Atmospheric pressure capacitively coupled plasma source for the direct analysis of non-conductive solid samples† Sorin D. Anghel,a Tiberiu Frentiu,b Emil A. Cordos,*b Alpar Simonc and Adrian Popescuc aDepartment of Physics, Babes-Bolyai University, M. Kogalniceanu 1, 3400 Cluj, Romania bDepartment of Chemistry, Babes-Bolyai University, Arany J 11, 3400 Cluj, Romania cResearch Institute for Analytical Instrumentation, P.O. Box 717, PO 5, 3400 Cluj, Romania Received 8th September 1998, Accepted 4th January 1999 An atmospheric pressure capacitively coupled device for the rf sputtering and direct analysis by AES of nonconductive solid samples was developed.It operates at 13.56 MHz, with Ar flow rates lower than 1 l min-1 and rf input powers between 20 and 50 W. With the aim of studying the expulsion mechanism and determining the analytical performance, the following materials were used: ZnO with known content of trace elements (Si, Pb, Cd and Na); four andesite standards with certified contents of Pb, Cu and Cr; and a synthetic ( laboratory made) andesite standard.The expulsion mechanism depends on the rf power. Below 38 W, the sputtering rate is nearly constant and the atomisation is produced only by sputtering. At powers higher than 38 W, thermal evaporation will be present in addition to rf sputtering. All measurements for evaluating the analytical performance were made at optimum working parameters (rf power=36 W, Ar flow rate=0.5 l min-1).The detection limits are 0.3, 0.7, 1.0 and 0.9 mg g-1 for Na, Pb, Si and Cd, respectively, in the ZnO matrix and 0.8, 1.0 and 0.5 mg g-1 for Pb, Cr and Cu, respectively, in the andesite matrix. The dynamic range for Pb is about three orders of magnitude. The relative standard deviations for Pb in the certified standards are between 1.7 and 12.7% and the recoveries compared with the certified values are between 95 and 104% for the concentration range 5.8–35.1 mg g-1.and its most challenging problems are related to non- Introduction conductive samples. Over the last few decades, plasma excitation sources for Low pressure plasmas have become sampling systems; in analytical atomic emission spectrometry (AES) have been well the 1980s, glow discharge (GD) techniques experienced rapid characterised as sensitive spectral sources. After considering growth.2,3 The increased interest in this technique is attributed all of the characteristics of plasma sources that make this almost totally to its performance in the analysis of solid technique very popular, one problem area still remains unre- samples. One of the most exciting areas of GD development solved: sample introduction process versus sample state.is the coupling of radiofrequency powered GD with various Therefore, sample introduction is one of the most fertile spectrometric techniques.4 In this way, rf sputtering has research areas in AES. become a sample introduction method and the analyses are As emphasised by Blades et al.,1 the sample must be no longer limited to conductive samples.presented to the spectral source in a form that the source can Capacitively coupled plasmas, sustained at atmospheric accommodate easily using a method that ensures good trans- pressure, have been used for direct solid sampling in a relatively port eYciency towards the spectral source, is optimum for the few cases. Pless et al.5 used a capacitively coupled microwave particular sample and provides the best opportunity for plasma for sampling and excitation of solid samples, placed obtaining the required information.in W cups and introduced into the plasma. This type of plasma For samples in the solid state, there are basically two routes. has also been used for the direct analysis of steel samples.6 The first consists in dissolving the solid with a suitable solvent Radiofrequency capacitively coupled plasmas (rf CCPs), susand presenting it to the spectral source in liquid form, using tained at atmospheric pressure, were used by Liang and the specific techniques for liquid sample introduction.In some Blades7 and Anghel et al.8 to analyse conductive samples (low particular cases (when the geometry and the mode of operation or medium alloyed steel ) by the AES technique. In these of the source permit), the sample atoms and ions can be sources, rf coupling was achieved in diVerent