JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 387 Shadow Spectral Filming A Method of Investigating Electrothermal Atomization Part 3.* Dynamics of Longitudinal Propagation of an Analyte Within Graphite Furnaces Albert Kh. Gilmutdinov Yu. A. Zakharov and A. V. Voloshin Department of Physics University of Kazan 18 Lenin Str. Kazan 420008 Russia The dynamics of formation and dissipation of an Ag absorption layer in electrothermal atomic absorption spectrometry was investigated using the shadow spectral filming technique with different operating modes (i.e. gas-stop and gas-flow conditions and wall and platform atomization). The use of chemically inert Ag as the test element allowed the characterization of the influence of physical factors (mass transfer and adsorption-conden- sation of the vapour on the wall) during atomization in a graphite furnace.It has been shown that under gas-stop conditions the cross-sectional structure of the Ag absorbed layer is practically uniform. This uniformity testifies to the high efficiency of diffusional mass transfer in graphite furnaces. The use of an internal gas flow leads to pronounced distortion of the uniformity for the duration of the atomization process there is a sharp decrease in the gas-phase concentration of Ag when going from the bottom of the furnace to the top. The results of imaging Hg vapour propagation along the length of the tube are presented. The results were obtained by replacing the graphite furnace by a quartz tube having the same geometry; the absence of the vapour near the cooler ends of the tube is shown.Analysis of all of the results allows the introduction of a new characteristic atomization parameter the disappearance temperature and a cascade mechanism of the analyte propagation in non- isothermal furnaces. The mechanism consists of a number of condensation-vaporization processes of the analyte as the temperature wave propagates along the length of the furnace during heating. The mechanism is important for relatively low heating rates when the velocity of diffusional propagation of the analyte vapour is higher than the velocity due to temperature propagation. Keywords Graphite furnace; electrothermal atomic absorption spectrometry; silver atomization; cascade mechanism of vapour propagation In previous parts of this the technique of shadow spectral filming (SSF) was described and the results of imaging of atomic and molecular formation and dissipation processes in graphite furnaces were presented.The results showed that the cross-sectional distributions of an analyte within a graphite and a metal-lined furnace can be distinctly non-uniform. The characteristics of the non-uniformities are strongly dependent on the particular species being atomized. For example A1 and Ga atoms are generally located near the graphite walls,’ while their molecules are only located along the central axis of the tube far from the walls.* Analysis of the SSF data allowed the demonstration of a number of peculiar features of electrothermal atomiza- tion; (2) the ‘sponge’ effect of the platform within the furnace;’ (ii) anisotropy of longitudinal and transverse propagation of the atomic vapours within a graphite furnace;’ (iii) ‘inverse’ atomization of T1 Ga and In; (zv) redistribution of A1 atoms under the platform;’ and (v) existence of not only atomic or molecular species in the furnace gas phase but also finely dispersed condensed particles at temperatures as high as 2300 “C.The distribu- tion of these particles has an unusual ‘donut’ structure.* All of these features testify to the complexity of atom formation processes and clearly indicate the presence of significant gas-phase and heterogeneous reactions while the analyte is transported in the furnace volume. These processes occur against a background of rapidly increasing furnace tempera- ture with pronounced temperature gradients in both the longitudinal and transverse direction^.^ In general the factors that affect the structure of atomic and molecular layers in graphite furnaces can be divided into two groups.Firstly there are the physical factors including sample evaporation diffusion natural and *For Part 2 of this series see ref. 2. forced convection physical adsorption-desorption and condensation-revaporization. These particular processes do not change the nature of the species and are character- istic of all elements in non-isothermal furnaces. Secondly there are the chemical factors which include solid and gas- phase reactions chemical adsorption and heterogeneous reactions. These processes change the nature of the species and are specific to each element.For example Sn atoms show a sharp decrease in the gas-phase concentration when going from the bottom of the furnace to the sample dosing hole.4 This was attributed to the gas-phase oxidation of Sn-0 molecules ingressing through the dosing hole. An example of heterogeneous reactions is provided in ref. 2 where the possibility of heterogeneous reduction of gaseous oxides by the graphite walls was proposed. Generally these two groups of factors act simultaneously creating difficulties for the interpretation of results. The influence of the chemical factors appears against the background