Characterization of a Magnetron Radiofrequency Glow Discharge With a Glass Cathode Using Experimental Design and Mass Spectrometry C. MOLLE M. WAUTELET J. P. DAUCHOT AND M. HECQ Universit6 de Mons-Hainaut Avenue Maistriau 23 B-7000 Mons Belgium A magnetron radiofrequency-powered glow discharge with a borosilicate glass cathode has been interfaced to a quadrupole mass spectrometer and an energy analyser. The influence of the electrical discharge parameters on the ion energies of Ar and Si and on the intensity of the signals has been studied as well as the influence of the position of the extraction orifice (along or perpendicular to the discharge axis). This study was carried out via an 'experimental design' procedure. For both orifice positions the ions entering the mass filter have minimum energy at 60 mTorr (1 Torr = 133.322 Pa) when the pressure is increased from 10 to 450 mTorr.The ion energy does not depend significantly on the power. When the orifice position is along the discharge axis the ion energy doubles when the orifice-cathode distance is decreased from 5 to 2.5 cm; above 5 cm the distance does not influence the energy. In the lateral position the ion intensity increases when the pressure goes from 10 to 130 mTorr; above this pressure the intensity remains constant. Maximum intensity is observed at 100 mTorr when the quadrupole is positioned along the discharge axis. For both positions the intensity rises by less than one order of magnitude with an increase in power of from 20 to 70 W. An important relationship between the parameters bas been found in the axial position an increase in the pressure leads to a rise in the ion intensity for a small orifice- cathode distance whereas for a larger distance maximum intensity is observed at 60 mTorr.Keywords Glow discharge mass spectrometry; magnetron radiof requency glow discharge; experimental design Glow discharge mass spectrometry (GDMS) is recognized as a powerful method for the elemental analysis of solids down to the pg kg-' range.' Most studies on this subject have dealt with d.c. discharges and required an electrically conducting material. Techniques that would enable the study of insulators to be carried out have been proposed such as mixing an insulating powder with a conducting matrix.lP6 Although this method works well with some materials most (e.g.glass and ceramics) cannot easily be rendered into a powdered form. Moreover the sputtering of pressed samples often leads to spectral interferences in GDMS owing to the matrix and to adsorbed species. This is a serious limitation mainly when working with low-resolution instruments.2*6 Insulating mate- rials can be analysed directly by radiofrequency (r.f.) dis- charge~.~-'' With r.f. discharges work can be carried out at lower pressure than in d.c. discharges so that the mean free path of the particles is increased and the redeposition of sputtered material is reduced.l2 With low-pressure discharges it is also expected that the formation of polyatomic species in the gas phase is reduced.13 A further improvement in the method is to use a magnetron discharge where the electrons are confined by a magnetic field.This allows the discharge to be maintained over a large pressure range (0.005-1 Torr; 1 Torr= 133.322 Pa) and a lower Journal of Analytical Atomic Spectrometry voltage and a higher current can be used.I4 The confinement of the electrons near to the cathode results in increased efficiency of the ionization of the gas and of the sputtering rate. However the geometry of the magnetic field leads to localized erosion of the cathode. When as in the present arrangement the magnetic field has a cylindrical symmetry the sputtering is not uniform and leads to annular erosion. An r.f. planar magnetron discharge in Ar has also been studied by optical emission spectr~metry'~ and mass spec- trometry in our laboratory.16 Heintz et al.17 have examined the influence of a magnetic field in an r.f.glow discharge on spatial emission features. A planar magnetron glow discharge device has been employed for mass spectrometry measurements in the d.c. rnode.l3 It is well established that glow discharges are influenced by various parameters (power pressure and sampling distance). These dependences are generally non-linear and interrelated. In the present paper empirical relationships between the measured parameters for GDMS (intensity and energy distri- bution of the species) and the external parameters (power pressure and distance from the cathode) are deduced by using an experimental design method.'