Analysis of Freshly Brewed Espresso Using a Retronasal Aroma Simulator and

Transcription

1 Food Sci. Technol. Res., 15 (3), , 2009 Analysis of Freshly Brewed Espresso Using a Retronasal Aroma Simulator and Influence of Milk Addition Masayuki akiyama 1*, Kazuya murakami 1, Michio ikeda 1, Keiji iwatsuki 1, Akira wada 2, Katsuya ToKuno 2, Masanobu onishi 2, Hisakatsu iwabuchi 2 and Yasuyuki sagara 3 1 Morinaga Milk Industry Co., Ltd., , Higashihara, Zama, Kanagawa, , Japan 2 San-Ei Gen F.F.I., Inc., , Sanwa-cho, Toyonaka, Osaka, , Japan 3 Department of Global Agricultural Sciences, Graduate School of Agricultural and Life Sciences, The University of Tokyo, Yayoi, Bunkyo-ku, Tokyo, , Japan Received August 29, 2008; Accepted February 18, 2009 Abstract: A quick (headspace: 1 min and retronasal aroma simulator (RAS): 2 min) sampling method for capturing volatiles released from coffee drinks by solid-phase microextraction (SPME) (fiber type; divinylbenzene (DVB)/carboxen/polydimethylsiloxane (PDMS)) has been applied to freshly brewed espresso and café latte (espresso and milk). Headspace volatiles and volatiles in the RAS effluent were collected from espresso and café latte, and examined by gas chromatography/mass spectrometry (GC/MS) and GC/olfactometry (GC/O, CharmAnalysis ). When milk was added to espresso in 60% by weight, aroma release (odor activities) was generally suppressed, both in headspace and RAS aromas. The relative charm value of sweet-caramel odor increased in the headspace aroma, while the phenolic odor increased markedly and the smoke-roast odor decreased in the RAS aroma. These results suggested that milk addition had different effects on the headspace and RAS aromas. Keywords: espresso, coffee, milk, retronasal aroma, solid-phase microextraction (SPME), gas chromatography/olfactometry (GC/O, CharmAnalysis ) Introduction Coffee is a widely consumed beverage with a complex, but pleasant aroma. Numerous studies have been conducted on coffee to date and more than 800 volatile compounds have been identified (Nijssen et al., ; Semmelroch and Grosch, 1995, 1996; Czerny et al., 1996, 1999; Grosch, 1998; Mayer et al., 2000; Sanz et al., 2002). Coffee is often enjoyed with milk which suppresses the bitterness, astringency, and acidity of the brew (Kim et al., 1995). An understanding of these effects is particularly important given the number and success of café latte-type coffee drinks on the Japanese ready-to-drink coffee market. Solid-phase microextraction (SPME) is a rapid, simple, inexpensive, and reproducible method used to collect released volatiles (Arthur and Pawliszyn, 1990; Yang and Peppard, 1994; Bicchi et al., 1997; Steenson et al., 2002; Jung *To whom correspondence should be addressed. m_akiyam@morinagamilk.co.jp and Ebeler, 2003; Charles-Bernard et al., 2005; López-Galilea et al., 2006). The short-duration SPME sampling method whose effectiveness for collecting fresh coffee volatiles has been reported in our previous studies (Akiyama et al., 2007, 2008) has been applied to sample headspace volatiles and volatiles in the effluent from a retronasal aroma simulator (RAS) released from freshly brewed espresso and café latte. A RAS was developed by Roberts and Acree (1995) for use as one of the various types of model mouths (van Ruth et al., 1994; Roberts and Acree, 1995; Elmore and Langley, 1996). In addition to other methods, these are useful tools for the study of retronasal aromas, such as breath-by-breath analysis (Graus et al., 2001) and oral breath sampling (OBS) (Roozen and Legger-Huysman, 1995). One of the objectives of this study was to compare aromas of the conventional headspace and the effluent from a RAS in coffee volatiles. Since flavor release changes in response to the addition of milk to coffee (Bucking and Steinhart, 2002; Denker et al., 2006), another objective of this

2 234 study was to assess the effect of milk addition on espresso. The changes in released espresso aromas in the headspace and the RAS effluent, and through the addition of milk, were therefore analyzed by GC/MS and GC/O. The obtained results provide useful information for the flavor design in the industrial production of coffee drink products. Materials and Methods Coffee and milk samples Guatemala SHB arabica coffee (Coffea arabica) beans were supplied by Unicafe Inc. (Tokyo, Japan), and dark roasted (L value, 18) using a Probat roaster (G-12, Emmerich, Germany). The degree of roasting was represented as an L value, which was determined using roasted ground coffee (particle size: < 500 µm) and a color meter (ZE-2000, Nippon Denshoku Industries Co., Ltd., Tokyo, Japan). The roasted coffees were divided into 1 kg portions and stored at 20 until use. All roasted coffee beans were held at room temperature for 2 h before grinding to allow their temperatures to equalize. Commercial, ultra-high-temperature (UHT)-pasteurized (135, 2 sec) milk (Morinaga Milk Industry Co., Ltd., Tokyo, Japan) was used in the milk addition experiments. The milk contained more than 3.5% fat and more than 8.3% solid of non-fat. Brewing espresso Grinding of roasted coffee (about 7.5 g) and brewing of espresso using ion-exchange hot water (pressure of 15 atm) was performed using a Saeco Royal Professional coffee machine (Nihon Saeco K. K, Kawasaki, Japan). The brewing using a fully automatic espresso machine was performed to standardize brewing conditions and preparation, taking approximately 30 sec from grinding to brewing. The espresso was collected in a 100 ml beaker, which was then covered with aluminum foil. The espresso was characterized as follows: volume, about 65 g; Brix, about 2.5 ; temperature of espresso, about 70 ; temperature of headspace air, approximately 55. SPME device and sampling conditions For sampling headspace and RAS volatile compounds of espresso, DVB/ carboxen/pdms fiber (50/30 µm thickness) was selected from six types of SPME fibers for use with the SPME device (Sigma-Aldrich Co., St. Louis, MO, USA) as described in m. akiyama et al. a previous study on freshly brewed drip coffee aroma (Akiyama et al., 2007). Before headspace and RAS sampling, the SPME fiber was reconditioned in the GC injection port according to the manufacturer's instructions (data sheet T ; Sigma-Aldrich Co.). When brewing of espresso was completed, sampling