JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 609 Determination of Uranium and Thorium in Aluminium with Flow Injection and Laser Ablation Inductively Coupled Plasma Mass Spectrometry Peter van de Weijer Peter J. M. G. Vullings Wilhelmina L. M. Baeten and Wim J. M. de Laat Philips Research Laboratories P. 0. Box 80.000 5600 JA Eindhoven The Netherlands In order to determine uranium and thorium at the sub-ng g-l level in aluminium the limit of detection (LOD) for continuous-flow nebulization inductively coupled plasma mass spectrometry (ICP-MS) is not sufficient when a sample solution with the usual maximum concentration of 1 mg ml-l is used. Therefore two alternative sample introduction techniques have been used flow injection (FI) and laser ablation (LA).With FI-ICP-MS the achievement of sub-ng g-l detection limits is hampered by the presence of 'spikes'. Although these spikes are also present with LA it is possible to obtain a 0.2 ng g-l LOD for uranium and thorium. This LOD is achieved artificially by rejecting all measurements containing spikes. Keywords Inductively coupled plasma mass spectrometry; orifice deposition; flow injection; laser ablation; spike phenomenon As the dimensions of integrated circuits become smaller their susceptibility to alpha-particle induced damage ('soft errors') increases. Reduction of the exposure to alpha particles requires materials of extremely high purity with respect to elements from the natural uranium and thorium radioactive decay series. The required concentrations for uranium and thorium are below 1 ng g-l.With continuous- flow nebulization inductively coupled plasma mass spectro- metry (ICP-MS) the usual 1 mg ml-l sample concentra- tion' results in sub-pg ml-l levels of uranium and thorium in solution. These concentrations are below the detection limits attainable (2 pg ml-l for uranium and 4 pg ml-l for thorium2). Chemical preconcentration might solve this problem but this is time consuming and at the ng g-l level there is a chance of contamination. Therefore an attempt was made to measure uranium and thorium in samples of aluminium which is used for wiring in integrated circuits with flow injection (FI) and laser ablation (LA) ICP-MS. Experimental The work described was performed with a PlasmaQuad PQ I1 Plus ICP-MS instrument (VG Elemental). The operating conditions are given in Table 1.During all experiments the Table 1 Typical operating conditions of the ICP-MS system Plasma- Power 1250 W Argon outer gas flow rate Intermediate gas flow rate Mass spectrometer- Sampling depth Sampler nickel Skimmer nickel Expansion pressure 2 . 3 ~ lo2 Pa Intermediate pressure t l x Pa Analyser pressure 2.4 x lo-' Pa 13.5 1 min-l 0.6 1 min-' 10 mm beyond load coil 1.0 mm orifice (Nicone) 0.75 mm orifice (Nicone) Liquid sample introduction- Nebulizer flow rate Sample uptake rate Nebulizer Meinhard Solid sample introduction- Nd:YAG laser energy Laser pulse width 10 ns Laser focus Repetition rate 10 Hz Camer flow rate 0.74 1 min-l 0.8 ml min-l 250 mJ pulse-' 3 mm (glass) or 10 mm (Al) below sample surface 0.95 1 min-l sensitivity of the ICP-MS instrument was at its specified value (2 x lo5 counts s-l for a 100 ng ml-l solution of llSIn).For survey analyses the ion lenses were tuned to an element in the mid-mass range. For the determination of uranium and thorium they were tuned to uranium. All calibrated glassware was soaked in 10% v/v nitric acid overnight. Standard solutions of uranium and thorium were prepared by dilution of commercially available standard solutions containing 1 mg ml-l of the elements (supplied by Spex Industries). Hydrochloric acid and nitric acid were Suprapure re- agents obtained from Merck. The water was ultrapure according to the International Organization for Standard- ization 3696 Class 1 water. For FI the aluminium samples were dissolved in a 1 + 3 + 5 mixture of nitric acid (65%) hydrochloric acid (37%) and water.This acid mixture also served as the wash solution between injections. For LA the samples were etched for 2 min in the same mixture in order to remove surface contaminants and an additional pre-ablation time of 2 min was used. The standards used to obtain the sensitivity factors for the elements in aluminium were Pechiney A1 raffine A-99 (1 199) No. 8928,8930 and 893 1 Pechiney Alliage Al-Cu A- U2GN 9143 and Alcoa SA2049/19. These five standards contained 18 elements in the range from 1 ng g-l to 10 mg g-l. For uranium and thorium a separate high-purity aluminium sample (R06) prepared in-house was used. The uranium and thorium contents in this sample as measured in