Multiple ion counting measurement strategies by SIMS – a case study from nuclear safeguards and forensics

A primary tool of nuclear safeguards authorities for detecting undeclared nuclear activities is environmental sample analysis containing uranium-bearing aerosol particles. SIMS (Secondary Ion Mass Spectrometry) has been established as a powerful technique for performing uranium isotope analysis on safeguards samples. Typically, a screening of the sample for locating the particles of interest is followed by micro-beam measurements of individual particles in order to obtain their precise isotopic composition. The uncertainty of these measurements, in particular for the minor isotopes 234 U and 236 U, is mainly limited by the counting statistics as the total amount of uranium available is very low, usually less than 1 pg. This uncertainty can be reduced when recording several isotopes simultaneously using a multi-collection system. This has been available commercially for many years and has been used for di ﬀ erent applications in geosciences on Large Geometry-SIMS (LG-SIMS) instruments. This paper will describe and discuss how isotope measurements of low content uranium materials can be optimized using a multi-ion counting system consisting of ﬁ ve discrete dynode electron multiplier (EM) detectors. The results obtained for uranium isotope ratios are presented for di ﬀ erent measurement protocols: static multi-collector, dynamic multi-collector and single collector measurements. Some of the fundamental properties of the detector system that a ﬀ ect the overall measurement uncertainty are discussed. Best practices are also suggested based on the current experience.


Introduction
When nuclear uranium materials are processed in industrial quantities, it is difficult to avoid the release to the immediate environment of aerosol particles containing the isotopic signature of the handled materials.This allows safeguards authorities with specialized sampling techniques and analytical laboratories to monitor the handling of nuclear materials at nuclear facilities and to perform forensic investigations to nd the origin of materials. 1,2[8][9][10] Recently, the performance of SIMS instrumentation for nuclear particle analysis has greatly improved thanks to the implementation of Large Geometry-SIMS (LG-SIMS), 11 the CAMECA IMS 1280 or 1280-HR, rather than the previously used Small Geometry-SIMS (SG-SIMS), the CAMECA IMS xF series.
LG-SIMS instruments feature numerous instrumental advantages, in particular removal of background interferences with minimal loss of transmission and parallel detection of all U isotopes using the multi-collection system.In addition, the productivity has signicantly increased with the implementation of automated particle measurement (APM) soware that performs fast screening of the sample to determine the exact location of the uranium particles and their enrichment level. 12,13n this paper we describe the performance obtained using the multi-ion counting system equipped with EM detectors for isotopic micro-beam measurements on individual particles.The use of a multi-ion counting system for the APM screening measurements has been described earlier and is not further discussed here. 12Some of the basic parameters of the EM detectors like HV adjustment, detector dri and aging effects, as well as detector inter-calibration methods are discussed in the context of improving the isotope measurement quality.Performance obtained using different measurement protocols (static multi-collection, dynamic multi-collection and monocollection) is compared for uranium particle analysis, and advantages and disadvantages of each method are discussed.
Though discussions in this paper are focused on the nuclear safeguards and forensics application, the ndings regarding the EM detector behaviour are also valid for other elds of work (e.g. in geo/cosmochemistry) and for other types of instruments (e.g.single or multi-collector TIMS and ICPMS -Inductively Coupled Plasma Mass Spectrometry).

A. Instrumentation
The LG-SIMS measurements were performed on a CAMECA IMS 1280-HR at the Institute for Transuranium Elements (ITU) in Karlsruhe, Germany.The IMS 1280-HR is a large geometry double focusing magnetic sector secondary ion mass spectrometer (magnet radius of 585 mm).The instrument can operate in both microprobe and microscope modes, since it provides scanning ion images as well as mass ltered stigmatic ion images.
The instrument is equipped with a multi-collection detector system, which allows simultaneous isotope measurements for elements from lithium up to uranium, see Fig. 1.This system consists of ve moveable collector units (L2, L1, C, H1, and H2).Each unit is equipped with a set of three slits for adjusting the mass resolution conditions, and with either a discrete dynode electron multiplier (EM) or a Faraday cup (FC) detector for higher signal intensities.For all measurements reported in this paper, the ve collector units were equipped with EM detectors (Hamamatsu R4146).A voltage was applied to the DSP2 stigmator aer the magnet to increase the mass dispersion in order to measure the 234 U, 235 U, 236 U, 238 U and 238 U 1 H species simultaneously on the ve multi-collection EM detectors.The instrument is also equipped with a mono-collection detector system with two FCs and one EM detector (AF150H from ETP Electron Multipliers).
Previous studies on U particle analyses have shown that the ion yield is higher using negative oxygen primary ion beam (O À or O 2

