Vesuvius is a high-risk volcano and the 1631 Plinian eruption is a reference event for the next episode of explosive unrest. A complete stratigraphic and petrographic description of 1631 pyroclastics is given in this study. During the 1631 eruption a phonolite was firstly erupted followed by a tephritic phonolite and finally a phonolitic tephrite, indicating a layered magma chamber. We suggest that phonolitic basanite is a good candidate to be the primitive parental-melt of the 1631 eruption. Composition of apatite from the 1631 pyroclastics is different from those of CO2-rich melts indicating negligible CO2 content during magma evolution. Cross checking calculations, using PETROGRAPH and PELE software, accounts for multistage evolution up to phonolite starting from a phonolitic basanite melt similar to the Vesuvius medieval lavas. The model implies crystal settling of clinopyroxene and olivine at 6 kbar and 1220°C, clinopyroxene plus leucite at a pressure ranging from 2.5 to 0.5 kbar and temperature ranging from 1140 to 940°C. Inside the phonolitic magma chamber K-feldspar and leucite would coexist at a temperature ranging from from 940 to 840°C and at a pressure ranging from 2.5 to0.5 kbar. Thus crystal fractionation is certainly a necessary and probably a sufficient condition to evolve the melt from phono tephritic to phonolitic in the 1631 magma chamber. We speculate that phonolitic tephrite magma refilling from deeper levels destabilised the chamber and triggered the eruption, as testified by the seismic precursor phenomena before 1631 unrest.
Somma-Vesuvius historic activity is characterised by occasional Plinian eruptions (79 A.D., 472 A.D., 1631 A.D.) separated by long periods of inactivity . Each Plinian eruption was followed by an open-conduit phase, producing either effusive, often eccentric, eruptions  or mixed (effusive and explosive) eruptions, of Strombolian to violent-Strombolian type and sub-plinian . Plinian eruptions produced evolved pyroclastics ranging from phonolitic tephrite to tephritic phonolite . In contrast, eccentric effusive vents poured out fluid phonolitic basanite or phonolitic tephrite lavas (as studied in this paper). This behaviour suggests differentiation in the magma chamber leading to the Plinian eruptions, as opposed to sudden discharge of near-primary magma rising from a deep reservoir. Both reactivation scenarios are needed to describe the volcanic hazard in the Vesuvius area.
A classical approach has been adopted here, with the deliberate omission of models generated by experiments using artificial charges. Hard data from rocks and minerals, collected through accurate stratigraphic work are used instead. Juvenile pyroclastic composition is used to constrain the possible parental magma, its liquid line of descent and chemical zoning in the magma chamber. Magma evolution software was used to verify the hypotheses created using the field and petrological data. With regards to the nomenclature of the rocks we use that approved by International Union of Geological Sciences .
All the hypotheses proposed for the Vesuvius Plinian eruptions are considered to explain the magma chamber evolution up to the triggering of the eruption: (I) fractional crystallisation and an increase in volatiles pressure leading to magma chamber wall breakage , (II) limestone Assimilation and Fractional Crystallisation (AFC) with consequent explosion , (III) feeder dyke formation with the injection of fresh hot magma in the plumbing system and consequent destabilisation of the chamber [8, 9]; (IV) a combination of the above mechanisms .
The behaviour of volatiles, especially CO2, which is particularly crucial in AFC models , has been investigated using apatite compositional variations.
The ultimate aim of this work is the formulation of a melt evolution model for the 1631 magma chamber and its eruption trigger, which represents one of the possible future hazardous scenarios in the case of explosive reactivation of the volcano . This possible unrest has to be evaluated and forecast with maximum effort through a multidisciplinary approach, as the next Vesuvius eruption will impact one of the world most populated and vulnerable volcanic areas.
2 The 1631 AD eruption and deposits
The 1631 event is classified as a relatively small Plinian eruption  following both the approach of  based on the eruptive column height of about 20 km and with the discharge rate (M0) peak of 8 × 107 kg/sec . Before this eruption, Vesuvius was at rest for about five centuries . The 1631 eruption caused more than 4,000 fatalities, significantly affected the Neapolitan region and influenced the evolution of natural sciences in the first half of the 17th century . Due to the proximity of Vesuvius to Naples, this eruption was described in hundreds of contemporaneous chronicles. Long-term seismic precursors are documented and consist of several local events felt in Naples from 1616 to 1630, ranging on the Mercalli Cancani Sieberg (MCS) intensity scale from III to VII . Notably two main seismic sequences are recorded during 1620 to 1622 and in 1626 with a damaging event that occurred on 10th March 1626 at 00:40 h (UT). In addition, on 2nd April 1630 Naples was rocked by a powerful quake described as fairly long and producing a considerable shaking of buildings . In 1631, from November to December, considerable seismic activity occurred in the Vesuvius area culminating in a grade VII MCS shake on 15th December at 23:00 h (UT), followed by an intense seismic swarm (> 30 felt shocks) . As a whole, two seismic swarms culminating in a grade VII MCS event occurred in the five years prior to the eruption. At the same time, short-term precursors, such as sensible ground-deformation, were accurately documented during the final week before the eruption [13, 14].
According to , the eruption started on 16th December 1631 at 07:00 h (UT) with the rapid growth of an eruptive column that lasted around 8 hours. Fallout of vesiculated lapilli and lithic clasts occurred until 18:00 (UT) on this day (Plinian phase). During the night between the 16th and 17th the volcano produced discrete explosions accompanied by lapilli relapses (Vulcanian phase). At 10:00 (UT) on December 17th several hot pyroclastic flows gushed out during a major collapse of the summit of Vesuvius (Nuèes Ardentes phase). These flows travelled to the nearby coast in a few minutes, destroying everything in their path (Fig. 1a). Afterwards a column of ash arose from the large depression left after the Nuèes Ardentes phase. In the night between the 16th and the 17th and in the following days extensive lahars and foods, resulting from heavy rains, affected the valleys of the volcano and spread across the plain North and Northeast of Vesuvius, producing further devastation. This eruptive sequence is not substantially different from that of the Plinian eruption of 79 A.D. 
The main depositional unit of the 1631 eruption consists of a thick Plinian fallout bank (Fig. 1b, Fig. 2c, 2d) composed of vesiculated white-greenish lapilli, crystals and lithic clasts, overlying a 30 cm thick paleosoil (Fig. 2d). At the mid-level of the fallout deposit there is a change in colour of the juvenile pyroclasts from white-greenish to grey-greenish (Fig. 2c). On this basis the Plinian fallout depositional unit is subdivided in two subunits: the “white fallout” and the “grey fallout” (Fig. 1b) . This deposit has been sub-divided into seven layers (from a to f in Fig. 1b) on the basis of planar discontinuities, gradation, grain size and component changes. The crystal-rich a layer (Fig. 1b and Fig. 2d) corresponds to the early stage formation of the Plinian column. Layers b, c, and d form the “white” sub-unit. The layer d is a mixed level with light-and dark-pumices, while the “grey” sub-unit is comprised of ei, es, and f layers. The f layer shows a large amount of lithic clasts and high-density lapilli. In the fallout-type sections of Scudieri and San Leonardo (2 and 8 in Fig. 1a), the Plinian fallout is overlain by a few centimetres of lapilli deposited during the Vulcanian phase (g layer in Fig. 1b).
The pyroclastic rocks produced by the Nuèes Ardentes phase confined to the higher slopes of Somma-Vesuvius were emplaced by diluted density currents (surges) . Concentrated pyroclastic density currents (pyroclastic flows) entered the valleys of the southern portion of the volcano and spread through the foothill areas into the sea (Fig. 1a) [13, 17]. The pyroclastic flows are up to 8 m thick, unwelded, massive and ash-rich, characterised by notable amounts of lava lithic clasts, clinopyroxene adcumulates, skarn fragments, and marls. Occasionally sparse remains of charred vegetation are found (Fig. 2b). In some localities (e.g., Pozzelle quarry) (site 45 in Fig. 1a) up to four flow-units are present (Fig. 2a and 2b).
