Comparison of two ICP-MS set-ups for measuring ⁹⁹Tc in large volume water samples

Abstract

Large volume fjord and seawater samples have been radiochemically prepared for ICP-MS analysis in order to test the robustness of the procedure and to carry out a comparison of two ICP-MS set-ups. A sector field instrument (MicroMass PT2) coupled with an ultrasonic nebuliser and a quadrupole ICP-MS (Perkin-Elmer Elan 6000) coupled with an electrothermal vaporisation (ETV) unit were used. The results showed that the radiochemical procedure was robust, removing Ru and Mo to acceptable levels, and that the two set-ups gave results that were in agreement. The correlation coefficient between the sets of 11 results was 1.0 ± 0.05. The importance of establishing the matrix effect when using an ETV is discussed.

---

Introduction

Technetium-99 is a long-lived (2.1 × 10⁵ years) beta-emitting radionuclide, present in the European seas primarily from reprocessing effluents discharged from Sellafield, UK, and Cap de la Hague, France. There is widespread interest in measuring Tc in marine and estuarine waters, in order to utilise it as a marine tracer and to understand its behaviour in the environment. Technetium-99 is currently measured in most environmental samples by beta-counting, following prior pre-concentration and removal of all other beta-emitting nuclides from the sample, since beta-counting is not nuclide-specific. Because of the low concentrations of ⁹⁹Tc in seawater, ⁹⁹Tc may need to be isolated from initial sample volumes of 500 l¹ when beta-counting is used. Therefore, although beta-analysis has a reasonable detection limit (around 3 mBq (5 pg)¹) and has yielded a large amount of useful data, there is a drive to use alternative methods with lower detection limits and without the uncertainty associated with completely removing all other beta-emitting nuclides from every sample.

Different types of mass spectrometry have been used to measure Tc, and the results have been generally very good. TIMS and accelerator mass spectrometry (AMS) have shown low instrumental detection limits, at approximately 10–20 fg,^2,3 but most work has been carried out using inductively coupled plasma mass spectrometry (ICP-MS), because of the relative availability of these instruments and the lower cost of running samples. Detection limits for ICP-MS range considerably according to the age and design of the instrument, the sample introduction system used and, possibly, the calculation used. Generally, quotes lie in the range from 3 pg l⁻¹(ref. 4) to 0.6 ng l⁻¹(ref. 5 and 6), which correspond to absolute detection limits of 0.02 mBq (30 fg)–3.8 mBq (6 pg), assuming that the samples were prepared in 10 ml where appropriate. Significantly better detection limits have been reported by Becker and Dietze⁷ using a sector field instrument with a torch shield and a MicroMist microconcentric nebuliser with a minicyclonic spray chamber of 2.5 fg. However, despite the number of papers reporting measurements of pure Tc solutions or high specific activity samples, there are very few which describe the use of the technique for the routine analysis of low-level samples and, moreover, those prepared from large initial volumes.

The main interferences in the mass spectra of ⁹⁹Tc are ⁹⁹Ru and, to some extent, ⁹⁸MoH, ⁵⁹Co⁴⁰Ar, ⁸⁷Sr¹²C, ⁸⁷Rb¹²C, ⁴³Ca¹⁶O⁴⁰Ar, ⁴⁰Ar₂¹⁸OH and ⁴⁰Ca¹⁸OH⁴⁰Ar. These can be removed in the chemical separation procedure, for example, by using TEVA.Spec resin (Eichrom Industries), but when working with large samples and using large amounts of reagents, absolute clean up can be difficult. Some ⁹⁹Ru can be tolerated, as it can be accounted for by measuring at m/z 101 for ¹⁰¹Ru; m/z 101 experiences very little, if any, interference and so this is a reasonable correction. However, despite being able to account for Ru, its presence affects the minimum detectable activity of ⁹⁹Tc in a sample, especially when working with the most sensitive techniques.⁸ Other potential problems involve the use of ^99mTc as the radiochemical yield monitor for the separations procedure, since low levels of ⁹⁹Tc build up in the ⁹⁹Mo generator and are carried into the yield monitor solution unless the generator is washed carefully prior to yield monitor elution.⁹ The problem increases when analysing low-level, large volume samples, since the complicated separation procedure often requires the addition of multiple aliquots. There is no stable isotope of Tc to use as an internal standard in mass spectrometry, but suitable elements include Rh⁸ and Nb,³ which can be selected according to the chemical composition of the sample. For example, when analysing seawater, Nb may not be favoured because of the Nb content of the samples. Beals⁵ proposed use of the long-lived ⁹⁷Tc isotope as both a yield monitor and internal standard, but difficulties arise in obtaining high purity ⁹⁷Tc at reasonable expense, as well as the isobaric interference from ⁹⁷Mo. Additionally, in our laboratory we have observed an unidentified polyatomic interference at m/z 97.

