- Email: [email protected]
International comparison of key volatile organic components in indoor air
Measurement 82 (2016) 476–481
Contents lists available at ScienceDirect
Measurement journal homepage: www.elsevier.com/locate/measurement
International comparison of key volatile organic components in indoor air Nicholas D.C. Allen ⇑, Paul J. Brewer, Richard J.C. Brown, Robert P. Lipscombe, Peter T. Woods Analytical Science Division, National Physical Laboratory, Teddington, Middlesex TW11 0LW, United Kingdom
a r t i c l e
i n f o
Article history: Received 28 July 2015 Accepted 13 January 2016 Available online 19 January 2016 Keywords: Inter-laboratory comparison Volatile organic compounds Thermal desorption Tube loading Indoor air
a b s t r a c t An international comparison was coordinated by the National Physical Laboratory to assess the analytical capabilities of laboratories for measuring ten selected volatile organic compounds. These were chosen to represent a wide range of non-methane hydrocarbons, terpenes and oxygenated hydrocarbons, which pose a significant risk to public health from their presence in indoor environments from emissions by building and consumer materials. The components were loaded onto sorbent tubes and distributed to twentyfour participating laboratories for analysis. It was found that laboratories predominantly reported analytical results that were greater than the reference value, however the majority of results were closer to the reference value when compared to the previous comparison carried out in 2010. Crown Copyright Ó 2016 Published by Elsevier Ltd. All rights reserved.
1. Introduction It has been well documented that air quality influences human health. Poor air quality has been linked to higher mortality rates, and other adverse health effects, such as bronchitis and cardiovascular problems and the phenomenon known as Sick Building Syndrome [1,2]. There is significant work being carried out globally to ensure that air quality is being monitored and regulated to minimise negative effects on public health [3–5]. While there has been extensive research to study the impact of components in outdoor air on public health, measurements of indoor air quality have been a lower priority. Currently in the European Union there are three obligatory national schemes for assessing emissions in indoor air from construction products, the most well-known being the German AgBB evaluation scheme, as well as a number of voluntary schemes. In the forthcoming years it is likely that more regulations will be introduced. According to ⇑ Corresponding author. Tel.: +44 (0)20 8943 6913. E-mail address: [email protected] (N.D.C. Allen).
the US Environmental Protection Agency (EPA), indoor air can be two to five times more polluted than outdoor air [6]. As the public are spending increasing lengths of their time indoors, it is necessary that careful monitoring is implemented and to achieve this, the development of a traceable measurement infrastructure is required. This is of increasing importance within the construction of new buildings and there are international and European Union Construction Products Directives including EN ISO/IEC 16000-11:2006 [7] and CEN/TC 351 (document number 0588) [8] associated with these and other air pollutants. With continual improvements being made to seal buildings in an effort to make them more energy efficient, the overall ventilation and air exchange rates with the outside have reduced, leading to heightened concentrations of pollutants trapped indoors [9]. It is essential that laboratories carrying out the analysis of volatile organic compounds (VOCs) found in indoor air ensure that results are comparable and traceable. Proficiency testing schemes are one way of showing compliance with quality control procedure and demonstrating comparability [10,11]. One such scheme for assessing
http://dx.doi.org/10.1016/j.measurement.2016.01.027 0263-2241/Crown Copyright Ó 2016 Published by Elsevier Ltd. All rights reserved.
477
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
global laboratory performance in air quality analysis is the Ambient, Indoor, Workplace Air and Stack Emissions Proficiency Testing Scheme (AIR PT Scheme) [12]. This is a combination of the former Workplace Analysis Scheme for Proficiency (WASP) [13] and Stack Emissions Proficiency Testing Scheme (STACKS) [14]. However, while such schemes provide an insight into the performance of analytical laboratories for measuring routine components, such as benzene, studies of components with emerging measurement requirements for indoor air quality are scarce. An international comparison was coordinated by the National Physical Laboratory (NPL), in which ten selected components commonly found in indoor air from construction material emissions, were either gas or liquid loaded onto sorbent tubes and sent for analysis to each of the participating laboratories. The exercise was aimed at laboratories (some of which are ISO 17025 accredited [15]) undertaking the analysis of VOCs from air samples collected on sorbent tubes and analysed using thermal desorption techniques as detailed in ISO 16017-1 [16] and 16017-2 [17]. This comparison provides a means of assessing the analytical capability of each laboratory [18,19]. The results of the comparison are presented and contrasted to a similar comparison carried out in 2010 [10]. 2. Experimental To keep anonymity each participant and sample set was ascribed a letter A to X, and the assignment was only known to the coordinators at NPL. Participants were supplied with Silco-treated TenaxÒ sorbent tubes: one blank and three loaded (despatched in packing containers) with benzene, toluene, butyl acetate, hexanal, m-xylene, a-pinene (both optical isomers were included at approximately an even ratio), styrene, 1,2,3-trimethylbenzene, 2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone. Four of the components selected (benzene, toluene, butyl acetate and 2-ethyl-1-hexanol) featured in a comparison organised by NPL in 2010. In addition to the components included in the 2010 comparison, six additional components were chosen that are commonly found in indoor air from building emissions but are not frequently analysed. Analysis of blanks confirmed that packaging and transport conditions did not introduce material onto the sorbent tubes. 2-Ethyl-1-hexanol and 1-methyl-2-pyrrolidone were liquid loaded due to their low vapour pressures. This was done prior to the gas loading of the other components, sorbent tubes were loaded with 2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone from a liquid reference standard following a similar method to that adopted by Martin et al. [20]. A nominal 5 lL aliquot of a nominal 40 lg L1 stock solution containing both components was injected from a 10 lL syringe through a modified injection port onto each sorbent tube. After each injection, compressed air scrubbed by a filter was passed over the sorbent tube for a minute at a flow rate of approximately 150 mL min1 to remove the solvent. The 10 lL dispensing syringe was weighed before and after injection to determine the loaded mass transferred onto each sorbent tube. The volume of the syringe was set by a Chaney adaptor to improve repeatability.
