AN IMPROVED AEROSOL GENERATION SYSTEM FOR THE PREPARATION OF XRF CALIBRATION FILTERS

Transcription

1 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol AN IMPROVED AEROSOL GENERATION SYSTEM FOR THE PREPARATION OF XRF CALIBRATION FILTERS C. Vanhoof, V. Corthouts and N. De Brucker Flemish Institute for Technological Research (Vito) Boeretang 200, 2400 Mol, Belgium ABSTRACT A fast and reliable method for the analysis of airborne loaded filters becomes more and more important for monitoring campaigns and inventory studies at sites with airborne pollution. X- ray fluorescence spectroscopy (XRF) can perform quantitative analysis on filters without any pretreatment. Quantitative determination of various metals on loaded filters requires the use of calibration curves set up with synthetic filters. For accurate analysis, the calibration standards have to be prepared with a comparable matrix composition as the real filter samples. Therefore, cellulose nitrate filters were loaded by means of aerosol generation with various metals. The performance of an optimized aerosol generation system, based on the ultrasonic nebulization of a multi-element solution, is presented in this work. A multi-element solution is converted into a fine dense aerosol by ultrasonic action. The produced aerosol is dried and collected onto a filter. Filters can be loaded sequentially with a maximum standard deviation of 5 % for various metal concentrations. The influence of loading time and alteration of the initial metal concentration on the obtained elemental concentration have been investigated. Finally, the validation of the XRF measurement method is described. The linearity and sensitivity of the calibration curves are calculated for the various metals. The XRF detection limits, repeatability, and accuracy are determined. The XRF analysis of real filter samples is confirmed by subsequent analysis of filter samples with ICP-OES after acid digestion. INTRODUCTION Nowadays the attention for environmental quality control is of growing interest. Monitoring of specified locations, potentially polluted, becomes a major issue. With the current analytical techniques such as inductively coupled plasma optical emission spectrometry, timeconsuming acid digestion procedures that are subject to contamination and loss of analytes, are required. Wavelength Dispersive X-ray Fluorescence spectrometry (WD-XRF) overcomes these problems and is capable of analyzing air loaded filters without any pretreatment. The XRF technique is susceptible to matrix and enhancement effects, but these are negligible for thin film samples. For accurate analysis, the XRF calibration standards have to possess a comparable matrix composition as the real filter samples. As standard filters are not commercially available, metal loaded filters for calibration need to be self-prepared. Several methods for the preparation of standards are already described in literature [1,2]. Haupt and Schaefer compared the droplet technique, based on the VDI Richtlinie 2267 (Association of German Engineers, Guideline 2267), with aerosol loaded filters [3,4]. The aerosol filter loading was achieved by atomizing a multi-element solution using a cross-flow nebulizer equipped with a Scott spray chamber. Experiments have shown that the use of aerosol loaded filters results in higher sensitivities compared to the droplet method.

2 This document was presented at the Denver X-ray Conference (DXC) on Applications of X-ray Analysis. Sponsored by the International Centre for Diffraction Data (ICDD). This document is provided by ICDD in cooperation with the authors and presenters of the DXC for the express purpose of educating the scientific community. All copyrights for the document are retained by ICDD. Usage is restricted for the purposes of education and scientific research. DXC Website ICDD Website -

3 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol In this paper, an optimized aerosol loading system is described. For an efficient generation of fine and dense aerosols, the aerosol generation was performed with an ultrasonic nebulization system instead of the pneumatic cross-flow analyzer, resulting in a higher aerosol yield. The produced aerosols were dried by heating at 150 C, followed by drying through a permeable membrane. The required loading levels on the filter, ranging between 11 and ng/cm 2, can be achieved by varying the initial concentration of nebulized solution and loading time. Furthermore, the XRF measurement method is extensively validated and the collected data are applied for the analysis of real filter samples. The XRF analysis of real filter samples is confirmed by subsequent analysis of filter samples with ICP-OES after acid digestion. EXPERIMENTAL The experimental set-up of the aerosol generation system is shown in figure 1. The aerosol generator is an ultrasonic nebulizer CETAC 5000-AT system equipped with a constant temperature cooling system. Ar flow Ultrasonic nebulizer Peristaltic pump Multi-element solution Vacuum pump Filter unit Figure 1: Flow chart of the aerosol loading process Multi-element standard solutions containing Cu, Zn, Pb, Ni, As, Cd, Cr, Co and Mo were prepared from corresponding single element Spex Plasma standard solutions. For each element acidified standard solutions (2% HNO 3 ) with equal elemental concentrations of 0.1, 1, 10 and 100 mg/l respectivily, were mixed mg/l calcium nitrate solution was added as a filler. For Sb, the standard solution was prepared independently from a Spex Plasma standard solution, without addition of acid or filler. With a peristaltic pump at a flow rate of 0.9 ml/min, an aqueous multi-element solution is pumped onto an air-cooled piezoelectric transducer where a fine and dense aerosol is formed. The Ar carrier flow, regulated at 160 ml/min, transports the wet aerosol through a heating tube at 150 C where the solvent is evaporated. The solvent vapor is separated from the aerosol by passing the stream through a chilled condenser. Additionally, the aerosol is dried by passing through a drying tube, consisting of a permeable membrane purged with dried nitrogen. The dried aerosol is then collected onto a cellulose nitrate membrane filter (Sartorius) with a diameter of 50 mm and a porosity of 0.8 µm. The pump rate is controlled so that the suction rate is 20-30% higher than the Ar flow rate. The XRF measurements were performed with a wavelength-dispersive Bruker SRS 3000 X- ray spectrometer provided with the SpectraPlus software. The system was equipped with a rhodium X-ray tube, a 3000 kw generator, a flow-proportional detector and a scintillation detector, and a 58-position sample changer. The spectrometer operated under vacuum atmosphere, a 34 mm channel mask, no primary beam filter is used. The specified parameter