electrode geoliberated directly from the matrix through thermal evapo- metries, but in both cases the sample was one of the discharge ration, sputtering or ablation.The idea of solid sampling is electrodes in direct contact with the plasma. The possibility of based on the fact that in devices where the discharge is in easy adaptability to solid sampling requirements and direct contact with the electrodes (or with the sample), material analysis of conductive solid samples was highlighted.The rf expulsion could take place. For this situation, the vaporised CCP discharge also has the advantage that it could be mainmaterial is entrained in a flow of gas and transported to the tained at low powers (from 300 down to 10W), at atmospheric plasma spectral source. pressure and with low gas consumption (usually1 l min-1).9 Without any doubt, direct solid sampling provides a good In previous publications from our laboratory, the developopportunity to avoid lengthy sample preparation procedures ment and characterisation of atmospheric pressure rf CCP torches for AES in various electrode configurations (tip–ring, tube–single ring, tube–double ring) were described.10–15 These torches have been used mainly for the analysis of liquid †Presented at the 8th Solid Sampling Spectrometry Colloquium, Budapest, Hungary, September 1–4, 1998.samples. The rf CCP achieved with these torches could be J. Anal. At. Spectrom., 1999, 14, 541–545 541Fig. 1 Experimental set-up.modified in order to carry out atmospheric pressure rf Fig. 2 The sputtering chamber. 1=Tungsten electrode, diameter sputtering of non-conductive solid samples. 4 mm, 45° (connected to rf ); 2=upper PTFE lid; 3=PTFE electrode In this paper, we present results obtained by applying an rf holder, 12 concentric holes, 1 mm id; 4=quartz chamber, 14 mm id, CCP discharge for solid sampling and direct analysis of non- 100 mm length; 5=PTFE sample pellet holder cup; 6=brass counter conductive samples, viz., oxides and silicates.The original electrode; 7=lower PTFE lid; 8=Ar inlet; 9=Ar outlet; 10=lateral tip–ring rf CCP design11 has been modified to provide the port, 5 mm id, 50 mm length. sampling and excitation conditions. This modified rf CCP torch allows both solid sampling of non-conductive samples electrode holder (12 concentric 1 mm id holes). The sample and direct analysis of such samples by AES. pellet was placed inside a PTFE sample holder cup on a plate brass counter electrode.The rf electrode and the sample holder were placed at the opposite sides of the quartz tube and fixed Experimental with two PTFE lids. The plasma viewing is ensured by the Instrumentation lateral port. The experimental set-up is shown schematically in Fig. 1 and Standards and sample pellet preparation details of the equipment used and the operating conditions are provided in Table 1. The experimental measurements were carried out on cylindrical sample pellets of two representative matrices: oxides (ZnO) The capacitively coupled plasma used for rf sputtering of non-conductive solid samples was obtained inside a tubular and silicates (andesite).For studying the analytical performance, the following quartz chamber (14 mm id, 100 mm length, with a lateral port of 5 mm id, 50 mm length) on a sharp tungsten electrode materials were used: ZnO powder (puriss. 99+%, from Research Institute for Semiconductor Materials, Bucharest, (4 mm diameter, 45°) (see Fig. 2). The plasma gas (Ar) was introduced inside the chamber via a gas inlet and a PTFE Romania) with trace elements (Si, Pb, Cd and Na) and as certified reference materials four andesite standards (Japanese, JA-1, JA-2, JA-3; and American, AGV-1) with certified con- Table 1 Instrumentation and operating conditions tents of Pb, Cu and Cr. Also, synthetic (laboratory made) Plasma generation Rf oscillator with inductive reaction and andesite matrices (57% SiO2, 18% Al2O3, 3% MgO, 7% CaO, grid tuned circuit: 13.56 MHz; 10–80 W. 3% K2O, 3% Na2O and 9% Fe2O3) were prepared with known Laboratory constructed contents of Pb (3, 10, 30, 100, 300 and 1000 mg g-1). Using Stabilised power supply BS 452 E type these synthetic and certified standards, calibration curves for (2×500 V; max. 400 mA). TESLA, Brno, Pb were constructed to establish the dynamic range and verify Czech Republic the