Determination of mono- and sesquiterpenes in water samples by membrane inlet mass spectrometry and static headspace gas chromatography

Talanta 49 (1999) 179 – 188

Determination of mono- and sesquiterpenes in water samples by membrane inlet mass spectrometry and static headspace gas chromatography Marja Ojala a,b,*, Raimo A. Ketola a, Timo Mansikka a, Tapio Kotiaho a, Risto Kostiainen c b

a VTT, Chemical Technology, P.O. BOX 1401, FIN-02044 VTT, Finland Uni6ersity of Helsinki, Laboratory of Analytical Chemistry, P.O. Box 55, FIN-00014 Uni6ersity of Helsinki, Finland c Uni6ersity of Helsinki, Department of Pharmacy, Di6ision of Pharmaceutical Chemistry, P.O. Box 56, FIN-00014 Uni6ersity of Helsinki, Finland

Received 4 June 1998; received in revised form 12 November 1998; accepted 16 November 1998

Abstract A membrane inlet mass spectrometric (MIMS) method is presented and compared with a static headspace gas chromatographic method (HSGC) for the determination of terpenes in water. The MIMS method provides a very simple and fast analysis of terpenes in water, detection limits being relatively low, from 0.2 mg l − 1 for monoterpenes to 2 mg l − 1 for geraniol. The analysis of terpenes by the HSGC (equipped with flame ionization detector, FID) method is more time-consuming and the detection limits (2 mg l − 1 for monoterpenes to 100 mg l − 1 for geraniol) are higher than with MIMS. However, the HSGC method has the advantage of determining individual mono- and sesquiterpene compounds, whereas MIMS provides only separation of different classes of terpenes. Both methods were applied to the analysis of mono- and sesquiterpenes in several condensation water samples of pulp and paper mills. © 1999 Elsevier Science B.V. All rights reserved. Keywords: Membrane inlet mass spectrometry; Headspace gas chromatography; Terpenes; Water samples

1. Introduction Mono- and sesquiterpenes are common constituents of volatile oils and several woods contain extractable terpenes, and they are found for ex* Corresponding author. Tel.: + 358-9-4565312; fax: + 3589-4567026. E-mail address: [email protected] (M. Ojala)

ample in the condensation waters of pulp and paper mills [1,2]. Monoterpenes exist as hydrocarbons or as oxygenated compounds such as aldehydes, alcohols, ketones, esters or ethers. These may be acyclic, monocyclic, dicyclic or tricyclic in structure [3]. The solubility of mono- and dicyclic terpene hydrocarbons in water is low (10–30 mg l − 1). However, monoterpenes containing the hydroxyl group have solubilities of 10–100-fold

0039-9140/99/$ - see front matter © 1999 Elsevier Science B.V. All rights reserved. PII: S 0 0 3 9 - 9 1 4 0 ( 9 8 ) 0 0 3 5 5 - 5

180

M. Ojala et al. / Talanta 49 (1999) 179–188

compared to terpene hydrocarbons [4]. Due to their low solubility in water, terpene hydrocarbons have relatively high Henry’s law constants [5,6] and therefore they partition predominantly into the atmosphere [7]. For this reason they have been analyzed more often in air [8 – 11] than in water. Ekman et al. [12] have evaluated wood extractives in pulp and water samples by gas chromatography – mass spectrometry. Using gas chromatography, the turpentine extracted from wood has been studied and 15 different terpenes were found [13]. Kimball et al. [14] have identified 22 terpenoids from wood using GC combined with mass selective detection. Volatile sesquiterpenes and sesquiterpenoids in environmental and geological samples have been analyzed by a capillary gas chromatograph combined with a mass spectrometer after extraction and fractionation [15]. Furthermore, gas chromatography or gas chromatography-mass spectrometry has been used in the analysis of terpenes in waste water samples [1,16,17]. Static headspace gas chromatographic methods have been widely used in analysis of terpenes from essential oils, plant products and in aroma constituents [18 – 20]. Furthermore, headspace methods have been used in quantitation of volatile constituents, for example, terpenes in juices [21 – 23]. Mathieu et al. [24] have used dynamic headspace to analyze volatile organic compounds in red coffee berries and they found many mono- and sesquiterpene compounds. However, according to our study only one article reporting the measurements of terpenes from water by static headspace method was found in the literature [25]. The detection limits obtained for terpene hydrocarbons were 5 mg l − 1, which are in good agreement with our results. Although GC-methods are widely used, they are time-consuming, so that faster and more sensitive methods are needed. In addition, conventional GC-techniques are not well suited for on-line measurement, and therefore there is a need to develop rapid and sensitive on-line methods for the analysis of contaminated water samples. Membrane inlet mass spectrometry (MIMS) meets all these requirements [26,27]. The analysis with MIMS involves a flow of the sample over a

