43 1 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Use of a Furnace With a Graphite Filter for Electrothermal Atomic Absorption Spectrometry* Dmitry A. Katskovt Rudolph Schwarzer Pieter J. J. G. Marais and Robert 1. McCrindle Centre for Applied Chemistry Technikon Pretoria Private Bag X680 Pretoria South Africa The development of an atomization technique for electrothermal atomic absorption spectrometric (ETAAS) analysis based on the filtration of the analyte vapour through heated graphite is presented in its historical context. Initially the method was applied to the analysis of solids. Despite the effectiveness of atomization the time taken to prepare each solid sample and graphite tube arrangement made single-element determi- nations impractical using the ETAAS technique.The proposed new version of the furnace with a graphite filter was designed for the analysis of liquids using commercial ETAAS instruments equipped with an autosampler programmable power supply and a system for background correction. The main advantages of the atomizer were discovered in the course of the determination of Al Bi Cd and Cu for different sample volumes and Cd Pb and Bi in the presence of NaCl and CuCI matrices. These advantages include a 1.6-2.8-fold increase in sensitivity; the possibility of increasing the volume of doped solutions up to 100 pl and at the same time reducing the drying period to 15 s; and obtainment of a lower level of spectral background and chemical interferences without chemical modification. The accuracy of the method was verified by the determination of Cd and Pb in whole blood and steel. Keywords Electrothermal atomic absorption spectrometry; furnace with graphite filter; matrix interference; analysis of blood and steel The effect of the enhancement of the analytical signal in electrothermal atomic absorption spectrometry (ETAAS) when a furnace made from porous graphite was replaced by a furnace lined with Ta foil has been described by L'vov.' The results obtained led to the conclusion that at high temperatures ordinary (not pyrolytic) graphite becomes transparent to the vapours of some metals.The subsequent interest displayed in the transportation of a metal vapour through graphite resulted because of some practical problems in the ETAAS analysis of solids.' When certain powdered chemicals or organic materials were analysed using the boat technique a lengthy procedure was required for the decomposition and pyrolysis of the matrix and the removal of adsorbed gases.Rapid heating of the atomizer caused pulsed ejection of sample powder or soot particles into the analytical zone together with the atomic vapour. The corresponding background signals restricted the limit of detec- tion of the analyte elements. The effect of transportation of metal vapours through graph- ite has been used in different atomizerss7 to prevent the ejection of analyte material into the analytical zone. The atomizer 'furnace with ring ~avity'~ consisted of two coaxial graphite tubes situated between removable graphite washers of special shape [Fig.l(u)]. The external tube was made of more dense graphite than the internal tube. The powder to be analysed was sampled into the ring space between these two tubes. When both tubes were heated electrically via the washers the vapour of the sample passed into the analytical zone of the central cavity through the graphite. About 100-150mg of a powdered mixture of the solid sample with graphite could be sampled. In spite of very low limits of detection for some highly-volatile elements ( z 1 x this atomizer has not really been used in practice because of the complexity of construction and poor reproducibility of analyt- ical results owing to the lack of control of the temperature and heating rate of the internal tube. The idea of the 'capsule-in-flame' atomizer [Fig.1 (b)) * Presented at the XXVIII Colloquium Spectroscopicurn Inter- nationale (CSI) York UK June 29-July 4 1993 and the CSI Post Symposium on Graphite Atomizer Techniques in Analytical Spectroscopy Durham UK July 4-7 1993. t On leave from the State Institute for Applied Chemistry Dobrolyobova pr. 14 St. Petersburg 197198 Russia. (b) 4 6 t- 5 / 9 4 4 Fig. 1 Atomizers with the sample vapour filtration (a) furnace with ring cavity; (b) capsule-in-flame; (c) capsule-in-furnace; ( d ) ring chamber tube; (e) furnace with graphite insert. 