Analysis of soil and sediment samples by laser ablation inductively coupled plasma mass spectrometry Scott A. Baker,† Melody Bi, Ricardo Q. Aucelio, Benjamin W. Smith and James D. Winefordner* Department of Chemistry, University of Florida, Gainesville, FL 32611, USA Received 1st June 1998, Accepted 8th October 1998 The analysis of soil and sediment samples using laser ablation inductively coupled plasma mass spectrometry (LA-ICP-MS) was studied. Solution-based calibration was used for the quantification of trace elements in these samples. In most cases, the measured concentrations were within ±20% of the certified values using 60Ni or 107Ag as the internal standard.Measurements with Ag were carried out to investigate whether an internal standard could be spiked into soils for quantification purposes. The influence of particle size on the applicability of sample spiking was briefly studied, and it was demonstrated that particle size could significantly influence measurements if only the surface constituents of the particle were ablated.Use of 43Ca or 44Ca as an internal standard produced poorer results owing to interferences at these masses. In some cases, such as with Sr, Ba, Y and Rb, the measured concentrations were low by a factor of 2–3. This could be remedied by using one of these elements as an internal standard for the others. The eVects of elemental speciation, organic content and particle size were investigated.Elemental speciation and organic content of the soils did not appear to significantly aVect the LA-ICP-MS measurements. Particle size, however, was found to influence the precision and sensitivity of the measurements. Samples with smaller particle sizes yielded higher signal levels and better precision. The elemental analysis of soils is important for several reasons, matched standardization in the analysis of complex matrices, such as soils, but reported no data for quantitative measure- one of the most important being the identification of contaminants and the establishment of levels of toxic elements present ments in soil samples.As an alternative to direct analysis of soils, fused mixtures of soil and diluent have been employed in the soil. A complicating factor in the determination of elemental constituents in soils is the refractory nature of these for LA-ICP-MS analyses.14 The production of fused beads oVers improved matrix matching; however, a limitation of this materials.1 Traditional methods of analysis rely on decomposition of the soil, typically through microwave or acid diges- methodology is the loss of volatile species as a result of the high fusion temperatures and increased sample preparation tion, and analysis of the resulting solution by flame or furnace AAS2 or ICP-AES.2,3 Direct methods of analysis are advanta- times. This paper reports on the use of LA-ICP-MS for the analysis geous because of the elimination of time-consuming sample preparation steps and the risk of sample contamination from of soil and sediment samples.Solution-based calibration was investigated for obtaining accurate quantitative measurements chemical reagents used in the digestion process. Several solid sampling techniques have been utilized for analyzing soil and in these complex matrices. The influences of sampling strategy, internal standards, instrumental parameters, sample particle sediment samples, including ETV-ICP-MS,4,5 GDMS,6 dc Arc-AES7 and XRF.2,8 XRF is commonly used to analyze sizes and composition were studied in order to characterize the strengths and limitation of this approach.soils and other environmental samples; however, the technique lacks the sensitivity to determine many low-level species of interest. Experimental Recently, laser sampling techniques have gained popularity Instrumentation for analyzing soil samples, since little to no sample preparation is required.Laser induced breakdown spectroscopy (LIBS) is The experiments were performed with a Finnigan MAT (San well suited for the field based analysis of soil samples, as Jose, CA, USA) SOLA ICP-MS and Finnigan MAT System evidenced by the number of recent publications on this appli- 266 laser ablation accessory. Typical operating conditions for cation.9–11 The technique, however, suVers from more severe the ICP-MS are given in Table 1. In these studies, a combined matrix eVects and poorer sensitivity than LA-ICP-MS.flow of nebulized solution and carrier gas from the ablation Relatively few applications of LA-ICP-MS for analyzing soils chamber was introduced to the ICP-MS at all times.15–17 This have been reported.12–14 Durrant and Ward12 analyzed several allowed the use of solutions in optimizing the ICP-MS Chinese reference soils by LA-ICP-MS using elemental operating conditions and provided a convenient means of responses derived from a single soil in the series and reported calibration (discussed later).that determined concentrations for about 80% of the elements The laser used was an Nd5YAG with an output beam at studied were within a factor of two of concentrations measured 266 nm. It was operated at 5 Hz with typical pulse