4 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol settings for the various elements are presented in table 1. For each element background intensities were measured at the left and the right site of the element peak position. The total measurement time for both background positions was 60 seconds, the same as the time set for measuring the element peak intensity. The filters were placed in a sample holder provided of a hollow Al backing cup. The effective measured filter area is 9.08 cm 2. With the described instrument settings a total elemental analysis can be performed only twice due to radiation damage of the filter. The heat production of the high power tube causes a degradation of the filters. However, no change in metal loading is observed during a complete measurement run, confirmed by ten repetitive measurements for each element on a filter. Spectral line Crystal Peak position ( ) Collimator ( ) Generator (kv/ma) BG1 ( ) BG2 ( ) CuKA1 LiF200 45,010 0,15 60/50 44,406 45,682 ZnKA1 LiF200 41,779 0,15 60/50 41,212 42,416 PbLB1 LiF200 28,260 0,15 60/50 27,674 28,708 CdLA1 Ge 74,556 0,15 30/100 73,736 75,254 NiKA1 LiF200 48,655 0,15 60/50 48,120 49,302 AsKB1 LiF200 30,448 0,15 60/50 29,904 CrKA1 LiF200 69,357 0,15 50/60 68,656 70,094 CoKA1 LiF200 52,795 0,15 60/50 52,180 53,476 MoKA1 LiF200 20,295 0,15 60/50 19,976 20,696 SbLA1 LiF ,385 0,46 30/ , ,966 BG1,2: background 1 and 2 Table 1: XRF measurement parameters The homogenity of the elemental deposition on the filter surface is determined by performing electron probe microanalysis (EPMA, Jeol JXA-8621 MX) equipped with an energy dispersive detector. The ICP-OES measurements were performed with a Perkin Elmer Optima Aerosol loaded filters were dissolved by microwave digestion in closed teflon bombs with HNO 3. For Sb, however, dissolution with HCl and HNO 3 was carried out in glass beakers placed on a hot plate until complete dissolution. Real filter samples were dissolved using a semi-open microwave digestion with HCl, HNO 3, HF and H 3 BO 3 [5]. RESULTS Aerosol filter loading: influence of element concentration and loading time With the instrumental set-up, described above, filters were loaded sequentially starting from a multi-element standard solution containing Cu, Zn, Pb, Ni, Cd, As, Cr, Co and Mo. Each loading process was repeated 5 times to define the in-between batch variation. The concentrations of the initial multi-element solutions were 1, 10 and 100 mg/l. The loading time was varied from 5 to 20 minutes for each solution. The influence of these factors on the aerosol filter loading of Cu is presented in figure 2. With increasing concentration and increasing loading time, the filter loading shows a linear increase. A similar profile is observed for all other elements. The results show that the in-between batch variation can be limited to 5% for loadings above 110 ng/cm 2, and to 10% for loadings below 110 ng/cm 2. These observations were confirmed in additional experiments, described below (see table 4), where the elemental concentration was determined by the XRF calibration curve based on aerosol loaded filters. Based on these observations, filters were aerosol loaded to achieve a concentration range between 11 and ng/cm 2 for each element. Subsequently, an ICP-OES analysis was performed. The yield was calculated as the ratio of the expected concentration and the ICP-