The remaining eight components were loaded from a Primary Reference Gas Mixture (PRGM), gravimetrically prepared and validated at NPL. The composition of the PRGM is shown in Table 1. The PRGM was passed over each sorbent tube in turn at a nominal flow rate of 10 mL min1 for 15 minutes. A minimised dead volume connector and a low volume restriction device (used to control sample flow) were employed. Each day the system was purged for two hours before tube loading commenced. The other end of the sorbent tube was connected to a flow meter (BIOS), which was used to record the flow during each tube loading. In total ninety-two sorbent tubes were loaded over four days. The mass loading of all of the eight components was determined using Eq. (1), where m is the mass of the analyte loaded (g), f is the PRGM flow rate (L min1) at STP, t is the loading time (min), Vm is the molar volume of an ideal gas at standard temperature and pressure (L mol1), x is the amount fraction of the component in the PRGM (mol mol1) and Mw is the molecular weight of the component (g mol1).
m¼
ftxMw Vm
ð1Þ
A set of gravimetric liquid standards, containing all of the ten VOC components, were prepared at nominal concentrations of 20, 40 and 60 g L1 in methanol (P99.9%, Sigma Aldrich). These were diluted with the same methanol to produce solutions with nominal concentrations of 60, 100, 200, 400, 600 and 1000 lg L1. Aliquots of 1 lL of the liquid standards were injected onto Silco-treated TenaxÒ sorbent tubes (as previously described for 2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone) and analysed using Thermal Desorption Gas Chromatography (TD-GC) Flame Ionisation Detection (FID) and Mass Spectroscopy (MS) as outlined in 16017-2 to create a calibration curve. This was used to validate the gravimetric loadings of a sample of eleven tubes taken from the batch loaded for the comparison. The batch of eleven tubes comprised the first and last tube loaded on each day, as well as an additional tube from each of the first three days. These measurements were also used to interrogate the homogeneity of the batch. For most components the standard deviation for analysis of the batch of eleven tubes was less than 2%. An indoor air method was created where the TenaxÒ sorbent tubes were desorbed for 15 minutes at 275 °C. The system was dry purged for three minutes and then
Table 1 The amount fraction of components in the primary reference gas mixture used for gas loading. Component
x (nmol mol1)
Benzene Toluene Butyl acetate Hexanal m-Xylene a-pinene Styrene 1,2,3-Trimethylbenzene Nitrogen
248.6 262.8 253.3 242.6 250.2 246.3 254.5 237.7 Balance
478
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
pre-purged for one minute before desorption, with material then held on a U-T6SUL cold trap (a trap split flow of 20 ml min1 was set) before being released onto a 60 m Zebron 1701 column (0.32 mm inner diameter, 1 lm film thickness). The GC temperature program started at 40 °C and ramped at a rate of 5 °C a minute to 100 °C. The method held the temperature at 100 °C for four minutes before ramping at a rate of 7.5 °C to 150 °C and finally increasing to 250 °C at a rate of 40 °C each minute. Tube loading for the 2010 comparison was similar to the methodology described above. Two components, 2-ethyl-1-hexanol and dodecane, were liquid loaded and four other components (benzene, toluene, butyl acetate and o-xylene) were gas loaded.
