摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 517 Background and Background Correction in Analytical Atomic Spectrometry Part 1 m Emission Spectrometry A Tutorial Review J. B. Dawson* and R. D. Snook Department of Instrumentation and Analytical Science University of Manchester Institute of Science and Technology P.O. Box 88 Manchester M60 lQD UK W. J. Price 15 Amberley Close Holne Cross Ashburton Devon TQ13 7JE UK Summary of Contents Introduction Background and General Principles of Background Correction in AES Indications of presence of background Measurement of background Instrumental considerations Background fluctuations ‘noise’ in AES Flames Arcs and sparks Plasma sources Source-generated Background in AES and its Correction D.c. plasmas Inductively coupled plasmas Microwave-induced plasmas Low-pressure discharges Photographically recorded spectra Instrument-generated Background in AES and its Correction Spectrograph Photographic emulsion Microdensitometry Computation Spectrometer Photoelectric detection of spectra Wavelength selection Non-dispersive systems Filter instruments Michelson interferometers Dispersive systems Background correction devices Photomultiplier tubes Solid-state detectors Detectors Signal and data processing Conclusion References Keywords Background correction; emission spectrometry; review Introduction the net signal to derive an analyte concentration. The recognition of the existence of a background and the effectiveness of its correction are essential in ensuring the accuracy and precision of an analysis.This review (Part comprehensively examines of background and procedures for background correction in atomic emission and its correction in atomic absorption spectrometry (Part 2) and atomic fluorescence spectrometry (AFS) (Part All measurements require an ‘origin’ ‘zero’ or ‘baseline’ as a reference starting point. In quantitative spectrochemical analysis spurious radiations from many causes can occur at all wavelengths. Background correction is therefore fre- implicit if not explicit in all serious determinations. In Its simplest form it is the subtraction of the signal generated by quentlY required to determine that starting Point and .is spectrometry (AES). A companion review of background the background from the before Of 3) will be published later.Matrix interference effects i.e+ modification of the response of the system to the analyte * To whom correspondence should be addressed at the follow- (sensitivity) by the components of its matrix will not be ing address 26 Harecroft Road Otley West Yorkshire LS2 1 2BQ considered. The subject matter in this review will be UK. presented under three main headings Background and518 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 General Principles of Background Correction in AES; Source- generated Background in AES and its Correction; and Instrument-generated Background in AESand its Correction. The literature on analytical atomic spectrometry has been accumulating for well over a century and currently leads to more than 1000 publications per annum.Background effects are recorded implicitly or explicitly in many of those papers hence it would be impractical for any work on this subject to quote every relevant reference. Therefore in this review representative references are quoted to give an entry into the literature rather than an exhaustive listing of all the material consulted. Attention will be drawn to earlier work where appropriate to encourage the reader to solve the problems of today by learning from the past rather than by repeating previous studies. This approach is particularly relevant where modern electronic and computing power can carry out operations that were lengthy or impractical in earlier work. Inevitably with such a voluminous literature we have not read in full all the papers consulted particu- larly those in foreign languages.In these instances abstracts have had to be relied upon to provide the necessary information. Further in compiling this review rather than attempting to foresee all eventualities and recommend solutions we have sought to provide the reader with the general information necessary for solving the specific problems that arise in his or her own laboratory. Thus fundamental concepts such as ‘resolving power’ ‘noise’ and ‘detection limits’ are presented in addition to the origins of background and means for its reduction or correction. Most sections are self-contained and may be taken in any sequence by the reader. Background and General Principles of Background Correction in AES Background signals occurring in emission analysis arise from emission of non-analyte species scattered radiation and the detector characteristics. In absorption and fluores- cence analysis the background signal is more complex. The instrumental techniques employed for background correc- tion in AAS and AFS are therefore substantially different from those used in AES.Indications of Presence of Background One of the most important and difficult stages in back- ground correction for a new analysis is the accurate identification and measurement of the background signal. When instruments with built-in background correction are in use this stage may ironically be more difficult than with a simpler system and could even lead to erroneous conclusions if the automatic background correction proce- dure is not fully understood by its user.Preparation of a direct-reading instrument for analytical use commonly requires adjustment of the electrical zero settings of the analytical and where available background channels with- out and with the atom source operating. Care should be taken to ensure that the optical conditions implicit in those settings are fulfilled e.g. setting a ‘zero’ with an emission source operating is a background correction procedure in itself. If the background is small and reproducible and the analyte signal relatively large background correction may be effected simply by subtraction of a previously measured ‘blank’ or by ‘zeroing’ the instrument using a blank sample. When the background is large and unstable relative to the analyte signal i.e.near the detection limit then sophisti- cated signal processing or even chemical separation of the analyte from its matrix may be necessary. The background component of the signal from a spectro- metric system may have many causes in the analytical process. These causes include contaminated reagents and apparatus carryover from one sample to another wave- length coincidence of radiation emitted or absorbed by atoms and molecules in the analytical vapour stray light in the spectrometer extraneous optical signals dark and non- linear responses of the detector electrical pick-up and non- linearity in the signal processing and display circuitry and finally reading and calculation errors. The detection and correction of the latter errors e.g. ‘rounding’ errors will not be considered in this review but the possibility of their existence should not be overlooked if anomalous results are being obtained. If the response function of the system is sigmoid e.g.that of the photographic plate the interpreta- tion of the combined background plus analyte signal is complex. At low and high concentrations the increase in apparent analyte signal produced by the background is greater than at mid-range concentrations where the gradient of the detector response curve is steeper. The existence of a background signal at the analytical wavelength may be detected by analysis of a representative sample known to contain a negligible amount of the analyte element a ‘blank’. When this approach is not possible alternatives include the analysis of reference materials and synthetic samples care being taken to ensure that the reagents employed are not contaminated.Discrepancies between known and observed analyte concentrations may indicate a background signal and merit further investiga- tion by alternative methods. Spectral scans in the neigh- bourhood of the analytical line and literature searches for possible line overlap from matrix elements and for the experience of others in carrying out the same analysis should serve to identify the probable source of the back- ground. The two most likely sources are other components in the sample itself and the atomization ‘cell’ compounded by inadequacies in the wavelength-selection system. Complete correction for background in either photogra- phic or photoelectric spectrometry can only be inferred if the plot of line intensity (or intensity ratio if an internal standard is used) against accurately known concentrations for a set of standard samples passes through the origin.Measurement of Background Direct measurement of the background occurring at the analytical wavelength at the same time as measurement of the analytical line is impossible in emission analysis. The nearest realization of this ideal is background correction in atomic absorption by means of the Zeeman effect and its polarization properties.’ When the background in emission is stable and the same from sample to sample and the same for the standard and blank materials then background correction can be effected by measurements at the analyti- cal wavelength alone.The samples standards and blank are measured in turn and then the blank is subsequently subtracted. When the background is variable from sample to sample accurate background correction requires mea- surement at additional wavelengths. For a continuum background measurements adjacent to the line may suffice. Where there is line overlap measurement of another line in the interfering element’s spectrum will be required from which if the relative intensities of the measured and overlapping lines are known an estimate of the latter’s contribution to the background at the analytical wavelength may be made. Usually background signals differ from one region of the atom source to another and therefore the same region should be used for both analysis and back- ground correction. However occasionally background cor- rection based on an average signal from the whole atom source may be satisfactory.When accurate measurement of the background is not possible the problem of background correction may be solved by separation and if possible concentration of theJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY. JUNE 1993 VOL. 8 519 analyte from its matrix. This can be achieved chemically chromatographically or at the point of atomization itself by selective volatilization with the aid of matrix modifiers. Inevitably sample pre-treatment adds to the time taken and complexity of the analysis and to the possibility of error through contamination or loss of the analyte. Thus an instrumental solution e.g. increasing the wavelength selec- tivity (resolution) of the system until comparable to atomic linewidths thereby separating the analyte wavelength from the background spectral features is usually sought. Instrumental Considerations When the background originates in the atom cell e.g.emission or absorption by plasma or flame gases and the sample is continuously fed to the cell then modulation periodic or transient of the sample feed may be used to discriminate against background in direct-reading sys- t e m ~ . ~ . ~ However if the background is a property of the sample itself this modulation affects the background also and this approach therefore requires careful evaluation. Modulation of the radiation from emission sources facili- tates signal processing and discrimination against electrical background signals. Continuously alternating measurement of the analytical line and background intensities is also used to effect background correction.For this purpose it is advantageous for the modulation frequency to be as high as practicable (>300 Hz) in order that measurements of the analytical and background signals are as near simultaneous as pos- sible. If the modulation produces a sinusoidal signal a ‘lock-in’ or ‘tuned’ amplifier may be used to detect the net line minus background signal. When the modulation generates a square signal waveform gating circuits and digital electronics may be used to measure line and background separately and to carry out the necessary background correction computation. In emission systems wavelength modulation by optical means is sometimes used to modulate the analytical line signal whereas in absorption and fluorescence electrical modulation of the power supply to the light source is used to modulate the intensity of the analytical line.Whatever system of modulation is em- ployed irregularities in it will contribute to fluctuations in the detected signal. If the fluctuations in the line and background signals are correlated then correction for these fluctuations is as near complete as possible when either a photographic plate or a photodetector array is used to record simultaneously the line and background or in a direct-reading system where separate channels are used to integrate both line and background signal. In a direct- reading system too real-time background subtraction can be effected.This however will not give better background correction than subtraction at the end of the completed exposure. If the background signal noise is dominated by random photon events no correlation with the analyte signal noise will exist and there will be no advantage in simultaneous measurement of analyte line and background other than the saving of time. When a scanning spectro- meter is used in emission analysis unless a high-frequency wavelength modulation is superimposed correlation be- tween signal and background noise is limited to low frequencies (< 1 Hz). On-line data acquisition with digital computers has greatly facilitated complex signal pro~essing.~ The signal-to- background ratio (S/B) of a spectral scan may be maximized in a manner analogous to one used for minimizing the relative standard deviation (RSD) of measurements of the peak in electrothermal atomization AAS.S This procedure uses digital filtration to achieve optimum precision in the measurement of peak and baseline signals.The spectrum scan data are combined either with a correlation function corresponding to the scanned line profile as produced by the spectrometeI-6 or more simply a linear sliding mean is calculated using a wavelength interval corresponding to half of the full width of the line at half the maximum value. This procedure is not adequate to cope with line overlap problems and more sophisticated spectrum stripping proce- dures are then required e.g. Kalman Background Fluctuations ‘Noise’ in AES If a background signal were absolutely constant correction for it would be much simplified.In reality it suffers from both random (stochastic) and systematic (flicker) fluctua- tions (‘white’ or ‘shot’ and ‘coloured’ or ‘proportional’ noise respectively) as does the analytical signal. These fluctuations are of greatest significance in determinations made near the detection limit and when high-precision analysis is attempted. An authoritative discussion of the definition estimation and use of detection limits has been prepared by the Analytical Methods Committee of the Royal Society of Chemi~try.~ The detection limit (qL) is defined as ( 1 ) where qL is the amount of analyte oM is the standard deviation of the blank measure K is a numerical constant determined by the desired confidence level and G is the gradient of the response curve near the detection limit.By rearrangement of eqn. (1) it can be seen that the RSD of a measurement at the detection limit (srL) is equal to 1IK KOM qL= G The above relationship will be returned to later following a theoretical analysis of the effect of background and noise on measurement reproducibility. In this analysis we shall follow an approach similar to that of Winefordner et a1.lo Our analysis will be limited to consideration of an induc- tively coupled plasma (ICP) source and a spectrometer with a photomultiplier as detector. Representative numerical values will be inserted into the theoretical expression so that the relative importance of the noise sources may be determined. An analogous approach could be employed to determine the effect of noise in the photographic or solid- state detection of spectra.A comprehensive study of SIB in ICP-AES has been published by Boumans et af.1Jv12 In the following simple analysis all signals will be expressed in terms of the photomultiplier output current i,. Thus i = is+ ib+ id + i where the subscripts indicate output currents arising from the spectral line s and its background b in the plasma and dark events d and insulator leakage 1 in the photomultiplier. Assuming the noise on these signal sources to be random and independent then the standard deviation of i a will be (3) The noise on the spectral line as and background $ signals has two components. One component is propor- tional to the signal and arises from random or pseudo- random fluctuations in the radiation output of the source. The RSDs for these effects will be designated k for the spectral line and kb for the background for a data collection interval of unit time.For a time interval t the RSDs become k s / f i and k b / f i . Sources of such noise in the plasma include fluctuations in analyte transport processes gas pressures plasma gas rotationI3 and discharge pulsa- tions. In the ICP these are the most important sources of520 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 n o i ~ e . ~ ~ J ~ The other component of noise originates in the intrinsic randomicity of the photon flux and the release of photoelectroncs from the photocathode which to a first approximation is proportional to the square root of the photoelectron flux from the photocathode.If an observa- tion time t is assumed then and where e= electron charge and g=gain of photomultiplier. At room temperature the dark current has two compo- nents. One id originates in the thermal release of electrons from the photocathode and other random events in the photomultiplier tube. The other component i arises from current leakage across the photomultiplier base. The noise on both currents is a function of the square root of the current but in the former it is modified by the tube gain hence Combination of the above expressions into eqn. (3) gives (4) The RSD s of an analytical determination can be ex- pressed in terms of the noise analysis as The factor 2 arises because the background is measured twice once as the blank and once with the sample.By assuming the following representative values for the para- meters in eqn. (4) the expression may be used to assess their relative importance under a variety of operating conditions e= 1.6 x C g=2 x lo7 (ref. 16) t = I s i,=O-10-4 A (ref. 16) ib= A id=2x A (ref. 16) i,= A (ref. 16) ks= 5 x (1 s sampling period) kb=2 x (1 s sampling period). It is apparent that the dark current ( i d = 2 x A) and the effect of the leakage current illg (=5 x A) are small compared with the current generated by the plasma background (ib= A) and therefore may be ignored. This reduces eqn. (5) to Eqn. (6) can be used to examine the relationship between srt and the SIB i.e. islib and is presented graphically in Fig. 1 . It is worth noting that this approach when combined with eqn.( 2 ) provides a continuous function linking analytical precision at high concentration to the detection limit of the instrumental system. It is also a means for identifying the relative contributions of signal and back- ground noise to measurement precision. Curve A was calculated using the numerical data given above and shows two near-linear sections one with near zero gradient and the other with a gradient of unity. The horizontal portion of the curve arises from the dominance of the proportional noise (k,=0.5%) of the spectral line. The sloping section reflects a constant level of background noise (kb=0.2%) and 1x102 10 1 1 ~ 1 0 - l I X I O - ~ 1x10-3 SIB Fig. 1 Theoretical dependence of precision on SIB of ICP source with photomultiplier detection calculated using eqn.( 6 ) and data presented in the text. A Proportional noise dominant; and B shot noise dominant ii decreasing analytical signal. The shape of this graph is typical of a proportional noise-dominated situation. The curve will be displaced horizontally and vertically for different values of the coefficients of variation k and k,. Curve B was derived from the same data by setting k and k to zero to simulate the situation when photoelectron ('shot') noise is dominant i.e. when i,k,/eg and i,kb/eg<< I . The lower section of B is dominated by the shot noise on the signal and has a gradient of 1/2 i.e. the precision is inversely proportional to the square root of the signal. The upper section of the curve again has a slope of unity resulting from the combination of constant noise on the background and a decreasing signal.Detection limits are frequently calculated using K= 2 as this value approxi- mates to the 90% confidence level. The corresponding RSD is 0.5. From Fig. 1 the values of SIB at the detection limit for the hypothetical proportional and shot noise-limited conditions are 1.25 x I 0-2 and 5 x l OM3 respectively. Elimi- nation of proportional noise from the background signal has lowered the detection limit by a factor of 2.5. The above analysis demonstrates that where it is possible to make accurate measurements of a stable background and to correct for it precise measurements are possible despite an S/B of less than 1 . Under these conditions if the background noise is predominantly proportional noise then a reduction in the background signal ib without reducing the analytical signal will give a proportional improvement in sfI is photoelectron noise limited then the improvement will be proportional to the square root of the reduction in ib.Further as the effect of random noise on analytical precision is inversely proportional to the square root of the observation time t increasing the latter will improve the precision. The use of extended observation times is limited however by instrumental drift and the demands of sample throughput. It is often assumed as it was indeed in the foregoing analysis that the noise in the analytical and background signals is uncorrelated yet as early as 1975 Alkemade et d . 1 7 J 8 noted that correlation may arise in the low-frequency component.Other workers19 using digital Fourier analysis found a strong component in several sources and cross- correlation was used to enhance the signal-to-noise ratio ( S I N ) 5-fold in flame photometry.20 Thus if the back-JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 52 1 ground or a parameter proportional to it can be measured simultaneously with measurement of the analytical line integration or correlation of the two signals for the duration of the observation time may be used to reduce the effect of non-shot noise (proportional noise) in the system and thereby improve the precision and decrease detection limits. A simple electronic circuit a ‘noise canceller’ which has been described by Hobbs,21 could fulfil such a function in direct-reading AES.Source-generated Background in AES and its Correction Background signals of several kinds commonly originate within the excitation source. As emission sources are independent of the rest of the spectroscopic system they will be reviewed separately and from the point of view of the nature of the background generated. Background in emission can arise from several causes spectral continua band or molecular spectra ‘spread’ of local intense lines impurity or interfering elements stray light etc. There are several ways in which the actual background intensity can be reduced. In some instances the conditions within the excitation source can be adjusted to enhance the analyte signal and suppress that of an interfer- ent controlled vaporization may be used to separate the analyte signal temporally from interferant or background signals and in the analysis of solution samples pre- concentration and/or interferent removal may be carried out prior to the analysis.The sources to be reviewed will include flames arcs plasmas and low-pressure discharges. Less common emission sources such as laser-induced plasmas and imploding thin films will not be considered here in view of their limited use and lack of information on their S/B and S/N characteristics. Flames Flame excitation of atomic line spectra is one of the oldest techniques in spectroscopy and is still in use on a modest scale today principally for the determination of the alkali metals. The problems of background effects found in flame systems using nebulized samples have many parallels with those found in the more recent developments in plasma excitation of nebulized samples.In practical analytical terms it is often the fluctuation of the background signal relative to the analytical signal that is the most important factor limiting the analysis. The design of the burner system is important in determining the nature of the flame background and ‘noise’. The pre-mixed system where the fuel and oxidant gases and the sample aerosol are mixed in a chamber prior to combustion in a laminar flow burner is ‘quieter’ with lower background than the more sample-efficient turbulent flow total consumption burner where the mixing process takes place directly in the flame. A background much reduced in comparison with the tra- ditional total consumption burner was achieved in a sample-efficient burner that combined pre-mixing of the fuel (acetylene) and oxidant (dinitrogen oxide) gases with direct injection of the sample A significant part of the background signal is produced by the flame itself and is independent of the sample hence modulation of the sample flow to the flame is a long- established technique for discrimination against back- ground in direct-reading instrument^.^*^*^^ The emission background of the flame is generated principally in the central primary reaction zone and in the secondary external diffusion reaction region the mantle where environmental oxygen contributes to the combustion of the fuel gas.The latter background is eliminated when the flame is shielded from entrainment of room air either by means of a solid barrier such as a quartz tube or a flow of inert gas.24+2s As much of the background from hydrocarbon-fuelled flames arises from carbon-containing combustion products the use of hydrogen as a fuel gas substantially reduces back- ground emission.For example changing from a dinitrogen oxide-acetylene flame to dinitrogen oxide-hydrogen made possible the determination of boron at 249.8 nm with a detection limit of 30 pg 1-1.26 Changes in fuel and oxidant gases also change the flame temperature an increase in which may produce more efficient atomization and excita- tion but it could also lead to a loss of ground-state atoms through ionization and incur a more intense and complex background emission.The chemical environment of the analyte and interferent elements also depends on the fuel and oxidant gases and may be adjusted to enhance or suppress free atom production. The presence of the cyano- gen radical in the dinitrogen oxide-acetylene flame as a scavenger for oxygen reduces the production of refractory oxides. Oxide formation can also be encouraged or reduced by adjusting the fuel gas ratio to produce an oxidizing or reducing environment. The former condition is required for analysis by molecular emission e.g. Ca OH bands -600 nm and the latter for atomic emission. Unfortunately excess of hydrocarbon fuel tends to lead to increased background signals hence the highest S/B is obtained under compromise flame gas conditions. Arcs and Sparks The direct current (d.c.) arc relies on electrical heating to vaporize the sample in the electrode into a low-voltage discharge; the effective excitation temperature is in the region of 4500-8000 K.It is a sensitive system as it can totally consume a relatively large sample but its reproduci- bility is poor. The arc itself has a low background but continuous radiation from incandescent electrodes and cyanogen bands with band heads at 42 1.6 388.3 and 359.0 nm degrading towards the ultraviolet (UV) region can be a problem when graphite electrodes are used. The greatest analytical sensitivity is found in the neighbourhood of the cathode the ‘cathode layer’ where emission intensities are 10-50 times greater than in the central column of the arc.27 Cyanogen bands can be eliminated by the use of ‘pencil self electrodes’ i.e.a pair of electrodes prepared from the metal sample instead of using a graphite or carbon counter electrode.28 To the same end graphite electrodes can be sparked (not arced!) in an atmosphere of pure oxygen.28 In the high-voltage capacitor discharge spark the sample is vaporized by electron and ion bombardment and a higher excitation temperature up to 10 000 K is achieved. As a consequence sample consumption is much lower and the background is higher. Together these characteristics lead to a lower sensitivity than that of the arc. Although consider- able variation from spark to spark occurs,29 the reproduci- bility is greater as the discharge is repeatedly extinguished and struck to different parts of the surface.The high-voltage alternating current (ax.) arc is interme- diate in performance between the d.c. arc and high-voltage spark in its performance. Over the years many variants of arcs sparks and spark-triggered arc hybrid systems have been reported. Some were designed to improve the S/B but more usually improved sensitivity and repeatability were sought . Early studies of the use of high-voltage discharges in metallurgical a n a l y ~ i s ~ ~ . ~ ~ demonstrated that the light em- itted during the first 5-10 p s consists primarily of continu- ous background and gaseous band spectra that die away while the emission of metallic spark lines continues for522 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 40-50 ps. Hence time resolution of individual discharges can be used to discriminate against part of the spark ba~kground.~~ In both arcs and sparks the line-to-back- ground ratio is a function of electrode separation and generally decreases with increasing separation.If however observations are made close to one electrode the separation effect is much less. Over many years magnetic fields have been applied to d.c. arcs with a view to improving their perf~rmance,~~ but they are still not widely used. Precision is improved by virtue of rotation of the discharge over the electrode surface. Improvements in line-to-background ratios have been attributed to an increased residence time of the analyte atoms and ions and to the exclusion of compounds and hot particles of carbon from the main excitation Operation of the arc in an inert atmosphere reduces the production of carbon dioxide and adjustment of the chamber pressure facilitates optimization of the line- to-background ratio.37 The presence of an inert gas (argon) reduces the intensity of the whole spectrum because of the much higher ionization potential in the gap.However enclosure of the discharge in an atmosphere of oxygen can overcome problems from the cyanogen bands mentioned earlier particularly if traces of barium are to be measured at 455.4 nm.28 In some metallurgical analyses a silver elec- trode rather than graphite has been used to overcome this problem. Spectroscopic bufferdcarriers are in general use in arc- based methods and improve repeatability and sensitivity by standardizing volatilization and excitation conditions.Mo- dification of the volatility of the analyte elements and their matrix may produce temporal separation of the analyte and background emissions and hence the possibility of an improved line-to-background ratio. On the other hand buffers may generate an increased spectral background and be contaminated by the analyte element and therefore require selection with some care. Plasma Sources In this section plasma sources will be considered to be those electrical discharges generated in a gas stream and operating at or near atmospheric pressure into which the sample may be continuously introduced as an aerosol vapour or gas. The three principal subject areas are d.c. plasmas inductively coupled plasmas (ICPs) and mi- crowave-induced plasmas (MIPS).D.c. plasmas These sources have evolved from the d.c. arc and fall into two broad categories those where the discharge is confined within a chamber (wall-stabilized arc)38*39 and others where the electrodes are in the The early designs of these sources (‘plasma jets’) have been comprehensively reviewed by F a ~ s e 1 ~ ~ and more recent developments by Ebdon and S p a r k e ~ . ~ ~ The variety of electrode configurations that have been investigated may be a reflection of the difficulty of injecting the sample-containing gas into an established arc Magnetic fields have been used with a view to improving the incorporation of the sample into the dis- charge. In an electrode assembly consisting of a graphite tube anode and a concentric tungsten wire cathode with an external parallel magnetic field the arc current channel rotated at 2-3 ~ H Z .~ ~ The resulting diffuse plasma was easily penetrated by an aerosol. Intense background was reduced by withdrawing the cathode wire into the anode tube. At present open systems appear to be the most widely used of the plasma sources possibly arising from commer- cial availability general ease of operation and the possibil- ity of optimization for individual sample types. Back- ground emission is high from the arc column and conse- quently analytical observations are usually made outside that region either in the tail flame or close to the zone where the sample is injected into the plasma and where the atom vapour density is greatest. A comprehensive approach to the optimization of the SIB by Sparkes and E b d ~ n ~ ~ using simplex optimization followed by univariate search was applied to the determination of magnesium (279.079 nm) in a kaolin slurry.The parameters selected were the horizontal and vertical viewing positions nebulizer gas flow rate plasma sleeve gas flow rate ammonia dispersant concentration lithium buffer concentration and the con- centration of the slurry. It was found that the most critical parameter by far was the viewing position while the optimum gas flow rates were similar to those used in the analysis of sample solutions. Similar results were obtained when the S/B for the determination of copper in serum was optimized by other workers.47 Inductively coupled plasmas As the emission source that has made the greatest contribu- tion to the development of analytical atomic spectrometry over the last decade the ICP has been the subject of intensive study. Remarkably the general form of the ‘torch’ has changed little since that originally developed by Reed48 in the early 1960s for the growing of crystals.Much of the research effort has been directed at techniques of sample introduction development of an understanding of the discharge process and modifications to torch configuration generator circuitry and operating conditions. One of the consistent aims has been to decrease detection limits which generally means increasing the SIB and improving back- ground correction procedures. The background emission of the ICP has several possible components comprising a continuum with coincident and overlapping atomic lines and molecular bands.Spectral scans by Ediger and F e r n a n d e ~ ~ ~ serve to illustrate these effects. In a low-power argon ICP there is a broad maximum at 450 nm decreasing to -2% of the maximum at 200 nm.sO Broad lines (half-widths 0.4 nm) have been attributed to emission from very short-lived (- s) autoionization The selection of the most appro- priate analytical line is the first step toward maximizing the S/B. Tables of analytical lines with corresponding possible overlapping lines have been compiled and proce- dures for wavelength selection for trace analysis by ICP- AES o ~ t l i n e d . ~ ~ ~ ~ ~ Improvements in the S/B can also be achieved by ‘tuning’ the plasma operating conditions and by modification of the sample.Some of the results of attempts to improve SIB ratios by such approaches will now be summarized. It has been reporteds4 that operation at 50 rather than 27.12 MHz increases the SIB 3-15-fold with further improvement claimed when a micro-torch was operated at 80 M H Z . ~ ~ This frequency effect has been attributed to a lower plasma temperature at the higher frequency which reduces the intensity of the background continuum. At still higher frequencies e.g. 148 MHz the S/B was found to be less than at 27 M H Z . ~ ~ Over the years modifications to ICP torch design have been examined with a view to reducing the background signal. These have included laminar in place of tangential gas flows and extended torch tubes to exclude entrainment of room air. The former reduces the noise on the back- grounds7 while the latter reduces and simplifies molecular band emission.s8~s9 Following early optimism,60 end-on (‘axial’) viewing of the plasma has not led to improvements in detection limits i.e.to higher S/B and S/N ratios,61 owing in part at least to the intense background emission ofJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 523 the toroidal plasma although this can be reduced by using a limiting aperture at the entrance slit of the spectrometer. As it is the noise on the background signal that deter- mines the detection limit of an analytical method deter- mining the source of that noise is central to any endeavour to improve detection limits. From studies of the noise power spectrum of the argon ICP it has long been recognized that background noise is dominated by low- frequency fluctuations that are aerosol borne and originate in the nebulizer-spray chamber system. 1 4 9 1 5 The noise power per unit frequency bandwidth is inversely propor- tional to frequency (f) usually referred to as ‘Ilfnoise’.Over the years much effort has been devoted to reducing aerosol-borne background noise in both flame and plasma systems; no one system has proved to be outstandingly superior to all other carefully designed devices and operated systems. Contamination of the aerosol by or its retention on components of the generation and transport system can occur. Care is required in the selection of the materials from which the components of the instrument that come in contact with the sample are constructed e.g.hydro- fluoric acid attack on silica and mercury retention by polypropylene.62 The matrix can have a significant effect on the intensity and nature of the background when the sample is intro- duced into the plasma.63 It may modify the aerosol generation processes and the excitation mechanisms in the plasma through interelement effects. As the effects in complex matrices are not the summed effects of the individual the background signal from a sample cannot be predicted reliably a priori but must be determined for each sample type to be analysed. Much information on the analysis of many types of samples by ICP-AES has been collected over the years. Consultation of that literature will serve to guide the analyst when faced with a new sample for analysis but the selected procedure should also be verified by the analyst personally.Gas composition its flow rate and power input are major parameters influencing the SIB and S/N of an ICP. Many workers have studied this problem but owing to the many factors that affect those parameters results and the conclu- sions drawn from them have not always been consistent. It has been reported65 that for readily excited analytical lines (CR I 425.453 nm and Ca 1442.673 nm) the SIB decreased continuously with increasing radiofrequency (r.f.) power regardless of flow rate whereas for harder to excite lines (Zn I 213.856 nm Ca I1 408.671 nm Ti I1 336.121 nm Fe I1 259.940 nm and Zn I1 202.548 nm) the maximum SIB was achieved as both r.f. power and flow rate increased.The detection limits experienced with water-cooled low-flow torches are worse than those with conventional torches owing to an increase in the background66 leading to a reduced SIB. High-power nitrogen-cooled plasmas have been studied since the mid- 1 9 7 0 ~ ~ ~ The highest S/B for ion lines was achieved with argon as the plasma and aerosol carrier gas and a nitrogen-argon mixture as the coolant gas.68 Improved precision found when using nitrogen cooling69 has been attributed to a reduced fluctuation of the background. When the operating conditions (gas flow rates power input observation height) of an ICP are being selected by for example simplex optimi~ation,~~ the aim is often to maximize the SBR; however minimization of interferences may well be more important when determina- tions are made at concentrations much above the detection limit.The background spectrum of the argon ICP in the near-infrared (NIR) region contains a large number of intense broad argon lines which limit the usefulness of this region for the determination of trace amounts of the halogens and other non-metals. Replacing argon as the plasma gas with helium with its ‘cleaner’ spectrum in the red and NIR regions reduces this problem.71 In view of the high cost of helium a low-flow torch has been designed for analytical use.72 The presence of the solvent in the ICP has a considerable effect on the properties of the discharge and hence on the signal and background emissions. Organic solvents produce a higher background than water owing to the presence of carbon-containing species.The addition of oxygen to the plasma gas when organic solvents are aspirated has been reported to reduce the background to levels comparable to those of aqueous samples.73 The effect of solvent in the plasma is eliminated when electrothermal vaporization (ETV) is used although the vaporization of the sample can modulate the background emission. Under optimum condi- tions the rate of increase of background emission with increasing r.f. power is reported to be less for the ETV technique than for aerosols.74 Attempts to reduce the effect of background emission by modulation of the analyte signal have had only limited success. Modulation of the sample flow to the plasma did not yield significant improvements in the S/B largely owing to the concomitant modulation of the background emis- ion.^^ Selective modulation of the emitted analytical resonance line has been effected by passing the radiation either alternately through and round a flame into which is aspirated a high concentration of the analyte element76 or through an electronically modulated hollow-cathode dis- charge of the analyte element.77 This line modulation approach improved SIB ratios compared with those ob- tained with the original relatively low-resolution systems.Selective line modulation has the potential for accurate correction of the mean background signal. The component of noise within the bandwidth of the electronic detection system arising from the total background radiation from the plasma flame or hollow-cathode discharge and transmitted within the wavelength bandwidth by the spectrometer will determine the SIN and not simply the background radiation within the much narrower bandwidth of the absorption line.Microwave-induced plasmas (MIPS) The development of the MIP as an emission source took place in parallel with that of the ICP78379 and has been reviewed by Zander and Hieftje.80 At present most MIPs are used for the analysis of gas mixtures particularly gas chromatographic (GC) effluents and operate at 50-100 W. When aerosols are fed into the MIP higher gas flow rates and power input to the plasma (> 100 W) are required to achieve stable operation. Desolvation or electrothermal vaporization reduces the gas and power requirements. For a toroidal MIP the centre of the discharge exhibited the lowest rotational temperature (2000-2700 K) but gave the highest S/B.For a diffuse helium plasma the highest rotational temperature was at the centre (2200-2600 K) while the S/B was constant over the whole discharge.E1 Pulse operation of the MIP has been reported to improve the S/B.82 Over the years there have been many attempts to increase the analytical utility of the MIP. However when the performances of the four different designs of torch were compared none was significantly better than the others.83 The detection limits in solutions using the MIP are generally worse than those for the ICP. In view of the limited utilization of solution MIP this review will con- sider only gaseous sample introduction systems. The most notable features of the MIP are that in its usual form it operates with a low flow rate of helium ( ~ 1 1 min-I) and thus is not expensive to run.Further the background emission by helium in the red and IR spectral range is simple and low. The plasma is very suitable for the determination of halogens ( b r ~ m i d e ~ ~ . ~ ~ and and non-metals owing to its background characteristics and524 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 the high excitation energies available. Plasma gas flow rate is the most important parameter determining the SIB. When the MIP is coupled to a GC system the analyte is temporally separated from its matrix hence background emission from that source is minimized. However contri- butions to the background emission can arise from some components of GC effluents.It was found for example that background emission from hydrogen derived from hydro- gen-containing compounds limited the detection of deuter- ium in GC effluent88 at the 656.28 nm line owing to incomplete resolution of the hydrogen and deuterium lines; the spectral continuum was negligible. The detection of oxygen using GC-MIP was found to be complicated by the presence of water vapour contamination in the plasma gas.89 On the basis of the relative S/B ratios it was recom- mended that oxygen-containing species should be moni- tored using the OH band at 308.9 nm in an argon plasma rather than the 0 I line at 777.194 nm in a helium plasma. Low-pressure discharges The analytical applications of low-pressure discharges first expored in the 1940~,~O are usually effected by means of a glow discharge (GD) or hollow cathode discharge (HCD) in which the analyte is sputtered from the electrode and excited by energetic rare gas ions.91 Their use is limited to the analysis of solid samples or gas streams (solutions can only be analysed after drying on the electrode) and by the instrumental complexities associated with vacuum oper- ation.The glow discharge has found a niche in metallurgical analysis as it is less sensitive to sample microstructure and can be used for depth profiling. As low-pressure discharges are ‘cool’ atomic lines suffer only minimal Doppler and collision broadening. The stability of the discharge leads to a good S/N whereas very low background emission produces a high S/B.The SIB is affected by changes in filler gas pressure applied voltage and operating current these parameters being interdepen- dent. For a given lamp and sample type if two parameters are specified the third is uniquely defined. The emission intensity from the lamp tends to be lower than from other electrically excited sources. Efforts to increase the output have included the application of magnetic9* and high- frequency electricg3 fields but relatively little practical improvement has been achieved by these means. Voltage modulation or pulsing of the discharge with synchronous detection has however led to increased analyte line intensity and improved discrimination against the back- ground light emitted by the plasma gas by an order of magnitude.94 Changes in the design of the sources have tended to result in improved analytical accuracy and p r e c i s i ~ n ~ ~ * ~ ~ rather than lower detection limits.Reduced-pressure discharges are efficient excitation sources but poor atomizers as their temperatures are low so more energetic systems have been studied. Atomization of the analyte from the electrode into the discharge by means other than sputtering have included ele~trothermal-~~*~~ and la~er-based~~ systems. These systems may generate a higher background but careful optical alignment and use of the far- UV region of the spectrum can minimize its effect. Further as the signals are transient some temporal discrimination against background is also possible. Instrument-generated Background in AES and its Correction The human eye was the first detector employed in spectro- chemical analysis followed by the photographic plate then photoelectric devices.This review of spectrometric systems will not consider the no longer used visual approach. However although the use of photographic recording of spectra is declining this subject wll be included as the increasing use of photoelectric array detectors of spectra is likely to encounter some of the problems found in photogra- phy. The characteristics of the instrumentation associated with photographic and photoelectric detection will be presented separately. It has to be said that the photographic plate has the unique advantage that it displays clearly and in a way that is approached only by wavelength modulation the relationship between an analytical line and underlying background radiation.Those who have worked with photo- graphic recording may well appreciate the problems of background correction more clearly than those who have used only photoelectric detection. Photographically Recorded Spectra The photographic plate is long established as a valuable means of recording spectra and for quantitative analysis. loo Its great advantage lies in its simultaneous acquisition of all spectral information within the wavelength range observed and the creation of a permanent record that can be re- examined at some future time. However it suffers from the serious disadvantage of an inability to generate instantane- ous results and of a complex and variable process for the conversion of a latent photographic image into an analyte concentration value. Nevertheless in view of the continu- ing use of photographic method~,~O~-~O~ it is appropriate to touch on some of the more important considerations affecting the generation of accurate background-corrected results; more detailed accounts of the general principles of spectrography can be found for example in refs.104 and 105. Present-day practice is based on the same principles whose application has been eased and improved by means of modern technology. In this section the contribution of the spectrograph and photographic processes to the back- ground signal will be considered. Attention will be directed to the factors affecting the measurement of the background and the incorporation of that information into the deriva- tion of the net signal generated by the analyte. As spectrography is a mature even elderly technique there have been very few papers on the subject published in the last decade or more; nevertheless many of the present generation of users of emission spectrometry have had to re-learn perhaps somewhat painfully the lessons learnt by their forebears such as the importance of dispersion and resolving power.Spectrograph The background signal has its origins in the continuum and line overlap of the source external stray light and light scattered from surfaces within the instrument. Lens and mirror errors spherical aberration chromatic aberration astigmatism camera and field curvature can also degrade the SIB. Up to the early 195Os many high-resolution spectrographs were based on concave diffraction gratings thereby eliminating the need for additional optical compo- nents although astigmatism was a limitation.However to achieve low reciprocal linear dispersion the instruments were large (2-3 m) because the groove density of the concave grating was relatively low 600 lines mm-l. Since that time plane gratings with focusing optics have domi- nated because higher groove densities (up to 3600 lines mm-I) have made more compact instruments possible. As a general rule the greater the resolving power (R) of the instrument the higher is the signal-to-background ratio that can be achieved. Resolving power is defined as R =A/M where A= wavelength of measurement and M=difference in wavelength between two spectral lines such that the diffraction maximum of one line falls on the first minimum of the other.The resolving power (R,) of aJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 525 prism is given by the expression Rp=b dp/dA where b=length of the base of the prism and dp/dA is the rate of change of refractive index with wavelength; for a grating R,=nm where n is the total number of grooves in the diffraction grating and m is the order of the spectrum. The angular dispersion ( D ) of a spectrograph (dqYdA) determines the ease with which the resolving power of the instrument can be exploited; @ is the angle of deviation of radiation of wavelength A. In the case of a prism D,=2 [sin( ;)I$ where (Y is the angle of the apex of the prism and for a grating mn D,= ~ we cos e where 8 is the angle of incidence and Wg is the width of the grating.It is worth noting that both the resolving power and dispersion of a spectrograph are determined by the proper- ties of the dispersing element alone; the associated optics play no part. Natural or fused quartz is required for UV wavelengths but its dispersion in the visible region is poor. ‘Large’ spectrographs therefore may have both a quartz and a glass optical system in order to cover the whole wavelength range satisfactorily. The collimator and camera optics determine the overall size of spectrograph and its reciprocal linear dispersion (t) L= IIFD where F is the focal length of camera lens. A theoretical optimum or critical slit-width (S,) below which the illumination drops very rapidly and resolution is improved only slightly can also be derived.The condition to be fullfilled is that the central maximum of the diffraction pattern of the entrance slit should just fill the limiting aperture of width W of the spectrograph. Then AF Sc=-=;lA W where A =aperture ratio of the instrument = F/ W. The critical slit-width can be also shown to be compar- able to the slit-width corresponding to the theoretical resolution of the instrument SR as follows For a grating normally illuminated 1 rnn Dp=- X- Re nm Wcos6 1 W N- N Hence For a prism (7) Let cr=60” and p= 1.5 then -a- DP 1 R bx0.66 When illuminated for minimum deviation W=b x 0.8. Hence =1.2 AA SRpX W x 0.83 This analysis may be extended to examine the experimen- tal spectral band width A& of spectrographs. Assuming Gaussian profiles the factors determining the bandwidth may be combined as quadratic function:’06 where A& = physical line width; AAo = theoretical resolution (aScL); A& = geometrical bandwidth =(actual slit width) x L; A& = contribution from optical aberrations.Table 1 presents theoretical estimates of resolving power reciprocal linear dispersion and optimum slit-width for several types of spectrographs. Experimental studies by Mermetlo6 and Boumans et al.12 of high-resolution instru- ments recorded values of AAo between 1 and 4 pm and for AAZ between 1 and 7 pm. These bandwidths are of the same order as physical line widths ML (1-10 pm). Instrument manufacturers do not usually publish the theoretical limits of the instrument but rather the aggregate of those para- meters as the ‘nominal’ or ‘practical’ resolution.Typical values of practical resolution lie in the range 10-30 pm. From the foregoing it can be seen that the minimum practical bandwidths of generally available instruments tend to be several times greater than physical widths of commonly used analytical spectral lines. The resolving power of the photographic emulsion ranges from 30 to 200 lines mm-’ with corresponding image line widths of 30-5 pm; usually the greater the resolving power the less sensitive is the emulsion. Hence the limits set by the photographic plate are comparable to those set by the spectrograph. For ease of operation and accuracy of measurement where background radiation is not a serious problem wider slits may be used with a consequent greater radiation throughput and shorter exposure times but with reduction in the SIB.Stray light within an instrument originates from reflec- tion of radiation out of the analytical beam by optical surfaces on to structures within the spectrograph which in turn scatter the radiation on to the photographic plate. The reflections may be at lens and prism surfaces as in Littrow systems or unwanted orders and ghosts from diffraction gratings. The reflections in a Littrow system are reduced first by use of a 30” back-silvered prism and further reductions are made either by placing a blackened bar across the centre of the lens or placing the lens at a small angle to the optical axis. Coating or ‘blooming’ of lenses also reduces reflected radiation.The energy in unwanted orders of a grating spectrum is reduced by ‘blazing’ i.e. generating a groove profile that reflects the radiation at an angle corresponding to a given wavelength and order. Scattering by dust and fluorescence from grease can also arise if the optical surfaces are contaminated. Mismatch between the acceptance angle of the spectrograph and the external light-gathering optics can result in ‘overfilling’ of the optical components and radiation falling directly on internal structure e.g. grating or prism mounts and hence being scattered on to the plate. A mismatch can also result in radiation being collected from outside the optimum (S/B) analytical zone of the emission source e.g. room526 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 Table 1 Representative values of resolving power reciprocal linear dispersion and optimum slit-width of prism and grating spectrographs Instrument values* Reciprocal Theoretical Equivalent resolving dispersion (L)l width (SJ bandwidth/ Theoretical linear optimum slit spectral power (R) nm mm-’ Pm Pm Prism- Medium quartz (F=0.7 m A = 11.7) 30 OOO? 1.5 3.5 Large Littrow (F= 1.7 m A=27) Glass dense flint 23 OOO$ 1.8 16.4 1 1 0003 3.3 16.4 Quartz 50 OOO? 0.5 8.2 Grating- 600 grooves mm-’ (F= 1.5 m A = 15) 60 000 1.2 4. s? 1200 grooves mm-’ (F= 3.4 m A = 25) 160 000 0.25 7. 5? 9.0$ 15.0% 3600 grooves mm-* (F=0.7 m A = 2 8 ) 100000 0.35 8.5t 17.0$ * F= focal length of collimator/camera lens; A =aperture ratio (f-number). ? Wavelength of radiation = 300 nm.3 Wavelength of radiation = 600 nm. 5.25 29.5 54.1 4.1 5.4 10.8 1.9 3.8 3.0 6.0 light incandescent electrodes which then adds to the spectral background. Photographic emulsion Background correction is complicated by the fact that the density of a photographic image is not directly proportional to its exposure (i.e. irradiance x time) over the whole of its density range. The Hurter and Driffield (H&D) character- istic curve of log (intensity) versus density is typically sigmoid with a ‘toe’ a more or less linear central portion and a ‘shoulder’. Thus photographic densities must in principle be converted into units of exposure before subtraction of the background exposure is effected. Implicit in this step is the assumption that the effects of the line and background exposures on the developed density of the plate are additive so that the line exposure may be calculated by subtracting the background exposure derived from density measurements made at an adjacent location on the plate.The sensitivity of photographic emulsions (i.e. the slope y of the linear portion of the H&D curve referred to above) is wavelength dependent hence measurements of background should be made close to the analytical line. If however the background is stable and well characterized over the spectral range used a single measurement modified by correction factors to give values at other wavelengths may be adequate. The response of the emulsion may be affected by reciprocity and intermittancy failure i.e. differing developed densities produced by the same exposure given at differing light intensities.The response of a photographic emulsion to a sharply focused spectral line and to the uniform illumination from a continuum spectrum or stray light may differ through scattering of light in the emulsion (halation) and local effects in the plate-developing process such as developer penetration and exhaustion in the neighbourhood of dense lines (Eberhard effect) and more complete development of weak lines than strong (intensity retardation effect). Thus the slope of the blackening curve may be greater for a narrow line image or the edge of a wide image than it is for the central portion of a blackened area or for background blackening an effect re-examined by Plsko in 1988.Io7 A very low level of background exposure may not appear on the developed plate but it can still contribute significantly to the density of a weak spectral line.A deliberately induced low background density (up to 0.1) can therefore be beneficial in work demanding the highest sensitivity as it overcomes the ‘inertia’ or initial low-sensitivity part of the H&D curve. As a consequence of the variability of the photographic process calibration graphs are usually based on the ratio of the intensity of the analyte line to that of an internal standard element. Calibration of the density response of an emulsion to light of known relative intensities may be achieved by interposing a stepped neutral filter or rotating sector close to the entrance slit. Uniformity of illumination over the full length of the slit is essential.Alternatively plate calibration can be made with a series of pre-calibrated lines in an iron spectrum. This avoids slit-length and reciprocity failure errors. Microdensitometry The first requirement of the microdensitometer is that it should not contribute errors to the measurement of the photographic plate. There should also be a linear relation- ship between the light transmitted by the spectral line image and the response of the photometer. One of the most important conditions is the exclusion of stray light. The presence of stray light will lead to an underestimate of the spectral line densities. Torok and Hafenscher1OB examined the effect of stray light on calibration graphs. To minimize stray light the width of the entrance slit of the densitorneter should be less than that of the projected image of the recorded line.Stray light generated within the emulsion by scattering of light by the developed silver grains (Schwarzchild-Villiger effect lo9) can lead to errors when spectral lines with densities greater than 1.5 are being measured. The microdensitometer requires stable high- quality optical mechanical and electronic systems to ensure that accurate information is recovered from the photographic plate. The characteristic curve of an emulsion as derived from microdensitometry of a spectral line may not match that shown by the manufacturer of the plate even if identical emulsion development conditions apper- tain because the latter are obtained by measurement of large areas of emulsion. The ‘zero’ of the instrument must be set on a ‘clear’ unexposed part of the photographic plate free from fog or background to avoid errors in background correct ion.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY.JUNE 1993 VOL. 8 527 Computation Torok and Zimmer (1972)114 and Heltai and Zimmer ( 1 980).115 Present-day developments are confined to redis- covering established procedures or applying computers to meter.ll6 Photoelectric Detection of Spectra The purpose of computation is to translate the densities of phic plate into measures of the relative exposures by the radiation producing the plate blackening. Computer control of the microdensitometer and data processing developed during the late 1970s greatly facilitated spectrography and ’Pectral lines and background measured On the Photogra- the control and data processing of the microdensito- background correction.l1° The non-linearity of the emulsion response and artefacts generated in the instrumeritation necessitate the use of calibration graphs for both emulsion and analyte responses.Few if any emission sources emulsions and development procedures used in conjunc- tion with spectrography are sufficiently stable and reprodu- cible to permit analyte calibration curves based on absolute measurements alone. For quantitative analysis the use of an exposure ratio with an internal standard element the ‘line intensity ratio’ is the norm hence background is present in at least four measurements i.e. two lines plus background and two backgrounds adjacent to the lines.If the spectral line is in a region of highly variable back- ground the background may be measured on either side of each line to give a total of six background measurements for each analytical line pair (although if the background pattern is the same from spectrum to spectrum in routine analysis it is not normally necessary to correct the internal standard line). However when a background is present it is important that it is accurately measured particularly when the analyte signal is weak. Improperly corrected back- ground emission leads to non-linear calibration graphs near the detection limit. In trace analyses where suitable internal standard lines may not be available at the appropriate wavelength (e.g. some non-ferrous alloys) the background intensity underlying the line can often be used as the internal standard.ZB-lll This is taken as the mean of two readings at equal distances on either side of the analytical line.As there is a logarithmic relationship between plate density and exposure background correction cannot be effected by simply subtracting the background density from that of the line plus background nor should the microtlensi- tometer be ‘zeroed’ on the background. Both of these procedures generate a parameter that is derived from the ratio of the exposure of the line plus background to that o f the background not the net analyte or standard signal that is required. The calculation of the correct line ratios can be a time-consuming operation therefore methods have been devised employing calculating boards and ‘subtraction’ logarithmic tables to assist in the process.