Anal. Chem. 1999, 71, 23-27
Comparison of Gas-Sampled and SPME-Sampled Static Headspace for the Determination of Volatile Flavor Components Michael E. Miller and James D. Stuart*
Department of Chemistry, U-60, University of Connecticut, 55 North Eagleville Road, Storrs, Connecticut 06269-3060
Traditional static headspace and headspace solid-phase microextraction (SPME) techniques were compared for their effectiveness in the extraction of volatile flavor compounds from the headspace of various juice samples. Each method was used to evaluate the responses of certain analytes from real samples and calibration standards in order to provide sensitivity comparisons between the two techniques. Experimental results showed traditional static headspace lacked the sensitivity needed to evaluate certain flavor volatiles, such as r-terpinene and linalool, and that further concentration of the headspace was necessary. Dramatic improvements in the extraction abilities of the SPME fibers over the traditional static headspace method were noted. Different SPME fibers were investigated to determine the selectivities of the various fibers to the different flavor compounds present in the juice samples. Of the various fibers investigated, the PDMS/DVB fiber proved to be the most useful for these analyses. Aging studies of juice samples were also performed which verified that degradation could be observed and quantified. The determination of volatile components in a mixture is a process widely used in many disciplines, such as environmental, food, forensic, fragrance, oil, pharmaceutical, and polymer analysis. The method of choice for many of these analyses is static headspace followed by GC or GC/MS analysis. This relatively simple sampling method is dependent upon the formation of equilibrium conditions in a closed system. Henry’s law can describe the equilibrium between a liquid or solid sample and the gaseous phase above. After equilibrium is established, an aliquot of the gaseous headspace is sampled by means of a gastight syringe or sample loop and then transferred to a gas chromatograph for analysis. A difficulty with this method is that samples often have key analytes present in complex matrixes at very low concentration, making it difficult to place enough analyte mass on column without first concentrating the sample. The full evaporation technique (FET) can address the complex matrix problem, by using equilibration at temperatures above the boiling points of the analytes of interest, to force the analytes out of the matrix and into the headspace.1,2 These conditions can pose problems for many flavor compounds, since the boiling points are (1) Markelov, M.; Guzowski, J. P.; Jr. Anal. Chim. Acta 1993, 276, 235-245. (2) Stuart, J. D.; Miller, M. E.; Williams-Burnett, M. L. J. Soil Contam. 1997, 6, 439-463. 10.1021/ac980576v CCC: $18.00 Published on Web 11/14/1998
© 1998 American Chemical Society
much too high (>170 °C) to safely use this technique on aqueous samples. Cryogenically trapping and concentrating the analytes in the GC inlet is another possibility, but this has difficulties when large amounts of water are present, since ice buildup can interfere with this process. Solid-phase microextraction (SPME) is a relatively new technique that is able to address the need for concentrating the analytes in the headspace.3 SPME uses a small (1 cm) piece of fused silica, on which a liquid phase, similar to a GC stationary phase, has been coated to absorb the desired analytes and concentrate them on the fiber. The selectivity of the extraction of target analytes in the gaseous phase can be significantly altered through the use of different liquid phases on the fiber.4 In this study, the analysis of volatile flavor compounds from fruit beverages was originally undertaken through the use of gassampled static headspace extraction.5 The low recovery of the flavor volatiles was a persistent problem with the gas-sampled static headspace method. SPME was then investigated. It was found that the SPME fiber’s ability to concentrate the analytes provided a dramatic improvement in the amount of analyte placed on column and produced a method capable of monitoring flavor compound oxidation. Different fibers were experimented with to determine an optimum fiber for the detection and quantification of the flavor volatiles. Real samples were then analyzed fresh (prior to the expiration date) and after 3 months, to determine the effect of aging on the samples. Dramatic differences between the fresh and aged profiles of the orange juice samples were noted, especially the oxidation of limonene to carvone, as noted in the literature.6 This differs from other reports that claim little or no oxidative loss of limonene occurs in complete orange juice samples.7 EXPERIMENTAL SECTION All experiments were carried out using a Varian 3400 gas chromatograph (Varian Instrument Group, Walnut Creek, CA), coupled to a Finnigan MAT 700 ITD ion trap mass spectrometer (Finnigan MAT, San Jose, CA). A splitless injection was made with the inlet temperature set at 220 °C, with the purge vent opening 1.00 min after injection. The column used was an HP-5 (3) Zhang, Z.; Yang, M. J.; Pawliszyn, J. Anal. Chem. 1994, 66, 844A-853A. (4) Yang, X.; Peppard, T. LC-GC 1995, 13, 882-886. (5) Miller, M. E.; Stuart, J. D.; Smith, S. R.; Widmer, W. Pittsburgh Conference, March 1997; Abstr. 277P. (6) Buckholz, L. L.; Daun, H. J. Food Sci. 1978, 43, 535-539. (7) Marsili, R. LC-GC 1986, 4, 358-362.
