor, S. E. Thomas and B. T. Jones, Appl. Spectrosc., 1998, 52, J. A. Rust, J. A. Nóbrega, C. P. Calloway and B. T. Jones, Spectrochim. Acta, Part B, 2005, 60, J. A. Rust, J. A. Nóbrega, C. P. Calloway Jr. and B.T. Jones, Spectrochim. Acta, Part B, 2006, 61, S. Miaokang and S. Yinyu, J. AOAC Int., 1992, 75, D. K. Combs and L. D. Satter, J. Dairy Sci., 1992, 75, Z. Wang, D. Li, P. Lu and C. Wang, J. Environ. Qual., 2001, 30, R. L. Goguel and D. A. St. John, Cem. Concr. Res., 1993, 23, F. Worrall and D. G. Pearson, Appl. Geochem., 2001, 16,

100 14. B. J. MacFaddeen, J. Labs-Hochstein, R. C. Hulbert Jr. and J. A. Baskin, Geology, 2007, 35, E. C. Lima, F. J. Krug, J. A. Nóbrega and A. R. A. Nogueira, Talanta, 1998, 47, J. C. J. Silva, E. D. Garcia, A. R. A. Nogueira and J. A. Nóbrega, Talanta, 2001, 55, A. Salido and B. T. Jones, Talanta, 1999, 50, J. A. Rust, J. A. Nóbrega, C. P. Calloway and B. T. Jones, Anal. Chem., 2005, 77, P. V. Oliveira, M. Castanho, J. A. Nóbrega and P. O. Luccas, Quim. Nova, 2000, 23, J. A. Rust, J. A. Nóbrega, C. P. Calloway and B. T. Jones, Anal. Sci., 2005, 21, X. Hou, Z. Yang and B. T. Jones, Spectrochim. Acta Part B,2001, 56, P. G. Dickens and M. S. Whittingham, Q. Rev. Chem. Soc., 1968, 22, D. M. Santos, P. O. Luccas, J. A. Nóbrega and E. T. G. Cavalheiro, Thermochim. Acta, 2000, 362, J. A. Nóbrega, J. Rust, C. P. Calloway and B. T. Jones, Spectrochim. Acta, Part B, 2004, 59, G. D. Christian and F. J. Feldman, Appl. Spectrosc., 1971, 25,

101 26. J. W. Robinson, E. M. Skelly Frame and G. M. Frame III, Undergraduate Instrumental Analysis, Marcel Dekker Inc., New York, 6th ed., 2005, pp J. V. Chauvin, D. G. Davis and L. G. H. Aargis, Anal. Lett., 1992, 25, Z. F. Queiroz, F. J. Krug, P. V. Oliveira, M. M. Silva and J. A. Nóbrega, Spectrochim. Acta, Part B, 2002, 57, P. V. Oliveira, F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz and F. R. P. Rocha, J. Braz. Chem. Soc., 2000, 11,

102 CHAPTER V MULTI-WAVELENGTH DETERMINATION OF COBALT BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY George L. Donati, Mário H. Gonzales, Joaquim A. Nóbrega and Bradley T. Jones The following manuscript was accepted for publication in Analytical Letters and its current status is in press. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati, Mário H. Gonzales and Joaquim A. Nóbrega. The manuscript was prepared by George L. Donati and Joaquim A. Nóbrega, and edited by Bradley T. Jones. 88

103 ABSTRACT A simple tungsten coil atomic emission device is used for the determination of Co in a polluted water sample. Multiple cobalt emission lines in the to nm range are detected simultaneously. Limits of detection at five different wavelengths range between 20 and 70 ng Co in a 25 µl sample volume. Summing the signals for the five different Co emission lines improves the detection limit to 10 ng. The combination of the five lines also results in improved precision (2.2 % relative standard deviation for a 4 mg/l Co solution) and accuracy (102 % for a polluted water reference material). Keywords: Tungsten coil atomizer; Atomic Emission; Cobalt; Multiple lines 89

104 INTRODUCTION Tungsten coil atomic emission spectrometry (WCAES) has recently been reported for a variety of elements (Rust, 2005; Rust, 2006; Donati, 2008). One of the primary advantages of the technique is its simplicity. The atomizer is composed of a low cost tungsten filament extracted from a mass-produced, commercially available light bulb. The specific heat of tungsten and the low mass of the coil allows for a fast heating rate using a small solid-state power source. The device also needs no cooling system to bring the atomizer back to room temperature after the atomization step. Therefore the system is ideally suited for field applications due to its inherent portability.the portability advantage of the tungsten coil atomizer has been demonstrated previously for atomic absorption spectrometry (Nóbrega, 1995; Sanford, 1996; Hou, 2001; Ribeiro, 2002). A WCAES system, however, is even easier to transport into the field, since no external radiation source is required. Two potential limitations must be overcome for the successful implementation of WCAES. First of all, the gas phase temperature in the vicinity of the coil is relatively low, even during the high temperature atomization step. At the currents typically applied to the filament during the atomization step, the maximum temperature of the coil surface approaches the melting temperature for W, 3300 K (Salido, 1999). The gas phase temperature, however, drops off rapidly with distance from the surface and may be as much as 1000 K lower in the observation zone (Queiroz, 2002). Lower gas phase temperatures result in 90

105 smaller excited state populations, and hence smaller emission signals. Secondly, the blackbody emission arising from the tungsten surface is very intense during the atomization step, since the filament, after all, is intended for use as a light bulb. This blackbody emission decreases with distance from the coil surface. Fortunately, the WCAES device employs a solid-state detector that collects timeresolved spectra, and the emission arising from the excited atoms often occurs prior to the time of maximum blackbody emission levels. Nevertheless, in the end a compromise is necessary: the viewing zone must be close enough to the coil surface to maintain an appreciable population in the excited state, but far enough away to minimize the collection of background radiation. Once this optimal position is identified, the technique is applicable to a wide range of elements (Rust, 2005; Rust, 2006), and sensitivities may be better than those reported for other atomic emission techniques (Donati, 2008). The determination of Co by WCAES is a challenge due to the combination of its low volatility and its relatively weak, though plentiful, atomic emission lines (Rust, 2006). Cobalt has been determined in a complex sample matrix (animal feces) with some success by tungsten coil atomic absorption spectrometry (Lopes, 1999). In the current work, careful optimization of instrumental parameters allows the sensitive detection of Co by WCAES. Both the accuracy and precision are improved by the simultaneous monitoring of multiple Co emission lines. 91

106 EXPERIMENTAL Instrumental A photograph of the WCAE atomization cell is presented in Figure 1. The atomizer is the tungsten coil filament produced for a 150 W, 15 V commercially available slide projector light bulb (Osram Xenophot HXL, Pullach, Germany). The fused silica bulb envelope is removed leaving the filament and the bulb base intact (Fig. 1B). The bulb base is mounted in a standard ceramic two-pronged power socket that is counter-sunk into a 25 mm diameter aluminum rod. The rod is approximately 75 mm in length, and it is mounted vertically using a standard ¼-20 screw attached to an optical rod and holder. The bulbs are mass produced to strict optical specifications, so the positioning of the filament from bulb to bulb is highly reproducible. Power is provided through a two-wire feedthrough in the bottom of the aluminum rod by a programmable, constantcurrent, solid state DC power source (BatMod, Vicor, Andover, MA, USA). The filament is housed inside a glass atomization cell with fused silica windows (Ace Glass Part No , Vineland, NJ, USA). The windows in Figure 1 have been removed for clarity. The cell is fastened to the aluminum rod using a compression o-ring fitting. A ground glass sample inlet port is positioned at a 120 degree angle to the bulb mount. A purge gas (10 % H 2 in Ar) enters the cell through the bulb base. The purge gas flows at approximately 1 L/min, and this prevents coil oxidation and promotes the reduction of oxides in the atomization 92

107 cloud. The purge gas also serves to quickly and efficiently cool the coil at the end of the atomization cycle. The atomization cell and tungsten coil atomizer have been described in detail elsewhere (Salido, 1999; Rust, 2005). The optimal viewing position is represented by the black dot in Figure 1B. The position is 1 mm to the right of the side surface of the coil and also 1 mm from the top surface of the coil. Since the purge gas enters the cell from the coil base and exits through the sample introduction port, the atom concentration is slightly higher above the coil. The radiation emanation from this point is collected with a fused silica lens and imaged (1:1) onto the entrance slit of a crossed Czerny-Turner spectrograph (MonSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). The spectrograph has a 156 mm focal length and is equipped with a 2400 gr/mm grating (52 x 40 mm), resulting in a reciprocal linear dispersion of 2 nm/mm at 400 nm. The detector is a thermoelectrically-cooled charge coupled device, CCD (Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA). The CCD detector consists of a two-dimensional array of 1340 x 100 pixels. Each pixel is 20 x 20 μm in size, so the image area on the CCD camera is 26.8 x 2 mm. The system provides a spectral window of approximately 55 nm, depending upon the selection of the central wavelength. A 25 μm fixed entrance slit width is employed. Thus, the typical resolution is approximately 2-3 pixel widths (~0.1 nm spectral bandpass). The CCD detector has user selectable integration times as low as 1 ms. For the WCAES system, the greatest signal to noise ratio (S/N) for Co is observed with an integration time of 500 ms. Six successive spectra covering a 3 93

108 s integration time are collected at the beginning of the atomization step to ensure that all emission signals are captured. Cobalt reaches maximum emission levels during the 2nd spectrum, returning to zero by the beginning of the 3rd spectrum. The 5 s atomization step (Table 1) provides an additional 2 s at high temperature. This serves to clean the coil and eliminate memory effects from sample to sample. Reference solutions and samples Cobalt, aluminum, iron, manganese, nickel, and vanadium reference solutions were prepared from dilution of single element stock solutions (1000 mg/l, SPEX CertiPrep, Metuchen, NJ, USA) with 2 % v/v HNO 3. The blank solution (2 % v/v HNO 3 ) was prepared from dilution of concentrated (14 M) trace metal grade HNO 3 (Fisher Scientific, Pittsburg, PA, USA) with distilled-deionized water (Milli-Q, Millipore Corp., Bedford, MA, USA). A water standard reference material from VHG Labs (Water Pollution Standard 1, Manchester, NH, USA) was used to evaluate the accuracy of the method. The water reference solution was diluted with 2 % v/v HNO 3 prior to Co determination. Method Optimization A micropipette was used to place a 25 µl sample aliquot directly onto the tungsten coil through the ground glass sample introduction port. The atomization 94

109 program (Table 1) consisted of seven heating steps. The first 4 steps effectively dried the sample while minimizing the potential for sample loss. This was accomplished by using successively lower currents. As the sample dried, the coil resistance increased resulting in slightly higher temperatures for a given constant current (Donati, 2008). Upon completion of the drying stage, the coil was returned to room temperature for 10 s (step 5). The detector was triggered at the beginning of the atomization step (step 6) and six successive spectra were collected covering a 55 nm spectral window. The final cooling step readied the coil for the deposition of the next sample aliquot. New tungsten coils were subjected to two surface conditioning heating cycles as described previously (Donati, 2008). The purge gas flow rate and atomization current were optimized simultaneously. The current was varied across a range of 7.2 to 12.8 A, and the flow rate was varied from 0.5 to 1.3 L/min. Eleven combinations of these two variables were evaluated in duplicate using a 4 mg/l Co solution. Potential matrix interference effects were investigated as well. Ten single element interference solutions were prepared each containing 4 mg/l Co and one other element: either 4 mg/l Fe, 4 mg/l Mn, 4 mg/l Ni, 10 mg/l V, 20 mg/l Al, 40 mg/l Fe, 40 mg/l Mn, 40 mg/l Ni, 100 mg/l V, or 200 mg/l Al. In addition, two multielement solutions were prepared, each containing 4 mg/l Co. One of these contained the 5 potential interferents at their lower level, and the other at their higher level. Cobalt emission signals for these solutions were compared with those for the interferent-free 4 mg/l Co solution. 95

110 RESULTS AND DISCUSSION WCAES signals As mentioned above, in the optimal configuration the image of the tungsten coil was positioned slightly off axis relative to the entrance slit of the spectrograph (Figure 1). With this configuration, the Co emission signals appeared and then returned to the baseline prior to background emission signal reaching its maximum level. Cobalt signals emerged at 0.7 s after the onset of the atomization step and were exhausted by1.2 s. A detector integration time of 0.5 s was employed to maximize this signal. All Co emission lines falling within the 55 nm spectral window were collected simultaneously. Blank spectra, collected during the atomization of 2 % v/v HNO 3, were subtracted from all sample spectra. A similar background correction strategy has been reported for the determination of Al and V in oil samples using continuum source WCAAS (Rust 2005), and for continuum source graphite furnace AAS applications (Welz, 2003). Figure 2 shows the background-corrected WCAES spectrum collected for a 4 mg/l Co solution. Notice that at least 12 Co emission lines are present in a relatively narrow 14 nm spectral window. The wavelengths were assigned using the NIST Spectral Database (Sansonetti, 2003). A comparison of the relative intensities of these lines was also performed. The intensity for the most intense Co line (345.4 nm) was divided by the intensity of the other 11 lines, and the result was compared with NIST values (Table 2). The general disagreement in 96

111 the relative intensities indicates a significant difference in temperature between the NIST atomization source, and the cooler W-coil atomization cloud. The multiple emission lines observed by WCAES allow simultaneous measurements with different sensitivities. Each line could be considered separately for quantitative measurements, or the signals may be combined. Multiple lines allow accuracy assessment as previously proposed for chloride determination using flow analysis spectrophotometry (Oliveira, 1997) and ICP OES measurements for geological samples (Cosnier, 2007). Improvements in the LOD have been reported for the determination of Mn using multiple wavelengths by continuum source atomic absorption spectrometry (Heitmann, 2007). This method is particularly useful for the determination of non-metals and molecular species (Resano, 2009). Atomization cycle optimization The tungsten coil is an open atomizer and its surface temperature varies throughout the heating cycle. For example, during the drying step the applied current resistively heats the coil and slowly removes the solvent. Upon complete removal of the solvent, the resistance of the coil increases, and the same applied current causes the W-coil to heat rapidly. This could cause premature loss of analyte atoms. The optimized heating cycle (Table 1) was established by monitoring the potential across the W-coil throughout the cycle. Abrupt potential 97

112 increases at constant applied current indicated that the coil was dry and ready for the next step (Donati, 2008). Purge gas flow rate and atomization current were optimized simultaneously. A 4 mg/l Co standard solution was run with several combinations of these settings, and the emission signal at nm was monitored. Figure 3 shows a maximum signal level at a flow rate of 0.90 L/min and an atomization current of 10.0 A. Notice however that these parameters have a relatively small effect on the emission signal over the range investigated. At low flow rates, the lack of sufficient hydrogen may affect the reducing atmosphere around the coil, resulting in lower signals. At high flow rates, the atomization cloud is diluted more rapidly, also resulting in lower emission signals. Similarly, lower atomization currents result in lower temperatures, and thus lower emission signals. Higher atomization currents result in rapid expansion, and lower signals as well. The result is a relatively broad plateau where flow rate and atomization current are near the optimum values. Matrix interference Ten single-element interference solutions and two five-element interference solutions containing 4 mg/l Co were analyzed as described above. Of these, only 3 produced a Co emission signal with more than 10 % error compared to the interference-free Co signal at nm. Those three contained 4 mg/l Ni, 10 mg/l V, and 40 mg/l Fe. In no case was the Co signal affected by 98

