Determination of sulfur isotope ratios and concentrations in water samples using ICP-MS incorporating hexapole ion optics Paul R. D. Mason,* Karsten Kaspers and Manfred J. van Bergen Faculty of Earth Sciences, Utrecht University, Budapestlaan 4, 3584 CD Utrecht, The Netherlands. E-mail: mason@geo.uu.nl Received 15th March 1999, Accepted 12th May 1999 Sulfur isotope ratios are diYcult to determine by quadrupole ICP-MS due to interfering O2+ and NO+ molecular ions of high signal intensity at isotopes 32S and 34S.Rf-only hexapole devices have recently been introduced into ICP-MS instrumentation to facilitate ion transfer from interface to analyser. By introducing a mixture of ‘reactive’ gases into the hexapole, a series of ion–molecule reactions can be induced to reduce or remove interfering polyatomic species. The eVects of various gas mixtures (He, H2 and Xe) on the transfer of sulfur ions through the hexapole and the breakdown of interfering O2+ and NO+ molecular ions at m/z=32 and m/z=34 were investigated. A rapid charge transfer reaction between O2+ and Xe gives at least a factor of 10 improvement in the S+/O2+ ratio.A further reduction in O2+ is achieved by the addition of H2. d34S variations were investigated in crater-lake waters and waters obtained from springs and rivers on the flanks of volcanoes in Java, Indonesia. Under optimum conditions (S=10–50 mg l-1), the 34S/32S measurement precision for standards and samples was <0.3% RSD.Mass bias errors were corrected by using a concentration-matched in-house standard of average North Atlantic sea-water (d34S=20.5‰). Results compare favorably against published data measured by standard gas source mass spectrometric techniques. The proposed technique is potentially useful as a survey tool due to the large d34S variation (±20‰) encountered in nature and the accuracy and reproducibility of the technique (±3–5‰). The determination of S isotope abundance and S concentration ture [eqn.(1)] is essential in many environmental and geological studies. The 32SO42-(aq)+H234S(g)< 34SO42-(aq)+H232S(g) (1) S-cycle aVects all living organisms and depending on chemical form S can be beneficial, useful or hazardous to man. Sulfur Anaerobic bacteria in ocean floor sediments promote this isotopes are usually measured to high precision on gaseous reaction and can cause significant shifts in 34S/32S due to the SO2 by gas source mass spectrometry or on solids by secondary large relative mass diVerence. 34S/32S variations in nature are ionization mass spectrometry (SIMS). Sulfur concentrations large and are usually reported as d34S relative to the Canyon are commonly determined independently by isotope dilution Diablo Troilite (CDT) [eqn. (2)] MS, X-ray techniques or inductively coupled plasma atomic d34S ‰=[(34S/32Ssample-34S/32Sstandard)/34S/32Sstandard]×1000 emission spectrometry. Sample preparation for isotopic analy- (2) sis of natural waters is time-consuming and more than one analytical technique is usually required for combined isotop- Oxidation usually increases d34S whilst reducing conditions ic–elemental investigations. A single, more rapid and easy to can significantly lower d34S.use technique would be beneficial for large-scale environmental Sulfur isotope ratios in natural water samples are typically or geological chemical surveying. Inductively coupled plasma measured by gas source mass spectrometry.Sulfate in solution mass spectrometry (ICP-MS) is a technique that could poten- is initially reduced to sulfide and precipitated as BaS or CdS. tially fulfil these requirements if current problems related to Sulfides are then converted to SO2 by reactions with CuO, instrumental spectroscopic interferences could be overcome. V2O5 or O2 at high temperature. SO2 gas is introduced into ICP-MS is a versatile technique as many types of sample the inlet of the mass spectrometer and isotope ratios can be introduction method can be utilized (e.g.direct analysis of measured to high degrees of accuracy and precision. solutions, introduction of solids by laser ablation), potentially Disadvantages of this technique include slow sample throughopening up new applications. put and the possibilities of errors due to large amounts of Sulfur is widely distributed in the environment by volcanism, sample handling. volatile emissions, precipitation, acid mine drainage and Sulfur isotope ratios are not routinely measured by ICP-MS, anthropogenic activity.Sulfate is released into the atmosphere as in a solution-loaded plasma there is a significant contrias SO2 by volcanic and biogenic activity. Sulfur can be bution from O-, N- and H-based polyatomic ions that on a concentrated and stored in its reduced form as sulfide in typical instrument with a quadrupole mass spectrometer often metallic minerals that occur in many natural rock types or saturate the counting system due to their prevalent abundance. which are the products of industrial processes.Sulfur concen- However, if these spectroscopic interferences can be reduced trations have a limited use when defining the S-cycle and a or eliminated then there is scope for using ICP-MS to investistudy of S isotopes enables sources, sinks and natural mixing gate this isotope system. Sulfur isotopes are strongly fractionprocesses to be characterised.Sulfur isotopes have been used ated in nature, exhibiting a relatively large 34S/32S ratio to investigate magma de-gassing and the influence of toxic variation and thus a high precision technique may not be crater-lake waters in drainage systems.1–5 essential. The most important isotopes of S are free from Sulfur isotopes are fractionated from one another by elemental isobaric overlap and isotopes 32S and 34S are suYcchanging redox conditions. 