Fallout deposits were mainly sampled in the two type-sections of San Leonardo and Scudieri (respectively 8 and 2; Fig. 1a). Samples were collected from each one of the seven layers composing the fallout deposit (Table 1). Sampling was repeated in two different sections to obtain robust data on chemical variation inside the fallout deposit. A total of 69 single juvenile scorias from pyroclastic flow units were handpicked and analysed (Table 1). Assuming that the Plinian fallout deposit is fully representative of the magma chamber conditions prior to the eruption, as demonstrated by the bulk rock geochemistry (section 3.2), apatite chemistry of 10 samples from the Plinian fallout were analysed (Table 3a). Apatite crystals, about 0.01 mm long, were separated by heavy liquid (sodium polytungstate) settling, purified using Frantz magnetic separators at different operative conditions and finally hand-picked.
the crater (km)
|45||4.2||N130°E||Pozzelle quarry||1st pyroclastic flow unit||F45PS|
|3rd and 4th pyroclastic flow unit||GIO7|
|53||3.8||N267°E||San Vito-Ercolandia||pyroclastic flow unit||F53J|
|67||6.5||N116°E||Mauro Vecchio quarry||pyroclastic flow unit||F67PS|
|75||5||N230°E||Torre del Greco-Viale della Gioventù||pyroclastic flow unit||F75PS|
|76||7.7||N128°E||Passanti quarry||pyroclastic flow unit||F76PS|
|57||7.3||N183∘E||Villa Inglese quarry||pyroclastic flow unit||GIO5|
|101||6.3||N275∘E||Portici Cemetery||pyroclastic flow unit||GIO9|
|56||4.5||N173∘E||Cappella Nuova-New road behind Viulo cone||pyroclastic flow unit||GIO6|
|2||4.6||N69∘E||Scudieri quarry||g||vulcanian fallout||TIL2-g||2g|
|f||Plinian fallout layer||TIL2-f||2f|
|es||Plinian fallout layer||TIL2-es||2es|
|es||Plinian fallout layer||TIL2-esD|
|ei||Plinian fallout layer||TIL2-ei||2ei|
|d||Plinian fallout layer||TIL2-dS||2d|
|c||Plinian fallout layer||TIL2-cS|
|b||Plinian fallout layer||TIL2-b||2b|
|a||Plinian fallout layer||TIL2-a||2a|
|39||3.5||N115∘E||Pastino-Azienda Agricola Fabbricini||dark-gray Plinian fallout subunit||TIL39-N||39n|
|light-gray Pliniana fallout subunit||TIL39-B||39b|
|8||5.4||N70∘E||Barri-San Leonardo||dark greenish-gray subunit||TIL8-N||8n|
|f||Plinian fallout layer||TIL8-fD|
|f||Plinian fallout layer||GIO10-f|
|f||Plinian fallout layer||GIO11-f|
|es||Plinian fallout layer||GIO10-es|
|es||Plinian fallout layer||GIO11-es|
|es||Plinian fallout layer||GIO12-es|
|es||Plinian fallout layer||TIL8-esS|
|ei||Plinian fallout layer||TIL8-ei|
|ei||Plinian fallout layer||GIO12-ei|
|ei||Plinian fallout layer||GIO11-ei|
|ei||Plinian fallout layer||GIO10-ei|
|d+c||Plinian fallout layer||GIO10-c+d|
|d+c||Plinian fallout layer||GIO11-c+d|
|d+c||Plinian fallout layer||GIO12-c+d|
|d||Plinian fallout layer||TIL8-dS|
|c||Plinian fallout layer||TIL8-cS|
|b||Plinian fallout layer||GIO10-b|
|b||Plinian fallout layer||GIO11-b|
|b||Plinian fallout layer||GIO12-b|
3.2 Bulk rock analyses
Representative juvenile components were crushed and powdered and used for major and trace elements analyses. Seventy-nine whole-rock chemical analyses were performed at Actlabs Laboratories, Canada via inductively coupled plasma mass spectrometry (ICP-MS). A set of CO2 determinations was obtained by Othmer-Fröhlich method at the Institute of Geosciences and Georesources, CNR-Pisa. Fluorine was determined by NaOH fusion and specific ion electrode, Cl was determined by water leaching and ICP-MS. Instrumental sensitivity for major elements is 0.01%, and between 0.01 and 1 ppm for trace elements. Reproducibility is between 2% and 5% for major elements and 2% for most of the trace elements. Data in Table 2a, 2b are rounded to three figures in accordance with instrumental sensitivity.
|PHONOLITE (white lapilli)||TEPHRITIC PHONOLITE (white lapilli)||PHONOLITIC TEPHRITE (drak-grey lapilli)||PHONOLITIC BASANITES|
|de La Roche’s index|
note: (a) = Pyroclastic flow; (b) = Plinian fallout; (c) = Lava; n.a. = not analyzed
|PHONOLITE (white lapilli)||TEPHRITIC PHONOLITE (white lapilli)||PHONOLITIC TEPHRITE (drak-grey lapilli)||PHONOLITIC BASANITES|
3.3 Scanning Electron Microscope (SEM) and Electron Micro Probe Analysis (EMPA)
Apatite crystals were preliminarily examined using an Energy Dispersive System (EDS) on a SEM, then were analysed for Ca, Sr, Fe, Mg, Mn, Na, La, Ce, Nd, Y, P, Si, S, F and Cl with a Cameca Sx50 microprobe run at 15 kVand 20 nA, at the London Natural History Museum (EMMA division) using wollastonite as a standard for Ca and Si, synthetic Sr and Ti oxide for Sr, pure metals for Fe and Mn, olivine for Mg, jadeite for Na, rare earth glass for Ce, La, Nd, Y, Durango apatite for P and F, celestite for S and halite for Cl. As F analysis may result in count acceleration , P and F were analysed with short counting times at the beginning of the analytical sequence. After data collection for the major elements, longer measurements were made for the minor and trace elements, since this second group of elements shows less variation with beam exposure time. The apatite formula A10(XO4)6(OH, F, Cl)2 was adopted.
3.4 X-ray diffraction for determination of lattice parameters (XRDSC), Infra Red (IR) and Raman
Selected single apatite crystals were investigated by XRDSC. IR and Raman spectroscopic investigations were performed to detect the volatile-related components in the structure, such as F-, Cl- and OH-. Lattice parameters were determined by a four-circle X-ray diffractometer (radiation MoKα, graphite-monochromatised). About 50 refections were adopted to calculate the parameters using a LAT program of Phillips with an estimated error lower than 0.001Å. The a and c cell parameters are listed in Table 3a and depicted in Fig. 4a–4c. The IR spectra using a KBr disc were scattered in a Fourier Transform IR instrument Bruker mod. IFS113V, collecting 50–100 scans in vacuum (P = 5 mbar) over the range 400–4000 cm-1. The oriented micro-Raman spectrum sample 39b was measured using a 180-degree back-scattering geometry with a Labram micro-spectrometer (instrument mod S.A.). A curve-fitting program was applied for the determination of frequencies of the weak bands for both IR and Raman spectra.
|atoms per formula unit (a.p.f.u.)|
4 Mineralogy of 1631 apatite
4.1 Apatite geochemistry
Owing to its particular crystal-chemical properties, apatite structure allows a number of substitutions at different sites. During fractional crystallisation the melt enriches in volatiles, such as H2O, H2S, SO2, CO2, HF and HCl, as the corresponding ionic species, OH-, S2-, F-, and Cl- are not incorporated into the lattice of crystallising solid phases. Some of these ionic species (e.g., OH-, F- and Cl-) can easily enter the apatite structure [19, 20]. Consequently, the compositional evolution of apatites mimics that of the magma from which they crystallised. In particular, the substitution among the F-, Cl- and OH- ions, which occupy the structural channel (Z site), is a sensitive indicator of the fugacity of corresponding volatiles in the magma [21, 22]. So, co-variations in apatite ionic species (e.g., OH-, Cl-, F-) related to volatile components (e.g., H2O, HCl, HF) and non-volatile elements (e.g., Rare Earth Elements (REEs)) in the magma suggest that a common process (e.g., fractional crystallisation) occurs during apatite crystallisation (Fig. 4a–4c).In addition, apatite is crucial to assess CO2 content in the crystallising melt.