113 more than 30 % due to a single element. When all 5 interfering elements were present together, the percent error in the Co signal was 35 % at the low interferent concentration and 50 % at the high one. When the signals for five different Co emission wavelengths were combined (Fig 2, A-E), the Co percent error was less than 10 % on 5 occasions, and higher than 30 % only twice: for the 40 mg/l Ni concentration and for the solution containing all 5 interfering elements at the lower concentrations. In no case was the percent error worse than 40 % for the summed Co emission lines. The less-intense Co emission lines, when observed individually, showed more susceptibility to interference effects. Given the relatively low resolution optical system, these interferences are likely a combination of spectral overlap and matrix effects in the solution or gas phases. Analytical figures of merit Reference solutions prepared in 2 % v/v HNO 3 were used to determine the analytical figures of merit for Co at 5 different Co emission lines (Table 3). These lines were least prone to the matrix interferences described above. A combination of the 5 lines less prone to matrix interferences was used to improve sensitivity, accuracy and precision. Limits of detection (LOD) were calculated as 3 times the standard deviation in the blank signal (n = 20) divided by the slope of the calibration curve. Absolute detection limits for Co varied from 20 to 70 ng, depending upon wavelength. A slightly lower LOD (10 ng) was observed with the five summed signals. The high resolution WCAES system reported earlier (Rust, 99

114 2006) provided a much lower LOD for Co (0.7 ng). This was possible since the strongest Co emission line (353.0 nm) was completely resolved from its nearest neighbors. The current system, which is intended for portability, is not capable of isolating the Co emission lines, and in fact, the strongest line at nm provides poor performance due to severe spectral overlap. The 4 mg/l Co solution was used to determine the method precision. The relative standard deviation (RSD, n = 10) was calculated and reported as percent. Values ranged from 3.8 to 12 % with a considerable improvement when using the five combined signals (2.2 %). The accuracy of the WCAES method was evaluated by determining Co in the polluted water standard reference material. This sample contained 100 mg/l Co. It also contained the following concomitants: 500 mg/l Al; 250 mg/l V; 100 mg/l As, Be, Cr, Cu, Fe, Pb, Mn, Ni and Zn; 25 mg/l Cd and Se; and 5 mg/l Hg. The sample was diluted with distilled-deionized water (Millipore) and analyzed by both the method of standard additions and the calibration curve method. For the standard addition method, no statistical difference was found between certified and determined Co values at a 95 % confidence level for any of the 5 wavelengths. The percent recoveries ranged from 92.6 to 106 %. With the traditional calibration curve method, recoveries were poorer overall, ranging from 80.6 to 109 %. However, the 5 combined lines gave a 102 % recovery in this case. Even though the matrix affected the accuracy at several single wavelengths, the use of multiple lines eliminated the need for the standard addition method in this case. In addition, the precision for the polluted water 100

115 sample using the combined lines was 3 % RSD, while that for the individual lines ranged from 6 % to 38 % RSD. The benefits of combining several different emission signals from a single element may be summarized as follows. The emission signals measured at different wavelengths add linearly, while their noises are combined quadratically. Therefore, even when a particular line adds only a small emission signal, the S/N for the sum may be slightly improved. Table 3 demonstrates this phenomenon. Adding the two strongest lines together (C and E) results in a reduction in both LOD and relative standard deviation at 4 mg/l Co. Adding the third, fourth, and fifth lines results is very small improvements in S/N. Summing the two strongest emission lines also improves the accuracy of the calibration curve method. While adding the third, fourth, and fifth lines improves accuracy only slightly, if at all, using measurements at multiple wavelengths improves the confidence of the analyst. A potential interferent at one particular wavelength will have a smaller effect when more emission lines are summed. 101

116 CONCLUSIONS A tungsten coil atomizer in conjunction with a small spectrograph and a solid-state detector provides time-resolved atomic emission measurements. This arrangement allows the simultaneous monitoring of multiple emission lines for Co, resulting in improved analytical figures of merit and decreased matrix interference effects. The performance is relatively insensitive to small variations in purge gas flow rate and atomization current. The instrument has the potential for portability and simultaneous multi-element determinations. ACKNOWLEDGEMENTS This material is based upon work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: NSF CBET and DHS DN-077-ARI J.A.N. is thankful to Fundação de Amparo à Pesquisa do Estado de São Paulo for traveling funds (Process 07/ ). 102

117 REFERENCES Cosnier, A., J. M. Mermet, S. Vélasquez and S. Lebouil Application of unique software tools dedicated to multiline analysis by CCD-based ICP-AES for geological samples. Spectroscopy, Special Supplement, October: Donati, G. L., J. Gu, J. A. Nóbrega, C. P. Calloway Jr. and B. T. Jones Simultaneous determination of the Lanthanides by tungsten coil atomic emission spectrometry. J. Anal. At. Spectrom. 23: Donati, G. L., B. E. Kron and B. T. Jones Simultaneous determination of Cr, Ga, In and V in soil and water samples by tungsten coil atomic emission spectrometry. Spectrochim. Acta Part B 64: Heitmann, U., B. Welz, D. L. G. Borges and F. G. Lepri Feasibility of peak volume, side pixel and multiple peak registration in high-resolution continuum source atomic absorption spectrometry. Spectrochim. Acta Part B 62: Hou, X., K. E. Levine, A. Salido, B. T. Jones, M. Ezer, S. Elwood and J. B. Simeonsson Tungsten coil devices in atomic spectrometry: absorption, fluorescence, and emission. Anal. Sci. 17: Jackson, K. W Electrothermal atomization for analytical atomic spectrometry. Chichester, UK: J. Wiley and Sons. Lopes, G. S., A. R. A. Nogueira, P. V. Oliveira and J. A. Nóbrega Determination of cobalt in animal feces by tungsten coil atomic absorption spectrometry. Anal. Sci. 15:

118 Nóbrega, J. A., M. M. Silva, P. V. Oliveira, F. J. Krug and N. Baccan Espectrometria atômica com atomização eletrotérmica em superficies metálicas. Quim. Nova 18: Oliveira, C. C., R. P. Sartini, E. A. G. Zagatto and J. L. F. C. Lima Flow analysis with accuracy assessment. Anal. Chim. Acta 350: Oliveira, P. V., F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz and F. R. P. Rocha Influence of Na, K, Ca and Mg on lead atomization by tungsten coil atomic absorption spectrometry. J. Braz. Chem. Soc. 11: Queiroz, Z. F., F. J. Krug, P. V. Oliveira, M. M. Silva and J. A. Nóbrega Electrothermal behavior of sodium, potassium, calcium and magnesium in a tungsten coil atomizer and review of interfering effects. Spectrochim. Acta Part B 57: Queiroz, Z. F., P. V. Oliveira, J. A. Nóbrega, C. S. Silva, I. A. Rufini, S. S. Sousa and F. J. Krug Surface and gas phase temperatures of a tungsten coil atomizer. Spectrochim. Acta Part B 57: Resano, M., J. Briceno and M. A. Brelarra Direct determination of phosphorus in biological samples using a solid sampling-high resolutioncontinuum source electrothermal spectrometer: comparison of atomic and molecular absorption spectrometry. J. Anal. At. Spectrom. 24: Ribeiro, A. S., M. A. Z. Arruda and S. Cadore Espectrometria de absorção atômica com atomização eletrotérmica em filamento de tungstênio: uma revisão crítica. Quim. Nova 25:

119 Rust, J. A., J. A. Nóbrega, C. P. Calloway Jr. and B. T. Jones Fraunhofer effect atomic absorption spectrometry. Anal. Chem. 77: Rust, J. A., J. A. Nóbrega, C. P. Calloway Jr. and B. T. Jones Advances with tungsten coil atomizers: continuum source atomic absorption and emission spectrometry. Spectrochim. Acta Part B 60: Rust, J. A., J. A. Nóbrega, C. P. Calloway Jr. and B. T. Jones Analytical characteristics of a continuum-source tungsten coil atomic absorption spectrometer. Anal. Sci. 21: Rust, J. A., J. A. Nóbrega, C. P. Calloway Jr. and B. T. Jones Tungsten coil atomic emission spectrometry. Spectrochim. Acta Part B 61: Rust J. A., G. L. Donati, M. T. Afonso, J. A. Nóbrega and B. T. Jones An overview of electrothermal excitation sources for atomic emission spectrometry. Spectrochim. Acta Part B 64: Salido, A. and B. T. Jones Simultaneous determination of Cu, Cd and Pb in drinking-water using W-coil AAS. Talanta 50: Sanford, C.L., S. E. Thomas and B. T. Jones A portable, battery-powered, tungsten coil atomic absorption spectrometer for Pb determinations. Appl. Spectrosc. 50: Sansonetti, J. E., W. C. Martin and S. L. Young Handbook of basic atomic spectroscopic data. Gaithersburg, USA: National Institute of Standards and Technology. Accessed in August 22,

120 Welz, B., H. Becker-Ross, S. Florek, U. Heitmann and M. G. R. Vale Highresolution continuum-source atomic absorption spectrometry what can we expect? J. Braz. Chem. Soc. 14:

121 FIGURE CAPTIONS Figure 1. Photographs of WCAES atomization cell (A), and close-up showing the viewing position (black dot) relative to the W-coil (B). Figure 2. Background-corrected emission spectrum for a 4 mg/l Co solution. The five emission lines used for quantification are designated A-E. Figure 3. Contour plot showing the effects of purge gas flow rate and atomization current on the relative Co emission signal at nm. The maximum value (100 %) occurs at 10 A and 0.9 L/min. 107

122 Figure 1. Photographs of WCAES atomization cell (A), and close-up showing the viewing position (black dot) relative to the W-coil (B). A 108

123 109 B

124 Figure 2. Background-corrected emission spectrum for a 4 mg/l Co solution. The five emission lines used for quantification are designated A-E. 110

125 Figure 3. Contour plot showing the effects of purge gas flow rate and atomization current on the relative Co emission signal at nm. The maximum value (100 %) occurs at 10 A and 0.9 L/min. 111

126 TABLE CAPTIONS Table 1. Tungsten coil heating cycle. Table 2. Atomic emission lines for cobalt. Table 3. Analytical figures of merit for cobalt determination by WCAES. 112

127 Table 1. Tungsten coil heating cycle Step Applied Current a Time Read (A) (s) No No No No No 6 b Yes No a. The temperature (T) in Kelvin of the dry coil surface may be estimated from the current (I) in amps: T 309 I (Salido, 1999). b. Atomic emission signals were collected during this step. 113

128 Table 2. Atomic emission lines for cobalt Emission line (nm) a Relative Intensity b NIST Relative Intensity a Experimental data (A) (B) (C) (D) (E) a. The five emission lines used for quantification are designated A-E. b. Relative intensities calculated as I / I λ where nm is the most intense line in the region (Sansonetti 2003). 114

129 Table 3. Analytical figures of merit for cobalt determination by WCAES Emission Limit of Precision b VHG Water Pollution Standard Wavelength detection (mg/l) (nm) (ng) a Certified Standard Addition Calibration Curve c (A) ± (B) ± (C) ± (D) ± (E) ± 7.5 C + E ± 4 A + C + E ± 3 A + B + C + E ± 3.1 A + B + C + D + E ± 3 a. Based on 3 standard deviations for the blank signal and a 25 μl sample volume. b. Precision reported as % relative standard deviation for a 4 mg/l Co solution (n = 10). c. VHG water pollution standard 1, WPS Found values are the mean ± 1 standard deviation (n = 5). 115

130 CHAPTER VI SIMULTANEOUS DETERMINATION OF Cr, Ga, In AND V IN SOIL AND WATER SAMPLES BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY George L. Donati, Benjamin E. Kron and Bradley T. Jones The following manuscript was published in Spectrochimica Acta Part B, volume 64, pages , 2009 (DOI /j.sab ), special issue dedicated to the 10th Rio Symposium on Atomic Spectrometry, held in Salvador, BA, Brazil. It is reprinted here with permission. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati and Benjamin E. Kron. The manuscript was prepared by George L. Donati and edited by Bradley T. Jones. 116

131 ABSTRACT Tungsten coil atomic emission spectrometry is employed for the simultaneous determination of Cr, Ga, In, and V. Both V and In are detected by this technique for the first time. The atomizer is a simple, inexpensive tungsten filament extracted from a mass-produced, commercially-available 150 W, 15 V microscope bulb. A 25 μl sample aliquot is placed directly on the coil and a small constant-current power source is used to carefully dry, ash and atomize the sample. Analytical signals are detected with a Czerny-Turner spectrograph and a charge coupled device detector. Multiple emission lines from all 4 elements are monitored simultaneously in a 54 nm spectral window. Concentration limits of detection are in the μg l -1 range for all elements, and the absolute limits of detection are 0.2, 2, 0.5, and 10 ng for Cr, Ga, In, and V, respectively. Even lower values may be obtained by combining the signals for the multiple emission lines of a single element. The method precision is typically better than 5.0 % relative standard deviation, and sometimes as good as 0.95 % (Ga). Standard reference materials of soil and water are used to check the method accuracy. After a simple acid extraction, the values determined by the method presented no significant difference from the reported values at the 95 % confidence level. Keywords: Atomic emission; Tungsten coil; Portability; Field analysis 117

132 1. INTRODUCTION The technological revolution brought several undeniable benefits to the society, but increasing concerns regarding waste management and environmental issues represent a challenge to the modern world. Contamination of soil, water and air by a wide range of different toxic substances is a growing environmental problem. Low prices and fast technological advances contribute to increase the demand for cell phones, personal computers, digital cameras, automobiles and other goods. Most of these products and their accessories eventually end up in landfills, representing an important source of contamination by metals such as Ni, Cr, Cd, Zn, and Cu [1]. In 2007 an estimate of 500 million computers became obsolete in the United States, and only a small fraction of those will be recycled [2]. Strict new laws have been established to regulate metal concentrations in the environment [3, 4]. Faster, more sensitive analytical methods are required to attend to these regulations. Field methods are of great interest since simultaneous assessment of contamination sources and respective remediation actions may be quickly deployed. Traditional atomic spectrometric techniques present limited portability and their application in the field is compromised by critical aspects such as equipment size, and gas, power and cooling requirements. Although some field methods have been reported [5-8], equipment simplification, size reduction, and sensitivity improvement would make these techniques more viable. A new, potentially portable method was introduced in 118

133 2005: tungsten coil atomic emission spectrometry (WCAES) [9, 10]. With this technique, eleven elements were simultaneously determined at the μg l -1 level using inexpensive tungsten filaments extracted from mass-produced, commercially-available 150 W miniature light bulbs. Power was supplied by a small, solid-state constant current source, and the analytical signals were monitored with a high resolution Czerny-Turner spectrograph and a charged coupled device (CCD) detector. More recently [11], the 1.33 m spectrograph was replaced by one with a 156 mm focal length, and the method was applied to the determination of 14 lanthanides with limits of detection in the sub-μg l -1 range. Improvements related to coil conditioning and heating cycles were presented and seven lanthanides were simultaneously determined in a soil sample following a simple acid extraction. In the present work, a step closer to a portable WCAES device is taken by simplifying the optical arrangement and applying the method to the simultaneous determination of Cr, Ga, In, and V in soil and water samples. These four test elements have increasingly been employed in the fabrication of electronics, batteries, and automobile parts. Thus they have become important sources of environmental contamination due to mining activities and waste residues. Soil and water contamination by these potentially toxic metals requires fast action to prevent disease and minimize ecological damage. A simple, fast, and low-cost WCAES field device could be employed in this arena. 119

134 2. EXPERIMENTAL 2.1. Instrumental Fig. 1 presents the schematic design of the WCAES instrumentation. The tungsten coil filament is used as both atomizer and excitation source for atomic emission. The coil is extracted from a commercially available 150 W, 15 V microscope light bulb (Osram Xenophot HXL, Pullach, Germany) by removing the fused silica envelope. The bulb base remains intact and is mounted in a standard two-pronged ceramic power socket. A glass atomization cell with fused silica windows is used to house the extracted filament (Fig. 1B), and a 10% H 2 -Ar purge gas mixture, flowing at 0.6 l min -1, is used to prevent coil oxidation and to cool the atomizer. A programmable solid state DC power supply (BatMod, Vicor, Andover, MA, USA) provides the constant current to resistively heat the coil during the heating cycles. Both atomization cell and tungsten coil atomizer have been described in detail previously [12, 13]. Radiation emitted by the atom cloud at high temperature is collected with a 25 mm diameter, 75 mm focal length fused silica lens. A 1:1 image of the cloud is focused on the entrance slit of a crossed Czerny-Turner spectrograph (MonoSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). The image of the coil is positioned approximately 1 mm off-axis from the spectrograph s 25 μm entrance slit (Fig. 1C) to minimize the blackbody radiation collected at high temperature. 120