this department by neutron activation analysis3 (NAA) were found to be 25 and 52 ng g-l.Results and Discussion Continuous-flow Nebulization ICP-MS As a test of the ICP-MS apparatus used the limit of detection (LOD) (3dsensitivity) of uranium and thorium using continuous-flow nebulization of a solution containing 100 pg ml-l of uranium (no matrix) was determined. By using 50 s integration times a 2 pg ml-l detection limit for both uranium and thorium was obtained. This is compar- able to the values reported in the literature.2 The 2 pg ml-l LOD in solution corresponds to a 2 ng g-l LOD in a solid if one assumes a sample solution containing 1 mg ml-l of the solid sample. Aluminium however is a difficult matrix in terms of orifice blocking. In Fig. 1 the uranium signal from a solution containing 100 ng ml-l of uranium and 1 mg ml-I of aluminium is shown.The signal decreases over time owing to the orifice becoming blocked by A1203. If it is assumed that the transport efficiency of the sample intro-610 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL. 6 0 10 20 Ti me/mi n 30 Fig. 1 238U signal from a solution containing 100 ng ml-l of U and 1 mg mi-1 of Al. Dwell time per channel is 5 s duced by the nebulizer is 1% the aluminium flow into the plasma is 0.13 pg s-l. Although the time scale of Fig. 1 exceeds the normal analysis time in continuous-flow nebuli- zation ICP-MS it demonstrates the severe problem of orifice blocking. It indicates that the usual sample concen- tration of 1 mg ml-l is still too high when analysing aluminium with continuous-flow nebulization ICP-MS.Flow Injection ICP-MS The idea of using FI was to inject higher concentrations of aluminium but with less orifice bl~cking.~ The samples were introduced by the use of a home-made injection loop. The transient signals were recorded using the single-ion moni- toring facility. The injection volumes were varied from 50 to 1000 pl. The resulting signals are shown in Fig. 2. A 250 pl injection volume was selected for the remainder of the experiments as the amplitude of the flow injection peak approaches the steady-state value of continuous-flow nebulization at this value. With this injection volume an LOD for both uranium and thorium in solution of 4 pg ml-l (no aluminium matrix) was obtained. Then the signal for 100 ng ml- of uranium was measured in the presence of 1 2.5 5 and 10 mg ml-l of aluminium as a function of time.The time interval between two injections was 2 min resulting in a wash time of approximately 1 min. As shown in Fig. 3 the only acceptable signals were those at 1 and 2.5 mg ml-* of aluminium. However the LOD of uranium in aluminium would only be 2-4 ng g-l under those circum- stances. Therefore the wash time was increased from 1 to 3 min. As a result the signal decrease was reduced substan- tially (Fig. 4). A similar observation has been made by Douglas and These workers found that for a solution containing 10 mg ml-l of calcium a wash time of four times longer than the analysis time is sufficient to remove all of the deposits on the sampler. Time/s Fig.2 238U signal from a solution containing 100 pg ml-I of U in single-ion monitoring mode for five different injection volumes A 50; B 100; C 250; D 500; and E 1000 pl I A t 4- f 0 I 0 10 20 30 40 Time/min Fig. 3 238U signal from a solution containing 100 ng m1-l of U and 1-10 mg ml-l of Al A 1; B 2.5; C 5; and D 10 mg 1-l. Each data point represents the height of the peak after injection. The time interval between the injections is 2 min resulting in a wash time of about 1 min t 4- 0 0 10 20 30 40 Time/mi n Fig. 4 Effect of wash time on the 238U signal from a solution containing 100 ng ml-l of U and 10 mg ml-' of Al A 1; and B 3 min If the slight decrease of the uranium signal over time for the 3 min wash time is accepted a satisfactory LOD for uranium in the solid (0.4 ng g-l) could be obtained.However when an effort was made to establish the LOD in the presence of 10 mg ml-l of aluminium a value of 20 pg ml-l (Le. 2 ng g-l in the solid) was found. The increase in the LOD caused by the presence of aluminium which prohibits sub-ng g-l measurements is due to spikes in the mass spectrum. These spikes are shown in Fig. 5. The spikes are single-channel events that only take place with a direct aluminium flow into the plasma. Their presence is independent of the mlz value even when no atoms or molecules are to be expected. These spikes were only observed when injecting aluminium solutions. They were not observed for solutions containing 10 mg mi-' of Na Mg Zn or Mo. For aluminium these spikes have also been observed