7

E. T. Contis et al. (Editors) Food Flavors: Formation, Analysis and Packaging Influences © 1998 Elsevier Science B.V. All rights reserved

69

Aroma analysis of coffee brew by gas chromatography-olfactometry K. D. Deibler, T. E. Acree, E. H. Lavin Depart, of Food Science & Technology, Cornell University, Geneva, NY 14456 Abstract During the study of coffee flavor, the processes of brewing, extraction and sampling cause losses of the aroma compounds present in coffee grounds. In this study, coffees from two brewing methods were extracted, serially diluted and each dilution sniffed twice using the gas chromatography-olfactometry (GCO) technique called CharmAnalysis . Among the hundreds of volatile chemicals present, 18 of the thirty most potent odorants were identified by comparing the mass spectra, odor activity and Kovat's retention indices with those of authentic standards. Our studies have verified the presence of previously identified aroma compounds among the most potent odorants in coffee and show the differences between the two brewing methods tested.

1. INTRODUCTION According to legend, coffee was discovered by an Arab goat herder named Kaidi. He noticed that his goats became frisky and danced around the fields after chewing on the berries from coffee bushes. After watching this, an abbot gave some of the berries to neighboring monks, who prayed all night without falling asleep. The first coffee drink, a steeped water broth, was consumed around the year 1000 AD. Arabs from the port of Al Mukkah (Mocha) on the Red Sea became the sole source for the world's coffee controlling the lucrative coffee market by only permitting the export of boiled or roasted beans. In the 1600's, smugglers broke the Arabian monopoly in coffee growing. They took seven seeds of unroasted coffee beans from the port of Mocha to the western Ghats of southern India. In the early 1700's, the Dutch began cultivating descendants of the original plants in Java [1, 2]. Today coffee is the second most important trade commodity, second to oil [3]. Coffee shops grew 20% annually from 1991 to 1995 with an expected four fold increase by 1999 making coffee shops the fastest-growing type of food and drink outlet in the United States [4]. However, coffee houses or bars are not a new phenomenon. New York colonists first brought coffee to their breakfast table in about 1668 to replace beer. Coffeehouses became the centers of cities' business.

70

political and social life during colonial times. Court trials and city council meetings were held in early coffee houses. Paul Revere plotted the American Revolution at the Green Dragon Coffee House in Boston [2]. Breakfast remains the most popular time of day for coffee consumption in the US [5]. Coffee sales in the United States reached $7.4 billion in 1995 with a 2 cup per person daily consumption [4]. Consumer tests show that the taste of coffee is the most important factor in purchasing coffee, thus understanding the aroma profile of coffee is imperative [6]. The two commercially consumed varieties of coffee come from Coffea arabica and Coffea canephora var. robust a. Most supermarket coffees are a blend of the two and most instant coffees are made from Robusta beans. Robusta beans are generally considered inferior to the more expensive Arabica beans. Coffee grows in the regions between the Tropic of Cancer and the Tropic of Capricorn. Many countries' economy depends on its sales of coffee beans. Beans grown at lower altitudes are believed to be of lower quality with less flavor. Where the coffee is grown is very important to the quality. Table 1 shows commonly accepted characteristics of beans grown in various regions [7].

Table 1 Characteristics of coffees from different regions of the world. GENERAL AREA African Arabian Peninsula Hawaii Caribbean Indonesian Central American South American

COUNTRY OR TYPE Tanzanian, Kenya, Ethiopian Yemen (Mocha)

CHARACTERISTICS Heavy body; bright and floral; excellent for blending Heavy body but more aroma than African coffees. Kona No body; some aroma Jamaica Blue Mountain Balance of body and aroma Java, Sumatra, Celebes Balance of body and aroma; spicy Nicaraguan, Mexican, Some body, lots of aroma; Costa Rican, Guatemalan hints of cocoa Colombian, Brazilian Some body; lots of aroma; nutty

