Characterisation and Correction of Instrumental Bias in Inductively Coupled Plasma Quadrupole Mass Spectrometry for Accurate Measurement of Lead Isotope Ratios IAN S. BEGLEY† AND BARRY L. SHARP Department of Chemistry, L oughborough University, L oughborough, L eicestershire, L E11 3TU, UK The accuracy and external precision of isotope ratio was found to impose limitations upon the accuracy and precision of measured isotope ratios. Mass dependent effects measurements can be limited by any of a number of instrumental bias factors which include mass bias, pulse pile- are well known in quadrupole based ICP-MS, the most common of these being mass bias.5–7 In this study, those forms up, background effects and mass scale shift.Measured isotope ratios are generally biased from their true value by the of instrumental instability and bias which limit the accuracy and external precision of isotope ratio measurements made by interaction of several of these instrumental bias factors.Attainment of accurate isotope ratio measurement requires quadrupole based ICP-MS are characterised. Methods for minimising the influences of the various instrumental bias careful consideration of all possible causes of instrumental bias and adoption of methods for their elimination or correction. factors have been applied to Pb isotope ratio measurement to yield high accuracy and precision. The various forms of instrumental bias observed in measurement of the Pb isotope ratios for NIST SRM 981 Natural Lead ( Isotopic) were minimised to reveal EXPERIMENTAL 204Pb5206Pb=0.0592±0.0002, 207Pb5206Pb=0.9148±0.0007, Instrumentation 208Pb5206Pb=2.1709±0.0015 (n=5) compared with certified values of 0.059042±0.000037, 0.91464±0.00033 and The ICP used employed a 27 MHz crystal-controlled supply 2.1681±0.0008, respectively.The thallium mass bias (Plasma-Therm, Kresson, NJ, USA, Model 2500F) with an concentration procedure was employed in obtaining this data. automatic impedance-matching network.The ICP was oper- All uncertainties given as 2s. ated at 1300W and the plasma, auxiliary and nebuliser gas flow rates were 12.0, 0.0 and 0.75 l min-1, respectively. A glass Keywords: Inductively coupled plasma mass spectrometry; concentric nebuliser (J.E. Meinhard, Santa Cruz, CA, USA, isotope ratio measurement; mass scale shift; mass bias; lead Model TR-30-CZ), peristaltic pump (Gilson, Villiers Le Bel, France, Model Minipuls 2) and water-cooled single pass spray The use of quadrupole mass analysers in ICP-MS fundamen- chamber, of the Surrey design,8 were used for sample introduc- tally limits the accuracy and precision to which isotope ratio tion.The plasma source was centred about the sampling orifice, measurements may be made. Quadrupole mass analysers are which was located approximately 12 mm from the load coil. specifically designed for, and hence perform optimally in, the The quadrupole mass analyser was a VG Micromass scanning of relatively large mass ranges.As such they are (Altrincham, Cheshire, UK) Model 12-12S, fitted within the inherently less stable than magnetic sector mass analysers and vacuum system of the ICP-MS instrument described by Date less well suited to highly precise and accurate isotope ratio and Hutchinson.9 The efficiency of ion transmission through measurement. Quadrupole mass analysers are none the less this vacuum system is lower than that given by commercially capable of providing sufficient accuracy to be useful in many manufactured ICP-MS instruments.The ICP-MS instrument isotope ratio studies. The precision of isotope ratio measure- used had a sensitivity of 2×106 counts s-1 per mg l-1. The ments made by quadrupole ICP-MS has at best approached original cones and lens stack were replaced with those of 0.1% RSD,1 and then, only for isotopes of similar abundance the type utilised in VG PlasmaQuad II instruments (VG (e.g. 107Ag5109Ag). For isotope ratios involving an isotope of Elemental, Winsford, Cheshire, UK). The detector was a low abundance, such as 204Pb, a precision of 0.2 to 1.0% RSD continuous dynode channel electron multiplier (Galileo, is typical.2 The accuracy and precision of 207Pb5206Pb ratio Sturbridge, MA, USA, Model 4870 V), positioned at right measurements for NIST SRM 981 Natural Lead (Isotopic), angles to the instrument axis. The mass resolution was set at made by ICP-MS, have generally been 0.2 to 0.3%.Accuracies 0.9 u, defined as the peak width at 10% of the peak height. for isotope ratio measurement of 0.25% are attainable for isotope ratios in the range 0.2 to 5,3 although 0.5% is more Reagents typical. Stock solutions were 10000 mg ml-1 Specpure ICP solutions In a previous study, we investigated methods for the effective (Johnson Matthey, Royston, Herts, UK). High purity water use of data acquisition parameters to gain enhancement of the used for dilution of stock solutions was obtained by passing precision of isotope ratio measurements made by ICP-MS.4 A demineralised water through a laboratory-reagent grade water measurement precision of 0.05% RSD was realised, using system (Liquipure, Bicester, Oxfordshire, UK), operated at optimised data acquisition parameters, for the 107Ag5109Ag 18MV.All working solutions were prepared so as to contain isotope ratio. However, even under optimised conditions, 1% (v/v) nitric acid of ultrapure reagent grade.instability of the mass spectrometer or its associated ion optics Reference material NIST SRM 981 was obtained from Promochem (Welwyn Garden City, Herts, UK). Working † Present address: Department of Chemistry, Scottish Crop Research Institute, Invergowrie, Dundee, DD4 6RA, UK. solutions of NIST SRM 981 were prepared with the aid of Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 (395–402) 395polypropylene beakers and volumetric flasks (Analytical pump noise.