摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 935 Operation Principles and Design Considerations for Radiofrequency Powered Glow Discharge Devices A Review R. Kenneth Marcus Department of Chemistry Howard L. Hunter Chemical Laboratories Clemson University Clemson SC 29634- 1905 USA One of the most promising of new spectrochemical devices is the r.f. powered glow discharge (GD). In addition to being a self-contained solids atomization/excitation/ionization source the device permits the direct analysis of non-conducting sample types such as glasses and ceramics. While only recently emerging on the atomic spectroscopy scene the r.f. GD has been known in the physics and semiconductor fabrication communities for over three decades. This review is intended to acquaint the reader with the underlying principles which permit the generation of an r.f.plasma at a non-conducting surface some of the plasma characteristics of r.f. GDs and the pertinent source design criteria for application in analytical atomic spectrometry. Keywords Glow discharge; radiofrequency; glow discharge atomic emission spectrometry; glow discharge mass spectrometry; review One of the biggest challenges remaining in the area of analytical atomic spectrometry is the development of more universal methods for the direct analysis of solid materials. While the success of flame and furnace atomic absorption spectrometry and inductively coupled plasma optical and mass spectrometry is undeniable there still remains great interest and effort directed toward improving solid sample dissolution techniques.It is the dissolution step that often limits the ultimate analytical utility of such methods. While dissolution procedures present a homogeneous easily manipulated (flow injection chromatography etc.) sample to the spectrochemical source they are limited by time constraints or loss of analytical quality owing to sample dilution or contamination. Regardless of the possible disadvantages aqueous solution nebulization continues to be the predominant method of sample introduction. Arc and spark sources have historically provided an aIternative to solution-based techniques for solids sampl- ing. Solid material analyses have evolved such that they now require greater powers of detection speed and preci- sion than afforded by standard arc and spark technologies.In addition spatially (depth) resolved elemental profiles are required to assess such systems as galvanized coatings or multi-layered automotive glass. It is for these reasons that glow discharge (GD) devices are receiving increased inter- est within the analytical community. These reduced pres- sure inert atmosphere plasmas rely on a cathodic sputter- ing step to atomize solid electrically conductive samples directly. Subsequent atomic excitation and ionization pro- cesses in the adjacent plasma (negative glow) render the devices useful for atomic absorption (AA) fluorescence (AF) emission (AE) and mass spectrometry (MS).1-2 The requirement that the sample matrix be electrically conductive has limited the application of GD devices to the analysis of metals alloys and metallic thin film systems.Unfortunately an appreciable fraction of all of the solid samples requiring elemental analyses are oxides (e.g. glasses ceramics and refractories) and are not directly amenable to GD analyses. For such samples mixing and compacting the material in a conductive host matrix (Cu Ag Ta graphite) can be a viable ~ p t i o n . ~ - ~ However this process is often undesirable owing to time constraints and target inhomogeneity in addition to sample dilution and possible contamination. These are the same obstacles faced in arc and spark discharge source analysis of oxide materials. Thus this author refers to the problem at hand as the ‘non-conductive barrier’ depicted in Fig. 1 inhibiting analyses by electrical discharges.These oxide (or possibly nitride carbide etc.) materials present yet another analyti- cal adversity as they are some of the most difficult matrices to place into aqueous solution. Given the trends and demands of new materials development it is obvious that methods of direct analysis of such materials are in growing demand. Described here are fundamental aspects of a family of GD devices which have shown the ability to overcome the ‘non-conductive barrier’. Although ‘re-introduced’ to the analytical community in the late 1980s,’ the concept of r.f. powering of GD devices has been known for over 30 year^.^^^ In fact this technology has been chosen almost exclusively over normal d.c. GDs for semiconductor device fabrication for the last 20 years.loJ1 While the most attractive feature of r.f.powering (at least to the analytical chemist) is the capability to sputter atomize non-conduc- tive materials directly the devices have a number of other attributes which suggest their general use for solids analysis. It is the intention of this review to familiarize the reader with the fundamental processes occurring in r.f. powered GDs to describe some of the characteristics which make them promising spectrochemical sources and to familiarize the reader with the wealth of literature (principally physics and engineering) available which describes fundamental studies of these devices. The operating characteristics of the - I Fig. 1 Non-conductive barrier to solids analysis by charged particle methods936 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 r.f. GD devices are illustrated as they have been imple- mented for direct solids elemental analysis in this labora- tory and in collaborative projects with other workers. Radiofrequency GD devices would seem to solve a very important set of problems facing the analytical chemist. It is therefore beneficial to become as familiar with the devices as possible taking information accumulated over the last 30 years and using it to achieve the goals facing the discipline. Historical Background The evolution of the use of high frequency plasmas in sputtering applications has an interesting history. Basically current fundamental knowledge comes from the seemingly unrelated fields of plasma and solid-state physics.The former dealing with the study of deleterious reactions in confined plasmasldischarges in the 1930s and the latter looking for means of determining sputtering characteristics of non-conductive materials in the 1960s. The end result is a clear example of serendipity in one area of science being fundamental to another. The etching of glass materials in intimate contact with gaseous plasmas has been known phenomenologically since the 1930s. In terms of substantiating the process Robertson and Clapp12 were perhaps the first to record the observation that material was removed from the walls of a glass tube when a high frequency discharge was initiated between two external electrodes. Discharges struck in Pyrex tubes having a thin silver film inner coating (filled at mTorr pressures of air) showed substantial removal of the metallic coating in the area immediately adjacent to the electrodes.Not only was the silver removed but the glass matrix itself was ‘altered’ even in the case where no metallic film was initially present. Following these observations Hay13 performed experiments which led to the conclusion that the removal of material was probably due to positive ion bombardment (i.e. sputtering). These studies indicated that (i) electrode removal decreased with increasing operating frequencies (above 3.0 MHz) (ii) removal was inversely related to operating pressure at a fixed frequency and (iii) the effect of pressure was less at high discharge frequencies. An impor- tant note for the discussion here is the fact that neither d.c.nor 60 Hz discharges caused removal of the silver film. In 1948 Lodge and StewartI4 compared high-frequency dis- charges formed by metal electrodes external to the glass vacuum envelope and d.c. glow discharges where the electrodes were located within the low pressure (0.05- 1.5 mm Hg) cell. The two types of discharges were visually comparable each exhibiting a Crooke’s dark space nega- tive glow and positive column. The exposed electrode surfaces were also similar suggesting that the high-fre- quency plasmas produced a high negative d.c. wall poten- tial which results in cathodic sputtering. Probably the single most important set of experiments related to high-frequency plasma etching of glass containers was described by Butler in 1961.15 Research in that laboratory was directed at studying the fundamental pro- cesses underlying the confinement of plasmas by superim- posed microwave radiation. Previous theories suggested magnetic or r.f.field confinement mechanisms. The studies described involved a filament arc generated in a glass enclosure on which a.c. fields were superimposed by externally located electrodes. A metallic ring electrode mounted within the discharge cell allowed measurement of induced wall potentials while a Langmuir probe apparatus permitted measurement of plasma ion and electron charac- teristics. Vital to the application of r.f. discharges in sputtering applications was the verification of a negative biasing of applied r.f. waveform which produced a net d.c. bias potential on the inner wall.Distortions of the temporal behaviour led to the use of square wave potentials which revealed the sequential bombardment of the walls by positive ions and electrons. The results of this alternative bombardment are discussed in detail in the following section. Mathematical models of the potential fields accu- rately depicted the voltage responses at the target surface and confirmed that confinement of the plasma was due to the development of a d.c. positive ion sheath. During the 195Os the measurement of sputtering yields was evolving for applications in the nuclear and semicon- ductor material areas in the field of solid-state physics. While conductive materials (metals and semiconductors) are readily sputtered by noble gas positive ions yield determinations on non-conductive elements and com- pounds is hindered by inability to maintain high ion energies owing to the accumulation of positive charges on the target surface.Simply there is no means of neutralizing the charge on the surface so incoming ions experience a repelling positive charge. Thus some means of charge compensation is required. (This problem continues to cause difficulties in the secondary ion mass spectrometry (SIMS) technique.16) In 1955 Wehner” proposed a rationale for sputtering non-conducting materials which was later de- monstrated18 by employing a conductive backing electrode to a dielectric target and applying a suitable high frequency potential. (In this case the high-frequency electrode was not the discharge sustaining electrode but was an auxiliary electrode within a d.c.-sustained two electrode system.) These workers reasoned that the d.c.biasing concept illustrated by ButlerIs would allow for the efficient charge compensation and near continuous bombardment of the target by positive ions. It was proposed that due to the high sputering rates the method could be used for relative yield measurements etch studies and sputtering of insulating layers. Cleaning of observation windows in etching systems without venting to the atmosphere was seen as another key application. Indeed in subsequent studies by Davidse and Maissel,19 the technique of depositing insulating thin films was developed into a practical method involving a rela- tively simple two electrode geometry (grounded diode) wherein the discharge was sustained from the target electrode in a similar manner as a d.c.-powered GD.Theory of Operation In a text dedicated to a thorough coverage of GD pheno- mena Chapman8 described the basic theory of GD oper- ation. In the discussion here the focus is on those aspects most pertinent to the use of the r.f.-powered devices in analytical spectroscopy. The result of a d.c. potential applied to a sample (electrode) made of a non-conducting material is considered first. Obviously the current neces- sary to sustain a discharge cannot flow and the resulting response of the insulator surface potential to the applied potential is analogous to the discharging of a capacitor as demonstrated in Fig. 2. When a high negative voltage (- V,) is applied to the insulator the surface potential drops initially to - V,.If positively charged species (ions pro- duced by breakdown of the plasma gas) are in the vicinity of the target they are accelerated to the surface. At this point the potential on the target increases with time to more positive potentials due to ion neutralization reactions at the cathode surface. The characteristic discharge time of this capacitor (sample) is of the order of 1 p s in reduced pressure environments. The short-lived discharge will exist until a minimum voltage threshold value is reached and the plasma is extinguished. Positive charge accumulation also impedes particle bombardment. In order to have continued ion bombardment of the surface and thus cathodic sputter- ing a source of electrons must be available to neutralize positive ions while maintaining a negative acceleratingJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY. OCTOBER 1993 VOL.8 937 - Time - Fig. 2 Capacitor-like response of an insulating surface to an applied high voltage potential. This approach has been practised directly in the field of SIMS wherein electron or 02- guns are used to flood the target surface with negatively charged species.i6 Wehner” correctly proposed that the application of a high- frequency potential at a rate that would overcome the decay rate of the ‘capacitor’ would allow plasma electrons to neutralize the positive charges that accumulated on the surface of the insulator during the positive portion of each voltage cycle. The most analytically important result of the application of a high-frequency a.c.potential is the self-biasing of the electrodes. It is this self-biasing that sustains the discharge and induces the sputtering process in the plasma. Consider the 2 kV peak-to-peak square wave potential (V,) and the resultant potential on the cathode surface (V,) shown in Fig. 3. As the potental is applied during the first half-cycle the surface becomes charged to - 1 kV followed by a decay to approximately -0.7 kV. As the second half-cycle begins the $2 kV potential reversal results in a + 1.3 kV potential on the surface. During this half-cycle electrons are acceler- ated to the insulator surface and discharging results analogous to that of the previous half-cycle. However owing to the greater mobility of the electrons relative to the positive ions the surface potential decays toward zero at a faster rate than the previous half-cycle thus reaching a value of +0.6 kV.It is important to note that the net current in this system is zero as an equal number of positive and negative charged species must pass in each respective half-cycle. As the second full cycle is initiated and the polarity of the electrode is switched the resulting +1 > 3 -1 +1 > 5 -1 r 2r 3r I Pb’ )i Time - Fig. 3 Establishment of a d.c. bias potential at an insulating surface under the influence of a high-frequency square wave potential (a) and the resultant potential on the cathode surface (b) potential is - 1.4 kV (+0.6-2 kV). After some number of cycles the waveform of Vb reaches a negative d.c. offset which is the self-bias that sustains the primary ion current.(ButleP obtained oscilloscope traces of this form in his plasma confinement studies. The d.c. bias potential is generally one-half the applied peak-to-peak voltage and is a function of the general electrode design and discharge operating parameters. The insulating target material is alternately bombarded by high energy ions and low energy electrons; however the dis- charge is for all intents and purposes continuous with respect to most analytical measurement systems. Most importantly the sample serves as a time-averaged cathode and is subject to the cathodic sputtering characteristic of GD devices. Throughout the discussions here it is impor- tant to keep the capacitive nature of the electrode processes (due to the differing mobilities of electrons and positively charged ions) in mind as this defines the d.c.bias which in turn ultimately has a pronounced effect on the atomization excitation and ionization characteristics. The described capacitive response of an insulating target will of course not be present for an electrically conductive cathode. In fact the bias potential in a non-capacitive (conductive electrode) system is zero. As with atmospheric pressure r.f. plasmas the r.f. potential in these devices is typically coupled to the insulating material via a matching (L-C) network which serves to maximize the power to the plasma by balancing the system impedance of the r.f. generator to that of the discharge. (The importance of power coupling is discussed later.) As such the capacitance of this circuitry provides the capacitive response for conductive targets.Fig. 4(a) illustrates the evolution of a d.c. bias at a copper target sputtered in an r.f. GD atomic emission device in this laboratory. Note that the time frame of the establishment of the bias potential is of the order of 250 ,us for this particular r.f. generator/matching box system. The frequency of this waveform reflects the biasing imparted due to the extended sampling time required to visualize this waveform. Based on the time scale approxi- mately 3400 complete r.f. cycles are required to establish the bias. Fig. 4(6) illustrates the evolution of the actual r.f. powering potential. This waveform was obtained by sampl- ing the back of a non-conductive sample and thus reflects the load imposed by the sample and capacitance of the matching circuitry.As seen here approximately 200 ps are required for the actual r.f. power to reach its targeted level. Thus the actual biasing phenomenon probably occurs over a much smaller number of cycles than observed in Fig. 4(a). Also shown in Fig. 4(c) is the corresponding surface potential for a Macor (Corning Glass Works Corning NY USA) machinable glass ceramic sample. This waveform was acquired by putting a thin metal wire against the sample surface exposed to the sputtering action and might not accurately reflect the actual surface potential but does serve to illustrate the biasing process. As can be seen both the r.f. peak-to-peak and d.c. bias potentials appear to be smaller for the case of a non-conductive target.This is not unexpected as power losses through non-conductive ma- trices are inevitable. Of note is the fact that the biasing of the sample surface for both the conductive and non- conducting samples seem to occur on the same time scale. Before discussing the design considerations of r.f. GD sources it will be instructive here to bring together the capacitive components of the plasmas.8.20 The electrical characteristics of an r.f. GD can be demonstrated with an equivalent circuit of the type shown in Fig. 5(a). As can be seen there are two primarily capacitive components of the circuit. The first bit of capacitance which the input voltage experiences is that of the matching network which will vary in order to match the overall impedance of the load to the 50 Q output of the generator.In the case of non-conductive938 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 0 -250 800 I 1 600 400 $ 2oo 3 -200 0 2 0 > -400 -600 -800 I I I I I 0 -50 -100 I I I I I J 0 0.1 0.2 0.3 0.4 0.5 Time/ms Fig. 4 Oscilloscope traces (a) the establishment of a d.c. bias potential at a copper target; (b) the applied r.f. power to the back of a non-conductive target; and (c) the subsequent evolution of the d.c. bias potential on the surface of a 4 mm thick Macor sample samples the capacitance of the target can be appreciable being a function of the material and its thickness. One could also assign a capacitive component to the discharge walls as they become coated with sputtered material but we consider here that the chamber walls are clean metal electrodes.There are more or less three impedance compo- nents to the system that of the cathode dark space (ZcdS) the plasma negative glow (Z,J and the anode dark space (Zw). Electrical equivalents of the impedance components for the target and wall sheaths are shown in Fig. 5(b). The diodes correspond to electron motion toward the respective target surfaces. As detailed previously the capacitive nature of the sheaths above the target (and walls) is induced by the I- ( b ) Matching network @I Samplebarget & 171 Chamber Fig. 5 Electronic equivalent circuit for (a) an r.f. glow discharge plasma and (6) the impedance component of the wallltarget sheathsto relative mobilities of positive ions and electrons.Finally the resistor represents the ion current through the dark spaces. The impedance of the negative glow is controlled by the density and mobility of electrons within the negative glow. Thus those conditions (such as increasing pressure) where electron densities increase effect an increase in plasma impedance. In most instances of plasma modelling or source design the negative glow impedance is assumed to be negligible or at least a constant (resistance). Atomization/Excitation/Ionization Characteristics The operating characteristics of any GD source that is current voltage and pressure ultimately define the atomi- zationlexcitationlionization characteristics. Briefly dis- charge gas ions must be accelerated to the cathode surface and arrive with sufficient energy so that the collision cascade can produce a sputtering event (i.e.atomization). The ion kinetic energy is thus affected by the discharge voltage (d.c. bias) and the source pressure which affects the ion and atom mean free paths. The majority of discharge electrons are in fact created as secondary products in the sputtering process. These electrons are accelerated from the cathode by the fall potential to the negative glow region. Within the negative glow these electrons cause excitation of sputtered and discharge gas species as well as performingJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 939 Table 1 Comparison of analytically pertinent plasma species' Parameter r.f. d.c. Sputtered atom density (copper)/atoms Electron den~ity/cm-~ Average electron energy/eV Metastable atom den~itylcm-~ Excitation temperature/K - 3000-6000 (refs.30 and 33) 101o-lO1l (ref. 26) l Or5- 1 016 (ref. 27) 4-6 (ref. 27) 10'o-lO'l (ref. 26) 1010-1013 (refs. 27 and 28) 1014 (refs. 28 and 30) 2-5 (refs. 31 and 32) 10'o-10'2 (refs. 28 and 29) direct ionization of these species. Long-lived (metastable) excited state discharge gas atoms can in turn affect ionization of sputtered species via a Penning-type process. Therefore the extent of atomic excitation and ionization in this region is also dictated by the discharge power. As the application of r.f.-powered GDs evolves fundamental studies relating changes in operating parameters to the density and energy of sputtered atoms electrons and ions and metastable atoms will become more important to developing better insights into analytical source design and operat ion.As the use of r.f. GD devices in this laboratory was just beginning a comparison of analytically relevant particle densities and energies found in the literature was per- formed.' While the vast majority of spectroscopic studies of r.f. GD devices have been pursued as a means of process monitoring a large number of more fundamental studies have been undertaken to gain a more thorough understand- ing of collisional processes occurring in the plasma negative glow.2* Gottscho and Miller2* have reviewed the optical techniques that have been applied for r.f. plasma diagnos- tics. These techniques include a~tinometry,~~ laser induced fluore~cence~~ and optogalvanic spectros~opy.~~ Charged species in these plasmas have been sampled mass spectro- metrically26 and by use of electrostatic probe^.*^*^^ The densities of the analytically relevant species in r.f.and d.c. GD devices are compared in Table 1. It should be kept in mind that comparisons among discharges of different geometries are difficult to make so that these figures should not be taken as absolute. As can be seen the values given for the two plasma types are in fairly close agreement. Therefore one would expect that the analytical characteristics of r.f. and d.c. GD devices should not be very different. Preliminary data indicate that this is indeed the case. A comparison of sputtering rates of metals for the r.f. atomic emission sources developed in this laboratory and conventional Grimm-type devices shows no real differences (0.1 - 1.5 mg min-I).The key difference is of course the fact that the r.f. powered devices are capable of direct atomization of insulating materials. Another advan- tage which has yet to be investigated may result from the lower operating pressures of the r.f. devices producing a higher degree of depth resolution for thin film analyses.34 In practice there exist a number of substantial differ- ences in the plasma energetics between d.c. and r.f. powered discharges. Comparisons of the two operating modes have been made with the same source (described in ref. 7) simply by changing the power supplies. (In principle any r.f.-designed discharge can operate in the d.c. mode but the converse is not necessarily true.) Direct comparisons are shown in Table 2.35 The ion and electron densities plasma potentials and electron energies were obtained by Langmuir probe measurement^.^^ The general trends between the two plasma types is that the r.f.plasmas are composed of much higher energy electrons while the ion and electron number densities are greater for the d.c. case. Based on a simple Boltzmann analysis these sorts of data suggest that the r.f. plasma probably has more electrons of suitable energy for atomic excitation than the d.c. device and that high-lying excited states would be more extensively populated in the former. A number of mechanisms have been proposed to explain the enhancement in electron energies in r.f. GDs,~ but the effect is most easily understood by picturing the electrons quickly oscillating as the potential is varied at 13.56 MHz.Additional energy is obtained by electrons when electron-atom collisions give them a directionality in phase with the instantaneous field change. Electrons in a d.c. plasma feel no such polarity changes in the effectively field-free negative glow. Source Design Considerations To a first approximation then we must be concerned with the design of sources that optimize the power delivered to the system so as to achieve the maximum atomization/exci- tation/ionization capabilities for the applications at hand. Considered here are the design parameters of source geometry r.f. power coupling schemes and operating frequency as they apply to analytical r.f. powered GD devices. Two source designs developed in this laboratory for atomic emission and mass spectrometry applications are used to demonstrate the design aspects.Other source designs could be developed based on the spectroscopic mode being applied (e.g. AA) and the relevant sampling requirements (size shape vacuum etc.) specific to the instrumentation at hand. A summary of r.f. GD source design criteria for plasma etching systems presented by Kumagi3' is an excellent resource in this area. Source Geometry The geometry of the discharge is critical in terms of isolating the bombarding species to the sample cathode and determining the energy of the bombarding ions as related to the d.c. bias. The first criterion is to maximize the ratio of the respective surface areas of the anode to the cathode.Only in highly asymmetric r.f. discharges is the d.c. bias located at a particular electrode. Specifically ion bombard- ment is isolated to the smaller electrode (termed by analogy Table 2 Comparison of charged particle characteristics in the same source3s Parameter r. f. d.c. Electron densi t y / ~ m - ~ Average electron energy/eV 4-7 0.7-1.0 Excitation temperature/K 5 000- 8 000 2500-4000 Electron temperaturelev 1.5-2.5 0.2-0.6 2 X 10"-6 x 10" 6~ 10'0-18X 10" Ion number density/cm-' 3x 10'0-12x 10'0 4 x 10'0-2ox 10'0 Plasma pot en t ial/eV 9-16 2-4940 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 Coupling Vln 1 - Asymmetric small electrode powered) -L - Symmetric Asym-metric (large electrode powered) d.c. Coupled I Capacitively coupled 1 Fig.6 Depictions of relative plasma potentials V,(t) (solid curves) excitation electrode voltages V(t) (broken curves) and d.c. bias voltages (Vdc) as a function of discharge coupling and geomet ry38 to d.c. sources as the cathode) with the negative bias approaching one-half of the r.f. peak-to-peak voltage in the ideal case. The actual distribution of electrode potential as it relates to both discharge symmetry and power coupling has been presented by Kohler et aL3* As shown in Fig. 6 six basic cases of symmetry and coupling can be described demon- strating the effect of geometry and mode of r.f. coupling (either directly or capacitively coupled) on both the d.c. bias voltage Vdc (broken h e ) and the plasma potential Vp (t) (solid line) (assuming a sinusoidal waveform).As illus- trated increasing the anode-to-cathode ratio increases the d.c. bias voltage and decreases the plasma potential. This distribution of voltage is due to differences in the capaci- tances of the dark space sheaths of the two electrodes. At frequencies above a few MHz the sheaths can be consi- dered to behave capacitively because the r.f. frequency is below the electron resonant frequency and above that of the ion resonant freq~ency.~~ That is to say that an electron can cross the respective sheaths in less than one r.f. cycle while a more massive ion experiences a number of r.f. oscilla- tions. An expression predicting the effect of sheath (electrode) capacitance in the development of a d.c. bias has also been presented by Kohler et In this model the d.c.bias ( v d c ) is related to the applied r.f. potential (V,) and the respective target and wall capacitances (C and Cw) by the relationship It is important to note that the respective capacitances must be based on the surface area of the electrode in contact with the plasma not necessarily their physical sizes. In addition this potential is the d.c. bias developed with respect to the counter electrode (which is typically grounded). A model presented by Koenig and Maisselzo also works on the premise of relative capacitance in describing the respective biasing. These workers developed the rela- tionship Vl/ V2 = (A2/Al)4 where V,/V2 is the ratio of the d.c. potential difference between the glow space and the electrodes (ie. the sheaths) and Az/Al is the ratio of the electrode surface areas.More recent studies by Coburn and Kay4’ show that this relationship can in some instances follow more closely the first power than the fourth. The observed differences would most likely seem to come from the estimation of the ‘active’ electrode areas. In any case optimization of d.c. bias potentials in analytical applications is obtained in those instances where the anode-to-cathode surface area ratio is as large as practically obtainable. Such optimization max- imizes the power delivered for sample atomization while minimizing the possibility of ablating the cell walls. Beyond optimizing the respective electrode areas it is an additional consideration to optimize the power density delivered to the sputtering target. The direct relationship between power density and discharge voltage is well known in conventional d.c.GD devices.4z Additionally discharge voltage is inversely related to source operating pressure for a constant current. These basic characteristics have been substantiated also with the ‘external’ sample mount geo- metry shown in Fig. 7 which was developed in this laboratory for r.f. GD atomic emission analysis.43 Analo- gous to the anode tube in a Grimm-type GD lamp the limiting orifice in this device acts as a shutter to restrict sputtering to the sample surface defined by the orifice diameter. Shown in Fig. 8(a) is the effect of r.f. power on measured d.c. bias potential for a range of orifice diameter^.^^ As with conventional sources higher power densities (smaller sput- tering areas) produce higher bias voltages. Results such as these suggest that apart from a sufficient anode-to-cathode ratio which isolates sputtering to the target the r.f.d f b- Female coaxial connector Copper conductor Male coaxial connector Glass insulator RG-213/U Coaxial cable (to matching network) 1 in H Fig. 7 External sample mount geometry developed for r.f. glow discharge atomic emission spectrometry4’JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 94 1 100 I I I I I I I 0 10 20 30 40 50 60 70 PowerMl Fig. 8 Dependence of d.c. self-bias potential on (a) r.f. generator output power for various orifice sizes A 2; B 4; C 6; D 8; E 10; and F 12 mm (source pressure=800 Pa) and (6) r.f. generator power and discharge pressure ( 1 Ton= 133 Pa) with the 6 mm diameter orifice disk44 discharge behaves very much like a d.c.discharge. The analogy between the d.c. and r.f. discharges is illustrated further in Fig. 8(b) wherein the effect of operating pressure on the d.c. bias (constant power) is comparable to that of d.c. sources. A grounded shutter located in close proximity to the cathode surface as shown in Fig. 7 is necessary to confine the sputtering to the sample target and is an integral component in all r.f. plasma deposition systems. The placement of the shutter is critical to prevent the plasma fall (dark space) from spreading to undesirable locations within the chamber and those components being sputtered into the plasma. To isolate the sputtering to the target the shutter should be located less than one dark space from the sample surface.Dark space thicknesses are inversely proportional to cell pressure and to a lesser extent the bias potential. Typical values range from 5 mm at pressures of 13 Pa to 0.5 mm at pressures towards 1.33 kPa. Use of insulating spacers (e.g. ceramic or Teflon) common in d.c. GD designs is tenuous as exposed insulator surfaces tend to become charged when exposed to the ref. potentials. Attempts in this laboratory to operate with the so-called floating anode geometry used in Grimm-type devices led to very erratic operation arcing and sputteringlvaporization of the shutter. Therefore it is necessary to operate the r.f. GD systems with only the hard-grounded anodehhutter in proximity to the sample surface with no floating or insulating surfaces in direct contact with the plasma volume. The basic design criteria involving isolation of the d.c.bias to the cathode surface have also been followed in the design of a direct insertion probe (DIP) for MS.45 The need to interface to the low pressure mass spectrometer without having to break the system vacuum necessitates the use of a vacuum interlock system to present the sample to the source which is housed under vacuum at all times. The tendency of GDMS source operation to be severely hindered by residual gases means that the cell itself cannot be directly exposed to the atmosphere. The use of DIPS in conjunction with vacuum interlocks is common practice in organic MS. The DIP shown in Fig. 9 is made of a 12.5 mm diameter stainless-steel body through which the power is passed to the r.f.high voltage feedthrough mounted at the end. A number of different sample holders can be readily mounted to the feed- through allowing for sample sizes up to 10 mm in diameter. Integral to the design here is the presence of a grounded steel cap which encircles the sample holder and acts as a shutter around the sample. The impact of the active electrode areas is now becoming a relevant point in analytical source design. For example the nature of the discharge support gas can effect plasma volume. R ~ t h ~ ~ has observed spatially resolved emission profiles between the electrodes of a closely spaced parallel plate (one powered and the other grounded) r.f. discharge. Profiles of non-reactive gases indicate that the symmetry of the discharge can be dramatically affected by the nature of the gas without physical changes in the geometry of the discharge chamber.Basically the volume of the plasma is affected by the thermal properties of the gas the pressure and the charged particle densities. In molecular gases the electron temperature is lower than that of atomic gases owing to vibrational and rotational energy losses not available to the latter. Thus diffusion is increased in the atomic discharges resulting in increased asymmetry of the discharge. As the plasma diffuses from the grounded plate into the chamber which is also at ground the discharge becomes highly asymmetric with respect to the powered electrode. Mass spectrometric studies in this laboratory also indi- cate the importance of the active cell volume.Early r.f. GDMS work employing the DIP device with a large volume six-way cross indicated that optimum ion signal was achieved when operating at low source pressures (13-40 Pa) with ion sampling at the negative glow-dark space interfa~e.~ Alternatively operation within a greatly reduced ion volume of a commercial instrument (VG 9000 and GloQuad) yields optimum ion intensities at pressures of the order of 250-400 Pa.47 In this situation ions are sampled from the bulk negative glow region. Such differ- ences most likely result from the ability of the plasma to fill / Stainless-steel tubing (1.27 cm o.d. Electrical Macor Steel id.) feedthrough sleeve sleeve \ \ \ Glass tubing Vacuum weld Sample Sample (1 .O cm o.d. holder 0.8 cm i.d.) w Fig.9 Direct insertion probe for r.f. powered GDMS4s942 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 the respective cell such that the ‘ionization volume’ is located adjacent to the ion extraction orifice. Addition- ally the relative anode sizes would be expected to affect the sheath characteristics at the grounded exit orifice and thus change ion extraction efficiencies. The analytical ramifications of these two pressure regimes are clearly evident in the mass spectra obtained. In the lower pres- sure case molecular species such as metal dimers (M,+) argides (MAP) and discharge gas dimers (Ar,+) are far more prevalent than in the higher pressure cell. While difficult to compare the absolute beam intensities for atomic analyte ions seem to be comparable.Therefore the latter cell design would be considered preferable because of the reduced likelihood of isobaric overlaps. Elucidation of the relationships of this sort between source geometry and analytical operating parameters will give r.f. GD practitioners more insight and ability in designing analytical devices. Radiofrequency Power Coupling The ability to efficiently couple the r.f. energy to the cathode and plasma is a integral part of optimum r.f. GD source operation. More specifically the electric coupling of the r.f. power to the target electrode such that a bias is generated at its surface is critical. Referring again to Fig. 6 the importance of capacitive coupling can be seen. Without a capacitive coupling either through a non-conductive sample or the external circuitry (i.e.the matching network) the d.c. potential difference between the electrodes cannot be maintained and the d.c. bias potential would be divided equally between the electrodes. Note that the bias potential is zero for every geometry where the r.f. power is d.c.- coupled. Therefore the potentials at the powered and grounded electrodes will be such that they alternate be- tween being the powered and counter electrode maintain- ing an a.c. plasma which is very inefficient at sputter atomization due to low ion energies and suffering from introduction of contaminant species from the chamber walls. The second consideration in the design of an analytical r.f. GD is the careful shielding of the powering cable. Plasma generators (e.g.ICP) are well known for their proficiency at broadcasting r.f. interference (rfi) when not properly grounded and shielded. In order to minimize if not eliminate rfi source designs should incorporate coaxial shielding extending from the power cable and sample holder to the sample surface. In both of the basic sources designed in this laboratory the external sample mount and DIP used in MS coaxial shielding to the sample surface is complete. Taking such precautions results in the addition of no detector noise beyond that inherent in the rest of the electronic background. An additional feature of coaxial shielding is that it protects against electrical shock. It must be kept in mind that r.f. potentials exhibit a creeping ‘skin effect’ and tend to leak through/over what would be considered infinite resistance in d.c.systems. Operating Frequency To date the vast majority of the analytical applications of r.f. powered GDs have been at an operating frequency of 13.56 MHz. Variations in plasma operating frequency beyond the levels of being able to sustain plasmas at insulating electrodes (> ~ 0 . 5 MHz) are expected to affect sputtering ion energies due to the number of oscillations and collisions an ion experiences while it crosses the cathode dark space. The coupling of r.f. energy to electrons in the plasma negative glow will depend likewise on the relative r.f. power and gas-phase collision frequencies. For these reasons the role of operating frequency will be an important field of investigation as the r.f.GD techniques evolve. Variations of plasma operating frequencies have been studied extensively with respect to plasma deposition systems. Donnelly et have studied the effects of plasma operating frequency on N,-Cl plasmas used for chemical etching. Vital to the role that frequency plays in cathodic sputtering is the relationship between the power frequency and the ion plasma frequency (ai) which is given by (3) where n is the ion number density e is the electron charge e is the permittivity of vacuum and rn is the ion mass.39 At frequencies less than mi the sheath has resistive character but is capacitive at higher frequencies. Thus at low frequencies ((0.5 MHz for N,) relatively large potentials are required to sustain the discharge as ions are accelerated across the sheath in less than one r.f.cycle actually n/2. At higher frequencies a time varying field is experienced and the ion energies are lower (X Vln). In terms of obtaining high ion kinetic energies and thus higher sputtering rates lower operating frequencies (in the 50 kHz range) show an increase in electric field strength. This increase in energy is of course countered by the eventual inability to operate a continuous plasma at a non-conducting surface. At high plasma frequencies the plasma power is less centered on the cathode but dissipated more in the negative glow leading to enhanced electron impact dissociation excitation and ionization as measured by emission and fluorescence experiment^.^^?^^ The enhanced molecular dis- sociation observed in the plasmas mentioned above would be an important aspect of r.f.GDMS. Unfortunately these sorts of studies have not been performed on what one would call ‘analytical’ r.f. glow discharges. It can be assumed that given the same fmdamental processes occurring in analyti- cal plasmas there may be different plasma frequencies which are optimum for the various analytical modes of operat ion. A preliminary study of the role of operation frequency in r.f. GDMS was undertaken in collaboration with the Oak Ridge National Laboratory (TN USA).49 The operating frequencies of 40.68 and 13.56 MHz were compared on a VG 9000 double sector spectrometer. A number of interest- ing points were revealed. First the d.c. bias values obtained from 40.68 MHz operation were routinely 20% lower than for 13.56 MHz operation for all of the combinations of power and pressure investigated.The lowering of the d.c. bias for the same powers and pressures is strong evidence of enhanced ionization efficiency in the higher frequency plasma. Secondly the lower bias voltage was evidently compensated for by a higher sputtering ion current V I Fig. 10 Overcoming the non-conductive barrier with r.f. powered GD devicesJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 943 producing a 3-fold enhancement in mass loss for a copper target as compared with the lower frequency case. This in turn translated into analyte ion beam intensities that were a factor of 3-5-fold higher for the 40.68 MHz plasmas. The final area of comparison the extent of molecular ions present in the mass spectrum did not indicate any substantive differences between the two operating frequen- cies.These limited studies definitely suggest much more detailed investigations into the role of excitation frequency in terms of basic analytical figures of merit and also the fundamental plasma processes. Conclusion Radiofrequency powered GD devices provide some unique opportunities to the analytical spectroscopist. The ability to directly sputter-atomize non-conductive materials is in itself a strong selling point. In addition the production of a much more energetic plasma glow yields some advantages over conventional d.c. powered sources. Throughout this review no ‘analytical data’ was presented. This was intentionally done so as to allow the reader to concentrate on the fundamental and practical aspects of the plasmas themselves and not to make continual comparisons with existing techniques.Such comparisons which are vital in the long run are best made in individual research papers or in a review devoted to that particular topic. In fact the ultimate utility of the devices will not be known for a few more years. It is hoped that this review will serve as a useful starting point for those entering into the field as well as fostering further interest and new ideas. During the course of this laboratory’s work in r.f. glow discharge spectrometries financial assistance from the National Science Foundation Jobin-Yvon Division of Instruments SA and VG Elemental along with research collaborations with Oak Ridge National Laboratory and the Westinghouse Savannah River Laboratory have been invaluable.Their sponsorships are gratefully acknowledged. The author acknowledges the helpful comments of Dr Douglas C. Duckworth of ORNL in reviewing this manu- script and the tireless efforts of his other graduate students (Dr Michael R. Winchester Dr Duencheng Fang Paula R. Cable Tina R. Harville Chris Lazik and Charles R. Shick) throughout the studies described here. 1 2 3 4 5 6 7 8 9 10 1 1 12 References Marcus R. K. Spectroscopy 1992 7(5) 12. Harrison W. W. Barshick C. M. Klingler J. A. Ratliff P. H. and Mei Y. Anal. Chem. 1990,62 943A. Dogan L. Laqua K. and Massman H. Spectrochim. Acta Part B 1972 27 65. Caroli S. Alimonti A. Zimmer K. Spectrochim. Acta Part B 1983 38 625. Marcus R.K. and Harrison W. W. Anal. Chem. 1987 59 2369. Winchester M. R. and Marcus R. K. Appl. Spectrosc. 1988 42 941. Duckworth D. C. and Marcus R. K. Anal. Chem. 1989,61 1879. Chapman B. N. Glow Discharge Processes Wiley New York 1980. Coburn J. W. and Harrison W. W. Appl. Spectrosc. Rev. 1981 17 95. Sugano T. Applications of Plasma Processing to VLSI Techno- / o n Wiley-Interscience New York 1985. Morgan R. A. Plasma Etching in Semiconductor Fabrication Elsevier Amsterdam 1985. Robertson J. K. and Clapp C. W. Nature 1933 132 479. 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 Hay R. H. Can. J. Research Sect. A 1938 16 191. Lodge J. I. and Stewart R. W. Can. J. Research Sect. A 1948 26 205. Butler H.S. Microwave Laboratory Report No. 820 Stanford University 196 1. Benninghoven A. Rudenauer F. G. and Wehner H. W. Secondary Ion Mass Spectrometry Basic Concepts Instrumen- tal Aspects Applications and Trends Wiley New York 1987. Wehner G. K. Adv. Electron. Electron Phys. 1955 VII 253. Anderson G. S. Mayer W. N. and Wehner G. K. J. Appl. Phys. 1962 33 2991. Davidse P. D. and Maissel L. I. J. Appl. Phys. 1966,37 574. Koenig H. R. and Maissel L. I. IBM J. Res. Devel. 1970 14 168. Coburn J. W. J. Vac. Sci. Technol. A 1986 4 1830. Gottscho R. A. and Miller T. A. PureAppl. Chem. 1984,56 189. Coburn J. W. and Chen M. J. Appl. Phys. 1980 51 3134. Gottscho R. A. and Mandich M. L. J. Vac. Sci. Technol. A 1985 3 617. Walkup R. E. Dreyfus R. W. and Avouris P. Phys. Rev. Lett.1983 50 1846. Eckstein E. W. Coburn J. W. and Kay E. Int. J. Muss Spectrom. Ion Phys. 1975 17 129. Cox T. I. Deshmukh V. G. I. Hope D. A. O. Hydes A. J. Braithwaite N. St. J. and Benjamin N. M. P. J. Phys. D. 1987 20 820. Marcus R. K. Ph.D. Dissertation University of Virginia Charlottesville VA 1986. Ferreira N. P. Strauss J. A. and Human H. G. C. Spectrochim. Acta Part B 1983 38 899. Brackett J. M. Mitchell J. C. and Vickers T. J. Appl. Spectrosc. 1984 38 136. Borodin V. S. and Kagan Yu. M. Sov. Phys. Tech. Phys. (Engl. Transl.) 1966 11 131. Mehs D. M. and Niemczyk T. M. Appl. Spectrosc. 1978,32 269. Mehs D. M. and Niemczyk T. M. Appl. Spectrosc. 198 1,35 66. Coburn J. W. Eckstein E. W. and Kay E. J. Appl. Phys. 1975,46 2828. Fang D. and Marcus R. K. paper presented at the 17th Annual Meeting of the Federation of Analytical Chemistry and Spectroscopy Societies (FACSS) Cleveland OH USA Octo- ber 7-12 1990. Fang D. and Marcus R. K. Spectrochim. Acta Part B 1990 45 1053. Kumagi H. Y. J. Vac. Sci. Technol. A 1986 4 1800. Kohler K. Coburn J. W. Horne D. E. Kay E. and Keller J. H. J. Appl. Phys. 1985 57 59. Tsui R. T. C. Phys. Rev. 1968 168 107. Kohler K. Horne D. E. and Coburn J. W. J. Appl. Phys. 1985 58 3350. Coburn J. W. and Kay E. J. Appl. Phys. 1972 43 4965. Fang D. and Marcus R. K. Spectrochim. Acta Part B 1988 43 1451. Winchester M. R. Lazik C. M. and Marcus R. K. Spectrochim. Acta Part B 199 1 46 483. Lazik C. M. and Marcus R. K. Spectrochim. Acta Part B 1992,47 1309. Duckworth D. C. and Marcus R. K. J. Anal. At. Spectrom. 1992 7 71 1. Roth R. M. Mat. Res. SOC. Symp. 1987 98 209. Duckworth D. C. Donohue D. L. Smith D. A. Lewis T. A. and Marcus R. K. Anal. Chem. 1993 65 in the press. Donnelly V. M. Flamm D. L. Bruce R. H. J. Appl. Phys. 1985 58 21 35. Duckworth D. C. Smith D. H. and Marcus R. K. paper presented at the 40th ASMS Conference on Mass Spectrometry and Allied Topics Washington DC USA May 31-June 5 1992. Paper 3/01575K Received March 18 I993 Accepted June 22 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800935