* It is shown that the measured parameters can be fitted by a quadratic polynomial as a function of the external parameters.The quadrupole mass spectrometer can be placed either laterally or axially relative to the axis of the discharge. A comparison of the results obtained under the two experimental set-ups is performed. EXPERIMENTAL A schematic drawing of the discharge chamber is shown in Fig. 1. The 1 in diameter cathode (US Gun Campbell CA USA) is powered by a 13.56 MHz Advanced Energy RFX 600 (Advanced Energy Industries Fort Collins CO USA) capaci- tively coupled with a matching netbox Advanced Energy ATX 600. The cathode is cooled by flowing octane in order to reduce the dielectric losses and can be displaced vertically over 6cm. In the planar magnetron device the cathode contains two concentric circular magnets which supply a f230 G (1 G = 1 x loP4 T) magnetic field above the glass sample.Torr by a Balzers TPH 450H turbo pump. The Ar gas flow is controlled by a Brooks mass flow meter. The discharge pressure is controlled by adjusting the pumping rate uia an MKS throttle valve and controller. An MKS capacitance manometer gauge measures the sputtering pressure in the range 1 x - 1 Torr. Lower pressures are measured with an ionization gauge. The mass spectrometer is a differentially pumped Balzers PPM 420 quadrupole equipped with cylindrical ion transfer optics the chromatic aberration of which is used for the energy analysis (Fig. 1). This arrangement makes mass and energy spectrometry possible the mass spectrum with a pre-selected ion energy as the parameter or the energy spectrum with the The vacuum chamber is pumped down to 1 x Journal of Analytical Atomic Spectrometry December 1995 VoZ.10 10391 m Fig. 1 Schematic representation of the r.f. glow discharge electrical interface and mass spectrometer system 1 gas inlet; 2 turbomolecular pump; 3 cathode magnetron; 4 extraction orifice; 5 ion lenses; 6 ion source; 7 quadrupole; 8 deflection; 9 secondary electron multiplier; 10 electrometer or ion count; 11 turbomolecular pump; 12 cooling tubing; 13 copper conductor; 14 female coaxial connector; 15 male coaxial connector (r.f. input type 'W); 16 RG-393 coaxial cable (61 cm length); 17 marked scale; and 18 cathode translator mass as the parameter. The bandpass of the ion optics is a constant of about 1.5eV over the whole energy range.The transmission of the ion optics is about 40% and the trans- mission of the quadrupole mass filter is about 35%.19 The extraction orifice 0.1 mm in diameter can be operated either floating or polarized relative to the ground or to a variable voltage power supply (Vb). Depending on this voltage ions having a specific initial energy will enter the mass filter Neutral gas particles can be detected by the use of a Balzers cross beam ionization source located between the energy analyser and the mass filter. The mass filtered ion current is deflected at go" amplified through a secondary electron multiplier and detected by an electrometer.20 In order to increase the measured intensities the extraction orifice is polarized at + 15 V. As mentioned previously two geometrical positions of the mass spectrometer were studied laterally and axially relative to axis of the discharge.The sample is made of 0.5 mm thick barium borosilicate glass (Corning Glass 7059). Experimental Design Procedure In principle several factors can influence the production of ions in the discharge and sampling by the mass spectrometer for example pressure ( p ) electric power (W) geometry of the discharge gas flow and thickness of