Food Chemistry 291 (2019) 49–58

Contents lists available at ScienceDirect

Food Chemistry
journal homepage: www.elsevier.com/locate/foodchem

Impact of drying process on chemical composition and key aroma T

components of Arabica coffee
Fareeya Kulapichitra, Chaleeda Borompichaichartkula, Inthawoot Suppavorasatita,
Keith R. Cadwalladerb,
⁎

a
Department of Food Technology, Faculty of Science, Chulalongkorn University, Bangkok 10330, Thailand
b
Department of Food Science and Human Nutrition, University of Illinois at Champaign-Urbana, 1302 West Pennsylvania Avenue, Urbana, IL 61801, USA

ARTICLE INFO ABSTRACT

Keywords: Influence of heat pump drying (HP at 40, 45 and 50 °C), tray drying (TD) and sun drying (SD) on the quality of
Coffee Arabica coffee was evaluated. Drying process did not affect the caffeine content, but influenced levels of some
Flavor amino acids. Sucrose content was higher in HP and TD than in SD green coffees. The perceived aroma of brewed
Drying coffee from SD was similar to HP, but differed from TD. Concentrations of 30 important odorants were compared
Chemical composition
for SD, HP (50 °C) and TD brewed coffees. 2-Furfurylthiol, a key odorant of coffee, was at the same level in SD
Sensory evaluation
and HP coffees and lowest in TD samples. Principal component analysis (PCA) separated SD from HP and TD,
Stable isotope dilution analysis
based on the concentrations of 23 odorants. Combined results of sensory and chemical analyses showed that in
comparison to SD, HP was superior to TD for preserving overall flavor quality.

1. Introduction The choice of the drying technique used is dependent on economic

Globally, coffee is the second most traded commodity after petro-
leum. Total world coffee consumption was estimated at > 9 million
tons in 2015–2016 with a consign close to US$21 billion (ICO, 2017).
Much of the increase in demand can be attributed to the coffee’s flavor
and other desirable effects associated with its consumption (De Melo
Pereira et al., 2019; Dong, Hu, Chu, Zhao, & Tan, 2017, 2019). Among
the various types of coffees, Arabica coffee (Coffea arabica) dominates
the world coffee market (70–80% market share) (Belitz, Grosch, &
Schieberle, 2009) and is preferred over other coffee species because of
its superior sensory properties (Belitz et al., 2009).
Processing methods and associated variables depend on the geo-
graphic location and the climate where coffee is produced and are
important factors in determining coffee quality; however, the influence
of these factors on the final quality of coffee is still not well understood
(De Melo Pereira et al., 2019). Among the many processing operations,
drying is one of the most important post-harvest processing steps that
should to be considered, since coffee subjected to any processing
method must be dried to a final water content of < 12% to inhibit
microbial spoilage and prevent chemical deterioration during storage
and distribution (Belitz et al., 2009; De Melo Pereira et al., 2019; Dong
et al., 2017; Flament, 2002).
factors and/or the type of processing methods employed, which further
influences the physicochemical properties of green coffee beans and
subsequently the crop quality. Coffee drying can be performed by sun
drying (SD) or by using various mechanical drying techniques (De Melo
Pereira et al., 2019).
Sun drying (SD) is commonly used by the industry; however, there
are some drawbacks, such as prolonged drying time, high labor cost and
the requirement for a large surface area for drying. Even with these
drawbacks, SD is usually preferred over mechanical drying since SD
tends to produce roasted coffees with superior sensorial properties. This
is due in part to the higher contents of sugars formed by pre-germina-
tion of the seed embryo during SD and because mechanical drying at
high temperatures may adversely affect the structure of green coffee
and overall final quality of coffee (Borem, Marques, & Alves, 2008;
Dong et al., 2017). Despite the potential impact of drying method, there
is a general lack of comprehensive studies related to how various drying
procedures affect coffee quality and, in particular, final brewed coffee
flavor.
Heat pump drying (HP) has been widely reported as an energy ef-
ficient process for drying operations (Dong et al., 2017, 2019; Goh,
Othman, Mat, Ruslan, & Sopian, 2011). The closed drying system of HP
is comprised of efficient heating and cooling systems, which enable

Corresponding author.
⁎

E-mail addresses: fareeya.k@student.chula.ac.th (F. Kulapichitr), Chaleeda.B@chula.ac.th (C. Borompichaichartkul), Inthawoot.S@chula.ac.th (I. Suppavorasatit),
cadwlldr@illinois.edu (K.R. Cadwallader).