*1z I n caSes of extreme difficulty fi.g. a faint analytical line on a steeply sloping and irregular background a plate-scanning tech- nique a ‘comparator display photometer,’ has been uscd.’ l 3 I n this system the output signal of a reciprocating scanning microphotometer was displayed on a long afterglow cath- ode-ray tube (CRT).The spectral region of interest was scanned at 12 cycles min-l and data for ‘line intensity ratio’ calculations were obtained by measurement of the CRT trace. Modern direct-reading scanning spectrometers with background correction are based on equivalent principles. The reagents used to prepare the calibration graph may contain a residual undetermined amount of the analyte element a ‘background’ or ‘blank’. If the calibration graph of exposure ratio i’ersz4.s concentration plotted on a linear scale does not pass through the origin extrapolation of the curve until it intercepts the negative abscissa provides an estimate of the residuum. This estimate should then be added to the known amount of analyte in the standard to give its true content.Mathematical procedures for background correctior and the preparation of calibration graphs have been developed and reported for many years e.g. Lundegardh ( 1 933),lo0 By eliminating the photographic plate with its complex response to radiation intermediate stages of chemical development and microdensitometric measurement direct- reading spectrometers developed in the 1 have the potential for faster and more accurate analysis.Back- ground always exists in one form or another but with a direct reader may well not be obvious to the operator. The need for and problems of background correction thus remain although they take a different form. In routine analysis if the background (although different under analyte and internal standard lines) is constant from spectrum to spectrum correction may not be essential as intensity ratio calibration graphs constructed from non- corrected line intensities will simply be offset and curved rather than linear. When a calibration procedure involving standardized materials of the same type as the samples is used (e.g. in routine metal analysis) the zero offsets necessary to make the calibration graph pass through the origin are themselves effectively the background correction.In addition to a background signal generated by radiation falling on the detector there is the dark current generated in the detector by the thermal release of electrons from the photocathode and drift and ‘noise’ in the electronic system (see also under Background fluctuations noise in AES). The response of most detectors is linear from dark current levels up to high intensities when saturation may occur. provided that the appropriate electronic systems and operating conditions are used; large and rapid changes in light intensity can however lead to a non-linear response. Generally background correction can be effected by a simple subtraction procedure usually in the form of ‘zero offset correction’ without the use of a detector response transform.The principal difficulty is that of identifying and accurately measuring the background to be subtracted. When measurements are required at analyte concentrations close to the detection limit inaccurate background correc- tion can lead to non-linearities in the response curve of direct-reading systems just as in photographic Although photoelectric systems are in principal capable of direct measurement of radiation intensities in practice time constants in the measuring circuit result in the dimensions of the data produced being those of irradiance i.4.. intensity x time. The time duration of the measurement is crucial in that to achieve comparable reproducibility unstable or weakly emitting sources require much longer measurement times than stable or intense sources.Arcs sparks and transient sampling systems are generally less stable sources than flames plasmas low-pressure discharges and systems in which the sample may be continuously introduced. The time constant is particularly important when attempting background correction in time-shared or transient response systems. Measurements of background in the neighbourhood of analytical lines may be made by either a static or dynamic system. The former is analogous to the approach used with the photographic plate where information is collected for an extended period and modest changes in source intensity are unimportant. In the dy- namic system the effectiveness of background correction is a function of the modulation frequency and the rate of change of the source intensity.A more detailed discussion of background correction in direct-reading systems will now be presented in three parts spectrometers detectors and data handling. As the use of528 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 the inductively coupled plasma is the most actively devel- oping aspect of AES at the present time the review will reflect this bias. Spectrometer The earliest (1 935) form of single-element direct-reading spectrometerI2O employed a flame monochromator and photocell for the determination of alkali and alkaline earth metals. Background correction was by means of subtraction of the signal generated by a ‘blank’ solution. Successive generations of flame photometers using either dispersive or non-dispersive wavelength selection employed the same method of background correction sometimes aided by means of an electrical or mechanical zero offset of the measurement system.The generally low wavelength resolu- tion of such instruments led to background correction problems when complex matrices were analysed. In 1965 Warrenl2I described a high-resolution flame photometer for the determination of copper and magnesium in blood serum designed to overcome such problems by directly measuring the background under the analytical line gener- ated by the sample. The instrument had a resolution of 10000 a bandpass of -0.01 nm and a scan rate of -0.005 nm s-l. The intensities of the line and of the background on either side of it were recorded on a chart recorder and thus the net line intensity could be derived.In spite of the system’s good resolution the detection limit for copper at the 327.396 nm line was degraded by background generated by the 327.421 nm OH band from 36 to 200 fig I-’. The operation of this instrument incorporated many of the features found in current background correcting systems which inevitably also suffer from equivalent limitations. The first reports of simultaneous multi-element direct- reading spectrometers appeared in the early 1940s. Boett- ner and BrewingtonlzZ described a two-channel system one channel being used for an internal standard element based on a grating spectrograph wherein the outputs of the photomultipliers were ratioed but no explicit background correction was attempted.The first direct-reading instru- ment incorporating background correction is attributable to Saunderson et al.123 The instrument was a grating spectro- graph with eleven photomultipliers mounted in place of the camera. By means of a multiple shutter each photomulti- plier could be exposed to one or two lines of the element or an adjacent portion of the background spectrum. Signals were integrated for 20 s and background correction (sub- traction) was carried out simultaneously. The same prin- ciples are incorporated into the polychromators of today the most significant developments being in the realms of detection electronics and computing. A major influence on the design of spectrometers has been the greatly increased use of atomic spectrometry by the analyst as distinct from the spectroscopist.This change created a demand for the compact ‘user-friendly’ ‘press- button’ instruments which characterize current develop- ments. We shall now present outlines of the techniques for background corrections that are incorporated into present day instruments in order that the reader may be made aware of the strengths and weaknesses of the equipment he or she is using. Wavelength selection. The purpose of the wavelength- selection device is to collect as much light as possible from the emitting source and to separate the analytical line as completely as possible from other lines continuum back- ground and stray light. The devices fall into two broad categories non-dispersive instruments which are generally simple systems of large aperture and low resolving power and dispersive instruments which are more complex and flexible in use and with much higher resolving powers but lower light-gathering efficiency. Both systems may be used in single or multi-element modes; in the latter instance the measurements of the analyte signals are made either simultaneously or sequentially.The background correction procedures differ considerably between the several categ- ories of instrument. Non-dispersive syst ems-jilt er instruments. The optical filters used in these systems may be either broad-band (-40 nm) optical absorbing material (coloured glass quartz or solutions) or narrow-band (- 10 nm) interference filters. The use of filters is usually restricted to the visible region of the spectrum where their resolving power is of the order of 40 although interference filters are available down to 200 nm or so.Simple flame-based instruments using filters or photode- tectors with a limited spectral response for the selection of the analytical line are adequate when the signal from the analyte is greatly in excess of the background and any residual background signal is well characterized so that correction can be readily effected. An example of such an analysis is the long-established determination of sodium and potassium in serum where correction is made by blank subtraction or by use of a calibration graph derived from synthetic solutions similar in composition to the sample. Filter systems are also adequate when the analyte element or a compound of it can be separated from its matrix by for example chromatography or the generation of a volatile compound.The excitation in such a system may be by flame or electrical discharge. Selective excitation of the analyte element to generate a fluorescence signal may also be used in a filter instrument to improve discrimina- tion against the background emission. Such systems and the associated background correction procedures will be considered in the review of techniques in atomic fluores- cence to be presented later. Non-dispersive systems-Michelson interferometers. Mi- chelson interferometers have been employed for Fourier transform spectrometry (FTS) in the UV and visible region of the ~ p e c t r u m . ~ ~ ~ - ~ ~ ~ The interferogram is produced by moving the mirror in one arm of the interferometer.The resultant fluctuations in the output of the photodetector are recorded and the spectrum recovered from it by applying a fast Fourier transform to a series of sampling points (up to 2 x lo6) on the interferogram. The advantages of using this type of instrument over conventional grating spectrometers are that all the spectrum is observed for all the time with extremely high resolution. For example typical resolving powers are of the order of lo6 compared with 104-105 for a good grating spectrometer. This resolving power is suffici- ently high to allow line width studies to be undertaken in most atomic line sources such as hollow cathode discharges and plasmas and to minimize line overlap and other spectral background problems.The use of a Fourier transform spectrometer also offers the advantage of higher light throughput because a circular aperture is used rather than a narrow slit as in a grating instrument. The high light throughput of the interferometer cannot however be translated into improvements in signal-to- noise ratios and hence in detection limits. This is because of the multiplex disadvantage which results from instabili- ties of the source the recording of all of the spectral lines with their associated noise in one scan of the interferometer and ‘jitter’ of the sampling interval.lZ8 Subsequent transfor- mation of the interferogram results in the noise from every part of the observed spectrum being distributed throughout the background spectrum. In effect the S/N depends on the length of the transformation i.e.the number of points in the transform and the complexity of the observed spectrum. Voigtman and Winefo~dner~~~ considered theJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE multiplex advantage and the effect of low-frequency noise in multiplex spectrometry and showed that shot noise becomes distributed uniformly across the spectrum whereas periodic fluctuations in the signal are localized around spectral lines. Experimental observation of these predic- tions was reported by Marra and H o r l i ~ k ' ~ ~ and later by Hobbs et and also Kolar and Williams.128 These workers showed that the presence of a few strong lines in the spectrum can degrade the S/N for weak lines in the spectrum thus degrading the detection limits in compari- son with those obtained with grating spectrometers.They also showed that the nature of the limiting noise changes with the concentration of the analyte and matrix. At low concentrations (Le. high dilutions) the noise is independent of the signal level yet at high signal levels the noise is proportional to the signal as was considered under Back- ground fluctuations noise in AES. Hobbs et ./.I2' investi- gated these dependences using an ICP interfaced with a high-resolution FTS. The model system chosen was the emission of Fe lines from a sample containing 10 ppm Fe in the spectral region 240-250 nm in the presence and absence of 1000 ppm of U. In this spectral bandpass there are 35 Fe I lines and 85 U I1 lines. The predicted degradation ofthe SNR when uranium was present was a factor of 3 whereas the experimentally determined degradation factor was 2.7.thus illustrating the reality of the multiplex disadvantage. Hobbs et .[.I2' also showed experimentally that detection limits are dependent on the concentration of other elements in solution. Thus as the solution is further diluted the apparent detection limit is improved. However the matrix is also diluted so the real detection limit is not improved. It seems therefore that the multiplex disadvantage presents a serious limitation to the use of high-resolution UV-FTS for analytical determinations unless methods to overcome source noise such as proportional fluctuation noise arising from the sample introduction and desolva- tion-atomization processes are successful. Even the source photon noise presents a practical limitation to the tech- nique. Other procedures to minimize the effect of the multiplex disadvantage include co-adding spectra and self- adaptive filtering131 for stepped-scan interferometers. Both of these techniques however involve a time penalty which detracts from the original advantage offered by FTS over for instance a slew-scanning spectrometer.Some reduction in the effect of redistributive noise has been achieved with a multi-mode spectrometer.132 The instrument consisted of a scanning Michelson interferometer a flat field grating and a linear photodiode-array detector. The spectrometer could be operated in the multi-channel (dispersive) multiplex (interferometric) or combined multi-channel-multiplex modes.Dispersive systems. General reviews of spectrometers for emission analysis have been presented by O l e ~ i k ' ~ ~ and Hill et al.,134 a set of criteria for the evaluation of spectrometers have been published by the Analytical Methods Committee of the Royal Society of C h e m i ~ t r y l ~ ~ ~ ~ ~ and a procedure for the optimization of ICP correction parameters was de- scribed by Ediger and F e r n a n d e ~ . ~ ~ In relation to hack- ground correction the spectrometer fulfils two functions first to generate a spectrum with the minimum of stray light and second to provide a means for measuring the unavoida- ble background. It may be either a polychromator system in which all analytical lines are measured simultaneously or a scanning monochromator in which the lines are measured sequentially.Generally reduction of the optical bandpass of the system until it is equivalent to the minimum width of the spectral line image of the analyte will increase the SBK by reducing the transmission of continuum background and of adjacent or overlapping spectral lines. In practice the slit- 529 1993 VOL. 8 widths employed tend to be determined by considerations of line intensity and practical instrumental resolution (see also under Spectrograph and Table I). The exit slit is usually set slightly wider than the entrance slit to reduce the effect of instrument instability. If however the slit widths employed are such that the corresponding optical bandpass is significantly greater than that of the minimum width of a spectral line image and if the only background is a continuum from the source then the dependence of the detection limit on slit-width can be derived as follows.If at the spectrometer entrance slit Pis the photon flux of the analytical line per unit of analyte and B is the photon flux of the background per unit spectral bandwidth then Np= PSHtE and NB= BSHSLtE= BSZHLtE (9) where Np N and ND (eqn. 10) are the number of photoelectron events generated in the detector in a time interval t by the analytical line background radiation and dark current respectively S and H are the width and height of the spectrometer slits L is the reciprocal linear disper- sion of the spectrometer and E is the efficiency of conversion of incident photons into photoelectron events. Eqn. (9) demonstrates that in photoelectric spectrometers a spectral line signal is proportional to slit-width while the background signal from a continuum is proportional to the square of the slit-width. Now the detection limit is q,= G' where K - probability constant gM =standard deviation of blank measure and G=sensitivity of the method (i.e.response per unit of analyte). Hence if E is a conversion factor from photoelectron events to photomultiplier current and k is the RSD of the proportional noise on the background for a unit time data collection period then G=&Np - -y[ BL( I+-) +m] N D ' (10) HtE t The effect of slit-width on detection limit as indicated by eqn. (10) can now be deduced. Of the three terms within the square root bracket the first is independent of slit-width the second is proportional to its square and the third is inversely proportional to its square.It follows that when the first term i.e. photon noise on the background signal is dominant changes in slit-width will not affect the detection limit. If however the second term i.e. proportional noise on the background is dominant increasing the slit-width will worsen the detection limit. Finally in the unlikely situation that the noise on the dark current is greater than that on the background i.e. the third term dominates then increasing the slit-width will improve the detection limit. When the lowest possible detection limit is required it is helpful to determine the relative values of background5 30 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 photon and proportional noise and dark current noise so that the optimum slit-width can be selected. The interactions between spectral bandwidth linewidth and overlapping lines with detection limits have been studied by Boumans and Vrakkix~g.~~ Where there is line coincidence or overlap with a broadened line between the analyte and matrix element within the chosen slit-width there is little that can be done instrumentally except either to choose an alternative analytical line or to measure the interfering element at another line and apply a proportional correction to the analyte measurement. The selection of the most suitable analytical line is a crucial step in minimizing the demands to be placed on the background correction procedure. To facilitate this process tables of analytical lines used in ICP-AES with possible interferents derived from computer simulation have been ~ o m p i l e d .l ~ ~ - ~ ~ ~ The dispersive element of a spectrometer may be a refractor (glass or quartz prism) or a diffractor (grating or echelle). Resolving powers of 500000 or higher have been used in studies of atomic spectra and line profiles whereas 30000 is the minimum practical resolution required for general analytical use. This value is a compromise between instrument compactness light throughput and minimiza- tion of line overlap and background continuum. A proce- dure for measuring the practical resolving power of mono- chromators in ICP-AES has been described by Mermet.Io6 Stray light and hence background correction problems may arise in dispersive systems form several sources the spectrometer may accept light form outside the emission source dirty optical components may scatter light optical surfaces may reflect light out of the optical path to be scattered by structures within the instrument or unwanted orders of diffracted spectra may be inadequately trapped.A comprehensive discussion of the origins and elimination of stray light in optical systems has been presented by G00dman.I~~ Even following the best possible practice there will inevitably be a residuum of stray light hence it is necessary to devise means instrumental or procedural whereby this source of background may be detected and measured. Early in the development of ICP-AES a number of possible solutions to these problems were p r o p o ~ e d .I ~ ~ J ~ ~ The properties of the dispersive device in relation to its effect on background will be briefly reviewed before considering in some detail techniques for background correction used in conjunction with such devices. The diffraction grating is preferred to the prism as the dispersing element in modern spectrometers except as an order sorter in echelle instruments. The non-linear disper- sion of the prism its cost of manufacture and difficulties in meeting the requirements of modern direct-reading spectro- metry contributed to the demise of prism-based instru- ments. The diffraction grating has the additional advantage of being usable in several orders thereby achieving in- creased dispersion and resolving power. The efficiency of a grating is wavelength and polarization dependent.The ruled diffraction grating is now tending to be superseded by holographically produced gratings. This development over- comes the problems of ‘Rowland’s ghosts’ spurious spectral lines produced by periodic defects in the ruling of the grating and also improves the stray light rejection by an order of magnitude.143 Overall the S/N of a holographic grating is better than that of a ruled grating although its efficiency and ease of blazing may be only comparable to or even worse than those of the latter. As an alternative to a single holographic grating the discrimination against stray light in a ruled grating system can be greatly increased (approximately 1 000-fold) by means of a double monochro- mator system i.e. two monochromators in series.The resolving power of the combined system is equal to that of the final monochromator. High resolution in a compact instrument can be achieved by use of an echelle This device is intermediate between a reflection echelon and a diffraction grating. It has a low groove frequency (50- 100 grooves mm-I) operates in high orders (20- 1 50) and has corresponding theoretical resolving powers ranging from 30000 to 450000. Overlapping spectral orders are separated by means of a prism that disperses the radiation in a direction orthogonal to that produced by the Cchelle. Line overlap and continuum background effects can be reduced by use of high resolving powers but the incorpora- tion of the order-sorting prism introduces additional optical surfaces from which stray light may be scattered; instru- ment instability can also be a problem.Background correction devices. Even with the highest quality dispersing optics some background radiation will remain. Therefore although it is good practice to minimize the background in reality it may be more economical to incorporate some form of background correction device. The nature of the background should be determined wherever possible from either a photographic or photoelec- tric survey of the spectrum. The measurement of back- ground is usually carried out at a wavelength adjacent to the analytical line. When the background is highly structured the precise location of that measurement should be selected with great care. If the background is a simple continuum and constant in the neighbourhood of the line a single measurement is adequate; when it is steadily changing with wavelength measurements on both sides of the line may be required.The background may be measured and subse- quently removed in the data processing system or it can be simultaneously subtracted electronically. Whichever proce- dure is used the fluctuations in the background may still pass through the system and reduce the precision of measurement. Notwithstanding the variety of dispersive devices used in analytical atomic spectrometry the optical procedures used for background correction fall into two broad categories those associated with static wavelength selection and others with scanning systems. Several options are available for background correction in static polychromator-like systems.The effect of stray light may be reduced by means of rejection including atomic v a p o ~ r s l ~ ~ or by using photodetectors that are insensitive to the wavelengths of the stray radiation (e.g. the ‘solar blind’ photomultiplier which ignores radiation above 400 nm). In order that measurements of background may be made as close as possible to the analytical line in the dispersed spectrum mirror tunnels146 and fibre have been used to collect radiation adjacent to the line and redirect it on to the photodetectors thereby overcoming the problems created by the relatively large dimensions of the latter. When the characteristics of the background are well established three fixed slits may be employed one for the spectral line and the others measur- ing background on either side of it.If the slits are in the exit plane then three may be required for each line although occasionally if the relative fluctuations of the background are constant over large regions of the spectrum fewer measurements of it may suffice.148 The number of exit slits and the associated problems of their alignment may also be reduced by using an array of three entrance slits each of which is illuminated in turn via a chopper.149 By this means the photodetectors receive sequentially line and back- ground radiation. The signals produced may be either synchronously decoded to give separate measures of line and background intensities or by a circuit tuned to the chopping frequency to give an output proportional to the total intensity less background at the line wavelength.The exact positioning of extra entrance slits however may be appropriate to only one analytical line and the use of a chopper reduces the available energy and consequently the overall SIN.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993. VOL. 8 5 3 1 Short-range wavelength scanning may be used to measure the background in the neighbourhood of analytical lines. This can be achieved by movement either of the entrance slit directlylsO or of its image by means of a rotatable quartz refractor plate mounted behind the entrance slit.' 5 1 The refractor plate may oscillate about a vertical axis152 to give a sinusoidally modulated signal of the net line intensity or be a segmented quartz chopper disc consisting of two pairs of plates of differing thickness rotating about a horizontal axis and set at 45" to the optical axis.153 The latter construction gives a step displacement of the line and background which improves the data collection efficiency.The same effect can be achieved by driving an oscillating refractor plate with a staircase waveform under computer c o n t r 0 1 . ~ ~ ~ ~ ~ ~ ~ If single element analysis only is being carried out using a mono- chromator the quartz refractor may be placed in front of the exit ~ 1 i t . I ~ ~ When quartz refractor plates are employed for wavelength scanning the dispersive properties of the material should be borne in mind; a given angular rotation of the plate will produce different wavelength scans at different central wavelengths.Compared with static direct reading of line and back- ground intensities wavelength- or time-modulated systems have several weaknesses. First to achieve accurate back- ground correction the emission source must be temporally and spatially stable over the period of the modulation. Second the time to accumulate the same amount of spectral data is increased considerably (2-3-fold at least). Third the photodetector and its associated electronic circuitry may be required to accommodate rapid and large changes in radiation intensities. Finally imperfections in the modula- tor will add to the 'noise' in the system. In multi-element analysis based on wavelength scanning using a monochromator with motorized drive of the dispersing element background measurements can be made on either side of and close to the analytical line during the scanning process.157 The scan itself may be either continu- ous in which case all parts of the spectrum are observed for equal time or slewed in which case irrelevant portions of the spectrum are traversed quickly with slow scan or wavelength modulation over the region of measurement. Additional rapid short-wavelength modulation may be superimposed on the spectrum scan by devices mounted at either the entrance or exit slits of the r n o n o ~ h r o r n a t o r .~ ~ ~ These devices are similar to those used in static dispersion instruments and serve to reduce the time interval between measurements of line and background intensities. Accurate analysis using scanning systems is only possible with continuously atomizing stable sources such as flames glow discharge lamps or plasmas.In practical terms satisfactory background correction is possible in both simultaneous and sequential systems particularly when linked to computer control of the instru- ment and data processing. The principal differences be- tween the two systems is the greatly increased time required for multi-element analysis by a sequential instrument and its unsuitability for use with unstable or intermittent sources such as arcs and sparks. Detectors The first practical photoelectric devices were constructed in the early 1890s and almost immediately were applied to the measurement of UV radiation from the sun. Since that time astronomers have always been in the forefront in exploiting photoelectric devices.Unfortunately the average spectrochemical laboratory has not been funded to pioneer such developments and consequently spectrochemists have had to wait until such devices became generally commercially available. The use of photoelectric cells in the photometry of the photographic plate began in the late 1920s followed by the direct reading of spectral line intensities in the mid- 1930s. This pattern of analytical spectrometry following commercial availability and relia- bility of photodetectors is still found today. The current commercial developments of sensitive photoelectric arrays and imaging devices with their potential for combining the spectral data collection capacity of the photographic plate with the immediacy of measurement of the photomultiplier provide a basis for some of the most promising research topics in analytical atomic spectrometry.In relation to background detection the photoelectric detector should have two desirable qualities. First it should not contribute to the background signal (a dark current) and second its response should be linear from low light levels (- 1 O3 photons cm-2 s-l) up to four or five orders of greater intensity. Additional qualities include an efficient. reprodu- cible rapid and sensitive response to light. no memory effects a wide spectral range physical compactness and robustness. Photoelectric devices fall into two categories photoemissive wherein radiation liberates electrons from the surface of the detector (photocells photomultipliers) and solid-state devices photoconducting (photodiodes) photovoltaic (barrier layer cells) and photocharge storage (charge-coupled devices).Photomultipliers are currently the established 'work horses' in spectrometry and the developments in charge-coupled devices hold out the promise of exciting opportunities for direct-reading multi- element analysis with accurate background correction. Photoinultiplier tubes. The characteristics of the photo- multiplier are well documented in manufacturer's literature and the reader is referred to such sources for fuller details.I6 Only aspects affecting background correction will be pre- sented here. The dark current of the photomultiplier is usually much smaller than the signal generated by the background radiation falling on it.Dark current originates in events within the photomultiplier and in electrical leakage in associated electrical circuitry; it is typically of the order of a few nanoamps in the anode circuit. At room temperature the two sources of dark current are comparable but the stochastic nature of the photomultiplier events produce fluctuations in that current up to a 1000 times greater than those in the leakage current. Cooling the photocathode to 0 "C can reduce the photomultiplier dark current by an order of magnitude. The leakage dark current is stable and maybe 'backed off electrically or discriminated against by taking advantage of the discrete nature of the primary interaction the release of a photoelectron whereby photon counting can be used as a means of processing the photomultiplier output.In some spectrometric applications e.g. when the photomultiplier is close to an exit slit only a limited central portion of the photomultiplier photocathode is illuminated. In this case a defocusing magnet may be placed against the cathode to direct photoelectrons from other regions of the cathode away from the first dynode thereby reducing the dark current.159 Following exposure to high-intensity radia- tion (l0000xdark current) the dark current may be doubled for a period of up to 100 ,us or more. If measurement of very low light levels is required the use of photon-counting techniques will minimize the effects of photomultiplier background. In most analytical atomic spectrometric applications however background signals generated by the emitting source and within the spectrometer are much greater than those generated within the photomultiplier; therefore the latter usually does not warrant separate treatment. If such treatment is required however modulation of the radiation falling on the photocathode with synchronous demodula- tion may be used to discriminate against dark current.The output signal of the photomultiplier requires elec-532 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 tronic processing before being usable for analytical interpre- tation. Such processing may include gating amplification integration and analogue and digital computation. Any of these stages may contribute to a baseline signal. Electronic zero suppression may be used to offset these effects but its effectiveness is dependent on the stability of the total instrumental system.Consequently a more reliable ap- proach can be that of recording the total background from all sources and applying this information as a correction to the analytical signal. Solid-state detectors. Solid-state devices may take the form of area detectors or one- or two-dimensional arrays. Area detectors are used in simple non-dispersive instru- ments where background correction is effected by simple subtraction of the background generated by a blank sample. The background generated by the detector its dark current is frequently very temperature dependent. Array detectors where a number of individual photosensitive surfaces effectively form a series of optical exit slits when incorpor- ated into dispersive instruments can measure line and adjacent background simultaneously.The feasibility of using one- or two-dimensional array detectors in analytical atomic spectrometry has been inves- tigated since 1967.160-163 The attraction of these systems is instantaneous simultaneous multi-wavelength measure- ment readily interfaced with computer data processing. This can combine some of the best features of photographic and photoelectric detection and provides data for accurate background correction. Photoelectric imaging devices are intrinsic integrators but they do not have the in-built high amplification capability of the photomultiplier. Photodi- odes require additional charge storage provision. Lack of sensitivity restricted spectral range inadequate resolution and high cost have in the past limited the exploitation of one- and two-dimensional detectors.When the ultimate in detection limit is not required the multi-element capability of array detectors can be exploited. The dimensions of the individual photosensitive surfaces and the dead zone between them limit the resolution of the system. This problem is reduced by using spectrometers of high linear dispersion. Over the past few years considerable technical progress has been made in the development of solid-state photoelec- tric devices and their associated electronic circuitry. This has reduced dark current read-out noise and cross-talk and improved quantum efficiency and wavelength range. Charge-coupled devices (CCDs) are capable of measuring light fluxes over a wide dynamic range with high quantum efficiency (up to 70Y0l~~) and appear to be the most promising of the solid-state detectors at the present time.165-168 Linear arrays with a large number of detector elements are available at minimal cost.These can be interfaced directly to signal acquisition boards and micro- computer data acquisition systems to provide full spectrum coverage in a given spectral range. Typical CCD lengths consist of 1024 elements in a standard 25.6 mm package and used with a spectrometer of modest reciprocal linear dispersion e.g. 1.6 nm mm-l a reasonable free spectral range can be covered (40 nm); 2048 element arrays are also available. Photodiode arrays (PDAs) and CCDs however can have significant read noise and dark currents at room tempera- ture which mitigate against their use in low-light situations i.e.they generate a relatively large background signal. Read noise is the noise introduced when a single charge packet is read out. This can range from 1200 electrons for a PDA to less than five electrons for a CCD array.169 The dark current of silicon array detectors originates in the thermally generated charge defects in the bulk silicon and in the surface silicon-silicon oxide interface with the conducting gates of the device. To overcome this problem CCDs are normally operated at low temperatures (down to 125 K) to reduce the dark current to approximately 0.01 e s-l ( 1.6 x A) which is insignificant for most spectros- copic applications. It is also worth noting that this is small when compared with the photocathode dark current count measured with a cooled bialkali photomultiplier tube (PMT).The lower threshold value for such a photocathode approximately -25 “C is about 1-2 counts cm-2 s - I (1.6 x 10-19-3.2 x 1 O-I9 A). This threshold is determined by pulses which originate from cosmic radiation and window radioactivity processes which are independent of the tube temperature. In practice therefore there is little merit in cooling a photomultiplier below about - 25 “C the temper- ature at which the current arising from these sources becomes significantly larger than the current arising from thermionic processes. Although the dark current of a CCD is much lower than that of a PMT this observation can be misleading as the dominant source of noise from a CCD is the read noise.Read noise can be reduced by using ‘binning’ techniques whereby the charges from a group of detector elements on the array are collected together and read as the sum of the charges. Thus only the noise from one read is observed. The penalty for this however may be loss of spectral resolution of both the spectral lines observed and background features. Also the response of these devices in the UV region of the spectrum is poor. Attempts to improve the sensitivity in the UV region have been based on coating the device’s photoactive area with a fluorescent ‘down converter,’ which absorbs UV photons but fluoresces proportionately in the visible region where the array is more sensitive. Improvements in UV quantum efficiency (up to 95%) and in SNR generally can also be obtained using CCD arrays which rely on virtual phase technology167 as described by Sweedler et al.170 In the manufacture of virtual phase CCDs ion diffusion regions are employed rather than gate elec- trodes with the result that fewer overlying gates are required.This maximizes the response of the device at all wavelengths which is especially important in the UV region where quantum efficiencies are poor. Exceptionally low read-out noise (< 1 electron rms) has been achieved with ‘skipper’ CCDs.I7l These devices operate on a non-destruc- tive read-out scheme hence read-out noise can be reduced by repeated scans of the pixels before resetting. Problems can also occur at high light levels e.g. when observing the continuum background of an ICP.The main problem then is ‘blooming’ of the array which occurs when individual detectors are saturated i. e. their charge capacity is exceeded. When this occurs charge spills over into adjacent detector elements and can interfere with the spectral information dispersed at the detector position in the spectrometer focal plane. Recently manufacturers have addressed this problem by providing extra gate structures as charge drains. However this incurs a trade-off in quantum efficiency owing to the presence of the extra gate structure. Small diode arrays with ten diodes have been combined with a single wide exit slit to measure line and adjacent background intensities which can then be subtracted e l e c t r ~ n i c a l l y . ~ ~ ~ J ~ ~ In this configuration the diode array served effectively as a series of slits whose widths are those of the photodiode.This arrangement may be employed in both polychromator and scanning monochromator instru- ments and may well be a more effective and economical use of array detectors than attempting to observe a complete spectrum simultaneously. A high-dispersion spectrometer (0.08 nm mm-I) was combined with a one-dimensional diode array for the determination of molybdenum by ICP- AES. 174 Continuous background correction reduced errors due to drift and low-frequency noise; a detection limit of 6 ng ml-1 was established. A multivariate technique was usedJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 533 to calibrate the system.175 Several groups of workers have combined the high two-dimensional dispersion of an echelle spectrometer with a two-dimensional charge-injec- tion device array ~ e n s o r .I ~ ~ J ~ ~ One of these instruments has been applied to the determination of calcium chromium copper iron nickel and lead and in the analysis of National Institute of Standards and Technology (formerly National Bureau of Standards) SRM 1643a by d.c. plasma excita- t i ~ n . ” ~ Although the ability of array devices to generate valuable data for use in background correction and for multi- element analysis renders them attractive detectors of radiation it is likely to be some time before their use in analytical atomic spectrometry is widespread as they will need to combine physical dimensions similar to those of a spectrometer slit with the sensitivity and ease of operation of a photomultiplier.The CCD arrays appear to have greater potential for applications in atomic spectrometry than photodiode arrays owing to their superior S/N. Signal and data processing Photon counting is the most effective technique for measur- ing very low light levels and producing digital data. However it is rarely used in optical analytical atomic spectrometry because radiation background signals are usually well above the few counts per second of which the system is capable of measuring and the cost of high speed (10 MHz) counting circuits is much higher than that of their analogue equivalents. The capabilities of current analogue electronic signal processing systems are such that if correctly chosen and operated they are not a limiting factor in background correction.In some instances however cost may limit their use; for example digital integration is more reliable than analogue integration but in a polychromator system the provision of many integrating analogue-to-digital (AID) convertors becomes much more expensive than employing say high-quality polystyrene feed-back capacitors. Follow- ing analogue integration using the latter the stored signal can be read out via a single A/D convertor into a computer for data processing including background correction. Ana- logue systems may be used for background correction either by subtraction of the signal in a background channel from that in the analyte channel or if the background is highly reproducible by applying an electrically generated backing- off signal.More complex analogue signal processing is necessary if real-time internal standardization is to be attempted. It has been that much of the noise in ICP emission signals may be attributed to overall atom vapour density fluctuations hence fluctuations in analyte and internal standard element and background emissions are correlated. If the background is negligible compared with the analyte and internal standard signals signal division i.e. real-time ratioing of analyte to internal standard signals is claimed to reduce noise ten-fold to give an RSD =0.1°/o.L80 When the background is not negligible it should be subtracted from the analyte and internal standard signals prior to calcula- tion of the analyte ratio.It has been observed that the background itself can be used to provide a satisfactory internal reference signal for photographically recorded d.c. arc spectra.’ 1 1 ~ 1 8 1 This ap- proach may be relevant to direct-reading ICP-AES in some situations. For example the spectral background was used as the upper value and the dark current of the detector as the lower value in a two-point method for correction of drift in an ICP emission spectrometer. 18* In instruments employ- ing background correction by means of optical modulation of the analytical line and its background analogue elec- tronic demodulation produces a background-corrected sig- nal which may be subsequently digitized. If the background is structured or widely variable within the scanned spectral range particular care is needed to ensure the accuracy of the background correction.As noted earlier in relation to spectrography graphic display or profiling of the spectrum in the neighbourhood of the analytical line is a great help in selecting the optimum background measurement wave- length. Digital signal processing backed by computer data pro- cessing has made possible accurate and highly sophisticated background correction. In addition computer control with feedback to the source and spectrometer can be used to ensure that the instrument is always operating under clearly defined conditions which helps to standardize the line to background relationship. Peak detection and off-peak back- ground selection programs are included in the control of scanning spectrometers.The signals generated by the spectrometer may be fed either continuously into the computer as from a scanning monochromator or FT spectrometer or batchwise at the end of the exposure as in a polychromator. The information may be a complete spec- trum from a scanning spectrometer or array detector or more commonly discrete information on analytical line intensities and their associated backgrounds. In some analyses subtraction of a mean background intensity from the line intensity is sufficient correction to give accurate results. Where complex and variable matrices are to be analysed advantage can be taken of the data-handling capacity of modern laboratory computers. Libraries of spectral data may be stored in the com- p ~ t e r l ~ ~ and used to identify unknown components of the sample and to provide data for background correction.The spectral information may originate in published tables of lines their wavelengths widths and relative intensities or from direct measurement of standard samples of known composition. The former information may be used to identify possible line and so assist in the selection of interference-free analytical lines185 or if that is not possible to provide a means for calculating the amount of interference signal to be subtracted from the analytical line intensity. This approach is particularly relevant when a wide variety of sample matrices are to be analysed. In order to characterize the background generated by a particular sample type and instrument the second approach is required. The individual spectra of the analyte and major component elements are excited observed and stored in the computer.These spectra are then combined in varying proportions to synthesize a spectrum identical with that of the analytical sample provided that no direct interactions between the components modify line intensi- ties. The information provided by both sources may be combined to determine the optimum wavelength and procedure for background correction. Many computer programs have been written to improve background correc- tion and the accuracy of analysis. Some of these will now be reported. The Kalman filter which was originally developed for use in engineering,L86 was applied to analytical atomic spectro- metry by Van Veen and De Loos-V~llebregt.~ The spectrum of the sample is compared with reference single-element spectra stored in the computer.Following initial estimates of analyte and interferent content an iterative procedure is used to combine the reference spectra to give background correction improve analytical accuracy reduce detection limits and locate unknown lines not found in the reference spectra. A similar but simpler approach is based on multivariate curve-fitting methods with least-squares sum- mation of the spectra of single-element solutions to repli- cate the sample ~ p e c t r u m . ~ ~ ~ ’ ~ ~ ~ Methods for the detection of line overlaps can be based on measurement of the apparent line width.IB9 In one methodIgO a cubic semi-spline534 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 algorithm is used to fit emission line intensity data to a smooth curve and so obtain full-width ratios from the sample spectrum for comparison with those of standard spectra. Following a similar approach Francois and Jans- sens191 developed an algorithm based on Zimmermann’s method for the calculation of the full width at half maximum intensity. When an isolated analytical line is scanned at 1 pm intervals its profile is usually symmetrical whereas over the same spectral region that of an interferent is either linear or asymmetric. This property has been used to derive a quadratic polynomial function for background correction. The same workers developed a fully automated procedure based on digital filtration. 192 They emphasized that an important condition for accurate background correction is the availability of a sufficient number of appropriate reference channels.The potential for sophisti- cated background correction procedures using computers appears to be almost limitless; what may prove to be the limit is the professional resources that can be spent on writing the programs and installing them with all the instrument- and matrix-specific data in the spectrometer’s computer! Fourier self-deconvolution has been proposed as a proce- dure for enhancing resolution by mathematically reducing the line width of Lorentzian line pr0fi1es.I~~ A general text on Fourier transforms is presented in ref. 194. The technique is used generally in infrared spectrometry where spectral line profiles are Lorentzian. The procedure relies on the principle that convolved line profiles can be described in the Fourier domain by a simple multiplication operation and can therefore be separated easily.Inverse transformation leaves the deconvolved and reduced line width. This operation can only be achieved if the true line profile is known and an appropriate weight function can be designed for the Fourier deconvolution. The line-narrowing factor K can be expressed as K= FWHH( U/)/FWHH( W,) where FWHH(U/) and FWHH(W,) are the full width at half-height of the assumed line shape Wand the output line shape W respectively. However the procedure only works well when the S/N ratios are high. There is a trade-off between the narrowing factor K and SIN value as K increases the S/N decreases. In general this trade-off is common to other mathematical procedure used to enhance resolution e.g.derivative techniques. Although these procedures can provide good resolution enhancement in situations where the line profile is well known and the S/N ratios are not dominated by source noise e g . in IR spectrometry they are little used in AES where the aim of improving resolution is to separate weak lines in the wings of strong lines and where source noise dominates the S/N characteristics of the measurement. Numerical derivative spectrometry has been used in ICP-AES to reduce the effect of spectral interference and wavelength positioning errors.195 Several-fold improvements in S/B ratios and detection limits were reported. Conclusion Background correction is implicitly or explicitly effected in all quantitative spectrochemical analyses.It is directed at the removal of the contribution of the background to the output signal and to generate a signal that is a function of the analyte emission alone. It is good practice to minimize the background signal by choice of the excitation system optimization of instrumental parameters and where pos- sible appropriate sample processing. The magnitude origin and spectrum of the irreducible background should also be determined whenever possible in order that the appropriate correction procedure can be selected. Ideal background correction requires the simultaneous collection of accurate data on analytical and background signals and a knowledge of the relationship between the background at the analytical wavelength and the wavelengths at which it was able to be measured.The determination of that relationship may be aided by means of spectrum analysis employing derivative or Fourier techniques. These tools are however generally more appropriate for the study of band rather than line spectra and as yet do not have a significant role in quantitative atomic spectrochemical analysis. Instrumental background correction procedures are in widespread use. They facilitate data acquisition and pro- cessing but can lead to reduced precision generate artefacts and encourage a false sense of security. Computer data processing makes possible the use of sophisticated correc- tion procedures based on a priori knowledge of the spectral characteristics of the background and thus facilitates accurate analysis over a wide range of sample types and concentrations.On the other hand when the matrix does not vary significantly in composition from one sample to the next background correction may be effected simply by calibrating the analysis with synthetic standards similar in composition to the sample. In practice the choice of background correction procedures may well be determined more by the exigencies of the moment e g . the nature of the samples the accuracy required the time available and the availability of the instruments and staff than by a rational appraisal of the optimum solution. It has been the aim of this review to provide a basis for making the best decision whatever the circumstances. From this review it is apparent that although the fundamental problems of and limitations imposed by background in quantitative spectrochemical analysis are those that have challenged spectroscopists for over a century technological developments and greater under- standing of fundamental processes have greatly facilitated improved background correction.These developments have included more stable sources with less background holographic gratings and coated lenses with less stray light sensitive photoelectric detectors with low dark currents electronic signal and data processing and the application of statistical techniques and noise analysis. It is safe to predict that much future development will be along these previous well trodden paths. The most likely direction of develop- ments will be with respect to array detectors and computer processing of spectral data to improve the efficiency of existing techniques.Lasers and electrothermal devices e.g. furnace atomic non-thermal excitation spectrometry (FANES) may find a place as producers of atomic vapours although they tend to generate relatively intense back- ground radiation. Improvements in light collection and throughput of optical systems and reduction of aberrations could speed up analysis. Reduction in detection limits and improvements in precision would be achieved if propor- tional noise were eliminated by improved sample introduc- tion and signal processing and thereby the photon noise limit reached. However the pressure for these improve- ments may not be overwhelming in that present methods meet most analytical needs and the limit to the analysis is often the quality of the sample and not its measurement.Where a large improvement in analytical performance is required it is likely to be met best by a change of technique. References 1 Prugger M. and Torge R. Ger. Pat. 1964469 1969. 2 Rudiger K. Gutsche B. Kirchof H. and Herrmann R. Analyst 1969 94 204. 3 Steele A. W. and Hieftje G. M. Appl.. Spectrosc. 1986,40 357. 4 Otto M. Analyst 1990 115 685. 5 Dawson J. B. Duffield R. J. King P. R. Hajizadeh-Saffar,JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 199. VOL. 8 535 6 7 8 9 10 1 1 12 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 M. and Fisher G. W. Spectrochirn. .4cta Part B 1988 43. 1133.Collins J. B. Ivaldi J. C. Salit M. L. and Slavin. W. A [ . Spectrosc. 1990. 11 109. Van Veen E. H. and De Loos-Vollebregt M. I. (-.. Spectrochirn. Acta Part B 1990 45 3 13. Van Veen E. H. and De Loos-Vollebregt M. T. ('. .4nd Chem. 199 1 63 I44 1 . Analytical Methods Committee Analvst 1987 112 199. Winefordner J. D. Avni R. Chester T. L. Fitzgerald J. J. Hart L. P. Johnson D. J. and Plankey F. W. Spectrochitn. Acta Part B 1976. 31 1. Boumans P. W. J. M. Spectrochim. Acta Part B 1991 46 431. Boumans P. W. J. M. Ivaldi J. C. and Slavin W. Spectrochim. Acta Part B 199 1 46 64 1. Belchamber R. M. and Horlick. G. Specfrochim. A d a Part B 1982 37 17. Salin E. D. and Horlick G. Anal. Chem. 1980 52 1578. Walden L. Bower J. N. Nichdel S. Bolton D. L.. and Winefordner J. D.Spectrochim. Acta Part B 1980 3s. 535. Photomultipliers Thorn EMI. Ruislip 1986. Alkemade C. T. J.. paper presented at the 5th International Conference on Atomic Spectroscopy Melbourne 197'5. Alkemade C. T. J.. Snelleman W. Boutilier G. I). and Winefordner J. D Spectrochim. Acta Part B 1980 35 26 1. Talmi Y. Crosmun R. and Larson N. M. ,4nal. Chem. 1976 48 3 19. Hieftje G. M. Bystroff R. I. and Lim R. .Anal. ('hem. 1973 45. 253. Hobbs P. C. D. Opt. Photon. NeMs 1991 April 17. Mossotti V. G. and Duggan M. .4ppl. Opt. 1968 7. 1325. Wolf R. E. Skogerboe R. K. and Rosentreter J. .I.. .1nd Chern. I99 1 63 540. Mavrodineanu R. and Boiteux H. Flame Specttnscopj Wiley New York. 1965 p. 168. Kirkbright G. F. and West T. S. Appl. Opt. 1968. 7 1305. Price W.J.. and Whiteside P. J. Commun. 18th Colloquium Spectroscopicum Internationale Grenohle 1 9 7 5 . G. A.M.S. Paris Vol. 3 p. 746. Strock L. W. Spectrum Analvsis Mith the C a r / m .4rc Cathode Laver. Hilger London 1936. Price W. J.. Spectrochim. .4cta 1955 7 118. Mork B. J. and Scheeline A. Spectrochim. Actu Part B 1989 44 1297. Crosswhite H. M. Steinhaus A. W.. and Deake G . H.. J . Opt. Soc. Am. 1951 41 299. Bardocz V. A. and Varsanyi F. Z . Naturforsch. 7 . d ,4 1955 10 1031. Miyami T. Fukui I.. and Imamura N. paper presented at the Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy Atlanta GA 1989. Myers A. T. and Brunstetter B. C. Anal. Chem. 1047 19 71. Todorovic. M. Vukanovic V. and Georgijevic V. !ipectro- chitn. Acta Part B 1969 24 571.Lummerzheim D. and Nickel H. Fresenius' %. Anal. Chem. 1969 245 267. Mihajilidi T. Simic M. Simic D. and Georgijevic J. J . Serb. Chem. SOC. 1987 52 215. Gordon W. A. and Chapman G. B. Spectrochim. Ac la Part B 1970 25 123. Scribner B. F. and Margoshes M. Spectrochim. Actu 1959 15 138. Pavlovic M. S. Pavlovic N. Z. and Marinkovic. M. J . Anal. At. Spectrom. 1989 4 581. Keirs C. D. and Vickers T. J. Appl. Spectrosc. I977,31,273. Meyer G. A. Spectrochim. Acta Part B 1987 42 333. Fassel V. A. Plenary Lectures and Reports 16th Colloquium Spectroscopicum Internationale Heidelberg 197 I Adam Hilger London p. 63 Ebdon L. and Sparkes S. T. ICP InJ Newsl. 1985 10 797. Le Blanc C. and Blades M. W. J. Anal. At. Spectrom. 1990 5 99. Slinkman D. and Sacks R. Anal. Chem.1990 62 1656. Sparkes S. T. and Ebdon L. J. Anal. At. Spectrom. !988,3 653. Bamiro F. O. Littlejohn D. and Marshall J. J. Anal. At. Spectrom. 1988 3 279. 48 49 50 51 52 53 54 5 5 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 Reed T. B. J . ,4ppl. Phys. 1963 34. 2266. Ediger R. D. and Fernandez F. J. At. Spectrosc.. 1980. 1. 1 . Tracy D. H.. and Myers S. A. Spectrochim. Acta Part H 1982 37 1055. Lovett R. J.. Federation of Analytical Chemistry and Spectroscopy Societies (FACSS) Philadelphia 1979 Ab- stracts. McLaren J. W. and Berman S. S. Spectrochim. Actu. Purt B 1985 40 21 7. Boumans P. W. J. M.. and Vrakking J. J. A. M. ,/. Anal. '-11. Spectrom. 1987 2 5 13. Boumans P. W. J. M. and Vrakking J. J. A. M. Spectro- chim..4cta Part B 1987 42 553. Marshall J. Anal. Proc. 1988 25 238. Webb B. D. and Denton M. B. Spectrochim. ,4cta Part R 1986 41 361. Davies J. and Snook R. D. .4nalyst 1985 110 887. Wallace G. F. At. Spectrosc. 1981 2 93. Montaser ,4. Clifford R. H. Sinex S. A. and Capar S . G.. J. Anal. .4t. Spectrom. 1989 4 499. Davies J. Dean J. R. and Snook. R. D. .4nalj~st 1985. 110. 535. De Loos-Volbregt M. T. C. Tiggelman J. J. and De Cialan. L. Spectrochim. .Icta. Part B 1988 43. 773. Krull I. S. Buchel D. S. Schleicher R. C. and Smith S. B. Analvst 1986. 111 345. Larson G. F. and Fassel V. A.. Appl. Spectrosc. 1979 33 592. Maessen F. J. M. J. Balke J. and De Boer J. L. M. Spectrochim. Acta Part B. 1982. 37 517. Umemoto M. and Hayashi K. Bunseki Kagaku 1990. 39 T35.de Loos-Vollebregt M. T. C. and Tiggelman J. J.. 5th Biennial National Atomic Spectroscopy Symposium Lough- borough 1990 63 Atomic Spectroscopy Group Royal Society of Chemistry London. Greenfield S. and McGeachin H. McD. ,4nal. C'him. .4ctu 1978 100 101. Zaray G. Broekaert J. A.. Boehmer R. (3.. and Leis. F. i'alanta 1987 34 629. Ohls K. ICP In$ Nebid. 1981. 7 6. Ebdon L. Cave M. R. and Mowthorpe D. J. .4nal. C'hrm. Acta 1980 115 179. Chan S. K.. and Montaser. A.. .4pp1. Spectrosc.. 1987 41. 545. Tan. H. M.. Chan S. K. and Montaser A.. .4nal. C'hem. 1988 60 2542. Magyar B. Leinemann P. and Vonmont H.. Spectrochim. Acta Part B 1986 41 27. Umemoto M. and Kubota M. Bunko h'enkvu 1990,39 19. Steele A. W. and Hieftje G. M. Appl. Spectrosc. 1986. 40 1127. Downey S.W. and Hieftje G. M. Anal. Chim. ..lc/a 1982 141 193. Mitchell J. C. Steele A. and Hieftje G . M.. Anal. Chim. Acta 1987 200 539. Cobine J. D.. and Wilbur D. A. J. Appl. Phy.r. 1951 22. 835. Mavrodineanu R. and Hughes R. C. Spectrochim. Acta 1963 19 1309. Zander A. T. and Hieftje G. M. Appl. Spectrosc. 1981 35 357. Heltai G. Broekart J. A. C. Burba P. Leis. F. Tschoepel P. and Tolg G. Spectrochim. Acta Part B 1990 45 857. Mohamed M. M. Uchida T. and Minami S. Appl. Spectrosc. 1989 43 129. Forbes K. A. Reszke E. E. Uden P. C. and Barnes R. M. J. Anal. At. Spectrom. 1991 6 57. Abdillahi M. M. Tschanen W. and Snook. R. D. Anal. Chim. Acta 1985 172 139. Abdillahi M. M. and Snook R. D. Analyst 1986 111 265. Alder J. F. Jin Q-.H. and Snook R. D. Anal. C'him. Acta 1980 120 147.Alder J. F. Jin Q.-H. and Snook R. D. Anal. C'him. Acta 1981 123 329. Goode S. R. Gemmil J . J. and Watt B. E. J . Anal. A t . Spectrom. 1990 5 483. Goode S. R. and Kimbrough L. K. J . Anal. At. Spectrom. 1988 3 915.536 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 i l l 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 McNally J. R. Jr. Harrison G. R. and Rowe E. J. Opt. SOC. Am. 1947 37 93. Caroli S. J. Anal. At. Spectrom. 1987 2 661. Pavlovic B. V. and Dobrosavljevic J. S. Spectrochim. Acta Part B 1989,44 1191. Caroli S. Senofonte O. Violante N. and Astrolago R. J. Anal. At. Spectrom. 1988 3 887. Wagastuma K. and Hirokawa K. Anal.Chem. 1989 61 2 137. KO J. B. Spectrochim. Acta Part B 1984 39 1405. Caroli S. Senofonte O. Violante N. Petrucci F. and Alimonti A. Spectrochim. Acta Part B 1984 39 1425. Falk H. Hoffmann E. and Ludke Ch. Prog. Anal. Spectrosc. 1988 11 417. Sturgeon R. E. Willie S. N. Luong V. T. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. Leis F. Sdorra W. KO J. B. and Nieman K. Mikrochim. Acta 1989 2 185. Lundegardh H. Die Quantitative Spektralanalyse der Ele- mente Band 11 Gustav Fischer Jena 1934. American Society for Testing and Materials Practices for Photographic Processing in Optical Emission Spectrographic Analysis ASTM Standard El 15-87 ASTM Philadelphia 1987. Zimmer K. Heltai G. Florian K. and Veress G. Kem. Ujubb. Eredmenyei 1989 70 1. Tellinghuisen J.Appl. Opt. 1991 30 1723. Mitchell R. L. The Spectrographic Analysis of Soils Plants and Related Materials Commonwealth Bureau of Soil Science Harpenden 1948. Nachtrieb N. H. Spectrochemical Analysis McGraw-Hill New York 1950. Mermet J.-M. J. Anal. At. Spectrom. 1987 2 681. Plsko E. paper presented at the VIIth Polish Spectro- analytical Conference and X CANAS Torun 1988. Torok T. and Hafenscher I. Spectrochim. Acta Part B 1978 33 283. Schwarzchild K. and Villiger W. Astrophys. J. 1906 23 287. Bournans P. W. J. M. and Bosveld M. Spectrochim. Acta Part B 1979 34 59. Harvey C. E. Spectrochemical Procedures Applied Research Laboratories Glendale 1950. Mayer A. and Price W. M. Chemical and Spectrographic Analysis of Magnesium and its Alloys Magnesium Elektron Manchester 1958 p.52 17. Price W. J. J. Appl. Phys. 1959 10 352. Torok T. and Zimmer K. Quantitative Evaluation of Spectrograms by Means of I-Transformation Heyden Lon- don 1972. Heltai G. and Zimmer K. Acta. Chim. Acad. Sci. Hung. 1980 103 171. Huang Z. Chen X. Zhao X. and Lu D. Lihua Jianyan Huaxue Fence 1987 23 19. Hasler M. F. and Dietert H. W. J. Opt. SOC. Am. 1944,34 751. Hasler M. F. Lindhurst R. W. and Kemp J. W. J. Opt. SOC. Am. 1948 38 789. Murphy M. P. Parsons M. L. Staten R. T. and Hinckle L. A. paper presented at the Pittsburgh Conference on Analytical Chemistry and Applied Spectroscopy Atlantic City NJ 1982. Jansen W. H. Heyes J. and Richter C. Z. Phys. Chem. Abb. A 1935 174 291. Warren R. L. Analyst 1965 90 549. Boettner R. L. and Brewington G. P. J.Opt. SOC. Am. 1944 34 6. Saunderson J. L. Caldicourt V. J. and Petersen E. W. J. Opt. SOC. Am. 1945 35 681. Brault J. W. J. Opt. SOC. Am. 1976 66 1081. Faires L. M. Anal. Chem. 1986 58 1023A. Stubley E. A. and Horlick G. Appl. Spectrosc. 1985 39 800. Hobbs P. Spillane D. E. M. Snook R. D. and Thorne A. P. J. Anal. At. Spectrom. 1988 3 543. Kolar W. and Williams R. Appl. Spectrosc. 1992 46 615. Voigtman E. and Winefordner J. Appl. Spectrosc. 1987 41 1182. Marra S. and Horlick G. Appl. Spectrosc. 1986 6 804. Williams R. Appl. Spectrosc. 1989 43 235. 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 Wingerd M. A. and Dessy R. E. Appl. Spectrosc.1990,44 1444. Olesik J. W. Chem. Anal. (N. Y.) 1987 90 466. Hill S. Dawson J. B. and Thorne A. P. J. Anal. At. Spectrom. 1990 5 15 1 R. Analytical Methods Committee Anal. Proc. 1986 23 109. Analytical Methods Committee Anal. Proc. 1987 24 3. Boumans P. W. J. M. and Vrakking J. J. A. M. Spectro- chim. Acta Part B 1986 41 1235. Boumans P. W. J. M. Vrakking J. J. A. M. and Heijms A. H. M. Spectrochim. Acta 1988 43 1365. Quian S. and Zhang W. Fenxi Ceshi Tongbao 1989,8,62. Goodman D. Opt. Photon. News 1992 January 52. Annu. Rep. Anal. At. Spectrosc. 1976 6 12. Fassel V. A. Katzenberger J. M. and Winge R. K. Appl. Spectrosc. 1979 33 1. Hayat G. S. and Pieuchard G. Diffraction Gratings Ruled and Holographic Jobin Yvon Longjumeau 1972. Dymott T. C Int. Lab. 1989 March 26.Koirtyohann S. R. and Lichte F. E. Can. J. Spectrosc. 1978 23 101. De Wit A. and Neugebauer R. J. Spectrochim. Acta Part B 1984 39 939. Haisch U. Laqua K. and Massmann H. Spectrochim. Acta Part B 1983 38 1227. Friedhoff P. Overkamp B. and Norris J. paper presented at the 10th Spektrometertagung The Hague 1974. Steiner M. C. and Boumans P. W. J. M. Abstract Handbook XXI Colloquium Spectroscopicum Internationale and 8th International Conference on Atomic Spectroscopy Cam- bridge 1979 99 Association of British Spectroscopists Cambridge. Pye Unicam Anal. Bull. 1980 3(3) 7. Snelleman W. Rains T. C. Yee K. W. Cook H. D. and Menis O. Anal. Chem. 1980 42 394. Zander A. T. ASTM Spec. Tech. Publ. 1981 No. 747 64. Michel R. G. Sneddon J. Hunter J. K. and Ottaway J. M. Analyst 1981 106 288.Layman L. R. and Lichte F. E. paper presented at FACSS Philadelphia 1979. O'Haver T. C Harnly J. M. Marshall J. Carroll J. Littlejohn D. and Ottaway J. M. Analyst 1985 110 451. Baxter D. C. Duncan I. S. Littlejohn D. Marshall J. Ottaway J. M. Fell G. S. and Ataman 0. Y. J. Anal. At. Spectrom. 1986 1 29. Burman J. Johansson B. Morefalt B. Norfelt K. and Olsson L. Anal. Chim. Acta 1981 133 379. Maines I. S. Mitchell D. G. Rankin J. M. and Bailey B. W. Spectrosc. Lett. 1972 5 251. Topp J. A. Schrotter H. W. Hacker H. and Brandmuller J. Rev. Sci. Instrum. 1969 40 1164. Langer O. Foit J. and Polej B. Handbook XIV Collo- quium Spectroscopicum Internationale Debrecen 1 967 799 Hungarian Academy of Sciences Budapest. Hasegawa S. and Endo K. Bunko Kenkyu 1967 15 155. Annu. Rep. Anal. At. Spectrosc. 1971 1 25. Littlejohn D. Jowitt R. Sparkes S. T. Thorne A. P. and Walton S. J. J. Anal. At. Spectrom. 1991 6 145R. Charge Coupled Devices Princeton Instruments Trenton NJ 1991. Vanderpant L. and Taylor S. Spectrosc. World 199 1,3(2) 12. MacKay C. D. Photon. Spectra 1992 February 113. Optoelectronic and Image Sensors Texas Instruments Dal- las TX 1990. Jones D. G. Anal. Chem. 1985,57 1057A. Sweedler J. V. Bilhorn R. B. Epperson P. M. Simms G. R. and Denton M. B. Anal. Chem. 1988,60 282A. Sweedler J. V. Jalkian R. D. and Denton M. B. Appl. Spectrosc. 1989 43 953. Janesick J. R. Appl. Opt. 1992 31 1893. Boumans P. W. J. M. Colloquium Spectroscopicum Inter- nationale Cambridge 1979 8 1 Association of British Spectroscopists Cambridge. Webb D. P. and Salin E. D. Talanta 1990 37 33. Brushwyler K. and Furuta N. and Hieftje G. M. Talunfu 1990 37 23. Glick M. Brushwyler K. R. and Hieftje G. M. Appl. Spectrosc. 199 I 45 328.JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 537 176 177 178 179 180 181 182 183 184 185 I86 187 Scheeline A. Bye C. A. Miller D. L. Rynders S. W.. and Owen R. C. Jr. Appl. Spectrosc. 1991 45 334. Bilhorn R. B. and Denton M. B. Appl. Spectrosc. 1390,44 1538. Sims G. R. and Denton M. B. Talanta 1990 37 I . Belchamber R. M. and Horlick. G. Spectrochim. Ac fa Part B 1982 37 1037. Myers S. A. and Tracy D. H. Spectrochim. Acta Part 13 1983 38 1227. Hahn-Weinheimer P. and Frank P. Preprints XV I Collo- quium Spectroscopicum Internationale. Heidelberg 197 1 Adam Hilger London vol. 1 p. 209. Gueldner D. Neue Hiitte I99 1 36 109. Legere G. Burgener P. and Froggatt N. Specttoscopy 1988 3( 19). 14. Burton L. and Blades M. W. Spectrochim. Acta Part H 1988 43 305. Webb D. P. and Salin E. D. J. Anal. At. Spectrom. 1989,4 793. Kalman R. E. J. Basic. Eng. 1960 82 35. Wangen L. Gallimore D. and Bentley G. paper prcsented at FACSS XV Boston 1988. 188 189 190 191 182 193 194 I95 Lacey R. F. .4ppl. Spectrosc. 1989 43 1 135. Boumans P. W. J. M. Spectrochim. Actu Part B 1989 44 1325. Farwell S. O. and Kagel C. T. Appl. Spectrosc. 1986 40 944. Francois J. P. and Janssens F. Spectrochim. Actu Part R 1990 45 177. Janssens F. and Francois J . P. Appl. Spectrosc. 1992 46 283. Kauppinen J. K. Moffat D. J . Hollberg M. R. and Mantsch H. H. Appl. Spectrosc. 1991 45 41 1. Marshall A. G. and Verdun F. R. Fourier Transforms in NMR Optical and Mass Spectrometry Elsevier New York 1990. Yang J. Piao Z. Zeng X. Zhang Z. Chen X. and Guan Q. Fenxi Huaxue 1992 20 153. Paper 2/04O03 D Received Julv 27 1992 Accepted January 15 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800517

