JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 263 I Direct Solid Sampling for Analysis with Inductively Coupled Plasma Using a Novel Electronic Spark Source* C. Webb C. B. Cooper 111 A. T. Zander J. T. Arnold and E. S. Lile Varian Associates Ginzton Research Center Palo Alto California 94303 USA S. E. Anderson Va rian Op tical Spectroscopy Ins trum en f s Melbourne A us tralia A solid sampling accessory using a spark discharge for introduction of analyte material into an inductively coupled plasma (ICP) torch has been developed. The device uses a novel electronic design which results in a dramatic simplification of the hardware compared with previously described arrangements. A simple cell consisting of a tungsten electrode adjacent to the sample in a flowing inert gas atmosphere completes the unit as an extremely compact benchtop accessory.This device has been optimized for stable operation over a minimum of 12 min; relative standard deviations for signals over that period are typically <3%. Precisions of 6 4 % were found for the determination of trace elements in National Institute of Standards and Technology Standard Reference Materials with limits of detection in the single digit ppm range. Keywords Solid sampling; spark; inductively coupled plasma Inductively coupled plasma atomic emission spectrometry (ICP-AES) and inductively coupled plasma mass spectrometry (ICP-MS) are widely employed analytical techniques. The most common method of sample introduction is by nebuliz- ation of liquids but this requires that solid samples must first be subjected to some dissolution procedure and methods of direct solid sample introduction are of obvious appeal.Additionally solvents can result in polyatomic interferences in ICP-MS.l Spark ablation has previously2+ been shown to be a useful technique for solid sampling and although it is limited to conducting samples techniques have been developed for mixing finely dispersed samples with conducting powders such as g r a ~ h i t e . ~ In this paper results obtained using a novel implementation of spark ablation as a solid sampling accessory (SSA j for ICP is reported. Developments in modern electronic components have permitted the design of a small simple circuit’ which is incorporated into a compact benchtop SSA unit representing a dramatic reduction in hardware compared with previous devices.The present emphasis is on the optimization of the operation of this device and the analytical results obtained from it. Experimental A schematic diagram of the cell is shown in Fig. 1. The sample is sealed to the cell (volume = 1 ml) by means of an O-ring and the spark source is connected to a tungsten electrode (diameter 1.5 mm) the height of which can be adjusted. Argon flows through the cell at a flow rate of z 11 min-’ and transports ablated material to the ICP torch; the argon gas flow rate was not found to be an important parameter. Sample cooling is not necessary. The spark source allows various parameters to be set allowing the adjustment of pulse-width frequency and current. In addition as discussed below separate pre-burn conditions can be set along with the duration of the pre-burn step.The samples have so far consisted of steel [National Institute of Standards and Technology (NIST) Standard Reference Materials (SRMs) 1261 1263 1264 1265 1269 and 12231 Fe-Ni alloy (NIST SRM 1159) and brass (NIST SRMs 1104,1107 and 1108). Typically for steel samples the analytical settings used were 3 mm tip gap 8 ps pulse-width 60 A current and 500 Hz frequency. The pre-burn settings differed in having * Presented at the XXVIII Colloquium Spectroscopicum Internationale (CSI) York UK June 29-July 4 1993. Tungsten elect rode t- SSA power supply Fig. 1 Schematic diagram of sample cell and spark source. The electronics allows selection of pulse parameters as follows pulse width 4-10 ps; current 20-80 A; and frequency 250-1200 Hz.A timed pre- burn with independent parameter settings is permitted. The tip gap is adjustable in the range 0-10 mm 80 A current and 1000 Hz frequency for the first 40 s. All of the data shown here were obtained with ICP-AES systems. Normal ICP operating conditions were 1.2 kW power input outer argon flow and intermediate flow rates were 13.5 and 1.5 1 min-l respectively. Integration times were 5 s; RSD infor- mation is based on ten readings. Sample preparation simply consisted of abrasion with sandpaper and an acetone rinse. The dry aerosol was transported to the ICP via a plastic (‘Tygon’) tube. The analytical lines used were Fe 240.488 nm (analytical data) or 259.940 nm (optimization data) Ni 231.604 Mn 257.610 V 292.402 Cr 267.716 Sn 242.949 Zn 206.200 and 220.353 nm.Results and Discussion Optimization The criterion for stable operation was that a steady signal could be maintained for 12 min while the signal stabilization264 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 time was minimized; 12min was chosen as a plausible upper bound for analysis time using a sequential spectrometer. This required proper selection of pulse parameters and in particular the tip gap was found to be critical. Craters on sample surfaces were examined using a stylus profilometer and were found to vary essentially as would be expected with respect to tip gap as shown in Fig. 2 where tip gap is a parameter. It should be noted that the 1 mm gap results in a crater which is already several percent of the gap after only 80s.Perhaps more importantly significant build-up of ablated material was also observed on the tip at the 1 mm setting. Accordingly an ICP signal monitored for example from the matrix Fe 260 nm line exhibited noise and drift at a 1 mm gap. This was not a