3. Results and discussion A summary of the reported results from 2010 and 2013 comparison are presented in Table 2 as a function of performance assessment. Values have been normalised to the gravimetric loadings. A standard deviation of performance assessment (rPA), with respect to the reference value, was set as 7.5% to allow comparison with other schemes, such as the WASP scheme. Reference values were assigned by taking the gravimetric loadings for all sorbent tubes, with the exception of three components (butyl acetate, hexanal and 1,2,3trimethylbenzene) in the 2013 comparison where a significant bias was observed between the gravimetric and validation values. For butyl acetate and hexanal the gravimetric value was larger than the analytical result, indicating retention of the components on the sorbent material. However, for 1,2,3-trimethylbenzene a larger analytical value compared to the gravimetric loading was observed which may be attributed to background levels in the system or some retention of the calibration standard. For these components, the reference values were assigned from the analytical values for the eleven tubes analysed at NPL. In the 2010 comparison, all reference values were assigned from the gravimetric loadings. To confirm the stability of the batch, a sample of three tubes were analysed after the submission deadline for the laboratory results. A difference of less than ±2% was observed between the mean reported analytical value for
the three tubes when compared to the batch of eleven tubes analysed at NPL to confirm homogeneity, for all components with the exception of 1-methyl-2-pyrrolidone (3%) and 1,2,3-trimethylbenzene (4%). The ten components have been subdivided into three categories for the purpose of this discussion: Set 1 – hydrocarbon components that are routinely measured (benzene, toluene and m-xylene), Set 2 – more challenging aromatic and unsaturated hydrocarbon components that are not routinely measured (a-pinene, styrene and 1,2,3trimethylbenzene) and Set 3 components containing one or more heteroatom, two of which were liquid loaded (hexanal, butyl acetate, 2-ethyl-1-hexanol and 1-methyl2-pyrrolidone). A greater number of results were reported for Set 1 components in contrast to Set 2 and 3. For the Set 1 components over half of laboratories reported results that fell within one rPA in the 2013 comparison. In the 2010 comparison, the proportion of measurements of benzene and m-xylene within one rPA was much lower (both 38%) than in 2013 suggesting a significant improvement in laboratory analysis. For example, of the eleven external participants that reported results for benzene in both the 2010 and 2013 comparison, only three of them had an average result closer to the reference value in the 2010 comparison whereas most of them performed better in 2013. Similarly, of the eleven external participants that reported results for toluene in the 2010 comparison, nine of the participants’ average result was closer to the reference value in the 2013 comparison. For Set 2 components, fewer results were within one rPA relative to Set 1. Results predominantly fell within two rPA, suggesting a greater analytical challenge. For the component in Set 3, results were mixed. For butyl acetate in the 2013 comparison, just over half of the reported results were within two rPA. Hexanal appears to be the most challenging component to analyse, as only 23% of results are within two rPA. As a significant number of laboratories underestimated the loading, this could suggest that the component was difficult to desorb. The reported results for the two liquid-loaded components (2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone) were closer to the reference value than hexanal. Over half of the laboratories were within two rPA for 2-ethyl-1-hexanol, a marked improvement on the 2010 comparison where
Table 2 Results for each component in the 2010 and 2013 comparisons. Component
Benzene Toluene m-Xylene o-Xylene a-pinene Styrene 1,2,3-Trimethylbenzene Hexanal Butyl acetate 2-Ethyl-1-hexanol 1-Methyl-2-pyrrolidone
x 6 ±7.5%
±7.5% < x 6 ±15%
±15% < x 6 ±22.5%
x > ±22.5%
Total measurements
2010
2013
2010
2013
2010
2013
2010
2013
2010
2013
5 8 – 5 – – – – 5 4 –
13 13 13 – 7 10 9 3 8 10 7
2 1 – 4 – – – – 1 1 –
2 5 2 – 6 4 6 2 4 1 3
1 0 – 1 – – – – 4 4 –
1 2 4 – 4 2 3 4 4 2 2
5 4 – 3 – – – – 2 2 –
7 4 5 – 3 6 4 13 7 7 7
13 13 – 13 – – – – 12 11 –
23 24 24 – 20 22 22 22 23 20 19
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
only 36% of laboratories achieved this (six of the nine external laboratories that reported average results closer to the reference value in the 2013 comparison), however only eleven laboratories reported results. The fewest laboratories attempted to analyse 1-methyl-2-pyrrolidone, however over half of those who did managed to report an average analytical result were within two rPA. Fig. 1 shows the 2010 and 2013 comparison results for benzene. Each point corresponds to one reported measurement and its relative difference to the reference value. In 2013, more laboratories report values closer to the reference value than in 2010, showing that laboratory performance in the most recent comparison has improved. Fig. 2 shows reported results for five components common to both the 2010 and 2013 comparisons. It shows that
479
for the majority of components, laboratories frequently give results that are higher than the reference value. This is more pronounced for benzene and toluene unlike the xylenes and 2-ethyl-1-hexanol, where there is a more even distribution of reported results around the reference value. The distribution of reported values for butyl acetate shows an improvement in 2013 with results closer to the reference value than in the 2010 comparison. In general, shows better similarity, with more results closer to the reference value. The median relative difference for all of the reported results in the 2013 comparison was calculated for each component and was plotted against carbon number in Fig. 3. Where more than one component had the same carbon number an average was taken. Fig. 3 shows the best agreement for components with a carbon number of six
Fig. 1. The relative difference of reported results from the reference value for benzene in the 2010 (open circles) and 2013 (closed circles). Where applicable, laboratories participating in both comparisons have been assigned the same number.
Fig. 2. In ascending order the average relative difference for each participating laboratory relative to the assigned reference for five components common to the 2010 (open circles) and 2013 (closed circles). Four results that exceed the y-axis values are excluded.
