Académique Documents
Professionnel Documents
Culture Documents
by
Niya Wang
A Thesis
Presented to
The University of Guelph
ABSTRACT
Niya Wang
Advisor:
ACKNOWLEDGMENTS
I am most grateful to Prof. Dr. Loong-Tak Lim for giving me the opportunity
to work in his group. I have always appreciated his far-sighted guidance,
continued support, and constructive evaluation throughout my research and in
many aspects of my life. Further, I am much indebted to my advisory committee
members Dr. Lisa Duizer, and Dr. Massimo Marcone for their unlimited
confidence on my research work and helps during the writing of the thesis.
Special thanks to Natural Sciences and Engineering Research Council of
Canada (NESRC) and Mother Parkers Tea & Coffee Inc., for their essential
financial support, without which this research will not be possible. Many thanks to
my Packaging and Biomaterials Group sisters and brothers: Ana Cristina Vega
Lugo, Solmaz Alborzi, Suramya Minhindukulasuriya, Roc Chan, Grace Wong,
Alex Jensen, Khalid Moomand, Qian Xiao, Xiuju Wang, and Ruyan Dai for their
assistance, friendship, patience, and bringing colourful life for these years. Many
thanks are also going to Dr. Yukio Kakuda, Dr. Sandy Smith, and Bruce Manion
for their technical assistance along the way.
I would like to take this opportunity to express my deepest gratefulness to
my parents, my husband Dr. Yucheng Fu, my son Stanley Fu, and other family
members for their infinite love, support and encouragement throughout these
years of my studies at Guelph.
iii
TABLE OF CONTENTS
ACKNOWLEDGMENTS.........iii
TABLE OF CONTENTS......iv
LIST OF FIGURES......vi
LIST OF TABLES..........viii
LIST OF ABBREVIATIONS.......ix
1 INTRODUCTION ............................................................................................... 1
2 LITERATURE REVIEW ..................................................................................... 4
2.1 THE GREEN COFFEE BEANS ............................................................................. 4
2.2 ROASTING OF COFFEE BEANS .......................................................................... 8
2.3 AROMA COMPOUNDS IN ROASTED COFFEE ...................................................... 14
2.4 FOURIER TRANSFORM INFRARED (FTIR) SPECTROSCOPY ................................ 19
2.5 CHEMOMETRICS ........................................................................................... 21
3 JUSTIFICATION AND OBJECTIVES .............................................................. 26
4 FEASIBILITY STUDY ON CHEMOMETRIC DISCRIMINATION OF ROASTED
ARABICA
COFFEES
BY
SOLVENT
EXTRACTION
AND
FOURIER
LIST OF FIGURES
Figure 1 Chemical composition of green, roasted, and brewed coffee (Barter
2004)..................................................................................................................... 9
Figure 2 Schematic diagram of a typical FTIR spectrometer ............................. 20
Figure 3 Vibrational absorbance due to common bands .................................... 20
Figure 4 Schematic diagram of PCA analysis .................................................... 24
Figure 5 Air temperature (in roast chamber) profiles of the fluidized bed hot air
coffee roaster ...................................................................................................... 30
Figure 6 Appearance of coffee extracts by dichloromethane, hexane, ethyl
acetate, acetone, ethanol, and acetic acid (the right vial represent the extracts by
Method #1) .......................................................................................................... 35
Figure 7 FTIR spectra of coffee extracts obtained with hexane, dichloromethane,
ethyl acetate, acetone, ethanol, or acetic acid using method #1 (with water) and
method #2 (no water) .......................................................................................... 38
Figure 8 Selected FTIR spectra of dark roast coffee extract obtained with
dichloromethane as a solvent (using method 1#)................................................ 41
Figure 9 PCA of FTIR data for hexane, dichloromethane, ethyl acetate, and
acetone extracts of medium roast coffee. Row A: Two-factor score plots. Row B:
Loading plots of PC1. Row C: Corresponding FTIR raw spectra ........................ 42
Figure 10 PCA of FTIR data for hexane, dichloromethane, ethyl acetate, and
acetone extracts of dark roast coffee. Row A: Two-factor score plots. Row B:
Loading plots of PC1. Row C: Corresponding FTIR raw spectra ........................ 43
Figure 11 PCA of FTIR data for dichloromethane extracts of coffee (from the
vi
same origin) with two degrees of roast. Row A: Two factor score plots. Row B:
Loading plots of PC1. Row C: Corresponding FTIR raw spectra ........................ 50
Figure 12 PCA of FTIR data for ethyl acetate extracts of coffee (from the same
origin) with two degrees of roast. Row A: Two factor score plots. Row B: Loading
plots of PC1. Row C: Corresponding FTIR raw spectra ...................................... 51
Figure 13 Changes in lightness, moisture content, pH value, and titratable acidity
of coffee beans processed to different roast stages (A). The same data are
plotted as a function of actual roast time (B). Roasting occurred isothermally at
210, 220, 230 and 240oC .................................................................................... 62
Figure14 PCA analysis for coffees during roasting. Column A: Two-factor score
plots. Column B: Loading plots of PC2. Column C: Representative FTIR spectra
at the start-of-second-crack ................................................................................ 69
Figure15 The expanded 2910-2850 cm-1, and 1800-1500 cm-1 regions of the
spectra of coffee roasted at 230oC...................................................................... 71
Figure16 PCA analysis for coffees collected at the same sampling point. Row A:
Two-factor score plots. Row B: Loading plots of PC1. Row C: Representative
FTIR spectra at 230oC ........................................................................................ 73
Figure 17 SEM micrographs of internal texture for coffee beans collected at
different stages of roast. The temperatures indicated on each row of micrographs
were the roast temperature ................................................................................. 76
vii
LIST OF TABLES
Table 1 Chemical composition of green Arabica and Robusta coffee beans
(g/100g) ................................................................................................................ 7
Table 2 Potent odorants in Arabica coffee from Colombia ................................. 16
Table 3 Physical properties of the investigated solvents (Pagni 2005) ............... 34
Table 4 Evaporation time of the coffee extracts.................................................. 36
Table 5 L* Value of Roasted Ground Arabica Coffee Beans .............................. 39
Table 6 Turkey method for L* value comparisons .............................................. 40
Table 7 SIMCA Classification Results for Coffees from Different Geographic
Origins ................................................................................................................ 52
Table 8 SIMCA classification results for coffees according to degree of roast ... 53
Table 9 Time Taken to Achieve Different Stages of Roasting at Four Different
Final Roast Temperatures .................................................................................. 58
viii
LIST OF ABBREVIATIONS
FTIR
ATR
PTR-MS
PAS
Photoacoustic spectroscopy
PCA
HCA
PLS
PCR
PLS-DA
KNN
K-nearest neighbour
SIMCA
PCs
Principal components
HS-SPME
NMR
GC-MS
GC
Gas chromatography
L*
Lightness
SEM
ix
1 INTRODUCTION
Coffee is one of the most popular beverages in the world. Nearly 25 million
farmers in 50 countries around the world depend on coffee for a significant part of
their livelihoods (Cague et al. 2009). Coffee is the most traded commodity
second after oil (Ponte 2002). Among coffee drinkers, the average consumption
in the United States is 3.2 cups of coffee per day versus 2.6 cups in Canada
(Canada 2003).
A good quality cup of coffee is depended on many factors, such as the
quality of green beans, the roasting conditions, the time since the beans are
roasted, and the type of water used for brewing. More than 800 volatile
compounds have been identified in roasted coffee, whereof around 30
compounds are responsible for the main impression of coffee aroma
(Baggenstoss et al. 2008).
The overall quality and chemical composition of green coffee beans are
affected by many factors, such as the composition of the soil and its fertilization,
the altitude and weather of the plantation, the cultivation, and the drying methods
used for the beans. Coffee plants are mainly grown in tropical and subtropical
regions of central and South America, Africa and South East Asia, in temperate
and humid climates at altitudes between 600 and 2500 m (Schenker 2000). The
genus coffee belongs to the botanical family of Rubiaceae and comprises more
than 90 different species (Davis 2001). However, only C. arabica, C. canephora,
and C. liberica are of commercial importance (Schenker 2000). As a result of
modem breeding techniques some hybrids of C. arabica and C. canephora have
1
recently been introduced with success. Usually roasted coffee beans from
different origins are blended at specific ratios to provide coffee of unique flavour
profiles. Often time, coffee beans are blended for the purpose of cost saving.
Coffee cherries are harvested each year when they are bright-red, glossy,
and firm. After removing the outer hull, the seeds inside of the cherry are
commonly called "green coffee beans". The quality of the green coffee beans is
dictated by a number of parameters, including bean size, color, shape, method of
drying, crop year, and presence of defects (crack, withered bean, bean in
parchment, mouldy bean, etc.).
The unique aroma profiles of coffee are closely related to the timetemperature profile used during roasting. The roasting profiles are chosen to
produce high quality coffee which are unique to specific brands and must be
strictly controlled to meet consumers expectations. Coffee producers rely on
sensory and physicochemical characteristic evaluations to assure that roasting
takes place at the target process parameters. Industrial scale roasting of coffee
beans is mainly achieved by conventional drum roasting, in which beans are
heated with hot gas in a horizontal drum, or vertical drums equipped with paddles.
Roasting time can range from 3 to 12 min, depending on the temperature used,
which is typically between 230 to 250oC. By contrast, fluidized bed roasting is
achieved by directing high velocity hot air towards the beans, usually from the
bottom of the roaster, to suspend the beans in turbulent air. The hot air
temperature ranges from 230 to 360oC (Eggers & Pietsch 2001). The roast
temperature determines both flavour formation and structural product properties.
