physicochemical changes of coffee beans during roasting

Physicochemical Changes of Coffee Beans During Roasting

by

Niya Wang

A Thesis

Presented to

The University of Guelph

In partial fulfilment of requirements

for the degree of

Master of Science

in

Food Science

Guelph, Ontario, Canada

© Niya Wang, April, 2012

ABSTRACT

PHYSICOCHEMICAL CHANGES OF COFFEE BEANS DURING

ROASTING

Niya Wang Advisor:

University of Guelph, 2012 Professor Loong-Tak Lim

In this research, physicochemical changes that took place during roast

processing of coffee beans using fluidized air roaster were studied. The results

showed that high-temperature-short-time resulted in higher moisture content,

higher pH value, higher titratable acidity, higher porous structure in the bean cell

tissues, and also produced more aldehydes, ketones, aliphatic acids, aromatic

acids, and caffeine than those processed at low-temperature-long-time process.

Fourier transform infrared (FTIR) spectroscopy and chemometric analysis

showed that clusters for principal components score plots of ground coffee,

extracted by a mixture of equal volume of ethyl acetate and water, were well

separated. The research indicated that variations in IR-active components in the

coffee extracts due to different stages of roast, roasting profiles, and

geographical origins can be evaluated by the FTIR technique.

iii

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.

iv

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

TRANSFORM INFRARED SPECTROSCOPY ................................................... 27

4.1 INTRODUCTION ............................................................................................. 27

4.2 MATERIALS AND METHODS ............................................................................ 29

4.2.1 Chemicals ............................................................................................ 29

4.2.2 Coffee Beans and Roasting Conditions ............................................... 29

4.2.3 Degree of Roast as Determined by Color Measurements ................... 30

4.2.4 Solvent Extraction of Ground Coffee ................................................... 30

4.2.5 ATR-FTIR Analysis .............................................................................. 31

4.2.6 Data Analysis ...................................................................................... 32

4.3 RESULTS AND DISCUSSIONS .......................................................................... 32

4.3.1 Optimization of Solvent Extraction for FTIR-ATR ................................ 33

4.3.2 Color Analysis ..................................................................................... 38

v

4.3.3 PCA Analysis of Solvent Extracts of Coffee Beans ............................. 40

4.3.4 PCA Analysis for Coffees According to Degree of Roast .................... 47

4.3.5 SIMCA Analysis ................................................................................... 52

5 EFFECTS OF DIFFERENT TIME-TEMPERATURE PROFILES ON COFFEE

PHYSICAL AND CHEMICAL PROPERTIES ...................................................... 54

5.1 INTRODUCTION ............................................................................................. 54

5.2 MATERIALS AND METHODS ............................................................................ 57

5.2.1 Chemicals and materials ..................................................................... 57

5.2.2 Green Beans and Roasting Conditions ............................................... 57

5.2.3 Degree of Roast as Determined by Color Measurements ................... 58

5.2.4 Moisture Content of Ground Coffee ..................................................... 58

5.2.5 pH Value .............................................................................................. 59

5.2.6 Titratable Acidity .................................................................................. 59

5.2.7 Solvent Extraction and ATR-FTIR Analysis of Ground Coffee ............. 59

5.2.8 Chemometric Analysis ......................................................................... 60

5.2.9 Scanning Electron Microscopy (SEM) Analysis ................................... 60

5.3 RESULTS AND DISCUSSION ........................................................................... 60

5.3.1 Evolution of physical and chemical properties during roasting ............ 60

5.3.2 Changes in coffee at various stages of roast ....................................... 66

5.3.3 Effects of roast temperature on changes in coffee .............................. 72

5.3.4 Microstructural analysis ....................................................................... 74

6 CONCLUSIONS AND FUTURE WORKS ........................................................ 78

7 REFERENCE ................................................................................................... 82

vi

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:


Figure 11 PCA of FTIR data for dichloromethane extracts of coffee (from the

vii

same origin) with two degrees of roast. Row A: Two factor score plots. Row B:


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

viii

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

ix

LIST OF ABBREVIATIONS

FTIR Fourier transform infrared

ATR Attenuated total reflectance

PTR-MS Proton transfer reaction-Mass spectrometry

PAS Photoacoustic spectroscopy

PCA Principal component analysis

HCA Hierarchical cluster analysis

PLS Partial least squares

PCR Principal component regression

PLS-DA Partial least squares-discriminant analysis

KNN K-nearest neighbour

SIMCA Soft independent modeling of class analogy

PCs Principal components

HS-SPME Headspace solid phase microextraction

NMR Nuclear magnetic resonance

GC-MS Gas chromatography-mass spectrometry

GC Gas chromatography

L* Lightness

SEM Scanning electron microscope

1

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

2

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 time-

temperature 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.

3

Different temperature profiles affect dehydration and the chemical reaction

conditions in the bean which control gas formation, browning and flavour

development. In general, the use of roasting temperature of greater than 200°C

is required in order to result in desirable chemical, physical, structural, and

sensorial changes in the coffee beans (Clarke & Macrae 1988; Schenker 2000;

Schenker et al. 2002; Baggenstoss et al. 2008). Color change and weight loss

are frequently used as a measure of the degree of roast, and both are directly

related to the final roasting temperature (Sivetz 1991; Illy & Viani 1995). Other

methods, such as the ratios of free amino acids (Nehring & Maier 1992), and

chlorogenic acids content (Illy & Viani 1995) have also been used.

Researchers have reported the effects of time-temperature profile on coffee

aroma properties. In general, low-temperature-long time roast processes result

in sour, grassy, woody, and underdeveloped flavour properties. In comparison,

high-temperature-short-time produced the higher quality coffee in terms of

producing more aroma volatiles and higher brew yield (Schenker et al. 2002;

Lyman et al. 2003). Reviewing these and other literature, one can conclude that

the complex changes in coffee during roasting do not solely depend on physical

parameters at the start and end point of the thermal process, but rather a path-

dependent phenomenon. Therefore, to gain insight into the changes of

physicochemical properties of coffee during roasting, the green beans must be

roasted under controlled conditions.

The overall objective of this study is to apply a chemometric technique, in

conjunction with Fourier transform infrared spectroscopy, to elucidate the effects

4

of time-temperature effects on physical and chemical properties of coffee from

different grown regions during fluidized-bed roasting.

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 world’s 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

5

command higher prices (Valdenebro et al. 1999). Table 1 provides a general

survey on the chemical composition of green Arabica and Robusta coffee beans

(Illy & Viani 1995).

Coffee cherries are harvested when they become bright-red, glossy, and

firm, either by selective hand-picking or non-selective stripping of whole branches

or mechanical harvesting. The hand-picking method is very time-consuming, but

results in a superior product quality because only ripe cherries are selected. After

harvesting, the coffee fruits are separated from the pulp, which is carried out by

dry or wet processing (Clarke & Macrae 1987; Illy & Viani 1995). The dry process

is simple and inexpensive. The whole cherries are dried under the sun in open air,

followed by the separation of the hull (dried pulp and parchment) mechanically to

yield the green beans. On the contrary, the wet process requires greater

investment and more care, but results in a superior coffee quality. In the wet

process, the pulp of the coffee cherries, which is made up of exocarp and

mesocarp, is removed mechanically, but the parchment remains attached to the

beans. After drying either under the sun or in a dryer, the parchment is removed

to produce the green coffee beans. Bean size, color, shape, processing method,

crop year, and presence of defects, are some of the parameters used to evaluate

the quality of green coffee beans (Banks 2002).

Green coffee beans have minimal flavour. However, upon roasting,

characteristic coffee aroma is developed due to the complex reactions that take

place in the beans. To develop a unique flavour, green coffee beans are roasted

according to different time-temperature profiles (Buffo & Cardelli-Freire 2004).

6

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 sulphur-

containing 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).

7

Table 1 Chemical composition of green Arabica and Robusta coffee beans

(g/100g)

Component 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

Minerals (as oxide ash) 4.2 4.4

Volatile aroma traces traces

Water 8 to 12 8 to 12

After harvesting, green coffee beans should be dried to 10-14.5% moisture

content and stored below 26oC under dry environment (50-75% RH) to maintain

the bean quality and to prevent the growth of mould (Gopalakrishna Rao et al.

1971; Kulaba 1981; Betancourt & Frank 1983). Under optimal storage conditions,

8

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).

2.2 Roasting of coffee beans

Green coffee beans provide neither the characteristic aroma nor flavour of

brewed coffee until they are roasted. Moreover, the roasting process increases

the value of coffee beans, by 100-300% of the raw material (Yeretzian et al.

2002). Roasting of coffee beans typically takes place at 200-240°C for different

times depending on the desired characteristics of the final product. Events that

take place during roasting are complex, resulting in the destruction of some

compounds initially present in green beans and the formation of volatile

compounds that are important contributors to the characteristic of coffee’s aroma.

The chemical compositions of green, roasted, and brewed coffee are shown in

Figure 1 (Barter 2004).

9

Figure 1 Chemical composition of green, roasted, and brewed coffee (Barter

2004)

Briefly, as temperature increases to about 100oC, green coffee beans

undergo moisture loss from 8-12% in green coffee beans to about 5% in the

roasted coffee beans (Illy & Viani 1998). The smell of the beans changes from

herb-like green bean aroma to bread-like, the color turns from green to yellowish,

and the structure changes from strength and toughness to more crumbly and

brittle. When the internal temperature of beans reaches 100oC, the color

darkened slightly for about 20-60 s due to the vaporization of water. At 160-

170oC, the beans become lighter in color for about 60-100 s. As roasting

cellulose(non Hyd)

18%

cellulose(Hyd) 13%

starches and pectins

13%

soluble carbohydrates

9%

water12%

non volatile acids7%caffeine

1%

protein12%

ash 3% oil

11%

trigonelline1%

Green coffee beans

caffeine1%

water2%

starches and pectins

14%

CO2

2%

cellulose(Hyd) 14%

cellulose(non Hyd)

17%trigonelline

1%

oil 13%

ash 4%

protein13%

non volatile acids7%


10%

Roasted coffee beans

oil 1%


37%

non volatile acids31%

caffeine6%

protein5%

ash 16%

trigonelline4%

Brewed solubles

10

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

11

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 240oC

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)

12

studied changes in chemical composition of Arabica coffee roasted at 230°C at

different process times. They reported that as the roasting time increased, the

following changes occurred: (1) sugar contents first increased, and then

decreased; (2) minimal change in caffeine content; (3) proteins decreased

continuously; (4) free fatty acids increased; and (5) unsaponifiable compounds

decreased (Fobe et al. 1968).

2.3 Changes in Chemical Compositions during Roasting

Roasting causes a net loss of matters in the forms of CO2, water vapor, and

volatile compounds. Moreover, degradation of polysaccharides, sugars, amino

acids and chlorogenic acids also occurred, resulting in the formation of

caramelization and condensation products. Overall, there is an increase in

organic acids and lipids, while caffeine and trigonelline (N-methyl nicotinic acid)

contents remained almost unchanged (Buffo & Cardelli-Freire 2004). The main

acids present in green beans are citric, malic, chlorogenic, and quinic acids.

