Fresenius J Anal Chem (2001) 371 : 798–805 DOI 10.1007/s002160101069
S P E C I A L I S S U E PA P E R
Erwin Rosenberg · Rainer A. Hallama · Manfred Grasserbauer
Development and evaluation of a calibration gas generator for the analysis of volatile organic compounds in air based on the injection method Received: 25 April 2001 / Revised: 23 July 2001 / Accepted: 28 July 2001 / Published online: 12 October 2001 © Springer-Verlag 2001
Abstract The development and operational evaluation of a calibration gas generator for the analysis of volatile organic compounds (VOC) in air is described. Details of the construction, as well as of the evaluation of the apparatus are presented here. The performance of the test gas generator is validated both by on-line GC analysis of the calibration gas produced and by off-line analysis of adsorptive samples taken from the generated calibration gas. Both, active and passive sampling have been used, and the results demonstrate the excellent accuracy and precision of the generated test gas atmosphere: For the 11 investigated organic compounds (aromatic and halogenated compounds), the found values were in most cases within 5% of the target value with a reproducibility of better than 3% RSD (as determined by the analysis of the sampled adsorbent tubes). Custom made adsorbent tubes were used for active and passive sampling and in both cases were analysed by thermal-desorption GC. Particularly the combination of passive sampling and thermodesorption-GC analysis offers significant advantages over the commonly used active sampling on activated charcoal, followed by CS2 desorption in terms of avoidance of hazardous solvents, potential for automation and improved detection limits. Both sampling techniques are capable for monitoring VOCs at concentrations and under conditions relevant for workplace monitoring.
Introduction Volatile organic compounds (VOCs) represent important trace constituents of natural, urban, industrial and work-
Dedicated to Professor Dr. Bernd Neidhart on the occasion of his 60th birthday E. Rosenberg (✉) · R.A. Hallama · M. Grasserbauer Institute of Analytical Chemistry, Vienna University of Technology, Getreidemarkt 9/151, 1060 Vienna, Austria e-mail:
[email protected]
place atmospheres. Their concentration may vary from a few ng m–3 to several µg m–3 or even mg m–3, depending on the actual case. Since the knowledge of their temporal and spatial distribution, as well as their individual and sum concentration is important to assess both their effect on man (e.g. in the case of workplace monitoring) as well as their environmental impact, analytical techniques are required to accurately determine VOCs at these concentration levels [1, 2, 3]. The accuracy of gas concentration measurement results, however, is inevitable linked to the possibility of accurately calibrating the respective instrumentation. The most commonly employed procedure to do so is to generate a test gas atmosphere that contains precisely known amounts or concentrations of individual VOCs in a suitable zero gas (usually nitrogen or hydrocarbon-free air). Many methods have been developed over the past years for the production of controlled test gas atmospheres [4, 5], and they all have particular advantages and limitations which have to be considered in each single case of application. It is beyond the scope of this introduction to discuss these techniques which may generally be classified in static methods (dilution of a concentrated standard or a pure compound to a defined gas volume [6, 7, 8]) and dynamic methods (such as the blending of calibration gases [9], the injection method, the permeation method [10, 11, 12], the diffusion method [13, 14, 15] and the saturation method [16, 17, 18, 19]) in detail. For a more detailed discussion on this topic, the reader is referred to the literature [4, 5, 20, 21, 22]. The generation of test gas atmospheres has been regulated by several national standards, such as, e.g., the German standard VDI 3940 [23]. We report here on the construction of a test gas generator based on the injection method and present data for its validation with a set of organic compounds that are relevant for both environmental monitoring (with concentrations usually in the low µg m–3 range) as well as for workplace monitoring where concentrations usually met are rather in the mg m–3 range. In view of the wide concentration range that must be covered by a test gas generator, an
799
Fig. 1 Schematic view of the self-built injection apparatus. Gas tank with synthetic air (1), needle valve (2), rotameter for monitoring the total gas stream (3), needle valve and rotameter for the dry gas stream (4), needle valve and rotameter for the humidified gas stream (5), humidifier (6), hygrometer (7), heated injection manifold (stainless steel tee sealed with a rubber septum (8)), injection syringe in syringe holder (9), micrometer drive (10), step motor (11), step motor control (12), mixing chamber (13), chamber with sampling ports (14), air flow calibrator (15), stainless steel capillary to the on line-GC (16)
apparatus working according to the injection principle is particularly advantageous.