Analytical Chemistry, Vol. 71, No. 1, January 1, 1999 23
MS capillary column, 25 m in length, 0.20-mm i.d., with a 0.33µm film thickness (Hewlett-Packard, Wilmington, DE). The oven program was 30 to 200 °C at 4 °C/min. The transfer line from the GC to the ITD was maintained at 220 °C and consisted of a 0.25mm-i.d. length of uncoated, deactivated fused silica. The ITD was set to scan over a mass range of 45-200 amu. Helium was used as the carrier gas at a flow rate of 1.0 mL/min. Gas-Sampled Static Headspace. For gas-sampled static headspace determinations, a Tekmar 7000/7050 static headspace unit with autosampler (Tekmar-Dohrmann, Cincinnati, OH) was used. This instrument was designed such that the headspace was sampled by pressurizing the equilibrated headspace with carrier gas to a pressure above atmospheric pressure and then allowing the pressurized headspace to vent to atmosphere via a sample loop, thereby filling the loop. Injection of the contents of the sample loop was accomplished by placing the sample loop in-line with the carrier gas, which then flows into the GC inlet port. Prior to analysis, samples were equilibrated at either 40 or 120 °C. The 120 °C sampling procedure represented an optimized set of headspace conditions. The 40 °C condition was used to evaluate the direct comparison of the two (gas-sampled and SPMEsampled) extraction techniques, using conditions that mimic one another as closely as possible. The sample was equilibrated for a period of 20.0 min, followed by mechanical agitation for a period of 10 min. The sample loop (0.50 mL) and transfer line (from autosampler to GC) were stainless steel, treated with a fused silica coating (Polar Option), chosen to ensure as much deactivation of the sample pathway as possible. The injection parameters were set as follows: For the 40 °C studies, vials were pressurized to 10 psi for 0.50 min, and the pressurized headspace equilibrated for 0.10 min. For the 120 °C studies, these steps were omitted since the static pressure in the vials after equilibration was sufficient to allow for the proceeding steps to take place without additional pressurization. It should be noted that all vials should be pressurized to the same pressure with carrier gas, to ensure reproducible results. This was omitted in our case to prevent septum leakage, which occurred when vials were left on the injector needle for the additional time needed to pressurize and equilibrate the sample. Once pressurized, sample headspace gases were vented to atmosphere via the 0.50-mL sample loop for 0.10 min, to fill the loop. The contents of the loop were then equilibrated for 0.10 min, and finally the sample loop was placed in-line with the carrier gas via a six-port valve for 1.00 min. This action injected the sample into the GC and simultaneously started the GC and MS detector for data acquisition. SPME-Sampled Static Headspace. For the SPME determinations, a manual SPME holder and fibers were used (Supelco, Bellfonte, PA). The fibers were conditioned as recommended by the manufacturer. Between uses, fibers were kept sealed from ambient air by piercing the tip of the SPME needle into a small piece of septum to prevent accidental contamination. The sampling procedure involved placing 10.0 mL of sample into a 20-mL VOA vial and sealing with a screw-top septum-containing cap. The SPME needle was then inserted through the septum and the vial placed in a water bath maintained at 40.0 °C. It should be noted that the vial was submerged only as far as necessary to submerge the liquid phase of the sample, to help keep the SPME fiber cool, which is a desired condition for SPME. This is because as the 24
Analytical Chemistry, Vol. 71, No. 1, January 1, 1999