During roasting the first three acids decrease while quinic acid increases as a

result of the degradation of chlorogenic acids (Ginz et al. 2000). Formic and

acetic acids yields increase up to the medium roasting degree and then begin to

fall as roasting is continued. According to Balzer (Balzer 2001), a rapid increase

in titratable acidity during roasting was observed from green to medium roast,

followed by a smaller decrease as roasting proceeded.

13

The reaction products formed are highly dependent on the roasting time-

temperature 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 (PTR-

MS) 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 180oC

(Yeretzian et al. 2002). Similar observations were observed by Hashim and

Chaveron, who concluded that methylpyrazine may be used as an indicator to

14

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.

2.4 Flavour compounds in roasted coffee

Chemical compounds present in roasted coffee can be roughly grouped into

volatile and non-volatile, some of the former being responsible for aroma and the

latter for the basic taste sensations of sourness, bitterness and astringency

(Buffo & Cardelli-Freire 2004). Russwurm reported that carbohydrates, proteins,

peptides and free amino acids, polyamines and tryptamines, lipids, phenolic

acids, trigonelline, and various non-volatile acids in the green coffee beans were

involved in the flavour formation during roasting (Russwurm 1970). For example,

chlorogenic acid contributes to body and astringency; sucrose contributes to

15

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).

16

Table 2 Potent odorants in Arabica coffee from Colombia

Sweet/caramel-like group

Methylpropanal

2-Methylbutanal

3-Methylbutanal

2,3-Butandione

2,3-Pentandione

4-Hydroxy-2,5-dimethyl-3(2H)-furanone

(HD3F)

5-Ethyl-4-hydroxy-2-methyl-3(2H)-

furanone (EHM3F)

Vanillin

Sulfurous/roasty group

2-Furfurylthiol

2-Methyl-3-furanthiol

Methional

3-Mercapto-3-methylbutyl-formiate

3-Methyl-2-butene-1-thiol

Methanethiol

Dimethyltrisulfide

Earthy group

2-Ethyl-3,5-dimethylpyrazine

2-Ethenyl-3,5-dimethylpyrazine

2,3-Diethyl-5-methylpyrazine

2-Ethenyl-3-ethyl-5-methylpyrazine

3-Isobutyl-2-methoxy-pyrazine

Smoky/phenolic group

Guaiacol

4-Ethylguaiacol

4-Vinylguaiacol

Fruity group

Acetaldehyde

Propanal

(E)-β-Damascenone

Spicy group

3-Hydroxy-4,5-dimethyl-3(5H)-furanone

(HD2F)

5-Ethyl-3-hydroxy-4-methyl-2(5H)-

furanone(EHM2F)

The non-volatile components in roasted coffee are made up of mainly the

following:

17

(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-O-

methylcafestol, is present in Robusta but not Arabica coffee, making it a suitable

18

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

19

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).

2.5 Fourier transform infrared (FTIR) spectroscopy

FTIR spectroscopy is a powerful tool for identifying types of chemical bonds

in a molecule by producing an infrared (IR) absorption spectrum. Interferometer

is one of the key components in a FTIR spectrometer. It consists of IR light

source, fixed mirror, moving mirror, beam splitter, and detector (Figure 2). The

principle of the FTIR spectroscopy is that the beam splitter splits the light beam

from the IR source and sends half of the IR radiation to the fixed mirror and the

other half to the moving mirror. The split beams recombine to form overlapping

radiation waves that interact with the sample, resulting in an infrared spectrum.

20

Figure 2 Schematic diagram of a typical FTIR spectrometer

The radiation emerging from an IR source passes through the

interferometer and to a sample before reaching a detector. Upon amplification of

signal, the data are transformed to the digital type by an analog-to-digital

converter and transferred to a computer for Fourier-transformation. FTIR

measures the absorbance of IR active species over a range of wavenumbers in

the IR region that are absorbed by a material. IR spectral regions can be divided

into three parts, which are near-IR (13000-4000 cm-1), mid-IR (4000-400 cm-1),

and far-IR (400-10 cm-1). The bonds involved in the near-IR are usually due to C-

H, N-H or O-H stretching. Typical vibrational absorbance for common bonds in

the mid-IR is shown in Figure 3.

Figure 3 Vibrational absorbance due to common bands

Sampling methods in FTIR include transmission, reflectance, and micro-

sampling (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

21

difficult to analyze by transmission method. Attenuated total reflectance (ATR)

spectroscopy uses total internal reflection phenomenon to analyze a sample. In

many applications, it successfully replaces constant path transmission cells and

salt plates used for the analysis of liquid and semi-liquid materials. Because of

the reproducible effective path length, ATR is well suited for both qualitative and

quantitative applications. Some other spectroscopy such as specular reflectance

spectroscopy, diffuse reflectance spectroscopy, and photoacoustic spectroscopy

(PAS) are also very useful in analyzing samples. Micro-sampling method is used

for very small samples (microgram or microlitre) by the help of an IR microscope.

If a microscope facility is not available, some other special sampling accessories

such as a beam condenser or a diamond anvil cell can be used (Stuart 2003).

Various FTIR techniques have been used for coffee research. For instance,

FTIR has been used for caffeine determination in roasted coffee in the mid-IR

range (Garrigues et al. 2000; Ohnsmann et al. 2002), for discrimination of coffee

varieties (Kemsley et al. 1995; Briandet et al. 1996b; Garrigues et al. 2000), and

for detection of adulteration in instant coffees by sugars, starch, or chicory

(Briandet et al. 1996a). Moreover, FTIR-ATR has been successfully used in the

analysis of brewed coffee to study the effects of roasting conditions on coffee

aroma. Lyman et al. investigated the 1800-1680 cm-1 region of IR spectrum,

which contains carbonyl vibration bands that can be used to correlate vinyl

esters/lactones, esters, aldehydes, ketones, and acids (Lyman et al. 2003).

2.6 Chemometrics

22

Chemometrics can be generally described as the application of

mathematical and statistical methods to improve chemical measurement

processes, and extract more useful information from chemical and physical

measurement data (Workman et al. 1996; Paul 2006; Fu 2011).

In general, there are three categories of chemometric analysis (InfoMetrix

2010):

(1) Exploratory data analysis is often used to reveal hidden patterns in

complex data by reducing the information to a more comprehensible form, to

expose possible outliers, and to indicate whether there are patterns or trends in

the dataset. Principal component analysis (PCA) and hierarchical cluster analysis

(HCA) are some of the exploratory algorithms.

(2) Continuous property regression is used to develop calibration models

that correlate the information in a set of known measurements to the property of

interest. Partial least squares (PLS) and principal component regression (PCR)

are two algorithms commonly used for regression and are designed to avoid

problems associated with noise in the data.

(3) Classification modeling is applied in scenarios where samples are

required to be classified into predefined categories or "classes". A classification

model is used to assign a sample's class by comparing the sample to a

previously analyzed data set, for which its categories are already known. PLS

discriminant analysis (PLS-DA), k-nearest neighbor (KNN) and soft independent

modeling of class analogy (SIMCA) are some of the primary chemometric

workhorses in classification modeling.

23

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.

24

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).

Figure 4 Schematic diagram of PCA analysis

Many researchers have used chemometrics to study various phenomena in

coffee. Briandet and others adopted PCA to analyze FTIR spectra of coffee

extracts. They showed that 100% correct classifications for both training and test

samples for Arabica and Robusta in Instant Coffee. They also applied PLS to

predict the relative Arabica and Robusta contents in their coffee samples by

analyzing the FTIR spectra (Briandet et al. 1996a). Bicchi and others

characterized different roasted coffees and coffee beverages by applying PCA to

25

chromatographic data obtained by headspace solid phase microextraction (HS-

SPME), 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.

26

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:

To analyze coffee from different geographical origins (Colombia, Costa

Rica, Ethiopia, and Kenya) processed to medium and dark roasts, using

FTIR spectroscopy and chemometric analysis.

Developing the understanding of the effects of time-temperature

conditions on the physicochemical properties (color, moisture contents, pH,

titratable acidity, and microstructure) of coffee from Brazil.

To study the physicochemical changes (color, moisture contents, pH,

titratable acidity, and microstructure) in coffee beans at different stages of

roast using FTIR spectroscopy and chemometric analysis.

27

4 FEASIBILITY STUDY ON CHEMOMETRIC DISCRIMINATION OF

ROASTED ARABICA COFFEES BY SOLVENT EXTRACTION AND

FOURIER TRANSFORM INFRARED SPECTROSCOPY

4.1 Introduction

Coffee is one of the most popular beverages in the world due to its unique

aroma, taste, and stimulating effects of caffeine. The quality of brewed coffee is

affected by many parameters. Depending on the species (Arabica, Robusta, or

Liberica) and method used to process the coffee cherries (dry vs wet), the overall

quality and chemical composition of coffee bean can vary considerably. By and

large, the Arabica coffees have more pronounced and finer flavor profiles that are

considered better quality and, accordingly, command a higher price than the

Robusta and Liberica coffees (Davis 2001). The composition of the soil and its

fertilization, the altitude and weather of the plantation, and the final cultivation

and drying methods used will all affect the green bean quality (Costa Freitas &

Mosca 1999). Roasting, the final processing step before grinding and brewing,

ultimately determines the organoleptic properties of the coffee beverage. During

the roasting process, the reactions that occur in the coffee bean are complex and

strongly dependent on the time-temperature profile used (Lyman et al. 2003;

Baggenstoss et al. 2008).

Grading of whole roasted coffee beans is relatively easy as compared to

ground coffee due to the presence of visual clues in the former (size, shape,

defect, etc.). By contrast, these indicators are absent for ground coffees;

28

therefore, sample discriminating can be difficult. Often time, sensory evaluation

and cupping are needed (Martin et al. 1999). Analytical methods have been

successfully used for component analysis of coffee, including mineral contents

(Krivan et al. 1993; Martin et al. 1999), volatile compounds (Silvia et al. 2004),

chlorogenic acids (Maria et al. 1998), fatty acids (Bertrand et al. 2008), and

amino acid enantiomers (Casal et al. 2003).

FTIR spectroscopy is a rapid and non-destructive technique that has been

used for investigating covalent bond vibration in coffee. This method has been

used to determine the caffeine content in roasted coffee (Garrigues et al. 2000;

Ohnsmann et al. 2002), to discriminate coffee varieties (Kemsley et al. 1995;

Briandet et al. 1996b; Garrigues et al. 2000), and to detect adulteration in instant

coffees (Briandet et al. 1996a). Because of the complexity of FTIR spectral data,

chemometric analysis [e.g., PCA and SIMCA] is often used to reduce the

dimensionality of spectral data to aid the extraction of useful information,

identification of natural data trends, and classification unknown samples

(Hendriks et al. 2005). Chemometric analysis has been successfully applied to

analyze FTIR spectral data of coffee, for instance in the chemical discrimination

of Arabica and Robusta coffees (Briandet et al. 1996b), quality control and

authentication of instant coffees (Charlton et al. 2002), and adulteration detection

of freeze-dried instant coffees (Briandet et al. 1996a).

