Correction of Matrix Effects in Quantitative Elemental Analysis With Laser Ablation Optical Emission Spectrometry C. CHALE� ARD, P. MAUCHIEN*, N. ANDRE, J. UEBBING†, J. L. LACOUR AND C. GEERTSEN‡ Commissariat a` l’Energie Atomique, L aboratoire de Spectroscopie L aser Analytique, Centre d’Etudes Saclay, Ba� t. 391, 91191 Gif sur Y vette, France A new approach to the quantification of the optical emission when ablation is performed with a laser irradiance lower than 1 GWcm-2.13 signals from a laser produced plasma in air at atmospheric pressure is described.It is based on a simple analytical model The analytical procedure commonly used for the correction of matrix effects15 is based on normalization of the analyte giving intensity of the emission lines as a function of the vaporized mass and the plasma excitation temperature. Under signal by a reference signal (generally the major element of the matrix). Such a normalization procedure is of no use when no the hypothesis of a stoichiometric ablation process, these two parameters are presumed to be responsible for the matrix reference element is available.This is frequently the case for on-line analysis, where the matrix could change with time as effects observed when the composition of the sample is changed. Under the experimental conditions chosen it is the process evolves, as for microanalysis because the matrix could change from one location to the other on the solid. demonstrated that the acoustic signal emitted by the plasma is proportional to the vaporized mass and that the excitation The purpose of the present work was to identify the dominant matrix effects appearing in LA-OES and to develop an plasma temperature can be controlled using the line ratio of a given element chosen as a ‘temperature sensor’.Normalization analytical procedure enabling correction of these effects to be made for quantitative multi-matrix analysis. of the emission signals by these two parameters allows for an efficient correction of the matrix effects.Results obtained on a series of aluminium alloys, steel and brass samples EXPERIMENTAL demonstrate, for the first time, the possibility of matrix independent analysis with LA-OES with an accuracy of a few The experimental arrangement shown in Fig. 1 is similar to percent. the one described previously5 and so will be discussed only briefly here. A XeCl excimer laser (Lambda-Physik EMG 201 Keywords: L aser ablation; optical emission spectrometry; MSC, Gottingen, Germany) emitting at 308 nm with a 28 ns matrix effects; elemental analysis pulse duration was used as the ablation source.The laser beam was spatially filtered before being focused Laser ablation is becoming a very popular technique in the on the sample in order to control the energy distribution of field of analytical atomic spectrometry. This is mainly attribu- the beam impinging on the sample throughout the experiments.table to the fact that the process is applicable to a very large For this purpose, the laser beam was focused by a 1000 mm range of materials and analytical situations. The most popular focal length lens on a diaphragm, and the central part of the applications result from the fruitful combination of LA with beam, passing through this diaphragm, was focused on the ICP-MS. sample surface (10° relative to the normal incidence) by a LA-OES is also a well known technique1,2 with a very high plano-convex lens with a focal distance of 250 cm.The resulting potential for elemental analysis. Its principle is based on the focal spot diameter was 500 mm (spot size about 2×10-3 cm2), measurement of the intensity of the atomic lines emitted by giving a power density of 1.4×109 Wcm-2 at the sample the plasma initiated above the surface, during the interaction surface. Such a large spot area is required to avoid problems of a high power density laser beam with a solid sample.This with the heterogeneity of the samples.5,10 purely optical technique offers the possibility of measurements The laser beam energy was frequently measured with a without sample contact, which is obviously of great interest power meter (Scientech 372, Boulder, CO, USA) and continufor many industrial applications. Several papers have demon- ously monitored by means of a calibrated photodiode. strated the potential of the technique for on-line measurement In order to ensure that all the experiments were carried out of meltingmaterial3 and for the control of industrialsamples.4–6 with a constant power density, even when the samples were Performances obtained have been summarized in several moved or replaced, it was necessary to control carefully the reviews.7–9 In a recent paper,10 potential of the technique for quantitative microanalysis with lateral resolution of a few microns has been demonstrated for the first time.However, for this application, as for one-line analysis where LA-OES has a unique potential, the major drawback comes from matrix induced effects.11–15 Some papers have shown either the matrix dependent character of the ablation process in binary samples with IR lasers16,17 or the non-stoichiometric evaporation process † On leave from Institut fu�r Spektrochemie und Angewandte Spektroskopie, Dortmund, Germany. Fig. 1 Experimental set-up. ‡ On leave from Pechiney Centre de recherches de Voreppe, France.Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 (183–188) 183distance between the focusing lens and the sample surface. vaporized mass and of the excitation temperature in the plasma. Hence, correction of the matrix effects could be This was done by means of a laboratory-made ultrasonic emitter–receiver, allowing positioning of the surface with a performed if both the vaporized mass and the plasma temperature are measured simultaneously with the analytical signal.precision of approximately 0.1 mm, a value compatible with the Rayleigh distance, evaluated to be 5 mm. This is obviously a very simple approach if one considers the number of parameters influencing the transformation of Emission from the plasma was collected by a set of two plano-convex lenses with a focal distance of 50 mm, resulting solid matter into a plasma but it was the purpose of the present work to simplify the problem in order to point out the in a demagnification ratio of 30. In this way, the whole volume of the plasma was collected by the 200 mm entrance slit-width key parameters of the ablation process.Nevertheless, these two key parameters must also be sup- of the spectrometer. This experimental configuration, using a very small collection angle, leads to a loss in sensitivity but posed to be dependent on the power density. As a consequence, it is essential to work under closely controlled laser parameters.could improve the reproducibility of the measurements. It is in fact known that laser produced plasmas are not spatially From this point of view, the ideal situation for laser ablation experiments would be a ‘top hat’ distribution of the laser beam and temporally homogeneous emission sources18,19 and the shot-to-shot variations in the atomic densities and temperature energy. Indeed, under such experimental conditions, the power density at the surface could be known precisely by just distribution inside the plasma could affect the reproducibility of the analytical measurements when only the central part of monitoring the laser energy and measuring the diameter on the crater.This is obviously not the case when the laser exhibits the plasma is probed. Detection was performed perpendicular to the surface, i.e., an uncontrolled energy beam distribution because each point of the laser surface does not receive the same amount of emission was collected in the expansion axis of the plasma.This configuration is the most compatible with in situ measure- energy. This is of particular importance with a multi-mode laser source for which the variation is generally very high. This ments because it is single ended and it is less sensitive to changes in the plasma-to-collecting lens distance that result can induce a highly complexion process in which the ablation regime may not be the same at each point on the when several shots are fired at the same place on the sample surface.A crater is formed on the surface, leading to a ablated surface. For these reasons, an attempt was made to improve the experimental set-up, by introducing the spatial displacement of the plasma. This displacement causes minimum perturbations if the plasma is probed in its expansion direction beam filtering system described under Experimental. The efficiency is demonstrated in Fig. 2. (i.e., perpendicular to the sample surface) because of the depth of focus of the detection optics. Both the energy distribution (integrated on the pulse duration) of the focused beam imaged on a charge coupled device The f/6 spectrometer (DILOR Model XY, Lille, France) was equipped with a gated intensified photodiode array (EG&G (CCD) camera, and the corresponding appearance of the crater in an aluminium sample before spatial filtering was done are RETICON S-series, Sunnyvale, CA, USA; 1024 photodiodes, 700 intensified photodiodes).The wavelength range recorded shown in Fig. 2(a). It can be seen that, without spatial filtering, the distribution of the beam is highly disturbed leading to a simultaneously was approximately 13 nm (at 300 nm) and the spectral resolution was about 0.15 nm. Gated electronics allows very irregular crater. The main problem comes from the fact that the beam energy distribution is known to change from the adjustment of both the delay after the laser pulse and the gate duration in the range 0.1–99 ms.According to the results pulse to pulse and also with time, as a function of the number of shots performed with the gas charge. It is also suspected obtained in previous work,5 emission from the plasma was measured 2 ms after the laser pulse using a 2 ms gate duration that the distribution would appear much more disturbed if the measurement was not integrated over the pulse duration. This (these values have been determined as giving the best SNR). Measurements of the emission intensity were made by integrat- problem corresponds to the known existence of temporal modes in an excimer laser cavity.ing a series of 20 shots fired in the same crater. Six independent measurements performed at different positions, chosen arbi- The results obtained when the laser is spatially filtered are shown in Fig. 2(b). As expected, the distribution of the beam trarily on the sample, were then