energies of by INAA, with many being significantly closer. HoVmann 0.7 mJ and pulse widths of 10 ns. The System 266 ablation et al.13 discussed the importance of internal standardization accessory was modified by using a separate computer to and the grinding of soil samples for achieving acceptable levels control the x–y–z translation stage. This allowed for transof precision.They also mentioned the importance of matrix- lation of the sample at approximately 15 mm s-1 while the laser was repetitively fired. Single spot and translation sampling were both used in this work and a comparison of the two will †Present address: United States Department of Agriculture, be made.Typical analysis times were 60–80 s (signals measured Agricultural Research Service, Beltsville Human Nutrition Research Center, Food Composition Laboratory, Beltsville, MD 20705, USA. over 300–400 laser shots). J. Anal. At. Spectrom., 1999, 14, 19–26 19Table 1 Typical ICP-MS operating conditions of particle sizes, sand (about 10 g) was ground in an alumina grinding vial (Spex Model 8003) for 1–2 min. The ground Rf power 1200 W material was then transferred to a sieve and four particle size Coolant gas flow rate 15 L min-1 fractions (<35, 35–60, 60–80 and >80 mm) were obtained.Auxiliary gas flow rate 0.9 L min-1 Samples of 1 g of the diVerent particle size fractions of sand Nebulizer gas flow rate 0.7 L min-1 Ablation chamber flow rate 0.3 L min-1 were placed in glass vials, spiked with solutions containing Solution uptake rate 1.0 mL min-1 Co, Ag, Y, Rb, W, Ba and Pb, and dried at 110 °C for several hours. The sand samples were then mixed with 10% (m/m) Scan conditions— Faraday scans (major elements)— cellulose binder (Spex CertiPrep) for 30 min and pressed into Scan range per isotope 1 u pellets for analysis.Number of passes 32 To determine whether diVerences in measured composition, Number of channels per u 16 precision and ablation yield were observed for diVerent particle Dwell time per channel 64 ms size fractions, NIST SRM 2704 and HPS Sandy Soil B samples Multiplier scans (minor and trace elements)— were sieved into two diVerent size fractions (<35 and Scan range per isotope 0.25 u 35–60 mm).Samples of 1 g of each fraction were pressed into Number of passes 128 pellets for analysis. Number of channels per u 16 ms The eVect of the organic content of the sample on the Dwell time per channel 2–4 ms LA-ICP-MS results was studied by adding cellulose binder to NIST SRM 2709 soil in diVerent proportions (10 and 20% Solution-based calibration m/m). The 1 g soil and binder samples were mixed for 30 min and then pressed into pellets.Additionally, an inorganic sample The solution-based calibration approach for quantitative (NIST SRM 1633 Coal Fly Ash) was studied for comparison. measurements with LA-ICP-MS, which utilizes dual sample With the coal fly ash reference material, a sample mixed with introduction or a ‘wet plasma’ configuration, has been cellulose binder (10% m/m) was prepared, and also one mixed described previously.15–17 Briefly, it involved measuring the with high-purity graphite (20% m/m).The 1 g samples were intensity of one or several isotopes for each element and an pressed into pellets at 35 MPa. internal standard and determining relative sensitivity factors Because elements exist in a variety of forms (e.g., carbonates, (RSFs) for each analyte. The RSFs (analyte/internal standard) silicates, oxides) in soils, the eVect of speciation on the measured from standard solutions were then used to determine LA-ICP-MS results was studied.Samples containing various the concentration of analytes in the solid samples based on compounds of Ba were prepared in a sand matrix. Compounds the expression of Ba (nitrate, oxide, chloride, carbonate, and sulfate) (Aldrich, Milwaukee, WI, USA) were added to 10 g of sand to produce concentration (analyte)solid= intensity (analyte)solid intensity (internal standard)solid samples with a total Ba concentration of about 750 ppm.In addition, Ni (as sulfate) was added at the same concentration (about 500 ppm) in each sample for use as an internal standard. × concentration (internal standard)solid RSFsolution Each of the mixtures was ground and homogenized in an alumina grinding vial for 30 min. A 0.9 g portion of the sample The success of this methodology depended on achieving identwas then mixed with 0.1 g of cellulose binder for 30 min before ical plasma conditions whether an ablated solid or nebulized being pressed into a pellet.To study the matrix dependence of solution was being analyzed, and that the ablated material speciation eVects, Ba (and Ni as internal standard) samples was representative of the bulk solid sample. were prepared in a graphite matrix. Direct grinding and mixing of the Ba and Ni compounds in graphite, however, did not Samples produce homogeneous samples. It was necessary to prepare concentrated mixtures (10% m/m) of Ba and Ni in sand and Several soil and sediment standard reference materials were analyzed to determine the applicability of solutions for cali- then dilute this mixture