ModiÆer effects in the determination of arsenic, antimony and bismuth by electrothermal atomic absorption spectrometry Leon Pszonicki* and Jakub Dudek Institute of Nuclear Chemistry and Technology, Dorodna 16, 01-231 Warsaw, Poland Received 23rd July 1999, Accepted 14th September 1999 The effect of palladium, mixed palladium±magnesium and magnesium modiÆers on the determination of arsenic, antimony and bismuth was tested. It was found that palladium modiÆer works correctly only in nitric acid solution.If the sample solution contains hydrochloric acid then palladium becomes a strong interferent and causes signiÆcant losses of the analyte. For arsenic and antimony, but not for bismuth, this shortcoming may be eliminated by use of mixed palladium±magnesium or magnesium modiÆer. Magnesium modiÆer was found to be superior to the mixed modiÆer since it is able to eliminate also the negative effects of perchloric acid and of the iron group elements. It is, however, completely ineffective in relation to bismuth when used individually or in a mixture with palladium.Palladium may be used as a modiÆer for the determination of antimony and bismuth in hydrochloric acid solution only when it is preliminarily reduced to the metal form. The mechanism of the modiÆer activity is discussed. A very effective technique for arsenic, antimony and bismuth determination by atomic absorption spectrometry is hydride generation.1 It is necessary, however, to keep the elements in solution in a low oxidation state, it is sensitive to many interfering sample components and, in the case of organic samples, it requires very good mineralization of the organic matrix.All these shortcomings have led to various methods using electrothermal atomic absorption spectrometry for the determination of these elements. Since the above elements and many of their compounds are volatile, chemical modiÆers must always be added to the samples to prevent losses during the preliminary phases of the atomization process.Frech2 applied chromium and nickel salts as modiÆers for the determination of antimony. Ediger3 used nickel nitrate for the determination of arsenic and Gladney4 for the determination of bismuth. Later, when palladium became popular as a very versatile modiÆer, it was applied to the determination of arsenic,5 antimony6 and bismuth.7 Welz and co-workers8±10 suggested improving the activity of the palladium modiÆer by using a mixture with magnesium nitrate.Qiao and Jackson11 showed, using scanning electron microscopy, that an increase in pyrolysis temperature is accompanied by the formation of large palladium droplets, by increasing migration of palladium towards the edges of the platform and by broadening of the absorption peaks of the analyte. Addition of magnesium nitrate, owing to its very low melting temperature, eliminates all these effects. During the pyrolysis phase, magnesium causes a homogeneous distribution of palladium and the analyte in the form of small droplets in the centre of the platform.These results suggest the physical nature of the activity of magnesium used in a mixed modiÆer. On the other hand, the application of magnesium nitrate individually as modiÆer for the determination of many elements in sea-water12 and a comparison of its activity with that of palladium and palladium±magnesium mixture suggests that its chemical activity should also be considered.Kildahl and Lund13 found palladium nitrate modiÆer to be superior to nickel nitrate and nickel sulfate in the determination of arsenic and antimony. Various workers have used palladium modiÆer either as the chloride or the nitrate.14,15 In general, however, the mechanism of its action is not discussed and remains unknown. The effect of chloride was usually investigated by using sodium chloride and in most cases it was found to be negligible.Bermejo- Barrera et al.12 did not Ænd any effect of sodium chloride on the determination of arsenic in sea-water in the presence of palladium nitrate, mixed palladium and magnesium nitrates and reduced palladium up to a concentration of 20 g l21. The investigation of the effect of hydrochloric acid, which may be present in the sample solution, has been little studied although