32S and 34S are most significantly iently abundant to be measured in the same acquisition by the counting system of a typical ICP-MS instrument.A limitation fractionated by a biogenic chemical reaction at low tempera- J. Anal. At. Spectrom., 1999, 14, 1067–1074 1067Table 1 Instrument operating parameters that remains however is the poor ionization eYciency of S in an Ar plasma due to its high first ionization energy. In Rf forward power 1300W addition, S is relatively light and thus is not transmitted by Cool gas flow rate 13.00 l min-1 typical instrumental ion optics as eVectively as heavier masses.Intermediate gas flow rate 1.00 l min-1 As a result, ICP-MS is a trace rather than an ultra-trace Carrier gas flow rate 0.85 l min-1 Sample uptake rate 0.5 ml min-1 method for determination of S, with expected detection limits Sampling depth 12 mm in the parts-per-billion range. Another potential problem to Sampling/skimmer cones Nickel consider is that S ions may be prone to matrix-induced space Extraction lens -650 kV charge eVects, as observed with light B and Li isotope systems.6 Intermediate lens 400 V Sulfur has been measured successfully by ICP-MS by reducing or eliminating the solvent load of the nebulised solution, Hexapole gases— Optimum He gas flow rate 1.0 ml min-1 through alternative sample introduction techniques, e.g.elec- Optimum H2 gas flow rate 4.0 ml min-1 trothermal vaporization.7 However, residual O2+ has been Optimum Xe gas flow rate 0.15 ml min-1 significant due to continuous air entrainment in the plasma Signal measurement parameters— leading to problems in the determination of 32S.Attempts to Acquisition mode Single ion measuring (peak hopping) measure S isotopes by high resolution ICP-MS8 have been Dwell time 30 ms per peak successful giving excellent precision (<0.2% RSD) and low Points per peak 1 detection limits down to 100 ng l-1. A relatively low resolution Total analysis time 10 min setting of 2000 gives adequate separation of 32S and O2+, conditions under which sensitivity and peak shape are only partially compromised.Multiple collector magnetic sector field Experimental instrumentation is expected to give further improvements in measurement precision for S isotopes but is expensive to install. Instrumentation Recent developments in ICP-MS technology have intro- The ICP mass spectrometer used for this study was a duced ‘collision’ or ‘reaction’ cells for ion focusing9–11 and for Micromass (Manchester, UK) Platform ICP.14 This instru- filtering of selected ions by reactions with gases during ment utilizes an rf-only hexapole to extract ions from the transmission from plasma to analyzer.12–15 H2, Xe, N2, O2, region behind the skimmer cone and to transport them into NH3 and various other gases have been introduced into the the quadrupole mass analyzer (Fig. 1). Electrostatic ion lenses rf-only multipole ‘reaction’ cell in small volumes to promote are used to guide ions into and out of the hexapole, which is ion–molecule reactions.16–19 mounted at an angle to the instrument axis, and for guiding This study was undertaken to investigate the feasibility of ions into the oV-axis quadrupole.A photomultiplier ‘Daly- using an rf-only hexapole ion-focusing device as a tool to type’ detector is used in configuration with an ion–electron remove suYcient O2+, O2H+ and NO+ polyatomic ions to be conversion dynode and a photon producing phosphor plate.20 able to measure S isotope ratios by ICP-MS.A series of In other respects the configuration is similar to that of a ion–molecule reactions between Xe, H2 and O2+ were induced standard commercial ICP-MS instrument. The instrument was to remove the interfering species and thus enable accurate operated with a standard sample introduction configuration measurement of 34S/32S ratios. To test the validity of the of Fassel quartz torch, Scott double pass quartz spray chamber technique some natural volcanic crater-lake and spring waters, and Meinhard nebuliser.Operating conditions are given in previously analyzed by gas source MS, were chosen, which Table 1 and discussed in more detail below. contained high concentrations of S that remained in the mg l-1 range after matrix dilution. Natural isotopic variation in the Samples and reagents sample set was large, enabling meaningful data to be acquired within the resolution of a quadrupole counting system.This Water samples were collected in the field and filtered through paper discusses the possible reactions and processes that lead 0.45 mm Millipore filters.4 They were acidified with HNO3 and to polyatomic ion reduction and comments on the potential transported to the laboratory in polyethylene bottles. Samples of this new technique for routine S isotopic and elemental were diluted by volume with Millipore ultrapure water between 10 and 100 times prior to ICP-MS analysis.For the waters analysis in the earth and environmental sciences. Fig. 1 Schematic diagram of mass spectrometer used in this study. 