Anal. Chem. 1994,66, 3543-3551
Simple Fiber-optic Interface for On-Line Supercritical Fluid Extraction Fourier Transform Infrared Spectrometry Daniel L. Heglund and David C. Tllotta' Department of Chemistty, University of North Dakota, Grand Forks, North Dakota 58202 Steven B. Hawthorne and David J. Miller Energy C? Environmental Research Center, University of North Dakota, Grand Forks, North Dakota 58202
A simple and inexpensive fiber-optic interface for coupling a supercritical fluid extractor to a FT-IR spectrometer is described. The interface, based on chalcogenide (AsSeTe) fiber optics and a stainless steel union cross, is rugged up to pressures of 400 atm, has a spectral window from 5000 to 800 cm-I, and is relatively easy to clean. It is shown that the SFEFT-IR interface allows qualitativeand quantitativeinformation to be obtainedfrom the extracts of real-worldsamples (caffeine in coffee and total petroleum hydrocarbonsin soil, respectively) without the use of either a flow restrictor or a collectionsolvent. The determination of total petroleum hydrocarbons (TPH) in soils by static on-line SFE-FT-IR yields values in good agreement with both off-line SFE and Soxhlet extraction. The detection limit for TPH, using a 48-s scan time, was determined to be 1.6 ppm (wt TPH/wt sample). The linear dynamic range of calibration for the TPH determinations was greater than 3 orders of magnitude.
Supercritical fluid extraction (SFE) has been shown to be an excellent alternative to conventional solvent extraction for the removal of organic compounds from solid Generally, SFE is fast ( 100%) and positive absorbances would be obtained at high analyte concentrations ( T C 100%). Although these absorbances are (29) Hecht, E. Opfics; Addison-Wesley Publishing: Reading, MA, 1990: p 60.
\
6.02-
;
0.01-
I
S 0
b
m.00-
1
d
n c e
-0.01-
-0.02-
b -0. 0 3 5 1900
I
I
I
I
I
I
1600
1700
1600
lS00
1480
1380
I 1200
I
I
I
llcl0
tu00
900
I 600
Wavenumbors
Figure 6. Infrared absorbance spectra of caffeine obtained (A) by subtraction of on-line spectrum of decaffeinated coffee SFE extract from on-line spectrum of caffelnated coffee SFE extract and (B) in CHpClp,The spectra have been offset for clarity.
proportional to analyte concentration (and can be used quantitatively as long as the calibration is performed under the same experimental conditions), they are sensitive to variations in the pressureor temperatureof the SFE-IR system. Thus, the experimentally obtained absorbance values were converted to corrected absorbance values by spectral normalization. The procedure used here is similar to other methods developed to normalize diffuse reflectance spectra for calibration p10ts.l~ Specifically, the transmittance spectrum is normalized to 100% transmittance using a region of the spectrum where neither the analyte nor the matrix absorbed radiation. The actual transmittance of a band that absorbs was then a fraction of the original value. For example, if the uncorrected transmittance at a specified nonabsorbing wavenumber was 140% and that of a C-H stretching band was presented as 70%, then the computed “corrected” transmittance is (70/140) X 100, or 50% corrected transmittance. This transmittance spectrum was then converted to a “corrected” absorbance spectrum for quantitation. For this work, the nonabsorbing transmittancevalue was obtained at 2755 cm-l. SpectralSubtraction. Spectral subtraction is useful in SFE in order to remove spectral interferences arising from both the C02 matrix and the coextracted compounds (assuming that spectral information concerning these coextracted compounds can also be obtained). Of course, in order to eliminate the bands of the C02 matrix, the spectra must be acquired under similar pressure and temperature conditions. An online SFE-FT-IR spectrum of caffeinated coffee after 30 min of SFE is shown in Figure SA in the region from 4000 to 800 cm-l. The infrared bands shown in this spectrum are due to caffeine, natural oils and aromas, the C02 background, and the water modifier (ca. 1600 cm-l). Figure 5B shows an online SFE-FT-IR spectrum of decaffeinated coffee in the same spectral region. Notice that many of the bands present in the caffeinated coffee spectrum are also present in the decaffeinated coffee spectrum. The caffeine bands in thecaffeinated
coffee spectrum can be spectrally isolated by subtracting the decaffeinated coffee spectrum from the caffeinated coffee spectrum. The resultant subtraction spectrum is shown in Figure 6A and has been plotted in the 1900-800-~m-~region. (The C-H stretching region is not shown in Figure 6A because of subtraction artifacts arising from the strong absorption of the coffee oils and aromas.) For comparison, Figure 6B shows an IR spectrum of caffeine in CH2C12. Note that the caffeine spectrum is nearly cleanly recovered by spectrally subtracting the two SFE coffee extract spectra (in fact, the SFE-FT-IR spectrum of caffeine has a nearly straight baseline as compared with the sloping base line of the caffeine reference spectrum). The subtle differences in the band positions of caffeine in the C02 matrix have been reported previously and do not arise from the subtraction process.17
Determination of TPH in Soil by On-Line SFE-Fl’-IR Spectroscopy. The ability to obtain quantitative information with the fiber-optic SFE system was tested by determining TPHs in soils. A totally static extraction (no CO2 flow once the SFE cell was filled) at 340 atm and 77-80 OC was utilized to remove the TPH components from the samples. Transfer of the extracted organics into the IR cell was facilitated by holding the IR cell at atmospheric pressure until the extraction was complete. Following extraction, the valve separating the SFE cell from the IR cell was opened, and the extracted components were transferred as the C02 depressurized to about 250 atm. Examples of the raw infrared spectra of the SFE extracts (Le., without transmittance correction or subtraction of the C02 background) are shown in Figure 7. These spectra were obtained at an interface pressure and temperature of 250 atm and 25 “C, respectively. Figure 7A shows an infrared spectrum of the gasoline-contaminated soil extract after 30 min of SFE, Analytical Chemistty, Vol. 66, No. 20, October 15, 1994
3549
140-
'a
120%
r
100-
r
a
"
80-
m 1 t t a
60-
I
; 40e
I
20-
4
O , t 4000
3600
3200
2800
2400
I
1200
2000
Wavenumba r s
%
100-
T P
a
80-
S
m
60t t a
n
B
40-
C
e
20---
0 a000
11
,,I.,
3600
1
I
I
I
3200
2800
2400
,
I
I
2000
1600
( 7 1200
Figure 7. On-line infrared spectra of (A) gasoline-contaminated soil SFE extract and (6)diesel fuel-contaminated soil SFE extract. SFE conditions: 340 atm and 77-80 O C for 30 min.
