ANALYST, MAY 1991, VOL. 116 437 Certification of a Reference Material for Aromatic Hydrocarbons in Tenax Samplers Stefaan Vandendriessche and Bernardus Griepink Commission of the European Communities, Community Bureau of Reference (BCR), Brussels, Belgium Jacobus C. Th. Hollander, Johannes W. J. Gielen and Fred G. G. M. Langelaan TNO, Division of Technology for Society, Dele, The Netherlands Kevin J. Saunders BP Research Centre, Sunbury-on-Thames, Middlesex TW16 7LN, UK Richard H. Brown OMHL3, Health and Safety Executive, 403 Edgware Road, London NW2 6LN, UK A homogeneous and stable reference material consisting of aromatic hydrocarbons sorbed on Tenax in stainless-steel sample tubes has been prepared and certified. An initial feasibility trial established that a homogeneous and stable batch could be prepared.Part of this batch was used in an intercomparison, which allowed the identification of various sources of error. A second test batch was used in a second intercomparison in an attempt to improve the analytical performance of the participating laboratories. A third batch (of 1000 tubes) was then prepared and certified on the basis of analyses carried out in ten laboratories. The certified values for benzene, toluene and m-xylene are, respectively, 1.053 k 0.014, 1.125 k 0.015 and 1.043 k 0.015 pg per tube. This Community Bureau of Reference (BCR) Certified Reference Material 112 is recommended for quality control and for calibration purposes. Keywords: Tenax; occupational hygiene; reference material; intercomparison; aromatic hydrocarbons National and international legislation prescribes that the exposure of individual workers to certain potentially harmful substances (defined as agents according to Directive 80/1107EEC)l shall be assessed.If the assessment indicates that exposure is likely to be in the region of the relevant exposure limit, regular monitoring must be carried out. Of the various techniques that are suitable for regular monitoring of personal exposure to a wide variety of organic vapours, a device commonly used is a sorption tube. In this technique a known volume of workplace air (usually from the worker's breathing zone) is drawn through the tube by means of a sampling pump, or a mass of analyte is collected by diffusion. The vapours collected are recovered by solvent or thermal desorption and determined, usually by gas chroma- tography (GC).The certified reference material (CRM) described [Community Bureau of Reference (BCR)*] is based on Tenax, a sorbent that is usually desorbed thermally. This is the first sorbent for which appropriate long-term stability has been demonstrated. A CRM to be used for quality control or calibration should ideally closely match the material being analysed. However, it would be impractical to have CRMs to cover all possible combinations of organic solvents. A choice of representative analytes must therefore be made, and for the first certifica- tion, benzene, toluene and rn-xylene have been chosen on the basis of their toxicity and wide usage throughout the world. Experimental Preparation of Homogeneously Charged Batches The tubes used for the preparation of the test batches and for the production of CRM 112 were commercially available stainless-steel tubes which were cleaned and filled with Tenax as shown in Fig.1. The tube dimensions were: length, 89 mm; o.d., 6.34 mm; and i.d., 5.0 mm. Proprietary brass caps with polytetrafluoroethylene (PTFE) ferrules were used to close the tubes. (Preliminary experiments had shown that the aluminium caps supplied with the tubes were inadequate for the purpose of long-term storage.) The bed of Tenax (100 mg, 60-80 mesh) was 32 mm long and was retained at one end by a metal gauze and at the other end by a plug of quartz wool, a metal gauze and a spring. All tubes were pre-conditioned for 16 h at 300 "C and with a helium flow of 30 ml min-1. Each tube contained, before charging, less than 1 ng (the limit of detection of the method is 1 ng) of benzene, toluene and m-xylene.Diffusion cells were used to blend known levels of benzene, toluene and rn-xylene vapours into a stream of clean air. A known volume of this air was drawn through each tube. The apparatus used is represented schematically in Fig. 2. The amounts of benzene, toluene and rn-xylene (approxi- mately 1 yg of each), the flow-rates and the volumes of air (approximately 1 1 per tube) were such that saturation or breakthrough did not occur. The following precautions were taken to ensure that all the tubes received the same mass of each vapour: (i) the temperature of diffusion cells was controlled to within k 0.02 "C; (ii) the mass flows of air were controlled to within k 0.2%; (iii) all tubes were charged in an uninterrupted period of time during which the atmospheric pressure (which influenced the rates of diffusion) varied only slightly ( e .g . , the finally certified batch was charged in a period of 44 h in which the lowest and highest atmospheric pressures were 100.1 and 102.0 kPa); ( i v ) the total vapour content of the air was continuously monitored by use of a photoionization detector; and ( v ) the amount of air drawn through each tube was fixed by an electronic timer which operated the valves. G H Fig. 1 Schematic representation of a sample of CRM 112. A, Stainless-steel