1068 J. Anal. At. Spectrom., 1999, 14, 1067–1074from Patuha, West Java, identical aliquots of the same samples have been analyzed by ICP-AES, ICP-MS and gas source MS with data reported in the literature.4 The sample from Ijen, a large and chemically homogeneous crater lake in the Kawah Ijen volcano, East Java, was collected in the same month as a sample analyzed by gas source MS reported in a separate study.21 Sulfur isotopic standards are not readily available in solution form.Most internationally recognized isotopic standards are prepared in an insoluble BaS, AgS or ZnS matrix as a consequence of the sample preparation procedures involved in gas source MS. Thus, in this study, it was necessary to find an inexpensive, readily available and well characterized solution standard.Sea-water contains abundant S (#1000 mg l-1) Fig. 2 EVect of addition of Xe into hexapole on ion transmission. Figure shows relative response of a 10 mg l-1 multi-element solution and has a relatively uniform isotopic composition. d34S in before and after the addition of 0.2 ml min-1 Xe to 5.0 ml min-1 He. average sea-water is 20.5‰.22 A 200 ml volume of open ocean sea-water collected from the North Atlantic Ocean near Madeira was stripped of metal cations using a SaYdex ion- multipole device.The number of ions transmitted by the exchange column and collected in a dilute HCl matrix for use multipole will therefore increase as gas partial pressure as an isotopic standard. increases until a maximum is reached and ion transmission starts to decay once more. Collisional focusing is dependent Analysis procedure on the mass of the focused ion and the mass of the collision gas. The energy distribution of ions of a given mass is reduced Diluted SPEX and Johnson-Matthey standard solutions were by over 90%, enabling improved performance of a subsequent used for instrument optimization and calibration for quantitatmass analyser.14 ive analysis.The solutions are certified for elemental concen- In this study, an optimum amount of gas for maximum tration but not for isotope ratios of S. The instrument was focusing of S+ ions corresponded to a He or H2 gas inlet rate optimized on 32S using a 10 mg l-1 single element standard of approximately 5 ml min-1 (partial pressure in the region solution and a blank acid solution. All gas lines leading into 1×10-3 mbar).The use of Xe in the hexapole had a dramatic the hexapole were carefully cleaned prior to analysis and the eVect on transmission of all masses due to the influence of its highest available commercial grade gases were used in all relatively heavy mass on axial ion kinetic energy. Relative cases. Gas input into the hexapole was controlled by mass transmission into the quadrupole was mass-dependent (Fig. 2) flow controllers. and for the addition of 0.2 ml min-1 (#5×10-5 mbar) only To determine the isotope ratio in an unknown sample, the 30% transmission for U and 5% transmission for Be was sample and standard were first diluted to the same concenmaintained. The transmission of S ions was also reduced to tration (usually in the range 20–50 mg l-1) using published 10–20% of the amount measured with only He in the hexapole, ICP-AES data for S concentrations in the samples.4 Two whilst background interferences (mostly O2+) were reduced to blanks were acquired before each block of sample and stanat least <1% of their original abundance (which was close to dards.A sample blank consisting of water and acids used to saturating the counting system). Thus, the improvement factor dilute and acidify the samples and a standard blank using was better than 10 for the S+/O2+ ratio. reagents from the preparation of the standard solutions were run.Standards were run immediately before and after each Ion–molecule reactions unknown sample. For each sample and standard, ten 1 min scans were acquired and statistics calculated for the ten repeat The ease with which ions can react with molecules has been measurements. All data were collected in single ion measuring known since the earliest history of mass spectrometry.26 (peak hopping) mode using one point per peak. Quadrupole Electron or proton transfer reactions (described in dwell and settle times were optimized to take account of Knewstubb27) were induced in this study by focusing the ion dominant instrumental noise frequencies leading to superior beam, extracted from the plasma, through a partial volume of short-term precision.23 The data for the standards, after blank a ‘reactive’ gas in the hexapole (e.g.H2, Xe). The reactions correction, were used to calculate a correction factor, which that are most likely to occur are those that are exothermic was then applied to the ratio measured for the unknown with large reaction rate constants.Reactions have been extensample to overcome mass bias error. sively studied under experimental thermal conditions where the reaction energy is defined as the diVerence in the sums of Results and discussion heats of formation of the products and neutrals (or diVerence in ionisation potentials for charge transfer reactions).Reaction The processes enabling the relatively interference-free rate and cross-section data available in the literature28–30 may measurement of the 32S and 34S isotopes are initially discussed thus provide some guidance as to which reactions take place. and the results of optimisation work for accurate and precise However, the hexapole used in this study was operated under isotope ratio measurements are shown. Data for crater-lake non-thermal conditions where the kinetic energy of the ions and spring waters from Java are presented and discussed in must also be taken into account.Non-thermal conditions tend terms of day-to-day precision and accuracy. to reduce the rate of exothermic and increase the rate of endothermic reactions. In addition, the ions entering the Ion focusing in a hexapole hexapole have high kinetic energies, much of which is lost in energy-damping