this acid is often used in a mixture with other acids for the dissolution of samples and is not always carefully removed.Such an effect on the determination of lead with palladium modiÆer16 indicates that palladium in the presence of hydrochloric acid may be partially transformed into the chloride. Palladium chloride decomposes during the pyrolysis phase and evolves free chlorine atoms, causing losses of lead. This observation affects the suggestion that palladium chloride may also be used as a modiÆer since in some situations, when during the drying of the sample a sufÆcient amount of hydrochloric acid is present, it can become an interferent.In general, one still observes many inconsistencies in the description of modiÆer effects. The mechanism of their action is usually unknown and seldom discussed. Most observations have a purely empirical character and concern a given element, a given type of sample or given atomisation conditions. Therefore, they are hardly transferable to the other analytical systems. The aim of these studies was to test systematically the behaviour of arsenic, antimony and bismuth during atomization in a graphite tube with a platform in the presence of palladium and palladium±magnesium modiÆers.The effect of nitric, hydrochloric and perchloric acid and their mixtures was also tested. Experimental Apparatus A Thermo Jarrell Ash (Franklin, MA, USA) SH 4000 atomic absorption spectrometer, equipped with a Model 188 controlled furnace atomizer (CTF 188) and Smith±Hieftje background correction system, was used.A Visimax II (Thermo Jarrell Ash) hollow-cathode lamp for arsenic, antimony and bismuth and pyrolytic graphite-coated graphite tubes with a pyrolytic graphite platform were used for all measurements. J. Anal. At. Spectrom., 1999, 14, 1755±1760 1755 This Journal is # The Royal Society of Chemistry 1999Reagents All solutions were prepared from high purity analytical-reagent grade compounds using ultra-pure water (resistivity 18 MV cm21) obtained with a Milli-Q water puriÆcation system (Millipore, Bedford, MA, USA).Hydrochloric and nitric acid used to prepare all solutions were puriÆed by sub-boiling distillation using a quartz subboiling apparatus (Kuerner Analysentechnik, Rosenheim, Germany). Perchloric acid (70%, Suprapur) was obtained from Merck (Darmstadt, Germany). Arsenic, antimony and bismuth stock standard solutions (10 mg ml21) were prepared by dissolution of As2O3, Sb2 O3 and Bi2O3 in nitric or hydrochloric acid and dilution with water to 1 mol l21 acid concentration. Palladium chloride solution (20 mg ml21 Pd) was prepared by dissolution of palladium chloride in hydrochloric acid and dilution with water to 1 mol l21 acid concentration.Palladium nitrate solution (20 mg ml21 Pd) was prepared by dissolution of palladium metal sponge in nitric acid and dilution with water to 1 mol l21 acid concentration. Magnesium, iron(III), calcium, nickel and cobalt nitrate stock standard solutions (50 mg ml21 of the element) were prepared by dissolution of the oxides in nitric acid and dilution with water to 1 mol l21 acid concentration.Magnesium and iron(III) chloride solutions (50 mg ml21 of the element) were prepared by dissolution of the oxides in water and dilution with hydrochloric acid to 1 mol l21 acid concentration. Sodium chloride solution (50 mg ml21 Na) was prepared by dissolution of sodium chloride in water. Procedure In all experiments, the integrated absorbance of arsenic, antimony and bismuth was measured.The applied measurement parameters are presented in Table 1 and the atomisation time±temperature program in Table 2. In the experiments, if not indicated otherwise, one of the following two types of solutions were always used: type 1, 50 ng ml21 of an analyte (arsenic, antimony or bismuth) in 1 mol l21 nitric or hydrochloric acid or in their 1z1 mixture (solutions without modiÆers); type 2, 50 ng ml21 of an analyte (arsenic, antimony or bismuth) and 300 mg ml21 of a modiÆer [palladium, magnesium, iron(III) or a mixture of palladium and magnesium] in 1 mol l21 nitric or hydrochloric acid or in their 1z1 mixture (solutions with modiÆers).When a mixture of palladium and magnesium was applied then the concentration of