https://doi.org/10.1016/j.foodchem.2019.03.152
Received 19 November 2018; Received in revised form 26 March 2019; Accepted 31 March 2019
Available online 01 April 2019
0308-8146/ © 2019 Elsevier Ltd. All rights reserved.
F. Kulapichitr, et al.
accurate control of the temperature (−20–100 °C) and relative hu-
midity (15–80%) of the drying gas, with a potential for heat recovery.
Due to several benefits of heat pump drying (HP), it is worthwhile to
investigate its potential as an alternative to SD in coffee processing. To
our knowledge, there are no published studies on the effect of HP on the
composition and flavor quality of Arabica roasted coffee. However, a
comprehensive study was recently conducted on the effects of various
drying methods on the bioactive components, fatty acid composition,
and volatile profiles of green Robusta coffee and on the volatile profiles
and taste properties of roasted Robusta coffee, by use of electronic nose,
electronic tongue, and gas chromatography-mass spectrometry
(GC–MS) (Dong et al., 2017, 2019); however, that study did not eval-
uate the effect of drying on the character impact odorants and the
overall perceived aroma of the roasted coffee brews.
The sensory properties and the chemical composition are important
quality markers of coffee, which can be influenced by the various coffee
processing methods (De Melo Pereira et al., 2019). Since these markers
can be influenced by the drying process, it is worthwhile to study how
HP, as a viable alternative to SD, affects them. Therefore, the present
study was aimed at investigating the impact of three different drying
methods on the selected chemical constituents of Arabica coffee, with
an emphasis on comparison of selected key aroma compounds.
2. Materials and methods
2.1. Raw materials
Arabica coffee cherries (Coffea arabica L. cv. Catimor) were collected
from Doi Chang, Chiang Rai Province, Thailand, during the November
2016 harvest. All drying processes were performed at the Faculty of
Engineering, Mahasarakham University, Thailand. The wet parchment
Arabica coffee beans, processed using a wet process method, were dried
in a heat pump dryer at either 40, 45 or 50 °C ( ± 1 °C) for 23, 16 or
11 h, respectively, with an air velocity of 0.6 m/s and a constant re-
lative humidity of 70%, until a moisture content of 10 ± 1% (w.b.)
was reached, as recommended by local coffee producers to prevent
deterioration, due to the high humidity in this region of Thailand
during storage of the green coffee parchment before exporting, dehul-
ling and roasting. The HP temperature range (40–50 °C) was chosen
since it was previously shown to preserve the bioactive contents and
aroma precursors in nuts and coffee (Dong et al., 2017, 2019; Goh et al.,
2011).
For tray drying (TD), parchment coffee beans were uniformly dis-
tributed over stainless steel trays (53 × 72 × 3 cm) in a 0.5-cm-thick
layer and dried at 50 ( ± 0.5) °C for 18 h in an electric convection hot-
air dryer (Kluay Nam Thai, Bangkok, Thailand) until a final moisture
content of 10 ± 1% (w.b.) was reached.
The SD drying process was done in accordance with customary
practices of local coffee producers in the region. SD dried coffee was
prepared by uniformly distributing the wet parchment coffee beans on a
fabric. Drying was performed in a ventilated open area, which allowed
natural air flow to dry the samples at an ambient temperature of
20–30 °C and approximately 70% relative humidity for 3 days until a
final moisture content of 10 ± 1% (w.b.) was reached.
For comparison purposes, the above drying treatments were con-
ducted using coffee beans obtained from the same lot and plantation.
The green coffee beans subjected to the above drying treatments, all
had moisture contents of 10 ± 1% (w.b.) and water activities of < 0.6
(Table 2). The dried coffee parchment beans were vacuum-packed in
laminated plastic pouches and stored at 4 °C prior to roasting.
To obtain green coffee beans, silverskins and hulls were gently re-
moved from parchment beans using a porcelain mortar and pestle. To
prepare green coffee powders for chemical analysis, green coffee beans
were ground for 1 min in a hi-speed blender equipped with a stainless
steel jar (model 7011HS; Waring Commercial, New Hartford, CT).
Green coffee powders from each treatment were kept in sealed
50
Food Chemistry 291 (2019) 49–58
laminated plastic pouches and stored at −40 °C prior to further ana-
lysis.
The green coffee samples (100 g whole beans) from each drying
treatment were roasted using a single batch coffee roaster (Sample Pro-
G2; Nexu International, Ltd, Hong Kong) for 15 min at 200 °C, in order
to prepare a medium roasted coffee. Once the roasting process was
terminated, cool air was immediately blown into the chamber for 4 min
to cool the roasted beans.
The roasted coffee beans were then removed from the roaster and
placed in a metal container to further cool the beans to room tem-
perature. The roasted samples were then vacuum-packed and stored at
4 °C in sealed aluminum-clad bags equipped with CO2 degassing valves.
Prior to analysis, samples were brought to room temperature and al-
lowed to reach equilibrium.
2.2. Determination of moisture contents and water activity in green coffee
Moisture contents and water activities (aw) of green coffee beans
were determined in triplicate. For moisture determination, approxi-
mately 10 g of whole green coffee beans were dried for 16 ± 0.5 h, via
an electrically heated drying oven with forced air ventilation at 105 °C.
Water activity (aw) was measured using an instrument based on the
chilled mirror dew point technique (Series 3 TE; Aqualab, Pullman,
WA).
2.3. Determination of amino acids
Total amino acids of green coffee were determined as previous de-
scribed (Kunarayakul, Thaiphanit, Anprung, & Suppvorasatit, 2018)
using a Waters Alliance 2695 series HPLC system (Waters Corp., Mil-
ford, MA) equipped with a column heater and Jasco fp-2020 fluores-
cence detector (Jasco Co., Tokyo, Japan; λEx 250 nm/λEm 395 nm). A
Hypersil Gold C18 column (4.6 mm × 150 mm, 3 μm; ThermoFisher
Scientific, Winsford, Cheshire, UK) set at 35 °C was used to separate the
analytes. For sample preparation, green coffee powder (100 mg) was
hydrolyzed (5 mL of 6 N HCl, 110 °C, 22 h), then derivatized using
AccQ-fluor reagent (Waters Corp.) at 55 °C for 10 min. An aliquot
(10 µL) of each sample was analyzed by HPLC using sodium acetate
buffer (pH 4.9) and 60% acetonitrile as eluents A and B. Amino acids
were identified by comparison of their retention times with those of a
mixed amino acid standard (Sigma-Aldrich, St. Louis, MO). Quantita-
tive analysis was performed using external standard calibrations of
peak areas versus concentrations of authentic standards. Analyses were
performed in triplicate.
2.4. Determination of sugars
The determination of total sugars in green coffees was performed
according to AOAC method 982.14 (AOAC, 2000). Finely ground green
coffee powder (2 g) was extracted twice with 50 mL of petroleum ether
to remove fat prior to analysis. Residual petroleum ether was removed
with a gentle stream of N2, then a volume (100 mL) of methanol/
deionized water (1:1) was added and the mixture weighed (original
weight). The mixture was incubated at 80 °C for 25 min, stirred, then
cooled to room temperature and alcohol was added to bring the mixture
back to its original weight. The extract was filtered through a 0.45-µm
nylon syringe filter and an aliquot (10 µL) was injected into a Zorbax
NH2 column (250 mm × 4.6 mm id., 5 µm; Agilent Technologies Inc.,
Palo Alto, CA) connected to an Agilent 1100 series HPLC equipped with
refractive index detector (Agilent Technologies Inc.). The flow rate of
mobile phase, consisting of acetonitrile:deionized water (70:30 v/v),
was 1.5 mL/min. Sugars were quantitated based on external standard
calibrations of peak areas versus concentrations for authentic standards
(fructose, glucose, and sucrose) (Sigma-Aldrich). Analyses were per-
formed in duplicate.
F. Kulapichitr, et al.
2.5. Determination of caffeine
Caffeine was extracted from green coffee powder samples using hot
water according to AOAC (1995). Carrez I and II solutions (Sigma-Al-
drich) were added to clarify the extract before it was filtered (no. 1
filter paper and then 0.45 μm filter cartridge). An aliquot (20 µL) of the
extract was analysed using a 1100 series HPLC equipped with a diode-
array detector (Agilent Technologies Inc.) and a Zorbax-Eclipse-XDB-
C18 column (4.6 mm × 250 mm, 5 μm; Agilent Technologies Inc.). Se-
parations were performed using an isocratic mobile phase of methanol:
water: acetic acid (80:19:1 v/v/v) and peaks were monitored at a wa-
velength 276 nm. Caffeine was quantitated based on external standard
calibration of peak areas versus concentration of caffeine (Sigma-Al-
drich). Analyses were performed in duplicate.
2.6. Sensory evaluation – R-index by ranking
Human subject protocol number 17,658 was approved by the
University of Illinois at Urbana-Champaign Institutional Review Board
(IRB) prior to conducting sensory analysis. An R-index by ranking test
was conducted to determine if aroma differences existed among roasted
coffee brews prepared from green coffee beans dried by the different
drying processes, as well as to identify representative coffee samples for
quantitative analysis of selected aroma compounds. Brewed coffees
were prepared from roasted coffee beans from each drying treatment
and then evaluated with respect to their overall aroma similarities
compared to a benchmark product (control), consisting of the brewed
coffees prepared from the roasted SD dried coffee beans. R-index by
similarity ranking test was applied for this comparison (Lorjaroenphon,
Cadwallader, Kim, & Lee, 2008).
Roasted coffee beans were brought to room temperature (∼25 °C)
and then transferred to a blender jar (4 oz., The Ball Brothers Glass
Manufacturing Co., Broomfield, CO), which was vacuum sealed to
minimize evaporation of volatiles within the bean shells. Beans were
frozen at −70 °C for 1 h and then ground using a Hamilton Beach®
Fresh-Grind Coffee Grinder (Southern Pines, NC), to obtain a fine
powder with a particle size between 1.77 and 2.36 mm. Fresh brewed
coffee samples were prepared from the various ground roasted coffee
samples at the same time and on the same day of sensory analysis using
a modification of the steep cup methodology (Lyman, Benck, Dell,
Merle, & Murray-Wijelath, 2003) using a ratio of 8.5 g ground coffee to
150 mL of water, according to published recommendations (gold cup
standard; Specialty Coffee Association of America, Santa Ana, CA).
Accordingly, coffee powder (22.7 g) was brewed with 400 mL of pre-
heated odorless water (98–100 °C) for 5 min, followed by immediate
cooling of the brew in an ice bath to approximately 30 °C. Each brewed
coffee (50 mL) was then transferred to a 125-mL Teflon sniff bottle
(Nalgene PTFE wash bottle without siphon tube; Nalge Nunc Interna-
tional, Rochester, NY), which was covered with aluminum foil to avoid
assessor bias and labeled with a 3-digit code.
A set of 5 test samples (brewed coffees) was evaluated, including
reference (SD coffee), another SD coffee (blind control), the three HP
dried coffees and the TD dried coffee. Sensory evaluation was per-
formed within 4 h of completing all sample preparations. The serving
order was random and balanced. Thirty students and staff (21 females
and 9 males; 18–54 years of age) participated in this test. Panelists were
asked to gently squeeze the bottle and sniff the expressed air, and to
rank the test samples on how similar they were to the control sample.
John Brown computation, as described by O’Mahony (1992), was used
to calculate the R-index, which is based on percentage of times a sample