the cathode. The traditional method of investigation is to change one variable at a time while keeping the others constant which requires a large number of experiments. The information obtained could be incomplete and the predictions are not reliable when care is not taken to consider the interactions between the factors. The statistical design (changing several variables from one run to the next) the so called 'experimental design' is more appropriate for process development and optimization studies.In general the statistical method is more efficient than the classical method and the interactions between parameters can easily be estimated. The experimental design procedure is divided into two parts firstly a screening design is applied in order to identify the factors which will have an influence; and secondly the response surfaces are calculated as a function of the selected factors.21 The responses considered here are the detector current and the voltage corresponding to the maximum ionic current- voltage characteristic plot. Typical ionic current-voltage curves are shown in Fig. 2.In previous work in this 1aboratory,l6 the screening design was applied to GDMS studies of an alu- minium base alloy with the quadrupole in the lateral position. It was shown that of the various factors the most important ones relevant to the intensity and energy of the ions are p W and the thickness of the cathode. In the present work it has been assumed that these same factors remain the most import- ant ones. However at high power erosion of the glass sample cannot easily be evaluated since its shape varies as a result of the formation of molten zones upon sputtering. Hence the thickness of the cathode was not introduced as a parameter into the experimental design. In our new apparatus a better cathode cooling system is in operation so that there are as yet no molten zones in the glass samples.On the other hand a new factor has to be taken into account in the case of the axial position of the mass spectrometer this is the distance (d) between the cathode and the extraction orifice. It is worth noting that in the lateral position the cathode position relative to the orifice is constant ( Z = 3.6 and Y = 5.5 cm Fig. 1). The 3.0 1.5 U Y g o i > .r f 10 .- E t a 5 0 10 20 + I + + + I m m + + - + + 3.0 0.5 a z 0 30 g 3.0 Fig. 2 Measured Ar' (M) and Si' (+) ionic current-voltage V characteristics at 450 mTorr and 45 W (a) for the lateral position; and (b) for the axial position (sample-orifice distance = 4 cm) of the mass spectrometer 1040 Journal of Analytical Atomic Spectrometry December 1995 Vol. 10orifice voltage kept constant at + 15 V was not taken as a factor in the experimental design because a variation of this voltage caused a relatively weak effect on the responses studied.Moreover an additional factor to the experimental design should increase greatly the number of experiments to be performed. In general the original values of the factors are not used directly. They are translated into coded values in such a way that the origin of the coded factor space is the centre of the experimental domain. The original factors ( U ) are transformed into coded factors (X) where U is the value of the factor j during experiment i Ui(0) is the value of the same factor at the centre of the experimental domain and AUj is the variation step. The coded values are dimensionless.The correspondence between the original values and the coded values is given in Tables 1 and 2 as well as the factor levels and the experimental results. In Table 2(b) the values given for the sample-orifice distance are those read at the marked scale on the cathode translator. The values of 6.5 and 2.5 in parentheses are the true distances between the orifice and the cathode. In general the observed responses can be fitted by a quadratic equation.22 For the lateral quadrupole position the following model is used to approximate the true relationships between the discharge parameters and the responses (ion intensity and energy) over the experimental domain Response = bo + blxli + b2x2i + bllxfi + b22~$i + b,2x,ix2i (2) where xli is the log of the pressure and x2i is the r.f.power for run i. For the axial quadrupole position the quadratic equation is Response = b + blxli + b2xZi + b ~ ~ + b ,xfi + bz2x& + b33~32i + b12XliX2i + b13XliX3i + b23X2iX3i ( 3 ) where xli is the log of the pressure x2i is the r.f. power and x ~ is the cathode-orifice distance for run i. When the factors are replaced by their coded counterparts in the empirical model the coefficients for bjk namely the effects (see Table 3 Table 1 Lateral quadrupole position (a) Factor levels and observed responses- Coded value of the factor Observed response Run Si+ energy/ Log number Log(p) Power v (Si+ intensity/A) 1 2 3 4 5 6 7 8 9 10 11 12 -1 1 0 0 - 1 - 1 1 - 1 0 0 1 1 - 1.4 0 0 - 1.4 0 0 1.4 0 0 0 0 1.4 20.4 17 21.1 18.3 17.5 18.6 28.4 17 17.3 24.3 16.9 16.7 -7.17 - 7.49 -8 -7.19 -7.12 - 6.8 - 9.21 - 7.62 - 7.21 -7.1 - 7.12 - 6.89 (b) Correspondence between coded values and original values- Factor Coded value Original value Unit Log( pressure) - 1.4 1 mTorr 1.4 2.6 1.4 70 R.f.power - 1.4 20 w and 4) can be compared with each other in order to deduce the relative importance of the various factors. To estimate the parameters a central composite design was used (Design Expert 4.0 Stat-Ease Minneapolis MN USA). Experiments at the centre of the experimental domain are added in order to verify the reproducibility of the results. When the mass spectrometer is in the lateral position the factorial part is a 22 design while it is a 23 design in the axial position. In both cases the matrices are rotatable and almost orthogonal.For each design the same glass sample was used for all experiments. The measured data are fitted by means of a least-squares regression program. Model adequacy is checked by analysis of the residuals on a normal probability plot and analysis of variance (ANOVA Design Expert 4.0). A normal probability plot of the residuals is a graph where the residuals are on the x-axis from smallest to largest and the y-axis is a normal probability scale which ‘straightens out’ the plot of a cumulat- ive normal distribution. If the residuals fall approximately along a straight line they come from a normal distribution. In this case the model is assumed to be valid. Some results from other statistical tests to check the validity of the model are shown in Tables 3 and 4.For example one test compares the error related to the use of this model (lack of fit of the mean square LoFMS) with the pure error (pure error mean square PEMS) obtained from replicated design points. This step is carried out in the ANOVA procedure with a statistical test on the ratio ( F ) of the LoFMS and PEMS variances. If the F value calculated is larger than the F value tabulated (for a probability level of 0.05) then the variance associated with the model is significantly larger than the experimental error and the model is not valid. The I; ratio is significantly large when the probability of a larger F value (Prob>F in Tables 3 and 4) is lower than 0.05. On the other hand the model is valid if Prob>F is larger than 0.05. Other tests are the coefficient of determination (R-squared) showing the pro- portion of variability in the data accounted for by the model and the coefficient of variation (CV) or the relative standard deviation of the residuals (standard deviation as a percentage of the mean of the response over all cases). For the models pertinent to the lateral position of the quadrupole (Table 3) the results of the statistical tests show that there is no significant difference between the experimental errors and the errors due to the lack of fit (Prob > F >0.05).The coefficient of determination is larger than 0.8 indicating that 80% of the variation in the observed values can be explained by the chosen model. The CV is less than 5% which means that the relative standard deviations of the residuals (difference between the experimental value and the fitted value) are less than 5% of the average response.It should be noted that the