problem at either 3 or 5mm however and the 3mm setting was considered more satisfactory for the remainder of this work since it gave a larger signal and shorter signal stabiliz- ation time than the 5 mm setting. The time required for signal stabilization is obviously important since it represents dead time during which no analysis can be performed. With the analytical conditions described above and no pre-burn the time required for signal stabilization was about 200s.It was demonstrated that this depended upon the sample surface since shutting off the spark once a stable signal was achieved and then re-starting after a 40 s delay resulted in a much faster signal rise with sparking occurring on to the pre-existing crater. It appears that the stabilization is only achieved after some initial conditioning of the surface. Suitable pre-burn conditions to allow accelerated surface conditioning were determined and it was found that an increased frequency was particularly effective in enhancing the rate of erosion. Whereas the crater depth reached= 13 pm in 160 s with the usual analytical settings mentioned above a crater of depth= 17 pm was achieved in only 40 s with the pre- burn settings. Based upon the previous observation that signal stabilization requires ~ 2 0 0 s with the analytical settings it is considered that a crater about this size corresponds to adequate surface conditioning.In Fig. 3 the signal rise characteristics both with and without a 40s pre-burn are compared and it can be seen that the signal reaches its steady-state value within about 80s with the pre-burn compared with about 200s without. Although the pre-burn condition used includes an increase in current various parametric studies have shown that it is the increase in frequency that is most effective in enhancing the rate of signal rise. The results of one such experiment for NIST SRM 1223 steel are shown in Fig. 4 where only the frequency the is varied. The signal increases initially much faster at higher frequencies than can be explained on a linear (effective time-compression) basis.Microscopy of craters sug- gests that melting is occurring and the supra-linear dependence on frequency is probably due to thermal effects. Comparison of integrated signals with crater depth (used for simplicity h i \ I I I W I 40 t 50 -3 -2 - 1 0 1 2 3 Surface positi on/m m Fig. 2 Sections obtained with a stylus profilometer of craters pro- duced on sample surfaces as a function of tip gap (a) 5 mm (160 s); (b) 3 mm (160 s); and (c) 1 mm (80 s). No pre-burn was employed here 0 100 200 300 400 500 600 Time/s Fig. 3 ICP signal intensity for Fe 260 nm is shown as a function of time (zero is spark-on time) A with and B without a 40 s pre-burn as described in the text 0 50 100 150 200 Time/s Fig.4 ICP signal intensity for Fe 260 nm is shown as a function of time. The effect of spark frequency A 244; B 325; C 500; D 1000; and E 1200 Hz on signal rise characteristics is illustrated in lieu of volume since width does not vary much with depth) shows good linearity which would seem to rule out expla- nations dependent on particle size variations. It should be emphasized that the time required for initial signal stabilization is strongly dependent upon requirements with respect to the length of time for which a stable signal must be held i.e. for the present purposes 12min. The 80s initial stabilization time is only 10Y0 of the total spark-on time. If analysis times shorter than 12 min are required then it would be possible (e.g. by using shorter tip gaps) to reduce the time required for signal stabilization also. As discussed above a basic dilemma is that it is desirable to keep the crater diameter small in order to minimize the time required for surface conditioning while keeping the tip gap large to maintain a stable signal.Although we believe that a good compromise utilizing the pre-burn described has been found a more direct approach involving a modification to the tip was also tested. It was noted that a concave tip where sparking occurs from the perimeter produces a ring shaped crater whose diameter is approximately twice that of the electrode. This shows that a spark from one point on the circumference is influenced by the electrostatic field produced by the overall shape of the tip. In order to take advantage of this the conventional pointed tip was used with a collar as illustrated in Fig.5. The crater size is very dependent upon the setback as defined in the figure but a range of setbacks can be found for which the crater diameter is decreased and the depth increased compared with a simple pointed electrode. The magnitude of the effect is small (% 10-15%) however and €or simplicity a simple pointed electrode was used to obtain the analytical data described below.265 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 - - ’\ Setback Fig.5 The ‘collared-tip’ arrangement used as an alternative to the more usual simple point arrangement Analytical Performance The analytical performance was evaluated using a Varian Liberty ICP atomic emission system during a relatively short period of time and it is possible that the ultimate performance will be improved.However the results obtained here are already very encouraging. In Fig. 6 stability data for several elements obtained from an NIST SRM 1223 steel sample are shown. The intensity information is plotted on a reduced scale; i.e. for each element the maximum is set equal to unity and the other data points scaled accordingly. The relative standard deviation values were calculated from the sets of data in Fig. 6 for each element taken over the 12 min period and are given in Table 1. Here the ‘norm’ suffix refers to