In this study, we employed ATR-FTIR to analyze coffee extracts prepared

using six organic solvents (dichloromethane, ethyl acetate, hexane, acetone,

ethanol, and acetic acid). Our objective was to investigate the feasibility of using

29

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).

4.2 Materials and Methods

4.2.1 Chemicals

Hexane was purchased from Sigma-Aldrich Ltd. (St. Louis, MO).

Dichloromethane, ethyl acetate, acetone, and acetic acid were purchased from

Fisher Scientific (Ottawa, Canada). Ethanol was purchased from Greenfield

Ethanol Inc. (Brampton, Canada).

4.2.2 Coffee Beans and Roasting Conditions

Wet-processed green coffee beans (Arabica variety) from Colombia, Costa

Rica, Kenya, and Ethiopia were purchased from Green Beanery (Toronto,

Canada). Green coffee beans (45 g) were roasted in a fluidized bed hot air

roaster (Fresh Roast SR 500, Fresh Beans Inc., Park City, UT). Two isothermal

roasting programs were used for preparing dark and medium roast coffees

(Figure 5). The roasted beans were stored in hermetic glass bottles in the dark at

15°C before grinding.

30

Figure 5 Air temperature (in roast chamber) profiles of the fluidized bed hot air

coffee roaster

4.2.3 Degree of Roast as Determined by Color Measurements

Roasted coffee beans were ground using a coffee grinder (Bodum Antigua

Electric Burr Grinder, Bodum, Inc., Copenhagen, Denmark) at the medium grind

setting. The color of the ground coffee was measured in the L*, a*, b* system

using a Konica Minolta CM-3500d spectrophotometer (Konica Minolta Sensing,

Inc., Osaka, Japan) in the reflectance mode. Before analysis, the instrument was

calibrated on a white standard tile. Measurements were taken in triplicate.

4.2.4 Solvent Extraction of Ground Coffee

After grinding, coffee grounds were extracted with dichloromethane, ethyl

acetate, hexane, acetone, ethanol, or acetic acid, following two extraction

25

50

75

100

125

150

175

200

225

250

0 50 100 150 200 250 300 350 400 450 500 550

Ro

asti

ng

Tem

pera

ture

, oC

Roasting Time, S

Medium roast profile

Dark roast profile

31

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 Prestige-

21; 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 μL) 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 cm−1 resolution. Each

spectrum was an average of 20 scans. For each extract, 3 FTIR spectra

32

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.r-

project.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.

4.3 Results and Discussions

33

4.3.1 Optimization of Solvent Extraction for FTIR-ATR

From the preliminary experiments, we observed that wetting of coffee grinds

prior to solvent treatment was necessary to enhance extraction. The effects of

deionised water, NH4OH solutions (0.25, 0.5, 0.75, and 2.5 M), and HCl solutions

(0.05, 0.1, 0.15, and 0.25 M) on coffee wetting were evaluated at different

volumes (0.5, 1.0, and 1.5 mL). The FTIR spectra data showed that there was no

significant difference between alkaline, acid, and neutral wetted samples (data

not shown). There was no difference between adding 0.5 mL and 1.0 mL

solutions, but spectral intensities were weakened when 1.5 mL of aqueous

solutions were added, probably due to the dilution effect. On the basis of the

positive preliminary experiment findings, 1.0 mL water was added during ground

coffee extraction.

The effects of water on the solvent extraction have been reported previously.

Yamamoto et al. studied the efficiency of various organic solvents for cadmium

dithizonate extraction from an aliquot of cadmium solution, and found that an

enhanced extraction efficiency of extraction was achieved by using a 50%

aqueous phase/solvent volume ratio (Yamamoto et al. 1972). Murphy et al.

reported that solvent at 53% (organic phase to water) was the most efficient in

maximizing the isoflavone extraction in soy food matrix (Murphy et al. 1999).

During the extraction of ground coffee, different phase separation

behaviours were observed among the tested solvents due to their different

physical properties (Table 3). When extraction was carried out in the presence of

water (Method #1), the solvent phase of dichloromethane was at the bottom,

34

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.

Table 3 Physical properties of the investigated solvents (Pagni 2005)

Solvent Solubility in water, at

20°C

Polarity

index (P)

Density,

g/mL

Dichloromethane Immiscible (1.3 v/v) 3.1 1.326

Hexane Immiscible(0.0013 v/v) 0.1 0.659

Ethyl acetate Slightly soluble (8 v/v) 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

35

Dichloromethane Extract

Hexane Extract

Ethyl acetate Extract

Acetone Extract

Ethanol Extract

Acetic acid Extract

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)

In order to increase the sensitivity of FTIR analysis, extracts were allowed

to evaporate on the ATR crystal to remove the solvent. The required times for

completing evaporation of solvent are summarized in Table 4. As shown,

dichloromethane, and hexane had relatively short evaporation times due to their

immiscibility with water. Intermediate evaporation time was observed for ethyl

acetate since it is partially soluble in water. By contrast, acetone, ethanol, and

acetic acid are completely miscible with water, resulting in relatively long

evaporation times. The extended evaporation time for these solvents may not be

desirable due to the potential loss of coffee volatile compounds (Mottaleb et al.

1997) .

36

Table 4 Evaporation time of the coffee extracts

Evaporation time (s)

Solvent With H2O No H2O

Dichloromethane extract 60 60

Hexane extract 60 60

Ethyl acetate extract 180 180

Acetone extract 600 60

Ethanol extract 760 280

Acetic acid extract 840 780

Selected FTIR spectra of solvent extracts obtained by methods #1 (with

water) and #2 (no water) are shown in Figure 7. The 3100 to 2750 cm-1 region in

the majority of spectra (except acetic acid, acetone, and ethanol extracts

obtained with extraction method #1) were typical for the fatty acid moiety of lipids

due to asymmetrical C-H stretching (2920 cm-1), symmetrical C-H stretching

(2850 cm-1), and methylene asymmetrical stretching band (weak shoulder at

2954 cm-1) (Innawong et al. 2004). In the presence of water, the absorbance

around 3676-3028 cm-1 for acetic acid, acetone, and ethanol extracts can be

attributed to the O-H stretching band. The 1800–800 cm-1 region contained

absorbance bands due to C=O (ester, aldehydes, and ketones) stretching, C-H

37

(methylene) bending (scissoring), and C-O (esters and alcohol), CH2

stretching/bending (Innawong et al. 2004). These regions contained fingerprint

information that may be important for discriminating coffee samples from different

origins.

Spectra from method #1 extracts were relatively more complex than those

from method #2 extracts, especially when dichloromethane and ethyl acetate

were used for extraction. For instance, dichloromethane extract from method #1

resulted in many additional peaks that were absent for those from method #2,

including 1487 cm-1 (C=C, C-H deformation), 1398 cm-1 (CH3 symmetric

deformation), 1323 cm-1 (symmetric vibrations of COO- groups), and 1284 cm-1

(Amide III band components of proteins) (Movasaghi et al. 2008). In terms of

band shape and intensity, different spectral features were observed in the 1720-

1203 and 1064-940 cm-1 regions. With method #1, water-induced swelling of the

coffee particles might have facilitated the extraction of additional compounds. A

similar enhancement in spectral features was observed for the dichloromethane

and ethyl acetate coffee extracts. For the hexane and acetic acid extracts,

minimal spectral differences were observed between methods #1 and #2. The IR

spectra of the hexane extracts were similar to lipid (Hennessy et al. 2009)

indicating that lipids may be the main components extracted when hexane was

used as a solvent. Overall absorbance values were considerably stronger for the

acetone and ethanol extracts probably due to the contribution from water present

in the extracts. The spectra of acetic acid extracts and pure acetic acid were

similar (data not shown), indicating that acetic acid is not an effective solvent for

38

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.

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)

4.3.2 Color Analysis

Ground coffee samples from different geographical regions could not be

distinguished readily by visual inspection. The L* (lightness) values of ground

coffee beans from different geographic regions (Colombia, Costa Rica, Ethiopia,

and Kenya) were similar among medium roast or dark roast samples (Table 5).

0.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

3.6

60012001800240030003600

Ab

so

rban

ce

Wavenumber, cm-1

With water

80010001200140016001800Wavenumber, cm-1

Enlarge region (1800-800 cm-1)

No water

No water

With water

Hexane extract

0.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

3.6

60012001800240030003600

Ab

so

rban

ce

Wavenumber, cm-1

With water

80010001200140016001800Wavenumber, cm-1

No water

With water

No water

Dichloromethane extract Enlarge region (1800-800 cm-1)

0.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

3.6

60012001800240030003600

Ab

so

rban

ce

Wavenumber, cm-180010001200140016001800

Wavenumber, cm-1

No water

With water

No water

With water

Enlarge region (1800-800 cm-1)Ethyl acetate extract

0.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

3.6

60012001800240030003600

Ab

so

rban

ce

Wavenumber, cm-1

With water

80010001200140016001800Wavenumber, cm-1

With water

No water

No water

Enlarge region (1800-800 cm-1)Ethanol extract

80010001200140016001800Wavenumber, cm-1

With water

No water

0.0

0.8

1.6

2.4

3.2

4.0

4.8

5.6

6.4

60012001800240030003600

Ab

so

rban

ce

Wavenumber, cm-1

No water

With water

Enlarge region (1800-800 cm-1)Acetic acid extract

0.0

0.4

0.8

1.2

1.6

2.0

2.4

2.8

3.2

3.6

60012001800240030003600

Ab

so

rban

ce

Wavenumber, cm-1

With water

80010001200140016001800Wavenumber, cm-1

No water

With water

No water

Acetone extract Enlarge region (1800-800cm-1)

39

Tukey pairwise comparison analysis (Table 6) confirmed that differences in L*

values were not significant between ground samples for dark or medium roasted

beans, implying that samples from the same degree of roast exhibited the same

lightness.

Table 5 L* Value of Roasted Ground Arabica Coffee Beans

Roast degree Coffee bean sample Lightness [L*]

Dark Colombian 19.83 ± 0.05

Costa Rican 19.61 ± 0.18

Ethiopian 19.46 ± 0.21

Kenyan 19.72 ± 0.06

Medium Colombian 25.21 ± 0.16

Costa Rican 25.35 ± 0.29

Ethiopian 25.64 ± 0.06

Kenyan 25.28 ± 0.09

40

Table 6 Turkey method for L* value comparisons

Comparison 95% SCI

(Dark roast)

95% SCI

(Medium roast)

Different

from 0?

Colombian VS. Costa Rican (-0.176, 0.616) (-0.326, 0.866) No

Colombian VS. Ethiopian (-0.026, 0.766) (-0.036, 0.896) No

Colombian VS. Kenyan (-0.286, 0.506) (-0.396, 0.536) No

Costa Rican VS. Ethiopian (-0.246, 0.546) (-0.176, 0.756) No

Costa Rican VS. Kenyan (-0.286, 0.506) (-0.396, 0.536) No

Ethiopian VS. Kenyan (-0.136, 0.656) (-0.106, 0.826) No

Dark roast: MSE = 0.022908, HSD (t, αF) = 0.396; Medium roast: MSE = 0.03175,

HSD (t, αF) = 0.466

4.3.3 PCA Analysis of Solvent Extracts of Coffee Beans

Typical FTIR spectra of dichloromethane extracts (method #1) of dark roast

coffee beans from various regions are presented in Figure 8. As shown, although

variances between spectra exist, the differences are subtle and data

interpretation is difficult. To extract relevant information from the data, PCA was

employed to reduce the dimensionality of the IR spectra and facilitate the

visualization of the inherent structure of the dataset (Figures 9 and 10).