The coffee bean is actually half of a bean found inside a fruit called the coffee cherry. The coffee cherry is ripe when the skin is red and has two green beans inside. The fruit is picked by hand since the fruits ripen at different times on the same bush. The fruit is fermented to loosen the beans, which are then removed, washed and dried. There are two methods of extracting the green seed from the fruit: the wet method and the dry method. The wet method produces a higher acidity and cleaner flavor than the dry method which produces an increased body and earthy flavors [3]. The green bean is the commodity primarily traded. It is roasted by a roastmaster at 180 °C which is primarily w h e n

71

the characteristic aromas are formed. Formation pathways of many coffee odorants at roasting conditions have been discussed by Holscher [8], Baltes [9], and Tressl [10]. During roasting the composition of the beans dramatically changes; sucrose content drops from 7.3% to 0.3%, chlorogenic acid drops from 7.6% to 3.5% and protein content goes from 11.6% to 3.1%. Free amino acid levels also change greatly [11]. The length of time for roasting affects the amount of caffeine in the beans; the darker the roast the less the caffeine. Roastmasters use both smell and sight to determine when the type of roast they desire has been achieved. The roast is differentiated based on color from a Light city roast, city roast, Brazilian to Viennese, French roast, Spanish -Cuban and espresso being the longest roast time and darkest bean [7]. Due to the high quantity of unsaturated oils (13%), coffee beans are highly vulnerable to autoxidation. Different brewing methods call for different sizes of grinds. Grinds for espresso are much finer than those used for the long slow method of percolation which use a course grind. Contact with light and moisture affect the composition of the coffee bean while stored. All of these factors make coffee flavor highly variable. 1.1. Coffee Aroma and Brewing Method The enticing aroma of coffee cannot be characterized by a single chemical component but is a combined response to many different chemical components. More than 800 volatile compounds have been found in roasted coffee [12, 13]. Only a small number of these volatile compounds contribute to the aroma. Aroma profiles of the green beans, roasted beans, brewed coffee, Coffea arabica, and Coffea canephora var. Robusta have been evaluated [8,14-21]. The Werner Grosch group has quantitated 22 important odorants in coffee brews by stable isotope dilution assay and identified 32 of the 38 odorants detected. Stable isotope dilution assay, aroma extraction dilution analysis (AEDA), odor active value (OAV) analysis, and gas liquid chromatographic analysis have been conducted on coffee [22, 23]. Extraction temperature, time and particle size are among the ways brewing methods profoundly affect coffee flavor. In the US market filtered coffee (extracted between 2 and 10 min) falls in between the extremes of espresso (extracted in seconds) to percolator coffee (extracted between 15 to 30 min) and it is used widely around the world. In this study a laboratory method for brewing filtered coffee was developed to allow both controlled brewing time followed by rapid cooling and solvent extraction. Because the technique involved cooling under reduced pressure, the potential for aroma loss was examined by comparing the gas chromatography-olfactometry (GCO) data from solvent extracts of rapidly cooled coffee with coffee cooled in an ice bath. This quick brew method produces an extractable brew ideal for GCO analysis. Using the experimental brew method, shorter brew times (down to seconds); immediate and rapid cooling; controlled contact time of water and grinds; and controlled brew time are easy to achieve. This quick brew method uses apparatus available in most modern chemistry laboratories. It allows for the extraction into water of the coffee aromas with a minimized loss of aroma with the water vapor.

72

The aroma profile of a cup of coffee is variable and can be influenced by bean origin, annual weather conditions, roasting method and time, grind size, freshness, and brewing procedures. By using this experimental brewing method, brewing time and temperature can be easily controlled and comparisons may easily be made between experiments while producing an extractable simulation of a typical cup of coffee.