1 and 2 Internal and external tubes; 3 washers; 4 powdered sample; 5 graphite contact; 6 capsule; 7 burner; 8 tube furnace; 9 graphite plug; 10 graphite insert; 11 graphite thread; and 12 dosing hole appeared to be more promising than its predecessor.In this m e t h ~ d ~ the powdered sample mixed with graphite was placed in a graphite cylinder (capsule) of about 20mm in length and with an external diameter of 3.5 111111 having an internal cavity of 1.5-2 mm in diameter and a length of about 18 mm. About 30mg of powder could be dosed into the capsule cavity. The capsule was held between two water-cooled stands which served as electric contacts. The burner for acetylene-air or acetylene-dinitrogen oxide flames was situated beneath the capsule. When the capsule was heated by an electric current the sample vapour was introduced into the flame which maintained a constant and sufficiently high temperature in the analytical zone.432 JOURNAL OF ANA An analytical practice described by Kruglikova6 made it possible to determine the main advantages of the filtration of the sample vapour through the heated graphite partition and also the disadvantages of the atomizer used.It appeared that a substantial portion of high- and even low-volatility carbide forming elements such as Ti and V could be transported rapidly through the heated graphite. Any interferences resulting from light scattering were absent; the chemical and spectral interferences were decreased substantially. When refractory metals were being determined some draw- backs of the atomizer4 were discovered. A longitudinal tem- perature gradient due to the cooling of the ends of the capsule caused some of the vapour to travel along the capsule cavity. This vapour then condensed at the ends of the capsule in a low-temperature region.This led to the evaporation process being hindered. Attempts to increase the evaporation rate by means of increasing the electrical power caused damage to the capsule. An attempt to eliminate this limitation and also to increase the sensitivity of determination led to the next design of atomizer the 'capsule-in-f~rnace',~ [Fig. 1 (c)]. A plugged cap- sule filled with a sample of powdered graphite was inserted into the cavity of the graphite furnace which was then heated through its contacts following a pre-set programme. It was discovered that with this atomizer the same range of elements as with the capsule-in-flame could be determined but with an approximately 50-fold increase in sensitivity. No traces of the destruction of the capsule which was made of carbon rod for atomic emission spectral analysis could be found when the capsule was used for the determination of Ti and V.The principal advantages of vapour filtration such as the suppression of background and interferences were demon- strated by the direct determination of Ag Cu and Ni in solid materials using a specially designed 'ring chamber tube'7 [Fig. l(d)]. The results obtained from different calibration methods using standard solutions (i.e. calibration curve addition of a standard solution to a known mass of solid sample of a standard solid reference material) were in a good agreement with certified values. When the developed techniqueM was applied to routine determinations the limitations of conventional AAS as a single-element method of analysis of solids were highlighted the labour and time consumed in making a capsule or another special atomizer which would have only a short lifespan; and dosing of the sample into the capsule or handling the powder or granulated sample in the furnace appeared to be too expensive for the determination of only one element.Apparently the atomizers developed based on the designs in Fig. 1 would be much better utilized in any versions of atomic absorption or atomic emission that are multi-element techniques. The next stage in the development of the method was to use the advantages of vapour filtration for the analysis of liquids. A laboratory-made cylindrical pyrolytic graphite coated graph- ite furnace was used in the experiments.**' An insert made from porous graphite in the form of a spool was placed in the central part of the furnace [Fig.1 (e)]. The liquid to be analysed was placed into the ring space between the furnace and insert walls by using the dosing hole. It was then dried and ashed in accordance with the pre-set programme. To prevent wetting of the graphite insert and penetration of liquid into the body of the furnace coils of graphite thread were rolled round the graphite insert. This served as a sample collector. At the atomization stage evaporation of the sample distributed on the collector was assumed to be delayed relative to the temperature of the insert. The effect of this delay resulted in the vapour entering the analytical zone at a higher temperature. This caused the thread to behave in a similar fashion to a L'vov platform.The following characteristics of the atomizer were estab- lished:8-'0 the range of elements that could be determined ,YTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 included the high- and medium-volatility metals; the sensitivity of determination for most elements was practically the same as for the ordinary commercial furnace when the integration mode of registration was used; the relative limit of detection was improved because of the large volume of analyte (100-150 pl); and the most important advantage of the atom- izer was the significant decrease in spectral and chemical interferences. The main