JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1991 VOL. 6 565 Interference of Copper Silver and Gold in the Determination of Selenium by Hydride Generation Atomic Fluorescence Spectrometry an Approach to the Studies of Transition Metal Interferences Alessandro D'Ulivo and Leonard0 Lampugnani Consiglio Nazionale delle Ricerche Istituto di Chimica Analitica Strumentale Via Risorgimento 35 56 I00 Pisa Italy Roberto Zamboni Dipartimento di Chimica e Chimica Industriale Universita di Pisa Via Risorgimento 35 56100 Pisa Italy An attempt has been made to elucidate the mechanism of liquid-phase interferences caused by Cull Ag' and AulIl on the evolution of hydrogen selenide. An empirical equation able to describe accurately the interference plots obtained by systematic studies has been found.The equation has also been derived using a physical model for the interferences based on the capture of the hydrogen selenide by the active surface of metal colloid aggregates growing in the solution and having a fractal structure. Keywords Hydride generation; non-dispersive atomic fluorescence spectrometry; selenium determination; interference The hydride generation (HG) technique combined with atomic spectrometry as a detection system can be consi- dered at present one of the most powerful analytical tools for the determination of many elements such as As Sb Bi Se Te Ge Sn and Pb.1-4 The main limitation for such a technique stems from interferences arising when some inorganic ions are present in the reaction matrix. Since the first systematic study by Smith5 of the interfer- ence from 48 foreign elements on the determination of As Sb Bi Se Te and Ge many other interesting studies have appeared.6-21 Moreover most of the papers concerning the analytical use of HG coupled with atomic spectrometry have included interference figures from a large number of foreign ions.A careful survey of the above literature shows wide discrepancies in the magnitude of the interference effects reported by different workers. Most of this non-homogene- ity appears to be due to the complex mechanism of action of the interfering species so that many experimental details may play an important role in the determination of the interference figures for a given HG apparatus. It is recognized that the magnitude of a given interference effect depends upon the type of HG system (batch continuous or flow injection) ernployed,l0J the chemical conditions em- ployed in the reduction ~ t e p ~ J ~ J ~ q ~ ~ the mixing sequence of the reagents9.I8 and the type of atomizer.I0J2J1 In order to elucidate the complex mechanism of action of interferences occurring in the liquid phase and caused by the transition metal ions several workers have made some interesting studies using specifically designed experiments.Kirkbright and Taddia,6 reporting on the interference of Cull Nil1 PtlV and PdIr on arsine generation observed the formation of a finely dispersed black precipitate. They proposed that the interference was caused by the reduction of the interfering ions to the metal by sodium tetrahydro- borate.The finely dispersed metals were then thought to adsorb and decompose the arsine on their surface also supposed by Smith.5 In order to support the hypothesis they performed an experiment in which the addition of Ni powder to the reaction solution caused the complete suppression of the As signal. Welz and Melcher13 proposed that metals in the ionic form hardly contributed to the interference effects in the determination of Se. They bubbled hydrogen selenide evolved from a batch HG system through a wash bottle containing the metal ions in acid solution. The resulting measured interferences were several orders of magnitude lower than those observed when the metal ions were present in the reaction vessel. Bax et a/.'* studied the interference of W1 NilI and Cot1 on the determination of Se using a continuous HG system.They employed a specially designed mixing apparatus in order to obtain five different mixing sequences of the analyte interferent and reagents. They showed evidence of the possibility of several interference mechanisms. In particular the elimination of the generated hydrogen selenide by the products (metal and/or metal boride) of the reaction between the interferent ions and sodium tetrahydroborate and the catalytic decomposition of sodium tetrahydroborate by the metal ions or by their reaction products appeared to play the major role. They concluded that the reaction of the metal ion with the hydrogen selenide did not contribute (Nil1 and Co") or hardly contributed (Cull) to the interferences. The main problem of the results described above is their applicability to