averaged.is much more homogeneous and its size is only dependent on the optical arrangement. The consequence is a regular and RESULTS AND DISCUSSION well defined crater size, as can be seen in the figure. From these results, one can consider that the experimental set-up Analytical Model developed allows for reproducible and well controlled ablation The analytical procedure proposed to quantify the emission conditions and will enable a quantitative study of the matrix lines is based on the hypothesis of a stoichiometric ablation effects to be carried out.process. It is a reasonable hypothesis if one considers the power density used in the experiments.13,20 Vaporized Mass Diagnostics The intensity of an emission line of a given element i can be written as: A promising approach to this problem was presented by Chen and Yeung,21 who used the acoustic wave produced by the Ii=KCiMple-E/kT (1) ablation process as an internal standard in their experiments. The measurements were done in a closed cell filled with He where K is a constant factor which takes into account the collection efficiency and the spectroscopic data for the line at 50 Torr (1 Torr=133.322 Pa).It was demonstrated that increasing the pressure leads to a complicated waveform, being considered, Mpl is the total mass of matter vaporized in the plasma plume, Ci is the concentration of the element i in resulting from reflection and mixing of the acoustic waves within the cell.The acoustic wave was also evaluated for the the plasma, equivalent to the concentration in the solid phase under stoichiometric ablation conditions, and e-E/kT is rep- normalization of signals generated by an ICP source, using laser ablation as the sampling technique.22 The results showed resentative of the excitation temperature assuming the plasma is in local thermodynamic equilibrium (k is Boltzmann’s very good correlation between the acoustic signal and the analytical signal but, as for the previous case, experiments constant).Assuming the validity of the proposed approach, eqn. (1) were performed in a closed sample cell. Finally, to our knowledge, no results so far have been shows that matrix effects observed in laser ablation experiments result only from the sample-to-sample variations of the published on the use of acoustic wave normalization at 184 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12(a) (b) Fig. 2 Effect of spatial filtering on laser beam (left) and crater (right). Left images correspond to single shot measurements while right images result from ten shots in the same crater: (a) without spatial filtering; and (b) with spatial filtering. atmospheric pressure in air, without a sample cell. This is a interest. This result contrasts with that published, for example, by Shannon and Russo23 who have shown that, in the same very interesting challenge because acoustic wave measurement is a very simple technique and well adapted to in situ measure- range of power density, two different regimes of interaction have to be considered.This difference illustrates the difficulty ments (no contact between the sample and the instrumentation is required at all), which constitutes, from our point of view, of comparing experiments carried out with different laser sources having different energy distributions. the main interest in LA-OES. A first set of experiments was carried out to evaluate the Detection of the acoustic wave was made by means of a microphone positioned approximately 20 cm in front of the feasibility of acoustic wave normalization under the present experimental conditions.In order to be quantitative, such an sample surface. The acoustic pressure resulting from the plasma expansion was converted into an electric signal whose ampli- evaluation requires controlled variation of the ablated mass to be provoked. The simplest way of doing this consists in varying tude was measured by means of a gated electronic device.The gate width and the temporal delay after the laser pulse were the laser power, but that a correlation exists between ablated mass and laser power has to be verified. adjusted to the first maximum of the signal, the delayed signals being attributed to reflection of the acoustic wave on both the The ablated mass was evaluated by weighing the sample before and after ablation experiments; 500 shots were integrated sample and the components positioned in its vicinity.The signal was recorded for each shot and stored in the computer to obtain a good measurement accuracy. The results obtained on an aluminium sample are presented in Fig. 3. It appears memory at the same time as the emission line intensities. The variation of the acoustic signal as a function of the that the ablated mass exhibits a linear dependence versus the laser irradiance incident at the surface in the range investigated ablated mass when the laser power, incident on an aluminium sample, was varied between 0.6 and 1.2 GW cm-2 is shown in (0.6–1.2 GW cm-2).This clearly demonstrates that a unique regime of interaction has to be considered in the range of Fig. 4. It can be seen that the acoustic signal (As) is proportional to the ablated mass (As=aMpl where a is a constant) in the power range investigated. The same linear behavior has been Fig. 3 Ablated mass as a function of the laser irradiance incident on Fig. 4 Detected acoustic signal as a function of the ablated mass. The laser irradiance, incident on an Al sample, was varied between an Al sample. The measurements result from integration over 500 laser shots. 