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 945 Characterization of a Helium Discharge for Hollow Anode Furnace Atomization Non-thermal Excitation Spectrometry Philip G. Riby* and James M. Harnlyt United States Department of Agriculture Beltsville Human Nutrition Research Center Nutrient Composition Laboratory Building 161 BA RC-East Beltsville Maryland 20705 USA A discharge in He has been characterized for hollow anode furnace atomization non-thermal excitation spectrometry (HA-FANES). The small surface area of the cathode and high current densities make it possible to stabilize He discharges up to 1200 hPa using a d.c. power supply operating in the constant current mode. Current-voltage characteristics were determined for pressures up to 800 hPa. Characterization at higher pressures was not possible owing to limitations of the power supply.The release of thermionic electrons from the cathode surface at temperatures in excess of 1750 K caused a dramatic decrease in the discharge potential during the atomization cycle. Increased ohmic heating of the cathode with increased currents caused a time shift in the release of thermionic electrons. Excitation temperatures obtained from Boltzmann plots of five He I lines decreased slightly as a function of increasing pressure and increased as a function of increasing current. Dramatic decreases in the discharge potential were observed for high concentrations (250 mg ml-l) of NaCI. The potential decreases were too short-lived to permit the accurate measurement of excitation temperatures. Integrated analytical signals increased linearly with increasing pressure.Keywords Furnace atomization non-thermal excitation spectrometry; atomic emission spectrometry; graphite furnace; helium discharge Hollow anode furnace atomization non-thermal excitation spectrometry (HA-FANES)1-3 is one of three techniques which combine the high atomization efficiency of the graphite furnace with the excitation capability of an electrical discharge. The concept of FANES was first introduced by Falk and co-~orkers~-~ when they described the use of the furnace as a cathode for a hollow cathode discharge (HC-FANES). The HC-FANES technique uses classic hollow cathode lamp operating pressures (6.7-40 hPa) and currents ( 10-50 mA) to sustain glow discharges in either Ar or He.The HA-FANES method was first suggested by Ballou et a1.I as a means of achieving a more stable discharge. The furnace is used as the anode and a carbon rod running the length of the centre of the furnace serves as the cathode. Optimum operating pressures and currents are considerably higher than those used for HC-FANES. Blades and co-workers8-*0 and Sturgeon and co-worker~I~-~~ com- bined the physical design of HA-FANES with an r.f. discharge in He at atmospheric pressure. This technique was named furnace atomization plasma excitation spectro- metry (FAPES). All three approaches HA- and HC-FANES and FAPES have detection limits comparable to conven- tional electrothermal atomic absorption spectrometry (ETAAS) and are inspired by the same goal the develop- ment of a simultaneous multi-element graphite furnace emission source.The HA-FANES method sacrifices the enhanced emis- sion intensities and electron densities associated with the hollow cathode geometry,18 but achieves comparable inte- grated intensities by operating at fill-gas (gas used to fill the vacuum cross) pressures almost an order of magnitude higher.2>3 In Ar the integrated analytical signals increased linearly with increasing pressure up to 266.7 hPa. Dis- charges at higher pressures were not possible in Ar. Between 200 and 266.7 hPa the detection limits for HA-FANES are comparable to those for HC-FANES and conventional ETAAS. The increase in the signal with pressure is * Present address School of Biological and Chemical Sciences University of Greenwich Wellington Street Woolwich London UK SE18 6PF.t To whom correspondence should be addressed. consistent with an expected analyte loss function controlled by diffusion. Higher pressures produce smaller diffusion coefficients and longer residence times in the furnace. This linear relationship between pressure and the integrated signal also suggests that the mechanism of excitation of the analyte is not adversely affected by the increase in pressure. To date all published reports on HA-FANES have employed a discharge in Ar. Use of a discharge in He would have the advantage of higher excitation states but based on observations of other He discharge^,^^^^^ the disadvantage of lower electron densities. Certainly the need for higher excitation energies exists for efficient excitation of high energy metals and non-metals.The HC-FANES and FAPES techniques use He almost exclusively. For HA-FANES it is anticipated that the increase in integrated intensity with pressure observed with Ar will also be seen for He. The stability of the electrical discharge regardless of the support gas is critical to the use of the HA-FANES technique. Any inconsistency in the ability of the discharge to excite the analyte throughout the atomization cycle will create the potential for analytical inaccuracies. The need for temporal discharge stability for non-thermal excitation is analogous to the need for temporal temperature stability for conventional ETAAS. In both cases a constant environ- ment is necessary in order to achieve reproducible results independent of the sample matrix.The two main sources of instability of the discharge are the generation of thermionic electrons at temperatures above 1800 K and easily ionizable elements (EIEs). In both cases charge carriers that can alter the discharge potential and the mechanisms of the excitation process are intro- duced into the discharge. It is well documented that in the constant current mode the presence of thermionic elec- trons causes a dramatic drop in the discharge potential for HA-FANES using Ar3 and for HC-FANES using either Ar or He.7 In both cases the drop in potential is associated with the cathodic temperature the wall temperature of HC- FANES and the central carbon rod temperature of HA- FANES. With FAPES an increase in reflected power is observed with the presence of thermionic electron^.^^ Sturgeon et a1.14 pointed out that the temperature of the central electrode not that of the furnace wall correlated with the onset of reflected power.This is consistent with the946 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 expected self biasing of the r.f. e1ectrode.l8 The increase in reflected power can be avoided with the use of an automated impedance matching circuit provided a reduced furnace heating rate (as compared to conventional ETAAS) is used.1° The use of the excitation temperature to characterize the thermodynamic nature of discharges is common. Unfortu- nately at lower pressures the lack of thermodynamic equilibrium (lack of agreement between excitation transla- tion vibration and rotation temperatures) reduces the significance of these measurements.Excitation tempera- tures however can be used to monitor changes in the relative populations of the excited states of an atomic species as a function of time i.e. the absolute temperature has no meaning but relative changes can be a diagnostic for the plasma stability. To avoid ambiguity in this study the computed slopes of the linear least squares fit of intensity as a function of transition energy is called the excitation environment. Falk et aL7 have reported that the excitation envirionment of HC-FANES is relatively constant over the atomization cycle for atomization temperatures between 500 and 2500 K for atomization pressures between 13.3 and 66.7 hPa and for Ar or He.Similar measurements are lacking for HA-FANES and FAPES. Hettipathirana and BladeslO determined excitation environments for FAPES based on intensity measurements integrated over the duration of the atomization cycle. Sturgeon et all6 mea- sured instantaneous excitation kinetic and ionization temperatures for FAPES. They concluded that local ther- modynamic equilibrium did not exist. In neither case were excitation environments measured as a function of time during the atomization cycle nor has the stability of the excitation environments been determined for any of the techniques in the presence of an EIE. This study investigates the use of an He discharge for HA- FANES at pressures of 66.7-800 hPa. Current-voltage characteristics and background spectra are documented.The variation of the discharge voltage and the excitation environment were determined as a function of atomization temperature pressure and the presence of an EIE. The behaviour of the analytical signal as a function of pressure was determined. Experimental Instrumentation The HA-FANES device consists of a pyrolytic graphite coated graphite integrated contact cuvette (ICC) (Rings- dorff Werke Bonn Germany) with a pyrolytic graphite coated 1 .O mm diameter centrally positioned carbon rod (Ringsdorff Werke) running the length of the furnace. The ICC was heated by an HGA 500 power supply (Perkin- Elmer Norwalk CT USA) which had been modified as previously de~cribed.~ Water cooling of the furnace was provided by a recirculating water-bath (Model FK Haake Saddle Brook NJ USA).A high voltage d.c. power supply (BHK 1000-0.2M Kepco Flushing NY USA) operating in the constant current mode maintained the discharge within the furnace. The furnace was housed in a 100 mm diameter six-way vacuum cross (Huntington Labs Mt. View CA USA). Pressure was monitored using two pressure gauges an analogue scale Convectron vacuum gauge (Series 275 Granville-Phillips Boulder CO USA) and a variable capacitance pressure sensor (Baratron Type 122A MKS Andover MA USA) and associated power supply and digital scale (Model PDR-C-lC MKS). The chamber was evacuated using a vacuum pump (Duo Seal Model 1376 Sargent-Welch Scientific North Linder IL USA) with a dry-ice trap. In previous work3 a Spex (Edison NJ USA) 1704 monochromator was used. This monochromator was used in this study (with 25 pm slit-widths) only to scan the background emission spectra of the He discharge from 200 to 400 nm.For the rest of the work in this study an Cchelle polychromator (Spectraspan 111 Spectrametrics Andover MA USA) was used in the single-element mode with photomultiplier tube (PMT) detection. Matching slits 50 pm wide and 500 pm high were used. The current from the PMT was amplified using a custom-built operational amplifier circuit consisting of a current-to-voltage converter and a non-inverting amplifier. Wavelength modulation was used for background correction as previously de~cribed.~ A quartz refractor plate (25.4 x 25.4 x 6 mm) was mounted on a galvanometer (Model G300PD General Scanning Watertown MA USA) and run by a scanner controller (CCX-101 General Scanning).A 100 Hz sinusoidal wave form was furnished to the scanner controller by a function generator. Emission signals were detected using a lock-in amplifier (Model 128A Princeton Applied Research Princeton NJ USA) operating in the 2f mode (200 Hz) with a 0.1 s pre-filter. The lock-in amplifier signals were digitized at 200 Hz using a 12 bit analogue-to-digital (ND) converter (Model DAC A1 13 Interactive Structure Bala Cynwyd PA USA) mounted in an Apple microcomputer (Apple I1 Plus Apple Computers Cupertino CA USA). The data acquisition program recorded the maximum intensity signal or peak height and summed the peak area over the entire atomization cycle and over any specified interval. Data were subsequently plotted on a strip-chart recorder (Model BD 40 Kipp & Zonen Amsterdam The Netherlands).This was achieved using an 8 bit D/A converter (Model AO-03 Interactive Structures). Discharge potentials were monitored on a 0-5 V scale using a custom- built potential divider (voltage divided 1 120). Procedures All experiments were performed in a 'static' mode i.e. there was no gas moving in or out of the furnace system during the experiments. The furnace system was evacuated to 0.13 hPa the valve to the vacuum was shut off the system was filled to the desired pressure with He the He inlet valve was shut off and the experiment(s) were run. The integrity of the vacuum system was sufficient to prevent any significant change in pressure over a 15 min period. Initial experiments were designed to investigate the dependence of the discharge current-voltage characteristics on pressure with the furnace off.Voltage readings were taken at currents ranging from 10 to 200 mA at pressures of 133.3 266.7 400 533.3 666.7 and 800 hPa He. A multimeter (Model 75 Fluke Everett WA USA) was connected to the 0-5 V output of a custom-built voltage divider3 to provide a digital readout. The voltages were then multiplied by 120 to take account of the effect of the divider. Physical characteristics were observed and photo- graphically recorded. The discharge voltage was monitored as a function of time and temperature while the furnace was run by connecting the voltage divider to the A/D converter of the computer. Data were acquired for 10 s a 3 s atomization cycle at 2000 "C and a 7 s cool-down period.In this manner the effects of pressure and current on the discharge voltage were measured. Excitation environments of the discharge were measured by using Boltzmann plots of five He I line intensities as described by Sturgeon et a1.16 The wavelength of the five emission lines (and the upper energy level of the observed transition) were 388.87 nm (23.01 eV) 396.47 nm (23.74 eV) 438.79 nm (24.04 eV) 492.19 nm (23.74 eV) and 501.57 nm (23.09 eV). The logarithm of (Id3/& plotted against energy has a slope of 1/2.303 kT where I is the intensity of the atomic line d is the wavelength g isJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993. VOL. 8 947 the statistical weight f i s the oscillator frequency k is the Boltzmann constant and T is the discharge temperature.To measure the excitation environment as a function of time during the atomization cycle the intensities of each of the five He lines were acquired during five separate atomizations. For each atomization 200 readings were taken over the 3 s atomization (2000 "C) and 7 s cool down cycle. The intensities were then entered into a spreadsheet (Microsoft Excel Microsoft Redmond WA USA) where the intensities were adjusted for spectrometer throughput and detector response. l9 The slopes were calculated using linear regression and the temperatures determined. The oscillator strength data were taken from values compiled by Sturgeon et d . l 6 The effect of the introduction of an EIE i.e. sodium on the excitation temperature was assessed by atomizing various concentrations of sodium and measuring the discharge voltage during the atomization and cooling cycle.Temperature computations were identical with those de- scribed above. Results and Discussion Physical Characteristics It was observed that stable discharges could be sustained in He at pressures as high as 1200 hPa. Fig. 1 shows the current-voltage plots for discharges in He from 133.3 to 800 hPa. These measurements were made with a 'cold' anode i.e. with the furnace off. Discharges at 933.3 1066.7 and 1200 hPa were possible but required currents greater than 200 mA (off-scale for the high voltage power supply) for the discharge to cover the whole cathode surface. Consequently at these pressure levels the current-voltage relationship could not be quantitatively characterized. It can be seen that at lower pressures of He (1 33.3-533.3 hPa) the highest current levels are accompanied by a sharp increase in the voltage.This is called the 'abnormal' region of the discharge and occurs after the discharge has ex- panded to completely cover the cathodic surface. Further current increase can only be achieved with higher current densities. As the pressure increases higher currents are necessary for the discharge to cover the cathode surface but as the pressure increases the required voltage necessary to increase the current density is not as great. At 666.7 and 500 z 6 g 400 c1 - 0 > 300 200 A I i C 0 100 Current/mA 200 Fig. 1 Discharge potential as a function of curent at pressures of A 133.3; B 266.7; C 400; D 533.3; E 676.7; and F 800 hPa of He.The furnace was at room temperature 800 hPa no sharp voltage increase is observed despite the fact that the cathode surface is completely covered at currents of approximately 120 mA. Fig. 2 shows photographs of discharges at 133.3,400 and 1200 hPa. At all three pressures the discharge completely sheaths the centrally positioned carbon rod. In all three cases the furnace was at room temperature. At 133.3 hPa the discharge is very diffuse. With increasing pressure the discharge contracts towards the cathode. This phenomenon is a result of the decrease in the electron mean free path and has been well characterized for hollow cathode discharges. l8 In the case of HA-FANES however the discharge contracts towards the central cathode.At pressures above 400 hPa the discharge continues to shrink slightly but the change is relatively minor. It can be seen that the discharge at 400 hPa is closer in appearance to the 1200 hPa discharge than the 133.3 hPa discharge. The main difference between the discharges at 400 and 1200 hPa is the slightly reddish colour of the cathode at 1200 hPa. This is a result of the ohmic heating of the cathode at the high currents (>200 mA) needed to sustain the discharge. The stabilization of a He discharge at atmospheric pressure has previously only been possible in FAPES with an r.f. generated discharge running at 20 to 100 W.8-17 The ability of HA-FANES to sustain a d.c. discharge above atmospheric pressure is likely to be a result of the small cathodic area and the high current density (Table 1).Although discharges at similar pressures might be sustaina- ble for HC-FANES the high current requirements the optical obstruction of the furnace wall and the high background from black-body emission make such operating conditions of little interest. Although the cathode of HA- FANES does hinder optical imaging the lag of the cathode temperature behind the wall temperature3 results in little background from black-body emission. The HA-FANES system has an operating power of -40 W close to that of the r.f. discharge. Unfortunately 200 mA was the limit of the power supply. It might be possible to sustain discharges at higher pressures but greater currents are required. Emission Spectra Fig. 3 shows the emission spectrum of HA-FANES obtained at 400 hPa of He with a current of 120 mA and with the furnace off.This spectrum was obtained using a scanning monochromator (see under Experimental). No corrections have been made for the wavelength dependent response of the monochromator or the PMT. It was observed that the intensity of the spectra is much more dependent on the current than on the pressure. There was little change in the intensity of the spectra between 133.3 and 400 hPa whereas an increase in the current from 80 to 120 mA produced a 50% increase in intensity. Compared with spectra previously obtained in the relative intensities of the CN and CH bands are significantly diminished. This might be due to reduced sputtering of the carbon from the cathode with He.Upon first examination the spectrum in Fig. 3 appears to be almost identical with that reported by Hettipathirana and BladeslO for FAPES with a sealed environment. They Table 1 Comparison of HA- and HC-FANES HA-FANES H C-FANES Parameter Ar He Ar He Cathodic area/mm2 69 69 528 528 Maximum current Current/mA 10-110 10-200 15-80 15-80 density/mA mm-2 1.59 2.90 0.15 0.15948 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 I I I 1 200 240 280 320 360 400 Wavelengthlnm Fig. 3 Discharge spectrum at 200-400 nm taken at 400 hPa of He and 120 mA. The furnace was at room temperature reported dominant bands from NH N2 and Nz+. Close examination however shows some important differences. Hettipathirana and BladeslO identified the bandhead at 392 nm as that of the first negative system of N2+.They confirmed the presence of N2+ by the strong bandheads at 429 and 469 nm. In Fig. 3 it can be clearly seen that for HA-FANES the bandhead in this region lies below 390 nm. This bandhead is most likely due to CN at 388.3 nm CH at 388.9 and the strong He line at 388.9 nm. It seems most likely that the bands ending at 3 14.5 336.0 and 359.0 nm are CH NH and CN respectively. Sturgeon et al.15 reported an absence of emission from the N2 bands in a sealed system and concluded that the ingress of air against the positive flow of He and Ar was the source of N2. Hettipathirana and BladesIO have suggested that the pres- ence of the Nz and N2+ bands might arise from N2 contamination of the He. In both sealed FAPES systems and in HA-FANES using either Ar or He NO bands are not detectable.This supports the conclusion of Sturgeon et al.I5 that the source of the NO bands in 'open' FAPES systems was atmospheric 02. The presence of OH NH CH and CN bands between 330 and 400 nm presents significant background spectra against which analytical signals must be detected. These spectra suggest the advantage of using a high-resolution spectro- 600 1 I 1 1 1 0 2.0 4.0 6.0 8.0 10.0 Time/s Fig. 4 Discharge potential at 66.7 hPa of He versus time for 3 s atomization at 2000 K at currents of A 25; B 30; C 40; and D 50 mA 500 400 5 300 n P 5 200 n .- C Y Q v) .- 100 A c - 0 2.0 4.0 6.0 8.0 10.0 Tirnels Fig. 5 Discharge potential at 266.7 hPa of He versus time for 3 s atomization at 2000 K at currents of A 50; B 70; C 90; D 100; E 110; F 130; G 150; and H 170 mA 300 1 3 200 Q 0) Q Q .- c c P 2 100 n r 0 0 2.0 4.0 6.0 8.0 10.0 Tirnels Fig.6 Discharge potential at 800 hPa of He versus time for 3 s atomization at 2000 K at currents of A 160; B 180; and C 200 mA meter for detection. The change in these bands as a function of temperature is of considerable interest since the molecu- lar bands present a complex and dynamically changing background. Unfortunately the scanning method used here is not suitable for acquiring background spectra during the furnace atomization cycle. Discharge Potential versus Time Figs. 4-6 show the discharge potential (voltage drop between the anode and cathode) as a function of time for discharges in 66.7 266.7 and 800 hPa of He at a series of currents.The currents were selected to cover the stable operating range at each pressure i.e. the lowest current was high enough to ensure complete coverage of the cathode by the discharge and the highest current was low enough that shorting arcs between the anode and cathode were not observed (i.e. stable operation was still maintained). Data were acquired at 20 Hz for 10 s the 3 s of the atomizationFig. 2 Hollow anode discharges at (a) 133.3; ( b ) 400; and (c) 1200 hPa of He. The furnace was at room temperature [to face page 9481JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY. OCTOBER 1993 VOL. 8 949 cycle and the first 7 s of the cool-down cycle The discharge was started 10 s prior to the atomization cycle (for this study) and was maintained throughout the 10 s period of data acquisition. Optical pyrometric measurements of the furnace wall temperature as a function of time have previously been rep~rted.~ With a 3 s atomization at a temperature of 2300 K the furnace requires approximately 1.25 s to reach the programmed temperature holds there for 1.75 s and then cools rapidly dropping below the lower measurement limit of the optical pyrometer (approximately 1300 K on the scale used) after approximately 3.5 s of total elapsed time.The response of the discharge potential (Fig. 4) shows that the cathodic temperature lags considerably behind that of the furnace wall since the drop in potential is not expected until the cathode reaches 1750 K. Unfortunately with this experimental arrangement it was not possible to measure accurately the temperature of the cathode directly with the optical pyrometer.The focusing optics were marginal for imaging the cathode through the dosing hole and reflected light from the much hotter furnace wall made any measure- ment suspect. The cathode is subject to heating from two different sources ohmic heating from the discharge and radiative heating from the furnace wall. Since the discharge is ignited prior to the start of the atomization cycle the cathode will be at a higher temperature than the furnace wall at the start of the cycle. The furnace wall will then heat rapidly and quickly exceed the cathodic temperature. Positioned in the centre of the furnace with no contact with the furnace wall the cathode is slow to heat much slower than a L’vov platform.Falk et a/.’ reported sharp potential drops between cathodic temperatures of 1750 and 1800 K. Although the furnace wall reaches 2300 K in 1.25 s Fig. 4 shows that the cathode for HA-FANES just reaches 1750 K after approximately 2.5 s. The cathode is also slower to cool than a platform. Unlike the furnace contacts the Macor mount of the cathode is not cooled by a circulating water supply. Heat dissipation of the cathode is through radiative processes and conduction through the electrical connec- tions. Macor is thermally non-conducting but the tempera- ture of the block was observed to increase steadily with routine atomization. The time between atomizations neces- sary to prevent an increase in the temperature of the Macor block was 5 to 10 min.The initial potential of the discharge is dependent on the current and correlates with the increase in cathode temperature induced by resistive heating. This is shown clearly in Fig. 4 where the initial temperature of a discharge at 66.7 hPa of He ranges from 290 to 420 V for currents of 25-50 mA. There is a marked increase in the potential during the first 2.5 s as the density of the He in the furnace is reduced with increasing temperature. Al- though the vacuum cross is held at a constant pressure there is a large dead volume outside the furnace. Thus the gas density inside the furnace decreases and the overall pressure in the system increases by 2.7-4.0 hPa. A dramatic drop in potential to 20-30 V occurs after 2.5 s as a result of the appearance of the thermionic electrons.When the furnace power shuts off after 3 s the potential of the cooling stage then proceeds to retrace itself in an asymmetric mirror image of the heating stage. The asym- metry is due to the cooling rate being slower than the heating rate. The close overlap of the sharp potential drop (at approximately 2.8 s) for each of the currents suggests that the ohmic heating at these currents is relatively small compared with the radiative heating. This was verified visually. There was no sign of heating of the cathode prior to the start of atomization. At a pressure of 266.7 hPa (Fig. 5) the initial potential ranges from 200 to 450 V. The initial increase in potential 2500 I ( 6 ) 3500 I 3000 2500 1 I I 1 I 1 I L > C ‘5 3000 w (d) 3500 3000 2500 1 1 1 1 I I 3000 I 2500 1 I I I I 1 I 0 2.0 4.0 6.0 8.0 10.0 Time/s Fig.7 Excitation temperatures versus time for 3 s atomization at 2000 K for a He discharge at (a) 66.7 hPa and 25 mA (b) 133.3 hPa and 50 mA ( c ) 266.7 hPa and 100 mA (d) 533.3 hPa and 1300 rnA and (e) 800 hPa and 180 mA prior to the sharp drop is reduced (compared with 66.7 hPa in Fig. 4) and the time of the onset of the potential drop is highly current dependent. The time of onset ranges from approximately 2.2 s for the lowest currents (50 and 70 mA) to approximately 1.5 s for the highest current (1 70 mA). The minimum potential for the lowest currents appears to be the same as that observed at 66.7 hPa (20-30 V). With increasing current the minimum steadily increases reaching a minimum of 50 V at 170 mA. Visually the cathode can be seen to glow red hot prior to the start of the atomization cycle with currents above 100 mA.At a pressure of 800 hPa (Fig. 6) the upper limit of the current is restricted by the high voltage power supply. A minimum current of approximately 120 mA is necessary for the discharge to cover the anode. The difference in ohmic heating between 160 and 200 mA seems minor. The onset of the potential drop for all three currents is about the same and is approximately the same as that for the highest currents at 266.7 hPa. Possibly with the higher pressures convective cooling of the cathode is more significant. The minimum potential of around 50 V is the same as that observed with the higher currents at 266.7 hPa.950 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 5 1 I 1 I 1 I 1 0 2.0 4.0 6.0 8.0 10.0 Time/s 4 A B 23E 1 L I I 1 I I 0 2.0 4.0 6.0 8.0 10.0 Time/s Fig. 8 He line intensities versus time for 3 s atomization at 2000 K at 66.7 hPa and 25 mA at A 388.9; B 501.6; C 492.2; D 396.5; and E 438.8 nm 0 2.0 4.0 6.0 8.0 10.0 Time/s Fig. 9 He line intensities versus time for 3 s atomization at 2000 K at 266.7 hPa and 100 mA at A 388.9; B 501.6; C 492.2; D 396.5; and E 438.8 nm Excitation Environment Versus Time The extreme drop in the discharge potential observed in Figs. 4-6 raises the question of the stability of He discharge with respect to temperature. If the excitation process is thermally dependent then a change in the analyte vaporiza- tion temperature due to a matrix component can lead to a significant difference in the magnitude of the analytical signal and can be a source of interference.Fig. 7 shows the excitation environment of HA-FANES measured as a function of time at different pressures. At each pressure an optimum current was used (see under Experimental). Values for the excitation environment were computed as described under Experimental from the recorded intensities of five He lines. Figs. 8-10 provide examples of the variation of the individual He lines as a function of time at 66.7 266.7 and 800 hPa. Each of the traces in Figs. 8- 1 0 were acquired during separate atomiza- tions. Like the discharge potential studies in the previous Fig. 10 He line intensities versus time for 3 s atomization at 2000 K at 800 hPa and 180 mA A 388.9; B 501.6; C 492.2; D 396.5; and E 438.8 nm section data were acquired for a 3 s atomization at 2300 K and the first 7 s of the cool-down cycle.The precision of the excitation environment measure- ments can be best evaluated by examining the precision of the computed values between 7 and 10 s into the atomiza- tion cycle. Although the temperature is still decreasing slowly the precision of the values of the excitation environment about a linear least squares fit is about 3%. The biggest source of imprecision arises from the fact that five separate atomizations must be used one for each transition. When the intensity is changing rapidly differ- ences in timing between individual atomizations can introduce significant error. It can be seen from Fig.7 that the excitation environment is relatively stable as a function of time and furnace temperature. The excitation environment varies most signi- ficantly at the lowest pressures. At 266.7 533.3 and 800 hPa the maximum variation was less than 10% (300 K). At 66.7 and 133.3 hPa the greatest variation was found between 2 and 5 s the interval during which the discharge potential goes through a dramatic decrease and recovery (Fig. 4). The regions of greatest variations in the excitation environment correspond to periods of most rapid change in the He line intensities. This suggests that variation between atomizations is significant and that the stability of the excitation environment may be better than that shown in Fig. 7. Figs. 8-10 all show that in general the emitted intensi- ties of the He lines were inversely proportional to the energy of the upper level of the observed transition (see under Experimental).The difference in the responses of the five He lines is most easily seen in Fig. 8. The initial (0-2.5 s) slopes of the 388.9 50 1.6 and 396.5 nm lines match closely and are less steep than the slopes of the 492.2 and 438.8 nm lines. With the appearance of the thermionic electrons the 388.9 492.2 and 438.8 nm lines exhibit the greatest change in intensity. There appears to be no correlation between the energy of the upper level of the transition and the response to the presence of thermionic electrons. This suggests that the discharge remains relatively stable during the atomiza- tion cycle. Figs. 9 and 10 show that the variation in the intensity of all the He lines as a function of time was almost identical.There are no dramatic variations between lines. Conse- quently the variations in excitation environment observedJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 95 1 in Fig. 7 for pressures of 266.7 hPa and higher were the result of subtle changes in the line intensities. This observation strengthens previous speculation that at these pressures variations between atomizations might limit the accuracy of the excitation temperatures. The stability of the excitation environment is somewhat deceiving at low pressures (66.7 hPa). Although the excita- tion environment remains relatively stable the concentra- tion of each of the excited species as indicated by the measured intensities (Fig.8) decreased by almost two orders of magnitude. This exceeds the decrease in density predicted for the increase in temperature (296-2473 K) by an order of magnitude. It appears that the flood of thermionic electrons depletes the population of the excited species but in a proportional manner such that the excitation environment remains the same. The ratio of the fill-gas atoms to the analyte atoms is thus decreased dramatically and could result in altered efficiency of excitation of the analyte. At higher pressures (>133.3 hPa) the ratio of excited species to the analyte is of much less concern (Figs. 9 and 10). The variation in intensity as a function of tempera- ture shows that the populations of excited species might actually increase during the atomization cycle.This sug- gest that the discharge is much more robust at higher pressures. It is interesting to note that values for the excitation environment measured here are in good agreement with those measured for FAPES by Hettipathirana and Blades’O and Sturgeon et a/.16 using He at atmospheric pressure. In each case the measured values were close to 3000 K. Although local thermodynamic equilibrium does not ex- ist,I6 the excitation process and the relative populations of excited species appear to be similar. Falk et a/.’ measured excitation environment values between 9000 and 11000 K for discharges in He at pressures between 13.3 and 66.7 hPa. He observed that the measured values decreased consistently with increasing pressure.The lowest excitation environment of 9000 K was measured at 66.7 hPa. The failure of HA-FANES excitation temperatures to agree with those of Falk et al. at 66.7 hPa is most likely due to this difference in geometry. The enhancement in electron densities and emission intensities obtained with the hollow cathode geometry (HC-FANES) as compared with a planar cathode geometry (HA-FANES) is well documented.’* 0 4 8 0 4 8 0 4 8 Time/s Fig. 11 Discharge potential versus time for (a) 1; (b) 10; ( c ) 100; (d) 250; (e) 500; and U 1OOOpg ml-I of Na (as the chloride salt) for 3 s atomization at 2000 K at 533.3 hPa of He and 130 mA Discharge Potential Versus EIE The rapid atomization of an EIE into the discharge can be expected to alter the discharge potential having the same effect on the discharge as the sudden appearance of the thermionic electrons. In this study Na as the chloride salt was chosen as the EIE.The effect of NaCl on the discharge potential was determined as a function of the concentration of Na (with fixed He pressure and current) as a function of the current (with fixed concentration of Na and pressure) and as a function of the pressure (with fixed concentration of Na and optimized current). First as shown in Fig. 1 1 the discharge potential was measured as a function of time for a series of solutions containing concentrations of Na ranging from 1.0 pg ml-’ (2.5 pg ml-1 of NaCl) to 1000 pg ml-I (2500 ,ug ml-1 of NaCl). The total mass of Na ranged from 10 ng to 10 pg. Like the previous studies data were acquired at 20 Hz for a 3 s atomization at 2300 K and the first 7 s of the cool-down cycle. The He pressure was 533.3 hPa and the discharge current was 130 mA.Fig. 11 shows that perturbations of the discharge poten- tial due to Na starts at concentrations as low as 10 pg ml-I. U 300 F 8 a 200 r 100 0 3 10 1 0 3 10 0 3 10 Time/s Fig. 12 Discharge potential versus time for 250 pg ml-I of NaCl (upper traces) and de-ionized water (lower traces) for 3 s atomization at 2000 K and 533.3 hPa of He at (a) and (e) 100; (b) and 140; ( c ) and (g) 160; and (d) and (h) 200 mA952 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 13.33 40.00 66.67 Pressurelh Pa 26.67 53.33 80.00 Fig. 13 Integrated intensity of A B; and B As as a function of the discharge pressure The trace for 1 pg ml-' of Na was identical with that of a de- ionized water blank (not shown). The NaCl vaporizes at a fairly low temperature (1 680 K) and therefore at an early time (0.8-1.8 s).The perturbation can be characterized as an initial increase and then a decrease in the potential within the first second of the atomization cycle. At concentrations of 500 and 1000 pg ml-l of Na there is a very rapid and significant decrease in the discharge poten- tial at about 0.8 s. At 1000 pg ml-l the potential appears to drop almost to zero. It is apparent that the cloud of easily ionized atoms provides a path for an arc between the anode and cathode. This could not be visually observed since the furnace is heating rapidly at this time. Attempts to measure excitation temperatures in the presence of 500 or 1000 pg ml-l of Na were unsuccessful.The rapid transient nature of the perturbation and the limitation of measuring the He line intensities during successive atomizations resulted in extremely erratic re- sults. The reproducibility of the intensity-time plots for a single He line was very poor over the time interval when the Na volatilized. Measurements at lower concentrations of Na gave temperature versus time plots that were similar to those in Fig. 7. The plots in Fig. 11 show that the effect of an EIE on the discharge potential is very short-lived with a full width at half maximum of approximately 0.2 s. An interference will occur only if there is a direct temporal overlap of the EIE and the analyte. For Na specifically interferences would be expected only for the most volatile elements those elements that volatilize between 0.5 and 1.0 s into the atomization cycle.Quantification of the interference will not be possible until specific analytes are determined in the presence of Na and other EIEs. A solution containing 100 pg ml-1 of Na was atomized at 2300 K in 533.3 hPa of He using discharge currents ranging from 100 to 200 mA. In Fig. 12 the Na solutions in the upper traces can be compared with de-ionized water in the lower traces. It can be seen that as the current increased the perturbation due to Na within the first second of the atomization cycle diminished. At 200 mA the difference between de-ionized water and 100 pg ml-' of Na is almost negligible. Finally the effect of Na on the discharge potential was determined as a function of the pressure.A solution of 100 pg ml-I of Na was atomized at 2300 K at 66.7 533.3 and 800 hPa. The plot of the discharge potential as a function of time obtained at 800 hPa and 160 mA was almost identical with that obtained at 533.3 hPa and 160 mA [Fig. 12(c)]. At 66.7 hPa a considerably lower current (40 mA) was used and the drop in the discharge potential with the volatiliza- tion of the Na was much more pronounced. This could be a result of the lower density of the fill gas which increases the EIE to fill gas ratio or the lower current (Fig. 12). It appears that the discharge in He is more robust at higher pressures and currents. Analytical Signal versus Pressure The analytical signals for As and B were determined as a function of temperature and pressure.At each pressure a range of currents was used. This current range was shifted to higher values as the pressure was increased in order for the discharge to completely cover the cathode (as discussed earlier). The atomization temperatures for As and B were 2 100 and 2300 K respectively at wavelengths of 278.0 and 249.7 nm. The time integrated signals of As and B were found to increase linearly with increasing He pressure as shown in Fig. 13. This is identical with the behaviour of the integrated analytical signals in the Ar di~charge.~J There was some variation in the integrated signals over the range of currents at each pressure consequently the average integrated signal was plotted in Fig. 13. Values are not presented for As at 800 hPa as very high blank levels made the results suspect.Non-zero intercepts of the intensity axis are seen for both elements. This is an artifact of the wavelength modulation process and the background emis- sion structure. The negative offsets were constant and additive in nature. The linear increase in the integrated As and B signals with pressure implies that the efficiency of excitation of the analyte is independent of the pressure. The diffusion coefficient decreases directly with increasing pressure. Consequently the analyte residence time will be propor- tional to the pressure. Since both the analytical emission signal and the analyte population increase linearly with increased pressure the number of excited analyte atoms must remain a constant fraction of the total population. Thus the efficiency of the excitation process is independent of pressure.The measured excitation temperatures (Fig. 7) suggest a similar conclusion since the average temperature is approximately 3000 K for pressures from 66.7 to 800 hPa. The best signal-to-noise ratios were found at pressures of 66.67 hPa for both elements. At this pressure the emis- sion-time profiles of As and B have full widths at half maximum of approximately 1 s. These widths are similar to those of conventional ETAAS and FAPES. The HA-FANES detection limits for As and B were 45 and 2 pg respectively as compared with 10 and 1000 pg respectively for conventional ETAAS. The HA-FANES detection limits were comparable for As and almost three orders of magnitude better for B.Conclusion Use of He as a fill gas provides HA-FANES with a very robust discharge. Analytically useful discharges have been characterized at pressures up to 800 hPa although higher pressures can be attained with a power supply capable of higher currents. Operating between 533.3 and 800 hPa at currents of 140 to 200 mA. the drop in discharge potential and the fluctuation in the excitation temperature with the appearance of thermionic electrons is significantly reduced.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 953 At these pressures and currents the perturbation of the discharge potential by an EIE (Na) is reduced. These higher operating pressures and currents are analytically significant since the integrated analytical signals increase linearly with increasing pressure. 9 Smith L. E. Liang D. C. Steel D. and Blades M. W. Spectrochim. Acta Part B 1990 45 493. 10 Hettipathirana T. D and Blades M. W. Spectrochim. Acta Part B 1992 47 493. 1 1 Sturgeon R. E. Willie S. N. Luong V. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. 12 Sturgeon R. E. Willie S. N. Luong V. and Berman S. S. Anal. Chem. 1990 62 2370. 13 Sturgeon R. E. Willie S. N. Luong V. and Berman S. S. J. Anal. At. Spectrom. 1990 5 635. References 1 Ballou N. E. Styris D. L. and Harnly J. M. J. Anal. At. Spectrom. 1988 3 1141. 2 Harnly J. M. Styris D. L. and Ballou N. E. J. Anal. A t . Spectrom. 1990 5 139. 3 Riby P. G. Harnly J. M. Styris D. L. and Ballou N. E. Spectrochim. Acta Part B 1991 46 203. 4 Falk H. Spectrochim. Acta Part B 1977 32 437. 5 Falk H. Hoffman E. and Ludke Ch. Spectrochim. Acta Part B 1981 36 767. 6 Falk H. Hoffman E. and Ludke Ch. Spectrochim. Acta Part B 1984 39 283. 7 Falk H. Hoffman E. and Ludke Ch. Prog. Anal. Spectrosc. 1988 11 417. 8 Liang D. C. and Blades M. W. Spectrochim. Acta Part B 1989,44 1059. 14 Sturgeon R.-E. Willik S. N. Luong V. and Berman S. S. J. Anal. At. Spectrom. 1991 6 19. I5 Sturgeon R. E. Willie S. N. Luong V. and Berman S. S. Appl. Spectrosc. 199 1 45 14 13. 16 Sturgeon R. E. Willie S. N. and Luong V. Spectrochim. Acta Part B 199 1 46 I02 1. 17 Sturgeon R. E. and Willie S. N. J. Anal. At. Spectrom. 1992 7 339. 18 Pillow M. E. Spectrochim. Acta Part B 1981 26 821. 19 Zander A. T. Miller M. H. Hendrick M. S. and Eastwood D. Appl. Spectrosc. 1 985 39 1. Paper 3/02896H Received May 20 1993 Accepted June 7 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800945

