- Email: [email protected]
Sequential determination of platinum, ruthenium, and molybdenum in carbon-supported Pt, PtRu, and PtMo catalysts by atomic absorption spectrometry
Talanta 63 (2004) 791–796
Short communication
Sequential determination of platinum, ruthenium, and molybdenum in carbon-supported Pt, PtRu, and PtMo catalysts by atomic absorption spectrometry Silvera Scaccia∗ , Barbara Goszczynska IDROCOMB, Hydrogen Project and Fuel Cells, ENEA, C.R. Casaccia, via Anguillarese 301, I-00060 Rome, Italy Received 8 August 2003; received in revised form 10 November 2003; accepted 11 December 2003 Available online 3 February 2004
Abstract A simple accurate and precise analytical method for the determination of platinum, ruthenium, and molybdenum in Pt, PtRu, and PtMo nanoparticles catalysts deposited on high-surface area carbon by flame atomic absorption spectrometry (FAAS) and graphite furnace atomic absorption spectrometry (GFAAS) is described. The complete digestion of samples (0.010–0.020 g), which contain noble metals (NMs) in the range between 0 and 30% in combination among them or with other non-NMs, is obtained under mild conditions using both concentrated HCl and HCl + HNO3 (1 + 1 (v/v)) mixture to boiling for 30 min in an open vessel. Carbon is separated from the solution by filtering it. Under optimized conditions of the flame, the poor sensitivity of platinum is enhanced 50-fold in presence of 1% (m V−1 ) ascorbic acid, whereas the analytical signal of ruthenium increased by the presence of co-existing platinum. Any kind of interference is observed on the analytical signal of molybdenum. Recovery test obtained by analyzing commercial powder catalysts ranged from 99 to 101%. The precision, expressed as relative standard deviation of five measurements, is better than 1%. Electrode catalysts, made by using the carbon-supported platinum-based powder catalysts, have been analyzed for the metal loadings onto the electrode by GFAAS after dissolution under the same conditions used for the powder catalysts. The precision, expressed as relative standard deviation of three measurements, is better than 2%. © 2004 Elsevier B.V. All rights reserved. Keywords: Platinum; Ruthenium; Molybdenum; Catalyst; Flame atomic absorption spectrometry; Graphite furnace atomic absorption spectrometry
1. Introduction Platinum is the state of the art anode electrode of the proton exchange membrane fuel cells (PEMFCs) for use in zero-emission electric vehicles application due to its effective electrocatalytic properties. Unfortunately, Pt catalyst is poisoned by the contaminant CO when hydrogen fuel produced from reformed hydrocarbons is used with consequent degradation of the cell. It has been recently recognized that elements such as Ru, Mo, Ge, Sn, Os, and W exhibit co-catalytic activity and inhibit the deactivation of platinum if present as monolayers absorbed over the Pt surface. Thus, efforts in the development of binary and ternary alloy active catalysts are made. The catalysts are usually obtained by chemical reduction of the appropriate noble metal (NM) salts and finely dispersed on high-surface area graphite powder. ∗
Corresponding author. Tel.: +39-6-30483815; fax: +39-6-30486357. E-mail address: [email protected] (S. Scaccia).
0039-9140/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2003.12.014
Carbon-supported platinum catalysts thus obtained are deposited on carbon paper disk electrode. From an economical and environmental point of view, it is desirable to load the electrode disk with an amount of NMs as low as possible (below 0.5 mg cm−2 ) with good electrochemical performances. Therefore, analytical methods to control precisely and accurately the metal content in carbon-supported platinum-based catalysts are needed. The characterization of platinum-based catalyst supported on carbon is often carried out by energy dispersive X-ray spectroscopy (EDX) because of relative simplicity of the sample preparation and rapidity of the measurements [1,2]. However, this technique is useful mainly for quality control because of the low detection limits, low precision, and overall because a small area of the specimen may be analyzed at any time, thus that inhomogeneities at small scale strongly will influence the bulk measurements. Atomic spectroscopy techniques are well-established analytical tools for the determination of NMs owing to the
792
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
high sensitivity and selectivity, accuracy and precision, and low cost. Among these, flame atomic absorption spectrometry (FAAS) is generally employed for NMs determinations in concentrates due to the poor sensitivity of the technique for these elements, but has the advantage of being less subjected to chemical and spectral interferences. Graphite furnace atomic absorption spectrometry (GFAAS) is more sensitive than FAAS, but it undergoes more severe interferences and lower reproducibility [3]. In the atomic absorption determination of NMs, the solid sample have to undergo wet chemical attack with strong acids like aqua regia, HCl, HNO3 , HBr, Br2 , H2 O2 , etc., in a hot plate or using high-pressure digestion systems owing to the high ionization potential and high boiling points of NMs. Exception is made for Ru, which forms a volatile (108 ◦ C) ruthenium tetraoxide during the high-temperature oxidizing dissolution process [4]. The correct result of the analysis depends on the completeness of transferring of analyte from solid sample onto solution, avoiding analyte losses due to sampling evaporation to dryness. Therefore, a sample dissolution procedure has to be well planned prior to AAS determination. Recently, reviews on the dissolution procedure for the determination of NMs in geologist and environmental samples have been reported [5,6]. More recently, Balcerzak [7] has written a review on the sample digestion methods for the determination of traces of precious metals prior to the spectrometric analysis. Various laborious procedures for the digestion of carbon-supported platinum-based catalysts, which make use of mixture of both HNO3 and HCl (1 + 6 (v/v)) and aqua regia [8–10], have been reported. The use of aqua regia for the mineralization of platinum metal by complexation with HCl on the self-protecting passivation layer of platinum obtained by the oxidant acid action is mandatory. It has been reported that hot HCl is able to dissolve some metallic platinum catalyst supported on MgO, previously calcined in reducing conditions, because nanoparticles of Pt are finely dispersed onto the support so that the atmospheric O2 is sufficient to guarantee a complete dissolution [11]. However, poor recoveries have been obtained with fresh/calcined and spent catalyst samples. In the present work, a simple accurate and precise method for the determination of platinum, ruthenium, and molybdenum in Pt, PtRu, and PtMo nanoparticles catalysts finely dispersed on high surface-activated carbon by flame and GFAAS is reported. The digestion of noble and non-NMs is achieved by using both concentrated HCl and a mixture of HCl+HNO3 (1+1 (v/v)) to boiling for 30 min. The dissolution procedure is fast, economic, and environment friendly because repeated evaporation of the acid to eliminate nitrates nitrosyl compounds is not required. Evaluation of the chemical interferences are made for both techniques. The proposed method may be also applicable for the certification of carbon-supported platinum catalyst used as reference materials for EDX analysis.
