The alcoholic beverage “pisco” is produced in Peru by distilling wine made from several varieties of grapes. Depending on the grape variety and the processes of fermentation and distillation used, both the chemical and the sensory characteristics can undergo changes that make it possible to obtain piscos of differing aromatic character. At the present time, the production conditions and equipment used are described in the Norma Técnica Peruana (NTP 211.001) [1]. According to the regulation governing pisco’s Denomination of Origin, water cannot be added to the wine or the liquor, nor is double distillation allowed. These two aspects distinguish pisco from other spirits produced in the Southern Cone. Following this regulation, the distillation of alcoholic must can be carried out at the end of fermentation, or with fermentation incomplete. In the latter case, the final product is called green must pisco. The producers who make these green must piscos depend on personal empirical knowledge acquired over time. Ordinarily, fermentation is stopped when a level of about 25 grams of sugar per litre is reached; this usually occurs between the fourth and the tenth day of fermentation.

Regarding the distillation process, the regulation stipulates that pisco production must be done exclusively by direct, discontinuous distillation, separating the head and tail so that only the middle part of the product is selected. The equipment permitted for this operation is the falca, the alembic, or the alembic with a wine heater. Both types of alembic should be made of copper or tin, whereas the falca is made of bricks and clay with the walls covered in a mix of lime and cement. It consists of a basin in which the most is heated over a wood fire. It is important to keep in mind that the form or design of the distillation equipment can also affect the end characteristics of the distillate. The falca, which is a much more rustic piece of equipment than the alembic, permits a lower level of condensation of alcoholic fumes, and more easily allows a greater amount of impurities to pass.

Lastly, it is worth noting that, in present-day Peru, the greater percentage of pisco is produced by artisan methods that follow traditional customs acquired by the producers, and a lower proportion is produced on an industrial scale, although this situation has been changing in recent years. The difference in production scale could perhaps be reflected in the aromatic composition of the end product.

Until now, there have been no studies of the influence of certain production processes on the aromatic composition of Peruvian piscos: namely, the point in time when distillation takes place, the distillation equipment, and the scale of production. Nevertheless, studies previously published by Cacho et al. [2,3] give detailed information about the aromatic profile of artisan piscos from the most representative varieties (the Quebranta as a non-aromatic pisco and the Italia as an aromatic pisco). The present work has used part of that information as a reference for the purpose of comparing the effect of the different production processes on the aromatic composition of these distillates. Other authors, such as Agosin et al. [4], Lillo et al. [5], Herraiz et al. [6] and Bordeu et al. [7] have focused their attention on the quantitative and sensory analysis of Chilean pisco. There are also many studies of the aromatic composition of wine distillates similar to pisco, like orujo, grappa, and calvados [8-14], but none allows for evaluation of the production process.

The principal objective of the present work is to test whether there are significant differences in the aromatic composition of the Quebranta and Italia varieties of pisco when comparing:

1) piscos obtained from completely fermented musts and incompletely fermented musts (green musts);

2) piscos produced by distillation in a falca and an alembic;

3) piscos produced on an artisan scale and an industrial scale.

The final part of the study is a sensory evaluation of the effect of the technological variations on the pisco’s final aroma.

Pisco samples

Fifty-one samples of pisco were analysed: 28 of the Italia variety and 23 of the Quebranta variety. Nine of the Quebranta piscos were produced from completely fermented musts using artisan methods and alembic distillation. The rest of the analysed Quebranta piscos were obtained by varying one of the three production stages. Specifically, six were obtained from incompletely fermented musts, eight by a-largescale (industrial) process, and three were distilled in a falca rather than an alembic. Of the Italia samples, nine were obtained under what we considered to be reference conditions (complete fermentation, artisan production, alembic distillation). Of the rest, seven came from incomplete fermentation of the must, five were obtained on an industrial scale, and seven were distilled in a falca. Table 1 provides information about all the groups of samples analysed.

Table 1: Groups of pisco samples analysed in this study according to their variety, fermentation process, distillation instrument and production scale.