by Makishima et aL6 A possible explanation for the presence of the spikes could be photon scattering on A1203 clusters in the expansion stage.These clusters could be present in that region due to occasional release of pieces of A1203 from the sampling cone when aluminium matrices are being injected. In order to test this explanation a stainless-steel wire was mounted at the back of the sampling cone (Fig. 6). This wire could serve as a continuous source of photon scattering which would result in an increase in the background signal. However the presence of the wire did not cause a change in the background level while the indium signal from a 100 ng ml-l solution was measured (although the sensitivity decreased by a factor of 3). Therefore the spikes are notJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 199 1 VOL.6 61 1 0 10 Tirne/min 20 Fig. 5 (a) Signal at mlz 238. The first three peaks are the response to three injections of a solution containing 100 pg ml-I of U. After 10 min three injections of a blank solution have been performed which show no response at mlz 238. (6) Signal at mlz 238. The first three peaks are the response to three injections of a solution containing 100 pg ml-l of U and 10 mg ml-I of Al. After 10 min 3 injections of a solution containing 10 mg ml-I A1 have been performed which do show a response at mlz 238. (c) Signals at rnlz 230. The first five peaks are the response to five injections of a solution containing 100 pg ml-l of U and 10 mg ml-' of Al. After 10 min five injections of a solution containing 10 mg ml-I of A1 have been performed which show a similar response to the first five injections.Sensitivities are the same in (a) (b) and (c) induced by photon scattering in the expansion stage. In order to test the possibility of photon scattering from other regions of the mass spectrometer the neutral and the ion beam were blocked with a quartz window behind the skimmer. In this instance no spikes (or background) could be observed while a 10 mg ml-l solution of aluminium was nebulized. A quartz window behind the quadrupole but in front of the multiplier had the same result. Thus the spikes are not induced by photons unless their wavelength is so short (for example at the resonance radiation of argon) that they are absorbed by the quartz window.De-tuning of the lens stack had no influence on the number of spikes thereby eliminating the possibility of (regular) ions. Therefore neutral particles (e.g. aluminium atoms or aluminium oxide molecules) clusters of neutral particles or photons of short wavelengths are the most likely explanation for the presence of spikes. (a) Fig. 6 (a) Side and (b) bottom view of the sampling cone with a 1 mm diameter stainless-steel wire to simulate spikes Laser Ablation ICP-MS The amount of aluminium flowing into the ICP when ablating aluminium samples is 0.3 pg s-l. This number is derived from the mass loss of the sample if a 10% efficiency of the sample introduction is assumed. Despite the larger aluminium flow with LA in comparison to that with continuous-flow nebulization a slower decrease in signal is observed.This can be seen by comparing Fig. 7 with Fig. 1. There are two reasons for this slower decrease. First the ablation process is interrupted after each 3 rnin period for 1 min in order to move the laser spot to a fresh part of the surface of the sample. In this way a decrease in the signal due to cratering is prevented. This means that there is a 1 rnin wash time whereas Fig. 1 corresponds to a continuous flow. Secondly there is reduced oxide formation due to the absence of water resulting in less orifice blocking. The ion signal with LA sample introduction however is far less stable than the signal with nebulization of a solution. As the problem of orifice blocking is less severe with LA it appeared to be a suitable technique for both semi-quantita- tive survey analysis and full quantitative analysis of aluminium samples.The response curve in semi-quantita- tive survey analysis of the elements in aluminium is different from that obtained by nebulizing a solution or by ablating a glass sample. This is shown in Fig. 8 in a qualitative way (the experimental conditions are not exactly the same for the three measurements). The difference in the response curves between glass and aluminium is probably due to the higher heat conductivity of aluminium in comparison with that of glass.7 As a result of this difference the temperature during ablation of glass is higher than during the ablation of aluminium. For aluminium the gasification is influenced by evaporation effects resulting in higher sensitivity factors for more volatile elements (such as lead and magne~ium).