each of them was equal to 300 mg ml21. ModiÆers were introduced into hydrochloric acid solution in the form of chlorides and into nitric or mixed nitric±hydrochloric acid solution in the form of nitrates.Using an Eppendorf pipette, 10 ml aliquots were always injected into the tube, corresponding to amounts of 0.5 ng of analyte and 3 mg of modiÆer. Results and discussion Although arsenic, antimony and bismuth belong to the same group of the periodic system and for determination of the Ærst two of them almost identical procedures are often proposed, their volatility and behaviour in the presence of modiÆers during atomisation in a graphite tube show signiÆcant differences.Therefore, the individual discussion of the observed phenomena for each of these elements is justiÆed. Arsenic Arsenic in nitric acid solution dropped on the platform in the graphite tube is thermally stable up to almost 600 �C (Fig. 1, curve A). Above this temperature one observes losses, increasing gradually with increase in temperature. This suggests that arsenic, present in the solution in the form of arsenic acid, is decomposed during the drying and early pyrolysis stages to oxide and successively reduced to elemental arsenic that sublimes at temperatures above 600 �C.Palladium nitrate added to the solution extends the range of arsenic stability up to 1100 �C (curve D). Such behaviour agrees with commonly known facts and justiÆes the application of palladium as modiÆer. In hydrochloric acid solution arsenic is present in the form of chloride and it is lost almost completely during the drying stage below 200 �C (curve B). This result corresponds to the previous observations of Galban et al.17 Krivan and Arpadian,18 using radiotracers and uncoated graphite tubes, obtained different results.They found that arsenic is stable during the drying of the sample and it is lost during pyrolysis in the temperature range 200± 600 �C. Their results suggest that the arsenic solution in hydrochloric acid contained some amount of nitrate ions, and this is discussed below. In our investigations, the addition of palladium modiÆer (in the chloride form) to the hydrochloric acid solution did not improve the situation (curve E), indicating that palladium in hydrochloric acid is not able to form any stable compound with arsenic at low temperature.The situation becomes much more complex when arsenic is in a mixed solution of nitric and hydrochloric acid (Fig. 1, curve C). In this solution arsenic is partially transformed into the chloride and lost during the evaporation and drying of the sample below 200 �C.The residue is in the form of arsenic acid and, therefore, the shape of curve C is similar to that of curve A. An unexpected phenomenon occurs when palladium modiÆer is added to the mixed solution (curve F). Instead of stabilisation of the measured arsenic signals at higher temperatures, one observes their signiÆcant suppression. The Table 1 Parameters for the determination of arsenic, antimony and bismuth Parameter Arsenic Antimony Bismuth Wavelength/nm 193.70 217.60 223.10 Bandpass/nm 2.0 0.4 0.4 Background correction Smith±Hieftje Smith±Hieftje Smith±Hieftje Signal pulse lamp current/mA 5.0 6.0 6.0 Background pulse lamp current/mA 3.5 2.5 3.0 Table 2 Temperature±time program Atomization Drying Pyrolysis 1 Pyrolysis 2 As Sb Bi Cleaning Temperature/�C 120 900 200 2100 2200 2100 2400 Ramp/s 40 10 10 0 0 0 – Hold/s 10 5 0 4 4 4 2 Purge Ar Low Medium Medium Off Off Off Medium 1756 J.Anal. At. Spectrom., 1999, 14, 1755±1760difference between curves C and F is particularly large above 400 �C. This range corresponds to the temperature of the decomposition of palladium chloride to elemental palladium and free chlorine (about 500 �C).It suggests that the evolved chlorine forms the volatile chloride with arsenic, which is lost during the pyrolysis stage. The relatively small difference between the curves below 400 �C is due to the fact that palladium chloride is not decomposed during the pyrolysis stage but at the beginning of the atomization stage.Under these conditions, however, within a very short time of operation and with the `gas stop' function switched on, only a small amount of the arsenic chloride