In this study, we compare two different ICP-MS set-ups for measuring Tc isolated from large volume (5–50 l) water samples, taken from a fjord in southern Norway and the Kattegat. The Tc concentrations in these waters are low (from 80 fg l⁻¹) and so the Tc must be pre-concentrated from the large, bulk matrix, and separated from potential interferences in order to carry out ICP-MS analysis. This study allows us to assess the suitability of a chemical procedure and the capability of both set-ups for analysing samples prepared from ‘difficult’ matrices. The ICP-MS set-ups used were a MicroMass PT2 ICP-sector field mass spectrometer (SFMS), or high resolution ICP-MS, coupled with an ultrasonic nebuliser (USN), and an Elan 6000 quadrupole ICP-MS coupled with an electrothermal vaporisation unit (ETV). The different sample introduction systems used gives the Elan 6000 an advantage, since up to 80% of the sample reaches the plasma as opposed to approximately 20% via an ultrasonic nebuliser.¹⁰ However, the ICP-SFMS magnet is too slow to scan in conjunction with the short, transient signal obtained from an ETV, meaning that both instruments are running at their, theoretically, most sensitive configurations with the introduction systems we had available.

Methods

Radiochemical procedure

All acids used were of analytical reagent grade and distilled nitric acid was used in the procedure for the TEVA.Spec columns and subsequent stages.

Technetium-99m (t_½ = 6.01 h) was used as the yield monitor during the chemical separations. A ⁹⁹Mo (t_½ = 2.75 d) generator was obtained from the isotope laboratory at Risø National Laboratory, and was washed thoroughly before the actual yield monitor solution was eluted with 0.9% NaCl. The washing consisted of five 24 ml rinses of 0.9% NaCl on obtaining the generator and on every Monday morning, followed by five washes prior to every yield monitor elution. The in-growth time between the final wash and the yield monitor elution was increased as the generator decayed, so as to collect sufficient ^99mTc. An aliquot of the yield monitor solution was gamma-counted three times in standard geometry in a well detector, and then the samples were spiked with aliquots of the same solution, approximately 1 kBq sample⁻¹. A portion equivalent to 20 kBq of each ^99mTc solution was prepared for analysis by ICP-MS to assess the ⁹⁹Tc content of the yield monitor itself. 10 ml of 0.1 M HNO₃ was added to the yield monitor solution and this was applied to a 0.5 ml TEVA.Spec column. The column was washed with 10 ml 2 M HNO₃ to remove the NaCl and the Tc was eluted with 4 ml 8 M HNO₃. The elution was gamma-counted in standard geometry and the yield calculated.

The water samples analysed were of different compositions, ranging from oxic seawater through to anoxic waters. Prior to analysis, the anoxic samples were thoroughly flushed with air until all smell of sulfide had been removed. The chemistry of these samples meant that it was necessary to break down any organic complexes or precipitate phases, to release the entrained Tc into the solution phase as the oxidised TcO₄⁻ ion. The first step was therefore to acidify the sample with 14.4 M HNO₃ to a pH less than 1.5 and to heat it for 2 h to >90 °C. This was carried out in a 70 l round-bottomed flask using a cluster of large immersion heaters. If the sample was less than 50 l, deionised water was added so that the heating elements of the immersion heaters were immersed. During storage, FeO_x(OH)_y had precipitated onto the walls of some sample containers, and so 12 M HCl was used to dissolve any precipitate, and this was added to the rest of the sample prior to heating. The ^99mTc yield monitor was added during heating, ensuring that the yield monitor was left to equilibrate for at least 30 min with agitation. After heating, saturated NaOH solution was added to precipitate Mg(OH)₂ (pH ≈ 10), and to remove the precipitable components of the sample. Tc, as TcO₄⁻, does not co-precipitate, although small losses occur from the solution remaining with the precipitate. The supernatant was siphoned off through a paper filter, acidified to pH 6 to prevent further precipitation, which may block the column, and then pumped through a BioRad AG1-X4 anion exchange column¹ (see Table 1 for column volume and flow rate for the different sample volumes; flow rates were set so that the samples were pumped through the columns overnight, this is achievable, convenient and gives a high uptake of TcO₄⁻ onto the resin). The columns were washed with five column volumes of 0.5 M HNO₃ and then the Tc eluted with five column volumes of 10 M HNO₃.

