Development of an Infrared Hollow Waveguide as a Sensing Device

The Effect of Pressure and Temperature on Mid-Infrared Sensing of Dissolved Hydrocarbons in Water. Charles Heath , Matthew Myers , and Bobby Pejcic...
1 downloads 0 Views 100KB Size
Download PDF
Anal. Chem. 1999, 71, 3740-3746

Development of an Infrared Hollow Waveguide as a Sensing Device for Detection of Organic Compounds in Aqueous Solutions Jyisy Yang,* Jhy-Woei Her, and Sheng-Hsi Chen

Chung-Yuan Christian University, Chung-Li, Taiwan

In this paper, a new detection method based on an infrared hollow waveguide is developed to detect semivolatile to nonvolatile organic compounds in aqueous solutions. The hollow waveguide is produced by chemical deposition of silver on the inner surface of a polyethylene tube. The surface of the silver layer is further coated with a hydrophobic film to attract organic compounds in aqueous solution. Samples were pumped through this hollow waveguide sampler and organic compounds were attracted onto the hydrophobic film. After removal of the residual water molecules in the hollow waveguide sampler, organic compounds can be sensed by conventional Fourier transform infrared (FT-IR) spectrometry. Theoretical aspects of this type of sampler are also presented. The derived analytical equations for this type of sampler were consistent with experimental data. Under the condition of constant hydrophobic film volume, high linearity (R2 equal to 0.9993) between the concentration of analyte and the detected signal was obtained for concentrations in the range from 2.5 ppm to 50 ppb. By co-adding 100 scans with 4 cm-1 resolution, the typical detection limit in this type of sensing method can be lower than 10 ppb. Several factors such as sampling flow rate, sampling time, and hydrophobic film volume were also investigated in this work. In recent years, Fourier transform infrared (FT-IR) spectroscopic methods have been increasingly applied to detect organic compounds in aqueous solutions. To eliminate the effects of water’s strong absorbance, the attenuated total reflection (ATR)1 effect is used as the method for detection of organic species in aqueous solutions. Different forms of ATR sensing crystals have been used to detect organic compounds in aqueous solutions.2-6 To increase the sensitivity, a solid-phase microextraction (SPME) technique7-9 is used in conjunction with the ATR crystal. A thin * To whom correspondence should be addressed. Fax: 886-345-63160. E-mail: [email protected]. (1) Harrick, N. J. Internal Reflection Spectroscopy; Wiley: New York, 1967. (2) Messica, A.; Greestein, A.; Katzir, A. Appl. Opt. 1996, 35, 2274-2284. (3) Simhony, S.; Katzir, A.; Kosower, E. M. Anal. Chem. 1988, 60, 1908-1910. (4) Gobel, R.; Krska, R.; Kellner, R.; Kastner, J.; Lambercht, A.; Tacke, M.; Katzir, A. Appl. Spectrosc. 1995, 49, 1174-1177. (5) Paul, P. H.; Kychakoff, G. Appl. Phys. Lett. 1987, 51, 12-14. (6) Newby, K.; Reichert, W. M.; Andrade, J. D.; Benner, R. E. Appl. Opt. 1984, 23, 1812-1815. (7) Zhang, Z.; Pawliszyn, J. Anal. Chem. 1993, 65, 1843-1852.