insignificant coefficients have been removed in a step- wise regression procedure (Design Expert 4.0). For the models pertinent to the axial position of the quadru- pole (Table 4) the statistical tests are calculated from the full model. For the Si’ intensity it can be considered that the model is valid. Indeed even if Prob> F is less than 5% the R-square and the CV values are good (>94 and <3% respectively). The Si’ energy model is less favourable but this model is kept because of the R-square value ( ~ 9 4 % ) and a random residuals distribution (normal probability plot). RESULTS AND DISCUSSION Energy Analysis It is known that two factors determine the kinetic energy of the ions sampled by the mass spectrometer the potential difference between the plasma (at the plasma potential V,) and the orifice (at + 15 V in the present work) and the extent of the collisions in the intermediate region between the plasma Journal of Analytical Atomic Spectrometry December 1995 Vol.10 1041Table 2 Axial quadrupole position (a) Factor levels and observed responses- Coded value of the factor Observed response Run Sample-orifice Si+ energy/ number Log(P) Power 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 -1 0 0 0 -1 0 0 -1 1 1 0 1 0 - 1 - 1.68 1.68 0 1 -1 1.68 0 0 1 0 0 1 1 1 0 -1 - 1.68 -1 0 0 0 -1 (b) Correspondence between coded values and original values- Factor Coded value Log( pressure) R.f.power Sample-orifice distance - 1.68 1.68 1.68 1.68 - 1.68 - 1.68 distance 1 0 0 - 1.68 1 0 0 -1 -1 1 1.68 1 0 - 1 0 0 0 -1 V 44.7 27.8 26.2 21.2 48.7 24.8 24.7 27.2 26.1 36 58.8 37.2 26.5 27.6 31.5 34.6 23.5 23.1 Original value 1 2.6 20 70 0 (6.5)* 4 (2.5) Log( Si+ intensity/A) - 7.45 - 6.42 - 6.63 - 7.72 - 7.03 - 6.53 - 6.66 - 7.07 - 7.43 - 6.49 - 6.64 - 6.57 - 7.07 - 7.46 - 7.88 - 6.74 - 6.64 - 7.68 Unit mTorr w cm * Values in parentheses are the true distances between the orifice and the cathode. Table 3 Lateral position. Coefficients of the polynomial and statistical tests Response Si+ energy/V Prob > F Response Log( Si+ intensity/A) Prob > I; b0 bl b2 bll b22 b0 bl b2 bll 17.18 - 1.3 -0.1 4.1 1 -0.67 0.106 R-squared 0.9728 cv 4.08% -7.16 0.52 1 0.282 - 0,376 0.212 R-squared 0.8238 cv 4.56% Table 4 Axial position.Coefficients of the polynomial and statistical tests Response bo bl b2 b3 bll hZ2 b33 b12 b13 b23 Si+ energy/V 24.8 1 - 1.51 0.56 9.22 2.9 0.81 5.36 -0.23 -1.82 0.03 Prob > F 0.021 R-squared 0.9381 CV 11.58% Response b0 bl b2 h3 bll b22 b33 b12 b13 b23 Log(%+ intensity/A) -6.613 0.202 0.164 0.287 -0.255 -0.055 -0.209 -0.06 0.25 -0.017 Prob > F 0.032 R-squared 0.9438 CV 2.38% and the sampling orifice. The magnitude of the potential difference between an insulated surface and the plasma is given for a Maxwellian electron energy distribution by:23 b - K = 2 p l n ( G ) kT 2Mi =5.372 (kT,) (for an Ar plasma) (4) where V is the floating potential T is the electron temperature e is the charge of the electron Mi and me are the ion and electron masses and k is the Boltzmann constant.In the case of an Ar plasma an electronic temperature of 1 eV leads to a potential difference of 5.372 V. However the electrons show departure from a Maxwellian distribution owing to a non- energetic equilibrium with the ions and neutral species and also with their own ensemble in low pressure r.f. discharge^;^^ therefore eqn. (4) is an approximati~n.~~ The voltage corresponding to the maximum energy distribution (Fig. 2) has been studied as a function of several factors (Tables 3 and 4). The calculated response curves for Si' are shown in Figs 3 and 4. It would appear that whatever the position of the spectrometer the energy is at a minimum at about p = 60 mTorr [Fig.3(u) and (b)]. In a previous study on an aluminium base alloy with the mass spectrometer orifice at the floating potential and located in the lateral position,16 the Al' energy decreased by about 12V (V voltage) when the pressure increased from 10 to 450 mTorr. This behaviour was expected because ( Vp - 5) correlates with the electron tempera- ture [eqn. (4)] and the electron