of the physical processes. Therefore it is desirable initially to isolate the effect of the physical processes. For this purpose an element that will be fairly inert to the chemical processes that occur within the graphite furnace should be investigated.The structure of the absorbing layer of a chemically-inert element can only be determined by the physical processes and a suitable element could be used to investigate this group of factors. Once the influence of the physical processes have been established it should be much easier to investigate the influence of the chemical processes. A suitable candidate for the test element is Ag. It is recognized that in graphite furnaces Ag undergoes a straightforward atomization mechani~m,~ Ag(s 1 )-+Ag(g). Spatially resolved results obtained by Holcombe et aL6 have shown that the cross-sectional distribution of Ag atoms in a CRA-type furnace is practically uniform. This testifies to388 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL.8 the inertness of the Ag atoms under graphite furnace conditions. Manning and Slavin' used Ag as a test element to control the stability of graphite furnace conditions. All of these facts favoured the use of Ag as a test element to investigate the group of physical factors. However the inertness of Ag is only relative. It is recognized that Ag suffers from severe matrix interferences in the presence of chloride species.8 Furthermore Salmon and Holcombe9 reported an increase in the appearance temperature of Ag of from 750 to 930 "C when 1% of oxygen was added to the argon sheath gas. Nevertheless the atomization mechanism of Ag under standard operating conditions (ie. a nitrate matrix and pure argon purge gas) appears to be the simplest and most understood mechanism.The first objective of the present work was to investigate the formation of the absorption layer of Ag under different conditions (Le. wall and platform atomization and gas-stop and gas-flow modes) using the SSF technique to character- ize the effect of the physical processes within a graphite furnace. The second objective was to investigate in greater detail the dynamics of the longitudinal propagation of the analyte in a tube atomizer. Although the SSF technique does not allow the direct imaging of the dynamics of the absorption layer in the longitudinal direction this imaging can be performed using physical modelling. The graphite tube which is not transparent to radiation in the optical region can be replaced by a quartz tube of the same geometry. The modified atomizer can then be turned 90" to allow imaging of the longitudinal processes.Mercury was chosen as the element to be used for the imaging of the dynamics of longitudinal propagation of an analyte within a tube atomizer. Experimental All the experiments reported in this paper were performed using the Saturn-3 atomic absorption spectrometer with the Graphite- 1 electrothermal atomizer and the SSF set-up described elsewhere. l The apparatus allows the recording of the image of the interior of the furnace in the light of the analysis line. An atomic vapour which is vaporized into the furnace volume creates a shadow picture that is recorded by a cine camera. Electrodeless discharge lamps VSB-2 (Vladikavkaz USSR) for Ag and Hg were used as primary sources.The following wavelengths were used for SSF Ag 328.1 nm and Hg 253.7 nm. A negative cine-film (TASMA Kazan USSR) with a speed of 450ASA and a contrast coefficient of 1.5 was used to record the non-stationary absorbing layers. The images were recorded at 12 frames s-I (Ag) and 16 frames s-l (Hg). Densities produced on the film by light that had passed through the absorbing layer were calibrated using a set of neutral-density filters and were converted into absorbance values. The filters were filmed under identical conditions just before the experiment. An automatic micro- photometer was used to make photometric measurements of the images. The size of the microphotometer probe beam was 0.5x0.5mm when the i.d.of the furnace image was equal to 6mm. Perkin-Elmer pyrolytic graphite coated graphite tubes and platforms were used in the study of the atomization of Ag and a quartz tube having the same geometry (28 mm in length 8 mm o.d. 4 mm i.d.) as the Perkin-Elmer tubes was used to study the longitudinal propagation of Hg. Standard GSORM-type nitrate solutions of the elements (OFXI Odessa USSR) were used. Test solutions were prepared by dilution of the stock solution. The Hg solution was stabilized by adding 200pg ml-1 of K2Cr20 and 3% HN03. Volumes of 1 and lop1 of the solutions were injected into the furnace with a microlitre pipette. The instrumental parameters for operation of the atom- izer when the Ag samples were deposited onto the wall and the platform are presented in Tables 1 and 2 respectively.It should be noted that relatively low heating rates of about 200K S - I were used in these experiments to exclude non- stationary processes of atom formation. The atomizer parameters for Hg atomization will be described later in the text. In all experiments argon gas of the highest available quality was used as the internal purge gas and as the sheath gas. The wall and platform temperatures were measured