Content uploaded by Christoph Hartmann
Author content
All content in this area was uploaded by Christoph Hartmann on Oct 20, 2014
Content may be subject to copyright.
rXXXX American Chemical Society Adx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
ARTICLE
pubs.acs.org/JAFC
Influence of Foam Structure on the Release Kinetics of Volatiles from
Espresso Coffee Prior to Consumption
Susanne Dold,
†,§
Christian Lindinger,
#
Eric Kolodziejczyk,
†
Philippe Pollien,
†
Santo Ali,
†
Juan Carlos Germain,
†
Sonia Garcia Perin,
†
Nicolas Pineau,
†
Britta Folmer,
^
Karl-Heinz Engel,
§
Denis Barron,
†
and Christoph Hartmann*
,†
†
Nestle Research Center, P.O. Box 44, 1000 Lausanne 26, Switzerland
§
Chair General Food Technology, Technische Universit€at M€unchen, Maximus-von-Imhof-Forum 2, 85350 Freising-Weihenstephan,
Germany
#
Rue du Village 5, 1052 Le Mont sur Lausanne, Switzerland
^
Nestle Nespresso S.A., Avenue de Rhodanie 40, 1007 Lausanne, Switzerland
ABSTRACT: The relationship between the physical structure of espresso coffee foam, called crema, and the above-the-cup aroma
release was studied. Espresso coffee samples were produced using the Nespresso extraction system. The samples were extracted with
water with different levels of mineral content, which resulted in liquid phases with similar volatile profiles but foams with different
structure properties. The structure parameters foam volume, foam drainage, and lamella film thickness at the foam surface were
quantified using computer-assisted microscopic image analysis and a digital caliper. The above-the-cup volatile concentration was
measured online by using PTR-MS and headspace sampling. A correlation study was done between crema structure parameters and
above-the-cup volatile concentration. In the first 2.5 min after the start of the coffee extraction, the presence of foam induced an
increase of concentration of selected volatile markers, independently if the crema was of high or low stability. At times longer than
2.5 min, the aroma marker concentration depends on both the stability of the crema and the volatility of the specific aroma
compounds. Mechanisms of above-the-cup volatile release involved gas bubble stability, evaporation, and diffusion. It was concluded
that after the initial aroma burst (during the first 23 min after the beginning of extraction), for the present sample space a crema of
high stability provides a stronger aroma barrier over several minutes.
KEYWORDS: espresso coffee, crema, above-the-cup volatile release, foam structure
’INTRODUCTION
The term aroma refers to perceptions of volatiles in the atmo-
sphere through the olfactory system. It is related to volatile organic
compounds (VOCs) that reach the olfactory epithelium in the
upper part of the nose. During food consumption, the VOCs
released to the headspace can enter the consumer’s nose, leading
to an aroma impression even before the start of eating or drinking.
Whereas in total more than 800 VOCs have been identified in
coffee aroma, several studies
14
have shown that only fewer than
50 of them can be considered as impact aroma compounds.
Espresso coffee is the beverage prepared by short percolation
(30 (5 s) of hot water (90 (5C) at high pressure (above
7 bar) through a layer of roasted and ground coffee (6.5 (1.5 g).
The extraction results in a polyphasic colloidal system. A foam
layer of small bubbles is formed on top of the aqueous solution
with dispersed fine coffee particles and microscopic oil droplets.
5,6
On the physicochemical point of view, the espresso coffee foam,
called crema, is a dynamic biphasic colloidal system. It is com-
posed of gas bubbles framed by liquid films called lamella. The
gas phase of the crema consists of air, water vapor from the
percolation process, carbon dioxide formed in the Maillard reaction
during coffee roasting and present in the roasted and ground
coffee grain asperities, and volatile aroma compounds that are
released from the liquid into the gas phase. Several factors affect
the formation of the crema. The carbon dioxide formed during
the Maillard reaction and the pressure applied during the extrac-
tion are the two main factors.
7,8
In the Nespresso system used in
current study the pressure during the extraction is determined
not only through the particle size and the packing of the coffee
but also through the membrane of the Nespresso capsule that
provides a back-pressure before rupture. The somewhat elastic
aluminum membrane ruptures upon swelling when the pressure
inside the capsule becomes high enough.
When gas is dispersed in water, the increase in interfacial area
increases the free energy of the system. According to the thermo-
dynamic dictum, all systems strive to reach a state of global energy
minimum. Liquid foams are not stable, but collapse by separating
into two phases to minimize the interfacial area and the system’s
free energy. This instability makes foams very difficult to study.
Most foams owe their existence to the presence of surface active
molecular compounds, which decelerate foam destabilization by
reducing the surface energy and by stabilizing the lamella films
against rupture.
9,10
The destabilization of espresso coffee foam
has been ascribed to the mechanism of drainage,
7
meaning that
the liquid between the bubbles drains to the bulk. The lamella
films become thin and eventually rupture. This leads to bubble
coalescence and, in the case of exposed films, to foam collapse,
Received: May 5, 2011
Revised: September 7, 2011
Accepted: September 10, 2011
Bdx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
including the release of entrapped gas and a loss of foam struc-
ture. Thus, the lamella films are of crucial importance for foam
stability. Their thickness is reduced not only by drainage but also
by evaporation, considering the temperature of espresso coffee.