in graphite to give final concentrations of about 750 ppm for both analytes.bration of these materials. These materials included NIST SRM 2704 (BuValo River Sediment), NIST SRM 2709 (San Joaquin Soil ), NIST SRM 2710 (Montana Soil ), NIST SRM Results 2711 (Montana Soil ), High Purity Standards (HPS) (Charleston, SC, USA) Sandy Soil B and HPS Loam A.For Single spot versus translational sampling analyses, the samples were pressed into pellets without binder at a pressure of 35 MPa. For the determination of RSFs, Single spot and translational sampling were studied for the analysis of soils and sediments. Signals were typically a factor solutions were prepared by dilution of a 10 ppm multi-element standard (High Purity Standards) with de-ionized water and of 4–5 larger when the sample was translated. This signal enhancement was greater than what has been observed in Optima HNO3 (Fisher Scientific, St.Louis, MO, USA) to a final acid concentration of 2%. work involving glass and ceramic samples, most likely because a deep crater was formed more quickly in these compacted A limitation of the use of solutions for calibration was that a knowledge of the concentration of the internal standard in particulate samples. A comparison of the results obtained with both sampling strategies is shown in Fig. 1 for NIST SRM the solid was required. To address this limitation, Ag was spiked into the soils as an internal standard. This was done 2704. These plots represent the analyte responses for (a) Ni and (b) Ba relative to Ca from a total of 25 scans obtained by adding 2.5 mL of a 100 ppm Ag standard to 1 g of soil (250 ppm Ag in soil ) and drying the samples at 110 °C for from 25 diVerent spots on the sample, or at diVerent lines (five scans per line) produced from translation of the sample.several hours. The soil was then transferred into a plastic vial and mixed with a Spex (Metuchen, NJ, USA) Mixer/Mill Copper and cobalt were also studied and produced similar results. The relative analyte responses from both sampling Model 8000 for 15 min to ensure homogeneity. The influence of particle size on the applicability of sample methods were similar in most cases; however, significant deviations from the mean (as much as a factor of five) were spiking was investigated by spiking sand samples (Mallinckrodt Baker, Paris, KY, USA) of diVerent particle observed in some scans when sampling at a fixed location.This must have resulted from local inhomogeneity in the soil size ranges with a multi-element solution. To generate a range 20 J. Anal. At. Spectrom., 1999, 14, 19–26have been related to diVerences in the particle size distribution for the NIST SRM 2704 sample compared with the other soils. The eVects of particle size on measurement precision will be discussed later.Selection of the internal standard was found to be an important factor in obtaining accurate results for soils. Initially, Ca was used as the internal standard because it was present at significant levels in all of the samples and possessed minor isotopes that could be measured with the multiplier detector. In addition, it has been successfully used in this laboratory for the analysis of glass samples by LA-ICP-MS with solution-based calibration. This was not the case with the soil samples.A comparison of the results for several elements (V, Co, Ni, Cu, Zn, Sr, Ag, Ba and Pb) in (a) NIST SRM 2709 and (b) NIST SRM 2704 soils using 43Ca as the internal standard is shown in Fig. 2. These plots have been normalized to the certified or information values; therefore, the accuracy of the solution calibration method can be directly assessed by comparing the normalized value with unity (a value of unity would signify that the measured and certified concentrations were identical ).For NIST SRM 2709, the measured concentrations of V, Co, Ni, Cu, and Zn were between 40 and 60% higher than the certified values for these elements, those for Ba and Sr were about 60% lower than the certified values and that for Pb was within 10% of the certified value. For NIST SRM 2704, diVerent results were obtained. Fig. 1 Relative analyte signals for (a) Ni and (b) Ba from NIST SRM There was a systematic increase in the measured concentration 2704 sediment sample using both single spot (A) and translational for all of the analytes studied in this sample.Systematic (B) sampling. changes were also observed for the other soils. Because errors in the measured concentrations were largely systematic, rather than random, it is believed that they were the result of an samples, and clearly indicated the need to sample a large interference by aluminum oxide at m/z 43.Based on these enough portion of the soil to ensure accurate and precise observations, 43Ca was deemed a poor choice for internal measurements. Measurement of these local inhomogeneities is standardization since the levels of aluminum in the sample required for accurate measurement of the bulk composition; aVected the signal at this mass. The use of 44Ca was also however, their presence can significantly aVect the precision investigated for internal