2. Experimental 2.1. Instrumentation A Varian SpectrAA 220FS (Victoria, Australia) flame atomic absorption spectrometer equipped with a GTA110 graphite furnace, an auto sampler and a deuterium lamp as background correction system was used. Hollow cathode lamps (Varian) of Pt, Ru, and Mo were used as sources. A pneumatic nebulizer with a glass impact bead was used. The instrumental parameters of the spectrometer were as follows: wavelengths and width bands at 265.9 and 0.2 nm for Pt; 349.9 and 0.2 nm for Ru; and 313.3 and 0.5 nm for Mo, respectively. Integration time was 10 s. Observation height varied between 6 and 12 mm. A slot burner for air–acetylene flame was used. Acetylene flow-rate was varied between 1.5 and 4.0 l min−1 and air flow was kept constant at 13.5 l min−1 . Pyrolytic graphite-coated graphite tubes were used. Argon (99.999%) gas was fluxed in the graphite tube. Peak height absorbance signals were measured by the AA instrument. The volume of the solution introduced into the graphite tube was 20 l. The samples were weighted with a micro balance (AND HM-202) with 0.001 g accuracy. 2.2. Reagents and standard solutions All acids were of analytical-reagent grade. Ultra high-purity water with a specific resistance of 18 M cm obtained from a Milli-Q water purification system (Millipore, Bedford, MA, USA) and filtered through a 0.22 m membrane filter was used throughout. Platinum and ruthenium stock solutions of 980 and 1005 mg l−1 in 5 wt.% HCl, respectively, were p.a. grade from Aldrich. Molybdenum stock solution of 1000 mg l−1 was obtained by dissolution of an appropriate amount of pure molybdenum lamina (p.a. Aldrich) in 3 ml of a hot mixture of HCl + HNO3 (1 + 1 (v/v)) for 30 min. Working standard 5–40 mg l−1 Pt, 4–10 mg l−1 Ru, and 10–50 mg l−1 Mo solutions were prepared daily by one step dilution of the standard stock solutions with 1 M HCl for FAAS measurements. Working standard 0.1–0.8 mg l−1 Pt, 0.08–0.32 mg l−1 Ru, and 0.0125–0.0375 mg l−1 Mo solutions were prepared daily by stepwise dilutions of the standard stock solutions with 1 M HCl for GFAAS measurements. To evaluate the interferences from the co-existing element matrix-matched standard solutions were prepared by adding the concomitant Pt, Ru, and Mo at the same concentration as present in the sample solution. Standard Pt solutions containing 1% (m V−1 ) ascorbic acid (Merck, Germany) were also prepared. 2.3. Samples 2.3.1. Powder catalysts Commercially available powder catalysts Pt/C, PtRu/C, and PtMo/C from E-Tek (Inc. Natick, MA, USA) were
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
793
chosen in a wide range of bulk composition to encompass a large variation of analyte concentrations and validate the method. In addition, three Pt/C and PtRu/C powder catalyst samples were prepared by chemical reduction with Na2 S2 O4 of H2 PtCl6 and H2 RuCl5 ·H2 O solutions in presence of Vulcan XC-72R carbon (Cabot) to metallic Pt and Ru. The obtained powders were annealed at 600 ◦ C in Ar for 10 h. The nominal compositions were 10 wt.% Pt/C, 10 wt.% PtRu/C (atomic ratio Pt:Ru 1:1), and 15 wt.% PtRu/C (atomic ratio Pt:Ru 1:1). 2.3.2. Electrode catalysts Commercial 20 wt.% Pt/C, 15 wt.% PtRu/C (atomic ratio Pt:Ru 1:1) and 30 wt.% PtMo/C (atomic ratio Pt:Mo 4:1) powder catalysts were used for the preparation of three electrode catalysts. In brief, powder catalyst (5 mg) was mixed with 1 ml aqueous solution of Nafion and glycerol to form a slurry. One hundred microlitres of the slurry was spread on a disk (weight 0.010 g and surface area 1 cm2 ) of carbon paper by a spray technique and dried at 70 ◦ C for 1 h. The metal loading was Pt(0.1), Pt(0.05) + Ru(0.025) and Pt(0.133) + Mo(0.016) mg cm−2 , respectively. Prior to dissolution procedures, the samples were dried at 200 ◦ C for 2 h in an electric furnace to remove moisture and stored in a desiccator prior to analysis. 2.4. Powder and electrode catalysts dissolution procedure Accurately weighted sample (0.010–0.020 g) was digested gently in a 10 ml covered Pyrex beaker with 3 ml of: (i) aqua regia (HCl +HNO3 , 3+1 (v/v)) for Pt/C, PtRu/C, and PtMo/C; (ii) concentrated HCl for Pt/C and PtRu/C; (iii) a mixture of HCl + HNO3 (1 + 1 (v/v)) for PtMo/C on an electric hot plate for 30 min. After cooling, the solution was filtered through a teflon filter (45 mm diameter) with a pore size of 0.8 m in a millipore filtration apparatus under vacuum to separate it from the carbon. The filter and residue were washed with 3 ml of 1 M HCl. The solution along with the washings were added to 25 ml volumetric flask and made up to volume with 1 M HCl. The final solution was analyzed for the Pt, Ru, and Mo after appropriated dilution with 1 M HCl. 2.5. Optimization of flame composition The optimized flame parameters were obtained by introducing 25 mg l−1 Pt, Ru, and Mo acid-containing standard solutions into the nebulizer for different flame composition at a given observation height over the burner head, being the others parameter kept constant. The effect of acetylene flow-rate was more significant as compared to those of observation height (Fig. 1). The maximum absorbance
Fig. 1. The effect of acetylene flow-rate on absorption signals of 25 mg l−1 Pt, 10 mg l−1 Ru, and 25 mg l−1 Mo at different observation heights: 4 mm (circle); 6 mm (square); 8 mm (triangle up); 10 mm (triangle down).
signal for platinum was achieved at the same observation height (between 4 and 6 mm) regardless the flame composition. The optimized flame composition was as follows: a stoichiometric acetylene–air flame for platinum; a rich acetylene–air flame for Ru and Mo. 2.6. Optimization of graphite furnace The optimized drying and pyrolysis temperature and holding time were chosen for all the analytes. The temperature programs used for the graphite tube are described in Table 1.