a real analytical system.As they were obtained by means of indirect experiments some perturba- tions are always present with respect to the original analytical system. The present work is based on the idea that the results of systematic studies in which the depletion of the analytical signal caused by a given concentration of interferent is used to characterize the interference figures for the analytical system should contain information on the mechanisms at the origin of the interferences. For this purpose the depressive effects of Cu Ag and Au on the determination of Se by HG with non-dispersive atomic fluorescence spectrometry (NDAFS) were measured for two different concentrations of Se. A wide range of interferent concentra- tions was investigated in order to obtain well-defined interference plots in the range 0-100% depression of the signal.The peak height peak area and peak shape of the AF signal were considered in order to collect more complete information about the interference phenomena. In this paper it is assumed that the interference plots can be accurately represented by an empirical equation which can be justified by the assumption of a physical model for the interference mechanism. The present study while suggest- ing an alternative method to report interference data by means of their analytical description represents a new approach to gaining direct information on the nature of the liquid-phase interference in the HG technique.566 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1991 VOL.6 Experimental Apparatus The batch HG vessel was made of borosilicate glass (Fig. 1). The atomizer was a miniature argon-hydrogen flame supported on a borosilicate glass tube similar to one used previously.22 The flame was obtained with flow rates of 0.30 1 min-' for hydrogen and 0.85 1 min-' for argon the latter also serving as the carrier gas passing through the HG vessel and sweeping the generated hydride to the atomizer. The HG vessel was connected to the atomizer by means of a 30 cm long borosilicate transfer line (6 mm o.d. 2 mm The NDAF spectrometer was the same as described previou~ly.~~ Reagents All reagents were Suprapure grade (Merck) unless otherwise specified. Standard stock solutions (1000 pg rn1-I) of the elements were prepared by using Titrisol solutions (Merck).The Au standard stock solution was prepared from the pure metal (Johnson Matthey) dissolved in hot aqua vegia [hydro- chloric acid-nitric acid (3 + 1 )] and appropriately diluted with hydrochloric acid. A 5% sodium tetrahydroborate solution was prepared by dissolving the reagent (SpectrosoL Merck) in 0.1 mol dm-3 sodium hydroxide solution and filtering through a 0.45 pm membrane. This solution was kept refrigerated and diluted to the required concentration just before use. Milli-Q water (Millipore) was used throughout. Procedure A 5 ml volume of 0.5 mol dm-3 hydrochloric acid was placed in the reaction vessel and spiked with microlitre volumes of a standard solution of the analyte and of the interfering element according to the measurement to be performed. A pipette containing 1 ml of 1% sodium tetrahydroborate solution was inserted in the injection port.Argon was then allowed to flow through the vessel while the sample was subjected to vigorous magnetic stirring. Fig. 1 Borosilicate reaction vessel used in the present work. 1 Ar carrier gas inlet; 2 Ar carrier gas outlet; 3 disposable micropipette tip for the NaBH injection; and 4 poly(tetrafluorethy1ene) (PTFE)-coated magnet fitted with PTFE rings Table 1 Summary of the experimental parameters Mean power for EDL* Power modulation for EDL* Photomultiplier Photomultiplier voltage RC time constant Atomizer Burner Observation height Argon flow rate Hydrogen flow rate Reaction vessel Transfer line Sample volume and acidity Reducing agent Hydride generation 7 w 7031 Hz square wave 100% amplitude modulation 0.5 duty cycle R759 Hamamatsu 750 V 0.5 s Miniature argon-hydrogen flame Borosilicate glass tube 9 mm i.d.8 mm above the burner top 0.85 1 min-' 0.3 1 min-' See Fig. 1 Borosilicate glass 30 cm long 6 mm o.d. 2 mm i.d. 5 ml 0.5 mol dm-3 HC1 1 ml of 1% NaBH solution Batch mode manual NaBH injection * EDL= Electrodeless discharge lamp. When the baseline had stabilized the sodium tetrahydro- borate solution was injected and the peak recorded. The gases evolved from the reaction solution were directly continuously supplied to the atomizer. A summary of the experimental parameters is reported in Table 1. Since the metals Cu Ag and Au are strong interfering elements they generate a memory interference effect also observed by Meyer et aL7 In order to avoid erroneous evaluation of the signal variation