‘Detector’ refers to internal noise (Johnson and shot noise) arising within the measurement circuits used in Supplies, Little Easton, Derbyshire, UK). All solutions were stored in polypropylene bottles (Analytical Supplies). detection and counting of ions. Designated as ‘Pick-Up’ are external and interference noise components, such as those associated with the ac line frequency and its harmonics. RESULTS AND DISCUSSION Methods for the identification and subsequent reduction of these noise types have been described previously.4 Sources of Instrumental Noise and Offset The impacts of ‘Background’, ‘Mass Bias’, ‘Pulse Pile-Up’ T ransfer function and ‘Mass Scale Shift’ combine to give an offset, which is equivalent to the bias observed in the measured ratio, prior to The measurement of an isotope ratio is depicted as a hypotheticorrection.Each of these offsets has a number of contributing cal transfer function in Fig. 1. The true ratio is envisaged as factors, for instance, ‘Spectral Interference’ and ‘Analyte being altered by both noise and offset (each being the overall Contamination’ are given as inputs to ‘Background’. That is sum of a number of contributing factors) to produce the to say, the background is determined by the sum of analyte measured ratio. contamination and spectral interference from molecular species The four sources of noise, given in Fig. 1, are all likely to and ions other than those of the analyte element.To help contribute to the imprecision of isotope ratio measurement. simplify the hypothetical transfer function, in this context, ‘Poisson’ refers to the uncertainty associated with the rate of analyte contamination relates to sample-to-sample memory arrival of ions at the detector, often referred to as the counting effects and/or contamination occurring during sample prep- statistic. ‘Ion Source’ relates to the various forms of internal aration, evident in the procedural blank, which if present may noise arising during sample introduction and plasma processes, cause the measured isotope signature to differ from that of the including audible-frequency noise associated with gas dynamic phenomena occurring in the plasma discharge and peristaltic sample material.A number of correction methods for minimis- Fig. 1 Hypothetical transfer function for isotope ratio measurement by inductively coupled plasma quadrupole mass spectrometry. 396 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12ation of offset are indicated in Fig. 1. Correction of mass bias, pulse pile-up and mass scale shift can be made post-acquisition, if data acquisition is undertaken by an appropriate method. Noise Instabilities associated with the ICP ion source may arise from either a change in energy transfer from the plasma to the sample, or variation in the efficiency of the nebulisation and transportation of sample.10 As these should influence both isotopes coherently, the direct influence of instability arising from the ICP ion source upon the isotope ratio, should be rendered insignificant by use of appropriate data acquisition parameters.4 Mermet and Ivaldi11 have shown that ratioing of simultaneously measured emission line intensities, obtained by ICP-AES, is effective in cancellation of flicker noise in cases for which a correlation coefficient of >0.95 exists between the line intensities. The data acquisition parameters previously found to be optimum for measurement of a single isotope ratio (dwell time=10.24 ms; points per peak=3; settle time=2 ms; number of sweeps=1600)4 were used for determination of the 207Pb5206Pb isotope ratio over a period of 4 h.By plotting the Fig. 2 Influence of isotope mass upon observed mass shift for cobalt, indium, cerium, lead and uranium, each at a concentration of signal intensity data obtained for the 206Pb and 207Pb isotope 50 ng ml-1.intensities against one another, a linear correlation coefficient of 0.9995 was obtained. This high degree of correlation indicates that noise arising from instability of the nebuliser and/or the ICP may be cancelled when measuring the 207Pb5206Pb isotope ratio. The contribution made by ion source to noise, as depicted in Fig. 1, can hence be assumed to be negligible for isotope ratio measurements made using appropriate data acquisition parameters.Mass scale shift The stability of the mass scale calibration (the relationship between measured mass, represented by a 32 bit channel address, and actual mass) is generally limited by the constancy of the dc and rf potentials applied to the quadrupole rod pairs.12 A change in the order of 