7,912
Besides the roasted and ground coffee, the second most im-
portant ingredient of espresso coffee is water. It constitutes >95% of
the beverage.
6
Calcium and magnesium ions in the water have been
found to decrease foam stability as already claimed by Navarini and
Rivetti.
13
Themechanismcanbeexplainedbyachangeinion
content, respectively by the interaction between the cations and
protein/polysaccharide complexes, leading to a destabilization of
the foaming mechanism.
6,9
The crema of espresso coffee is highly appreciated by con-
sumers for its sensory properties, notably its appearance prior to
consumption. Once in the mouth it contributes to the creaminess
and smoothness of the drinking experience. Crema is tradition-
ally believed to act as an aroma-sealing lid that traps the volatilized
compounds and doses their emission into the atmosphere.
6,7,14
In the current contribution, we propose that the crema structure
affects and supports volatile release.
Given the importance of aroma in coffee consumption and the
consumers’appreciation of crema, the objective of this work was
to investigate the influence of the physical structure of espresso
coffee foam on headspace volatile release. Several parameters of
crema structure were analyzed with regard to their impact on
volatile release above the cup.
’MATERIALS AND METHODS
Materials. Two different blends of commercial Nespresso capsules
were used: blend A (A) (lot 92333786PD 17:49 B) and blend B (B)
(lot MOUON-13 07:39 A). Both types of capsules contained pure
Arabica coffee. The main sensory attributes for blend A are roasted and
cocoa notes; blend B has roasted and fruity notes. For coffee extraction
commercial bottled waters of different mineral contents were used:
Acqua Panna (Pan) (lot L9245087701) for its mediumlow miner-
alization (calcium, 32.9 mg/L; magnesium, 6.5 mg/L) and Contrex
(Con) (lot 92473019N6) for its high mineralization (calcium, 468.0 mg/L;
magnesium, 74.5 mg/L). For all extractions a commercial Nespresso
machine C190 Plus was used.
Foam Structure Analysis Methods. The elapsed time before the
start of foam structure analysis and the elapsed time before the start of
the volatile release analysis methods was synchronized. The starting
point of all measurements, t= 0 min, was set as being 1 min after the start
of coffee extraction.
Foam Volume. Foam volume was determined using a digital caliper
(Garant IP65-CI50, Hoffmann Group, Munich, Germany). Measure-
ments were carried out in a commercial Nespresso glass cup (Essenza
espresso cup ref 3301/2/B). The distance from the top of the cup to a
certain fill level was linked to the respective fill volume by doing a
calibration with distilled water of 20 C. Volumes were plotted against
filling heights, and a third-order polynomial was fitted as calibration
curve. For foam volume analysis, 40 g of coffee beverage was extracted
into the cup. The distance from the top of the cup to the foam surface
and from the top of the cup to the liquid surface was measured with
the digital caliper. Corresponding filling volumes were calculated using
the third-order polynomial fit. The foam volume was calculated as the
difference of the two volumes. The measurement was done at t= 0 min
and at t= 5 min (equaling 1 and 6 min after the start of coffee extraction).
Ten repetitions were done for each sample. Averages and standard
deviations were calculated.
Foam Drainage. Foam drainage was determined using computer-
assisted microscopic image analysis. An amount of 25 g of coffee beverage
was extractedinto a plastic vessel. The plastic vessel had a transparent and
plane side wall, which allows a cross-sectional view of the crema and the
liquid volume beneath the crema. The plastic vessel was placed into a dark
chamber facing a digital microscopy camera (Infinity Y2-1C, Lumenera
Corp., Ottawa, Canada). Microscopic images were taken every 15 s,
starting at t= 0 min (1 min after the start of coffee extraction). In total
40 images were taken. The microscopic images were analyzed using the
image processing software Colibri developed in-house. For each picture,
the area (level) of the liquid underneath the foam was calculated. Data
were normalized by equating the area at t= 0 min to 100%. This allowed a
relative comparison between the drainage kinetics of thedifferent samples.
Triplicates were done for each sample. Averages and standard deviations
were calculated for each data point.
Foam Surface Area Fraction. The quality of the foam surface
was analyzed using computer-assisted microscopic image analysis. An
amount of 40 g of coffee beverage was extracted into a commercial
Nespresso porcelain cup and placed into a dark chamber under a digital
microscopy camera (Infinity Y2-1C, Lumenera Corp.) and a luminous
ring (Schott S40-55, Schott AG, Mainz, Germany). Five samples were
extracted and analyzed for each of the four investigated coffee beverages.
For each of the five samples, a series of 20 microscopic images of the
complete illuminated foam surface was taken at intervals of 30 s starting
at t= 0 min (1 min after the start of coffee extraction). Positioning of the
camera and coffee sample remained the same throughout the measure-
ments. The microscopic images were analyzed using the image proces-
sing software Colibri (noncommercialized in-house development). Each
color image was transformed into a black/white image. For each black/
white image the area fraction was calculated as the ratio between the area of
white parts of the crema surface in the transformed image and the total area
of the crema surface. Averages and standard deviations have been calculated
from the five repetitions for each time point and for each coffee beverage.