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY. JUNE 1993 VOL. 8 539 Measurement of Inductively Coupled Plasma Infrared Atomic Emission of Carbon and Oxygen From Alcohols Using Fourier Transform Infrared Spectrometry C. A. Morgan B. W. Smith and J. D. Winefordner* Department of Chemistry University of Florida Gainesville FL 3261 1 USA Atomic carbon and oxygen infrared (IR) emission from a series of alkan-1-01s is presented. The measurements were made over the spectral range 3000-9000 cm-l. Two types of plasmas were employed (pure argon and argon-nitrogen) and atomic carbon and oxygen IR emission lines are tabulated for both plasmas. The measurements were obtained using a Fourier transform interferometer equipped with a deuterium triglycine sulfate (DTGS) and an lnSb detector. It was found that the 40.68 MHz inductively coupled plasma used in this study did not generate enough power to sustain a stable plasma on introduction of the higher molecular mass alcohols.However enhancement of the atomic carbon and oxygen IR emission from methanol ethanol and propanol was achieved with the argon-4% nitrogen mixed plasma. The background emission from the pure argon plasma was characterized and a table of relative intensities and wavelengths of the argon atomic IR emission is presented. Keywords Alcohols; inductively coupled plasma Fourier transform infrared spectrometry; atomic emission In recent years interest has been shown in developing applications for the inductively coupled plasma (ICP) in the near- to mid-infrared region.'-'' The ICP provicles an efficient vaporization.atomization and emission cell capable of multi-element determination with few matrix interference effects. Several studies have been carrkd out using atomic infrared (IR) emission in the near-IR (NIR) region with the ICP as the emission source. Fry arid co- workers1-* studied the NIR emission of N 0. F C1 Br S C and H from organic compounds excited in an argoii ICP. Lieu et al.9 determined oxygen in silver halides using oxygen NIR lines. Blades and Hauserlo determined sulfur NIR emission lines below 10000 cm-I. Stublei and Horlick'' studied emission lines of C H 0 and S in the range 14 000-4000 cm-I to determine the analytical iLseful- ness of the sulfur NIR emission lines. As can be seer from these studies the focus has been on the spectral characteri- zation of non-resonance NIR emission lines of nonmetals whose resonance emission lines lie in the more diffic-ult to analyse vacuum ultraviolet region.In this work the atomic IR spectroscopic characteristics of organic compounds specifically a series of alkan - 1 -oh introduced into two types of plasmas a pure argon plasma and an argon-nitrogen plasma were studied. The analytical region of interest ranges from 9000 down to 3000 Lm-'. The instrumental set-up includes the coupling of a 40 MHz ICP with a Fourier transform infrared (FTIR) spectro- meter. In addition to reporting both carbon and oxygen atomic IR emission of the series of alcohols under Ijtudy the background emission of the pure argon plasma is characterized. Coupling the ICP with the FTIR spectrometer offered many advantages.Together with the high atomilTation efficiency of the ICP the FTIR spectrometer provided fast highly accurate and highly reproducible wavelength mea- surements in addition to high resolution capabilities I 2 Experimental The ICP system used as a source in this experiment was manufactured by Baird (Bedford MA USA). This source consisted of a Model 3000-D 2.5 kW r.f. generator (40.68 MHz) and an automatic matching network. The air-cooled demountable torch (Plasma-Therm Kresson NJ USA) was * To whom correspondence should be addressed. held concentric to the three-turn copper coil of the r.f. generator by a Teflon clamp attached to an x-y- transla- tional stage. All samples were delivered with a peristaltic pump into a Type A Meinhard concentric glass nebulizer and Scott spray chamber at a rate of 0.5 ml min-' and subsequently introduced as aerosols into the axial channel of the plasma.The Fourier transform interferometer used in this study was a BOMEM Model DA3.02 commercial laboratory instrument with a resolution capability of 0.04 cm-l. I t was equipped with a deuterium triglycine sulfate (DTGS) and an InSb detector. All data reported in Tables 2-4 were obtained using the more sensitive InSb detector. For all samples studied 64 interferograms were signal averaged. All spectra were calculated from single-sided. unaliased interferograms after application of strong13 apodization. The operating parameters for both the ICP and the interferometer were optimized for the analysis of organic compounds with respect to ICP power gas flow rates sample uptake rate percentage of nitrogen observation zone.spectrometer scan speed apodization and gain. The resulting optimized conditions are given in Table 1 . Table 1 Operating conditions for ICP and FTIR spectrometers Value Parameter ICP- R.f. frequencylMHz Power input t o ICP/kW Reflected po wed W Coolant gas flow ratell min-l Plasma gas flow rate/l min-' Carrier gas flow rate/l min-' Sample uptake rate/ml min-l N in plasma gas (O/O) Observation zone/mm above load coil Spectral rangdcm-' Resolution/cm- ' Scan speedlcm s-' Apodization Base gain Positional gain FTIR- Argon- Pure argon nitrogen plasma plasma 40.68 1.3 0- 100 11.5 0.6 0.9 0.5 5 40.68 1.5 I40 14.0 0.8 0.9 0.5 4 10 3000-9000 0.707 0.10 Strong 4 64540 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 Radiation from the plasma was optically coupled to the interferometer with a pair of off-axis paraboloidal reflectors (M1 and M2) arranged so that the radiation from the plasma gathered by the 119 mm focal length f/l mirror (Ml) was focused on to the curve surface of the 83.6 mm focal length f/2 mirror (M2). Subsequently a 1 1 image of the plasma from M2 was focused 1 .O cm in front of the 10 mm circular emission port of the interferometer. The reflector M1 was positioned perpendicular to and 18.5 cm from the centre of the plasma; M2 was positioned directly in front of but 35 cm away from M1. M2 was also oriented perpendicular to and 9.4 cm away from the emission port of the interferometer.An RG 1000.S 1000 nm cut-off filter (Corion Holliston MA USA) was placed in front of the entrance aperture in order to minimize measurement of stray light. The light entering the interferometer was collected by an internal plane mirror reflected to an f/4 collimating mirror directed to a KBr beamsplitter and focused through a 5 mm aperture on to either a DTGS or InSb detector. The resulting interferogram was transformed using a dedicated DEC PDP- 1 1 computer. The gases for the mixed plasmas were blended using a Model 5878 thermal mass flow controller (Brooks Instruments Veenendaal The Netherlands) Successful ignition of the argon-nitrogen plasma required special techniques. The torch had to be raised so that the lip was approximately 3 mm above the load coil and the plasma gas flow rate had to be increased to 1.5 1 min-l.The impedance matching network had to be manually tuned and the power had to be increased to maintain stability. The most stable argon-nitrogen plasma was maintained with a mixture of 4% of nitrogen with the argon coolant gas. Results and Discussion Background Emission in a Pure Argon Plasma The background emission from the pure argon plasma was observed over the range 3000-10000 cm-l using a resolu- tion of 0.707 cm-l. Although the FTIR spectrometer used in this study had a 0.040 cm-' resolution capability a compromise resolution was selected that offered baseline separation of most Ar I emission lines (Fig. 1) in addition to rapid scan times ( t 7 s per scan). Table 2 lists the observed Ar I infrared emission in terms of wavenumbers and wavelengths. Relative intensities were normalized with respect to the 7287.73 cm-l argon line.The data of Outred14 served as the reference for identification and classification of the argon lines. Argon emission lines not previously reported in the literature for ICP-AES are marked with asterisks. 1.95 * ; 1.44 + .- u) Q w .5 0.93 aa .- * - 0.42 -0.10 7300 7350 7400 7450 7500 Wavenumberlcm-' Fig. 1 cm-I Spectrum of pure argon plasma in the range 7300-7500 Analyte Emission in a Pure Argon Plasma A series of alkan-1-01s were introduced into a 40.68 MHz pure argon inductively coupled plasma. Initial attempts were made difficult because these alcohols tended to destabilize the plasma and constant manual tuning of the ICP impedance matching network became necessary in order to prevent extinction.To ensure that the entire series of alcohols was observed under identical conditions samples of each of the alcohols were nebulized sequentially into the same plasma. Higher relative molecular mass alcohols were measured first proceeding to the lower relative molecular mass alcohols with appropriate tuning of the impedance matching network. Ethanol and methanol altered the impedance of the pure argon plasma drastically always causing extinction of the plasma. However if the process was started with propan- 1-01 the impedance mis- match could be offset by manual tuning. Careful subse- quent introduction of ethanol and then methanol along with further tuning of the impedance matching network proved successful.On introduction of any of the alcohols a bright green tongue was observed in the axial channel of the plasma extending several millimetres above the load coil. This tongue had the characteristic green colour that has been attributed to the well known C2 Swan band. Although the emission was intense in the visible region it was too weak in the IR region to cause any interference with the atomic IR emission. Both carbon and oxygen atomic IR emission was observed for all eight alcohols and is reported in Table 3. Table 2 Background emission lines from pure argon plasma in the range 3000- 10 000 cm-I Wavenumberkm-I Wavelengthhm Energy levels/cm-I J 3356.164* 3474.432* 3896.306* 39 19.885* 3979.272* 41 72.18 1* 4193.1 17* 432 1.837* 4528.633* 4642.726* 4763.8 79* 482 1.938* 4842.097* 4849.770 4920.740* 4992.162* 5007.286* 2979.59 2878.16 2566.53 2551.09 25 13.02 2396.82 2384.86 23 13.83 2208.17 2 153.90 2099.13 2073.82 2065.22 206 1.95 2032.2 1 2003.14 1997.09 120229-12 3557 1 1 3468-1 16942 1 1 4975-1 1 8870 1 1 3643-1 1 7562 1 1 3020-1 1 6999 10 7496-1 1 1667 1 1 2750-1 16942 107496-1 1 1818 107289-1 1 1818 107496-11 2138 107054-1 1 1818 1 1 5366-120188 1 1 1818-1 16660 107289-1 12138 10 8722- 1 1 3643 1 1 1667-1 16660 107131-11 2138 1-2 2-3 1-0 1-0 3-2 1-0 4-3 1-1 2- 1 1-2 0- 1 1-2 1-1 2-2 0- 1 0- 1 1-2 Intensity 1.43 1.66 2.07 2.10 4.23 1.62 8.59 1.35 1.64 2.14 3.8 1 1.75 1.34 8.1 1 1.80 1.06 1.06JOURNAL OF ANALYTICAL ATOMIC SPECTROME'TRY JUNE 1993.VOL. 8 54 1 Table 2-can ti n ued Wavenumber/cm-l Wavelength/mm 5044.726* 5365.41 9* 5383.603* 5424.936 5428.053* 5580.567 5589.175* 5609.130 573 1.087 590 1.5 1 8 5972.155 6040.940 6052.094 6082.592 6 178.70 I 6252.160 6287.643 6490.763 65 13.028 6533.577 6588.759 6644.622 6824.159 683 1.665 6849.433 70 12.586 7093.8 17 7 187.169 7 22 9.820 7230.167 7287.730 7308.924 7339.127 735 I .635 7365.662 738 I .460 7403.336 7405.820 7457.168 7479.375 7500.033 7509.353 75 15.735 7532.549 7555.778 7558.310 7566.269 7685.647 7716.214 7730.197 7809.203 7843.48 1 785 I .523 7870.566 8005.8 8 3 8026.1 12 8037.090 8049.840 8060.843 809 1.089 8099.533 8227.563 8234.962 8253.804 8370.488 8520.5 54 8567.837 8702.299 8737.486 9359.509 9366.773 954 1.449 9549.063 1982.26 1863.78 1857.49 1843.34 1842.28 1791.93 1789.17 1782.80 1744.87 1694.47 1674.43 1655.37 1652.32 1644.03 16 18.46 1599.44 1590.42 1540.65 1535.38 1530.55 15 17.73 1504.97 1465.38 1463.77 1459.97 1426.00 1409.67 1391.24 1383.16 1383.09 1372.16 1368.19 1362.56 1360.24 1357.65 1354.74 1350.74 1350.28 1 340.99 1337.01 1333.32 133 1.67 1330.54 1327.57 1323.49 1323.04 132 1.65 1301.12 1295.97 1293.62 1280.54 1274.94 1273.63 1270.55 1249.08 1245.93 1244.23 1242.26 1240.56 1235.92 1234.63 1215.42 1214.33 121 1.56 1 194.67 1 173.63 1167.15 1149.12 1144.49 1068.43 1067.60 1048.05 1047.22 Energy levels/cm-' 1 1 2138-1 17183 1 1 1818-1 I7183 I 1 4821-120207 1 1 4805- 12 0230 1 1 4821-120249 10087-1 1 1667 1 1 4641-120229 I 1 4641-120249 10 7289-1 1 3020 106237-1 12138 10 7496- 1 1 3468 1 14147-120188 106087-1 12138 1 14147-120230 10 7289-1 1 3468 108722-1 14975 I 1 5366- 12 1654 1 1 3716-120207 1 13719-120229 1 1 4716-120249 10 7054- 1 1 3643 10 7054-1 1 3643 108722-1 1 5366 1 1 3426- 12 0249 I 1 4821-12 1653 1 1 4641-12 1653 107054-1 14147 1 1 3020-1 2 0207 1 1 3020- 12 0249 106237-1 13468 10 5462-1 1 2750 10 7496-1 I 4805 106087-1 13426 10 7290-1 1 4641 107496-1 14861 106087-1 1 3468 105617-1 13020 106237-1 13643 1 1 2750-1 2 0207 106237-1 13716 1 12750-120249 107131-1 14641 10 7289- 1 1 4805 107289-1 14821 10 6087-1 1 3643 10 5462-1 1 3020 104102-1 1 1667 107289-1 14975 104102-1 1 1818 107131-1 14861 I0 5647- I 1 3426 107131-1 1497s 105617-1 13468 107496-1 15366 10 5462-1 1 3468 105617-1 13643 104102-1 12138 I 1 2138-120188 106087-1 14147 1 1 2138-120229 105619-1 13716 1 1 4326-12 1653 107131-11 5366 105462-1 I3716 1 1 1818-120188 1 1 1667-120188 - - - 1 1 3426-122686 - - - J 2-2 1-2 3-4 2-2 3-3 1-0 2-3 2-3 2-3 2-2 1-2 1-2 1-2 1-2 2-2 0- 1 1-2 3-4 3-3 3-4 0- 1 0- 1 2-3 3-4 2-2 2-3 0- 1 3-4 3-4 2-2 3-4 1-2 1-2 2-2 1-0 1-2 3- 3 2- I 4-4 2-3 4-4 1-2 2-2 2-3 1-1 2-3 1-0 2- 1 1-1 1-0 2-2 1-1 2-2 1-1 3-2 2- 1 1-2 2-2 1-1 2-3 2-3 2-3 1-1 3-3 1-2 0- 1 - - - 2- 1 - - - Intensity 4.86 1.01 I .42 6.30 2.68 8.50 1.08 4.27 2.02 3 1.77 3.13 5.78 3.95 9.72 1.24 6.06 7.7 1 1.94 2.01 8.00 1.96 4.13 4.46 6.6 1 3.97 6.5 1 9.77 14.8 1 3.25 5.61 100.0 34.28 40.09 5.97 3.24 1.98 80.99 5.49 3 1.45 75.78 3.36 48.98 1.04 30.56 10.19 15.59 10.51 18.71 22.87 6.52 22.73 3.19 5.06 12.2 1 35.37 18.93 20.10 2.07 18.50 8.22 6.57 1.71 5.7 1 13.43 4.26 1.77 7.67 1.06 1.13 0.82 3.80 1.03 2.70 * Argon emission lines not previously reported in the literature for ICP-AES.542 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 Table 3 Atomic IR emission from a series of alkan-1-01s in a pure argon plasma (expressed as a percentage of the strongest observed carbon line intensity uncorrected for overall instrument response). The alcohols are represented by the number of carbons they contain (from Cl = methanol to C8 = octan-1-01). Blank spaces in table indicate no observation of the transition above background noise Wavenumberkm- 5069.10 5282.14 5456.83 5479.87* 5486.64 5547.50* 5626.26 5668.23 57 10.89 5765.92 5 7 70.9 3 5787.14 5800.74 59 18.92 6874.52 6932.86 6942.72 7275.05 7294.39 7372.80 7644.98* 7925.47 8404.07 8505.91 8509.60 8574.22 Wavelengthhm C 1972.20 11.7 1892.65 6.7 1832.07 11.2 1823.36 8.9 1822.1 1 22.6 1802.6 1 14.9 1776.89 16.4 1763.74 15.2 1750.56 16.8 1733.86 48.9 1732.35 14.8 1727.5 17.7 1723.45 13.8 1689.04 100 1454.25 10.3 1442.0 1 11.6 1439.97 6.7 1374.56 1370.92 - 1355.97 1 - 1308.05 - 1261.41 10.5 1 189.58 7.1 1 175.65 29.6 1175.14 9.4 1166.29 5.2 - c2 25.4 - - 20.9 19.2 18.0 - - 41.0 24.8 - - 100 33.7 23.8 18.4 8.5 19.7 18.3 12.0 55.6 15.8 9.7 - - c3 - - - - 37.8 34.4 22.2 22.0 35.1 70.1 - - 22.0 33.5 30.2 20.6 100 - - 13.8 26.3 12.6 39.0 - - - c4 - - - - 26.6 - - 25.2 19.7 50.4 - - - 100 18.1 11.7 8.5 4.4 18.6 8.1 18.4 8.6 42.3 21.3 8.7 - * Atomic oxygen lines (all others are atomic carbon lines).Table 4 Atomic IR emission from a series of alkan-1-01s in an argon-nitrogen plasma (expressed as a percentage of the strongest observed carbon line intensity uncorrected for overall instrument response).The alcohols are represented by the number of carbons they contain from C1 = methanol to C3 = propan-1-01. Blank spaces indicate no observation of the transition above background noise Wavenumberkm - I W avelengthhm Energy levelskm - I 4364.37 2290.65 73 975-78 340 4702.41 2125.98 79 3 10-84 01 3 4713.13 2121.15 79 323-84 036 471 7.61 21 19.14 79 3 18-84 036 4755.37 2102.31 73 975-78 73 1 5069.10 1972.19 72 6 10-77 679 5282.14 1892.65 78731-84013 5305.16 1884.44 78 73 1-84 036 5456.83 1832.06 78 529-83 986 5479.87* 1824.36 97 488-102 968 5486.64 1822.1 1 78529-84016 55 1 1.24 18 13.98 69 744-75 255 5543.30 1803.48 69 7 101-75 253 56 12.02 1781.40 78 307-83 9 19 5626.26 1776.89 78 2931-83 9 19 5668.23 1763.73 78 3 18-83 986 57 10.89 1750.56 78 2 151-83 926 5765.92 1733.85 78 249-84 0 1 5 5770.93 1732.35 78 2 10-83 986 5787.14 1727.49 78 199-83 986 5800.74 1723.44 7 8 2 1 fi-84 0 1 6 59 18.92 1689.03 72 6 10-78 529 6239.86 1602.16 77 679-83 9 19 6246.42 1600.48 77 67!)-83 926 6874.52 1454.25 61 98’1-68856 6922.24 1444.22 7 1 385-78 307 6932.86 1442.0 1 7 1 385-78 3 18 6942.72 1439.96 7 1 364-78 307 7298.44 1369.78 64 086-7 1 385 736 1.03 1358.35 70 743-78 104 7372.80 1355.97 70 74 3-78 148 7594.46* 13 16.40 88 630-96 225 7925.47 1261.41 71 385-79310 7933.40 1260.15 71 385-79318 7953.88 1256.90 7 1 364-79 3 18 7958.26 1256.2 1 7 1 364-79 323 7966.28 1254.95 71 352-79318 8404.07 1 189.58 69 744-78 148 C1 17.5 9.6 9.4 16.5 7.2 5.8 13.1 22.5 8.1 15.6 14.5 14.4 16.5 45.1 13.4 17.6 11.4 6.1 4.5 14.5 2.3 12.5 6.7 3.7 9.8 7.4 3.8 12.3 4.0 5.2 3.1 5.3 - - - - 100 - c2 11.8 16.1 8.3 12.6 18.5 6.7 5.1 11.0 5.5 23.0 8.7 6.9 15.4 14.0 15.0 17.4 45.2 12.5 15.2 11.0 6.4 5.9 14.4 2.1 12.4 6.9 3.2 8.1 8.9 4.0 12.6 4.1 3.6 4.1 3.1 5.4 - 100 c3 13.9 18.2 13.1 10.9 23.0 7.8 5.9 10.7 22.5 7.5 15.8 15.1 13.9 15.3 44.3 13.2 17.1 10.1 6.0 6.0 15.4 12.4 7.4 9.7 8.5 12.2 4.6 - - - 100 - - - - 4.0 5.7JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY.JUNE 1993 VOL. 8 543 Table 4 -continued 8406.07 84 15.50 8427.27 8488.40 8505.9 1 8509.60 8566.9 1 8574.22 8596.97 1 189.29 1 187.96 1 186.30 1 1 77.75 1 1 75.33 1 174.82 1 166.96 1 165.96 1 162.88 69 710-78 116 69 689-78 104 69 689-78 1 16 69710-78 199 69 744-78 249 69 689-78 199 70 743-79 3 10 69 744-78 3 18 69 7 10-78 307 * Atomic oxygen lines (all others are atomic carbon lines).- - - 2.1 1.7 25.5 24.4 12.8 12.4 8.1 7.3 5.6 5.3 2.7 2.3 - - - 4.2 2.5 1.9 25.6 12.2 7.1 6.6 2.7 - Analyte Emission in an Argon-Nitrogen Plasma As the argon-nitrogen plasma offered greater heat transfer to the analyte p a ~ - t i c l e s ~ ~ - ~ ~ an experiment was performed to determine if there would be enhancement of the atomic carbon and oxygen IR emission from the alcohols under study using such a plasma. The argon-nitrogen plasma was experimentally difficult to ignite and required higher powers to keep it stable. It was found to run best with a mixture of 4% of nitrogen in the argon coolant gas flow and produced a plasma that was much thinner slightly taller and much less bright than a pure argon plasma.Impedance matching was more difficult with this plasma; many impedance adjustments were required before successful sample introduction codd be achieved. It was only possible to introduce methanol ethanol and propan- 1-01; the higher molecular mass alco- hols required powers in excess of those attainable wnth the 40.68 MHz ICP. The cooler argon-nitrogen plasma suppressed the back- ground emission and resulted in a 2-fold increase in sensitivity for atomic carbon and oxygen IR emission. Table 4 identifies 47 atomic carbon and oxygen emission lines observed using the argon-nitrogen plasma ;is the source. Conclusions The ICP-AES-FTIR spectrometry technique has been shown to be analytically useful for the identification of atomic IR emission lines in the alkan-1-01 series.In addition an almost 2-fold enhancement of the intensii y and number of atomic IR emission lines was achieved using an argon-nitrogen plasma as the atom source. Continued experimentation with a higher power ICP would no doubt broaden the scope of this study to other organic molecules of higher molecular mass and more complex matrices. References 1 Northway S. J. and Fry R. C. Appl. Spectrosc. 1980,34 332. 2 Northway S. J. Brown. R. M. Jr. and Fry R. C. Appi. Spectrosc. 1980 34 338. 3 Fry R. C. Northway S. J. Brown R. M. Jr. and Hughes S. K. .4nal. Chem. 1980 52 1716. 4 Brown R. M. Jr. and Fry R . C. Anal. C'hern. 198 I 53 532. 5 Hughes S. K. and Fry R. C . Anal. C'hern. 1981 53 1 1 I I . 6 Hughes S. K. Brown R. M. Jr. and Fry R. C. ,4ppl. Spectrosc. 1981 35 396. 7 Hughes S. K. and Fry R. C . Appl. Spectrosc. 1981 35 493. 8 Schleisman A. J. J. Fately W. G. and Fry R. C.. J . Phys. Chem. 1984 88 398. 9 Liu X. Shen X. and Zhang Y . K. Microchem. J.. 1988 37 306. 10 Blades M. W. and Hauser P. .4nal. C'him. Acta. 184 157 163. 1 1 Stubley E. A. and Horlick G. Appl. Spectrosc. 1984 38 162. 12 Griffiths P. R. and Haseth J. A.. Fourier Transform Infrared Spectroscopy Wiley New York 1986. 13 Norton R. H.. and Reinhard B.. J. Opt. Soc. A m . 1976 66 259. 14 Outred M. J. Phjx Chem. ReJ Data 1978 7 81-198. 15 Montaser A. Fassel V. A. and Zalewski J. '4ppI. Spectrosc. 1981 25 3. 16 Ishii I. and Montaser A. J. Anal. At. Spectrum. 1990 5 57. 17 Montaser A. Ishii I. Palmer B. A. and Layman. L. R. Spectrochirn. Acta Part B 1990 45 603. 18 Ishii I. Golightly D. W. and Montaser A J. Anal. ,4t. Spectrom. 1988 3 965. Paper 2/042 19C Received August 5 1992 Accepted December 16 1992