480
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
and seven. Heavier and lighter components are shown to result in larger deviations from the reference value. Gas was loaded onto TenaxÒ sorbent tubes and these are usually recommended for VOCs with a carbon number of six or greater [21], therefore this may account for the trend observed with the lighter components. The median relative difference in general increases for a carbon number of eight or greater. This could be a consequence of laboratories using an unsuitable response factor for these components. Fig. 4 shows the absolute relative difference from the reference value of all laboratories in the 2013 comparison by component versus the difference in gravimetric and analytical measurements of the sample of eleven tubes analysed at NPL. The correlation shows that components that were challenging to recover resulted in larger deviations of laboratories from the reference value. The difference between the median analytical value and the median
gravimetric value was greatest for the oxygenated VOCs butyl acetate (15 ng) and hexanal (33 ng), whereas the difference for all other components was less than 10 ng. Likewise the largest mean absolute relative differences reported were for butyl acetate (19%), hexanal (33%) and styrene (20%). This may be attributed to these components being more difficult to desorb from the sorbent tubes or as a result of using an inaccurate response factor in the calibration. This highlights the need for better sampling methods to improve the state of the art for these components. For the homogeneity analysis, NPL used liquid reference standards to validate the gravimetric loadings. Therefore some of the difference between the gravimetric and reported result may be attributed to the difference between gas and liquid loading. It was found that for the homogeneity analysis the mean difference between the liquid standards and the reported gas loading on the tube was less than 5%
Fig. 3. The median relative difference of all reported results in 2013 comparison versus carbon number.
Fig. 4. The absolute mean relative difference of all reported results for each gas loaded component in 2013 against the absolute difference between the median gravimetric value and the median certified value from validation measurements.
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
for all of the components except styrene (6%), butyl acetate (9%) and hexanal (22%). These challenging components should be where future focus lies in the development of reference standards for indoor air. 4. Conclusions The comparison has given an important insight into the trends in the capability and performance of analytical laboratories for measuring VOCs for indoor air emissions. In 2013 the proportion of laboratories reporting results within one rPA increased relative to the comparison carried out in 2010. Participants involved in both of the comparisons predominantly showed average analytical results closer to the reference value in the 2013 comparison when contrasted to the 2010 comparison, highlighting that the state of the art has improved. For the majority of components, laboratories overestimated the reference values. Further collaborations are imperative to help us to better understand what caused the biases observed and whether it is a function of the analytical technique adopted. More challenging components included oxygenated VOCs; here more work is required to develop indoor air reference standards and sampling techniques. Similarly, new methods and sorbents suitable for analysing different classes of components, are required in an effort to understand the poor performance observed in this comparison for oxygenated VOCs. To ensure that improvements continue to be made it is essential that future comparisons take place. With increasing monitoring of VOCs in indoor air and new legislation being drafted such as prEN16516 (Construction products—Assessment of release of dangerous s ubstances—Determination of emissions into indoor air) studies, such as this comparison, are essential in helping to provide an infrastructure for indoor air quality to support new legislation. Acknowledgements The authors would like to acknowledge the funding of the work by the UK Government’s National Measurement System Chemistry and Biology Knowledge Base Programme. The authors would also like to thank all of the laboratories that participated in the comparison. References [1] P. Wolkoff, G.D. Nielsen, Organic compounds in indoor air—their relevance for perceived indoor air quality?, Atmos Environ. 35 (26) (2001) 4407–4417. [2] J. Sundell, On the history of indoor air quality and health, Indoor Air 14 (Supplementary 7) (2004) 51–58.
481
[3] T. Brennan, Indoor Environmental Quality and Climate Change, US Environmental Protection Agency, Washington, DC, 2010. [4] L. Molhave, G. Clausen, B. Berglund, J. de Ceaurriz, A. Kettrup, T. Lindvall, M. Maroni, A.C. Pickering, U. Risse, H. Rothweiler, B. Seifert, M. Younes, Total volatile organic compounds (TVOC) in indoor air quality investigations, Indoor Air 7 (4) (1997) 225–240, http://dx. doi.org/10.1111/j.1600-0668.1997.00002.x. [5] British Standard, BS EN 482:2012, Workplace exposure – general requirements for the performance of procedures for the measurement of chemical agent. [6] U.S. Environmental Protection Agency, Office of Air and Radiation, Report to congress on indoor air quality, volume II: assessment and control of indoor air pollution, EPA 400-1-89-001C, 1989. (last accessed 28.07.15). [7] EN ISO/IEC 16000-11:2006, Determination of the emission of volatile organic compounds from building products and furnishing – sampling, storage of samples and preparation of test specimens. [8] Harmonized EU VOC-classes – classification system for VOC emissions of construction products used in the indoor environment – CEN/TC 351, document number 0588. [9] A.P. Jones, Indoor air quality and health, Atmos. Environ. 33 (28) (1999) 4535–4564. [10] PT. Woods, Proficiency Testing for the Analysis of Hydrocarbons using Passive Adsorption Tubes – in support of the EU Construction Products Directive & Related UK Regulations: In Support of the EU Construction Products Directive & UK Regulations - Presentation to the Meeting on ‘‘New Construction Products Regulations” – British Measurement and Testing Association held at Building Research Establishment, 17 November 2010 [11] EN ISO/IEC 17011:2004, Conformity assessment – general requirements for accreditation bodies accrediting conformity assessment bodies. [12] AIR PT, ambient, indoor, workplace air and stack emissions proficiency testing scheme LGC standards proficiency testing. Issue 2, September 2014. [13] P.R. Stacey, M.D. Wright, A proficiency testing scheme for aromatic hydrocarbons in air by the manual thermal desorption-GC method: a comparison of laboratory performance with the uncertainty requirements of the European Union Ambient Air Directive, J. Environ. Monit. 3 (4) (2001) 425–431. [14] Stack Emissions Proficiency Testing Scheme (STACKS), (last accessed 28.07.15). [15] EN ISO/IEC 17025-1:2006/Cor1, General requirements for the competence of testing and calibration laboratories. [16] EN ISO/IEC 16027-1:2000, Indoor, ambient and workplace air – Sampling and analysis of volatile organics by sorbent tube/thermal desorption/capillary gas chromatography – Part 1: Pumped sampling. [17] EN ISO/IEC 16027-2:2003. Indoor, ambient and workplace air – sampling and analysis of volatile organics by sorbent tube/thermal desorption/capillary gas chromatography – Part 2: Diffusive sampling. [18] G.M. Roberts, E. Woolfenden, The application of thermal desorption for the analysis of trace toxic compounds in areas of human occupation, Indoor Air 1–5 (2005) 3859–3863. [19] E. Woolfenden, P. Simpson, P. Hughes, Improving the determination of VOCs and semi-VOCs in building materials and car trim components using direct thermal desorption, Indoor Air 1–5 (2005) 2303–2308. [20] N.A. Martin, N.L.A.S. Barber, J.K. Black, R.P. Lipscombe, C.A. Taig, A comparison of gas- and liquid-loaded sorbent standards for the calibration of measurements of volatile organic compounds, Atmos. Environ. 41 (35) (2007) 7666–7671. [21] E. Woolfenden, Sorbent-based sampling methods for volatile and semi-volatile organic compounds in air: Part 1: Sorbent-based air monitoring options, J. Chromatogr. A 1217 (16) (2010) 2674–2684.