Volatile Release Analysis Methods. The above-the-cup aroma
intensity of the samples was measured online by proton transfer reaction
mass spectrometry (PTR-MS). The setup consisted of a headspace
sampling oven linked to a proton transfer reaction mass spectrometer.
Identification of the chemical compounds contributing to the ion trace
markers monitored by PTR-MS was done. The VOCs were identified by
using Tenax trapping and gas chromatography (GC) with simultaneous
and parallel detection by PTR-MS and time-of-flight mass spectrometry
(TOF-MS). The volatile release analysis methods used for this work
were described in detail by Lindinger et al.
1517
Online Headspace Volatile Release. A double-jacketed, water-
heated sample cell (glass vial) was placed inside an oven at a temperature
of 100 C with active air circulation. A water bath at a temperature of
50 C was connected to the sample cell to keep the sample at constant
temperature. The sample cell was connected to the fix-installed top of
the cell by a clamp and sealed by a silicone O-ring. Three tubes were
installed in the cover of the cell. The first tube was supplied by the
preheated purge gas, the second provided a thermocouple to measure
the sample temperature, and through the third one the sample gas was
pushed out of the vial. Before analysis by PTR-MS, the sample gas was
diluted with dry air. Two flow controllers provided the dilution gas and
the purge gas, keeping the dilution ratio constant. Only a small fraction
of sample gas (40 standard cubic centimeters per minute (sccm)) was
used for analysis by PTR-MS (high-sensitivity PTR-MS, Ionicon Analytik
GmbH, Innsbruck, Austria). The major part was removed through an
exhaust line. Dry synthetic air was used as purge gas and dilution gas. The
flow rate of the purge gas was maintained at 300 sccm, the one of the
dilution gas at 3000 sccm.
To analyze the impact of crema, volatile release was measured for each
of the four beverages with crema (wc) and without crema (woc). For the
preparation of a sample without crema, a glass funnel covered with two
paper filters (KIMTECH Science precision wipes, Kimberly-Clark
Corp., Dallas, TX) was put on the sample cell and 40 g of coffee brew
Cdx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
was extracted and filtered. On average, a residue of 3 g of crema mixed
with coffee was held back, and 37 g of coffee brew was collected into the
sample cell. The elapsed time from the start of the extraction to the end
of filtration (45 s) and then to connection of the sample cell to the
headspace sampling oven (15 s) was maintained constant at 60 s. To
prepare a sample with crema, the coffee was directly extracted into the
sample cell. On average, 40 g coffee brew was collected. As for the coffee
preparation without crema, the sample cell was connected to the oven
setup 60 s after the start of the coffee extraction. Therefore the time delay
(45 s) caused by filtering in the case of extraction without crema did not
influence the comparability of the measurements.
The concentration (parts per billion) in the gas phase of 58 selected
ion traces was obtained during 10 consecutive minutes after connection
of the sample cell. The 58 ion traces chosen showed to be the most
discriminating over the mass range m/z20160 for coffee headspace
analysis. For each analysis three individual headspace release measure-
ments were done. Between consecutive measurements, the sample cell
was equilibrated in the oven setup for 30 min. Ion trace release profiles
were obtained by calculating average concentrations and standard
deviations for each data point (each 12.4 s).
Identification of Volatile Organic Compounds. To analyze
the chemical compounds contributing to the ion traces monitored by
PTR-MS, entrapment on Tenax traps was used. Tenax traps were
desorbed with an automatic thermodesorption unit (ATD), which was
connected to a GC column. Tenax traps were desorbed at 300 C for
10 min on the ATD unit (ATD Turbo Matrix 350, PerkinElmer Inc.,
Boston, MA) and purged with a helium flow of 20 sccm to the ATD trap.
The volatile compounds were cryofocused at 30 C for 10 min,
desorbed at 320 C for 3 min, and injected into the GC (Agilent 6890N
series GC, Agilent Technologies Inc., Santa Clara, CA) at 200 C.
A 60 m DB-Wax column (J&W scientific Inc., Folsom, CA) with an
internal diameter of 0.32 mm and a film thickness of 0.5 μm was used.
The temperature was kept at 20 C for 5 min, increased at a rate of
4C/min to 240 C, and maintained for 10 min. Helium at 3 sccm was
used as flow carrier gas. After thermodesorption and GC separation,
the eluent was split (1:1) to two detectors that were supplied parallel
and simultaneously: TOF-MS (Pegasus III TOF-MS, LECO Corp., St.
Joseph, MI) for identification based on the fragmentation patterns and
PTR-MS (high-sensitivity PTR-MS, Ionicon Analytik GmbH) to iden-
tify the ion trace found online and to semiquantify contribution
compounds by their GC peaks for a selected ion corresponding to the
same ion trace in the online mode. To allow a proper coupling of GC
with PTR-MS, the effluent gas of the GC column had to be mixed with
moist air prior to splitting. As the mobility of ions depends on the buffer
gas and its humidity, these parameters had to be kept constant to obtain
the same PTR-MS fragmentation pattern. In the online mode dry
synthetic air was used as buffer gas for the humid headspace sample.