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 955 Furnace Atomization Plasma Excitation Spectrometry Effects of Sodium Chloride and Sodium Nitrate on Lead and Silver Emission T. D. Hettipathirana and M. W. Blades* Department of Chemistry 2036 Main Mall University of British Columbia Vancouver British Columbia Canada V6T 1Zl The effect of alkali halide and nitrate salts on the emission intensity from Ag and Pb excited using furnace atomization plasma excitation spectrometry has been studied. Compared with the emission signal obtained when there is no interferent the presence of either NaNO or NaCl as a comcomitant in the sample causes an interference effect by decreasing the emission intensity over the power range 14-40 W. Keywords Helium plasma; graphite furnace; interference; easily ionized element; furnace atomization plasma excitation spectrometry Furnace atomization plasma excitation spectrometry (FAPES) uses an emission spectroscopic source in which a plasma formed in the interior of a graphite furnace acts as the excitation medium.'-l0 For FAPES as for electro- thermal atomic absorption spectrometry (ETAAS) samples are deposited onto the graphite furnace wall using a micropipette or other similar sample injector. The sample is subjected to drying and thermal pre-treatment cycles after which the plasma is ignited by applying radiofrequency (r.f.) power to a graphite electrode which is located on the furnace axis.Using electrothermal heating of the graphite furnace analytes are then vaporized from the furnace wall into the gas phase where contact with the plasma can dissociate molecules and excite atoms and ions.The excitation temperature of the plasma has been reported to be between 3500 and 4500 K for r.f. input powers of An important characteristic of any spectrochemical source furnace or plasma is its susceptibility to chemical interferences. Non-spectral interferences in graphite fur- naces which have been studied using AAS can be classified into two distinct types (i) condensed-phase interferences i.e. volatile compound formation incomplete vaporization due to occlusion formation of refractory compounds or change in the rate of analyte supply; and (ii) vapour-phase interferences i.e. a shift in the dissociation equilibria due to stable compound formation shifts in ionization equili- bria or changes in the rate of analyte removal.For plasma sources such as the inductively coupled plasma (ICP) and direct current plasma (DCP) non- spectral interferences are broadly classified as transport (viscosity effects) vaporization dissociation and ionization interferences. Additional sources of interference can occur as a result of either the analyte or interferent altering the plasma properties. Examples of this would be changes in collisional excitation rate due to changes in the electron energy distribution or density and ambipolar diffusion due to shifts in the spatial distribution in plasma constituents. One family of ubiquitous interferents which afflicts both plasmas and graphite furnaces is the alkali salts for example NaCl.The effects of metal and alkali chlorides including NaC1 on the atomic absorption signals for Cu Pb and Ni have been studied.I1-l3 In general it has been found that the integrated Pb absorbance decreased as the amount of NaCl was increased. This effect was attributed to the formation of volatile halides which were lost during the ashing stage. Welz et a1.I' used a dual-cavity platform to study the 20-50 W.3*'0 * To whom correspondence should be addressed. interference effects of NaCl on Pb. They observed that while some Pb is lost as volatile PbCl during the ashing stage Pb was also lost due to analyte vapour expulsion from the volatilization of the NaCl matrix. It appears that the most significant effect of NaCl in ETAAS is not due to the Na but rather to the C1 acting as a carrier. The effects of Na as an easily ionizable element (EIE) have also been extensively studied in flame AAS and ICP,IS DCP,16 and microwave induced plasma17 emission spectro- metry.In flames enhancement of emission intensity from neutral atomic lines is observed as a result of shifts in the ionization equilibrium toward the atomic form. In the ICP both enhancement and depressions in emission signals have been observed. The root cause for this modification of the analyte signal in ICP atomic emission spectrometry (ICP- AES) has not been adequately explained and is still the subject of considerable debate. For DCP-AES the combined effects of shifts in analyte ionization equilibrium and modification of the plasma properties are observed. To date detailed studies on the effects of Na on the FAPES signal have not been reported.Intuitively one would expect that FAPES would experience both kinds of NaCl interference effect i. e. furnace and plasma. However the presence of the plasma may act to dissociate molecular halides and thus help to alleviate some forms of vapour- phase interference experienced in furnaces. Sturgeon et al.* reported a reduction in Pb emission when NaCl was present in the sample. While they did not come to a conclusion about the nature of the effect of NaCl on Pb in FAPES experimental observations were attributed to plasma quenching or condensation of the NaCl onto the central electrode. Smith et al.' observed a slight enhancement of peak height at low concentrations and a depression at higher concentrations for both NaN03 and NaCl.Although the results were very preliminary it was suggested that at low concentrations of interferent the signal was enhanced by a shift in the ionization equilibrium and at higher concentrations the plasma was being quenched. It should be noted that in this study analyte and interferent solutions were prepared in dilute HN03. In addition the FAPES source used was based on a free-running oscillator operating at 27 MHz and the power delivered to the plasma was not measured. One feature of FAPES which distinguishes it from most other common plasma methods is the transient nature of the analyte vaporization and excitation. For spectrometric methods based on differential vaporization and transient signals such as ETAAS and FAPES analyte and interferent can have a different temporal behaviour.Therefore gas- phase interferences might not be present in some instances simply because the potential interferent has different956 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 temporal vaporization characteristics from the analyte. Therefore it is necessary to observe the temporal trace at all points in time in order to understand the effects of Na and to attempt to distinguish between ETAAS and plasma type interference effects. The experiments described in this publication were carried out to study the possible interference effects from Na that can be found in FAPES. No attempt was made to use higher ashing or atomization temperatures or other chemical or instrumental means to reduce the effect of the interferent. It is hoped that studies of this kind will indicate possible ways to achieve this end.Experimental A full description of the instrumentation can be found el~ewhere.~J~ The main components of the FAPES system used in this study were a pyrolytic graphite coated inte- grated contact graphite furnace (Ringsdorff BadGodesberg Germany) and a pyrolytic graphite coated co-axial graphite electrode (Ringsdorff). The plasma was generated and sustained using an Advanced Energy (Fort Collins CO USA) Model RFX-600 13.56 MHz r.f. generator an Advanced Energy Model ATX-600 automatic tuner and an Advanced Energy (Fort Collins) Model 50 1 7-000-G impe- dance matching network. The furnace was heated using an IL-555 (formerly Instrumentation Laboratory now Thermo-Jarrell Ash Waltham MA USA) graphite furnace power supply.The temperature of the graphite furnace was monitored using an Ircon Series 1100 (Ircon IL USA) optical pyrometer which viewed the graphite furnace through the sample deposition hole. The appearance time was defined as the time taken by the absorbance or the emission signal to reach the average baseline plus two standard deviations of the baseline noise. A 0.35 m Model 270 Czerny-Turner monochromator (Schoeffel-MacPherson MA USA) with a holographic grating with 2400 lines mm-' was used. All signals were detected using a Hamamatsu (Middlesex NJ USA) Model R955 photomultiplier (PMT) with an entrance slit of 75 pm wide and 1 mm high. All signals were digitized with 12-bit resolution using a 16 channel Model ADM 12-1 0 (Quatech Akron OH USA) analogue-to-digital card capable of 250 Hz with 2000 points per channel and stored using a 12 MHz IBM PC/AT compatible computer for further processing.Reagents All analyte solutions were prepared using serial dilution (with distilled water) of a 1000 mg 1-1 stock solution prior to analysis unless otherwise noted. Lead and silver solutions were prepared using analytical-reagent grade AgNO and Pb(NO,)* (BDH Poole Dorset UK). Sodium solutions were prepared from analytical-reagent grade NaCl (BDH) and NaNO (Anachem Portland OR USA). Nitric acid (1% v/v) was prepared using analytical-grade reagent (BDH). Procedure A 5 pl aliquot of solution was deposited onto the furnace side wall using an Eppendorf0.5-10 ml UltraMicro pipette.The FAPES source assembly was purged with He (Union Carbide Toronto Canada) and the furnace temperature was increased to 470 K for 45 s to dry the sample and held at 470 K for another 45 s to ash the sample. Within the next 10 s the plasma was ignited and then vaporized to reach the temperature of 2050 K within 5 s for all determinations. The blank determinations were carried out by depositing the same volume of distilled water 1% v/v HNO solution or respective interferent in water. Replicate measurements were carried out for each determination. Resonance lines of Pb 283.30 nm and Ag 328.07 nm were used unless otherwise stated. Sodium emission was measured at 330.23 nm. Results and Discussion As the furnace heats during the atomization cycle its load impedance changes necessitating the use of an automatic impedance matching network to maintain the reflected power at a minimum.In order to assess the effects of NaCl addition on the magnitude of the reflected power and the ability of the r.f. matching network to compensate for changes in reflected power an experiment was carried out in which the reflected power was measured for a constant applied power of 20-40 W. The results of the experiment at 20 W are presented in Fig. 1 which is a plot of reflected power for atomization of distilled water and 162 ng of NaCl in distilled water. The temperature profile for the atomiza- tion cycle is also shown. The average reflected power is about 1 W and although there are excursions around this value they appear to be random rather than systematic.There is no significant difference between the reflected power with and without the presence of NaCl. It should be noted that the heating rate of the furnace for this FAPES system is rather slow when compared with that for ETAAS. While this may compromise the analytical performance for some elements it might be an advantage for maintaining a stable plasma using crystal controlled servo-driven capaci- tor impedance matching. The relatively slow heating rate enables the servo-motors to track changes in impedance leading to a constant value for forward and reflected power at furnace temperatures up to 2200 K. Fig. 2 provides the temporal emission signals for 162 ng of Na as NaCl and for 162 ng of Na as NaNO,. There is differing behaviour for the Na signal derived from NaCl and NaN03.For NaNO there is a single well defined peak with an appearance temperature of 1080 K which is consistent with the published value of the appearance temperature for Na.18 For NaCl there are two peaks a sharp early peak and a peak later in time which has a similar character to Na temporal emission from dosed NaNO,. Campbell and OttawayI8 have suggested that in ETAAS Na can be formed from carbon reduction of Na,O 4NaN0,-2Na20 + 4N02 + O2 Na,O + C-2Na + CO (1) ( 2 ) In the presence of NaC1 Na might also be formed from dissociation of volatilized NaCl. The existence of two peaks for the Na signal from NaCl is indicative of Na being formed by two different mechanisms. The early peak is most likely the result of vaporization and gas-phase 10 I 1 2200 Time/s Fig.1 Reflected power for distilled water (solid line) and 162 ng of Na as NaCl (dotted line) in distilled water at 20 W; graphite furnace heating curve (broken line)JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 957 L = 1.0 t 2 0.8 0 - > 0.6 0.4 w .- v) C .- C . 0.2 l u o fn .- E 0 2.0 4.0 6.0 8.0 Timels Fig. 2 Temporal traces of FAPES emission intensity resulting from the deposition of 162 ng of Na as A NaNO,; and B NaCl 0 2.0 4.0 6.0 8.0 Ti m e/s Fig. 4 Temporal traces of FAPES emission intensity resulting from the deposition of A 0.25 ng of Ag; B 0.25 ng of Ag and 162 ng of Na as NaCl; and C 0.25 ng of Ag and 1 62 ng of Na as NaNO 0 2.0 4.0 6.0 8.0 Timels Fig. 3 Temporal traces of FAPES emission intensity resulting from the deposition of A 0.5 ng of Pb; B 0.5 ng of Pb and 162 ng of Na as NaCl; and C 0.5 ng of Pb and 162 ng of Na as NaNO dissociation of volatile NaCl and the latter peak and that from Na dosed as NaNO is most likely from Na (or Na20) which has condensed on the central electrode and re- vaporized in the gas phase.Fig. 3 shows the temporal emission signals for 0.5 ng of Pb as Pb(NO,)* 0.5 ng of Pb and 162 ng of Na as NaCl and 0.5 ng of Pb and 162 ng of Na as NaNO at 20 W r.f. input power. The effect of both NaNO and NaCl on the Pb emission signal is to reduce the integrated emission relative to the signal with no interferent. The appearance and peak temperatures for the Pb signal with NaCl as the interferent are the same as those for Pb alone.In fact when normalized the signal traces overlap at all points in time within the error of the measurement. This points to a reduction in the excitation ability of the plasma and/or a reduction in vaporization efficiency as the cause for the reduction in signal. It should be noted that Na atomizes later in time than Pb such that the emission signals overlap only during the later stages as Pb is diffusing out of the excitation volume. It is possible that undissociated NaCl is present in the plasma volume at earlier times however the dissociation energy of NaCl is 4.2 eV hence if NaCl is present in the plasma some emission from Na should be observed particularly since the dosing amount is very high. When 0.5 ng of Pb is deposited on the wall and 162 ng of NaCl are deposited on the central electrode the same integrated emission intensity is obtained compared with that obtained with Pb alone.It has previously been reported that Pb deposited on the electrode has about the same appearance time as that deposited on the wall when a 20 W plasma is used.9 These observations suggest that the effect of NaC1 Na or C1 is not due to a post-atomization gas- phase interaction or a reduction in the excitation ability of the plasma. This points to a loss of Pb as PbCl which is either partially dissociated in the plasma (the dissociation energy of PbCl is 3.1 eV) or lost from the furnace in the undetected molecular form before excitation can take place as the most likely cause for the reduction in signal. The effect of NaNO on the Pb emission signal is more complex then that observed for Pb in NaCI.The signal appears at a later time and it evolves much more quickly. Apparently the supply of Pb is much faster in NaNO then in NaCl. Eklund and Holcombe19 have shown that nitrate can reduce the signal for metals in graphite furnaces as a result of oxidation of the metal by NOz and O2 formed from NO according to eqn. (1) during ashing and atomization. A depressive effect and narrowing of the absorbance peak width of Pb has been reported for ETAAS when NaNO is present in the sample.20 This was attributed to an increase in the partial pressure of oxygen formed from the decompo- sition products of NaNO,. It is also interesting to note from the same study that in the presence of CO the Pb peak height increases without any apparent peak shape differ- ence when NaNO is present.This effect was attributed to the decrease in oxygen partial pressure. Holcombe et aL2' reported the formation of a double peak for Pb with a 100- fold excess of NO (as NaNO,) and delayed atomization in the presence of a 1000-fold excess of NO,. The late shift was attributed to an increase in the sticking coefficient of Pb to the oxygenated surface. However in a later study Bass and HolcombeZZ attributed shifts in absorbance signals to the presence of gas phase CO and C02 and refuted the idea of blocking of active sites by oxygen. Presumably in the present experiment PbO is formed from oxidation of Pb by decomposition products of NaNO and PbO dissociates (the dissociation energy of PbO is 3.8 eV) to provide the measured emission signal. The delay in onset of the Pb emission signal is probably because the PbO is bound up in the NaNO matrix and is prevented from undergoing reduction by carbon. When 0.5 ng of Pb is deposited on the wall and 162 ng of NaNO are deposited on the central electrode a slight as yet unex- plained reduction in emission is observed but the peak shape appearance temperature and peak temperature are the same as those obtained when Pb only is deposited on the wall.This is not all that surprising considering the relative temporal emission data from Pb and Na. The signal from Na dosed as NaNO (Fig. 2) does not appear until 2.5 s by which time the Pb signal has almost disappeared (Fig. 3). The data in Fig. 4 are the temporal emission signals for 0.25 ng of Ag (as AgNO,) 0.25 ng of Ag and 162 ng of Na as NaC1,0.25 ng of Ag and 162 ng of Na as NaNO at 20 W r.f.input power. The effect of both NaCl and NaNO on the958 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 1.2 1 .o k 0.8 ,P 0.6 $ 0.4 0 Q Y W W W 5 0.2 - 0” I I 1 1 0 1x10’ 2x10’ 3x10’ Amount of Nahg Fig. 5 Plot of the interference effect of different concentrations of Na as NaCl on emission from 0 Ag and 0 Pb. The vertical axis is normalized to 1.0 for no interferent o-8 I I I 40 0 10 20 30 PowerMl Fig. 6 Plot of the effect of r.f. power on the interference of 162 ng of Na as A NaCI; and B NaN03 on Ag emission intensity. The vertical axis is normalized to 1.0 for no interferent Ag signal is to decrease the integrated emission relative to the case with no interferent.The magnitude of the reduc- tion in Ag emission is much greater for NaN03 compared with NaCl. The effect of both Na-containing interferents is to shift the peak temperatures to lower values. Apparently in the presence of Na (C1 and NO3) either Ag is released from the furnace wall more quickly or both C1 and NO3 are acting as carriers. An additional feature of the Ag temporal response is that that the emission profile for Ag overlaps significantly with the profile for Na emission when NaCl is used; in fact there is an almost direct overlap of the profiles. However because the Na emission from NaNO does not have this early peak one can surmise that there is probably much less atomic Na in the gas phase when Ag is also present with the nitrate interferent.There are several possible scenarios for an interpretation of the data. The NaNO suppresses emission from Ag to a greater extent then NaCl hence NO is a major cause of Ag signal suppression. The most likely explanation for this observation is oxidation of Ag by NO2 and O2 formed from NO ashing and atomization as mentioned previ0us1y.l~ An alternative explanation is that the presence of Na suppresses emission from Ag with both the nitrate and chloride but when C1 is present it acts as a ‘releasing agent’. For the effect of NaCl on Ag alone the formation of gas-phase AgCl (with a dissociation energy of 3.2 eV) and subsequent decrease in the gas phase Ag free atom population is also a possibility. There is some consistency in the observations for Ag and Pb.Fig. 5 is a plot of the interference effect in which the emission intensity for the analyte alone is arbitrarily given a ” 10 20 30 PowerMl 40 Fig. 7 Plot of the effect of r.f. power on the interference of 162 ng of Na as A NaCl; and B NaNO on Pb emission intensity. The vertical axis is normalized to 1.0 for no interferent L w o 2.0 4.0 6.0 8.0 Time/s Fig. 8 Temporal traces of FAPES emission intensity resulting from the deposition of A 0.5 ng of Pb; and B 0.5 ng of Pb and 162 ng of Na as NaCl at 14 W r.f. input power value of 1 and the integrated intensity is plotted for different concentrations of Na as NaCl. It can be seen that even small concentrations of Na cause a marked depression in the Ag or Pb emission.As the concentration of the interferent increases the interference reaches a fairly stable level although the Ag emission appears to experience a small decrease in the severity of the effect at higher concentrations. The fact that Ag and Pb behave similarly points to a common effect causing signal suppression. If one overlooks the details of the temporal evolution of the signal the data in Fig. 5 indicate that the presence of NaCl affects the properties of the plasma decreasing its ablity to excite Ag and Pb. Unfortunately this interpretation of the data is not completely consistent with the experiments in which analyte is dosed on the wall and interferent is dosed on the electrode. However it should be recognized that there may be a spatial component to the phenomena observed which is not appreciated at this time.The effect of the magnitude of the r.f. power on the interference effect of NaCl and NaNO on Ag is provided in Fig. 6 and on Pb in Fig. 7. For Ag in NaCl the effect is essentially independent of the magnitude of the r.f. power thus the interference effect is constant not proportional. A rational interpretation of this observation is that the interference effect is related to losses during atomization or perhaps pre-atomization losses due to the presence of the plasma before the atomization cycle. For Ag in NaNO the interference effect becomes less significant as the r.f. power is increased such that at 40 W it is of an equivalent magnitude for both NaCl and NaN03. This strongly suggests a dissociation effect in the plasma; as the power and hence temperature of the plasma increases the Ag,O is dissociated to a greater extent.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 959 For Pb in NaCl and NaNO the magnitude of the interference effect is essentially the same over the power range 20-40 W although it becomes a little less significant as r.f. power is increased within this range. Similar to Ag these data suggest that the interference effect is partly due to losses during atomization or pre-atomization. A curious feature of the plot in Fig. 7 is that at 14 W there is an enhancement in the emission for Pb in both NaCl and NaNO,. Fig. 8 is a plot of the Pb emission signal with and without 162 ng of Na as NaC1. It is clear that the signal is enhanced in contrast to the observation at higher powers.A similar effect is seen with NaN03 as the concomitant. There is a slight shift in the temporal profile later in time which suggests that the effect is related to an enhancement in the analyte population or an enhancement in the excitation ability of the plasma. Conclusions The effect of sodium chloride and nitrate on the emission intensity from Ag and Pb excited using FAPES has been studied. Compared with the emission signal obtained when there is no interferent the presence of either NaNO or NaCl as a comcomitant in the sample causes an interference effect by decreasing the emission intensity over the power range 14-40 W. There are subtle features associated with the temporal behaviour of the signal which suggest that NaCl and NaNO also act as carriers leading to early atomization of these elements although the carrier mecha- nism is different for the two interferents.For Ag in NaCl the interference effect is related to loss of Ag as undissociated AgCl during atomization or pre- atomization. For Ag in NaNO the effect is caused by the loss of Ag as Ag,O during the atomization cycle which apparently is not dissociated in the plasma. For Pb in NaCl and NaNO the magnitude of the interference effect is essentially the same over the power range 20-40 W. Shifts in the analyte equilibrium between atomic and ionic forms caused by Na acting as an EIE are not observed although the only situation in which there is near complete temporal overlap of the atomic populations of Na and analyte is for atomization of Ag in NaC1.The possible effects of a non-homogeneous distribution of plasma and analyte have not been investigated in this study. Certainly for other plasma sources it has been amply demonstrated that interference effects demonstrate spatial inhomogeneities. The results of a study on this topic will be reported in a future paper. The authors would like to thank the Natural Sciences and Engineering Research Council of Canada for financial support. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 References Sturgeon R. E. Willie S. N. Luong V. T. and Dunn J. G. Appl. Spectrosc. 199 1 45 14 13. Sturgeon R. E. Willie S. N. Luong V. T. and Berman S. S. J. Anal. At. Spectrom. 1991 6 19. Sturgeon R. E. Willie S. N. and Luong V. T.Spectrochim. Acta Part B 1991 46 1021. Sturgeon R. E. Willie S. N. Luong V. T. and Berman S. S. Anal. Chem. 1990 62 2370. Sturgeon R. E. Willie S. N. Luong V. T. and Berman S. S. J. Anal. At. Spectrom. 1990 5 635. Sturgeon R. E. Willie S. N. Luong V. T. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. Smith D. L. Liang D. C. Steel D. and Blades M. W. Spectrochim. Acta Part B 1990 45 493. Liang D. C. and Blades M. W. Spectrochim. Acta Part B 1989,44 1059. Hettipathirana T. D and Blades M. W. J. Anal. At. Spectrom. 1992 7 1039. Hettipathirana T. D. and Blades M. W. Spectrochim. Acta Part B 1992 47 493. Czobik E. J. and Matousek J. P. Anal. Chem. 1978 50 3. Slavin W. and Manning D. C. Anal. Chem. 1979 51 261. Slavin W. Carnrick G. R. and Manning D. C. Anal. Chem. 1984 56 163. Welz B. Akman S. and Schlemmer G. J. Anal. A t . Spectrom. 1987 2 793. Blades M. W. and Horlick G. Specrochim. Acta Part B 1981 36 881. LeBlanc C. and Blades M. W. J. Anal. At. Spectrom. 1990,5 99. Matousek J. P. Orr B. J. and Selby M. Spectrochim. Acta Part B 1986 41 41 5. Campbell W. C. and Ottaway J. M. Talanta 1974 21 837. Eklund R. H. and Holcombe J. A. Anal. Chim. Acta 1979 109 97. Cedergren A. Frech W. and Lundberg E. Anal. Chem. 1984 56 1362. Holcombe J. A. Rayson G. D. and Akerlind N. Jr. Spectrochim. Acta Part B 1984 37 319. Bass D. A. and Holcombe J. A. Anal. Chem. 1988,60 578. Paper 2/04 7791 Received September 4 I992 Accepted April 23 I993