Reagents and standards

Dichloromethane and methanol of LiChrosolv quality were supplied by Merck (Darmstadt, Germany), and absolute ethanol by Panreac (Barcelona, Spain), all of ARG quality. Pure water was obtained from a Milli-Q purification system (Millipore, Bedford, MA). Semi-automated solid-phase extraction (SPE) was carried out with a VAC ELUT 20 station supplied by Varian (Walnut Creek, CA). The LiChrolut EN resins and polypropylene cartridges were obtained from Merck (Darmstadt, Germany).

The chemical standards used for identifications were supplied by Aldrich (Steinheim, Germany), Fluka (Buchs, Switzerland), Poly- Science (Niles, USA), Lancaster (Strasbourg, France), and Alfa Aesar (Karlsruhe, Germany). An alkane solution (C8–C28), 20 mg L^-1 in dichloromethane, was employed to calculate the linear retention index (LRI) of each analyte.

Chemical quantitative analyses by direct injection of the distillate in a gas chromatograph (GC-FID)

This analysis was carried out following the procedure proposed by López-Vázquez et al. [12]. This method consists of direct injection of the sample in the chromatographic system after the addition of certain internal standards (4-methyl-2-pentanol and 4-decanol). In this way, 18 majority volatile compounds were able to be quantified in the pisco samples. The calibration method is described in detail in a previous paper [15].

Chemical quantitative analyses by solid phase extraction (SPE) followed by injection in a gas chromatograph (GC-Ion Trap-MS)

This analysis was carried out using the method proposed and validated by López et al. [16] with some modifications. This method consists of obtaining a representative extract of the sample via a solid phase extraction process (SPE), to which a known quantity of internal standards (4-hydroxy-4-methyl-2-pentanone and 2-octanol), is added, and afterward is analysed by GC-MS. This method allowed for the quantification of 45 volatile compounds, present in lower concentrations, which were not able to be quantified by the GC-FID method. Calibration information about all the quantified compounds is described in detail in two previous papers [15,16].

Sensory analysis

Sensory panel: The sensory panel was formed by twelve judges aged 23-40. All the judges had some previous experience in sensory analysis.

Triangle tests: For the purpose of comparing, from a sensory point of view, the different sample groups analysed in this study, and evaluating whether the chemical changes resulting from the distinct production processes affected the final aroma of the piscos, different triangle tests were carried out comparing representative samples from each group according to the Spanish Norm AENOR 87-006-92 [17]. For this purpose, eight representative samples were prepared (four of each variety), mixing equal portions of piscos produced under the same conditions.

For this test, the panellists were presented with three coded samples, two of which were identical, and the other, different. The panellists were asked to identify the odd sample of the three. This sensory test is employed frequently for detecting the existence of small differences amongst samples. Once the test was concluded, the results obtained were interpreted, for which a significance level of 95% was required in all cases.

Statistical analysis

For the data obtained from the chemical quantitative analysis, three different one-factor analysis of variance (ANOVA) tests were carried out to look for discriminant odorants. Also, Discriminant Analysis was carried out using SPSS software (version 15.0) from SPSS Inc. (Chicago).

Chemical differences

Tables 2 and 3 show the results obtained from quantitative analysis of 64 volatile compounds in 23 and 28 piscos of the Quebranta and Italia varieties, respectively, produced using differing procedures. In both tables the samples have been collected into four groups. The first was considered to be the reference group, containing the piscos resulting from the distillation of fresh must with complete fermentation, distilled in an alembic using artisan methods; these being the ones most often found on the market.

Table 2: Compounds analyzed by GC-FID and GC-MS in Quebranta pisco samples produced by complete and incomplete fermentation of wine, on an industrial or nonindustrial scale, and distilled in alembic and falca.

Table 3: Compounds analysed by GC-FID and GC-MS in Italia pisco samples produced by complete and incomplete fermentation of must, on an industrial or artisan scale, and distilled in alembic and falca.

The second group includes green must piscos (with incomplete fermentation), distilled in an alembic using artisan methods. Comparing this group with the reference group, evaluation was made of the effect on the volatile composition of these piscos obtained from most of complete fermentation or incomplete fermentation (green must).

The third group consists of piscos of complete fermentation, distilled in an alembic and produced on an industrial scale. Thus, this group served for evaluating the effect of different production scales.

Finally, the fourth group includes piscos of complete fermentation, distilled in a falca using artisan methods. Thus, the samples of this last group served for evaluating the influence of distillation in an alembic or a falca.

