摘要:

JOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1399 Spatial Imaging of the Furnace Atomization Plasma Emission Spectrometry Source Victor Pavski and Chuni L. Chakrabarti Centre for Analytical and Environmental Chemistry Ottawa-Carleton Chemistry Institute Department of Chemistry Carleton University 7 725 Colonel By Drive Ottawa Ontario K7S 5B6 Canada Ralph E. Sturgeon* Institute for Environmental Research and Technology National Research Council of Canada Ottawa Ontario K7A OR9 Canada The spatial distribution of background species and analyte in a He FAPES source have been determined using a charge-coupled device (CCD) imaging system. Imaging of the He 667.82 nm support gas line reveals the presence of two distinct plasmas an intense luminous zone located around the centre electrode and a diffuse plasma near the tube walls.In a graphite tube at room temperature the plasma system exhibits a significant dependence on the d.c. bias of the centre electrode for a constant forward r.f. power of 50 W. Polyatomic background species exhibit similar overall distributions and response to changes in the centre electrode d.c. bias. The ingress of atmospheric nitrogen into the plasma can be indirectly detected at elevated tube wall temperatures. Images obtained for Ag atomized into 30 W and 50 W plasmas point out the role of the centre electrode as a primary condensation and secondary re-volatilization site. Condensation and re-evaporation is alleviated when Ag is atomized into a 70 W plasma as the result of plasma-induced heating of the centre electrode.Keywords Furnace atomization plasma emission spectrometry; imaging; charge-coupled device; centre electrode; d.c. bias r,f. power Furnace atomization plasma emission spectrometry (FAPES) a combined source in which samples are electrothermally atomized into an r.f. plasma formed inside a graphite furnace is emerging as an attractive technique for simultaneous multi- element analysis.'" Since its inception in 1989,495 a number of studies have been completed on analytical applications of FAPES6-10 as well as fundamental investigations into the characteristics of the plasma''-14 and analyte atomization p r o c e s s e ~ . ~ . ~ ~ ~ ~ ~ Despite this attention the spatial distribution of the plasma and its response to changes in operating param- eters remains largely unknown.Hettipathirana and Bladed3 have obtained spatially-resolved emission intensities of plasma gas species across the width of the graphite tube by translating their source in discrete increments in front of a linear photodi- ode array detector. Such a mapping approach is however limited to the recording of thin two-dimensional strips of information and cannot reliably be used for obtaining spatial and temporal distributions of excited analyte species during atomization of transient species since a series of images of sequential analyte atomizations which may not be reproduc- ible would have to be gathered. A particularly successful approach to the determination of analyte atom distributions within electrothermal atomizers has been the shadow spectral filming technique of Gilmutdinov and c~-workers'~-~~ in which atomic or molecular absorption across a two-dimensional cross-sectional area of a graphite tube is obtained by recording the attenuation of the primary light source (i.e. hollow cathode or electrodeless discharge lamp) through the atomizer onto 16mm film.This technique has recently been further refined by the replacement of the film camera with a charge-coupled device (CCD) camera as a detector,20 thus permitting near real-time quantitative analysis of either steady-state or transient phenomena. Obtaining spatially- and temporally-resolved distributions of analyte atoms and plasma gas species within the FAPES source is of importance because strong radial inhomogeneity of the plasma is indicated.13 Such information would be useful * To whom correspondence should be addressed.NRCC No. 37578. not only for diagnostic purposes but also may serve as an aid in determining an optimal observation point for analytical determinations. Additionally the FAPES plasma may also be utilized to gain insight into phenomena affecting Massmann- type graphite furnace atomizers in general such as the ingress and distribution of components of the ambient atmosphere. Experimental Instrumentation FAPES atomizer Much of the instrumentation used in this study has been described in detail in previous publication^.^.'^ The FAPES workhead consisted of a modified Perkin-Elmer Model HGA-2200 atomizer into which a 1 mm pyrolytic graphite coated graphite electrode was co-axially centred.A mount with four set-screws was attached to one end of the furnace housing to permit precision positioning of the centre electrode. Radiofrequency power is supplied to the centre electrode by an Advanced Energy Ind. (Ft. Collings CO USA) Model RFX-600 13.56 MHz generator tuned with a Model ATX-600 impedance matchbox. The d.c. bias voltage on the centre electrode was controlled in approximately 12 V increments over a range of -88 to +88 V using an r.f. blocking box (Advanced Energy Ind. Model 5018-000-C) and eight 12 V motorcycle batteries connected in series.14 The magnitude of the bias voltage and the reflected power were read from the appropriate ports on the ATX-600 tuner. A Perkin-Elmer Model 76B graphite furnace controller was used as the furnace power supply.The maximum power heating rate was employed for all analyte atomizations. The temperature of the graphite furnace was measured with a calibrated Ircon Series 1100 automatic optical pyrometer (Ircon Niles IL USA) focused onto the graphite tube wall through the dosing hole. The output of the pyrometer was collected and stored on a Nicolet Series 4094 digital oscillo- scope (Nicolet Instruments Canada Mississauga ON Canada). Although pyrolytic-coated graphite tubes ( Perkin-Elmer Norwalk CT USA) were used for most experiments a tube1400 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 coated with a 50 ym layer of pyrolytic boron nitride (Ringsdorff Werke Bonn Germany) and a tube fabricated entirely from machinable glass ceramic (Macor; Corning Glass Works Corning NY USA) were also used.A 2.80 x 0.40 x 0.06 cm Macor strip and a 2.50 x 0.50 x 0.05 cm graphite strip designed to fill the length of a graphite tube were used as L'vov platforms in some experiments. Imaging spectrometer The basic components of the imaging spectrometer have been outlined in a previous paper,20 with the salient modifications being the removal of the primary light source and collimating lens the replacement of the conventional graphite furnace atomizer with one modified for FAPES,' and the addition of the r,f. generator components. A block diagram of the appar- atus appears in Fig. 1. A Photometrics CH250/A thermoelec- trically-cooled camera (Photometrics Tuscon AZ USA) containing a Thomson TH1883 chip (384 x 576 pixels) was used.The operation of the CCD is controlled by a Photometrics AT200 16-bit camera controller board installed in a Raven 486/33 microcomputer (33 MHz 80486DX central processing unit). The output of the CCD is digitized by a 500kHz A/D converter and the resulting 12-bit image is displayed using Image200 Image Acquisition Software (Version 2.0.0; Photometrics). For obtaining images of steady-state ('static') phenomena in the plasma such as support gas and background polyatomic species the full spatial resolution of the CCD was utilized.21 A sub-region on the chip of sufficient size to contain the image was defined (typically 380 x 280 pixels) an exposure time was selected to provide good signal to noise (0.4 to 4 s depending on the wavelength region) and the camera was manually triggered.For obtaining time-resolved images the following modifications were made. A co-axial cable connecting the 'Peak Read' port of the graphite furnace power supply and the serial port of the computer was used to trigger the camera at the start of an atomization cycle. The CCD itself was operated in frame-transfer mode,21 wherein the chip is divided into two regions of equal size. An image is acquired on the lower half of the chip (image array) and then quickly transferred into the upper half of the chip (storage array) for subsequent read-out. To prevent blurring (aliasing) of the image during frame transfer a variable-frequency chopper is positioned between the atomizer and the focusing lens with the frequency output of the chopper connected to the AT200 camera controller board.The timing of the frame transfer is controlled by the frequency output of the chopper so that the transfer occurs while the emission from the atomizer is blocked by a chopper blade. The ultimate time resolution of the CCD is limited by its readout rate (200000 pixels s-') and thus the image size. In order to obtain time resolution sufficient to accurately image FAPES emission transients the size of the active image Chomer ADerture CCD chiD Tuner I Lens Interference filter Power supply 1 Power R.f. generator Fig. 1 Block diagram of the CCD-based optical imaging spectrometer array was reduced by the use of 4 x 4 binning in which the output of 4 serial and 4 parallel pixels is combined. Although such an approach also decreased the spatial resolution of the CCD the resolution obtained (=0.01 mm2) is quite adequate and was similar to that used earlier for the imaging of transient species during atomization in ETAAS.20 Images of transient species during emission were obtained at a rate of 50ms per image.A 1:1 image of the centre of the atomizer interior is formed on the CCD camera by a 150mm focal length bi-convex UV grade synthetic fused silica lens. Light exiting the lens then passes through an iris diaphragm to provide depth-of-field and an interference filter (Andover Corporation Lawrence MA USA) to isolate the spectral region of interest. The interference filter is mounted in a precision post rotator (Melles Griot Irvine CA USA) which is attached to the CCD camera by means of a flexible bellows (Ealing Electro-Optics Holliston MA USA) to prevent stray light from striking the working surface of the CCD.The interference filters used in this study to isolate the desired spectral lines are tabulated in Table 1 along with their associated bandpasses and trans- mission efficiencies at peak maximum. Table 1 is organized such that the most prominent band of a particular transition is listed first followed by weaker transitions which fall within the bandpass of the interference filter used. Information on the band system and wavelengths for the polyatomic species is taken from Pearse and Gaydon.22 Wavelengths for the atomic lines are from Corliss and B ~ z r n a n ~ ~ and Reader and C ~ r l i s s . ~ ~ Data on the energy levels for atomic and molecular systems are from Radzig and S m i r n ~ v .~ ~ Reagents High purity He (99.995%; Liquid Carbonic Scarborough ON Canada) was passed through the internal purge gas inlet of the graphite furnace at a flow rate of 200 ml min-' and served as the FAPES plasma gas. High purity Ar (99.995%; Liquid Carbonic Scarborough ON Canada) was used as the external sheath gas and was maintained at a flow rate of 11 min-'. A stock solution (lOOOmgI-l) of Ag was prepared by the dissolution of the high purity metal. Working standards were obtained by dilution of the stocks with high-purity (18.3 Mi2 cm resistivity) distilled de-ionized water acidified to 1% v/v with HNO (Ultrex grade; J. T. Baker Canada Toronto ON Canada). Procedure For obtaining images of the He plasma and background polyatomic features (COf N ?-NO and OH) the furnace power supply was not used and a 50 W plasma was formed while the (unloaded) graphite furnace was at room temperature.The d.c. bias on the centre electrode was varied to determine its effect on the spatial distribution of the plasma and back- ground polyatomic species. An exposure time sufficient to provide good signal-to-noise was selected and the CCD camera manually triggered. This procedure was repeated using a boron nitride-coated tube and a Macor tube. Spatial distributions of the above plasma gas species were also determined for graphite boron nitride and Macor tubes into which Macor and graphite strips were inserted. To obtain images of ingress of the atmosphere into the FAPES plasma a 50 W plasma was first formed with a 0 V d.c.bias imposed on the centre electrode. The furnace was then ramped from room temperature to 2000°C at a rate of 1250°C s-'. At the onset of atomization a signal pulse was sent from the peak read port of the furnace controller to the serial port of the computer and the CCD camera triggered automatically. A series of 10 images each taken with an exposure time of 1 s were then successively collected. Images of FAPES emission transient species were obtainedJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1401 Table 1 Spectral features monitored and interference filters used Species He I N (second positive system) CN (violet system) CO+ (first negative system) ?-NO (third positive system) OH B2C-X2C A2C2-X2 TI A2C+-X211 3.19-0 Transition Energy levels/eV A/nm 'D2-'PI0 23.07-21.22 667.82 C311,-B3 llg 11.05-7.38 380.49 (0,2) 371.05 (2,4) 375.54 (1,3) 385.79 (4,7) 389.46 (3,6) B2 C-X2 C 388.34 (0,O) 385.09 (4,4) 385.47 (3,3) 386.19 (2,2) 387.14 (1,l) 229.96 (0,l) 221.45 (1,l) 222.03 (8,6) 222.27 (5,4) 224.04 (2,2) 225.43 (6,5) 226.86 (3,3) 229.37 (10,8) 229.82 (4,4) 232.52 (1,2) 245.25 (2,3) 236.25 (6,6) 238.15 (3,4) 259.56 (0,3) 251.64 (2,5) 252.36 (2,5) 255.00 (1,4) 255.90 (1,4) 263.07 (2,6) 263.91 (2,6) 267.14 ( 1 3 ) 268.00 (1,5) 306.36 (0,O) 306.72 (0,O) 307.80 (0,O) 308.90 (0,O) 294.52 (3,2) 295.12 (3,2) 296.24 (3,2) 2pl,20-2s1,2 3.66-0 338.29 5.69-0 5.45-0 4.05-0 Transmission Central wavelength efficiency of and FWHM of filter at interference peak filter/nm maximum (%) 651; 25 62 380; 10 43 340; 10 230; 10 260; 10 300; 10 43 12 24 23 340; 10 32 as follows.Solutions (10 pl) of 4.0 pg ml-' Ag were manually pipetted into the atomizer. A 100°C 30s drying stage and 400 "C 30 s charring step were used. The graphite tube was than allowed to cool to room temperature and the plasma ignited. The analyte was-then atomized at 2700°C (using maximum power heating) for 5 s after which the r.f. power was turned off. The CCD was operated in frame transfer mode and a sequence of 100 images was obtained from the onset of the atomization trigger. The analytical line monitored and interference filter used for Ag are listed in Table 1. The contri- bution of plasma background emission to the Ag distribution was assessed by recording a sequence of images from a 'blank' atomizataion of 10 pl of 1% HN03 under identical furnace temperature and plasma conditions. This sequence was then subtracted from the analyte sequence to yield the background- corrected analyte distribution.The effect of r.f. power on the distribution of Ag emission was ascertained using 30 50 and 70 W plasmas for which the d.c. bias on the centre electrode was maintained at 0 V. Results and Discussion Optical resolution in the imaging system described in this study is obtained through the use of interference filters and therefore the spectral resolution that can be achieved (typically 20nm) is poor by the standards of conventional atomic emission spectroscopy. Hence a judicious matching of inter- ference filter with spectral line is required and appropriate caution must be exercised in the interpretation of results.Wavelength traces of background structures formed in a Massmann-type FAPES atomizer using He as the plasma gas under conditions similar to those used in this work have been recorded by Sturgeon et aL7 They reveal that the bandpasses of the interference filters used in this work are sufficient to isolate predominantly only the polyatomic background spec- tral features desired (see Table l) with the exception of the N second positive system which also contains the 388.89 nm He I line and several prominant transitions of the CN violet system. The He I line lies on the far wing of the bandpass of the interference filter used (380+ 10 nm) with the result that transmission efficiency is poor (about 1 %) so the contribution of this He line to the measured N distribution is negligible.As well CN emission would only be expected to contribute significantly to the N distribution at elevated temperatures when the vapour pressure of carbon is such as to form a significant amount of CN. For the imaging of Ag emission transient species plasma background lines arising from the firing of a 'blank' graphite tube were subtracted from the analyte distribution using the CCD imaging software as described under Experimental. Obviously this procedure1402 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 would not compensate for the presence of any polyatomic or dimeric analyte species contained within the spectral bandpass of the interference filter used.It is generally accepted however that Ag is reduced on the graphite surface and atomization occurs from direct volatilization of the metal without the production of volatile intermediate polyatomic oxide s p e ~ i e s . ~ ~ ~ ~ The 340 f 10 nm filter used effectively isolates the Ag 338.29 nm analysis line from other Ag atomic or dimer transition^,^^ and although weak Ago bands occur at 349.34 349.01 348.77 and 348.42 nm,22 these lie on the far edge of the interference filter bandpass where the transmission efficiency is poor ( z 1 %). If present they would probably not be detected. Effect of Centre Electrode d.c. Bias on He Distribution Fig. 2 shows the spatial distribution across the central axis of the graphite tube for the He 667.82nm line obtained using a 50 W plasma with a 0 V d.c.bias on the centre electrode and the tube wall at room temperature. The image in Fig. 2(u) is a false-coloured representation of the actual CCD image where white represents high intensity and blue represents low inten- sity. The three-dimensional plots of Fig. 2(b) are graphs of pixel intensity over the cross-section of the graphite tube. As a result of the depth-of-field of the optics all intensities are integrated along the length of the graphite tube so no longi- tudinal information can be obtained. From Fig. 2(u) it is evident that there is an intense region of plasma surrounding the centre electrode along with a diffuse outer ring close to the tube walls with comparatively little emission in the inter- mediate region.The intensity of the plasma surrounding the centre electrode is approximately 20 fold greater than that of the outer ring. It should be stressed at this time that the outer ring does not terminate at the walls but is recessed from them by a ‘dark space’ of roughly 90 pm thickness. This estimate was made by heating the graphite tube to approximately 2000°C in the absence of a plasma and recording the incan- descent image of the tube wall through the same interference filter used to isolate the He 667.82nm line. This image was then added (using the Image200 software) to an image obtained for the He plasma with the tube at room temperature. A gap between the thermal emission image of the inner tube wall and the terminus of the outer plasma ring was evident; its magni- tude was estimated from the image display on the computer screen given that the internal diameter of the tube is 6 mm.It may be argued that the outer ring occurs as a result of reflection from the graphite tube walls of light from the plasma surrounding the centre electrode or that it may arise from photoelectrons ejected from the tube walls interacting with the plasma support gas. Subsequent experiments have shown that the second plasma exists as a consequence of the reversal of polarity of the r.f. field between the centre electrode and the tube wall. In a study of an atmospheric pressure r.f. glow discharge formed between two parallel-plate copper electrodes of equal area Hotz28 found that the (cathode) glow shifted position and was associated solely with the electrode which acted as the cathode during a given half-cycle.Because of the rapidity of this process with respect to the integration time of the human eye the glow appeared to be simultaneously located at both electrodes.28 This situation also occurs with the co-axial cylindrical electrode geometry employed in FAPES. The weaker intensity of the outer ring plasma is a consequence of the 40-fold-increase in surface area of the outer electrode (graphite tube; 6 mm diameter) relative to the centre electrode (1 mm diameter). Experiments were undertaken to determine the effects of the d.c. bias on the centre electrode on the structure of the plasma. A constant forward power of 50 W was maintained. The effect of a change in the d.c. bias of the centre electrode on the plasma is also of practical interest from an analytical standpoint since the d.c.self-bias of the centre electrode has been shown to change significantly during the course of an emission transient species momentarily attaining values of approximately - 120 V at elevated temperature from an initial (room temperature) potential of about - 10 V.I4 Fig. 3 represents the images and corresponding three- dimensional intensity plots of emission from He I (667.82 nm transition) in a graphite tube as the d.c. bias of the centre electrode is varied. The distribution for the - 12 V bias is of interest in that it is most representative of a self-bias FAPES system operating with the tube wall at room temperature. As the d.c. bias of the centre electrode is made more negative [Fig.3 (a) and (b)] there is a small but significant increase in the intensity of the plasma surrounding it. Although not particularly evident from the 2D distributions this increase is clearly revealed in the 3D plots when they are compared with the 0 V bias [Fig. 2 (b)]. The outer ring intensity decreases slightly as the d.c. bias is made more negative; unfortunately the large scale used in this series of figures precludes a clear representation of this condition. These intensity changes would support the belief that the outer ring does not result from reflection or excitation by emitted photoelectrons. Were this Fig. 2 of 50 W and a 0 V bias imposed on the centre electrode (a) false-coloured image and (b) three-dimensional response surface Cross-sectional distribution of the He 667.82 nm line intensity obtained with the graphite tube at room temperature a forward r.f.powerJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 I I I 1403 ... I l l ... Ill ... I l l ... (d) I I I I l l Fig. 3 Effect of changes in the d.c. bias potential of the centre electrode on the cross-sectional distribution of intensity from the He 667.82 nm line. The forward r.f. power was 50 W and the graphite tube was at room temperature. (a) False-coloured images at (i) - 12 (ii) -47 (iii) - 83 V bias; (h) three-dimensional response surface at (i) - 12 (ii) -47 (iii) -83 V bias; (c) false-coloured images at (i) + 12 (ii) + 35 (iii) +83 V bias; and ( d ) three-dimensional response surface at (i) + 12 (ii) + 35 (iii) + 83 V bias1404 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 the case an increase in the intensity of the centre plasma (and hence an increase in the flux of photons to the tube wall) would also bring about an increase in the intensity of the outer ring. The observed increase in the centre electrode plasma with increasing negative d.c. bias can be explained by considering that a larger negative d.c. bias on the centre electrode implies that it behaves as the cathode for a greater period of time during a given half-~ycle.