135 The spectrograph has a 156 mm focal length and is equipped with a 2400 grooves per mm grating (52 x 40 mm), resulting in a reciprocal linear dispersion of 2 nm/mm at 400 nm. A thermoelectrically-cooled CCD (Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA) serves as the detector. The CCD consists of a two-dimensional array of 1340 x 100 pixels. Each pixel is 20 x 20 μm in size, so the image area on the CCD camera is 26.8 x 2 mm. Depending upon the wavelength selection, the system provides a spectral window of approximately 54 nm. Selection of integration times as low as 1 ms are possible in the CCD detector software. The best signal-to-noise ratio values for all elements studied are observed with a 500 ms integration time per spectrum. Six successive spectra are collected at the beginning of the atomization step, corresponding to the first 3 s of high current application. Most elements reach maximum emission levels during the third spectrum, and all signals return to zero by the end of the fifth spectrum. Therefore, a 5-s atomization step is long enough to ensure that all emission signals are recorded, leaving an additional 2 s of high coil temperature to remove any remaining sample residue. 121

136 Fig. 1. Schematic design of the WCAES instrument: (A) Block diagram; (B) Alternate view of the atomization cell; (C) Coil image position relative to the spectrograph s slit. A Lens C C D 10 % H 2 / Ar Constant current D/A Converter B Sample introduction Atomic cloud Silica window W coil 10 % H 2 / Ar Power supply 122

137 2.2. Sample preparation All reference solutions were prepared from dilution of single element stock solutions (1000 mg l -1, SPEX CertPrep, Metuchen, NJ, USA) with distilleddeionized water (Milli-Q, Millipore Corp., Bedford, MA, USA). Two reference materials were used to test the accuracy of the method: Montana Soil from the National Institute of Standards and Technology (NIST SRM # 2711, Gaithersburg, MD, USA), and Water Pollution Standard 1 from VHG Labs (WPS1-100, Manchester, NH, USA). Three different tap water samples (one from North Carolina, one from California, and one from the District of Columbia) were collected and spiked with 100 μg l -1 of Cr, 2000 μg l -1 of Ga, 500 μg l -1 of In, and 4000 μg l -1 of V. Soil samples were prepared as follows. In a plastic extraction container, 2.50 g of soil (weighed accurately) is combined with 2.50 ml of concentrated (14 mol l -1 ) nitric acid (Fischer Scientific, Pittsburg, PA, USA). The mixture was left to react for 1 min, and then diluted with 2.50 ml of distilled-deionized water. The container was placed in an aluminum hot block at 100 ºC and the extraction was carried out for 1.5 h. To prevent complete sample drying and consequent analyte loss, two additional 1-mL aliquots of distilled-deionized water were added to the container, one 35 min and the second 70 min after the beginning of the cycle. The mixture was allowed to cool for 40 min and then filtered with coarse paper (Fisherbrand). The filtrate was diluted to a final volume of 25.0 ml with distilleddeionized water. 123

138 For Cr determination, 120 μl of the final solution was further diluted to 10.0 ml with distilled-deionized water, and the solution was analyzed directly. Due to severe matrix interferences with Ga determination, an additional sample digestion step was applied. An aliquot of 20.0 ml of the final solution obtained in the first digestion procedure was transferred to a new container and combined with 1.0 ml of concentrated HNO 3 and 0.50 ml of 30 % v/v H 2 O 2 (Acros, Morris Plains, NJ, USA). The mixture was returned to the hot block and heated for an additional 1 h at 80 ºC. The resulting solution was diluted with distilled-deionized water to a final volume of 25.0 ml, and Ga was directly determined by WCAES. The water standard reference sample was diluted with distilled-deionized water and no additional treatment was required before WCAES determination. Spiked water samples were prepared by diluting single element stock solutions (SPEX) with water collected directly from the tap in different regions of the United States Heating cycle A sample aliquot of 25 μl was placed directly onto the tungsten coil using a micropipette (Eppendorf μl, Brinkman Instruments Inc., Westbury, NY, USA). With the purge gas flowing at 0.6 l min -1, the heating cycle presented in Table 1 was applied. Solvent vaporization was achieved during the two first steps. During these steps, the potential necessary to produce the constant coil current was monitored continuously. This allowed for better control of the drying process 124

139 [11]. An abrupt increase in the applied voltage during the last few seconds of the second step indicated that the coil had reached dryness (causing an increase in resistance). Since a dry coil can reach high temperatures even by applying relatively low currents [14], a heating cycle with gradually lower currents was used to prevent analyte loss [11, 14, 15]. Steps 3 and 4 accounted for the sample pyrolysis, and a cooling step (step 5) ensured reproducibility since the temperature at the beginning of the high temperature atomization step was always the same (near room temperature). The atomization step (step 6), with a constant current of 10 A, provided enough energy to both atomize and excite analyte atoms, generating atomic emission. The detector was trigged at the beginning of this step and six successive spectra were recorded. Finally, another cooling step readied the coil for the next sample aliquot. Two different conditioning programs were applied to new coils to improve sensitivity. Those programs are discussed in detail elsewhere [11]. 125

140 Table 1. Tungsten coil heating cycle Step Applied Current a / A Time / s Read No No No No No Yes No a. The temperature (T) in Kelvin of the dry coil surface may be estimated from de current (I) in amps: T 309 I [14] Gas flow rate and atomization current Both purge gas flow rate and applied current for atomization (step 6) may affect the magnitude of the analytical emission signal for each element studied. The combined effects of these parameters were evaluated with a factorial approach. The purge gas flow rate was optimized in the range of 0.5 to 1.3 l min -1, and the applied current for atomization in the range of 7.2 to 12.8 A. Aqueous solutions of Cr, Ga, and In (5.0 mg l -1 ), and V (20 mg l -1 ) were repetitively analyzed by WCAES, and the emission signal for each element was 126

141 used to determine the optimal experimental conditions. In total, 11 different random pairs (flow rate: atomization current) were tested. The emission signals for the 4 elements varied by no more than a factor of 3 from highest to lowest. The method of the least squares was used to determine the coefficients for each independent variable, and a second degree equation relating those variables was established. The calculated optimal atomization current for Cr and V was 10 A. On the other hand, both Ga and In presented better results when the atomization current applied was at the extremes (7.2 or 12.8 A). The purge gas flow rate had little significant effect on the magnitude of the emission signal, with the exception of Ga, which presented better results with the gas flow either at 0.5 or 1.3 l min -1. Considering the lower sensitivity for V, and seeking a compromise between analytical signal and tungsten coil lifetimes, a purge gas flow rate of 0.6 l min -1 and an applied atomization current of 10 A were adopted for the simultaneous determination of all 4 elements by WCAES. 127

142 3. RESULTS AND DISCUSSION 3.1. Emission spectra The intense blackbody radiation emitted by the tungsten coil during the atomization step can saturate the detector before any analytical signal is collected. To prevent this from happening, the coil image was positioned approximately 1 mm to the side of the spectrograph entrance slit (Fig. 1C). Background correction was performed by subtracting the spectra collected during the atomization of a blank solution from the corresponding spectra collected during the atomization of a sample [16, 17]. Fig. 2 presents background corrected spectra for Cr, Ga, In, and V in the nm region. More than 10 emission lines for V, 3 for Cr, and 2 for each Ga and In were observed. With this arrangement, the maximum emission signals for Ga and In were observed on the second spectrum collected, as shown in Fig. 2A (0.5 to 1.0 s after the onset of the high temperature atomization step). The most intense signals for Cr and V were observed on the third spectrum ( s, Fig. 2B). As expected, emission signals for more refractory metals tend to appear later in the atomization step. With the integration time of 0.5 s, spectral interference caused by concomitants could be corrected to some extent with this crude temporal separation. 128

143 3.2. Analytical figures of merit Analytical figures of merit were determined with aqueous solutions of each metal (Table 2). Limits of detection (LOD) were calculated as 3 times the standard deviation in the blank signal (n = 20), divided by the slope of the calibration curve. Values as low as 7, 90, 20, and 400 μg l -1 for Cr, Ga, In, and V, respectively, were determined. Since a 54 nm spectral window was monitored for all 4 elements, multiple lines for the same element may be summed, reducing the LOD and increasing both precision and accuracy [10, 11]. Adding the intensities for different emission lines clearly increased the slope of the calibration curves, but it also increased the measured blank noise. The limit of detection was expected to improve only if the limiting source of blank noise was shot noise or detector noise (but not in the case of flicker noise). Table 2 demonstrates that using summed signals resulted in limits of detection 1.6, 1.2, and 1.4 times lower than those observed for the most intense single emission line for Cr, Ga, and V, respectively. For In, no significant sensitivity improvement was observed by combining the emission intensities at and nm. In this case, the sensitivity improvement obtained by summing only two emission lines was not enough to counterbalance the noise increase, particularly with the nm line. Considering the small sample volume required for WCAES (25 μl), and summing multiple lines for each element, the best absolute LODs were 0.1, 2, 0.5, and 8 ng for Cr, Ga, In, and V, respectively. 129

144 The precision of the method was calculated for each element and reported as the relative standard deviation (RSD, n = 10). Aqueous solutions in the upper range of the calibration curves were used for this determination: Cr 100 μg l -1, Ga 5000 μg l -1, In 2000 μg l -1, and V 20,000 μg l -1. Most RSD values were below 5.0 %, with the lowest being 0.95 % (Ga). Signal-to-noise ratios (S/N) were calculated as the reciprocal of the relative standard deviation (RSD -1 ) and were in the range. The typical linear dynamic range (LDR) for each element was one to two decades for a single emission line (Table 2). These values were obtained using calibration solutions containing all 4 elements. Regardless of the relative intensity of individual emission lines for a given element, the calibration curves deviated from linearity at roughly the same concentration level, at the upper end of the calibration curve: Cr 100 μg l -1, Ga 5,000 μg l -1, In 2,000 μg l -1, and V 20,000 μg l -1. The LDR may be extended at the lower end by using multiple lines for the same element. The accuracy of the method was evaluated by determining all 4 elements in commercially-available reference materials of soil and water. These results are presented in Table 3. Not all emission lines could be used to determine all 4 elements in both standard reference materials due to interferences caused by concomitants. However, no statistical difference was found between the values determined by WCAES and the values reported for both standards, for all 4 elements. To evaluate the performance of the method in real samples and, since Ga and In are not present in the water standard reference sample, 3 tap water samples were spiked with 100 μg l -1 Cr, 2000 μg l -1 Ga, 500 μg l -1 In, and

145 μg l -1 V and determined by WCAES (Table 3). For Cr and V, recoveries were in the % range, while recoveries varying from 32 to 128 % were observed for Ga and In. The poor recoveries observed for Ga and In in some water samples may be related to non-spectral interferences. In this case, those elements may have formed volatile compounds with some concomitants and were lost during the drying step of the atomization cycle. One hypothesis is the formation of Ga and In chloride molecules, resulting from the high concentrations of this anion in some water samples. The stability of these chlorides has been reported before and they may even be used for indirect determination of chloride by graphite furnace molecular absorption spectrometry [18]. Despite some variation in the recoveries of Ga and In for the spiked samples, it is clear that all these elements could be determined with adequate accuracy and precision using a simple extraction and/ or digestion procedure that could easily be employed in the field. Extensive literature is available on interference effects of concomitants on the analytical signals observed with tungsten coil atomic absorption spectrometry (WCAAS) [19, 20]. While such effects may be similar in WCAES, a more detailed study is required to understand and minimize them. 131

146 Fig. 2. Background corrected spectra for simultaneous multi-element WCAES determination of Cr (1.0 mg l -1 ), Ga (5.0 mg l -1 ), In (1.0 mg l -1 ), and V (20 mg l -1 ): (A) Spectrum collected from s after the beginning of the atomization step; (B) Spectrum collected from s after the beginning of the atomization step (A) Ga V In Relative Intensity Ga In V Ca Cr V 2000 V V Wavelength (nm) (B) V Relative Intensity Cr 8000 V Ca Ga V 3000 In V V Ga In Wavelength (nm) 132

147 Table 2. WCAES analytical figures of merit Element Emission Limit of LDR a Precision b S/N c Wavelength (nm) detection (μg l -1 ) Cr lines summed Ga lines summed In lines summed V lines summed a. Linear dynamic range in decades. 133

148 b. Percent relative standard deviation for an aqueous solution: 100 μg l -1 Cr, 5000 μg l -1 Ga, 2000 μg l -1 In, and μg l -1 V. c. Signal-to-noise ratio for an aqueous solution: 100 μg l -1 Cr, 5000 μg l -1 Ga, 2000 μg l -1 In, and μg l -1 V. 134

149 Table 3. WCAES accuracy for the determination of Cr, Ga, In, and V in soil and water samples Element Emission Montana soil Polluted water Spiked tap water c wavelength (μg g -1 ) a (mg l -1 ) b (% recovery) (nm) Reported Found Reported Found NC CA DC Cr ± ± ± ± ± Ga ± In ± V ± ± a. NIST SRM # Found values are the mean ± 1 standard deviation (n = 3). 135

150 b. VHG SRM # WPS c. NC = North Carolina water; CA = California; DC = District of Columbia. 136

151 4. CONCLUSIONS WCAES provides adequate accuracy and precision to determine Cr, Ga, In, and V in soil and water samples. Using a tungsten filament extracted from a common bulb, and a small power supply, this method represents a simple, economical alternative to more traditional atomic spectrometric methods. No radiation source or cooling system is required. By replacing the power supply with an automotive battery, and the high resolution monochromator with a smaller detector, WCAES can be used in the field following a simple acid extraction for fast, accurate analyses. The sample volume required (25 μl) is another advantage. Absolute limits of detection are at or below the ng level. Sensitivity may be further improved by summing multiple lines for a single element. The possibility of determining several elements simultaneously, at multiple wavelengths, increases the analytical output and creates alternatives to reduce matrix interferences either by temporal separation of the analytical signal or by choosing a near analytical line less prone to interference. Precision, accuracy, and linear dynamic range can be improved by monitoring multiple emission lines for a single element. Finally, the interferences observed in WCAES demand further investigation in order to understand the processes involved and to expand the method applications. 137

152 ACKNOWLEDGEMENTS This material is based upon the work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET

153 REFERENCES [1] C. Xiaoli, T. Shimaoka, C. Xianyan, G. Qiang, Z. Youcai, Characteristics and mobility of heavy metals in an MSW landfill: Implications in risk assessment and reclamation, J. Hazard. Mater. 144 (2007) [2] B. K. Gullett, W. P. Linak, A. Touati, S. J. Wasson, S. Gatica, C. J. King, Characterization of air emissions and residual ash from open burning of electronic wastes during simulated rudimentary recycling operations, J. Mater. Cycles Waste Manag. 9 (2007) [3] U. S. E. P. Agency, National primary drinking water regulations, U. S. Environmental Protection Agency, Washington, D. C., [4] CONAMA, Legislation CONAMA number 357/2005, Ministério do Meio Ambiente, Conselho Nacional do Meio Ambiente, Brasília, Brazil, [5] J. D. Batchelor, S. E. Thomas, B. T. Jones, Determination of cadmium with a portable, battery-powered tungsten coil atomic absorption spectrometer, Appl. Spectrosc. 52 (1998) [6] X. Hou, B. T. Jones, Field instrumentation in atomic spectroscopy, Microchem. J. 66 (2000) [7] C. L. Sanford, S. E. Thomas, B. T. Jones, Portable, battery-powered, tungsten coil atomic absorption spectrometer for lead determinations, Appl. Spectrosc. 50 (1996) [8] K. Song, Y. I. Lee, J. Sneddon, Recent developments in intrumentation for laser induced breakdown spectroscopy, Appl. Spectrosc. Rev. 37 (2002)