was started by exposing an SPME fiber to the headspace in each glass beaker covered with aluminum foil. For determination of the appropriate SPME sampling time, the SPME fiber was exposed to the headspace of espresso for 0.5, 1, 2, and 4 min, respectively, after finishing brewing of espresso. Based on these findings results, the SPME fiber was exposed to the headspace air for 1 min; and the RAS air for 2 min, and the collected volatiles were analyzed by GC/MS and GC/O. After sampling, the fiber was placed into the injection port of the GC/MS or GC/O and thermally desorbed for 10 min at 250. Each SPME sample was conducted in triplicate. Headspace SPME sampling from freshly brewed espresso and café latte Water and milk was mixed in a 300 ml beaker in the ratio shown in Table 1, and heated using a microwave oven (NE-N1, Matsushita Electric Industrial Co., Ltd., Osaka, Japan). Into the heated mixture of water and milk (ca. 70, total 130 g), the espresso was brewed, and the beaker was covered with aluminum foil after mixing the contents. The SPME fiber was pushed out of its housing (stainless steel housing which was set on the aluminum foil as shown in Fig. 1) immediately after brewing of the espresso, and exposed to the headspace air. RAS parameters and SPME sampling from freshly brewed espresso and café latte using RAS A method using RAS (Roberts and Acree, 1995) was used to sample retronasal aroma. The RAS consisted of a 1-L stainless steel blending container which had a water jacket for controlling the temperature (38 ), a voltage controller and high torque-speed motor to precisely control the rotation speed of the shear blade (650 rpm) to simulate a model mouth, and a controlled nitrogen gas supply (1000 ml/min) as a carrier gas to sweep over the freshly brewed espresso (milk 0%) or café latte (milk 20% (w/w) or milk 60% (w/w)) using the RAS system. The RAS volatiles were trapped by exposing the SPME fiber to the effluent gas (Fig. 1). Table 1. Sample composition of espresso (milk 0%) and café latte (milk 20% (w/w) and milk 60% (w/w)). (w/w) Espresso Milk Water Milk solid (non-fat) (SNF) Milk fat Coffee a (in terms of green beans) Milk 0% 65 g 130 g 5.0% Milk 20% 65 g 39 g 91 g 1.7% 0.7% 5.0% Milk 60% 65 g 117 g 13 g 5.0% 2.1% 5.0% a Weight of green coffee beans is 1.3 times of weight (7.5g) of roasted coffee beans used for brewing espresso.

3 Retronasal Aroma of Espresso 235 Brewing head of espresso machine Freshly brewed espresso was dripped directly into a beaker. SPME device SPME device Flow controller Flow meter Water Water jacket AC Variable speed Water Sealed blending assembly Gear assembly Headspace Retronasal Aroma Simulator (RAS) Fig. 1. Sampling method and apparatus used to trap headspace and RAS volatile compounds of freshly brewed espresso and café latte (espresso and milk). GC/MS analysis GC/MS analyses were performed in triplicate on a 5973 mass selective detector (Agilent Technologies, Santa Clara, CA, USA) with a fused silica capillary column DB-WAX (60 m 0.25 mm, 0.25 μm film thickness, Agilent Technologies). The flow of the helium carrier gas was 1.6 ml/min. The oven temperature was programmed at an initial 50 for 2 min, followed by an increase of 3 /min to 220, and held at 220 for 20 min. The injection port, equipped with a 0.75 mm i.d. liner (Sigma-Aldrich Co.) was maintained at 250. The inlet was operated in the splitless mode, and the injection purge on the GC was off for the initial 1 min of the manual SPME injection. GC/MS analyses were conducted under the same conditions where quantitative comparisons of data were necessary. GC/O analysis The dilution GC/O analysis using different lengths of SPME fiber was developed by Deibler et al. (1999), and has been successfully applied to GC/O analyses of the aromas of ground roasted coffee, coffee brew, and the coffee aroma during grinding (Deibler et al., 1999; Akiyama et al., 2003a, 2003b, 2005, 2007, 2008). Using four different lengths of SPME fibers (2.5 mm, 5 mm, 10 mm, and 20 mm), up to 1/8 dilution GC/O analysis was accomplished. In the present study, a 1/64 dilution was achieved by a recently developed dilution method which involves varying the split ratio (Deibler et al., 2004) and using two different lengths of SPME fiber (Table 2) to obtain a more detailed assessment of odor activity than that which is possible by the conventional 1/8 dilution method using varying fiber lengths. Volatile compounds in the headspace of glass beakers and the effluent gas from the RAS of freshly brewed espresso and café latte were thus collected using SPME fibers cut to different lengths (20 mm and 2.5 mm), respectively. After sampling, the fibers were placed into the injection port of the GC/O, and thermally desorbed for 10 min at 250 at different dilution ratios (1, 1/2, 1/4, and 1/8) using four different settings (1:0, 1:2.7, 1:12.9, and 1:51.0) as shown in Table 2. GC/O analyses were conducted in triplicate using CharmAnalysis on a 6890 GC (Agilent Technologies) modified by DATU, Inc. (Geneva, NY, USA) (Acree et al., 1984). A fused silica capillary column DB-WAX (15 m 0.32 mm, 0.25 μm film thickness; Agilent Technologies) was used, and the flow of the helium carrier gas was 3.2 ml/min. The oven temperature was set to an initial temperature of 40, followed by an increase at 6 /min to 230 where the temperature was held for 20 min. The injection and detector ports were maintained at 250 and the injection purge on the GC was off for the initial 1 min. The retention time of each compound was converted to Kovats indices using C 6 -C 28 n-alkanes (Tokyo Kasei Kogyo Co., Ltd., Tokyo, Japan). GC/O evaluation During GC/O analyses using Charm- Analysis, a sniffer sniffs the air coming from the olfactometer and holds down the mouse button when he or she detects an odor in the air, and the computer records the perception, the time and duration of the sensation. Odor activities of Table 2. Settings for GC/O dilution analysis by varying length of SPME fiber and GC split ratio. Target injection fraction for factor of two dilution series 1 1/2 1/4 1/8 1/16 1/32 1/64 SPME fiber length 20 mm 20 mm 20 mm 20 mm 2.5 mm 2.5 mm 2.5 mm Calibrated split ratio settings * 1 : 0 1 : : : : : : 51.0 * Split ratios expressed as the proportion injected into the column : proportion in exhaust.