Contents lists available at ScienceDirect
Measurement journal homepage: www.elsevier.com/locate/measurement
International comparison of key volatile organic components in indoor air Nicholas D.C. Allen ⇑, Paul J. Brewer, Richard J.C. Brown, Robert P. Lipscombe, Peter T. Woods Analytical Science Division, National Physical Laboratory, Teddington, Middlesex TW11 0LW, United Kingdom
a r t i c l e
i n f o
Article history: Received 28 July 2015 Accepted 13 January 2016 Available online 19 January 2016 Keywords: Inter-laboratory comparison Volatile organic compounds Thermal desorption Tube loading Indoor air
a b s t r a c t An international comparison was coordinated by the National Physical Laboratory to assess the analytical capabilities of laboratories for measuring ten selected volatile organic compounds. These were chosen to represent a wide range of non-methane hydrocarbons, terpenes and oxygenated hydrocarbons, which pose a significant risk to public health from their presence in indoor environments from emissions by building and consumer materials. The components were loaded onto sorbent tubes and distributed to twentyfour participating laboratories for analysis. It was found that laboratories predominantly reported analytical results that were greater than the reference value, however the majority of results were closer to the reference value when compared to the previous comparison carried out in 2010. Crown Copyright Ó 2016 Published by Elsevier Ltd. All rights reserved.
1. Introduction It has been well documented that air quality influences human health. Poor air quality has been linked to higher mortality rates, and other adverse health effects, such as bronchitis and cardiovascular problems and the phenomenon known as Sick Building Syndrome [1,2]. There is significant work being carried out globally to ensure that air quality is being monitored and regulated to minimise negative effects on public health [3–5]. While there has been extensive research to study the impact of components in outdoor air on public health, measurements of indoor air quality have been a lower priority. Currently in the European Union there are three obligatory national schemes for assessing emissions in indoor air from construction products, the most well-known being the German AgBB evaluation scheme, as well as a number of voluntary schemes. In the forthcoming years it is likely that more regulations will be introduced. According to ⇑ Corresponding author. Tel.: +44 (0)20 8943 6913. E-mail address: [email protected] (N.D.C. Allen).
the US Environmental Protection Agency (EPA), indoor air can be two to five times more polluted than outdoor air [6]. As the public are spending increasing lengths of their time indoors, it is necessary that careful monitoring is implemented and to achieve this, the development of a traceable measurement infrastructure is required. This is of increasing importance within the construction of new buildings and there are international and European Union Construction Products Directives including EN ISO/IEC 16000-11:2006 [7] and CEN/TC 351 (document number 0588) [8] associated with these and other air pollutants. With continual improvements being made to seal buildings in an effort to make them more energy efficient, the overall ventilation and air exchange rates with the outside have reduced, leading to heightened concentrations of pollutants trapped indoors [9]. It is essential that laboratories carrying out the analysis of volatile organic compounds (VOCs) found in indoor air ensure that results are comparable and traceable. Proficiency testing schemes are one way of showing compliance with quality control procedure and demonstrating comparability [10,11]. One such scheme for assessing
http://dx.doi.org/10.1016/j.measurement.2016.01.027 0263-2241/Crown Copyright Ó 2016 Published by Elsevier Ltd. All rights reserved.
