Sampling and sample preparation for analysis of aromas and fragrances
Transcription
Sampling and sample preparation for analysis of aromas and fragrances
Trends Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 Sampling and sample preparation for analysis of aromas and fragrances Fabio Augusto, Alexandre Leite e Lopes, Cla´udia Alcaraz Zini Some recent advances in sampling and sample-preparation technologies for fragrance analysis are addressed in this review. Procedures, such as analytical distillation (vapor distillation and simultaneous distillationextraction), headspace-manipulation methods (static and dynamic headspace analysis and headspace solid-phase microextraction) and direct extraction methods (such as liquid-liquid, solid-phase and supercritical fluid), will be discussed and critically evaluated. Contemporary applications of these techniques to the study of natural and synthetic aromas will be presented. # 2003 Published by Elsevier Science B.V. Keywords: Aroma and fragrance analysis; Sample-preparation techniques Fabio Augusto*, Alexandre Leite e Lopes Institute of Chemistry – State University of Campinas (Unicamp), C.P. 6154, 13083907 Campinas, Sa˜o Paulo, Brazil Cla´udia Alcaraz Zini Institute of Chemistry – Universidade Federal do Rio Grande do Sul (UFRGS), 90040-060 Porto Alegre, Rio Grande do Sul - Brazil *Corresponding author. Tel.: +55-19-3788-3057; Fax: +5519-3788-3023 (faculty); E-mail: [email protected]. br 160 1. Introduction The term ‘‘odor’’ refers to biological, physical and psychological e¡ects caused by the interaction between chemical stimulants ^ aromas and fragrances ^ and olfactive systems of living creatures [1]. The study of the composition of fragrances and aromas (F&A) is relevant in several ¢elds, as follows. In food science, chemicals or their blends associated with odors discerned as desirable or agreeable are related to a multivariate set of sensorial responses know as ‘‘£avor’’ [2]. Flavors are the predominant elements in the perception of food and beverage quality [3], and, therefore, its acceptance by consumers. Also, ‘‘o¡-£avors’’ ^ connected to volatile substances with unpleasant odors ^ may be caused by microbial contamination of foodstu¡s [4]. Investigation of these substances can be important in food-safety studies. Odor can be a signi¢cant environmental parameter in some situations, since it can be related to the human perception of comfort and to the sanitary conditions of an indoor atmosphere [5], to the contamination of air resulting from agricultural [6] and industrial [7] activities, as well as to the quality of natural and treated water [8]. Chemical characterization of fragrances released by vegetables and animals is important to several branches of industry and science. Natural essential oils from plants are traditional base ingredients for perfumes and related toiletry products. However, the rarity of some plants and the possible dermatotoxic e¡ects caused by some natural products create demand for synthetic alternative mixtures. The design of these ‘‘synthetic’’ essential ingredients depends fundamentally on reliable analytical data on the composition of their natural counterparts [9]. Moreover, knowledge about the composition and emission dynamics of £oral scents and similar biogenic mixtures of volatile organic compounds is fundamental in several biological studies because of their many di¡erent roles in plant reproductive processes, defense against predators and intra-species communication [10]. The development and application of methodologies for the determination of the chemical composition of aromas and similar mixtures is a challenging task. 0165-9936/03/$ - see front matter # 2003 Published by Elsevier Science B.V. doi:10.1016/S0165-9936(03)00304-2 Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 This is a consequence of some inherent general properties of such samples: (a) The concentration of relevant analytes in fragrance samples can be extremely low. For example, the literature values for low-odor threshold concentrations (which can be de¢ned the minimum concentration for an odor-producing substance that has a 50% probability to be perceived as a faint odor by a human subject [1]) for selected substances important in food, biological and environmental aromas, such as ethyl 4-methyl-4-pentenoate, 2-iodophenol, butyric acid and 2-ethyl-3,5-dimethylpyrazine, are respectively 0.06 nL/L, 1 mg/L, 240 nL/L, 0.007 nL/L [11^13]. The typical perceptible concentration for some odorants can, as shown, be less than 1 ng/L of the substance. As a result, the corresponding analytical procedures must provide extremely high sensitivities, adequate for detection and quanti¢cation of these species at such low levels. (b) Apart from the typically low concentration of the odor-active analytes, most aromas are extremely complex blends of substances. Kotserides and Baumes [14] reported 48 di¡erent organic volatile substances as odorants with impact in Bordeaux wines, Cabernet Sauvignon and Merlot. The same complexity can be found in £oral scents; Bayrak and Akgu«l [15] identi¢ed more than 60 volatile organic compounds in samples of essential oils from Turkish rose (Rosa damascena, Miller). (c) For aromas generated by biological sources, such as plants and animals, further analytical di⁄culties arise from the dynamic nature of such systems. For example, the composition of samples obtained from detached parts of plants may not correspond to the mixture released by undisturbed live organisms [16,17]; this creates a demand for methods allowing in-vivo sampling. Also, the fact that production and emission of plant volatiles can be a¡ected or triggered by factors, such as light, environmental temperature, stress and the presence of trace atmospheric pollutants [18,19], can introduce further complications. (d) Some odorants have limited chemical stability, as a result of photolysis, oxidation and other reactions; e.g., the atmospheric chemical lifetime of monoterpenes under daylight conditions was estimated to range from less than 5 min (a-terpinene) to 3 hours (a- and b-pinene, sabinene) [18]. As a result, chemical characterization of F&A ordinarily demands state-of-art techniques for sampling and sample preparation, analyte separation, detection and quantitation. Usually, the application of an analytical separation technique in the ¢nal steps [2] is involved. Gas chromatography coupled to mass spectrometry (GC-MS) and other similar detection schemes, such as infrared absorption spectrometry [20], are the techniques normally employed for F&A chemical analyses. Trends The coupling of olfactometric detection to gas chromatography (GC-O) is also extremely relevant for qualitative and quantitative chemical analysis of fragrances [21]; several instrumental and methodological variants of GC-O are described in the literature, such as CHARMAnalysis and aroma extraction dilution analysis (AEDA) [22]. Along with the chromatographic techniques, the use of ‘‘electronic noses’’ (arrays of electrochemical sensors that generate an electric signal that emulates the expected response from the human olfactory system) has been growing in recent years. The development of these fascinating devices involves the arrival of suitable new materials, such as piezoelectric crystals and synthetic conductive polymers, as well as sophisticated data-processing and chemometric techniques [23]. Their more remarkable features, from the analytical standpoint, are the possibility of fast, direct, qualitative and quantitative evaluation of F&A with limited or no preliminary sample-preparation procedures ^ although the low sensitivities provided by the devices presently available still prevent their use in solving most F&A analytical problems. Apart from applications involving measurement with the above-mentioned electronic noses and similar techniques, adequate isolation and pre-concentration of the odor-active analytes are mandatory and critical steps in the methodologies for chemical characterization of F&A [2]. Several sorbent-extraction approaches ^ classical liquid-liquid extraction (LLE), solid-phase extraction (SPE), solid-phase microextraction (SPME) and supercritical £uid extraction (SFE) ^ have been employed for sample preparation in F&A analyses, as well as dynamic headspace (DHS) and static headspace (SHS) methods and procedures based on analytical distillation. In the following sections, the application of some of these techniques to procedures for chemical characterization of £avors and F&A will be addressed. 2. Analytical distillation procedures Distillation is usually carried out in two slightly di¡erent ways, as follows. In the ¢rst, the matrix to be extracted is mixed or suspended with water in a suitable vessel ¢tted with a condenser and, while the mixture is boiled, a condensate phase is collected (hydrodistillation). After the process, an organic, water-insoluble fraction (for vegetable materials, the essential oil) can be separated from the water. In the second procedure, steam is passed through a vessel containing the matrix-water mixture (steam distillation) to yield a similar condensate. http://www.elsevier.com/locate/trac 161 Trends Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 2.1. Vapor distillation Steam-assisted distillation and hydrodistillation are traditional procedures for isolation of volatile aromarelated compounds from odoriferous samples, such as food and detached parts of plants. Being simple and straightforward procedures, they are still extensively applied for F&A characterization either alone or combined with other sample-preparation procedures. For example, Saritas et al. [24] isolated the essential oils from aromatic lichens of various genera (Mnium, Plagiomnium and others) by hydrodistillation of fresh and dried plant parts. Using GC-MS combined with 13 C-NMR and other techniques, they were able to identify several volatile terpenoid and aliphatic compounds in these hydrodistillates, including two unreported volatile sesquiterpenes ((+)-10-epi-muurola-4,11diene and 10,11-dihydro-a-cuparenone). 2.2. Simultaneous distillation-extraction (SDE) A widespread distillation-based sample-preparation method for chemical analysis of F&A is simultaneous distillation-extraction (SDE), also known as the Lickens-Nickerson method. Fig. 1 shows one of the proposed designs for the Lickens-Nickerson SDE apparatus. An amount of the fragrance-generating sample (food, beverage, sliced plant tissue, etc.), along with distilled water for dry samples, is contained in £ask 2, and £ask 3 receives a suitable volume of an extracting solvent denser than water (dichloromethane, chloroform, etc.). Flasks 2 and 3 are heated; the water and solvent vapors are conducted to the extractor body, where they are allowed to condense over the surface of the cold tube 4. Figure 1. Likens-Nickerson simultaneous distillation-extraction apparatus (modified from Perpe`te et al. [26]): 1 – body; 2=sample flask; 3=extracting solvent flask; 4=cold tube; 5=inlet for purge gas; 6=cold water inlet and outlet. 