Evaluation of VOC measurements in the EXPOLIS study

Abstract

Personal exposures and microenvironment concentrations of 30 target VOCs were measured for 401 participants living in five European cities as a part of the EXPOLIS (Air Pollution Exposure Distributions within Adult Urban Populations in Europe) study. Measurements in Basel used an active charcoal (Carbotech) adsorbent as opposed to the Tenax TA used in the other study centres. In addition, within each centre, personal and microenvironment VOC sampling required different sampling pumps and, because of different sampling durations, different sampling flow rates. Thus, careful testing of the sampling and analysis procedures was required to ensure accuracy and comparability of collected data. Monitor comparison tests using Tenax TA showed a mean VOC concentration ratio of 0.95 between the personal and microenvironment monitors. The LODs for the target VOCs using Tenax TA ranged from 0.7 to 5.2 µg m⁻³. The LODs for the 14 target compounds quantifiable using Carbotech ranged from 0.9 to 3.2 µg m⁻³. Tenax TA field blanks showed no remarkable contamination with the target VOCs, except benzaldehyde, a known artefact with this adsorbent. Thus, the diffusion barrier system used prevented contamination of Tenax TA samples by passive diffusion during non-sampling periods. Duplicate and parallel evaluations of the Tenax TA and Carbotech showed an average difference of <17% in VOC concentrations within the sampling methods, but a systematic difference between the methods (Tenax TA ∶ Carbotech concentration ratio = 1.18–2.36). These field evaluations and quality assurance tests showed that interpretation and comparison of the results in any VOC monitoring exercise should be done on a compound by compound basis. It is also apparent that carefully planned and realised QA and QC (QA/QC) procedures are needed in multi-centre studies, where a common sampling method and laboratory analysis technique are not used, to strengthen and simplify the interpretation of observed VOC levels between participating centres.

---

Introduction

There are three key issues to be decided before the start of any monitoring campaign for volatile organic compounds (VOCs): the adsorbent, the sampling equipment and the analytical technique. The choice of adsorbent is intimately linked to the analytical technique, and also defines which of the hundreds of the VOCs in air can be quantitatively determined. The sampling technique can be either passive, based on thermal diffusion of the molecules to the medium, or active, pumping a known air volume through the adsorbent. The selected technique has an effect on the weight and complexity of the sampler, the required sampling time and the sensitivity, accuracy and precision of the measurements.

A number of VOC surveys utilising different adsorbents as well as different sampling techniques have been conducted. The VOC TEAM (Total Exposure Assessment Methodology) study, conducted in the 1980s, provided a comprehensive database of the exposure of adult populations to 20 target VOCs for more than 1000 study participants of 10 cities in the US. The full-scale study (Phase II) actively sampled personal exposures and outdoor air concentrations for each participant using Tenax GC adsorbent tubes, which were analysed by gas chromatograph (GC) with mass selective detection (MSD).^1–4 In possibly the broadest indoor air survey, a Canada-wide survey of 757 randomly selected homes,⁵ the samples were collected passively (OVM-3500 badges) and analysed by GC equipped with MSD. In a German Environmental Survey (GerEs II) conducted in 1990/91,⁶ personal exposures to 74 VOCs were measured for 113 participants during one week of passive sampling (OVM-3500 badges), followed by analysis by GC equipped with a flame ionisation detector (FID).

Until recently there has not been a European VOC exposure database that would allow an assessment of cost-effective ways to reduce VOC exposure. Prior to the EXPOLIS study, GerES II provided the only European database of VOC exposure based on representative population samples. Policy decisions, therefore, have been partly based on studies with small, non-representative samples or on data from North America, even though application of chemicals, especially consumer products and pesticides, differs significantly between European countries, as well as between Europe and the US or Canada.

The EXPOLIS study⁷ addresses this lack of information. In the VOC component of the study, personal exposure and microenvironment (residential indoor, residential outdoor and workplace indoor) concentrations of VOCs were investigated for 401 participants living in five European cities (Athens, Basel, Helsinki, Milan and Prague) between September 1996 and March 1998. Although in total up to 323 different compounds were identified in the samples, calibration, quality assurance and quality control (QA/QC) measures were focussed on a core set of 30 target VOCs selected on the basis of their environmental and health significance, and utility as markers of pollution sources. VOC measurements in Basel used an active charcoal sorbent (Carbotech), while Tenax TA was used in other study centres. In addition, different sampling pumps and flow rates were used for personal and microenvironment sampling. Therefore, careful testing of sampling and analysis procedures was required to ensure accuracy and comparability of collected data. This paper describes sampling and analysis of VOCs as a component of the EXPOLIS study, including limits of detection, blind laboratory performance evaluation, Tenax TA–Carbotech method comparison, personal–microenvironment monitor comparison and field blank and duplicate evaluation.

