Thermo-Oxidative Decomposition of Lovage (Levisticum officinale) and Davana (Artemisia pallens) Essential Oils under Simulated Tobacco Heating Product Conditions

Two flavour samples, lovage root oil (Levisticum officinale) and davana oil (Artemisia pallens) were obtained from Hertz Flavors GmbH & Co. KG (Reinbek, Germany). The samples were stored in amber essential oil dropper bottles at 4 °C in the refrigerator prior to analysis. All experiments were performed within a 10-day time window after first breaking the seal on the bottles, to minimize oxidation during refrigerated storage.

The gas chromatography/mass spectrometry (GC/MS) experiments were performed using an Agilent 6890 A/5973 GC/MS (Agilent Technologies, Santa Clara, CA, USA). The essential oil samples were dissolved in dichloro-methane to obtain 30 mg/mL concentration on the day of the measurements. Aliquots of 1 μL solution were injected to the GC/MS operating with helium carrier gas with a split ratio of 100:1.

The separation of the compounds was carried out on a DB-1701 capillary column (30 m × 0.25 mm, 0.25 μm film thickness). The GC oven was programmed to hold at 40 °C for 7 min before increase to 280 °C at a rate of 10 °C/min. The mass spectrometer was operated in electron impact ionization mode at 70 eV in a range of m/z 29–400. The identification of the compounds was performed using NIST 2011 and Wiley 2009 mass spectral libraries. For the correct identification of the isomers, literature data (4, 12–13, 19) were also taken into account. Five parallel experiments were carried out for both samples.

The pyrolysis experiments were carried out using a Pyroprobe 2000 (CDS Analytical, Oxford, PA, USA) coupled on-line to the Agilent 6890/5973 GC/MS. The pyrolyser is equipped with a platinum coil and a quartz tube. Aliquot of 0.05 μL essential oil sample was dispensed on a piece of quartz wool placed in the middle of the quartz tube. The Pyroprobe was immediately inserted into the pyrolysis chamber preheated to 250 °C. The sample was then heated to 300 °C (calibrated temperature) and kept there for 5 min in 9% oxygen and 91% nitrogen atmosphere using 276 mL/min flow rate. The temperature was chosen as 300 °C in order to mimic the heating conditions for the low-temperature tobacco heating products (30). The gas mixture purged the volatile molecules to the DB-1701 capillary column and then the carrier gas was switched to helium. A solvent delay of 7 min was applied to protect the mass spectrometer from oxygen. The parameters of GC/MS analysis were the same as for the analysis of the essential oil solutions. Five parallel runs were performed on both samples.

Figure 1(a) presents the chromatogram of the lovage root oil solution. The identification of the compounds, the relative intensities (area %) and the standard deviations calculated from five parallel experiments are listed in Table 1. Compounds having relative intensities higher than 0.1% were considered for evaluation. The quantification of the oil components by using reference compounds was not the purpose of this study. Therefore, the relative intensity data calculated from the integrated total ion current values were used for evaluating the composition of the essential oils. The area percent-values of the chromatograms provide a good estimate of the changes in the compositions during the thermo-oxidative decomposition.

Figure 1

The main component of lovage root oil was (Z)-ligustilide [50], which had 58% relative intensity calculated as area-percentage. (E)-Ligustilide [56] represented only 4%, while the relative intensity of 2-pentylcyclohexa-1,3-diene [10] and (Z)-3-butylidenephthalide [45] amounted to 6 and 4%, respectively. Eleven other compounds had relative intensity between 1 and 2%, altogether representing 15% of the oil. These components belonged to various types of compounds: monoterpenes, sesquiterpenes, alkylbenzenes, isobenzofuranone derivatives, and long-chain alcohol and ester. All the other compounds were minor components of relative concentrations between 0.1 and 1%, altogether amounting to 13%.

The components of lovage oil have been categorized into classes, as shown in Table 2. It can be established that isobenzofuranones were the most significant constituents of lovage oil including alkyl- and alkenyl-dihydroisobenzofuranones, tetrahydroisobenzofuranones, and phthalides, altogether representing a relative intensity of 74%. Terpenoids constituted 9.5% of the oil; slightly more than half of them were monoterpenoids, with the rest being sesquiterpenoids. Hydrocarbon terpenes dominated among the terpenoids, oxygen-containing terpene alcohols and acetate constituted only about 11 and 24% of monoterpenoids and sesquiterpenoids, respectively. Cyclohexadiene and benzene with aliphatic side chains represented another important group of hydrocarbon components constituting about 7% of the oil. Lovage oil contained long-chain aliphatic compounds as well, such as fatty acid, fatty acid esters, and alcohol.