collisions, but of which a substantial pro- An rf-only multipole device24,25 when filled with inert gas at relatively high operating pressures (10-3–10-4 mbar) can portion remains as ions are focused into the quadrupole.Thus, reaction rate data are used here simply as a qualitative guide focus ions through a process termed collisional focusing.9 Significant amounts of axial kinetic energy are lost, corre- to indicate which reactions are possible. Scanned spectra for the mass region from m/z=30 to m/z= sponding to the energy losses predicted from collision crosssections, and ions are focused towards the centre of the 36 are shown in Fig. 3. Fig. 3(a) shows the case where the J. Anal. At. Spectrom., 1999, 14, 1067–1074 1069abundance of O2+ increased slightly, whilst many other polyatomic ions (including ArO+ and O+) decreased and Ar+ decreased dramatically [Fig. 3(b)]. The slight increase in O2+ may be explained by O2 impurities (ionised by charge transfer with Xe+) in the H2 gas or gas lines. Under routine, solutionloaded plasma conditions, S isotopic measurements are not possible due to a saturation of the counting system at m/z=32.The addition of small volumes of Xe gas into the hexapole cell had a dramatic eVect on signal response [Fig. 3(c)]. Count rates at all masses were significantly reduced, with the exception of the large 14N16O+ peak that remained at m/z=30. With a He–Xe mixture a peak at m/z=32 of the order of 500 000 counts s-1 remained, reflecting a contribution from residual 16O16O+ and 14N18O+.The addition of H2 to the He–Xe mixture promoted a further reduction in background count rates to <100 000 counts s-1 at m/z=32. When a solution containing 10 mg l-1 S was introduced [Fig. 3(d)] a spectrum for S, significantly above the background (>10 times), was observed. In order to assess, qualitatively, the reactions that had been taking place, other parts of the mass spectrum were investigated before and after the addition of Xe as a reactive gas. By observing the peaks produced during the addition of Xe it is possible to suggest a series of reactions that may have been important.Table 2 summarises a suggested reaction series in the hexapole cell involving Xe and H2 gases and some abundant ions. Thermal reaction rate data from the literature are shown, where available, indicating that such reactions are tentatively supported by experimental work.29,30 Xe is reactive, undergoing endothermic charge transfer with O2+ in experimental work at 290 K.32 The non-thermal environment of the hexapole in this study may contribute suYcient kinetic energy to the reactant ion to increase dramatically the rate of this reaction above the rate predicted by experimental thermal data.The Xe background was typically very low when Xe was not used in the hexapole, showing that plasma Xe contamination, a problem frequently encountered with Ar plasma gases, was minimal. Major products of reactions with Xe included Xe+ and XeH+ accompanied by more minor XeO+, XeOH+ and XeH2O+ (Fig. 4, Table 2). The formation of Xe+ and XeH+ has been observed in previous studies12 and attributed to charge and proton transfer reactions between Xe and Ar+ and ArH+. Much of the Xe+ and XeH+ was also the result of charge transfer reactions with O2+ and our data show an identical isotopic abundance with the results produced by Rowan and Houk.12 When H2 was substituted for He with Fig. 3 EVects of addition of various gases into hexapole on response between m/z=30 and m/z=36.The spectra are for a 2% v/v ultrapure the Xe in the hexapole, the overall response for Xe+ and HNO3 solution unless stated otherwise. (a) Introduction of He only XeH+ fell and proportionally more XeH+ was formed. H2 (inlet flow rate of 5 ml min-1). (b) Introduction of a 1.0 ml min-1 appears to have promoted the reduction of O2+. This is not He–4.0 ml min-1 H2 mixture. (c) Addition of Xe to the 1.0 ml min-1 supported by thermal reaction rate data, but in the non- He–4.0 ml min-1 H2 mixture.(d) As in (c) but with aspiration of 10 mg l-1 S single element solution in 2% v/v HNO3. Table 2 Suggested reaction series when using Xe and/or H2 as a reactive gas. Thermal reaction rate constant data, where available, are from Anicich and Huntress29 and Anicich30 eVects of ion–molecule chemistry have been minimized. He, which does not react with most charged species (exceptions Thermal reaction rate constant k are H2+, D2+ and He+), was the only gas introduced into Reaction at 290–300 K/cm3 s-1 the hexapole. Impurities that can react with the ion beam, such as N2 and O2, were present in the high grade He at the O2++XeAXe++O2 5.5×10–11 O2++XeAXeO++O — 0.1 ml l-1 level (Grade 6.0, Air Products) and could have made OH++XeAXeH++O 9.2×10–10 some small but relatively insignificant modifications to this H2O++XeAXe++H2O 8.0×10–10 part of the mass spectrum.Large peaks were observed at all Xe++H2AXeH++H <2.0×10–11 masses, but most notably at m/z=30, 32, 33 and 34 correspond- XeO++H2AH2O++Xe — ing to NO+ and O2+ polyatomic ions.This is similar to a XeOH++H2AH3O++Xe — background spectrum for a dilute HNO3 solution, observed Ar++XeAXe++Ar 4.3×10–13 ArH++XeAXeH++Ar — on ICP-MS instruments with standard electrostatic ion optics, Ar++H2AArH++H 8.6×10–10 where ion–molecule reactions do not take place.31 ArH++H2AH3++Ar 8.9×10–10 H2 gas has previously been used in collision cells as a OH++H2AH2O++H 8.6×10–10 reactant to remove ArX+ species13 (where X=O, H, N, Ar, H2O++H2AH3O++H 8.3×10–10 Cl, etc.).As H2 was added to the He in our experiments, the 1070 J. Anal. At. Spectrom., 