0.05% in the amplitude of the rf potential, for a m/z of 100, would cause a mass shift of approximately 0.05 u. The daily shift in mass scale calibration for quadrupole mass analysers employed in ICP-MS is typically±0.10 u, however, ambient temperature change could easily cause additional shift.13 As the resolution of the quadrupole mass analyser is inversely proportional to the ratio of the Fig. 3 Mass spectrum for Pb obtained in the scanning mode across 512 data channels. Spectrum has been smoothed using Fourier spectral dc to rf potential, any shift in mass scale calibration caused by analysis with Blackman windowing to remove random noise.change in one, but not both, of the applied potentials would also cause the resolution to change. A multi-element solution comprising Be, Mg, Co, In, Ce, Pb Fourier smoothed mass spectrum for Pb shown in Fig. 3 was utilised to simulate the effect of mass shift. The 206Pb and and U, each at a concentration of 50 ng ml-1, was used to examine the mass scale calibration, at regular intervals over a 208Pb isotope peaks were removed from the mass spectrum and cross correlated to identify the central channel addresses period of 5.5 h.Mass scans were undertaken across the mass range 4.7 to 240.3 u, utilising 4096 data acquisition channels, for which the degree of overlap between isotope peaks was maximum. Both isotopes were integrated across a width of to give a step size of approximately 0.058 u. Such a low mass resolution was insufficient to allow direct monitoring of mass 0.4 u about these central channel addresses to obtain the value of the 208Pb5206Pb isotope ratio in the absence of mass shift.scale shift. Regression analysis was performed upon each mass scan to provide a linear relationship between channel number By moving the 0.4 u wide integration windows across both isotope peaks in equal increments of 0.0134 u [1 digital-to- and actual mass. The linear equations produced for each mass scan allowed calculation of the observed mass for each isotope analogue converter (DAC) step], the apex of the peaks moved away from the centre of the integration window and drift in and, hence, the mass shift relative to the initial mass scan.Fig. 2 shows the mass shift observed for the 59Co, 115In, 140Ce, mass location was simulated. From the data presented in Fig. 4, it may be seen that a mass shift of 0.1 u caused the 208Pb and 238U isotopes over the period of the experiment. It may be seen that the magnitude of the mass shift increased observed isotope ratio to change by approximately 0.5%.Movement of the integration window to higher masses with isotope mass. To a first approximation, doubling of the isotope mass caused the severity of the mass shift to increase caused a rise in the observed isotope ratio while the reverse was true for movement to lower masses (Fig. 4). Thus, it may two fold, as may be seen by comparison of the 59Co and 140Ce or the 115In and 238U isotopes in Fig. 2. be concluded that the observed change in the 206Pb5208Pb isotope ratio is a consequence of a quantisation error in To assess the influence of mass scale shift upon isotope ratio measurement, while retaining a constant mass resolution, the selection of mass by the multi-channel analyser.In this instance, Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 397Fig. 5 Relative inaccuracy in isotope ratio A5B arising from incorrect Fig. 4 Influence of mass shift upon the 206Pb and 208Pb isotope selection of the dead time used in dead time correction.A dead time intensities and the 208Pb5206Pb isotope ratio obtained by simulation of 20 ns has been applied whereas a 25 ns dead time provides accurate of mass shift based upon movement of the integration windows across correction. the isotope peaks. Background the quantisation error results in the nominal centre of the Initially, it is necessary to establish whether background makes 206Pb isotope peak lying slightly to the left (lower mass) of the a significant contribution to the offset of the measured isotope centre of the initial integration window, relative to that of the ratio from its true value.This may be done by scanning the 208Pb isotope peak. mass region of interest for a procedural blank. In the absence of background, be it from spectral interference, analyte contamination or a combination of both, the isotope ratio(s) Dead time should be normally distributed about unity. If the isotope At count rates of above approximately 0.4 MHz, continuous ratios for the procedural blank deviate significantly from unity, dynode electron multipliers, such as the Galileo 4870V used then background can be assumed to have a measurable impact herein, count fewer events than actually occur.The interval upon the offset (high background count rates, for example during which the electron multiplier is ‘hung-up’ is termed the across a number of masses, yielding a ratio of unity, would dead time. In addition, the counting logic circuitry also exhibits still affect the measured ratio, but this situation is normally a finite response time, sometimes referred to as sag,14 which indicative of instrumental problems).From the values of the may result from pulse broadening in