41

Figure 8 Selected FTIR spectra of dark roast coffee extract obtained with

dichloromethane as a solvent (using method 1#)

7501000125015001750200022502500275030003250350037504000

1/cm

-0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

Abs

HIGH T-COLOM 2-2HIGH T-KENYA 1-1

HIGH T-ETHIOPIAN 1-1HIGH T-COSTA 1-1

FTIR Measurement

Costa Rican

Colombian

Kenyan

Ethiopian

42

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

A

B

C

-0.003

-0.002

-0.001

0.000

0.001

0.002

-0.010 -0.005 0.000 0.005 0.010

2n

d P

rin

cip

al

Co

mp

on

en

t

1st Principal Component

Colombian Costa Rican Ethiopian Kenyan

Hexane Extract

-0.003

-0.002

-0.001

0.000

0.001

0.002

-0.005 -0.003 0.000 0.003 0.005

3n

d P

rin

cip

al

Co

mp

on

en

t



Dichloromethane Extract

-0.003

-0.002

-0.001

0.000

0.001

0.002

-0.005 -0.003 0.000 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t




-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.002 -0.001 0.000 0.001 0.002

3n

d P

rin

cip

al

Co

mp

on

en

t



Acetone Extract

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000

PC 1 Loading (89.4%)

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000

PC 1 Loading (24%)

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2850

1741 1678

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

16681548

1028

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Wavenumber, cm-1

2850

2920

1741 1726

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2920

2850

1236

1550

16431697

1743

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


43

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

Hexane Extract

Dic

hlo

rom

eth

an

e


Ace

ton

e

-0.003

-0.002

-0.001

0.000

0.001

0.002

-0.010 -0.005 0.000 0.005 0.010

2n

d P

rin

cip

al

Co

mp

on

en

t



-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


-0.002

-0.001

0.000

0.001

-0.002 -0.001 0.000 0.001 0.002

3n

d P

rin

cip

al

Co

mp

on

en

t



-0.15

-0.05

0.05

0.15

0.25

8001600240032004000

PC 1 Loading (60%)

-0.002

-0.001

0.000

0.001

0.002

-0.004 -0.002 0.001 0.004

2n

d P

rin

cip

al

Co

mp

on

en

t



-0.002

-0.001

0.000

0.001

0.002

-0.010 -0.005 0.000 0.005 0.010

3n

d P

rin

cip

al

Co

mp

on

en

t



0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2920

2850

1741 1726

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

1550

1683

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2920

2850

16471697

1236

1550

1741

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

1514

1662

1649

1481

A

B

C

Dichloromethane Extract Acetone Extract

44

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

45

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 O-

H 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

46

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 1743-

1741, 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,

47

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 (1600-

1700 cm-1) and fatty acid esters (1700-1800 cm-1) were being extracted from the

dark roast coffee.

4.3.4 PCA Analysis for Coffees According to Degree of Roast

Roasting results in many physical changes and chemical reactions in the

coffee beans. Depending on the extent of the roast, which is time-temperature

dependent, the quality and sensory properties of the resulting coffees can vary

considerably. Medium roast coffee has a more full-bodied flavour, a balance of

taste and aroma, and carries citrus taste. In comparison, dark roast coffee has a

heavier sweet taste, with a lingering aftertaste of chocolate (Schenker et al. 2002;

Lyman et al. 2003).

Dichloromethane and ethyl acetate extracts were tested for the

discrimination of dark and medium roast coffees. As shown in Figures 11 and 12

(Row A), two-component score plots resulted in well-separated clusters

corresponding to dark coffees (right clusters) and medium coffees (left clusters)

from the four origins. The loading plots for dichloromethane extracts showed that

all coffee samples, except the Columbian coffee, exhibited strong loading scores

at 2920, 2850, and 1743 cm-1 due to CH2 asymmetrical stretching of methyl

groups, C-H symmetrical stretching of methyl groups, and C=O stretching of

48

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 (1714-

1705 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

49

cm-1), and lipids/proteins (2935-2847 cm-1) (Lyman et al. 2003; Movasaghi et al.

2008; Wang et al. 2009).

50

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

-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t


Dark roast Medium roast

-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t



-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t



-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t




A

-0.1

0.0

0.1

0.2

8001600240032004000

PC 1 loading (75.9%)

-0.1

0.0

0.1

0.2

8001600240032004000


-0.1

0.0

0.1

0.2

8001600240032004000


-0.1

0.0

0.1

0.2

8001600240032004000


B

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

1481

1510

1550

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2850

17432920

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2850

1743

1465

2920

1678

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2850

1743

1695

2920

860

829C

51

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

C

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2920

2850

1741

2355-2347

16741647

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2928-2916

2850

1741

16741701

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2920

2850

1741

1674

1030

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rban

ce

Wavenumber, cm-1

2920

2850

1741

1701

1674

1647

1550

1236

-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t


Dark roast medium roast

Kenyan

-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t



Colombian

-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t



Costa Rican

-0.002

-0.001

0.000

0.001

0.002

-0.003 -0.001 0.001 0.003

2n

d P

rin

cip

al

Co

mp

on

en

t



Ethiopian

A

-0.1

0.0

0.1

0.2

8001600240032004000


-0.1

0.0

0.1

0.2

8001600240032004000

PC 1loading (59.5%)

-0.1

0.0

0.1

0.2

8001600240032004000


-0.1

0.0

0.1

0.2

8001600240032004000

PC 1 loading (74%)

B

52

4.3.5 SIMCA Analysis

Following the successful application of PCA techniques to discriminate

selected coffee samples according to their geographical origin and degree of

roast, SIMCA classification was employed to predict the origin and degree of

roast for unknown samples.

Table 7 shows the results of the prediction performance for coffee from

different origins based on SIMCA models at the 5% significance level. Except for

the dichloromethane extract for the Ethiopian medium roast sample, all other

samples were correctly assigned to the country of origin during model validation.

Similar validation results were obtained for the prediction degree of roast within

each coffee (Table 8). Overall, ethyl acetate is a more optimal solvent for the

discrimination of coffee origins and roasting degrees. Ethyl acetate is also a

common solvent used for decaffeinating coffee and tea leaves (Dusseldorp et al.

1990; Deventer et al. 1992).

Table 7 SIMCA Classification Results for Coffees from Different Geographic

Origins

Solvents Roast

degree

Country of Origin Correct Classification, %

Colombia Costa

Rica Ethiopia Kenya

Dichloromethane Dark 100 100 100 100

Medium 100 100 33 100

Ethyl acetate Dark 100 100 100 100

Medium 100 100 100 100

53

Table 8 SIMCA classification results for coffees according to degree of roast

Solvents Origin

Roast Degree Correct

classification, %

Dark Medium

Dichloromethane

Colombia 100 100

Costa Rica 100 100

Ethiopia 100 100

Kenya 100 100

Ethyl acetate

Colombia 100 67

Costa Rica 100 100

Ethiopia 100 100

Kenya 100 100

In summary, the aforementioned solvent extraction and chemometric

analysis methodologies may be useful for the coffee industry as a rapid and

reasonably accurate tool for classification of coffee according to origin and

degree of roast. Also, the methods may be useful for routine quality evaluation

and complement sensorial/cupping procedures. Further investigation on more

coffee varieties and study using bigger sample size are necessary to improve the

model robustness. Furthermore, it will be interesting to incorporate sensory

evaluation of coffee in future studies, which is a well-known practice in coffee

industry for discerning coffees.

54

5 EFFECTS OF DIFFERENT TIME-TEMPERATURE PROFILES ON

COFFEE PHYSICAL AND CHEMICAL PROPERTIES

5.1 Introduction

Green coffee beans provide neither the characteristic aroma nor the taste of

a cup of coffee. To reveal their flavour, green coffee beans need to be roasted.

Roasting is one of the most important steps in coffee processing that leads to the

development of the desired aroma, taste, and color of the final brewed product. In

general, the use of roasting temperature of greater than 200oC is required in

order to result in desirable chemical, physical, structural, and sensorial changes

in the coffee beans (Schenker 2000; Schenker et al. 2002; Baggenstoss et al.

2008).

The time and temperature conditions applied during roasting have a major

impact on the physical and chemical properties of roasted coffee beans. Geiger

et al. reported that CO2, a by-product formed due to Strecker reactions and the

degradation of organic compounds, increased greatly towards the end phase of a

high-temperature-short-time process (260oC, 170 s), while the CO2 formed was

much lower when a low-temperature-long-time (228oC, 720s) process was

employed (Geiger et al. 2005). Schenker et al. found that roasting process that

involved a ramping temperature profile (150 to 240oC in 270 s; 240oC for 55 s)

resulted in the formation of a greater quantity of aroma volatiles than a low-

temperature-long-time process (isothermal heating at 220oC for 600 s) (Schenker

et al. 2002). Baggenstoss also reported that high-temperature-short-time roasting

55

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;

56

Fukuyama 1999; Hennessy et al. 2009). Although gas chromatography-mass

spectrometry (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.

57

5.2 Materials and Methods

5.2.1 Chemicals and materials

Ethyl acetate was purchased from Fisher Scientific (Ottawa, Canada),

ethanol from Greenfield Ethanol Inc. (Brampton, Canada), and sodium hydroxide

from Sigma Aldrich (Ontario, Canada). Wet-processed green coffee beans

(Arabica) from Brazil were donated by Mother Parkers Tea & Coffee Inc.

5.2.2 Green Beans and Roasting Conditions

Green coffee beans (45 g) were roasted in a fluidized bed hot air roaster

(Fresh Roast SR 500, Fresh Beans Inc., Park City, UT). Four isothermal roasting

programs were used for roasting (Table 9). Coffee beans were collected at six

roast stages (start-of-first-crack, end-of-first-crack, 48 s-after-first-crack, start-of-

second-crack, end-of-second-crack, and 48 s-after-second-crack) and air-

quenched. The roasted bean samples were stored in hermetic glass bottles in the

dark at 15oC before grinding.

58

Table 9 Time Taken to Achieve Different Stages of Roasting at Four Different

Final Roast Temperatures

Final

Roasting

temperature

(°C)

Time taken (min)

start-of-

first-

crack

end-of-

first-

crack

48s-

after-

first-

crack

start-of-

second-

crack

end-of-

second-

crack

48s-

after-

second-

crack

210 3.8 5.3 6.1 16.9 19.1 20.7

220 2.7 3.6 4.4 8.8 11.2 12

230 2.4 3.3 4.1 4.9 5.9 6.7

240 1.4 2.6 3.4 3.6 3.9 4.7

5.2.3 Degree of Roast as Determined by Color Measurements

Roasted coffee beans were ground using a coffee grinder (Bodum Antigua

Electric Burr Grinder, Bodum, Inc., Copenhagen, Denmark) at the medium grind

setting. The colour of the ground coffee was measured in the L*, a*, b* system

using a Konica Minolta CM-3500d spectrophotometer (Konica Minolta Sensing,

Inc., Osaka, Japan) in the reflectance mode. Before analysis, the instrument was

calibrated on a white standard tile. Measurements were taken in triplicate.