Table 1 Column volumes of BioRad AG1-X4 and flow rates used for different initial sample volumes of sea- or fjord water

	Initial sample volume/l
	5	25	50
Volume of BioRad AG1-X4/cm^3	21	59	79
Flow rate/ml min^−1	6	30	55

---

In the method development stage we found that the BioRad resin contained a large amount of Ru, which bleeds from the resin when concentrated nitric acids are applied to it. Since Ru is the major interference in mass spectrometry, it was necessary to pretreat the resin with 14.4 M HNO₃, leave it to stand for 2 h, and wash it thoroughly with nitric acid and deionised water to remove most (approximately 95%) of the Ru. Using 10 M rather than 14.4 M HNO₃ to elute the Tc also reduces the amount of Ru leached by a factor of approximately 15. During the course of sample preparation, an additional wash of a solution of NaOH, EDTA and NaClO was used for some samples. This is routinely used at Risø to remove radioactive Ru isotopes prior to Tc elution and so seemed a quick way of achieving Ru decontamination prior to the final clean up, described below. However, the samples prepared in this way actually showed stable Ru levels enhanced by a factor of approximately 30, seriously affecting the uncertainty on the Tc measurements. It was therefore decided that the BioRad column should simply be used as a pre-concentration step. It is important not to use only radioactive Ru to test for Ru decontamination in method development for ICP-MS as it overlooks any contamination from the reagents and laboratory.

The 10 M HNO₃ elutions from the Bio-Rad column were heated down to 0.25–0.5 ml and then 25 ml deionised water (18.2 MΩ) was added. The solution was applied to a 2 ml TEVA.Spec column, which was then washed with 40 ml 2 M HNO₃ to remove Ru and Mo, and finally the Tc was eluted with 7 ml 8 M HNO₃, as described by Tagami and Uchida.¹¹ When the initial volume of the samples was between 25 and 50 l, this step was repeated. Since ^99mTc has a short half-life (6.01 h), it was necessary to use two successive freshly prepared ^99mTc yield monitor aliquots in the analysis of the 25–50 l samples. The first yield monitor was used for the pre-concentration stage, and the second for the TEVA.Spec columns, and sufficient time was left between these steps for the first yield monitor to decay fully. The final elution of Tc was evaporated down and prepared in standard geometry for gamma-counting. The yield of the procedure was calculated using the radioactive decay equation to decay-correct, and comparing the activity of the remaining yield monitor at time zero with the activity added. The samples were then prepared for ICP-MS (see below).

Samples

Two sets of samples were used in this study, all of which were prepared using the radiochemical procedure described above and were then stored in the 5 ml gamma-counting standard geometry used to obtain the procedural yield. The first set consisted of 11 water samples from two stratified fjords, and the Tc fractions were initially prepared for analysis on the Elan 6000 with ETV, as described by Skipperud et al.¹² They were evaporated very slowly to dryness at <80 °C in AA cups, then 100 μl of 1% HNO₃ was added to each and they were left to stand for 1 h. Finally, 100 μl of 1 M NH₃OH was added and the sample was agitated to ensure thorough mixing and dissolution of the sample. The total volume was therefore only 200 μl, maximising the amount of sample Tc reaching the plasma per 20 μl injection. Smaller total volumes can be used, but they are not very convenient to handle and may make it difficult for the autosampler to collect all three aliquots accurately. The resultant solution is alkaline, which helps stabilise the volatility of Tc, minimising losses in the drying stages of the heating program prior to the vaporisation temperature of 2600 °C.¹² Three 20 μl aliquots were analysed on the Elan 6000 with ETV. The remaining 140 μl of the samples were dried slowly at <80 °C, redissolved in 10 ml of 1% HNO₃ with 50 ng l⁻¹¹⁰³Rh internal standard, and measured on the PT2 with USN.

