Introduction

The high complexity of some matrices such as food, environmental, or biological samples could complicate their analytical determination. From an instrumental and data treatment point of view, chromatographic coelutions lead to mixed mass spectra that might difficult the identification of the compounds, especially when hard ionization techniques such as electron ionization (EI) are used. Moreover, depending on the ion source selected, ion suppression effects could become critical, thus hindering the detection of analytes. Thereby, powerful analytical techniques that ensure both a high chromatographic resolution and a selective and sensitive determination are required when dealing with the characterization of complex samples. In this way, comprehensive two-dimensional gas chromatography (GC × GC) coupled to mass spectrometry has become a powerful technique to enhance the targeted analysis of complex samples or to identify differences and quality biomarkers between different samples [1,2,3,4,5,6]. In GC × GC analysis, the transferring process from the first dimension (1D) to the second dimension (2D) should be fast and reliable. Among the modulators used, valve-based modulators are gaining more relevance since they simplify the methodology, helping on the automatization since cryogenic solvents (i.e., liquid N2) are no longer required and reducing the costs compared with the most common thermal modulators (cryogen-based longitudinally modulated cryogenic system and the 2- and 4-jet modulators) [7, 8]. These valve-based modulators are generally classified in forward (FFF) and reverse flow fill/flush (RFF) modulators. FFF modulators elute the 1D effluent accumulated in the collection channel in the same direction than the loading step, while, in RFF modulators, the 1D effluent is eluted in the opposite direction, thus helping the focusing of the analytes in the 2D column especially when analyzing samples containing a wide dynamic concentration range [9]. Although these flow modulators are able to provide a good peak capacity and sensitivity, one of the main issues regards with their compatibility with mass spectrometry (MS) due to the high 2D flow rates (20 mL min−1). Thus, flow splitting before the MS entrance using vacuum ionization techniques such as EI is generally required, reducing the sensitivity of the method [10].

Traditionally, GC has been coupled to MS by vacuum ionization techniques like EI, allowing the identification of the ionized compounds by MS libraries (i.e., NIST). However, as mentioned above, these hard ionization sources produce a heavy fragmentation that may hamper the identification of the analytes in complex sample matrices. In this sense, atmospheric pressure ionization (API) sources can offer great potential in GC-MS determinations since these soft ionization techniques usually preserve the molecular or quasi-molecular ion. Moreover, their flexibility for the coupling of GC with advanced high-resolution mass spectrometry (HRMS) systems, usually limited to liquid chromatography applications [11, 12], enhances the capabilities of these sources for non-targeted applications. Among API sources, plasma-based sources are gaining more attention because they have shown a great ionization efficiency for a wider range of compounds than other API techniques [11]. Although plasma-based sources have been widely used for ambient ionization mass spectrometry [13], their application using GC-HRMS is still quite limited. Recently, our research group has developed a tube plasma ionization (TPI) source for GC-qTOF-MS coupling which allows the determination of a wide range of compounds [14]. Moreover, the operation of this source at atmospheric pressure conditions could help to improve the performance of flow modulators for GC × GC analysis since the 2D flowrate is negligible compared with the high flows required in the source, thus avoiding the split of the eluate before the MS determination. Thereby, this analytical platform could be a good strategy to simplify the analysis of complex samples.

In this work, the characterization of the aroma profile of commercial vermouth has been comprehensively evaluated by GC × GC-qTOF-MS using an RFF modulator. Due to the complex volatile profile of vermouth, advanced extraction and analytical tools are required. Thus, both the extraction process and GC × GC separation, including flow modulator parameters, have been carefully optimized. Additionally, the GC × GC-HRMS coupling has been done by using the novel TPI source to achieve a selective and wide-scope ionization and obtain the aroma profile identification of vermouth.

Materials and methods

Reagents and standards

Dichloromethane for HPLC grade (99.8%) was purchased from Thermo Fisher Scientific (Dreieich, Germany), while MilliQ water was obtained from a Sartorius ultrapure water system (Goettingen, Germany) (resistivity 18.2 M Ω cm−1). 1,4-Dichlorobenzene (≥ 99.9%), used as recovery standard, and Supelco 37-component FAMES mix (200–600 mg L−1), used as internal standard, were acquired from Merck (Darmstadt, Germany). The working solutions were prepared by adequate dilution of the recovery and internal standards in dichloromethane and stored at  − 20 °C before the analyses. Sodium chloride for analysis (100%) was obtained from Bernd Kraft (Duisburg, Germany). Moreover, helium Alphagaz 1 (≥ 99.999%), used as the carrier gas, argon Arcal Prime Smartop (≥ 99.999%), employed as the discharge gas, and nitrogen Alphagaz 1 (≥ 99.999%), used as the auxiliary gas in the TPI source, were supplied by Air Liquide (Oberhausen, Germany).

