Determination of Toxic Elements in Liquid Hazardous Waste Using High-resolution Energy-dispersive X-ray Fluorescence Spectrometry P. A. RUSSELL*a AND R. JAMESb aOxford Instruments, Industrial Analysis Group, Abingdon, Oxfordshire, UK OX14 1TX bRechem International, Gwent, UK NP45DQ The analysis of liquid hazardous waste (LHW) prior to Process Control/Acid-generating Elements ( Br, Cl, I, P, S) disposal using high-temperature incineration or as an These elements are required analytes for the effective control alternative fuel is required for process and regulatory control.of the composition of combustion off-gases. The feed composi- The typical requirements of the industry sector are rapid tion must be controlled within the capacity of gas treatment screening to support decisions on the most appropriate facilities to ensure that emission limits are not breached. Where treatment of the waste. This paper reports the use of high- waste-derived fuels are manufactured to a specification, the resolution EDXRF spectrometry for the determination of level of confidence in the measured value must be high to halides and toxic heavy elements using a rapid technique ensure satisfactory performance.combined with a unique sample preparation methodology. Calibrations were developed using traceable certified 1000 mg l-1 aqueous standards and pure organic solvents. Heavy Metals (Cr, Cu, Ni, V, Zn) Samples were stabilized in an alumina matrix and measured against a suitable calibration.The benefits of the proposed These elements are required analytes to ensure environmental technique are as follows: (i) ease of obtaining calibration performance meets regulatory requirements for releases to all standards; (ii ) removal of sample history, i.e., control of media. These elements are commonly occurring (e.g., as wear matrix effects; (iii) minimization of analyte loss during sample metals/trace elements in fuel oils) and are often regulated in preparation; (iv) wide range of matrix types measurable using discharges to air and water.a single calibration, i.e., clean solvents to turbid sewage sludge; (v) accuracy of measurement, typically within 10% relative in the concentration range 10–100 mg kg-1 with a precision of Toxic/Volatile Metals (As, Cd, Hg, Pb, Sb, Se, Sn, Tl ) better than 5% relative; and (vi) speed of analysis, for >20 elements typically <15 min from receipt of sample. The These elements are required analytes owing to their potential results presented show high-resolution EDXRF to be ideally presence in trace amounts sufficiently high to result in unacsuited to the analysis of LHW owing to good heavy element ceptable atmospheric emission.The elements are sufficiently detection in the atomic number range 30–82 (Zn–Pb). volatile to evade many of the treatment stages that are effective Detection limits are in the range 3–17 mg kg-1 for heavy for heavy metals and, in addition, are potentially more hazardelements and below the working calibration range for Cl, P ous in their own right.The volatility of these metals has and S. These limits satisfy typical requirements for process historically resulted in techniques involving ashing, heating or and regulatory control, which are of importance in the range digestion being found inadequate for determinations in waste >0.1% m/m for P, S and Cl and >50 mg kg-1 for the matrices owing to volatile loss.heavier elements. Keywords: T oxic elements; liquid hazardous waste; high- Precision, Accuracy and Appropriate Analysis resolution energy-dispersive X-ray fluorescence spectrometry; matrix modification The typical level of accuracy and precision achieved in the analysis of liquefied waste is generally considered to be acceptable if better than 20% relative. The contribution from interelement effects to the accuracy of an analysis can be significant The analysis of waste for the purposes of disposal risk assess- in LHWs.The use of matrix modification methods reduces ment features some unique difficulties resulting from the these inter-element or matrix effects and their accuracy needs extremely wide range of matrices encountered. The types of typically to be <5% for the matrix modification technique to elemental analytes and relevant ranges of measured values be of value. The alumina method proposed in this paper relies selected for waste assessment usually reflect requirements of on the principle of little or no inter-element effects being environmental regulation and waste disposal licensing.1 The present in the final specimen, which is demonstrated by the key purposes of these measurements are (i) to ensure com- linearity of the calibrations for all elements studied.Dilution pliance with regulatory requirements (e.g., for waste-derived into the linear calibration ranges is used to maintain the fuels), (ii) to allow appropriate selection of disposal strategy, minimization of matrix effects while still achieving appropriate (iii) to confirm the waste composition and (iv) to ensure detection limits for the elements of interest for both process appropriate process control where a treatment or incineration and regulatory control.The reduction of matrix effects enables disposal option is selected. The target analytes can be con- calibrations to be developed from pure element standards sidered to fall into three groups, (i) process control/acid- prepared in aqueous or organic solvent.generating elements, (ii) heavy metals and (iii) toxic/volatile Table 1 identifies the main areas of potential error encountered in any analytical technique used for the analysis of LHWs. metals, as follows. Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 (25–32) 25Table 1 Key sources of error in the analysis of liquid hazardous waste Estimated error Systematic Means of error control proposed Source of error magnitude (%) (Y/N) in this work Phase distribution of sample 0–typically 50 N Addressed by mixing with alumina in one stage Analyte losses 0–100 N No preparation/digestion stages ensuring no analyte loss Chemical interferences 0–100 N No preparation/digestion stages preventing uncontrolled chemical reactions Instrumental interferences 1–15 Y XRF spectra simple.Alumina normalizes matrix effects and reduces inter-element effects to <5% Linearity over measured range 0–25 Y Simple sample dilution after s/w out of range warnings Stability of calibrations 0–15 Y Very stable technique and measures a static sample identical to calibration standards REVIEW OF EXISTING METHODOLOGIES FOR explosives and, although this may be known, they cannot be easily and/or safely exposed to thermal digestion sample LHW ANALYSIS preparation techniques.Bomb digestion is used, followed by A brief review of the literature reveals a surprising lack of either potentiometric titration for determination of halogens work being carried out to solve some of the difficult sample or ion chromatography13 for determination of halogens, PO42- preparation problems encountered in the analysis of LHWs.and SO42-. The bomb technique relies on complete conversion The most recent review of environmental analysis by Cresser of organic compounds into inorganic forms, i.e., halogenated et al.2 reported only five references relating to LHWs, all of solvents to X- (X=Br, Cl, F, I) and subsequent quantitative which were on the determination of a few elements in waste transfer of these components to the final measurement tech- water.In general, dissolution techniques used for AAS or AES nique. Typical analysis times for these methods are 30–45 min. analysis of soils and waters are also used for LHWs. Various Common interferences occur in the measurement of I in the reviews2–6 of atomic spectrometry show an emphasis on extrac- presence of sulfate using ion chromatography and the forma- tion, preconcentration and digestion methods of sample prep- tion of stable metal halides during digestion yields low recover- aration.A number of key problems occur when using extensive ies. In the analysis of LHWs a flame test should be carried dissolution and/or extraction procedures for LHW analysis. out before attempting bomb digestion.If the flame test indicates (i) The effects of organic solvents on AAS/AES element a very high calorific value, i.e., potentially explosive, then the sensitivities7–9 and the variable and unpredictable nature of bomb technique would not be appropriate and an alternative LHWs preclude even direct introduction of clean solvents, i.e., technique must be used. Frequently there is no alternative containing no undissolved material. This drives the need for digestion technique available. sample preparation methods based on digestion or preconcentration to produce a common matrix form.XRF is capable of measuring samples in their ‘as-received’ state and therefore should be an ideal technique for the analysis (ii) Potential loss of analytes such as Cd, Hg, Pb, Tl and Se during digestion10 and/or atomization11 will always affect the ofLHW in all its forms, i.e., sludges to clean solvents. Decreases in analysis time and cost, in addition to lowering potential confidence of measurements made when digestion preparation techniques are employed.analyte losses through eliminating extensive sample preparation, demonstrate the obvious advantages to this approach. (iii) Sample preparation time is an issue for industrial incineration plants. Methods that employ lengthy clean-up West et al.14 reported on various methods for the analysis of electroplating bath sludges. They first dried the samples, then preparation procedures which, although ultimately producing high-quality analyses, are not cost-effective.Peraniemi et al.12 prepared lithium metaborate fusions or pressed pellets for WDXRF analysis. This approach relies on a preconcentration reported on a preconcentration techniquefor the determination of P in waste water using EDXRF. The sample preparation drying stage followed by digestion (fusion) or pelletization to allow a traditional XRF analysis to be used. As the sludge time was quoted as between 3 and 6 h plus overnight drying.The spectrometer available to the researchers was unable to matrix is aqueous and of a predictable