Experimental section

VOC sampling design and methodologies

Active VOC sampling using programmable pumps was selected in the EXPOLIS study, as the sampling design was to collect microenvironment VOC samples in residential or workplace microenvironments only when study participants were present in each microenvironment. Thus, simultaneous sampling for personal exposures and microenvironment concentrations would be comparable, and samples would not be affected by VOC concentrations that occur in the specified microenvironments when the participant was not present. Passive sampling would have entailed the re-capping of VOC badges, which would have been more problematic. The vacuum of the PM_2.5 sampling pump was used to draw the VOC samples via a T-joint (Fig. 1) for both personal and microenvironment sampling.⁷ Some general characteristics of VOC sampling in EXPOLIS including sampling and analysis techniques in each centre, quantification of target compounds, sampling flow rates and sampling volumes are presented in Table 1. Measurements in Basel used active charcoal (Carbotech AG, Basel, Switzerland) adsorbent as opposed to Tenax TA (Chrompack, Middelburg, Netherlands) used in other study centres.

Fig. 1 Layout of PM_2.5 and VOC sampling procedure using a common pump for both samples. VOC sampling flow rates were adjusted in Tenax TA sampling by stainless-steel capillary tubes placed in-line between the sampling pump and the sampling tube. A diffusive flow of VOCs from the air to the sample tube during the non-actively sampled periods was minimised by a diffusion barrier.

Table 1 Some characteristics of Tenax TA and Carbotech VOC sampling in EXPOLIS

Parameter	Tenax TA	Carbotech
a In Helsinki, n = 183 for PEM and n = 477 for MEM. b n = 50 for PEM and n = 133 for MEM.
Used in EXPOLIS	In Athens, Helsinki, Milan and Prague	In Basel
Sampler type	Stainless-steel tube containing 250 mg of Tenax TA adsorbent (60–80 mesh)	Glass tube containing two stacks of active charcoal (25 mg each, 0.05–0.1 mm particle size) stabilised with 4 silver nets
Preparation of sampling tubes before sampling and analysis after sampling	Centrally by VTT Chemical Technology in Espoo, Finland	By Carbotech AG in Basel, Switzerland
Analysis technique	Thermal desorption with GC-MSD/FID analysis	CS2 desorption with GC-MSD analysis
Quantified EXPOLIS target VOCs	All 30 compounds	14 of 30 compounds
Sampling flow rate control	By stainless-steel capillary tube (PEM: 200 × 0.25 mm id; MEM: 320 × 0.25 mm id) placed in line between the pump and the sampling tube	By stainless-steel valve placed in line between the pump and the sampling tube
Mean sampling flow rate/mL min^−1	PEM: 0.81 (s 0.07)^a	PEM: 13.6 (s 2.9)^b
	MEM: 2.07 (s 0.35)^a	MEM: 26.6 (s 9.2)^b
Mean sampling volume/L	PEM: 2.30 (s 0.22)^a	PEM: 39.0 (s 8.3)^b
	MEM: 2.83 (s 0.92)^a	MEM: 38.4 (s 13.1)^b
Mean relative difference between pre- and post-sampling flow rates (%)	PEM: 2.1 (s 10.2)^a	PEM: 1.7 (s 10.5)^b
	MEM: 9.7 (s 16.1)^a	MEM: 31.8 (s 75.7)^b

---

Personal VOC exposures were measured by personal exposure monitors (PEMs).⁷ The PEM equipment (including also a PM_2.5 sampler and a CO monitor), carried by each participant, was packed into an aluminium briefcase. Aluminium was chosen because it is lightweight, durable and free of VOC emissions. Microenvironment VOC concentrations were measured by microenvironment monitors (MEMs) (including also a PM_2.5 sampler), packed into portable sound-absorbent containers made of MDF board.⁷ No significant emissions of any target VOCs were observed when testing the container material at VTT Chemical Technology (Espoo, Finland). MEMs were placed in central representative locations and at least 1 m away from the wall in the sampled microenvironments. Sampling heights were approximately 0.5 m from the floor/ground level and the sampling pumps were programmed to run in the home environments for the non-working hours and at the workplaces for the working hours according to information given by each participant.

Target air sample volumes were 2–3 L for Tenax TA and 30–50 L for Carbotech sampling. VOC sampling flow rate adjustment techniques are presented for both methods in Table 1. Diffusive flow of VOCs from the surrounding air to the Tenax TA sampling tube during the non-sampled periods was minimised by a diffusion-barrier system. The diffusion-barrier consisted of a stainless-steel capillary (200 × 0.50 mm id) placed in-line before the Tenax TA tube (Fig. 1). Airflow through each sampling tube was measured before and after each sampling period with a bubble flow meter (Mini Buck Calibrator M-1, A. P. Buck Inc., Orlando, FL, USA). The sample volume was calculated using the average of the pre- and post-sampling flow rates. Personal exposure sampling times were approximately 48 h, but microenvironment sample collection times varied depending on the schedule of each participant, typically 26–30 h in residential indoor and outdoor environments and 14–18 h in workplaces. Temperature and relative humidity were measured (Extech 445922, Extech Inc., Waltham, MA, USA) when delivering and collecting the equipment to/from the microenvironment sampling locations. All sampling data were recorded on data collection sheets and, subsequently, entered into the EXPOLIS database.⁷

VOC sample analysis and calibration

Preparation of the Tenax TA tubes before sampling and analysis of the tubes after sampling were performed centrally in Finland (Table 1). Initial conditioning of the Tenax TA sampling tubes was performed at 280–300 °C for 2 h using a helium gas flow of 30–50 mL min⁻¹. Subsequent conditioning of tubes at 260 °C for 6 min was performed just before mailing. Clean, tightly sealed sampling tubes were mailed in closed containers prior to sampling from Finland to each study centre by courier.