sheet polydimethylsiloxane membrane which extracts non-polar, medium polar and low molecular weight organic compounds from the matrix. After the extraction step the organics pervaporate through the membrane into the ion source of a mass spectrometer. The total analysis takes only a few minutes. The MIMS method has previously been applied to the analysis of volatile organic compounds (VOCs) in environmental water and air samples [28–34] as well as in the monitoring of chemical and biochemical processes [35–37]. Membrane inlet mass spectrometry has also been applied for continuous on-line waste water monitoring for VOCs [38,39]. This study compares the membrane inlet mass spectrometric (MIMS) method for the analysis of terpenes in water with the static headspace gas chromatographic (HSGC) method.

2. Experimental

2.1. Reagents and samples The following reagents were obtained from Fluka: a-Pinene [80-56-8]; b-pinene [127-91-3]; limonene [138-86-3]; a-terpinene [99-86-5] D-3carene [13466-78-9]; myrcene [123-35-3]; camphene [79-92-5]; longifolene [475-20-7]; a-cedrene [11028-42-5]; linalool [78-70-6]; and geraniol [10624-1]. A stock solution (1 g l − 1) of each standard compound was prepared in methanol. The calibration standards (0.1 mg l − 1 –1 mg l − 1) were prepared by dilution with deionized water (Millipore, Milli-Q Plus). Authentic waste water samples from pulp and paper mills were used in comparison of the MIMS and HSGC methods. Many of the samples were unhomogenous due to their high terpene concentration and the low solubility of terpenes in water. Samples were diluted with deionized water before analysis as follows: Samples 10 and 11 were diluted by 1:250, sample 12 by 1:10 and sample 13 by 1:2 using deionized water for MIMS analysis. In addition, samples 10 and 12 were diluted by 1:10 and sample 11 by 1:50 for HSGC-analysis. All other samples were analyzed without dilution (Table 3).

M. Ojala et al. / Talanta 49 (1999) 179–188

181

Table 1 Standard terpene compounds and their electron impact mass spectra measured by MIMS Compound

MWa

Ions m/z (RA)b

Monoterpenes a-pinene Camphene b-Pinene Myrcene D-3-carene a-terpinene Limonene

136 136 136 136 136 136 136

93(100), 91(44), 92(37), 77(31), 79(25), 121(14), 105(13), 67(10), 53(10), 94(10), 136(6) 93(100), 121(59), 79(48), 91(41), 67(32), 77(30), 107(30), 94(18), 53(16), 95(16), 136(11) 93(100), 69(34), 91(32), 79(26), 77(26), 94(13), 121(13), 67(13), 92(12), 53(12), 136(6) 93(100), 69(76), 91(26), 79(20), 77(18), 53(15), 67(15), 92(11), 94(11), 53(15), 136(3) 93(100), 91(53), 77(38), 79(35), 92(29), 80(22), 121(20), 105(16), 136(12), 94(12) 121(100), 93(91), 136(61), 91(58), 77(36), 79(30), 119(29), 105(23), 107(12), 65(11) 68(100), 67(74), 93(65), 79(36), 53(28), 91(27), 94(25), 92(24), 77(23), 121(22), 136(14)

Monoterpene alcohols Linalool Geraniol

154 154

71(100), 55(84), 93(69), 69(50), 57(40), 56(40), 67(30), 80(27), 53(24), 121(19) 69(100), 93(24), 68(19), 67(18), 53(14), 91(13), 55(10), 79(10), 77(9), 123(9)

Sesquiterpenes Longifolene a-cedrene

204 204

161(100), 91(92), 105(76), 93(68), 107(67), 94(64), 79(63), 119(57), 133(54), 95(52), 204(26) 119(100), 93(40), 105(33), 91(26), 161(18), 69(18), 204(16), 77(15), 120(15), 121(14)

a b

MW is molecular weight, molecular ions are bolded. RA is relative abundance.