formed is able to leave the tube and be lost. The rest is decomposed again to free atoms and measured. Curve A in Fig. 2 demonstrates the effect of various amounts of hydrochloric acid added to a solution of arsenic in 1 Mnitric acid in the presence of palladium. It shows that even a very low concentration of hydrochloric acid (0.05 M) added to the solution causes the loss of almost all the arsenic during pyrolysis.Curve B was obtained under identical conditions, except that palladium nitrate was not added to the sample solution but introduced directly onto the platform in the tube and pyrolysed at 900 �C to the form of elemental palladium. Then the sample was dropped into the tube and atomized in the usual way. In this case, the losses caused by hydrochloric acid during the drying stage are relatively small, up to about 30% (compare also the difference between curves A and C at 200 �C in Fig. 1). They achieve a constant level at a concentration of hydrochloric acid of 0.2 M. The difference between curves A and B in Fig. 2 is due to the formation of palladium chloride in the sample solution containing free hydrochloric acid. Free chlorine, evolved in the decomposition of this chloride during pyrolysis, causes almost total loss of arsenic.The results obtained in the presence of preliminarily pyrolysed palladium chloride provide further conÆrmation of the mechanism proposed above (Fig. 3). When the sample without modiÆer is introduced into the tube on the palladium chloride preliminarily pyrolysed at temperatures below 600 �C, one observes signiÆcant suppression of the signal. This suppression decreases with increase in the temperature of preliminary pyrolysis and it disappears at 600 �C when all the palladium has been reduced to the metal.A similar suppression effect occurs when arsenic is preliminarily pyrolysed with palladium from nitric acid solution at 900 �C. Then the process is stopped, an additional amount of palladium in the form of chloride is added to the tube and the normal atomization process is carried out from the beginning (Fig. 4, curve A). The suppression effect is proportional to the amount of added palladium chloride up to 3 mg and above that level it quickly approaches saturation.When the additional palladium is added in the form of palladium nitrate the suppression is not observed (curve B). The last two experiments indicate unambiguously that palladium chloride, present in the sample during the pyrolysis stage, always causes losses of arsenic even if the latter is already bound in a refractory arsenic±palladium compound. It should be emphasised that the above effects are observed only in the presence of palladium chloride or palladium nitrate and free hydrochloric acid.Addition of chlorides of various metals to the arsenic and palladium nitrate modiÆer solution in nitric acid is without effect, even if the amount of the introduced chloride ions is several times larger than that introduced in the form of free acid. This probably results from the fact that the afÆnity of chloride ions to these metals is higher than that to palladium and, therefore, during the drying Fig. 1 Pyrolysis curves for arsenic: A, in nitric acid; B, in hydrochloric acid; C, in a mixture of nitric and hydrochloric acid (1z1); D, in nitricE, in hydrochloric acid with palladium nitrate; and F, in a mixture of nitric and hydrochloric acid with palladium nitrate.Fig. 2 Effect of hydrochloric acid concentration on the arsenic signal: A, in nitric acid with palladium nitrate; B, in samples with various concentrations of HCl introduced into the tube on palladium preliminarily pyrolysed at 900 �C; and C, in nitric acid with palladium and magnesium nitrate. Fig. 4 Effect of additional portions of palladium added to the preliminarily pyrolysed arsenic with palladium: A, palladium added as chloride; and B, palladium added as nitrate. Fig. 3 Effect of preliminary pyrolysis temperature of PdCl2 on arsenic solution in nitric acid. J. Anal. At. Spectrom., 1999, 14, 1755±1760 1757stage, when water and free acids are being removed, palladium chloride is not formed.Moreover, all these chlorides are either volatile as chloride or decompose at high temperatures during atomization