Optimisation of a current generation ICP-QMS and benchmarking against MC- 1 ICP-MS spectrometry for the determination of lead isotope ratios in 2 environmental samples

6 The precision and accuracy of lead (Pb) isotope measurements obtained from quadrupole-based mass 7 spectrometers (ICP-QMS) are considered to be limited by a number of factors originating in different 8 components of the instruments. In this study, experimental and instrumental protocols were 9 optimised for determining lead isotope ratios in urban soil digests. Experimental measures included 10 individual dilution of all samples and isotopic standards (SRM-981, NIST) to a single Pb concentration 11 intended to produce an intensity which was high enough to negate blanks and interferences but low 12 enough to ensure the detector operated only in pulse counting mode. Instrumental protocols included 13 batch dead time correction, optimisation of dwell time and the number of scans employed and 14 correction of mass discrimination by sequential application of both internal ( 203 Tl/ 205 Tl ratio) and 15 external (SRM-981, NIST) standards. 16 This optimised methodology was benchmarked against multi-collector mass spectrometer (MC-ICP- 17 MS) measurements of Pb isotope ratios using replicate digest solutions of the same soil; but after 18 these had been subjected to Pb separation using an ion-exchange procedure. On the assumption that 19 MC-ICP-MS measurements are more accurate, small additive and multiplicative differences were 20 observed in only the 4 th decimal place. ANOVA was used to compare the precisions of the two 21 techniques demonstrating equal precisions c. 0.08% for 207 Pb /206 Pb, suggesting a sample heterogeneity 22 limitation. By contrast, for 207 Pb /204 Pb, the worst-case ratio, ICP-QMS had a 10-fold poorer precision, 23 despite negligible interference from 204 Hg, implying an instrumental limitation. The study concludes 24 that ICP-QMS can provide valuable


27
Lead isotopes can be measured by a range of mass spectrometry techniques. Until fairly recently, 28 thermal ionization mass spectrometry (TIMS) was the preferred choice for precise (0.001-0.01% RSD) 29 measurement of Pb isotope abundances (1, 2). However, TIMS involves laborious sample preparation 30 steps, such as separation of the analyte from the matrix into a highly purified form, involving extensive 31 chemical treatment, and stringent optimisation of vaporisation and ionisation conditions of samples, 32 with long analysis times (~45 minutes). These measurements are not only time consuming but also 33 incur cost, rendering TIMS unsuitable when analysing large numbers of samples (3,4). Inductively 34 coupled plasma mass spectrometers (ICP-MS) are more appropriate for routine use with large sample 35 numbers and are operationally simpler than TIMS (5); sample introduction is at atmospheric pressure 36 and ionization of most elements is readily achieved. They are widely used for multi-element and 37 isotopic analysis and are routinely capable of a precision (<0.05% RSD) comparable to that of TIMS 38 for Pb isotopes. The term 'ICP-MS' covers a range of instruments which differ in their mass filter and 39 detector systems. These include; those that are quadrupole-based (ICP-QMS) or high resolution, 40 magnetic sector field-based (HR-ICP-MS) and those which can utilize single-collector (SC), multi-41 collector (MC), array detector (AD) or 'time-of-flight' (TOF) detectors (4, 6). 42 Quadrupole-based inductively coupled plasma -mass spectrometers (ICP-QMS) are the most 43 commonly used ICP-MS instruments principally because they are relatively inexpensive to produce 44 and support, and are fast and simple in terms of sample preparation, handling and operation. As a 45 result, the number and availability of ICP-QMS instruments far outweigh MC-ICP-MS. However, in 46 isotope studies, specifically when high precision ratio measurements are required, MC-ICP-MS is 47 preferred over ICP-QMS and is considered the benchmark for accurate and precise isotope data (7). 