Sample and sample preparation

Vermouth (18% v/v ethanol) was bought in a local store in Essen (Germany) in 2022. The sample was stored at  − 20 °C before extraction and analysis to avoid any loss of the most volatile analytes. A headspace solid-phase microextraction (HS-SPME) was considered for the recovery of VOCs from the vermouth. The HS-SPME was carried out following the procedure described by Zhu et al. [15] with slight modifications. Briefly, 2 mL of vermouth were diluted with 8 mL of MilliQ. Then, the sample was spiked with 20 µL of an appropriate concentration of the internal standards (1,4-dichlorobenzene and FAMES mix) to reach a concentration in the final extract of 5 µg L−1 and 4–12 µg L−1, respectively. Then, 2.5 g of sodium chloride were added, and the sample was equilibrated for 30 min at 45 °C. Then, the sample was extracted using a divinylbenzene/carboxen/polydimethylsiloxane (DVB/CAR/PDMS) fiber (1 cm, 50/30 μm; Supelco, Bellefonte, PA, USA) for 30 min at the same temperature with constant stirring. Finally, the fiber was injected into the GC injector for thermal desorption for 3 min, and it was left in the injector for another 30 min.

Instrumentation

The analyses of the vermouth were conducted on an 8890 GC gas chromatograph, equipped with an 7693A ALS autosampler and a reversed flow modulator (Agilent Technologies, Santa Clara, CA, USA). The GC was coupled to a flame ionization detector (FID, Agilent Technologies) for both sample extraction and chromatographic separation optimizations. Additionally, for the identification of the aroma profile, the gas chromatograph was coupled to a 6546 LC/Q-TOF high-resolution mass spectrometer (Agilent Technologies) interfaced with the homemade TPI source described elsewhere [14]. Briefly, the TPI source is based on an inverse low-temperature plasma configuration. A stainless-steel needle electrode is positioned inside a quartz tube, also known as dielectric, where the argon discharge gas is passing through. When fast changes of high voltage are applied to the pin electrode, the dielectric is polarized, and the gas ignites by means of a dielectric barrier discharge (DBD). The GC × GC separation was carried out using a DB-5MS (30 m × 0.25 mm ID; 0.25 µm, Agilent Technologies) as 1D column, SLB-IL60 (5 m × 0.25 mm ID; 0.20 µm, Merck Supelco, Darmstadt, Germany), as 2D column, and a deactivated silica capillary (4.2 m × 0.1 mm ID, Agilent Technologies) as restrictor. The He carrier gas flow rates were set at 0.5 mL min−1 for the 1D and 20 mL min−1 for the 2D, respectively. The temperature program was fixed as follows: from 40 °C with 5 °C min−1 to 300 °C (hold for 10 min) (run time: 62 min). The injector temperature was set at 250 °C (splitless, 3 min). The modulation period and the injection time were set at 2.5 s and 0.12 s, respectively. Moreover, the transfer line temperature was fixed at 280 °C.

TPI was operated using argon as discharge gas at 200 mL min−1. The plasma working conditions consisted of a voltage of 2.5 kV, a frequency of 12.5 kHz, and a pulse width of 2 µs. Moreover, an N2 auxiliary gas at 400 mL min−1 was used to guide the ions through the plasma region and toward the MS inlet [16]. The dry gas flow rate and temperature were set at 13 L min−1 and 250 °C to avoid any condensation issues in the source, while the spray shield was not used for GC × GC-HRMS analyses. The fragmentor, skimmer, and octapole 1 RF Vpp voltages were set at 125, 65, and 750 V, respectively. The MS acquisition was done in full-scan data-dependent acquisition in both positive and negative ion mode. Both full-scan and MS/MS acquisitions were performed from 20 to 750 m/z using an acquisition rate 20 and 15 spectra s−1, respectively. Regarding the MS/MS acquisition, a maximum of 2 precursors were chosen per cycle (precursor threshold: 10,000 counts), and no isotope model was applied to allow the potential fragmentation of both [M]+• and [M+H]+ ions. The collision energy was set according to a linear curve in a mass-dependent manner ranging from 11 eV at m/z 20 to 47.5 eV at m/z 750. The qTOF was calibrated daily to ensure mass accuracy with the ESI-L Low concentration tunning mix solution (1:10 v/v) using the Agilent Dual Jet Stream electrospray source from Agilent Technologies. The GC × GC was controlled using the OpenLab software, while the qTOF control, data acquisition, and processing were done using the Agilent MassHunter Workstation 10.0 software.