An internal standard is not necessary when using the Elan 6000 with ETV as the signal is stable, and three replicates are analysed to account for any small variation. However, when using the ICP-SFMS with USN, the signal noise is greater than with the standard pneumatic sample introduction, and the signal can decrease over time as particulate matter builds up on the cones. Furthermore only one replicate is analysed, so the internal standard is very important. ¹⁰³Rh, as suggested by Fifield et al.,⁸ was selected as the best internal standard for these samples, because it fulfils all the necessary criteria; it is close to Tc in both mass and first ionisation potential (7.46 vs. 7.28 eV), it has a minimal memory effect, the background at m/z 103 is very low and there are very few interferences, and our samples contained very little Rh. Moreover, each TEVA.Spec column was found to remove 98% of the Rh present, so by adding a significant amount of internal standard (50 ng l⁻¹) there will be very little error, either from additional Rh from within the sample or from measurement error.

The second set of samples consisted of five sub-samples (5 l) of Kattegat seawater. These were prepared radiochemically and analysed on the two set-ups to assess the reproducibility of the total analytical procedure and to establish matrix effects on both set-ups. Although this seawater is not directly representative of the fjord sample compositions, it tested the sensitivity of the set-ups to a sample composition that is at least similar to the bulk of our samples, and that had been through the same chemical preparation. The unique samples of low Tc atom concentrations from the fjord could not be directly tested for matrix effects, as the maximum amount of sample had to be retained in order to measure them on both ICP-MS set-ups.

Different approaches were used for testing the matrix effects on the two set-ups. For the PT2 with USN, one Kattegat sample (prepared in 10 ml) was subdivided into five 2 ml aliquots and four of these were spiked with 100 μl of Tc standards of increasing concentration, so that a calibration plot could be drawn with the sample concentration as the ‘blank’ (method of standard additions). The aspiration and counting time was reduced in line with the smaller volumes (2.1 ml) of these sub-samples. The gradient of this calibration curve was compared with that of the standard calibration curve, to establish if the signal was enhanced or suppressed by the sample matrix. For the Elan 6000 with ETV, our fjord samples were dissolved in a very small volume, 200 μl total volume per sample. Testing the matrix effect was therefore not as accurate as for the PT2 with USN, since we wanted to look at the matrix effect when the Kattegat samples were also dissolved in 200 μl, and so could not practically create a parallel calibration curve by subdividing and spiking a sample. Instead, the samples were dried slowly after analysis and then redissolved in 140 μl of a spiked solution.

Running protocols

Elan 6000 ICP-MS with ETV. The Elan 6000 with ETV was run using the parameters in Tables 2 and 3, utilising the development work of Skipperud et al.¹² The samples were run semi-manually, with washes, blanks and standards run regularly.

Table 2 Data acquisition parameters and ICP conditions for the Elan 6000 ICP-MS with ETV

Dwell time/ms (^99Tc; ^100Mo; ^101Ru)	15; 5; 10
Read delay/s	0.7
Sweeps per reading	1
Readings per replicate	100
Number of replicates	3
Points per peak	1
ICP rf power/W	1100
Plasma Ar flow rate/l min^−1	0.98
Auxiliary Ar flow rate/l min^−1	0.3
Nebuliser Ar flow rate/l min^−1	0.68
Sample uptake volume/μl	20

---

Table 3 ETV heating program

Cell temperature/°C	Ramp time/s	Hold time/s
80	10	10
120	5	10
2600	0	3
2650	1	4

---

PT2 ICP-SFMS with USN. The PT2 ICP-SFMS procedure had been developed prior to the comparison exercise, while the chemical method was also being developed. The ICP-SFMS method is similar to the method previously published for the measurement of trace levels of Pu in environmental samples.¹³ The operating conditions are shown in Table 4. The memory effect, or carry-over between samples, was eliminated using washes of 0.8 M HNO₃, as described by Richter et al.⁶ This brought the background down to the original level, and so we included a wash “sample” between every sample and standard. The wash time used was therefore the same as the time taken to analyse a sample (≈8 min), and is longer than deemed necessary by Richter et al.⁶ However, this reliably reduced the memory effect and allowed unknown Tc concentrations to be analysed in randomly ordered samples. Standards were run every 3–5 samples.