after each interference measurement the bottom part of the vessel was washed with nitric acid (1 + I) then several analyte blanks were run until the signal was restored to the original level.In many instances poisoning of the cell occurred then the bottom part of the vessel had to be replaced with a clean one. Overnight cleaning with aqua regia was found to be effective in recovering the poisoned cells. Results and Discussion General Consideration Dedina,12 in a classification of the interferences occurring in the HG technique divided them into liquid-phase interfer- ences and gas-phase interferences. For transition metals according to current literature,12-16J8 the interference ef- fects occur in the liquid phase.Therefore the discussion contained in the present paper is based on the assumption that the metals considered as interferents give rise only to liquid-phase interferences. The HG-NDAFS apparatus employed in the present study has a concentration detection limit (3a of the blank) of 0.006 ng ml-1 with a dynamic range extending up to 80 ng m1-1.22 Both the selenium concentrations 0.5 and 50 ng ml-l chosen to perform interference experi- ments were within the dynamic range and gave rise to signal levels well above the 0.02 ng ml-l blank level. Interference Measurement and Data Presentation The depressive effects of Cu Ag and Au on the AF signal obtained by the reduction of 0.5 and 50 ng ml-1 of SeIV are reported in Figs.2 3 and 4 respectively.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER I99 1 VOL. 6 567 10 1 o2 1 o3 1 o4 1 o5 Cu"/ng mi-' Fig. 2 Effect of Curl concentration on the AF signal depression of Se. Peak height measurements for A 0.5; and B 50 ng ml-1 S P . Peak area measurements for C 0.5; and D 50 ng mi-' SeiV 100 8? 1 2 50 0 Fig. 3 I # 1 1 I 1 10 1 o2 1 o3 10' 1 o5 Ag'lng ml-' Effect of AR' concentration on the AF signal depression of Sey Peak height measurements for A 0.5; and B 50 ngml-I SerV. Peak area measurements for C 0.5; and D 50 ng ml-1 Se'" Only for Cu was the magnitude of the interference independent of the Se concentration and therefore the threshold above which the interferent effect starts was controlled essentially by the Cu concentration.However in all instances two distinct interference plots were obtained depending on whether peak height area measurements were performed. This discrepancy is correlated to the modifica- tion of the peak shape consequent upon the increasing interference effect as shown in Figs. 5 6 and 7. The Cu interference began its depressive effect at the tail of the peak and then its action extended progressively to the 100 - v a 50 0 1 10 1 o2 1 o3 1 o4 1 o! Au"'/ng ml-' Fig. 4 Effect of Aul" concentration on the AF signal depression of Se. Peak height measurements for A 0.5; and B 50 ng ml-' SeV. Peak area measurements for C 0.5; and D 50 ng ml-1 SeV A 1 A 0 15 30 Time/s Fig. 5 Signal shapes obtained by the reduction of (a) 0.5; and (b) 50 ng ml-1 of Setv in the presence of various concentrations of Cu" A=O; B=O.l; C=0.2; and D=0.5 ,ug ml-1 Cur'.The NaBH solution was injected at t=O s 0 15 30 45 Time/s Fig. 6 Signal shapes obtained by the reduction of 50 ng ml-L of SeIV in the presence of various concentrations of Au? A=O; B=0.5; C=2.0; and D=5.0,ug ml-I Au'l'. The NaBH solution was injected at t=O s whole peak (Fig. 5). The modification of the peak shape with the increasing interference effect was essentially the same at low and high concentrations of Se. At low concentrations of Se the interference behaviour of Ag and Au was similar to that of Cu in terms of peak shape modification. However Ag and Au began to interfere at a lower concentration than Cu. The interference plots at low concentrations of Se show that the peak area depression is never less than the peak height depression.This reflects the sequence of peak shape modifications described above. At high concentrations of Se the interference behaviour568 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 of Ag and Au was different from that observed at low concentrations of Se. The peak shape modification due to the interferences involved a more pronounced tailing of the peak with respect to the peak without interference (Figs. 6 and 7). This behaviour was reflected in the interference plots where at a given interference concentration the peak area depression was never larger than the peak height depression. For Ag the singular inversion in area depres- sion shown by the interference plot was a consequence of the signal splitting into two peaks (Fig.7). A first consideration of the present systematic studies led to the conclusion that the assumption of Meyer et al.' that the magnitude of interferences