2.2. Mass spectrometric conditions Samples were analyzed using a quadrupole

Table 2 Detection limits (S/N= 3) of selected terpenes by static HSGC and MIMS Compound

Detection limit (mg l−1) HSGC

Monoterpenes a-pinene Camphene b-pinene Myrcene D-3-carene a-terpinene Limonene Monoterpene alcohols Linalool Geraniol Sesquiterpenes Longifolene a-cedrene

MIMS/SIM

3 3 3 2 2 2 2

0.2 0.5 0.2 0.2 0.5 0.5 0.5

30 100

0.5 2

5 5

2 0.5

mass spectrometer (Balzers QMG 421C, Balzers Aktiengesellschaft, Balzers, Liechtenstein) with a mass range 1–500 amu. This quadrupole instrument was equipped with an open cross-beam electron impact (70 eV) ion source. The data were collected either by scanning full mass spectrum (mass range m/z 50–210 amu), or by using a selected ion monitoring mode (SIM). The latter was used to determine detection limits. A sheet membrane inlet built at VTT Chemical Technology [40] was used. The membrane material was polydimethylsiloxane (Specialty Silicone Products, Ballston SPA, NY, USA) with a thickness of 110 mm and contact area of 28 mm2. During operation of the system, a water stream was continuously sucked through the membrane inlet via a peristaltic pump (Ismatec IPS4, Ismatec SA, Switzerland), with a typical flow rate of 10 ml min − 1, and with aliquots of sample solution (20–30 ml) injected into this stream. Using a heat exchanger, which was heated with a circulating water bath (Lauda M3, MGW, Germany), the water stream was heated to 70°C before the membrane inlet. Identification and quantitation of terpenes in water samples were done using a calculation pro-

182

M. Ojala et al. / Talanta 49 (1999) 179–188

Fig. 1. (a) Electron impact mass spectrum of a mixture of a-pinene, b-pinene and D-3-carene (about 50 mg l − 1 of each component). (b) Electron impact mass spectrum of a mixture of longifolene and a-cedrene (about 50 mg l − 1 of each component). (c) Electron impact mass spectrum of a mixture of linalool and geraniol (about 100 mg l − 1 of each component).

gram (Solver) designed at VTT Chemical Technology [41,42]. This calculation program uses a modified algorithm of the general deconvolution method, which assumes that the intensity of any mass-to-charge ratio (m/z) is a linear function of the concentration of the chemical compounds which contribute to that particular m/z.

2.3. Gas chromatographic conditions The system used in this experiment was a gas chromatograph (HP 5890 Series II, Hewlett Packard, Palo Alto, CA, USA) equipped with two FIDs and a headspace sampler (HP 7694). Two columns were used, a SPB-1 (30 m× 0.32 mm×

M. Ojala et al. / Talanta 49 (1999) 179–188

183

Fig. 1. (Continued)

1.0 mm, Supelco, Supelco Park, Bellefonte, PA 16823, USA) and a DB-1701 (30 m × 0.32 mm× 1.0 mm, J and W Scientific, Folsom CA, USA). The temperature program used was optimized for the separation of terpenes and it was 45°C, 5 10$C min − 1 min “ 210°C, 10 min. The temperatures of the injector and detectors were 220 and 250°C, respectively. In addition, the temperatures of sample vials and transferline between the headspace autosampler and the GC were 80 and 120°C, respectively.