162 http://www.elsevier.com/locate/trac In this operation, aroma analytes are removed from the matrix by the water vapor and transferred to the organic phase when the liquids condense together on the cold tube. Both water and solvent are collected in the extractor body after their condensation and return to the corresponding £asks, allowing continuous re£ux. A modi¢ed apparatus allows use of solvents lighter than water, such as pentane or ethyl acetate, as extractors. Among others, this device was employed by Schlotzhauer et al. [25] to study insect attractors in the fragrance of £owers from Japanese honeysuckle (Lonicera japonica). Large amounts of linalool and a-farnesene were found in samples collected during daylight periods. Perpe'te et al. [26] combined LLE and SDE to ¢nd the key compound responsible for the unpleasant o¡-£avor in alcohol-free beers (1-methylpropionaldehyde), which was not previously found in reports using other sample-preparation techniques, such as DHS analysis. 2.3. Other distillation methods Atmospheric-pressure distillation-based sample-preparation methods may have some serious disadvantages when applied to F&A characterization, since the temperatures usually necessary for the operation can lead to degradation of some analytes. For example, after studying the aroma from sa¡ron (Crocus sativus L.), Tarantilis and Polissou [27] found that the most important odorant in these samples, safranal, is oxidized during conventional steam distillation, forming artifacts, such as 2,6,6-trimethyl-1,3-cyclohexadien-1-carboxylic acid. Other compounds found in these distillates were also determined as artifacts resulting from the thermal degradation and oxidation of carotenoids from the sa¡ron matrix. The same problems can be observed in SDE: Siegmund et al. [28] discovered that 5,6-dihydro-2,4,6-trimethyl-4H-1,3,5-dithiazine, usually pointed to as an important aroma-active compound in cooked and cured meat aromas, is an artifact formed during the distillations performed according the Lickens-Nickerson SDE method. Vacuum distillation, which can be performed under milder conditions, can overcome these problems and is an alternative to regular distillation methods. Moio et al. [29] combined vacuum distillation and LLE to identify the key odorants from Gorgonzola cheese. After GC-MS and GC-O analysis of the distillates/extracts, 2-nonanone, 1-octen-3-ol, 2-heptanol, ethyl hexanoate, methylanisole and 2-heptanone were determined to be the most relevant odorants in the aromas of natural and creamy Gorgonzola cheeses. Another interesting application of low-pressure distillation was described by Bouchilloux et al. [30]. Using highvacuum distillation (temperatures between 35 C to 45 C and pressures between 0.5 Pa and 1.5 Pa) and simultaneous derivatization with p-hydroxymercuribenzoic Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 acid, the authors were able to determine trace levels of some powerful odorous thiols in red wines of Bordeaux. The analytes were identi¢ed and quanti¢ed in the samples as their volatile organomercury derivatives, in concentrations ranging from 0.25 mg/L to 10 mg/L (3-mercapto-2-methylpropanol), 10 ng/L to 5 mg/L (mercaptohexanol) and 1 ng/L to 200 ng/L (3-mercaptohexyl acetate). 3. Headspace methods Procedures based on the manipulation of the headspace (HS) in contact with odorous materials are popular and very suitable for chemical analysis of F&A [31]. Di¡erent approaches have been employed, either using direct HS analysis or by collection of the odorants in the HS using sorbent devices or cold traps. These techniques and their applications for the chemical characterization of F&A are discussed in the following paragraphs. 3.1. SHS The simplest way to assess the chemical composition of an aroma is direct analysis (by GC or another convenient technique) of a portion of the air in contact with the odor source, without any other sample-treatment step. However, since little or no pre-concentration of the analytes is involved and, considering the typical trace or ultratrace levels of the components, it is not usually feasible to apply SHS to chemical-fragrance analysis [32]. For example, of 118 references indexed in an extensive literature review on HS analysis of £oral fragrances [33], only nine employed SHS as the sampling or sample-preparation procedure. Nonetheless, when techniques and devices with adequate detection limits are available, and depending on the concentration of the analytes, SHS can be particularly suitable because of its inherent simplicity. For example, Clarkson and Cooke [34] employed GC-MS and GC-AED (atomic emission detection) systems equipped with high-volume injection ports to analyze aroma compounds in the HS of commercial cigarettes. Injecting 1 mL of the HS air in contact with the samples, the authors pointed out that, for these samples, both analysis time and detection limits were better than those obtained using a commercial DHS analyzer. 3.2. DHS A large number of reports describing the use of DHS methods can be found in the literature regarding the chemical characterization of F&A. Because of the constant depletion of the analytes from the sample or from the adjacent atmosphere, the general approach on DHS potentially provides improved analytical sensitivity when compared to SHS and other equilibrium extraction procedures [2]. Desorption of trapped analytes for Trends subsequent analysis can be performed either with small volumes of adequate solvents [35] or using on-line automated thermal desorption (ATD) devices [36], the later being eminently suitable for routine procedures. Experimental set-ups such as that depicted in Fig. 2 [37] can be used, either in the laboratory or under ¢eld conditions, to collect fragrance compounds emitted by live plants in dynamic conditions. A vase containing a live plant or some parts of it can be isolated from the environment by a glass bell or similar device. A controlled £ow of puri¢ed air is passed through the chamber, carrying the aroma compounds to a tube containing a sorbent or to a cold tube, where the volatiles are accumulated for further desorption and analysis. Using a similar apparatus with polyvinyl-acetate bags as the isolating device instead of the glass bell to emulate the usual