Radio Frequency Glow Discharge-Optical Emission Spectrometry For

Anal. Chem. 2004, 76, 1039-1044

Radio Frequency Glow Discharge-Optical Emission Spectrometry For Direct Quantitative Analysis of Glass Beatriz Ferna´ndez,† Nerea Bordel,‡ Rosario Pereiro,† and Alfredo Sanz-Medel*,†

Department of Physical and Analytical Chemistry, Faculty of Chemistry, University of Oviedo, Julian Claverı´a, 8, 33006 Oviedo, Spain, and Department of Physics, Faculty of Sciences, University of Oviedo, 33006 Oviedo, Spain

Direct solid quantitative analysis of glass by radio frequency (rf) glow discharge-optical emission spectrometry (GD-OES) by using special silicate glasses of very different composition (e.g., the SiO2 content varied from 10 to 70%, the CaO content from 0 to 25%, the Na2O content from 0 to 15%, the K2O content from 0 to 19%, etc.) for calibration is investigated in detail. The effect of sample thickness is studied by using calibration standards and samples of different thickness than that considered as “standard” in our experiments (3 mm). Considering the variable concentrations of easily ionized elements present in the different glass specimens, measurements of the Mg(II) 280.27-nm/Mg(I) 285.21-nm line intensity ratio for the calibration standards and samples containing magnesium were carried out to check the plasma robustness. Results showed that this indicator value remained virtually constant for all the glasses assayed. The proposed quantification scheme is based on the “constant emission yield” concept in which the measured analytical signals were corrected for thickness differences relative to the 3-mm “standard thickness”. In addition, better fittings to linear plots were obtained when using discharge Ar emissions as the reference signal. Results for the simultaneous determination of the glass major components (SiO2, Na2O, CaO, MgO, Al2O3, and K2O) by rf-GD-OES in two samples of different thickness (3 and 1.1 mm) showed good agreement with the expected results. Glow discharges (GDs) with detection either by optical emission spectrometry (OES) or mass spectrometry (MS) have proved to be very powerful techniques for the direct analysis of solids, including both homogeneous samples (bulk analysis) and coatings (in-depth profile analysis).1-3 Powering the spectrochemical source with direct current (dc), only electrically conductive solid materials can be sputtered (and thus directly analyzed). The * To whom correspondence should be addressed. Fax: 34-985103474. E-mail: [email protected]. † Department of Physical and Analytical Chemistry. ‡ Department of Physics. (1) Marcus, R. K., Broekaert, J. A. C., Eds.; Glow Discharge Plasmas in Analytical Spectroscopy; John Wiley & Sons Ltd.: Chichester, U.K., 2003. (2) Payling, R., Jones, D. G., Bengtson, A., Eds.; Glow Discharge Optical Emission Spectrometry; John Wiley & Sons Ltd.: Chichester, U.K., 1997. (3) Angeli, J.; Bengtson, A.; Bogaerts, A.; Hoffmann, V.; Hodoraba, V.-D.; Steers, E. J. Anal. At. Spectrom. 2003, 18, 670-679. 10.1021/ac035113q CCC: $27.50 Published on Web 01/21/2004