794
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
Table 1 Heating programs of the graphite furnace Step
Temperature (◦ C)
Hold time (s)
Gas flow-rate (l min−1 )
Dry 1 2 3
85 95 120
5.0 40.0 10.0
3.0 3.0 3.0
Pyrolysis 1 1000 2 1000 3 1000
5.0 1.0 2.0
3.0 3.0 0.0
Atomization 1 2700 (Pt); 2600 (Ru, Mo) 2 2700 (Pt); 2600 (Ru, Mo) 3 2700 (Pt); 2600 (Ru, Mo)
1.3 2.0 2.0
0.0 0.0 3.0
3. Results and discussion 3.1. Determination of platinum, ruthenium, and molybdenum in Pt/C, PtRu/C, and PtMo/C powder catalysts by FAAS 3.1.1. Interferences studies The presence of high concentration of inorganic acids into the sample solution does not provoke any chemical interferences on the analytical signal of all the analytes under the recommended operating conditions of the flame. The poor sensitivity of platinum in the air–acetylene flame is enhanced twice time in presence of 1% (m V−1 ) ascorbic acid. The effect of the reducing agent on the atomization efficiency may be related to the formation of volatile platinum compounds in acid media and under reducing conditions as reported recently in Ref. [12]. No interference effects of co-existing Ru and Mo on the absorbance signal of platinum have been evidenced. A beneficial effect of the co-existing platinum on the Ru analytical signal has been observed. The influence on the ruthenium signal has been examined over the concentration up to 1000 mg l−1 platinum. A positive interference of about 50% is observed likely due to the formation of inter-metallic compounds with platinum. Any interferences on the analytical signal of molybdenum from the co-existing Pt element has been observed.
3.1.2. Sample digestion of powder catalysts with different acids Different sample digestion procedures for carbonsupported platinum-based catalysts have been evaluated. Besides digestion with aqua regia it has been also used concentrated HCl for acid decomposition of Pt/C and PtRu/C in order to reduce losses of Ru by RuO4 formation during digestion with oxidant acid. Recovery test has been carried out on commercial carbon-supported catalysts in order to evaluate the efficiency of the proposed digestion procedures. In Table 2, the results of the analysis of Pt/C and PtRu/C catalysts digested with both aqua regia and concentrated HCl are reported. The recovery is within 99–100%, being the FAAS gathered results in good agreement with the certificated values for all the acidic mixture used. A mixture of HCl + HNO3 (1 + 1 (v/v)) has been tested for the digestion of carbon-supported Pt and PtMo/C catalysts. In Table 2, the results of the analysis of Pt/C and PtMo/C catalysts digested in different acids are also reported. 3.1.3. Analytical figures of merit Calibration curves have been constructed against five acid-containing standard solutions and the figures of merit are listed in Table 3. For ruthenium determination calibration with standard additions has been also conducted to take in account the positive interference of co-existing platinum. The curves are linear with the correlation coefficients r2 better than 0.999 and the intercepts significantly do not deviate from zero at 95% confidence level. In order to obtain a higher precision of the measurements, the sample solutions have been diluted to a level near the centroide of the calibration graphs [13] (see Table 3). The precision, expressed as relative standard deviation of five readings, is better than 1% (see Table 2). 3.1.4. Powder catalyst samples analysis Three carbon-supported platinum-based powder catalysts, prepared in our laboratories, have been analyzed according to the proposed method. In Table 4, the analytical results for platinum and ruthenium in Pt/C and PtRu/C powder catalysts are listed. As it can be seen, a good agreement of the results with the nominal values has been found.
Table 2 FAAS results (wt.%)a of Pt, Ru, and Mo in commercial Pt/C, PtRu/C, PtMo/C powder catalysts after digestion in different acids Catalyst
Pt
Ru HCl + HNO3 (1 + 1 (v/v))
HCl 10 wt.% 20 wt.% 30 wt.% 15 wt.% 20 wt.% 30 wt.% a
Pt/C Pt/C Pt/C PtRu/C (atomic ratio Pt:Ru 1:1) PtRu/C (atomic ratio Pt:Ru 1:3) PtMo/C (atomic ratio Pt:Mo 4:1)
Mean ± S.D. of five measurements.
9.95 20.5 29.9 9.90 7.80
± ± ± ± ±
0.2 0.2 0.3 0.1 0.07 26.7 ± 0.2
Aqua regia 90.2 20.0 ± 0.2 29.8 ± 0.6 9.7 ± 0.1 7.75 ± 0.06 26.5 ± 0.2
Mo
HCl
Aqua regia
5.15 ± 0.05 11.7.0 ± 0.1
4.90 ± 0.05 11.0 ± 0.1
HCl + HNO3 (1 + 1 (v/v))
Aqua regia
3.25 ± 0.03
3.20 ± 0.03
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
795
Table 3 Characteristic parameters of the analytical calibration curves obtained for the determination of platinum, ruthenium, and molybdenum by FAAS
Correlation coefficient (n = 5) Equation of the regression lines (Abs vs. concentration) (mg l−1 ) Centroid of the regression line (x, y) sb (standard deviation of the slope) sa (standard deviation of the intercept) Working linear range (mg l−1 ) Dilution of sample solution (v/v) r2
Pt
Ru
Mo
0.9994 y = 0.001227x + 7 × 10−5
0.99998 y = 0.002830x + 1 × 10−6
0.99994 y = 0.00248x + 1 × 10−4
14, 0.017 5 × 10−6 9 × 10−5 4–25 1+9
5, 0.015 3 × 10−6 1 × 10−6 4–10 1+4
30, 0.072 1 × 10−5 4 × 10−4 10–50 1+4
Table 4 FAAS results (wt.%)a of Pt and Ru in Pt/C and PtRu/C powder catalysts after digestion in concentrated HCl Catalyst
Pt
10 wt.% Pt/C 10 wt.% PtRu/C (atomic ratio Pt:Ru 1:1) 15 wt.% PtRu/C (atomic ratio Pt:Ru 1:1) a
Ru
HCl
Aqua regia
HCl
Aqua regia
9.95 ± 0.2 7.6 ± 0.08 10.2 ± 0.1
9.90 ± 0.2 7.7 ± 0.07 10.1 ± 0.1
3.3 ± 0.05 4.6 ± 0.05
3.2 ± 0.04 4.5 ± 0.04
Mean ± S.D. of five measurements.