~’ Fig. 3(c) and (a) illustrate the effects of increasing positive d.c. bias on the centre electrode. In general with increasing positive bias the intensity of the centre electrode plasma decreases and there is enhanced emission in the region between the centre electrode and the tube wall.Most significantly however plasma intensity at the tube wall increases as positive bias is increased to +83 V. These effects can be explained by considering that a positive d.c. bias applied to the centre electrode will cause the counter-electrode (graph- ite tube) to behaves as a cathode for a greater period of time during a given half cycle such that the plasma will be enhanced near its surface. The increases in intensity of this plasma coupled with the decreases in intensity of the centre electrode plasma with increasing positive bias are consistent with the above hypothesis. At a low positive bias of 12 V [Fig. 3(c)(i)] two regions of plasma are established at the wall possibly reflecting the presence of a ‘negative glow’ and a ‘cathode glow’ at the wall.It may be that a ‘negative glow’ and a ‘cathode glow’ exist at the centre electrode but are not resolved from one another owing to the intensity of the centre electrode plasma and the small surface area of the centre electrode. Because of the larger surface area of the tube wall relative to the centre electrode these regions then become resolved at the wall as the polarity shifts and the field reverses. As the d.c. bias is increased above +25 V however the wall plasma merges into a single glow most probably owing to increases in the intensities of both the ‘cathode glow’ and ‘negative glow’. To further investigate the source of the outer ring plasma biasing experiments were repeated using a pyrolytic boron nitride tube and a Macor tube.Representative examples of the distributions obtained with these tubes are provided in Fig. 4. Although the intensity of the centre plasma in the boron nitride tube [Fig. 4(4 and (b)] was found to be of similar magnitude to that obtained for a graphite tube the outer ring plasma was more intense ( z 4 fold). This suggests a greater contribution of electrons to the ring plasma as a consequence of the low work function for secondary electron emission from this surface. For the Macor tube the intensities of both the centre electrode and ring plasmas were about half of that of the corresponding plasmas in a graphite tube [Fig. 4(c) and (41. As well the plasma intensity appears more uniformly distributed over the cross-section of the tube.Since the Macor tube is an insulating material the corresponding counter electrode for this system effectively becomes the furnace hous- ing some further distance beyond the tube. This decreases the electric field gradient and by consequence the intensity of both plasmas. In contrast to a graphite tube the spatial distribution of the plasma formed in a boron nitride or Macor tube remains largley unchanged by alteration of the d.c. bias of the centre electrode. No ‘cathode glow’ at the walls is observed with the application of a positive bias voltage and the relative intensities of the centre electrode plasma and the outer ring remain essentially the same regardless of the applied d.c. bias. The marginal change associated in plasma structure with changes in d.c.bias for these insulating materials can be explained as follows. Because an insulating surface is incapable of conducting a charge its potential will float to the potential of the centre electrode and thus the changes noted for the graphite tube will not be observed for such surfaces i.e.. there will be no change in the time a given electrode acts as the cathode during a given half-cycle. Although Cvov platforms have been used in FAPES without any analytical difficulties,’ the effect of the platform on the distribution of the centre electrode and outer ring plasmas has not been studied. For this purpose a 2.5 cm long graphite strip and a 2.8 cm Macor strip were placed inside graphite boron nitride and Macor tubes. Macor and boron nitride were selected to provide insulating surfaces which unlike graphite would not readily take the potential of the tube wall.Fig. 5 Fig.4 The effect of tube material on the distribution of the He 667.82nm line intensity forward r.f. power 50 W and tube wall at room temperature. (a) Boron nitride-coated tube - 50 V d.c. bias; (b) boron nitride-coated tube -t 50 V d.c. bias; (c) Macor tube - 37 V d.c. bias; and (d) Macor tube + 12 V d.c. biasJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1405 Fig. 5 False-coloured images illustrating the effect of a strip on the distribution of the He 667.82 nm line intensity in various tube materials forward r.f. power; 50 W d.c. bias of the centre electrode OV; and tube wall at room temperature. (a) Graphite tube and graphite strip; (b) graphite tube and Macor strip; (c) boron nitride-coated tube and Macor strip; and (d) Macor tube and Macor strip shows the images obtained with a 50 W 0 V bias plasma for a graphite boron nitride and Macor tube.As images of the boron nitride and Macor tubes were acquired using different camera exposure conditions absolute intensity comparisons for these systems cannot be made. For all tube materials there is an absence of the outer plasma ring beneath the strip. The outer ring instead terminates at the strip and extends around the central plasma. Although there is weak emission beneath the strip no distinct plasma structure is evident. Fig. 5(u) shows the results obtained with a graphite tube and graphite strip whereas in Fig. 5(b) results for a Macor strip are illustrated. For the graphite strip there is enhanced emission between the electrode and the strip while the intensity of the outer ring plasma is similar to that obtained in a graphite tube in the absence of a strip (z 100).The presence of the Macor strip in the graphite tube has the effect of increasing the overall intensity of the outer ring plasma (to x 500) while decreasing the emission between the electrode and the strip. The enhanced emission between the electrode and the graphite strip is a likely consequence of the enhanced field gradient as the electrode-tube wall distance is decreased. No difference in the characteristics of graphite or Macor strips in boron nitride or Macor tubes was noted presumably because the potential of these insulating tube materials float with respect to the centre electrode potential as noted earlier.Macor strips inserted into boron nitride or Macor tubes [Fig. 5(c) and (d)] exhibit the same general distribution as was observed for the graphite tube-Macor strip system in that the outer plasma ring termin- ates at the strip and then tends to deflect around the centre electrode. Fig. 6(u) and (b) shows the false coloured images and 3D intensity plots obtained in a graphite tube for background polyatomic species CO' N NO and OH at 50 W power with a OV d.c. bias on the centre electrode and the graphite tube wall at room temperature. No relative intensity inter- comparisons can be made for these images as a consequence of the different exposure times used. In general the distri- butions of these species are analogous to those obtained for He and their responses to changes in applied d.c.bias are also similar. However the ratio of the centre electrode plasma intensity to the outer ring plasma intensity differs for the different species as does the extent of emission between the centre electrode and the tube wall. The distribution of COf [Fig. 6(b)] is the most similar to that observed for He and is also in good agreement with the distribution measured by Hettipathirana and Blades.13 Nitrogen NO and OH exhibit progressively increasing radial emission intensity and progress- ively decreasing centre electrode plasma intensities relative to the intensities of the outer ring. A similar broad distribution for OH has also been noted by Hettipathirana and Blades.13 These emission patterns follow the relative excitation energies for these species i.e.OH (4.05 eV) NO (5.45 eV) N (11.05 eV) CO' (19.39 eV)." Atmospheric Ingress into the FAPES Plasma The ingress of components of atmospheric air into graphite furnace atomizers has been the subject of considerable investi- g a t i ~ n . ~ O - ~ ~ Sturgeon et u2.l' have shown that most of the nitrogen-containing background species in an He FAPES plasma can be eliminated by using an enclosed integrated contact cuvette (ICC) furnace. This would suggest that the ingress of components of the atmospheric air into 'open' Massmann-type FAPES atomizers is a significant source of nitrogen- and oxygen-containing polyatomic background species. In view of this attempts were made to observe the ingress of nitrogen by imaging the nitrogen second positive system (see Table 1).Fig. 7 presents a selection of such images obtained at 1 s intervals as a dry unloaded furnace was ramped from room temperature to 2000T at a rate of 1250 "C s-'. Fig. 7(u) shows the N distribution at the onset of the ramp which can be regarded as being at essentially room temperature. This image is typical of that presented in Fig. G(a)(ii) and exhibits no unusual structure. No significant structural features are observed until 4 s into the ramp when the graphite tube first reaches its steady-state temperature of 2000°C. At this point a 'plume' appears extending from the dosing hole down to the bottom of the graphite tube radially filling much of the tube volume.The emission intensity at the dosing hole is approximately twice that near the tube bottom. This distribution remains essentially unchanged in structure over the next 6 s although its overall intensity increases as is evident from Fig. 7(c). It is noteworthy that the distribution resulting from ingress of air obtained experimentally in this study is in general form quite similar to the theoretical distribution calculated by Gilmutdinov et ~ 1 . ~ ~ for the ingress of oxygen into graphite tube atomizers. Although images of N ingress were obtained at elevated temperatures attempts to reproduce this structure at room temperature were unsuccessful. Increasing the forward power to 80 W or altering the d.c. bias from + 25 to - 38 V to change the plasma properties had no effect even in the absence of an external purge gas flow and with a low internal plasma gas flow rate (z 120 ml min-').It was found that a minimum tube wall temperature of 1900 "C was required (80 W 0 V bias plasma) for the 'plume' to be observed. These results are somewhat surprising in light of the findings of Sturgeon and Falk32 who indirectly showed that the ingress of atmospheric O2 into a Massmann-type graphite furnace atomizer at room temperature occurred even in the presence of both internal and external purge gas flows. A likely explanation for the results of this work is that the He FAPES plasma is not sufficiently energetic with the tube wall at room temperature to allow the ingress of N to be imaged. Several strong bands of the violet system of CN fall within the bandpass of the interference filter used to isolate the N Second Positive System and its relatively low excitation energy (3.27 eV; see Table 1)1406 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 I ... I l l iv Fig. 6 Spatial distribution of polyatomic background species in a graphite tube. The spectral features isolated and interference filters used are listed in Table 1. Forward power 50 W; d.c. bias of the centre electrode. 0 V; and tube wall at room temperature. (a) false-coloured images of (i) CO+ (ii) N2 (iii) NO (iv) OH; and (b) three-dimensional response surfaces of (i) COf (ii) N (iii) NO (iv) OH make its excitation likely. As the furnace temperature rises the internal partial pressure of carbon increases such that N entering through the sample injection hole readily reacts to form CN which then becomes excited in the plasma.This would account for the fact that the ‘plume’ is observed only at elevated furnace temperatures (2 1900 “C). It should also be noted that the results described above were obtained with a ‘worn’ electrode whose length had been eroded by He+ sputter- ing and volatilization of carbon during repeated heating cycles. Attempts t o reproduce the nitrogen ingress images with a fresh electrode traversing the entire length of the tube yielded irreproducible results. This is explicable in light of previous work’ which has shown that excitation temperatures in FAPES increase with decreasing electrode length; thus a short electrode (terminating a few mm beyond the sample injection hole) is required to provide a sufficiently energetic plasma for N excitation and/or an elevated partial pressure of carbon (for CN formation) which derives from a worn electrode surface.Additional factors may also come into play to account for formation of the ‘plume’. The fact that a minimum tube wall temperature of 1900 “C is required before nitrogen ingress is observed may suggest that thermionic electrons emitted at this temperature may play a role in enhancing the excitation capability of the plasma. Additionally the ‘plume’ is onlyJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 fa) 1407 Fig. 7 electrode 0 V. See Table 1 for the interference filter used and spectral features isolated. (a) time = 0 s; (b) time = 4 s; (c) time = 10 s False-coloured images illustrating ingress of atmospheric nitrogen into a graphite tube forward power 50 W and d.c.bias of the centre observed after the graphite tube has reached a steady-state temperature possibly because there would be a concomitant expulsion of the internal plasma gas out of the dosing hole in the initial stages of the heating ramp which would serve as an added barrier to the influx of nitrogen. As the tube reaches its steady-state temperature and the expulsion of the internal gas ceases the barrier to the influx of nitrogen is removed causing the concentration of nitrogen in the atomizer to increase sufficiently to enable the ‘plume’ to be observed in the (more energetic) plasma. Atomization of Ag Fig. 8 displays sequences of images obtained for 40 ng Ag atomized from the tube wall into 30 50 and 70 W plasmas (0 V bias).The numbers beneath each image denote its actual position in the atomization sequence. For 30 W [Fig. 8(u)] two distinct emission events are evident. Early in time (images 2-4) there is weak Ag emission near the centre electrode probably as a result of violent shattering of the AgNO crystal lattice on heating with transfer of analyte onto the cooler centre electrode as the tube temperature is too low at this time to promote atomization of Ag. Weak Ag emission at the wall where the sample was originally deposited also occurs. Subsequently (images 4-7) enhanced Ag emission originates at the tube wall as atomic Ag is transferred from the wall to the cooler centre electrode.The extremely weak Ag emission noted in images 8-11 and localized at the centre electrode suggests that the electrode has not yet reached a temperature sufficient to promote release of the condensed Ag. Release of Ag from the centre electrode begins to occur in image 12 and expands to fill the entire tube volume in images 13-14 after which dissipation begins. Imai and Sturgeon16 have reported the temperature of the centre electrode to be 450 K for a 30 W plasma and found that plasma-assisted heating of the tube wall is essentially negligible for forward r.f. powers of 30 W and less. Therefore the early condensation-transfer of Ag onto the electrode surface shown in images 2-6 appears reasonable as does the subsequent volatilization of Ag from the centre electrode.For a 50 W plasma two distinct emission events are also observed [Fig. 8(b)]. Silver emission occurs early in time exclusively at the wall (image 3) and then in the tube volume as Ag atomic vapour transfers from the wall to the centre electrode (images 5-7). No Ag emission is observed in images 8-9 owing to insufficient heat at the electrode; Ag condensed on the electrode is subsequently released in images 10-14. Unlike atomization into a 30 W plasma [Fig. 8(u)] no weak Ag emission was observed at the electrode either early or late in the transient emission most probably as a consequence of increased plasma-induced heating of the electrode at 50 W (600 K uersus 450 K16). In contrast to the 50 and 30 W cases atomization into a 70 W plasma yields only one distinct emission event [Fig.S(c)]. The FAPES plasmas operated at forward r.f. powers of >50 W have been found to exhibit substantial plasma-induced heating of both the tube wall and the centre electrode,16 with the result that double peaks associated with analyte condensation an re-volatilization pro- cesses are reduced or removed entirely. This is borne out by the sequence of images in Fig. 8(c) which show Ag emission beginning directly underneath the centre electrode (image 2) rapidly filling the tube volume (images 3-12) and then dissipating. Similar condensation-revaporization effects have been noted for Cr in the low pressure hollow anode furnace atomization non-thermal excitation spectrometry (HA-FANES) system.34 Conclusions Two plasmas exist in the FAPES source lending it substantial radial inhomogeneity.In addition to an intense plasma sur- rounding the centre electrode there exists a diffuse outer plasma near the tube walls. The spatial distribution of the plasma formed in a graphite tube is significantly influenced by the d.c. bias of the centre electrode. The ingress of atmospheric nitrogen can be followed by observation of CN emission at elevated temperature when atmospheric N combines with carbon volatilized from the graphite tube. Images obtained for the atomization of Ag into plasmas operated at forward r.f. powers of 30 50 and 70 W illustrate that while condensation and re-evaporation from the centre electrode is significant at the lower plasma powers plasma-induced heating of the centre electrode and the tube wall at 70 W is sufficient to prevent the centre electrode from acting as a condensation site.Use of a CCD imaging system provides reliable two- dimensional information useful in understanding the effect of operating parameters on the spatial distribution of species in the FAPES source. Imaging the source at different operating pressures and graphite tube diameters should provide greater insight into the nature of the plasmas present. The potential perturbation of the spatial distribution of plasma species and analytes by easily ionized elements (EIEs) and the effect of d.c. bias control in the presence of EIEs also merit future investigation. The authors thank J. W. C. Johns and P. Boivin of the National Research Council of Canada (NRCC) for the loan of the bulk of the interference filters.The assistance of J. Logan and M. Grenier of the Carleton University Electronics Shop in writing some of the data manipulation software and mod- ifying part of the existing FAPES equipment is gratefully acknowledged as is that of V. Luong of the Institute for Environmental Research and Technology (IERT) NRCC. The authors also wish to thank Photometrics Ltd. for assistance in obtaining the CCD camera and the Natural Sciences and Engineering Research Council of Canada for partial financial support of this project.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 fa) 1407 Fig. 7 electrode 0 V. See Table 1 for the interference filter used and spectral features isolated. (a) time = 0 s; (b) time = 4 s; (c) time = 10 s False-coloured images illustrating ingress of atmospheric nitrogen into a graphite tube forward power 50 W and d.c.bias of the centre observed after the graphite tube has reached a steady-state temperature possibly because there would be a concomitant expulsion of the internal plasma gas out of the dosing hole in the initial stages of the heating ramp which would serve as an added barrier to the influx of nitrogen. As the tube reaches its steady-state temperature and the expulsion of the internal gas ceases the barrier to the influx of nitrogen is removed causing the concentration of nitrogen in the atomizer to increase sufficiently to enable the ‘plume’ to be observed in the (more energetic) plasma. Atomization of Ag Fig. 8 displays sequences of images obtained for 40 ng Ag atomized from the tube wall into 30 50 and 70 W plasmas (0 V bias).The numbers beneath each image denote its actual position in the atomization sequence. For 30 W [Fig. 8(u)] two distinct emission events are evident. Early in time (images 2-4) there is weak Ag emission near the centre electrode probably as a result of violent shattering of the AgNO crystal lattice on heating with transfer of analyte onto the cooler centre electrode as the tube temperature is too low at this time to promote atomization of Ag. Weak Ag emission at the wall where the sample was originally deposited also occurs. Subsequently (images 4-7) enhanced Ag emission originates at the tube wall as atomic Ag is transferred from the wall to the cooler centre electrode. The extremely weak Ag emission noted in images 8-11 and localized at the centre electrode suggests that the electrode has not yet reached a temperature sufficient to promote release of the condensed Ag.Release of Ag from the centre electrode begins to occur in image 12 and expands to fill the entire tube volume in images 13-14 after which dissipation begins. Imai and Sturgeon16 have reported the temperature of the centre electrode to be 450 K for a 30 W plasma and found that plasma-assisted heating of the tube wall is essentially negligible for forward r.f. powers of 30 W and less. Therefore the early condensation-transfer of Ag onto the electrode surface shown in images 2-6 appears reasonable as does the subsequent volatilization of Ag from the centre electrode. For a 50 W plasma two distinct emission events are also observed [Fig.8(b)]. Silver emission occurs early in time exclusively at the wall (image 3) and then in the tube volume as Ag atomic vapour transfers from the wall to the centre electrode (images 5-7). No Ag emission is observed in images 8-9 owing to insufficient heat at the electrode; Ag condensed on the electrode is subsequently released in images 10-14. Unlike atomization into a 30 W plasma [Fig. 8(u)] no weak Ag emission was observed at the electrode either early or late in the transient emission most probably as a consequence of increased plasma-induced heating of the electrode at 50 W (600 K uersus 450 K16). In contrast to the 50 and 30 W cases atomization into a 70 W plasma yields only one distinct emission event [Fig.S(c)]. The FAPES plasmas operated at forward r.f. powers of >50 W have been found to exhibit substantial plasma-induced heating of both the tube wall and the centre electrode,16 with the result that double peaks associated with analyte condensation an re-volatilization pro- cesses are reduced or removed entirely. This is borne out by the sequence of images in Fig. 8(c) which show Ag emission beginning directly underneath the centre electrode (image 2) rapidly filling the tube volume (images 3-12) and then dissipating. Similar condensation-revaporization effects have been noted for Cr in the low pressure hollow anode furnace atomization non-thermal excitation spectrometry (HA-FANES) system.34 Conclusions Two plasmas exist in the FAPES source lending it substantial radial inhomogeneity.In addition to an intense plasma sur- rounding the centre electrode there exists a diffuse outer plasma near the tube walls. The spatial distribution of the plasma formed in a graphite tube is significantly influenced by the d.c. bias of the centre electrode. The ingress of atmospheric nitrogen can be followed by observation of CN emission at elevated temperature when atmospheric N combines with carbon volatilized from the graphite tube. Images obtained for the atomization of Ag into plasmas operated at forward r.f. powers of 30 50 and 70 W illustrate that while condensation and re-evaporation from the centre electrode is significant at the lower plasma powers plasma-induced heating of the centre electrode and the tube wall at 70 W is sufficient to prevent the centre electrode from acting as a condensation site. Use of a CCD imaging system provides reliable two- dimensional information useful in understanding the effect of operating parameters on the spatial distribution of species in the FAPES source.Imaging the source at different operating pressures and graphite tube diameters should provide greater insight into the nature of the plasmas present. The potential perturbation of the spatial distribution of plasma species and analytes by easily ionized elements (EIEs) and the effect of d.c. bias control in the presence of EIEs also merit future investigation. The authors thank J. W. C. Johns and P. Boivin of the National Research Council of Canada (NRCC) for the loan of the bulk of the interference filters. The assistance of J.Logan and M. Grenier of the Carleton University Electronics Shop in writing some of the data manipulation software and mod- ifying part of the existing FAPES equipment is gratefully acknowledged as is that of V. Luong of the Institute for Environmental Research and Technology (IERT) NRCC. The authors also wish to thank Photometrics Ltd. for assistance in obtaining the CCD camera and the Natural Sciences and Engineering Research Council of Canada for partial financial support of this project.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1409 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 References Banks P. R. Liang D. C. and Blades M. W. Spectroscopy 1992 7 36. Gilchrist G. F. R. Celliers P.M. Yang H. Yu C. and Liang D. C. J. Anal. At. Spectrom. 1993 8 809. Anders Ohlsson K. E. Sturgeon R. E. Willie S. N. and Luong V. T. J. Anal. At. Spectrom. 1993 8 41. Liang D. C. and Blades M. W. Spectrochim. Acta Part B 1989 44 1059. Sturgeon R. E. Willie S. N. Luong V. T. Berman S. S. and Dunn J. G. J. Anal. At. Spectrom. 1989 4 669. Smith D. L. Liang D. C. Steel D. and Blades M. W. Spectrochim. Acta Part B 1990 45 493. Sturgeon R. E. Willie S. N. Luong V. and Berman S . S. Anal Chem. 1990 62 2370. Sturgeon R. E. Willie S. N. Luong V. T. and Berman S. S. J. Anal. At. Spectrom. 1990 5 635. Sturgeon R. E. Willie S. N. Luong V. T. and Berman S. S. J. Anal. At. Spectrom. 1991 6 19. Sturgeon R. E. and Willie S. N. J. Anal. At. Spectrom. 1992 Sturgeon R. E.Willie S. N. Luong V. T. and Dunn J. G. Appl. Spectrosc. 1991 45 1413. Sturgeon R. E. Willie S. N. and Luong V. T. Spectsochim. Acta Part B 1991 46 1021. Hettipathirana T. D. and Blades M. W. Spectrochim. Acta Part B 1991 47 493. Sturgeon R. E. Luong V. T. Willie S. N. and Marcus R. K. Spectrochim. Acta Part B 1993 48 893. Hettipathirana T. D. and Blades M. W. J. Anal. At. Spectrom. 1992 7 1039. Tmai S. and Sturgeon R. E. J. Anal. At. Spectrom. 1994 9 493. Gilmutdinov A. Kh. Zakharov Yu. A Ivanov V. P. and Voloshin A. V. J. Anal. At. Spectrom. 1991 6 505. Gilmutdinov A. Kh. Zakharov Yu A. Ivanov V. P. Voloshin A. V. and Dittrich K. J . Anal. At. Spectrom. 1992 7 675. 7 339. 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 Gilmutdinov A. Kh. Zakharov Yu. A. and Voloshin A. V. J . Anal. At. Spectrom. 1993 8 387. Chakrabarti C. L. Gilmutdinov A. Kh. and Hutton J. C. Anal. Chem. 1993 65 716. Epperson P. M. Sweedler J. V. Bilhorn R. B. Sims G. R. and Denton M. B. Anal Chem. 1987 60 32712. Pearse R. W. B. and Gaydon A. G. The Ident$cation of Molecular Spectra 4th edn. Chapman and Hall London 1976. Corliss C. H. and Bozman W. R. Experimental Transition Probabilities for Spectral Lines of Seventy Elements U.S. Department of Commerce Washington D.C. NBS Monograph 53 1962. Reader J. and Corliss C. H. Wavelengths and Transition Probabilities for Atoms and Atomic Ions. Part I . WuvelengthA U.S. Department of Commerce Washington D.C. NBS Monograph 68 1980. Radzig A. A. and Smirnov B. M. Reference Data on Atoms Molecules and Ions Springer-Verlag Berlin 1985. Frech W. Lundberg E. and Cedergren A. Prog. Anal. At. Spectrosc. 1985 8 257. Smets B. Spectrochim. Acta Part B 1980 35 33. Hotz R. F. J. Appl. Phys. 1970 41 1500. Chapman B. Glow Discharge Processes Wiley New York 1980. L'vov B. V. and Ryabchuk G. N. Spectrochim. Acta Part B 1982 37 673. Sturgeon R. E. Siu K. W. M. and Berman S. S. Spectrochim. Acta. Part B 1984 39 213. Sturgeon R. E. and Falk H. Spectrochim. Acta Part B 1988 43 421. Gilmutdinov A. Kh. Chakrabarti C. L. Hutton J. C. and Mrasov R. M. J. Anal. At. Spectrom. 1992 7 1047. Riby P. G. Harnly J. M. Styris D. L. and Ballou N. E. Spectrochim. Acta Part B 1991 46 203. Paper 4/01 374C Received March 8 1994 Accepted June 21 1994