© 2004 American Chemical Society

analysis of nonconductors by dc-GDs requires either sample mixing with a high purity conductive binder4 (in such case, some of the associated problems are the dilution of the analyte with the binder, as well as the risk of contamination of the sample during the process of preparation) or the use of a secondary cathode.5 Using radio frequency (rf) GDs, both conductor and insulators can be directly sputtered,6-8 thus expanding enormously the field of potential applications of the GDs. The direct analysis of nonconducting materials with the rf-GD should be its great advantage versus the dc-GD counterpart. However, a lot of work remains to be done in the area of the quantification of nonconductors (e.g., glasses, insulators, ceramics, etc.). Issues such as the lack of commercially available rf-GDs capable of measuring or estimating both the dc-bias voltage and the electrical current generated when measuring nonconductors, the influence of sample thickness,9,10 or the not yet fully understood role played by certain elements commonly found in insulators (e.g., oxygen or nitrogen) seem to account for the present reluctance of researchers to investigate proper quantification schemes11 for nonconducting samples. Some interesting approaches to overcome such limitations could be the measurement of the true power in the plasma,12 the use of the argon emission signal to compensate the wrong measurement of power,13 etc. Concerning homogeneous glass samples, Pan et al.14 have investigated the elemental analysis of glass powder samples (pressed without a binder to form sample disks 1 mm thick) by rf-GD-OES, and data on limits of detection and sample-to-sample signal reproducibility have been provided. Furthermore, Anfone (4) Brenner, I. B.; Laqua, K.; Dvorachek, M. J. Anal. At. Spectrom. 1987, 2, 623-627. (5) Schelles, W.; de Gendt, S.; Van Grieken, R. E. J. Anal. At. Spectrom. 1996, 11, 937-941. (6) Marcus, R. K. J. Anal. At. Spectrom. 2000, 15, 1271-1277. (7) Payling, R.; Aeberhard, M.; Delfosse, D. J. Anal. At. Spectrom. 2001, 16, 50-55. (8) Hodoroaba, V.-D.; Unger, W. E. S.; Jenett, H.; Hoffmann, V.; Hagenhoff, B.; Kayser, S.; Wetzig, K. Appl. Surf. Sci. 2001, 179, 30-37. (9) Lazik, C.; Marcus, R. K. Spectrochim. Acta, Part B 1993, 48, 1673-1689. (10) Parker, M.; Marcus, R. K. Spectrochim. Acta, Part B 1995, 50, 617-638. (11) Report on round robin exercise with nonconductors. In Final General Meeting of the EU-GDS Network, Wiener Neustadt (Austria), March 2-6, 2002. (12) Marshall, K. A.; Casper, T. J.; Brushwyler, K. R.; Mitchell, J. C. J. Anal. At. Spectrom. 2003, 18, 637-645. (13) Anfone, A. B.; Marcus, R. K. J. Anal. At. Spectrom. 2001, 16, 506-513. (14) Pan, X.; Marcus, R. K. Mikrochim. Acta 1998, 129, 239-250.

Analytical Chemistry, Vol. 76, No. 4, February 15, 2004 1039

et al.13 investigated the analytical characteristics of the rf-GD-OES for the elemental analysis of bulk solid glass specimens. Analyte signal intensity variations due to sample thickness were studied and quantitatively corrected using as reference signal the response of Ar (I) emission lines. Unfortunately, the same matrix composition was used throughout those experiments, thus restricting the applicability of the proposed methodology to the use of matrixmatched standards. The constant emission yield concept has been widely employed for multimatrix calibrations15-19 addressed to the analysis of conducting materials with GD-OES. However, the use of emission yields has not been investigated so far for the quantification of nonconductors. In this paper, the capabilities of a rf-GD-OES for the quantification of bulk glass solid specimens of different thicknesses and different matrix compositions have been investigated for the first time by resorting to this constant emission yield concept using for calibration a set of silicate glasses with a wide range of concentrations of silicon, boron, sodium, calcium, magnesium, potassium, aluminum, etc. EXPERIMENTAL SECTION The commercial rf-GD-OES instrument JY5000RF from Jobin Yvon Emission, Horiba Group (Longjumeau Cedex, France) was used.20 The main body of the GD chamber was made of stainless steel. In all cases, a 4-mm-i.d. annular anode made of copper alloy was employed. The optical path of the spectrometer was nitrogenpurged and operated over the wavelength range 119-777 nm. A 0.5-m Paschen Runge polychromator was employed (at present, the polychromator used has photomultiplier tubes in 31 channels). In addition, the system was equipped with a Czerny-Turner monochromator (0.64 m), which allows increasing the instrument’s capabilities to any desired wavelength within the spectral range. Details of the polychromator and the monochromator are given elsewhere.20,21 Table 1 collects the analytical emission lines monitored throughout the experiments. For all the experiments, net analytical signals (i.e., background subtracted) were considered. The background intensity was measured at a fixed position (0.2-0.1 nm) beside the corresponding analytical emission wavelength. The forward rf power was the electrical parameter measured with the rf-GD-OES instrument used. The operational mode “constant pressure and constant forward power”20,21 was used throughout the experiments. The pressure was kept fixed at 450 Pa, and the forward power employed for the calibration curves was 20 W. The reflected power was, in all cases, lower than 1 W. Penetration depths into the different materials were estimated by measuring with a profilometer (Perth-o-meter S5P, Mahr Perthen) two profile traces in different directions across the center of each crater. Sputtering rates were evaluated as mass loss per sputtering time unit (measuring penetration depths per time unit (15) Bengtson, A.; Ha¨nstro ¨m, S.; Lo Piccolo, E.; Zacchetti, N.; Meilland, R.; Hocquaux, H. Surf. Interface Anal. 1999, 27, 743-752. (16) Bengtson, A. Spectrochim. Acta, Part B 1994, 49, 411-429. (17) Weiss, Z. J. Anal. At. Spectrom. 2003, 18, 584-589. (18) Weiss, Z.; Smı´d, P. J. Anal. At. Spectrom. 2000, 15, 1485-1492. (19) Pisonero, J.; Pe´rez, C.; Pereiro, R.; Bordel, N.; Sanz-Medel, A. J. Anal. At. Spectrom. 2001, 16, 370-375. (20) Ferna´ndez, B.; Bordel, N.; Pe´rez, C.; Pereiro, R.; Sanz-Medel, A. J. Anal. At. Spectrom. 2002, 17, 1549-1555. (21) Ferna´ndez, B.; Bordel, N.; Pereiro, R.; Sanz-Medel, A. J. Anal. At. Spectrom. 2003, 18, 151-156.