Experimental Description of the device A schematic representation of the test gas generator is given in Fig. 1. The individual components are discussed in their order in the flow path of the test gas. The zero gas is supplied from a compressed gas cylinder of nitrogen (Messer Griesheim, Austria, purity >99.999%) and regulated by a needle valve (Hoke Micro-mite, Hoke, obtained through Burde, Vienna, Austria). Using a manually actuated needle valve was found to produce a more constant flow than when using a mass flow controller (range 0 – 2 L min–1; Aalborg, Denmark) which was drifting significantly during operation. The zero gas stream was measured by a rotameter (range 20– 200 NL h–1, Krohne DK 48/N, obtained through SMK, Vienna, Austria). The connection between the gas cylinder and the rotameter was 1/8’’ copper tubing, after this first rotameter transparent PVC tubing was used. In order to be able to accomplish different relative humidities in the test gas, the total gas flow was split into two parts of which one was led through a water-filled saturation chamber while the second part of the gas stream by-passed it. The dimensions of the saturation chamber and the dwell time of the zero gas were sufficiently large so that the gas left the saturation chamber with close to 100% relative humidity (RH). The RH of the test gas was adjusted by setting the relative ratios of the dry and the vapour-saturated gas flow, but was also measured subsequently by an electronic hygrometer. The humidified zero gas stream is directed into the injection manifold. This is basically a stainless steel 1/4 T-piece (Swagelok, obtained by Vaust, Gablitz, Austria) with a short piece of copper tubing. Two arms of the T-piece are connected to the copper and PVC tubing, respectively, through which the calibration gas flows. The third arm of the T-piece is sealed with a silicone rubber septum through which the syringe needle is pierced. The T-piece and the copper tubing are heated to 130 °C by a heating tape (2 m of 45 W/m; Horst Laborgeräte, Lorsch, Germany) in order to support the evaporation process of the liquid standards that are injected into the gas stream. The temperature is controlled via a temperature controller (HT 100, Horst Laborgeräte) and a PT 100 thermocouple that is inserted between the heating tape and the T-piece. The actual injection device consists of an in-house-built manifold and drive that can accommodate commercial gas-tight syringes. Different fittings have been
made from aluminium to house the syringes of different size (from 10 to 2500 µL volume). For the generation of a test gas, the syringes have to be mounted in the particular fitting, and the needle of the syringe has to be inserted in the T-piece through the septum and brought to such a position that the needle tip is located exactly in the gas stream passing through the T-piece. The syringe is then actuated by a motor-driven micrometer screw. This micrometer screw is driven by a stepper motor that makes 100 steps per revolution, where one revolution moves the micrometer screw exactly 0.5 mm forward. That means that for a stroke length of 50 mm (which is typical for most commercial syringes) 10000 steps of the stepper motor are required. The stepper motor is controlled via a custom made controller that is interfaced to the serial I/O interface of a personal computer. A self-written QBasic program allows to control all relevant functions via the PC (start/stop, injection speed, reversal of movement, etc.). For limitations lying in the software, the maximum duration between the steps of the stepper motor was ca. 8 seconds which means that the volume of the chosen syringe may be introduced over a maximum period of 80000 s, corresponding to more than 22 h which safely allows to simulate workplace sampling, e.g. during an 8 h-working shift. After the injection manifold, the test gas passes a homogenisation chamber (V=approx. 1 L) before entering the sampling manifolds (one of smaller dimensions intended for active sampling, and one of larger diameter where passive samplers can be exposed). After leaving the sampling manifolds, the gas stream is controlled by a gas flow calibrator (Mini-Buck, M-5, obtained through Supelco, Austria). By controlling precisely the parameters of test gas generation, it was thus possible to calculate test gas concentrations from the amount of substance injected per unit of time (known exactly from the liquid standard composition and the injection speed) and the gas flow through the injection apparatus. Thus, ‘true values’ could be assigned and recovery rates be calculated. The stability of the calibration gas was controlled by an on line-GC. This was achieved by a type 6890 GC (Hewlett–Packard, Palo Alto, CA, USA) which was equipped with a flame ionisation detector (FID) and connected to the sampling manifold by a 1/8 stainless steel capillary. The gaseous sample was introduced through a heated six-port-valve with a 0.5 mL sampling loop. Prior to each GC measurement, the test gas was sucked by a membrane pump through the capillary and the sample loop, then the pump was switched off and the valve actuated to transfer the test gas to the GC column head. Although absolute calibration of the detector was not necessary to assess the stability of the generated test gas, calibration was carried out with a propane calibration gas (19.35 mg m–3 in N2, supplied by Messer Griesheim, Austria) that was also introduced into the sampling chamber of the calibration gas apparatus. GC runs were started every 20 min during the operation of the test gas generator. Materials Test substances The substances investigated in this work are listed in Table 1. Name, formula, molecular weight, threshold limit values (TLV) in Austria [24] (which are for this range of compounds basically the same as the German MAK values [25] except for 1,4-dichloroben-