with a photodiode sensor using a procedure described elsewhere.' Results and Discussion Formation and Dissipation of the Ag Absorption Layer Wall atomization Fig. l(a) shows the absorbance profile for 0.3 ng of Ag when lop1 of the sample were dosed onto the wall. The SSF imaging of the atomization process is presented in Fig.1 (b). It can be seen that vaporization of Ag starts from the bottom of the furnace where the sample was deposited (frames 22-26) and that the cross-sectional distributions of Ag atoms during the initial part of the atomization process are non-uniform the gas-phase concentration of Ag de- creases monotonically when going from the bottom to the top of the furnace. During the dissipation of the atomic cloud stage (frames 42-54) the cross-sectional structure of the absorbing layer is practically uniform. Quantitatively these results are illustrated by Fig. l(c) and ( d ) presenting the absorbance distributions along the vertical diameter of the furnace prior to (c) and after ( d ) the peak absorbance. This is the region that is isolated by the monochromator slit in a conventional spectrometer.The results presented above correlate well with earlier results of Holcombe et aL6 obtained for CRA-type furnaces. These data show the expected behaviour of Ag atoms within graphite furnaces the initial non-uniformities of the atomic distributions can be explained by diffusional propagation of the vapour from the bottom of the furnace where the atoms source is located. The results presented here are for atomization under gas- stop conditions. In this instance however the structure of the absorbing layer is determined by the superposition of both vaporization and removal processes. To separate the Table 1 Instrumental parameters for the Graphite-1 system with Ag samples deposited on the wall Internal Temperature1 Ramp Hold gas flow/ Step "C time/s time/s ml min-' 1 90 5 50 150 2 400 5 20 150 15 0*( 300) 3 1800 5 *Gas stop mode. Table 2 Instrumental parameters for the Graphite-1 system with Ag samples deposited on the platform Internal Temperature/ Ramp Hofd gas flow/ Step "C time/s time/s ml min-1 1 260 1 60 150 2 400 5 30 150 3 1800 5 15 0*(300) *Gas stop mode.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL.8 t E i= 0.9 0.7 Q) C $ 0.5 s 9 0.3 L 4 O.' 389 0 3 6 I 2 3 4 5 6 0 3 6 Height/mm Time/s HeighVmm Fig. 1 Absorbance versus time profile for 0.3 ng of Ag using wall atomization (a). Absorbance versus height contours at 0.17 s time intervals prior to (c) and after (d) the peak absorbance. (b) Images of the processes recorded using the SSF technique. The figures over the frames (for this figure and subsequent figures) are their numbers beginning at the onset of atomization. The images were recorded at 12 frames s-' 1 .oo 0.75 P) c $ 0.50 s 9 0.25 0 5 10 15 Time/s 1600 c Fig.2 Dynamics of the formation and dissipation of the absorption layer of the Ag atoms using wall atomization for 2.5 ng of the metal (a) recorded under gas- stop conditions; (b) recorded with an internal gas flow rate of 300 mi min-'; (c) absorbance versus time profiles for the corresponding gas stop (broken line) and gas flow (solid line) conditions. Curve T represents the change in the wail temperature390 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 supply function from the removal function Van den Broek and de Galanlo proposed that atomization should be performed under conditions of internal gas flow.Under these conditions the mean residence time of the atoms is sharply decreased and the recorded absorption pulse repre- sents the supply function. Furthermore the internal gas flow is also used in practice analytically to increase the dynamic range of a determination' and in thermochemical studies.12J3 An investigation of the influence of internal gas flow on the formation of the absorption layer was recently performed by Huie and Curran.I4 Their results however dealt with atomization of Na. In this case the structure of the absorbing layer is strongly influenced by interactions of Na atoms with the graphite walls and therefore it is difficult to extract from the data the influence of the internal gas flow alone.Fig. 2 presents the results of SSF imaging of Ag atom formation and dissipation processes in (a) gas-stop and (b) gas-flow (300 ml min-l) conditions when atomizing 2.5 ng of the metal. Fig. 2(a) illustrates the same features of the dynamics of the absorption layer as the previous Fig. l(b). However the presence of an internal gas flow in the furnace changes the absorption layer structure remarkably [Fig. 2(b)]. The absorption layer in this instance is strongly non- uniform with a sharp decrease in atom density when going from the bottom of the furnace to the top. There is essentially no absorbance near the sample dosing hole throughout the whole atomization process (frames 23-53). The most probable reason for these features is a sharp increase in the velocity of the internal gas flow when going from the bottom of the furnace to the top.Indeed near the bottom of the furnace the velocity of the vertical compo- nent of the gas flow is close to zero. This is because the two opposing horizontal flows of argon gas meet in the middle of the furnace. This