1040 Analytical Chemistry, Vol. 76, No. 4, February 15, 2004

Table 1. Analytical Emission Lines Selected wavelength (nm) B(I)a Mg(I)a Si(I) Cu(I) Zn(I) Zr(II) Ni(I) P(I)b S(I)b Ti(I) Fe(I)

249.67 285.21 288.15 324.75 334.50 339.19 341.47 178.28 180.73 365.35 371.99

wavelength (nm) Ca(II) Al(I)a Ar(I) Sr(II) Nb(I) Cr(I) Cd(I)b Mn(II)b Na(I) K(I)a

393.36 396.15 404.44 407.77 416.46 425.43 228.80 257.60 589.59 766.49

a Measured with the monochromator. b Measured on its second order.

and considering the crater diameter and the material density15). The mean of three sputtered replicates was used to calculate each penetration depth. Glasses used as calibration standards and samples can be divided into three groups. A2, B1, C1, E1, and F1 are special silicate glasses for XRF monitoring from Breitla¨nder (Hamm, Germany) (the set also contained a D1 sample, but it was not used in our experiments because it tended to crack under the GD operation conditions). S-620 corresponds to the standard reference material 620 from NIST (Gaithersburg, MD). Finally, HGa and HGb have the same chemical composition (HG) but different thicknesses. They were from Saint-Gobain Cristaleria, S.A. (Avile´s, Spain). The silica content in the HG samples was determined by gravimetric analysis; sodium by atomic absorption spectrometry; and calcium, magnesium, aluminum, and potassium by X-ray fluorescence. In addition, a quartz sample 0.75 mm thick was investigated. Chemical composition and thickness of standards used for calibration are collected in Table 2. For analysis with rf-GD-OES, the materials were polished to a mirror finish using metallographic grinding paper (120, 600, and 1200 grit) and then cleaned with ethanol. Samples were cooled at 0 °C by a cold liquid circulating between the sample and the rf power input. High-purity argon (99.999% minimum purity) from Air Liquide (Oviedo, Spain) was employed as discharge gas. RESULTS AND DISCUSSION Sputtering Rates. Figure 1 shows the sputtering rates measured for the six calibration standards under the selected operation conditions. As can be seen, there are important differences in the rates between the six specimens, all of them 3 mm thick, with the exception of HGa (2.8 mm thick). The observed standard sample with the lowest sputtering rate (C1) corresponds to the sample with the highest B2O3, Al2O3, and P2O5 concentrations, while the glass with the highest sputtering rate was E1 (standard with the highest Na2O concentration). The sputtering rate for a quartz sample 0.75 mm thick was 0.194 ( 0.03 µg‚s-1. Elemental Matrix Effects in the Plasma. The strong influence of elements with low ionization potentials on analyte emissions (the so-called “easily ionized elements”, EIEs) has been traditionally investigated in plasma sources, such as the inductively coupled plasma (ICP) coupled to OES.22 In this direction, a