drawbacks of the design in Fig. l(e) were disco- vered when the graphite insert was placed into an electrother- rnal atomizer of an Hitachi 2-9000 AA instrument equipped with autosampler and programmable power supply instead of the laboratory-made furnace." It appeared that the accuracy of the determinations was inferior to that obtained with the commercial furnace.The graphite insert was heated not only by radiation but partly by the electrical current. The resistance of the contacts between the insert and the furnace changed after each firing and after touching the insert with the sampler nipple. As a result the heating rate of the insert varied and was impossible to control. This drawback displayed itself most often when the heating programme of maximum power was used to increase the delay between the heating of the thread and that of the furnace and insert. The object of the following investigation was to optimize the atomizer design so that it could be used in commercial instrumentation and to highlight the advantages of the method in the determination of elements in everyday materials.In accordance with this aim the design of the atomizer was refined'' and some of the analytical characteristics for the determinations of Cd Pb Bi Cu and A1 were investigated and compared with those found using a furnace equipped with a platform.l2 The properties of the atomizer were also investi- gated by carrying out the following determinations Cd Bi and Cu in the presence of a large excess of NaCl as the matrix; Cd and Pb in the presence of CuC1,; and Cd and Pb in whole blood; and Pb in steel. Experimental Instrumentation Atomic absorption measurements were made using a Perkin- Elmer Model 5000 spectrometer with deuterium lamp back- ground correction and HGA-500 accessories an AS-40 auto- sampler a Perkin-Elmer 56 recorder and hollow cathode lamps operated at their recommended currents.A Keller micro-pyrometer Model PB06 was used for measurement of the stable temperature of the furnace and the filter either for visual observations or for determining their heating kinetics. Commercial atomizer furnace with a platform (FP) Pyrolytic graphite coated graphite tube atomizers (Perkin- Elmer) with a solid pyrolytic graphite platform were used in comparative experiments. Laboratory-made atomizer furnace with filter (FF) The design of the atomizer is shown in Fig. 2. The filter (1) was made from a carbon rod used for spectral emission analysis and was inserted into a pyrolytic graphite coated graphite tube furnace (2a and 2b).Graphite thread (3) was placed in the ring cavity between the furnace and the filter. Two types of commercially available graphite furnaces [ Perkin-Elmer and Pyrocarbo (South Africa)] of length 28 mm were used with the standard set of graphite accessories (4). The only feature which distinguished these furnaces from the originals were four 1 mm wide slits (5) of 4 mm length on both ends cut equidistant from each other. The filter having the same length as the furnace had the shape of a spool with bulges at both ends. The externalJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 433 . 4 4 Fig. 2 Furnace with graphite filter 1 graphite filter; 2a and 2b pyro- lytic graphite coated furnaces; 3 graphite thread; 4 electrical contacts; 5 cuttings; 6 dosing hole; 7 opening in the contact; 8 auto-sampler tip; and AA and BB sectional views diameter of the bulges was exactly the same as the internal diameter of furnace cavity (about 6mm) which provided a tight contact between the furnace ends and the filter.The external and internal diameters of the central part of the filter were 4 and 2.5 mm respectively. The length of the central part was 13 mm. To provide maximum light flux through the filter cavity the internal diameter of the filter in the area of the bulges was increased to 4.5 mm. The graphite thread of about 50mm in length was rolled onto the central part of the filter before being inserted into the furnace or pushed into the ring cavity through the dosing hole (6) after the furnace and the filter had been joined.Rejnement of atomizer and position of tip of autosampler After assembling the atomizer (Fig. 2) it was placed in its normal working position between the electrical contacts (4) with the dosing hole (6) against the hole in the contact (7). The positions of the dosing hole and the tip of the auto- sampler (8) was adjusted to provide for the sample liquid to be injected into the ring cavity. (If a drop of liquid landed near the dosing hole after the sample had been inserted it could act as a plug and prevent steam being emitted during the drying step. This would result in the sample bubbling out.) To prevent the sample being placed on the filter surface near the dosing hole the tip of the autosampler was slightly bent as is shown in Fig.2. Preparation of atomizer The atomizer was fired with a step-by-step increase in tempera- ture until 2900 "C was reached. Each