4.2 XRDSC and Vibrational spectroscopy
Vibrational Spectra of the 1631 apatite samples clearly show the OH absorption bands at 3540 cm-1 (stretching mode) and in the region of 725–750 cm-1 (libration mode), in agreement with the OH content obtained from formula calculations (Fig. 3a). By IR we find the difference between Fap and OH-bearing Ap and classified it to F-Hap. However, OH content is estimated by difference in channel site, assuming F + OH+ Cl = 2 [23, 24]. Based on the EMPA data, the tetrahedral sites are fully occupied by phosphate ions in the 1631 samples and there is no evidence for the presence of carbonate ions. Accordingly, in the Raman spectra of the studied samples, only PO4 vibrational bands at 961 (ν1), 471–474 (ν2), 1040–1090 (ν3) and 566–604 cm-1 (ν4) can be resolved and there are no carbonate bands. For example, in the representative Raman spectrum of sample 39b (Fig. 3b), one ν1 peak at 964 cm-1, two ν2 peaks at 430 and 446 cm-1, four ν3 peaks at 1029, 1048, 1059, and 1078 cm-1 and four ν4 peaks at 580, 590, 607, and 616 cm-1 were resolved, similar to those of the apatite end members . Finally, Si and S contents of the 1631 samples should be less than 1% because it is not possible to resolve peaks related to structural SiO4, SO4, CO3 anions by these Raman spectra.
In the 1631 apatites the A-site is chiefly filled by Ca with a slight substitution of Mg, Sr and light REEs (less than 0.1 apfu), while the X-site is mainly occupied by P with a weak substitution of Si and S(generally less than 0.1 apfu). The most distinctive substitution occurs in the channel, with F varying from 1.05 to 1.82 apfu, Cl from 0.02 to 0.24 apfu and OH from 0.16 to 0.80 apfu (Table 3a).
4.4 Apatite data implications
Apatites from the vent-opening deposit levels (sample 2a in Fig. 3c, d and Table 3a) and the first emitted Plinian fallout pumices (sample 2b in Fig. 3c, 3d and Table 3a) have the lowest molar fraction of F-Ap, and the highest OH-Ap values, whereas apatites from the Vulcanian phase deposit (sample 2g in Fig. 3c, 3d, Table 3a) have the highest F-Ap and the lowest OH-Ap fraction (Fig. 3c). The molar fractions of F-Ap and OH-Ap change quite regularly along the sequence, except for sample 2f, which shows lower F-Ap and higher OH-Ap than adjacent samples, consistent with its probable formation position at the contact with the skarn shell of the magma chamber. On the other hand, there is no significant co-variation of Cl-Ap with stratigraphic position. The variation of a and c cell parameters, together with the c/a ratio, are characteristic of OH-F binary apatites (Table 3a and 3b). Only F, Cl and OH are meaningful in terms of correlation with c/a with magmatic evolution along the eruptive sequence (4 a, b and c). According to , who established chemical characteristics for apatites from different alkaline rocks and carbonatites, the 1631 compositions, in terms of F-Cl-OH, are typical of mafic alkaline silicate rocks and are clearly different from those of apatites crystallising from CO2-rich melts (Fig. 3d). Summing up, the chemistry of 1631 apatites excludes their crystallisation from either a saturated melt or a carbonate-contaminated melt and highlight a general increase of volatiles (such as OH) from the bottom to the top of the magma chamber, which is coherent with a fractional crystallisation process.
5.1 Juvenile lapilli petrography
Juvenile lapilli of the 1631 eruption are vesiculated, showing a porphyritic texture in which there is a continuous variation in crystal size (i.e., seriate texture) (Fig. 5a) and variable mineral phases/glass ratios, with crystal content between 20 and 36 vol.% .
Lapilli pheno-modal composition is generally leucite phonolitic tephrite but many samples show clearly that K-feldspar dominates on plagioclase and are leucite tephritic phonolite or even phonolite. Phonolitic tephrite samples contain more abundant leucite and magnetite. Main phases are: glass (44.0–74.0 vol.%), clinopyroxene (5.6– 34.1 vol.%), leucite + (nepheline) (3.7–12.2 vol.%), plagioclase (2.0–9.5 vol.%), mica (0.7–9.1 vol.%), K-feldspar (0.4–5.4 vol.%), magnetite (0.4–3.2 vol.%), apatite (0.5– 2.6 vol.%), garnet (0.4–0.7 vol.%) and olivine (1.0–2.0%). A feldspatoid of the cancrinite group (microsommite) was also observed . K-feldspar forms corroded skeletal or euhedral phenocrysts (Fig. 5c). Larger corroded crystals show kaolinitic alteration along cleavages, with faint zonation and/or perthites. Abundant, euhedral fresh K-feldspar laths are in the groundmass. Plagioclase is zoned, generally subhedral and forms seriate phenocrystals. Leucite is generally fresh, euhedral in the ground-mass with concentric magnetite inclusion, or forms large fragments up to 2 cm (Fig. 2e), showing characteristic sector polysynthetic twinning and K-feldspar and clinopyroxene inclusions (Fig. 5b). Nepheline is limited to small intergranular patches. Clinopyroxene occurs as large subhedral or euhedral zoned crystals, showing corrosion features, whereas smaller groundmass clinopyroxene crystals have skeletal features. Biotite-phlogopite fakes show kink bending and fringed endings. Some large mica crystals show corrosion and reaction rims, which indicate a possible xenocrystic origin. Mica contains abundant clinopyroxene and is possibly disaggregated from larger ultramafic nodules of mica-clinopyrossenite (Fig. 5d). Olivine forms small subhedral grains in the ground mass or rare resorbed phenocrysts. EMPA of olivine testifies mafic glass and apatite inclusions but not carbonate inclusions. Garnet is present as small euhedral crystals immersed in the groundmass. It shows red-brown polarisation colours and slight birefringence typical of Ti-rich garnet melanite. Apatite occurs as an inclusion in clinopyroxene (Fig. 5e, 5f) and K-feldspar, or as discrete euhedral crystals in the groundmass (Fig. 5g) up to 250–300 µm showing clinopyroxene inclusions (Fig. 5h). Magnetite forms discrete euhedral crystals or crystal aggregates in the groundmass. Glass is clear, colourless to brown-yellow.
5.2 Skarns lithic fragments
Skarns are an interesting part of the lithic clasts from the 1631 eruption and sometimes show multiple complex contacts among igneous rock, skarn and marble (Fig. 6). The marble is medium-grained, quite pure periclase/brucite marble (Fig. 6a). The skarns are banded and their contact with porphyritic tephritic phonolite is marked by a phlogopite and/or clinopyroxene reaction rim, 1–10 mm thick (Fig. 6a). The skarn bands consist of millimetrethick alternating undulated or convolute lamination of essential forsterite, spinel and calcite, roughly parallel to the boundaries of the magmatic rock rim (Fig. 6b). Skarn blocks vary from sharp fragmental shape to rounded corroded shapes infiltrated by the melt (Fig. 6c). A detailed mineralogical description of these skarns is found in [26, 27]. Observation of composite igneous, skarn and marble contacts in the 1631 ejecta testifies the very limited thickness of the skarn shell.
5.3 Bulk rock geochemistry
All 1631 rocks are generally silica under-saturated and have a potassic character. White lapilli (tephritic phonolite to phonolite) have an average SiO2 content of 50.4 wt.%, up to a maximum of 53.4 wt.%. Grey lapilli (phonolitic tephrite) are always SiO2-undersaturated with minimum SiO2 of 44.9 wt.% and average of 48.1 wt.%. Notable differences in average major oxides characterises tephritic phonolite with respect to phonolitic tephrite lapilli: TiO2 is 0.66 and 0.82 wt.%, Al2O3 is 18.8 and 16.7 wt.%, MgO is 2.60 and 4.93 wt.%, CaO is 7.76 and 10.7 wt.%, respectively, while Na2O+K2O is 11.4 (12.7 in the phonolitic samples) and 9.12 wt.% and P2O5 is 0.41 and 0.66 wt.%, respectively. The CO2 content of most samples is below detection limit (< 0.1 vol.%) apart from two samples whose CO2 concentration is between 0.5 and 0.7 vol.%, corresponding to 0.2–0.3 wt.% CaCO3. In fact, very small fragments of sedimentary limestone are present as accidental lithics in the pyroclastic rocks that cannot be completely removed. SO3 contents are generally less than 0.5 wt.%.