1999, 14, 1067–1074Table 3 Stability of S standard solutions in various matrix types and at diVerent concentrations Counting Acid matrix 34S/32S 1s RSD (%) statistics 10 mg l-1 S standard— De-ionised water 0.044 0.00018 0.41 0.22 2% v/v HNO3 0.042 0.00012 0.30 0.25 2% v/v aqua regia 0.046 0.00014 0.29 0.34 2% v/v HCl 0.047 0.00009 0.19 0.22 5% v/v HCl 0.047 0.00009 0.18 0.25 50 mg l-1 S standard— 5% v/v HCl 0.050 0.00008 0.16 0.18 Optimization of ICP-MS Plasma parameters such as rf forward power, torch positioning and carrier gas flow were carefully optimized prior to each experiment.Minor fine-tuning was carried out to the ion lens settings, but the adjustments were relatively insignificant. Fig. 4 Some products of ion–molecule reactions with Xe.Count rates for both S+ and background ions were optimized to give the highest S+/background ion signal, alternately using a 10mg l-1 S single element standard and a blank, both in a 2% v/v ultrapure HNO3 matrix. thermal environment of the hexapole the rate of this reaction Experiments were performed with a sheathing bonnet34 may have been substantially increased. between the outer end of the plasma torch and the ion XeO+, XeOH+ and XeH2O+ were observed in the mass extraction aperture to observe if any further gain in reduction region m/z=140–156.A close assessment of the isotopic of the O2+ polyatomic ion could be achieved due to a reduction abundance of the observed peaks reveals that XeOH+ and in entrainment of atmospheric gases. No real gain in perform- XeH2O+ were formed more readily or removed less readily ance was observed, suggesting that much of the O2+ may be by subsequent reactions than XeO+. When H2 was added to originating from the solvent carrying the sample into the the He–Xe gas mixture, the XeO+, XeOH+ response was plasma.dramatically reduced. Meanwhile, H3O+ increased to a much The eVects of diVerent acid matrix types and analyte greater degree than would be expected if this light polyatomic concentrations on S isotope response and ratio measurement ion was only the product of reactions between H2, OH+ and precision were investigated (Table 3). Optimum S+ response H2O+. Thus, we postulate reactions between XeO+, XeOH+ and precision on the 34S/32S ratio, measured on the synthetic and H2 as shown in Table 2.However, similar eVects may be S standard solution, was achieved with a HCl matrix, although caused by reactions between Xe and contaminant H2O or O2 the eVects were not very significant. Slightly higher S+ count in the H2 gas. rates in the HCl matrix were found to give marginally improved A disadvantage of adding H2 is increased production of detection limits. 34S/32S ratio precision was improved further SH+ at m/z 33 and 34, which could be detrimental to accurate by increasing the HCl concentration from 2 to 5% v/v. The isotope ratio measurements. 33S/32S and 33S/34S ratios are HCl matrix was also considered suitable for standards and larger than expected assuming that the single element S samples as it matched the matrix of the cation-stripped sea- standard (Johnson-Matthey) has a close to natural average water isotopic standard.The production of NO+ polyatomic isotopic abundance. The contribution of 33S1H on 34S is ions was subdued in HCl with respect to HNO3, due to a diYcult to assess as ratio accuracy was aVected by many lowering of N loading in the core of the plasma, leading to sources of instrumental mass bias (discussed below) but slightly reduced background count rates at m/z=32. appeared to occur at the 3–5% level. Assuming that the rate Residual background signals during washout periods, after of production of 33S1H is constant and only a minor compoprolonged aspiration of S-bearing solutions, were no poorer nent of the response at m/z=34, it should be adequately than routinely experienced for many elements by ICP-MS.corrected for against the reference standards used for cali- Washout times of 2–3 min proved suYcient to return to bration. The ease of production of SH+ is surprising as no background count rates. reaction was recorded in experimental studies of ion–molecule Calibration lines for 32S and 34S are shown in Fig. 5. A high reactions.33 However, it is possible that a reaction was facilidegree of linearity was observed when using Xe in the hexapole tated under the non-thermal conditions encountered in this cell. The lower limit of detection (3s) for S when using H2 in study or that S+ reacted with H2O impurities in the H2. the gas mixture was of the order of 20–50 mg l-1 and with Other reaction products were observed including complexes only a He and Xe mixture was 100–300 mg l-1.Sulfur concen- with Cl+ made possible by the use of HCl as a sample trations of the order of 50 mg l-1 were measured routinely to introduction matrix. Most notable were XeCl+ and ClH2+. an internal precision (repeatability of ten measurements) of The production of these new polyatomic ions appeared to be better than 2% RSD using either 34S or 32S for calibration. of little significance for the modification of background count Reproducibility of S concentration measurements performed rates for the measurement of S isotopes.on diVerent occasions with diVerent tuning conditions was The majority of the reaction products were not transmitted noticeably poorer (6–15% RSD). into the mass analyser due to a potential barrier applied between the hexapole and the quadrupole. The potential Precision of S isotope ratio measurements barrier could be adjusted manually to obtain the most favourable transmission of