3740 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

film of hydrophobic materials is coated on the internal reflection element (IRE). The coated film serves two general purposes: the exclusion of water molecules from the penetration depth and the attraction of organic species near the sensing crystal. Sensitivity can be increased in this way. Several hydrophobic materials, such as low-density polyethylene (LDPE),10,11 poly(vinyl chloride),12 polyisobutylene (PIB),13 and silicone,14 have been studied for their application to signal enrichment for organic compounds in aqueous solutions. Although present ATR-IR spectroscopic methods provide direct and convenient ways to detect organic compounds in an aqueous solution, the sensitivity is still limited because of the short penetration distance of evanescent waves out of the ATR crystal. In recent years, sensing devices based on waveguides have been developed to detect various organic species in gas phase or liquid solution.15-18 For example, Dasgupta et al.15 used gas-permeable liquid core waveguides to detect a variety of gases. Fluoride polymers are typical materials to form the molecule-permeable waveguides. The strong IR-absorbing nature of polymers limits the extension of this type of method into the IR regions. To maintain the advantage of the FT-IR spectroscopic method and to increase the sensitivity of FT-IR spectrometry in detection of organic compounds in aqueous solution, a new detection method based on an IR hollow waveguide19-21 is proposed herein. Since the method to produce an IR hollow waveguide was proposed by Harrington et al.,19 several applications have been found, especially in guiding CO2 laser radiation for medical applications.22 In this (8) Zhang, Z.; Pawliszyn, J. J. High Resolut. Chromatogr. 1993, 16, 689. (9) Zhang, Z.; Pawliszyn, J. Anal. Chem. 1995, 67, 34-43. (10) Krska, R.; Kellner, R.; Schiessel, U.; Tacke, M.; Katzir, A. Appl. Phys. Lett. 1993, 63, 1868-1870. (11) Heo, J.; Rodrigues, M.; Saggese, S.; Sigel, G. H., Jr. Appl. Opt. 1991, 30, 3944-3951. (12) Ertan-Lamontagne, M. C.; Lowry, S. R.; Seitz, W. R.; Tomellini, S. A. Appl. Spectrosc. 1995, 49, 1170-1173. (13) Gobel, R.; Krska, R.; Kellner, R.; Seitz, R. W.; Tomellini, S. A. Appl. Spectrosc. 1994, 48, 678-683. (14) Blair, D. S.; Burgess, L. W.; Brodsky, A. M. Anal. Chem. 1997, 69, 22382246. (15) Dasgupta, P. K.; Genfa, Z.; Porruthoor, S. K.; Caldwell, S.; Dong, S.; Liu, S.-Y. Anal. Chem. 1998, 70, 4661-4669. (16) Altkorn, R.; Koev, I.; Gottlieb, A. Appl. Spectrosc. 1997, 51, 1554-1559. (17) Tsunoda, K.; Nomura, A.; Yamada, J.; Nishi, S. Appl. Spectrosc. 1990, 44, 163-165. (18) Liu, S.-Y. U.S. Patent 5,412,750, May 2, 1995. (19) Gregory, C. C.; Harrington, J. A. Appl. Opt. 1993, 32, 5302-5309. (20) Abel, T.; Hirsch, J.; Harrington, J. A. Opt. Lett. 1994, 19, 1034-1036. (21) Croitoru, N.; Dror, J.; Gannot, I. Appl. Opt. 1990, 29, 1805-1809. (22) Harrington, J. A.; Gregory, C. C. Opt. Lett. 1990, 15, 541-543. 10.1021/ac9903252 CCC: $18.00

© 1999 American Chemical Society Published on Web 08/03/1999

Here, K is the partition coefficient, Corg is the concentration of analyte in the hydrophobic layer, and Caq is the concentration of solute in aqueous solution. In terms of number of analyte molecules in the hydrophobic layer, the above equation can be derived as follows:

no ) KVfVsCo/(Vs + KVf)

Figure 1. (A) Adsorption model of hollow waveguide sampler. (B) Detection of analytes adsorbed inside the hollow waveguide sampler.