The gas from the GC column in the offline mode was dry helium. To
compensate for this, the PTR-MS reaction chamber was fed by two inlets
in the offline mode. One came from the GC column carrying the dry
helium. The other one came from the PTR-MS headspace oven setup,
where the sample cell was filled with 100 mL of purified water to mimic
the humidity of the headspace sample gas in the online mode.
The identification of the chemical compounds contributing to the ion
traces monitored by PTR-MS was done by comparing GC relative
retention times (RI) and fragmentation patterns detected by TOF-MS
to commercial and internal databases (Wiley Registry of Mass Spectral
Data, 9th edition, Wiley, European Distribution Centre, PO22 9NQ
UK; FMD, internal Flavor Molecules Database).
’RESULTS AND DISCUSSION
Physical Foam Structure. Foam Volume. Foam volumes of
the four espresso coffee samples were determined with a digital
caliper. Results are shown in Figure 1. A ttest (two-sided, unpaired,
α
risk
= 0.01) was applied to check for significant differences. The
risk level α
risk
= 0.01 was the threshold level we fixed to confirm the
significance of a difference. The pvalues produced from the ttests
were compared to α
risk
. If the observed pvalue was below α
risk
,we
concluded that the difference was significant.
The initial foam volumes (t= 0 min) are not significantly
different between the beverages prepared with Acqua Panna and
those prepared with Contrex. At t= 5 min, the foam volumes of
espresso coffees prepared using different waters show significant
differences. Higher foam volumes were measured for the espres-
so coffees prepared with Acqua Panna than for those prepared
with Contrex. For both types of coffee capsules, the foam of
espresso coffees prepared with mediumlow mineralized Acqua
Panna shows higher stability in volume than the foam of the
corresponding espresso coffees prepared with the highly miner-
alized Contrex.
Foam Drainage. The amount of liquid draining from the
espresso coffee foam was obtained by measuring the increase of
the area-projected volume of liquid phase underneath the foam.
Data were normalized by equating the area ofliquid phase present
at t= 0 min (equals 1 min after the start of coffee extraction) to
100%. Figure 2 shows averages and standard deviations obtained
of triplicates for the four espresso coffee samples. The espresso
coffees prepared with Contrex show a higher increase of the area of
liquid underneath the foam than the corresponding beverages
prepared with Acqua Panna. This observation applies to each data
point within the study period of 10 min. For both types of coffee
capsules, blends A and B, more liquid drains out of the foams of
espresso coffees prepared with Contrex than of the foams of
espresso coffees prepared with Acqua Panna.
Foam Surface Area Fraction. Figure 3 shows average area
fractions and standard deviations obtained of five measurement
repetitions for each of the four different espresso coffee beverages.
All four espresso coffee samples show the same area fraction of
Figure 1. Foam volumes of espresso coffees prepared from blends
A and B, both using Acqua Panna (Pan) and Contrex (Con) at t= 0 min
and at t= 5 min (equaling 1 and 6 min after the start of coffee extraction)
(averages and standard deviations of 10 repetitions shown).
Ddx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
approximately 0.6 at t= 0 min. The temporal evolution of the
area fraction shows clear differences between the beverages
prepared with Acqua Panna and those prepared with Contrex.
At elapsed times >6 min after the start of the coffee extraction,
the area fractions of both espresso coffees prepared with Acqua
Panna remain at higher values compared to those of the espresso
coffees prepared with Contrex. Consequently, the kinetics of the
area fraction depends on water ion content. The area fraction
indicates the intensity of light reflected by the crema surface. The
visual perception “white”in the black/white images results from
reflected light. According to basic electromagnetic theory, the
capacity to reflect light depends on the thickness of the lamella
films located at the foam surface.
10
Thus, the results of the area
fraction can be interpreted as follows: All four espresso coffees
show approximately the same value for the area fraction at t=0
min, meaning that the initial film thickness of the lamella films
located at the top of the crema does not differ considerably
between samples. The faster decrease of the area fraction for
espressos prepared with water of higher mineral level compared
to espressos prepared with water of lower mineral level indicates
a more pronounced thinning of the lamella films located at the
foam surface.
Generally, a decrease in foam volume can be assigned to phase
segregation, implying the loss of gas into the surrounding air and
the loss of fluid into the liquid bulk due to drainage. The latter
implicates thinning of the lamella films, which in turn leads to
film rupture and foam collapse, including a loss of entrapped gas
and a loss in foam volume.
7,9,10
Thus, the observations on foam
volume, foam drainage, and foam surface quality support each
other and demonstrate that the foams of espresso coffees prepared
with water of high mineralization show a lower stability than the
foams of the corresponding espresso coffees prepared with water
of low to medium mineralization.