A general picture of major and trace element variations is depicted by their covariance with the solidification index (S.I. = 100 [MgO / (MgO + FeO + Fe2O3+ Na2O + K2O)]) of  (Fig. 7a–7f, Fig. 8a–8h). The solidification index is preferred in this context to the more commonly used SiO2 or Mg# index [Mg / (Mg + Fe2+)], because the S.I. explains a much larger proportion of chemical variance, consequently highlighting the possible addition, if any, of external Ca and Mg (limestone) to the “magmatic” component.
Juvenile fragments from pyroclastic flows show a bimodal composition virtually identical to the light and dark lapilli in the Plinian fallout. Notably, in the white lapilli there is a more evolved sample that detaches from the rest in all the diagrams but compares with four lapilli samples from the pyroclastic flow samples (Fig. 7a–7f; Fig. 8a–8h). Although these evolved samples are ca. 5% of the dataset, data are consistent enough to justify a different rock-type (Fig. 9).
In the Total Alkali vs. Silica (TAS) diagram (Fig. 9a), the 1631 rock samples are somewhat dispersed in four different classification fields but most samples plot in the phono tephrite and tephri phonolite plotting fields. Several 1631 pyroclastics have water content > 2% and the use of the TAS diagram it is not advisable above this threshold . Classification cannot simply be done based on the TAS diagram and a double check is required. The de La Roche’s classification diagram is a neat addition and offers more value because its suitability for volatile-rich and alkaline rocks  (Fig. 9b). In addition, it has the two-fold advantage of being semi-normative and able to explain a much larger chemical variance as it involves all the major oxides and carbon . Chemical composition of the 1631 rocks in the de La Roche diagram follows a narrow path basanite → tephrite → phonolitic tephrite → phonolite. In addition, this diagram shows that the tephrite and phonolitic tephrite compositions are pretty well separated and basanite and phonolite compositions are also represented.
Normative calculation (CIPW) (Table 2a) for 1631 samples provides additional information determining the ideal mineralogy of these porphyritic rocks. The degree of silica saturation produces lc + ne saturation and abundant diopside formation. About 50% of the samples express some normative olivine, from 1.2% to 10.2% with an average of 4.2%, whereas modal olivine ranges from only 1 to 2%. Normative or + ab / wo + en + fs + fo + fa ratio works as a differentiation index and when plotted against normative lc + ne indicates that compositions with higher mafic content, for each rock types, correspond to the less evolved rocks having higher foids content (Fig. 10a, 10b). From a normative point of view, the 1631 samples are well defined in the Fo–(Ne + Ks)–Or triangle of the Rm–Rs–Ri diagram (Fig. 10c) . More primitive compositions plot in the Fo– (Ne + Ks)–Lc sub-triangle but most compositions move towards Lc–Or tie-line in the Fo–Lc–Or sub-triangle. When the CIPW norms of 1631 samples are plotted in the conventional alkali feldspar - quartz - plagioclase - foids (AQPF) normative diagram(Fig. 10d), they spread from foidite (virtually Or-free) to tephritic phonolite. In general, normative diagrams are consistent with chemical classification and indicate a Fo-bearing composition (tephrite or basanite) evolving to an Or-rich term (phonolite). Combining chemical, normative and modal compositions, the following classification for the 1631 rocks types is adopted: leucite phonolitic tephrite, leucite tephritic phonolite and leucite phonolite. Phonolitic basanites are excluded upon  criteria because modal olivine is << 10 vol.% in the 1631 samples.
In tephritic phonolite and phonolitic tephrite, Ba averages 1778 and 1652 ppm, Sr averages 1035 and 921 ppm and the Cr + Ni sum averages 70.2 and 104 ppm, respectively. General chemical evolution of the melt is obtained by plotting large lithophile Elements (LILE) and high field strength elements (HFSE) vs. solidification index (Fig. 8). A considerable decrease in the Cr + Ni sum coupled with a significant increase in Th and La, Zr and especially incompatible Rb is apparent (Fig. 8). Again three different chemical rock-types are defined.
The most primitive 1631 samples TILES(4)D has MgO = 5.42 wt.%, Mg# = 78 and Cr + Ni = 123 ppm. These values are comparable with those reported in the literature for the 1631 most primitive sample, GV2, from  that contains MgO =5.82 wt.%, Mg# =77and Cr + Ni= 148 ppm (Table 2a and 2b).
The most primitive Vesuvius rocks may be represented by medieval lavas PA10 or PA28 that have MgO = 7.77– 8.01 wt.%, Mg# = 67–83 and Cr + Ni = 273–324 ppm and normative olivine (Fo + Fa) ~ 10% (Table 2a and 2b). Representative Vesuvius basanite (PA28) displays higher LILE but lower Zr-Hf and Ta-Nb with respect to selected Mediterranean rocks classified as basanites in both the TAS and de La Roches’s diagrams and having ~10% of olivine in the CIPW norm (Table 2a and 2b, Fig. 11a). 1631 rocks conform to the geochemical pattern of hypothetical basanitic parental magma (PA28) but have a slightly higher LILE/HFSE ratio as expected from evolved daughter rocks (Fig. 11a). Total REEs average 286 ppm in 1631 leucite phonolite, 282 ppm in leucite tephritic phonolite and 270 ppm in leucite phonolitic tephrite. The average LREE/HREE ratio is 45 in 1631 phonolite and 1631 tephritic phonolite and 43 in 1631 phono tephrite, as expected in rocks evolved from each other. In addition, a crossover occurs at Nd level, indicating that phonolites are enriched in LREE and notably in La.
The compositional ranges for the three 1631 rock-types substantially overlap with average orogenic Mediterranean basanites (Fig. 11b). There is a slightly negative Eu anomaly in the order of that observed in average Mediterrean basanites.
Both in middle ages and in modern times, Vesuvius issued olivine-bearing lavas during intra-Plinian activity by vents aligned WSW–ENE on the southern fank of the volcano [2, 17]. These Vesuvius phonolitic basanites (Table 2a and 2b) are here considered to have a primitive magmatic composition (i.e., the Vesuvius parental melts). These lavas have modal olivine ~10%, MgO content > 7 wt.%, high Mg# (83– 86) and high Cr + Ni (usually in the 200–270 ppm range), which are very reasonable figures for primitive, mantle-derived melts . CaO content is up to 12.9%, which is relatively high for alkaline basalts that usually have between 6 and 12%, but more similar to average Mediterranean basanites, which have 11.43 wt.% (Tab. 2a). Italian kamafugitic rocks contain much higher CaO (up to 15.4) but are intrinsically related to a carbonatitic component, which is not present at Vesuvius .
If we look at ‘immobile couples’ Hf/Zr and Ta/Nb ratios of all the rocks considered are distributed in a narrow range of variation (Fig. 12) around an average composition of Hf/Zr ratio of 0.023 and Ta/Nb ratio of 0.061, comprising 193 anorogenic Mediterranean samples which classify consistently as basanite . Genuine basanites are rare in Italy and restricted to the Veneto Province  and Hyblean plateau . In addition, there are some samples from the Roman Region . Vesuvius basanitic lava PA28 and average historical Vesuvius lavas plot in the high Hf/Zr and low Ta/Nb quadrant with respect to average Mediterranean basanites and are more similar to Hyblean Distric and Roman Region basanites; whereas, Veneto district basanites have much higher Ta/Nb and thus represent a different mantle source (Fig. 12). Notably, average 1631 phonolitic tephrite is also very close to the average mediterrean basanites (Table 2a and 2b), suggesting a similar genetic condition.