horizontal flow is then converted into a vertical flow. The vertical flow rapidly accelerates and reaches a maximum value near the dosing hole. It can be shown that when the volume velocity of the internal gas flow is 300 ml min-l the linear velocity of the flow near the dosing hole is approximately equal to 1.5 m s-l for a dosing hole radius of 1 mm. Fig. 2(b) also illustrates another feature of the atom formation process. In these experiments a small sample volume of 1 pl was used.Frames 23 and 53 clearly show that atomization starts and terminates at the same place on the furnace wall. This clearly indicates that the atom source is localized. Atomic absorption pulses recorded by the spectrometer in the conventional mode and corresponding to the atomi- zation processes imaged above are presented in Fig. 2(c). The solid line indicates the atomization using 300 ml min-l internal gas flow and represents approximately the supply function. It can be seen from the curve that vaporization of Ag from the sample is finished after 7 s of heating. On the other hand the broken line which indicates the use of gas- stop conditions clearly shows the presence of Ag atoms in the furnace volume for longer than 15 s. Use of the internal purge gas leads to cooling of the furnace walls and therefore it can be suggested that in the case of gas-stop conditions the sample should be completely vaporized in less than 7 s.This situation raises the question about the presence of the atomic vapour in the furnace volume for a long time after the sample vaporization has finished. Indeed the diffusion coefficient D of Ag atoms at a temperature of 1800°C 3.5cm2 s-l. The atom residence time 7 can be estimated approximately using the known relationship z= P/8D where 1 is the length of the absorption layer. For the experimental conditions used in the present work this relationship gives a residence time of about 0.3s. The presence of the dosing hole in the centre of the furnace leads to a decrease in this estimate.I6 This means that Ag atoms should be completely absent from the furnace volume less than 1 s after the completion of sample vaporization. Consequently the absorbance profile [broken line in Fig.2(c)] should go to zero not later than 8 s after the atomization is started. However this is not the case. The most probable reason for the long residence time of the atoms within the furnace volume is because of the strong temperature gradients within the atomizer in the longitudinal direction and adsorption-condensation of the vapour on the cooler parts of the tube. Additional information about this point is provided below. Plagorm atomization For the atomic absorption spectrometric determination of Ag in a graphite furnace platform atomization is recom- mended.' The results of SSF imaging of the atomization of 2.5 ng of Ag from the platform are shown in Fig.3(a). It can be seen from both the images and also more quantitatively from Fig. 3(c) and (d) that the cross-sectional distributions of Ag atoms are more uniform than in the case of wall atomization (Fig. 2). The probable reason for this is an increase in the velocity of diffusional mass transfer as a result of increasing the gas-phase temperature when the sample is vaporized from the platform. An additional reason is the decreased distance between the platform and the top of the atomizer. The sharp rise in the absorption pulse [Fig. 3(b)] indicates that platform atomization of Ag is fairly fast. Nevertheless the cross-sectional distributions of the atoms are practically uniform [Fig.3(c) and ( d ) ] except for the earliest moments in the atomization. This means that the diffusional processes which lead to the smoothing of the non-uniformities are very efficient. It should be noted that these results were obtained for relatively low tempera- tures that do not exceed 1800 "C. Therefore it can be concluded from these results that when using platform atomization chemically inert elements at high tempera- tures must be distributed uniformly over the cross-section of the furnace. Conversely cross-sectional non-uniformities of analyte distributions strongly indicate pronounced gas- phase or heterogeneous interactions. This evidence sup- ports previous conclusions which attribute the non-unifor- mities that have been in graphite furnaces to these types of interactions.Fig. 3(a) also illustrates the anisotropy of longitudinal and transverse propagation of an analyte vapour that was postulated previous1y.l Frames 58-63 clearly show the absence of atoms under the platform when the upper part of the furnace is completely filled. This means that Ag atoms propagate faster in the furnace cross-section than in its longitudinal section. This can be concluded because the distance between the platform surface and the top of the furnace is approximately equal to the distance between the edge of the sample on the platform and the edge of the platform. If the transport of the analyte is only due to diffusion filling of the cross-section of the furnace volume would occur simultaneously with the appearance of atoms under the platform.The most probable reason for the delay in atom propagation in the longitudinal direction is the remarkable