analytical method.Stabilised power supply I 4104 type (40 V; 5 A). IEMI, Bucharest, Romania All the oxides were of high purity (Johnson Matthey Sputtering chamber and Laboratory constructed (Fig. 2) Chemicals, Royston, Herts., UK). plasma support gas High purity Ar, 0.1–1 l min-1. Azo-Mures, The raw materials necessary for pellet preparation were Tg. Mures, Romania mixed with 30 ml of ethanol in a mortar and stirred until Optics 110 mm focal length, 30 mm diameter homogeneous. This procedure was continued until the whole fused silica lens amount of the added alcohol had evaporated, and this pro- Monochromator Computer driven scanning type (1 step= 0.002 nm), 1 m focal length, Czerny– cedure was repeated twice.Drying at 105 °C was followed by Turner mount, with 2400 grooves mm-1 7 h of sintering in a furnace at 1000–1100 °C. After a new set diVraction grating and 20 mm slits (internal of grinding and sifting steps, the sample was homogenised wavelength calibration with an Si hollow again.Some powder obtained in this way was pressed in a cathode lamp). Research Institute for steel press at a pressure of 87×105 Pa for 10 min. The resulting Analytical Instrumentation, Cluj-Napoca, analytical sample pellets had a cylindrical shape, with a Romania Detector 9781 R photomultiplier tube operated at diameter of 11 mm, and were used for analytical 700 V. Thorn EMI, Ruislip, Middlesex, determinations. UK Driving, data acquisition Digital data acquisition and monochroma- Results and discussion and processing tor driving carried out by an IBM PC equipped with a laboratory constructed Atomisation mechanism of the sample pellets. Parameter interface (64 ms data acquisition time); optimisation data processing with appropriate in-house software.Research Institute for Analytical The main characteristic of our capacitively coupled plasma is Instrumentation, Cluj-Napoca, Romania that it is an intrinsic part of the tuned circuit of the rf generator 542 J.Anal. At. Spectrom., 1999, 14, 541–545Fig. 3 The plasma and the tuned circuit of the rf oscillator. Rg, Cg= automatic negative group; L=adjustable coupling coils; C=condenser of the oscillating circuit. (Fig. 3). The oscillator is of the inductive reaction type having the tuned circuit placed in the control grid network of the active element (pentode tube) and it is presented in detail elsewhere.16 The presence of the adjustable inductive coupling in the reaction circuit oVers the possibility of maximisation of the rf power transferred to the plasma.The automatic grid bias network (RgCg) develops a negative voltage on the grid in the range 150–200 V. Its value is a function of the anode positive bias. Over this negative dc component the rf component is superimposed, the amplitude of oscillation being 15–20% greater than the dc potential of the control grid. Because the oscillator works under resonant conditions, the amplitude of the rf oscillations on the sustaining electrode of the plasma is in the range 2000–3000 V.Owing Fig. 5 EVect of rf power on the emission intensities at an Ar flow rate to the dc component, the rf wave is translated towards negative of 0.5 l min-1 for (a) 20 mg g-1 Na (588.99 nm) from ZnO matrix and values, its form being asymmetric with respect to ground. (b) 35.1 mg g-1 Pb (405.78 nm) from AGV-1 andesite standard. During a full cycle of the rf wave, the plasma sustaining electrode has at negative potential a time interval longer than ation.At suYciently high rf powers (38 W in the case of an a half-cycle. This fact, combined with the lower mobility of argon plasma), the heating of the sample is so great that it the positive ions relative to the electrons, causes an accumutends to show incandescence. We assume that the increase in lation of positive charge greater than the negative charge close the sample temperature causes thermal evaporation on its to the sustaining plasma electrode.This accumulation of surface. This can be a supplementary atomisation mechanism positive charge has two consequences: (a) an excess of negative which will cause an increase in the number of sample atoms charge that appears toward the free end of the plasma where in the plasma. These observations are supported by the plots the sample is placed and (b) electrostatic shielding of the in Fig. 4 and 5. sustaining