A possible explanation for the lower stability measured for
espresso coffees prepared with water of high mineral content is
the potential disturbing effect of the ions on the interactions
between the foaming fractions. Surfactant molecules can bind to
proteins and polysaccharides, forming surfactantbiopolymer
complexes, which may have functional characteristics different
from those of the individual components. The interaction can be
based on different mechanisms, whereby the most important are
electrostatic and hydrophobic interactions. These have a strong
influence on physicochemical properties of the system, which, in
the case of espresso coffee, are directly linked to the physical crema
structure. It is possible that the ions of the high mineralized water
disturb these electrostatic interactions of surfactantbiopolymer
complexes and change the stability of the surface physical
network.
8,19
Volatile Release above the Cup. The headspace volatile
release was measured online using PTR-MS. Each of the 58
monitored ion traces shows its own release pattern. Within the
framework of this study, some markers were selected for further
investigation and discussion. Three selection criteria were applied:
First, the signal-to-noise ratio of the ion trace release profile
should be >10.
18
Second, the selected markers should represent
potential coffee aroma compounds. This selection criterion was
applied by using an internal database (FMD, internal Flavor
Molecules Database). However, compounds such as 2-furfurylthiol,
Figure 3. Area fraction (calculated by dividing the area of white parts of
the crema surface in the black/white images by the total area of the
crema surface) starting at t= 0 min (equals 1 min after the start of coffee
extraction) for blends A and B, both prepared with Acqua Panna (Pan)
and Contrex (Con) (averages and standard deviations of five repetitions
shown).
Figure 2. Increase of area projected volume of liquid phase underneath
espresso coffee foam starting at t= 0 min (equals 1 min after the start of
coffee extraction) for blends A and B, both prepared with Acqua Panna
(Pan) and Contrex (Con) (averages and standard deviations of tripli-
cates shown).
Edx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
3-mercapto-3-methylbutylformate, methional, β-damascenone,
and Furaneol known to have a high aroma impact are not
covered by the markers, because they are present in coffee
headspace at concentrations too low to be detectable by online
techniques.
14
Third, ion traces associated with acetic acid were excluded, as
this molecule showed disturbing effects due to interaction of the
molecule with the tubing system of the headspace sampling
setup. By applying these three selection criteria, 8 of 58 ion traces
were chosen for further analysis for blend A (m/z45, 59, 69, 73,
75, 81, 87, and 95), and 7 of 58 ion traces (m/z45, 59, 69, 73, 75,
81, and 87) were chosen for blend B.
Figures 4 and 5 exemplarily show the release profiles of m/z
45, 59, and 81 for blends A and B, prepared with Contrex and
Acqua Panna, both with crema (wc) and without crema (woc).
Table 1 shows the chemical components contributing to m/z
45, 59, and 81. Each marker represents several chemical com-
pounds. Thus, the release profiles need to be considered as the
superposition of release profiles of all contributing compounds.
The three markers m/z45, 59, and 81 represent volatiles with
both very high and very low volatilities. The other investigated
markers are similar to or between the extremes represented by
these three marker characteristics.
Within the first 2.5 min after the start of the coffee extraction,
the presence of crema generally resulted in an above-the-cup
volatile concentration significantly higher than that of the liquid
coffee phase without crema. This was shown for both coffee
blends and for all investigated release patterns. After this initial
burst, the impact of crema is more differentiated. Depending on
the investigated group of chemical compounds and on the type
of water used, the crema on top of the liquid coffee phase can
either act as long-term enhancer of the volatile release above cup
or have no effect or act as a long-term barrier of volatile release.
These findings demonstrate that the concept of crema being an
aroma-sealing lid
7,14
cannot be generalized. When the impacts of
foams with different stability patterns starting at t= 2.5 min are
compared, it can be observed within the given sample space that
the volatile concentration above the cup is higher for espresso
coffees prepared with water with high mineral content (here:
Contrex). For the corresponding liquid coffee phases without
crema, no considerable influence of the type of water used for coffee
preparation could be evidenced. With regard to this finding, it is
suggested that espresso coffees with a crema of lower stability
provide higher above-the-cup volatile concentrations at longer
times after the extraction (>2.5 min). To confirm this prelimin-
ary conclusion, the data of physical foam structure and volatile
release were used for a statistical correlation study.
Correlation of Physical Foam Structure and Above the
Cup Volatile Release. For espresso coffee samples prepared
with crema (A Pan wc, A Con wc, B Pan wc, and B Con wc), a
normalized principal component analysis (PCA) was done
considering the markers m/z45, 59, 69, 73, 75, 81, and 87.
Variables (i.e., markers) are represented as lines. The smaller the
angle between two variables, the better is the correlation. Indivi-
duals (i.e., coffee samples) are represented as dots. When a coffee
sample is positioned at large distance from the center (reference)
and close (orthogonal projection) to the vector of a marker, the
coffee sample has a high value for this marker. In addition, the
data obtained for the structure parameters foam volume
(FoamVol), foam drainage (FoamDrain), and foam surface area
fraction (FoamSAF) are added as supplementary variables (i.e.,
these variables are added a posteriori on the PCA map).