The use of fractional crystallisation and mass balance modelling by petrologic software requires pressure (P) and temperature (T) inputs, which have to be obtained via mineralogical and geological constraints. Clinopyroxene, olivine ± Cr-spinel are the first phase to crystallise in primitive Vesuvius liquids  plus leucite at pressures > 2.5 kbar. The Fo content of olivine is used to estimate the crystallising melt temperature . Olivines from 1631 samples average Fo96 and typically contain Cr2O3 + NiO between 0.01 and 0.03 wt.% (Table 3b); this could be a reaction product found in the skarns. Olivine from Vesuvius lavas have Fo content up to 90 mol.% . According to correlation curves, Fo90 crystallising melt would have a maximum potential temperature of 1300°C . This is in accordance with the liquidus T at 6 kbar modelled by software at 1280°C and is lower with respect to Italian primitive kamafugites, which are associated with carbonatites . Mineralogy of 1631 deposits suggests that evolved 1631 magma rested in a magma chamber where low-pressure phases, such as leucite and vesuvianite can crystallise (max 2.5 kbar) [41, 42]. Coexistence of K-feldspar and leucite in the 1631 deposits constrains the temperature of the magma chamber at ~1090°C, based on the classical experimental phase equilibria results of  for the system NaAlSi3O8-KAlSi3O8-H2O. Recently, a pre-eruption temperature of 977 ± 30°C was computed for the 1631 eruption by  using the CaO content of the most felsic ground-mass glass and CaO geothermometry. The minimum temperature of the 1631 magma crystallising inside the magma chamber is assumed to range from 1170 to 850°C based on homogenising temperatures of melt inclusions . Leucite may be metastable during K-feld cristallisation, in this case a temperature between 900 and 850°C is reasonable at a late stage of melt evolution. These temperatures are also in agreement with δ34S and Sulphur contents measured in 1631 volcanics by . This pre-eruption temperature correspondstoa water pressure of ~ 1.25kbar along the incongruent melting curve of K-feldspar , whereas the uncertainties regarding the temperature estimated by  converts to a pressure range of ~ 0.8 to 2.6 kbar. At water contents larger than 1 wt.%, the phonolitic melts has a very low viscosity even at temperature < 850°C . A low pressure superfcial chamber located about 1.5 km beneath the volcano is consistent with the caldera-like collapse of the main Vesuvius cone that occurred in the aftermath of the Nuées Ardentes phase of the 1631 eruption  and is confirmed by the comparison of the nature and distribution of lithic clast populations within the 1631 deposits and the stratigraphic sequence beneath the volcano . The intense seismicity that occurred from 19th November to 16th December restricted to the villages surrounding the volcano also ft with a very superficial source.
In addition to literature data, we focused on the identification of a parental magma, which should rely both on compositional and textural constraints of the 1631 and the Vesuvius rocks. Among the Vesuvius phono-basanitic lavas with Mg#≌ 80, PA28 sample is a suitable candidate for determining the parental liquid of the 1631 eruption.
Starting from this composition, quantitative melt differentiation calculation was performed by mass balance and fractional crystallisation modelling using the Petrograph software  and equilibrium line of descent by the PELE software , assuming pressure between 6 and 0.5 kbar in accordance with field and mineralogical data. At higher pressure and lower quartz-magnetite-fayalite redox buffer (QMF) crystallisation models are invariably dominated by crystallisation of clinopyroxene, which appears early in the sequence, whereas olivine is subordinate (Fig. 13a, Supplementary materials 1). In fact, at lower pressures olivine crystallisation would be insufficient and plagioclase would disappear at equilibrium. Results of the crystal fractionation modelling suggest that at 6 kbar, Vesuvius phonolitic basanite can fractionate enough olivine (4.8%) and clinopyroxene (18%) to evolve to phonothephritic composition (TILES) at ~1220°C with a QMF + 1. Plagioclase, nepheline and alkali feldspar crystallise in higher proportion and with lower olivine/plagioclase ratios in phono-tephritic composition (TILES) with respect to phonolitic basanite (PA28) at the same P and QMF but at a lower T < 1140°C (Fig. 13b). Such differentiation conditions are not possible at lower pressure and higher QMF. Further 3.6 wt.%olivine, plus 13.8 wt.%clinopyroxene and 1.3 wt.% of Ti-Magnetite fractionation would move the liquid to a tephritic phonolite (F67PS(6)) with a T decrease of ~90 °C, which virtually consumes all the olivine in the liquidus (Fig. 13c). Calculations indicate that the subsequent crystallising phases are invariably represented by leucite plus nepheline, clinopyroxene, magnetite, alkali feldspar and apatite, in excellent agreement with the mineral associations of the 1631 rocks.In the range of 1100–940°C, QMF +2 and 2.5 kbar, tephritic phonolite affected by crystal settling of 15.6 wt.% clinopyroxene, 8 wt.% leucite, 1.8 wt.% Ti-magnetite and 0.8 wt.% apatite would evolve to phonolite (Fig. 13d, Supplementary materials 2). At 0.5 kbar, modelling indicates that liquid becomes undersaturated in leucite at about 840°C and massive alkali feldspar crystallisation ~40 wt.% starts (Fig. 13d). The calculated liquid lines of descent ft well with the magmatic evolution from phonolitic basanite to phonolite, the latter representing the extreme differentiates at low T and P and high QMF. Therefore, modelling results indicate that “normative” passage from foidite to phonolite depends mainly on specific cpx, olivine crustal settling at a specific depth and temperature, that plagioclase and nepheline remain in the melt and that leucite may float from tephritic phonolite to phonolite (Table 4), where it may be unstable (Fig. 13d).
|Density of crystal|
|6 kbar||1||3,40 diopside||81,2|
|6 kbar||1||3,40 diopside||51,7|
|2.5 kbar||0,1||3,19 apatite||0,25|
|0.5 kbar||0,1||3,19 apatite||0,39|
As aconsequence, 1631 melt shows a sharp decrease of TiO2, CaO, MgO, P2O5 and Cr + Ni during differentiation, a phenomenon well depicted by the Ca#/S.I. or Mg#/S.I. (Fig. 7e–7f). A negligible Eu anomaly (Fig. 11b) suggests that plagioclase did not contribute to the crystal settling.
Settling velocities of heavy minerals in such low-viscosity, volatile-rich melts is quite high and suggest a rapid melt differentiation (Table 4). Viscosity, volatile content and temperature was experimentally calculated for the 1631 phonolitic melt portion [47, 51, 52]. We calculated similar values for the main 1631 rocktypes and crystal settling velocity using Stokes’ law (Table 4). Different melt densities of calculated dense rock values suggest that in the magma chamber there were at least two liquids with different physical and chemical properties, which favoured relatively isolated, immiscible strata in the chamber. We speculate that the more primitive phonolitic tephrite was injected into the magma chamber at a time relatively close to the eruption. This explains why three different rock types were erupted at different times (phases) of the Plinian fallout, which represent a gradual magma chamber discharge from top to bottom.
As a whole, the liquid line of descent depicted in Figures 8, 9, 10 and 13 suggests that: (i) the parental 1631 magma (first appearing inside the magma chamber) is a phonolitic tephrite produced by olivine fractionation in a deep reservoir of a phonolitic basanite (similar to the most primitive Vesuvius erupted lavas and Hyblean and Roman Region basanites); (ii) the residual liquid evolved inside a crustal magma chamber to leucite tephritic phonolite and finally to leucite phonolite composition, at subsurface conditions.
5.5 Carbon dioxide
Carbon dioxide in both 1631 apatites (see Section 3.3) and glass inclusion is negligible . 1631 apatite composition are classified as mafic alkaline and are clearly different from those crystallising from CO2-rich carbonate melts (Fig. 3d). At crustal pressure, the CO2 production by melt/country rock reactions is likely limited to skarn formation. Geologic and textural evidence of a thin skarn shell in the 1631 eruption ejecta testify the small volume of in situ metasomatism and/or thermo-metamorphism at the magma chamber walls but do not imply per se substantial assimilation. Fresh sedimentary lithic clasts in the 1631 pyroclastic rocks are highly indicative of limited magma/crust interaction and thus potential assimilation. Skarn assemblages suggest temperatures of 650–700°C for the skarn forming processes, which typically include two types of metasomatic reactions, i.e., formation of spinel-forsterite-calcite endoskarns by desilication of aluminosilicate bodies at the contact of dolostone wall rocks and reaction of pre-existing endoskarns with new infuxes of magma and fluid as documented by [26, 27]. According to these authors, these metasomatic processes are promoted by CO2-rich fluids and act as sinks of CO2, producing calcite as part of the skarn mineral assemblage. It cannot be excluded that changes in temperature and CO2 fugacity may drive these reactions in the opposite direction, destroying carbonate minerals and producing CO2 . The potential CO2 produced is expected to escape towards the surface considering that: (i) the skarn shell prevents CO2 movement from country rocks into the magma chamber and (ii) dissolution of CO2 in the magma canbe considered negligible at the relatively low pressure of the 1631 magma chamber (see above) based on the experimentally determined CO2 solubility in melts (51).