electrode of the plasma.Consequently, an internal Fig. 4 shows the dependence of the expulsion rate on the rf dc electric field appears in the plasma. When the distance power absorbed by the plasma. The expulsion rate was calcu- between the plasma electrode and the sample is 2–4 mm, this lated as the diVerence between the ZnO sample mass before field is suYciently intense to accelerate the positive ions and after exposure to collisions with ions. Nine replicate towards the sample and to induce sputtering of the sample, sample pellets were used.At powers lower than 25 W, the causing sample atomisation. Sputtered atoms are then availsputtering rate is low since the intensity of the dc internal field able to enter the plasma for subsequent excitation and ionis- Fig. 6 EVect of Ar flow rate on the Pb (405.78 nm) emission intensity Fig. 4 Expulsion rate (mass loss per minute) as a function of rf power for a ZnO sample pellet. for the AGV-1 andesite standard (35.1 mg g-1) at an rf power of 36 W.J. Anal. At. Spectrom., 1999, 14, 541–545 543Table 4 Concentrations and statistics for Pb in standard andesite Table 2 BEC and LODs for trace elements in ZnO and andesite matrices matrices Pb concentration/mg g-1 Statistics Element l/nm BEC/mg g-1 LOD/mg g-1 Matrix Na 588.99 11.4 0.3 ZnO Andesite Certified Determined standard value value c: ±sc tcalculated ttabulated (n; 95%) Pb 405.78 21.4 0.7 ZnO Si 251.61 33.3 1.0 ZnO Cd 228.81 28.1 0.9 ZnO JA-1 5.8 5.5±0.7 0.96 2.78 JA-2 19.3 19.9±0.6 2.23 2.78 Pb 405.78 23.1 0.8 Andesite Cr 425.43 31.5 1.0 Andesite JA-3 6.7 7.0±0.7 0.96 2.78 AGV-1 35.1 34.8±0.6 1.12 2.78 Cu 324.75 14.5 0.5 Andesite Student factor |tcalculated|=(c: -m)Óm/sc.c: =Average concentration of m=5 successive measurements; sc=standard deviation for m=5; m= certified concentration; n=m-1 degrees of freedom. is low. In the range of 25–38W, the sputtering rate is nearly constant, which means that the intensity of the accelerating field remains nearly constant.An explanation of this phenomenon could be the complementary eVects of the increases in ology, respectively, both elaborated by Boumans et al.17,18 the rf voltage amplitude and in the electrostatic shielding. In The relative standard deviation of the background (RSDB) this power range the sample atomisation is produced only by was calculated for 10 successive measurements of the the sputtering process. At powers higher than 38 W, thermal background.evaporation will be present because of heating of the sample As can be seen in Table 2, the detection limits are1 mg g-1. by energised ions. Hence an increase in the atomisation These results are similar to other results obtained for the direct rate occurs. analysis of solid samples.4 In Fig. 5 the eVect of rf power on the emission intensities The dynamic range was determined for Pb from andesite of Na (588.99 nm) from a ZnO matrix and Pb (405.78 nm) matrices.It is about three orders of magnitude. The correlation from AGV-1 andesite is shown. coeYcient for the calibration curve is about 0.999 for this As can be seen, each plot is composed of two straight lines dynamic range and the standard deviation of its slope is 2.8%. with diVerent slopes. The first line corresponds to the For the statistical comparison of the two calibration curves sputtering mechanism and the second to the combination of (F-test and t-test), the null hypothesis was used.19 The statistics sputtering and thermal evaporation.Observing that the of the calibration curves for Pb in andesite matrices are increase in the slope takes place at an rf power of 37–38W, presented in Table 3. which is the same as when the atomisation rate begins to As shown in Table 3, the calculated values for F and t are increase, the thermal evaporation assumptions seem to be smaller than those tabulated, so there are no significant plausible. diVerences between the two curves for a probability of 95%.In Fig. 6, the eVect of the gas flow rate on the Pb Therefore, the Pb concentrations in the certified standards (405.78 nm) emission line intensity from AGV-1 andesite at were determined using the calibration curve obtained with an rf power of 36W is shown. The plot has a maximum at a the synthetic standards. Subsequently, the recoveries and the gas flow rate that can be considered optimum (0.5 l