4 236 volatile compounds obtained by GC/O dilution analyses were recorded as charm values (Acree et al., 1984), and the relative importance of component odorants was expressed in terms of the odor spectrum value (OSV), which is the normalized charm value modified by an approximate Stevens law exponent (n = 0.5) to the most potent odorant detected (Acree, 1997). OSVs are calculated using the formula, OSV = ((charm value)/(charm value) max ) 1/2 100, and are used to approximate the relative importance of odorants by accounting for the exponential nature of olfactory psychophysics; charm values indicate the true odor activity measurement and are a linear function of concentration. Each charm value was rounded off to two significant figures in order to reflect the actual resolution of the dilution analysis. Acidic, butteryoily, green-blackcurrant, green-earthy, nutty-roast, phenolic, smoke-roast, soy sauce, sweet-caramel, and sweet-fruity were the aroma descriptions used in all GC/O experiments to describe potent odorants. These descriptions were chosen from the results of a single preliminary free choice GC/O analysis using a lexicon of words commonly used for coffee evaluation (Akiyama et al., 2003a). Identification of the volatile compounds Standard compounds used were obtained from Tokyo Kasei Kogyo Co., Ltd., Sigma-Aldrich Co., CTC Organics (Atlanta, GA, USA), or Oxford Chemicals Ltd. (Hartlepool, England). The preparation of 2,3-dihydro-6-methylthieno[2,3c]furan, and 1-(3,4-dihydro-2H-pyrrol-5-yl)-ethanone were conducted as described in the literature, respectively (Buchi et al., 1971; De Kimpe et al., 1993). Volatile compounds were identified by comparing their mass spectra and Kovats indices using C 6 -C 28 n-alkanes to those of standard compounds and to those from the literature (Baltes and Bochmann, 1987a, 1987b, 1987c). Also, tentative identifications of some potent odorants found only by the GC/O analyses were made by comparing their Kovats indices and aroma properties to those of standard compounds and to those from the literature (Holscher et al., 1990). When compounds were detected only by GC/O analysis, GC/O analyses were repeated using two columns with different polarities (DB-WAX and HP-5 (15 m 0.32 mm, 0.25 μm film thickness; Agilent Technologies)), and their Kovats indices and odor characteristics were compared with those of standard compounds to verify that they were accurately identified. Statistical analysis Analysis of variance (ANOVA), Tukey HSD analysis, and t tests were performed using a statistics package (SPSS 9.0J for Windows, SPSS Japan Inc., Tokyo, Japan). Results and Discussion Parameters for headspace- and RAS-SPME sampling m. akiyama et al. The SPME technique is widely regarded as a simple, rapid, sensitive, and highly reproducible sampling method. However, in order to obtain suitable data using this technique, it is important to set up sampling conditions that can be subjected to experimental analysis. To do this effectively, SPME fiber type and exposure time of the fiber to the headspace in a glass beaker or the effluent gas from RAS needed to be carefully optimized. Selection of the most appropriate SPME fiber is very important for sampling volatiles in the same proportion as the targeted headspace volatiles. In a previous study (Akiyama et al., 2007), out of six types of fibers tested (PDMS, carbowax/dvb, PDMS/DVB, polyacrylate, carboxen/pdms, and DVB/carboxen/PDMS) for the analysis of freshly brewed drip coffee volatiles, DVB/carboxen/PDMS fiber was carefully selected for its adsorption capabilities, affinities, and sensitivity to coffee volatiles. In addition, these fibers were associated with markedly high similarities in the chromatograms obtained by the selected SPME method and conventional reduced-pressure steam distillation. The DVB/carboxen/PDMS fiber was therefore also used in this study to assess freshly brewed espresso aroma. To determine the appropriate SPME sampling time for studying headspace aroma, the SPME fiber was exposed to the headspace of freshly brewed espresso for 0.5, 1, 2, 3, and 4 min, respectively. Since the purpose of this study was to capture headspace volatiles of just-brewed espresso, sampling by SPME fiber was started when coffee brewing by espresso method was completed. The peak s of 40 important coffee volatiles sampled by SPME fiber are listed in Table 3. To identify more constituents, it is favorable to collect volatiles in larger quantities by exposing the SPME fiber for a longer time. However, a longer sampling time can lead to changes in the freshness of coffee aroma, or a constitutional difference from real headspace caused by competition phenomena on the SPME fiber. In the case of 2-min headspace sampling, no. 5 2-methylbutanal and no. 6 3-methylbutanal (Table 3) showed smaller increases in peak abundance and decreased linearity, which led to a change in the component proportion of headspace volatiles. From these results, 1 min exposure, which maintained high coefficients of determination (R 2 > 0.80) between each peak and exposure times, was selected as the optimal sampling time for headspace aroma. In this manner, in the case of 3-min RAS sampling, no. 4 2-butanone and no. 8 2,3-butanedione (Table 4) exhibited reduced increases in peak abundance and decreased linearity. Based on these findings, optimum sampling times were determined to be 1 min for the headspace aroma and 2 min for the RAS aroma. Influence of milk addition on headspace and RAS volatile