À
) compared to positive oxygen primary mode (O 2 + ). 11 However, positive mode has been selected for the present measurements because of the higher primary beam density and current stability.The selected operating conditions for the measurements are shown in Table 1.

B. Reference materials
Uranium Certied Reference Materials (CRM) U010 and U100 from New Brunswick Laboratory (DOE, USA) were used for mass calibration and mass bias determinations, as well as for determining the precision and accuracy of uranium isotope measurements. 235U enrichment is $1 atom% for CRM 010 and $10 atom% for CRM U100. 234U and 236 U abundances for CRM U100 are higher than that for CRM 010 (more than 10 times for 234 U, and more than 5 times for 236 U). 14 The standards were prepared in a clean room by transferring the particles onto a planchet (SPI Pyrolitic Graphite Planchet, 25.4 Â 3.2 mm, #428GP-AB) using a vacuum impactor. 15 Isotopic micro-beam measurements of uranium particles Micro-beam measurements were performed on individual particles using a small focused primary beam rastered over a 10 Â 10 mm area around each particle.Uranium isotope ratios are corrected for detector dead time and mass bias.A hydride correction is also applied to the 236 U/ 238 U ratio.11 The detector dead time was measured by setting the preamplier threshold to zero which saturates the pulse counting system as all noise pulses are counted.The inverse of this maximum count rate gives an estimate of the dead time value, see values in Table 2.
Fig. 1 CAMECA IMS 1280-HR ion optics, a projection area with a detection system after the magnet.DSP2 is an octopole device that is used both as a stigmator and a deflector.It can be used to increase the mass dispersion.The detector dark noise was measured at mass 5 for a period of 60 minutes.The dark noise was in the range of 2.8 Â 10 À4 count per s to 3.3 Â 10 À3 count per s for the 5 Hamamatsu detectors.The single collection EM had a dark noise of 2 Â 10 À3 count per s.For the measurements reported in this paper there was no subtraction of detector dark noise.
Multi-collector measurements are also corrected for the detector yield values reported in Table 2, whereas a dri time correction is applied to single collector measurements. 16n this paper, uranium isotope data are presented in atom%, as commonly used in safeguards applications.
Three different measurement protocols have been investigated in this paper.The selected settings for these measurements are described in Table 3: (a) Peak jump (PJ) mono-collection measurements where 234 U, 235 U, 236 U, 238 U and 238 U 1 H species are recorded sequentially on the EM ETP detector.The mass peak switching is performed by changing the magnetic eld.
(c) Dynamic multi-collection measurements consisting of simultaneous isotope detection as in (b), combined with magnetic peak jumping for measuring 235 U and 238 U species using the same detector (L1), see Fig. 2.