48 This is principally due to simultaneous measurement of all relevant isotopes, which effectively cancels 49 out the effects of 'plasma flicker', and the flat-topped peaks produced in MC-ICP-MS, with their 50 inherent resistance to instrument drift. This is in contrast to sequential measurement of isotopes on 51 leptokurtic peaks in ICP-QMS (8-10). The sequential nature of the analysis in any single collector 52 instrument renders the technique susceptible to variations in, for example, sample aerosol transport, 53 RF power fluctuations and physical perturbations within the plasma itself (plasma flicker). In essence, 54 any signal instability results in non-sample related variation in relative isotope abundances from one 55 microsecond to the next. To achieve the ultimate high accuracy and precision that MC-ICP-MS is 56 capable of, does pose some extra constraints, these include the need to separate the element of 57 interest not just to ensure minimising any spectral interferences but also to minimise matrix effects, 58 Soil samples were air-dried and sieved to <2 mm and a small subsample of the archived soil (c. 2.0 g 123 in triplicate) was ground in an agate ball mill (Model PM400, Retsch GmbH & Co., Germany). Complete 124 digestion of 0.2 g samples was achieved in PFA vials in a 48-place Teflon-coated graphite block digester 125 ( Model A3, Analysco Ltd, Chipping Norton, UK) using a mixed acid digestion process i.e., HNO3 (68%), 126 HClO4 (70%) and HF (70%). The dried digestate was dissolved in 68% HNO3 and diluted to 50 mL with 127 Milli-Q water (18.2 MΩ cm). Prior to analysis of total Pb concentration by ICP-QMS a further 1:10 128 dilution with 2% HNO3 was undertaken. Internal standards were introduced to the sample stream via 129 a separate T-line including Ge (50 µg L -1 ), Rh (20 µg L -1 ) and Ir (10 µg L -1 ) in 2% HNO3 and multi-element 130 standards (0, 20, 40, 100 μg L -1 ) were used to determine the Pb concentration (CLMS-2; CertPrep™). 131 Three procedural blanks were also included in each digestion batch for quality assurance. All samples 132 were analysed with the ICP-QMS operating in 'kinetic energy discrimination' (KED) mode, with He as 133 the cell gas, to eliminate polyatomic interferences e.g. 191  The precision of Pb isotope ratio measurements generally improves with increasing Pb concentration 136 as counting statistics improve (19). However, at high concentrations the ion counter detector used 137 with the ICP-QMS employed in this study trips to an analogue signal, resulting in non-linearity caused 138 by measuring isotopes in different detector modes, with added uncertainty due to the application of 139 a gain correction between the two detector types (20, 23). Thus, individually tailored dilutions of each 140 sample digest were carried out, with 2% HNO3, using previously quantified total concentrations of Pb. 141 The objective was to work with near identical concentrations (~15 µg L -1 ) of all samples and reference 142 materials (SRM-981, NIST Pb wire). This approach had the advantages of (i) reducing the effect of any 143 error in the dead time correction factor and (ii) ensuring that the intensity of the 204 Pb signal was as 144 high as possible (c. 27 X 10 3 cps) while keeping the 208 Pb signal below the detector trip point (c. 1 X 10 6 145 cps). The blank contribution for each isotope although subtracted during analysis was only 0.76% of 146 the sample Pb and had an insignificant effect on results. 147

(ii)
Lead separation method for Pb isotope measurement by MC-ICP-MS.

