JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 273 Development of a Proposed International Standard for Determining Arsenic in Workplace Air Using Hydride Generation Atomic Absorption Spectrometry* Robert D. Foster and Alan M. Howe Occupational Medicine and Hygiene Laboratory Health and Safety Executive Broad Lane Sheffield UK S3 7HQ Proposed Draft International Standard DIS 11 041 described and developed by the International Organisation for Standardisation (ISO) involves sampling air by drawing it though a membrane filter with a paper back- up pad both of which have been treated with sodium carbonate solution. Particulate arsenic compounds and arsenic(ii1) oxide vapour are collected but not arsine. The filter and pad are digested using nitric acid sulfuric acid and hydrogen peroxide.The resultant solution is boiled down to sulfuric acid fumes and then made up to volume with water. Aliquots of the sample solution are prepared for hydride generation by the addition of hydrochloric acid and potassium iodide. Analysis by both continuous flow and flow injection hydride generation atomic absorption spectrometry (HG-AAS) is described. The method exhibits a recovery of >95%. It can be adapted to determine nanogram to milligram levels of collected arsenic making it suitable for testing compliance with the widely differing limit values stipulated in government health legislation of different countries e.g. 0.1 mg m-3 UK Maximum Exposure Limit (MEL) 0.01 mg m-3 US Occupational Safety and Health Administration (OSHA) Permissible Exposure Limit (PEL) and 0.002 mg mP3 US National Institute of Occupational Safety and Health (NIOSH) Short Term Limit Value.The International Standard will supplement existing internationally adopted analytical standards such as NIOSH 7901 which employs electrothermal atomic absorption spectrometry and NIOSH 7900 which employs HG-AAS but which for safety reasons requires the use of a fume cupboard suitable for handling perchloric acid. It should be useful to a wide range of laboratories including those with limited AAS facilities. This paper describes the testing protocol used to validate the method and gives results obtained to date including some obtained by laboratories using the analytical method in an inter-laboratory exercise. Keywords Arsenic and arsenic(ii1) oxide; workplace air sampling; hydride generation; atomic absorption spectrometry; method validation The International Organisation for Standardisation (ISO) is a federation of 92 national standards organizations.The British Standards Institute (BSI) is the member body for the UK. The I S 0 co-ordinates the exchange of information on international and national standards in all fields except those of electrical and electronic engineering liaising with some 450 international organizations in an attempt to encourage consistency of stan- dards. International Standards are developed by nearly 200 I S 0 Technical Committees that are responsible for specific technical areas. Work is carried out through 630 sub- committees. It is estimated that some 20000 engineers scien- tists and administrators participate in the drafting of standards either by technical testing of the procedures or by participation in working group meetings to represent the consolidated views and interests of industry government labour and individual consumers in the standards development process.The number of published International Standards is now approaching 10000 most of which are available in English French and Russian. The format of the Standards is carefully prescribed and wording is kept as simple as possible to ease the task of translation for those reading in other than their first language. In 1986 the Occupational Medicine and Hygiene Laboratory (OMHL) of the UK Health and Safety Executive (HSE) was approached by Sub-committee 2 (SC2) Workplace Atmospheres of Technical Committee 146 (TC146) Air Quality to provide a Convenor for Working Group 2 (WG2) Inorganic Particulate Matter.The OMHL was chosen because of its involvement in the organization of developing analytical methods published in the MDHS Series (Methods for the Determination of Hazardous Substances). These are distributed through HM Stationary Office and are widely used throughout the UK and other parts of the world to monitor the exposure * Presented at the XXVIII Colloquium Spectroscopicurn Internationale (CSI) York UK June 29-July 4 1994. of workers to air contaminants in the workplace. Shortly afterwards the Working Group resolved to produce an International Standard for determination of arsenic in work- place air using continuous flow (CF) or flow injection (FI) hydride generation atomic absorption spectrometry (HG- AAS).Much of the practical work required to develop the procedure was subsequently carried out at the OMHL. However the laboratories of similar organizations in other countries were also involved in evaluating the method notably Institut National de Recherche et de Securite (INRS) in France and Instituto Nacional de Seguridad Higiene en el Trabajo (INSHT) in Spain. Occupational exposure to inorganic compounds of arsenic can occur through inhalation ingestion or by absorption through the skin. Chronic exposure usually arises by inhalation and can produce respiratory tract effects notably perforation of the nasal septum skin effects such as aczematous and folicular dermatitis together sometimes with abdominal symp- toms.' The nervous system and liver can also be affected.Inorganic arsenic compounds are recognized as lung and skin carcinogens. Exposure of workers to arsenic