stage. The only exception was found for iron(III) chloride, which at 315 �C is transformed into iron(II) chloride with evolution of free chlorine. Its effect in the presence of hydrochloric acid is similar to that of palladium chloride. In this case, however, the situation is complex since iron, at least in the nitrate form, is also a good modiÆer for arsenic determination.However, when it is present together with palladium it causes serious suppression of the arsenic signal. This last problem will be discussed later. It was mentioned above that arsenic±palladium compounds, formed during the pyrolysis stage, are sensitive to the action of chlorine evolved from palladium chloride. The exact investigation of these compounds was carried out in the following way. Arsenic together with palladium nitrate in the nitric acid solution was dried and pyrolysed in the tube at various temperatures in the usual way.At the end of the pyrolysis stage the process was stopped and the gaseous hydrochloride or gaseous chlorine was introduced to the tube using a plastic syringe with a quartz needle. Then the pyrolysis was repeated at a constant temperature of 400 �C with the `gas stop' function switched on, followed by the atomization stage. The results obtained in the presence of gaseous hydrochloride are represented by curve A in Fig. 5.The arsenic± palladium compound (probably palladium pyroarsenite, PdAs2O5) obtained during the drying stage below 200 �C is almost completely lost during the second pyrolysis in the atmosphere of gaseous hydrochloride. When the temperature of the Ærst pyrolysis is increased, the compound is transformed into a form more resistant against gaseous hydrochloride and above 600 �C losses of arsenic are not observed. As the second pyrolysis is carried out in an atmosphere of gaseous chlorine, the losses of arsenic are very high and independent of the type of arsenic±palladium compound formed (curve B).These results conÆrm in an indirect way the previous hypothesis that the losses of arsenic occurring during the pyrolysis stage at temperatures above 400 �C can be caused only by free chlorine evolved in the decomposition of palladium chloride. All problems resulting from the losses of arsenic in the presence of some amount of hydrochloric acid can be completely removed by using the mixed palladium±magnesium nitrate modiÆer, as demonstrated by curve C in Fig. 2. Comparison of the pyrolysis curves obtained with palladium and palladium±magnesium modiÆers shows that with the latter arsenic is stabilised towards higher temperature only to an insigniÆcantly greater extent (Fig. 6, curves A and B) than with palladium alone. The above data seem to conÆrm the recommendation of a mixed palladium±magnesium modiÆer as being very versatile. 8±10 However, the question arises of what the role of magnesium nitrate is. Does it support only the activity of palladium or does it play a more independent role? The pyrolysis curve for arsenic (Fig. 6, curve C) obtained in the presence of magnesium nitrate alone, without palladium, has the same character as that obtained with palladium alone (curve A). An identical curve is obtained for arsenic in hydrochloric acid solution when magnesium nitrate is used.The addition of palladium improves only insigniÆcantly the stability of arsenic compounds in the range above 1000 �C, as demonstrated by the difference between curves B and C. These experiments prove that magnesium nitrate plays the role of an independent modiÆer for arsenic and its activity is comparable to that of the mixed modiÆer in all types of solution. The identical behaviour of the magnesium modiÆer and the mixed magnesium±palladium modiÆer indicates that, at least in the mixed nitric±hydrochloric acid medium, magnesium plays the main role.It forms with arsenic, during the evaporation of the sample solution, a compound resistant to gaseous hydrochloride and chlorine during the pyrolysis stage in the temperature range up to 1000 �C (Fig. 5, curves C and D). Moreover, the excess of magnesium nitrate decomposes at a low temperature to magnesium oxide, which can act as a trap for free chlorine evolved during pyrolysis. Magnesium nitrate alone is superior to