~ It is surprising however that the sensitivity for the even more volatile cadmium is so low.For glass gasification is dominated by the ablation process resulting in a gas-phase composition that reflects the composition of the solid. 0 10 20 30 40 Ti m e/rn i n Fig. 7 Signal for 238U from an A1 standard which contains approximately 1 ,ug g-l of U. In order to avoid ablation from a deep crater the laser ablation was interrupted for 1 min after each 3 min period and the laser moved to a different place on the sample612 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1991 VOL. 6 0 100 200 mtz Fig. 8 Response curves obtained with the relative sensitivity factors obtained for the solution at a nebulizer flow of 0.74 1 min-'. (a) Solution (without a matrix) nebulizer flow 0.74 1 min-I; (b) NIST SRM 612 Trace Elements in Glass nebulizer flow 0.95 1 min-l; and (c) aluminium Pechiney 8930 nebulizer flow 0.95 1 min-' The response curve in Fig.8(c) is not suitable for performing semi-quantitative survey analysis of elements in aluminium. The high mlz portion of the curve is dominated by the highly volatile lead. This could lead to unrealistic values e.g. for the actinides. Therefore the relative sensitivity factors for a number of elements using six aluminium standards were determined. Table 2 provides a comparison of these sensitivity factors with those obtained from the solution using the same carrier gas flow rate (0.95 1 min-l). Both sets of results are normalized to manganese.By using 1 min scans the concentration of elements can be measured in the range from 100 ng g-l to 0.1 g g-l; zinc was used as an internal standard. Fig. 9 gives an example of such measurements. The precision of the measurements with laser ablation ICP-MS is substantially lower than that of continuous-flow nebulization ICP-MS. Defining the blank signal as the signal with laser off is not correct. The laser pulse not only results in a signal from the sample but also in a signal from other parts of the sample introduction system (including any memory) which is probably induced by the shock wave of the laser-induced plasma. As an illustration the signal for 52Cr which was corrected for the laser-off signal is shown in Fig.10. At low concentrations of chromium a contribution from Arc occurs which results from ablation of the poly(tetrafluoroethy1ene) bottom of the sample chamber. In order to test the detection limit of LA-ICP-MS two aluminium samples were used in which uranium and thorium were measured by NAA:3sample R06 contains 25 k 3 ng g-l of uranium and 52 k 10 ng g-l of thorium; and sample R07 contains 0.6k0.3 ng g-l of uranium and 1.3k0.1 ng g-l of thorium. The R06 was used as a standard whereas R07 was considered to be a sample. A peak-jumping procedure was used in dual mode; the analogue mode was used to measure the aluminium matrix J 0 1 2 3 4 5 6 7 Log(Cu concentrationtng g-') Fig. 9 Plot of signal for Cu using Zn as an internal standard as a function of the concentration in A1 standards A 63Cu; and B T u Table 2 Relative sensitivity factors (intensity ratios of the signals normalized to concentration and abundance) for a number of elements as obtained in 1 min scans of six aluminium standards compared with the sensitivities for solutions (without aluminium) at the same plasma conditions (nebulizer flow 0.95 1 min-I) Relative intensity factor Relative intensity factor Solution Solution without Aluminium without Aluminium Element aluminium standard Element aluminium standard Be Mg Ti V Cr Mn Fe c o Ni c u 0.18 0.95 0.92 0.93 0.93 1 .oo 1.25 1.03 0.82 0.66 0.24 1.53 0.52 0.53 0.6 1 1 .oo 0.55 0.4 1 0.35 0.38 Zn Ga Zr Cd Sn Sb Pb Bi Th U 0.32 0.97 0.68 0.48 0.78 0.3 1 0.6 1 0.49 0.43 0.4 1 0.62 1.08 0.42 0.7 1 0.98 0.57 1.93 1.47 0.58* 0.59* *Obtained from peak jumping.JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1991 VOL.6 613 o 1 2 3 4 5 6 7 8 Log(Cr concentration/ng 9-7 Fig. 10 Plot of signal for T r using Zn as an internal standard as a function of the concentration in Al standards which served as an internal standard. The pulse-counting mode was used to record the uranium and thorium signals. The peak-jump parameters are shown in Table 3. The results given in Table 4 show the sub-ng g-l detection capability of LA-ICP-MS. These results were obtained at the beginning of the attempts to measure uranium and thorium in aluminium at such low concentrations. At that time measurements were not hampered by spikes. After many hours of introducing aluminium into the ICP