work, we extended the hollow waveguide to sense organic compounds in aqueous solution by joining it with the principle of SPME. The inner surface of the waveguide is coated with a hydrophobic polymer which serves as the SPME coating. When aqueous samples are passed through the tube, organic species can be adsorbed into the SPME layer for further analyses by FTIR spectrometry. Figure 1A shows the basic diagram of the structure of the hollow waveguide sampler for the sampling of aqueous solution. After a certain amount of sample is pumped through the sampler, the adsorbed analytes in this hollow waveguide sampler can be detected by directing IR radiation into one end of the sampler as shown in Figure 1B. Unlike ATR methods, for which analytes need to diffuse into the depth of penetration, any organic species adsorbed onto the SPME layer of the hollow waveguide sampler can be detected by the reflection of the IR radiation inside the hollow waveguide. This reveals that the path length of IR radiation is significantly increased in the hollow waveguide sampler, and consequently, the sensitivity can be greatly increased. This type of sampler can also provide several other advantages as compared with the ATR method: (1) It is highly possible to perform mass analysis because the sampling and detection steps are separate. (2) There is no limitation on sample volumes. (3) High sampling flow rates can be used because no back-pressure is built in the hollow sampler. (4) Mass production of the sampler is possible because the cost of producing a piece of a hollow waveguide is much lower than that of producing an ATR crystal. (5) There is higher sensitivity in detection than with ATR because of the long path length of radiation and higher sampling flow rate. THEORETICAL TREATMENTS The hollow fiber sampler employs a hydrophobic film to trap any organic analytes pumped through the hollow fiber, and then the analytes are detected using FT-IR spectrometry in internal multireflection mode. The principle of trapping organic compounds from aqueous solution that is passed through the tube is adaptable from the principle of SPME. The partition coefficient in a twophase system is of the most concern in SPME, and it can be expressed as the ratio of the concentrations of solute in hydrophobic and aqueous phases and as shown in eq 1:

K ) Corg/Caq

(1)

(2)

Here, no is the number of analyte molecules in the organic layer, Vf is the volume of the hydrophobic layer, Vs is the volume of aqueous solution, and Co is the initial concentration of the analytes in the aqueous sample. In SPME, the partition coefficient is assumed to be large and KVf . Vs. Therefore, the above equation can be simplified to

no ) VsCo

(3)

Because partition coefficients for most organic compounds are not large enough to meet the condition of KVf . Vs under normal circumstances, in the application of eq 3 one typically keeps the Vs and Vf constant to meet the requirement for quantitative analysis. When a flow cell is used to detect analytes through the principle of SPME, eq 3 is no longer valid because the sample volume is greatly increased. Assuming the volumes of hydrophobic film and sample are 10-3 and 100 mL, respectively, the partition coefficient should be larger than 107 in order to meet the requirement of KVf . Vs (within 1% error). This large K value is not generally obtainable for organic compounds in aqueous solutions. By assuming the sample volume is large enough and much larger than KVf, eq 3 can be expressed in the following simple form:

no ) KVfCo

(4)

This equation indicates that the larger the volume of hydrophobic film, the larger the signal that can be obtained. It also indicates that the variation of the analytical signal caused by different film thickness can be completely removed because the original concentration is directly proportional to no/Vf. As indicated by Ai,23 SPME is operated in nonequilibrium for most of the detection. On the basis of Fick’s first law of diffusion, the author also derived the relationship between the number of analyte molecules and the adsorption time as follows:

n ) no[1 - exp(-at)]

(5)

where t is the equilibrium time and a can be expressed as eq 6.

a ) 2A(m1m2KVf + m1m2Vs)/(m1VsVf + 2m2KVsVf)

(6)

Here, A is the surface area of the SPME phase and m1 and m2 are mass transfer coefficients in the sample matrix and in the SPME polymer, respectively. Parameter a is a measure of how fast an adsorption equilibrium can be reached in the SPME process. Providing the volume of sample is much larger than the (23) Ai, J. Anal. Chem. 1997, 69, 1230-1236

Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

3741

value of KVf, eq 6 can be simplified for a hollow waveguide sampler to be as follows:

a ) 2A(m1m2)/(m1Vf + 2m2KVf)

(7)