Table 2. Chemical Composition and Thickness of Calibration Standards calibration standards (%) A2 B1 C1 E1 F1 HGa thickness (3 mm) (3 mm) (3 mm) (3 mm) (3 mm) (2.8 mm) SiO2 B2O3 Na2O CaO MgO Al2O3 P2O5 K2O MnO FeO Fe2O3 CoO NiO CuO ZnO As2O3 BaO F TiO2 Cr2O3 SrO ZrO2 Nb2O5 CdO Sb2O3 Nd2O3 UO3 La2O3

36.8 3.50 0.30 0.70 3.50 11.9 0.60 2.50 31.9 0.40

51.0

10.0 22.5 9.0

50.9 6.0 15.0 5.0

35.5 15.0 1.0 0.5

1.5

25.0

3.0 19.0

6.60 0.5 1.0 0.50

19.3

5.0

1.0 0.5

0.50

2.5 5.0

58.2 3.0 1.0 3.0 1.0 2.0

70.9 13.9 9.38 4.53 0.58 0.28 0.092

1.0 0.5 1.0 2.0 0.5 3.0 0.8 4.0 0.5

0.5

Figure 2. Mg(II) 280.27-nm/Mg(I) 285.21-nm line intensity ratio for the samples containing magnesium. Error bars correspond to results from three replicates. The Mg(II) 280.27-nm line was measured with the monochromator.

0.2 0.3 4.0 1.0

0.015

1.0 1.0 0.5

0.80 1.0 0.6 1.0 0.5

Figure 1. Sputtering rates obtained for the calibration standards at 450 Pa and 20 W of forward power.

parameter that has been frequently used to check the plasma excitation robustness in ICP-OES is to measure ionic to atomic emission intensity ratios for a selected element present in the sample. The Mg(II) 280.27-nm/Mg(I) 285.21-nm line intensity ratio has been widely recommended for such purposes22,23 in ICP-OES. The concentrations of sodium, potassium, calcium, and magnesium in our studied glasses varied by large intervals (0-15% Na2O, 0-19.3% K2O, 0-25% CaO, and 0-4.5% MgO), depending upon the standard material. Therefore, it was considered of interest to investigate the values of Mg(II) 280.27-nm/Mg(I) 285.21-nm line intensity ratios in such calibration standards and samples containing magnesium. Results observed are collected in Figure 2, and they show clearly that the Mg(II)/Mg(I) ratios were kept constant for all the glass specimens assayed (indicating a quite robust low-pressure plasma). (22) Todolı´, J. L.; Gras, L.; Hernandis, V.; Mora, J. J. Anal. At. Spectrom. 2002, 17, 142-169. (23) Mermet, J. M. Anal. Chim. Acta 1991, 250, 85-94.

Figure 3. Effect of the matrix and the thickness of the samples on the Ar(I) 404.44-nm emission line: (a) signal intensity versus forward power, (b) slope of the plots from Figure 3a versus the inverse of sample thickness.

Emission Yields Correction. In our search for a simple quantification scheme, the effect of forward power (P) on the Ar line emission intensity (IAr) was investigated for the sample S-620 nm, for the sample F1 (in this case, 5.35 mm thick), and at four different thickness of the HG sample. As expected, a linear relationship was obtained for all the samples investigated (see Figure 3a).

IAr,M,th ) aM,th + mM,thP

(1)

IAr,M,th is the argon emission intensity in a matrix M of thickness th and m, the slope for such sample. Obviously, for a fixed forward power, the thicker the sample, the lower the argon signals.12 Additionally, as can be seen in Figure 3b, a linear graph was observed when plotting the slopes obtained in Figure 3a versus Analytical Chemistry, Vol. 76, No. 4, February 15, 2004

1041

Figure 4. Effect of the matrix and the thickness of the samples on the Si(I) 288.15-nm emission line: (a) signal intensity versus forward power, (b) slope of the plots from Figure 4a versus the inverse of the sample thickness.

the inverse of the corresponding sample thickness (for subsequent work, the slope of the latter plot will be referred to as “slopeAr”). Therefore, in our studies, m can be considered as matrixindependent.

mth ) b + slopeAr(1/th)

(3)