the amplifier, the limited isotope ratios obtained, it should be possible to identify rise and fall times of the discriminator, or the maximum clock whether the background arises from analyte contamination rate of the counter.Most pulse counting systems show dead and/or is due to spectral interference. times of 10 to 30 ns. If the value(s) of the isotope ratio(s) of interest for the If the count rate is <0.4 MHz then the probability Pn of n procedural blank are similar to those expected for the analyte, ion current pulses occurring during the dead time D is governed then analyte contamination is liable to add to the offset. It by Poisson statistics. If the count has a Poisson distribution, may be possible to take action to identify and reduce analyte the measured count rate N may be corrected for dead time D contamination arising from sample preparation or within the utilising the equation: sample introduction system.Otherwise, blank subtraction of all subsequent data may be necessary, however, this may result N¾= N (1-ND) (1) in deterioration of the precision of isotope ratios. If the value(s) of the isotope ratio(s) of interest for the procedural blank are indicative of isobaric interference, which cannot be removed where N¾ is the estimated count rate.by pre-treatment of samples, it may be necessary to apply It is necessary to precisely determine the dead time D to background correction. Generally, an additional isotope which minimise the dependence of isotope ratios upon analyte conprovides a measure of the level of isobaric interference on the centration and isotopic abundance. In Fig. 5, the dead time isotope(s) of interest will be included in the data acquisition has been under-estimated by 5 ns, a dead time correction of procedure.Arithmetic calculation can then be used to deter- 20 ns has been applied, whereas a 25 ns dead time provides mine the proportion of the total ion count arising from the accurate correction. It may be seen, from Fig. 5, that isotope isobaric interference. ratios can become biased by several parts per thousand as a consequence of inaccurate selection of the dead time utilised in dead time correction.Since the gradients of the lines shown Mass bias in Fig. 5 increase as the ratio of A to B increases, it may be assumed that the further the measured isotope ratio lies from Mass bias derives from the differential transmission of ions of different mass from the point at which they enter the sampling unity the more susceptible the ratio is to inaccurate correction for dead time. Hence, accurate measurement of the 204Pb5206Pb device until they are finally detected by the electron multiplier.The majority of mass bias effects occur in the interface region isotope ratio in natural abundance materials is more difficult to achieve than is accurate 207Pb5206Pb isotope ratio measure- of the ICP-MS, but some additional bias may occur in the quadrupole mass filter, unlike in sector mass analysers where ment, by pulse counting. 398 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12transmission in the magnetic and electrostatic sectors is largely likely to show similar mass bias effects to those for Pb. Variation in the 194Pt5196Pt and 194Pt5195Pt isotope ratios free of bias.Mass bias may be measured and hence corrected by use of an external standard of known isotopic composition observed during 80 acquisitions, each of a duration of 2 min, is shown in Fig. 6. The change in the 194Pt5196Pt isotope ratio or a fixed, constant ratio within the sample. External mass bias correction involving use of an external standard has been is approximately double that observed in the 194Pt5195Pt isotope ratio, over the same period.If a linear relationship widely used in ICP-MS. Estimation of mass bias by use of an external standard is advantageous as: (i) mass bias can be were to exist between mass bias and mass shift the variation in the 194Pt5196Pt isotope ratio would be anticipated as being measured at the correct masses (position upon the mass response curve is equivalent for standard and sample) and twice that observed in the 194Pt5195Pt isotope ratio.The best fitting linear relationship for the data given in Fig. 6 has a (ii) mass bias can be measured at the relative isotope abundances to be found in samples (pulse pile-up effects are gradient of 1.82 (r2=0.67). The similarity between the predicted (2) and observed (1.82) gradients suggests that linear correction generally equivalent for standard and sample, although this does not apply for isotope dilution).for mass bias is at least as valid as either of the other methods for correction of mass bias. Internal mass bias correction using Measurement of mass bias and analysis of samples are however displaced in time and are therefore susceptible to eqn. 2 resulted in improvement in accuracy, but reduction in the precision of the Pt isotope ratios as a consequence of shifts variation in the magnitude of the mass bias. Regular monitoring and updating of the mass bias correction is necessary to arising from sources other than mass bias.prevent long term instability. Additionally, differences between the mass response curves for the standard and