ISSN:0267-9477    

DOI:10.1039/JA9930800539

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 545 Signal Enhancement of Lead and Thallium in Inductively Coupled Plasma Atomic Emission Spectrometry Using On-line Anodic Stripping Voltammetry Jack R. Pretty* and Joseph A. Caruso-f Department of Chemistry University of Cincinnati Cincinnati OH 45221 -01 72 USA An on-line anodic stripping voltammetric (ASV) flow cell was employed for the preconcentration and signal enhancement of transition metals which exhibit relatively high detection limits in inductively coupled plasma atomic emission spectrometry but which are also responsive in ASV. Signal enhancements of over 60-fold for 20 ml samples of thallium(i) and lead(ii) were obtained in much less time than enhancements previously reported for cadmium using this method.The precision was acceptable and plots of peak heights and areas versus mass of analyte were linear for thallium and lead. Plots of enhancement factor versus sample volume were also linear for both metals. Keywords Inductively coupled plasma atomic emission spectrometry; anodic stripping voltamme try; signal enhancement; preconcentration; lead and thallium determination Since its introduction for analytical purposes inductively coupled plasma atomic emission spectrometry (ICP-AES) has proved to be an excellent method for the identification and determination of the majority of elements.'-3 This is partly due to the relative ease with which simultaneous multiple element analyses can be performed.' The ICP-AES detection limits are suitable for a wide variety of applica- tions.However for several elements of considerable analyt- ical interest the detection limits are still relatively owing to the lower emission intensities that they exhibit in the plasma. Among the most important of these elements are lead thallium bismuth tin and mercury. The detection limits for such analytes are often too poor to permit accurate determination of the levels present in biological and environmental system^.^ Superior detection limits can generally be achieved with techniques such as inductively coupled plasma mass spectrometry but the instrumental cost is considerably higher greater expertise is required on the part of operators and sample matrices can exert complex influences on the analyte ~ignal.~Jj-~ Various techniques have been used to enhance the analyte signals in ICP-AES including ultrasonic nebuliza- tion,2 hydride generation l o preconcentration on resin col- umns with e l ~ t i o n ~ J l precipitation with collection and redissolution,12 direct sample insertion' and electrother- mal vaporization (ETV).2J4 The degree of enhancement reported for these approaches ranges from under one to over three orders of magnitude. The greatest enhancements are predictably achieved with systems that circumvent the limitations of solution nebulization and that deliver analyte essentially quantitatively to the plasma (hydride ETV insertion).Additional enhancement is obtained when preliminary analyte preconcentration is combined with quantitative delivery.Il The degree to which matrices affect analytical results or convenience of operation varies con- siderably as does the over-all instrumental complexity and sample throughput of these procedures.Many of the transition metals that have relatively poor ICP-AES detection limits are responsive in anodic stripping voltammetry (ASV).4J5J6 Among the advantages of the latter technique is the potential for preconcentration of trace levels of analyte from a substantial sample volume into a much smaller volume (that of the working elec- trode).l5J6 Previous reports have described the use of ASV flow cells based on a working electrode of reticulated *Present address National Institute of Occupational Safety and ?To whom correspondence should be addressed. Health 4676 Columbia Parkway Cincinnati OH 45226 USA. vitreous carbon (RVC) placed on-line with the ICP for the purpose of preconcentrating electroactive analytes" while eliminating undesirable matrix constituents.18 The ASV cell designed in these laboratories has been successfully inter- faced with both ICP-AES and ICP-MS units.18-20 Accurate determination of several important metals in certified samples has been demonstrated using the ASV flow system with ICP-MS detection and large amounts of common matrices that are generally detrimental in ICP have been eliminated to allow improved analyte determination (al- though certain matrices can exert effects on the electroche- mical procedure).Early work also demonstrated the high degree of signal enhancement that could be attained using this ce11,18 as analyte from up to 20 ml of sample was deposited at the electrode then released into about 300 pl of solution for delivery to an ICP. Following improvements to the system manifold20 and tests with a variety of electrode surfaces and electrolytes the use of on-line ASV for improved detection of transition metals which have relatively poor sensitivity in ICP-AES was explored.Thallium and lead both of continuing interest owing to their toxicity and widespread occurrence in biological and environmental systems were chosen as the test analytes because of their well characterized behaviour in ASV.1sJ6i2' Experimental Instrumentation The ICP-AES instrument and microcomputer control sys- tem have been described previously.18 The operating conditions for the ICP-AES instrument are listed in Table 1.Details of the construction and operation of the ASV flow cell have been reported along with the method for plating of the RVC working electrode with mercury.'* The flow system manifold incorporated design modifications which are detailed in a separate report.20 A PAR 174 potentiostat (Princeton Applied Research Princeton NJ USA) was used to control the working electrode potential. Sample injection was performed using a PTFE valve (No. 5041 Rheodyne). Sample loop volumes were 1.00 ml (Rheodyne) and 2.04 and 4.95 ml (made in our laboratory from the same PTFE tubing as used in the ASV system manifold). The larger sizes were used to minimize the number of loop re-loadings necessary to deliver larger sample volumes. Short lengths of poly(tetrafluoroethy1ene) (PTFE) tubing built into the injection valve connected it to each end of the sample loop.The volume of liquid contained in this connecting tubing was found to be I00 pl,546 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 ~~ ~~~ Table 1 Operating conditions for ICP-AES instrument Radiofrequency power Forward 1.0 kW Reflected t 5 w Coolant 14.4 1 min-' Auxiliary 0.7 1 rnin-l Nebulizer 1.05 1 min-' Thallium 276.787 nm Lead 220.353 nm Gas flow rates Observed wavelengths Observation height Monochromator slits 15 mm above load coil 50 pm (entrance and exit) and this extra volume was included with each sample injection (e.g. four loadings of the 4.95 ml loop also introduced an additional 400 pl of sample into the ASV cell for a total of 20.2 ml). These extra volumes are reflected in the values quoted in the tables for volumes and mass of analyte.Exact volumes of loops and tubing were deter- mined by filling them with water and draining the contents into a vessel on a balance. Data Processing The software which provided monochromator control also allowed the calculation of peak area and height values. The integration time for each ICP-AES data point was 1 s. To assign a baseline intensity value for use in peak integration and further calculations 20 points on each side of a peak were averaged and the two means were averaged. Integra- tion limits were set by placing cursors on either side of the peak at points on the baseline which most nearly matched the average baseline value and which were as close as possible to the sides of the peak as observed on the system CRT display. The software then integrated the peak area and displayed the result.The highest peak data point can be found using another software command while the peak is still within the cursors. The mean baseline value is subtracted from the reported maximum to yield peak height. No units are provided with the displayed signal values. The software also includes a Savitsky-Golay smoothing program which was utilized in the present work. Values of 3 0 of the baseline (for use in determining analyte detection limits) were determined from 20 or more succes- sive baseline data points as mentioned above. Reagents Distilled de-ionized water supplied by a purification system (Barnstead Boston MA USA) was used for all solutions.Doubly distilled nitric acid (No. 621 GFS Chemicals Columbus OH USA) was used for the sample electrolyte and the supply electrolyte delivered to the ASV cell (both 0.1 moll-I) and for the bypass solution which was supplied to the ICP whenever the cell output was diverted (1% v/v). The mercury plating solution was 1 mmol 1 - I in mercury(xx) nitrate (99.999% Alfa Products Morton Thiokol Danvers MA USA) prepared in 0.1 mol 1-l ammonium nitrate of 99.999% purity (No. 25,606-4 Aldrich Milwaukee WI USA). Stock solutions (10 pg ml-l) of thallium and lead were prepared from standard solutions of each element (Fisher Scientific Cincinnati OH USA) and diluted as needed to prepare samples. All samples of less than 100 ng ml-I analyte concentration were prepared fresh daily.Results and Discussion Optimization of Flow Rate and Timing Previous experiments involving signal enhancement (of cadmium in 0.1 mol 1 - I NH,NO,) had used relatively low rates of sample delivery to the ASV cell in order to ensure high deposition efficiency. This resulted in considerable enhancement of the ICP-AES signal (up to a factor of 50 for 20 ml samples) but inordinately long times were required for larger samples.18 This was clearly unacceptable for routine application and one goal of the present work was to improve throughput for large sample volumes. Several experiments with analytes such as copper and arsenic in dilute nitric acid electrolyte19 have indicated that substan- tial enhancement could be achieved at higher flow rates than those used previously.Hence the reduction in analyte deposition efficiency at faster flow rates could be more than offset by the larger sample volume that could be run per unit time. It was presumed that the high conductivity of nitric acid solutions promoted efficient deposition at faster flow rates so 0.1 mol I-' HN03 was used as the sample and cell supply electrolytes in experiments with lead and thallium. Although both analytes may be deposited on the surface of a bare carbon working electrode a mercury- coated electrode was used as results for earlier experiments had indicated that this electrode often gave superior precision. I9vZ0 In all optimization and enhancement studies the ASV cell was flushed with clean electrolyte for 2 min after passage of the sample before the deposited analytes were stripped.Using 4 ml volumes of 1 pg ml-I thallium as test samples preliminary experiments were carried out to determine how rapid a sample delivery flow rate could be used without an inordinate decrease in the analyte deposition efficiency as demonstrated by the size of the ICP-AES signal peak. At each flow rate the effect of allowing different lengths of time for complete passage of the 4 ml sample volume through the ASV cell was examined. It was found that a sample delivery rate of 1.58 ml min-l gave peak heights equal to those achieved at slower flow rates and better than those achieved at faster flow rates. At this deposition flow rate allowing 4 min per sample (equivalent to 1 min ml-I) gave the best response per unit time; allowing less time per volume reduced the peak heights substantially whereas allowing more time produced very minor peak height gains which were offset by the lower sample throughput.The need to allow more time for sample passage than might seem apparent from the flow rate is due to broadening of the sample plug as it moves through the ASV flow system manifold. The flow and time parameters listed above were used for the remaining signal enhancement studies. Count- ing the time for reloading the 5 ml sample loop a 20 ml sample required 22 rnin to be run through the ASV cell. In the previously reported work with cadmium more than I h had been required for the same volume.18 Thallium Thallium was deposited at the electrode at - 1 .O V and a final potential of -0.2 V was used for stripping. Analyte levels ranging from 1 pg ml-L to 50 ng ml-l were run using various total volumes.Because of this a total sample mass of 1010 ng of T1 was delivered to the ASV cell in two ways as 10.1 ml of 100 ng ml-L T1 and as 20.2 ml of 50 ng ml-L TI. Therefore two data points were obtained for this mass of analyte. This provided a greater number of points for the generation of plots and allowed a comparison of the values obtained for the same mass of thallium using different sample volumes and total deposition times. For the pair of 1010 ng values the peak heights matched very well whereas the peak areas differed slightly (Table 2). AnJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 547 Table 2 Signal enhancement of thallium in ICP-AES.Peak heights (arbitrary units) and areas (s) RSD values (To) in paren- theses. Three replicate runs made for all samples; sample concen- trations in ng ml-L with volumes (ml) given in parentheses Sample 1* (4.28) 500 (4.28) 100 (5.05) 500 (10.10) 100 (10.10) 100 (20.20) 50 (20.20) Mass of TI/ ng 4280 2140 505 5050 1010 2020 1010 Mean peak height 1.67 (1.2) 0.84 (6.3) 0.20 (10) 2.05 (1.4) 0.44 (4.9) 0.83 (4.5) 0.44t Mean peak areals 1.16 (2.2) 0.550 (1.7) 0.146 (4.1) 1.5 (1.5) 0.316 (1.7) 0.615 (1.2) 0.336 (4.9) *Value given in pg ml-I. ?All replicate runs were calculated to have the same value. unexpected result was obtained for 2020 and 2 140 ng; the peak height values are approximately in the expected proportions (within the level of precision) but the peak area found for the lower mass of analyte is significantly greater than that found for the higher mass.The parameters for plots of peak height and area versus mass of thallium were as follows (units for height are arbitrary and area is given in seconds) peak height versus mass of analyte correlation coefficient =0.999 13 slope=0.3955 s pg-I T I intercept=0.0208; peak area versus mass of analyte correlation coefficient =0.996 14 slope=0.276 s pg-l T1 intercept=0.0235 s. Both plots are essentially linear the peak-height plot being the better of the two according to the correlation coefficients. The peak area anomaly mentioned above was probably partly respon- sible for this. The small non-zero intercept values may be the result of the uncertainty in assigning an exact baseline value owing to the presence of baseline noise (see under Data Processing).This can influence the integration of the smallest peaks (it is also one probable reason for the poorer RSD values associated with some of the smaller peaks). No RSD value is reported for the peak height of 20 ml of 50 ng ml-I T1 because the data processing software gave the same value for all three replicate runs an artifact of the digitization process. For the ICP-AES system used in this work the continu- ous nebulization detection limit of thallium (comparing 3 0 of the baseline to the height of the signal for 1 pg ml-I T I ) was 6.4 x lo2 ng ml-I at the emission wavelength of 276.79 nm. Running various volumes of supporting electrolyte blank through the ASV cell gave no thallium peak signals in ICP so the ASV-ICP detection limit could not be calcu- lated as three times the standard deviation of electrolyte blank peak area.Based on the height of the ASV-ICP peaks obtained by preconcentrating 20 ml of 50 ng ml-I T1 the detection limit calculated using 30of the baseline was 1 1 ng ml-I for a 20 ml sample volume. After application of a nine-point smoothing program the 30 detection limit was 233 ng ml-1 by continuous nebulization and 4.5 ng ml-I for 20 ml samples using the ASV cell. The use of larger sample volumes would further lower the ASV-ICP detection limits. Lead Lead was also deposited at - 1.0 V with a final stripping potential of -0.2 V. As with thallium running different volumes of supporting electrolyte blank through the ASV cell gave no visible analyte peak signals above the baseline in ICP-AES.Various volumes of lead sample at levels ranging from 1 pg ml-I to 50 ng ml-I of lead were delivered to the cell so that pairs of data points resulted for two masses of analyte ( 1 010 and 5050 ng). For both of these pairs the peak heights and areas matched very well (Table 3). Only one 20.2 ml volume of 100 ng ml-I of lead (2020 ng) was made owing to time constraints and the height and area values for this single run were omitted when the plots for peak height and area versus analyte mass were gener- ated. The RSD values for peak heights and areas ranged from low to moderately large with values for area often much lower than the corresponding values for height and lower analyte masses generally exhibiting larger RSD values (Table 3).As observed for thallium the plot of mass analyte versus peak area is less linear than that for mass versus peak height. The parameters for the plots were as follows (units for height are arbitrary and area is give in seconds) peak height versus mass of analyte (omitted 2.15 for 2020 ng Pb) correlation coefficient = 0.99979 slope= 0.962 s pug-' Pb intercept =O. 139; peak area versus mass of analyte (omitted 1.99 for 2020 ng Pb) correlation coefficient=0.99689 slope=0.701 s pg-I Pb intercept=0.286 s. Table 3 illustrates several peak area discrepancies. Whereas the peak heights for 2020 and 2525 ng of lead were in proportion the peak area of the lower mass was slightly higher than that of the other. The area value for 1100 ng lead was also unexpectedly high although the peak height was proportional (within the level of precision) to that of other analyte masses. This was the only example in which a sample volume of about 1 ml was used and the high value could indicate that the brief deposition time required for this small volume allowed a slightly greater proportion of lead to be retained at the electrode prior to stripping (less chance for the analyte to redissolve back into the dilute nitric acid supply electrolyte flowing through the cell) relative to that retained using larger samples and longer deposition times. However if this were so the mean value of peak height should also have been affected.Both peak height and peak area plots exhibited fairly high positive intercept values.The reasons for this are unknown as runs of the electrolyte blank produced no significant peak signals. One possibility is that the ASV cell was not functioning at maximum efficiency for the higher masses of analyte yielding disproportionately small peaks for larger masses of analyte and causing the slopes of the plots to be artificially lowered. This could occur if the amount of lead introduced into the cell exceeded the amount of lead that could dissolve in the mercury film although the large surface area available with the mercury-coated RVC work- ing electrode makes this appear unlikely. Thallium is more soluble in mercury than lead so exceeding the solubility limit would be less likely for the former.15 However the good linearity of the peak height curve for lead does not indicate a disproportionate deposition efficiency for higher masses of analyte.For the ICP-AES system used in this work the detec- tion limit of lead with continuous nebulization (deter- mined as above for thallium using 30 of the baseline) was 4.9 x lo2 ng ml-I without smoothing. The 30 detection limit based on the height of ASV-ICP peaks produced from 20 ml of 50 ng ml-I Pb was 7.9 ng ml-I. After nine- point data smoothing the respective values were 221 and 3.5 ng ml-I. Signal Enhancement Capability The performance of the ASV-ICP system in signal enhance- ment is summarized in Table 4 by comparing the continu- ous nebulization signal heights for the indicated analyte levels with peak heights obtained with various volumes of the same analyte levels using the ASV flow system.The enhancement factor is defined as the ratio of the height of the ASV-ICP peak signal for a given volume of analyte to the height of the ICP-AES continuous nebulization signal produced by the same analyte level. For analyte levels that548 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Table 3 Signal enhancement of lead in ICP-AES. Peak heights (arbitrary units) and areas (s) RSD values (O/O) in parentheses. Three replicate runs made for all samples except as indicated; sample concentrations ic ng ml-I with volumes (ml) given in parentheses Sample Mass of Pbhg Mean peak height Mean peak areds 1* (1.10) 1* (5.05) 500 (5.05) 500 (10.10) 100 (10.10) 50 (10.10) 50 (20.20) (2 runs) 100 (20.20) (1 run) *Values given in pg m1-I 1100 5050 2525 5050 1010 505 1010 2020 1.15 (7.1) 5.01 (2.0) 2.55 (1.6) 41.99 (2.7) 1.16 (4.6) 0.59 (5.6) 1.14 1.17 2.15 1.26 (0.75) 3.84 (0.96) 1.93 (0.88) 3.84 (0.64) 0.977 (0.55) 0.527 (3.80) 0.967 1.07 1.99 were above the continuous nebulization detection limits direct calculation of enhancement factors was possible.For levels that were below the detection limit the values listed in Table 4 for continuous nebulization signal height were calculated based on the heights of signals observed for detectable levels and these calculated height values were used in determining the enhancement factors. If the enhancement factor is plotted against sample volume the slope of the plot can be taken as the signal enhancement capability of the ASV flow system defined as enhancement factor per millilitre of sample.For both thallium and lead the resulting plots were reasonably linear although both also exhibit small negative intercepts for unknown reasons (Table 4). The slopes of the plots were 3.31 for both analytes while the average enhancement factor was 65.6 for samples of just over 20 ml volume attesting to the efficiency of the ASV flow cell in the signal enhancement of these metals. The performance of the systems is illustrated by Figs. 1 and 2. Both figures present continuous nebulization signals for levels of analyte that could be observed without the ASV cell alongside ASV-ICP peaks resulting from 20 ml vol- umes of analytes at levels that were well below the continuous nebulization detection limits of the available ICP-AES unit.Fig. 1 shows two replicate peaks for 20 ml of 50 ng ml-l of thallium and Fig. 2 shows a single run for 20 ml of 100 ng mi-l of lead. Although no data smoothing was used on these displays the ASV-ICP peaks are visible well above the baseline noise. As indicated numerically in Table 4 the peak heights are well above the continuous nebuliza- tion signals for 1 ,ug ml-l of the respective analytes. Conclusions The ICP-AES unit used in this work could not provide fully optimized analyte signals. Further the available photomul- tiplier tube was unresponsive at the most sensitive thallium emission wavelength (1 90.864 nm) (ref. 4) and a less intense line was monitored instead. For these reasons the detection limits quoted for the analyte metals with continuous nebulization do not match those which can be obtained under more carefully controlled conditions. However the degree of signal enhancement obtained for a given sample volume with the on-line ASV cell is not affected by these factors and interfacing the system with a more sensitive ICP unit would result in proportionally better ASV-ICP Table 4 Signal enhancement capability of ASV flow system ICP-AES signal heights obtained by continuous nebulization compared with heights of peaks obtained by preconcentration of indicated sample volumes.Continuous nebulization values for analyte levels below detection limits are calculated as described in the text. Signal heights in arbitrary units Anal yte Continuous nebulization Concentration signal height Thallium 10 pg ml-I 1 pg m1-l 500 ng ml-I 100 ng m1-l 50 ng ml-I Lead 10 pg m1-I 1 pg ml-' 1.32 0.131 0.066 (calc.) 0.013 (calc.) 0.0066 (calc.) 3.43 0.342 500 ng ml-* 0.17 100 ng m1-I 50 ng ml-' 0.034 (calc.) 0.017 (calc.) Sample volume in ASV-ICP/ml 4.28 4.28 10.10 5.05 10.10 20.20 20.20 1.10 5.05 5.05 10.10 10.10 20.20 10.10 20.20 ASV-ICP Peak height 1.67 0.84 2.05 0.20 0.44 0.83 0.44 1.15 5.0 1 2.55 4.99 1.16 2.15 0.59 1.16 Enhancement factor versus sample volume Thallium correlation coefficient=0.99854 slope:= 3.3 1 ml-l intercept= - 1.31 Lead correlation coefficient=0.99549 slope= 3.3 1 ml-l intercept= - 1.01 Enhancement factor 12.7 13 31 15 34 64 67 3.37 14.6 15.0 29.4 34.1 63.2 35 68.2JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 549 4 j/ 0 990.00 Time/s 1980.00 Fig.1 Peaks resulting from preconcentration of 20 ml of 50 ng ml-I T 1 in ASV-ICP-AES. Continuous nebulization 'signals shown for comparison .4 500 ng ml-I T1; B 1 pg ml-' T 1 ; and C. 10 pg ml-' T 1. Display has not been smoothed. Further details in text detection limits than those reported here. The ASV cell itself was not designed specifically for maximum signal enhancement efficiency. Appropriate design modifications or the use of a working electrode with a higher surface area (allowing more efficient analyte deposition at faster flow rates) could further improve the degree of enhancement and sample throughput. The available ICP-AES detection system allowed only one element at a time to be monitored and accordingly only single-analyte test solutions were utilized in this work.However the deposition potential of - 1.0 V and final stripping potential of -0.2 V were suitable for both lead and thallium. With a suitable ICP-AES unit as detector simultaneous element determinations could be performed with the ASV flow system by depositing both elements at once from the sample then stripping them back into solution together. A similar approach would hold for most of the transition metals that deposit efficiently at a mcrcury film electrode (zinc cadmium copper indium etc.) so long as suitable deposition and stripping potentials are used. In determining elements by ASV alone the simultaneous deposition of metals with similar crystalline structures can lead to the formation of alloy-like intermetallic states within mercury film electrodes.16 These generally occur 0 350.00 Time/s 700.00 Fig.2 Peak resulting from preconcentration of 20 ml of 100 ng ml-I Pb in ASV-ICP-AES. Continuous nebulization signals shown for comparison A 500 ng ml-I Pb; B 1 p g ml-' Pb; and C 10 pug ml-' Pb. Display has not been smoothed. Further details in text when one of the metals is present at a high concentration within the film and in a much higher proportion than the others. The stripping current peaks of intermetallic states can lie at unexpected potential values and analyte determi- nation by measurement of these peak heights and areas is inaccurate. In the present application intermetallic forma- tion is theoretically of less importance; as long as all deposited metals are completely stripped from the electrode at the end of the potential scan for delivery to the ICP detector undistorted element-specific determinations are possible.A s stated previously the successful determination of selected analytes in certified samples has been demon- ~ t r a t e d . ' ~ - * ~ Several experimental factors in this work (such as use of a mercury-coated electrode and dilute nitric acid electrolyte) are similar to the conditions successfully em- ployed for the determination of copper in a urine matrix.20 It is likely that lead and thallium will prove equally amenable to determination by ASV-ICP although this has not yet been tested with certified samples. When sufficient sample volumes are available the system can provide a substantial reduction in detection limits while simultane- ously eliminating detrimental sample matrix effects. The efficiency of the on-line ASV-ICP system for signal enhancement of lead and thallium has been demonstrated.The inconsistencies observed for analyte peak area values are of some concern particularly given the generally superior behaviour of the corresponding peak height data. The reasons for these variations have not yet been eluci- dated. Possible considerations include the accuracy with which peaks have been integrated and the question of whether data points were taken often enough to trace adequately the most rapid changes of the ICP-AES peak signal. Alternati\,ely undetected shifts in various system parameters over the course of extended experimental sessions could affect the final results.The ASV flow system is still a prototype in some respects and imperfect perform- ance is not unexpected. Experiments similar to those described here have been performed with tin and mercury but although substantial enhancements were achieved these analytes require the use of more specialized electrode or electrolyte systems for the best anodic stripping response and the results to datc lie outside the scope of this paper. Although these signal enhancement studies have focused on elements with high ICP-AES detection limits the approach is equally valid for the signal enhancement of electroactive analytes in ICP- MS although the magnitude of the blank peak signals produced by trace electrolyte contamination has so far limited exploration of this option.The potentiostat and x-y recorder were provided by the National Forensic Center of the Food and Drug Adminis- tration Cincinnati OH. J.R.P. notes the suggestion of Karen Wolnick of the National Forensic Center that signal enhancement of thallium merited examination. The assis- tance of Robert Voorheis Paul MacKensie William Brauntz Arthur Case and Douglas Hurd of the electronics and machine shops of the Department of Chemistry University of Cincinnati and the advice contributed by Elmo A. Blubaugh are also gratefully acknowledged. Our particular thanks are due to Timothy Davidson who conceived the project and participated in the planning stages prior to his departure from the laboratory. References I Fassel V. A. and Kniseley R. N. Anal. C'hem. 1974 46 1 I IOA.2 Barnes R. M. CRC Crit. Rev. Anal. C'hein.. 1978 7 203. 3 Olesik J. W. Anal. C'hcwi. 1991 63 12A.5 50 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 4 Winge R. K. Peterson V. J. and Fassel V. A. Appl. Spectrosc. 1979 33 206. 5 Wunsch G. Knobloch S. Luck J. and Blodorn W. Spectrochim. Acta Part B 1992 47 199. 6 Hieftje G. M. and Vickers G. H. Anal. Chim. Acta 1989 216 1. 7 Tan S. and Horlick G. J. Anal. At. Spectrom. 1987 2 745. 8 Tan S. and Horlick G. Appl. Spectrosc. 1986 40 445. 9 Vaughan M. A. and Horlick G. Appl. Spectrosc. 1986 40 434. 10 Hwang J. D. Huxley H. P. Diomiguardi J. P. and Vaughn W. J. Appl. Spectrosc. 1990 44 491. 1 1 Moss P. and Salin E. D. Appl. Spectrosc. 1991 45 1581. 12 Akagi T. and Haraguchi H. Anal. Chem. 1990 62 81. 13 Abdullah M. Fuwa K. and Haraguchi H. Appl. Spectrosc. 1987 41 715. 14 Nisamaneepong W. Caruso J. A. and Ng K. C. J. Chromatogr. Sci. 1985 23 465. 15 Vydra F. Stulik K. and Julakova E. Electrochemical Stripping Analysis Ellis Horwood Chichester 1976. 16 Wang J. Stripping Analysis Principles Instrumentation and Applications VCH Deerfield Beach FL 1984. 17 Ogaram D. A. and Snook R. D. Analyst 1984 109 1597. 18 Pretty J. R. Evans E. H. Blubaugh E. A. Shen W.4 Caruso J. A. and Davidson T. M. J. Anal. At. Spectrom. 1990 5 437. 19 Pretty J. R. PhD Dissertation University of Cincinnati I99 1 . 20 Pretty J. R. Blubaugh E. A. Evans E. H. Caruso J. A. and Davidson T. M. unpublished work. 21 Hoeflich L. K. Gale R. J. and Good M. L. Anal. Chem. 1983 55 1591. Paper 2/00 75 7F Received February 12 I992 Accepted January I I993