is ranked less similar to the control. The percentage R-index was
compared to the critical value (n = 30) for a two-tailed test at α = 0.05,
to determine significant difference from the control.
51
Food Chemistry 291 (2019) 49–58
2.7. Determination of volatile compounds
2.7.1. Chemicals
Unless otherwise stated, all reagent chemicals were obtained from
Sigma-Aldrich. Dichloromethane (DCM), methanol, hydrochloric acid
(HCl), sodium hydroxide (NaOH, anhydrous), sodium sulfate (Na2SO4,
anhydrous) were purchased from Fisher Scientific (Fair Lawn, NJ).
Florisil was purchased from US Silica Company (Berkeley Springs, WV),
Ultrahigh purity (UHP) nitrogen and UHP helium were obtained from
Airgas (Rantoul, IL). Odorless water was prepared by boiling glass-
distilled deionized water in an open flask to two thirds of its original
volume.
2.7.2. Unlabeled reference standards
All authentic reference standards including n-alkane standards
(C5–C30) were purchased from Sigma-Aldrich, except for 4-vi-
nylguaiacol which was obtained from Lancaster Synthesis (Ward Hill,
MA). 2,6-Diethylpyrazine was synthesized using a published procedure
(Fang & Cadwallader, 2013).
2.7.3. Isotopically labeled standards
(2H2)-3-Methylbutanal, (2H3)-2-isobutyl-3-methoxypyrazine, (2H3)-
guaiacol were obtained from Sigma-Aldrich. The following compounds
were synthesized according to published procedures: (2H2)-2-methyl-
propanal (Lapsongphon, Yongsawatdigul, & Cadwallader, 2015), (2H2)-
3-methylbutanoic acid (Steinhaus & Schieberle, 2005); (2H4)-β-da-
mascenone (Kotseridis, Baumes, & Skouroumounis, 1998); (2H2-3)-1-
octen-3-ol (Lin, Welti, Vera, Fay, & Blank, 1999); (2H4)-hexanal
(Steinhaus, Sinuco, Polster, Osorio, & Schieberle, 2009); (2H3)-2-me-
thylpyrazine, (2H5)-2-ethylpyrazine, (2H3)-2,3-dimethylpyrazine, (2H3)-
2,6-dimethylpyrazine, (2H10)-2,6-diethylpyrazine, 2-(2H3)-3,5-tri-
methylpyrazine, ( H3)-2,3-diethyl-5-methylpyrazine and (2H5)-3-ethyl-
2
2,5-dimethylpyrazine (Fang & Cadwallader, 2013); (2H3)-4-vi-
nylguaiacol (Scheidig, Czerny, & Schieberle, 2007); and (2H2)-2-fur-
furylthiol, (2H6)-dimethyl trisulfide and (2H3)-methional (Sen & Grosch,
1991).
2.7.4. Synthesis of labeled standards
2.7.4.1. (1,2-13C2)-2,3-Pentanedione. The target compound was
synthesized by adapting the method previous described by Schieberle
and Hoffman (1997) for (13C4)-2,3-butanedione. 3-Benzyl-5-(2-
hydroxyethyl)-4-methylthiazolium chloride (100 mg, 0.4 mmol),
0.5 mL of triethylamine and 5 mL of DCM were added to a 22-mL vial
equipped with a magnetic stir bar. After sealing the vial with a Teflon-
coated silicon septum cap, a solution of 157 mg of (13C2)-acetaldehyde
(200 μL, 4 mmol; Sigma-Aldrich) and 80.5 mg of propanal (100 μL,
2 mmol) in 1 mL of DCM was added via a syringe through the septum.
The mixture was stirred for 1 h at room temperature, after which it was
extracted with 2 M NaOH (3 × 5 mL) and then brine (2 × 5 mL). The
organic layer was dried over 2 g of anhydrous Na2SO4 and then added
to a stirred suspension of pyridinium chlorochromate (PCC, 1.1 g,
5 mmol) in 3 mL DCM. The suspension was stirred for 1.5 h at room
temperature and then passed through a 1.5 g bed of Florisil to yield a
crude mixture containing the target compound. The concentration of
the labeled diketone (MS (EI), of (13C2)-2,3-pentanedione: 45 (100%),
57 (38%), 102 (M+, 4%)) was determined by GC–FID using external
standard calibration against unlabeled 2,3-pentanedione.
2.7.4.2. (3,4-2H2)-2-Methylbutanal. A working solution of 2-methyl-
(3,4-2H2)-butanal was synthesized in two steps, beginning with the
synthesis of the deuterium labeled alcohol, followed by its oxidation to
the corresponding aldehyde. 2-Methyl-(3,4-2H2)-1-butanol was
synthesized according to the method previously described for the
synthesis of 3-methyl-(3,4-2H2)-1-butanol with slight modification
(Steinhaus & Schieberle, 2005), as follows: Wilkinson’s catalyst
(0.15 g), 2-methyl-3-buten-1-ol (0.950 g, 11.0 mmol) were placed in a
F. Kulapichitr, et al.
pressure reactor equipped with stirring bar and rubber septum. The
reactor was flushed for 5 min with deuterium gas (40 psi; UHP grade
99.995%; isotopic enrichment 99.7%; Matheson Tri-Gas, Parsippany,
NJ) using a needle, which was placed below the solution. The spent
catalyst was removed by centrifugation after the reaction was complete.
2-Methyl-(3,4-2H2)-1-butanol was obtained after purification by
vacuum distillation. Yield of 2-methyl-(3,4-2H2)-1-butanol: 0.470 g
(49.5%). MS–EI, m/z (%): 58 (1 0 0), 43 (88), 42 (81), 45 (81), 57
(73), 44 (55), 41 (35), 40 (28), 39 (17), 76 (1, M + ).
2-Methyl-(3,4-2H2)-1-butanol (50 mg; 0.55 mmol) in 2 mL of 1,2-
dichloroethane was added in one portion to a stirred solution of PCC
(0.43 g; 0.002 mol) in 5 mL of 1,2-dichloroethane. After stirring for
1.5 h at room temperature, the reaction mixture was passed through a
column of Florisil (10 g), followed by an additional 10 mL of 1,2-di-
chloroethane to rinse the column. The final solution (approximately
15 mL) was used directly in SIDA. Concentration of the labeled alde-
hyde (MS(EI), m/z (%): 59 (1 0 0), 58 (53), 42 (52), 43 (31), 57 (14), 60
(8), 88 (8, M + )) was determined by GC–FID, using external standard
calibration against unlabeled 2-methylbutanal.
2.7.4.3. (1,1,1-2H3)-Propanal. The target compound was synthesized
from (1,1,1-2H3)-1-propanol (Sigma-Aldrich) by PCC oxidation, as
described previously for (3,4-2H2)-2-methylbutanal. Concentration of
the labeled aldehyde (MS(EI), m/z (%): 61 (100, M+), 60 (27), 41 (6),
62 (4), 59 (4), 40 (5)) was determined by GC–FID, using external
standard calibration against unlabeled propanal.
2.7.4.4. (2H5)-4-Ethylguaiacol. The target compound was synthesized
using a modification of the procedure described by Rayne and Eggers
(2007) for (2H3)-4-ethylguaiacol. First, 4′-hydroxy-3′-methoxy-(2H3)-
acetophenone was synthesized from unlabeled 4′-hydroxy-3′-methoxy-
(2H3)-acetophenone (HMAP) by hydrogen–deuterium exchange as
follows: 0.72 g (4.3 mmol) of unlabeled HMAP, 10 mL of dry pyridine,
5 mL of 2H2O (250 mmol) and 3 drops of 50% (w/v) aqueous NaO2H1
(sodium deuteride) were added to a 40-mL amber vial. The vial was
sealed with a Teflon screw cap and heated at 60 °C for 72 h. The mixture
was then poured into 50 mL of water and the solution acidified (approx.
pH 2) using 4 N HCl. The mixture was extracted with ethyl acetate
(3 × 20 mL), washed with water and the solvent removed to yield the
intermediate product (0.68 g of (2H3)-HMAP; MS(EI), m/z (%): 151
(1 0 0), 169 (47, M + ), 123 (18), 152 (17), 170 (9, M + 1), 108 (7), 46
(7), 52 (6), 168 (5))
(2H3)-HMAP (250 mg; 1.5 mmol), 50 mg of palladium on carbon
(catalyst) and 5.0 mL of 2H1-methanol (Sigma-Aldrich) were placed in
the aforementioned pressure reactor equipped with stirring bar and
rubber septum. The reactor was flushed for 5 min with UHP deuterium
gas using a needle placed below the solution. After 24 h the spent
catalyst was removed by centrifugation. The supernatant was diluted
with 100 mL of water and extracted with ether (3 × 25 mL). The ether
extract was then extracted with 1 M NaOH (3 × 25 mL) to recover the
product. Finally, the basic solution was acidified (approx. pH 2) and
extracted with ether (3 × 25 mL) to yield 110 mg of (2H5)-4-ethyl-
guaiacol: MS(EI), m/z (%): 139 (1 0 0), 157 (43, M + ), 140 (16), 124
(10), 96 (8).
2.7.5. Determination of selected volatile compounds by SIDA-HS-SPME-
GC/(TOF)MS
Three drying treatments were selected for volatile analysis based on
the results of the sensory analysis. In order to better understand the
effects of drying methods on coffee flavor, stable isotope dilution ana-
lysis (SIDA) coupled with headspace–solid-phase microextraction–gas
chromatography–mass spectrometry (HS–SPME–GC–MS) was used to
determine selected key aroma compounds known to be important in
coffee aroma. The target volatiles, isotopically labeled internal stan-
dards, selected MS ions and response factors (Rf) are listed in Table 1.
Coffee powders (2.8 g) were brewed with 50 mL of hot odorless
52
Food Chemistry 291 (2019) 49–58
water (98–100 °C) for 5 min. Coffee brews were cooled immediately in
an ice bath and stirred for 10 min. In the case of 2-furfurylthiol de-
termination, immediately after the cooling step, cysteine (900 mg) was
added to prevent the binding of the thiol to non-volatile coffee matrix
constituents (Sun et al., 2018).
Coffee brews were transferred to 50-mL glass test tubes (Corning
Inc., New York, NY), sealed with a PTFE-lined cap and centrifuged at
1700 g (IEC HN-SII centrifuge; Damon/IEC Division, Needham, MA) for
10 min to remove coffee particles. A 2-mL aliquot of the supernatant
was transferred to a 20-mL screw top headspace vial equipped with
PTFE coated stir bar and a PTFE-lined silicon closure (Supelco Co.,
Bellefonte, PA, USA). For the determination of other target volatiles
besides 2-furfurylthiol, after this step known concentrations of labeled
internal standards (Table 1) were spiked through the septum and then
the sample was mixed well prior to instrumental analysis. To avoid the
potential for sample carryover, a blank (clean empty vial) was analyzed
between each sample.
The sample vial was analyzed using a CombiPal autosampler (Leap
Technologies, Inc., Fort Lauderdale, FL) connected to a GC–MS system
(7890A GC (Agilent Technologies, Inc.)/time-of-flight mass spectro-
meter (TOF–MS; Pegasus 4D; LECO®, St. Joseph, MI)). Sample was pre-
incubated at 60 °C for 10 min at an agitator speed of 250 rpm; then a 2-
cm SPME fiber (50/30 μm DVB/ CarboxenTM/PDMS StableFlexTM;
Supelco) was exposed to the headspace (HS) of the sample vial for
30 min at the same temperature.
Volatile compounds were desorbed into the GC–MS system by hot
splitless injection (260 °C; 4 min valve delay; then 20 min post injection
fiber conditioning in the inlet). A Stabilwax® GC column
(30 m × 0.25 mm id × 0.25 µm film thickness; Restek, Bellefonte, PA)
was used to separate the volatiles. Helium was used as the carrier gas at
a constant flow rate of 1 mL/min. Oven temperature was programmed
from 40 °C to 225 °C at 4 °C/min, with initial and final holding times of
5 min and 30 min, respectively. The MSD conditions were as follows:
capillary direct interface temperature, 230 °C; ionization energy, 70 eV;
mass range, m/z 35 to 300; electron multiplier voltage
(Autotune + 200 V); scan rate, 50 scans/s.
2.7.6. Compound identification
Compounds were identified by comparison of their mass spectra and
retention indices (RI) with authentic reference standards. The RI of
each compound was calculated using the retention time (RT) of that
compound compared with the RTs of a series of standard n-alkanes
(C5–C30).
2.7.7. Quantitation of selected volatile compounds
Peak areas for selected ions of target analytes/IS were determined
using Leco Chroma TOF software (version 3.34). The concentration of a
target compound was calculated by the following equation:
Concentration (t ) = concentration (i ) × Rf
×
area of ion (t )/ area of ion (i)
mass (t )/ mass (i )
where i = labeled internal standard, t = target compound, and
Rf = response factor. The Rf for a target analyte consisted of the inverse
of the slope of the calibration plot of area ratio versus mass ratio, and
was based on analysis of five levels of standard compound (unlabeled)
against the labeled internal standard. For 3-mercapto-3-methylbutyl
formate, the Rf was determined against 1-heptanethiol spiked in brewed
coffee prepared from SD.
2.7.8. Determination of odor-activity values (OAVs)
The odor-activity value (OAV) of a compound was calculated by
dividing its concentration in coffee by its published odor detection
threshold in water (Semmelroch & Grosch, 1996).
F. Kulapichitr, et al. Food Chemistry 291 (2019) 49–58