sample arising Lead Isotope Ratio Measurement from matrix effects are not accounted for by external mass bias correction. Noise reduction Internal mass bias correction is beneficial in that it: (i) pro- The data acquisition parameters selected for high accuracy Pb vides near continual monitoring of change in mass bias and isotope ratio measurement are summarised in Table 1.Data (ii) can correct for non-spectral interferences caused by acquisition was undertaken in the peak jumping mode because matrix effects. a greater proportion of the available acquisition time could be Internal mass bias correction is however based upon mass and isotope abundance which differ from those of the isotope(s) of interest. Additionally, internal mass bias correction generally requires the element of interest to have three or more isotopes and for at least two of these isotopes to have not undergone any form of isotopic fractionation in nature.There are three algorithms which may be applied in correction of mass bias: these are based upon linear, power law and exponential relationships between mass bias and mass difference. 16 If mass bias is assumed to be a linear function of mass difference, to the precision of the data, mass bias correction may be performed using eqn. 2: (A/B)corr=(A/B)obs(1+an) (2) where (A/B)corr is the mass bias corrected ratio of isotopes A and B, (A/B)obs is the observed ratio of isotopes A and B, a is the bias per unit mass and n is the mass difference between isotopes A and B (u). If mass bias is, however, a power law function of mass difference, eqn. 3 may be utilised to correct for mass bias: (A/B)corr=(A/B)obs(1+a)n (3) Finally, should an exponential relationship exist between mass bias and mass difference, mass bias correction may be performed using eqn. 4: Fig. 6 Relationship between the 194Pt5196Pt and 194Pt5195Pt isotope ratios measured over 80 consecutive, 2 min integrations. The line (A/B)corr=(A/B)obs expan (4) representing the best fitting linear relationship has a gradient of 1.82. Taylor et al.16 found the power law and exponential functions to be more successful in correction of mass bias than the linear Table 1 Data acquisition parameters selected for use in Pb isotope function for U isotope ratio measurement, by multiple ratio measurement based upon noise spectral analysis4 collector ICP-MS.Acquisition mode peak jumping Although all three algorithms may yield equivalent mass Quadrupole rest mass 200 u bias factors under normal circumstances it is feasible that Quadrupole settle time 2 ms inaccuracy may arise from use of an inappropriate algorithm Detector dead time 32.5 ns in estimation of mass bias by internal mass bias correction.As Discriminator level 40 internal mass bias correction is most useful in instances in which the magnitude of the mass bias varies with time, it was Number of isotopes 4 Dwell time per channel 10.24 ms proposed that the appropriateness of an algorithm could be Points per peak 3 determined by monitoring variation in the mass bias of two Acquisition time 120 s isotope ratios over time. Replicate measurements 10 Platinum was selected to identify whether mass bias was a Abundance sensitivity* <5 ppm linear function of mass as it supplied two isotope ratios which are fixed and constant in nature and are close to unity.* Overspill for isotopes differing by 1 u using three points per peak positioned at m-0.05, m and m+0.05 u. Platinum is also only 10 u removed from Pb, and as such, is Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 399spent in accumulation of counts on the isotopes of interest, were somewhat higher, averaging 0.12% RSD for 205Tl5203Tl and 0.13% RSD for 207Pb5206Pb (n=5).allowing the theoretical counting statistic to be minimised. A dwell time of 10.24 ms was favoured, for all isotopes, independent of their isotopic abundance, to give a high frequency cut- Mass calibration off of 49 Hz.4 The narrow mass range being studied allowed Lying at the upper end of the mass range, the Pb isotopes are use of a 2 ms settle time, compared to the more commonly susceptible to mass scale shift resulting from instability of the used 10 ms.quadrupole mass analyser (see Fig. 2). As previously men- Three acquisition points per peak were utilised because it tioned, use of three data acquisition points per isotope allowed provided a crude method of identifying mass shift, and hence mass scale shift to be monitored. In instances where mass shift suspect isotopic data. Additionally, three points per peak was observed, re-calibration of the mass scale was undertaken allowed the maximum accumulated count to approximate prior to analysis of the sample which followed.The isotopes 48×106 ion counts, thrice that attainable using a single point utilised in mass calibration were 203Tl, 205Tl, 206Pb, 207Pb, per peak, since the multi-channel analyser limits the maximum and 208Pb. count per channel to 16×106 ion counts. Use of three points The nature of mass scale shift is such that the larger the per peak was also beneficial to the attainment of a sizeable difference in mass of the measured isotopes, the greater the count for 204Pb, which with a natural abundance of 