temperature decreases when the pressure increases.26 From Figs. 3 and 4 it can be seen that the Si' energy drops by about 1OV when the pressure varies from 10 to 60mTorr. This decrease could be also attributed to a dependence of T and therefore & on the pressure. The increasing Si' energy above 60 mTorr cannot be explained for the moment. It could be that there is an effect of the orifice polarization on the plasma when the pressure is 1042 Journal of Analytical Atomic Spectrometry December 1995 Vol.1024.00 4 SO.OO\ I - Fig. 3 Calculated response curves for the Si' voltage (V,) uersus pressure and r.f. power (a) for the lateral position; and (b) for the axial position (sampleorifice distance = 4 cm) of the mass spectrometer Fig.4 Calculated response curves for the Si+ voltage (V,) versus pressure and sample-orifice distance in the axial position (r.f. power = 45 W) higher. Fig. 3(a) and (b) shows that the ion energies vary only slightly with the electric power. This is also anticipated from eqn. (4) because Cook and Das26 have shown by Langmuir probe studies that the electron temperature does not vary with r.f. power. For the axial position (Fig.4) the ion energy depends strongly on the sample-orifice distance ( d ) . The importance of this factor (Jb3 I = 9.22) is shown in Table 4 compared with factor p (lbll=l.51). The energy varies from about 30V (V voltage) for d=6.5 cm to about 65 V at d=2.5 cm while a variation in p leads to a maximum energy variation of about 7V. Given the fact that the ion energy is a function of the plasma potential (V,) Vp increases relative to the potential of the orifice when d decreases. This result conforms [eqn. (4)] to the results of Rossnagel and K a ~ f m a n ~ ~ who have shown that in a d.c. magnetron discharge T increases as the distance from the cathode decreases. From Fig.4 it appears that when d is larger than about 5 cm the energy of the Si+ ions remains almost constant.The same result was observed previously16 during the screening design in the lateral position for a conducting sample where the sampling distance Y (Fig. 1) ranged from 4.5 to 5.5 cm. Moreover when d is larger than 5 cm (Fig. 4) the energy of the ions is similar to that in the lateral position [Fig. 3(a)]. During the study of conducting targets,16 at high pressure the maximum in the distribution of Ar+ ions was shifted by 10 V towards a lower voltage as compared with the Al+. In the present case at 450mTorr the maxima of the Ar' and Si' distributions (Fig. 2) are separated by at the most 4 V. This could be due to the fact that when the voltage of the orifice is fixed at + 15 V the intermediate region thickness in front of the orifice is lower than in the case of a floating aperture.28 In this case the resonant charge exchanges between Ar' ions and Ar atoms are much less probable.Ion Intensity In the following 'intensity' ( I ) refers to the value of the measured current at the maximum of the energy distribution curves (Fig. 2). The intensity of the mass spectrometer signals depends on the density of the ions in the plasma but also on the effectiveness of the sampling of the discharge and the transmission in the mass spe~trometer.~~ Ion density and sampling efficiency depend also on the experimental conditions of the discharge and on geometrical fact01-s.~' The intensity at the maximum of the Si+ energy distribution has been studied as a function of several factors (Tables 3 and 4). The calculated response curves are shown in Figs.5 and 6. The variation of I with p appears to be important in both of the two positions. In the lateral position [Fig. 5(a)] I increases when p varies from 10 to 130 mTorr and remains constant above 130 mTorr. In the axial position [Fig. 5(b)] I goes through a maximum at about p = 100 mTorr. Whatever the position of the quadru- pole I increases by less than one order of magnitude with increasing W. In the axial position an interaction between p and d (Fig. 6 ) is observed. This effect (Jb13 1 =0.25 Table 4) is more important than that of p (Ib 1 =0.202). At low d I increases with