3. Results and discussion The analytical performance characteristics of the MIMS and HSGC methods were determined using the following standard compounds; monoterpenes: a- and b-pinene, limonene, a-terpinene, D-3-carene, myrcene and camphene, sesquiterpenes: longifolene and a-cedrene; and alcoholic derivatives of acyclic terpenes: linalool, and geraniol. Electron impact (EI) mass spectra were measured for all these compounds using MIMS. Ten ions with the highest abundance and the molecular ions are presented in Table 1 and the mass spectra of the quantitation standard mixtures (a)

a monoterpene mixture consisting of a- and bpinene and D-3-carene (about 50 mg l − 1 of each component); (b) a sesquiterpene mixture, longifolene and a-cedrene (about 50 mg l − 1 of each component), and (c) terpene alcohol mixture, linalool and geraniol (about 100 mg l − 1 of each component) are presented in Fig. 1. All the measured EI mass spectra (except linalool) agreed well with the EI mass spectra published in a reference mass spectral library [43]. The molecular ions m/z 136 for monoterpene hydrocarbons and m/z 204 for sesquiterpene hydrocarbons were recorded, but the molecular ion m/z 154 for linalool and geraniol was not observed. The EI spectra of monoterpenes were rather similar (Table 1), the base peak for nearly all monoterpenes was m/z 93 and the other fragment ions were nearly identical. EI spectra of sesquiterpenes, longifolene and cedrene, had many common fragment ions and the base peaks were m/z 161 and m/z 119, respectively. Due to the similarities in their mass spectra the identification of individual terpenes is very difficult without chromatographic separation. However, the mass spectra of mono- and sesquiterpenes are clearly different. Due to these differences the quantitation could be carried out by the MIMS method as monoterpene and sesquiterpene classes and it was performed using

M. Ojala et al. / Talanta 49 (1999) 179–188

184

Table 3 The terpene concentrations (mg l−1) of analyzed authentic water samples using the HSGC and MIMS methods Sample

1 2 3 4 5 6 7 8 9 10 11 12 13

Monoterpenes

Sesquiterpenes

Total

HSGC

MIMS

HSGC

MIMS

HSGC

MIMS

48 32 61 36 530 1500 730 1900 860 23 000 53 000 12 000 2700

160 98 110 74 410 1500 440 1800 610 37 000 49 000 8500 1100

100 72 110 67 370 580 260 540 330 37 00 6100 2500 670

69 62 70 49 370 650 320 750 450 5700 4800 2000 470

150 100 170 100 900 2100 1000 2400 1200 27 000 59 000 15 000 3400

230 160 180 120 780 2100 760 2600 1100 42 000 53 000 11 000 1600

the whole mass spectrum of a standard mixture (Fig. 1), in the calculations. Detection limits with MIMS were measured using selected ion monitoring (SIM), and the base peak of each individual compound was used as the ion in measurement. The detection limits varied from 0.2 to 2 mg l − 1 (Table 2). It is worth of notice that the detection limits measured by MIMS using the whole spectrum were about one order of magnitude higher compared to those measured by MIMS in a selected ion monitoring mode. The SIM results are reported here, because it is the typical method used for detection limit measurements in connection with MIMS and because the SIM method can be used to quantitate low levels of terpenes. Detection limits measured with HSGC were relatively low (2 – 5 mg l − 1), except for the more polar compounds linalool and geraniol, which had detection limits of 30 and 100 mg l − 1, respectively (Table 2). The relatively high detection limit for linalool could be expected to be due to its low Henry’s law coefficient, 0.002 kPa m3 mol − 1 [6]. The differences in detection limits can also be seen in the chromatogram presented in Fig. 2, which shows a gas chromatogram of a standard terpene mixture with about 20 mg l − 1 of each terpene compound. On the basis of this chromatogram it is easy to see that the analysis time with the HSGC method is relatively long,

about 30 min. Whereas, with the MIMS method the analysis time was rather short, about five min, as in our earlier studies [31–34]. The repeatability and linear dynamic ranges of the MIMS method were tested using six compounds: a-pinene; a-terpinene; limonene; myrcene; longifolene; and geraniol. The measurements were done using at least four different concentrations (from 0.1 to 300 mg l − 1) and triplicate analysis of every single concentration. Because of the poor solubility of terpenes in water, their linear dynamic ranges were measured only up to concentrations of about 300 mg l − 1 with both the methods. The repeatability of the MIMS method was good [relative standard deviations (RSD)= 2.6–18%]. The repeatability of the HSGC technique was also good, the RSD being lower than 5% for every terpene compound (not determinated for geraniol). The RSD values for the HSGC method were determined using four different concentrations of standard compounds and triplicate analysis of each. The linear dynamic ranges of the MIMS method were three orders of magnitude for all compounds, except for longifolene and geraniol, which had ranges of two orders of magnitude (correlation coefficients varied from 0.992 for geraniol to 1.000 for a-terpinene). The linear dynamic ranges by the HSGC method were only two decades and the correlation coefficients