800 Table 1 List of substances investigated in this work (TLV, threshold limit value in Austria [24]; Bp, boiling point; D, diffusion coefficient) Compound
2-Bromo-2-chloro-1,1,1-trifluoroethane 1,4-Dichlorobenzene (solid, m.p.=53 °C) Dichloromethane Tetrachloroethene 1,1,1-Trichloroethane Trichloroethene Benzene Ethylbenzene Styrene Toluene m-Xylene o-Xylene p-Xylene
Formula
F3C-CHClBr C6H4Cl2 CH2Cl2 Cl2C=CCl2 CH3CCl3 ClHC=CCl2 C6H6 C6H5C2H5 C6H5CH=CH2 C6H5CH3 C6H4(CH3)2 C6H4(CH3)2 C6H4(CH3)2
zene), boiling point and diffusion coefficient (at 25 °C) are given in this table [26]. The substances are obtained from Merck, Aldrich and Fluka with purities of more than 99% in all cases. Adsorbent tubes All adsorbent tubes used were custom made of Pyrex glass and had an outer diameter of 6 to 6.2 mm, a wall thickness of ca. 1 mm and a length of about 12 cm. They were filled with the adsorbents up to a bed length of 8 cm. Silanized glass wool plugs were used to hold the adsorbents in place in the tubes. The tubes were sealed with brass caps (Swagelok, 1/4) and Teflon ferrules (Swagelok, 1/4″) or Vespel-Ferrules (Supelco, 1/4). For active sampling, tubes filled with 400 mg Carbopack B (for aromatic compounds) or a combination of 150 mg Carbopack B and 450 mg Carboxen 569 (for halogenated compounds) were used. Tubes filled with about 200 mg Tenax were used for passive sampling of aromatic compounds as well as of halogenated compounds. Prior to sampling, all tubes were cleaned in a stream of 30–50 mL min–1 He (purity >99.999%, Messer Griesheim) at 300 °C for two hours. This conditioning time seemed to be sufficient, considering the fact that due to the high split ratio only ca. 0.5% of the analytes eventually remaining on the adsorption tube after conditioning is transferred to the GC column and is detected. Sampling procedures Samples were taken by both pumped and diffusive sampling [27] on adsorption tubes (glass tubes filled with adsorption material, as detailed above) and analysed by thermal desorption-gas chromatography. The calibration gas generator has also been evaluated for active sampling on active charcoal tubes followed by solvent desorption with CS2 and subsequent GC analysis. Since these results, however, are not relevant for the evaluation of the test gas generator, the comparison of thermal and solvent desorption of the sampling tubes will not be discussed here, but is given elsewhere [28]. Active sampling For active sampling, a defined volume of test gas was sucked through the adsorption tube by a sampling pump. The adsorption tubes were connected to the sampling chamber of the test gas apparatus. The gas flow through the adsorption tube was measured before and after sampling by an air flow calibrator (Mini-Buck,
Mol. weight (g mol–1) 197.39 147.00 84.93 165.83 133.40 131.39 78.11 106.17 104.15 92.14 106.17 106.17 106.17
TLV (mL m–3)
(mg m–3)
5 75 100 50 200 50 5 100 20 100 100 100 100
40 450 360 345 1080 270 16 440 85 380 440 440 440
Bp (°C)
D (cm2 s–1)
50 174 41 121 74 87 80.1 136 145 110.6 139 144 138
0.076 0.067 0.1037 0.0797 0.0794 0.0875 0.0859 0.0693 0.0701 0.0763 0.067 0.0727 0.0672
M-5) at the inlet of the pump. The sampled volume was calculated as the product of the average gas flow through the adsorption tube and the sampling time. The rotameters on the pump are useful for monitoring the drift of the gas flow during sampling. Active sampling was done over 2–2.5 h with a sampling rate of 6–8 mL min–1, resulting in a sample volume of approx. 1 L, to avoid sample breakthrough. Thermal desorption tubes show a significant flow resistance, which makes them difficult to use with commercially available sampling pumps Since these pumps are designed for sampling with active charcoal tubes, they can be operated only at very small flow resistance, because otherwise the pressure drop exceeds the pre-set threshold and the pump is shut down. Besides, it was very difficult to achieve a constant flow of less than 10 mL min–1 with these pumps. Therefore, a custom-made pump was constructed: It consisted of a membrane pump, an expansion vessel and four needle valves with four rotameters. It allows to sample simultaneously on four channels and to adjust very small gas flows, the constancy of which can be monitored during sampling by rotameters. Passive sampling Similar adsorbent tubes were used for passive sampling as for active sampling. The uptake rate of the diffusive samplers was controlled through the judicious design of the diffusion path which was constructed in-house from stainless steel tubes of precisely known inner diameter. Passive sampling was also done by inserting the open end of the samplers with the diffusion path into one of the ports of the sampling chamber. All sampling was carried out at ambient temperature in the airconditioned laboratory (21±1 °C). Gas chromatographic analysis Calibration and analysis Analysis was carried out by an OI 4460 A instrument (OI Analytical, College Station, TX, USA) modified for thermal desorption operation which was coupled to a HP 6890 gas chromatograph with electronic pressure control (Hewlett Packard, Palo Alto, CA, USA) with a flame ionisation detector. The analytes were desorbed at 300 °C from the adsorption tube onto an internal (focusing) trap which was kept at room temperature. The trap (OI #6) was packed with Tenax, silica gel and active charcoal. After this focusing step the analytes were desorbed from the trap and transferred to the split injector of the gas chromatograph. A SPB-Octyl column
