Universal calibration for analysis of organic solutions of medium and low volatility by inductively coupled plasma-atomic emission spectrometry† Assad A. Al-Ammar, Rajesh K. Gupta and Ramon M. Barnes* Department of Chemistry, Lederle Graduate Research Center Towers, University of Massachusetts, Box 34510, Amherst, MA 01003–4510, USA Received 29th October 1998, Accepted 26th February 1999 A novel chemometric technique is described to facilitate the use of a single organic solvent matrix standard to calibrate inductively coupled plasma atomic emission spectrometry (ICP-AES) for the accurate determination of trace elements in another organic solvent or complex mixture of several organic solvents. Analysis errors arising from the diVerence in solvent matrix vapor loading and chemical composition between standard and sample are corrected.The technique is intended for organic solvent mixtures for which the plasma vapor loading is tolerable with conventional sample introduction techniques (i.e., conventional nebulizer–spray chamber arrangement without a solvent desolvator).Simultaneous measurement of atomic and ionic spectral lines of the same analyte is required. A correction factor is estimated from its linear correlation with the line intensity ratio. For elements without a sensitive atom line, the correction is estimated from the correlation between their ionic line intensities and the ratio of ionic-to-atomic line intensities of another analyte in the same sample.Experimental tests with seven trace elements (Al, Be, Ca, Cu, Fe, Mg, and Mn) in diVerent organic solvent mixtures (xylene, dichloromethane, hexane, carbon disulfide, acetone and 1,2,3,4-tetrahydronaphthalene) demonstrate the eVectiveness of universal calibration. The determination of trace elements in hydrocarbons is import- ated with DIN. Furthermore, a major disadvantage of DIN is the plasma cooling that occurs when organic solutions are ant in the petroleum and petrochemical industry.Elements that are aggressive catalysts, poisons or corrosive in the nebulized. This degrades the detection limit for most elements by as much as an order of magnitude compared with conven- cracking furnace are harmful even at mg L-1 levels owing to the large volumes of hydrocarbons processed. Also, shipments tional sample introduction methods.6 The aim of this investigation was to demonstrate universal of purchased feed or outgoing products frequently require monitoring of trace contaminants.However, the determination calibration with a conventional nebulizer–spray chamber without desolvation. The nebulizer–spray chamber arrangement, of trace elements in refinery and chemical plant streams is complicated by the continuous variability in composition and in contrast to the use of a desolvator, permits the analysis of volatile analyte species. However, the organic liquids that can volatility, which makes the use of a single organic solvent matrix standard for universal calibration impractical.be analyzed are limited to those that can be tolerated by the plasma (i.e., organic compounds with medium or low vola- Universal calibration is defined as using standards prepared in a solvent to measure samples in diVerent solvent matrices tility). Nevertheless, these compounds constitute a large class encountered in the petroleum and petrochemical industry. than the calibration standard.Various approaches have been applied to achieve universal Furthermore, petroleum compound mixtures with medium or low volatility containing small and varying percentages of calibration. Botto and Zhu1,2 used a membrane desolvator to strip organic solvent from the aerosol with an eYciency of highly volatile compounds also can be analyzed. A chemometric technique was developed to correct for the 99.9%. Cryogenic desolvation at temperatures below -80 °C also was used to remove organic vapor.3 Since these measurement errors arising from the diVerence in matrix loading, chemical composition and nebulization eYciency approaches are based on solvent stripping, volatile analyte compounds, which usually form the largest fraction of the between the sample and standard matrices.Universal calibration can be achieved. total analyte concentration, also are lost. Consequently, solvent removal methods are of limited practical use. Botto and Zhu1 also evaluated