By substitution of eq 4 and eq 7 into eq 5, the following equation can be obtained for a hollow waveguide sampler:

n ) KVfCo[1 - exp(-at)]

(8)

This equation can be used to monitor the analytical signals in nonequilibrium conditions. Equation 8 can also be simplified into eq 4 for equilibrium conditions (large value of at). One important drawback is that the film volume in the sampler affects the analytical signal in both the equilibrium and nonequilibrium situations. For example, the amount of analyte in equilibrium conditions is proportional to film volume as in eq 4, and n/Vf should be directly proportional to the original concentration of analyte in the sample solution. In nonequilibrium situations, film volume also affects the value of the coefficient a as in eq 7. EXPERIMENTAL SECTION Materials and Reagents. Polyethylene (PE) tubes were used as the substrate of the hollow waveguide and were obtained from a local supplier. The internal and external diameters of the PE tube were 1.5 and 2.2 mm, respectively. PE tubes were cut to 1 m in length before chemical deposition of the inner surface was performed. Chemicals to form the plating solutions were reagentgrade. These chemicals include silver nitrate and glucose (Prochem Inc., Rokford, IL), sulfuric acid, chromic acid, SnCl2, and PdCl2 (Acros Organics, New Jersey). Polyisobutylene (PIB) was obtained from Aldrich (Milwaukee, WI). Toluene (TEDIA, Fairfield, Ohio) was used to dissolve PIB and 1-chloronaphthalene (1-CN), chlorobenzene (CB), 2-chlorotoluene (2-CT), and trichloroethylene (TCE) were used as probe molecules to represent different volatility compounds which were obtained from Merck (Schuchardt, Germany). Production of a Hollow IR Waveguide Sampler. The inner surfaces of the PE tubes were chemically coated with a thin layer of silver, according to a method and formula described in the literature.24 The PE tubes were first cleaned by soaking in ethanol for 3 min. The PE surface was further treated with a sulfochromic bath (K2Cr2O7, 79 g/L; H2SO4, 863 mL/L) for 1 min. Palladium chemisorption by the one-step process (sensitization/activation process) was carried out by immersing the substrate for 3 min in the mixed solution (12 g/L SnCl2, 0.25 g/L PdCl2, 60 mL/L HCl). After rinsing with deionized water, a metallization operation was performed by separately pumping a silver solution (35 g/L AgNO3, 100 mL/L NH4OH, 25 g/L NaOH) and a reduction solution (4.5 g/L glucose, 4 g/L tartaric acid, 100 mL/L ethanol) through the PE tubes. Each solution was pumped for 3 min. The pumping speed was controlled at 10 mL/min. The produced hollow waveguide possessed a transmission of infrared radiation around 10 dB/m in this work. The produced waveguide was cut to 17 cm long and further coated with PIB on the inner surface of the (24) Charbonnier, M.; Alami, M.; Romond, J. M. J. Electrochem. Soc. 1996, 143, 472-480.

3742 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

Figure 2. (A) Equipment for sampling analytes in aqueous solution. (B) System to remove the residual water molecules. (C) Setup for detection of adsorbed analytes.