154 [9] J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Tungsten coil atomic emission spectrometry, Spectrochim. Acta, Part B 61 (2006) [10] J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Advances with tungsten coil atomizers: continuum source atomic absorption and emission spectrometry, Spectrochim. Acta, Part B 60 (2005) [11] G. L. Donati, J. Gu, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Simultaneous determination of the Lanthanides by tungsten coil atomic emission spectrometry, J. Anal. At. Spectrom. 23 (2008) [12] J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Fraunhofer Effect Atomic Absorption Spectrometry, Anal. Chem. 77 (2005) [13] A. Salido, B. T. Jones, Simultaneous determination of Cu, Cd and Pb in drinking-water using W-Coil AAS, Talanta 50 (1999) [14] P. V. Oliveira, M. Catanho, J. A. Nóbrega, P. O. Luccas, Avaliação de programas de aquecimento para espectrometria de absorção atômica com atomização eletrotérmica em filamento de tungstênio, Quím. Nova 23 (2000) [15] G. L. Donati, K. E. Pharr, C. P. Calloway, Jr., J. A. Nóbrega, B. T. Jones, Determination of Cd in urine by cloud point extraction-tungsten coil atomic absorption spectrometry, Talanta 76 (2008) [16] J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Analytical characteristics of a continuum-source tungsten coil atomic absorption spectrometer, Anal. Sci. 21 (2005)

155 [17] B. Welz, H. Becker-Ross, S. Florek, U. Heitmann, M. G. R. Vale, Highresolution continuum-source atomic absorption spectrometry what can we expect?, J. Braz. Chem. Soc. 14 (2003) [18] L. H. J. Lajunen, Spectrochemical Analysis by Atomic Absorption and Emission, 1st ed., RSC, Cambridge, 1992, pp [19] Z. F. Queiroz, F. J. Krug, P. V. Oliveira, M. M. Silva, J. A. Nóbrega, Electrothermal behavior of sodium, potassium, calcium and magnesium in a tungsten coil atomizer and review of interfering effects, Spectrochim. Acta, Part B 57 (2002) [20] P. V. Oliveira, F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz, F. R. P. Rocha, Influence of Na, K, Ca and Mg on lead atomization by tungsten coil atomic absorption spectrometry, J. Braz. Chem. Soc. 11 (2000)

156 CHAPTER VII DOUBLE TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY: SIGNAL ENHANCEMENT AND A NEW GAS PHASE TEMPERATURE PROBE George L. Donati, Clifton P. Calloway, Jr and Bradley T. Jones The following manuscript was published in the Journal of Analytical Atomic Spectrometry, volume 24, pages , 2009 (DOI /b904636d), and is reprinted with permission. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati. The manuscript was prepared by George L. Donati and edited by Bradley T. Jones. 142

157 ABSTRACT A new double atomizer arrangement for tungsten coil atomic emission spectrometry is described. Two small constant current power supplies, a Czerny- Turner spectrograph, and a CCD detector are used to determine refractory elements in water samples. The analytical figures of merit for Ba, Sr, Ti, and V are reported and compared with the single coil arrangement. The use of two atomizers provides more energy for the atomization and excitation of the analytes and consequently improves sensitivity. Addition of a second coil improves limits of detection for refractory elements by a factor of 40 (V) and 5 (Ti). Using 25 μl sample volumes, the limit of detection for V is 10 µg l -1 at the nm emission line, and 400 µg l -1 for the nm Ti emission line. Vanadium is determined in a polluted water certified reference material, and the results agree with the certified value (250 mg l -1 ) at the 95 % confidence level. For Ba and Sr, which emit strongly with a single coil, the double coil detection limits are improved by less than a factor of two: 0.07 μg l -1 Ba and 0.4 μg l -1 Sr. Emission signals for Ag, Cu and Sn are observed for the fist time with a WCAES system. A new method to determine the gas phase temperature by atomic emission spectrometry is also presented. Emission intensities for Dy (418.7 and nm) and Eu (462.7 and nm) are used to calculate Boltzmann temperatures for both the single and double coil arrangements. Temperature values calculated by this method agree with the traditional optical two-line absorption method using Sn (284.0 and nm). 143

158 INTRODUCTION Metallic atomizers have been employed in electrothermal atomic spectrometric methods for more than three decades. While not as popular as graphite furnaces, metal devices fashioned into tubes, boats, wires, and coils have functioned as both atomizers and vaporizers in a variety of applications. 1,2 Compared to graphite tubes, metal devices provide higher heating rates, lower power requirements, carbide-free species formation, and simple cooling processes. The material of choice for metal atomizers is tungsten, due to its high melting point (3680 K), low specific heat (0.132 J g -1 K -1 ), and low vapor pressure at temperatures above 2000 K (2.62 x Pa at 2000 K, and 1.59 x 10-6 Pa at 2400 K). 1,3 After their original appearance in the 1970s, metallic atomizers drew very little attention until the development of faster electronics, so the rapid transient analytical signals produced by these devices were easier to monitor. In the subsequent 20 years, tungsten coil atomic absorption spectrometry (WCAAS) has been feasible, with results comparable to graphite furnace atomic absorption spectrometry (GFAAS) in most cases. 4,5,6,7 In 2005, tungsten coil atomic emission spectrometry (WCAES) was reported for the first time. 8 This technique was applied to the determination of Ca, Ba, and Sr, with limits of detection (LOD) below the µg l -1 level. Recently, two other works have described WCAES determination of all 14 lanthanides 9 and eleven other elements. 9, 10 The WCAES method presents some interesting advantages compared to more traditional techniques. The low cost, portability, 144

159 multi-element capacity, and small sample volume requirements make WCAES a viable alternative to popular analytical methods such as flame atomic absorption spectrometry (FAAS) and inductively coupled plasma optical emission spectrometry (ICP OES). On the other hand, WCAES is not without its limitations. Refractory elements, or elements with high boiling points (Os, Pt, B, Ta), are difficult to detect with this method, as well as those elements with particularly high excitation energies (As, Se, Be, Cu, Sn). 11,12 Since the tungsten coil is an open system, WCAES exhibits a significant negative temperature gradient with distance from the atomizer surface. 13 The optical two-line absorption method based on the Boltzmann equation has been applied to the determination of the gas phase temperature near the tungsten coil surface. 14,15 Using the Sn atomic absorption lines at and nm, the gas phase temperature was found to drop at a rate of 250 C per mm away from the coil surface. This lower gas phase temperature may decrease WCAES sensitivity and increase matrix interferences, potentially reducing both precision and accuracy. 16,17 In the present work, two tungsten coils extracted from 15 V, 150 W light bulbs are simultaneously used as atomizers to increase the energy available for the atomization and excitation of the analytes. Two small solid-state constant current power supplies, a crossed Czerny-Turner spectrograph, and a charge coupled device (CCD) detector complete the system. Barium, Sr, Ti and V are the test elements determined by both the single and double coil arrangements. The optical two-line absorption method is used to determine the gas phase 145

160 temperatures in both cases. 14 In addition, a new method for gas phase temperature determination based on atomic emission is developed and the results are compared to the Sn atomic absorption method. Double tungsten coil atomic emission spectrometry (DWCAES) is presented as an alternative to determine refractory elements and to improve WCAES sensitivity, precision, and accuracy. The method is applied to the determination of V and Ti in water samples. 146

161 EXPERIMENTAL Instrumentation Two tungsten coils extracted from commercially available 15 V, 150 W light bulbs (Osram Xenophot HLX 64633, Augsburg, Germany) were housed inside a laboratory-designed glass cell (Ace Glass Part No. D131703, Vineland, NJ, USA). The cell was 10 cm long with four Ace #25 ports. One of the ports was positioned vertically downward. This port was fitted to the 25 cm diameter aluminum mount holding one of the W-coils. This mount was attached to the optical bench. Two ports were positioned 180 apart from each other, in the horizontal plane, forming a T-shape with the base mount port. These two ports were equipped with fused silica windows that allowed for a straight through optical view. The fourth port was also in this horizontal plane, angled 90 to each of the other three, and pointing backwards. This port held the second W-coil mount. The coils were therefore positioned 90 from each other at the center of the optical path. A fifth smaller 10/18 ground glass joint was positioned at the center of the cell, angled up and forwards at 30 above horizontal. This was the sample introduction port. Two programmable solid state DC power supplies (BatMod, Vicor, Andover, MA, USA) were used to provide a constant current to resistively heat the coils during the heating cycle. A 10 % H 2 / Ar gas mixture flowed at 0.9 l min -1 inside the cell to protect the coils from oxidation during the high-temperature atomization step. Additionally, the gas mixture provided a 147

162 reducing atmosphere during the atomic cloud generation and also served as a cooling agent for the atomizers. Analyte atomic emission was collected from the 3 mm space between the two coils. This image was projected onto the entrance slit of a crossed Czerny- Turner spectrograph (MonSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA) by a 25 mm diameter, 75 mm focal length fused silica lens. The spectrograph had a 156 mm focal length and was equipped with a 2400 gr/mm grating (52 x 40 mm), resulting in a reciprocal linear dispersion of 2 nm/mm at 400 nm. The emission signals were detected by a thermoelectricallycooled CCD (Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA) with a two-dimensional array of 1340 x 100 pixels. Each pixel was 20 x 20 μm in size, so the image area on the CCD camera was 26.8 x 2 mm. A 25 µm fixed entrance slit width was used for all measurements. For atomic absorption experiments, a Sn hollow cathode lamp (Fisher Scientific) was positioned on the front end of the cell, and a second 25 mm diameter, 75 mm focal length lens focused its emission between the coils, and eventually onto the same entrance slit. A schematic diagram of the DWCAES system is presented in Figure 1. The image of the lower W-coil was projected approximately 1 mm off of the entrance slit in order to minimize the background radiation (Figure 1C). Results from DWCAES were compared with the single coil arrangement. For this arrangement, a glass cell with only three windows was used (Ace Glass Part No ). Details of the 3-window cell can be found in the literature. 18,19 148

163 Reference solutions and samples All reference solutions were prepared from dilution of single element stock solutions (1000 mg l -1, SPEX CertPrep, Metuchen, NJ, USA) with distilleddeionized water (Milli-Q, Millipore Corp., Bedford, MA, USA). Two different water samples (tap water and thawed snow) were collected and spiked with 2.0 mg l -1 Ti, and 0.4 mg l -1 V to assess the method accuracy. To check V accuracy, a reference material (Water Pollution Standard 1 WPS1-100, VHG Labs, Manchester, NH, USA) was also analyzed. The reference material was diluted with distilled-deionized water and no additional treatment was required before DWCAES determination. Safety considerations Material safety data sheets were consulted before using each chemical reagent. Essential safety precautions were taken in each step of the analysis. Aqueous waste was stored in glass containers prior to being sent for adequate treatment and disposal. Gas phase temperature determination The optical two-line method was applied to determine the gas phase temperature for both double and single coil arrangements. 14 A 2 mg l -1 Sn 149

164 solution was employed, and the absorption signals at and nm were monitored for temperature determination based on the Boltzmann equation. A similar method based on the same equation was developed for temperature determination using atomic emission signals. Solutions of 2 mg l -1 Dy, and 0.1 mg l -1 Eu were used to determine the gas phase temperature in both one and two coil arrangements. Dysprosium emission lines at and nm, and Eu emission lines at and nm were employed. These concentrations were chosen based on the linear dynamic range (LDR) for each element. The highest concentration on the LDR was chosen to ensure maximum emission signals. Atomization cycle The atomization cycle was composed of a series of eight heating steps (Table 1). A µl micropipette (Eppendorf, Brinkmann Instruments Inc., Westbury, NY, USA) was used to place 25 µl of sample directly onto the lower coil. During the first four heating steps, a constant current was passed through the sample coil only. Most solvent vaporization took place during the first two steps. To ensure better control of the vaporization process and prevent sample loss, the potential across the lower coil was monitored during the heating cycle. A sudden increase in potential signaled that the sample coil had reached complete dryness. Heating steps with progressively lower currents prevented premature drying, and reduced the potential for analyte loss during the 150

165 drying/ashing processes. 9 A cooling period (step 5) improved atomization repeatability since the starting temperature prior to atomization was always the same (near room temperature). In step 6, the upper coil was heated to maximum temperature beginning 0.5 s prior to sample atomization. This short period was sufficient to ensure that the upper coil reached full power immediately prior to sample atomization. In this manner, the analyte atoms were introduced into the maximum possible atomization temperature. In step 7, the upper coil remained at full power, while the lower coil was heated to atomization temperature. Finally, a 20-s cooling period prepared the coils to the next heating cycle. Contrary to previous results, no conditioning program was necessary for new W-coils. 9 No significant difference was observed for emission signal intensities produced by either conditioned or unconditioned coils. Integration times as low as 1 ms could be selected in the CCD detector software. For this work, the optimum signal-to-noise ratio (S/N) was observed with an integration time of 500 ms. Eight successive 500-ms spectra were collected (4.0 s), during heating steps 6 and 7. The highest emission signals were observed during the third and fourth spectra for all elements. Peak intensity values were used for all calculations. Each lower (sample) coil withstood approximately 200 heating cycles before replacement. This coil lifetime is similar to that previously reported for single coil methods. 9 The upper coil, which remained sample-free, lasted twice as long (400 cycles). 151

166 RESULTS AND DISCUSSION Gas Phase Temperature One of the biggest limitations of WCAES is that the energy provided by the atomizer is not high enough to atomize and/or excite some elements. 12 Equation 1 describes the relationship between the current (I, in amps) applied to the coil and the temperatures (T, in K) observed at the coil surface. 19 T 309 I (1) Based on this equation, it should be possible to reach temperatures as high as the tungsten melting point (3680 K). In fact, at currents in excess of 11 A, melting of the W-coil is observed. However, the gas phase temperatures observed around the coil can significantly differ from the surface temperature. Because the W-coil is an open atomization system, gas phase temperatures as low as 1800 K are typical only 2 mm away from the coil surface. 15,20 Since the analyte atoms do not spend much time at the coil surface, they may not absorb enough energy to reach the excited state and subsequently they may not emit radiation. Furthermore, due to the high levels of blackbody radiation emanating from the coil surface during the atomization, the region nearest the coil surface does not offer the optimum S/N. Therefore, some elements with high transition 152

167 energies (ΔE 350 kj mol -1 ), or refractory elements such as V, Ti, Os, B, or Ta, produce little or no emission signal. 12 For the current study, the gas phase temperatures approximately 2 mm above the coil surface for the single coil arrangement, and between the two coils for the double coil arrangement, were determined by applying the optical two-line method. 14 In this method, gas phase temperatures were determined by computing the atomic absorption signals for a single element at two closely spaced wavelengths. These values were inserted into an equation derived from the Boltzmann distribution. Specifically, the measured absorbance values for Sn at and nm were inserted into Equation T ( K) = A (2) log(4.9 ) A The high temperature step of the atomization cycle in Table 1 provided a measured gas phase temperature of 1780 K with a single W-coil, and 2220 K for the double coil arrangement. The additional energy provided by the second W- coil facilitated the observation of emission signals for Ag at nm (ΔE = kj mol -1 ), Cu at and nm (ΔE = and kj mol -1 ), Sn at nm (ΔE = kj mol -1 ), and Na at nm (ΔE = kj mol -1 ). 21 None of these lines were observed with the single coil arrangement (Figure 2). 153

168 Based on the same principle, an equation relating the effective temperature of the gas phase, T eff, and the emission intensities for two closely spaced lines for a single element may be derived. T eff E2 E1 = (3) g 2 f 2 I k log( ) g f I Where E 1 and E 2 are the electronic energies of the upper levels from which the two transitions arise; k is the Boltzmann constant; g 1 f 1 and g 2 f 2 are the products of statistical weight and oscillator strengths for the two transitions; and I 1 and I 2 are the emission intensities. Unfortunately, values for gf are not readily available in the literature. However a semi-empirical relationship can be used in this case (Equation 4). 22 The Einstein probability (ga) which is easily found in the literature, 21 may be used to predict gf using a simple relationship (where λ is the wavelength in Å). ga gf λ 16 = x (4) 2 For Dy and Eu, the theoretical g 2 f 2 /g 1 f 1 values calculated from Equation 4 were and 1.313, respectively. These values could differ from experimental values due to inaccuracies in Equation 4, or because of slight changes in optical throughput and detector responsivity between the two wavelengths for emission. These considerations cancel for absorption measurements, where ratios of 154

