Anal. Chem. 2000, 72, 4310-4316
Multielement Analysis of Polyethylene Using Solid Sampling Electrothermal Vaporization ICP Mass Spectrometry Frank Vanhaecke,* Martin Resano,† Marieke Verstraete, Luc Moens, and Richard Dams
Laboratory of Analytical Chemistry, Ghent University, Institute for Nuclear Sciences, Proeftuinstraat 86, B-9000 Ghent, Belgium
Next to laser ablation (LA) also electrothermal vaporization (ETV) from a graphite furnace as a means of sample introduction opens possibilities for direct analysis of solid samples using inductively coupled plasma mass spectrometry (ICPMS). In this paper, it is demonstrated that solid sampling ETV-ICPMS is very well suited for the determination of metal traces in polyethylene. A limited multielement capability is often cited as an important drawback of ETV-ICPMS. However, by studying the effect of monitoring an increasing number of mass-to-charge ratios on the signal profile (integrated signal intensity and repeatability) of selected analyte elements, the multielement capability of (solid sampling) ETV-ICPMS was systematically evaluated, and the results obtained suggest that, with a quadrupole-based ICPMS instrument, at least 11 elements can be determined “simultaneously” (from the same vaporization step), in essence without compromising the sensitivity or the precision of the results obtained. In this work, the “simultaneous” determination of Al, Ba, Cd, Cu, Mn, Pb, and Ti in a polyethylene candidate reference material has been accomplished, despite the large variation in analyte concentration (from 5 ng/g for Mn to 500 µg/g for Ti) and in furnace behavior (volatility) they exhibit. To avoid premature losses of Cd during thermal pretreatment of the samples, Pd was used as a chemical modifier. Two different calibration methodss external calibration using an aqueous standard solution and single standard additionswere studied and the results obtained were compared with those obtained using neutron activation analysis (NAA) and/or with the corresponding (candidate) certified values (if available). Single standard addition was shown to be preferable (average deviation between ICPMS result and reference value < 3%), althoughsexcept for Basacceptable results could also be obtained with external calibration. While inductively coupled plasma mass spectrometry (ICPMS) is mainly intended for the trace analysis of sample solutions, direct analysis of solid materials (solid sampling) can be accomplished * Corresponding author. E-mail:
[email protected]. Fax: +32-92646699. † While on leave from the Department of Analytical Chemistry, University of Zaragoza, E-50009 Zaragoza, Spain.
4310 Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
when using laser ablation (LA) for sample introduction.1,2 In this way, wet chemistry sample pretreatment and the associated drawbackssrisk of contamination and/or analyte losses, sample dilution, and the sometimes laborious and time-consuming digestion procedures requiredscan be avoided. LA-ICPMS may be particularly useful in those situations in which information about several elements is required and a high sample throughput is more important than obtaining very accurate and/or precise results. Additionally, the application range of LA-ICPMS was recently considerably extended as a result of the introduction of ultraviolet (UV) lasers.3 Direct analysis of plastics could already be accomplished using an infrared (IR) laser,4 but application of a UV laser allows an improved spatial resolution and a better control of the ablation process.5 In LA-ICPMS analysis of plastics, calibration was accomplished using either (solid) synthetic standards with a matrix composition similar to the materials being analyzed4,6 or aqueous standard solutions to which a suitable chromophore was added to modify the absorption of the laser photons.7 The coupling of a graphite furnace to an ICP mass spectrometer also results in a method that permits the direct analysis of solid samples. In this work, the utility of electrothermal vaporization (ETV) ICPMS for multielement trace metal analysis of polyethylene was studied. However, when surveying the literature devoted to ETV-ICPMS, it is rather surprising that the large majority of papers have only dealt with the “simultaneous” determination (from the same vaporization step) of 1-3 elements, and only very few report on the determination of a larger number.8-11 In solid sampling ETV-ICPMS, the situation is even (1) Durrant, S. F. J. Anal. At. Spectrom. 