does not depend on analyte concentration cannot be generally valid. It is also note- worthy that the signal peak height and signal peak area measurements may lead to a different numerical evaluation of the same interference effect. If the role played by these parameters is ignored different interference figures can be obtained by different operators using the same HG appara- tus under the same experimental conditions. However the interference plots shown in Figs. 2 3 and 4 indicate the difficulty in reporting complete and concise information about the interference behaviour using a given HG apparatus.The set of experimental points contained in the plots are tedious to present in a report either in table form or in a pattern image. Therefore an attempt has been made to give a mathemati- cal description of interference plots by fitting the experi- mental data with an equation describing the well-known saturation function C" Ba+cu p=- where [ 100 p=AS(o/o)] is the relative AF signal depression measured experimentally c is the interferent concentration expressed in ng ml-l and B and a are the equation parameters optimized by the fitting procedure. In particular B represents the interference concentration giving a 50% signal depression while a is related to the slope of the rising portion of the interference plot.The results of the non-linear least squares fits for all six sets of the experimental data are presented in Table 2. The quality of the fits on the bases of both the sum of residuals and the coefficient of correlation can be considered more than satisfactory. Figs. 8 and 9 show the worst and the best of such fits respectively. The variable p (O<p=Gl) of eqn. (1) is obtained by experimental measurement of the AF signal F. If Fo is the t - (0 C 0 In LL .- a 60 s H B Time - Fig. 7 Signal shapes obtained by the reduction of 50 ng ml-l SeIV in the presence of various concentrations of Ag! A=O; B=O.l; C=0.2; D=0.3; E=0.5; F=0.7; and G=1.0 pug ml-I Ag'. The NaBH solution was injected at t=O s 0.4 0.2 0 I 10 lo2 lo3 lo4 lo5 lo6 Ag'/ng mlF' Fig. 8 Experimental data (0) and fitted curve (-) obtained for peak area measurements for Ag interference on the reduction of 50 ng ml-l SeIV 1 .o 0.8 0.6 a 0.4 0.2 0 I 1 10 lo2 lo3 lo4 105 10' Au"'/ng ml-' Fig.9 Experimental data (0) and fitted curve (-) obtained for peak height measurements for Au interference on the reduction of 50 ng ml-' SelV AF signal in the absence of interference and F is the AF signal in the presence of interference then If So is the total Se concentration S the concentration of Se leaving the solution in the presence of the interferent and b the slope of the calibration graph then Fo= bSo and F= bS. If s* is the concentration of Se remaining in the solution which is unable to give a signal because of the interference process then p also represents the fraction of Se retained in the solution following the interference p=s*ls (3) Eqn.(1) should possess some physical significance related to the nature of the interference phenomena but for the moment it can be considered as an empirical function giving concise and complete information on the interfer- ence figures using a given HG apparatus with the aim of reducing the fragmentation and discrepancies in the litera- ture data. Assumption of a Physical Model for the Interferences In order to discuss the physical interpretation of the interference plots some assumptions based on the experi- mental evidence or literature data about the nature of the interfering agent and the interference mechanism must be made.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL.6 569 Table 2 Values of parameters obtained in the non-linear least squares fit Peak area- Sell/ Interferent ng ml-I B a CU" 0.5 168k13 1.3k0.11 Ag' 50 293k45 0.77k0.10 AulI1 0.5 55 k 5.6 0.95 k0.08 CU" 50 165i 13 1.5 k0.17 Ag' 0.5 40k3.6 1.6 k0.19 Au"' 50 3500k 138 1.31 50.06 Peak height- SeIV/ Interferent ng ml-I B a CU" 0.5 430k32 1.4 k 0.13 CU" 50 350k33 1.4 k0.16 Ag' 0.5 86k7.3 1.2k0.12 Ag' 50 160k 17 1.3-tO.16 AuI" 50 2050k94 0.48 k 0.02 Au'" 0.5 1 2 1 k l l 1.3-tO.13 z (Error)2 162 200 222 58 1 207 45 c (Error)* 168 27 1 160 47 1 235 43 r2 0.9975 0.9970 0.9960 0.9873 0.9965 0.9979 r2 0.9969 0.995 1 0.9964 0.99 17 0.9953 0.9988 Chemical and physical nature of the interferent The interfering species are supposedly formed by the reaction between the metal ions and tetrahydroborate to give presumably the metal^*^*^^ in the form of particles capable of capturing hydrogen selenide by means of active sites exposed on their ~urface.~.