5.2.4 Moisture Content of Ground Coffee

Gravimetrical determination of moisture content of ground coffee was

carried out using oven dehydration method according to the Swiss Food Manual

59

(SWFOH 1973). Samples of roasted beans were ground finely, and then dried at

103°C 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.

60

5.2.8 Chemometric Analysis

Raw FTIR spectra were exported as ASCII format, organized in Excel

spreadsheets, and then analyzed using Pirouette v.4.0 software (Woodinville,

WA). A principal component analysis (PCA) was adopted to reduce the

dimensionality of complex FTIR data, as well as to facilitate the visualization of

the data structure. Second derivative and mean-center were applied to FTIR

spectra to reduce baseline variation and enhance spectral features.

5.2.9 Scanning Electron Microscopy (SEM) Analysis

Coffee bean samples were cut perpendicular to the crevice to expose the

internal bean structure. The bean samples were coated with 20 nm of gold in a

sputter coater (Model K550, Emitech, Ashford, Kent, England). The pore

structure of coffee beans was examined using a scanning electron microscope

(SEM S-570, Hitachi High Technologies Corporation, Tokyo, Japan) at an

accelerating voltage of 10 kV. Micrographs were collected using Image capture

system (Quartz PCI, Version 7, Quartz Imaging Inc., Vancouver, BC).

5.3 Results and Discussion

5.3.1 Evolution of physical and chemical properties during roasting

Roasting causes the buildup of pressure within the coffee bean due to the

formation of steam and other gases. The increased internal pressure causes the

bean to expand and crack, resulting in a phenomenon known as the “first crack”.

As heating continues to higher temperatures (>160°C), the beans become darker

61

and generate a large amount of CO2, with concomitant degradation of

carbohydrates and amino acids. When the pressure buildup exceeds the strength

of the cellulosic cell wall, rapid popping of coffee bean occurs – a roasting event

commonly known as the “second crack”. In practice, these cracking phenomena

are often used to evaluate the progress of roast processing of coffee beans.

Accordingly, these events were chosen as the reference sampling points in the

present study.

In Figure 13A, the evolution of color, moisture content, pH value, and

titratable acidity of coffee samples collected at different stages of roast are

presented. Figure 13B shows the corresponding changes of the same properties

as a function of actual roast time.

10

20

30

40

50

Lig

htn

es

s, L

*

(i)

3

4

5

6

7

8

9

Mo

istu

re c

on

ten

t, %

(ii)

0.00

0.20

0.40

0.60

0.80

1.00

1.20

Tit

rata

ble

ac

idit

y, g

/L

Stage of roasting

(B)

(iv)

5.0

5.2

5.4

5.6

5.8

6.0

pH

va

lue

Stage of roasting

(iii)

A

10

20

30

40

50

Lig

htn

es

s, L

* (i)

3

4

5

6

7

8

9

Mo

istu

re c

on

ten

t, %

(ii)

0.00

0.20

0.40

0.60

0.80

1.00

1.20

To

tal a

cid

ity, g

/L

Stage of roasting

210 C

220 C

230 C

240 C

(B)

(iv)

5.0

5.2

5.4

5.6

5.8

6.0

pH

va

lue

Stage of roasting

(iii)

62

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

Lightness. Color change is one of the most important modifications in

coffee bean during roasting (Illy & Viani 1995; Eggers & Pietsch 2001) caused by

non-enzymatic browning reactions such as Maillard reaction and caramelization

(Massini et al. 1990; Parliment Thomas 2000). Lightness (L*) is often used as a

measurement of the degree of roast (e.g., light, medium, and dark), which is

directly related to the roasting time and temperature (Sivetz 1991; Illy & Viani

1995; Griffin 2001-2006). The L* values at different stages of roast are

summarized in Figures 13(i) and 13B(i). As shown, at the end-of-first-crack, all

bean samples achieved a similar degree of roast (i.e., medium roast), with L*

10

20

30

40

50

-2 1 4 7 10 13 16 19

Lig

htn

es

s, L

*

Time, min

(i)

3

4

5

6

7

8

9

-2 1 4 7 10 13 16 19

Mo

istu

re c

on

ten

t, %

Time, min

(ii)

4.8

5.0

5.2

5.4

5.6

5.8

6.0

-2 1 4 7 10 13 16 19

pH

va

lue

Time, min

(iii)

0.00

0.20

0.40

0.60

0.80

1.00

1.20

-2 1 4 7 10 13 16 19

Tit

rata

ble

ac

idit

y, g

/L

Time, min

(iv)

B

10

20

30

40

50

Lig

htn

es

s, L

* (i)

3

4

5

6

7

8

9

Mo

istu

re c

on

ten

t, %

(ii)

0.00

0.20

0.40

0.60

0.80

1.00

1.20

To

tal a

cid

ity, g

/L

Stage of roasting

210 C

220 C

230 C

240 C

(B)

(iv)

5.0

5.2

5.4

5.6

5.8

6.0

pH

va

lue

Stage of roasting

(iii)

63

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).

Moisture content. Figures 13A(ii) and 13B(ii) show the changes of

moisture content during roasting. During the early phase of roasting (up to the

end-of-first-crack), considerable decrease in moisture content was observed in

the coffee beans. The rapid loss in moisture during the initial phase of roasting

can be attributed to the vaporization of water as the temperature of the beans

increased above the boiling point of water. From the end-of-first-crack to the

start-of-second-crack, there was an increase in moisture content due to the

generation of water as a result of pyrolytic cleavage of carbohydrates, and

64

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.

pH value and titratable acidity. The changes in pH value and titratable

acidity at different stages of roast are presented in Figures 13A(iii), 13A(iv),

13B(iii), and 13B(iv). As shown, at all roasting temperatures investigated, a

significant decrease in pH value (P<0.05) was observed as the coffee beans

were roasted up to the start-of-first-crack. The decrease in pH was caused by the

formation of formic, acetic, glycolic, and lactic acids as the coffee beans were

processed to medium roast (Ginz et al. 2000). Beyond the start-of-first-crack,

minimal changes in pH value were detected up to 48 s-after-first-crack, thereafter,

a dramatic increase in pH was observed. The rapid increase in pH value may be

attributed to the destruction of organic acids formed and those that were present

initially (citric acid, malic acid, and chlorogenic acids), as the coffee beans were

taken to darker roasts (Ginz et al. 2000). Before the start-of-second-crack, the pH

65

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

66

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

240°C, 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.

5.3.2 Changes in coffee at various stages of roast

Roasting is a complex process that results in substantial physical and

compositional changes in coffee beans. These changes are both time and

temperature 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

67

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., 210oC). 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 CH2

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 240oC, 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).

68

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 230°C, more absorbance with large loading score (>0.1 and <

-0.1) were observed than those roasted at 210 and 240°C, especially at 1741 cm-

1 (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 Lüllmann 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 & Lüllmann 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 & Lüllmann 1993; Baggenstoss et al. 2008).

69

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

-0.005

-0.003

-0.001

0.001

0.003

0.005

-0.006 -0.002 0.002 0.006 0.01 0.014

2n

d P

rin

cip

al

Co

mp

on

en

t


Green coffee beansstart-of-first-crackend-of-first-crack48s-after-first-crackstart-of-second-crackend-of-second-crack48s-after-second-crack

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nce

Wavenumber, cm-1

2850

29201697

1660

1707

1741

1718

1514

1556

1545

1739

230 C

-0.005

-0.003

-0.001

0.001

0.003

0.005

-0.006 -0.002 0.002 0.006 0.01 0.014

2n

d P

rin

cip

al

Co

mp

on

en

t


Green coffee beans

start-of-first-crack

end-of-first-crack

48s-after-first-crack

start-of-second-crack

end-of-second-crack

48s-after-second-crack

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nce

Wavenumber, cm-1

2920

2850

1514

1660

2839

1739

240 C

-0.005

-0.003

-0.001

0.001

0.003

0.005

-0.006 -0.002 0.002 0.006 0.010 0.014

2n

d P

rin

cip

al

Co

mp

on

en

t



-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nce

Wavenumber, cm-1

15142850

29201697

1660

1707

1741

1718

1739

220 C

210 C

-0.005

-0.003

-0.001

0.001

0.003

0.005

-0.006 -0.002 0.002 0.006 0.010 0.0142

nd

Pri

nc

ipa

l C

om

po

ne

nt



A

-0.15

-0.05

0.05

0.15

0.25

8001600240032004000


B

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nce

Wavenumber, cm-1

2850

1697

1741

1514

1660

29201739

C

70

The IR spectra at 230oC were overlaid to examine the differences at

different roast stages (Figure 15). Compounds that accounted for the observed

variances include lipids (2920-2850 cm-1), unsaturated ester/lactone (1780-1762

cm-1), aliphatic esters (1755-1740 cm-1), aldehydes (1739-1724 cm-1), ketones

(1725-1705 cm-1), aliphatic acids (1714-1705 cm-1), aromatic acids (1700-1680

cm-1), and caffeine (1650-1600 cm-1) (Bellamy 1975; Keller 1986; Socrates 1994;

Briandet et al. 1996b; Hennessy et al. 2009). As shown in Figure 15, the spectral

differences were mainly related to the changes in concentration of these

compounds. For instance, the content of unsaturated ester/lactone (1780-1762

cm-1), and caffeine (1650-1600 cm-1) decreased from the start-of-first-crack to the

start-of-second-crack, and then stabilized. By contrast, aliphatic esters (1755-

1740 cm-1) and aldehydes (1739-1724 cm-1) gradually increased from the start-

of-first-crack to the start-of-second-crack, and then decreased due to thermal

degradation. Ketones (1725-1705 cm-1), aliphatic acids (1714-1705 cm-1), and

aromatic acids (1700-1680 cm-1) firstly decreased from the start-of-first-crack to

the start-of-second-crack, and then increased to the end-of-second crack.

Thereafter no detectable change was observed on further roasting. Finally, the

content of protein was stable from the start-of-first-crack to 48 s-after-first-cracks,

but significant decrease was observed from 48s-after-first-crack to the start-of-

second-crack.

