Characterization and Reduction of Silver Matrix Induced Effects in the Determination of Gold Iridium Palladium Platinum and Rhodium by Graphite Furnace Laser-induced Fluorescence Spectrometry ERIC MASERA PATRICK MAUCHIEN" AND BERNARD REMY CEAILaser Analytical Spectroscopy Group CEN Saclay DCCIDPEISPEAISPS 91 191 Gij-sur- Yvette France YANNICK LERAT Kodak European Research Analytical Laboratory Kodak Path6 CRT 71 102 Chalon-sur-SaBne France The determination of ultratraces of Au Ir Pd Pt and Rh in silver nitrate by graphite furnace laser-induced fluorescence (GF-LIF ) spectrometry is subject to strong interference by the formation of condensation. Two kinds of condensation are clearly identifiable diatomic molecules ( AgH Ag,) and condensed particles. In order to find the best analytical parameters for the determination of Au Ir Pd Pt and Rh in silver nitrate by GF-LIF a detailed study of the origin and spa tio-temporal behaviour of these condensations was carried out.Solutions for matrix interference reduction are proposed these include the use of neon as a purge gas a preliminary evaporation of part of the matrix prior to atomization and the use of a transverse heated atomizer. Keywords Graphite furnace laser-induced fluorescence spectrometry two dimensional imaging condensation phenomena silver matrix; gold; iridium; palladium platinum; rhodium Graphite furnace laser-induced fluorescence spectrometry (GF-LIF) is very well adapted to ultratrace determinations in dense matrices because of its high sensitivity and its ability to atomize high salt content samples.'92 For these reasons GF-LIF was chosen for quantitative determination in the ng g-' range of precious metals (Au Ir Pd Pt and Rh) in silver nitrate.The limits of detection obtained for precious metals in silver nitrate depend upon the intrinsic sensitivity of the technique relative to the element and upon matrix interference. Limits of detection down tong I-' can be achieved for precious metals in pure ~ a t e r ~ . ~ thus dilution of silver nitrate with distilled water to a concentration of 1-10 g 1-' is necessary owing to the low levels of precious metals in silver nitrate (ng g-'). When 20 pl of a l o g 1-1 silver nitrate solution are introduced into the furnace the resulting mass of silver after decomposition of the matrix is 130pg.Such a large amount of silver can generate interferences in the determination of precious metals. Indeed the first attempts carried out for the determination of gold in silver nitrate showed that for 130pg of vaporized silver a large non-specific fluorescence signal was superim- posed with the analytical signal of gold. The intensity and the molecular nature of this background was not compatible with the limits of detection required in our study. * To whom correspondence should be addressed. Journal of Analytical Atomic Spectrometry A better understanding of all the processes occuring during the atomization is necessary in order to be able to eliminate the perturbations induced by the silver matrix. Therefore a first assembly of a laser and a gated charge coupled device (CCD) camera was made for the spatio-temporal study of the species present during the atomization.Set-ups dedicated to furnace imaging have already been used suc~essfully~~~ for the investigation of some analyte and matrix distribution in the furnace. Non-uniform atom distri- butions during the atomization of the elements of the alu- minium group were pointed out by Gilmutdinov et al.,' using electrodeless discharge lamps a monochromator and a classical cine camera. Using hollow cathode lamps narrow bandpass filters and a CCD camera more detailed information on aluminium was obtained by Chakrabarti et aL6 The aluminium atoms and the aluminium oxide distributions were determined and atom condensations and oxidations were proved to take place during the atomization of 20 pg of aluminium.Using a dye laser as a probe enables the acquisition of pictures of species distributions in the fluorescence mode as well as in the absorption mode. Measurements performed in the absorption mode allow the visualization of the atom distributions and the condensations of the matrix. The fluores- cence mode is more selective and allows the two-dimensional imaging of each particular species. Thus atoms molecules and condensed particles can be clearly distinguished using the appropriate excitation wavelengths. In a previous paper,7 the first results of this study were presented. Details were given of condensation phenomena occuring in the centre and at the ends of the furnace during the atomization. Two kinds of condensation were identified.Diatomic mol- ecules mostly AgH and Ag are formed during the atomization even for low silver-content samples. These molecules induce strong spectral interferences as their fluorescences overlap the fluorescence of analyte atoms. Condensed particles are observed at the cooled ends of the furnace and in its central part. These condensations strongly affect the gold fluorescence intensity and distribution for very high amounts of silver relative to the case of gold in a pure water matrix. The modifications in the gold fluorescence distribution are attri- buted to the trapping of gold atoms on the condensed particles and to pre- and postfilter effects (absorption and scattering of the incident laser beam and absorption of the gold fluorescence respectively).In order to find the optimum silver matrix atomization Journal of Analytical Atomic Spectrometry March 1996 Vol. 11 (21 3-223) 21 3parameters giving the highest analytical performances for the determination of precious metals a detailed study of the condensations was carried out. The spatio-temporal evolution of the species formed during atomization and their dependance on the atomization temperature and amount of silver are presented in this paper. The origin of the condensation is discussed. The effect on the LIF analytical signal for gold of the condensation is then discussed. Solutions for matrix interference