1999, 14, 1385-1403. (2) Gu ¨nther, D.; Jackson, S. E.; Longerich, H. P. Spectrochim. Acta, Part B 1999, 54, 381-409. (3) Jeffries, T. E.; Perkins, W. T.; Pearce, N. J. G. Analyst (Cambridge, U.K.) 1995, 120, 1365-1371. (4) Marshall, J.; Franks, J.; Abell, I.; Tye, C. J. Anal. At. Spectrom. 1991, 6, 145-150. (5) Hemmerlin, M.; Mermet, J. M. Spectrochim. Acta, Part B 1996, 51, 579589. (6) Bertucci, C.; Zydowicz P. J. Phys. IV 1996, 6, 853-862. (7) Boue-Bigne, F.; Masters, B. J.; Crighton, J. S.; Sharp, B. L. J. Anal. At. Spectrom. 1999, 14, 1665-1672. (8) Olson, L. K.; Vela, N. P.; Caruso, J. A. Spectrochim. Acta, Part B 1995, 50, 355-368. (9) Berryman, N. G.; Probst, T. U. Fresenius’ J. Anal. Chem. 1996, 355, 783788. (10) Fairman, B.; Catterick, T. J. Anal. At. Spectrom. 1997, 12, 863-866. (11) Byrne, J. P.; Chapple, G. At. Spectrosc. 1998, 19, 116-120. 10.1021/ac0002700 CCC: $19.00
© 2000 American Chemical Society Published on Web 08/19/2000
more pronounced, as most work reported focused on the determination of a single element only.12,13 This strongly contrasts with the articles devoted to ETV ICP optical emission spectrometry (ETV-ICPOES), which sometimes describe the determination of more than 10 elements.14,15 A possible reason for the limited use of the multielement capabilities of the “detector” in ETV-ICPMS is the rather limited suitability of the quadrupole filtersthe most commonly used mass spectrometer in ICPMS instrumentations to deal with the transient signals that the graphite furnace produces. It has, for instance, been stated that in order to avoid severe signal distortion, a maximum of four elements is recommended.16,17 However, to the best of our knowledge, no experimental study aiming at an evaluation of the limits of the multielement capabilities of the quadrupole filter to deal with transient graphite furnace signals has been carried out so far. Therefore, a systematic study, aiming at obtaining information concerning the maximum number of nuclides that can be monitored with the experimental setup used, without being confronted with a substantially deteriorated sensitivity and/or precision, was carried out. Subsequently, the simultaneous determination of Al, Ba, Cd, Cu, Mn, Pb, and Ti in a polyethylene sample using solid sampling ETV-ICPMS was accomplished. The analysis of polyethylene was selected as an application in which a relatively large number of elements has to be monitored during the manufacturing process, while the results have to be obtained as fast as possible. The elements selected were present in a very wide concentration range, ranging from about 5 ng/g for Mn to about 500 µg/g for Ti, and they show significant differences in terms of furnace behavior (volatility). EXPERIMENTAL SECTION Instrumentation. A Perkin-Elmer HGA-600MS electrothermal vaporizer coupled to a Perkin-Elmer Sciex Elan 5000 ICP mass spectrometer was used. The electrothermal vaporization unit was equipped with a model AS-60 autosampler. Pyrolytic graphitecoated tubes and cups (cup-in-tube technique for solid sampling18) were used throughout. Both the left and the right-hand contact cylinder were modified (a cutout was made) to permit convenient insertion and removal of the cup. The HGA-600MS was interfaced to the ICP via an 80-cm length (6-mm i.d.) Teflon tubing. The operation of the HGA-600MS was completely computer-controlled. During the drying and pyrolysis steps of the temperature program, opposing flows of argon gas (300 mL/min), originating from both ends of the graphite tube, removed all the vapors through the dosing hole. Before starting the vaporization step, a graphite probe was pneumatically activated to seal the dosing hole. Once the graphite tube was sealed, a valve located at one end of the HGA (12) Moens, L.; Verrept, P.; Boonen, S.; Vanhaecke, F.; Dams, R. Spectrochim. Acta, Part B 1995, 50, 463-475. (13) Kurfu ¨ rst, U., Ed. Solid Sample Analysis; Springer-Verlag: Berlin Heidelberg, 1998; pp 191-246. (14) Schaeffer, U.; Krivan, V. Anal. Chem. 1999, 71, 849-854. (15) Zaray, G.; Kantor, T.; Wolff, G.; Zadgorska, Z.; Nickel, H. Mikrochim. Acta 1992, 107, 345-358. (16) Lamoureux, M. M.; Gre´goire, D. C.; Chakrabarti, C. L.; Goltz, D. M. Anal. Chem. 1994, 66, 3208-3216. (17) Yu, L.; Koirtyohann, S. R.; Rueppel, M. L.; Skipor, A. K.; Jacobs, J. J. J. Anal. At. Spectrom. 1997, 12, 69-74. (18) Voellkopf, U.; Grobenski, Z.; Tamm, R.; Welz, B. Analyst (Cambridge, U.K.) 1985, 110, 573-577.