waveguide. PIB was first dissolved in toluene to form 0.5-5% (w/ v) solutions. A funnel was placed on the top of the hollow waveguide to allow the passage of 25 mL of PIB solutions. After drying, a 2-cm-long segment of the hollow waveguide sampler was removed. The coated sampler was purged with a 150 mL/min nitrogen flow for 10 min to remove the solvent. Scans were acquired to obtain the spectra of PIB in the coating. On the basis of the integrated peak intensities around 900 cm-1, the averaged relative standard deviation of the PIB film thickness for four replicated runs in five different PIB solutions was 8.82%. To reduce errors caused by the coating, samplers with similar film thickness were selected for the following experiments. On the basis of the observation of infrared signals, it is clear that the toluene can be completely removed after nitrogen purging. The homogeneity of the PIB coating was also examined by obtaining the spectra for every 2.5-cm piece on the hollow waveguide samplers. The average relative standard deviation on the integrated peak intensity around 900 cm-1 was only 1.2%. Procedure of Sampling and Detection. Probe organic compounds were dissolved in water to form the desired concentrations. A microbellows pump (Ikea Co., Tokyo, Japan) was used to draw solution through the sampler, and the setup of the equipment is shown in Figure 2A. Sampling with suction was used to prevent any adsorption of analytes by the tubes of the bellows pump. After sampling, residual water was present in the sampler so nitrogen gas was used to reproducibly remove the residual water molecules. In the setup, a gas flow meter equipped with a flow regulator was used (Cole-Parmer Instrument Co., Niles, IL). Figure 2B shows the equipment used in this work for detection of analytes. Two nonimaging infrared concentrators (modified from Spectral-Tech fiber probe accessory) were placed at each end of the hollow waveguide sampler for concentrating IR radiation from the sample compartment of an FT-IR spectrometer to the hollow waveguide and to redirect the transmitted IR

Figure 4. Effect of nitrogen flow rate (9, 50 mL/min; 2, 200 mL/ min) on the removal of residual water molecules and desorption of 1-CN. 2.5 ppm 1-CN was used as a sample solution. The sample was pumped through the sampler (5% PIB coated) in a sampling flow rate of 4 mL/min for 10 min. Triplicate runs were performed for all the data points.

Figure 3. (A) Noise level of noncoated, 5% PIB coated PE hollow waveguide. Spectra were obtained by co-adding 100 scans with 4 cm-1 resolution. (B) Typical spectra of the PIB coating and 1-CN (2.5 ppm solution and 5% PIB coated) detected with the hollow waveguide sampler. Spectra were obtained by co-adding 100 scans with 4 cm-1 resolution.

radiation back to the MCT detector. A Nicolet Magna 550 FT-IR spectrometer equipped with a medium-range mercury-cadmiumtelluride (1 mm2) detector was used to detect the adsorbed analytes. The typical noise levels of 5% PIB-coated and noncoated hollow waveguides produced in this work are shown in Figure 3A. Each spectrum was obtained by co-adding 100 scans with 4 cm-1 resolution. As can be seen in this figure, the noise level was increased after coating with a thin layer of PIB, but it was still around 1 mAU in the region below 1000 cm-1 for the PIB coated sampler. Several noisy regions can be found in the spectrum due to the absorption of infrared radiation by the PIB coating. PIB has small absorption features in the region lower than 1000 cm-1 as can be seen in Figure 3B. This allows the application of this polymer to the detection of organic compounds in aqueous solution and makes it especially useful for the detection of halogenated compounds. Two small peaks located around 900 cm-1 are used as an indication of the film thickness of the PIB coating. A typical spectrum of 1-CN detected (2.5 ppm) by the method proposed here is also shown in Figure 3B. Two strong absorption peaks located around 700 cm-1 can be found in the spectrum. A peak located at 766 cm-1 was selected as an indication of the amount of 1-CN in this work. RESULTS AND DISCUSSION Removal of Residual Water Molecules in Sampler. Water is a strong infrared absorber. Any residual water after sampling