procedure employed in ¢eld sampling of £oral scents, Raguso and Pellmyr [38] performed an extensive study of the methodological aspects of sampling and pre-concentration of £oral fragrances by DHS with solvent desorption and chromatographic analysis. Live £owers of Clarkia breweri were used as fragrance source, as well as synthetic mixtures of typical £ower odorants. It was found that Porapak Q was more e⁄cient than Tenax TA or Tenax TA/charcoal mixtures as sorbent for the evaluated fragrances. However, the relative abundance of the detected compounds ^ a useful parameter in biochemical and ecochemical studies ^ did not depend on the composition of the sorbent. The trapped compounds were desorbed using either diethyl ether, dichloromethane or hexane; the last solvent provided better recoveries. In the range studied (0.5^1.0 L/min), the e¡ect of the stripping air£ow rate on the recovery of the trapped analytes generally was not signi¢cant when compared to the Figure 2. Chamber for DHS isolation and collection of volatiles emitted by living plants (modified from Vercammen [37]). http://www.elsevier.com/locate/trac 163 Trends Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 e¡ects of other sources of variation on the results between experiments, such as variation in the composition of the volatile emissions between di¡erent specimens of the same plant. Although it can require more complex and expensive hardware (prices up to US$ 15,000) and cannot be applied to fragrances containing temperature-sensitive compounds, DHS with TD (DHS-TD) has several advantages over solvent desorption: simpler procedure; improved detection limits; and; no interference of solvent peaks in the chromatograms. Ageloupoulos et al. [39] monitored the temporal emission pattern of the volatile compounds (unsaturated C6-alyphatic alcohols and aldehydes) released by single intact and damaged leaves of potato (Solanum tuberosum) and broad bean (Vicia faba) by DHS (using a specially designed sampling chamber) coupled to GC-MS. The volatiles were trapped in Tenax TA cartridges and desorbed either thermally (inserting the cartridges inside the programmed-temperature injector of the GC-MS system) or by a solvent (diethyl ether). It was determined that TD, which provided better detectability and therefore required reduced sampling times, allowed the time-emission pro¢les to be assessed for the monitored compounds. Vercammen et al. [37] also employed a commercial on-line ATD apparatus coupled to a GC-MS to compare sorption tubes packed with Tenax TA and with pure grinded polydimethylsiloxane (PDMS), as well as SHS SPME, for DHS collection of fragrance compounds released by live roses and jasmine. PDMS was found to be more e⁄cient for the isolation of volatile odorants released by these plants, resulting also in cleaner chromatograms. Analytical methodologies involving DHS-TD for characterization of the aroma of foods and beverages are also common. To study the release of volatile £avor compounds by leaves of Japanese pepper (Xanthoxylum piperitum), a spice employed in traditional oriental cuisine, Jiang and Kubota [40] collected the volatiles in the HS of crushed, mechanically disturbed or intact leaves in Tenax TA cartridges. The analytes were thermally desorbed in the injector port of a GC-MS, where they were cryofocused, separated, detected and quanti¢ed (Fig. 3). It was determined that the number of detected volatile compounds increased from 12 (from intact leaves) to 22 (mechanically disturbed leaves) and then to 36 (crushed leaves). For the disturbed leaves, it was found that the main compounds responsible for the characteristic aroma were limonene (0.56 mg/g) and (Z)-3-hexen-1-ol (0.63 mg/g). For crushed leaves, higher concentrations of C6-aldehydes, especially (Z)-3-hexenal (17.88 mg/g), and hexanal (6.77 mg/g), imparted an undesirable grassy odor to the samples. Introduction of artifacts produced by degradation of the sorbent can be a major di⁄culty in DHS-TD; e.g., Canac-Arteaga et al. [41] found that the interaction of water vapor, volatile compounds and the trapping 164 http://www.elsevier.com/locate/trac Figure 3. GC-MS chromatographic profiles after DHS-ATD collection of volatiles released from Japanese pepper (Xanthoxylum piperitum) leaves (modified from Jiang and Kubota [40]). material (Tenax TA) can produce artifacts during the analysis of aromas of dehydrated cheese and Parmesan cheese. Retention of water by the trapping material can also be a potential problem, especially when it is necessary to cryofocus the analytes in the chromatographic column before separation (moisture can freeze and clog the column). Carbon molecular sieves (Carbosieve, Carboxen) can retain large amounts of water, as opposed to non-polar polymeric sorbents, such as Tenax and graphitized carbon blacks. Polar polymers, such as Porapak T and N, can also hold appreciable amounts of water, which can, however, be easily removed prior to analyte desorption by a current of dry air [42]. Introduction of products from the thermal degradation of adsorptive materials can be avoided in DHS procedures by using cryotrapping. Stra¤nsky¤ and Valterova¤ [43] determined the emission pro¢le of compounds related to the putrid fragrance from Hydrosme rivieri £owers during their £owering period. A single, live specimen was enclosed in a glass cylinder, which was purged by 15 mL/min of pure air. The compounds released were trapped in a U-shaped tube containing methanol at Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 78 C. It was found that emission of dimethyl di- and tri-sulphides started after 70 h of £ower enclosure, peaking at 118 h and ending after 142 h (Fig. 4). Before 70 h of enclosure, only n-alkanes were detected. To eliminate water co-volatilized with the