ISSN:0267-9477    

DOI:10.1039/JA9940901399

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 141 1 Mechanisms Controlling the Direct Solid Sampling of Silicon from Gold Samples by Atomic Absorption Spectrometry with Electrothermal Atomization Part 1 Analyte Migration to the Solid Surface Michael W. Hinds Royal Canadian Mint 320 Sussex Drive Ottawa Ontario Canada KIA OG8 Garrett N. Brown and David L. Styris Pacific Northwest Laboratory Box 999 Richland WA 99352 USA The results of electrothermal atomic absorption spectrometry (ETAAS) investigations of Si vaporization from solid gold samples indicate that Si is released only after melting of the gold matrix and the migration of Si to the surface is assisted by the formation of convection cells in the melt. This convection is driven by the thermal gradients in the sample.Diffusion alone in either the solid or the melt is too slow to account for the relatively short time for total Si release before a significant portion of the gold matrix is depleted. Silicon that is carried by the induced convection to the surface is readily vaporized. It is shown that the fiquidus lines of the binary phase diagrams associated with a gold solvent can be used to predict the existence of analyte carryover in the gold. Consequently these diagrams offer a means of identifying a prior; candidate elements that may be determined by solid sampling ETAAS in gold via aqueous solution Cali bration. Keywords Graphite furnace atomizer; electrothermal atomic absorption spectrometry; solid sample analy- sis; silicon; atomization mechanism Solid sample analysis by direct weighing into a graphite atomizer typically requires matrix matched standards.' The use of aqueous standards for such analysis has been demon- strated.Headridge and Riddlington2 and Irwin et aL3 used aqueous standards for determining Bi and volatile elements in nickel-based alloys. More recently Hinds and Kogan4 found that aqueous standard calibration was acceptable for the determination of Si in solid gold samples (although it has limited utility). In general it is not clear that the atomization mechanisms are identical for an analyte from an aqueous solution and from a solid gold sample or that these mechanisms are dissimilar and only fortuitously produce comparable numbers of analyte atoms in the analytical volume. There is currently little understanding of the processes that control analyte release and atomization associated with con- densed metal samples.The determinations of volatile elements in nickel-based alloys by electrothermal atomic absorption spectrometry (ETAAS) used aqueous standards either mixed with a chemical modifier or pipetted onto residual nickel solid spheres in the a t ~ m i z e r . ~ The resulting variations in the absorbance peak profiles are thought by some to reflect differences in chemical forms of the analyte or the extent of analyte occlusion within the solid sample5 Nikolaev,6 using thermodynamic calculations explained that observations of elements released from metal samples were a result of a distillation process. Extensive absorbance signal tailing should be expected in the direct analysis of solid metals because of the time required for analyte species to migrate through the bulk to the surface.However this tailing was not observed in the measurements of Si in solid gold samples4 or in the measurement of elements in nickel-based alloy^.^,^ Instead relatively sharp absorption peaks were observed. This suggests that analyte atoms migrate to the surface prior to atomization or that there is an enhancement of analyte transport to the surface. The purpose of this work was to address these possibilities and to provide an understanding of the observed behaviour of silicon being released from solid gold samples. Experimental Instrumental Two different atomic absorption spectrometers were used with a Perkin-Elmer (Norwalk CT USA) HGA 500 graphite furnace during the course of this work a Perkin-Elmer Model 5000 and a Perkin-Elmer Model 3100.Data were collected by a personal computer interfaced to the spectrometers via a DAS8 12-bit analogue-to-digital converter (Keithley Metrabyte Taunton MA USA). Data collection hardware and software were modelled after the work of Allen and Jackson.' Continuum-source background-corrected integrated absorbance values were used throughout this work. The outside wall temperature was measured with a computer-interfaced optical pyr~meter.~ The atomizer was a pyrolytic graphite coated graphite tube. All of the atomic absorption measure- ments were made during wall atomization and unless specified otherwise the temperature programme described in Table 1 was used throughout.A Perkin-Elmer Model 41 OOZL atomic absorption spec- trometer was used for experiments that required re-use or inspection of the solid gold sample after a particular heating cycle. The associated transversely heated graphite furnace housing design permitted convenient recovery of solid samples from inside the atomizer. The atomizer can be removed without tilting as is the case for the HGA 500 atomizer. The silicon hollow-cathode lamp was operated at 40mA. Table 1 Temperature programme used for atomizing Si in solid gold samples. Wall atomization Step Temperature/"C Ramp time/s Hold time/s Dry 200 15 10 Pyrolysis 1400 1 15 Cool-down 20 1 5 Atomize 2650 0 6* Clean-out 2650 1 3 * Read 1s before atomization step and gas flow stopped during atomization.1412 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 The less sensitive non-resonance line at 221.1 nm was used with a slit-width of 0.2nm. A non-resonance line of Au (274.8 nm) was also measured with a gold hollow cathode (15 mA operating current) and with the slit-width set at 0.7 nm. Reagents Water used in these experiments was distilled and de-ionized using a Nanopure TI system (Barnstead/Thermolyne Dubuque IA USA). Calibration solutions were diluted from a 1000 yg ml-1 stock standard solution made from sodium silicate (High Purity Standards Charleston SC USA). Samples The gold sample with silicon impurity was a reference material prepared at the Royal Canadian Mint (RCM) by adding specific amounts of elemental impurities to high-purity gold (99.999%).The particular reference material used in this study was designated FAU8 and was found to contain 27.8 f4.8 pg 8-l of Si (+95% confidence interval). The pro- duction and characterization process has been detailed else- where,' so only a brief description will be presented. Weighed amounts of the solid metals were added to molten high-purity gold and the resulting mixture was cast. The cast bar was homogenized by rolling and annealing processes. Homogeneity of the silicon was checked by taking gold shavings from several positions on the bar. The shavings were acid washed to remove surface contamination. Samples were dissolved in aqua regia and silicon was determined by ETAAS and inductively coupled plasma atomic emission spectrometry ( ICP-AES).4 There was no indication of silicon segregation. Spot checks by spark ablation ICP-AES measurements about the bar also indicated that there was no evidence of silicon inhomogeniety.Sample Preparation and Solid Sample Introduction Solid pieces of gold were cut from shavings by using a stainless- steel knife. Pieces between 0.2 and 0.5 mg were selected by weighing with a Mettler UM3 electronic balance (0.0001 mg resolution). The selected gold sample was transferred first into a 10 ml plastic cup and then with the aid of fine curved- nose stainless-steel forceps placed in the graphite furnace. This placement was facilitated by positioning a funnel (pipette tip cut 7 mm from the tapered end) in the dosing hole. Results and Discussion Successful application of aqueous standards to the direct analysis of Si in solid Au requires that during the atomization cycle the aqueous and the solid samples each contribute equal numbers of free analyte atoms and equal numbers of stably bound analyte atoms bound in polyatomic species or as sorption (trapped) species in the atomizer.For discussion purposes this requirement is expressed through the equalities relating the 0 number of analyte atoms deposited (No) to the sum of the total number of free analyte atoms (Nat) and the number of analyte atoms (niNi) contained in each of the ni molecules or trapping sites of type i; Ni is the number of analyte atoms in the type i molecule or type i trapping site. Using (as) and (s) to denote aqueous and solid and setting No(,q) equal to No(s) for the present experiments it is seen that Nateaq) + C niNi(aqj = Nat(sj + C njNj(s) (1) 1 j where the Ni and N j need not be identical.The Nat(aq) and N,t(sj for Si are identical through the empirical equality of the Si absorbance peak areas. It follows that the second and last terms in eqn. (1) must also be equal for Si species. That these equalities should actually occur is unexpected. After all the N values are characterized by the time integral of the convolution of the system response function and the supply function for the given speciesg As the associated supply functions are probably unique to the particular sample form the values of N for the two samples should differ. Peak profiles (Fig. 1) for Si from an aqueous solution and Si from a solid gold sample are not exactly coincident but this effect is reproducible.Most of the atomization from the solution sample occurs under near-isothermal conditions within the atomizer. Silicon atomization from the solid sample occurs 0.1 s later than from the aqueous solution. It had been sug- gested that Si atoms from both sample types experienced similar temperature environments and therefore would have similar residence times.4 However even if atomization from the two samples occurs over identical regions of the tempera- ture-time profiles; the respective convolution integrals are not necessarily identical. Their equality in the present experiments remains unexplained. The background absorbance was relatively low extensive tailing was not observed and subsequent atomization cycles showed no further Si absorbance.It appears that all of the silicon atomizes from the solid sample during the first 2.5 s of the atomization step (Fig. 1). The physical form of the solid gold sample was observed at each step of the temperature programme through the dosing hole. There was no apparent change during drying but during pyrolysis at 1200°C the irregular shape of the sample was transformed into a spheroidal shape. Consequently the surface tension of the liquid gold was sufficiently large that the gold could not wet the graphite. The spheroid decreased in size but remained in the same position during each atomization cycle. Similarly Irwin et aL3 noted that nickel spheres formed and accumulated in the atomizer during solid sample ETA AS determinations of bismuth in nickel alloy samples.The mass loss of a 0.5mg gold sample exposed to one atomization cycle was determined by using a transversely heated graphite atomizer (platform removed) and by differen- tial weighing of the sample. As noted previously the furnace assembly for this atomizer permitted convenient recovery of the remaining gold sample for further analysis. It should be noted however that software limited the temperature to 2600°C. Under these conditions there was a 25% decrease in sample mass. This is an important observation because after one firing there is no evidence of Si in the remaining gold sample. In a separate experiment on a 0.5 mg solid gold sample the 274.8 nm non-resonance line of gold was monitored during consecutive atomization cycles.The absorbance signals observed for consecutive firings from various atomizer con- figurations are shown in Fig. 2. These curves indicate trends and vary depending on sample shape and placement within the atomizer. For transversely heated atomizers there was no deviation from the baseline after the fourth firing. The efficient removal of Au from the atomizer is probably due to the spatial isothermality of the atomizer which prevents condensation. 2.0 1- I 2800 -I 1.5 Time/s Fig. 1 B 0.4 mg solid gold sample; C outside tube wall temperature Peak profiles for 10 ng Si from A an aqueous standard andJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1413 16 1 I 0 2 4 6 a 10 No. of atomizer firings Fig.2 Effect of the number of atomizer firings on the integrated absorbance of Au at a non-resonance wavelength (274.8nm) for a 0.5 mg solid gold sample from A transversely heated atomizer (plat- form removed); B Massman atomizer (wall atomization); and C Massman atomizer (platform atomization) When Massman-type atomizers (with and without platforms) were used the residual gold was retained for more than four firings; gold condenses at the cooler ends of such atomizers.Condensed gold is visible on the tube ends after one firing. The condensed gold re-atomizes and partially re-condenses with each firing and cool-down phase. Falk et al." reported that during the atomization cycle the temperatures at the end regions of the Massman atomizer increase initially to maximum values corresponding to given positions along the length.Then during stabilization of the temperature of the central portion of the atomizer these end-region temperatures decrease mono- tonically towards lower values. Condensation at a given end-region position can occur at this lower temperature. Re-atomization from this position can occur early in a new cycle when the temperature is maximum. Similar effects were observed by Frech et a2.l' for chemical modifiers condensing at the tube ends. Fewer gold atoms (smaller peak area com- pared with wall atomization) were vaporized from repeated platform atomizer firings probably owing to the slower heating rate of the platform compared with wall atomization. Silicon absorbance (221.1 nm) and gold absorbance (274.8 nm) were monitored for 0.5 mg gold samples atomized in a Massman atomizer. The results are illustrated in Fig.3. The appearance of the silicon signal follows that of gold by about 0.3 s and ends before an appreciable fraction of the gold sample has vaporized. This latter observation indicates that all of the silicon in the solid sample is released and would necessarily have to be transported to the near-surface of the gold sample prior to or during atomization. This conclusion was addressed by repeating the experiments on a solid gold sample having silicon distributed only on the surface. A 10 11 aliquot of a 10 ppm Si solution was dried on to a 0.4 mg high- purity gold sphere (formed by heating the gold to 1400°C in a graphite furnace). The sphere with the resulting surface deposit of Si was placed in the graphite furnace and heated (temperature programme as in Table 1).As shown in Fig. 4 3 1 I z 2 -P m J 0 1 2 3 4 Time/s Fig. 3 Peak profiles for A Si atomizing from a 0.5 mg solid gold sample and B Au at a non-resonance wavelength (274.8nm) for a 0.5 rng solid gold sample 0.6 I I 0.4 8 0.3 $ 0.2 0.1 0 -0.1 I I I I I I L 0 0.5 1 .o 1.5 2.0 2.5 3.0 Time/s Fig. 4 Absorbance peak profiles for Si A occluded in and B adsorbed onto 0.4 mg high-purity gold spheres. Temperature programme included at pyrolysis step at 1400 "C there is little difference between the peak shapes resulting from Si deposited on or occluded within gold spheres. This similarity suggests that prior to atomization occluded Si is transported to the surface region of the sample or surface silicon is transported into the bulk gold.Complete exhaustion of silicon from the molten gold during the atomization cycle indicates that the silicon either co-volatilizes with the gold or rapidly diffuses to the surface of the gold where it vaporizes. If silicon co-volatilizes all of the silicon would have to be contained in the vaporized surface region determined by the outer 8% of the gold sphere radius. It is difficult however to conceive of a diffusion mechanism that would cause all of the silicon in the sphere to diffuse to and concentrate in this outer layer of the gold. For such a mechanism to be effective it would be necessary to overcome the concentration gradient responsible for Fickian-type diffusion. Were such a mechanism controlled by bulk solvent properties alone it would also be operating during metallurgi- cal processing involved in the macro sample preparation.It is evident however that preferential migration of silicon to the surface of the samples prior to the pyrolysis cycle has not occurred. Random sampling from the bulk gold samples indicated a uniform dispersal of Si throughout the bulk (as noted under Experimental). The mechanism responsible must therefore be associated with size and molten form of the sample during the atomization cycle. A plausible mechanism for the above effect is the rapid diffusion of silicon to and vaporization from the surface of the molten gold during the atomization cycle. The chemical potential needed to drive such diffusion is the negative concen- tration gradient created by the vaporization of silicon reaching the surface of the molten gold droplet.Maintenance of this negative gradient places a limiting condition on the relative rates at which the silicon solute and the gold solvent vaporize. This condition is obtained by setting a vanishing upper limit on the total derivative of volume density of N ( t ) silicon atoms in a sphere of radius R. The result ( 1/N) IdN( t)/dtl> (1/R) IdR/dtl indicates that the magnitude of the fraction time rate of change of N ( t ) must exceed the magnitude of the fraction time rate of change of R. This condition is met in the present experiments as only 25% of the gold is vaporized during the same period in which 100% of the silicon is depleted. To evaluate the feasibility of such a mechanism we use the radial diffusion equation for a sphere with the diffusing species vaporizing from the surface; the solution to the equation was given by Crank.12 This solution to the diffusion equation indicates that the diffusion coefficient for Si in gold must be no smaller than 3 x lop2 cm2 s-' assuming a 0.2 mm diameter gold sphere and its depletion of diffusing species within a 0.7 s interval during atomization; this period of depletion is indi- cated in the data in Fig.1. The diffusion coefficient for Si in1414 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 solid copper at 727"C has been reported by Norwich13 to be nine orders of magnitude smaller than the above minimum. Gold should exhibit Fickian diffusion properties similar to those of copper; both are face-centred cubics with similar interstitial spacings between atomic radii.Therefore the rapid diffusion of Si cannot occur in solid gold. The diffusion coefficient for molten gold can also be calcu- lated from copper diffusion data if the data are representative of the copper liquid phase. Using the empirical liquid copper self-diffusion results from Henderson and Yang,14 it was found that even at 27OO0C the self-diffusion coefficient D(cu-cu) is two orders of magnitude below the minimum (noted above) necessary to explain the Si depletion The diffusion coefficient D(si-cu) for the diffusion for Si in liquid copper at this temperature will be slightly greater than D(cu-cu). The correction is based on D(si-cu) being inversely proportional to the square root of the reduced mass for Si and Au.The corresponding diffusion coefficient for Si diffusing in liquid gold should only be 60% smaller than D(si-cu) considering the differences between the atomic masses of gold and copper. Consequently bulk diffusion even in liquid gold cannot account for the observed rapid transport of Si in the gold samples. Bulk diffusion might be enhanced however possibly through another mechanism field-aided diffusion of Si in the high electric fields of the space charge formed when oxygen adsorbs on SiO. Since the Fermi level of gold is greater than that of Si dilute amounts of Si in the present experiments cannot be competitive with the gold in providing the necessary electronic states. Consequently Si contained in the gold lattice remains neutral and its motion is uninfluenced by the above elec- tronic fields.We are left with only one other possibility the influence that the surface tension of the small molten gold sample has on the transport of Si. It is well established that inhomogeneit- ies in surface tension can induce surface movements that in turn induce convective motion in the underlying fluid.15 In the presence of a temperature gradient transverse to the surface the convective cells are driven by heat flow into the body of the fluid. Solute in the molten samples can therefore be carried to the surface by this convective flow. A combination of vaporization of some fraction of the solute (Si) as it reaches the surface and transport of the remaining solute to the surface from the bulk will result in solute depletion from the sample.For the present experiments the surface tension is certainly a basic property of the molten Au sphere and this tension is a strong function of temperature (approximately 5 x N m-l OC-'). Minor temperature inhomogeneities can yield large local variations in this surface tension.16 These variations can then induce convective flow in the molten sphere. The atomizer provides a conductive and radiative heat source to drive the flow if a temperature gradient is established across the bulk. However does such a temperature gradient exist within the sample? To address this question assume that a small cube- shaped sample rests on the graphite surface. Uniform contact between surfaces is not necessary. Assuming conductive heat paths through the contact points to the sample radiative heating from the atomizer walls and radiative losses from the sample through the ends and dosing hole to the cooler 'outside' surfaces the rate of change in enthalpy of the sample can be expressed as p VC,d T/dt = kad T/dx + OE,A ( K4 + T4) - 06 A ( T4 - Tb4) (2) where p is the sample density I/ is the sample volume C is the heat capacity T is the sample temperature p is the density A and E are the sample radiative area and emissivity respect- ively and the subscripts s a and b refer to the sample atomizer and background respectively.The temperature derivative refers to the vertical temperature gradient across the sample where x is in the direction perpendicular to the planar surfaces of contact o is the Stephen-Boltzmann constant a is the sample to atomizer contact area and k is the thermal conduc- tivity of the gold sample.The first term on the right-hand side of eqn. (2) is the thermal heat conductively of the gold. The second and third terms express the heat radiated from the atomizer to the sample and the heat radiated from the sample to the background respectively. The heat flow by gas-phase convective processes can be neglected in this approximation as radiation terms dominate at the high temperatures associated with the molten gold samples. The steady-state temperature gradient is then given by dT/dx=(a&,A,/ka) (T4- T 4 ) + ( ~ ~ A k ~ ) (T4- Tb4) (3) This gradient is negative if (&,A + &,A,) T4 < &,A K4 + G4 (4) As A,>>A this inequality can be simplified to This is the condition necessary to establish the negative temperature gradient required for induced convection flow in the liquid sample.As the sample temperature ( T ) can never be greater than the atomizer temperature ( K ) a vertical tempera- ture gradient must exist at temperatures significantly high to establish radiative heating and radiative cooling of the sample. Once again this gradient can drive the motion of the convective cells that are induced by the surface tension inhomogeneities. Silicon absorbance profiles from Si occluded in the gold reference material FAU8 with and without pyrolysis at 1400 "C are compared in Fig. 5. The principal difference between these two samples is that the pyrolysed sample is annealed and recrystallized whereas the sample not undergoing pyrolysis begins the atomization cycle from a work-hardened state. Differences in peak areas were due to the difficulty involved in obtaining samples of identical mass.It was rationalized that this difference in the samples might manifest itself as differences in the silicon peak profiles as trapping sites for impurities will be dissimilar for the two samples. To the contrary the Si profiles were similar. The substantial differences that can exist in the gold lattice imperfection densities associated with these two pre-treatments did not evidently affect silicon release. 'Therefore the silicon must be released after the lattice defects (impurity traps) are annihilated i.e. after melting occurs. The cool-down step between pyrolysis and atomization was increased to 40s to ensure that ambient temperatures were reached prior to atomization. In experiments in which a very short cool down step was used (Table l) the atomization cycle was initiated close to the pyrolysis temperature.Therefore the silicon signal was shifted early in time compared with that obtained without pyrolysis. Atomization of Si deposited on gold surfaces was also investigated. As shown in Fig. 6 elimination of the pyrolysis 0.6 /- 1 0.5 0.4 0.3 0.2 0.1 0 -0.1 I I I 1 I I 0 0.5 1 .o 1.5 2.0 2.5 3.0 Tirne/s Fig. 5 Comparison of peak profiles for silicon occluded in 0.3 mg gold samples; A with and B without a pyrolysis step. Temperature programme included a 40 s step at 20 "C prior to atomizationJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 1415 0.7 2 0.5 2 0.3 a e a 0.5 1 ' . ' -0.1 1 1 I I I I 1 0 0.5 1 .o 1.5 2.0 2.5 3.0 Timeis 0.1 Fig. 6 Comparison of peak profiles for silicon adsorbed on to 0.4 mg high-purity gold spheres; A with and B without a pyrolysis step. Temperature programme included a 40s step at 20°C prior to atomization I - I .I 1 ,\ o w . - step affected the absorbance profile of Si deposited and dried on the surface of a gold sphere. A slight delay of the peak maximum was observed for the Si absorbance profile from the non-pyrolysed sample compared with the same experiment with a pyrolysis step although the signals appeared at the same time. A delay might be expected if the non-pyrolysed sample contained a higher percentage of more refractory silicon dioxide (compared with the monoxide) than the pyrolysed sample.Surface tension changes could also explain the delay. The surface impurity from the silicon deposition will decrease the surface tension of the liquid gold. This decrease will enhance wetting of the graphite and will thus extend the liquid sample-graphite interface available for the direct interaction with the condensed silicon. It can be hypothesized that this extension influences analyte release. The above experiment was repeated with silicon in aqueous solution but without solid gold present; the resulting absorbance signals were identical (Fig. 7). Comparison with Metallurgical Data It is interesting to compare these results with analyte release characteristics that might be interpreted from the Si-Au binary phase diagrams.Whether such diagrams can be useful for these purposes is debatable. Certainly at the temperatures of release the liquid-metal phase is stable at all analyte concentrations. Consequently at such temperatures the binary phase diagram is insufficient to predict the release behaviour. However the liquidus line may be useful for these purposes even though it represents considerably lower temperatures of the liquid metal. This line indicates the tendency for the analyte to remain in solution. Consequently it exhibits the temperature dependence associated with the analyte-matrix bond energies albeit at ;; 0.9 I I -0.1 I I I I 0 0.5 1 .o 1.5 2 .o 2.5 3.0 Time/s Fig.7 Comparison of peak profiles for silicon originating from an aqueous standard; A with and B without a pyrolysis step.Temperature programme included a 40 s step at 20 "C prior to atomization temperatures considerably lower than the observed appearance temperatures. Although these energies reflect the surface ten- sions and the heats of vaporization that are of interest here it remains to be shown that the indicated temperature depen- dences can be utilized to predict behaviour at more elevated temperatures. The phase diagram for a binary mixture of silicon and gold17 indicates that for low concentrations of Si near the gold melting-point the bonding of Si in the gold lattice decreases with increasing temperature. Hence Si should be readily released from the bulk gold at more elevated temperatures. This argument can be extended to the release of Sn from molten gold.The phase diagram for Sn-Au near the gold melting-pointi8 is similar (although more complex) to that for the Si-Au system. The absorbance profile for Sn atomizing from a solid gold sample (Fig. 8) has the same characteristics as observed for Si atomization from gold a sharp peak no extensive tailing and no observable levels of analyte remaining in the gold sphere from a second atomization cycle. This suggests that the mechanism of transportation of Sn in gold is similar to that suggested for Si and that this similarity is predictable from the gross similarities in the liquidus lines of the respective phase diagrams. In contrast the liquidus line in the phase diagram for the Fe-Au system'' shows that Fe bonding in the gold matrix is independent of temperature for low concentrations near the gold melting-point.Therefore relative to the release of Si and Sn a slow release of Fe might be expected. This is observed in the absorbance profile for Fe atomizing from a liquid gold sample (Fig. 9). This profile shows a marked tailing. The Fe absorbance peak appears earlier and is of larger magnitude than those observed for Si and Sn because the sensitivity for Fe is approximately an order of magnitude larger. Results from a second (follow-up) atomization cycle indicate that measurable amounts of Fe are retained in the gold sphere remnant of the initial cycle (Fig. 9). This also supports the above interpretation of the slope of the Fe-Au liquidus line. Consequently this line indicates that the use of aqueous standards for direct determinations of Fe in a solid gold matrix is an ineffective approach.The Pd-Au liquidus line2* indicates that the Pd-matrix bond energy increases with increasing I I 0 2 4 6 8 Time/s Fig. 8 Peak profile of Sn (286.3 nm) atomizing from a 0.5 mg solid gold sample 1 .o a S m e v) 2 0.5 0 1 2 3 4 5 6 Timeis Fig. 9 Peak profiles of Fe (248.8 nm) atomizing from a 0.4 mg solid gold sample; A first and B second atomization cycle1416 JOURNA OF ANALYTICAIL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 I 0 1 2 3 4 5 6 Time/s Fig. 10 Peak profiles A 2 ng Pd (247.6 nm) atomizing from a solu- tion integrated absorbance = 0.501 s; B 4 ng Pd atomizing from 0.2 mg solid gold sample (first atomization cycle) integrated absorbance = 0.449 s; and C Pd atomizing from the remaining solid gold sample during second atomization cycle integrated absorbance = 0.170 s temperature near the melting-point of the gold matrix.Therefore Pd absorbance might be observed in a second firing and tailing on the release of Pd might also be expected. Fig. 10 shows the absorbance profile for Pd atomizing from a gold sample; the tailing feature is evident and Pd atomization was delayed compared with atomization from an aqueous solution. Significant absorbance was also observed during the second atomization cycle so some Pd was retained in the remainder of the gold sample. Interestingly Pd profiles from Pd in an aqueous solution also reveal tailing which is due to the refractory nature of Pd. It is probable that the tailing from the solid gold sample is due to both the interaction between Pd and gold and the refractory nature of Pd.From the above discussion it appears that the liquidus line of the binary phase diagrams can be used to predict rough features of analyte release from solid gold samples. In particu- lar these diagrams can predict whether to expect full release of the analyte from the gold melt. This information can be used to develop a convenient and accurate calibration scheme for the direct determination of trace elements in such samples ix. solution calibration or matrix-matched standards. Whether such predictions might be extended to other metal matrices is speculative. Conclusions ETAAS results and liquid-metal diffusion data indicate that for ETAAS of solid gold samples Si vaporization occurs primarily from the surface of the matrix in the molten phase.The rapid exhaustion of Si and perhaps Sn results from transport of the analyte in the convective cells of the matrix. These cells induced by surface tension inhomogenieties and the temperature gradients in the molten sample move the analyte between the bulk and surface regions. This mixing combined with vaporization of the species near the matrix surface results in depletion of analyte before much of the gold matrix is vaporized. The efficiency of analyte release is related to analyte-matrix bond energies and can be approximated from the binary phase diagrams. For analytes occluded in gold metal the slopes of the liquidus lines in these diagrams can be indicative of how these energies might change with tempera- ture.These slopes can therefore be used to predict whether analyte carryover in the gold matrix should be expected. Further efforts are required to determine if such an interpret- ation of the phase diagrams can be extended to other binary systems. 'The authors express gratitude to Drs. G. Miiller-Vogt H. Cochran and 0. Bassran for bringing liquid metal convec- tion cell formation to their attention to Dr. G. Ocampo for helpful discussions on the metallurgy of gold and to Dr. K. W. Jackson for the graphite furnace temperature measurements. 'This work was supported in part by the Director of Research Office of Basic Energy Sciences Chemical Sciences Division of the US Department of Energy and performed under contract DE-AC76RLO 1830.1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 References Langmyhr F. J. and Wibitoe G. Prog. Anal. At. Spectrosc. 1985 8 193. Headridge J. B. and Riddington I. M. Mikrochim. Acta 1982 2 197. Irwin R. Mikkelsen A. Michel R. Dougherty J. P. and Preli F. R. Spectrochim. Acta Part B 1990 45 903. Hinds M. W. and Kogan V. J. Anal. At. Spectrom. 1994,9,451. Hassell D. C. Rettberg T. M. Fort F. A. and Holcombe J. A. Anal. Chem. 1988 60 2680. Nikolaev G. I. Zh. Anal. Khim. 1973 28 454. Allen E. and Jackson K. W. Anal. Chim. Acta 1987 192 355. Kogan V. Hinds M. Ocampo G. and Valente G. in Precious Metals 1993 ed. Mishra R. International Precious Metal Institute Press Allentown PA USA 1993. Paveri-Fontana S. L. Tessari G. and Torsi G. Anal. Chem. 1974,46 1032. Falk H. Glismann A. Bergann L. Minkwitz G. Schubert M. and Skole I. Spectrochim. Acta Part B 1985 40 533. Frech W. Li K. Berglund M. and Baxter D. C. J. Anal. At. Spectrom. 1992 7 141. Crank J. The Mathematics of Diffusion Clarendon Press Oxford 2nd edn. 1976 p. 96. Norwich A. S. J. Appl. Phys. 1951 22 1182. Henderson J. and Yang L. Trans. AM. Inst. Min. Metall. Pet. Eng. 1958 212 72. Sternling C. V. and Scriven L. E. AIChE J. 1959 5 514. Allen B. C. in Liquid Metals ed. Beer S. Z. Marcel Dekker New York 1972 p. 188. Phase Diagrams of Binary Gold Alloys ed. Okamoto H. and Massaiski T. B. ASM International Metals Park OH 1987 p. 267. Phase Diagrams of Binary Gold Alloys ed. Okamoto H. and Massaiski T. B. ASM International Metals Park OH 1987 p. 280. Phase Diagrams of Binary Gold Alloys ed. Okamoto H. and Massaiski T. B. ASM International Metals Park OH 1987 p. 103. Phase Diagrams of Binary Gold Alloys ed. Okamoto H. and Massaiski T. B. ASM International Metals Park OH 1987 p. 221. Paper 4/03733B Received June 20 1994 Accepted August 18 1994