5 Retronasal Aroma of Espresso 237 Table 3. Changes in the peak of typical headspace coffee volatile compounds relative to exposure time of SPME fiber (DVB/carboxen/PDMS), and linear relationship between peak of typical headspace coffee volatile compounds and exposure time of the SPME fiber. No. Ret. time (min) Compounds 0.5 min 1 min 2 min 3 min 4 min Peak ( 10 3 ) a R 2b Methyl acetate 3, , , , , Methylfuran 6, , , , , Methylfuran , , Butanone 3, , , , , Methylbutanal 8, , , , , Methylbutanal 9, , , , , ,5-Dimethylfuran 1, , , , , ,3-Butanedione 4, , , , , ,3-Pentanedione 8, , , , , ,3-Hexanedione 2, , , , , Methyl-1H -pyrrole 5, , , , , Ethenyl-5-methylfuran , , , , Ethyl-1H -pyrrole , , , , Pyridine 49, , , , , Pyrazine 1, , , , , (Methoxymethyl)furan 2, , , , , Dihydro-2-methyl-3(2H )-furanone 5, , , , , Methylpyrazine 15, , , , , ,5-Dimethylpyrazine 5, , , , , ,6-Dimethylpyrazine 5, , , , , Ethylpyrazine 5, , , , , Ethyl-6-methylpyrazine 4, , , , , Ethyl-5-methylpyrazine 3, , , , , Trimethylpyrazine 3, , , , , Ethyl-2,5-dimethylpyrazine 2, , , , , Furancarboxaldehyde 29, , , , , [(Methylthio)methyl]furan 3, , , , , Furfuryl formate 11, , , , , (2-Furanyl)-ethanone 8, , , , , H -Pyrrole 5, , , , , Furfuryl acetate 48, , , , , Methyl-2-furancarboxaldehyde 16, , , , , ,2'-Methylenebisfuran c 2, , , , , Methyl-1H -pyrrole-2-carboxaldehyde 3, , , , , Furanmethanol 62, , , , , (2-Furanylmethyl)-5-methylfuran c , , , , ,3-Dihydro-6-methylthieno[2,3c ]furan 2, , , , , (2-Furanylmethyl)-1H -pyrrole 4, , , , , Methoxyphenol 1, , , , , Ethenyl-2-methoxyphenol 1, , , , , a Peak s are average values of three measurements. Fractions are considered to be within acceptable limits as determined by integration and rounded off. b Coefficient of determination. c Tentative identifications obtained by comparing mass spectra and retention indices with data in the literature (Baltes and Bochmann, 1987a). compounds of freshly brewed espresso To obtain headspace and RAS volatile composition of freshly brewed espresso aroma and to investigate the influence of milk addition on headspace and RAS volatile compounds, the headspace and RAS volatile components of freshly brewed espresso and café latte (espresso with milk in 20% or 60% (w/w), Table 1) were captured using the SPME sampling methods for analysis by GC/MS. The peak s of the identified 40 compounds, and their relative standard deviations (RSDs) analyzed under the same GC/MS condition are shown in Table 5. In three repeated GC/MS analyses, the RSDs of peak s for 40 typical volatile coffee compounds in the headspace and RAS were within 10%, demonstrating that the sampling method was reproducible. The peak of 40 volatiles in the GC/MS analysis of espresso with milk in 0%, 20%, and 60% (w/w) were statistically analyzed by a Tukey HSD multiple comparison test. Significant differences (* p < 0.05, ** p < 0.01) of peak s among milk 0%, 20%, and 60% (w/w) were then assessed in 22 headspace volatile compounds and are shown in Table 5. Twenty-one out of the 22 compounds showed a significant decrease when milk was added (20% and 60% (w/ w)). Whereas, of the headspace compounds, only the no. 14 pyridine showed a significant increase in peak after the addition of milk in 20 or 60% (w/w), which could be caused by the increase in ph due to milk addition. The observed retardation effect was therefore caused by the milk addition, corroborating previous findings in which the addition of different milk or vegetable products typically reduced the amount of volatiles in the headspace of the coffee beverages being tested, and the effect was typical for each additive (Kim et al., 1995; Bücking and Steinhart,

6 238 m. akiyama et al. Table 4. Changes in the peak of typical RAS coffee volatile compounds relative to exposure time of SPME fiber (DVB/carboxen/PDMS), and linear relationship between peak of typical RAS coffee volatile compounds and exposure time of the SPME fiber. No. Ret. time (min) Compounds 0.5 min 1 min 2 min 3 min 4 min Peak ( 10 3 ) a R 2b Methyl acetate 5, , , , , Methylfuran 13, , , , , Methylfuran 1, , , , , Butanone 6, , , , , Methylbutanal 20, , , , , Methylbutanal 21, , , , , ,5-Dimethylfuran 2, , , , , ,3-Butanedione 8, , , , , ,3-Pentanedione 15, , , , , ,3-Hexanedione 4, , , , , Methyl-1H -pyrrole 18, , , , , Ethenyl-5-methylfuran 1, , , , , Ethyl-1H -pyrrole 1, , , , , Pyridine 53, , , , , Pyrazine 3, , , , , (Methoxymethyl)furan 3, , , , , Dihydro-2-methyl-3(2H )-furanone 5, , , , , Methylpyrazine 8, , , , , ,5-Dimethylpyrazine 3, , , , , ,6-Dimethylpyrazine 4, , , , , Ethylpyrazine 4, , , , , Ethyl-6-methylpyrazine 3, , , , , Ethyl-5-methylpyrazine 2, , , , , Trimethylpyrazine 2, , , , , Ethyl-2,5-dimethylpyrazine 2, , , , , Furancarboxaldehyde 29, , , , , [(Methylthio)methyl]furan 6, , , , , Furfuryl formate 17, , , , , (2-Furanyl)-ethanone 8, , , , , H -Pyrrole 9, , , , , Furfuryl acetate 61, , , , , Methyl-2-furancarboxaldehyde 14, , , , , ,2'-Methylenebisfuran c 4, , , , , Methyl-1H -pyrrole-2-carboxaldehyde 3, , , , , Furanmethanol 28, , , , , (2-Furanylmethyl)-5-methylfuran c 1, , , , , ,3-Dihydro-6-methylthieno[2,3c ]furan 2, , , , , (2-Furanylmethyl)-1H -pyrrole 5, , , , , Methoxyphenol 2, , , , , Ethenyl-2-methoxyphenol , , , , a Peak s are average values of three measurements. Fractions are considered to be within acceptable limits as determined by integration and rounded off. b Coefficient of determination. c Tentative identifications obtained by comparing mass spectra and retention indices with data in the literature (Baltes and Bochmann, 1987a). 2002). Therefore, the changes in the concentration of coffee volatiles with an additive can be attributed to several effects: lipids readily adsorb most volatiles, whereas proteins and carbohydrates can interact by adsorption, entrapment, and encapsulation (Kinsella, 1990). In case of RAS compounds, 29 compounds exhibited a significant decrease in peak when milk was added in 20 or 60% (w/w). As with headspace volatiles, only the peak of no. 14 pyridine was significantly increased by adding milk. The 22 compounds in the headspace that showed the significant differences in peak s, were common to the RAS aroma. Comparison of volatile compounds in the headspace and RAS of freshly brewed espresso and café latte Differences between headspace and RAS aromas for each of 40 potent peak compounds detected by GC/MS analysis were statistically analyzed by t tests using the relative of each peak versus total peak of the 40 compounds for the same milk content in order to compare the profiles of volatile headspace and RAS aroma components. Significant differences (* p < 0.05, ** p < 0.01) were observed in 36 compounds for espresso (milk 0%), 32 compounds for café latte (milk 20% (w/w)), and 34 compounds for café latte (milk 60% (w/w)) in Table 5. From these results, it was found that the highly-volatile compounds, such as no. 2 2-methylfuran; no. 5 2-methylbutanal; no. 6 3-methylbutanal; no. 9 2,3-pentanedione; and no methyl-1H-pyrrole, had larger relative peak s in RAS aroma than in headspace aroma, irrespective of milk content (gray background in Table 5). Unlike conventional headspace sampling, the coffee sample was agitated in a RAS vessel, which could be one reason why highly volatile compounds were more abundant in the RAS effluent.

