Anal. Chem. 2001, 73, 2405-2411
Determination of Vapor Pressure of Low-Volatility Compounds Using a Method To Obtain Saturated Vapor with Coated Capillary Columns Lars Rittfeldt
Swedish Defence Research Agency, Department of NBC Defence, FOI, SE-901 82 Umeå, Sweden
The vapor pressures of O-ethyl S-2-diisopropylaminoethyl methylphosphonothiolate (VX), O-isobutyl S-2-diethylaminoethyl methylphosphonothiolate (RVX), and 2,4-dinitrotoluene (2,4-DNT) were determined with the gas saturation method in temperatures ranging from -12 to 103 °C. The saturated vapor was generated using a fused-silica column coated with the compound. This column was placed in a gas chromatograph, and the vapor pressure was determined directly from the detector signal or by sampling on Tenax tubes that were subsequently analyzed. From the linear relationships obtained by plotting log P vs 1/T, the enthalpies of vaporization (∆Hvap) and the vapor pressures at selected temperatures were determined. The vapor pressure of VX at 25 °C was 0.110 Pa and the ∆Hvap 77.9 kJ‚mol-1. The corresponding results for RVX were 0.082 Pa and 76.6 kJ‚mol-1. The vapor pressure of 2,4-DNT at 72 °C (melting point) was determined to 6.0 Pa, and the enthalpies of the solid and the liquid state were 94.2 and 75.3 kJ‚mol-1, respectively. Using capillary columns to generate saturated vapors has three major advantages: short equilibrium time, low consumption of sample, and safe handling of toxic compounds. In the assessment of the fate and behavior of chemicals in the environment, the vapor pressure is one of the most important properties. Vapor pressure data are indispensable in the modeling of substance distribution between atmosphere, soil, and water. The vapor pressure also governs the persistence, volatilization, and transport of a chemical introduced into the environment. Unfortunately, there is a lack of reliable vapor pressure data of low-volatility and toxic chemicals especially at subzero temperatures. The reason is that most determinations of vapor pressure are done at elevated temperatures. Also, extrapolation to ambient or subzero temperatures introduces errors. The gas saturation method has been widely used to measure low vapor pressures and is considered sufficiently accurate at pressures down to ∼10-6 Pa. Measurements below 0 °C have however seldom been undertaken using this method. In our institute, the gas saturation method was employed in the beginning of 1970s1 to measure the volatility of the highly toxic nerve agent, O-ethyl S-2-diisopropylaminoethyl methylphos(1) Frostling H. Acta Chem. Scand. 1974, A28, 83-85. 10.1021/ac010015v CCC: $20.00 Published on Web 05/05/2001
© 2001 American Chemical Society
phonothiolate (VX). The saturated vapor was obtained from an aerosol of the compound kept in a drum that was rotating to reduce the precipitation of particulate. In principle, the use of an aerosol is a good technique to obtain a good surface/load ratio that will promote a fast equilibrium partition. The equipment is however very complicated to handle especially for highly toxic compounds when rigorous precautions have to be taken into consideration. In more recent works,2-7 a packed column was used to generate saturated vapors. To obtain a large evaporation surface, the test compound in the column was either finely ground or coated on an inert material. Wania et al.5 used a comparatively small saturator column, and ∼40 mg of the compound was needed for the coating of the glass beads. On the other hand, Buchanan et al.7 used 7 g of neat VX in their saturator. Experiments involving such a large amount of a highly toxic compound impose a great risk and cannot be carried out in an ordinary laboratory. Use of a large amount also necessitates that the column is conditioned for quite a long time before reliable measurements can be performed. The method presented in this paper involves a novel method to generate saturated vapor by a coated fused-silica column. Less than 1 mg of the test substance is needed for the coating; hence, a prepared column can be safely handled even with a toxic compound. When gas is passed through the column it will have a close interaction with the inner surface and saturation will be reached almost instantly. By using a gas chromatograph to monitor the column effluent, the vapor pressure of the compound can be determined over a wide temperature range. Sampling on Tenax tubes may be applied for measuring vapor pressures below 10-3 Pa and qualitative analysis of the vapor. The accuracy of the method was confirmed for VX and 2,4dinitrotoluene by reproducing literature data for vapor pressures, enthalpies of sublimation, and vaporization. EXPERIMENTAL SECTION Materials. VX and O-isobutyl S-2-diethylaminoethyl methylphosphonothiolate (RVX) were synthesized in-house (Figure 1). (2) Pella P. A. Anal. Chem. 1976, 48, 1632-1637. (3) Westcott, J. W.; Simon, C. G.; Bidleman, T. F. Environ. Sci. Technol. 1981, 15, 1375-1378. (4) Sonnefeld, W. J.; Zoller, W. H. Anal. Chem. 1983, 55, 275-280. (5) Wania, F.; Shiu, W. Y.; Mackay, D. J. Chem. Eng. Data 1994, 39, 572-577. (6) Da Silva, J. P.; Da Silva, A. M. J. Chem. Eng. Data 1997, 42, 538-540. (7) Buchanan, J. H.; Buettner, L. C.; Butrow, A. B.; Tevault, D. E., ECBC-TR068, Research and Technology Directorate, 1999.
