Coffee production is constantly evolving in terms of agronomic techniques, selection of new varieties that adapt to different climatic conditions, and expressing optimal organoleptic characteristics. Moreover, different varieties of Arabica respond differently to the treatments of the drupes in post-harvest, showing their best characteristics, according to the technology applied: “natural coffee” and “washed coffee”. The two varieties discussed here represent new varietal selections, which correspond to new Arabica coffee crus. The results obtained by the three extraction methods of coffee are reported below.
3.1. Crema Analysis
It is well-known that the foam stability and texture represent a very important qualitative parameter in espresso coffee both when it is prepared at the bar, thanks to high-pressure coffee machines, and when it is obtained by means of devices that use capsules. For this reason, the stability, visual perception, and size of bubbles in espresso coffees obtained through the two extrusion methods at 19 bar (Nespresso®, N) and 15 bar (Dolce Gusto®, DG) in the presence of capsules were studied. The same coffee selections extracted through the moka did not show foam (due to the low pressure of the water vapor), so they were not considered in this phase of the research.
The coffee foam is a thick layer of very small gas bubbles consisting of water vapor due to the percolation process and carbon dioxide that is formed with the Maillard reaction during coffee roasting [
16]. Coffee foam, texture, and persistence play a very important role in qualitative and sensorial terms in espresso coffee even if the factors that influence the formation and stability of the foam are still unclear [
17]. Coffee foam is a dynamic and unstable system, which tends to collapse after a short time. In this study, the observation of the persistence and thickness of the foam collar was carried out for 10 min.
As reported in
Figure 1, the foam of the Guatemala and Brazil coffees seemed more stable when obtained with the extraction method at 19 bar (N) compared to the same selections extracted at 15 bar (DG).
Therefore, when a higher pressure (19 bar, N) is used in the extraction of the coffee from the capsules, a more stable foam is obtained; on the contrary, at 15 bar (DG), the lower extraction pressure is responsible for a greater instability of the foam, which after 10 min tends to disappear completely regardless of Guatemala (DGG) and/or Brazil (DGB) coffee origins. However, as shown in
Figure 1, NG coffee (washed Guatemala) has greater persistence than Brazil coffee extracted under the same conditions, and this could be due to the different composition of soluble carbohydrates, proteins, and lipids present in the composition of washed Guatemala coffee (data not shown). These molecules could affect the greater stability of the foam, which remains stable at 40% of the initial value even after 10 min of observation. Therefore, as shown in
Figure 1, the extraction of the capsules at 19 bar (N) allows to obtain a more stable foam, and this feature is more present in washed Guatemala coffee.
Coffee crema could be defined as a continuous liquid phase surrounding a dispersed phase of gas bubbles. Each bubble belonging to the foam has a shape that gives a minimal surface area and depends, for thermodynamic reasons, on the presence of nearby bubbles. The liquid phase solubilizes biopolymers, with the function of surfactants, which participate in the formation of lamellae, i.e., the liquid films that trap the gas phase and stabilize the bubbles against coalescence [
18].
Figure 2A shows the average score obtained by means of sensory analysis on the appearance of the crema (foam), expressed as a visual attribute linked to fineness, appearance, uniformity, and fine creaminess. The Guatemala (NG) and Brazil (NB) coffees extracted from the capsules at 19 bars showed a crema with greater finesse; on the contrary, DGG and DGB cremas had a coarser texture with a lower average score than the one extracted from the capsules at higher pressure (average score around 4–4.5).
As reported by Luidwig (2014) [
19] and Angeloni (2019) [
20], not only could the characteristics related to the varieties and geographical origin of the coffee affect the quality of the beverage but also physical variables, such as extraction time, temperature, etc. The molecules responsible for the formation of foam present in coffee and that stabilize the interfacial layer of bubbles and emulsions are mostly lipids and proteins that create a surface physical network. Therefore, the main differences found in the stability and drainage of the foam depend on the structural characteristics of the bubbles.
Figure 2B shows a stereomicroscope image of the coffee crema at T0 (in this case, the NG sample). Considering that the foams of the other coffees observed under the stereomicroscope did not show evident differences, the other images are not shown.
As shown in
Figure 2B, the bubbles are spherical, and at sight, they are apparently uniform. As reported by Chen and Dickinson (1995) [
21] and McClements (2005) [
22], the high pressure during the extraction of the coffee from the capsules emulsifies part of the oils, which are responsible for the creamy perception; in addition, during the extraction of the espresso, the foam is stabilized by the interactions between surfactants and biopolymers, which, as a whole, affect the entire structure and are closely linked to the perception of mouthfeel and the flavor of the coffee.
Since foams are transient systems, which tend to collapse over time, the stereomicroscope images were studied at T0 and after 10 min (T10), with the aim of observing the microscopic variations of the foam bubbles during drainage. In fact, in the initial stage, the bubbles are spherical, and it is more objective to measure their diameter. Later, bubbles increase their size and acquire a polyhedral shape due to the coarsening of the bubbles and the drainage of the liquid [
15]. Over the time, the bubbles became larger and exhibit polydispersity. This phenomenon may be due to the internal pressure of the bubbles of smaller diameter, which determines the flow of gas to the larger ones. This leads to a further reduction of the diameter of the smaller bubbles until they disappear and a consequent increase of the big ones [
15].