standardization. Similar systematic of analyses.From the results in Fig. 1, it appeared that diVerences in the measurement accuracy were also observed translation of the sample eVectively averaged out the inhomowith this isotope, most likely resulting from interferences due geneities during a typical 1 min analysis, which should result in more precise measurements. The absence of large deviations in the relative analyte signal (analyte/Ca) was almost certainly due to the larger mass ablated with sample translation.The average precision (RSD) of the relative intensities for the four analytes studied was 8.8% with translation sampling and 37% with single spot sampling. Since higher sensitivity and better precision were obtained with translation of the sample, this mode of sampling was used for all subsequent measurements. Solution-based calibration Solution-based calibration required the selection of an appropriate internal standard with a known concentration in the sample.Because of inhomogeneity in the soil samples, measurement of a minor isotope of a major matrix constituent (Ca) was initially used to provide an acceptable level of precision. Using an isotope of Ca (43Ca or 44Ca), which was present at levels of 1–3% in the soils studied, as the internal standard resulted in typical precision values (RSD, n=10) of <10%. The precision values were not consistent among the soils analyzed. Measurements on the NIST SRM 2704 sample exhibited the poorest level of precision, with RSD values consistently around 10–12%.With the other soils, the RSD was typically around 6–7%. Measurement precision is obviously dependent on the concentration of analyte in the sample and the amount of material ablated; however, this could not explain the poorer precision observed with the NIST SRM 2704 sample. This sample contained similar levels of trace Fig. 2 Results for the analysis of NIST SRM 2709 soil and NIST elements and Ca as the other samples.In addition, a similar SRM 2704 sediment samples using 43Ca as the internal standard. mass (about 50 ng per shot) was ablated for all of the soils Measured concentrations for analytes have been normalized by their certified or information values. studied. The diVerences in precision that were observed might J. Anal. At. Spectrom., 1999, 14, 19–26 21to SiO+ or CO2+ at m/z 44. Using an isotope of silicon (28Si, 29Si or 30Si) as the internal standard worked well for the analysis of minor elements in soil; however, none of the silicon isotopes could be measured on the multiplier detector, which was required for measuring trace elements, because of saturation of the detector at these masses.The inability to measure both large (>106 counts s-1) and small (<106 counts s-1) ion signals during a measurement is a limitation of the present detection system, which employs a dual detector (electron multiplier and Faraday cup) design in which only one detector can be selected at a time.It is important to mention that this limitation is due to the present ICP-MS instrumentation and that silicon isotopes for internal standardization could be used with other ICP-MS systems. Because of the interferences on Ca isotopes and the lack of any other suitable internal standard that could be used over the whole series of soils, a minor sample constituent was investigated as an internal standard. The use of trace elements as internal standards produced more accurate measurements for the soil samples.This is illustrated in Table 2, which compares the measured concentrations for several analytes in NIST SRM 2704, 2709 and 2711 soils using the solution calibration method with both 43Ca and 60Ni as internal standards. The RSF values using Ni as the internal standard were determined by dividing the analyte/Ca RSF by the Ni/Ca RSF. Significantly more accurate results were obtained using Ni as the internal standard, indicating that solution calibration was useful even for a complex matrix like soil.A limitation of the technique, however, was that the concentration of an element in the sample that could serve as the internal standard had to be known. This could be addressed by using a complementary technique to determine the concentration of an element in the sample or by spiking in a known amount of some element that was not in the sample at an appreciable concentration relative to the amount added.A suitable element for spiking could be chosen by doing a survey scan over the sample of interest. The utility of sample spiking for quantification purposes was investigated by adding Ag (initially present at levels of <5 ppm) to NIST SRM 2709 and HPS Loam A soils to produce final concentrations of 250 ppm. Results from the use of solution-based calibration with Ag as the internal standard are provided in Table 3 and demonstrate that reasonable accuracy could be obtained in most cases.The potential influence of particle size on the applicability of sample spiking will be discussed later. For the analytes studied, consistent patterns concerning the accuracy of solution calibration were observed. Most analytes (V, Cr, Mn, Co, Ni, Cu