temperature of the fiber increases, the partition coefficient decreases.8 The SPME holder was secured and the fiber extended into the headspace, and the fiber was allowed to equilibrate for 30 min. The fiber was then retracted, removed from the vial, and placed immediately into the inlet of the GC. Injection was accomplished by extending the fiber in the heated inlet for ∼6 min, while the injector operated in the splitless mode for 1.00 min. The additional 5 min of exposure time in the injector port allowed the fiber to be cleaned of any compounds that may not be desorbed in the initial minute. Preliminary studies indicated that the above procedure allowed for reproducible, quantitative transfer of target analytes into the injector port of the GC/MS. Standard Preparation. Concentrated standard solutions were prepared from neat compounds by dilution of 10.0 µL of each compound into 4.0 mL of purge and trap grade methanol. A 5.0µL aliquot of the concentrated standard was then diluted into 50.0 mL of distilled water to make a working standard, from which all analytical standards were produced. These standards were used to verify and quantify the identities and concentrations of 13 compounds that were identified by retention index and mass spectral interpretation of the total ion chromatograms resulting from the extraction of the juice samples. Retention index and mass spectral data were compared to values published by Adams.9 Compounds identified and quantified by standards were as follows: R-pinene, β-pinene, myrcene, R-phellandrene, limonene, β-ocimene, γ-terpinene, terpinolene, linalool, 4-terpineol, R-terpineol, carvone, and valencene. Concentrations of the compounds in the standard solutions ranged from 1 to 200 ppb. Another standard mixture, prepared in the same way, was made containing nine compounds: myrcene, limonene, valencene, linalool, R-terpineol, acetophenone, carvone, benzaldehyde, and n-nonanal. These compounds were chosen as representative of the terpenes, alcohols, ketones, and aldehydes present in many of the samples. Beverages Analyzed. The fruit juices used in these experiments were orange juice (Tropicana Season’s Best, Veryfine), grapefruit juice (Veryfine), lemon juice (Real Lemon), and lime juice (Real Lime). All were store bought and refrigerated at 4 °C during storage. All samples were chosen because they were advertised as being made from real fruit juice, to ensure a wide variety of flavor analytes being present. RESULTS AND DISCUSSION Comparison of Gas-Sampled and SPME-Sampled Static Headspace. Originally, these experiments were to be performed exclusively by gas-sampled static headspace. In these preliminary experiments, a variety of juices and alcoholic fruit beverages were tested. It was found that very few compounds were detected in the headspace of the samples by this method. Seven compounds in total were detectable from four different beverages. Figure 1A shows a representative chromatogram obtained by the gassampled static headspace technique. The detected compounds include R-pinene, myrcene, and limonene. These compounds, though detected, were not considered to be analytically useful, since the concentrations of R-pinene and myrcene were often well below the minimum detection limits of the technique. The conditions employed to obtain the chromatogram shown in Figure (8) Zhang, Z.; Pawliszyn, J. Anal. Chem. 1995, 67, 34-43. (9) Adams, R. P. Identification of Essential Oil Components by Gas Chromatography/Mass Spectroscopy; Allured Publishing: Carol Stream, IL, 1995.
Table 1. Comparison of Gas-Sampled vs SPME-Sampled Static Headspace Techniques gas-sampled static headspace
SPME-sampled static headspace
analyte
DL (ppb)
slopea
intercept
R2
DL (ppb)
slopeb
intercept
R2
benzaldehyde myrcene limonene acetophenone linalool n-nonanal R-terpineol carvone valencene
2890 93.0 5.88 2210 319 178 354 285 0.477
1.692 5.183 11.18 1.393 9.474 19.09 6.256 3.668 104.9
-4044 1643 713.4 -1465 -2376 -2214 -1249 -1019 4241
0.9774 0.9961 0.9994 0.9998 0.9911 0.9955 0.9952 0.9999 0.9999
4.72 4.80 0.516 26.9 2.44 3.02 5.52 14.4 0.275
3542 20550 28623 4000 12434 31698 3496 5506 35620
26939 56771 1333125 -2831.5 -29521 -53964 -18637 12251 544217
0.9939 0.9999 0.9993 0.9996 0.9998 0.9993 0.9984 0.9990 0.9976
a
Average range, 0.5-11 ppm. b Average range, 1.0-400 ppb.
Figure 1. Representative chromatograms of flavor volatiles by gassampled static headspace (A) and SPME-sampled headspace (B). Identified peaks: (1) R-pinene, (2) myrcene, (3) limonene, (4) R-terpinene, (5) terpinolene, (6) linalool, (7) 4-terpineol, (8) R-terpineol, and (9) n-nonanal (refer to the Results and Discussion section for experimental details).