Three different PCAs were done. The first one using data
obtained at tburst, the time corresponding to the highest aroma
release intensity, which occurs around 45 s after the start of the
headspace release measurements (Figure 6). The second PCA was
Figure 4. Volatile release profiles of PTR-MS ion traces m/z45, 59, and
81 for samples prepared from blend A with Contrex (Con) and Acqua
Panna (Pan), both with crema (wc) and without crema (woc) (t= 0 min
equals 1 min after starting coffee extraction) (averages and standard
deviations of triplicates).
Fdx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
done on data obtained at t= 5 min, equaling 6 min after the start of
the coffee extraction (Figure 7). The third PCA was done on the
calculated differences between the data obtained at tburst and the
data obtained at t=5min(tburst t= 5 min) (Figure 8).
The PCA at tburst (Figure 6) shows that the essential infor-
mation is reflected by the first principal component (93%,
horizontal axis). It is almost one-dimensional. The volatile
release of all seven markers is highly negatively correlated with
the foam surface area fraction. From the results of the PCA for
this time point it can be concluded that the thinner the lamella
films located at the foam surface, the higher is this initial aroma
burst. There is no correlation with the foam drainage or the foam
volume at this short time after extraction. It cannot be observed
either that foams of different stabilities cause different initial
above-the-cup aroma intensities. Therefore, we suggest that the
rupture of the thin exposed lamella films leads to bubble collapse
and release of entrapped gas.
7,9,10
Thus, the initial burst could
be explained as follows: The VOCs entrapped in the gas phase of
the foam bubbles are released into the headspace as the bubbles
at the crema surface explode due to rupture of thin exposed lamella
films. Thinning of films at the top of foams is caused either by
evaporation of the liquid phase or through gravity-driven drainage.
10
As no correlation with foam drainage was shown and considering
the temperature of freshly brewed coffee, we concluded that
evaporation is a major influencing phenomenon here. At the high
exchange surface area of the crema, volatiles will evaporate from
the liquid phase to the gas phase above the cup and will thus be
available for perception. Rate-limiting effects are unlikely for the
liquid phase, because the liquid is exposed to natural convection.
Consequently, in the presence of crema the mechanisms of
volatile release are lamella rupture and evaporation. The high
exchange surface area of crema allows a burst of both high and
low volatile aroma compounds. In the absence of crema the
mechanism of volatile release is purely driven by evaporation at
the liquid-gas interface.
The PCA of data obtained 6 min after the start of the coffee
extraction (t= 5 min) (Figure 7) is almost one-dimensional, too,
but shows correlations other than the PCA at tburst. The volatile
release is strongly negatively correlated with the foam volume
and negatively correlated with the foam surface area fraction. It
can be concluded that the thinner the lamella films at the foam
surface and the lower the foam volume, the higher is the volatile
concentration above the cup. Large foam drainage can also be
considered as contributor to a high above-the-cup aroma in-
tensity because the corresponding vector in the PCA has a small
angle with most marker vectors. A low foam volume, thin lamella
films at the foam surface, and drainage of the liquid out of the
foam into the bulk are all indicators for foam destabilization.
9,10,12
Accordingly, for all coffee samples, the PCA for t= 5 min reveals
that the volatile release of espresso prepared with the highly
mineralized water (Contrex) is globally higher than the volatile
release of espresso prepared with low mineralized water (Acqua
Panna). Thus, this PCA confirms the assumption that was
already made on the basis of the results of the foam structure
analysis and the volatile release analysis. After the initial aroma
burst, a crema of low stability gives higher above-the-cup volatile
concentrations than a crema of high stability in the present
sample space. This can be explained to a large extent by the fact
that foam bubbles rupture due to thin lamella films and release
the trapped volatiles into the surrounding air. For a more stable
crema the VOCs remain trapped in the gas bubbles for a longer
time. For the initial aroma burst it was suggested that evaporation
leads to thinning of lamella films. From the correlation observed
between above-the-cup volatile concentration and foam destabi-
lization for the situation after the initial aroma burst, the thinning
of lamella films can be attributed to drainage.
Figure 5. Volatile release profiles of PTR-MS ion traces m/z45, 59, and
81 for samples prepared from blend B with Contrex (Con) and Acqua
Panna (Pan), both with crema (wc) and without crema (woc) (t= 0 min
equals 1 min after the start of coffee extraction) (averages and standard
deviations of triplicates).
Gdx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
In a third PCA, differences in the volatile release profiles
between tburst and t= 5 min were correlated to differences in the
foam structure between tburst and t= 5 min. The PCA for the
difference Diff(Figure 8) illustrates that the decrease of volatile
concentration of markers m/z73, 87, 45, and 59 is positively
correlated with the decrease of foam drainage and negatively
correlated with the decrease of foam surface area fraction.
