摘要:

Comparison of Ultraviolet Laser Ablation and Spark Ablation of Metals and Alloys for Analysis by Axially Viewed Inductively Coupled Plasma Atomic Emission Spectrometry M. GAGEAN AND J. M. MERMET L aboratoire des Sciences Analytiques (CNRSUMR 5619), Ba�t. 308, Universite� Claude Bernard—Lyon I, F-69622 V illeurbanne Cedex An excimer UV laser system and a spark system have been selection was, therefore, not necessarily the most suitable for the analysis of metals and alloys. Line selection is summarized used for the ablation of metals and alloys as a means of sampling for axially viewed ICP-AES.Ablation efficiency and in Table 1. A refractor plate was used for background correction. Because of the design for trace determination, background sensitivity expressed as S/B are presented for different types of materials. Ablation efficiency depends on the type of material correction led to correct values of the net line intensity for low element concentrations, i.e.equal to or below 1 ppm. For and cannot be compensated for by using an internal standard. Cast iron and steel are among the materials that are the most higher concentrations, the wavelength excursion range was not necessarily large enough, and the background measurement difficult to ablate. At least for bulk analysis, spark ablation is a good compromise between investment cost and sensitivity. could include part of the line wing. A power of 950W and a plasma gas flow rate of 15 l min-1 were used for both spark The pre-ablation time was found to be shorter for spark ablation than for laser ablation.Laser ablation requires less and laser ablation. Most previously published work dealing with laser ablation surface preparation and may be preferred for micro-analysis. was based on the use of a Nd5YAG laser operating at the Keywords: Metals; alloys; laser ablation; spark ablation; fundamental wavelength of 1064 nm. It has been shown16 that inductively coupled plasma; atomic emission spectrometry; the use of lasers operating in the UV region of the spectrum axial viewing minimized possible selective volatilization and enhanced ablation efficiency.In this work, a Lambda-Physik LPX 110i The elemental analysis of conductive solid materials such as 308 nm XeCl excimer laser was used. The operating conditions metals and alloys is usually performed by analytical spectro- were a repetition rate of 10 Hz and an energy in the 175– scopic methods such as spark emission spectrometry.In con- 215 mJ range. Although the maximum repetition rate could be trast, ICP-AES has gained wide acceptance for the analysis of adjusted to 100 Hz, a 10 Hz rate was selected in order to samples in the form of solutions. However, there has been a increase the lifetime of the gas mixture used in the excimer trend towards the expansion of the field of application of ICP- laser and to avoid a load excess in the plasma. These conditions AES by using direct solid sampling methods such as spark did not result in the best possible S/B, but were chosen for ablation for conductive materials or laser ablation for any type routine work.The 12 mm×24 mm rectangular laser beam was of material. This makes it possible to avoid complex and focused on the target via a plano-convex lens with a focal tedious sample dissolution procedures, thereby resulting in a length of 20 cm. The elliptical spot size was 0.5 mm×1 mm, significant reduction in the sample preparation time, particu- resulting in an average fluence of 10 J cm-2.The target was larly for laser ablation. moved with respect to the laser beam by means of two Not only does laser ablation enable bulk analysis but it also computer-driven dc motors. The displacement corresponded enables the micro-analysis of samples with high lateral reso- to a 5 mm stroke and the speed was set up at 1 mm s-1. lution, usually below 100 mm.However, the investment cost is Surface preparation consisted of a rough polishing with grade significantly higher than that of spark ablation. Spark ablation 400 polishing paper. ICP-AES seems to be an interesting alternative1–9 to laser Two different ablation cells were used: one for samples with ablation ICP-AES10–15 for the bulk analysis of conductive a diameter of 16 mm, and another for samples with a diameter materials such as metals and alloys. The purpose of this work of 100 mm.A 0.65 l min-1 Ar carrier gas flow rate was used was, therefore, to evaluate the sensitivity of UV laser ablation to take up the ablated material to the ICP. Few modifications ICP-AES expressed as S/B for metals and alloys, and to study the influence of the material on the ablation efficiency of Table 1 Line selection of the Thermo Jarrell Ash 61 E Trace ICP selected elements. The S/B values were compared with those system obtained using spark ablation ICP-AES. An axially viewed plasma was used in order to obtain the best possible sensitivity.Element Line/nm Element Line/nm Ag I 328.068 Mn II 257.61 EXPERIMENTAL Al I 396.15 Ni II 231.60 As I 189.04 P I 178.287 A Thermo Jarrell Ash ICP 61 E Trace ICP system was used. Bi I 223.06 Pb II 220.353 The major feature of the system was the axial viewing of the Cd II 226.50 Sb I 206.838 Cr II 267.716 Si I 251.61 observation zone in the plasma in order to improve the S/B Cu I 324.75 Sn II 189.99 values, and consequently, the LOD.The dispersive system was Co II 228.62 Ti II 334.94 a polychromator with photomultiplier tubes for detection. Fe II 259.94 Zn I 213.86 Originally, the TJA 61 E Trace system was designed principally Mg II 279.55 for the determination of trace elements in water. The line Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (189–193) 189Table 3 Influence of the type of material on the volume and mass of were made to the laser ablation system to adapt it for the ICP ablated material per second using a XeCl excimer laser (10 Hz, 200 mJ) system.The ablated material was transported along a polyamide tubing directly to the standard cyclonic spray chamber Material Volume/mm3 s-1 Mass/mg s-1 of the system using the inlet normally used for the pneumatic Steel 0.44×106 3.0 nebulizer. The id of the tubing was 4 mm. The amount of Brass 1.32×106 10.8 ablated material was measured by using a Rank Talysurf Aluminium 5.8×106 13.6 profilemeter.Glass 8.2×106 19.4 In order to obtain a stable signal and high values for S/B, Polymer 28.4×106 50.0 the pre-integration time was typically at least 5 min. This can be explained not only by the need to perform preablation but also by the large dead volume of the ablation cell. Ten From the results shown in Table 3, it can be seen that the replicates, with an integration time of 20 s each, were used. It efficiency of ablation depended strongly on the type of material.should be noted that if low RSD values of the signal had to Steel and cast iron were always more difficult to ablate than be obtained, a pre-integration time of at least 10 min was any other material. Generally, metals or alloys, in particular found to be necessary. brass or aluminium (Table 3) were more difficult to ablate A standard Thermo Jarrell Ash spark ablation system was than materials such as glass or polymer (PVDF).This variation used. The electronic controlled waveform source makes it in ablation efficiency has consequences on the S/B and LOD.possible to control the voltage level by step (40, 60, 80 or It is interesting to note that the results differed from those 100%) and the number of breaks per half-cycle (2, 3, 4 or 5). obtained for pure metals16 where similar ablation efficiencies In contrast to laser ablation, the adjustment has been optimized were found, regardless of the metal. Alloys do not behave like for each material in order to obtain the best possible S/B their pure constituents.values. A cyclonic spray chamber with a double entrance was used, one for the tubing coming from the spark ablation system, the other one for a concentric pneumatic nebulizer. As SBR Values for laser ablation, the id of the tubing was 4 mm. It should be Because of the various ablation efficiencies, variation was noted that the deposition of particles wase important for expected in the S/B values for the different types of materials spark ablation than for laser ablation.A pre-ablation time of used in this work. Results are summarized in Table 4 for Al, 20 s was used. The integration time was set up either to 1 or Cu, Fe, Mn, Ni, Sn and Zn. The variation in the S/B values is 10 s, depending on the signal intensity. A series of 10 replicates rather complex. Usually, S/B were improved between materials was also used. Surface preparation was carried out with the difficult to ablate such as cast iron and materials easier to same polishing paper as used for laser ablation.The operation ablate such as Pb, but not necessarily to the same extent. For of the spark ablation system was driven by the software of the instance, the S/B values for Cu and the Ni were improved by ICP system. a factor of 18 and 15, respectively, whereas for Zn the S/B was Several types of sample were selected: cast iron, steel, bronze, lower for a Pb matrix with a cast iron matrix.Clearly, Zn brass, zamack (zinc alloy), aluminium, Cu–Al alloy, and lead. behaved differently from Ni, particularly in the bronze sample. Details of their origin and composition are given in Table 2. It is interesting to compare these values with those obtained using the same ICP system and pneumatic nebulization of solutions. For instance, for Ni, the S/B values were in the RESULTS AND DISCUSSION 10–20 range for Ni solutions of 1 mg l-1 using a concentric Amount of Laser Ablated Material pneumatic nebulizer. Assuming a nebulizer efficiency of 3% for an uptake rate of 1.5 ml min-1, 1 mg l-1 corresponds to The amount of ablated material was estimated by measuring the volume of the crater and by using the density of the 45 ng min-1 of Ni.In the case of steel, 100 ppm results in 21 ng min-1 of Ni (Table 3) assuming a transport efficiency of material. Results expressed as the volume and the mass of ablated material per second (for 10 Hz and 200 mJ) are 100%.Practically, this transport efficiency was lower, i.e., in the 30–50% range. It can be considered then that the ratio of summarized in Table 3. Table 2 Composition (ppm) of the samples used in this work Cast iron Steel Bronze Brass Zinc Al Cu–Al MRDF Sollac Ctif Ctif NIST Pe�chiney Ctif Pb Element XXI 31506 UE 15–1 UZ34/1 630F 601 2154V 2288 Ag 34 Al 100 320 200 1100 43000 As 1200 Bi 170 Cd 48 Co 11.7 Cr 770 110 31 7.1 Cu 60 9760 26.4 Fe 1700 7200 230 32.5 30500 Mg 70 300 28 Mn 2500 1890 650 2150 1060 7.8 1200 Ni 170 2500 2600 27 7.5 4100 2 P 460 1200 Pb 83 50 Sb 1150 Si 10000 10 1050 2900 21.6 150 Sn 80 30 10100 40 8.5 100 Ti 270 1260 6.8 Zn 170 8100 22.3 100 190 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Table 4 Effect of the type of material on the S/B values of Al, Cu, Fe, Mn, Ni, Sn, and Zn. The SBRs were normalized to 100 ppm. An XeCl laser was used with an energy of 200 mJ and a repetition rate of 10 Hz.The ICP system was the TJA 61E Trace Element Cast iron Steel Bronze Brass Zinc Al Cu–Al Pb Al 1.7 7.7 3 5 0.8 Cu 6 26 6.8 46 108 Fe 4.9 5 15 6.6 Mn 4.6 30 12.9 8.9 19.5 7 Ni 1.45 1.7 2.5 4 4.1 4.4 2.8 22 Sn 1.8 2.0 2.2 0.14 Zn 32.6 1.1 40 24.3 11 the Ni S/B obtained using solutions to that obtained using the Ni and Mn concentration could vary over a large range of values. ablation may be, in the first approximation, explained by the ratio of the amount of Ni reaching the plasma.A consequence of the variation in the S/B as a function of the material would be the difficulty of using one type of Limits of Detection for Laser Ablation material as a standard for calibration of any other type of material, even using an internal standard. For instance, use The LOD were determined following Boumans’ approach17 where both the S/B and the RSD of the background, RSDB , of Mn as an internal standard for Fe and Ni would not result in a similar compensation for each material (Fig. 1). The line were used: intensity ratios were normalized to 1 ppm. From Fig. 1, it can be seen that the Fe5Mn and the Ni5Mn ratios depended LOD= 3cRSDB S/B strongly on the type of material. The Ni5Mn ratio varied in the 0.15–1.3 range, i.e., a variation of almost a factor of 10, where c is the concentration used for the experiment. while the Fe5Mn variation was in the 0.36–0.57 range. It can The RSDB values were particularly low compared with be concluded that the ablation efficiency depended both on previously published work10–15 where the values were typically the type of matrix and on the element.It should also be noted in the 2–5% range. For instance, the average RSDB values that some errors in the determination of the net line intensities were 0.62, 0.47, 0.59, 0.71, 0.53, 0.46 and 0.46% for steel, cast could occur as mentioned in the experimental section, since iron, brass, zinc alloy, bronze, aluminium and lead, respectively.They were close to those obtained for aqueous solutions using the same ICP system, i.e., 0.2–0.3%. It was found that the LOD were usually in the sub-ppm range (Table 5). The best results were obtained for the Al and Pb matrices. Moreover, Cu was very often a sensitive element. These LOD were improved when compared with the LOD obtained by using the conventional radial viewing. The average improvement factor was a factor of 5 for the steel sample and of 20 for the aluminium one. The overall improvement can be explained by both the axial viewing and the use of a UV laser with a high available energy, such as the excimer laser.These values compared well with those previously obtained using ICP-AES,10–15 MIP-AES,18 dc plasma-AES19 or ICP-MS.20–22 Comparisons for Cu10,12,14,19,20 and Mn10,11,12,20 are summarized in Figs. 2 and 3, respectively. Note that even if the LOD in this work can be considered as excellent, they could be Fig. 1 Effect of using Mn as an internal standard to compensate for significantly improved by using a higher repetition rate (i.e., the intensities of Fe and Ni in various materials.Intensities were 100 Hz). As mentioned above, this would, however, reduce the normalized to 1 ppm in each case. Fe–Mn: solid bar; Ni–Mn: crosshatched bar. lifetime of the gas mixture used to run the excimer laser. Table 5 LOD (ppm) observed for various matrices using laser ablation. An XeCl laser was used (10 Hz, 200 mJ) along with a TJA 61E Trace ICP system Element Cast iron Steel Brass Zinc alloy Bronze Aluminium Lead Ag 0.015 Al 0.8 0.5 0.35 2.6 0.1 As 2.1 0.65 Bi 0.85 0.5 Cd 0.2 1.7 Cu 0.2 0.05 0.3 0.03 0.01 Cr 2.2 2.8 0.5 0.6 Co 0.36 Fe 0.35 0.15 0.2 0.2 Mn 0.4 0.15 0.25 0.05 0.09 Mg 0.2 0.002 Ni 1.0 1.1 0.45 0.5 0.4 0.3 0.04 Si 0.4 0.8 0.45 0.15 Sn 0.8 1.0 0.9 1.0 Ti 0.2 0.3 0.1 Zn 0.04 1.4 0.03 0.1 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 191Table 8 Comparison of the S/B values (normalized to 100 ppm) obtained by laser and spark ablation (80% maximum voltage, 3 breaks per half cycle, integration time of 10 s) for brass samples Element Laser ablation Spark ablation Al 5.0 21.1 As 2.7 2.4 Fe 5.0 6.0 Mn 12.9 19.1 Ni 4.0 4.5 Si 2.2 4.2 Sn 2.0 1.1 Table 9 Comparison of the S/B values (normalized to 100 ppm) obtained by laser and spark ablation (80% maximum voltage, 4 breaks per half cycle, integration time of 1 s) for Zn-alloy samples Fig. 2 Comparison of the LOD for Cu obtained by laser ablation in this work (for cast iron and steel, solid bars) with previously published Element Laser ablation Spark ablation work (cross-hatched bars). Al 0.8 0.6 Cd 11.3 14.6 Cr 4.2 7.1 Fe 15.0 30 Mg 160 190 Mn 8.9 10.8 Pb 1.6 0.6 Sn 2.2 2.0 Table 10 Comparison of the S/B values (normalized to 100 ppm) obtained by laser and spark ablation (80% maximum voltage, 3 breaks per half cycle, integration time of 10 s) for Cu-Al samples Element Laser ablation Spark ablation Mn 7.5 5.5 Ni 3.0 1.0 Fig. 3 Comparison of the LOD for Mn obtained by laser ablation Si 4.8 1.9 in this work (for cast iron and steel, solid bar) with previously Zn 23.8 74.4 published work (cross-hatched bar). Comparison with Spark Ablation Table 11 Comparison of the S/B values (normalized to 100 ppm) A first lookt Tables 6–11 indicates that the S/B values obtained by laser and spark ablation (60% maximum voltage, 3 breaks per half cycle, integration time of 10 s) for Pb samples obtained using laser and spark ablation were comparable, with some notable exceptions.In the case of cast iron, the Al S/B Element Laser ablation Spark ablation was significantly higher for spark ablation, while the Mn S/B Ag 87 188.7 was far lower (Table 6). For bronze (Table 7) and brass Bi 5 9.0 (Table 8) samples, it was found that the Al S/B value was also Ni 22.1 60.1 far higher when spark ablation was used. This behaviour for Table 6 Comparison of the S/B values (normalized to 100 ppm) obtained by laser and spark ablation (80% maximum voltage, 3 breaks Al was not found in the case of the Zn-alloy sample (Table 9), per half cycle, integration time of 1 s) for cast iron samples where the Al S/B were similar for both types of ablation.Spark ablation was found to be slightly less efficient than laser Element Laser ablation Spark ablation ablation for the Cu–Al sample (Table 10), except for Zn. In Al 1.7 24.6 contrast, spark ablation was more efficient for the Pb sample Cr 0.65 0.6 (Table 11).Clearly, the majority of the elements behave simi- Mg 7.4 19.6 larly for both types of ablation, with some exceptions for Mn 4.0 0.3 elements such as Al and Mn. However, there was no general Sn 1.8 1.9 Ti 7.8 15.0 rule concerning elements such as Al and Mn. Their behaviour Zn 32.6 76.5 seemed to depend on the matrix. These results confirmed that Mn was not an adequate internal standard.Due to the software-based compulsory flushing procedure Table 7 Comparison of the S/B values (normalized to 100 ppm) with a wet aerosol between each spark ablation replicate, obtained by laser and spark ablation (80% maximum voltage, 3 breaks estimation of the RSDB led to values in the range of 5–20%, per half cycle, integration time of 10 s) for bronze samples probably because the equilibrium time between the wet and Element Laser ablation Spark ablation the dry aerosols was too short.A manually-driven operation of the spark system led to values in the range 0.5–1%, which Al 3.0 51.5 Fe 4.9 13.7 corresponded to LOD in the ppm range or below. This range Mn 30 34.2 can be compared to those given in previously published work Ni 2.5 4.0 where the LOD were in the 1–10 ppm for low-alloyed steel,3 Si 3.6 12.6 1–5 ppm for steel,5 1–30 ppm for Al–Si alloys,2 0.24–24 ppm Zn 1.1 0.7 for low-alloyed steel and 1–9 ppm for Al alloys.4 192 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 124 Prell, L. J., and Koirtyohann, S. R., Appl. Spectrosc., 1988, 42, 1221. CONCLUSIONS 5 Go� mez Coedo, A., Dorado Lo�pez, M. T., Jimenez Seco, J. L., Both spark and laser ablation provides good sensitivity in and Gutieriez Cobo, I., J. Anal. At. Spectrom., 1992, 7, 11. 6 Go� mez Coedo, A., Dorado Lo�pez, M. T., Gutieriez Cobo, I., and ICP-AES. The advantages of spark ablation ICP-AES are a Escudero Baquero E. E., J. Anal. At. Spectrom., 1992, 7, 247.low investment cost, ease of use and a relatively short pre- 7 Go� mez Coedo, A., Dorado Lo�pez, M. T., Escudero Baquero, E., integration time. The highest sensitivity permitted by the and Gutieriez Cobo, I., J. Anal. At. Spectrom., 1993, 8, 827. method can be achieved on a routine basis. However, spark 8 Go� mez Coedo, A., Dorado Lo�pez, M. T., Rivero, C. J., and ablation may require more careful surface polishing. Spark Gutieriez Cobo, I., J. Anal. At. Spectrom., 1993, 8, 1023.ablation is only suitable for bulk analysis because of the large 9 Webb, C., Cooper III, C. B., Zander, A. T., Arnold, J. T., Lile, E. S., and Anderson, S. E., J. Anal. At. Spectrom., 1994, 9, 263. spot diameter that is unavoidable. Laser ablation requires little 10 Thompson, M., Goulter, J. E., and Sieper, F., Analyst, 1981, surface preparation and may provide slightly better sensitivity 106, 32. and LOD than spark ablation when it is necessary, by using 11 Kawaguchi, H., Xu, J., Tanaka, T., and Mitzuike, A., Bunseki a higher repetition rate.It should be noted that this is true Kagaku, 1982, 31, E185. only for an excimer laser and not for an Nd5YAG laser 12 Ishizuka, T., and Uwamino, Y., Spectrochim. Acta, Part B, 1983, working in the UV (third or fourth harmonic) because of a 38, 519. 13 Tremblay, M., Leong, M., and Winefordner, J., Spectrosc. L ett., limited available energy. Laser ablation exhibits some limi- 1987, 20, 311.tations such as a long pre-integration time, which can be 14 Richner, P., Borer, M., Brushwyler, K., and Hieftje, G. M., Appl. minimized by a better cell design, a high investment cost, Spectrosc., 1990, 44, 1290. particularly with an excimer laser and the need for mainten- 15 Gagean, M., and Mermet, J. M., in Progress in Analytical ance. However, laser ablation can be used for the analysis of Chemistry in the Steel and Metals Industry, ed. Nauche, R., samples with complex shapes, the determination of elements European Commission, Luxembourg, 1996, p. 146. 16 Geertsen, C., Briand, A., Chartier, F., Lacour, J. L., Mauchien, P., in layers and inclusions, and, obviously, for non-conductive Sjo�stro�m, S., and Mermet, J. M., J. Anal. At. Spectrom., 1994, 9, 17. materials. 17 Boumans, P. W. J. M., Anal. Chem., 1994, 66, 459A. 18 Ishizuka, T., and Uwamino, Y., Anal. Chem., 1980, 52, 125. The authors are grateful to Thermo Jarrell Ash for the loan 19 Mitchell, P., Ruggles, J., and Sneddon, J., Anal. L ett., 1985, of the 61 E Trace ICP and the spark ablation systems. 18, 1723. 20 Arrowsmith, P., Anal. Chem., 1987, 59, 1437. 21 Huang, Y., Shibata, Y., and Morita, M., Anal. Chem., 1993, REFERENCES 65, 2999. 22 Raith, A., Hutton, R. C., Abell, I. D., and Crighton, J., J. Anal. 1 Human, H. G. C., Scott, R. H., Oakes, A. R., and West, C. D., At. Spectrom., 1995, 10, 591. Analyst, 1976, 101, 265. 2 Aziz, A., Broekaert, J. A. C., Laqua, K., and Leis, F., Spectrochim. Paper 6/05457I Acta, Part B, 1984, 39, 1091. Received August 5, 1996 3 Lemarchand, A., Labarraque, G., Masson, P., and Broekaert, J. A. C., J. Anal. At. Spectrom., 1987, 2, 481. Accepted November 13, 1996 Journal of Analytical Atomic Spectrometry, Februa