1A included an equilibration of the vial for 10 min, followed by shaking for 5.0 min, both at a temperature of 140 °C. These conditions were deemed to be too extreme to generate data about a sample without worry of sample degradation since the sample was obviously discolored after analysis. In an attempt to place more analyte mass on the column, cryogenic trapping, on-column injections, larger sample loops, and multiple injections were explored with little success. Although certain of these techniques did aid in getting more analyte on the column, chromatographic resolution suffered greatly. Due to the low flow rate and smalldiameter capillary column demanded by the ion trap mass spectrometer, the large injections degraded chromatographic performance to such a point that the techniques mentioned above were unusable. It should be noted that many of these failures are most likely the result of the system employed here. The use of larger bore, thicker film columns with higher carrier gas flow rates
would allow for more efficient use of these techniques, resulting in some improvement of the results. The use of NaCl to provide a “salting-out” effect on the analytes was also explored. The addition of 1.5 g of NaCl yielded no significant effect on the recoveries of the nonpolar analytes, as reported elsewhere.10 SPME-sampled static headspace was then used to concentrate the analytes of interest in these determinations. Initial comparison studies were performed using a 100-µm-film poly(dimethylsiloxane) (PDMS) fiber, considered appropriate for the analysis of VOCs.11 This fiber showed dramatic increases in the recoveries of the volatile flavor components in the sample. Figure 1B shows a representative chromatogram of the SPME-sampled static headspace recoveries. In initial comparisons of the two techniques, samples were run under similar conditions, specifically 40 °C equilibration temperature and 30.0-min total equilibration time. The other sampling parameters were optimized for each technique. Preliminary studies showed the SPME-sampled recoveries generated peaks ∼1400 times larger than the equivalent gassampled headspace technique. In later rigorous comparison studies, standards were run using both techniques under optimized conditions and the slopes of the calibration plots compared to evaluate the relative sensitivities. For the SPME-sampled experiments, the 65-µm PDMS/DVB fiber was used. Experiments were run in concentration regimes that corresponded to the respective lower limit of detection for the method, specifically, the gas-sampled experiments were run in the part-per-million range, while the SPME-sampled experiments were run in the partper-billion range. Calibration studies on the SPME-sampled technique were also run at the part-per-million range, but at this concentration level, the calibration curves for the various standards were nonlinear. This suggests that the two techniques are complementary. The results are given in Table 1. Of the nine compounds analyzed, the average sensitivity increase from gassampled to SPME-sampled static headspace was ∼1800, as estimated by comparing the slopes of the corresponding calibration curves. This result compares favorably to the value of 1400 obtained earlier. Other authors have noted lesser improvements in sensitivity when comparing gas-sampled and SPME-sample static headspace, but their work dealt with more volatile compounds in simpler matrixes, allowing the gas-sampled method to (10) Steffen, A.; Pawliszyn, J. J. Agric. Food Chem. 1996, 44, 2187-2193. (11) Pawliszyn, J. Solid-Phase Microextraction: Theory and Practice; Wiley-VCH: New York, 1997.
Analytical Chemistry, Vol. 71, No. 1, January 1, 1999
25
Figure 2. Comparison of 100-µm PDMS, 85-µm polyacrylate, 75-µm Carboxen, 65-µm PDMS/DVB, and 65-µm CW/DVB fibers. All recoveries are scaled to the 100-µm PDMS, which is assigned 100% for comparison purposes.
Figure 3. Changes in recoveries due to sample degradation. Large increases in trans-carveol and carvone arise from the oxidation of limonene and other terpenes. Note: Concentration axis is plotted in log format to allow all information to be shown in a single figure.
perform better.12 Reproducibility of this technique was found to be good with an average relative standard deviation (RSD) of 11.3%. Comparison of Different Fibers’ Recoveries. Five different fibers were evaluated to determine which fiber most effectively extracted flavor volatiles from fruit juice samples. The five fibers (12) Jelen, H. H.; Wlazly, K.; Wasowicz, E.; Kaminski, E. J. Agric. Food Chem. 1998, 46, 1469-1473.
26 Analytical Chemistry, Vol. 71, No. 1, January 1, 1999
studied are shown in Table 2. These fibers were used to extract analytes from the headspace of aliquots of the same sample for comparison of the relative recoveries of 13 analytes. The results of the experiments on the 100-µm PDMS fiber, 85-µm polyacrylate fiber, 75-µm Carboxen fiber, 65-µm PDMS/DVB fiber, and the 65-µm CW/DVB fiber are summarized in Figure 2. These results show that, of the fibers evaluated, the PDMS/DVB fiber proved to be the most effective at extracting flavor volatiles overall,
Table 2. Types of Fibers Used
fiber
phase
film thickness (µm)
PDMS polyacrylate Carboxen PDMS/DVB CW/DVB
polydimethylsiloxane polyacrylate porous carbon PDMS/divinylbenzene Carbowax/divinylbenzene
100 85 75 65 65