0.6 and 1.2GW cm-2. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 185observed for different matrices such as nickel, steel and glass. from Boltzmann’s law: These results validate the use of acoustic signals as a diagnostic tool for evaluating the amount of material removed by the r= E2-E1 ln I1 I2 +ln g2A2 g1A1 (2) ablation process.The low value of the slope giving the acoustic signal as a function of the ablated mass should be noted. This is obviously where E1,2 are the upper energy levels of the lines, I1,2 are the a limitation in the precision that could be expected for the measured intensities, g1,2 are the degeneracy factors of the normalization of analytical signals, but it is partially compen- atomic states and A1,2 are the transition probabilities. sated for by the good reproducibility of the acoustic measure- From eqn. (1) the normalized intensity can then be written ments (a few percent).as: Moreover, the observation of the linearity between the acoustic wave generated by the plasma and the ablated mass Ii Ase-Eu/r=1 aKCi (3) leads to the very interesting conclusion that, for the matrices tested, most of the material removed by the laser ablation where As=aMpl is the acoustic signal intensity and Eu the process is vaporized under the experimental conditions.This upper energy level of the analytical line (i ) being considered. is confirmed by the results presented in Fig. 5, which show the Owing to uncertainties existing for the values of the spectro- linear dependence between the acoustic signal and the intensity scopic parameters of the lines, the proposed procedure will of the 403.08 nm Mn line emitted by the plasma produced lead to minimum imprecision if only one element is chosen as from aluminium, when the laser power is varied as described the ‘temperature sensor’.This means that it must be present previously. As the emission signal is proportional to the atomic both in the reference samples used to produce the calibration density of the element, it is possible to conclude that the largest curve and in the unknown sample. Copper was chosen because part of the material removed by the laser is atomized.it is present in many metallic samples and it has two close This important result probably has to be attributed to the lines (510.55 and 515.32 nm) originating from two separate very high temperature of the plasma produced by laser–surface levels (30784 and 49937 cm-1, respectively) and so is well interaction in the GW cm-2 power range.24 adapted to temperature measurements. A first set of experiments was made to evaluate the normalization procedure when the laser energy was varied.The results Plasma Excitation Temperature Diagnostic shown in Fig. 6 indicate that the excitation temperature is only very weakly dependent on the laser irradiance. This is the It is known that a laser produced plasma is not a homogeneous reason why a linear dependence was previously found (see medium.18 Hence, measurement of the excitation temperature Fig. 5) between the acoustic signal, which is proportional to when the whole volume is probed leads to a value that is not the ablated mass, and the emission signal.particularly useful for the physical interpretation of the Qualitatively, these results show that, in the range measurements. For analytical applications the problem is 0.6–1.2 GW cm-2, only a small part of the laser energy is different. Quantification is always based on the use of a absorbed in the plasma by the inverse bremstrahlung effect.10 calibration curve and, as a consequence, only a relative Provided this condition is fulfilled, an increase in the laser coefficient proportional to the excitation temperature is energy produces a proportional increase in the ablated mass required for normalization of the signals.This will be an without modification of the excitation temperature. This seems integral part of the slope of the calibration curve provided the to demonstrate that, contrary to the generally accepted view, experimental conditions remain the same.vaporization–atomization and excitation can be, to a certain The procedure used in the present work was based on the extent, considered as independent phenomena. measurement of the ratio between two lines of a given element used as a ‘temperature sensor’. The two lines chosen originate from two separate energy levels in order to give a ratio that is Analytical Evaluation sensitive to variations in the excitation temperature. Assuming The proposed procedure has been evaluated under real analyt- that the populations of the atomic levels follow an exponential ical conditions by plotting calibration curves for some elements law, a temperature normalization coefficient (r) was calculated in particular matrices.For each sample, six measurements, from the intensities using the following relationship derived taken at different positions on the surface, were averaged. Each measurement corresponds to the integration of 500 laser shots in the same crater. The intensity of the analyte line, the acoustic signal and the two copper lines used for the calculation of the Fig. 5 Intensity of the Mn (403.08 nm) line as a function of the Fig. 6 Ratios of the intensity of the Mn lines (404.145403.08 nm) as acoustic signal. The laser irradiance, incident on an Al sample, was varied between 0.6 and 1.2GW cm-2. a function of the laser irradiance in the range 0.6–1.2 GW cm-2. 