palladium used individually as modiÆer when the sample solution contains hydrochloric acid.It should be emphasized, however, that magnesium is active only as the nitrate and, therefore, it may be applied only when an amount of nitrate ions is present in the sample solution. The dependence of magnesium activity on the concentration of nitrate ion is shown in Fig. 7 and indicates that magnesium achieves its full activity as modiÆer already at a low concentration of nitrate ions of about 0.05 mol l21.Exactly the opposite situation is observed for palladium, which loses its modifying properties and becomes an interferent in the presence of a small amount of hydrochloric acid (Fig. 1, curve F and Fig. 2, curve A). Another acid often used in the preparation of samples for analysis, particularly for the mineralization of organic matter, is perchloric acid. It is well known as a strong interferent in electrothermal AAS. Its effect on the arsenic signal cannot be eliminated by application of either palladium (Fig. 8, curve A) or the mixed palladium±magnesium modiÆer (curve B). Application of magnesium nitrate removed this effect completely (curve C). It was mentioned earlier that in nitric acid medium iron is also a good modiÆer for arsenic. The pyrolysis curve (Fig. 6, curve D) obtained in its presence is similar to those in the presence of palladium, magnesium and the mixture of palladium and magnesium.However, if some amount of Fig. 5 Effect of the preliminary pyrolysis temperature on the arsenic signal obtained in the presence of gaseous hydrochloride or chlorine at 400 �C: A, arsenic with palladium in the presence of hydrochloride; B, arsenic with palladium in the presence of chlorine; C, arsenic with magnesium in the presence of hydrochloride; and D, arsenic with magnesium in the presence of chlorine. Fig. 6 Pyrolysis curves for arsenic in nitric acid: A, with palladium nitrate; B, with palladium nitrite and magnesium nitrate; C, with magnesium nitrate; and D, with iron(III) nitrate. 1758 J. Anal. At. Spectrom., 1999, 14, 1755±1760hydrochloric acid is present in the sample solution, one observes the suppression of the measured signals and the atomization peaks become widened and deformed, which makes them difÆcult to interpret. This effect may be eliminated by addition of magnesium nitrate. If iron modiÆer is applied in combination with palladium, then always, independently of the type of sample solution, a signiÆcant suppression and widening of the atomization peaks occur and the magnitude of the observed deformation is proportional to the amount of iron used.Addition of magnesium nitrate does not remove this effect. It may be partially eliminated only by increasing the atomization temperature (Fig. 9), but its total elimination does not occur. This suggelusmn;iron±palladium triple compounds are formed and they are much more refractory than those of arsenic±palladium.An identical effect is observed in the presence of cobalt and nickel. The above phenomenon indicates the limitation of the application of the palladium and mixed palladium±magnesium modiÆers to samples containing the metals of the iron group. This limitation does not concern the magnesium modiÆer. Antimony The behaviour of antimony is similar to that of arsenic, with one exception. Without modiÆers it is stable at a low temperature in hydrochloric acid medium and its pyrolysis curves obtained for nitric and hydrochloric acid solutions are almost identical (Fig. 10, curves A and B). This suggests that antimony exists in both media in the form of antimonous acid. In nitric acid the addition of palladium stabilizes it up to 1000 �C (curve C). Addition of palladium to hydrochloric acid solution causes signiÆcant losses of antimony at temperatures above 400 �C (curve D). The small losses observed in this case for pyrolysis at 200 �C are due to the activity at the beginning of the atomization stage of the palladium chloride that was not decomposed during pyrolysis.Curve E obtained with palladium in the mixed nitric and hydrochloric acid solution is similar, but the losses of antimony at 200 �C are smaller since only part of the palladium was transformed into chloride. All observed interferences