ISSN:0267-9477    

DOI:10.1039/JA9930800955

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 96 1 Use of the Hildebrand Grid Nebulizer as a Sample Introduction System for Microwave-induced Plasma Spectrometry* Henryk Matusiewicz Politechnika Poznanska Department of Analytical Chemistry 60-965 Poznan Poland The Hildebrand grid nebulizer (GN) has been successfully interfaced to a novel highly efficient microwave plasma cavity at atmospheric pressure for the introduction of urine. High matrix solutions from a spray chamber are fed directly into the cavity with no desolvation apparatus. The GN performed well when the signal-to- background ratio was optimized for solution by adjusting the nebulizer gas glow solution uptake rate gas pressure impact bead position and forward plasma power. Errors due to the suppression of the Cu Fe and Zn signal intensities by matrix elements in urine were eliminated using calibration by standard additions coupled with background measurements.Results are presented for the analysis of Human Reference Urine Seronorm. Within the precision of measurement (7-1 1 O/O for replicate determinations) analytical results for the test elements Cu Fe and Zn were in agreement with recommended values. The nebulizer exhibited no clogging problems. Keywords Hildebrand grid nebulizer; microwave-induced plasma spectrometry; sample introduction; urine analysis One of the active areas of analytical research is that of sample introduction which has been and will remain one of the key factors in the successful application of atomic spectrometry. Sample introduction for atomic spectrometry in general has been reviewed by Sneddonl and nebulization into microwave-induced plasmas (MIPS) has received some attention over the past years.There are several methods of sample introduction into an MIP and many different types of nebulizers can be used for this purpose including those of the concentric or cross-flow design the glass-frit nebu- lizer Babington-type nebulizers and ultrasonic nebulizers. One recent approach to nebulization utilizes the Hilde- brand dual platinum grid nebulizer (GN) available from Leeman Labs (Lowell MA USA) which was developed with several goals in mind to be simple to construct; to provide freedom from clogging; to have a uniform depth of liquid at the point of aerosol generation; and to have an efficiency and washout characteristics comparable to those of the cross-flow nebulizer. In the past few years the GN has been used as a sample introduction device for inductively coupled plasmas (ICPs).The design of the nebulizer2v3 and the characteristics of the aerosol it produce^^-^ have been described by other researchers. It was concluded that the grid type generated the largest amount of aerosol (although this is not the only criterion of excellence in a nebulizer) and appeared to provide both high efficiency and robustness for sample handling. Additionally the nebulizer based on a dual platinum grid (as introduced by Leeman) has gas flow requirements similar to those of an MIP. Thus applications of this device have included its use for high salt content solutions and for solutions containing elevated levels of dissolved solid~.~l~J Other workers have described its use for organic solvents,2 selected hydride-forming elernent~,~ clay mineral analysis using slurry nebulizationIO and bone and biological tissues The objective of this study was to investigate the suitability of introducing high salt and high dissolved solids content matrices using grid nebulization for direct sample analysis with the MIP.No previous attempt has been reported for the grid nebulization of any samples into an MIP system; hence preliminary investigations were per- formed to elucidate the applicability of this technique for the rapid analysis of urine. *Presented in part at the 1993 European Winter Conference on Plasma Spectrochemistry Granada Spain January 10- 15 1993.This paper reports for the first time on the joint use of a highly efficient microwave plasma cavity assembly12 coupled to the Hildebrand dual platinum GN. Analytical potential is discussed. Experimental Instrumentation The plasma system consisted of a 700 W 2.45 GHz stabilized generator Model MPC-01 a (Enterprise for Im- plementation of Scientific and Technological Progress Plazmatronika Wrodaw Poland). The microwave cavity and generator used have been described previously.I2 This system was operated without modification and it should be noted that any reference to power refers to that measured at the generator which is the actual power coupled to the plasma. The microwave cavity assembly was mounted on a pedestal which could be moved both vertically and horizontally so as to adjust the observation position.A 75x6.3 mm o.d.x2.5 mm i.d. boron nitride (BN) discharge tube was located centrally in the discharge hole (id. 6.5 mm) of the rectangular cavity. The position of the tip of the BN tube was approximately flush with the end of the cavity. The plasma source was imaged by a quartz lens (f= 160 mm) onto an intermediate diaphragm. An image of the aperture of the diaphragm (d=2.0 mm) was focused by a second quartz lens (f= 76 mm) onto the entrance slit of a 0.5 m Ebert monochromator (Jarrell-Ash Franklin MA USA) having wavelength range 200-450 nm grating constant 1/2242 mm width 54 mm and reciprocal linear dispersion 0.8 nm mm-I. Entrance and exit slit-widths were set at 30 pm. The position of observation was selected to be at the centre of the plasma.By rotating a quartz plate ( I mm thick) in front of the exit slit alternate measurements of the line and background intensities (wavelength shift _+ 0.02 nm) were made. Copper Fe and Zn atomic emissions were measured at the 324.7 372.0 and 213.8 nm lines respectively. The instrument was peaked at the resonance lines using appro- priate electrodeless discharge lamps (powered by an EDL Power Supply Perkin-Elmer Germany) as sources. Analyte emission signals obtained in the single element mode were recorded using output signals from the photomultiplier tube (type EM1 9781A) which were processed through a type 6 10 000 potentiometric recorder (Beckman Fullerton CA USA).962 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 MIP I) Teflon 2 1 cm H Top view Fig. 1 Schematic diagram (to scale) of the proposed cone-type glass spray chamber The gas flow rate was controlled with a precision pressure gauge and monitored with mass flow controllers Models ERG 500 and ERG 2000 (DHN Warsaw Poland) with digital readout (power supply Model ERG 3M3 DHN). Pure argon (99.99%) was used as plasma gas. Aerosols were produced from solutions and introduced into the MIP torch using a Hildebrand dual Pt GN (Leeman Labs) containing two 1 00-mesh platinum screens and mounted in a spray chamber fitted with an impact bead. Details of this GN system have been reported earlier.2~3~6~9 Solution delivery through 0.76 mm i.d. poly(viny1 chloride) (PVC) tubing was controlled by a Gilson Minipuls 3 peristaltic pump (Gilson Middleton WI USA) with tubing tension adjustment.Sample uptake rates were verified by aspirating a solution for 60 s and weighing the solution before and after sampling; reproducibility of uptake (six replicates) was within 1% for uptake rates of 1 .O ml min-l. The aerosol was directed into an in-house designed Pyrex conical spray chamber (internal volume= 50 ml). The nebulizer was placed at an angle of about 10" to the horizontal near the bottom of the chamber along with a glass bead spoiler to avoid flooding of the chamber face which would have impaired analytical performance. Fig. 1 depicts this sample introduction system consisting of a GN impact bead and nebulizer chamber. The pressure inside the system is slightly above atmospheric and kept constant by a liquid column contained within a U-shaped overflow tube.Aerosols were directly transported to the discharge tube via an 1 1 cm length of interconnecting Teflon tubing (i.d. 6 mm). Reagents and Controls Standard solutions of the elements were prepared from Titrisol solutions (Merck Darmstadt Germany) contain- ing 2 g 1-' of the element. Serial dilutions were made with high-purity doubly distilled water in order to prepare working standards. A synthetic urine (a simulated urine solution) was prepared by dissolving 14.1 g of NaCl 2.8 g of KCl 17.3 g of urea 1.9 ml of ammonia solution (25%) 0.6 g of CaCl 0.4 g of MgS04 1 g of glycine 0.2 g of glucose 1 g of creatinine and 10 ml of HCl in 1 1 of doubly distilled water. This was the 'blank' solution.Validation of the methods described in this work was performed using Seronorm lyophilized human reference Table 1 Optimum operating conditions used with grid nebuliza- tion (Hildebrand dual Pt GN) of synthetic urine for MIP-AES using an argon plasma Torch type Forward power/W Microwave frequency/GHz Viewing mode Nebulizer flow rate/l min-' Solution uptake rate (pumped)/ Nebulizer pressurehar Grid spacing/mm Impact bead distance/mm Wavelength moni torednm ml min-* BN tube (75 x 6.3 mm 0.d. x 2.5 mm i.d.) 140 2.45 Axial 0.8 0.9 3.5 2 4 Cu 324.7 Fe 372.0 Zn 213.8 material (Trace Elements in Urine) obtained from Ny- comed As Diagnostics Oslo Norway. The certified elemen- tal values were obtained after reconstitution. Optimization Procedure Preliminary optimum operating conditions nebulizer gas flow rate and solution uptake rates for the GN were established as described by Hight and Rader." The spectrometer was peaked at the selected element line using emission from an appropriate electrodeless discharge lamp.Aerosols produced from solutions by the GN were transported to the plasma through the spray chamber by using argon as the carrier gas which at the same time served as the plasma supporting gas. The emission intensities from each solution (blank and then sample) were measured for 15 s. Net analyte emission was calculated by taking the sequential difference of measured emission intensities for sample and blank (synthetic urine). Optimization was performed by means of univariate searches using the signal- to-background noise as the optimization criterion.Optim- ized operating conditions are summarized in Table 1. Results and Discussion The study included an investigation of the suitability of the Hildebrand dual Pt GN for solution introduction into the MIP optimization of the nebulizer and plasma operating conditions and application of grid nebulization for direct urine analysis. Optimization of Operating Parameters Before any analyses were performed parameters governing the suitability of the GN for introduction of solution samples into the MIP source were studied. Parameters that were varied in these optimizations included sample intro- duction rate (as determined by the peristaltic pump speed) plasma gas flow rate nebulizer gas pressure position of impact bead and power applied by the microwave genera- tor.Initial experiments indicated that the copper 324.7 nm emission line could be used as a suitable reference to establish compromise conditions for multi-element deter- minations using solution sample nebulization. Conditions were optimized while nebulizing a 'synthetic urine' solution spiked with 1 pg ml-I of Cu. These observations were used to perform a univariate search in which four of the five parameters were held constant while the fifth parameter was varied. Simultaneous optimization of all operating condi- tions by simplex procedures was not undertaken in this initial investigation. The grid spacing was kept constant at 2 mm as the literature6 indicated that this spacing provided the most effective generation of a fine aerosol regardless of sample delivery rate.The parameters investigated wereJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 963 those expected to exert the greatest effect on the signal-to- background ratio (SIB). Plasma powers were investigated in the ranges 100-200 W. Argon flow rates and sample pumping rates were varied in the ranges 0.3-1.5 1 min-' and 0.4-1.2 ml min-l respectively. Also nebulizer gas pressure and position of impact bead were varied in the ranges 2.5-4.5 bar and 3-10 mm respectively. The opti- mum operating parameters for the plasma are shown in Table 1. Of the five parameters studied the nebulizer gas pressure and position of impact bead were found to be the most critical. The intensity of the copper 324.7 nm emission line increased only marginally when the rate of sample delivery was increased from 0.4 to 1.2 ml rnin-l suggesting that there is little or no analytical advantage in using sample delivery rates above 0.9 ml rnin-l.This value was therefore chosen for most routine analyses. Conceiva- bly higher delivery rates could overload the nebulizer if the inter-grid space became nearly completely filled with liquid. This condition could result in a loss in aerosol production as droplets of aerosol formed at the first grid encountered and coalesced with the bulk liquid. To reduce the thickness of bulk liquid at the point of aerosol formation would require greater nebulizer gas flow rates (and gas pressures much greater than 4 bar). With the optimum sample introduction system power and plasma gas flow rate were optimized.The general trend is that the SIB ratio increased with an increase in argon carrier flow rate and with moderate applied powers. Therefore for routine work an argon carrier flow rate of 0.8 1 m1n-I and forward power of 140 W were selected. This corresponds to a gas pressure of 3.5 bar. The role of the impact bead appears to be important. It should be emphasized here that studying the bead distance from the second nebulizer (downstream) grid (for pulse damping) is an obvious critical parameter to nebulization/transport efficiency. The optimum impact bead distance from the second (front) grid was 4 mm. This was almost the closest the bead could be positioned to the GN-spray chamber assembly. The furthest away the bead could be positioned was 10 mm.Signal intensity dropped off quickly as impact bead distance was increased beyond the optimum position of 4 mm. Thus the bead when positioned to give optimum signal is functioning as a nebulizer in its own right. No nebulization efficiency (transport efficiency) study was undertaken because this aspect has been dealt with by others6 The Hildebrand GN is 6.24% bulk efficient for distilled de-ionized water with a double-pass concentric glass spray chamber. It was concluded6 that the Hildebrand dual Pt GN gave superior performance over the concentric glass nebulizer and V-groove Babington-type nebulizer reflecting the higher nebulization/transport efficiency of this device. Although the transport efficiency reported6 is for a different system than the one used in this study i.e.a spray chamber with an impact bead will indeed result in a different tertiary aerosol than a double-pass concentric one it was however roughly assumed that the transport effici- ency trend would be approximately the same. A memory effect is one of the potential problems associated with this nebulization ~ystem.~.~JI The sample flowing between the dual Pt grids toward their centres comes in contact with a relatively large surface area. Under these conditions memory effects not encountered with the cross-flow might be expected. The GN affords the oppor- tunity to trap solution between the grids thereby prolong- ing the wash-out period. Several experiments were con- ducted to evaluate the magnitude of memory effects for this nebulizer including the associated spray chamber.Copper was used as the test element at a spike concentration of 1 pg ml-l using the synthetic urine solution. It was decided not to use an extended rinse time in the present study (when the rinse solution was pumped at the analytical rate or by using the 'rabbit switch' between samples to increase the rinse rate) because the time required to eliminate carryover for a real urine matrix was very long; after about 15 min the value was within 1-3 standard deviations of the initial blank value. In this study it was found that with the Hildebrand GN the signal decreased to 1% of its original value after 15 s by rinsing the nebulizer and spray chamber with the synthetic urine solution at 3 ml min-'.Therefore the Hildebrand GN exhibited virtually no memory effects with delivery of synthetic urine solutions (aerosol had been stabilized) but requires a much longer washout time (about 15 min) for solutions of real matrix content such as urine (not synthetic). Clearly aerosol mixing in the spray cham- ber controls the rate at which a steady emission signal is achieved. When nebulizing high dissolved solids content or high matrix solutions an important consideration is the effect of clogging. In no instance was blockage of the Hildebrand GN peristaltic tubing or torch encountered during uptake of either synthetic urine solution or urine samples. In addition the torch showed no significant build-up of solid residues or carbon during the measurement period and visual inspection showed that clogging of the nebulizer did not occur. However salt crystals were deposited in the spray chamber during prolonged sampling (continual nebu- lization for 1 h of the synthetic urine solution).Deposition was prevented by alternating nebulization with a 5 min 5% nitric acid wash cycle. Analysis of Urine Control Sample The urine matrix is quite complex. It contains many inorganic and organic compounds which produce a rich emission spectrum with many narrow and broad emission lines and cause spectral interference in the form of line overlaps and elevated background. Besides spectral inter- ference interference from ionization can also be expected. This is caused by the presence of species which are more easily ionized than the argon plasma gas.This may lead to suppression of the analyte signals as has been reported for the analysis of high salt content solutionsi3 and a complex sample rnatrix.l4 Considering the diversity and complexity of the urine composition it is reasonable to use background correction matrix matching (the simulated urine solution for calibration) and/or standard additions for quantifica- tion. Therefore the method of two standard additions to each sample was chosen and employed for calibration and it was necessary to apply background correction on one or both sides of the emission peaks for copper iron and zinc. To validate the grid nebulization MIP-atomic emission spectrometry (AES) technique solutions of certified com- mercially available urine controls (Seronorm Trace Ele- ments in Urine) were prepared for Cu Fe and Zn determinations.The calibration data based on the back- ground corrected emission intensity signals of the metals in spiked urine solution are shown in Table 2. All experimen- tal data fell within the accepted certified values. The precision obtained using the Hildebrand GN was good for the solutions analysed in this study. The overall precision of the method expressed as the relative standard deviation (RSD) of nine replicate measurements for three separate samples was 7- 1 I% acceptable for applications in biology and medicine. These results suggest that the GN is well suited to the analysis of samples containing high concentrations of dissolved solids and high matrix solution. Conclusions With the use of a grid-type nebulizer and the highly efficient resonant cavity direct introduction of urine samples into a medium-power argon MIP at atmospheric pressure can be964 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 Table 2 Grid nebulization MIP-AES analysis of urine controls (Seronorm Trace Elements in Urine) batch No. 108 RSD (Oh) Concentration*/pg I-’ Recommended valuei/ Element (n=9) (n=6) Pug I-‘ c u 10.8 50+ 5 45 Fe 8.9 7 0 k 6 72 Zn 6.8 700 f 48 640 *Mean k standard deviation. TMean value from the Nordic Quality Control Program. achieved with no desolvation. The use of the method of standard additions and background correction was essential to compensate for severe suppression of trace element intensity in urine solutions although calibration with matrix matching would be preferable for routine analyses.Sample preparation is entirely eliminated when urine samples are analysed. The major disadvantage in using the GN was the large volume of solution required to eliminate carryover from the previous analysis. The wash-out time was shortened by rinsing with the synthetic urine solution. Several future studies are suggested by the data including work using the grid-type nebulizer for direct sample introduction for application with slurries and with helium MIP-AES. Financial support by the State Committee for Scientific Research (KBN) Poland Grant No. PB6 18/P3/92/02 ‘Application of Microwave Techniques to Analytical Chemistry’ and by Deutscher Akademischer Austausch- dienst (DAAD) Fellowship while in Germany is gratefully acknowledged. The assistance of Professor G. Tolg and Dr P. Tschopel in obtaining the necessary equipment is also gratefully acknowledged. The experimental work was car- ried out at the Max Planck Institut fur Metallforschung Dortmund Germany. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 References Sample Introduction in Atomic Spectrometry Analytical Spec- troscopy Library ed. Sneddon J. Elsevier Amsterdam 1990. Brotherton T. Barnes B. Vela N. and Caruso J. J. Anal. At. Spectrom. 1987 2 389. Brotherton T. and Caruso J. J. Anal. At. Spectrom. 1987 2 695. Varnes A. W. J. Anal. At. Spectrom. 1988 3 803. Koropchak J. A. and Aryamanya-Mugisha H. J. Anal. ,4t. Spectrom. 1989 4 291. Smith T. R. and Denton M. B. Appl. Spectrosc. 1990,44,21. Brotherton T. J. Shen W. L. and Caruso J. A. J. Anal. At. Spectrom. 1989 4 39. Dymott T. Lab. Pract. 1991 40 79. Watling R. J. and Collier A. R. Analyst 1988 113 345. Laird D. A. Dowdy R. H. and Munter R. C. J. Anal. At. Spectrom. 1990 5 5 15. Hight S. C. and Rader J. I. Analyst 1991 116 1013. Matusiewicz H. Spectrochim. Acta Part B 1992 47 122 1. Jin Q Zhang H. Duan Y. Yu A. Ren Y. Zhang X. Lu H. and Yu S. Microchem. J. 1991 44 153. Brown P. G. Haas D. L. Workman J. M. Caruso J. A. and Fricke F. L. Anal. Chem. 1987 59 1433. Paper 3/0 1438.7 Received March 12 I993 Accepted June 2 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800961