7 Retronasal Aroma of Espresso 239 Table 5. Volatile compounds found in the headspace and the RAS effluent of freshly brewed espresso (milk 0%) and café latte (milk 20% (w/w) and milk 60% (w/w)) using the solid-phase microextraction sampling method. No. Ret. time (min) Compounds Milk 0% Headspace RAS Peak ( 10 3 ) a ( RSD%) b Milk 20% (w/w) Milk 60% (w/w) Peak ( 10 3 ) a ( RSD%) b Effect of milk Effect of milk c Milk 0% Milk 20% (w/w) Milk 60% (w/w) Comparison of headspece and RAS d Milk 0% Milk 20% (w/w) Milk 60% (w/w) Methyl acetate 3,298 (6.91) 3,253 (9.79) 3,207 (8.11) 10,867 (2.87) 11,445 (4.42) 11,540 (4.13) * ** ** Methylfuran 10,221 (9.35) 5,580 (9.44) 4,442 (8.65) ** (0-20, 0-60) 31,002 (2.89) 25,015 (7.33) 17,515 (1.27) ** ** ** ** Methylfuran 632 (9.05) 476 (7.97) 341 (9.35) * (0-20,20-60),** (0-60) 1,963 (2.03) 1,594 (6.57) 1,124 (7.27) ** ** ** ** Butanone 3,759 (7.25) 3,744 (8.61) 3,677 (9.80) 11,165 (3.48) 11,611 (3.28) 11,681 (2.01) ** ** ** Methylbutanal 10,230 (9.70) 8,458 (9.11) 8,130 (5.61) * (0-60) 45,171 (3.61) 45,196 (9.14) 38,062 (1.14) * (0-60, 20-60) ** ** ** Methylbutanal 10,356 (8.75) 8,734 (8.94) 8,264 (5.05) * (0-60) 43,298 (1.28) 42,706 (9.26) 36,351 (3.33) * (0-60, 20-60) ** ** ** ,5-Dimethylfuran 1,776 (8.46) 587 (9.13) 421 (6.23) ** (0-20, 0-60) 5,858 (4.20) 3,305 (6.57) 1,945 (4.90) ** ** ** ** ,3-Butanedione 5,324 (9.26) 5,229 (9.39) 5,296 (7.24) 12,156 (0.65) 12,651 (2.17) 12,713 (0.28) * (0-20, 0-60) * ** ** ,3-Pentanedione 10,979 (9.83) 9,676 (5.22) 10,071 (9.11) 28,507 (5.77) 29,280 (7.19) 26,595 (2.75) ** ** ** ,3-Hexanedione 3,647 (9.23) 2,827 (7.82) 2,566 (9.36) * (0-20), ** (0-60) 11,555 (8.92) 12,421 (3.53) 9,836 (5.37) * (20-60) ** ** ** Methyl-1H -pyrrole 9,512 (5.57) 7,476 (5.91) 7,435 (8.39) ** (0-20, 0-60) 41,019 (1.57) 39,162 (3.88) 30,436 (1.26) ** (0-60, 20-60) ** ** ** Ethenyl-5-methylfuran 1,796 (6.84) 445 (9.40) 271 (8.01) ** (0-20, 0-60) 5,528 (9.63) 2,103 (7.03) 834 (5.69) ** ** ** ** Ethyl-1H -pyrrole 1,420 (7.02) 922 (7.54) 794 (7.32) ** (0-20, 0-60) 6,378 (0.64) 5,357 (3.72) 3,453 (4.10) ** ** ** ** Pyridine 64,203 (9.41) 82,131 (5.00) 84,372 (3.69) ** (0-20, 0-60) 110,217 (1.44) 164,634 (3.45) 159,758 (6.49) **(0-20, 0-60) ** ** Pyrazine 2,351 (1.35) 2,256 (3.57) 2,388 (1.86) 9,071 (5.15) 7,919 (3.66) 6,507 (5.65) * (0-20), ** (0-60, 20-60) ** ** ** (Methoxymethyl)furan 3,582 (7.57) 3,081 (8.48) 2,910 (9.87) 11,242 (6.15) 11,226 (5.01) 8,965 (4.14) ** (0-60, 20-60) ** ** ** Dihydro-2-methyl-3(2H )-furanone 7,297 (9.42) 7,374 (8.48) 7,689 (4.65) 14,123 (6.02) 14,899 (4.97) 14,037 (5.33) Methylpyrazine 21,357 (8.45) 23,450 (9.75) 24,459 (1.87) 28,171 (4.87) 31,726 (5.59) 29,458 (6.75) * ** ** ,5-Dimethylpyrazine 9,087 (4.22) 9,210 (9.70) 9,258 (3.23) 14,179 (5.68) 15,754 (4.63) 13,473 (8.61) ** ** ** ,6-Dimethylpyrazine 10,437 (4.51) 10,674 (9.34) 10,747 (3.21) 15,726 (6.14) 17,385 (4.35) 15,059 (9.21) ** ** ** Ethylpyrazine 10,058 (4.57) 9,918 (9.51) 9,802 (3.58) 19,508 (7.12) 20,579 (9.90) 17,335 (6.77) * Ethyl-6-methylpyrazine 9,422 (4.18) 9,170 (7.94) 8,477 (4.13) 14,047 (9.34) 14,846 (4.68) 11,533 (7.66) * (0-60, 20-60) ** ** ** Ethyl-5-methylpyrazine 7,071 (6.13) 6,483 (7.52) 5,881 (3.77) * (0-60) 12,211 (9.47) 11,301 (9.51) 9,316 (7.20) * (0-60) * ** Trimethylpyrazine 6,402 (4.49) 6,208 (9.18) 5,704 (3.65) 8,741 (9.16) 9,317 (5.44) 7,338 (7.49) * (20-60) ** ** ** Ethyl-2,5-dimethylpyrazine 6,046 (5.88) 5,591 (8.71) 4,393 (4.49) * (20-60), ** (0-60) 8,842 (9.40) 8,649 (5.64) 6,105 (9.68) ** (0-60, 20-60) ** ** ** Furancarboxaldehyde 44,539 (6.25) 45,334 (7.77) 45,469 (2.45) 83,525 (6.96) 88,981 (1.00) 80,029 (3.97) ** ** ** [(Methylthio)methyl]furan 8,451 (5.99) 4,863 (8.72) 3,042 (9.34) ** 29,478 (9.19) 18,305 (2.56) 8,357 (8.34) ** ** ** ** Furfuryl formate 21,983 (7.81) 20,096 (8.78) 17,946 (5.77) * (0-60) 56,772 (7.09) 57,594 (0.96) 43,647 (7.47) ** (0-60, 20-60) ** ** ** (2-Furanyl)-ethanone 15,790 (8.29) 15,845 (7.23) 14,580 (5.28) 28,357 (9.09) 29,317 (4.11) 25,169 (4.06) ** ** ** H -Pyrrole 7,863 (7.31) 7,790 (6.80) 7,744 (1.86) 20,139 (0.41) 20,769 (1.53) 18,843 (5.68) * (20-60) * ** ** Furfuryl acetate 104,067 (6.42) 92,219 (8.51) 76,306 (7.39) ** (0-60) 216,583 (6.18) 210,063 (1.86) 151,882 (7.77) ** (0-60, 20-60) ** Methyl-2-furancarboxaldehyde 31,997 (4.34) 32,167 (9.35) 30,327 (4.25) 59,680 (9.92) 63,947 (2.65) 51,021 (7.39) * (20-60) * ** * ,2'-Methylenebisfuran e 6,042 (5.68) 1,874 (8.88) 981 (6.82) ** 23,708 (5.94) 6,720 (8.96) 2,325 (4.40) ** ** ** ** Methyl-1H -pyrrole-2-carboxaldehyde 7,379 (4.22) 6,841 (9.67) 6,259 (0.76) * (0-60) 14,603 (5.20) 14,343 (4.55) 10,962 (9.28) ** (0-60, 20-60) Furanmethanol 85,893 (7.50) 89,519 (7.92) 89,738 (2.94) 88,357 (6.50) 97,591 (2.68) 88,258 (4.47) ** ** ** (2-Furanylmethyl)-5-methylfuran e 2,311 (0.35) 599 (7.85) 223 (9.04) ** 7,068 (8.19) 1,386 (6.68) 488 (8.66) * (20-60), ** (0-20, 0-60) ** ,3-Dihydro-6-methylthieno[2,3c ]furan 5,864 (1.41) 2,499 (9.14) 1,335 (8.34) ** 14,322 (7.30) 5,029 (7.16) 1,932 (6.55) ** ** * (2-Furanylmethyl)-1H -pyrrole 11,579 (2.77) 5,578 (9.08) 3,071 (8.38) ** 27,048 (5.31) 11,171 (7.17) 4,274 (9.68) ** * ** Methoxyphenol 4,535 (6.34) 3,983 (7.21) 3,140 (8.21) * (20-60), ** (0-60) 8,377 (8.96) 7,047 (8.72) 4,550 (8.26) ** (0-60, 20-60) * Ethenyl-2-methoxyphenol 3,565 (2.45) 2,315 (8.54) 1,342 (8.60) ** 3,540 (8.65) 2,267 (9.31) 947 (7.49) ** ** ** ** 592,121 (6.24) 564,477 (7.70) 532,498 (3.77) 1,202,102 (5.03) 1,204,612 (2.23) 993,653 (4.41) ** (0-60, 20-60) a Peak s are average values of three measurements. Fractions are considered to be within acceptable limits as determined by integration and rounded off. b Relative standard deviation. c Significant differences (** p < 0.01) among three ratios of added milk, and significant differences (* (0-20) p < 0.05, ** (0-20) p < 0.01) between milk 0% and milk 20% (w/w), (* (20-60) p < 0.05, ** (20-60) p < 0.01) between milk 20% (w/w) and milk 60% (w/w), and (* (0-60) p < 0.05, ** (0-60) p < 0.01) between milk 0% and milk 60% (w/w) by Tukey HSD multiple comparison tests. d Significant differences (* p < 0.05, ** p < 0.01) between two sampling methods by t tests. Gray background: peak- (PA) ratio of RAS versus total PA > PA ratio of headspace. e Tentative identifications obtained by comparing mass spectra and retention indices with data in the literature (Baltes and Bochmann, 1987a).

8 240 Influence of milk addition on headspace and RAS odorants of freshly brewed espresso The SPME methods for sampling the headspace and RAS volatiles were applied to GC/O analyses in order to investigate potent odorants, and their odor activities and characteristics in freshly brewed espresso without milk (milk 0%) or with milk (20% or 60% (w/w)). Thirty-three and 36 odorants were detected in the headspace and RAS volatiles, respectively. The charm values (Acree et al., 1984) and OSVs (Acree, 1997) analyzed under the same GC/O conditions are listed in Tables 6 and 7, respectively. m. akiyama et al. The influence of milk addition (20% and 60% (w/w)) on each odorant was investigated. Changes in each charm value for the 33 potent headspace odorants detected by GC/O analysis with milk addition were statistically analyzed by a Tukey HSD multiple comparison test. Significant differences (* p < 0.05, ** p < 0.01) in odor activities of the odorants of espresso with milk in 0%, 20%, and 60% (w/w) were observed in 16 compounds and shown in Table 6. The 16 compounds in headspace odorants showed a significant decrease in charm values when milk (20% and 60% (w/w)) was added to espresso. Similarly, total charm values and charm values Table 6. Potent odorants found in the headspace volatiles of freshly brewed espresso (milk 0%) and café latte (milk 20% (w/w) and milk 60% (w/w)) using the solid-phase microextraction sampling method. No. Description Component Retention Charm values a (OSVs b ) indices Milk 0% Milk 20% (w/w) Milk 60% (w/w) Effect of milk c 1 Acidic 3-Methylbutyric acid (22) 0 (0) 0 (0) ** (0-20, 0-60) Total acidic ** (0-20, 0-60) 2-1 Buttery-oily 2-Methylpropanal (22) 31 (17) 43 (26) * (0-60), ** (0-20) and 3-Methylbutanals (50) 170 (39) 150 (49) -3 2,3-Butanedione (40) 190 (42) 180 (54) -4 2,3-Pentanedione (13) 20 (13) 17 (17) -5 (E )-2-Nonenal (21) 16 (12) 3 (7) ** (0-20, 0-60) Total buttery-oily Green-blackcurrant 3-Mercapto-3-methylbutyl formate (59) 170 (39) 88 (38) * (0-60) Total green-blackcurrant * (0-60) 4-1 Green-earthy 2-Methoxy-3-(1-methylethyl)pyrazine (25) 36 (18) 20 (18) -2 2-Methoxy-3-(1-methylpropyl)pyrazine (33) 53 (22) 21 (18) * (0-20, ** 0-60) Total green-earthy * (0-60) 5-1 Nutty-roast 1-(3,4-Dihydro-2H -pyrrol-5-yl)-ethanone (0) 0 (0) 9 (12) -2 2-Ethyl-3,5-dimethylpyrazine (53) 250 (48) 200 (57) ** (0-60) -3 2,3-Diethyl-5-methylpyrazine (47) 120 (33) 82 (36) * (0-20), ** (0-60) -4 Unknown (28) 79 (27) 53 (29) -5 Unknown (22) 26 (15) 10 (13) * (0-20), ** (0-60) Total nutty-roast ** (0-20, 0-60) 6-1 Phenolic 2-Methoxyphenol (90) 520 (69) 410 (81) * (0-60) -2 4-Ethyl-2-methoxyphenol (52) 300 (52) 170 (52) ** (0-60, 20-60) -3 4-Ethenyl-2-methoxyphenol (6) 0 (0) 0 (0) -4 3-Methyl-1H -indole (54) 130 (34) 45 (27) ** (0-20, 0-60) Total phenolic * (0-20), ** (0-60) 7-1 Smoke-roast 3-Methyl-2-butene-1-thiol (83) 290 (51) 210 (58) -2 2-Furanmethanethiol (100) 830 (87) 620 (100) * (0-60) -3 2-((Methylthio)methyl)furan (40) 56 (23) 16 (16) ** (0-20, 0-60) -4 3-Mercapto-3-methyl-1-butanol (6) 0 (0) 0 (0) -5 Unknown (6) 0 (0) 0 (0) Total smoke-roast * (0-60) 8 Soy sauce 3-(Methylthio)propanal (39) 140 (36) 110 (42) Total soy sauce Sweet-caramel 2-Hydroxy-3-methyl-2-cyclopenten-1-one (5) 0 (0) 0 (0) -2 4-Hydroxy-2,5-dimethyl-3(2H )-furanone (62) 320 (54) 330 (73) -3 2-Ethyl-4-hydroxy-5-methyl-3(2H )-furanone (29) 64 (24) 63 (32) -4 3-Hydroxy-4,5-dimethyl-2(5H )-furanone (96) 1100 (100) 510 (91) Total sweet-caramel * (0-60) 10-1 Sweet-fruity Linalool (8) 0 (0) 0 (0) -2 Benzeneacetaldehyde (20) 25 (15) 10 (13) * (0-20), ** (0-60) -3 (E )-Beta-damascenone (56) 150 (37) 63 (32) ** (0-20, 0-60) Total sweet-fruity ** (0-20, 0-60) 11-1 Others 1-Octen-3-one (11) 3 (5) 2 (6) * (0-20, 0-60) -2 Dimethyl trisulfide (9) 0 (0) 2 (6) Total others Total * (0-20), ** (0-60) a Each charm value is rounded off to two significant figures to reflect the actual resolution of the dilution analysis, and represented as an average of three measurements. b Odor spectrum value (OSV) is the normalized charm value modified with an approximate Stevens law exponent (n = 0.5). c Significant differences (* (0-20) p < 0.05, ** (0-20) p < 0.01) between milk 0% and 20% (w/w), (** (20-60) p < 0.01) between milk 20% (w/w) and 60% (w/w), (* (0-60) p < 0.05, ** (0-60) p < 0.01) between milk 0% and 60% (w/w) and (** p < 0.01 ) between milk 0% and 20% (w/w), 0% and 60% (w/w), and 20% (w/w) and 60% (w/w) by a Tukey HSD multiple comparison test.

9 Retronasal Aroma of Espresso 241 Table 7. Potent odorants found in the RAS volatiles of freshly brewed espresso (milk 0%) and café latte (milk 20% (w/w) and milk 60% (w/w)) using the solid-phase microextraction sampling method. No. Description Component Retention indices Charm values a (OSVs b ) Milk 0% Milk 20% (w/w) Milk 60% (w/w) Effect of milk c 1 Acidic 3-Methylbutyric acid (19) 0 (0) 0 (0) * (0-20, 0-60) Total acidic * (0-20, 0-60) 2-1 Buttery-oily 2-Methylpropanal (41) 100 (28) 100 (32) ** (0-20, 0-60) and 3-Methylbutanals (64) 430 (58) 330 (58) ** (0-20, 0-60) -3 2,3-Butanedione (62) 390 (55) 350 (60) ** (0-20, 0-60) -4 2,3-Pentanedione (27) 76 (24) 53 (23) * (0-60) -5 (Z )-2-Nonenal d (16) 1 (3) 0 (0) * (0-60) -6 (E )-2-Nonenal (29) 28 (15) 8 (9) ** (0-20, 0-60) Total buttery-oily ** (0-20, 0-60) 3 Green-blackcurrant 3-Mercapto-3-methylbutyl formate (63) 170 (36) 130 (36) * (0-20, 0-60) Total green-blackcurrant * (0-20, 0-60) 4-1 Green-earthy 2-Methoxy-3-(1-methylethyl)pyrazine (39) 98 (27) 13 (12) ** -2 2-Methoxy-3-(1-methylpropyl)pyrazine (39) 93 (27) 42 (21) ** (0-20, 0-60) Total green-earthy ** 5-1 Nutty-roast Unknown (10) 5 (6) 0 (0) * (0-20), ** (0-60) -2 1-(3,4-Dihydro-2H -pyrrol-5-yl)-ethanone (0) 1 (3) 7 (8) -3 2-Ethyl-3,5-dimethylpyrazine (60) 250 (44) 220 (47) * (0-20, 0-60) -4 2,3-Diethyl-5-methylpyrazine (45) 160 (35) 97 (31) * (0-20), ** (0-60) -5 Unknown (27) 100 (28) 81 (29) -6 Unknown (21) 34 (16) 21 (15) ** (0-20, 0-60) Total nutty-roast * (0-20), ** (0-60) 6-1 Phenolic 2-Methoxyphenol (100) 1300 (100) 980 (100) -2 4-Ethyl-2-methoxyphenol (78) 470 (60) 480 (70) -3 4-Ethenyl-2-methoxyphenol (10) 0 (0) 0 (0) -4 3-Methyl-1H -indole (29) 32 (16) 5 (7) * (0-60) Total phenolic Smoke-roast 3-Methyl-2-butene-1-thiol (100) 510 (63) 190 (44) * (20-60), ** (0-20, 0-60) -2 2-Furanmethanethiol (93) 760 (76) 700 (85) * (0-20, 0-60) -3 2-((Methylthio)methyl)furan (38) 34 (16) 14 (12) ** (0-20, 0-60) -4 Unknown (8) 0 (0) 0 (0) -5 3-Mercapto-3-methyl-1-butanol (15) 0 (0) 0 (0) ** (0-20, 0-60) -6 Unknown (18) 8 (8) 0 (0) * (0-60) Total smoke-roast ** (0-20, 0-60) 8 Soy sauce 3-(Methylthio)propanal (45) 220 (41) 170 (42) Total soy sauce Sweet-caramel 2-Hydroxy-3-methyl-2-cyclopenten-1-one (0) 0 (0) 2 (5) -2 4-Hydroxy-2,5-dimethyl-3(2H )-furanone (32) 130 (32) 87 (30) -3 2-Ethyl-4-hydroxy-5-methyl-3(2H )-furanone (19) 29 (15) 16 (13) -4 3-Hydroxy-4,5-dimethyl-2(5H )-furanone (72) 640 (70) 350 (60) Total sweet-caramel Sweet-fruity Linalool (14) 3 (5) 0 (0) * (0-20, 0-60) -2 Benzeneacetaldehyde (28) 57 (21) 27 (17) ** -3 (E )-Beta-damascenone (84) 340 (51) 120 (35) * (0-20), ** (0-60) Total sweet-fruity * (0-20), ** (0-60) 11-1 Others 1-Octen-3-one (13) 8 (8) 1 (3) * (20-60), ** (0-20, 0-60) -2 Dimethyl trisulfide (26) 40 (18) 29 (17) * (0-20, 0-60) Total others * (0-20), ** (0-60) Total ** (0-20, 0-60) a Each charm value is rounded off to two significant figures to reflect the actual resolution of the dilution analysis, and represented as an average of three measurements. b Odor spectrum value (OSV) is the normalized charm value modified with an approximate Stevens law exponent (n = 0.5). c Significant differences (* (0-20) p < 0.05, ** (0-20) p < 0.01) between milk 0% and 20% (w/w), (** (20-60) p < 0.01) between milk 20% (w/w) and 60% (w/w), and (* (0-60) p < 0.05, ** (0-60) p < 0.01) between milk 0% and 60% (w/w) by a Tukey HSD multiple comparison test. d Tentative