Anal. Chem. 1997, 69, 4917-4922
Quantitative Determination of Semivolatile Organic Compounds in Solution Using Trap-and-Release Membrane Inlet Mass Spectrometry Frants R. Lauritsen* and Raimo A. Ketola†
Physical Biochemistry Group, Institute of Biochemistry, Odense University, DK-5230 Odense M, Denmark
This paper discusses the use of trap-and-release membrane inlet mass spectrometry (T&R-MIMS) for the quantitative determination of semivolatile organic compounds in real samples. We found that the T&R-MIMS technique is particular sensitive to relatively polar, semivolatile organic compounds. For example, the detection limits for the acids acetylsalicylic acid and phenoxyacetic acid were lowered by a factor of 100 as compared with those possible with standard MIMS, and caffeine was detectable only with the T&R-MIMS method. The detection limits were in the parts-per-billion range, and the dynamic range was 3 orders of magnitude. As a practical example of the application of the T&R-MIMS technique, we used it for the quantitative analysis of caffeine in ground coffee and tea leaves. Good agreement between T&R-MIMS and HPLC determinations was found, and the reproducibility of the whole analytical system for caffeine determination (extraction procedure and T&R-MIMS determination) was within 10% as relative standard deviation. However, for coffee, a large background from the essential oils prevented low-level work, such as the determination of residual caffeine in decaffeinated coffee. Obviously, the analysis of many complex matrixes will require the use of tandem mass spectrometry. Membrane inlet mass spectrometry (MIMS) has become an established technique for the determination of volatile organic compounds in environmental samples1,2 and in microbial broths.3-5 In addition, its application to air monitoring is a rapidly growing area.6-8 The reason for this is the simplicity, high sensitivity, fast * Author to whom correspondence should be addressed. E-mail:
[email protected]. Fax: -45-65932309. † On leave from VTT Chemical Technology, P.O. Box 1401, FIN-02044 VTT, Finland. (1) Cooks, R. G.; Kotiaho, T. In Pollution Prevention in Industrial Processes: The Role of Process Analytical Chemistry; Breen, J. J., Dellarco, M. J., Eds.; ACS Symposium Series 508; American Chemical Society: Washington, DC, 1992; pp 126-154. (2) Kotiaho, T. J. Mass Spectrom. 1996, 31, 1. (3) Kotiaho, T.; Lauritsen, F. R.; Choudhury, T. K.; Cooks, R. G.; Tsao, G. T. Anal. Chem. 1991, 63, 895A. (4) Degn, H. J. Microbiol. Methods 1992, 15, 185. (5) Lauritsen, F. R.; Lloyd, D. In Mass Spectrometry for the Characterization of Microorganisms; Fenselau, C., Ed.; ACS Symposium Series 541; American Chemical Society: Washington, DC, 1994; pp 91-106. (6) Cisper, M. E.; Garrett, A. W.; Cameron, D.; Hemberger, P. H. Anal. Chem. 1996, 68, 2097. (7) Gordon, S. M.; Callahan, P. J.; Kenny, D. V. Rapid Commun. Mass Spectrom. 1996, 10, 1038. (8) Ketola, R. A.; Ojala, M.; Sorsa, H.; Kotiaho, T.; Kostiainen, R. Anal. Chim. Acta 1997, 349, 359. S0003-2700(97)00570-2 CCC: $14.00
© 1997 American Chemical Society
analysis time, and on-line monitoring capabilities of the technique. In a recent comparison9 between three analytical methods (MIMS, purge-and-trap GC/MS, and static headspace gas chromatography) for the analysis of volatile organic compounds in water samples, MIMS was found to be superior or equal to the other techniques with respect to analysis time, detection limits, and reproducibility. The major drawback is the lack of chromatographic separation of components. During the last 5 years, many new membrane inlet designs have been reported. Degn et al.10 used hydrophilic membranes to detect water in organic solvents, and Lauritsen et al.11 used porous membranes and solvent chemical ionization for the determination of organic compounds in complex organic matrixes. The demand for still lower detection limits led to the development of the combined membrane inlet/jet separator system by Dejarme et al.12 and several types of inlets using an enrichment device13-15 between the membrane inlet and the mass spectrometer. In recent work from Cooks’s group,16 a chemically modified cellulose membrane was used for the selective introduction of aldehydes