Anal. Chem. 2001, 73, 5651-5654
Headspace Solvent Microextraction Aaron L. Theis, Adam J. Waldack, Susan M. Hansen, and Michael A. Jeannot*
Department of Chemistry, St. Cloud State University, St. Cloud, Minnesota 56301-4498
A hanging microliter drop of 1-octanol is shown to be an excellent preconcentration medium for headspace analysis of volatile compounds in an aqueous matrix by gas chromatography (GC) or gas chromatography/mass spectrometry (GC/MS). Model compounds benzene, toluene, ethylbenzene, and o-xylene (BTEX) are conveniently and rapidly preconcentrated in the microdrop. An internal standard, decane, is present in the organic extracting solvent, and linear calibration curves of relative peak area versus aqueous concentration are obtained for the four model compounds. Detailed kinetic studies reveal that the overall rate of mass transfer is limited by both the aqueous-phase stirring rate and the degree of convection within the organic phase. The very low vapor pressure of 1-octanol results in minimal evaporation of the microdrop during the extraction time. This system represents an inexpensive, convenient, and precise sample cleanup and preconcentration method for the determination of volatile organic compounds at trace levels. Direct headspace analysis of volatile organic compounds (VOC) in aqueous or other matrixes has been utilized extensively for years as a means of directly determining VOC without interference from the sample matrix.1,2 In such determinations, the headspace is typically sampled directly with a microsyringe, usually for analysis by gas chromatography; however, such determinations require large Henry’s Law constants and are, therefore, applicable only to extremely volatile compounds. Semivolatile compounds, or VOC present at trace levels require a more efficient sampling procedure to collect a detectable amount of analyte. Techniques such as purge-and-trap,1,2 membrane extraction,3 and solid-phase microextraction4 are excellent analytical techniques that are successfully employed to achieve this goal; however, they each require a specialized apparatus with some type of solid or polymeric sorbent to collect the analyte. Solvent microextraction using a drop of solvent suspended from the tip of a syringe needle in a stirred aqueous sample solution has been successfully demonstrated in conjunction with quantitation by GC.5-9 Additionally, the use of a drop as a gas * Corresponding author. Phone: 320-255-2046. Fax: 320-203-6041. E-mail:
[email protected]. (1) Drozd, J.; Novak, J. J. Chromatogr. 1979, 165, 141-165. (2) Nunez, A. J.; Gonzalez, L. F.; Janak, J. J. Chromatogr. 1984, 300, 127-162. (3) Yang, M. J.; Harms, S.; Luo, Y. Z.; Pawliszyn, J. Anal. Chem. 1994, 66, 6, 1339-1346. (4) Pawliszyn, J. Solid-Phase Microextraction, Theory and Practice; Wiley: New York, 1997. (5) Jeannot, M. A.; Cantwell, F. F. Anal. Chem. 1997, 69, 235-239. (6) de Jager, L. S.; Andrews, A. R. J. Analyst 2000, 125, 1943-1948. 10.1021/ac015569c CCC: $20.00 Published on Web 10/30/2001
© 2001 American Chemical Society
sampling interface for the determination of atmospheric NH3, SO2, NO2, and Cl2 has been demonstrated.10-12 The present study demonstrates the feasibility of the use of a very small volume of solvent to achieve preconcentration in headspace analysis of VOC in an aqueous matrix. The technique is very inexpensive when compared to sorbent-based techniques, because the drop is completely renewable at negligible cost. Good precision and sensitivity are obtained with short analysis times (e.g., 5 min). A further objective of this work is to characterize the kinetic (mass transfer) behavior of the system. EXPERIMENTAL SECTION Apparatus. A Hamilton 7105 plunger-in-needle syringe (Hamilton, Reno, NV) with point no. 2 was used to suspend the drop of 1-octanol and inject it into the GC or GC/MS. Samples were stirred in 1.0 mL conical vials containing Teflon-lined septa (Wheaton, Millville, NJ) using an electronic magnetic stirrer (LTE Scientific, Greenfield, Oldham, England). The syringe was clamped in a fixed position relative to the vial in order to consistently place the needle tip in the headspace of the vial. For the kinetic studies, the vial was thermostated at 25 °C with a water bath. Where convection in the organic phase was desired, the bristles of an electric toothbrush were used to gently agitate the needle and attached drop of 1-octanol. In this case, a piece of aluminum foil was used in place of the septum to allow for efficient transfer of vibrations