Table 5 Characteristic parameters of the analytical calibration curves obtained for the determination of platinum, ruthenium, and molybdenum by GFAAS
Correlation coefficient r2 (n = 3) Equation of the regression lines (Abs vs. concentration) (mg l−1 ) Centroid of the regression line (x, y) sb (standard deviation of the slope) sa (standard deviation of the intercept) Working linear range (mg l−1 ) Dilution of sample solution (v/v)
Pt
Ru
Mo
0.99998 y = 0.613x + 1 × 10−3 0.4, 0.2 1 × 10−3 3 × 10−3 0.1–0.8 1+4
0.99992 y = 1.01x + 1 × 10−4 0.16, 0.16 6 × 10−3 9 × 10−4 0.08–0.32 1+4
0.9999 y = 7.0x + 2 × 10−4 0.02, 0.14 3 × 10−2 1 × 10−4 0.0125–0.0375 1 + 24
3.2. Determination of platinum, ruthenium, and molybdenum in Pt/C, PtRu/C, and PtMo/C electrode catalysts by GFAAS 3.2.1. Interference studies The determination of 4 ng Pt, 4 ng Ru, and 0.4 ng Mo with the co-existing elements have been made using the graphite furnace program given in Table 1. Any interference effects of the elements with each other during their determination have been evidenced. The pyrolysis/atomization curves are similar for platinum and ruthenium and any loss of analyte due to the acid-interference have been seen. A decrease of sensitivity of the molybdenum signal in the graphite tube has been observed with the increasing of the number of firings. This
is due to the formation of the molybdenum carbide during atomization [14]. It has been reported the beneficial effect of platinum on the sensitivity and reproducibility of the analytical signal of molybdenum in the graphite furnace. In fact, in presence of NMs the formation of inter-metallic compound of PtMo likely occurs, which dissociating rapidly during atomization avoids the formation of molybdenum carbide, thus improving the reproducibility of measurements [15]. 3.2.2. Analytical figures of merit Calibration curves have been constructed against three acid-containing aqueous standard solutions and the figures of merit are listed in Table 5. The curves are linear with the correlation coefficients r2 better than 0.9999 and the intercepts
Table 6 GFAAS results (mg)a of Pt, Ru, and Mo in commercial Pt/C, PtRu/C, and PtMo/C deposited onto carbon paper electrodes after digestion in different acids Metal loading (mg)
Pt HCl
Pt(0.1) Pt(0.05) + Ru(0.025) Pt(0.133) + Mo(0.016) a
0.100 ± 0.004 0.049 ± 0.001
Mean ± S.D. of three measurements.
HCl + HNO3 (1 + 1 (v/v))
0.130 ± 0.003
Ru
Mo
HCl
HCl + HNO3 (1 + 1 (v/v))
0.0240 ± 0.0005
0.0150 ± 0.0003
796
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
do not significantly deviate from zero at 95% confidence level for all the analytes. For the determination of Mo, 50 ng Pt are co-injected into the graphite tube as a chemical modifier resulting in a good precision and better linearity of the calibration curve. The use of the chemical modifier results also in an enhancement of the tube lifetime, lowering the analysis cost. 3.2.3. Electrode catalyst samples analysis The analysis of the electrode catalysts made by using commercial powder catalysts with a nominal metal loadings of 0.1 mg cm−2 Pt, 0.05 + 0.025 mg cm−2 Pt + Ru, and 0.133 + 0.015 mg cm−2 Pt + Mo by GFAAS were in a good agreement with the found values, as shown in Table 6.
4. Discussion and conclusions The determination of NMs (Pt and Ru) and non-NMs (Mo) in carbon-supported platinum-based powder catalysts after digestion in different acids has been evaluated by analyzing the sample solutions by FAAS. It has been found that the use of both concentrated HCl and HCl + HNO3 (1 + 1 (v/v)) mixture for decomposition of Pt, Ru, and Mo nanoparticles finely deposited on carbon allows to dissolve accurately the noble and non-NMs, under mild conditions compared to the usual digestion method with aqua regia. Conversely, the determination of the Pt, Ru, and Mo loaded
onto carbon paper electrode disks, made by using powder catalysts, is carried out after dissolution of samples in different acids by GFAAS due to the higher sensitivity compared to the FAAS. In fact, metal loadings on electrode catalysts fall under the detection limit of quantification of FAAS for all the analytes studied. References [1] E.M. Crabb, M.K. Ravikumar, Electrochim. Acta 46 (2001) 1033. [2] N. Fujiwara, K. Yasuda, T. Ioroi, Z. Siroma, Y. Miyazaki, Electrochim. Acta 47 (2002) 4079. [3] C.R.M. Rao, G.S. Reddi, Trends Anal. Chem. 19 (2000) 565. [4] M. Hoashi, R.R. Brooks, R.D. Reeves, Anal. Chim. Acta 232 (1990) 317. [5] J.G. Sen Gupta, Talanta 40 (1993) 791. [6] R.R. Barefoot, J.C. Van Loon, Talanta 49 (1999) 1. [7] M. Balcerzak, Anal. Sci. 18 (2002) 737. [8] M. Merdivan, R.S. Aygun, N. Kulcu, At. Spectrosc. 18 (1997) 122. ´ ecicka, D. Bystro´nska, Anal. Lett. 32 (1999) [9] M. Balcerzak, E. Swi˛ 1799. ´ ecicka, E. Balukiewicz, Talanta 48 (1999) 39. [10] M. Balcerzak, E. Swi˛ [11] S. Recchia, D. Monticelli, A. Pozzi, L. Rampazzi, C. Dossi, Fresenius J. Anal. Chem. 369 (2001) 403. [12] P. Pohl, W. Zyrnicki, J. Anal. At. Spectrom. 16 (2001) 1442. [13] J.C. Miller, J.N. Miller, Statistics for Analytical Chemistry, Ellis Horwood, Chichester, 1984, p. 140. [14] B. Welz (Ed.), Atomic Absorption Spectrometry, second revised ed., VCH, Weinheim, 1985, p. 306. [15] E.A. Piperaki, N.S. Thomaidis, I. Demis, J. Anal. At. Spectrom. 14 (1999) 1901.