step was 400-500 "C with a manual hold time of 4-5 s. There were two reasons for this conditioning firstly to clean the atomizer and secondly to ensure that a reliable electrical contact between the furnace and the filter had been achieved. At high temperature under the longitudinal pressure from the cone contacts the slotted ends of the furnace were pressed firmly onto the ends of the filter. After preparation the atomizer was ready for use. For a well prepared atomizer at any step in the temperature pro- gramme both the filter and the furnace were heated at the same rate.When the maximum power was applied for rapid heating of the atomizer a small delay in the temperature of the thread relative to that of the furnace and the filter was observed. The five atomizers prepared were examined and used at different stages of the research. The stabilized temperature of the external wall was approximately 100-150°C higher than indicated by the power supply. The temperature of the filter was also higher at about 1100 as against 1000°C and 2500 as against 2300 "C for the furnace. Reagents The samples of Al Bi Cd Cu and Pb were made by diluting standard solutions (Titrisol from Merck). Nitric acid (1-2%) was added to each solution. Solutions of NaCl and CuCl of different concentrations were prepared from the solid salts of analytical-reagent grade.The certified reference material of whole blood (Seronorm N 010010 Nycomed) was used without dilution. The certified reference material of solid steel (BCS N335) was dissolved in a mixture of nitric and hydro- chloric acids (0.4 and 4%) and then the solution was diluted up to 0.7% by mass. General Procedure The analytical characteristics of the newly designed atomizer (FF) were compared with those found when using a furnace with a platform (FP) in two sequential sets of experiments. Analytical lines the currents of the hollow cathode lamp and the monochromator slit-widths were the same as recommended in the manufacturer's man~a1.l~ For both atomizers the pro- gramme of maximum power was used in the atomization step. For the FP the temperature and the drying charring atomiz- ation and cleaning times as defined by Slavin et all4 were used.For the FF the optimal conditions were also determined in sequential experiments. These are summarized in Table 1. It should also be noted that the role and value of gas flow through the cavity of the atomizer remained unclear. There was no definite relationship between the gas flow rate as indicated on the power supply and the magnitude or shape of the analytical signal. Consequently the gas flow rates given in Table 1 were not critical. All the final dilutions of the analyte additions of matrix or reference solutions were performed directly in the atomizer by means of two-step dosing with the aid of the autosampler. For each analytical signal the shape was recorded and integrated absorbance was registered.Results and Discussion Dosing Volume and Drying Time For the stabilized temperature platform furnace (STPF) in the conventional mode of operation an analyte volume of about 20pl and a drying time for this volume of about 40 s at a temperature of 250 "C is re~0mrnended.l~ An increase above this optimum in the volume dosed led to the platform being overwhelmed and the liquid spreading over the surface of the furnace. Fast drying at a temperature higher than rec- ommended caused the analyte solution to boil and sputter. As a result when the conditions as given were not applied the shape of the analytical signals was distorted and the analysis became inaccurate. To illustrate this inaccuracy some traces of the atomic absorption signals from 0.1 ng of Cd in an FP are given in Fig.3(a). Different volumes of water (10-50 pl) were added to 10 p1 of analyte in sequential experiments. For all the samples the drying conditions as given above were used despite the fact that the sample volumes exceeded those recommended by a factor 2-3. It is clear from Fig. 3(a) that an increase in sample volume was accompanied by the appearance of some additional atomic absorption signals preceding the analytical signal. An accept- able volume for a single dosing was confirmed to be less than 20-30 pl. The results of the same experiment with the FF are given in Fig. 3(b). The range of the volumes of water added was434 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Table 1 Experimental conditions as used for the determination of metals in the furnace with graphite filter Furnace heating programme Drying Charring - Internal Dosed volume/ Temperature/ Rate/ Hold time/ Temperature/ Rate/ Hold time/ gas flow/ P1 "C "C s-l S "C "C s-l S ml min- Metal Cd Pb Bi A1 c u 5-100 250 5 10 - - - - - 5-20 5-20 - 500 5 15* 50t 500 5 15* 500 5 15 700 5 15 700 5 15 - - - - Atomization Cleaning Dosed Internal volume/ Temperature/ Rate/ Hold time/ Temperature/ Rate/ Hold time/ gas flow/ Metal PI "C "C s-l S "C "C s-l S ml min-' Cd 5-100 1000 0 5 Pb 1200 0 5 Bi - 1200 0 5 A1 5-20 2400 0 5 c u 5-20 2400 0 5 - 1500 1 3 5ot 17001 1 3 1700 1 3 2800 1 3 2800 1 5 - - - - * In the analysis of whole blood the charring time used was 60 s.t Gas stopped at the atomization step.