D. Fundamentals of the EM detection system
EM detectors work in a direct pulse counting mode.A secondary ion striking the rst dynode (conversion dynode) of the EM induces secondary electron emission.These electrons are then accelerated and multiplied through the successive dynode stages of the EM.The gain of the detector is dened as the mean number of output electrons per secondary ion (typically in the range of 10 8 ).For most of the secondary ions reaching the detector, a charge pulse is produced at the last dynode output.A preamplier converts the charge pulses into voltage pulses and amplies them.Then a discriminator compares the amplied voltage pulses to a given detection threshold and rejects the smaller pulses to cut off the system noise.The amplitude of the output voltage pulses is not constant but distributed according to a pulse-height distribution (PHD), see Fig. 3.
It is known that for EM detectors, the gain of the detector can be affected by short term dri issues and long term aging phenomena.Below is a list of three main components of the detector dri and aging: (1) Long term aging of the last dynodes that needs to be compensated for by progressively increasing the EM high voltage throughout the EM's lifetime in order to keep the gain constant.The lifetime of an electron multiplier depends on the total number of ions collected (total integrated charge), therefore, frequent high intensity measurements shorten the EM lifetime.The vacuum environment also plays a major role in determining the overall life of the detector.This deterioration is known to be caused by carbon deposition on the last dynodes. 172) Long term deterioration of the rst dynode due to deposition of impinging ions, see Fig. 4 and 5.As the detector gets older, this effect can be observed by scanning the ion beam along X and Y directions over the detector rst dynode, as the gain becomes lower at the usual position where the ion beam hits the detector.This effect is reduced in time if the impinging  ion beam is enlarged or scanned over the rst dynode. 18In our case, aer a deterioration was observed, the lifetime of the detectors has been increased by shiing the ion beam along Y so that it hits a fresh area on the rst dynode, see Fig. 6.This can be done by applying a small offset along the Y direction using main deector DSP2.This brings back a correct mass peak atness along X and also slightly improves the detector yield, resulting in a lower EM HV.
(3) Short term dri behaviour where the detector yield changes depending on the intensity of the applied ion beam and dris back to its original position if le without an ion beam.Interestingly, this dri is in the opposite direction compared to the long term aging effect.As this effect is rarely discussed or described, particular attention is given to it in this paper.A previous study showed that detector non-linearity is affected when switching a high intensity beam into a detector used for lower intensity measurements. 19This could possibly be related to short term dri issues.Examples are shown below of different short term dri measurements.Experience shows that the amplitude and duration of the short term dri depends on the detector age.When the detector is new, it shows a fairly high degree of short term yield dri, and its stability improves as it gets older.When a highly stable detector is required one can age a new detector by sending a high beam dose over the rst dynode until it stabilises.
In CAMECA instruments, a quick iterative automated routine is used to adjust the EM HV in order to keep the PHD curve within a given template.Practically, the routine does not process the PHD curve, but adjusts EM HV so that when the EM threshold is set to Thr2, typically 250 mV, the signal is half of the signal obtained using Thr1, typically 50 mV, see Fig. 3. Generally, Thr1 corresponds to the working threshold for removing noise pulses and is kept xed all along the EM lifetime.5 LG-SIMS 238 U scanning ion images on deposition area 1 and deposition area 2 shown in Fig. 4, showing that the depositions contain uranium.This automatic HV adjustment routine was recently included in the analysis acquisition soware, and thus can be performed at the start of each micro-beam measurement.The rst results obtained using this new soware capability are shown in Section B.
For uranium measurements using a high sample extraction voltage (8 kV for the present measurements), new multi-collection Hamamatsu detectors typically start with a voltage at about 1600 V which gradually increases up to about 2200-2400 V.The starting voltage for the mono-collection ETP detector is typically higher around 1850 V.The EM HV values used for the present measurements are shown in Table 2.

