Improving Precision and Accuracy of Isotope Ratios from Short

Mar 31, 2016 - These two methods were applied using postacquisition computational algorithm via a VBA (visual basic for application) Macro, developed ...
0 downloads 10 Views 1MB Size
Download PDF
Article pubs.acs.org/ac

Improving Precision and Accuracy of Isotope Ratios from Short Transient Laser Ablation-Multicollector-Inductively Coupled Plasma Mass Spectrometry Signals: Application to Micrometer-Size Uranium Particles Fanny Claverie,*,† Amélie Hubert,‡ Sylvain Berail,† Ariane Donard,†,‡ Fabien Pointurier,‡ and Christophe Pécheyran† †

Laboratoire de Chimie Analytique Bio-Inorganique et Environnement, IPREM UMR 5254, CNRS, Université de Pau et des Pays de l’Adour, 2 Avenue du Président Angot, 64053, Pau Cedex 9, France ‡ CEA, DAM, DIF, F-91297 Arpajon, France S Supporting Information *

ABSTRACT: The isotope drift encountered on short transient signals measured by multicollector inductively coupled plasma mass spectrometry (MC-ICPMS) is related to differences in detector time responses. Faraday to Faraday and Faraday to ion counter time lags were determined and corrected using VBA data processing based on the synchronization of the isotope signals. The coefficient of determination of the linear fit between the two isotopes was selected as the best criterion to obtain accurate detector time lag. The procedure was applied to the analysis by laser ablation-MC-ICPMS of micrometer sized uranium particles (1−3.5 μm). Linear regression slope (LRS) (one isotope plotted over the other), point-by-point, and integration 234 U/238U ratios. Relative internal precisions of 0.86 to 1.7% and 1.2 respectively, using LRS calculation, time lag, and mass bias corrections. 235 U/238U ratios with good accuracy (relative difference with respect to

M

were due to the differences between the response times of the various Faraday detectors, referred to as “detector time lags”.2−6 Several methods have been developed to correct or minimize the isotope drift. By rapidly changing the signal intensity using copper solutions with different concentrations, Hirata et al.5 observed that the isotope ratio (65Cu/63Cu) was correlated to the rate of change in intensity (V s−1) and estimated a correction factor for the Faraday amplifier. They reduced the relative isotope variation in LA-MC-ICPMS analysis from 3 to 5‰ down to below 1‰. In another study, conducted by Epov et al.,7 analytical methodology was developed for speciesspecific isotopic composition of Hg by GC-MC-ICPMS analysis using a standard−bracketing protocol combined with the use of δ-notation relative to standard reference material (NIST SRM-3133). They managed to reduce the limitations induced by the drift and the short signal and obtained an external 2SD precision of 0.56‰ for δ202Hg. The time response between the two Faraday detectors used to measure the isotopes has been calculated on lead isotopic ratios by

ulticollector-inductively coupled plasma mass spectrometry (MC-ICPMS) is ideally suited for precise isotope ratio measurement. Recently, the development of hyphenated methods has opened a new realm of possibility with the use of online chromatography, solid sampling, traps, etc. Among them, laser ablation (LA) allows direct solid sample analysis with no (or little) sample preparation (avoiding contamination), high sample throughput, and high sensitivity thanks to the generation of short (from a few seconds to a few minutes) transient signal. However, measuring isotope ratios on short and intense signals can be challenging and leads to poorer precision compared to conventional continuous sample introduction.1 Isotope ratio drift during short transient signals has been observed during MC-ICPMS experiments. It is characterized by a systematic isotope change along the signal. Ten years ago, Krupp et al.2 observed this phenomenon with gas chromatography coupled to MC-ICPMS (GC-MCICPMS). They studied potential causes such as chromatographic fractionation, change in instrumental mass bias, analyte concentration, and peak shape but identified the fast change of intensity of the signal as the most probable reason for the isotope drift. This hypothesis was shared by several authors who established that the biases © XXXX American Chemical Society

methods were tested to calculate the 235U/238U and to 2.4% were obtained for 235U/238U and 234U/238U, A relative external precision of 2.1% was obtained for the reference value below 1%).