nature, this approach, although time consuming, gave good results. However, it would determine P directly from a filtrate (which would have reduced the preparation time) owing to spectral overlap from neigh- be inappropriate for organic-based sludges and suffers from an inability to determine halogens or S in organic form. Lucke bouring elements. (iv) Spectral interferences, especially from Cl and other et al.15 dried sewage sludge prior to analysis and results were reported on the basis of spikes made to this dried matrix.halogens, are encountered in AAS,3 ETAAS,11 ICP and ICP-MS.9 Vanhoe et al.9 described a number of mass spectral Results from determinations of elements such as Cd, Pb and Hg made using this approach may be low owing to losses of interferences encountered in the ICP-MS analysis of biological samples due to Cl, Ar and O species combinations.Biological these volatile elements during the drying stage. A few researchers have studied the use of simple matrix samples can be considered to be similar in nature to LHW sample types because of their high concentrations of C, O and modification methods for XRF analysis which avoid heating and/or digestion. Seiber,16 as reported in a review by Marshall often Cl. Species combinations of Cl–Ar–O were reported as potential interferents with many environmentally sensitive et al.,5 used a binding powder obtained from Chemplex Industries to stabilize grease and oil samples prior to XRF elements, e.g., As, Cr, Se and V.Suppression of these interferences again relied upon complicated extraction or digestion analysis. In a Chinese paper by Liu et al.,8 reported in a review by Bacon et al.,17 MgO powder was used to stabilize oil techniques. (v) The use of bomb digestion can be highly dangerous samples followed by low-temperature heating (270 °C) prior to the determination of Cr, Cu, Fe, Mn, Ni, V and Zn by EDXRF when unknown organic solvents are mixed with acids or oxygen.Many LHWs are explosive or may be precursors to spectrometry. Carbon is also used as a simple matrix modifier 26 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12in a newly released ASTM method18 for the determination of such as Cd are linear at trace concentrations, i.e., the mass of alumina+sample used to fill a sample cup does not affect the toxic metals in liquefied waste prior to its use as an alternative fuel for cement kilns.The carbon matrix method also uses sensitivity of the heavy elements, e.g., the infinite thickness of Cd in alumina is approximately 2.3 cm compared with 26.4 cm low-temperature heating similar to that of Liu et al.8 The heating of the sample in both of the above techniques precludes in a carbon matrix. (iv) Compared with a calibration based on liquids of varying the determination of halogens, S and P, and as these are critical elements the techniques have limited use in process density, there is no need to carry out a background correction, i.e., ratio to Compton scatter.For elements determined from control analysis for a high-temperature incinerator. The use of a light matrix, i.e., carbon or MgO, is an their X-ray lines at energies above the Compton peak, background correction is not possible using the Compton scatter important aspect in the development of simple matrix modifi- cation techniques for use in the analysis of LHW by XRF.peak. (v) Mixing a sample with alumina ensures that liquid phases The use of low-power X-ray tubes in many EDXRF spectrometers reduces potential losses of analyte from samples cannot separate prior to or during analysis and that undissolved solids are correctly incorporated into the analysis, i.e., through heating. Excellent detection of elements with X-ray lines in the energy range 8–40 keV makes EDXRF the preferred do not settle on the support film, thus introducing a bias.This is especially important in the analysis of sewage sludge. choice of XRF for the analysis of LHW. Existing methods for inorganic elemental analysis of LHW (vi) The sample used for analysis is safer to handle in the spectrometer, i.e., leakages cannot occur in the sample support and their limitations are summarized in Table 2. film owing to small punctures or chemical attack during analysis.FUNDAMENTALS OF THE ACTIVATED (vii) The sample mass used for analysis is relatively large ALUMINA METHOD FOR XRF (5 g) compared with alternative techniques. All techniques used for the analysis of liquefied waste suffer from a common problem, i.e., matrix variation and its effect EXPERIMENTAL on analyses based on multiple standards calibrations. Using Sample Preparation high-resolution EDXRF the use of activated alumina mixed with the sample minimizes this effect.Materials A key feature of this technique is the use of a well known Calcined alumina (1500°C) (Merck/BDH, Poole, Dorset, UK), principle for the analysis of complex matrices by XRF. For 50–60 ml wide-mouthed HDPE sample bottles, a minimum of many years the technique of preparing glass fusion beads using two 1 cm diameter stainless-steel ball-bearings or similar heavy lithium borate mixes for