Results and discussion
A. Short term dri measurements Fig. 7 shows a short term dri pattern recorded for the single ETP EM using the automatic EM high voltage (EM HV) routine.The voltage values are the result of repeated EM HV adjustments.The dri pattern was measured aer a period of about 14 hours with no signal on the detector.At rst, there is a rapid detector dri corresponding to a voltage drop, and then the detector stabilises.The detector dris back within about an hour to the starting HV value aer the ion beam is switched off.
A similar test is shown in Fig. 8 for a new Hamamatsu detector, using two different ion beam intensities.The results show that the short term dri increases with the ion beam intensity.
The detector dri can be measured more precisely by physically connecting the detector to two pulse counting channels in parallel.One channel is set to the normal 50 mV threshold (Thr1) and the second channel set at 250 mV (Thr2).The ratio of the intensities I Thr1 /I Thr2 provides a continuous measurement of the PHD shi, similarly to what is performed by the CAMECA EM HV adjustment routine.For short term dri, the threshold ratio usually drops below the expected value around 2, which means the pulse height distribution moves towards high amplitude values.This is equivalent to an EM HV increase so that the short term dri must be compensated by decreasing the EM voltage, as shown in Fig. 7 and 8.
In Fig. 9, measurements using parallel counting channels with different thresholds have been performed on two detectors, C (almost new) and H1 (signicantly aged).The dri measurements show that the new detector (C) dris signicantly more and for a longer period of time compared to the older one (H1).Fig. 10 shows similar dri measurements performed on detectors L1 ( 235 U) and H1 ( 238 U) while simultaneously measuring the isotopic ratio 235 U/ 238 U on CRM U100, as in Fig. 11.Fig. 10 shows that detector L1 dris more than H1 despite a signal 10 times lower ( 238 U count rate of $300 kcts s À1 ).Fig. 7 Drift pattern obtained by performing repeated EM HV adjustments for the ETP detector of the mono-collection system, using a 238 U ion beam intensity of 325 kcts s À1 .The detector stabilises after 80 s and a voltage decrease of 19 V.This detector has gone through a significant aging thus the voltage is rather high.A linear t to the 235 U/ 238 U data in Fig. 11 indicates however that the isotope ratio dri is small, not larger than $0.5& during the 30 minutes of measurement.A long series of measurements would require repeated EM HV adjustments in-between the measurements to compensate for the detector dri.This has been conrmed by a 10 hour long measurement session on the CRM U100 sample, using the auto EM HV adjust function for detector H1 ( 238 U, count rate of $130 kcts s À1 ) but not for L1 ( 235 U).This set of measurements showed an overall increase in the 235 U/ 238 U ratio of 3& over the total 10 hours.
It is important to take these short term dri patterns into account, in particular when performing multi-collection analyses.For uranium measurements one should avoid putting high intensity beams on the minor isotope detectors prior to sample measurements as this will push the detector away from its normal working yield.The C detector which presents a large dri (Fig. 9) can be used for the measurement of the 236 U minor isotope, as the dri is expected to be negligible for such a low intensity signal.The detector used for the 238 U measurements has a very good stability making it well suitable for the measurements at high intensities.The dri patterns have also to be taken into account for selecting the method for detector yield inter-calibration.As an example, sequentially measuring a given high intensity ion beam on the different detectors does not provide accurate results due to the large difference in ion beam intensities when performing real U isotope measurements.A recommended method for detector yield inter-calibration is described in Section B.2.

