Thanks to the relatively low dilution ratio (1:10) and to the use of high purity reagents, it was possible having good results from a large set of analytes. Indeed, concentrations were higher than LOQ for all the analytes indicated in Table 2
. All data (ranges in Table 4
) resulted to be compatible with the known ranges of elements in wine [17
]. The precision was better than 5% for most elements and not lower than 20% even for ultra-trace elements such as heavy lanthanides.
In the following sections, we will discuss the possibility of using the elemental distribution, or part of it, to distinguish between Barbera d’Asti, Barbera d’Asti superiore
wines. It must be remembered that Barbera d’Asti
and Barbera d’Asti superiore
are indeed parts of the same designation, i.e., Barbera d’Asti
DOCG, therefore they are produced in the same geographic areas; in addition, the territory of Nizza
designation is totally contained inside that of Barbera d’Asti
. Therefore, differences among these wines may be expected, rather than from soil, because of oenological practices and, in particular, of ageing (see Table 1
Our previous work on the use of lanthanides distribution as traceability markers [14
] clearly indicated that the original fingerprinting given by soil is lost during the winemaking process. The same conclusion arose from other past works: Jakubowski et al. [18
] in 1999 questioned the fact that rare earth elements (REE) distribution could be considered as reliable fingerprint for the geographic provenance of a wine. Nicolini et al. [19
] and Castiñeira et al. [20
] both advised that fining treatment with bentonite could lead to fractionation of the original trace element distribution in white wines. Rossano et al. [21
] in their study on the influence of clarification, filtration, and storage on the concentration of REE in white wines, found that these processes provided a range of effects ranging from an overall increase to fractionation resulting in small increase of light REEs. As to red wines, Mihucz et al. [22
] and Tatár et al. [23
] found similar behaviours respectively in Romanian and Hungarian red wines.
The cited studies were mainly focused on the variation of absolute
concentrations of lanthanides, or on the variation of their distribution along the wine chain without any reference to soil. In the present study we wanted to deepen the relationship between soil and wine, by comparing their distributions after normalisation to Ce according to the formula [Lanthanide]Ce-normalised
. Normalisation allows a better comparison between samples (soil and wine) whose concentrations differ by 2–3 orders of magnitude. The lanthanides distributions of all our wine samples follow the Oddo-Harkins rule (Figure 2
a, Ce-normalised data for Nizza
wines, shown in logarithmic scale in order to highlight the differences on the heavy lanthanides that could not be properly appreciated under a linear scale). The behaviour of some lanthanides, however, is apparently unusual. In particular, the content of Nd, Dy, Er and Yb is higher than expected. This cannot be ascribed to isobaric interferences in the determination by ICP-MS: 144
Nd is isobaric with 144
Sm but its interference is automatically subtracted via software and the only known polyatomic interference is from 96
] which can be safely excluded being the level of Ru in our samples under LOD; 163
Dy has positive interference from 147
Sm accounts for only 15% of total Sm; 174
Yb has positive interference from 158
Gd accounts for 25% of total Gd) but the Gd/Yb ratio is ranging from 0.304 to 3.618, so no correlation seems to exist. The behaviour of 167
Er could be explained in terms of positive interference from 151
, as 151
Eu has, in turn, interference from 135
, but no correlation exists indeed between 167
Er and 151
, nor between 167
Er and 135
The behaviour of Eu is widely variable but this is due to the fact that both Eu isotopes, 151Eu and 153Eu, suffer from positive interference from Ba oxides (135Ba16O+ and 137Ba16O+ respectively); as this interference cannot be resolved with the instrument used in this study (a low resolution quadrupole mass spectrometer), the signal of Eu depends indeed on the content of Ba which is highly variable.
By contrast, the lanthanides distributions determined in the corresponding samples of soil, collected at every location of Nizza
producers (Figure 2
b), are highly homogeneous and closely follow the Oddo-Harkins rule with a general lowering trend of heavy lanthanides. This is the expected behaviour, considering the very small size of the production area of Nizza
To evaluate numerically the different behaviour of lanthanides in wines and soils, as far as Ce-normalised data are concerned, the average RSD (calculated on all lanthanides except Ce) was 55.2% in wines but only 10.0% in soil samples.
In the end, it must be accepted the fact that the winemaking processes had heavily influenced the lanthanides distribution, possibly as a consequence of the use of clarifying materials such as bentonite, as it was already cited in our previous work on Moscato d’Asti [25
]; bentonites are indeed used by nearly all the producers of Nizza
wine. According to these results, it is apparent that lanthanides cannot act as traceability markers as they are not representative of the original fingerprint, i.e., the distribution in soil. Not surprisingly, an attempt of distinguishing between Barbera d’Asti, Barbera d’Asti superiore
wines on the base of Ce-normalised data of lanthanides, using pattern recognition techniques, was unsuccessful (data not shown).