2
2 LITERATURE REVIEW
2.1 The green coffee beans
The overall quality and chemical composition of green coffee beans are
affected by many factors, such as the composition of the soil and its fertilization,
the altitude and weather of the plantation, and the final cultivation and drying
methods used. Coffee plants are grown in tropical and subtropical regions of
central and South America, Africa, and South East Asia, mainly in regions with
temperate and humid climates (Schenker 2000). Brazil is by far the largest
grower and exporter of green coffee beans in the world followed by Vietnam,
Colombia, Indonesia, Ethiopia and India producing nearly 2.5 million tons of
green coffee beans per year (Franca & Oliveira 2009).
The genus coffee belongs to the botanical family of Rubiaceae and
comprises more than 90 different species (Davis 2001). However, only Coffea
Arabica (Arabica), Coffea canephora (Robusta), and Coffea liberica are of
commercial importance (Schenker 2000). Arabica accounts for approximately 64%
while Robusta accounts for about 35% of the worlds production; other species
with not much commercial value like Coffea liberica and Coffea excelsa represent
only 1% (Rubayiza & Meurens 2005). Due to its more pronounced and finer
flavour qualities, Arabica is considered to be of better quality and accordingly
The aroma profile of roasted ground coffee is also related to the origin and
variety of the beans. In general, blends with greater Arabica content tend to carry
more fruity notes due to the aldehydes, acetaldehyde, and propanal, while the
pyrazines give the earthy odor. In comparison, Robusta beans carry stronger
roasty and sulphury note due to the presence of greater amount of sulphurcontaining compounds (Sanz et al. 2002). Thus, Arabica is often added for the
aroma effect while Robusta is used for enhancing the body, earthy and phenolic
notes of the coffee blend (Parliment & Stahl 1995). Besides contributing to
balanced flavour profiles, Robusta coffee is often blended with Arabica for cost
reduction purpose. Robusta beans are lower in cost since the crops are more
hardy to grow (more resistant to infestation) and easier to harvest (grown in
regions of low elevation) than the Arabica counterpart.
Defective beans (black or brown, sour, immature, insect-damaged, split),
which represent about 11-20% of coffee production, can impact the flavour of the
roasted products. Mazzafera compared the chemical composition of defective
beans and non-defective beans. The researcher found that non-defective beans
were heavier, had higher water activity, and lower titratable acidity than the
defective beans. The content of sucrose, protein, 5-caffeoylquinic acid, and
soluble phenols were also higher in non-defective coffee beans (Mazzafera 1999).
Nevertheless, the antioxidant level in the defective beans, especially chlorogenic
acids, remains high which may be a good source of antioxidant or radical
scavenger for other food applications (Nagaraju et al. 1997).
Arabica coffee
Robusta coffee
Polysaccharides
49.8
54.4
Sucrose
8.0
4.0
Reducing sugars
0.1
0.4
other sugars
1.0
2.0
Lipids
16.2
10.0
Proteins
9.8
9.5
Amino acids
0.5
0.8
Aliphatic acids
1.1
1.2
Quinic acids
0.4
0.4
Chlorgenic acids
6.5
10.0
Caffeine
1.2
2.2
Trigonelline
1.0
0.7
4.2
4.4
Volatile aroma
traces
traces
Water
8 to 12
8 to 12
green coffee beans may be stored for more than 3 years (Bucheli et al. 1998).
Usually, green coffee beans are packaged in natural jute, sisal or burlap bags,
although high quality beans may be packaged in high barrier synthetic vacuum
packages fabricated from synthetic thermoplastic polymers. Cupping is a method
to detect the early stages of coffee deterioration. Bucheli and others (Bucheli et al.
1996) reported that glucose was a sensitive marker for green coffee bean quality
during storage. Glucose is present only in trace amount of good quality green
coffee, and the content will increase when deterioration occurs (Wolfrom & Patin
1965; Bucheli et al. 1996).
starches and
pectins
13%
water
12%
cellulose
(Hyd)
13%
starches and
pectins
14%
protein
13%
cellulose
(non Hyd)
18%
protein
12%
ash
3%
oil
11%
ash
4%
oil
13%
trigonelline
1%
Brewed solubles
trigonelline
1%
CO2
2%
cellulose
(Hyd)
14%
cellulose
(non Hyd)
17%
trigonelline
caffeine 4%
6%
non volatile
acids
31%
ash
16%
oil 1%
protein
5%
soluble
carbohydrates
37%
continues at this temperature, Maillard and pyrolytic reactions start to take place,
resulting in gradually darkening of the beans (Hernandez et al. 2007). The
buildup of water pressure, along with the large amount of gases generated
causes the cellulose cell wall to crack, giving rise to the so called first crack. As
heating continues at the roasting temperature (160-170oC), the coffee becomes
darker and more rapid popping of coffee bean occurs (second crack) as the
carbon dioxide (CO2) buildup exceeds the strength of the cellulosic walls of the
bean. Finally, after roasting, the fresh roasted coffee beans are quickly cooled to
stop roasting (Yeretzian et al. 2002).
The final quality of roasted coffee is influenced by the design of the roasters
and time-temperature profiles used. Although heat transfers during roasting can
involve conduction, convection, and radiation, convection by far is the most
important mode of heat transfer that determines the rate and uniformity of
roasting (Baggenstoss et al. 2008). Coffees roasted in fluidized-bed roaster that
is almost exclusive based on convective heating can result in low density and
high yield coffee (Eggers & Pietsch 2001). On the other hand, coffees roasted in
drum roaster that involves mainly conductive heat transfer have less soluble
solids, more degradation of chlorogenic acids, more burnt flavour, and higher
loss of volatiles than the fluidized bed roasters (Nagaraju et al. 1997).
The effects of time-temperature profile on coffee aroma properties have
been reported by Lyman et al. (Lyman et al. 2003). They observed that the
medium roasted process (6.5 min to the onset of the first crack and 1.0 min to the
onset of the second crack) resulted in good balance of taste and aroma with
10
citrus flavour. However, the sweated process (4.5 min to the first crack and 6.5
min to the second crack) resulted in non-uniform bean color and the coffee was
sour, grassy, and underdeveloped. Reducing the heating rate further by using
the baked process (11 min to the first crack and 18 min to the second crack)
produced coffee of flat, woody with low brightness and acidity (Lyman et al.
2003). In another study, Schenker et al. reported that LHC process (150 to 240 oC
in 270 s; 240oC for 55 s) resulted in the formation of the highest quantities of
aroma volatiles, while the long time low temperature (LTLT) approach (isothermal
heating at 220oC for 600 s) generated the lowest aroma volatiles. Moreover, the
distribution of the 13 volatile compounds monitored was considerably different
depending on the roasting profiles used (Schenker et al. 2002).
Depending on the extent of heat treatment, coffee can be largely
categorized as light, medium or dark roasts. Light roast process tends to give
non-uniform bean color with sour, grassy, and underdeveloped flavour, while
medium roast process produces a balanced taste and aroma with citrus flavour.
By contrast, dark roast process produces coffee of low acidity sensory profiles
(Lyman et al. 2003). Physical characteristics such as temperature, color, and
weight-loss are often used as indicators of roast degree. However, these
parameters only allow assessment of the flavour profile for coffee roasted under
narrow process conditions (Sivetz 1991; Illy & Viani 1995). Other analytical
methods for quantifying the degree of roast include ratio of free amino acids
(Nehring & Maier 1992), alkylpyrazines (Hashim & Chaveron 1995), and
chlorogenic acids content (Illy & Viani 1995). Fobe and others (Fobe et al. 1968)
11
12
The reaction products formed are highly dependent on the roasting timetemperature profile used. Excessive roasting produces more bitter coffee lacking
satisfactory aroma, whereas very short roasting time may be insufficient to
develop full organoleptic characteristics (Yeretzian et al. 2002; Lyman et al. 2003;
Buffo & Cardelli-Freire 2004). Although the majority of phenolic antioxidants
naturally occurring in coffee bean are lost during roasting, the formation of other
antioxidants from Maillard reactions during roasting can enhance the antioxidant
activity of coffee. Compared to medium roast coffee, dark roast coffee exhibited
lower radical scavenging activity than medium roasted coffee due to the
degradation of polyphenol, and thus the antioxidant activity will also depend on
roasting severity and type of coffee (Giampiero Sacchetti 2009).
The profile of organic compounds generated during roasting is very
dynamic and complex. Using Proton transfer reaction-Mass spectrometry (PTRMS) technique, Yeretzian et al. (Yeretzian et al. 2002) simultaneously monitored
the evolution of 8 volatile compounds at isothermal conditions as a function of
time. They observed a distinctive increase in acetic acid, methyl acetate, and
pyrazine concentrations in the headspace, all occurred at the same time.
Concomitantly, there was a rapid decrease in water vapor and methanol
concentrations. Moreover, these peaks shifted in synchronous manner with the
roasting condition. For instance, at 190oC, the above observed changes took
place at 19 min but shifted to 30 min when the beans were roasted at 180 oC
(Yeretzian et al. 2002). Similar observations were observed by Hashim and
Chaveron, who concluded that methylpyrazine may be used as an indicator to
13
monitor the roasting progress of coffee beans (Hashim & Chaveron 1995). It has
been suggested that the pressure buildup within intact bean cells is comparable
to inside an autoclave, which can further complicated the chemical reactions
occurred in coffee bean during roasting (Buffo & Cardelli-Freire 2004).