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001 2405
Figure 1. Chemical structures of VX and RVX.
These nerve agents are synthesized in a special laboratory that is built mainly for synthesis of highly toxic compounds. The last steps of the synthesis are done in a glovebox inside fume hoods to maximize the security. Full body protection including breathing protection is worn when the agents are handled, which includes synthesis work as well as preparation of solutions and columns. For the destruction of the agents, a 50:50 solution of 10% sodium hydroxide and ethanol is used. The purity of the agents determined by gas chromatography was better than 96%. VX showed some degradation when a column coated with VX was kept at ambient temperature for several days. The 2,4-dinitrotoluene used had a purity of 97% and was purchased from Aldrich. Methylene chloride (Merck, p.a.) was used as solvent. The preparation of columns was done with 1% solutions of the test compounds. Calibration solutions of known concentrations of the test compounds and n-undecane (Sigma, p.a.) were used to determine relative response factors. A certified calibration gas containing 10.1 (0.2 ppm propane in nitrogen was used for the calibration experiments. Instrumentation. An HP 5880A gas chromatograph (GC) equipped with a flame ionization detector (FID) was used to test the properties of the columns. The instrument was not furnished with any subambient equipment. Temperatures below ambient was achieved by placing pieces of dry ice in the oven. This procedure lowered the accuracy of the oven temperature with ∼(0.5 °C. The carrier gas (nitrogen) was supplied via a mass flow controller (Brooks, 5850S) kept at ambient temperature. The controller was calibrated for standard conditions (273.15 K, 101 325 Pa) in the range 0.5-10 mL‚min-1. The accuracy specification states the error limit to be (0.2% of full-scale flow plus (0.7% of actual flow. The Tenax (TA) minitube system of CCAI was used in the experiments requiring enrichment or qualitative analysis. The system was either adapted to an HP 5890 GC and a TRIO-1 mass spectrometer or to a Varian 3500 GC equipped with a FID. An E.T.I. precision thermometer purchased from Pentronic AB was used to measure the GC oven temperature. The instrument was found to be correct within 0.1 °C at 0.0 (ice water) and at 100.8 °C (boiling water under 1022 mbar pressure). Preparation of Column. An untreated fused-silica tube (∼3 m long, inner diameter of 0.32 mm) was washed with methylene chloride by suctioning the solvent through the column using a water aspirator. The pressure was then adjusted so that a small plug of solvent passed through the column in a few seconds. During the passage, a thin film of solvent will deposit on the column wall. The film thickness, governed by the speed of the plug,8 may be estimated by checking how much solvent is retained when a plug passes through a solvent-free column. After these (8) Novotny, M.; Blomberg, L. J. Chromatogr. Sci. 1970, 8, 390-393.