Table 1
Selected ion (m/z) and response factors (Rf) used in stable isotope dilution analysis.
No.a Compound Ionb Labeled internal standard Ionb R2c Rfd

1 Methanethiol 47 [2H3]-Methanethiol 51 1.00 0.60

2 Propanal 58 [2H3]-Propanal 61 0.99+ 1.70
3 2-Methylpropanal 72 [2H2]-2-Methylpropanal 74 1.00 0.60
4 2-Methylbutanal 86 [2H2]-2-Methylbutanal 88 0.99+ 0.52
5 3-Methylbutanal 86 [2H2]-3-Methylbutanal 88 0.99+ 0.58
6 2,3-Butanedionee 86 [2H2]-3-Methylbutanal 88 0.99+ 0.07
7 2,3-Pentanedione 100 [13C2]-2,3-Pentanedione 102 0.99+ 0.69
8 Hexanal 72 [2H4]-Hexanal 76 0.99+ 0.16
9 2-Methylpyrazine 94 [2H3]-2-Methylpyrazine 97 0.99+ 1.00
10 2,6-Dimethylpyrazine 108 [2H3]-2,6-Dimethylpyrazine 111 0.99 1.30
11 2-Ethylpyrazine 108 [2H5]-2-Ethylpyrazine 113 0.99+ 1.50
12 2,3-Dimethylpyrazine 108 [2H3]-2,3-Dimethylpyrazine 111 0.99+ 1.10
13 Dimethyl trisulfide 126 [2H6]-Dimethyl trisulfide 132 0.99+ 0.52
14 Trimethylpyrazinee 122 [2H5]-3-Ethyl-2,5-dimethylpyrazine 141 0.99+ 0.68
15 2,6-Diethylpyrazine 135 [2H10]-2,6-Diethylpyrazine 144 1.00 0.69
16 3-Ethyl-2,5-dimethylpyrazine 135 [2H5]-3-Ethyl-2,5-dimethylpyrazine 141 0.99+ 0.68
17 2-Furfurylthiol 114 2-[2H2]-Furfurylthiol 116 0.99+ 0.58
18 1-Octen-3-ol 57 [2H3]-1-Octen-3-ol 60 0.99+ 0.70
19 2-Ethyl-3,5-dimethylpyrazinee 135 [2H5]-3-Ethyl-2,5-dimethylpyrazine 141 0.99+ 0.68
20 Methional 104 [2H3]-Methional 107 0.99+ 1.18
21 2,3-Diethyl-5-methylpyrazine 150 [2H3]-2,3-Diethyl-5-methylpyrazine 153 0.99+ 1.10
22 3-Mercapto-3-methylbutylformatee 102 1-Heptanethiol 98 0.97 0.69
23 2-Isobutyl-3-methoxypyrazine 124 [2H2]-2-Isobutyl-3-methoxypyrazine 127 0.99+ 0.92
24 3-Methylbutanoic acid 87 [2H2]-3-Methylbutanoic acid 89 0.99+ 0.86
25 (E)-β-Damascenone 190 [2H4]-(E)-β-Damascenone 194 1.00 0.72
26 Guaiacol 124 [2H2]-Guaiacol 127 0.99+ 0.92
27 4-Ethylguaiacol 152 [2H5]-4-Ethylguaiacol 157 0.99+ 1.1
28 4-Vinylguaiacol 150 [2H3]-4-Vinylguaiacol 154 0.99+ 0.89
29 Indole 117 [2H4]-Indole 121 1.00 0.45
30 3-Methylindole (skatole) 131 [2H3]-3-Methylindole 134 1.00 0.91

a
Numbers correspond to those in Table 5.
b
Selected ion used for quantitation.
c
Coefficient of determination for calibration plot.
d
Response factor.
e
Isotope unavailable, a structurally similar compound was used as internal standard.

Table 2
Drying time, moisture content 1, water activity (aw) 1, contents (mg/100 mg d.w.b.)2 of sugars and caffeine in green coffee beans processed by different drying
treatments.
Properties and compositions HP TD SD

40 °C 45 °C 50 °C

Drying time (h) 23 16 11 72 18

Moisture content (% w.b.) 9.66 ± 0.49A 10.48 ± 0.32A 10.00 ± 0.070A 10.71 ± 0.27A 9.10 ± 0.67A
Water activity (aw) 0.59 ± 0.05A 0.59 ± 0.04A 0.59 ± 0.01A 0.59 ± 0.01A 0.56 ± 0.03A
Sucrose 4.85 ± 0.34B 4.43 ± 0.21D 4.55 ± 0.40C 4.96 ± 0.32A 3.96 ± 0.25E
Glucose/fructose ND3 ND ND ND ND
Caffeine 1.26 ± 0.020A 1.20 ± 0.12A 1.22 ± 0.10A 1.25 ± 0.060A 1.21 ± 0.040 A

A,B,C,D,E
Values with different capital letters in a row are significantly different (p < 0.05).
1
Mean ± S.D. from triplicate analyses.
2
Mean ± S.D. from duplicate analyses.
3
Not detected.

2.8. Statistical analysis
Mean values from chemical analyses were analyzed for significant
differences at p ≤ 0.05 by one-way ANOVA and least significant dif-
ference (LSD). Statistical analyses were performed using SPSS Statistics
software version 13.0 (SPSS Inc., Chicago, IL). Multivariate analysis of
the significant key volatile compounds among three types of brewed
coffees was carried out by principal component analysis (PCA) based on
correlation matrix using XLStat v.2018.2.10 (Addinsoft, Paris, France).
A PCA score plot was used to evaluate the distribution of the significant
key volatiles determined from univariate analysis (one-way ANOVA)
relative to the overall variability of the data sets obtained from the
three types of brewed coffees. PCA computed those volatile variables
into new variables, and PC dimensions/factors indicated the overall
variance of data sets. A biplot was constructed, where the direction and
distance of each variable (vector) were used to indicate the contribution
of each volatile compound variable with respect to the overall volatile
profile.
3. Results and discussion
3.1. Effect of drying process on contents of caffeine, sugars and amino acids
in green coffee
Sucrose is the most abundant sugar in green coffee and during
roasting it is hydrolyzed to glucose and fructose, which in turn take part