1.4%, influence upon the isotope ratio, as was shown in Fig. 2. For is a primary restriction to the theoretical counting statistic. this reason, the lightest and heaviest isotopes being measured To permit internal mass bias correction, the 203Tl and 205Tl were ratioed to give a monitor ratio for correction of mass isotopes were included in the data acquisition procedure.The scale shift. Any integration which gave a monitor ratio out by number of isotopes to be measured thus increases from four 2s of the mean value (n=10), for integrations within the to six. Seven if the nature of the sample material had necessi- determination were rejected as outliers. This test was under- tated correction for isobaric interference from 204Hg on the taken iteratively, never more than twice, until all remaining minor Pb isotope. Utilising three acquisition points per peak monitor ratios occurred within 2s.Similar procedures for data and a dwell time of 10.24 ms, sequential measurement of six reduction are routinely used in isotope ratio measurements isotopes proved to be ineffective in minimisation of instrumen- made by sector based mass analysers. Statistical outliers were tal noise. With an elapse time of 195 ms, the ratioing frequency found in the monitor ratio (e.g. 203Tl5207Pb), but no outliers of isotopes is inadequate for reduction of low frequency noise, were observed in either the Tl or Pb isotope ratios. as the band-pass extends to frequencies as low as 2.6 Hz.4 The external precision of 207Pb5206Pb isotope ratio measure- To attain a measurement precision approaching counting ment before and after removal of outliers, based upon use of statistic levels for Tl mass bias corrected Pb isotope ratios, it 203Tl5207Pb as a monitor ratio, for five samples of 1000 ng ml-1 was found necessary to measure the Pb isotope ratios on an of NIST SRM 981 are compared in Table 2.Removal of individual basis. Hence, both Tl isotopes and two of the four statistical outliers resulted in improvement in measurement Pb isotopes were measured within any given data acquisition precision, while the mean measured value of the 207Pb5206Pb procedure. For both the 205Tl5203Tl and Pb isotope ratios, isotope ratio remained unchanged, to within the precision of noise reduction extended to a frequency of 5.2 Hz.4 the data.Similar reduction in the external precision of the Between the lower and upper cut-off frequencies given by 204Pb5206Pb and 208Pb5206Pb isotope ratios was obtained by the ratioing of isotopes and the dwell time, respectively, noise removal of statistical outliers based upon use of 203Tl5206Pb reduction occurs as the result of the summation of sweeps.4 and 203Tl5208Pb as monitor ratios, respectively. While isotope peak profiles are added coherently in the multichannel analyser upon summing of sweeps, noise components Dead time correction are reduced by smoothing.All noise frequencies other than those associated with the sweep rate and its harmonics are The instrumental dead time for Pb isotope ratio measurement attenuated by the summation of sweeps. A total of 916 sweeps was determined by calculation of the observed mass bias for were acquired and summed in the multi-channel analyser the 205Tl5203Tl and 208Pb5207Pb isotope ratios.These isotope during the acquisition time. As a result, the width of the ratios were chosen as they have similar values, 205Tl5203Tl= bandpass at the reciprocal of the sweep time (7.64 Hz for a 2.3871 and 208Pb5207Pb=2.3704, for NIST SRM 981. The data sweep time of 0.131 ms) and associated harmonics (15.28 Hz, acquisition parameters, with the exception of the dead time, 22.92 Hz, etc.) was just 0.0072 Hz.4 given in Table 1 were utilised. The observed mass bias factors The precision of isotope ratio measurements is influenced for the 205Tl5203Tl and 208Pb5207Pb isotope ratios were found by noise and instabilities originating in both the ICP and the to converge at a dead time of approximately 32.5 ns.quadrupole mass filter. Noise originating in the ICP is coherent in all isotopes of a given element, therefore, these coherent Table 2 Improvement in external precision given by removal of noise sources can be removed from measured isotope ratios statistical outliers for the 207Pb5206Pb isotope ratio, based upon use by use of suitable data acquisition parameters.If the quantity of 203Tl5207Pb as a monitor ratio, for a solution containing of noise originating in the quadrupole mass filter were negli- 1000 ng ml-1 of Tl and NIST SRM 981 gible, the precision of isotope ratio measurements would approximate the theoretical counting statistic. Noise originat- 207Pb5206Pb ing in the quadrupole mass filter can be minimised by use of Sample Uncorrected Mass scale shift corrected a fixed m/z. Using the data acquisition parameters given in Table 1, the measurement precision for the Tl and Pb isotope 1 0.9155 0.9153 (8)* 2 0.9155 0.9149 (8) ratios, obtained for a fixed mass position upon the apex of the 3 0.9141 0.9143 (9) 206Pb isotope peak (count rate 0.5 MHz), by switching the 4 0.9146 0.9146 (10) quadrupole mass analyser into remote preventing it from 5 0.9141 0.9148 (9) ‘jumping’, averaged 0.08% RSD for 205Tl5203Tl and 