Received: December 18, 2015 Accepted: March 31, 2016

A

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry Gourgiotis et al.8 When using flow injection analysis-MCICPMS and GC-MC-ICPMS, a time lag of less than 10 ms was obtained and the postacquisition synchronization of the signals was performed accordingly, which leads to better repeatability (improved by a factor 13) and better measurement uncertainties (improved by a factor of 14−20). Pettke et al.6 presented two correction methods that allow the difference in amplifier response among Faraday detectors to be accounted for. They also demonstrated that the integration of the entire signal is the most suitable method to obtain accurate isotope ratios from transient signals. In this sense, integration of the whole peak, by summing the total number of baselinesubtracted counts for each isotope, has been preferred by Cottle et al.9 in order to overcome the delay (∼0.2 s) observed between Faraday and ion counter detectors. However, a different data processing, which has not been tested in studies reported by Pettke et al.6 and by Cottle et al.,9 for precise and accurate determination of isotope ratios on transient signals using hyphenated MC-ICPMS has recently been evaluated.10−12 The linear regression slope (LRS) method developed by Fietzke et al.13 and based on the fit of both isotopes over the entire collecting data (the slope of this linear fit representing the isotope ratio) has been shown to provide improved accuracy and precision. The LRS method presents several advantages over the integration and point-by-point methods. As with the integration method, even if all data points are taken into account for the calculation, the highest intensity signals have the highest influence on the calculated isotope ratio. Additionally, as with the point-by-point method, LRS provides an internal standard deviation, which is very useful information when dealing with single particle analysis where one unique test is possible. The precise and accurate measurement of major and minor isotopes of single uranium particles are of great interest for safeguarding purposes.14 Nowadays, uranium isotope ratio determination from single particles, referred to as “particle analysis”, is routinely performed by two analytical methods: thermal ionization mass spectrometry (TIMS) and secondary ionization mass spectrometry (SIMS). However, low ionization efficiency and the presence of isobaric interferences, for TIMS and SIMS, respectively, limit their use for minor isotope measurement. Laser ablation coupled to ICPMS is a promising technique for this application. Ablation of particles produces short transient signals, which present a high signal-to-noise ratio leading to a low limit of detection.15,16 Recently, LA-MCICPMS has been used to achieve higher precision on isotope ratio measurements.10,17 Lloyd et al.17 reached relative external precision of 0.22% for 235U/ 238U in uranium oxide particles (9 μm diameter ablation) by LA-MC-ICPMS. Comparison of isotope ratio calculation by three different methods (point by point, linear regression slope, and integration) for the measurement of 235U/238U was performed by Kappel et al.10 in several micrometric uranium particles made of isotopically certified reference materials. To our knowledge, only one study has reported the measurement of minor isotopes (234U and 236 U) in particles by LA-MC-ICPMS.18 Those particles were made of silicate doped with uranium. However, to date, internal precision obtained for analysis of individual micrometer-sized uranium particles is higher than 1% for the 235U/238U ratio and several % for the minor isotope ratios (234U/238U and 236 U/238U), which is beyond the figures obtained with the well-established techniques for particle analysis (SIMS and TIMS).

The aims of this work are to study and correct the isotope ratio temporal bias observed on short transient signals produced by LA-MC-ICPMS in order to achieve the best accuracy and internal precision. Because of large differences of abundances between minor and major uranium isotopes, different types of detectors (ion counter or Faraday cup) were used. An isotope ratio temporal bias correction method was developed and then applied to uranium micrometer-sized particles. For the first time, a detector time lag correction was applied to the short transient signals produced by laser ablation of individual uranium micrometric particles (1−3.5 μm diameter range) and measured by MC-ICPMS. Both 234 U/238U and 235U/238U atomic ratios were measured.

■

EXPERIMENTAL SECTION Instrumentation. A femtosecond laser ablation system (Alfamet, Novalase S.A., France) fitted with a diode-pumped KGW-Yb crystal was used. It operates at an IR wavelength of 1030 nm and delivers 360 fs pulses at high repetition rates from 1 Hz to 10 kHz and low energy (0.1−100 μJ at the sample surface). In order to reach sufficient fluence, a 50 mm focal length objective was fitted in the laser machine to provide a 17 μm laser spot. The narrow laser beam offers high spatial resolution, enabling the ablation of very small objects. A galvanometric scanning beam device (from Scanlab) allows fast movement of the laser beam with high positioning precision. Combined with the movement of the sample, complex trajectories can be performed in order to virtually enlarge or change the laser beam shape.19,20 Our method is intended to be used for small particles. Thus, only low intensity signals are obtained, especially for low abundance isotopes. The ablation conditions used to ablate the IRMM 541 glass were chosen to obtain signal intensities (∼0.1 V for 238U) far below the upper limit of the MC-ICPMS detectors (10 V). 40 μm diameter craters were obtained using three concentric circles at a scanner speed rate of 0.5 mm s−1 and a repetition rate of 500 Hz (resulting in 900 shots in 2 s) for IRMM 541 ablation and 0.9 mm s−1 and 300 Hz (resulting in 60 shots in 0.2 s) for particle ablation. The fluence was set to 28 J cm−2 in both cases. It should be noted that the impact of the laser at the surface of polycarbonate disks coated with a thin polymer layer has been studied in a previous paper.15 It has been shown that using high repetition rate below 1 kHz leads to well-defined geometry crater without melted rims. The number of pulses applied to the sample enables the entire ablation of the particle. A “Nu plasma” MC-ICPMS (Nu Instruments, U.K.) was used for detection. Wet plasma conditions gave the best signal stability and were used throughout the study. A 2% HNO3 (sub-boiling distillation of Instra grade HNO3, Baker) solution was simultaneously introduced via a 200 μL min−1 microconcentric nebulizer and a 20 mL cinnabar spray chamber (both from Glass Expansion) coupled to a two-inlet torch (made to order by Nu Instrument). Data were acquired using the time resolved analysis (TRA) mode with an integration time of 0.2 s. More details concerning MC-ICPMS parameters are given in Table 1. The gains between the Faraday preamplifiers were determined using the instrumental routine on a daily basis. The gains between Faraday and ion counters, also called yields, were measured using a 0.5 μg L−1 uranium solution at the beginning and at the end of each day and were found to be constant (relative standard deviation below 1.1% on 3 days of analysis). A dynamic analysis was performed, B