Results and discussion

Vermouth consists of the maceration of dried herbs, barks, seeds, and leaves from aromatic and bitter herbs in white wine and neutral spirits for several weeks. This ethanolic maceration gives rise to the extraction of volatile and polar compounds present in the herbs that increase the already rich wine chemical profile in relevant bioactive compounds [17, 18]. Although wine is well characterized, there is very few information about the chemical composition of vermouths, and more studies are required to evaluate the properties of these beverages [19]. Indeed, the volatile profile of wine-related beverages defines the quality and the sensations produced during its consumption [20]. Thus, here we propose the use of GC × GC-TPI-qTOF analytical platform to help on the elucidation of the volatile profile of a vermouth sample.

Method set-up optimization

The extraction of volatile organic compounds (VOCs) is a critical and challenging step considering the large variety of different volatile molecules that form wine and vermouth profiles as well as the low concentrations in which most of these compounds are usually found. Liquid-liquid extraction (LLE) and solid-phase extraction (SPE) have commonly been used for the recovery of the VOCs from wine-related matrices [21]. However, in recent years, the use of headspace solid-phase microextraction (HS-SPME) has been increasing since it is cheaper, faster, it requires less sample manipulation than LLE and SPE, and it does not use organic solvents, which follows the principles of green chemistry [22]. Among the fibers used, DVB/CAR/PDMS fiber which contains three adsorbents is generally selected, extending the range of VOCs extracted from food items [21, 23, 24]. In this work, the extraction of VOCs from the vermouth was adapted from Zhu et al. [15] as mentioned above. To ensure a constant recovery of the analytes from vermouth, the recovery yield determined as the ratio between the response of internal standard spiked at 1 mg L−1 in a blank and a vermouth sample was established as a quality control (n = 3). 1,4-Dichlorobenzene was selected since dichlorobenzenes has been widely used as internal standard for aroma analysis [15, 25, 26]. Under these conditions, the recovery was 98 ± 4% which ensures the high quality of the results. Additionally, as can be seen in the 1D chromatogram obtained by GC-FID (Fig. 1), the VOC profile of the vermouth is very complex, thus requiring the use of more powerful separation techniques such as GC × GC.

Fig. 1
figure 1

1D GC-FID chromatogram of vermouth using a DB-5MS (30 m × 0.25 mm ID; 0.25 µm) after HS-SPME (DVB/CAR/PDMS) extraction

Full size image

For the GC × GC separation, a non-polar DB-5MS (poly(5% phenyl/95% dimethyl arylene siloxane), 30 m × 0.25 mm ID; 0.25 µm) was selected in the 1D, while a highly polar ionic liquid (IL) phase consisting of the SLB-IL60 (1,12-di(tripropylphosphonium)dodecane bis(trifluoromethylsulfonyl)imide; 5 m × 0.25 mm ID; 0.2 µm) was chosen as 2D column. The use of an ionic liquid (IL) stationary phase in the 2D provides a low separation correlation with the selected 1D; at the same time, it has a higher thermal stability (up to 300 °C) compared with other polar stationary phases such as polyethylene glycol-based columns (up to 260/270 °C) [27, 28]. The separation of the VOC profile was achieved using common 1D (0.5 mL min−1) and 2D (20 mL min−1) He flow rates for reverse fill/flush modulators, whereas the temperature program was set as follows: 40 to 300 °C (held 10 min) at 5 °C min−1. One of the critical points when working with valve-based modulators is the optimization of the modulation parameters.