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 96 5 Determination of Lead by Electrothermal Vaporization Microwave-induced Plasma Atomic Emission Spectrometry After Flow-through Electrolytic Deposition in a Graphite Tube Packed With Reticulated Vitreous Carbon Ernest Beinrohr Department of Analytical Chemistry Slovak Technical University CS-8 12 37 Bratisla va Slovakia Ewa Bulska University of Warsaw Department of Chemistry 02-093 Warsaw Poland Peter Tschopel and Gunther Tolg* Laboratorium fur Reinsstoffanaiytik Max-Pianck-institut fur Metaiiforschung Bunsen-Kirchhoff-Strasse 13 0-4600 Dortmund 1 Germany A simple method is described for the determination of lead in salt solutions by microwave-induced plasma atomic emission spectrometry (MIP-AES). Traces of Pb were preconcentrated electrochemically in a flow system incorporating a flow-through cell with a graphite tube packed with reticulated vitreous carbon serving as the working electrode.After washing the cell with water the tube was placed into the graphite furnace coupled to the capillary of an argon MIP operated in a surfatron. After the tube had been dried the plasma was ignited and the evaporation temperature was applied to the tube. The evaporated deposit was transported by argon gas into the plasma for excitation. A detection limit of 0.2 ng ml-' of Pb in 1 ml of preconcentrated sample solution was achieved. Keywords Electrothermal vaporization; microwave-induced plasma atomic emission spectrometry; lead determination; flo w-through electrochemical preconcentra tion Low power argon microwave-induced plasmas (Ar-MIPs) have become attractive as spectroscopic sources owing to their simple construction operation and low costs.How- ever owing to the low power the stability of the plasma deteriorates when even small amounts of sample material are introduced. Wet aerosols produced by nebulizers can only be introduced if the amount of the liquid is minimized through desolvation or the power applied to the plasma is increased. However nebulization of the sample solution has not proved to be fully satisfactory and gaseous sample introduction seems to be therefore the most efficient method. Electrothermal vaporization as a sample introduction technique for MIP has been described by several workers.' Recently a method for in situ trapping of volatile hydrides in a pre-heated graphite furnace followed by analyte evaporation and determination by MIP atomic emission spectrometry (MIP-AES) was demonstrated to be a power- ful analytical technique for hydride forming elements.* Some metal ions can be deposited electrolytically on a graphite tube and determined after evaporating and atom- izing the deposit simply by heating the tube.3 Flow electrolysis has proved to be an effective tool for analyte- matrix separation for some electroactive elements in electri- cally conducting solution^.^ In principle it enables the analyte species to be separated not only from the interfering matrix components but also from the solvent and therefore it might become an advantageous sample pre-treatment procedure for MIP spectroscopy. To achieve high electro- chemical recoveries while depositing from flowing solu- tions the mass transfer of the analyte species to the electrode surface should be as high as possible.Porous working electrodes made from graphite tubes packed with reticulated vitreous carbon (RVC) can fulfill this condition. The main goal of this work was to optimize the conditions for an effective electrolytic separation of the * Also at Institut fur Spektrochemie und angewandte Spektro- skopie Bunsen-Kirchhoff-Strasse I I D-4600 Dortmund 1 Germany. analyte by making use of the above electrode and to accommodate the graphite furnace assembly containing the packed tube to the Ar-MIP. The choice of analyte was of less importance. Experimental Reagents Reagents of analytical-grade purity and doubly distilled water were used.The electrodeposition of Pb was carried out in 0.1 mol 1-l potassium nitrate solutions with the pH adjusted to 4.8 with acetate buffer. Apparatus Spectrometer The spectrometer was equipped with a 0.5 m Ebert monochromator (Jarell-Ash) with a wavelength range of 200-450 nm. The grating was constant 112242 mm; width 54 mm reciprocal linear dispersion 0.8 nm mm-I and slit- width 30 pm. A photomultiplier EM1 978 1/A was used. The transient signals were recorded on a potentiometric recor- der (Beckman type 610 000). The monochromator was aligned on the Pb I 283.3 nm analyte line by using a Pb hollow cathode lamp source. Cavity A cylindrical TMolo cavity with a special mounting chuck for the capillary to produce a 3-filament (3-F) MIP was The surfatron used was of the design of Selby and Hieftje.6 Suprasil quartz capillaries (Hereaus Quartz- schmelze Germany) with 6 mm 0.d.and 4 mm i.d. were used. Microwave generator The microwave generator was an EMS-Microtron Mark 111 type EMS 6000 with a frequency of 2.45 GHz and maximum forward power of 200 W was used. The reflected966 (a) JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 Pt electrode Sample out 1 Sample in 1 " ' I Counter space Counter electrode Tube RVC lonex membrane Quartztube Tube RVC ( b ) Surfatron Furnace Fig. 1 Diagram of (a) the flow-through cell and (b) GF-MIP assembly power was adjusted to a minimum after ignition of the plasma. Electrothermal atomizer The CRA power supply Model 61 (Varian Techtron Mulgrave Australia) was used.The graphite tube posi- tioned between three graphite supporting rods was directly connected to the quartz capillary of the MIP.* Flow system and electrochemical cell The flow system consisted of a Gilson Minipuls 2 peristaltic pump an injection valve (LATEK Eppelheim Germany) and a three-electrode flow-through cell (Fig. 1) with a glassy carbon counter and Ag/AgCl reference electrodes as de- scribed re~ently.~ The graphite tube working electrode of 18 mm length 5 mm 0.d. and 3 mm i.d. was machined from a high density graphite rod (Quality RWO Ringsdorf Bonn- Bad Godesberg Germany). The tube was packed with RVC of 100 pores per linear inch porosity (Electrosynthesis East Amherst NY USA). A Jaissle 1000-B potentiostat was used for the electrolysis.The potentials throughout this paper are expressed versus the Ag/AgCl reference electrode used in the cell. Procedure The packed graphite tube was cleaned before use by heating at 2000 "C for 5 s in the atomizer while purging argon gas through it and after cooling it was fixed into the cell body. The cell was flushed with ethanol to remove air bubbles from the RVC filling. Doubly distilled water was then purged through the system. The deposition potential was set to the graphite tube and the sample solution with pH adjusted to 4.8 with acetate buffer was injected into the water flow. After the electrodeposition the cell was washed with water for another 2 min. The cell was disassembled and the tube was fixed between the supporting rods of the furnace (Fig.1). The tube was dried at a temperature of about 80 "C while purging argon gas through it. On evaporating the solvent the plasma was ignited and left to stabilize for about 1 min. The tube was then heated to 2000 "C with the highest temperature ramp in order to evaporate the deposit. The argon carrier gas purging through the tube introduced the sample vapour into the MIP and the signal Table 1 Furnace parameters Temperature/ Hold time/ Ar flow rate/ Step "C S 1 h-1 Drying 80 60 10 45 Plasma ignition Ambient - Atomization 2000 3 45 Cleaning > 2000 3 45 was recorded. Signal height was evaluated. The experimen- tal parameters of the furnace are listed in Table 1. Results and Discussion Electrochemical Properties of the Cell The electrochemical properties of the cell with the packed graphite tube working electrode were tested with potassium hexacyanoferrate(r1) solutions in 1 mol 1-I potassium chloride supporting electrolyte.Electrochemical conversion efficiencies of over 90% were observed for flow rates of up to 1 ml min-l. At flow rates of 3-15 ml min-' the conversion efficiency decreased to 70-30%. According to the hydrodynamic voltammograms7 the cell exhibited Nernstian behaviour up to flow rates of 1.5 ml min-' with an effective charge number of about 0.95. After heating the packed graphite tube working electrode in the atomizer at 2000 "C for 10 s no significant deterioration in the electrochemical properties of the electrode was observed. Only the background currents were higher than those with tubes containing untreated RVC material.The packed tubes could be used 20-30 times without any deterioration. Optimization of the Plasma Preliminary experiments in this work were performed with an Ar-MIP sustained in the Beenakker-type TMolo cavity. However the plasma was found to be unstable during the evaporation step. The evaporated species probably some residues of the sample matrix not completely removed from the graphite tube during the washing procedure caused severe interferences and disturbed or even extinguished the plasma. However no significant improvement of the plasma stability was observed even if the tube was rinsed for 10-20 min with water. The sample matrix penetrating inside the tube wall probably could not be completely removed and evaporated at the heating step and thus caused interference.A surfatron producing a more stable plasma was used instead. As has been described previously the MIP gener- ated in a surfatron is easier to operate and to tune and offers better detection limits compared with the TMoro cavity.8 Moreover the plasma discharge is more tolerant to interfer- ences caused by easily ionized element~.~*I~ The different plasma configurations in a surfatron can be obtained under a wide range of operating conditions such as gas flow rate forward power and capillary position.* Of all plasma forms investigated the best detection limits were obtained with the 3-F MIP which was therefore used in further experi- ments. The plasma was stable during the evaporation step unless the contact between the cell and the plasma capillary was tight.However tubes used more than 20-30 times could not be fixed sufficiently firmly thereafter owing to some burning of the tube ends during the atomization step. The operating conditions of the 3-F MIP were optimized with the electrothermal atomizer connected to the plasma capillary. Sample aliquots of 5 pl were injected onto the RVC filling inside the tube. The sample was dried evaporated and transported into the plasma with the carrierJOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY. OCTOBER 1993 VOL. 8 967 10 c 2t rA I O L 1 I 1 1 1 400 800 1200 1600 2000 Tern perat u rePC Fig. 2 Dependence of the signal response on the evaporation temperature for A 5 ng of Pb injected directly into the graphite tube and B 1 ml of 5 ng m1-I Pb sample in 0.2 mol 1-1 Na,SO preconcentrated at -2 V and at a flow rate of 1 ml min-I I I 1 1 1 1 I 1 0 2 4 6 8 10 12 14 Flow rate/ml min-' Fig.3 Dependence of the signal response on the flow rate during deposition sample volume 1 ml; Pb concentration 8.3 ng ml-I in 0.1 mol I-' KNO at pH 4.8; and deposition potential - 3 V ~~ ~~ ~~~ ~ Table 2 Working parameters for the 3-F Ar-MIP in a surfatron Parameter Range Optimum Forward power/W 50- 1 50 120 Gas flow ratell h-' 15-70 45 Observation zone*/mm 0-2.0 1 .o * Distance from the centre of the capillary (centre at 0 mm). gas. The optimum working parameters found for the 3-F Ar-MIP are listed in Table 2. Fig. 2 illustrates the dependence of the signal response on the heating temperature for Pb samples directly injected into the tube and for those electrolytically preconcentrated. The optimum temperature range was found to be between 1700 and 2000 "C.Higher temperatures than this decreased the lifetime of the tubes owing to their unsatisfactory shielding with the nitrogen gas. In the case of a surfatron sodium or potassium ions present in the sample solution were found not to interfere with the plasma discharge and therefore it was possible to determine Pb also by direct injection of 5-10 p1 of sample solution into the packed tube. Table 3 Detection limits of Pb with various plasma sources Technique* PN-MIP PN-ICP GF-MIP GF-ICP GF-T-MIP GF-MIP (surf)? GF-MIP (surf)? A/ nm 283.3 405.8 405.8 405.7 405.7 283.3 283.3 Detection limit/ ngIm1-I 380 272 50 130 10 40 0.2 Ref.This work 12 13 13 14 This work This work * PN=pneumatic nebulization; GF=graphite furnace; T= tor- t Direct injection of 5 jd of sample to the furnace. $ 1 ml of sample solution preconcentrated. oidal; and surf surfatron. Electrodeposition and Determination The roles of the deposition potential and the flow rate in the preconcentration efficiency were investigated. The flow rate significantly influences the deposition recovery and hence the signal sensitivity. The highest sensitivities were ob- served at flow rates below 2 ml min-' (Fig. 3). In order to have reasonable preconcentration times also for sample volumes of more than 1 ml a flow rate of 1 ml min-' was chosen as a compromise. However only a precise control of the flow rate can ensure reproducible deposition yields and analytical results.As is implied in Fig. 4 a deposition potential of -3 to - 4 V ensures the highest signal sensitivity presumably owing to higher electrodeposition efficiencies under these conditions. However these potentials are much more negative than expected from voltammetric and coulometric measurements provided in deoxygenated solutions with mercury-coated electrodes.Il The presence of oxygen in the sample solutions and the use of an electrode with a carbon surface without mercury coating could be a plausible explanation for these observations. The calibration curve method was used for calibration. Standard solutions of Pb in 0.1 moll-' potassium nitrate at pH 4.8 were treated by the proposed procedure. A linear calibration response was observed up to aboui 10 ng of Pb in the injected standard solutions.The parameters of the calibration curve were as follows intercept -0.3; slope 18.0 ng-I; and correlation coefficient 0.9997. The blank values corresponded t o 0.5 ng ml-1 of Pb and the absolute detection limit (3s level) for Pb was found to be968 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 ~ ~~ Table 4 Analyses of salt solutions Pb foundng ml-I Sample solution Sample volume/ ml Proposed method ASCWC* 0.5 mol 1-l Na2S0 5.0 0.27 4 0.02-f 0.24 f 0.03 0.1 mol 1-l KN03 1 .o t 0 . 2 0.15 f 0.03 5.0 0.18 f 0.031- - * Anodic stripping coulometry with collection 2.4 ml of sample solution ana1y~ed.l~ t Significance level=0.05; n= 7. 0.2 ng. It should be pointed out that many factors contribute to the detection limit attainable with an MIP system.The nature of the matrix operating parameters type of support- ing gas and flow rate and applied microwave power all influence the signal sensitivity and repeatability. However of major importance is the kind of sample introduction into the plasma. The best detection limits are obtained with gaseous sample introduction following electrothermal eva- poration. The preconcentration of the sample significantly improves the detection limit owing to the enrichment of the analyte and the removal of the interfering matrix (Table 3). The precision of the results expressed as the relative standard deviation (RSD) for three complete analyses with different tubes for 1-5 ml sample solutions containing 3- 13 ng ml-I of Pb was found to be 2-4%. A complete analysis including sample preconcentration and determination by MIP-AES took 5-15 min depending on the volume of sample taken for preconcentration. The proposed method was tested by analysing sodium sulfate and potassium nitrate solutions.The accuracy of the method was checked by anodic stripping coulometry with collection.1s The results are listed in Table 4. Interferences owing to solvent and matrix components such as easily ionizable elements can be significantly suppressed by this technique. However some matrix constituents undergoing electrochemical reduction can cause problems arising from decreased electrolysis yields of the analyte and/or interferences as a result of some electrodeposition of major matrix components.E. B. thanks the Alexander von Humboldt-Foundation in Bonn for the financial support enabling this research to be carried out and E. Bulska thanks the Max-Planck-Gessel- schaft (Miinchen Germany) for the grant of a research fell0 ws h i p . References 1 Matusiewicz H. Spectrochim. Acta. Rev. 1990 13 47. 2 Bulska E. Broekaert J. A. C. Tschopel P. and Tolg G. Anal. Chim. Acta 1993 271 171. 3 Beinrohr E. Lee M. L. Tschopel P. and Tolg G. Fresenius’ J. Anal. Chem. 1993 346 689. 4 Beinrohr E. Fresenius’ J. Anal. Chem. 1990 338 735. 5 Kollotzek D. Tschopel P. and Tolg G. Spectrochim. Acta Part B 1984 39 625. 6 Selby M. and Hieftje G. M. Spectrochim. Acta Part B 1987 42 285. 7 Curran D. J. and Tougas T. P. Anal. Chem. 1984 56 672. 8 Bulska E. Broekaert J. A. C. Tschopel P. and Tolg G. Anal. Chim. Acta in the press. 9 Galante L. J. Selby M. and Hieftje G. M. Appl. Spectrosc. 1988 42 559. 10 Richts U. Broekaert J. 4 . C. Tschopel P. and Tolg G. Talanta 199 1 38 863. 1 1 Beinrohr E. Nemeth M. Tschopel P. and Tolg G. Fresenius’J. Anal. Chem. 1992 343 566. 12 Winge R. K. Peterson V. J. and Fassel V. A. Appl. Spectrosc. 1978 33 206. 13 Aziz A. Broekaert J. A. C. and Leis F. Spectrochim. Acta Part B 1982 37 381. 14 Heltai G. Broekaert J. A. C. Burba P. and Leis F. Spectrochim. Acta Part B 1990 45 857. 15 Beinrohr E. Nemeth M. Tschopel P. and Tolg G. Fresenius’ J. Anal. Chem. 1992 344 93. Paper 3/01 92lE Received April 5 1993 Accepted June I 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800965

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 969 Diagnostic Investigations of Aerosols with Varying Water Content in Inductively Coupled Plasma Mass Spectrometry* Norbert Jakubowski lngo Feldmann and Dietmar Stuewer lnstitut fur Spektrochemie und ange wandte Spektroskopie Postfach 10 13 52 0-440 13 Dortmund Germany Measurements of axial and radial ion intensity distributions and also of ion energy characteristics by a simple bias potential technique giving a representation of the kinetic energy distributions have been applied to study the influence of water vapour on the behaviour of aerosols in inductively coupled plasma mass spectrometry. The full range of moisture content has been covered from dry aerosols generated by solid evaporation to wet aerosols formed by the addition of an increasing amount of water vapour as well as vice versa from pneumatically generated wet aerosols to dry aerosols by applying dehumidification techniques.As an alternative to humidification of a dry aerosol hydrogen addition was considered. Comparison of results was performed on the basis of identical values for power and nebulizer flow rate while sampling distance and bias potential were always individually optimized. A strong dependence on the content of water or hydrogen was observed offering a promising opportunity to influence the analytical performance while providing the basis for a consistent interpretation of the results. Exploitation of the method in analysis requires that sampling distance and bias potential be taken into account in optimization procedures in addition to nebulizer flow rate and power to which optimization procedures have been restricted in most applications so far. Keywords Inductively coupled plasma mass spectrometry; aerosol; water content; bias potential; sampling distance Increasing acceptance demonstrates the merits of induc- tively coupled plasma mass spectrometry (ICP-MS) for trace and ultratrace analysis but also forces research to improve elemental sensitivities and detection limits and to overcome inherent limitations in particular spectral and non-spectral interferences.A promising and in many cases successful approach is the application of new and perhaps unconventional sample introduction systems pro- ducing different types of aerosols. Actually the lowest detection limits for microanalysis are realized by dry aerosols as for instance produced by electrothermal vapori- zation (ETV)’ or laser ablation.2 For the direct analysis of conducting solids spark ablation might become a low-cost alternative to laser a b l a t i ~ n .~ Several techniques are prom- ising for wet aerosols ultrasonic neb~lization,~ hydraulic high pressure nebulization (HHPN)5 and direct injection nebulization.6 However in the case of these nebulization techniques optimum performance can only be achieved by reduction of the water content because otherwise the plasma cannot withstand the high water intake from the nebulizer. This is the reason why desolvation as is common in ICP atomic emission spectrometry (ICP- AES),7 is of increasing interest in ICP-MS. Simultaneously desolvation can be helpful in improving detection limits and to reduce interferences that can be caused by solvents or chemical pre-treatment. Optimum values for the analyt- ical figures of merit are to be expected for totally dry aerosols but a complete dehumidification could not be realized with conventional desolvation systems.Therefore application of cryo-cooling systems8 or permeable foils9 have been considered recently to be more effective for dehumidification. In previous work different sample introduction tech- niques were investigated producing dry aerosols by ETVO and spark ablation3 and wet aerosols by pneumatic nebuli- zation (PN) and by HHPN5 including desolvation the * Presented in part as the Alan Date Memorial Award lecture at the 4th Surrey Conference on Plasma Source Mass Spectrometry July 199 1 ; preliminary results published in J.Anal. At. Spectrom. 1991 6 249. effect of which has also been studied for On the basis of this work a comprehensive study of aerosols with systematic variation of the water content in ICP-MS has now been performed by application of different diagnostic techniques covering the full scale from dry to wet aerosols by humidification as well as from wet to dry aerosols by dehumidification. It can be expected that the same pheno- mena will be observed although in reverse direction for both humidification and dehumidification. However it is necessary to be aware that when applying desolvation the dehumidification efficiency will be limited by the water vapour pressure so that the full range can only be covered in the reverse experiment i.e.with humidification of an initially dry aerosol. The influence of moisture has already been studied extensively in ICP-AES. Addition of water vapour exerts a direct influence on the physical properties of the plasma mainly electron density excitation temperatures and ioni- zation temperature. 1 2 ~ 1 3 ~ 1 4 9 1 5 Nowak et al.” observed shrink- ing of the plasma when water was added. In a wet aerosol the maximum electron density of the plasma is higher but electron density decreases more rapidly with increasing observation height along the axis in comparison to a dry aerosol. A plausible explanation could be improved heat transfer in the plasma owing to the presence of hydrogen generated by dissociation of water m o l e c u l e ~ .~ ~ J ~ On the other hand reduction of water from an aerosol generated by a pneumatic nebulizer results in increasing spectral intensi- ties17J8 and an axial shift of the intensity maxima.14 This demonstrates that control of the water intake to the plasma provides the opportunity to exert considerable influence on physical properties which can be exploited in favour of the analytical figures of merit. Aerosol conditioning was considered for application in ICP-MS at an early stage in the development of the te~hnique.l~-*~ Water reduction was investigated in ICP-MS mainly with the aim that it might be an efficient means of reducing spectral interferences. As a consequence some basic work has been performed concerning particular aspects of However these studies do not give a comprehensive overview nor can the experience gained in ICP-AES be transferred to ICP-MS.In compari-970 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 son with ICP-AES any change of the plasma conditions will give rise to more significant consequences in ICP-MS because ion sampling is not averaged over larger areas and thus offers considerably higher local resolution. Further- more ion energies are an important parameter in ICP-MS which can be influenced by any changes of the plasma conditions. A change of ion energies has been observed in the application of desolvation by different ~ o r k e r s . ~ l - ~ ~ In practical operation the water intake to the plasma can be controlled by cooling of the spray chamber or application of a desolvation system.The present investigations were enabled by the special features of the laboratory-built ICP-MS system which has been described previou~ly.~~ This system can serve not only for spectrometrical separation and registration but was especially designed to also provide diagnostic measurement techniques. Full x-y-z-positioning capability enables mea- surement of axial and also radial ion intensity distributions. These contain not only valuable diagnostic information but can also be utilized to determine the optimum value for the sampling distance (dJ i.e. the distance between the top of the load coil and the entrance of the sampling orifice. The second special feature is the response of the whole ion optical arrangement to a variable reference bias potential.As result of this simple technique the so-called ion energy characteristic (IEC) is obtained which represents the ion kinetic energy distribution as has been shown in previous investigation^.^^ In addition the bias potential technique25 provides not only valuable diagnostic information but can also be used to optimize the ion transmission by the choice of the optimum bias potential (U,,). Experimental For these investigations the laboratory-built ICP-MS sys- tem with a specific 'long' torch was used. The distance from the injector orifice to the end of the torch is 25 mm and the diameter of the components are injector tube 1 mm; inner tube 13 mm; and outer tube 19 mm. In normal operation an outer gas flow rate of 20 1 min-' was used; no intermediate gas was necessary.An aerosol generation system was assembled providing sample introduction by ETV as well as by PN and continuous control of moisture content by drying or moistening. This enables the compari- son of different aerosol generation systems with otherwise constant operational conditions. A schematic diagram of the experimental arrangement is shown in Fig. 1. I Fig. 1 Schematic diagram of experimental arrangement (a) solid vaporizer; (b) electrothermal vaporizer; (c) pneumatic nebulizer with cooled spray chamber; (d) water vaporizer; (e) desolvation system; and cf) ICP-MS system For generation of a wet aerosol a GMK type Babington pneumatic nebulizer has been used for which a nebulizer flow rate (NFR) of 1 .O 1 min-' gives optimum operation.A sample solution with a concentration of 2.5 pg ml-' of Fe is nebulized with a sampfe uptake rate of 1 ml min-'. The resulting water intake to the plasma at room temperature without cooling of the spray chamber was 300 pg s-' corresponding to an efficiency of 1.8% as was determined by the well known silica gel trap technique. Cooling of the spray chamber enables lowering of the temperature down to + 1 "C with a resulting reduction of the water intake rate down to 170 pg s-l. Furthermore a desolvation system enabling a lower limit of water intake of about 60 pg s-l has been described e1sewhere.I Two different evaporation systems are used for the generation of dry aerosols. A simple glass flask with a volume of about 10 ml serves to produce a dry aerosol by evaporating ferrocene [bis( q-cyclopentadienyl)iron(xx)] an organic Fe compound with a very low boiling-point (249 "C).This Fe compound can be vaporized using temperatures up to the boiling-point without dissociation. For these investigations operation at room temperature was sufficient to achieve a stable and reproducible iron intake of 80 ng s-I measured by collecting the evaporated material in 4.5 moll-I HN03 with subsequent determination by ICP-MS. For ETV of samples with high boiling-points a tungsten wire loop is provided which corresponds to a system developed for atomic absorption spectrometry (AAS).26 In the course of this work different techniques were considered for humdification and dehumidification experi- ments.The first idea was to use the pneumatic nebulizer itself for control of the moisture content. However in combination with a cooled spray chamber the accessible interval was restricted to values of water intake rate from 170 up to 300 pg s-'. Therefore a desolvation unit was also included by which the accessible range could be extended down to 60 pg S - I . Below this limit humidification experiments could only be performed by applying a simple low volume vaporizer. By variation of the heating tempera- ture a water intake rate ranging from 20 ,ug s-' at room temperature up to 212 pg s-l at 90 "C could be achieved. In order to remove the influences from changes in potential as a result of varying water content the bias potential was adjusted to give maximum transmission in all measurements of radial or axial ion intensity distributions which requires extremely careful alignment of the sampling orifice in the optical axis.The dynode multiplier was operated in an analogue mode throughout these investiga- tions in order to register ion intensity distributions and ion energy characteristics as the direct output of an x-y- recorder. Results and Discussion Operational Optimization Investigations with the aim of comparing different types of aerosols are complicated by the fact that in order to achieve optimum analytical performance each aerosol requires its own operational optimization. For operational parameters in ICP-MS it is necessary to distinguish between two types of parameter. The conventionally optimized parameters power and nebulizer flow rate exert a direct influence on atomization and ionization in the plasma and from this point of view they are 'ion generation' parameters.The d as well as U and any ion optical lens setting influence the ion transfer from the plasma through the sampling orifice and skimmer to the mass analyser but not the ion generation processes in the plasma; thus they are 'ion transfer' parameters as opposed to ion generation parameters. In the ICP-MS work published so far the role of the generationJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 97 1 0' I I I I I I I I I 6 7 8 9 10 11 12 13 14 15 16 Sampling distance/mm Fig. 2 Axial intensity distributions of Fe+ ions with different NFR A 0.9; B 1.0; C 1 . 1 ; and D 1.2 1 rnin-'.Forward power 1 kW and Ub=5 V parameters has been studied much more carefully than have the transfer parameters of which only d has been consi- dered while little or no attention has been paid to the bias p~tential.*~J~+*~ With this difference between generation and transfer parameters in mind it should be borne in mind that any change in the generation parameters requires at least subsequent optimization of the transfer parameters if real analytical optimization is required. For comparability of experimental results obtained with different aerosols or sample introduction systems it is always necessary to ensure physical similarity between them and therefore the generation parameters have to be kept constant in such experiments while the transfer parameters must be optim- ized individually.On the basis of these considerations the investigations commenced with a series of experiments in order to obtain a general overview of the mutual interde- pendence of the operational parameters to be taken into account. For variation of the NFR Fig. 2 shows the axial intensity distribution obtained with a solution containing 2.5 pg ml-I of Fe and with an incident power of 1 kW. The optimum d is shifted from below 6 mm to above 10 mm with the NFR ranging from 0.9 to 1.2 1 min-I. This shift of the ion intensity maximum to a greater d with increasing NFR can be ascribed to a higher mass flow rate in the aerosol. An increase of the NFR from 1.0 to 1.1 1 min-l leads to a shift of the maximum by about 3 mm demonstrating that the optimum ds is highly sensitive to minor changes of the NFR.The intensity is only slightly dependent on the NFR for values below 1.1 1 min-l but a considerable loss can be observed for higher values. A possible explanation is plasma cooling with high flow rates leading to the suggestion that the loss could be balanced by an increase of the forward power which will therefore be the next topic to be investigated. Concerning the influence of the incident power as the second ion generation parameter Fig. 3 shows correspond- ing axial intensity distributions of Fe+ ions obtained for four different values of the power with the NFR kept constant at 1.2 1 min-I for which a significant loss in sensitivity was observed in Fig. 2. However as demon- strated by the results in Fig.3 this can be compensated for as can any other value of the NFR if a sufficiently high power setting is chosen. An increase of the power enhances the efficiency of atomization and ionization so that the intensity maximum appears at a d that is about 1 mm lower for a power increase of 0.1 kW while the maximum intensity remains nearly constant. This demonstrates also that a change of the power requires adjustment of the d if - 4000 Y .- C 2 2 3000 c .- - $ 2 2000 .- u) Q) CI - 1000 0 ' ' ' 1 1 1 I l l 1 5 6 7 8 9 10 1 1 12 13 14 15 16 Sa m pl i ng d i stance/m m Fig. 3 Axial intensity distributions of Fe+ ions with different forward powers A I . 1; B I .2; C 1.3; and D 1.4 kW. NFR= 1.2 1 min-' -2.0 -1.0 0 1 .o 270 l 4 Radial posit ion/m m Fig. 4 Spatial intensity distributions of Fe+ ions for a slightly moistened aerosol; z=axial distance from edge of load coil maximum sensitivity is to be maintained which can be realized in the combined optimization of both.From these measurements the axial ion intensity distri- bution and in particular the optimum value for the d is found to be very sensitive to the NFR as well as to the incident power. Nevertheless not very much attention has been paid to optimization of the d in the routine applica- tion of ICP-MS so far. This can only be partially attributed to the fact that this type of optimization is a tedious procedure with commercial instruments in their current state of development. It could also be because most people working with ICP-MS have entered the field from applica- tion of ICP-AES where the influence of the ds (or observa- tion height as it is usually called in AES) is significantly lower.However it should be realized that ICP-MS provides a rather high local r e s o l ~ t i o n ~ ~ which in the case of the instrument used here amounts to about 0.5 mm whereas such effects are mainly averaged out with an emission spectrometer which registers radiation from a wider volume of the plasma. The higher resolution of ICP-MS offers the opportunity to get a representation of the spatial ion intensity distribu- tion in the region of the aerosol channel as demonstrated in Fig. 4. Measurements were performed at different d values varying with an increment of 1 mm. For each d the radial intensity distribution was measured with an increment of 0.2 mm.With respect to these measurements it is necessary to be aware that in comparison to AES local differences in the plasma become much more apparent in ICP-MS.972 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 5 I 1 .- $ 4 C =I $ 3 0 -10 -5 0 5 10 15 Bias potentialN Fig. 5 Energy characteristics of Fe+ ions with different NFR A 0.5; B 0.55; C 0.6; D 0.7; and E 0.8 I min-'. Forward power 1 kW and d,=7 mm In comparison to ICP-AES the situation is more compli- cated in ICP-MS because MS deals with ions and thus influences from different ion kinetic energies must also be taken into account. This is demonstrated by Fig. 5 showing IEC measurements for Fe+ ions obtained by PN as above with different values of the NFR. With increasing NFR the distribution maximum increases in intensity and is shifted towards the upper limit of the potential.This can easily be explained by potential distribution measurements which were performed earlier using a tungsten wire probe.25 In the aerosol channel of a wet aerosol a positive potential exists while in the surrounding plasma the potential is negative. For low values of the NFR the aerosol channel does not break through the outer plasma and the ion intensity maximum appears at a negative bias potential. Above a certain threshold of the NFR ions are sampled from the aerosol channel itself the positive potential of which is reflected in the IEC measurement. This demonstrates that IEC measurement by application of the bias potential technique is a simple and powerful means not only of optimizing transmission and the resulting sensitivity effec- tively but also of obtaining information concerning the plasma potential at which ions are mainly generated.More correctly differences in the bias voltages corresponding to the intensity maxima reflect the differences in the local potentials at which ions are generated.28 For higher values of the NFR there is no further increase of the bias potential which is in agreement with the idea of a fully developed aerosol channel. However a significant increase in the maximum intensity was observed which reflects once more the aforementioned effect of a shift in the axial intensity distribution achieved by increasing NFR. As a consequence of these results NFR and incident power as ion generation parameters were kept constant for all measurements that were used for a direct comparison of different aerosols in these investigations while bias poten- tial and d as ion transfer parameters were always optimized individually.Wet and Dry Aerosols The measurements in Fig. 5 have already demonstrated that the IEC maximum appears at a positive potential in the case of a wet aerosol. A comparison of a pneumatically generated wet aerosol and a dry aerosol generated by solid evaporation is shown by the IEC measurements presented in Fig. 6(a) and (6). In addition to Fe+ the intensity distribution of Ar+ is also shown. For the dry aerosol [Fig. 6(a)] the ion intensity maxima occur at negative values of the bias potential whereas for a pneumatically generated 2000 (a) c .- a Y 600 - 400 200 > ' ' ' ' ' \ ' I I 8 8 I ' I 8 ' 8 "Fe' ' 8 0' I I I I I 1 I -20 -15 -10 -5 0 5 10 15 20 Bias potentialN Fig.6 Energy characteristics for different aerosols (a) dry aerosol generated by solid vaporization d,(Fe)= 7 mm and ds(Ar)=9.5 mm; and (6) pneumatically generated wet aerosol d,(Fe)= 10 mm and d,(Ar)= 18 mm. Forward power 1 kW and NFR= 1.2 1 min-I wet aerosol [Fig. 6(6)] the IEC maximum is found at about + 6 V in the case of Fe+. This makes it obvious that the presence of water significantly influences the plasma in the aerosol channel. Additional information can be extracted from measure- ments of radial intensity distributions for both types of aerosol that are presented in Figs. 7 and 8. The figures show the distributions of the ion species Fe+ Ar+ and ArH+ obtained with a dry (Fig.7) and a wet (Fig. 8) aerosol for two characteristic values of the d each and the appertaining bias potentials. In the case of the dry aerosol a loss of Ar ions occurs in the central aerosol channel which has a diameter of about 1.5 mm. The distribution of ArH+ is nearly uniform over the plasma with peaks at the outer edge while the total intensity increases with the sampling distance. For the wet aerosol a moderate intensity maxi- mum exists for Ar+ and a more pronounced maximum for its hydride in the region of the aerosol channel the intensity of which decreases with increasing d,. This can be inter- preted by the different sources of the hydrogen supply for the plasma. In a dry aerosol the only hydrogen source is the surrounding atmosphere; in a wet aerosol the dissociation of water molecules is an additional hydrogen source that becomes predominant in the aerosol channel and causes the appearance of an additional hydride maximum in this region.Comparison of the radial distributions for Ar+ ions in Figs. 7 and 8 demonstrates shrinking of the plasma owing to the presence of water; the diameter of the argon plasma is about 2 mm greater in the case of the dry aerosol. This effect has also been reported for ICP-AES13 and corre- sponds to the observation that the reflected power is changed by the introduction of water.22973 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 I;, y r + 57Fe+ I I ,.I .. p.7,-,- % 0/-- . _ '\ .j ; /'..... +.. ; .. ........... ,>*. .... ................. 2 ........ \ I 0 -8 -6 -4 -2 0 2 4 6 8 Radial position/mm Fig. 7 Radial intensity distributions of three ion species from a dry aerosol for (a) ds=8 mm and (b) ds= 1 1 mm; &= -8.5 V and intensity of ArH+ is magnified 20-fold 8 6 4 A a 2 C .- C > L .- S O n L 2 8 . . 57Fer . . . . . . . . ....................... I ""';' 0 -8 -6 -4 -2 0 2 4 6 8 " I I I Radial position/mm Fig. 8 Radial intensity distributions of three ion species from a wet aerosol for (a) d,= 9 mm and (b) d,= 13 mm; &,= 10.5 V Humidified Aerosols In order to obtain a more detailed insight into the effects that can be attributed to the presence of water in an aerosol humidification of dry aerosols as well as dehumidification A v) C .- 5 10000 2 I f C .- - > C a 5000 C - 0 B -C A D E 5 7 9 11 13 15 17 19 Sampling distance/m m Fig.9 Axial intensity distributions of Fe+ ions for aerosols with varying moisture content A dry aerosol generated by solid vaporization &,= - 12 V; B slightly moistened aerosol water intake 19 pg s-I &= - 6 V; C moistened aerosol water intake 40 pg s-l &= -4 V; D moistened aerosol water intake 84 pg s-* &,= -2 V; and E moistened aerosol water intake 176 pg s-l &= 0 V. Forward power 1 kW and NFR= 1.2 1 min-' of wet aerosols was studied so that the transition between the two extremes in both directions could be observed. As the first step of this programme for generation of humidi- fied aerosols the dry aerosol of the solid vaporizer with ferrocene was used and was moistened afterwards with an increasing water load. Variation of the water intake rate was achieved by increasing the temperature of the water supply.In the measurements of the axial ion intensity distributions the bias potential was always optimized individually while typical ion generation parameters of a wet aerosol were preserved throughout the series of experiments. A value of 1.2 1 min-' was chosen for the NFR as before which represents a compromise between the optimum value of 1 .O 1 min-' for the wet aerosol and 1.4 1 min-l for the dry aerosol and keeps the ionization maximum in the observa- tion range. Axial ion intensity distributions for varying moisture addition are shown in Fig. 9. For the dry aerosol (A) the maximum ion intensity obviously appears very close to the top of the load coil at the value of d which is below the accessible measurement range of the system.Humidifica- tion by water vapour with a low intake rate (B) shifts the maximum of the ion intensity towards a higher d9 Further increase of the water intake (C D E) causes an additional axial shift of the ionization maximum and this is accom- panied by a loss of sensitivity. A possible explanation is that an increasing part of the energy in the plasma is used for dissociation of water molecules with the consequence that the ionization maximum appears later in the gas flow i.e. at an increasing d,. However this cooling of the plasma can be compensated for by increasing the power in the same way as already discussed. The final distribution (Fig.9 E) is fairly similar to that of a pneumatically generated wet aerosol for which the ion intensity maximum appears at a d of about 13 mm for the chosen NFR of 1.2 1 min-I. Summarizing the observed phenomena the appearance of the plasma is illustrated in Fig. 10. In comparison to the dry aerosol [Fig. 1 O(a)] two effects concerning the outward shape of the plasma are combined in the case of the humidified aerosol [Fig. 1 O(b)] radial shrinking as men- tioned in the last paragraph and axial shift of the back edge from the injector in the direction of the load coil and correspondingly of the intensity maximum which amounts to nearly 7 mm in the case demonstrated. It is obvious from the results discussed that erroneous974 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 Fig. 10 Sketch of plasma shrinking as observed by visual inspection; the arrow indicates the intensity maximum. (a) dry aerosol; and (6) dry aerosol with slight humidification 8000 I I y 6000 .- C 3 z L I -20 -15 -10 -5 0 5 10 0 Bias potentialN Fig. 11 Energy characteristics of Fe+ ions with humidification by a water vaporizer at different temperatures ("C) optimized sampling distance (mm) and corresponding water intake (pg s-l) derived from measurements with the silica gel trap technique A dry 6.0 mm; B 25 "C 7.5 mm 19 pg s-'; C 45 "C 9.5 mm 40 pg s-l; D 65 "C 11.5 mm 84 pg s-l; and E 85 "C 14 mm 176 pg s-l. Forward power 1 .O kW and NFR= 1.2 1 min-' conclusions can be drawn if only measurements without optimization of the d are compared. Taking only into account for instance the intensity values for a d of 12 mm from Fig.9 the conclusion would be that humidification is an effective means of achieving a considerable intensity gain or else that dehumidification of a wet aerosol yields a loss of intensity. For commercial instruments ds is in most cases fixed at a rather high value and this is the reason why a higher NFR was required in many investiga- tions for optimum sensitivity with dry aerosols and why maximum intensity was obtained with a lower power. This must be ascribed to the shift of the ionization maximum because both an increase in the NFR and a decrease in the power cause a shift of the ionization maximum towards higher d values as can be seen from the measurements made here. As for ds a shift appears also for the bias potential as can be seen from the optimized values given in the caption of Fig.9. This is demonstrated in more detail in Fig. 11 showing IECs obtained in a corresponding series of mea- surements. A slight humidification by switching the water vaporizer to the exit of the solid vaporizer system at room temperature results in a small increase in V accompanied by a considerable increase in the ion intensity. The latter must be ascribed mainly to the fact that the ionization maximum was still outside the observation range in the case of the dry aerosol. A further increase in the water intake by increasing the temperature of the vaporizer leads to a decrease in the intensity while V is shifted more and more to positive values.The strong influence of the water content on U gives a clear indication that the bias potential technique should also be included in optimization if optimum analytical performance is really desired. The observation of a strongly positive potential for a pneumatically generated wet aerosol is in agreement with results reported by different groups.21*24*31 In some recent paper^,^^-^^ Hieftje and co-workers have shown that indeed positive potentials are generated in a plasma sheath at the sampling orifice and as a consequence in the interface region if a grounded wall comes into contact with the plasma which is ascribed to the different mobilities of ions and electrons. Electrons diffuse faster to the grounded wall so that a sheath region close to the wall is impoverished of electrons resulting in a positive sheath potential.By Langmuir probe measurements these workers could show that the region of the positive potential extends up to the interface. These potentials are very sensitive to alterations of the sample introduction system and depend strongly on the water intake to the plasma. Such changes of potentials and ion energies by changing solvent load are more significant for the analytical performance than are gas- kinetic processes35 or different coil configuration^.^^ They depend strongly on plasma interactions in the interface region. In the case of the instrument used here such effects may be more pronounced than for commercial instruments. In any case careful control of the relevant operational parameters (e.g. solvent load NFR pressure in the interface and geometrical lay-out of the interface) is recommended because they execute an immediate influ- ence on Coulombic fields in the interface and thus on the ion transport proper tie^.^^ The fact that addition of the water can force an originally dry aerosol to display the same operational features as an originally wet aerosol generated by PN has been used by Hirata and co-workers for ET3'J8 and spark ablation of non- conducting samples.3g Water addition by a merging sys- enabled application of operational conditions as normally used with PN offering the advantage that no special optimization is required for ETV which would be a much more demanding task in operation because of the transient signals obtained in ETV.Furthermore it can be suggested that the addition of water to dry aerosols might be helpful in facilitating calibration in the case of direct analysis of solids by techniques such as Iaser ablation or spark ablation.With the addition of water to aerosols obtained from solids it might be possible to perform calibration by nebulization of standard solutions the advantage of which is obvious. The applicability of this suggestion has already been proved for laser ablati~n.~' The experiments with humidified aerosols just discussed indicate that the addition of a small amount of water (below 30 pg s-l) to a dry aerosol when keeping power d and NFR constant can be exploited to achieve a gain in intensity. As could be seen from further experiments the addition of water might also be useful in improving the stability of the analytical signals.This could be considered as an advan- tage but the disadvantage of the appearance of additional interferences must also be taken into account. Hydrogen Addition It can be presumed that hydrogen is mainly responsible for the effects caused by humidification of a dry aerosol. Therefore the influence of the addition of hydrogen to an ETV generated aerosol was studied in more detail. TwoJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 975 8 ( a) I C : I I I 1 I 1 5 6 7 8 9 10 11 12 13 14 Sampling distance/mm 8 ( b ) I B #- I "5 6 7 8 9 10 11 12 13 14 Sampling distance/m m -15 -10 -5 0 5 10 Bias potentialN Fig. 12 Intensity of Fe+ ions from a dry aerosol by varying the amount of hydrogen added to the aerosol gas flow; foward power 1 .O kW.(a) Axial intensity distribution for NFR= 1.2 1 min-l A 0 ml min-' V,= -6 V; B 2.5 ml min-I &= - 3 V; C 5 ml min-' ub' - 1 V; D 20 ml min-I Us=O V; and E 30 ml min-* &,= 1 V. (b) Axial intensity distribution for NFR=1.4 1 min-' A 0 ml min-l &= - 6 V; B 2.5 ml min-' &,= - 2 V; C 5 ml min-' V; D 10 ml min-l Vb=-1.5 V; E 20 ml min-I u b = - 1.5 V; F 30 ml min-l &= - 1 V; and G 40 ml min-I &=-0.5 v. (c) Ion energy characteristics by varying the amount of hydrogen added corresponding to (a) ds=6 mm; and NFR= 1.2 1 min-I series of measurements with different values of NFR were performed. The results are presented in Fig. 12 showing the axial intensity distributions of Fe+ obtained with various amounts of hydrogen being added in Fig.12(a) and (6) and additionally IEC measurements for NFR= 1.2 1 min-' in Fig. 12(c). The reflected power which shows a tendency to increase with the addition of greater amounts of hydrogen was always tuned to a minimum. Instability of the plasma does not occur unless hydrogen addition is increased to more than 40 ml min-I. Shrinking of the plasma and axial shift appear with hydrogen addition as well as with humidification as outlined in the preceding paragraph. As a consequence in Fig. 12(a) it can be seen that a certain hydrogen load is necessary to shift the intensity maximum into the observa- tion region. For the measuremets in Fig. 12(6) a higher value of the NFR is chosen ensuring that the maximum is in the observation region even without hydrogen addition.The initial axial shift of the intensity maximum by the higher hydrogen load is accompanied by an increase in the maximum value. Finally the axial position comes to a limit while the maximum value decreases again. In comparison to the measurements with a lower NFR represented in Fig. 12(a) the intensity gain by hydrogen is now fairly small and does not exceed a factor of 12. This shows once more that confusion can arise if results are interpreted that do not consider variation of the d,. Comparison of these results with Fig. 9 makes it obvious that these measurements confirm the similarity between humidification and hydro- gen addition offering a more detailed consideration by including the d dependence. The observation of a gain in intensity as seen in Fig.12(a) is in agreement with the literature. In ETV-ICP-AES signal enhancement of ion lines by the addition of hydrogen up to a factor of 10 has been observed for instance by M e r ~ n e t . ~ ~ In ETV-ICP-MS the use of hydrogen for improvement of the sensitivity was first mentioned some years ago.Io Recently Shibata et ~ 1 . ~ ~ investigated the effect of a hydrogen-argon carrier gas in ETV-ICP-MS in some detail. They also found a strong increase of analyte ion intensities and explained it by charge transfer collisions in the interface region as an additional reaction channel for ionization. In previous unpublished work with ETV by a tungsten wire loop (cJ Fig. l) an improvement of sensitivity by hydrogen addition was also observed which led to the present work with the object of emphasizing the positive influence of hydrogen addition.With the progress of the investigations the conclusion was reached that the same improvement could be achieved by careful optimiza- tion of NFR and ds without the addition of hydrogen. De-humidified Aerosols It can be seen from the humidification experiments that a strong signal depression for the analyte ions appears with higher moisture content (cf Figs. 9 and 12). Conversely this leads to the presumption that in the case of aerosols with high water content dehumidification might be help- ful in increasing the analyte ion intensities resulting in higher elemental sensitivities and better detection limits. Of the two different techniques provided for dehumi- dification in the present experimental arrangement cool- ing of the spray chamber has the advantage of lower cost and greater simplicity whereas the desolvation system is advantageous because of the greater flexibility and higher efficiency. Typical results obtained with dehumidification by the cooled spray chamber are represented in Fig.13 showing the axial intensity distributions for two different tempera- tures of the spray chamber with optimized bias potential [Fig. 13(a)] and the IECs for four temperatures [Fig. 13(6)] measured at the optimum d for the spray chamber with cooling at 1 "C. The results demonstrate that dehumidifica- tion of a wet aerosol leads to a decrease of both the optimum d and the bias potential with a simultaneous increase in the analyte ion intensity. This is the reverse of the preceding experience that humidification of a dry aerosol causes an increase in the optimum d and of the bias potential while the ion intensity decreases according to the expectation that the behaviour of an aerosol is reversible with respect to varying moisture content.These results are in agreement with the findings of Hutton and Eaton,2' who976 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 v) C U .- .- e Sampling distance/m m -5 0 5 10 15 20 Bias potentialN Fig. 13 Measurements for wet aerosols with dehumidification by cooling of the spray chamber at different temperatures; d,= 1 1 mm. ( a ) Axial intensity distributions of Fe+ ions for cooling tempera- tures A 1 "C (Ub=6 V water intake 170 pg s-l); and B 20 "C (r/,=9 V water intake 300 pg s-l).(b) Energy characteristics of Fe+ ions for cooling temperatures A 1 "C B 5 "C; C 10 "C; and D 20 "C used a Meinhard nebulizer with a cooled spray chamber and observed a gain in intensity a change of ion kinetic energy and a reduction of oxide interferences. Although the expected trend is generally demonstrated by these measurements the intensity gain as a result of dehumidification might be even more effective if water reduction were more effective. This can be achieved by the application of a desolvation system which was also studied in the course of the present investigations. However owing to the high practical significance of the results they have been published separately.' * The results obtained with desolvation are similar to those for a cooled spray chamber.In summary an intensity gain up to a factor of 2-5 a shift of the optimum d by about 2-3 mm a change of the bias potential by about -6 V and a decrease of the oxide ion intensity to about 40% in the case of BaO was obtained. However even with a desolvation system the water intake to the plasma is limited by the water vapour pressure to a minimum value of about 60 pg s-l. With respect to the fact that there is really no detailed and consistent picture of the processes contributing to the observed phenomena the discussion has been restricted here to some important aspects. A more detailed discussion would not only have to take into account the total amount of the water intake but also the size distribution of the aerosol droplets because it is known from preliminary measurements that a decrease of the cooling temperature favours the appearance of smaller droplet diameters.Of course the discussion gets more complex if the influence of water droplets is investigated as has been shown in the work of Olesik and F i ~ t e r . ~ ~ The water intake to the plasma that can be realized by desolvation is not far away from the region where maxi- mum intensity is achieved. Further reduction of the water intake requires more sophisticated techniques,8 of which cryogenic cooling promises to be the most efficient. This has especially been proved for the reduction of the solvent load in the analysis of organic solutions for improvement of detection limits by use of ethanol-water and for application of liquid chromatography. Nevertheless the development of a new and fully satisfying dehumidification system would be of considerable interest.Conclusion The measurements made show that besides the ion genera- tion parameters power and nebulizer flow rate the ion transfer parameters sampling distance and bias potential also exert a strong influence on the analytical performance in JCP-MS. A multiparametric optimization procedure would be indispensable to the achievement of optimum analytical performance. For a true comparison of different aerosols or different working conditions it is at least necessary that careful optimization of the ion transfer parameters is performed for each set of the ion generation parameters in order to avoid misinterpretations.The behaviour of aerosols is strongly influenced by the content of water vapour; humidification of a dry aerosol by higher amounts of water causes intensity depression in the same way as dehumidification of a wet aerosol can be useful to obtain a gain in intensity. On the other hand for the instrument used here a strong influence of water on the optimum bias potential expanding from below - 10 V for dry aerosols to above + 10 V for wet aerosols was observed. Optimization of the ion transfer parameters which has been neglected so far in most cases might be an effective means of realizing better analytical performance depend- ing on the instrument. In any case it is a necessary prerequisite for operational comparisons to take these parameters into account because otherwise neither com- parability of the results nor transferability of the analytical experiences is warranted.The authors gratefully acknowledge support from Finnigan MAT Bremen Germany The work has been supported financially by the Minister fur Wissenschaft und Forschung des Landes Nordrhein-Westfalen and by the Bundesmin- ister fur Forschung und Technologie. References 1 Hulmston P. and Hutton R. C. Spectroscopy 1991 6 35. 2 Denoyer E. R. Fredeen K. J. and Hager J. W. Anal. Chem. 1991 63 445A. 3 Jakubowski N. Feldmann I. Sack B. and Stuewer D. J. Anal. At. Spectrom. 1992 7 121. 4 Montaser A. Tan H. Ishii I. Nam S.-H. and Cai M. Anal. Chem. 199 1,63 2660. 5 Jakubowski N. Feldmann I. Stuewer D. and Berndt H. Spectrochim. Acta Part B 1992 47 119.6 Wiederin D. R. Smith F. G. and Houk R. S. Anal. Chem. 191 63 219. 7 Boumans P. W. J. M. and De Boer F. J. Spectrochim. Acta Part B 1976 31 355. 8 Wiederin D. R. Houk R. S. Winge R. K. and D'Silva A. P. Anal. Chern. 1990 62 1155. 9 McLaren J. W. Lam J. W. and Gustavsson A. Spectrochim. Acta Part B 1990 45 109 1. 10 Jakubowski N. and Stuewer D. 1989 Winter Conference on Plasma Spectrochemistry Reutte Austria paper P2-83. 11 Jakubowski N. Feldmann I. and Stuewer D. Spectrochim. Acta Part B 1992 47 107. 12 Caughlin B. L. and Blades M. W. Spectrochim. Acta Part B 1987 42 353.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 977 13 Nowak S. van der Mullen J. A. M. van Lammeren A. C. A. P. and Schram D. C. Spectrochim. Acta Part B 1989 44 41 1. 14 Murillo M.and Mermet J.-M. Spectrochim. Acta Part B 1989,44 359. 15 Alder J. F. Bombelka R. M. and Kirkbright G. F. Spectrochim. Acta Part B 1980 35 163. 16 Tang Y. Q. and Trassy C. Spectrochim. Acta Part B 1986 41 143. 17 Olesik J. W. and Den S. J. Spectrochim. Acta Part B 1990 45 325. 18 Nixon D. E. J. Anal. At. Spectrom. 1990 5 5 3 1. 19 Date A. R. and Gray A. L. Analyst 1981 106 1255. 20 Houk R. S. Fassel V. A. and Svec H. J. Dyn. Mass Spectrom. 1981 6 234. 21 Hutton R. C. and Eaton A. N. J. Anal. At. Spectrom. 1987 2 595. 22 Zhu G. and Browner R. F. J. Anal. At. Spectrom. 1988 3 781. 23 Lam J. W. and McLaren J. W. J. Anal. .4t. Spectrom. 1990 5 419. 24 Tsukahara R. and Kubota M. Spectrochim. Acta Part B 1990 45 58 1. 25 Jakubowski N. Raeymaekers B. J. Broekaert J.A. C. and Stuewer D. Spectrochim. Acta Part B 1989 44 2 19. 26 Berndt H. and Schaldach G. in CAS-5. Colloquium Atomspektrometrische Spurenanalytik ed. Welz B. 1989 Bodenseewerk Perkin-Elmer Uberlingen p. 109. 27 Ross B. S. Chambers D. M. Vickers G. H. and Yang P. J. Anal. At. Spectrom. 1990 5 35 1. 28 Vickers G. H. Wilson D. A. and Hieftje G. M. Spectrochim. Acta Part B 1990 45 499. 29 Schmit J. P. and Chtaib M. Can J. Spectrosc. 1987 32 56. 30 Douglas D. J. and French J. B. J. Anal. At. Spectrom. 1988 3 743. 31 Olivares J. A. and Houk R. S. Appl. Spectrosc. 1985 39 1070. 32 Chambers D. M. Poehlman J. P. Yang P. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 741. 33 Chambers D. M. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 761. 34 Chambers D. M. Ross B. S. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 785. 35 Fulford J. E. and Douglas D. J. Appl. Spectrosc.. 1986 40 971. 36 Ross B. S. Yang P. Chambers D. M. and Hieftje G. M. Spectrochim. Acta Part B 1991 46 1667. 37 Hirata T. and Masuda A. J. Anal. At. Spectrom. 1990 5 627 38 Hirata T. Akagi T. Shimiru H. and Masuda. A. Anal. Chem. 1989,61 2263. 39 Hirata T. Akagi T. and Masuda A. Analyst 1990 115 1329. 40 Hirata T. J. Anal. At. Spectrom. 1990 5 589. 4 1 Moenke-Blankenburg L. Schumann T. Gunther D. Kuss H.-M. and Paul M. J. Anal. At. Spectrom. 1992 7 251. 42 Matousek J. P. and Mermet J.-M. 1992 Winter Conference on Plasma Spectrochemistry San Diego USA paper THS. 43 Shibata N. Fudagawa N. and Kubota M. Spectrochim. Acta Part B 1992 47 505. 44 Olesik J. W. and Fister J. C. 111 Spectrochim. Acta Part B 1991 46 851. 45 Huang B. Yang J. Pei A. Zeng X. and Boumans P. W. J. M. Spectrochim. Acta Part B 1991 46 407. Paper 3/01 196H Received March 1 1993 Accepted June 1 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800969