can occur during the manufacture and use of arsenic compounds when arsenic oxides are produced during the smelting of metal ores and when metal alloys containing arsenic are heated strongly. Arsenic is alloyed with lead and copper to increase hardness and heat resistance in such processes as the production of lead shot and the manufacture of lead battery grids. Arsenic com- pounds are used in the pottery industry [arsenic(~~~)chloride]; in glass making [arsenic(m)oxide As,O,]; as pigments [cop- per(n)acetoarsenite Cu(COzCH3).3Cu(As02)z]; as insecti- cides (arsenic(m)oxide copper(1r)arsenite [Cu( As02),-H20] copper(I1)acetoarsenite and calcium arsenate [Ca,(AsO,),]}; as fungicides [copper(~~)arsenite]; as herbicides [sodium arsenite (NaAsO,) and calcium arsenate]; as wood preserv- atives (copper and chrome arsenites); in the production of anti-274 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 fouling paints for ships [copper(~~)acetoarsenite]; as a rodenti- cide [arsenic(m)oxide and copper(~i)arsenite]; in the elec- tronics industry in the production of semiconductors (galium arsenide GaAs); and alloyed with selenium to coat photocopier drums.Cancer of the respiratory tract has been reported in excess frequency in workers known to be exposed to dust and fumes containing arsenic whilst engaged in the smelting of metal ores and the production of insecticides. The task of drafting a sampling and analytical method commenced with a literature search.Manufacturers of HG-AAS systems were also approached for information on available instrumentation. A number of methods exist for the measurement of inorganic compounds of arsenic in workplace air. Sampling procedures include those which draw air through untreated membrane filters*5 to collect particulate dust con- taining arsenic and those using sodium carbonate or sodium hydroxide treated membrane filters or without7 treated back-up pads) or treated quartz fibre filters’ in order to collect arsenic(I1I)oxide vapour in addition. These filters are subjected to a wet digestion and the resultant sample solutions are analysed using various instrumental techniques.These include electrothermal AAS5,6 or inductively coupled plasma atomic emission spectrometry ( ICP-AES)4.8 for the direct analysis of the sample solutions and discrete batch HG-AAS.3,7 There was a requirement for a method that employs the techniques of CF or FT-HG-AAS which are increasingly and now pre- dominantly used for the determination of arsenic in preference to discrete batch HG-AAS. There was also a requirement for an effective sample filter digestion procedure for use with HG that does not involve the use of perchloric acid and hence does not require a fume cupboard adapted to scrub out perchloric acid fumes which otherwise present a fire hazard. A working draft was prepared in which the sample digestion procedure was based on US National Institute of Occupational Safety and Health (NIOSH) 7901; with modifications to provide a final solution suitable for HG-AAS rather than electrothermal AAS.After scrutiny and approval by the mem- bers of the Working Group the method was approved as a Committee Draft and a detailed evaluation of sample digestion procedures commenced. Experimental Instrumentation The standard is written assuming that the instrumental tech- nique employed will be HG-AAS. However the method can be used with very little adaptation using HG with atomic fluorescence spectrometry (AFS) or ICP-AES. The two types of HG systems described in the proposed Draft International Standard (DIS) and their expected output are illustrated in Figs. 1-4. Continuous flow systems generate a constant atomic absorption signal and work by pumping a continuous stream of acid test solution and sodium tetrahydroborate solution to a mixing piece.Arsenic in the test solution is reduced to arsine gas which is swept by a stream of argon into a heated silica or quartz absorption cell mounted in the beam of an AA spectrometer. Flow injection systems use a switching valve to inject a discrete volume of test solution into an acid blank stream to produce a transient AA signal. The 197.2 nm arsenic line is used for AA measurements unless the highest sensitivity is required. The 193.7 nm arsenic line is approximately twice as sensitive as the 197.2 nm arsenic line but use of the 197.2 nm line is preferable since the calibration obtained at this wave- length has a greater linear range.The instrumentation used at OMHL in validating the method was a Perkin-Elmer 5100 AA spectrometer using an arsenic hollow cathode lamp without background correction. An electrically heated quartz gas cell was used in conjunction with a Perkin-Elmer FIAS 200 HG system. The system was deliberately de-sensitized by using smaller diameter peristaltic pump tubing than usual (1.14 mm tetr To heated silica or quartz cell mounted in AA wectrometer I F- Gaseous- liquid dm Two channel reactants peristaltic pump a Fig. 1 Schematic diagram of a continuous flow HG system / I I 0 22.50 45.00 Tim e/s Fig. 2 AAS output when valve is operated to change the continuous flow to the mixing piece from acid blank to acidified sample solution; the broken line indicates time at which measurement begins Heated silica or quartz cell ( a ) Autosampler 0000000 mounted in AA spectrometer ....... .. ........... ; ....... ... . . . . . . ........ ............ . . ..;,.:.. .. ; .. :. .. ..