Short communication
Sequential determination of platinum, ruthenium, and molybdenum in carbon-supported Pt, PtRu, and PtMo catalysts by atomic absorption spectrometry Silvera Scaccia∗ , Barbara Goszczynska IDROCOMB, Hydrogen Project and Fuel Cells, ENEA, C.R. Casaccia, via Anguillarese 301, I-00060 Rome, Italy Received 8 August 2003; received in revised form 10 November 2003; accepted 11 December 2003 Available online 3 February 2004
Abstract A simple accurate and precise analytical method for the determination of platinum, ruthenium, and molybdenum in Pt, PtRu, and PtMo nanoparticles catalysts deposited on high-surface area carbon by flame atomic absorption spectrometry (FAAS) and graphite furnace atomic absorption spectrometry (GFAAS) is described. The complete digestion of samples (0.010–0.020 g), which contain noble metals (NMs) in the range between 0 and 30% in combination among them or with other non-NMs, is obtained under mild conditions using both concentrated HCl and HCl + HNO3 (1 + 1 (v/v)) mixture to boiling for 30 min in an open vessel. Carbon is separated from the solution by filtering it. Under optimized conditions of the flame, the poor sensitivity of platinum is enhanced 50-fold in presence of 1% (m V−1 ) ascorbic acid, whereas the analytical signal of ruthenium increased by the presence of co-existing platinum. Any kind of interference is observed on the analytical signal of molybdenum. Recovery test obtained by analyzing commercial powder catalysts ranged from 99 to 101%. The precision, expressed as relative standard deviation of five measurements, is better than 1%. Electrode catalysts, made by using the carbon-supported platinum-based powder catalysts, have been analyzed for the metal loadings onto the electrode by GFAAS after dissolution under the same conditions used for the powder catalysts. The precision, expressed as relative standard deviation of three measurements, is better than 2%. © 2004 Elsevier B.V. All rights reserved. Keywords: Platinum; Ruthenium; Molybdenum; Catalyst; Flame atomic absorption spectrometry; Graphite furnace atomic absorption spectrometry
1. Introduction Platinum is the state of the art anode electrode of the proton exchange membrane fuel cells (PEMFCs) for use in zero-emission electric vehicles application due to its effective electrocatalytic properties. Unfortunately, Pt catalyst is poisoned by the contaminant CO when hydrogen fuel produced from reformed hydrocarbons is used with consequent degradation of the cell. It has been recently recognized that elements such as Ru, Mo, Ge, Sn, Os, and W exhibit co-catalytic activity and inhibit the deactivation of platinum if present as monolayers absorbed over the Pt surface. Thus, efforts in the development of binary and ternary alloy active catalysts are made. The catalysts are usually obtained by chemical reduction of the appropriate noble metal (NM) salts and finely dispersed on high-surface area graphite powder. ∗
Corresponding author. Tel.: +39-6-30483815; fax: +39-6-30486357. E-mail address: [email protected] (S. Scaccia).
0039-9140/$ – see front matter © 2004 Elsevier B.V. All rights reserved. doi:10.1016/j.talanta.2003.12.014
Carbon-supported platinum catalysts thus obtained are deposited on carbon paper disk electrode. From an economical and environmental point of view, it is desirable to load the electrode disk with an amount of NMs as low as possible (below 0.5 mg cm−2 ) with good electrochemical performances. Therefore, analytical methods to control precisely and accurately the metal content in carbon-supported platinum-based catalysts are needed. The characterization of platinum-based catalyst supported on carbon is often carried out by energy dispersive X-ray spectroscopy (EDX) because of relative simplicity of the sample preparation and rapidity of the measurements [1,2]. However, this technique is useful mainly for quality control because of the low detection limits, low precision, and overall because a small area of the specimen may be analyzed at any time, thus that inhomogeneities at small scale strongly will influence the bulk measurements. Atomic spectroscopy techniques are well-established analytical tools for the determination of NMs owing to the
792
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
high sensitivity and selectivity, accuracy and precision, and low cost. Among these, flame atomic absorption spectrometry (FAAS) is generally employed for NMs determinations in concentrates due to the poor sensitivity of the technique for these elements, but has the advantage of being less subjected to chemical and spectral interferences. Graphite furnace atomic absorption spectrometry (GFAAS) is more sensitive than FAAS, but it undergoes more severe interferences and lower reproducibility [3]. In the atomic absorption determination of NMs, the solid sample have to undergo wet chemical attack with strong acids like aqua regia, HCl, HNO3 , HBr, Br2 , H2 O2 , etc., in a hot plate or using high-pressure digestion systems owing to the high ionization potential and high boiling points of NMs. Exception is made for Ru, which forms a volatile (108 ◦ C) ruthenium tetraoxide during the high-temperature oxidizing dissolution process [4]. The correct result of the analysis depends on the completeness of transferring of analyte from solid sample onto solution, avoiding analyte losses due to sampling evaporation to dryness. Therefore, a sample dissolution procedure has to be well planned prior to AAS determination. Recently, reviews on the dissolution procedure for the determination of NMs in geologist and environmental samples have been reported [5,6]. More recently, Balcerzak [7] has written a review on the sample digestion methods for the determination of traces of precious metals prior to the spectrometric analysis. Various laborious procedures for the digestion of carbon-supported platinum-based catalysts, which make use of mixture of both HNO3 and HCl (1 + 6 (v/v)) and aqua regia [8–10], have been reported. The use of aqua regia for the mineralization of platinum metal by complexation with HCl on the self-protecting passivation layer of platinum obtained by the oxidant acid action is mandatory. It has been reported that hot HCl is able to dissolve some metallic platinum catalyst supported on MgO, previously calcined in reducing conditions, because nanoparticles of Pt are finely dispersed onto the support so that the atmospheric O2 is sufficient to guarantee a complete dissolution [11]. However, poor recoveries have been obtained with fresh/calcined and spent catalyst samples. In the present work, a simple accurate and precise method for the determination of platinum, ruthenium, and molybdenum in Pt, PtRu, and PtMo nanoparticles catalysts finely dispersed on high surface-activated carbon by flame and GFAAS is reported. The digestion of noble and non-NMs is achieved by using both concentrated HCl and a mixture of HCl+HNO3 (1+1 (v/v)) to boiling for 30 min. The dissolution procedure is fast, economic, and environment friendly because repeated evaporation of the acid to eliminate nitrates nitrosyl compounds is not required. Evaluation of the chemical interferences are made for both techniques. The proposed method may be also applicable for the certification of carbon-supported platinum catalyst used as reference materials for EDX analysis.