tube; B, metal gauze; C, s ring; D, quartz wool plug; E, grooves for O-rings; F, bed of Tenax 800 mg); G, Swagelok caps (brass); and H, PTFE ferrules438 ANALYST, MAY 1991, VOL.116 The volume flow-rates were measured at regular intervals using calibrated mercury sealed piston flow meters. The diffusion rates of the cells were known accurately from gravimetric determinations which were made weekly over a period of more than 3 months. The purity of the liquids in the diffusion cells was greater than 99.9%. On the basis of the measured flow-rates, rates of diffusion and times taken to charge the tubes, the amounts of vapour received by each tube could be calculated with an estimated uncertainty [a combina- tion of precision (95% confidence limits) and bias] of 0.015 pg of each compound for individual tubes and 0.010 pg for the means of a batch.Assuming quantitative sorption on the Tenax, the amounts charged to each sample of CRM 112 were calculated to be: 1.054 pg of benzene; 1.123 pg of toluene; and 1.039 pg of rn-xylene. Homogeneity and Stability Tests The homogeneity and stability of each batch studied were tested in the following manner: (a) a number of tubes (13 for the test batches, 40 for the reference material) were selected systematically so as to include all sampling units (see Fig. 2) and to include at least one in every four of each series of 12 tubes that were loaded simultaneously. These tubes were analysed in one laboratory on the same day. ( 6 ) Tubes stored under different conditions, i.e., in a refrigerator (04 "C), at room temperature (19-23 "C) and at approximately 40 "C, were analysed at regular time intervals (in the same labora- tory, the tubes stored under different conditions all being analysed on the same day).All these tests were carried out at the laboratory which prepared the batches. Intercomparisons For each round of analysis (two preliminary intercomparisons and the certification exercise), each participating laboratory analysed from four to ten tubes. The participating laboratories were as follows: Akzo Research (Arnhem, The Netherlands); BP Research (Sunbury-on-Thames, UK); Directoraat-Gene- raal Van de Arbeid (Voorburg, The Netherlands); Dow Chemical (Nederland) (Terneuzen, The Netherlands); Eolas (Dublin, Ireland); Health and Safety Executive, Occupational Medicine and Hygiene Laboratory (London, UK); ICI Chemicals and Polymers plc.(Runcorn, UK); Koninklijke- Shell Laboratorium (Amsterdam, The Netherlands); Labora- tory of the Government Chemist (London, UK); Arbejdsmil- joinstituttet (Hellerup, Denmark); Rhone-Poulenc Indus- trialization(Decines-Charpieu, France); State Laboratory (Dublin, Ireland); and Universita Degli Studi di Urbino (Urbino, Italy). In the following text, numerical codes (not related to the above alphabetical order) are used to refer to the analytical laboratories. M FC Mixing Critical orifice Sampling I unit x 12 Fig. 2 Schematic representation of the vapour generating and tube charging apparatus; up to 12 tubes were charged in parallel. (MFC = mass flow controller) All the laboratories but one used a procedure based on thermal desorption and gas chromatographic separation and detection (represented schematically in Fig.3) as follows. (a) A carrier gas stream was passed through the tube to be analysed, which was held at approximately 250 "C (laboratory 9: 275 "C; laboratory 6: 280 "C) for 5-10 min (laboratory 10: 3 min; laboratory 6: 4 min; laboratory 8: 25 min). The gas stream was then passed into a cold trap which contained a small amount of Tenax (or similar sorbent) and which was held at approximately -30 "C (laboratory 8: -25 "C). (6) For the second step of the procedure (injection into the gas chromatograph), a valve was switched so that the gas flow was directed into the GC column and the cold trap was heated to 250-300 "C. (c) In the third step of the procedure (gas chromatographic separation and detection) the valve was switched to the original position and (unless the isothermal mode was used) the temperature programme of the chromato- graph was started. Most of the laboratories used a capillary column (length, 25-60 m; i.d., 0.20-0.32 mm) coated with a non-polar stationary phase [ e .g . , poly(dimethylsiloxane)]. Laboratory 6 used a packed column (2 m x 2 mm i.d.). A wide range of temperature programmes were used (chosen so as to complete the chromatogram in 5-20 min). A flame ionization detector was used in all instances. The procedures used in laboratories 6, 11 and 12 differed significantly from that outlined above. Solvent desorption and mass spectrometric detection were used in laboratory 11. The Tenax powder was removed from the tube and transferred with 2 ml of hexane into a 5 nil flask which was occasionally shaken.After 30-60 min at room temperature, samples from the solution were injected into the gas chromatograph with a syringe. No cold-trap was used in laboratory 6; the vapours were sorbed directly on the GC column which was held at room temperature. Laboratory 12 transferred the Tenax from the tubes supplied into 18 cm tubes before desorption. Certification of CRM 112 In the third round of analysis, the certification round, the participants had all re-evaluated their method of calibration. An example re-evaluation is given by Wright.3 