For these markers, the decrease in volatile release is stronger
for the espresso coffees prepared with Acqua Panna than for the
espresso coffees prepared with Contrex. Correspondingly, for a
more stable crema with less drainage and thicker lamella films, we
observed a stronger decrease in volatile release with respect to its
maximum level of intensity. However, for three other groups of
chemical compounds (markers m/z81, 69, and 75), the PCA
indicates a much weaker correlation between foam stability and
volatile release as compared to the four markers mentioned
above. The decrease of the volatile release for the markers m/z81
and 69 tends to correlate strongly with the decrease in foam
volume, whereby it has to be considered that m/z75 is only
weakly represented in this PCA. It is suggested that the different
correlations of the markers are related to the volatility of the
different compounds (Table 1). The low-volatile compounds are
less present in the gaseous phase of the crema, and thus their
concentration above the cup is less correlated to the destabiliza-
tion process of the crema.
Four effects are proposed to take place depending on the
volatility of the aromas and the stability of the crema: For high-
volatile compounds (K=10
3
10
2
) and a crema of low
stability, the crema acts as an enhancer for aroma release. High
volatiles are abundant in the gas phase of the foam bubbles and
released upon their rupture. Examples for this effect are the
release patterns of Contrex m/z45 and Contrex m/z59. For low-
volatile compounds (K=10
4
10
3
) and a crema of low
stability, the creation of new interface upon bubble rupture
increases the diffusion of low-volatile aromas. The release pattern
of Contrex m/z81 shows this behavior. For high-volatile
compounds and a crema of high stability, the crema acts as a
barrier that entraps volatiles. Acqua Panna m/z45 and Acqua
Panna m/z59 are examples for this effect. For low-volatile
aromas and a crema of high stability, the aromas will be released
from the interface through a diffusion process, as, for example,
shown by the release pattern of Panna m/z81.
Table 1. Chemical Compositions of the PTR-MS Ion Traces
a
m/z45, 59, and 81 Obtained by GC-MS Coupling
16
for Blends A and
B, Both Prepared with Acqua Panna and Contrex
b
m/z45 (fruity, green, malty notes) m/z59 (buttery, green notes) m/z81 (roasted notes)
compound K
air/water
compound K
air/water
compound K
air/water
acetaldehyde 1.63 10
2
acetone n/a 2-furanmethanol 1.19 10
4
2-methylbutanal 8.36 10
2
2,3-butanedione 5.98 10
3
2-furfuryl formate n/a
3-methylbutanal 7.91 10
2
propanal 1.80 10
2
2-furfuryl acetate 6.12 10
3
2-methyltetrahydrofuran-3-one n/a pyridine n/a
pyrazine n/a
a
Tentatively identified using commercial and internal databases.
b
Sensory descriptors (internal data bases) of odorant compounds identified in the
chemical composition of PTR-MS ion traces were assigned directly to the corresponding masses. Air/liquid partition coefficients K
air/water
(without unit)
were measured by stripping with pure air an aqueous solution containing individually the dissolved chemical compounds at mg/L levels. Depletion of
VOC’s relative concentration in the gas phase measured at 60 C by PTR-MS allowed determination of partition coefficients.
20
Figure 6. PCA of foam structure parameters and volatile release of
selected markers (vectors) for tburst (105 s after the start of coffee
extraction) for the four different espresso coffee samples with crema
(points).
Figure 7. PCA of foam structure parameters and volatile release of
selected markers (vectors) for t= 5 min (6 min after the start of coffee
extraction) for the four different espresso coffee samples with crema
(points).
Hdx.doi.org/10.1021/jf201758h |J. Agric. Food Chem. XXXX, XXX, 000–000
Journal of Agricultural and Food Chemistry ARTICLE
Conclusion. These results show for the first time the influence
of crema stability properties on the volatile release profiles above
the cup. In summary, during the first 2.5 min after the start of the
espresso coffee extraction, the presence of crema provides a
volatile burst above the cup, no matter if the crema is of high or
low stability. Six minutes after the start of the extraction, a crema
of low stability gives higher above-the-cup volatile concentrations
than a crema of high stability. After the initial aroma burst, a
crema of low stability provides a more long-lasting high above-
the-cup aroma intensity.
It is still an open question whether these release kinetics and
their differences can be perceived by a consumer or not, and
beyond, which kinetics would then be the most liked. It should be
kept in mind that in an average consumption behavior the first sip
would occur during the burst of the aromas above the cup
(during the first 2.5 min), and most probably the cup is fully
consumed after 5 min. To provide a sensory validation, time-
resolved sensory studies and consumer preference studies are
required.
’AUTHOR INFORMATION
Corresponding Author
*Phone: +41 21 785 8102. E-mail: Christoph.Hartmann@rdls.
nestle.com.
’ABBREVIATIONS USED
VOCs, volatile organic compounds; A, blend A;B, blend B; Pan,
Acqua Panna; Con, Contrex; PTR-MS, proton transfer reaction
mass spectrometry; GC, gas chromatography; TOF-MS, time-of-
flight mass spectrometry; sccm, standard cubic centimeters
per minute; ppb, parts per billion; ATD, automatic thermode-
sorption; wc, with crema; woc, without crema; PCA, principal
component analysis; RI, retention index (time); FoamVol, foam
volume; FoamDrain, foam drainage; FoamSAF, foam surface area
fraction; α
risk
, risk level of significance used in the ttest.