3.3. Other Trace- and Ultra-Trace Elements
Despite the unsuccessful attempt of using lanthanides to distinguish between Barbera d’Asti
, Barbera d’Asti superiore
wines, we wanted to explore the behaviour of the other trace- and ultra-trace elements. Indeed, many authentication studies on wines generically exploit the whole of trace elements rather than only lanthanides [9
]. Hopfer et al. [29
], as an example, were able to classify Californian wines according to their vineyard origin and their processing winery with respect of soil elemental content and viticultural practices.
It is well known that winemaking treatments can affect the mineral content of wine. Clarification with bentonites has strong effects in varying the original metal distribution [30
], as already pointed out with reference to lanthanides. Fermentation with different yeast strains markedly affects the content of alkaline, alkaline-earth and transition metals [31
]. In a recent study, Catarino et al. [32
] followed the trend of elements during winemaking, highlighting the role of the different steps in modifying the original elemental composition in soil.
Pohl reviewed the possible sources of metals [17
] in wine, indicating the primary source as the natural contribution from soil, regulated by the climatic condition during grapes growth; a secondary source in the external impurities coming from environment, outside and inside the cellar work; a third source in the oenological practices. Other sources of variation can be the following:
pH of soil;
type of rootstock;
vine growing system;
type of cultivar;
time of harvest (it can change from one zone to another and from a farm to another, even at short distances)
type of collection (manual and/or mechanical)
Transfer time (from vineyard to cellar) and temperature conditions
Different types of processing that the product can undergo depending on the objectives of the company grape pressing (time, duration, temperature)
use of yeasts (usually different from a farm to another)
duration of maceration and therefore of extraction from skins;
further processing steps (ageing in steel, barrique—type of wood and provenance—or bottles);
conservation conditions (temperature, relative humidity, etc.).
Another factor to be considered is of course the thermopluviometric trend, but in this work all wine samples were from the same vintage.
After evaluating the role of lanthanides, in our study we used all the elements determined by ICP-OES and ICP-MS to verify the possibility of discriminating between Barbera d’Asti, Barbera d’Asti superiore and Nizza wines. The dataset was composed of 57 variables (the elements determined) and 51 samples (wines of the three designations). Principal Components Analysis (PCA) was used; data were transformed into z-scores before analysis. However, no satisfactory results were obtained (data not shown).
Better results were obtained after dividing the samples into two groups, the first containing Barbera d’Asti
wines and the second containing Barbera d’Asti superiore
wines, i.e., the less aged wines against the more aged ones. A preliminary test by means of Analysis of Variance (ANOVA) indicated that Li, Rb, Sr, B, and Tl were the variables with the higher discriminating power within this scheme (alpha = 0.05). We then carried out PCA analysis using only these five variables: the results of PC1 vs. PC2 score plot (Figure 4
), accounting for 70.13% of total variance, suggests that a discrete discrimination is achievable between the younger Barbera d’Asti
(blue circles in figure) and the more aged Barbera d’Asti superiore
wines (red circles in figure).
The information arising from the loadings (black arrows in figure) indicates that Barbera d’Asti superiore and Nizza wines have a higher content of B, Li and Sr, while Barbera d’Asti wines have a higher content of Rb and Tl. Although alkaline and alkaline-earths elements are considered good indicators of geographical origin, in the present study their role must be considered in the light of oenological practises, being the origin of the samples nearly the same or at least too close to be discriminated (it must remembered that the samples of Barbera d’Asti and Barbera d’Asti superiore analysed in this study come from producers of Nizza). Three factors must be considered:
The alcoholic content: Catarino et al. [32
] showed that the concentration of Rb is inversely proportional to alcohol %, which is in agreement with our data if we consider that the average alcohol % is 14.2 for Barbera d’Asti
wines and 14.7 for Barbera d’Asti superiore
The widespread use of bentonites by producers of these wines: Catarino et al. [30
] showed that this treatment causes a strong fractionation of the original elemental distribution in musts; in particular Li, Sr and Tl were found to increase after bentonites treatment, while B and Rb decreased. However, bentonites are widely used in the production of all Barbera
The main difference between Barbera d’Asti
and Barbera d’Asti superiore
is ageing, which involves a more or less prolonged contact with barriques. Kaya et al. [33
] studied the effect of wood aging on the mineral composition of wine; Sr was found to increase significantly in wines aged in wood, while for Li, Rb, and Tl no significant effect was registered. These results partially confirm the differences found in our study with concern to Sr, which is higher in Barbera d’Asti superiore
than in Barbera d’Asti
In the end, it is possible that the elemental differences arisen in this study be a combination of all the factors above described. The role of Tl is hard to be explained, considering that this metal must be included in the group of contaminant elements of wine [34
]. Even the role of B is still to be accounted for.