the analysis of geological and other stable material, Chemplex Industries (Tuckahoe, NY, USA) or materials has been used to great effect.This principle relies on Spex Industries (Edison, NJ, USA) 31.5 mm X-ray sample reducing, by dilution, all samples to a common matrix form cups, 4 mm Prolene film (Chemplex Industries) and 4 mm and thus largely eliminating the sample history. The alumina Hostaphan high-purity polyester film (Oxford Instruments, method uses this principle of diluting a sample to a common Abingdon, Oxfordshire, UK) were used.matrix form with the following additional benefits in respect of the analysis of LHW: (i) The alumina is activated. This will stabilizehighly volatile Standard solutions solvents, reducing losses that may occur due to thermal and Standard solutions used were for AAS 1000 mg l-1 pure ana- X-ray heating in the spectrometer during analysis [a common lyte aqueous standard solutions of As, Cd, Cr, Cu, Fe, Hg, Ni, problem with pure solvent(s) samples especially in high wattage Pb, Sb, Se, Sn, Tl, V and Zn, for P, triethyl phosphate, for S, spectrometers], thus avoiding an obvious area of bias.dithioglycol, for Cl, trichlorobenzene, for Br, (ii) The diluting matrix, i.e., alumina, is significantly heavier 1-bromonaphthalene and for I, iodobenzoic acid, and low in atomic number than its fusion bead equivalent. This gives molecular mass polyethylene glycol (PEG 400). the benefit of reducing background in the XRF spectrum compared with that produced by a low-Z organic or aqueous solution in addition to substantially reducing inter-element Preparation Procedure for Standards and Samples and matrix effects.(iii) The density of an alumina mixed sample compared Weigh 15±0.01 g of alumina into a 60 ml wide-mouthed polyethylene bottle, add 5±0.01 g of sample to the bottle, with the original liquid form of the sample helps to ensure that the alumina sample is infinitely thick with respect to the place two 1 cm diameter stainless-steel ball-bearings or similar mixing device in the bottle, seal the bottle with the screw-cap K-series fluorescence X-rays emitted from heavy elements such as Cd or Sb.This ensures that calibrations for heavy elements and shake vigorously until the sample and alumina are com- Table 2 Summary of existing methods and limitations Technique Reference Elements Sample preparation Key limitations Classical 13 Metals Presence of unknown metals; colour; difficult to Digestion, pH and buffering control, titrations control pH Bomb calorimetry 10, 13 Halides, P, S Addition of reactive agents Potentially determines only organic halides not total owing to complexing of inorganic halides with metal species; oxidation prevents determination of I ICP-AES, ICP-MS 3–5, 7, 9, 10 Halides, P, S, Digestion Halide interference; heating during preparation trace metals causes loss of important analytes, e.g., Pb, Tl, Hg; particulates; large conc.of alkali metals can cause significant interference; organic solvents AAS 3–5, 10, 11 Trace metals Digestion Halides interference; particulates; organic solvents XRF 8, 12, 14–16, Halides, P, S, As received, fusion, digestion Homogeneity; loss of analytes in fusion or 18 trace metals digestion process Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 27pletely mixed (approximately 30 s). Tapping the bottle on a another. Tables 3 and 4 list the parameters used in this work for the two calibrations studied.The instrument used was an hard surface will aid the mixing process. Transfer sufficient sample to fill a standard 31.5 mm diameter vented Chemplex Oxford Instruments ED2000 EDXRF spectrometer fitted with a silver target X-ray tube. The benefit of the Ag target tube sample cup fitted with a suitable X-ray transmission film for the elements to be measured. Tap the cup gently on a flat compared with the more commonly used Rh target tube are twofold: improved excitation of Cl by Ag L lines; Rh L lines surface (analysis face down) to compact the sample and remove any air gaps.The sample is now ready for analysis. The do not excite Cl and produce serious spectral interference in energy-dispersive spectrometers at low Cl concentrations; and analysis is carried out using a calibration based on the same alumina to sample ratio. a heavier Z number target, i.e., Ag, gives more high-energy excitation for improved heavy element sensitivity. Intensities were obtained using the method of least-squares Calibration peak fitting to library spectra19 using XpertEase software (an Oxford Instruments proprietary EDXRF software package for Analytical standards should be prepared gravimetrically by Microsoft Windows operating systems).blending the pure element standards into a suitable calibration suite as determined by the final analytical requirements. A requirement for two calibrations was identified based on the process and regulatory controls needed.These are identified Routine Analysis as: ‘light elements and halogens’ and ‘toxic elements.’ The concentration ranges for each calibration are as follows: light An important advantage of this technique is the ability to dilute a sample into the concentration range of the calibration. elements and halogens, P, S and Cl 0.05–5%, Br 0–1% and I 0–0.2%; and toxic elements, all elements 0–600 mg kg-1. If the result reported from an analysis is outside the calibration range, then a dilution of the sample can be made by changing Standards can be single elements or mixtures. Standard solutions can then be mixed at a 15 g alumina to 5 g standard the mass of sample placed on the alumina but maintaining the alumina to sample ratio. This is carried out either by adding ratio to produce the final calibrants. Note: more than one standard can be added to a single 15 g the required balance up to 5 g as PEG or similar material to a reduced mass of sample, or by taking a sub-sample of the alumina measure so long as the total mass of standard equals 5 g.This will maintain the alumina to sample ratio whilst waste–alumina mix and mixing it with a PEG–alumina blank at a suitable diluting ratio, e.g., 1 g of waste–alumina sample allowing mixtures of incompatible standards/elements to be manufactured. made up to 10 g using the blank gives a 10-fold dilution. A simple post-analytical calculation will give the true concen- Empirical calibrations using a suite of standards as described above were used to develop the two methods required, i.e., tration of the analyte in the original sample. Bulk sampling will depend on two main factors, i.e., viscosity light elements and halogens and toxic elements.Standard concentrations were limited to a maximum of 600 mg kg-1 for and homogeneity. Larger samples can be taken to ensure a suitable sample size if homogeneity is poor.The alumina to the heavy trace elements as this was the maximum concentration which could be obtained from 1000 mg kg-1 stock sample ratio is maintained by increasing the mass of alumina used. Liquids that are highly viscous can be transferred by AAS aqueous standard solutions whilst minimizing the number of individual standards required, i.e., a maximum of three thoroughly mixing and then pouring or spooning an accurately weighed amount of approximately 5 g. Again the alumina to elements in the highest standards were used.Serial dilutions of pure analyte standards were used to produce each element sample ratio is maintained by adding three times as much alumina as sample used. Both of the above approaches will calibration. Mixed analyte standards were used to assess the spectral interferences of the spectrometer and check for cali- allow direct reporting of the final results from the calibration. bration linearity. Note: samples are diluted into the linear range using a suitable diluent, i.e., distilled water or PEG, e.g., 5000 mg kg-1 Method Validation Pb solution is diluted into the 0–600 mg kg-1 calibration range by using 15 g of alumina to 0.5 g of sample+4.5 g of The use of reference standards for validating this method is diluent.This gives a 10-fold dilution and maintains the alumina not possible. The term liquid waste implies that the material to sample ratio. for analysis cannot be classified in terms of a specific matrix and as such no standard reference material is available. Method validation is therefore based entirely upon spiking real waste Recommended Spectrometer Conditions samples with known concentrations of analytes (similar to the technique of standards additions).Experiments were carried Specific parameters for obtaining the optimum performance for ranges of elements will vary from one spectrometer to out to determine the following: Table 3 Instrumental conditions for toxic elements Condition Path Voltage/kV Current/mA Primary beam filter Elements measured Live time/s 1 Air 15 1000 Thick Al V–Fe 100 2 Air 35 65 Thin Ag Ni–Br 100 3 Air 45 100 Thick Ag Hg–Pb 150 4 Air 50 500 Thick Cu Cd–I 150 Table 4 Instrumental conditions for light elements and halogens Condition Path Voltage/kV Current/mA Primary beam filter Elements measured Live time/s 1 He 5 1000 None P–K 150 2 He 35 65 Thin Ag Br 50 3 He 50 500 Thick Cu I 50 28 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12(1) Blank measurement of the alumina sample preparation.Accuracy A sample containing only PEG was used to check for any Table 7 gives the results of the measurement of calibration bias in the calibrations at the zero concentration level. standards. The accuracy figure shows the match of given versus (2) Accuracy of the calibrations. Standards were run against calculated concentration for each element. the calibrations to assess accuracy and to check for bias in the calibrations due to either matrix or spectral effects.Matrix/matrix spike recoveries (3) Analyte spiked recoveries using real waste samples. Three types of actual waste solutions were selected from routine Tables 8–10 for the toxic elements and Tables 11–13 for the test samples taken at an incineration plant: clear solutions, light elements and halogens give results for the measurement turbid solutions, i.e., containing significant solids not in of the spiked samples separated into the three matrix