After sampling the Tenax TA tubes were couriered back to Finland for analysis. VOCs were desorbed from the tubes with helium gas (50 mL min⁻¹) at 260 °C for 6 min into a cold trap. Subsequent flash desorption was followed by 1 ∶ 1 split into two non-polar capillary columns of GC (Hewlett-Packard 5890 Series II+, Hewlett-Packard GmbH, Waldbronn, Germany). One column was connected to FID and the other to MSD (Hewlett-Packard 5972). VOCs were identified from the MSD total ion chromatogram by a Wiley 275 software library. Peaks on the FID chromatograms were identified on the basis of retention times of the pure compounds (high purity standard reference materials). The masses of each target compound were computed using response factors from calibration standards. The response factors of halogenated compounds were calculated from the MSD total ion chromatogram due to their low response in FID. Xylenes (m-, p- and o-) and trimethylbenzenes were quantified using the response factor of toluene. One standard reference tube containing known amounts of pure target compound standard solutions was analysed for every seven samples.

Preparation of the Carbotech tubes before sampling and analysis of the tubes after sampling were performed in Switzerland. Conditioning of the Carbotech tubes was performed by rinsing them twice with 0.5 mL of carbon disulfide (CS₂) and drying for 10 min under a pure stream of nitrogen prior to use. After sampling, 6 µL of an internal standard (IS) solution of 1-chloro-n-hexane (250 ng µL⁻¹) was added with a syringe to both sections of active charcoal. VOCs were desorbed from each section of the Carbotech tubes with two sequential rinses of CS₂ (250 µL and 50 µL, respectively). The two rinses were combined into one vial. The vials were closed and stored at 4 °C until analysis (in 1–7 d). Analysis was made by GC (MFC 500, Carlo Erba Instruments, Milan, Italy) equipped with an autosampler (A200S, Zwingen, Switzerland) and a tandem capillary column (injector side: HP5, detector side: Innowax). The tandem column was connected to MSD by a 0.5 m deactivated fused silica (0.2 mm id) transfer-line. Respective peaks were quantified by linear regression using calibration curves based on eight calibration solutions containing known amounts of pure standard solutions and internal standard (diluted from stock solutions). A separate calibration curve was run for every 10–20 Carbotech samples. The VOCs were identified from the MSD ion chromatogram by reference spectra and retention time. To accept the result, the quality of the match had to exceed 500 (ITD fit).

Method detection limits

A common definition of the analyte concentration giving a signal level (y_LOD) equal to the blank signal (y_B) plus three standard deviations (s_B) of the blank was used to define the limit of detection (LOD) of each target VOC for Tenax TA samples:⁸When the FID response was treated as a dependent variable and concentration as an independent variable the intercept from the calibration standard solutions was used as an estimate of analytical noise (y_B). The standard error of the regression line was used as the estimate of the standard deviation of the blank (s_B). The LOD was the x value for y = y_LOD. The LODs for the Carbotech samples were reported as the value of the lowest detected calibration standard by the laboratory performing the analysis. A value of half the respective LOD for each method was used in statistical analyses for samples that were below the LOD.⁹

Blind performance evaluation

Blind performance evaluation of the VOC measurements using Tenax TA by VTT Chemical Technology together with 17 other laboratories was conducted previously as part of the “Second inter-laboratory comparison of small chamber measurements”.¹⁰ Participants brought at least two samplers, one of which was used as a blank. The other was loaded with air from a 0.28 m³ chamber, which had been loaded with painted plates the previous day. Loaded samples were analysed by the participants in their respective laboratories—including VTT. Samples were analysed for 2-(2-butoxyethoxy)ethanol and two monoisobutyric esters of 2,2,4-trimethylpentane-1,3-diol. The isomer esterified in position 3 was named in evaluation as Texanol-1 and the one esterified in position 1 as Texanol-2. More detailed methodologies and results may be found in De Bortoli et al.¹⁰

Separate blind performance evaluation of the VTT analysis laboratory was performed as part of the EXPOLIS QA/QC procedures. Six clean Tenax TA sampling tubes were mailed to EC-JRC Environment Institute/Air Quality Unit (Ispra, Italy), two for each spiking concentration and two for blanks. 1 µL of a liquid mixture of six compounds (toluene, hexanal, tetrachloroethene, 2-ethylhexan-1-ol, limonene and naphthalene) in methanol was injected into each tube (excluding blanks). The methanol solvent was evaporated from the sample tubes with 1 L of high purity He. Two different concentrations of each compound were injected into the separate sampling tubes, the first reflected concentrations that were much higher than expected in the EXPOLIS field samples, the second represented estimated field sample concentrations. After loading tubes were mailed back to VTT for analysis. The whole procedure took less than one week.