Chemical reactions happened during coffee roasting are very complex,
which have not been fully understood. Based on the literature reviewed, we can
conclude that the quality of roasted coffee cannot be solely described in terms of
physical parameters at the start and end point of roasting, but rather it is
dependent on the path taken during the roasting process. To reach a specific
flavour profile, not only that precise control of roasting time and temperature is
needed, the variety/quality of green beans, cooling, and degassing conditions are
expected to be important as well.
14
color, aroma, bitterness, and sourness; minor protein components like free amino
acids are highly reactive by interacting with reducing sugars, which make the
Maillard reaction happen; triogenlline generates pyridine and may be
consequently be responsible for some objectionable flavours; and caffeine can
be contributed to the bitterness (Flament 2002).
Maillard reactions have been identified to be the major pathway in the
formation of volatile compounds in coffee roasting (Shibamoto 1991). In the
Maillard reaction, reducing sugars such as glucose or fructose react with free
amino acids to form N-substituted glycosylamine adducts, which are then
rearranged to aminoketones and aminoaldoses by Amadori and Heynes
rearrangements. A complex reaction cascade of Amadori and Heynes
rearrangement products leads to numerous volatile compounds and complex
melanoidins.
More than 800 volatile compounds have already been identified in roasted
coffee, among which, about 40 compounds are responsible for the characteristic
aroma of coffee (Belitz et al. 2009). Some of these compounds are summarized
in Table 2, showing the odorant groups that they are being categorized to
(Semmelroch et al. 1995; Czerny et al. 1999; Mayer et al. 2000).
15
Sulfurous/roasty group
Methylpropanal
2-Furfurylthiol
2-Methylbutanal
2-Methyl-3-furanthiol
3-Methylbutanal
Methional
2,3-Butandione
3-Mercapto-3-methylbutyl-formiate
2,3-Pentandione
3-Methyl-2-butene-1-thiol
4-Hydroxy-2,5-dimethyl-3(2H)-furanone Methanethiol
(HD3F)
Dimethyltrisulde
5-Ethyl-4-hydroxy-2-methyl-3(2H)furanone (EHM3F)
Vanillin
Earthy group
Smoky/phenolic group
2-Ethyl-3,5-dimethylpyrazine
Guaiacol
2-Ethenyl-3,5-dimethylpyrazine
4-Ethylguaiacol
2,3-Diethyl-5-methylpyrazine
4-Vinylguaiacol
2-Ethenyl-3-ethyl-5-methylpyrazine
3-Isobutyl-2-methoxy-pyrazine
Fruity group
Spicy group
Acetaldehyde
3-Hydroxy-4,5-dimethyl-3(5H)-furanone
Propanal
(HD2F)
(E)--Damascenone
5-Ethyl-3-hydroxy-4-methyl-2(5H)furanone(EHM2F)
16
(1) Proteins, peptides and amino acids: Crude protein content is relatively
stable during roasting, while the free amino acids decrease by 30%, with dark
roast espresso reaching up to 50% (Belitz et al. 2009). Protein content plays an
important role in espresso coffee as it affects the foamability of the beverage that
the foamability increased generally with increase total protein concentration until
a maximum value is reached (Nunes et al. 1997). The composition of the amino
acids vary dependent on their thermal stability and reactions involved. For
instance, changes in glutamic acid content are less dramatic as compared to
cysteine and arginine. The latter amino acids tend to deplete rapidly during
roasting due to their involvement in Maillard browning reactions (Illy & Viani
2005).
(2) Carbohydrates: Only traces of free mono and disaccharides in green
coffee remain after roasting. Cellulose, hemicellulose, arabinogalactan and
pectins play important roles in the retention of volatiles and contribute to coffee
brew viscosity. It is reported that in espresso coffee, the foam stability is related
to the amount of galactomannan and arabinogalactan (Nunes et al. 1997).
(3)
Non-volatile
lipids
and
lipid-solubles:
Triglycerides,
terpenes,
tocopherols and sterols contribute to brew viscosity. The lipid fraction tends to be
stable and survive the roasting process with only minor changes. Linoleic and
palmitic acids are the predominant fatty acids in coffee. Cafestol and kahweol are
diterpenes that degrade by the roasting process. Another diterpene, 16-Omethylcafestol, is present in Robusta but not Arabica coffee, making it a suitable
17
indicator for detecting Robusta content in coffee blend (Speer et al. 1991; Belitz
et al. 2009).
(4) Caffeine: Caffeine is of major importance with respect to the
physiological properties of coffee, and also in determining the strength, body and
bitterness of brewed coffee. The caffeine content of green coffee beans varies
according to the species that Robusta coffee contains about 2.2%, and Arabica
about 1.2%. Environmental and agricultural factors appear to have a minimal
effect on caffeine content. During roasting there is no significant loss in terms of
caffeine (Ramalakshmi & Raghavan 1999). However, caffeine content per 177
mL (6 oz) of coffee range from 50 to 143 mg, depending on the mode of
preparation(Rogers & Richardson 1993; Bell et al. 1996). Bell and others (Bell et
al. 1996) reported that more coffee solids, larger extents of grinding, and larger
volumes of coffee prepared at a constant coffee solids to water ratio led to
significantly higher caffeine content. Home-grinding yielded caffeine content
similar to that of store-ground coffee, and boiled coffee had caffeine contents
equal to or greater than filtered coffee (Bell et al. 1996).
(5) Acids: Acids are responsible for acidity, which together with aroma and
bitterness is a key contributor to the total sensory impact of a coffee beverage.
Carboxylic acids, mainly citric, malic and acetic acids are responsible for acidity
in brewed coffees. Arabica coffee brews are more acidic (pH 4.85-5.15) than
Robusta brews (pH 5.25-5.40) (Vitzthum 1975).
(6) Melanoidins: The final products of the Maillard reaction between amino
acids and monosaccharides, are the brown-coloured substances that impart to
18
roasted coffee its characteristic color, possess antioxidant activity, and affect on
the flavor volatiles (Hofmann & Schieberle 2001; Del Castillo et al. 2002; Vignoli
et al. 2011).
19
Sampling methods in FTIR include transmission, reflectance, and microsampling (Stuart 2003). The transmission method is based on the absorption of
IR radiation as it passes through a sample. It can be used to analyze solid, liquid,
and gaseous sample. The reflectance method can be used for samples that are
20
2.6 Chemometrics
21
Chemometrics
can
be
generally described
as the
application
of
Among the chemometric analyses used, PCA by far is the most commonly
used. It is a linear and non-parametric pattern recognition technique which
reduces multidimensionality by correlating data to two or three dimensions (Anil
et al. 2004). The goal of PCA is to visualize the inherent data structure and reveal
how different variables change in relation to each other. This is achieved by
transforming correlated original variables into a new set of uncorrelated
underlying variables, known as principal components (PCs), using the covariance
matrix. The new variables are linear combinations of the original ones. The
principle of PCA can be illustrated using a simple dataset, where the 3 variables
needed to describe the dataset are represented by three axes in the data-space
(Figure 4). PC1 has a direction that takes into account as much variance in the
data as possible. PC2, orthogonal to PC1, has a direction where the second
largest variance occurs. The objects are then projected down to the plane of the
two PCs. A large data-set may therefore be represented by only a few PCs,
which describe a large part of the variance in the data as a linear combination of
the original variables. PCA is very useful for solving pattern recognition problems
arising from chromatographic and spectroscopic data (Hagman & Jacobsson
1990).
On other hand, PLS is a useful multivariate regression technique for
correlating two or more blocks of data with each other, or predicting a value of
one block by using the data from the other block that is easier to measure
(Gerlach et al. 1979). PLS can handle more than one dependent variable and is
not significantly influenced by the correlation between the independent variables.
23
In addition, it can tolerate missing values in the data-matrix (Geladi & Kowalski
1986). In the PLS method, X (independent) variables are related to a block of Y
(dependent) variables through a process where the variance in Y-block
influences the calculation of PCs of X-block (Hagman & Jacobsson 1990).
chromatographic data obtained by headspace solid phase microextraction (HSSPME), and the results showed that coffees from different origins can be
successfully separated (Bicchi et al. 1997). In another study, Charlton and others
applied PCA to analyze Nuclear magnetic resonance (NMR) spectra from 98
coffee samples obtained from three different producers (Charlton et al. 2002). In
their study, 99% of the samples were correctly classified accordingly to their
manufacturers. Also, blind testing of the PCA model with a further 36 samples of
instant coffee resulted in a 100% success rate in identifying the samples from the
three manufacturers.
25
3 OBJECTIVES
Currently, integrated studies are lacking on elucidating the effects of bean
variety and roast degree, under different time-temperature conditions, on the
physical and chemical properties of coffee. The objectives of this study are:
26
27
infrared spectral data of these extracts, in conjunction with PCA and SIMCA, to
discriminate four Arabica ground coffees from different origins (Colombia, Costa
Rica, Ethiopia, and Kenya) that had been roasted to two roast degrees (medium
or dark).
29
250
Dark roast profile
Roasting Temperature, o C
225
200
175
150
125
100
75
50
25
0
50
100
150
200
250
300
350
400
450
500
550
Roasting Time, S
Figure 5 Air temperature (in roast chamber) profiles of the fluidized bed hot air
coffee roaster
30
procedures. In the first procedure (method #1), 0.2500 g of ground coffee was
accurately weighed into a glass vial, and 1 mL deionized water was added to wet
the sample. The glass vial was shaken for 1 min with an IKA-VIBRAX-VXR
vibrator (Janke & Kunkel, Inc., Staufen, Germany) at the 200 dial setting; 1 mL of
organic solvent was added and the mixture was shaken for an additional 5 min.