- k5i%iEiJ=$ Sample loop . . . . . . Pump 1 FI valve11 I Mixina Diece 1 (&Gas ressure regulator Pu (b) FI Valve functions Fill Injection Fig. 3 Schematic diagram of (a) the FT-HG system; and (b) FI valve functions 0 7.50 Time/s 15.00 Fig.4 AAS output when valve is operated to inject a fixed volume of acidified sample solution into the acid blank flowJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 275 id.) thereby reducing the flow rate of the acid blank stream to approximately 5 ml min-'. The flow rate of the reductant stream was also approximately 5 ml min-'.The system was used in FI mode with a sample loop volume of 200 pl and in simulated CF mode by extending the sample loop of the sample valve to 1500 pl. The latter over-large volume resulted in a stepped change in the AA spectrometer output rather than a peak. Reagents The proposed DIS specifies that reagents used should be of the highest purity consistent with minimum arsenic content. Solutions used in the method include solution for filter treat- ment (1 moll-' sodium carbonate in 5% v/v glycerol solution); blank acid used in HG and in the preparation of samples and standard solutions [dilute hydrochloric acid (1 + l)]; solution used in the preparation of sample and standard solutions for reducing As5+ to As3+ for a more rapid reduction to arsine in the hydride generator (100 g 1-' potassium iodide); solution for use in preparation of sample solutions [dilute sulfuric acid (1 + 9)]; reductant solution for HG [between 2 and 20 g 1-' of sodium tetrahydroborate in 0.1 mol 1-1 sodium hydroxide solution (the concentration used should be in accordance with the requirements of the HG system 2 g 1-' was used for work at the OMHL)].Sample Collection Procedure Particulate arsenic and arsenic compounds and arsenic(II1) oxide vapour are collected by drawing a known volume of air through a sodium carbonate impregnated cellulose ester mem- brane filter and a sodium carbonate impregnated back-up paper pad mounted in a sampler designed to collect the inhalable fraction of airborne particles.In most workplace situations where exposure to arsenic can occur (e.g. in the refining of base metals welding and other hot metal processes) a significant proportion of the arsenic is present in the form of arsenic(1rr)oxide vapour.' This vapour is collected by reac- tion with sodium carbonate on an impregnated cellulose ester membrane filter and impregnated back-up paper pad. As203 + Na2C03+2NaAs02 +C02 A collection efficiency of 1.00 has been reported for similarly prepared filter-pads for laboratory generated particulate arsenic aerosols' and 0.98 for arsenic(I1r)oxide vapour.' Arsenic in the form of arsine passes through the filters and is not collected. Metal arsenides that react with water vapour to yield arsine are similarly not estimated.Procedures The following procedures are modified extracts from the proposed DIS. Filter Preparation Place the cellulose ester membrane filters on a clean poly(tetra- fluoroethylene) (PTFE) sheet or similar inert flat surface in an arsenic-free environment. Establish the volume of sodium carbonate solution required to just wet the entire surface of a filter after it has been allowed to spread for a few minutes. Dispense this volume of sodium carbonate solution onto each filter and allow to dry at room temperature in an arsenic-free environment. Prepare sodium carbonate impregnated back-up pads in a similar fashion. Sample Digestion Place sample filters and back-up pads into beakers. If appro- priate wash dust from the inside of samplers into the beaker using a minimum volume of water.Add 5 ml of concentrated nitric acid and 1 ml of concentrated sulfuric acid to each beaker cover with a watch-glass and heat to approximately 175°C on the hot-plate in a fume cupboard. When the initial vigorous reaction has subsided slide back the watch glasses so that the beakers are only partially covered. Continue to heat each beaker until the solution volume has been reduced to approximately 1 ml and then remove from the hot-plate. Allow the solutions to cool and then carefully add 2ml of hydrogen peroxide to each beaker. Replace the beakers on the hot-plate again covering with the watch-glasses and when the initial vigorous reaction has subsided slide back the watch- glasses so that the beakers are only partially covered. Continue to heat until dense white fumes of sulfur trioxide are evolved (raise the temperature of the hot-plate to 200°C if necessary).If the solution becomes discoloured owing to charring of residual organic material carefully add hydrogen peroxide dropwise until a clear solution is obtained and then evaporate again until the appearance of dense white fumes. Remove the beakers from the hot-plate and allow the solutions to cool. Carefully rinse the watch glass and the sides of each beaker with a small volume of water transfer each solution quantitat- ively into a 10ml one-mark calibrated flask. Filter through a cellulose (paper) filter which has been pre-washed with dilute sulfuric acid and then with water if there is any evidence of undissolved particulate matter. Dilute to the mark with water.Preparation of Solutions for Hydride Generation Prepare blank and sample test solutions for analysis. Transfer an aliquot V ml of the sample digestion solution and (5- V,) ml of dilute sulfuric acid to a 25 ml calibrated flask. Add 12.5 ml of concentrated hydrochloric acid and 2.5 ml of potassium iodide solution and make up to volume with water. Allow 1 h for reduction of pentavalent arsenic to take place before analysis. National occupational exposure limits for arsenic vary considerably and therefore the volume of the aliquot of sample solution used can be varied up to a maximum of 5ml according to the detection limit