ISSN:0267-9477    

DOI:10.1039/JA9940901411

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1417 Approach to the Determination of Lead by Vapour Generation Atomic Absorption Spectrometry Les Ebdon Phillip Goodall and Steve J. Hill Analytical Chemistry Research Unit Department of Environmental Sciences University of Plymouth Drake Circus Plymouth UK PL4 8AA Peter Stoc kwel I P S Analytical Arthur House Sevenoaks UK TN15 6QY K. Clive Thompson LabServices Yorkshire Water Charlotte Road Sheffield UK S2 4EQ A new method for the determination of lead by vapour generation atomic spectrometry is described. This is based upon continuous flow methodology and involves the derivatization of lead in the presence of an oxidant with sodium tetraethylborate to yield volatile tetraethyllead. Simplex optimization was applied to this system and to conventional hydride generation.Characteristic concentrations of 0.36 ng cm-3 (Pb 283 nm) and 0.145 ng cm-3 (Pb 217 nm) for alkyl generation atomic absorption spectrometry were observed. These are approximately 4-5 times better than the equivalent hydride system and superior to all reliable literature values for characteristic concentrations of lead hydride generation atomic absorption spectrometry. Using the 283 nm lead line limits of detection (3s) obtained for the hydride and alkyl generation systems were 0.62 and 0.07 ng ~ m - ~ respectively. The determination of lead in a standard reference water (NISI SRM 1643c) is reported and good agreement obtained between the certificate value (35.3+0.9 ng cm-3) and the experimentally derived value (34.8 1.4 ng cm -3).Keywords Lead; vapour generation; derivatization; continous flow; tetraethyllead It is widely accepted that the determination by atomic spec- trometry of many of the Group 14-16 elements can be enhanced significantly by the formation of volatile covalent hydrides. Despite the environmental importance of lead the generation of lead hydride has not become as popular as the generation of the hydrides of As or Se because of the relative difficulty of the hydride technique for the determination of lead and the availability of excellent alternatives such as electrothermal atomic absorption spectrometry (ETAAS) which were adequate for existing environmental legalisation. Recent proposals for reducing levels of lead in the environment mean that ETAAS becomes a problematical method at these new levels.For example recent recommendations by the World Health Organization suggest that the maximum permissable levels of lead in drinking water will be reduced from 50 to 10 ngcmP3 of Pb and this is likely to be taken up by the European Union. The analytical uses of lead hydride with AAS detection were first reported by Thompson and Thomerson.' Improvements in lead hydride yield and consequently sensitivity by the application of auxiliary oxidants were made by Fleming and Ide2 using dichromate-tartrate and Vijan and Wood3 using hydrogen peroxide. A number of systems involving different oxidants and chelating agents have been reported including potassium ~ermanganate,~ cerium( IV),5 persulfate6 and he~acyanoferrate(II1)~ with a variety of chelating agents such as lactate tartrate malate citrate and oxalate. A number of studies of oxidant systems have been ~ndertaken,~.~** and depending upon the exact nature of the oxidant system characteristic concentrations ranging from about 0.4-3.2 n g ~ m - ~ have been reported although most fall within the range of 0.8-2 ng ~ m - ~ .The exception to this seems to be a report by Li et aL9 who observed characteristic concentrations of 0.04 ng cmP3 for a Ce(II1)-oxalic acid oxidant. The vapour generation of lead by derivatization with sodium tetraethylborate (NaTEB)l0>l1 has been used as a pre- concentration strategy for ETAAS by trapping tetraethyllead (TEL) in the graphite tube at 400°C prior to atomization." A characteristic mass of 28 pg was obtained and the determi- nation was apparently free of interference up to a 5000-fold excess of Fell' Ni" Mn'l As"' Zn11.12 Sodium tetraethylborate has been utilized in vapour generation systems for cadmium by D'Ulvio and Chen13 and more extensively by Ebdon et In addition NaTEB has been used as a ethylating reagent in speciation analysis using gas chromatographic techniques." In this paper we report an alternative method of vapour generation for lead based upon derivatization with NaTEB to TEL.Major improvements in sensitivity were obtained by the use of an oxidizing agent (hydrogen peroxide). This new approach was compared with conventional continuous flow lead hydride generation after rigorous optimization of the two approaches using Simplex routines.Experimental Instrumentation Derivatization of lead was performed under continuous flow methodology using a commercial hydride generator (PS Analytical Sevenoaks UK). Atomic absorption measurements were accomplished using a flame heated quartz tube atom cell (SP9 Pye Unicam Cambridge UK). The operating conditions for both the hydride system and the atomic absorption spec- trometer are shown in Table 1. Control of the reactor collection and processing of the analytical signals were accomplished via a PC using commercial software (Touchstone Spinoff Technical Systems Benfleet UK). Gas-liquid separation was accomplished using glass U-tube separators of either the Type A configuration (PSA H003-G101 Hydride Type) or Type B (PSA H003-G102 Mercury Type).Reagents Sodium tetrahydroborate. Analytical reagent grade solutions stabilized by the addition of an alkali (NaOH 0.1 mol dmP3) were prepared freshly each day and filtered prior to use. Hydrogen Peroxide Solution 27.5% v/v. Analytical grade Aldrich Chemical Company Gillingham UK. Sodium Hydroxide (Aristar) Nitric Acid (AnalaR) and Hydrochloric Acid (AnaZaR). BDH Poole UK. Sodium Tetraethylborate. Prepared in these lab~ratories~~1418 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 Table 1 Operating conditions for lead vapour generation Hydride generator Sample flow rate/cmV3 min-' Blank flow r a t e / ~ m - ~ min-' Reagent flow r a t e / ~ m - ~ min-' Delay time/s Rise time/s Analysis time/s Memory time/s Atomic Absorption Spectrometer Flame type Path length/cm Lamp current/mA Bandpass/nm Wavelength/nm 8 8 4 10 30 60 60 Air/acetylene stoichiometric 10 6 0.5 283.3 or 217.0 0 90 Time/s 180 but further purified by crystallization and recrystallization from sodium-dried diethyl ether at - 78 "C.Solutions were prepared freshly each day and purged vigorously with argon prior to use to remove ether of crystallization. Standard Lead Solution 1000 pg ~ 1 1 2 ~ ~ . Aldrich Chemical Company. Results and Discussion Gas-Liquid Separation The interface of a continuous flow vapour generator with an atom cell is normally accomplished using some form of gas- liquid separation device such as a glass U-tube. These are of two basic designs Type A used in hydride generation and Type B used in cold vapour (CV) mercury determinations.The essential difference between the two designs is whether the carrier gas is sparged through the reaction mixture (Type B) to produce an intimate gas-liquid mixture or acts solely as a means of transferring an evolved vapour into an atom cell (Type A). The former are used almost exclusively with CV-mercury generation using Sn" reductants. The latter type are associated normally with hydride generation chemistries involving reagents such as sodium tetrahydroborate. In common with previously reported work on cadmium vapour generation,14 no detectable lead vapour was evolved in the ethylation system when either the Type A U-tube or a mem- brane gas-liquid separator were used as interfaces. The Type B separator was successful in isolating a volatile lead compound under conditions which will be discussed later. Unlike the cadmium vapour generation system useful analytical signals were obtained with argon gas flow rates more akin to those normally employed in CV mercury systems (e.g.200-400 cm3 min-l). This reflects both the less reactive nature of TEL and its higher vapour pressure when compared to diethylcadmium. Initial Experiments on Tetraethyl Diethylcadmium Lead Vapour Generation The use of NaTEB to ethylate Pb2+ in aqueous solution was investigated by mixing at a T-Piece a reagent stream of aqueous NaTEB (1 % m/v 8 cm3 min-l) with an aqueous lead standard (50 ng cm-3 Pb2+ 8 cm3 min-I). The combined streams were passed to a Type B gas-liquid separator and purged with argon at a flow rate of 800cm3 min-l.The effluent from the gas-liquid separator was passed to an AA spectrometer for detection of lead. A very small absorbance was observed (Fig. 1A). This small absorbance was unambigu- ously assigned to lead by repeating the measurement at the 217 nm line and the possibility of non-specific absorbance eliminated by the imposition of deuterium background correc- tion and measurement at a non-absorbing line (280.2 nm). The formation of TEL from an aqueous solution of Pb" requires an oxidation step. Reaction of Pb" with NaTEB is Fig. 1 Vapour generation AAS of tetraethyllead ([Pb] = 50 ng cmP3) spectrometry A no oxidant; and B with oxidant known to follow a disproportionation route to the TEL product. l2 Pb2+,,,I+ [B(Et),]- =Pb(Et),+B(Et) ( 1 4 2Pb(Et)z=Pb + Pb(Et)4 ( 1B) The sensitivity and peak shape obtained with lead alkyl vapour generation was poor.It was thought that conversion of lead( 11) to lead(1v) using an oxidizing reagent such as hydrogen per- oxide might provide a route which did not involve the dispro- portionation step. Addition of hydrogen peroxide to the sample/blank solutions resulted in a large increase in absorbance due to lead and improved peak shape (Fig. 1B). The following stoichiometries are suggested for the derivatiz- ation of lead@) by NaTEB in the presence of an oxidant. or Pb2+(a,,+2H+(aq~+H202=Pb(IV)(a,~+2H20 (3a) This situation mirrors the development in hydride generation between the original paper by Thompson and Thomerson (1) and the papers by Fleming and Ide (2) and Vijan and Wood (3). Hydrogen peroxide was used as the oxidant because of its convenience and low contamination with respect to lead.Pb(IV)(,,) + 4 [B( Et)4] -(,,)=PbEt + 4BEt3 (3b) Optimization of Lead Vapour Generation Chemistry The lead alkyl generation system was compared with hydride generation to assess if there were any advantages to be gained either in terms of sensitivity or robustness of the experimental conditions. The two chemistries were first optimized using the modified simplex technique. l 5 9 l 6 Four variables were optimized namely purge gas flow reagent oxidant and sample/blank acid concen- trations. The figure of merit was absorbance using peak height measurements. In all cases a blank with the same composition in terms of oxidant concentration and acidity as the sample was used.In both hydride and alkyl generation cases the reagent was dissolved in aqueous 0.1 mol dmP3 NaOH. The individual reagent solutions were prepared by dilution of a concentrated stock solution with aqueous 0.1 mol dm-3 NaOH. Prior to determining the response at each individual vertex the reagent and blank solutions were pumped through the vapour generation system for at least 5 min to ensure that an equilibrium was obtained. Sample/blank flows were mixed in a 2:l ratio with respect to reagent flows. The optimizations were continued until the system was judged to be incapable of further improvement. This conditionJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 was assumed to be achieved when the standard deviations of the five highest response factors were within twice the estimated precision of the methods [i.e.relative standard deviation (RSD) = 6%]. The five highest responses generated during the optimization are shown in Table 2 (hydride) and Table 3 (alkyl). The optimum conditions are assumed to be the mean of each set of variables. To determine the critical parameters and confirm the accuracy of the optimization univariate searches were performed around the simplex derived optima by holding three parameters constant at the mean shown in Tables 1 and 2 and varying the fourth parameter. These are shown in Figs. 2(u)-5(u) (hydride) and Figs. 2(6)-5(b) (alkyl). The mean value optimization of the individual parameters is indicated by an arrow whilst the vertical lines represent the extremes of the sets of variables at the stopping conditions.Even a cursory examination of Figs. 2-5 reveals that the lead alkyl generation chemistry is under conditions yielding maximum sensitivity more sensitive to experimental conditions than the corresponding hydride. This is compensated for by the higher efficiency and consequently higher sensitivity of the lead alkyl generation. The critical experimental parameter and perhaps the most difficult to control is the sample acidity which is of paramount importance in both hydride and alkyl vapour generation. In the alkyl generation case the optimum sample acidity corresponds to complete neutralization of the alkaline matrix of the reagent [Fig. 2(b)]. However at the expense of approximately halving the sensitivity lowering the sample acidity to x20 mmol dm-3 accesses a region where the analytical signal is relatively independent of sample acidity [Fig.2(u)]. Even under these non-optimum conditions the lead alkyl system is still significantly more sensitive than the equivalent hydride system. Use of the alkylation chemistry above 50mmol dmM3 of acid i.e. [NaOH] c[H+] is not recommended even though TEL formation is still highly efficient. Partial hydrolysis of NaTEB yields volatile B(Et) (b.p.=95”C) whose presence in the atom cell scatters the incident radiation leading to greatly increased noise on the analytical signal and baseline. Boron was also deposited upon the surface of the quartz T-cell. Continuum background correc- tion reduced but did not eliminate completely this non- specific scatter.The analytical signal is relatively independent of reagent concentration in the ranges of 0.5-2% m/v NaTEB and Table 2 Stopping conditions for the simplex optimization of lead hydride chemistry; [Pb] = 50 ng c r ~ - - ~ Purge gdm- m m ~ l d m - ~ min-’ (YO v/v) Absorbance 38.1 134 490 1.25 0.135 46.8 88 451 1.12 0.123 48.9 115 750 1.32 0.119 40.9 118 670 1.38 0.119 38.8 115 660 0.95 0.117 [NaBH$/ [H+]/ gas flow/cm3 [HzOz] Mean 42.7 114 604 1.20 0.123 Table 3 Stopping conditions for the simplex optimization of lead alkyl chemistry; [Pb] =20 ng cm-3 “aTEBI/ CH + I/ g dmP3 mmol dm- 4.4 1 52 7.17 44 5.9 1 54 8.03 48 8.63 45 Mean 7.23 49 ~ Purge gas flow/ cm3min-l 3 10 230 260 260 260 270 CHZ0Zl (YO v/v) Absorbance 0.25 0.275 0.38 0.265 0.40 0.255 0.43 0.270 0.44 0.270 0.38 0.213 0.4 0.3 0.2 0.1 Q c 1419 - z 0 20 40 60 80 100 120 140 160 180 200 0 $ 0.5 a (b) 1 0.4 0.3 0.2 0.1 0 10 20 30 40 50 60 70 80 [Acidl/mmol dm-3 Fig.2 Univariate search of sample acidity around the simplex derived optima for (a) hydride and (b) alkyl lead vapour generation AAS see text for explanation of arrow and vertical lines 0.6 I (a) 0.5 0.4 0.3 0.2 0.1 8 - ; o 0 0.5 a 0.4 0.3 0.2 0.1 1 2 3 4 5 6 7 8 9 10 Isodium tetra hydraborate] (“YO) 0 0.5 1 .o 1.5 2.0 [Sodium tetraethylboratel (%) Fig. 3 Univariate search of reagent concentration around the simplex derived optima for (a) hydride and (b) alkyl lead vapour generation AAS see text for explanation of arrow and vertical lines 5-8% m/v NaBH,. The lower optimum concentrations of NaTEB offsets partially the higher cost of this reagent when compared with NaBH,.Lower oxidant concentrations are required for the alkyl system when compared with the hydride system and both systems demonstrate clearly defined and relatively broad optimums (Fig. 4). The milder oxidizing con- ditions at the reaction T-piece reflects the stabilization of lead (IV) by TEL when compared with plumbane. The lower1420 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 0.4 0.3 0.2 0.1 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 [Hydrogen peroxide1 (%) Fig. 4 Univariate search of oxidant concentration around the simplex derived optima for (a) hydride and (b) alkyl lead vapour generation AAS see text for explanation of arrow and vertical lines 0 . 4 ~ 0.3 o.2 0.1 I 0 200 400 600 800 Purge gas flow/cm3 min-' 1000 Fig. 5 Univariate search of purge gas flow rate around the simplex derived optima for (a) hydride and (b) alkyl lead vapour generation AAS see text for explanation of arrow and vertical lines oxidant concentration for the alkylation system is beneficial in terms of reduced blank levels.Overall the alkylation chemistry is significantly more sensi- tive than the hydride. If the absolute sensitivity of the method is compromised in favour of a more stable behaviour with respect to sample acidity [Fig. 4(b)] then the experimental conditions are more robust than the equivalent hydride system. Table4 Experimental conditions for the determination of lead in NIST 1643c by alkyl vapour generation AAS Reagent Purge gas [ NaTEB]/ alkalinity/ [Oxidant] flow/ g dm-3 mol dmP3 (% v/v) cm3 min-l 7.5 0.55 0.275 300 Vapour generator settings Delay/s 10 Read/s 60 Rise/s 30 Decay/s 40 Reagent stream flow rate/cm3 min-' 4 Sample/blank flow rate/cm3 min-' 8 Detection Limits and Calibration The detection limits and characteristic concentrations of lead by vapour generation AAS were determined for the alkyl and hydride chemistries at the simplex derived optima.The detec- tion limit was defined as 3sb. The characteristic concentrations were 1.8 and 0.36 n g ~ m - ~ respectively for the hydride and alkyl generation systems whilst detection limits were 0.62 and 0.07 n g ~ m - ~ . The use of compromise conditions for alkyl vapour generation ie. [H"] = 20 mmol dm-3 resulted in a decrease in characteristic concentration to 0.75 ng ~ 1 1 1 ~ ~ .The use of the Pb 217 nm line gave superior sensitivity improving the characteristic concentration by a factor of approximately 2.4 ie. to 0.76 ng cm-3 (hydride) and 0.145 ng cmF3 (alkyl). Determination of Lead in a Reference Water Lead was determined in a National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1643c Trace Elements in Water by alkyl generation AAS. The sample was diluted by a factor of two to allow determination of lead on the rectilinear portion of the calibration curve. The reference material (NIST 1643c) is preserved in approximately 0.5 mol dmV3 nitric acid Le. [H+]=0.25 mol dmW3 in the diluted sample. Consequently the reagent alkalinity was increased to 0.55 mol dmP3 NaOH to allow for this increased acidity when compared with previously determined conditions (Table 3).Experimental conditions for the determination of lead in NIST 1643c are shown in Table 4. Lead was determined in NIST 1643c by alkyl vapour generation AAS using direct calibration. The lead content of this water was found to be 34.8 1.4 ng cm-3 (n = 5 uncertainty based upon 20) compared to the certificate value of 35.3 kO.9 ng ~ r n - ~ . Conclusions The alkyl generation method is sufficiently sensitive to form the basis of a method for the determination of lead in drinking waters at the new levels recommended by the World Health Organization. 1 2 3 4 5 6 7 8 9 10 References Thompson K. C. and Thomerson D. R. Analyst 1974,58 1138. Fleming H. D. and Ide R.G. Anal. Chim. Acta 1976 83 67. Vijan P. N. and Wood G. R. Analyst 1976 101 966. Castillo J. R. Mir J. M. Martinez C. Val J. and Colon M. P. Mikrochim. Acta 1985 1 253. Li J. Liu Y. and Lin T. Anal. Chim. Acta 1990 231 151. Jin IS. and Taga M. Anal. Chim. Acta 1982 143 229 Thao R. and Zhou H. Fenxi Huaxue 1985 13 283. Madrid Y. Meseguer J. Bonilla M. and Camara C. Anal. Chim. Acta 1990 238 181. Li J.-X. Liu Y.-M. and Li T.-Z. Anal. Chim. Acta 1990,231,151. Honeycutt J. B. and Riddle J. M. J. J. Am. Chem. Soc. 1961 83 369.1420 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 0.4 0.3 0.2 0.1 0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 [Hydrogen peroxide1 (%) Fig. 4 Univariate search of oxidant concentration around the simplex derived optima for (a) hydride and (b) alkyl lead vapour generation AAS see text for explanation of arrow and vertical lines 0 .4 ~ 0.3 o.2 0.1 I 0 200 400 600 800 Purge gas flow/cm3 min-' 1000 Fig. 5 Univariate search of purge gas flow rate around the simplex derived optima for (a) hydride and (b) alkyl lead vapour generation AAS see text for explanation of arrow and vertical lines oxidant concentration for the alkylation system is beneficial in terms of reduced blank levels. Overall the alkylation chemistry is significantly more sensi- tive than the hydride. If the absolute sensitivity of the method is compromised in favour of a more stable behaviour with respect to sample acidity [Fig. 4(b)] then the experimental conditions are more robust than the equivalent hydride system. Table4 Experimental conditions for the determination of lead in NIST 1643c by alkyl vapour generation AAS Reagent Purge gas [ NaTEB]/ alkalinity/ [Oxidant] flow/ g dm-3 mol dmP3 (% v/v) cm3 min-l 7.5 0.55 0.275 300 Vapour generator settings Delay/s 10 Read/s 60 Rise/s 30 Decay/s 40 Reagent stream flow rate/cm3 min-' 4 Sample/blank flow rate/cm3 min-' 8 Detection Limits and Calibration The detection limits and characteristic concentrations of lead by vapour generation AAS were determined for the alkyl and hydride chemistries at the simplex derived optima.The detec- tion limit was defined as 3sb. The characteristic concentrations were 1.8 and 0.36 n g ~ m - ~ respectively for the hydride and alkyl generation systems whilst detection limits were 0.62 and 0.07 n g ~ m - ~ .The use of compromise conditions for alkyl vapour generation ie. [H"] = 20 mmol dm-3 resulted in a decrease in characteristic concentration to 0.75 ng ~ 1 1 1 ~ ~ . The use of the Pb 217 nm line gave superior sensitivity improving the characteristic concentration by a factor of approximately 2.4 ie. to 0.76 ng cm-3 (hydride) and 0.145 ng cmF3 (alkyl). Determination of Lead in a Reference Water Lead was determined in a National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1643c Trace Elements in Water by alkyl generation AAS. The sample was diluted by a factor of two to allow determination of lead on the rectilinear portion of the calibration curve. The reference material (NIST 1643c) is preserved in approximately 0.5 mol dmV3 nitric acid Le.[H+]=0.25 mol dmW3 in the diluted sample. Consequently the reagent alkalinity was increased to 0.55 mol dmP3 NaOH to allow for this increased acidity when compared with previously determined conditions (Table 3). Experimental conditions for the determination of lead in NIST 1643c are shown in Table 4. Lead was determined in NIST 1643c by alkyl vapour generation AAS using direct calibration. The lead content of this water was found to be 34.8 1.4 ng cm-3 (n = 5 uncertainty based upon 20) compared to the certificate value of 35.3 kO.9 ng ~ r n - ~ . Conclusions The alkyl generation method is sufficiently sensitive to form the basis of a method for the determination of lead in drinking waters at the new levels recommended by the World Health Organization. 1 2 3 4 5 6 7 8 9 10 References Thompson K. C. and Thomerson D. R. Analyst 1974,58 1138. Fleming H. D. and Ide R. G. Anal. Chim. Acta 1976 83 67. Vijan P. N. and Wood G. R. Analyst 1976 101 966. Castillo J. R. Mir J. M. Martinez C. Val J. and Colon M. P. Mikrochim. Acta 1985 1 253. Li J. Liu Y. and Lin T. Anal. Chim. Acta 1990 231 151. Jin IS. and Taga M. Anal. Chim. Acta 1982 143 229 Thao R. and Zhou H. Fenxi Huaxue 1985 13 283. Madrid Y. Meseguer J. Bonilla M. and Camara C. Anal. Chim. Acta 1990 238 181. Li J.-X. Liu Y.-M. and Li T.-Z. Anal. Chim. Acta 1990,231,151. Honeycutt J. B. and Riddle J. M. J. J. Am. Chem. Soc. 1961 83 369.