into the mass spectrometer, even from very complex mixtures. A particular interesting paper was published by Yakovlev et al.,17 who created the first MIMS system capable of detecting dissolved ions. They used a strong electric field to drag ions from the solution through a porous membrane into the vacuum. New designs of the traditional membrane inlet with silicone membranes have resulted in the development of small inlets giving spacial resolution by Thomas and Lloyd18 and Baumgardner et al.19 The new membrane inlet systems have recently been reviewed.20 Whereas the analysis of volatile organic compounds in aqueous samples has become routine for the MIMS system, the analysis of semivolatiles (boiling point above 250 °C) has not. This is because the membrane inlets cannot be operated at temperatures much higher than 70 °C before bubble formation in front of the (9) Ketola, R. A.; Virkki, V. T.; Ojala, M.; Komppa, V.; Kotiaho, T. Talanta 1997, 44, 373. (10) Degn, H.; Bohatka, S.; Lloyd, D. Biotechnol. Tech. 1992, 6, 161. (11) Lauritsen, F. R.; Kotiaho, T.; Choudhury, T. K.; Cooks, R. G. Anal. Chem. 1992, 64, 1205. (12) Dejarme, L. E.; Bauer, S. J.; Cooks, R. G.; Lauritsen, F. R.; Kotiaho, T.; Graf, T. Rapid Commun. Mass Spectrom. 1993, 7, 935. (13) Shoemaker, J. A.; Bellar, J. W.; Eichelberger, J. W.; Budde, W. L. J. Chromatogr. Sci. 1993, 31, 279. (14) Rivlin, A. A. Rapid Commun. Mass Spectrom. 1995, 9, 397. (15) Mendes, M. A.; Pimpin, R. S.; Kotiaho, T.; Eberlin, M. Anal. Chem. 1996, 68, 3502. (16) Xu, C.; Patrick, J. S.; Cooks, R. G. Anal. Chem. 1995, 67, 724. (17) Yakovlev, B. S.; Talrose, V. L.; Fenselau, C. Anal. Chem. 1994, 66, 1704. (18) Thomas, K. L.; Lloyd, D. FEMS Microbiol. Ecol. 1995, 16, 103. (19) Baumgardner, J. E.; Quinn, J. A.; Neufield, G. R. J. Mass Spectrom. 1995, 30, 563. (20) Lauritsen, F. R.; Kotiaho, T. Rev. Anal. Chem. 1996, 15, 237.
Analytical Chemistry, Vol. 69, No. 23, December 1, 1997 4917
membrane causes highly instable signals. At temperatures above 100 °C, the signal falls almost to baseline level because of the large volumetric expansion as the water starts to boil.21 The low inlet temperature limits the vaporization of the semivolatiles from the membrane surface and results in long membrane response times (>5 min) for such compounds. Until recently, compounds with a boiling point between 200 and 300 °C were best detected by the so-called direct insertion membrane probes, where the membrane is mounted inside22 or in the immediate vicinity21,23 of the ionizing region. With these inlets, problems with chromatographic effects on vacuum surfaces from the “cold” membrane surface to the ionizing region are almost eliminated. Recently, we introduced a completely new way of conducting the MIMS experiment, the so-called trap-and-release MIMS24 (T&R-MIMS). In this method, semivolatile organic compounds are preconcentrated inside the membrane, before they are thermally released into the ion source by heat radiation from the filament. The system uses a standard membrane inlet with a silicone tube passing straight through the ion source. A long slit in the ion source parallel to both the tubular membrane and the filament allows heat radiation from the filament continuously to bombard the membrane surface. During a sampling period, the membrane is kept cold by the sample liquid flowing through the inside of the silicone tube. However, during a short interruption of the liquid flow, the membrane is rapidly heated to more than 300 °C, and organic compounds dissolved in the membrane are released into the ion source. In this fashion, a desorption peak is obtained. A similar system has been presented by the group of Matz.25,26 The major difference between their system and ours is, that our inlet is an integral part of the ion source, whereas theirs forms a separate unit mounted at a distance from the ion source. The physical principles of operation of the two systems are the same. It is the purpose of this paper to demonstrate the performance of the second-generation T&R-MIMS system in a practical application. The second-generation system27 deviates from the first by the way it induces the fast heating of the membrane. Instead of simply interrupting the liquid flow in