and Zn) could be determined with reasonable accuracy (typically ±20%) using a single solution for calibration; however, the measured concentrations for several analytes (Rb, Sr, Ba and Y) were consistently lower (by around a factor of 2–3) than their certified values in the soils.More accurate measurements of the latter elements could be obtained if one element in this group was used as the internal standard for the other elements. The low results obtained for Rb, Sr, Ba and Y might be partially due to matrix eVects resulting from high levels of eYciently ionized elements (EIEs), such as Na and Al. EIEs generally cause a decrease in ICP-MS intensities, and are most severe for elements with low ionization energies.18 Since Rb, Sr, Ba and Y possessed the lowest ionization energies of the elements studied, the EIE eVect could help explain the reduced sensitivity observed for these elements in soil and sediment matrices.In general, matrix eVects in ICP-MS are diYcult to measure and quantify. They can often be minimized for particular analytes through optimization of the ICP-MS operating conditions or selection of appropriate internal standards;19 however, for multi-element determinations, this is often not feasible Table 2 Measured concentrations (ppm) in soils using solution calibration with 43Ca or 60Ni as the internal standard Soil V Cr Co Cu Zn Sr Ag Ba Pb NIST SRM 2704 Measureda (43Ca) 210±36 326±48 33.6±5.2 249±31 940±140 119±12 351±48 570±100 Measureda (60Ni) 87±15 135±22 13.9±2.6 103±16 390±72 49.0±8.2 146±34 235±58 Certifiedb 95±4 135±5 14.0±0.6 98.6±5.0 438±12 (130) 414±12 161±17 NIST SRM 2709 Measureda (43Ca) 160±12 146±12 21.6±1.3 52.3±3.9 150±9 107±5 0.39±0.10 369±27 21.1±4.2 Measureda (60Ni) 116±10 106±11 15.6±1.4 37.9±3.8 109±9 77.5±5.8 0.28±0.08 268±23 15.3±3.2 Certified 112±5 130±4 13.4±0.7 34.6±0.7 106±3 231±2 0.41±0.03 968±40 18.9±0.5 NIST SRM 2711 Measureda (43Ca) 106±9 59.8±6.7 18.4±2.3 182±18 471±56 139±9 7.2±1.3 439±40 3150±210 Measureda (60Ni) 57.8±4.7 32.5±4.8 9.2±1.0 110±16 255±38 75.6±5.6 3.9±0.8 240±30 1570±140 Certifiedb 81.6±2.9 (47) (10) 114±2 350.4±4.8 245.3±0.7 4.63±0.09 726±38 1162±31 aConfidence intervals at the 95% level, n=5 measurements.bValues in parentheses are non-certified information values provided by NIST. since all elements do not behave identically in the ICP-MS. 22 J. Anal. At. Spectrom., 1999, 14, 19–26Table 3 Measured concentrations (ppm) for analytes in NIST SRM 2709 and HPS Loam A soil samples using Ag as the internal standard. Ag was spiked into each of the soils at 250 ppm Sample Co Ni Cu Zn Sr Ba NIST SRM 2709 Measureda 12.0±0.5 73.0±4.9 33.2±1.0 103±4 80.9±4.2 371±34 Certified 13.4±0.7 88±5 34.6±0.7 106±3 231±2 968±40 HPS Loam A Measureda 14.7±2.7 21.4±2.1 19.6±2.4 20.6±3.5 164±26 Certified 14.4±0.4 20.9±1.0 12.4±1.4 60.3±2.6 448±28 aConfidence intervals at the 95% level, n=5 measurements. Therefore, improvements with respect to one analyte may adversely aVect the measurement of another analyte.In the analysis of steels by solution nebulization ICP-MS, Vaughan and Horlick20 reported that matrix eVects could be minimized by a slight decrease in the nebulizer flow rate corresponding to the maximum analyte signal.This strategy was investigated in this work to determine whether more accurate results could be obtained for Rb, Sr, Y and Ba. Decreasing the nebulizer flow rate from 0.7 L min-1, the flow that produced the highest sensitivity, aVected the RSFs for both nebulized solutions and laser ablated solids in a similar manner (Fig. 3). Both (a) Co, an element for which relatively accurate results could be obtained, and (b) Sr are included for comparison. Several other analytes were studied and the same trends were observed. In Fig. 3, the data have been normalized to the solutiondetermined RSF at 0.7 L min-1. Results from these measurements indicated that adjustment of the nebulizer flow rate could not be used to improve the accuracy of the measurements, since the analytes were aVected similarly, regardless of sample introduction method or matrix (soil or 2% HNO3).Similarly, adjustment of the ICP rf power was investigated to determine if this variable could be selected to improve the accuracy of measurements for all analytes in the soils. The results for (a) Co and (b) Sr are presented in Fig. 4. In these plots, Ni was used as the internal standard and the RSF values were normalized to the solution-based RSF value at 1200 W Fig. 4 EVect of rf power on relative sensitivity factors (RSFs) for (a) Co and (b) Sr with solution nebulization of a 10 ppb solution (B) and laser ablation (A) of NIST SRM 2709 soil.RSF values have been normalized to the solution-based RSF at 1200 W, which represented the rf power producing maximum analyte signals. (optimum sensitivity) of rf power. Changing the rf power aVected the RSF values slightly; however, the changes were similar for both the solution and the laser ablated NIST SRM 2709 soil sample. Therefore, no improvement in the accuracy of the measurements was observed as the rf power was changed over the range 1050–1350 W.Although