The organic phase was then transferred to another vial and allowed to rest for 10
min before ATR-FTIR analysis. In the second procedure (method #2), a similar
procedure was used except that water was not added prior to solvent extraction.
All extractions were performed in triplicate.
4.2.5 ATR-FTIR Analysis
The coffee extract was scanned using an FTIR spectrometer (IR Prestige21; Shimadzu Corp., Tokyo, Japan) equipped with a deuterated triglycine sulfate
detector and a KBr beam-splitter. A MIRacle ATR accessory equipped with a
diamond crystal (Pike Technologies, Madison, WI) was used for sampling. The
background spectrum was collected using an empty ATR cell. To collect each IR
spectrum, coffee extract (6 L) was placed onto the ATR crystal, and the solvent
was allowed to evaporate until no further changes through consistently controlling
evaporation time during the experiment in IR spectrum were observed. This
technique removed interference from the solvent signals and increased the
sensitivity of chemometric analysis. The times required for complete evaporation
of solvent were different due to the different solubilities of each solvent in water.
Samples were scanned from 600 to 4000 cm 1 at 4 cm1 resolution. Each
spectrum was an average of 20 scans. For each extract, 3 FTIR spectra
31
replicates were scanned. Between samples, the ATR crystal was carefully
cleaned with 95% (v/v) aqueous ethanol solution, and dried with lint-free tissue
paper. The spectral baseline was examined visually to ensure that no residue
from the previous sample was retained on the crystal. All spectra were recorded
at room temperature (23 0.5 C).
4.2.6 Data Analysis
Statistical comparison of color values of ground coffee samples was
conducted based on Tukey pairwise comparisons using R software (www.rproject.org). For chemometric analysis, FTIR spectra were exported as ASCII
format, organized in Excel spreadsheets, and then analyzed using Pirouette v.4.0
software (Woodinville, WA). During PCA, second derivative and mean-center
were applied to FTIR spectra to reduce baseline variation and enhance spectral
features. Nine spectra (3 extracts for each coffee and 3 replicate spectra for each
extract) for each coffee were divided into two groups: 6 spectra from the first two
extracts were used to calibrate the SIMCA model, while the remaining 3 spectra
from the third extract were used for validation to evaluate the prediction accuracy
of the calibrated SIMCA model. The optimum number of PCs in each class was
selected on the basis of the lowest number of PCs giving minimum value of
variance.
32
while hexane, ethyl acetate, acetone, ethanol, and acetic acid phases were on
top (Figure 6). Three layers (solvent, water, ground coffee phases) were
observed when dichloromethane, hexane, and ethyl acetate were used as a
solvent because they were immiscible or slightly soluble in water. The three
layers observed were likely caused by the different densities of ground coffee,
water, and solvent. However, for acetone, ethanol, and acetic acid extractions,
only two phases were observed since these solvents were miscible with water.
For coffee extracted by method #1, coffee grinds were all in one layer. On the
other hand, in the presence of organic solvent alone (Method #2), the extract
layers were hazy, and tended to contaminate with grind particulates. This may be
due to the fact that when the samples were wetted with water, the entrapped air
in the ground coffee matrices was readily displaced by the solvents, thereby
reducing the buoyancy of the grind particulates.
Polarity
Density,
20C
index (P)
g/mL
Dichloromethane
3.1
1.326
Hexane
Immiscible(0.0013 v/v)
0.1
0.659
Ethyl acetate
4.4
0.895
Acetone
Miscible (infinitely)
5.1
0.786
Ethanol
Miscible (infinitely)
5.2
0.789
Acetic acid
Miscible (infinitely)
6.2
1.049
Solvent
34
Dichloromethane Extract
Acetone Extract
Hexane Extract
Ethanol Extract
35
With H2O
No H2O
Dichloromethane extract
60
60
Hexane extract
60
60
180
180
Acetone extract
600
60
Ethanol extract
760
280
840
780
(esters and
alcohol),
CH2
37
coffee extraction. On the basis of the evaporation time data and FTIR spectral
features observed, dichloromethane, hexane, ethyl acetate, and acetone extracts
obtained via method #1 were selected for subsequent analyses.
3.6
Hexane extract
3.6
2.8
2.8
2.4
2.4
No water
2.0
No water
1.6
No water
2.0
No water
1.6
1.2
With water
1.2
0.8
With water
With water
0.4
0.8
With water
0.4
0.0
0.0
3600
3000
2400
1800
1200
Wavenumber, cm -1
600
1800
3.6
1600
1400
1200
Wavenumber, cm-1
1000
3600
800
3000
2400
1800
Wavenumber, cm-1
1200
600
1800
Acetone extract
3.6
3.2
1600
1400
1200
Wavenumber, cm -1
1000
800
3.2
2.8
2.8
No water
2.4
No water
2.0
No water
Absorbance
Absorbance
Dichloromethane extract
Absorbance
3.2
Absorbance
3.2
2.4
No water
2.0
1.6
With water
1.2
With water
1.6
1.2
With water
With water
0.8
0.8
0.4
0.4
0.0
0.0
3600
3000
2400
1800
Wavenumber, cm-1
1200
600
1800
1400
1200
Wavenumber, cm-1
1000
3600
800
3000
6.4
Ethanol extract
3.6
1600
2400
1800
1200
Wavenumber, cm -1
600
1800
1400
1200
1000
Wavenumber, cm -1
800
1600
5.6
3.2
4.8
2.8
No water
2.4
2.0
With water
1.6
Absorbance
Absorbance
4.0
No water
No water
3.2
No water
2.4
With water
1.2
1.6
With water
0.8
With water
0.8
0.4
0.0
0.0
3600
3000
2400
1800
1200
Wavenumber, cm-1
600
1800
1600
1400
1200
Wavenumber, cm-1
1000
800
3600
3000
2400
1800
1200
Wavenumber, cm-1
600
1800
1600
1400
1200
Wavenumber, cm-1
1000
800
38
Lightness [L*]
Dark
Colombian
19.83 0.05
Costa Rican
19.61 0.18
Ethiopian
19.46 0.21
Kenyan
19.72 0.06
Colombian
25.21 0.16
Costa Rican
25.35 0.29
Ethiopian
25.64 0.06
Kenyan
25.28 0.09
Medium
39
95% SCI
Different
(Dark roast)
(Medium roast)
from 0?
(-0.176, 0.616)
(-0.326, 0.866)
No
(-0.026, 0.766)
(-0.036, 0.896)
No
(-0.286, 0.506)
(-0.396, 0.536)
No
(-0.246, 0.546)
(-0.176, 0.756)
No
(-0.286, 0.506)
(-0.396, 0.536)
No
(-0.136, 0.656)
(-0.106, 0.826)
No
Comparison
Dark roast: MSE = 0.022908, HSD (t, F) = 0.396; Medium roast: MSE = 0.03175,
HSD (t, F) = 0.466
40
0.16
Kenyan
Abs
Ethiopian
Costa Rican
0.14
0.12
0.1
0.08
0.06
0.04
0.02
-0
4000
3750
FTIR Measurement
3500
3250
3000
2750
2500
2250
2000
1750
1500
1250
1000
750
1/cm
Figure 8 Selected FTIR spectra of dark roast coffee extract obtained with
dichloromethane as a solvent (using method 1#)
41
Hexane Extract
Costa Rican
Kenyan
Colombian
Ethiopian
Colombian
Kenyan
0.000
-0.001
-0.002
-0.005
0.000
0.005
0.001
0.000
-0.001
-0.002
-0.003
-0.005
0.010
Acetone Extract
Ethiopian
Kenyan
Colombian
-0.003
0.000
0.003
0.005
0.000
-0.001
-0.002
-0.003
-0.005
Ethiopian
Kenyan
-0.003
0.000
0.003
0.001
0.000
-0.001
-0.002
-0.003
0.005
-0.002
-0.001
0.000
1st Principal Component
PC 1 Loading (24%)
PC 1 Loading (89.4%)
Costa Rican
0.002
0.001
Costa Rican
0.002
0.001
-0.003
-0.010
Costa Rican
0.002
0.002
Dichloromethane Extract
Ethiopian
Colombian
0.001
0.002
PC 1 Loading (26.4%)
PC 1 Loading (59.4%)
0.25
0.25
0.25
0.25
0.15
0.15
0.15
0.15
0.05
0.05
0.05
0.05
-0.05
-0.05
-0.05
-0.15
4000
-0.15
4000
3200
2400
1600
800
0.20
3200
2400
1600
800
0.16
0.12
0.12
1741
Absorbance
0.08
2920
1741 1726
2850
0.04
0.00
4000
3200
2400
1600
800
0.08
0.00
4000
1600
-0.15
4000
800
2400
1600
2400
1600
800
1028
2920
0.12
1668
1743
2850
1550
0.08
800
0.00
4000
Wavenumber, cm -1
1548
0.16
0.12
0.08
0.04
0.04
3200
3200
0.20
0.16
0.04
Wavenumber, cm -1
2400
1678
Absorbance
0.16
3200
0.20
0.20
2850
-0.15
4000
Absorbance
-0.05
3200
2400
Wavenumber, cm -1
1600
800
0.00
4000
3200
2400
1600
800
Wavenumber, cm -1
Figure 9 PCA of FTIR data for hexane, dichloromethane, ethyl acetate, and acetone extracts of medium roast coffee.