normalization by the matrix Fe; both stability and precision data are shown. The precision data in Table 1 refer to a data set from ten sequential craters located at different positions on an NIST SRM 1261 steel sample; the sample was repeatedly re-surfaced (using sandpaper and an acetone rinse) and following the pre- burn procedure outlined above intensity information was obtained. Normalization results in an appreciable overall 1.2 r 1 2 0.8 I U i I 0.6 I I I 0 5 10 15 Ti melmin Fig.6 Stability data obtained for various elements in NIST SRM 1223 steel D Fe; 0 Ni; 0 Mn; and x V.The intensity information is plotted on a reduced scale i.e. for each element the maximum is set equal to unity and the other data points scaled accordingly Table 1 Relative standard deviations for stability and precision. The stability values refer to the data in Fig. 6 and are for a single crater over a period of 12 min.The precision values are for 10 separate craters on one sample. The suffix ‘norm’ refers to normalization by Fe matrix Element Fe Ni Mn V Ni norm Mn norm V norm Stability 2.6 3.3 1.9 1.5 1 .o 3.1 2.8 Precision 3.6 4.0 3.7 3.2 1.5 1.5 1.6 improvement for the precision data but not for the stability data. This could reflect small but significant changes in the presentation of the sample to the cell; e.g. variations in surface roughness could cause minute but variable leaks perturbing gas flow patterns in the cell. However all RSDs are 4% or better which compares well in the context of solid sampling generally.’ Analytical accuracy was tested by running a number of both steel and brass SRMs in two series. Fig. 7 shows the comparison of the measured values with the stated values for Cr in steel; other elements behave similarly.Some of the steels were run more than once and in those cases duplicate points are plotted. In the first run of the series NIST SRM 1264 was arbitrarily taken as a ‘standard’ and the sensitivity factors derived from this sample were applied to the data from the remaining runs. General agreement is good over a wide range of concentrations though deviation occurs at lower concentrations because back- ground corrections other than subtraction of the spark-off background were not made. There was generally little need to use matrix normalization except for NIST SRM 1263 (points at stated concentration 1.31 %) which exhibited a noticeably lower erosion rate than the other samples under fixed conditions.Since only three brass samples were used data obtained using the highest concentration for each element are plotted in Fig. 8 to derive the sensitivity factor and the highest point for each element is effectively constrained to lie on the line while the two lower points verify the accuracy of the method. I I I I J I X ~ O - ~ i x i o - * I X I O - ’ I 10 1 x 1 0 2 SRM concentration (%) Fig.7 Calculated analysis data for Cr in steels are plotted against SRM data sheet values A Cr and; B Cr (normalized). ‘Normalized’ refers to data which have been normalized by the matrix (Fe). These data refer to a series of runs with the various steel samples NET SRM 1261 1263 1223 1269 1265 1159 (Fe-Ni alloy) where the first run of the series with NIST SRM 1264 steel was used as the ‘standard’ to derive sensitivity factors for the remaining runs RSD (Yo) I lo-‘ 1 x 10-2 1 1 XI02 SRM concentration (%I Fig.8 Calculated analysis data for various elements in three brass samples are plotted against SRM data sheet values 0 Ni; 0 Sn; 0 Mn; x Fe;+ Zn; and A Pb (see text for further explanation). Since there was no information on Mn at the two lower concentrations these points were plotted at an assumed detection limit of 1 ppm266 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 Detection limits were calculated for one of the steel samples and found to be generally in the single digit ppm range. Preliminary data using the solid sampling device for ICP-MS analysis indicate that low concentrations of La (700 ppb) and Nd (700 ppb) in steel are readily measured.Additional detail will be the subject of a future publication." Conclusion A new compact solid sampling accessory for ICP based upon a spark source which utilizes novel electronics has been developed. The optimization of this device has been described noting considerations in the compromise choice of tip gap and the advantage of using a pre-burn at higher pulse frequency for rapid conditioning of the sample surface. Analytical data have been obtained for a number of steel and brass SRMs showing good performance with respect to stability precision detection limits and analytical accuracy. References 1 Alves L. C. Wiederin D. R. and Houk R. S. Anal. Chem. 1992 64 1164. 2 3 4 5 6 7 8 9 10 Routh M. W. and Tikkanen M. W. in lnductively Coupled Plasmas in Analytical Atomic Spectrometry eds. Montaser A. and Golightly D. W. VCH New York 1987 p. 455-468. Human H. G. C. Scott R. H. Oakus A. R. and West C . D. Analyst 1976 101 265. Aziz A. Broeckaert J. A. C. Laqua K. and Leis F. Spectrochim. Acta Part B 1984 39 1091. Jakubowski N. Feldmann I. Sack B. and Stuewer D. J. Anal. At. Spectrom. 1992 7 121. Broekaert J. A. C. Leis F. Raeymaekers B. and Zaray Gy. Spectrochim. Acta Part B 1988 43 339. Steffan I. and Vujicic G. Spectrochim. Acta Part B 1991 47 61 Arnold J. A. Zander A. T. Lile E. S. and Cooper C. B. in the press. Routh M.H. and Tikkanen M. H. in Inductively Coupled Plasmas in Analytical Atomic Spectrometry eds. Montaser A. and Golightly D. W. VCH New York 1987 pp. 441 and 463 Webb C. Plantz M. Cooper C. B. paper presented at the 1994 Winter Conference on Plasma Spectrochemistry San Diego CA USA January 10-14 1994. Paper 3104821 G Received August 10 1993 Accepted October 21 1993