accurate measurements were not obtained in all cases, solution calibration provided a relatively simple means of estimating analyte concentrations in the sample. In all cases, the measured analyte concentrations were within a factor of three of their certified concentrations using 60Ni as the internal standard and, in most cases, much better accuracy was achieved.Lower concentrations were consistently measured for Rb, Sr, Y and Ba; however, the use of one of these elements as an internal standard for the rest of the group produced more accurate results. To investigate potential sources of inaccuracy in soil samples, element speciation, sample organic content and particle size eVects were studied. Speciation eVects Since metals exist as a variety of compounds in soil samples, Fig. 3 EVect of nebulizer flow rate on relative sensitivity factors the eVect of speciation on LA-ICP-MS measurements was (RSFs) for (a) Co and (b) Sr with solution nebulization of a 10 ppb examined.In a study involving the detection of Ba and Pb in solution (B) and laser ablation (A) of NIST SRM 2709 soil. RSF soils by LIBS, it was determined that the chemical form (oxide, values have been normalized to the solution-based RSF at 0.7 L min-1, carbonate, sulfate, chloride, nitrate) of the analyte aVected the which represented the nebulizer flow rate providing maximum analyte signals.sensitivity of LIBS measurements.11 To determine if this might J. Anal. At. Spectrom., 1999, 14, 19–26 23help explain the reduced sensitivities observed for Ba, Sr, Rb the level corresponding to the maximum for a 1% HNO3 solution. and Y in this work, the eVect of speciation on Ba in a sand matrix was examined. The results for these studies are provided Results for the analysis of NIST SRM 2709 soil which contained 1.2% carbon, with no binder added, 10% cellulose in Table 4.The first column of data compares the RSF (Ba/Ni) values obtained from the five sand samples. The BaO sample binder added and 20% cellulose binder added were compared (Table 5). No significant changes in the measured analyte produced a lower value (statistically significant at the 95% confidence level ); however, the decrease was minor in compari- concentrations were observed with the addition of cellulose binder at levels up to 20%.Similarly, a coal fly ash sample son with the factor of 2–3 lower concentration measured with solution calibration. To determine if the observed decrease was studied to determine if diVerent results were obtained when 10% cellulose or 20% graphite was used to prepare the could be attributed to the formation of Ba2+ or BaO+, the signals for these isotopes were included in determining RSFs sample pellet. The results for these analyses (Table 6) indicated that the addition of an inorganic (graphite) or organic (cellu- (column 2 in Table 4).The same trend was observed; BaO produced a slightly lower RSF than the other compounds. lose) binder produced similar results. The accuracy of these measurements, based on solution calibration with Ni (98 ppm The RSF values given in the last column of Table 4 were measured from the sand pellets by solution nebulization to in sample) as the internal standard, was around 10–20% for all analytes studied.ensure that the samples contained the same levels of Ba and Ni. For the solution measurements, a 5 mg portion of each sand sample was diluted in 20 mL of 5% HNO3 and then Analysis of particle size fractions heated to dissolve the Ba and Ni species. The results indicated Two diVerent particle size fractions (<35 and 35–60 mm) of that the pellets contained reproducible levels of Ba and Ni. NIST SRM 2704 and HPS Sandy Soil B were analyzed to The sulfate pellet was not included because BaSO4 was insoldetermine whether analytes were distributed similarly in both uble in the dilute HNO3 solution.fractions, and to determine how particle size aVected the The eVects of speciation were also examined in a graphite precision of LA-ICP-MS measurements when analyzing par- matrix. Only the oxide, sulfate and carbonate forms of Ba ticulate samples. The results for NIST SRM 2704 indicated were studied, with Ni (as sulfate) as the internal standard.No that the concentrations of the eight trace elements studied significant diVerences between the RSFs (Ba/Ni) were were identical, within the experimental uncertainty, in both observed. The RSF (Ba/Ni) values were 1.45±0.19 for the particle size fractions. The ablated mass was slightly higher oxide, 1.48±0.25 for the sulfate and 1.54±0.27 for the carbon- (about 20%) for the pellet consisting of particles less than ate. Based on speciation studies with Ba in both sand and 35 mm compared with the 35–60 mm pellet.Also, the precision graphite matrices, it appeared that the form of the analyte did of the measurements was significantly better for the smaller not significantly aVect the accuracy of LA-ICP-MS particle sample (average RSD=7.3%) than the larger particle measurements. sample (average RSD=14%). The unsieved sample was analyzed for comparison, and the precision of measurements on EVect of organic content