Tenax TA–Carbotech comparison

Tenax TA and Carbotech methods were compared in Helsinki to estimate the comparability of field results (including all the field procedures). In this comparison, seven parallel 24 h microenvironment samples were collected using Tenax TA and Carbotech in an office building in downtown Helsinki. Collection, storage, shipment and analysis were performed according to normal EXPOLIS VOC field procedures. Carbotech tubes were mailed to/from Helsinki separately for each parallel measurement.

PEM–MEM comparison

The comparability of personal and microenvironment VOC monitors (Tenax TA only) was tested in a 45 h study in an office building in downtown Helsinki. In the comparison, three PEMs and ten MEMs with Tenax TA tubes were run in parallel.

Field blanks and duplicates

Field blanks were collected for a target of 10% of all VOC samples. Similarly, target for field duplicate samples was 10% of all samples. MEM duplicates were randomly assigned to study participants, but due to practical constraints duplicate PEM measurements were performed by researchers carrying two PEMs simultaneously.

Results and discussion

VOC sampling design and methodologies

Table 2 shows the theoretical VOC contamination with and without the diffusion-barrier in a typical EXPOLIS microenvironment (workplace) monitoring arrangement (Tenax TA). In these conditions the diffusion-barrier reduces the diffusive flow by 99.75% compared to its absence and to 0.05% of the actively collected VOC mass.

Table 2 Theoretically calculated actively sampled and passively diffused^a VOC masses for typical EXPOLIS Tenax TA microenvironment (workplace) sampling arrangement without and with diffusion-barrier used

		Actively sampled VOC mass/ng	Passively diffused VOC mass/ng	Passively diffused/actively sampled (%)
a The passive VOC transfer through a layer of gas is calculated from Fick's law assuming: VOC concentration in air = 500 µg m^−3, active VOC sampling flow rate = 2 mL min^−1, active sampling time = 16 h (2 × 8 h), passive sampling time = 40 h, and diffusion coefficient = 0.07 cm^2 s^−1.
Without diffusion barrier—
Tube:	l = 50 mm, lid = 5 mm	960.0	197.9	20.61
With diffusion barrier—
Tube:	l = 200 mm, id = 0.5 mm	960.0	0.5	0.05

---

The mean VOC sampling flow rates and volumes are presented both for Tenax TA (in Helsinki) and Carbotech sampling in Table 1. Other centres using Tenax TA reported similar flows and volumes as in Helsinki. The smaller volume collected with Tenax TA in the personal measurements may have slightly increased the number of samples with non-detections. In practice, however, the number of samples in which compounds were detected showed good agreement between personal and indoor samples in both Tenax TA and Carbotech (Table 3).

Table 3 EXPOLIS target compound list and percentage of samples above the sample detection limit (LOD)

Compound	CAS number	Tenax TA/Helsinki					Carbotech/Basel
		LOD/ng	Samples above LOD (%)				LOD/ng	Samples above LOD (%)
			Personal (n = 183)	Indoor residential (n = 181)	Outdoor residential (n = 156)	Workplace (n = 140)		Personal (n = 50)	Indoor residential (n = 47)	Outdoor residential (n = 47)	Workplace (n = 39)
a Cannot be determined with Carbotech method. b Only qualitative analysis with Carbotech method.
Alkanes—
Hexane	110-54-3	7.46	15	9	12	18	39	42	34	28	31
Nonane	111-84-2	2.23	67	76	45	38	40	62	55	0	39
Decane	124-18-5	3.77	70	75	16	38	36	96	89	28	67
Cyclohexane	110-82-7	6.27	13	13	2	4	—^a
Undecane	1120-21-4	2.16	89	91	20	54	42	88	81	0	51
Aromatics—
Benzene	71-43-2	2.89	83	68	76	61	72	70	60	28	44
Toluene	108-88-3	6.05	99	100	80	88	128	100	100	79	95
Ethylbenzene	100-41-4	3.36	79	82	26	48	61	80	68	26	33
m(p)-Xylene	108-38-3	6.05	98	97	58	75	80	100	98	79	87
o-Xylene	95-47-6	6.05	42	48	12	22	47	94	79	40	54
Styrene	100-42-5	3.40	19	27	3	6	—^b
Naphthalene	91-20-3	2.92	4	10	1	1	—^a
Propylbenzene	103-65-1	1.78	36	37	10	16	46	26	19	0	13
Trimethylbenzenes	95-63-6	6.05	53	57	16	27	120	84	66	23	49
Alcohols—
2-Methylpropan-1-ol	78-83-1	3.68	45	70	3	18	—^a
Butan-1-ol	71-36-3	3.80	85	92	5	38	—^a
2-Ethylhexan-1-ol	104-76-7	7.16	37	49	3	16	—^b
Phenol	108-95-2	7.94	4	9	3	6	—^a
Octan-1-ol	111-87-5	5.37	0	2	0	0	—^a
Esters—
2-Butoxyethanol	111-76-2	7.23	12	23	0	8	—^b
Alkanals/Alkanons—
Hexanal	66-25-1	12.93	60	85	1	16	—^b
Benzaldehyde	100-52-7	2.65	89	95	83	86	—^b
Octanal	124-13-0	8.33	58	71	6	9	—^b
Halogenated—
Trichloroethene	79-01-6	4.34	4	2	0	4	47	26	21	2	13
Tetrachloroethene	127-18-4	4.00	8	7	0	5	66	22	17	6	18
1,1,2-Trichloroethane	79-00-5	4.34	0	1	0	0	43	0	0	0	0
Miscellaneous—
D-Limonene	138-86-3	3.55	95	96	5	57	—^b
1-Methyl-2-pyrrolidinone	872-50-4	11.51	2	1	0	1	—^b
3-Carene	13466-78-9	2.06	66	75	14	24	—^b
α-Pinene	80-56-8	8.23	71	89	12	25	—^b