system by both LA and FI the measurement of uranium and thorium in R07 by laser ablation was repeated; the spikes were then observed.However uranium and thorium could be measured at the ng g-l level when rejecting those measurements which included spikes in one of the five digital-to-analogue converter (DlA) steps (see Table 5). The detection limit as determined at mlz 230 while ablating the aluminium sample or at mlz 232 and 238 with the laser off is 0.2 ng g-l for both elements. This of course is a rather unconventional way of determining the LOD. It is how- ever the only way that our purpose could be achieved which is a sub-ng g-l measurement of uranium and thorium. If the spiked measurements were not rejected the result of the measurement would be unrealistically high.If the spiked blanks at mlz 230 are not rejected the calculated detection limit can be a factor 2-5 higher. The LOD calculated at mlz 232 and 238 with laser off is not hampered by spikes as there is no flow of aluminium into the plasma. In view of the unconventional determination of the detection limits of LA-ICP-MS it can be concluded tenta- tively that the performance of LA-ICP-MS is better than that of FI-ICP-MS. Rejection of spiked measurements in FI-ICP-MS is not possible as one measurement lasts about 1 min even if smaller injection volumes are chosen. In a 1 min period there is always at least one spike. A shorter measurement could be obtained by a synchronized peak- jump measurement during the FI peak. However in that instance the required LOD cannot be obtained.The intention was to measure uranium and thorium in aluminium samples that contained approximately 1 0 mg g-l of copper and 10 mg g-l of silicon. Unfortunately standards for this matrix were not available. In order to test whether the sensitivity factors of the elements were influ- enced by the presence of copper and silicon the following experiment was performed. By using the high-purity alumi- nium sample R06 as a standard an effort was made to measure the concentration of a number of elements in an Table 3 Peak-jump parameters for the determination of uranium and thorium in aluminium Quadrupole settle time 10 ms Peak jumping dwell time Number of points per peak Number of sweeps per peak Detector type Dual 20 ps for Al 20 ms for U and Th 5 100 Peak jumping D/A steps 5 Table 4 Comparison of the results obtained by LA-ICP-MS with those obtained with NAA (concentrations % la in ng g-') Element NAA LA-ICP-MS U 0.6 % 0.3 0.7 k 0.2 Th 1.3 k 0.1 1.3 k 0.2 Table 5 Comparison of the results obtained by LA-ICP-MS while rejecting measurements with spikes with those obtained with NAA (concentrations k la in ng g-l) Element NAA LA-ICP-MS U 0.6 k 0.3 0.8 % 0.3 3 out of 6 runs Th 1.3k0.1 1.1 k0.3 4 out of 6 runs Table 6 Comparison of the concentration for a number of elements in Al-Si-Cu as obtained with LA-ICP-MS using an aluminium standard with those obtained by NAA (concentrations % la in ng g-') R06 by Al-Si-Cu by Al-Si-Cu by Element NAA NAA LA-ICP-MS s c 7625 55k3 77+ 15 Cr 150k 10 97k5 103 f 8 As 170k20 620+ 120 520 k 60 108% 13 Sb 180k20 60%6 Hf 4.3 k 0.5 1.6k0.2 1.1 %0.2 Th 52% 10 <0.2 (0.3 U 25k3 (1 (0.3 Al-Si-Cu sample. The concentration of these elements is known by measurement with NAA.3 The results presented in Table 6 show that the concentrations recovered are almost within experimental error.Apparently the presence of 10 mg g-l of copper and silicon does not have a dramatic effect on the aluminium matrix in terms of the sensitivity factors of the impurity elements. Conclusion The measurement of uranium and thorium in aluminium is hampered by the presence of spikes. At the same aluminium flow into the plasma spikes occur more often with sample introduction as aqueous solution than by LA suggesting that A1203 is involved. By rejecting measurements with spikes it is possible to measure sub-ng g-l concentra- tions with LA-ICP-MS. The LOD for both elements is 0.2 ng g-*. References 1 2 Hieftje G. M. and Vickers G. H. Anal. Chim. Actu 1989 Date A. R. and Gray A. L. Applications of Inductively 216 1.614 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY DECEMBER 1991 VOL. 6 Coupled Plasma Mass Spectrometry Blackie Glasgow 1989. Hanssen J. M. G. Jansen R. M. W. and Jaspers H. J. J. personal corn m unicat ion. Hutton R. C. and Eaton A. N. J. Anal. At. Spectrom. 1988 3 547. Paper 1 /00380A Douglas D. J. and Kerr L.A. J. Anal. At. Spectrom. 1988 3 Received January 28th 1991 749. Accepted July 30th 1991 6 7 Makishima A. Inamoto I. and Chiba K. Appl. Spectrosc. 1990,44 91. Hager J. W. Anal. Chem. 1989 61 1243. 3 4 5