Regarding the injection time, 0.12 s were enough to achieve good and narrow modulated peak shapes, whereas to maximize the separation and avoid wrap-around effects, two modulation times (2.5 and 3 s) were considered. The evaluation of the effect of these two modulation periods was carried out by comparing the practical peak capacity (2Dnc,practical) [29]. As can be seen in Table 1, the 2Dnc,practical calculated for the analysis done using 2.5 s modulation time was 1.7 times higher (2Dnc,practical = 7006) than the one using 3 s (2Dnc,corrected = 4140). This fact is related to the peak capacity reduction produced when higher modulation times are used. In this case, an undersampling effect occurs, and the obtained separation of peaks in the 1D can be lost during the modulation process giving rise to broader non-separated peaks. As can be seen in Table 1, while the peak capacity of the 2D (2nc) for both modulation times was very similar (24 and 20 for the analysis carried out at 2.5 and 3 s modulation time, respectively), the peak capacity of the 1D (1nc) was much lower when a higher modulation period was used (1nc = 231) in comparison with the shorter modulation time (1nc = 374). Thereby, the modulation time was set at 2.5 s. Thus, the GC × GC setup optimized in this work provided very high peak capacity values and high orthogonality for the separation of the vermouth aromatic profile.

Table 1 Peak capacity achieved using both modulation times
Full size table

GC × GC-HRMS coupling via TPI for the analysis of vermouth

One of the main gaps of flow modulators has been their coupling with MS systems. The high flow rates required in the 2D (ca. 15–30 mL min−1) are not compatible with those MS instruments working with vacuum ionization techniques such as electron ionization [30]. To sort out this problem, Tranchida et al. [10] proposed the use of a longer external accumulation loop for collecting the 1D eluate. This homemade flow modulator leads to a reduction of the 2D flow rate up to 6–8 mL min−1 and an improvement of the peak shapes. However, when using current available flow modulators, a split of the flow is required before the connection of the GC × GC with the mass spectrometer, leading to a loss of sensitivity of the whole setup. In contrast, when working with API techniques such as TPI, this 2D flow rate is negligible compared with other gases used in the source such as the N2 auxiliary gas and dry gas (ca. 0.2–13 L min−1) [31]. Therefore, the split is no longer required, which allows to detect all the eluate coming from the GC, and consequently, it does not reduce the sensitivity of the method.

Under these setup conditions, Fig. 2 shows the GC × GC-HRMS chromatograms using the TPI source in positive and negative ion modes. In these results, it can be observed that most of the compounds were only ionized in positive TPI mode, which shows a more complex profile. However, although in this case the positive mode provided better ionization efficiency, the possibility to switch the polarity in API sources provides additional information and advantages to carrying out the identification of VOCs, especially those compounds that are ionized in both ionization modes.

Fig. 2
figure 2

GC × GC-HRMS chromatogram of vermouth (1:4 v/v diluted with H2O) using a DB-5MS (30 m × 0.25 mm ID, 0.25 µm) as 1D column and SLB-IL60 (5 m × 0.25 mm ID, 0.20 µm) as 2D column (modulation period: 2.5 s; injection time: 0.12 s) in a positive and b negative TPI mode

Full size image

Regarding the separation and the peak shape, it was observed that some of the peaks showed a significant peak tailing in the 2D. This effect could be related to the high concentration of some of the compounds present in the sample that are transferred to the 2D short column producing an overloading of the column. At this point, different dilutions of the sample were injected using the optimized method as can be seen in the GC × GC chromatograms of the vermouth sample with different dilution factors (see electronic supplementary material (ESM), Fig. S1). The higher the dilution was, the less peak tailing was observed in the chromatogram.

However, injecting the sample in a low concentration produced a loss of sensitivity of the less concentrated compounds due to the large range of metabolite concentrations present in the sample. Figure S2 shows the loss of sensitivity of one of the low concentrated compounds after the 1:100 dilution. This peak presents a good peak shape tailing and adequate S/N intensity when the concentrated sample is injected. However, if a dilution is done in the sample, the signal is close to the noise with the risk of losing the information of this compound during the filtering of the data treatment. Therefore, the analysis of the sample diluted in a 1:4 v/v ratio with water was used for the chemical identification of the VOCs in vermouth. In addition, modulated peaks showed very narrow peaks widths (ca. 0.36 s) which require very fast mass analyzers to ensure a good quality of the acquired data. Generally, it is assumed that a well-defined 1D GC peak may have at least 8–10 scans for quantification purposes, although this number could be slightly lower for qualitative analysis. To fulfill this criterion, the scan rate was set at 20 spectra s−1 for the full-scan acquisition. As can be seen in Fig. 3 for an average intense peak eluted at the middle of the temperature program, there are around 7 scans per modulated peak. Considering that the peaks are modulated at least 3–4 times, around 20–30 spectra per 1D peak can be expected, which is high enough to carry out the identification of VOCs in vermouth. Additionally, the outstanding detection capability that TPI has shown for a wide range of analytes [14] may ensure the detection of flavoring compounds even at low concentration ranges. The repeatability of the HS-SPME GC × GC-TPI-qTOF-MS methodology has been evaluated by measuring the area of the internal standard (1,4-dichlorobenzene) in the sample (n = 3). The relative standard deviation (RSD) was 8.2% which shows the good performance of the determination.