analytes, Kolb et al. [44] proposed modi¢cation of a commercial on-line DHS-cryotrapping device by incorporating a water trap (glass-lined steel tube packed with 65% LiCl on Chromosorb W/AW). The authors reported good results in the application of this technique to several processes, including the characterization of aroma from ground co¡ee. 3.3. HS-SPME SPME, a fast, simple and convenient sample preparation method introduced in 1990 [45], has gained increasing popularity for F&A analysis, especially as an alternative to DHS methods. It is specially suitable for qualitative and quantitative analysis of fragrance compounds released by odorous samples, only requiring exposure of a ¢ber to the HS above the sample for a suitable period of time, followed by direct TD on the heated injection port of a GC [46]. For these reasons, HS-SPME is being presently considered as the best available choice for sample preparation in fragrance and F&A analysis [2]. Because of the small dimensions of the sampling device and the simplicity and speed of the extraction procedure, HS-SPME is able to collect fragrances from live plants with minimum disturbance of the specimen, under both laboratory and ¢eld conditions. Vereen et al. [47] employed HS-SPME and GC-MS to study volatile compounds released by intact and mechanically damaged leaves of Fraser ¢rs (Abies fraseri) in the ¢eld. Branches of ¢r were enclosed in 100 mL Tedlar bags Figure 4. Profile of the emission of volatiles from a single Hydrosme rivieri flower (modified from Stra´nsky´ and Valterova´ [43]) after DHS-cryotrapping sampling and GC-FID analysis. Compounds: *=dimethyl disulphide; &=dimethyl trisulphide; ^=n-decane; *=n-undecane; &=n-dodecane; ~=n-tridecane. Trends and the air inside was extracted with 100 mm PDMS ¢bers over periods of up to 4 h. After 5 min of extraction, monoterpenes such as 3-carene predominate; after 3 h, the major component in the chromatograms is bornyl acetate, with minor amounts of heavier compounds (e.g., camphor and borneol). Two species (b-phellandrene and g-terpinene), detected in the fragrance exhaled from damaged leaves, were assigned as wound-response compounds. However, inadequate precision (RSD > 20%) and slow equilibration times for heavier compounds prevented quantitative application of HS-SPME to these samples. Zini et al. [48] employed HS-SPME and GC-ITMS (ion trap mass spectrometry) to assess the emission pro¢les from intact and mechanically damaged leaves of living Eucalyptus citriodora trees, using the sampling chamber shown in Fig. 5 and extractions with PDMS/DVB ¢bers performed every 30 min for continuous periods of between 8 h and 10 h. The main compounds identi¢ed were isoprene, citronellal, citronellol and b-caryophyllene. Di¡erent patterns of dependence between extracted amounts and leaf-enclosure times were observed; e.g., for rose oxide (cis-4-methyl-2-(2-methyl1-propenyl)-tetrahydropyran), a maximum in the area versus time curves appears after 300-400 min of leaf enclosure, while, for citronellal, the peak areas decayed exponentially from the start of the experiments. HS-SPME also has been extensively employed in recent applications related to food £avors. Roberts et al. [49] studied methodological and operational aspects of the application of HS-SPME to analysis of volatile £avor compounds. For co¡ee aroma, the best sensitivity overall was achieved using ¢bers coated with 65 mm PDMSDVB; this ¢ber was also found to be the most adequate for heavier polar odorants, such as vanillin. CarbowaxDVB ¢bers were found to be suitable for organic acids. Figure 5. Glass chamber for SPME sampling of volatiles emitted from single, live leaves (based on Zini et al. [48]). 1=Silanized glass cylindrical body; 2 – silanized glass lid; 3 – SPME sampling holes topped with silicone septa; 4 – SPME holder+fiber; 5 – DC power supply for microfan; 6 – microfan; 7 – Teflon support for microfan; 8 – Teflon tape seal. http://www.elsevier.com/locate/trac 165 Trends Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 Non-polar odorants were detectable at the mg/L level, while polar compounds could generally be detected at the mg/L level only; e.g., in the extraction of co¡ee aroma with PDMS-DVB ¢ber for 5 min followed by GC-MS analysis, they found 7.7 mg/L vanillin, 116 mg/L furaneol, 2.3 mg/L ethylguaiacol, 5.7 mg/L guaiacol, 40 mg/L 4-vinylguaiacol, 0.2 mg/Lb-damascenone, 0.4 mg/L 2,3-diethyl-5-methylpyrazine and 1 mg/L 2-ethyl-3,5-dimethylpyrazine. In a similar study conducted by Augusto et al. [50], different SPME ¢bers were compared for the characterization of the aromas of industrialized pulps of Brazilian tropical fruits ^ cupuassu (Theobroma grandi£orum, Spreng.), caja¤ (Spondias lutea, L.), siriguela (Spondias purpurea, L.) and graviola (Anona reticulata, L). The best e⁄ciencies overall for these aromas were obtained with Carboxen-PDMS ¢bers, and especially for low-molar mass compounds. Another area where HS-SPME is quickly becoming favored is in studies of the aroma of alcoholic beverages, essential in the assessment of the quality of such products and in optimizing their production. Sala et al. [51] developed a HS-SPME methodology for determination of the aroma-active heterocycles, 3-alkyl-2-methoxypyrazines, in musts from Cabernet Sauvignon and Merlot grapes. Levels as low as 0.1 ng/L of the analytes could be detected after extraction of the sample HS with PDMS-DVB ¢bers and GC-NPD (nitrogen-phosphorous detection) analysis of the extracts. Weber et al. [52] showed that chromatographic pro¢les of aroma obtained by HS-SPME and GC-MS can be useful in classifying samples of wine according to their production area, grape variety and vintage. Nonato et al. [53] compared HS-SPME and conventional LLE in the analysis of secondary aroma