the membrane inlet, an air plug is passed through it. In this fashion, energy is not spent in a heating of the sample liquid inside the membrane, but only in heating of the membrane material itself. The result is a narrower desorption peak. As a practical demonstration of the technique, we have chosen the quantitative determination of caffeine in coffee and tea. With its low vapor pressure and high water solubility, caffeine determination represent the limits of the technique. In addition, the caffeine content in coffees and teas is well known, and standard analytical methods exist.28 EXPERIMENTAL SECTION Description of the Trap-and-Release MIMS System. The technical details of the trap-and-release MIMS system have been (21) Bier, M.; Kotiaho, T.; Cooks, R. G. Anal. Chim. Acta 1990, 231, 175. (22) Bier, M.; Cooks, R. G. Anal. Chem. 1987, 59, 597. (23) Lauritsen, F. R. Int. J. Mass Spectrom. Ion Processes 1990, 95, 259. (24) Leth, M.; Lauritsen, F. R. Rapid Commun. Mass Spectrom. 1995, 9, 591. (25) Matz, G.; Kesners, P. Anal. Mag. 1995, 23, M12. (26) Matz, G.; Lennemann, F. J. Chromatogr. A 1996, 750, 141. (27) Lauritsen, F. R.; Leth, M. In Proceedings of the 44th ASMS Conference on Mass Spectrometry and Allied Topics, Portland, OR, May 12-16, 1996; pp 867-868. (28) Maier, H. G.; Tillack, B.; Ho ¨ller, U. Dtsch. Lebensm.-Rundsch. 1996, 92, 283.
4918 Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
Figure 1. Schematic diagram of the T&R-MIMS system.
described earlier.24 The system consists of a 20 mm long silicone tubing (10 mm effective length) mounted between two 1/16 in. stainless steel tubes, such that liquid flowing into the system through one steel tube passes through the silicone membrane and leaves the system through the other steel tube (Figure 1). The silicone tube passes straight through the ion source of the mass spectrometer. A long slit in the ion source parallel to the membrane lets radiation (heat and electrons) from the filament continuously bombard the silicone membrane. The silicone membrane is kept cold by a continuous flow of either the sample solution or water. With the help of a three-way valve (Figure 1), an air plug of well-defined length (duration) can be introduced into the liquid flow. When the air plug passes through the silicone membrane, it is no longer cooled, and the temperature rises rapidly to more than 300 °C.27 As a result, semivolatile organic compounds dissolved in the membrane are released into the ion source, and a desorption profile is observed. More than 1000 analysis cycles can be done with a single membrane before significant degradation in performance is observed. To increase the solubility of organic compounds in the silicone membrane, the membrane inlet was always operated at room temperature or lower. The mass spectrometer was a QMG 400 single-quadrupole mass spectrometer (Balzers, Liechtenstein) with a mass range of 1-500. The ion source was a cross-beam electron impact ion source, and the ionization was performed using 50 eV electrons. The experiments were made with an electron emission current of 1.0 mA. The membrane was a poly(dimethylsilicone) membrane (Technical Products Inc., Decatur, GA) with a wall thickness of 216 µm (i.d. 0.020 in. and o.d. 0.037 in.). It was soaked (expanded) in heptane and then fitted to the steel tubes. Following evaporation of the heptane, a strong seal between the membrane and the steel tubes was obtained. Quantitative Analysis of Caffeine. Trap-and-Release MIMS. Here, 2.00 g of roasted coffee or a bag of tea (weight around 2.0 g) was displaced in a 250 mL Erlenmeyer flask, and 100 mL of boiling water was added to the flask. After 10 min, the solution was filtered through a cheesecloth, and 1.0 mL of an internal standard, 1-naphthalenemethanol (400 mg/L) was added to the solution in order to correct for instrumental drift. The sample solution was cooled to 0 °C in an ice-water bath before analysis. Caffeine was analyzed using three different standard concentrations (100, 300, and 500 mg/L) and an internal standard method. Quantitation was achieved by comparing the signals of caffeine and 1-naphthalenemethanol in the samples with the signals obtained from the standard solutions prepared in a similar way.