2406
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
preparatory tests, the remaining solvent was completely evaporated to make the column ready for coating according to the “dynamic method”. This was done by dipping the end of the column for ∼0.5 s into the test compound (e.g., VX) solution. In this way, a 40-50-µL solution plug was made to pass through the column, which was then gently purged with pure gas to evaporate the solvent, leaving the test compound on the column wall. As a safety precaution, a few centimeters were cut off from the column ends in order to avoid contact with the toxic compound. The concentration of the solution, as well as the speed of the liquid plug passing through the column, govern the column film thickness. One percent of the compound in methylene chloride and a speed of 5 dm‚s-1 give the film a thickness of ∼0.1 µm, which corresponds to a load of ∼0.1 mg‚m-1. Vapor Generation. The coated column was placed in the gas chromatograph and then connected to the mass flow regulator, which gives a stable and accurate flow in the range from 0.5 to 10 mL‚min-1. The flow is set to a value that matches the volatility of the test compound and the detector sensitivity. High gas flow will consume more substance and might also overload the detector. If needed, the saturated flow from the column can be diluted with makeup gas to adapt it to the detection device. Determination of Vapor Pressure. Two techniques were used to measure the concentration of the test compound in the column effluent. Vapor pressures were then derived from the concentrations obtained using the ideal gas law. Gas Chromatographic Detectors. The flame ionization detector, nitrogen phosphorus detector (NPD), and photoionization detector (PID) are suitable for these studies as they are sensitive and have a wide linear range. The use of a GC also gives good control of the column temperature, thus allowing the vapor pressure to be determined as a function of temperature. In the experiments presented, an HP-5880A with a FID was used. The test column was connected to the detector via a glass “quick fit” and a deactivated fused-silica liner (0.15 m, i.d. 0.53 mm). The liner is necessary to prevent the column end from being heated by the detector and having a temperature different from that of the oven. The concentration of the test compound can be derived directly from the detector signal once the sensitivity of the detector is known. The FID signal is proportional to the mass flow of compound entering the detector. Response factors are therefore expressed in units of (pA(ng‚s-1)-1), which is equal to (pAs‚ng-1) or “peak area” per nanogram. It is thus possible to determine the response factor of a compound in an ordinary gas chromatographic analysis. The accuracy of this technique is however poor and therefore the detector sensitivity was determined as follows: The detector response factor of propane was first determined by recording the detector signal for the calibration gas at different gas flows. The response factor was calculated for gas flows at 273.15 K and 101325 Pa since the mass flow controller was calibrated for these conditions. The flow controller was found to give a constant flow virtually independent of operating temperature and outlet pressure. In the next step, detector responses of the test compound (e.g., VX) and n-undecane were compared. This was done in a gas chromatographic analysis where the relative response of the test compound is derived from the ratios of peak area and amount. Finally, the obtained FID response factor (pA(ng‚s-1)-1) of the
Figure 2. Principal scheme of the test equipment used to sample on Tenax tubes.
test compound is adjusted to correlate with the calibration gas (propane) instead of n-undecane. This adjustment is ∼3% and is derived from predicted detector sensitivities of compounds in a homologous series.9,10 Sampling of Vapors on Tenax. Sampling on Tenax followed by thermal desorption and gas chromatographic analysis was done in order to improve the limit of detection by several magnitudes but also to obtain qualitative analyses of the vapor from the column. The technique allows for separate analysis of the test compound, and therefore, a possible influence of impurities can be corrected for by the use of Raoult’s law. The sampling was done according to Figure 2 with the Tenax tube connected to the column via a deactivated fused-silica liner and an open split connector purged with pure gas. The sample pump (connected to the Tenax tube) was set to a flow 10-20% higher than the column flow, thus securing that all flow from the column enters the tube. The adsorbent tube was kept at a higher or the same temperature as the column because a lower temperature would promote formation of aerosol and thereby lower the sampling efficiency. The Tenax tube sampling was mostly employed at ambient or subambient temperatures. Tubes to be analyzed were mounted in a cassette, which was placed in the thermal desorption unit. The tubes were then sequentially inserted, heated, and analyzed. The detector response for the test compound was calibrated by analyzing known amounts in spiked Tenax tubes. These were prepared by making 5-µL injections of test compound standards into the glass wool of the tubes. Volatile solvents (e.g., pentane or methanol) were used for making up the standards, thus allowing their removal from the adsorbent while clean air was sampled from a gas cylinder. RESULTS AND DISCUSSION When a newly prepared test column is mounted in the gas chromatograph, it needs to be gently conditioned before commencing the actual measurements. During this period, the detector response may be considerably higher due to the presence of volatile impurities. Figure 3 shows a typical detector signal curve for a test compound having volatile impurities. This example is however run under mild conditions (1 mL‚min-1 and 30 °C) which prolong the time to reach a stable level. Provided none of the (9) Frostling, H. Aerosol Sci. 1970, 1, 341-356. (10) Jalali-Heravi, M.; Fatemi, M. H. J. Chromatogr., A 1998, 825, 161-169.