S+ analyte ions and reaction products.Good precision is a critical parameter for isotope ratio data In this manner, ion–molecule reactions were used as a tool to to be usefully applied in environmental or geological studies. In quadrupole ICP-MS, sequential measurements are suscep- filter out interfering molecular ions. J. Anal. At. Spectrom., 1999, 14, 1067–1074 1071variations, (iv) incorrect blank subtraction, (v) concentrationdependent mass discrimination, (vi) detection system nonlinearity, and (vi) short-term variation in magnitude of background polyatomic ion transmission.Some of these eVects are discussed in detail below. Three of the most critical tuning parameters that aVect mass bias were adjusted whilst other instrumental parameters were set at their optimum settings (Fig. 6). Nebuliser flow rate aVects the region of the plasma sampled by the ion extraction orifice. A higher flow rate changes the temperature of the portion of sampled plasma and consequently changes the ionization eYciency of S+ and the rate of production of O2+. 34S/32S ratios can vary by as much as 25% through small changes in plasma conditions. Lens potential was generally found to have a relatively minor eVect on mass bias unless set below -600 V. At such a setting, sensitivity was considerably poorer and 34S response fell below the detection limits of the technique, thus modifying the 34S/32S ratio. Therefore, it was critical to ensure that 34S remained significantly above the background count rates whilst 32S did not saturate the detection system.The type and amount of gases introduced into the hexapole Fig. 5 Calibration lines for 32S and 34S. cell had the greatest eVect on mass bias. As discussed above, the addition of H2 can significantly aVect the transmission of tible to instabilities of ionization in the plasma, ion extraction background polyatomic species at both m/z=32 and m/z=34.and operation of the quadrupole.34 For isotope ratios involving In addition, SH+ may increase the signal at m/z=34 relative isotopes of contrasting abundance, a precision of between 0.2 to m/z=32. The addition of H2, whilst further reducing and 1.0% RSD is typical.31 For the S isotope ratios in this background count rates at m/z=32, was detrimental to 34S/32S study it was necessary to achieve the lowest possible RSD, at accuracy. Measured 34S/32S approached the isotopic composithe lower end of this range, in order to be able to resolve tion of the standard most closely with no H2 in the hexapole diVerences in the water samples to be tested.With careful cell. Large deviations in 34S/32S by as much as 40% were optimization of instrument parameters23 it was possible to observed as H2 was added. It was necessary to ensure prior to obtain isotope ratio precision approaching values predicted by each experiment that no added impurities were present in the counting statistics.Xe gas supply due to leaks or insuYcient evacuation of gas Single ion measuring or peak hopping mode was used as lines when changing bottles. Owing to the high cost of Xe gas, more time was proportionally spent in the accumulation of the Xe bottle was isolated from the ICP-MS instrument when counts on the isotopes of interest. 32S and 34S were the only not in use. Each time the gas lines were used it was necessary isotopes measured in order to maximise counting time.Mass to flush for several minutes to remove residual atmospheric scale shift resulting from instabilities in the quadrupole was not found to be a major problem with the instrument used in this study, but calibration across the mass range was checked on a regular basis. No significant gain in short-term precision was observed with the sheathing bonnet, suggesting acoustic noise and pump pulsation peaks to be a minor source of noise.34 Laboratory temperature stability and cleanliness was high (Class 10 000 clean room).The dwell time selected for each isotope had a major eVect on the short-term precision that was obtainable. Instrumental noise frequencies were observed at diVerent dwell time settings and these were optimized at 30 mS per peak after all other sources of noise (e.g. peristaltic pump noise) had been reduced as far as possible. In this study, 34S/32S short-term repeatability was generally excellent, with RSDs often approaching values predicted by counting statistics (Table 3).An important factor in the routinely high degrees of precision could have been the detection system, as it has been shown in previous studies that analogue ‘Daly’-type detectors can give improved isotope ratio repeatability at high count rates.35 Better results were also obtained after the instrument had been conditioned by continuous aspiration of the acid matrix for several minutes prior to analysis. Accuracy of S isotope ratio and S concentration measurements Accurate isotope analysis may be aVected by incorrect calibration, instrumental mass bias or matrix-induced mass bias.Sources of error are often cumulative and include: (i) short-term drift in instrument response, (ii) poor characteriz- Fig. 6 EVect of tuning parameters and analyte concentration on measured 34S/32S isotope ratio. ation of reference standards, (iii) matrix-dependent response 1072 J. Anal. At. Spectrom., 1999, 14, 1067–1074with ultrapure water as discussed above.Thus, we have eliminated the eVects of sample handling as far as possible. Accuracy and precision of data collected for these real samples (which have a complex matrix) more closely represent the data quality that can be expected in analysis of routine samples