All syringes and volumetric glassware used were calibrated gravirnetrically and the analytical balances were checked with certified weights. The results were calculated from the GC detector response using calibration graphs based on three or more calibration solutions. These solutions were prepared by mixing accurately known volumes (laboratories 3 and 7) or masses (other laboratories) of pure benzene, toluene and rn-xylene with an accurately known amount of solvent (methanol, cyclohexane or carbon disulphide), in systems designed so as to minimize losses through evaporation. Successive dilution was avoided.The purity of the products used was verified. In laboratories 6 and 12, microlitre amounts of standard solutions were injected directly into the gas chromatograph for calibration. In other laboratories, similar amounts of standard solutions were added to clean Tenax tubes by injection in a flow of clean gas (injection of liquid on the Tenax powder was avoided) and these tubes were analysed in the same way as the samples. Thermal desorber C Fig.3 Schematic representation of thermal desorption-gas chromatography equipmentANALYST, MAY 1991, VOL. 116 439 I I I 1 /- + C Q) c. C 1.05 8 .? 1 .oo - (0 C .- w- 0 C 0 tj 0.95 2 .- LL 1.05 1 .oo 0.95 I I 1 0 3 6 14 Time/months Fig. 4 Stability of a batch of charged tubes similar to CRM 112. Mean k one standard deviation of a set of 5-13 analyses. ( a ) Benzene; (b) toluene; and (c) rn-xylene. 0, storage at 0 4 OC; 0, storage at ambient temperature; and A, storage at 40 "C Results and Discussion Homogeneity and Stability Tests The homogeneity test yielded in all instances relative standard deviations (RSDs) of <1.5%; this corresponds with the RSD typical for the thermal desorption-gas chromatography pro- cedure used by Brown et al.4 It was concluded that no inhomogeneity could be detected.The stability test outlined above was applied only to the first test batch. The results are presented graphically in Fig. 4. Significant systematic differences which correlated to the volatility of the compound or to the storage temperature were not observed; the long-term variation in the results is thus of analytical origin and not caused by losses through evapora- tion, chemical reactions or irreversible sorption. The stability of CRM 112 has been monitored further by the periodic analysis of samples stored at room temperature (as the entire batch); no instability was detected after 25 months. Intercomparisons Figs.S(a) and (b), 6(a) and (b), and 7(a) and ( 6 ) present the intercomparison results for individual laboratories in graph- ical form where the variable plotted is the ratio of the value found to the known value, and the means and standard deviations are shown. In each instance, the mean values were close to the amounts with which the tubes were charged, but the confidence intervals were unsatisfactory as they were much larger than expected on the basis of the within-laboratory repeatability of the analysis. 1.50 1.25 1 .oo 0.75 0.50 * € 1 1 1 1 1 1 1 1 1 1 1 1 1 i i , a 3 ; f 0.50 (Illllllllllllj 1.50 1.25 1 .oo 0.75 0.50 ! 1 1 1 1 1 1 1 1 1 1 1 1 1 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 Laboratory n urn ber Fig. 5 Results of three successive intercomparisons for benzene on Tenax (mean value and standard deviation for each laboratory).'Known value' = value calculated from charging data. (a) First round; (b) second round; and ( c ) third round After each of the preliminary intercomparisons, the results were discussed with all the participating laboratories. They explained in great detail how their analyses had been carried out. Consequently, various potential sources of error were identified and eliminated. Considerable progress was made between the second and the third round of analysis. This improvement resulted from measures being taken, from which some laboratories expected only a minor improvement (e.g., calibration of volumetric glassware, preparation of calibration solutions in sealed systems in order to avoid losses through evaporation, gravimetric dilution and dispensing, weighing of the syringe before and after injection, and calibration and maintenance of the analytical balance).As all participating laboratories were considered to be highly competent analytical laboratories, other laboratories are expected to experience similar difficulties and therefore some typical sources of error are listed below. Errors due to the handling of the tubes Errors arising from the handling of the tubes included: loss of sorbent when applying force to remove the cap from the grooved end of the tubes; and losses of vapour (or contamina- tion risk) upon contact with the atmosphere when the tubes were opened in order to transfer the sorbent into another container for solvent desorption, or similar risks when caps440 ANALYST, MAY 1991. VOL.116 1.50 I (a) I 1.25 1 R 0.75 t 1.50 I(bl 4 1 > 1.25 0.50 \ 0.50 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 Laboratory number Fig. 6 Results of three successive intercomparisons for toluene on Tenax (mean value and standard deviation for each laboratory). 