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 979 Determination of Trace Metals in Concentrated Brines Using Inductively Coupled Plasma Mass Spectrometry On-line Preconcentration and Matrix Elimination With Flow Injection Les Ebdon Andrew Fisher Howard Handley* and Philip Jones Plymouth Analytical Chemistry Research Unit Department of Environmental Sciences University of Plymouth Drake Circus Plymouth UK PL4 8AA An on-line preconcentration matrix elimination procedure was developed for the analysis of concentrated brines by inductively coupled plasma mass spectrometry. Elements of interest (alkaline earths and first row transition metals) were concentrated quantitatively and isolated from the sodium chloride matrix using a chelating ion- exchange column. Two columns were evaluated an in-house dynamically coated chelating exchange resin and a commercially available chelating resin.The method was applied to the determination of the metals in concentrated brines and the analysis of the Open Ocean Sea-water Reference Material NASS-3 (National Research Council of Canada). Linear calibrations were obtained in the concentration range of interest (t200 pg I-') and excellent agreement with the certified reference values was obtained. Keywords Brine analysis; matrix elimination; inductively coupled plasma mass spectrometry; flow injection; on- line preconcen tration During the last decade the introduction of new technology into the chlor-alkali industry has led to a shift away from mercury and diaphragm cells which were relatively tolerant to feed brine quality towards membrane cells.Membrane cells although more efficient than mercury and diaphragm cells are also more susceptible to trace metal impurities in the feed brines. Relatively high concentrations of certain cations in the feed brines can reduce current efficiency and lead to shortened membrane life-times.* The determination and quantification of these trace elements is thus a prerequisite for process control. Previously Ebdon et uL3 described the use of an automated high-performance liquid chromatography sys- tem for the determination of the alkaline earths (which are of particular interest) in feed brines utilizing a novel chelation ion-exchange column for preconcentration- matrix elimination prior to analysis by ion chromato- graphy. There have been many reports in the literature regarding the use of chelation and ion-exchange columns for sample clean-up prior to analysis by either electrother- mal atomic absorption spectr~metry,~~~ flame atomic ab- sorption spectrometry (FAAS),6-9 inductively coupled plasma atomic emission spectrometry10 and inductively coupled plasma mass spectrometry (ICP-MS).I1-l4 Over half of these have dealt with the determination of trace metals in sea-water and a similar number have described the use of on-line analysis.Little work has been carried out however on the analysis of concentrated brines and the determina- tion of the alkaline earths. In this paper the use of two chelation ion-exchange columns is described for matrix elimination and on-line preconcentration of the alkaline earths and first row transition metals in concentrated brines prior to determina- tion by ICP-MS.The columns used were an in-house chelating ion-exchange column and a commercially avail- able chelating column. The performance of the two columns was assessed. Total residual sodium was determined by FAAS (Model 4000 atomic absorption spectrometer Perkin-Elmer Beac- onsfield Buckinghamshire UK). The plasma operating conditions and quadrupole conditions are outlined in Tables 1 and 2 respectively the conditions for the FAAS work are stated where appropriate in the text. A four channel peristaltic pump (Gilson Minipuls 3 Anachem Luton Bedfordshire UK) was used for delivery of all reagents and samples and to pump the drain from the spray chamber.Three inert switching valves (Dionex Camberley Surrey UK) were used for directional flow of reagents controlled via an auto-ion controller (Dionex) for unattended oper- ation. The valve configuration is outlined in Fig. 1. The columns were maintained at 60 "C in a water-bath through- out the experiments. All reagents were of analytical-reagent grade unless stated otherwise. Distilled de-ionized water was obtained from a Milli-Q water system (Millipore Bedford MA USA). Concentrated brines were obtained from ICI Chemicals and Polymers (Runcorn Cheshire UK). Stock standard solutions (Merck Poole Dorset UK) were diluted daily to yield 10 mg 1-l multi-element working solutions which were used in turn to spike concentrated brine samples to yield final concentrations in the range 0-200 pg 1-I.Throughout this paper the term brine refers to saturated sodium chloride solution (approximately 30% m/v sodium chloride). Table 1 Plasma conditions Forward power/W 1350 Coolant gas flow rate/l min-' Intermediate gas flow ratell min-' Nebulizer gas flow ratell min-' 13.0 0.6 0.8 Instrumentation and Reagents All elemental determinations were carried out using a VG Plasmaquad I1 (FI Elemental Winsford Cheshire UK). Table 2 Quadrupole conditions No. of channels 2048 No. of sweeps 100 Dwell time/ps 320 Points per peak 5 Skipped regionslu 20-23.5 27.5-4 1.5 70.0-85.0 * Present address Laboratory of the Government Chemist Queens Road Teddington Middlesex UK TW 1 1 OLY.980 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 1 mol I-' HNO I m Acid loop t-gstr-$;stt-rute Waste Metpac Sample ICP-MS Distilled water 6 u 4 Brine sample 1 mol I-' HNO Acid loop Distilled water I u + Brine sample 1 mol I-' HNO Fig. 1 Block diagram of the switching valves preconcentration column and sample loops used during ICP-MS studies (a) switched for loading the sample loop; (6) for preconcentration and matrix elimination; and (c) for elution of retained metals Procedure Preparation of the Xylenol Orange preconcentration column The preparation of this column has been described in an earlier p~blication.~ Preconcen t rat ion Procedure The preconcentration manifold was configured such that distilled de-ionized water was pumped continuously to the nebulizer via channel one of the peristaltic pump see Fig.1. The brine samples were loaded into the sample loop via channel two of the pump. A sample loop was used to load the brine samples in order to improve the precision. During this step the column was washed with an aliquot of dilute ammonium solution to adjust the pH of the column to that of the brine (pH 11). This pH was also found to be the optimum for retention of the alkaline earths during a previous study.3 The valve was then switched so that the ammonium solution carried the brine sample to the preconcentration column and the trace metals were re- tained. A further 5 ml of ammonium solution were then pumped across the column to remove the matrix from the connecting tubing and column dead volume. The nitric acid eluent (containing 50 pg 1-' of indium as internal standard) was loaded in the eluent loop via a syringe.Once the brine sample had been loaded onto the column the valves were switched such that the water stream carried the nitric acid slug to the column eluting the retained metals and carrying them to the nebulizer. If the internal standard had been added to the sample then it is unlikely that it would have been retained on the column to the same degree as the analytes in question. If on the other hand the nitric acid eluent (with internal standard) had been pumped continu- ously to the nebulizer the transient signal of the analytes would have been ratioed to a constant signal of the internal standard. By configuring the system in this way both the analytes and internal standard are measured as transients and both are present in the same concentration of residual sodium chloride eluting from the column.Determination of Residual Sodium Chloride In order to determine the residual sodium chloride present in the eluate from the column the above procedure was carried out and the eluate from the column was connected to a flame atomic absorption spectrometer. Owing to the high concentrations of sodium present it was necessary to decrease the sensitivity of the spectrometer by turning the burner head through 45" (rotation through a full 90" which would have been preferred was not possible) and also by using the less sensitive 303.3 nm line for the measurements. An air-acetylene flame was used with gas flow rates as recommended by the manufacturer. The preconcentration procedure outlined above was repeated with various vol- umes of ammonium solution being used to wash the column after loading the brine sample. Results and Discussion Residual Sodium Concentration Both the preconcentration columns were washed with various volumes of dilute ammonium solution (1-10 ml) following the preconcentration of 1.2 ml of brine.From this it was seen that the total sodium present was less than 3000 mg 1 - I after a 5 ml wash volume. Further increments in wash volume had very little effect on this level. This can be accounted for by the fact that the resin used in the column is converted into the sodium form during preconcentration of the brine. Clearly then no amount of washing will remove this residual sodium until acid is passed through the column which subsequently converts the resin back to the hydrogen form. Clearly the acid cannot be used as a wash step as it also elutes the retained metals.This level of sodium (0.3%) however does not present a problem especially when introduced to the ICP-MS system as a flow injection slug. Xylenol Orange Column Initial work was centred on the Xylenol Orange preconcen- tration column. Although the main elements of interest in the brines were the alkaline earths the transition metals were also determined thus making full use of the capabili- ties of ICP-MS. Linear calibrations were obtained from the alkaline earth metals although the results for the transition metals were somewhat disappointing. This might be ex- plained by virtue of the difference in the optimum pH for retention of each of the metals studied.15 For magnesium calcium strontium and barium the optimum pH is around 10.5.For the first row transition metals such as cadmium cobalt and zinc however the optimum pH is 5-6 and for copper 4-6. The optimum pH for retention of vanadium is 1.8. Indeed indium which was used as the internal standard is retained at a pH of 3-4.5 highlighting the problem of spiking the internal standard to the brine samples. Follow- ing this procedure magnesium strontium and barium were determined in the concentrated brine at the 9 50 and 8 pg 1 - I level respectively which compares well with other determinations. Calcium was not determined owing to the polyatomic interferences from Ar+ ArH2+ and C02+ at mlz 40 42 and 44 respectively.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 98 1 Metpac CC-1 Column The Metpac CC-1 column is a commercially available chelating exchange column (Dionex) with an iminodiacetic acid (H2ida) functionality. The column was used in the same manner as the Xylenol Orange column. Linear calibrations were obtained for the following elements magnesium strontium manganese copper cobalt nickel vanadium chromium and zinc over the concentration range 0-200 pg 1-l. Non-linear calibrations were obtained over the same concentration range for aluminium and titanium. The residual sodium present in the eluate was such that the 24Mg isotope could not be used due to overlap of the 23Na isotope. The removal of chloride was so complete however that no interferences were observed on 51V or 53Cr from 35C1160+ and 37Cl160+ respectively.Calcium could not be determined reliably at these levels owing to the presence of Ar+ ArH2+ and C02+ at rnlz 40 42 and 44 respectively. Using the preconcentration-matrix elimination proce- dure outlined above a 1.2 ml sample aliquot was loaded onto the column while 2.5 ml of acid was used to elute the retained metals. Clearly there is then a net dilution of a factor of two using this method. When the work was initiated this was not a problem since ICP-MS has sufficient sensitivity to determine the analytes at these concentra- tions and the main purpose of the column was for matrix elimination. As no reference materials were available in a 30% sodium chloride matrix a standard Open Ocean Sea- water Reference Material NASS-3 (National Research Council of Canada) was used.However the determination of trace metals in sea-water samples requires significant improvements in the limit of detection if accurate analysis is to be achieved. This requires a larger sample size to be loaded the limit of detection being dependent on the preconcentration factor achieved and the quality of the reagent blanks. Analytes in sea-water (10 ml) were preconcentrated by passing the water through the column at a flow rate of 2 ml min-l for 5 min. In this way five analytes could be determined virtually simultaneously. Longer preconcentra- tion periods would be necessary to determine analytes such as cadmium cobalt lead and manganese which are present at extremely low concentrations.After the preconcentration period the column was washed in the normal way to remove the chloride matrix and then the analytes were eluted for detection by nitric acid (0.2 ml 1 mol 1 - I ) containing indium as the internal standard. A decreased acid volume was used to minimize contamination. The isotopes of the analytes determined are detailed in Table 3. Copper was determined at rnlz 65 rather than the more abundant isotope at rnlz 63 to avoid the interference of 23Na40Ar+. The results obtained from the analyses are shown in Table 3. In general the results are in reasonable agreement with the certificate values. Comparison Although the two columns have essentially the same functional group (H,ida) they behave somewhat differently from one another. This might be due in part to the environment of the functionality itself in the case of Xylenol Orange this is adjacent to a benzene ring.The other ~ Table 3 Results for the analysis of Certified Reference Material Open Ocean Sea-water NASS-3; values given k the 95% confi- dence limit ( n = 5) Isotope Analyte determined c u 65 Mo 98 Ni 58 U 238 Zn 64 In 115 Certificate value/ Value obtained/ 0.109+0.011 11.5 k 1.9 0.257 k0.027 0.293 k 0.03 3.00 k0.15 2.94 k 0.13 Pg I-' Pg I-' 0.1 15 f 0.01 0 10.2 k 2. I 0.178 k 0.025 Internal standard - 0.190 k 0.020 factor which effects the performance is the distribution of the functionality across the resin. The Xylenol Orange dye was coated onto an anion-exchange resin (Dowex 1 X8). The coating of the dye may not have been even throughout and may have been retained by ionic interactions physical trapping and non-polar-non-polar interactions.This also effects the environment of the functionality. The Xylenol orange column was found to be efficient at retaining the alkaline earths at the pH used although the first row transition metals were only slightly retained leading to poor recoveries and poor reproducibility. The Metpac CC-1 column however retained all the metals studied yielding good precision and linear calibrations with the exception of aluminium and titanium. The structure of the Metpac CC-1 column is confidential but the functionality is obviously in a different environment yielding better recoveries for the transition metals at higher PH. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 References Brooks W. N. Chem. Br. 1986 1095. Wada H. Asakura K. Rattaiah G. V. and Nakagawa G. Anal. Chim. Acta 1988 214 439. Ebdon L. Handley H. W. Jones P. and Barnett N. W. Mikrochim Acta. I 9 9 1 11 39. Sturgeon R. E. Berman S. S. Desaulniers A. and Russell S. Talanta 1980 27 85. Nakashima S. Sturgeon R. E. Willie S. N. and Bermann S. S. Fresenius' Z. Anal. Chem. 1988 330 592. Riley J. P. and Taylor D. Anal. Chim. Acta. 1968,40 479. Olsen S. Pessenda L. C. R. RGiiEka J. and Hansen E. H. Analyst 1983 108 905. Fang Z. and Welz B. J. Anal. At. Spectrum. 1989 4 543. Hirata S. Honda K. and Kumamaru T. Anal. Chim. Acta 1989 221 65. Hirata S. Umezaki Y. and Ikeda M. Anal. Chem. 1986,58 2602. McLaren J. W. Mykytiuk A. P. Willie S. N. and Berman S. S. Anal. Chem. 1985 57 2907. Beauchemin D. and Berman S. S. Anal. Chem. 1989 61 1857. Boomer D. W. Powell M. J. and Hipfner J. Talanta 1990 37 127. Shabani M. B. and Masuda A. Anal. Chem. 1991,63,2099. Indicators ed. Bishop E. Pergamon Press Oxford 1972 p. 358. Paper 3/01 7256 Received March 25 1993 Accepted May 2 7 I993