$ For the analysis of steel the cleaning temperature was 2700 "C. 10 20 30 10 s H :I I 30 40 50 60 70 80 90 Time - Fig.3 Atomic absorption signals for the determination of Cd after the addition of different volumes of water (in p1) to lop1 of sample containing 0.1 ng of Cd. (a) In the furnace with a platform; and (b) in the furnace with a graphite filter increased from 10 to 100 pl and the drying time was reduced to 15 s for all the volumes dosed. In this experiment all the signals had the same shape and magnitude. In a series of ten determinations of Cd in the samples containing equal amounts of analyte but whose volume varied by a factor ten the relative standard deviation (RSD) for the integrated absorbance was only 1.7%. It appeared that in spite of the thick stream of steam from the dosing hole during the drying period in the case of large volumes there were no significant losses of analyte.This might be explained by the large adsorbing surface in the ring cavity provided by the graphite thread. Some experiments performed to compare the atomizers with and without graphite thread confirmed this theory. The reproducibility of the atomic absorption signals in the case of large dosing volumes was much poorer for the atomizer without the thread. Sensitivity and Detection Limit The integrated absorbances for Cd Pb Bi Cu and A1 were compared for the same amounts of analyte in the FP and FF. The results show an increase in absolute sensitivity for determi- nations in the FF relative to FP by a factor of 2.8 for Cd 2.2 for Pb 1.6 for Bi 2.6 for A1 and 1.6 for Cu.These data made it possible to estimate the atomic vapour losses which occurred through the dosing hole. Taking into account that the absorb- ing layer lengths are approximately the same for the FF and the FP and the volume of their analytical zone is in a ratio of 1:3 it is clear that theoretically in the absence of vapour losses through the dosing hole the gain in absolute sensitivity for the FF should be a factor of about 3. In practice more than 7% of the vapour was lost during the determination of Cd and 45% for Cu. Apparently the value of this loss depends on the transportation rate of the metal vapour through the graphite. This estimation made clear the importance of finding a material for the filter with a high diffusion permeability. The different increases in sensitivities of the elements investigated indicated that the rate of diffusion of the metal vapour through graphite depends on the chemical interaction between the element and graphite at high temperatures. Taking into account both the increase in dosing volume of the liquid and analytical signal for the FF it appeared that the relative sensitivity of the determination could be improved on average about 10-fold.It should be noted that the gain in sensitivity cannot be completely accompanied by a similar improvement in detection limit because of some decrease in the light flux through the atomizer. For the configuration of the FF the light flux through the analytical zone was 1.6 times less than through the FP.Consequently the magnitude of shot noise had to increase ( 1.6)O.' = 1.25 times. Supposing the limit of detection to be determined by the relationship of the magnitudes of signal to shot noise it was improved in the FF by an average 10/1.25=8 times. This estimate could be correct if there is no light coming into the spectral instrument from the end of heated filter. It might not be valid in the determination of refractory metals. The problem relating to the general improve- ment in the limits of detection must be solved on the basis of the optimization of the filter configuration taking into account the original geometry of the light beam length and diameterJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VO1 of the analytical zone thickness of the walls and volume of the ring cavity space.Matrix Interferences The results of the determination of Cd Bi and Cu in the presence of increasing amounts of NaCl when background correction was applied are shown in Fig. 4 for FF and FP. The dependence of the magnitude of the background on the analytical lines of Cd and Bi is also plotted against the matrix concentration in Fig. 4(a) and (b). For all three elements the extension of the linear parts of the plots which characterize freedom from spectral inter- ferences is much wider in the case of the FF than in the case of the FP. To resolve this problem of the behaviour of the plots the Cd signals in the presence of large amounts of NaCl were recorded without background correction for both fur- naces (Fig.5). In Fig. 5 for the FP the analytical signal for 0.1 ng of Cd almost overlapped with the background of the molecular absorbance