Figure 3. Initial registration of the FID signal for a capillary column coated with VX when flow is set to 1.0 mL‚min-1. Oven temperature is 30 °C.
impurities have volatility similar to that of the test compound it normally takes less than 10 min to condition the column at higher flow and temperature. After the conditioning period, the column effluent should be entirely saturated with the test compound and the detector signal should be stable as soon as temperature and flow have stabilized. As seen in Figure 4, the FID response is proportional to the column flow at isothermal conditions. Changing the temperature and keeping the column flow constant, the detector signal follows the vapor concentration. Figure 5 shows how the FID responds to 5 °C increases in oven temperature. It is apparent that the detector response closely follows the oven adaptation to the set temperature as if the column would be a temperature sensor. When a coated column is used, the test compound evaporates mainly from the gas inlet side. The compound will vaporize gradually until depletion from that side and the “clean” zone extends toward the detector. Eventually the length of the remaining coating will be too short so that the equilibrium concentration cannot be reached. The detector response will then drop, as seen in Figure 6. The VX concentration in this experiment was ∼1100 mg‚m-3 and thus ∼11 µg‚min-1 evaporates. Since the column was coated with ∼100 µg‚m-1 the front of the coating moves ∼1 dm‚min-1. Hence, as the detector signal drop occurs within 2 min, Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
2407
Figure 4. Registration of FID signal for a capillary column coated with VX when flow is increased from 0 to 10 mL‚min-1 in steps of 2 mL‚min-1. Oven temperature is 30 °C.
Figure 6. Registration of FID signal for a capillary column coated with VX when column flow is set to 10 mL‚min-1 and all the coating evaporates after ∼10 min. Oven temperature is 75 °C.
values can be used for calculating vapor concentrations (1) or pressures (2). The logarithms of the vapor pressure are then plotted as a function of the inverse absolute temperature.
Ctc ) Sdet(KtcFcarr)-1
(1)
Ctc is the concentration of test substance at 273.15 K and 101325 Pa (ng‚mL-1), Sdet is the detector signal (pA), Ktc is the detector response factor (pAs‚ng-1) of the test compound derived in the calibration experiments, and Fcarr is the carrier flow of nitrogen at 273.15 K and 101 325 Pa (mL‚s-1).
Ptc ) PcolVtc(Vtc + carr)-1
Figure 5. Registration of FID signal for a capillary column coated with VX when oven temperature is increased from 35 to 50 °C in steps of 5 °C. Column flow is 3 mL‚min-1.