decrease in the temperature of the atomizer towards the ends of the tube. This can lead to adsorption- condensation of the vapour on the graphite walls at distances shorter than the length of the platform. The platform atomization experiments were also re- peated for the internal gas-flow conditions. Fig. 4(b) shows SSF imaging of the dynamics of Ag atom formation and dissipation using an internal gas flow of 300 ml min-I for the atomization of 50ng of the metal. For comparison the same atomization process recorded using gas-stop condi- tions is shown in Fig. 4(a) which illustrates the same features of the atom formation processes as can be seen in Fig.3(a) where the results were obtained for a much lowerJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 t E i= 0 3 6 2 7 Height/mm Ti m e/s E 1 i= 0 3 6 Heig ht/m m 39 1 Fig. 3 (6) Absorbance versus time profile for 2.5 ng of Ag using platform atomization. Absorbance versus height contours at 0.17 s time intervals prior to (c) and after ( d ) the peak absorbance. (a) Images of the processes recorded using the SSF technique 1 .oo 0.75 al C 4 0.50 B 2 0.25 0 1600 Time/s Fig. 4 Dynamics of the formation and dissipation of the absorption layer of& atoms using platform atomization for 50ng of the metal (a) recorded under gas-stop conditions; (b) recorded at an internal gas flow rate of 300 ml min-I; (c) absorbance versus time profiles for the corresponding gas-stop (broken line) and gas-flow (solid line) conditions.Curves T and Tp represent the change of wall and platform temperatures respectively392 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 sample concentration. Fig. 4(b) presents results similar to the previous case for wall atomization [Fig. 2(6)] frames 70-1 10 show strong cross-sectional non-uniformities of the absorb- ing layer for the duration of the atomization process with a sharp decrease in atom concentration when going from the platform to the top ofthe atomizer. As in the previous case of wall atomization atomic absorbance near the sample dosing hole is practically absent at all times during atomization. However Fig. 4(b) provides additional interesting informa- tion.As can be seen from frame 75 the atoms appear in the region under the platform very early in the atomization and remain under the platform until the end of atomization (frames 75- 120). Comparison of this observation with Fig. 4(a) shows the surprising result that when an internal gas flow is used the atoms reach the region under the platform even faster than in the case of gas-stopconditions [compare frames 60-69 in Fig. 4(a) and frames 70-75 in Fig. 4(b)]. This indicates that the internal gas flow (at a volume velocity of 300 ml min-' is not laminar in the central part ofthe atomizer where the two flows of argon gas coming from the tube ends meet. The earlier appearance of the atoms underthe platform in this case can be explained only by the presence of a turbulent flow of argon gas in the central regionofthe furnace which can transport the atoms under the platform.An internal gas flow of 300 ml mine' is typical during the ashing stage in electrothermal atomic absorption spectro- metry. Obviously the results presented above can be extrapolated to the ashing stage. This allows the suggestion to be made that the sample matrix which is supposed to be carried away by the internal gas flow might partially remain under the platform. This could potentially lead to interfer- ences during the atomization stage. Fig. 4(c) shows the recording of the atomization processes using a conventional atomic absorption spectrometer. The solid line refers to internal gas flow conditions and the broken line refers to gas stop conditions.The differences between these two curves are similar to the previous case of wall atomization. Therefore the reason for the differences can again be attributed to the strong temperature gradients along the length of the tube. Imitation of the Sponge' effect In Part 1 of this work' the inverse atomization ofT1 In and Ga atoms was described. This feature of atomization consists of the appearance of analyte atoms from the top of the atomizer when the sample is dosed onto the platform. Subsequently the atoms vaporized from the top of the furnace travel to the platform and disappear near its surface. The conclusion can be drawn that the atoms are 1 .o 0.8 a 0.6 C m +? s 2 0.4 0.2 0 Time/s Fig. 5 Imitation of the 'sponge' effect using Ag as a test element.