ISSN:0267-9477    

DOI:10.1039/JA9930800545

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 5 5 1 Determination of Uranium and Thorium in Basalts and Uranium in Aqueous Solution by lnductively Coupled Plasma Mass Spectrometry Elizabeth H. Bailey Anthony J. Kemp and K. Vala Ragnarsdottir Department of Geology University of Bristol Wills Memorial Building Queens Road Bristol BS8 1 RJ UK Methods are described for the simultaneous determination of uranium and thorium in basaltic rocks at concentrations from mg kg-1 (ppm) to ,ug kg-1 (ppb) levels and for uranium in solution at femtogram levels. The technique for basalts does not require preconcentration and samples are taken into solution using a conventional rapid acid digestion method. The results obtained agree in general within 10% of published values for eight basalt reference materials.In addition optimum conditions with respect to acidic media and storage vessel were ascertained as 5% HNO and high-density polyethylene respectively. The method is applicable to a wide range of silicate rocks and other geological materials. For the determination of uranium at sub-ng ml - concentrations such as occur in natural waters sample introduction by electrothermal vaporization was investigated using several matrices. Complexation of uranium with ethylenediaminetetraacetic acid (EDTA) prior to analysis proved successful at levels down to 0.5 fg mi-I. Routine determination of such low concentrations is thus feasible. Keywords lnductively coupled plasma mass spectrometry; uranium; thorium; basalt; electrothermal vaporiza- tion Uranium and thorium are geologically incompatible ele- ments (i.e.they become enriched in the melt rather than the solid phase) that become concentrated in evolved rocks such as granites. Their concentrations in basaltic rocks are low and traditional techniques employed for their determi- nation are generally time consuming labour intensive and require large sample sizes. The techniques include fluorime- try and extractive spectrophotometry,' delayed neutron assay,2 non-destructive gamma-ray spectrometry3 and in- strumental neutron activation a n a l y ~ i s . ~ Similarly. the concentrations of the elements in natural waters are extremely low and often below the detection limits of the technique being e m p l ~ y e d . ~ . ~ The determination of ura- nium and thorium at low levels is required in the aerospace and nuclear industries and also in environmental samples where these and other elements are monitored r ~ u t i n e l y .~ . ~ Inductively coupled plasma mass spectrometry (ICP-MS) offers a rapid method for the determination of uranium and thorium in a variety of sample types (from aqueous solutions to silicate rocks) and over a wide range of c o n c e n t r a t i o n ~ ~ - ~ ~ (from jig ml-l to fg ml-I levels). The feasibility of electrothermal vaporization (ETV)I3 for the determination of uranium and plutonium at femtogram levels has been demonstrated previously. 1 4 1 5 The first part of this study was aimed at the simultaneous determination of uranium and thorium in basaltic rocks and waters with emphasis on the effects of acidic media and storage vessel.The acidic media considered were S% vlv nitric acid 5% vlv hydrochloric acid 5% vlv aquii rcgia (hydrochloric acid-nitric acid 5+2) and a 5% vlv 5 + 2 nitric acid-hydrochloric acid mixture for the analysis of basaltic rocks. The types of sample storage vessels consi- dered were high-density polyethylene (HDPE) polyethyl- ene poly(propy1ene) and borosilicate glass. For the lower levels of uranium present in waters 5% vlv nitric acid 5% vlv aqua regia 0.0 1 mol 1 - l ethylenediaminetetraacetic acid (EDTA) and 0.01 mol I-' citric acid were investigated. The use of organic acids reflects attempts to retain the uranium in solution by complexation and to minimize losses to the vessel walls etc. Uranium(v1) is much more soluble than uranium(1v) in natural waters.Calculations in progress in our laboratory using the density model of Anderson et al.I6 indicate that UOz is 14 orders of magnitude less soluble than U 0 3 at low temperatures and pressures and a pH of 7. Over the past three decades several attempts have been made to measure uranium(1v) solubility a c ~ u r a t e l y . ' ~ ~ ~ ~ However owing to problems of oxidation contamination and other factors the results are questionable and often conflicting; at best the solubility is known at very low pH (1-4) (Fig. 1). Poor solubility data exist over the pH range 4-8 typical for natural waters with values ranging over two orders of magnitude. This lack of good data is principally due to the limits of detection (LODs) for measuring uranium by neutron activation analysis and by normal nebulization ICP-MS (approximately 1 0-lo rnol 1 - I ) .Calculations using the best currently available thermodynamic datalYqZ0 in the density mode116 indicate that the solubility of uranium(1v) might be lower than 10-l' mol I-'. Electrothermal vapori- ~ a t i o n ~ l - ~ ~ however offers a possible solutioil to this problem with reported absolute detection limits in the region of 1 fg. The second part of this study was focused on the determination of femtogram levels of uranium. Electrother- mal vaporization offers many advantages over solution n e b ~ l i z a t i o n ~ ~ including greater sensitivity smaller sample sizes and the ability to analyse solutions with high salt and organic contents. The lower concentrations of oxygen and hydrogen in the p l a ~ m a ~ ~ ~ ~ avoids the formation of large amounts of oxide and hydroxide and the use of a Freon gas mixed with the argon carrier gas minimizes carbide format ion.The method developed here is to be used to determine the solubility of uranium(1v) under hydrothermal conditions - 1 1 1 2 3 4 5 6 7 PH I O x .i 8 9 10 11 Fig. 1 200; and 0 300 "C Experimental uranium solubility values (ref. 5 ) x . 100; .,552 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 6x105 5x105 4x105 - LO CL :z 3x105 :3 .# 2x105 1x10~ 0 - [250-2000 bar (1 bar= 1 x lo5 Pa) 250-500 "C] in experi- ments currently in progress. The experiments are per- formed with controlled oxygen fugacity foZ (using a hydro- gen membrane) to ascertain that the solubility of U02 and not that of a more oxidized species is being investigated.The method of analysis was designed to minimize deterior- ation of the solutions in the time interval between sampling and analysis. To do this organic acids were investigated as good complexing ligands for the storage of uranium samples alongside more traditional acidic media and for their behaviour in the furnace. An investigation was undertaken to determine a suitable ETV graphite furnace heating profile for such analysis and to optimize the ICP-MS instrumental conditions including the injection volume and Freon and nebulizer flow rates. - - - - - - Experimental Spectrometer Analyses were performed on a VG Elemental PlasmaQuad PQ I1 instrument after optimization for uranium and thorium.A Gilson Minipuls 3.0 peristaltic pump was used to introduce samples into the plasma during normal nebulization. Sample introduction for solutions containing femtogram amounts of uranium was effected via a VG Elemental Mk 2 ETV unit. The Freon gas employed in ETV analysis was Freon 23 (trifluoromethane) available from Distillers Special Gases. It was controlled by the mass flow controller an integral part of the ETV unit. Materials and Reagents A list of reagents used in this investigation together with their quality is given in Table 1. Sample Preparation To investigate the effects of the storage vessel and acidic media on the determination of uranium and thorium solutions containing 100 ng ml-l of the elements in each acid type were prepared in each of the four different vessels.All preparations were carried out under laminar flow in a clean environment. Bismuth ( 100 ng ml-I) was added as an internal standard to each. The solutions were then analysed on the PlasmaQuad instrument. The results indicated that the solutions stored in HDPE vessels with a 5% nitric acid medium consistently gave up to a factor of two higher counts for both uranium and thorium (Figs. 2 and 3). Calibration standards in 5% v/v nitric acid were prepared to contain 10 25 50 and 100 ng ml-I of the elements and 100 ng ml-l of bismuth as an internal standard from Aldrich stock 1000 ng ml-l atomic absorption standard solutions. Calibration graphs for both uranium and thorium are known to be linear over several orders of magnitude and this was confirmed for up to 100 ng ml-I.Optimum instrumental operating conditions are given in Table 2. A long acquisition time (131 s) was chosen to give the best possible sensitivity. 6x105 I High density polyethylene 1 Polyethylene 0 Poly( p r opy lene) 5x1 O5 4x105 2 3x105 2 6 2x105 1x10~ I 0 5% HNO 5% HCI 5% aqua regia 5% 5+2 HN0,-HCI Fig. 2 Effect of acid matrix and vessel on thorium storage I High density polyethylene Polyethylene El Poly(propy1ene) 0 Borosilicate glass 5% HNO 5% HCI 5% aqua regia 5% 5+2 HN0,-HCI Fig. 3 Effect of acid matrix and vessel on uranium storage Bismuth (209Bi) was used as an internal standard for this analysis. Its purpose was to compensate for non-spectros- copic interferences arising from matrix elements in the sample solution.It has been shown that elements of similar masses and ionization potential give similar signal re- sponse~.~' Hence bismuth was considered to be a suitable internal standard for uranium and thorium determination. Use of *IsIn as an internal standard has been shown to give erroneous results for uranium and thorium.25 Solutions containing reference materials2* (Table 3) were prepared in the following manner (1) the powdered samples were first dried for a minimum of 12 h at 110 "C; (2) 200 mg of each sample was weighed into 50 ml PTFE beakers and moistened with 1-2 ml of 18 Mi2 water; (3) 3 ml of 40% hydrofluoric acid and 2 ml of concentrated nitric acid were added. The beakers were placed on a hot-plate at 210 "C and the solution was evaporated just to dryness before being removed from the hot-plate and allowed to Table 2 Instrumental conditions for the determination of ura- nium and thorium in basaltic rocks Table 1 Reagents used in the investigation Reagent Quality Nitric acid Aristar Hydrofluoric acid (40%) Aristar Hydrochloric acid Aristar Water 18 M a quality Bismuth 1000 mg I - l * Thorium 1000 mg l-l* Uranium 1000 mg I-'* * AAS standard solution.Mass range/u Peristaltic pump rate/ml rnin-l Nebulizer gas flow rate/l min-' Coolant gas flow rate/l min-l Auxiliary gas flow rate/l min-l Forward powerIW Dwell timelps Number of sweeps Acquisition time/s Nebulizer Internal standard Acidic medium Storage vessel 200-245 0.8 0.7 14 0.75 1350 320 100 131 Meinhard 100 ng ml-1 of Bi 5% HNO 125 ml narrow-necked HDPEJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 553 Table 3 International standards used Standard Source BCR- 1 BHVO- 1 BR BE-N JB- 1 JB- 1 a JB-2 JB-3 US Geological Survey (USGS) Reston VA USA USGS Centre de Recherches Petrographiques et Geochimiqnes (CRPG) Vandoeuvre France CRPG Geological Survey of Japan (GSJ) Ibaraki Japan GSJ GSJ GSJ Table 4 Instrumental operating condition for ETV Gas flow rate Coolant/l min-I 14 Auxilaryll min-' 0.75 Carriedml min-' Freonhl min-I 0.6-0.85 (optimum = 0.75) 0.2-0.6 (optimum = 0.4) Nebulizer Meinhard Forward r.f. power/W 1350 Sample cone Nickel Aperture diametedmrn 1 .O Skimmer cone Nickel Aperture diametedmrn 0.9 Reflected power/W <5 Injection volume/pl 10-50 cool; (4) concentrated nitric acid ( 5 ml) was added to each beaker the contents of which were then topped up to 40 ml with 18 Mi2 water; (5) the beakers were returned to the hot- plate until the cake had dissolved and then were subse- quently cooled; (6) the contents were transferred into Nalgene calibrated flasks and diluted to 100 ml with I8 Mi2 water; and (7) the solutions were finally transferred into HDPE storage containers.Using these solutions a wide range of elements found in basaltic rocks can be deter- mined. The solutions for uranium and thorium determination were prepared from these stock solutions in the following manner aliquots (5 ml) of these solutions were pipetted into 10 ml calibrated flasks 100 pl of 10 mg 1-1 of bismuth solution were added to each and the contents were diluted to volume with 5% v/v nitric acid and the solutions were then transferred into HDPE vessels for immediate analysis. Acid digestion was chosen in this study because other methods of sample digestion such as lithium metaborate fusion lead to high total dissolved solids and hence analytical problems.Acid digestion lowers the amount of total dissolved solids by volatilizing Si as SiF during the first heating stage. For a more detailed comparison of dissolution techniques see refs. 1 1 and 12. Electrothermal Vaporization Solutions were prepared to contain femtogram amounts of uranium from 1000 mg I-' Aldrich stock solutions using micropipettes and then stored in HDPE bottles. Prior to analysis of the solutions the IGP was optimized on 209Bi by adjusting the torch position nebulizer gas flow rate and lens potential using a wet plasma.This was necessary because attempts to optimize on argon or bismuth peaks using a dry plasma were in general difficult and unsuccessful because a stable signal for bismuth could not always be obtained and maintained for a sufficient length of time. A similar problem has been observed when attempts have been made to optimize on the mercury peak using a dry ~1asma.l~ Despite previous demonstrationz9 that water has a signifi- cant effect on the analytical characteristics of an argon Table 5 ETV furnace temperature profile Stage Temperature/"C Time/s Process 1 50 2 Drying 2 150 60 Drying 55 Ashing 3 525 4 2600 t l Vaporization 5 Pause 15 Cleaning plasma optimization using a wet plasma was sufficient to achieve LODs at the femtogram level for uranium. Bismuth rather than uranium was used for optimization to avoid any possibility of memory effects when subsequently determin- ing uranium at femtogram levels. The instrument operating conditions used are listed in Table 4.All samples were measured in the single-ion monitoring mode. A furnace heating profile was first developed. This process is typically one of trial and error adjusting the ashing and vaporization temperatures to ensure that the element is not vaporized prior to analysis that all of the element is vaporized during analysis and to obtain good peak shapes. Temperatures can often be estimated by studying the physical and chemical properties of the element concerned. For example uranium has a low volatility so the vaporization temperatures need to be high whereas mercury requires lower ashing temperatures to minimize losses prior to analysis due to its higher volatility.Typical features of such a profile for uranium must therefore include long drying and ashing stages followed by rapid vaporization to a temperature of at least 2500 "C. The profile depicted in Table 5 was developed based on these element properties and was used throughout the study. Low-level uranium solutions readily deteriorate and are therefore difficult to analyse. Consequently a range of matrices were studied both for their ability to retain uranium in solution and for their performance during heating in the furnace. As nitric acid is recommended as the best matrix with normal nebulization in ICP-MS this acid was the first choice for investigation. Analyses of aqua regia 0.01 mol I-' EDTA and 0.01 rnol I-* citric acid matrices were also carried out.The choice of the organic acids as complexing ligands reflects the need to find a medium able to retain low concentrations of uranium in solution for periods of hours rather than minutes. Oxygen- containing organics (e.g. diethyl ether isobutyl methyl ketone and tributyl phosphate) have been shown to extract uranium and thorium successfully from aqueous solu- t i o n ~ ~ ~ - ~ * of nitric and hydrochloric acid indicating that strong complexation is occurring in the organic media. The injection volume and Freon and carrier gas flow rates were initially varied individually to give the highest count rate then a series of varying combinations were investi- gated.Injection volume Typical injection volumes for ETV analysis are in the range 10-100 pl. For a particular temperature profile i t is expected that a linear relationship with detection will exist until the volume injected is too great to be dried and ashed in the alloted time and a plateau is reached. Freon jlow rate Addition of a Freon gas such as trifluoromethane to the gas flow through the ETV unit is necessary for elements with low volatility in order to improve transport of the analyte to the plasma. An improvement of one to two orders of magnitude in LOD has been demonstrated using Freon 23 (trifl~oromethane).~3 The Freon decomposes at the high554 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 Table 6 Data obtained from repeated injection of a 1 ng ml-I of uranium standard without blanks between samples Peak height Sample No. Sample type absorbance 1 Blank 4 2 Standard 670 3 Standard 653 4 Standard 674 5 Standard 696 6 Standard 684 7 Standard 603 8 Blank 4 Mean of Standards 663.33 temperatures it encounters in the furnace to liberate fluorine radicals. These radicals then react with uranium to form fluoride complexes of much higher volatility than that of the uranium alone. The concentration of Freon required to react with all the analyte present in the tube will therefore be a factor of the volume of solution injected and the concentration of the analyte in this solution. Therefore high Freon flow rates combined with small solution volumes should provide the best detection limits.Carrier gas flow rate Varying the flow rate of the carrier gas through the furnace alters the transport properties of the analyte to the plasma. The shorter the distance between the plasma and furnace the more efficient the transport of the analyte to the plasma is expected to be. To this end the instrument was modified to give the shortest (aproximately 700 mm) most direct route between the plasma and the furnace. Results by some workers indicate that the sensitivity increases with increas- ing carrier gas flow rate until a plateau is reached.31 Other workers observed a peak in flow rate.34 All measurements were repeated a minimum of three times and usually five times. At first the pattern of sample blank sample blank etc. was repeated but if sufficient cleaning time is included in the furance heating profile it was found to be unnecessary to analyse a blank after every sample.Holding the temperature at its maximum value for 15 s appears sufficient to clean the graphite tube (Table 6) and blanks need only be analysed after every third or even fifth sample. Results Basaltic Samples The results obtained for thorium in the basaltic rocks (using conventional solution nebulization) are in very good agree- ment with the recommended values with the exception of JB-2 and BHVO-1 (Table 7). A repeat of the procedure for these standards again gave a value with large error. It is believed that for JB-2 this might be due to the lower concentration of thorium compared with that in the other reference materials. It must also be remembered that recommended values are the average of values determined by a range of techniques.Adsorption of thorium on the glassware and sample cone in the ICP system initially gave difficulties due to memory effects but aspiration of 5% v/v nitric acid for 2 min between samples was found to overcome this problem. The uranium results all fall within the 10% accepted error difference. The higher percentage difference obtained for the analyses for uranium compared with those for thorium are thought to be a result of the lower concentrations of the former. No memory effect was observed for uranium. Electrothermal Vaporization Preliminary ETV work indicated the need for rigorous cleaning of the unit in addition to ICP glassware and cones prior to each set of analyses and often in the course of a day's continuous use in order to avoid a decrease in sensitivity due to the build-up of carbon in the system. Tubing connecting the ETV unit to the plasma should be replaced on a daily basis to avoid contamination especially when analysing solutions with an organic matrix as a layer of carbon builds up on the inner surface.The condition of the graphite tube is also important and it must be replaced when signs of pitting are evident particularly on the underside of the tube (i.e. 180" from the aperture). Deterioration of the graphite tube is usually indicated when the peak shapes become ragged and poorly defined. The LOD was calculated from the recorded data accord- ing to the following equation:35 ItJ 4 3 LOD (ng ml-l) = 3 x RSD x - x c where the subscripts a and b refer to analyte and blank respectively RSD is the relative standard deviation I is the integrated peak area and c is the concentration of the analyte (ng ml-I).Absolute limits of detection can then be calculated volume injected 011) 1000 Absolute LOD (ng) = x LOD Levels of uranium in solutions prepared to contain 0.01 ng ml-1 or less of uranium in 5% v/v nitric acid or aqua regia were found to be unmeasurable when analysed within a few hours of initial preparation probably owing to losses to the vessel walls. However uranium in 0.01 mol 1-I EDTA and 0.01 mol 1-1 citric acid solutions prepared to contain 100 fg ml-' of uranium was still detectable up to 2 Table 7 Results obtained for the basaltic samples and comparison with recommended values Thorium/mg kg-' Uranium/mg kg-' This Suggested Difference This Suggested Difference - Standard study value (To) study value (O/O) JB- 1 JB-la JB-2 JB-3 BHVO- 1 BCR- 1 BR BE-N 9.99 8.63 0.26 1.32 1.25 5.45 10.88 10.91 9.20 8.80 0.33 1.30 1.08 5.98 11.00 11.00 8.59 1.93 21.21 1.54 15.74 8.86 1.09 0.82 1.83 1.58 0.17 0.5 1 0.44 1.67 2.63 2.59 1.70 1.60 0.16 0.46 0.42 1.75 2.50 2.40 7.65 1.25 6.25 10.87 4.67 4.57 5.20 7.92JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 555 2800 2600 v) 8 2400 - 5 = 2200 Table 8 Effect of sample matrix on absolute LODs - - - - Matrix Absolute LOD*/fg 5% Nitric acid 857 5% Aqua regia 409 0.01 mol 1 - I EDTA 33.7 * 30 pl injection of 1 ng ml-I uranium solution. Each value represents the average of five injections. weeks after initial preparation.Absolute LODs for a 1 ng ml-I uranium solution were an order of magnitude better when EDTA was used as a complexing agent than with nitric acid and aqua regia (Table 8). The ashing of EDTA in the graphite furnace creates substantial amounts of carbon. This carbon is deposited on 4000 h 3000 t A v) 0 2 2000 1000 0 10 20 30 40 50 Injection volume/p I Fig. 4 Variation in count rate with injection volumes; Freon flow rate 0.4 ml min-I 200 0.2 0.3 0.4 0.5 Freon flow rate/ml min-' Fig. 5 Effect of Freon flow rate on area counts per second (ACPS). Injection of 30 pI of 100 fg m1-I uranium solution 3000 I 1 2oooY 1800 I I I 0.60 0.70 0.80 Carrier gas flow rate/ I min-' Fig. 6 Effect of carrier gas flow rate on area counts per second (ACPS). Injection of 30 ,uI of 100 fg ml-I uranium solution and Freon flow rate 0.4 ml min-' the sample and skimmer cones during a typical day's analysis and can cause a severe decrease in sensitivity.Re- cleaning of cones can become necessary during a long series of analyses when using organic matrices but is of course a factor of the volume of solution injected for each analysis. Investigation of volumes in the range 10-50 pl did not indicate one optimum volume. Injection of a 100 fg ml-I solution in the range 10-50 pl exhibited as expected approximately linear behaviour. A plateau in LODs was not observed in this range of volumes indicating that all the sample is dried and ashed satisfactorily in the time allotted in the profile (Fig.4). Injections of 30 pl of solutions containing 100 fg ml-1 of uranium with variable Freon flow rates indicated that 0.4 ml min-l was the optimum value for the chosen temperature profile (Fig. 5). A clear optimum carrier gas flow rate of 0.75 1 min-l was observed for this temperature profile (Fig. 6). Discussion The rapid determination of uranium and thorium in basaltic rocks is possible using the solution nebulization method outlined here. The HDPE vessels proved to instigate less adsorption of uranium and thorium onto their walls than the other vessels investigated. Further 5% v/v nitric acid gave the best results when count rates are considered. The speed and sensitivity of the method make it advantageous over previously more cumbersome and less sensitive methods.The method is not applicable to more evolved igneous rocks as the digestion technique employed will not adequately dissolve minerals such as zircons which often contain high concentrations of both uranium and thorium. Using EDTA as the sample matrix repeatable detection of uranium can be achieved at femtogram levels with sample introduction by ETV (Table 9). Careful optimiza- tion of the parameters allowed absolute LODs of less than 1 fg (0.2-0.5 fg) to be achieved. The best set of conditions to achieve repeatable analysis were as follows an EDTA sample matrix a Freon flow rate of 0.4 ml min-l a carrier gas flow rate of 0.75 1 min-l and the temperature profile given in Table 5. The improved LODs with the organic matrix compared with an aqueous matrix can be attributed to their different properties when in contact with a graphite surface and their improved complexing ability.Organics have a tendency to spread more than aqueous media when in contact with a graphite surface and hence the drying and ashing stages in the furnace might be more efficient when using organic matrices owing to the larger surface area of solution in contact with the furnace. A number of other matrices are to be examined that might be suitable for use in hydrothermal experiments on uranium 3000 2000 0 9 5 1000 0 0.2 0.4 0.6 ~ ~ ~ O ~ p p ~ O Concentration/ng ml-' Fig. 7 matrix over the range 0.001-1 ng ml-I of uranium Linear behaviour exhibited in a 0.01 mol I-' Na,CO,5 56 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 Table 9 Repeatability of 10 pl injections containing 10 fg of uranium in 0.01 mol I-’ EDTA and blanks Sample Sample Sample Blank Blank Peak height Peak area (ACPS) 12 323 15 348 15 33 1 10 186 8 I79 Mean of samples 13.5 335 Standard deviation 1.5 10.47 Precision 11.1 1 3.13 Mean of blanks 9.0 182.5 Standard deviation 1 .o 12.25 Precision 11.1 1 6.7 1 LOD Absolute LOD 1.2 fg 0.12 fg solubility. In particular preliminary work with a 0.01 mol 1-l Na2C03 matrix which forms a soluble carbonate with uranium shows linear behaviour in the furnace over the range 0.001-1 ng ml-I (Fig. 7). Deviation from linearity is observed when concentrations approach 10 ng ml-l sug- gesting incomplete vaporization and transportation of analyte concentrations exceeding 1 ng ml-’ in a 0.01 mol 1-l Na2C03 matrix.The problem of carbon build-up on the cones is eliminated with such a matrix and the solutions remain stable for several days. Optimization of several other parameters on the instru- ment is still possible. Forward power will play a major role in determining the properties of the plasma and hence variations in this value might improve detection limits significantly. Work on the determination of thallium using ETV has indicated that a peak is observed with increasing r.f. (forward power).34 Further the sampling distance (distance between the ETV unit and the plasma) can be varied and might lead to a further improvement in LODs and repeatability. The main problem with this technique is the substantial amount of carbon generated by the ashing of organic acids in the ETV furnace and its subsequent deposition on the cones. A method of solving this problem would be to introduce a venting mechanism whereby gases produced in the furnace prior to the vaporization stage are released into the atmosphere.The authors are grateful for the guidance and advice they received from Dr. P. Hulmston (formerly of VG Elemental Winsford Cheshire). Dr. M. R. Palmer is thanked for his comments on an early version of the manuscript. E.H.B. acknowledges an NERC studentship. References Kamai Y. Imai N. and Terashima S. Geostand. Newsl. 1986 10 73. Millard H. T. in Descriptions andAnalysis ofEight New USGS Rock Standards ed. Flanaghan F. J. US Geological Survey Professional Paper 1976 No. 840 61. Komura K. Tan K. and Ueno K.Geostand. Newsl. 1988 12 371. Potts P. J. and Rogers N. W. Geostand. Newsl. 1986. 10 121. Parks G. A. and Pohl D. C. Geochim. Cosmochim. Acta 1988 52 863. 6 Red’kin A. F. Savel’yeva N. I. Sergeyeva E. I. Omel’yan- enko B. I. Ivanov 1. P. and Khodakovsky I. L. in Experiment 89 Informative Volume ed. Zharikov V.A. USSR Academy of Sciences Institute of Experimental Mineralogy Moscow 1990 p. 79. 7 Beck G. L. and Farmer 0. T. 111 J. Anal. At. Spectrom. 1988 3 771. 8 Toole J. Hursthouse A. S. McDonald P. Sampson K. Baxter M. S. Scott R. D. and McKay K. in Plasma Source Mass Spectrometry ed. Jarvis K. E. Gray A. L. Jarvis I. and Williams J. Special Publication No. 85 Royal Society of Chemistry Cambridge 199 I p. 155. 9 Totland M. Jarvis I. and Jarvis K. E. Chem. Geol.1992 95 35. 10 Jarvis K. E. Chem. Geol. 1990 83 89. 11 Vollkopf U. Guensel A. Paul M. and Wiesmann H. in Applications of Plasma Source Spectrometry ed. Holland G. and Eaton A. N. Royal Society of Chemistry Cambridge 1991 p. 162. 12 Kantipuly C. J. Longerich H. P. and Strong D. F. Chem. Geol. 1988 69 17 1. 13 Park C. J. Van Loon J. C. Arrowsmith P. and French J. B. Anal. Chem. 1987 59 2191. 14 Hall R. J. B. James M. R. Wayman T. and Hulmston P. in Plasma Source Mass Spectrometry ed. Jarvis K. E. Gray A. L. Jarvis I. and Williams J. Special Publication No. 85 Royal Society of Chemistry Cambridge 1990 p. 145. 15 The Analysis of Uranium at ppq levels by ETV-ICP-MS Application Note VG Elemental Winsford 199 1. 16 Anderson G. M. Castet S. Schott J. and Mesmer R. E.Geochim. Cosmochim. Acta 199 1 55 1769. 17 Applications of Plasma Source Mass Spectrometry ed. Holland G. and Eaton A. N. Royal Society of Chemistry Cambridge 1991. 18 Plasma Source Mass Spectrometry ed. Jarvis K. E. Gray A. L. Jarvis I. and Williams J. Special Publication No. 85 Royal Society of Chemistry Cambridge 1990. 19 Langmuir D. Geochim. Cosmochim. Acta 1978 42 547. 20 Lemire R. J. and Tremaine P. R. J. Chem. Eng. Data 1980 25 361. 21 Dittrich K. Mohamad I. and Nielsbergall K. in Plasma Source Mass Spectrometry ed. Jarvis K. E. Gray A. L. Jarvis I. and Williams J. Special Publication No. 85 Royal Society of Chenistry Cambridge 1990 p. 18. 22 Matusiewicz H. J. Anal. At. Spectrom. 1986 1 171. 23 Park C. J. and Hall G. E. M. J. Anal. At. Spectrom. 1987 2 473. 24 Shen W.-L. Caruso J. A. Fricke F. L. and Satzger R. D. J. Anal. At. Spectrom. 1990 5 451. 25 Park C. J. and Hall G. E. M. Geol. Surv. Can. Pap. 1986 26 Hall G. E. M. Pelchat J.-C. Boomer D. W. and Powell M. J. Anal. At. Spectrom. 1988 3 791. 27 Igarashi Y. Kawamura H. and Shiraishi K. J. Anal. At. Spectrom. 1989 4 571. 28 Givindaraju K. Geostand. Newsl. Spec. Issue 1989 13. 29 Gregoire D. C. J. Anal. At. Spectrom. 1988 3 309. 30 Sato T. J. Appl. Chem. 1965 15 489. 31 Sato T. J. Appl. Chem. 1966 16 53. 32 Takeda K. Yamaguchi T. Akiyama H. and Masuda T. Analyst 199 1 116 501. 33 Kirkbright G. F. and Snook R. D. Anal. Chern. 1979 51 1938. 34 Park C. J. and Hall G. E. M. J. Anal. Ar. Spectrom. 1988 3 335. 35 Hulmston P. and Hutton R. C. Spectroscopy 1991 6 35. NO. 86-1B 767. Paper 2/00262K Received January 17 I991 Accepted December 14 1991

ISSN:0267-9477    

DOI:10.1039/JA9930800551

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 557 Speciation of Eight Arsenic Compounds in Human Urine by High-performance Liquid Chromatography With Inductively Coupled Plasma Mass Spectrometric Detection Using Antimonate for Internal Chromatographic Standardization Erik H. Larsen National Food Agency Central Laboratory 7 9 MdrkhH Bygade DK-2860 S@borg Denmark Gunnar Pritzl National Environmental Research Institute 399 Frederiksborgvej OK-4000 Roskilde Denmark Steen Honore Hansen Royal Danish School of Pharmacy 2 Universitetsparken DK-2 100 Copenhagen Denmark Four anionic and four cationic arsenic compounds in urine were separated by anion- and cation-exchange high- performance liquid chromatography and detected by inductively coupled plasma mass spectrometry (ICP-MS) at m/z 75.The species were the anions arsenite arsenate monomethylarsonate and dimethylarsinate and the cations arsenobetaine trimethylarsine oxide arsenocholine and the tetramethylarsonium ion. Hexahydroxyanti- monate(tti) was co-chromatographed with the arsenic anions but detected at m/z 121 and used as an internal standard for their qualitative analysis. Arsenite was prone to oxidation to arsenate in urine but was stable after at least 4-fold dilution of the urine with water. Arsenite was unstable in both urine samples and standard mixtures when diluted with the basic (pH 10.3) mobile phase used for anion chromatography. This could not be prevented by adding ascorbic acid as antioxidant. The argon chloride interference at m/z 75 was eliminated by chromatographic separation of the chloride present in the sample from the arsenic analytes. The CIO+ ion detected at m/z 51 and 53 was used to monitor the retention time of chloride in the anion-exchange system.The chloride eluted about 100 s after the last analyte peak and the intensity of ArCI+ was negligible even after injection of a 1% NaCl solution (less than 200 ions s-l). The recovery of all arsenic species in urine was close to 100%. The chromatographic peaks were evaluated by their peak heights and calibration was carried out by the method of standard additions. The calibration graphs were linear for all species (r>0.999). The limits of detection were 3-6 ng crn4 for the cations and 7-10 ng C M - ~ for the anions in urine. Keywords Arsenic speciation; human urine; ion chromatography; inductively coupled plasma mass spectrome- try; internal standardization Human intake of arsenic mainly occurs via food consump- tion and/or occupational exposure.After absorption in the gastro-intestinal tract or in the lungs the arsenic is elimi- nated from the body predominantly via the urine. The daily intake of arsenic with a normal Danish diet amounts to an average of 1 18 pg per day' About 86 pg of this daily intake originates from seafood and seafood products. From many studies reported in the literature it has been established that the most common by far arsenic species present in seafood is arsenobetaine (AsB) but other similar compounds like the tetramethylarsonium (TMAs) ion are present in smaller amounts.2 The molecular species of arsenic present in other types of food is not well documented.Radiotracer studies have shown that 73AsB is readily absorbed and almost totally excreted within approximately 3 d in rodents. The labelled AsB appeared unmetabolized in the urine of mice rats and rabbits after oral and intra- venous admini~tration.~ Radiotracer studies with 74AsB given to human subjects showed that the administered activity was almost completely eliminated via the urine. However no data on the speciation of the administered labelled arsenic in urine were given.4 Other studies how- ever indicated that AsB does not undergo biotransforma- tion in human^.^ In contrast to the reported unmetabolized urinary excretion of AsB occupational,6 environmental7 or cases of food exposure8 to inorganic arsenic have resulted in the presence of both inorganic and metabolized (methylated) forms of arsenic in urine from humans.The methylated forms found in the urine were dimethylarsinate (DMA) and monomethylarsonate (MMA). Trimethylarsine oxide (TMAO) a less often studied arsenic species has been found as a methylation product in human urine after intake of DMA.9 The various arsenic species possess different toxicological properties in humans ranging from AsB which is relatively non-toxic,I0 to inorganic arsenic which is a suspected human carcinogen." Therefore an adequate toxicological evaluation of arsenic exposure should be based on data concerning the chemical specia- tion. Additionally knowledge of the pattern of arsenic species (e.g. AsB versus arsenic anions) present in human urine provides information on the source of the arsenic exposure.The separation and detection of arsenic species present in urine has been performed by high-performance liquid chromatography (HPLC) coupled with hydride generation atomic absorption spectrometry (HGAAS)I2 or inductively coupled plasma mass spectrometry (ICP-MS).I3J4 Other less selective methods employ HG without previous separa- tion of the species in order to monitor total hydride- forming arsenic in urine which is used as a measure of arsenic exposure from environmental and occupational S O U ~ C ~ S . ' ~ ~ ' ~ Arsenobetaine from seafood cannot be deter- mined by hydride-based techniques12J6 unless a thorough base digestion has been carried out in advance.Although being highly and equally sensitive to all the arsenic species mentioned electrothermal atomic absorption spectrometry (ETAAS) can only be used as an off-line detector for liquid chromatography due to its discrete mode of operation.I7 The ICP mass spectrometer fulfils the ideal requirement of being an on-line real-time chromatographic detector that is highly sensitive to both anionic and cationic arsenic species. Urine is a sample type that often causes analytical problems because of its high salt content in addition to a558 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 range of organic constituents. A chromatographic system using a low ion strength mobile phase is particularly prone to disturbances. The urine matrix can also cause column overload which results in peak splitting or broadening of the analyte signals.Furthermore the chloride content of urine has been shown to cause an interference problem due to the formation of the 40Ar35C1+ ion at rnlz 75 (arsenic). Strong dilution of the urine samples might therefore be necessary in order to circumvent this pr0b1em.l~ Recently a method for the separation and detection of seven arsenic species using anion- and cation-exchange chromatography was published. l 8 The purpose of the present paper is to report further developments of the method with regard to the number of arsenic species covered and the adaptation and application of the method to the speciation of eight arsenic species that might be simultaneously present in human urine. Taking advantage of the multi-element capability of the ICP-MS detector an additional purpose was to investigate the possible co- chromatography of 0x0-anions of elements other than arsenic with a view to chromatographic internal standardi- zation.The problem of argon chloride interference on the arsenic determinations due to the high chloride content of urine is also addressed. Experimental Instrumentation A Waters 6000A HPLC pump and a U6K injector valve (Waters Associates Milford MA USA) were used in conjunction with the cation- and anion-exchange columns and mobile phases shown in Table I . The injector allowed introduction of variable sample volumes which was useful during the development of the method. The 0.10 mol dm-3 sodium carbonate buffer originally used18 for the anion- exchange chromatography was replaced with a 0. ! 0 mol dm-3 amm.onium carbonate buffer assuming that the latter would be less perturbing to the ICP mass spectro- meter.The polystyrene-divinylbenzene based polymeric anion-exchange column used in this study was similar in performance to that used previously. l8 Guard columns packed with the same stationary phases as the analytical columns were purchased from the suppliers of the analytical columns. The anion-exchange separations were run at 50 "C using an electrically heated HPLC column oven (Microlab h h u s Denmark). The chromatographic system was con- nected to the ICP-MS instrument by 20 cm of &. in poly(tetrafluoroethy1ene) capillary tubing (0.5 mm 1.d.) from the column outlet to the inlet hole of the standard cross flow nebulizer.The chromatographic flow rate of 1 cm3 min-l was compatible with the nebulizer uptake rate of the ICP-MS instrument which was a Perkin Elmer-SCIEX Elan 5000 (Perkin Elmer-SCIEX Thornhill Ontario Canada).The instrument was run in the graphics mode which allowed a continuous real-time monitoring of the arsenic-containing substances eluting from the chromato- graphic system. The mass spectrometer was set to sample ion intensities (by peak jumping) at one or more of mlz 5 1 (35C1160+); rnlz 53 (37Cli60+); mlz 75 (75A~+); rnlz 77 (40Ar37C1+ and/or 77Se); mlz 82 (82Se+) and rnlz 121 (lzlSb+). The ion intensities at rnlz 5 1 53,77 and 82 were of diagnostic value only in the investigation of the possible occurrence of 40Ar35C1+ interference on mlz 75.Prior to the HPLC-ICP-MS runs an aqueous arsenic standard solution was aspirated in order to check the sensitivity of the instrument. Optimized settings (Table 1) normally gave a signal of approximately 2000 ions s-l for a 10 ng cm-3 arsenic standard (Asv). The effect of increasing the dwell time per mass on the signal-to-noise (SIN) ratios of the chromatographic peaks was investigated and dwell times of 500 750 and 1000 ms were compared. During the cation- exchange HPLC-ICP-MS runs only rnlz 75 was monitored which made the use of a dwell time of 1000 ms possible without sacrificing peak resolution. This resulted in approx- imately a 1.5 times better SlN ratio compared with the 500 ms dwell time. For the detection of arsenic after anion chromatography an optimal 750 ms dwell time per mass was used because both mlz 75 and rnlz 12 1 were monitored.By using this dwell time a chromatographic point was acquired every 1500 ms whereby even the most narrow chromatographic peak in the anion-exchange system was determined by at least 20 points. Quantification was carried out by the method of standard additions. The chromatogra- phic peaks were evaluated by their heights from the baseline recorded prior to the elution of the first peak. For AsB however the peak heights were measured from the trough- to-trough baseline. Standard Substances and Chemicals Reference solutions of each of arsenous acid (As"') monoso- dium salt arsenic acid (Asv) disodium salt MMA disodium salt DMA monosodium salt AsB arsenocholine (AsC) bromide and TMAs iodide were obtained as described elsewhere.171L8 A solution of TMAO was kindly donated by Dr.W.R. Cullen (University of British Columbia British Columbia Canada). An estimate of the purity of the TMAO solution was obtained by running it through the cation- and anion-exchange HPLC-ICP-MS systems. A single peak Table 1 Operating conditions of the chromatographic and the ICP-MS systems Chromatography Anion-exchange column Mobile phase Cation-exchange column Mobile phase Flow rate 1 cm3 min-1 Injected volume ION 120 (125 x 4.6 mm i.d.) Interaction Chromatography (Mountain View CA USA) 0.10 mol dm-3 of NH4HC:03 at pH 10.3 with NH40H Ionosphere-C (1 00 x 3 mrri i.d.) Chrompack International (Middelburg The Netherlands) 0.1 mol dm-3 pyridinium ion at pH 2.65 with HCOOH 100 mm3 (anions) or 50 1311113 (cations) ICP-MS R.f.power Argon flow rates Plasmdcoolant Auxiliary Nebulizer Dwell time per mass Sweeps per replicate Nebulizer Spray chamber temperature 1350 W 14.80 dm3 min-l 0.80 dm3 min-l 0.98 dm3 min-I 1000 ms (cations) and 750 ms (anions) 1 Cross flow with sapphire bits 20.0 "C559 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 appeared in the cation-exchange chromatogram at a reten- tion time of 75 s. In the anion system a single peak eluted with the void volume at 90 s. The concentration of the TMAO solution was determined by ETAAS. l 7 Intermediate standard mixtures of the species except arsenite were prepared from the reference solutions at 5 pg ~ m - ~ each in water. The final dilute aqueous standard mixtures were prepared daily from these standard mixtures.Arsenite was added each day to the final solutions since prolonged storage even in the cold and dark caused oxidation of this species to arsenate. The chemicals used were of Merck pro analysi quality (Merck Darmstadt Germany) and the water was produced in a Millipore Super-Q apparatus (Millipore Bedford MA USA). All mobile phases were passed through a 0.45 ,ug filter and degassed before use. Samples and Sample Preparation Samples of first void (morning) urine were used throughout for the development and testing of the method and were stored in the dark at 4 "C before use. An aliquot of 12 cm3 of the urine was cleaned by slowly passing it through a silica- based SepPak CI8 disposable cartridge.The first 10 cm3 were discarded and the last 2 cm3 were kept for ana1y~is.l~ Before injection into the chromatographic system the urine was passed through a 0.45 pg filter under pressure from a 25 cm3 syringe. The cleaned and filtered urine was diluted 1 + 3 with water and spiked with standards of AslI1 AsV MMA DMA TMAO AsB AsC and TMAs. The samples were also spiked with hexahydroxyantimonate(n1) as the chromato- graphic internal standard. All solutions and spikings were carried out with calibrated micropipettes or Hamilton syringes in disposable polyethylene sample cups with lids. Results and Discussion Sample Preparation Free silanol groups of the silica stationary phase in chromatographic columns have been shown to act as ion exchanging sites for cationic arsenic species.This feature has been used analytically for HPLC separations of the cationic arsenic species.'* However during the sample preparation of urine using the silica-based SepPak C18 cartridges any retention of the analytes was unwanted. It was therefore investigated whether the silica-based material of the SepPak cartridges retained any of the arsenic analytes that were spiked in diluted urine (1 +3) at two concentra- tion levels (13 and 63 ng ~ m - ~ respectively of each species). The spiked urine was analysed before and after passage through the cleaning step by both anion- and cation-exchange HPLC-ICP-MS. Complete recovery of all eight cationic and anionic arsenic species was obtained at both concentration levels tested. Speciation of Arsenic Anions in Urine An advantageous feature of the carbonate mobile phase is its high pH value of 10.3 at which arsenous acid (pK 9.2) becomes ionized and is therefore retained on the ion- exchange column. In other anion-exchange chromatogra- phic systems operated well below the pK value arsenous acid is normally not retained and consequently elutes with the void volume.14 In the present anion-exchange system however none of the anionic analytes elute with the void volume.Therefore the problem of peak overlap from other unretained arsenic species co-injected with the sample e.g. AsB but not accounted for in the development of the separation technique has been eliminated. The chromato- grams of Asn1 Asv MMA and DMA co-injected with the cationic species are shown in Fig.1. The isocratic elution of 4 v) . (b' 5-81 4 Time/s Fig. 1 Anion-exchange HPLC-ICP-MS of arsenic species spiked in (a) aqueous solution and (b) in urine diluted ( 1 +3). A 0.5 ng amount of each species was injected. Peaks 1 DMA; 2 As"'; 3 MMA; 4 AsV and 5-8 AsB TMAO AsC and TMAs (eluted unresolved in the void volume). Peak 9 (broken line) corresponds t o 0.12 ng of hexahydroxyantimonate internal standard the anionic arsenic species takes about 5 min and retention times for the four anionic analytes in urine are similar to those of the aqueous standard mixture. Internal standardization of arsenic anions using the hexa- h ydroxyan t im ona t e(Iu,) ion Internal standardization normally requires an arsenic com- pound different from the analytes that will co-chromato- graph with a retention time within those of the arsenic analytes.Such a compound could be hard to find consider- ing the relatively compact chromatograms shown in Fig. 1. Therefore the use of 0x0-anions of other elements that are chemically related to arsenic anions were considered. Being a multi-element chromatographic detector the ICP mass spectrometer is ideally suited for this approach of internal standardization. A suitable substance would have to elute in one sharp peak within the retention times of the arsenic species and its detection should be free of interferences. Several 0x0-anions were tested as candidate substances and the hexahydroxyantimonate(rIr) ion Sb(OH)6- was se- lected because of its chemical similarity to arsenate. The pK value of 4.4 (ref.20) indicates full ionization of the antimonate at pH 10.3 of the mobile phase. The chromato- graphy of the hexahydroxyantimonate(rI1) ion under the conditions given for the separation of arsenic anions is shown in Fig. 1. The antimonate elutes with a retention time close to that of DMA. The narrow peak shape similar to that of the arsenic analytes indicates that the chromato- graphy of the Sb(OH)6- ion is efficient. Isobaric or polyatomic overlaps at mlz 121 (antimony) in ICP-MS are not likely to present a problem in the analysis of a urine matrix. Finally the urine samples tested with this method did not show any detectable inherent antimony. Therefore the hexahydroxyantimonate(I1r) ion fulfils the requirements of an internal standard. Preliminary experiments were undertaken in an attempt to find a suitable compound of an element other than560 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 arsenic for internal standardization in the chromatography of arsenic cations. However the compounds tested (metal ions) gave rise to strong chromatographic tailing and did therefore not qualify as internal standards. Research is now in progress to find more suitable substances for this purpose. Chemical stability of the analytes The urine analyses showed that under certain conditions the arsenite peak was reduced in size or disappeared completely. A decrease in height of the arsenite peak was always accompanied by an increase in the arsenate peak which indicates that arsenite was oxidized to arsenate. In standard mixtures freshly made up in water [Fig.2(a)] this problem never occurred which indicates that the urine matrix itself plays a role in the conversion of arsenite. The arsenite was completely converted too if the urine was diluted with the basic mobile phase [Fig. 2(b)] rather than with water prior to injection into the chromatographic system. This latter undesirable effect was also seen for standard mixtures made up in the mobile phase and illustrates that the oxidation was pH dependent. This is in accordance with the general information from a diagram showing redox potential versus pH for the arsenic-water system which indicates that with increasing pH in this case from the neutral pH of water to the pH 10.3 of the mobile phase the potential necessary to oxidize arsenite to arsen- ate decreases.An attempt to prevent the oxidation of arsenite by adding 1 O/o ascorbic acid as an antioxidant to the 120 100 80 100 I v) g 80 0 .- '$ 60 40 0 .- a 20 0 80 loo I (c' 60 40 20 c 5-8 I 0 100 200 300 Ti me/s Fig. 2 Effect of type of diluent on conversion of As'" in anion- exchange HPLC-ICP-MS. Arsenic species spiked in (a) aqueous solution (b) urine diluted ( 1 + 3) with mobile phase and (c) urine diluted ( 1 + 5) with 1% ascorbic acid. Amount of each arsenic species injected was (a) 9 ng (b) 9 ng and (c) 6 ng. For peak identification see Fig. 1 . Hexahydroxyantimonate(rr1) internal standard (2 ng) (broken line) was injected in (a) and (6) 1.4 ng in (c) urine [Fig. 2(c)] was unsuccessful and resulted in total disappearance of the arsenite peak and an irregular peak shape between the peaks corresponding to MMA and AsV was seen.Furthermore the peak from antimonate was diminished to almost zero. A series of dilutions of urine indicated that at least 4-fold dilution of the urine with water was necessary in order to prevent the full or partial conversion of As111 in urine. The curve of added arsenite standards in urine was linear (r=0.9990) up to 90 ng and was also parallel to the aqueous standard curve which indicates that the arsenite was stable in this concentration range. There was no evidence of any conversion of MMA and DMA. A rgon chloride interference The interference from the polyatomic ion 40Ar35C1+ at mlz 75 (arsenic) has been shown to occur in ICP-MS for a number of samples with a high content of chloride including ~ r i n e .* ~ - ~ ~ The ArCl ion is formed by combination of chlorine from the sample with argon from the plasma. In order to suppress the formation of the ArCl ion addition of a few percent nitrogen to the argon plasma has been successfully employed.22 Another approach was to separate the arsenic analytes from the chloride by HG which has been used for total arsenic determination^^^ and in arsenic speciation24 in chloride rich matrices. In speciation work involving liquid chromatography the chloride has been separated chromatographically from the arsenic ana- lytes.14v21 In order to investigate whether the argon chloride interferes with the detection of arsenic anions in this study a 1% m/v sodium chloride solution spiked with arsenic species was analysed.The salt concentration corresponded to twice the average chloride concentration in urine and was selected in order to account for natural variations of the chloride content of urine.25 Undiluted first void (morning) urine also spiked with arsenic species was also tested. The chloride that was monitored as the 35Cl'60+ ion at mlz 51 and as the 37C1160+ ion at mlz 53 eluted more than 100 s later than the last analyte peak (arsenate) and was therefore well separated from the anionic arsenic species as shown in Fig. 3. The ratio between the two monitored intensities of C10+ in both chromatograms is approximately 3 1 which corresponds to the natural abundance of the two chloride isotopes. The chromatographic peak detected as the C10+ ion is therefore assigned to chloride present in the sample.Furthermore as indicated in Fig. 3 the degree of Arc]+ formation at mlz 75 is limited to less than 200-300 ions s-l for both samples tested. A small increase in the intensity at mlz 77 was also observed (not shown in Fig. 3) and was caused by the formation of 40Ar37C1+. However a signal at this mlz value could also be attributed to 77Se but this possibility was excluded since no 82Se was detected at the simultaneously monitored rnlz 82. The low intensity of the argon chloride interference is in contrast to the findings in a similar study in which the injection of urine diluted 5-fold gave rise to a large signal from ArCl+.21 The reason for this is not known but it is worth noting that two different ICP-MS instruments were used.As an additional feature Fig. 3(a) also shows an example of the oxidation of arsenite to arsenate that takes place in undiluted urine as discussed in the previous section. By comparing the ratio of the chromatographic peaks corresponding to arsenite and arsenate in the undiluted urine in Fig. 3(a) with the ratio of the same peaks in the aqueous standard mixture in Fig. 2(a) it is clear that the arsenite peak height is reduced and the arsenate peak height is increased in the urine. In the cation- exchange chromatographic system the chloride eluted with the void volume and therefore did not cause any interfer- ence problems in the speciation of cationic arsenic species in urine.56 1 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 35 30 25 20 15 10 a 5 N O c I K 0 .- I\ I \ 3sc~0+ 37c~0+ - I I I 25t h 0 100 200 300 400 500 600 Time/s Fig. 3 Monitoring of the roAr3SC1+ interference in anion-exchange HPLC-ICP-MS of (a) eight arsenic species spiked in undiluted urine and (6) seven arsenic species spiked in 1% NaCl aqueous solution. Injected amounts of each of the species 1-4 were 4.5 ng in (a) and 0.9 ng in (b). Chloride monitored as 3sCl'60+ at m/z 5 1 and as 37C1160+ at m/z 53 gave rise to a small peak of 40Ar3SC1+ at mlz 75 in (b). For peak identification see Fig. 1 3 C I31 v) .- 15 10 5 0 ( b ) 6 I 50 100 150 200 Time/s Fig. 4 Cation-exchange HPLC-ICP-MS of arsenic species spiked in (a) aqueous solution and (b) urine diluted ( 1 + 3). Amount of each species injected 0.44 ng.Peaks 1 DMA; 2 As1''; 3 MMA; 4 AsV; 5 AsB; 6 TMAO; 7 AsC; and 8 TMAs Speciation of Arsenic Cations Including Trimethylarsine Oxide in Urine The separations of the cationic arsenic species AsB TMAO AsC and TMAs in aqueous solution and in urine are shown in Fig. 4. The co-injected DMA was used to indicate the void volume of the chromatographic system. The retention times of each of the different arsenic species are identical in both aqueous solution and urine. However the peak of AsB in urine is broader than the similar peak in aqueous solution which results in a reduced peak height. When the urine was diluted further and spiked at the same level the distortion disappeared. This behaviour can be explained by temporary column overload caused by a large amount of matrix constituents eluting close to the void volume where the elution of AsB also takes place.The shape of the peaks of TMAO AsC and TMAs that elute later in the chromato- gram were not disturbed even when the undiluted urine was analysed. Previous work on separation of seven arsenic species1* did not include TMAO. As shown in the chromatograms in Fig. 4 this arsenic species eluted between the peaks for AsB and AsC in the cation-exchange system. It is speculated that the cationic character of TMAO is due to protonation of the As=O bond which takes place in the acidic mobile phase (pH 2.7). The protonation most probably leads to the formation of the cationic compound (CH,),As+OH [or its corresponding salt (CH,),As+O-] in analogy with the protonated form of DMA at low pH.18 When injected into the anion-exchange system TMAO coeluted with the other cationic species in the void volume and its presence did therefore not disturb the chromatography of the anions.Under the conditions given in Table 1 the isocratic separation of the cations lasts about 2 min. The retention times obtained for the four arsoniurn ions present in the urine matrix and in aqueous solution are identical which demonstrates the robustness of the system. Method Characteristics The performance of the ion-exchange HPLC-ICP-MS speciation method has been characterized by recovery linear response range correlation coefficient (r) of standard curves limits of detection and repeatability in Table 2. The recoveries are close to 100% relative to aqueous standard curves except for AsB.When calculating the recovery of AsB relative to a standard additions calibration graph in urine the recovery of this compound is 100%. This reflects that some peak broadening and consequently a reduction in peak height for AsB in urine does occur. It was therefore necessary to calibrate by the method of standard additions. The values of r are better than 0.999 for all compounds and reflect a linear relationship between standards added to the urine and the ICP-MS signal. The repeatability was esti- mated for a blank urine (no detectable arsenic present) spiked at 2 ng cme3 of each species. The values obtained are surprisingly favourable taking into consideration that this spike level is close to the limits of detection for most compounds in the diluted urine.Arsenic Speciation In Urine After Consumption of Fish In order to make use of the analytical method for speciation of arsenic present in urine after the consumption of seafood one human subject ingested 250 g of meat from the flatfish species plaice (Pleuronectes pfutessu) caught in the northern part of The Kattegat Sea. Prior to consumption of the fish the subject did not ingest any food of marine origin for 1 week; rice and poultry were also excluded from the diet as these food items are known to contribute to the arsenic intake.' The first void of urine (330 cm3) was collected the next morning 11.5 h after the consumption of the fish. At this time the concentration of arsenic in the urine was expected to be at its peak value.26 The results of the HPLC-ICP-MS analyses are indicated in Fig.5. The intensity scale was selected to show the presence of the562 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Table 2 Method parameters for arsenic speciation in urine* by HPLC-ICP-MS LODS Urine (diluted 1 +3) Urine Arsenic Recovery? Linear rangel - (undiluted)/ Repeatability9 compound (Oh) ng cm-3 r ng ng absolute ng cm-3 (O/O) A n ions- DMA 102 2-1 50 0.9991 (n=6) 1.6 0.16 6.5 A P 96 2-90 0.9990 (n=5) 2.3 0.23 9.2 MMA 97 2-200 0.9999 (n=6) 2.0 0.20 8.0 AsV 107 2-90 0.9999 (n=5) 2.6 0.26 10.4 Cations- AsB 100 2-200 0.9994 (n=6) 1.5 0.075 6.0 TMAO 108 1-100 0.9998 (n=5) 0.7 0.035 2.8 AsC 106 1-100 0.9997 (n=5) 0.8 0.040 3.2 TMAS 106 1-100 0.9994 (n=5) 0.7 0.033 2.7 8.9 6.8 5.4 6.7 23 6.9 5.8 5.5 *All data are based on analyses of urine diluted with water (1 + 3) iind spiked at 2 5 10 38 89 163 and 490 ng ~ m - ~ of each arsenic TAverage values for recovery of spikes over entire linear range and calibrated versus an aqueous standard curve except AsB which was $Limit of detection (LOD) is estimated as the concentration (or amount) of arsenic compound giving a peak that is three times higher §Relative standard deviation from six repeated injections of urine spiked at 2 ng CM-~.anion and at 2 9 38 58 95 and 180 ng calibrated versus a standard additions curve in urine. than the amplitude of the baseline noise. of each arsenic cation. small peaks in both chromatograms. The strongly domi- nating chromatographic peak from AsB caused some overlap especially with the neighbouring smaller peak of TMAO and made further optimization of the cation- exchange chromatographic conditions necessary.This was carried out by running the chromatograms after a series of dilutions of the 100 mmol dm-3 pyridine mobile phase. A concentration of 35 mmol dm-3 proved optimal for separation of the cationic compounds present in the urine. The identity of the peaks in the modified chromatographic system Fig. 5(a) was confirmed by spiking with standards of the same compounds. In this system AsB eluting immediately after the void volume showed a split peak probably due to column overload caused by the urine matrix. A sharp peak of AsB with similar shape to that of the aqueous standard was obtained after dilution of the urine with water (1 + 10) and the AsB was quantified at 0 100 200 300 2500 ng in the original urine.In addition to AsB smaller amounts of TMAO and TMAs were also present as shown in the cation-exchange chromatogram of Fig. 5(a). The concentrations of these substances in the original urine were 21 and 30 ng ~ m - ~ respectively. The anion-exchange chromatogram in Fig. 5(c) shows the presence of peaks corresponding to DMA and As1'' at concentrations of 11 and 20 ng ~ m - ~ respectively. The sum of the small anionic and cationic peaks except that of AsB amounts to 2.5% m/v and AsB to 97.5% m/v of the arsenic compounds accounted for by this analysis. A first void urine collected the day before the ingestion of the fish was also analysed. The chromatograms of this blank urine in Fig. 5(b) and (d) show that none of the eight arsenic compounds covered by this method could be detected.The arsenic species detected in the urine therefore originated from the ingested fish The hexahydroxyantimonate internal standard in Fig. - 0 100 200 300 Time/s Fig. 5 Speciation of arsenic in urine after consumption of fish by (a) cation exchange and (c) anion exchange HPLC-ICP-MS (injected volume 50 mm3). Chromatograms (b) and (6) correspond to a blank urine analysed by the cation and anion techniques respectively. Hexahydroxyantimonate internal standard (0.35 ng) (broken line) was co-injected in (c) and (6). For peak identification see Fig. 4JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 563 5(c) (broken line) supports the identity of the first eluting anionic peak DMA which is not completely separated from AsB.At all the concentration levels tested in this work the retention time of the hexahydroxyantimonate ion likewise that of DMA was highly reproducible. Therefore the use of the antimony compound as an internal standard for qualitative analysis of the arsenic anions is feasible. The speciation of arsenic in samples of seafood is now under study using HPLC-ICP-MS with special emphasis on the presence of cationic compounds other than arsenobetaine. Conclusions An analytical method for the speciation of four anionic and four cationic arsenic compounds in human urine is pre- sented. The two chromatographic systems used proved robust towards possible chromatographic disturbances caused by the urine matrix.Internal standardization for qualitative analysis of the arsenic anions was carried out by co-chromatography of antimonate that was additionally detected by ICP mass spectrometer taking advantage of its multi-element capability. The possible interference from argon chloride on the detection of arsenic did not pose a problem. The chloride present in the urine was separated chromatographically from the arsenic analytes and when the chloride eluted from the column the argon chloride formed was insignificant in intensity at mlz 75. After intake of arsenic via a fish meal AsB TMAO TMAs DMA and arsenite were detected and quantified in the urine. Arseno- betaine accounted for about 97.5% of the total arsenic species determined in the urine by this method. References Food Monitoring in Denmark.Nutrients and Contaminants 1983- 1987 National Food Agency of Denmark 1990. Publi- cation No. 195 ISBN 87-503-862 1-2. Cullen W. R. and Reimer K. J. Chem. Rev. 1989 89 713. Vahter M. Marafante E. and Dencker L. Sci. Total Environ. 1983 30 197. Brown R. M. Newton D. Pickford C. J. and Sherlock J. C. Hum. Exp. Toxicol. 1990 9 41. Cannon J. R. Edmonds J. S. Francesconi K. A. and Langsford J. B. in Management and Control of Heavy Metals In The Environment CEP Consultants Edinburgh 1979 pp. Offergelt J. A. Roels H. Buchet J. P. Boeckx M. and Lauwerys R. Br. J. Ind. Med. 1992 49 387. 283-286. 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Kalman D. A. Hughes J. van Belle G. Burbacher T. Bolgiano D. Coble K. Mottet N. K. and Polissar L.Environ. Health Perspect. 1990 89 145. Crecelius E. A. Environ. Health Perspect. 1977 19 147. Marafante E. Vahter M. and Norin H. Envall J. Sand- strom M. Christakopoulos A. and Ryhage R. J. Appl. Toxicol. 1987 7 I I I. Food and Agriculture Organization (FAO) World Health Organization (WHO) Toxicological Evaluation of Certain Food Additives and Contaminants The 33rd Meeting of the Joint FAO/WHO Expert Committee on Food Additives. Geneva 1989. Food and Agriculture Organization (FAO) World Health Organization (WHO) 1983 WHO Food Addit. Ser. No. 18. Chana B. S. Smith N. J. Anal. Chim. Acta 1987 197 177. Heitkemper D. Creed J. Caruso J. and Fricke F. L. J. Anal. At. Spectrom. 1989 4 279. Sheppard B. S. Shen W-L. Caruso J. A. Heitkemper D. T. and Fricke F. L. J. Anal. At. Spectrom. 1990 5 431. Norin H. and Vahter M. Scand. J Work. Environ. Health 1981 7 38. Miirer A. J. L. Abildtrup A. Poulsen 0. M. and Christen- sen J. M. Talanta 1992 39 469. Larsen E. H. J. Anal. At. Spectrom. 199 I 6 375. Hansen S. H. Larsen E. H. Pritzl G. and Cornett C. J. Anal. At. Spectrom. 1992 7 629. Hakala E. and Pyy L. J. Anal. At. Spectrom. 1992 7 191. Stability Constants of Metal-Ion Complexes. Section I Inor- ganic Ligands. Compiled by Lars Gunnar Sillen. Special Publication No. 17. The Chemical Society. London 1964 p. 207. Sheppard B. S. Caruso J. A. Heitkemper D. T. and Wolnik K. A. Analyst 1992 117 971. Branch S. Ebdon L. Ford M. Foulkes M. and O’Neill P. J. Anal. At. Spectrom. 1991 6 151. Branch S. Corns W. T. Ebdon L. and O’Neill P. J. Anal. At. Spectrom. 1991 6 155. Story W. C. Caruso J. A. Heitkemper D. T. and Perkins L. J. Chromatogr. Sci. 1992 30 427. Report of the Task Group on Reference Man International Commission on Radiological Protection No. 23 Pergamon Press 1975 p. 378. Foa V. Colombi A. Maroni M. Buratti M. and Calzaferri G. Sci. Total Environ. 1984 34 241. Paper 2/06692K Received December 17 1992 Accepted February 25 1992