min-1).RSDs of concentrations were also calculated. In Table 4, the At this gas flow rate, the number of sputtered atoms entering results and statistics for Pb determination (five successive the plasma, where the excitation and atomisation processes measurements) in certified standards are given. take place, is maximum. At gas flow rates greater than the Table 4 shows that for each standard, tcalculated<ttabulated. optimum, the residence time of atoms in the plasma decreases Therefore, for a probability of retaining 95% and n=4 degrees and, as a result, the net intensities of the emission line of freedom, the null hypothesis is valid and there are no also decrease.systematic errors between the certified and determined values of the concentration using the plotted calibration curve. Analytical performance The RSDs for Pb in andesite are 12.7% for JA-1, 3.1% for JA-2, 10% for JA-3 and 1.7% for AGV-1. The recoveries for All measurements for evaluating the analytical performances were made at an rf power of 36W and an Ar flow rate of Pb compared with the certified values are 95±12% for JA-1, 103±3% for JA-2, 104±10% for JA-3 and 99±2% for AGV-1. 0.5 l min-1. In Table 2, the background equivalent concentration (BEC) The determination errors are much higher at lower concentrations, but for concentrations higher than 10 mg g-1 Pb, they and the limits of detection (LODs) are presented. They were calculated with the 3s method and the BEC–RSDB method- are very good, being around 2–3%.Table 3 Statistics for calibration curves for Pb in andesite matrices Intercept Slope Andesite No of Calibration curve equationa matrix points Fcalc Ftab b ttab d Fcalc Ftab b ttab d y=(a±sa)+(b±sb)c type plotted s12/s22 F(n1; n2) tcalc c (n; 95%) s22/s12 F(n2; n1) tcalc c (n; 95%) -4×10-3 ± 28.5+(7.1±0.2)c (r=0.999) Synthetic n1=6 5.74 F5;3=14.88 6×10-5 2.31 6.250 F3;5=7.764 0.45 2.31 -5×10-3 ± 11.9+(7.2±0.5)c (r=0.997) Certified n2=4 standard aa=Intercept; sa=standard deviation of the intercept; b=slope; sb=standard deviation of the slope; r=correlation coeYcient.bn1=n1-1, n2= n2-1 degrees of freedom. ct=(x: 1-x: 2)/sÓ(1/n1+1/n2); s=Ó[(n1-1)s12+(n2-1)s22]/(n1+n2-2) where x: 1 and x: 2=intercepts and slopes of the calibration curves (a1; a2; b1; b2); s1 and s2=standard deviations of the intercepts and slopes of the calibration curves. dn=n1+n2-2 degrees of freedom. 544 J. Anal. At. Spectrom., 1999, 14, 541–54511 E.A. Cordos�, S. D. Anghel, T. Frent�iu and A. Popescu, J. Anal. References At. Spectrom., 1994, 9, 635. 12 T. Frent�iu, S. D. Anghel, A. M. Rusu, M. Ponta and E. A. Cordos�, 1 M. W. Blades, P. Banks, C. Gill, D. Huang, Ch. LeBlanc and Fresenius’ J. Anal. Chem., 1996, 355, 254. D. Liang, IEEE Trans. Plasma Sci., 1090, 19, 1991. 13 E. A. Cordos�, T. Frent�iu, A. Fodor, M. Ponta, A. M. Rusu and 2 W. W. Harrison, C. M. Barshick, J. A. Klinger, P. H. RatliV and S. Negoescu, ACH Models Chem., 1995, 132, 313. Y. Mei, Anal. Chem., 1990, 62, 943A. 14 T. Frent�iu, S. D. Anghel, A. Simon, A. Popescu, A. M. Rusu, 3 D. C. Duckworth and R. K. Marcus, Anal. Chem., 1989, 61, 1879. S. Negoescu and E. A. Cordos�, ACH Models Chem., in the press. 4 R. K. Marcus, T. R. Harville, Y. Mei and Ch. R. Shick, Anal. 15 T. Frent�iu, S. D. Anghel, A. Simon, A. Popescu and E. A. Cordos�, Chem., 1994, 66, 902A. ACH Models Chem., in the press. 5 A. M. Pless, A. Croslyn, M. J. Gordon, B. W. Smith and 16 E. Ta¢ taru, S. D. Anghel and I. I. Popescu, Rev. Roum. Phys., 1991, J. D. Winefordner, Talanta, 1997, 44, 39. 36, 29. 6 W. R. L. Masamba, B. W. Smith and J. D. Winefordner, Appl. 17 P. W. J. M. Boumans, Spectrochim. Acta, Part B, 1991, 46, 431. Spectrosc.., 1992, 46, 1741. 18 P. W. J. M. Boumans, J. C. Ivaldi and W. Slavin, Spectrochim. 7 D. Liang and M. W. Blades, Spectrochim. Acta, Part B, 1989, Acta, Part B, 1991, 46, 641. 44, 1049. 19 J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, 8 S. D. Anghel, T. Frent�iu, A. M. Rusu, L. Bese and E. A. Cordos�, Wiley, New York, 2nd edn., 1988, ch. 3, pp. 53–62 and ch. 5, Fresenius’ J. Anal. Chem., 1996, 355, 252. pp. 101–115. 9 M.W. Blades, Spectrochim. Acta, Part B, 1994, 49, 47. 10 S. D. Anghel, T. Frent�iu, A. Simon, E. Darvasi, A. M. Rusu and E. A. Cordos�, Fresenius’ J. Anal. Chem., 1996, 355, 250. Paper 8/07036I J. Anal. At. Spectrom.,