B. Isotope measurements on CRM samples
Sets of uranium isotope ratio measurements on standard particles CRM U010 and CRM U100 are described below.These measurements are well suited for comparing the performance of different analytical protocols as the particles are large and free of any matrix materials.
For all results, "Mean" is the average isotope abundance obtained from many analyses, each analysis being performed on a different particle."SD" is the absolute standard deviation, RSD% the relative standard deviation (in percent), and "Bias%" the deviation (in percent) from the certied values.
B.1.PJ mono-collection measurements.For the single collector measurements, 234 U, 235 U, 236 U, 238 U and 238 U 1 H species were recorded in a peak jump sequence, see Table 3. Measurements were performed at count rates of 100-350 kcts s À1 on 238 U.The results are shown in Table 4 for CRM U010 and CRM U100 samples.Data are corrected for a mass bias of À0.00296 per amu.The automatic EM HV adjust routine was not used here because all species are measured on the same detector thus the ratios are not affected by detector yield variations.
As expected, the measurement precision on the CRM U100 is better than on the CRM U010 due to the higher counting rates for 234 U, 235 U, and 236 U and thus improved Poisson statistics.The measurement uncertainty is mainly determined by the combination of the Poisson statistics and the dri time correction for the 235 U measurement.A possible improvement could be foreseen by applying fast electrostatic mass peak switching, in order to reduce the uncertainty component from the dri time correction.
The mono-collection PJ protocol provides good quality data as long as the particles are large enough and the primary beam remains reasonably stable.It should also be pointed out that it is the simplest method to implement.B.2. Static multi-collection measurements.In these measurements, all isotopes were recorded simultaneously using the multi-collection system, see Table 3 and Fig. 2. Measurements were performed at count rates of 100-350 kcts s À1 on 238 U. Data are corrected for a mass bias of À0.00232 per amu.
The detector yield for L2 ( 234 U), L1 ( 235 U) and C ( 236 U) versus H1 ( 238 U) was derived from a set of CRM U100 measurements by matching the measured values to the nominal isotope ratios.The H2 detector yield ( 238 U 1 H) versus H1 was derived from a peak jump measurement using a count rate of about 50 kcts.It is important to have the ion beam hitting the rst dynode at the same geometrical position for all detectors during these measurements, as any shi will change the relative detector yield.The results of the measurements are shown in Table 5 for CRM U010 and CRM U100 samples.For CRM U010 measurements, the automatic EM HV adjustment routine was applied on detector H1 ( 238 U).This routine was also applied for CRM U100 measurements, but in this case the auto HV adjust was performed every h analysis on detector L1 ( 235 U) because of the higher 235 U count rate for this sample ($10 times higher).In fact, a series of preliminary measurements performed during a time period of 10 hours on this sample showed that without EM HV adjustment a dri of the L1 detector is observed (see also Section A and Fig. 10 and 11).
The results show that the use of static multi-collector measurements improves the measurement precision compared to single collector measurements, which is mainly due to the improved Poisson statistics (for a similar analysis duration).The auto HV adjust function works well but should be improved in order to allow adjustment of more than one detector for each analysis, typically detectors L1 ( 235 U) and H1 ( 238 U) that receive most intense signals.
It should be noted that these measurements were performed over short periods of time, thus uncertainties are not affected by long term variations.
B.3.Dynamic multi-collection measurements.For these measurements, all isotopes were measured simultaneously in a rst sequence followed by a peak jump in order to measure the 235 U and 238 U signals using the same detector (L1), see Table 3 and Fig. 2. Measurements were performed at count rates of 100-350 kcts s À1 on 238 U. Data are corrected for a mass bias of À0.00232 per amu.The automatic EM HV adjust routine was applied for all measurements on detector H1 ( 238 U).The results are shown in Table 6 for CRM U010 and CRM U100 samples.These data demonstrate that excellent measurement precision can be maintained over long periods of time for 234 U, 235 U and 236 U.The precision level for 234 U and 236 U measurements is slightly worse compared to the static multi-collection protocol but better than that for single collector measurements (for the same analysis duration).This method has the advantage that the 235 U/ 238 U ratio is not affected by detector yield variations since the two isotopes are measured on the same detector.
The dynamic measurement protocol also provides a continuous monitoring of the mass bias stability, whereas separate mass bias measurements are required for static multi-collection mode.

C. Isotope measurements on real samples
Interlaboratory comparisons have shown that the results obtained for uranium isotope measurements on standard U particle materials vary not only between different measurement techniques but also between laboratories using the same technique. 20One reason for the spread in quality between laboratories using the same technique could be related to the handling of some of the detector issues presented in this paper.
Measurements on inspection samples at ITU using the SIMS dynamic multi-collection method described above show that measurement uncertainties (1SD) on samples with large particles (micrometers in diameter) are in the range of $0.1-0.3% for 235 U atom% and $0.4-3% for 234 U atom%, in an enrichment range from depleted to 20%.The detection limit for the 236 U atom% depends on the enrichment level due to the hydride correction; detection levels of 1-20 ppm (3SD) are typically obtained in an enrichment range from depleted to 20%.It must be pointed out that the uncertainties obtained on real U particle samples vary signicantly from sample to sample as the particle sizes are not identical.

Conclusions
Uranium isotope measurements are challenging because of the large difference in isotopic abundance, and also due to the demand of high precision and accuracy on 235 U as well as on the minor isotope measurements.Parallel detection using multi-collection is preferred to sequential mono-collection mode as the uncertainty determined by Poisson (counting) statistics is signicantly reduced for the same analysis time.It should be noted that for particle analysis, the analysis duration is limited by the available particle volume.Therefore, the multi-collection mode provides much better precision for small particles as it maximizes the number of ions collected before the particle is sputtered away.
However, precision using multi-collection may be limited by detector dri and aging effects that require inter-calibration of the detector yields.A recent improvement in controlling the detector dri is the implementation of the automated EM high voltage adjustment during analysis acquisition.
A compromise between the single collector measurements and the static multi-collector measurements is the use of dynamic multi-collector measurements that provides high precision measurements of the 235 U/ 238 U ratio at levels of 1& in a robust way.The results shown in this paper demonstrate that these measurements provide highly precise and accurate measurements for all U isotopes.
The static multi-collector measurement protocol has the potential to provide the lowest uncertainties (close to the Poisson statistics) even at high signal intensities if detector dri issues and yield variations can be well controlled and/or corrected.The long term aging and short term dri issues can be handled reasonably well if detectors are sufficiently stable (having gone through some aging); the EM HV adjust is used frequently and detectors are not stressed with high intensities beams prior to measurements at lower intensities.Another problem however is the deterioration of the rst dynode with time which in our case seems to be related to the accumulation of uranium on the dynode surface.This is an issue as the detector yield then becomes very sensitive to the precise ion beam shape and position on the rst dynode.Defocusing or scanning the beam before it hits the rst dynode might be helpful for minimizing the rst dynode deterioration; however one would need to carefully verify that this does not affect measurement precision or accuracy.
Studies on EM detector dri effects and suggested measurement protocols based on current experience are presented in this paper.Further investigations are needed to better understand the causes of the short term dri and rst dynode deterioration in order to more precisely control or compensate for these effects.