reduction are proposed according to the origin and nature of the condensations involved in the perturbations. The solutions for the reduction of matrix interference are simple due to the necessity to perform routine determinations of precious metals in silver nitrate.Three methods are proposed. The first method concerns the modification of the gas phase different purge gases are tested in order to determine the best thermal and collisional conditions for silver atomization. The second method concerns molecular formation. Since the condensations are completely (condensed particles Ag,) or partially (AgH) linked with the matrix it is possible to reduce their formation by reducing the amount of silver prior to atomization. This reduction of the amount of atomized silver is done using a high temperature pre-heating step. This solution is possible only when the analyte is far more refractory than the matrix and results are discussed according to the physico-chemical properties of the precious metals The third method concerns the use of a Transverse Heated Graphite Atomizer (THGA) which is sup- posed to reduce matrix condensations.Silver atomizations have been performed to evaluate the perturbations induced by the silver matrix in the THGA. EXPERIMENTAL Instrumental Set Up Both the set-up used for the acquisition of pictures and the conventional GF-LiF set-up are described in a previous paper.7 The THGA used is a prototype built in collaboration with the Department of Analytical chemistry Chalmers University of Technology Goteborg Sweden. It has been described by Lundberg et aL8 Temperature Program The same HGA 500 power supply (Perkin-Elmer USA) was used for experiments with the EHGA (End Heated Graphite Atomizer) or the THGA concept.Atomizations were carried out in pyrolytic graphite-coated graphite tubes with a one second ramp time. Unless specified drying was carried out at about 100 "C under 300 ml min-' internal argon flow and ashing was carried out at 1000°C under the same argon flow. Atomizations were carried out at 2200 "C in the gas stop mode. Spectroscopic Data Information on condensed particles was obtained using a CCD camera operated in the absorption mode.7 As condensed particles scatter all wavelengths no special care had to be taken for dye laser tuning. Molecular compounds were studied using either the CCD set-up working in the fluorescence mode or the classical GF-LIF ~ e t - u p . ~ AgH was excited with the laser tuned on a rotational transition of the (0-0) vibrational transition from the electronic systems X to C.Fluorescence was detected at 333 nm corre- sponding to the (0-0) AgH vibrational transition (system A to X). For CCD imaging a 130nm passband filter (maximum transmission 80% at 350 nm) enabled integration of the strong- est fluorescence bands of AgH with good rejection of the laser scattering. 21 4 Journal of Analvtical Atomic Svectrometrv. March 1996 Ag was excited at 264.9 nm (X to C electronic systems). This excitation wavelength gave a strong Ag fluorescence signal at 285 nm (probably B to X electronic transition). For CCD imaging the same passband filter as for AgH was used. This did not enable integration of the strongest band of Ag (low transmission under 300 nm) but enabled rejection of the excitation wavelength with good efficiency.Gold atoms were excited with the laser tuned on the 242.795 nm line. The fluorescence intensity was measured at 312.3 nm using the spectrometer or with the CCD camera using a 10 nm bandwidth interferential filter centred at 313 nm. Laser Probing In the experiments devoted to the acquisition of pictures the laser beam was extended with a divergent lens in order to probe the whole furnace volume. When the classical GF-LIF system was used a 3 mm diameter laser beam was adjusted in the middle of the fur- nace section. Presentation of the Pictures Pictures obtained in the fluorescence mode are presented in 'false' colours. Black corresponds to no fluorescence emission and white to maximum fluorescence emission.Intermediate colours are those of the rainbow. We have chosen to present pictures with as much colour contrast as possible the pictures have been normalized so that all the colours from black to white appear in all the pictures. The raw digital data represen- tative of the true fluorescence intensity of the studied species are presented on diagrams. One vertical and two horizontal sections named respectively VS HS1 and HS2 are extracted from the picture. Fluorescence intensity is presented uersus position in the furnace (expressed in millimetres). Fig. 1 is a scheme of the outline of the furnace showing the location of the extracted sections and positioning of the labels in the furnace (- 3 + 3 and the zero label in the middle of the furnace).Pictures were taken at a 25 Hz repetition rate. Four pictures (tl-t4) representative of an atomization were chosen to illus- trate the spatio-temporal evolution of each compound (Au Ag AgH) studied in the fluorescence mode. Pictures of the condensed particles were obtained in the absorption mode. They are presented in black and white; where dark areas correspond to absorption of the incident laser radiation. For these pictures the corresponding times VS * . . . . . . -3 . . . . . . . Q -3. . . . . . . . . . . . . +3 . . . . . . +3 . . . . . . . . . . HS 1 HS2 Fig. 1 Outline of the furnace and location of the extracted sections VS HS1 and HS2 shown in Figs. 4 6 8 and 13. Labels +3 -3 are expressed in millimetres Vol. 11( t in seconds) are indicated below the pictures.The instant t = 0 corresponds to the beginning of the ramp time. argued to explain the initiation of the condensation cloud. Firstly a slight temperature gradient probably exists between the graphite walls and the argon gas phase and the large silver RESULTS AND DISCUSSION Formation of Condensed Particles At the beginning of the atomization the gas expansion leads to the expulsion of a minor part of the sample through the injection hole and in a second step to the diffusion