increasing p while at high d I goes through a maximum at about p = 60 mTorr. Such a maximum of I near to 60 mTorr was also observed in the case of Al' during the study of an aluminium base alloy.16 From Fig.6 the variation of I with d can also be observed at low p I goes through a maximum near to d = 4 cm; at high p I decreases when d increases. An explanation of these observations is not obvious due to the particular geometry of the device when the sampling orifice is both in the discharge axis and near to the cathode. Indeed the orifice could be considered as the anode and therefore to change the orifice-cathode distance could vary the magnitude of the electric field. Moreover in the axial position the aperture is in a magnetic field free region. It is worth noting that the cathode dark space in a magnetron discharge is very thin (below 1 mm) whatever the pressure. For any value of d the orifice is always in the negative glow.At low d values the increasing Si+ intensity with pressure could be attributed to a Journal of Analytical Atomic Spectrometry December 1995 Vol. 10 1043$00 Fig.5 Dependence of log Si+ intensity on pressure and r.f. power (a) for the lateral position; and (b) for the axial position (sample-orifice distance=4 cm) of the mass spectrometer 2.600 0 Fig.6 orifice distance in the axial position (r.f. power = 45 W) Dependence of log Si' intensity on pressure and sample- larger electron confinement and therefore to a larger ioniz- ation when the pressure increases. This effect though less important is also observed at d=6.5 cm but only for a variation of pressure of from 10 to 60 mTorr; above 60 mTorr as the collisions are more important the ion flow towards the orifice is reduced.This also explains the decreasing intensity E+05 5 E+04 5 E+03 5 E+O2 5 a E+Ol 5 10 15 20 25 30 +a L a 75 . 70 . . C E+04 - 1 5 E+02 5 E+01 130 135 E+02 1 .o 0.9 0.8 0.7 0.6 0.5 0.4 4 35 40 45 50 55 60 L 80 85 90 95 100 140 145 150 155 + P a 200 201 202 203 204 205 206 207 208 209 210 m/Z Fig.7 R.f.-GDMS spectra of NIST SRM 1412 Glass Multicomponent (r.f. power = 45 W; Ar pressure = 100 mTorr). Elemental concentrations Si 19.81; Li,w2.09; B 1.41; Na 3.48; Al 3.98; K 3.44; Sr 3.85; Ba 4.18; and Pb 4.08% at high pressure when d increases and at lOmTorr when d varies from 4.5 to 6.5 cm. In the present work all of the measurements were made with the orifice voltage at + 15 V. As already indicated relative to a floating orifice the signal is increased but the improvement 1044 Journal of Analytical Atomic Spectrometry December 1995 Vol.10is rather weak. In addition owing to sputtering of the glass target an insulating film is deposited on the orifice. Thus in order to maintain a constant voltage the orifice must be cleaned periodically. This is why for all future experiments a floating orifice will be used. The maximum intensity is about the same for both positions of the quadrupole ( 4 x lop7 A in the axial position and 2.6 x lo-’ A in the lateral position) but the background in the axial position is about twice that in the lateral position (1 x and 5 x A respectively). Initially it appeared to be surprising that about the same amount of ions was collected for both configurations because the distance from the orifice to the target is shorter in the axial position.However it is worth noting what is fitted to the polynomial is the peak maximum and not the surface of the energy distribution. The distribution is about ten times larger in the axial configuration than in the lateral configuration the integrated signal is thus much higher in the axial configuration. Limits of detection calculated by the method described by bourn an^,^^ are less than 100 pg kg-’. In order to demonstrate the analytical capability of the magnetron discharge the mass spectra of National Institute of Standards and Technology (NIST) Standard Reference Material (SRM) 1412 Glass Multicomponent is shown in Fig. 7. The mass spectra are characterized as predominately atomic but contributions from metal oxides can be seen.Compared with spectra published by Shick