to the drop. A sketch of the apparatus is shown in Figure 1. Separation, identification, and quantitation were performed on the following instruments: A Shimadzu QP-5000 gas chromatograph/mass spectrometer equipped with a Restek XTI-5 capillary column (30 m, 0.25-mm i.d., 0.25-µm d.f.; Restek, Bellefonte, PA), electron impact ionization, and a quadrupole mass analyzer was used for the analytical quantitation studies. The column, injector, and interface temperatures were 80, 250, and 280 °C, respectively. The inlet pressure was 35 kPa, and the total flow rate was 45 mL/ min, with the inlet operated in split mode, resulting in a split ratio of 62. The quadrupole was scanned from 29 to 350 amu at 2 scans/ s, except for the detection limit studies in which single-ion monitoring (SIM) was used. In the case of SIM, the quadrupole was fixed at m/z 78 for benzene and m/z 91 for toluene, ethylbenzene, and o-xylene. All analytes and the internal standard eluted before the solvent (1-octanol), and the data acquisition was stopped after 10 min (before the elution of the solvent). A Hewlett(7) Psillakis, E.; Kalogerakis, N. J. Chromatogr. A. 2001, 907, 211-219. (8) de Jager, L. S.; Andrews, A. R. J. J. Chromatogr. A. 2001, 911, 97-105. (9) Liu, W.; Lee, H. K. Anal. Chem. 2000, 72, 4462-4467. (10) Liu, S.; Dasgupta, P. K. Anal. Chem. 1995, 67, 2042-2049. (11) Cardoso, A. A.; Dasgupta, P. K. Anal. Chem. 1995, 67, 2562-2566. (12) Genfa, Z.; Dasgupta, P. K. Anal. Chem. 2000, 72, 3165-3170.
Analytical Chemistry, Vol. 73, No. 23, December 1, 2001 5651
Figure 1. Sketch of the apparatus used for headspace solvent microextraction (not to scale).
Packard 5890 gas chromatograph equipped with a Supelco SPB-1 capillary column (30 m, 0.32-mm i.d., 0.25-µm d.f.; Supelco, Bellefonte, PA) and flame ionization detector was used for the kinetic studies. The column, injector, and detector temperatures were 100, 250, and 280 °C, respectively. The inlet was operated in split mode with a pressure of 10 kPa and a total flow of 100 mL/min. In practice, because of the relatively large bore of the microsyringe needle, it was necessary to replace the GC septum on a daily basis to ensure reliable operation. For the determination of diffusion coefficients via the Taylor dispersion method, a Sage Instruments (Orion Corp., Cambridge, MA) model 351 syringe pump was used to deliver 1-octanol through 762 cm of loosely coiled 0.051-cm (nominal)-i.d. stainless steel tubing at a flow rate of 0.045 mL/min. Samples were injected via a Rheodyne 7125 injector with a 5-µL loop (Rheodyne, Rohnert Park, CA). Sample dispersions were monitored at the tubing outlet at 260 nm uisng a Hitachi model 100-10 flow-cell spectrophotometer and recorded on a strip chart recorder. The temperature of the tubing was maintained at 25.0 °C using a constanttemperature bath. Reagents. Reagent or HPLC grade 1-octanol, n-decane, benzene, toluene, ethylbenzene, and o-xylene (Aldrich, Milwaukee, WI) were used as received. Avoid skin and eye contact and breathing vapors, particularly with benzene, a known carcinogen. Extraction Procedure. A fixed concentration of decane internal standard (0.100% v/v) was prepared in the extracting solvent, 1-octanol. Aqueous standard mixtures of benzene, toluene, ethylbenzene, and o-xylene (BTEX) were freshly prepared at various concentrations ranging from 5 ppm to 1 ppb (v/v or w/v). A 0.500-mL aliquot of the mixture was placed in the 1-mL vial with a stir bar. The Hamilton 7105 syringe was rinsed and primed at least 20 times with the solvent/internal standard. After the uptake of 1.00 µL of solvent, the needle was used to pierce the vial septum, and the syringe was clamped into place such that the tip of the needle was located in a consistent position in the headspace. The syringe plunger was depressed, exposing the drop, and stirring commenced for various times at various stirring rates. After stirring, the drop was retracted and injected into the GC or GC/MS for analysis. The analytical signal was the peak area ratio of the analyte to the internal standard. Determination of Diffusion Coefficients. Diffusion coefficients of benzene, toluene, ethylbenzene, and o-xylene in 1-oc5652
Analytical Chemistry, Vol. 73, No. 23, December 1, 2001
Figure 2. Mass chromatograms for benzene, toluene, ethylbenzene, and o-xylene at aqueous concentrations of (a) 10 ppb, (b) 1 ppb. The break in the baseline between benzene and toluene is from the change in the ion being monitored (m/z 78 for benzene; m/z 91 for the other compounds).