0.06% Mean 0.9147 0.9148 RSD for 207Pb5206Pb (n=3).These compared favourably with Standard deviation 0.0007 0.0004 the theoretical counting statistic, which for the first determination had a value of 0.05% RSD. The RSDs of 10 replicate * Number of integrations remaining following removal of statistical outliers. integrations forming a determination, for normal operation 400 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12Mass bias correction Interference effects Lead suffers from only one isobaric interference, 204Pb being The weaknesses of the thallium correction method are: the acquisition time necessary to measure the Tl isotopes, which overlapped by 204Hg.Working solutions of 1000 ng ml-1 of NIST SRM 981 were found to contain Hg in insufficient could otherwise be spent accumulating counts on the Pb isotopes; and the reduction in precision of the Pb isotope quantities to necessitate measurement of 202Hg (29.8%) for correction of 204Hg (6.8%) on 204Pb.The 201Hg5202Hg isotope ratios which can result from correction. It is necessary to be mindful of the need to precisely correct for dead time and ratio has a fixed value of 0.571, which is sufficiently removed from unity to be indicative of Hg contamination. The ion assure that noise is minimised in both Tl and Pb isotopes if the thallium correction method is to be used effectively. It is counts observed on masses 201 and 202 u were found to be comparable at 60 counts s-1 for a solution containing worthy of note that Walder and co-workers19,20 have successfully used the thallium correction method with simultaneous 1000 ng ml-1 of NIST SRM 981, and therefore no correction for the effect of the less abundant 204Hg isotope was considered measurement of the isotopes using Faraday detectors to reproducibly measure 207Pb5206Pb in NIST SRM 981 to within to be necessary.To measure the level of Pb contamination present in the Tl 0.001%.20 Mass bias corrections were performed using the thallium stock solution, the 206Pb and 207Pb isotopes and their ratio were determined for solutions containing 500, 750 and correction method, validated by Ketterer et al.,17 by spiking each of the Pb samples with Tl to a concentration of 1000 ng ml-1 of Tl.The ion count on mass 206 u varied from 59 counts s-1 for 500 ng ml-1 of Tl to 63 counts s-1 for 500 ng ml-1. The 205Tl5203Tl isotope ratio has a fixed and known value of 2.3871.Measured mass bias factors for both 1000 ng ml-1 of Tl. The 207Pb5206Pb isotope ratio had a value of 0.9 at all three Tl concentrations, which is suggestive of the 205Tl5203Tl isotope ratio and the 207Pb5206Pb isotope ratio for 1000 ng ml-1 of NIST SRM 981, having a certified value a small quantity of Pb contamination being present in the solutions. As the level of Pb contamination associated with of 0.91464, may be seen from Table 3 to be of the order of 0.73% u-1.the Tl content of solutions was considered to be below that which would allow precise correction, no corrective action was The Tl correction method assumes mass bias to be a power law function of mass.2,17–20 Since a linear relationship was taken. A solution containing 1% (v/v) nitric acid in high purity water was used as the procedural blank and measured using observed between mass bias and mass for Pt isotope ratio measurement (Fig. 6), the use of both linear and power law the same data acquisition parameters as the samples. Background correction was performed by subtraction of the algorithms in mass bias correction was investigated. Table 4 shows that the values of the 207Pb5206Pb isotope ratio obtained ion counts obtained for the procedural blank from those for subsequent samples. following Tl mass bias correction using either algorithm are equivalent, to within the precision of the data. The power law algorithm (eqn. 4) was used in all subsequent mass bias Accuracy and precision corrections. Isotopic data obtained for five consecutive determinations of a solution containing 1000 ng ml-1 of NIST SRM 981, is given in Table 5. Samples were analysed for 20 min, the analysis Table 3 Measured mass bias factors for the 205Tl5203Tl and comprising ten measurements each of 2 min duration. Each of 207Pb5206Pb isotope ratios, for solution containing 1000 ng ml-1 of Tl the Pb isotope ratios was measured individually using the and NIST SRM 981 optimum data acquisition parameters as detailed in Table 1.The measured isotope ratios for NIST SRM 981 are shown to Bias factor (% u-1) be in good agreement with the certified ratios, following mass Sample 205Tl5203Tl 207Pb5206Pb bias, mass scale shift and dead time corrections. The measurement precision (% RSD) for the Pb isotope ratios, given in 1 0.66 0.75 Table 5, is that obtained for 1s of the mean of the five replicate 2 0.69 0.78 3 0.74 0.69 determinations.The counting statistic, however, represents the 4 0.74 0.74 uncertainty associated with measurement for a single inte- 5 0.77 0.70 gration. An external measurement precision of 0.038% RSD Mean 0.72 0.73 was obtained for the 207Pb5206Pb isotope ratio following RSD* (%) 0.087 0.038 correction of offsets, compared with 0.12% RSD for the uncorrected data. * Calculated for 1 standard deviation. Table 5 