Table 4 Data acquisition parameters and ICP conditions for the PT2 ICP-SFMS with USN

a For each point in each peak.
Dwell time^a/ms (at m/z 99, 101, 103)	4
Sweeps per reading	80
Readings per replicate	1
Number of replicates	1
Points per peak	15
ICP rf power/W	1350
Plasma Ar flow rate/l min^−1	12.5
Auxilary Ar flow rate/l min^−1	0.8
Nebuliser Ar flow rate/l min^−1	0.7
Solution uptake rate/ml min^−1	1.15

---

Results and discussion

Radiochemical methodology

The results show that the chemical preparation successfully reduced Ru in the samples to a level where it did not interfere significantly with the Tc measurement. For example, on the Elan 6000 with ETV the counts at m/z 99 from Ru showed a mean of 313 ± 115 counts, whereas the Tc contributed 1600–20000 counts. However, this amount of Ru increases the minimum detectable activity in a sample by approximately 10%. The analyses of the ^99mTc yield monitor solutions, 20 times the activity used for each sample, showed that the amount of ⁹⁹Tc per yield monitor aliquot was negligible. All of these were below the detection limit, and so each yield monitor aliquot must have contained less than approximately 10 μBq (16 fg).

The chemical yield of the first part of the procedure for the 25–50 l samples (heating, precipitation and anion exchange column) showed an average of 80.5 ± 11% (1s) over the samples analysed in this study, with a general trend of slightly higher yields being obtained for the smaller samples. This is presumably because of lower losses during the Mg(OH)₂ precipitation since a greater proportion of the supernatant could be removed from the precipitate when the sample had been diluted (see Methods section). The average yield from the two TEVA.Spec separations was 96 ± 1%.

Stability

The Elan 6000 with ETV showed great stability over the course of the first run, with the external standards showing a standard deviation of only 3.4%. The PT2 ICP-SFMS with USN is less stable per se, as the USN introduces a certain amount of noise. However, using the Rh internal standard compensates for this, reducing the standard deviation of the external standards from 12.2% to 3.5%. ¹⁰³Rh therefore seems to be a better internal standard in the PT2 with USN than reported for AMS, which shows within-run variation of 12% between standards.⁸ On the second run on the Elan 6000 with ETV, however, there was a slow upward drift throughout the run, followed by a rapid decrease in signal when the tip sealing the ETV had corroded. This illustrates the day-to-day variability encountered in machine performance. However, the upward drift was slow enough to be monitored by the external standards, and did not necessitate the use of an internal standard.

Interferences

Potential interferences are listed in the Introduction. There are more potential interferences when using an ETV than a USN because of the formation of carbides from the graphite of the furnace, but the interferences from hydrides and oxides should be lessened. The probabilities of the ⁹⁸Mo¹H and the ⁵⁹Co⁴⁰Ar polyatomic interferences forming in the plasma of the PT2 ICP-SFMS were found experimentally. Increasing concentrations of Mo and Co were measured, monitoring the change in signal both at m/z 99 and 98 or 59, respectively. Once the signal at m/z 98 or 59 started to trip, the signal was estimated from the signal for lower concentrations, and the concentration was raised until the polyatomic interference was clearly visible. The formation probablities were calculated to be: ⁹⁸Mo¹H/⁹⁸Mo = 3 × 10⁻⁵ and, ⁵⁹Co⁴⁰Ar/⁵⁹Co = 2 × 10⁻⁶.