et al.,” the ratio of the analyte signal to the background noise is higher with the present device. CONCLUSIONS In the present paper an experimental design procedure has been developed for the parametric evaluation of an r.f. mag- netron glow discharge. It has been shown that it is possible to obtain a model expression that translates the influence of a parameter (pressure power and sampling distance) on the ion intensity and energy rapidly. Interaction between the param- eters is also additional information available in the experimen- tal design approach. For example in the axial position it was found that an interaction between pressure and distance has a significant influence on the ion intensity. The observed response surfaces are consistent with previous studies on magnetron glow discharges.Variations in the ion energy can be explained by changes in the electron temperature. Future studies will involve the analysis of glass and ceramic materials. The limits of detection are in the pg kg-’ range but these could be lowered if isobaric interferences were removed. Quantification of the analysis will require mu1 tivariate calibration. This work was supported by the Ministere de la Region Wallonne Direction GCnerale des Technologies et de la Recherche and IRSIA (Institut pour TEncouragement de la Recherche Scientifique dans 1’Industrie et 1’Agriculture). REFERENCES 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 Raith A.Vieth W. Huneke J. C. and Hutton R. C. J. Anal. At. Spectrom. 1992 7 943. Tong S. L. and Harrison W. W. Spectrochim. Acta Part B 1993 48 1237. Caroli S. Senofonte O. Del Monte Tamba M. G. Cilia M. Brenner I. B. and Dvorochek M. Spectrochim. Acta Part B 1993 48 877. Duckworth D. C. Barshick C. M. Bostick D. A. and Smith D. H. Appl. Spectrosc. 1993 47 243. Duckworth D. C. Barshick C. M. and Smith D. H. J. Anal. At. Spectrom. 1993 8 875. Woo J. C. Jakubowski N. and Stuewer D. J. Anal. At. Spectrom. 1993 8 881. Duckworth D. C. and Marcus R. K. Anal. Chem. 1989,61,1879. Duckworth D. C. and Marcus R. K. Appl. Spectrosc. 1990 44 649. McLuckey S. A. Glish G. L. Duckworth D. C. and Marcus R. K. Anal. Chem. 1992,64 1606. Duckworth D. C. Donohue D. L. Smith D.H. Lewis T. A. and Marcus R. K. Anal. Chem. 1993 65 2478. Shick C. R. Jr. Raith L. A. and Marcus R. K. J. Anal. At. Spectrom. 1993 8 1043. Harrison W. W. and Bentz B. L. Prog. Anal. At. Spectrosc. 1988 11 53. McCaig L. Shi Z. Holbrook T. Brewer S. and Sacks R. Appl. Spectrosc. 1992 12 1762. Waits R. K. in Thin Film Processes eds. Vossen J. L. and Kern W. Academic Press New York 1978. Mehdi T. Legrand P. B. Dauchot J. P. Wautelet M. and Hecq M. Spectrochim. Acta Part B 1993 48 1023. Molle C. Springael S. Wautelet M. Dauchot J. P. and Hecq M. Appl. Spectrosc. 1995 49 872. Heintz M. J. Galley P. J. and Hieftje G. M. Spectrochim. Acta Part B 1994 49 745. Goupy J. L. Methods for Experimental Design Elsevier Amsterdam 1993. Hofler K. and Peter G. FL 9496 Balzers Liechtenstein personal communication. Hofler K. Methods of Plasma Characterization Balzers Technical Note Liechtenstein. Box G. E. P. Hunter W. G. and Hunter J. S. Statistics for Experimenters Wiley New York 1978. Feneuille D. Mathieu M. and Phan-Tan-Luu R. Mkthodologie de la recherche exptrimentale LPRAI Aix-Marseilles 1980. Turban G. Intkractions Plasmas Froids Matkriaux Les Editions de Physique Paris 1988 p. 79. Godyak V. A. Piejak R. B. and Alexandrovich B. M. Plasma Sources Sci. Technol. 1992 1 36. Cox T. I. Deshmukh V. G. Hope D. A. O. Hydes A. J. Braithwaite N.St. and Benjamin N. M. P. J. Phys. D Appl. Phys. 1987 20 820. Cook J. G. and Das S. R. Can. J. Phys. 1991 69 236. Rossnagel S. M. and Kaufman H. R. J. Vac. Sci. Technol. 1987 A4 2276. Jouan P.-Y. These de Doctorat Nantes 1992. Helm H. Mark T. D. and Lindinger W. Pure Appl. Chem. 1980 52 1739. Hecq M. Hecq A. and Fontignies M. Anal. Chim. Acta 1983 155 191. Boumans P. W. J. M. Anal. Chem. l994,66,459A. Paper 5/03076E Received May 15 1995 Accepted August 10 1995 Journal ojAnalytica1 Atomic Spectrometry December 1995 Vol. 10 1045