ISSN:0267-9477    

DOI:10.1039/a605457i

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Use of Laser Ablation Inductively Coupled Plasma Mass Spectrometry for Fingerprinting Scene of Crime Evidence R. J. WATLING*, B. F. LYNCH AND D. HERRING Chemistry Centre of Western Australia, 125 Hay St, East Perth, Western Australia 6004 The requirement to uniquely characterise and compare analysis at afuture date if required by an independent authority physical evidence from crime scenes is a major task in forensic for corroborative purposes. Furthermore, if the analysis is to science.Laser ablation inductively coupled plasma mass be elemental in nature, the technique should allow investigation spectrometry (LA–ICP-MS) was investigated for its potential of a wide range of elements to enable a maximum discrimito provide data on relative trace elemental compositions to nation between samples of the same type. High sensitivity achieve this aim. Glass and steel samples were examined as LA–ICP-MS has been used to determine the elemental associthey frequently occur as physical evidence and represent two ations in samples as small as 50 mm in diameter and to provide distinctly dissimilar sample types.A fine focus Nd5YAG laser ‘fingerprint’ data on the trace and ultra-trace element assemwas used enabling specimens of approximately 50 mm in blages present in glass and steel. These associations of elements, diameter to be examined. Ablation protocols and optimum rather than the traditional element concentrations, form the compromise sets of laser parameters were established for the basis of the approach for the analysis of scene of crime material analysis of both sample types using both free running and by LA–ICP-MS. Q-switched mode of laser operation.Mass spectra acquired Glass particles are often recovered as physical evidence. under these conditions were reproducible and were generated in When glass is broken, it shatters into fragments and minute a fraction of the time required for the conventional solution particles that are often distributed over a very large area.1,2 In analyses.Sixty-two glass samples were examined of which addition, small glass particles can become embedded in shoes thirty-one were float glasses, four were sheet glasses and or adhere to and be retained on clothing, especially woollen twenty-seven were container glasses. The steel samples garments.3 Movement may cause some of the small glass examined were drillings from sixty-nine sources and included particles retained on garments to be transported from clothing steel from safes, firearm barrels, tools, angle iron, rods and to other property, for example, a car or house.The same is crowbars. The LA–ICP-MS method is at present an essentially true in incidents of safe breaking due to the generation of qualitative technique and relies on comparison of trace element swarf or other cutting debris.4 assemblages or ratios. Samples can be conveniently compared Methods of sample identification and analysis currently by direct overlay of spectra or interpretive software can be available for glass and steels include the determination of used.Software facilitating the inter-comparison of three various physical properties such as the refractive index of elements simultaneously (ternary plots) in large groups of glasses,5 classical wet chemistry,6 optical microscopy,7,8 scan- samples was used to establish both the reproducibility of the ning electron microscopy energy dispersive X-ray spec- ‘fingerprint’ and the uniqueness of the inter-element trometry,9 XRF,10 FAAS,11 ICP-AES,10 and ICP-MS.4,12 associations.Results have shown that robust analytical Commonly, differentiation between glass particles has relied procedures have been developed which reliably discriminate on measurement of values of refractive index in association both steel and glass samples and could have direct application with determination of major and minor element composition for the examination of a wide range of other crime scene by scanning electron microscope–microprobe (both non- evidence.destructive techniques). This produces a conclusive accurate Keywords: L aser ablation inductively coupled plasma mass differentiation between two glass fragments that exhibit differspectrometry ; scene of crime evidence; forensic evidence; trace ent refractive indices. However, difficulties arise when element fingerprinting attempting to discriminate between two modern window glasses that, for example, have essentially the same gross major and minor elemental composition and which have refractive Many criminal activities result in the generation of debris, or indices falling in a very narrow range. Australian float glasses other material which becomes available to investigating auth- tend to lie within the range 1.5189 to 1.5194 for example.orities, as physical evidence of the crime. However, the gener- Although many glasses have similar gross elemental composi- ation of traditional analytical and forensic chemical data is tion, even similar glasses could have different trace and ultra- often costly and time consuming and the resources of police trace element signatures.Variations in the levels of these forces, in most countries, have become so stretched that only elements occur when the glass is manufactured. An obvious the more serious of crimes are able to justify the effort and major source of variation lies in the trace and ultra-trace cost necessary for in-depth investigation.Laser ablation induc- elemental composition of raw materials consumed in the tively coupled plasma mass spectrometry (LA–ICP-MS) offers process, which is directly dependent on the location where the the potential of producing fast, definitive and cost effective raw minerals are mined. Similarly, steels used in the manufac- forensic chemical evidence for use in identifying and comparing ture of safes have essentially the same composition, but as physical evidence relating a suspect to the scene of certain with the glasses, the levels of trace elements will vary in crimes.different steels. It is essential that the method of examination be accurate, Analytical techniques that require the sample to be taken highly sensitive and applicable to very small amounts of into solution prior to analysis, have several shortcomings. material as is often the case in forensic examinations.In Dissolution itself can often be time-consuming and, depending addition it is highly desirable that the analytical technique used also be relatively non-destructive, so as to facilitate further upon the thermolability of the analyte, certain elements such Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (195–203) 195Fig. 1 Variability in laser crater dimensions with relation to laser voltage and number of repeat shots.as mercury, cadmium, selenium, arsenic, thallium, gallium, acid is essential for the dissolution of glass and unless the rare earth element fluorides are broken down as a final part of the germanium and bismuth can be lost by volatilisation. Furthermore certain other elements and their compounds, such dissolution procedure then they will not report quantitatively to the final solution and therefore not be determined accurately. as zirconium, hafnium, tantalum, niobium, titanium, thorium and uranium, may be chemically inert and their dissolution Once dissolved, however, the stability of the analytes can vary significantly, with such elements as niobium, hafnium, either incomplete or not possible under the chemical regime used.Of particular significance in this respect are the rare tantalum, tin and titanium requiring acid, preferably hydrochloric, concentrations above 20% v/v to stay in solution and earth elements, which form insoluble fluorides.Hydrofluoric 196 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12prevent hydrolysis. Potential contamination from reagents and incorporation of airborne particulate material can cause additional problems. This aspect is particularly significant when determining low analyte concentrations as such contamination can potentially mask a unique feature that may facilitate differentiation. All of these aspects conspire to increase the analytical blank levels and reduce the number of elements that are significant for identification purposes.LA–ICP-MS is a technique that uses the solid sample for analysis and thereby eliminates the need for sample dissolution. The technique of LA–ICP-MS was investigated to acquire a complete elemental profile, specifically at trace and ultratrace element levels, and thus, with the aid of interpretational software, ‘fingerprint’ steel and glass samples. However, LA–ICP-MS, using the commonly available 1064 nm Nd5YAG system used in this study, is not as yet a completely quantitative technique and consequently it was essential to develop a new approach to handling and interpretation of the data obtained.On the basis of this, the concept of comparison of the interelement associations was investigated. Comparative analysis of mass spectra provides a particularly rapid and convenient means of identifying trace element variations. The successful application of this technique largely eliminates previous ambiguity and facilitates distinction between similar glass and similar steel samples.Fig. 2 Comparison of the spectral fingerprint of three repeat samples of the same steel. EXPERIMENTAL cloud that is carried in the nebuliser gas stream to the ICP Laser Ablation Inductively Coupled Plasma Mass Spectrometry torch. Here particles are volatilised to atoms and subsequently LA–ICP-MS13–21 is a modification of ICP-MS in which a laser ionised at the high temperatures experienced in the plasma.is used for initial sample volatilisation prior to its introduction Analysis is performed using a quadrupole mass filter which, into the plasma. This modification facilitates optimisation of when set in scanning mode, acquires a complete mass spectrum. each process of the overall multi-element technique. The mass spectrum is an indication of an elemental assemblage Importantly, this enables the introduction of solids directly which is unique for the particular sample being analysed.into the plasma, thereby eliminating time-consuming and Consequently, it can be likened to a ‘fingerprint’, enabling the difficult dissolution, and the risks of contamination from comparative analysis of samples based on the trace element reagents. The system often has the added advantage of rela- signature they exhibit.23–26 tively high sample throughput. Optimisation of laser parameters is, however, extremely difficult, a situation that has Analytical protocols probably considerably influenced the lack of enthusiasm with which the technique has been received by the international Quantitative analysis using the Nd5YAG laser requires stan- analytical community.dards with the same composition as the sample under investigation. However, this presents some difficulties. Firstly, this Apparatus demands that the exact sample composition be known prior to analysis, so that standard materials of the same composition A VG (Ion Path, Winsford, Cheshire, UK) LaserLab accessory can be matched and analysed for calibration purposes.linked to a VG Turbo Plus inductively coupled plasma mass Secondly, there is no guarantee of the availability of standards spectrometer was used for all experiments. The laser used in with that same composition and therefore the technique cur- this investigation was a 500 mJ pulsed Nd5YAG laser. It rently has only a limited quantitative application. Finally, operates at a fundamental wavelength of 1064 nm.The system pulse to pulse variations in laser energy and variations in laser was modified with an S-Option high sensitivity interface and coupling efficiency, dependent upon sample orientation and a high resolution lens system was incorporated into the laser colour, will lead to differing amounts of sample reaching the beam to facilitate focussing to a spot size of 20 mm. The VG plasma.22,27 Such variation could only partially be compen- Turbo Plus ICP-MS instrument was used with the pulse sated for by the use of an internal standard.Hence the counting detector in the scan mode. qualitative nature of this analytical technique must be stressed. The laser can be operated in either the Q-fixed (free running) However, provided the material ablated is representative of or Q-switched mode. These two pulse modes allow very that sample, then the pattern of elemental assemblages depicted different sampling actions.22 The energy of the laser radiation, in the mass spectrum can be compared between samples.These when operated in the fixed-Q pulse mode is absorbed directly data can then be used to differentiate between samples of the by the surface of (thermally) conducting samples. This coupling same type. produces deep narrow craters. The Q-switched pulse mode is applied to non-conducting samples. The energy in this mode generates an ionising plasma above the focal point on the Sample Preparation sample.This laser-induced plasma couples with the surface of the sample and ‘eats’ into it, in a manner similar to that in Samples up to 15 mm in diameter were embedded in plastic blocks, of 25 mm diameter and 10 mm thickness, which were GD. The two modes of laser operation enable LA, using the IR Nd5YAG laser, of a diverse range of sample types. When polished to expose the surface of the sample. Samples smaller than this, such as small glass shards and cutting debris, were fired correctly, the laser ablates a representative sample of material from the solid surface.This forms a fine particulate mounted in cyanoacrylate ‘superglue’ on the surface of the Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 197Fig. 3 Variation in laser crater dimentions with relation to frequency and ablation time at 550 V. plastic blocks. These blocks perform a variety of functions. National Institute of Standards and Technology (NIST) SRM 610 Trace Elements in Glass, which contains a significant They raise the sample from the stage of the cell and provide a convenient means of storage and transport.In addition, concentration of lanthanum (139U), and is particularly useful when setting up the instrument as this element occurs in the samples prepared in such a manner can be conveniently analysed using alternative techniques, such as SEM and energy mid mass region. As the sample was ablated, under Q-switched mode, the sensitivity of all ion lens settings and gas flow rates dispersive X-ray analysis, for major and minor element concentrations.were optimised. The choice of this material, for laser optimisation prior to steel analysis, and not a steel standard is because The instrument was optimised prior to analysis using 198 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12craters, required to be produced to ensure optimum reproducible lasing and ablation conditions, was established.The Q-fixed pulse mode was used for the analysis of steels. The use of this mode for metallic samples has been documented in the literature.22,28 Laser energy (voltage applied to the laser flash lamp) was varied between 500 and 1000 V, and encompassed the entire energy range necessary for the ablation of the steels in this investigation. The number of laser shots at each site was varied from 1 to 5. Shots at each site were fired in quick succession rather than at regular intervals over 1 s as, for the fundamental aspect of ensuring reproducibility of laser crater production, it was necessary to observe the formation of each crater as it occurred to determine whether or not appropriate lasing had been initiated.The various ablation conditions produced a range of craters, the general physical features of which remained similar for equivalent conditions between the four steels under investigation. An example of the craters produced, using each of the ablation protocols for a representative steel, is shown in Fig. 1. Coupling, due to the absorption of energy at the surface of the steel from the laser light at 500 V, was not sufficient to remove significant material. When the laser energy was increased to 600 V, coupling caused the ejection of a significant volume of material, represented by the pit volume of the crater produced. Further increases in energy correspond to an increase in the amount of material ablated, as indicated by the size of the increasing craters.The diameter of the 700 V craters were bigger than craters produced with a laser energy of 600 V. At 800 V, the crater diameters were bigger than those at 700 V. Fig. 4 Reproducibility of inter-element association patterns for two white glasses of similar refractive index. optimisation requires the production of a steady state signal and continuous ablation of metals coats the inside of the laser cell and contaminates the system for a considerable time.The use of glass standards also ensures that uniform ablation is undertaken and that a uniform amount of material is delivered to the plasma. The fact that glass and the steel sample being analysed under one of the investigated protocols, are not the same and that plasma loading will be different does not matter as the technique detailed is not based on the quantitation of analytes, merely their association. The use of this NIST standard has proved to be robust and convenient and facilitates reproducible instrument optimisation and maintenance of batch to batch signal reproducibility.Laser Optimisation Before relevant mass spectral data could be collected for actual use, optimum laser operating parameters needed to be established for each sample type under investigation. Conditions established for glass samples were necessarily different from the conditions established for steel. Steel Three steel and three glass samples of varying composition, were tested to establish an optimum universal ablation protocol for each sample type.Experimentation, to establish each protocol, involved varying parameters such as the laser pulse mode, the laser power or energy, and the repetition rate of laser shots. This formed the basis of the initial investigation. Fig. 5 Comparison of the inter-element association patterns for white glasses with similar refractive index. On the basis of this work the form and nature of the laser Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 199However, the diameter of the craters did not increase signifi- cantly above 800 V laser energy. At the extreme conditions experienced at the higher laser energies of 900 and 1000 V, the material erupted from the sample and was dispersed over a wide area, including the inside of the ablation cell. This was of particular concern as it greatly increased the risk of contamination for future samples.It is imperative that material is not deposited on the optical window of the ablation cell. This is because subsequent laser pulses may revaporise this material and interfere with the analysis of ensuing samples. In addition, at these extreme conditions, the mass of molten material, fragments and vapour ablated from the surface, is composed of substantially bigger particles. As such, they are more susceptible to gravitational deposition, covering the ablation cell, the walls of the interface tubing and being deposited in the switching solenoid mechanism used to redirect gas flows during analysis.In addition, incomplete atomisation of large particles can lead to deposition on, and clogging of, the sample and skimmer cones. This can cause a change in ion vectoring which will lead to erratic and suppressed ion transport and sensitivity loss and significantly effect reproducibility. A second effect of these bigger particles, is their ability to act as depositional loci, removing, by adsorption, smaller particles that would otherwise have reached the analyser.29 It must also be considered that the higher energy conditions require relatively larger samples to ensure reproducibility of conditions from sample to sample and as this cannot always be assured the lowest acceptable voltages should always be used.In this way small and large samples will be analysed under identical protocols.Laser energies, suitable for reproducible ablation, therefore encompass the 700 to 800 V range. A laser energy of 800 V was eventually selected, on the basis of increased sensitivity for a variety of analytes. Under these conditions, excellent reproducibility of the analyte signal was obtained. Fig. 2 represents three offcuts of the same steel Fig. 6 Comparison of the inter-element association patterns for sample analysed on three separate occasions over a three brown glasses produced by the same factory on different days.month period. The effects of the repetition rate of laser shots at each irradiation site observed remained somewhat similar at each different sampling conditions, and was applied to four different coloured glass samples with refractive indices between 1.5164 laser energy. At 3, 4 and 5 shots, splatter of sample about the laser hole, and in the cell, was produced, which increased and 1.5189. The physical characteristics of the craters, produced within each of the ablation conditions, remained similar irres- the risk of analyte contamination for subsequent sample analyses.These repetition rates were therefore avoided, leaving the pective of the glass composition tested. The variation between the craters produced can be rep- option of 1 or 2 shots per site. The first laser shot fired at the steel heats the cold surface, and will proportionately resented for an average glass sample, for the 550 V ablation matrix, in Fig. 3. In each photograph the experimental crater remove relatively more volatile elements than refractory ones. This may bias the elemental composition of the spectrum. The referred to is the one closest to the centre. The largest variation observed in the crater dimensions was due to the variation in release of the more refractory elements, however, such as barium, molybdenum or cobalt for example, will depend on the repetition rate of laser shots. Ablation conditions with a laser repetition rate of 5 Hz were insufficient to ablate an the surface properties of the sample.So firing just one shot may lead to the irreproducible release of such refractory adequate amount of glass for analysis. This effect is indicated by the physical dimension of craters (Fig. 3). The increased elements from different samples of steel. Consequently, two shots per site was selected, so that elements were released from pulse rate of 10 Hz generated significantly 4× more material.However, the most analytically useful mass of glass particles an already molten pool of material, contained within a very hot zone in the steel. This diminished the consequences of was produced at 15 Hz. An ablation time of 60 s was finally chosen. Factors that variations in individual surface properties, such as crystallinity and smoothness, and ensured that refractory elements were influenced this decision included the error associated with a 1 s variation in an ablation time of 30 s, as compared to the sampled reproducibly irrespective of the steel sample under investigation.29 same variation in a 60 s ablation, as well as the effects of a 90 s ablation period with respect to the thickness of the sample analysed.Longer ablation periods would require thicker Glass samples and these would not always be available. Sample thickness was an important consideration in For the ablation of glass samples, the laser was operated in the Q-switched pulse mode.This followed the conventional determining a relevant ablation condition. Samples supplied for analysis that are particularly thin, for example thin shards application of this pulse mode to non-conducting samples.22,29,30 Laser power was varied in increments of 100 V of glass, require ablation that provides enough material to the mass analyser, without total destruction of the sample. between 550 and 850 V. The repetition rate of laser shots was extended from 5 to 15 Hz, for periods of 30, 60 and 90 s at Consequently, a low laser power is desirable, but not essential. The laser voltage used in the glass analysis was selected on each irradiation site.This experimental matrix generated 36 200 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Fig. 7 Comparison of the inter-element association patterns for steel from five different safes. the basis of the sensitivity of specific analyte signals required In the diagrams detailed here it is also possible to increase or decrease the value of a component by a specific order of for unequivocal spectral comparison.A final glass ablation protocol of 750 V at 15 Hz for 60 s was found to be an ideal magnitude. If this has been done the actual amount of multiplication is added as a suffix to the analyte isotope on the compromise for all glass samples analysed. diagram. In this way it is possible to plot ternary associations of analytes that differ considerably in their relative concen- RESULTS AND DISCUSSION tration within the matrix under investigation and show the resulting association at a sensible position within the diagram Reproducibility studies rather than in the corner associated with the highest concen- Two white float glasses were chosen for this study.Each tration analyte. sample was ablated ten times over a 24 h period on two separate occasions 5 d apart. This test not only indicates the reproducibility of the analytical procedure in-run, but also on Discrimination of Clear (Colourless) Glasses a day to day basis.Representative results are detailed in Fig. 4. The elements Y, Sr, Ba, La and Ce were chosen. Ternary plots Fig. 5 details the results for five different glass samples analysed on ten different occasions over a two month period. On the indicate an excellent precision for both with in-run and day to day data and confirm that the technique is robust for glass basis of these ternary plots for Mo, Ni, Y and Zr, it is easily possible to differentiate between the five glass samples, all of analysis. The overlap of data for the two runs and the isolation of the data into discrete clusters for the two glasses confirms which have refractive indices that differ by less than 0.1% relative.White glass 2 appears to have a more variable trace that it is possible to discriminate between these two glasses when more conventional techniques cannot differentiate element association in the ternary plots than any of the other glasses and this probably relates to spatial variations in the between them.Ternary plots represent the direct comparison of the relation- matrix. From these data it is possible to see the concept of both inter-element association patterns and ternary plotting. ship between three components in a system. For data to plot in any corner of the triangle, representing the ternary associ- Associations of elements may sometimes not differentiate between samples.In these cases there are two explanations; ation of three components, would indicate that the concentration of that component indicated at the corner would be firstly that there is in fact no differentiation at all, (the samples may be the same) or secondly, there is no difference in the 100% relative to the other two components. Positioning a data point anywhere else within the ternary plot indicates the inter-element associations between two different samples.In cases where no differences can be observed it is simply a matter relative percentile inter-association of the three components defined by the diagram. of using new elements to compare associations. Eventually, if Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 201no differentiation can be found the conclusion that the samples are either the same or made from exactly the same material must be made.It must be remembered that concentrations of matrix analytes are in ppm and ppb levels and an exact replication of non-identically sourced material is unlikely. Often it will be seen that some inter-element associations do not produce differentiation while some do. Under these circumstances, it is essential to compare both types of data when determining commonality of source material, lack of differentiation, taken in association with ability to differentiate is often a conclusive identifier.In all, with the data set used, it is possible to produce over 27000 inter-element plots for comparative purposes. These plots are easy to interpret and aid considerably the interpretation of closely similar spectra. This concept is taken into the interpretation of all spectra produced by this methodology, not simply on its own but in association with the spectral plots to establish commonality of material for both glasses and steels. Discrimination of Brown Bottle Glasses Ten replicates of eleven samples of brown glass, a convenient weapon used in many late night brawls, were analysed over a two month period using the same ablation protocol as for clear glass.Representative results are given as the ternary plots (Fig. 6). The brown glass samples were taken from bottle glass manufactured at the same factory on different days. This exercise represents an extreme test of the application of the technology to samples that are essentially identical and are often impossible to tell apart using conventional analytical protocols.Results for Ba, Rb, Sr and Ce, indicate that while the inter-element associations form a relatively closely packed basic matrix, the components of this matrix are made up of isolated tight clusters representing the repeat analyses of a single sample. Within the overall major cluster there is definite isolation of the individual sample clusters indicating that these can be distinguished.The extremely tight associations within an individual cluster indicate that the reproducibility of the technique is excellent. As indicated, the changes of association of the inter-elemental clusters, when analytes are changed, adds substantially to the application of the methodology for distinguishing between samples. Discrimination of Steel Samples Data representing the analysis of steel from five different safes, on three separate occasions over a three month period, is shown in Fig. 7. These data have been plotted on the same diagrams to give an indication of reproducibility of analytical protocol, robustness of the technique and reproducibility of the analyte signal. Data indicate close agreement between repeat analyses. In using ternary plots it must be remembered Fig. 8 Comparison of the inter-element association patterns for steel that, while under some conditions two samples of differing from various origins. provenance may not be discriminated, under other conditions these samples will be discriminated.Samples with the same provenance will not be discriminated under any variation of the ternary plot analytes. An example of this can be seen for the process of sawing off material, such as sawing off a shotgun or whether two pieces of steel came from the same original safes 3 and 5, which in Fig. 7, upper left, upper right and lower left, are closely associated with each other while in Fig. 7, sample. The excellent reproducibility of the technique for these matrices is again demonstrated, together with its clear potential lower right, they are spatially widely separated. By comparing similar materials over a time period, the analytical concept of distinguishing between similar samples.Again the use of the ternary plots can be demonstrated with associations and lack and protocol is carefully tested and under these conditions is clearly able to distinguish between the various samples.of associations indicating the exact manner in which these diagrams can be used to determine provenance of material. In Fig. 8, gun barrels, crow bars and pliers are compared. While it is not intended to indicate that there should be Obviously more work is required to build a data base of all this information, however, this study has clearly been able to similarity between these sample groups, it is again necessary to demonstrate variability with time to test the robustness of develop an analytical technique that will, if utilised, be of major significance to determining the provenance of scene of the technique under different conditions.Often it is necessary to establish whether or not debris from clothing came from crime evidence. 202 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 125 Locke, J., and Underhill, M., Forensic Sci. Int., 1985, 27, 247. CONCLUSIONS 6 Standard Methods of Chemical Analysis, ed. Welcher, F. J., D.Van Using LA–ICP-MS coupled with pattern recognition software Nostrand, Canada, 1963, vol. 2, Part B, p. 2229. 7 Locke, J., and Elliot, B. R., Forensic Sci. Int., 1984, 26, 53. it has been possible to develop an analytical protocol and 8 Marcouiller, M., J. Forensic Sci., 1990, 35, 554. methodology that facilitates identification of the provenance 9 Andrasko, J., and Maehly, A. C., J. Forensic Sci., 1978, 23, 250. of glass and steel debris from crime scenes. This technology is 10 Koons, R.D., Fiedler C., and Rawalt, R. C., J. Forensic Sci., robust, on a day to day basis, and while requiring a high level 1988, 33, 49. of analytical skill to develop and operate successfully can, with 11 Catterick, T., and Wall, C. D., T alanta, 1978, 25, 573. appropriate training, be conveniently taught to less skilled 12 Zurhaar, A., and Mullings, L., J. Anal. At. Spectrom., 1990, 5, 611. 13 Denoyer, E. R., Fredeen, K. J., and Hager, J. W., Anal Chem., operators.The analytical data is amenable to storage in an 1989, 61, 445A. electronic database, meeting search and match requirements 14 Moenke-Blankenburg, L., in L aser Micro Analysis, ed. obviously necessary when dealing with large amounts of data. Winefordner, J. D., and Kolthoff, I. M., John Wiley & Sons, The mass spectra obtained contain a sufficiently large number Canada, 1989. of analyte elements for comprehensive and definitive inter- 15 Reed, S. J. B., Chemical Geol., 1990, 83, 1.sample comparison. Elemental assemblages, displayed as a 16 Chenery, S., Hunt, A., and Thompson, M., J. Anal. At. Spectrom., 1992, 7, 647. mass spectrum, form the basis of differentiation between 17 Williams, J. G., and Jarvis, K. E., J. Anal. At. Spectrom., 1993,8, 25. samples. The presence, absence and relative abundance of 18 Pearce, J. G., Perkins, W. T., Abell, I., Duller, G. A. T., and Fuge, elements in particular elemental associations (ternary discrimi- R., J.Anal. At. Spectrom., 1992, 7, 53. nation diagrams or plots) provide a unique means of display 19 Crain, J. S., and Gallimore, D. L., J. Anal. At. Spectrom., 1992, and comparison of the trace element signature or ‘fingerprint’. 7, 605. Ternary plots confirm good precision on sample re-analysis. 20 Perkins, W. T., Pearce, N. J. G., and Fuge, R., J. Anal. At. Spectrom., 1992, 7, 611. The relative abundances of analytes, contained within the trace 21 Perkins, W. T., Pearce, N. J. G., and Jeffries, T. E., Geochim. element signature of one sample when compared with the same Cosmochim. Acta, 1993, 57, 475. element relationships in a different sample, allowed discrimi- 22 Jarvis, K. E., Gray, A. L., and Houk, R. S., Handbook of nation between similar samples of steels and glasses. Inductively Coupled Plasma Mass Spectrometry, Blackie, Glasgow, The technology can be automated to facilitate high sample 1991, ch. 10. throughput. Sample preparation can be as simple or as detailed 23 Watling, R. J., Rapid Commun. Mass Spectrom., 1996, 10, 130. 24 Watling, R. J., Herbert, H. K., Barrow, I. S., and Thomas, A. G., as required by the operator and definitive analytical data can Analyst, 1995, 120, 1357. be obtained in as little as 5 min. 25 Watling, R. J., Herbert, H. K., and Abell, I. D., Chem. Geol., 1995, 124, 67. The authors wish to thank the National Institute for Forensic 26 Watling, R. J., Herbert, H. K., Delev, D., and Abell, I. D., Science (Australia) for funding this research and the Western Spectrochim. Acta, Part B, 1994, 49, 205. Australian Police Forensic Branch for their support and assist- 27 Dittrich, K., andWennrich, R., Prog. Anal. Spectrosc., 1984, 7, 139. 28 Yasuhara, H., Okano, T., and Matsumura, Y., Analyst, 1992, ance in obtaining research materials. 117, 395. 29 Franks, J., Marshall, J., Brown, I., and Garden, L., Anal. Proc., REFERENCES 1992, 29, 23 30 Moenke-Blankenburg, L., Schumann, T., Gu�nther, D., Kuss, 1 Hickman, D. A., Anal. Chem., 1984, 56 (7), 844A. H. M., and Paul, M., J. Anal. At. Spectrom., 1992, 7, 251. 2 Walls, H. J., Forensic Science, Sweet and Maxwell, London, 2nd edn., 1968, p. 30. Paper 6/05143J 3 Brewster, F., Thorpe, J., Gettinby, G., and Caddy, B., J. Forensic Received July 23, 1996 Sci., 1985, 30, 798. 4 Vaughn, M.-A., and Horlick, G., J. Anal. At. Spectrom., 1989, 4, 45. Accepted September 7, 1996 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 2