this sample was intermediate between the two fractions, with an average RSD of 10%.The original material consisted of To determine whether the organic content of the sample aVected the measurements, both soil and coal fly ash samples about 75–80% (by mass) of particles less than 35 mm and the remainder in the 35–60 mm range. were studied. The organic content was studied not only to determine if this might help explain the observed discrepancies The HPS Sandy Soil B sample produced diVerent results in terms of particle composition.The results indicated that the for Ba, Sr, Rb and Y, but also to determine if the addition of cellulose binder, used in pellet preparation, aVected the accu- smaller particles (<35 mm) had approximately three times more Pb and Ag than the larger particles (35–60 mm). All of racy of results. Previous studies using solution nebulization ICP-MS have indicated that the addition of small amounts of the other elements studied contained identical levels of trace elements.These results are not easily explained, but might be organic solvents could either increase or decrease analyte signals. Allain et al.21 found that the addition of glycerol or related to the preparation of this standard reference material. In this soil, the concentrations of most trace elements were methane significantly enhanced the signals for some analytes (As, Au, Se, Te and Hg), whereas other analytes were essen- enriched 10–100-fold by spraying the sample with an aerosol mist.The samples were then dried, ground, sieved and blended. tially unaVected. These workers attributed the signal enhancements with addition of carbon to a modification of ionization If, for instance, the elements were added sequentially rather than simultaneously, preferential adsorption of Pb and Ag on equilibrium over a limited energy range (9–11 eV). The ionization energy of carbon (11.2 eV) is slightly above this range. the smaller particles might have occurred.The behavior of these samples, in terms of measurement precision, was similar Longerich22 reported ICP-MS signal suppression for several elements when various organic solvents were added to a 1% to that observed with the NIST SRM 2704 sample. Measurements on the smaller particle sample were charac- HNO3 solution. The decrease in sensitivity could be recovered, however, when the nebulizer gas flow rate was reduced from terized by higher precision (average RSD=5.5%) than the Table 4 Measured relative sensitivity factors (RSFs) for Ba species in a sand matrix using Ni as the internal standard.Ba and Ni concentrations in the sand were approximately 750 and 500 ppm, respectively LA-ICP-MS ICP-MS: Ba species RSF (Ba/Ni)a,b RSF (Batotal/Ni)a,c RSF (Ba/Ni)a,d Oxide 0.89±0.09 1.18±0.10 1.66±0.02 Nitrate 1.12±0.16 1.64±0.17 1.70±0.02 Carbonate 1.14±0.07 1.50±0.09 1.68±0.04 Chloride 1.33±0.09 1.75±0.09 1.56±0.01 Sulfate 1.19±0.05 1.58±0.05 aConfidence intervals at the 95% level, n=5 measurements.bLaser ablation results using Ba+ signal only. cLaser ablation results using Ba+, Ba2+ and BaO+ signals. dSolution results using Ba+ signal only. 24 J. Anal. At. Spectrom., 1999, 14, 19–26Table 5 Measured relative sensitivity factors (RSFs) for analytes in NIST SRM 2709 with and without the addition of cellulose binder RSFa V Cr Mn Co Ni Cu Zn Rb Ba No cellulose 32.8±2.7 3.24±0.30 35.6±4.1 36.5±2.1 7.04±0.36 16.4±0.9 3.48±0.17 11.6±0.9 1.41±0.16 10% cellulose 34.3±1.0 3.45±0.38 35.9±1.9 37.4±1.0 7.01±0.25 16.7±0.8 3.51±0.25 12.1±0.7 1.40±0.11 20% cellulose 36.2±1.4 3.55±0.26 36.3±2.3 40.4±3.6 7.69±0.50 18.6±1.4 3.73±0.33 14.0±0.6 1.39±0.08 aConfidence intervals at 95% level, n=5 measurements, 43Ca used as the internal standard.Table 6 Measured concentrations (ppm) from NIST SRM 1633 (Coal Fly Ash) using solution calibration with graphite and cellulose as binders V Cr Mn Cu Zn Cd Certified 214±8 131±2 493±7 128±5 210±20 1.45±0.06 Measureda (20% graphite) 239±12 153±14 571±21 143±3 223±9 1.65±0.27 Measureda (10% cellulose) 244±9 160±8 579±43 153±10 235±15 1.83±0.26 aConfidence intervals at the 95% level, n=5 measurements. larger particle sample (average RSD=12%).The RSD of the was close to the maximum output of the laser system. In addition, particle sizes at which preferential vaporization measurements for the unsieved sample was, on average, 7.1%. Based on these studies, the precision of measurements was became significant would depend on the type of sample being analyzed.For more refractory materials, diVerences in relative significantly aVected by sample particle size. Samples consisting of larger particles produced poorer precision and lower analyte signals would be expected to occur at even smaller particle sizes. Based on the observed diVerences in relative ablation yields than samples consisting of smaller particles.To investigate whether particle size influenced the relative analyte signals for the spiked sand samples, it was concluded that particle sizes in the sample could significantly aVect the signals for analytes deposited on the surface of particles with respect to an analyte in the particle, several