Fig. 3
figure 3

Peak shape of an average modulated peak (m/z 155.0710) achieved by GC × GC-TPI-HRMS (qTOF)

Full size image

Thus, this analytical platform provides a clear separation and selective and sensitive detection of the VOC profile of vermouth in 60 min, which can be very useful not only for characterization purposes but also to develop high-throughput authentication approaches.

Identification of VOCs by GC × GC-TPI-MS/HRMS

The developed methodology was applied to the VOC profile characterization of the vermouth sample. The diluted sample (1:4 v/v) was extracted by HS-SPME (DVB/CAR/PDMS) and analyzed by GC × GC-qTOF-MS (n = 3) in both, positive and negative TPI modes. The confidence of the identifications was established according to the criteria reported by Schymanski et al. [32]. Although structure assignments based on the monoisotopic mass, the isotopic cluster match, and the ring double bond equivalent (RDBE) enhance the confidence level, the assignment of probable structures (level 2) requires literature or MS library match. However, this is one of the main gaps of GC-API-MS methodologies since the number of MS libraries available is quite low [11, 33], or in the case of new API sources such as TPI, they do not even exist. For that reason, the establishment of tentative candidates (level 3) was achieved by the comparison of experimental MS/HRMS spectra with those predicted by in silico methods [33, 34] using MetFrag [35]. To propose one structure as a tentative candidate, a minimum MS/HRMS match score of 80% was considered. Moreover, to ensure the correct assignment, the mass accuracy was fixed at 5 ppm for precursor molecules, while for product ions, it was set at 5 and 10 ppm for m/z higher and lower than 100 Da, respectively. In addition, the linear retention indexes (LRI) were determined and compared with those proposed in the literature and mass spectral libraries (i.e., NIST) by means of the absolute difference (ΔLRItheo) to increase the confidence in the identification. The proposed annotations were also filtered considering their relevance for food and aroma analysis using the FoodDB database [36]. Following all these conditions, compounds tentatively identified in the vermouth sample, corresponding with the peaks clearly identified in the 2D plot, are shown in Table 2. The most abundant classes of compounds identified were monoterpenes, terpenoids, and sesquiterpenoids, carboxylic acids and carboxylate esters, alkylbenzenes, phenols and derivatives, and aromatic aldehydes. Most of these compounds are flavor agents which might be strongly related to the organoleptic properties of the vermouth. All the compounds detected in positive mode presented a very clear [M+H]+ ion as the base peak of the mass spectra, as shown in Fig. 4 (compound identified as (4E,6E)-allocimene), except for the compound piperonal, that showed abundant [M+H]+ and the [M]+•. As can be observed, the use of TPI as ion source significantly decreased the in-source fragmentation, thus improving the selectivity of the methodology compared with those approaches based on the EI source (Fig. 4b). Moreover, the possibility to carry out tandem mass spectrometry experiments (MS/HRMS) using the GC × GC-TPI-qTOF-MS approach provided structural information that helps to increase the confidence on the annotation (i.e., (4E,6E)-allocimene MetFrag score: 84.3%). Although most of the analytes were only identified in positive TPI mode by means of the protonated molecule, some compounds could also be determined in negative ion mode. For instance, carvacrol led to the deprotonated molecule while Edulan I showed the [M-CH3]‒ ion as the base peak of the mass spectra. All this information, together with the well-defined retention times in both dimensions and the LRI, may increase the confidence level in the identification of these compounds.