ISSN:0267-9477    

DOI:10.1039/a605143j

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Slurry Sampling Electrothermal Atomic Absorption Spectrometry: Results From the Second Phase of an International Collaborative Study N. J. MILLER-IHLI Food Composition L aboratory, Beltsville Human Nutrition Research Center, Beltsville, MD 20705, USA. E-mail:Miller-Ihli@bhnrc.usda.gov The second phase of an international collaborative study cantly from particle size effects and it offers longer residence designed to evaluate the current state-of-the-art for solid times with a correspondingly higher atomization efficiency.sampling was completed. Samples were sent to more than 20 There are two principal methods for solids analysis using laboratories who participated in the first phase, eight of which ETAAS: direct analysis of the solid material and the analysis reported complete enough sets of data to be considered. Each of a slurry or suspension from which a representative aliquot collaborator was sent four powdered materials and one pre- may be injected into the furnace.Bendicho and de Loosmade slurry for analysis. Different elements were determined Vollebregt1 concluded in their review that slurry sampling is in the various samples including Cu, Cr, and Pb. Average preferable to direct solids analysis because slurry sampling can performance for all elements determined in all materials was be fully automated and because slurry concentrations may be excellent with the overall mean concentrations±uncertainties easily adjusted to ensure that a reasonable quantity of material computed for all collaborators overlapping the reported is injected into the furnace for analysis.This author’s expertise reference ranges in all instances. Some of the laboratories in solid sampling using ETAAS lies with the analysis of slurries demonstrated consistently good expertise whereas others or suspensions. Slurry sampling offers several benefits over utilized analysis conditions which led to poorer performance. direct solids analysis, including combining the benefits of solid Based on the mean reference concentration for the various and liquid sampling, utilization of conventional liquid sample analytes in the reference materials analyzed, the average handling apparatus, straightforward automation, flexibility in performance for all laboratories was 100±7% and the range slurry preparation, ease of use of chemical modifiers and the of recoveries was 78–107%.The importance of using advantage that slurries may be prepared in advance.1,2,9 To secondary wavelengths and the importance of using sufficiently ensure that a representative aliquot of slurry is injected into large amounts of solid to be representative of the bulk the furnace for analysis, the slurry must either be stabilized material is demonstrated. Possible problems affecting prior to analysis or mixed vigorously.Stabilization with thixoanalytical accuracy are discussed, including the use of mini- tropic agents has proved to be problematic because of problems flows, inappropriate amounts of matrix modifier, high thermal with reproducible sample delivery,9,10 but homogenization pretreatment temperatures, low atomization temperatures, using mechanical mixing devices such as magnetic stir-bar short atomization and/or read times, expulsion losses, mixing,11 vortex mixing,12,13 gas bubbling14 or ultrasonic agi- inadequate background correction and sample sizes which were tation1,2,6–8,15,16 has proved useful.The ultrasonic mixing too small. The effect of sedimentation errors due to high approach is the only one which has been commercialized density materials is discussed. Analytical results for a pre- (USS-100, Perkin-Elmer, Norwalk, CT, USA).17 made slurry were not found to be significantly different from Because of the significant potential that ETAAS offers for results for slurries of the same material which were prepared solid sampling and because of its more recent widespread use by the collaborators.Data are presented which show that for important applications, an international solid sampling analyte extraction into the liquid phase of the slurry is not a collaborative study was planned to evaluate the current state- prerequisite for accurate slurry analyses. of-the-art. More than 25 laboratories responded favorably to Keywords: Slurry; sedimentation; particle size; electrothermal a request to participate in a two-phase study.The first phase, atomic absorption spectrometry; solid sampling; non-resonance which was the subject of a previous report,18 focused on the wavelength; ultrasonic slurry sampling determination of Pb and Cr in sediments. Samples were sent to 28 laboratories and data were received from 18collaborators with 16 providing slurry analysis results and two providing Solid sampling is gaining in popularity and one of the analytical direct solids analysis results. Of those who employed slurry techniques receiving a great deal of attention continues to be sampling, 12 utilized the commercial USS-100 autosampler electrothermal (graphite furnace) AAS (ETAAS), which has accessory, one used a hand-held ultrasonic probe, one used a received numerous reviews.1–3 Recently, the use of furnaces combination of hand-held ultrasonic probe mixing and mag- has been extended to electrothermal vaporization (ETV) netic stirring and two used vortex mixing followed by hand ICP-MS,4,5 where they are used as vaporizers, but that will pipetting.not be discussed here. Frequently reported benefits of solid The results from Phase 118 provided some very interesting sampling include decreased sample preparation time, decreased insights into the current state-of-the-art of solid sampling. The chance of analyte loss due to retention by an insoluble residue data from the large number of analysts reporting slurry results or premature volatilization, decreased use of acid and a suggested that this technique is sufficiently mature that it may corresponding decrease in the production of waste, decreased be used for routine analyses.A review of the data suggested likelihood of sample contamination and the ability to characthat it is important to employ what is often referred to as terize materials using micro-amounts of solid.1,6–8 The afore- STPF conditions.19 One of the most significant problems mentioned benefits have been demonstrated in literature encountered by analysts concerned sampling.Because of the reports and the suitability of ETAAS highlighted because, unlike nebulization techniques, ETAAS does not suffer signifi- Food Composition Laboratory (FCL) characterized level of Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (205–212) 205sample homogeneity of the sediments analyzed, the suggestion analyte concentration range (e.g., 50–150 mg g-1 Pb in fly ash).Participants also received information outlining considerations was made for collaborators who performed slurry analyses to use a minimum of 2–5 mg per 1 ml of diluent. Analysts who for optimizing slurry analysis conditions focusing on several key points. followed this recommendation but chose the most sensitive wavelengths for analyses had to do one or more things to The first point outlined the importance of ensuring that a reasonable amount of solid was injected into the furnace for reduce sensitivity. Some utilized very small sample analysis volumes (e.g., 2–5 ml compared with the recommended 20 ml analysis.Analysts were encouraged to consider the use of a minimum of 2 mg of solid to prepare 1 ml of slurry and they volume) and some utilized mini-flows during the atomization, both of which proved problematic. The preferred method was were encouraged to use 20 ml injection volumes for all analyses. Collaborators were reminded that the use of alternative, less to use the recommended larger sample size for analysis and to select an alternative, less sensitive wavelength, if available, for sensitive wavelengths might prove desirable, and they were told that a 2 ml blank ‘slug’ used after the slurry would help the analysis.Other problems identified during Phase 1 included the use of too little chemical modifier, selection of a thermal ensure that all of the slurry was injected into the furnace for analysis. The second point outlined the importance of optimiz- pretreatment temperature which was too high, leading to preatomization losses, use of ‘fast furnace’ programming with no ing ETAAS analysis conditions focusing on analysis time, thermal pretreatment and atomization temperatures and the thermal pretreatment step, which led to apparent losses due to expulsion during the rapid heating during the atomization use of modifiers.Analysts were encouraged to consider the use of an alternative line rather than using a mini-flow to reduce step, the selection of atomization temperatures which were too low, leading to incomplete atomization, and the selection of sensitivity.They were also reminded of the potential benefit of oxygen ashing in removing organic material during the thermal atomization times which were too short. In this work, Phase 2 data are reported and reviewed and pretreatment step. The third point outlined the need for optimizing the slurry the usefulness of ultrasonic slurry sampling is evaluated.Analytical performance characteristics are used as a key to mixing conditions. Analysts were told that typical ultrasonic mixing times are 20–25 s and they were reminded of the benefit identify possible problems with ETAAS analysis conditions and the importance of a systematic approach for solid sampling of extending the mixing time for the first analysis to ensure that slurries are effectively ‘pre-mixed’, particularly those slurr- is discussed.Key issues include material density and particle size, analyte extraction and the suitability of this approach for ies containing low density material, which tends to float on the surface until particles are effectively wetted. Finally, Phase 2 routine analyses. Results from Phases 1 and 2 are both considered in making a final overall evaluation of the current participants were given a summary of possible analysis conditions, which are summarized in Table 2.Collaborators were state-of-the-art of solid sampling. told that they were not in any way obliged to use the conditions outlined. Conditions were provided simply to provide some EXPERIMENTAL guidance for less experienced slurry analysts. Participants were encouraged to prepare four slurries of different concentration Study Protocol for analysis to help evaluate any possible trends of concen- Inquiries regarding participation in Phase 2 were sent to all tration dependence as a function of sample mass analyzed.of the original participants in the International Collaborative When the Phase 2 samples were sent out from the Food Study. Phase 2 samples were sent to more than 20 laboratories Composition Laboratory (FCL) of USDA, data report forms but only ten laboratories returned data and only eight provided were included which asked for specific information regarding data which were complete enough to be included in the instrumentation used, background correction method, method evaluation.All collaborators who returned data are acknowl- and time used for slurry mixing, identification of SRMs run as edged in Table 1. Each of the collaborators brought their own controls and detailed instrumental conditions (e.g., wavelength, expertise to the study and a wide range of approaches were source, modifier, sample and standard volumes, thermal pre- utilized to do the determinations. All participants chose to treatment and atomization temperatures, read time, atomiz- perform slurry analyses.ation from wall or platform, range of standards and Collaborators were sent information regarding the samples concentration of SRM 1643c determined). Collaborators were which included the general type of sample [e.g., coal fly ash asked to report four values for each of the materials received (CFA), marine material (MBM), glass (GLS), botanical mate- for analysis. Collaborators were also asked to attach a copy rial (BOT) and pre-made slurry (SLR)] and the approximate of the analysis conditions as printed out by the instrument so that detailed information such a slit widths, lamp currents, gas Table 1 Phase 2 collaborators: slurry sampling flow rates, pyrolysis and atomization conditions and volumes could be verified.This proved particularly useful when Affiliation Name reviewing the data in those instances where inaccurate results were reported. J. Padmos and G. de Loos- Delft University of Technology Six samples were sent to collaborators. The first was ident- Vollebregt (The Netherlands) ified as SRM 1643c Acidified Water, which collaborators were F.E. Greene and N. Miller- Beltsville Human Nutrition asked to read as a QC check and to record the concentration Ihli Research Center, USDA (USA) determined when reporting analysis conditions. The samples V. Krivan Ulm University (Germany) U. Richter and W. Dannecker Institut fu�r Inorganische und sent for slurry analyses were all reference materials but were Angewandte Chemie (Germany) not identified as such to the collaborators. The first material, D.Anderson and E. S. Frame General Electric CRD (USA) coal fly ash (CFA), was NIST SRM 1633a (National Institute D. Stillwell and C. Musante Connecticut Agricultural of Standards and Technology, Gaithersburg, MD, USA). The Experimental Station (USA) second material was a marine material (MBM), DOLT-1, I. Lopez Garcia University of Murcia (Spain) which is a dogfish liver tissue (National Research Council of M.Hoenig Institute for Chemical Research (Belgium) Canada, Ottawa, Ontario). The third material was a glass D. Butcher Western Carolina University (GLS) reference material produced by the US Geological (USA) Survey (USGS) (Denver, CO, USA) and designated GSD. The P. Esser Anneliese Zementwerke fourth material was a botanical material (BOT), which was Aktiengesellschaft (Germany) NIST SRM 1547 Peach Leaves.The final material (SLR) was 206 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Table 2 Suggested analysis conditions summary (analyses indicated by brackets are optional) Material Slurry preparations Coal fly ash (CFA) Pb: 5 mg in 1 ml, 10 mg in 1 ml, 20 mg in 1 ml, 100 mg in 10 ml Cr: 1 mg in 1 ml, 2 mg in 1 ml, 2.5 mg in 1 ml, 20 mg in 10 ml [Cd]: 1 mg in 1 ml, 2 mg in 1 ml, 5 mg in 1 ml, 20 mg in 10 ml Marine material (MBM) Pb: 10 mg in 1 ml, 20 mg in 1 ml, 30 mg in 1 ml, 200 mg in 10 ml [Se]: 1 mg in 1 ml, 3 mg in 1 ml, 5 mg in 1 ml, 30 mg in 10 ml Glass (GLS) Cr: 2 mg in 1 ml, 5 mg in 1 ml, 10 mg in 1 ml, 50 mg in 10 ml [Pb]: 5 mg in 1 ml, 10 mg in 1 ml, 20 mg in 1 ml, 100 mg in 10 ml Botanical material (BOT) Pb: 10 mg in 1 ml, 20 mg in 1 ml, 30 mg in 1 ml, 200 mg in 10 ml Cu: 5 mg in 1 ml, 10 mg in 1 ml, 20 mg in 1 ml, 100 mg in 10 ml [Cr]: 10 mg in 1 ml, 20 mg in 1 ml, 30 mg in 1 ml, 200 mg in 10 ml Pre-made slurry (SLR) Contains ~200 mg in 10 ml; diluent 2% HNO3+0.005% Triton X-100 Suggested conditions for quantification for various materials— Pb 261.4 nm 0.5–5.0 mg ml-1 standards CFA [GLS] Pb 283.3 nm 5–100 ng ml-1 standards MBM BOT SLR Cr 429.0 nm 25–500 ng ml-1 standards CFA GLS Cr 357.9 nm 1–100 ng ml-1 standards [BOT] [SLR] Cd 228.8 nm 0.5–5.0 ng ml-1 standards [CFA] Cu 324.8 nm 1–100 ng ml-1 standards BOT SLR Se 196.0 nm 1–40 ng ml-1 standards [MBM] a pre-made slurry which was made using SRM 1547 Peach Slurry Preparation Leaves. The purpose of sending this material as both a solid A variety of approaches were used for slurry preparation but and as a pre-made slurry was to evaluate if there were any nearly all collaborators followed the suggestion of preparing systematic differences seen in the reported concentrations for four slurries using different masses of powdered sample.Most the solid. collaborators prepared three of their slurries in 1 ml of diluent Each of the solid materials provided to collaborators was and then prepared their fourth slurry using a larger mass of presented in a powdered form.This was done because the solid in 10 ml. The majority used a slurry diluent consisting of purpose of the study was to evaluate the instrumental capabili- 2–5% nitric acid combined with 0.005% Triton X-100 (octyl- ties for lid sampling, focusing on possible errors associated phenoxypolyethoxyethanol) (Rohm and Haas, Philadelphia, with slurry sampling, atomization conditions, calibration stra- PA, USA).For the glass sample, some collaborators used water tegies, mixing times, use of chemical modifiers, possible losses as a diluent. Again, any unusual diluents used which potentially and the effect of sample size. The intent of this study was not had a significant effect on analytical results are identified and to evaluate sample homogeneity or possiblesources of contami- discussed in the Results and Discussion section.nation due to sample grinding. Should the reader be interested in issues related to grinding techniques, potential contamination, particle size and homogeneity, related research results Calibration and Quantification have been discussed in detail previously.2,6,7,10 Laboratories 1–6 and 8 accomplished quantification using calibration against aqueous calibration curves. In most instances, platform atomization was selected and integrated Instrumentation absorbance measurements were used for quantification.There was a wide range of instrumentation used by the Laboratory 7 most often utilized the method of standard collaborators. Spectrometers used by the eight collaborators additions. In every laboratory, 3–5 readings were made for whose results are summarized here included the following: four each of the four slurries. 5100 PC (Perkin-Elmer) equipped with Zeeman-effect background correction (laboratories 1, 2, 5 and 6), one 4100ZL (Perkin-Elmer) equipped with Zeeman-effect background cor- RESULTS AND DISCUSSION rection (laboratory 3), one 3030 (Perkin-Elmer) equipped with Determination of Pb, Cu, and Cr in BOT Zeeman-effect background correction (laboratory 4), one 1100B (Perkin-Elmer) equipped with deuterium are back- One of the goals of Phase 2 was to discern if there would be a systematic difference in reported results for a material which ground correction (laboratory 7), and one 400 Zeeman (Varian, Palo Alto, CA, USA) equipped with Zeeman-effect background was presented to collaborators as both a powder, from which collaborators were to prepare their own slurries for analysis, correction (laboratory 8).Laboratories 1–6 used the USS-100 (Perkin-Elmer) mixing device in conjunction with their instru- and as a pre-made slurry. The material selected for this test was a botanical material BOT, which was actually SRM 1547 ment’s autosampler for automated ultrasonic mixing of slurries combined with automated sample injection for analysis.Peach Leaves reference material produced by NIST.20 The pre-made slurry was prepared by weighing approximately Laboratory 7 utilized magnetic stirring and manual sample injection into the furnace. Laboratory 8 utilized a Vibra-Cell 200 mg of material in a 10 ml volume of 2% nitric acid diluent containing 0.005% of Triton X-100. The moisture content was (Sonics and Materials, Danbury, CT, USA) hand-held ultrasonic probe for mixing prior to autosampler injection for determined in FCL to be 3.8%.Each collaborator received a single 10 ml slurry preparation. analysis. Different chemical modifiers were used by collaborators and Fig. 1(a) contains a plot of Cu analytical data as a function of laboratory number. Circles represent the data for slurries these are not summarized here. When the selection of a particular modifier or the amount of modifier used significantly prepared directly from the solid material in the collaborator’s laboratory and squares represent data for the pre-made slurry.affected the analytical results, this is noted in the Results and Discussion. In fact, any analysis conditions contributing to Slurries prepared by collaborators typically ranged from 5 to 20 mg in 1 ml and included a single larger volume slurry of unexpected results are identified and discussed. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 207sis of slurries made in the four laboratories who reported data was 1.12±0.15 mg g-1 as received, which is also in excellent agreement with the reference concentration. The data for the collaborator-made slurries was not statistically significantly different from the data for the pre-made slurry (t=0.39). Only one laboratory (7) reported a value which fell outside the ±20% range as compared with the mean reference value for either slurry. Fig. 1(c) contains a plot of Pb analytical data as a function of laboratory number.Slurries prepared by collaborators typically ranged from 10 to 50 mg in 1 ml and included a single larger volume slurry of 100–200 mg in 10 ml. The certified Pb concentration is 0.87±0.03 mg g-1 dry mass or 0.84±0.03 mg g-1 Pb as received. Although data for both the pre-made slurry and the collaborator-made slurries for laboratory 3 were high, a review of their individual data and analysis conditions did not provide any insight into the potential source of such a bias.The mean reported concentration for the premade slurry for six laboratories was 0.89±0.10 mg g-1 as received, which is in excellent agreement with the reference concentration range. The mean reported concentration for the analysis of slurries made in the seven laboratories who reported data was 0.90±0.26 mg g-1 as received, which is also in excellent agreement with the reference concentration range. The data for the collaborator-made slurries was not statistically significantly different from the data for the pre-made slurry (t=0.88).Only one laboratory (3) reported a value which fell outside the±20% range as compared with the mean reference value for either slurry. It is clear from the data for this botanical material that accurate slurry determinations may be made using either Fig. 1 Results from the analysis of BOT (NIST SRM1547): (a) Cu collaborator-made or pre-made slurries. Consideration of the results; (b) Cr results; and (c) Pb results.Circles represent data for characteristics of the material analyzed, combined with the slurries prepared by collaborators and squares represent data for a pre-made slurry. The certified contents are 3.6±0.4 mg g-1 Cu as analytical conditions used, may provide some insight regarding received and 0.84±0.03 mg g-1 Pb as received. The reference concen- the high degree of success achieved. SRM 1547 Peach Leaves tration for Cr is 1 mg g-1.Dashed lines represent the uncertainty of is a very homogeneous material which was jet milled at NIST the certified value. Dotted lines represent ±10% of the mean reference and classified to a particle size of approximately 75 mm.20 The value. Error bars represent ±1 standard deviation (Cu, n=3–4 slurry, density was experimentally determined at FCL to be n=4 pre-made slurry; Cr, n=3–4 slurry, n=4 pre-made slurry; Pb, 1.4 g cm-3, so sedimentation errors were not likely. Analyte n=4 slurry, n=4 pre-made slurry).distribution studies were done at FCL by preparing slurries and using ultracentrifugation to sediment particles so that the concentration in the supernatant could be determined. The 40–100 mg in 10 ml. The error bars reflect the standard deviation for the 3–4 replicate slurry preparations analyzed or the amount of analyte extracted into the liquid phase of a slurry prepared at FCL using a diluent containing 5% nitric acid 3–4 replicate aliquots sampled in the case of the pre-made slurry.The reference concentration is denoted by a horizontal and containing 0.005% Triton X-100 was Pb 93%, Cu 82% and Cr<5%. Clearly, analyte extraction was not a prerequisite line and the dashed horizontal lines indicate the specified reference uncertainty. The certified Cu content is for accurate analyses and particle sizes up to 75 mm could easily be tolerated using slurry sampling. 3.7±0.4 mg g-1 dry mass, which is equivalent to 3.6±0.4 mg g-1 Cu as received.The mean reported concentration for the pre-made slurry for seven laboratories was Determination of Pb in MBM 3.73±0.17 mg g-1 as received, which is in excellent agreement with the reference concentration range. The mean reported Another goal of Phase 2 was to look at a wide range of sample matrices in an effort to see if collaborators considered the concentration for the analysis of slurries made in the six laboratories who reported data was 3.69±0.42 mg g-1 as importance of optimizing thermal pretreatment and atomization conditions to avoid analyte loss. The second material received, which is also in excellent agreement with the reference concentration range.The data for the collaborator-made slurr- selected for the study was a marine biological material, MBM, which was actually DOLT-1, a dogfish liver tissue produced ies were not statistically significantly different from the data for the pre-made slurry (t=0.22).None of the laboratories by the National Research Council of Canada.21 The moisture content was determined at FCL to be 3.5%. reported data outside the ±20% range as compared with the mean reference value for either slurry. Fig. 2 contains a plot of the Pb analytical data as a function of laboratory number. Slurries prepared by collaborators typi- Fig. 1(b) contains a plot of Cr analytical data as a function of laboratory number. Slurries prepared by collaborators typi- cally ranged from 5 to 30 mg in 1 ml and included a single, larger volume slurry of 50–200 mg in 10 ml.The certified Pb cally ranged from 10 mg to 40 mg in 1 ml and included a single larger volume slurry of 40–200 mg in 10 ml. The Cr content is content is 1.36±0.29 mg g-1 dry mass, which is equivalent to 1.31±0.28 mg g-1 as received. The mean reported concen- not certified but the reference concentration is approximately 1 mg g-1 Cr as received. The mean reported concentration for tration for six laboratories was 1.28±0.41 mg g-1 as received, which is in excellentagreement with the reference concentration the pre-made slurry for four laboratories was 1.09±0.04 mg g-1 as received, which is in excellent agreement with the reference range.Only two laboratories (6 and 7) reported values which fell outside the certified range. concentration. The mean reported concentration for the analy- 208 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Determination of Pb and Cr in CFA Another goal of Phase 2 was to determine the suitability of solid sampling for the determination of elements of varying volatility in refractory matrices. The material selected for this test was a coal fly ash CFA, which was actually SRM 1633a produced by NIST.22 The moisture content was determined by FCL to be negligible. Fig. 3(a) contains a plot of the Pb analytical data as a function of laboratory number. Slurries prepared by collaborators typically ranged from 1 to 40 mg in 1 ml and included a single, larger volume slurry of 50–100 mg in 10 ml.The certified Fig. 2 Pb results from the analysis of MBM (DOLT-1). The certified Pb content is 72.4±0.4 mg g-1 dry mass. The mean reported content is 1.31±0.28 mg g-1 as received. Dashed lines represent the concentration for eight laboratories was 70.6±12.1 mg g-1, uncertainty of the certified value. Error bars represent ±1 standard which is in very good agreement with the reference concen- deviation (n=4).tration range. The one laboratory which reported a value which fell outside the±20% range based on the mean certified value was laboratory 8. A review of their reported analysis conditions revealed that four times less than the recommended A review of the apparently low values combined with a minimum amount of sample was used for analysis (e.g., 5ml of detailed review of the analysis conditions revealed that labora- a 2 mg ml-1 slurry).The choice to use the small mass was no tory 6 utilized a mixed modifier consisting of 0.01 mg of doubt due to the fact that this collaborator chose to use the Mg(NO3)2 combined with 0.2 mg of phosphate. Their slurries more sensitive 283.3 nm line rather than the recommended were prepared using from 18 mg in 2 ml to 195 mg in 10 ml, 261.4 nm analysis wavelength. Additionally, laboratory 8 util- which provided equivalent analyte masses in the furnace of ized a 1000 °C thermal pretreatment step with a combination 0.24–0.53 ng of Pb.Individual data reported for four slurry palladium nitrate, phosphate and magnesium nitrate modifier. preparations were 0.83±0.04, 0.61±0.18, 0.57±0.14 and It is possible that both the amount used and the 1000 °C char 1.22±1.01 mg g-1 as received. The large variability in the temperature may have contributed to analyte losses due to reported results and the poor characteristic mass (23 pg), premature volatilization. If the data from laboratory 8 are combined with the very low values reported as compared with excluded, the mean reported concentration for seven labora- reports from other collaborators using similarly prepared tories was 73.9±8.4 mg g-1, which is in excellent agreement slurries, suggest that instrumental conditions were not optim- with the reference concentration range.ized. It is possible that because of the large mass of material Fig. 3(b) contains a plot of the Cr analytical data as a in the furnace the amount of modifier was not sufficient.function of laboratory number. Slurries prepared by collabor- Nevertheless, no source of error could be pinpointed which ators typically ranged from 1 mg in 1 ml to 40 mg in 2 ml and suggested that the data from laboratory 6 should be excluded. included a single, larger volume slurry of 20–100 mg in 10 ml. Laboratory 7 reported data which were significantly high The certified Cr content is 196±6 mg g-1 dry mass.The mean compared with all other reported data. A review of the analysis reported concentration for eight laboratories was conditions revealed that laboratory 7 used very small samples 170±33 mg g-1, which is low but in reasonable agreement for analysis (20 ml injections of slurries made using as little as with the reference concentration range. Three laboratories (1, 16 mg in 5 ml of diluent) and they used a 100 ml min-1 mini- 4 and 8) reported values which were significantly low (e.g., less flow, which is known to provide non-isothermal atomization than 80% recovery based on the mean certified reference conditions and which could result in different degrees of concentration).None of the three laboratories used a chemical analyte expulsion for aqueous standards compared to complex samples. Laboratory 7 also utilized a much different slurry diluent than the other collaborators. The diluent was made up of 8% v/v hydrogen peroxide plus 20% v/v ethanol, 0.01% Triton X-100 and 0.1% m/v ammonium dihydrogenphosphate solution, and slurries were mixed using ultrasound for 5 min and magnetic stirring for 15 min.Aliquots were then handpipetted into the furnace for analysis. Individual results for four slurry preparations were 2.18±0.33, 1.89±0.22, 2.00±0.24 and 2.17±0.20 mg g-1 Pb as received. In a detailed report provided with Phase 2 results, this collaborator also provided data for different dilutions of one of the slurries and obtained a mean value of 2.17±0.13 mg g-1.The same collaborator also provided a method of additions result of 2.08±0.19 mg g-1 for the analysis of five slurries. Clearly, this collaborator had a systematic error in all analyses. It is interesting that their reported characteristic mass was 11 pg for Pb, which is in reasonable agreement with the manufacturer’s expected value for a deuterium arc background corrected system. Most likely, the source of this collaborator’s high results was related to the fact that deuterium arc background correction did not provide adequate compensation for background due to the sample Fig. 3 Results from the analysis of CFA (SRM 1633a): (a) Pb results; matrix which was not removed during the 400 °C thermal and (b) Cr results. The certified contents are 72.4±0.4 mg g-1 Pb as pretreatment step, leading to high results. The use of a received and 196±6 mg g-1 Cr as received. Dotted lines represent 100 ml min-1 mini-flow during atomization may have also ±10% of the mean reference value.Error bars represent ±1 standard deviation (n=8). been problematic. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 209modifier in their analyses and all of them used wall atomization This most likely cannot be attributed to sample inhomogeneity since the reference material is made from molten glass, and it and a 2500–2600 °C atomization temperature. Laboratories 4 and 8 both used a relatively short read time of 5 s during the most likely is not related to Kurfurst’s ‘nugget’ theory24 discussed in the Phase 1 report.18 It is interesting that labora- atomization, which may have been too short based on the author’s experience.In addition, laboratory 4 reported only tories 1 and 4 used less than the recommended minimum amount of sample and chose to use the more sensitive 357.9 nm an 85% recovery for the water sample 1643c, suggesting a systematic error, most likely with the calibration standards.Cr analysis line rather than the recommended 429.0 nm line. As was previously the case, laboratory 8 used no thermal Laboratory 4 also used a 50 ml min-1 mini-flow of argon during the atomization step, which may have led to expulsion pretreatment step, which again proved problematic, most likely owing to losses due to expulsion during the rapid heating losses. Laboratory 8 used no thermal pretreatment step (e.g., ‘fast furnace’ conditions), which we concluded during Phase 1 during the atomization step.As in the determination of Cr in coal fly ash, laboratory 4 reported a value for 1643c which led to apparent losses due to expulsion during the rapid evolution of both matrix and analyte vapors during the was approximately 15% low compared with the mean certified reference concentration, and also a characteristic mass of atomization step. If the data from laboratories 1 and 8 are excluded, the mean reported concentration for six laboratories 8.6 pg, which is significantly better than the manufacturer’s expected value.Both of these facts suggest that perhaps there was 186±19 mg g-1, which is in good agreement with the reference concentration range. was a systematic error due to the standards which most likely contained higher analyte concentrations than intended. If the data from laboratories 1, 4 and 8 are excluded, the mean Determination of Cr in GLS reported concentration for five laboratories was 41.6±6.5 mg g-1, which is in good agreement with the certified The final material distributed for analysis was a glass material reference concentration range.Areview of the data and analysis produced by the USGS and identified as GSD.23 This particular conditions use by laboratory 6 did not reveal anything which material was a very high density material (2.6 g cm-3 determight have contributed to the low results. Because of the very mined in FCL) and we felt it would be interesting to evaluate high density of this material, it is likely that the most significant the collaborators’ ability to perform slurry analyses using such problems related to representative sampling because of serious a high density material with a known particle size distribution.sedimentation errors as predicted by Stoke’s law.25 A more Fig. 4(a) contains a plot of Cr analytical data as a function of detailed evaluation of GSD and potential sources of error in laboratory number.Slurries prepared by collaborators typianalyzing this material are the subject of another report.25 cally ranged from 1 to 10 mg in 1 ml and included a single, larger volume slurry of 20–50 mg in 10 ml. The reference concentration based on data in USGS Professional Paper CONCLUSIONS No. 101323 is 46.5±9.2 mg g-1. The mean reported concen- The analytical data returned by the eight collaborators for tration for eight laboratories was 36.8±8.5 mg g-1, which Phase 2 proved very useful in evaluating the current state-of- certainly overlaps the reference concentration range, but a the-art of slurry sampling.Table 3 contains a summary of large percentage of the reported values appeared to be signifi- reported slurry results including all data. No data were cantly low. excluded, including those data for which possible sources of Specifically, laboratories 1, 4, 6 and 8 reported low values. error were identified. Enough data were reported for the Fig. 4(b) contains a plot of concentration determined as a requested analytes in each of the five materials to provide an function of sample mass injected into the furnace. Those adequate basis for concluding that the average performance collaborators who used less than the suggested minimum using slurry sampling was very good. It should be noted that sample mass of 0.04 mg reported systematically low values. in all instances, the mean value±uncertainty for the reported results overlaps the certified concentration range.If one is to consider the average attainable accuracy for the collaborators in Phase 2, based on the mean reference concentration for the various analytes in the reference materials analyzed,the average performance was 100±7% and the range of average recoveries was 78–107%. Admittedly, laboratories occasionally reported values for particular materials which were significantly high, or more often low, but evaluation of the analysis conditions most often provided insight into the cause of the bias.With the exception of the glass material, there were few values reported by collaborators which fell outside the ±20% range based on the mean reference value. This clearly demonstrates the relative maturity of slurry ETAAS analyses in the hands of the collaborators. Another interesting point is that seven of the eight laboratories whose data are summarized here utilized ultrasonic slurry sampling and, of the seven, six utilized the USS-100.These data suggest that ultrasonic agitation reliably provides good mixing of slurry samples prior to analysis. This study provided very useful data which may be combined with data from Phase 1 to obtain a more general view of the strengths and weaknesses of slurry sampling. When data were combined from the two phases, the data for the pre-made slurry were not considered since there was no difference Fig. 4 Results from the analysis of GLS (GSD): (a) Cr results; and between results obtained for the pre-made slurry as compared (b) Cr results as a function of sample mass injected for analysis.The with slurries made by the collaborators. One of the purposes certified content is 46.5±9.2 mg g-1. Dashed lines represent the uncer- of this study was to see which sample types were problematic tainty of the certified value. Error bars represent±1 standard deviation (n=4). and to evaluate if a particular element proved routinely to be 210 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Table 3 Summary of reported slurry results (concentration mg g-1 in solid as received) Summary Pb Certified Pb Summary Cr Certified Cr Summary Cu Certified Cu Material reported content reported content reported content NIST SRM 1547 Peach Leaves 0.90±0.26 0.84±0.03 1.12±0.15 (1)† 3.69±0.42 3.6±0.4 n* 7 4 6 NIST SRM 1547 pre-made slurry 0.89±0.10 1.09±0.04 3.73±0.17 n 6 4 7 NRCC DOLT-1 Dogfish Liver 1.28±0.41 1.31±0.28 ND — ND — n NIST SRM 1633a Coal Fly Ash 70.6±12.1 72.4±0.4 170±33 196±6 ND — nUSGS GSD Borosilicate Glass ND — 36.8±8.5 47.1±6.6 ND — n 8 * n=Number of laboratories reporting data.† Reference value only (not certified). Fig. 6 Results expressed as percentage recovery (based on the mean Fig. 5 Results expressed as percentage recovery (based on the mean reference concentration) plotted as a function of percentage of analyte reference concentration) plotted as a function of material density.extraction into the liquid phase of the slurry. problematic. Based on the data for the glass material GSD, there was concern that high density materials may prove each material (see Table 3). Here, the average percentage recovery for all the laboratories is based only on the mean difficult to analyze. Fig. 5 contains a plot of average percentage recovery based on the mean reference concentration as a certified reference concentration for convenience in plotting the data to show trends.These data show that significant function of the density of the material analyzed. Remember that computation of the average percentage recovery for all analyte extraction is not necessary for reasonable accuracy. Consider the determination of Pb in SRM 1633a, where only the laboratories is based on the mean certified reference concentration and does not take into consideration the uncer- 5% was extracted yet the average recovery reported by the collaborators was 98% based on the mean reference concen- tainty specified by the manufacturer.It is interesting that all of the materials which led to lower average recoveries had tration. It is also clear that reasonably high analyte extraction is not a guarantee of good accuracy, as evidenced by the data densities of 2.2 g cm-3 or greater. High density alone, however, cannot be blamed for low recoveries because there were reported for Pb in PACS-1, where microsampling limitations are seen due to the inhomogeneous distribution of Pb in this instances where good average accuracy was obtained for some of the high density materials, such as Pb in SRM 1633a (98% sediment.These data combined with the density data suggest that higher density materials with particle diameters sufficiently recovery; density=2.2 g cm-3), Cr in SRM 1633a (87% recovery; density=2.2 g cm-3), Pb in SRM 2704 (103% recovery; large to result in rapid sedimentation are most difficult to accurately analyze.density=2.6 g cm-3) and Cr in SRM 2704 (96% recovery; density=2.6 g cm-3). The reason why some high density mate- Fig. 7 contains a log–log correlation plot which summarizes all of the slurry data, with the exception of the pre-made rials are not problematic is that sedimentation will be affected by not only the density of the solid, but also by the particle slurry, for Phases 1 and 2. A total of six materials were analyzed and, with the exception of Pb in PACS-1, all mean size and the density of the diluent as indicated by Stoke’s law.25 Without the benefit of detailed particle size information reported concentrations for the collaborators overlapped the reference ranges, taking into consideration not only the mean it is difficult to confirm which materials are most likely to settle rapidly, but data from the collaborators suggest that reference value but also the reported uncertainty in the characterization of analyte content of these materials.For the com- PACS-1 and GSD would have faster settling rates than SRM 1633a and SRM 2704. This is discussed in greater detail in bined Phases 1 and 2, based on the mean reference concentration for the various analytes in the reference materials our report on the analysis of high density materials.26 Fig. 6 contains a plot of average accuracy obtained by the analyzed, the average performance was 100±5% and the range of recoveries was 78–102%.collaborators as a function of the percentage of analyte extracted. Again, remember that the mean These studies have highlighted the strengths of slurry sampling ETAAS based on the wide range of experience of concentrations±uncertainties calculated for the collaborators overlapped the reference range for each element reported in collaborators world-wide. The importance of using secondary Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 211REFERENCES 1 Bendicho, C., and de Loos-Vollebregt, M. T. C., J. Anal. At. Spectrom., 1991, 6, 353. 2 Miller-Ihli, N. J., Anal. Chem., 1992, 64, 964A. 3 Jackson, K. W., and Qiao, H., Anal. Chem., 1992, 64, 50R. 4 Gregoire, D. C., Miller-Ihli, N. J., and Sturgeon, R. E., J. Anal. At. Spectrom., 1994, 9, 605. 5 Moens, L., Verrept, P., Boonen, S., Vanhaecke, F., and Dams, R., Spectrochim. Acta, Part B, 1995, 50, 463. 6 Miller-Ihli, N. J., J. Anal. At. Spectrom., 1994, 9, 1129. 7 Miller-Ihli, N. J., Fresenius’ J. Anal. Chem., 1990, 337, 271. 8 Miller-Ihli, N. J., Fresenius’ J. Anal. Chem., 1993, 345, 482. 9 Stephen, S. C., Littlejohn, D., and Ottaway, J. M., Analyst, 1985, 110, 1147. 10 Miller-Ihli, N. J., J. Anal. At. Spectrom., 1988, 3, 73. 11 Doc¡ekal, B., J. Anal. At. Spectrom., 1993, 8, 763. Fig. 7 Correlation plot (log–log) summarizing all slurry data from 12 Epstein, M. S., Carnrick, G. R., Slavin, W., and Miller-Ihli, N. J., Phases 1 and 2. Anal.Chem., 1989, 61, 1414. 13 Hinds, M. W., and Jackson, K. W., At. Spectrosc., 1991, 12, 109. wavelengths and sufficiently large amounts of solid to be 14 Bendicho, D., and de Loos-Vollebregt, M. T. C., Spectrochim. Acta, Part B, 1990, 45, 679. representative of the bulk material has been demonstrated. 15 Dobrowolski, R., Spectrochim Acta, Part B, 1996, 51, 221. Possible problems leading to inaccurate results have been 16 Hoenig, M., and Cilissen, A., Spectrochim. Acta, Part B, 1993, highlighted, including the use of mini-flows, inappropriate 48, 1303. amounts of chemical modifier, high thermal pretreatment 17 Carnrick, G. R., Daley, G., and Fotinopoulos, A., At. Spectrosc., temperatures, wall atomization, low atomization temperatures, 1989, 10, 170. short atomization and/or read times, expulsion losses, inad- 18 Miller-Ihli, N. J., Spectrochim. Acta, Part B, 1995, 50, 477. 19 Slavin, W., and Manning, D. C., Spectrochim. Acta, Part B, 1989, equate background correction and sample sizes which were 44, 1245. too small. This study demonstrates that low values may be 20 Standard Reference Material 1547 (Peach L eaves), NIST obtained when sedimentation errors occur when analyzing Certificate of Analysis, National Institute of Standards and high density materials. The overall performance for the wide Technology, Gaithersburg, MD, 1991. range of analytes (Pb, Cr and Cu) and the wide range of 21 Reference Material DOLT-1 (Dogfish L iver), NRCC Certificate, materials suggests that slurry sampling has become a suffic- National Research Council of Canada, Ottawa, 1986. 22 Standard Reference Material 1633a (Coal Fly Ash), NIST iently mature technique that it may be used for quantitative Certificate of Analysis, National Institute of Standards and analysis as long as the slurry analysis conditions are optimized. Technology, Gaithersburg, MD, USA, 1979. Data from these studies highlight the fact that significant 23 U.S. Geological Survey Glass Reference Standards for the T race analyte extraction into the liquid phase is not a prerequisite Element Analysis of Geological Materials—Compilation of for accurate slurry analyses but that increased extraction can Interlaboratory Data, Professional Paper No. 1013, US Geological lead to improved measurement precision. Survey, Washington, DC, 1976. 24 Kurfurst, U., Pure Appl. Chem., 1991, 63, 1205. 25 Majidi, V., and Holcombe, J. A., Spectrochim. Acta, Part B, 1990, The author gratefully acknowledges the assistance of F. E. 45, 753. Greene. 26 Miller-Ihli, N. J., Spectrochim. Acta, Part B, in the press. Mention of trademark or proprietary products does not constitute a guarantee or warranty of the product by the Paper 6/06058G US Department of Agriculture and does not imply their Received September 3, 1996 approval to the exclusion of other products that may also Accepted October 24, 1996 be suitable. 212 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12