(a) Absorbance versus time profiles for gas-stop (broken line) and gas flow (solid line) conditions. Curves T and T indicate wall and platform temperature respectively. Images of the processes recorded using the SSF technique under (b) gas flow and (c) gas stop conditionsJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 393 Table 3 Instrumental parameters for the rod atomizer with the quartz tube Internal Temperature/ Ramp Hold gas flow/ Step "C time/s time/s ml min-' 1 90 10 50 I50 2 300 15 30 I50 3 1000 5 5 0 adsorbed-condensed on the much cooler platform. There- fore it can be suggested that the platform acts as a 'sponge' inside the furnace. To support this hypothesis the inverse atomization and 'sponge' effect was simulated in the present work using Ag as a test element.For this simulation 50ng of the metal were dosed onto the platform and the furnace was heated for 40s at a temperature of 1100 "C under gas-stop conditions. Then after cooling the furnace the platform was removed and a new platform was inserted. The subsequent heating of the furnace with this empty platform using an internal gas flow of 150 ml min-l produced an absorbance signal consisting of two peaks [solid line in Fig. 5(a)]. It is interesting to note that the appearance time for the first peak coincides with that for the wall atomization peak and that the second peak coincides with the appearance time of the peak produced by platform atomization. Similar experiments in which the sample was dosed onto the wall and then atomized in the presence of the empty platform have been carried out previously by Matousek and P0we1l.l~ Recently Fonseca et a1.I8 and L'vov et a1.19 used the same approach to investi- gate physico-chemical processes in graphite furnaces.In all instances double absorbance peaks were recorded. How- ever in contrast to Fig. 5(a) the heights of the first peaks reported ear1ierl7-l9 were much smaller than those of the second. The reason for the difference is that in the previous publication^,'^-^^ the sample was vaporized into the region below the platform while in the results presented in Fig. 5 the sample was vaporized into the region above the platform. In the former case atoms producing the first absorbance peak are located in the small area under the platform and therefore their cross-sectional distribution is strongly non-uniform.As shown by Gilmutdinov et al.,*O this type of non-uniformity leads to a remarkable decrease in the measured absorbance value. Fig. 5(b) shows the SSF image of the double atomization process. The images were recorded using the same experi- mental conditions as were used to obtain the solid curve in Fig. 5(a). Frames 22-32 show the initial 'inverse' flux of the atoms from the top of the furnace. It can be seen that the atoms disappear near the platform. The temperature of the platform during the formation of the first peak is almost 400 "C lower than the wall temperature [Fig. 5(a)]. Therefore this allows the proposal to be made that the disappearance of the Ag atoms near the platform is because of their adsorption-condensation on the much cooler platform.Subsequently as the platform is heated the adsorbed-con- densed atoms revaporize into the furnace volume creating the second peak on the absorbance curve on Fig. 5(a). Frames 44-49 of Fig. 5(b) provide the imaging of this secondary process. Frame 38 corresponds to the minimum between the two peaks in Fig. 5(a). It can be seen from Fig. 5(b) that the images of the process are similar to previous images of T1 In and Ga atomization;I that is the images can be considered an accurate imitation of the 'sponge' effect. The same experiment was also carried out using gas-stop conditions. The results of conventional recording of the process are presented in Fig. 5(a) by the broken line. The appearance of the absorbance signal in this instance coincides with the appearance time for wall atomization and the absorbance signal goes off-scale quickly.The SSF record corresponding to this situation is presented in Fig. 5(c). Because large amounts of Ag were atomized in the gas- stop mode the SSF pictures are quickly saturated (frame 32). However at the initial stage of the atomization (frames 22-28) the same 'sponge' effect can be seen with a decreased density of the atoms near the cooler platform. Similar results are also obtained when the sample is dosed onto the wall and then atomized in the presence of an empty platform. It should also be noted that measurements taken under similar experimental conditions without the empty platform have produced only a single absorbance peak.This provides additional support for attributing the second peak on Fig. 5(a) to the secondary atomization of Ag from the platform. The results of the imitation of the 'sponge' effect presented above dealt with imaging a process in the atomizer cross-section. However they can also be applied to the analysis of the analyte propagation in the longitudi- nal direction. Indeed the presence of a platform that is much cooler than the surrounding furnace walls creates cross-sectional temperature gradients of a few hundred degrees3 200-800 "C [see also Fig. 5(a)] within the furnace. The same temperature situation is characteristic of the longitudinal section of the furnace in conventional HGA- type furnaces that were used in this work the temperature of the central part is much higher than the temperature of the ends3 Therefore the experiments that were presented in Fig.5(b) and (c) can also be considered an imitation of the longitudinal propagation of analyte. The atoms that have been vaporized from the top part of the furnace travel to the surface that has a significantly lower temperature. This is similar to the situation that exists in the case of normal vaporization and analyte propagation in the longi- tudinal direction. Therefore features such as analyte vapour disappearance near the cooler platform can also be expected along the longitudinal section of the furnace as