the influence of chemical composition on solvent matrix loading.They discovered that aromatic com- Theory and derivation pounds suppressed analyte signals more than aliphatic com- DiVerences between sample and calibration solution matrices pounds. Chlorinated and oxygenated organic compounds can cause variations in transport eYciencies and plasma matrix suppressed analyte signals less than simple hydrocarbons of loading. A large error can be expected in estimating the analyte comparable volatilities.Direct injection nebulization (DIN) concentrations. Accordingly, to make universal calibration also can be applied for universal calibration, because it eliminpossible a correction must be estimated for each analyte in ates the variation in matrix loading that originates from the every sample matrix. The analyte signal measured for a sample diVerence in volatility.4,5 However, matrix loading that results must be multiplied by a corresponding correction factor to from the diVerence in chemical composition cannot be eliminconvert it to the value expected if the sample matrix were the same as the standard. The corrected signal can then be used †Presented in part at the Fifth Rio Symposium on Atomic to obtain the correct concentration by reference to the cali- Spectrometry, Cancu� n, Mexico, October 4–10, 1998, and the 1999 bration standard. Thus, the correction factor eliminates the European Winter Conference on Plasma Spectrochemistry, Pau, France, January 10–15, 1999.diVerence in matrix loading and transport eYciency between J. Anal. At. Spectrom., 1999, 14, 801–807 801sample and standard. The aim of this work was to develop a correction factor with the measured spectral lines intensities. This approach was used previously to develop a chemometric chemometric method to establish a mathematical relationship that links the correction factor to matrix loading and transport technique called common analyte internal standardization (CAIS).It was used to correct for the signal drift9 and eYciency. This relationship must be expressed as variables that can be measured, for example, by inductively coupled inorganic matrix eVect (NaCl, H2SO4, HNO3, etc.) in ICPAES. 8 plasma atomic emission spectrometry (ICP-AES). Since only spectral line intensities are measured in AES, this dependence Vapor loading should use spectral line intensities as variables. The diVerence in transport eYciency has been corrected From the diVerence in the magnitude of intensity depression with an internal reference element.1 For this purpose, one (or enhancement) between an atomic and ionic line of the internal reference element is suYcient to correct for all the same element, a mathematical function that relates vapor analytes in a sample.2 The transport eYciency can be corrected loading to a correction factor can be developed starting from by multiplying the analyte intensity by the ratio of intensities eqn.(1). of the internal reference in the standard and the sample. The For an ionic line correction should be applied after correcting the intensities of the internal reference and the analyte for matrix loading. The Ii=Ii0 exp(Ai-k1iC1-k2iC2-…-kniCn) (2) problem that remains to be solved in this study is to correct and for an atomic line for matrix loading by finding an algebraic relationship between matrix loading and a correction factor.Ia=Ia0 exp(Aa-k1aC1-k2aC2-…-knaCn) (3) Dividing eqn. (2) by eqn. (3) and rearranging terms yields Organic matrix loading (Ia0/Ii0)(Ii/Ia)=exp[Ai-Aa+C1(k1a-k1i)+…+Cn(kn a-kni)] To find the correction function, a mathematical analysis should (4) be based on an exact algebraic representation of the analyte intensity, I, and the matrix concentration, C. The eVect of From eqn. (2) inorganic matrices (such as, NaCl, Cl2) on analyte intensity Ii/Ii0=exp(Ai-k1iC1-k2iC2-…-kniCn) (5) in ICP-AES resembles an exponential decay curve.7 Correspondingly, we postulate that organic matrix vapor After subtracting eqn.(4) from eqn. (5), Ii0/Ii can be expressed loading also depresses the analyte intensity according to an as exponential decay curve given by the equation Iio/Ii=(Ia0/Ii0)(Ii/Ia)+exp[-(Ai-k1iC1-k2iC2-…-kniCn)] I=I 0 exp(A-k1C1-k2C2-...-knCn) (1) -exp[Ai-Aa+C1(k1a-k1i)+…+Cn(kna-kni)] (6) Eqn. (1) describes the situation where the organic matrix of Eqn.