ISSN:0267-9477    

DOI:10.1039/a606058g

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

REVIEW Slurry Nebulization in Plasmas LES EBDON*, MICHAEL FOULKES AND KAREN SUTTON† Department of Environmental Sciences, University of Plymouth, Drake Circus, Plymouth, Devon, UK PL 4 8AA SUMMARY OF CONTENTS however, of an intractable nature, such as silicaceous minerals, refractory compounds and ceramics. Methods based on, for Introduction example, fusion and digestion using acids can result in incom- Plasma Techniques Used in Slurry Analysis plete dissolution of the sample, evaporative losses of the more Importance of Slurry Particle Size on volatile elements and contamination problems.Dissolution is Analytical Performance also time consuming and the sample preparation time often Slurry Preparation exceeds the analysis time. Hazardous and concentrated acids Grinding Methods such as HF and HClO4 may be required to digest samples. In Dispersion of Slurries addition, achieving a true multi-element dissolution for certain Magnetic Stirring and Vortex Mixing samples may be difficult, as is the case for geological materials Ultrasonic Agitation which contain a variety of minerals. The direct introduction Solid Particle Size Distribution Measurements of solids or slurries into plasmas would circumvent these Influence of Slurry Concentration difficulties and markedly reduce sample preparation time by Modifications to the Sample Introduction System combining matrix destruction and analyte atomization and Nebulizers excitation in a single step.Analytical plasmas are, in general, Spray Chambers higher temperature sources than conventional combustion Torches flames, a prerequisite if complex solid matrices are to be Use of a Sheathing Gas efficiently atomized, ionized and excited. Use of a Tandem Source The introduction of solid samples into the plasma may be Use of Flow Injection implemented using a number of well established techniques. Use of an Inverted Geometry Plasma One such device, known as a ‘swirl cup’, directly injects Calibration Techniques powders into the plasma.1 Argon gas is injected downwards Aqueous Calibration into an agitated cup containing the powdered sample. A cloud Use of Internal Standards of powder is formed by the gas displacement and is then led Use of Standard Additions to an outlet and directed to the plasma.The centrally injected Use of Empirical Correction Factors gas flow displaces the powder from the base of the cup on to Use of Intrinsic Internal Standardization the surrounding walls.Once the sample loading is sufficiently Use of Standard Slurries great, the powder drops back into the base of the cup, into Optimization the path of the gas jet. To overcome this problem, the cup is Use of Mixed Gas Plasmas agitated by a mechanical oscillator. Matrix Effects Another technique for the introduction of solid samples into Fundamental Studies plasmas employs a ‘fluidized bed chamber’.2,3 Argon gas flows Aerosol Formation, Transportation and Loss Temperature Measurements through a sintered glass disc on to which the powdered sample Applications of Slurry Sample Nebulization into Plasmas is deposited.Mechanical vibration again enables an efficient References and homogenous cloud of sample to be formed. A cyclone spray chamber is often situated above the fluidized bed to Keywords: Slurry sample nebulization ; inductively coupled eliminate larger particles or agglomerates from the main cloud plasma atomic emission spectrometry ; inductively coupled by the action of a centrifugal force.There is a possibility, plasma mass spectrometry ; direct current plasmas; microwave- however, that particle segregation according to density will induced plasmas; electrothermal vaporization ; review occur, which yields erroneous analytical results. The use of a ‘direct sample insertion device’ (DSID) is a commonly used method of solid sample introduction.4–6 Of all INTRODUCTION the methods used to introduce solids directly into the plasma, this technique has the highest analyte transport efficiency.A It is well established that the method of sample presentation graphite rod is used as a sample elevator to introduce the solid into plasmas is a critical step in an analytical procedure. The directly into the torch in the region of the induction coil below most conventional method of introducing samples to plasmas the main plasma body. Direct inductive heating of the carbon is by nebulization of dissolved samples.Traditionally, samples occurs and the sample vaporizes directly into the plasma. This have been prepared as solutions using digestion steps such as direct insertion method may give low limits of detection and fusion or acid dissolution. This facilitates introduction, cali- a wide dynamic range, but suffers from matrix effects and the bration and homogenization. Many analytical samples are, need for closely matching standards.The analytical signal is highly dependent on the position of the device in the plasma † Present address: Department of Chemistry, University of Cincinnati, P.O. Box 210037, Cincinnati, OH 45221–0037, USA. with respect to the torch central axis and, as a consequence, Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (213–229) 213poor reproducibility may occur. The use of computer con- overcome. This review aims to collate the papers published on the introduction of slurries into various plasmas.It is hoped trolled stepper motors have improved the reproducibility of sample insertion. that the review will provide instruction and insight to the analyst approaching this method of sample introduction for ‘Spark discharges’ may also be used for solid sample introduction into the ICP.7–9 The sample, after spark elutriation, is the first time. The review does not include publications on slurry nebulization in AAS. not transported directly from the spark sampling chamber to the ICP but is first carried into an intermediate chamber where the aerosol is nebulized along with water, and then introduced PLASMA TECHNIQUES USED IN SLURRY into the injector tube of the torch using a secondary argon gas ANALYSIS flow.The result is a more homogenous aerosol. ‘Microarcs’ are used in a similar manner. The argon gas flows over the Slurry nebulization has been implemented using a variety of plasma techniques.The most common is ICP-AES with numer- microarc electrodes after samplevaporization and is introduced to the ICP. The microarc ICP combination exhibits detection ous papers reporting a wide variety of studies.13–71 ICP-MS has been increasingly used in recent years, although less work limits and working curves comparable to those of solution sample introduction techniques. Matrix and ionization inter- has been published in this area68–84 as it is a younger spectroscopic technique.DCP emission spectrometry was used for ferences effects are absent or may be overcomeeasily. Precision, however, may be affected by the sample electrode material and earlier studies into slurry sample introduction14,85–91 but recent analytical trends have favoured the use of modern ICP-AES its shape. ‘Laser ablation’ is increasing in popularity as a method of and ICP-MS instrumentation. This may be attributed to the low limits of detection attainable in the hotter ICP, fewer direct solid sample introduction.10–12 The power from a focused laser is used to vaporize an area of material directly from a matrix effects and instrument availability.One paper has even reported the introduction of slurries into an MIP,92 despite solid surface. The current availability of a wide range of lasers, in particular Nd5YAG lasers and lasers of higher power, has the significantly lower kinetic temperature of this plasma compared with most ICPs. made laser ablation a popular technique for use with ICP-MS.Minimal loss of sample occurs during the vaporization stage, ETV has developed into a popular tool for the determination of trace elements in recent years. This technique permits the which is important, for example, in the analysis of materials where complete sample destruction is unfavourable. The lack possible separation of matrix constituents using particular temperature and time components before transportation of the of homogeneity of samples is a disadvantage in many applications but is an advantage when spatial analysis is required.analyte into the ICP via a stream of argon. Several papers on slurry ‘nebulization’ ETV coupled with ICP-AES have been Another approach to solid sample introduction involves the nebulization of (usually aqueous) suspensions of fine powders published,93–95 in addition to the use of ETV–ICP-MS,96,97 which offers superior limits of detection. into the ICP and is termed ‘slurry nebulization’.Slurry sample nebulization is an alternative technique to solution nebuliz- A survey of solid sampling in ICP-MS has been written by Bauman,98 who compared slurry sample nebulization with ation and has received particular attention in recent years. The technique offers many of the same advantages of other solid other solid sample introduction techniques such as ETV, direct sample insertion and laser ablation. Darke and Tyson99 also sample introduction techniques, such as elimination of complex dissolution procedures and the avoidance of hazardous compared ETV and laser ablation with slurry nebulization. A study of the literature with regard to the application of the chemicals. In addition, the technique is simple to implement, is inexpensive and requires little instrument modification.The slurry technique to biological materials has been made by de Benzo et al.100 most significant advantage of slurry nebulization is that it can be calibrated using aqueous solutions in an analogous way to solution nebulization.As other solid sampling techniques are IMPORTANCE OF SLURRY PARTICLE SIZE ON plagued by problematic calibration procedures, slurry nebuliz- ANALYTICAL PERFORMANCE ation has attracted the interest of analytical spectroscopists in a wide range of application areas. It is well known that the particle size distribution of a slurry is the limiting factor controlling analytical recovery.56 Slurry Fig. 1 shows the number of publications on slurry nebulization into plasmas published from 1981 to 1995 and reflects nebulization into plasmas requires that both the analyte transport efficiency of the slurry particle through the sample intro- its popularity as an alternative to conventional aqueous sample introduction. In the last decade, the technique has been adapted duction system and the atomization efficiency of that particle in the plasma must be identical with those of a solution.56 If for different samples and many of the early difficulties of achieving a homogeneous and stable dispersion have been these criteria are fulfilled then simple aqueous calibration may be used and precision of analytical results may be attained. Ebdon et al.38 measured transport and atomization efficiencies directly from emission intensity data.Magnesium of known concentration was absorbed on a silica-based cationexchange material and the magnesium and silicon emission signals were compared with those of the equivalent solution.The magnesium recovery was used as a measure of the slurry transport efficiency relative to the solution and was found to depend on the slurry particle size. A transport efficiency of 3–4% was obtained for particles of 10 mm diameter. This transport efficiency improved to 14–15% when the particle size was 5 mm. Neither particle size range was found to model the atomization of solutions under the plasma conditions used.Vien and Fry88 discussed the effect of slurry particle size on analytical recovery for plant tissue samples using DCP emission spectrometry. They stated that for aerosol transport through a modified DCP spray chamber, particles larger than Fig. 1 Number of papers published per year from 1981 to 1995 on slurry nebulization into plasmas. 23 mm will not be delivered to the plasma. Instead, they settle 214 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12out of the aerosol stream and exit the spray chamber via the corresponded to reduced particle size.For a basalt rock slurry, elemental intensity ratios increased sharply with grinding time. drain. Using a diffractometer, the median plant particle diameter was measured to be 89 mm, which exceeded the spray Ebdon et al.63 examined the effect of increased grinding time on analytical recovery. Increasing the grinding time of a chamber cut-off limit. By milling the sample using an abrasive milling aid, the particle size distribution was much improved dolomite slurry yielded more particles of 2.52 mm diameter and less, resulting in a higher transport efficiency.However, and the elemental recovery for Cu was found to be 94%. No correction factors, standard addition, matrix matching or grinding times above 2 h appeared to introduce contamination from the grinding media, particularly of Fe and Cr. lithium buffering were used. Fernandez et al.47 measured this particle size distribution Isozaki et al.43 found that the slurry particle size coupled with nebulizer gas flow rate influences sensitivity and accuracy effect on the ‘atomization efficiency’ of Ca, Mg, Al, Fe, Mn, Ti, Na and K in slags using ICP-AES.The sample was ground in the determination of iron in silicon nitride by ICP-AES. Approximately 90% of the slurry particles were found to be in a ball-mill for increasing periods of time. It was observed that suspensions prepared from different particle size ranges below 2.3, 8.5 and 4.4 mm for three different silicon nitride samples.It was concluded that the particle size for this produced different emission intensities for the elements in question. The emission signal was found to increase with particular sample should be below 10 mm. Broekaert et al.74 discussed the possibilities and limitations of the slurry tech- grinding time but not to the same extent for the various elements, indicating that the slag was a multicomponent nique for ICP-AES and ICP-MS.They found that the particle size of the slurry is of major importance and concluded that material. Raeymaekers et al.31 collected solid Al2O3 particles in the slurry particle size should be below the 5–10 mmlevel. Lobin�ski et al.50 looked at the effect of particle size on the solution/ aerosol on membrane filters at the top of the torch injector in the absence of an ICP discharge. The mean diameters of the slurry emission ratio for ten ZrO2 powders using ICP-AES.For samples with particles smaller that 10 mm a full recovery particles were found to be below 17 mm. Emission signals for an Al2O3 powder, a refractory oxide of low grain size, showed was obtained. Mochizuki et al.79 illustrated that sensitivity, precision and accuracy were dependent on the slurry particle good agreement with those of an Al solution. Ebdon et al.38 performed fundamental studies of refractory size for rare earth element determination in silicate rocks.The emission count rate was shown to increase with increasing samples using ICP-AES. Particle size effects were studied using alumina. Slurries of known composition with size distributions grinding time. The role of particle size on the efficiency of vaporization exceeding 90% (by volume) less than 8 mm, but of various size fractions, were analysed for their aluminium content. The finest and atomization of botanical sample slurries using fluorination –slurry sampling ETV–ICP-AES has been investigated.95 particle size alumina slurry (100% less than 2.5 mm) gave a 99% recovery, indicating that the transport and atomization Recoveries were obtained by comparing elemental intensities from the plant slurries with those obtained when the samples efficiencies of the slurry are close to those of a solution.For samples difficult to atomize, a particle size of less than 5 mm were acid digested. The recovery increased with decreasing particle size.When the average particle size was less than (the bulk being less than 3 mm) was optimum. Conditions which brought about a longer residence time in the plasma 170 mm the recovery was approximately 100%. Similar results were obtained by Fuller et al.13 for the analysis of ores using also aided omization. In the work of Darke et al.,70 calculated recoveries for slurry nebulization ICP-AES. This work concluded that the use of in situ fluorination yields excellent recoveries for larger SARM 5, a rock material, were low for a wide range of elements.A high proportion of the particles were in the range particle sizes. Hence there is widespread agreement on the criticality of 7–10 mm and it was concluded that reduction of the particle size should improve the ICP-AES results. particle size in slurry nebulization studies. With ICP sources, the deviations seen between slurries and solutions of equivalent Ebdon and Collier30 used a variety of spray chambers and torches with different diameter injectors to study particle size concentration are attributable primarily to poorer transport for larger slurry particles.The conclusion of the most rigorous effects in kaolin slurry nebulization. The use of a 3 mm id injector tube allowed the analysis of kaolin slurries with studies is that slurry particles larger that 5 mm (in some studies 2 mm) do not reach the plasma and this is responsible for the particles up to 8 mm in diameter.The recoveries were comparable to those for equivalent solutions. It was suggested that loss of signal. Such conclusions have favoured attrition in particle size reduction techniques and the use of dispersants even when particles larger than 8 mm reached the plasma they were only partially atomized or passed through the plasma to prevent agglomeration of such small particles. without being atomized. In a separate paper,28 the same workers illustrated the effects of nebulizer design, particle size, dispersants and viscosity in the analysis of kaolin by slurry SLURRY PREPARATION nebulization.Grinding Methods Recently, Goodall et al.56 introduced the size occupancy diameter (SOD) model, which assumes that the maximum It is now accepted that, unless the sample from which a slurry is to be prepared is of a naturally small mean diameter (as allowable particle size is that which allows the occupation of every aerosol droplet by one solid particle.For slurries to give with clays), then it must be ground in order to reduce the solid particle size. The transport efficiency of the slurry particle recoveries comparable to those of solutions it was stated that the particle size distribution of the slurry should not exceed through the sample introduction system and the behaviour of the particle in the plasma in terms of atomization and excitation 2.9 mm for a material of density 1 g ml-1 or 1.5 mm for a material of density 7 g ml-1.The solid particle size in the must be identical with those for a solution of equivalent concentration if aqueous calibration is to be used. The particle slurry was stated to be the fundamental limiting factor for efficient slurry nebulization into the ICP. size is the limiting parameter for efficient slurry nebulization and the maximum particle size of the slurry must be such that Ebdon and Collier28 found a wide particle size range to be the major cause of low recoveries for a kaolin sample using a single solid particle may occupy an aerosol droplet.To achieve a particle size distribution that would yield results ICP-AES. The particle size distribution of post-spray-chamber kaolin was measured and a cut off point of 2 mm for particle similar to those of an equivalent, equimolar aqueous solution, a wide range of grinding techniques have been employed. size was calculated. Halicz and Brenner25 also found that increased signal ratios The technique of choice and the actual grinding time depend Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 215on the sample being prepared. Table 1 lists the various mills required to produce a homogeneous slurry of the correct size range, often overnight for convenience. The resultant finely and size reduction methods that have been used to date, with varying degrees of success, to grind a wide range of materials. ground slurry can then be transferred to a calibrated flask through a coarse sieve which retains the grinding material, The particle size ranges that result are also tabulated.The choice of grinding or milling agent is important and will affect and diluted to volume with dispersant. The technique is inexpensive and offers the advantage of grinding a number of the analytical accuracy. The grinding material should be harder than the material being ground and should not contain samples simultaneously, depending on the type of flask shaker used.The beads have measure of hardness (MOH) values of elements that will interfere with an analysis.101 The ‘bottle and bead’ method has been widely used to 7–8. Zirconia beads may introduce contamination during the grinding step and this may be assessed by analysing ‘grinding prepare slurries from a wide variety of materials, as shown by Table 1. The technique involves weighing approximately blanks’ which have undergone a similar procedure to that of the slurry.Contamination after 11 h of grinding with zirconia 0.1–1.0 g of sample into a 30 cm3 polypropylene bottle and adding 10 g of virtrified zirconia beads. Aqueous dispersant is beads is usually significant for zirconium (4.4 mg ml-1), hafnium (200 ng ml-1), aluminium (400 ng ml-1), titanium added so the beads are covered and the bottle is shaken on a wrist-action laboratory flask shaker for the amount of time (150 ng ml-1) and iron (100 ng ml-1).102 The contamination Table 1 Grinding methods used in the preparation of slurries for introduction into plasmas Grinding Resulting method Sample Grinding time particle size/mm Comments Ref.Geological materials 40 min <60 1.00 g of sample ground with 10 g of beads 64 Bottle and bead method Geological/ 15 h (overnight) <5 Aqua regia added to prevent sample loss 81 refractory materials Refractory samples — <10 Beads: sample mass ratio=1051. Dispersant 38 added prior to grinding Firebrick 2 h 2–2.5 0.5 g l-1 Na4P2O7 added 56 Plant material 4 h <2 Material ashed before grinding 56 Coals Overnight <8 Varied in efficiency for different samples 77 Ores and minerals 1 h <8 Technique compared with micronizing mill 27 Ores 30 min <6 Acetone added 13 Rocks, sediments, 10 h (overnight) <10 Studies in sample charring performed using 63 sewage sludge sewage sludge Soils 3 h <8 Dispersant added before grinding 33 Soils and sediments 12 h <50 Lithium (as chloride) added as ionization 90 buffer Soils, zeolite, 24 h <3 Materials may be reduced to <8 mm with 76 catalysts 2 h grinding Slags 3 h <10 Sample ground in ball mill for 3 h and 47 sieved before bottle and bead method Carbon black 10 h (overnight) <1 Extremely small particle size observed using 67 optical microscopy Agricultural samples Overnight <5 Recovery 95% of equivalent aqueous 87 solution NIST Total Diet 10 h (overnight) <2 2 g of sample used with 10 g of zirconia 65 beads Spinach, garlic and 1 h — Agate spheres used with various amounts of 93 pollen PTFE Micronizing mill Silicate rocks 8 h <3 4 g of sample ground in mill 79 Ores and minerals 10–15 min <8 Produced a similar slurry distribution to 1 h 27 grinding by bottle and bead method Coal 20–30 min 6 Sample weighed directly on to agate 22 grinding pellets Coal 30 min <25 Samples ground to <38 mm in Tema disc 23 mill initially Coal 10 min <20 Dispersant added to ten times original mass 86 of coal powder Biological samples 30 min 3 0.5 g sample and 5 pieces of nylon-coated 80 stainless-steel beads used Whole plant tissue 30 min 5.7 Homogenizer used for 4 min prior to milling. 88 Alumina milling aid added. Mixing mill Plant tissue 15 min 27–154 Samples oven dried at 80°C for 6 h before 44 grinding Estuarine sediments 30 min <2 1.5 ml agate vial and agate ball used 45 Puck-type grinder Geological and — <10 80% of particles <10 mm, majority <5 mm 39 refractory materials Puck-type grinder Geological materials 10 min in puck- <2 Samples ground initially to 75 mm in puck- 25 then rotary type then type grinder then for various periods in grinder 40 min in rotary grinder rotary grinder Grinding mill Silicon carbide 10 min <38 4 g of sample ground 57 Vibration pot mill Geochemical 16 h (overnight) 7–10 Beads: sample mass ratio=1051 70 materials 216 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12introduced by the grinding medium is dependent on the length may be difficult to grind. The attrition materials used in mills and grinders such as zirconia and tungsten carbide may have of time employed and on the hardness of the sample. Fortunately, for many materials, grinding times of 2–4 h are comparable MOHvalues resulting in poor grinding efficiencies and even contamination if long periods of size reduction are sufficient (Fig. 2). For materials with MOH values of less than 7, grinding periods are shorter and the resultant contamination employed.In addition, rock samples are invariably composed of a variety of minerals which generally have different grain from the beads is decreased. Size reduction may also be carried out using a micronizing sizes, MOH values and friabilities.81 Grinding may therefore reduce the particle size of some of the minerals to a lesser extent mill and has been used extensively in the preparation of slurries (Table 1).The grinding vessel is a polypropylene jar of 125 ml than others. This will result in poor analytical accuracy for some elements after aspiration of the slurry into the plasma. volume with a threaded-capped polythene lid. Agate grinding elements are packed into the jar in an ordered array and Fractionation in the sample introduction system may also occur for samples where particle size varies with sample composi- 0.1–2.0 g of the solid sample is added along with 15 ml of the dispersant of choice.The jar is then clamped into the process tion.38 Generally, grinding procedures which produce more uniform particle size, e.g., ball mills, are therefore preferred to timer and vibrated for a suitable period of time to obtain the particle size required (10–30 min). The major elemental com- techniques which may produce a wide range of particle sizes. A review by de Benzo et al.100 discusses the importance of position of agate is silicon (as silica) with minor impurities of aluminium, sodium, iron, potassium, calcium and magnesium preparing a homogeneous and representative sample for biological materials. They argue that trace elements in biological (normally less than 0.02%).The MOH of agate is 6–7 and the material is extremely resistant to abrasion during the grind- samples may be evenly distributed in the lattices of the sample matrix or they may be present as discrete particles found ing process. A ‘mixing mill’ is an impact grinder and blender, suitable irregularly throughout the matrix. Trace elements are present as salts in most biological tissue and liquids or are bound to for the preparation of slurry samples.It has a three-dimensional action that provides a vigorous grinding motion, which impacts proteins, vitamins and enzymes. Total sample dissolution ensures elemental homogeneity but for biological slurry on the sample vial at a rate of over 100 collisions per second so that the sample can be rapidly ground.Stainless-steel, samples the homogeneity depends on the distribution of the element of interest in the original material and the method of tungsten carbide and agate vials may be used. Grinding solid samples using stainless steel results in Fe, Cr and Ni contami- grinding. Low degrees of precision are obtained if there is difficulty obtaining a homogeneous slurry, but this may be nation. In addition, halide-releasing compounds are not suitable for use with steel and may corrode the grinding material.overcome, for instance, by taking all the tissue from one part of the plant or animal. Tungsten carbide is harder and heavier than steel and therefore grinding times for slurry preparation may be reduced. Tungsten Biological materials that contain a high degree of fibrous material may be resistant to grinding. In addition, some plants and cobalt (a binding agent) are causes of contamination. Vibration mills consist of a torus-shaped or cylindrical- may contain components that can appreciably increase the viscosity of slurries at relatively low sample loadings.87 shaped shell.The solid is contained in the shell, together with the beads for grinding, and vibrated. The grinding material Charring the samples by placing them in a muffle furnace at a temperature of 200–300°C breaks down this fibrous structure and sample collide with each other and the shell, resulting in the breakdown of solid particles.The mills can be run dry or and allows grinding of the material to the required particle size range.87 This technique was found to yield a favourable wet and steel, agate and tungsten carbide are among the grinding materials used. particle size distribution in the analysis of a sewage sludge,63 although a reduced grinding efficiency and analytical recovery A ‘puck-type grinder’ reduces the particle size of slurry samples by the action of a spinning puck and a ring inside a were observed for copper.It was concluded that the element was associated with the denser fraction of the slurry and grinding container. The grinding container may be made of hardened steel, tungsten carbide, agate or alumina ceramic. therefore particles were deposited in the sample introduction system. Losses due to the volatility of certain species must also The last material is almost pure aluminium oxide with trace amounts of silicon, calcium and magnesium.The material is be considered if the charring technique is used and consequently low temperatures are recommended. resistant to abrasion and is lightweight, although brittle. It is known that materials containing a large proportion of brittle, hard composites can be ground efficiently and make Dispersion of Slurries excellent slurries if a correct dispersant is chosen. This is the case for some geological samples such as rocks, sediments and Preparing slurries in aqueous solution alone is unsuitable for the majority of samples owing to flocculation effects which soils.However, some materials which have a high measure of hardness (MOH) value, such as silicates and other refractories, result in rapid sedimentation of the finely powdered material. It is therefore essential to prepare a stable and homogeneous slurry that will achieve a stable, homogeneous aerosol for introduction into the plasma, yielding accurate and precise analytical results.This is achieved by employing stabilizing agents, commonly termed ‘dispersants’ or ‘surfactants’. Table 2 shows the different dispersants that have been used in the preparation of slurries from a wide variety of samples. If the surface of the solid in the preferred liquid carrier is lyophobic, the powder will be difficult to disperse. If it is lyophilic the powder disperses easily.103 To obtain a suitable dispersion of lyophobic particles, stabilizing agents are added to wet the surface so that the particle becomes lyophilic. There have been few in-depth studies in the literature regarding the stabilizing effect of dispersants.Farin�as et al. published an excellent study on the colloidal stability of ceramic suspensions for the nebulization of slurries Fig. 2 Particle size distribution for dolomite. for ICP-AES.61 They reported a case study using alumina Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 217Table 2 Dispersants used to stablize slurries for nebulization into plasmas Dispersant Concentration Sample Ref.Aerosol OT 0.5 g l-1 Plant material 56 0.1% m/v Coal 69, 77 0.5% m/v Sewage sludge 63 Ammonia 0.35% v/v Slags 47 0.35% v/v Kaolin clay 28, 30, 89 Darvan-7 0.5% m/m Ceramics 61 Darvan-C 0.5% m/m Ceramics 61 Dolapix PC-33 0.5% m/m Ceramics 61 Glycerine+0.5 M HCl 40% ZrO2 36 Glycerol+Kodak photoflow 40% v/v Ceramic, geological and refractory materials 39 2% v/v Sodium hexametaphosphate 0.1% m/v Refractory samples, sulfide ore 27,38 Sodium hexametaphosphate+ 0.1% m/v TiO2 29 monoisopropanolamine 0.02% m/v Tetrasodium pyrophosphate 0.05% m/v Soils, catalysts, geological samples,irebrick 76 0.1% m/v Geochemical materials silicon nitride 70, 81 1% m/v Soils 33 0.5 g l-1 Firebrick, sediment, sewage sludge 43, 56, 63 Triton X-100 0.005% v/v Coal, total diet, oyster tissue, lobster hepatopancreas 96, 97 0.01% v/v Soils, milk, plant tissue 87 Sediments 45 0.05% v/v Tea leaves 59 0.1% v/v Milk powder 75 Soils, sediments 82, 90 Silicate rocks 79 Silicon carbide 57 0.5% v/v Whole plant tissue 88 Coal 15, 22, 86 Total diet 65 Carbon black 67 1% v/v Coal 23, 24 Biological samples 80 Sewage sludge 63 Xylene 100% Airborne particulate matter 16 HCl 1% m/m Silicon nitride 60 2 mol l-1 Clays 17 H2SO4 2% v/v Biological samples 41 HNO3 1% m/m Zeolites 94 10% m/m Lobster hepatopancreas, marine sediment 92 0.01 M Citrus and tomato leaves 44 1 M Marine sediment 52 HNO3+NaCl 1 M Clays 48 0.1 M NaCl 0.1 M Clay minerals 40 None Animal tissue 85 Geological materials 25, 55, 64 Clays 25 Si3N4 35, 42 Al2O3 74 SiC 74 Foods 93 ZrO2 50 Ores and flotation feeds 18 slurries when different pH conditions and stabilization agents At a particular concentration of potential-determining ion, the positive and negative surface activities will be equal and were used.The paper shows the different types of stabilizing additives available for ceramic suspensions and also the best the overall potential at the surface of the particle will be zero.In this scenario, no double layer exists and agglomeration of use of them. The authors explain the electrostatic stabilizing mechanism in detail and state that in a ceramic suspension, the particles is observed at a concentration known as the isoelectric point. This demonstrates the important role a H+ and OH- ions determine the potential. In the case of Al2O3 , the charge on the surface of any slurry particle is dispersant plays when added to a slurry.The potential at the surface of a slurry particle is known as the zeta potential. A negative, resulting in the attraction of protons to the surface and creating a gradient of concentration (of protons) from the slurry is only stable when the zeta potential is high and significantly removed from the isoelectric point. The addition particle surface to the liquid. Conversely, the OH- ion concentration is decreased near the surface of the particle and of a dispersant causes the zeta potential to change and ultimately acts to stabilize the slurry.Clearly the concentration increases with distance from the surface. A charged layer, known as the electrical double layer, is formed because of the of a dispersant must be controlled, as an excess can be detrimental to the slurry stability and be a cause of coagulation. potential gradient set up. The surface potential of the particle may be changed according to the Nernst equation by changing Triton X-100 was found not to stabilize alumina slurries and no real dispersion was achieved with Kodak photoflow in the concentration of the potential determining ion.The relative adsorption of ions on the surface of the slurry particle will glycerol. PKV-5088 was unable to prevent sedimentation of alumina despite being commonly used to disperse non-oxide alter as a result. 218 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12ceramics. Three dispersants, Dolapix PC-33, Darvan-C and recoveries of Mo were found to be very good for all the certified reference materials analysed. Darvan-7, were found to produce the desired dispersion of an alumina slurry. The intensity and precision of the ICP-AES measurement were found to be directly related to the stability Magnetic Stirring and Vortex Mixing offered by the dispersant system. The authors stated that the dispersant additive may yield the required stability providing Magnetic stirring of a slurry is frequently performed, after preparation and dispersion, whilst the slurry is being pumped an electrostatic stabilizing mechanism by either (i) controlling the pH (using potential determining ions), (ii) using inorganic to the sample introduction system.47,59 This ensures a homogeneous distribution of the solid material in the liquid phase electrolytes or (iii) using a polyelectrolyte where long-chain polymers adsorbed on the surface of particles prevent contact and prevents sedimentation of particles.Errors caused by sedimentation are directly related to particle size and the between them. In the same paper, the influence of pH on the slurry stability presence of dense solid particles of varying composition may cause differential sedimentation. was examined by plotting the zeta potential versus pH for Al2O3. The pH was altered by the addition of ammonia solution Many geological materials contain magnetic, iron containing minerals such as magnetite (Fe3O4).The use of magnetic and the point of maximum zeta potential evaluated. The optimum pH for the preparation of the homogeneous alumina stirring would result in migration and even adhesion of some slurry particles to a magnetic stirring bar.102 Vortex mixing slurry was 4–5 and it was concluded that pH is the parameter that controls stability. As the stability of the slurry increases, so may be used in these instances as an alternative and effective method to ensure homogeneity of slurry samples before too does the emission intensity and the RSD of the values are lower. Fuller et al.13 described how the pH of an ore slurry was aspiration into the ICP.64 adjusted to between 6 and 7 to maintain stability.Van Borm et al.49 state that ‘the stability of a slurry is a Ultrasonic Agitation crucial prerequisite for efficient and reproducible nebulization’. The influence of pH on the stability of 1% m/v Al2O3 and 1% Slurries have been homogenized using ultrasonic agitation as an alternative to grinding the sample, or in addition to m/v SiC slurries was given as an example.The stability was measured using a microelectrophoresis technique in an alter- grinding. Samples that do not require grinding are well documented. Ebdon and Collier28 prepared slurries of 0.2% m/v nating field. The Al2O3 slurry was found to be stable between pH 3 and 7 and the addition of HCl increased stability.Adding unground kaolin samples, dispersed with 0.35% m/v NH3. Slurries were placed in an ultrasonic bath for 30 min and ammonia solution to an SiC slurry in water caused the pH to change from 3.8 to 10 with a concurrent increase in stability. particles were found to be less that 2 mm in diameter. Laird et al.48 analysed air dried clay, of various size fractions, using Lobin�ski et al.50 attempted to stabilize a ZrO2 slurry by optimizing the pH.Two optima were found at pH 1–2 and ICP-AES by sonicating 0.05 g of the sample at 40 W in 100 ml of either 1 M HNO3 or 0.1 M NaCl. 10–11. The addition of an acidic medium gave a broader optimum and so in further preparations all slurries were Biological samples have been prepared in a study by Qin et al.95 Slurries were dispersed with an ultrasonic vibrator for acidified to pH 2. Ebdon and Collier28 investigated three dispersants commonly 20 min and shaken vigorously prior to sampling.Miller-Ihli and co-workers96,97 have prepared several biological samples used in the stabilization of kaolin slurries: Calgon (sodium hexametaphosphate), Dispex and aqueous ammonia solution. for slurry ‘nebulization’ ETV–ICP-MS. After diluting the samples, sonication was carried out for 30 min. The mean Aqueous ammonia was found to be the most effective in dispersing the irregular, flat, plate-like kaolin particles which particle size of lobster hepatopancreas96 was below 50 mm.Lobster hepatopancreas was again analysed together with a are known to be prone to ‘edge to centre’ electrostatic attraction forces. Equivalent slurry and aqueous atomisation efficiencies certified marine sediment by Matusiewicz and Sturgeon92 using MIP-AES. Fine powders of CRM materials were suspended for Mg were obtained. Dilute ammonia solution (0.35%) was also found to be the most effective dispersant in the analysis of in a 10% solution of nitric acid to make a 1% m/v slurry. Suspensions were pre-treated by sonication prior to analysis slags in a study by Fernandez et al.47 Triton X-100, sodium pyrophosphate d ammonia were compared and ammonia for 5 min using an ultrasonic cell disrupter in order to disintegrate particle agglomerates and to complete the dispersion.was found to produce the most stable emission signals for all the elements under investigation. Ebdon et al.63 compared Periodic sonication using an ultrasonic probe was used to agitate the sample while it was pumped into the MIP in order Aerosol-OT and Triton X-100 to determine whether reduced analytical recoveries were caused by agglomeration of slurry to maintain a uniform slurry.Broekaert et al.74 analysed a selection of ceramic powders particles in the analysis of a sewage sludge. The best recoveries were found for Pb, Cu, Zn and Mn when a 1% Triton X-100 (Al2O3, SiC and ZrO2) by ICP-AES. The refractory material was added to water and placed in an ultrasonic bath for aqueous dispersant was used as the dispersing medium.Some early studies concerning the viability of slurry nebuliz- 10 min. The authors apparently did not find it necessary to add a surfactant to stabilize the slurries, which is suprising in ation as an alternative sample introduction technique did not realise the importance of the addition of dispersant as a slurry the light of the studies reported above in the section Dispersion of Slurries.The particle size distribution was measured to be stabilizing agent. Halicz and Brenner25 prepared slurries of silicate materials in the absence of a dispersant. Poor analyte between 0.2 and 5 mm. Za�ray and co-workers35,42,60 analysed silicon nitride samples signal intensity ratios were obtained and calibration functions were not consistent. Further work55,64 concluded that slurries by ICP-AES. Suspensions, dispersed in 1% m/m HCl or 0.5% sodium hexametaphosphate, were agitated ultrasonically for of particle size less that 1.5 and 2.0 mm did not behave similarly to aqueous aerosols.Again, a surfactant was not introduced 5 min before nebulization. The particle size distribution was measured to be not greater than 0.8 mm. However, a negative to stabilize the slurry and the poor recoveries compared with solutions of equivalent concentration are, perhaps, not surpris- deviation in results of 20–40% for Al, Fe, Ca, Mg and Ti was observed in one silicon nitride sample using slurry nebulization ing if there was agglomeration.Biological samples have also been prepared for analysis compared with an equivalent analysis where the sample had been prepared as a solution.60 This was attributed to a lower using ETV–ICP-AES without the use of dispersants.93 However, PTFE was used as a fluorinating agent to eliminate nebulization efficiency for the slurry than for an aqueous solution. Isozaki et al.43 prepared silicon nitride in a similar memory effects and appears to act as a solid dispersant as the Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 219way, the sample being dispersed in sodium hexametaphosphate with a range of slurry concentrations. The instrumentation required is relatively expensive. and sonicated for 1 min. In the work of Fernandez and co-workers,47 slag samples Suspensions of slurries may also be analysed using a Coulter Counter. This technique determines the size and number of were initially ground to less than 10 mm using a wet grinding technique. Slurries were then dispersed using 0.35% v/v particles dispersed in a conducting medium by drawing them through a small orifice with an electrode on either side.As a ammonia solution and then placed in an ultrasonic bath for 15 min. particle passes through the sample orifice, a change in resistance is measured which then generates a voltage pulse. This Airborne particulate matter was prepared as a slurry by Sugimae and Mizoguchi.16 Particles were suspended in xylene change in potential difference is directly proportional to the volume of the particle in the slurry.The pulses are then and dispersed using an ultrasonic immersion device for 5 min. Ultrasonic agitation continued during sample nebulization amplified and quantified to give a size distribution of the slurry. As the particles are drawn through the sample orifice into the plasma for ICP-AES analysis.It has been reported that the analyte of interest may be under pressure, loose flocs are broken down to give a primary dispersion which may not reflect the ‘true’ dispersion. The use partially extracted into the liquid phase, owing to the ultrasonic action, when slurries are prepared in an acidic medium. Huang of a proprietary dispersant is required in this method and this may differ from that proposed for the analysis which may and Shen36 analysed ZrO2 powder by slurry sampling ICPAES.The ultrafine powder was dispersed in 40% glycerine introduce errors. Scanning electron microscopy (SEM) is also a common and 0.5 M HCl using ultrasonic agitation. After 1.5 h agitation, a homogeneous suspension had formed and the particle size technique used to measure the mean particle size of a slurry. A beam of electrons scans across the sample in a series of was less than 0.1 mm. Fe and Mg were determined by aspirating the slurry into the ICP.Subsequently, the flasks were centri- parallelpaths. The electrons interact with the sample producing secondary electron emission, backscattered electrons and fuged to settle the particles and the clear supernatant fluid was analysed to determine the dissolved Fe and Mg concentrations X-rays. These signals are then displayed on the screen of a cathode-ray tube and photographic images may be obtained. as a result of ultrasonic treatment. ‘Some Fe and a considerable amount of Mg’ were found to be leached into solution.The technique has a large depth of focus (about 300 times that of an optical microscope) and can yield significant information Manickum and Verbeek59 determined the concentration of Al, Ba, Mg and Mn in tea leaves by ICP-AES slurry nebuliz- about the size, shape and surface characteristics of a particle. Unfortunately, SEM is only applicable to solid, conducting ation. The sample was sonicated for 5 min with a solution containing 10 M HCl and Triton X-100. It was found that surfaces.Optical microscopy can also be used to check the dispersion between 50 and 60% of the total Al and Ba and between 70 and 77% of the total Mg and Mn were extracted during this of a slurry. It is useful to observe slurries in this manner immediately prior to sample introduction into the plasma to 5 min dispersion period. It was noted that similar amounts of analyte elements are leached in the usual 5 min brewing time ensure that loose agglomeration of particles has not occurred.Given the widespread availability of optical microscopes and of tea made with boiling water. the ease of use, it is suprising that this simple precaution is not more widely employed. SOLID PARTICLE SIZE DISTRIBUTION MEASUREMENTS INFLUENCE OF SLURRY CONCENTRATION The particle size distribution of a fine powdered sample may be determined by examining a suspension of that material. The concentration of the slurry is an important factor to consider during preparation.Slurries can be diluted, but only Various methods have been used to measure the particle size distribution of slurries to ensure suitability of the slurry before within a limited range as the precision may be degraded with slurries that are highly diluted. This is because of the smaller analysis by plasma techniques. Often, the method by which the particle size is measured depends on the instrumentation number of particles in the total volume which remain after dilution has been performed.Ebdon and Wilkinson23 prepared available, so there is no rigid methodology to follow for a particular sample. Table 3 lists the techniques that are coal slurries in the concentration range 1–25% m/v for ICPAES analysis. The effect of sample pumping rate with varying documented in the literature along with the particle size that was measured for the particular sample. concentration was optimized. For an argon plasma, the signal to background ratio increased linearly with increasing slurry The technique of photosedimentometry combines photoelectric measurement and gravitational settlinof particles.A concentration up to 20% m/v. The effects of viscosity and non-Newtonian rheological phenomena should be considered narrow beam of light is passed through the slurry and on to a photocell. For a homogeneous suspension, the concentration for very concentrated slurries. Williams et al.76 prepared slurries of refractory compounds of solid particles in the beam of light will be the same as the concentration throughout the suspension.As particles begin for nebulization into an ICP mass spectrometer. Slurries of 1 g per 100 ml were originally prepared but it was found that, to fall and leave the light beam, they are replaced by particles settling from above. The attenuation of the beam of light is after 30 min of nebulization, partial blockage of the sampling cone occurred due to the deposition of particles.Two dilutions directly related to the surface area of the particles in the light beam and, from this relationship, the particle size distribution (0.05 g and 0.0001 g per 100 ml) of the slurry were therefore made prior to the analysis for minor/trace level elements and can be determined. While this technique was once widely employed, it has waned in popularity now that more rapid for major elements. No blocking problems persisted after dilution.Ebdon et al.77 made similar dilutions for minor trace methods are available. Laser diffraction methods quantify the diffracted light from elements and major elements in the analysis of coal by ICP-MS to prevent cone blockage. An alternative way of reducing the a suspension. The angle of diffraction from each particle in the slurry is directly proportional to the particle size. The intensity sample loading on the cones is to use flow injection to sample the slurry (see the section Use of Flow Injection, below).of the diffracted light is a measure of the surface area of the particle. In the droplet size analysers commercially available, Mochizuki et al.79 determined rare earth elements in silicate rocks using ICP-MS. Again, the use of slurries with a high a He–Ne laser is commonly used and different instruments cover different particle size ranges. Such methods are rapid, solids concentration resulted in deposition on the sampling cone and the analytical signal fell with time.It was stated that do not require the use of dispersants and can be employed 220 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Table 3 Techniques used to measure the particle size distribution of slurries before nebulization into plasmas Particle size measurement Particle size method Sample distribution/mm Comments Ref. Al2O3 <0.3 No particles larger that 16 mm observed 37, 49 Automated electron probe microanalysis TiO2 powder <5 Number of particles sized=500 29 ZrO2 powder <1.4 Before analysis, slurries filtered over 50 Nuclepore membrane Coulter Counter Carbon black <1 Optical microscopy used to check dispersion 66 Firebrick 2–2.5 Optical microscopy showed aggregation of 56 Plant material <2 5–10 particles for plant material Geochemical materials 7–10 Particle size effects observed 70 Kaolin Various fractions from 0.5 Particle size distribution measured for various 28, 30 up to 50 fractions Ores <6 — 13 Refractory samples <10 Slurry does not model atomization of an 38 equivalent solution Rocks, sediments, sewage <10 Optical microscopy used to ensure efficient 63 sludge dispersion of slurries Soils, zeolites, catalysts <3 — 76 Total diet <2 Optical microscopy used to check for 65 agglomeration Whole coal <25 Various size fractions analysed 23 Whole coal Two samples <2 and two Various certified coals analysed 69 <10 Laser diffractometer Coal <8 Particle size distributions measured for 77 various coals Coal 6 Non-intrusive aerosol droplet size 22 measurements also performed using the diffractometer Geological samples <2 Particle size analyser used 55 Whole plant tissue 5.7 Diffractometer could not distinguish between 88 milling aid and plant tissue Refractory oxide powders <20 Sub-mm condensates and remolten larger 31 particles identified Photosedimentometer Slags <10 — 47 SEM Geological materials <2 Silicate rocks and glass samples analysed 25, 64 Botanical samples <400 Samples prepared by sieving 95 Optical microscopy Coal <20 Particles obtained from different grinding 86 (430× magnification) techniques differentiated Submicron particle sizer Biological samples 3 Flour powder analysed successfully whereas 80 liver and pine needles posed problems X-ray diffraction Rock samples <0.2–45 Particle size and mineralogy detected by XRD 48 the upper limit of concentration depends on the sample matrix consists of three basic parts: a nebulizer, which produces a fine aerosol of solution and, if of the correct design, slurry par- being analysed.For silicate rocks, slurries of concentration 0.004–0.1% m/v were prepared, stirred and continuously nebul- ticles;104 a spray chamber, which removes relatively large particles;105 and a torch, which enables the sample to travel ized for 100 min. Aspiration of slurries of 0.1% m/v resulted in a decrease in count rate over time.No sampler cone orifice to the plasma via an injector. Other parts may be added to the sample introduction system to improve sample through- clogging was observed with slurries of 0.05% m/v. The same group80 also investigated the slurry concentration effect for put and transport efficiencies if desired. These may include desolvation devices, tandem sources, the use of a sheathing biological samples. The effect of stability on count rates of Mn, Rb, Mo and Pb was tested by continuously nebulizing gas or flow injection. 