identifications obtained by comparing aroma properties and retention indices with data in the literature (Holscher et al., 1990). from several potent odorants in Table 6, such as no methoxy-3-(1-methylpropyl)pyrazine (green-earthy odor); no ethyl-3,5-dimethylpyrazine (nutty-roast odor); no ,3-diethyl-5-methylpyrazine (nutty-roast odor); no ethyl-2-methoxyphenol (phenolic odor); no methyl- 1H-indole (phenolic odor); no ((methylthio)methyl)fu ran (smoke-roast odor); and no (E)-beta-damascenone (sweet-fruity odor) all showed significant differences (p < 0.01) after the addition of milk. In the case of RAS odorants, significant differences in the charm values were also observed between milk content of 0% and 20% (w/w), 0% and 60% (w/w), or 20% and 60% (w/ w) in 25 odorants that were shown to be significantly different (* p < 0.05, ** p < 0.01) in Table 7. Twenty-five of a total of 36 RAS odorants detected by GC/O analysis exhibited a significant decrease in the charm values after the addition of

10 242 m. akiyama et al. milk, indicating that the influence of milk addition on RAS aroma was similar to that observed with headspace aroma. However, in addition to the headspace odorants showing significant differences, some highly-volatile odorants of the other RAS odorants, such as no and 3-methylbutanals (buttery-oily odor); no ,3-butanedione (buttery-oily odor); no methoxy-3-(1-methylethyl)pyrazine (greenearthy odor); and no methyl-2-butene-1-thiol (smokeroast odor) shown in Table 7 also showed significant decrease (p < 0.01) in the charm values when milk was added. Influence of milk addition on headspace and RAS odor descriptions of freshly brewed espresso The odorants of espresso, espresso with milk (20% (w/w)), and espresso with milk (60% (w/w)), detected by GC/O were classified into the 11 odor descriptions listed in Tables 6 and 7. The influence of milk addition on each odor description of the headspace aroma and RAS aroma for espresso could be characterized as follows: the total charm values of each odor description showed a tendency to decrease in all odor descriptions when milk was added to espresso. In particular, significant decreases of the charm values were observed in acidic, nuttyroast, phenolic, and sweet-fruity odors (p < 0.01), and greenblackcurrant, green-earthy, smoke-roast, and sweet-caramel odors (p < 0.05) of the headspace aroma (Table 6). In the case of the RAS aroma, charm values of buttery-oily, greenearthy, nutty-roast, smoke-roast, and sweet-fruity odors (p < 0.01), and acidic and green-blackcurrant odors (p < 0.05) decreased significantly when milk was added (Table 7). Comparison of the headspace odorants and RAS odorants of freshly brewed espresso and café latte Headspace aroma and RAS aroma were compared by the relative activities (OSVs) of the odorants detected by GC/O analysis (Tables 6 and 7). Several potent odorants (OSV above 30): no methyl-1H-indole (phenolic odor), no hydroxy- 2,5-dimethyl-3(2H)-furanone (sweet-caramel odor), and no hydroxy-4,5-dimethyl-2(5H)-furanone (sweet-caramel odor) showed larger OSVs in the headspace aroma of espresso (milk 0%) than in the RAS aroma (Tables 6 and 7). On the other hand, no and 3-methylbutanals (buttery-oily odor), no ,3-butanedione (buttery-oily odor), no ethyl-2-methoxyphenol (phenolic odor), and no (E)- beta-damascenone (sweet-fruity odor) had larger OSVs in the RAS aroma of espresso (milk 0%) than in the headspace aroma. When milk (20% or 60% (w/w)) was added to espresso, no methyl-1H-indole (phenolic odor), no hydroxy-2,5-dimethyl-3(2H)-furanone (sweet-caramel odor), and no hydroxy-4,5-dimethyl-2(5H)-furanone (sweetcaramel odor) showed larger OSVs in headspace aromas of the café lattes than in the RAS aromas. Whereas, no methoxyphenol (phenolic odor) had larger OSVs in the RAS aromas of the café lattes than in the headspace aromas (Tables 6 and 7). Comparison of GC/O profiles of headspace aroma and RAS aroma from espresso and café latte The odorants detected by GC/O were classified into 11 odor descriptions (Tables 6 and 7). The relative charm value of each description to total charm values was shown in Fig. 2. From the GC/ O results of headspace and RAS aromas of freshly brewed espresso (milk 0%), the relative charm values of buttery-oily and sweet-fruity odors of headspace aroma were smaller than those in the RAS aroma, while the relative charm value of Nutty-roast Green-earthy Green-blackcurrant Phenolic Buttery-oily Smoke-roast Soy sauce Sweet-caramel Sweet-fruity Milk 60% (w/w) Headspace RAS Headspace Milk 0% Acidic Others RAS % 20% 40% 60% 80% 100% (Charm value / total charm value) 100 Fig. 2. Comparison of the headspace aroma and RAS aroma profiles from espresso (milk 0%) and café latte (milk 60% (w/w)).