this experiment implies that less than 0.2 m of coating is needed to reach equilibrium. The column flow was 10 mL‚min-1, which corresponds to ∼2 m‚s-1 flow rate. The time to reach equilibrium is thus ∼0.1 s. This agrees well with a theoretical value obtained for a 0.32-mm column with laminar gas flow when the diffusion constant for the test compound is assumed to be ∼0.05 cm2‚s-1. Once a test column has been conditioned in the gas chromatograph and a stable detector response is obtained, the column may be used for further experiments. The measurement of detector signal versus temperature provides the basic data. To reduce systematic errors, this is done both for the rising and the lowering of temperatures in the investigated interval. In this way, a subcooled state of the compound will also be revealed. The detector signal values can be printed directly in picoamperes with a resolution of 0.01 pA. After a new temperature has been set, the FID signal was monitored until its variation is negligible. The value recorded for each temperature is compensated for the background signal, i.e., the detector signal obtained from pure carrier gas. Using the response factor of the test compound, determined in separate calibration experiments, the detector signal 2408
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
(2)
Ptc is the vapor pressure of the test compound and Pcol is the total pressure at the end of the column. Vtc is the volume of test compound and Vcarr is the volume of carrier gas flowing during a specific time period, e.g., 1 min. Vtc is calculated from the Ctc at 273.15 K and 101 325 Pa using the ideal gas law. The additional mass flow due to the evaporated test compound is negligible as it is less than 0.1% in most experiments. Pcol is close to ambient pressure, but depending on the flow restriction in the liner to the detector, the pressure increase might not be negligible for high column flows. In the calibration procedure, the FID signal was recorded for different carrier gas flows of the calibration gas. It turned out that the calculated response factors varied over the flow range. This is caused by a change in detector sensitivity when the sum flow of carrier and makeup gas is varied. The detector response is therefore expected to change with an altered carrier flow, which is demonstrated in Figure 7. When the concentration or vapor pressure of a compound was calculated, the corresponding response factor was used. Therefore, as long as the conditions are the same, in the calibration procedure and the experiment, the differences in detector sensitivity as well as deviations in column flow will be leveled out. The relative detector response factors for the test compounds were determined in gas chromatographic analyses and the results are presented in Table 1.
Table 1. Relative Detector Response (FID), Equations of Vapor Pressure, Enthalpies of Vaporization, and Sublimation Obtained in Capillary Column GC Experiments for VX, RVX, and 2,4-DNT
a
compound
relative FID responsea
temp (°C)
line equation log P (Pa)
∆Hvap or ∆Hsub (kJ‚mol-1)
concn 25 °C (mg‚m-3)
VX (l) RVX (l) 2,4-DNT (s) 2,4-DNT (l) 2,4-DNT (a)
0.52 ( 0.03 0.52 ( 0.03 0.47 ( 0.02 0.47 ( 0.02 0.47 ( 0.02
-12 to 103 -10 to 103 32-73 32-103 27-73
-4068T-1 + 12.687 -4003T-1 + 12.342 -4921T-1 + 15.03 -3934T-1 + 12.177 -4309T-1 + 13.261
77.9 76.6 94.2 75.3 82.5
11.9 8.9 2.5 7.0 4.7
Relative to propane.
Figure 7. Registration of FID response as a function of makeup gas flow. Carrier gas flows were 2 and 10 mL‚min-1 of 10.1 ( 0.2 ppm propane in nitrogen.
The results of the FID measurements for columns coated with VX and RVX, respectively, are presented in Figure 8 and in Table 1. Using the integrated Clausius-Clapeyron equation, the enthalpy of vaporization (∆Hvap) can be calculated from the regression line. This can be done if ∆Hvap is constant within the investigated temperature interval. As seen from Figure 8, there is a slight nonlinearity in the data and an Antoine or second-order polynomial fit would model the data better. This is what could be expected for such a wide temperature interval, and therefore, the derived ∆Hvap values are not exactly valid over the entire temperature interval. The equation of the line is
log P ) -A/T + B
(3)
A ) ∆Hvap/2.303R
(4)
where
Results obtained for VX, presented in Table 1, are in good agreement with literature values. Buchanan et al.7 reported the concentration to be 12.6 mg‚m-3 at 25 °C. As the data almost overlap each other, the value of ∆Hvap should be close to the one of the present work. This is however lower than the value of 100 kJ‚mol-1 previously reported by Frostling.1 The vapor concentration 11.9 mg‚m-3 of VX at 25 °C is ∼30% lower than the vapor
Figure 8. Vapor pressure registered for columns coated with VX respective RVX versus the inverse of the absolute temperature (-12 to +103 °C). Column flows were 2 or 10 mL‚min-1.