148
Variations in mass bias and small interferences are known to impact on data quality in MC-ICP-MS 149 isotope ratio determinations (24). Therefore, prior to the measurement of Pb isotopes in soil digests, 150 a preliminary separation of Pb was performed using an anion exchange (AG1-X80) column (25). 151 Aliquots, containing an estimated 200 ng of Pb from each digest, were heated overnight to dryness in 152 PFA vials on a hotplate. An aliquot of HBr (2 mL) was then added to each vial and left overnight again 153 to solubilise the sample. Sample Pb was separated using anion exchange resin columns (2 mL, AG1-154 X80). The resin was pre-cleaned with MQ quality (>18 MΩ cm) water followed by 6 M HCl and then 155 equilibrated by washing the column with 2 mL 1 M HBr. All samples were then passed through 156 individual columns adsorbing the sample Pb as PbBr4 2ions. Major cations were then eluted from the 157 columns with further HBr. The sample Pb was then eluted using 8 mL of 6 M HCl back into its original 158 vial. Concentrated nitric acid (0.2 mL) was added to each vial as an oxidising agent and the contents 159 evaporated to dryness on a hotplate, overnight, at 90˚C. Prior to analysis by MC-ICP-MS each sample 160 was diluted to c.5-10 ng mL -1 and spiked with Tl (c. 5-10 ng mL -1 ) for instrumental mass bias correction 161 with the aim of achieving less than the maximum 10V signal on 208 Pb and 205 Tl. 162

ICP-QMS 164
At the University of Nottingham Pb isotope ratios were determined by current generation ICP-QMS 165 (Model iCAP Q; Thermo Scientific, Bremen, Germany). All samples were analysed with the ICP-QMS 166 operating in 'kinetic energy discrimination' (KED) mode, with He as the cell gas, to reduce potential 167 polyatomic interferences. Additionally, in the past it has been suggested that the use of a collision gas 168 may "thermalize" the ion beam and improve isotope ratio precision (26). Samples were introduced 169 from an autosampler incorporating an ASXpress™ rapid uptake module (Cetac ASX-520, Teledyne 170 Technologies Inc., Omaha, NE, USA), with a 5 mL sample loop, through a perfluoroalkoxy (PFA) 171 Microflow PFA-ST nebuliser (Thermo Fisher Scientific, Bremen, Germany). The typical operational 172 parameters selected for measurement of Pb isotopic ratios are given in Table 2. The settle/idle time  173 was not optimised as this is hard-wired into the software of this instrument and not readily available 174 to the user. 175 using the instruments array of Faraday detectors. The collector configuration used is illustrated in 183 Table 3. Each individual acquisition consisted of 3 blocks of 25 cycles, with a 5-second integration time 184 per cycle, following a 60 second de-focused baseline (c. 7 min. per acquisition). The measured Tl ratio 185 was used to make an exponential correction for instrument induced mass bias effects on both Pb and 186 Hg isotopes, based on an assumed 205 Tl/ 203 Tl ratio of 2.3890. Hg was measured at amu 202 and a 187 proportionate amount subtracted from the ion beam at amu 204, based on an assumed 204 Hg/ 202 Hg 188 ratio of 0.229883. 189 The precision and accuracy of the method was assessed through repeat analysis of a SRM-981, NIST 190 Pb reference solution, (also spiked with Tl The analytical errors for each of the sample ratios were propagated relative to the reproducibility of 196 the SRM-981, NIST to take into account the errors associated with the normalisation process. 197

198
Evaluation of detector dead time correction factor 199 A dead time correction factor (dtcf) was determined, according to a method described by Nelms,200 Quétel (17), for the isotopic ratios: 204 Pb/ 208 Pb, 206 Pb/ 208 Pb and 207 Pb/ 208 Pb. Values of dtcf were 201 measured from intensity ratios determined across a range of Pb standard solutions (i.e. 0, 2, 4, 6, 202 8....22 µg L -1 ) prepared from SRM-981, NIST in 2% HNO3. Data from samples providing higher count 203 rates for 208 Pb were used so that the detector was more likely to be close to saturation; we also wanted 204 to investigate the validity of measurements up to the point where the detector tripped to analogue 205 signals. 