tanol were determined by the Taylor dispersion method.13,14 The radius of the tubing was determined to be 0.0268 cm from the measured flow rate, elution time, and tubing length. Peak widths were determined by the intersection of the tangents to the points of inflection with the baseline. Elution times were determined by the intersection of the tangent lines. The diffusion coefficient (D) was calculated according to
σt2 )
a 2t 24D
(1)
where σt2 is the variance (σt is 1/4 the width at the baseline), a is the radius of the tubing, and t is the elution time. In all cases, three replicate determinations were performed. For the octanol diffusion studies in which it was necessary to calculate the diffusion coefficient from the extraction data, it was necessary to directly calibrate the response of the GC using standard solutions of BTEX (1 × 10-4 M to 5 × 10-3 M) in the same 1-octanol/internal standard solution used for the extractions with the same syringe and priming procedure. RESULTS AND DISCUSSION Quantitative Analysis of Model Compounds. Replicate 5-min headspace extractions of a series of BTEX standards in water (1-5 ppm (w/v)) were performed to generate analytical calibration curves at a stirring speed of 1000 rpm. Even though the system is not at equilibrium (discussed below), the response is linear for all four analytes (R2 ) 0.99 for benzene and 0.98 for toluene, ethylbenzene, and o-xylene). Linear response under nonequilibrium conditions for headspace solid-phase microextraction (SPME) has been demonstrated elsewhere.15 As is the case for SPME, it is not necessary to know the actual concentrations of BTEX in the organic phase. To estimate the detection limits of this system, the GC/MS was employed using single-ion monitoring (SIM). Figure 2 shows the mass chromatograms for the four compounds at 10 ppb and 1 ppb (v/v) with a stirring rate of 1500 rpm and a stirring time of 5 min. Thus, these compounds can be determined at the low parts(13) Taylor, G. Proc. R. Soc. A 1953, 219, 186-203. (14) Cussler, E. L. Diffusion, Mass Transfer in Fluid Systems; Cambridge University Press: Cambridge, 1984; Chapters 4, 5. (15) Ai, J. Anal. Chem. 1998, 70, 4822-4826.
Table 1. Concentration-Based Equilibrium Distribution Constants for Benzene, Toluene, Ethylbenzene, and o-Xylene at 25 °C
a
Kow Khwb Kohc
benzene
toluene
ethylbenzene
o-xylene
135 0.224 603
490 0.269 1820
1410 0.310 4550
1320 0.186 7100
a Refs 18-20. b From Henry’s Law constant data.21 KhwKoh.
c
From Kow )
per-billion level using this method, well below guidelines established by the Environmental Protection Agency (EPA) for drinking water.16 It is noteworthy that the solvent, 1-octanol, elutes after the analytes and internal standard. This is a consequence of the need for a solvent with a low vapor pressure so as to preclude the evaporation of the drop during the extraction. 1-Octanol has a very low vapor pressure (0.07 mm Hg at 25 °C).17 To avoid saturation of the MS detector, data acquisition must be stopped prior to the elution of the solvent. Equilibrium Distribution Constants. The distribution of analyte between the three phases may be described by the following equilibrium constants,
Khw ) Ch/Cw
(2)
Koh ) Co/Ch
(3)
Kow ) Co/Cw ) KhwKoh
(4)
where Khw is the headspace-water distribution constant, Koh is the octanol-headspace distribution constant, and Kow is the (overall) octanol-water distribution constant. Literature values for these equilibrium constants are shown in Table 1. All four analytes have relatively small Khw (small Henry’s Law constants). It might, therefore, appear to be reasonable to assume that equilibrium between the water and headspace is established quickly. On the other hand, all four analytes have relatively large Koh equilibrium constants, indicating the need for a large flux of analyte from the headspace to octanol (and hence, depletion of the headspace). The rate-limiting step is not immediately obvious from the above considerations. Control of Extraction Kinetics. Mass transfer in the headspace is assumed to be a fast process, because diffusion coefficients in the gas phase are typically ∼104 times greater than corresponding diffusion coefficients in condensed phases.14 Furthermore, convection is induced in the headspace by the stirring of the aqueous phase. Therefore, the rate of extraction is limited by slow mass transfer in aqueous phase or slow mass transfer in the octanol or both. In general, it was noted that variations in stirring speed resulted in moderate changes in the analytical signal (amount extracted); however, even under stagnant conditions, significant amounts of analyte were extracted. To understand the rate-limiting step(s) in the overall extraction process, plots of the peak area ratio vs the stirring time were generated for BTEX (16) U.S. Environmental Protection Agency Office of Water website (http:// www.epa.gov/safewater/mcl.html). (17) Stull, D. R. Ind. Eng. Chem. 1947, 39, 517.