the vapour propagates to the cooler furnace ends. Physical Modelling of Longitudinal Analyte Propagation The SSF technique' allows the imaging of the dynamics of atomization processes only in the atomizer cross-section.However the conventional graphite furnace can be re- placed by a quartz tube of the same geometry that will be transparent to the radiation from the primary source in the transverse direction. A schematic diagram of the atomizer modified in this way is shown in Fig. 6. The quartz tube (1) has a length of 28 mm and the i.d. and 0.d. are 4 and 8 mm respectively. The tube has a dosing hole of 1.5 mm and a rectangular orifice of 3 x 5 mm at the bottom. The tube was mounted on a heated graphite rod (2) that has a rectangular prominence of 3 x 5 x 1 mm upon which the quartz tube can rest. The rod was fixed between the atomizer contacts and heated using the temperature programme given in Table 3.The tube was fixed between the atomizer contacts (3) using a graphite gasket (4) so that the argon sheath gas could pass within the tube. The sample was dosed onto the graphite surface within the atomizer through the dosing hole and therefore Hg was atomized from the graphite surface and propagated subsequently in the quartz tube volume. The tube was heated by the rod and had approximately the same temperature distribution as the rod the maximum tempera- ture is at the centre of the rod and the temperature decreases monotonically towards the rod ends. Therefore the system can be used for physical modelling of longitudi- nal analyte propagation within a graphite tube. Before the SSF measurements were made the tempera- ture programme was optimized using the atomizer in a conventional spectrometer with a deuterium background corrector.Fig. 7(a) shows the absorbance profile obtained from the spectrometer with the atomizer in its normal394 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 position. To obtain SSF images of the longitudinal propaga- tion of Hg vapour the atomizer was then rotated by 90". The longitudinal structure of the Hg cloud on the rising and falling portions of the absorbance versus time profile is shown in Fig. 7(b) and (c) respectively. Frames 18 and 34 clearly show the prop-agation of the Hg vapour in the longitudinal direction. However even 3 s after atomization has started the atoms have not reached the tube ends (frame 34). The reason for this is adsorption-condensation of the vapour on the cooler ends of the tube.It can be concluded from these results that the actual length of the atomic layer is determined by the extension of the high- temperature zone of the tube. It is interesting to note that frames 43-1 12 which correspond to the decreasing portion of the absorbance peak show localization of the atom source near the tube ends in the form of two symmetrical rings. That the position of the rings ceases to change with time corresponds to the fact that the atomizer has reached its final temperature distribution and the longitudinal gradients of the temperature have stopped changing. Cascade Mechanism of an Analyte Transfer All of the results presented above in Fig. 5(b) and (c) (frames 25 and 28) Fig. 7(b) (frames 18 and 34) and also the 1 3 1 Fig.6 Schematic diagram of the modified atomizer for longitudi- nal imaging of the analyte propagation 'sponge' effect for T1 Ga and In atoms reported in Part 1 of this work,' clearly show the disappearance of analyte vapour near the cooler graphite or quartz surfaces. This allows the introduction of a new characteristic temperature for the description of electrothermal atomization the disappearance temperature Tdapp. This parameter can be defined as the temperature below which an analyte cannot exist near the surface as a vapour. As with the appearance temperature Tapp the disappearance temperature should be considered as being semiquantitative because the defini- tion depends on the particular value of the absorbance from which the vapour will be considered absent near the surface.Obviously Tdapp is a function of the physico- chemical properties of the analyte and the surface the temperature gradients and the furnace geometry. Therefore a mechanism of analyte transport in non- isothermal furnaces that is based on the above experimental results and the Tdapp definition can be proposed (Fig. 8). There are two different situations under which this mecha- nism can be considered low and high furnace heating rates. Low furnace heating rate In this case the velocity of the analyte diffusional propaga- tion is greater than the velocity of propagation of the temperature front along the tube length. When the tempera- ture of the central part of the furnace where the sample is located reaches Tap the rate of sample vaporization becomes high enough to detect atoms in the gas phase.The atoms diffuse from the sample towards the ends of the tube within a medium having a monotonically decreasing temperature (Fig. 8). When the atoms reach the region in the furnace where the temperature is equal to Tdapp they adsorb-con- dense onto the surface (solid lines in Fig. 8). Then as the furnace temperature is increased the thermal wave also propagates along the tube length so that the adsorbed-con- densed