53
F. Kulapichitr, et al. Food Chemistry 291 (2019) 49–58

in aroma formation via Maillard and Strecker degradation reactions Table 4

(Belitz et al., 2009; De Melo Pereira et al., 2019; Flament, 2002). As was R-index values based on degree of similarity ranking of aroma for
previous mentioned, mechanical drying can result in coffee of inferior brewed coffees prepared from green coffee beans processed by dif-
flavor compared with SD, since the latter process allows for further ferent drying treatments.
metabolism (via germination) and formation of important aroma pre- Drying treatment (green coffee) R-Index JBa
cursors (Borém, Isquierdo, & Taveira, 2014; Kleinwächter & Selmar,
HP at 40 °C 54.98
2010). Sucrose content was markedly lower in beans dried by SD beans
HP at 45 °C 53.35
compared with HP and TD beans (Table 2), which can be explained by a HP at 50 °C 31.67
greater degree of sugar catabolism caused by prolonged sun drying TD at 50 °C 73.33*
process (Borém et al., 2014; Kleinwächter & Selmar, 2010). Among all
a
treatments, the content of sucrose was the highest (4.96 mg/100 mg Calculated using John Brown computation against SD dried
d.w.b.) in the TD (50 °C) beans and was well within the expected range coffee (control) expressed as percent [n = 30 (21 females and 9
reported for Robusta coffee beans (3–7%) (Belitz et al., 2009). males, 18–54 years old). Critical R-index value = 67.10 (n = 30, two-
tailed test).].
Caffeine is an important contributor to the bitter taste of coffee
* Significantly different from control at α = 0.05.
(Perrone, Donangelo, & Farah, 2008). In this study, the contents of
caffeine did not differ among the coffee beans subjected to the various
3.2. Effects of the different drying processes on the overall perceived aroma
drying treatments (p < 0.05) (Table 2). This can be explained by the
of the brewed coffees
relatively stable nature of caffeine, as it is known to be very resistant to
degradation under the various drying conditions used in this study
The R-index by ranking (degree of aroma similarity) test was carried
(Dong et al., 2017). The levels of caffeine reported here are in agree-
out, in order to determine if significant differences existed with respect
ment with previous studies, which indicated that the caffeine content of
to the perceived aromas among brewed coffee from beans prepared by
Arabica coffee beans can range from 0.8 to 1.4% d.w.b. (Belitz et al.,
SD (control) compared with brewed coffees from HP and TD. The re-
2009).
sults (Table 4) indicated that only the brew made from TD dried coffee
Coffee beans contain numerous amino acids and the actual profile
differed from the SD control. Among the HP brewed coffees, the one
depends on many variables, including how the beans are processed
produced from HP (50 °C) beans was most similar to SD brewed coffee,
(Dong et al., 2017). Contents of individual amino acids differed only
since it received the lowest R-index value. The R-index results cannot be
slightly among dried coffee beans (Table 3), with beans subjected to HP
readily explained by the chemical composition results (Tables 2 and 3),
(50 °C) containing slightly higher levels of most amino acids. It was
since no clear trends were observed among the treatments. However,
evident that the higher temperatures used in HP and TD had only a
the sensory results clearly indicated that on the basis of perceived
moderate effect on the amino acids, possibly because only a short
aroma, HP drying technology may be superior to TD with respect to
drying time was required for HP and TD compared with SD. Among HP
final coffee aroma quality. This may be because HP requires only a
samples, the 50 °C condition resulted in the increase of certain amino
short drying time and mild drying conditions (50 °C), and thus produces
acids and showed a significant difference in main amino acids – glu-
a coffee with a perceived aroma closer to that of SD coffee. To further
tamic and aspartic acids (Belitz et al., 2009) compared with SD
evaluate aroma differences among coffees, SD, HP (50 °C) and TD
(p < 0.05). This may relate to shorter drying time, which results in
brewed coffees were subjected to instrumental analysis, to compare
lower rate of proteolytic reaction within the beans in comparison with
these coffees with respect to concentrations of selected key odorants.
the longer HP at lower temperature. Glutamic acid and aspartic acid in
this study were around 2 mg/100 mg d.w.b. and 1 mg/ 100 mg d.w.b.,
which were close to the range previously reported for green Robusta 3.3. Effects of drying process on important aroma components of brewed
coffee processed by HP (1.66 mg/100 mg d.w.b.) (Dong et al., 2017). coffees
Glutamine and asparagine are converted to glutamic acid and aspartic
acid during analysis; therefore the contents of glutamine and aspar- A robust and accurate analysis method based on SIDA combined
agine are not represented in Table 3. with HS–SPME–GC–(TOF)MS was used to accurately quantitate 30

Table 3
a
Amino acid contents (mg/100 mg d.w.b.) of green coffee beans processed by different drying treatments.
Amino acid HP TD SD

40 °C 45 °C 50 °C

AB C
Aspartic acid 1.25 ± 0.086 1.09 ± 0.052 1.36 ± 0.077A 1.28 ± 0.061AB 1.17 ± 0.013BC
Serine 0.614 ± 0.039A 0.612 ± 0.035A 0.624 ± 0.039A 0.613 ± 0.037A 0.633 ± 0.037A
Glutamic acid 2.34 ± 0.14AB 2.22 ± 0.11B 2.48 ± 0.11A 2.40 ± 0.14AB 2.24 ± 0.13B
Glycine 0.679 ± 0.036A 0.667 ± 0.025A 0.718 ± 0.036A 0.708 ± 0.043A 0.699 ± 0.036A
Histidine 0.269 ± 0.014B 0.268 ± 0.015B 0.291 ± 0.018AB 0.278 ± 0.015AB 0.302 ± 0.017A
Arginine 0.682 ± 0.036A 0.652 ± 0.036A 0.711 ± 0.042A 0.682 ± 0.038A 0.693 ± 0.038A
Threonine 0.438 ± 0.022A 0.424 ± 0.011A 0.434 ± 0.0020A 0.428 ± 0.024A 0.448 ± 0.016A
Alanine 0.492 ± 0.028A 0.444 ± 0.022B 0.507 ± 0.030A 0.462 ± 0.013AB 0.494 ± 0.025A
Proline 0.548 ± 0.032A 0.549 ± 0.026A 0.553 ± 0.019A 0.542 ± 0.016A 0.561 ± 0.033A
Tyrosine 0.304 ± 0.013AB 0.261 ± 0.015C 0.319 ± 0.020A 0.285 ± 0.0070BC 0.304 ± 0.0030AB
Valine 0.571 ± 0.028A 0.512 ± 0.0030B 0.578 ± 0.010A 0.560 ± 0.036A 0.588 ± 0.033A
Isoleucine 0.405 ± 0.022A 0.380 ± 0.014B 0.402 ± 0.0010AB 0.386 ± 0.0050AB 0.392 ± 0.0020AB
Leucine 0.879 ± 0.049A 0.826 ± 0.0050B 0.896 ± 0.0090A 0.862 ± 0.0030AB 0.864 ± 0.0040AB
Lysine 0.655 ± 0.0037ABC 0.623 ± 0.0038C 0.666 ± 0.0031AB 0.687 ± 0.040A 0.642 ± 0.0019BC
Phenylalanine 0.533 ± 0.013B 0.497 ± 0.0060C 0.572 ± 0.015A 0.565 ± 0.034AB 0.545 ± 0.0028AB

A,B,C,D
Values with different capital letters in row are significantly different (p < 0.05).
a
Means ± S.D of triplicate analyses (< 15% RSD).