The production of free CO2 at deeper levels, through deep magma degassing or mantle degassing is very likely, considering the high flux of partly magmatic CO2 discharged today in Central-Southern Italy, including the Vesuvian area [54, 55]. Nevertheless, the deeply originated CO2 did not dissolve into the shallow 1631 magma chamber for the reasons noted above.
5.6 Limestone assimilation and AFC model
A substantial country-rock assimilation (>>10%) and consequent fractional crystallisation (AFC model) was assumed in the past literature at Vesuvius, originating from the classical work of . This assumption was based on: (I) the widespread presence of limestones, dolostones and marls in the substratum of the volcano, (II) the presence of skarns as ejecta in the volcanics and (III) the abundant CO2 emission at the surface. Assimilation of cold rocks on a large scale requires large amounts of heat and a proportionally larger amount of very hot magma. Felsic melts generated by underplating at the mantle crust boundary is likely but is a different assimilation model or better melting-mixing model [57–59]. Many authors strongly criticised this model being widely used for rocks similar to those of Vesuvius and raised objections based on thermodynamic grounds and geochemical evidences [60–62]. However, some authors, using an experimental approach, still assume that: (I) the Vesuvius primitive/parental melt is a trachy-basalt with MgO of ~ 5 wt.% and temperature of ~ 1100°C and (II) it assimilated significant amounts of limestone, thus chemically regressing to SiO2-undersaturated tephritic compositions [7, 63]. This topic requires further discussion.
Geochemical features of the 1631 Vesuvius magma point to a mafic alkaline SiO2-undersaturated parental melt, with a CaO content of 12.8 to 14.0 wt.%. Notionally, this composition may be obtained by desilication reactions driven by carbonate addition, such as:
which may convert the basaltic assemblage (plagio-clase dominated) into a nepheline-melilite bearing paragenesis. In fact, these reactions form gehlenite and wollas-tonite, which are among the main constituents of melilite and diopside, respectively. The two main decarbonation reactions referred by  are either:
if olivine is present, or:
if olivine is absent. Therefore, the main expected consequences of the progressive digestion of carbonates are destruction of forsteritic olivine, plagioclase and silica and production of CO2 and diopsidic clinopyroxene.
It must be noted, however, that the experimental runs of  diverge on several aspects from the natural Vesuvius system and in particular from the 1631 magma chamber and related products. Therefore experimental results cannot be neatly applied to the 1631 case.
Secondly, in the TAS and de la Roche’s diagrams, carbonate assimilation and related processes cause either a decrease in SiO2 at constant Na2O + K2O or a decrease in SiO2 accompanied by an increase in Na2O + K2O as recognised by  (see their Fig. 2 and related discussion). Both trends are roughly perpendicular to the 1631 evolution trend, as shown in Figure 9a and 9b, indicating that there is a marked contrast between the chemical changes driven by carbonate assimilation and the data from the 1631 natural system. The only common point between the experimental runs and the most primitive samples of the 1631 eruption is due to the artificial addition of olivine in the experiments.
6.1 Vesuvius parental melt
Although basanite modal compositions are not present among the 1631 samples, Vesuvius basanites exist, even if they were rarely reported in the Vesuvius literature, as already pointed out by . We found that emission of basanitic lavas is characteristic of eccentric fissural eruptions that occurred during middle ages and also in modern times of Vesuvius activity (Table 2a and 2b, Fig. 9). From the modal point of view they are phonolitic tephrite (Ol < 10 vol.%) or phonolitic basanite (Ol ≥ 10 vol.%). From the chemical point of view (according to the diagram of , 1986, Fig. 9b) these rocks are confirmed to be basanites. Most primitive samples [e.g., GV2 or Tiles(4)D in Table 2a and 2b] of the 1631 eruption are phonolitic tephrite but chemically they are close to basanitic composition, with MgO of 5.42 wt.%, Mg# of 69 and Cr + Ni of 115 ppm. Normative calculation (CIPW of Fig. 10d) and the software-calculated liquid line of descent at equilibrium often indicates foidites at low pressure crystallisation; however, this is in contrast with essential plagioclase and k-feldspar in the modal composition of 1631 rocks.
6.2 Magma differentiation
The 1631 rocks (phonolite, tephritic phonolite and phonolitic tephrite) are likely produced by means of crystal fractionation processes in a primitive phono-basanitic melt, similar to that erupted during medieval and modern Vesuvius activity and having a composition similar to many other Mediterranean anorogenic basanites. Melt evolution developed through fractional crystallisation of olivine, clinopyroxene and plagioclase, plus nepheline and leucite at lower pressure and temperature. This evolution is reproducible by mass balance calculation and synthetic liquid line of descent with appropriate mineralogy. A first hypothesis is that the parental basanite melt might under-plate at the base of the crust, at a depth of approximately 20 km . We speculate that the primary basanite melt may have experienced an early stage of clinopyroxene and olivine separation and evolution to phonolitic tephrite evidenced by adcumulates (Fig. 13c) . A further stage is upward migration and stationing in a crustal reservoir, possibly located ca. 8–9 km under Vesuvius  and coinciding with the base of the sedimentary cover (Fig. 14). At that pressure a reasonable temperature of 1190°C allows relatively moderate clinopyroxene and olivine fractionation and the melt to evolve significantly to thepritic phonolite (Fig. 13). Injection of tephritic phonolite melt into a subvolcanic magma chamber may have occurred through repeated refilling . Crystallisation of K-feldspar and other minor phases may produce a layer of immiscible phonolitic liquids at pressures << 1 kbar and a T of about 840°C (Fig. 13). Leucite would be metastable in these conditions but its ability to float towards the top of the chamber explains melt undersaturation, which was previously interpreted as limestone assimilation and magma desilication. The observed and calculated crystallisation patter does not require assimilation; crystal fractionation is a sufficient condition.A further injection of hot phono-tephritic melt inside the 1631 magma chamber was likely favoured by local tectonic events accounted for by relatively intense volcano tectonic earthquakes felt in Naples during the five years prior to the eruption .
6.3 Magma chamber breakage and eruption scenario
At the wall-rock/magma interface of the magma chamber, represented by the skarn carapace, internal volatile-supported pressure closely balances the external (overburden) pressure as suggested by the relatively long-term magma differentiation without magma chamber deformation. This delicate equilibrium was probably destabilised by the last refilling of fresh magma from the deep reservoir producing sudden breakage of the country rocks accompanied by a notable number of local shakes in the two months before the 1631 eruption. This mechanism is consistent with the presence of both long- and short-term precursors before the 1631 eruption and distribution of recent instrumental ipocentral locations (Fig. 14) . The inferred scenario for the next explosive eruption at Vesuvius involves the formation of a magma chamber, produced by the rising and stationing at shallow depth of a magma batch, possibly accompanied by volcano tectonic earthquakes felt over a large area. The melt in this chamber would stratify and crystallise rapidly, producing volatile pressures that might be able to trigger an explosive eruption in times that are essentially dependent on the frequency and volume of the subsequent feeding events. Vesuvius seismicity suggests two possible volcano seismogenetic levels similar to those hypothesised for the 1631 precursors; however, seismic activity is low and no sequences comparable to those of the 1616 to 1630 period have since occurred. A further scenario is possible, either in the absence of the magma chamber, or in its presence if the uprising phono-tephritic or phono-basanitic magma does not enter the chamber, of discharging directly to the surface through fractures dissecting the low flanks of the volcano. This is a comparatively less severe scenario but low-level vent locations and related outpouring lava would be locally disastrous on the very densely populated coastal band.
Massimo Guidi and Ilaria Baneschi (IGG-CNR) performed CO2 bulk rock analyses. Frances Wall and EMMA staf at NHM is acknowledged for their kind assistance during apatite EMP analyses. Olivine analyses are courtesy of Geosciences Department of Padua University. We are grateful to Luigi Marini for the useful discussions on CO2 role, and to Giada Iacono-Marziano for valuable suggestions about improvement of an earlier version of the manuscript.