2. Experimental 2.1. Instrumentation A Varian SpectrAA 220FS (Victoria, Australia) flame atomic absorption spectrometer equipped with a GTA110 graphite furnace, an auto sampler and a deuterium lamp as background correction system was used. Hollow cathode lamps (Varian) of Pt, Ru, and Mo were used as sources. A pneumatic nebulizer with a glass impact bead was used. The instrumental parameters of the spectrometer were as follows: wavelengths and width bands at 265.9 and 0.2 nm for Pt; 349.9 and 0.2 nm for Ru; and 313.3 and 0.5 nm for Mo, respectively. Integration time was 10 s. Observation height varied between 6 and 12 mm. A slot burner for air–acetylene flame was used. Acetylene flow-rate was varied between 1.5 and 4.0 l min−1 and air flow was kept constant at 13.5 l min−1 . Pyrolytic graphite-coated graphite tubes were used. Argon (99.999%) gas was fluxed in the graphite tube. Peak height absorbance signals were measured by the AA instrument. The volume of the solution introduced into the graphite tube was 20 l. The samples were weighted with a micro balance (AND HM-202) with 0.001 g accuracy. 2.2. Reagents and standard solutions All acids were of analytical-reagent grade. Ultra high-purity water with a specific resistance of 18 M cm obtained from a Milli-Q water purification system (Millipore, Bedford, MA, USA) and filtered through a 0.22 m membrane filter was used throughout. Platinum and ruthenium stock solutions of 980 and 1005 mg l−1 in 5 wt.% HCl, respectively, were p.a. grade from Aldrich. Molybdenum stock solution of 1000 mg l−1 was obtained by dissolution of an appropriate amount of pure molybdenum lamina (p.a. Aldrich) in 3 ml of a hot mixture of HCl + HNO3 (1 + 1 (v/v)) for 30 min. Working standard 5–40 mg l−1 Pt, 4–10 mg l−1 Ru, and 10–50 mg l−1 Mo solutions were prepared daily by one step dilution of the standard stock solutions with 1 M HCl for FAAS measurements. Working standard 0.1–0.8 mg l−1 Pt, 0.08–0.32 mg l−1 Ru, and 0.0125–0.0375 mg l−1 Mo solutions were prepared daily by stepwise dilutions of the standard stock solutions with 1 M HCl for GFAAS measurements. To evaluate the interferences from the co-existing element matrix-matched standard solutions were prepared by adding the concomitant Pt, Ru, and Mo at the same concentration as present in the sample solution. Standard Pt solutions containing 1% (m V−1 ) ascorbic acid (Merck, Germany) were also prepared. 2.3. Samples 2.3.1. Powder catalysts Commercially available powder catalysts Pt/C, PtRu/C, and PtMo/C from E-Tek (Inc. Natick, MA, USA) were
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
793
chosen in a wide range of bulk composition to encompass a large variation of analyte concentrations and validate the method. In addition, three Pt/C and PtRu/C powder catalyst samples were prepared by chemical reduction with Na2 S2 O4 of H2 PtCl6 and H2 RuCl5 ·H2 O solutions in presence of Vulcan XC-72R carbon (Cabot) to metallic Pt and Ru. The obtained powders were annealed at 600 ◦ C in Ar for 10 h. The nominal compositions were 10 wt.% Pt/C, 10 wt.% PtRu/C (atomic ratio Pt:Ru 1:1), and 15 wt.% PtRu/C (atomic ratio Pt:Ru 1:1). 2.3.2. Electrode catalysts Commercial 20 wt.% Pt/C, 15 wt.% PtRu/C (atomic ratio Pt:Ru 1:1) and 30 wt.% PtMo/C (atomic ratio Pt:Mo 4:1) powder catalysts were used for the preparation of three electrode catalysts. In brief, powder catalyst (5 mg) was mixed with 1 ml aqueous solution of Nafion and glycerol to form a slurry. One hundred microlitres of the slurry was spread on a disk (weight 0.010 g and surface area 1 cm2 ) of carbon paper by a spray technique and dried at 70 ◦ C for 1 h. The metal loading was Pt(0.1), Pt(0.05) + Ru(0.025) and Pt(0.133) + Mo(0.016) mg cm−2 , respectively. Prior to dissolution procedures, the samples were dried at 200 ◦ C for 2 h in an electric furnace to remove moisture and stored in a desiccator prior to analysis. 2.4. Powder and electrode catalysts dissolution procedure Accurately weighted sample (0.010–0.020 g) was digested gently in a 10 ml covered Pyrex beaker with 3 ml of: (i) aqua regia (HCl +HNO3 , 3+1 (v/v)) for Pt/C, PtRu/C, and PtMo/C; (ii) concentrated HCl for Pt/C and PtRu/C; (iii) a mixture of HCl + HNO3 (1 + 1 (v/v)) for PtMo/C on an electric hot plate for 30 min. After cooling, the solution was filtered through a teflon filter (45 mm diameter) with a pore size of 0.8 m in a millipore filtration apparatus under vacuum to separate it from the carbon. The filter and residue were washed with 3 ml of 1 M HCl. The solution along with the washings were added to 25 ml volumetric flask and made up to volume with 1 M HCl. The final solution was analyzed for the Pt, Ru, and Mo after appropriated dilution with 1 M HCl. 2.5. Optimization of flame composition The optimized flame parameters were obtained by introducing 25 mg l−1 Pt, Ru, and Mo acid-containing standard solutions into the nebulizer for different flame composition at a given observation height over the burner head, being the others parameter kept constant. The effect of acetylene flow-rate was more significant as compared to those of observation height (Fig. 1). The maximum absorbance
Fig. 1. The effect of acetylene flow-rate on absorption signals of 25 mg l−1 Pt, 10 mg l−1 Ru, and 25 mg l−1 Mo at different observation heights: 4 mm (circle); 6 mm (square); 8 mm (triangle up); 10 mm (triangle down).
signal for platinum was achieved at the same observation height (between 4 and 6 mm) regardless the flame composition. The optimized flame composition was as follows: a stoichiometric acetylene–air flame for platinum; a rich acetylene–air flame for Ru and Mo. 2.6. Optimization of graphite furnace The optimized drying and pyrolysis temperature and holding time were chosen for all the analytes. The temperature programs used for the graphite tube are described in Table 1.