801 Table 2 Conditions used for thermal desorption Step
Condition/Parameter
Value
Desorption of the adsorbent tubes Temperature of the focusing trap Desorption temperature Desorption time Drying step
25 °C 300 °C 5 min No
Desorption from the focusing trap Desorb preheat Temperature of the focusing trap Time
No 220 °C 2 min
Bake Temperature of the internal trap Time Auxiliary temperatures Valves and transfer line
220 °C 4 min 180 °C
(30 m×0.32 mm×1 µm, Supelco) was used for the resolution of aromatic compounds, whereas a HP 624 column (30 m× 0.32 mm×1.8 µm, Hewlett–Packard) was used for the resolution of halogenated compounds. The parameters for thermal desorption are listed in Table 2, those for gas chromatographic analysis are listed in Table 3. Since the commercially available thermodesorption kit for the OI 4460 instrument was not fully satisfactory for our purposes, the heating block for the thermal desorption tube was modified as follows: The heater was made longer (80 mm instead of 65 mm); therefore, adsorbent tubes with a bed length of up to 80 (instead of 65) mm can be used, which is important to avoid early sample breakthrough. For calibration, an empty steel tube was mounted onto the apparatus instead of an adsorption tube, and 2 µL of a standard mixture of the components which were to be analysed were injected through the septum. For this purpose, a stainless steel T-piece was used to connect the stainless steel tube to the He gas flow for desorption. The third arm of the T-piece was sealed with a septum that could be pierced for the injection of the calibration standard into the empty steel tube. The stainless steel tee with
Table 3 Conditions used for GC analysis Condition/Parameter
Value
Carrier gas Injector temperature Injector pressure Total gas flow Split ratio Split flow Gas saver mode Column flow (set-point) Linear velocity of the carrier gas in the column Oven program
He (purity>99.999%) 250°C 6.4 psi (0.44 bar) 243 mL min–1 200:1 239 mL min–1 Off 1.2 mL min–1 22 cm s–1
Detector temperature Hydrogen flow (FID) Air flow (FID) Nitrogen flow (makeup gas, FID)
45°C (4 min)Æwith 15° min–1 to 135°C (0 min)Æwith 90° min–1 to 225°C (hold 2 min) 250°C 40 mL min –1 450 mL min–1 30 mL min –1
the septum was arranged on top of the empty tube, ensuring that the standard was injected in the heated zone of the thermal desorption unit where the analytes evaporated immediately. When injecting manually (and thus comparatively slowly) into a heated zone, the evaporation of the liquid standard mixture from the syringe needle during injection into the thermal desorption unit has to be considered. Thus, a correction for the actual volume that is injected in this way was made to minimise the calibration error. After usually six calibrations the samples were measured by mounting the adsorbent tubes instead of the empty tubes onto the thermal desorption unit and starting an analysis cycle. When adsorption tubes with very small loadings were measured after tubes with high sample loadings, blank values were measured in between to avoid sample carry-over. On line-GC measurements. Hewlett–Packard 6890 GC with FID. Sample introduction from the end of the calibration gas sampling manifold through a gas sampling valve with a 0.5 mL sampling loop kept at 105 °C installed before the split/splitless injector. Split 1:20. Column flow: 5.0 mL at 80 °C. Oven program: 80 °C (1 min) with 25 ° min–1 to 200 °C (hold 1.2 min). Column: HP 5 (Hewlett–Packard), 30 m×0.53 mm×1.5 µm.
Results and discussion A series of experiments was carried out that allowed the evaluation of both the proper functioning of the calibration gas generator as well as of the feasibility of the entire analytical procedure (including sampling and analysis) for selected volatile organic compounds by active or passive sampling, respectively. Only for active sampling (which was the more suitable technique, particularly for shortterm sampling) the detailed results are reported below. They are consistent with the results obtained for passive sampling for which only the halogenated compounds seem to exhibit an underestimation of up to 15%, depending on the sampling conditions and the adsorbent used. It shall be emphasised here that the proprietary design of the passive samplers allows their analysis by thermal desorption which in part is described elsewhere [29]. The performance of both active and passive sampling with the self-made samplers, followed by thermal desorption-GC analysis has been assessed on the basis of the pertinent standards [30, 31, 32, 33].
Characteristics of test gas generation The flexibility and versatility of the test gas generator built based on the injection method stems from the fact that the operating parameters can easily be changed to accommodate a wide concentration range for the calibration gas. The mass concentration of a substance in the calibration gas stream can be calculated by use of the equation: cm(g ) =
10 6 vsir × q × cm × 60 dV/dt
(l )
(1)
where cm(g) is the mass concentration of the analyte in the calibration gas (mg m–3), vsyr is the velocity of movement of the syringe barrel (mm/h), q is the cross sectional area of the syringe barrel (to be calculated as syringe volume/length of scale) (mm2), cm(l) is the mass concentration
802 Table 4 Stability of test gas generation under different experimental conditions, assessed by on line-GC analysis of a four component-standard (benzene, toluene, 1,1,1-trichloroethane and tetrachloroethylene).