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Na2O, 8% CaO, and 3% Al2O3. Its nominal U mass fraction is 50 ppm with an amount ratio of 235U/238U of (7.277 ± 0.007) × 10−3 (at the 2σ level) corresponding to natural abundance. NU particles were produced specifically for particle analysis (AEA Technology, Harwell, U.K.). They were deposited on a transparent polycarbonate disk (25 mm diameter) and fixed into a coating of an organic compound called “collodion” (mixture of ether and nitrocellulose). They were directly located with the camera of the laser ablation system. Although the isotopic composition of these particles is not certified, they were produced from an NU bulk material and reference values of (7.248 ± 0.016) × 10−3 for 235U/238U and (6.1 ± 1.3) × 10−5 for 234U/238U obtained by TIMS from a suspended solution were provided.22 The size of the particles ranged from 0.1 to 3.5 μm diameter with most of them measuring between 0.6 and 1.4 μm, leading to an average diameter of 1 μm.22 The uranium standard solution of 0.5 μg L−1 was prepared from IRMM 184 SRM (IRMM, Geel, Belgium). It contained NU with a certified 235U/238U isotope ratio of (7.2623 ± 0.0022) × 10−3 (at the 2σ level). Nebulization of this solution was used to determine the gains between amplifiers of the Faraday cups and ion counter detectors as presented previously. Correction of Time Lags between Detectors. In order to correct for the time lags between detectors, two methods were employed. The first method was derived from the one developed by Gourgiotis et al.8 and applied to flow injection analysis (FIA) and gas chromatography-MC-ICPMS coupling using only Faraday detectors. It consisted of minimizing the linear regression slope of the isotope ratio calculated point-bypoint as a function of time (Risotope1/isotope2 = f(t) ≈ kt + b) by synchronizing the transient isotope signals. Indeed, if the isotope signals are synchronized (no time lag), this slope (k) should be equal to 0. A second method based on the maximization of the coefficient of determination R2 of the plot of isotope 1 versus the isotope 2 was developed. It relies on the fact that when the isotope signals are perfectly synchronized and under ideal detection conditions (e.g., no noise), this linear regression slope should give a coefficient of determination equal to 1. These two methods were applied using postacquisition computational algorithm via a VBA (visual basic for application) Macro, developed with Microsoft Excel software. From a selected zone surrounding the peak signal (10 s on each side of the peak borders, defined manually and corresponding typically to 150 data points per isotope), the macro applied a background correction to the raw data. The first 10 s of the signal (i.e., at least 10 s prior to ablation) were used to calculate the average blank signal for each isotope. Then, to allow precise time resolution adjustment, the macro performed a linear interpolation of the data between each successive point of the background corrected signal on this selected area. The number of interpolated values between two acquisition points was set to 200 for Faraday cup/Ion Counter (FC/IC) time lag determination (i.e., resulting in one data point every 200 ms/ 200 = 1 ms) and 1500 for Faraday cup/Faraday cup (FC/FC) time lag determination (i.e., resulting in one data point every 200 ms/1500 = 0.13 ms). The determination of the time lag was then performed iteratively: the isotopes were shifted one over the other by one increment unit (i.e., by step of 1 and 0.13 ms for FC/IC and FC/FC configuration, respectively) until the optimal value of the criterion was obtained (the slope as close as possible of 0 for method 1 and the higher coefficient of determination for method 2). Once the optimal time lag was

Table 1. LA-MC-ICPMS Operating Conditions and Detector Configurations laser ablation system parameters wavelength pulse duration fluence He flow rate ablation strategy

1030 nm 360 fs 28 J cm−2 580 L min−1 3 concentric circles at 300 Hz and 0.9 mm s−1 (particles) 500 Hz and 0.5 mm s−1 (IRMM) Nu plasma HR MC-ICPMS parameters

plasma gas flow rate auxiliary gas flow rate nebulizer gas pressure rf power interface cones acceleration voltage instrument resolution integration time

C1 C2

L1 238 U

L2 238

U

13 L min−1 0.8 L min−1 24.0 psi 1300 W Ni wet plasma cones (type A) 6000 V low 0.2 s detector configuration IC0 236 U

L3 235 U

IC1 234 U 235 U

IC2 234

L4

U

which consisted of measuring the same isotope with different detectors, by modifying the intensity of the magnetic field. For instance, concerning configuration 2 (see Table 1), 238U was measured on L2 during the first cycle and then on IC0, IC1, and IC2 on the following cycles. Ratios of the signals were calculated for each pair of detectors and applied off-line to the corresponding isotope ratios of the ablated material (IRMM 541 and particles). For example, for configuration 2, the 235 U/238U ratio was corrected using the ratio IC1/L2 obtained during the gain determination. A mass bias correction was applied using exponential law. Particles of natural uranium measured at the beginning and at the end of each analytical session were used to determine an average mass bias factor. This mass bias factor was then applied to the other particles measured during the session. All uncertainties in the present paper are given as combined standard uncertainties. It should be noted that throughout this paper, the terms “internal precision” and “external precision”, often used in isotopic analysis, were used instead of “repeatability precision” and “reproducibility precision” as defined in ref 21. The developed method will be dedicated to the analysis of particles which can be made of natural uranium (NU, 235U/238U = 7.26 × 10−3), enriched uranium (EU, higher 235U abundance than in NU) and depleted uranium (DU, lower 235U abundance than in NU). Two configurations were tested to select the best way to measure precise uranium isotope ratios for micrometric particles. The configurations are presented in Table 1. The first configuration allows the detection of 238U and 235U with similar detectors, i.e., Faraday cups (FC), and 234U and 236U with ion counters (IC), whereas the second configuration allows the detection of the highest abundant isotope 238U with FC and the isotopes of lower abundance (234U and 235U) with IC. Samples and Reagents. The evaluation of the method has been performed using two sample types: IRMM 541 glass and particles of natural uranium (NU particles). The IRMM 541 (from the Institute for Reference Materials and Measurements, Geel, Belgium) is a glass material composed of 75% SiO2, 14% C