169 sample and reference intensities are calculated at each wavelength. Therefore, empirical values were calculated based on experimental data: the gas phase temperatures determined by the Sn absorption method, at 10 A atomization current with a single coil. Using Equations 3 and 4, the specific equations relating T eff and emission intensities for Dy and Eu were determined (Equations 5 and 6, respectively). The semi-empirical corrected values for g 2 f 2 /g 1 f 1 are included in these equations. T eff = I (5) log(0.59 ) I T eff = I (6) log(1.42 ) I A comparison between the gas phase temperatures determined by the traditional atomic absorption method with Sn lines (Equation 2) and the ones obtained with Equations 5 and 6 are presented in Table 2. Despite some relatively high standard deviations, the gas phase temperature may be estimated with atomic emission signals. This significantly simplifies the measurement process since no absorption lamp is required, and gas phase temperature could be determined for samples naturally containing these elements. Potentially, a large number of elements could be examined for temperature determinations of this type, so that other samples might naturally contain a temperature detecting 155

170 element. Alternatively, samples could be spiked with Eu or Dy for simultaneous temperature determination. Analytical Figures of Merit Four test elements were chosen to compare emission signals using one and two W-coils: two refractory elements (V and Ti), and two strong emitters (Ba and Sr). Figures of merit are summarized in Table 3. The number of W-coils did not affect the precision or the LDR for any element. However, significant improvements in S/N and limit of detection (LOD) were observed for the refractory elements using two W-coils. For the double coil arrangement, LODs as low as 400 µg l -1 for Ti (399.9 nm), and 10 µg l -1 for V (437.9 nm) were determined. By summing the emission signals of all five different vanadium lines, a slightly lower LOD was observed (9 µg l -1 ). The double coil arrangement improved the LOD for V by a factor of 40 at the most sensitive line (437.9 nm). This remarkable improvement in the V signal is demonstrated in Figure 3. The LOD for Ti improved by a factor of 5 at the most sensitive line (399.9 nm), while those for Sr and Ba improved only slightly (by a less than a factor of 2). Nevertheless, the DWCAES LOD for Ba is the lowest reported to date for a W- coil emission system. To check the method accuracy for refractory elements, a certified reference material (WPS1-100) containing 250 mg l -1 V was analyzed by DWCAES. After simple dilution with distilled-deionized water, the V content was 156

171 determined with 95 % confidence limits at two wavelengths: 246 ± 22 (437.9 nm) and 258 ± 25 (439.5 nm). Other V emission wavelengths produced less accurate results due to spectral interferences. The method proposed was also applied for the determination of V and Ti in two spiked water samples: tap water and thawed snow (Table 4). These samples were spiked with 0.4 mg l -1 V and 2.0 mg l -1 Ti and analyzed by DWCAES. For Ti, the spiked amount was only a factor of five higher than the LOD. Nevertheless, recoveries ranging from 58 to 117 % were obtained, with the best results for Ti at nm. The variability in the recoveries at different wavelengths reflected the relative S/N levels. At the higher wavelength (399.9 nm) the background emission of the coil was higher than at the other three wavelengths, so poorer recovery was observed (58 %). Since all wavelengths were measured simultaneously with different recovery values, other potential causes for poor accuracy (analyte loss or contamination) are difficult to justify. Spectral interferences, on the other hand, might vary at different wavelengths. Recoveries for V were closer to 100 % since the spiked values were ten times higher than the LOD. 157

172 CONCLUSIONS A new W-coil arrangement with two atomizers is a viable alternative for the determination of refractory elements. DWCAES may be most appropriate in cases with low sample volume requirements or the need for simultaneous determination of more than one analyte. The system is small, so field methods requiring portability at low cost could benefit from the design. An increase in the energy available for the atomization and excitation of the analytes is responsible for significant sensitivity improvements especially for refractory elements. Compared to the single coil arrangement, the method provided limit of detection reductions of a factor of 40 for V, and 5 for Ti. The double coil arrangement also enabled the observation of emission signals from elements with transition energies higher than 350 kj mol -1. Emission signals for Ag, Cu and Sn, were observed for the first time with a WCAES system. Pretreatment of new W-coils was not necessary, so atomizer lifetime should increase. Determination of gas phase temperatures with atomic emission signals is also a viable alternative to the traditional atomic absorption method. 158

173 ACKNOWLEDGEMENTS This material is based upon work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET and Grant Award Number 2008-DN-077-ARI

174 Table 1. Tungsten coil heating cycle Step Applied Current / A Time / s Read Upper coil Lower coil No No No No No Yes Yes No 160

175 Table 2. Gas phase temperature measurements Method Element Arrangement Atomization Temperature / K current c / A WCAAS a Sn 1 coil ± 40 1 coil ± coils ± 130 WCAES b Dy 1 coil ± coil ± coils ± 390 Eu 1 coil ± coil ± coils ± 840 a. Temperature determination by tungsten coil atomic absorption spectrometry using the Sn absorption lines at and nm, n = 7. b. Temperature determination by tungsten coil atomic emission spectrometry using the Dy emission lines at and nm, or the Eu emission lines at and nm. c. Current applied to the sample coil during the atomization step (Step 7). The other conditions in the heating cycle are the same as presented in Table

176 Table 3. DWCAES analytical figures of merit Element Emission λ / nm Limit of detection / μg l -1 LDR a Precision b 1 coil 2 coils 1 coil 2 coils 1 coil 2 coils Ba Sr Ti V a. Linear dynamic range in decades, beginning at the detection limit. b. Precision (repeatability) reported as % relative standard deviation for an aqueous solution at the highest point of the LDR. 162

177 Table 4. DWCAES determination of Ti and V in spiked water samples Element Emission λ / nm Spiked water a % Recovery Tap Snow Ti V a. Concentrations added: 2 mg l -1 Ti, 0.4 mg l -1 V. 163

178 FIGURE CAPTIONS Figure 1. DWCAES arrangement: schematic diagram (A), close-up of the atomization cell (B), and coil images projected on the spectrograph entrance slit (C). Figure 2. DWCAES emission signals for some elements with transition energies higher than 350 kj mol -1 : Ag, Cu, Sn = 1000 mg l -1 ; Na = 100 mg l -1 Figure 3. Emission signals for a 2.0 mg l -1 V solution using one (lower trace) and two (upper trace) tungsten coils. 164

179 Figure 1 A 165

180 Figure 2 166

181 Figure 3 167

182 REFERENCES 1. J. A. Nóbrega, M. M. Silva, P. V. Oliveira, F. J. Krug and N. Baccan, Quím. Nova, 1995, 18, G. Zhong, H. Luo, Z. Zhou and X. Hou, Appl. Spectrosc. Rev., 2004, 39, D. R. Lide, Ed., CRC Handbook of Chemistry and Physics. 88th ed. 2008, Boca Raton: CRC, G. L. Donati, K. E. Pharr, C. P. Calloway, Jr., J. A. Nóbrega and B. T. Jones, Talanta, 2008, 76, G. S. Lopes, A. R. A. Nogueira, P. V. Oliveira and J. A. Nóbrega, Anal. Sci., 1999, 15, A. S. Ribeiro, M. A. Z. Arruda and S. Cadore, Quím. Nova, 2002, 25, K. A. Wagner, K. E. Levine and B. T. Jones, Spectrochim. Acta, Part B, 1998, 53, J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr. and B. T. Jones, Spectrochim. Acta, Part B, 2005, 60, G. L. Donati, J. Gu, J. A. Nóbrega, C. P. Calloway, Jr. and B. T. Jones, J. Anal. At. Spectrom., 2008, 23, J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr. and B. T. Jones, Spectrochim. Acta, Part B, 2006, 61,

183 11. K. W. Jackson, Electrothermal Atomization for Analytical Atomic Spectrometry. 1st ed, ed. K.W. Jackson. 1999, Chichester: John Wiley & Sons J. A. Rust, G. L. Donati, M. T. Afonso, J. A. Nóbrega and B. T. Jones, Spectrochim. Acta, Part B, 2009, 64, H. Berndt and G. Schaldach, J. Anal. At. Spectrom., 1988, 3, R. E. Sturgeon and C. L. Chakrabarti, Spectrochim. Acta, Part B, 1977, 32, Z. F. Queiroz, P. V. Oliveira, J. A. Nóbrega, C. S. Silva, I. A. Rufini, S. S. Sousa and F. J. Krug, Spectrochim. Acta, Part B, 2002, 57, P. V. Oliveira, F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz and F. R. P. Rocha, J. Braz. Chem. Soc., 2000, 11, Z. F. Queiroz, F. J. Krug, P. V. Oliveira, M. M. Silva and J. A. Nóbrega, Spectrochim. Acta, Part B, 2002, 57, J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr. and B. T. Jones, Anal. Chem., 2005, 77, A. Salido and B. T. Jones, Talanta, 1999, 50, C. G. Bruhn, F. E. Ambiado, H. J. Cid, R. Woerner, J. Tapia and R. Garcia, Anal. Chim. Acta, 1995, 206, NIST. Basic atomic spectroscopic data. [cited 10/29/2008]; Available from: F. O. Borges, G. H. Cavalcanti and A. G. Trigueiros, Braz. J. Phys., 2004, 34,

184 CHAPTER VIII INDIRECT DETERMINATION OF IODIDE BY TUNGSTEN COIL ATOMIC EMISSION SPECTROMETRY George L. Donati, Joaquim A. Nóbrega, Clésia C. Nascentes and Bradley T. Jones The following manuscript was published in the Microchemical Journal, volume 93, pages , 2009 (DOI /j.microc ), and is reprinted with permission. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati, Joaquim A. Nóbrega and Clésia C. Nascentes. The manuscript was prepared by George L. Donati, Joaquim A. Nóbrega and Clesia C. Nascentes, and edited by Bradley T. Jones. 170

185 ABSTRACT The spectroscopic determination of iodide is a difficult challenge, especially in small sample volumes. The strongest transition lines for this element lie in the vacuum ultraviolet region of the spectrum, so most conventional instruments produce very weak signals. This work describes a tungsten coil atomic emission procedure for the indirect determination of iodide. A 25 μl aliquot of a solution containing a known amount of indium is deposited on the tungsten coil and dried with a simple heating program. Once the coil is dry, 25 μl of an iodide solution is added to the coil. The solution is dried and vaporized at high current. The atomic emission signal for In at nm is monitored. In the presence of iodide, InI is formed and the In emission signal is attenuated. This attenuation is proportional to iodide concentration with a method detection limit of 0.6 mg l -1 iodide using an In concentration of 10 mg l -1, and 3 mg l -1 iodide using an In concentration of 50 mg l -1. Linear calibration curves span a range of two orders of magnitude. Analysis of a deionized water sample spiked with 50 mg l -1 iodide gives a recovery of 100 % and a precision of 5.5 % relative standard deviation. Analysis of a tap water sample spiked with 50 mg l -1 iodide gives a recovery of 140 % and a precision of 7.1 % relative standard deviation. The poor accuracy for the tap water analysis may arise from the reaction of In with other halides in the sample. This is the first report of determination of a halogen using the tungsten coil atomizer. 171

186 Keywords: Iodine; Indium; Tungsten-coil; Electrothermal vaporization; Indirect detection; Atomic emission spectrometry 172

187 1. INTRODUCTION 1.1 Background Iodine is rarely determined using spectroanalytical techniques due to its particular chemical and spectroscopic properties. Characteristics such as high volatility, difficult dissociation, and large excitation energy often require special conditions for the determination of this element by traditional spectrometric methods [1]. The high energy gap between iodine s lowest excited state and its ground state corresponds to analytical resonance lines lying in the vacuumultraviolet spectral region (178.3, 183.0, 184.5, and nm). These low wavelengths represent a significant challenge for most detection techniques. Bermejo-Barrera et al. have reviewed iodide detection procedures based on atomic absorption spectrometry (AAS) and they have referenced only 40 manuscripts published in 30 years [1]. The review discusses both direct and indirect AAS methods. As mentioned above, direct detection requires instrumental adjustments that allow operation at low wavelengths [2]. On the other hand, indirect detection requires one or more chemical reactions between iodide and a different species which can be detected directly by AAS. These indirect methods are often based on the tendency of non-metals to form stable, volatile, diatomic molecules [3]. For example, many of the halogens have been determined following their reaction with Al, Ga, In or Tl. In many cases, however, selectivity becomes a critical issue, since diatomic spectra are not as wavelength 173

188 specific as atomic spectra. In 1989, iodine was determined as thallium iodide by graphite furnace molecular absorption spectrometry or GFMAS [4]. A simple extraction procedure was adopted for the determination of iodide in water and plant leaves. Similarly, chlorine was determined in drinking and ground waters after the formation of AlCl [5]. Aluminium monochloride was characterized by strong absorption bands in the range nm, so a lead hollow-cathode lamp was used as radiation source and the analytical signal was monitored at the Pb nm resonance line. More recently high-resolution, time-resolved instrumentation has been employed in GFMAS, and well defined absorption spectra for InF, InCl, AlF and AlCl have been reported [6]. Bromine has also been determined by AlBr and CaBr molecular absorption [7]. Tungsten coil atomizers have been employed successfully in both atomic absorption and emission measurements [8,9]. The tungsten coil is an open atomizer characterized by a pronounced gradient in the gas phase temperature [10,11]. Gas phase temperature may drop by as much as 250 o C for every millimeter away from the coil surface. This temperature gradient is an undesirable characteristic that may cause severe interferences due to processes occurring in the gas phase [12]. The low temperature gas phase may promote the formation of molecular species. In this case, analyte concentration could be determined by the formation of a diatomic molecule and its subsequent absorption at a specific wavelength. Alternatively, the reaction of the analyte with a probe element may result in the attenuation of the original atomic emission signal for the probe. In this work, iodide is determined by its reaction with In 174

189 forming InI. The InI molecule is then directly detected by molecular absorption of gaseous InI at the nm line emitted by a vanadium hollow-cathode lamp. In addition, a corresponding reduction in the In emission signal, as monitored by tungsten coil atomic emission spectrometry (WCAES), is directly proportional to iodide concentration. These are the first detection techniques reported for a halogen using a tungsten coil atomizer. This communication demonstrates the proof of concept that molecules formed with a tungsten coil atomizer may be used in quantitative spectrometric methods for the determination of anions. 1.2 Formation of InI Indium has been successfully employed for the indirect determination of iodide by molecular absorption methods [3], and it also exhibits a strong WCAES signal [13]. Successful formation of this gaseous diatomic species requires careful optimization of the heating cycle for the tungsten coil. The heating program must be designed to prevent both analyte volatilization during the drying step and InI dissociation during the atomization step. The formation of gaseous InI will result in the reduction of the In emission signal (Eq. (1)) and the simultaneous formation of an InI molecular absorption signal (Eq. (2)). The intensity of either of these signals at a constant In concentration should be proportional to the original I - concentration. In (g) In * (g) In (g) + hν (1) 175

190 In (g) + I (g) InI (g) (2) In addition, the presence of a purge gas (90 % argon plus 10 % hydrogen) and the presence of oxygen in the atomization chamber (from the atmosphere or from the sample solution) add complexity to the potential thermochemical processes that must also be considered: In (g) + O 2(g) InO (g) + O (g) (3) InO (g) + H 2(g) In (g) + H 2 O (g) (4) I (g) + ½ H 2(g) HI (g) (5) These processes depict a potential competition between InO and I for the hydrogen molecules available during the atomization step. Thus, the presence of iodine atoms could reduce indium emission even without the formation of InI. 176