(6) is useful because it indicates that the correction the universal calibration solution diVers from the organic factor Ii0/Ii can be calculated from the experimentally deter- matrix of the sample. The sample matrix (major matrix) may mined ratio Ii/Ia. Another helpful version of eqn. (6) can be exist in pure form or mixed with small concentrations, C1, obtained by dividing eqn. (3) by eqn. (2) followed by sub- C2, …, Cn, of other organic solvents (minor matrices) that are tracting the resulting equation from eqn.(5) and rearranging: more volatile than the sample matrix. An example is premium grade gasoline containing approximately 1% of pentanes and Ii0/Ii=(Ii0/Ia0)(Ia/Ii)+exp[-(Ai-k1iC1-k2iC2-…-kniCn)] butanes. In eqn. (1), I0 represents the intensity of the analyte -exp[Aa-Ai+C1(k1i-k1a)+…+Cn(kni-kna)] (7) dissolved in the organic matrix of the universal calibration solution. I represents the intensity of the same concentration The ratio Ii/Ia [eqn.(6)] or its inverse, Ia/Ii [eqn. (7)] is not of the analyte dissolved in sample matrix (major matrix), aVected by the change in transport eYciency, since atomic and which contains concentrations C1, C2, …, Cn of other volatile ionic lines are aVected by transport eYciency to the same organic compounds. If C1, C2, …, Cn are zero, then I= extent. Therefore, these ratios are true specific measures of I0 expA represents the intensity of the analyte dissolved in matrix vapor loading.pure sample matrix. A and k1, k2, …, kn are constants related Although eqns. (6) and (7) are similar, the results of a to the organic compounds in the sample. numerical simulation and experiments indicate that these two The validity of eqn. (1) was tested experimentally. The equations behave diVerently. The plots of correction factor eVects of small additions of CH2Cl2 to a m-xylene matrix and Ii0/Ii against Ii/Ia [eqn. (6)] or Ia/Ii [eqn. (7)] give straight lines.hexane to a 1,2,3,4-tetrahydronaphthalene matrix were meas- However, the linear regression correlation coeYcient was ured for several atomic and ionic lines of Be, Ca, Cu, Fe, Mg higher with eqn. (7) than with eqn. (6). Accordingly, the plot and Mn. Small hexane or CH2Cl2 concentration additions of the correction factor against Ia/Ii is preferred to correct for (0–6%) do not aVect the transport eYciency of the matrix. matrix vapor loading. However, a marked increase in vapor loading is expected, because of the high volatility of the additives compared with Application the xylene or tetrahydronaphthalene matrices.The measure- Analysis method. An analytical scheme was developed to ments were followed by evaluating the applicability of the correct for the matrix vapor loading, so that universal cali- exponential decay of eqn. (1). Straight lines were obtained for bration can be applied. The approach employs several control all additives studied with correlation coeYcients 0.98, in solutions and one universal calibration solution.The control excellent agreement with the logarithmic form of eqn. (1): solutions and the calibration solution are prepared from the ln I=ln I 0+A-k1C1-k2C2-…-knCn (1¾) same organic matrix and contain exactly the same concentrations of analytes. In contrast to the calibration solution, Furthermore, the magnitude of signal suppression was found to be related to the ionization and excitation potential as the control solutions contain, in addition to the major matrix, varying percentages of another organic minor matrix that has reported earlier for an inorganic matrix eVect.7,8 Atomic lines are less depressed than ionic lines.This diVerence in behavior higher volatility compared to the major matrix (e.g., hexane, chloroform). Variable amounts of the minor matrix are added between the atomic and ionic lines toward organic matrix vapor loading is used as a key factor in the next section to to the control solutions to create changeable high vapor loadings.However, the amounts of the minor matrix added develop an exact algebraic relationship applied to estimate the 802 J. Anal. At. Spectrom., 1999, 14, 801–807should be small, so that the transport eYciency is unchanged atomic-to-ionic line combination can be used as internal reference. When no fitting analyte can be found in the sample, for all the control solutions and equal to that of the calibration solution. Therefore, the