2. MATERIALS AND METHODS 2.1. Brewing Methods 2.1.1. Quick Brew The coffee grind to water ratio most commonly reported in the literature (0.035, [24, 25]) was used. Approximately 50 g of a blend of Brazilian, Guatemalan, and Colombian roasted Arabica coffee beans were ground in a Krups Type 203 for 8 sec to achieve a particle size range of 300-500 jim. The apparatus used for the experimental brew is shown in Figure 1. Distilled deionized water (1250 mL, 95 °C) was filtered over the roasted and ground coffee (45.0 g) on a UF-50 filter (Mr. Coffee., Inc., Bedford Heights, OH) in a 10 cm diameter Buchner funnel attached to a 20 cm water cooled condenser collected in a 2000 mL ice bath cooled vacuum flask. The condenser and exposed glassware other than the funnel were insulated and chilled with frozen chill packs. A minimal vacuum was pulled (0.13 atm) to achieve an increased flow rate (3.5 mL/sec) and reduced brewing time (6 min). 2.1.2. Conventional Brew Water temperature, grind size, water volume, filter paper, and grind quantity were held constant for the conventional brew method. The conventional brew had a flow rate of 1.0 mL/sec. After brewing, the coffee was chilled in an ice bath to 35 °C. 2.2. Aroma Extraction and Dilution Analysis The aroma extraction procedures are summarized in Figure 2. The brew (1.0 L) was successively extracted with a nonpolar solvent, Freon 113^^ (666 mL) and a polar solvent, ethyl acetate (666 mL). This successive extraction with two solvents (non-polar and polar) produces a greater volatile recovery than would have been achieved using a single solvent. Each solvent extraction was first stirred gently with a magnetic stirrer for 30 minutes, then separated in a separatory funnel, and dried by filtering over MgS04. The extracts were concentrated 243 times at 0.5 atm for freon and at 0.8 atm for ethyl acetate in a rotary evaporator. The concentrates were then diluted in increments of 3-fold. Gas chromatography-olfactometry (GCO) using CharmAnalysis was conducted on the dilutions down to the concentration in which no aroma could be detected [26]. A GCO run consisted of a 1 vil injection into a 0.25 mm x 10 m column coated with 0.52 micron OVIOI methyl silicone in an HP5890 gas

73

Water at 95X Coffee Grinds 45.0 g

Coffee Filter

Cold Packs Water

Vacuum 4 in Hg Coffee Ice Bath

Figure 1. Diagram of quick brewing method.

chromatograph modified by DATU, Inc. The temperature was held at 35 °C for 3 min, programmed at 6 °C/min to 225 °C. The injector temperature was 200 °C and the detector was held at 225 °C. Retention times of all odor active compounds were recorded on a Macintosh^^^ computer and converted to retention indices by linear interpolation of the retention times of a series of 7-18 carbon paraffin standards run under identical conditions and detected with a flame ionization detector (FID) [27]. The retention times of the n-paraffins were measured before each series of analyses and periodically between GCO analyses to account for any changes in the column. The OVIOI column was used because it elutes most odorants at the lowest possible temperature and can be temperature-programmed at high rates to minimize sniffer fatigue [26]. The same human subject was used for all GCO analyses. Multiple measures of each GCO analysis were conducted on 2 replicates of the brewing methods and extraction procedure, comprising a total of sixteen sets of dilutions. The corresponding data from the two solvents were grouped together to total all the aromatic components of the coffee brew. The resulting dilution analyses were converted into Charm units (the areas of the peak in the Charm chromatogram) a unitless ratio proportional to

74

the amount of eluting stimulus divided by its odor-detection threshold [27]. Odor spectra were generated from the Charm data using an exponent of 0.5 and normalizing to the most potent odorant. Chemical identification of odor ants in the coffee samples was based on an exact match of odor character and retention index with that of an authentic standard [27] [28]. Gas chromatography-mass spectrometry (GC-MS) correlation of authentic standards verified the chemical identification. The GC-MS was conducted at 70 electron volts on a mass range of 33-300 M / Z in an HP5970. The HP5890 GC was programmed to heat isothermally at 35 °C for 3 min and then increase at 4 °C/min to 240 °C. The same column type used for GCO except twice the length (20m) was used for GC/MS. The injector temperature was 200 °C and the detector was held at 250 °C.