477
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
global laboratory performance in air quality analysis is the Ambient, Indoor, Workplace Air and Stack Emissions Proficiency Testing Scheme (AIR PT Scheme) [12]. This is a combination of the former Workplace Analysis Scheme for Proficiency (WASP) [13] and Stack Emissions Proficiency Testing Scheme (STACKS) [14]. However, while such schemes provide an insight into the performance of analytical laboratories for measuring routine components, such as benzene, studies of components with emerging measurement requirements for indoor air quality are scarce. An international comparison was coordinated by the National Physical Laboratory (NPL), in which ten selected components commonly found in indoor air from construction material emissions, were either gas or liquid loaded onto sorbent tubes and sent for analysis to each of the participating laboratories. The exercise was aimed at laboratories (some of which are ISO 17025 accredited [15]) undertaking the analysis of VOCs from air samples collected on sorbent tubes and analysed using thermal desorption techniques as detailed in ISO 16017-1 [16] and 16017-2 [17]. This comparison provides a means of assessing the analytical capability of each laboratory [18,19]. The results of the comparison are presented and contrasted to a similar comparison carried out in 2010 [10]. 2. Experimental To keep anonymity each participant and sample set was ascribed a letter A to X, and the assignment was only known to the coordinators at NPL. Participants were supplied with Silco-treated TenaxÒ sorbent tubes: one blank and three loaded (despatched in packing containers) with benzene, toluene, butyl acetate, hexanal, m-xylene, a-pinene (both optical isomers were included at approximately an even ratio), styrene, 1,2,3-trimethylbenzene, 2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone. Four of the components selected (benzene, toluene, butyl acetate and 2-ethyl-1-hexanol) featured in a comparison organised by NPL in 2010. In addition to the components included in the 2010 comparison, six additional components were chosen that are commonly found in indoor air from building emissions but are not frequently analysed. Analysis of blanks confirmed that packaging and transport conditions did not introduce material onto the sorbent tubes. 2-Ethyl-1-hexanol and 1-methyl-2-pyrrolidone were liquid loaded due to their low vapour pressures. This was done prior to the gas loading of the other components, sorbent tubes were loaded with 2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone from a liquid reference standard following a similar method to that adopted by Martin et al. [20]. A nominal 5 lL aliquot of a nominal 40 lg L1 stock solution containing both components was injected from a 10 lL syringe through a modified injection port onto each sorbent tube. After each injection, compressed air scrubbed by a filter was passed over the sorbent tube for a minute at a flow rate of approximately 150 mL min1 to remove the solvent. The 10 lL dispensing syringe was weighed before and after injection to determine the loaded mass transferred onto each sorbent tube. The volume of the syringe was set by a Chaney adaptor to improve repeatability.
The remaining eight components were loaded from a Primary Reference Gas Mixture (PRGM), gravimetrically prepared and validated at NPL. The composition of the PRGM is shown in Table 1. The PRGM was passed over each sorbent tube in turn at a nominal flow rate of 10 mL min1 for 15 minutes. A minimised dead volume connector and a low volume restriction device (used to control sample flow) were employed. Each day the system was purged for two hours before tube loading commenced. The other end of the sorbent tube was connected to a flow meter (BIOS), which was used to record the flow during each tube loading. In total ninety-two sorbent tubes were loaded over four days. The mass loading of all of the eight components was determined using Eq. (1), where m is the mass of the analyte loaded (g), f is the PRGM flow rate (L min1) at STP, t is the loading time (min), Vm is the molar volume of an ideal gas at standard temperature and pressure (L mol1), x is the amount fraction of the component in the PRGM (mol mol1) and Mw is the molecular weight of the component (g mol1).
m¼
ftxMw Vm
ð1Þ
A set of gravimetric liquid standards, containing all of the ten VOC components, were prepared at nominal concentrations of 20, 40 and 60 g L1 in methanol (P99.9%, Sigma Aldrich). These were diluted with the same methanol to produce solutions with nominal concentrations of 60, 100, 200, 400, 600 and 1000 lg L1. Aliquots of 1 lL of the liquid standards were injected onto Silco-treated TenaxÒ sorbent tubes (as previously described for 2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone) and analysed using Thermal Desorption Gas Chromatography (TD-GC) Flame Ionisation Detection (FID) and Mass Spectroscopy (MS) as outlined in 16017-2 to create a calibration curve. This was used to validate the gravimetric loadings of a sample of eleven tubes taken from the batch loaded for the comparison. The batch of eleven tubes comprised the first and last tube loaded on each day, as well as an additional tube from each of the first three days. These measurements were also used to interrogate the homogeneity of the batch. For most components the standard deviation for analysis of the batch of eleven tubes was less than 2%. An indoor air method was created where the TenaxÒ sorbent tubes were desorbed for 15 minutes at 275 °C. The system was dry purged for three minutes and then
Table 1 The amount fraction of components in the primary reference gas mixture used for gas loading. Component
x (nmol mol1)
Benzene Toluene Butyl acetate Hexanal m-Xylene a-pinene Styrene 1,2,3-Trimethylbenzene Nitrogen
248.6 262.8 253.3 242.6 250.2 246.3 254.5 237.7 Balance
478
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
pre-purged for one minute before desorption, with material then held on a U-T6SUL cold trap (a trap split flow of 20 ml min1 was set) before being released onto a 60 m Zebron 1701 column (0.32 mm inner diameter, 1 lm film thickness). The GC temperature program started at 40 °C and ramped at a rate of 5 °C a minute to 100 °C. The method held the temperature at 100 °C for four minutes before ramping at a rate of 7.5 °C to 150 °C and finally increasing to 250 °C at a rate of 40 °C each minute. Tube loading for the 2010 comparison was similar to the methodology described above. Two components, 2-ethyl-1-hexanol and dodecane, were liquid loaded and four other components (benzene, toluene, butyl acetate and o-xylene) were gas loaded.