eVect of only vapor loading is meas- an appropriate amount of the internal reference element is added to the sample.However, in contrast to the conventional ured and corrected. The control solutions, calibration solution, samples and method of internal reference addition, the amount added need not be exact, because only the ratio Ia/Ii needs to be measured. blank are measured under the same experimental conditions. From the measurement results for the control solutions and This ratio is independent of the amount added.Also, in contrast to the conventional method of internal reference the calibration solution, a straight-line correction curve is constructed by plotting the vapor loading correction factor, addition, only one internal reference is needed for all the elements in the sample. Ii0/Ii, against Ia/Ii of eqn. (7). Ii0 is the intensity of the analyte in the calibration solution as defined by eqn. (1).The measured value of Ia/Ii for the sample is then used to estimate the vapor Experimental loading correction factor for the sample with aid of the constructed correction curve. The sample correction coeYcient Instrumentation is then multiplied by the measured sample intensity, Ii, to A commercial ICP-AES system (Optima 3000, Perkin-Elmer, transform it to Ii0, which corresponds to the predicted intensity Norwalk, CT, USA) was used for all the experiments.The of the analyte in the sample if it exists in the matrix of the experimental operating parameters (Table 1) were selected calibration solution. The corrected intensity is then used to based on the optimization reported in the manufacturer’s calculate the vapor loading-corrected concentration by referinstrument manual. Experimental wavelengths are listed in ence to the calibration solution. This concentration is further Table 2. corrected for transport eYciency, if the sample and the calibration solution transport eYciencies are diVerent from each Table 1 Operating parameters for the ICP-AES measurements other, by using an internal reference as mentioned at the beginning of the section.ICP system Optima 3000 prototype Rf power 1.3 kW Frequency (free running) 40 MHz Analyte as internal reference. This scheme applies to the ICP torch Type 2 quartz slotted extension analytes with a pair of sensitive atomic and ionic lines, which Torch injector Ceramic alumina is true for most elements.However, for some elements either Outer argon flow rate 15 L min-1 only an atomic line or ionic line can be found with appropriate Intermediate argon flow 2.0 L min-1 sensitivity. In this situation obtaining the correction curve for rate the analyte is possible by using the ratio Ia/Ii, of an other Observation height 15 mm Central argon flow rate 0.8 L min-1 analyte in the sample. This approach, which specifies an Nebulizer Concentric glass (Glass Expansion, analyte to act as a surrogate internal reference, seems to be Hawthorn, Victoria, Australia, reasonable, because the ratio Ia/Ii is a general measure for Model 38493) vapor loading no matter to which analyte it belongs.Sample pump rate 0.8 mL min-1 To apply this concept for an analyte with only ionic lines Sample pump tubing Viton, orange–orange (0.035 in id) of suitable intensity, a mathematical correlation should first Spray chamber Glass Scott double pass coolant jacketed (Spectro Analytical Instruments, be found to relate the analyte correction factor, Iiom/Iim to the Fitchburg, MA, USA) internal reference correction factor, Iior/Iir. This correlation Spray chamber temperature 0 °C can be obtained from the equation Integration time (auto) 10 s Background correction ±0.04 nm (Iior-Iir)/Iir=(Iiom-Iim)/Iim (8) Drain Pumped Eqn.(8) is the equation for the conventional method of internal standardization in ICP-AES. As recognized by several Table 2 Analyte and emission wavelengths investigators,10,11 eqn.(8) lacks general applicability. In this mathematical derivation a more general form is Element Wavelength/nma (Iior-Iir)/Iir=H(Iiom-Iim)/Iim (9) Al I 309.271 where H is a constant the magnitude of which depends on the Be I 234.861, II 313.042 Ca II 393.366, II 396.847 nature of the analyte and the internal reference. Cu I 324.754, I 327.396 From eqns. (7) and (9), the equation Fe II 259.940 Iiom/Iim={(Iior/Iaor)(Iar/Iir)+exp[-(Air-k1irC1r-k2irC2r Mg II 279.553, I 285.213 Mn II 257.610 -…-knirCnr)] -exp[Aar-Air+C1r(k1ir-k1ar)+… (10) aI are atom lines and II are ion lines.