of the major part towards the water cooled tube ends.' When about 130 pg of silver are volatilized the resulting density of the silver atoms is similar to the initial argon gas density. If the silver atoms meet a colder atmosphere during expansion or diffusion the saturation pressure is reached and condensations appear.These condensations have been dis- cussed for several elements (Au Cu Ag Mn) by Frech and L'vovlo.ll who observed non-specific signals in atomic absorp- tion spectrometry (AAS) measurements not only at the ends of the tube but also in the centre. In the following part a description is given of the condensed particle phenomena observed in the furnace with and without the use of a L'vov platform. Explanations are based on assumptions best supporting the experimental results obtained with the CCD camera. Central condensed particles This type of condensation was observed only when atomiza- tions were performed without a L'vov platform. Pictures of condensations in the centre of the furnace recorded in the absorption mode are presented in Fig.2. In the first picture condensation appears as a diffuse disc centred in the furnace. The disc then becomes more and more dense. In the third picture two different parts of the condensation cloud are clearly visible a very dense central part surrounded by a more diffuse cloud. The dense central part then falls down towards the bottom of the furnace where it dissociates. The diffuse external cloud dissociates before touching the bottom of the graphite tube. The problem of central condensations has already been described in a previous paper,7 but this phenomenon was not completely understood at that time. Pictures of the condensed particles were obtained by recording the black-body emission. Complementary information relative to pictures in absorption was obtained by studying the emission of condensed particles.Pictures in emission showed the same central positioning of the condensed particles but also gave prominence to the presence of a condensation tail going from the injection hole to the centre of the furnace. According to the pictures of the central condensed particles taken in emission and in absorption two major causes can be pressure leads to condensation i n the colder regions. Secondly entrance of cold argon in the furnace (external argon flow) can justify the presence of the condensation tail from the injection hole to the centre of the f ~ r n a c e . ~ After the cloud formation is initiated a chain reaction effect appears. The gas phase heating process occurs mainly by conduction the cloud acting as a thermal shield.Consequently the density of its central part is increased by the trapping of more and more silver atoms. When atomizations are performed with an inserted platform no condensed particles appear. The platform concept was created to achieve better isothermal conditions during the atomization. The sample is mostly heated by radiation and slightly by conduction leading to a delayed atomization which occurs at more isothermal conditions. Peripheral condensed particles It is well known that the EHGA suffers from a non-uniform longitudinal temperature. Even after the temperature equilib- rium has been reached the ends of the furnace walls are about 1000 K colder than its central part. Consequently condensation appears at the ends of the furnace.The expansion of the peripheral condensation in the EHGA when 130pg of silver are atomized under argon atmosphere at 2200 "C is shown in Fig. 3. Condensation occurs at the ends of the tube but only part of those observed in the pictures take place in the furnace volume. Simple naked eye obser- vations of the window support volume proves that the vertical centred stripes (observed until t = 6 s) do not expand in the furnace volume but at the location of the graphite contacts and in the extension tube volume. This makes the results very consistent with gas phase temperature measurements performed at the ends of the furnace by Welz et al.' At the water cooled ends of the furnace a radial temperature gradient is created from the middle of the tube section to the graphite walls.Consequently silver located close to the walls of the furnace condenses at the cooled ends while no condensation appears in the middle of the tube section. Silver is then drained out of the tube volume and condenses in the extension tube volume in the form of verti- cal stripes. Molecular Compounds Only a few reports on molecular fluorescence as a background in analytical LIF spectroscopy can be found in the literature. Sjostrom12 has reported the fluorescence of NaCl as a spectral interference in the determination of gallium. Liang et aL2 have Fig. 2 Pictures recorded in the absorption mode showing the forma- tion of condensed particles in the centre of the furnace during the atomization of 130 pg of silver from the walls at 2200 "C.Times are indicated in seconds Fig. 3 Pictures recorded in the absorption mode showing the forma- tion of condensed particles at the ends of the furnace during the atomization of 130 pg of silver from a platform at 2200 "C. Times are indicated in seconds Journal of Analytical Atomic Spectrometry March 1996 Vol. 11 215t2 t3 t4 Fig. 4 Pictures recorded in the fluorescence mode showing the spatioteniporal distribution of Ag molecules formed during the atomization of 130 pg of silver from a platform at 2200°C (colour key see under Experimental). The excitation wavelength is 264.9 nm. The diagrams below the pictures indicate the raw digital data over three sections of the furnace (cf. Fig. 1). Black line tl red line t black dashed line t and green line t4 reported a molecular fluorescence background in the determi- nation of antimony and tellurium in tap water.Nevertheless most of the backgrounds reported in analytical LIF spec- trometry originate from blackbody radiation stray light and concomitant laser ~cattering.'*'~.'