workhead directed the carrier argon flow originating from the far end of the graphite tube directly to the argon plasma at a flow rate of 1000 mL/min. To prevent the cup from being “launched” from the tube when switching from the internal flow (300 mL/ min) to the carrier flow (1000 mL/min), the HGA-600MS was equipped with an adjustable pneumatic delay (Martonair BM/ 1430).19 A microbalance (Sartorius M3P, Germany) with a readability of 1 µg was used for weighing the samples. Samples and Standards. Samples. The polyethylene sample analyzed was a candidate reference material. In the polymer industry, the concentration of several elements sadditives, catalyst residues, and contaminantsshas to be determined in the materials produced in order to achieve the required level of production control and quality management. Also, from an environmental point of view, reliable analysis of plastics is an important issue, as an EC-directive (94/62/EC) regulates the maximum amounts of heavy metals (Cd, Cr, Hg, and Pb) allowed in plastics. The element concentrations that have to be determined range from the percent down to the nanogram per kilogram level.20 Analysis of certified reference materials (CRMs) is a traditional way of checking the level of accuracy with which an analysis can be performed. Unfortunately, until recently, no CRMs of this type were available. Therefore, the Standards, Measurements & Testing (SM&T) Program of the European Community started a campaign, aiming at the certification of the contents of As, Br, Cd, Cl, Cr, Hg, Pb, and S in two consumable synthetic polymer materials. These reference materials were produced by doping a polyethylene base material with the aforementioned elements at two concentration levels (high level CRM 680 and low level CRM 681). At the time of writing this paper, these CRMs are not yet commercially available, but the “candidate certified values” have been submitted to the BCR Certification Committee. More information on the production of these CRMs (“PERMs”spolymer elemental reference materials) can be found elsewhere.20 Standards. A standard solution, containing appropriate amounts of Al, Ba, Cd, Cu, Mn, Pb, and Ti, was prepared from commercially available 1 g/L single-element standards by dilution with 0.14 M HNO3. HNO3, 14 M, was purified by sub-boiling distillation in quartz equipment and 100-fold diluted using Milli-Q water to obtain 0.14 M HNO3. Water was doubly distilled and further purified using a Milli-Q water purification system (Millipore, MA). The chemical modifier solution was prepared by dilution of a commercially available 1 g/L Pd (as Pd(NO3)2) stock solution. Procedure for Solid Sampling ETV-ICPMS. Low-level PERM (BCR CRM 861) is available in the form of grains, each of these weighing about 10 mg. A ceramic knife was used to cut the polyethylene grains into smaller pieces (of about 1 mg). Thereafter, the material was loaded into a sample cup using a pair of tweezers, coated with PTFE to prevent contamination, and its mass determined by means of a microbalance. The sample cup was finally inserted into the furnace for subsequent analysis using the insertion tool provided by Perkin-Elmer. Finally, 20 µL of 250 mg/L Pd solution were added using the autosampler. The operating conditions are summarized in Table 1. (19) van Alphen, A. Akzo Nobel Research Report RGA F98003. (20) Van Borm, W.; Lamberty, A.; Quevauviller, P. Fresenius’ J. Anal. Chem. 1999, 365, 361-363.
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4311
Table 1. Instrumental Operating Conditions and Data Acquisition Parameters for the ETV-ICPMS Measurements ICP-Mass Spectrometer RF power/W 1000 Ar gas flow rates/L/min plasma 12 auxiliary 1.2 carrier 1.0 sampling cone nickel, 1.0 mm aperture diameter skimmer nickel, 0.75 mm aperture diameter lens voltages a HGA-600MS Electrothermal Vaporizer sample mass 0.5-1.0 mg chemical modifier Pd (5 µg) vaporization program: temperature (°C) ramp (s) hold time (s) drying step 120 10 30 pyrolysis step 800 10 30 vaporization step 2700 0 10 Data Acquisition scanning mode peak hop transient dwell time per acquisition point 40 ms acquisition points per spectral peak 1 sweeps per reading 1 readings per replicate 25 27Al+, 137Ba+, 114Cd+, 63Cu+, signals monitored 55Mn+, 208Pb+, 49Ti+ total measurement time 10 s a Optimized using pneumatic nebulization; no further tuning required when switching to ETV.