and Figure 7B shows an infrared spectrum of the diesel fuelcontaminated soil extract after 30 min of SFE. Both spectra show strong C-H stretching vibrations in the region of 3000 cm-1. In addition, it should be noted that both spectra show the presence of water. The spectrum in Figure 7A shows both the broad 0-H stretching vibration at ca. 3400 cm-l (which overlaps the strong C02 absorption at 3700 cm-l) and the bending vibration at ca. 1600 cm-l. The spectrum in Figure 7B, which was obtained from a sample with a lower concentration of water, shows only the bending vibration at ca. 1600 cm-l. Quantitation was accomplished by monitoring the 2932cm-l C-H stretching absorption (with base line correction, as described above) and employing a slightly unconventional application of Beer's law. Strict application of Beer's law 3550
Analytical Chemistry, Vol. 66,No. 20, October 15, 1994
employs units of concentration that are dependent upon the volume or mass of the solution (e.g., molarity or ppm). However, when using a variable density solvent such as CO2, conventionalconcentration units are awkward because changes in pressure (at constant cell volume) change the volume or mass of the solvent (and hence, the volume of solution) in the cell. In order to simplify the on-line experimental procedure, the TPH was determined in units of ppmv using the system dead volume (the entire system volume less the volume occupied by the sample) rather than the mass or volume of the C02. When the dead volume of the system is utilized rather than the volume of the C02 solvent, the concentration of the solute is independent of the system temperature or pressure. Parts per million by volume is converted to ppm (wt solute/wt sample) with
~~
~~
Table 1. Comparison of On-Line SFE-R-IR wlth Off-Line SFE-IR and Soxhlet Extractlon Determinations of Total Petroleum Hydrocarbons for Selected Sol1 Samples
TPH, ppm (RSD, % ) b
where Cppmis the concentration in pg/g of soil, Cppmv is the concentration in pL/L, D is the density of the EPA reference oil (0.83 g/mL), vd is the dead volume of the system (in L), and W is the weight of the soil sample. The factor of 1000 in the numerator is used for unit conversion. The total dead volume (the volume of the system with a sample in the SFE cell) was determined at room temperature using the volume information on the control panel of the ISCO pump. First, the initial (empty) volume of the system was determined by measuring the amount of COZat 185 atm that was needed to fill the entire system. This volume is needed in order to determine the dead volume of the IR cell and the interconnecting plumbing between the SFE and IR cell. The outflow of COZfrom the pump was deemed to be at equilibrium when the CO2 flow rate (according to the pump readout) was C0.05 mL/min, which generally required 1-3 min. (Note that temperature equilibration would be faster if the water jacket available with the 260D pump was utilized.) Next, the IR cell was sealed-off from the SFE cell, and the dead volume of the empty SFE cell was determined. The two dead volumes were then subtracted (the dead volume of the empty system less the dead volume of the SFE cell) to yield the dead volume of the IR cell and interconnecting plumbing. Since this volume is constant for a given experimental arrangement, this measurement needs to be performed only once and noted. Next, the dead volume of the SFE cell with sample was determined. This was accomplished by simply noting the amount of C02 at 185 atm required to fill the SFE cell with a sample in it. Finally, the total dead volume was obtained by adding the dead volume of the IR cell and interconnecting plumbing to the dead volume of the SFE cell with sample. This procedure for determining the system dead volume was verified using the known volume of a series of glass beads (determined by water displacement). A plot of the volume of the glass beads determined by water displacement versus the volume of the glass beads determined by COZdisplacement yielded a straight line 0,= 0.205 0 . 9 8 6 ~ and ) a regression coefficient (r) of 0.995. The reproducibilities of the volume measurements performed in this fashion were typically