 Arnò V., Principe C., Rosi M., Santacroce R., Sbrana A., Sheridan M.F., Eruptive history. In: Santacroce R. (Ed.), Somma-Vesuvius. CNR Edizioni, Roma, 1987, 114(8), 53-104Search in Google Scholar
 Principe C., Tanguy J.C., Arrighi S., Paiotti A., Le Goff M., Zoppi U., Chronology of Vesuvius activity from A.D. 79 to 1631 based on archeomagnetism of lavas and historical sources. Bull. Volcanol., 2004, 66, 703-72410.1007/s00445-004-0348-8Search in Google Scholar
 Arrighi S., Principe C., Rosi M., Violent strombolian and sub-plinian eruption at Vesuvius during the post-1631 activity. Bull. Volcanol., 2001, 63, 126-15010.1007/s004450100130Search in Google Scholar
 Ayuso R.A., De Vivo B., Rolandi G., Seal II R.R., Paone A., Geochemical and isotopic (Nd-Pb-Sr-O) variations bearing on the genesis of volcanic rocks from Vesuvius, Italy. J. Volcanol. Geotherm. Res., 1998, 82, 53-7810.1016/S0377-0273(97)00057-7Search in Google Scholar
 Iacono Marziano G., Gaillard F., Pichavant M., Limestone assimilation by basaltic magmas: an experimental re-assessment and application to Italian volcanoes. Contrib. Mineral. Petrol., 2008, 155, 719-73810.1007/s00410-007-0267-8Search in Google Scholar
 Civetta L., Galati R., Santacroce R., Magma mixing and convective compositional layering within the Vesuvius magma chamber. Bull. Volcanol., 1991, 53, 287-30010.1007/BF00414525Search in Google Scholar
 Morgan D.J., Blake S., Rogers N.W., De Vivo B., Rolandi G., Davidson J.P., Magma chamber recharge at Vesuvius in the century prior to the eruption of A.D. 79. Geology, 2006, 34 (10), 845-84810.1130/G22604.1Search in Google Scholar
 Barberi F., Davis M.S., Isaia R., Nave R., Ricci T., Volcanic risk perception in the Vesuvius population. J. Volcanol. Geotherm. Res., 2008, 172(3-4), 244-25810.1016/j.jvolgeores.2007.12.011Search in Google Scholar
 Rosi M., Principe C., Vecci R., The 1631 Vesuvius eruption. A reconstruction based on historical and stratigraphical data. J. Volcanol. Geotherm. Res., 1993, 58, 151-18210.1016/0377-0273(93)90106-2Search in Google Scholar
 Principe C., Marini L., Evolution of the Vesuvius magmatic-hydrothermal system before the 16 December 1631 eruption. J. Volcanol. Geotherm. Res., 2008, 171, 301-30610.1016/j.jvolgeores.2007.12.004Search in Google Scholar
 Principe C., The 1631 eruption of Vesuvius: Volcanological concepts in Italy at the beginning of the XVII century. In: Morello N. (Ed.), Volcanoes and History. Proceedings of the 20th INHIGEO Symposium, Naples, Aeolian Islands, Catania, September 19-25 1995, 525-54, Brigatti Editore, Genova, 1998Search in Google Scholar
 Paolillo A., Principe C., Bisson M., Gianardi R., Giordano D., La Felice S., Volcanology of the South-Western sector of Vesuvius, Italy. J. Maps, 2016, 12, 425-440, DOI 10.1080/17445647.2016.1234982DOI 10.1080/17445647.2016.1234982Search in Google Scholar
 Stormer J.C., Pierson M.L., Tacker R.C., Variation of F and ClX-ray intensity due to anisotropic diffusion in apatite during electron microprobe analysis. Am. Mineral., 1993, 78, 641-648Search in Google Scholar
 Greenwood J.P., Itoh S., Sakamoto N., Warren P., Taylor L., Yurimoto H., Hydrogen isotope ratios in lunar rocks indicate delivery of cometary water to the Moon. Nat. Geosci., 2011, 4, 79-8210.1038/ngeo1050Search in Google Scholar
 McCubbin F.M., Steele A., Hauri E.H., Nekvasil H., Yamashita S., Hemley R.J., Nominally hydrous magmatism on the Moon. Proc. Natl. Acad. Sci. U. S. A., 2010, 107(25), 11223-1122810.1073/pnas.1006677107Search in Google Scholar PubMed PubMed Central
 Piccoli P., Candela P., Apatitein felsic rocks:amodel for the estimation of initial halogen concentration in the Bishop Tuf (Long Valley), and Tuolumne intrusive suite (Sierra Nevada Batholith) magmas. Am. J. Sci., 1994, 294, 92-13510.2475/ajs.294.1.92Search in Google Scholar
 Liu Y., Stoppa F., Tonucci L., Mingsheng P., Study on 1H NMR Spectroscopy of Fluor-hydroxylapatite. J. Chin. Ceram. Soc., 2002, 30, 42-44.Search in Google Scholar
 Liu Y., Comodi P., Sassi P., Vibrational spectroscopic investigation of phosphate tetrahedron in fluor-, hydroxy-, and chlorapatites. N. Jb. Miner. Abh., 1998, 174(2), 211-222, doi:10.1127/njma/174/1998/211doi:10.1127/njma/174/1998/211Search in Google Scholar
 Comodi P., Liu Y., Stoppa F., Woolley A.R., A multi-method analysis of Si-, S- and REE-rich apatite from a new find of kalsilite-bearing leucitite (Abruzzi, Italy). Mineral. Mag., 1999, 63(5), 661-67210.1180/002646199548826Search in Google Scholar
 Stoppa F., Liu Y., Chemical composition and petrogenetic implications of apatites from some ultra-alkaline Italian rocks. Eur. J. Mineral., 1995, 7, 391-40210.1127/ejm/7/2/0391Search in Google Scholar
 Pascal M.-.L, Di Muro A., Fonteilles M., Principe C., Zirconolite and calzirtite in banded forsterite-spinel-calcite skarn ejecta from the 1631 eruption of Vesuvius: inferences for magma-wallrock interactions. Mineral. Mag., 2009, 73(2), 333-35610.1180/minmag.2009.073.2.333Search in Google Scholar
 Pascal M.-L., Fonteilles M., Boudouma O., Principe C., Qandilite from Vesuvius skarn ejecta: conditions of formation and miscibility gap in the ternary Spinel-Qandalite-Magnesioferrite. Can. Mineral., 2011, 49(2), 459-48510.3749/canmin.49.2.459Search in Google Scholar
 Kuno H., Yamasaki K., Iida C., Nagashima K., Differentiation of Hawaii magmas. Jpn. J. Geol. Geogr., 1957, 28, 179-218Search in Google Scholar
 de La Roche H., Classification et nomenclature des roches ignées: un essai de restauration de la convergence entre systématique quantitative, typologie d’usage et modélisation génétique. Bull. Soc. Geol. Fr., 1986, 2, 337-35310.2113/gssgfbull.II.2.337Search in Google Scholar
 Wilson B.M., Igneous petrogenesis a global tectonic approach. Spinger, 2007Search in Google Scholar
 Stoppa F., Schiazza M., An overview of monogenetic carbonatitic magmatism from Uganda, Italy, China and Spain: Vol-canological and geochemical features. J. South Am. Earth Sci., 2013, 41, 140-15910.1016/j.jsames.2012.10.004Search in Google Scholar
 Gasperini D., Bosch D., Braga R., Bondi M., Macera P., Morten L., Ultramafic xenoliths from the Veneto Volcanic Province (Italy): Petrological and geochemical evidence for multiple metasomatism of the SE Alps mantle lithosphere. Geochem. J., 2006, 40, 377-40410.2343/geochemj.40.377Search in Google Scholar
 Bianchini G., Clocchiatti R., Coltorti M., Joron J.L., Vaccaro C., Petrogenesis of mafic lavas from the northernmost sector of the Iblean District (Sicily). Eur. J. Mineral., 1998, 10, 301-31510.1127/ejm/10/2/0301Search in Google Scholar
 Gasperini D., Blichert-Toft J., Bosch D., Del Moro A., Macera P., Albarede F., Upwelling of deep mantle material through a plate window: evidence from the geochemistry of Italian basaltic volcanics. J. Geophys. Res., 2002, 47, 30-3810.1029/2001JB000418Search in Google Scholar