Row A: Two-factor score plots. Row B: Loading plots of PC1. Row C: Corresponding FTIR raw spectra
42
Dichloromethane Extract
Hexane Extract
Ethiopian
Kenyan
Colombian
0.002
Ethiopian
0.000
-0.001
-0.002
-0.005
0.000
0.005
Colombian
-0.001
Ethiopian
Kenyan
Colombian
-0.001
0.000
0.001
0.000
-0.001
-0.002
-0.004
0.002
Costa Rican
Kenyan
-0.002
0.001
0.001
0.000
-0.001
-0.002
-0.010
0.004
-0.005
PC 1 Loading (60%)
Ethiopian
0.002
0.001
PC 1 Loading (85.8%)
Costa Rican
0.002
0.000
-0.002
-0.002
0.010
Acetone Extract
0.001
-0.003
-0.010
Costa Rican
0.001
3nd Principal Component
Dichloromethane
Acetone
Costa Rican
Colombian
PC 1 Loading (51.3%)
0.000
0.005
1st Principal Component
0.010
PC 1 Loading (83.7%)
0.25
0.25
0.25
0.25
0.15
0.15
0.15
0.15
0.05
0.05
0.05
0.05
-0.05
-0.05
-0.05
-0.05
3200
2400
1600
-0.15
4000
800
0.20
3200
2400
1600
-0.15
4000
800
0.20
3200
2400
1600
0.16
0.12
2850
0.08
1550
0.12
0.08
0.04
0.04
0.00
4000
0.00
4000
3200
2400
Wavenumber, cm -1
1600
800
3200
2400
1600
800
Wavenumber, cm -1
800
1514 1481
0.16
1741
2920
1550
2850
0.08
0.00
4000
1600
1649
1236
0.12
0.04
2400
1647
0.16
Absorbance
1741 1726
Absorbance
Absorbance
2920
1683
3200
0.20
0.20
1697
0.16
-0.15
4000
800
Absorbance
-0.15
4000
0.12
1662
0.08
0.04
3200
2400
1600
800
0.00
4000
Wavenumber, cm -1
3200
2400
1600
800
Wavenumber, cm -1
Figure 10 PCA of FTIR data for hexane, dichloromethane, ethyl acetate, and acetone extracts of dark roast coffee. Row
A: Two-factor score plots. Row B: Loading plots of PC1. Row C: Corresponding FTIR raw spectra
43
For the medium roast coffee samples, FTIR data for dichloromethane and
ethyl acetate extracts resulted in well-separated clusters in PCA score plots,
which corresponded to the four countries of origin (Figure 9, row A); however,
cluster patterns were less discernible for hexane and acetone extracts. For the
dark roast samples, the PCA score plots of FTIR data for all the solvent extracts
showed clear distinctive groupings for coffee beans from the four countries of
origin (Figure 10, row A). Overall, separation distances between clusters were
greater for the dark roast samples than for the medium roast counterparts,
implying that the IR active-active components that were distinctive to the bean
origin tended to develop when the beans were roasted to a darker degree.
Roasting of green coffee beans results in the formation of large quantities of
aroma compounds, and aroma profiles are dependent on the time and
temperature regimes applied during the roast process (Lyman et al. 2003). The
greater cluster separation for dark roast samples observed in the current study
may be due to a larger number of aroma compounds produced in the dark
roasted coffee (Byers 2003).
The different clustering behaviors observed for extracts prepared from
different solvents could be attributed to the different polarities of the solvents
used. The polarity index for hexane, dichloromethane, ethyl acetate, and acetone
are 0.1, 3.1, 4.4, and 5.1, respectively (Byers 2003). Thus, hexane is non-polar
and extracts only non-polar compounds from coffee. On the other hand, acetone
is relatively more polar and tends to extract polar compounds. For
dichloromethane and ethyl acetate, both polar and non-polar compounds are
44
extracted. The polarity effect can be observed in the original spectra (Figures 9
and 10, row C). The spectral region from 3676-3028 cm-1 is mainly due to the OH stretching band from water. As shown, the absorbance intensity in this region
progressively became stronger for hexane, dichloromethane, ethyl acetate, and
acetone in ascending order. This result is consistent with the polarity for these
solvents.
To further investigate regions of spectra that contribute to the variance of
samples, the loading plots for a corresponding PC were inspected. Here we
focused on PC1 since it explained the maximum variance existing in the dataset
(Figures 9 and 10, row B). The percent variance accounted by PC1 was also
indicated on each loading plot. Regions of each spectrum with a relatively large
loading score (>0.1) were highlighted as red dotted lines. As shown, the loading
plots for hexane extracts were markedly different than those of the other three
solvent extracts, due to the non-polar nature of hexane. The loading plots of
hexane extracts for medium and dark roasts were similar, except that
absorbance at region 1741-1726 cm-1, which is due to C=O stretching band
mode of fatty acid esters, was higher and wider in the medium roast compared
with the dark roast coffee (Yoshida et al. 1997). For dichloromethane extracts,
the most prominent difference in loading plots for dark and medium roast coffees
was in the region of 2920-2850 cm-1, which can be attributed to CH2
asymmetrical stretching vibrations of hydrocarbon methyl groups (Eliane
Nabedryk 1982). The medium roast coffees exhibited significant loading score
around this region, but negligible for dark roast coffees. A similar trend was
45
observed for the region around 1741-1678 cm-1 The minimal changes observed
for these spectral regions for the dark roast samples could be caused by a
decrease in protein and lipids due to the Maillard reaction and pyrolytic cleavage,
respectively (De Maria et al. 1994; Yeretzian et al. 2002).
For ethyl acetate extracts, loading plots for medium and dark roast coffees
were comparable, indicating that the compounds extracted by ethyl acetate from
medium and dark roast coffees were similar, although subtle differences did exist.
The main regions that contribute to the differences between samples are 17431741, 1647-1643, and 1697 cm-1. The band at 1697 cm-1 is due to isolated
carbonyl stretching of C=O bonds, and the band at 1647 cm -1 is due to
conjugated carbonyl stretching of C=O bonds of caffeine compounds (Falk et al.
1990). Garrigues et al. (Garrigues et al. 2000) and Ohnsmann et al. (Ohnsmann
et al. 2002) also utilized absorbance at 1659 and 1704 cm -1 to determine the
caffeine content in coffee and tea, respectively. In these cited studies, the C=O
bands investigated shifted to higher frequencies due to the different solvent used
(i.e., chloroform). Based on this information, it is conceivable that the separated
clusters observed were partly caused by the different caffeine contents of among
the various coffee samples.
Other important vibration bands that contributed to the separated clusters
for dichromethane extracts were at 1705 cm -1 (C=O stretching vibrations of
ketones), 1655 cm-1 (C=O stretching of caffeine compounds), 1599 cm-1 (-NH
group), and 1548 cm-1 (N-H bending of peptide groups). These bands were also
detected in ethyl acetate and acetone extracts with some shifts (1701, 1651,
46
1604, and 1552 cm-1 for ethyl acetate; 1699, 1647, 1599, and 1558 cm -1 for
acetone) (Magidman 1984; Mishra & Kumar 2002). For hexane extracts, the most
prominent spectral difference between the medium and dark roast coffees is that
the latter showed a stronger overall absorbance, implying that more lipids (16001700 cm-1) and fatty acid esters (1700-1800 cm-1) were being extracted from the
dark roast coffee.
and
ethyl
acetate
extracts
were
tested
for
the
aliphatic esters (Hennessy et al. 2009; Wang et al. 2009). For the Colombian
coffee, the bands that correspond to significant loading scores at 1550, 1510,
and 1481 cm-1 can be attributed to N-H bending of peptide groups, C=N
stretching of amino groups, and benzene absorption bands, respectively (Mishra
& Kumar 2002; Zhang et al. 2005; Barua et al. 2008).
For ethyl acetate extracts, the loading plots revealed that spectral regions
that contributed to cluster separation were mainly at 2850-2920 cm-1 due to CH2
asymmetrical stretching and C-H symmetrical stretching of methyl groups
(Hennessy et al. 2009) as well as 1650-1750 cm-1 due to C=O stretching
vibrations and C=N stretching (Paradkar & Irudayaraj 2002). For coffee, this
region has been assigned to a number of important compounds, including
aromatic acids (1700-1680 cm-1), aliphatic acids (1714-1705 cm-1), ketones
(1725-1705 cm-1), aldehydes (1739-1724 cm-1), and aliphatic esters (1755-1740
cm-1) (Bellamy 1975; Keller 1986; Socrates 1994). Absorbance in the 2850-2920
cm-1 region was mainly due to lipids (Hennessy et al. 2009).
Overall, roasting coffee from a medium to a dark degree causes increases
in esters/lactones (1788 cm-1), aldehydes/ketones (1739-1722 cm-1), ketones
(1725-1705 cm-1), aromatic acids (1700-1680 cm-1), and aliphatic acids (17141705 cm-1), but a decrease in caffeine content (1700-1692 cm-1, and 1647-1641
cm-1) (Lyman et al. 2003; Movasaghi et al. 2008; Wang et al. 2009). Others have
also observed decreases in the amount of lipids (around 1736, 1740, 1745, and
1750 cm-1), polysaccharides and hemicelluloses (1739 cm-1), esters (1751-1740
48
cm-1), and lipids/proteins (2935-2847 cm-1) (Lyman et al. 2003; Movasaghi et al.