required. For the lowest detection limit an aliquot volume of 5ml should be used; an aliquot volume of 1 ml is suggested when arsenic in air concentrations are to be compared with a limit value in the region of 0.01 mg m-3 of arsenic; and for limit values in the region of 0.1 mg m-3 of arsenic an aliquot volume of 0.1 ml is suggested.The final test solution matrix is 1 + 1 hydrochloric acid 1 +49 sulfuric acid and 10 g 1 -' potassium iodide. Calibration Prepare calibration solutions from commercially available arsenic standard solutions diluted in the test solution matrix to cover the range 0-50 ng ml-' of arsenic when the 197.2 nm arsenic line is used or 0-25 ngml-' of arsenic when the 193.7 nm arsenic line is used. The upper limit of the working range is dependent upon the performance characteristics of the HG system used and other instrumental factors that affect sensitivity and the linearity of the calibration.In general it is best to work in the linear range of an AA calibration where absorbance is proportional to the concentration of arsenic in solution. A certain amount of curvature can be tolerated. Hydride generation AAS calibrations are more curved than flame AAS calibrations and discretion should be exercised in assessing whether re-calibration over a lower concentration range is necessary. Analysis Set up and operate the HG and AA spectrometer in accordance with the manufacturer's instructions.276 + Vote acceptance IS0 working group 2 -* JOURNAL OF ANALYTrCAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 1 Propose modifications and revise IS0 committee draft i Validation of the Standard The stages in the validation of the method described briefly below are illustrated in the flow diagram given as Fig.5. Test analytical method Filter digestion H y d r ide genera t i o n/AA Testing of the initially proposed method The digestion procedure described in the first Committee Draft was based closely on that in NIOSH 79016 and did not involve sulfuric acid the sample solution was taken to dryness before being re-dissolved in hydrochloric acid. The method was tested both within the OMHL and in the laboratories of similar organizations. Although the procedure could be made to work with care it was found that any lack of thoroughness in boiling off residues of nitric acid led to the formation of iodine from the potassium iodide added to reduce arsenic to the trivalent state prior to HG. The method specified in the first draft used treated mem- brane filters without treated back-up paper pads.A collection efficiency of 94% for arsenic(r1I)oxide in air was reported by Costello et aL9 for membrane filters which were prepared in a similar fashion and this was considered by the Working Group to be adequate for air monitoring purposes. However investigations at INRS indicated that for high levels of arsenic(rr1)oxide in air the use of a treated back-up paper pad with the treated membrane was necessary if an unacceptable level of breakthrough was not to occur. The sampling and digestion methods were therefore modified and incorporated in an improved version of the draft. 4 Method satisfactory? Testing of the mod$ed analytical procedure After preliminary proving at OMHL the method was evalu- ated fully.Analytical recovery was determined using treated Validate method Analytical ruggedness recovery precision etc. Test method for on-site sampling 25 mm filter-pads dosed with arsenic solutions to simulate sampling an atmosphere containing arsenic at the UK 'Maximum Exposure Level (MEL) of 0.1 mg m-3 for 4 h and at 0.1 0.5 2.0 and 5.0 times this limit. The same exercise was :repeated for 37mm filters dosed to simulate the US Occupational Safety and Health Administration Permissible Exposure Limit (PEL) of 0.01 mg m-3. Recovery and precision of measurement at the different dosing levels were determined with the sample solution aliquot volume optimized for measurement of arsenic at the limit value. Measurements were made using both CF-HG and FI-HG techniques.Ruggedness testing was carried out to investigate the degree to which deviations could be made from the recommended procedures and conditions (volumes of reagent hot-plate temperature rates of evaporation of acids etc.) before the method failed to perform satisfactorily. The extent of transition metal interference was also briefly investigated. A number of transition metals mainly those of Groups 8 and 11 cause signal depression in the determination of arsenic by HG-AAS. This interference has been shown to be associated with the reduction of metal ions to the free The most severe interferences are caused by nickel copper and cobalt but with the reagent concentrations used in this method the signal depression caused by 10 pg ml-I of these three metals is less than lo% for a solution containing 10 ng ml- of arsenic.12 Flow injection systems are much less affected by metal interferences than CF systems.r Generate test samples Conduct inter- laboratory exercise Interlaboratory Exercise Laboratories were recruited through the OMHL Workplace Analysis Scheme for Proficiency (WASP) quality assurance scheme from laboratories listed by the National Measurement Process results and statistics Literature search Devise sampling and analytical method for As in workplace air Prepare IS0 committee draft Fig. 5 Preparation of proposed IS0 Standard 11041 Arsenic in Workplace AirJOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL. 9 277 Accreditation Service (NAMAS) as accredited for occupational hygiene air monitoring and from the customers of instrument manufacturers.The