followed by the PDMS fiber, then the CW/DVB fiber and polyacrylate fiber, and finally the Carboxen fiber. These data confirmed that the 65-µm PDMS/DVB fiber performed the best for these samples and, therefore, was used for all subsequent comparison experiments. This fiber performed the most effective extractions, for this analysis, due to the fact that the fiber coating is composed of a mixed coating containing PDMS, a liquid phase that favors the absorption of nonpolar analytes, as well as DVB, a porous solid that favors the adsorption of the more polar analytes. It can be seen that there is little difference between the PDMS fiber and the PDMS/DVB fiber in extracting the nonpolar analytes (terpenes), but the more polar aldehydes and ketones are extracted, on average, a factor of 3 times greater by the PDMS/ DVB fiber as measured by peak area. Also, the CW/DVB and the polyacrylate fibers that are more selective to polar analytes did show enhanced recoveries of the polar analytes, but were less effective in recovering the nonpolar analytes. Finally, the Carboxen fiber showed recoveries that varied independent of analyte polarity. This is due to the differing adsorption mechanism of the Carboxen fiber. The Carboxen fiber is composed of a porous carbon that is best for small molecules (C2-C6 analytes).13 Therefore it would be expected to and did show poor sensitivity to the analytes of interest in this study. Comparison of Various Beverages by SPME Sampling. The beverages used in this study were analyzed for differences in the flavor profiles of various fruit juices. The various juices did yield different chromatograms, but no analyses were performed to quantify the differences. The experiments were performed simply to verify that differences could be detected between samples. It was observed that all samples contained large amounts of limonene, which was expected since limonene accounts for up to 95% of the compounds found in citrus oils.14 Many other compounds, up to 20, were present at various levels, these being some of the flavor compounds that differentiate the various fruits. When compared to the Tropicana orange juice sample, many differences between the samples were apparent. These differences include variations in the R-pinene to β-pinene ratio, as well as differences in the levels and composition of the oxidized terpenes. Compounds, such as the terpineol isomers varied depending on the sample. For example, 1-terpineol was present only in the lime juice sample, while the 4-terpineol to R-terpineol ratio was much (13) Supelco Web Site, “Frequently Asked Questions” (http://www.supelco.sial.com/supelco/spme/spmefaq.htm). (14) Lee, H. S.; Widmer, W. W. from the Forty-Fourth Annual Citrus Processors’ Meeting, October 28, 1993, University of Florida, Gainesville, FL. (15) Dieckmann, R. H.; Palamand, S. R. J. Agric. Food Sci. 1974, 22, 498-503.
greater in the grapefruit and orange juice samples than it was in the lime and lemon juice samples. Measurement of Orange Juice Oxidation. The SPME method in use here was able to detect terpene oxidation, which is an important analysis to those evaluating flavor extracts in the beverage industry. Two samples of the same orange juice brand were analyzed, one being newly purchased, the other having been stored 3 months in a refrigerator. The 3-month aging time was derived from the fact that the sample purchased at the outset of the experiments was used and then compared to a sample purchased 3 months into the experiments. The aged sample was stored in a refrigerator at 4 °C, sealed in a Nalgene bottle with minimal headspace. The storage conditions were not meant to rigorously mimic real storage; the only intent was to verify that changes in the sample could be detected. The changes in the sample were monitored by comparing the peak areas obtained for each of the compounds in the two samples. Figure 3 shows the resulting data. The changes include loss of the terpenes and certain alcohols originally present in the sample. Over time, these compounds were oxidized and formed other alcohols and ketones, in particular carvone.6 Suggested reaction pathways for these reactions were proposed by Dieckmann and Palamand and are based on the reaction of the terpenes with molecular oxygen.15 CONCLUSIONS Static headspace analysis of volatile compounds is a widely used technique. However, for many analyses, the gas-sampled static headspace method lacks the sensitivity needed to perform adequately. SPME has the ability to perform these analyses where gas sampling falls short. In this study, the comparison of gassampled and SPME-sampled static headspace showed that SPME determinations of flavor compounds are, on average, ∼3 orders of magnitude more sensitive under the conditions employed in this study. This increased sensitivity allows fast, accurate determinations of flavor compounds in a convenient, easy to perform analysis. Monitoring of juice oxidation was performed, as well as the profiling of different juice samples. Different fibers were investigated with the best fibers found to be 100-µm PDMS and 65-µm PDMS/DVB. These fibers gave excellent recoveries of the terpenes, with the advantage of the 65-µm PDMS/DVB being better able to recover the more polar alcohols, ketones, and aldehydes. ACKNOWLEDGMENT The authors thank Tekmar-Dohrmann for the loan of the Tekmar 7000/7050 and for financial support of some of the work presented here. Standards were graciously supplied by Dr. Wilbur Widmer, Jr., of the Florida Department of Citrus. We appreciate the valuable discussions with Richard Gaines of the Coast Guard Academy, New London, CT, and Robert Shirey of Supelco, Bellfonte, PA. Received for review May 27, 1998. Accepted October 15, 1998. AC980576V
Analytical Chemistry, Vol. 71, No. 1, January 1, 1999
27