---

Flow resistance of the PM_2.5 filters increased with loading of particles during the sampling. Consequently, there was a potential for the flow through the VOC sampling tubes to increase towards the end of each sampling period, in microenvironment measurements, as the mass flow controlled pump drawing air for both PM_2.5 and VOC samples was compensating for increasing flow resistance. Thus, it was possible for VOC concentrations to be somewhat over represented towards the end of each sample relative to the beginning. Measured differences in pre- and post-sampling flow rates for microenvironment measurements, however, were relatively small (Table 1) with both Tenax TA and Carbotech due to the low PM_2.5 masses collected. Personal exposure measurements used a volumetric flow controlled pump, which kept the pump speed constant as the particle load on PM_2.5 filter increased. Consequently, flow rate changes for personal samples were minimal both with Tenax TA and Carbotech as personal filters were loaded during the sampling period (Table 1).

Method detection limits

Table 3 shows the CAS numbers, chemical grouping, defined sample LODs and the percentages of samples above the LODs for both Tenax TA samples in Helsinki and Carbotech samples in Basel. For Tenax TA, the defined LODs ranged from 1.78 (propylbenzene) to 12.93 ng (hexanal) (equivalent to 0.7–5.2 µg m⁻³ with an assumed sampling volume of 2.5 L). For Carbotech LODs ranged from 36 (decane) to 128 ng (toluene) (equivalent to 0.9–3.2 µg m⁻³ with an assumed sampling volume of 40 L) for the 14 target compounds that could be quantified with the method. As a comparison, LODs for 26 target VOCs ranged between 1.6 and 5.9 µg m⁻³ in the study by Fellin and Otson (using OVM-3500 badges).⁵ In the VOC TEAM study, LODs were reported to be generally on the order of 1 µg m⁻³.^1,2,4,11

In our study, the number of samples in which specific VOC compounds were detected should be compared cautiously and on a compound by compound basis as the relative sensitivities of the Tenax TA and Carbotech methods depend on the compound analysed. Thus, higher percentage detection in samples could refer to lower detection limits for specific compounds or increased levels in the environment.

Blind performance evaluation

Measurement of 2-(2-butoxyethoxy)ethanol by VTT at two different concentrations using Tenax TA tubes were within 20% of mean concentrations measured by 10 participating laboratories (using both Tenax TA and active charcoal tubes, and both FID and MSD analyses), in the “Second inter-laboratory comparison of small chamber measurements”.¹⁰ Similarly, VTT measurements of Texanol-1 were within 12%, and Texanol-2 within 10% of mean concentrations measured by participating laboratories. Results obtained with charcoal adsorption and solvent elution were consistently lower than those obtained by Tenax TA adsorption and thermal elution.

The results of the separate EXPOLIS blind performance evaluation for Tenax TA deviated from applied concentrations by more than 10% only for hexanal (−13.2%) in the concentrated solution and for toluene (+11.0%), hexanal (−10.4%), 2-ethylhexan-1-ol (+10.6%) and tetrachloroethene (+23.8%) in the dilute solution. The relative percent differences (RPDs) between the duplicate samples were less than 10% both for the concentrated and the dilute solution.

Tenax TA–Carbotech comparison

The results of the field comparison of Tenax TA and Carbotech for those target compounds that were quantified by both methods are shown in Table 4. Median Tenax TA ∶ Carbotech concentration ratios ranged from 1.18 (ethylbenzene) to 2.36 (undecane) with a mean of 1.53 (s 0.47) for nine compounds with at least two parallel measurements above LODs in both samples. Thus, the Carbotech method gave VOC concentrations that were systematically lower than those of the Tenax TA method.