These two tables show the minimum, maximum, and average levels for each of the four sample groups. Different analyses of the variance of a single factor (ANOVA) were carried out in order to evaluate: 1) the “time point of distillation” factor, 2) the “production scale” factor, and 3) the “distillation equipment” factor. The compounds that showed significant differences (p<0.05) in these ANOVA analyses are shown with distinct letters (a,b,c).

Time point of must distillation: As can be seen in Table 2, few significant differences were found between the Quebranta pisco samples resulting from completely or incompletely fermented must (comparison of the first two columns of the table). In fact, only five of the 64 analysed compounds showed significant differences via ANOVA. These are acetaldehyde, butyl acetate, ethyl hexanoate, β-damascenone, and γ-nonalactone. In every case, the highest levels corresponded to the green must pisco samples. It is notable that both the ethyl hexanoate and the acetaldehyde were found in concentrations greater than the olfactory threshold, in both the green musts and the piscos with complete fermentation, as explained below, which leads us to think there is a possible sensory effect from both aromatic compounds in the characteristic aroma of the Quebranta variety. Furthermore, with most of the studied families (esters, acetates, terpenes, cinnamates, volatile phenols, and lactones), slightly higher levels were observed in the samples of green must pisco.

With respect to the Italia variety (Table 3), the differences that were found to relate to complete versus incomplete fermentation corresponded to six compounds. Nerol, ethyl cinnamate, farnesol, and acetovanillone (detected in low concentrations) were shown to be significantly higher in the samples that were produced from green must; while other compounds like 2,3-butanediol and phenyl acetaldehyde showed the lowest levels when fermentation was incomplete. It is worth noting that the majority of these discriminant compounds were found in low concentrations, much lower than their corresponding olfactory thresholds, as will be presented below, which could be an indicator of the low aromatic potential of these compounds on an individual level. Nevertheless, it is possible to state that some important terpenes, such as linalool, nerol, and geraniol, were at slightly higher levels in the green musts. Linalool was found in concentrations of up to 14.96 mg L^-1 in the green must samples (nearly double the maximum value found in the samples produced from completely fermented musts). Nevertheless, for this pisco variety, higher concentrations of the majority of analysed volatiles were not so clearly observed in the green musts.

Distillation equipment: Some of the compounds in the Quebranta variety, such as neryl acetate and 4-vinylphenol, were shown to be significantly higher in the piscos distilled in an alembic, whereas other compounds, such as butyl acetate, α-terpinolene, ethyl dihydrocinnamate, o-cresol, and δ-decalactone, had significantly greater values in samples distilled in a falca.

In general terms, the samples distilled in a falca were characterized by higher concentrations of terpenes, cinnamates, vanillines, and, to a lesser extent, of higher alcohols. Especially noteworthy was the presence of methyl vanillate and ethyl vanillate in concentrations reaching 121 μg L^-1 and 92 μg L^-1, respectively, in some samples.

The samples distilled in an alembic showed higher levels of the ester family and also of some volatile phenols, such as 4-vinylphenol, whose average values using an alembic were higher than those of the falca by a factor of eight.

With respect to the Italia variety, eight of the nine differentiating compounds showed significantly higher values in the samples distilled in a falca, of which three were terpenes (t-limonene oxide, c- and t-linalool oxide) and three were volatile phenols (guaiacol, o-cresol and m-cresol). On the other hand, the samples distilled in an alembic showed significantly higher values only of 2,3-butanediol. In general terms, once again there were slightly higher observed levels of analysed volatiles in the falca-distilled samples.

Production scale: In samples of the Quebranta variety of pisco, 16 of the 64 analysed compounds showed significant differences when a comparison of production scale, either industrial or artisan, was made (Table 2). Samples produced by artisan methods showed significantly higher contents of some alcohols, such as 2- and 3-methyl-1-butanol and β-phenylethanol, some volatile phenols (guaiacol, m-cresol and 4-vinylphenol), and some acetates and vanillins. Nevertheless, the samples from industrial-scale production were generally richer in nearly all the analysed compounds. Of these, it is especially noteworthy that the observed levels of acetaldehyde, phenyl acetaldehyde, of some esters, and of ethyl cinnamate were significantly higher than those found in the artisan piscos.