+Cnr(knir-knar)]}-1+H] can be derived. Numerical simulations based on eqn. (10) Table 3 EVect of 2% m/v of various volatile compounds in a xylene indicate that the plots of the correction factor for the matrix on the analyte line signal intensity analyte, m, against the ratio, Iar/Iir, for the internal reference Signal change (%) element, r, provide straight lines.Hence the same correction scheme based on eqn. (7) can be applied. Analyte and CH2Cl2 Hexane CS2 Acetone If the analyte has only atomic lines of proper sensitivity, an wavelength/nm (bp 40 °C) (bp 69 °C) (bp 46 °C) (bp 56 °C) equation similar to eqn. (10) can be derived that correlates the analyte correction coeYcient measured by using atomic Be I 234.861 -17 -19 +23 -2 Be II 313.042 -36 -45 +13 -19 lines with the atomic-to-ionic line ratio of the internal refer- Mg I 285.213 -4 +3 +44 +11 ence element.Mg II 279.553 -50 -51 +24 -21 In the modified correction scheme based on eqn. (10), Fe II 259.940 -42 -46 +35 -18 adding an internal reference element to the sample is not Ca II 393.366 -43 -41 +42 -13 necessary. Any other analyte in the samples with suitable J. Anal. At. Spectrom., 1999, 14, 801–807 803Table 4 Experimental values of A in xylene and k in 2% m/v of various volatile compounds in xylene k Analyte and wavelength/nm A CH2Cl2 Hexane CS2 Acetone Be I 234.861 -0.39 0.099 0.11 -0.108 0.01 Be II 313.042 -0.94 0.23 0.30 -0.048 0.12 Mg I 285.213 -0.05 0.011 -0.011 -0.19 -0.054 Mg II 279.553 -0.92 0.35 0.36 -0.10 0.124 Fe II 259.940 -0.91 0.28 0.31 -0.144 0.10 Ca II 393.366 -0.66 0.28 0.26 -0.17 0.081 Simulation values— Atom Aa=-0.387 ka=0.04 Ion Ai=-0.94 ki=0.1 Fig. 1 Log(signal intensity) for the Mg 279 nm line as a function of Fig. 3 Numerical simulation of correction factor, Iio/Ii, as a function percentage of dichloromethane.of Ia/Ii. Fig. 2 Log(signal intensity) for the Al 309 nm line as a function of percentage of hexane. Fig. 4 Correction factor Ii0/Ii for Be as a function of Ia/Ii for Be. The six points correspond to Be in pure xylene and 2% hexane, 4% hexane, 2% CS2, 2% acetone and 2% CH2Cl2, in xylene. Error bars=2%. Reagents Commercial atomic emission calibration solutions prepared from a non-volatile metal organic salt in hydrocarbon oil act as the volatile minor matrix.Three control solutions were prepared containing 4, 8 and 12% m/m hexane. Several samples (VHG Laboratories, Manchester, NH; 100 mg L-1) were used to prepare three types of test solutions: sample, calibration containing 4 or 6 mg L-1 of each element were prepared in m-xylene (bp 139 °C, 99% purity; Aldrich) as a major matrix and control. Each of the control solutions contained a mixture of Be, Mg, Ca, Fe,Mn, Cu and Al.The control and calibration containing 2% m/m of various highly volatile organic compounds to act as the minor volatile matrix. These volatile solutions contained 4.0 mg L-1 of each element. The calibration solution was prepared in a 1,2,3,4-tetra- minor matrices were CS2 (bp 46 °C), hexane, acetone (bp 56 °C) and CH2Cl2 (bp 40 °C). hydronaphthalene matrix (tetralin, bp 217 °C, 98% purity; Aldrich, Milwaukee, WI, USA). The control solutions were To match the diVerence in transport eYciency of the calibration solution and the sample, each sample and the cali- prepared in 1,2,3,4-tetrahydronaphthalene as the major matrix to which hexane (bp 69 °C) was added in various amounts to bration solution contained 4 mg L-1 indium as an internal 804 J.Anal. At. Spectrom., 1999, 14, 801–807Verification The exponential decay of eqn. (1) was tested under the experimental and instrumental arrangements given in Table 1 that are typical of those widely used for routine analysis.The eVect of 0–5.8% CH2Cl2 in m-xylene and 0–12% hexane in 1,2,3,4-tetrahydronaphthalene was measured for several ionic and atomic lines and results plotted as ln I against the CH2Cl2 or hexane additive. Numerical simulation Although eqns. (6) and (7) closely resemble each other, their behaviors were tested by numerical simulation under conditions that simulate typical operations used in routine analysis. In this numerical test the sample was taken to be an organic matrix that diVers in volatility from that of the calibration solution.The