206 The dtcf was initially set to zero ns. An isobaric correction factor for 204 Hg was determined from 207 measurement of the signal at m/z = 202 ( 202 Hg). This correction was negligible for samples as Hg would 208 be efficiently evaporated from the sample solutions during the high temperatures of the digestion 209 procedure. Intensity values (cps) were measured using a quadrupole dwell time of 10 ms with 200 210 sweeps and 20 separate integrations to give 4000 quadrupole visits to each isotope per sample and a 211 total signal integration time per isotope of 40 seconds. Initially, a blank correction was made to all raw 212 Pb isotope intensities (Isample -Iblank). Sample intensity values were corrected using Eq. 1, following 213 optimisation of the dtcf value (28). 214 I corr = (I sample − I blank ) 1 − (I sample − I blank )dtcf 10 −9 (1) 215 The value of dtcf was optimised by determining the slope of individual isotope intensity ratios 216 (i.e. 204 Icorr/ 208 Icorr) against the intensity signal for 208 Pb (cps) and varying the value of dtcf until the value 217 of the slope squared (slope 2 ) was minimised. This was undertaken using the Microsoft Excel Solver 218 function. The optimised values of dead time calculated for Pb isotopic ratios were 37.4, 36.7 and 36.3 219 ns for 204 Pb/ 208 Pb, 206 Pb/ 208 Pb and 207 Pb/ 208 Pb ratios respectively. The values of dtcf calculated did not 220 change systematically with mass difference between the isotopes and they were independent of mass 221 and analyte concentration as illustrated in Fig. 1

227
Prior to mass bias correction, Pb isotope intensities were corrected for an isobaric interference from 228 204 Hg, through measurement of 202 Hg, and for operational blanks. The resulting data was then 229 corrected for mass bias using both proxy internal and external standards. Thallium (2 µg L -1 ) was used 230 as an internal standard because its isotopic masses (i.e., 203 Tl and 205 Tl) are sufficiently close to those 231 of the Pb isotopes that it can be used as a proxy for continuous monitoring of mass discrimination in 232 individual samples (22) Three equations (i.e., linear, power and exponential models; equations 2-4 below) are generally used 236 to estimate the mass discrimination correction factor (or 'K-factor') for a pair of Pb isotopes when Tl 237 is used as an internal standard (15,30). Most previous studies conducted using MC-ICP-MS have 238 favoured a power (Eq 3.) or exponential (Eq 4.) equation over a linear relation (Eq. 2) on grounds of 239 precision for internal standardisation (1, 15, 30). However, a linear model (Eq. 2) is more commonly 240 used for mass discrimination correction for quadrupole mass analysers as this has been found to be a 241 sufficiently accurate model as their precision is rarely good enough to discriminate between models 242 (31). 243 (4)  246 In equations 2 -4, Δm is the mass difference between two Pb isotopes; K is the K-factor for the ratio 247 of those two isotopes and ε is the bias per unit mass determined for the ratio 205 Tl/ 203 Tl. Preliminary 248 data for Pb isotopes were processed using the linear (Eq. 2), power (Eq. 3) and exponential (Eq. 4) 249 equations to assess the difference in mass bias correction factor, but very similar results were 250 observed using the three approaches. Therefore, the simplest linear correction was used in this study 251 and applied to all samples and the NIST SRM-981 Pb reference material. 252 Thallium-corrected data was then further normalised using the NIST SRM-981 Pb reference material 253 to allow for any element-specific drift in instrument sensitivity during the experiment. The refined K-254 factor for each sample was calculated using interpolated values of the K-factor for SRM-981, NIST 255 samples (15 µg L -1 ) run every 10 th sample. The magnitude of the K-factor correction using Tl varied 256 between 0.5% for the 207 Pb/ 206 Pb ratio to 2% for the 208 Pb/ 204 Pb ratio, reflecting the greater correction 257 for mass bias for wider spaced isotopes. The subsequent SRM-981 K-factor corrections varied between 258 0.25% and 0.5% and were not mass dependent, probably reflecting differences in Pb-Tl responses, 259 that others have addressed using artificially modified Tl isotope ratios. A table of these K-factors for 260 both correction stages are given in Electronic Appendix Table 1.