'Known value' = value calculated from charging data. (a) First round; (b) second round; and (c) third round were removed a long time before the determinations were carried out. Errors in the calibration Errors in calibration resulted from: partial evaporation of volatile compounds of interest, or of the volatile solvent, between preparation and final use of the standard solutions (especially when syringes were being manipulated); relying on uncalibrated syringes or other glassware (errors of the order of 10% can be expected); and ignoring the water content of the solvents used.Errors due to malfunctioning instruments Errors caused by malfunctioning instruments were due to leaks in the connections of the tube in the thermal desorber; the use of a malfunctioning analytical balance (if the calibra- tion is not checked periodically, an error may remain undetected as short-term repeatability may remain normal); and the use of an ionization detector that was not optimized (checking is required whenever the response factors for benzene, toluene and m-xylene differ by more than 2%). 1.50 1.25 1 .oo 0.75 0.50 1 1 I I I I I I I 1 1 I I 1 1.50 1 P) 3 m - 1.25 ; Y u 1.00 C 3 0 .c al - 3 0.75 >" 0.75 0.501 I I I I I I I I ' I 1 2 3 4 5 6 7 8 9 1 0 1 1 1 2 1 3 Laboratory number Fig.7 Results of three successive intercomparisons for rn-xylene on Tenax (mean value and standard deviation for each laboratory). 'Known value' = value calculated from charging data. (a) First round; (b) second round; and (c) third round Certification The results of the certification (effectively a third intercom- parison) are presented in Figs. 5(c), 6(c) and 7 ( c ) . For this intercomparison, the ten laboratories applying thermal desorption achieved excellent agreement. Only one result (laboratory 3, benzene) was identified as a straggler (Dixon test). As this result was associated with a calibration error, the result was rejected. The data obtained from laboratory 11 (which used solvent desorption) were also rejected, on the grounds either that solvent desorption had been used or that losses through volatilization had occurred when the Tenax was being transferred from the tube into the flask, or that desorption had been incomplete. The mean of laboratory means and the halfwidth of its 95% confidence interval for each component are as follows: benzene, (1.053 k 0.014) pgper tube; toluene, (1.125 k 0.015) pg per tube; and rn-xylene, (1.043 k 0.015) pg per tube. If the interlaboratory standard deviation is significantly larger (as calculated using the F-test) than the intralaboratory standard deviation, the presence of remaining systematic errors in the participating laboratories is indicated. It is generally assumed that these errors are randomized, i.e., they do not cause aANALYST, MAY 1991, VOL.116 44 1 systematic error in the mean of means, if the interlaboratory standard deviation does not exceed the intralaboratory standard deviation by more than a factor of 2-3; in the present instance, the factor is <1.0 for each analyte. In addition, the values measured correspond closely with the masses with which the tubes were charged (the difference being 0.1, 0.2 and 0.4% for benzene, toluene and m-xylene, respectively). These results were therefore considered to provide sufficient basis for certification. Conclusions The data presented in this paper demonstrate that a batch of Tenax tubes charged with benzene, toluene and m-xylene can be prepared and analysed to an uncertainty of better than 2%, provided the analytical laboratories apply due care and attention to detail.Initial intercomparisons gave results with much larger errors and some of the potential sources of such errors have been identified. The absence of a detectable systematic error is a sufficient reason to declare the batch certified to BCR specifications and the batch is offered for sale for calibration and quality control. The uncertainty obtained renders the material useful for these purposes. The certification of this reference material is the first major achievement in the BCR efforts to provide means of quality assurance in occupational health monitoring. Work similar to that described here is in various stages of progress for: (i) aromatic hydrocarbons on active charcoal (for solvent desorption); (ii) chlorinated C2 hydrocarbons on Tenax; and (iii) esters and ketones on Tenax. Several other possible projects are in the discussion stage; feasibility studies on amines (including triethylamine), alde- hydes (including formaldehyde) and isocyanates are planned for the near future. References Council of the European Communities, Directive 88/1107/EEC, Official Journal of the European Communities, 1980, No. L327/8. Vandendriessche, S., and Griepink, B., The Certification of Benzene, Toluene and rn-Xylene Sorbed on Tenax in Tubes, Report EUR 1 2 3 0 8 ~ ~ , Commission of the European Communi- ties, Luxembourg (Office for Official Publications of the European Communities, 1989). Wright, M. D., BCR Certification of Reference Materials- Organics on Tenax-a Critical Examination of Sources of Calibration Error, Health and Safety Executive (HSE) Internal Report IR/L/IA/90/3, HSE, London, 1990. Brown, R. H., Cox, P. C., Purnell, C. J., West, N. G., and Wright, M. D., in Identification and Analysis of Organic Pollutants in Air, ed. Keith, L. H., Ann Arbor Science Publishers/Buttenvorth, Woburn, MS, 1984. Paper 010.53936 Received November 29th, 1990 Accepted December 12th, 1990