sample matrix was assumed to contain various concentrations (from 0 to 5%) of a minor organic Fig. 5 Correction factor Ii0/Ii for Mg as a function of Ia/Ii for Mg. matrix that has a volatility much higher than that of the major The six points correspond to pure xylene and 2% hexane, 4% hexane, matrix. In this simulation, constant A and k values were 2% CS2, 2% acetone and 2% CH2Cl2 in xylene.Error bars=2%. experimentally determined (Table 3) when the exponential decay behavior of the organic matrix was examined (Ai, Aa, ka and ki in Table 4). Also, the ionic spectral lines were assumed to be more aVected by the matrix than atomic lines corresponding to experimental observations. Results and discussion The method was tested by measuring trace concentrations of Al, Be, Ca, Cu, Fe, Mg and Mn in simulated samples prepared from organic matrices that diVer in chemical composition, volatility, surface tension and viscosity from the matrix of the calibration solution.Also, several highly volatile organic compounds were mixed with the major sample matrix to simulate the composition of petroleum compounds that include medium and low volatility products containing small but variable percentages of highly volatile organic compounds (e.g., pentanes and butanes). The presence of a variable composition of highly volatile organic compounds causes changeable vapor loadings and a universal standardization strategy is therefore Fig. 6 Correction factor Ii0/Ii for Mg as a function of Ia/Ii for necessary. The highly volatile organic compounds used in this Mg. The four points correspond to Mg in pure 1,2,3,4- study diVered not only in volatility but also in chemical tetrahydronaphthalene and in 4, 8 and 12% hexane in 1,2,3,4- composition, thus testing the ability of the proposed method tetrahydronaphthalene. Error bars=2%.to correct for both vapor loading and chemical composition. Exponential response reference. Pure xylene and 1,2,3,4-tetrahydronaphthalene were measured as blanks. The eVect of highly volatile compound additions on analyte The eVect of 0–5.8% CH2Cl2 in m-xylene matrix was meas- (Be, Fe, Mg, Ca) signals is reported in Table 3 for four ured for several ionic and atomic lines of 4 mg L-1 Be, Ca, additives to a xylene matrix. The suppression (or enhancement) Cu, Fe, Mg and Mn (Table 2) to test eqn.(1). The experiment is less for atomic than ionic lines. The applicability of the was repeated using 0–12% hexane in a 1,2,3,4-tetraexponential decay of eqn. (1) was evaluated by plotting ln I hydronaphthalene matrix. against the percentage of CH2Cl2 or hexane additive. Examples selected from the straight line plots obtained are given in Method Fig. 1 for Mg with CH2Cl2 and Fig. 2 for Al with hexane in xylene. From these plots, values of A and k were calculated, The instrument was first stabilized for drift by allowing it to run for 1 h, after which the samples, calibration solutions, as illustrated for xylene in Table 4.The A and k values of Be, Ca, Fe and Mg ion lines are similar but significantly diVerent blanks and control solutions were measured. The procedure for matrix correction was then followed in exactly the way from those of Be and Mg atom lines. The A and k values for atomic and ionic lines (Ai, Aa, ka, and ki Table 4) used for discussed in the Theory and derivation section.The elements measured were subdivided into three groups simulations were estimated from these data. The magnitude of signal suppression (or enhancement) is according to the way in which the correction experiment was conducted. The first group, Mg and Be, was corrected by related to the ionization and excitation potential as reported for an inorganic matrix eVect.7,8 Atomic lines are less depressed using their own atomic/ionic line intensities.The second (Fe, Mn and Ca) and third (Al and Cu) groups were corrected by than ionic lines. This diVerence in behavior between atomic and ionic lines toward the organic matrix vapor loading eVect using the ratio Ia/Ii of Mg (i.e., Mg was regarded as the internal reference) to simulate elements that have no atomic is pivotal in developing the relationship applied to estimate the correction factors for measured spectral line intensities. or ionic lines of proper sensitivity, since the only commercial calibration solutions available for this study