Figure 2. T&R-MIMS desorption profiles of toluene (m/z 91 monitored), 1-naphthalenemethanol (m/z 158 monitored), and caffeine (m/z 194 monitored) obtained during the passage of a 50 s air plug.
We found that the maximal desorption signal from caffeine was obtained after 20 min of sampling. This corresponds to the time it takes to reach a steady state,24 where the number of caffeine molecules entering the membrane equals the number of molecules leaving it again either to the sample solution or into the vacuum. The sample flow rate was 1.0 mL/min. An increase in sample flow rates did not result in higher desorption intensities, and a reduction in flow rates resulted in an insufficient cooling of the membrane. To reduce memory effects, each sample was followed by a 130 s long distilled water plug and a 50 s air plug. This cleaning process was repeated twice. High-Performance Liquid Chromatography (HPLC). The analysis was carried out using a Model 510 liquid chromatograph (Waters Millipore, Milford, MA) equipped with a 20 µL sample loop and a Model 441 absorbance detector (at a wavelength of 214 nm). The column used was a Hypersil 5 C18 column (4.0 mm × 300 mm, Phenomenex, Torrance, CA), and the mobile phase was acetonitrile/water (8:92 v/v) at a flow rate of 1.0 mL/min. Other Conditions. Standard MIMS. The membrane inlet mass spectrometry system used in the comparison of detection limits between T&R-MIMS and standard MIMS was a sample cell5 with a flat sheet silicone membrane operated at 70 °C. The system uses the same type of mass spectrometer as the T&R-MIMS system, except that it has a home-built ion source, which makes it possible to fully integrate the ion source with the sample cell. The system was the most sensitive of six operating MIMS systems in our group with respect to a determination of semivolatile organic compounds. Reagents. Toluene (Mallinkrodt, Inc., Paris, KY), caffeine (The British Drug Houses, England), 1-naphthalenemethanol (Aldrich Chemical Co., Milwaukee, WI), and acetonitrile (Rathburn, Scotland, HPLC-grade) were used as received. Stock solution of toluene (1000 mg/L) was made in methanol (Prolabo, Denmark), and stock solutions of other reagents were made in distilled water. Further dilutions were made in distilled water. RESULTS AND DISCUSSION Desorption Profile. Figure 2 shows the desorption profiles for three compounds with widely different melting and boiling points. In this experiment, a standard solution containing 170 µg/L toluene (melting point (mp) -95 °C, boiling point (bp) 110 °C), 3.6 mg/L 1-naphthalenemethanol (mp 64 °C, bp 301 °C), and 270 mg/L caffeine (mp 238 °C, bp not reported) was passed through the inlet for 20 min before the trapped molecules were
released during the passage of a 50 s air plug. As expected the small and volatile compound toluene reached a steady state flow through the membrane before the air plug was passed through the system, as evidenced by its elevated level prior to 15 s. 1-Naphthalene methanol did not diffuse so fast through the cold membrane (room temperature), and it did not reach a steady state level within the 20 min before the passage of the air plug. Caffeine is expected to have a membrane response time comparable to that of 1-naphthalenemethanol, but because of its extreme low vapor pressure it was not observed at all during the 20 min sampling period. At the time when the air plug reaches the membrane, the signals from both toluene and 1-naphthalenemethanol increase immediately because of the elevated temperature in the membrane, whereas the signal from caffeine has a delay of about 6 s before it rises. The delay in the caffeine signal is probably the result of an interruption in the membrane heating at 100 °C until residual liquid inside the membrane is vaporized.27 Caffeine, with its extremely low vapor pressure, probably does not evaporate from the silicone membrane at temperatures below 100 °C. The measured widths of the desorption profiles at halfheight was 8.5, 12.7, and 12.3 s for toluene, 1-naphthalenemethanol, and caffeine, respectively. The difference is a result of a slower diffusion of 1-naphthalenemethanol and caffeine than toluene in the silicone membrane. A simple parameter for the signal improvement (the gain) of the T&R-MIMS method as compared to standard MIMS can be defined as the ratio between the maximal signal obtained during the desorption profile and the signal obtained just before the air plug enters the membrane inlet. For the three compounds measured in Figure 2, the gains were calculated as 5, 100, and . 