278
To optimise the number of scans (peak visits by the quadrupole), 10 separate runs with 1000 scans 279 per run was selected and the number of points per peak was set to one (dwell time 10 ms). where IRMC and IRQuad are isotope ratio measurements from the multi-collector and quadrupole 290 instruments and df represents degrees of freedom. 291 The data processed using individual groups of 1000 scans showed relatively high and variable residual 292 standard deviation values for all isotopic ratios (grey circles). By contrast, data processed in a 293 cumulative way showed better precision and consistency for all three isotopic ratios. Precision 294 generally increased with increasing number of scans as shown in Fig 3. (a-d). For the current study we 295 decided to use 10,000 scans to try to minimise variation due to counting statistics, although Fig .3  296 suggests this could potentially be reduced to 6000 scans to save time without significant loss of 297 precision. For example, if the precision of the ratios is solely based on counting (Poissonian) based 298 statistics, then the standard deviation of the number of counts measured is the square root of the 299 mean number of counts. This would result in a √10 = 3.162 improvement in ratio precision, when using 300 10,000 ratios rather than 1,000. Whereas acquiring 10,000 ratios rather than 6,000 would produce a 301 more modest precision improvement of √(10,000/6,000) = 1.291. This potentially being swamped by 302 other normally distributed noise sources. 303 Following optimisation of the Pb isotopic ratios measurement protocol, the processed data for diluted 307 soil sample digests analysed with ICP-QMS (10,000 scans) were compared to equivalent data for 308 chemically separated Pb measured by MC-ICP-MS. Each of the 3 digestions for an individual sample 309 were treated as a separate data point to achieve the most direct comparison of the two techniques. 310 An excellent correlation (r=0.999 and r=0.998; n=36) was observed for 206 Pb/ 207 Pb and 208 Pb/ 207 Pb 311 isotopic ratios with slope values of 1.0008 and 1.0006 respectively shown in Fig 4. (a and b). The 312 correlation observed for the 208 Pb/ 204 Pb isotopic ratio was also good, albeit a little weaker than for the 313 other ratios (Fig 4. c; r= 0.987, slope=1.0017). This is understandable given that: (i) there is a greater 314 dependence on mass bias because of the mass difference of 4 amu; (ii) 204 Pb (isotopic abundance = 315 1.4%) produces a lower signal giving poorer counting statistics; (iii) the large difference in isotopic 316 abundance results in increased vulnerability to errors in dead time correction and (iv) the need for Hg 317 correction on 204 Pb, which further compromises the measurement uncertainty. Of these potential 318 effects, the dominant and unavoidable one will be number (ii), poorer statistics from the isotopic 319 abundance of 204 Pb c.1.4% compared to the next smallest isotope 207 Pb c.22.1%; this being a factor of 320 c.3.4. All of the other effects are either minimised by optimisation, (i) and (iii), or, in the case of (iv), 321 insignificant for samples prepared in the current study by HF digestion. 322   With such strong correlations it is instructive to examine the difference between the 323 accuracy/precision of the benchmark MC-ICP-MS and the ICP-QMS data (Fig 5. a-d), assuming that 324 the MC-ICP-MS has no error for the purposes of this comparison. There may be two sources of error 325 between the datasets: additive i.e. a simple offset value constant for all samples, and multiplicative 326 i.e. a change in value proportional to the isotope ratio measured. For 206 Pb/ 207 Pb (Fig. 5a) there was a 327 clear increase in error with increasing isotopic ratio, indicating a multiplicative bias between ICP-QMS 328 and MC-ICP-MS. For 208 Pb/ 207 Pb and 208 Pb/ 204 Pb this was not apparent. A regression analysis of ICP-329 QMS on MC-ICP-MS was made and tested for significance of the slope being different to 1 for all ratios 330 shown in Table 4. All slope values were statistically different from 1 except 208 Pb/ 206 Pb; however only 331 in the 4 th significant figure. To account for the effect of the very narrow range of isotope ratio values, 332 an intercept between the ICP-QMS and MC-ICP-MS data was estimated by correcting the ICP-QMS 333 data for the slope, then calculating the difference between the two data sets. The median difference 334 was calculated (Table 4) to avoid outliers biasing the estimate. In terms of impact, the offset for 335 208 Pb/ 204 Pb ratio was the most significant with an absolute value of 0.02, equivalent to approximately 336 0.04% of the median 208 Pb/ 204 Pb ratio. 337 As each original soil sample was digested and analysed separately 3 times there was an opportunity 338 to test 'fitness for purpose' for the whole method from digestion to ICP-QMS and MC-ICP-MS 339 measurement. This was achieved using analysis of variance (ANOVA) to separate the total variance of 340 the dataset into the analytical variance and the true variance between samples; where the analytical 341 variance is a combination of the digestion variance and the measurement variance.