~ In our previous paper,7 we suggested that two molecular species Ag and AgH are generated during the atomization of silver. These molecules induce an important background overlapping the LIF analytical signal. A detailed study of these compounds was performed using the following diagnostics spatiotemporal behaviour of the molecules (CCD laser imaging technique); evolution of the molecular fluorescence intensities versus amount of atomized silver; and evolution of the molecular fluorescence intensities versus atomization temperature.For all the experiments the L'vov platform was used. Ag molecules Silver is vaporized at a temperature lower than those of the precious metals of interest. As a result of this atomization of silver matrix at analytical conditions creates a high silver pressure and consequently the formation of Ag molecules. Spatiotemporal evolution of the Ag molecules. The global evolution of Ag fluorescence when 130 pg of silver are atom- ized in the furnace is shown in Fig. 4. Picture t3 shows that the Ag fluorescence distribution is not homogeneous over the furnace section. The maximum fluorescence intensity is located just above the platform and a minimum intensity appears close to the graphite walls. The reduction of the Ag fluorescence at the vicinity of the walls may come from molecular dissociation on active sites of the heated carbon.Another explanation could be an increase of the fluorescence yield due to better thermal and collisional conditions achieved in the centre of the furnace. Evolution of Ag signal as a function of amount of silver [Fig. 5(a)]. An Ag fluorescence signal is detected even for 20 pg of silver nitrate. When the amount of silver is increased the height of the peak (maximum concentration) and its time duration increase (longer residence time). -Q ? 0 11' I ,'\ -' ' 130 pg 2 4 6 .- E O .- E 0.2 a I@) 2 0.1 5 $ 2 0 0 5 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 a 1 t -0.7 ' I I I I I I 0 2 4 6 8 Time/s Fig. 5 (a) Evolution of the Ag fluorescence intensity with silver iimount. Atomization from a platform at 2200°C.Peak areas are 260 1500 2500 and 4750 (arbitrary units) for 13 65 130 and 380 pg of silver respectively. (b) Evolution of the Ag fluorescence intensity with the atomization temperature for 130 pg of silver vaporized from a platform 21 6 Journal of Analytical Atomic Spectrometry March 1996 Vol. 11Evolution of Ag signal as a function of atomization temperature [Fig. 5(b)]. For a given amount of atomized silver (130 pg) signals registered at different temperatures (from 1800 to 2800°C) show two different Ag formation rates. For 2200 2000 and 18OO0C an equilibrium is reached (plateau) in the supply and removal of Ag,. For temperatures above 2200°C (Ag boiling point) the Ag fluorescence curves are more sharp. At the same time the heights of the peaks decrease possibly because of thermal dissociation and an increase in the rate of diffusional losses.AgH molecules The origin of the AgH molecules is not as clear as for Ag molecules. Among other hydride compounds AIH molecules have already been identified by Ohlsson et al.” during atomiza- tions in graphite furnaces. Ohlsson et a1.l’ concluded that AlH molecules are formed in the gas phase and that the sample solution is probably the main source of H,. A hydrogen partial pressure of 200-600 Pa was experimentally determined in the atomizer. In our experiments four assumptions can be made to explain the presence of the hydrogen pressure considering the furnace structure and the operating conditions pollution of the purge gas; entrance of air through the injection hole; pollution induced by the dilution water; and desorption of hydrogen by the graphite walls.Experiments have been performed in order to test these different hypothesis. Hydrogen free argon was produced by flowing argon through heated copper oxide (800 K) and then through a cold trap (173K) to condense water. No noticeable difference in the AgH fluorescence amplitude appeared when purified argon was used relative to untreated argon. The second and third assumptions were checked using deuterium as a tracer. A high partial pressure of deuterium oxide vapour was created around the furnace when deuterated water was vaporized in highly divided drops close to the injection hole. No AgD fluorescence appeared in the furnace when the external argon protection flow was on.In the same way when silver nitrate was diluted in deuterated water no tl t2 AgD fluorescence appeared and the AgH fluorescence was not reduced. As the three most probable external hydrogen contributions have been shown to not be responsible for the strong AgH concentration it must therefore be concluded according to our original assumptions that the main part of the hydrogen comes from the graphite crucible itself. When the graphite is brought up to 2000-2400°C hydrogen can desorb from the solid graphite or from the pyrolytic coating made of pyrolysed hydrocarbons. As a consequence it is impossible to avoid the presence of hydrogen in the graphite tube during the atomiz- ation step. Spatiotemporal evolution of the AgH Juorescence (Fig.6). The spatiotemporal evolution of AgH showed the same non- homogeneous distribution as for Ag,. The maximum fluorescence intensity appeared in the centre of the furnace with a decrease of intensity towards the graphite walls. Evolution of AgH signal as a function of amount of silver [Fig. 7(a)]. From the peak heights obtained it is apparent that an instantaneous equilibrium between dissociation and formation of AgH was achieved for 65 pg of silver. For silver amounts above 65 pg lack of hydrogen was the limiting factor for AgH formation. When the amount of silver was increased the residence time and the vaporization time increased. Consequently the integrated AgH fluorescence increased. Evolution of AgH signal as a function of atomization temperature [Fig.7(b)]. Four atomization temperatures were tested from 2000 to 2600°C. As the atomization temperature increased the fluorescence signal occured in a shorter time because the diffusion speed increased. But as hydrogen was certainly emitted