can affect the infrared signal. Therefore, residual water molecules present in the surface of the SPME phase after sampling should be removed completely. In this work, a stream of nitrogen gas was used to purge the hollow waveguide of residual water molecules. Passing a stream of nitrogen gas can eliminate the water molecules, but it can also partially remove the adsorbed organic compounds in the SPME phase. To remove the water molecules completely and to retain adsorbed organic molecules in the SPME phase as much as possible, the nitrogen flow should be optimized. The hollow waveguide sampler was first exposed to a 2.5 ppm 1-CN solution in a 4 mL/min sampling flow rate for 10 min. The sampler was transferred to the device shown in Figure 2B to purge with two different nitrogen flow rates (50 and 200 mL/min). The spectra of 1-CN were acquired after purging for a certain amount of time. By plotting the 1-CN signals at 766 cm-1 with different purging times, the curves shown in Figure 4 are obtained. As can be seen in this figure, the highest nitrogen flow rate that can effectively remove the residual water molecules is also the most effective in removing the adsorbed 1-CN. With the smaller nitrogen flow rate (50 mL/min), a large error was found in signals for the first 3 min due to incomplete removal of the residual water molecules. To retain organic analytes in the hollow waveguide sampler as much as possible and to completely remove the residual water, a 50 mL/min nitrogen flow rate was selected and 4 min of purging was employed in the following experiments. On the basis of the results of Figure 4, the regeneration of the SPME coated hollow waveguide can be effectively done by passing nitrogen gas through it at a high flow rate. Generally, passing nitrogen at a flowrate of 200 mL/min for 10 min can remove 75% of the adsorbed 1-CN as is shown by the data in Figure 4. However, to further increase the speed of regeneration of the hollow waveguide sampler, a 500 mL/min nitrogen flow was employed for all of the following tests. By regenerating the same hollow waveguide sampler 20 times using 500 mL/min, the obtained standard deviation of PIB signals was 1.5%. This result indicated that the PIB coating of the hollow waveguide is very stable and that the regeneration was also highly efficient. Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

3743

Figure 5. Relationship between 1-CN signal and adsorption time for 24 mL/min (2) and 4 mL/min (9) sampling rate. 2.5 ppm 1-CN and 5% PIB coated hollow waveguide were used as the sampler.

Therefore, a 500 mL/min nitrogen flow rate is used as the regeneration flow rate. Effect of Flow Rate on Adsorption. The sampling flow rate affects the linear flow rate through the hollow waveguide sampler and, hence, affects the adsorption efficiency. Although a fast sampling flow rate reduces the time available for diffusion into the PIB film, a larger volume can be applied to the sampler in a given period of time with one. Sampling flow rates first studied were 4 and 24 mL/min. A 1-CN solution at 2.5 ppm and a 5% PIB coated hollow waveguide sampler were used in these experiments. A plot of the 1-CN peak intensity at 766 cm-1 against the flow rate of sampling is shown in Figure 5. As can be seen in this figure, higher signals were obtained using higher sampling flow rates. This is simply caused by the larger amount of sample being pumped through the sampler. An exponential curve shape was obtained in these runs which may be caused by saturation of the PIB film after the passage of large volumes of sample. Comparing the signal at 10 minutes for both sampling flow rates, we can see that the adsorption efficiency for the higher sampling flow rate is much lower than that for the lower sampling flow rate. For example, if the sample is pumped for 10 minutes, the total volume of sample is 240 mL for the 24 mL/min sampling rate but 40 mL for the 4 mL/min sampling rate. Around 6 times the volume of sample is processed at the higher sampling flow rate as compared with that processed at the lower rate, but the signal is only increased to 3 times that for the lower rate. This effect can be seen in Figure 5. To further study this effect, the 2.5 ppm 1-CN solution was passed through the 5% PIB coated hollow waveguide sampler for 10 min using different sampling flow rates. A plot of the 1-CN peak intensity at 766 cm-1 against the sampling flow rate is shown in Figure 6. As can be seen in this plot, the adsorption efficiency decreases gradually. Although the adsorption efficiency drops as the sampling flow rate is increased, the analytical signal is increased as the sampling flow rate is increased. Considering the stability of PIB coating, the chance of stripping the PIB coating from the silver layer is increased because of the absence of strong interactions between the PIB and the silver layer of the waveguide. However, no stripping of the PIB coating was observed in this work, so a sampling flow rate of 15 mL/min was selected for following experiments, for simplicity. 3744 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

Figure 6. Effect of sampling flow rate on adsorption rate of 1-CN in aqueous solutions at constant adsorption time. 2.5 ppm 1-CN solution and a 5% PIB coated hollow waveguide were used.