Typical Error Bars
In evaluating the quality of data produced by the two techniques the first comparison is the within-343 sample analytical precision (as measured by RSD %) of the ICP-QMS with that of MC-ICP-MS (Table 5). 344 If we assume that the MC-ICP-MS inherently has a better measurement precision but with the within 345 sample precision essentially being equal between the techniques, then we must assume that the 346 digestion process has limited our analytical precision. The ANOVA, via the F-statistic, can be further used to test, for each isotope ratio, the power of the 360 analytical technique to discriminate between samples. The greater the ratio of the "between samples" 361 variance and the "within sample" variance which is the F-statistic, the better the discrimination i.e. 362 when the ratio falls below the Fcrit value we have no discrimination. Binary plots were used to attempt soil Pb source apportionment for the three sets of data derived 373 from (i) MC-ICP-MS, (ii) ICP-QMS measured using 10,000 scans, and (iii) ICP-QMS using just the first 374 run of 1,000 scans as an example of acquisition parameters used previously in the literature (33). The 375 first comparison was made using a conventional binary plot ( 208 Pb/ 207 Pb vs 206 Pb/ 207 Pb) as shown in 376 Fig. 6. The MC-ICP-MS data (dark circles) all fell on a single straight trend line, expected from a simple 377 2 isotopic end-member system. Similarly, data analysed by ICP-QMS with 10,000 scans (grey circles) 378 fell very close to the multi-collector data. By contrast, data from ICP-QMS using 1000 scans (white 379 circles) clearly deviated from the multi-collector data and the single mixing line in the binary plot. The 380 standard error at each data point (n=3) was also large compared to ICP-QMS data analysed using 381 10,000 scans where reproducibility was very high. 382 A further comparison was made using a second binary plot (Fig. 7) involving the 204 Pb isotope ( 208 / 204 Pb 383 versus 206 Pb/ 207 Pb) in an attempt to identify the presence of further end member sources not revealed 384 by Fig. 6; as Ellam (34) has pointed out plots based only on 206,207,208 Pb ratio have a limited ability to 385 discriminate more than 2 end-members. Again, all data points determined by MC-ICP-MS fell on a 386 single straight mixing line, providing no evidence of a 'third source' of Pb. The data determined by ICP-387 QMS using 10000 scans (grey circles) again fell very close to the multi-collector data. However, a few 388 points visibly deviated from the single mixing line and the standard errors were very small, suggesting 389 a small systematic deviation. The largest systematic deviations and standard errors (n=3) were 390 observed for ICP-QMS data using 1,000 scans (white circles). It is clear that although a straight line 391 through the data would indicate 2 dominant sources, the substantial scatter around the mixing line 392 could be erroneously interpreted as indicating multiple (> 2) sources of Pb in the current dataset. This 393 highlights the necessity for longer total acquisition time than previously reported (33, 35) and a fuller 394 understanding of causes of error in determination of Pb isotope ratios. 395 Figure 6. A binary plot using isotopic ratios 208 Pb/ 207 Pb vs 206 Pb/ 207 Pb for data derived from MC-ICP-MS and ICP-QMS with either 10,000 or 1,000 scans. Note error bars for MC-ICP-MS and ICP-QMS (10000 scans) are smaller than symbols for most samples.  In summary, although ICP-QMS is not capable of the absolute precision or the full source 429 discrimination power of MC-ICP-MS, it is clearly capable of providing 'fit-for-purpose' Pb isotope ratio 430 data on environmental samples, such as soils, where sample heterogeneity can be a limiting factor. 431

432
SN is an employee of Thermo Fisher Scientific but the study was independently designed and 433 executed by the other authors. 434