ISSN:0267-9477    

DOI:10.1039/JA9930800557

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 565 Laser Vaporization Inductively Coupled Plasma Mass Spectrometry a Technique for the Analysis of Small Volumes of Solutions R. Krishna Prabhu S. Vijayalakshmi T. R. Mahalingam K. S. Viswanathan and C. K. Mathews Radiochemistry Programme lndira Gandhi Centre for Atomic Research Kalpakkam 603 102 India The laser vaporization process was demonstrated to be a suitable technique for introducing solutions into the inductively coupled plasma (ICP) similar to the introduction of solids using a laser ablation process. The technique was found to have several advantages over the conventional nebulization method of sampling into the ICP e.g. a large dynamic range of concentrations that can be analysed (5 ng ml-l-mg ml-l) small sample volumes required for an analysis (200 PI) and no collection of drain solution.Details of the technique optimization of the different experimental parameters and its utility as an analytical tool are described. Analysis of a National Institute of Standards and Technology Standard Reference Material NlST SRM 1606 Stainless Steel for a number of elements showed good agreement with certified values with relative standard deviations in the range 2-8%. Suppression of the analyte signal due to matrix effects is also discussed. Keywords Laser vaporization; inductively coupled plasma mass spectrometry; solution sample sample introduction Laser ablation inductively coupled plasma mass spectrome- try (ICP-MS) has gained considerable popularity in recent years for the analysis of solid sample^,^-^ for various reasons.The solids can be sampled directly without any need for dissolution thus minimizing sample preparation. Further the amount of the sample injected into the plasma can be controlled by suitably adjusting the laser intensity which widens the dynamic range of the technique. Concen- trations of analytes ranging from percentage to parts per million (ppm) levels can be measured without the need for any dilution or preconcentration usually resorted to with solutions.6 The technique is also suitable for the analysis of both conductive and non-conductive solid samples. Recently the possibility of adapting this technique for the analysis of solutions has been explored. Our interest in doing so was twofold. Firstly this technique would use a considerably smaller volume of solution for an analysis than a nebulizer and would not generate waste drain solution.This would be particularly attractive in the analysis of solutions containing radioactive elements. Sec- ondly by controlling the amount of solution injected into the plasma (by controlling the laser intensity) an increase in the dynamic range of the ICP-MS over that usually obtained using a nebulizer was expected. In this paper the development of a technique to analyse small volumes of solutions using laser vaporization ICP-MS is reported. Experimental Instrumentation All ICP-MS measurements were made using an Elan 250 system (SCIEX). Typical instrumental operating and mea- surement conditions are given in Table l. An HY400 Nd:YAG laser (Lumonics) was used for the vaporization process.The operating parameters of the laser were as follows pulse width (temporal) 10 ns; wavelength 532 nm; typical pulse energies 1-2 mJ per pulse (Q- switched mode); and repetition rate 10 Hz. Vaporization Cell The laser ablation cell shown in Fig. 1 was designed and fabricated in-house originally for studies on laser ablation of solids. The cell was then modified to accept solution samples. About 200 pl of the test solution were placed in a small poly(tetrafluoroethy1ene) (PTFE) boat and a graphite wheel (ultra F purity Ultra Carbon) partially dipped in the Table 1 Typical instrumental operating and measurement condi- tions Operating conditions- Coolant argon flow rate Auxiliary argon flow rate Sampling depth 23 mm 12 1 min-l 1.6 I rnin-' Lens settings B P E s2 50 (5.0 V) 89 ( - 18.0 V) 19 (-11.4 V) 16 (-3.4 V) Measurement conditions- Resolution Low ( 1 k O .1 u) High (0.6 kO.05 u) Mu1 t i-elemen tal (peak hopping) Scanning mode Measurement time I s Measurements per peak 1 Dwell time 0.025 s Repeats per integration 10 test solution was rotated using a stepper motor. The film of solution that adhered to the wheel then passed through the laser beam as the wheel rotated. The laser beam was focused using a 100 mm focal length Suprasil lens with the focal point being about 10 mm below the wheel surface from where the vaporization occurred. Tight focusing at the wheel surface was avoided to prevent any laser ablation of the graphite wheel. The solution vapours that were pro- duced as a result of the laser heating were then swept away by a stream of argon (flow rate 2 1 min-I) into the ICP through a 1.75 m long Tygon tube.This length of tubing was the shortest that could be used given the physical con- straints of the set-up. A two-way stop-cock was introduced in the argon line so that samples could be changed without having to cut off the argon supply to the ICP. Reagents and Chemicals All chemicals used were of Specpure grade (Johnson Matthey) and distilled de-ionized water (1 8 MR) obtained with a Milli-Q system (Millipore) was used to prepare solutions of the required concentrations. Results and Discussion First the ion intensities of the different elements in the test solutions were measured for a period of about 10 min to566 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 Argon in Graphite disc Solution Fig. 1 Schematic diagram of the laser vaporization cell 35 20 6 n 0 X v) r - z 5 8 0.35 - 0.20 0.05 Time/m in 63c u Fig. 2 Plot of ion intensity against time for Pb (1 pg ml-I) and Cu (2.5 pg ml-I) over a 10 min period. Laser intensity used was 1.5 mJ per pulse determine the temporal characteristics of the signal. The test solution consisted of Pb Cu Li and Cd. As the results obtained with each of these elements were similar only those of Pb and Cu will be presented here. Fig. 2 shows the ion intensities as a function of time for Cu (2.5 pg ml-I) and Pb (1 pg ml-l) with the laser operated at 1.5 mJ per pulse. It can be seen that for the experimental configuration that we used the ion intensities reached a steady value about 2 min after the laser was turned on.The time taken to reach the steady-state value varied from element to ele- ment but 2 min appeared to be typical by which time constant ion intensity values were obtained for most elements. In the measurement of concentrations all ion intensity data were recorded only after their values had stabilized. Once stabilized the scatter in the ion intensity values for a given run was about 5%. However between runs the ion intensities showed a maximum scatter of 14 * 0 v) 3 c x 10 +d 8 c 6 - 2 1 2 3 4 28 1 1.2 0.8 < -F 0.4 1 2 3 4 Laser energy/mJ per pulse Fig. 3 Plot of ion intensity and intensity ratio against laser energy for (a) 63Cu (2.5 pg m1-I) and (b) 2osPb ( 1 pg ml-I) T1 ( 1 pg ml-I) and Ga ( 1 pg ml-I) were used as internal standards for Pb and Cu respectively to calculate the intensity ratios about 15%.Once the laser was turned off the ion intensities dropped instantaneously to background values for all the elements in the test solution except Cd at high concentra- tions. For Cd alone at concentrations of about 1 mg ml-I the ion signals persisted for several minutes. Low back- ground values were then restored only after the cell and tubing had been thoroughly cleaned. Effect of Laser Energy Fig. 3 shows the variation of ion intensities as a function of laser energy. It can be seen that the ion intensities increased as the laser energy (energy per pulse) was increased because the amount of the solution vaporized and injected into the plasma was higher at higher laser energies. This effect can be utilized to control the amount of solution sampled into the ICP.A low laser energy can be used when sampling concentrated solutions whereas when sampling dilute solutions the laser energy can be increased. However when laser energies of about 3 mJ per pulse and above were used condensation of water on the lens surface due to excessive vaporization of the test solution was observed. The condensation resulted in a monotonic decrease in the ion intensities over a period of time as the laser intensity reaching the solution vaporization region was attenuated. The problem of water condensation therefore set an upper limit on the laser energy that could be used. Therefore in subsequent work the laser energies were kept below 3 mJ per pulse typically 1.5 mJ per pulse.Effect of Argon Flow Rate The important parameter governing the transport of the vaporized solution into the plasma is the flow rate of the argon carrier gas sweeping through the sample cell. Fig. 4 shows the ion intensities of four elements as a function of the carrier argon flow rate. A flow rate of about 2 1 min-'JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 567 20 c I Y 16 7 0 * 12 \ C > v) .- 4- g 8 .- C - 4 0 0.8 1.4 2.0 2.6 3.2 Ar flow rate/l min-' Fig. 4 Plot of ion intensity against carrier Ar flow rate for A Li; B Cu; C Pb; and D Cd 286.0 171.6 57.2 2 v) 3 c s 165 C - 99 33 5 15 25 2 6 10 Concentration/pg m1-l Fig. 5 Plot of ion intensity against concentration for (a) (r=0.958) and (b) 208Pb (r=0.909) over the concentration range 0.25-25 pg ml-I was found to be optimum.At lower flow rates the signal intensity decreased owing to the reduced transport effici- ency. At higher flow rates the analyte was diluted in the carrier gas stream causing a reduction in the ion signal intensity. Tthe optimum argon flow rate of 2 1 min-' was therefore used for all further experiments. Concentration Calibration A plot of the ion intensities of Pb and Cu in the test solution as a functon of concentration (over the concentration range 0.5-25 pg ml-l) is shown in Fig. 5. A laser energy of I .5 mJ per pulse was used in this experiment. The straight line is the least-squares fitted line. As only raw ion intensity values have been plotted against concentration the fit is rather poor owing to the scatter in the ion intensity values between different runs.To correct for these fluctuations internal standards were used in the test solutions. Each of the test solutions contained a constant concentration of an internal standard such as Ga TI or Sb. The ion intensity of the 3.20 1.92 a a \! -0 0.64 5 15 25 1.5 0.9 5 4- 0.3 Concentration/ pg m I-' Fig. 6 Plot of intensity ratio against concentration for (a) 63Cu (r=0.997) and (b) 208Pb (r=0.985) over the concentration range 0.25-25 pg ml-I T1 ( 5 pg m1-I) and Ga ( 5 pg ml-I) were used as internal standards for Pb and Cu respectively to calculate the intensity ratios element of interest was divided by the ion intensity of the internal standard for a given run.This corrected for any short-term fluctuations in the experimental conditions. It can be seen from Fig. 3 that even when the laser energy was deliberately changed by more than a factor of three the intensity ratio (Ie,e,e,tlIi, std,) remains virtually constant even though the ion intensities of the elements varied markedly. In choosing the internal standards elements with comparable masses to those of the elements of interest were used. A plot of the intensity ratio against concentration of Pb and Cu is shown in Fig. 6 for the same data as those shown in Fig. 5. It can be seen that the fit is excellent with a correlation coefficient of 0.99. Similar results with other elements such as Li and Cd were obtained. Next the possibility of using this technique for higher ( 100- 1000 pg ml-*) and lower ( 10- 100 ng ml-I) concentra- tion ranges was examined.To handle the higher concentra- tion range the laser intensity was reduced so that less material was injected into the plasma. The laser intensity was reduced by placing an aperture in the path of the laser beam. (Owing to physical constraints of the set-up the intensity of the laser after the aperture could not be measured but it is probably in the region of 1 mJ or less.) A plot of the ion intensity ratio against concentration in this higher concentration range also yielded fits similar to that shown in Fig. 6 with a correlation coefficient of 0.99. Solutions at higher concentrations can therefore be ana- lysed without any dilution such as is necessary in solution nebulization. Similar results were obtained for the concen- tration range 10-100 ng ml-l for which a laser energy of 1.5 mJ per pulse was used.For the elements studied it is estimated that the detection limits (calculated as the concentration corre- sponding to three standard deviations of the blank) is about 5 ng ml-l. This limit could not be improved upon by increasing the laser energy because as was pointed out earlier condensation of water on the lens surface set an upper limit on the laser energy that could be used. The vaporization cell is being re-designed in order to eliminate568 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 ~ ~~ ~ Table 2 Determination of elements in NIST SRM 160b by laser vaporization-ICP-MS Concentration (O/O) Certified By concentration By standard Element Is0 t ope value calibration* additions* Pb Pb V c o c u c u Mn Mo Ni Cr 206 208 51 59 63 65 55 90 60 52 0.00 1 0.00 1 0.047 0.10 0.I72 0.172 I .64 2.38 12.2 18.4 0.0010 (8.3) 0.001 1 (7.0) 0.044 ( 5 . 5 ) 0.104 (4.0) 0.168 (3.5) 0.168 (3.8) 1.60 (4.4) 2.30 (2.2) 12.36 (1.3) 18.33 (1.9) 0.001 1 (4.0) 0.0012 (4.4) 0.049 (8.7) 0.103 (4.8) 0.173 (3.1) 0.174 (4.8) - - - - *The concentrations given are the means of four values and the values in parenthesis are the relative standard deviations of the mean. condensation of water on the lens surface. Another pos- sibility that could be explored to improve the detection limits would be to use a higher repetition rate of the laser as the amount sampled per unit time would then be higher.An excimer laser might be suitable for this technique. Analysis of Certified Standard A certified National Bureau of Standards (NBS) [now National Institute of Standards and Technology (NIST)] Standard Reference Material (SRM) 160b Stainless Steel Cr-Ni-Mo (AIST 3 16) was analysed to check the validity of this technique. Eight elements whose concentrations in the standard ranged from ppm to percentage levels were determined in order to check the usefulness of this technique for the analysis of solutions containing elements over a wide concentration range. The solid standard was dissolved to yield a solution in which the concentration of the reference material was 10 mg ml-I. The concentrations of the eight elements in this solution were then determined.The concentrations of Pb V Co and Cu (which ranged from 0.001 to 0.2%) were determined both by the use of a concentration calibration graph and by the method of standard additions. The concentrations of Mn Mo Ni and Cr (which ranged from 1.6 to 18%) were determined by the use of a concentration calibration graph alone. In these experiments high resolution was used in order to avoid interference from Fe flanking the two sides of the 55Mn peak. The laser intensity was also reduced by using an aperture when handling the higher concentrations. The results together with the certified values are given in Table 2. It can be seen that the experimental results are in fairly good agreement with the certified values. Matrix Effects When analysing the NIST SRM it was observed that the ion intensities of the analytes were lower than those obtained with pure solutions of the analytes.This is because of the presence of a high concentration of Fe constituting the matrix in the solution of the NIST SRM. Suppression of ion intensities due to such matrix effects is well known in ICP-MS.7-8 To clarify this point further the ion intensities of Cu and Ga in solutions containing various amounts of Fe (up to 20 mg ml-*) were recorded. Fig. 7(a) shows the percentage suppression for Cu and Ga as a function of Fe concentration in the solution. The Cu and Ga intensities were progressively suppressed with increase in the Fe concentration with the suppression amounting to about 80% for an Fe concentration of 20 mg ml-I. The analyte signal suppression is given by analyte signal suppression (O/O) = x 100 1 analyte signal analyte signal (without matrix)-( with matrix analyte signal without matrix However the ratio of the ion intensities of Cu and Ga remained almost constant over the range of Fe concentra- tions studied.The internal standard method (which uses ratios of intensities) therefore corrects for the matrix suppression. Similar results were obtained with Pb and T1 as shown in Fig. 7(6). As the absolute ion intensities show a scatter of about 15% between two individual runs it is not possible at 0.6 4 -sJ -u = 0.4 0.2 0 - 0.6 % -!+ - 0.4 - 0.2 10 15 20 25 0 5 Concentration of Fe/mg ml-’ Fig. 7 Plot of analyte signal suppression and intensity ratio against concentration of Fe for ( a ) Cu and Ga and (h) Pb and TIJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE present to establish what proportion of the matrix effect is due to the plasma and what part if any is due to the laser sampling process.The scatter in the absolute ion intensities also results in a comparable uncertainty in the values of percentage analyte suppression. However above a matrix concentration of 5 mg ml-l the suppression exceeds this uncertainty and the matrix effects are therefore obvious. A more detailed study of the matrix effects interelement effects and relative sensitivity factors for the laser vaporiza- tion process (compared with the nebulization process) is in progress. Conclusions The technique described here enjoys a number of advan- tages over the nebulization technique.First the amount of solution required for an experimental run is very small (200 pl). This volume is required for the sampling graphite wheel to be properly dipped and wetted in the test solution. However of this volume only a negligible fraction is sampled into the plasma in a typical experimental run lasting 2-5 min. By comparing the ion intensities observed in this experiment and in the nebulization method (and assuming a sample introduction efficiency of 1% for the nebulization technique9) it was estimated that in the laser vaporization technique the sample is introduced into the plasma at a rate of about 2 pl min-l with the remainder left behind in the PTFE boat. In the nebulization technique the solution is consumed at a rate of about 1 ml min-l and generally about 2 ml of the solution are required for a typical experimental run. Of this only about 1% of the solution consumed is actually sprayed into the plasma while the remainder is drained as waste.These features of the laser vaporization method namely a small analyte volume and no collection of drain solution make it particularly attractive in the analysis of radioactive solu- tions where handling of large volumes and collection of waste are undesirable. The results presented show that the range of concentra- tions analysed is 5 ng ml-l-1 mg ml-l thus offering a large dynamic range. With the nebulizer solutions at even 5 pg ml-l levels would be difficult to handle and dilution is usually resorted to which is both time consuming and a source of errors. Another technique that handles small volumes of analyte solutions is electrothermal vaporization (ETV).l0-l2 In fact ETV handles even smaller volumes ( I p1)l0 than laser vaporization.Both of these techniques also lead to a reduced oxide formation compared with the nebulization technique.1° For example with La and Ce it was found that the intensity of the MO+ peak (where M=La or Ce) is 1993 VOL. 8 569 about 1 O/o of M+ with laser vaporization compared with 5Oh with the nebulization technique. The sensitivity offered by ETVl0-l2 is at present superior to that offered by laser vaporization. However as mentioned earlier the use of higher repetition rates for the laser would increase the sensitivity of the laser vaporization technique. Laser vapor- ization however has a wider dynamic range making possible the analysis of higher concentrations of analytes without dilution.Another feature of the laser vaporization process is the steady-state ion signal observed as opposed to the transient pulse obtained with ETV. This feature of the laser vaporization technique makes it more convenient to undertake multi-element determinations. The laser va- porization set-up is also amenable to containment in a glove-box for handling radioactive solutions. Laser vaporization has been shown to be a convenient and efficient technique to sample solutions into the ICP. With only minor changes in the experimental set-up the technique can handle both solids (laser ablation) and solutions. Further work is necessary to optimize the set-up in order to handle samples on a routine basis. Matrix effects and interelement effects also need to be studied and such studies are now in progress. We thank A. Veerapandian for his assistance in the design and fabrication of the laser vaporization cell. 1 2 3 4 5 6 7 8 9 10 11 12 References Yasahura H. Okano T. and Matsumura Y. Analyst 1992 117 395. Hager J. W. Anal. Chem. 1989 61 1243. Gray A. L. Analyst 1985 110 551. Arrowsmith P. Anal. Chem. 1987 59 1437. Arrowsmith P. and Hughes S. K. Appl. Spectrosc. 1988,42 1231. Ishizuka T. and Uwamino Y. Spectrochim. Acta Part B 1983 38 519. Gillson G. R. Douglas D. J. Fulford J. E. Halligan K. W. and Tanner S. D. Anal. Chem. 1988,60 1472. Vijayalakshmi S. Krishna Prabhu R. Mahalingam T. R. and Mathews C. K. At. Spectrosc. 1992 13 61. Browner R. F. in Inductively Coupled Plasma Emission Spectroscopy-Part 2 ed. Boumans P. W. J. M. Wiley New York 1987 ch. 8. Gregoire D. C. J. Anal. At. Spectrom. 1988 3 309. Park C. J. and Hall G. E. M. J. Anal. At. Spectrum. 1987 2 473. Shibata N. Fudagawa N. and Kubota M. Anal. Chem. 199 1 63 636. Paper 2/05 063C Received September 22 I992 Accepted January 8 I993

ISSN:0267-9477    

DOI:10.1039/JA9930800565

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 57 1 Gas Chromatographic Determination of Phosphorus Sulfur and Halogens Using a Water-cooled Torch With Reduced-pressure Helium Microwave- induced Plasma Mass Spectrometry W. Charles Story* and Joseph A. Carusot Department of Chemistry University of Cincinnati Cincinnati OH 45221 -01 73 USA A system is described that utilizes a reduced-pressure water cooled microwave-induced plasma torch to interface a gas chromatograph with a plasma mass spectrometer. The detection limits vary with the compound used. For phosphorus in triethyl phosphate the detection limit is 86 pg and for chlorine in chlorotoluene it is 22 pg. A seven component mixture was analysed simultaneously for compounds containing chlorine (m/z35 and m/z 37) phosphorus (m/z 31) sulfur (m/z 32) bromine (m/z 79 and m/z 81) and iodine (m/z 127).A Supelco pesticide mixture was also analysed with simultaneous detection of chlorine and sulfur. Sub-nanogram detection limits were obtained for all elements studied. Keywords Determination of phosphorus sulphur and halogens; gas chromatography; water-cooled torch; reduced pressure microwa ve-induced plasma; mass spectrometry In a previous study the first experiments to determine phosphorus and sulfur using our plasma mass spectrometer were described. These determinations are difficult since polyatomic interferences at typical phosphorus and sulfur mlz values are virtually ubiquitous. This paper reports on several modifications to the plasma spectrometer inter- face that result in improved analytical capabilities.The purpose of this experiment is to investigate a plasma mass spectrometer interface that would take advantage of the multi-element detection capabilities of our plasma mass spectrometer and permit the selective detection of phos- phorus and halogens through the use of a helium micro- wave-induced plasma (MIP). Considerable effort has been spent in the development of various methods for element-selective detection in gas chromatography (GC) eluate^.^.^ A number of early papers were published in the 1960s describing the deter- mination of phosphorus using atomic emission spectro- met~-y.~-~ More recent work by Rivikre et ~ 1 . ~ 9 ' ~ has improved on the earlier results. Substantial work has been carried out in our laboratories on the determination of halogenated compounds by atomic emission spectrometry as well as mass spectrometry (MS).lI-l4 Quimby and Sullivan15 have developed a commercially available system that is versatile and well accepted and Barnes and Reszke16 have investigated different MIP sources.In our first study,' the detection limits obtained were higher than reported for some of the other methods commonly used. It was decided that further improvements in the experimental design were needed. One component that needed upgrading was the torch. A water-cooled torch has now been developed to improve the sensitivity of the system. Furthermore to reduce the phosphorus interaction with the torch walls the addition of a small percentage of hydrogen to the plasma gas was also investigated. Hydrogen addition serves two purposes as a reagent gas to scavenge the phosphorus before it reacts with the hot quartz of the torch; and as an aid in reducing the polyatomic-ion formation.Reducing polyatomic-ion formation results in lowering of background signal intensities. Another area of the system where improvement is possible is the interface between the GC transfer line and the plasma torch. There was unacceptable band broadening with the previous system and an attempt was made to ~~ *Present address Environmental Health Research and Testing ?To whom correspondence should be addressed. 3235 Omni Drive Cincinnati OH 45232 USA. reduce this by minimizing the dead volume in the interface. This was realized by incorporating a tantalum tube at the transfer tube outlet to help transfer the analyte directly to the plasma thereby reducing possible losses to the torch wall.The effect of each of these changes on the determination of phosphorus and sulfur is discussed. Figures of merit for phosphorus and chlorine were determined and a polychlori- nated pesticide standard mixture was analysed. Experimental Experimental details are given in the previous study' except for the following differences. The torch (Precision Glas- blowing of Colorado) consisted of a 2 mrn central channel with a 1.5 mm thick water layer containing a baffle to direct the water flow. The outer diameter of the torch was in. A collar was added to the end of the torch to permit connection to the vacuum system with a 1 in ultra-Torr fitting.A schematic drawing of this torch is shown in Fig. 1. A new adaptor was constructed to connect the GC transfer line to the torch. This connector consisted of a stainless-steel piece machined with a & in fitting for the transfer line connection. A in diameter tantalum tube was press fitted into the opposite side to transport the GC eluates to the plasma. The reagents used have been described previously' except for a pesticide mixture that was obtained from Supelco. The sample is a Supelpreme-HC pesticides mixture containing 200 pl ml-I of 18 different polychlorinated pesticides in a 50-50 mix of hexane and toluene. Some of the common pesticides included in the mixture are aldren dieldrin endrin 4,4'-DDT endosulfan sulfate and endosulfan. Results and Discussion The first objective was to characterize the effect of water cooling on the observed background signals.A series of experiments were performed where the background signal at rnlz 3 1 was observed as a function of plasma power and cooling temperature. The plasma gas was pure helium at a helium flow rate of 200 ml min-*. The resulting curves are shown in Fig. 2. The most obvious feature in Fig. 2 is the similarity between all the curves. It would appear that at least in the temperature range studied small temperature differences do not have a noticeable affect on the back- ground intensity. The other important information ob- tained from this temperature study is how quickly the background intensity increases with forward power. Also of572 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 i in End view i in Water out 2 mm i.d. - fin Water in Fig. 1 Schematic diagram of water cooled torch 8x1 o5 A 2 6x105 .- C 3 2 L c 4x105 Q > v) - *-' .- - i!! 2x105 0 Fig. 2 100 200 300 400 500 Power/W Plot of signal intensity at m/z= 3 1 as a function of plasma forward power and cooling water temperature with a He flow rate of 200 ml min-l. A 25; B 20; C 15; D 10; and E 5 "C interest here is that the power at which the background exceeds 100000 counts s-l is only about 220 W. When hydrogen is added to the plasma gas the background at 400 W is still much lower than 100 000 counts s-l. It is believed that the shape of this background curve is such that there is a threshold power where ablation of the inner torch wall begins.The presence of OH functional groups in the quartz of the torch provides a ready source of oxygen for the formation of NOH+ (mlz= 3 1 ) . This would explain why the background is very low below 200 W followed by rapid background increases. A second study was performed to characterize the effectiveness of the water cooling in preventing torch ablation. Scans were collected that included rnlz 30 and 3 1. These masses correspond to NO+ and NOH+ respectively. It is believed that the NO+ would result from residual gases present in the helium. The NOH+ species could possibly come from the OH ablation of the inner torch wall. The results of this study are shown in Fig. 3. The graph for the signal intensity at rnlz 3 1 has been multiplied by a factor of 5 to make it visible on the same scale as the mlz 30 signal.What can be seen is a relatively steady increase in the intensity at mlz 30 which is to be expected if the contribution to the signal results from the gas phase. If the level of nitrogen and oxygen in the plasma is constant then as the plasma energy increases there should be a smooth increase in the amount of NO+ formed. The signal at rnlz 31 remains relatively constant with the exception of the 80 r v) $ 60 3 8 n s! 40 \ c > v) .- 20 - 0 n 0 d Z = 3 0 fW"=31 J 150 170 190 210 230 250 PowerAN Fig. 3 Plot of signal intensity at mlz=30 and m/z=3 1 at different plasma forward powers at 10 "C data point at 190 W. Why the intensity at 190 W is higher than the others is still unclear. It is possible that as the plasma changes length with changes in power an optimal set of conditions for the formation of the polyatomic species arises.If one temporarily disregards the data point at 190 W the signal at mlz 3 1 remains relatively constant as the power is increased. This indicates that the rnlz 3 1 signal is due to species being ablated from the torch and that the water cooling helps to moderate the ablation. Optimal operating conditions for the system were deter- mined by simplex optimization using a program obtained from Professor Ebdon's group at the University of Plym- outh Plymouth UK. The program performs an optimiza- tion for up to nine variables between limits for each variable established by the user. The program was used for the simultaneous optimization of plasma power plasma gas flow hydrogen flow and GC carrier flow.Since malathion or diazinon takes more than 20 min to elute this was a time consuming process using simplex (thirteen vertices were processed during this optimization study) but it would have taken longer if a univariate optimization for the same four variables was carried out. A summary of the optimum operating conditions is given in Table 1. The addition of the hydrogen gas to the torch resulted in a background intensity at 400 W that was less than 10000 counts s-l in intensity. Compared with the background signals displayed in Fig. 2 this represents a significant improvement. After optimiza- tion a determination of the figures of merit for phosphorus sulfur and chlorine was performed. A mixture of seven compounds was prepared in methanol.The seven compo- nents are chlorobenzene bromononane iodobenzene tri- ethyl phosphate chlorotoluene diazinon and malathion. The scans were collected using the TRA software supplied by VG instruments. This software collects data in either a peak jump or scan mode. The user selects the masses to be monitored prior to the analysis. When the data are Table 1 Optimum plasma operating conditions Instrument Generator Frequency/MHz Cavity Power/W He flow rate/ml min-' 3% Hydrogen flow rate/ml min-l Hydrogen (Oh) Pressurelmbar Torch cooling temperature/"C Sampler-skimmer distance/mm VG PlasmaQuad Micro-Now 420B 2450 Beenakker TMO 10 450 245 100 0.87 1.0 (0.75 Torr) 10 4 (6 mm normally)JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 573 collected the mass spectrometer collects a set of data and stores it in RAM.These data are collected using the peak jump mode of the software with a dwell time of 6 4 0 p for each point. With the seven masses selected this resulted in a data acquisition time of 0.46 s for each scan. These scans are called time slices and about 1000 time slices were collected for each chromatogram. The data are stored on disk in the form of integrated intensity after all of the time slices have been collected. The individual chromatograms were extracted from the binary data file by a QuickBASIC program called VGDATA written for this purpose by the author. In addition to the single-ion chromatograms a total ion chromatogram can be obtained by adding the intensities of the individual masses monitored.The total-ion chroma- togram was extracted using the same VGDATA program. Fig. 4 shows the total-ion chromatogram that is obtained from a 32 ng injection of each of the seven components in the mixture. The injection masses given indicate the amount of the monitored element present in the injection. In this chromatogram the total ion beam chromatogram is displayed and the transfer line temperature was 325 "C. It is obvious from the peak shapes that there is a problem with the detection of compounds of high volatility. One possible explanation is a problem with the injection. Since a splitless injector was installed in the GC there was little adjustment that could be made in the injection process. Another problem could be in the transfer line.There is a large stainless-steel surface in the connector between the transfer line and torch on which some decomposition might take place at elevated temperatures. The temperature of the transfer line was decreased to 250 "C and the injection repeated. The results of this injection are shown in Fig. 5. As can be seen in comparison to Fig. 4 the peaks for the more volatile components are much sharper and roughly equal in intensity to the peak heights for the other components. There is a second problem that needs to be addressed in this chromatogram. The intensity of all seven components should be equal since equal amounts of the elements monitored were injected. At first glance only six compounds might be noticed in Fig. 5. The seventh component is iodobenzene which is much lower in inten- sity than the other compounds.It was necessary to investigate the reason for this discrepancy. The ion lenses in the VG PlasmaQuad spectrometer do not deliver a unifcim response across the mlz range. The response curve is somewhat humpbacked in appearance when solution nebulization is used with an argon induc- tively coupled plasma (ICP). The masses with the greatest sensitivity are from rnlz 100- 130. At either end of the range 120 ' - I u) 100' 2 8 0 c 3 n C I ? F I 0 1 2 3 4 5 6 Time/min Fig. 4 Chromatogram of a seven component mixture with a 32 ng injection of each component A chlorotoluene; B iodobenzene; C bromononane; D triethyl phosphate; E chloronapthalene; F diazinon; and G malathion. Transfer line temperature 325 "C 400 1 I I A c F 0 1 2 3 4 5 6 Time/min Fig.5 Chromatogram of seven component mixture with a 32 ng injection of each component A chlorotoluene; B iodobenzene; C bromononane; D triethyl phosphate; E chloronapthalene; F diazinon; and G malathion. Transfer line temperature 250 "C 500 lB - 400 4 in u) C c 300 0 m 2 2 200 \ .- in al c - 100 0 Time/min Fig. 6 Chromatogram of seven component mixture with a 32 ng injection of each component with ion lenses tuned at m/z= 127 A chlorotoluene; B iodobenzene; C bromononane; D triethyl phosphate; E chloronapthalene; F diazinon; and G malathion. the response falls off. This mass bias is not a problem with routine operation as the response difference is not that great and the calibration curve will normalize any differ- ences.When a dry helium MIP is used for the excitation source the result is a plasma where the kinetic energies of the ions are different from those in a wet argon ICP. This difference in kinetic energy leads to a much greater mass bias effect in an experiment such as this. With this in mind it was necessary to confirm that the decreased intensity for iodine was due to the mass bias effect. Without continuous solution introduction it is difficult to optimize the ion lenses to a particular mass. It is easy to accomplish this at mlz 31 because one simply needs to optimize the intensity on the ever present background signal. For iodine at rnlz 127 there is no significant background signal to optimize the ion-lens tuning. To obtain a steady signal for optimiza- tion less than 1 pl of neat iodobenzene was introduced into the gas chromatograph.The steady-state signal was more than adequate for optimizing the ion lens tuning on rnlz 127. After 2 h the background intensity had finally decreased to a level where an injection of 32 ng was possible. The total-ion chromatogram obtained with this new set of ion lens tuning conditions is shown in Fig. 6. As can easily be seen the iodobenzene peak is the most intense peak in this chromatogram. If one wanted to use a helium plasma for multi-element detection over a broad mass574 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY7 JUNE 1993 VOL. 8 I l Total I\ ion Time - Fig. 7 Single-ion chromatograms from seven component mixture with a 32 ng injection of each component range with equal sensitivity the mass bias effect would need to be minimized. One of the advantages offered by plasma MS is the ability to monitor a number of mlz values simultaneously.In the case of an unknown sample it might be possible in a single injection to identify a number of different elements. An example of this can be seen in the single-ion chromatograms extracted from the seven component mixture. Fig. 7 displays the individual-ion chromatograms for the masses monitored. Two features should be noted from these chromatograms. The first is that for elements with more than one isotope it is possible to confirm the presence of that element by looking at the isotope ratios obtained from the scan. It can be seen that for chlorine 35 and 37 the 3:l relationship is clear.In the case of sulfur its major isotope is at m/z 32 which is severely interfered with by the presence of 02+ and it is possible to select another isotope with less interference. This is analogous to selecting a different line in optical spectroscopy. The second feature to be seen in this chromatogram is the high selectivity possible with mass spectrometric detection. There is no hint of the chlorine peaks in the sulfur chromatogram at m/z 34 even though chlorine at m/z 35 is only 1 u away. Rapid monitoring of many different elements and confirmation by the determination of isotope ratios is a powerful aspect of the mass spectrometer. Using the peak jump TRA software figures of merit were determined for two different elements simultaneously. A summary of these results is displayed in Table 2.The detection limits are determined by the level of signal detected at three standard deviations above 600 1 0 2 4 6 8 10 12 Ti me/m i n Fig. 8 Supelco pesiticide mixture with 80 ng of each compound injected. Upper chromatogram single-ion chromatogram of m/z= 35 (chlorine signal). Lower chromatogram single-ion chro- matogram of m/z= 34 (sulfur signal) x 15 background (n= 10). The detection limits reported are for absolute amounts of element injected. The detection limit obtained for malathion is higher than the limits reported for the other phosphorus-containing compounds because mala- thion tends to adsorb to the sides of the glassware used to prepare the samples resulting in a decreased detection limit at low concentrations. Generally speaking the detection limits available with this technique are in the 20-100 pg range.While these limits are somewhat higher than MS detection limits for most metals it must be kept in mind that these elements are some of the most difficult to determine by ICP-MS and these detection limits with the reduced pressure helium MIP are better than could other- wise be obtained by conventional ICP-MS. The peak widths at half maximum peak height are approximately 10 s giving detectable amounts of 2-10 pg s-l. Of course the analysis of 'real' samples is always the end goal of this type of research. Toward this end a standard pesticide mix was obtained from Supelco. This mixture is representative of many of the compounds analysed by the US Environmental Protection Agency (EPA) SW-846 Method 8080 for the determination of chlorinated pesti- cides.This mixture was shipped as 2000 ng ml-l. After dilution to 80 ng ml-l in methanol the mixture was injected and analysed. Since the percentage of chlorine in each compound is slightly different the relative peak heights will vary somewhat. The W l and 34S single-ion chromatograms for this injection are shown in Fig. 8. The poor peak shape for the later eluting compounds can be attributed to the inability of the gas chromatograph to heat the oven to the maximum temperature needed for this particular analysis. It can be seen that the sensitivity of this method for chlorine is easily in the picogram region. There are three different endosulfan compounds in this mixture and with the multi- Table 2 Figures of merit for phosphorus and chlorine Detection limits*/pg Pesticide Element Slope (log-log) r2 (n= 10) Triethyl phosphate P 1.1866:3 0.98 19 86 Diazinon P 1.20830 0.9962 97 Malathion P 1.2698'3 0.991 5 229 Chlorotoluene c1 1.1852'7 0.9990 22 Chloronapthalene c1 1.0873.3 0.9986 36 * Detection limits calculated from smallest sample injection on calibration curve.RSD=4.7% based on peak area for three replicate 8 ng injections.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE element detection abilities of the mass spectrometer it is a simple matter to pick out which of the peaks correspond to the endosulfan compounds by looking at the m/z 34 chromatogram. The sulfur 34S chromatogram is located below the total ion chromatogram in Fig. 8. Based on the isotope abundance of 34S and the percentage of sulfur in these three compounds the amount of sulfur detected is approximately 350 pg.A calculation of the isotope ratio for chlorine can be used to confirm that chlorine is actually being detected. The values for the chlorine isotope ratios calculated from the first peak of the chromatograms displayed in Fig. 8 gives a value of 2.94 1 for 35C1:37C1 versus the 3.02:l actual value. Conclusions In comparison to the non-water cooled quartz torches used in the previous experiments the results for the helium plasma system are greatly improved. The addition of hydrogen as a reagent gas has helped to reduce the background intensity and reduce the amount of phosphorus reaction that occurs. This system shows good sensitivity for the determination of phosphorus sulfur chlorine and bromine in GC eluates.The possibility exists that even greater sensitivity can be achieved for organotins and other organometallic compounds. These experiments serve to highlight the need for a better design of the ion lenses to reduce the mass bias effect that makes the determination of iodine or other high relative molecular mass elements in the same scan as chlorine very difficult. The analysis of multi- element mixtures demonstrates the ability of the mass spectrometer to determine isotope ratios on GC peaks with excellent selectivity. It is conceivable that with some changes in ion-lens design and plasma configuration commercial plasma mass spectrometric GC detection for non-metals will be a common analytical tool in the not too distant future.1993 VOL. 8 575 The authors are grateful to the National Institute of Environmental Health Sciences for support of this work through grants numbered ESO 3221 and ESO 04908. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 References Story W. C. Olson L. K. Shen W.-l. Creed J. T. and Caruso J. A. J. Anal. At. Spectrom. 1989 5 467. Montasser A. and Golightly D. W. Inductively Coupled Plasmas in Analytical Atomic Spectrometry VCH Weinheim 1987 ch. 13. Bulska E. J. Anal. At. Spectrom. 1992 7 201. McCormack A. J. Tong S. C. and Cooke W. C. Anal. Chem. 1965,37 1470. Bache C. A. and Lisk D. J. Anal. Chem. 1965 37 1477. Bache C. A. and Lisk D. J. Anal. Chem. 1966 38 783. Bache C. A. and Lisk D. J. Anal. Chem. 1966 38 1757. Bache C. A. and Lisk D. J. Anal. Chem. 1967,39 787. Riviere B. Mermet J.-M. and Deruaz D. J. Anal. At. Spectrom. 1987 2 705. Riviere B. Mermet J.-M. and Deruaz D. J. Anal. At. Spectrom. 1989 4 5 19. Satzger R. D. Fricke F. L. Brown P. G. and Caruso J. A. Spectrochim. Acta Part B 1987 42 705. Mohamad A. H. Creed J. T. Davidson T. M. and Caruso J. A. Appl. Spectrosc. 1989 43 1127. Creed J. T. Davidson T. M. Shen W.-l. Brown P. G. and Caruso J. A. Spectrochim. Acta Part B 1989 44 909. Creed J. T. Davidson T. M. Shen W.4 and Caruso J. A. J. Anal. At. Spectrom. 1990 5 109. Quimby B. D. and Sullivan J. J. Anal. Chem. 1990 62 1027. Barnes R. M. and Reszke E. E. Anal. Chem. 1990,62,2650. Paper 2/04864G Received September 9 I992 Accepted February 9 I992