Figure 7. Theoretical calculation of responses in eq 8. (A) Plot of film volume against sampling time keeping all other parameters in eq 8 constant. (B) Plot of sampling times against film volumes keeping all other parameters in eq 8 constant.

Effect of SPME Film Thickness on Adsorption. Film thickness affects the analytical signals as derived in eq 4 or 8. In equilibrium situations, the analytical signals should be proportional to the film thickness as shown in eq 4. In nonequilibrium situations, an exponential curve shape should be obtained. To demonstrate these phenomena, the theoretical calculation based on eq 8 is shown in Figure 7A for different SPME phase volumes. As can be seen in this plot, the larger the SPME phase volume, the higher the analytical signal that can be obtained. In practical terms, however, the SPME phase volume cannot be infinitely extended for two reasons. First, noise level was increased as the

Figure 8. (A) Relationship between signal of 1-CN and adsorption time for different film thicknesses. The integration of the PIB band located at 900 cm-1 (from 989 to 894 cm-1) was used to indicate the PIB film thickness. 2, 5% PIB (integrated band intensity is 53); [, 1.5% PIB (18); 1, 0.5% PIB (5.8). (B) Effect of PIB thickness on adsorption rate of 1-CN in aqueous solutions with 10-min sampling times. Integration of the PIB band located at 900 cm-1 (from 989 to 894 cm-1) was used to indicate the PIB film thickness.

SPME phase volume was increased. This is due to light scattering on the SPME phase and also due to the absorption of infrared radiation by the PIB phase. On the other hand, the increase of the noise level in the larger SPME phase volume offsets some advantage in increasing the analytical signals. Examining the noise level (averaged in the region of 1200-650 cm-1) for noncoated, 3% coating, and 5% coating, the obtained average noise levels are around 0.21, 0.36, and 0.81 mAU, respectively. Second, the reproducibility of thick SPME coating was relatively low because of the high viscosity for high percentage PIB solutions. Therefore, a 5% PIB coating was selected as being practical for following experiments. This reveals that the use of very thick coatings of the SPME phase is limited. To further examine the relationship of analytical signals and SPME phase volumes at fixed sampling times, the analytical signals were calculated on the basis of the parameters used in Figure 7A for 10-, 20-, and 30-min sampling times. The results are plotted in Figure 7B. As can be seen in this plot, the signal is not proportional directly to the SPME phase volume. It can also be seen that the longer the adsorption time, the more linear will be the relationship obtained (close to the situation in eq 4). To study the feasibility of derived equations, a 2.5 ppm 1-CN solution was used to detect the signals using different PIB coated hollow waveguide samplers. Results are shown in Figure 8A for the relationship of sampling time and detected 1-CN signals for three