exercise was carried out broadly in accord- ance with I S 0 IS 5725.13 The exercise was not restricted to laboratories using the instrumentation specified in the DIS i.e. CF- or FI-HG-AAS. Some laboratories were asked to adapt the procedures with the minimum modification using HG-AFS HG-ICP-AES or discrete batch (DISCRT) HG-AAS. All laboratories were asked to complete a question- naire to obtain details of their instrumentation and comments on the comprehensibility and ease of use of the DIS. Test samples included filters dosed with known levels of arsenic and those generated from an artificially generated atmosphere of particulate arsenic(II1)oxide. Blank filters were also supplied.Laboratories were asked to analyse the sample solution obtained for each test filter three times and to report blank corrected replicate results. A mean result for each test sample was obtained from the replicate results and from this the percentage of the expected result was calculated. In order that statistics for the accuracy of the method should not be affected by preferential selection of results participants were cautioned that although the levels of the arsenic on the different batches of filters might appear similar they should not assume that they were the same. However in fact four identical samples dosed with arsenic at ‘US’ and ‘UK’ levels were provided to each laboratory. Dosed membrane3lter-paper pads (sets ‘UK’ and ‘US’) The test samples were produced from treated membrane filter-paper pads which were prepared as a number of batches identified by letters.They were dosed with arsenic at levels that simulated sampling at the UK MEL (approximately 50 pg on 25 mm membrane pads) and US PEL (approximately 5 pg on 37mm membrane pads). Two separate mailings of dosed filters at each level were prepared and sent out together with blank filters from the same batches. As far as possible the concentration dosed on the filters of all batches was the same. The volume of arsenic solution spotted onto the filters was measured carefully compensating for small evaporation losses by using the procedure given in BS7653 and determined to a 2 x RSD (relative standard deviation) error of < 1%. The ‘expected’ result was based on the certificated concentration of the standard used to prepare the dosing solution and the accurately determined volume dispensed by the micropipette used for the dosing.No significant difference between the levels spotted onto different batches of filters was noted for either the US batches or the UK sets of treated filter-pads. Inspection of the results obtained by ourselves and from the other laboratories in the exercise justifies the use of these values. The level of arsenic in the blank filters was demonstrated to be very low (<0.02 pg) for all batches. Test air samples from an atmosphere of arsenic(m)oxide (set Test air samples were prepared using cassette-type air samplers containing 25 mm treated membrane filters and paper pads at the laboratory of the National Institute of Occupational Health in Oslo Norway under the direction of Dr.Y. Thomassen. An atmosphere of arsenic(1n)oxide was generated and the air was drawn through 80 samplers simultaneously. It was impossible to control exactly the level of arsenic in the air but the collection of 5-10 pg in each sampler was aimed at. Air drawn through individual samplers was restricted by a critical orifice and the sampling rate was measured carefully. The amount of arsenic in individual samplers should vary with the measured flow rate. A number of the samplers were analysed at OMHL to establish that this was the case and to obtain a factor by which the flow rates through individual samplers could be multiplied in order to calculate ‘expected’ results for arsenic “N W”) collected.Blank filter cassettes were also provided. Several were analysed at OMHL and in all cases the blank was <0.02 pg of arsenic. Field Sampling During the final stages of validation it is intended to test the analytical procedures on samples that have been collected during on-site sampling in workplaces where arsenic is present in the air. Results and Discussion The detailed results of the method validation will be presented in a back-up report.14 The following is a summary of the conclusions so far. Method Performance Bias and recovery Laboratory experiments indicate that the method does not exhibit significant bias. The mean recovery for doped filters in the range 4.8-96 pg of arsenic has been determined to be 100.7% for a sample solution aliquot of 0.1 ml and 98.8% for a sample solution aliquot of 1 ml using CF-HG-AAS; and 99.3% for a sample solution aliquot of 0.1 ml and 102.7% for a sample solution aliquot of 1 ml using FI-HG-AAS.Repeatability The component of the coefficient of variation of the method that arises from analytical variability CV(analysis) is depen- dent upon a number of factors including the volume of the sample solution aliquot used in preparation of the test solution and whether CF- or FI-HG-AAS is used. The CV(ana1ysis) is at a minimum when the concentration of arsenic in the test solution is in the mid-range of the calibration and in laboratory experiments it has been estimated to be about 1% using CF-HG-AAS and about 3% using FI-HG-AAS for measure- ments made at 197.2 nm on test solutions with an arsenic concentration in the range 10-40 ng ml-’.This gives a measure of the repeatability of the method. (The repeatability of an AA method15 at a given level is the closeness of agreement between successive results obtained using the same method on identical material submitted