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry obtained, averages of data points were done in order to obtain the exact same numbers of data points per second than before the interpolation (in our case, one data point every 0.2 s). Finally, the macro calculated the isotope ratio by LRS, the isotope ratio being the slope of the linear fit. Higher degrees of interpolation (up to 4000 corresponding to a time resolution of 0.05 ms) were used in this study. IC/IC detector time lags were studied using the 234U/235U ratio obtained with configuration C2. The optimal criterion was not always reached, probably due to the very low time lags of IC/IC detectors (below 3.5 ms) combined with the background noise and the low signal intensities. Concerning, FC/IC and FC/FC detector pairs, higher degrees of interpolation did not bring more information and were not used further as the calculation time was drastically increased. In our conditions (use of 200 or 1500 interpolated data between two acquisition points), the calculation time was short, ranging from 1 to 2 s.

■

RESULTS AND DISCUSSION Isotope Ratio Temporal Bias Model. Laser ablation leads to short transient signals, which are affected by the isotope ratio temporal bias as previously mentioned. Figure 1 shows a typical signal profile obtained when ablating a ∼1 μm uranium particle. The signal profile shows an asymmetrical triangular shape, with a fast rise and a slower descent.

Figure 2. Simulated signal profiles, point-by-point related isotope ratios and corresponding linear regression slopes of the two isotopes when acquisition of isotope 2 is delayed (acquisition with the detector with the highest time lag) (a and b) and when isotope 1 is delayed (c and d). Red dots on the LRS in parts b and d correspond to ratios calculated from the rise in the isotope signals and blue dots to the decrease in the signals.

from both cases, included five points out of the line (up for Figure 2b and down for Figure 2d), which correspond to the ratios of the 5 consecutive points of the fast signal rise. The steeper the slope of the signal, the stronger the effect of the time lag. Although the bias induced by differences in amplifier time responses among Faraday detectors has been studied in the past, ion counter to Faraday time lag has, to our knowledge, been little explored. This is probably related to the fact that this effect is visible for transient signals only and has no influence when isotope ratios are calculated by peak area integration.18 Time Lag Correction. Time lag correction was performed using the developed macro, based on the two regression criteria (k as close as possible to 0 with method 1, and R2 as close as possible to 1 with method 2). However, as demonstrated above, the asymmetric triangular shape of the time-shifted signals did not generate a linear isotope ratio temporal bias and therefore, method 1 was not appropriate. In addition to the nonlinear profile of the isotope ratio, the definition of the left and right borders of the peak of interest was very critical for the time lag determination. A low signal typically observed at the beginning of the peak produced noisy isotope ratios that drastically influenced the slope of isotope ratios as a function of time, which thus limit accurate or even reproducible determination of time lag values. In contrast, method 2 directly uses the LRS of one isotope against the other which drastically reduces the influence of lower signals on the slope value (isotope ratio) and R2. Therefore, method 2 avoided the possible error induced by the subjective choice of the peak borders. It should however be noticed that, for most of the tested ablation signals, similar time lags were obtained with both methods, although the slope criteria method required more caution and was more time-consuming. Method 2 was, thus, used throughout this study.

Figure 1. Signals obtained after ablation of a uranium particle using C2 configuration.

In contrast to Gaussian shape signals, the point-by-point isotope ratios obtained from shifted triangular signals cannot be assimilated into a linear plot. Figure 2 simulates signals from two isotopes (named 1 and 2) of same intensity (IR = 1) when isotope 2 is delayed (Figure 2a) or isotope 1 is delayed (Figure 2c). Stronger isotope ratio temporal bias is obtained when the signal is rising, followed by a second lower isotope ratio temporal bias when the signal is dropping. The isotope ratio profiles as a function of time provided by this model are similar to the ones observed for the ablation of particles (Figure 1). A mathematical demonstration is given in the Supporting Information showing that the point-by-point ratio of two time-shifted triangular shaped curves can be assimilated to a 1/ (1 + X) type function for each side of the peak (rise and fall, respectively). Note that depending on the time lags of the detectors used for acquisition of isotopes 1 and 2, the increasing part of the peak shows isotope ratio systematically above (Figure 2a, highest time lag for detector used for measuring isotope 2) or below (Figure 2c, highest time lag for detector used for measuring isotope 1) the expected ratio of 1. The opposite trend is observed for the decreasing part of the peak. The LRS of the isotope 1 versus isotope 2 plot, obtained D

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

IRMM 541 was performed to avoid the irregularity of the signals observed when ablating particles whose size and shape can differ. Ablations were performed with the same conditions (laser parameters and position in the ablation cell) and during the same day. While ablations of individual particles were performed in separate LA-MC-ICPMS acquisitions, five to ten ablations of IRMM 541 were performed within the same acquisition in order to study the internal and external variability of the time lags. As shown in Figure 4, identical time lags were

A typical example of the time lag correction on particles is presented below considering the 235U and 238U isotopes (Figure 3). The profile of the isotope ratios as a function of time

Figure 4. Time lags obtained for the 235U/238U ratio measured with C2 detector configuration (IC1/L2) for repeated ablation of IRMM541 under the same conditions. Five to ten ablations were performed for each acquisition.