Accuracy and precision of Pb isotope ratio measurement for Table 4 Comparison of linear and power law algorithms for correc- NIST SRM 981 tion of mass bias observed in the 207Pb5206Pb isotope ratio by thallium mass bias correction Sample 204Pb5206Pb 207Pb5206Pb 208Pb5206Pb 1 0.0593 0.9153 2.1706 207Pb5206Pb 2 0.0591 0.9150 2.1715 3 0.0592 0.9143 2.1715 Mass bias algorithm 4 0.0593 0.9146 2.1698 5 0.0593 0.9148 2.1712 Sample Uncorrected Linear Power law Mean 0.0592 0.9148 2.1709 1 0.9215 0.9150 0.9155 RSD* (%) 0.12 0.038 0.033 2 0.9218 0.9149 0.9155 Counting statistic (%) 0.15 0.048 0.041 3 0.9209 0.9136 0.9141 4 0.9214 0.9140 0.9146 Certified value 0.059042 0.91464 2.1681 5 0.9211 0.9135 0.9141 2s 0.000037 0.00033 0.0008 Mean 0.9213 0.9142 0.9147 * Calculated for 1 standard deviation for isotope ratios measured Standard deviation 0.0004 0.0007 0.0007 to five decimal places. Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12 401CONCLUSIONS ing this work and the Engineering and Physical Sciences Research Council for provision of a studentship to I.S.B.Application of the methodology outlined for high accuracy isotope ratio measurement to the determination of the 207Pb5206Pb isotope ratio for five samples of NIST SRM 981 revealed 0.9148±0.0004 (1s), compared with a certified value REFERENCES of 0.91464. The bias from the certified value is 0.013%, well 1 Ince, A. T., Williams, J. G., and Gray, A. L., J. Anal. At. Spectrom., within 1s (RSD=0.043%). To the best of our knowledge, the 1993, 8, 899.accuracy and precision of the Pb isotope ratios, given in 2 Ketterer, M. E., J. Anal. At. Spectrom., 1992, 7, 1125. Table 5, are equivalent to or better than those published 3 Fassett, J. D., and Paulsen, P. J., Anal. Chem., 1989, 61, 643A. elsewhere for measurement by quadrupole based ICP-MS. 4 Begley, I. S., and Sharp, B. L., J. Anal. At. Spectrom., 1994, 9, 171. Measurements made using multiple collector, magnetic sector 5 Gillson, G. R., Douglas, D.J., Fulford, J. E., Halligan, K. W., and Tanner, S. D., Anal. Chem., 1988, 60, 1472. based ICP-MS instruments are of a superior accuracy and 6 Vaughan, M. A., and Horlick, G., Spectrochim. Acta, Part B, precision.19,20 Utilising a standard nebuliser and spray chamber 1990, 45, 1301. configuration, the multiple-collector ICP-MS instrument used 7 Ross, B. S., and Hieftje, G. M., Spectrochim. Acta, Part B, 1991, by Walder et al.19 provided a precision of 0.011% RSD and 46, 1263.bias of -0.058% for 207Pb5206Pb. These are not far removed 8 Williams, J. G., in Handbook of Inductively Coupled Plasma Mass from the accuracy and precision given herein. In this study, Spectrometry, ed. Jarvis, K. E., Gray, A. L., and Houk, R. S., Blackie and Son Ltd., Glasgow and London, 1992. each sample analysis used approximately 15 mg of Pb (uptake 9 Date, A. R., and Hutchison, D., J. Anal. At. Spectrom., 1987, 2, 269. of 0.75 ml min-1 of 1000 ng ml-1 Pb for 20 min), compared 10 Carre�, M., Poussel, E., and Mermet, J.-M., J. Anal. At. Spectrom., with 200 ng of Pb for multiple-collector ICP-MS.19 The rela- 1992, 7, 791. tively large amount of Pb required reflects the low count 11 Mermet, J.-M., and Ivaldi, J. C., J. Anal. At. Spectrom., 1993, 8, 795. sensitivity of our ICP-MS instrument, at 2×106 counts s-1 12 Bley, W. G., Vacuum, 1988, 38, 103. per mg l-1 being 10–20 fold below that of modern commer- 13 Batey, J., VG Elemental, Winsford, Cheshire, UK, personal communication. cial systems. 14 Russ, G. P., and Bazan, J. M., Spectrochim. Acta, Part B, 1987, The bias, expressed as a percentage difference from the 42, 49. certified value, for the 208Pb5206Pb and 204Pb5206Pb isotope 15 Vincent, C. H., Random Pulse T rains, T heir Measurement and ratios, given in Table 5, was 0.13 and 0.30% (n=5), respectively. Statistical Properties, Peter Peregrinus, London, 1973. Certified values for NIST SRM 981 are 2.1681 for 208Pb5206Pb 16 Taylor, P. D. P., De Bie`vre, P., Walder, A. J., and Entwistle, A., and 0.059042 for 204Pb5206Pb. As is the case for these measured J. Anal. At. Spectrom., 1995, 10, 395. 17 Ketterer, M. E., Peters, M. J., and Tisdale, P. J., J. Anal. At. Pb isotope ratios, accuracy generally decreases as the true Specom., 1991, 6, 439. value of the isotope ratio deviates from unity and the mass 18 Longerich, H. P., Fryer, B. J., and Strong, D. F., Spectrochim. separation between isotopes increases despite best attempts to Acta, Part B, 1987, 42, 39. remove bias. This study demonstrates the high precision and 19 Walder, A. J., Platzner, I., and Freedman, P. A., J. Anal. At. accuracy which can be achieved for analysis of standard Spectrom., 1993, 8, 19. solutions by careful planning of experiments. Use of the given 20 Walder, A. J., Koller, D., Reed, N. M., Hutton, R. C., and Freedman, P. A., J. Anal. At. Spectrom., 1993, 8, 1037. methodology is likely to provide the best possible isotope data attainable by ICP-MS for any given sample type. Paper 6/05078F Received July 22, 1996 The authors would like to thank the British Geological Survey Accepted November 11, 1996 for loan of the ICP-MS instrument, VG Elemental for support- 402 Journal of Analytical Atomic Spectrometry, April 1997, Vol. 12