During oxidative pyrolysis at 300 °C, the relative composition of lovage oil altered, as the pyrolysis-gas chromatogram (pyrogram) illustrates in Figure 1(b). The differences in the relative intensity data of the as-received and the pyrolysed oil are shown in Table 1. It should be noted that the evaporation of the compounds played a considerable role due to the volatility of the oil components, the moderate pyrolysis temperature, and the high gas flow rate. Nevertheless, the relative intensities of numerous compounds were changed and a few new products were formed during pyrolysis. Changes in composition were assessed for statistical significance using two tailed, two-sample t-tests with an alpha risk of 0.05. For instances, where a compound was absent in either sample group, significance was evaluated using a one-sample t-test with an alpha risk of 0.05.

The significance of the changes in the area-percent values of the components is listed in Table 1. The relative intensity of the main component, (Z)-ligustilide [50] reduced from 58.1 to 51.6% (0.01 ≥ p > 0.001). The levels of (Z)-3-butyl-idenephthalide [45] and other alkyl- and alkylidenephthalides were enhanced during pyrolysis. Scheme 1 presents the reaction routes proposed for the conversion of (Z)-ligustilide [50] during low-temperature oxidative pyrolysis. The cyclohexadiene ring of (Z)-ligustilide was transformed (aromatised) via dehydrogenation leading to the formation of (Z)-3-butylidenephthalide [45]. Similarly to our previous studies (30, 31, 32), it is likely that oxygen radicals initiated the cleavage of the CSH bonds, resulting in the release of water molecules, because the pyrolysis temperature was considered low for dehydrogenation with the formation of hydrogen molecules. Our previous study on flavour ingredients (30) demonstrated that the dehydrogenation reactions were more pronounced in the oxidative than in the inert atmosphere. (Z)-3-Butylidenephthalide [45] could be further oxidized at the alkylidene group forming phthalic anhydride [29]. The scission of the furanone ring of (Z)-ligustilide [50] also took place as the increased yields of valerophenone [23] and 1-phenyl-1-pentene [14] indicated. The elimination of carbon dioxide from the furanone ring followed by hydrogen transfer could result in the formation of 1-phenyl-1-pentene [14]. The removal of CO groups and hydrogen transfer led to the release of valerophenone [23]. Valerophenone could also be formed via the oxidation of 1-phenyl-1-pentene [14]. Similar reactions were observed in the case of alkyl cinnamate compounds during oxidative pyrolysis: the double bonds were oxidised forming oxo-compounds (32). Cinnamaldehyde underwent transformation to benzalde-hyde at as low a temperature as 60 °C, as described by Friedman (33).

Scheme 1

As Table 2 shows, the relative intensity of dihydroisobenzofuranones generally decreased due to the aromatisation of the cyclohexadiene rings; however, the cyclohexene rings of alkyl and alkenyl-tetrahydroisobenzofuranones did not undergo aromatisation. The relative intensity of fatty alcohol, acid and esters decreased, while the intensity of shorter-chain alkenes and alkenyl-oxo compounds increased, indicating oxidation reactions and chain scissions. Small changes took place in the distribution of cyclic monoterpenes during low-temperature pyrolysis (Table 1). The total intensity of the bicyclic monoterpenes like α- and β-pinene [1 and 4] was reduced, while the relative yield of monocyclic monoterpenes slightly increased. The four-membered ring of pinenes was cleaved next to the tertiary carbon atom, resulting in the formation of limonene [5] and γ-terpinene [8] similarly to what we observed in lime essential oil previously (31).

The chromatogram of davana oil solution can be seen in Figure 2(a) and the identification of the compounds is found in Table 3. Furthermore, Table 3 lists the relative intensities and standard deviations considering the peaks having > 0.1 area-percent either in the as-received oil or in the pyrolysate. In the as-received oil, 75 peaks were identified, two peaks [42 and 63] among those contained two unresolved compounds and one peak [24] contained three unresolved compounds. Five peaks contained two slightly resolved compounds each [16–17, 19–20, 32–33, 34–35, and 43–44] and their intensity ratios were estimated based on the selected ion chromatograms. Additionally, 29 unidentified peaks were taken into account in the area-percent calculations, which had a relative intensity of > 0.1%. As Table 3 indicates, the majority of the constituents belonged to furan derivatives of different oxygen contents as well as to sesquiterpenes and oxygen- and sulphur-containing sesquiterpenoids. (+)-Davanone [59] was the main component of davana oil representing 29.5 area-percent of the total ion chromatogram. Other davanone isomers [54, 55 and 56] represented only a total of 3.4 area-percent. The second-most intense peak in the chromatogram was bicyclogermacrene [41], amounting to 11.7 area-percent. The intensity of bicyclogermacrene was relatively high, while that of the two bicycloelemene isomers [16 and 18] was very low (below 0.2 area-percent) indicating that the as-received davana oil had been kept fresh (24).