compounds (e.g., esters and heavy alcohols) in Brazilian sugar-cane spirits (cachacc¸ a). The precision in quantifying the target analytes with SPME was better than with the standard LLE procedure. Over and above these reports, a new HS-SPME approach for food £avor analysis was recently proposed by Pe¤re's et al. [54]. To characterize the ripening of Camembert cheese, the HS in contact with a 2 g sample contained in 10 mL £asks was exposed for 10 min to a 75 mm Carboxen-PDMS SPME ¢ber. In preliminary assays, the extracted materials were analyzed by GC-MS to identify odorant compounds. For subsequent experiments, the extracted compounds were thermally desorbed from the ¢ber and directly introduced into the ionization chamber of a MS without chromatographic separation. More than 60 compounds were identi¢ed in the HS-SPME extract by GC-MS; the Carboxen-PDMS ¢ber was found to be especially e⁄cient for collecting sulfur compounds (e.g., dimethyl sul¢de, methyl thioacetate), which are important impact odorants for this cheese variety. The HS-SPME-MS data obtained without chromatographic separation were statistically 166 http://www.elsevier.com/locate/trac processed and found to be valuable for qualitative chemometric classi¢cation of the samples. 4. Direct extraction procedures Methods involving application of LLE, SPE or SFE directly to the samples to isolate odorants are found occasionally in the recent literature. Although HS-based methods are usually more suitable for fragrance analysis, in some cases, direct extraction of the target analytes from the matrixes is necessary; e.g., species with high odor impact and in extremely reduced concentrations can be important to some F&A, but may not be detected using HS methodologies if their volatility is not high enough to provide a suitable concentration in the sample HS [1]. 4.1. LLE Despite its simplicity, the modern tendency is to replace LLE by other techniques, because high-purity solvents are required for trace analysis and because of the need to reduce the environmental and health risks associated with their manipulation. Also, in the case of biological F&A, LLE cannot be applied to live samples. However, there are several contemporary reports on use of this technique for F&A analysis, especially for collection of preliminary data [55]. Lo¤pez and Go¤mez [56] addressed some operational parameters on the application of LLE to extract aroma compounds from wines. Several solvents ^ diethyl ether, n-pentane, Freon-11, n-hexane, 1:1 ether/pentane, 1:1 ether/hexane and dichloromethane ^ were compared for extractions of odoriferous terpenic substances from ‘‘arti¢cial wine’’ (12% v/v ethanol in water). It was concluded that dichloromethane and 1:1 ether/pentane are the best solvents for extracting these analytes from wine. In a related study, continuous LLE was employed by Rocha et al. [57] to extract free and releasable odor volatiles from Portuguese Bairrada white grapes employed in the local wine industry. A volume of 250 mL of raw or enzimatically treated must was continuously extracted for 25 h with 75 mL of dichloromethane at 50 C. The organic phase was dried and concentrated by vacuum distillation prior to the GC-MS analysis. Chemometric analysis of the data allowed differentiation of several grape varieties and of the enzyme treatment studied. Conventional batch LLE is also still employed occasionally in F&A analysis; recent examples include the identi¢cation of odorants in sugar-beet juice [58]. 4.2. SPE SPE can be directly applied to isolate odorants from liquid or lique¢able odoriferous samples, such as Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 beverages, fruit pulps and tissues. A typical contemporary application of SPE to aroma analysis was presented by Wada and Shibamoto [59], who studied the direct extraction of odorants from Chablis red wine using Porapak Q columns. Samples spiked with selected aroma analytes (2-methyl-1-propanol, 3-methyl-1butanol, 2-phenylethanol, ethyl hexanoate, ethyl octanoate, diethyl butanedioate and hexanoic, octanoic and decanoic acids) were passed through 30-cm glass columns packed with the sorbent. Diethyl ether, dichloromethane and pentane were evaluated to desorb the extracted materials, which were analyzed by GC-FID and GC-MS. Dichloromethane was found to be the best desorbing solvent, recovering up to (103.0 3.7) % of the spiked analytes. Analysis of unspiked samples revealed 67 volatile odor compounds in this variety of red wine. SPE has also been applied to characterize butter aroma. Adahchour et al. [60] assessed the use of SPE cartridges packed with di¡erent sorbents ^ C18, C8, NH2 and CN-bonded silica, as well as SDB-1 (PS-DVB copolymer) ^ to extract aroma compounds from this material. Butter samples were melted and the aqueous fraction, after separation of the fat by centrifugation, was passed through the SPE cartridges. The cartridges were eluted by methyl acetate, which was dried and analyzed by GC-MS. Best overall recoveries were obtained with SDB1 cartridges (average recovery ca. 80%); detection limits for impact odorants, such as vanillin and diacetyl, were in the low-pg range. For samples such as fruit pulps, SPE can be combined with isolation techniques, such as distillation, as exempli¢ed by Boulanger et al. [61]. After distillation of cupuassu pulp and extraction with XAD-2 resin, the main odorants were identi¢ed as linalool, a-terpineol, 2-phenylethanol, myrcene, limonene, ethyl 2-methylbutanoate, ethyl hexanoate, butyl butanoate, 2,6-dimethyl-oct-7-en-2,6-diol, (E)- and (Z)-2,6-dimethyl-octa-2,7-dien-1,6-diol and methoxy2,5-dimethyl-3(2H)-furanone. Stan¢ll and Ashley [62] presented an interesting application of SPE to analysis of £avor-related alkylbenzenes in cigarette smoke. Tobacco from