794
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
Table 1 Heating programs of the graphite furnace Step
Temperature (◦ C)
Hold time (s)
Gas flow-rate (l min−1 )
Dry 1 2 3
85 95 120
5.0 40.0 10.0
3.0 3.0 3.0
Pyrolysis 1 1000 2 1000 3 1000
5.0 1.0 2.0
3.0 3.0 0.0
Atomization 1 2700 (Pt); 2600 (Ru, Mo) 2 2700 (Pt); 2600 (Ru, Mo) 3 2700 (Pt); 2600 (Ru, Mo)
1.3 2.0 2.0
0.0 0.0 3.0
3. Results and discussion 3.1. Determination of platinum, ruthenium, and molybdenum in Pt/C, PtRu/C, and PtMo/C powder catalysts by FAAS 3.1.1. Interferences studies The presence of high concentration of inorganic acids into the sample solution does not provoke any chemical interferences on the analytical signal of all the analytes under the recommended operating conditions of the flame. The poor sensitivity of platinum in the air–acetylene flame is enhanced twice time in presence of 1% (m V−1 ) ascorbic acid. The effect of the reducing agent on the atomization efficiency may be related to the formation of volatile platinum compounds in acid media and under reducing conditions as reported recently in Ref. [12]. No interference effects of co-existing Ru and Mo on the absorbance signal of platinum have been evidenced. A beneficial effect of the co-existing platinum on the Ru analytical signal has been observed. The influence on the ruthenium signal has been examined over the concentration up to 1000 mg l−1 platinum. A positive interference of about 50% is observed likely due to the formation of inter-metallic compounds with platinum. Any interferences on the analytical signal of molybdenum from the co-existing Pt element has been observed.
3.1.2. Sample digestion of powder catalysts with different acids Different sample digestion procedures for carbonsupported platinum-based catalysts have been evaluated. Besides digestion with aqua regia it has been also used concentrated HCl for acid decomposition of Pt/C and PtRu/C in order to reduce losses of Ru by RuO4 formation during digestion with oxidant acid. Recovery test has been carried out on commercial carbon-supported catalysts in order to evaluate the efficiency of the proposed digestion procedures. In Table 2, the results of the analysis of Pt/C and PtRu/C catalysts digested with both aqua regia and concentrated HCl are reported. The recovery is within 99–100%, being the FAAS gathered results in good agreement with the certificated values for all the acidic mixture used. A mixture of HCl + HNO3 (1 + 1 (v/v)) has been tested for the digestion of carbon-supported Pt and PtMo/C catalysts. In Table 2, the results of the analysis of Pt/C and PtMo/C catalysts digested in different acids are also reported. 3.1.3. Analytical figures of merit Calibration curves have been constructed against five acid-containing standard solutions and the figures of merit are listed in Table 3. For ruthenium determination calibration with standard additions has been also conducted to take in account the positive interference of co-existing platinum. The curves are linear with the correlation coefficients r2 better than 0.999 and the intercepts significantly do not deviate from zero at 95% confidence level. In order to obtain a higher precision of the measurements, the sample solutions have been diluted to a level near the centroide of the calibration graphs [13] (see Table 3). The precision, expressed as relative standard deviation of five readings, is better than 1% (see Table 2). 3.1.4. Powder catalyst samples analysis Three carbon-supported platinum-based powder catalysts, prepared in our laboratories, have been analyzed according to the proposed method. In Table 4, the analytical results for platinum and ruthenium in Pt/C and PtRu/C powder catalysts are listed. As it can be seen, a good agreement of the results with the nominal values has been found.
Table 2 FAAS results (wt.%)a of Pt, Ru, and Mo in commercial Pt/C, PtRu/C, PtMo/C powder catalysts after digestion in different acids Catalyst
Pt
Ru HCl + HNO3 (1 + 1 (v/v))
HCl 10 wt.% 20 wt.% 30 wt.% 15 wt.% 20 wt.% 30 wt.% a
Pt/C Pt/C Pt/C PtRu/C (atomic ratio Pt:Ru 1:1) PtRu/C (atomic ratio Pt:Ru 1:3) PtMo/C (atomic ratio Pt:Mo 4:1)
Mean ± S.D. of five measurements.
9.95 20.5 29.9 9.90 7.80
± ± ± ± ±
0.2 0.2 0.3 0.1 0.07 26.7 ± 0.2
Aqua regia 90.2 20.0 ± 0.2 29.8 ± 0.6 9.7 ± 0.1 7.75 ± 0.06 26.5 ± 0.2
Mo
HCl
Aqua regia
5.15 ± 0.05 11.7.0 ± 0.1
4.90 ± 0.05 11.0 ± 0.1
HCl + HNO3 (1 + 1 (v/v))
Aqua regia
3.25 ± 0.03
3.20 ± 0.03
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
795
Table 3 Characteristic parameters of the analytical calibration curves obtained for the determination of platinum, ruthenium, and molybdenum by FAAS
Correlation coefficient (n = 5) Equation of the regression lines (Abs vs. concentration) (mg l−1 ) Centroid of the regression line (x, y) sb (standard deviation of the slope) sa (standard deviation of the intercept) Working linear range (mg l−1 ) Dilution of sample solution (v/v) r2
Pt
Ru
Mo
0.9994 y = 0.001227x + 7 × 10−5
0.99998 y = 0.002830x + 1 × 10−6
0.99994 y = 0.00248x + 1 × 10−4
14, 0.017 5 × 10−6 9 × 10−5 4–25 1+9
5, 0.015 3 × 10−6 1 × 10−6 4–10 1+4
30, 0.072 1 × 10−5 4 × 10−4 10–50 1+4
Table 4 FAAS results (wt.%)a of Pt and Ru in Pt/C and PtRu/C powder catalysts after digestion in concentrated HCl Catalyst
Pt
10 wt.% Pt/C 10 wt.% PtRu/C (atomic ratio Pt:Ru 1:1) 15 wt.% PtRu/C (atomic ratio Pt:Ru 1:1) a
Ru
HCl
Aqua regia
HCl
Aqua regia
9.95 ± 0.2 7.6 ± 0.08 10.2 ± 0.1
9.90 ± 0.2 7.7 ± 0.07 10.1 ± 0.1
3.3 ± 0.05 4.6 ± 0.05
3.2 ± 0.04 4.5 ± 0.04
Mean ± S.D. of five measurements.