54
F. Kulapichitr, et al.
volatile compounds previously reported to be important contributors to
the characteristic aroma of coffee beverages (Belitz et al., 2009;
Semmelroch & Grosch, 1996). Aroma compounds containing different
functional groups and representative of several different formation
pathways were chosen for quantitation. These included 10 pyrazines
(nos. 9–12, 14–16, 19, 21 and 23), 5 sulfur-containing compounds
(nos. 1, 13, 17, 20 and 22), 5 aldehydes (nos. 2–5 and 8), 3 guaiacols
(nos. 26, 27 and 28), 3 ketones (nos. 6, 7 and 25), 2 indoles (nos. 29
and 30), 1 alcohol (no. 18) and 1 acid (no. 24). Among these, 23
compounds were determined to have OAVs > 1, with 5 having
OAVs > 100 and 5 with OAVs > 1000. The aroma profiles were si-
milar for the three types of brewed coffees (SD, HP and TD), with some
notable exceptions discussed below.
3.3.1. Sulfur-containing compounds
Among the compounds analyzed in this study, the sulfur-containing
compounds were the most likely group to undergo changes during
sample preparation and analysis, due to evaporation, oxidation, de-
gradation and/or possible interactions with other volatile or non-vo-
latile compounds in the brewed coffee matrix (Weerawatanakorn, Wu,
Pan, & Ho, 2015). In an effort to avoid such losses, a coffee brewing
procedure was developed, which included an immediate (ice-bath)
cooling step and the addition of cysteine to prevent the rapid covalent
binding of volatile thiols, in particular no. 17, to non-volatile coffee
matrix components (e.g., hydroxyhydroquinones; Rowe, 2009; Sun
et al., 2018).
2-Furfurylthiol (no. 17) had the highest OAV among the volatile
compounds analyzed in all brewed coffees (SD, HP and TD). This
odorant has been reported to contribute an important and characteristic
sulfurous, roasted coffee-like note to freshly brewed coffee
(Semmelroch & Grosch, 1996). Its formation in coffee is believed to
occur during roasting via the Maillard reaction between hydrogen sul-
fide (from cysteine) and 2-furfural (from arabinose) (Poisson,
Schmalzried, Davidek, Blank, & Kerler, 2009; Weerawatanakorn et al.,
2015).
The concentrations for 2-furfurylthiol found in this study are 2–3
times higher than those previously reported by Semmelroch and Grosch
(1996) for Arabica (19.1 µg/L) and Robusta (39 μg/L) coffees; however,
these researchers did not add cysteine to prevent 2-furfurylthiol binding
during their analysis. The compound is possibly a key odorant for dis-
criminating the perceived aromas of the brewed coffees, since SD and
HP contained similar levels of 2-furfurylthiol (p-value = 0.207), while
TD contained the lowest content. This is in agreement with the sensory
evaluation results shown in Table 4 and discussed earlier in Section 3.2.
With respect to the other sulfur-containing compounds (nos. 1, 13,
20 and 22) analyzed in this study, only methional (no. 20) markedly
differed among the different brewed coffees, with SD containing ap-
proximately a 2 and 1.5 times higher level than HP and TD, respec-
tively. Methional forms as a result of the Strecker degradation of me-
thionine and α-dicarbonyl compounds during roasting (Semmelroch &
Grosch, 1996). According to Table 5, the concentrations of methional in
all samples were higher than previously reported in brewed coffee
(5.7 μg/L; Semmelroch & Grosch, 1996). This odorant is heat labile and
can be readily degraded to methanethiol (no. 1) (Flament, 2002).
Considering that methanethiol was found at much lower OAVs in this
study compared to what was previous reported in brewed coffee
(OAV = 1050; Semmelroch & Grosch, 1996), it is possible that me-
thional was better stabilized in the present study, due to careful sample
preparation (roasted beans were frozen for 1 h at −70 °C before
grinding) and a strict procedure was used to prepare and immediately
cool down the coffee brews.
3.3.2. Aldehydes
Strecker aldehydes (nos. 3–5), which are known to contribute
characteristic malty notes to coffee products, were present at high OAVs
(> 1000) in all brewed coffees. Their great contribution to the overall
55
Food Chemistry 291 (2019) 49–58
aroma of the brewed coffees is in good agreement with previous lit-
erature (Belitz et al., 2009; Semmelroch & Grosch, 1996). Although the
concentrations for these aldehydes differed among the brewed coffees,
there was no consistent trend which could indicate a potential differ-
ence in the perceived malty aroma of the products. Furthermore, there
was no apparent correlation between the levels of the Strecker alde-
hydes found in the brewed coffees and the relative contents of their
corresponding precursor amino acids in the green coffee beans.
However, many factors in addition to precursors can influence the
Maillard reaction during roasting of coffee, such as reaction tempera-
ture, time, pressure, pH and initial moisture content from postharvest
dehydration (Dong et al., 2019; Ho, Hwang, Yu, & Zhang, 1993; Ledl &
Schleicher, 1990).
3.3.3. Ketones
The three ketones (nos. 6, 7 and 25) analyzed in the brewed coffees
have been previously shown to contribute to the aroma of roasted
coffee products (Semmelroch & Grosch, 1996). The compounds 2,3-
butanedione (diacetyl; no. 6) and 2,3-pentanedione (no. 7), which
impart buttery and creamy notes to coffee, were present at moderate
OAVs; however, no clear trend was observed among the three types of
brewed coffees that might explain any perceived aroma differences.
Scheidig et al. (2007) previously reported that improper storage of
green coffee beans can lead to an increase in the floral and cooked-
apple-like compound β-damascenone (no. 25). However, despite its
apparent high aroma impact (OAVs > 3800), no clear trend was ob-
served among the brewed coffees that might explain any perceived
aroma differences.
3.3.4. Pyrazines
Pyrazines have been reported to contribute nutty and earthy aroma
notes to coffee products (Flament, 2002; Pickard, Becker, Merz, &
Richling, 2013). Among the 10 pyrazines quantitated in this study, only
5 had OAVs > 1, including 4 alkylpyrazines (nos. 15, 16, 19 and 21)
and 1 methoxypyrazine (no. 23). Pyrazines nos. 15, 16, 19 and 21 were
found at higher levels in SD than in both HP and TD. Among these, nos.
19 and 21 were earlier reported as potent odorants in brewed coffee
(Pickard et al., 2013; Pickard, Becker, Wilms, & Richling, 2014), and
the concentrations shown in Table 5 are similar to those earlier re-
ported by Semmelroch and Grosch (1996) for a Robusta coffee brew
(9.3 µg/L). However, on the basis of their relatively low OAVs com-
pared with 2-furfurylthiol and the Strecker aldehydes, alkylpyrazines
may make only a modest contribution to the aromas of the brewed
coffees. On the other hand, the earthy-smelling compound 2-isobutyl-3-
methoxypyrazine (no. 23), which has been previously reported in both
green coffee beans (Scheidig et al., 2007) and roasted coffee brews
(Semmelroch & Grosch, 1996), was found at high OAVs in all brewed
coffees and at about 2 fold higher concentrations in SD and HP than in
TD. This difference, especially considering the relatively high OAVs for
no. 23, could account at least in part to the perceived (sensory) aroma
differences detected between SD and HP versus TD brewed coffees.
Compound no. 23 usually can be formed from the reaction between
leucine and glyoxal in raw coffee (Flament, 2002). The difference
among treatments in with respect to content of no. 23 cannot be ex-
plained by the composition results (i.e., sugar and amino acid profiles),
which were similar among treatments.
3.3.5. Guaiacols
Guaiacols are responsible for characteristic smoky and clove-like
notes in roasted coffee products (Semmelroch & Grosch, 1996). The
OAVs varied greatly among the three guaiacols (nos. 26–28) analyzed
in this study. The low OAV for 4-ethylguaiacol (no. 27) indicated this
compound may be of much less aroma importance than guaiacol (no.
26; OAV ≥ 130) and 4-vinylguaiacol (no. 28; OAV = 350). Only minor
concentration differences existed among all guaiacols, indicating that
drying process had little or no effect on the levels of these compounds in
F. Kulapichitr, et al. Food Chemistry 291 (2019) 49–58

Table 5
Retention indices, concentrations and odor-activity values (OAVs) of selected aroma compounds in brewed coffees prepared from green beans processed by three
different drying treatments.
a
No. Compound RI Odor Threshold (µg/L)c Concentration (ug/L)b OAV

SD HP (50 °C) TD SD HP (50 °C) TD

1 Methanethiol 689 0.2 23.7 ± 0.14 B 25.6 ± 1.4 A 24.0 ± 0.050 B 120 130 120
2 Propanal 804 10 997 ± 1.3 B 1120 ± 160 A 1040 ± 0.86 AB 100 110 100
3 2-Methylpropanal 820 0.70 2810 ± 0.32 A 2440 ± 0.060 B 1820 ± 0.25 C 4000 3500 2600
4 2-Methylbutanal 925 1 1790 ± 0.55 A 1430 ± 0.53 B 1378 ± 68 B 1800 1400 1300
5 3-Methylbutanal 929 0.35 1370 ± 1.2 C 2080 ± 1.5 A 1630 ± 1.6 B 3900 5900 4700
6 2,3-Butanedione 995 15 1730 ± 0.15 B 1480 ± 0.23 C 1980 ± 0.01 A 120 99 130
7 2,3-Pentanedione 1073 30 1930 ± 0.15 A 1660 ± 0.14 B 1420 ± 0.24 C 64 55 47
8 hexanal 1092 10 2.19 ± 0.15 A 1.54 ± 0.0010 B 2.21 ± 0.070 A 0.22 0.15 0.22
9 2-Methylpyrazine 1282 105,000 795 ± 0.070 A 767 ± 0.060 B 686 ± 0.18 C 0.0075 0.0073 0.0065
10 2,6-Dimethylpyrazine 1343 1500 1240 ± 0.17 C 3120 ± 0.17 A 1710 ± 0.73 B 0.83 2.1 1.1
11 2-Ethylpyrazine 1349 21,000 702 ± 0.22 A 589 ± 0.050 B 618 ± 0.070 C 0.033 0.028 0.029
12 2,3-Dimethylpyrazine 1362 2500 932 ± 0.11 B 1420 ± 1.0 A 856 ± 0.01 C 0.37 0.57 0.34
13 Dimethyl trisulfide 1404 0.01 0.138 ± 0.001 A 0.114 ± 0.0060 B 0.109 ± 0.0050 B
14 11 11
14 Trimethylpyrazine 1418 2500 72.1 ± 0.001 A 53.2 ± 0.21 B 53.5 ± 0.015 B 0.79 0.58 0.58
15 2,6-Diethylpyrazine 1444 6 85.8 ± 0.010 A 53.4 ± 0.17 B 52.0 ± 0.05 C 14 8.8 8.6
16 3-Ethyl-2,5-dimethylpyrazine 1447 8.6 142 ± 0.030 A 98.5 ± 14 B 96.2 ± 14 B 16 11 11
17 2-Furfurylthiol 1449 0.01 60.5 ± 0.002 A 64.3 ± 0.001 A 55.4 ± 0.001 B 6000 6400 5500
18 1-Octen-3-ol 1450 1 2.74 ± 0.001 A 2.51 ± 0.002 B 2.52 ± 0.11 B 2.7 2.5 2.5
19 2-Ethyl-3,5-dimethylpyrazine 1457 2 38.0 ± 0.06 A 25.8 ± 0.03 B 25.4 ± 0.88 B 19 13 13
20 Methional 1474 0.2 30.7 ± 0.004 A 13.1 ± 0.001 C 16.3 ± 0.001 B 150 66 82
21 2,3-Diethyl-5-methylpyrazine 1490 0.09 10.4 ± 0.01 A 4.09 ± 0.15 B 4.03 ± 0.14 B 120 45 45
22 3-Mercapto-3-methylbutyl formate 1498 0.0035 1.66 ± 0.11 B 1.98 ± 0.002 A 1.69 ± 0.05 B 470 570 480
23 2-Isobutyl-3-methoxypyrazine 1537 0.005 4.14 ± 0.001 A 4.21 ± 0.003 A 2.66 ± 0.001 B 830 840 530
24 3-Methylbutanoic acid 1673 560 2720 ± 0.08 C 3147 ± 0.10 A 2980 ± 1.5 B 4.9 5.6 5.3
25 E)-β)-Damascenone 1823 0.00075 2.87 ± 0.005 B 3.01 ± 0.045 A 2.98 ± 0.043 A 3800 4000 3900
26 Guaiacol 1843 2.5 336 ± 0.34 C 355 ± 0.45 B 387 ± 0.51 A 130 140 150
27 4-Ethylguaiacol 2050 50 152 ± 0.13 B 152 ± 0.13 B 179 ± 0.01 A 3.0 3.0 3.6
28 4-Vinylguaiacol 2190 20 7010 ± 0.73 A 7010 ± 0.36 A 7020 ± 0.36 A 350 350 350
29 Indole 2385 90 120 ± 0.030 A 119 ± 4.1 A 120 ± 0.030 A 1.3 1.3 1.3
30 3-Methylindole (skatole) 2490 3 5.71 ± 0.74 A 4.75 ± 0.17 B 5.86 ± 0.85 A 1.9 1.6 2.0