types.suspension, and biphasal solutions, i.e., containing two Using eqn. (2), a recovery figure for each analyte in each distinctly immiscible liquid phases. matrix type was determined. The results, referred to as a matrix spike/matrix spike duplicate (MS/MSD), are shown below. For each matrix type a sub-sample was spiked with a known concentration of analyte. The spiked sample was prepared MS/MSD recovery (%)=[(C2-D1C1 )/C3]×100 (2) using the alumina technique and measured.where D1=dilution factor due to matrix spike addition= 1-(mass of spike)/(total mass of sample+spike), C1=calcu- RESULTS lated concentration of matrix without spike, C2=calculated concentration of matrix+spike and C3=given concentration Calibration Details and Standard Errors of matrix spike. Table 5 shows the 3s lower limit of detection and regression details for all elements included in this study. The lower limits Method Reproducibility of detection (LLD) for each analyte in Table 5 are based on the following equation: A number of replicate analyses were made on a waste sample to demonstrate the reproducibility of the sample preparation on the two calibrations, i.e., toxic elements and light elements LLD= 3Óbg net peak × 1 ÓT ×conc.(1) and halogens. (1) A single measurement from each of 10 repeat sample preparations was made for Cl content. A repeat of this where bg=background intensity under analyte peak (cps), net process was made on newly prepared samples 48 h later.(2) A peak=fitted peak intensity of analyte (cps), T=count time (s) single measurement from each of 10 repeat sample preparations and conc.=concentration of analyte. was made for a waste sample spiked with 54.3 mg kg-1 Cd. A single Cl analysis of the waste sample measured on an Blank, Accuracy and Matrix Spike Results Oxford XR400 EDXRF spectrometer, at a separate site and by a second operaror, using the alumina sample preparation Errors shown in the following tables were taken from the technique, is shown in the last column of Table 14.results output of the instrument and are nominally ±2s.19 These represent the total error attributed to spectrum processing and counting statistics. Details of the calculations used DISCUSSION in the error calculations were given by Statham.19 The recovery results show that all but three, i.e., 93%, of the results from the toxic elements calibrations lie within ±15% Blank relative of the spiked value; 81% are within 10% relative and after correcting for possible error based on the 2s errors Table 6 gives the results of the measurement of the PEG shown, a further 12% lie within the ±15% range.Three blank sample. measurements lie outside this range. These anomalous values are explained as follows: first, the determination of Tl in the turbid waste gave 43% recovery; this anomaly was due to spectral interference from a high Br content in this sample of Table 5 Calibration results (measurement times according to Tables 3 and 4) 0.58%; and second, the absence of Pb in the single-phase solvent and low Cu recovery in the biphasal solution are Element Units LLD/mg kg-1 Std.error conc. attributed to errors in spike preparation. Repeat analysis of P % n/a* 0.132 these samples was not possible owing to a lack of sample. S % n/a 0.056 All of the results from the light elements and halogens Cl % n/a 0.066 calibration lie within ±15% relative of the spiked values and Br mg kg-1 5 147 most are within ±10% relative.I mgkg-1 7 35 During the running of these experiments a number of V mgkg-1 8 6.1 possible areas of bias were observed. These effects are Cr mg kg-1 8 9.6 Fe mg kg-1 14 7.9 summarized as follows: Ni mg kg-1 16 8.5 (i) Under certain conditions the Prolene film used in the Cu mg kg-1 17 7.7 determination of the light elements and halogens was seen to Zn mg kg-1 11 5.6 relax and become crinkled.This was found to be associated As mg kg-1 5 8.2 with samples containing high concentrations of certain chlori- Se mg kg-1 5 5.3 nated compounds. The exact nature and type of compound Cd mg kg-1 3 1.9 Sn mg kg-1 6 2.5 was not known and was outside the requirements of this study. Sb mg kg-1 5 4.3 This film relaxation was associated with halogens migrating Hg mg kg-1 7 7.5 through the film, often resulting in spurious results if the Tl mg kg-1 4 13.1 sample was not measured immediately. For this reason, it was Pb mg kg-1 4 5.4 recommended that samples are measured within 30 min of preparation.This was typically the way the technique is used * n/a: These elements were calibrated at concentrations significantly higher than their respective detection limits. at the incinerator sites where the alumina method has been Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 29Table 6 Blank measurement results V Cr Fe Ni Cu Zn As Se Cd Sn Sb I Hg Tl Pb mg kg-1 1.2 0.0 17.3 0.0 0.0 1.6 1.9 2.6 0.0 1.3 0.0 3.9 1.9 2.2 0.0 mg kg-1 error 0.5 0.7 0.2 3.6 0.04 0.4 7.6 2.5 2.5 0.4 0.9 10 2.9 3.9 1.0 Table 7 Accuracy measurement results P S Cl Se/ As/ Br/ Cd/ Sn/ Sb/ I/ (% m/m) (% m/m) (% m/m) mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 Calc.conc. 0.98 