Figure 3. 238U and 235U signal intensities and point-by-point 235U/238U isotope ratios for an NU particle measured with configuration 2 (238U measured with an FC and 235U with an IC) (a) before correction of the time lag and (c) after correction of the time lag. The resulting LRSs of 235U over 238U are plotted (b) before correction and (d) after correction of the time lag, red dots being ratios corresponding to the rise of the isotope signals and blue dots to the decrease of the signals.

obtained for ablations recorded within the same acquisition (less than 5% RSD). However, the time lags vary significantly from one acquisition to another but not as a function of time. It is assumed that the start of the acquisition has an influence on the time lag but no explanation has so far been found. It should be noted that this experiment was performed with and without tau correction (option available on the MC-ICPMS software), which is usually selected when using FC/FC detectors. When the tau correction was not selected, a similar trend was obtained although two acquisitions produced two outlying values (2 ablations out of 65 presented significantly different withinacquisition time lags). Taking into account time lags measured with and without MC-ICPMS instrumental tau correction, detector time lags for IRMM 541 (37−358 ms) were in the same range as those obtained with particles (58−377 ms). This experiment clearly demonstrates that a fixed time lag correction cannot be applied for all ablation peaks. The calculation of the time lags must be done for each ablation or at least each MC-ICPMS measurement. Analytical Performance with Ion Counter and/or Faraday Cup Detection. In order to compare both configurations (FC/FC and IC/FC), NU particles were ablated and corresponding signals were measured using detector configurations C1 and C2. Each configuration was used on separate days to allow MC-ICPMS optimization. The signals obtained were corrected individually for the yield (gain between IC and FC) and for the time lag. For both configurations, mass bias correction was performed and the associated propagated uncertainties were calculated. As can be seen in Figure 5, the detector time lag correction allowed drastically improved internal and external precision of the 234 U/238U and 235U/238U ratios, when IC/FC configuration (C2) was used. Within our conditions, the benefits of applying the time lag correction with the FC/FC configuration (235U/238U recorded with C1) appeared to be of limited interest with respect to external and internal precision.

without and with time lag correction (Figure 3a and 3c) was in good agreement with the model described above. Without time lag correction (Figure 3a), the isotope ratio value was very high at the beginning of the peak and then decreased to reach a plateau (below the expected value). However, with time lag correction, the isotope ratio was constant over the entire peak width and better stability was obtained. It can be noticed that the corrected signals appear slightly smoother than before the correction. When averaging the interpolated signals to obtain the original number of data points per second, signals from both detectors do not correspond to the same analysis time than before. Therefore, depending on the shift and the data averaged, it can induce a smoothing but this is not necessarily the case. As can be seen in Figure 3b,d, the coefficient of determination was drastically improved (from 0.9919 to 0.9996, i.e., a relative increase of 0.78%). In contrast, the slope of the LRS (i.e., the isotope ratio) was moderately affected by the time lag correction (from 6.78 × 10−3 to 6.81 × 10−3, i.e., a relative increase of 0.44%) even though, the data corresponding to the signal rise provided higher ratios corresponding to the 7 red points above the LRS. Note that no mass bias correction was performed here. Time Lag Variability. Correction of the detector time lags (FC/FC and IC/FC) has been performed specifically for particle analysis. Amplifier time lags (FC/FC detectors) were typically on the order of tenths of milliseconds, which is in good agreement with the study by Gourgiotis et al.8 However, much larger time lag values (ranging between 58 and 377 ms), up to 2 orders of magnitude higher than FC/FC configuration, were obtained for IC/FC configuration. To study this phenomenon in detail, a series of successive ablations of E

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

when switching from one configuration to the other. This was the case for internal precision that varied within a range of 1.1% to 1.8% for both configurations. The external precision was however significantly higher with configuration 2 (2.0% versus 1.0%). The method developed here was dedicated to the determination of isotope ratios of NU, EU, and DU particles. In the context of safeguards, great importance is paid to minor isotopes 234U and 236U. Therefore, the use of ICs was preferred and configuration 2 was used in the next section after time lag correction. Application to Uranium Particles. A total of 59 NU particles were ablated and analyzed over 2 days using configuration 2 to evaluate the potential of this new time lag correction on the measurement of isotope ratios. In total, 20 of the 59 particles were used for mass bias correction (the 5 first and last particles of the sessions). As previously mentioned, diameters of these particles were between 0.1 and 3.5 μm (with most of them ranging from 0.6 to 1.4 μm diameter), which is the typical size range encountered in real-life samples. Therefore, the prepared synthetic sample can be regarded as an acceptable approximation to real-life samples, at least from the particle size point of view. Because of the laser ablation optic system used, particles below 1 μm could not be located and analyzed. Assuming the case of dense spherical particles (density of 10.5 g cm−3 (UO2)), most of the analyzed particles would roughly correspond to 5−200 pg of 238U, 35−1500 fg of 235 U, and 0.3−11 fg of 234U (taking into account 1 and 3.5 μm diameter particles, respectively). This is however an overestimation of the mass since these particles are probably less dense than expected (investigation using a SEM equipped with a focused ion beam suggests that particles contain voids and are more or less porous). Because of this low amount of material, precise isotope ratio determination is challenging, especially when measuring a low-abundance isotope (234U). Figure 6 shows the isotope ratio values and the corresponding internal precisions obtained when applying the time lag correction. It should be highlighted that the sizes of the particles were fairly nonhomogeneous. As a consequence, the signal intensity ranged from 0.04 to 4.48 V for 238U with an average of 0.6 V (over 59 particles). This led to very low signal intensities for the minor isotope. Ablation of the small particles generated very short transient signals lasting on average 10 s (from 2.4 to 25.6 s). Short and low intensity signals are not ideal for high precision isotope ratio measurement. It should be noted that no correlation was found between the accuracy and precision of the isotope ratios and the signal intensity (see Figure S-6 of the Supporting Information). It is assumed that the characteristics of the ablated particle (size, porousness, etc.) are responsible for the important differences in term of intensity, stability, and duration of the signals from one particle to the other. However, even within these unfavorable conditions, external precision of 2.1% was obtained for 235 U/238U. Internal standard deviations were drastically improved by a factor of up to 8.4 and 6.9 for 235U/238U and 234 U/238U, respectively, depending on the particles, when time lag correction was applied. The method enabled good internal precision and accuracy even when very low coefficient of determination was obtained before correction (from R2 of 0.5242 and 235U/238U of (5.60 ± 0.39) × 10−3 to R2 of 0.972 and U235/U238 of (7.10 ± 0.10) × 10−3 after correction).