0.5% and 1.0% m/v slurries of bovine liver. Spectra were taken at 5 min intervals. It was found that slurries of 0.5–1% m/v Nebulizers could be aspirated for 100 min with no decrease in sensitivity from orifice clogging. Refractory compounds are deposited on The efficient nebulization of a sample is the most critical step the sampling cone and cause the sensitivity of the analysis to in the sample introduction system.There are many examples decrease with time. The low content of refractory inorganic in the literature of various nebulizers and their customized compounds in biological samples allows the aspiration of counterparts which are all designed to improve transport relatively high slurry concentrations. efficiencies and are suited to particular samples. Conventionally, acid dissolution procedures produce solu- The nebulization of slurries into a plasma requires a nebultions with a 1% sample content.Clearly, the ability to use izer that is unlikely to clog when solid particles are passed slurries with a sample content of 20% or more yields major through it. The most popular designs for slurries are based on advantages in trace analysis. the V-groove ‘Babington-type’ nebulizer.13,23,28,37 The sample is pumped down a groove containing a small aperture through which argon gas flows at high velocity. This causes the sample MODIFICATIONS TO THE SAMPLE to shatter into an aerosol of small droplets.The high pressure INTRODUCTION SYSTEM of the argon gas ensures efficient nebulization and prevents blocking of the small orifice. Examples of typical aerosol size The effective introduction of a slurry sample into a plasma is a critical step if recoveries comparable to those for equivalent distributions are illustrated and discussed in detail in a paper by Ebdon et al.34 There are many variations of this V-groove solutions are to be obtained.The sample introduction system Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 221nebulizer, including one design with an integral impact bead30 a Scott double pass water-cooled spray chamber constructed from glass in the analysis of coal by ICP-MS. Cooled spray and another design based on a cooled Babington nebulizer,83 both of which serve to reduce noise levels in the resultant chambers offer advantages in noise reduction as they act to lower the vapour loading in the plasma so precision is spectra.Algeo et al.19 introduced a modified Babington principle improved. The particle size of the aerosol is due to the slurry particle nebulizer with sample heater for use in the analysis of metals in lubricating oils by ICP-AES. The heating action on the and its surrounding jacket of solvent. One simple way to reduce aerosol particle size is to remove this jacket by desolv- droplet occurred as the sample was introduced into the nebulizer and served to increase the aerosol output.It also reduced ation. This technique has been used by Hartley et al.82 to improve transport and atomization efficiency in ICP-MS. The emission intensity variation that arose due to differences in oil type and viscosity. Metallic particles in lubricating oils were condenser was cooled using Peltier coolers and the slurry, which was dried before entering the plasma, appeared to also measured by Saba et al.,14 who investigated the efficiencies of various nebulizer–spray chamber combinations.A ceramic increase in atomization and transport efficiency. It was also noted that desolvation resulted in a decrease in the ionization nebulizer and horn-like glass spray chamber were found to be the most efficient of all the systems studied. temperature of the plasma presumably owing to the decreased hydrogen levels in the plasma. The Ebdon nebulizer28,30,56,77,82 is a commonly used Vgroove variation of the Babington nebulizer.It has been shown The influence of spray chamber temperature was measured by Pollman et al.83 Their spray chamber design enabled the to give excellent performance in slurry nebulization23,24,28 and is of a robust one-piece design. The diameter of the gas orifice temperature to be maintained constantly between 5 and 25°C. They found that in the analysis of Al2O3 slurries by ICP-MS, is 0.2 mm to allow a high-velocity gas flow to be used for nebulization.the RSD of the analyte signal improved upon cooling the spray chamber. A V-groove Babington-type nebulizer has been used successfully with MIPs.92 This design of nebulizer contained a very Gervais and Salin45 developed a heated spray chamber followed by a condenser for slurry sample introduction ICP- small (0.1 mm diameter) gas orifice, a solution orifice of 0.8 mm diameter and an impact bead to break up large droplets. Both AES. Both the nebulizer and spray chamber are heated with a heating coil–thermocouple arrangement.An internal con- the Hildebrand grid and maximum dissolved solids nebulizers have been described for use in ICP-AES slurry work.48 The denser follows the spray chamber. No sample overloading was observed and detection limits were improved by a factor of 10. former contains a fine platinum grid upon which the slurry is laid and the aerosol is formed when the argon passes through McCurdy and co-workers22,86 designed a spray chamber configuration without impact beads or baffles.The perform- the grid, whereas the latter is based on the V-groove design. Fuller et al.13 used a cross-flow slot-type nebulizer for the ance of the modified spray chamber was compared with that of an unaltered spray chamber in the analysis of coal by ICP- continuous aspiration of slurries over 10% m/v into the DCP with success. AES. The impactor bead was found to have a minor effect on the sample droplet size passing through the system.Mohamed Rademeyer et al.46 have developed a rotating disc nebulizer for solution and slurry sample introduction ICP-AES. The et al.85 designed a novel, single pass, glass spray chamber for use with a V-groove nebulizer for a DCP. nebulizer is not based on pneumatic or ultrasonic principles but around the action of a planar spinning disc. The disc is An in-house spray chamber was designed to complement a V-groove nebulizer for MIP spectrometry.92 A glass bead rotated at high frequency and particles are placed on the disc.The disc is contained within container walls. The high fre- spoiler was used in the spray chamber to prevent flooding and deleterious results. The fraction of sample reaching the plasma quency spinning causes the particles to be swept outwards resulting in an area of reduced pressure at the disc surface. If was measured to be 3.2%, which was argued to be consistent with the general performance of spray chambers.The use of a gas flow is introduced above the centre of the disc, the low pressure will be counteracted and the gas spirals outwards and this spray chamber design gave recoveries comparable to certified values. However, as 10% nitric acid was used to upwards, taking the sample droplets/particles with it. Turbulence occurs, resulting in mixing of the sample with the prepare the slurries, solid slurry particles partially dissolved and elements of interest were extracted into the aqueous phase.gas and the sample is introduced to a spray chamber. This factor contributed to the good analytical recoveries and it is not clear if the action of the spray chamber would be as Spray Chambers satisfactory if a non-acidic dispersant had been used to disperse the slurries. As with nebulizers, there have been many spray chambers designed and used exclusively for slurry nebulization introduc- A ‘reduced volume’ spray chamber has been designed33 which excelled in performance over a conventional Scott tion into plasmas.The spray chamber separates the droplets that emerge from the nebulizer. Large droplets are rejected double pass spray chamber. The spray chamber has minimal volume and does not introduce excessive spectral noise. Ebdon and removed via a drain whilst small droplets are allowed to pass to the plasma torch. Approximately 1–2% of the sample and Wilkinson23,24 reported using a recycling spray chamber to conserve sample solution when using high sample flow reaches the plasma; the rest is pumped out of the spray chamber.There are two basic designs of spray chamber used rates. The waste from a conventional double pass chamber was collected into a slurry reservoir and continuously presented for solution and slurry analysis, the ‘double pass’ design, which consists of a chamber with an inner barrel, and the ‘single to the nebulizer for aspiration into the ICP. Ebdon and Collier30 compared the designs of four spray pass’, which is a single chamber sometimes containing a baffle or impact bead which aids the production of small droplets.chambers; a conventional double pass, a cyclone, a ‘straightthrough’ and ‘direct’ spray chamber. The cyclone is a vortex Ebdon et al.63 used a Scott double pass Ryton spray chamber that gave good recoveries for slurries of rocks and soils. It was type chamber which has been used previously for the aspiration of solutions into the ICP.The ‘straight-through’ prototype observed that a considerable dead volume existed and so a customized glass double pass spray chamber was made. This design had minimum impact surfaces, thus allowing larger particles to reach the plasma. The ‘direct’ spray chamber gave improved recoveries for a sewage sludge sample. Cooled spray chambers have also been used to improve allowed large particles to enter the torch injector directly and minimized impaction processes. It was found that the spray recoveries. Williams et al.76 used a single pass water-cooled spray chamber, maintained at 13°C.Ebdon et al.77 employed chambers filtered out the majority of particles, particularly the 222 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12larger ones. The ‘straight-through’ spray chamber was chosen stream. They illustrated that the use of flow injection is as applicable to slurries as it is to solutions and offers many for further studies.While there has been very considerable research into this advantages, including the ability to introduce a very concentrated slurry into the plasma via a carrier stream, resulting in area of sample introduction, none of the marginal improvements reported can be classified as ‘breakthroughs’. It would long-term stability of results. The particular advantages of FI-slurry nebulization for ICP-MS have already been noted appear that a single system based on a Babington-type nebulizer, e.g., an Ebdon nebulizer, and a conventional double pass (see the section Influence of Slurry Concentration).spray chamber, provides a satisfactory sample introduction system for virtually all samples. Use of an Inverted Geometry Plasma Algeo et al.19 described initial investigations with an inverted Torches geometry plasma for the analysis of metals in lubricating oils by ICP-AES. They found that with a conventionally oriented Torches have been modified for slurry sample introduction in sample introduction system, a high proportion of large metal the work of Verbeek and Brenner32 and Varga et al.35 In the particulates settled out in the spray chamber and were removed former study, an ‘extended torch’ was designed which extended via the drain.In an attempt to overcome this effect, the ICP 6.5 cm above the injector tip. Emission for slurry nebulization was turned upside down so that the spray chamber was was found to be spread over a greater range of height when situated above the torch. However, this caused plasma instabil- compared with a Trassy–Mermet torch. No appreciable ity due to condensates accumulating on the walls of the spray improvement in signal-to-background ratios (SBRs) was chamber and dripping into the torch.A new design of spray obtained. Varga et al.35 stated that an extended torch can chamber was made to prevent this and the accuracy of results eliminate carbon contamination from the air in the analysis of was improved owing to fewer particle fractionation problems Si3N4 by ICP-AES.occurring in this region of the sample introduction system. Ebdon and Collier30 investigated the effect of various torch injector tubes on analyte sample transport. Capillary tubes of 2.2, 3 and 4 mm id were used along with a 2 mm i.d. jet injector CALIBRATION TECHNIQUES tube. The authors found that the internal diameter of the Aqueous Calibration injector dictated the mean particle droplet size of the aerosol entering the plasma.The 4 mm and jet injector tubes were For a slurry to be efficiently nebulized, vaporized and atomized, found to be unsuitable for slurry atomization as the sample the particle size distribution, as previouslymentioned, is crucial. was reported to ‘cone out’, resulting in a reduction of sensitivity A narrow particle size distribution with a mean particle size of the signal. The 3 mm injector tube was found to be the of less than 2 mm will ensure sample transport and recoveries most suitable for slurry nebulization. A 3 mm injector was also comparable to those for the equivalent solutions for most used in the work of Foulkes et al.27 and the same group63 samples.If this is achieved then simple aqueous calibration described the use of a demountable torch with 2 mm alumina may be used successfully. This technique has been used by injector that gave accurate analytical results for a range of most analysts who have achieved the desired mean particle certified reference materials. Jarvis and Williams78 highlighted size and stability of the slurry.For example, Totland et al.81 the importance of using a wide bore injector for slurry nebuliz- used aqueous calibration standards in the analysis of geological ation into the ICP. They used a 3 mm diameter injector that samples by ICP-MS. Stock solutions (1000 mg ml-1) of plati- resulted in a decrease in sample velocity through the injector num group elements and gold were serially diluted to produce aperture and hence a longer residence time in the plasma.multielement standards in the required range with no need for matrix matching. Ebdon et al.63 prepared standard solutions in the same way. Clearly when using dispersants which are Use of a Sheathing Gas surfactants, the nebulization efficiency can be altered signifi- Halicz and Brenner25 used a demountable Trassy–Mermet cantly and, not suprisingly, it was found that improved analyt- torch with a sheathing gas tube between the torch and a ical accuracy was achieved if the standards were matrix nebulizer that enhanced aerosol injection into the plasma for matched to the samples by adding equivalent amounts of the analysis of geological samples by ICP-AES.dispersant. Use of a Tandem Source Use of an Internal Standard Karanassios et al.52 employed an oxygen–hydrogen flame Analytical accuracy and precision have also been markedly sample introduction system for marine sediment slurry analysis.improved by employing an internal standard such as yttrium64 The flame reduced the mean particle size of the slurry particle and scandium25 in ICP-AES. In ICP-MS aluminium has been for those particles greater than 10 mm in diameter. This was used with success.79 Ebdon et al.77 reported the use of two stated to eliminate the need for fine grinding before introduc- internal standards in ICP-MS analyses, rhodium for a tion into the ICP completes vaporization.Sustaining the flame semiquantitative analysis and indium for a fully quantitative and plasma at the slightly different pressures required and study. The internal standard may be used to compensate for providing an effective interface were found to be relatively reduced transport efficiencies and inefficient atomization, par- straightforward. However, many difficulties were encountered, ticularly in the analysis of refractory elements in geological including plasma instability, non-uniform heating, extinction samples. It is important that any internal standard used in of the flame and deposition of analytes on the expansion aqueous solution is compatible with the selected dispersant.chamber resulting in memory effects. Use of Standard Additions Use of Flow Injection It is possible with the slurry technique to calibrate by standard additions with aqueous solutions.50,60,75 However, this tech- Ambrose et al.33 reported the use of flow injection (FI) ICPAES for the analysis of soil slurries with a low volume spray nique again assumes identical transport characteristics between solutions and slurries.Lobin�ski et al.50 used the technique in chamber. Flow injection is a technique in which a known sample volume is injected into a continuous mobile carrier the analysis of ZrO2 by ICP-AES. Silicon nitride powder was Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 223analysed by Za�ray et al.60 using standard additions of a The same method was used by Ebdon and Collier28 for a 0.5% m/v kaolin slurry, but in this case detection was by ICP-AES. multielement (Al, Fe, Ca, Mg and Ti) stock solution. Ebdon and Wilkinson23,24 used a variable step size simplex procedure to optimize for the determination of Mn in coal Use of Empirical Correction Factors slurries by ICP-AES. Plasma gas flow, auxiliary gas flow, nebulizer gas flow, rf power and viewing height above the load Alternative calibration methods have been used when analysing slurries, again to improve analytical accuracy.Empirical cor- coil were the variables investigated. Both an all-argon plasma and an argon–nitrogen plasma were optimized. The SBR was rection factors have been employed20,36 where the intensity of an elemental line for a slurry analysis has been lower than used as the criterion of merit. Gervais and Salin45 optimized their procedure for the analy- that of an equivalent solution.Conventional internal standardization will not suffice because of the differences in trans- sis of slurries by ICP-AES using a heated sample introduction system. Simplexoptimization was used to optimize temperature portation and dissociation behaviour between the analytes present in the slurry particles and the internal standard. Thus and viewing height. The signal-to-blank ratio for six elements was used as the criterion of merit and a variation of 5% a series of slurries may be analysed by dissolution in addition to slurry nebulization and correction values computed empiri- between the five best vertices was used as an indicator of the optimum being found.The simplex method was also used for cally which can later be applied to unknown slurries. Such a method relies on similar behaviour in all the slurries the optimization of ZrO2 slurries for the determination of Al.50 Forward power, sample uptake rate, nebulizer pressure, investigated.plasma gas flow rate and viewing height were the variables under investigation. Use of Intrinsic Internal Standardization Ebdon et al.63 carried out single element and multielement optimization of viewing height, nebulizer gas flow and rf power Variations in transport efficiency between aqueous standards and slurry samples may be overcome by using an intrinsic in the simultaneous multielement analysis of five certified reference materials.Instrumental on-board algorithms and internal standard.21,28 An element in the slurry sample is selected to be monitored. The analyte line signals are then simplex optimization were compared and it was found that the on-board algorithms provided optimum conditions that ratioed to the signal for the selected element. If the intrinsic internal standard element is present in a constant amount in yielded superior results in a shorter time period. Freon assisted ETV of slurries for ICP-AES analysis has all the samples, it can be used to correct for varying transport efficiency.Silicon has been used effectively in the analysis of been optimized for a 1 kW plasma and at a viewing height of 15 mm above the load coil.94 Integrated signals were used as kaolin by ICP-AES.28 The use of an intrinsic internal standard may, however, be invalidated if an acidic medium is used and the figure of merit, as the authors believed that this function is independent of the vaporization kinetics.The signal to partial digestion of the slurry occurs. baseline noise ratio was measured for the refractory elements Cr, V and Ti. It was concluded that low observation height Use of Standard Slurries and a reduced plasma power are most suitable when an argon furnace carrier is employed. Standard slurries have been used as an alternative to aqueous calibration in some analyses of solid samples.16,47,64 Standard Za�ray et al.42 identified, using three dimensional plots, the optimum plasma operating conditions with respect to rf power reference materials are normally used as, for example, in the analysis of airborne particulatematter by ICP-AES.16 Standard (0.8 kW) and aerosol gas pressure (2 bar) for the determination of Al, Ca, Fe and Mg in silicon nitride powder.At the optimum suspensions were prepared by homogenizing a known amount of the reference material. Fernandez et al.47 used a reference conditions for the elemental analyses, the atomization efficiencies ranged from 77 to 100%.slag in the determination of a range of elements. A range of geological standard reference materials has also been used.64 Standards were made from materials including SiO2 rich granites, marine sediments and coals. USE OF MIXED GAS PLASMAS The use of standard slurries for calibration is problematical.56 The nature of the material being used for calibration must be The thermodynamic properties of the ICP itself may be altered by modifying one or all three of the gas flows used.A mixed identical with that of the unknown slurry sample. For heterogeneous standard slurries (e.g., coal), elements will not be gas is defined as an Ar plasma where a molecular gas (e.g., N2 or O2 ) or a noble gas (e.g., He) has been added. It is thought distributed uniformly and the particle size distributions and densities of elements within the sample may not be comparable that when the gas mixture induces a thermal ‘pinch’, causing the plasma volume to decrease and energy to increase within to those for the slurry being analysed.the plume, slurry atomization may be enhanced. The excitation temperature has been measured to be around 5500 K for a OPTIMIZATION 1.5 kW Ar plasma, increasing to 6800 K upon addition of 5% O2.27 Optimization of operating variables is increasingly being used to enhance the analysis of slurries using plasmas. Some slurries Refractory materials may be prepared as slurries of the desired particle size but the recoveries may still be low as these may present difficulties in volatilization and atomization and the optimization of plasma conditions may overcome or materials are notoriously difficult to vaporize and atomize in the ICP.The use of a mixed gas plasma may be beneficial in improve upon these problems. Sparkes and Ebdon89 used simplex optimization to improve these cases. Verbeek and Brenner32 investigated the performance of Ar, Ar–O2 (5% O2) and Ar–N2 (10% N2) plasmas for the determination of Mg in a kaolin slurry by DCP-AES.The SBR, corrected for slurry concentration, was used as the figure the analysis of geological samples by ICP-AES. Homogeneous samples were used so that identical chemical composition of merit. Seven parameters were optimized: horizontal and vertical viewing positions, nebulizer gas flow rate, plasma gas could be ensured and compositional differentiation eliminated.Line profiles were related to the thermochemical properties of flow rate, concentration of ammonia dispersant, concentration of lithium buffer and the concentration of the slurry. A the oxides or nitrides. The normalized SBR was calculated for Ti, Cu, Eu, Si and Mn using the three plasmas as a function univariate search was carried out for all seven parameters after optimization to illustrate the role of each of the parameters. of viewing height. It was found that the maximum SBR for all 224 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12the elements under investigation occurred lower in the mixed 0.1 M NaCl to the matrix would affect the elemental recoveries. It was concluded that the addition of 0.1 M NaCl as a matrix gas plasmas than in the Ar only plasma. Identical mixed gas plasmas were used to enhance sensitivity solution and restricting the particle size of the slurry to less than 2 mm reduced matrix interferences and aided the dis- in the analysis of refractory materials and coals, by eliminating particle size and sample composition effects.39 The analytical sociation and excitation of clay particles within the plasma resulting in improved Al recoveries. sensitivity of the Ar–O2 plasma was different to that of the Ar plasma for the determination of Si and Fe.For Ti and Al, the Jerrow et al.90 analysed six soils and sediments by slurry nebulization DCP-AES for seven elements. Approximately calibration curves were not statistically different for the three plasmas. However, it was concluded that the accuracy and 0.5 g of the powdered samples was ground using the bottle and bead method and diluted to 100 ml using 0.1% Triton precision of the elemental analysis were improved on addition of O2 and N2.X-100. The recoveries for Ba, Sr, Ca and Al were found to be poor for three of the samples. It was concluded that lithium Xhoffer et al.53 investigated the effect of the addition of N2 and O2 to the plasma gas flow on exhaust particles during the or potassium buffers, added to the matrix to reduce EIE effects, adversely affected the recoveries owing to cation exchange of analysis of silicon carbide by ICP-AES.Both N2 and O2 were found to have no effect on the shape and morphology of sub- the analyte element ion from the slurry particles into the solution phase. micron metre aerosols. The use of O2 appeared to affect the chemical composition of the exhaust SiC particles whereas N2 Sparkes and Ebdon89 found that for slurries, containing lithium at a concentration of 5 g l-1, analysed by DCP-AES, did not.Ebdon and Goodall51 employed a hydrogen modified ICP an increase in slurry concentration resulted in a decrease in the excitation temperature and electron number density. This for the AES analysis of refractory particles. An improvement in accuracy, corresponding to the elimination of atomization was not observed if lithium was absent and the effect was attributed to an increase in the rate of radiative cooling of the interferences, was the resultant effect.An increase in the plasma rotational temperature from 2200 to 3900 K was measured. plasma. A partial limitation to the use of the DCP for slurry nebulization was suggested as the addition of 5 g l-1 lithium This was attributed to an increase in energy transfer from the toroidal to the annular region of the ICP, due to the higher as a buffer limits thlinear working range of slurry concentrations to approximately 10% m/v for a refractory kaolin thermal conductivity of hydrogen.The analytical sensitivity and detection limits were not affected on the addition of sample. High sample loadings therefore cannot be tolerated. The effect of high salt matrices on analyte response for the hydrogen as it was concluded that transportproblems predominated over atomization difficulties. analysis of milk powder by ICP-MS has been investigated.75 Sodium or calcium at 500 mg ml-1 was added to a 10 ng ml-1 A desolvation system was used by Hartley et al.82 prior to introduction of the slurry aerosol into the ICP to increase Pb solution and isotope ratio measurements were performed.The precision in the presence of Na or Ca was lower than that transport efficiency and improve recoveries. This decreased the amount of hydrogen in the plasma and therefore hydrogen obtained for a pure lead solution, and this was attributed to transport effects either in the nebulizer or at the ICP-MS (1.5% v/v) was added to the Ar plasma to raise the ionization temperature from 5500 to 8400 °C for the analysis of slurries interface. Accuracy was not affected.by ICP-MS.82 Thermal conductivity and energy transfer processes were improved, as were the recoveries for the slurries. Ebdon and Goodall54 examined the effect of adding hexa- FUNDAMENTAL STUDIES fluoroethane (Freon 116) to an Ar ICP.The gas is commonly Aerosol Formation, Transportation and Loss used for ETV techniques. Unusual plasma spectrochemistry was observed together with non-linear calibration curves. This Transportation of a slurry sample into a plasma is facilitated by the formation of an aerosol by the nebulizer and by the was attributed to the formation of molecular fluoride species and clearly dictated against the use of Freon 116 as a potential filtering action of the spray chamber.This physical process is crucial to analytical accuracy and precision. If aqueous cali- vaporization aid. In contrast, Za�ray et al.60 found no degradation in linearity when using Freon-12 (CCl2F2) to bration standards are used then the transportation of slurries through the sample introduction system must mimic that of assist in the atomization of silicon nitride slurries. They found that the addition of 10 ml of Freon-12 reduced the differences solutions. Fundamental investigations have examined the behaviour of slurries in terms of aerosol formation, transpor- between slurry and solution signals.tation of the aerosol through the sample introduction system and any loss processes that occur. MATRIX EFFECTS Ebdon et al.34 have made extensive comparative studies of solution and slurry behaviour in the ICP. Laser particle sizing, Little work has been reported concerning liquid component matrix effects and associated interferences when slurry samples a non-intrusive technique, was employed to measure the slurry aerosol particle size distribution, between 1 and 40 mm, at are analysed in plasmas.Laird et al.40 measured the effect of matrix solution composition on matrix interferences for clay various stages in the sample introduction system. It was found that the majority of particles emitted from an Ebdon V-groove minerals using ICP-AES. Four size fractions (<0.2, 0.2–2, 2–10 and 10–45 mm) were used to prepare the slurries.Portions of nebulizer were below 20 mm in diameter and an increased nebulizer gas flow rate resulted in the formation of finer the clay samples (0.05 g) were sonicated for 30 s in 100 ml of the matrix solution. They believed that matrix inferences are particles. The particle size distribution was found to be modi- fied by both double and single pass spray chambers indicating problematic owing to the different forms of the aqueous standards and sample suspensions. Slurries were prepared in that ballistic, turbulent and recirculatory deposition processes occurred.The aerosol emitted from a 2 mm injector showed a HNO3 which acted partially to dissolve some of the clay particles in suspension. Cations were displaced from the clay reduction in the particle size distribution, i.e., smaller particles passed through whereas larger particles did not. A 3 mm framework by hydronium ions in the matrix solutions. The elements displaced (and therefore in the matrix solution) were injector allowed more of the larger particles to enter the plasma.Aerosols of 1% m/v slurries did not behave differently transported to the plasma more readily and the recoveries were improved. The authors were able to rationalize in this from those of equivalent solutions under identical conditions and it was concluded that aerosol formation, transportation way the differential elemental recoveries obtained. An experiment was also performed to ascertain whether the addition of and loss processes are the same for both systems.Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 225Table 4 Applications of slurry nebulization into plasmas Sample Elements Plasma Dispersant determined technique used Comments Ref. Fe, Si, Al, Ti, Ca, Zn, Airborne particulate ICP-AES 100% xylene Particulates collected on polystyrene fibre 17 matter Pb, Cu, Mn filters Animal tissue Cu, Zn, Ca, P DCP-AES — Babington slurry nebulizer used 85 Biological materials Mg, Ca, K, Na, Fe, Mn, ICP-AES H2SO4 Carbonaceous slurries prepared 41 Cu, Zn Biological samples Mg, Al, Mn, Fe, Co, ICP-MS 1% v/v Triton Detection limits range from 0.0001 (U) to 80 Cu, Zn, As, Se, Br, X-100 0.52 mg g-1 (Zn) Rb, Sr, Mo, Ag, Cd, Sb, Pb, U Botanical samples B, Cr, Ti, Mo ETV–ICP-AES — Aqueous calibration used 95 Carbon black S ICP-AES 0.5% v/v Triton Good accuracy and precision observed 67 X-100 Ceramic Powder Fe, Mg ICP-AES — Aqueous standards without matrix 36 matching used for calibration Al, Ca, Si, Fe ICP-AES Various Effect of dispersant on slurry stability 61 dispersants discussed investigated Al, B, Na, Mg, Ca, Ti, ICP-AES and ICP- — Good agreement between two plasma 74 V, Cr, Mg, Mn, Fe, MS techniques and with aqueous samples Co, Ni, Cu, Zn, Ga, Zr, Ba, La, Ce Citrus and tomato Mn, Mg, Ca ICP-AES — RSDs between 0.5 and 2.0%.Matrix 44 leaves matching avoided Clays Sc, Si, Ca, Fe, Al, Mn, ICP-AES — Grinding time and particle size controlled 25 Ti, Mg reliability of analytical calibration functions Si, Al, Ca, Mg, Fe ICP-AES 0.1 M NaCl Particle size effects and matrix solution 40 composition on matrix interferences determined Al, Si, Ca, Fe, Mg, K ICP-AES 1 M HNO3+0.1 Aqueous calibration used 48 M NaCl Coal S ICP-AES 0.5% Triton Reduced nebulizer flow rate used with 22 X-100 spray chamber geometry modification Mn ICP-AES 0.5% Triton Mixed gas plasmas and simplex 15 X-100 optimization used Mn ICP-AES 1% Triton X-100 Optimum conditions established using 23 simplex optimization Cu, Fe, Mn, Ni ICP-AES 1% Triton X-100 Results compared with those obtained by 24 ashing and digestion S, Al, Ca, Fe, Mg, Si, ICP-AES 40% glycerol Various mixed gas plasmas used 39 Ti, Mn and 2% Kodak photoflow Al, Fe, Ca, Ti, Cr, Mn, ICP-AES 0.1% Aerosol Good agreement with certified values 69 Cu, Sr, V, Zn, Co, Ni OT obtained using aqueous calibration Al, Ti, Fe, Be, Mg, Cr, ICP-MS Mn, Co, Ni, Cu, Ge, Rb, Ba, Ce, Sm, U 64 elements by ICP-MS 0.1% Aerosol Rhodium internal standard used 77 semiquantitative OT analysis. 16 elements by fully quantitative determination Cr, Cu, Mg, Mn, Ni, Pb DCP-AES 0.5% Triton Near unity response factors observed 86 X-100 Cu, Cr, Ni, Pb, Mn ETV–ICP-MS 0.005% Triton Standard additions used 86 X-100 V, Mn, Ni, Cu, Pb ETV–ICP-MS 0.005% Triton 10 mg Pd added as a physical carrier with 97 X-100 O2 ashing Firebrick Ti, Fe, Al, Ca, Mg ICP-AES 0.05% Na4P2O7 Hydrogen modified ICP used 51 Ti, Fe, Al, Mg ICP-AES 0.05% Na4P2O7 Freon 116 modified Ar ICP used 54 Al, Ca, Fe, Mg, Ti ICP-AES 0.5% Na4P2O7 Influence of particle size on analytical 56 accuracy discussed Foods Mo ETV-ICP-AES — Samples shaken for 15–20 min before 93 analysis Geological and Ba, Ce, Co, Cr, Cu, Ni, ICP-AES — Yttrium added as an internal standard 64 related non- Pb, Sr, V, Zn, Si, F conducting Al, Ca, Mg, Ti, Mn materials Geological materials Si, Al, Fe, Ca, Ti, Na, ICP-AES — Dispersants not added.Found that 55 P, S, Ba, Be, Ce, Co, slurries <1.5 mm do not behave as Cr, Cu, Eu, La, Mn, solutions Ni, Sr, V, Y, Yb, Zn Be, B, Cr, Ge, As, Se, ICP-MS 0.05% Na4P2O7 13 reference materials analysed. Precision 78 Nb, Mo, Ag, Cd, Sn, 5–10% Sb, Te, Ta, W, Bi, Th, U Ru, Rh, Pd, Os, Ir, Pt, ICP-MS 0.05% Na4P2O7 RSDs 2–30% for most elements. Good 81 Au agreement with certified values Industrial catalysts Al, Si, Mg, Cu, Mn, Fe, ICP-MS and ICP- 0.05% Na4P2O7 Solid deposition on sampling cone of 76 V, Cr, Co, Ni AES ICP-MS gave low recoveries Kaolin Mg, Fe ICP-AES 0.35% m/v Effect of particle size, dispersants and 28 ammonia viscosity discussed solution 226 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Table 4 (continued) Sample Elements Plasma Dispersant determined technique used Comments Ref. Fe, Mg, Ti, Si ICP-AES 0.35% m/v Various nebulizers, spray chambers and 30 ammonia injector tubes employed solution Mg DCP-AES Aqueous Simplex optimization used 89 ammonia solution Lobster Co, Cr, Ni, Pb ETV–ICP-MS 0.005% Triton Standard additions used 96, 97 hepatopancreas X-100 Cd, Cu, Fe, Zn MIP-AES 10% HNO3 Good agreement with certified results 92 Milk powder Pb ICP-MS 0.1% Triton Lead isotope ratios measured for isotope 75 X-100 dilution analysis P, Ca, Na, K, Mg, Cu, DCP-AES 0.01% Triton Lithium enhancement buffer used 87 Fe, Zn, Mn X-100 Ores and flotation Au, Pt, Pd, Ru, Rh ICP-AES — Elements absorbed on ion exchange resin 18 feeds then made into slurry Oyster tissue V, Mn, Ni, Cu, Pb ETV–ICP-MS 0.005% Triton Oxygen ashing employed.Pd added as a 97 X-100 physical carrier Pine needles Al, Cr, Mn, Fe, Co, Ni, ICP-MS 1% Triton X-100 Poor accuracy obtained owing to fibrous 80 Cu, As, Br, Rb, Sr, Cd, nature of material Sb, La, Ce, Eu, Pb, Th,U Plant materials Ca, Mg, Mn, P ICP-AES 1 g l-1 Aerosol Sample ashed before grinding 56 OT Cu, Mn, Pb, Zn DCP-AES 0.01% Triton Carbonization procedure used for fibrous 87 X-100 material Refractory oxide Al, Cu, Fe, Mg, Nb, Si, ICP-AES 0.01% SHMP Slurry technique compared with direct 29 powders Zr and mono- insertion and electroerosion techniques iso- propanolamine Refractory samples Cu, Zn, Pb, Sn, Ag, Fe, ICP-AES 0.1% SHMP Fundamental studies carried out 38 Mo, As, Zr Rice flour Mn, Fe, Co, Ni, Cu, Zn, ICP-MS 1% Triton X-100 Acceptable accuracy and precision 80 As, Se, Br, Rb, Mo, Cd, obtained Pb Rocks Si, Ca, Mg, Fe, Al, Ti, ICP-AES — Grinding time and particle size effects 25 Mn investigated Al, Fe, Cr, Mn, Ba, Cd, ICP-AES 0.5% Na4P2O7 Good accuracy for all elements measured 63 Cu, Zn, Sr, Pb Al, K, Na, Si, Ba, Mn, ICP-AES and ICP- 0.1% Na4P2O7 Poor recoveries owing to large particle size 70 Sr, Ti, Ca, Mg, Cu MS of slurry Slags Si, Ca, Mg, Al, Fe, Mn, ICP-AES 0.35% ammonia Calibration performed using reference slag 47 Ti, Na, K Sulfide ores Cu, Zn, Pb, Sn, Ag, Fe, ICP-AES 0.1% SHMP Various instrument operating parameters 27 Mo, As, Zr used Soils Ca, Fe, Mg, Mn, Cu, V ICP-AES 1% Na4P2O7 Flow injection used 33 Si, Al, Fe, Mg, Mn, Cr, ICP-AES and ICP- 0.05% Na4P2O7 Solid deposition on sampling cone of 76 V, Ni, Cu, Co, Zn MS ICP-MS gave low recoveries Na, Ba, Sr, Ca, Mg, Al, DCP-AES 0.1% Triton EIE effects observed 90 Ti, Si X-100 Sediments Al, Fe, Mg, Co, Cr, Cu, ICP-AES 0.01% Triton Heated sample introduction system used 45 Mn, Ni, Pb, V, Zn X-100 Cu, Mn, Zn ICP-AES 1 M HNO3 Tandem source used 52 Al, As, Zn, Cr, Cu, Mn, ICP-AES 0.5% Na4P2O7 Good accuracy for all elements measured 63 Pb, V V, Cr, Mn, Co, Ni, Cu, ICP-MS Triton X-100 Desolvation device used to improve trans- 82 Zn, Sr, Ba, La, Pb port efficiency Na, Ba, Sr, Ca, Mg, Al, DCP-AES 0.1% Triton EIE effects observed 90 Ti, Si X-100 Cu, Zn MIP-AES 10% HNO3 Good agreement with certified values 92 Sewage sludge Al, As, Zn, Cr, Cu, Mn, ICP-AES 0.5% Aerosol OT Samples charred before grinding.Poor 63 Pb, V and 1% Triton accuracy for some elements X-100 Silicate rocks La, Ce, Pr, Nd, Sm, Eu, ICP-MS 0.1% Triton RSD 0.8–6.3% for REE contents of 79 Gd, Tb, Dy, Ho, Er, Tm, X-100 0.13–38 mg ml-1 Yb, Lu Silicon carbide Fe, Al, B, Cu, Ni, V, Na, ICP-AES 0.1% Triton RSD 0.8–2.1%. Internal standard used 57 Cr, Ca, Co, Ti X-100 Silicon nitride C, Al, Ca, Fe, Mg ICP-AES — Low power Ar ICP used 35, 42 Fe ICP-AES 0.5% Na4P2O7 Concluded that particle size should be 43 <10 mm for good recoveries Al, Fe, Ca, Mg, Ti ICP-AES 1% HCl Freon 12 halogenation agent added 60 Tea leaves Al, Ba, Mg, Mn ICP-AES 0.05% Triton RSD 0.3–1.8% 59 X-100 Total diet Na, K, Ca, Mg, Fe, Zn, ICP-AES 0.5% Triton Results for all elements within certified 65 Mn, Cu X-100 range Cu, Cr, Ni, Pb ETV–ICP-MS 0.005% Triton Standard additions used 96 X-100 V, Mn, Ni, Cu, Pb ETV–ICP-MS 0.005% Triton Oxygen ashing employed. 97 X-100 Whole plant Cu DCP-AES 0.5% Triton Alumina added to aid grinding 88 X-100 Zeolite Zn, Pb, Cd, Mn, V, Cu, ETV-ICP-AES 1% HNO3 Freon assisted graphite furnace 94 Ti, Ca vaporization Zirconium oxide Al, B, Ca, Cu, Fe, Mg, ICP-AES — Detection limits of 0.03–10 mg g-1 50 Mn, Na, Ti, V, Y obtained. Results compare well with those for equivalent solutions Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 227Aerosol droplet size distributions of Al2O3 slurries (1%m/v) into plasmas has increased significantly and it is now possible routinely to determine the elemental composition of finely were measured using a cascade impactor and collected above the injector.31 It was found that a Babington nebulizer pro- powdered and dispersed slurries of many sample types using a high solids nebulizer and simple aqueous calibration.The duced slurry aerosols equivalent to solution aerosols. It was also found that particles larger than 17 mm did not reach the challenge for slurry nebulization remains in the area of rapid particle size reduction without contamination.ICP so samples need to be homogeneous. An ‘evaporation model’ was developed to predict the analytical performance of slurry nebulization into an ICP. The authors used the model to describe differences between solution and slurry evaporation REFERENCES and thus to predict when aqueous calibration may be used. 1 Hoare, H. C., and Mostyn, R.A., Anal. Chem., 1967, 39, 1153. Droplet size distributions have also been measured by Van 2 Dagnall, R. M., Smith, D. J., West, T. S., and Greenfield, S., Borm et al.49 along with analyte transport efficiencies and Anal. Chim. Acta, 1971, 54, 397. analyte mass transport rates. They stated that differences 3 De Silva, K. N., and Guevremont, R., Spectrochim. Acta, Part between solutions and slurries in terms of these three processes B, 1990, 45, 997. 4 Salin, E. D., and Horlick, G., Anal.Chem., 1979, 51, 2284. are dependent on the nebulizer gas flow, the characteristics of 5 Sommer, D., and Ohls, K., Fresenius Z. Anal. Chem., 1980, the instrument and, most importantly, the type and state of 304, 97. the slurry. 6 Zhang, L.-X., Kirkbright, G. F., Cope, M. J., and Walton, J. M., Xhoffer et al.53 measured exhaust aerosols from Al2O3, ZrO2 Appl. Spectrosc., 1983, 37, 250. and SiC ceramic suspensions using electron energy loss spec- 7 Scott, R. H., Spectrochim.Acta, Part B, 1978, 33, 123. troscopy. The size, shape and surface characteristics of the 8 Keilson, J. P., Deutsch, R. D., and Hieftje, G. M., Appl. Spectrosc., 1983, 37, 451. slurry particle changed during the time spent in the sample 9 Aziz, A., Broekaert, J. A. C., Laqua, K and Leis, F., Spectrochim. introduction system and were found to be particularly depen- Acta, Part B, 1984, 39, 1091. dent on the type of plasma gas used, the type of slurry sample 10 Laqua, K., in Analytical L aser Spectroscopy, ed.Omenetto, N., and the distance from the impactor part of the spray chamber Wiley, New York, 1979, ch. 2, p. 47. to the torch. 11 Thompson, M., Goulter, J. E., and Sieper, F., Analyst, 1981, The fundamental parameters of aerosol particle size were 106, 32. 12 Denoyer, E. R., Fredeen, K. J., and Hager, J. W., Anal. Chem., discussed by Goodall et al.56 For efficient introduction into 1991, 63, 445A. the plasma, the slurry particle size is known to be of major 13 Fuller, C.W., Hutton, R. C., and Preston, B., Analyst, 1981, importance. They stated that the particle size distribution of 106, 913. the slurry should not exceed a value determined by the density 14 Saba, C. S., Rhine, W. E., and Eisentraut, K. J., Anal Chem., of the particle. A size occupancy diameter model was intro- 1981, 53, 1099. duced which illustrates that for slurry nebulization to be 15 Wilkinson, J. R., Ebdon, L., and Jackson, K. W., Anal.Proc., 1982, 19, 305. comparable to that of solutions, one slurry particle should be 16 Sugimae, A., and Mizoguchi, T., Anal. Chim. Acta, 1982, 144, 205. contained within a single aerosol droplet. 17 Spiers, G. A., Dudas, M. J., and Hodgins, L. W., Clays Clay Miner., 1983, 31, 397. 18 Watson, A. E., and Moore, G. L., S. Af r. J. Chem., 1984, 37, 81. Temperature Measurements 19 Algeo, J. D., Heine, D. R., Phillips, H. A., Hoek, F. B. G., Plasma processes may be investigated by using the excitation Schneider, M.R., Freelin, J. M., and Denton, M. B., Spectrochim. Acta, Part B, 1985, 40, 1447. temperature (Texc), the ionization temperature (Tion) and the 20 Page, J. R., and Jewel, K. E., Soil Sci., 1985, 139, 211. electron number density (ne) as diagnostic tools. Sparkes and 21 Mackey, J. R., and Murphy, W. J., Zeolites, 1985, 5, 233. Ebdon89 used these parameters to explain slurry related phen- 22 McCurdy, D. L., and Fry, R. C., Anal. Chem., 1986, 58, 3126.omena in the DCP concerning matrix effects caused by the 23 Ebdon, L., and Wilkinson, J. R., J. Anal. At. Spectrom., 1987, 2, 39. addition of easily ionized elements to the slurry matrix. Ebdon 24 Ebdon, L., and Wilkinson, J. R., J. Anal. At. Spectrom., 1987, and Foulkes71 measured the rotational (gas kinetic) tempera- 2, 325. 25 Halicz, L., and Brenner, I. B., Spectrochim. Acta, Part B, 1987, ture of the ICP and employed temperature mapping to 42, 207. investigate the processes occurring when a slurry particle enters 26 Brenner, I.B., Lang, Y., Le Marchand, A., and Grosdaillon, P., a plasma. When solutions and slurries of up to 1% m/v were Am. L ab., 1987, 19, 17. introduced into the ICP, no significant differences in the 27 Foulkes, M. E., Ebdon, L., and Hill, S., Anal. Proc., 1988, 25, 92. rotational temperature were observed. The presence of water 28 Ebdon, L., and Collier, A. R., J. Anal. At. Spectrom., 1988, 3, 557. from the aerosol appeared to raise the temperature, in contrast 29 Broekaert, J.A. C., Leis, F., Raeymaekers, B., and Za�ray, G., Spectrochim. Acta, Part B, 1988, 43, 339. to humidified argon. When slurries of fine particles of com- 30 Ebdon, L., and Collier, A. R., Spectrochim. Acta, Part B, 1988, pounds, which have melting points comparable to and boiling 43, 355. points in excess of the rotational temperature, were nebulized, 31 Raeymaekers, B., Graule, T., Broekaert, J. A. C., Adams, F., and excellent recoveries were observed.This confirmed that Tscho�pel, P., Spectrochim. Acta, Part B, 1988, 43, 923. transport and atomization efficiencies for solution and slurry 32 Verbeek, A and Brenner, I. B., J. Anal. At. Spectrom., 1989, 4, 23. introduction into the ICP are comparable. 33 Ambrose, A. J., Ebdon, L., Foulkes, M. E., and Jones, P., J. Anal. At. Spectrom., 1989, 4, 219. 34 Ebdon, L., Foulkes, M. E., and Hill,S., Microchem. J., 1989, 40, 30. APPLICATIONS OF SLURRY SAMPLE 35 Varga, I., Za�ray, G., Sze`pvolgyi, J., and Ko`nya, G., Mikrochim. INTRODUCTION INTO PLASMAS Acta, 1989, 3, 381. 36 Huang, M., and Shen, X.-E., Spectrochim. Acta, Part B, 1989, Some applications of slurry sample introduction into plasmas 44, 957. are given in Table 4. These applications are listed alphabetically 37 Van Borm, W. A., and Broekaert, J. A. C., Anal. Chem., 1990, 62, 2527. for sample type, and the elements determined, the analytical 38 Ebdon, L., Foulkes, M.E., and Hill, S., J. Anal. At. Spectrom., technique used and dispersant employed to stabilize the slurries 1990, 5, 67. are given. A wide range of applications is clearly evident, 39 Long, G. L., and Brenner, I. B., J. Anal. At. Spectrom., 1990, 5, 495. showing that slurry nebulization is applicable widely through- 40 Laird, D. A., Dowdy, R. H., and Munter, R. C., J. Anal. At. out environmental, geological, biological, industrial and food Spectrom., 1990, 5, 515. analysis. 41 Fagioli, F., Landi, L., Lacatelli, G., Righini, F., Settimo, R., and Magarini R., J. Anal. At. Spectrom., 1990, 5, 519. In the last 15 years, the understanding of slurry nebulization 228 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 1242 Za�ray, G., Farkas, A., and Varga, I., Acta Chim. Hung., 1991, 73 Jarvis, K. E., Chem. Geol., 1992, 95, 73. 74 Broekaert, J. A. C., Lathen, C., Brandt, R., Pilger, C., Pollman, D., 128, 489. Tscho�pel, P., and To� lg, G., Fresenius’ J.Anal. Chem., 1994, 43 Isozaki, A., Ogawa, M., Shibagaki, M., and Morita, Y., Anal. 349, 20. Sci., 1991, 71, 1249. 75 Dean, J. R., Ebdon, L., and Massey, R., J. Anal. At. Spectrom., 44 Carrio�n, N., Ferna�ndez, A., Eljuri, E. J., Murillo, M., and 1987, 2, 369. Franceschetto, M., At. Spectrosc., 1991, 12, 162. 76 Williams, J. G., Gray, A. L., Norman, P., and Ebdon, L., J. Anal. 45 Gervais, L. S., and Salin, E. D., J. Anal. At. Spectrom., 1991, 6, 41. At. Spectrom., 1987, 2, 469. 46 Rademeyer, C. J., Collins, C. S., and Butler, L. R. P., J. Anal. 77 Ebdon, L., Foulkes, M. E., Parry, H. G. M., and Tye, C. T., At. Spectrom., 1991, 6, 329. J. Anal. At. Spectrom., 1988, 3, 753. 47 Fernandez, M. L., Fairman, B., and Sanz-Medel, A., J. Anal. At. 78 Jarvis, K. E., and Williams, J. G., Chem. Geol., 1989, 77, 53. Spectrom., 1991, 6, 397. 79 Mochizuki, T., Sakashita, A., Iwata, H., Ishibashi, Y., and 48 Laird, D. A., Dowdy, R. H., and Munter, R. C., Soil Sci. Soc. Gunji, N., Anal. Sci., 1989, 5, 311. Am. J., 1991, 55, 274. 80 Mochizuki, T., Sakashita, A., Iwata, H., Ishibashi, Y., and 49 Van Borm, W. A. H., Broekaert, J. A. C., Klockenka�mper, R., Gunji, N., Fresenius’ J. Anal. Chem., 1991, 339, 889. Tscho�pel, P., and Adams, F. C., Spectrochim. Acta, Part B, 1991, 81 Totland, M., Jarvis, I., and Jarvis, K. E., Chem. Geol., 1993, 46, 1033. 104, 175. 50 Lobin�ski, R., Van Borm, W., Broekaert, J. A. C., Tscho�pel, P., 82 Hartley, J. H. D., Hill, S. J., and Ebdon, L., Spectrochim. Acta, and To� lg, G., Fresenius J. Anal. Chem., 1992, 342, 563. Part B, 1993, 48, 1421. 51 Ebdon, L., and Goodall, P., J. Anal. At. Spectrom., 1992, 7, 1111. 83 Pollman, D., Pilger, C., Hergenro�der, R., Leis, F., Tscho�pel, P., 52 Karanassios, V, Usypchuck, L., Moss, P., and Salin, E. D., and Broekaert, J. A. C., Spectrochim. Acta, Part B, 1994, 49, 683. J. Anal. At. Spectrom., 1992, 7, 1243. 84 Balaram, V., Curr. Sci., 1995, 69, 640. 53 Xhoffer, C., Lathen, C., Van Borm, W., Broekaert, J. A. C., 85 Mohamed, N., Brown, R. M., and Fry, R. C., Appl. Spectrosc., Jacob, W., and Van Grieken, R., Spectrochim. Acta, Part B, 1981, 35, 153. 1992, 47, 155. 86 McCurdy, D. L., Wichman, M. D., and Fry, R. C., Appl. 54 Ebdon, L., and Goodall, P., Spectrochim. Acta, Part B, 1992, Spectrosc., 1985, 39, 984. 47, 1247. 87 Sparkes, S., and Ebdon, L., Anal. Proc., 1986, 23, 410. 55 Halicz, L., Brenner, I. B., and Yoffe, O., J. Anal. At. Spectrom., 88 Vien, S. H., and Fry, R. C., Appl. Spectrosc., 1988, 42, 381. 1993, 8, 475. 89 Sparkes, S. T., and Ebdon, L., J. Anal. At. Spectrom., 1988, 3, 563. 56 Goodall, P., Foulkes, M. E., and Ebdon, L., Spectrochim. Acta, 90 Jerrow, M., Marr, I., and Cresser, M., Anal. Proc., 1992, 29, 45. Part B, 1993, 48, 1563. 91 Jerrow, M., Marr, I., and Cresser, M., J. Anal. At. Spectrom., 57 Goulter, J., T IZ Int. Powder Bulk Mag., 1992, 116, 41. 1992, 7, 1117. 58 Chen, C., and McCreary, T. W., Appl. Spectrosc., 1994, 48, 410. 92 Matusiewicz, H., and Sturgeon, R. E., Spectrochim. Acta, Part 59 Manickum, C. K., and Verbeek, A. A., J. Anal. At. Spectrom., B, 1993, 48, 723. 1994, 9, 227. 93 Hu, B., Jiang, Z., and Zeng, Y., J. Anal. At. Spectrom., 1991, 6, 623. 60 Za