191 2. EXPERIMENTAL 2.1. Instrumental The instrumental arrangement employed for the indirect determination of iodide using a tungsten coil atomizer is shown in Figure 1 [13]. The tungsten filament was extracted from a commercially available 150 W, 15 V light bulb (Osram Xenophot HXL, Pullach, Germany). The fused silica envelope was removed leaving the base intact. The base was mounted in a standard twopronged ceramic power socket. The coil was housed in a glass cell with fused silica windows, and a purge gas (10 % H 2 90 % Ar) flowing at 1.0 l min -1 was used to cool the atomizer and prevent its oxidation during atomization at high temperatures. Solvent vaporization and sample atomization were achieved by resistively heating the atomizer with a programmable solid state constant current power supply (BatMod, Victor, Andover, MA, USA). Radiation in the observation zone was collected by a 25 mm diameter, 75 mm focal length fused silica lens and focused on the entrance slit of a crossed Czerny-Turner spectrograph (MonoSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). For emission measurements, the coil image was positioned approximately 1 mm off-axis from the 25 µm spectrograph slit. This minimized the detection of blackbody radiation originating from the tungsten coil at high temperatures. Indium emission was observed at nm. The spectrograph had a focal length of 156 mm, a 2400 grooves mm -1 grating, and a 2 nm mm -1 reciprocal linear 177

192 dispersion at 400 nm. The analytical signal was detected by a thermoelectricallycooled CCD (Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA), with a two-dimensional array of 1340 x 100 pixels. Depending upon the wavelength region, the system provided a spectral window of approximately 54 nm. Selection of integration times as low as 1 ms was possible. An integration time of 200 ms provided the optimum signal-to-noise ratio (S/N) for emission measurements. A 4 s acquisition time was employed to ensure the collection of all analytical transient signals, so 20 successive spectra were stored for each run. Emission signals used for calibration curves were collected at two different times after the onset of atomization: 1.0 s for maximum sensitivity and 1.4 s for a longer linear dynamic range. For molecular absorption measurements, a V hollow cathode lamp (Perkin Elmer, Norwalk, CT, USA) was used to monitor InI absorbance at nm. An additional fused silica lens with 25 mm diameter and 75 mm focal length focused the HCL radiation approximately 1 mm above the tungsten coil atomizer, and the same lens used for emission measurements carried the unabsorbed radiation to the detection system. In this case, a 50 ms integration time provided the best S/N, so 80 successive spectra were stored during the 4 s acquisition time Reference and sample solutions Stock solutions of 1000 mg l -1 iodide and 10,000 mg l -1 indium were prepared by dissolving NaI and In(NO 3 ) 3, respectively (Fischer Scientific, Fair 178

193 Lawn, NJ, USA), with distilled-deionized water (Milli-Q Millipore Corp., Bedford, MA, USA). All reference solutions for calibration curves were prepared by diluting the iodide stock solution with distilled-deionized water. Two water samples were spiked with 50 mg l -1 iodide: one deionized water sample and one tap water sample. These samples were used to test the accuracy and precision of the technique Heating cycle The tungsten coil heating cycle is listed in Table 1. As previously described [14], the applied constant current (I, A) is proportional to the temperature (T, K) of the dry coil surface such that: T = 309 I The heating steps were identical for absorption and emission measurements except for the high temperature Step 6. A higher temperature was required for atomic emission than for molecular absorption. Solvent vaporization occurred during Steps 1 and 2, and a cautiously controlled sample pyrolysis was carried out in Steps 3 and 4. This prevented both iodide vaporization and InI dissociation prior to the measurement step. The potential across the atomizer was continuously monitored during the heating cycle to ensure efficiency in the drying and pyrolysis processes, and to prevent premature analyte losses [9]. A sudden, sharp increase in the atomizer voltage during the last few seconds of Step 2 indicated complete solvent vaporization. As demonstrated previously [15], a dry coil may reach high temperatures even during the application of relatively low 179

194 currents. Thus, a heating cycle with gradually lower applied currents during the drying and pyrolysis steps was used to prevent analyte loss [9,15,16]. A cooling step (Step 5) was used to ensure repeatability since the temperature immediately prior to the atomization step was always near room temperature. The high current in Step 6 generated enough energy to form the appropriate gas phase species. For the emission measurements, a compromise current prevented the dissociation of InI molecules while providing enough energy to promote In emission (8 A). The detector was triggered at the beginning of Step 6, and a final cooling step prepared the atomizer for the next cycle. In all experiments, a 25 µl sample aliquot was deposited on the tungsten coil with a micropipette (Eppendorf µl, Brinkman Instruments Inc., Westbury, NY, USA). For emission experiments, fresh tungsten coil atomizers were pre-treated to improve both sensitivity and repeatability. The pretreatment conditioning programs are described in detail elsewhere [9] Analytical signal optimization Analytical parameters including purge gas flow rate, atomization current, and sample introduction sequence were optimized separately for absorption and emission measurements. In both cases, gas flow rates in the range l min -1 and atomization currents from 5 to 10 A were tested. The sample introduction sequence was optimized as well. Initially, a single 25 µl aliquot of a pre-mixed solution of In(NO 3 ) 3 and NaI was introduced onto the coil. Then the sequential 180

195 introduction of 10-µl aliquots of In and I - solutions was tested, applying the solutions in either order. Finally, a 25 µl aliquot of In solution was introduced and a drying cycle was applied (Steps 1-4 in Table 1). This was followed by the deposition of a 25 µl aliquot of I - solution, and the complete heating cycle in Table 1 was applied. The reverse of this last approach was also evaluated: drying of the I - solution followed by deposition of the In solution. Blanks were run for all cases, where the I - solution was replaced with distilled-deionized water. 181

196 3. RESULTS AND DISCUSSION 3.1. Molecular absorption As mentioned above, halides have been determined previously by indirect diatomic molecular absorption spectrometry. In order to demonstrate that iodide reacts with indium on the tungsten coil (and hence should result in a reduction of the In emission signal), iodide was determined indirectly by InI molecular absorption using a V hollow cathode lamp (318.4 nm). This diatomic molecule was chosen based on its spectroscopic and physical data as described in the literature [3], and the availability of an adequate radiation source. The heating program and the solution introduction sequence were optimized for the formation of InI molecules either in the condensed or gas phase. The high volatility of iodine necessitated relatively low-temperature drying steps. Additionally, the rapid expansion of the atomic cloud during the high-temperature vaporization step reduced the probability of collisions between indium and iodine in the gas phase. Once formed, the InI had to persist in the gas phase without decomposing as it traveled through the optical path of the absorption system. Vaporization currents (Step 6 in Table 1) in the range 5-10 A were evaluated and the best results were obtained with a value of 6.0 A. Currents lower than 6 A did not provide enough energy to vaporize the sample, so no appreciable number of InI molecules were detected. On the other hand, currents higher than 6 A also resulted in lower absorption signals, perhaps due to 182

197 molecular dissociation occurring at high temperature. Purge gas flow rates ranging from 0.8 to 1.2 l min -1 were evaluated and the best results were obtained at 1.0 l min -1. Higher flow rates may have enhanced the expansion of the atomization cloud, reducing the likelihood of In/I collisions. Lower flow rates may have provided fewer of the H 2 molecules necessary for efficient atomization. The solution introduction order was critical for molecule formation. The best absorption signals were obtained when the indium solution was introduced first and a drying cycle was applied (Steps 1-4 in Table 1). Then the iodide solution was applied to the dry coil and the complete heating cycle in Table 1 was applied. Lower molecular absorption values were observed when the iodide solution was introduced first. In this case, a larger amount of iodine atoms may have been lost during the first drying step due to the absence of indium on the tungsten atomizer. The presence of In could have reduced iodine volatilization by forming InI in the condensed phase. In addition, a portion of the In(III) solution introduced onto the coil prior to I - introduction may have been converted to indium oxides during the drying cycle, and these oxides also may have diminished analyte volatilization before the atomization step. Figure 2 shows the molecular absorption profile for gaseous InI formed under these optimized conditions. The absorbance peak maximum occurred 1 s after the onset of the atomization current, followed by a gradual return to the baseline at the relatively low vaporization temperature. With these optimized conditions a quantitative correlation between iodide concentration and molecular absorption was obtained using a 1000 mg l -1 In solution. Absorbance signals of and were obtained with I - solutions 183

198 of 500 and 1000 mg l -1, respectively. Lower iodide concentrations were difficult to measure due to poor repeatability. Unfortunately, the relatively poor absorptivity of InI at nm renders this technique impractical for samples containing low I - concentrations Atomic emission On the other hand, tungsten coil atomic emission spectrometry (WCAES) produces a very strong signal for In [13]. The population of excited atoms in a tungsten coil atomizer depends upon the temperature of the gas phase, the quantum properties of the element, and the chemical environment in which the atom cloud is generated. As mentioned above, the tungsten coil is an open atomizer with a pronounced temperature gradient. The gas phase temperature is known to decrease as the atom cloud expands around the tungsten surface. This temperature gradient could be exploited for indirect analytical applications. Lower atomizer temperatures provide more favorable conditions for the generation of molecules in the gas phase, and they also minimize the dissociation of those molecules once formed. Consequently, the presence of iodide should decrease the indium atomic emission signal, and this attenuation should be proportional to iodide concentration. Indeed, the maximum emission signal observed for a 10 mg l -1 indium solution decreases upon the addition of 5 mg l -1 I - (Figure 3). As expected, the formation of InI in either the condensed or 184

199 gas phases resulted in a reduction of excited In atoms in the atomic cloud and a consequent drop in the emission signal. Previous studies demonstrated that the optimum applied current for maximum indium emission is 10 A [13]. However, no attenuation of the In emission signal was observed by adding different concentrations of iodide at this current. The high gas phase temperature at 10 A must have inhibited the formation of the InI molecule. On the other hand, no In emission signal was observed when atomization currents lower than 6 A were employed. In this case, the energy provided by the atomizer was insufficient to promote atomic emission. Therefore, a compromise vaporization current was chosen. Varying the vaporization current in the range 5-10 A resulted in a maximum S/N at 8 A. As with the molecular absorption arrangement, the optimal purge gas flow rate was 1.0 l min -1. The best sample introduction sequence again was indium application followed by a drying cycle, then iodide application followed by the complete heating program. Under these conditions, linear calibration curves were obtained for iodide using either 10 or 50 mg l -1 In concentrations (Figure 4). The curves were plotted as percent attenuation of the In emission signal versus I - concentration. Maximum iodide sensitivity (Figure 4A) was observed for a 10 mg l -1 In concentration when the emission signal was collected 1.0 s after the onset of atomization (Figure 3). The limit of detection (LOD) in this case was 0.6 mg l -1 iodide. The linear range extended only up to 50 % attenuation, or 7 mg l -1. At 50 % attenuation, the observed signal fell 5 % below the signal expected for the 185

200 extrapolated linear fit. The resulting linear dynamic range was only slightly more than one order of magnitude. The linear dynamic range was increased by simply plotting the emission signals for the same runs collected 1.4 s after the onset of atomization (Figure 4B). The detection limit in this case was only slightly worse using a 10 mg l -1 In concentration: 0.7 mg l -1 iodide. The linear range again extended up to 50 % attenuation, or 70 mg l -1. The linear range was extended even further by using a higher In concentration (50 mg l -1 ). This resulted in an LOD of 3 mg l -1 iodide and an upper limit of 160 mg l -1. Figure 4A shows that the highest calibration sensitivity (m) for the I - determination was 7.27 % / (mg l -1 I - ) using 10 mg l -1 In. This means that 1 mg l -1 I - reacted with 7.27 % of the 10 mg l -1 In present, or mg l -1 In. So the mass fraction for the reaction of the two species was In/I. Taking the atomic masses into account, the mole fraction for the reaction was slightly less than one: In/I. This quotient is called the Transfer Ratio (TR) for indirect detection methods in chromatography, and typically is smaller than expected [1]. TR may be used to convert In concentration to I concentration in the LOD equation [18]: 3 σ 3 N 3 C In In LODI = = = ( 6) m S C / TR ) TR ( S / N) /( In In I I In Substituting for C In (10 mg l -1 ), the S/N for repetitive measurements of the In solution in the absence of I - (66.7), and TR I (0.727) gives an iodide LOD of 0.6 mg l -1. This LOD is depicted by the triangle in Figure 4A. The LOD might be 186

201 improved by using lower In concentrations, increasing the S/N for the In emission signal, or increasing TR. In practice, however, changes in these values may offset each other [18]. The accuracy and precision for the indirect method were tested by repetitive analysis of two water samples spiked with 50 mg l -1 iodide: one deionized water and the other tap water. The recovery for the deionized water sample was 100 % and the precision was 5.5 % relative standard deviation (n = 5). For the tap water the recovery was 140 % and the precision was 7.1 % RSD. These precision values are typical for WCAES measurements. The overestimation of the iodide level in the tap water sample is likely caused by the presence of other halides in the sample (chloride or fluoride). These species may also react with In to form gaseous diatomic molecules. For samples with appreciable levels of other halides, a separation step may be required prior to the WCAES analysis. This preliminary step has also been suggested for indirect determination by molecular absorption spectrometry [4]. In conclusion, these results demonstrate the feasibility of measuring iodide indirectly by tungsten coil atomic emission spectrometry. Iodide is difficult to determine spectroscopically, and this method provides a detection limit lower than 1 part per million. Ongoing investigations will examine the selectivity of the procedure, the use of alternative emission probe elements (which may provide higher S/N for lower concentrations), and application to different non-metal species. 187

202 ACKNOWLEDGEMENTS This material is based upon a work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET and Grant Award Number 2008-DN-077-ARI J.A.N. is thankful to the Department of Chemistry (Federal University of São Carlos) for his leaving license, and to the National Council of Science and Technological Development (CNPq) and the State of São Paulo Research Foundation (FAPESP) for funds. C.C.N. is thankful to the Department of Chemistry (Federal University of Minas Gerais) for her leaving license. 188

203 FIGURE CAPTIONS Figure 1. Schematic diagram of the tungsten coil instrumentation. The inset shows a close-up view of the monochromator entrance slit with the coil image positioned for emission measurements. Figure 2. Molecular absorption profile observed for InI at nm using a V hollow cathode lamp as the radiation source. The vaporization step begins at time 0. Solution concentrations were 1000 mg l -1 In and 500 mg l -1 I -. Figure 3. Atomic emission profiles for 10 mg l -1 In at nm in the absence of I - (squares) and in the presence of 5 mg l -1 I - (diamonds). The vaporization step begins at time 0. The maximum iodide sensitivity is observed at 1.0 s after the onset of atomization, and the longest linear dynamic range is achieved at 1.4 s. Figure 4. Analytical calibration curves for the indirect determination of iodide by tungsten coil atomic emission spectrometry. The Indium atomic emission signal was measured at nm in all cases. The highest Iodide sensitivity is observed with a 10 mg l -1 In solution and a measurement time of 1.0 s after the onset of atomization (A). The limit of detection is depicted by the triangle. Longer linear dynamic ranges are observed at a measurement time of 1.4 s with either 10 of 50 mg l -1 In (B). 189

204 Figure 1 Lens Lens HCL C C D 10 % H 2 / Ar Constant current power supply D/A Converter Spectrograph slit Slit aperture Coil image 190

205 Figure Absorbance at nm Time (s) 191

206 Figure m n t a l a n ig S n io s is m E In e tiv la e R Relative In Emission Signal at nm Maximum Sensitivity Longer Linear Range Time (s) 192

207 Figure 4A tive In Emission Signal at nm Rela n io s is m E m iu d In f o n tio a u n te A t n e rc e P A 10 mg l -1 In Iodide Concentration (mg l -1 ) 193

208 Figure 4B Percent Attenuation of Indium Emission 30 n io s is m E m iu d In f o 20 n tio a u n te A 10 t n e rc e P B 10 mg l -1 In 50 mg l -1 In Iodide Concentration (mg l -1 ) 194

209 Table 1. Tungsten coil heating cycle. Step Applied current (A) Time (s) a 6 b or 8 c a. Data acquisition occurred during the first 4 s of Step 6. b. Vaporization current for molecular absorption. c. Atomization current for atomic emission. 195