A,B,C
Values with different capital superscript letters in row are significantly different (p < 0.05).
a
Retention index on Stabilwax® column.
b
Means from triplicate analyses (RSD < 15%).
c
Odor detection threshold values in water µg/L; nos. 1–3, 5, 7, 16–18, 21, 23 from Semmelroch and Grosch (1996); nos. 4, 10, 12, 13, 15, 26, 29, 30 from
Rychlik, Schieberle, and Grosch (1998); nos. 6, 8, 24, 25, 27, 28 from Guth and Grosch (1993); 9, 11, 20 from Koehler, Mason, and Odel (1971) and 14, 19, 22 from
Flament (2002).

the brewed coffees. This finding is not in agreement with a previous
study that showed an increase in guaiacols in Robusta roasted coffee
prepared by HP drying (Dong et al., 2019).
3.3.6. Miscellaneous compounds
Four additional compounds were analyzed in addition to those
previously discussed. These were chosen to represent compounds de-
rived via lipid oxidation/degradation, e.g., hexanal (no. 8) and 1-octen-
3-ol (no. 18) and those derived from amino acid degradation via the
Maillard reaction, e.g., indole (no. 29) and 3-methylindole (skatole, no.
30) from tryptophan (Holscher & Steinhart, 1993; Flament, 2002). Al-
though these compounds have been reported as coffee aroma con-
stituents, they have not been generally regarded as characterizing or
high impact compounds. This is further supported by the low OAVs
determined for these compounds in this study. In addition, the con-
centrations of these compounds differed only slightly across the three
types of coffee brews, which further demonstrates that drying process
had only a slight impact on their concentrations in the brewed coffees.
3.4. Principal component analysis (PCA)
In an attempt to further discriminate the three types of coffees on
the basis of their aroma profiles, the concentration results for the 23
volatile compounds with OAVs > 1 were subjected to PCA. The PCA
score plot shown in Fig. 1A is based on the first two principal compo-
nents, PC1 (36.38%) and PC2 (21.94%), which were responsible for a
cumulative variance of 58.33%.
The score plot was able to separate and segregate the natural SD
dried coffee and the mechanical dried coffees from a distinct volatile
characteristic to a clear cluster group and indicated the close relation-
ship between the HP and TD coffees. The distances of each cluster
correlated to the proximity in the expression of significant volatile
constituents, where SD was a significant discriminant on the positive
side of PC1 axis while HP and TD were separated on the negative side of
this PC axis. The visualization of PC1 against PC2 showed the close
distance and relationship between HP and TD. This indicates the
sharing of same aroma compound characteristics between the two
mechanically dried coffees, due to the lack of significant differences in
the concentrations of volatile compounds nos. 2, 4, 13–14, 16, 18, 19,
21, 25, 28–29.
The PCA biplot (Fig. 1B) further evaluates how the volatile com-
pound variables were distinguished between the three brewed coffees,
where the direction and length of each variable (Eigen vector) indicates
the relative contribution. All brewed coffees were characterized by high
positive coefficients (> 0.7) for compounds nos. 13–16, 19–21 in PC1
and negative coefficients of no. 3 (−0.81) in PC2. The direction of the
Eigen vectors of these volatiles indicated the dominant characteristics
of SD via the higher concentration of certain volatiles that clearly dis-
tinguishes it from the mechanically dried coffees (HP and TD). The
direction of Eigen vectors and negative coefficients for nos. 24 (−0.63)
and no. 5 (−0.7) in PC1 showed the distinct volatile characteristics of
HP as well as the positive correlation to PC2 of no. 30 (0.702). Mean-
while, the positive coefficients for nos. 26 (0.82) and 27 (0.91) in PC2
were related to the volatile content of TD.