186 Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12excitation temperature parameter (r) were all recorded simultaneously.The first evaluation was made on a series of aluminium alloys. The calibration curve obtained by plotting the net intensity of the 510.55 nm Cu line as a function of the concentration is shown in Fig. 7. It can be seen that, even without normalization, the emission intensity is proportional to the analyte concentration. This results from the fact that both the atomized mass and the excitation temperature were the same for the samples.This was confirmed experimentally by measurement of the two coefficients. If one considers the different matrix compositions of the samples, this indicates that matrix effects are low under the experimental conditions used. Indeed, in the case of spark emission spectrometry, a widely used technique in the metallurgical industry, analysis of the same aluminium alloys requires a calibration curve to be plotted for each alloy owing to the pronounced matrix effects inherent with this technique.The same low matrix effects have been observed for Mn. The calibration curve obtained by plotting the net intensity of the 414.14 nm Mn line as a function of concentration in a series of samples (steel, aluminium and nickel) is shown in Fig. 8. According to eqn. (1), the linearity of the response indicates that both the ablated mass and the excitation temperature are the same for the three matrices. Significant matrix effects have been observed for samples with high Zn concentrations.This is illustrated in Fig. 9(a), which shows the net 510.55 nm Cu line emission intensity as Fig. 9 (a) Net intensity of Cu (510.55 nm) line as a function of the a function of the concentration in three samples (Zn, Al–Zn Cuconcentration. The measurements were performed on three different and Al–Cu alloys). As can be seen, owing to pronounced matrices: Zn, Al–Zn, Al–Cu. (b) Normalized intensity of Cu as a matrix effects, the net intensity is not proportional to the function of the Cu concentration under the same conditions as in (a).The intensity of the Cu line is normalized by both acoustic signal and concentration. excitation temperature coefficients. The values of the acoustic signal and temperature normalization coefficients (determined as previously) for the three samples are shown in Table 1. The acoustic signal, representative of the vaporized mass, is approximately the same for the two aluminium samples (Al–Zn and Al–Cu) while it is 20% higher for the Zn alloy.On the contrary, the excitation temperature coefficient is the same for the two samples with high Zn concentrations (Zn and Al–Zn samples) and significantly lower for the sample without Zn. These results confirm that the processes of sample vaporization and excitation in the plasma do not depend on the same parameters. The amount of vaporized material appears to be dependent on the major element in the matrix, while the excitation temperature depends on the presence of Zn in Fig. 7 Net intensity of Cu (510.55 nm) line as a function of the Cu the sample. concentration. The measurements were performed on different Al After normalization of the net line intensity by both acoustic alloys. signal and excitation temperature normalization coefficients, a calibration curve is obtained, as shown in Fig. 9(b). Even if one considers the low number of points, the linearity of the response demonstrates the validity of the procedure and of the parameters used to diagnose the variations induced by matrix effects.In order to complete these results, the procedure was evaluated for a wider range of materials and concentrations. Copper was selected as the analyte (concentration range 0–90%) in Al alloys, steels and brass samples. Table 1 Acoustic signal and excitation temperature normalization coefficients determined for Zn, Al–Zn and Al–Cu alloys Alloy Acoustic signal coefficient (As) Texc coefficient (r) Zn 6302 10069 Fig. 8 Net intensity of Mn (414.14 nm) line as a function of the Mn Al–Zn 5337 10095 concentration. The measurements were performed on different Al–Cu 5449 8704 matrices: steel, aluminium and nickel. Journal of Analytical Atomic Spectrometry, February 1997, Vol. 12 187the main advantage of LA-OES over existing techniques is its ability to perform in situ measurements. Pronounced matrix effects have been observed for samples with Zn concentrations exceeding a few percent.It has been observed that the major element in the sample appears to be responsible for the amount of vaporized material, while the presence of Zn increases the excitation temperature. It was demonstrated that the normalization of the net intensity by the acoustic signal, representative of the vaporized mass, and a temperature normalization coefficient, calculated from a ratio of two lines of a fixed element, allows for a multimatrix calibration curve with a satisfactory level of precision (accuracy of a few percent).It has to be pointed out that the precision of the results is not affected by the precision of the temperature determinations provided the same lines are used for the calculation of the normalization coefficient. As a result, the only limitation of the proposed procedure lies in the fact that an element used as a ‘temperature sensor’ must be present in all samples. In many cases, however, this is an acceptable analytical constraint.REFERENCES 1 Brech, F., and Cross, L., Appl. Spectrosc., 1962, 16, 59. 