210 REFERENCES [1] P. Bermejo-Barrera, M. Aboal Somoza, A. Bermejo-Barrera, Atomic absorption spectrometry as an alternate technique for iodine determination ( ), J. Anal. At. Spectrom. 14 (1999) [2] J. M. Mansfield, T. S. West, R. M. Dagnall, Determination of iodine by atomicabsorption spectrometry using the platinum-loop technique, Talanta 21 (1974) [3] L. H. J. Lajunen, Spectrochemical Analysis by Atomic Absorption and Emission, 1st ed., RSC, Cambridge, 1992, pp [4] Y. Zhi-He, H. Hui-Ming, Determination of iodine by molecular absorption spectrometry of thallium iodide in the graphite furnace, Spectrochim. Acta Part B 44 (1989) [5] P. Parvinen, L. H. J. Lajunen, Determination of chloride in drinking and ground water by AlCl molecular absorption spectrometry using graphite furnace atomic absorption spectrometer, Talanta 50 (1999) [6] M. R. Shepard, B. T. Jones, D. J. Butcher, High-resolution time-resolved spectra of indium and aluminum atoms, fluorides, chlorides, and oxides in a graphite tube furnace, Appl. Spectrosc. 52 (1998) [7] M. D. Huang, H. Becker-Ross, S. Florek, U. Heitmann, M. Okrus, High-resolution continuum source electrothermal absorption spectrometry of AlBr and CaBr for the determination of bromine, Spectrochim. Acta Part B 63 (2008)

211 [8] X. D. Hou, K. E. Levine, A. Salido, B. T. Jones, M. Ezer, S. Elwood, J. B. Simeonsson, Tungsten coil devices in atomic spectrometry: absorption, fluorescence and emission, Anal. Sci. 17 (2001) [9] G. L. Donati, J. Gu, J. A. Nóbrega, C. P. Calloway, B. T. Jones, Simultaneous determination of the lanthanides by tungsten coil atomic emission spectrometry, J. Anal. At. Spectrom. 23 (2008) [10] Z. F. Queiroz, P. V. Oliveira, J. A. Nóbrega, C. S. Silva, I. A. Rufini, S. S. Sousa, F. J. Krug, Surface and gas phase temperatures of a tungsten coil atomizer, Spectrochim. Acta Part B 57 (2002) [11] J. A. Rust, J. A. Nóbrega, C. P. Calloway, B. T. Jones, Analytical characteristics of a continuum source tungsten coil atomic absorption spectrometer, Anal. Sci. 21 (2005) [12] P. V. Oliveira, F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz, F. R. P. Rocha, Influence of Na, K, Ca and Mg on lead atomization by tungsten coil in atomic absorption spectrometry, J. Braz. Chem. Soc. 11 (2000) [13] G. L. Donati, B. E. Kron, B. T. Jones, Simultaneous determination of Cr, Ga, In and V in soil and water samples by tungsten coil atomic emission spectrometry, Spectrochim. Acta Part B, Special Issue Dedicated to the 10th Rio Symposium on Atomic Spectrometry, 64 (2009) [14] K.E. Levine, K.A. Wagner, B.T. Jones, "A low-cost, modular electrothermal vaporization system for inductively coupled plasma atomic emission spectrometry," Appl. Spectrosc., 52 (1998)

212 [15] P. V. Oliveira, M. Catanho, J. A. Nóbrega, P. O. Luccas, Avaliação de programas de aquecimento para espectrometria de absorção atômica com atomização eletrotérmica em filam ento de tungstênio, Quím. Nova 23 (2000) [16] G. L. Donati, K. E. Pharr, C. P. Calloway, Jr., J. A. Nóbrega, B. T. Jones, Determination of Cd in urine by cloud point extraction-tungsten coil atomic emission spectrometry, Talanta 76 (2008) [17] S.M. Cousins, P.R. Haddad, W. Buchberger, Evaluation of carrier electrolytes for capillary zone electrophoresis of low-molecular-mass anions with indirect UV detection, J. Chromatogr. A 671 (1994) [18] P. Doble, P.R. Haddad, Indirect photometric detection of anions in capillary electrophoresis, J. Chromatogr. A 834 (1999)

213 CHAPTER IX A NEW ATOMIZATION MICRO-CELL FOR TRACE METAL DETERMINATIONS BY TUNGSTEN COIL ATOMIC SPECTROMETRY G. L. Donati, R. B. Wildman and B. T. Jones The following manuscript was prepared for submission to Analytica Chimica Acta. Stylistic variations are due to the requirements of the journal. All of the presented research was conducted by George L. Donati and Robert B. Wildman. The manuscript was prepared by George L. Donati and edited by Bradley T. Jones. 199

214 ABSTRACT A new metallic atomization cell is used for trace metal determinations by tungsten coil atomic absorption spectrometry and tungsten coil atomic emission spectrometry. The atomizer is a tungsten filament extracted from 15 V, 150 W microscope light bulbs. A small solid-state power source, a Czerny-Turner spectrograph, and a thermoelectrically-cooled charged coupled device detector are used in this potentially portable system. Different protective gas mixtures are evaluated to improve atomic emission signals. Ar, N 2, CO 2 and He are used as solvents, and H 2 and C 2 H 2 as solutes. A 10% H 2, 90% Ar mixture provided the best results. Parameters such as protective gas flow rate and atomization current are also optimized. The optimal conditions are used to determine the figures of merit for both methods and the results are compared with values found in the literature. The new cell provides a better control of the radiation reaching the detector and a small, more isothermal environment around the atomizer. A more concentrated atomic cloud and a smaller background signal result in lower limits of detection using both methods. Cu (324.7 nm), Cd (228.8 nm) and Sn (286.3 nm) determined by tungsten coil atomic absorption spectrometry presented limits of detection as low as 0.6, 0.1, and 2.2 µg L -1, respectively. For Cr (425.4 nm), Eu (459.4 nm) and Sr (460.7 nm) determined by tungsten coil atomic emission spectrometry, limits of detection of 4.5, 2.5, and 0.1 µg L -1 were calculated. The method is used to determine Cu, Cd, Cr and Sr in a water standard reference material. Results for Cu, Cd and Cr presented no significant difference from 200

215 reported values in a 95% confidence level. For Sr, a 113% recovery was obtained. The method is a simple, inexpensive, sensitive alternative to more traditional atomic spectrometric methods. The use of a new, small metallic cell combined with compact, simple equipment represents a step toward a portable, reliable and sensitive method for trace metal analysis. Keywords: Tungsten coil; Micro-cell; Atomic absorption; Atomic emission; Trace metal analysis; Portability 201

216 1. INTRODUCTION Atomic absorption and atomic emission spectrometry have become some of the most important methods for trace metal analysis in the last few decades. Characteristics such as high sensitivity, simplicity, wide linear range, robustness, precision and accuracy have placed methods such as flame atomic absorption spectrometry (FAAS), graphite furnace atomic absorption spectrometry (GFAAS), and inductively coupled plasma optical emission spectrometry (ICP OES) among the most preferred tools for the determination of metals at low levels in a wide range of samples [1]. Despite all this success, traditional atomic spectrometric methods are relatively expensive and present portability limitations related to equipment size, power, gas and cooling requirements. Thus, increasing new demand for less expensive, portable methods has driven research for new strategies to be used in field applications [2-5]. An interesting alternative to those methods is the use of metallic atomizers in electrothermal atomic absorption spectrometry. Tubes, boats, wires and coils have been successfully used as atomizers and vaporization devices for decades [6]. Due to higher heating rates, lower power requirements and simple cooling processes, metallic atomizers may represent the perfect choice for portable devices. Recently, a portable car battery-powered tungsten coil atomic absorption spectrometry (WCAAS) system was used to determine Pb and Cd in the field [2-3]. In 2005, a new, simple, inexpensive, potentially portable method was introduced by Rust et al. [7]. Tungsten coil atomic emission spectrometry (WCAES) uses a filament extracted 202

217 from mass produced 15 V, 150 W projector light bulbs as atomizer, and a small solid-state power source to produce atomic emission signals which are monitored by a high resolution Czerny-Turner spectrograph and a charged coupled device (CCD) detector. In that work, Ca, Ba and Sr were determined in water samples with limits of detection (LOD) below the µg L -1 level. Further improvements were made on the system and the determination of a wide range of elements was possible at low µg L -1 levels [8-11]. Although presenting several advantages over more traditional methods, some limitations prevent a more disseminated use of atomic spectrometric methods based on tungsten coil atomizers. Because this is an open system, severe matrix interference is observed in the determination of different analytes. The gas phase temperature can drop by as much as 200 ºC for each millimeter away from the coil surface [12], which may cause significant negative effects on the atomization processes. As had occurred with GFAAS before the introduction of the stabilized temperature platform furnace (STPF) method [13], problems related to recombination and molecular absorption due to a non-isothermal environment are responsible for severe spectral and chemical interferences. On the other hand, determinations using analytical signals collected too close to the coil surface may be impractical. An intense background signal due to blackbody radiation emitted by the coil during the atomization step can saturate the detector before any analytical signal is collected. This problem becomes even more evident in WCAES determinations [8-11]. Some strategies have been applied to minimize such limitations in tungsten coil-based methods. The use of a light- 203

218 blocking aperture between the atomization cell and the spectrograph entrance slit [9], as well as the positioning of the coil image slightly off axis to the detector [10-11] have been reported to successfully reduce background signal in WCAES. Additionally, application of different modifiers [14], separation methods [15], and special system arrangements [16] have been reported to reduce matrix interferences and increase sensitivity in WCAAS determinations. In this work, a new atomization micro-cell is used to increase sensitivity and reduce background signal in both WCAAS and WCAES determinations. The aluminum cell s reduced volume provides both a more isothermal environment and a more concentrated atomic cloud, which may prevent recombination and molecular absorption, as well as increase the intensity of the analytical signal. Two different cap designs provide lower background signals due to positionadjustable coil-blocking ports. Respectively, pinhole-size and slit-shape ports are used for atomic absorption and atomic emission determinations. Different gas mixtures may be introduced into the atomization cell through two gas entrances at its base. Mixtures of Ar, N 2, CO 2, or He with different concentrations of H 2 are evaluated as protective gas in WCAES. Argon or H 2 /Ar 1:9 mixed with different concentrations of C 2 H 2 are also evaluated. Important parameters affecting sensitivity such as protective gas flow rate and atomization current are optimized. Tungsten coils extracted from 15 V, 150 W microscope light bulbs are used as atomizers. Power is supplied by a small solid-state constant current source. A Czerny-Turner spectrograph and a charged coupled device (CCD) detector complete the system arrangement. Copper, Cd and Sn are determined by 204

219 WCAAS, while Cr, Eu and Sr are simultaneously determined by WCAES. The new arrangement with a windowless atomization micro-cell is more portable, rugged, and potentially more sensitive than previous systems. The method is applied to the determination of Cu, Cd, Cr and Sr in a water standard reference material. 205

220 2. EXPERIMENTAL 2.1. Instrumentation A schematic diagram of the equipment arrangement for both WCAAS and WCAES is presented in Fig. 1. The atomizer is a tungsten coil extracted from commercially available microscope light bulbs (Osram Xenophot HLX 64633, Augsburg, Germany). The fused silica envelope was removed, leaving the filament and bulb base intact. The bulb base was mounted in a standard ceramic two-pronged power socket which was countersunk about 1.5 cm into the base of a laboratory-designed aluminum cell. The cell, fashioned from a 2.5-cm diameter Al rod, is 9 cm long with a 1.8 cm i.d. and two gas entrances at its base. The hollow, base-closed cell is attached to the optical bench and a snug-fitting cap is inserted at the opposite end to house the atomizer. The power socket presents two 2-mm orifices to allow the protective gas mixture to flow into the atomization cell cap. Different cap designs were used for WCAAS and WCAES (Fig. 1B). Both caps are 3.8 cm long, with 1.4 cm i.d. and 1.6 cm o.d. An Al 2.5-cm diameter, 3-mm thick ring was fused outside the cap to prevent it from totally sliding into the atomization cell. A screw passes through the ring and serves to adjust the ports height relative to the detector, which can block part of the coil blackbody radiation and reduce background signal. A 0.5-cm high, 0.4-cm wide sample introduction port is positioned 1.3 cm above the base of the cap s ring, in a 90º angle from the other two ports. The WCAAS cap has two 1.5-mm diameter 206

221 ports separated by a 180º angle, positioned approximately 2 mm off-axis from the cell center and 0.8 cm above the base of the cap s ring. The WCAES cap has two slit-shape ports separated by a 180º angle, sitting 0.4 cm above the base of the cap s ring, at the cell center. The slit facing the detector (front slit) is 0.2 cm high, 0.6 cm wide, and the back slit is 0.3 cm high, 1.2 cm wide. With this design, additional adjustment and better control of the amount of radiation being blocked are possible by rotating the cap in different angles. Fig. 1 A small programmable solid-state DC power source (BatMod, Vicor, Andover, MA, USA) was used to provide a constant current and resistively heat the coil during the heating cycle. A gas mixture flows inside the cell to protect the coil from oxidation, provide a reducing atmosphere and cool the atomizer during the high-temperature atomization step. Analyte atomic emission exiting the cap slit was collected by a 25 mm diameter, 75 mm focal length fused silica lens and the 1:1 image was projected onto the 25-µm entrance slit of a crossed Czerny- Turner spectrograph (MonSpec 18, Scientific Measurement Systems Inc., Grand Junction, CO, USA). The spectrograph has a 156 mm focal length and is equipped with a 2400 gr/mm grating (52 x 40 mm), resulting in a reciprocal linear dispersion of 2 nm/mm at 400 nm. The emission signals were detected by a thermoelectrically-cooled CCD (Spec-10, Princeton Instruments, Roper Scientific, Trenton, NJ, USA) with a two-dimensional array of 1340 x 100 pixels. For 207

222 WCAAS, either a hollow cathode lamp (Cu HCL, Photron, Narre Warren, Australia), or an electrodeless discharge lamp (Cd EDL, Perkin Elmer, Norwalk, CT, USA; or Sn EDL, Westinghouse, Philadelphia, PA, USA) was positioned on the opposite side of the cell, and a second 25 mm diameter, 75 mm focal length lens focused its emission into the cap, passing through the 1.5-mm ports, and eventually onto the same spectrograph s entrance slit. For additional background reduction in WCAES, the partially blocked coil image was projected approximately 0.5 mm off of the spectrograph entrance slit (Fig. 1C). The CCD detector software provides user selectable integration times as low as 1 ms. For Cu, the best signal-to-noise ratio (S/N) was obtained with a 30-ms integration time. For Cd and Sn, an integration time of 20 ms presented the best results. In all WCAAS determinations, several spectra were collected, adding to a 3-s atomization step. For WCAES, the best S/N values were obtained with 0.5-s integration times. In this case, 8 spectra were collected, totalizing 4 s of atomization Standards All reference solutions were prepared from dilution of single element stock solutions (1000 mg l -1, SPEX CertPrep, Metuchen, NJ, USA) with distilleddeionized water (Milli-Q, Millipore Corp., Bedford, MA, USA). A standard reference material from the National Institute of Standards and Technology (Trace elements in water, SRM #1643e) was used to check the method accuracy. 208

223 The reference material was diluted with distilled-deionized water and no additional treatment was required before WCAAS or WCAES determinations Atomization cycle A µl micropipette (Eppendorf, Brinkmann Instruments Inc., Westbury, NY, USA) was used to place 25 µl of sample directly onto the coil and a 7-step atomization cycle (Table 1) was applied to promote sample vaporization, pyrolysis and atomization. Heating steps with progressively lower currents and monitoring of the potential across the coil during the heating cycle ensured a better control of solvent vaporization and prevented sample loss during the drying/ashing processes [9]. Most solvent vaporization took place during the first two steps, with sample pyrolysis and some matrix elimination occurring in steps 3 and 4. A cooling period (step 5) was introduced to improve atomization repeatability. In this case, variations from cycle to cycle and even from sample to sample would be minimized since the starting temperature prior to atomization was always the same (near room temperature). In step 6, a high current generated the atomic cloud which was used in WCAAS determinations. For WCAES, the energy provided by the atomizer at this high current step was also enough to promote analyte excitation and atomic emission. Finally, a 20-s cooling period readied the atomizer for the next heating cycle. A conditioning heating program for brand new tungsten coils was used to improve the method 209