However, it is not always desirable to put very high concentrations of an element into the instrument, and so it is often sufficient to know that a polyatomic will not form at the concentrations present in your samples. Therefore, when using the Elan 6000 with ETV, the concentrations of ⁵⁹Co, ⁹⁸Mo, ⁸⁷Rb and ⁸⁷Sr were increased until they caused the machine to trip at the respective mass. No extra counts were observed at m/z 99. None of our samples caused the Elan 6000 to trip when measured at m/z 59, 98 or 87, and so polyatomic signals would not interfere with the measurement of these samples. Comparing this with the formation probability found using the PT2 set-up of ⁹⁸Mo¹H/⁹⁸Mo = 3 × 10⁻⁵, we should have only seen only 12 extra counts, which would have been indistinguishable from the background variation. The formation probability of ⁹⁸Mo¹H can therefore be said to be less than or approximately equal to this value.

The interferences from ⁴⁰Ar₂¹⁸O¹H, ⁴⁰Ca¹⁸O¹H⁴⁰Ar, and ⁴³Ca¹⁶O⁴⁰Ar were not investigated. Even if ⁴⁰Ar₂¹⁸O¹H forms to a significant extent, it is a constant interference for any given set-up, as the Ar flow rate and the amount of water introduced to the plasma are constant. Therefore, any counts from this species are included in the background measurement at m/z 99. Ca²⁺ will pass through the anion exchange and TEVA.Spec resins and so the majority will be removed from the samples in the radiochemical preparation. Regardless of this, any interference from ⁴⁰Ar₂¹⁸O¹H would far exceed that from ⁴⁰Ca¹⁸O¹H⁴⁰Ar, since Ar species constitute the vast majority of the ion beam. The background at m/z 99 in both systems is approximately 200 counts (Table 5, see below), thus, even if this was all due to ⁴⁰Ar₂¹⁸O¹H, any ⁴⁰Ca¹⁸O¹H⁴⁰Ar interference at m/z 99 would be negligible. Additionally, ⁴³Ca¹⁶O⁴⁰Ar involves the 0.135% abundance ⁴³Ca isotope and, thus, was not envisaged to represent a problem in the analysis of these samples.

Table 5 Detection limits (DL) of the PT2 with USN and the Elan 6000 with ETV^a

	PT2/sample^−1	Elan/measurement^−1		Elan/sample^−1
		(1)	(2)	(1)	(2)
a Elan 6000 with ETV detection limits given for two separate occasions.
Typical background/counts measurement^−1	200	200	300
Typical sensitivity/counts mBq^−1	700	8000	12000
Typical DL (3 × sqrt)/mBq	0.089	0.006	0.004	0.06	0.04
Typical DL (3 × 50 counts)/mBq	0.29	0.018	0.013	0.18	0.13
Typical DL (3 × sqrt)/fg	140	10	6	96	64
Typical DL (3 × 50 counts)/fg	460	30	20	290	210

---

Matrix effects

The PT2 with USN showed no matrix effect for either the raw Tc data or that corrected by the Rh internal standard. The gradients of the standard additions and calibration curve slopes were within 1s of one another, as shown in Table 6.

Table 6 Testing the matrix effect on the PT2 with USN

Rh correction?	Curve type	Gradient	1s
No	Calibration curve	2358	27
	Standard additions	2373	51
Yes	Calibration curve	8.03 × 10^−3	9 × 10^−5
	Standard additions	8.30 × 10^−3	2 × 10^−4

---

The Elan 6000 with ETV, however, showed an average signal enhancement of 1.32 ± 0.02 for the Kattegat samples. Interestingly, the ^99mTc yield monitor solutions measured using the ETV showed a much greater signal enhancement of 2.2 ± 0.15. when they were tested for a matrix effect. The matrix therefore has a very distinct effect on the transmission of Tc into the plasma and must be assessed very carefully when using an ETV.

Detection limits, background and sensitivity

Typical sensitivities and detection limits and are shown in Table 5. They are given in counts mBq⁻¹, mBq sample⁻¹ and fg sample⁻¹ to avoid the confusion of using ng l⁻¹, given that the sample volumes were 10 ml and 200 μl for the PT2 with USN and the Elan 6000 with ETV, respectively.