Table 4 Tenax TA–Carbotech and PEM–MEM method comparisons

Compound	Tenax TA–Carbotech comparison (n = 7)		PEM–MEM comparison with Tenax TA (n = 3 PEMs and 10 MEMs)
	n ^a	Median Tenax TA ∶ Carbotech concentration ratio	PEM ∶ MEM concentration ratio (ratio of means)
a Number of pairs where compound found above detection limit in both Tenax TA and Carbotech sample. b Cannot be determined with Carbotech method. c Only qualitative analysis with Carbotech method. d ND = not determined above LOD in all PEM and MEM samples.
Hexane	7	2.31	0.96
Nonane	0	—	0.84
Decane	0	—	0.88
Cyclohexane	—^b		1.07
Undecane	2	2.36	0.88
Benzene	5	1.51	0.91
Toluene	7	1.32	0.87
Ethylbenzene	7	1.18	0.89
m(p)-Xylene	7	1.26	0.93
o-Xylene	7	1.24	0.80
Styrene	—^c		ND^d
Naphthalene	—^b		ND
Propylbenzene	3	1.21	0.86
Trimethylbenzene	7	1.36	0.86
2-Methylpropan-1-ol	—^b		ND
Butan-1-ol	—^b		ND
2-Ethylhexan-1-ol	—^c		ND
Phenol	—^b		ND
Octan-1-ol	—^b		ND
2-Butoxyethanol	—^c		ND
Hexanal	—^c		ND
Benzaldehyde	—^c		0.57
Octanal	—^c		1.85
Trichloroethene	0	—	ND
Tetrachloroethene	0	—	ND
1,1,2-Trichloroethane	0	—	ND
D-Limonene	—^c		ND
1-Methyl-2-pyrrolidinone	—^c		ND
3-Carene	—^c		ND
α-Pinene	—^c		1.08

---

PEM–MEM comparison

The results of PEM–MEM comparison study for target compounds with concentrations above LODs in all PEM and MEM samples are presented in Table 4. For these 15 VOCs the mean concentrations measured by personal monitors are compared to mean concentrations measured by microenvironment monitors. The PEM ∶ MEM concentration ratio ranged from 0.57 (benzaldehyde) to 1.85 (octanal) with a mean of 0.95 (s 0.28). The results of the study showed that PEM–MEM comparisons for octanal and especially benzaldehyde, which is a known artefact with Tenax TA as a result of reaction of the sorbent with strong oxidising agents,¹² should be avoided. The PEMs typically slightly underestimated the concentrations relative to the MEMs for other compounds in these parallel tests. It must be noted, however, that the MEMs ran for a considerably longer period of time and sampled larger sample volumes (mean 5.8 L) in this comparison than in the microenvironment monitoring in the EXPOLIS study.

Field blanks and duplicates

The median field blank contamination of the Tenax TA (n = 74; Helsinki: n = 60, Athens: n = 9 and Prague: n = 5) and Carbotech (n = 40) samples were below the LOD for all quantifiable target compounds except hexane in Carbotech. In the case of the Carbotech sampling of hexane, the median contamination was 43 ng, the equivalent of 1.1 µg m⁻³ with assumed sampling volume of 40 L. The 95th percentile of the field blank contamination was below the LOD for all quantifiable target compounds except benzaldehyde (5.5 ng, the equivalent of 2.2 µg m⁻³ with assumed sampling volume of 2.5 L) in Tenax TA sampling. The low contamination levels in the Tenax TA field blanks showed that the diffusion-barrier system successfully prevented sample contamination.

Previous studies have also reported low background concentrations of VOCs in cleaned Tenax TA sampling tubes.^13,14 De Bortoli et al.¹³ reported benzene and toluene contamination to contribute most to the background in Tenax TA sorbent. In EXPOLIS, the background levels of these compounds were negligible and did not impair the measurement of these compounds in the study. Other studies have shown artefact formation in Tenax sampling tubes for benzaldehyde, acetophenone and phenol.^15–19 In EXPOLIS, Tenax TA field blanks did not demonstrate consistent artefact formation across samples for phenol, and acetophenone was not among our target compounds. Benzaldehyde contamination was found on 38% of the field blanks above LOD, however, and care should be taken in interpreting these results. Hexane showed consistent contamination (in 83% of the field blanks above LOD) with the Carbotech method. For this reason the results were corrected by subtracting median field blank levels.