224 precision and WCAES sensitivity. Details on the conditioning program may be found in the literature [9]. Table Optimization studies Different protective gas mixtures were used to improve WCAES analytical signals. Mixtures of Ar, N 2, CO 2 or He with H 2 (0, 10, 15, 20, 25, 30, and 40%) or C 2 H 2 (5 and 10%) were tested and their effect as protective gas on Ca, Cr, Eu and Sr emission signals, as well as on the atomizer lifetime, was evaluated. A mixture of H 2 /Ar 1:9 with C 2 H 2 (5 or 10%) was also studied. In another study, a multielement solution containing Cr, Eu (200 µg L -1 ) and Sr (20 µg L -1 ) was subjected to solvent vaporization and sample pyrolysis (steps 1-4 in Table 1) using a protective gas composed of either 0 or 10% H 2 in Ar before atomization with a 1:9 H 2 /Ar gas mixture. The objective was to evaluate the effect of H 2 on potential solid phase reactions occurring on the coil surface prior to atomization [12,17]. Important parameters for both WCAAS and WCAES senstivities, such as protective gas flow rate and atomization current, were also evaluated. Gas flow rates ranging from L min -1 and atomization currents from 6-10 A were studied and their effects on both WCAAS and WCAES analytical signals for Cu, 210

225 Cd, Sn (atomic absorption) and Ca, Cr, Eu and Sr (atomic emission) were evaluated. 211

226 3. RESULTS AND DISCUSSION 3.1. Background signal reduction A non-isothermal environment is probably the most critical limitation for atomic absorption spectrometric methods based on tungsten coil atomizers. Severe matrix effects due to recombination and molecular absorption in a cooler gas phase frequently reduce WCAAS precision and accuracy [12,15]. To overcome this problem, a metallic small-volume atomization cell with windowless ports was used. The new micro-cell design may provide a less pronounced gas phase temperature gradient, minimizing matrix interference and improving precision, accuracy and sensitivity. Additionally, a significant part of the background radiation emanating from the coil was blocked by the pinhole-size, windowless ports positioned approximately 1 mm away from the atomizer. A height adjustment screw (Fig. 1B) also allowed fine positioning and signal collection in different places around the coil. Three test elements were determined by WCAAS using the new atomization cell design. Copper was determined at two different wavelengths, i.e and nm, with limits of detection below µg L -1. The temporal background correction method [15,18] was applied for Cu determinations. Since the analytical signal was monitored in 30- ms intervals, the BG signal was recorded 60 ms before the appearance of the Cu peak at the analytical wavelengths. Then, the BG was subtracted from the Cu absorption peak providing the net analytical signal. Cadmium and tin were also 212

227 determined by WCAAS and LODs at low µg L -1 levels or even lower were obtained. For Cd and Sn, the near line BG correction method [15,19-20] was used. In this method, two resonance lines from the EDL radiation source are monitored: the analytical line, which is sensitive to both analyte and BG absorptions, and a near BG sensitive line, which is not sensitive to the analyte and thus reflects only the BG absorption signal. The analytical lines used were and nm, while the BG absorption lines were and nm for Cd and Sn, respectively. Fig. 2 presents an absorption spectrum obtained with a 2 µg L -1 Cd solution using the near line BG correction method. The solid gray line represents the nm analytical line, the dotted line represents the nm BG sensitive line, and the solid black line represents the net absorption signal for Cd. Fig. 2 WCAES greatest limitation is the high background interference caused by the coil s intense blackbody radiation at high temperatures. An alternative to reduce such interference is to carry out atomic emission measurements away from the coil. However, this strategy results in less intense analytical signals due to a significant drop in gas phase temperatures, which may make the determination of some elements impossible [6]. To overcome these limitations, the new micro-cell design provides a small volume area around the atomizer, which may represent a more isothermal environment allowing measurements 213

228 farther from the coil. The slit-shape ports and the height adjustment screw (Fig. 1B) allow a fine adjustment of the position and amount of radiation reaching the detector. Additionally, more blackbody radiation can be blocked by turning the atomization cell s cap through different angles. Fig. 3 shows pictures of the metallic atomization cell with (3B) and without the cap (3A). It can be seen that significant BG reduction and consequent method sensitivity improvement is possible simply by rotating the cap through different angles. Fig. 3 Three test elements were simultaneously determined by WCAES using the new micro-cell: Cr, Eu and Sr. Limits of detection at the µg L -1 level and lower were obtained by the method. Fig. 4 presents a BG corrected spectrum obtained in a 50 nm spectral window. The net emission signal was calculated by recording the blank spectrum (distilled-deionized water) and subtracting it from a multielement solution spectrum. Fig Protective gas and atomization current optimization An important parameter in both WCAAS and WCAES is the nature of the protective gas flowing inside the atomization cell during the heating cycles. While 214

229 several works have studied the effects of different mixtures on WCAAS analytical signal [12,17,21], no study has been done for WCAES. The new metallic cell used in this work presents the advantage of having two gas entrances at its base, which facilitate gas mixing on site, inside the cell. Mixtures containing Ar, N 2, CO 2, or He with H 2 (0-40%) or C 2 H 2 (5 or 10%) were evaluated to improve WCAES signal for Ca, Cr, Eu and Sr. It was observed that mixtures presenting even small concentrations of CO 2 or C 2 H 2 could severely damage the atomizer without producing any measureable analytical signal. This may be related to carbon deposition on the coil surface, which would contribute to rapid metal degradation and a high BG signal due to smoke. In fact, a thick layer of a black substance (most likely a mixture of carbon and tungsten carbide) was observed on atomizers heated in CO 2 or C 2 H 2 atmospheres. For the other mixtures evaluated, a trend related to H 2 concentration was observed in all cases. The analytical signal increases as the H 2 concentration increases up to a maximum, and then it drops as more H 2 is added to the mixture (Fig. 5A). At optimum H 2 /solvent ratio conditions, the best results for all elements evaluated were obtained using Ar. An analytical signal trend following the order He < N 2 < Ar was also observed (Fig. 5B). Those results are in accordance to these gases thermal conductivities: 0.152, , and W m -1 K -1 for He, N 2 and Ar, respectively [22]. An atmosphere with He would dissipate heat from the atomizer much faster than an Ar mixture. Thus, less energy would be available for atomic excitation and emission, and a less intense analytical signal would be observed. Thermal conductivity may also explain why higher H 2 concentrations cause a signal 215

230 depression for all elements studied. From Fig. 5A, it can be observed that H 2 is necessary for atomic emission, probably due to its reducing properties [12,17,21]. However, above a certain concentration, H 2 relatively high thermal conductivity (0.182 W m -1 K -1 ) [22] would overcome the benefits of its reducing effect on atomic cloud generation and negatively affect the analytical signal. This fact may be confirmed by Cr emission intensities presented in Fig. 5A. Since this is a more refractory element, it would be expected to show a sharper signal depression with higher H 2 concentrations. In fact, the optimal H 2 % v/v for Cr is 10%, while 20 % is the optimal H 2 concentration for all other elements studied. Fig. 5 In another study, a multielement solution containing Ca, Cr, Eu and Sr was submitted to the heating cycle presented in Table 1 with and without the presence of 10% v/v H 2 during solvent vaporization and sample pyrolysis (steps 1-4). Smaller analytical signals were observed for all elements studied when pure Ar was used. From these results, it can be supposed that H 2 also plays an important role in solid phase reactions occurring on the coil surface prior to atomization. Atomization processes in electrothermal atomic spectrometry are little understood, but the effect of H 2 as a reducing agent may explain the sensitivity improvements observed with a 1:9 H 2 /Ar protecting gas mixture [12,17,21]. 216

231 Protecting gas flow rates and atomization currents were also optimized for both WCAAS and WCAES for all elements studied. Gas flow rates ranging from L min -1 and atomization currents from 6-10 A were evaluated. The best analytical signals were obtained with gas flow rates of 0.6 L min -1 for Sr and Sn, 1 L min -1 for Cu and Cd, and 1.2 L min -1 for Cr and Eu. On the other hand, atomization currents of 7 A for Cu, 8 A for Eu, 9 A for Cd, Sn and Cr, and 10 A for Sr presented the best S/N values Analytical figures of merit Using the optimal conditions described above, the analytical figures of merit for each element were determined using standard aqueous solutions. Considering the relatively small difference in signal intensity for Eu and Sr using either 20% or 10% v/v H 2 in Ar (Fig. 5A), and the mixture commercial availability, all determinations were carried out using the latter. Table 2 presents the analytical figures of merit for Cu, Cd and Sn determined by WCAAS, and Cr, Eu and Sr determined by WCAES using the new atomization micro-cell. Limits of detection are comparable, and sometimes lower than the values reported in the literature for WCAAS. For Sn, a 540-fold lower LOD is obtained with the new method. For WCAES, LODs are compared to results obtained previously with a similar system employing a conventional glass atomization cell. Slightly better results are observed with the new method for most elements. 217

232 Table 2 To check the method accuracy, Cu, Cd, Cr and Sr were determined in a water standard reference material (NIST SRM #1643e). Table 3 presents the results. A standard addition method was applied for Cu and no statistical difference was observed between determined and reported values for Cu, Cd and Cr at a 95% significance level. For Cr determined at 429.0, a much higher value than reported was obtained. This fact may be explained by high concentrations of concomitants presenting emission lines in the same region: V at nm, Mo at nm [25]. On the other hand, a 113 % recovery was observed for Sr, which may reflect a high concentration of alkali-earth concomitants in the sample. The interference processes are not totally understood, but elements such as Ca and Ba can significantly affect Sr emission signal, limiting this element determination in several samples [7,26]. Table 3 218

233 4. CONCLUSIONS Atomic spectrometric methods based on tungsten coil atomizers may be considered an important strategy to develop reliable portable systems for trace metal analysis. The use of a new atomization micro-cell has provided a more isothermal environment, with the atomic cloud concentrated in a smaller volume, and a better control of the background radiation reaching the detector. The new method presented better limits of detection for most elements (four out of six) determined by WCAAS (i.e. Cu and Sn) and WCAES (i.e. Cr and Sr). It is simple, inexpensive, compact, sensitive and potentially portable. The protective gas mixture composed of H 2 dissolved in Ar presented the best results for all elements determined by WCAES. Considering its commercial availability, the best choice is a 1:9 volume ratio H 2 /Ar mixture. The use of solvent gases with lower thermal conductivity such as Xe ( W m -1 K -1 ) and Kr ( W m -1 K -1 ) [22] could improve the WCAES analytical signal. However, the cost of replacing Ar by such expensive gases would make it prohibitive for routine methods. 219

234 ACKNOWLEDGEMENTS This material is based upon work supported by the National Science Foundation and the Department of Homeland Security through the joint Academic Research Initiatives program: CBET and Grant Award Number 2008-DN-077-ARI The authors would like to thank Mr. Robert Morris from the Department of Physics at Wake Forest University for all his help constructing the metallic cells used in this work. 220

235 REFERENCES [1] J. D. Ingle, Jr., S. R. Crouch, Spectrochemical Analysis, 1st ed., Prentice-Hall, Englewood Cliffs, 1988, 590p. [2] C. L. Sanford, S. E. Thomas, B. T. Jones, Appl. Spectrosc. 50 (1996) [3] J. D. Batchelor, S. E. Thomas, B. T. Jones, Appl. Spectrosc. 52 (1998) [4] X. Hou, B. T. Jones, Microchem. J. 66 (2000) [5] K. Song, Y. I. Lee, J. Sneddon, Appl. Spectrosc. Rev. 37 (2002) [6] J. A. Rust, G. L. Donati, M. T. Afonso, J. A. Nóbrega, B. T. Jones, Spectrochim. Acta B 64 (2009) [7] J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Spectrochim. Acta B 60 (2005) [8] J. A. Rust, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, Spectrochim. Acta B 61 (2006) [9] G. L. Donati, J. Gu, J. A. Nóbrega, C. P. Calloway, Jr., B. T. Jones, J. Anal. At. Spectrom. 23 (2008) [10] G. L. Donati, B. E. Kron, B. T. Jones, Spectrochim. Acta B 64 (2009) [11] G. L. Donati, C. P. Calloway, Jr., B. T. Jones, J. Anal. At. Spectrom. 24 (2009) [12] P. V. Oliveira, F. J. Krug, M. M. Silva, J. A. Nóbrega, Z. F. Queiroz, F. R. P. Rocha, J. Braz. Chem. Soc. 11 (2000)

236 [13] W. Slavin, D. C. Manning, G. R. Carnrick, At. Spectrosc. 2 (1981) [14] J.A. Nóbrega, J. A. Rust, C. P. Calloway, Jr., B. T. Jones, Spectrochim. Acta B 59 (2004) [15] G.L. Donati, K. E. Pharr, C. P. Calloway, Jr., J. A. Nóbrega, B. T. Jones, Talanta. 76 (2008) [16] P. Wu, Y. Zhang, R. Liu, Y. Lv, X. Hou, Talanta. 77 (2009) [17] M. Suzuki, K. Ohta, Anal. Chem. 57 (1985) [18] B. Welz, H. Becker-Ross, S. Florek, U. Heitmann, M. G. R. Vale, J. Braz. Chem. Soc. 14 (2003) [19] K.A. Wagner, K. E. Levine, B. T. Jones, Spectrochim. Acta B. 53 (1998) [20] A. Salido, B. T. Jones, Talanta. 50 (1999) [21] M. Suzuki, K. Ohta, Prog. Analyt. Atom. Spectrosc. 6 (1983) [22] R. C. Weast, M. J. Astle, W. H. Beyer, CRC Handbook of Chemistry and Physics, 68th ed., CRC, Boca Raton, 1988, pp. E11-E13. [23] C.G. Bruhn, J. Y. Neira, G. D. Valenzuela, J. A. Nóbrega, Talanta. 48 (1999) [24] X. Hou, B. T. Jones, Recent Res. Develop. Appl. Spectrosc. 3 (2000) [25] NIST. Basic atomic spectroscopic data. [cited 06/03/2009]; Available from: [26] M. Suzuki, K. Ohta, Talanta 28 (1981)

237 FIGURES AND TABLES CAPTIONS FIGURES Figure 1. Schematic diagram of WCAAS and WCAES instrumentation. A- System arrangement overview. B- WCAAS and WCAES atomization cell caps. C- Atomizer projection on the spectrograph entrance slit for WCAES. Figure 2. WCAAS Cd absorption spectrum: 2 µg L -1 solution monitored at nm, using the nm line for background correction. Figure 3. WCAES atomization cell. A- Cell base showing the atomizer at 0 A. B- Cell with the cap on and the atomizer at 3.5 A. Figure 4. WCAES multielement simultaneous determination. Concentrations: Cr and Eu 200 µg L -1 ; Sr 20 µg L -1. Calcium is always present as concomitant in the blank. Figure 5. Effect of the protecting gas mixture on WCAES analytical signal. A- Effect of H 2 volume percentage in Ar. B- Comparison between Ar, N 2 and He mixtures at the best conditions for each element. Concentrations: Cr and Eu 200 µg L -1 ; Sr 20 µg L -1. Calcium is always present as concomitant in the blank. 223

238 TABLES Table 1. Tungsten coil heating cycle Table 2. WCAAS and WCAES analytical figures of merit Table 3. WCAAS and WCAES accuracy 224

239 Figure 1 A Lens Lens Radiation source C C D Protecting gas i Protecting gas i B Sample introduction WCAAS cap Constant current l Sample introduction D/A Converter WCAES cap C Spectrograph slit Coil image Slit aperture 225

240 Figure nm (BG) nm (Analytical line) Absorbance BG corrected signal Time (s) 226

241 Figure 3 A B 227