There is a significant amount of daily variation in the sensitivity of an instrument, depending on the optimisation of the machine and factors such as the time of the last service. Equally, the sensitivity can be improved by maximising the amount of time counting on a particular mass. For example, if Tc alone is measured, then the time spent counting the available atoms will be more than double when m/z 99 and 101 are measured. There is a finite time for the magnet to settle when it changes mass, hence the more masses counted, the more time is “wasted”, as well as there being less time available for measuring at m/z 99. Including an internal standard therefore raises the detection limit for Tc, although it is necessary when using an USN. The detection limit is affected by the sensitivity and the background, and so the cleanliness of the machine is also very relevant, and does vary somewhat from day to day. Therefore, it is extremely difficult to give an accurate, let alone exact, detection limit of any set-up.

The practical detection limit, quoted as three times the standard deviation of the background (3s), of measuring large batches of samples is even harder to assess, since the background may vary as a run progresses, and the signal will fall if the cones block slightly with build up of deposits. Moreover, for Tc, it depends on the amount of Ru in the sample. With this in mind, the data in Table 6 are given for the background amount of Ru (the signal at m/z 101 in the blanks and standards), at the beginning of a run, when the calibration curve is made. The background variation was determined in two ways, firstly as the square root of the initial background signal, and then as ±50 counts, which was seen to be a typical variation in background over the course of a run. The second detection limits are therefore the more realistic for a normal set of analyses. The sensitivity is monitored by external standards run at regular intervals throughout the course of a run, and these did not fall significantly for either machine. The data reflect the same conditions as used when analysing the samples, i.e., the same number of sweeps, the same dwell times and the same masses measured. They are a function of both the sample introduction system and the instrument itself and, when given per sample, are on the basis that samples for the Elan 6000 with ETV were dissolved in 200 μl solution, despite the fact that only 60 μl was used.

Given the above discussion, the detection limits and sensitivity of the two overall set-ups are remarkably similar, with both capable of measuring samples with a Tc content an order of magnitude lower than is possible to measure by beta-counting (≈3 mBq (5 pg)¹). Comparing these detection limits with those in the literature, they are certainly among the better values for ICP-MS, but are at least one order of magnitude higher than quoted for AMS³ and TIMS.² The detection limits on our PT2 instrument could probably be improved by approximately a factor of 10 by utilising a shielded torch, and reduced even further by using a MicroMist nebuliser, as demonstrated by Becker and Dietze.⁷

Comparison of the results from the two set-ups

Initially, a scatter plot was drawn of the results obtained for the fjord samples on the two ICP-MS set-ups, with no correction for the matrix effect when using the ETV. The correlation coefficient (R²) of the scatter plot was 0.978, with a 1s standard deviation of 4.9%, showing that the results are consistent between the two set-ups. However, the slope of the line was 1.29, with the results from the Elan 6000 with ETV being 29% higher than obtained using the PT2 with USN. This agrees well with the matrix effect observed for the Kattegat water samples, which showed no matrix effect when using the USN but a 30% signal enhancement when using the ETV. Thus, when working with an ETV the matrix effect must be monitored closely and, for these samples, the data must be divided by a correction factor of 1.3 to obtain accurate results. However, once the matrix effect has been accounted for, the two set-ups show good agreement. Fig. 1 is the scatter plot of the corrected results, of slope 1.0 ± 0.05, and Fig. 2 shows the residuals of the scatter plot, demonstrating that there is no systematic error.

Fig. 1 A scatter plot of the results for the fjord samples measured on both set-ups.

Fig. 2 The residuals of Fig. 1.

The Kattegat seawater used to test the reproducibility of the measurements had a concentration of 1.2 mBq l⁻¹. The PT2 with USN measurements showed a standard deviation of 6.1% over three samples and the Elan 6000 with ETV, 5.8% over two. After correcting the data from the Elan 6000 with ETV for the matrix effect (dividing by 1.3), the standard deviation for the five samples was lower than those from the individual set-ups, at 5.2%. This shows that the two set-ups agree within the precision of the individual set-ups and procedural reproducibility. Considering that the external standards showed standard deviations of 3.5% on the two set-ups (see above), there is an additional uncertainty associated with measuring samples, of about 2%. This is easily explained through the uncertainty associated with gamma-counting the ^99mTc yield monitor solution before and after the radiochemical procedure. One aliquot of the solution is used to measure the initial activity, and then further aliquots are added to the samples, and so there is a pipetting uncertainty that also contributes to this.