ISSN:0267-9477    

DOI:10.1039/a604914a

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Vanadium, Rhodium and Platinum in Automotive Catalytic Converters Using Inductively Coupled Plasma Mass Spectrometry With Spark Ablation OLEG V. BORISOVa, DAVID M. COLEMAN*a AND R. O. CARTER IIIb aDepartment of Chemistry, Wayne State University, Detroit, MI 48202, USA bScientific Research L aboratory, Ford Motor Company, PO Box 2053, Dearborn, MI 48121, USA A high voltage unidirectional spark source was used as a Automotive catalytic converter samples are refractory in nature.Problems arising from sample preparation by micro- sampling tool for ICP-MS analysis of automotive catalytic converters. Electrical conductivity of the samples, required for wave acid digestion have previously been reported.13 It was determined that different polyatomic species can contribute to spark experiments, was provided by mixing samples with powdered graphite. Spark discharge parameters, such as peak measurement uncertainty and that close matrix matching is needed.This is not always possible due to variations in loading current and repetition rate, which influence spark sampling efficiency, were studied and optimized. Prior work involving of potentially interfering elements from sample to sample. This is particularly essential when low and trace concentrations liquid nebulization ICP-MS analysis of series of ‘palladiumonly wavelength dispersive X-ray fluorescence secondary of analyte elements are to be determined. In this case the magnitude of signals arising from interfering species can be standards’ was compromised owing to spectral interferences arising primarily from polyatomic oxide species.Additional comparable with that of the analyte. In the present work spark ablation, using a unidirectional separation techniques were required in order to separate analyte from matrix elements in cases where interferences high-voltage spark discharge, was used. The main goal of the study was to evaluate the feasibility of spark sampling as a could not be resolved.Spark ablation, with its dry plasma, was shown to result in a significant reduction in the formation of tool for direct sample introduction into an ICP-MS instrument. A simultaneous objective was to determine Pt, Rh and V in polyatomic species and simplification of the analytical procedure. Analytical results achieved by ICP-MS with liquid catalytic converters. Results for solid sampling and liquid sample introduction for ICP-MS are compared.nebulization and with solid sampling are compared. Limitations of the spark ablation technique owing to high voltage rf interference on the ICP-MS instrument and spark EXPERIMENTAL discharge wander around the sample surface are discussed. Instrumentation Keywords: Inductively coupled plasma mass spectrometry; precious metals; automotive catalytic converter; spark ablation The spark source is a research-grade, high-voltage, thyratrontriggered, adjustable-waveform source that has been described elsewhere.14 A typical spark current waveform is shown in Inductively coupled plasma mass spectrometry is an increas- Fig. 1. The ICP-MS instrument used in this study was an Elan ingly popular technique for elemental analysis. Most conven- 5000 (Perkin-Elmer SCIEX, Thornhill, Ontario, Canada). The tional ICP-MS methods require the sample to be in solution operating conditions used are summarized in Table 1. Carrier form. Sample dissolution steps are probably the most time argon flow was adjusted to achieve a stable spark discharge.consuming and are the major source of errors in the entire This was monitored visually and by oscilloscope (current analysis scheme. Dissolution procedures suffer from such prob- waveform). lems as solvent contamination, conflicting requirements for dissolution of various elements and analyte dilution. Moreover, liquid sample introduction into the ICP-MS instrument increases the risk of various interferences, including matrixand digestion acid-induced interferences primarily due to the formation of polyatomic oxide species and doubly charged ions.These common interferences have been reviewed by Evans and Giglio1 for ICP-MS analysis. Nevertheless they are often difficult to overcome. In contrast, direct solid sampling avoids sample dissolution steps, hence resulting in significant decreases in the formation of polyatomic oxide ions due to decreased water vapor loading and the absence of dissolution acids.Among the most popular direct solid sampling methods are laser, spark and arc ablation techniques, which have been used in conjunction with ICPAES and ICP-MS systems for the analysis of conductive and non-conductive materials.2–12 Laser and spark ablation are both efficient at producing a fine aerosol from the surface of interest. This aerosol is carried by an argon gas flow through Fig. 1 Typical unidirectional high voltage spark current waveform. a gas transport system to the mixing chamberof an ICP system.Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (231–237) 231Table 1 Instrumentation and operating conditions Reagents and Sample Preparation L iquid nebulization ICP-MS system, Elan 5000— Rf power/W 1000 The concentrated nitric, hydrochloric and hydrofluoric acids Outer plasma gas flow rate/l min-1 14.0 used were J. T. Baker Ultrex II Ultrapure (Phillipsburg, NJ, Intermediate flow rate/l min-1 0.900 Aerosol carrier flow rate/l min-1 0.987 USA).Crystalline boric acid was J. T. Baker Ultrex (99.98% Ion lens settings (relative voltages) pure). Calibration standards were diluted from single-element B 42 1000 mg l-1 stock solutions (Spex Plasma Standards, Spex E1 15 Industries, Edison, NJ, USA). All dilutions were carried out P 46 with de-ionized water (resistivity 18MV cm). An In standard S2 45 solution was added to all samples, standards and blanks as an Sampler cone Nickel, 1.14 mm orifice Skimmer cone Nickel, 0.89 mm orifice internal standard.The dissolution procedure (termed Analytical Procedure 313) included microwave digestion of the Spark source, unidirectional high voltage— samples in a mixture of nitric, hydrochloric and hydrofluoric Peak current/A 30 Repetition rate/kHz 2 acids. Carrier gas flow/l min-1 0.987 Spark gap/mm 2.0±0.1 Spark ablation ‘Pre-burn time’/s 60 Owing to the non-conductive nature of the catalytic converter samples, SpectroStandard powdered graphite (Chemplex, Stuart, FL, USA) was mixed with them as part of the pelleting process, a non-conductive material to graphite ratio of 0.5 was found to be optimal.Platinum(IV) (99.99% pure, Aldrich Chemicals, Milwaukee, WI, USA), vanadium(V) (99.5% pure Baker analysed reagent, J. T. Baker) oxides and NIST SRM 2556 Recycled Pellet (NIST, Gaithersburg, MD, USA) were used for instrument calibration. SRM 2556 was ignited for 2 h at 500°C and stored in desiccator prior to analysis.Platinum oxide underwent a series of consecutive dilutions, mixing with powdered cordierite substrate, to give a final concentration of about 100 mg g-1 for Pt. Vanadium oxide underwent a series of consecutive dilutions, mixing with graphite, to give a final Vconcentration of about 1000 mg g-1 in the synthetic standard. A 2 g portion of a catalytic converter sample (100 mesh) was mixed with 4 g of graphite powder in a ball mixer–mill for 5 min and pressed into a pellet (2.54 mm in diameter) with gradually applied pressure (1950 psi; 1 psi=6894.76 Pa) in a hydraulic press.Standards were prepared by mixing and Fig. 2 Typical sample chamber for spark ablation experiments. pelleting X g of platinum(IV) synthetic standard or SRM 2556 with (2-X) g of substrate and 4 g of graphite. Blanks were prepared by mixing and pelleting 2 g of cordierite with a washcoat with 4 g of graphite. In the case of V measurements, A 600 W microwave digestion oven MDS-81D (CEM, standards were prepared by mixing and pelleting Xg of Indian Trails, NC, USA) was used for sample digestion. synthetic standard with 2 g of cordierite with a washcoat and The sample chamber for spark ablation experiments consists 4 g of graphite; pure graphite pellets were taken as blanks of a cylindrical quartz body sealed by a Teflon endpiece on owing to the presence of V in cordierite with a washcoat.one side and by the pelleted sample on the other side.A typical spark cell assembly used in this work is shown in Fig. 2. The RESULTS AND DISCUSSION counter-electrode consists of a hollow brass body with a Spark Ablation Optimization and Parametric Studies circular array of holes on the top surface and a polished thoriated tungsten tip. Argon flow, introduced to the chamber Monitoring the ICP-MS ion count can serve as an excellent through the counter-electrode, carries material ablated from diagnostic tool in determining spark sampling efficiency and the sample surface to the ICP-MS instrument through 1 m of in optimizing spark parameters.Some information on sampling fluoropolymer tubing (6.35 mm id). mechanisms can also be obtained. Owing to the transient nature of the spark sampling process, the detected ICP-MS signal is the average of few hundred individual spark discharges Samples depending on the spark repetition rate (typically 0.1–2 kHz) and data acquisition time.It appears that high spark repetition The active components of automotive catalytic converters are precious metals including Pt, Pd and Rh that are finely rates of 2 kHz (further increases in repetition rate results in thyratron tube saturation) and relatively long integration times dispersed on an inert substrate such as c-alumina, which is in turn dispersed on a honeycomb ceramic substrate, usually (about 1 s) are optimal. However, the above methodology alone has limitations.A relatively wide distribution of ablated cordierite. More detailed descriptions of catalytic converters can be found elsewhere.13,15 In the current study ‘Pd-only particle sizes can generate signal spikes. Such behavior is increased whenever there is a high degree of spark wander wavelength dispersive X-ray fluorescence (WDXRF) secondary standards’ with Pt and Rh in the range of 1–100 mg g-1 were about the sample surface. In order to stabilize the signal, the particle size distribution introduced to the ICP is limited by used.These were first ignited for 2 h at 500 °C. They were further dried in a conventional oven for 2 h at 110°C and the spray chamber (which also acts as a pneumatic ‘buffer’ in stabilizing the aerosol carrier gas flow). In most cases the spray stored in a dessicator prior to analysis. Samples were analysed by both liquid nebulization and spark ablation. chamber was left in position. Ablated material carried by the 232 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Fig. 3 Dependence of the time profile measurements for 195Pt, 48Ti, 103Rh, and 220.00 on the presence (a) and absence (b) of the spray chamber when ablating SRM 2556. argon flow, before entering the torch, was passed through the spray chamber wherein excessively large particles were Fig. 4 Time profile measurements for 195Pt, 103Rh, and 220.00 deposited. Signal dependence with the absence and the presence achieved at three spark discharge peak currents: (a) 20; (b) 30; and (c) of the spray chamber while ablating SRM 2556 Used 40 A.Automotive Catalytic Converter is illustrated in Fig. 3. It is evident that the presence of the spray chamber significantly improves measurement precision. After the initial offset, the signal rises rapidly and levels out for the duration of the ‘burn’. Although this process is faster when the spray chamber is not present, the signal is much more stable when the spray chamber is used.As evidenced in Fig. 3, decay of the signal is also more rapid in the absence of the spray chamber. When the spray chamber is used, sample carryover is a potential source of memory effects; this was not found to cause significant problems when the system was purged with argon for 1 min between determinations. In both cases [Fig. 3(a) and (b)] ignition of the spark source is accompanied by an increase in instrument background level, which is maintained during all the experiments.This phenomenon can be monitored by measuring the signal at m/z 220 where no elemental or polyatomic ions are present (formation of 208Pb12C was not observed under the specified operating conditions). The influence of spark rf noise on the ICP-MS signals has previously been observed by Hirata et al.16 In the present case such an interference was even more prominent. In order to study the dependence of spark parameters on Fig. 5 Dependence of the signal intensity for 48Ti, 195Pt, and 105Pd sampling characteristics, spark peak current and spark rep- on the spark discharge peak current.etition rate were studied. The time profile measurements for three different spark discharge peak currents while sampling SRM 2556 are illustrated in Fig. 4. The achievement of (arbi- ICP-AES signal versus ablation laser power density. This similarity is not unexpected because spectral emission in ICP- trarily defined) pseudo-steady-state signal levels is rapid when higher peak current settings are used (indicated by arrows in AES and ion count in ICP-MS are related to the amount of material removed from the sample surface, which in turn Fig. 4). On the other hand, measurement precision is lower at higher peak currents. As expected, higher currents result in depends on the energy applied. Between spark peak current settings of 10 and 30 A, significant increases in the ICP-MS increased sample uptake which is accompanied by formation of deposits on the chamber walls.In severe cases this can signals are observed for all three elements. Beyond these values the slope of the signal versus spark current curve does not result in spurious arcing within the chamber. In Fig. 5, signal dependence on peak current is shown for three major elements change significantly. The observed change in slope is probably due to mechanistic changes in the spark ablation process. present in the catalytic converter.The resultant curves are similar to those obtained by Fernandez et al.3 who plotted Again similar behavior is observed for all three elements. As Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 233For example, in the series of ‘Pd-only’ catalytic converters under investigation, concentrations of interfering species such as Sr vary over four orders of magnitude. The same can be said for Hf and Zr. This fact creates serious matrix matching problems.Among the major interferences are ClO+ and BAr+ on V51, SrO+ and SrOH+ on Rh103, and HfO+ on Pt isotopes. It is generally agreed that conditions that result in the lowest levels of polyatomic oxides are low aerosol carrier gas flow and high rf generator power.1 Similar behavior was observed in the present experiments. Although significant reduction of the interfering species was achieved through low aerosol carrier flow rates (0.8 l min-1) and high rf power (1300W), complete elimination of these ions is not possible.A further step was the development of a reliable method for the determination of V, Rh and Pt using liquid nebulization, especially when determination of the presence or absence of an analyte is to be judged. Spectral interferences such as 35Cl16O+ and 179Hf16O+ on 51V+ and 195Pt+, respectively, can be corrected using simple equations, i.e., by measuring the signals at m/z 53 Fig. 6 Dependence of the signal intensity for 48Ti, 195Pt and 105Pd for 37Cl16O+ and 193 for 177Hf16O+ and back-calculating the on the spark discharge repetition rate.contribution from 35Cl16O+ at m/z 51 and 179Hf16O+ at m/z 195. Unfortunately, since signals from 10B40Ar+ at m/z 50 and 88Sr16O+ at m/z 104 are small compared with highly abundant noted, at high peak current settings (above 30 A) the problems Ti+, Pd+ and ZrO+ in the sample, this methodology is not of material deposition on the spark chamber walls and arcing feasible when interferences from 11B40Ar+ and 87Sr16O+ on increases.It is generally believed that mechanical ablation is 51V+ and 103Rh+ are considered. This phenomenon is illus- the major process responsible for sampling. The similar trated in Figs. 7(a) and 8(a). In this case the use of simple behavior for the three elements suggests that preferential equations to compensate for spectral interferences can result vaporization (fractionation) of the elements does not occur.in the introduction of large errors. However, more detailed studies, including elements with a Another approach to overcome spectral interference prob- wide range of melting-points, are required before accurate lems is to calculate relative yields of the interfering species. conclusions can be drawn. For example, BO+5B+ and SrO+5Sr+ ratios are easily In Fig. 6 the dependence of signal intensity as a function of obtained from measurements in standard B and Sr solutions spark repetition rate is shown.As expected, an increase of the under experimental conditions identical to the sample analysis. spark discharge frequency, while maintaining a fixed peak These ratios can be used to correct experimental data for current, results in a proportional increase in the amount of analyte elements. In recent work by the present workers this material removed by spark ablation. An almost identical approach was employed. It assumes that experimentally response is observed for Ti, Pd and Pt.In both these parametric obtained ratios are the same in the standard solution and in experiments the baseline level, defined at m/z 220.00, was the sample solution in the presence of matrix elements. measured and subtracted from each point. Based on these results it is preferable to use lower, but adequate, spark energy and high repetition rates in order to provide sufficient and stable ablation. Optimized values for Spark ablation spark source operation are given in Table 1.The ‘pre-burn The dry plasma, resulting from direct spark ablation, results time’ is that required to achieve pseudo-steady-state in significant decreases in the intensity of O-, H-, N- and operation (Fig. 4). Ar-containing species.17 Fig. 7 is a mass spectrum in the V range for both liquid nebulization and spark ablation. In the case of solid sampling, the absence of V spectral interferences Analytical Results is evident [Fig. 7(b)]. As expected, a digestion blank consisting L iquid nebulization of a mixture of hydrochloric, nitric, hydrofluoric and boric acids results in spectral overlap [Fig. 7(a)]. Utilization of the Digestion of the catalytic converter, prior to liquid sample nebulization, resulted in clear solutions with good recoveries spark source for solid sampling of catalytic converters for the determination of V is undoubtedly advantageous. Likewise, for precious and most other metals (except for some rare earth elements as was shown in earlier studies).13 A major problem Rh and Pt spectral information for liquid nebulization [Fig. 8(a)] and for spark ablation [Fig. 8(b)] can be compared. which accompanies liquid nebulization is the formation of interfering species.1 It was found that determination of rela- Although more detailed studies of polyatomic ionic formation under the dry plasma conditions are needed, current results tively high concentrations of analytes (about the same order of magnitude as an interfering parent element) is not signifi- show no evidence of such interferences for solid sampling and significant simplification of the resultant spectra.Note a sig- cantly affected since the relative yield of polyatomic and doubly charged ions is about 1–2%. Close matrix matching was nificant reduction in the ZrO+ signal intensity at m/z 107 (where no Pd isotopes are present) and no evidence of HfO+ successfully used.13 The idea of matrix matching is to use cordierite with a washcoat, which contains all the elements at species in the Pt mass range [Fig. 8(b)] under solid sampling conditions. Future research will investigate the formation of levels similar to the catalytic converter, except for precious metals, in order to correct for minor interferences. However, polyatomic ions in the spark ablation experiments. Expansion of the Rh spectral range is shown in Fig. 9. The when analyte concentrations are about 100–1000 times lower then those of the interfering parent elements, as it is in ‘Pd- measured signal intensity corresponds to 1.15 mg g-1 of Rh in the SRM pellet and about 8 mg g-1 of the element in the pellet only’ catalysts, induced spectral interferences cannot be neglected.In addition, loadings of potential interfering elements prepared from catalytic converter sample 1 (based on WDXRF data). The Pt spectral range is shown in Fig. 10. Spark ablation on the catalytic converter vary widely from sample to sample. 234 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Fig. 7 Portion of mass spectrum in vanadium range for $ blank and # cordierite with a washcoat obtained by liquid nebulization (a) and spark ablation (b) ICP-MS. of the SRM and sample 8 pellets correspond to 15.6 and about A five-point calibration with equation 5 mg g-1 of Pt, respectively. Y=-230.82+170.45×X where Y is signal intensity (counts s-1), X is platinum concen- Instrumental Calibration tration (mg g-1) and the correlation coefficient 0.998 was used L iquid nebulization for the determination of Pt in SRM 2556.The negative offset in the calibration equation is due to spark source rf interference Using the external instrument calibration technique, standard on the ICP-MS system. The certified Pt value is 697.4±2.3 mg V, Rh and Pt solutions were added to the digestion reagent g-1 (stated uncertainties are 99% confidence intervals of the blank. Elemental equations and relative yields of interfering single method mean).In the course of this analysis, SRM 2556 species were used to correct for stated interferences. Based was found to contain 689.3±12.7 mg g-1. Given that the spark on five-point instrument standardization, the calibration ablation ICP-MS results are the mean values of two different equations listed in Table 2 were obtained. burns (ten replicates each) on two different spots on the sample surface with confidence limits calculated at 95% confidence Spark ablation levels, both values are in good agreement.One of the features which distinguishes spark ablation from A series of pellets containing different amounts of a synthetic conventional liquid nebulization is the periodic occurrence of Pt standard were prepared and used for instrument calibration. ‘spikes’ or anomalies in the signal. This is illustrated in Fig. 11 where nine replicates of the Pt isotopes were measured. The presence of such of spikes is generally due to occasional spark Table 2 Calibration equations for the determination of V, Rh and Pt obtained with liquid nebulization wander, hence ablation of different spots on the sample surface with slightly different particle sizes.A second characteristic Intercept/CPS Slope/CPS Correlation feature of spark ablation is the background offset and its drift. Element (ln CPS)-1 (ln CPS)-1 (mg g-1)-1 coefficient Such behavior is due to residual rf interference caused by the V 0 0.0113 0.999 high-voltage spark.The effects can be minimized by: collection Rh 0 9.744×10-4 1.000 of more replicates, using longer integration times and per- Pt 0 3.634×10-3 1.000 forming two-point background correction (which normally Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 235Fig. 10 Portion of mass spectra in platinum range for cordierite with a washcoat (a), SRM 2556 (b), and sample 8 (c). Fig. 8 Portion of mass spectra for sample 2 obtained by liquid nebulization (a) and spark ablation (b) ICP-MS.Fig. 11 Portion of mass spectra in platinum range (nine replicates are collected) showing spikes on the signal due to the spark wander. concentration range of these elements in the samples of interest Fig. 9 Portion of mass spectra in rhodium range for cordierite with only. Owing to the fact that amounts of Pt in SRM 2556 are a washcoat (a) SRM 2556 (b) and sample 1 (c). approximately 14 times higher than that of Rh, only five points of the ten-point calibration curve for Pt are shown. is not used in ICP-MS measurements owing to the low background levels).Comparison of Results This methodology was employed for elemental analysis of a series of ‘Pd-only’ catalytic converters. Final instrument cali- The results for a series of ten ‘Pd-only’ catalytic converters achieved by WDXRF, ICP-MS with liquid nebulization and bration was achieved by adding different amounts of synthetic V standard to the cordierite with a washcoat (standard spark ablation ICP-MS are compared in Table 3.Presented ICP-MS data are the mean values of five replicates and two additions technique) for V calibration and by preparing a series of solid standards from SRM 2556 (external calibration different burns (ten replicates each) on two different surface spots for liquid nebulization and spark ablation, respectively, technique) for Rh and Pt calibration. Resultant calibration curves with correlation coefficients 0.990, 0.996 and 0.991, with confidence limits calculated at 95% confidence levels.In the case of spark ablation experiments, improvement of the respectively, for V, Rh, and Pt are shown in Fig. 12 for the 236 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12capability of ICP-MS allows for such measurements to be done. As was discussed previously, low detection limits attainable by liquid nebulization ICP-MS are spoiled by spectral interferences.Spark ablation ICP-MS is relatively free from spectral interferences due to different O-, H- and N- species but has its own disadvantages as mentioned above. Nevertheless, both ICP-MS sampling techniques produced some non-zero results for these samples, except in one case for Rh. Vanadium was found to be present in cordierite with a washcoat in amounts of 57.6±1.2 mg g-1 by liquid nebulization and 58.6±8.3 mg g-1 by spark ablation ICP-MS and was not analyzed further in catalytic converter samples.Again, as for Rh and Pt, both sampling techniques provided similar results, although, liquid nebulization had somewhat better precision. Conclusions and Future Development As a result of preliminary studies, spark ablation was shown to be a promising alternative to liquid nebulization for ICP-MS analysis. Reduction of sample preparation times and significant reduction of polyatomic species were achieved. Owing to the reasons of background shift and its drift in the course of the measurements, as well as spark wander Fig. 12 Calibration curves for vanadium (a), rhodium (b) and plati- effects, detection limits attainable by spark ablation are at num (c).least an order of magnitude higher then those for liquid nebulization. Detection limits for Rh and Pt achieved in recent work are, respectively, 0.02 and 0.04 mg g-1 for liquid Table 3 Pt and Rh determination in the automotive catalytic converter samples nebulization and 1.0 and 0.5 mg g-1 for spark ablation.Development of a new ‘zero-bias’ spark source in this ICP-MS laboratory, which is characterized by excellent stability, low voltage, high attainable repetition rates, negligible rf inter- Sample WDXRF Liquid nebulization Spark ablation ference and absence of spark wander is believed to improve Rh/mg g -1— analytical performance further, including lowering of detection 1 25 20.99±0.51 13.39±0.92 limits. Preliminary results are very encouraging. 2 ND* 0.260±0.053 0.58±0.28 3 15 13.50±0.28 8.19±0.80 The authors are grateful to Frank Kunz of the Ford Motor 4 51 42.73±0.90 47.8±2.7 Company for collaboration, supplying sample materials, 5 ND 0.690±0.074 1.03±0.51 providing WDXRF data and for helpful discussions. 6 5 5.260±0.096 4.7±1.1 7 ND 0.552±0.029 0 8 51 43.31±0.73 42.4±1.9 REFERENCES 9 10 9.854±0.090 6.38±0.93 10 20 17.94±0.30 10.3±2.0 1 Evans, E. H., and Giglio, J. J., J. Anal. At.Spectrom., 1993, 8, 1. 2 Cousin, H., and Magyar, B., Microchim. Acta, 1994, 113, 313. Pt/mg g -1— 3 Fernandez, A., Mao, X. L., Chan, W. T., Shannon, M. A., and 1 15 15.92±0.26 15.88±0.71 Russo, R. E., Anal. Chem., 1995, 67, 2444. 2 ND 2.11±0.02 2.76±0.39 4 Carr, J. W., and Horlick, G., Spectrochim. Acta, Part B, 1982, 37, 1. 3 25 26.04±0.10 22.87±0.90 5 Lichte, F. E., Anal. Chem., 1995, 67, 2479. 4 ND 2.22±0.02 1.64±0.60 6 Ivanovic, K. A., Coleman, D. M., Kunz, F. W., and Schuetzle, D., 5 ND 2.61±0.09 3.85±0.66 Appl.Spectrosc., 1992, 46, 894. 6 10 11.65±0.17 12.53±0.52 7 Coedo, A. G., Dorado, T., Rivero, C. J., and Cobo, I. G., J. Anal. 7 41 40.70±0.78 40.3±1.3 At. Spectrom., 1993, 8, 1023. 8 5 7.77± 0.08 7.60±0.86 8 Coedo, A. G., Lo�pez, T. D., Cobo, I. G., and Baquero, E. E., 9 30 30.96±0.32 30.0±1.0 J. Anal. At. Spectrom., 1992, 7, 247. 10 20 21.00±0.24 18.27±0.64 9 Coedo, A. G., Lo�pez, M. T. D., Seco, J. L. J., and Cobo, I. G, J. Anal. At. Spectrom., 1992, 7, 11. * ND=not determined. 10 Coedo, A. G., Dorado, M. T., and Cobo, I. G., J. Anal. At. Spectrom., 1993, 9, 223. 11 Coedo, A. G., Dorado, T., Escudero, E., and Cobo, I. G., J. Anal. signal-to-noise ratio was achieved by collection of ten replicates At. Spectrom., 1993, 8, 827. and 1 s integration times for each point of the spectral peak. 12 Jiang, S.-J., and Houk, R. S., Spectrochim. Acta, Part B, 1987, Spectra obtained by these means were corrected for back- 42, 93. ground shift using a two-point background correction pro- 13 Borisov, O. V., Coleman, D. M., Oudsema, K. A., and Carter, R. O., J. Anal. At. Spectrom., in the press. cedure. As can be seen, both ICP-MS sample introduction 14 Coleman, D. M., and Walters, J. P., Spectrochim. Acta, Part B, techniques provide similar results, which in most cases are 1976, 31, 547. close to the WDXRF data. Some deviations of the results for 15 Beary, E. S., and Paulsen, P. J., Anal. Chem., 1995, 67, 3193. Rh are connected with problems associated with the determi- 16 Hirata, T., Akagi, T., and Masuda, A., Analyst, 1990, 115, 1329. nation of a mononuclidic element. Moreover, WDXRF data 17 Jakubowski, N., Feldmann, I., Sack, B., and Stuewer, D., J. Anal. are not certified values and are given for comparison purposes At. Spectrom., 1992, 7, 121. only. The major concern was the samples with ‘zero’ (according to WDXRF data) amounts of Pt and Rh. One of the known Paper 6/05142A disadvantages of WDXRF is insufficient trace metal analysis Received July 23, 1996 Accepted October 23, 1996 detection limits. On the other hand, the excellent detection Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12