contained no These results also indicate the importance of the chemical composition of the matrix on the degree of depression or such elements.J. Anal. At. Spectrom., 1999, 14, 801–807 805Table 5 Trace element concentrations in xylene and in volatile organic–xylene mixtures using standards in a 1,2,3,4-tetrahydronaphthalene matrix Concentration/mg L-1 Standard Mg II 279.553 Be II 313.042 Fe II 259.940 deviation Solution Concentration nm nm nm Mean of mean error Xylene (pure) Expected 6.00 6.00 6.00 Uncorrected 2.50 3.20 3.72 Error (%) -58 -47 -38 -48 10 Corrected 6.00 5.70 6.10 Error (%) 0.0 -5.0 1.7 -1.1 3.5 2% CS2 in xylene Expected 4.00 4.00 4.00 Uncorrected 0.80 0.94 0.93 Error (%) -80 -77 -77 -78 2.0 Corrected 4.20 3.80 4.20 Error (%) 5.0 -5.0 5.0 1.7 5.8 2% acetone in xylene Expected 4.00 4.00 4.00 Uncorrected 0.98 1.15 1.22 Error (%) -76 -71 -70 -72 3.1 Corrected 3.80 4.10 3.70 Error (%) -5.0 2.5 -7.5 -3.3 5.2 2% CH2Cl2 in xylene Expected 4.00 4.00 4.00 Uncorrected 0.98 1.05 1.08 Error (%) -76 -74 -73 -74 1.3 Corrected 4.10 4.30 4.20 Error (%) 2.5 7.5 5.0 5.0 2.5 2% hexane in xylene Expected 4.00 4.00 4.00 Uncorrected 0.84 0.98 0.95 Error (%) -79 -76 -76 -77 1.8 Corrected 3.90 4.00 4.30 Error (%) -2.5 0.0 7.5 1.7 5.2 Ia/Ii ratio (nm/nm) Mg 285/Mg 279 Be 234/Be 313 Mg 285/Mg 279 Number of replicates=3.Standard deviation ranged from 1 to 3%.Table 6 Trace element concentrations in xylene using calibration solutions in a 1,2,3,4-tetrahydronaphthalene matrix Concentration/mg L-1 Concentration/mg L-1 Ia/Ii Element and ratio wavelength/nm Expected Uncorrected Error (%) Corrected Error (%) (nm/nm) Mg II 279.553 4.0 2.0 -50.0 4.4 10.0 Mg 285/Mg 279 Be II 313.042 4.0 2.5 -37.5 4.7 17.5 Be 234/Be 313 Ca II 393.366 4.0 3.5 -12.5 4.0 0.0 Mg 285/Mg 279 Ca II 393.366 6.0 4.6 -23.3 5.6 -6.7 Mg 285/Mg 279 Ca II 396.847 4.0 3.7 -7.5 3.9 -2.5 Mg 285/Mg 279 Ca II 396.847 6.0 4.7 -21.7 5.4 -10.0 Mg 285/Mg 279 Mn II 257.610 6.0 4.0 -33.3 6.1 1.7 Mg 285/Mg 279 Cu I 327.396 4.0 3.6 -10.0 4.9 22.5 Mg 285/Mg 279 Cu I 327.396 6.0 4.5 -25.0 6.2 3.3 Mg 285/Mg 279 Cu I 324.754 4.0 3.6 -10.0 5.0 25.0 Mg 285/Mg 279 Cu I 324.754 6.0 4.5 -25.0 6.5 8.3 Mg 285/Mg 279 Al I 309.283 4.0 2.2 -45.0 4.1 2.5 Mg 285/Mg 279 Al I 309.283 6.0 3.6 -40.0 6.9 15.0 Mg 285/Mg 279 Mean error±SD -26.2±14.1 6.7±10.9 Number of replicates=3.Standard deviation ranged from 1 to 3%. enhancement of the analyte signal. Oxygenated or chlorinated coeYcient=0.76), while the plot of the correction factor against Ia/Ii retains its linear character. Accordingly, it is hydrocarbons cause less signal depression than ordinary hydrocarbons with comparable volatility, in agreement with the preferable to use the plot of the correction factor against Ia/Ii to correct for the matrix vapor loading. results obtained by Boorn et al.12 Numerical simulation Simulated samples The experimental results from the simulated samples were Results of the numerical simulation indicate that plots of the correction factor Ii0/Ii, against Ii/Ia [eqn.(6)] or Ia/Ii [eqn. (7)] used to test the linearity of the relationship between the correction factor and the ratio Ia/Ii. All the constructed curves give straight lines. However, the straight lines obtained from the plot of the correction factor against Ia/Ii [eqn.(7)] (Fig. 3) were linear regardless of the diVerences in volatility and chemical composition. Representative examples of these exper- give better correlation coeYcients than the lines obtained from the plots of the correction factor against Ii/Ia. This diVerence imentally obtained plots are illustrated in Fig. 4–6 for Be and Mg in xylene and 1,2,3,4-tetrahydronaphthalene. This agrees becomes more pronounced when the major and minor solvents of the sample have much higher volatility than the calibration with the results of the numerical simulations with eqn.