The relative percent differences between duplicate samples for the target VOCs exceeding LODs are summarised in Table 5. For Tenax TA, the median RPD for the PEM duplicates (n = 15; Helsinki: n = 12 and Athens: n = 3) ranged from 2.4 (3-carene) to 30.3% (2-ethylhexan-1-ol) (mean 11.4%, s 7.7%). The median RPD for the MEM duplicates (n = 51; Helsinki: n = 40, Athens: n = 5 and Prague: n = 6) ranged from 3.2 (2-butoxyethanol) to 54.3% (phenol) (mean 11.6%, s 10.7%). For Carbotech, the median RPD for the MEM duplicates (n = 11) ranged from 7.6 (ethylbenzene) to 31.4% (undecane) (mean 16.4%, s 9.0%). The precision of duplicate measurements for both Tenax TA and Carbotech was comparable to previous VOC exposure studies using Tenax GC, i.e., around 5–30%.^{1–4,11,20–24}

Table 5 Median relative percent difference (RPD%)^a for duplicate samples

Compound	Tenax TA				Carbotech
	Personal (n = 15)		Microenvironment (n = 51)		Personal (n = 1)		Microenvironment (n = 11)
	n ^b	Median	n ^b	Median	n ^b	Median	n ^b	Median
a Calculated as [|value1 − value2| / (value1 + value2) / 2] × 100. b Number of pairs where compound found above detection limit in both samples. c Cannot be determined with Carbotech method. d Only qualitative analysis with Carbotech method.
Hexane	6	2.8	12	9.5	1	48.4	7	27.4
Nonane	7	11.4	30	10.7	0	—	0	—
Decane	9	11.5	28	7.0	0	—	3	18.0
Cyclohexane	8	11.0	15	3.6	—^c		—^c
Undecane	11	6.7	31	6.0	0	—	2	31.4
Benzene	15	9.0	36	7.3	0	—	5	14.9
Toluene	15	7.2	48	4.8	1	22.7	8	8.0
Ethylbenzene	14	6.4	29	4.2	0	—	5	7.6
m(p)-Xylene	15	5.6	46	8.4	1	31.3	9	12.3
o-Xylene	14	8.8	31	6.6	0	—	6	10.0
Styrene	5	10.2	15	11.4	—^d		—^d
Naphthalene	5	5.4	12	12.5	—^c		—^c
Propylbenzene	6	7.1	16	7.5	0	—	1	11.9
Trimethylbenzene	13	10.2	30	12.5	0	—	5	9.5
2-Methylpropan-1-ol	1	27.3	21	8.4	—^c		—^c
Butan-1-ol	4	17.0	24	6.1	—^c		—^c
2-Ethylhexan-1-ol	3	30.3	16	12.4	—^d		—^d
Phenol	0	—	3	54.3	—^c		—^c
Octan-1-ol	0	—	1	4.3	—^c		—^c
2-Butoxyethanol	2	25.4	4	3.2	—^d		—^d
Hexanal	7	3.1	28	20.3	—^d		—^d
Benzaldehyde	14	14.7	41	13.7	—^d		—^d
Octanal	9	22.5	25	8.7	—^d		—^d
Trichloroethene	0	—	6	36.4	0	—	1	29.7
Tetrachloroethene	1	10.4	8	11.2	0	—	0	—
1,1,2-Trichloroethane	0	—	0	—	0	—	0	—
D-Limonene	8	13.8	27	8.3	—^d		—^d
1-Methyl-2-pyrrolidinone	0	—	0	—	—^d		—^d
3-Carene	5	2.4	17	18.0	—^d		—^d
α-Pinene	11	5.4	33	6.4	—^d		—^d

---

Conclusions

Active sampling on active charcoal (Carbotech) (CS₂ desorption with GC-MSD analysis) and active sampling on Tenax TA (thermal desorption with GC-MSD/FID analysis) had sufficient sensitivity and selectivity for the basic needs of the EXPOLIS study. The Carbotech method, however, quantified only 14 of the 30 target compounds, and had lower precision and accuracy. Higher analytical detection limits with Carbotech (ng sample⁻¹) were compensated by higher sampling rates. Thus, the LODs (µg m⁻³ of air) were comparable for both methods, i.e., 0.7–5.2 µg m⁻³, depending on the method and compound, and to a lesser degree, on the sample.

In the Tenax TA–Carbotech field comparison, Carbotech showed systematically lower levels than Tenax TA, with observed median Tenax TA ∶ Carbotech concentration ratios ranging from 1.18 (ethylbenzene) up to 2.36 (undecane). Comparison of the results of the two methods should be done with caution considering these differences. Comparison between personal and microenvironment monitors using Tenax TA gave average concentration ratios from 0.80 to 1.08 for all detected target VOCs, except benzaldehyde (0.57), a known Tenax TA artefact, and octanal (1.85). Based on this, PEM–MEM comparisons for these two compounds should be avoided. The median RPD of the observed concentration differences between duplicate samples in the Tenax TA method was below 15% for most of the target compounds and exceeded 30% only for 2-ethylhexan-1-ol (PEM), phenol (MEM) and trichloroethene (MEM). In the Carbotech samples the median RPDs for MEM duplicates were somewhat higher and for PEM only one duplicate was made.