spiked sand accuracy of LA-ICP-MS results when using sample spiking to introduce an internal standard into the sample. If the particles samples were analyzed. The same amount of each analyte was added to four diVerent sized fractions of sand (<35, 35–60, are too large to be eYciently ablated, only the surface concentration will be measured and the use of a spiked internal 60–80 and >80 mm).All analyte signals were normalized to 44Ca, which was present in the sand matrix. This was necessary standard will lead to systematically low results for elements present in the bulk particles. because analyte signals from the pellet consisting of particles <35 mm were more than twice as large as those obtained from the pellet containing particles >80 mm because of the larger Conclusions ablation yield.The results (Fig. 5) indicated that the relative Reasonably accurate (±20%) results for most trace elements signals for all analytes increased significantly for particles in soil and sediment samples could be obtained using a single larger than 35 mm. This was due to the ineYciency of the solution standard for calibration. Using a spiked internal ablation process at the laser energies used (0.7 mJ) in vaporizstandard was demonstrated to be useful for particulate samples ing larger particles of sand. The deposited analytes were and enhances the utility of the LA-ICP-MS technique, since concentrated on the surface of the particles; therefore, they potential internal standards can be rapidly identified with a were preferentially vaporized relative to the bulk of the particle survey scan over the sample of interest. When using a spiked and produced higher relative intensities.The onset of this element as the internal standard, care must be taken to ensure process should be shifted to larger particle diameters as the that particle sizes are suYciently small so that they are laser energy is increased; however, the energy used (0.7 mJ) eYciently ablated relative to the deposited spike. In the analysis of soils and sediment, systematically low results were obtained for Rb, Sr, Y and Ba when using Ni or Ag as the internal standard.This was probably due, in part, to matrix eVects associated with large amounts of easily ionizable elements present in the soils. The eVects of elemental speciation and sample organic content were investigated and no significant diVerences in the accuracy of the measurements were observed as a result of these factors. Particle size was demonstrated to aVect the precision of measurements and influenced the ablation yield. Measurements performed on samples composed of smaller particles were more precise and produced higher signal levels.This work was supported by an Air Force OYce of Scientific Research–University Research Initiative (AFOSR–URI) grant (F49620–93–1-0349). References 1 J. C. Van Loon, Selected Methods of Trace Metal Analysis: Biological and Environmental Samples, Wiley, New York, 1985. 2 S. M. Pyle, J. M. Nocerino, S. N. Deming, J. A. Palasota, J. M. Fig. 5 EVect of particle size on relative analyte signals for a sand Palasota, E. L. Miller, D. C. Hillman, C. A. Kuharic, W. H. Cole, P. M. Fitzpatrick, M. A. Watson and K. D. Nichols, Environ. Sci. matrix. Analytes were spiked into the sand as solutions and 44Ca was used as the internal standard. Technol., 1996, 30, 204. J. Anal. At. Spectrom., 1999, 14, 19–26 253 C. J. Warren, B. Xing and M. J. Dudas, Can. J. Soil. Sci., 1990, 13 E. HoVmann, C. Lu�dke and H. Stephanowitz, Fresenius’ J. Anal. Chem., 1996, 355, 900. 70, 617. 14 X. Guo and F. E. Lichte, Analyst, 1995, 120, 2707. 4 G. Galbacs, F. Vanhaecke, L. Moens and R. Dams, Microchem. 15 S. Chenery and J. M. Cook, J. Anal. At. Specom., 1993, 8, 299. J., 1996, 54, 272. 16 E. F. Cromwell and P. Arrowsmith, Anal. Chem., 1995, 67, 131. 5 M. J. Liaw and S. J. Jiang, J. Anal. At. Spectrom., 1996, 11, 555. 17 S. A. Baker, M. J. Dellavecchia, B. W. Smith and J. D. 6 J. Teng, C. M. Barshick, D. C. Duckworth, S. J. Morton, D. H. Winefordner, Anal. Chim. Acta, 1997, 355, 113. Smith and F. L. King, Appl. Spectrosc., 1995, 49, 1361. 18 J. W. Olesik, Anal. Chem., 1996, 68, 469A. 7 C. A.Mahan, J. Anal. At. Spectrom., 1997, 12, 247. 19 Handbook of Inductively Coupled Plasma Mass Spectrometry, ed. 8 S. J. Goldstein, Environ. Sci. Technol., 1996, 30, 2318. K. E. Jarvis, A. L. Gray and R. S. Houk, Blackie, Glasgow, 1992. 9 R. Barbini, F. Colao, R. Fantoni, A. Palucci, S. Ribezzo, H. J. L. 20 M.-A. Vaughan and G. Horlick, J. Anal. At. Spectrom., 1989, van der Steen and M. Angelone, Appl. Phys. B, 1997, 65, 101. 4, 45. 10 K. Y. Yamamoto, D. A. Cremers, M. J. Ferris and L. E. Foster, 21 P. Allain, L. Jaunault, Y. Mauras, J.-M. Mermet and T. Appl. Spectrosc., 1996, 50, 222. Delaporte, Anal. Chem., 1991, 63, 1497. 11 A. S. Eppler, D. A. Cremers, D. D. Hickmott, F. J. Ferris and 22 H. P. Longerich, J. Anal. At. Spectrom., 1989, 4, 665. A. C. Koskelo, Appl. Spectrosc., 1996, 50, 1175. 12 S. F. Durrant and N. I. Ward, Fresenius’ J. Anal. Chem., 1993, 345, 512. Paper 8/04060E 26 J. Anal. At. Spectrom., 1999, 14, 19–26