5 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol OES values. For the elements Pb, Cd, Co, As and Mo the recoveries were better than 85% covering the full concentration range. The elements Cu, Zn, Ni and Cr show recoveries above 85% for concentrations higher than 80 ng/cm 2, but for the lower concentrations recoveries of 50% were obtained. Experiments on blank filters have shown that these outliers are due to concentration fluctuations of these elements on blank filters. As an example the yield calculated for Mo and Co is shown in figure Aerosol loading of Cu Aerosol yield of Mo and Co Concentration (ng/cm 2 ) Concentration (ng/cm 2 ) Yield (%) Time loading (min) Cu-1 mg/l Cu-10 mg/l Cu-100 mg/l (sec.axis) Concentration (10 2 ng/cm 2 ) Mo Co Figure 2: Influence of the elemental concentration and loading time on the aerosol loading Figure 3: Aerosol yield of Mo and Co The elemental distribution on the filter surface was evaluated by performing an elemental mapping of the filter area with EPMA. Preliminary measurements for Pb and Zn confirmed the homogeneous distribution of the filter loading. Aerosol filter loading of Sb The addition of Sb to the multi-element solution is not possible due to the precipitation of Sb complexes in a nitric acid solution. Therefore, a 100 mg/l Sb solution was prepared in 2% HCl, including a 1000 mg/l Ca(NO 3 ) 2 filler solution. The Sb on the aerosol loaded filter shows a very weak XRF signal compared to the other elements loaded on filters. To evaluate the Sb aerosol loading process, the influence of addition of acid and filler was examined. Sb and Cu standard solutions were prepared in aqueous and acid medium, with and without addition of Ca(NO 3 ) 2. The XRF intensities of Sb and Cu were measured on 3 independent loaded filters. As shown in table 2, addition of acid and filler has no influence on the Cu intensity. The intensity of the Sb peak, however, decreases drastically, i.e. 77 %, when acid is added. Even addition of filler results in an intensity decrease of 19 %. The combination of acid and filler finally diminishes the intensity of Sb to 83%. Intensity Sb Yield Intensity Cu Yield kcps % kcps % 100 mg/l Sb and Cu mg/l Sb and Cu, % HCl 100 mg/l Sb and Cu, mg/l Ca(NO 3 ) mg/l Sb and Cu, 1000 mg/l Ca(NO 3 ) 2, 2% HCl Table 2: Aerosol loading of Sb on filters These losses are due to the formation of volatile complexes of antimony during the aerosol process, which are then deposited on the walls of the aerosol system.

6 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol Therefore, the preparation of aerosol-loaded filters for Sb was carried out with an aqueous Sb solution, without addition of acid or filler. The stability of the Sb solution was checked with ICP-OES. No variation occur in a period of 5 days. Particle size distribution of the aerosols The influence of the addition of a Ca(NO 3 ) 2 filler on the particle size distribution of the aerosols was examined. The particle size was determined with a Scanning Mobility Particle Sizer (TSI, Model 3071) followed by a Condensation Nucleus Counter (TSI, Model 3020). Particles between µm and 0.75 µm were measured. The particle size is measured gradually with intervals of µm for the small particles and increases to 0.1 µm for the large particles. Multi-element solutions containing 0.1, 1, 10 and 100 mg/l of each element (Cu, Zn, Pb, Ni, Cd, As, Cr, Co and Mo) were prepared with and without addition of filler. These solutions were nebulized and the output of the aerosol generator was connected to the particle size analyzer. The particle size distribution is calculated as a function of the cumulative mass percentage. Increasing the concentration of the nebulizing multi-element solution from 0.1 to 10 mg/l without filler results in an increase of the aerodynamic mean diameter (i.e. at the 50% point of the cumulative curve) of the aerosols, from 0.08 µm up to 0.24 µm. At higher concentration levels, 10 mg/l and 100 mg/l, there is a stagnation of the particle size diameter. Addition of 1000 mg/l filler to the solution results in an increase of the particle size mean diameter up to 0.4 µm for the various concentration levels, except for the 100 mg/l solution. In this solution, the total elemental concentration equals the concentration of filler. The standard deviation, which is a measure for the degree of dispersion, is determined by the difference in size between the mean diameter and the size corresponding to the 84% and 16% point of the cumulative curve [6]. The standard deviation of the aerosol particle size distribution, is 1.5 µm for the different nebulizing solutions without filler. Addition of a filler to the nebulizing solution results in a lowering of the standard deviation to 1.3 µm, indicating a narrowing of the particle size distribution. The significance of the difference is statistically confirmed by performing the F-test for variances. The measurements show that the addition of filler to the nebulizing solution results in the formation of aerosols with a uniform and narrow particle size distribution comparable to the fine fraction of real aerosol particles. Validation of the calibration curve on the basis of aerosol loaded filters Filters were loaded with the aerosol generation system, according to the method described. Concentrations for Cu, Zn, Pb, Ni, As, Cd, Cr, Co, Mo and Sb between 11 and 3300 ng/cm 2 were obtained. Aerosol-loaded filters were analyzed with WD-XRF using the instrumental set-up shown in table 1. Consequently, the theoretical loading of the various metals on the filters was determined by ICP-OES. The statistical evaluation of the linear calibration function, performed according to ISO , proves a linear behavior for all elements [7]. The performance characteristics including the calibration range, coefficient of variation (V xo ), limit of detection (L d ) and limit of quantitation (L q ), are given in table 3. These limits are calculated on the basis of calibration curves, set up with 10 standards around the detection limit region, and a confidence interval of 95% [8]. For Cu, Zn, Ni, Cr, Co, Mo and Sb the limits of quantitation are below 55 ng/cm 2. For Pb, Cd and As the limits of quantitation vary between 110 and 253 ng/cm 2. The coefficient of variation for all elements is less than 10%. The repeatability of the XRF measurement for the element Cu was determined by 10 successive measurements of the same filter. With increasing concentrations, the relative standard deviation decreases from 11%, at the lowest concentration level, to 0.8% for the upper concentration level of the calibration range.