56
F. Kulapichitr, et al.
Fig. 1. Effect of drying methods on 23 significant volatile compounds (from
one-way ANOVA) in brews prepared from dried and roasted coffee beans: (A)
PCA score plot (PC 1 against PC 2) and (B) PCA biplot (PC 1 against PC 2) (HP:
heat pump, 50 °C; SD: sun dried; TD: tray dried, 50 °C.) (Volatile compound
numbers correspond to those in Table 5.)
Although the PCA was able to separate SD from HP and TD brewed
coffees (Fig. 1A), the sensory results indicated that panelists could not
differentiate between SD and HP on the basis of their perceived aromas.
As discussed earlier, this might relate to the lack of significant differ-
ence in the content of the most important character impact odorant of
brewed coffee, 2-furfurylthiol, between SD and HP. Furthermore, even
PCA showed a low correlation for no. 17 in PC1 and PC2 (–0.101 and
–0.45 respectively).
4. Conclusions
This research is the first to compare the chemical, overall perceived
aroma characteristics and selected key aroma compounds of Arabica
coffees prepared by different drying treatments. The findings of this
study indicated that the use of HP produced a coffee with a similar
chemical profile to that of conventional SD coffee. With respect to
perceived aromas, the SD brewed coffee did not differ from any of the
HP brewed coffees, but did differ from the TD brewed coffee. This same
trend was observed for 2-furfurylthiol, where the concentration did not
differ between SD and HP (50 °C) brewed coffees and was lowest in the
TD brewed coffee. PCA helped visualize the results of ANOVA and more
clearly segregated SD from HP and TD coffees based on the variance of
23 selected volatile compounds. HP at 50 °C could serve as a viable
alternative to SD considering its short drying time and the similar
physicochemical properties and perceived aroma characteristics of HP
brewed coffees compared to those of SD.
57
Food Chemistry 291 (2019) 49–58
Acknowledgements
This study was financially supported by a research grant from the
90th Anniversary of Chulalongkorn University Fund (Ratchadapisek
Sompoch Endowment Fund; GCUGR1125594017D) and the Thailand
Research Fund (TRF), Royal Golden Jubilee Ph.D. Program, under
Grant No. PHD/0099/2558.
Statement on conflict of interest
The authors declare no conflict of interests.
References
AOAC 1995. Official method, 980.14, Theobromine and caffeine in cacao products. In
AOAC (Ed.), Official methods of analysis of the association of official analytical
chemists 17 (ed.). Gaithersburg: AOAC International.
AOAC. 2000. Official method, 982.14, Glucose, fructose, sucrose, and maltose in pre-
sweetened cereals. In AOAC (Ed.), Official methods of analysis of the association of
official analytical chemists 17 (ed.). Gaithersburg: AOAC International.
Belitz, H. D., Grosch, W., & Schieberle, P. (2009). Food Chemistry (4 ed.). Berlin,
Heidelberg: Springer.
Borem, F. M., Marques, E. R., & Alves, E. (2008). Ultrastructural analysis of drying da-
mage in parchment Arabica coffee endosperm cells. Biosystems Engineering, 99, 62–66.
Borém, F. M., Isquierdo, E. P., & Taveira, J. H. S. (2014). Coffee Processing. In F. M.
Borem (Ed.). Handbook of Coffee Post-Harvest Technology (pp. 49–68). (1 ed.). G.E:
Norcross.
De Melo Pereira, G. V., de Carvalho Neto, D. P., Magalhaes Junior, A. I., Vasquez, Z. S.,
Medeiros, A. B. P., Vandenberghe, L. P. S., & Soccol, C. R. (2019). Exploring the
impacts of postharvest processing on the aroma formation of coffee beans - A review.
Food Chemistry, 272, 441–452.
Dong, W., Hu, R., Chu, Z., Zhao, J., & Tan, L. (2017). Effect of different drying techniques
on bioactive components, fatty acid. composition, and volatile profile of Robusta
coffee beans. Food Chemistry, 234, 121–130.
Dong, W., Hu, R., Long, Y., Li, H., Zhang, Y., Zhu, K., & Chu, Z. (2019). Comparative
evaluation of the volatile profiles and taste properties of roasted coffee beans as af-
fected by drying method and detected by electronic nose, electronic tongue, and HS-
SPME-GC-MS. Food Chemistry, 272, 723–731.
Fang, M., & Cadwallader, K. R. (2013). Convenient synthesis of stable deuterium-labeled
alkylpyrazines for use in stable isotope dilution assays. Journal of Agricultural and
Food Chemistry, 61(15), 3580–3588.
Flament, I. (2002). Coffee flavor chemistry (1st ed.). New York: John Wiley & Sons.
Goh, L. J., Othman, M. Y., Mat, S., Ruslan, H., & Sopian, K. (2011). Review of heat pump
systems for drying application. Renewable and Sustainable Energy Reviews, 15(9),
4788–4796.
Guth, H., & Grosch, W. (1993). Odorants of extrusion products of oat meal. Changes
during storage. Zeitschrift fur Lebensmittel-Untersuchung und Forschung, 196, 22–28.
Ho, C. T., Hwang, H. I., Yu, T. H., & Zhang, J. (1993). An overview of the Maillard
reactions related to aroma generation in coffee. The 15th ASIC Colloquium
(Montpellier) (pp. 519–527). ASIC: France.
Holsher, W., & Steinhart, H. (1993). Formation pathways for primary roasted coffee
aroma compounds. In T. H. Parliment, M. J. Morello, & R. J. McGorrin (Eds.).
Thermally Generated Flavors (pp. 206-217), ACS Symposium Series 54. Washington
D.C.: American Chemical Society.
International Coffee Organization (ICO). Historical data on the global coffee trade.
(2017). http://www.ico.org/historical/1990%20onwards/PDF/1a-total-production.
pdf (accessed in 1412.17.).
Kleinwächter, M., & Selmar, D. (2010). Influence of drying on the content of sugars in wet
processed green Arabica coffees. Food Chemistry, 119, 500–504.
Koehler, P. E., Mason, M. E., & Odel, G. V. (1971). Odor threshold levels of pyrazine
compounds and assessment of their role in flavor of roasted foods. Journal of Food
Science, 36(5), 816–818.
Kotseridis, Y., Baumes, R., & Skouroumounis, G. K. (1998). Synthesis of labeled (2H4)β-
damascenone, (2H2)2-methoxy-3-isobutylpyrazine, (2H3)α-ionone, and (2H3)β-io-
none, for quantification in grapes, juices and wine. Journal of Chromatography A, 824,
71–78.
Kunarayakul, S., Thaiphanit, S., Anprung, P., & Suppavorasatit, I. (2018). Optimization of
coconut protein deamidation using protein-glutaminase and its effect on solubility,
emulsification, and foaming properties of the proteins. Food Hydrocolloids, 79,
197–207.
Lapsongphon, N., Yongsawatdigul, J., & Cadwallader, K. R. (2015). Identification and
characterization of the aroma-impact components of Thai fish sauce. Journal of
Agricultural and Food Chemistry, 63, 2628–2638.
Ledl, F., & Schleicher, E. (1990). The Maillard reaction in food and in the human body-
new results in chemistry, biochemistry and medicine. Angewandte Chemie, 102(6),
597–626.
Lin, J., Welti, D. H., Vera, F. A., Fay, L. B., & Blank, I. (1999). Synthesis of deuterated
volatile lipid degradation products to be used as internal standards in isotope dilution
assays. 2. vinyl ketones. Journal of Agricultural and Food Chemistry, 47, 2822–2829.
Lorjaroenphon, Y., Cadwallader, K. R., Kim, H., & Lee, S. Y. 2008. Evaluation of key
odorants in chipotle pepper by quantitative analysis, calculation of odor activity
F. Kulapichitr, et al.
values and omission studies. In: Proceedings of the 8th Wartburg symposium on
flavor chemistry & biology: Recent highlights in flavor chemistry and biology.
Garching, Germany: Deutsche Forschungsanstalt für. pp.191–200.
Lyman, D. J., Benck, R., Dell, S., Merle, S., & Murray-Wijelath, J. M. (2003). FTIR-ATR
analysis of brewed coffee: Effect of roasting conditions. Journal of Agricultural and
Food Chemistry, 51, 3268–3272.
O’Mahony, M. (1992). Understanding discrimination tests: A user-friendly treatment of
response bias, rating and ranking R-index tests and their relationship to signal de-
tection. Journal of Sensory Studies, 7, 1–47.
Perrone, D., Donangelo, C. M., & Farah, A. (2008). Fast simultaneous analysis of caffeine,
trigonelline, nicotinic acid and sucrose in coffee by liquid chromatography-mass
spectroscopy. Food Chemistry, 110(4), 1030–1035.
Pickard, S., Becker, I., Merz, K. H., & Richling, E. (2013). Determination of the alkyl-
pyrazine composition of coffee using stable isotope dilution-gas chromatography-
mass spectrometry (SIDA-GC-MS). Journal of Agricultural and Food Chemistry, 61(26),
6274–6281.
Pickard, S., Becker, I., Wilms, H., & Richling, E. (2014). Alkylpyrazine contents of coffee
beverages using stable isotope dilution gas chromatography-mass spectrometry. LWT
– Food Science and Technology, 58(1), 188–193.
Poisson, L., Schmalzried, F., Davidek, T., Blank, I., & Kerler, J. (2009). Study on the role of
precursors in coffee flavor formation using in-bean experiments. Journal of
Agricultural and Food Chemistry, 57(21), 9923–9931.
Rayne, S., & Eggers, N. J. (2007). 4-Ethylphenol and 4-ethylguaiacol in wines: Estimating
non-microbial sourced contributions and toxicological considerations. Journal of
Environmental Science and Health, Part B, 42, 887–897.
Rowe, D. (2009). Chemistry and technology of flavours and fragrances. Hoboken: John Wiley
& Sons.
Rychlik, M., Schieberle, P., & Grosch, W. (1998). Compilation of odor thresholds, odor
58
Food Chemistry 291 (2019) 49–58
qualities and retention indices of key food odorants. Deutsche Forshungsanstalt für
Lebemsmittelchemie and Institut für Lebemsmittelchemie der Technischen. Garching,
Germany: Universität München.
Scheidig, C., Czerny, M., & Schieberle, P. (2007). Changes in key odorants of raw coffee
beans during storage under defined conditions. Journal of Agricultural and Food
Chemistry, 55, 5768–5775.
Schieberle, P., & Hoffman, T. (1997). Evaluation of the character impact odorants in fresh
strawberry juice by quantitative measurements and sensory studies on model mix-
tures. Journal of Agricultural and Food Chemistry, 45, 227–232.
Sen, A., & Grosch, W. (1991). Synthesis of six deuterated sulfur containing odorants to be
used as internal standards in quantification assays. Zeitschrift Fur Lebensmittel-
Untersuchung Und-Forschung, 192, 541–547.
Semmelroch, P., & Grosch, W. (1996). Studies on character impact odorants of coffee
brews. Journal of Agricultural and Food Chemistry, 44(2), 537–543.
Steinhaus, M., & Schieberle, P. (2005). Characterization of odorants causing an atypical
aroma in white pepper powder (Piper nigrum L.) based on quantitative measurements
and orthonasal breakthrough thresholds. Journal of Agricultural and Food Chemistry,
53, 6049–6055.
Steinhaus, M., Sinuco, D., Polster, J., Osorio, C., & Schieberle, P. (2009). Characterization
of the key aroma compounds in pink guava (Psidium guajava L.) by means of aroma
re-engineering experiments and omission testing. Journal of Agricultural and Food
Chemistry, 57, 2882–2888.
Sun, Z., Hayat, K., Yu, J., Karangwal, E., Duhoranimana, E., Zhang, X., & Xia, S. (2018).
Quantification of free 2-furfurylthiol in coffee brew using a prefabricated coffee
model. Food Analytical Methods, 11(3), 654–662.
Weerawatanakorn, M., Wu, J. C., Pan, M. H., & Ho, C. T. (2015). Reactivity and stability
of selected flavor compounds. Journal of Food and Drug Analysis, 23, 176–190.