’REFERENCES
(1) Blank, I.; Sen, A.; Grosch, W. Aroma impact compounds of
Arabica and Robusta coffee. Qualitative and quantitative investigations.
In Proceedings of the 14th International Colloquium on the Chemistry of
Coffee, San Francisco, CA; ASIC: Paris, France, 1991; pp 117-129.
(2) Holscher, W.; Steinhart, H. Investigation of roasted coffee
freshness with an improved headspace technique. Z. Lebensm.-Unters
Forsch. 1992,195,33–38.
(3) Semmelroch, P.; Grosch, W. Analysis of roasted coffee powders
and brews by gas chromatography-olfactometry of headspace samples.
Lebensm.-Wiss. Technol. 1995,28, 310–313.
(4) Pollien, P.; Krebs, Y.; Chaintreau, A. Comparison of a brew and
an instant coffee using a new GC-olfactometric method. In Proceedings of
the 17th International Colloquium on the Chemistry of Coffee, Nairobi,
Kenya; ASIC: Paris, France, 1997; pp 191-196.
(5) Navarini, L.; Ferrari, M.; Suggi Liverani, F.; Liggieri, L.; Ravera,
F. Dynamic tensiometric characterization of espresso coffee beverage.
Food Hydrocolloids 2004,18, 387–393.
(6) Petracco, M. Beverage preparation: brewing trends for the new
millennium. In Coffee. Recent Developments, 1st ed.; Clarke, R. J., Vitzthum,
O. G., Eds.; Blackwell Science: Abingdon, U.K., 2001; pp 140164.
(7) Petracco, M. The cup. In Espresso Coffee The Science of Quality,
2nd ed.; Illy, A., Viani, R., Eds.; Elsevier Academic Press: Amsterdam,
The Netherlands, 2005; pp 290315.
(8) Piazza, L.; Gigli, J.; Bulbarello, A. Interfacial rheology study of
espresso coffee foam structure and properties. J. Food Eng. 2008,
84, 420–429.
(9) Damodaran, S. Protein stabilization of emulsions and foams.
J. Food Sci. 2005,70 (3), 54–66.
(10) Weaire, D.; Hutzler, S. The Physics of Foams, 1st ed.; Oxford
University Press: Oxford, U.K., 1999.
(11) Navarini, L.; Barnaba, M.; Suggi Liverani, F. Physicochemical
characterization of espresso coffee foam. In Proceedings of the 21st
International Colloquium on the Chemistry of Coffee, Montpellier, France;
ASIC: Paris, France, 2006; pp 320-327.
(12) Walstra, P. Principles of foam formation and stability. In Foams:
Physics, Chemistry and Structure, 1st ed.; Wilson, A. J.,Ed.; Springer-Verlag:
Berlin, Germany, 1989; pp 116.
(13) Navarini, L.; Rivetti, D. Water quality for espresso coffee. Food
Chem. 2010,122, 424–428.
(14) Maeztu, L.; Sanz, C.; Andueza, S.; De Pe~na, M. P.; Bello, J.; Cid,
C. Characterization of espresso coffee aroma by static headspace GC-MS
and sensory flavor profile. J. Agric. Food Chem. 2001,49, 5437–5444.
(15) Lindinger, C.; Pollien, P.; Ali, S.; Yeretzian, C.; Blank, I. Un-
ambiguous identification of volatile organic compounds by proton-transfer
reaction mass spectrometry coupled with GC/MS. Anal. Chem. 2005,
77, 4117–4124.
(16) Lindinger, C.; Labbe, D.; Pollien, P.; Rytz, A.; Juillerat, M. A.;
Yeretzian, C.; Blank, I. When machine tastes coffee: instrumental
approach to predict the sensory profile of espresso coffee. Anal. Chem.
2008,80, 1574–1581.
(17) Lindinger, C.; Pollien, P.; De Vos, R. C. H.; Tikunov, Y.;
Hageman, J. A.; Lambot, C.; Fumeaux, R.; Voirol-Baliguet, E.; Blank, I.
Identification of ethyl formate as a quality marker of the fermented off-
note in coffee by a nontargeted chemometric approach. J. Agric. Food
Chem. 2009,57, 9972–9978.
(18) Gross, J. H. Mass Spectrometry, 1st ed.; Springer-Verlag: Berlin,
Germany, 2004.
(19) Chen, J.; Dickinson, E. Protein/surfactant interfacial interactions
Part 2. Electrophoretic mobility of mixed protein + surfactant systems. Coll.
Surf. A: Physicochem. Eng. Asp. 1995,100,267–277.
(20) Pollien, P.; Jordan, A.; Lindinger, W.; Yeretzian, C. Liquid-air
partitioning of volatile compounds in coffee: dynamic measurements
using proton-transfer-reaction mass spectrometry. Int. J. Mass Spectrom.
2003,228,69–80.
Figure 8. PCA of foam structure parameters and volatile release of
selected markers (vectors) for the difference Diff(values for tburst
values for t= 5 min) for the four different espresso coffee samples with
crema (points).