ISSN:0267-9477    

DOI:10.1039/a605142a

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Determination of Platinum, Palladium, Rhodium and Titanium in Automotive Catalytic Converters Using Inductively Coupled Plasma Mass Spectrometry With Liquid Nebulization OLEG V. BORISOVa , DAVID M. COLEMAN*a , KRISTINE A. OUDSEMAa AND R. O. CARTER IIIb aDepartment of Chemistry, Wayne State University, Detroit, MI 48202, USA bScientific Research L aboratory, Ford Motor Company, P.O. Box 2053, Dearborn,MI 48121, USA ICP-MS was used for the determination of Pt, Pd, Rh and Ti distribution of the precious metals on the support.Moreover, the limited detection limits of the method do not permit trace in automotive catalytic converters. A microwave digestion procedure employing a mixture of H2O, HF, HCl and HNO3 and ultra-trace analysis. This became important after the introduction of ‘palladium-only’ catalytic converters which was studied and compared with a procedure using a mixture of H2SO4, H3PO4 and aqua regia. The use of the latter results in contain between 1 and 100 mg g-1 of platinum and rhodium.4 Another disadvantage of the WDXRF method is that the significant losses of Pt and Ti.The influence of sample size on the recoveries of Pt, Rh, Ti and some REEs is discussed. It is palladium line interferes with the ‘best’ rhodium line. For these reasons, in addition to the lack of meaningful standards shown that a sample amount larger than 0.12 g results in incomplete recoveries of REEs with no significant loss of the (covering the wide range of concentrations of the precious metals and additives), WDXRF is used primarily for quality other elements.Instrument calibration using the standard addition method and matrix matching are discussed. It is control. Since introduction of the ICP-MS technique, trace and ultra- shown that close matrix matching using cordierite, the substrate, with a washcoat, the catalyst support, provides trace analysis of various geological materials has shown great progress. Several potential advantages include low detection spectral interference correction for polyatomic oxide species.The results obtained are in good agreement with those limits, relatively simple spectra, high sensitivity and large dynamic range. ICP-MS has been widely used for the determi- obtained by wavelength-dispersive X-ray fluorescence analysis. nation of precious metals in geological materials.5–8 Most of Keywords: Inductively coupled plasma mass spectrometry; the methods use a fire assay procedure to recover precious platinum, palladium, rhodium and titanium; automotive elements.The nickel sulfide procedure, although simple for catalytic converter collecting precious elements, results in some losses.6 In another fire assay method, precious metals are fused in a lead button which is subjected to cupellation with silver to yield a bead of Since 1975, catalytic converters have been widely implemented precious metals. Unfortunately, this method is not ideal for for controlling automobile exhaust emissions.The purpose of automotive catalysts owing to the high concentration of the catalytic converter is to oxidize completely hydrocarbon alumina present.9 species to carbon dioxide and water. Later modifications of Sample preparation methods include an aqua regia leach catalytic converters resulted in the introduction of ‘three-way’ prior to determination of precious metals in geological catalysts which oxidize carbon monoxide and hydrocarbons samples.5 Aqua regia is employed owing to the demonstrated and also reduce nitrogen oxides to molecular nitrogen. From solubility of most elements.The method, although less labori- this perspective, platinum group metals (mainly platinum, rhoous than fire assay procedures, produces incomplete recoveries dium and palladium) are known to be catalytically active. The of some elements and was recommended only for preliminary precious metals in different ratios are deposited on large surface studies. Treatment of the sample with a mixture of sulfuric area c-alumina (washcoat), which in turn is dispersed on a and phosphoric acids, which is extremely effective in dissolving honeycomb consisting of aluminum magnesium and/or calcium alumina, was employed prior to addition of aqua regia in aluminum silicates (usually cordierite, representing serpentine analysis of the catalytic converters.10 This modified leach, group ceramics).1 Cerium and other REEs are added to the washcoat during catalyst production.The use of cordieritebased catalytic converters is justified by their favorable properties, such as low temperature coefficient of thermal expansion and excellent thermal resistance, along with reasonable mechanical properties.2 A typical catalytic converter is shown in Fig. 1. The use of expensive platinum group elements in catalyst production fostered the development of an accurate method of determination.Wavelength-dispersive (WD) XRF, often employed, has advantages of relatively simple sample preparation and rapid measurement.1 On the other hand this method suffers from several disadvantages. Catalytic converters, especially spent ones, because of high-temperature exposure, can undergo partial conversion from c-alumina to a-alumina. The various forms in which the alumina may be present can result in high and fluctuating background levels.3 Fig. 1 Typical automotive catalytic converter. Another major difficulty is connected with the inhomogeneous Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (239–246) 239developed to solubilize precious elements without dissolving WDXRF data were generated on a Siemens (Madison, WI, USA) SRS 300 wavelength-dispersive spectrometer using the the cordierite substrate, provided good recoveries of rhodium and palladium. The behavior of platinum and other major standard procedures established at the Ford laboratory.elements was not studied. Moreover, platinum dioxide, which may be formed during catalyst preparation, is not soluble in Sample Dissolution Procedures aqua regia.11 On the other hand, hydrofluoric acid, which is superior in the dissolution of different silicates, was shown to Catalytic converter samples were dissolved by two different procedures. Commercial cordierite [10SiO2 4Al2O3 4 provide optimum results for some elements of interest when it was used for automotive catalyst dissolution.12 The use of (MgFeO) H2O] and cordierite with dispersed c-alumina (washcoat) underwent the same digestion procedure as the hydrofluoric acid, however, can result in loss from solution of REEs and some other elements which form only sparingly sample and were used for matrix matching.PFA Teflon digestion vessels (CEM) fitted with pressure relief valves were soluble fluorides. In order to bring these fluorides into solution and to deactivate excess of unreacted hydrofluoric acid, boric used.The vessels rotated on a Teflon turntable during digestion, to ensure uniform exposure of each sample to the microwave acid is usually added. This results in a high saline concentration which may cause complications in the subsequent spectro- radiation. Prior to each digestion, 20 ml of nitric acid were added to each vessel and the microwave generator was operated metric determination of individual components. Recently, Beary and Paulsen13 described the development of a method at 25% power for 20 min.Subsequently the vessels were rinsed with deionized water and dried. Samples (100 mesh powder for the ICP-MS certification of Pt, Pd, Rh and Pb in two spent automotive catalytic converters using a combination of ignited beforehand for 2 h at 500°C) were further dried in a conventional oven for 2 h at 110°C and stored in a desiccator. HNO3 and HF and either HClO4 or HCl for sample dissolution. In this work, all of the elements except for mononuclidic This drying time was adequate to reach constant mass.Recovery studies were performed by varying the sample Rh were quantified by isotope dilution ICP-MS. Owing to the wide variety of catalytic converters with amount from 0.05 to 0.25 g. different additives and ratios of precious metals, different dissolution techniques may be required. Different approaches Reagents to correcting for spectral interferences may also vary from sample to sample.In this work, we attempted to compare Concentrated nitric, hydrochloric, hydrofluoric and sulfuric acids used were of Ultrex II Ultrapure grade (J. T. Baker, sample preparation using two different microwave digestion techniques, namely a modified aqua regia leach and dissolution Phillipsburg, NJ, USA). Analytical-reagent grade concentrated phosphoric acid (J. T. Baker) verified to be free from precious with hydrofluoric acid. The recoveries of precious metals and other major elements achieved with both procedures were metals was used.Crystalline boric acid was J. T. Baker Ultrex (99.98% pure). Calibration standards were diluted from single- examined. Different instrument calibration techniques with and without matrix matching were tested. Based on these element 1000 mg l-1 stock solutions (Spex Plasma Standards, Spex Industries, Edison, NJ, USA). All dilutions were made studies, a final analytical procedure was adopted. The results obtained were compared with those obtained by WDXRF. with deionized water (resistivity of at least 18 MV cm).An indium standard solution was added to all samples and diges- Another purpose of this work was to obtain reliable results to facilitate further study using direct solid sampling with a tion blanks as an internal standard. Two different dissolution procedures were initially evaluated. high-voltage uni-directional spark. It is expected that this method will minimize most of the errors connected with sample dissolution, including loss of the elements of interest and Evaluation Procedure 1 possible contamination.This dissolution procedure was adopted and modified from a method reported by Kingston and Jassie.14 Four millilitres of EXPERIMENTAL nitric acid, and 5 ml of a HCl–HF (7+3) were added to each vessel, containing 0.1–0.2 g of sample. The vessels were sealed Instrumentation with a cap and relief valve using a capping station. The The ICP-MS instrument was an Elan 5000 (Perkin-Elmer microwave oven generator was operated at 100% power for SCIEX, Thornhill, Ontario, Canada).Samples were introduced 2.5 min and at 65% power for 20 min. The vessels were cooled by means of a crossflow pneumatic nebulizer. ICP-MS and 50 ml of 0.35 m l-1 boric acid solution was added to each conditions were optimized to maximize the detector response vessel to dissolve insoluble fluorides and to allow the sub- while aspirating a 10 mg l-1 multi-element solution.The sequent use of conventional glassware. The vessels were re- operating conditions used are summarized in Table 1. sealed and the microwave generator was operated at 100% A 600 W MDS-81D microwave digestion oven (CEM, power for 10 min. The resulting solutions were transferred into Indian Trails, NC, USA) was used for sample preparation. 100 ml calibrated flasks and diluted to volume with deionized water. The samples were filtered using Whatman No. 42 Table 1 Instrumentation and operating conditions filter-paper. Plasma conditions— Rf power 1000W Evaluation Procedure 2 Plasma gas flow rate 14 l min-1 Auxiliary gas flow rate 0.9 l min-1 A 2.5 ml volume of sulfuric acid and 2.5 ml of phosphoric acid Nebulizer gas flow rate 0.987 l min-1 were added to the vessels containing the powder samples (0.1–0.2 g).10 The vessels were sealed and the microwave Mass spectrometer conditions— generator was operated at 50% power for 10 min.This was Ion lens settings (relative voltages) repeated four times, allowing the vessels to cool to room B lens 42 E1 lens 15 temperature between cycles. A 20 ml volume of aqua regia was P lens 46 then added to each vessel. A vigorous reaction ensued and S2 lens 45 upon completion of the reaction (approximately 10–15 min) Sampler cone Nickel, 1.14 mm orifice the vessels were sealed and the microwave generator was Skimmer Nickel, 0.89 mm orifice operated at 45% power for 15 min.This was done three times, 240 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12allowing the vessels to cool to room temperature between can be divided into two groups: matrix-induced spectral overlap arising from oxide and polyatomic ions (spectral cycles. Following dissolution, some particles remained. The samples were filtered using Whatman No. 42 filter-paper. interferences) and matrix-induced signal intensity changes (non-spectral interferences).The isotopes used in this study, Digestion Evaluation Procedure 2 was subsequently refined for particular samples into the final Analytical Procedure 3, along with possible spectral interferences, are listed in Table 2. In order to investigate matrix-induced interferences from the which is detailed later. species listed in Table 2 and their possible contribution to quantitative analysis errors, semi-quantitative analysis was RESULTS AND DISCUSSION carried out first. The idea of semi-quantitative measurements available with the ICP-MS software is based on the utilization Instrument Calibration Techniques of a ‘general response curve’ generated from a multi-element Two major techniques were used for instrument calibration: standard solution.Using this response curve and the intensities standard addition and external standardization. In external obtained from the sample elements, the software algorithm standardization, calibration of the instrument was performed provides an estimate of concentrations for all elements.In by addition of different amounts of stock single-element stan- this work the multi-element reference solution consisted of dard solutions to the ‘blanks’: (1) an acid digestion blank 100 mg l-1 each of Li, Be, Se, Y, Rh, Nb, Re, Tl and Bi. Semi- which contained all the acids, but no sample, underwent the quantitative analysis of catalytic converter material indicate same digestion procedures as the catalytic converter samples; 50, 5400, 6000, 200, 100 and 20 mg g-1 of Cu, Zr, Ca, Hf, Sr (2) a solution prepared from digestion of cordierite; and (3) a and Y, respectively.solution prepared from digestion of cordierite with a washcoat. In a separate series of experiments, the possible formation ‘Blanks’ (2) and (3) are believed to provide corrections for of interfering species (listed in Table 2) induced by the matrix minor matrix-induced interferences such as interferences from was investigated.Single-element standard solutions containing doubly charged ions and polyatomic species, in addition to 100 mg l-1 of Cu, Zr, Ca, Hf, Sr and Y were used to calculate matrix suppression effects. The standard addition technique the relative yields of interfering species under the present consists in the addition of different amounts of stock standard operating conditions. Based on these results and estimated solutions to the sample of the catalytic converter, i.e., spiking concentrations of the parent interfering elements, the contri- the sample.This technique is not capable of correcting for butions of the indicated polyatomic and doubly charged ions matrix- and acid-induced interferences. In both cases the to the concentrations of the ions of interest were calculated as concentration of In as an internal standard was maintained at shown in Table 3. It is clear that interferences due to the 40 mg l-1 after the final dilution.formation of polyatomic and doubly charged species are small The concentrations of precious metals in the catalytic con- and do not significantly influence the elements of interest. verter sample were of major importance in this work; neverthe- However, interferences induced by the mineral acids used in less, other major and minor elements were also included in the the digestion procedures, especially polyatomics resulting from study. One of the major concerns in the determination of La sulfuric and phosphoric acids, cannot be neglected.This type and Pr was the fact that highly abundant Ce in the sample of interference is well documented. Digestion blanks resulting can cause interferences on the much less abundant La and Pr from the two procedures are shown in Fig. 3. Digestion blanks by peak broadening. However, as follows from Fig. 2, this was from Evaluation Procedure 1 resulted in only minor inter- not the case and all three peaks with mass 139 (La), 140 (Ce), ferences on titanium, which were successfully overcome by 141 (Pr) were well resolved even when the instrument was blank subtraction [Fig. 3(a)]. A different situation was operated at low resolution. observed when interferences from SO and PO polyatomic ions A significant problem in ICP-MS analysis is the occurrence on titanium were considered when Evaluation Procedure 2 of different matrix-induced interferences.15 These interferences was used [Fig. 3(b)]. These major interferences cannot be corrected by simple digestion blank subtraction. The resulting intensities from interfering ions were 30–50% of those for titanium isotopes in the sample.Furthermore, the precision of the measured intensities of interfering species is low and cannot be explained by counting statistics alone. This is possibly due to the fact that the formation mechanisms of the polyatomics and elemental ions are different. The presence of the major 31P16O and 32S16O polyatomic ions with mass 47 and 48, respectively, significantly affects the corresponding titanium isotopes and results in increased RSDs for corresponding titanium isotopes.This is clearly illustrated by comparing the RSDs for 20 replicates of the measured signal for titanium isotopes obtained by aspirating 500 mg l-1 of pure titanium Table 2 Isotopes used for analysis, isotopic abundances and possible spectroscopic interferences Possible interferences Induced by acids Induced by matrix used in the Isotope Abundance (%) components digestion procedure 48Ti 73.940 Zr2+ NO2, PO, CCl, SO 49Ti 5.510 CaH PO, CCl, SO, HSO 103Rh 100 SrO, CuAr, SrOH Fig. 2 Portion of mass spectra of cerium, lanthanum and praseodym- 105Pd 22.230 SrO, YO ium in the catalytic converter sample. La, Ce, and Pr concentrations 195Pt 33.800 HfO in the sample are 2000, 60 000 and 100 mg g-1, respectively. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 241Table 3 Relative yields of the possible interferencing species and their estimated contributions to the concentrations of elements of interest Estimated contribution to the Analyte Interfering species Relative yield (%) analyte concentration (mg g-1) 48Ti 96Zr2+ 0.04 0.06 49Ti 48Ca1H 1.30 0.20 103Rh 87Sr16O 0.11 0.01 105Pd 87Sr18O, 88Sr17O 0.12 0.006 103Rh 86Sr16O1H 0.14 0.01 103Rh 63Cu40Ar 0.05 0.02 105Pd 89Y16O 1.90 0.4 195Pt 179Hf16O, 177Hf18O 1.70 0.6 Fig. 3 Reagent blanks in (a) Evaluation Procedure 1 and (b) Evaluation Procedure 2.Fig. 4 Spectra of cordierite (a), cordierite with washcoat (b) and catalytic converter (c). standard solution with those obtained from Evaluation Procedure 2 digestion blanks spiked with 500 mg l-1 of standard titanium solution (Table 4). As a result, Evaluation Comparison and Evaluation of the Two Digestion Procedures Procedure 2 is not capable of providing optimum titanium Analytical results for the major elements attainable by the two analytical data.Nevertheless, these titanium measurements digestion procedures were obtained using semi-quantitative were made for comparative and evaluative purposes. analysis and the results are given in Table 5. It is clear that A catalytic converter spectrum, along with those for cordier- Evaluation Procedure 1 provides higher recoveries for most ite and cordierite with a washcoat, obtained using Evaluation Procedure 1 are shown in Fig. 4. It is evident that the cordierite and the cordierite with a washcoat do not contain appreciable Table 5 Semi-quantitative analysis of catalytic converter samples amounts of precious metals, La or Pr, validating its use for (0.1 g) as the result of two digestion procedures (Evaluation Procedure matrix matching in further analysis.However, possible matrix 1 with HNO3–HF–HCl and Evaluation Procedure 2 with H2SO4–H3PO4–aqua regia) effects cannot be corrected for titanium owing to its presence in the cordierite substrate and hence in all the samples. Concentration/mg g-1 Element Evaluation Procedure 1 Evaluation Procedure 2 Table 4 RSD values (based on 20 replicates) of the measured signal for titanium isotopes in the absence and presence of the Procedure 2 Ti 2900 900 digestion blank V 40 10 Fe 1200 70 RSD (%) Ni 60 40 Zr 5000 20 500 mg l-1 Blank spiked with Rh 360 340 Ti isotope Abundance (%) Ti standard 500 mg l-1 Ti Pd 2 2 La 270 1000 46 7.93 2.52 3.18 47 7.28 2.56 11.5 Ce 10 900 56 500 Pr 12 30 48 73.9 1.51 3.11 49 5.51 2.14 3.56 Hf 130 1 Pt 2300 190 50 5.34 2.65 3.32 242 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12elements in the sample, except La, Ce and Pr. The lower values was collected on ashless filter-paper, rinsed several times with hot hydrochloric acid, rinsed with deionized water and dried. for Ce and some other REEs in the digestion by Evaluation Procedure 1 may be due to the formation of insoluble complex The filter-paper containing the residue was then placed in a porcelain crucible, and ashed at 450 °C for 30 min in a muffle fluorides by these elements as a result of employing hydrofluoric acid.Addition of boric acid results in dissolution of furnace. The residue, comprising about 35–40% of the initial sample mass, was transferred to the Teflon digestion vessel fluorides, but may be incomplete owing to insufficient amounts of digestion acids and/or inadequate heating times. and digested by Evaluation Procedure 1.The resultant solution was clear. Analysis of this solution showed that this residue Precious metals, the major concern in our work, show similar results, except for Pt using Evaluation Procedure 2. was primarily Si containing significant amounts of Al, Mg, Ti, La, Ce, Pt and some other elements. Quantitative studies have This was not unexpected because platinum dioxide, which could be present in the sample, is not soluble in aqua regia.shown that in the case of 0.2 g of initial sample 1840 mg g-1 Ti, 1230 mg g-1 Ce and 230 mg g-1 Pt remain trapped with More detailed studies of the recoveries of precious metals and titanium achieved by the two digestion procedures in the the Si. catalytic converter sample were made using true quantitative analysis. Various sample amounts taken for digestion were studied. The dependence of the analytical results on the sample Further Development of the Digestion Procedure amount is shown in Fig. 5. The results for Rh obtained by the Based on these results, it is evident that Evaluation Procedure two procedures are similar [Fig. 5(a)]. There is almost no 2 is not capable of providing reliable results for most of the dependence of Rh concentration values on variations in the elements used in this work. Further work on this procedure sample amount. A different behavior was observed for Pt and was discontinued. Ti [Fig. 5(b) and (c)]. Titanium, being a matrix element, can In order to improve the recoveries of Ce and other REEs, serve as an indicator of dissolution completeness.For Evaluation Procedure 1 was slightly modified. The resultant Evaluation Procedure 1, when hydrofluoric acid was used, the method (Analytical Procedure 3) was used in all further work. highest results for Ti, and also for other matrix elements (in Water was added to the samples prior to acid addition in Table 5), were achieved. In comparison, Evaluation Procedure order to improve the sample wetability.Longer digestion times 2 provided lower recoveries of Ti, which are also dependent were also used, as discussed below. on the sample amount. Titanium is believed to be present in the catalytic converter sample in the form of Ti dioxide, which only slowly dissolves in hot sulfuric acid. Therefore, complete Analytical Procedure 3 transfer of Ti to the solution is unlikely. It is reasonable, therefore, that increases in sample size decrease the Ti A sample (0.05–0.25 g) was placed in the digestion vessel, 1 ml of deionized water was added and mixed with the sample until recoveries.A similar picture was observed for Pt [Fig. 5(b)]. Again, a uniform suspension was obtained, then 4 ml of nitric acid and 5 ml of HCl–HF (7+3) were added to each vessel. The Evaluation Procedure 1 produced higher results than Evaluation Procedure 2. This suggests an assumption that vessels were sealed with a cap and relief valve using a capping station.The microwave generator was operated at 100% power incomplete recovery of Pt from the sample occurs in the presence of platinum dioxide. for 2.5 min, at 25% power for 60 min, and at 65% power for 20 min. The vessels were cooled and 50 ml of 0.35 mol l-1 Residue left as the result of Evaluation Procedure 2 digestion boric acid solution were added to each vessel. Finally, the vessels were re-sealed and digested at 100% power for 10 min and at 35% power for 50 min.Analytical Procedure 3 provided clear solutions when 0.04–0.12 g samples were taken; some turbidity was observed with larger sample amounts owing to an insufficient amount of digestion acids. When turbidity occurred, the solutions were filtered, but small portions (10 ml) were left unfiltered. These unfiltered portions were diluted and treated identically with the filtered portions prior to introduction into the instrument by means of a conventional cross-flow nebulizer.Owing to the very small size of undissolved particles in these suspensions, clogging of the nebulizer was not a problem. This was done to determine which elements were not completely dissolved from the catalytic converter when larger sample sizes were selected. The analytical results obtained with the use of Analytical Procedure 3 using ICP-MS semi-quantitative analysis are given in Table 6 for a 0.2 g sample. It is clear that results for Ce and other REEs are significantly improved compared with the Evaluation Procedure 1 (Table 5).Because Ce is one of the major components of the catalyst, its concentration as a function of sample amount (semiquantitative measurements) is shown in more detail in Fig. 6. Introduction of an unfiltered portion of the sample in the form of a fine suspension into the ICP-MS resulted in uniform recoveries of Ce. On the other hand, filtering of the sample resulted in losses of Ce, probably in the form of complex fluorides.A similar pattern was observed for La and Pr. Fig. 5 Analytical results for Rh (a), Pt (b) and Ti (c) achievable by Quantitative analyses for Pt, Rh and Ti were carried out to two digestion procedures: $, Evaluation Procedure 1; &, Evaluation Procedure 2. determine the element recoveries from filtered and unfiltered Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 243Table 6 Semi-quantitative analysis of the catalytic converter sample additions were made to the digestion blank and the catalytic (0.2 g) as the result of digestion with Analytical Procedure 3 converter.Intermediate values were observed when additions (H2O–HNO3–HF–HCl) were made to the cordierite. This behavior is not surprising since cordierite with a washcoat contains all the potential Concentration/mg g-1 interfering elements (Table 3) except for the precious metals. Element Filtered sample Unfiltered sample It can therefore serve as an excellent matrix matching material for the correction of spectral and non-spectral interferences.A Ce 31700 57000 digestion blank or plain cordierite cannot provide complete La 540 1060 Pr 30 50 matrix matching. High results obtained from the standard Rh 350 350 addition method also show the presence of spectral inter- Pd 6 6 ferences and this technique is not capable of interference Pt 2300 2300 correction. Based on these results, further work was carried Ti 2900 2900 out via an external calibration method using cordierite with a washcoat as a matrix matching component.The results for Ti show no substantial difference between filtered and unfiltered portions of the samples; both calibration techniques provide similar results. Comparison with WDXRF Data The analytical results for precious metals, REE and Ti which were obtained in the course of this work were compared with WDXRF data and the results are given in Table 11. The WDXRF data are the averages of two runs on two different days, rather than certified values.The ICP-MS results are the mean values of triplicate samples digested using Analytical Procedure 3. Instrument calibration was carried out using the external calibration technique with additions to the cordierite with a washcoat. Comparison of the results for precious metals obtained by the WDXRF and ICP-MS techniques clearly shows they are similar. However, the Pd levels are below the WDXRF detection limit.The ICP-MS detection limits, defined as the concen- Fig. 6 Semi-quantitative analysis of cerium achieved for unfiltered tration of analyte that produces a signal equal to three times portion of the sample ($) and filtered portion of the sample (&) as a the standard derivation of the digestion blank (ten measure- function of sample amount. ments), were calculated and are given in Table 12 along with the concentrations of the elements of interest in the sample samples.Instrument calibration techniques, identical with those after final dilution. As follows from Table 12, low detection discussed earlier, were used. Unfortunately, the external Ti limits allow for the accurate measurement of the indicated calibration using additions to cordierite and cordierite with a elements. washcoat cannot be performed owing to the presence of Ti in The results obtained for the selected REEs (Table 11) are both matrix matching substances. consistently lower when obtained by ICP-MS.The results for Results for Rh, Pt and Ti obtained by these means are given Ti, on the other hand, are significantly higher. in Table 7. As can be seen, the Rh and Pt results are consistently For further validation of the method, spent automotive lower for addition to the cordierite with a washcoat than in catalytic converter (monolith) SRM 2557 with certified values all other calibration techniques for both filtered and unfiltered for Rh and Pt of 135.1±1.9 and 1131±11 mg g-1, respectively, samples.To compare filtered and unfiltered pairs for these was digested using Analytical Procedure 3 and analyzed using calibration techniques (Table 7), the F-test was employed in the external calibration technique with additions to cordierite. order to determine whether or not the corresponding standard The results obtained for Rh and Pt were 132.3±2.8 and deviations are statistically different. In other words, the test 1129±14 mg g-1, respectively (mean values for five replicates determines if two sample standard deviations belong to the with 95% confidence limits).As can be seen, the two sets of same population. In this case the ratio of s21/s22 (where s1>s2 results are in good agreement. and represent standard deviations for filtered and unfiltered samples, respectively) adequately follows the F-distribution with degrees of freedom f1 and f 2. The results are presented in CONCLUSIONS Table 8. As can be seen, F<F(0.95, 10, 10) and two mean values for filtered and unfiltered pairs (Table 7) can therefore Verification of the ICP-MS technique for the determination of precious metals in an automotive catalytic converter showed be compared using the t-test to determine whether or not the two mean values differ owing to statistical fluctuations in the that the technique is capable of providing accurate and reliable results which are in good agreement with WDXRF data.The data arising from indeterminate errors and belong to the same general population with the same mean value.Results of the ICP-MS technique has low detection limits, which make it superior for trace and ultra-trace determinations of precious t-test are illustrated in Table 9. As can be seen, there are no significant differences between the mean values [t<t(0.99, 20)]. metals. For this reason, it is critical that a digestion technique be employed that results in complete sample dissolution.It This allows the filtered and unfiltered populations for Rh and Pt to be combined as one general population. was shown that attempts to leach precious metals from the sample, rather than dissolve the latter, result in poor recoveries The results of such combination values for Rh and Pt are given in Table 10. It is evident that the mean results for Pt for Pt. This may be even more crucial in the case of trace amounts of Pt. It was shown that Analytical Procedure 3, and Rh were the lowest when additions were made to the cordierite with a washcoat. They were the highest when employing a mixture of HF, HNO3, HCl and H2O, provided 244 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12Table 7 Quantitative analysis data for rhodium, platinum and titanium (mg g-1) obtained by three calibration techniques (see text) Calibration technique Cordierite with a washcoat Blank addition Cordierite addition addition Sample addition Metal Sample mass/g Filtered Unfiltered Filtered Unfiltered Filtered Unfiltered Filtered Unfiltered Rh 0.0404 368 370 367 369 363 365 370 371 0.0454 364 366 363 364 359 360 370 368 0.0509 365 364 363 363 359 359 371 371 0.0609 365 366 363 365 359 361 370 369 0.0708 366 368 365 366 361 362 363 365 0.0910 367 367 365 366 361 362 363 366 0.1202 363 366 362 364 358 360 365 369 0.1517 365 369 363 367 360 363 362 368 0.1705 363 365 362 364 358 360 360 367 0.2009 365 366 363 364 359 360 371 373 0.2209 363 367 362 366 358 362 371 375 Mean 365 367 363 365 360 361 367 369 s 1.56 1.67 1.38 1.63 1.68 1.75 4.34 3.17 95% confidence limit 1.0 1.1 0.9 1.1 1.1 1.2 2.9 2.1 Pt 0.0404 2370 2400 2330 2370 2290 2320 2340 2340 0.0454 2340 2370 2300 2330 2260 2290 2290 2300 0.0509 2320 2320 2290 2290 2250 2250 2340 2310 0.0609 2340 2330 2300 2300 2260 2260 2340 2320 0.0708 2340 2330 2310 2290 2260 2240 2350 2300 0.0910 2290 2360 2210 2320 2230 2280 2250 2360 0.1202 2250 2330 2210 2290 2270 2250 2300 2430 0.1517 2290 2350 2250 2310 2260 2270 2370 2400 0.1705 2280 2320 2250 2280 2210 2240 2380 2410 0.2009 2280 2320 2250 2290 2200 2250 2350 2400 0.2209 2270 2310 2240 2280 2200 2240 2400 2450 Mean 2310 2340 2270 2300 2250 2260 2340 2370 s 37.5 27.2 41.0 27.0 30.1 25.3 43.0 54.8 95% confidence limit 25 18 27 18 20 17 29 37 Ti 0.0404 2860 2830 — — — — 2850 2850 0.0454 2870 2830 — — — — — — 0.0509 2970 2830 — — — — 2730 2780 0.0609 2790 2850 — — — — — — 0.0708 2800 2840 — — — — 3080 3050 0.0910 2880 2850 — — — — 3090 3020 0.1202 3040 2850 — — — — — — 0.1517 3020 2910 — — — — 3010 3110 0.1705 2860 2880 — — — — — — 0.2009 2940 2890 — — — — — — 0.2209 2920 2900 — — — — 3050 3020 Mean 2900 2860 — — — — 2970 2970 s 82.0 29.7 — — — — 145.9 127.7 95% confidence limit 56 20 — — — — 150 130 Table 8 F-test results of the comparison of the standard deviations Table 9 t-test results of the comparison of the mean values for filtered and unfiltered sample pairs [t(0.99, 20)=2.85] for the filtered and unfiltered sample pairs [F(0.95, 10, 10)=2.97] Calibration technique Calibration technique Cordierite with a Cordierite with a Cordierite washcoat Sample Cordierite washcoat Sample Element Blank addition addition addition addition Element Blank addition addition addition addition Rh 2.52 2.57 2.45 1.50 Rh 1.12 1.22 1.21 2.05 Pt 2.12 2.47 1.47 1.59 Pt 2.46 2.47 1.37 1.45 clear solutions in the case of 0.1 g samples or smaller.It is likely that an increased microwave digestion time would result dissolution by using a uni-directional high voltage spark source in improved recoveries.as a direct solid sampling device for ICP-MS. The preliminary The interferences observed in this work did not cause results are very encouraging. significant problems. This should be studied further in the case of low and trace determinations of precious metals such as in ‘palladium-only’ catalysts. The authors are grateful to Frank Kunz of the Ford Motor Company for collaboration, supplying sample material, In parallel work now in progress, we are attempting to eliminate the time-consuming and error-prone steps of sample providing WDXRF data and helpful discussions.Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 245Table 10 Mean values for rhodium and platinum (mg g-1) obtained REFERENCES from the combination of filtered and unfiltered sample pairs 1 Kallmann, S., and Blumberg, P., T alanta, 1980, 27, 827. 2 Kingon, A. I., and Davis, R. F., in Engineered Materials Handbook, Calibration technique Volume 4. Ceramics and Glasses, technical chairman Schneider, S. J., ASM International, Metals Park, OH, 1991, pp. 758–761. Cordierite with a 3 Kallamann, S., T alanta, 1976, 23, 579. Cordierite washcoat Sample 4 Thayer, A., Chem. Eng. News, 1993, 8, 6. Element Blank addition addition addition addition 5 Juvonen, R., Kallio, E., and Lakomaa, T., Analyst, 1994, 19, 617. Rh 366±1 364±1 360±1 368±2 6 Jackson, S. E., Fryer, B. J., Grosse, W., Healey, D. C., Longerich, Pt 2320±16 2290±17 2250±13 2350±22 H. P., and Strong, D. F., Chem. Geol., 1990, 83, 119. 7 Date, A. R., Davis, A. E., and Cheung, Y. Y., Analyst, 1987, 112, 1217. Table 11 Analytical data (mg g-1) obtained for the catalytic converter 8 Date, A. R., and Gray, A. L., Spectrochim. Acta, Part B, 1985, sample by ICP-MS and WDXRF 40, 115. 9 S¡ulcek, Z., and Povondra, P., Methods of Decomposition in Element ICP-MS WDXRF Inorganic Analysis, CRC Press, Boca Raton, FL, 1989, pp. 212–217. Rh 360±1 396 10 Oudsema, K. A., PhD T hesis, Wayne State University, 1993. Pd 7.4±0.2 (5) 11 Gmelins Handbuch der anorganische Chemie, Verlag Chemie, Pt 2250±13 2258 Berlin, 1939, System No. 68, Platin, Teil C1, p. 46. La 1340±23 1474 12 Brown, J. A., Jr., Kunz, F. W., and Belitz, R. K., J. Anal. At. Pr 532±1 830 Spectrom., 1991, 6, 393. Ti 2970±48 1500 13 Beary, E. S., and Paulsen, P. J., Anal. Chem., 1995, 67, 3193. 14 Introduction to Microwave Sample Preparation: T heory and Practice, ed. Kingston, H. M., and Jassie, L. B., American Table 12 Method detection limits obtained in the analysis of the Chemical Society, Washington, DC, 1988, pp. 41–42. catalytic converter by liquid nebulization ICP-MS and range of 15 Tan, S. H., and Horlick, G., J. Anal. At. Spectrom., 1987, 2, 745. concentrations of the elements during the actual analysis Element Method detection limit/mg l-1 Concentration range/mg l-1 Paper 6/05159F Received July 23, 1996 Rh 0.0015 5–15 Pd 0.26 0.5–3 Accepted October 30, 1996 Pt 0.012 20–90 La 0.049 10–55 Pr 0.020 2–10 Ti 0.15 20–100 246 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12