~ The interaction between the surface and the adsorbed molecule can be assumed to be an equilibrium reaction characterized by K,.The microscopic structure of the interferent particles can be supposed to be highly porous with highly irregular boundaries. This assumption is based on evidence from several experiments on metal colloid aggregation in solu- Weitz and co-workers26,28 reported several transmission electron microscopic images of Au colloid aggregates obtained from NaAuCl reduction with triso- dium citrate. The aggregates show a highly disordered structure typical of fractal o b j e ~ t s ~ ' ? ~ ~ and characterized by a mass fractal dimension D ( 1 d X 3 ) obtained by light scattering methods.The fractal dimension of the aggregates is scale invariant i.e. the aggregates grow in a self-similar mode. The dynamics of the aggregation process is depen- dent on the experimental conditions used; fast and slow aggregation processes develop aggregates with different fractal dimensions.26728 The fast aggregation process which seems to be the one that takes place under our experimental conditions is obtained in the presence of a high concentration of electrolytes in solution. In the course of fast aggregation the mean cluster radius R depends both on the initial metal concentration c and time t according R a (ct) lID (4) Considering that m the mass of a cluster is proportional to the r a d i u ~ ~ ~ .~ ~ at a fixed time then the number of the clusters per unit volume of solution is conserved independently of the initial metal concentration c . ~ O maRD ( 5 ) Interference mechanism As shown previously the peak shape changes with the interferent concentration. This explains why the ratio of the peak area to peak height also changes with the interferent concentration. It is also apparent that the analysis of the peak shape modification can contribute to a better insight of the interferent processes and to a better definition of the physical model. Before discussing the phenomena that modify the peak shape it must be realized which processes control the peak shape in the absence of the interferent.The signal shape in the absence of an interferent (Figs. 5 and 6 curve A) can be thought of as being due to the combination of several different processes namely (i) the formation of hydrogen selenide (ii) the transfer of the hydrogen selenide from the liquid to the gaseous phase (ziz) the broadening and delay of the peak due to the dead volume and (iv) the broadening and delay of the peak due to anomalous interactions of the hydrogen selenide31 with the glass surface. Process (i) according to literature data,32 should be very fast compared with the peak lifetime. Considering that the addition of sodium tetrahydroborate is completed in less than 0.5 s this process should not contribute significantly to the peak shape. The contribution of process (iv) is difficult to evaluate.Considering that the cumulative dead volume of the reaction vessel the transfer line and the atomizer was about 90 ml the appearance of a peak maximum should be expected after about 6.5 s instead of the 5-6 s observed experimentally. This small difference can be explained by the fact that during the injection of sodium tetrahydro- borate the argon flow rate was suddenly augmented by the hydrogen evolved from the solution. This phenomenon was also recognizable on the signal as a small shoulder appear- ing just prior to the peak maximum. No anomalous delay in peak appearance as reported by D e d i ~ ~ a ~ l was observed in the batch HG system reported here. So if any anomalous delay occurs it should be of little relevance using the present apparatus.Therefore it can be assumed that the peak shape of the AF signal is due to a combination of processes (ii) and (iii). In the present HG setup any modification of the signal shape caused by the interferent can be correlated to the circumstance that the interference phenomena occurs in whole or in part during the release of the hydride from the reaction solution. The peak shape can therefore be considered to be the result of two dynamic processes having the same starting time the release of hydrogen selenide from the solution and the formation of active sites able to trap the generated hydrogen selenide.570 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 1991 VOL. 6 It can be seen from Fig 5 that the rising portion of the peak remains practically the same in the presence of 0.1 pg ml-l of Cu. During this time hydrogen selenide has already formed and starts to leave the solution without any apparent interference.After this early period the formation of the active sites and the capture of the hydride begins. This evidence seems to exclude the possibility of catalytic decomposition of the sodium tetrahydroboratel* with the consequent