Figure 5. (a) 235U/238U and (b) 234U/238U isotope ratios obtained after LA-MC-ICPMS of uranium particles using configurations 1 and 2, with and without time lag corrections. Averages ratio and the associated expanded uncertainties (1 SD) are represented in dashed and full lines, respectively.

The time lag correction enabled better external precision (RSD) using configuration 2 from 2.6% to 1.2% for the 235 U/238U ratio and 3.6% to 2.0% for the 234U/238U ratio. In the same way, the internal precision was improved by a factor of up to 4.6. For 234U/238U recorded with IC/FC configuration (C1), the external precision was greatly improved (from 15% to 1% after time lag correction) as well as the internal precision (from 11.5% to 1.8% before correction and 1.5% to 1.8% after time lag correction). Analytical performances with configurations 1 and 2 were roughly similar when time lag correction was applied. However, when considering the 235U/238U ratio, the external precision (RSD) was slightly improved with configuration 1 (0.9% versus 1.2% for configurations 1 and 2, respectively) while the internal precision was significantly better with configuration 2 (between 0.74% and 0.91% with configuration 2 and between 1.2% and 1.7% with configuration 1). This demonstrates that for the analyzed particles (from 1 to 3.5 μm diameter), the use of an IC to measure the 235U provided better internal precision when time lag correction was applied (not true without time lag correction). The signal-to-noise ratio enhancement provided by the IC for measurement of 235U resulted in better counting statistics, thus improving the internal precision of the 235U/238U ratio. Regarding accuracy, the time lag correction had no significant effect when both 235U and 238U were detected on FC as the measured isotope ratios were in good agreement with the expected value whether the time lag correction was applied or not. In contrast, when configuration 1 was used, the corrected values were closer to the expected value ((234U/238U)time lag corrected = (5.46 ± 0.088) × 10−5, versus ( 234 U/ 238 U) before correction = (4.99 ± 0.19) × 10 −5 and (234U/238U)expected= (6.1 ± 1.3) × 10−5). When considering the 234U/238U ratio, both configurations 1 and 2 use FC to measure 238U and IC to measure 234U. No significant difference in terms of performance was expected F

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

the isotope signals over the whole peak is not affected by isotope ratio temporal bias. External precision of 2% was obtained for both 235U/238U and 234U/238U ratios using this method. In contrast, point-by-point data processing was the method most affected by the time lag, as accuracy, external and internal precisions for uncorrected signals were very poor. The high increase of the isotope ratios during the rapid (less than 2 s) signal rise had a great impact here. When the time lag correction was applied, the point-by-point method showed only slightly poorer performance compared to the LRS method (also time lag corrected) in terms of external uncertainty for 235 U/238U (2.6% vs 2.1%) and 234U/238U (2.9% vs 2.5%). However, internal deviations were 1 order of magnitude lower using LRS data processing. In terms of accuracy, very similar results were obtained for 235U/238U and 234U/238U using the three methods, all gave ratios within the uncertainty of the expected value. It should be highlighted that the LRS and integration methods were slightly more accurate for 235U/238U (less than 1% below the true value) and 234U/238U than the point-by-point method. Therefore, the LRS method combined with the time lag correction appears to be the best method for calculation of isotope ratio from short transient signals produced by LA-ICPMS since it provides similar performances (external precision and accuracy) as the integration method with the additional information on the internal precision. Although external repeatability of 0.2% to 1.4% have been obtained for 235U/238U in LA-MC-ICPMS analysis of particles by Lloyd et al.17 and Kappel et al.,18 it is worth mentioning that much larger quantities of uranium were sampled. Kappel et al.18 analyzed glass particles of uranium of 10−20 μm diameters that contain 100−800 pg of uranium. While, Lloyd et al.17 analyzed larger particles (>20 μm) but sampled only the equivalent of 9 μm diameter which corresponded to an amount of 4 ng of uranium. In both cases, larger particles than those analyzed in this study were sampled (1−3.5 μm). The amount of uranium ablated in our study was 20−800 times less (our particles containing 5−200 pg of uranium as previously stated). Smaller particles of 0.3−5 μm diameters were analyzed by Kappel et al.10 with the FC/FC configuration. However, as the particles could not be individually identified with the optical observation

Figure 6. Effect of the detector time lag correction on isotope ratio and internal precision for (a) 235U/238U and (b) 234U/238U ratios.