100 for toluene, 1-naphthalenemethanol, and caffeine, respectively. The low gain of toluene reflects the fast diffusion of this compound through the membrane. It takes 5 min to fill the membrane with toluene (steady state flow condition) and about 30 s to empty it. The maximal theoretical gain is, therefore, about 10. In a previous paper,23 we showed that there was a direct correlation between the membrane response time and the gain: the longer the response time, the larger the gain. The gain parameter cannot be directly related to an improvement in detection limits. The low operating temperature of the membrane (room temperature or lower) during the trapping period of the T&R-MIMS experiment gives much smaller signals than would be observed in a standard MIMS experiment with a membrane operated at 70 °C. Further, the elevated temperature of the membrane (up to 300 °C) during the desorption profile27 causes a general increase in the background, making low-level work difficult. A result of this is that volatile organic compounds like toluene are best measured with a standard MIMS system. However, with semivolatile organic compounds, large improvements in detection limits can be obtained. This is displayed in Table 1, which shows a comparison of measured detection limits for a variety of semivolatiles obtained with T&R-MIMS and standard MIMS. The improvement factor in detection limits obtained with the T&R-MIMS system is from 5 (fluoranthene) to . 100 (caffeine). It is interesting that the improvement factors are largest (50 or higher) for the compounds that are the most difficult to measure with standard MIMS. These are typically relatively polar compounds, compounds that do not dissolve very well in the membrane. With the T&R-MIMS system, detection Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
4919
Table 1. Comparison of Detection Limits for Standard MIMS and T&R-MIMS
compound DDT phenoxyacetic acida 4-phenylphenol phenanthrene fluoranthene acetylsalicylic acida caffeine
detection limitc (µg/L) ion bp water (°C) solubilityb monitored (m/z) standard T&R 260 285 305 340 385 135d 238d
1 3 2 1 1 3 2
235 107 170 178 202 120 194
1000 10000 100 4 25 20000 nde
25 100 2 0.5 5 250 600
a pH adjusted to 2. b I ) insoluble; 2 ) slightly soluble; 3 ) soluble; 4 ) very soluble. c Standard MIMS: S/N ) 3. T&R-MIMS: the concentration that causes a 50% increase in signal as compared to a blank. d Melting point. e Not detectable.
limits for relatively polar semivolatiles are lowered from typically parts-per-million to parts-per-billion levels. In the experimental setup described here, the membrane is continuously bombarded with light and electrons from the filament, also during the sampling period. This means that there must be a temperature gradient across the membrane, with the vacuum side warmer than the sample side. An obvious possibility to improve the detection limits would, therefore, be to turn off the filament during the sampling period or at least to deflect the electrons away from the ion source. In this fashion, the membrane temperature at the vacuum side is lowered, and a larger sorption of sample molecules should result. We did test the system with an electron trap (active during the sampling period) mounted near the filament. However, this gave no differences in performance of the system, and it was concluded that most of the radiation energy hitting the membrane comes from the light and not from the electron bombardment. Only few attempts were done to test the possibility of turning the filament off during the sampling period, mainly because it is our experience with the Balzers quadrupole systems that the lifetime of the filament is drastically reduced when the filament is frequently turned on and off. Quantitative Aspects. A requirement for quantitative determinations is a system with a low memory effect. The T&R-MIMS system is particularly sensitive to memory effects, since the part of the membrane that binds it to the steel capillaries is not sufficiently heated during the desorption step. Sample molecules dissolved in this “cold” part of the membrane will, therefore, diffuse back to the center of the membrane and be released in the next desorption. The result is a carryover from one sample to the next, and cleaning between samples becomes necessary. In our T&R-MIMS system, two parameters can be used in a cleaning process: (a) the duration of a flushing period with pure water and (b) the use of a cleaning air plug. Figure 3 shows a schematic drawing of the experimental sequence used in our search for the optimal cleaning process. The duration of the cleaning air plug turned out to be much more important than that of the flushing period. In Table 2, we show the results of varying the duration of the cleaning air plug. In this experiment, a standard solution of 500 mg/L caffeine was sampled for 20 min before the caffeine was released from the membrane with a 50 s air plug. Following the sample analysis, the system was flushed with pure water for 130 s before a cleaning air plug of varying duration was passed through the membrane 4920 Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
Figure 3. Experimental procedure used in the evaluation of the memory effect. Table 2. Percent Residual Signal after the Use of a Cleaning Air Plug of Varying Duration duration of the cleaning step(s)
memory effect (%)