2008; Wang et al. 2009).
49
Colombian
Costa Rican
Dark roast
Medium roast
0.001
0.000
-0.001
-0.002
-0.003
-0.001
0.001
0.001
0.000
-0.001
-0.002
-0.003
0.003
Kenyan
Medium roast
Dark roast
-0.001
0.001
0.001
0.000
-0.001
-0.002
-0.003
0.003
0.001
0.000
-0.001
-0.002
-0.003
0.003
-0.001
0.001
0.003
PC 1 loading (46.6%)
PC 1 loading (67.8%)
PC 1 loading (75.9%)
-0.001
0.001
Medium roast
0.002
0.002
Ethiopian
Dark roast
Medium roast
0.002
Dark roast
0.002
PC 1 loading (29.2%)
0.2
0.2
0.2
0.2
0.1
0.1
0.1
0.1
0.0
0.0
0.0
0.0
2400
1600
-0.1
4000
800
0.20
3200
0.16
0.12
Absorbance
Absorbance
-0.1
4000
800
1510
0.08
1550
0.04
2400
Wavenumber, cm-1
1600
800
1600
-0.1
4000
800
0.08
0.12
2920
2400
1600
800
1678
0.00
4000
1600
800
1743
2850
0.08
Wavenumber, cm-1
2400
0.16
1465
2850
0.04
3200
3200
0.20
1743
0.16
1743
2850
0.12
0.00
4000
2400
2920
0.04
3200
3200
0.20
2920
1481
0.00
4000
1600
0.20
0.16
2400
Absorbance
3200
Absorbance
-0.1
4000
0.12
1695
829
0.08
860
0.04
3200
2400
1600
800
0.00
4000
Wavenumber, cm -1
3200
2400
1600
800
Wavenumber, cm -1
Figure 11 PCA of FTIR data for dichloromethane extracts of coffee (from the same origin) with two degrees of roast. Row
A: Two factor score plots. Row B: Loading plots of PC1. Row C: Corresponding FTIR raw spectra
50
Colombian
Costa Rican
Medium roast
Dark roast
0.002
0.001
0.000
-0.001
-0.002
-0.003
-0.001
0.001
0.001
0.000
-0.001
-0.002
-0.003
0.003
Dark roast
-0.001
0.001
0.000
-0.001
-0.002
-0.003
0.003
-0.001
0.001
0.001
0.000
-0.001
-0.002
-0.003
0.003
PC 1loading (59.5%)
PC 1 loading (44.6%)
medium roast
0.002
0.001
Kenyan
Medium roast
0.002
0.002
Ethiopian
Dark roast
Medium roast
Dark roast
-0.001
0.001
0.003
PC 1 loading (46.7%)
PC 1 loading (74%)
0.2
0.2
0.2
0.2
0.1
0.1
0.1
0.1
0.0
0.0
0.0
0.0
2400
1600
1674
0.20
-0.1
4000
800
-0.1
4000
800
1674
2850
0.08
0.04
2355-2347
3200
2400
Wavenumber, cm -1
1600
800
0.12
2400
1600
-0.1
4000
800
0.08
2850
0.04
0.00
4000
2400
1600
800
1741
2920
0.08
0.00
4000
2850
800
1236
1741
0.12
2920
0.08
1550
2850
0.04
3200
2400
1600
800
Wavenumber, cm -1
Wavenumber, cm-1
1600
1674
1701 1647
0.16
0.12
0.04
3200
2400
1030
1741
2928-2916
3200
0.20
1674
0.16
1741
2920
3200
0.20
0.16
0.12
0.00
4000
1600
1701
0.20
Absorbance
Absorbance
2400
1647
0.16
3200
Absorbance
3200
Absorbance
-0.1
4000
0.00
4000
3200
2400
1600
800
Wavenumber, cm -1
Figure 12 PCA of FTIR data for ethyl acetate extracts of coffee (from the same origin) with two degrees of roast. Row A:
Two factor score plots. Row B: Loading plots of PC1. Row C: Corresponding FTIR raw spectra
51
Roast
degree
Costa
Colombia
Rica
Ethiopia
Kenya
Dark
100
100
100
100
Medium
100
100
33
100
Dark
100
100
100
100
Medium
100
100
100
100
Dichloromethane
Ethyl acetate
52
classification, %
Origin
Dark
Medium
Colombia
100
100
Costa Rica
100
100
Ethiopia
100
100
Kenya
100
100
Colombia
100
67
Costa Rica
100
100
Ethiopia
100
100
Kenya
100
100
Dichloromethane
Ethyl acetate
53
54
led to beans of lower density, higher volume, less roast loss, and lower moisture
content as compared to the low-temperature-short time process (Baggenstoss et
al. 2008). Lyman et al. roasted green coffee beans under various process
conditions to study the effect of roasting on brewed coffee (Lyman et al. 2003).
Using a medium roast process (6.5 min to the onset of the first crack and 1.0 min
to the onset of the second crack), Lyman et al. observed that coffee of balanced
taste and aroma with citrus flavour was produced. However, using the so-called
sweated process (4.5 min to the first crack and 6.5 min to the second crack),
coffee beans of non-uniform bean color with sour, grassy, and underdeveloped
were resulted. In comparison, the baked process (11 min to the first crack and
18 min to the second crack) produced coffees that were flat, woody with low
brightness and acidity (Lyman et al. 2003). Based on the these observations,
one can conclude that the quality of roasted coffee does not solely depend on the
physical parameters at the start and end point of roasting, but rather it is
dependent on the time-temperature conditions applied during the roasting
process.
Chemometrics employ statistical and mathematical techniques to convert
complex spectral and chromatographic data into information with reduced
dimensionality
to
facilitate
interpretation
(Socrates
1994).
Chemometric
methodologies, such as PCA, PCR, PLS, ANN have been successfully applied in
process monitoring, detection of product adulteration, quality evaluation,
screening of defective green coffee beans, and shelf-life study (Eliane Nabedryk
1982; Keller 1986; Socrates 1994; Dovbeshko 1997; Yoshida et al. 1997;
55
(GC-MS),
gas
chromatography
(GC),
and
sensory
array
instruments (electronic noses and tongues) have been used for studying the
aroma compounds in roasted coffee (Rahn & Konig 1978; Parliment & Stahl 1995;
Kantor & Fekete 2006), analyses involving these techniques are time-consuming
and some of these instruments are complicated. FTIR- ATR spectroscopy is a
simple technique which is rapid and provides an overall infrared fingerprint of the
specimen. Using an FTIR-ATR technique, Lyman et al. investigated 1800-1680
cm-1 region of the IR spectrum of coffee brews. The carbonyl stretching region
provided compositional information that can be used to correlate vinyl
esters/lactones, esters, aldehydes, ketones, and acids (Lyman et al. 2003). In our
previous study, FTIR-ATR spectroscopy was used as a rapid tool for
discriminating geographical origin of roasted coffee extracts and studying the
changes in chemical compositions when coffees were roasted to different degree
(Wang et al. 2011).
In the study, we applied a chemometric technique to analyze FTIR-ATR
fingerprints of coffee beans at different stages of roast. The objectives of this
study are to: (1) develop the understanding of the effects of time-temperature
conditions on the physical and chemical properties of coffee; (2) to analyze
coffees roasted to the same stage by different roasting profiles using FTIR-ATR
technique.
56
57
start-of-
end-of-
temperature
first-
first-
(C)
crack
crack
210
3.8
5.3
220
2.7
230
240
48s-
48s-
start-of-
end-of-
second-
second-
crack
crack
6.1
16.9
19.1
20.7
3.6
4.4
8.8
11.2
12
2.4
3.3
4.1
4.9
5.9
6.7
1.4
2.6
3.4
3.6
3.9
4.7
afterfirstcrack
aftersecondcrack
58
(SWFOH 1973). Samples of roasted beans were ground finely, and then dried at
103C for 5 h. Measurements were taken in triplicate.
5.2.5 pH Value
A 2.00 g ground coffee was accurately weighed into a 200 mL glass bottle,
and 100 mL of deionized water was added in. The glass bottle was boiled for 10
min. Then, 50 mL of the filtered extract was used for pH value determination with
a pH meter. Measurements were taken in triplicate.
5.2.6 Titratable Acidity
A 10.00 g of ground coffee was accurately weighed into a 200 mL glass
bottle, and 75 mL 80% ethanol was added to wet the sample. The glass bottle
was shaken for 16 h under magnetic stirring. After that, 50 mL of the filtered
extract was titrated against 0.1 N NaOH solution (Horwitz 2000). Measurements
were taken in triplicate.
5.2.7 Solvent Extraction and ATR-FTIR Analysis of Ground Coffee
After grinding, coffee grounds were extracted with ethyl acetate following
the extraction procedure: 0.2500 g of ground coffee was accurately weighed into
a glass vial, and 1 mL of deionized water was added to wet the sample. The
glass vial was shaken for 1 min; 1 mL of ethyl acetate was added, and the
mixture was shaken for an additional 5 min. The organic phase was used for
ATR-FTIR analysis in the region of 600 to 4000 cm-1 at 4 cm-1 resolution (Wang
et al. 2011). All extractions were performed in triplicate.
59
are often used to evaluate the progress of roast processing of coffee beans.
9
50
Accordingly, these events (i)were chosen 8as the reference sampling points in the
40
present study.