ISSN:0267-9477    

DOI:10.1039/JA9930800571

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 577 Determination of Trace and Ultra-trace Amounts of Germanium in Environmental Samples by Preconcentration in a Graphite Furnace Using a Flow Injection Hydride Generation Technique Guanhong Tao* and Zhaolun Fangt Flow Injection Analysis Research Centre Institute of Applied Ecology Chinese Academy of Sciences Box 4 1 7 1 1001 5 Shenyang China A method was developed for the trace and ultra-trace determination of germanium by flow injection (FI) hydride generation with subsequent trapping and electrothermal atomization in a graphite furnace pre-coated with palladium. The samples were propelled into the FI system to be merged on-line with a tetrahydroborate reductant. The hydride and hydrogen gases evolved were separated from the liquid phase in a gas-liquid separator and transferred to a palladium-coated graphite furnace pre-heated to 400 "C to collect the analyte which was then atomized at 2500 "C.The sensitivities and interference effects using different concentrations of hydrochloric acid were compared. The tolerance to interferences was improved at high acidities of 3 mol I-' HCI and also in the FI system when compared with a batch in situ collection system. With the reported system germanium was determined at a sampling frequency of 18 h-' with a detection limit (34 of 0.004 pg I-' in 0.15 mol I-' HCI medium and 0.03 pg I-' in 3 mol I-' HCI medium using a 4.5 ml sample. The precision (relative standard deviation) was 2.0% at the 0.3 pg I-' level in 0.15 mol I-' HCI medium and 2.5% at the 1.5 pg I-' level in 3 mol I-' HCI medium.The proposed method was applied to the determination of germanium in garlic tap water ginseng and geological reference samples. Keywords Germanium; in situ preconcentration flow injection hydride generation; electrothermal atomic absorption spectrometry environmental sample Recently the trace or ultra-trace determination of german- ium in environmental samples particularly of a biological nature has become increasingly of interest owing to its potential function in the retardation of human ageing and inhibition of growth of cancer ce1ls.l Reported methods used for the determination of germanium include spectro- photometry and atomic absorption spectrometry.2-21 The most often applied spectrophotometric method is that using phenylfluorone ~ i t h ~ ? ~ or without4 a surface-active reagent. A detection limit of 48 pg 1-1 was reported in a recent application of the m e t h ~ d .~ Germanium shows relatively poor sensitivity in flame atomic absorption spectrometry (FAAS) methods owing to its tendency to form very stable oxide species in the flame,6 whereas electrothermal atomic absorption spectrometry (ETAAS) methods often suffer from low sensitivity caused by losses of volatile germanium compounds (e.g. GeO and GeCl,) during the ashing ~ t a g e . ~ * ~ Through the use of various modifiers characteristic masses of 8-43 pg have been r e p ~ r t e d ~ - l ~ corresponding to a characteristic concentration of 0.4-2 pg 1-* with a 20 pl sample which is not sufficiently sensitive for the determi- nation of germanium in most biological samples.Although hydride generation atomic absorption spectrometry (HGAAS) has proved to be a sensitive method for most hydride-forming elements the sensitivity of HGAAS proce- dures for germanium is exceptionally poor compared with those for other hydride-forming elements.I4 The detection limits achieved are in the range 0.01-0.5 mg 1-1 despite different attempts to improve the atomization conditions including the use of a heated quartz cell15 and dinitrogen oxide-acetylene,16 arg~n-hydrogenl~ and nitrogen-hydro- gen18 flames and also using inductively coupled plasma atomic emission spe~trometry.'~ Therefore the relative sensitivities are no better than those obtained by ETAAS. More recently it was demonstrated that hydride-forming elements can be collected on the surface of a heated graphite tube pre-coated with palladium before introduc- *On leave from the Research Centre for Eco-Environmental ?To whom correspondence should be addressed.Sciences Chinese Academy of Sciences Beijing China. tion of the gaseous hydrides to improve greatly the detection limits in the determination of hydride-forming elements by ETAAS.20-24 Zhang et a1.22 optimized the operating conditions and successfully applied this in situ preconcentration technique to the determination of ger- manium and other hydride-forming elements in geological materials and natural water samples (detection limits were not reported). Doidge et al.23 employed a palladium and Triton X- 100 modifier which was reported to improve both the sensitivity and reproducibility of the determination of germanium owing to the even distribution of palladium over the tube surface.A detection limit of 4.5 ng 1-1 of germanium in the buffered sample volume (3.7 ml) was achieved with a precision of 3.5- 10% [relative standard deviation (RSD)] but no applications of the method to real samples were reported. Haug and Ju24 applied this method to the analysis of a synthetic solution that contained germanium and gallium in NaCl-NaOH solution. A detec- tion limit of 3 ng 1-l in 10 ml of buffered sample was obtained with an RSD of 2-3%. In the above studies either b a t c h - w i ~ e ~ ~ ? ~ ~ or continuous-flow22 approaches were used for the hydride generation. Flow injection (FI) techniques have been shown to be capable of significantly enhancing the performance of HGAAS methods by providing higher sample throughputs better selectivities higher absolute sensitivities and large savings in reagent c o n s ~ m p t i o n .~ ~ ~ ~ ~ In this work an attempt was made to implement FI-HGAAS techniques in the in situ preconcentration of germanium in a graphite tube to produce a more automated system and to improve further the performance of the in situ preconcentration procedure particularly in applying the method to the determination of ultra-trace amounts of germanium in environmental samples. Experimental Apparatus A Perkin-Elmer Model 2 100 atomic absorption spectro- meter with deuterium-arc background corrector equipped with a germanium hollow cathode lamp operated at 7 mA,578 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 NaBH Sample Table 1 Furnace temperature for in situ preconcentration and atomization 24 4.5 - w - - - -- w - ~-- 4.5 -.-.-.. Step Temperature/"C Ramp time/s 1 . Drying 90 10 120 10 2. Pyrolysis/pre-treatment 800 10 3. Collection 400 10 4. Atomization 2 500 0 5. Cleaning 2650 I P1 Go Inner argon Hold time/s flow rate/ml min-' 20 300 20 300 20 300 80 100 5 0 5 300 v t - was used with a Perkin-Elmer (Uberlingen Germany) Model HGA-700 graphite furnace. A wavelength of 265.2 nm was used with a bandpass of 0.7 nm. Pyrolytic graphite coated polycrystalline graphite tubes were used with simi- larly coated L'vov platforms. The recommended graphite furnace temperature programme is given in Table 1.Integrated absorbance (peak area) was used for evaluating the results throughout this work. The analyte absorption peaks were recorded by high-resolution graphics and printed out using an EX-800 printer (Epson Japan). ( a ) 1.5% ( b ) 1.5% u P2 stop cm cm HCI = 1 P2 Go Fig. 1 Flow injection hydride generation manifold for ETAAS in situ preconcentration (a) collection; and (6) washing and sample changing. V Injector valve; P1 and P2 peristaltic pumps; W waste; SP gas-liquid separator; L reaction coil; and GF graphite furnace A Perkin-Elmer Model FIAS-200 FI system with an injector valve having five channels on the stator and four channels on the rotor was used with the hydride generation accessory. The FI manifold for the hydride generation is shown in Fig.1. The rotation speed of the two multi- channel peristaltic pumps their stop and go intervals and the actuation of the injector were programmed on and automatically controlled by a separate PC computer (Epson Model 4030 1 A) independent of the spectrometer. Tygon pump tubing was used to deliver all solutions. All reaction coils and connections were made with 0.7 mm i.d. poly(tet- rafluoroethylene) (PTFE) tubing (Zhaofa Research Institute for Laboratory Automation Shenyang China). A Chemi- fold plastic connector block (Perkin-Elmer) furnished with W-confluence points was used for merging of the reagent and carrier stream. The glass gas-liquid separator was of W-configuration (4 cm high x 1 cm i.d.) and half-filled with glass beads of 3 mm diameter.The outlet of the separator consisting of a 50 cm length of PTFE tubing was connected to a 5 cm x 1.5 mm o.d.xO.5 mm i.d. quartz capillary with a short piece of silicone-rubber tubing. The quartz capillary was loaded on a laboratory-made rotating arm that could be used to swing the tip of the capillary into the sampling port of the graphite tube and functioned as a hydride introduction probe during the in situ collection (preconcentration) stage. The HG reaction vessel for the batch method was constructed from a 25 cm x 15 mm i.d. test-tube furnished with a rubber stopper that was fitted with a small rubber septum. Two PTFE tubes were inserted through the stopper one serving as the argon carrier gas inlet reaching to the bottom of the test-tube and the other as the gas outlet via the quartz capillary to the graphite furnace extending a short length of 5 mm beyond the stopper.The reductant was introduced through the rubber septum using a hypoder- mic syringe into the reaction vessel containing the acidified sample and argon was allowed to bubble through the reaction mixture to release the hydrides. Reagents A 1000 mg 1 - I germanium stock standard solution was prepared by dissolving 0.1442 g of GeOz in 10 ml of 1 mol I-* sodium hydroxide solution and diluted to 100 ml .with 3 mol 1-I hydrochloric acid. A series of working :standard solutions were prepared by two-stage dilutions of the stock standard solution. Sodium tetrahydroborate solution (0.5-3.0% m/v) was freshly prepared daily by dissolving NaBH (Merck Darms- tadt Germany) in 0.05 mol 1-I sodium hydroxide solution and filtered before use. This concentration of base was usually sufficient to keep the reductant stable for 1 d.Nitric acid and hydrochloric acid were of ultra-pure reagent grade (Beijing Chemical Factory Beijing China) ;and demineralized water was used throughout. All other chemicals were of analytical-reagent grade. No blank signalsJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 579 were detectable in any of the reagents under the conditions used in this study. Palladium solution 500 mg 1 - I was prepared by dissolv- ing 0.1660 g of palladium chloride in 10 ml of nitric acid and diluting to 100 ml with de-mineralized water. Procedures The graphite furnace temperature programme in Table 1 also shows the various stages of operation for the in situ collection.Before each hydride generation cycle 20 pl of 500 mg 1-' Pd were pipetted into the furnace and dried. The furnace was heated to 800 "C and kept constant at that temperature for 20 s to complete the pre-treatment then cooled to 400 "C the deposition temperature of the analyte. For the introduction of the first sample the FI system was actuated for 20 s in the position shown in Fig. l(b) simultaneously with the initiation of the drying cycle and then remained in the standby mode (both pumps stopped) until the beginning of the collection sequence. For all subsequent samples the FI system was kept in the standby mode during the entire period before initiation of hydride collection. After the furnace temperature had reached 400 "C in about 10 s the tip of the quartz capillary probe was inserted through the sample introduction port at the centre of the graphite tube by manually swinging over the arm on which the probe was fixed and was kept in contact with the opposite internal wall throughout the collection period.The FI system was actuated immediately after insertion of the probe with the injector valve in the position shown in Fig. l(a). This initiated the collection sequence during which the acidified sample (4.5 ml min-I) was merged with the tetrahydroborate reductant (4.5 ml min-I) both propelled by pump P 1. The reaction mixture passed through a 50 cm length of reaction coil and was guided into the gas-liquid separator where the waste was pumped out at a flow rate at least double that of the total inflow rate.The germane formed was separated and transferred by the argon carrier and generated hydrogen gas (or the latter alone when argon was not used) into the palladium-coated graphite tube and germanium was sorbed on the tube walls following decomposition. After a 60 s generation (collec- tion) period the valve was automatically actuated to the washing and sample change position. A hydrochloric acid rinsing solution propelled by pump P2 was run for 10 s to expel the residual sample in the reaction coil to the separator in order to prepare the system for the next sample and to collect the remaining fraction of germane evolved from the residual sample after the termination of the sampling period. Meanwhile the next sample was intro- duced into the sample uptake lines.The FI system was stopped automatically at this point. After another interval of 10 s for the withdrawal of the quartz probe from the graphite tube giving a total of 80 s at the 400 "C collection temperature the graphite furance was heated to the atomization temperature at 2500 "C using maximum power with interrupted internal argon flow and the absorbance signals were recorded. Finally the furnace temperature was increased to 2650 "C for 3 s to clean the furnace. For the batch hydride generation procedure after intro- ducing a defined volume (4.5 ml) of the acidified sample solution the reaction vessel was sealed with the stopper and 4.5 ml of tetrahydroborate reductant were injected through the septum with a syringe.A reaction and transfer time of 60 s was used to sweep the hydride vapour from the reaction vessel to the graphite tube after the initiation of the hydride generation reaction. A steady argon flow of 100 ml min-' was maintained during the process. The furnace temperature programme was the same as that used in the FI-HGETAAS system. Sample Pre-treatment Geological samples (0.500 g) were treated with 5 ml of a nitric acid-hydrofluoric acid mixture (7 + 3 v/v) in PTFE beakers. The vessels were heated on a hot-plate at 140 "C and gently boiled nearly to dryness. A further 2 ml of nitric acid were added and the solutions were again taken nearly to dryness. After cooling the digests were diluted to 50 ml. A 1 ml volume was taken and diluted further to 100 ml with 0.15 mol 1 - I HCl.Dried garlic or ginseng samples (1 .OOO g) were soaked with 5 ml of an HN03-HF mixture (7+3 v/v) in PTFE beakers for 5-6 h and then gently heated on a hot-plate keeping the temperature below 140 "C to near dryness. A further 5 ml of nitric acid were added and the solution was taken nearly to dryness again. After cooling the garlic digests were diluted to 50 ml with 0.15 moll-' HCl and the ginseng digests to 50 ml with 3 mol 1 - I HCl. Tap water samples were filtered and acidified to 0.15 mol 1 - I in HC1 with 3 mol 1 - I HCl. Method Development The operational parameters of the FI preconcentration system (including the furnace collection temperature and conditions for furnace coating) were optimized indepen- dently of the ETAAS atomization procedure.An atomiza- tion temperature of 2500 "C as recommended by the instrument manufacturer was used as a preliminary value during the optimization of the FI parameters and later was further optimized under the optimized FI conditions using a univariate method. The integrated absorbance was taken as the main figure of merit in the optimization with regard to precision. For the FI-HG in situ collection system a univariate approach was also used for optimization. Preli- minary efforts using simplex optimization were discourag- ing owing to within-day variations in the reactivity of the tetrahydroborate reductant. The parameters optimized included sample acidity tetrahydroborate concentration flow rates of the reaction mixture and stripping gas reaction coil length and collection temperature.The ranges of parameters studied were determined both from those often reported in batch HG in situ collection and FI-HG procedures and from practical limitations mainly associ- ated with effectiveness in gas-liquid separation. Medium values of the parameter ranges were used in preliminary univariate studies and the parameters were gradually adjusted to close-to-optimum values for the final univariate studies. The figures of merit used for the optimization of the FI- HG system was mainly integrated absorbance as the principle objective of the study was to develop a sensitive method for the determination of germanium at low concen- tration levels. However the tolerance of interferences often encountered in environmental samples and the precision of the results were also taken into consideration. The precision of the method was estimated using 1 1 replicates of standard solutions at medium and high concentration levels within the analytical range.The sample pre-treatment step was not included in the precision studies. Calibration The FI-HG in situ collection ETAAS system was calibrated using each five standards in the range 0.10-0.50 pg 1-l for a 0.15 mol 1-1 HC1 medium and 0.5-2.5 pg 1 - I for a 3 mol 1-1 HCl medium. Calibration points were based on the aver- ages of triplicate measurements of the standards and blanks. The calibration graph fitting was effected by linear regression.5 80 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Results and Discussion FI Manifold Design The FI manifold used in this study ( c f Fig.1) differs considerably from previous designs used for FI-HGAAS systems without trapping. The analytical readouts of HGAAS systems without trapping of hydrides are directly associated with transient signals produced by the hydride generation reaction while the readouts (usually peak height) are little influenced by the absolute amount of sample introduced above 0.5 ml. For the FI-HG in situ collection system studied here the situation is almost exactly the opposite in that output signals generated by the hydride generation reaction are not monitored directly while readouts are directly influenced by the amount of sample introduced during the trapping sequence. Such differences are reflected in the design of the manifold in that time-based sampling was used instead of volume-based sampling in order to improve the sensitivity and detection limits by increasing the amount of sample introduced in a defined time period.This approach also facilitated adjust- ment of the analytical range through sample volume variations to meet various demands by changing the sampling period under a defined sample flow rate. In order to avoid carryover when using the time-based approach a hydrochloric acid rinsing solution propelled by a separate pump P2 was used to expel the residual sample in the reaction coil to the separator and transfer the residual germane to the graphite tube following the termination of the sampling period. The concentration of the hydrochloric acid was made identical with that in the sample.The rinsing pump was stopped during the collection period in order to save acid. Optimization of Experimental Parameters for the FI In Situ Collection System Acidity conditions The efficiency of the generation of germane is strongly dependent on the acidity at which the reaction is performed and also on the acid species. Various acidity conditions for the hydride generation determination of germanium have been proposed by different ~ ~ r k e r ~ . ~ ~ ~ ~ * ~ ~ Thompson and Pahla~anpour~~ observed a distinct maximum of the analyt- ical signal with approximately 0.1 mol 1-l HCl similar results also being obtained with other acids. Castillo et found that a 0.1 mol 1-1 solution of acetic acid and sodium acetate (with a pH of 4.75) gave optimum sensitivity for the generation of germane.Hydrochloric acid at a concentra- tion of 3 mol 1-l was used as the reaction medium by Zhang et a1.22 In this work an attempt was made to investigate the effect of HC1 acidity up to a concentration of 4 moll-'. The curve in Fig. 2 plotted from integrated absorbance ob- tained under different acidities shows a maximum between 0.10 and 0.18 moll-'. This observation was similar to that reported by Thompson et aL2' However the procedures with low acidities were reportedly less tolerant to interfer- ences than those with higher acidities.29 This phenomenon has been confirmed also under the FI conditions used in this study following an investigation of the performances with two concentrations of HCl (0.15 and 3 mol 1-l) and discussed later under Interferences.The low-acidity conditions producing optimum sensitiv- ity were used for samples with relatively simple matrices such as tap or surface water samples. For samples with relatively complex matrices such as ginseng the 3 rnol 1-I HCl medium was required to achieve satisfactory recover- ies with some sacrifice of sensitivity. The reduced sensitiv- ity with higher acidities is at least partially due to the lower reaction rate under such conditions (see under Reaction coil length) leading to decreased collection efficiencies (see v) 0.10 -z 0 C Q 0.08 e 2 0.06 0 t a 2 0.04 w + = 0.02 I 1 O ' Oll [HCII/mol I-' Fig. 2 Effect of hydrochloric acid concentrations on the german- ium signal expressed as integrated absorbance. The flow rates of both sample and reductant were 2.5 ml min-I.The concentration of Ge and NaBH were 0.5 pg 1-I and 1 .Oo/o m/v respectively. Other FI conditions are as shown in Fig. 1 0.12 0.08 v) 5 0.04 C Q e s o n 0.16 W c E 2 0.12 a C - 0.08 0.04 I ~ 0 0.5 1 1.5 2 2.5 3 3.5 [NaBH,] (% m/v) Fig. 3 Effect of NaBH concentrations on the germanium signal for (a) 0.3 pg I-' of Ge in 0.15 rnol I-' HCI; and (b) 1.5 pg I-' of Ge in 3 rnol I-' HCI. Other FI conditions are as shown in Fig. 1 under Analytical Performance). In the 3 mol 1-1 HCI medium apparently a large percentage of the analyte is pumped to waste before being converted into the hydride. Tetra h ydrobora te concentration The relationships between the tetrahydroborate concentra- tion and the peak areas obtained at 0.15 moll-' HC1 acidity using 0.3 pg 1-1 Ge under FI conditions shown in Fig.1 are demonstrated in Fig. 3(a). The sodium hydroxide concen- tration (0.05 mol 1-l) in this study was kept constant for different reductant concentrations. As the peak areas remained almost constant when the tetrahydroborate con- centration was higher than 1.25% 1.5% NaBH was used in subsequent studies. As shown in Fig. 3(b) for samples in 3 mol 1-' HC1 matrices the relationships between the reductant concen- tration and the peak areas were different from those in 0.15 mol 1-l HCl. Although the sensitivity increased with anJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 58 1 increase in concentration of tetrahydroborate the reaction became so violent when this concentration was higher than 2% that gas-liquid separation could no longer be ensured.Therefore 1.5% NaBH solution was also used in subse- quent studies. The gas-liquid separation might be im- proved at higher tetrahydroborate concentrations by using larger separators but this was not investigated in this study. Flow rate parameters for sample reagent and carrier gas The effect of varying the sample and reagent flow rates on the integrated absorbance was investigated while keeping the samp1e:reagent flow rate ratio (1:I) and sampling time (60 s) constant (Fig. 4). The sample (carrier):reductant flow rate ratio used in this study was lower than that normally used in FI-HGAAS systems (5 l) because higher tetrahy- droborate concentrations are required for the generation of germane.The lower ratio was used to provide sufficient flexibility in the investigation of higher reductant concen- trations for optimization studies. Ratios lower than 1 1 were not attempted in order to avoid excessive dilution of the sample. As expected for both 0.15 and 3 mol I-' HCl media increasing the flow rate resulted in increases in the integrated absorbance. However owing to the introduction of larger sample volumes the reaction became more violent at higher flow rates and some liquid drops were brought into the hot graphite tube by the generated gases deteriorat- ing both the repeatability and the lifetime of the graphite tube. In order to shorten the collection time to increase the sample throughput and at the same time ensure effective gas-liquid separation a total sample and reductant flow rate of 9.0 ml min-' (sample reductant ratio I I ) was used for both acid media.In most FI-HGAAS methods that are based on peak- height measurements carrier gases are used to strip and transport the hydrides into the The flow rate and flow stability of the carrier gas usually have a significant effect on the sensitivity and repeatability of the method. In a preliminary stage of this work argon carrier gas was also 0.12 0.08 v) \ $ 0.04 C Q e 5 o n Q 0.16 c (21 L a g 0.12 - 0.08 0.04 I " 2 4 6 8 10 12 Total flow rate/ml min-' Fig. 4 Effect of sample and NaBH total flow rates on the germanium signal for (a) 0.3 pg I-' of Ge in 0.15 mol I-' HCl; and (b) 1.5 pg 1 - I of Ge in 3 mol I-' HCI (sample:reductant 1 1). Other FI conditions are as shown in Fig.1 used for the in situ preconcentration system and the effect of its flow rate was investigated. However the results showed that effects from the carrier gas flow rate were negligible. Presumably this is because the amount of hydrogen generated during the reaction was sufficient for transporting the germane into the furnace and the readouts were not based on transient signals produced during the hydride generation reaction but on integrated readings from the electrothermal atomization process. The repeatabiliy and height/width of the transient peak forms from the HG process (which are not monitored) are not important as long as the sample volumes and reaction conditions remain constant and the hydrides are introduced into the graphite furnace without loss.Therefore in subsequent studies no carrier gas was used. This produced about a 10% higher sensitivity apart from a saving in the argon gas supply without noticeable effects on the precision and sample throughput. The increase in sensitivity without the carrier gas might be the result of the prolonged time of interaction of the hydride with the palladium surface. Reaction coil length The effect of the reaction coil length investigated over the range 10- 120 cm is shown in Fig. 5. The shortest length of 10 cm corresponded to the tube length connecting the reagent merging point and the gas-liquid separator. For the 0. I5 moll-' HCl reaction medium under the experimental conditions shown in Fig. 1 the response remained almost constant beyond 50 cm implying that the reaction time was sufficient beyond this point. As shorter reaction times are beneficial for suppressing interfering reactions that are slower than the main reaction a coil length of 50 cm was chosen for the generation of the hydride to enhance the selectivity.With a 3 mol I-* HCl reaction medium the hydride generation reaction was slower and the sensitivity might be improved by using longer reaction coils [Fig. 5(6)]; however for the sake of convenience the same coil length as that used for lower acidity was applied in this study and more detailed investigations on the kinetics of the hydride 0.12 0.08 -$ 0.04 0 C m e 2 0 20 40 60 80 100 m 0.12 r 1 L 0 E 0.08 0.04 L 1 I 1 1 1 0 20 40 60 80 100 120 Length of reaction coil/cm Fig. 5 Effect of the lengths of reaction coil on the integrated absorbance for (a) 0.3 pg I-' of Ge in 0.15 rnol I-' HCI; and (6) 1.5 pg I-' of Ge in 3 rnol I-' HCI. Other conditions are as shown in Fig.1582 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 0.12 0.10 v) \ C Q n L 0.08 n 0.06 0.04 0.02 5 0 c. CI - - - - - - - 0 500 1000 1500 2000 2500 Tern peraturePC Fig. 6 Effect of sorption and atomization temperatures on the integrated absorbance of germanium generation reaction under high acidity conditions which appear to be complicated will be made in a later study. Optimization of temperature parameters for ETAAS The optimum operating conditions for ETAAS are given in Table 1. Deposition of the analyte occurs at relatively low temperatures and as mentioned in the introduction the efficiency of collection is enhanced by coating the graphite tube with palladium.In this study the effects of germane collection using 20 pl of 100-1000 mg 1 - I Pd for coating were investigated. Concentrations above 200 mg 1-1 were found to be sufficient for trapping and stabilizing the collected product (identical absorption signals were ob- tained above this concentration). A concentration of 500 mg 1-1 of Pd was used in subsequent studies. The relationships between the integrated absorbance and the thermal sorption/atomization temperatures are shown in Fig. 6. The analyte was effectively absorbed on the palladium-treated graphite surface over the temperature range 300- 1000 "C which presumably reflects the break- down and surface interaction temperature of the hydride.The optimum atomization temperature was shown to be in the range 2300-2500 "C. A temperature of 400 "C is recommended for sorption and 2500 "C for atomization owing to the better precision. Interferences Interference studies for two reaction media were performed for some ionic species frequently reported to create interfer- ences in HGAAS methods. The results are given in Tables 2 and 3. In this study up to 0.4 mg 1-l of Cull and 0.03 mg 1-l of Nil1 could be tolerated without any significant adverse effect on the integrated absorbance using 0.15 mol 1-1 HCl (Table 2). Other potential interferents such as Fell Fell1 and CoI1 did not interfere seriously in the hydride generation process up to concentrations of 3.0 3.0 and 1.5 mg l-l respectively. Because the optimum acidity for the generation of stannane was about 0.1 mol l-l SnlI was the most serious interferent among the hydride-forming species studied but still tolerated up to a concentration of 0.06 pg I-'.Presumably the interference from SnlI originates either in the gas or in the liquid phase owing to competition for the reductant. Using 3 mol 1-' HC1 the transition metal ions investigated produced no significant effects on the recovery up to relatively high concentrations of at least several pg 1-l. The effect of acidity on nickel is particularly dramatic as can be seen in Fig. 7. The tolerance for this interferent has Table 2 Effect of potential interferents on the recovery of germanium in 0.15 mol 1-1 HCl reaction medium.All samples contained 0.3 ,ug 1-1 of germanium. All results are averages of three measurements with 1-2% RSD Concentration/ Interferent mg 1 - l Nil1 0.03 0.05 0.1 CU" 0.4 0.6 1.2 2.4 CO" 0.6 1.5 4.5 Recovery (W 101 93 81 99 90 85 75 95 96 75 Fell 3.0 102 10.0 73 Fell1 3.0 96 6.0 75 ZnI1 1.5 4.5 A P 0.1 0.5 AsV 1.6 Sn1I 0.03 0.06 0.15 Bill1 0.06 1 .o 95 85 96 70 101 98 92 80 93 89 SelV 0.25 98 0.5 92 SeV1 0.06 100 1 .o 97 Sb1I1 0.03 98 0.06 94 1 .o 85 improved by two orders of magnitude with the higher acid concentration. The higher tolerances for transition metals at higher acid concentrations is presumably due to a reduction in the local formation of the interfering precipitate associated with these metals. However the interference from all hydride- forming species studied was enhanced at the higher acidity probably because more tetrahydroborate is consumed in the production of hydrogen at the higher acidity and competi- tion from hydride-forming species in reacting with the reductant becomes more pronounced.However the con- centration of the tetrahydroborate could not be increased further because of difficulties in achieving effective gas- liquid separation. Arsenic and selenium create much stronger interferences in the lower valency states for selenium particularly at higher acidities. However the latter acid conditions are used only for samples with complex matrices which require an acid digestion after which selenium should be converted into the higher valency state. Judging from the tolerated concentrations of other hydride-forming species and their concentration ranges in most environmental samples serious interferences from such elements are not likely to be encountered in the analysis of real samples.When compared with the batch in situ collection system described under Experimental the FI hydride generation in situ preconcentration system exhibited decreased transition metal interferences in the determination of germanium. The interfering effects of nickel on hydride generation ofJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 583 Table 3 Effect of potential interferents on the recovery of germanium in 3 mol 1 - I HCl reaction medium. All samples contained 1.5 pg 1-* of germanium. All results are averages of three measurements with 1-2% RSD Concentration/ Recovery Interferent mg 1-I (O/O) Nil1 10.0 1 00 50.0 85 Cull 1.5 98 6.0 93 7.5 90 CO" 7.5 103 50.0 88 Fell 50.0 99 Fell1 60.0 98 SeIV 0.1 95 0.6 66 SeV1 7.5 101 SnlI 0.5 0.75 As"' 0.10 0.15 AsV 0.10 0.15 SblI1 0.10 0.15 96 90 95 81 97 86 96 86 Bit11 0.05 95 0.10 80 0.15 65 A I? 0001 0.01 0.1 1 10 100 [Nickell/rng I-' Fig.7 Comparison of interfering effects of nickel on hydride generation of germanium with a batch method using A 0.15 and C 3 mol I-' HCl and an FI system using B 0.1 5 and D 3 rnol I-' HCI germanium using batch and FI systems with 0.15 and 3 mol 1-1 HCl are shown in Fig. 7. The interference-free range is increased by about an order of magnitude using the FI approach for both acidities. A similar effect was also observed for CulI and Fell but not as distinct as with Nil1.~~ Table 4 Characteristic performance data of FI-HG in situ preconcentration ETAAS system I. 11. Acidity of sample Sample consumption Sampling frequency Calibration graph (0-0.5 pg 1-1) RSD Detection limit (30) Acidity of sample Sample consumption Sampling frequency Calibration graph (0-0.5 1-I) RSD Detection limit (30) 0.15 mol I-' HCl 4.5 ml 18 h-I A=-0.001+0.40~ 2.0% (n= 11) (0.3 pg 1 - I ) 2.1% ( n = l l ) (0.5 pg I-]) 0.004 pg 1-I 3 mol 1-I HCI 4.5 ml 18 h-I (r= 0.9996) A =0.007 + 0.067~ (r=0.9997) 2.5% (n= 11) (1.5 pg 1-I) 2.3% (n= 11) (2.0 pg 1-I) 0.03 pg 1-' The greater freedom from transition metal interferences in FI (compared with the batch system) might be due to a kinetic discrimination effect. The reaction time for the hydride generation process in the FI system is precisely controlled by the flow rate and line length so that the slower interfering reaction can be suppressed.Another reason for the greater tolerance of the FI system to interferences might be that the high flow rate of the sample and reagent through the reaction conduit and gas-liquid separator leaves little possibility for accumulating reduced metal or metal boride deposits therein. Such deposits are considered to be a main source of interferences in HGAAS systems by some workers. 17330*31 Analytical Performance Characteristic data for the performance of the FI-HG in situ preconcentration ETAAS system are summarized in Table 4. The characteristic mass of the FI and batch HG in situ collection methods are compared with that of the direct ETAAS method to evaluate the collection efficiency of the in situ collection procedures (Table 5) taking the efficiency of direct ETAAS as 100%.The collection efficiencies of the batch and FI methods were almost identical at the lower acidity reaching over 80°/o. The efficiency of the FI method at the higher acidity however was significantly lower than that of the batch procedure at similar acidities while both were lower than those obtained at lower acidities. Presum- ably this is due to the slower reaction rate at the higher acidity as discussed previously rather than to loss of analyte in the in situ collection itself. The higher efficiency of the batch method seems to be associated with the longer reaction time available. However even in batch systems under high acidities the reaction may not be complete before total dissociation of the reductant which explains the relative low efficiency (40%) of the batch procedure.Table 5 Characteristic mass and collection efficiency for batch and FI-HG in situ preconcentration ETAAS system FI hydride Batch hydride furnace furnace Conventional furnace* 0.18 mol 1 - I 3 mol 1 - I 0.18 rnol 1 - I 3 mol 1-' Characteristic mass/pg per 0.0044 s 40 49 295 47 101 Collection efficiency (Oh) 100 82 14 85 40 *Using the method described in the 1iteratu1-e~~ under the proposed conditions.584 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Table 6 Determination of germanium in geochemical reference samples using the FI-HG in situ preconcentration ETAAS system Found/ Recommended/ Material g-' Pg g-' GSR- 1 1.9f0.1 2.0 GSS- I I .3 + 0.I 1.3 GSS-5 2.5 4 0. I 2.6 *Mean values -t standard deviations (n= 3). Table 7 Determination of germanium in environmental samples using the FI-HG in situ preconcentration ETAAS system; results given in pg g-I except for tap water samples given in pg I-' Sample* Concentration Recovery (O/o)? Tap water 1 0.037 95+3 Tap water 2 0.0 10 9942 Garlic 0.0024 106k4 Ginseng 1 0.01 5 101 -t3 Ginseng 2 0.020 104k2 Ginseng 3 0.05 1 104+_2 *Tap water and garlic were analysed in 0.15 rnol I-' HCI medium and spiked with 0.1 p g I-' of germanium for recovery tests. For tap water 2 22.5 ml of sample were used. Ginseng samples were analysed in 3 rnol I-' HCI and spiked with 1 pg I-' of germanium in the digests for recovery tests.The tests were run in triplicate. ?Mean values f standard deviations (n= 3). The results of sample analyses are given in Tables 6 and 7. Determinations of germanium in ginseng were per- formed in a 3 mol 1-' HCl medium as satisfactory recoveries could not be obtained with 0.15 mol 1 - I HCl. Conclusions The combination of the flow injection technique with hydride generation in situ preconcentration ETAAS mainly using commercially available equipment produced a sensi- tive efficient and automated system that is particularly characterized by its improved tolerance to interferences. Such features are essential for the successful application of the method to the analysis of environmental samples of a biological nature which often have complicated matrices.Although the results here relate only to germanium the system can also be extended to the determination of other hydride-forming elements with minor modifications. The authors are grateful to Bodenseewerk Perkin-Elmer Uberlingen Germany for partial financial support to the Flow Injection Analysis Research Centre and for the loan of AAS and FI equipment and to Professor Ni Zheming Sun Lijing and Xu Shukun for valuable comments and discus- sions. 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 References Pang Z. Wang B. and Wei J. Yaowu Fenxi Zhazhi 1990 10 300. Burns D. T. and Dadgar D. Analyst 1980 105 75. Donaldson E. M. Talanta 1984 31 997. Aznbrez J. Moneo P. Vidad J. C. and Palacios F. Analyst 1985 110 747.Nukatsuka I. Takahashi K. Ohzeki K. and Ishida R. Analyst 1989 114 1473. Johnson D. J. West T. S. and Dagnall R. M. Anal. Chim. Acta 1973 67 79. Sohrin,Y. Isshiki K. and Kuwamoto T. Talanta 1987 34 341. Gao Y,-q. and Ni Z.-m. Acta Chim. Sin. 1982 40 1022. Chen Z.-s. and Chen Y.-l. Fenxi Huaxue 1991 19 1405. Mino Y. and Shimomura S. Anal. Chim. Acta 1979 107 253. Dittrich K. Mandry R. Mothes W. and Judelevic J. G. Analyst 1985 110 169. Kolb A. Muller-Vogt G. Wendl W. and Stoebel W. Spectrochim. Acta Part B 1987 42 951. Carnrick G. R. and Barnett W. B. At. Spectrosc. 1984 5 213. Nakahara T. Prog. Anal. At. Spectrosc. 1983 6 163. Thompson K. C. and Thomerson D. R. Analyst 1974 99 595. Halicz L. .4naI.vst 1985 110 943. Smith A. E. Analyst 1975 100 300. Hahn M. H. Mulligan K. J. Jackson M. E. and Caruso J. A. Anal. Chim. Acta 1980 118 115. Wolnik K. A. Fricke F. L. Hahn M. H. and Caruso J. A. Anal. Chem. 198 I 53 1030. Drasch G. Meyer L. and Kauert G. Fresenius' Z. Anal. Chem. 1980 304 14 1. 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. Acta Part B 1989 44 751. Doidge P. S. Sturman B. T. and Rettberg T. M. J. Anal. At. Spectrom. 1989 4 25 1. Haug H. O. and Ju C.-h. J. Anal. At. Spectrom. 1990 5 215. Fang Z . 4 in Flow Injection Atomic Spectroscopy ed. Burgu- era J. L. Marcel Dekker New York 1989 ch. 4. Tyson J. F. Spectrochim. Acta Rev. 1991 14 169. Thompson M. and Pahlavanpour B. Anal. Chim. Acta 1979 109 251. Castillo J. R. Lanaja J. and Aznarez J. Analyst 1982 107 89. Meyer A. Hoofer C. Tolg G. Raptis S. and Knapp G. Fresenius' Z. Anal. Chem. 1979 296 337. Kirkbright G. F. and Taddia M. Anal. Chim. Acta 1978 100 145. Bye R. Talanta 1986 33 705. Lin W.-y. He Y.-z. Luo J.-h. Huang L.-j. Lin K. Chen J.-h. and Wu S.-r. Fenxi Shiyanshi 1991 10 30. Paper 210562 7E Received October 21 I992 Accepted January 11 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800577