Figure 3. Plot of peak area ratios of benzene (b), toluene (9), ethylbenzene (2), and o-xylene ([) to decane (internal standard) versus stirring time at 1500 rpm. The aqueous phase concentrations were all 1 ppm (v/v). Error bars represent (1 standard deviation of 4-5 measurements. The solid lines represent fits to an equation of the form y ) a(1 - exp(-kt)). Table 2. Extraction Rate Constants for Benzene, Toluene, Ethylbenzene, and o-Xylene at 500, 1000, 1500, and 1500 rpm with Agitation of the Drop k (min-1) compd
500 rpm
1000 rpm
1500 rpm
1500 rpm, agitated
benzene toluene ethylbenzene o-xylene
0.55 ( 0.05 0.44 ( 0.06 0.31 ( 0.06 0.27 ( 0.05
0.59 ( 0.05 0.46 ( 0.03 0.32 ( 0.04 0.30 ( 0.03
0.66 ( 0.03 0.50 ( 0.03 0.40 ( 0.03 0.36 ( 0.03
1.1 ( 0.1 0.73 ( 0.06 0.50 ( 0.04 0.45 ( 0.04
under the following stirring conditions: 500, 1000, 1500, and 1500 rpm with agitation of the drop. A typical plot is shown in Figure 3. The precision obtained varies from 9 to 11% RSD for the 1-min extractions down to 1-2% RSD for the 5-min extractions in which the system is closer to equilibrium. Even under conditions of gentle mechanical drop agitation, the drop remains stable throughout the extraction. Extraction rate data such as that shown in Figure 3 are fit to an equation of the form y ) a(1 - exp(-kt)), a simple first-order kinetic model that adequately fits extraction rate data both for two-phase systems22 and three-phase systems.15,23,24 The parameter a represents the equilibrium value (peak area ratio in this case). The parameter k represents a rate constant for the extraction process. Detailed theoretical interpretations have been presented elsewhere.15, 22-24 Extraction rate constants (k) are shown in Table 2 for BTEX extractions at 500, 1000, 1500, and 1500 rpm with drop agitation, along with error estimates from the curve fits. There is clearly an enhancement in the rate of extraction with faster stirring, which suggests that aqueous-phase mass transfer is a limiting step in (18) Fujita, T.; Iwasa, J.; Hansch, C. J. Am. Chem. Soc. 1964, 86, 6, 5175-5180. (19) Iwasa, J.; Fujita, T.; Hansch, C. J. Med. Chem. 1965, 8, 150-153. (20) Hansch, C.; Leo, A. Substituent Constants for Correlation Analysis in Chemistry and Biology; Wiley: New York, 1979. (21) NIST Chemistry Webbook (http://webbook.nist.gov/chemistry/). (22) Jeannot, M. A.; Cantwell, F. F. Anal. Chem. 1996, 68, 2236-2240. (23) Ma, M.; Cantwell, F. F. Anal. Chem. 1998, 70, 3912-3919. (24) Ma, M.; Cantwell, F. F. Anal. Chem. 1999, 71, 388-393.
Analytical Chemistry, Vol. 73, No. 23, December 1, 2001
5653
Table 3. Surface Concentrations of Benzene, Toluene, Ethylbenzene, and O-Xylene in 1-Octanol Obtained Experimentally and Assuming No Resistance to Mass Transfer in the Aqueous and Headspace Phases. compd
exptl Co,s (mol/cm3) × 106
theor Co,s (mol/cm3) × 106 a
benzene toluene ethylbenzene o-xylene
0.69 0.90 1.2 1.2
1.51 4.77 11.5 10.9
a K C ; assuming no resistance to mass transfer in the aqueous ow w or headspace phases.