The samples of the Italia variety showed fewer important differences than those observed in the Quebranta piscos (as can be seen in Table 3). In fact, only three compounds allowed for a statistical differentiation of the two production processes. These were guaiacol and the cinnamates (ethyl cinnamate and ethyl dihydrocinnamate), which showed higher values in samples produced on an industrial scale. In general, it can be concluded that, for the Italia variety, the aromatic profiles observed in the two types of processes were very similar, according to the quantitative data. In fact, the compounds that showed significant differences were found in low concentrations, such that, on first impression, they do not appear sufficient to allow for sensory differentiation.

Graphic representation of quantitative results: The main conclusion from these quantitative results is that the variations attributed to the evaluated processes (the point in time when distillation takes place, the distillation equipment, and the scale of production) depend mostly on the grape variety, since the differentiating compounds are not the same in the two varieties, except for o-cresol and δ-decalactone, which showed higher levels in the falca-distilled samples in both the Quebranta and the Italia varieties; and ethyl cinnamate, which showed higher concentrations in samples produced on an industrial scale in both varieties.

In order to see graphically the differences amongst the samples produced by the distinct processes, a discriminant analysis was carried out on all the quantitative data, grouping the samples according to production process. Figure 1 shows a graphic representation of Quebranta variety samples, located in the space defined by the two obtained canonical functions, of which canonical function 1 explained 92 % of the variance and allowed for a clear separation of three sample groups. This graphic clearly shows that what is taken as the reference group (piscos with complete fermentation, alembic distillation, and artisan production) is well removed from the rest of the groups, all of which underwent some type of technical variation in the production process. Therefore, it is clear that each of the evaluated parameters causes differences in the volatile composition of the resulting piscos. Figure 1 allows for the conclusion that the greatest differences in the volatile composition of the Quebranta piscos were obtained when changing either the distillation equipment (falca/alembic) or the production scale (artisan/industrial), while the differences were less marked between green must and completely-fermented must samples.

Figure 1: Canonical discriminant analysis of quantitative data from Quebranta piscos.

In the same way, Figure 2 shows the distribution resulting from a discriminant analysis applied to samples of the Italia variety. Function 1, which was able to explain 84.5% of the total variance, showed in this case a more clear separation between the four sample groups. Here, the greatest distance appeared between the reference samples and those produced on an industrial scale.

Figure 2: Canonical discriminant analysis of quantitative data from Italia piscos.

Sensory analysis

Sensory contribution of individual aromatic compounds: Tables 4 and 5 show the thresholds of individual olfaction for 35 selected compounds, representing each of the principal chemical families which were determined in a previous work [18]. Using these thresholds, aroma values were estimated for each of the compounds in the Quebranta and Italia piscos, respectively.

Table 4: Individual odour thresholds (mg L^-1) and average odour activity values (OAV_av) of different compounds obtained in Quebranta pisco samples produced by different processes.

Table 5: Individual odour thresholds (mg L^-1) and average odour activity value (OAV_av) of different compounds obtained in Italia pisco samples produced by different process.

Based on these results, the compound with the highest aromatic relevance, showing the maximum OAV in both varieties, was ethyl isovalerate. Levels of this ethyl ester were on average 34 and 37 times higher than the threshold in the industrially-produced Quebranta samples and in the falca-distilled Italia samples, respectively. This fact would explain the greater aromatic intensity attributed to these Quebranta samples during sensory analysis, when compared with the artisan-produced samples. Nevertheless, this fact would not explain observations made with regard to the Italia variety, where no significant differences were seen when either a falca or an alembic were used for distillation (Table 6).

Table 6: Triangle test results.

Another distinctive compound of the Italia variety is linalool. Table 5 shows how the average concentrations are eight times greater than the olfactory threshold in samples produced from green must, when compared with those produced by complete fermentation. Another outstanding compound is 3-methyl-1-butanol, with median OAV of between 3 and 5 in both varieties, but with the highest obtained values in the falca-distilled Quebranta variety. Finally, ethyl hexanoate also stands out, showing OAV >1 in 48 of the 51 analysed samples, with median OAV of between 1.7 and 3.1 in both varieties, with the highest value being found in the industrial-scale Quebranta.