1 Liter Coffee Sample added 666 mL Freon 1 13TM stirred gently 30 min separated and dried over MgS04 Freon 113TM

- added 666 mL ethyl acetate - stirred gently 30 min - separated and dried over MgS04 Water

Ethyl Acetate

Discan

3

^Concentrate 243X ^ by rotovap

f Serial Dilutions by factor of 3X

Concentrate 243X^ by rotovap

Serial Dilutions^ by factor of 3X

f CharmAnalysisTM ^

r CharmAnalysisTM ]

(Ethyl Acetate CHARM]

(Freon 113 CHARM)

i

CHARM GROUP TOTAL

Figure 2. Flow summary of solvent extraction of the coffee brews.

75 3. RESULTS AND DISCUSSION The thirty most potent aroma chemicals detected in the coffee extracts (spectral values greater than 1.0%) are listed in Table 2. The 18 odorants identified were also among the most potent odorants detected in coffee by Aroma Extraction Dilution Analysis (AEDA) in three other studies [22-24, 29]. In all three studies 2-furfurylthiol and P-damascenone were among the top three most potent odorants found in coffee. As shown in Table 2, the odorants were the same in both brews although they were ranked somewhat differently. Table 3 shows that quantitative GCO data is very noisy since the ranking variation in spectral values contributed by multiple measures (Al, A2) is almost as great as the variation contributed by the replicate samples (Al, Bl). Therefore, the ranking data should be accepted as approximations and perhaps listed as "most potent groups," not individual compounds. These errors partially result from using a human subject as a GC detector. To compare the yield of the two methods, total Charm (sum of the peak areas in the Charm chromatogram) for each grouped chromatogram was logarithmicly transformed (for normalization) and compared using analysis of variance (ANOVA). A significant difference between the extracted aromas from the two methods was detected at p=0.03. The experimental brew produced 100% greater total Charm than the conventional brewing method. The challenge with comparison of individual chemical responses is that the system is over defined; there are more variables (intensity measurements) than there are cases (brewing methods and replications). It would not be reasonable to increase the number of cases due to cost and time of each experiment. Spectral data was used since cluster analysis strongly indicated an increase in charm values between the duplicate. Zero charm values were replaced with a calculated upper limit equal to 3 s where s was the standard deviation in the blank. For this data s was taken as the median standard deviation, 2.3. Any charm value below 6.9 was thus replaced with 6.9 [27]. The spectral data was arcsine square-root transformed. Six chemicals (methional, E-2-nonenal, sotolon, guaiacol, 5-methyl-6,7dihydrocyclopyrazine, and Furaneol) were selected because they all varied in the same direction. The selection was required to reduce the number of variables. A factor analysis using Statistica resulted in three factors with an eigenvalue greater than 1.0 and also exhibited an apparent cut off on a Scree plot. The resulting factors explaining 86% of the variation were varimax rotated (Table 4). Multivariate analysis of variance (MANOVA) was conducted considering brewing method and duplication with factor 2 and 3. There was an overall intensity increase in the data from the first run to the duplicate. Based on the MANOVA and the factor analysis, it can be concluded that there is a 280% (p=0.03) increase in concentration of methional comparing the conventional brewing method to the quick brewing method. Sotolon demonstrated a 167% increase and cis-2-nonenal demonstrated a 100% decrease at a significance level of 15%. Using discriminate analysis, methional and cis-2-nonenal showed a significant change (p=0.5).