Syringe Injection volume speed (µL) (mm h–1)
Benzene
10 10 25 25 100 100 250 250
10.6 2.2 2.7 2.5 1.3 2.8 3.8 1.5
2.5 9.5 2.5 9.5 2.5 9.5 2.5 9.5
Toluene 1,1,1-Trichloroethane
Tetrachloro- No. of ethylene data points evaluated
Equilibration time (h)
1.0 2.1 1.8 2.0 1.0 2.8 4.1 1.3
1.1 2.2 2.4 2.1 1.1 3.4 4.5 1.2
2.1 1.25 2.4 2.25 1.4 1.5 3 2.25
RSD (%)
of the analyte in the liquid standard (g cm–3) (for pure substances, cm(l) may be replaced by the density ρ of this substance in (g cm–3)), and dV/dt is the zero gas vector (cm3 min–1). The factor 106/60 results from the conversion of the different units. As is apparent from Eq. (1), for a constant gas flow (which was typically approximately 1 L min–1) the concentration in the test gas can easily be varied by either exchanging the syringe or the injection speed with which the standard is dosed into the gas stream. To cover a wide range of different instrumental settings (and resulting test gas concentrations), four different syringe volumes (10, 25, 100 and 250 µL) and two different injection speeds (2.5 and 9.5 mm h–1) were investigated. The stability of the test gas generation was investigated by on line-GC analysis of the test compound set (benzene, toluene, 1,1,1-trichloroethene and tetrachloroethylene). The results of the on line-GC-analysis are given in Table 4. They indicate that the stability of the generated test gas is, with two exceptions, within 3% RSD. Considering that this RSD value includes both the fluctuation of the test gas generation and the variability of the response of the GC–FID analysis (which usually already accounts for about 1% RSD), the stability of the test gas concentration can be considered excellent, provided that a long enough time span is allowed for the system to equilibrate. Depending on the operating parameters of the apparatus, this equilibration time may range from approximately one hour to about three hours to reach satisfactory constancy of the test gas concentration.
Use of the calibration gas generator for the preparation of controlled test gas atmospheres The test gas generator was designed to allow the generation of test gas atmospheres from compounds that are liquid at room temperature, but with the use of gas-tight syringes even gaseous substances could in principle be used for the generation of test gases. In this work, compounds covering a wide range of volatility (bp=20 °C ... 192 °C) have been investigated. In addition to the evaluation of the stability of the test gas generation by the on line-GC, both active and passive sampling were conducted to demonstrate the practical ap-
2.3 2.7 2.1 2.3 1.1 2.3 4.0 1.4
8 10 8 10 11 10 5 8
plicability of the developed system. Recovery rates r and their standard deviations s calculated from four to six replicate measurements were used for this purpose. Recoveries were calculated by comparing the experimentally found analyte masses with those that were expected according to the equations for analyte mass uptake for active sampling: mi , active = ci × V
(2)
where mi,active is the mass of analyte i collected on the adsorbent bed by active sampling (mg), ci the concentration of analyte i in the test gas (mg m–3), and V the volume sampled (m3), and for passive sampling: mi , passive = Di × ( A / L ) × ci × t
(3)
where mi,passive is the mass of analyte i collected on the adsorbent bed by passive sampling (mg), Di is the diffusion coefficient of analyte i in the test gas (m2/s1), A is the cross-sectional area of the diffusion path (m2), L is the length of the diffusion path (m), ci is the concentration of analyte i in the test gas (mg m–3), and t is the sampling time (s) The results for the aromatic and halogenated compounds investigated here for active sampling under different conditions are listed in Table 5. No extensive data is presented here for passive sampling, although these data are suited to further support the conclusion of the feasibility of the presented test gas generator and also demonstrate the practicality of the proposed passive sampler design. Concentration levels in Table 6 are given in relation to the Austrian threshold limit value (TLV), since a method potentially suitable for workplace monitoring is required to be validated in the working range of concentrations from 0.1 to 2 times the TLV and to meet certain criteria for the relative overall uncertainty (ROU). The relative overall uncertainty, calculated as (|r – 1| + 2s) ×100 [%], is far below 30% in all cases (for both sampling methods), and therefore meets the requirements for ROU in the European standard EN 482 [30]. Thus, also the analysis of samples produced by the test gas generator attests to its feasibility for the production of test gas atmospheres of precisely known concentrations.