M. Ojala et al. / Talanta 49 (1999) 179–188

185

Fig. 2. A gas chromatogram of a standard terpene mixture (about 20 mg l − 1 of each compound). Due to its high detection limit geraniol can not be seen in the chromatogram. It elutes between linalool and longifolene. Results are obtained using SPB-1 column. DB-1701 column could not separate longifolene and cedrene. 1. a-pinene, 2. camphene, 3. b-pinene, 4. myrcene, 5. D-3-carene, 6. a-terpinene, 7. limonene, 8. linalool, 9. longifolene, 10. a-cedrene.

varied from 0.995 for a-pinene to 1.000 for a-cedrene (measurement range 3 – 300 mg l − 1). Authentic water samples were also analyzed for terpenes using the MIMS method and the results were compared with those obtained by the HSGC method. As an example of the measured results Fig. 3a shows the mass spectrum for the authentic water sample 13 (Table 3) and Fig. 3b shows the headspace gas chromatogram of the same sample. The same major ions as in the standard spectra 1a and 1b can be seen in this spectrum. Ion ratio between m/z 119 and 121 indicates that the concentration of monoterpenes is higher than that of sesquiterpenes as can also be seen in the gas chromatogram (Fig. 3b). According to the mass spectra and the gas chromatogram of the sample (Fig. 3), the concentration of linalool and geraniol was very low and they have not been quantitated. The gas chromatogram (Fig. 3b) shows good sep-

aration between terpene compounds and furthermore, several unknown peaks can be seen. Main compounds in the samples were a- and b-pinene and D-3-carene, in addition camphene, myrcene, limonene, longifolene and a-cedrene were found. The quantitation of the mono- and sesquiterpene content in the authentic water samples by the MIMS method was performed using spectra 1a and 1b as standards for the Solver quantitation and identification program [41,42]. The mass spectra (Fig. 1c) for terpene alcohol mixture was not used since concentrations of these were very low as proved by the measured chromatogram and mass spectra. The Solver program capability for quantitative analysis of terpenes in water samples was first thoroughly evaluated. All the eleven terpene compounds (Table 1) were used in testing. The first tests were made with solutions of only three

186

M. Ojala et al. / Talanta 49 (1999) 179–188

Fig. 3. (a) Electron impact mass spectrum of the authentic water sample 13 (Table 3, diluted by 12 before analysis). (b) The gas chromatogram of the authentic water sample 13 (Table 3, diluted by 12 before analysis). Results are obtained using SPB-1 column. DB-1701 column could not separate longifolene and cedrene.

monoterpenes i.e. a- and b-pinene and D-3-carene. Results of these measurements showed that the individual terpenes could be separated well from

three component mixtures at concentration range 20–100 mg l − 1. The relative standard deviations of the concentrations varied in the range 8–22%

M. Ojala et al. / Talanta 49 (1999) 179–188

187

Table 4 Results obtained in testing the quantitative capabilities of the Solver program Compound

Concentration (mg l−1)

Analyzed concentration (mg l−1)

Mean

Monoterpenes a

213 425 850

171 401 806

188 435 949

202 479 963

187 439 906

Sesquiterpenes b

63.9 127 256

32.4 107 206

52.5 141 290

61.7 175 330

48.9 141 275

31 24 23

Terpene alcohols c

75 150 300

59.7 137 252

73.0 156 294

81.9 165 311

71.5 153 286

16 9.1 11

RSD (%) 8.2 8.9 9.6

Error (%) −12 3.2 6.6 −24 11 7.5 −4.6 1.8 −4.4

a-pinene, camphene, b-pinene, myrcene, D-3-carene, a-terpinene and limonene. Longifolene and cedrene. c Linalool and geraniol. a