pressure reported by Franke,11 but 30% higher than the published value.12 The 30% deviation in vapor pressure corresponds to a temperature change of ∼2.5 °C. RVX is closely related to VX. The two compounds have identical molecular weights and are about equally toxic. Measurement results for a column coated with RVX are presented in Figure 8 and Table 1. The data points of this experiment deviate more (R2 ) 0.9958) from the line obtained than the data of VX do (R2 ) 0.9992). The reason for this is the “data shift” occurring at ∼60 °C. Other experiments undertaken for RVX had lower data deviation but they did not span over that wide temperature interval. As the obtained line equations for RVX were about the same, the data from these experiments were not included in the figure. Data for RVX have not been reported earlier, but the results are reasonably close to those of VX considering their structural similarity. In gas chromatographic analyses, the retention times of VX and RVX were found to be 14.37 and 14.50 min, respectively. Regarding VX as a standard with well-established vapor pressure, the retention times can be used to estimate the vapor pressure of RVX.13,14 This approximation supports that RVX should have slightly lower vapor pressure than VX. In the measurements undertaken with VX and RVX, there was no need to sample vapor on Tenax to improve the detection limit. This was nevertheless done at ambient temperature in order to (11) Franke, S. Manual of Military Chemistry. Chemistry of Chemical Warfare Agents; Deutscher Milita¨rverlag: Berlin (East), 1967; Vol. 1. (12) Handbook of Physical Properties of Organic Chemicals; Howard, P. H., Meylan, W. M., Eds.; CRC Press. Inc.: Boca Raton, FL, 1997; p 1156. (13) Bidleman, T. F. Anal. Chem. 1984, 56, 2490-2496. (14) Donovan, S. F. J. Chromatogr., A 1996, 749, 123-129.
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
2409
Figure 9. Vapor pressure registered for columns coated with 2,4dinitrotoluene versus the inverse of the absolute temperature (27103 °C). Column flows were 2 or 10 mL‚min-1.
check the vapor purity as well as method accuracy. These data (three for each compound) deviate from the lines obtained similarly to the data presented in Figure 8. At vapor pressures close to 10-3 Pa, the FID signal is only 5% above the background level for clean carrier gas. The precision will consequently be reduced in parallel to lowering the pressure. The detector signal is however very stable at these levels and by changing the flow back and forth between zero and 10 mL‚min-1 the precision can be improved. The vapor pressure of 2,4-dinitrotoluene (2,4-DNT) was determined in three different experiments employing the FID method in addition to four determinations with sampling on the Tenax tubes. The nitro compound has a melting point of 71-72 °C. It is thus solid under the conditions for the column preparation. The “dynamic method” to prepare capillary columns is found to give uniform liquid coatings.8 This is supported in the result presented in Figure 6, where the signal drop being almost linear indicates a uniform VX coating. An equivalent result has been obtained for 2,4-DNT, thus suggesting that the “dynamic method” can be used to produce solid coatings, which are sufficiently uniform. The results from the measurements of vapor pressure versus temperature are presented in Figure 9 and Table 1. As can be seen, the data below the melting point are split up into three lines where the middle one was obtained from newly prepared and conditioned columns while the lower line is derived from columns that had been stored in the freezer for more than 1 h. The upper line, which extends above the melting point, is from columns charged with liquid 2,4-DNT. As can be seen, the compound can be subcooled to 32 °C. With further cooling to 27 °C, the vapor pressure decreases to a value close to the middle line. The upper and lower lines coincide with the expected15 vapor pressures of a solid and subcooled compound given in relation 5.
ln P(s) ) ln P(l) + 6,81 (1 - Tmp/T)
(5)
There is no obvious explanation to the middle line, but it is probably related to the compound’s state of aggregation. It may (15) Lyman, W. J.; Reehl, W. F.; Rosenblatt, D. H. Handbook of Chemical Property Estimation Methods; American Chemical Society: Washington, DC, 1990; p 14-4.