ISSN:0267-9477    

DOI:10.1039/JA9940901417

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1423 Determination of Lead in Wine Other Beverages and Fruit Slurries by Flow Injection Hydride Generation Atomic Absorption Spectrometry with On-line Microwave Digestion Carmen Cabrera Departamento de Nutricidn y Bromatologi Facultad de Farmacia Universidad de Granada 18071 Granada Spain Yolanda Madrid and Carmen Camara* Departamento de Quimica Analitica Facultad de Quimicas Universidad Complutense de Madrid 28040 Madrid Spain A simple and rapid flow injection-lead hydride generation atomic absorption spectrometry (FI-HG-AAS) method was optimized for the determination of lead in wine other beverages and fruit. Lead hydride was generated in HN0,-H202 medium using NaBH as reducing agent. To determine lead in beer juice and fruit a microwave oven was coupled on-line to the FI-HG-AAS system. For fruit the lead hydride was generated from slurries of the fresh sample. No matrix effect was found in the determination of lead.The method enabled the direct determination of lead in untreated samples with use of an aqueous calibration graph. The detection limit was 10 pg I-' in wine and other beverage samples and 1 .O ng in fruit. Keywords Lead; atomic absorption spectrometry; hydride generation; flow injection; wine; beverages; fruit The toxic effect of lead at low concentrations in the environ- ment is a cause of increasing concern and explains the growing interest in determining its concentration in foodstuffs.' The trend in dietary behaviour in the last few years has been one of increased consumption of fruit and of beverages such as wine beer soft drinks and juices.Although the amount of lead in these samples is very low the daily dietary intake may be physiologically significant. Therefore the determination of lead in such samples requires the use of techniques which provide high sensitivity and low detection limits. Traditionally electrothermal atomic absorption spec- trometry (ETAAS) has been applied in such cases,2 but numer- ous interferences have been rep~rted.~ An alternative method for the determination of lead at low concentrations in the samples of interest is hydride generation atomic absorption spectrometry (HG-AAS). The use of flow injection HG-AAS (FI-HG-AAS) for lead determination has been described by several workers4*' and has been shown to provide a number of advantages over conventional batch operation including higher sensitivity (thereby enabling the use of small sample volumes) lower reagent consumption higher sample through- put and easy automation. As a rule HG-AAS requires the mineralization of solid samples prior to analysis which increases both analysis time and the risk of sample c~ntamination.~.~ Pre-treatment of powdered solid samples by slurrying in liquid medium over- comes these problems and has the advantages of rapid analysis simplicity and reduction of blank level^.^.^ Slurry formation FI-HG-AAS appears to be an attractive method for the determination of lead in complex matrices because it combines simple sample treatment high sensitivity and selectivity and high sampling frequency.In the present work an FI-HG-AAS system was optimized to determine lead in wine other beverages and fruit. Lead hydride was generated in HN03-H202 medium using NaBH as reducing agent.' Lead hydride was directly generated from liquid samples and from the slurries of fresh fruit samples. The concentrations of HNO and H20z were optimized for each type of sample. In order to increase the lead hydride generation efficiency in beer juice and fruit samples a microwave (MW) oven was coupled on-line to the FI-HG-AAS system upstream from the H202 and NaBH addition step. The accuracy * To whom correspondence should be addressed. precision and selectivity of the proposed method were evalu- ated. The results obtained were compared with those obtained by ETAAS.Experimental Apparatus A Perkin-Elmer Model 2380 atomic absorption spectrometer equipped with an electrodeless discharge lamp operated at 10 W from an external power supply was used for all determi- nations. A spectral bandwidth of 0.7 nm was selected to isolate the 217.0 nm lead line. The signals were recorded on a Perkin- Elmer Model 56 recorder set at the 10 mV range. Background correction was not used. The FI manifold is shown in Fig. 1. It comprises of a four- channel peristaltic pump (Gilson HP4) a six-way valve (Omnifit) for sample injection a mixing and reaction coil (Teflon tubing 0.5 and 0.8 mm internal diameter) and a U-tube gas-liquid separator (Philips). A domestic Balay MW oven (650 W power) and an ice-bath were incorporated into the system for beer juice and fruit samples (Fig.2). A loop of poly (tetrafluoroethylene) (PTFE) tubing was placed inside the MW oven through the ventilation holes. Lead hydride was carried by argon to the quartz atomization cell heated by an acetylene-air flame. A Perkin-Elmer HGA-400 graphite furnace with pyrolytic tubes and L'vov platform at 283.3 nm and a spectral bandwidth of 0.7nm were used for method validation. A deuterium arc background corrector and an electrodeless discharge lamp were used. An aluminium block and Pyrex tubes were used for solid sample mineralization. Sample I Ar Fig. 1 Diagram of the FI-HG-AAS manifold for determination of lead in wine and soft drinks1424 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 Sample Peristaltic Pump Fig.2 Diagram of the FI-MW-HG-AAS manifold for determination of lead in beer juice and fruit slurries Reagents All the reagents were of analytical-reagent grade or higher purity and de-ionized water from a Milli-Q system (Millipore) was used. A standard 1OOOmg 1-l lead(I1) solution (Merck) was used. Working solutions were prepared each day by diluting appropriate aliquots of the standard solution. A 6% m/v sodium tetrahydroborate solution was prepared by dis- solving sodium tetrahydroborate(1Ir) powder (Aldrich) in de-ionized Milli-Q water and stabilizing in 1 YO sodium hydrox- ide (Merck). Solutions were prepared daily and filtered before use." Nitric acid solutions of several concentrations were prepared by appropriate dilution of 65% v/v nitric acid (Carlo Erba).Hydrogen peroxide solutions of varying concentrations (v/v) were prepared by appropriate dilution of the stock solution. Triton X-100 (Serva) and 30% m/v silicone antifoam- ing emulsion in water (Fluka) were also used. Sample Preparation Slurry procedure The fruit samples were thoroughly washed or peeled; 1.Og portions of powdered and homogenized fresh sample were weighed accurately and placed in small polyethylene bottles with 10.0 g of blown zirconia spheres (Glen Creston Stanmore UK) after which 5.0 ml of 1.0% m/v Triton X-100 solution were added. The bottles were shaken for about 10 min in a flask shaker until a slurry formed. This grinding procedure ensured that 90% of the particles had a diameter of less than 25 pm. This particle size was small enough to determine lead by HG-AAS.7 Slurries were separated from the zirconia spheres using a Buchner funnel and transferred into a calibrated flask.A few drops of 30% m/v silicone antifoaming emulsion in water were added before the slurry was diluted. The resulting suspensions were analysed directly by ETAAS and HG-AAS. Nitric acid mineralization in an aluminium block A 5 ml portion of wine or other beverages was treated with 5ml of 65% v/v nitric acid in a pyrex tube placed in an aluminium block and heated at 160°C for 3 h. The solutions were left to cool to room temperature transferred into a calibrated flask and diluted to the mark with deionized Milli-Q water. The resulting solutions were analysed by ETAAS. Sample Analysis Lead was directly determined by FI-HG-AAS in liquid and slurried fruit samples.The soft drinks and beer samples were degased in an ultrasonic bath prior to analysis in order to increase the precision of the determinations. For all analyses 100 pl of liquid or slurried sample were injected into a continu- ous flow of nitric acid. In beer juice and slurried fruit samples the acidic solution was introduced into a MW oven and then cooled in an ice-bath to reduce the amount of water vapour (Fig. 2). The resulting solution was mixed with H202 solution and 6% m/v NaBH for lead HG. Lead hydride was carried by argon to the quartz atomization cell heated by an acety- lene-air flame for atomization. Analytical peaks were recorded as peak height. The optimum conditions for lead HG in each sample are summarized in Table 1.Ten distinct portions were taken from each sample and each portion was then analysed in triplicate. The same procedure was followed for the blanks. An aqueous calibration graph and the standard additions method were used. Results and Discussion Effect of Nitric Acid Hydrogen Peroxide and Sodium Tetrahydroborate Concentrations on FI Lead HG From Samples To determine the conditions giving the maximum sensitivity the effects of various chemical and physical parameters on lead HG from each type of sample were evaluated. For this study samples were spiked with 20-50 ng of lead stock solution in order to quantify the sample matrix effect. The effect of different nitric acid concentrations on the efficiency of HG was tested for each type of sample. Nitric acid is the most widely used reagent for lead HG in hydrogen peroxide medium.A concentration range of 5-50% v/v nitric acid was chosen for this study. The results plotted in Fig. 3 show that the absorbance increases rapidly with increasing HN03 concentration and that the optimum HN03 concen- tration depends on the type of sample being 10% v/v HN03 for wine and soft drinks and 30-40% v/v for beer juice and slurried fruit. Nitric acid concentration was the chemical parameter which most strongly affected the maximum lead hydride yield. This acid provides adequate acidity for lead HG and is an appropriate medium for sample mineralization. Hence the optimum nitric acid concentration depends on the complexity of each sample. The effect of hydrogen peroxide concentrations from 2 to 20% v/v on the efficiency of lead HG was also tested at the optimum HNO concentrations mentioned in the preceding paragraph.The results presented in Fig. 4 show that maximum sensitivity was obtained with 2% v/v hydrogen peroxide for wine and soft drinks. However for beer juice and fruit it was necessary to increase the H202 concentration to 15% v/v. This could be attributed to hydrogen peroxide decomposition by reducing agents present in the samples. The influence of H202 concentration was more marked for beer samples due to the presence of high concentrations of compounds such as sugars and organic and aromatic compounds. The effect of sodium tetrahydroborate on the HG yield is more important for lead than for other elements such as selenium or arsenic. The absorbance signal increased with NaBH concentration up to 6% where it levelled off.A concentration of 6% m/v NaBH was chosen as optimum for further experiments. The ability to choose this working concen- tration represents a considerable saving in analysis cost in comparison with the methods proposed by several workers for similar samples4~" or with batch mode analy~is.~ Also the low concentration of reducing agent provides lower blanks resulting in a lower detection limit. The optimum HN03 H202 and NaBH concentrations for each type of sample considered are summarized in Table 1. In order to evaluate the selectivity of the method for each sample standard additions graphs were prepared for blanks and for samples and used to calculate the blank-to-sample slope ratio for each sample (Table 2).No matrix effect was observed for wine and soft drinks as evidenced by the slope values of nearly 1. With beer juice and slurried fruit samples the method lacked the necessary sensitivity and selectivity for lead analysis. Effect of On-line Microwave Treatment on Lead HG Efficiency To overcome the matrix effect in the beer juice and slurried fruit samples a MW oven was coupled to the FI manifold as a simple way of sample pre-treatment before lead HG. Sample hold time in the MW oven was optimized by testing differentJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1425 0.16 0 (0 2 0.12 42 v) 2 0.08 0.04 Table 1 Chemical and physical parameters optimized for lead determination in wine other beverages and fruit by FI-HG-AAS and FI-MW- HG-AAS - - - - Lead HG* HNO HZOZ NaBH Sample ("/"v/v) (%v/v) (%m/v) Wine? 10 2 6 Soft drink? 10 2 6 Beer? 30 15 6 Juice? 40 15 6 Fruit 40 15 6 MW oven treatment Internal diameter Coil length/rn of tubingfmm Hold time/s _. 1.5 1.5 5.0 0.5 0.5 0.8 - 14 14 45 * Injection volume of 100 pl.t No pre-treatment. $ Slurry of edible portion of fresh sample. o.20 I A 10 20 30 40 50 Fig.3 Effect of nitric acid concentration on the efficiency of lead hydride generation in 6% m/v NaBH and the optimum hydrogen peroxide concentration for each type of sample (see Table 1) A wine; B soft drink; C juice; D beer; and E fruit slurry PTFE tube lengths and internal diameters (Table 1). The MW sample treatment enhanced the efficiency of lead HG by about 50%.In addition this treatment produced sample mineraliz- ation and accelerated the reaction kinetics thus increasing the reaction rate and giving rise to higher and sharper lead peaks. Furthermore selectivity was considerably improved by the MW oven step (blank sample slope ratio values of nearly 1 Table 2) which enabled lead determinations in this kind of sample to be read from a simple aqueous calibration graph. Sample Analysis The proposed method was applied in the determination of lead in some liquid (wine and other beverages see Table 3) 0.20 0.16 a 2 0.12 e s II 4 0.08 0.04 c A A D I I I I I I 0 4 8 12 16 20 [H,O,l (%I Fig. 4 Effect of hydrogen peroxide concentration on the efficiency of lead hydride generation in 6% m/v NaBH and the optimum nitric acid concentration for each type of sample (see Table 1) A wine; B soft drink; C juice; D beer; and E fruit slurry and solid samples (fruits see Table4).The former were ana- lysed directly without mineralization or preconcentration and the latter after slurrying the fresh sample (edible portion). Both types of sample were analysed by FI-HG-AAS or FI-MW- HG-AAS under the experimental conditions summarized in Table 1. The results were read off an aqueous calibration graph. The proposed method was validated by comparison with the ETAAS technique. Lead was determined in the liquid sample with and without mineralization and in slurried solid samples. An aliquot of 10 pl was injected into the tube and run under optimized ~0nditions.l~ Lead determination in wines by ETAAS requires the application of the standard additions method.Lead determination in other beverages (juice beer and soft drinks) requires sample pre-treatment by acid min- Table 2 Analytical characteristics for lead determination in wine other beverages and fruit by FI-HG-AAS and FI-MW-HG-AAS Detection Quantification Characteristic Precision Slope ratio Slope ratio Sample limit9 limit5 mass ("/.)TI without MW with MW Wine* 10 pg 1-1 40 pg I-' 0.8 ng 5-6 Soft drink* 10 pg 1-' 40 fig 1-' 1.0 ng 6-8 Beer? 10 pg I-' 45 pg 1-1 1.1 ng 5-6 Juice? 10 pg 1-1 45 pg 1-' 1.5 ng 6-9 Fruit$ 1 ng g-' 3.2 ng 2.0 ng 3-8 1.10 1.05 3.50 2.50 2.50 - 1.20 1.15 1.10 * No pre-treatment. ? Pre-treated in MW oven. $ Slurrying and MW pre-treatment. 9 Calculated according to IUPAC rules. 7 Ten replicate determinations on each of three different samples.1426 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 Table 3 Lead concentrations in wine and other beverages determined by FI-HG-AAS or FI-MW-HG-AAS and by ETAAS Lead/pg Sample Red wine White wine Orange juice (natural)$ Orange juice (commercial)$ Pineapple juice (commercial)$ Orange soft drink Lemon soft drink Beer$ HG 59+1 9 6 f l 60+ 1 50+ 1 98f1 66? 1 41f1 ndt GF 5 8 f l 9 8 f l ndt 62f 1 55+1 100fl 63f1 3 5 f l * Mean value f standard deviation at the 95% confidence level t Not detectable. $ Pre-treated in MW oven. (Student’s t ) n= 10. Table4 Lead concentrations in fruit (microgram per gram of fresh weight of edible portion) determined by FI-MW-HG-AAS and by ETAAS Lead/pg g-’* Sample Scientific name HG GF Kiwi Actinidia deliciosa 0.38 & 0.02 0.31 fO.O1 Banana Musa paradisiaca 0.22 * 0.02 0.20 f 0.02 Strawberry Fragaria sp 0.21 + 0.02 0.20 f 0.01 Mango iVanganifera laurina 0.39 0.01 0.40 + 0.01 Custardapple Annona Cherimola 0.20 k 0.03 0.22 + 0.01 Pineapple Ananas sativus 0.30 f 0.01 0.31 + 0.01 * Mean value f standard deviation at the 95% confidence level (Student’s t ) n = 10.eralization. In the fruit samples lead was directly determined from slurries of fresh samples. For mineralized and slurried samples the standard additions method can be dispensed with. The results for the two methods are presented in Tables 3 and 4. Comparison by the F-test revealed no significant differences at the 95% confidence level so it was concluded that the method can be applied as an alternative to ETAAS.Analytical Characteristics of Lead Determination by FI-HG-AAS or FI-MW-HG-AAS The characteristic mass and precision of the analytical method were evaluated and the detection and quantification limits were calculated according to IUPAC rules.12 The results are summarized in Table 2. Conclusions By modifiying the chemical and physical conditions of lead HG and sample pre-treatment the proposed method could also be used for lead determination in other similar samples. The authors wish to thank Direcion General de Ciencia Tecnologia (DGCYT) for financial support under contract PB 92/0218 and Max Gorman for revising the manuscript. 1 2 3 4 5 6 7 8 9 10 11 12 13 References Concon J. M. Food Toxicology Contaminant Additives Dekker New York 1988. Welz B. Atomic Absorption Spectrometry VCH Weinheim 1985. Aroza I. Bonilla M. Madrid Y. and Camara C. J. Anal. At. Spectrom. 1989 4 163. Baluja C. and Gonzalez A. Talanta 1992 39 329. Welz B. and Guo T. Spectrochim. Acta Part B 1992 47 645. Dabeka R. W. and Makenzie A. D. J. Spectrosc. 1986 31 44. Madrid Y. Bonilla M. and Camara C. J. Anal. At. Spectrom. 1989 4 167. Madrid Y. Bonilla M. and Camara C. Analyst 1990 115 563. Madrid Y. and Camara C. Analyst 1994 119 1647. Knechtel J. R. and Fraser J. L. Analyst 1978 103 104. Cacho J. Ferreira V. and Nerin C. Analyst 1992 117 31. Long L. G. and Winefordner J. D. Anal. Chem. 1983,55 713A. Cabrera C. Lorenzo M. L. Gallego C. and Lopez M. C. Anal. Chim. Acta 1991 246 375. Paper 4/03457K Received June 8 1994 Accepted July 26 1994