For all statistical analyses on each experimental condition (coffee variety and extraction conditions), the diameter of about 450 bubbles of coffee foam were measured (150 bubbles for three proofs). Statistically, the greater the number of bubbles, the lower the variability (many research papers report on very high bubbles number analyzed, also thousands). However, Barik et al. [
15], to determine the minimum number of bubbles required for the statistical analysis, studied
d32 for 50, 100, 150, 200, 250, 300, 350, and 400 bubbles obtained from the optical images. As reported by the author, the value of
d32 does not change for the bubbles number over 200. In this experiment, 150 bubbles sizes of the coffee foam from capsule extraction were considered for each proof; the statistical analysis carried out on the bubbles size was useful to study the foam systems of the espresso coffees obtained by capsules.
As shown in
Figure 2C, the distributions of the bubble diameters change from T0 to T10. In all the observed cases, the box plots at T0 show a rather contained interquartile gap as well as the width of the range. This rather limited variability of the diameter of the coffee bubbles at T0 indicates that the foam is rather uniform even at a microscopic level, with small bubbles and diameters similar to each other. However, if the mean diameters are observed (
Figure 2D), the NB sample presents bubbles, on average, very small (26 µm) at T0, statistically similar to the DGG sample (27 µm, T0); statistically higher values are observed in NG (36 µm, T0), while the highest value is present in the DGB sample (44 µm, T0). Therefore, in general, it can be stated that at T0, the bubbles of the coffee foam, at a microscopic level, are of uniform size in the different samples; moreover, the means of the diameters present significant differences for NG and DGB coffees that show larger bubbles. Furthermore, at T0, it is not possible to distinguish the effect of the coffee and the extraction pressure from the capsules.
After 10 min of observation (T10), the diameters of the bubbles change in size; large and small bubbles coexist in the foams, a phenomenon that leads to disproportionation of the bubbles and their instability. In all cases observed at T10 (
Figure 2C), the interquartile gap of the box plots is much larger than T0, and the mean values are far apart from the median ones. Furthermore, the ranks are very broad. This graphic/statistical elaboration highlights that after 10 min from its formation, the coffee foam collapses, showing its dynamic and transitory character. In the observed coffee foam, large and small bubbles coexist, which make it unstable; furthermore, following Oswald’s maturation [
15], the large bubbles grow at the expense of the small ones, and the system collapses, as observed also in
Figure 1A,B. Therefore, the values of the bubbles diameters change greatly, and thus, the behavior of the coffee foam differs in function of the applied pressure.
In
Figure 2C, the differences at T10 can be observed on the various samples; in particular, the coffee foam that has less variability is NG, while the other three samples have quite similar variability in terms of dispersibility of the size of the bubbles and therefore of the width of the box plots. The average diameters at T10 (
Figure 2D) show higher values than at T0. Except for NG, which does not show statistically significant differences in the two observation times (T0 36 µm and T10 38 µm), all the other samples have mean values at T10 that are statistically higher than T0. In particular, NB at T10 has a diameter of about 39 µm and DGG 59 µm, while DGB has the highest value (81 µm). In summary, it can be stated that the coffees extracted at 15 bars after 10 min of observation have a much higher mean value of the bubbles; this is an indication of disproportionation and greater instability of the foam, a result also confirmed by the observation of stability (
Figure 1A,B). It could be stated that the lower extraction pressure of the coffee solubilizes those molecules responsible for the stability of the foam bubbles, such as proteins and polysaccharides, to a lesser extent. Thus, even if at first, the beverage has a uniform foam with small bubbles, in a short time, they collapse, probably because they are not stabilized by the surfactant molecules, which in this case, would be in a lower concentration than the extraction that took place at 19 bars.
3.2. Sensory Analysis
The sensory parameters were chosen to cover many important quality criteria for a good cup of coffee. The first impression of a cup of espresso coffee is the crema, evaluated in terms of volume/firmness and texture. The next impression is the smell perceived directly on the nose and measured through the analysis of the head space of the coffee infusion [
23,
24,
25].
It is known that the volatile organic compounds in coffee, identified so far, are more than 1000 (VOC), but only 20 represent the relevant aromatic compounds that identify the smell of coffee [
26,
27,
28]. The other descriptors concern the flavor, which corresponds to a combination of the impression of post-nasal aroma and taste. The coffee body is another very important descriptor and is often due to the presence of total solids and occasionally also the lipids or fatty acid content. Both caffeine and chlorogenic acids are often related to bitter taste. The astringency may be due to the chlorogenic acids extracted from coffee [
29] although other molecules could also cause the astringent sensation [
30].