ISSN:0267-9477    

DOI:10.1039/JA9930800979

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 983 Isotope Ratios of Calcium Determined in Calcium-46 Enriched Samples From Infants by Automated Multiple-collector Thermal Ionization Mass Spectrometry Judith R. Turnlund William R. Keyes and Karen C. Scott Western Human Nutrition Research Center United States Department of Agriculture Agricultural Research Service PO Bos 29997 Presidio of San Francisco CA 94 129 USA Richard A. Ehrenkranz Yale University School of Medicine Department of Pediatrics PO Box 3333 New Haven CT 06510 USA A high-precision method was developed for automated multiple-collector determination of 46Ca enrichment using magnetic sector thermal ionization mass spectrometry (TIMS). Calcium was separated from biological samples by precipitation as calcium oxalate.The calcium oxalate was washed then dissolved in nitric acid and isotope ratios were determined in 10 pg samples. The W a 46Ca and %a isotopes were collected simultaneously. The 46Ca:48Ca ratios were measured and corrected for fractionation by iterative normalization using the Va:@Ca ratio in order to achieve the required precision. Blocks of ten ratios were measured with an internal or within-run precision of 0.14% relative standard deviation (RSD) in urine samples and 0.10% RSD in faecal samples. The external or between-run precision for nine replicates was 0.07% RSD for urine and 0.09% RSD for faecal samples. Enrichment of samples collected following the feeding of 46Ca to pre-term infants ranged from 8 to 179 A0/o excess in faecal samples and from 19 to 91 A0/o excess in urine samples.If another isotope in addition to 46Ca is to be enriched the identical analytical method can be applied by collecting 42Ca %a %a and 46Ca simultaneously using the two unenriched isotopes to correct for fractionation. Keywords Thermal ionization mass spectrometry; stable isotopes; calcium; clinical samples The purpose of the research reported herein was to develop an automated multiple-collector method for high-precision thermal ionization mass spectrometry (TIMS) determina- tion of 46Ca in bioligical samples including some with little calcium. In addition a relatively rapid method was re- quired to maximize sample throughput. The stable isotope 46Ca was first used to study calcium metabolism in children in 1967.' Neutron activation analysis (NAA) was used to quantify the isotope in this and in other studies including a study of premature infants.2 Thermal ionization MS has also been used to analyse 46Ca samples in studies of calcium metabolism.Both magnetic sector3 and quadrupole4 TIMS have been used successfully and with greater precision and accuracy than NAA. While magnetic sector TIMS is capable of the highest precision analysis is relatively slow. This limitation has been mini- mized with the introduction of TIMS instruments with multiple collectors and computer control. Automated multiple-collector methods for a number of elements have been developed and rep~rted.~ The 46Ca isotope has a very low natural abundance 0.003% and an automated mul- tiple-collector method for measuring 46Ca ratios has not previously been reported.The method described in this paper was applied to the analysis of urine and faecal samples collected from pre-term human infants given a meal extrinsically labelled with 46Ca. The range of enrichments observed is reported. Results of the infant study conducted to evaluate calcium absorption and excretion in very low birth weight infants fed on human milk fortified human milk or one of three infant formulas will be reported elsewhere. Experimental Preparation and Administration of '6Ca solutions Calcium carbonate containing calcium with 34.9 1 at.-% 46Ca (Oak Ridge National Laboratory Oak Ridge TN USA) was dissolved in HC1(37%) and then diluted with de- ionized water. The 46Ca concentration and enrichment were determined by isotope dilution with 44Ca.The solution had an elemental calcium concentration of 2 1.08 pg g-I and a 46Ca concentration of 7.986 pg g-l. The dose of 46Ca given to an infant was based upon an estimate of the daily dietary intake of calcium assuming a human milk or formula milk intake of 150 ml kg-l d-' and provided approximately a 5-fold enrichment of the dietary intake of 46Ca during 1 d. The calcium content of the human and formula milk ranged from 300 to 950 pg 1-1 of Ca. An extrinsically labelled meal was prepared by adding an accurately measured volume of the 46Ca isotope solu- tion equilibrating for 2-3 h and then administering by a naso-gastric tube. Faecal and Urine Sample Collection Faecal and urine samples were obtained during 50 nutri- tional balance studies performed with 4 1 appropriate for gestational age premature infants [birthweight 1267 f.258 g gestational age 29.8 f 1.9 weeks (mean k SD) 4-83 post- natal days of age].6 Stool and urine were collected as previously described7 after 72 h of dietary intake. Stools passed during the collection period were pooled weighed and homogenized with de-ionized water. An aliquot was frozen for calcium isotope analysis. Urine excreted during the collection period was pooled the total volume was determined and an aliquot was frozen for later calcium isotope analysis. Sample Preparation Calcium was separated from urine and ashed faecal samples by conversion of the mineral to its oxalate form with subsequent precipitation.8 The calcium concentrations ranged from 0.4 to 5.4 mg g-l of Ca in faecal homogenates and from 7.0 to 194.0 pg ml-I of Ca in urine.The urine samples were spun in a centrifuge (4 min 2000g) to remove any particulate matter. Approximately 1 ml of a saturated ammonium oxalate solution adjusted to p H 2 10 with NH40H was added to 5 ml of urine. Homogenized faecal samples (1-2 g) were dried ashed in a mume furnace9 and984 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 dissolved in a minimum volume of 6 moll-' ultrapure HCl (Seastar Chemicals Seattle WA USA). Ammonium oxa- late solution (5 ml) was added to each faecal sample. A minimum of 15 h after the addition of ammonium oxalate urine and faecal samples were placed on ice for 15-30 rnin in order to increase oxalate precipitation then spun in a centrifuge for 5 rnin at 2000g.The precipitate was washed twice with de-ionized water then dissolved in 2% ultrapure HN03 (Seastar Chemicals) and diluted to 2 mg ml-* of Ca. Precautions were taken to avoid contamination with cal- cium. Tubes used in sample preparation were acid washed then rinsed in de-ionized water. All crucibles used to ash faecal samples were acid washed rinsed with de-ionized water refluxed for 20 rnin with approximately 0.5 ml of a 1 + 1 solution of concentrated HN03 and de-ionized water and rinsed in de-ionized water. Powder-free vinyl gloves a hair cover and shoe covers were worn during sample preparation procedures. Samples were prepared in a clean room supplied with high efficiency particular air (HEPA)- filtered air and under positive pressure.Calcium blanks were determined by measuring the isotope ratios of a solution highly enriched in 44Ca before and after it was subjected to sample preparation procedures. A set of four enriched standards were prepared by weighing the 46Ca solution into a calcium chloride standard solution containing 1000 ppm of Ca (Atomic absorption standard Mallinckrodt Paris KY USA) to achieve levels of enrichment that would include the expected enrichments of the samples. These samples were processed using oxalate precipitation and dissolved in HN03 as described above. The isotope abundances of the 46Ca and natural calcium solutions are shown in Table 1. Table 1 tions Isotope abundances of natural and 46Ca-enriched solu- Abundance (at.-%) Isotope Natural* 46Ca-enrichedt 40 96.94 59.06 42 0.65 0.60 43 0.14 0.14 44 2.09 3.71 46 0.0032 34.9 1 48 0.19 1.58 * Measured values.t Based on certified values (Oak Ridge National Laboratory). Isotope Ratio Determinations and Calculations Isotope ratios were determined with a computer-controlled magnetic sector thermal ionization mass spectrometer (Finnigan MAT Model 26 1 Bremen Germany) equipped with a reference pyrometer and two sample pre-heat positions with computer control and digital readout. The collector system contains five Faraday collectors a fixed central collector between four outer collectors mounted on moveable rods and manually adjustable with digital dials. The software supplied with the instrument was rewritten in- house to provide the additional capabilities described previously,s as well as the isotopic dilution and iterative normalization calculations specific to the 46Ca analysis.Prior to loading samples filaments were prepared. The sample magazine holding 1 3 pairs of double-filament assem- blies with filaments of zone-refined rhenium [O.OO 1 x 0.030 in (0.025 mmx0.76 mm) H. Cross Weehawken NJ USA] was heated automatically under vacuum in a bakeout device (Finnigan Option 567) at a filament current of 4.5 A for 15 rnin to remove impurities and adsorbed gases. Samples were loaded onto filaments in a laminar flow bench (Baker EdgeGARD Laminar Flow Bench Model EG- 4252 Sanford ME USA). Pipette tips used for loading were rinsed before use with de-ionized water and discarded after each use. A 5 pl drop of a sample solution containing 10 pg of Ca was placed on the evaporation filament.Using the automatic sample preparation device supplied with the mass spectrometer the samples were heated at 1 A for 3 min then at 1.5 A for 1 min and then the current was automatically increased gradually at a rate of 0.5 A min-' to 2.0 A and turned off. A 250 W heat lamp placed 10 cm above the sample magazine was used to aid in evaporation. After all samples were loaded the sample magazine was inserted into the mass spectrometer the vacuum pumps were turned on and the instrument pumped down to a pressure of 3 x Pa. Automatic analysis of the samples included heating each sample in each of two pre-heat positions. After the first sample was pre-heated the pre-heating steps for subsequent samples occurred during analysis of the previous sample.This in effect reduced total analysis time by 20 rnin per sample. The sample heating procedure was programmed to heat the filaments until specified values were reached as shown in Table 2. The pilot mass was first set to mlz 40 to centre and focus the ion beam and then changed to mlz 44 for final heating. When the 44Ca ion current resulted in an amplifier voltage of 7 V data acquisition began. The typical ionization filament temperatures at each step are shown in Table 2. ~ Table 2 Sample heating procedures prior to 46Ca analysis Time/ min O§ 1 O§ 20 20 25 35 45 50 5 5 Pilot mass - 40 40 40 44 44 44 44 Filament heated* I I I E I E E E E Control parameter FC FC FC FC ICll FC FC IC IC Control valuet 2.0 A 2.5 A 2.8 A 0.3 A 2.0 v 4.0 A 1.5 A 3.0 V 7.0 V Ionization filament (I) temperature$/"C 1250 1350 1450 1450 1500 1750 1750 1750 1750 * I =ionization filament and E=evaporation filament.t Filament current (FC) or ion beam voltage at the pilot mass. $ Typical temperature reached at control value. 8 Done in pre-heat positions. 7 IC= ion current.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 985 The ion beam intensities were integrated for 8 s with 5 s idle time between measurements. Means of ten measure- ments of each ratio were calculated. The individual ratio measurements were subjected to the Dixon outlier test for extreme mean,1° which was included in the software and ratios which failed the test were eliminated. Data for a sample were discarded if one of the ratios had a relative standard deviation (RSD) greater than 0.3%.Isotope dilution calculations using iterative normalization to cor- rect for fractionati~n,~ were carried out during analysis as programmed in-house. Each set of 12 samples analysed contained an additional unenriched sample of the same matrix that had been carried through all sample preparation procedures with the samples. The mean ratios of these unenriched samples were used as the natural ratios in enrichment and isotope dilution calculations. The amount of enriched 46Ca present in a sample was determined by isotope dilution using the total calcium content of the sample (determined by atomic absorption spectrometry) and the isotope ratios of the enriched and unenriched samples and the enriched 46Ca solution using eqn.(1). The derivation of the equation has been described previously. where A = isotopic abundance j=enriched isotope (46) k=reference isotope (48) m=total Ca M=mass of Ca (mg) n=natural Ca (unenriched) R = isotopic ratio cor- rected for fractionation with 44Ca:48Ca ~=~~Ca-enriched tracer and W=atomic mass. Fractionation corrections were made through iterative normalization using the 44Ca:48Ca ratio to improve the between-run precision of the 46Ca:48Ca ratio as described previou~ly.~ Briefly Ms calculated from eqn. ( l ) and M" were used to calculate an expected value of the 44Ca:48Ca ratio. This value and the measured value were used to normalize the 46Ca:48Ca ratio which was then used in eqn. (1) to calculate a corrected Ms.The procedure was repeated for a total of four interactions. Enrichment (A% excess) ( E ) was calculated using eqn. (2). R"jk - R"jk oo E= R"jk Results and Discussion Recovery of calcium from urine was low and variable when 5 ml of oxalate solution were used for 5 ml of urine. The low recovery was probably due to the low calcium content. Calcium recovery from the oxalate precipitation was not affected by the length of time allowed for the oxalate solution to react with the faecal ashes but urinary calcium recovery was as low as 15% and variable when reaction time was limited to 10 min. After conditions were optimized the recovery of calcium following oxalate precipitation was greater than 90% for both faecal and urine samples. Contamination with natural calcium was unacceptably high and variable initially ranging from 50 ng to 13 pg per sample.Since some urine samples contained as little as 20 pg of calcium it was essential to decrease contamination. After conditions were optimized and the extra precautions described under Sample Preparation were taken to avoid contamination the blanks ranged from 36 to 100 ng. This was less than 0.1% of the calcium content of all but a few urine sample aliquots which ranged in calcium content 1501 1 0 10 20 30 40 50 60 70 Ti m e/m i n Fig. 1 the time intervals shown in Table 2 Ion beam intensity over time for 44Ca. The arrows indicate from 20 to 600 pg. Calcium content of faecal aliquots ranged from 500 pg to 8 mg. Under the conditions for isotope ratio determinations described in Table 2 the ion beam intensity at rnlz 44 over time during the warm-up and data collection periods is illustrated in Fig. 1.Each 10 pA of ion-beam current produces 1 V of amplifier output. The heating times listed in Table 2 are indicated by the arrows. The first two 10 min steps carried out in the pre-heat positions served to eliminate any potassium and/or organic material present. In the next step the main peak at rnlz 40 was used to focus and centre the ion beam and to attain a pre-set intensity. The ionization filament current was increased before increasing the evaporation filament current since the efficiency of ionization is determined by the tempera- ture of the ionization filament. Ion-beam intensity varied in response to the filament current applied.The next step at 25 min was used to compensate for this by adjusting the ionization filament current until an amplifier output of 2 V at mlz 40 was reached. This procedure provided enough beam for focusing but avoided excessive signal. At 35 min the pilot mass was changed to rnlz 44 and the ionization filament current was raised to 4 A producing a large signal that declined rapidly. The evaporation fila- ment current was then increased to 1.5 A the ion beam intensity of mlz 44 rose again and the beam was re- focused. The evaporation filament current increase at 50 min was used to achieve 3 V of amplifier output at rnlz 44 and control was passed to the data collection module at 55 min for the final heating step which increased the signal to 7 V at mlz 44 (10 mV at rnlz 46).After final focusing and baseline and relative gain measurements the ion beam intensity was stable and increasing iso- topic ratios were stable and data collection began at 60 min. In order to assess the degree of fractionation and the effect of fractionation corrections on precision and accuracy of the ratios 20 sets of ten ratio measurements of a natural faecal reference sample were taken. They were measured over a 75 min period and yielded uncorrected 46Ca:44Ca ratios which increased with time from 0.0 0 1 5 146 to 0.0 0 1 5 353 demonstrating mass fractiona- tion of 1.4%. The normalized 46Ca:44Ca ratio was 0.0 0 1 5 142 k 0.0 000 0 1 5 (0.1 Oh RSD) showing the benefit of normalization. A single set of ten normalized ratio measurements under these conditions achieved sufficient precision and reduced sample analysis time so a single set was measured in subsequent samples.Because mass fractionation of the isotopic ratios was observed the 44Ca:48Ca ratio was used to correct the 46Ca:48Ca ratio through a linear iterative normalization correction. The theory and principles of fractionation corrections have been described in detail.I2 Table 3 shows the improvement of precision of the 46Ca:48Ca ratio due to iterative normalization for a set of nine replicates of a986 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 Table 3 Effect of iterative normalization using 44Ca:48Ca= 1 1.273 (ref. 12) on precision of 46Ca:48Ca ratio in unenriched urine and faecal samples Reference urine sample Unenriched (n=9) infant samples* Parameter Before After Before After 06Ca:48Ca 0.01700 0.01704 0.01708 0.01710 Standard deviation 0.000 1 3 0.0000 1 0.000 1 6 0.00004 External precision (TO)$ 0.76 0.07 0.94 0.23 Internal precision (Oh)? 0.15 0.14 0.13 0.12 * Series of unenriched urine and faecal samples from pre-term infants run on different days over a 9 week interval (n= 14).t Average within-run RSD of ten measurements. $ Between-run RSD. urine reference sample analysed on one day and for the natural faecal and urine samples analysed over a 9 week period along with the enriched samples. There was little improvement in the internal precision from 0.15 to 0.14% RSD for the urine reference sample but major improvement in external precision from 0.76 to 0.07% RSD. A comparison of expected and measured ratios of the set of solutions made from natural and 46Ca-enriched calcium are shown in Table 4.The measured ratios agree with expected values within the precision of measurements. Table 5 shows the enrichment and amount of 46Ca in faecal and urine samples collected over 72 h from infants after being fed 46Ca. The enrichments ranged from 8 to 179 A% excess in faecal samples and 19 to 91 A% excess in urine samples. The usual procedure for measuring the calcium isotopes 40Ca 42Ca 43Ca and 44Ca (ref. 5) was not suitable for the reliable measurement of the 46Ca isotope. For those isotopes 1 pg of calcium in the nitrate form was sufficient for measurements with external RSDs of 0.1% for the 42Ca:40Cand 44Ca:40Ca ratios and 0.2% for 43Ca:44Ca.The 40Ca 42Ca 43Ca and 44Ca beams were measured simultane- ously. Under the conditions of analysis for those more abundant isotopes the fractionation state was stable be- tween analyses and isotope ratios of natural samples were Table 4 Expected and measured 46Ca:48Ca ratios of 46Ca-enriched calcium (measured in pg of 46Ca per g of Ca) 46Ca:48Ca 46Ca added Expected Measured 0 - 0.01710 3.45 0.0 1864 0.01 863 10.5 1 0.02 172 0.02 177 3 1.87 0.03 127 0.03 126 93.30 0.05885 0.05868 close to the reference values so fractionation corrections were not needed. Owing to the very low abundance of 46Ca a number of changes were made to our procedures to achieve sufficient ion-beam intensity and stability at mlz 46 for accurate measurements of 46Ca ratios. The amount of calcium loaded onto the evaporation filament was increased to 10 pg.This provided sufficient sample to allow for the extra heating needed to increase the ion-beam intensity at rnlz 46 to an amplifier output of at least 10 mV. At this level the intensity at mlz 40 would be 325 V far beyond the 12.5 V amplifier limit. This necessitated omitting the peak at rnlz 40 and measuring only rnlz values of 44 46 and 48 simultaneously. The shielding in the multiple collector was sufficient to prevent interference from scattered ions from the large rnlz 40 beam. The calcium ion-beam intensity (Fig. 1) rose during heating to an initial peak fell to a lower level and then rose to a higher rising plateau as has been observed previ0us1y.I~ While the initial peak intensity was sufficient for accurate measurement of other calcium isotopes the final rising plateau region was necessary to achieve sufficient intensity and duration for reliable measurement at mlz 46.When measured by simultaneous collection of calcium isotopes nearly all variability due to fluctuations in ion- beam intensity and sample evaporation rate was elimi- nated. The internal precision was limited primarily by the counting statistics of the low intensity 46Ca ion beam. Since the rate of fractionation during a single set of ten ratio measurements was low normalization had little effect upon the internal precision as is seen in Table 3. External precision or repeatability of the ratios is more important than internal precision. Differences in fractionation states from analaysis to analysis directly affect the external precision.Fractionation was more variable and iterative normalization improved external precision markedly. The multiple collector method described herein was used to measure enrichment of 46Ca only. The idefitical multiple Table 5 Enrichment and amount of 46Ca in faecal and urine samples collected over 72 h after feeding 46Ca &Ca Faecal* (n = 50) Urinary* (n = 48) Enrichment (Aoh excess) 8-179 (80236) 19-91 (49-t 17) Amountlpg 0.4-19 (8.725.3) 0.03-1.82 (0.39k0.38) Dose (%) 2-59 (27 2 15) 0.3-6.5 (1.3 2 1.4) Infant body weightlgt (n=50) 1102-1950 Dietary Ca intake/mg kg-' d-' 35.0-205.8 %a fed/pg 7.99-43.92 * Range (mean 2 SD). -f The body weight of each infant was averaged over the course of the study. The range of those averages is reported here.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 987 ~~~ Table 6 T a isotope ratios measured using TIMS Ref. Sample type ( n ) 46Ca:44Ca* 46Ca:48Ca* This work Urine composite (9)t 0.001 5 1 14 * 0.0000008 0.00 1 5 168 * 0.0000020 14 Faecal/urine$ - 0.0 I 754 -+ O . O O O O ~ ~ ~ 15 CaCQ (15)11 0.00 1 5290 2 0.0000006 - 16 CaCO (7)** 0.00 I504 5 0.000029g 12 CaCO3tt 0.00 152 5 0.0000 I - 0.0 I 7039 f 0.000008 0.0 17 100 +. 0.000022 Infant faecal and urine ( 14)t - * Mean +- 2 x standard error of the mean. t One set of ten ratio measurements per analysis. $ Three sets of ten scans per analysis; n not reported. 9 Mean k SD. fl Quadrupole TIMS. 11 Sixty sets of ten double scans per analysis. ** One set of five double scans per analysis.tt Twenty sets of ten scans per analysis; n not reported. collector method can be used for high-precision measure- ments of enrichments of 46Ca and a second isotope (42Ca 43Ca or 44Ca) by collecting these isotopes simultaneously. In that case three ratios are measured and the ratio of the two unenriched isotopes is used to correct the other ratios for fractionat ion. The enrichments of samples following the dosing with 46Ca (Table 5) were variable and were as low as 8 Ao/o excess. This range of enrichment demonstrates what might be expected in infant studies and must be taken into account when designing experiments. At this level of enrichment high-precision isotope ratio measurements are required because a change of only 0.1% in the ratio results in a shift of 1.3% in the calculated 46Ca excretion.The 46Ca isotope ratios and precision reported here are compared with previously published TIMS data in Table 6. No comparable 46Ca data by NAA analysis were available but users of NAA have typically reported precisions of 1-5%.14 The external precision achieved for the urine sample set 0.07% RSD is considerably better than the 0.5% RSD achieved with quadrupole TIMS and is compar- able to the best precision reported by Jungck e? al.15 using magnetic-sector TIMS. Those data were based on ratio measurements in 60 sets of ten double scans ( 1200 scans) of the isotopic masses per sample and used a CaCO standard. Our ratios were based on a single set of ten ratio measurements per sample measured in calcium separated from human urine.Thus the procedure described in this paper made it possible to achieve the same high external precision from a small number of ratio measurements as can be made by an exhaustive set of measurements without a multiple-collector. Supported in part by grants from the National Institute of Child Health and Human Development Grant 4HD 17498 and the Children’s Clinical Research Center RR- 00125 NIH. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 References Bethard W. F. Schmitt R. A. Olehy D. A. Kaplan S. A. Ling S. M. Smith R. H. and Dalle Molle E. in Nuclear Activation Techniques in the Life Sciences International Atomic Energy Agency Vienna 1967 pp. 533-542. Ehrenkranz R. A. Ackerman B. A. Nelli C. M. and Janghorbani M. Pediatr. Res. 1985 19 178. Moore L. J. Machlan L. A. Lim M. O. Yergey A. L. and Hansen J. W. Pediatr. Res. 1985 19 329. Yergey A. L. Vieira N. E. and Covell D. G. Biumed. Environ. Mass Spectrom. 1987 14 603. Turnlund J. R. and Keyes W. R. J. Micrunutr. Anal. 1990,7 117. Ehrenkranz R. A. Gettner P. A. Nelli C. M. Sherwonit E. A. Williams J. E. Ting B. T. G. and Janghorbani M. Pediatr. Res. 1989 26 298. Ehrenkranz R. A. Gettner P. A. and Nelli C. M. J. Pediatr. Gastroenterol. Nutr. 1989 8 58. Smith D. L. Atkin C. and Westenfelder C. Clin. Chim. Acta 1985 146 97. Turnlund J. R. Smith R. G. Kretsch M. J. Keyes W. R. and Shah A. Am. J. Clin. Nutr. 1990 52 373. Dixon W. J. Ann. Math. Stat. 1950 22 68. Turnlund J. R. King J. C. Gong B. Keyes W. R. and Michel M. C. Am. J. Clin. Nutr. 1985 42 18. Russell W. A. Papanastassiou D. A. and Tombrello T. A. Geochim. Cosmochim. Acta 1978 42 1075. Moore L. J. and Machlan L. A. Anal. Chem. 1972,44,2291. Janghorbani M. Sundaresan A. and Young V. R. Clin. Chim. Acta 1981 113 267. Jungck M. H. A. Shimamura T. and Lugmair G. W. Geochim. Cosmochim. Acta 1984 48 265 1. Moore L. J. in Stable Isotopes in Nutrition eds. Turnlund J. R. and Johnson P. E. American Chemical Society Washington D.C. 1984 pp. 1-26. Paper 2/065 79G Received December 10 I992 Accepted May 5 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800983