may be removed by the use of either magnesium nitrate or mixed palladium±magnesium modiÆer.Unlike arsenic (compare Fig. 2, curve B), antimony is stable during the evaporation of hydrochloric acid solution up to 400 �C (Fig. 10, curve B). Therefore, it may be determined in such a solution with palladium modiÆer pyrolysed preliminarily at a temperature above 600 �C and reduced to the metal. Palladium in this form is resistant to hydrochloric acid introduced into the tube with the sample and palladium chloride, which causes the losses of antimony, cannot be formed.All this concerns, however, only the pure hydrochloric acid medium. If antimony is introduced into the tube on the pyrolysed palladium in the mixed nitric and hydrochloric acid medium, the palladium metal may be partially dissolved. Then some amount of palladium chloride is formed, giving losses of antimony proportional to the concentration of hydrochloric acid in the mixed medium. Perchloric acid and the iron group elements interfere in the determination of antimony with palladium or mixed palladium ±magnesium modiÆer in the same way as described for arsenic. These interferences do not occur when magnesium nitrate alone is applied. Bismuth The effect of the investigated modiÆers on bismuth is radically different from their effects on arsenic and antimony.Bismuth without modiÆers is stable during pyrolysis up to 600 �C Fig. 7 Effect of nitric acid concentration on the signal of arsenic atomized with magnesium from hydrochloric acid.Fig. 8 Effect of perchloric acid on the arsenic signal in nitric acid: A, with palladium nitrate; B, with palladium and magnesium nitrate; and C, with magnesium nitrate. Fig. 9 Atomization peaks of arsenic with palladium in nitric acid with 1 mg ml21 Fe(III) as nitrate: A, without iron at 2100 �C; B, with iron at 2100 �C; C, with iron at 2300 �C; and D, with iron at 2500 �C. Fig. 10 Pyrolysis curves for antimony: A, in nitric acid; B, in hydrochloric acid; C, in nitric acid with palladium; D, in hydrochloric acid with palladium; and E, in a mixture of nitric and hydrochloric acid (1z1) with palladium.J. Anal. At. Spectrom., 1999, 14, 1755±1760 1759(Fig. 11, curves A and B). Palladium added to the nitric acid solution suppresses its signal by about 30% but the atomization peaks obtained are very well shaped and reproducible. This indicates the formation of a refractory bismuth±palladium compound stable up to above 1100 �C (curve C).The observed suppression of the bismuth signal cannot be explained on the basis of our earlier experiments. The fact that the peaks obtained do not change their shape or magnitude with increase in the atomization temperature up to 2500 �C indicates that it cannot be explained simply as the result of a trapping effect, as described by Frech et al.19 Addition of palladium to hydrochloric acid solution causes losses of bismuth in the range between 400 and 600 �C (curve D), similarly as was observed for antimony.For bismuth, however, the atomization peaks are very badly shaped and irreproducible. Exactly the same phenomenon is observed for mixed nitric±hydrochloric acid solution. Unlike arsenic and antimony, the application of the mixed palladium±magnesium modiÆer or magnesium nitrate alone for bismuth does not improve the situation. Bismuth may be determined in hydrochloric solution with palladium modiÆer only when palladium is preliminarily pyrolysed to the metal form, as described for antimony.This does not concern, however, mixed nitric±hydrochloric acid solution. Addition of perchloric acid or the elements of the iron group to bismuth solution in nitric acid containing palladium modiÆer gives the same interferences as those observed for arsenic and antimony. For bismuth, however, these effects cannot be eliminated by replacement the palladium modiÆer with magnesium. In general, it may be stated that magnesium nitrate, applied either alone or in combination with palladium, is absolutely inert in the relation to bismuth. Conclusion Palladium is a really good modiÆer for the determination of arsenic, antimony and bismuth only when the sample to be analysed is dissolved in nitric acid.If the solution contains some amount of hydrochloric