ISSN:0267-9477    

DOI:10.1039/a605159f

出版商:RSC    

年代:1997

数据来源: RSC

摘要:

Evaluation of a Direct Current Arc Charge Injection Device Spectrograph for Direct Analysis of Soils CYNTHIA A. MAHAN Chemical Science and T echnology Division, Inorganic T race Analysis Group, CST-9, MS G740, L os Alamos National L aboratory, L os Alamos, NM 87545, USA The feasibility of using a dc arc charge injection device (CID) trochemical techniques resulted in vast improvements in the analytical figures of merit over the dc arc methods. Although spectrograph for the semiquantitative determination of metals in soils has been demonstrated. Ar–O2 gas in a proportion of solution plasma techniques have replaced the dc arc, it remains a useful analytical tool and is employed primarily for the 70+30 was delivered through a Stallwood jet at 5 ml min-1 while 20 mg of sample were burned using 15 A of current and analysis of impurities in metals.Recent improvements in solid state multichannel detectors and the need for rapid, direct a constant 4 mm electrode gap.SPEX multi-element graphite powders were used to map wavelengths on the CID sensor and analytical methods has led to a renewed interest in the dc arc technique. The charge injection device (CID) detector rep- construct working curves. The CRM SO-1 was diluted with graphite powder and analysed. The analysis was accomplished resents relatively new technology and it is this device that has the potential to improve analytical performance over the using the full-frame capability of the CID, which permitted the evaluation of several analytical lines for the target elements.traditional dc arc spectroscopic technique.4,5 The CID is a two-dimensional solid state integrating detector where individ- Several elements, including Ag, Be, Cd, Hg, Sb and Se, were present in the diluted sample at less than 5 mg g-1, and results ual pixels (detector elements) are electronically and independently interrogated. They possess a high quantum efficiency corroborated these low concentrations. Quantitative results were obtained for P and Pb, semiquantitative results were over a wide range of wavelengths and are sensitive to UV radiation, where many metals of analytical importance emit.achieved for Ba, Mn and Sr, and results for B, Co, Cr, Cu, Ni, V and Zn were qualitative. Detection limits were The enhancement in UV sensitivity is one of the latest major improvements in the device that has captured the interest of determined by burning three samples of pure graphite powder.Detection limits for Ag, B, Ba, Be, Cd, Co, Hg, Mn and Ni atomic spectroscopists.4–11 The dc arc CID spectrograph combines the strengths of solid sampling with the quantitative were 2 mg g-1; for Cr, P, Pb, Sb, Se, Sr, V and Zn 5 mg g-1, and for As, Cu and Tl 10 mg g-1. The relative advantages of a multi-element solid state detection device. Potential analytical benefits over PMT-based ICP-OES instru- precision of the method was examined by burning three separate 33 mg g-1 multi-element SPEX standards.Most ments include full elemental fingerprinting of the sample, the ability to detect weak spectral lines in the midst of strong elements showed RSDs of less than 20% with the exception of As, Cd, Co, Ni, P and Zn. Ni and Zn showed the greatest matrix signals, simultaneous background correction, elimination of liquid wastes, and minimal sample preparation. imprecision at about 70% and 55% RSD, respectively.The use of Li2CO3 to buffer the arc temperature improved the precision This study evaluates the potential of a dc arc CID spectrograph for the analysis of soils. Instrumental parameters includ- for most elements, especially Ni and Zn. ing gas flow rate and applied current were investigated and Keywords: Direct current arc ; solids analysis; trace metal optimized. A semiquantitative analysis is described using analysis; charge injection device diluted graphite powdered standards to construct calibration graphs, and results for a CRM, SO-1, are presented.Detection limits and precision of the method are reported. The effect of The pioneering work of Goldschmidt, Mannkopff and colspectrochemical buffering on precision was also investigated. leagues at Go�ttingen in the early 1930s led to the widespread In addition, several lines were evaluated as potential internal use of dc arc emission spectroscopy for the analysis of soils, standards. ores, and plant and biological materials.Several excellent publications by Ahrens, Slavin, and Reeves and Brooks are a testament to the quality and amount of work that was achieved using the dc arc emission technique.1–3 These treatises contain EXPERIMENTAL comprehensive descriptions on the performance of the dc arc Instrumentation including volatilization behavior of the elements, specifics on photographic and PMT detection, research results on a variety A Thermo Jarrell Ash (Franklin, MA, USA) dc arc AtomComp 2000 spectrograph was used.A block diagram of the echelle of spectroscopic carriers and buffers, empirical studies on electrode configurations and gap widths, applied current and spectrometer is shown in Fig. 1. Source light enters through a 53 mmentrance aperture and then through ashutter mechanism voltage effects, the effect of various gases and gas mixtures on analytical results, background interference studies and corre- which controls the exposure time of each burn.The spectrograph consists of a 0.381 m echelle polychromator equipped sponding correction routines, etc. In short, the dc arc is a well characterized, mature emission spectrochemical technique. The with a 54.3 grooves mm-1 grating with a blaze angle of 45.5°. The quartz prism (17.5°) disperses wavelengths along the analytical shortcomings of the traditional dc arc are well documented and include narrow linear dynamic range, matrix horizontal axis and the echelle grating disperses the orders along the vertical axis, producing a two-dimensional spectrum interferences, and poor precision.2 The dc arc was the predominant analytical method for trace which is imaged on the CID 38 array detector.The sensor consists of a 512×512 matrix of pixels, each 28 mm2, resulting elements in geological materials until the emergence of atomic absorption and plasma methods in the 1960s and early 1970s. in 262 144 detector elements and a photoactive area of 14.3×14.3 mm.Dark current is minimized by cooling the The steady state emission signal achievable with plasma spec- Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (247–254) 247sensor so that higher orders and shorter wavelengths are on the bottom of the sensor. Wavelengths increase from right to left within the individual orders. The Na D lines, 589.592 and 588.995 nm {order 45}, are located in the upper left part of the image inside the green boxes, and the Be doublet is near the center of the image marked by the red boxes.The resonance Ba 455.403{58} nm line marked by the yellow box is located just below and to the right of the Na lines. Argon emission lines and faint cyanogen bands are evident across the upper portion of the CID image located between the Na lines and the Be doublet. Each desired wavelength must be calibrated (i.e., mapped) prior to analysis. Initial wavelength calibration involves profi- ling a pair of Hg lines, 253.652 and 312.566 nm, using a controlled exposure of a Hg pen lamp.The exact column and row position of the most intensely illuminated pixels are stored. All wavelengths are then referenced to these Hg coordinates. Fig. 1 Block diagram of AtomComp 2000 (courtesy Thermo Jarrell Therefore, any movement in the spectrum can be corrected by Ash). performing a simple Hg profile. Most elements were mapped after burning a SPEX 49 element series G7 blended graphite detector in a refrigerated housing to a temperature of -80 °C.powder admixture. However, in cases where the element lines Argon flow is maintained at 4 l min-1 through the polychrom- showed complex backgrounds or potential interfering lines, ator housing to prevent condensation from forming on the then single element standards (SPEX) were arced to ensure camera window. The optical block is maintained at a constant that the lines were mapped correctly. temperure (37.8±0.1°C) with the use of thermal blankets Signal processing of the emission intensities is accomplished (Watlow controller, Winona, MN, USA).The computer- by defining a sub-array of pixels which represents each wave- controlled constant current source can be set to deliver up to length, and referencing the position of the most intense pixel 36 A of current. The arc stand consists of a Stallwood jet within that sub-array to the Hg lines. Generally, the sub-array which provides a controlled laminar atmosphere surrounding readout widths are 15 pixels and heights are 3 pixels, although the discharge which helps to stabilize the arc and minimize these default parameters can be altered within certain guide- CN band interferences.The Stallwood jet and electrode jaws lines. The center 3×3 pixel array represents the analytical line are cooled using a WAC050 water recirculation unit. Signal with the most intense pixel centered within this sub-array. processing, wavelength calibration, imaging, analysis, and other Background correction points are chosen on an element by instrument and data collection and reduction operations are element inspection and are generally 2 pixels wide by 3 pixels performed using ThermoSPEC CID software.Instrument high. Fig. 3 shows a magnified image of the Cd 228.802 {115} specifications are given in Table 1. sub-array obtained by burning a 10 mg g-1 multi-element SPEX standard. There are two modes for signal processing, depending on Wavelength mapping and signal processing whether full or partial frame images are desired.A partial Spectral features are imaged on the CID sensor and represented image is acquired in analysis mode, whereas a full-frame image digitally in the form of an echellogram. Fig. 2 is a segment of is obtained in research mode. Full-frame images are generally a full-frame echellogram obtained by burning 20 mg of a used when qualitative fingerprinting of the sample is desired. 100 mg g-1 SPEX standard containing 49 elements. Several In this acquisition mode, every pixel is ‘turned on’. Charge is maps which serve to mark the location of elemental emissions collected at each detector based on the user-defined integration are shown. Echelle grating orders increase vertically on the time; thus, the exposure time for all pixels is identical. The Table 1 Instrument specifications Echelle grating: 54.3 grooves mm-1 Image: 151 Blaze angle: 45.5° f/l: 381 mm Prism: quartz (17.5°) Speed: f/8.5 Wavelength range: 190–800 nm Fig. 3 Magnified image of Cd 228.802 sub-array. The cursor (+) is positioned at column 237, row 309. Information from the cursor dialogue box includes the count: 1440 counts, raw: 702 counts, bkg: 65 counts, and net: 637 counts. The count represents the intensity (total counts) of the pixel where the cursor resides. The raw intensity is the average intensity (counts) of the 3×3 sub-array. The bkg is the average Fig. 2 Partial image of a full-frame echellogram using 100 mg g-1 signal of the background and the net intensity represents the background corrected average intensity of the 3×3 Cd sub-array. Intensity SPEX powder containing 49 elements; highlighting the Na D lines (589.592 and 588.995 nm) marked by green boxes, the Be doublet readings may be time-normalized so that images using different integration times can be compared. A two-dimensional intensity plot (313.107, 313.042)marked by red boxes, and the Ba 455.403wavelength marked by the yellow box.illustrates the entire 15×3 Cd sub-array. 248 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12collected charge on each pixel can only be read after the RESULTS AND DISCUSSION integration period, so random access integration (RAI) cannot Instrumental Optimization be used in research mode.8,9 Pixels are therefore subject to saturation. All experimental studies described here, except for Gas composition and flow rates the applied amperage optimization, were conducted using the Initially, a 90+10 Ar–O2 mixture was used for the arcing full-frame imaging capability of the CID sensor. environment.However, the sample to sample precision was In analysis mode, non-destructive readout (NDRO) and poor as a result of incomplete sample consumption. A 70+30 RAI are used to query the 3×3 sub-array pixels that define mixture of Ar–O2 was tested and resulted in improved pre- each analytical line. A queue of analytical lines is established cision and hence served as the sheath gas for the remainder of and cycled throughout the read process.Every pixel in the the study. The Ar–O2 gas flow rate through the Stallwood jet 3×3 array is read, and if the most intensely illuminated pixel was also investigated by monitoring the intensities of several reaches 75% of the full well capacity (5800 counts), then the elements as a function of flow rate.Fig. 4 is a plot of the entire sub-array (3×15) is read, cleared and cycled to the emission intensities for several elements at 3, 4, 5, 6 and bottom of the queue for additional signal collection until the 7 ml min-1. The optimum gas flow rate was 5 ml min-1 and user-defined integration times are met. The total signal is was used for the remainder of this study. the sum of the collective readings. If the most intensely illuminated pixel has not reached 75% of the full well capacity, then the analytical line is recycled to the bottom of the queue.Applied amperage This process continues until 75% of the full well capacity is The applied amperage was optimized by monitoring the emis- reached or the integration gate is closed. RAI and NDRO sion intensities while arcing 20 mg of a 100 mg g-1 multi- were used to collect the transient emission signals during the element SPEX mixture using 5, 9, 13, 15 and 17 A. Transient applied amperage optimization studies.emission signals were obtained for several elements that represented a range of volatilities including Hg, Ag, Cu, Cr, Fe and B. Plots of the first derivative of the signal intensity with time Reagents and Materials (dI/dt) versus time for these elements are shown in Fig. 5(a)–( f ). Commercially prepared diluted graphite powdered standards An x-axis break was inserted at 21 s in order to highlight the containing 49 elements (G7 series, SPEX Industries, Edison, initial part of the plot.The inset graphs show the line (square), NJ, USA) were used to map element lines, and construct the left (circle) and right (triangle) background intensities for each working curves used for the semiquantitative analysis. The element as a function of time using 15 A of applied current. same 49 element mix (L5 series, SPEX) in a lithium carbonate A complete burn of the sample is ideal in dc arc methods. matrix was used as a buffered standard to assess precision.In addition, the applied current should produce a smooth, Elemental composition of the powdered standards is given in reproducible arc with the analyte emission occurring over a Table 2. Single element standards (SPEX) were used to verify narrow time band and representing the total analyte concen- calibration of low intensity elements or elements with spectral tration. An applied current of 5 A produced a weak, wandering interferences. CRM SO-1 (Canada Centre for Mineral and arc around the lip of the anode cup.The emission profiles for Energy Technology) was also analysed. Pure graphite powder most elements using a 5 A current were very broad and (200 mesh) (Carbone of America, Ultra Carbon Division, Bay exhibited low intensities. Also, many of the elements produced City, MI, USA) was used as a standard blank and as a diluent emission signals late in the burn (i.e., after 21 s). Generally, the for samples and reference materials. The graphite electrodes arc produced sharper time-resolved signals and appeared more (Carbone) consisted of an undercut anode (ASTM Type S-15) stable with increasing current.which served as the sample container, and a pointed counter The integrated time-resolved signals for the same elements cathode electrode (ASTM Type C-1). at different current settings were not equivalent, with the 5 A Procedure The CRM was accurately weighed and diluted with graphite powder, in a ratio of 9.551, graphite5CRM, into a small polystyrene vial containing two methacrylate balls.The sample was subsequently blended for 2 min using a laboratory mill (Wig-L-Bug, SPEX Industries). Single element and Li2CO3 standards were mixed with graphite and processed as described above. Standards and samples were tightly packed into the anode. The packed anodes held 20.0±0.5 mg of the graphite admixtures as determined by statistical analysis of the mass of eight separately packed electrodes. Packed electrodes were set in the water-cooled rhodium jaws inside the dc arc box.A current of 5–36 A was passed between the two electrodes causing vaporization of the admixture into the high temperature arc column. Table 2 SPEX G7 series diluted powder standards Ag2O Al2O3 As2O3 H3BO3 BaCO3 BeO Bi2O3 NH4Br CaF2 CdO CeO2 NH4Cl Co2O3 Cr2O3 CsNO3 CuO Fe2O3 Ga2O3 GeO2 HgO NH4I In2O3 K2CO3 Li2CO3 MgO MnO2 MoO3 Na2CO3 Nb2O5 NiO NH4H2PO4 PbO RbCl Sb2O3 SeO2 SiO2 SnO2 SrCO3 Ta2O5 TeO2 ThO2 TiO2 Fig. 4 Emission intensities as a function of Ar–O2 gas flow through Tl2O3 U3O8 V2O5 WO3 ZnO ZrO2 the Stallwood jet. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 249Fig. 5 Derivative plots of intensities at 5 (+), 9 (#), 12 (Q), 15 (2) and 17 (+) A for (a) Hg 253, (b) Ag 328, (c) Cu 324, (d) Cr 425, (e) Fe 238 and ( f ) B 249. Inset graphs show the emission signal intensities of the line (&), and left (#) and right (+) background pixels.burns exhibiting the lowest total emission intensities. The produce the same sharp transient signals as the 15 A burn. This was due in part to the high energy which caused sample energy of the arc was apparently not sufficient to produce complete vaporization of the sample or excitation of the atoms sputtering from the anode cup at the start of the 17 A burn. In addition, these plots were acquired from single shots and and ions that were produced. There was a significant increase in total emission intensity when 9 A of current was used and incremental increases for the 13, 15 and 17 A runs.To verify differences in arc temperature, the emission intensities for the Ca I 422.672 nm line and the Ca II 396.846 nm line were measured at 5, 9 and 15 A using a 100 mg g-1 standard. The ratios of the atom-to-ion line intensities at the three current settings are shown in Fig. 6. The ratio decreases as the arc current increases showing an increase in concentration of ions relative to the concentration of atoms, at the higher arc temperatures produced by higher currents.Several of the transient signals shown in Fig. 5 show an increase in emission intensity around 43 s. This is due to the rim of the anode cup burning away and residual elements entering the arc column. There does not appear to be a concomitant increase in the background signals as evidenced Fig. 6 Intensity ratio of Ca I 422 to Ca II 396 nm lines for 5, 9 and 15 A using 100 mg g-1 SPEX standards.in the inset plots. For most elements, the 17 A current did not 250 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12therefore the sample-to-sample precision probably played a working curve comprised of only two points. A less intense line could easily have been substituted, but the more intense role in the observed differences. However, it is clear that both the 15 and 17 A currents produce optimum transient signals. line was used owing to the low concentration of Be present in the CRM.Although several different lines for most elements Therefore, a 15 A current setting was used for the remainder of this study. All of the elements studied thus far exhibit the were investigated, the wavelengths given in Table 3 represent lines that displayed the greatest sensitivity, fewest interferences same nominal appearance times owing to the effective buffering capacity of the graphite matrix. As expected, the more refrac- and produced the best analytical results. Detection limits (3s/m, n=3) are also reported in Table 3, tory elements were present in the arc for longer times.All elements that were studied were completely vaporized within and were determined by burning pure graphite powders using the same operating conditions as described for the calibration 20 s. These intensity versus time plots will be used to determine the optimum integration window for each element when a standards.The low levels or absence of several elements including Ag, As, Be, Cd, Sb, Se and Tl in the CRM has been general quantitative method is developed and performed using RAI in analysis mode. corroborated. Initially, the Mn 257.610 nm and Sr 407.771 nm lines were used to determine the CRM concentrations owing to the Semiquantitative Analysis sensitivity and lack of interferences associated with these lines. However, the center pixels for both of these sub-arrays pro- Several SPEX standards and a CRM (SO-1) were analysed duced nearly 8000 counts which is at the saturation level. The using the full-frame capability of the CID detector.The advanmeasured concentrations using these lines were 132 mg g-1 for tage of using research mode was that every potential target Mn and 88 mg g-1 for Sr. The full wavelength capability of analyte line could be examined for sensitivity and interferences. the CID permitted the use of the less sensitive lines reported In addition, background correction points were changed and in Table 3, and resulted in improved recoveries for both optimized using the same exposure.The disadvantage of using elements. this readout mode was that signal-to-noise ratios were not Most of the measured element concentrations in SO-1 were optimized, and elements that were present in high concenbiased high relative to the expected values. Interferences from trations, or very sensitive lines, were subject to saturation. the sample matrix, especially Fe, were thought to contribute Twenty milligrams of a graphite blank, 10 and 100 mg g-1 to these high biases.The concentrations of the matrix constitu- SPEX standards, and CRM (SO-1) were arced in an Ar–O2 ents in the 1+9 diluted SO-1 sample are given in Table 4. In atmosphere using a 15 A current and a constant 4 mmelectrode order to confirm the Fe interferences, a sample was prepared gap. The detector exposure time for all samples was 20 s.from a SPEX single element Fe standard, diluted with graphite ‘Typical’ emission lines for each of the elements were inspected and analysed using the same operating conditions as described on the full-frame echellograms to assess relative sensitivities, for the calibration standards. Fe showed a significant inter- interferences, and background structure. Favorable emission ference on B, Hg, Ni, Pb and Zn with the most severe lines were mapped according to the process described under interference on the Hg line.Fig. 7 is a bitmap image of the Hg Experimental. Background correction points were chosen by sub-array and illustrates the extent of interference from the Fe visual inspection of each element sub-array. 253.682 nm shoulder on the Hg line. The plots are presented A linear least-squares fitting routine was used to construct calibration graphs for Ag, As, B, Ba, Be, Cd, Co, Cr, Cu, Hg, Mn, Ni, P, Pb, Sb, Se, Sr, Tl, V and Zn. Element concentrations Table 4 Matrix concentration in SO-1 (1+9 dilution).Units in the CRM SO-1 were determined from these working curves are mg g-1 and are reported in Table 3, along with values for the y- Si 25700 Mg 2310 intercepts, slopes, and correlation coefficients. CRM results, as Al 9300 Na 1970 measured in the diluted sample, are given, and the certified Fe 6000 Ca 1800 values represent the dilution-corrected results. The Be channel K 2680 Ti 530 was saturated using the 100 mg g-1 standard; hence, the Table 3 CRM SO-1 (1+9 dilution).Units are mg g-1 Wavelength/nm Measured Certified Detection Fe Corrected {order} Element concentration concentration limits interference concentration Y -intercept/slope/R 328.068{80} Ag I <0.2 0.01 0.2 * <0.2 -0.06/1.508/1.0000 234.984{112} As I <7.0 0.2 7.0 * <7.0 -0.69/0.263/0.9987 249.678{105} B I 12 2.0 1.6 3 9 -1.4/0.847/0.9995 455.403{58} Ba II 134 92 2.0 * 134 -3.8/1.858/0.9992 313.042{84} Be II <0.1 0.2 0.1 * <0.1 † 214.438{122} Cd II <1.3 0.014 1.3 * <1.3 1.3/1.030/0.9997 345.350{76} Co I 9 3.4 2.0 * 9 0.19/0.226/0.9999 425.435{62} Cr I 48 17 2.7 * 48 -0.31/0.317/0.9998 327.396{81} Cu I 44 6.4 8.8 * 44 1.7/0.191/0.9858 253.652{104} Hg I 15 0.002 1.1 20 <1.1 0.49/0.509/0.9998 403.307{102} Mn I 113 93 0.8 * 113 0.44/2.597/1.0000 305.082{86} Ni I 46 9.9 1.7 5 41 -0.17/0.101/0.9995 213.618{123} P I 56 65 2.2 * 56 -1.6/0.747/0.9991 405.783{65} Pb I 9 2.2 5.0 7 2 -0.31/0.141/0.9991 217.581{121} Sb I <4.2 0.03 4.2 * <4.2 -1.5/0.478/0.9981 196.090{134} Se I <4.8 0.01 4.8 * <4.8 -0.92/0.464/0.9993 346.446{76} Sr II 56 34 4.2 * 56 -0.29/0.100/0.9984 535.046{49} Tl I <8.1 0.06 8.1 * <8.1 -0.66/0.346/0.9993 311.071{85} V II 58 15 3.7 * 58 -0.44/0.118/0.9974 206.200{127} Zn II 76 15 2.2 8 68 -0.48/0.727/0.9999 * Negligible effect.† Two-point standardization. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 251Fig. 7 Bitmap image of the Hg 253 nm line, CRM SO-1 (green), Fe standard (yellow), 100 mg g-1 SPEX multi-element standard (white), 10 mg g-1 (grey), graphite blank (red). Fig. 9 Cu 324 and 327 wavelengths with spline curve and linear leastsquares fit. as a spline fit of 15 data points. Each data point represents the summed intensities (counts s-1) of the three row pixels for each column in the 15 column sub-array. An Fe standard poor linear fits. These atomic lines have been reported to suffer from self-absorption.1 A spline curve drawn through the three (yellow) is shown as a reference.A concentration of 15 mg g-1 Hg was determined in the diluted SO-1 sample, although the calibration data points and a linear fit is shown in Fig. 9, and highlights the rollover evident at the high concentration level. certified value was well below the expected detection limit. Clearly, the Fe shoulder is the cause of the apparent Hg If Cu calibration graphs emulate the spline curve, the estimated concentrations would be 25 and 20 mg g-1 for the 324 and concentration. The normalized interfering Fe contribution on several elements, expressed as concentration, is also given in 327 nm lines, respectively.These results are a factor of 2 better than results using the linear fit curves. Self-absorption is Table 3. The apparent element concentrations were corrected by subtracting the Fe contribution from the observed concen- difficult to identify using the CID sensor because of the line location in conjunction with the finite size of the pixels.In a trations. The corrected Hg concentration was below the expected detection limit. Although detection limits may be quantitative analysis method, working curves constructed from several standards would help to identify the problem of self- improved through smaller dilutions, the matrix elements are then present in much higher concentrations, so interferences absorption. It is obvious that Cu determinations using these two lines will limit the linear dynamic range of the method.caused by matrix constituents are exacerbated. The measured concentrations for Cu, Cr, Ni, V and Zn in Several additional Cu lines including 213.598, 219.598, 223.008 and 224.700 nm were also investigated. However, these SO-1 were also biased high relative to the expected values, although Fe showed only minor contributions to these signals. lines either showed very low sensitivities, or exhibited highly structured backgrounds that were difficult to correct.Working Each elementsub-array was scrutinized, and additional element lines were examined to help determine the cause of the high curves established from these Cu lines did not improve the analytical results for SO-1. recoveries. Bitmap images of the element sub-arrays for Cu, Cr, Ni, V and Zn are shown in Fig. 8. Each element map was The Ni I 341.476, Ni I 349.296 and Ni I 305.082 nm lines were investigated.Working curves for all three lines showed verified using single element standards and all analytical lines were appropriately centered in the 15×3 sub-arrays. reasonable linearity and resulted in SO-1 concentrations in the diluted sample of 45, 41 and 46 mg g-1, respectively. Both the Cu 324 and 327 nm lines appeared free from spectral interferences and showed excellent signal-to-noise However, the expected concentration in the diluted sample was 10 mg g-1.Element sub-arrays for the three lines are ratios. However, the working curves for both lines resulted in Fig. 8 Bitmap images of Ni, Zn, Cu, Cr and V lines for SO-1 (green), 100 mg g-1(red), 10 mg g-1 (black), and graphite blank (blue). 252 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12shown in Fig. 8. Background elevation is evident on all three wavelengths for the CRM(green), with the most severe problem on the Ni 349 nm line.This line also shows significant structured background on both sides of the peak. None of the Ni lines is very sensitive. The 10 mg g-1 SPEX standard (black) is only slightly distinguishable from the blank using the 349 nm line, while the 305 and 341 nm lines are slightly more sensitive. Attempts to optimize background correction and pixel widths of the lines did not result in significant improvements in the Ni concentration. Visual inspection of the peak area of the Fig. 10 %RSDs obtained for 33 mg g-1 SPEX standard; n=3.CRM relative to the peak area of the 100 mg g-1 standard supports the calculated Ni concentrations and suggests that background correction is adequate. In addition, all three lines associated with most elements including Ni and Zn could not resulted in the same nominal concentrations. be improved using In as internal standard. The potential for Additional Cr and V lines were also investigated; however, using Ar 772.421, O 777.543 and CN 358.59 nm lines to there were no significant differences in the concentration values compensate for differences in the arcing environment between obtained.Bitmaps of the Cr and V sub-arrays for the lines the SPEX standards and the CRM SO-1 was investigated. The that were calibrated are also depicted in Fig. 8. RSDs of these lines were <5%, which were calculated from The measured Zn concentration in SO-1 was found to be the replicate analyses of the 33 mg g-1 multi-element SPEX nearly a factor of 5 higher than expected in the 1+9 diluted standard. The background-corrected average intensities for the CRM.Several additional Zn lines were evaluated in order to graphite blank, and the 10 and 100 mg g-1 SPEX standards help elucidate the problem. The Zn concentrations determined for the Ar, O and CN lines were 1086±181, 776±88 and from the 202.548, 206.200 and 213.856 nm lines after optimizing 2614±117 counts, respectively. The greatest variability was background corrections were 74, 75 and 253 mg g-1, respectobserved using the Ar line (16.7%), and the best precision was ively.The expected concentration for the sample was 15 observed using the CN line (4.5%). However, the SO-1 sample mg g-1 in the 1+9 diluted sample. Fig. 8 shows the three Zn showed significant deviations in line intensities at all three sub-arrays. Whereas the atom line appeared to be the line of wavelengths relative to the SPEX standards. The Ar, O and choice, the greatest analytical bias was observed using this CN line intensities for SO-1 were 140, 461 and 3270 counts, line. The center analytical pixel for the 213 nm line had a total respectively.Interestingly, the Ar and O atomic line intensities of 7700 counts which is near the saturation level, but the were reduced, whereas the CN emission intensity increased resulting error would have much less effect on the data than relative to the SPEX standards. The difference in arc tempera- what was found.tures between the standards and the sample was probably There was no attempt to buffer the temperature of the arc responsible for the observed differences in the emission intensit- for this analysis. A higher arc temperature would be expected ies. Clearly, analysis and internal standard line pairs need to using the predominately graphite standards, whereas a cooler react similarly under different arcing environments in order to arc temperature would result from the high concentration of achieve improved precision.The excitation potential of the matrix elements present in the CRM. Differences in arc temelements is probably the most important factor when choosing peratures between the standards and sample may have contribline pairs. uted to the enhanced atomic emission intensities for Cr, Cu, Results of the precision, internal standard and semiquantit- Ni and Zn. The lower arc temperature of the CRM would ative analysis studies suggest that differences in arc temperature lower the per cent.ionization of some elements, thus enhancing between standards and samples may be responsible for the the atomic lines. However, according to Saha’s equation, a poor analytical results for some elements. A 330 mg g-1 49 large difference in arc temperature would not have a very large element SPEX standard in an Li2CO3 matrix was ball-milled effect on the percentage of Zn ionized, owing to the high with graphite powder in a ratio of 159 to prepare a 33 mg g-1 ionization potential of Zn (9.4 V).Hence, it is not clear why multi-element standard. Three separate aliquots of the diluted both ion and atom lines resulted in concentrations above the standard were arced using the same conditions described for expected levels. For both V and Zn, the ion lines were calibrated the previous replicate study. Most elements showed improved in the high temperature graphite arc and therefore a lower precision except for Cd which actually degraded from 30 to concentration would be expected in the CRM owing to the 60%.The RSDs for Co, Ni, P and Zn were 11, 2.2, 14 and lower ionization potential of the arc. The opposite effect was actually observed. A potential problem of the dc arc technique is poor precision. To examine the relative precision of this study three separate 33 mg g-1 multi-element SPEX standards were arced using the same experimental parameters used in the previous analysis.The percent RSDs for the elements of interest are shown in Fig. 10. Ag, Ba, Be, Mn, Pb, Sb and Tl had RSDs 10% (33±3 mg g-1), and B, Cr, Cu, Hg, Se, Sr and V had RSDs 20% (33±7 mg g-1). As, Cd, Co and P had RSDs 40%. Ni and Zn showed the greatest imprecision at about 70 and 55% RSD, respectively. The satisfactory precision achieved for many of the elements in conjunction with acceptable recoveries suggest that the sample preparation steps are reproducible.The method of internal standardization is often used in spectrochemical analysis to improve the precision of the technique. Indium was present in the 33 mg g-1 multi-element SPEX standards at a concentration of 1000 mg g-1. The RSD for In (325.609 nm) was 2.0%, which was more precise than Fig. 11 Recovery biases (±%); error bars represent the uncertainty of the certified values at the 95% confidence level. for every element studied except for Be. Hence, the precision Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 25324%, respectively. Spectrochemical buffering appears to stabil- Dave Wayne and Dan Gerth at Los Alamos National Laboratory for assistance in preparing this manuscript. ize the arcing environment, improving the precision for the most elements. Analytical results can be defined as quantitative, within 10% REFERENCES of the concentration present; semiquantitative, within 50% of the concentration present; and qualitative, within a factor of 1 Ahrens, L.H., Spectrochemical Analysis, Addison Wesley, 20 of the concentration present. The biases, relative to the Cambridge, MA, 1950. certified concentrations, for the dilution-corrected CRM are 2 Slavin, M., in Chemical Analysis, ed. Elving, P. J., and Kolthoff, I. M., Wiley, New York, 1971, vol. 36. presented in Fig. 11. Quantitative results were obtained for P 3 Reeves, R. D., and Brooks, R. R., in Chemical Analysis, ed. Elving, and Pb, semiquantitative results were achieved for Ba, Mn P.J., Winefordner, J. D., and Kolthoff, I. M., Wiley, New York, and Sr, and results for B, Co, Cr, Cu, V and Zn were qualitative. 1978, vol. 51. 4 Pilon M. J., Belmore R., Schleicher, R. G., Fields, R. E., Norris, J. A., and Denton, M. B., in Proceedings of the International CONCLUSIONS Precious Metals Institute, March 10–12, 1991, New Orleans, L A, Full wavelength coverage of the CID permitted the investi- (1991), pp. 147–160. 5 Seely, L., CID vs Photographic Emulsion, US Geological Survey, gation of a multitude of elemental lines. The use of alternative National Research Program, Water Resource Division, Denver, lines improvedthe analytical results for Sr and Mn. In addition, CO. investigation of additional lines for Ni, Cr, V and Zn substan- 6 Sweedler, J. V., Bilhorn R. B., Epperson, P. M., Sims, G. R., and tiated the high biases found for these elements. The problem Denton, M. B., Anal. Chem., 1988, 60, 282A. was evidently not due to line selection, rather with varying arc 7 Epperson, P. M., Sweedler, J. V., Bilhorn, R. B., Sims, G. R., and temperatures. Denton, M. B., Anal. Chem., 1988, 60, 327A. 8 Bilhorn, R. B., and Denton, M. B., Appl. Spectrosc., 1989, 43, 1. This study successfully demonstrated the potential of the dc 9 Pilon, M. J., Denton, M. B., Schleicher, R. G., Moran, P. M., and arc CID system for the semiquantitative determination of Smith, S. B., Jr., Appl. Spectrosc., 1989, 44, 1613. inorganic metals in soils. A more rigorous analysis scheme 10 Barnard, T. W., Crockett, M. I., Ivaldi, J. C., Lundberg, P. L., employing spectroscopic buffering in conjunction with internal Yates, D. A., Levine, P. A., and Sauer, D. J., Anal. Chem., 1993, standardization, the use of RAI in analysis mode, optimized 65, 1231. integration times, and use of interelement correction factors 11 Barnard, T. W., Crockett, M. I., Ivaldi, J. C., and Lundberg, P. L., Anal. Chem., 1993, 65, 1225. will be pursued in the future, and should result in improved analytical performance. Paper 6/05049B Received July 22, 1996 I thank Michael Pilon and Charlie Hodges at Thermo Jarrell Ash for their technical assistance and helpful discussions. Also, Accepted October 2, 1996 254 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12

ISSN:0267-9477    

DOI:10.1039/a605049b

出版商:RSC    

年代:1997

数据来源: RSC