Comparison with the values obtained using point-by-point and peak integration techniques were also performed. Signals were first background corrected. Then, point-by-point isotope ratios were calculated and averaged from the beginning of the ablation (of the peak) until 95% or 0.01 V (in the case of very low peak intensity) of the 238U signal was reached. The same peak definition was selected for the peak integration method. The sum of the signals was calculated for each isotope. Integrated signals of 234U and 235U were divided by the integrated signal of 238U to obtain the corresponding isotope ratios. Mass bias correction and uncertainty propagation were applied in the same way as for LRS. Table 2 presents the results obtained using the different methods, with and without time lag corrections. The time lag correction was not applied when using the peak integration method since the ratio of the sum of

Table 2. Results of the LA-MC-ICPMS Analysis of 39 NU Particles and the Associated Combined Standard Uncertaintiesa LRS 235

U/238U

average ratio external SD (RSD) R2 internal SD (RSD)

234

U/238U

average ratio external SD (RSD) R2 internal SD (RSD) a

uncorrected 7.09 × 10−3 4.78 × 10−4 (6.7%) from 0.5242 to 0.9913 from 1.73 × 10−4 (2.3%) to 4.08 × 10−4 (7.7%)

uncorrected

point-by-point time lag corrected

uncorrected

7.18 × 10−3 1.53× 10−4 (2.1%)

time lag corrected

8.99 × 10−3 1.45 × 10−2 (161%)

from 0.9719 to 0.99997 from 6.08 × 10−5 (0.9%) to 1.18 × 10−4 (1.7%) LRS

7.16 × 10−3 1.83 × 10−4 (2.6%)

uncorrected

time lag corrected

5.58 × 10−5 3.15 × 10−6 (5.6%)

5.59 × 10−5 1.39 × 10−6 (2.5%)

6.52 × 10−5 9.94 × 10−5 (152%)

5.80 × 10−5 1.68 × 10−6 (2.9%)

from 0.4396 to 0.9853 from 1.83 × 10−6 (3.1%) to 3.95 × 10−6 (8.5%)

from 0.9299 to 0.9997 from 6.46 × 10−7 (1.2%) to 1.33 × 10−6 (2.4%)

from 3.67 × 10−5 (127%) to 3.11 × 10−3 (636%)

from 3.52 × 10−6 (5.8%) to 2.37 × 10−5 (38%)

235

7.19 × 10−3 1.46 × 10−4 (2.0%)

from 2.98 × 10−3 (65%) to from 1.60 × 10−4 (2.2%) to 5.66 × 10−1 (647%) 1.51 × 10−3 (22%) point-by-point

time lag corrected

Expected isotopic ratios are (7.248 ± 0.016) × 10−3 for

peak integration

U/238U and (6.1 ± 1.30) × 10−5 for G

234

peak integration 5.55 × 10−5 1.12 × 10−6 (2.0%)

U/238U (1σ uncertainties). DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry system attached to the LA system (because most of the particles were too small and the surface of the silicon disk used by the author was dark), line or raster scans were applied instead of single particle analysis. In some cases, two or more peak maxima were observed, which most likely corresponded to several adjacent particles entering almost simultaneously into the plasma. The authors obtained a relative expanded uncertainty (k = 2) of 2.3% for the 235U/238U ratio over 118 particles using LRS. Although the study reported that point per point and peak integration methods gave lower relative uncertainties (1.1% and 1.5%), LRS was considered to be advantageous from the statistical point of view and because it allowed access to internal precision.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Phone: +33 5 40 17 50 35. Notes

The authors declare no competing financial interest.