5.2
50
(i)
40
(iii)
30
0.80
9
0.60
0.40
0.00
(B)
C
C
C
C
(ii)
4
3
1.20
10
6.0
1.00
5.8
pH value
210
220
230
240
0.20
20
5.0
(iv)
5.6
Stage of roasting
5.4
(iii)
5.2
5.4
Lightness, L*
5.6
1.00
Moisture content, %
5.8
10
6.0
30
20
(ii)
(iv)
0.80
Stage of roasting
0.60
0.40
(B)
0.20
0.00
5.0
Stage of roasting
Stage of roasting
61
Lightness, L*
50
(i)
40
30
(iii)
1.00
0.80
9
0.60
0.40
0.20
20
0.00
10
6.0
(iv)
Moisture content, %
3
1.20
-2
10
Time, min
13
16
210
220
230
240
(B)
6
(ii)
C
C
C
C
5
4
3
19
-2
10
Time, min
1.20
(iii)
5.8
13
16
19
(iv)
1.00
pH value
5.6
5.4
5.2
Stage of roasting
5.0
4.8
-2
7
10
Time, min
13
16
19
0.80
0.60
Stage of roasting
0.40
0.20
0.00
-2
10
13
16
19
Time, min
62
values ranging between 25 and 28, regardless of the roast temperatures used. At
the start-of-second-crack, the L* value decreased to approximately 20, as the
beans attained dark roast. This result indicates that the first and second cracks
correlated well with lightness of the roasted coffee, and these milestones could
be used as one of the parameters to evaluate the degree of roast in coffee beans.
During the first phase of roasting (before the second crack), variation of
lightness between beans was observed. However, when the beans were roasted
to the start-of-second-crack and onwards, the lightness differences between
beans reduced. This observation can be attributed to the decreased rate of
lightness change as the beans were processed to darker roast. As shown in
Figure 13B(i), the slopes for L* value versus roast time plots decreased when the
beans were roasted from the start-of-second-crack onwards. Moreover, the
change in slope is more prominent for beans roasted at lower temperatures (e.g.,
210 and 220oC) than those processed at higher temperature (240oC).
degradation of chlorogenic acids, other organic acids, and lipids (Yeretzian et al.
2002). Beyond the start-of-second-crack, the decreasing moisture content
observed could be caused by the further rupturing of cell walls, allowing the
escape of water as the roasting continued. No significant difference (P<0.05) in
final moisture content was observed for the final products regardless the roaster
temperature used, averaged at 3.53-3.66% dry basis. Samples roasted at 240oC
had higher moisture contents at various stages of roast as compared to those
roasted at lower temperatures (Figure 15A(ii)). The observed higher moisture
contents for the former may be attributed to the limited diffusion of water from the
bean to the surrounding air due to the short processing time involved.
value was lower for samples roasted at higher temperature as compared to those
roasted at lower temperatures. This trend is likely due to faster rate of the
formation of acids (due to degradation of glucose, fructose, and sucrose) at
higher temperatures than that at lower temperatures. However, as the roasting
process continued to the start-of-second-crack, the trend was reversed due to the
greater destruction of formic and acetic acids at higher temperatures (Ginz et al.
2000).
Acidity is one of the attributes commonly associated with high quality
coffees. The perceived acidity of coffee is a result of proton donation of acids to
receptors on the human tongue. Many researchers observed a linear correlation
between the pH value and the perceived acidity of coffee (Sivetz & Desrosier
1979; Griffin & Blauch 1999). However, some recent findings on perception
transduction mechanisms of acid taste suggested that undissociated forms of
acid molecules are important for acid perception Therefore, titratable acidity in
coffee brews could be a more reliable indicator for correlating the coffee acidity
than the pH value (Voilley et al. 1981; Brollo et al. 2008). As shown in Figures
13A(iv) and 13B(iv), a rapid increase of titratable acidity was observed as green
coffee beans were roasted to the end-of-first-crack (medium roast). This change
in titratable acidity may be due to the formation of total aliphatic acids to a
maximum level (Clarke & Macrae 1988). Consistent with the pH value, further
roasting from a medium to dark roast resulted in a decrease in titratable acidity,
due to the destruction of organic acids (e.g., citric, malic, lactic, pyruvic, and
acetic acids) (Clarke 1986). In general, medium-roast Arabica coffee brews have
65
a pH value ranging between 4.9 and 5.2, which is in agreement with the result
from the present studies (pH 5.2, 5.16, 5.1, and 5.14 for 210, 220, 230, and
240C, respectively) (Clarke & Macrae 1988). These observations suggested that
by using different roast temperature and time profiles, the titratable acidity in
roast coffees could be manipulated. For example, the maximum titratable acidity
can be obtained by using 210oC roast temperature before second crack started,
and then using 240oC for continued roasting.
dependent.
In
general,
high-temperature-short-time
roasting
produced more soluble solids, less degradation of chlorogenic acids, lower loss
of volatiles, less burnt flavour, larger volume increase, larger CO2 desorption, and
higher oil migration, as compared to low-temperature-long time roasting process
(Nagaraju et al. 1997; Schenker 2000).
In the present study, coffee beans were processed at 210, 220, 230, or
240oC to achieve a similar degree of roast (all dark roast), based on the L* value.
Bean samples were collected at six roast stages, ground into powder, extracted
in solvent, and then analyzed with FTIR spectroscopy. Typical FTIR spectra of
coffee extracts are presented in Figure 14 (column C). In order to determine
spectral variances due to the temperature treatment, PCA was employed to
reduce the dimensionality of the IR spectra to facilitate the visualization of the
66
dataset. As shown in Figure 14 (column A), for all the temperatures tested, the
score plots displayed clusters that were separated according to different stages
of roast. There is a clear trend for the clusters to move upwards along the PC-2
axis as the roasting progressed through different stages. The clusters for green
coffee were separated far from the other clusters, indicating that the chemical
compositions within the green beans were considerably different as compared to
the roasted beans, as expected.
For PC1, no separation was observed between coffees roasted to different
stages when low roast temperatures were used (e.g., 210 oC). However, at higher
temperature (e.g., 240oC), the clusters had a tendency to spread along the PC1
axis as the roasting process progressed. Further analysis of PC1 loading plots
(data not shown) revealed that frequencies that contribute to the spectral
difference for beans roasted at 210 and 240oC occurred at around 2920-2850
cm-1, which is the absorbance range due to asymmetric and symmetric CH 2
stretching modes. This shows that important changes in aliphatic hydrocarbon
contents had occurred during roasting, especially the lipids. This is consistent
with the observation that during the roasting experiment, spots of oil were
observed on the surface of beans that were roasted at 240 oC, but not those
roasted at 210oC. The high-temperature-short-time process used at 240oC might
have increased the rate of oil diffusion from the bean core to the surface
(Puhlmann & Habel 1989).
To further investigate regions of spectra that contribute to the variance of
samples, the loading plots for PC2 were also inspected (Figure14, Column B).
67
The percent variance accounted by PC2 was indicated on each loading plot.
Regions of spectrum with large loading score (>0.1 and < -0.1) mainly appeared
at 2920, 2850, 1739, and 1660 cm-1, which are due to CH2 asymmetrical
stretching of methyl groups, C-H symmetrical stretching of methyl groups, C=O
stretching of polysaccharides/hemicelluloses, and C=C stretching band of lipids
and fatty acids, respectively (Shetty 2006; Hennessy et al. 2009). For coffees
roasted at 220 and 230C, more absorbance with large loading score (>0.1 and <
-0.1) were observed than those roasted at 210 and 240C, especially at 1741 cm1
(fatty acid esters), 1718-1707 cm-1 (ketones), 1697 cm-1 (aromatic acids), and
1514 cm-1 (amino groups). These compounds are important in determining the
overall coffee organoleptic qualities. For example, esters provide softer and
fruitier aromas, while aldehydes/ketones result in sharp odours, ranging from
woody, cucumber, cooked fruit, and nuts. On the other hand, acids are important
contributors to aromas similar to vinegar, chocolate, and burnt caramel attributes
(Ginz et al. 2000; Lyman et al. 2003). Silwar and Lllmann reported that the real
flavour of roasted coffee appeared at 220-230oC. Beyond this point, the flavor
was judged to be slightly over-roasted at 240oC and over-roasted when
processed at 250 to 260oC (Silwar & Lllmann 1993). For coffees roasted at
210oC, some key compounds that contribute to aroma such as furans, pyrazines,
strecker aldehydes, and 2- and 3-methylbutanal might have not been fully
developed (Silwar & Lllmann 1993; Baggenstoss et al. 2008).