76 Table 2 Aroma occurrences resulting from CharmAnalysis of two brewing methods of coffee. Retention times were converted to retention indices (RI) by linear interpolation of the retention times of the series of 7-18 carbon paraffin. CHEMICAL STIMULANT

sotolon P-damascenone 2-furfurylthiol 4-vinylguaiacol 2-methyl-3furanthiol vanillin guaiacol furaneol methional 3-methoxy-2isobutyl pyrazine unknown unknown 2,4,5trimethylthiazole Abhexon unknown unknown unknown 4-ethyl guaiacol 5-methyl-6,7dihydrocyclopentapyrazine unknown unknown 2-ethyl-3,5-din\ethylpyrazine cis-2-nonenal unknown unknown unknown unknown unknown 2-isopropyl-3methoxypyrazine 2,3,5trimethylpyrazine

RI

EXPERIMENTAL CHARM OSV

CONVENTIONAL CHARM OSV

81 98 100 62 89

DESCRIPT

toast fruit toast cloves nuts

1057 1349 881 1279 844

46200 41123 37226 22327 19701

100 94 90 70 65

13937 20266 21092 7998 16740

1335 1066 1033 863 1160

18899 16159 15152 14221 8378

64 59 57 55 43

10773 12641 7064 3950 2989

1502 1252 965

5331 5285 4973

34 34 33

2016 1117 3035

31 burnt 23 floral 38 plastic

1156 990 1403 1222 1250 1110

2977 2493 2059 2001 1692 1613

25 23 21 21 19 19

4086 1118 2006 856 2027 983

44 23 31 20 31 22

1285 850 1045

1507 1280 907

18 17 14

808 1107 1295

20 cloves 23 stinky 25 burnt

1132 1206 1142 908 984 803 1076

866 865 656 589 547 480 464

14 14 12 11 11 10 10

1585 576 495 392 357 351 403

27 17 15 14 13 13 14

10

449

971

461

CAS NUMBER

28664-35-9 23726-93-4 98-02-2 7786-61-0 28588-74-1

71 vanilla 121-33-5 90-05-1 plastic n 58 caramel. 3658-77-3 3268-49-3 43 potato 24683-00-9 38 plants

13623-11-5

honey 698-10-2 plastic spice honey 2785-89-9 spice cotton 23747-48-0 candy

18138-04-0

18829-56-6 toast licorice cereal nutty plastic skunk green 25773-40-4

15 toast

14667-55-1

77

Table 3 Comparison of spectral results from GCO multiple measures (1 and 2) and brewing replicates (A and B) for the experin\ental brewing method extracts for the ten most potent components. Data are combined results from ethyl acetate and Freon 113^^ fractions. AROMA CHEMICAL A l 2-furfurylthiol p-damascenone 2-methyl-3-furanthiol sotolon guaiacol vanillin 4-vinylguaiacol furaneol methional 3-methoxy-2-isobutyl pyrazine

A2

100 19 56 lb 6 32 62 5 37 19

71 31 20 67 29 75 100 28 35 19

Bl 66 34 50 100 23 33 57 27 23 15

B2 100 35 13 31 11 40 7 14 11 10

STDev multiple measures 34 12 7 30 18 13 36 21 18 6

STDev replicates

STDev All

32 9 39 30 21 33 45 21 6 2

18 7 21 29 11 20 38 11 12 4

Table 4 Factor loading (variance maximized rotated) from factor analysis of selected chemicals' arcsine square-root transformed spectral data. AROMA CHEMICAL

FACTOR 1

FACTOR 2

FACTOR 3

methional furaneol sotolon guaiacol 5-methyl-6,7dihydrocyclopyrazine E-2-nonenal

-0.8 0.9 -0.1 0.9 0.6

0.0 -0.2 0.7 0.2 0.6

0.96 0.1 0.5 -0.3 -0.2

0.0

0.8

0.0

4. CONCLUSIONS CharmAnalysis and AEDA detect the same important aroma chemicals in coffee but variability in the data makes it difficult to obtain exact orders of importance. The experimental brewing method described here should minimize errors by providing better control of time and temperature. Although quantitative GCO is more error prone than other chemical measurements, it is useful for understanding the affects of various treatments on coffee aroma and provides direction for more precise chemical analysis such as isotope dilution analysis. 5. ACKNOWLEDGMENTS We are grateful for the financial and sample support from Nihon Tetra Pak.