2.09 0.53 93 393 4779 197 47 201 587 Error 0.02 0.02 0.005 4 10 15 4 3 7 13 Given conc. 1.024 2.095 0.598 100 400 4838 200 50 200 600 Accuracy (%) 96 100 89 93 98 99 98 94 100 98 Hg/ Tl/ Pb/ V/ Cr/ Fe/ Ni/ Cu/ Zn/ mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 mg kg-1 Calc. conc. 99 96 94 109 6 589 53 55 95 Error 9 13 7 7 3 0.2 8 3 5 Given conc. 100 100 100 100 10 600 50 50 100 Accuracy (%) 99 96 94 109 60 98 106 110 95 Table 8 Toxic elements spiking results: biphasal waste V Cr Fe Ni Cu Zn Se MS/MSD recovery (%) 96 111 84 86 73 105 109 Error/mg kg-1 6 5 1 9 4 5 4 Spike conc./mg kg-1 97.4 98.3 97.1 98.2 99.3 102.0 104.6 As Cd Sn Sb Hg Tl Pb MS/MSD recovery (%) 109 108 103 84 103 98 99 Error/mg kg-1 9 3 3 6 9 10 7 Spike conc./mg kg-1 109.7 47.1 53.5 53.3 101.8 102.5 183.7 Table 9 Toxic elements spiking results: single-phase waste V Cr Fe Ni Cu Zn Se MS/MSD recovery (%) 110 114 109 104 105 90 108 Error/mg kg-1 6 6 2 10 5 5 4 Spike conc./mg kg-1 116.8 109.5 109 105.7 109.9 105.3 103.8 As Cd Sn Sb Hg Tl Pb MS/MSD recovery (%) 108 100 95 112 98 77 Error Error/mg kg-1 9 3 3 6 9 10 Spike conc./mg kg-1 105 45.5 55.5 54.6 97.0 96 Table 10 Toxic elements spiking results: turbid waste V Cr Fe Ni Cu Zn Se MS/MSD recovery (%) 102 110 104 118 94 105 90 Error/mg kg-1 6 6 2 11 5 5 5 Spike conc./mg kg-1 103.4 102 106.4 104.5 103.5 106.6 93.6 As Cd Sn Sb Hg Tl Pb MS/MSD recovery (%) 121 92 103 108 114 43 108 Error/mg kg-1 10 3 3 6 11 13 10 Spike conc./mg kg-1 94.8 44.8 53.2 53.2 96.8 97.7 177 Table 11 Light elements and halogens spiking results: biphasal waste hence the effects of any matrix interaction at the Prolene film are minimal.The toxic elements calibration uses Hostaphan Cl film and is not affected in the same way. MS/MSD recovery (%) 98 (ii) High concentrations of Br can affect the performance of Error (% m/m) 0.008 Tl owing to the fitting of the Tl peak in the spectrum. The Spike conc. (% m/m) 1.654 performance for Tl will depend on detector resolution and accuracy of peak-fitting routines used to determine the countrate of individual elements in a spectrum.The type of sample, introduced. Heavy elements are not affected by this problem i.e., the major heavy elements present, will govern to a large as the bias appears to be due to loss and/or matrix interactions extent the accuracy of any determination of trace heavy toxic of light element analytes at the Prolene/alumina interface. The elements. This is not considered a limitation as the requirement of the method is to provide data upon which risk assessment determination of Br and I in the alumina is from the bulk, 30 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12Table 12 Light elements and halogens spiking results: single-phase waste P S Cl Br I MS/MSD recovery (%) 111.3 120 110 95 100 Error (% m/m) 0.02 0.01 0.007 0.0007 0.0032 Spike conc. (% m/m) 0.9156 0.926 1.279 0.0599 0.856 Table 13 Light elements and halogens spiking results: turbid waste P S Cl Br I MS/MSD recovery (%) 98 91 89 104 100 Error (% m/m) 0.01 0.009 0.005 0.0006 0.0034 Spike conc.(% m/m) 0.849 0.805 0.858 0.042 0.845 Table 14 Sample preparation reproducibility results 10 repeats: day 1 10 repeats: day 3 Second laboratory Mean s* RSD (%) Mean s* RSD (%) (single result) Cl (% m/m) 1.45 0.027 1.9 1.43 0.037 2.6 1.38 Cd/mg kg-1 56 1.3 2.2 * s=1s standard deviation. in terms of process control and environmental regulations can CONCLUSIONS be monitored. High concentrations and/or element(s) close to, The alumina matrix modification technique provides a reliable, or exceeding, decision threshold concentrations with respect fast and robust method of sample preparation prior to analysis to their control or regulated limits will govern the disposal by high-resolution EDXRF spectrometry. The typical levels of route of any waste.In the presence of elements with high accuracy achieved for both major acid-producing elements and concentrations, in terms of decision thresholds, the inability environmentally sensitive toxic trace elements were within 15% to determine accurately trace elements such as Tl, i.e., relative using spiked real waste samples.The precision of total <100 mg kg-1, will not affect, in most instances, the disposal measurement was better than 2.5%. route or whether to accept or reject a consignment. The wide elemental range, low power and non-destructive Perhaps the most useful spectroscopic feature of the alumina nature of EDXRF spectrometry strongly support its being the matrix technique is the control of spectrum background.Fig. 1 ideal instrument for LHW screening. The technique outlined shows two spectra, one a blank and the other containing in this paper shows that it is ‘fit-for-purpose,’ by exhibiting 1.25% Br. The backgrounds of these two spectra are almost traceable analysis, relative accuracy better than 20% and rapid identical. The effect of the variations in sample matrix can be analysis (<15 min) for the determination of elements in all considered to be negligible in the alumina matrix.In order to LHW. maintain and control the background component