than if reference standards of a simple matrix had been chosen. The samples were analyzed repeatedly, on diVerent occasions, with some subtle changes to tuning conditions during routine optimization for each experiment.This approach gives an indication of the reliability and reproducibility of the technique over time. d34S variations and S concentrations for crater-lake and Fig. 7 Linear dynamic range of analogue ‘Daly-type’ scintillation detector. spring and river waters from Patuha volcano (West Java) and Kawah Ijen volcano (East Java) are shown in Table 4 and Fig. 9. Under optimum conditions (S=10–50 mg l-1), the gases and to prevent introduction into the hexapole cell. 34S/32S measurement precision or repeatability for samples Fluctuating levels of H2 in the hexapole cell clearly have a during a 10 min scan was <0.3% RSD, close to data achieved large eVect on isotope ratio accuracy and must be avoided for matrix-free standard solutions. External reproducibility where possible. (1s) of mean data for the water samples from day-to-day for Concentration-dependent mass bias is evident from Fig. 6. d34S was typically±2–3‰.d34S for BRT1/7 in experiment 5 34S/32S ratios increased significantly as the concentration of (Table 4) was outside this range and the reason for this the S standard solution was increased. A 1–2% relative increase spurious result is not clear. Fig. 9 shows within-run repeat- in transmission eYciency was observed for 34S over 32S from ability, within-day reproducibility and inter-experiment 10 to 50 mg l-1. The measured 34S/32S ratio deviated from the reproducibility.natural isotopic composition with increasing concentration. It The data demonstrate that d34S can be measured at best to is diYcult to explain this eVect as related to ion-beam space a precision of 2‰ in real samples, approaching what can be charge eVects, which would be expected to have a reverse expected from counting statistics. However, some samples can eVect on relative transmission with increasing concentration.6 be problematic, as demonstrated by sample ALN1/5, which The eVect necessitates careful concentration matching of was measured with an internal precision of only 6‰.Accuracy standards and samples. was variable and results agree to between 1 and 6‰ of Experiments were performed to establish the linear dynamic previously published data. Mean results for five experiments, range of the detector using single element Tm solutions in conducted on separate occasions, smooth out the noise, and 2% v/v HNO3 (Fig. 7). The analogue detection system appears mean data for samples BT1/5 and ALN 1/5 agree to within to be not as prone as pulse counting systems to counting loss 1‰ of published results. However, the result for sample at high count rates.However, some counting loss was observed BRT1/7 diVers significantly from the gas source MS result. above 1×108 counts s-1, most probably related to the specifi- This may be due to a calibration error caused by a matrix cations of the counting electronics, and all data collected at component in BRT1/7, contamination or isotopic modification count rates above this value were discarded.Dead time correcin one of the two aliquots used by the two techniques, or an tions were not applied to the isotope ratio data collected in error during gas source MS. The result for sample IJM1/1 this study. agrees well with published data, although the sample analyzed Mass bias errors were corrected using the in-house standard in this study was not from the same aliquot analyzed in the of matrix-stripped North Atlantic sea-water (d34S=20.5‰).published work.5 The reproducibility of standards was acceptable but sometimes d34S variations between samples from Patuha volcano can erratic, especially after the introduction of samples with an be related to the dynamics of the local drainage system. The apparently heavier matrix (e.g. some spring waters) and was springs at Barutunggul and Alun-alun are the main source for a major limitation to accurate analysis (Fig. 8). However, the crater-lake waters in local streams and rivers. Diluted acid mean results from this study compare favorably against published data measured by gas source MS. Results for crater-lake and spring waters Sulfur isotope ratios were determined in some natural volcanic crater-lake and spring water samples using the methods outlined above. The chosen samples have all been previously measured for d34S by gas source MS and were re-analyzed to test the accuracy of the proposed ICP-MS technique.No sample preparation was necessary other than simple dilution Fig. 9 Results for crater-lake and spring waters from Patuha volcano showing within-run, within-day and external reproducibility and Fig. 8 Typical reproducibility of S isotopic standards. accuracy. J. Anal. At. Spectrom., 1999, 14, 1067–1074 1073Table 4 d34S and S concentration results for volcanic crater-lake and spring waters from Patuha and Kawah Ijen volcanoes, Java, Indonesia d34S (‰) Sample Source Published Mean Expt 1 Expt 2 Expt 3 Expt 4 Expt 5 PT1/5 Crater-lake acid water 19.0a 18 17 21 19 15 19 ALN1/5 Spring water 1.6a 2.1 0.5 5.8 2.8 0.6 1.0 BRT1/7 Spring water 15.5a 12 10 12 11 10 17 IJM1/1 Crater-lake acid water 22.5b 21 21 S/mg l-1 Sample Source Published Mean Expt 1 Expt 2 Expt 3 Expt 4 Expt 5 PT1/5 Crater-lake acid water 4192a 3489 3802 4074 3566 3896 2108 ALN1/5 Spring water 239a 198 180 178 164 193 276 BRT1/7 Spring water 210a 182 166 170 158 158 260 aData from Sriwana et al.4 bData from Delmelle et al.21 6 D.C.Gre� goire, B.M. Acheson and R. P. Taylor, J. Anal. At. waters from the crater-lake transport many toxic elements and Spectrom., 1996, 11, 765. deposit them further downstream into the local environment. 