Figure 2

The oil components were categorised into groups according to their chemical structure; and the summed intensities of the compounds in the groups are listed in Table 4. The furan derivatives were grouped into davanones, hydroxyisodavanones, davana ethers and the remaining compounds of smaller molecular mass were referred to as ‘other furan derivatives’.

The furan derivatives constituted about 48% of davana oil, while sesquiterpenes and sesquiterpenoids were the other main components amounting to about 27%. Davana oil contained several aliphatic and aromatic esters; however, most of them at low concentrations with the exceptions of geranyl acetate [25] and methyl [37] and ethyl cinnamates [34 and 47], representing about 1, 1 and 8%, respectively. Ethyl trans-cinnamate [47] was the third-most intense peak in the chromatogram, amounting to 6.3%.

During oxidative pyrolysis at 300 °C, the majority of davana oil components evaporated unchanged due to the moderate temperature and the high flow rate. Nevertheless, there were substantial alterations in the intensity of several compounds, as Figure 2b and Table 3 demonstrate. The significance of the intensity increase of the main constituent, (+)-davanone [59] appears marginal as determined by t-test. However, the second-most important component of davana oil, bicyclogermacrene [41] underwent a substantial decrease in its relative intensity from 11.7 to 6.5%, while the relative concentrations of other sesquiterpenes increased (Table 4). These observations can be explained by rearrangement reactions of bicyclogermacrene [41] as illustrated in Scheme 2a. Since the intensity of a bicycloelemene isomer [18] was increased from 0.18 to 2.03% and another isomer [16] was formed as a result of pyrolysis, these appear logical candidates for the products from conversion of bicyclogermacrene. It is proposed that the rearrangement reaction of bicyclogermacrene [41] took place by a pericyclic reaction referred to as sigmatropic shift. A σ-bond was formed between the tertiary and secondary carbon atoms at the double bonds, a double bond was shifted, and a single bond was broken leading to the formation of bicycloelemene, as demonstrated in Scheme 2a. Other rearrangement reactions of bicyclogermacrene [41] occurred to a smaller degree. The scission of the strained 3-membered ring probably took place with the creation of a new double bond resulting in α-humulene [35]. A σ-bond could be formed between the two secondary carbon atoms at the double bonds leading to the formation of ledene [39]. The intensity of isobicyclogermacrene [38] was increased from 0.23 to 0.66%. Both isobicyclogermacrene and bicyclogermacrene are stereoisomers of 3,7,11,11-tetramethylbicyclo[8.1.0]undeca-2,6-diene, so the transformation between them can be explained by ring-opening and - closing reactions. The relative intensity of a few sesquiterpene alcohols, e.g., [58] and [64] slightly increased, indicating that sesquiterpenes undergo certain oxidation reactions during pyrolysis. The depletion of bicyclogermacrene and the increased concentration of bicycloelemene and sesquiterpene alcohols are in agreement with the observations of Sawamura and co-workers (24, 25), who detected similar changes in the composition of yuzu oil during aging.

Scheme 2

Five davana ether stereoisomers [44, 46, 48–49] were found in the chromatogram, which had altogether 10.4% relative intensity in the as-received davana oil (Tables 3 and 4). The intensity of each isomer was reduced, representing only a total of 5.8 area-percent in the pyrolysate. On the other hand, the intensity of other furan derivatives increased. These changes can be mostly explained by the oxidation of davana ethers, as illustrated in Scheme 2b. cis-Arbusculone [8] was not detected and trans-arbusculone [11] had a minor concentration in the as-received oil; however, their amount was increased by 0.12 and 0.17 area-percent in the pyroly-sate, respectively. The relative intensity of 5,5-dimethylfuran-2(5H)-one [7] was almost tripled during pyrolysis. It is assumed that these compounds were formed in an oxidation reaction as depicted in Scheme 2b. The electron withdrawing capability of the oxygen atom made the double bond between the two furanic rings of davana ether more reactive; hence it could take up two oxygen atoms, while the CSC bond was cleaved, resulting in the formation of arbusculone and dimethylfuranone. The increased yield of 5-ethenyl-5-methyldihydrofuran-2(3H)-one [13] could be explained by scission of davana ether and oxygen uptake of the tetrahydrofuranic ring. The formation of artedouglasia oxide [51, 53, 60] can be rationalised by the following mechanism: An O-atom reacted with the reactive double-bond followed by the opening of the furanic ring and the ring closure took place to the tertiary carbon atom.

Several components of davana oil did not change significantly during pyrolysis. Although the relative intensity of a few esters changed, the total intensity of esters did not alter (Table 4).