depletion in the hydride reaction yield. Considering that the same behaviour is observed inde- pendently of the Se concentration used it is reasonable to suppose that the interference is controlled also by the rate of formation of the active sites. The same conclusion can be made for Ag and Au at low Se concentration levels. For Ag and Au at high Se concentration levels the hydrogen selenide seems at first to be captured and then subsequently partially released in a slow process.This experimental evidence on the basis of the hypothetical model can be explained by the progressive decrease of the active surface consequent upon the growth of aggregates with time and the destruction of some of the active sites available. In conclusion the total number of hydride molecules retained in solution should be considered proportional to the cumulative number of active sites present at the end of the aggregation process of the colloid. For copper it can be supposed that the number of active sites always remains high or that the process of capture leads to the chemical decomposition of the adsorbed hydride. Mathematical description of the model In order to describe quantitatively the model just outlined it is assumed that at a given time after the beginning of the cluster growth the concentration c of the interfering metal ion produces a number n of clusters per unit volume of solution.Each of the clusters possess a mean diameter 2R mean surface area a and a mean mass m. If the clusters are considered to be fractal objects they possess fractal dimen- sions 0 and D for the active surface area and the mass respectively. In the particular instance where the clusters can be considered to be Euclidean objects then D,=2 and D=3. The dependence of the surface area a and mass rn on the radius R of a single cluster will be a = constJPa and m = constJP where const and const are dimensionless constants de- pending on the fractal geometry of the clusters.The dependence of the surface area on the mass will be a = const,m' (6) where const,= const const,-a a= D,/D If c the mass per unit volume of solution of the interfering metal ion generates a mass per unit volume of solution of the interferent c it can be said that c,=h,c where h is a constant related to the type of chemical reaction and its yield. From eqns. (4) and ( 5 ) an increase in c would produce an increase in m maintaining the same number n of clusters per unit volume of solution. Considering the mass of a single cluster m= h,c/n the total surface area per unit volume of solution A = n,a will be (7) and A = (const,n,(l -")h,")c * If it can be supposed that the fractal structure of the aggregates remain unchanged at least in the range of interferent concentrations used in this study then the term in parentheses is a constant and A = const 1P (8) If there are d active sites per unit of surface area the number of active sites M per unit volume of solution will be M,,=d,A and where const2 = constld,.through the following process M = const2P ( 9 ) If the hydrogen selenide is captured by the active sites H,Se+Q=Q* where Q is a free active site and Q* is the active site after the capture of hydrogen selenide the equilibrium constant of the process will be Ki = W/(SM) where W and Mare the number of occupied and free active sites present in the unit volume of solution. Knowing that W=F and assuming W<<M then M=Mo and eqn. (10) becomes PIS= KiMo (1 1) Considering the mass balance So=S+F and the rela- tionship in eqn.(3) eqn. (1 1) becomes p=Mo/(Mo+ l/Ki) Using eqn. (9) and postulating that (const2Ki)-' =Ba then eqn. (1) is obtained. Physical meaning of B and a According to the above mathematical description the parameter B the characteristic concentration of the inter- ferent giving a 50% signal depression is given by B = (const,n,(l -a)h,*d,Ki)- (13) It is thus dependent on chemical and physical parameters strictly related to the nature both of the analyte and of the interfering species the adsorption energy of the analyte on the particle surface the surface structure of the particles generated their number and so on. As B is greatly dependent on experimental conditions eqn. (1 3) justifies the discrepancies existing in the literature about the magnitude of the interference effects reported.The parameter a is related to the fractal structure of the particle as it is 'felt' by the fractal interrogator in this instance the hydrogen selenide molecule and also to the nature of the interaction between the interrogator and the cluster surface. The experimental values of D reported in the litera- t ~ r e ~ ~ are in the ranges 1.95<0,<3.04 for physisorption 1 .6<0,<2. 