According to our tests and QA/QC results, the Tenax TA method was superior to the Carbotech method in the context of the EXPOLIS study. Field evaluations and quality assurance test showed that comparison of VOC results obtained by different sampling and analysis methods and analysed in different laboratories should be done on a compound-by-compound basis. It is also apparent that carefully planned and realised QA/QC procedures are needed in multi-centre studies to strengthen and simplify interpretation and comparison of observed VOC levels between participating centres.

Acknowledgements

This work is a part of European Union 4th Framework RTD Program funded multi-centre study; Air Pollution Exposure Distributions of Adult Urban Populations in Europe (EXPOLIS). EXPOLIS was supported by the following contracts: EU contract ENV4-CT96-0202, and ERB IC20-CT96-0061 (Prague); Academy of Finland contract No. 36586 (Helsinki); Swiss Ministry for Education and Science contract BBW No. 95.0894 (Basel); French National Environment Agency (ADEME); Union Routière de France and Grenoble Communauté de Communes (Grenoble); other national research funds; and intramural funding from the participating institutes.

References

1. L. A. Wallace, E. D. Pellizzari, T. D. Hartwell, R. Whitmore, C. Sparacino and H. Zelon, Environ. Int., 1986, 12, 369.
2. T. D. Hartwell, E. D. Pellizzari, R. L. Perrit, R. W. Whitmore, H. S. Zelon, L. S. Sheeldon and C. M. Sparacino, Atmos. Environ., 1987, 21, 1995.
3. L. A. Wallace, E. D. Pellizzari, T. D. Hartwell, C. Sparacino, R. Whitmore, L. Sheldon, H. Zelon and R. Perritt, Environ. Res., 1987, 43, 290.
4. L. Wallace, W. Nelson, R. Ziegenfus, E. Pellizzari, L. Michael, R. Whitmore, H. Zelon, T. Hartwell, R. Perritt and D. Westerdahl, J. Exposure Anal. Environ. Epidemiol., 1991, 1, 157.
5. P. Fellin and R. Otson, Atmos. Environ., 1994, 28, 3581.
6. K. Hoffmann, C. Krause, B. Seifert and D. Ullrich, J. Exposure Anal. Environ. Epidemiol., 2000, 10, 115.
7. M. J. Jantunen, O. Hänninen, K. Katsouyanni, H. Knöppel, N. Kuenzli, E. Lebret, M. Maroni, K. Saarela, R. Sram and D. Zmirou, J. Exposure Anal. Environ. Epidemiol., 1998, 8, 495.
8. J. C. Miller and J. N. Miller, Statistics for Analytical Chemistry, Ellis Horwood, Chichester, 1986, pp. 96–100..
9. R. W. Hornung and L. D. Reed, Appl. Occup. Environ. Hyg., 1990, 5, 46.
10. M. De Bortoli, S. Kephalopoulos and H. Knöppel, European Collaborative Action ‘Indoor Air Quality and Its Impact On Man’ Determination of VOCs Emitted From Indoor Materials And Products. Second Interlaboratory Comparison Of Small Chamber Measurements, European Commission JRC Environment Institute, Ispra, Italy, 1995, Report 16 EUR 16284 EN..
11. T. D. Hartwell, R. L. Perritt, E. D. Pellizzari and L. C. Michael, Atmos. Environ., Part A, 1992, 26A, 1519.
12. R. G. Lewis and S. M. Gordon, in Principles Of Environmental Sampling, ed. L. H. Keith, ACS, Washington, DC, 1996, pp. 401–470..
13. M. De Bortoli, H. Knöppel, E. Pecchio, H. Schauenburg and H. Vissers, Indoor Air, 1992, 2, 216.
14. R. H. Brown, Analyst, 1996, 121, 1171.
15. R. S. Hutte, E. J. Williams, J. Staehelin, S. B. Hawthorne, R. M. Barkley and R. E. Sievers, J. Chromatogr., 1984, 302, 173.
16. E. Pellizzari, B. Demian and K. Krost, Anal. Chem., 1984, 56, 793.
17. J. F. Walling, J. E. Burngarner, D. J. Driscoll, C. M. Morris, A. E. Riley and L. H. Wright, Atmos. Environ., 1986, 20, 51.
18. H. Rothweiler, P. A. Wäger and C. Schlatter, Atmos. Environ., Part B, 1991, 25B, 231.
19. X.-L. Cao and C. N. Hewitt, Environ. Sci. Technol., 1994, 28, 757.
20. L. A. Wallace, E. D. Pellizzari, T. D. Hartwell, V. Davis, L. C. Michael and R. W. Whitmore, Environ. Res., 1989, 50, 37.
21. C. C. Chan, S. H. Lin and G. R. Her, Air Waste, 1993, 43, 1231.
22. C. C. Chan, S. H. Lin and G. R. Her, Atmos. Environ., 1994, 28, 2351.
23. N. J. Lawryk and C. P. Weisel, Environ. Sci. Technol., 1996, 30, 810.
24. S.-O. Baek, Y.-O. Kim and R. Perry, Atmos. Environ., 1997, 31, 529.

---