ISSN:0267-9477    

DOI:10.1039/JA9940901423

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1427 Determination of Lead at Low Concentrations in Food Samples by Electrothermal Atomic Absorption Spectrometry R. Tahvonen and J. Kumpulainen Laboratory of Food Chemistry Food Research Institute Agricultural Research Centre of Finland FIN-31 600 Jokioinen Finland A method for the determination of Pb in biological materials by electrothermal atomic absorption spectrometry was tested at a wavelength of 217.0 nm and with 0.1 YO Pd as the chemical modifier. A Varian SpectrAA 400 Zeeman spectrometer equipped with a Photron Pb Superlamp was employed. As the contents of Pb in most biological materials are near the limit of detection peak height measurement and the method of standard additions were used. The validity of the method was tested using the following ARWCL-coded reference materials wheat flour potato powder milk powder and animal muscle in addition to various other commercial reference materials.Recovery tests were made on some reference materials. The accuracy of the method was satisfactory for the tested materials. High lamp intensity and the most sensitive Pb wavelength resulted in better characteristic masses; the characteristic mass in peak height mode at 283.3 nm with a hollow cathode lamp was usually 5-8 pg and at 217.0 nm with the Superlamp 3-5 pg depending on the sample matrix. Low backgrounds were measured in the sample materials. Ammonium phosphate cannot be employed as the chemical modifier when using the 217 nm line as phosphate species cause spectral interferences at that wavelength.Keywords Lead; food samples; chemical modification; palladium; graphite furnace; electrothermal atomic absorption spectrometry The level of Pb pollution in the and conse- quently in foods3-'0 has decreased particularly in Europe and the USA due to the implementation of stricter control meas- ures governing industrial and automobile exhaust emissions in recent years. However Lockitch" and Goyer12 have pointed out in their recent reviews that even very low Pb exposure can have several kinds of adverse effects and Carrington et concluded that the acceptable blood Pb or exposure levels defined are considerably lower than those actually measured in many human populations. Hence there is still a clear need to determine increasingly lower Pb concentrations in foods and other biological mate- rials.Determination of Pb in biological matrices is not an easy task even at moderate concentrations. The General Referee Report for lead of the Association of the Official Analytical Chemistsi4 concluded in the 1993 Report that a more accurate and sensitive electrothermal atomic absorption spectrometry (ETAAS) method is especially needed for the determination of Pb in foods. Slavin and Manning15 introduced the so-called stabilized temperature platform furnace (STPF) concept for ETAAS. Koirtyohann et all6 were the first to successfully employ the STPF concept and NH4H2P04 chemical modification for the determination of Pb in food matrices. Since then the basic principle has been applied in many laboratories including ours for the determination of Pb in However the method has a number of problems. First the sensitivity is not sufficient for all types of foods because the most sensitive Pb resonance line 217 nm cannot be employed because spectral interferences due to various phosphorus-oxygen compounds and other species are encountered when Zeeman-effect back- ground correction is used.Zeeman-effect background correc- tion is particularly valuable for the determination of Pb in complex inorganic matrices that produce large backgrounds. In addition we have noticed some accuracy problems when determining Pb in milk and meat samples in which the results have tended to be systematically too high.6 Tsalev et ~ 1 . ' ~ have reviewed the use of chemical modifiers and reported that NH4H2P04 and/or Mg(N03)2 have been employed most frequently for food matrices.In much recent work with biologi- cal materials Pd alone or in combination with some other reagents has been The charring temperature with Pd modifiers can be 1000°C or higher thus enabling the efficient reduction of interferences. However all these workers employed the less sensitive wavelength 283.3 nm for determi- nations of Pb. Welz et ~ 1 . ~ ~ have found that when the higher atomization temperatures needed with the Pd modifier are employed the characteristic mass (m,) values are higher due to significantly higher diffusional losses. In order to be able to decrease the limit of determination we have established a method employing the 217 nm wavelength which is a more sensitive line than 283.3 nm and Pd as the chemical modifier and the results are reported in the present paper. Materials and Methods The feasibility of using the 217 nm wavelength in determi- nations of Pb was tested employing a Varian SpectrAA 400 Zeeman spectrometer equipped with an autosampler and a Photron Pb Superlamp.The mixed chemical modifier consisted of 1000mg 1-' of Pd 1% hydroxylamine hydrochloride and 2.5% HCl.23 Our previous method employing 1% (NH4)H2P04 as the modlfier and the 283.3 nm wavelength served as the reference method. Some determinations were made using the mixed Pd modifier at 283.3 nm as in another earlier method. Pyrolytic graphite coated graphite tubes and pyrolytic graphite platforms (part numbers 63-10OO11-00 and 63-100013-00 Varian) were employed during the deterrni- nations.All measurements were made using two replicates. Determinations were carried out in the peak height mode employing the method of standard additions. A typical furnace programme is presented in Table 1. We analysed several reference materials (RMs) coded Agricultural Research Centre Central Laboratory (ARC/CL) potato powder ARC/CL animal muscle ARC/CL wheat flour and ARC/CL milk powder for which recommended values have been established based on interlaboratory comparison studies and other required certification p r o c e d ~ r e s . ~ ~ - ~ ~ These materials have been composed from typical commercial prod- ucts and thus ideally represent actual food samples both in terms of matrix composition and concentration levels.In addition various commercial reference materials National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1549 Non-fat Milk Powder NIST SRM 1577b Bovine Liver and Community Bureau of1428 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 ( a ) Table 1 modifier 0.1% Pd) Typical furnace programme (sample ARC/CL wheat flour (6) 0 0.70 ( C) Parameter Value Instrument mode Calibration mode Measurement mode Lamp current/mA Slit width/nm Slit height Wavelength Sample introduction Time constant Measurement time/s Replicates Background correction Maximum absorbance Absorbance Standard additions Peak height 8 1 .o Normal 217.0 Sampler automixing 0.1 1 .o 2 On 1.40 Furnace parameters Step "C Time/s 1 min-' Gas type command Temperature/ Gas flow rate/ Read 1 2 3 4 5 6 7 8 9 10 11 220 280 400 1 loo* 1 loo* 100 100 2400-f 2400t 2650 2650 18.0 30.0 10.0 5.0 20.0 6.5 5.0 1.3 3.0 2.0 3.0 3.0 3 .O 3.0 3.0 3.0 3.0 0.0 0.0 0.0 3.0 3.0 Normal Normal Normal Alternate Alternate Alternate Alternate Alternate Alternate Normal Normal No No No No No No No Yes Yes No No Sampler parameters (volumes in p1) Blank - 10 10 10 Addition 1 5 10 10 Addition 2 10 10 10 10 Sample - Standard Sample Blank Modifier - - - - Multiple inject Hot inject Temperature Inject rate Pre-inject Yes Yes 140°C 5 No * <900 with phosphate modifier.-f <2200 with phosphate modifier. Reference (BCR) 184 Bovine Muscle were analysed. Some samples were spiked with standard solutions to determine the recoveries. Samples (1-2 g) were digested in 0.5 1 glass tubes covered with a separatory funnel using 10ml of concentrated HNO (p.a.-plus Riedel-de-Haen) on an aluminium heating-block at 50 "C overnight.On the following morning the temperature was slowly increased to the boiling-point for about 2 h to complete the digestion and evaporate the volume to about 2ml. The solution was then diluted to 50ml in a volume bottle. All containers were washed with three acids [lo% HNO followed by 10% HCl and 3% HNO (p.a.-plus)] and rinsed with distilled water. Palladium concentrations of 0.05 and 0.2% were tested with potato powder animal muscle and milk powder RMs to determine the optimum amount of Pd. A repeatability test was performed by analysing an ARC/CL wheat flour sample 19 times in succession. Results and Discussion Figs.1 and 2 show the Pb peaks of milk samples+additions at 283.3 and 217 nm using (NH4)H2P04 and Pd as chemical modifiers. Serious interferences can be seen at 217 nm with the phosphate modifier. Both the sample solution and the graphite tubes used were the same at both wavelengths but the tempera- (a) VOL. 9 3000 0 I 3000 i i 0.70 0.70 3000 ( C ) -0.05 0 ' 0 7.3 Time/s Fig. 1 Lead peaks at (a) 283.3 nm; (b) 217 nm; and (c) 217 nm [with addition of 5 pl of Pb standard solution (20 pg 1-')I of a milk sample using phosphate modifier [ 1 YO (NH,)H,PO,] I /--I - 0.05 - 0.05 0 -. L E E c P t-" Fig. 2 Lead peaks at (a) 283.3 nm; (b) 217 nm; and (c) 217 nm [with addition of 5 pl of Pb standard solution (20 pg 1-')I of a milk sample with the mixed Pd modifier (0.1 7'0 Pd)JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 ture programme was optimized for both modifiers. Background absorption was much smaller at 217nm when using the Pd modifier (Fig. 2) and the absence of negative peaks may have been due to drastically lower phosphate concentrations in the samples. The higher temperatures used with the Pd modifier enabled the removal of more of the interfering matrix. With the Pd modifier the best peak shape was found with a sample solution containing 20 pg of Pd in the furnace (Fig. 3). The characteristic mass of Pb however is higher with such a large amount of Pd. In this sample the mass of milk powder was 2.274 g per 50 ml of solution. It is advisable to keep the sample masses of milk samples slightly lower than 2 g and the Pd mass 1Opg to obtain the best sensitivity at the 217nm wavelength.Serious interferences were also found when analysing potato samples with the phosphate modifier at 217 nm (Fig. 4). When the Pd modifier was used the shape of the peak was excellent (Fig. 5 ) . Some interferences were also visible with animal muscle samples when the phosphate modifier was used at 217 nm. Good peak shapes were obtained with the Pd modifier. According to Massmann et ~ 1 . ~ ~ the most frequently observed background interferences caused in flames as well as in furnaces by the structured absorption of molecules are those of the pyrolysis products of -SO4 -NO and -PO4. Diatomic mol- ecules such as SO NO and PO show very sharp rotational structures in the ultraviolet wavelength region.It is very likely that a large number of these compounds can arise during the digestion of biological samples and during the sample pre- treatment (drying ashing) in the furnace. Slavin and Manning15 have tested the background absorption of several matrices at the 217.0 and 283.3nm Pb lines. These background signals may be due to scattering by solid particles or to molecular absorption. However they did not attempt to distinguish between the two effects in that study. The background for phosphate and sulfate materials was much smaller at 283.3 nm. The 217.0nm line may be preferable if the solutions contain mostly chloride. According to Slavin and Manning15 use of 0.70 -0.05 0.70 aJ c 6 3 * 8 a a -0.05 0.70 - 0.05 I I 3000 0 3000 P 2 t? -.+ E" c 0 3000 0 0 7.3 Time/s Fig. 3 Lead peaks at 217 nm of a milk sample using (a) 5 p1; (b) 10 p1; and (c) 20 p1 [with addition of 5 pg of Pb standard solution (20 pg l-')] of the mixed Pd modifier (0.1 % Pd) 0.70 I (a) - 0.05 4 0 7.3 Time/s Fig. 4 Lead peaks at (a) 283.3 nm; (b) 217 nm; and (c) 217 nm [with addition of 5 pl of Pb standard solution (20 pg I-')] of a potato sample using phosphate modifier [ 1% (NH,)H,PO,] 0.70 -0.05 0.70 al C m e z a il ( a ) I I -0.05 p I- -0.05 0 7.3 Time/s Fig. 5 Lead peaks at (a) 283.3 nm; (b) 217 nm; and (c) 217 nm [with addition of 5 p1 of Pb standard solution (20 pg l-')] of a potato sample with the mixed Pd modifier (0.1% Pd)1430 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 L'vov platform removes the interferences from chlorides but not those from sulfates or phophates.They suggested that the probable reason is that phosphate and sulfate interferences result from a shift of the appearance temperature for Pb. Background measurement by Zeeman techniques may cause serious systematic errors if the spectral background is due to sharp rotational lines of molecules that display a Zeeman effect.28 Many small molecules have a Zeeman effect. It seems that the backgrounds and the interferences are more serious at 217 nm whereas at 283.3 nm phosphate modifiers can be employed successfully (Figs. 1 3 and 5). Also the use of the higher ashing temperatures allowed when employing the Pd modifier may further remove the interferences (Figs. 2,3 and 5 ) . The mechanisms of how Pd modifiers work in the furnace have been studied by many investigators.According to Shan and WangZ9 a very stable compound or alloy between Pb and Pd is formed which is an intermediary in the formation of Pb atoms. Volynsky et ~ 1 . ~ ' have concluded that the high efficiency and relative universality of Pd modifiers can be explained not only by the easy formation of Pd metal from its compounds but also by the unique catalytic properties of Pd. According to Yang et the reduced Pd and analyte form a stable intermetallic solid solution. According to Dabeka,32 citric acid magnesium nitrate and ammonium dihydrogen phosphate prevent the formation of refractory Pb peaks in the presence of Pd. Therefore it is conceivable that if Pd is used as a chemical modifier the Pb sensitivity in samples and standards will differ because the samples may have constituents (e.g. traces of citrate phosphate or magnesium nitrate) which prevent the formation of the refractory species.To correct for the possible refractory behav- iour of Pb the method of standard additions was used for quantification. The absorbances were too low to measure in the peak area mode which is often recommended as it is believed to eliminate the errors caused by different peak appearance times or peak shapes of different chemical forms of the analyte. Results obtained for the RM samples are presented in Table 2. Even the very low Pb concentrations of the ARC/CL milk powder ARC/CL wheat flour and ARC/CL potato powder as well as in the NIST SRM 1549 were successfully determined.Absorbance values were usually lower than 0.020 even with double injection in which 20 pl sample volumes were deposited on the platform. Results of the presently analysed reference materials were very close to the certified values. Table 3 shows the repeatability results for the ARC/CL wheat flour reference material. The relative standard deviation (RSD) was < 10% even with 10 pl sample volumes and absorbance readings below 0.010. The characteristic mass m of Pb when using the Pd modifier and a Superlamp at the 217nm wavelength is 3-5pg and at 283.3nm 5-8pg depending on the sample matrix. Values of digestion blanks and limits of detection (LODs) and quantitation (LOQ) are presented in Table 4. The limit of Table 3 Repeatability results of an ARC/CL wheat flour reference material sample* No.of determin- ations 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. Mean SD RSD RSD Mean Readings Pb/ Concentration/ (YO) absorbance 1 2 CLi3kg-l CLg1-l 23.2 19.9 20.1 19.9 21.2 22.9 22.9 22.3 20.7 21.2 19.2 24.0 16.4 19.2 26.0 21.2 21.4 21.4 23.1 21.4 pg kg-' 2.12 pg kg-' 9.91 % 1.06 0.91 0.92 0.91 0.97 1.05 1.05 1.02 0.95 0.97 0.88 1.14 0.75 0.88 1.30 0.97 0.98 0.98 1.06 1.7 9.9 13.4 5.9 10.8 10.6 43.2t 19.4 19.6 6.3 4.0 23.9t 2.7 7.9 19.9 28.2 16.3 17.0 4.8 0.0 10 0.008 0.008 0.008 0.009 0.009 0.009 0.009 0.009 0.009 0.008 0.010 0.007 0.008 0.012 0.009 0.009 0.009 0.009 0.009 0.010 0.008 0.009 0.009 0.007 0.009 0.008 0.008 0.009 0.010 0.009 0.012 0.007 0.010 0.008 0.007 0.010 0.008 0.008 0.008 0.008 0.012 0.009 0.007 0.007 0.008 0.007 0.010 0.013 0.010 0.007 0.010 0.008 0.010 0.008 0.010 0.009 ~~~ * The furnace programme corresponded to that presented in Table 1 except multiple injections were not employed.t Measurement was repeated if the RSD of the replicates was > 20% when the concentration was higher than 1.00 pg 1-' (quality control options). Table 4 Blank values measured with different sample calibrations Mean SD LOD (3s) LOQ (10s) Pb/pg I-' -0.09 - 0.06 -0.08 - 0.08 -0.01 0.01 - 0.03 - 0.04 0 0.01 0.02 -0.01 - 0.03 0.039 0.118 0.393 Table 2 were 92-98% Pb contents of RMs determined at 217 nm using 1% Pd modifier. Recoveries determined with wheat flour and animal muscle RMs Reference Material ARC/CL potato powder set 1 ARC/CL potato powder set 2 ARC/CL wheat flour ARC/CL pork meat ARC/CL skim milk powder NIST 1549 Non-fat Milk Powder NIST 1577b Bovine Liver BCR 184 Bovine Muscle Mean/ 19.1 23.7 21.0 89.9 16.7 19.9 CLg kg-' 131 266 SD/w kg- ' 2.71 1.44 2.86 13.6 - 2.8 1 8.90 22.3 RSD 14.2 6.1 13.6 15.1 14.1 6.8 8.4 (%I -* Certified value/ 26 5 2.8 26 f 2.8 18+7 89+ 13 16+3 19f3 129+4 239+11 kg-' ~~ ~~ * RSD of the same solution in a different calibration 16-17%.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 1431 Table 5 Statistics of determinations of lead (pg kg-') in milk samples (for duplicates 1 and 2) using phosphate modifier and Pd modifier Phosphate modifier (283.3 nm) ~~~ Pd modifier (217 nm) Sample 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 21 28 29 30 Mean 1 17.9 27.8 25.5 16.2 15.1 4.3 8.2 7.5 0.2 12.0 10.4 10.3 2.8 0.8 0.4 26.6 13.5 22.1 16.7 14.9 10.6 4.4 16.5 13.4 18.5 15.5 2.5 4.1 13.6 14.5 12.2 2 19.3 29.9 17.6 11.5 7.1 6.1 13.3 10.1 1.5 13.8 11.7 6.1 1.9 7.2 25.0 9.7 19.6 9.6 15.3 11.8 7.3 31.8 19.1 21.0 15.4 3.8 2.5 12.3 15.4 27.7 13.5 Mean 18.6 28.9 21.6 13.9 11.1 5.2 10.7 8.8 0.9 12.9 11.0 8.2 2.4 4.0 12.7 18.2 16.5 15.9 16.0 13.3 8.9 18.1 17.8 17.2 17.0 9.6 2.5 8.2 14.5 21.1 12.9 SD 1.0 1.5 5.6 3.3 5.6 1.3 3.5 1.9 0.9 1.2 0.9 3.0 0.6 4.5 17.4 11.9 4.3 8.8 1 .o 2.2 2.3 19.4 1.8 5.4 2.2 8.3 0.0 5.8 1.3 9.4 4.5 RSD 5.6 5.2 26.0 23.7 50.5 24.3 33.0 21.1 108.1 9.5 8.2 36.7 25.9 113.5 136.9 65.6 26.0 55.5 6.1 16.2 26.2 107.0 10.4 31.3 13.0 86.4 0.0 70.2 9.0 44.4 39.9 1 14.4 17.4 22.2 13.3 13.6 16.5 15.6 14.6 15.1 16.5 21.6 20.2 15.1 15.3 13.4 16.8 13.3 12.2 14.8 11.6 13.7 23.9 16.9 17.8 19.2 13.1 14.9 15.4 19.5 16.9 16.2 2 12.3 17.1 19.6 16.0 17.8 13.3 18.1 18.4 18.4 20.9 23.8 16.3 16.6 12.1 13.3 13.5 18.7 11.9 19.6 15.4 13.3 24.5 11.4 15.9 14.1 12.3 11.1 10.7 12.5 23.1 16.1 Mean 13.4 17.2 20.9 14.7 15.7 14.9 16.9 16.5 16.8 18.7 22.7 18.3 15.9 13.7 13.4 15.1 16.0 12.1 17.2 13.5 13.5 24.2 14.1 16.9 16.6 12.7 13.0 13.1 16.0 20.0 16.1 SD 1.5 0.2 1.8 1.9 3.0 2.3 1.8 2.7 2.3 3.1 1.6 2.8 1.1 2.3 0.1 2.4 3.8 0.2 3.4 2.7 0.3 0.4 3.9 1.3 3.6 0.6 2.7 3.3 4.9 4.4 2.2 RSD 11.1 1.4 8.8 13.0 18.9 15.2 10.5 16.3 13.9 16.6 6.9 15.1 6.7 16.7 0.5 15.7 23.9 1.8 19.8 19.9 2.1 1.8 27.8 8.0 21.7 4.5 20.7 25.5 31.0 21.9 13.9 detection for a 2 g sample mass is 3 pg Pb per kg dry matter and the LOQ is 10 yg Pb per kg dry matter.The results of Pb determinations using the phosphate modi- fier at 283.3 nm and with Pd modifier at 217 nm in real milk samples collected from Finnish dairies are compared in Table 5. As milk samples contain very low concentrations of Pb they can be considered to be nearly Pb-free samples. The mean RSDs of all samples already indicate the superiority of the method using the Pd modifier and the 217 nm line. The results can be used to calculate the LOD statistically estimating the RSD for the method assuming that the concentration affects it according to the following equation s = ( + k ) 100% The LOD of Pb in milk with phosphate modifier was 7.9 pg kg-l when determined with the phosphate modifier at 283.3 nm and 1.5 pg kg-l (LOQ 4.5 pg kg-') when determined with Pd modifier at 217nm.The optimum amount of hydroxylamine hydrochloride was 1 % in the present study. According to Beach23 higher amounts of the reducing agent may be necessary if a sample contains high concentrations of strongly oxidizing species. Beachz4 recommends concentrations of HN03 lower than 5%. References Riihling A. Brumelis G. Goltsova N. Kvietgus K. Kubin E. Liiv S. Magnusson S. Makinen A. Pilegaard K. Rasmussen L. Sander E. and Steinnes E. Atmospheric Heavy Metal Deposition in Northerno Europe 1990 Nord 1992:12. Nordic Council of Ministers. (Arhus Aka-Print A/S) 1992 41 pp. Makela-Kurtto R. Ervio R. and Sippola J. Agric. Sci. Finland 1993 2 337. 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Tahvonen R.and Kumpulainen J. Fresenius' J. Anal. Chem. 1991 340 242. Tahvonen R. and Kumpulainen J. Food Addit. Contam. 1993 10 245. Tahvonen R. and Kumpulainen J. Food Addit. Contam. 1994 in the press. Tahvonen R. and Kumpulainen J. Food Addit. Contam. 1994 11 415. Ellen G. Van Loon J. W. and Tolsma K. Zeitschrft Lebensm. Unters. Forsch. 1990 190 34. Elias R. W. International Conference on Heavy Metals in the Environment eds. Lindberg S. E. and Hutchinson T. C. New Orleans 1987 CEP Consultants Ltd Edinburgh pp. 197-202. The National Food Agency of Denmark Food monitoring in Denmark. Nutrients and contaminants 1983-1 987. Levnedsmid- delstyrelsen Sundhedsministeriet. Publication No. 195 September 1990. Von Weber O. Industriell Obst und Gemuseverwertung 1984 69 215.Lockiche G. Clin. Biochem. 1993 26 371. Goyer R. A. Environ. Health Persp. 1993 100 177. Carrington C. D. Sheehan D. M. and Bolger P. M. Food Addit. Contam. 1993 10 325. Capar S. G. J. AOAC Int. 1993 76 142. Slavin W. and Manning D. C. Anal. Chem. 1979 51 261. Koirtyohann S. R. Kaiser M. L. and Hinderberger E. J. J. Assoc. Off. Anal. Chem. 1982 65 999. Tsalev D. L. Slaveykova V. I. and Mandjukov P. B. Spectrochim. Acta Rev. 1990 13 225. Lynch D. and Littlejohn D. J. Anal. At. Spectrom. 1989 4 157. Kunwar U. and Littlejohn D. Talanta 1990 37 555. Granadillo V. A. Navarro J. A. and Romero R. A. J. Anal. At. Spectrom. 1993 8 615. Bermejo-Barrera P. Aboal-Somoza M. Soto-Ferreiro R. M. and Dominquez-Conzalez R. Analyst 1993 118 665. Welz B. Schlemmer G. and Mudakavi J. R. J. Anal. At. Spectrom. 1992 7 1257.1432 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 23 Beach L. Analytical Methods for Zeeman Graphite Tube Atomizers ed. Rothery E. 1986 Varian Techtron Pty. Limited. Publication No. 85-100650-00. 24 Kumpulainen J. Paakki M. and Tahvonen R. Fresenius’ 2. Anal. Chem. 1988 332 685. 25 Kumpulainen J. Paakki M. and Tahvonen R. Fresenius’ J. Anal. Chem. 1990 338 423. 26 Kumpulainen J. and Tahvonen R. Fresenius’ J. Anal. Chem. 1990 338,461. 27 Massmann H. El Gohary Z. and Gucer S. Spectrochim. Acta Part B 1976,31 399. 28 Massmann H. Talanta 1982 29 1051. 29 Shan X-q. and Wang D-x. Anal. Chim. Acta 1985 173 315. 30 Volynsky A. Tikhomirov S. and Elagin A. Analyst 1991 116 145. 31 Yang P.-y. Ni Z.-m. Zhuang Z.-x. Xu F.-c. and Jiang A.-b. J. Anal. At. Spectrom. 1992 7 515. 32 Dabeka R . W. Anal. Chem. 1992 64 2419. Paper 4/035081 Received June 10 1994 Accepted September 9 1994