for the test under the same conditions i.e. same operator same equipment same set of reagents same laboratory.) The over-all uncertainty of the method as defined by the Comite European de Normalization ( CEN),I6 has been shown to be within the specification of 30% prescribed by CEN7 for the over-all uncertainty of measurements for comparison with limit values assuming that the coefficient of variation of the method that arises from inter-specimen sampler variability CV(inter) is negligible and that the coefficient of variation of the method that arises from pump flow rate variability CV(flow) is limited to 5%.Detection limit and working range Detection limits for the determination of arsenic are dependent upon the analytical line at which absorbance measurements are made and upon the HG system and AA spectrometer used. However the qualitative and quantitative instrumental detec- tion limits for arsenic defined as three times and ten times the standard deviation of a blank determination have been esti- mated to be approximately 0.3 and 1 ng ml-’ respectively for the 197.2 nm arsenic line.14 For an air sample volume of 960 1 and a sample solution aliquot of 5 ml this corresponds to arsenic in air concentrations of 0.015 and 0.05 pg m-3 respect- ively.The working range of the method is approximately 100 ng to 125 pg of arsenic per sample for absorbance measure-Table 1 Results (%) for IS0 interlaboratory exercise on determining arsenic ‘UK dosed treated 25 mm test filter-pads. Expected result 50.54 pg of As. The significance of the statistics given is explained in the text Laboratory identification number h 4 00 36 CF AAS 3 CF AAS 4 CF AAS 14 CF AAS 18 24 CF CF AAS AAS 26 30 35 37 20 22 1 15 21 29 7 16 CF CF CF FI FI FI DISCRT DISCRT CF CF FI CF AAS AAS AAS AAS AAS AAS AAS AAS AFS AFS ICP ICP 8 CF AAS Hydride generator Instrument Mean result obtained from 3 analyses expressed as a percentage of expected result Filter 1 (%) Filter 2 (YO) Filter 3 (%) Filter 4 (O//.) Mean result for batch of filters (YO) Mean filter blank/pg of As CV of mean results for filters (%) CV of single replicate analysis (Yo) Average CV of three analyses (%) 98.02 100.75 99.60 99.3 1 99.42 0.01 1.13 1.14 1.28 53.59* 106.54 108.36 108.99 107.96 1.9 1.18 1.23 3.19 98.60 101.57 104.37 102.23 101.69 0.50 2.34 2.52 3.32 112.45 105.53 102.89 104.21 106.27 0 4.01 5.14 5.27 106.57 100.91 103.69 104.26 103.86 0.28 2.24 2.32 0.80 102.29 104.68 105.41 106.83 101.81 110.98 102.30 110.08 102.95 108.14 0 0.04 1.61 2.70 1.87 2.88 1.03 2.24 98.48 99.20 102.05 96.87 99.15 6.36 2.18 3.55 2.95 99.86 101.15 99.23 98.58 99.70 0.33 1.10 1.38 0.94 100.13 100.67 100.13 96.65 99.40 1.08 1.86 1.96 2.63 98.08 99.29 100.63 99.32 99.33 0.0 1 1.05 2.20 2.04 71.14 75.75 75.03 70.74 73.16 3.54 3.73 1.35 - 0.4 17.23 18.72 16.57 16.41 17.23 0.043 6.10 6.29 2.24 102.05 33.95* 93.35 121.00 105.47 -0.55 13.41 15.78 9.49 - 94.90 102.30 93.25 94.52 93.25 100.63 80.13 99.15 93.80 0 0 4.13 1.02 1.46 0.88 1.34 - 97.41 100.35 100.38 101.80 99.99 0.45 1.84 1.92 0.85 79.49 72.27 87.21 87.90 81.72 9.01 9.02 0.92 - 92.80 92.81 93.95 92.44 93.00 0.57 0.71 1.43 2.33 * Figure not included in statistical analysis.Table 2 Results for the IS0 interlaboratory exercise for determining arsenic ‘US dosed treated 37 mm test filter-pads. Expected result 5.02 pg of As. The significance of the statistics given is explained in the text Laboratory identification number 36 CF AAS 3 CF AAS 4 CF AAS 8 CF AAS 14 CF AAS 18 CF AAS 24 CF AAS ~~ 26 CF AAS _ _ _ _ _ _ ~ ~ ~ 1 15 DISCRT DISCRT AAS AAS 21 CF AFS 29 CF AFS 7 FI ICP 16 CF ICP 30 CF AAS 35 CF AAS 37 FI AAS 20 FI AAS 22 FI AAS Hydride generator Instrument Mean result obtained from 3 analyses expressed as a percentage of expected result Filter 1 (YO) Filter 2 (YO) Filter 3 (YO) Filter 4 (%) Mean result for batch of filters (Yo) Mean filter blank/pg of As CV of mean results for filters (YO) CV of single replicate analysis (%) Average CV of three analyses (%) 99.60 101.09 100.49 99.63 100.20 0 0.72 0.94 0.73 107.85 110.65 104.46 106.92 107.47 0.22 2.38 2.90 7.13 104.06 106.39 105.52 105.26 105.31 0 0.91 1.64 1.65 104.52 105.99 102.00 103.13 103.91 0 1.66 3.14 3.27 97.54 96.61 96.41 95.87 96.61 0 0.72 0.84 0.72 99.36 101.49 98.43 11 1.97 102.81 0 6.07 6.10 1.27 104.32 104.19 101.53 104.79 103.71 0.13 1.42 1.56 1.78 47.27 95.08 102.99 101.33 99.17 0.63 3.67 6.65 5.19 i02.00 102.13 101.26 100.53 101.48 0 0.73 1.78 1.65 98.21 99.54 98.80 96.61 98.29 0.11 1.26 1.73 1.25 95.88 101.86 99.73 98.87 99.09 -0.01 2.50 3.40 2.82 T i 71 l I .l L . 73.32 72.32 73.99 72.84 - 0.05 1.39 1.63 0.88 135.26 102.46 49.10* 85.96 97.89 0.04 10.65 10.69 1.64 73.78 72.95 123.79 74.87 87.85 27.47 29.46 12.74 - 1.24 93.51 97.01 92.91 96.91 95.08 0 2.29 1.75 - 101.w 98.74 102.73 100.00 100.62 0 1.67 1.74 0.51 86.96 91.95 86.49 91.15 89.14 0.56 3.15 3.24 1.32 93-08 95.48 88.16 90.75 91.87 0.02 3.41 3.54 2.14 110.58 117.55 107.46 112.57 112.04 -0.12 3.78 4.65 5.70 c1 F % M * Figure not included in statistical analysis. Table 3 Results of IS0 interlaboratory exercise on determining arsenic ‘NW test samplers.Arsenic oxide sampled onto treated 25 mm membrane-pads. The significance of the statistics given is explained in the text Laboratory identification number 36 CF AAS 3 CF AAS 4 CF AAS 8 CF AAS 14 CF AAS 18 CF AAS 24 CF AAS 26 CF AAS 30 CF AAS ~ 37 FI AAS 20 FI AAS 22 FI AAS 1 15 21 DISCRT DISCRT CF AAS AAS AFS 29 CF AFS 7 FI ICP 16 CF ICP 35 CF AAS Hydride generator Instrument Sampler No. 1 - Expected value/pg of As Mean result of 3 analyses/pg of As % of expected result CV of