3. Results and discussion A summary of the reported results from 2010 and 2013 comparison are presented in Table 2 as a function of performance assessment. Values have been normalised to the gravimetric loadings. A standard deviation of performance assessment (rPA), with respect to the reference value, was set as 7.5% to allow comparison with other schemes, such as the WASP scheme. Reference values were assigned by taking the gravimetric loadings for all sorbent tubes, with the exception of three components (butyl acetate, hexanal and 1,2,3trimethylbenzene) in the 2013 comparison where a significant bias was observed between the gravimetric and validation values. For butyl acetate and hexanal the gravimetric value was larger than the analytical result, indicating retention of the components on the sorbent material. However, for 1,2,3-trimethylbenzene a larger analytical value compared to the gravimetric loading was observed which may be attributed to background levels in the system or some retention of the calibration standard. For these components, the reference values were assigned from the analytical values for the eleven tubes analysed at NPL. In the 2010 comparison, all reference values were assigned from the gravimetric loadings. To confirm the stability of the batch, a sample of three tubes were analysed after the submission deadline for the laboratory results. A difference of less than ±2% was observed between the mean reported analytical value for
the three tubes when compared to the batch of eleven tubes analysed at NPL to confirm homogeneity, for all components with the exception of 1-methyl-2-pyrrolidone (3%) and 1,2,3-trimethylbenzene (4%). The ten components have been subdivided into three categories for the purpose of this discussion: Set 1 – hydrocarbon components that are routinely measured (benzene, toluene and m-xylene), Set 2 – more challenging aromatic and unsaturated hydrocarbon components that are not routinely measured (a-pinene, styrene and 1,2,3trimethylbenzene) and Set 3 components containing one or more heteroatom, two of which were liquid loaded (hexanal, butyl acetate, 2-ethyl-1-hexanol and 1-methyl2-pyrrolidone). A greater number of results were reported for Set 1 components in contrast to Set 2 and 3. For the Set 1 components over half of laboratories reported results that fell within one rPA in the 2013 comparison. In the 2010 comparison, the proportion of measurements of benzene and m-xylene within one rPA was much lower (both 38%) than in 2013 suggesting a significant improvement in laboratory analysis. For example, of the eleven external participants that reported results for benzene in both the 2010 and 2013 comparison, only three of them had an average result closer to the reference value in the 2010 comparison whereas most of them performed better in 2013. Similarly, of the eleven external participants that reported results for toluene in the 2010 comparison, nine of the participants’ average result was closer to the reference value in the 2013 comparison. For Set 2 components, fewer results were within one rPA relative to Set 1. Results predominantly fell within two rPA, suggesting a greater analytical challenge. For the component in Set 3, results were mixed. For butyl acetate in the 2013 comparison, just over half of the reported results were within two rPA. Hexanal appears to be the most challenging component to analyse, as only 23% of results are within two rPA. As a significant number of laboratories underestimated the loading, this could suggest that the component was difficult to desorb. The reported results for the two liquid-loaded components (2-ethyl-1-hexanol and 1-methyl-2-pyrrolidone) were closer to the reference value than hexanal. Over half of the laboratories were within two rPA for 2-ethyl-1-hexanol, a marked improvement on the 2010 comparison where
Table 2 Results for each component in the 2010 and 2013 comparisons. Component
Benzene Toluene m-Xylene o-Xylene a-pinene Styrene 1,2,3-Trimethylbenzene Hexanal Butyl acetate 2-Ethyl-1-hexanol 1-Methyl-2-pyrrolidone
x 6 ±7.5%
±7.5% < x 6 ±15%
±15% < x 6 ±22.5%
x > ±22.5%
Total measurements
2010
2013
2010
2013
2010
2013
2010
2013
2010
2013
5 8 – 5 – – – – 5 4 –
13 13 13 – 7 10 9 3 8 10 7
2 1 – 4 – – – – 1 1 –
2 5 2 – 6 4 6 2 4 1 3
1 0 – 1 – – – – 4 4 –
1 2 4 – 4 2 3 4 4 2 2
5 4 – 3 – – – – 2 2 –
7 4 5 – 3 6 4 13 7 7 7
13 13 – 13 – – – – 12 11 –
23 24 24 – 20 22 22 22 23 20 19
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
only 36% of laboratories achieved this (six of the nine external laboratories that reported average results closer to the reference value in the 2013 comparison), however only eleven laboratories reported results. The fewest laboratories attempted to analyse 1-methyl-2-pyrrolidone, however over half of those who did managed to report an average analytical result were within two rPA. Fig. 1 shows the 2010 and 2013 comparison results for benzene. Each point corresponds to one reported measurement and its relative difference to the reference value. In 2013, more laboratories report values closer to the reference value than in 2010, showing that laboratory performance in the most recent comparison has improved. Fig. 2 shows reported results for five components common to both the 2010 and 2013 comparisons. It shows that
479
for the majority of components, laboratories frequently give results that are higher than the reference value. This is more pronounced for benzene and toluene unlike the xylenes and 2-ethyl-1-hexanol, where there is a more even distribution of reported results around the reference value. The distribution of reported values for butyl acetate shows an improvement in 2013 with results closer to the reference value than in the 2010 comparison. In general, shows better similarity, with more results closer to the reference value. The median relative difference for all of the reported results in the 2013 comparison was calculated for each component and was plotted against carbon number in Fig. 3. Where more than one component had the same carbon number an average was taken. Fig. 3 shows the best agreement for components with a carbon number of six
Fig. 1. The relative difference of reported results from the reference value for benzene in the 2010 (open circles) and 2013 (closed circles). Where applicable, laboratories participating in both comparisons have been assigned the same number.