of 2 pg of NaCI. It is clear from Fig. 4 that in the case of the FP the magnitude of the background signals grew proportionally with the mass of the sample matrix. If an arbitrary assumption that the background correction system of the instrument functions efficiently if the magnitude of background signals does not exceed the same of atomic signal is made it is possible to determine the spectral inter- ference free limit (SIFL). In accordance with the plots in Fig. 4 for the FP this limit for an NaCl matrix in the determination of 0.05 ng of Cd was about 2 pg. For the determinations of Cd in the FP with chemical modification as described in the literature,14 the SIFL for NaCl was found to be about 200 pg.In the FF the background signal was distributed in time relative to the atomic absorbance signal of Cd across a wider region [Fig. 5(b)]. At the instant that the Cd signal appeared O . t t 0 Amount of NaCl added/vg Fig. 4 Integrated atomic absorbance (solid lines) and magnitude of background signals (broken lines) for the determination of (a) 0.05 ng of Cd (b) 10 ng of Bi and (c) 0.3 ng of Cu in the presence of increasing amounts of NaCl with the furnace equipped with a filter (F) and with a platform (P). The arrows mark the SIFL as determined in the text for an FP 0.6 a 0.4 e 0 a 0 m L) 0.2 0 9 43 5 - 0 5 (b) ltomization Cleaning 0 5 10 Jirne/s Fig.5 Appearance of the peak when determining without back- ground correction 0.1 ng of Cd in the FP in the presence of (a) 2 pg of NaCl and (b) in the FF in the presence of 400 pg of NaCl the magnitude of background was much less than the atomic absorption in spite of a 200-fold increase in the amount of matrix when compared with Fig.5(a) (400 pg). For both Cd and Bi the magnitude of the background at the instant when the atomic signal appeared grew non-proportionally with the amount of matrix [Fig. 4(a) and (b)]. An increase in back- ground occurred only at the stage when the furnace was cleaned i.e. when the temperature was increased. It is thus possible to assume that when the filter was used the SIFL was at least 200 times larger than for the FP without chemical modification and according to data in the literat~re,'~ at least twice as large as those obtained with chemical modification. The results given in Fig.4 for Cd Bi and Cu led to the conclusion that vapour of NaCl passed through the graphite at a slower rate than did the atomic vapour. This peculiarity in the behaviour of atomic and molecular species made it possible to use the background correction for a wide range of matrix concentrations. The increase in the SIFL in the case of the determination of Pb in CuCl (Table 2) was evidence that the nature of the effect is characteristic for molecular vapour. Determination of Cd and Pb in Reference Materials To highlight the problems that could arise in practice the determination of Cd and Pb in whole blood and Pb in steel was performed without any special pre-treatment or chemical modifiers." In the determination of Cd and Pb in whole blood 10-20 pl samples were dosed into the FF.To destroy the organic matrix and remove the smoke from the FF a charring temperature of 500°C was required for about 1 min. Calibration curve and standard additions methods were used. In the case of Pb the results obtained by using a calibration Table 2 Spectral interferences in the furnaces investigated Interference free limit/pg Analyte Matrix Platform* Platform? Filter Cd NaCl 200 60 > 400 - 90 > 1000 Bi cu NaCl 20 20 >400 Cd CuCl - > 100 > 100 Pb CuCl - 10 > 1000 NaCl * From the literature ref. 14. This work.436 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Table 3 Determination of Pb and Cd in reference materials Results Certified Con tent Reference content/ determined/ Method of Metal material pg I-' pg 1-' determination Cd Blood* 2.7 2.5 Standard additions (2.4-3.0) 2.5 Calibration curve Pb Blood* 42 41.8 Standard additions (3 7-46) 41.8 Calibration curve Pb Steel? 100 105 Standard additions * Senonorm N 010010. t A 0.7% solution of steel BCS M335 in HCI (4%) and HNO ( 0.4 yo).curve and those from the standard additions methods were practically the same in spite of the high temperature of charring. These results are within the specifications of the certificate (Table 3). The following peculiarities were discovered when Cd was determined. In each of a sequential series of experiments when whole blood and a reference solution of Cd in 0.3% solution of nitric acid in water was dosed the magnitude and area of the signals changed in sequential firings. The Cd signals from the reference solution decreased when it was dosed after the blood and in contrast the signals of Cd from blood increased when the blood followed the reference solution.Both the decreasing and increasing magnitudes reached their limit after 4-5 firings. The difference between the initial and final signals was 15-20% of the magnitude and 5-7% of the area. Neither the