78

6. REFERENCES 1. S. Braun, Buzz, The Science Lore of Alcohol and Caffeine. 1996, New York: Oxford University Press. 2. Krups, The Encyclopedia of Coffee and Espesso From Bean to Brew. 1995, Chicago: Trendex International, Inc. 160. 3. R. J. Clarke, and R. Macrae, Chemistry. Coffee. Vol. 1. 1985, New York: Elsevier Applied Science. 306. 4. Research Alert, Oct. 18 (1996) 5. C. A. National, Automatic Merchandiser, (1995) 38. 6. E. Maras, Automatic Merchandiser, (1996) 28. 7. T. Neuhaus, The Informed Baker.1996, Ithaca, NY: Cornell University. 8. W. Holscher, and H. Steinhart, Thermally Generated Flavors, Maillard, Microwaves, and Extrusion Processes, T. Parliment, M. Morello, R. McGorrin, Editor, (1994), American Chemical Society, 207-217. 9. W. Baltes, and G. Bochmann, Z. Lebensm Unters Forsch, 185 (1987) 5-9. 10. R. Tressl, Thermal Generation of Aromas,, (1989), ACS, 293-301. 11. I. Flament, and C. Chevallier, Chemistry and Industry, (1988) 592-596. 12. N. Imura, and O. Matsuda, Nippon Shokuhin Kogyo Gakkaishi, 39 (1992) 531-535. 13. A. Stalcup, K. Ekborg, M. Gasper, D. Armstrong, J. Agric. Food Chem., 41 (1993) 1684-1689. 14. W. Holscher, O. G. Vitzthum, H. Steinhart, The Cafe Cacao, XXXIV (1990) 205-212. 15. I. Blank, Sen, A., W. Grosch, Z. Lebensm Unters Forsch, 195 (1992) 239-245. 16. C. A. B. De Maria, L. Trugo, R. Moreira, C. Werneck, Food Chemistry, 50 (1994) 141-145. 17. W. Grosch, Trends in Food Sci. & Tech., 4 (1993) 68-72. 18. N. K. O. Ojijo and P. B. Coffee Research Foundation, Ruiru, Kenya., Kenya Coffee, vol. 58 (685) (1993) p.1659-1663. 19. N. qijo, Kenya Coffee, 58 (1993) 1659-1663. 20. O. Vitzthum, C. Weisemann, R. Becker, H. Kohler, The Cafe Cacao, XXXIV (1990) 27-32. 21. A. Williams, and G. Arnold, J. Sci. Food Agric, 36 (1985) 204-214. 22. P. Semmelroch, G. Laskawy, I. Blank, adn W. Grosch, Flavour and Fragrance J., 10 (1995) 1-7. 23. P. Semmelroch, and W. Grosch, J. Agric. Food Chem., 44 (1996) 537-543. 24. I. Blank, A. Sen, and W. Grosch, ASIC. 14 Colloque, San Francisco, (1991) 117-129. 25. T. Lee, R. Kempthorne, J. Hardy, J. of Food Sci., 51 (1992) 1417-1419. 26. T. Acree, J. Barnard, D. Cunningham, Food Chemistry, 14 (1984) 273-286. 27. T. Acree, and J. Barnard, Trends in Flavour Research, H. a. D. G. v. d. H. Maarse, Editor, (1994), Elsevier, 211-220. 28. L. Ettre, Chromatographia, 7 (1974) 38-46. 29. P. Semmelroch, and W. Grosch, Leben. Wiss.u-Technol, 28 (1995) 310-313.