GC × GC-TPI-qTOF-MS significantly increases both the sensitivity and the selectivity of the detection of aroma compounds. However, especially for some minor components of vermouth, the intensity could be still too low to acquire a good quality MS/HRMS spectrum. Even though TPI generally preserves the molecular/quasi-molecular ion, in some cases, in-source collision-induced dissociation (CID) fragmentation could be achieved (Fig. S3), depending on the lability of the compound [14]. When this happens and no or bad quality MS/HRMS spectrum was acquired, this fragmentation can be used to carry out in silico fragmentation comparison.

Table 2 VOCs tentatively identified in the vermouth sample
Full size table
Fig. 4
figure 4

Mass spectra of (4E,6E)-allocimene obtained by a TPI ( +) in full scan, b EI (obtained from the NIST library), and c MS/HRMS experiments after TPI ( +)

Full size image

In these cases, a minimum MS/HRMS match of 90% was required to ensure the correct annotation. Using this strategy, the confidence level on the identification of cis-chrysanthenyl propionate, methyl eugenol, ginsenol, geranyl acetone, epicubebol, α-irone, methyl dodecanoate, eugenyl formate, and β-Copaen-4α-ol could be increased without having a high-quality MS/HRMS acquisition.

The proposed aroma compounds have been related to different trends considering their main odor/tasting descriptors (see ESM, Table S1). On one hand, monoterpenes, terpenoids, and sesquiterpenoids are strongly related to floral and fruity aroma descriptors. Regarding aryl-aldehydes, those found in the vermouth sample (furfural, 5-methylfurfural, benzaldehyde, and heptadienal) are the main responsible for the almond odor. These compounds are present in natural sources like coffee, cacao, nuts, or some berries; however, furfural and methyl furfural can also be formed during the aging of the vermouth by Maillard and/or caramel reactions [37]. Alkylbenzenes and phenols showed different trends from sweet and citrus aroma compounds (i.e., styrene, 3-methyl-1-phenyl-3-pentanol, cymene, methyl eugenol, and vanillin), to spicy (i.e., eugenol and 4-vinylguaiacol), or even medicinal odor (i.e., chavicol and cresol). On the other hand, carboxylate esters are generally related to fruity aromas such as ethyl-2-methylpentanoate, so-called manzanate, diethyl malate, methyl pentanoate, pentyl acetate, ethyl furoate, ethyl levulinate, and ethyl sorbate, among others. Other compounds that may contribute to the fruity aroma are alcohol, esters (citronellyl aceate), and heterocyclic compounds like fructone, nonalactone, or piperonal. Thus, the different kinds of herbs, barks, seeds, and leaves used for the maceration will lead to a unique VOC profile of the vermouth analyzed. Therefore, the monitoring of these profiles using advanced analytical platforms such as GC × GC-TPI-qTOF-MS could help to identify potential biomarkers to differentiate between types of vermouths or even deal with potential authentication issues.

Conclusions

The proposed GC × GC-TPI-qTOF-MS methodology has shown a great potential to the analysis of complex samples such as vermouth. The use of a reverse fill/flush modulator allows an easier automatization of the GC × GC separation at the same time; it keeps a good peak shape for the modulated peaks. The column combination consisted of DB-5MS and SLB-IL60 in first and second dimension, respectively. The use of this ionic liquid column in the 2D provided a high orthogonality at the same time; it helps to reach higher temperatures (up to 300 °C) compared to other polar stationary phases such as polyethylene glycol-based columns (ca. up to 260/270 °C). Moreover, the combination of these powerful separation technique with an API source such as TPI helps to overcome the loss of sensitivity observed in GC × GC-EI-MS systems using flow modulators since there is no need to split the flow prior to the analysis with atmospheric pressure mass spectrometers. On the other hand, the use of API sources, such as TPI, significantly simplifies the data analysis of this complex sample since they largely preserve the molecular or quasimolecular ion, thus avoiding complex mixed MS spectra for coeluting compounds that could hamper the correct identification. Therefore, the combination of the high separation power of GC × GC, with the soft TPI ionization and the possibility to carry out MS/HRMS experiments with a high mass accuracy, provided a powerful, high-throughput, and reliable methodology to achieve the characterization of complex samples such as VOC profile of vermouth with a high level of confidence. Using this method, the different classes of compounds were tentatively identified in the sample, including monoterpenes, terpenoids, sesquiterpenoids and carboxylic acid, and carboxylate esters as the most abundant class, followed by alkylbenzenes and phenols and aryl-aldehydes. This aroma fingerprinting might be useful not only to characterize different vermouths but also to control the quality of the products as well as to deal with potential authentication issues.