Table 1 shows sensory parameters of the six coffee samples (NG, NB, DGG, DGB, MG, and MB); as reported in the
Table 1, the two coffees (Brasile and Guatemala) differ significantly in a few parameters, such as “sour” and “body”. Additionally, in this case, the difference between perceptions is dictated by the three methods of coffee extraction; in fact, many of the significant differences in the sensory parameters are due to the extraction, with the espresso technique at 19 bars and 15 bars, and to the method with the moka. The values of the sensory parameters, shown in
Table 1, were plotted in two spider plots (the comments on the two graphs in
Figure 3 refer to the mean values of the parameters in
Table 1; Tukey HSD test was applied).
Figure 3A,B shows the sensory profiles of the two coffees, Guatemala (G) (
Figure 3A) and Brazil (B) (
Figure 3B), extracted at 15 bar (DG) and 19 bar (N) and with the moka. As shown in
Figure 3 A,B, the extraction of coffee with the moka generates a sensory profile with generally lower values of the descriptors than the other two extractions.
Considering the spider plot of Guatemala coffee (G), it can be observed that the descriptors vary in intensity depending on the pressure used (15 and 19 bars). In particular, the appearance of the texture of the crema, sour, sweetness, bitterness, balance, and body do not seem to be influenced by the extraction pressure in DGG and NG, in accordance with Gloess et al. (2013) [
31].
On the contrary, the thickness of the crema is higher in NG, while DGG has a more intense smell as well as the perception of astringency, aromatic intensity, and persistence.
In this context, the Guatemalan sample extracted with the moka method (MG) is an exception, which presents most of the different descriptors with respect to DGG and NG, and the values are clearly lower; in this last case, the extraction method is responsible for a lower number of compounds in the coffee [
31].
The sensory analysis of Brazil coffee, obtained with the three extractions (NB, DGB, and MB) also showed some differences (
Figure 3B). Analyzing the spider plot, the DGB and NB samples have statistically similar sensory profiles except for the “crema” and “odor” parameters. Moreover, in this case, the extraction of coffee with the moka (MB) had lower scores and a less expanded sensory profile than the other two. The only similar parameters in the three extractions are “balance” and “bitter”, which were not affected by the coffee preparation method used.
Therefore, as can be observed in
Figure 3A,B, the sensory profiles of the coffees obtained at 19 and 15 bars (N and DG, respectively) generally show high and more similar values.
Figure 3 A,B also shows that, if the sensory profiles of the coffees extracted with the capsules at different pressures are analyzed, they show differences related to the origin of the coffee, too (Guatemala and Brazil). In particular, the Guatemala variety changes the sensory profiles according to the extraction method in all three cases (DG, N, and M), while the Brazil coffee extracted at 15 and 19 bars maintains roughly the same sensory characteristics.
As reported by Andueza et al. (2007) [
3] and Adhikari et al. (2019) [
32], even with slightly varying the coffee parameters (extraction method, consumption temperature, etc.) there is a quantitative and qualitative variation of volatile and organoleptically active molecules, with consequences for the sensory characteristics of the beverage. In this research, the observed differences are probably due to the distinct extraction of some molecules responsible for the flavor and the characteristic volatility of some odorous molecules following the three methods applied. All of this leads to sensory profiles with distinct shapes (
Figure 3) and statistically significant differences in many of the sensory parameters studied (
Table 1).
The principal component analysis (PCA) summarizes and synthesizes the differences relating to the two selections of Arabica coffee extracted in the three ways and the sensory characteristics.
Figure 4 shows that more than 91.6% of the total variance is explained by the first (81.26%) and second (10.35%) components. The multivariate analysis was able to separate the espresso coffees extracted with the two types of pressure in the first positive PC 1 quadrant compared to the negative quadrant, in which the two coffees extracted with the moka are present. Furthermore, PC 2 highlights how NG is placed on the positive part of the quadrant together with the sensory characteristics of bitter, balance, and smell. Furthermore, NB, DGG, and DGB coffees are placed in the negative quadrant of PC 2 and are more associated with the sensory characteristics of sour, aromatic persistence, and sweetness, while the texture character of the crema is located between the positive and negative quadrant of PC 2.
In summary, the PCA (
Figure 4) shows how M Brasile and M Guatemala extracted with moka differ from those obtained by capsules at higher pressures (N and DG). With the extraction by moka, it is also possible to distinguish the two varieties (M Brazil, quadrant PC2 positive and M Guatemala, quadrant PC2 negative).
Furthermore, N Guatemala (PC2 positive quadrant) differs from N Brazil (PC2 negative quadrant), highlighting different sensory characteristics of the two varieties of Arabica with the same extraction method. N Guatemala stands out for its sensory parameters of aroma and balance, while N Brasile stands out for sourness, persistence of the aroma, for sweetness, and for its aromatic intensity.
On the contrary, the two varieties, Guatemala and Brazil, when extracted with the DG method (15 bar), are superimposed in the PC2 negative quadrant (between the negative and positive quadrants, PC2); they also share the parameters of texture, sweetness, body, astringency, and aroma intensity. In this last case, the DG extraction method (15 bar) makes the two coffees very similar to each other.