7 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol Calibration V x0 L d L q Calibration V x0 L d L q range range (ng/cm 2 ) (%) (ng/cm 2 ) (ng/cm 2 ) (ng/cm 2 ) (%) (ng/cm 2 ) (ng/cm 2 ) Cu As Zn Cr Pb Co Cd Mo Ni Sb Table 3: Performance characteristics of the XRF method The repeatability of the aerosol loading process is determined by analyzing 5 sequentially loaded filters at 3 concentration levels. Element Conc. (n=5) RSD Conc. (n=5) RSD Conc. (n=5) RSD (ng/cm 2 ) % (ng/cm 2 ) % (ng/cm 2 ) % Cu Zn Pb < Cd < Ni As Cr Co Mo < Table 4: Determination of the repeatability The relative standard deviation is less than 10%, and at higher concentrations even less than 5% (table 4). The accuracy was determined by analyzing 6 real filter samples with XRF, followed by ICP- OES analysis. The correlation between the XRF and ICP-OES analysis for the elements Zn and Pb are given in figure 4. A similar correlation is obtained for all other elements. The data obtained from the aerosol calibration curve show a very good agreement with the data of ICP- OES. Deviations below 15%, mainly 5%, were achieved. Loading of Pb on filters Loading of Zn on filters 2 ) Pb-ICP (ng/cm y = x R 2 = Zn-ICP (ng/cm 2 ) y = 0.988x R 2 = Pb-XRF (ng/cm 2 ) Zn-XRF (ng/cm 2 ) Figure 4: Comparison between XRF and ICP-OES analysis of real filter samples

8 Copyright(c)JCPDS-International Centre for Diffraction Data 2000,Advances in X-ray Analysis,Vol CONCLUSION Aerosols loaded filters, prepared by ultrasonic nebulization, are suitable to be applied as standard filters to set up calibration curves for analysis of real airborne loaded filters. The optimized aerosol generation system is fast and easy to manipulate, and capable of preparing homogeneous loaded filters, comparable to real filter samples. Sequentially loading of membrane filters with identical process parameters, results in filter aerosol loading with a deviation less than 5 %. Various concentration levels on the filter can be obtained by altering the concentration of the metals in the nebulizing solution and the loading time. Analyses of real filter samples with ICP-OES prove the validity of the XRF measurement method, which is a fast and non-destructive analytical technique for the analysis of filter samples. Acknowledgements The authors wish to thank W. Brusten and J. Broeckx for performing the ICP-OES measurements and N. Bleux for his useful information on the measurements of the particle size distribution. REFERENCES [1] Galloo J.C. et Guillermon R., Analusis, 1989, v. 17, no 10, [2] Sulkowski M., Hirner A.V., X-ray spectrometry, 1996, vol 25, [3] Haupt O., Klaue B., Schaefer C., Dannecker W., X-ray spectrometry, 1995, volume 24, [4] VDI-Handbuch 'Reinhaltung der Luft', Band 4, Richtlinie 2267, Blatt 11 und 12, Beuth Verlag GmbH, Berlin, [5] PrEN (E) CEN/TC 292/WG3 N100 May 1998 [6] R. Dennis, Handbook on Aerosols, CGA corporation, Masschusetts, [7] International Organization for Standardization, ISO , Water Quality - Calibration and evaluation of analytical methods and estimation of performance characteristics; Part 1: statistical evaluation of the linear calibration function, International Organization for Standardization, Geneva. [8] Funk W., Dammann V., Vonderheid C. und Oehlmann G., Statistische methoden in der wasseranalytik; begriffe, strategien, anwendungen, VCH, Weinheim, 1985.