803 Table 5 Recovery rates r and their standard deviations s (n=4 to 6) for active sampling and thermal desorption analysis of apolar compounds under different conditions. (TLV=threshold limit values) Concentration/TLV: Rel. Humidity (%) Storage time (weeks) Compound
0.1 50 0 r
Benzene Toluene Ethylbenzene Xylene (sum of isomers) Styrene Dichloromethane 2-Bromo-2-chloro-1,1,1-trifluoroethane 1,1,1-Trichloroethane Trichloroethene Tetrachloroethene 1,4-Dichlorobenzene
1.020 1.022 1.038 1.040 1.067 1.033 1.006 1.031 1.028 1.026 1.039
s
1 50 0 r
0.010 0.010 0.011 0.011 0.026 0.015 0.010 0.013 0.010 0.013 0.011
1.014 1.016 1.029 1.028 1.021 0.984 0.984 0.995 0.960 0.979 0.962
s
2 50 0 r
0.025 0.026 0.028 0.028 0.028 0.016 0.013 0.012 0.012 0.013 0.018
1.012 1.019 1.039 1.040 1.034 0.969 0.960 0.969 0.945 0.963 0.948
s
2 20 0 r
0.010 0.011 0.016 0.019 0.018 0.018 0.009 0.010 0.011 0.010 0.004
0.981 0.976 0.972 0.971 0.971 0.996 0.972 0.965 0.918 0.938 0.962
s
2 80 0 r
0.015 0.019 0.021 0.021 0.021 0.007 0.007 0.006 0.001 0.003 0.004
0.970 0.972 0.981 0.983 0.991 0.987 0.971 0.987 0.937 0.960 0.942
s
2 80 2.5 r
s
0.021 0.023 0.026 0.028 0.028 0.019 0.014 0.016 0.014 0.016 0.016
1.001 1.004 1.011 1.012 1.017 0.990 1.022 0.993 0.885 0.908 0.939
0.016 0.017 0.021 0.023 0.024 0.015 0.024 0.007 0.012 0.008 0.009
Table 6 Recovery rates r and their standard deviations s (n=4 to 6) for passive sampling on Tenax TA and thermal desorption analysis of apolar compounds under different conditions. (TLV=threshold limit values) Concentration/TLV: Humidity/%: Storage time/weeks: Compound
0.1 50 0 r
Benzene Toluene Ethylbenzene Xylene (sum of isomers) Styrene Dichloromethane 2-Bromo-2-chloro-1,1,1-trifluoroethane 1,1,1-Trichloroethane Trichloroethene Tetrachloroethene 1,4-Dichlorobenzene
1.021 1.012 1.030 1.064 1.092 0.850 0.900 0.906 0.874 0.864 1.010
s
1 50 0 r
0.041 0.031 0.021 0.018 0.028 0.018 0.032 0.023 0.006 0.018 0.021
0.962 0.978 1.019 1.044 1.070 0.855 0.904 0.909 0.873 0.865 0.996
s
2 50 0 r
0.036 0.036 0.039 0.039 0.044 0.016 0.021 0.019 0.021 0.022 0.019
0.961 0.965 0.999 1.029 1.090 0.808 0.848 0.873 0.844 0.843 0.958
Comparison of active and passive sampling The use of the above presented test gas generator allowed the comparison of the two proprietary designs of thermally desorbable sampling tubes for both active and passive sampling, since such a comparison can only be carried out when a controlled test gas atmosphere is available which offers identical conditions for both active and passive sampling. For active sampling, recoveries are excellent for the investigated compounds and fall with only few exceptions outside the interval 100±5%. Particularly the halogenated compounds seem to be partially lost when stored for a period of more than two weeks. In contrast to this, aromatic compounds can be recovered from Tenax adsorbent tubes without problems even when sampled at medium (50%) and high (80%) relative humidities and stored for 2.5 weeks. In contrast to this, passive sampling on Tenax adsorbent tubes was not equally effective for aromatic and halogenated compounds: While the aromatic compounds
s
2 20 0 r
0.047 0.033 0.031 0.026 0.040 0.006 0.012 0.014 0.013 0.016 0.027
0.964 0.970 0.980 0.997 1.034 0.858 0.899 0.918 0.884 0.885 1.052
s
2 80 0 r
0.021 0.023 0.023 0.023 0.025 0.009 0.010 0.010 0.012 0.012 0.020
1.018 1.034 1.059 1.081 1.114 0.855 0.893 0.917 0.899 0.886 1.036
s
2 80 2.5 r
s
0.030 0.028 0.026 0.026 0.026 0.010 0.013 0.011 0.010 0.011 0.019
0.978 0.991 1.012 1.038 1.083 0.847 0.884 0.888 0.866 0.857 0.994
0.042 0.041 0.042 0.045 0.050 0.020 0.017 0.015 0.020 0.019 0.017
show again very satisfactory recoveries (all values within 100±5% with only styrene showing frequently higher values), the recoveries for halogenated compounds deviated significantly from 100%: Except for 1,4-dichlorobenzene, significant underestimates of up to 19% were observed. This indicated that Tenax is too weak an adsorbent for quantitative trapping of the analytes in passive sampling. This observation may be explained by back-diffusion of the analytes (desorption of the analyte from the adsorbent bed when exposed to an atmosphere that does not contain or contains the analyte at lower levels) and has been described in the literature [34]. Tenax is thus only recommended for passive sampling of aromatic and higher boiling compounds. Nevertheless, the relative standard deviation of the measurement was in all cases under 5% (for n=4 to 6), which can be considered excellent since it includes both the variability of the sampling procedure as well as that of the test gas generation which was estimated earlier at ca. 3%.