The OAV of the volatile phenols, lactones, some terpenes, acetates, alcohols, acetic acid, β-damascenone, farnesol, and phenyl acetaldehyde were less than 0.1 units in all the analysed samples, which suggests, at least on the surface, that these compounds do not individually have any great sensory impact.

Sensory differences (triangle tests): The aromatic changes introduced by the different production processes were evaluated by means of triangle tests on each variety, following the previouslydescribed procedure (2.5.2.). The results are shown in Table 6.

As can be seen, the samples produced from green must were significantly different, from a sensory point of view, from the samples produced using completely fermented must; this was true for both grape varieties. According to the panel of tasters, the Quebranta green must pisco presented a sweeter and more terpenic aroma, with greater aromatic intensity, which may be explained by the higher observed levels of terpenes and acetates. The Italia green must pisco, on the other hand, was defined as having a more floral and raisin aroma. This may be partially related to the higher (though not significantly higher) levels of compounds like β-damascenone, of acetates such as isoamyl acetate and β-phenylethyl acetate, linear esters, and terpenes like linalool, known to give sweet and floral aromas.

Regarding the samples distilled in a falca or an alembic, a sensory distinction between these two methods could be made only for the Quebranta variety: the falca-distilled Quebranta samples were said to have a more intense aroma, which some judges describes as an aroma of nuts. There were no significant sensory differences noted in the Italia variety samples, as related to the distillation equipment; this corroborates the few quantitative differences observed. Finally, the Italia variety was found to have more significant sensory differences (p<0.001), when industrial and artisan production scales were compared. The artisan production sample was described as being sweeter and with a greater aromatic intensity. This result was concordant with the distribution obtained in Figure 2, although the ANOVA did not show great quantitative differences. Nevertheless, it is possible that the higher level of some terpenes, such as linalool and gerianol, in the artisan samples contributed to this aromatic intensity.

In the case of the Quebranta variety, the industrial production samples showed a greater aromatic intensity, which could be attributable to the higher levels of esters and terpenes found in the samples produced on an industrial scale.

It is possible to deduce from the results obtained in this study that chemical differences do indeed result from modifications in the production processes of pisco. Green must piscos can be seen to have slightly higher levels of most analysed volatiles when compared with piscos made from completely fermented musts, especially in the case of Quebranta piscos. On the other hand, falca-distilled piscos show slightly higher levels of volatiles as compared to those distilled in an alembic; the production scale also affects the chemical profile of both varieties of pisco. In any case, when the individual behaviour of the compounds in both grape varieties is evaluated, distinct behaviours were observed.

Nearly all the studied technological variations had a sensory effect, with the exception of the change in distillation equipment in the Italia variety. Only a few compounds showed the same behaviour in the distillates of the studied grape varieties. These were o-cresol and δ-decalactone, which showed higher levels after falca distillation in both varieties, and ethyl cinnamate, which had significantly higher levels when made on an industrial scale in both the Quebranta and the Italia varieties.

Lastly, an evaluation was made of the individual aromatic effect of 35 odorants in these analysed pisco samples, in terms of their aroma values (OAV). Especially noteworthy is ethyl isovalerate, which reached aroma values of up to 34 aroma units in the industrial-scale Quebranta, while reaching only 20 aroma units in the artisan samples. In the Italia variety piscos, this ester reached its highest aroma value in the falca distillates (37 aroma units), while reaching only 14 aroma units in the alembic distillates. Other compounds that showed aroma values higher than 1 were: 2-methyl-1-butanol, 3-methyl-1-butanol, β-phenylethanol, and acetaldehyde in Quebranta piscos, and 2-methyl- 1-butanol, linalool, and acetaldehyde in the Italia variety. Therefore, all these aromatic compounds must be kept in view as having a possible effect on the overall aroma of these pisco varieties.

The authors gratefully thank the producers who have supplied the different Pisco samples and the Consejo Regulador del pisco in Peru. Liliana Moncayo thanks the Aragon Institute of Engineering Research (I3A) for a grant. We also thank the members of the sensory panel from the Laboratory for Flavour Analysis and Enology who participated in the sensory analysis.