Figure 4. Plot of peak area ratios of benzene (b), toluene (9), ethylbenzene (2), and o-xylene ([) to decane (internal standard) versus stirring time at 1500 rpm. The aqueous phase concentrations were all 1 ppm (v/v). Error bars represent (1 standard deviation of 4-5 measurements. The solid lines represent fits to an equation of the form y ) kt1/2.
the overall extraction process. However, slow mass transfer in the aqueous phase alone does not account for the overall rate, because a significant enhancement is observed when convection is induced in the octanol by vibration of the needle. Thus, both the stirring rate of the aqueous phase and the degree of convection in the organic drop influence the observed overall rate of extraction. Although the aqueous phase is stirred and convection is certainly present in the headspace as well, the octanol drop was found to be stagnant through observation of small particles of charcoal placed in the octanol (unless, of course, the needle was agitated directly). Therefore, mass transfer into the octanol is by diffusion alone, and furthermore, according to the data in Table 2, this represents a slow step in the overall extraction process. To determine if diffusion in octanol alone controls the rate of extraction at high stirring speeds (i.e., 1500 rpm), the data in Figure 3 were fit to a semi-infinite diffusion model shown below,25
n ) 2ACo,sD1/2π-1/2t1/2
(5)
where n is the number of moles extracted after time t, A is the surface area of the drop (estimated to be 0.04 cm2),5 Co,s is the surface concentration of the BTEX compounds in the drop, and D is the diffusion coefficient in octanol. If mass transfer in the aqueous and headspace phases is fast, then Co,s is equal to KowCw (Cw remains approximately constant during short (e5 min) time intervals.) It is reasonable to approximate the diffusion here as linear and semi-infinite when times are small.25 For example, the root-mean-square distance traveled by a molecule ((2Dt)1/2), even after 5 min, is only ∼0.05 cm, significantly smaller than the diameter of the drop (∼0.12 cm). The data fit to an equation of the form y ) kt1/2 is shown in Figure 4. The data fit a t1/2 dependence reasonably well, although the fit is poorer than that shown in Figure 3. To fit the data to eq 5, the peak area ratios were converted to moles extracted via direct calibration of the GC using standard solutions of BTEX in 1-octanol with internal standard. (25) Crank, J. The Mathematics of Diffusion, 2nd ed.; Oxford University Press: New York, 1975.
5654 Analytical Chemistry, Vol. 73, No. 23, December 1, 2001
Diffusion coefficients determined by the Taylor dispersion method for benzene, toluene, ethylbenzene, and o-xylene in 1-octanol were 4.3 ( 0.2, 3.9 ( 0.4, 3.38 ( 0.08, and 2.98 ( 0.04 × 10-6 cm2/s, respectively. Octanol surface concentrations calculated from the fit to eq 5, using these diffusion coefficients, are shown in Table 3, along with those predicted for the case of no resistance to mass transfer in the aqueous or headspace phases (i.e., Co,s ) KowCw). The estimated errors for the surface concentrations are about (25% considering the uncertainty in the drop area and the literature values for Kow. Clearly the Co,s values obtained from the fit to the semi-infinite diffusion model in octanol are much smaller than those predicted by neglecting resistance in the other phases, by almost an order of magnitude in some cases. The surface of the octanol is not at equilibrium with the aqueous phase, even at fast stirring rates. This is strong evidence that diffusion of solutes into 1-octanol is not the only slow step in the overall mass transfer process. There is another slow mass transfer step elsewhere in the system, presumably in the aqueous phase. Significance and Potential of the Technique. Headspace solvent microextraction (SME) reported herein and headspace solid-phase microextraction (SPME)4 appear to have similar capabilities in terms of precision and speed of analysis; however, headspace SME appears to offer two distinct advantages over headspace SPME. First, the choice of solvents is virtually unlimited, as compared to the number of phases currently available for SPME. Second, the cost of microliters of solvent for SME is negligible compared to the cost of commercially prepared SPME fibers. On the other hand, SPME offers the advantage that there is no solvent peak in the chromatogram, and splitless “injection” can be employed. Related current and future work in this laboratory is focused on exploring the use of other solvents, especially polar solvents, for the analysis of a wider range of analytes. The potential to couple SME with other separation methods such as high performance liquid chromatography (HPLC) and capillary electrophoresis (CE) is also under consideration. ACKNOWLEDGMENT This research was supported by an award from Research Corporation and funding from St. Cloud State University. Received for review July 18, 2001. Accepted September 30, 2001. AC015569C