(7). The simulated samples were measured using a universal solution matrix. Then, the plot of the correction factor against Ii/Ia shows more curvature (curved line with correlation calibration solution prepared in 1,2,3,4-tetrahydro- 806 J. Anal. At. Spectrom., 1999, 14, 801–807naphthalene. The proposed chemometric technique was matrices. The ratio of the atomic-to-ionic line intensities is sensitive to vapor loading and chemical composition of the applied to correct the results for the diVerence in vapor loading and chemical composition between the samples and the stan- matrix.Therefore, the ratio can be used as a mathematical measure for the eVect of these two variables. The chemometric dard matrices. The results of these measurements before and after correction are summarized in Table 5 for Mg, Be and technique based on this observation is capable of facilitating the use of the universal standard calibration strategy on a Fe in xylene and in Table 6 for Mg, Be, Ca, Mn, Cu and Al in xylene.Without correction the signals were suppressed by routine basis. The developed chemometric technique is general, because it can be applied also to elements that have only an 10–80%. Addition of up to 2% of other organic compounds accentuates the errors. Corrected value errors ranged from 0 ionic or an atomic line with suitable sensitivity, but not both.This can be achieved with the ratio of atomic-to-ionic line to 25%, and only for two Cu lines (at 4 mg L-1) were the corrected values worse than the uncorrected values (Table 6). intensities of other analyte in the sample. The new analysis and correction schemes are simple and involve no sample pre- The reported results indicate clearly the eYcacy of this chemometric technique both by using the Ia/Ii, ratio of spectral lines treatment. Hence they are amenable to programmed operation and autosampling.Studies are under way to extend the practi- of the same analyte (Mg and Be) or another analyte in the same sample (Al, Ca, Cu, Fe and Mn). cal application of this chemometric technique to other elements and matrices. The results of applying the proposed corrections to the simulated samples indicate that the transport eYciencies of several matrices (xylene+small percentage of other volatile Acknowledgments compounds) and the standard (1,2,3,4-tetrahydronaphthalene) The authors thank the Perkin-Elmer Corporation for providing are virtually the same despite the marked diVerences in their the Optima 3000 prototype system.This investigation was physical properties. This is expected from the results obtained supported by the ICP information Newsletter, Inc. (Hadley, by Boorn et al.,12 who indicated that even with compounds as MA, USA). diVerent in physical properties as m-xylene (bp 139 °C, viscosity 0.62 cP, density 0.86 g mL-1; and surface tension 29 N m-2) and nitrobenzene (bp 211 °C, viscosity 2.03 cP, References density 1.2 g mL-1 and surface tension 44 N m-2) only a small 1 R. I. Botto and J. Zhu, J. Anal. At. Spectrom., 1994, 9, 905. diVerence exists in their transport eYciencies. Accordingly, the 2 R. I. Botto and J. Zhu, J. Anal. At. Spectrom., 1996, 11, 675. results obtained in this work are concerned with only the 3 D. Wiederin, R. S. Houk, R. Winge and A. D’Silva, Anal. Chem., correction for the diVerence in volatility and chemical composi- 1990, 62, 1155. tion between the sample and the standard matrices. 4 J. Christodoulou, M. Kashani, B. M. Keohane and P. J. Sadler, The eVect of diVerent solvents on the emission background J. Anal. At. Spectrom., 1996, 11, 1031. 5 J. McLean, H. Zhang and A. Montaser, Anal. Chem., 1998, 70, intensity at the wavelengths studied was negligibly small. As 1012. an example, when tetrahydronapthalene solvent was replaced 6 T. Avery, C. Chakrabarty and J. Thompson, Appl. Spectrosc., by xylene, the changes in the background, expressed as equival- 1990, 44, 1690. ent concentration in mg L-1, were Ca (396 nm) 0.01, Ca 7 M. Thompson and M. H. Ramsey, Analyst, 1985, 110, 1413. (393 nm) 0.004, Cu (324 nm) 0.002, Mg (279 nm) 0.0001 and 8 A. S. Al-Ammar and R. M. Barnes, Spectrochim. Acta, Part B, Mg (285 nm) 0.007. For other analyte lines, the background 1998, 53, 1583. 9 A. S. Al-Ammar and R. M. Barnes, At. Spectrosc., 1998, 19, 18. equivalent concentration values were smaller. 10 X. Romero, E. Poussel and J. M. Mermet, Spectrochim. Acta, Part B, 1997, 52, 487. 11 J.-M.Mermet and J. C. Ivaldi, J. Anal. At. Spectrom., 1993, 8, 795. Conclusion 12 A. Boorn, M. S. Cresser and R. Browner, Spectrochim. Acta, Part Two spectral lines, an ionic and an atomic, of the same element B, 1980, 35, 823. can be used to correct for the diVerence in vapor loading and chemical composition between the sample and the calibration Paper 8/08389D J. Anal. At. Spectrom., 1999, 14, 801–807 807