71

Figure15 The expanded 2910-2850 cm-1, and 1800-1500 cm-1 regions of the spectra of coffee roasted at 230oC

0.00

0.04

0.08

0.12

0.16

0.20

2800285029002950

Ab

so

rba

nc

e 2920-2850

1500152515501575160016251650167517001725175017751800

Wavenumber, cm-1

start-of-first-crack

end-of-first-crack

48s-after-first-crack

start-of-second-crack

end-of-second-crack

48s-after-second-crack

1700-16801714-1705

1725-1705

1739-1724

1755-1740

1780-1762

1650-1600

1550

72

5.3.3 Effects of roast temperature on changes in coffee

Overall, clusters for the score plots were well separated according to

different roasting temperatures (Figure 16, column A). The separation distance

between 220 and 230oC tended to be closer during the first three stages of roast,

indicating that the IR-active components are similar in the coffee beans roasted

at these two temperatures. From the loading plots (Figure 16, column B), it is

evident that coffees collected at the start-of-first-crack were markedly different

from those of the other five stages, in that the former exhibited less number of

bands with large loading score, which could be due to the fewer compounds

extracted during the short process time (Lyman et al. 2003). Loading plots for

coffees collected at the end-of-first crack and 48s-after-first-crack were

comparable, suggesting that the compounds present for samples roasted at

different temperatures were similar. The main regions that contributed to the

separation in score plots of end-of-first-crack and 48s-after-first-crack were 1724-

1726, 1699-1701, and 1676 cm-1, due to aldehydes/ketones, aromatic acids, and

lipids, respectively (Bellamy 1975; Keller 1986; Socrates 1994). Other spectral

regions that contributed to cluster separation were 1660 cm-1 due to C=C

stretching band of fatty acids (Shetty 2006), 1650 cm-1 due to Amide I of protein

(Dovbeshko 1997), 1550 cm-1 from Amide II of proteins (Fukuyama 1999), and

1514 cm-1 caused by N=C stretching of amino groups (Barua et al. 2008).

73

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

-0.25

-0.15

-0.05

0.05

0.15

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nc

e

Wavenumber, cm-1

1550

1541

1650

1660

3649

1691

1556

1514

-0.003

-0.001

0.001

0.003

0.005

-0.005 -0.003 -0.001 0.001 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t


210 C-2 220 C-2 230 C-2 240 C-2

en

d-o

f-fi

rst-

cra

ck

-0.25

-0.15

-0.05

0.05

0.15

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nc

e

Wavenumber, cm-1

1514

1699-17011660

1724-1726

1541

3649

1676

1650

15561566

1689-1691

-0.003

-0.001

0.001

0.003

0.005

-0.005 -0.003 -0.001 0.001 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t


210 C-3 220 C-3 230 C-3 240 C-3

48

s-a

fte

r-fi

rst-

cra

ck

-0.25

-0.15

-0.05

0.05

0.15

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nc

e

Wavenumber, cm-1

2850

2920

1670

1741

1724

1514

-0.003

-0.001

0.001

0.003

0.005

-0.005 -0.003 -0.001 0.001 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t


210 C-4 220 C-4 230 C-4 240 C-4

sta

rt-o

f-s

ec

on

d-c

rack

-0.25

-0.15

-0.05

0.05

0.15

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nc

e

Wavenumber, cm-1

2850

2920

1741

1724

1514

1689

1670

-0.25

-0.15

-0.05

0.05

0.15

8001600240032004000


0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nc

e

Wavenumber, cm-1

1741

17241514

1689

16701643

1550

1540

2920

2850

-0.003

-0.001

0.001

0.003

0.005

-0.005 -0.003 -0.001 0.001 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t


210 C-5 220 C-5 230 C-5 240 C-5

en

d-o

f-s

eco

nd

-cra

ck

-0.003

-0.001

0.001

0.003

0.005

-0.005 -0.003 -0.001 0.001 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t


210 C-6 220 C-6 230 C-6 240 C-6

48

s-a

fte

r-s

eco

nd

-cra

ck

-0.003

-0.001

0.001

0.003

0.005

-0.005 -0.003 -0.001 0.001 0.003 0.005

2n

d P

rin

cip

al

Co

mp

on

en

t


210 C-1 220 C-1 230 C-1 240 C-1

sta

rt-o

f-fi

rst-

cra

ck

A

-0.25

-0.15

-0.05

0.05

0.15

8001600240032004000


B

0.00

0.04

0.08

0.12

0.16

0.20

8001600240032004000

Ab

so

rba

nc

e

Wavenumber, cm-1

2850

2920

1665

1741

C

74

At the start-of-second-crack and end-of-second-crack, the separation

distances between clusters increased considerably, indicating that there is an

increased differentiation in the IR-active components for coffee samples

processed at different temperatures. However, roasting beyond the second crack

(48s-after-second-crack), the cluster separation decreased, especially between

samples processed at higher temperatures (230 and 240oC). These results imply

that the second crack is an important milestone during coffee roasting, at which

the IR-active components will differ depending on the roast temperature

employed.

5.3.4 Microstructural analysis

Scanning electron microscope (SEM) revealed considerable changes in

microstructure of coffee beans as they were processed to different stages during

roasting (Figure 17). As shown by the micrographs, the unroasted beans

displayed a relatively compact morphology as compared to the heated samples.

When the beans were heated to the start-of-first-crack, a few pores started to

form. As the roasting progressed to the end-of-first-crack, porous networks were

established in the beans. Continued roasting to 48s-after-first-crack did not

induce appreciable microstructural changes. However, when the samples were

roasted to the second crack, further expansion of the pores is evident from the

micrographs, as exemplified by the reduced thicknesses of the walls surrounding

the pores. Furthermore, the pores were more polyhedral in shape than those

appeared in earlier roast stages, suggesting that there was an increased internal

75

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-after-

second-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 low-

temperature-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.

76

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

77

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 FTIR-

chemometric 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.

78

6 Conclusions and Future Works

This research studied the physicochemical changes that took place during

roast processing of coffee beans using a modified commercial fluidized bed

roaster.

An analytical method was developed for discriminating roasted coffee

samples from different geographical regions and roast degrees, using solvent

extraction and FTIR-chemometric techniques. Experimental results showed that

a mixture of equal volume of the solvent and water resulted in an optimal extract

that provided important spectral information for discriminating different samples.

PCA analysis of the FTIR spectra of extracts prepared from solvents of different

polarity (dichloromethane, ethyl acetate, hexane and acetone) allowed us to

study changes in chemical compositions between medium and dark roast

coffees. From the FTIR spectra, we observed that roasting coffee from medium

to dark 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 (1714-1705 cm-1), but a decrease in caffeine content (1700-

1692 cm-1, and 1647-1641 cm-1), lipids (around 1736, 1740, 1745, and 1750 cm-

1), and polysaccharides and hemicelluloses (1739 cm-1). The SIMCA models

developed, based on dichloromethane and ethyl acetate extracts, allowed

accurate discrimination of coffees from different countries of origin, as well as

their roast degrees. The extraction and chemometric analysis methodologies

developed here may be useful for the coffee industry as a rapid and accurate tool

for classification of coffee according to origin and degree of roast. Also, the

79

methods may be useful for routine quality evaluation and complement

sensorial/cupping procedures.

To study the physicochemical changes in coffee bean during various stages

of roast, Brazilian beans were roasted at different temperatures (210, 220, 230,

and 240oC) until dark roast products were achieved. Coffee beans samples,

collected at different milestones during the thermal process, were analyzed for

pH value, titratable acidity, moisture content, and color lightness. When roasted

beyond the start-of-second-crack (i.e., typical milestone that reflects the

doneness of dark roast coffee), we observed that coffee beans processed at

high-temperature-short-time resulted in higher moisture content, higher pH value,

higher titratable acidity, higher porous structure in the bean cell tissues, and also

produced more aldehydes, ketones, aliphatic acids, aromatic acids, and caffeine

than those processed at low-temperature-long-time process. FTIR-chemometric

analysis of ethyl acetate extract of roasted coffees showed that for all the

temperatures tested, the score plots displayed clusters that were separated

according to different stages of roast as well as according to different roasting

temperatures. From this study, we can conclude that coffee beans processed to

the same extent of roast (as determined from bean lightness and roast cracking

milestones) but using different time-temperature roasting profiles do not

necessarily have the same physical and chemical properties. FTIR-ATR analysis

of ethyl acetate extract of coffee contains useful absorbance information for

studying the effect of roasting conditions on IR-active compounds. The FTIR-

80

chemometric technique potentially can be used for detecting process deviation in

production roasters.

Using the modified commercial fluidized bed roaster, in conjunction with

chemometric, FTIR and other analytical methodologies, the project has achieved

the objective of elucidating the effect of time-temperature on physicochemical

changes that took place in various coffee beans. While there is a significant gain

in knowledge from this project, there are still a number of unanswered questions

that are yet to be answered. The followings are several topics that are worth

further investigation:

After roasting, coffee beans are quenched to remove the residual heat

quickly. This process can trap significant amount of CO2 in the bean,

thereby lengthens the required time for CO2 degassing. This is a critical

step that must be carried out before packaging of roasted coffee to

prevent packaging failure due to pressure build up within the package.

Conceivably, depending on the method of cooling used, post-roasting

carbon dioxide degassing time may be shortened, or even eliminated. For

instance, spraying roasted coffee with a controlled amount of water under

agitation will remove the residual heat from the coffee beans rapidly due to

the latent heat of vaporization of water. The humidified air may increase

the rate of CO2 degas. Potentially, this process may be incorporated as

part of the roasting regime towards the end of the roast cycle before

ejecting the beans from the roaster. Alternatively, slower cooling at

temperatures above ambient in an enclosed space will increase the

81

diffusivity of CO2, potentially shortening the duration of the degassing step.

Further investigation involving these types of innovative process

inventions to shorten or eliminate CO2 degas will simplify the degas

storage and packaging requirements of roasted coffee.

In this project, we have demonstrated that FTIR-chemometric technique is

useful for the evaluation of coffee properties. Further investigations

involving more coffee varieties and larger sample size will improve the

robustness of the models.

Due to the time constraint, sensory studies were not conducted in this

project. By coupling the IR fingerprints with sensory feedback from expert

cuppers, it is possible to derive PLS and SIMCA models that complement

the sensory evaluation, creating quantifiable spectral data for quality

assurance, validation, product development, training and other purposes.

A good quality cup of coffee is depending on many factors, such as the

degassing methods of coffee after roasting, grind particle size, brewing

time/temperature profiles, quality of brew water, and storage/time

conditions (temperature, O2, and relative humidity). Integrated studies

involving these factors, in conjunction with sensory evaluation, will improve

the understanding on the quality attributes of brewed coffee.

82

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 Time-temperature 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 gru¨nem 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.

83

BRIANDET, R. KEMSLEY, E. K. & WILSON, R. H. (1996a). Approaches to Adulteration Detection in Instant Coffees using Infrared Spectroscopy and Chemometrics. Journal of the Science of Food and Agriculture 71(3), 359-366.

BRIANDET, R. KEMSLEY, E. K. & WILSON, R. H. (1996b). Discrimination of Arabica and Robusta in Instant Coffee by Fourier Transform Infrared Spectroscopy and Chemometrics. Journal of Agricultural and Food Chemistry 44(1), 170-174.

BROLLO, G. CAPPUCCI, R. & NAVARINI, L. (2008). Acidity in Coffee: Bridging the Gap Between Chemistry and Psychophysics. In 22nd Colloquium: Coffee Aroma & Flavour Chemistry Campinas, São Paulo, Brazil: Association for Science and Information on Coffee.