disassembled cigarettes was ‘‘smoked’’ in a smoking machine and the particulate material produced collected with ¢lter pads. The collected particulate was then suspended in hexane and the target analytes extracted from these solutions with CN-silica SPE cartridges. After elution with a hexane/toluene/tetrahydrofuran mixture and concentration of the extract by vacuum evaporation, alkylbenzene odorants were detected and quanti¢ed by GC-MS. Detection limits per cigarette ranged from 1.1 ng (methyleugenol) to 27.8 ng (eugenol). This method was tested to quantify aroma alkylbenzenes in commercial brands sold in the USA, and levels of target analytes up to several micrograms per cigarette were found. Trends 4.3. SFE Compared to the direct extraction methods discussed above, SFE presents some advantages for fragrance analysis. For example, since the critical point for CO 2 (the most popular and most convenient £uid for SFE) is 31.1 C at 7.38 MPa, extractions can be carried out under milder temperature conditions and without need of further aggressive procedures, such as distillation of excess solvent, as is common in LLE and SPE [63]. Volatile compounds from £owers, leaves and stems of guaca (Spilanthes americana, Mutis ^ a South American plant with medicinal and insecticide properties) ^ were isolated by Stashenko et al. [64] using both SFE and SDE, and analyzed by GC-MS, GC-FID and GC-NPD. SFE was performed with CO 2 at 40^45 C and 7.24^7.53 MPa, using up to 30 g of vegetable sample. Extraction time was 2 h and, after depressurization, the collected analytes were dissolved in 2 mL CH2Cl2. Compared to SDE, a larger number of compounds could be detected at levels above 100 ppb in the SFE extracts (e.g., 67 analytes with SFE and 43 analytes with SDE, for £oral extractions); the overall extraction yield for SFE was up to twice that for SDE. SFE was found to be especially e¡ective in isolating sesquiterpenes and nitrogenated compounds from these samples. There are also several SFE applications reported for food-£avor analysis. Leunissen et al. [65] developed a SFE-GC-MS method for fast analysis of roasted peanut aroma. For aroma isolation, 2.00^2.50 g of ground, roasted peanuts were extracted for 10 min with CO2 at 50 C and 9.6 MPa. The conditions were adjusted to minimize the extraction of greasy materials. The CO2 was passed through a silica trap at 5 C to collect the extracted materials; the trapped analytes were desorbed by dichloromethane and analyzed by GC-MS. The characteristic aroma compounds were found to be methylpyrrole, hexanol, hexanal, 2-furanmethanol, 2-furancarboxyaldehyde, benzeneacetaldehyde and several alkylpyrazines. Quanti¢cation limits for these analytes were estimated to be in the range between 80 ng/g (methylpyrrole) and 0.4 ng/g (ethyl- and 2,3,5-trimethylpyrazine). Supercritical CO2 is a poor extractor for polar substances; therefore, for such analytes, addition of modi¢ers or use of other £uids, such as supercritical water, is advisable [2]. Kubatova et al. [66] also compared water^SFE and CO2^SFE with hydrodistillation for the separation of aromatic essential oils from savory (Satureja hortensis) and peppermint (Mentha piperita). Extraction of savory with supercritical water for 40 min removed almost 100% more thymol and carvacrol and 150% more borneol and linalool than hydrodistillation. For CO2 extractions, yields of borneol and linalool were similar than those for hydrodistillation; however, for thymol and carvacrol the e⁄ciency of CO2^SFE was only half that of http://www.elsevier.com/locate/trac 167 Trends Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 hydrodistillation. Water-SFE was found to be highly selective for polar oxygenated £avor compounds, when compared with CO2^SFE or hydrodistillation. 5. Conclusion As in any analytical procedure, a suitable choice of sample-preparation technique is essential for accurate, reliable characterization of the chemical composition of F&A. However, because of the peculiarities of these samples ^ especially the common presence of thermally or chemically labile analytes in trace or ultratrace concentrations ^ selection of the analyte-isolation and pre-concentration technique, as well as careful optimization of the corresponding operational parameters, are of paramount importance. Some of the general techniques discussed in the previous sections ^ namely distillation procedures, LLE and SPE ^ are still employed occasionally for F&A analysis. However, for more critical analysis at least, the present tendency is to replace them with methodologies that are simultaneously less aggressive to analytes and capable of dealing with ultra-low concentrations of analytes in samples. Most of the state-of-the-art, contemporary applications in F&A analytical chemistry are focused on several variations of DHS sampling or, more recently, on HS-SPME. Despite the fact that these techniques have drawbacks (e.g., expensive hardware for DHS-ATD, limitations on the extracted masses for HS-SPME), for nearly all cases in F&A analytical chemistry, either DHS or HS-SPME can obtain dependable results. Perhaps the next frontier to be breached in this area will be ¢eld sampling and analysis. The great majority of analytical work on plant fragrances is still performed, totally or in part, in the laboratory and/or under invitro conditions (i.e., using detached parts of the plants as the source of fragrance). Nonetheless, adequate, realistic characterization of the composition of organic volatile emissions from plants can be made only in their natural habitats, and with minimal biological, chemical and physical disturbance. There is therefore a need for portable or semi-portable instruments (such as GC-MS) and sampling procedures based on small, rugged hardware. For sample preparation, SPME seems to be the technique that most closely ¢ts these demands. References [1] N. Nuener-Jehle, F. Etzweiler, in: P.M. Mu«ller, D. Lamparsky (Editors), Perfumes: Art, Science and Technology, Blackie Academic & Professional, London, UK, 1991, p. 153. [2] A. Sides, K. Robards, S. Helliwell, Trends Anal. Chem. 19 (2000) 322. 