13 for chemisorption on a dispersed metal catalyst and 0.7 1 <D,<5.8 for heterogeneous catalysis on dispersed metals. In the present study the peak height measurements give a values more homogeneous than those given by peak area measurements. This is probably because the peak height measurements are performed at a time (after the start of the reaction and corresponding to the peak maximum) when the 'ripening' of the interfering clusters gives a better agreement with the assumptions made in the mathematical description.For peak area measurements a represents a mean value evaluated on the whole peak lifetime and might be affected by changes in the morphology of the interfering particles that are likely to occur during a longer time span. Even if no specific data relative to the hydrogen selenideJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY OCTOBER 199 1 VOL. 6 57 1 interaction with colloids of Group IB metals were found in the literature the values of CY reported in Table’2 seem consistent with its physical meaning and with the values of D reported above. Conclusion For liquid-phase interferences caused by Group IB transi- tion metals on the generation of hydrogen selenide in a batch HG system the magnitude of the interference effects as a function of the interferent concentration is related to the mechanism of the interference effect.A relatively simple physical model for the interferences based on the trapping of the hydrogen selenide at the surface of the interferent particles supposedly grown in the solution and having a highly folded surface that is fractal in nature seems to be in good agreement with the experimen- tal results. In fact the relevant equation describes accurately the interference plots and justifies in terms of chemical-physi- cal parameters the different behaviour of the metals considered and also the discrepancies among the interfer- ence data reported in the literature.Considering that the dynamics of aggregation for colloi- dal particles is a complex process dependent on many chemical and physical parameters it is reasonable to suppose that the experimental conditions play a role in determining the magnitude of the interferences not only through the control of the chemical reaction responsible for the production of the interfering species but also through the control of the dynamics of their aggregation into colloidal particles and of their resulting morphology. References 1 Robbins W. B. and Caruso J. A. Anal. Chem. 1979 51 889A. 2 Godden R. G. and Thomerson D. R. Analyst. 1980 105 1137. 3 Nakahara T. Prog. Anal. At. Spectrosc. 1983 6 163. 4 Dedina J. Prog. Anal. At.Spectrosc. 1988 11 251. 5 Smith A. E. Analyst 1975 100 300. 6 Kirkbright G. F. and Taddia M. Anal. Chim. Acta 1978 100 145. 7 Meyer A. Hofer Ch. Tolg G. Raptis S. and Knapp G. Fresenius Z. Anal. Chem. 1979 296 337. 8 Verlinden M. and Deelstra H. Fresenius 2. Anal. Chem. 1979 296 253. 9 Pierce F. D. and Brown H. R. Anal. Chem. 1976 48 693. 10 Pierce F. D. and Brown H. R. Anal. Chem. 1977 49 1417. 11 Hershey J. W. and Keliher P. N. Spectrochim. Acta Part B 1986 41 713. 12 Dedina J. Anal. Chem. 1982 54 2097. 13 Welz B. and Melcher M. Analyst 1984 109 569. 14 Welz B. and Melcher M. Analyst 1984 109 573. 15 Welz B. and Melcher M. Analyst 1984 109 577. 16 Welz B. and Shubert-Jacobs M. J. Anal. At. Spectrorn. 1986 1 23. 17 Petrick K. and Krivan V. Fresenius Z. Anal. Chem. 1987 327 338. 18 Bax D. Agterdenbos J. Worrel E. and Kolmer J. B. Spectrochim. Acta Part B 1988 43 1349. 19 Yamamoto M. Yamamoto Y. and Yamashige T. Analyst 1984 109 1461. 20 Welz B. and Melcher M. Spectrochim. Acta Part B 198 1,36 439. 21 Dittrich K. and Mandry R. Analyst 1986 111 277. 22 DUlivo A. J. Anal. At. Spectrom. 1989 4 67. 23 DUlivo A. Festa C. and Papoff P. Talanta 1983 30 907. 24 Creighton J. A. Blatchford C. G. and Albrecht M. G. J. Chem. SOC. Faraday Trans. 2 1979 75 790. 25 Creighton J. A. Alvarez M. S. Weitz D. A. andGaroff S. J. Phys. Chem. 1983 87 4793. 26 Weitz D. A. and Oliveria M. Phys. Rev. Lett. 1984 52 1433. 27 Matsushita M. in The Fractal Approach to Heterogeneous Chemistry Surfaces Colloids Polymers ed. Avnir D. Wiley Chichester 1989 p. 161. 28 Weitz D. A. Huang J. S. Lin M. Y. and Sung J. Phys. Rev. Lett. 1985 54 1416. 29 Pfeifer P. and Avnir D. J. Chem. Phys. 1983 79 3558. 30 Weitz D. A. Huang J. S. Lin M. Y. and Sung J. Phys. Rev. Lett. 1984 53 1657. 31 Dedina J. Fresenius Z. Anal. Chem. 1986 323 771. 32 Agterdenbos J. and Bax D. Anal. Chim. Acta 1986 188 127. 33 Farin D. and Avnir D. in The Fractal Approach to Hetero- geneous Chemistry Surfaces Colloids Polymers ed. Avnir D. Wiley Chichester 1989 p. 27 1. Paper 0/05149G Received November 16thl 1990 Accepted May 29thl 1991