Every analysis consisted of the determination of the blank value (by measuring three times the modifier solution alone), 5 replicate measurements of sample material (0.5-1 mg) and (i) 5 replicate measurements of the standard solution (external calibration), or (ii) 5 replicate measurements of sample material to which 10 µL of multielement standard solution was added (single standard addition), depending on the calibration approach. Hence, determination of the seven elements could be accomplished in approximately 30 minutes (no chemical sample pretreatment required), regardless of the calibration method used. For 5 replicate measurements of solid sample or of solid sample to which an aliquot of standard solution was added (standard addition), the median of the five results obtained was taken as the representative value instead of the average value. In this way, the detrimental influence of possible outliers caused by sample inhomogeneity is minimized, as has been discussed in ref 21. Procedure for NAA. In the framework of the certification campaign, organized by the SM & T Program of the European Community, the polyethylene candidate reference material studied has also been analyzed at our laboratory22 by means of neutron activation analysis (NAA) with k0-standardization 23 using an IRMM-530 Al-0.1%Au monitor.24 For this purpose, 1.5-g samples were irradiated in the nuclear reactor Thetis, present at our (21) Belarra, M. A.; Resano, M.; Castillo, J. R. J. Anal. At. Spectrom. 1999, 14, 547-552. (22) De Corte, F.; De Wispelaere, A., k0-INAA of PERM materials, report sent to the SM&T Program in the framework of the PERM certification campaign. (23) De Corte, F.; Simonits, A. J. Radioanal. Nucl. Chem. 1989, 133, 43-130. (24) Ingelbrecht, C.; Peetermans, F.; De Corte, F.; De Wispelaere, A.; Vandecasteele, C.; Courtijn, E.; D’Hondt, P. Nucl. Instrum. Methods Phys. Res., Sect. A 1991, 303, 119-122.
4312
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
Laboratory. Subsequently, the γ-radiation emitted by the induced radioisotopes was measured using a high-purity Ge detector. Spectrum analysis and concentration calculation were accomplished using the KAYZERO/SOLCOI software package (version 4, DSM Research, The Netherlands). RESULTS AND DISCUSSION Preliminary Experiments. The wide variation in analyte concentrationsfrom 5 ng/g for Mn to 500 µg/g for Tispresented an extra difficulty that had to be coped with. The OmniRange device of the Elan 5000swhich can selectively and reproducibly reduce the sensitivity of the mass spectrometer by varying the ion transmission efficiency (P lens setting) at the exact time at which the signal at a given mass-to-charge ratio is being measureds was used to bring excessively high signal intensities to within the working range of the detector. Additionally, as a result of the high Ba and Ti concentrations, monitoring of less abundant isotopes (137Ba and 49Ti, respectively) was required. In this way, the signal intensities for the analytes differed from one another by a maximum factor of 3 to 4 only, while the difference in concentration amounted to 5 orders of magnitude. The refractory character of Ti requires the use of the maximum vaporization temperature (2700 °C) during 10 seconds. For the pyrolysis step, a temperature of at least 700 °C is necessary to ensure adequate matrix removal. When using lower temperatures, the plasma was observed to douse during the vaporization step on various occasions, as a result of too much material being transported into the plasma. A pretreatment temperature of 700 °C poses no problem for most of the elements. For Cd, the volatility of which is well-known, problems were observed when no appropriate measures were taken. For an aqueous Cd standard, premature losses were observed at temperatures higher than 400 °C. Hence, to accomplish the determination of Cd, the use of a chemical modifier was necessary. Pd as a Chemical Modifier. The efficiency of Pd as a universal modifier has been reported on.25 Moreover, Pd has been previously used in order to achieve the determination of Cd by solid sampling GFAAS26 when similar problems were encountered. Hence, in this work its performance was also tested (added as Pd(NO3)2). In a first series of experiments, the suitability of using 1 µg of Pd was evaluated. Under these conditions, premature losses of Cd could not be completely avoided when working with solutions showing an analyte concentration similar to that in the solid samples. This fact is probably to be attributed to the relatively high Cd concentration (approximately 20 µg/g). Hence, the addition of 5 µg of Pd was tested. As can be seen in Figure 1, under these conditions, Pd can “stabilize” Cd at temperatures up to 800 °C for both solid samples and aqueous solutions. The similar behavior observed for solid samples and aqueous solutions is striking and opens possibilities for external calibration using aqueous standards, as will be discussed later on. The addition of Pd does not only affect the Cd signal, Pd can also act as a physical carrier, improving the sensitivity for other analytes.27 The use of Pd did not result in significant differences (25) Schlemmer, G.; Welz, B. Spectrochim. Acta, Part B 1986, 41, 1157-1165. (26) Belarra, M. A.; Resano, M.; Rodrı´guez, S.; Urchaga, J.; Castillo, J. R. Spectrochim. Acta, Part B 1999, 54, 787-795. (27) Ediger, R. D.; Beres, S. A. Spectrochim. Acta, Part B 1992, 47, 907-922.