2 Moenke-Blankenburg, L., Spectrochim. Acta Rev., 1993, 15, 1. 3 Carlhoff, C., Lorenzen, C. J., Nick, K. P., and Siebeneck, H. J., in In-Process Optical Measurements, ed. Spring, K. H., Proc. SPIE, SPIE Publications, Bellingham, 1989. 4 Lorenzen, C. J., Carlhoff, C., Hahn, U., and Jogwich, M., J. Anal. Fig. 10 (a) Net intensity of Cu (510.55 nm) line as a function of the At. Spectrom., 1992, 7, 1029.Cu concentration for a large range of concentrations (0–90%). The 5 Andre, N., Geertsen, C., Lacour, J. L., Mauchien, P., and measurements were performed on different matrices including Al Sjo�stro�m, S., Spectrochim. Acta, Part B, 1994, 49, 12 and 1363. alloys, steels and brass samples. One can observe that both the Al and 6 Sabsabi, M., Cielo, P., Boily, S., and Chaker, M., Proc. SPIE-Int. steel samples appear on the same calibration curve.This single slope Soc. Opt. Eng., 1993, 199, 2069. indicates the absence of matrix effects for the materials concerned. (b) 7 Ready, J. F., Effects of High Power L aser Radiation, Academic Normalized intensity of Cu as a function of the Cu concentration Press, New York, 1971. under the same conditions as in (a). The intensity of the Cu line is 8 Von Allmen, M., L aser Beam Interactions WithMaterials, Springer normalized by both acoustic signal and excitation temperature Series in Material Science, Springer, Berlin, 1987.coefficients. 9 Radziemski, L. J., and Cremers, D. A., L aser-induced Plasmas and Applications, Marcel Dekker, New York, 1989. As shown in Fig. 10(a) the calibration curve obtained from 10 Geertsen, C., Lacour, J. L., Mauchien, P., and Pierrard, L., the net 510.55 nm Cu line intensity exhibits three different Spectrochim. Acta Part B, 1996, 51, 1403. slopes, illustrating the matrix effects for the series of samples. 11 Kagawa, K., Kawai, K., Tani, M., and Kobayashi, T., Appl. Spectrosc., 1994, 48, 198.However, it has to be noted that both the Al and steel samples 12 Aguilera, J. A., Aragon, C., and Campos, J., Appl. Spectrosc., appear on the same calibration curve indicating the absence 1992, 46, 1382. of matrix effects for the materials concerned. 13 Chan, W. T., and Russo, R. E., Spectrochim. Acta, Part B, 1991, Normalization of the signal by both acoustic signal and 46, 1471. excitation temperature coefficient leads to the calibration curve 14 Wisbrun, R., Schechter, I., Niessner, R., Schro�der, H., and Kompa, shown in Fig. 10(b). A single slope calibration curve is K. L., Anal. Chem., 1994, 66, 2964. 15 Wisbrun, R., Niessner, R., and Schro�der, H., Anal. Meas. Instrum., obtained indicating the compensation of matrix effects by the 1993, 1, 17. normalization factors with a precision compatible with the 16 Ko, J. B., Sdorra, W., and Niemax, K., Fresenius’ Z. Anal. Chem., reproducibility of the individual measurements (an RSD of 1989, 335, 648. about 5%). 17 Ciocan, A., Hiddemann, L., Uebbing, J., and Niemax, K., J. Anal. At. Spectrom., 1993, 8, 273. 18 Geertsen, C., Lacour, J. L., and Mauchien, P., paper presented at CONCLUSION the E-MRS 1995 Spring Meeting, Symposium F, COLA ’95, Strasbourg, France, May 22–26, 1995. The results demonstrate that spatial filtering of the laser beam 19 Autin, M., Briand, A., Mauchien, P., and Mermet, J. M., allows for better control of the laser ablation parameters. The Spectrochim. Acta, Part B, 1993, 48, 851. first consequence is a well defined crater shape and long term 20 Russo, R. E., Appl. Spectrosc., 1995, 49, 14A. stability of the experimental conditions. Very low second order 21 Chen, G., and Yeung, E. S., Anal. Chem., 1988, 60, 2258. 22 Pang, H., Wiederin, D. R., Houk, R. S., and Yeung, E. S., Anal. matrix effects have been observed during analytical measure- Chem., 1991, 63, 390. ments performed on a series of different Al alloys. This 23 Shannon, M. A., and Russo, R. E., paper presented at the E-MRS demonstrates that neither the amount of atomized material, 1995 Spring Meeting, Symposium F, COLA ’95, Strasbourg, nor the excitation temperature of the plasma are affected by France, May 22–26, 1995. some change in the matrix composition. 24 Vertes, A., Dreyfus, R. W., and Platt, D. E., in IBM Research The single slope calibration curve obtained in the case of Report, RC 18520, 1992. Mn in very different samples, i.e., Al alloys, steel and Ni samples, indicates that under the experimental conditions Paper 6/04456E chosen, even first order matrix effects can be very low. This is Received June 26, 1996 an important analytical feature, especially if one considers that Accepted September 30, 1996 188 Journal of Analytical Atomic Spectrometry, February 1997,