7 H. Naka and D. C. Gre�goire, J. Anal. At. Spectrom., 1996, 11, 359. The S isotope variations measured in this study show that the 8 T. Prohaska, C. Latkoczy and G. Stingeder, European Winter crater-lake waters have been modified during percolation Conference on Plasma Spectrochemistry, Pau, France, January 10–15, 1999.through the caldera walls of the volcano. Water–rock inter- 9 D. J. Douglas and J. B. French, J. Am. Soc. Mass Spectrom., 1992, action or the precipitation of sulfate phases has led to S 3, 398. isotopic fractionation. This percolation process has impli- 10 V.I. Baranov and S.D. Tanner, European Winter Conference on cations for transport processes of heavy elements and can Plasma Spectrochemistry, Pau, France, January 10–15, 1999.provide constraints on dilution eVects with meteoric water. 11 E. R. Denoyer, S. D. Tanner and U. Voellkopf, Spectroscopy, 1999, 14, 2. 12 J. T. Rowan and R. S. Houk, Appl. Spectrosc., 1989, 43, 976. Conclusions 13 G. C. Eiden, C. J. Barinaga and D. W. Koppenaal, J. Anal. At. Spectrom., 1996, 11, 317. Interfering O2+ background ions at m/z=32 m/z=34 can 14 P. Turner, T. Merren, J.Speakman and C. Haines, in Plasma Source Mass Spectrometry. Developments and Applications, ed. be dramatically reduced through ion–molecule reactions with G. Holland and S. D. Tanner, Royal Society of Chemistry, Xe gas in a hexapole ion focusing device, confirming the Cambridge, Spec. Publ., 1997, vol. 202, p. 28. previous work of Rowan and Houk.12 The addition of H2 gas 15 N. Jakubowski, I. Feldmann and D. Stuewer, European Winter further promotes the reduction of O- and N-based background Conference on Plasma Spectrochemistry, Pau, France, January polyatomic interferences. However, H2 addition leads to 10–15, 1999. 16 J. Batey and S. Nelms, European Winter Conference on Plasma greater degrees of instrumental mass bias for measured 34S/32S Spectrochemistry, Pau, France, January 10–15, 1999. ratios. The transmission of S+ analyte ions was simultaneously 17 R. S. Houk, European Winter Conference on Plasma reduced, but by a factor of 10 less than the reduction of O2+.Spectrochemistry, Pau, France, January 10–15, 1999. Under these conditions, the transmission of S+ is suYcient 18 S. D. Tanner and V. I. Baranov, European Winter Conference on Plasma Spectrochemistry, Pau, France, January 10–15, 1999. for accurate and precise concentration isotope ratio measure- 19 U. Voellkopf, V. I. Baranov and S. Tanner, European Winter ments in the 10–50 mg l-1 range. Conference on Plasma Spectrochemistry, Pau, France, January Sulfur isotope ratios, determined for Indonesian crater-lake 10–15, 1999. and spring water samples, showed acceptable degrees of 20 J.Speakman, P. J. Turner, A. N. Eaton, F. Abou-Shakra, internal precision (<0.3% RSD measured on ten repeats of Z. Palacz and R. C. Haines, European Winter Conference on Plasma Spectrochemistry, Pau, France, January 10–15, 1999. 34S/32S ratio) and accuracy (34S within 1–6‰ of published 21 P. Delmelle, A. Bernard, M. Kusakabe, T. Fischer and B. Takano, data). Accuracy was limited by short- to medium-term repro- J. Volcanol. Geotherm. Res., 1999, in the press. ducibility of standards and samples and some sample matrix 22 K. Bruland, Chem. Oceanogr., 1983, 8, 157. eVects. ICP-MS incorporating hexapole ion optics is a 23 I. S. Begley and B. L. Sharp, J. Anal. At. Spectrom., 1997, 12, 395. 24 I. Szabo, Int. J. Mass Spectrom. Ion Processes, 1986, 73, 197. potentially useful survey tool for d34S determinations in 25 D. Gerlich, Adv. Chem. Phys., 1992, LXXXII, 1. environmental and geological samples. 26 J. J. Thomson, Rays of Positive Electricity, Longmans, London, 1933. Norbert Jakubowski is thanked for a preliminary review. We 27 P. F. Knewstubb, Mass Spectrometry and Ion-Molecule Reactions, are grateful to Jeroen Kraan and Gijs Nobbe for assistance Cambridge University Press, Cambridge, 1969. during laboratory work. The Utrecht ICP-MS laboratory is 28 E. W. McDaniel, V. Cerma�k, A. Dalgarno, E. E. Ferguson and jointly funded by the Netherlands Organization of Scientific L. Friedman, Ion-Molecule Reactions,Wiley, New York, 1970. Research (NWO/GOA) and the University of Utrecht. 29 V. G. Anicich and W. T. Huntress, Astrophys. J. Suppl. Ser., 1986, 62, 553. 30 V. G. Anicich, J. Phys. Chem. Ref. Data, 1993, 22, 1469. 31 K. E. Jarvis, A. L. Gray and R. S. Houk, A Handbook of References Inductively Coupled Plasma Mass Spectrometry, Blackie, 1 Y. Kiyosu and M. Kurahashi, Geochim. Cosmochim. Acta, 1983, Glasgow, 1992. 47, 1237. 32 E. E. Ferguson, D. Smith and N. G. Adams, Int. J. Mass 2 S. N. Williams, N. C. Sturchio, V. M. L. Calvache, F. R. Mendez, Spectrom. Ion Processes, 1984, 57, 243. C. A. Londozo and P. N. Garcia, J. Volcanol. Geotherm. Res., 33 D. Smith, N. G. Adams and W. Lindinger, J. Chem. Phys., 1981, 75, 3365. 1990, 42, 53. 34 A. L. Gray, J. G. Williams, A. T. Ince and M. Liezers, J. Anal. At. 3 B.W. Robinson and S. H. Bottrell, Appl. Geochem., 1997, 12, 305. Spectrom., 1994, 9, 1179. 4 T. Sriwana, M. J. Van Bergen, J. C. Varekamp, S. Sumarti, 35 L-Q. Huang, S-H. Jiang and R. S. Houk, Anal. Chem., 1987, B. Takano, B. J. H. Van Os and M. J. Leng, J. Volcanol. Geotherm. 59, 2316. Res., 1999, in the press. 5 P. Delmelle and A. Bernard, J. Volcanol. Geotherm. Res., 1999, in the press. Paper 9/02037C 1074 J. Anal. At. Spectrom., 1999, 14, 1067–10