of the sample The following matrix types are suitable for analysis using matrix, each element is restricted to the linear response range, high resolution EDXRF spectrometry with the activated alumi. e., <1000 mg kg-1 for trace elements. This is usually of the ina method of sample preparation: industrial waste solvents, order of 1–2% m/m but as suitable standards are often not sewage sludge, liquefied waste fuels, paints and inorganic available at these concentrations the range is restricted to pigments in liquid form, electroplating solutions and waste oils. 0–1000 mg kg-1 where pure aqueous solutions are readily A single calibration for the determination of light elements obtainable. A wider concentration range is permitted for low and halogens and a second for the determination of heavy atomic number elements as these do not significantly effect the trace elements can be used for most types of LHW samples.spectral background. The mechanism that allows any hazard- The simple alumina sample preparation technique allows for ous liquid waste to be determined using the alumina technique rapid analysis, with typically a total time, including sample is to dilute into the calibration range. preparation, of <15 min for halogens and light elements determination and an additional 15 min for the heavy trace elements.Based on sample preparation consumables and instrument overheads (helium, liquid nitrogen and power) the cost was estimated at £1.00 per analysis for a minimum of five halogens and light elements, £0.80 for a minimum of 16 heavy element traces and £1.35 for a minimum of 21 element analyses using both calibrations. The proposed alumina matrix modification technique is tailor-made for the determination of most elements encountered in LHW, unlike alternative instrumental techniques which must undergo some kind of aggressive (digestion) pretreatment.This absence of pre-treatment minimizes unpredictable analyte losses, producing a high degree of analytical confidence. The implementation of the alumina matrix modification technique can improve the reliability and accuracy of results Fig. 1 Spectra for a sample of real waste containing Br at 12.45% such that regulators and regulatory bodies could increase the (diluted 10-fold to bring the concentration into the calibration range) superimposed on a PEG blank sample.number of elements determined beyond the current minimum. Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12 316 Cresser, M. S., Armstrong, J., Cook, J. M., Dean, J. R., Watkins, The acceptance of the alumina technique coupled with high- P., and Cave, M., J. Anal. At. Spectrom., 1995, 10, 9R. resolution EDXRF spectrometry as an industry standard 7 McCrindle, R.I., and Rademeyer, C. J., J. Anal. At. Spectrom., would ensure that a near total elemental analysis of all LHW 1995, 10, 399. would become possible. Currently, a lack of appropriate avail- 8 Lui, Y. W., Li, D. L., Fan, Q. M., and Wei, C. L., Guangpuxue Yu able techniques and matrix difficulties restrict the type and Guangpu Fenxi, 1992, 12, 83. 9 Vanhoe, H., Goossens, J., Moens, L., and Dams, R., J. Anal. At. number of analyses carried out on any particular waste. The Spectrom., 1994, 9, 177. method has been extensively used at one UK incineration site 10 Applied Zeeman Graphite Furnace Atomic Absorption after the UK Environmental Agency (EA) approved its use for Spectrometry: Chemical L aboratory T oxicology, eds., Minoia, C., the determination of Br. and Caroli, S., Pergamon Press, Oxford, 1992, p. 79. 11 Tserovsky, E., Arpadjan, S., and Karadjava, I., J. Anal. At. Spectrom., 1993, 8, 85. Thanks are due to Jeanette Gravell and David James for the 12 Peraniemi, S., Vepsalainen, J., Mustalahti, H., and Ahlgren, M., preparation of samples and standards and to Andy Ellis for Fresenius’ J. Anal. Chem., 1992, 344, 118. his advice. 13 Meltsh, B., Muenzberg, I., and Janssen, A., L aborPraxis, 1995, 19 (4), 64 and 67. 14 West, H., Cawley, J., and Wills, R., Analyst, 1995, 120, 1267. 15 Lucke, N., Wehner, B., Thi Hong Lan, T., and Kalla, E., Acta REFERENCES Hydrochim. Hydrobiol., 1991, 19, 275. 16 Seiber, J. R., Adv. X-Ray Anal., 1993, 36, 155. 1 James, R., paper presented at Chemspec Europe 95 BACS 17 Bacon, J. R., Ellis, A. T., McMahon, A. W., Potts, P. J., and Symposium, 1995. Williams, J. G., J. Anal. At. Spectrom., 1994, 9, 267R. 2 Cresser, M. S., Garden, L. M., Armstrong, J., Dean, J. R.,Watkins, 18 ASTM Method D5839, 1997 Annual Book of ASTM Standards, P., and Cave, M., J. Anal. At. Spectrom., 1996, 11, 19R. ASTM, Philadelphia, PA, USA, 1997, vol. 11.04. 3 Marshall, J., Carroll, J., Crighton, J. S., and Barnard, C. L. R., 19 Statham, P., Anal. Chem., 1977, 49, 2149. J. Anal. At. Spectrom., 1993, 8, 337R. 4 Marshall, J., Carroll, J., Crighton, J. S., and Barnard, C. L. R., Paper 6/03605H J. Anal. At. Spectrom., 1994, 9, 319R. ReceivedMay 23, 1996 5 Marshall, J., Carroll, J., and Crighton, J. S., J. Anal. At. Spectrom., 1995, 10, 359R. Accepted October 9, 1996 32 Journal of Analytical Atomic Spectrometry, January 1997, Vol. 12