replicate analyses (Yo) Expected value/pg of As Mean result of 3 analyses/pg of As % of expected result CV of replicate analyses (%) Average of expected (YO) Mean filter blank (pg) Sampler No. 2- 6.86 7.24 105.55 0.52 7.11 6.85 96.3 1 13.95 7.1 1 7.3 1 102.85 0.72 7.04 6.90 98.10 1.09 6.93 7.54 108.73 0.33 7.00 5.17 73.88 1.91 6.90 6.61 95.90 1.77 7.18 6.79 94.64 0.47 7.04 7.36 104.59 1.73 6.86 7.03 102.46 0.49 7.07 3.17 44.87 8.84 6.93 7.07 7.21 4.41 5.57 6.79 63.62 78.76 94.14 17.40 2.79 1.11 6.83 5.50 80.62 2.00 7.1 1 4.76 67.02 1.70 7.42 7.75 104.34 3.37 7.11 4.28 60.22 9.41 L a Q P 6.93 7.14 102.97 0.16 104.26 0 7.04 7.48 106.25 0.78 101.28 0.11 7.00 4.83 68.93 2.50 67.98 - 7.04 7.3 1 103.93 0.82 103.39 0 7.11 7.18 100.97 1.19 104.85 6.36 7.07 7.18 101.48 0.99 101.48 0 7.07 5.95 84.13 1.18 79.01 0.13 7.04 6.75 95.92 3.06 95.91 0 7.11 6.70 94.27 1.22 94.46 0 7.11 7.28 102.43 1.98 103.51 0.11 7.28 5.35 73.50 10.14 66.86 - 0.05 6.86 - 7.25 3.73 - 6.84 54.32 - 94.32 9.14 - 0.89 58.97 78.76 94.23 -1.24 0 0 7.1 1 5.75 80.85 0.27 80.74 0.08 7.04 7.66 108.90 4.54 106.62 -0.12 7.18 6.98 97.17 1.52 97.64 0 6.93 6.94 100.12 2.65 101.29 0 7.04 5.96 84.65 0.42 64.76 -0.14 c 0 r QJOURNAL O F ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 279 ments made at the 197.2nm arsenic line on test solutions prepared using sample solution aliquots in the range 5-0.1 ml. Interlaboratory Exercise Results The exercise is not at present complete but preliminary returns are encouraging. The results obtained for the analysis of the test filters by 17 laboratories which had provided results and comments on the procedure at the time of the CSI in York in July 1993 are given in Tables 1-3. A further one laboratory found itself unable to carry out the procedure using its equipment. To maintain confidentiality results are tabulated against laboratory identification numbers notified to the laboratories concerned.Results obtained by OMHL using FI-HG-AAS are given against laboratory 36 and using CF-HG-AAS are given against laboratory 37. Three different coefficients of variation (CVs) which have been calculated for the results are given in the tables. 1. CV of mean results. This was calculated from the mean results obtained by triplicate analysis by the laboratory for the test filters of the set. This result gives the best measure of variation which is owing to sample preparation rather than instrumental variability. 2. CVof single replicate analysis. This is an estimate of the CV which would have been obtained for the set had the instrumental analysis been carried out once for each test filter solution rather than three times.3. Average CVfor replicates. For this CVs between replicate analyses of the same test filter sample solutions were calculated then averaged for all the filters of the set. This has been calculated to estimate the variability between determinations which is due to instrumentation rather than to the sample preparation procedure. It is a little premature to carry out a detailed over-all breakdown of statistics. More laboratory results are expected and the origin of some anomalous results needs to be investi- gated to establish whether they are genuine outliers or ‘rogue results’. In any case with so many different instruments involved the value of over-all statistics could be limited. However inspection of results on a laboratory by laboratory basis is fairly instructive.Results obtained using CF- and FI-HG-AAS as specified in the proposed DIS have generally been close to those expected based upon the amounts of arsenic dosed onto the filters and the levels calculated for arsenic(m) oxide collected in the filter cassettes. The only useful statistics are those obtained for CF-HG- AAS since ten of the 19 laboratories reported using this technique. These laboratories also tended to be those determin- ing arsenic routinely on a daily basis and consequently were more experienced than other participants. They consistently obtained results within 10% of the expected result for dosed test filters. The CV has been calculated for each level of dosing in order to obtain a measure of the reproducibility of the method.(The reproducibility of an AA method at a given level1’ is the closeness of agreement between individual results obtained using the same method on identical material submit- ted for the test but under different conditions i.e. different operators different equipment different laboratories different times.) For measurements made at 197.2nm using CF-HG- AAS the reproducibility determined from all mean filter results was found to be 3.7% for 37 mm filter-pads dosed with 5 pg of arsenic and 3.4% for 25 mm filter-pads dosed with 50 pg of arsenic. The analytical recoveries obtained for 5 and 50 mg were 101.9 and 102.9% respectively. Results obtained for samplers were expected to be less consistent since it was anticipated that there would be some variability in the collec- tion of arsenic(II1)oxide in the samplers and because of possible failure to recover all arsenic from the internal surfaces of the samplers.In practice however results were little different to those obtained for dosed filters. A reproducibility of 4.3% and a mean recovery of 100.7% was determined from the results obtained for all samplers analysed using CF-HG-AAS (except those from laboratory 24). It is interesting to note that the method was capable of adaptation by laboratories using the allied techniques of HG-AFS HG-ICP-AES and discrete batch HG-AAS but the numbers of sets of results were too small to yield statistical analysis (only 2-3 for each