acid it may become a strong interferent and, in extreme cases, it can cause the total loss of these elements during the pyrolysis stage. This effect is due to the transformation of palladium, at least partially, into the chloride that decomposes at about 500 �C, evolving free chlorine atoms.At this temperature chlorine is able to form very volatile chlorides with the analyte atoms, even if they are already bound in analyte±palladium compounds. In relation to arsenic and antimony this effect can be easily eliminated by use of a mixed palladium±magnesium modiÆer or magnesium nitrate modiÆer instead of palladium. Both of them work correctly when only some amount of nitrate ions is present in the sample solution. However, they are ineffective for bismuth in the mixed nitric±hydrochloric acid solution.The fact that magnesium alone is an equally good modiÆer as the palladium±magnesium mixture indicates that it works by itself. The suggestion of some workers that its role is limited to supporting the palladium activity seems to be only marginally correct. The protection of arsenic against its losses during the evaporation of the sample solution containing a large excess of hydrochloric acid indicates that magnesium forms with arsenic a compound (probably magnesium arsenite) already in the solution.This compound is resistant to hydrochloric acid and later, during the pyrolysis stage, it is transformed into a more refractory form (probably mixed magnesium arsenic oxide) that is resistant to chlorine gas evolved in the decomposition of palladium chloride. The completely inert behaviour of magnesium towards bismuth shows that the hypothesis of the action of magnesium oxide as a trap for chlorine atoms is unjustiÆed. In such a case magnesium should protect also the elements with which it is not able to form compounds in solution, e.g., bismuth.Magnesium alone used as modiÆer for the determination of arsenic and antimony is superior to the mixture of magnesium with palladium. It is able to protect these elements not only against the effect of hydrochloric acid but also against the effect of perchloric acid and the iron group elements when they are present in the sample. Antimony and bismuth can be determined in hydrochloric acid solution using palladium modiÆer only innitrate or chloride should be prelimi narily pyrolysed in the tube to the metal form, then the sample solution should be added and atomized in the usual way.Acknowledgement Financial support of this research by the Committee for ScientiÆc Research as Project No. 3 TO9A141 11 is greatly appreciated. References 1 D. L. Tsalev, J. Anal. At. Spectrom., 1999, 14, 147. 2 W. Frech, Talanta, 1974, 21, 565. 3 R. D. Ediger, At. Absorpt. Newsl., 1975, 14, 127. 4 E. S. Gladney, At. Absorpt. Newsl., 1977, 16, 114. 5 X.-Q. Shan, Z.-M. Ni and L. Zhang, Anal. Chim. Acta, 1983, 151, 179. 6 X.-Q. Shan and Z.-M. Ni, Acta Chim. Sin., 1981, 39, 575. 7 L.-Z. Jin and Z.-M. Ni, Can. J. Spectrosc., 1981, 26, 219. 8 B. Welz, G. Schlemmer and J. R. Mudakavi, J. Anal. At. Spectrom., 1988, 3, 93. 9 B. Welz, G. Schlemmer and J. R. Mudakavi, J. Anal. At. Spectrom., 1988, 3, 695. 10 B. Welz, G. Schlemmer and J. R. Mudakavi, J. Anal. At. Spectrom., 1992, 7, 1257. 11 H. Qiao and K. W. Jackson, Spectrochim. Acta, Part B, 1991, 46, 1841. 12 P. Bermejo-Barrera, J. Moreda-Pineiro, A. Moreda-Pineiro and A. Bermejo-Barrera, J. Anal. At. Spectrom., 1998, 13, 777. 13 B. T. Kildahl and W. Lund, Fresenius' J. Anal. Chem., 1996, 354, 93. 14 T. M. Rettberg and L. M. Beach, J. Anal. At. Spectrom., 1989, 4, 427. 15 V. I. Slaveykowa, F. Rastegar and M. J. F. Leroy, J. Anal. At. Spectrom., 1996, 11, 997. 16 L. Pszonicki and A. M. Essed, Chem. Anal. (Warsaw), 1993, 38, 759. 17 J. Galban, E. Marcos, J. Lamana and J. R. Castillo, Spectrochim. Acta, Part B, 1993, 48, 53. 18 V. Krivan and S. Arpadian, Fresenius' Z. Anal. Chem., 1989, 335, 743. 19 W. Frech, L. Ke, M. Berglund and D. C. Baxter, J. Anal. At. Spectrom., 1992, 7, 141. Paper 9/905984I Fig. 11 Pyrolysis curves for bismuth: A, in nitric acid; B, in hydrochloric acid; C, in nitric acid with palladium; and D, in hydrochloric acid with palladium. 1760 J. Anal. At. Spectrom., 1999, 14, 1755±1760