Interestingly, similar equations were obtained for the analytes, provided that the working interval for the forward power was from 10 to 35 W. Illustrative results for silicon and calcium are shown in Figures 4 and 5 (as in the case of argon, the slopes of the plots presented in Figures 4b and 5b will be called “slopeSi” and “slopeCa” respectively, and in general, “slopean”). Thus, to correct the measured intensities in a given sample for its thickness (th′) different from the selected standard (th) for the calibration, equations similar to 3, but for each analyte instead of Ar, can be applied,

Ian,th′ ) a′th + bP′ + slopean(1/th′)P′

(4)

Icorr an,th ) ath + bP + slopean(1/th)P

(5)

where Ian,th′ refers to the measured emission intensity for a given analyte in a given sample of thickness th′ measured at a forward power, P′; Icorr an(th) should be the “corrected” emission intensity at the reference forward power, P, and standard thickness, th. 1042

Analytical Chemistry, Vol. 76, No. 4, February 15, 2004

Thus, the corrected emission intensity for a given analyte can be obtained by dividing eqs 4 and 5 and rearranging.

Icorr an (th,P) )

bP + slopean(1/th)P

I (th′,P′) bP′ + slopean(1/th′)P′ an bP + slopean(1/th)P

a′(th′) + a(th) (6)

bP′ + slopean(1/th′)P′

(2)

From eqs 1 and 2, the following expression can be obtained.

IAr,M,th ) aM,th + bP + slopeAr(1/th)P

Figure 5. Effect of the matrix and the thickness of the samples on the Ca(II) 393.36-nm emission line: (a) signal intensity versus forward power, (b) slope of the plots from Figure 5a versus the inverse of the sample thickness.

For a fixed forward power, eq 6 can be simplified to

Icorr an (th,P) )

b + slopean(1/th)

I (th′,P) b + slopean(1/th′) an b + slopean(1/th)

a′ + a(th) (7) b + slopean(1/th′) (th′)

To obtain the emission yields for the different analytes at the selected operating conditions (20 W forward power and 450 Pa), the well-known equation in GD-OES for electrically conductive samples that relates the measured emission intensity with the analyte concentration15,19 was used,

IM,E,λ ) Rλ,EqMCM,E

(8)

where IM,E,λ is the intensity of the line λ of the element E in the matrix M, qM is the sputtering rate of the matrix M, Rλ,E is the emission yield of the line λ of the element E, and CM,E is the concentration of the element E in the matrix. According to eq 8, the emission yield Rλ,E can be calculated as the slope of the plot of IM,E,λ versus the product qMCM,E (sputtered mass of the element E).

Figure 6. Calibration curves obtained with rf-GD-OES at the selected operating conditions (450 Pa and 20 W of forward power). For corrections made in the y axis, see the text. (a) Si(I) 288.15 nm, (b) Ca(II) 396.36 nm, (c) Na(I) 589.59 nm.

It is commonly agreed that the emission yield for a given analytical emission line is less matrix-dependent in conductor than in insulator analysis. Actually, emission yields in insulators could be greatly affected by physical parameters from the sample itself, such as the sample thickness. To overcome the thickness effect, some corrections for the measured IM,E,λ should be applied. In our experiments, the emission intensity for a given analytical line was corrected by either eq 6 or eq 7; 3 mm was considered to be the standard thickness for the calibration. Figure 6 collects the calibration curves at 20 W of forward power for silicon, calcium, and sodium, where the corrected IM,E,λ obtained from eq 7 was used for the calibration standard of 2.8 mm (see Table 2). In addition, better fittings were observed when the argon signal was used as the reference signal. Figure 7 shows the calibration curves for silicon, calcium, and sodium when the corrected IM,E,λ was divided by the corrected argon signal. As can be seen, the data fit rather well to straight lines (the slope of such straight lines will be referred to as the “corrected emission yields”). Quartz was discarded as the calibration standard because the corr experimentally obtained Icorr Si /IAr value for such material was ∼39% lower than the value expected from extrapolating the calibration curve of Figure 7a.

Figure 7. Calibration curves obtained with rf-GD-OES at the selected operating conditions (450 Pa and 20 W of forward power) using argon as reference signal. (a) Si(I) 288.15 nm, (b) Ca(II) 396.36 nm, (c) Na(I) 589.59 nm.