Table 2. Effect of Adding 5 µg of Pd on the Peak Area for the Different Analyte Elements Present in a Standard Solution (Amount Introduced, 1 ng)a
without Pd with Pd a
Cd
Pb
100 (9) 3170 (510)
100 (2) 228 (36)
Ba
Ti
peak area (arbitrary units) 100 (9) 100 (10) 101 (3) 87 (1)
Cu
Mn
Al
100 (4) 69 (3)
100 (4) 50 (2)
100 (7) 34 (1)
Standard deviations are presented in brackets. The elements have been arranged according to the observed variation in intensity. n ) 4.
Figure 1. 114Cd+ signal intensity (peak area) as a function of the pyrolysis temperature for an aqueous standard solution of Cd (absolute amount of Cd introduced into the furnace, 20 ng) and solid samples (results normalized to 1 mg of sample, corresponding to ∼22 ng of Cd) of low-level PERM with 5 µg of Pd. Each result represents the average value of three measurements (solution) or the median of five measurements (solids).
in terms of signal profile. The relative differences in the sensitivity of the elements, on the other hand, are shown in Table 2. The enormous increase in the Cd signal intensity cannot be totally attributed to a physical effect (improvement of transport efficiency), but is the result of a chemical mechanism (Pd-Cd interaction), preventing premature losses. For the other elements, Pd appears to affect the sensitivity (transport efficiency) only to a limited extent, as the variation in signal intensity observed ranges from a 2-fold increase to a 3-fold decrease. It is logical that Pd does not improve the transport efficiency for the least volatile elements, because the vaporization of Pd occurs at lower temperatures than the vaporization of these analyte elements. However, the reason for the signal suppression is not clear. This might indicate some interaction between the analytes and the modifier that prevents the formation of (a) species with a higher transport efficiency for these elements. The use of relatively high amounts of Pd has already been shown to lead to a reduction in sensitivity.27-28 As the study of the nature of the Pd-analyte interaction is beyond the scope of this article, no further experiments were carried out in this context. Multielement Capability. As was previously pointed out, it is often assumed that measuring more that four transient ETV signals with a duration of a few seconds only with a quadrupolebased instrument may result in spectral skew (distortion).16,17 However, the lack of information on the actual impact of this skew on the analytical results is rather surprising. Hence, the effect of the number of elements measured on both the sensitivity and the precision for solution and solid sampling ETV-ICPMS was studied. (28) Resano, M.; Verstraete, M.; Vanhaecke, F.; Moens, L.; van Alphen, A.; Denoyer, E. R. J. Anal. At. Spectrom., 2000, in press.
This study was started using solution ETV-ICPMS. For this evaluation, Cu and Pb were chosen as target elements, because they exhibit more narrow signals (of shorter duration) than some of the other elements (e.g., Ba and Ti) and, subsequently, they should be affected to a larger extent by the decrease of the number of measurement points per transient signal. To carry out this study, the effect of monitoring 1, 3, 7, 11, 15, or 30 mass-tocharge ratios was evaluated. For each number of mass-to-charge ratios monitored, three series of 15 replicate measurements were carried out on three different days. The dwell time was set at 40 ms as a compromise value, ensuring both a sufficient number of measurement points (when monitoring a few nuclides) and an efficient use of the total measuring time.29 The total measuring time was set to 10 seconds, but the actual duration of the Cu and Pb signals is no longer than three seconds. The software of the ICP mass spectrometer used provides several ways to convert the raw data obtained for a nuclide into the signal intensity reported to the user. For transient signals, the options counted or integrated signal intensity can be used. When using the counted signal, the signal intensity reported equals the summation of the counts actually observed. When selecting the integrated signal intensity, on the other hand, the software calculates the best-fitting curve through the experimental data points and the surface under this curve is reported to the user. The choice between counted and integrated signal intensity affects the variation of the sensitivity as a function of the number of nuclides monitored to a very important extent. When increasing the number of nuclides monitored, there will be less data points describing the signal profile for each of the target elements. Selfevidently, this reduction in the number of data points per peak automatically leads to a substantial reduction in the counted signal intensity. Under identical experimental conditions, the decrease in integrated signal intensity, on the other hand, will be considerably less important as, in each case, an estimation of the total signal (surface under signal profile) is still made. In fact, under ideal circumstances, the integrated signal intensity may even remain unchanged on decreasing the number of data points. In the present work, the integrated signal intensity was used throughout. The difference in the signal shapes caused by increasing the number of mass-to-charge ratios measured can be seen for Pb in Figure 2. For Cu, a completely similar behavior was observed. The average values for the integrated signals and the RSDs obtained under the different conditions are shown in Table 3. The results show that on changing from 1 to 11 mass-to-charge ratios monitored, the sensitivity was only reduced by ∼35%, while even when measuring 30 mass-to-charge ratios (only 3 data points to describe each signal profile), the signal intensity (peak area) was (29) Denoyer, E. R. At. Spectrosc. 1994, 15, 7-16.