0 10 20 30 40 50
28.1 26.0 17.4 8.7 6.1 1.7
inlet system. After the cleaning step, the membrane was again flushed with pure water for 130 s before an air plug of 50 s duration was passed through the system. The memory effect was then defined as the ratio between the area of the desorption profile from the last 50 s to the area of the desorption profile from the sample itself. The table shows a dramatic effect of the duration of the cleaning air plug on the memory effect. Without the cleaning air plug, the memory effect was almost 30%, and then it dropped to about 2% with a 50 s air plug. It is interesting to note that the duration of the air plug has only a little effect up to 20 s, whereafter a fast drop is observed. This behavior probably reflects the temperature profile of the membrane during the passage of the air plug. At first the temperature rises rapidly to 100 °C, where it stays for a short while until all residual liquid inside the membrane has evaporated.27 In this period, which takes about 10 s, the temperature is too low to vaporize the residual caffeine. For safety reasons, we did not use air plugs longer than 50 s. Prolonged heating of the membrane causes fragmentation of the membrane, and ions from these fragments start to appear in the mass spectrum. Instead, we chose to use successive cleaning procedures with 50 s air plugs. We found that, with two successive cleaning procedures, the memory effect for caffeine could be kept below 2%. This cleaning process is quite effective, considering the high concentration of the original standard solution (500 mg/ L). With other compounds of higher volatility than caffeine, the cleaning step was even more effective, e.g., the residual amount of 1-naphthalenemethanol in the second cleaning step was only 1.6% after sampling of 4 mg/L of 1-naphthalenemethanol for 20 min (the signal height was equal to that of 300 mg/L caffeine).
Figure 4. Desorption profiles from 10 successive determinations of a standard solution of 100 mg/L caffeine (m/z 194 monitored). The profiles are normalized to the average height of all 10 peaks (average height, 100). The x-scale (cycles) is noncontinuous.
The repeatability and stability are very important parameters of analytical methods. To test the T&R-MIMS system, we analyzed a standard solution of 100 mg/L caffeine 10 times according to a procedure, where the standard solution was passed through the inlet system for 20 min before an air plug of 50 s was used to release the caffeine. Between each analysis, the system was cleaned two times with 130 s of pure water, followed by an air plug of 50 s. The ion chromatogram obtained during the experiment is presented in Figure 4. The whole measurement took more than 4 h, but only the chromatogram of caffeine from the release part of the samplings is presented in Figure 4. The relative standard deviation was calculated as 4% and 3% using peak heights and areas, respectively. Linearity is another very important factor of an analytical system. In order to test this for the trap-and-release system, we measured the intensity of the desorption profiles over a broad range of concentrations. We found that the T&R-MIMS system was linear over 3 orders of magnitude. For example, for caffeine the system was linear in the concentration range of 1-1000 mg/L (standard solutions containing 1, 3, 5, 10, 30, 50, 100, 300, 500, and 1000 mg/L caffeine analyzed) with a correlation coefficient as high as 0.998. Quantitative Analysis of Caffeine in Tea and Coffee. To test the capabilities of the T&R-MIMS technique with real samples, we analyzed the caffeine content in a typical cup of tea or coffee. Figure 5a shows a mass spectrum of a cup of tea obtained during the desorption of the sample. The spectrum represents the average of six scans of the whole spectrum each, recorded with a scan rate of 50 ms/amu. The molecular ion of caffeine (m/z 194) and the most abundant fragment ion (m/z 109) are clear, as are the peaks of m/z 147 (fragment from the membrane) and 149 (phthalate plastizicers). The particular tea used (Earl Gray, Twinings of London) contains Bergamot flavor, a common tea additive, and the main constituents of that substance (linalool, linalyl acetate, and limonene) produce the ions at m/z 105, 107, 119, 121, and 136. Overall, the spectrum is quite simple, and the caffeine content can be selectively determined through singleion monitoring of the molecular ion. Figure 5b shows a mass spectrum of a typical cup of coffee recorded in a similar fashion as the tea spectrum in Figure 5a. The spectrum is very complex as compared with the tea spectrum. Oils present in coffee give a huge background of ion clusters separated by 14 mass units. Still, the dominant ions (m/z 194 and 109) from caffeine are clear and can form the basis for a
Figure 5. T&R-MIMS mass spectra of a typical cup of (a) tea and (b) coffee. The spectra represent the average of six successive scans obtained during the release period.