ISSN:0267-9477    

DOI:10.1039/JA9940901427

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1433 Application of Experimental Designs in Optimization of a Hydride Generation Quartz Furnace Atomic Absorption Spectrometry Method for Selenium Determination Gaetane Lespes Fabienne Seby Pierre-Marie Sarradin and Martine Potin-Gautier Laboratoire de Chimie Analytique Universite de Pau Avenue de I'Universite 64000 Pau France An optimization of the hydride generation quartz furnace atomic absorption spectrometry (HG-QFAAS) method used for selenium trace determination was performed. Experimental designs at two levels including seven factors were used allowing the evaluation of the influence of each analytical variable on peak area repeatability and signal-to-noise ratio (S/N). For peak area the most significant factors are the helium and oxygen flow-rates.The NaBH flow-rate and hydride generator temperature have a strong significant effect on repeatability. No factor is really essential for the S/N. The detection limit is 1.5 ng of Se'" ( 3 ~ ~ ) . The absorbance is found to be linear over the range 1.5-60 ng. A 3.72% relative standard deviation is obtained for six determinations of 10 ng of Se'" standard solution. A Community Bureau of Reference (BCR) certified reference material (CRM) 402 White Clover was studied in order to control the quality of the determination. The total selenium contents were determined using the HG-QFAAS optimized method and then compared with the results given by an electrochemical method (differential pulse cathodic stripping voltammetry).Keywords Selenium; hydride generation; atomic absorption; optimization; experimental design For the past few years the biological importance of selenium has given rise to a great deal of interest because the concen- tration range between essential and toxic states is very Its toxicity and bioavailability depend on its phys- ico-chemical form. Inorganic selenium can exist in four different oxidation states selenate (Se") selenite (Se") elemental sel- enium (Se') and selenide (Se-"). In the environment oxidized mineral forms can be assimilated and reduced by microbial activity in various organic corn pound^^.^ such as selenoamino- acids (selenomethionine selenocysteine selenocystine) and their derivatives. By metabolization these organic species can be transformed into methylated compounds e.g.dimethylsel- enide (DMSe) dimethyldiselenide (DMDSe) which are volatile species.' ,6 Consequently it is very important to quantify the amount of selenium in both environmental and biological samples. Many analytical methods can be used fl~orimetry,~ hydride generation atomic absorption ~pectrometry,~~~ electrochemis- try,l0*l1 neutronic activation12 and electrothermal atomic absorption ~pectrometry.'~ The first three techniques need selenium to be in the selenite form (SeIV) and the sensitivity ranges and detection limits are similar for these three methods. Hydride generation quartz furnace atomic absorption spec- trometry (HG-QFAAS) is a very useful method because of its rapidity selectivity reliability and accuracy in allowing the determination of trace selenium in very complex matrices (ie.a simple method with a specific detection little matrix effects good repeatability and sensitivity). It is interesting to optimize the HG-QFAAS method in order to confirm results obtained with other analytical methods for determination of trace selenium in environmental ~ a m p l e s . l ~ - ~ ~ Several means of optimization can be used according to the accuracy required but some methods are too time-consuming (e.g. methods of response areas) or too restrictive in the number of parameters that can be selected (e.g. simplex procedure). In order to avoid these drawbacks and to assess from a reduced number of experiments the influence of many factors with accuracy the factorial design method was chosen in this study.This work has a double aim first the systematic examination of each adjustable parameter of the process all the operating variables usually considered by authors being reviewed and second the simultaneous optimization of the factors by a statistical method allowing a better control of the analytical procedure. The optimized method was applied to the analysis of a certified reference material. The HG-QFAAS and differential pulse cathodic stripping voltammetry (DPCSV) results were compared. Experimental Apparatus The HG-QFAAS apparatus used was previously developed in our laboratory for organotin speciation (Fig. 1).18-20 It includes a hydride generator vessel (1) with a magnetic stirrer a glass column (4 mm i.d.) (2) the geometry and packing of which can be modified a quartz tube atomizer (140 mm length 12 mm i.d.) (3) in a furnace (the geometry of the quartz tube has been optimized bef~re'~-~l) a VARIAN SpectrAA-10 atomic absorption spectrometer with a Perkin-Elmer furnace MHS 1 and a Shimadzu CR4A-Chromatopac integrator with a PC16N 1/0 card for the automatization of the whole process. Reagents A reducing solution was prepared with sodium tetrahydrobo- rate (NaBH,) (Fluka variable mass percentage) and stabilized by the addition of 1% m/v NaOH (Merck Suprapure).The solution was stored at 4 "C after decantation. The analytical medium was prepared with 37% m/v HC1 (Merck analytical- 3 Helium - Hydrogen 1 2 Fig. 1 1 HG vessel; 2 glass column and 3 quartz tube atomizer Schematic diagram of the automated HG-QFAAS apparatus:1434 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 reagent grade). The lOOOmgl-' Se" stock solutions were prepared with sodium selenite ( Na2Se03 5H,O Merck pro analysi) and stored at 4°C in Teflon bottles. The working solutions 10 mg 1-' and 100 pg 1-' were prepared daily. All solutions were prepared with 18 M!2 cm suprapure water (Millipore system MilliQ following MilliRO). The purity of the gases was He 2 99.995% H 3 99.995% and 0 2 99.5%. Procedures Analytical method The principle of the HG-QFAAS method has already been detailed's-20 and can be summed up as follows. The aqueous sample (selenite) is introduced with HCl into the hydride generator vessel (the total volume of the generator vessel was previously optimized).Then the reducing agent (NaBH,) is added using a peristaltic pump and SeH is formed from Se". The volatile hydride is flushed by the carrier gas and trapped in the column cooled by liquid nitrogen (cryogenic trapping). After a purging time the cold trap is removed and the column heated. The SeH is carried in the quartz tube (heated electri- cally by the furnace) and is atomized in an H2-02 flame. Absorption is measured at 196.0nm using a hollow cathode lamp. The recording and the treatment of the selenium signal are performed by an integrator. Experimental designs The study was performed with 50 ng of SeIV and the experimen- tal design method was used according to Goupy. To reduce the number of experiments fractional designs at two levels were used. The main factors were parameters with no or little interaction between them.The additional factors were estab- lished with some second or third order interactions. After the second order the interactions were considered to be insignificant. Each experiment was performed four times giving an average value taken for a representative value. The precision of the result was calculated with the standard deviation s (n=4). These values permitted the evaluation of the factor effect using the experiment matrices according to Goupy. The precision of the effect was three times the standard deviation (3s). Total selenium determination in White Clover To control the accuracy and reliability of selenium determi- nation by HG-QFAAS a Community Bureau of Reference (BCR) certified reference material (CRM) 402 White Clover was used which was certified for its total selenium ~ o n t e n t .' ~ The process used was similar to that described for the DPCSV method.15 First the sample was dried at 70 "C for 24 h. 100 mg of White Clover was digested with 10 ml of HNO (65%)+5 ml H202 (32%) + 2 ml HC104 (70%) and the residue after drying was dissolved in 10ml of suprapure water. Second all the selenium was reduced into the selenite form (Se") in 6 mol I-' HCl at 90 "C using a bath of hot water for 45 min. For the Se" determination 0.2 ml of the reduced solution was used. Standard additions of 3 6 and 9ng of Se" were made (respectively 30 60 and 90 pl of the 100 pg 1-' solution). Results and Discussion Factor Screening In a preliminary study the variation fields [high (+) and low (-) levels] of the different factors were defined to review the contribution of each variable to the HG-QFAAS method in order to plan the designs.All these parameters are listed in Table 1 according to the procedure steps. Some of them (the column geometry the column packing or the acid nature) did not vary in a continuous way and could only be included with difficulty sometimes in a design. They were optimized during the preliminary study. Acid nature In the HG methods described by different a ~ t h o r s * ~ ~ - ~ ~ HCl is always used. Protons are required for the SeH formation and the presence of chloride promotes the kinetics of the reaction.23 Hydrochloric acid seems to reduce interference^,^ and it has been proved to be the best analytical medium to perform the SeIV reduction.26 Acid concentration The HCl concentration varies in the literature from 1 to 6 mol l-1.8*23*27 Th e determination of variation field for this factor is especially important as the water vapour production depends upon it.Various concentration values were tested between 1 and 8 moll-'. Over 5 moll-' the addition of a reducing agent induced an important release of vapour satu- rated with acid disturbing the baseline and plugging up the trapping column with ice. So the concentrations used for experimental designs were 1 and 4 moll-'. NaBH amount and solutionfiow rate The amount of NaBH needed seems to depend on the total volume of the solution (sample and acid) present in the hydride generator vessel. The mass can vary from 0.09 g in 15 m12* to 0.2 g in 50 ml,23 for the same Se" content.The addition rate of the NaBH solution introduced into the generator ranges from 0.1 to 0.6 gmin-1.23 It is not clear that too large an amount of NaBH disturbs the analysis,23 but it can produce a stronger hydrogen release and a widening of the Se peak.27 For this work a 10% NaBH aqueous solution (stabilized by 1% m/v NaOH) was used. The solution flow rate in the generator was fixed at 2 ml min-' (0.2 g min-') and the solu- tion volume added was studied as a function of the flow time. The SeH peak area grew to a maximum at 2-4 ml(0.2-0.4 g). The peak area was steady for volumes >4 ml. As the release of water vapour increased with the NaBH addition while the repeatability was bad it was preferable to reduce the required volume of the NaBH solution and this parameter was set at 2 ml (0.2 g NaBH,).The NaBH solution flow rate range was studied from 0.6 to 3 ml min-l (0.06-0.3 g min-'). Under 1 ml min-l no appreciable effect was observed on the baseline or peak area but the flow time and the total analysis time were lengthened. The low level was chosen at 1 ml min-' (0.1 g min-l). At flow rates > 2 ml min-' an important release of water vapour was induced and the hydride peak disappeared confirming the observations of D e d i ~ ~ a . ~ The high level was fixed at 2 ml min-' (0.2 g min-'). Hydride generator temperature A high temperature is known to promote the hydride release from the liquid phase but causes problems with water in the cooled trap. The temperature was tested from 0 to 55°C.No peak appeared at low temperatures whereas for values > 50 "C the problems previously described were induced. The temperature range for experimental designs was consequently chosen between 20 and 50°C. Column packing ,4ccording to the analysis step the packing has to improve the SeH trapping. It should also reduce the formation of ice plugging the column. Quartz and drying agents (e.g. calcium ~ h l o r i d e ) ~ ~ ~ ~ have been used in other studies but did not give good results because of adsorption and contaminationJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1435 Table 1 Parameters of the HG-QFAAS method Reaction Acid nature Acid concentration NaBH amount NaBH solution flow rate Hydride generator temperature Trapping Column packing Column geometry (length and form) Purging time Temperature gradient Carrier gas flow rate Detection Furnace temperature Oxygen flow rate Hydrogen flow rate problem^.'^ Another author has used an empty column with a water trapping pre-column at O "C.* In this study chromatographic packing was used first (Johns-Manville Chromosorb W HP 80/100 mesh).However the flow-rates of the carrier gas used during the preliminary study produced high pressures in the column modifying the packing. Cleaning and drying stages had to be inserted between two analyses in order to prevent the adsorption of SeH but the process was too time consuming. A pre-column at - 5 "C connected with an empty column was then tested with no appreciable result for the water trapping and finally only an empty column was used to trap the hydride.Column geometry The total column length (between the hydride generator vessel and the quartz tube) seemed to have no significant influence on the signal or on the retention time (this last parameter being mainly dependent on the hydride elution rate). Nevertheless the length of the column immersed in liquid nitrogen during the cryogenic trapping stage remained an important ~arameter.'~.~~ A 525 mm long column with 110 mm immersed was tried and gave an effective trapping but the cryogenic trap needed a continuous liquid nitrogen addition. A 705mm long column with 180mm immersed was used requiring only one addition of liquid nitrogen at the beginning of the analysis. The process appeared less restrictive and the efficiency of the trapping the peak area and the signal-to-noise ratio (S/N) remained identical.Whatever the carrier gas flow- rate was these observations were established confirming that the cold trapping seemed to be effective with an immersion length of 30-90 mm.23 The column form and especially the shape of the bottom was examined because of the water being trapped there. A narrow and round U-form increased both the column plugging and the dissolution of the hydride in the aqueous phase leading to bad repeatability. A square base of sufficient width (35 mm) removed these problems while keeping an effective hydride trapping. The column geometry (length and form) was entirely deter- mined during the preliminary study. An empty glass tube with a square form a 705 mm total length a 35mm base width and a 180 mm immersed length was used.Purging time This parameter could be significant particularly because losses of SeH on the glass tubing during transport to the atomizer have been The purging time could also have a great influence on the accumulation of SeHz in the trap because the release from the generator is fairly rapid. Various times were tested from 0 to 420s. Between 0 and 30 s a clear increase in the peak area was observed. Over 30 s the peak area had no significant increase which indicated no noticeable loss of SeH,. Nevertheless a bad repeatability was observed probably caused by the formation of ice. The risk of plugging the column was increased by a long purging time. In this way the variation field appeared to be very narrow and its influence was clearly established.To reduce the number of factors the purging time was fixed at 30 s. Temperature gradient With an empty column the gradient had to be as small as possible in order to have a sufficient retention time. The inside of the column in the liquid nitrogen trap was at -50°C.21 After purging the liquid nitrogen was removed and the column was left in ambient temperature leading to an actual tempera- ture gradient of 25"Cmin-' and a retention time of about 30s. When the column was heated so that the gradient was 40 "C min-l baseline disruptions or hydride dissolution in the aqueous phase were observed. These problems were avoided with the weakest gradient because the water was removed sufficiently slowly.Consequently this parameter was left steady at 25 "C min-l. Carrier gasflow rate The gas flow rates depend on the apparatus geometry so it was difficult to compare the different values given by authors. The carrier gas flow rate was studied between 100 (lowest adjustable value) and 600 ml min-' and established that the peak area had the same order of magnitude from 100 to 400 ml min-l. Above this last value the area appreciably decreased confirming the previous observations on organotin c o m p o ~ n d s . ' ~ ~ ~ ~ The helium flow-rate factor was considered with the low and high levels having the respective values of 100 and 400 ml min-l. Furnace temperature According to literature the furnace temperature range varies between 700 and 900"C.9~26-32 Fo r organotin determi- the furnace is held at 950°C.This value is the upper limit of temperature because above it the quartz tube is soon worn and damaged.21,32 The temperature values 700 and 800"C were tested and seemed to give equivalent results with regard to the peak area and the background noise level. But with a low furnace temperature the baseline stabilization was difficult to obtain and required a long equilibration time. Therefore for the designs the furnace temperature was kept within the range 800-950 "C. OxygenJEow rate The oxygen flow rate was studied between 40 (lowest adjustable value) and 100 ml min-'. Above 80 ml min-l whatever the hydrogen flow rate the peak area decreased. So the variation field was fixed between 40 and 80 ml min-'. Hydrogen flow rate The preliminary study was performed between 200 and 400mlmin-' the flame in the quartz tube being unstable above 400 ml min- '.Under 300 ml min- ' whatever the oxygen flow-rate the peak area decreased. Consequently the low and high levels were respectively 300 and 400 ml min- '. Optimization The seven factors taken into account in the experimental designs are listed in Table 2. The SeH peak area [for 50 ng of1436 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 Table 2 Experimental field definition Factor (1) HC1 concentration/moll-' (2) NaBH solution flow-rate/ml min-' (3) Hydride generator temperaturerc (4) Furnace temperaturerc ( 5 ) Helium flow rate/ml min-l (6) Oxygen flow rate/ml min-' (7) Hydrogen flow rate/ml min-' Low level (-) 1 1 20 800 100 40 300 High level (+) 4 2 50 950 400 80 400 really significant.22 Then the four average values for the RSD response were calculated and the effect of each interaction was established.The results are presented in Table 6. SeIV given in area unit (a.~.)] the repeatability [relative standard deviation (RSD) %] calculated from each experiment made four times and the S/N (peak area-to-background noise average area) were the three responses studied. The criteria of optimization were defined according to a maximum area a minimum RSD and a maximum S/N. The three main factors were determined to be (l) [HCl]; (2) Q,; and (3) Tp because in the preliminary study their interactions seemed to be weaker. The four additional factors were (4)-( 7). First a 27-4 fractional design was used such that all the factors were connected with interactions (effects of factors and interactions were 'confounded').The interpretation of the design results was not easy because of the connection between each factor and some interactions. In order to remove any uncertainty and to determine the origin of the effects with accuracy a complementary design with the same number of experiments had to be used. It was such that the last factors were established with the opposite second order interactions. Therefore by combining the two designs the effects of each single factor could be evaluated. The matrices used for these two designs are described in Table 3. The average values obtained for the three responses studied are given in Table 4.The effects and their precisions for the factors and groups of interactions are presented in Table 5. The influence of each factor or group of interactions was estimated in comparison with the precision given in this table; if the effect was higher than the precision it was considered to be significant. Because of the combination of the two designs the confounded effects between factors and interactions disappeared and all the second order interactions appeared in groups. As the group of inter- actions 16+27+35 had a negative significant effect for the repeatability a matrix of four complementary experiments had to be used in order to know which of these interactions were Peak area Examining the first column in Table 5 three factors appear to have a really significant effect [HCl] (1); QHe (5) (positive influences); and Qo (6) (negative influence).The hydrogen flow-rate QH2 appears to be not really significant in the variation field studied. The generator temperature Tg. is found to have no influence on the peak area even though it is usually expected to promote the SeH release. This result can be explained because the temperature at its high level was not sufficient to increase the release noticeably but avoided a large production of water. Only three significant effects can be considered to reconstruct a 23 duplicated design (2 x 8 experiments) from the two pre- vious designs in order to establish a mathematical model allowing an improvement of the adjustment.22 From this design the effects of the three factors and of their interactions (can be calculated (Table 7).Each experimental response is an average obtained from the two experiments mentioned in the first column. From the evaluation of the effects the response can be represented by the following equation y (Peak area)= 3.4 + 0.3 [ HCl] + 0.4 QHe - 0.4 Qo + 0.2 [ HCl] Qo where [HCl] QHe and Qo take the values + 1 or - 1 according to the experimental matrices (Table 3) of the first two designs. 'This equation allows the calculation of the responses shown in the last column of Table 7. These calculated results are in good agreement with the experimental average responses Table 3 Experimental matrices Levels Main factors Additional factors (1) (2) (3) (4= 123) (5= k12) (6= f23) (7= 13) Experiment No. CHCll QI Tg Tf Q H ~ Qo QHz Average Initial design- 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Complementary design- + - + + - + + - + + - + + + + - + + + + - + + + + + + - - + + + + + + -JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 1437 80 - - 70 - C .- 5 60- E 50- -. An - Table 4 Calculated average responses for initial and complementary design 2.2 Responses Initial Complementary Experiment No. 1 2 3 4 5 6 7 8 Peak area (a.u.) 2.89k0.14 2.90k0.21 2.97 & 0.12 4.30f0.14 4.41 k0.35 3.37k0.11 2.39 & 0.12 3.29 k 0.22 RSD (n=4) (%) 4.3 & 0.2 7.3 f 0.5 4.0 f 0.2 3.3 & 0.2 8.0 f 0.6 3.2 & 0.1 5.0 f 0.3 6.7 & 0.4 SIN 65 f 10 2 4 f 3 41 &2 7 2 f 9 55f9 58+3 4 7 f 7 73f13 Experiment No. 9 10 11 12 13 14 15 16 Peak area (a.u.) 3.28 f 0.25 4.17f0.21 3.00&0.17 3.31 f 0.19 2.08 & 0.26 4.17 0.48 4.31 + 0.23 3.90 f 0.16 RSD (n =4) (%I 7.6 f 0.6 5.1 f0.2 5.5 f 0.4 5.7 f 0.3 12.8 f 2.0 11.6 + 0.8 5.3 f0.3 4.1 +_ 0.2 SIN 82f 14 59f4 53 & 10 55f12 51 f 10 75+ 10 78f 16 61+8 Table 5 Effects on the three responses studied Evaluation of effect Origin of effect 1 [HCl] 2 Qr 3 % 4 T 5 Q H ~ 6 Qo 7 Q H ~ 15 + 36 +47 25 + 37 + 46 17 + 26 + 45 16+27+ 35 12+ 34+ 67 14+ 23 + 57 13+24+ 56 Average Precision Peak area (a.u.) 0.26 0.02 0.07 - 0.08 0.40 - 0.42 -0.13 -0.11 -0.10 - 0.02 0.16 0.01 - 0.04 - 0.06 3.50 fO.10 RSD (Yo) -0.35 - 1.27 0.87 0.47 0.01 0.61 0.65 0.24 - 0.79 -0.37 0.8 1 0.35 0.00 - 0.04 6.22 f 0.25 SIN 0.4 0.9 3.0 - 7.2 6.9 - 3.9 0.7 2.0 3.5 1 .o 1.3 4.9 1.8 3.1 56.5 f 4.5 Table 6 Calculated average responses and effect on RSD response in the complementary design Responses Experiment RSD (n = 4) Evaluation of the effect on RSD (YO) No.(%) Origin 17 5.7 k 0.3 (16) [HClIQo - 0.32 18 3.4 If 0.2 (27) QrQH - 0.04 19 4.1 k0.2 (35) TgQHe 1.02 20 3.1 k 0.2 Average 3.91 Precision - + 0.25 confirming the validity of the model. The iso-response curves can be then plotted and are presented in Fig. 2 thus visualizing the influences. The peak area increases a lot when QHe (5) is at a high level. The Qo2 (6) also has a strong effect with the peak area increasing appreciably as the flow-rate decreases. The impor- tance of this effect remains the same whatever the helium flow- rate the interaction between these two factors (56) being nil. The influence of [HCl] (1) is weaker and not systematically linear a 4mol1-' concentration has quite a strong effect rapidly decreasing and remaining practically constant from 2 to 1 moll-'.This non-linear form shows that the interaction between the [HCl] and Qo2 (16) has a significant influence. It is so obvious that with a concentration close to 4 mol 1-' the 3.0 3.8 3.3 3.5 1 2 I I I 1 3 4 [HCl]/rnol I -' (1) Fig. 2 Isoresponse curves for peak area Table 7 The 23 duplicated design for peak area response Experiment No. 3 9 6 16 5 15 4 10 7 13 2 12 1 11 8 14 Evaluation of the effect Levels 0.3 0.4 - 0.4 -0.1 0.2 Peak area (a.u.) response Experimental average Calculated (56) + + - - + + -0.1 3.13 3.64 4.36 4.23 2.24 3.11 2.95 3.73 Average 3.4 Precision 0.1 3.3 3.5 4.1 4.3 2.2 3.0 3.0 3.81438 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 peak area increases all the more as Qo decreases. The improve- ment of the response is linear when only [HCl] ( 1) and QHe (5) are considered the interaction (15) being insignificant. In comparison with Table 5 no other interaction is significant. The largest peak area is thus achieved with a strong [HCl] a high QHe value and a low Qo value and is summarized in Table 8. In order to adjust the values of the factors Q was studied over the range 30-50 ml min-l by 5 ml min-l'iteps while other parameters were kept constant at their optimal values. The peak area was found maximum between 35 and 40 ml min-l. Repeatability (RSD) In Table 5 four factors are shown to have a very significant effect on RSD Q (2) (positive influence); Tg (3) (negative influence); Q ( 6 ) (negative influence); and QH (7) (negative influence).Two other factors appear to have a significant; but weaker; effect [ HCl] ( 1 ) (positive influence) and Tf (4) (nega- tive influence). As almost all the factors have a significance a complete design cannot be reconstructed. This is no problem because the designs and the complementary experiments used allow the evaluation of the useful effects for the analytical method. The only group of interactions having a strong negative effect (16+27+ 35) is clearly explained in Table 6 the inter- action between Tg and QHe (35) is at the origin of this effect. Other groups of interactions or other interaction have a positive significant effect 15 + 36 + 47 (strong influence) 17+26+45 and 16 (weak influences).The group 12+34+67 has a slight negative effect but this is not considered a drawback. The best repeatability is achieved with the factor levels summed in Table 8. The adjustment of the values was per- formed as previously Qo varying between 30 and 50 ml min-'. The optimal repeatability was obtained in the range 40-50 ml min- ' . Signal-to-noise (SIN) Only two factors listed in Table 5 have a weak significant effect Tf (4) (negative influence) and QHe (5) (positive influence). They may be considered to reconstruct a 22 quadrupled (4 x 4 experiments) design presented in Table 9 using the two pre- vious designs. Each experimental response is an average obtained from the four experiments mentioned in the first column. The response is represented by the following equation y (S/N) = 59.5 - 7.5 T,+ 7QHe where Tf and QHe take the values + 1 or - 1 according to the experimental matrices (Table 1) of the first two designs.The responses calculated from this equation are very close to the experimental average responses which proves the validity of the model. The iso-response curves are plotted in Fig. 3. It is clearly established that the signal-to-noise is improved simply by a decreasing furnace temperature and an increasing Table 9 The 22 quadrupled design for SIN response Experiment No. 6 7 9 1 3 2 3 13 16 1 4 14 15 5 8 10 11 Evaluation of the effect Levels S/N response + + - + - + + + _ - - - -7.5 7.0 1.0 Experimental average Calculated 61 60 44 45 73 74 60 59 Average 59.5 Precision 3.0 45 59 60 74 I I I 7 100 200 300 400 QH,/ml min-' (5) Fig.3 Isoresponse curves for SIN helium flow-rate the interaction between these factors being nil. These results are listed in the third column Table 8. Determination of the operating conditions These conditions were inferred from Table 8 and are presented in the final column. There are no difficulties in this determi- nation because no factor or no interaction have opposing effects. For Q, the optimum peak area and repeatability are over the ranges 35-40 and 40-50 ml rnin-l respectively; a 40 ml min-' flow-rate was chosen. Subsequently these con- ditions are confirmed to be optimum for SeIV determination in a NaBH mass range between 0.2 and 0.4 g. In order to use the optimized method the performances were determined using the same working solutions.The detec- tion limit (3s) is 1.5 ng of Se" and the absorbance is found to be linear over the range 1.5-60 ng. A 3.72% RSD deviation is obtained for six determinations of long of Se" standard solution. Selenium Determination in CRM 402 'The BCR CRM 402 White Clover was certified as having 16.70 & 0.25 pg g- total selenium content. Previously this CRM Table 8 Influential factors and operating conditions volume (sample + acid) in the hydride generator vessel 50 ml; purging time 30 s; temperature gradient 25 "C min-l; and absorption wavelength 196 nm Influential factor ( 1 ) HCI concentration (2) NaBH solution flow-rate (3) Hydride generator temperature (4) Furnace temperature (S) Helium flow-rate (6) Oxygen flow-rate (7) Hydrogen flow-rate Peak area Repeatability SIN Operating conditions ++ + ++++ ++ _ _ 4 moll-' 2 ml min-' 20 "C - 800 "C + 400 ml min- 40 ml min-l 300 ml min-lJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL.9 1439 was the subject of an intercalibration exercise organized between ten French laboratories by the MRT (French Research and Technology Office)." Our laboratory participated in this study for the electrochemical analysis part and the results gave The total selenium was determined in White Clover after the following two steps sample digestion and reduction to obtain the total selenium in SeIV form. This process had been optimized before in our 1ab0ratory.l~ The selenium amounts found were 6.9 k0.6 pg 8-l (RSD = 8.3% n = 6).The last two precisions were determined with the standard deviations. Considering the precision the determination of total sel- enium contents in the clover sample is in good agreement with the certified value and the precision is acceptable. The HG-QFAAS and DPCSV results are very close. The digestion stage does not interfere because the reagents used are com- pletely evaporated without volatilization of selenium. Nevertheless after the reduction step it is necessary to remove the chlorine formed,33 by degassing with nitrogen for 5 min in order to avoid the HG-QFAAS SeIV peak inhibition. Moreover the White Clover is also certified for As( ~50.09 pg g-') Co (~50.2 pg g-') and Mo (e6.9 pg g-') contents and contains significant amounts of Cr ( z 5.2 pg g-') Fe (w244 pg g-'QH,) Ni ( ~ 5 8.25 l g g-') and Zn (w25 pg g-').The results obtained suggest no interferences from these elements. 6.9k0.4 pg 8-l (RSD=6.0% n = 7 ) . Conclusion This method appears to be suitable for selenium speciation. It should allow a better understanding of selenium repartition in natural samples. A comparison of DPCSV and HG-QFAAS methods for environmental samples is now in progress in our laboratory. References Dubois F. and Belleville F. Path. Biol. 1988 36 1017. Chappuis P. Les Oligo-Cltments en Mkdecine et Biologie Medicales Internationales Paris 1991 ch 3 and 7. Cutter G. A. and Bruland K. W. Limnol. Oceanogr. 1984 29 1179. Abrams M. M. and Burau R. G. Commun. Soil Sci. Plant. Anal. 1989 20 221. Chau Y. K. Wong P. T. S. Silverberg B. A. Luxon P.L. and Bengert G. A. Science 1976 192 1130. Cooke T. D. and Bruland K. W. Environ. Sci. Technol. 1987 21 1214. 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 Nakaguchi Y. Hiraki K. Tamari Y. Fukunaga Y. Nishikawa Y. and Shigematsu T. Anal. Sci. 1985 1 247. Masscheleyn P. Delaune R. D. and William H. Spectrosc. Lett. 1991 24 307. Chan C. C. Y. and Sadana R. S. Anal. Chim. Acta. 1992,270,231. Batley G. E. Anal. Chim. Acta 1986 187 109. Campanella L. Ferri T. and Morabito R. Analusis 1989,17,507. Orvini E. Lodola L. Gallorini M. and Zerlia T. Heavy Metals Enuironment September 1981 3rd International Conference Amsterdam 657. Hoenig M. Analusis 1991 19 41. Potin-Gautier M. Astruc M. and Vermeulin P. Analusis 1991 19 M57. Potin-Gautier M. Astruc M. Vermeulin P. and Clement A. Analusis 1992 20 M73. Seby F. Potin-Gautier M. and Castetbon A. J. Fr. Hydrol. 1993 24 fasc. 1 81. Quevauviller P. Vercoutere K. and Griepink B. Anal. Chim. Acta 1992 259 281. Astruc A. Lavigne R. Desauziers V. Pinel R. and Astruc M. Appl. Organomet. Chem. 1989 3,267. Desauziers V. Leguille F. Lavigne R. Astruc M. and Pinel R. Appl. Organomet. Chem. 1989,3 469. Sarradin P. M. Pannier F. Astruc A. and Astruc M. J. Fr. Hydrol. 1993 24 fasc. 1 69. Sarradin P. M. PhD Thesis University of Pau 1993. Goupy J. in La MCthode des Plans D'expkriences ed. Dunod Bordas Paris 1988. Dedina J. Fresenius Z'Anal. Chem. 1986 323 771. Dedina J. Prog. Anal. Spectrosc. 1988 11 251. Ebdon L. and Sparkes S. T. Microchem. J. 1987 36 198. Nakata F. Yasui Y. Matsuo H. and Kumamaru T. Anal. Sci. 1985 1 417. Sanz J. De Marcos S. Galban J. and Gallarta F. Analusis 1993 21 27. Mayer D. Haubenwallner S. Kosmus W. and Beyer W. Anal. Chim. Acta 1992 268 315. Xu S. K. Zhang S. C. and Fang Z . L. Chinese Sci. Bull. 1989 35 526. Krivan V. Petrick K. Welz B. and Melcher M. Anal. Chem. 1985 57 1703. Chan C. C. Y. Anal. Chem. 1985 57 1482. Agterdenbos J. Van Noort J. P. M. Peters F. F. and Bax D. Spectrochim. Acta. Part B 1986 41 283. Brimmer S. P. Fawcett W. R. and Kulhavy K. A. Anal. Chem. 1987,59 1470. Paper 4/02 7 70A Received May 10 1994 Accepted July 22 1994