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4313
Figure 2. ETV-ICPMS signal profiles for Pb as a function of the number of mass-to-charge ratios monitored: (a) 1, (b) 3, (c) 7, (d) 11, (e) 15, (f) 30. Table 3. Differences in Sensitivity and Precision as a Function of the Number of Mass-to-Charge Ratios Monitoreda no. of mass-to-charge ratios monitored (Points measured to define the signal, assuming the signal duration to be 3 s) 1 (30)
3 (17)
7 (8)
11 (6)
15 (5)
30 (3)
Solution ETV-ICPMS (Amount of Each Element Introduced, 5 ng) Cu Pb
1.07 (0.03) 1.22 (0.08)
Cu Pb
4.3 (1.7) 6.0 (3.3)
integrated signal intensity (in million counts) and standard deviations for n ) 15 0.83 (0.08) 0.83 (0.04) 0.66 (0.02) 0.49 (0.02) 0.88 (0.04) 0.90 (0.05) 0.85 (0.04) 0.54 (0.07) 5.9 (2.8) 6.1 (2.1)
average RSD% values for 3 replicate analyses 5.8 (0.8) 4.8 (2.4) 6.2 (0.1) 4.2 (1.4)
18.5 (5.4) 10.7 (3.3)
average RSD% values for 3 replicate analyses 16.5 (0.6) 12.5 (2.3) 10.3 (2.0) 13.2 (1.1)
0.38 (0.02) 0.17 (0.01)
4.5 (1.3) 10.9 (2.0)
7.1 (0.7) 10.7 (0.7)
15.1 (0.2) 21.4 (0.6)
15.7 (0.5) 17.7 (1.1)
Solid Sampling ETV-ICPMS Cu Pb a
16.2 (1.1) 12.2 (1.1)
Standard deviations given in brackets.
still ∼15-35% of that observed under single ion monitoring conditions. In fact, this reduction in sensitivity is only a result of the fact that on reducing the number of points to describe the signal profile, the probability of missing the peak’s maximum is increased. It is even more interesting to study how the number of data points describing the signal profile affects the precision of a determination. For Cu, the standard deviation (RSD% for n ) 15) for the measurements remains almost constant and, only when 4314
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
monitoring 30 mass-to-charge ratios, some deterioration of the precision was observed. In addition, analysis of variance (ANOVA) showed that this difference in RSD is not significant at the 95% confidence level. For Pb, the RSD value remained similar as long as e11 mass-to-charge ratios are monitored, and a significant (ANOVA) deterioration of the precision could only be observed when measuring g15 mass-to-charge ratios. Moreover, the RSDs observed when monitoring (i) 15 and (ii) 30 mass-to-charge ratios were established to be similar. Conclusively, it can be stated that,
for solution ETV-ICPMS, it is possible to simultaneously measure at least 11 elements without significant sacrifices in terms of precision and sensitivity. In solid sampling analysis, a similar or even less pronounced influence might be expected. It has been shown that the nonhomogeneity of the small samples (typically ∼1 mg) introduced into the graphite furnace is the cause for the high RSD values usually reported.30 Hence, it is not clear if the contribution of the distortion of the signal will be significant at all when compared with this large inhomogeneity term. To study this fact, a similar experiment as described for solution ETV-ICPMS was carried out using solid sampling ETV-ICPMS. The results are also displayed in Table 3. For both elements studied, the situation appears to be similar to that for solution ETV-ICPMS. For Pb, an increase in the RSD% could be noticed when monitoring g15 mass-to-charge ratios. ANOVA revealed that this difference is statistically significant at the 95% level. However, it should be stressed that, for screening purposes, the difference between working with a precision of ∼10% (1-11 elements) or ∼20% (1530 elements) would not be very important. For Cu, all RSD values are quite similar and ANOVA revealed no significant differences. In general, the results can be considered as very promising, as they demonstrate the suitability of the method for determining simultaneously (from the same vaporization step) at least 11 elements, of course provided that a suitable temperature program can be found. For determinations involving a higher number of elements, a decrease in the precision of the results might be expected. A more detailed study concerning this point may be worth carrying out and will be the aim of further studies. In this context, it can be mentioned that, with an ICPMS instrument, equipped with a time-of-flight (TOF) mass spectrometer, there is no limitation at all. In this case, the analyte elements are simultaneously extracted from the ion source and introduced into the mass analyzer, while