Table 5 Characteristic parameters of the analytical calibration curves obtained for the determination of platinum, ruthenium, and molybdenum by GFAAS
Correlation coefficient r2 (n = 3) Equation of the regression lines (Abs vs. concentration) (mg l−1 ) Centroid of the regression line (x, y) sb (standard deviation of the slope) sa (standard deviation of the intercept) Working linear range (mg l−1 ) Dilution of sample solution (v/v)
Pt
Ru
Mo
0.99998 y = 0.613x + 1 × 10−3 0.4, 0.2 1 × 10−3 3 × 10−3 0.1–0.8 1+4
0.99992 y = 1.01x + 1 × 10−4 0.16, 0.16 6 × 10−3 9 × 10−4 0.08–0.32 1+4
0.9999 y = 7.0x + 2 × 10−4 0.02, 0.14 3 × 10−2 1 × 10−4 0.0125–0.0375 1 + 24
3.2. Determination of platinum, ruthenium, and molybdenum in Pt/C, PtRu/C, and PtMo/C electrode catalysts by GFAAS 3.2.1. Interference studies The determination of 4 ng Pt, 4 ng Ru, and 0.4 ng Mo with the co-existing elements have been made using the graphite furnace program given in Table 1. Any interference effects of the elements with each other during their determination have been evidenced. The pyrolysis/atomization curves are similar for platinum and ruthenium and any loss of analyte due to the acid-interference have been seen. A decrease of sensitivity of the molybdenum signal in the graphite tube has been observed with the increasing of the number of firings. This
is due to the formation of the molybdenum carbide during atomization [14]. It has been reported the beneficial effect of platinum on the sensitivity and reproducibility of the analytical signal of molybdenum in the graphite furnace. In fact, in presence of NMs the formation of inter-metallic compound of PtMo likely occurs, which dissociating rapidly during atomization avoids the formation of molybdenum carbide, thus improving the reproducibility of measurements [15]. 3.2.2. Analytical figures of merit Calibration curves have been constructed against three acid-containing aqueous standard solutions and the figures of merit are listed in Table 5. The curves are linear with the correlation coefficients r2 better than 0.9999 and the intercepts
Table 6 GFAAS results (mg)a of Pt, Ru, and Mo in commercial Pt/C, PtRu/C, and PtMo/C deposited onto carbon paper electrodes after digestion in different acids Metal loading (mg)
Pt HCl
Pt(0.1) Pt(0.05) + Ru(0.025) Pt(0.133) + Mo(0.016) a
0.100 ± 0.004 0.049 ± 0.001
Mean ± S.D. of three measurements.
HCl + HNO3 (1 + 1 (v/v))
0.130 ± 0.003
Ru
Mo
HCl
HCl + HNO3 (1 + 1 (v/v))
0.0240 ± 0.0005
0.0150 ± 0.0003
796
S. Scaccia, B. Goszczynska / Talanta 63 (2004) 791–796
do not significantly deviate from zero at 95% confidence level for all the analytes. For the determination of Mo, 50 ng Pt are co-injected into the graphite tube as a chemical modifier resulting in a good precision and better linearity of the calibration curve. The use of the chemical modifier results also in an enhancement of the tube lifetime, lowering the analysis cost. 3.2.3. Electrode catalyst samples analysis The analysis of the electrode catalysts made by using commercial powder catalysts with a nominal metal loadings of 0.1 mg cm−2 Pt, 0.05 + 0.025 mg cm−2 Pt + Ru, and 0.133 + 0.015 mg cm−2 Pt + Mo by GFAAS were in a good agreement with the found values, as shown in Table 6.
4. Discussion and conclusions The determination of NMs (Pt and Ru) and non-NMs (Mo) in carbon-supported platinum-based powder catalysts after digestion in different acids has been evaluated by analyzing the sample solutions by FAAS. It has been found that the use of both concentrated HCl and HCl + HNO3 (1 + 1 (v/v)) mixture for decomposition of Pt, Ru, and Mo nanoparticles finely deposited on carbon allows to dissolve accurately the noble and non-NMs, under mild conditions compared to the usual digestion method with aqua regia. Conversely, the determination of the Pt, Ru, and Mo loaded
onto carbon paper electrode disks, made by using powder catalysts, is carried out after dissolution of samples in different acids by GFAAS due to the higher sensitivity compared to the FAAS. In fact, metal loadings on electrode catalysts fall under the detection limit of quantification of FAAS for all the analytes studied. References [1] E.M. Crabb, M.K. Ravikumar, Electrochim. Acta 46 (2001) 1033. [2] N. Fujiwara, K. Yasuda, T. Ioroi, Z. Siroma, Y. Miyazaki, Electrochim. Acta 47 (2002) 4079. [3] C.R.M. Rao, G.S. Reddi, Trends Anal. Chem. 19 (2000) 565. [4] M. Hoashi, R.R. Brooks, R.D. Reeves, Anal. Chim. Acta 232 (1990) 317. [5] J.G. Sen Gupta, Talanta 40 (1993) 791. [6] R.R. Barefoot, J.C. Van Loon, Talanta 49 (1999) 1. [7] M. Balcerzak, Anal. Sci. 18 (2002) 737. [8] M. Merdivan, R.S. Aygun, N. Kulcu, At. Spectrosc. 18 (1997) 122. ´ ecicka, D. Bystro´nska, Anal. Lett. 32 (1999) [9] M. Balcerzak, E. Swi˛ 1799. ´ ecicka, E. Balukiewicz, Talanta 48 (1999) 39. [10] M. Balcerzak, E. Swi˛ [11] S. Recchia, D. Monticelli, A. Pozzi, L. Rampazzi, C. Dossi, Fresenius J. Anal. Chem. 369 (2001) 403. [12] P. Pohl, W. Zyrnicki, J. Anal. At. Spectrom. 16 (2001) 1442. [13] J.C. Miller, J.N. Miller, Statistics for Analytical Chemistry, Ellis Horwood, Chichester, 1984, p. 140. [14] B. Welz (Ed.), Atomic Absorption Spectrometry, second revised ed., VCH, Weinheim, 1985, p. 306. [15] E.A. Piperaki, N.S. Thomaidis, I. Demis, J. Anal. At. Spectrom. 14 (1999) 1901.