over the whole tube length and by the graphite contacts the AgH signal occured over a long time despite the silver diffusion. This study showed that even for low silver content samples Ag and AgH fluorescence signals appear. At the atomization temperatures and silver nitrate concentrations studied the intensities of these fluorescence signals was strong. As these signals are temporally and spatially superimposed with the t3 t4 Fig. 6 Pictures recorded in the fluorescence mode showing the spatiotemporal distribution of AgH molecules formed during the atomization of 130 pg of silver from a platform at 2200 “C (colour key see under Experimental).The excitation wavelength is 242.7 nm. The diagrams below the pictures indicate the raw digital data over three sections of the furnace (cf. Fig. 1). Black line t red line t2 black dashed line t3 and green line t4 Journal of Analytical Atomic Spectrometry March 1996 Vol. 11 21 70 -0.2 -0.4 -0.6 -0.8 3 -1 c .- c E 2 Y -1.4 0- .- v) nr - c .c 0 a 5 -0.2 s -04 -06 3 - -0.8 -1 -1 2 -1 4 - 1 6 L - - r ~ I I I I r 1 0 2 4 6 8 10 Timels Fig.7 (a) Evolution of the AgH fluorescence intensity with silver amount. Atomization from a platform at 2200 "C. Peak areas are 950 2200 3450 and 5100 (arbitrary units) for 13 65 130 and 380pg of silver respectively.(b) Evolution of the AgH fluorescence intensity with the atomization temperature for 130 pg of silver vaporized from a platform analyte fluorescence signal the interference occuring may limit the analytical performance. Consequences of the Condensations on the Determination of Precious Metals Condensed particles and molecular formation induced inter- ference were studied in the case of gold determination. Three types of spectral interferences were created by the silver matrix diatomic molecular fluorescence; condensed particles black- body radiation and laser scattering on condensed particles. The major problem was found to originate from diatomic molecules as their fluorescence overlapped the LIF analytical signal. Condensed particles do not emit fluorescence but only continuous radiations.As the conventional GF-LIF set-up uses a gated detection this continuous background had little effect on the analytical signal. The major spectral problem induced by the condensed particles was strong laser scattering. For elements with excitation and fluorescence wavelengths close to each other strong perturbations can be expected. The spatial interferences observed in the presence of the silver matrix can be attributed to trapping and filters effects. Condensed particles have an important part in this phenomena. Spatial Interferences densed particles at the ends of the tube. When important silver amounts are atomized condensations appear prior to the total diffusion of the gold atoms creating pre- and post-filter effects and a trapping effect.For the same reasons if atomizations are performed without a platform the central condensed particles induce a darkening in the gold fluorescence. Fig. 8(a-c) summarizes these different situations with (a) a homogeneous evolution (b) an atomization perturbed by cen- tral condensed particles and (c) perturbations induced by peripheral condensed particles. For silver amounts giving weak condensations the effect of silver on gold is a decrease in sensitivity. When increasing silver amounts (Fig. 9) from 13 up to 380 yg the signal of 30pg of gold is divided by a factor of two. It appears more and more delayed and trapping quenching and filter effects by silver affect the shape of the fluorescence profile of gold. Spectral Interferences The dye laser used in our experiments is a multimode laser with a 0.2 cm-' ( 6 GHz) spectral bandwidth. The non-linear crystal (BBO) used to produce the UV wavelengths roughly doubled the spectral bandwidth of the dye laser.The Voigt profile of the 242.8 nm gold line has a spectral width of about 4 GHz in the furnace. Consequently maximum sensitivity is achieved in the optical saturation rate as the laser line covers the whole bandwidth of the gold line. The small bandwidth of the dye laser allows a good elemental selectivity. But the selectivity of the laser in excitation might be insufficient for molecular bands as their spectra present numerous transitions. Fig. 10 illustrates this problem. Rovibronic transitions (R and Q branches [0-0) vibrational transition of the [X-C] electronic systems) of AgH were recorded by scanning the laser wavelength in the silver vapour. The fluorescence collected at 333 nm exhibits a maximum each time the laser wavelength corresponds to a molecular transition.All the molecular trans- itions observed in Fig. 10 were identified according to a paper by Ringstrom et In Fig. 10 the 312.3 nm fluorescence signal of a gold hollow cathode lamp was also recorded under the same scanning conditions. It appears that two rovibronic transitions of AgH are very close to the gold line. Consequently excitation of the gold atoms at 242.8 nm (41 174.3 cm-') leads to the excitation of AgH molecules. Fig. 11 shows the resulting fluorescence spectra emitted in the furnace between 250 and 350nm when the laser is tuned on the 242.8 nm gold line. The bands observed above 325 nm are clearly attributable to AgH molecules.Part of the signal around 250 nm is due to incomplete rejection of laser scattered light. The signal at longer wavelengths can be due neither to laser scattering nor to AgH fluorescence and the peak at about 285 nm irnplie~'~ an Ag fluorescence. Laser scanning experiments have proved that Ag is indirectly excited by means of an energy transfer from AgH towards Ag,. The molecular background observed at 312.3 nm for gold determination in silver has two components AgH fluorescence and Ag fluorescence. As a consequence a gold blank cannot be simply evaluated when analysis is performed with the standard additions method. Moreover the background associ- ated noise deteriorates the detection limit of the analyte.This study of the molecular background occuring for gold determi- nation