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 989 Evaluation of Linear Photodiode Array Detection for Continuum Source Atomic Absorption Spectrometry With Electrothermal Atomization Clare M. M. Smith* Robbin Nichol and David Littlejohn Department of Pure and Applied Chemistry University of Strathclyde 295 Cathedral Street Glasgow UK GI IXL A 256 pixel linear photodiode array (LPDA) detector was fitted to a Spectrametrics echelle spectrometer for measurements by continuum source atomic absorption spectrometry with electrothermal atomization (ET- CSAAS). By selection of appropriate pixel regions intensity values could be obtained for calculation of absorbance signals integrated over wavelength as well as time. Detection lmits obtained sequentially for Cd Cr Cu Mn Mo and Pb with the LPDA were similar in magnitude (4-24 pg) to those obtained with mechanical wavelength modulation and detection by a photomultiplier tube using the same atomizer and spectrometer system.The calibration graphs obtained with the LPDA however were more useful analytically as integration of the absorbance signals across the line profile increased the linear range. Four National Bureau of Standards (NBS) [now National Institute of Standards and Technology (NIST)] Standard Reference Materials (SRMs) 1577a Bovine Liver 1575 Pine Needles 1572 Citrus Leaves and 1566 Oyster Tissue were analysed by ET- CSAAS with LPDA detection. The concentrations obtained for Cd Cr Cu Mn Mo and Pb were generally in good agreement with the certified values. Keywords Continuum source atomic absorption spectrometry; electrothermal atomization; linear photodiode array defection; reference material Instrumentation for continuum source atomic absorption spectrometry (CSAAS) frequently incorporates wavelength modulation with photomultiplier tube (PMT) detection to measure the intensity of the source radiation across an absorption profile.’-3 There are some advantages however in replacing the wavelength modulation refractor plate and PMT with a linear photodiode array (LPDA) detector.Firstly improvements in signal-to-noise (S/N) ratios are expected owing to the greater quantum efficiency of the LPDA in the UV region compared with that of a PMT.4q5 Also a multiplex advantage can be obtained by simulta- neous measurement of intensities by the LPDA over the absorption line In addition Moulton et al.4+5 have shown that pulsing the xenon arc continuum lamp in- creases transiently the intensity of the lamp emission and thereby improves the S/N in CSAAS.Wavelength modula- tion involving a refractor plate and PMT cannot be used with pulsed lamps however an LPDA detector is suitable for this application. A number of workers have reported applications of LPDA detectors for flame-CSAAS citing improvements in detection limits4*’ and the possibility of simultaneous multi- line measurement~~-~ as the main advantages of the detec- tor. The LPDA devices are capable of rapid scan frequen- cies and so can be used for measurement of the rapid transient signals associated with electrothermal (ET) atomi- zation.Harnly et al.435J have investigated the optimum conditions for measurement of ET-CSAAS signals with an LPDA. Improvements in S/N were achieved by increasing the entrance slit-width and incorporating more pixels into the absorbance calculation. A model was developed which predicts the absorbance noise as a function of slit-width and the number of pixels used. The effcts of changing these parameters on the detection limits of a number of elements were established. The use of a large entrance aperture increasing the source intensity by pulsing the current and pixel averaging help reduce the relative contribution of read-out noise which is the limiting noise with the LPDA detector. Read-out lag a carry-over effect caused by the * Present address Nutrient Composition Laboratory Beltsville Human Nutrition Research Centre BARC-East US Department of Agriculture Beltsville MD USA.delay between atomizations can be removed by ignoring the initial few scans of the array when processing the data obtained for an atomization event. In this work an analytical evaluation of the advantages of LPDA detection in ET-CSAAS is described. The ability to monitor simultaneously the entire wavelength region of the absorption line profile has been used to produce absorbance versus wavelength versus time spectra of a number of elements on a sequential basis. By selection of pixels that cover the absorption profile it has been possible to generate three-dimensional integrated absorbance measurements rather than the normal absorbance-time measurements that are obtained with square wave wavelength modulation and PMT detection when only the absorbance at the line centre is used.The calibration graphs and detection limits obtained with both detection systems have been compared for Cd Cr Cu Mn Mo and Pb. The concentrations of these elements in four reference materials have been determined in order to indicate the suitability of LPDA detection for trace analysis by ET-CSAAS. Experimental Instrumentation A Spectrametrics SM I11 echelle spectrometer and a Philips SP9 graphite furnace atomizer were used with a Cermax 300 W xenon arc lamp. Details of the system have been described The spectrometer was equipped with a Hamamatsu S292 PMT and a 5 mm thick quartz plate for measurement of intensities by mechanical wave- length modulation.The refractor plate was positioned behind the entrance slit of the monochromator and was mounted on a G 300 PD optical scanner torque motor operated with a scanner controller (General Scanning Westertown MA USA). Computer generated wave forms were used to drive the motor via the scanner controller. When LPDA detection was required a Hamamatsu S2304 256 pixel array mounted in a removable cartridge was positioned in the focal plane in place of the conventional exit slit cartridge. This arrangement facilitated easy and quick interconversion of the detection modes. With the LPDA detector movement of the refractor plate was not required.990 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 SP 9 furnace PSU and controller Continuum source furnace D I Torquemotor I I controller n I I IBM 286 PC AID and DIA converters Fig.1 Block diagram of PMT system SP 9 furnace PSU and controller I Continuumsource f u d I - n I Echelle I * I . monochro lator e Amplifier and (r)l T D convertei I b control - IBM 286 PC Fig. 2 Block diagram of LPDA system Table 1 SP 9 atomizer programme for the determination of elements in reference materials by CSAAS 20 pl sample volume; and 10 p1 modifier solution containing I pg of Pd(N03)* Temperature/ Ramp/ Stage "C Time/s "C s-' Pd dry Pd pre-heat Sample injection Dry Pyrolysis Pre-atomize Atomization Pre-clean Clean 120 1300 120 800 100 2500 100 2700 - 20 20 20 1000 40 20 30 1000 1 5 2000 6 2 2000 - - - - Data acquisition was achieved by use of an IBM ATe microcomputer equipped with a monitor and Epson FX-85 printer. In-house software was written to provide data acquisition and graphical display of atomization peaks and background absorbances reprocessing of data selection of integration limits variation of bracketed smoothing rou- tines and output to either the printer or a file.Block diagrams of these systems for both modes of detection are given in Figs. 1 and 2. The amplifier circuit was constructed using an Lf351N input amplifier to minimize input bias currents and also give a high gain bandwidth product. The switchable gains used high quality low tolerance (0.1 O/o) resistors. Signal pathways were earthed throughout. The amplified signal was taken to an AD574 analogue-to-digital converter and the conversion was controlled by the trigger signal from the LPDA board.The timebase of the LPDA board was generated by a clocWtimer circuit whose fre- quency was determined by the computer. The Philips SP9 atomizer was equipped with pyrolytic graphite coated electrographite tubes. A compromise atom- izer programme developed for simultaneous multi-element ET-CSAAS measurements was used for analysis of refer- ence materials (Table 1 ). Individual optimum atomization temperatures (Cd 1900 "C; Cr 2700 "C; Cu 2500 "C; Mn 2400 "C; Mo 2700 "C; and Pb 2100 "C) were used when calculating characteristic mass and detection limit values for both detection modes. Chemical modification was achieved by addition of 1 pg of Pd(N03)2 prior to injection of the standard or sample solution with a micropipette.Radiation from the xenon arc lamp was focused through the atomizer with a 100 mm diameter lens of 220 mm focal length. A second lens 50 mm in diameter and of 250 mm focal length provided collection optics for the monochro- mator. The image at the entrance slit had a magnification factor of 1. Measurements were made for the following elements Cd (228.8 nm) Cr (357.9 nm) Cu (324.7 nm) Mn (279.5 nm) Mo (313.3 nm) and Pb (283.3 nm). For mechanical wavelength modulation and PMT detection the entrance and exit slits of the echelle spectrometer were matched at heights of 300 pm and widths of 50 pm. When the LPDA detector was used the entrance slit-height was set at 500 pm and the entrance slit-width was selected to allow a level of radiation suitable for AAS measurements whilst prevent- ing saturation of the array.For Cr the slit-width was set at 50 pm. For the other analytes a slit-width of 100 pm was selected. LPDA Detector Procedure The LPDA was scanned 80 times in each 5 s atomization period. For the 256 pixels this implied that each pixel was read out every 244 ps. The first six scans of each set of 80 were discarded in any calculation in order to remove the effects of read-out lag. Prior to analytical measurements an optical blank reading was taken to correct subsequent readings for dark current. Background corrected absorbance values were calculated for each pixel covering the absorpton profile by calculating log(l,,/l) where I is the background intensity and I is the transmitted intensity for each pixel. A series of absorbances are therefore obtained and summed to give an integrated absorbance. Standard Solutions and Reagents Standard solutions were prepared from 1000 pg ml-I stock solutions of Cd Cr Cu Mn and Pb (as nitrate) and Mo (as molybdate) supplied by BDH Chemicals (now Merck) Poole Dorset UK.Each standard was prepared to contain 2% vlv nitric acid. A 100 pg ml-I Pd(N03)2 solution was prepared by dilution of a 10% m/v Pd(N03)2 standard supplied by Johnson Matthey Royston UK. Aristar HN03 was obtained from BDH Chemicals. Analysis of Reference Materials Four National Bureau of Standards (NBS) [now National Institute of Standards and Technology (NIST)] Standard Reference Materials (SRMs) 1577a Bovine Liver 1572 Citrus Leaves 1566 Oyster Tissue and 1575 Pine NeedlesJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY.OCTOBER 1993 VOL. 8 99 1 TJ Wavelength interval (0.38 nm) Fig. 3 3-Dimensional intensity plot (intensity versus and time) - wavelength t Wavelength interval (0.38 nm) Fig. 4 3-Dimensional absorbance plot (absorbance versus wave- length and time) were analysed. The samples were prepared for analysis by microwave digestion as described previ~usly,~ using a domestic microwave oven (650 W) and poly(tetrafluor0- ethylene) (PTFE) reaction vessels supplied by Valtech Plastics Thirsk UK. A mass of 0.1 g of SRM was placed in a reaction vessel and 4 ml of concentrated HN03 were added. Five vessels could be placed in the microwave oven each time. The digests were heated at full power to a temperature of 90 "C which was held for 3 min.The vessels were then placed in an ice-bath and allowed to cool for 10 min. The vessels were opened in a fume cupboard and the contents diluted to 10 ml with distilled water. Analyte concentrations in the diluted samples were determined by reference to the CSAAS signals obtained from the analysis of standard solutions prepared as described above. The SRM and standard solutions were not matched in acid concentrations. Results Peak Volume Measurement Prior to the analysis of samples a high concentration standard was atomized. A three-dimensional intensity display was produced as shown in Fig. 3. This display was observed to identify the scan which exhibited the broadest absorption profile. This individual scan was then displayed and wavelength limits selected to allow integration of the absorbances calculated for each pixel covering the absorp- tion line profile.Half the number of pixels chosen for the AAS measurements were automatically selected on either side of the profile to provide the background intensity. The selection was made such that a small number of pixels between the absorption profile and the background region were not used. The intensities in the background areas were averaged to provide the I value. Log (1Jl) was calculated for each pixel across the absorption profile to give a series of absorbance values. These absorbances were summed to provide a wavelength integrated absorbance. The absorb- ances for a number of scans could be plotted to provide a Table 2 Characteristic mass values for ET-CSAAS using LPDA (all integrated absorbance) Characteristic mass/pg Element This work Ref.8 Cd 3.0 2.2 Cr 7.5 - c u 20 10 Mn 12 Mo 30 - Pb 4.6 - - three-dimensional absorbance display as shown in Fig. 4. From Figs. 3 and 4 the concept of wavelength integrated absorbance is clear; peak volume integrated in terms of both time and wavelength is calculated. The number of pixels required to cover the absorption profile depends on the analyte in question the slit-widths used the resolution of the spectral order and the pixel width. The wavelength interval covered by the LPDA on this order of the echelle was 0.38 nm. Figs. 3 and 4 illustrate an artifact of the use of an LPDA. The curved shape of the background in these figures shows an intensity distribution across the array.A reduction in the intensity response is observed from the centre to the ends of the array. As the spectral efficiency decreases towards the ends of an order9 of the echelle spectrometer each order will show a finite slope when viewed in a direct horizontal direction. The LPDA is positioned horizontally in the focal plane and so this spectral distribution is evident on scans of the array. This apparent curvature will result in negative blank values if left uncorrected as the transmitted intensity across the absorption profile will be greater than the I value measured at the sides of the profile. By scanning the array at the start of each measurement period and correcting for both the curvature of the response and the fixed pattern noise associated with the pixels,8 these signal biases which can cause problems when making analytical measurements are eliminated.Characteristic Mass and Detection Limit Values Characteristic mass values were defined as the mass of analyte expected to give a signal equivalent to an absor- bance signal of 0.0044. As previously mentioned the signals are integrated with respect to time and also in terms of wavelength. The wavelength integration is achieved by summing the absorbances measured by the number of pixels covering the absorption profile. The values obtained for Cd Cr Cu Mn Mo and Pb at the individual optimum atomization temperatures are given in Table 2. The results obtained with the LPDA and PMT detectors are not comparable as they are based on measurements from three- dimensional and two-dimensional signal displays respec- tively. Table 2 also gives the characteristic mass values of Cd and Cu reported by Harnly,8 with an identical LPDA detector.The value for Cd obtained in this work is almost identical to that obtained by Hardy8 and the Cu value is worse by a factor of 2. The analyte detection limits obtained with the PMT and LPDA detectors are given in Table 3. The detection limits were calculated initially on the basis of three times the standard deviation of the blank signal noise. However with the LPDA it was observed that the noise recorded from small absorbance signals was less than that from blank atomizations. Indeed it was possible in practice to obtain better detection limits than those predicted from the calculations based on three times the standard deviation of the blank signal noise.Hence Table 3 also contains992 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 Table 3 Detection limits obtained for ET-CSAAS using LPDA and PMT detection (all integrated absorbance) Detection limitlpg LPDA This work Element I* IIt Ref. 8 PMT* Cd 25 1 1 0.6 9 6 Cr 22 9 c u 19 17 0.8 23 Mn 9 4 - 1.5 Mo 43 24 - 20 Pb 21 10 - 23 - * 30 blank signal noise. $ Varian SpectrAA 40. 30 noise of signal for low mass of analyte. Line source ETAAS* $ 5 3 2 6 12 10.5 detection limits calculated for each of the analytes based on three times the standard deviation of the noise of signals obtained for analyte masses about 10 or 20 times the estimated detection limit values. For most of the analytes the detection limits calculated on the basis of the blank signal noise are 2-fold poorer than the values obtained by the alternative procedure. In this study the detection limits obtained with the LPDA were of a similar magnitude or slightly poorer than those obtained with PMT detection.These results do not compare well with those achieved by Harnly,8 who obtained detection limits of 0.6 and 0.8 pg for Cd and Cu respectively with a similar LPDA; a 20-fold improvement on the results obtained in this work. The major cause of this discrepancy would appear to have been noise sources involved in the electronics of this system that were eliminated in the system used by Harnly.8 Table 3 also gives the detection limits obtained by line source ETAAS with a Varian SpectrAA-40 atomizer at the same wavelengths and with the same individual optimum atomization temperatures.For most of the elements the line source detection limits are 2-fold better than those obtained by CSAAS. The Philips SP9 atomizer used for the CSAAS measurements has copper contacts hence the detection limits for Cu can be adversely affected by contamination. The expected improvement in S/N values due to the use of LPDA detection comes about only from reduced noise. The choice of detector has no bearing on the magnitude of the absorbance signal as long as identical atomizers are used with either PMT or LPDA detection. Therefore the lower noise level obtainable with the LPDA should allow an increase in S/N to be achieved. The read-out noise level estimated by the manufacturer of the LPDA is 0.096 mV.As read-out noise is the limiting noise source (as opposed to the shot noise for PMT detection) it is necessary to ensure that the noise level is as close to this figure as possible by minimizing all other noise components. Noise sources including analogue-to-digital noise quanti- zation noise and computer noise also become more impor- tant when using LPDA detection as opposed to PMT detection owing to a difference in gain when using these two detectors. With a PMT the gain obtained (which can be of the order of lo4 or lo5) renders these noise sources insignificant. This gain level is not obtained with an LPDA detector. Consequently noise sources that could be ignored with PMT detection become very important when using an LPDA and thorough noise studies must be carried out to achieve the S/N improvements possible with this detector.The fact that better detection limits were obtained from the - 909 - ~ 283.1 283.3 Wavelengthhm 283.5 Fig. 5 Single scan (intensity versus wavelength) at the Pb wavelength of 283.3 nm showing anomalous drop-out in integra- tion area. Peak integration area is shown by area I3 to C and background absorbance areas are shown by areas A to B and C to D 250 I 1 v) $ 200 C m e n 150 U 9 100 h UJ w 50 - 0 200 400 600 800 1000 [Cdl/pg I-' Fig. 6 Calibration graph for Cd using A PMT; and B LPDA detection v) 50 8 ' 40 e n m 30 . U w 2 20 P 4- - 10 A 0 [Wpg I-' Fig. 7 Calibration graph for Cr using A PMT; and B LPDA detection atomization of low analyte masses rather than from the blank signal noise is indicative of problems within the system.A number of other problems encountered in this work contributed to the poorer than expected detection limits obtained with LPDA detection. At the amplification ap- plied it was impossible to use the maximum entrance slit dimensions reported as optimum in the work of Harnly et aL4v5 A more serious problem is illustrated in Fig. 5 which shows one scan of the array in the intensity mode when the central wavelength is 283.3 nm. This scan of the array exhibits an anomalous drop in intensity over a small number of consecutive pixels. This occurred at least once for each set of 80 scans of the array made during anJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 993 Table 4 Determination of elements in NBS SRMs by ET-CSAAS and LPDA detection Concentration*/pg g-' ~~ SRM 1577a SRM 1575 Bovine Liver Pine Needles Element Certified Found Certified Found Cd 0.44 0.38 f 0.03 (0.05 NDt c u 158 149f 10 3.0 2.5 f0.2 Mn 9.9 9.1 f 0.7 675 662 & 21$ Pb 0.135 0.11 f 0.01 10.8 10.0 f 0.7 Cr - ND 2.6 2.4 f 0.3 Mo 3.5 3.1 f 0.3 - ND * Mean k one standard deviation; n = 6.t ND=not detected. f Determined after 1 0-fold dilution of initial sample solution. SRM 1572 Citrus Leaves SRM 1566 Oyster Tissue Certified Found 0.003 ND 0.80 0.60f0.08 16.5 16.1 f 1.1 23.0 21.2f 1.6 0.017 ND 13.3 13.1 2 1.2 Certified Found 3.5 3.3 f 0.3 0.69 0.6 1 -+ 0.04 to. 2 0.17 f 0.02 0.48 0.42 ? 0.03 63 52+-4 17.5 15.9f 1.1 1 80 - 60- 0 200 400 600 800 1000 [Cul/pg I-' Fig.8 Calibration graph for Cu using A PMT; and B LPDA detection 0 100 200 300 400 500 400 800 1200 1600 2000 0 [MoYpg I-' Fig. 10 Calibration graph for Mo using A PMT; and B LPDA detection 0 200 400 600 800 lo00 [Mnl/pg I-' [Pbl/pg I-' Fig. 9 Calibration graph for Mn using A PMT; and B LPDA detection detection Fig. 11 Calibration graph for Pb using A PMT; and B LPDA atomization event. If the irregular intensity drops occur in regions of the array selected for calculation of I. or I the precision of replicate measurements is impaired. As this was a random phenomenon it was difficult to correct. Calibration Graphs As the ET-CSAAS detection limits obtained for each of the analytes with both detectors are of a similar magnitude it was considered valid to compare the linear ranges of the calibration graphs.The results are illustrated in Figs. 6-1 1 for Cd Cr Cu Mn Mo and Pb. The individual optimum atomization temperatures mentioned previously were used for each of the elements. The figures indicate that integrating the signals by wavelength as well as time improves the linear range significantly for all of the elements. With mechanical modulation and PMT detec- tion only the absorbance at the line centre is used and so deviations from linearity occur at comparatively low concentrations. This extension of the linear range of the calibration graphs is similar to that obtained by Miller-Ihli et a1.I0 using wavelength modulation to measure the absorbance at several points across a single absorbance profile.994 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL.8 Analysis of Reference Materials with LPDA Detection The compromise atomizer programme given in Table 1 was used for analysis of the NBS SRMs. This programme was developed in a previous study for determination of the elements by simultaneous multi-element ET-CSAAS.3 The concentrations given in Table 4 are quoted as the mean f one standard deviation (n= 3). The sample solutions prepared as described under Experimental were analysed without dilution for all of the elements with the exception of Mn in SRM 1575 Pine Needles. In this case the sample solution had to be diluted 10-fold prior to analysis. In a previous study," with PMT detection dilution of the sample solution was required for determination of Cu in SRM 1577a owing to the poor linearity of the calibration graph.With LPDA detection however dilution was unnecessary and a single solution of SRM 1577a could be used for the determination of all the elements considered. In general reasonable comparison was obtained between the certified values and the derived concentrations for each of the analytes. The poorest comparisons were obtained for Cu and Mn in SRM 1566 Oyster Tissue. The standard deviation values quoted in Table 4 are of a similar magnitude to those obtained when the reference materials were analysed using the same compromise atomizer pro- gramme and simultaneous multi-element CSAAS with PMT detection." Despite the good performance of the technique the ET- CSAAS detection limits for Cd and Mo were not low enough to allow detection of these elements in SRM 1575 (Cd) and 1572 (Cd and Mo) where the analyte concentra- tions were less than 0.05 pg g-l.Conclusions There are analytical and instrumental advantages in using an LPDA detector for ET-CSAAS. The detection limits obtained although similar to those achieved with a PMT detector and mechanical wavelength modulation were not improved as expected which emphasizes the need for detailed noise studies. With LPDA detection it is possible to integrate the absorption signal in 3-dimensions by summing the absorbances obtained for pixels covering the line profile. This increases the linear range of the ET- CSAAS calibration graphs and thereby improves the appli- cation of the technique in trace element determinations as illustrated by the analysis of reference materials. Several software modifications are required to eliminate problems caused by the LPDA artifacts such as pixel sensitivity variations. The cause of the intensity drops observed for some pixels is not properly understood but might be an intrinsic feature of the LPDA operation and hence unavoidable. References 1 Harnly J. M. Anal. Chem. 1986 58 933A. 2 Marshall J. Ottaway J. M. and Littlejohn D. Anal. Chim. Acta 1986 180 357. 3 Littlejohn D. Egila J. N. Gosland R. M. Kunwar U. K. Smith C. M. M. and Shan X.-q. Anal. Chim. Acta 1991,250 71. 4 Moulton G. M. O'Haver T. C. and Harnly J. M. J. Anal. At. Spectrom. 1989 4 673. 5 Moulton G. M. O'Haver T. C. and Harnly J. M. J. Anal. At. Spectrom. 1990 5 145. 6 Jones B. T. Mignardi M. A. Smith B. W. and Winefordner J. D. J. Anal. At. Spectrom. 1989 4 647. 7 Fernando R. Calloway C. P. Jr. and Jones B. T. Anal. Chem. 1992,64 1556. 8 Harnly J. M. personal communication 199 1. 9 Zander A. T. Miller M. H. Hendrick M. S. and Eastwood D. Appl. Spectrosc. 1985 39 1. 10 Miller-Ihli N. J. O'Haver T. C. and Harnly J. M. Anal. Chem. 1984,56 176. 1 1 Nichol R. Smith C. M. M. and Littlejohn D. in preparation. Paper 3/01 36 I H Received March 8 1993 Accepted June 2 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800989

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 99 5 ln Sjfu Concentration of Selenium and Tellurium Hydrides in a Silver-coated Graphite Atomizer Ni Zhe-ming He Bin and Han Heng-bin Research Center for Eco-Environmental Sciences Academia Sinica P. 0. Box 2871 Beijing China A method is described for the sensitive determination of Se and Te by hydride generation with subsequent trapping and atomization in the graphite atomizer coated with silver. The atomization temperature for the Se and Te trapped on the silver surface occurred at 1800 "C lower than that from the palladium surface (~2000 "C). The trapping temperatures for Se and Te are 200-400 "C and 200-800 "C respectively. The characteristic masses are found to be 17 and 18 pg for Se and Te respectively. The recommended method can be applied to the determination of Se and Te in biological samples with concentrations greater than 20 ng g-l.Keywords Electrothermal atomic absorption spectrometry; in situ concentration; silver-coated graphite tube; selenium and tellurium hydride; biological samples Hydride generation (HG) coupled with in situ concentra- tion of the hydride in a palladium-coated graphite tube followed by atomization is recognized as one of the most sensitive techniques for the determination of trace levels of hydride forming elements. The method has been success- fully applied to the determination of organic and inorganic As in water,' Se in water and urine,2 Sb and Se in water,3 Bi Ge and Te in stream sediment standard reference ma- t e r i a l ~ ~ and Sn in steel and biological sample^.^ The mechanism of sequestration of the hydrides of As Se Bi Sb and Sn was studied by Sturgeon et aL6 They found that the catalytic reactivity of Pd promoted low-temperature de- position of the hydrides by dissociative chemisorption and the fact that reduced Ag Au Cu or Ni is incapable of sequestering the analyte hydride further supports their assumption. In an attempt to study the trapping capability of metals other than the platinum group it was found that although arsenic hydride was not efficiently trapped on the silver coated graphite tube the hydrides of Se and Te can be well sequestered on the surface at a temperature of 200-400 "C for Se and 200-800 "C for Te.Since Pd is more expensive than Ag the latter can be a less expensive alternative for the trapping procedure.This paper presents the analytical performances of the HG method with in situ concentration in a graphite tube treated with silver nitrate followed by atomization. The procedure was validated by the analysis of standard samples. Experimental Apparatus A Perkin-Elmer Model 4000 atomic absorption spectro- meter and a Model HGA-400 graphite furnace with deuterium background correction were employed for the measurement of Se and Te absorbances under the condition of 'gas stop'. High intensity lamps (General Research Institute of Non-Ferrous Metals Beijing China) were used as line sources. A spectral slit-width of 0.7 nm was used to isolate the 196.0 and 214.3 nm lines for Se and Te respectively. Pyrolitic graphite coated graphite tubes were used throughout.The sample introduction port in the middle of the tube was enlarged to a diameter of 2.5 mm with a drill bit for the insertion of the quartz tube connected with the outlet of the hydride generator HG-100.' Hydride generation was accomplished in a continuous mode in a quartz cell by using two channels of a peristaltic pump to deliver the sample and sodium tetrahydroborate solutions to the generator. The mixing and reaction of these solutions were accomplished simultaneously in the cell. A piece of 4 mm i.d. silicone rubber tube was used for the connection between the outlet of the hydride generator and the quartz delivering tube the tip of which was inserted through the enlarged sample introduction port of the graphite tube and held in contact with the opposite interior wall during the trapping of the hydrides in the palladium- or silver-coated graphite tube.Furnace temperatures were measured by using the output of a phototransistor. The phototransistor was calibrated against an optical pyrometer (Ircon UX-10) which was sighted through the injection hole at the bottom of the furnace. Reagents An Se stock solution 1000 pg mi-' was prepared by weighing 0.1405 g of SeOz (analytical-reagent grade Beijing Chemical China) in 25 ml of 4 moll-' HC1 and diluted to 100 ml with de-ionized water. A Te stock solution 1000 pg ml-l was prepared by dissolving 0.1251 g of Te02 in 5 ml of concentrated HCl and diluted with de-ionized water to 100 ml. These stock solutions were diluted and acidified daily to form the final standard solutions in the ng ml-' concentra- tion range.Sodium tetrahydroborate solution 2% m/v was prepared daily or more frequently by dissolving NaBH in de-ionized water and was used without filtration or stabilization. Palladium chloride solution (0.100 mg ml-I of Pd) was prepared by dissolving PdC12 in dilute HN03 and diluting with de-ionized water. Silver nitrate solution (0.100 mg ml-I of Ag) was prepared by dissolving the reagent in de-ionized water. All chemicals were analytical-reagent grade. Sample Preparation A sample of 0.200 g was mixed with 5 ml of concentrated HNO and 0.5 ml of concentrated HClO in a 50 ml flask. A small funnel was placed in the neck of the flask to avoid excessive loss of acid during sample decomposition.The temperature was maintained at about 95 "C for 2 h until about 0.5 ml of HClO remained in the flask. The Se and Te were reduced to Sew and Tew by adding 20 ml of 6 moll-' HCl and kept at 95 "C for 20 min. Then the solution was diluted to 25 ml with de-ionized water. Aliquots of 2.5 ml of the solution were placed in 10 ml calibrated flasks. To the solutions were added 2.5 5.0 and 10.0 ng of Se or 0.4 1.0 and 2.5 ng of Te respectively to allow for the use of the standard additions technique. The calibrated flasks were then diluted to the mark. The final acidities of the solutions996 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 Table 1 Furnace programme for Se and Te Step Tempera t ure/"C Ramp/s Holds Procedure 1 150 5 50 Inject AgN03 solution 2 300 5 120 Insert quart tube HG 3 1800 0 5 Remove quartz tube 4 2650 1 3 Clean the graphite tube and drying atomization 0.30 t I - - - 8 0.20 e 2 2 0.10 C 0 D i t I 1 1 200 600 lo00 1400 1800 2200 TemperaturePC Fig.1 Trapping and atomization curves of Se on silver- and palladium-coated tubes A Se on silver trapping curve atomiza- tion temperature 1800 "C); B Se on silver atomization curve (trapping temperature 300 "C); C Se on palladium trapping curve (atomization temperature 2200 "C; and D Se on palladium); atomization curve (trapping temperature 400 "C) 0.30 8 0.20 .f! 2 2 0.10 m 0 0 2.5 Time/s 5.0 Fig. 3 Absorption profiles for Se from A silver- and B palladium- coated tubes A B 1 0 D I I I 1 I I 200 600 lo00 1400 2000 2400 TemperaturePC Fig. 2 Trapping and atomization curves of Te on silver- and palladium-coated tubes A Te on silver trapping curve (atomiza- tion temperature 1800 "C); B Te on silver atomization curve (trapping temperature 300 "C); C Te on palladium trapping curve (atomization temperature 2200 "C); and D Te on palladium atomization curve (trapping temperature 400 "C) were 1 mol 1-* in HCl for Se and 3 moll-' for Te.A blank solution containing only the acids was prepared in the same manner. No Se or Te was detected in the blank. Procedure The recommended experimental conditions for HG and stripping were as follows uptake rate of sample solution and 2% m/v NaBH 4.5 ml min-I; and carrier gas flow rate 580 ml min-l for Se and 820 ml min-l for Te. The furnace programme for the collection and atomization of Se and Te is listed in Table 1.Results and Discussion Sorption and Atomization Temperature The effect of sorption temperature on the absorbance of analyte in both palladium- and silver-coated graphite tubes has been studied. The results are shown in Figs. 1 and 2. As 2.5 Time/s 5.0 Fig. 4 Absorption profiles for Te from A silver- and B palladium- coated tubes can be seen the optimum temperature ranges for trapping of Se and Te on the silver-coated tube are narrower as compared with a palladium-coated tube Se being more obvious. However in the trapping procedure the stabiliza- tion temperature of the analyte on the substrate seems to be not as important a factor as the pyrolysis temperature in the direct injection method. In the former case a large amount of salt which would interface with the atomization of the analyte has already been separated by the HG procedure and therefore a high ashing temperature is not so critical.The atomization curves were constructed by holding the trapping temperature at an optimum value and varying the atomization temperature. The results indicate that atomiza- tion of Se and Te is complete above 1600 "C which is lower than that observed in a palladium-coated graphite tube (> 2000 "C). The lower atomization temperature is advan- tageous when the lifetime of the the graphite tube is considered. The graphite tube could last for 120 firings for a silver-coated tube as compared to 90 firings when a palladium-coated tube is used. Figs. 3 and 4 present the absorbance profiles for Se and Te from silver and palladium surfaces respectively.It is clearly shown that the absorbance peaks of Se and Te from silver appear 0.2 s earlier in time than those from the palladium surface. This is to be expected as the analyte adsorbed on silver (m.p. = 962 "C) will vaporize at a lower temperature than that on palladium (m.p. = 1552 "C).JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 997 Table 2 Determination of Se and Te in reference samples; each result is mean &standard deviation of three determinations Concentration of analyte/pg g-I Se Te Sample Found Certified Found Certified Mussel (GBS 08571) 3.63 k 0.04 3.65 f 0.17 (0.020 - Rice Flour (NBS 1568) 0.40 +- 0.04 0.4 & 0.1 ~ 0 . 0 2 0 - Garlic Powder 0.1 1 1 k0.017 - (0.020 - Oe30 * 8 ; 0.20 2 9 / B 1 I I I 1 0 1 2 3 4 5 [HCll/mol I-' 0.10 Fig.5 Effect of acid concentration on the peak absorbance of A Se and B Te Effect of Acid Concentration The influence of acid concentration in the generation of selenium and tellurium hydride and trapping on a silver- coated graphite tube was studied. Results shown in Fig. 5 indicate that the recovery of Se is complete when the acid concentrations are kept between 1-4.5 mol 1 - I HCl while for Te the acidity should be maintained above 2.5 mol 1-1 HCl to achieve complete recovery. This result is in agreement with that reported earlier in conventional HG atomic absorption spectrometry (HGAAS).8*9 In a subse- quent experiment the acidity was kept at 3 mol 1-1 for the generation of tellurium hydride and 1 mol 1 - I for selenium hydride to avoid pollution due to the escape of excessive acid fumes.Effect of Silver Concentration Different amounts of silver were added to trap the analyte hydrides in the graphite furnace. The results show that 100 pl of 100 pg ml-I or 10 pg of Ag were sufficient to trap nanograms of selenium and tellurium hydride which is double the amount of palladium usually employed to stabilize the analyte hydrides. Analytical Figures of Merit The characteristic masses defined as the mass of analyte which provides a defined peak absorbance of 0.0044 for Se and Te on a silver-coated tube were found to be 17 and 18 pg respectively similar to those reported previously using a palladium-coated t ~ b e . ~ * ~ - ~ ~ The detection limits based on the variability of the blank (30) were 16 pg for Se and 15 pg for Te the latter being lower than that obtained from an uncoated tube,I1 but higher than the data reported from a modified Varian Techtron CRA-90 atomizer.12 The regres- sion equations for the calibrated curves were constructed for both analytes under optimum conditions by the least squares method; y = 0.2656~- 0.0082 with a correlation coefficient of 0.9997 and y=0.242~ with a relative coeffici- ent of 0.9999 were obtained for Se and Te respectively where x=analyte mass (ng) and y=peak height absorbance A,.The precision of the method was evaluated by replicate determinations of 1 ng of the analytes. The relative standard deviations of ten replicate determinations were 2.6% for Se and 2.1% for Te.Sample Analysis To assess the applicability of the proposed method well established reference materials are required to carry out a comparative analysis. Despite the availability of the certi- fied values of Se in various biological and environmental samples it is difficult to find standard materials with known Te concentrations. Since Te is present at very low concen- trations in environmental and biological samples and direct determination of Te by electrothermal AAS (ETAAS) is complicated by its volatility and matrix interferences a separation and concentration procedure is usually used before determination. Tellurium in urine has been sepa- rated by solvent extraction with isobutyl methyl ketone (IBMK),I3 the limit of detection defined as twice the standard deviation of the blank reported was 5 ng for Te which is relatively high.Weibust and Langmyhr14 at- tempted to determine Te in human blood and garlic by solid sampling using Pd as a chemical modifier but the concentrations were found to be below the detection limits of 3 ng ml-I and 104 ng g-l respectively. The determination of Se and Te was carried out in three reference samples Research Centre for Eco-Environmental Sciences Academia Sinica and China National Bureau of Oceanology Mussel GBW 08571; National Bureau of Standards (NBS) (now National Institute of Standards and Technology) 1568 Rice Flour; and Garlic Powder (Beijing) using the proposed procedure and the results are shown in Table 2. As can be seen the data for Se agree well with the certified values while the concentrations of Te are below 20 ng g-' of the sample.In order to investigate the applicabil- ity of the proposed method to the above samples a recovery test was carried out and the results (Table 3) indicate that recoveries of Te ranged from 98-1 14% and concentrations Table 3 Recovery of Te (ng g-l) added to samples; each result is mean +- standard deviation of three determinations Sample Added Found Recovery (O/O) Mussel (GBW 0857 1) 10 9.8 20.5 98k5 125 142 f 5 114k4 (NBS 1568) 10 9.8 k 0.5 98k5 125 125k3 look2 Rice Flour Garlic Powder 10 9.8 k0.5 98k5 125 122-t3 9 8 k 2998 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1993 VOL. 8 of Te as low as 20 ng g-l in the sample can be determined by the present procedure. Conclusion The proposed method to trap selenium and tellurium hydride in a silver-coated graphite atomizer followed by electrothermal atomization can be used for the determina- tion of trace amounts of Se and Te in biological samples.Silver can be used as an alternative for palladium in the trapping procedure for selenium and tellurium hydrides. This work was supported by the Chinese Academy of Sciences under Contract No. KM 85-47. References 1 Han H.-b. Liu Y.-h. Zhang S.-z. and Ni Z.-m. J. Environ. Sci. 1993 5 99. 2 Ni Z.-m. He B. and Han H.-b. Can. J. Spectrosc. in the press. 3 4 5 6 7 8 9 10 11 12 13 14 Zhang L. Ni Z.-m. and Shan X.-q. Spectrochim. Acta Part B 1989,44 339. Zhang L. Ni Z.-m. and Shan X.-q. Spectrochim. Acfa Part B 1989,44 751. Zhang L. McIntosh S. Carnrick G. R. and Slavin W. Spectrochim. Acta Part B 1992 47 701. Sturgeon R. E. Willie S. N. Sproule G. I. Robinson P. T. and Berman S. S. Spectrochim. Acta Part B 1989 44 667. Yan X.-p. and Ni Z.-m. J. Anal. At. Spectrom. 199 I 6,483. Nakahara T. Prog. Anal. At. Spectrosc. 1983 6 163. Fleming H. D. and Ide R. G. Anal. Chim. Acta 1976,83,67. Doidge P. S. Sturman B. T. and Rettberg T. M. J. Anal. At. Spectrom. 1989 4 25 1. Andreae M. O. Anal. Chem. 1984 56 2064. Yoon B. M. Shim S. C. Pyun H. C. and Lee D. M. Anal. Sci. 1990 6 561. Kobayashi R. and Imaizumi K. Anal. Sci. 1990 6 83. Weibust G. and Langmyhr F. J. Anal. Chim. Acta I98 I 128 23. Paper 3/01630G Received March 22 1993 Accepted May 27 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800995

出版商:RSC    

年代:1993

数据来源: RSC