b

and the error of the concentration determination varied in the range 2 – 12%. However, when individual terpenes were quantitated from eleven component mixtures the result of the Solver program was poor, errors of the concentration determination varied in the range −16 to + 100%. The reason for this is believed to be the fact that the terpene compounds used in testing have very similar EI mass spectra (Table 1). The next step was to test if different classes of terpenes can be reliably identified and quantitated by the Solver program. For these measurements mixtures of aand b-pinene and D-3-carene were used as monoterpene hydrocarbon standards, mixtures of longifolene and cedrene were used as sesquiterpene hydrocarbon standards and mixtures of linalool and geraniol were used as terpene alcohol standards. The reference library, which the Solver program is utilizing in the calculations, contained only these three mixtures [42]. The results of these measurements are presented in Table 4. From Table 4 it can be noticed that the determination of monoterpene concentrations are repeatable and accurate and that the determination of terpene alcohols was rather good. Whereas, the determination of sesquiterpene concentrations was more inaccurate (RSD 20 – 30% and maximum relative error − 24%). These results show that quantitation of different terpene classes with the Solver program works well and that the program can be utilized in the analysis of unknown samples. The .

results presented in Table 3 for the authentic water samples are mean values of triplicate measurements, except for samples 1–4 and 10 only duplicate (RSD varied from 2 to 25%). The total amount of terpenes in the authentic water samples (Table 3) were estimated by HSGC method using an external standard method (four different concentrations from 20 to 300 mg l − 1) and all the eleven individual compounds (Table 2) as standards. The amount of sesquiterpenes was calculated as the sum of longifolene, a-cedrene and the amount of unknown compounds eluted after limonene (number 7 in Fig. 3b). The concentration of these compounds were calculated by comparing their peak areas to the peak area of a-pinene. The HSGC results presented in Table 3 are means of duplicate determinations. The total amount of mono- and sesquiterpenes with both methods are in the same order. The differences between results with both methods varied from 0 to 60%, the average differences was 26%. The main reason for the differences between the results obtained by the different analytical methods is believed to be the high concentration of terpenes in the samples, which caused some unhomogeinity and the need to dilute samples before analysis. Furthermore, some non-terpenic compounds could have been included in HSGC quantitation because the identification of unknown compounds was not possible using GC-FID.

188

M. Ojala et al. / Talanta 49 (1999) 179–188

In conclusion, quantitation by MIMS is fast (only a few minutes) and the results are reliable and respectable addition, MIMS is believed to be suitable for on-line terpene analysis, for example for on-line industrial waste water monitoring. The HSGC method is more time-consuming but better selectivity can be obtained due to good separation of the individual compounds.

References [1] D. Wilson, B. Hrutfiord, Pulp Pap. Can. 76 (1975) 91. [2] T.P. Churilova. G.D. Khamuev, Gig. Sanit. (1989) 78. [3] L.F. Fieser, M. Fieser, Organic Chemistry, 3rd ed., Maruzen, Tokyo, Japan, 1956, p. 935. [4] J.D. Weidenhamer, F.A. Macias, N.H. Fischer, G.B. Williamson, J. Chem. Ecol. 19 (1993) 1799. [5] J. Hine, P.K. Mookerjee, J. Org. Chem. 40 (1975) 292. [6] J. Li, E.M. Perdue, in: Preprints of papers presented at the Two Hundredth and Ninth ACS National Meeting Anachem, CA, 35 (1995) 134. [7] D. Mackay, W.Y. Shiu, J. Phys. Chem. Ref. Data 10 (1981) 1175. [8] T. Hoffmann, Fresenius J. Anal. Chem. 351 (1995) 41. [9] D.D. Riemer, P.J. Milne, C.T. Farmer, R.G. Zika, Chemosphere 28 (1994) 837. [10] K. Eriksson, J.O. Levin, M. Rhen, R. Lindahl, Analyst 119 (1994) 85. [11] F. Juettner, J. Chromatogr. 442 (1988) 157. [12] R. Ekman, B. Holmbom, Nordic Pulp Paper Res. J. 4 (1989) 16. [13] S.V. Kossuth, J.W. Munson, Tappi 64 (1981) 174. [14] B.A. Kimball, R.K. Craver, J.J. Johnston, D.L. Nolte, J. High Resol. Chromatogr. 18 (1995) 221. [15] V.O. Elias, B.R.T. Simoneit, J.N. Cardoso, J. High Resol. Chromatogr. 20 (1997) 305. [16] D.-K. Nguyen, A. Bruchet, P. Arpino, J. High Resol. Chromatogr. 17 (1994) 153. [17] L.V. Kosyokova, S.A. Konokova, Gidroliz. Lesokhim. Prom-st. (1978) 9. [18] T.L. Potter, J. Essent. Oil. Res. 7 (1995) 347. [19] G. Buchbauer, L. Jirovetz, M. Wasicky, A. Nikiforoy, Fresenius J. Anal. Chem. 347 (1993) 465. [20] H. Vuorela, J. Pohjola, C. Krause, R. Hiltunen, Flavour Fragrance J. 4 (1989) 117. [21] M.G. Moshonas, P.E. Shaw, J. Agric. Food Chem. 45 (1997) 3968. [22] J.S. Paik, A.C. Venables, J. Chromatogr. 540 (1991) 456.