804
Evaluation of other adsorbents for passive sampling The limited suitability of Tenax TA for passive sampling of halogenated compounds prompted us to carry out additional experiments to screen the applicability of other adsorbents for this purpose. Carbopack B was tested as adsorbent for passive sampling of aromatic compounds at a concentration of two times the threshold limit value and a relative humidity of 77%. The recovery rates (1.032 to 1.085) were very similar to those obtained on Tenax TA. Similarly, a slight increase of the recovery rates with increasing boiling point (5 to 10% from benzene to styrene) could be observed. Carbopack B was also investigated for passive sampling of halogenated compounds. The recovery rates were between 75% (dichloromethane) and 99% (1,4-dichlorobenzene), obviously again increasing with decreasing volatility. The problems encountered with Carbopack B (low recovery rates for halogenated compounds with one and two carbon atoms) were exactly the same as with Tenax TA, even though Carbopack B has an approximately three times higher specific surface area than Tenax TA (100 m2/g instead of 35 m2/g). Therefore, Carboxen 569, which is a stronger adsorbent (specific surface area: 485 m2/g) than Carbopack B, was tested for passive sampling of halogenated compounds. The recovery rates of the saturated halocarbons were improved (92 to 96%), but those of the halogenated olefins were not (80 to 86%). On the other hand, desorption efficiency of less volatile compounds like 1,4-dichlorobenzene decreased significantly (recovery rate of about 40%). It can thus be concluded that neither of the investigated adsorbents is suitable for passive sampling of organic compounds covering a wide range of volatility. While the stronger adsorbents trap the more volatile compounds to a sufficient degree, the higher boiling compounds can only be recovered to an unsatisfactory extent; on the other hand, the weaker adsorbents are only suitable for the less volatile compounds because the more volatile ones are not trapped completely. Investigation of blank values Particularly in the analysis of air samples, careful consideration of blank values is required. Very often, it is not the sensitivity of the analytical method which determines the achievable detection limits, but rather how or whether blank values can be controlled to an acceptable degree. An investigation of blank values was, therefore, carried out also as part of this study. For this purpose, a number of adsorbent tubes that had not been sampled were analysed. The peak areas in the resulting chromatograms (Ablank) were related to those of chromatograms which were obtained by analysing tubes that were sampled at the lowest concentration level (0.1 TLV) by active (Aactive) or passive sampling (Apassive), respectively. The quotient Ablank/Aactive was below 0.003, and the quotient Ablank/Apassive was below 0.03 for all substances except of benzene (0.30) and xylene (0.05). This situation is first due to the fact that the amount of substance which is sampled on the
tube is very small with passive sampling; and secondly, the TLV of benzene is the lowest of the investigated compounds. It has to be emphasised that the blank values determined in this study are not due to adsorption tube blanks but due to memory effects of the internal trap of the thermal desorption unit which could not be further reduced under practical operating conditions. Detection limits for active and passive sampling Evaluation of the chromatograms obtained under typical conditions indicated that peaks with a peak area of 0.2 pAs could be detected, which corresponds to 0.01 ng carbon on the column and 2 ng on the adsorbent tube (at a split ratio of 1:200). For active sampling (sampling volume: 1 L), the detection limit is, therefore, 2 µg C m–3 (in fact, the concentration should be given as “effective carbon” when detected by FID). This is about three orders of magnitude higher than the detection limits usually reported for thermal desorption when used, e.g., in environmental applications. However, it was not the aim of this work to reach low detection limits, but to meet the ROU requirements of the regulation EN 482 [30] at the concentration levels relevant for workplace monitoring, i.e. in the range of 0.1 to 2 TLV. To reach lower detection limits, the parameters of the method (split ratio, sampling volume, trapping temperature) have to be adjusted accordingly. For passive sampling, the detection limits depend on the length and cross-section of the diffusion path, the diffusion coefficient and the sampling time. The value for L/A (quotient of length and cross-section of the diffusion path) was calculated as 75 cm, and the diffusion coefficients are in the order of 0.1 cm2 s–1. Therefore, the estimation of the uptake rate Ui=Di×A/L results in 0.00133 cm3 s–1≈4.8 cm3 h–1. In seven hours, the analytes from an equivalent volume of ca. 33.6 cm3 air can be sampled by passive sampling. Therefore the detection limit may be estimated at 2 ng/33.6 cm3 air, which corresponds to approx. 60 µg m–3 for passive sampling. It is interesting to compare these values with those usually achieved in air monitoring by active sampling and solvent desorption. When using solvent desorption, the chromatographic peaks are narrower, making therefore the detection of peaks with an area of 0.1 pAs possible. This peak area corresponds to 0.005 ng carbon on the column or 0.5 ng carbon injected (assuming a split ratio of 1:100). When 5 mL carbon disulfide are used for desorption of the analytes from the active charcoal bed and 1 µL is injected, at least 0.5×5000=2500 ng (i.e. 2.5 µg) carbon must be sampled on the adsorbent tube. For a sampling volume of 10 L, the detection limit is, therefore, 2.5 µg/ 10 L, i.e. 250 µg m–3 effective carbon.