introduction into the atomizer of a 5% solution of nitric acid in between dosing the series of blood and reference solutions nor a 10 s exposure of the atomizer to a temperature of 2500°C at the cleaning step after each atomization period could eliminate the short-term memory effect.An investigation of the nature of the effect is of considerable interest but is not within the scope of this paper. The only suggestions that might be made is that the effect could have been due to adsorption of anions onto the ash particles or the cooling of the ends of the furnace where the vapour condensed. The results of analysis calculated from the data obtained for the last five signals in a series of ten firings for blood and reference solution coincided with the certified concentration of Cd in the material under investigation. Exactly the same result was obtained when the method of standard additions was used (Table 3). In the determination of Pb in steel some peculiarities were also noted. Most of the Pb contained in the sample was atomized at the low temperature as shown in Table 1 without any particular problems.To remove the residues of Pb an unexpectable high furnace temperature was required. This high temperature was apparently associated with the formation at the high temperature of Pb compounds or solid solutions with steel and graphite as the components. For this removal the cleaning temperature was increased to 2700 "C and the method of standard additions was used for the determinations. The results obtained are in good agreement with the data of the certificate (Table 3). gation than the standard furnace with platform when used in a commercial atomic absorption instrument with background correction autosampler and programmable power supply. The advantages found were an increase in absolute sensitivity of determination and volume of analyte dosed; a decrease in the determination time when considering the reduction of the drying period; and an extension of the interference free limit.The results of the analysis of some reference materials demon- strated the accuracy of the determinations without the use of chemical modification. There remain some problems to be solved and the technique needs to be refined. For further improvement of the analytical characteristics of the atomization method discussed a complete study is suggested which would include the following items investigation of metal and molecular vapour diffusion through graphite of different density porosity and chemical activity to find the best material for the filter; modelling of the vapour transport processes in the atomizer to find the optimal con- figuration of the filter; transformation of the general design to prevent storage of matrix material at the ends of the furnace; clarification of the function of the graphite thread and finding the optimal length of the threads; investigation of long-term stability of the results in the presence of chemically active matrices; and determination of refractive and carbide forming elements.Even considering these issues the results obtained are very promising for analytical applications. These and other problems need to be clarified and the vapour filtration method could be used as the basis not only for AA analyses but with other related technique inductively coupled plasma (ICP) electrothermal atomization (ETA) ICP mass spectrometry ETA furnace atomic non-thermal emission spectrometry etc. when electrothermal evaporation of the samples can be applied. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 References L'vov B. V. Atomic Spectrochemical Analysis Nauka Moscow 1966 and Hilger London 1970. Langmyhr F. J. Talanta 1977 24 277. Katskov D. A. Kruglikova L. P. L'vov B. V. and Polzik L. K. Zh. Appl. Spectrosc. 1974 20 739. Katskov D. A. Kruglikova L. P. and L'vov B. V. Zh. Anal. Chem. 1975 30 238. Katskov D. A. Grinshtein I. L. and Kruglikova L. P. Zh. Appl. Spectrosc. 1980 32 536. Kruglikova L. P. Ph.D. Thesis State Institute for Applied Chemistry Leningrad 1975. Shmidt K. R. and Falk H. Spectrochim. Acta Part B 1987,42,3. USSR Patent No. 1448251 priority 1.1.1987. Katskov D. A. Vasil'eva L. A. and Grinshtein I. L. Abstracts of X I CANAS Moscow 1990 p. 48. Vasil'eva L. A. Grinshtein I. L. and Katskov D. A. Zh. Appl. Spectrosc. 1993 48 1345. Provisional R.S.A. patent No. 93/4657 29.6,1993. Katskov D. A. Marais P. J. J. G. McCrindle R. I. and Schwarzer R. XXVIII Colloquium Spectroscopicurn Internationale York UK June 29-July 4 1993. Perkin-Elmer Analytical Methods for Atomic Absorption Spectrophotometry Norwalk CT USA 1976. Slavin W. Manning D. C. and Carnrick G. R. At. Spectrosc. 1981 2 5. Tzalev D. L. Slaveykova V. I. and Manjukov P. B. Spectrochim. Acta Rev. 1990 13 225. Conclusion The electrothermal atomizer developed yielded better analyt- ical characteristics for the elements and matrices under investi- Paper 3 f06416F Received October 22 1993 Accepted December 20 1993