ISSN:0267-9477    

DOI:10.1039/JA9940901433

出版商:RSC    

年代:1994

数据来源: RSC

摘要:

JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 1441 Aboal-Somoza Manuel 469 Absalan G. 45 Adams F. 151 Adriaenssens Eddy 1389 Aizpun B. 1279 Akman Suleyman 333 Aller A. J. 871 Alonso Jose Ignacio Garcia Alvarado Jorge 1223 Alves Luis C. 399 Amarasiriwardena Dula 199 Anderson David R. 67 Anderson S. E. 263 Anderson Stan T. G. 1107 Anghel Sorin D. 635 Anzano Jesus M. 125 Argentine Mark D. 199 1121 Arnold J. T. 263 Arriagada Lorna 93 Arruda Marco A. Z. 657 Avila Akie K. 543 Baaske Bernd 867 Back M. H. 45 Barciela-Alonso Carmen 469 Barinaga Charles J. 1053 Barnes Ramon M. 199 981 1121 1299 1371 Barrios Carlos 535 Barshick Christopher M. 83 Barth Peter 773 Baxter Douglas C. 297 Baxter Malcolm J. 727 Bayne Charles K. 83 Beary Ellyn S. 1363 Beauchemin Diane 509 1341 Becerra JosC 535 Begley Ian S.171 Belarra Miguel A. 125 BeneS Petr 303 Bermejo-Barrera Adela 469 Bermejo-Barrera Pilar 469 483 Bernasconi G. 151 Berndt H. 861 Berndt Harald 39 193 Betti Maria 385 Bettinelli Maurizio 805 Biffi Claudio 443 Blades M. W. 1311 1323 Blanco Gonzalez E. 281 Blanco E. 1279 Bloxham Martin J. 935 Boge Edward M. 369 Bolland David T. 1255 Botelho Gloria M. A. 1263 Botto Robert I. 905 Boughriet A. 1135 Bowins Robert J. 1233 Branch Simon 33 Brandt R. 1063 Bratter Peter 1293 Brenner I. B. 737 Briand Alain 17 Broekaert JosC A. C. 1015 1063 Brown Garrett N. 1411 Brown Nicole V. 363 Bruhn Carlos G. 535 Bruno Sergio N. F. 341 BudiE Bojan 53 Burakov V. S. 307 483 1209 1371 483 CUMULATIVE AUTHOR INDEX JANUARY-DECEMBER 1994 Byrne John P. 913 Cabon J. Y.477 Cabrera Carmen 1423 Cai Xiangjun J. 697 Cairns Robert O. 881 Camara Carmen 291 1423 Campbell Michael 187 Campbell Michael J. 1379 Campos Reinaldo C. 341 1263 Carey Jeffrey M. 975 Carrion Nereida 205 217 Caruso Joseph A. 145 957 975 Castillano Theresa M. 1335 Castillo Juan R. 125 311 Cervera Maria Luisa 651 Chakrabarti Chuni L. 45 913 Charnley Norman R. 1185 Chartier FrCderic 17 Cheam Venghout 3 15 Chirinos JosC 237 Cimadevilla Enrique Alvarez- Cleland Sandra L. 975 Clive Thompson K. 1417 Cobo I. G. 223 Coedo Aurora G. 223 11 11 Conver T. S. 899 Cooper 111 C. B. 263 Cordos Emil A. 635 Cornejo Silva G. 93 Cornelis Rita 945 Craig Jane M. 1341 Crain Jeffrey S. 1273 Crews Helen M. 615 727 cserfalvi Tamas 345 Cujes Ksenija 285 Curtius Adilson J. 341 543 Dadfarnia Shayessteh 7 Dahl Kari 1 Dams Richard 23 177 187 815 1075 1243 Dean John R.615 de Boer Jan L. M. 1093 De Kimpe Jurgen 945 de la Guardia Miguel 651 Demesmay Claire 1379 Dennis John 727 Denoyer Eric R. 927 Deram L. 1135 Deruaz D. 61 Desrosiers Roland 315 Donard Olivier F. X. 1143 Doner Guleren 333 Dorado M. Teresa 223 11 11 Du Xiaoguang 629 Duan Yixiang 629 Durrant Steven F. 199 Ebdon Les 33 611 615 939 Elgersma Jaap W. 619 Elisa Soares M. 1269 Eljuri Elias 205 Elmahadi H. A. M. 547 Emteborg Hikan 297 Epler Katherine S. 79 Evans E. Hywel 939 Evans R. Douglas 985 Evans Susan 1249 Fadda Sandro 519 Fariiias Juan C. 841 1335 919 1399 Cabal 117 1263 1417 Farrer Humphrey N. 1107 Fassett Jack D. 1363 Fecher Peter A. 1021 Feinendegen Ludwig E. 791 Feldmann Ingo 1007 Fell Gordon S.457 Feng Xinbang 823 Fernandez de la Campa M. R. Fernandez Alberto 205 217 Fernandez Maria Luisa 1279 Ferreira Margarida A. 1269 Fischer Johann L. 623 Fischer W. 257 375 Fiiera M. 1285 Fisher Andrew S. 611 Florian K. 257 Fonesca Rodney W. 167 Foster Robert D. 273 Foulkes Michael E. 615 Frentiu Tiberiu 635 Gallego Mercedes 657 663 691 Geertsen Christian 17 Ghazy Shaban E. 857 Giessmann Ulrich 1007 Giglio Jeffrey J. 1335 Gijbels Renaat 1389 Gilmutdinov Albert Kh. 643 Giovanonne Bruno 1209 Glatz Jean-Paul 1209 Golloch Alfred 867 971 Goltz Douglas M. 919 GomiSEek Sergej 285 Gonzalez Urcesino 535 Goodall Phillip 1417 Goode Scott R. 73 965 Goossens Jan 177 187 Gower Stephen A. 363 369 Gras Nuri T. 535 Gray Alan L. 1179 Greb Ulrich 1075 Greenfield S. 565 Greenway Gillian M.547 Gregoire D. Conrad 393 605 Griffin Steven T. 697 Grohs James 927 Guqer Seref 797 Hadgu Negassi 297 Halls David J. 1177 Haraldsson Conny 1229 Harnly James M. 419 Harrison W. W. 991 1039 Hatterer Andre 525 Hauptkorn Susanne 463 Heitmann U. 437 Hernandez Cordoba Manuel 553 1167 Hese A 437 Heumann Klaus G. 1351 Hiernaut Tania 385 Hill Steve J. 935 1417 Hinds Michael W. 451 1411 Hladkq Z. 1285 HlavBEek I. 245 251 HlavaEkovA I. 245 251 Hoffmann Erwin 685 1237 Holcombe James A. 167 415 Hollenbach Mark 927 Horlick Gary 593 823 833 Houk R. S. 399 Hoult Gavin 7 Howe Alan M. 273 23 1 913 919 Hu Yanping 213 701 Huang Benli 779 Huang Zhuoer 11 Hudnik Vida 53 Hughes Dianne M. 913 Hutton J. C. 45 Hutton Robert C. 385 881 Hywel Evans E. 1335 Ignacio Garcia Alonso Jose Imai Shoji 493 759 765 Inagaki K.1161 Ince Ahmet T. 1179 Isaevich A. V. 307 Ito Tetsumasa 1001 Itriago Ana 205 Jacksier Tracey 1299 Jackson Jason G. 167 Jakubowski Norbert 193 1007 Janssens K. 151 Jaramillo Victor H. 535 Jepkens Brigitte 193 Jin Qinhan 629 851 Jones Delwyn G. 369 Jung Gerhard 1075 Kabil Mohamed A. 857 Kaneko Tetsuya 1273 Kantor Tibor 707 Karagozler A. Ersin 797 Katskov Dmitry A. 321 431 Kimber Graham M. 267 Kirschner Stefan 971 Kitagawa Kuniyuki 1273 Kloner A. 737 Kmetov Veselin 443 Koch Lothar 385 1209 1217 Kogan Valentina V. 451 Koirtyohann S. Roy 997 Koirtyohann S. R. 1305 Kojima Isao 1161 Kolihova Dana 303 Kondo S. 1161 Koppenaal David W. 1053 Koropchak J. A 899 Krahenbuhl Urs 1249 Kratzer Karel 303 Krieger Brian L. 267 Krivan Viliam 463 773 Kroft Marilyn 927 Krug F.J. 861 Krushevska Antoaneta P. 199 981 1121 Kubova Jana 241 1173 Kujirai Osami 751 Kumamaru Takahiro 89 Kumpulainen J. 1427 Kurfurst U. 531 Laborda Francisco 727 Lacour Jean-Luc 17 Lamoureux Marc M. 919 Larsen Erik H. 1099 Laser Bernd 1075 Lazik C. 45 Le Bihan A. 477 Lechner Josef 315 Lee Julian 393 Leis F. 1063 Lespes Gaetane 1433 Li Yongquan 679 Li Zhikun 679 Liang Yan Zhong 669 Liezers Martin 1179 1217 10151442 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1994 VOL. 9 Lile E. S. 263 Lim Jong-Soo 1357 Littlejohn David 1255 Liu X. R. 833 Lonardo Robert F. 1195 Long James V. P. 1185 Lopez Garcia Ignacio 553 1167 Lopez Jose C. 651 Lopez-Gonzalvez M. Angeles Lord 111 Charles J. 599 Lourdes Bastos M. 1269 Ludke Christian 685 1237 Luecke Werner 105 LyvCn Benny 1229 Ma Yizai 679 Madrid Yolanda 1423 Mamich Stephen 927 Manickum Colin K.227 Manninen Pentti K. G. 209 Manzoori Jamshid L. 337 Marais Pieter J. J. G. 321 431 Marchante Gayon Juan Manuel 117 Marcus R. Kenneth 1029 1045 Marcus R.K. 45 Marin Sergio R. 93 Marshall William D. 1153 Martin Fabienne M. 1143 Martinez-Garbayo Maria Paz Martinsen Ivar 1 Massey Robert C. 615 Masuda Kimihiko 1385 Mauchien Patrick 17 McAllister Trevor 427 McCrindle Robert I. 321 431 McLeod Cameron W. 67 751 McNutt Robert H. 1233 Mendes Paulo C . S. 663 Mermet Jean-Michel 17 61 Mezei Pal 345 Michel Robert G. 501 1195 Milagros Gomez M. 291 Miller-Ihli Nancy J. 605 1129 Milton Dafydd M.P. 385 Minnich Michael G. 399 Misakov P. Ya. 307 Mixon Paul D.697 Moenke-Blankenburg Lieselotte 1059 Moens Luc 177 187 815 1075 1243 Mohl Carola 791 Momplaisir Georges-Marie Montaser Akbar 1357 Montoro Rosa 651 Moreda-Piiieiro A. 483 Moreda-Piiieiro J. 483 Moreno Rodrigo 841 Mori Toshio 159 Mostafa Mohamed A. 857 Moulton Gary P. 419 Mousty Francis 719 Munthe John 1229 Murillo Miguel 205 217 237 Nagengast Anton 1021 Nagulin K. Yu. 643 29 1 71 3 125 1087 217 841 1203 1153 Nakahara Taketoshi 159 Nakamura Yoshisuke 751 Nam Sang-Ho 1357 Naoumidis A. 375 Naumenkov P. A. 307 Nevoral Vladislav 241 Ni Zhe-ming 669 Nickel H. 257 375 NiedergesaiB Rainer 1071 Nobrega J. A 861 Nolte Joachim 1059 O’Haver Thomas C. 79,419 Ohman Peder 1229 Ohorodnik S. K. 991 Ohta Masahide 1273 Okamoto Yasuaki 89 Okamoto Yukio 745 - Okochi Haruno 751 0116 Micheline 1379 Olson Lisa K.975 O’Neill Peter 33 Outred Michael 381 Outridge Peter M. 985 Ozdemir Yuksel 797 Pagliosa Giorgio 1209 Pak Yong-Nam 1305 Palacios M. Antonia 291 Papaspyrou Manfred 791 Patriarca Marina 457 Paulsen Paul J. 1363 Pauwels J. 531 Pavski Victor 1399 Payling Richard 363 369 Peachey Russell M. 267 Pepelnik Rudolf 1071 PQez- Arantegui J. 3 11 Perez Parajbn Juan M. 111 Petit de Peiia Yaneira 691 Petrucci G. A. 131 Petty John D. 267 Pilger C. 1063 Platzner Isaac 719 PolakoviEova Jozefa 1173 Polettini Albert0 L. 769 Pollmann D. 1063 Popescu Adrian 635 Potin-Gautier Martine 1433 Potts Philip J. 1185 Poussel E. 61 Prange Andreas 1071 Pretorius Warren G. 939 Prudnikov Evgeniy D. 619 Pyrzynska Krystyna 801 Quentmeier Alfred 355 QuerrC G. 311 Rademeyer Cornelius J.623 Radziuk Bernard 1 Raikov S. N. 307 Raith Angelika 1045 Rasmussen Gert 385 Recknagel Sebastian 1293 Reed Nicola M. 881 Reija Carmen 651 Reyes Olga 535 Richner Peter 985 Rivoldini Alessandro 519 RobCrt Robbie V. D. 1107 Robles L. C. 871 Rodriguez Aldo A. 535 Romon-Guesnier Sabine 199 RonEeviC Sanda 99 Rosenberg Rolf J. 713 Rosick Ullrich 1293 Rottmann L. 1351 1087 Rowland Stephen J 939 Rubio J. 151 Rummeli Mark H. 381 Sala JosC V. 719 Salbu Brit 1 Saleemi Abdollah 337 Salit Marc L. 997 tSalud Seremi 535 Santelli Ricardo E. 663 Sanz-Medel Alfredo 11 1 1 17 !Sarradin Pierre-Marie 1433 !Schaldach Gerhard 39 Scheie Andrew J. 415 Schneider Germar 463 Schoknecht G. 437 Scholze Horst 1237 Schumann Thomas 1059 Schwarzer Rudolph 43 1 Schwuger Milan J. 791 Seby Fabienne 1433 Segal I.737 Sekerka Ivan 315 Selby Mark 267 Sena Fabrizio 1217 Sharp Barry L. 171 Sheppard Brenda S. 145 Shick Charles R. Jr. 1045 Shimamura Tadashi 1385 lihtepan Aleksander M. 321 Silva M. M. 861 Silva R. B. 861 Siroki Marija 99 Sjostrom Sten 17 Skole Jochen 685 Slowick Jeffrey J. 951 Smit Henri C. 619 Smith B. W. 131 1039 Smith Clare M. M. 419 Smith David H. 83 Smith Fraser O. 267 Smith Monty R. 1053 Smith Trevor A. 67 SpEvaEkova VEra 303 Steers Edward B. M. 381 Steffan I. 785 1117 Stephens Roger 675 Stevenson C . L. 131 Stockwell Peter 1417 Sitrelko Vladimir 241 1173 Stuewer Dietmar 193 1007 1015 Sturgeon Ralph E. 493 605 759 765 1399 S turup Stefan 1099 Styris David L. 1411 Su Evelyn G. 501 Sugawa Kazumitsu 89 Sutton Robert L. 1079 Swenters Karin 1389 Sy T.437 Tahvonen R. 1427 Takahashi Katsuyuki 751 Takahashi Takako 1385 Takaku Yuichi 1385 Tan Yanxi 1153 Telgheder Ursula 867 971 Thoby-Schultzend orff Dominique 1209 Thomas Christopher L. 73 965 Thomassen Yngvar 1 Thompson K. Clive 7 1417 Tittarelli Paolo 443 805 231 281 1279 Tittes Wolfgang 1007 1015 Tolg Gunther 1015 1063 Tomlinson Medha J. 957 Trincherini Pier R. 719 Tsalev Dimiter L. 405 Tschopel P. 1063 Tsuge Shin 1273 Turak Elvan E. 267 Turk Gregory C. 79 997 Uchida Hiroshi 1001 Uden Peter C. 951 Ulens Katia 1243 Valckrcel Miguel 657 663 691 ValdCs-Hevia y Temprano van der Velde-Koerts Trijntje van Straaten Mark 1389 Van Winckel Stefaan 1243 Vandevelde Leon 1243 Vanhaecke Frank 187 Vanhoe Hans 23 177 187 815 Varga Imre 707 Veber Marjan 285 Verbeek Alistair A.227 Verlinden Jozef 1389 Verrept Peter 1075 Versieck Jacques 23 Visas Pilar 553 1167 Vincze L. 151 Vogl J. 1351 Vujicic G. 785 11 17 Wade Jeffery W. 83 Walden W. O. 1039 Walter Serge 525 Wang Jiansheng 957 Wang Jiazhen 679 Wang Jin 1153 Wang Mohui 1195 Wang Xiaohui 679 Wang Xiaoru 779 Wang Ying 851 Wartel M. 1135 Webb C. 263 Weir D. G. 1311 1323 Weiss Zdengk 351 Wiederin Daniel R. 399 Williams J. C. 697 Williams John G. 1179 Williams Jr. J. C. 697 Willie S. N. 759 Wiltshire Guy A. 1255 Winefordner J. D. 131 1039 Worsfold Paul J. 611 935 Wrobel Katarzyna 117 281 Xiao Grace 509 Yang Chenlong 779 Yang Pengyuan 779 Yang Wenjun 851 Yu Lijian 997 Yuan Xianglin 851 Yuzefovsky Alexander I. 1195 Zakharov Yu. A. 643 Zander A. T. 263 Zaray Gyula 707 Zhang Hanqi 851 Zhang Zhanxia 213 701 Zheng Jianguo 213 701 Zhu Jim J.905 Zhuang Zhixia 779 Zilkova Jana 303 Zilliacus Riitta 713 M. C. 231 1093 1203

ISSN:0267-9477    

DOI:10.1039/JA9940901441

出版商:RSC    

年代:1994

数据来源: RSC