quantitation. However, for coffee the oil background prevents us from detecting concentrations below 10 mg/L. A cup of decaffeinated coffee would typically contain residual caffeine at concentrations up to 10 mg/L. A selective determination of caffeine in such samples with the T&R-MIMS technique will require the use of tandem mass spectrometry. In general, we expect that the T&R-MIMS technique would improve considerably with respect to both selectivity and sensitivity if it were combined with tandem mass spectrometry. Table 3 shows the results of quantitative determinations of the caffeine content in various roasted coffee (all ecological) and tea brands obtained from a local supermarket. The concentrations are given both as the measured concentrations in the liquid (mg/ L) and as the calculated amount of extractable caffeine in the pure coffee or tea (mg/g). All values represent the average of two duplicate measurements of the same sample, and the relative standard deviations were below 10%. The results agree very well with reported values for the caffeine content in ground coffee and in tea bags.28,29 Good agreement between T&R-MIMS and HPLC (data included in Table 3) determinations was found. Generally, the deviation between the T&R-MIMS and the HPLC values was slightly higher for the tea samples than for the coffee samples. This reflects problems with precipitations in some of the tea samples when they were cooled prior to the T&R-MIMS analysis. The reproducibility of the whole analytical T&R-MIMS method for caffeine analysis (extraction of the coffee and tea samples and determination by T&R-MIMS) was tested through the analysis of six separate samples of the same coffee brand (C). We found that the method was sensitive to variations in the temperature of the cooled samples. However, when the temperature of the sample was kept within a few degrees, the relative standard deviation was calculated as 4.5%. (29) Blauch, J. L.; Tarka, S. M., Jr. J. Food, Sci. 1983, 48, 745.
Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
4921
Table 3. Quantitative Analytical Results of Some Coffee and Tea Brands
a
sample
concn of caffeine in extract (mg/L)
A B C D E
277 276 278 288 277
F G H I J K L
584 558 691 603 520 606 592
caffeine content of the original brand (mg/g)
HPLC result (mg/L)
differencea (%)
Coffee 13.9 13.8 13.9 14.4 13.9
286 285 290 282 287
3.1 3.2 4.1 -2.1 3.5
Tea 28.8 27.6 33.4 29.9 26.0 30.1 28.9
620 598 702 630 503 577 565
5.8 6.7 1.6 4.3 -3.4 -5.0 -4.8
Difference is (HPLC result - T&R-MIMS result)/HPLC result × 100 (as a %).
Overall the performance characteristics of the two methods (T&R-MIMS and HPLC) were very similar. The detection limit for caffeine with the HPLC method was estimated to be a little lower than that of T&R-MIMS (0.1 mg/L as compared to 0.6 mg/ L), but the T&R-MIMS method is still at a very infantile stage, and we expect detection limits to be improved by at least an order of magnitude. With both methods, the analysis time was about 20 min with a sample throughput of approximately 3/h. The only real drawback of the T&R-MIMS method is the need for larger amounts of sample, 10-20 mL, as compared with HPLC, where only 0.5 mL or less is required to flush the sample loop. In conclusion, we have demonstrated that the T&R-MIMS technique can be used for the quantitative determination of semivolatile organic compounds in solution. Detection limits for the semivolatiles were in the parts-per-billion range, and the linearity of the technique was 3 orders of magnitude. The reproducibility of the method is very high. For example, the levels
4922 Analytical Chemistry, Vol. 69, No. 23, December 1, 1997
of caffeine which were extracted and analyzed from ground coffee and tea leaves agreed within 7% of the levels determined with HPLC. We expect that the T&R-MIMS method, with some optimization, could find applications such as the determination of pharmaceuticals in urine samples or pesticides in environmental samples. However, such applications will probably require the use of tandem mass spectrometry in order to improve selectivity and sensitivity. ACKNOWLEDGMENT The Danish Research Academy is appreciated for its financial support. Received for review June 2, 1997. Accepted September 22, 1997.X AC970570Q X
Abstract published in Advance ACS Abstracts, November 1, 1997.