BUCHELI, P. MEYER, I. PASQUIER, M. & LOCHER, R. (1996). Determination of soluble sugars by high performance anion exchange chromatography (HPAE) and pulsed electrochemical detection (PED) in coffee beans upon accelerated storage. In The 10th FESPP meeting pp. L-12. Italy: Plant Physiology Biochemistry.

BUCHELI, P. MEYER, I. PITTET, A. VUATAZ, G. & VIANI, R. (1998). Industrial Storage of Green Robusta Coffee under Tropical Conditions and Its Impact on Raw Material Quality and Ochratoxin A Content. Journal of Agricultural and Food Chemistry 46(11), 4507-4511.

BUFFO, R. A. & CARDELLI-FREIRE, C. (2004). Coffee flavour: an overview. Flavour and Fragrance Journal 19(2), 99-104.

BYERS, J. A. (2003). Polarity index of solvents. Phenomenex catalog, www.Phenomenex.com.

CAGUE, R. MILLARD, M. & GIBSON, D. (2009). Beyond the Bean: Redefining Coffee Quality. pp. 1-18. Environment and Natural Resource Management and Agribusiness Practice Networks.

CANADA, C. A. O. (2003). 2003 CANADIAN COFFEE DRINKING STUDY Tonronto.

CASAL, S. ALVES, M. R. MENDES, E. OLIVEIRA, M. B. P. P. & FERREIRA, M. A. (2003). Discrimination between Arabica and Robusta Coffee Species on the Basis of Their Amino Acid Enantiomers. Journal of Agricultural and Food Chemistry 51(22), 6495-6501.

CHARLTON, A. J. FARRINGTON, W. H. H. & BRERETON, P. (2002). Application of 1H NMR and Multivariate Statistics for Screening Complex Mixtures: Quality Control and Authenticity of Instant Coffee. Journal of Agricultural and Food Chemistry 50(11), 3098-3103.

84

CLARKE, R. J. (1986). The Flavour of Coffee. Developments in Food Science 3B, 1-47.

CLARKE, R. J. & MACRAE, R. (1987). Coffee. London: Elsevier Applied Science

CLARKE, R. J. & MACRAE, R. (1988). Coffee: Physiology. London: Elsevier Applied Science.

COSTA FREITAS, A. M. & MOSCA, A. I. (1999). Coffee geographic origin -- an aid to coffee differentiation. Food Research International 32(8), 565-573.

CZERNY, M. MAYER, F. & GROSCH, W. (1999). Sensory Study on the Character Impact Odorants of Roasted Arabica Coffee. Journal of Agricultural and Food Chemistry 47(2), 695-699.

DAVIS, A. P. (2001). Two new species of Coffea L. (Rubiaceae) from northern Madagascar. ADANSONIA 23(2), 337-345.

DE MARIA, C. A. B. TRUGO, L. C. MOREIRA, R. F. A. & WERNECK, C. C. (1994). Composition of green coffee fractions and their contribution to the volatile profile formed during roasting. Food Chemistry 50(2), 141-145.

DEL CASTILLO, M. D. AMES, J. M. & GORDON, M. H. (2002). Effect of Roasting on the Antioxidant Activity of Coffee Brews. Journal of Agricultural and Food Chemistry 50(13), 3698-3703.

DEVENTER, G. KAMEMOTO, E. KUZNICKI, J. HECKERT, D. & SCHULTE, M. (1992). Lower esophageal sphincter pressure, acid secretion, and blood gastrin after coffee consumption. Digestive Diseases and Sciences 37(4), 558-569.

DOVBESHKO, G. I., GRIDINA, N.Y., KRUGLOVA, E.B., AND PASHCHUK, O.P. (1997). FTIR spectroscopy studies of nucleic acid damage. Talanta (53), 233-246.

DUSSELDORP, M. V. KATAN, M. B. & DEMACKER, P. N. M. (1990). Effect of Decaffeinated versus Regular Coffee on Serum. American journal of Epidemiology 132(1), 33-40.

EGGERS, R. & PIETSCH, A. (2001). Coffee: Recent Developments. Oxford: Blackwell Science Ltd.

ELIANE NABEDRYK, M. G., AND JACQUES BRETON (1982). Orientation of Gramicidin a Transmembrane Channel Infrared Dichroism Study of Gramicidin in Vesicles. BIOPHYSICAL JOURNAL 38, 243-249.

FALK, M. GIL, M. & IZA, N. (1990). Self-association of caffeine in aqueous solution: an FT-IR study. Canadian Journal of Chemistry 68, 1293-1299.

FLAMENT, I. (2002). Coffee flavor chemistry. Geneva: John Wiley & Sons, LTD.

85

FOBE, L. A. NERY, J. P. & TANGO, J. S. (1968). Influence of the roasting degree on the chemical composition of coffee. In Third International Colloquium on the Chemistry of Coffee pp. 389-397. Trieste: Association for science and information on coffee.

FRANCA, A. S. & OLIVEIRA, L. S. (2009). Alternative uses for coffee husks - a solid waste from green coffee production. In Chemical, Biological and Environmental Engineering (Ed L. Kai), pp. 21-24. Singapore: Word Scientific Publishing Co. Pte. Ltd.

FU, Y. (2011). INTERACTIONS OF WINE WITH PACKAGING AND PREDICTION OF QUALITY PARAMETERS. University of Guelph.

FUKUYAMA, Y., YOSHIDA, S., YANAGISAWA, S., AND SHIMIZU, M. (1999). A study on the differences between oral squamous cell carcinomas and normal oral mucosas measured by Fourier transform infrared spectroscopy. Biospectroscopy (5), 117-126.

GARRIGUES, J. M. BOUHSAIN, Z. GARRIGUES, S. & DE LA GUARDIA, M. (2000). Fourier transform infrared determination of caffeine in roasted coffee samples. Fresenius' Journal of Analytical Chemistry 366(3), 319-322.

GEIGER, R. PERREN, R. KUENZLI, R. & ESCHER, F. (2005). Carbon Dioxide Evolution and Moisture Evaporation During Roasting of Coffee Beans. Journal of Food Science 70(2), E124-E130.

GELADI, P. & KOWALSKI, B. R. (1986). Partial least-squares regression: a tutorial Analytica Chimica Acta (185), 1-17.

GERLACH, R. W. KOWALSKI, B. R. & WOLD, H. (1979). Partial Least Squares Path Modelling with Latent Variables. Analytica Chimica Acta 112, 417-421.

GIAMPIERO SACCHETTI, C. D. M., PAOLA PITTIA, DINO MASTROCOLA (2009). Effect of roasting degree, equivalent thermal effect and coffee type on the radical scavenging activity of coffee brews and their phenolic fraction. Journal of Food Engineering 90, 74-80.

GINZ, M. BALZER, H. H. BRADBURY, A. G. W. & MAIER, H. G. (2000). Formation of aliphatic acids by carbohydrate degradation during roasting of coffee. European Food Research and Technology 211(6), 404-410.

GOPALAKRISHNA RAO, N. BALACHANDRAN, A. NATARAJAN, C. P. & SANKARAN, A. N. (1971). Variations in moisture and colour in monsooned coffee. Journal of Agricultural and Food Chemistry 8, 174-176.

GRIFFIN, M. (2001-2006). Coffee Roast Colors and Characteristics. Coffee Research Institute.

86

GRIFFIN, M. J. & BLAUCH, D. N. (1999). Determination of the relationship between phosphate concentration and perceived acidity in coffee. In 18th International Scientific Colloquium on Coffee pp. 118-126. Helsinki, Finland: Association for Science and Information on Coffee.

HAGMAN, A. & JACOBSSON, S. (1990). Applications of chemometrics for characterization of macromolecules. Drug development and industrial pharmacy 16(17), 2527-2545.

HASHIM, L. & CHAVERON, H. (1995). Use of methylpyrazine ratios to monitor the coffee roasting. Food Research International 28(6), 619-623.

HENDRIKS, M. M. W. B. CRUZ-JUAREZ, L. BONT, D. D. & HALL, R. D. (2005). Preprocessing and exploratory analysis of chromatographic profiles of plant extracts. Anal. Chim. Acta. 545, 53-64.

HENNESSY, S. DOWNEY, G. & ODONNELL, C. P. (2009). Confirmation of Food Origin Claims by Fourier Transform Infrared Spectroscopy and Chemometrics: Extra Virgin Olive Oil from Liguria. Journal of Agricultural and Food Chemistry 57(5), 1735-1741.

HERNANDEZ, J. A. HEYD, B. IRLES, C. VALDOVINOS, B. & TRYSTRAM, G. (2007). Analysis of the heat and mass transfer during coffee batch roasting. Journal of Food Engineering 78(4), 1141-1148.

HOFMANN, T. & SCHIEBERLE, P. (2001). Chemical Interactions between Odor-Active Thiols and Melanoidins Involved in the Aroma Staling of Coffee Beverages. Journal of Agricultural and Food Chemistry 50(2), 319-326.

HORWITZ, W. (2000). Coffee and tea. In Official methods of Analysis of Association of Official Analytical Chemists International. Gaithersburg, MD, USA: AOAC Press.

ILLY, A. & VIANI, R. (1995). Espresso Coffee: The Chemistry of Quality Londres, UK: Academic Press.

ILLY, A. & VIANI, R. (1998). Espresso coffee. San Diego: Academic press.

ILLY, A. & VIANI, R. (2005). Espresso coffee: the science of quality. San Diego: Elsevier Academic Press.

INFOMETRIX (2010). Chemometrics.

INNAWONG, B. MALLIKARJUNAN, P. IRUDAYARAJ, J. & MARCY, J. E. (2004). The determination of frying oil quality using Fourier transform infrared attenuated total reflectance. Lebensmittel-Wissenschaft und-Technologie 37(1), 23-28.

87

K.SPEER R.TEWIS & A.MONTAG (1991). 16-O-Methylcafestol a Quality Indicator for Coffee. In 14th ASIC colloquium Francisco.

KANTOR, D. B. & FEKETE, A. (2006). Characterization and Quantification of Coffee by Using an Electronic Tongue. In The 2nd CIGR Section VI International Symposium on Future of Food Engineering pp. 26-28. Warsaw, Poland.

KELLER, R. J. (1986). The Sigma Library of FT-IR Spectra. St. Louis, MO: Sigma Chemical Co.

KEMSLEY, E. K. RUAULT, S. & WILSON, R. H. (1995). Discrimination between Coffea arabica and Coffea canephora variant robusta beans using infrared spectroscopy. Food Chemistry 54(3), 321-326.

KRIVAN, V. BARTH, P. & MORALES, A. F. (1993). Multielement analysis of green coffee and its possible use for the determination of origin. Microchimica Acta 110(4), 217-236.

KULABA, G. W. (1981). Coffee processing research: a review. Kenya Coffee 46, 351-360.

LYMAN, D. J. BENCK, R. DELL, S. MERLE, S. & MURRAY-WIJELATH, J. (2003). FTIR-ATR Analysis of Brewed Coffee: Effect of Roasting Conditions. Journal of Agricultural and Food Chemistry 51(11), 3268-3272.