many thousands of spectra can be monitored per second.31 Hence, even for signals of very short duration, ICP-TOFMS permits the monitoring of signal profiles for many elements without spectral skew. Anyway, it is clear that the distortion of the signals observed with our quadrupole-based ICPMS instrument should not affect the goal of the present works the simultaneous determination of Al, Ba, Cd, Cu, Mn, Pb, and Ti in polyethylene. The signal profiles obtained when analyzing 0.5 mg of PERM are shown in Figure 3. As a result of the fairly simple matrix and the possibility of complete matrix removal prior to analyte vaporization, unimodal peaks were observed for all elements, while the peak shapes were not affected upon addition of aqueous standard solution.32 Influence of the Sample Mass. In a previous paper,28 the difference between solid sampling GFAAS and solid sampling ETV-ICPMS in terms of mass range was pointed out. For solid sampling GFAAS, a linear relation between signal intensity and sample mass is only observed in a limited range. For solid sampling ETV-ICPMS, this range was shown to be significantly wider, such that time-consuming optimization can be avoided. This (30) Belarra, M. A.; Resano, M.; Castillo, J. A. J. Anal. At. Spectrom. 1998, 13, 489-494. (31) Mahoney, P. P.; Ray, S. J.; Hieftje, G. M. Appl. Spectrosc. 1997, 51, 16A28A. (32) Vanhaecke, F.; Boonen, S.; Moens, L.; Dams, R. J. Anal. At. Spectrom. 1995, 10, 81-87.
Figure 3. ETV-ICPMS signal profiles for 0.50 mg of solid sample (low-level PERM) using 5 µg of Pd.
Figure 4. Normalized signal intensity (peak area/sample mass) as a function of the sample mass for Pb.
fact and the possibility of the OmniRange device to adjust the sensitivity permit determination of analytes in a wide concentration range. In Figure 4, the relation between the peak area for Pb (normalized to 1-mg sample mass) and the sample mass is shown. Similar results were obtained for the other elements. The normalized signal was seen to be constant until a sample mass of about 4 mg was exceeded. Hence, the working range is about one order and a half of magnitude, confirming previous results.28 The signal suppression that occurs with higher amounts of sample could be demonstrated by monitoring the Ar dimer (40Ar2+) ion signal during the measurements.33 This suppression is caused by the relatively high concentration of inorganic compounds present in the sample that will reach the plasma during the vaporization step. All organic compounds, on the other hand, can be removed during the pyrolysis step. Analysis Results. Finally, the determination was carried out using the conditions shown in Table 1. Two calibration methods permitting a high sample throughput were tested: (i) external calibration using an aqueous standard and (ii) single standard addition, which was proven to be practicable and successful in previous work.32-35 Every determination consisted of the measure(33) Vanhaecke, F.; Galba´cs, G.; Boonen, S.; Moens, L.; Dams, R. J. Anal. At. Spectrom. 1995, 10, 1047-1052. (34) Boonen, S.; Vanhaecke, F.; Moens, L.; Dams, R. Spectrochim. Acta, Part B 1996, 51, 271-278. (35) Hu, Y.; Vanhaecke, F.; Moens, L.; Dams, R.; Geuens, I. J. Anal. At. Spectrom. 1999, 14, 589-592.
Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
4315
Table 4. Results Obtained for the Determination of Al, Ba, Cd, Cu, Mn, Pb, and Ti in Low-Level PERMa Al
Ba
standard addition RSD% external calibration RSD%
18.3 ( 1.4 (8.0) 16.8 ( 1.4 (8.8)
286 ( 21 (7.9) 418 ( 36 (9.2)
RSD%
17.90 ( 0.22 (1.1)
284.0 ( 2.7 (0.9)
Cd
Cu
solid sampling ETV-ICPMS 21.6 ( 1.1 13.0 ( 1.0 (5.7) (8.6) 23.0 ( 0.9 13.0 ( 1.1 (4.2) (9.0) NAA22 21.72 ( 0.31 (1.4)
13.41 ( 0.39 (2.8)
candidate certified values 21.73 ( 0.33
Mnb 4.59 ( 0.38 (9.0) 5.26 ( 0.45 (9.2)
Pb 13.8 ( 0.5 (4.3) 13.0 ( 0.9 (7.3)
5.03 ( 0.75c (12)
Ti 503 ( 37 (8.0) 527 ( 34 (7.0) 517.1 ( 3.3 (0.6)
13.76 ( 0.30
a Every experimental result is the average of seven (solid sampling ETV-ICPMS) and six (NAA) determinations, respectively. Candidate certified values are the result of an international certification campaign, organized by the SM & T Program of the European Community. All results are expressed in micrograms per gram, except for Mn (nanograms per gram), and the uncertainties are expressed as 95% confidence intervals and as relative standard deviations (RSD% in brackets). b (ng g-1). c n ) 5.