Quantification of Glass Samples. The corrected emission yield concept was applied to the quantification of the major components (SiO2, Na2O, CaO, MgO, Al2O3, and K2O) in two glass samples of different thickness measured at a forward power of 20 W. The sputtering rates of the samples were mathematically calculated using the emission yields obtained from the calibration curves and by measuring the emission intensities of all the sample components.15,18 In the case of sample S-620, which was 3 mm thickness, it was not necessary to make corrections for the measured analyte emission intensities. However, adequate corrections using eq 7 were applied to the sample HGb (1.1 mm). Results are collected in Table 3 for both samples and, as can be seen, they are in good agreement with concentrations given in Table 2. In addition, to test the validity of eq 6, sample HGb was also measured at 15 W (forward power), and a forward power of 25 W was used for sample S-620. Results are also shown in Table 3. Again, rather good agreement with the expected concentrations values can be observed. CONCLUSIONS This study shows that the use of the constant emission yield concept (for constant pressure and constant forward power Analytical Chemistry, Vol. 76, No. 4, February 15, 2004

1043

Table 3. Quantitative Analysis of Two Different Samplesa Rf-GD-OESb accepted compositionb

SiO2 Na2O CaO MgO Al2O3 K2O a

20 W

15 W

25 W

HGb

S-620

HGb

S-620

HGb

S-620

70.9 ( 0.2 13.9 ( 0.1 9.38 ( 0.05 4.53 ( 0.04 0.58 ( 0.02 0.28 ( 0.01

72.08 ( 0.08 14.39 ( 0.06 7.11 ( 0.05 3.69 ( 0.05 1.80 ( 0.03 0.41 ( 0.03

68.1 ( 1.3 15.5 ( 0.1 10.5 ( 0.3 4.5 ( 0.3 1.04 ( 0.05 0.23 ( 0.02

73.0 ( 1.0 11.7 ( 0.9 6.2 ( 0.4 4.8 ( 0.2 3.2 ( 0.2 0.45 ( 0.05

68.1 ( 1.2 15.7 ( 0.2 10.3 ( 0.2 4.6 ( 0.2 1.05 ( 0.06 0.25 ( 0.03

72.1 ( 1.1 11.4 ( 0.9 8.5 ( 0.5 5.1 ( 0.2 2.3 ( 0.3 0.27 ( 0.06

Concentrations are expressed as percentages. b The sample HGb was 1.1 mm thick, and the sample S-620 was 3 mm thick.

conditions) could be a good approach for the quantification of bulk glass samples, provided that some corrections for the thickness of the specimens along with argon normalization is carried out. Although the results collected in Table 3 are not as accurate as typically necessary for quantitative analysis, two considerations should be taken into account: first, the special silicate glasses for XRF monitoring provided by Breitla¨nder were not certified (as informed by the manufacturer, the analytical values stated should be considered as guiding analytical data). In addition, the quantified six elements do not constitute the 100% content of the samples, and therefore, some errors should be expected. In any case, the results shown here should be considered to be preliminary, but they show that the quantification of samples conventionally known as insulators with commercial rf-GD-OES equipment can be possible, and we hope this work pushes the GD analytical community to go further in this direction, both for bulk analysis and for coatings containing multicomponent nonconducting layers. On the other hand, eq 6 could be of great interest to extend the proposed approach to particular specimen samples when

1044 Analytical Chemistry, Vol. 76, No. 4, February 15, 2004

forward powers different from those selected for the calibration should be applied (for example, to apply lower forward power in order to avoid the cracking of a particular sample or higher forward power in order to achieve a proper sputtering). ACKNOWLEDGMENT The authors thank Saint-Gobain Cristalerı´a S.A. (Avile´s, Spain) for financial and technical support. In addition, financial support from Plan I + D + I Principado de Asturias 2001/2004 through project FC-02-PC-CIS01-11 is gratefully acknowledged. Finally, Beatriz Ferna´ndez acknowledges a grant from the Ministerio de Educacio´n, Ciencia y Deporte within the program Becas de postgrado para la formacio´n de profesorado universitario (ref. AP2002-2689).

Received for review September 23, 2003. Accepted November 27, 2003. AC035113Q