Fig. 2
Fig. 2 For static multi-collection measurements only sequence 1 is performed.For dynamic multi-collection measurements, the 234 U/ 238 U, 236 U/ 238 U and 238 U 1 H/ 238 U isotope ratios are computed from signals measured during sequence 1, whereas the 235 U/ 238 U ratio is obtained from signals recorded on L1 detector, sequences 1 ( 235 U) and 2 ( 238 U).

Fig. 4
Fig. 4 Scanning electron microscope image on the first dynode of a Hamamatsu detector.Two dark areas of approximately 300-400 mm in diameter are observed, most likely a slight charging effect due to a deposition of material that is less conductive.Two different ion beam Y impact positions have been used on this detector, which explains the presence of two deposition areas.The horizontal axis of this image corresponds to the Y direction of the detector.

Fig.
Fig.5LG-SIMS 238 U scanning ion images on deposition area 1 and deposition area 2 shown in Fig.4, showing that the depositions contain uranium.

Fig. 6
Fig. 6 Mass scan at mass 238 U with and without an offset on the DSP2 deflector, which corresponds to two different Y positions for the ion beam impact on the first dynode.A significantly better peak flatness is obtained using a new impact position (with DSP2 offset).

Fig. 3
Fig. 3 Typical detector pulse height distribution (PHD) from a Hamamatsu EM detector.Thr1 and Thr2 pulse amplitude threshold values used by the EM HV routine are typically set at 50 mV and 250 mV, respectively.

Fig. 8
Fig. 8 Drift patterns for an almost new Hamamatsu detector at two different ion beam intensities.The first drift curve is obtained using 670 kcts s À1 and the second with 170 kcts s À1 .The time between the two measurements (with no impinging ion beam) allowed the detector to drift back to its previous voltage.

Fig. 9
Fig. 9 Drift patterns for detectors C and H1 obtained from the intensity ratio for two different threshold values I 50 mV /I 250 mV , using a 238 U count rate around 300 kcts s À1 for both detectors.The intensity ratio of detector C can be brought back to its original value by reducing the EM voltage of about $35 V.

Fig. 11
Fig. 11 Relative 235 U/ 238 U deviation from the average value in multicollection ratio measurement performed while recording the detector drift patterns shown in Fig. 10.Each data point corresponds to a measurement of 4 seconds.

Table 3
Time settings for the different measurement protocols.The total acquisition time includes the waiting times required for magnetic field stabilisation 234U ¼ 4 s, 235 U ¼ 2 s, 236 U ¼ 4 s, 238 U ¼ 2 s, and 238 U 1 H ¼ 2 s Total acquisition time ¼ 768 s (40 cycles) Static multi-collection measurements Integration time ¼ 4 s Total acquisition time ¼ 705 s (167 cycles) Dynamic multi-collection measurements Integration time for sequence 1 ¼ 4 s and for sequence 2 ¼ 1 s Total acquisition time ¼ 710 s (100 cycles)

Table 4
Results in atom% from 41 measurements on CRM U010 and 54 measurements on CRM U100 using a single ETP detector.The measurements were made over a period of 2 weeks (U010) and 1 month (U100)

Table 6
Results in atom% from 68 measurements on CRM U010 and 43 measurements on CRM U100.The measurements were made over a period of 2 months