atoms revaporize and propagate further towards the tube ends where at Tdapp they will adsorb-condense again (broken lines in Fig. 8). As the furnace is heated the process can be repeated several times. Therefore for low heating rates of non-isothermal furnaces the actual mechanism for analyte transfer is a cascade mechanism whereby the analyte Fig.7 Absorbance versus time profile for 100 ng of Hg using atomization from the wall. Shadow spectral filming images of the processes ( h ) prior to and (c) after the peak absorbance. Figures beside the frames art their numbers beginning at onset of atomization. The images were recorded at 12 frames s-IJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY APRIL 1993 VOL. 8 395 Fig. 8 Qualitative representation of the cascade mechanism of analyte transfer. ( a ) Distribution of the temperature along the tube length (I,) at initial (solid line) and following (broken line) moments of time; and (0) longitudinal section of the tube atoms propagate to the ends of the fLirnace as a result of numerous condensation- vaporization processes. In this case the length of the atomic layer at any moment oftime is equal to the length I of the furnace region where the temperature is higher than Tdapp.The atom residence time in this case will be determined mainly by the time that is required for the furnace ends to reach the Tdapp value. In the experiments with Ag described above relatively low heating rates were used [Fig. 5(a) curve T,] when the final atomization temperature was only reached more than 10 s after heating was started. This explains the anomalously high residence time of Ag atoms in Figs. 2 and 4. High -furnace heating ratp The velocity of propagation of the Tdapp temperature front along the furnace length is higher than the velocity of analyte propagation in this situation.The analyte that is vaporized in the central hot part of the furnace will always exist in a region where the temperature is greater than TdaDP and the secondary adsorption-desorption processes will not play an important role. The atom residence time 5 can be estimated in this instance by the equation z=12/8D where I is the actual length of the furnace. Additional evidence supporting the above cascade mechanism is presented in Fig. 9. This figure presents the graphite furnace top (a) and lower (b) parts that were obtained by cutting the furnace along its length after an atomization cycle. The particles in the bottom of the furnace (b) are the residue of alumina particles after the furnace had been heated to 2300 "C under an internal gas flow of 150 ml min-".The symmetrical light-coloured rings beside the particle are deposits of alumina as a result of vapour condensation. Composition of the particles was established using X-ray diffraction. The rings are located close to the particle because the internal gas flow prevents analyte propagation to the ends of the tube. Furthermore the cold gas entering the furnace through its ends increases the longitudinal temperature gradients. This also leads to the earlier condensation of the vapour i n the central part of the furnace. Heating of the furnaces to higher temperatures leads to the deposit rings moving from the central part of the furnace towards its ends in accordance with the above cascade mechanism described above. Fig. 9 Photographs of the furnace interior ( a ) top half of the furnace with the sample dosing hole; and (b) lower half of the furnace with the residue of alumina particles and condensation deposits A.Kh. G. and J. C. Hutton for assistance in preparing the manuscript . 1 2 3 4 5 6 7 8 9 10 1 1 12 13 14 15 16 17 18 19 20 References 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. Welz B. Sperling M. Schlemmer G. Wenzel N. and Marowsky G. Spectrochim. Acta Part B 1988 43 1187. Rayson G. D. and Holcombe J. A. Anal. Chim. Acta 1982 136 249. Smets B. Spectr-ochim. Acta Part B 1980 35 33. Molcombe J. A.. Rayson G. D. and Akerlind N. Spectro- chim. Acta Part B 1982 37 319. Manning D. C. and Slavin W. Spectrochim. Acta Part B 1987 42 755. Sedykh E. M. Belyaev E. I. and Sorokina E. A. Zh. Anal. Khim. 1980 35 2162. Salmon S. G. and Holcombe J. A. Anal. Chem. 1982 54 630. Van den Broek W. M. G. T.. and De Galan L. Anal. Chern. 1977,49 2176. Welz B. Atomic Absorption Spectrometry VCH Weinheim 1985. L'vov B. V. and Rayunov P. A. Zh. Anal. Khim. 1985 49 614. L'vov B. V. and Yatsenko L. F. Zh. Anal. Khim. 1985 49 626. Huie C. W. and Curran C . J. Jr. Appl. Spectrosc. 1990 44 1329. L'vov B. V. Spectrochim. Acta Part B 1990 45 533. Guell 0. A. and Holcombe J. A. Spectrochim. Acta Part B 1989 44 185. Matousek J. P. and Powell H. K. L. Spectrochim. Acta Pard B t988 43 1167. Fonseca R. W. Guell 0. A. and Holcombe J. A. Spectro- chim. Acta Part B 1990 45 1257. L'vov B. V. Novichichin A. V. and Polzik L. K. Spectro- chim. Acta Part B 1992 47 289. Gilmutdinov A. Kh. Abdullina T. M. Gorbachev S. F. and Makarov V. L. Spectrochim. Acta Part B 1992 47 1075. NOTE-Refs. 1 and 2 are to Parts 1 and 2 of this series respectively. The authors thank Professor C. L. Chakrabarti as this paper was written during a stay at his laboratory by Paper 2/032626 Received June 22 I992 Accepted November 10 1992