Fig. 2. In ascending order the average relative difference for each participating laboratory relative to the assigned reference for five components common to the 2010 (open circles) and 2013 (closed circles). Four results that exceed the y-axis values are excluded.
480
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
and seven. Heavier and lighter components are shown to result in larger deviations from the reference value. Gas was loaded onto TenaxÒ sorbent tubes and these are usually recommended for VOCs with a carbon number of six or greater [21], therefore this may account for the trend observed with the lighter components. The median relative difference in general increases for a carbon number of eight or greater. This could be a consequence of laboratories using an unsuitable response factor for these components. Fig. 4 shows the absolute relative difference from the reference value of all laboratories in the 2013 comparison by component versus the difference in gravimetric and analytical measurements of the sample of eleven tubes analysed at NPL. The correlation shows that components that were challenging to recover resulted in larger deviations of laboratories from the reference value. The difference between the median analytical value and the median
gravimetric value was greatest for the oxygenated VOCs butyl acetate (15 ng) and hexanal (33 ng), whereas the difference for all other components was less than 10 ng. Likewise the largest mean absolute relative differences reported were for butyl acetate (19%), hexanal (33%) and styrene (20%). This may be attributed to these components being more difficult to desorb from the sorbent tubes or as a result of using an inaccurate response factor in the calibration. This highlights the need for better sampling methods to improve the state of the art for these components. For the homogeneity analysis, NPL used liquid reference standards to validate the gravimetric loadings. Therefore some of the difference between the gravimetric and reported result may be attributed to the difference between gas and liquid loading. It was found that for the homogeneity analysis the mean difference between the liquid standards and the reported gas loading on the tube was less than 5%
Fig. 3. The median relative difference of all reported results in 2013 comparison versus carbon number.
Fig. 4. The absolute mean relative difference of all reported results for each gas loaded component in 2013 against the absolute difference between the median gravimetric value and the median certified value from validation measurements.
N.D.C. Allen et al. / Measurement 82 (2016) 476–481
for all of the components except styrene (6%), butyl acetate (9%) and hexanal (22%). These challenging components should be where future focus lies in the development of reference standards for indoor air. 4. Conclusions The comparison has given an important insight into the trends in the capability and performance of analytical laboratories for measuring VOCs for indoor air emissions. In 2013 the proportion of laboratories reporting results within one rPA increased relative to the comparison carried out in 2010. Participants involved in both of the comparisons predominantly showed average analytical results closer to the reference value in the 2013 comparison when contrasted to the 2010 comparison, highlighting that the state of the art has improved. For the majority of components, laboratories overestimated the reference values. Further collaborations are imperative to help us to better understand what caused the biases observed and whether it is a function of the analytical technique adopted. More challenging components included oxygenated VOCs; here more work is required to develop indoor air reference standards and sampling techniques. Similarly, new methods and sorbents suitable for analysing different classes of components, are required in an effort to understand the poor performance observed in this comparison for oxygenated VOCs. To ensure that improvements continue to be made it is essential that future comparisons take place. With increasing monitoring of VOCs in indoor air and new legislation being drafted such as prEN16516 (Construction products—Assessment of release of dangerous s ubstances—Determination of emissions into indoor air) studies, such as this comparison, are essential in helping to provide an infrastructure for indoor air quality to support new legislation. Acknowledgements The authors would like to acknowledge the funding of the work by the UK Government’s National Measurement System Chemistry and Biology Knowledge Base Programme. The authors would also like to thank all of the laboratories that participated in the comparison. References [1] P. Wolkoff, G.D. Nielsen, Organic compounds in indoor air—their relevance for perceived indoor air quality?, Atmos Environ. 35 (26) (2001) 4407–4417. [2] J. Sundell, On the history of indoor air quality and health, Indoor Air 14 (Supplementary 7) (2004) 51–58.
481
[3] T. Brennan, Indoor Environmental Quality and Climate Change, US Environmental Protection Agency, Washington, DC, 2010. [4] L. Molhave, G. Clausen, B. Berglund, J. de Ceaurriz, A. Kettrup, T. Lindvall, M. Maroni, A.C. Pickering, U. Risse, H. Rothweiler, B. Seifert, M. Younes, Total volatile organic compounds (TVOC) in indoor air quality investigations, Indoor Air 7 (4) (1997) 225–240, http://dx. doi.org/10.1111/j.1600-0668.1997.00002.x. [5] British Standard, BS EN 482:2012, Workplace exposure – general requirements for the performance of procedures for the measurement of chemical agent. [6] U.S. Environmental Protection Agency, Office of Air and Radiation, Report to congress on indoor air quality, volume II: assessment and control of indoor air pollution, EPA 400-1-89-001C, 1989.