MAGIDMAN, J. M. P. A. P. (1984). Analysis of the Aroma of the Intact Fruit of Coffea arabica by GC-FT-IR. Applied Spectroscopy 38(2), 181-184.

MARIA, M. J. PABLOS, F. & GONZALEZ, A. G. (1998). Discrimination between arabica and robusta green coffee varieties according to their chemical composition. Talanta 46(6), 1259-1264.

MARTIN, M. J. PABLOS, F. & GONZALEZ, A. G. (1999). Characterization of arabica and robusta roasted coffee varieties and mixture resolution according to their metal content. Food Chemistry 66(3), 365-370.

MASSINI, R. NICOLI, M. C. CASSARÀ, A. & LERICI, C. R. (1990). Physico-chemical changes of coffee beans during roasting. Italian Journal of Food Science 2, 123-130.

MAYER, F. CZERNY, M. & GROSCH, W. (2000). Sensory study of the character impact aroma compounds of a coffee beverage. European Food Research and Technology 211(4), 272-276.

MAZZAFERA, P. (1999). Chemical composition of defective coffee beans. Food Chemistry 64(4), 547-554.

88

MISHRA, G. S. & KUMAR, A. (2002). Preparation of heterogeneous vanadrum (VO2+) catalyst for selective hydroxylation of cyclohexane by molecular oxygen. Catalysis Letters 81(1-2), 113-117.

MOTTALEB, M. A. COOKSEY, B. G. & LITTLEJOHN, D. (1997). Evaluation of gel permeation chromatography to Fourier transform infrared spectrometric detection using a modified thermospray for analysis of polystyrene samples. Fresenius' Journal of Analytical Chemistry 358(4), 536-538.

MOVASAGHI, Z. REHMAN, S. & REHMAN, I. U. (2008). Fourier Transform Infrared (FTIR) Spectroscopy of Biological Tissues. Applied Spectroscopy Reviews 43(2), 134 - 179.

MURPHY, P. A. SONG, T. BUSEMAN, G. BARUA, K. BEECHER, G. R. TRAINER, D. & HOLDEN, J. (1999). Isoflavones in Retail and Institutional Soy Foods. Journal of Agricultural and Food Chemistry 47(7), 2697-2704.

NAGARAJU, V. D. MURTHY, C. T. RAMALAKSHMI, K. & SRINIVASA RAO, P. N. (1997). Studies on roasting of coffee beans in a spouted bed. Journal of Food Engineering 31(2), 263-270.

NEHRING, U. P. & MAIER, H. G. (1992). Indirect determination of the degree of roast in coffee. Zeitschrift für Lebensmitteluntersuchung und -Forschung A 195(1), 39-42.

NUNES, F. M. COIMBRA, M. A. DUARTE, A. C. & DELGADILLO, I. (1997). Foamability, Foam Stability, and Chemical Composition of Espresso Coffee As Affected by the Degree of Roast. Journal of Agricultural and Food Chemistry 45(8), 3238-3243.

OHNSMANN, J. QUINTÁS, G. GARRIGUES, S. & DE LA GUARDIA, M. (2002). Determination of caffeine in tea samples by Fourier transform infrared spectrometry. Analytical and Bioanalytical Chemistry 374(3), 561-565.

PAGNI, R. (2005). Solvents and Solvent Effects in Organic Chemistry, Third Edition (Christian Reichardt). Journal of Chemical Education 82(3), 382-null.

PARADKAR, M. M. & IRUDAYARAJ, J. (2002). A Rapid FTIR Spectroscopic Method for Estimation of Caffeine in Soft Drinks and Total Methylxanthines in Tea and Coffee. Journal of Food Science 67(7), 2507-2511.

PARLIMENT, T. H. & STAHL, H. D. (1995). What makes that coffee smell so good? Chemtech 8, 38-47.

PARLIMENT THOMAS, H. (2000). An Overview of Coffee Roasting. In Caffeinated Beverages pp. 188-201. American Chemical Society.

PAUL, G. (2006). Practical Guide To Chemometrics. Boca Raton: FL: CRC Press.

89

PITTIA, P. ROSA, M. D. & LERICI, C. R. (2001). Textural Changes of Coffee Beans as Affected by Roasting Conditions. LWT - Food Science and Technology 34(3), 168-175.

PONTE, S. (2002). The 'Latte Revolution'? Regulation, Markets and Consumption in the Global Coffee Chain. World Development 30(7), 1099-1122.

PUHLMANN, R. & HABEL, G. (1989). Examinations to test the grinding behavior of roasted coffee. Lebensmittelindustrie 36(4), 161-163.

RAHN, W. & KONIG, W. A. (1978). GC/MS Investigations of the Constituents in a Diethyl Ether Extract of an Acidified Roast Coffee Infusion. Journal of High Resolution Chromatography & Chromatography Communications, 69-71.

RAMALAKSHMI, K. & RAGHAVAN, B. (1999). Caffeine in Coffee: Its Removal. Why and How? Critical Reviews in Food Science and Nutrition 39(5), 441-456.

ROGERS, P. J. & RICHARDSON, N. J. (1993). Why do we like drinks that contain caffeine? Trends in Food Science & Technology 4(4), 108-111.

RUBAYIZA, A. B. & MEURENS, M. (2005). Chemical Discrimination of Arabica and Robusta Coffees by Fourier Transform Raman Spectroscopy. Journal of Agricultural and Food Chemistry 53(12), 4654-4659.

RUSSWURM, H. J. (1970). Fractionation and analysis of aroma precursors in green coffee. In Fourth International Colloquium on the Chemistry of Coffee pp. 103-107. Amsterdam: Association for Science and Information on Coffee.

SANZ, C. MAEZTU, L. ZAPELENA, M. J. BELLO, J. & CID, C. (2002). Profiles of volatile compounds and sensory analysis of three blends of coffee: influence of different proportions of Arabica and Robusta and influence of roasting coffee with sugar. Journal of the Science of Food and Agriculture 82(8), 840-847.

SCHENKER, S. HEINEMANN, C. HUBER, M. POMPIZZI, R. PERREN, R. & ESCHER, R. (2002). Impact of Roasting Conditions on the Formation of Aroma Compounds in Coffee Beans. Journal of Food Science 67(1), 60-66.

SCHENKER, S. R. (2000). Investigations on the Hot Air Roasting of Coffee Beans. Swiss Federal Institute of Technology.

SEMMELROCH, P. LASKAWY, G. BLANK, I. & GROSCH, W. (1995). Determination of potent odourants in roasted coffee by stable isotope dilution assays. Flavour and Fragrance Journal 10(1), 1-7.

SHETTY, G., KEDALL, C., SHEPHERD, N., STONE, N., AND BARR, H. (2006). Raman spectroscopy: evaluation of biochemical changes in carcinogenesis of oesophagus. British Journal of Cancer (94), 1460-1464.

90

SHIBAMOTO, T. (1991). An Overview of Coffee Aroma and Flavour Chemistry. In In 14th International Scientific Colloquium on Coffee pp. 107-116. San Francisco, USA: Association Scientifique Internationale pour le Café.

SILVIA, R. LAURA, M. ANTONIO, B. CONCEPCION, C. & MANUEL, A. C. (2004). Screening and distinction of coffee brews based on headspace solid phase microextraction/gas chromatography/principal component analysis. Journal of the Science of Food and Agriculture 84(1), 43-51.

SILWAR, R. & LÜLLMANN, C. (1993). Investigation of aroma formation in Robusta coffee during roasting. Cafe Cacao 2(37), 145-152.

SIVETZ, M. (1991). Growth in use of automated fluid bed roasting of coffee beans. In In 14th International Scientific Colloquium on Coffee USA.

SIVETZ, M. & DESROSIER, N. W. (1979). Coffee Technology. Westport: AVI Publishing Company.

SOCRATES, G. (1994). Infrared Characteristic Group Frequencies. New York: John Wiley.

STUART, B. H. (2003). Infrared spectroscopy: Fundamentals and applications. Hoboken, NJ ; Etobicoke, Ont. : John Wiley & Sons

VALDENEBRO, M. S. LEÓN-CAMACHO, M. PABLOS, F. GONZÁLEZA, A. G. & MARTÍN, M. J. (1999). Determination of the arabica/robusta composition of roasted coffee according to their sterolic content. Analyst 124, 999-1002.

VIGNOLI, J. A. BASSOLI, D. G. & BENASSI, M. T. (2011). Antioxidant activity, polyphenols, caffeine and melanoidins in soluble coffee: The influence of processing conditions and raw material. Food Chemistry 124(3), 863-868.

VITZTHUM, O. G. (1975). Chemie und Bearbeitung des Kaffees. Berlin Springer.

VOILLEY, A. SAUVAGEOT, F. SIMATOS, D. & WOJCIK, G. (1981). Influence of Some Processing Conditions on the Quaility of Coffee Brew. Journal of Food Processing and Preservation 5(3), 135-143.

WANG, J. SOOJIN, J. BITTENBENDER, H. C. GAUTZ, L. & LI, X. Q. (2009). Fourier Transform Infrared Spectroscopy for Kona Coffee Authentication. Journal of Food Science 74, C385-C391.

WANG, N. FU, Y. & LIM, L.-T. (2011). Feasibility Study on Chemometric Discrimination of Roasted Arabica Coffees by Solvent Extraction and Fourier Transform Infrared Spectroscopy. Journal of Agricultural and Food Chemistry 59(7), 3220-3226.

91

WOLFROM, M. L. & PATIN, D. L. (1965). Carbohydrates of the Coffee Bean. IV. An Arabinogalactan1. The Journal of Organic Chemistry 30(12), 4060-4063.

WORKMAN JR., J. J. MOBLEY, P. R. KOWALSKI, B. R. & BRO, R. (1996). Review of Chemometrics Applied to Spectroscopy: 1985-95, Part I Applied Spectroscopy Reviews 31(1&2), 73-124.

YAMAMOTO, Y. KUMAMARU, T. HAYASHI, Y. & KANKE, M. (1972). Effect of solvent extraction on the atomic-absorption spectrophotometry in determination of ppM levels of cadmium with dithizone. Talanta 19(8), 953-959.

YERETZIAN, C. JORDAN, A. BADOUD, R. & LINDINGER, W. (2002). From the green bean to the cup of coffee: investigating coffee roasting by on-line monitoring of volatiles. European Food Research and Technology 214(2), 92-104.

YOSHIDA, S. MIYAZAKI, M. SAKAI, K. TAKESHITA, M. YUASA, S. SATO, A. KOBAYASHI, T. WATANABE, S. & OKUYAMA, H. (1997). Fourier transform infrared spectroscopic analysis of rat brain microsomal membranes modified by dietary fatty acids: Possible correlation with altered learning behavior. Biospectroscopy 3(4), 281-290.

ZHANG, W. WANG, X. & FU, X. (2005). In situ FTIR Investigation of Magnetic Field Effect on Heterogeneous Photocatalytic Degradation of Benzene over Pt/TiO2. Chinese Chemical Letters 16(9), 1275-1278.

physicochemical changes of coffee beans during roasting

Documents