68
0.005
0.003
0.001
0.12
0.005
0.003
0.001
-0.05
0.002
0.006
0.010
0.014
-0.15
4000
0.25
3200
2400
1600
800
PC 2 Loading (15.3%)
0.12
0.006
0.010
0.014
0.25
3200
2400
1600
800
2920
800
1718
1660
1697
1707
1739
1741
2850
1514
3200
2400
1600
Wavenumber, cm-1
0.20
0.16
0.15
0.12
2920
2850
-0.05
-0.005
-0.006 -0.002
0.002
0.006
0.01
0.014
-0.15
4000
0.25
2400
1600
0.00
4000
800
PC 2 Loading (19.9%)
0.006
0.01
0.014
2920
2850
0.12
0.05
0.08
0.002
3200
2400
1600
Wavenumber, cm-1
800
1660
0.16
0.15
-0.05
-0.005
-0.006 -0.002
1514
0.20
-0.001
-0.003
1718
1660
1697
1707
1739
1741
1545
0.04
3200
800
1556
0.08
-0.003
0.00
4000
PC 2 Loading (11.8%)
-0.001
0.001
3200
2400
1600
Wavenumber, cm-1
0.04
Absorbance
0.002
-0.15
4000
0.05
0.003
0.00
4000
0.16
0.15
-0.05
-0.005
-0.006 -0.002
0.005
1514
0.08
-0.003
0.001
2850
0.20
0.05
0.003
1660
1697
1739
1741
0.04
-0.001
0.005
2920
0.08
Absorbance
220 C
0.16
0.15
0.05
-0.005
-0.006 -0.002
230 C
0.20
-0.001
-0.003
240 C
0.25
PC 2 Loading (19.2%)
Absorbance
Absorbance
210 C
-0.15
4000
1739
2839
1514
0.04
3200
2400
1600
800
0.00
4000
3200
2400
1600
Wavenumber, cm-1
800
Figure14 PCA analysis for coffees during roasting. Column A: Two-factor score plots. Column B: Loading plots of PC2.
Column C: Representative FTIR spectra at the start-of-second-crack
69
70
0.20
1725-1705
1700-1680 1650-1600
1755-1740
1714-1705
0.16
start-of-first-crack
end-of-first-crack
48s-after-first-crack
Absorbance
1739-1724
2920-2850
start-of-second-crack
0.12
end-of-second-crack
48s-after-second-crack
0.08
1550
1780-1762
0.04
0.00
2950
2900
2850
2800
1800 1775 1750 1725 1700 1675 1650 1625 1600 1575 1550 1525 1500
Wavenumber, cm-1
Figure15 The expanded 2910-2850 cm-1, and 1800-1500 cm-1 regions of the spectra of coffee roasted at 230oC
71
72
1665
-0.05
-0.15
-0.25
4000
240 C-2
0.001
-0.05
1600
210 C-3
220 C-3
230 C-3
0.005
-0.25
4000
0.05
0.001
-0.05
-0.15
-0.003
-0.005
-0.25
4000
220 C-4
230 C-4
0.12
1556
1550
0.08
1541
0.04
3200
2400
1600
3649
0.00
4000
800
0.005
3200
2400
1600
Wavenumber, cm-1
0.16
0.12
1541
0.08
0.04
2400
1600
3649
0.00
4000
800
800
1514
1689-1691
1676
1699-1701
1660
1650
1566
1724-1726 1556
0.20
3200
800
1514
1691
1660
1650
0.16
0.15
0.003
210 C-4
2400
1600
Wavenumber, cm-1
PC 1 Loading (51.0%)
240 C-3
-0.003
-0.001
0.001
0.003
1st Principal Component
3200
0.20
Absorbance
-0.003
-0.001
0.001
0.003
1st Principal Component
2850
0.08
0.00
4000
800
-0.15
-0.001
3200
2400
1600
Wavenumber, cm-1
800
PC 1 Loading (47.5%)
240 C-4
0.20
0.15
1670
0.003
0.05
0.001
-0.05
-0.001
-0.15
-0.003
-0.005
-0.25
4000
0.005
-0.003
-0.001
0.001
0.003
1st Principal Component
210 C-5
220 C-5
230 C-5
0.005
240 C-5
3200
2400
1600
-0.001
-0.15
-0.003
-0.005
-0.25
4000
230 C-6
0.005
1514
2400
1600
Wavenumber, cm-1
800
1670 1643
1689
0.16
2920
0.12
1724
1514
1741 1540
2850
1550
0.08
0.04
3200
2400
1600
0.00
4000
800
3200
2400
1600
Wavenumber, cm-1
800
PC 1 Loading (45.0%)
240 C-6
0.20
0.15
0.003
0.05
0.001
-0.05
-0.001
-0.15
-0.003
-0.005
-0.25
4000
-0.003
-0.001
0.001
0.003
1st Principal Component
3200
0.20
-0.05
220 C-6
2850
0.08
0.00
4000
800
0.15
0.001
210 C-6
0.12
1724
1741
0.04
0.05
-0.003
-0.001
0.001
0.003
1st Principal Component
2920
PC 1 Loading (45.7%)
0.003
0.005
0.16
Absorbance
2400
0.15
0.05
0.005
3200
Absorbance
230 C-2
1741
0.12
PC 1 Loading (68.4%)
-0.001
48s-after-first-crack
220 C-2
2920
0.04
Absorbance
end-of-first-crack
210 C-2
0.005
0.003
0.005
start-of-second-crack
-0.003
-0.001
0.001
0.003
1st Principal Component
0.16
Absorbance
0.001
-0.003
-0.005
0.20
0.15
-0.001
PC 1 Loading (47.5%)
240 C-1
0.05
-0.003
-0.005
end-of-second-crack
B
230 C-1
0.003
0.005
48s-after-second-crack
220 C-1
0.005
0.16
Absorbance
start-of-first-crack
0.005
210 C-1
2920
1670
1689
1724
1741
0.12
2850
1514
0.08
0.04
3200
2400
1600
800
0.00
4000
3200
2400
1600
Wavenumber, cm-1
800
Figure16 PCA analysis for coffees collected at the same sampling point. Row A:
Two-factor score plots. Row B: Loading plots of PC1. Row C: Representative
FTIR spectra at 230oC
73
pore pressure that causes the pores to compress against each others, likely due
to the evolution of carbon dioxide. The carbon dioxide build up may have also
resulted in ruptured walls exhibited for some pores. Further heating to 48 s-aftersecond-crack resulted in morphologies that were less uniform than the previous
stages, possibly due to the artefacts caused by the disruption of brittle cell wall
during sample preparation. Further examination of micrographs for samples
processed to the last two roasting stages revealed the presence of droplets on
the surface of the pore walls. The droplets may be attributable to the coalesced
oil that formed during the extended roasting.
In general, high-temperature-short-time roasting processes tend to produce
beans of higher porous structure in the bean cell tissues as compared to lowtemperature-long-time processes. The former processing condition is also known
to result in higher extraction yield during solvent extraction (Pittia et al. 2001).
However, at the magnification level employed during the SEM analysis and
roasting conditions used, no correlation can be made between the pore size and
roasting temperature. The lack of correlation could be due to the sample cutting
procedure involved, which does not guarantee the sectioning through the
maximum diameter of pores. Density measurement will provide a more accurate
assessment of sample porosity.
75
Figure 17 SEM micrographs of internal texture for coffee beans collected at different stages of roast. The temperatures
indicated on each row of micrographs were the roast temperature
76
In summary, this study showed that coffee beans processed to the same
extent of roast (as determined from bean lightness and roast cracking milestones)
by using different time-temperature roasting profiles does not necessarily
produce coffees of similar physical and chemical properties. FTIR-ATR spectra of
ground coffee obtained by different time-temperature combinations can be useful
for studying the effect of roasting conditions on IR-active compounds. The FTIRchemometric technique adopted in this study could serve as a rapid tool for
discriminating coffees with different roasting degrees, and coffees roasted by
different roasting profiles.
77
79
81
7 REFERENCE
ANIL, K. D. DAVID, C. S. & MICHAEL, T. (2004). Applications of electronic noses
and tongues in food analysis. International Journal of Food Science &
Technology 39(6), 587-604.
BAGGENSTOSS, J. POISSON, L. KAEGI, R. PERREN, R. & ESCHER, F.
(2008). Coffee Roasting and Aroma Formation: Application of Different Timetemperature Conditions. Journal of Agricultural and Food Chemistry 56(14),
5836-5846.
BALZER, H. H. (2001). Acids in coffee.
BANKS, M. (2002). The World Encyclopedia of Coffee. London: Anness.
BARTER, R. (2004). A short introduction to the theory and practice of profile
roasting. Tea and Coffee Trade Journal 68, 34-37.
BARUA, A. G. HAZARIKA, S. HUSSAIN, M. & MISRA, A. K. (2008).
Spectroscopic Investigation of the Cashew Nut Kernel (Anacardium occidentale).
The Open Food Science Journal 2, 85-88.
BELITZ, H.-D. GROSCH, W. & SCHIEBERLE, P. (2009). Coffee, Tea, Cocoa.
Vatican Springer Berlin Heidelberg.
BELL, L. N. WETZEL, C. R. & GRAND, A. N. (1996). Caffeine content in coffee
as influenced by grinding and brewing techniques. Food Research International
29(8), 785-789.
BELLAMY, L. J. (1975). The Infrared Spectra of Complex Molecules. London,
England: Chapman & Hall Ltd.
BERTRAND, B. VILLARREAL, D. LAFFARGUE, A. POSADA, H. LASHERMES,
P. & DUSSERT, S. (2008). Comparison of the Effectiveness of Fatty Acids,
Chlorogenic Acids, and Elements for the Chemometric Discrimination of Coffee
(Coffea arabica L.) Varieties and Growing Origins. Journal of Agricultural and
Food Chemistry 56(6), 2273-2280.
BETANCOURT, L. E. & FRANK, H. K. (1983). Bedingungen des mikrobiellen
Verderbs von grunem Kaffee. Deutsche Lebensmittel-Rundschau 79, 366-369.
BICCHI, C. P. PANERO, O. M. PELLEGRINO, G. M. & VANNI, A. C. (1997).
Characterization of Roasted Coffee and Coffee Beverages by Solid Phase
Microextraction-gas Chromatography and Principal Component Analysis. Journal
of Agricultural and Food Chemistry 45(12), 4680-4686.
82
AND
85
86
87
89
90
91