 Pichavant M., Scaillet B., Pommier A., Iacono-Marziano G., Cioni R., Nature and evolution of primitive Vesuvius magmas: An experimental study. J. Petrol., 2014, 55(11), 2281-231010.1093/petrology/egu057Search in Google Scholar
 Beattie P.D., Ford C.E., Russell D.G., Partition coefficients for olivine-melt and orthopyroxene-melt systems. Contrib. Mineral. Petrol., 1991, 109(2), 212-22410.1007/BF00306480Search in Google Scholar
 Cigolini C., Thermobarometry of phlogopite-bearing dunitic enclaves from Mount Vesuvius: preliminary estimates. Atti R. Ac-cad. Sci. Torino, Cl. Sci. Fis., Mat. Nat., 1997, 131, 33-56Search in Google Scholar
 Nisbet E.G., Cheadle M.J., Arndt N.T., Bickle M.J., Constraining the potential temperature of the Archaean mantle: A review of the evidence from komatiites. Lithos, 1993, 30, 291-30710.1016/0024-4937(93)90042-BSearch in Google Scholar
 Cundari A., Ferguson A.J., Petrogenetic relationships between melilitite and lamproite in the Roman Comagmatic Region: the lavas of S. Venanzo and Cupaello. Contrib. Mineral. Petrol., 1991, 107(3), 343-35710.1007/BF00325103Search in Google Scholar
 Normand C., Williams-Jones A.E., Physicochemical conditions and timing of rodingite formation: evidence from rodingite-hosted fluid inclusions in the JM Asbestos mine, Asbestos, Québec. Geochem. Trans., 2007, 8, doi:10.1186/1467-4866-8-11doi:10.1186/1467-4866-8-11Search in Google Scholar
 Ehlers E.G., The interpretation of geological phase diagrams. Freeman, W.H., & Co, San Francisco, 1972Search in Google Scholar
 Cortini M.A., Lima A., Devivo B., Trapping temperatures of melt inclusions from ejected Vesuvian mafic xenoliths. J. Volcanol. Geotherm. Res., 1985, 26, 167-17210.1016/0377-0273(85)90051-4Search in Google Scholar
 Marini L., Chiappini V., Cioni R., Cortecci G., Dinelli E., Principe C., Ferrara G., Effect of degassing on sulfur contents and δ34S values in Somma-Vesuvius magmas. Bull. Volcanol., 1998, 60(3), 187-19410.1007/s004450050226Search in Google Scholar
 Romano C., Giordano D., Papale P., Mincione V., Dingwell D.B., Rosi M., The dry and hydrous viscosities of alkaline melts from Vesuvius and Phlegrean Fields. Chem. Geol., 2003, 202(1-2), 23-3810.1016/S0009-2541(03)00208-0Search in Google Scholar
 Brocchini D., Principe C., Castradori D., Laurenzi M.A., Gorla L., Quaternary evolution of the southern sector of the Campanian Plain and early Somma-Vesuvius activity: insights from the Tre-case 1 well. Mineral. Petrol., 2001, 73, 67-9110.1007/s007100170011Search in Google Scholar
 Petrelli M., Poli G., Perugini D., Peccerillo A., Petrograph: a New Software to Visualize, Model, and Present Geochemical Data in Igneous Petrology. Geochem. Geophys. Geosyst., 2005, 6, Q07011, doi: 10.1029/2005GC000932doi: 10.1029/2005GC000932Search in Google Scholar
 Boudreau A.E., PELE - a version of the MELTS software program for the PC platform. Comput. Geosci., 1999, 25, 201-203, https://nicholas.duke.edu/people/faculty/boudreau/ DownLoads.html10.1016/S0098-3004(98)00117-4Search in Google Scholar
 Chiodini G., Marini L., Russo M., Geochemical evidence for the existence of high-temperature hydrothermal brines at Vesuvio volcano, Italy. Geochim. Cosmochim. Acta, 2001, 65, 2129-214710.1016/S0016-7037(01)00583-XSearch in Google Scholar
 Gambardella B., Cardellini C., Chiodini G., Frondini F., Marini L., Ottonello G., Vetuschi Zoccolini M., Fluxes of deep CO2 in the volcanic areas of central-southern Italy. J. Volcanol. Geotherm. Res., 2004, 136, 31-5210.1016/j.jvolgeores.2004.03.018Search in Google Scholar
 Rogie J.D., Kerrick D.M., Chiodini G., Frondini F., Flux measurements of non volcanic CO2 emission from some vents in central Italy. J. Geophys. Res., 2000, 105(B4), 8435-844510.1029/1999JB900430Search in Google Scholar
 Rittmann A., Die geologisch bedingte Evolution und Differentiation des Somma-Vesuv magmas. Zeit. Vulkan., 1933, 15(1/2), 8-94Search in Google Scholar
 Lavecchia G., Stoppa, F., The tectonic significance of Italian magmatism: an alternative view to the popular interpretation. Terra Nova, 1996, 8, 435-44610.1111/j.1365-3121.1996.tb00769.xSearch in Google Scholar
 Wiesmaier S., Trol V.R., Carracedo J.C., Ellam R.M., Bindeman I., Wolf J.A., Bimodality of lavas in the Teide-Pico Viejo succession in Tenerife: The role of crustal melting in the origin of recent phonolites. J. Petrol., 2012, 53, 2465-249510.1093/petrology/egs056Search in Google Scholar
 Watkinson D.H., Wyllie P.J., Phase equilibrium studies bearing on the limestone assimilation hypothesis. Geol. Soc. Am. Bull., 1969, 80(8), 1565-157610.1130/0016-7606(1969)80[1565:PESBOT]2.0.CO;2Search in Google Scholar
 Bailey D.K., Carbonate volcanics in Italy: numerical tests for the hypothesis of lava-sedimentary limestone mixing. Period. Mineral., 2005, 74(3), 205-208Search in Google Scholar
 Bell K., Kjarsgaard B., Discussion of Peccerillo (2004) “Carbonate-rich pyroclastic rocks from Central Apennines: carbonatites or carbonate-rich rocks? Period. Mineral., 2006, 75(1), 85-92Search in Google Scholar
 Woolley R.A., Bailey D.K., Castorina F., Rosatelli G., Stoppa F., Wall F., Reply to: “Carbonate-rich pyroclastic rocks from central Apennines: carbonatites or carbonated rocks? A commentary”. A. Peccerillo. Period. Mineral., 2005, 74(3), 183-194Search in Google Scholar
 Iacono Marziano G., Gaillard F., Scaillet B., Pichavant M., Chiodini G., Role of non-mantle CO2 in the dynamics of volcano degassing: The Mount Vesuvius example. Geology, 2009, 37, 319-322, doi: 10.1130/G25446A.1doi: 10.1130/G25446A.1Search in Google Scholar
 Cigolini C., Petrography and thermobarometry of high-pressure ultramafic ejecta from Mount Vesuvius, Italy: inferences on the deep feeding system. Period. Mineral., 2007, 76(2-3), 5-24Search in Google Scholar
 Milia A., Torrente M.M., The possible role of extensional faults in localizing magmatic activity: a crustal model for the Campa-nian Volcanic Zone (eastern Tyrrhenian Sea, Italy). J. Geol. Soc. (London, U.K.), 2011, 168, 471-48410.1144/0016-76492010-121Search in Google Scholar
 Chiarabba C., Jovane L., DiStefano R., A new view of Italian seismicity using 20 years of instrumental recordings. Tectono-physics, 2005, 395, 251-26810.1016/j.tecto.2004.09.013Search in Google Scholar
 Sun S., McDonought W.F., Chemical and isotopic systematic in ocean basalt: implication for mantle composition and processes. In: Saunders A.D., Norry M.J. (Eds.), Magmatism in the Ocean Basins, Geological Society, 1989, 42, 313-34510.1144/GSL.SP.1989.042.01.19Search in Google Scholar
 Cassinis R., Scarascia S., Lozej A., The deep crustal structure of Italy and surrounding areas from seismic refraction data; a new synthesis. Boll. Soc. Geol. Ital., 2003, 122 (3), 365-376Search in Google Scholar
Supplementary Material 1
Supplementary Material 2
© 2017 F. Stoppa et al.
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 3.0 License.