analytical tech- nique). It is evident that data obtained using these techniques feature more anomalous results than those for CF-HG-AAS.This probably reflects the fact that these laboratories were not in general involved in the day-to-day routine determination of arsenic and were commonly operating their instrumentation in an unfamiliar way. Even so it is clear that it is possible to obtain the expected result for all of these techniques. Overall a wide range of instruments was used by participants in the interlaboratory exercise all of which seem to perform well with the method including systems manufactured by Perkin-Elmer Philips/Unicam PS Analytical and Varian. Close to 100% of the expected values can be obtained using CF- FI- or discrete batch HG-AAS or HG-ICP-AES or HG-AFS. Coefficients of variation for the mean results obtained for a set of filters by a laboratory depend upon the technique used but are commonly in the range of 1-4% whether determined for single replicate analysis or triplicate analysis. The CVs between replicate determination of the same sample solution again depend upon technique but are typically less than 3%.Feedback from participants on the use of the method has been instructive. Most analysts reported finding the method easy to understand and to carry out. Some minor changes to the wording and procedures of standard are indicated however. For instance it is clear that the use of a zero standard in the calibration must be stressed as must the need to include the same concentration of acids in the standards as is present in the sample solutions.Sulfuric acid concentration is particularly influential since although only 2% v/v is present in the sample and standard solutions this causes the AA output signal/arsenic concentration to be significantly lowered. Conclusions Conclusions about the effectiveness of proposed DIS from the investigations carried out so far indicate that fairly small changes need to be made to the current draft of the I S 0 standard before it is in a state to merit publication. Postscript Following the preparation of this paper the drafts of the MDHS and I S 0 standards have been modified slightly so as to no longer require that membrane filters used to collect fume should be treated with sodium carbonate solution. The change has been made because sodium carbonate-treated cellulose ester membrane filters were found to become friable after several weeks storage.Untreated 0.8 pm pore size cellulose ester membrane filters with sodium carbonate-treated paper back-up pads were demonstrated to be effective in collecting > 98% of arsenic when sampling workplace air containing mixed particulate and arsenic(rr1)oxide vapour fume. l4 References 1 Health and Safety Executive Toxicity Review 16 Inorganic Arsenic Compounds HMSO London 1986. 2 World Health Organisation (WHO) or International Agency For Research On Cancer (IARC) JARC Monographs on the Eualuation of the Carcinogenic Risk of Chemicals to Humans Volume 23- Some Metals and Metallic Compounds IARC Lyons 1980. 3 US National Institute for Occupational Safety and Health NIOSH Manual of Analytical Methods Method 7900-Arsenic and Compounds (Revision I 1987) US Government Printing280 JOURNAL OF ANALYTICAL ATOMIC SPECTROMETRY MARCH 1994 VOL.9 4 5 6 7 8 9 10 11 Office Washington DC 3rd. edn. 1984 DHHS Publication US National Institute for Occupational Safety and Health NIOSH Manual of Analytical Methods Method 7300-Elements (ICP) US Government Printing Office Washington DC 3rd. edn. 1984 DHHS Publication No. 84-100. US Occupational Safety and Health Administration OSHA Analytical Methods Manual Method I D 105 Inorganic Arsenic in Workplace Air USDOL/OSHA Salt Lake City 2nd. edn. 1991. US National Institute for Occupational Safety and Health NIOSH Manual of Analytical Methods Method 7901 -Arsenic Trioxide US Government Printing Office Washington DC 3rd. edn. 1984 DHHS Publication No. 84-100. Health and Safety Executive MDHS 41 Arsenic and Inorganic Compounds of Arsenic in Air HMSO London 1989. Demange M. Vien I. Hecht G. and Hery M. Mise au point d’une method de prelevement du trioxyde de diarsenic (Development of a method for sampling arsenic trioxide) Institut National de Recherche et de Securite Vandoeuvre 1992 ND 1872-146-92. Costello R. J. Eller P. M. and Delon Hull R. Am. Ind. Hyg. Assoc. J. 1983 44 21. Bax D. Agterdenbos J. Worrell E. and Beneken Kolmer J. Spectrochim. Acta Part B 1988 43 1349. Welz B. and Schubert-Jacobs M. J. Anal. At. Spectrom. 1986 1 23. NO. 84-100. 112 Anderson R. K. Thompson M. and Culbard E. Analyst 1986 111 1143. I3 International Organisation for Standardisation (TSO) International Standard 5725 Precision of Test Methods- Determination of Repeatability and Reproducibility for a Standard Test Method by Inter-laboratory Tests ISO Geneva 2nd. edn. 1986. 14 Foster R. D. and Howe A. M. Development and Validation of a Method for the Determination of Arsenic and Inorganic Compounds of Arsenic in Air (Excluding Arsine) Using Hydride Generation Atomic Absorption Spectrometry Analysis OMHL HSE Sheffield 1993 IR/L/IS/93/05. 15 International Organisation for Standardisation (ISO) International Standard 6955 Precision of Test Methods- Determination of Repeatability and Reproducibility for a Standard Test Method by Inter-laboratory Tests ISO Geneva 1982. 16 ComitC Europten de Normalization (CEN) Workplace Atmospheres-General Requirements for the Performance of Procedures for the Measurement of Chemical Agents CEN Brussels 1992 prEN 482. Paper 3/05445D Received September 10 1993 Accepted October 15 1993