[23] M.G. Moshonas, P.E. Shaw, Food Sci. Technol. 25 (1992) 236. [24] F. Mathieu, C. Malosse, A.-H. Cain, B. Frerot, J. High Resol. Chromatogr. 19 (1996) 298. [25] L.V. Kosyukova, Y.L. Belyaeva, Zh. Anal. Khim. 33 (1978) 794. [26] F.R. Lauritsen, T. Kotiaho, Rev. Anal. Chem. 15 (1996) 237. [27] T. Kotiaho, F.R. Lauritsen, T.K. Choudhury, R.G.Cooks, G.T. Tsao, Anal. Chem. 63 (1991) 875A. [28] R.G. Cooks, T.Kotiaho, in: J.J. Breen, M.J. Dellarco (Eds.), ACS Symposium Series 508, Pollution Prevention in Industrial Processes: The role of Process Analytical Chemistry, American Chemical Society, 1992, ISBN 08412-2478-1, pp. 126 – 154. [29] P.S.H. Wong, R.G. Cooks, M.E. Cisper, P.H. Hemberger, Env. Sci. Technol. 29 (1995) 215A. [30] T. Kotiaho, R.A. Ketola, M. Ojala, T. Mansikka, R. Kostiainen, Am. Environ. Lab 9 (1997) 19. [31] M. Ojala, R.A. Ketola, V. Virkki, H. Sorsa, T. Kotiaho, Talanta 44 (1997) 1253. [32] M. Ojala, R.A. Ketola, T. Mansikka, T. Kotiaho, R. Kostiainen, J. High Resol. Chromatogr. 20 (1997) 165. [33] R.A. Ketola, V.T. Virkki, M. Ojala, V. Komppa, T. Kotiaho, Talanta 44 (1997) 373. [34] V. Virkki, R. Ketola, M. Ojala, T. Kotiaho, V. Komppa, A. Grove, S. Facchetti, Anal. Chem. 67 (1995) 1421. [35] N. Srinivasan, R.C. Johnson, N. Kasthurikrishnan, P. Wong, R.G. Cooks, Anal. Chim. Acta 350 (1997) 257. [36] H. Degn, J. Microbiol. Methods 15 (1992) 185. [37] F.R. Lauritsen, D. Lloyd, in: C. Fenselau (Ed.) ACS Symposium Series 541, Mass Spectrometry for Characterization of Microorganisms, American Chemical Society, 1992, ISBN 0-8412-2478-1, pp. 126 – 154. [38] T. Kotiaho, R. Kostiainen, R.A. Ketola, et al., Process Control Qual. 11 (1998) 71. [39] M.A. LaPack, J.C. Tou, C.G. Enke, Anal. Chem. 63 (1991) 1631. [40] R.A. Ketola, M. Ojala, H. Sorsa, T. Kotiaho, R.K. Kostiainen, Anal. Chim. Acta 349 (1997) 359. [41] V. Virkki, R. Ketola, J. Juuja¨rvi, E. Oja, H. Sorsa, M. Ojala, T. Kotiaho, Development of calculation program for the analysis of multicomponent spectra, in: Presentation at XXIX Conference of the Finnish Physical Society, Jyva¨skyla¨, Finland, March 1995. [42] R.A. Ketola, M. Ojala, V. Komppa, T. Kotiaho, J. Juuja¨rvi, J. Heikkonen, Rapid Commun. Mass Spectrom. (submitted). [43] EPA/NIH Mass Spectral Data Base, U.S. Government Printing Office, Washington, 1978.