168 http://www.elsevier.com/locate/trac [3] G.V. Civille, J. Food Qual. 14 (1991) 1. [4] J.W. Arnold, S.D. Senter, J. Sci. Food Agric. 78 (1998) 343. [5] J. Pejtersen, H. Brohus, C.E. Hyldgaard, J.B. Nielsen, O. Valbjorn, P. Hauschildt, S.K. Kjaergaard, P. Wolko¡, Indoor Air 11 (2001) 10. [6] J.R. Miner, J. Animal Sci. 77 (1999) 440. [7] L.Y. Chen, P.T. Jeng, M.W. Chang, S.H. Yen, Environ. Sci. Technol. 34 (2000) 1166. [8] J. Noblet, L. Schweitzer, E. Ibrahim, K.D. Stolzenbach, L. Zhou, I.H. Su¡et, Water Sci. Technol. 40 (1999) 185. [9] R. Kaiser, in: P.M. Mu«ller, D. Lamparsky (Editors), op.cit., p. 213. [10] N. Dudareva, E. Pichersky, Plant Physiol. 122 (2000) 627. [11] G.R. Takeoka, R.G. Buttery, L.C. Ling, R.Y. Wong, L.T. Dao, R.H. Edwards, J.J. Berrios, Lebensm.-Wiss.-Technol. 31 (1998) 443. [12] A.M. Dietrich, S. Mirlohi, W.F. Costa, J.P. Dodd, R. Sauer, M. Homan, J. Schultz, Water Sci. Technol. 40 (1999) 45. [13] R.G. Buttery, L.C. Ling, J. Agric. Food Chem. 46 (1998) 2764. [14] Y. Kotseridis, R. Baumes, J. Agric. Food Chem. 48 (2000) 400. [15] A. Bayrak, A. Akgu«l, J. Sci. Food Agric. 64 (1994) 441. [16] J. Takabayashi, S. Takahashi, M. Dicke, M.A. Posthumus, J. Chem. Ecol. 21 (1995) 273. [17] J.B.F. Gervlieet, M.A. Posthumus, L.E.M. Vet, M. Dicke, J. Chem. Ecol. 23 (1997) 2395. [18] J. Kesselmeier, M. Staudt, J. Atmos. Chem. 33 (1999) 23. [19] J.P.F.G. Helsper, J.A. Davies, H.J. Boumeester, A.F. Krol, M.H. van Kampen, Planta 207 (1998) 88. [20] J. Auger, S. Rousset, E. Thibout, B. Jaillais, J. Chromatogr. A 819 (1998) 45. [21] P. Pollien, L.B. Fay, M. Baumgartner, A. Chaintreau, Anal. Chem. 71 (1999) 5391. [22] W. Grosch, Trends Food Sci. Technol. 4 (1993) 68. [23] R.I. Stefan, J.F. van Staden, H.Y. Aboul-Eneim, Crit. Rev. Anal. Chem. 29 (1999) 133. [24] Y. Saritas, M.M. Sonwa, H. Iznaguen, W.A. Ko«nig, H. Muhle, R. Mues, Phytochem. 57 (2001) 443. [25] W.S. Schlotzhauer, S.D. Pair, R.J. Horvat, J. Agric. Food Chem. 44 (1996) 206. [26] P. Perpe'te, S. Collin, J. Agric. Food Chem. 47 (1999) 2374. [27] P.A. Tarantilis, M.G. Polissiou, J. Agric. Food Chem. 45 (1997) 459. [28] B. Siegmund, E. Leitner, I. Mayer, W. Pfannhauser, P. Farkas, J. Sadecka, M. Kovac, Food Res. Technol. 205 (1997) 73. [29] L. Moio, P. Piombino, F. Addeo, J. Dairy Res. 67 (2000) 273. [30] P. Bouchilloux, P. Darriet, R. Henry, V. Lavigne-Cruege, D. Dubourdieu, J. Agric. Food Chem. 46 (1998) 3095. [31] H. Steinhart, A. Stephan, M. Bu«cking, J. High Resolut. Chromatogr. 23 (2000) 489. [32] R.J. Stevenson, X.D. Chen, O.E. Mills, Food Res. Int. 29 (1996) 265. [33] J.T. Knudsen, L. Tollsten, L.G. Bergstro«m, Phytochem. 33 (1993) 253. [34] P. Clarkson, M. Cooke, Anal. Chim. Acta 335 (1996) 253. [35] S.A. Rankin, F.W. Bodyfelt, J. Food Sci. 60 (1995) 1205. [36] E. Valero, E. Miranda, J. Sanz, I. Martinez-Castro, Chromatographia 44 (1997) 59. [37] J. Vercammen, P. Sandra, E. Baltussen, T. Sandra, F. David, J. High Resolut. Chromatogr. 23 (2000) 547. [38] R.A. Raguso, O. Pellmyr, Oikos 81 (1998) 238. [39] N.G. Agelopoulos, A.M. Hooper, S.P. Maniar, J.A. Pickett, L.J. Wadhams, J. Chem. Ecol. 25 (1999) 1411. [40] L. Jiang, K. Kubota, J. Agric. Food Chem. 49 (2001) 1353. [41] D. Canac-Arteaga, C. Viallon, J.L. Berdague, Analusis 28 (2000) 550. [42] J. Gawlowski, T. Gierczak, A. Jezo, J. Niedzielski, Analyst (Cambridge, UK) 124 (1999) 1553. Trends in Analytical Chemistry, Vol. 22, No. 3, 2003 [43] K. Stra¤nsky¤, I. Valterova¤, Phytochem. 52 (1999) 1387. [44] B. Kolbe, G. Zwick, M. Auer, J. High Resolut. Chromatogr. 19 (1996) 37. [45] J. Pawliszyn (Editor), Applications of Solid Phase Microextraction, RSC, Cambridge, UK, 1999. [46] A.J. Matich, in: J. Pawliszyn (Editor), op.cit., p. 349. [47] D.A. Vereen, J.P. McCall, D.J. Butcher, Microchem. J. 65 (2000) 269. [48] C.A. Zini, F. Augusto, E. Christensen, B.P. Smith, E.B. Carama‹o, J. Pawliszyn, Anal. Chem. 73 (2001) 4729. [49] D.D. Roberts, P. Pollien, C. Milo, J. Agric. Food Chem. 48 (2000) 2430. [50] F. Augusto, A.L.P. Valente, E.S. Tada, S.R. Rivellino, J. Chromatogr. A 873 (2000) 117. [51] C. Sala, M. Mestres, M.P. Marti, O. Busto, J. Guasch, J. Chromatogr. A 880 (2000) 93. [52] J. Weber, M. Beeg, C. Bartzsch, K.H. Feller, D.D. Garcia, M. Reichenbacher, K. Danzer, J. High Resolut. Chromatogr. 22 (1999) 322. [53] E.A. Nonato, F. Carazza, F.C. Silva, C.R. Carvalho, Z.L. Cardeal, J. Agric. Food Chem. 49 (2001) 3533. Trends [54] C. Pe¤re's, C. Viallon, J.L. Berdague¤, Anal. Chem. 73 (2001) 1030. [55] M. Mestres, O. Busto, J. Guasch, J. Chromatogr. A 881 (2000) 569. [56] E.F. Lo¤pez, E.F. Go¤mez, Chromatographia 52 (2000) 798. [57] S. Rocha, P. Coutinho, A. Barros, M.A. Coimbra, I. Delgadillo, A.D. Cardoso, J. Agric. Food Chem. 48 (2000) 4802. [58] P. Pihlsga‡rd, M. Larsson, A. Leufve¤n, H. Lingnert, J. Agric. Food Chem. 48 (2000) 4844. [59] K. Wada, T. Shibamoto, J. Agric. Food Chem. 45 (1997) 4362. [60] M. Adahchour, R.J.J. Vreuls, A. van der Heijden, U.A.Th. Brinkman, J. Chromatogr. A 844 (1999) 295. [61] R. Boulanger, J. Crouzet, Flav. Frag. J 15 (2000) 251. [62] S.B. Stan¢ll, D.L. Ashley, J. Agric. Food Chem. 48 (2000) 1298. [63] Q. Lang, C.M. Wai, Talanta 53 (2001) 771. [64] E.E. Stashenko, M.A. Puertas, M.Y. Combariza, J. Chromatogr. A 752 (1996) 223. [65] M. Leunissen, V.J. Davidson, Y. Kakuda, J. Agric. Food Chem. 44 (1996) 2694. [66] A. Kubatova, A.J.M. Lagadec, D.J. Miller, S.B. Hawthorne, Flav. Fragr. J. 16 (2001) 64. http://www.elsevier.com/locate/trac 169