different film thicknesses of sampler. The curve fits based on eq 8 are also plotted in this figure. As can be seen in this figure, the equilibrium situation can be reached around 30 min but a fast signal increase is obtained in the first 10 min. Assuming that equilibrium is reached after 10 min of sampling, the analytical signal of 1-CN should be proportional to the film thickness as derived in eq 4. By plotting the 1-CN signal with the film thickness for a 10-min sampling time, the curve can be fit well using eq 8, as can be seen in Figure 8B. This curve does not strongly deviate from a linear curve, revealing that for sampling times larger than 10 min, the effect of film thickness can be partially removed. For example, plotting the analytical signal/film thickness ratio versus the original concentration of the analytes yields a reasonably linear relationship. If a longer adsorption time is used, eq 8 can be simplified to eq 4, so that a linear plot of analytical signal against SPME phase volume can be obtained. This can further simplify the process of SPME coating if reproducibility of coating is not obtainable. Therefore, in the following section, an examination of the linearity of construction of a standard curve, a nonequilibrium situation was used, i.e., a 10-min sampling time. Relationship between Volatility, Concentration, and IR Signal. To study the limitations of this method in analysis of organic species in aqueous solutions, four different volatility compounds were used including 1-CN, chlorobenzene, 2-chlorotoluene, and trichloroethylene. The literature25,26 values of the vapor pressures at 25 °C for 1-CN, 2-CT, CB, and TCE are 0.017, 3.55, 12.05, and 74.27 Torr, respectively. The 1-CN is used as a representative of less volatile compounds such as polyaromatic hydrocarbons, polycyclic biphenyls, or some chlorinated pesticides (DDT type compounds). According to literature values, these compounds have lower or similar vapor pressures to 1-CN. 2-CT and CB were used to investigate which compounds could be used with this method. TCE was used to demonstrate the limitation of this method in the detection of highly volatile compounds. Using the method developed above, runs for six concentrations (2500, 1250, 500, 250, 125, and 50 ppb) of the probe molecules were performed. Each concentration was run in triplicate. In these experiments, a 5% PIB coated hollow waveguide sampler was used with 10 min of sampling time. The spectra were collected by coadding 100 scans with 4 cm-1 resolution. The linear regression result of 1-CN showed an R-squared coefficient of 0.9993, and its detection limit, three times the signal-to-noise ratio of the lowest concentration signals, was 10 ppb. This reveals that, for low volatility compounds, the developed method is highly suitable. For compounds with medium vapor pressure, the obtained R-squared coefficients were 0.9978 and 0.9993 for 2-CT and CB, respectively. On the basis of the lowest detectable concentration signals, the detection limits for these two compounds were 77 and 395 ppb. These results indicate that compounds with vapor pressures lower than 12 Torr at 25 °C are also detectable, but the detection limit is slightly higher. The relatively high detection limit for chlorobenzene is believed to be due to its relatively higher volatility than those of other compounds, which can cause adsorbed molecules be partially removed during the process of removing residual water (refer to Figure 4). In attempting to (25) Schwarzenbach, R. P.; Gschwend, P. M.; Imboden, D. M. Environmental Organic Chemistry; John Wiley & Sons: New York, 1992. (26) Boublik, T.; Fried V.; Hala, E. The Vapor Pressure of Pure Substances; Elsevier Science Publishers: Amsterdam, The Netherlands, 1984.

Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

3745

examine highly volatile compounds by this method, IR signals can also be obtained, but detectable concentration is increased to the ppm region. Also, gas-phase TCE was observed. This reveals that the TCE in the hollow waveguide sampler was partially vaporized. This limits the application of the sampling method to an examination of compounds with vapor pressures larger than 70 Torr. Also, detection of compounds with vapor pressures between 12 and 70 Torr is questionable by this method. CONCLUSION In this paper, a new method based on a hollow waveguide sampler is developed. In this method, the sampler was constructed by coating an SPME phase onto the inner surface of an IR hollow waveguide. By properly controlling the factors such as film volume, sampling time, and sampling rate, the PIB-coated hollow waveguide sampler can provide high sensitivity in detection of organic compounds in aqueous solutions, especially for chlorinated organic compounds. The film thickness largely affects the analytical signals as expressed in eqs 4 and 8 for both equilibrium and

3746 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999

nonequilibrium conditions. The influences of sampling flow rate are less important if sampling time is kept constant. The results regarding detection of different concentrations of probe molecules in aqueous solution show that this detection method is highly sensitive and possesses high linearity for less volatile compounds (lower than 12 Torr). For highly volatile compounds, this method is restricted as shown by the analytical results for TCE. ACKNOWLEDGMENT The authors would like to thank the National Science Council of the Republic of China for financially supporting this work under the Contract NSC86-2113-M-033-008. The authors would also like to thank Dr. Andrew J. Lange for his help in preparation of this manuscript.

Received for review March 26, 1999. Accepted June 7, 1999. AC9903252