in silver shows that matrix induced effects can be very penalizing in GF-LIF because directly or indirectly induced molecular fluorescence can cover a wide range of wavelengths. In a previous paper' it was shown that spatial interferences (modification of the gold fluorescence distribution over the furnace section relative to the case of gold determination in pure water matrix) in the gold fluorescence signal appear only when high silver amounts (about 500 pg) are atomized. These spatial interferences are attributable to the formation of con- Reduction of Matrix Interferences Use of neon as purge gas Neon has a thermal conductivity three times higher than argon.Consequently a temperature equilibrium between the 218 Journal of Analytical Atomic Spectrometry March 1996 Vol. 11tl t2 t3 t4 Fig. 8 Pictures recorded in the fluorescence mode showing the spatiotemporal distribution of gold atoms in silver. The excitation wavelength is 242.8 nm. (colour key see under Experimental). (a) Atomization from a platform of 20 ng of gold in 130 pg of silver at 2200 "C. The diagrams below the pictures indicate the raw digital data over three sections of the furnace (cf. Fig. 1). Black line t red line t black dashed line t and green line t,. (b) Atomization from the walls of 20 ng of gold in 130 pg of silver at 2200 "C (beginning of the atomization); and (c) atomization from a platform of 40 ng of gold in 1000 pg of silver at 2000 "C (end of the atomization) densed particle elimination is a larger silver evacuation through the ends of the tube.Results obtained with a mixture of neon and helium are not as good as for pure neon. Helium has a high thermal conduc- tivity but a poor heat capacity. Moreover it is far lighter than air and probably rapidly escapes out of the furnace in the gas stop mode thus permitting the entrance of cold external gases. AgH fluorescence emitted from the A to X system is reduced six times by a neon atmosphere relative to an argon atmos- phere. On the other hand no change appears in the Ag fluorescence when using neon instead of argon. Collisional processes are probably involved in this phenomenon and work is now in progress to seek a better understanding of these Removal of the condensed particles under a neon atmosphere is a very interesting point for analytical applications using graphite tubes.For LIF spectrometry the improvement mainly from the reduction of the trapping and filter effects. For AAS spectral interferences can be lowered using neon. cn c 3 the injection hole. Consequently less silver diffuses towards h c .- L. c 2. m .- 0 2 4 6 8 results. Time/s Fig. 9 Evolution of the gold fluorescence intensity (30 pg of gold) with silver amount. Atomization from a platform at 2200°C. Peak areas are 900 730 650 and 470 (arbitrary units) for 13 65 130 and 380 pg of silver respectively graphite walls and the gas phase is more rapidly reached. For this reason central condensed particles are no longer visible with the CCD camera when atomizations are performed using neon.The disappearance of peripheral condensations is more surprising. A temperature gradient probably still exists in a neon atmosphere between the ends of the furnace and its central part. A possible explanation for the peripheral con- High temperature ushing step An elimination of the silver matrix prior to atomization is possible only if it does not imply strong losses of analyte in the same time. The precious metals we wanted to determine have very different refractory properties and all of them are more refractory than silver. Gold is the most volatile element and iridium the most refractory element. Platinum is an intermediate case. These three elements are representative of Journal of Analytical Atomic Spectrometry March 1996 Vol.11 21 941 174.3 cm-' I 41300.5 cm-' I R5 R6 Fig. 10 Scanning of the dye laser between 41 350 and 41 150 cm-' in a silver vapour produced in the furnace showing AgH molecular transitions (R and Q branches). The fluorescence of AgH is collected at 333 nm. The position of the 242.8 nm (41 174.3 cm-') gold line is indicated Gold \ Aa7 (induced by ASH) I IU \ + J \ I \ 1 ' round due to ASH + A@ I 1 I I I I I I I 250 300 Fluorescence wavelength/nrn 350 Fig. 11 Fluorescence spectra of the species in the furnace between 250 and 350 nm with the laser tuned on the 242.8 nm gold line (peaks heights are indicative) Table 1 Analytical parameters for the study of Ir P t and Au in silver the various situations that have to be faced for precious metal determinations.Table 1 presents the atomization conditions (temperature use of inserted platform amount of silver) the nature of the background observed at the wavelength of the analyte and the refractory properties of the element. There is a 1729K gap in temperature between iridium and gold boiling points. Table 2 presents results of an analysis performed for ashing temperatures between 800 "C (minimum temperature for total AgNO decomposition) and 1800 "C. The background (bk) was obtained with free analyte samples. The resulting net fluorescence signals (background subtracted) obtained when precious metals were added to the initial solutions are pre- sented. The signal over background ratio is indicated for each ashing temperature. The higher it is the better the analytical conditions.Iridium. The background continuously decreased when increasing the ashing temperature from 1000 to 1800°C. At Boiling point/K L'vov platform Atomization temperature/'C Origin of the background Amount of atomized silver/pg Iridium 4810 no 2600 molecular 260 Platinum 4097 no 2500 molecular (Ag,) 130 Gold 308 1 Yes 2400 molecular (AgH + Ag,) 260 Table2 Effect of high temperatures ashing step on the backgrounds and on the analytical signals for Ir Pt and Au. The corresponding backgrounds are substracted Mineralization temperature/"C 800 1000 1200 1400 1600 1800 Ir Pt Au Background Net signal (S) for Background Net signal (S) for Background Net signal (S) for (bk) 5 Pg Ir (S)/(bk) (bk) 5 Pg Pt (S)/(bk) (bk) 10 P8 Au (S)/(bk) - - - 580 610 1 710 1160 1.6 4400 900 0.2 650 5 80 0.9 530 1330 2.5 3 700 1300 0.3 760 680 0.9 260 1540 5.9 750 1200 1.6 510 590 1.1 60 1440 24 100 900 9 50 40 0.8 25 1220 49 - - - - - - 220 Journal of Analytical Atomic Spectrometry March 1996 Vol.11the same time the net fluorescence signal of 5 added picog- rammes of iridium first increased until a temperature of 1400 "C was applied and then slowly decreased for higher temperatures. It still remained higher than for 1000°C. For ashing tempera- tures higher than 1400"C iridium was carried out with silver. The best ashing temperature was found to be 1600-1800°C with an improvement of 15-30 in the resulting signal over background ratio relative to direct atomization of the matrix without a pre-heating step. Platinum.The platinum showed similar behaviour to iridium. The large improvement obtained for an ashing temperature of 1600°C was principally due to the reduction of the intense background initially measured for atomization without a high temperature pre-heating step. The net fluorescence signal of platinum in silver increased when the ashing temperature was increased from 1000 to 1200°C. On the contrary when the same experiments were performed in a pure water matrix the fluorescence signal of platinum decreased indicating a loss of atoms during ashing. Consequently in the silver matrix measurements competition was shown to exist between the loss of analyte during the ashing step and the reduction of interference (quenching trapping filters effects) during atomization.Gold. The background observed for gold was reduced only when high ashing temperatures were used (about 1600 "C). Indeed the contribution of AgH for the background was hydrogen and not silver limited. This means that the back- ground was strongly lowered only when most of the silver was preliminarily eliminated. But at the same time owing to the volatile property of gold the gold fluorescence signal strongly decreased. For ashing temperatures below 1600 "C the compe- tition between reduction of the interferences and desorption of gold during the ashing step gave a reasonably constant gold signal. For most refractory elements elimination of a large part of the silver matrix prior to atomization is a solution for the reduction of matrix interferences. For the volatile elements (Au and probably Pd) it leads to strong losses of analytes.This study also showed that the use of a high temperature ashing step for background reduction depends not only on the analyte properties but also on the nature of the background which has to be precisely known. In the case of AgH for example a reduction of the amount of atomized silver gives little improvement on the background intensity (in the case of gold analysis). THGA concept The transverse heated graphite atomizer used in our experi- ments was described by Frech and Lundberg.8 Its performance for matrix interference reduction is briefly described in the following part. This type of furnace reduces matrix condensations thanks to its mode of heating as the longitudinal thermal gradient is suppressed by the lateral heating process.Performances of both atomization techniques are discussed in the case of gold in silver nitrate to using GF-LIF spec- troscopy. Experiments were performed with or without plat- form. The temperature program used for these experiments was similar to the one used for EHGA experiments. Formation of condensed particles in THGA. The THGA was kept under an inert atmosphere by an argon flow blown under the graphite crucible. This argon flow was equivalent to the external flow in EHGA; it was maintained during all the analysis steps (drying ashing atomization). In the THGA concept peripheral condensed particles were no more visible for the CCD camera because of the homogeneous temperature along the tube and because of the external argon flow draining matrix products out of the probing volume.In contrast to EHGA the THGA was not clamped in graphite contacts and after diffusion silver was not confined in a cooled area but diffused out of the tube and was subsequently blown out by the argon external gas flow. Central condensed particles were no more visible in THGA experiments. Even when no platform was used condensations were reduced to the point where they were no more visible with the CCD camera laser imaging technique working in the absorption mode. The THGA probably helped in the reduction of condensation as it was not closed at each end by the MgF windows. Using the AAS technique Frech et ~ 1 . ' ~ observed the forma- tion of condensations in a commercial THGA equipped with longitudinal Zeeman effect background correction. This atom- izer had a relatively small tube housing and cooled magnet poles were positioned close to the tube ends.Therefore these condensations were probably particles at the ends of the tube. These contradictory results may come from differences in the furnace design and in the argon flow patterns. Another expla- nation may be a higher sensitivity in AAS experiments. In our experiments condensed particles were strongly reduced in the THGA relative to the EHGA. Molecular formation in the THGA. Diatomic molecules (Ag AgH) were formed in the THGA as in the case of the EHGA. Roughly the same fluorescence intensities were obtained for Ag and AgH using EHGA or THGA. This is not surprising when the origin of the hydrogen (desorption from the graphite volume) and the partial pressure of silver created during atomization are taken into account.Fig. 12 shows the temporal evolution of the Ag and AgH compounds when 200 pg of silver nitrate are introduced into the THGA furnace. Distributions of gold atoms in the THGA. Very interesting results concerning free atom distribution have been obtained by studying atomization of gold in silver nitrate without using a platform. Fig. 13(a) shows atomization of 40ng of gold in 130 pg of silver. Pictures were recorded in the fluorescence mode. They show that when a one second ramp time tempera- ture program is