Overall, the Elan 6000 with ETV and the PT2 ICP-SFMS with USN are both capable of measuring Tc at low levels and with good precision. The PT2 ICP-SFMS was one of the first ever constructed, and lower detection limits have been quoted in the literature for similar machines, indicating that ICP-MS technology is improving all the time. Equally, the Elan 6000 is a relatively new instrument and shows much better results than usually quoted for quadrupole machines. The detection limit of the Elan 6000 with ETV could be improved by not counting on Mo, for example, or by dissolving the sample in a smaller volume. However, such changes could only improve the sensitivity by a factor of two or so, and would create different problems, such as how to assess for matrix effects for that sample without analysing parallel samples. Since Ru is not entirely removed from the samples, the detection limits and sensitivity of these set-ups are probably as good as could be required for measuring low-level, large volume samples by ICP-MS. However, the use of matrix modifiers may reduce the Ru interference when using ETV, as described by Song and Probst.¹⁴ With AMS, it has been demonstrated that 99% of the Ru in a sample can be prevented from reaching the detector for a 50% loss in ⁹⁹Tc signal.⁸ This is not possible using ICP-MS, but represents a way in which the low detection limits of AMS may be utilised without Ru masking the benefits.

Both set-ups have detection limits around an order of magnitude lower than beta-counting, and a series of low activity samples have been measured on both with excellent agreement between the two sets of measurements. The sample activities, while not being at the detection limit, would have been generally unmeasurable by beta-counting, demonstrating that ICP-MS is a viable, reliable and sensitive method for carrying out low-level analysis of ⁹⁹Tc isolated from large volume samples.

Acknowledgements

We thank the NKS for supporting this collaboration, Lise Stower-Roston for technical support when using the Elan 6000 and Lis Vinther Kristensen for technical support when using the PT2.

References

1. Q. Chen, H. Dahlgaard and S. P. Nielsen, Anal. Chim. Acta, 1994, 285, 177.
2. P. Dixon, D. B. Curtis, J. Musgrave, F. Roensch, J. Roach and D. Rokop, Anal. Chem., 1997, 69, 1692.
3. B. A. Bergquist, A. A. Marchetti, R. E. Martinelli, J. E. McAninch, G. J. Nimz, I. D. Proctor, J. R. Southon and J. S. Vogel, Nucl. Instrum. Meth. Phys. Res., Sect. B, 2000, 172, 328.
4. S. H. Sumiya, S. H. Morita, K. Tobita and M. Kurabayashi, J. Radioanal. Nucl. Chem., 1994, 177, 149.
5. D. M. Beals, J. Radioanal. Nucl. Chem., 1996, 204, 253.
6. R. C. Richter, S. R. R. Koirtyohann and S. S. Jurisson, J. Anal. At. Spectrom., 1997, 12, 557.
7. J. S. Becker and H. J. Dietze, J. Anal. At. Spectrom., 1999, 14, 1493.
8. L. K. Fifield, R. S. Carling, R. G. Cresswell, P. A. Hausladen, M. L. di Tada and J. P. Day, Nucl. Instrum. Meth. Phys. Res., Sect. B, 2000, 168, 427.
9. J. L. Más, M. García-León, J. P. Bolívar and C. Sánchez-Angulo, J. Anal. At. Spectrom., 2000, 15, 1369.
10. A. Montaser, M. G. Minnich, J. A. McLean, H. Liu, J. A. Caruso and C. W. McLeod, in Inductively Coupled Plasma Mass Spectrometry, ed. A. Montaser, Wiley-VCH, New York, 1998, pp. 83–264.
11. K. Tagami and S. Uchida, J. Radioanal. Nucl. Chem., 1999, 239, 643.
12. L. S. Skipperud, L. Stower-Roston and D. H. Oughton, J. Anal. At. Spectrom., submitted.
13. S. Stürup, H. Dahlgaard and C. H. Nielsen, J. Anal. At. Spectrom., 1998, 13, 1321.
14. M. Song and T. U. Probst, Anal. Chim. Acta, 2000, 413, 207.

---

Footnote

† Present address: Department of Earth Sciences, Dartmouth College, 6105 Sherman Fairchild Hall, Hanover, NH, 03755-3571, USA.

---