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 585 Examination of Separation Efficiencies of Mercury Vapour for Different Gas-Liquid Separators in Flow Injection Cold Vapour Atomic Absorption Spectrometry with Amalgam Preconcentration C. P. Hanna* P. E. Haigh and J. F. Tysonf Chemistry Department University of Massachusetts Amherst MA 01 003 USA S. Mclntosh Inorganic Analysis Division Perkin-Elmer Corporation 76 1 Main A venue Norwalk CT 06859 USA A comparison has been made of the separation efficiency of three designs of gas-liquid separator when used in a flow injection (FI) manifold for the determination of Hg by cold vapour atomic absorption spectrometry. The manifold used with each device was separately optimized for maximum sensitivity. This involved studies of the effects of reagent flow rates argon purge gas flow rate injection time and post-injection purge time.A significant difference with respect to both peak height and integrated signal sensitivity (by a factor of approximately 3) between the performance of a miniature design and that of two larger volume designs was obtained. No significant differences in precisions were observed. For the miniature design the use of either tetrahydroborate or tin@) reductant was investigated. No difference in peak height sensitivity was found but the integrated signal sensitivity for the tetrahydroborate was 36% lower. The efficiency of separation was measured by comparison of the signal obtained from a known mass of Hg vapour introduced via an amalgam preconcentration unit and the signal obtained from a known mass of Hg in solution introduced via the FI manifold and amalgam preconcentration unit.The efficiencies were found to be 101 k 4% and 103 k 6% for peak height and integrated signal respectively. Keywords Flow injection cold vapour atomic absorption spectrometry; gas-liquid separator; efficiency study The use of continuous flow (CF) and flow injection (FI) analysis for the determination of Hg by cold vapour atomic absorption spectrometry (CVAAS) has been the subject of study for a number of yea~s.I-'~ The various methods described can differ greatly with little conformity between them. While some methods determine the total Hg present in a sample,1-3,5,8,11,12,16,17 others appear to render only information about the amounts of inorganic Hg in the ~ a r n p l e .~ * ~ J ~ J ~ Some investigators use tin(@ as the re- ductant,1-s17*8,11-16 while others use sodium or potassium tetrahydr~borate.~-~J~J~ Sodium or potassium tetrahydro- borate is presumably used because of the rapid reaction kinetics and the ability to reduce organic Hg compounds to elemental Hg. However recent findings have shown that sodium tetrahydroborate does not reduce all organic Hg compounds to the same extent.I8 One component of CF- and FI-CVAAS that demonstrates virtually no conformity is that of the gas-liquid separator (GLS) employed in the system. Some practitioners of FI- CVAAS have utilized a microporous poly(tetrafluor0ethy- lene) (PTFE) membrane material as a diffusion medium for the separation of the Hg vapour from ~ o l u t i o n .~ * ~ J ~ While these membrane separators have yielded excellent results their mechanical stability and resilience over time in addition to their uniformity in composition are areas that require attention before wide-scale acceptance of them is achieved. Most investigators of CF- and FI-CVAAS have used some sort of open chamber (typically made of glass) in which the reaction products are separated by employing an inert purge gas. The designs of the separators are as variable as are the methods described. For example investigators have used devices which range from miniaturized Vijan-type U- tubesI0 and open chambers into which reaction products and purge gas are added separately,I4 to chambers in which the flow of reaction products is directed to the surface of a sintered glass frit for p~rging.~ The systems that employ *Present address Inorganic Analysis Division Perkin-Elmer TTo whom correspondence should be addressed.Corporation 761 Main Avenue Norwalk CT 06859 USA. these GLSs all exhibit varying degrees of sensitivity with no apparent agreement being reached as to the optimum GLS design. Only two investigations have been made into the comparison of GLS designs for their effect on separa- tion capability in which PTFE membrane separation in FI hydride generation for inductively coupled plasma atomic emission detection was e~amined.'~.*~ It has been n~ted,~JO however and is generally agreed upon that the reduction of the internal volume of the GLS is of greater importance in FI systems than in CF systems.Since FI involves the injection of a discrete sample a large volume GLS will lead to greater dispersion of the analyte zone prior to detection whereas a CF analyte zone will ultimately reach a maxim- ized steady state regardless of GLS volume. A recent development in CF- and FI-CVAAS technology has been that of an amalgam accessory for the determina- tion of Hg. The principle behind the amalgam accessory is that the liberated Hg vapour is trapped on the surface of gold-platinum gauze or gold-covered sand packed into a quartz tube. This packed area of the quartz tube is then rapidly heated and the released Hg vapour is conducted to an atomic spectrometer for detection. Less than 50 pg of Hg can be detected in an optimized system when AAS is applied.*' Since the kinetics of Hg desorption are consistent from one heating cycle to the next the signal then becomes dependent only upon the mass of Hg introduced onto the trapping surface and is independent of the kinetic processes occurring before the trapping surface.Thus when the amalgam accessory is used in conjunction with CF- and FI- CVAAS procedures the sensitivity is dependent only upon the efficiency of separation achieved with the GLS being used. Therefore large gas-phase dilution factors for the Hg vapour due to large volume GLSs or high purge gas flow rates can effectively be reversed. Maximum sensitivity will then be approached if a particular GLS approaches 100% separation efficiency. The aim of this work is to illustrate the variation of separation efficiencies for Hg vapour when three GLSs of different design are used.It is hypothesized that if a given GLS achieves 100% separation efficiency using an optim-586 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 I + AAS RM ized manifold with amalgam preconcentration the sensitiv- ity obtained is maximized. By using amalgam preconcentra- tion the kinetic process occurring in the manifold which would otherwise lead to a decrease in sensitivity are now decoupled from the atomic spectrometer. Manifold condi- tions for each GLS were optimized prior to comparisons. The most efficient conditions with the most efficient GLS were then used to determine the separation efficiency relative to a source of Hg vapour of known vapour phase concentration.- AAS Experimental Instrumentation An FIAS 200 flow injection atomic spectrometry system supplied by Perkin-Elmer was used throughout the study. This system consisted of pumps pump tubing connecting tubing (1.0 mm i.d. Teflon) injection valve manifold connections and argon regulation (0-250 ml min-I). This system was used in conjunction with a Perkin-Elmer 3100 atomic absorption spectrometer. A hollow cathode lamp (Perkin-Elmer) drawing 6 mA of current was used as the atomic line source and the spectrometer was tuned to the 253.7 nm line with a spectral bandpass of 0.7 nm and a low slit setting. A Perkin-Elmer amalgam system was used for the trapping of the liberated Hg vapour. This system supplies both an argon purge gas for the manifold and an argon carrier gas for the desorbed Hg.The trapping medium is a 1.2 cm long plug of rolled gold-platinum gauze inserted into a length of quartz tube (0.3 cm i.d.). The gold gauze was rapidly heated by activating two 10 W tungsten filament lamps facing one another around the gauze. The system was cooled by compressed air which was delivered under controlled timing. Analysis parameters such as pump speed argon flow rate purge time and injection time were controlled through Perkin-Elmer FIAS software run on a Digital DECStation PC. Data collection and data treatment were also controlled through the Perkin-Elmer FIAS software. RM Reagents All solutions were prepared with distilled de-ionized water produced by an E-Pure System (Barnstead). Hydrochloric acid carrier stream solutions were prepared by diluting an appropriate amount of concentrated hydrochloric acid (ACS grade 36.5% m/m Fisher Scientific) to concentra- tions expressed as O/o v/v with distilled de-ionized water.Tin(n) chloride reductant solution ( 10% m/v) was prepared L W (b) Ar 7 by dissolving 50.0 g of tin(@ chloride dihydrate (Fisher Scientific) in 50 ml of concentrated hydrochloric acid and diluting to 500 ml with distilled de-ionized water. Sodium tetrahydroborate reductant solution ( 1 O/o m/v) was prepared by dissolving 5.0 g of sodium tetrahydroborate powder (Fisher Scientific) in 500 ml of 0.05% m/v sodium hydrox- ide (from pellets Fisher Scientific) solution. These reduc- tant solutions were purged with argon for 30 min prior to use. Standard solutions were prepared by diluting a 1000 mg 1-' standard solution of Hg" (Fisher Scientific) to a concentration of 20 ng ml-".This standard was preserved with 10% hydrochloric acid or with 0.5% nitric acid-0.005% potassium dichromate. Gas-Liquid Separators Three differently designed gas-liquid separators were cho- sen for examination. The first GLS was supplied by Perkin- Elmer in the FIAS 200 system for Hg and hydride generation analysis procedures (referred to as PE). This GLS is shown in Fig. l(a). The PE consists of a cylindrical chamber of 0.8 cm i.d. and 3.0 cm height one third filled with 0.3 cm diameter glass beads. The reaction products and purging argon enter through a side arm with the argon and Hg passing through an opening in the top of the GLS and the spent liquid being pumped to waste through a second side arm.The second and third GLSs were obtained from PS Analytical (Sevenoaks Kent UK). One was a GLS designed specifically for sodium tetrahydroborate reductions in hydride generation and Hg cold vapour generation22 (re- ferred to as PSAl) and the other was designed for tin@) chloride reductions exclusively in Hg cold vapour genera- tion (referred to as PSA2). The PSAl is shown in Fig. l(b) and PSA2 is shown in Fig. l(c). The PSAl consists of a cylindrical chamber with a U-tube drain attached at the bottom so that a constant level of solution is sustained in the chamber. Two glass tubes feed into the chamber one for reaction products and one for the argon purge gas. The PSA2 is similar to PSAl in its cylindrical chamber and U- tube design. However the reaction products in PSA2 enter through a side arm of the chamber and a glass tube for the argon purge opens below the surface of the liquid.Procedure Each GLS was optimized separately from maximum separ- ation efficiency which was determined as the maximum signal arising from applying the amalgam trapping acces- sory. Each GLS was put in line with a constant manifold shown in Fig. 2 and the variables of reagent flow rate argon Ar-l Fig. 1 Design of the three gas-liquid separators (GLSs) examined (a) Perkin-Elmer GLS designed for sodium tetrahydroborate and tin(I1) chloride reductions; (b) PS Analytical GLS designed for sodium tetrahydroborate reductions; and (c) PS Analytical GLS designed for tin@) chloride reductions. W = waste and RM =reaction mixtureJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 587 I Amalgam- Sample AAS Ar W U Fig. 2 Manifold used for examination of GLS efficiencies. The broken lines indicate slight modifications in the manifold to accommodate the PE GLS since argon is added directly to the manifold for this GLS and not for the others purge flow rate injection time (the amount of time the injection valve remained in the 'inject' position with the argon purge gas activated TJ and post-injection purge time (the amount of time after the injection valve was returned to the 'load' position and the argon purge continued TJ were optimized by applying a single cycle alternating variable search procedure. The effect of reagent flow rate and argon purge flow rate were also examined for their effect on the signal without the amalgam trapping acces- sory using the PE GLS. This part of the optimization utilized 10% v/v hydrochloric acid as the standard solution preservative and as the carrier stream and 10% m/v tin(@ as the reductant.Once the optimum conditions were determined for each GLS they were compared directly with one another at their respective optimized conditions. The most efficient GLS was then used to examine the effects of carrier acidity on the signal and the blank values the contribution by the standard solution preservative to the signal and the blank values and how the signal obtained with this GLS compared with that obtained from a known mass of Hg vapour introduced into the amalgam system. This known mass was introduced by using the apparatus 0.5 0.4 - 0.3 P) m E 0.2 e 0.1 m c r - 3 0 1 1 I 1 I i y 0.20 n 5 m C 0.15 0.10 0.05 0 40 60 80 100 120 Ar Au-Pt Hg syringe auze Fig.3 Design of apparatus used for introducing a known mass of Hg vapour to the amalgam system. E 125 ml Erlenmeyer flask; I 2 mm i.d. PTFE inlet tube for pressure equilibration; T ther- mometer; and L 500 pl sample loop. This loop is filled with the mercury-saturated headspace of the Erlenmeyer by retracting the plunger of a 5 ml syringe. This known mass of mercury vapour is then injected into an argon carrier and then onto the Au-Pt gauze shown in Fig. 3 which is similar to systems previously d e s ~ r i b e d . ~ ~ - ~ ~ The effect of the type of reductant used was then examined by employing 1% m/v sodium tetrahydro- borate solution with the optimized conditions determined above using the most efficient of the three GLSs and comparing the efficiency obtained with that achieved using the tin@) reductant.Results and Discussion Effect of Reagent Flow Rate For all of the GLSs examined the speed of the pump used for propelling the reagents was varied from 40 to 120 rev min-l (5-15 ml min-l total flow rate) with the argon flow rate constant at 250 ml min-l injection time constant at 25 s and the post-injection purge time constant at 30 s. 2.5 I 1 l e 5 t 0 I I I I I I 0.4 0.2 - 40 60 80 100 120 Pump speedhev min-' Fig. 4 Effect of pump speed on the efficiency of separation for a 500 pl sample of 20 ng ml-I of Hg". Argon flow is constant at 250 ml min-' Ti is constant at 25 s and Tp is constant at 30 s (a) peak height measurements and (6) integrated signal.A PE signal; B PE blank; C PSA 1 signal; D PSA 1 blank; E PSA2 signal; and F PSA2 blank. The effect of pump speed when the amalgam accessory is not used with the PE GLS is also shown for (c) peak height measurements and (6) integrated signal588 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 0.5 0.4 0.3 Q g 0.2 e 0.1 0 E .- m o m !%I 100 150 200 250 300 50 100 150 200 250 0 Ar flow 1 1 I I 50 100 150 200 250 rate/ml min-' Fig. 5 Effect of argon purge flow rate on the efficiency of separation for a 500 pl sample of 20 ng ml-I of Hg". Pump speed is constant at 40 rev min-1 (5 ml min-I) Ti is constant at 25 s and Tp is constant at 30 s (a) peak height measurements and (b) integrated measurements.A PE signal; B PE blank; C PSAl signal; D PSAl blank; E PSA2 signal; and F PSA2 blank. The effect of argon purge flow rate when the amalgam accessory is not used with the PE GLS is also shown for (c) peak height measurements and (d) integrated signal For all three GLSs examined both peak height and integrated signal sensitivity increased with decreased pump speed while the associated blank values decreased for decreased pump speed. This trend is shown in Fig. 4(a) and (b). These observations can be explained in two ways (i) for a fixed time the decreased pump speed allows for an increased residence time of the sample in the manifold prior to separation thus the reduction of Hgll to Hg has more time to go to completion and more Hg vapour is liberated as an end result; and (ii) for a fixed time the decreased pump speed results in less background Hg from the carrier and reductant streams entering the GLS thus decreasing blank values. When the amalgam accessory is not used and the reagent flow rate is varied in the same range the signal is effected in an entirely different way as shown in Fig.4(c) and (4. While the integrated signal remains about the same throughout the range peak height shows a substantial decrease for lower reagent flow rates. The increased dispersion of the sample zone at lower reagent flow rates thus leads to a shorter signal that is more spread out over time. This effect is therefore effectively reduced by using amalgamation prior to detection. Effect of Argon Purge Flow Rate For all of the GLS examined the effect of argon purge flow rate was examined from 50 to 250 ml min-I.The pump speed was constant at 40 rev min-I injection time was constant at 25 s and post-injection purge time was constant at 30 s. The maximum argon purge flow rate examined was 250 ml min-l for two reasons (i) due to back-pressure limitations the requirement of adding argon to the mani- fold and not the GLS for the PE GLS resulted in a maximum argon flow rate of 250 ml min-l; and (ii) and argon flow rate greater than 250 ml min-' in the two PSA GLSs resulted in liquid being violently forced out of the draining end of the U-tube resulting in virtually no reaction products collecting in the GLS chambers. For all of the GLSs peak height and integrated signal sensitivity increased with increased argon flow rate as did the blank values obtained.This trend is shown in Fig. 5(a) and (b). While this argon increase led to a decrease in sensitivity for a system not using amalgam trapping due to increased analyte zone dispersion in the gas phase and decreased residence time in the atom cell Fig. 5(c) and (4 this is reversed by trapping and refocusing the analyte zone prior to desorption and detection. Effect of Injection Time (K) The effect of the length of time the injection valve was kept in the 'inject' position Ti was investigated to ensure that the entire sample zone was emerging from the sample loop. The manufacturer's recommended T of 25 s ( Tp= 30 s total purge time 55 s) and a Ti of 50 s (Tp=5 s total purge time 55 s) were investigated.It was found that there was virtually no difference in the signal blank or blank-subtracted signal for increasing T . It was decided to leave the value of Ti at 25 s. Effect of Post-injection Purge Time (2'') The effect of post-injection purge time Tp was examined from 5 to 30 s for the PE and PSAl GLS and from 2 to 30 s for the PSA2 GLS. While there was a general decrease in both signal and blank values obtained for decreased Tp there was a maximum in the blank-subtracted signals for each GLS. Thus an optimized Tp corresponds to a time where the maximum amount of the sample zone is entering the GLS with a minimum amount of the Hg background- containing carrier stream. The PE GLS produced a blank- subtracted signal maximum for a Tp of 5- 10 s and the PSA2 GLS produced a maximum at Tp= 10 s.However the PSAl GLS had a higher T for maximum blank-subtracted sensitivity at 30 s presumably due to the longer time period required for the complete emergence of the sample zone into the GLS because of the glass tube which extends into the PSAl GLS chamber. This trend is shown in Fig. 6. Comparison of PE PSAl and PSA2 GLSs The three GLSs examined were directly compared with one another at their respective optimized parameters of pumpJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE >. c .- .L 100 c .- va S al E 90 0 a¶ r Y a. .- 80 a¶ .- c - 70 a al .- 4- - d B - a" 7 0 ' I I I 1 0 10 20 30 TP Fig. 6 Effect of T on the efficiency of separation for a 500 p1 sample of 20 ng ml-1 of Hg". Pump speed is constant at 40 rev min-I (5 ml min-') Ti is constant at 25 s and argon flow is constant at 250 ml rnin-'.The maximum blank-corrected signal for each GLS is expressed as 100% relative sensitivity (a) peak height measurements and (b) integrated signal for A PE; By PSA I ; and C PSA2 0 2 4 6 8 1 0 1 2 HCI carrier (% v/v) Fig. 7 Effect of carrier acidity on the efficiency of separation for a 500 pl sample of 20 ng ml-1 of Hg" using the PE GLS. Pump speed is constant at 40 rev rnin-' (5 ml min-I) Ti is constant at 25 s T is constant at 10 s and argon flow is constant at 250 ml min-I. The maximum blank-corrected signals for A peak height; and By integrated absorbance are expressed as 100% relative sensitivity speed argon flow and Tp. The blank-corrected signals for each GLS demonstrate that the PE GLS is 2.75 times more efficient than the PSAl GLS and is 3.20 times more efficient than the PSA2 GLS.The associated blank-cor- rected signals obtained and precision data are shown in Table 1. Effect of Carrier Acidity The PE GLS was used to examine the effect of varying carrier acidity from 0 (distilled de-ionized water) to 10% v/v hydrochloric acid. While there was a steady decrease in the blank values obtained for decreased carrier acidity the signal remained constant and then dropped slightly for a water carrier. This trend resulted in a maximum blank- subtracted sensitivity for a 1 % hydrochloric acid carrier. This trend is shown in Fig. 7. 1993 VOL. 8 589 Table 1 Comparison of signals obtained with three different GLSs for optimized conditions of reagent and argon flow rates and injection and post-injection purge times. The injected sample was 500 pl of 20 ng ml-1 of Hg1I in 10% v/v hydrochloric acid.The carrier was 10% v/v hydrochloric acid and the reductant was 10% m/v tin(@ chloride in 10% v/v hydrochloric acid. All peak heights and integrated signals shown are blank-corrected Peak height Integrated signal GLS Mean RSD (O/O) Meads RSD (O/O) PE 0.3506 1.4 (n=5) 1.7406 1.2 (n=5) PSA 1 0.1275 1.8 (n=4) 0.6320 2.8 (n=4) PSA2 0.1095 1.4 (n=4) 0.5436 2.1 (n=4) Effect of Standard Solution Preservative Used The results for the variation of blank signal as a function of carrier acidity indicated that the standard solution preser- vative employed might be a source of excess background Hg.A Hg standard solution preservative described by Welz et all7 involves adding 1 ml of a 50% nitric acid-0.5% K2Cr207 solution to every 100 ml of aqueous Hg standard giving a preservative concentration of 0.5% nitric acid and 0.005% K2Cr207 at the point of analysis. This resulted in no decrease in signal values and a slight decrease in blank values compared with the signals and blanks obtained with the 10% hydrochloric acid preservative and 1% hydro- chloric acid carrier. Effect of Reductant Type Used The 10% m/v tin@) solution reductant employed for the optimization procedure was replaced with 1 Yo m/v sodium tetrahydroborate solution to examine its effect on separa- tion efficiency. It should be noted that a different manifold of identical design to that shown in Fig.1 was used. This was necessary since the introduction of sodium tetrahydro- borate into a manifold used for tin(@ reductions will result in the precipitation of elemental tin. Using the optimized parameters found with tin@) there was no difference in peak height sensitivity relative to the tin(@ reductant when the sodium tetrahydroborate reductant was used with the PE GLS. However there was a 36% decrease in integrated signal sensitivity for the sodium tetrahydroborate reductant compared with that obtained with the tin@) reductant for the PE GLS. A major difference between the use of tetrahydroborate and the use of tin is that the acid decomposition of the excess tetrahydroborate generates copious amounts of hydrogen. It is possible that the constant formation of hydrogen even with the argon purge deactivated leads to greater overall pressure in the system.This would result in decreased residence time of the Hg vapour in the atom cell and a decrease in integrated signal for the sodium tetrahydroborate reductant without neces- sarily leading to a decrease in peak height. Thus it it difficult to make direct comparisons between these two reductants due to the presence of the hydrogen. While the peak height sensitivity might have been unchanged it has been noted26 that using sodium tetrahy- droborate in determinations of Hg with amalgam concen- tration can lead to poisoning of the trapping medium from adsorption of metal hydrides generated from background elements (i.e. arsenic and selenium) in the sample.This results in fewer surface active sites for the liberated Hg vapour to form an amalgamation thus leading to lower trapping efficiency and lower sensitivity. It has also been noted27 that the presence of transition metals [specifically copper(rr)] can lead to depression of the H g signal when using sodium tetrahydroborate as a reductant. It was590 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Table 2 Optimized conditions for Hg vapour generation using the PE GLS by flow injection analysis. All flow rates shown are in ml min-I. The sample was 500 p1 of 20 ng ml-I of Hg". The optimized injection time (T,) was 25 s and the optimized post- injection purge time (T,) was 10 s Reagent Concentration Flow rate Ar purge - 250 Carrier 1% v/v HCl 3.5 Reductan t 10% m/v SnCl 1.5 Standard in 10% v/v HCl preservative 0.005% m/v K2CrZ07 - in 0.5% v/v HN03 deduced that the elemental Hg adsorbs on the surface of the finely divided copper precipitate that is also formed in the reduction process.For optimum sensitivity and the relative freedom from interference effects compared with sodium tetrahydroborate tin(@ was determined to be the better reductant. Optimized parameters are shown in Table 2. Comparison of the PE GLS Using a Calculated Mass of Hg With Amalgam Preconcentration The apparatus shown in Fig. 3 was used to introduce a calculated mass of Hg vapour into the gold-platinum gauze prior to thermal desorption and detection. By making injections of the Hg-saturated air onto the gold gauze along with the analysis of argon blank values the system is calibrated in terms of the mass of Hg introduced.The mass of Hg introduced is calculated by using the data for Hg vapour pressures reported by Weast et al.28 At a tempera- ture of 24 "C 500 pl of saturated air contains 9.2 ng of Hg vapour. A 500 pl sample of 20 ng ml-l of HgI1 (10.0 ng) was then injected into the optimized FI-CVAAS system with an amalgam concentration unit with the signals and the blanks being measured. The results obtained from the system containing the amalgam trap calibrated by the introduction of Hg vapour of calculated mass demon- strated that the PE GLS was 103+6% (95% confidence interval) efficient by integrated signal measurements and 10 1 2 4% (95% confidence interval) efficient by peak height measurements. Conclusions It has been shown under optimized conditions that the efficiency of Hg vapour separation in FI-CVAAS is variable and is dependent upon the design of the gas-liquid separator.The most efficient gas-liquid separator examined in this study was shown to achieve complete separation based upon a Hg vapour mass calibration of an amalgam concentration accessory. As it is known that the measures taken in the optimization process to ensure maximum separation efficiency (e.g. lower reagent flow rates higher argon purge flow rates) would lead to a decrease in sensitivity if amalgam preconcentration was not used the sensitivity of an FI-CVAAS manifold using amalgam preconcentration reaches a maximum value which is based upon the efficiency of the vapour separation process.It is thus apparent that the amalgam preconcentration process successfully decouples the kinetics of the FI-CVAAS mani- fold from the atomic spectrometer. This study also shows that a decrease in internal volume of the gas-liquid separator does not necessarily lead to incomplete gas-liquid separation as some workers have Financial support for this work and the provision of equipment by The Perkin-Elmer Corporation is gratefully acknowledged. 1 2 3 4 5 6 7 8 9 10 I 1 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 References El-Awady A. A. Miller R. B. and Carter M. J. Anal. Chem. 1976 48 110. Agemian H. and Chau A. S. Y. Anal. Chem. 1978 50 13. Jirka A. M. and Carter M. J. Anal. Chem. 1978 50 91. Oda C. E. and Ingle J. E. Jr. Anal. Chem. 1981 53 2030. Goto M.Shibakawa T. Arita T. and Ishii D. Anal. Chim. Acta 1982 140 179. de Andrade J. C. Pasquini C. Baccan N. and Van Loon J. C. Spectrochim. Acta Part B 1983 38 1329. Morita H. Kimoto T. and Shimomura S. Anal. Lett. 1983 16 1187. Anderson P. J. At. Spectrosc. 1984 5 101. Fang Z. Xu S. Wang X. and Zhang S. Anal. Chim. Acta 1986 179 325. Fang Z. Zhu Z. Zhang S. Xu S. Guo L. and Sun L. Anal. Chim. Acta 1988 214 41. Goto M. Munaf E. and Ishii D. Fresenius' 2. Anal. Chem. 1988,332 745. Birnie S. E. J. Automatic Chem. 1988 10 140. Pasquini C. Jardim W. F. and de Faria L. C. J. Automatic Chem. 1988 10 188. Nakahara T. Kawakami K. and Wasa T. Chem. Express 1988 3 651. Nakahara T. and Wasa T. Microchem. J. 1990 41 148. Munaf E. Takeuchi T. and Haraguchi H. Fresenius' 2. Anal. Chem. 1992 342 154. Welz B. Tsalev D. L. and Sperling M. Anal. Chim. Acta 1992 261 91. Baxter D. C. and Frech W. Anal. Chim. Acta 1990,236 377. Barnes R. M. and Wang X. J. Anal. At. Spectrom. 1988 3 1083. Wang X. and Barnes R. M. J. Anal. At. Spectrom. 1988 3 1091. Welz B. Melcher M. Sinemum H. W. and Maier D. At. Spectrosc. 1984 5 37. Thompson M. Pahlavanpour B. Walton S. J. and Kirk- bright G. F. Analyst 1978 103 568. Kaseke C. T. M.Sc. Thesis Loughborough University of Technology UK 1977. Dumarey R. Temmerman E. Dams R. and Hoste J. Anal. Chim. Acta 1985 170 337. Temmerman E. Vandecasteele C. Vermeir G. Leyman R. and Dams R. Anal. Chim. Acta 1990 236 371. Welz B. and Schubert-Jacobs M. Fresenius' 2. Anal. Chem. 1988 331 324. McIntosh S. unpublished data. CRC Handbook of Chemistry and Physics ed. Weast R. C. CRC Press Cleveland 52nd edn. 197 1. Fang Z. in Flow Injection Atomic Spectroscopy ed. Burguera J. L. Marcel Dekker New York 1989 ch. 4. Paper 2/06 7 79J Received December 12 1992 AcceDted Januarv 22. 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800585

出版商:RSC    

年代:1993

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Determination of Total Mercury by Single-stage Cold Vapour Atomic Spectrometric Detection Lian Liang and Nicolas S. Bloom* Brooks Rand Ltd. 3950 6th Avenue NW Seattle WA 98107 USA 59 1 Gold Amalgamation With ~~~ ~ The two-stage gold amalgamation technique with elemental mercury vapour detection was compared with a technique employing a single gold trap. When peak area was measured and special attention paid to the gold trap orientation the one-stage amalgamation procedure provided the same precision accuracy and detection limits as the benchmark two-stage technique (Fitzgerald and Gill 1979). The overall analysis time however was reduced from about 10 to 2 min per sample. An absolute detection limit (20 of trap blanks) of less than 1 pg of HgO was attained using an atomic fluorescence detector while relative standard deviations of 2.8 1.5 and 2.2% were obtained for 500 1000 and 2000 pg of Hg respectively.Recoveries of close to 100% with RSDs of t3% were obtained in the determination of certified reference materials intercalibration hair samples and lakewater using this more rapid procedure. Keywords Total mercury determination; gold amalgamation preconcentration; signal integration; cold vapour atomic spectrometric detection Two-stage gold amalgamation with gas-phase detection is a sensitive precise and accurate technique which has been successfully used in the determination of mercury in a variety of matrices. Atmospheric mercury is collected and concentrated onto a gold trap called the ‘field’ or ‘sampling’ trap and then transferred by thermal desorption to a second generally smaller trap called the ‘analytical’ or ‘permanent’ trap.The first trap can be considered as a part of the chemical separation technique while the second trap is a permanent component of the detector. Mercury in aqueous samples and solid-substrate digestates is isolated using SnCl or NaBH reduction and gas-phase stripping with pre-collection onto the first The mercury is finally introduced into the spectrometer by thermal desorp- tion into a suitable carrier-gas stream which transports it to the detection cell. Any one of several cold vapour spectro- metric techniques can be used including atomic absorp- tion1-496*7 atomic fluore~cence~~~ microwave plasma emis- sion9J0 or phofoacoustk1lJ2 The dual amalgamation technique has the following major advantage which has made it the standard technique for ultra-low level mercury analysis over the past two decades all samples are finally introduced into the spectro- meter from a single well characterized trap.This results in extraordinary precision and accuracy. Although the same is achievable with the single-trap procedure,6 this necessitates the use of only one gold trap/bubbler per analyser resulting in a rate of analysis which is determined by the sample purge time. Either peak height or peak area analysis can be used and neither the size nor the orientation of the sampling trap has any effect upon the ultimate signal characteristics. Using two-stage amalgamation however a new sample can be introduced in-line only when the analytical trap is cooled to room temperature. The total length of time necessary to analyse a sample once the mercury is collected on the initial trap is approximately 6-10 min depending upon trap heating and cooling conditions.Since in the past 20 years signal processing techniques have progressed greatly making low-cost peak integration available to most laboratories the use of the single-stage amalgamation method instead of the two-stage amalgama- tion technique should be investigated. In this work two- stage and single-stage gold amalgamation were compared. It was found that peak height results varied between indivi- dual traps by greater than +50% while peak areas were * Present address Frontier Geosciences Inc.8057 Corliss Avenue North Seattle WA 98 103 USA. identical within +2% when special attention was paid to the orientation of the trap in the analytical scheme. For best results and single sharp peaks the gold trap should be oriented with respect to the gas flow direction in the same manner as during sample collection (z.e.’ upon thermal desorption the gas carries the mercury past the length of the trap before release to the detector). With these considera- tions single-stage amalgamation reduces analytical time to approximately 2 min with no sacrifice in precision or accuracy. Experimental Instrumentation A cold vapour atomic fluorescence spectrometer was used as a mercury detector. The design and installation of this instrument is described in detail by Bloom and Fitzgerald.* The carrier gas was ultra-high purity argon maintained at a flow rate of 25 ml min-’. The photomultiplier tube high voltage was 450 V.A chromatographic peak integrator (Laboratory Data Control CI-4000) was used to provide peak area calculations. Peak heights were determined by manually measuring the recorded peak trace with a milli- metre ruler. In the configuration used the absolute detec- tion limit of the instrument was about 0.2 pg of HgO. Reagents and Materials A commercial mercury standard (1 000 mg 1-I) was diluted 1 +999 with 5% BrCl (ref. 2) in distilled deionized water (DDW) to form a working stock solution. From this a 10 pg 1-1 working standard solution was prepared by I +99 dilution with 5% BrCl in DDW. All standard dilutions were stored in Teflon bottles.Analytical-reagent grade SnCl (200 g) was dissolved in 200 ml of analytical-reagent grade HC1 and then diluted to 1.0 1 in a Teflon bottle. Prior to use mercury was removed from the solution by purging for several hours with mercury-free nitrogen at 500 ml min-l. A 150 ml flat-bottom bottle with a 24/40 tapered fitting served as the reaction vessel. A special 24/40 purge cap was used which allowed the purge gas to be fed into a central fritted tube and then exhausted through a side-port to the sampling trap (Fig. 1). A column (3 cm long by I cm i.d. Teflon) of 8-12 mesh soda-lime granules was placed at the outlet to remove potentially damaging bubbler solution droplets.2 A gold coated sand trap was placed at the bubbler inlet to remove traces of mercury from the purge gas.592 JOURNAL OF AlNALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL.8 ( a ) Purge gas in ( b ) A B 6 Carrier gas in To analyser 3 Heating wire Fig. 1 Orientation of the gold trap during (a) sample collection and (b) analysis to obtain repeatable results by the single-stage procedure The traps used for collecting mercury were prepared by packing a 2.5 cm length of gold-coated sand in the centre of a 10 cm long by 4.0 mm i.d. (6.4 mm 0.d.) quartz tube. The sand was held in place using plugs of quartz wool. The gilded sand' was prepared by sputter-coating acid-washed ashed (900 "C) quartz beach sand (40-80 mesh). Traps were initially conditioned and blanked by heating to 500 "C for 5 min with a flow of mercury-free nitrogen.All fittings and connectors were made using 0.125 or 0.250 in 0.d. Teflon fluorinated ethylene propylene (FEP) tubing with appropri- ate compression tubing fittings and adapters. All gases were of ultra-high purity further purified by gold-coated sand traps to remove traces of mercury. Water was 18 MQ de-ionized pre-treated by reverse osmosis. The water routinely contained less than 0.1 ng 1 - I of total mercury. Procedure Procedures for the determination of atmospheric mercury aqueous mercury and for the digestion of various solid matrices are described in detail el~ewhere.'-~ A comparison of the single- and two-stage amalgamation techniques using the following procedures is described here. Standards and samples to be analysed were added to the reaction vessel containing 50-100 ml of DDW with 5 ml of HCl and 0.5 ml of SnC1 to reduce the Hg" to HgO.The mercury vapour was then swept from solution with nitrogen at 450 ml min-l through the soda-lime trap onto the gold- coated sand (sampling) trap. For the two-stage amalgama- tion procedure the mercury collected on the sampling trap was transferred via thermal desorption at 450 "C (a very dull red glow after 2 min heating time) into the argon carrier gas and ultimately to the second (analytical) gold trap. The traps are heated using an 85 cm coil of 22 gauge Nichrome resistance wire at a potential of about 9.8 V a.c. In a similar manner mercury collected on the analytical trap is subse- quently desorbed into the detector cell. After each trap is heated for 2 min the coil is switched off and the trap cooled with an air blower.The next sample can be introduced only after the analytical trap has cooled to near room tempera- ture ( a 2 min). For the single-stage amalgamation the mercury collected on the sampling trap is released to the analyser directly by heating the trap for 2 min. After heating the trap can be removed as soon as it is cool enough to handle (X 1 min). The next sample can then be analysed immediately. Special attention must be paid to the trap orientation maintaining the same direction of flow during both sampling and analysis (Fig. 1). When using the single-stage amalgamation technique peak area integration must be used as the peak elution shape is a characteristic of the individual traps. Results and Discussion Effect of Trap Orientation on Peak Profile Three similar traps were used to collect 1000 pg standards by the purge and trap technique.The mercury was then thermally desorbed into the detector using a variety of trap orientations. As can be seen from Fig. 2 the direction in which the mercury enters and leaves the trap is critical to the determination of the peak characteristics. When the mercury is collected into the same end of the trap as was filled with gold and then eluted from that end a characteristic repeatable double peak is eluted [Fig. 2(a)]. It is not clear why this occurs but might have to do with stray gold being deposited on the walls during the filling procedure. If the mercury is collected into the trap from the t - Q C cn v) .- I a) 0.38 D 0.73 i b) 0.55 1.23 ! Time - Fig.2 Effect of gold trap orientation on the signal peak profile for aliquots of the same sample (a) sample collected into and released from the end in which the trap was filled with gold sand; (b) sample collected into and released from the trap end opposite from that in which gold sand was poured; and (c) the trap is oriented in the same way for both collection and analysis. Integrated areas are (a) 3039; (b) 3475; and (c) 3599 units. Elution time is given for each peak in secondsJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 6 ) 593 1.09 b t Time - Fig. 3 Effect of gold trap length on the signal peak profile for aliquots of the same sample (a) 5 cm of gold sand (area 567 1 units); and (b) 2.5 cm gold sand (area 55 19 units). Elution time is given for each peak in seconds opposite end a sharp single peak with very short elution time is observed [Fig.2(b)]. Finally if the mercury is collected from either side but eluted through the length of the trap and out of the end opposite to the collection end then a sharp clear peak of good definition is always obtained [Fig. 2(c)]. This peak is somewhat delayed com- pared with the others because it has to pass through the entire gold trap. Notice that all of the peaks resulted in similar areas although they differed dramatically in appearance. In spite of the delayed peak however passage of the mercury through the length of the trap results in the best repeatability. Effect of Trap Length on the Peak Profile Traps of two lengths of gold coated sand (2.5 and 5.0 cm) were compared for the determination of mercury in hair digestate [Health and Welfare Canada (HWC) interlabora- tory intercomparison samples]. Sample aliquots contained 1.14 ng of Hg in 175 pg of hair matrix.These profiles are illustrated in Fig,. 3. The 5.0 cm trap resulted in a considerably shorter broader peak than the 2.5 cm trap although the peak areas were otherwise indistinguishable (RSD=1.9%). In general the shorter trap should be preferred and perhaps the dimensions be even further optimized. The larger trap has a greater holding capacity for mercury and so might be preferable in some cases. Comparison of the Single- and Dual-trap Methods for Routine Samples Results obtained using the single and dual amalgamation methods were compared using both peak height and peak area for the determination of mercury in intercalibration hair sample digestates (Table 1).In this experiment the 2.5 cm traps were utilized. The peak area results for the single- trap method were found to be similar to those found using the ‘lab standard’ two-stage amalgamation procedure and near the centre of the HWC intercalibration study range of corrected mean and standard deviation. On the other hand the results calculated using peak height and the single-stage amalgamation procedure were often out of the range of the intercalibration mean and of much poorer repeatability. In a further comparison 15 water samples in a mercury concentration range of 0-9 ng 1-* collected from an urban mercury-contaminated lake were analysed by both methods following cold oxidation with BrCl (ref.2). The regression equation of single= 1.01 dual+O.OI (r2=0.994) indicates that the two methods result in essentially identical results with very high precision. In addition studies of spike recoveries were conducted on the same sample set resulting in the same quantitative and repeatable results by either method (Table 2). Finally the single-stage amalgamation method was ap- plied to the analysis of certified standard tissue and sediment samples (Table 3). The results are indistinguish- able from the certified values and from results produced by the laboratory using two-stage amalgamation with a mix- ture of both peak height and peak area determination. Same-day precision on sextuplicate analyses was better than -t 2%.Conclusions In the two-stage amalgamation procedure mercury from various samples is collected onto different gold-coated sand traps and then thermally transferred into a single ‘analyti- cal’ trap which is essentially a part of the detector. All mercury samples are then transferred from this second trap to the detector thus standardizing the peak shape and making quantification by peak height possible. In a case Table 1 Comparison of results (n= 3-6 analyses per point) for the determination of mercury in hair samples as part of an interlaboratory intercomparison exercise by HWC. Expected values are the mean k SD of all participating laboratories after rejection of 20 outliers (9-1 2 laboratories per mean) Mercury concentration/pg g-’ Peak area Peak height Expected HWC No.Single-stage Two-stage Single-stage mean value 91-3-1 19.0 k 0.4 1 19.7 k 0.45 13.71t 1.97 19.8k 1.08 9 1-3-2 13.8k0.03 14.1 k0.28 1 1.9 k 2.20 13.0 t 1.93 9 1-3-3 8.90k0.32 9.20k0.31 8.97 1- 2.39 8.60 & 1.27594 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY JUNE 1993 VOL. 8 Table 2 Comparison of the single and two-stage amalgamation techniques with peak area measurement for the determination of spike recoveries (pg of Hg) in water from a contaminated urban lake ( n = 3 analyses per point 20 ml aliquots of water) Mercury recovered Two-stage Single-stage Spike additiodpg Pg (%) Pg (Oh) 0 489k 12 - 499 & 7 - 500 101 1 k 17 104 1010+38 102 I000 1511% 1 1 102 1540k 17 104 Table 3 Analytical results for total mercury in acid-digested certified reference materials NRCC (National Research Council of Canada) Dogfish Muscle DORM- 1 and NRCC Estuarine Sediment BEST- 1.Note that two-stage values are a mixture of peak height and peak area measurements over an 18 month period as part of internal quality assurance Total mercury concentration/pg g-I Reference material Single-stage ( n = 6) Two-stage ( n = 23) Certified DORM-1 0.772k 0.004 0.795 k 0.041 0.798 2 0.074 BEST- 1 0.089 k 0.002 0.087 k 0.003* 0.092 * 0.008 * n = 3 . where the laboratory has no access to an electronic peak integrator the two-stage amalgamation procedure is essen- tial if precision results are to be obtained with multiple gold traps and a chart recorder. However it is time consuming compared with the single-stage method.Thus if many samples are to be analysed the savings in time might quickly pay for a dedicated peak integrator. Our studies using a variety of matrices seem to indicate no additional protection from matrix interferences (if indeed present) when converting from single- to two-stage amalgamation. With the use of electronic peak integration as presented here two-stage amalgamation becomes unnecessarily ineffi- cient. Single-stage amalgamation yields similar precision and accuracy and is more rapid. However it must be emphasized that if irregular peaks which can lead to poor precision are to be avoided attention must be paid to proper orientation of the gold-coated sand trap in relation to the detector. We would like to thank Sharon K. Goldblatt of Frontier Geosciences for her considerable editorial assistance in the preparation of this manuscript. References 1 Fitzgerald W. F. and Gill G. A. Anal. Chem. 1979,51 17 14. 2 Bloom N. S. and Crecelius E. A. Mar. Chem. 1983 14,49. 3 Gill G. A. and Fitzgerald W. F. Mar. Chem. 1987 20 227. 4 Horvat M. and Lupsina V. Anal. Chim. Acta 1991,243 71. 5 Nakahara T. Tanaka T. and Musha S. Bull. Chem. SOC. Jpn. 1978 51 2020. 6 Yamamoto J. Kaneda Y. and Hikasa K. Int. J. Environ. Anal. Chem. 1983 16 1. 7 Anderson D. H. Evans J. H. Murphy J. J. and White W. W. Anal. Chem. 1971 43 151 1. 8 Bloom N. S. and Fitzgerald W. F. Anal. Chim. Acta 1988 208 151. 9 Nojiri Y. Akira O. and Fuwa K. Anal. Chem. 1986 58 544. 10 Ballantine D. S. and Zoller W. H. Anal. Chem. 1984 56 1288. 1 1 Patterson J. E. Anal. Chim. Acta 1984 164 119. 12 Patterson J. E. Anal. Chim. Acta 1982 136 321. Paper 2/05 763H Received October 28 1992 Accepted January 20 1993

ISSN:0267-9477    

DOI:10.1039/JA9930800591

出版商:RSC    

年代:1993

数据来源: RSC