ment of five solid samples and (i) five aliquots of the aqueous standard (external calibration) or (ii) five solid samples to which an aliquot of standard solution was added (single standard addition). For both approaches, approximately 30 min is sufficient to carry out the simultaneous determination of the seven elements. The results are presented in Table 4, in which the results obtained using NAA,22 a very powerful solid sampling method, that can of course not be used for rapid routine applications, and the candidate certified values for the material analyzed have also been presented. With regard to the accuracy of the results, it can be concluded that, in (almost) all cases, the result obtained using single standard addition is closer to the reference value (the NAA result, or for Cd and Pb, the value proposed for certification) than the result obtained using external calibration. Additionally, when using single standard addition, no significant differences (95% confidence level) could be established between the ETV-ICPMS results and the corresponding reference values. When using external calibration on the other hand, the experimental results for Ba and Cd (borderline case) differ significantly from the reference value. The overestimated results for Ba obtained when using external calibration suggest that remaining matrix components improve the transport efficiency for Ba. Hence, single standard addition can be considered as a more universal approach, while it does not negatively affect the sample throughput. Nevertheless, it can be pointed out that the results obtained when calibrating against an aqueous standard can be considered as surprisingly good, as in previous works much larger deviations were reported.25,32,36,37 However, in the present case, the sample matrix was fairly simple and could be suffciently removed during pyrolysis. More complex matrixes may give origin to more pronounced signal suppression and, hence, larger bias when using external calibration. With regard to the precision, the results for both calibration approaches were very similar, illustrating that the precision is governed by the nonhomogeneity of the small samples vaporized and is thus hardly affected by the calibration method used. The corresponding F-tests did not reveal any significant difference. If (36) Wang, J. S.; Carey, J. M.; Caruso, J. A. Spectrochim. Acta, Part B 1994, 49, 193-203. (37) Hinds, M. W.; Gre´goire, D. C.; Ozaki, E. A. J. Anal. At. Spectrom. 1997, 12, 131-135.
4316 Analytical Chemistry, Vol. 72, No. 18, September 15, 2000
the precision observed for the single standard addition method is compared with that of the NAA results, the differences are clearly significant at the 95% confidence level in all cases, except for Mn. Of course, it should be taken into account that, for NAA, the sample intake was 1.5 g, while with ETV-ICPMS, only 0.5-1 mg of sample is consumed per firing. For Mn, the precision of the ETV-ICPMS result is better than that of the NAA value, although the difference is not statistically significant at the 95% confidence level. This fact may be explained by the low Mn concentration (about 5 ng/g), which makes its measurement using NAA not simple (low signal-to-background ratio) and shows the high suitability of solid sampling ETV-ICPMS to deal with element concentrations at the ultratrace level. The absolute limit of detection (LOD) for Mn was established to be 0.3 pg, which corresponds to a relative value of 0.3 ng/g for a typical sample mass of 1 mg. CONCLUSIONS It has been demonstrated that when using single standard addition for calibration, accurate results for the content of Al, Ba, Cd, Cu, Mn, Pb, and Ti in a polyethylene sample can be obtained using solid sampling ETV-ICPMS. Hence, calibration is more straightforward than with LA-ICPMS, while it is also possible to isolate the analyte elements from the matrix. Additionally, it has also been shown that the multielement capabilities of ETV-ICPMS may have been underestimated until now, as it is possible to monitor at least 11 elements, without making an important sacrifice in terms of precision and sensitivity. Of course, in contrast to LA-ICPMS, solid sampling ETV-ICPMS cannot be used for spatially resolved analysis and its application range will be more limited. ACKNOWLEDGMENT The authors would like to sincerely acknowledge Perkin-Elmer for lending them the HGA-600 MS ETV-unit and A. van Alphen (Accordis, The Netherlands) for mofifying this unit as described in the paper. Thanks are also due to Dr. Frans De Corte (Laboratory of Analytical Chemistry, Ghent University) for the permission to use his NAA analysis results for the low-level PE candidate CRM as reference values. Received for review March 7, 2000. Accepted July 6, 2000. AC0002700