Conclusion The construction of a test gas generator from components which are readily available in most analytical laboratories
805
or can be easily constructed in house was demonstrated. The performance of such a test gas generator suitable for both active and passive sampling of volatile organic compounds was validated by a set of 11 test compounds and fulfils the requirements of the EN 482 in terms of the relative overall uncertainty for sampling and analysis. The test gas generator can thus conveniently be used to validate both sampling and analysis of VOCs covering a wide range of volatilities and under different sampling conditions. This was demonstrated by comparing the results of active and passive sampling followed by thermal desorption-GC for the determination of the test compound set. It can be concluded from this study that the results of passive sampling are comparable to those of active sampling, and from the practical point of view that passive sampling is also more convenient and particularly for long-term sampling better suited than active sampling. It should also be noted that, in contrast to the general perception, the range of applicability of adsorptive sampling and thermal desorption-GC analysis is not limited to relatively apolar compounds, but has been extended to the analysis of polar compounds as has been reported elsewhere [29]. Acknowledgment Financial support of this work by the Allgemeine Unfallversicherungsanstalt, Vienna, is gratefully acknowledged. The authors thank H. Kronfuß for technical assistance in part of the measurements.
References 1. Brown RH, Curtis M, Saunders KJ, Vandendriessche S (1992) (eds) Clean air at work, Spec Publ 108. The Royal Society of Chemistry, Cambridge 2. Bloemen HJT, Burn J, (eds) (1993) Volatile organic compounds in the environment. Blackie Academic and Professional, London 3. Atkinson R (1990) Atmos Environ 24A:1–41 4. Barratt RS (1981) Analyst 106:817–849 5. Nelson GO (1992) Gas mixtures – preparation and control. Lewis, London 6. Rhoderick GC (1997) Fresenius J Anal Chem 359:477–483 7. Miller WR, Rhoderick GC (1995) Fresenius J Anal Chem 351: 221–229 8. Rhoderick GC, Miller WR (1993) J Chromatogr A 653:524– 531 9. Müller S, Oehme, M (1990) J High Resol Chromatogr 13:34– 39 10. Scaringelli FP, O’Keeffe AE, Rosenberg E, Bell JP (1970) Anal Chem 42:871–876
11. Burg WR, Birch SR, Cuddeback JE, Saltzman BE (1976) Environ Sci Technol 10:1233–1237 12. Schmidt WP, Rook HL (1983) Anal Chem 55:290–294 13. Miguel AH, Natusch DFS, (1975) Anal Chem:1705–1707 14. Torres L, Mathieu J, Frikha M, Namiesnik J (1981) Chromatographia 14:712–713 15. Gautrois M, Koppmann R (1999) J Chromatogr A 848:239– 249 16. Flammenkamp E, Ludwig E, Risse U, Kettrup A (1993) Fresenius J Anal Chem 347:230–232 17. Barkó G, Hlavay J (1998) Fresenius J Anal Chem 360:119–121 18. Schick B, Risse U, Kettrup A (1999) Fresenius J Anal Chem 364:709–713 19. Tolnai B, Barkó, Németh R, Hlavay J (2000) Microchem J 67: 157–161 20. Flammenkamp E, Risse U (1994) Herstellungsverfahren für Prüfgase. In: Grein H (ed) Luftanalytik – Analytische Methoden zur Prüfung gesundheitsschädlicher Arbeitsstoffe der DFG. Verlag Chemie, Weinheim, 9. Lieferung 21. Leichnitz K (1983) Pure Appl Chem 55:1239–1250 22. Namiesnik J (1984) J Chromatogr 300:79–108 23. Verein Deutscher Ingenieure (1980) Messen von Gasen, Prüfgase; VDI-Richtlinie 3490, Blätter 1–17, VDI, Düsseldorf 24. Austrian Ministry for Employment and Social Affairs (1995) Maximale Arbeitsplatzkonzentrationen und Technische Richtkonzentrationen (MAK-Werte-Liste), Amtliche Nachrichten Arbeit – Gesundheit – Soziales, Wien 25. Deutsche Forschungsgemeinschaft (ed) (2000) MAK- und BAT-Werte-Liste 2000. Maximale Arbeitsplatzkonzentrationen und Biologische Arbeitsstofftoleranzwerte. Mitteilung 36. Wiley-VCH, Weinheim 26. Drägerwerk AG (1991) Dräger Probenahme-Handbuch. Dräger, Lübeck 27. Blome H, Hennig M (1985) Staub – Reinhaltung der Luft 45: 505–508 28. Hallama R, Rosenberg E, Grasserbauer M (1997) GIT Fachz Lab 10:993–998 29. Hallama R, Rosenberg E, Grasserbauer M (1998) J Chromatogr A 809:47–63 30. Comité Europeén de Normalisation (1994) EN 482, Workplace atmospheres – general requirements for the performance of procedures for the measurement of chemical agents, Brussels 31. Comité Europeén de Normalisation (1995) EN 689, Workplace atmospheres – guidance for the assessment of exposure by inhalation to chemical agents for comparison with limit values and measurement strategy, Brussels 32. Comité Europeén de Normalisation (1995) EN 838, Workplace atmospheres – diffusive samplers for the determination of gases and vapours – requirements and test methods, Brussels 33. Comité Europeén de Normalisation (1997) EN 1076, Workplace atmospheres – pumped sorbent tubes for the determination of gases and vapours – requirements and test methods, Brussels 34. Sunesson A-L, Sundgren M, Levin J-O, Eriksson K, Carlson R (1999) J Environ Monit 1:45–50