used trapping and filters effects are very important in the THGA. The two lateral linear graphite contacting bridges create a cold line on each side of the furnace where silver condenses.In these regions no gold atoms fluorescence is emitted until the temperature equilibrium is reached. The pictures in absorption [Fig. 13(b)] have been recorded under the same analytical conditions. They represent the evolution of gold atoms between t = O and the instant tl in Fig. 13(a). When the temperature is high enough atoms trapped on 0 2 4 6 8 Time/s Fig. 12 Fluorescence profiles of AgH (black dashed line) Ag (grey line) and 30 pg of gold (black line) during the atomization of 130 pg of silver from the walls in a THGA at 2200°C Journal of Analytical Atomic Spectrometry March 1996 Vol. 11 221tl t2 t3 t4 Fig. 13 @)Pictures recorded in the fluorescence mode showing the spatiotemporal distribution of gold atoms (40 ng) in 130 pg of silver vaporized from the walls at 2200°C in a THGA.The excitation wavelength is 242.8 nm (colour key see under Experimental). The diagrams below the pictures indicate the raw digital data over three sections of the furnace (cj’. Fig. 1). (b) Pictures recorded in absorption of the 242.8 nm gold line showing the spatiotemporal distribution of gold atoms in the THGA between the beginning of the ramp time and t in Fig. 13(u) condensations are suddenly released (pictures t t 3 ) . Due to a dissymmetry in the graphite crucible positioning or because of an increase in temperature of the cooling water (circular cooling system) gold atoms trapped on each side of the tube were not released simultaneously (time between these two successive releases is roughly 0.1 s).This two-step atomization can be seen in Fig. 12. The fluorescence intensity reaches a plateau (1-2.5 s) until gold atoms trapped on the sides of the furnace are released. This study showed that the same problems are encountered in both kinds of atomizers but to a different extent. The sides of the THGA are equivalent to the cooled ends of the EHGA. Condensations occur in both furnace concepts depending upon their geometry. When atomization is performed with a platform this non- uniform gold fluorescence distribution does not appear any- more because atomization of silver is delayed. In this case the heating process has been long enough to reduce the tempera- ture gradient created by transverse heating and atomization occurs under better isothermal conditions.Consequently for determinations in a dense matrix a platform should be used as far as possible. CONCLUSION Several phenomena induced by the atomization of large amounts of silver were characterized. The origin of all the condensation observed during the atomization was determined. The CCD camera and the use of LIF spectrometry as well as 222 Journal of Analytical Atomic Spectrometry March 1996 Vol. 11 absorption measurements were used as diagnostic tools for the study of the phenomena occuring in the furnace. A good understanding of the perturbations induced by the matrix enabled solutions to be developed for the reduction of matrix interference in the determination of ultratraces of precious metals. The use of neon instead of argon strongly decreases the formation of condensation clouds.To our knowledge neon has never been reported as a purge gas for the reduction of interferences in graphite furnace determinations and a com- parison between argon and neon in AAS would be interesting. The use of a high temperature ashing step is convenient for the determination of elements which are far more refractory than the matrix. Nevertheless in LIF spectrometry determi- nations care must be taken about the origin of the background as it may not be reduced in proportion to the evaporated matrix during the ashing step. The loss of analyte atoms can consequently be too high in proportion to the background reduction. Experiments in the THGA must be continued. As the THGA is not clamped in graphite contacts the removal of silver during a high temperature ashing step is certainly more efficient than in the EHGA.Moreover as it is heated over its entire length deposition of silver at the ends of the tube during the ashing step is unlikely thus reducing the amount of silver vaporized during the atomization. Satisfactory conditions for precious metals determinations using GF-LIF are now available and work is in progress to determine the limits of detection of Au Ir Rh Pt and Pd in silver.REFERENCES Dougherty J. P. Costello J. A. and Michel R. G. Anal. Chem. 1988 60 336. Liang Z. Lonardo R. F. and Michel R. G. Spectrochim. Acta Part B 1993 48 7. Remy B. Verhaeghe I. and Mauchien P. Appl. Spectrosc. 1990 44 1633. Masera E. Mauchien P. and Lerat Y. Spectrochim. Acta Part B submitted for publication. Gilmutdinov A. Kh. Zakharov Yu. A. Ivanov V. P. and Voloshin A. V. J . Anal. At. Spectrom. 1991 6 505. Chakrabarti C. L. Gilmutdinov A. Kh. and Hutton C. J. Anal. Chem. 1993,65 716. Masera E. Mauchien P. and Lerat Y. J. Anal. At. Spectrom. 1995 10 137. Lundberg E. Frech W. and Hardy J. M. J. Anal. At. Spectrom. 1988 3 11 15. Welz B. Sperling M. Schlemmer G. Wenzel N. and Marowsky G. Spectrochim. Acta Part B 1988 43 1187. 10 11 12 13 14 15 16 17 18 Frech W. L'vov B. V. and Romanova N. P. Spectrochim. Acta Part B 1992 47 1461. L'vov B. V. and Frech W. Spectrochim. Acta Part B 1992 48 425. Sjostrom S. J. Anal. At. Spectrom. 1990 5 261. Irwin R. L. Butcher D. J. Takahashi J. Wei G.-T. and Michel R. G. J. Anal. At. Spectrom. 1990 5 603. Butcher D. J. Irwin R. L. Takahashi J. Su G. Wei G. T. and Michel R. G. Appl. Spectrosc. 1990 44 1521. Ohlsson K. E. A. Cedergren A. and Frech W. Spectrochim. Acta Part B 1992 47 1525. Ringstrom U. and Aslung N. Arkiu Fysik 1965 32 19. Rosen B. in Spectroscopic Data Relative to Diatomic Molecules Pergamon Press N.Y. 1970. Frech W. and L'vov B. V. Spectrochim. Acta Part B 1993 48 1371. Paper 5/05086C Received August 1 1995 Accepted October 16 1995 Journal of Analytical Atomic Spectrometry March 1996 Vol. 11 223