2410
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
derive from a sudden transition into a glassy state formed from subcooled liquid or on very rapid evaporation of solvent from a solution. This state should have a less rigid structure than the monoclinic crystals of 2,4-DNT. The “Tenax sampling method” was used at ambient temperature (21,8 °C); the results of four samples from two columns were 1.52 ( 0.05 mg‚m-3. These samples were taken from two different columns that had been kept in a freezer for more than 1 h. The corresponding concentration obtained from the equation for solid 2,4-DNT is 1.65 mg‚m-3. This is an acceptable discrepancy considering that the value is extrapolated from the equation. Also, taking into account the impurity of the compound and using Raoult’s law to correct, the adjusted concentration would be 1.57 ( 0.05 mg‚m-3. Calculating the concentration at the same temperature from the results of Pella2 gives 1.55 mg‚m-3. Accuracy Considerations. The gas chromatographic methods presented for measuring vapor pressure as a function of temperature have good reproducibility and the data scatter from the linear relationship is minimal (R2 ) 0.999). Systematic errors cannot however be ruled out. For instance, the gas chromatograph used was unexpectedly found to have an inaccurate oven temperature. The error was 3.3 °C at a set temperature of 100 °C, and close to ambient the error was 2.1 °C. This has a considerable effect on the vapor pressure results and a minor influence on the enthalpy of vaporization. For concentrations below 0.5 mg‚m-3, the background signal of the carrier gas is significantly higher than the signal caused by the test compound. Thus, a small relative error in background response is magnified in the calculation of the detector response of the compound. Such data should therefore be verified with the Tenax sampling method. Another source of systematic error is the detector response factor determined for the test compound. It has to be considered that the sensitivity of the detector (FID) varies with different gas flows, e.g., that of hydrogen, makeup gas, or carrier. Most of these problems can however be avoided by using a calibration gas to establish variations of the detector response under various conditions. The vapor pressure versus the temperature has been determined for some additional chemicals, i.e., naphthalene, 2-methylnaphthalene, phenanthrene, and 2,4,6-trintrotoluene. Though only single experiments were undertaken, the results compare well with literature data.16-17 There may however be a tendency in data to give lower enthalpies than what is found in the literature, and the vapor pressures at lower temperatures are in consequence slightly overestimated. The reason for this is not fully understood. It might be due to variations in the detector performance, and therefore, a calibration experiment was recently undertaken at higher concentrations, obtained from a calibration gas of 1007 ppm propane in nitrogen. It was found that the detector had only minor deviations from the linear relationship. Another reason, to obtain too low enthalpies, could be that the pressure at the column end increases with rising temperature due to an increase in flow restriction. Hence the vapor will not be saturated at ambient pressure. In future experiments, a method to measure the pressure at the column end will be implemented. (16) Oja, V.; Suuberg, E. J. Chem. Eng. Data 1998, 43, 486-492. (17) Shiu, W.-Y.; Ma, K.-C. J. Phys. Chem. Ref. Data 2000, 29, 41-130.
CONCLUSION Using capillary columns to generate saturated vapors has three major advantages: short equilibrium time, low consumption of sample, and safe handling of toxic compounds. The vapor pressure of a compound versus the temperature may be directly determined in a gas chromatograph using a detector with known performance, e.g., FID. This technique is applicable to vapor pressures from 10-3 to 103 Pa. Using the Tenax sampling method to monitor the column effluent makes it possible to measure vapor pressures as low as to 10-6 Pa. Once the vapor pressure has been determined as a function of temperature, capillary columns can be used for the production of accurate test atmospheres. It is easy to put together a flexible and advanced vapor generator. Changing column or recoating a column can be done in a matter of minutes. By connecting more
than one column to a gas flow, several compounds can be mixed in almost any proportions. ACKNOWLEDGMENT The author thanks Go¨sta Lindberg for synthesizing the agents as well as Associate Professor Bo Nilsson, Dr. Gertrud Puu, Edvard Karlsson, Lars Trogen, and Lars Ha¨gglund for valuable discussions. The author also thanks Swedish Defence Research Agency, FOI, Department of NBC Defence, Sweden for founding and supporting this project.
Received for review January 5, 2001. Accepted March 14, 2001. AC010015V
Analytical Chemistry, Vol. 73, No. 11, June 1, 2001
2411