■

REFERENCES

(1) Krupp, E. M.; Pécheyran, C.; Pinaly, H.; Motelica-Heino, M.; Koller, D.; Young, S. M. M.; Brenner, I. B.; Donard, O. F. X. Spectrochim. Acta, Part B 2001, 56, 1233−1240. (2) Krupp, E. M.; Donard, O. F. X. Int. J. Mass Spectrom. 2005, 242, 233−242. (3) Dzurko, M.; Foucher, D.; Hintelmann, H. Anal. Bioanal. Chem. 2009, 393, 345−355. (4) Günther-Leopold, I.; Waldis, J. K.; Wernli, B.; Kopajtic, Z. Int. J. Mass Spectrom. 2005, 242, 197−202. (5) Hirata, T.; Hayano, Y.; Ohno, T. J. Anal. At. Spectrom. 2003, 18, 1283−1288. (6) Pettke, T.; Oberli, F.; Audétat, A.; Wiechert, U.; Harris, C. R.; Heinrich, C. A. J. Anal. At. Spectrom. 2011, 26, 475−492. (7) Epov, V. N.; Rodriguez-Gonzalez, P.; Sonke, J. E.; Tessier, E.; Amouroux, D.; Bourgoin, L. M.; Donard, O. F. X. Anal. Chem. 2008, 80, 3530−3538. (8) Gourgiotis, A.; Bérail, S.; Louvat, P.; Isnard, H.; Moureau, J.; Nonell, A.; Manhès, G.; Birck, J.-L.; Gaillardet, J.; Pécheyran, C.; Chartier, F.; Donard, O. F. X. J. Anal. At. Spectrom. 2014, 29, 1607− 1617. (9) Cottle, J. M.; Horstwood, M. S. A.; Parrish, R. R. J. Anal. At. Spectrom. 2009, 24, 1355−1363. (10) Kappel, S.; Boulyga, S. F.; Dorta, L.; Günther, D.; Hattendorf, B.; Koffler, D.; Laaha, G.; Leisch, F.; Prohaska, T. Anal. Bioanal. Chem. 2013, 405, 2943−2955. (11) Epov, V. N.; Berail, S.; Jimenez-Moreno, M.; Perrot, V.; Pecheyran, C.; Amouroux, D.; Donard, O. F. X. Anal. Chem. 2010, 82, 5652−5662. (12) Rodríguez-Castrillón, J. A.; García-Ruiz, S.; Moldovan, M.; García Alonso, J. I. J. Anal. At. Spectrom. 2012, 27, 611−618. (13) Fietzke, J.; Liebetrau, V.; Günther, D.; Gürs, K.; Hametner, K.; Zumholz, K.; Hansteen, T. H.; Eisenhauer, A. J. Anal. At. Spectrom. 2008, 23, 955−961. (14) Donohue, D. L. Anal. Chem. 2002, 74, 28 A−35 A. (15) Hubert, A.; Claverie, F.; Pécheyran, C.; Pointurier, F. Spectrochim. Acta, Part B 2014, 93, 52−60. (16) Pointurier, F.; Pottin, A.-C.; Hubert, A. Anal. Chem. 2011, 83, 7841−7848. (17) Lloyd, N. S.; Parrish, R. R.; Horstwood, M. S. A.; Chenery, S. R. N. J. Anal. At. Spectrom. 2009, 24, 752−758. (18) Kappel, S.; Boulyga, S. F.; Prohaska, T. J. Environ. Radioact. 2012, 113, 8−15. (19) Ricard, E.; Pécheyran, C.; Sanabria Ortega, G.; Prinzhofer, A.; Donard, O. F. X. Anal. Bioanal. Chem. 2011, 399, 2153−2165. (20) Ballihaut, G.; Claverie, F.; Pécheyran, C.; Mounicou, S.; Grimaud, R.; Lobinski, R. Anal. Chem. 2007, 79, 6874−6880. (21) Vanhaecke, F., Degryse, P., Eds. Isotopic Analysis: Fundamentals and Applications Using ICP-MS; Wiley-VCH Verlag GmbH & Co. KGaA: Weinheim, Germany, 2012. (22) Tushingham, J. W. A. The Preparation of a Low Enriched Uranium Particle Standard for Environmental Analysis. UK Safeguards R&D Program, report SRDP-R259, IAEA Ref. UK A011058, AEAT3300, March 1998.

■

CONCLUSION Ablating single micrometric uranium particles leads to triangular shaped short transient MC-ICPMS signals which are affected by the isotope ratio temporal bias induced by the phenomenon of detector time lag. In this study, we observed and corrected for the difference of time responses between two Faraday detectors and, more importantly, between a Faraday cup and an ion counter detector. The development of a rapid data processing method (1−2 s) based on a VBA macro was used to determine and correct this time lag using two different criteria. The use of the criterion which consists of obtaining a coefficient of determination for the line obtained by plotting data point intensities of one isotope versus the other isotope as close as possible to unity was preferred. The method was successfully applied to the LA-MC-ICPMS analysis of individual micrometric natural uranium particles (diameters between 1 and 3.5 μm) deposited on polycarbonate disks. The time lag correction enables great improvement in accuracy and internal and external precisions when using both Faraday cup and ion counter detectors. The effect of the time lag correction when a Faraday/Faraday detector pair was used was less significant. The combination of the developed time lag correction with the LRS data processing method provided very low internal precisions (0.9−1.7% for the 235U/238U isotope ratio and 1.2% to 2.4% for the 234U/238U isotope ratio). These results are especially noteworthy for the minor 234U isotope which is of very low abundance in the analyzed particles (between 0.3 and 11 fg of 234U per particle). Until now, internal precisions obtained by various authors for the isotopic analysis of individual micrometric uranium particles have been higher than 1% for the 235U/238U ratios and at best, several % for the 234 U/238U ratios, using single or multiple collector ICPMS. This work shows that, provided appropriate corrections are made on the raw signals acquired simultaneously with Faraday cups (for the most abundant isotope, namely, 238U) and ion counters (for the lowest abundance isotopes, here 235U and 234U) of a MCICPMS, internal precisions below 1% on isotope ratios can be obtained for the analysis of micrometer-sized particles (1−3.5 μm diameter). This is of great interest for safeguard purposes, to discriminate between different categories of uranium isotopic enrichment and to access smaller particles.

■

Mathematical demonstration of the isotope ratio temporal bias profile in the case of Gaussian shape signal and triangular shape signals (PDF)

ASSOCIATED CONTENT

S Supporting Information *

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acs.analchem.5b04802. H

DOI: 10.1021/acs.analchem.5b04802 Anal. Chem. XXXX, XXX, XXX−XXX