A New Magma Type in the Continental Collision Zone. The Case of Capraia Island (Tuscany, Italy)

The Tuscany Magmatic Province consists of a Miocene to Pleistocene association of a wide variety of rock types, including peraluminous crustal anatectic granites and rhyolites, calcalkaline and shoshonitic suites and ultrapotassic lamproites. In addition to the magma types already recognised, the occurrence of a new, distinct magma type at Capraia and Elba islands and in mafic enclaves in the San Vincenzo rhyolites has been suggested by recent studies. This particular type of magma, represented by intermediate to acidic calcalkaline rocks showing high Sr, Ba, and LREE, is restricted to the northwestern sector of the province and to a time interval of about 8 to 4.5 Ma. New data obtained on rocks from Capraia Island have allowed for the verification of the occurrence of this new magma type, the exploration of its origin and a discussion of its possible geodynamic significance. The high-Sr-Ba andesite-dacite rocks occurring in the Laghetto area at Capraia display a composition that is intermediate between adakitic and calcalkaline rocks. It is suggested that they represent a distinct type of magma that originated at mantle pressure by melting of the lower continental crust, followed by mixing with other Capraia magmas. The geodynamic model that best explains the composition of the studied rocks is the thickening of the continental crust during continental collision, followed by extension that favoured melting of the lower crust.


Introduction
The Tuscany Magmatic Province (TMP) is one of the most extensively studied and best known magmatic worldwide areas. It consists of Miocene to Pleistocene (about 8.5 Ma to 300 ka) intrusive and effusive rocks, cropping out in the Tuscan Archipelago, as seamount on the northern Tyrrhenian Sea floor, and at several places in southern Tuscany and northern Latium. In addition, a small peralkaline mafic sill, occurring at Sisco (Corsica) and showing an older age of 14.5 Ma [1], is considered as an early part of the Tuscany Province, essentially because of its petrogenic lamproitic affinity close to some occurrences from southern Tuscany (Orciatico, Montecatini Val di Cecina and Torre Alfina; e.g., [2] and references therein).
Petrological and geochemical studies recognised the occurrence of several types of primary magmas, both of mantle and crustal origin. Mantle-derived magmas show a range of mafic compositions, from calcalkaline and shoshonitic to ultrapotassic lamproitic; crustal magmas consist of peraluminous granites and rhyolites. A wide variety of hybrid compositions results from the mixing of different types of mantle-derived melts as well as between these and crust-derived end-members [3].
Despite the high number of existing studies on the Tuscan Magmatic Province, there are still several petrological, volcanological and geodynamic aspects that have been not adequately considered in the past and that deserve further attention. In particular, the number of end-member magma compositions occurring in the area and their significance for the structural evolution of the northern Apennines and back-arc area deserves clarification. Such a subject is also of interest to other complex orogenic settings. Based on the available data, CA-SHO and LMP mafic end-members were likely formed by melting of anomalous mantle rocks contaminated by variable amounts of subducted sediments (see [7][8][9] and others). The compositions falling in between CA and LMP mafic magmas might result from mixing between the two end-members. An alternative hypothesis is that the entire range of mafic magmas is primary and was generated in a mantle source containing variable proportions of veins and geochemically barren mantle rocks entering into the melt (e.g., [10]). oldest volcanic centre in the Tuscan Magmatic Province, and its activity is slightly younger to contemporaneous to the magmatism of Elba.
Aldighieri et al. [4] reconstructed the volcanic evolution of the island (Figure 1). They grouped the deposits of the first phase of activity (Capraia Supersynthem) in five different synthems. The deposits of the second activity phase were included in the Punta dello Zenobito Synthem. Accordingly, it is possible to distinguish the following: (a) the oldest volcanic products (lava flows, breccia and scoria) represented by the Punta del Trattoio Synthem (southwestern cliff, not visible on the map in Figure 1); (b) the andesitic domes and deposits (block, ash and pyroclastic flow) belonging to the Monte Rosso Synthem, visible along the western sector of the island; (c) the thick lava domes associated with lava flows and pyroclastics of Monte Campanile Synthem on the eastern side of the island; (d) lava flows and breccias of Monte Castello Synthem, which cover most of the island; (e) the lava flows forming the Monte Rucitello Synthem, which partially fill a large depression due to a flank collapse of the eastern part of the Monte Castello Synthem. In the same period, a new flank collapse, possibly due to the intrusion of a cryptodome, occurred toward the east; lava flows, breccias, scoriae and small lava domes were emplaced in the northern sector of the island as well as two lava domes in the central area. After a long quiescence period, the eruptive activity resumed in the southern area of the island (Punta dello Zenobito Synthem).
The Capraia stratovolcano is predominantly composed of massive high-K calcalkaline andesite and dacite lavas, with minor ash and block flows and subordinate pyroclastic flows, emplaced during the first phase of activity. The second eruptive phase mostly consists of strombolian scoriae and a few lava flows and dykes of Punta dello Zenobito centre which show a shoshonitic basaltic composition [25,26,28].
Prosperini [28] and Poli et al. [24] distinguished six different groups of rocks: (1) Older Series, belonging to the main and older emission stage, whose lava flows cover the largest part of the island; (2) Monte Campanile and Porto volcanites, cropping out in the central part of the eastern coast; (3) Zurletto pyroclastics, probably originating from an external submerged centre, which crop out in the eastern coast; (4) San Rocco and Piano volcanites, erupted by small eruptive centres, disposed in a NE-SW direction; (5) Laghetto volcanites, cropping out in a narrow depressed area on the top of the volcano; (6) Zenobito volcanites, in the southern part of the island, forming a neck of porphyritic lava surrounded by stratified scoria deposits. A petrographic and geochemical overview of the Capraia lavas is reported in Prosperini [28] and Poli and Perugini [25].
The bulk of the rocks making up the stratovolcano have mostly a high-K andesitedacite composition with a slight enrichment in potassium (K 2 O = 2.5-4.0 wt%). The basaltic-andesites from Punta dello Zenobito display moderate concentrations of CaO (around 7-8 wt%) and Na 2 O (around 3 wt%) and variable MgO, Ni, and Cr content.

Materials and Methods
Among the Capraia products, a group of rocks shows Sr, Ba and LREE contents distinctively higher than other rocks with similar silica contents [3,6]. They come from the "Laghetto volcanic group" [24,28]. This group has been sampled with the aim of providing further data and investigating the origin of high Sr-Ba rocks.
In the Laghetto area ( Figure 2), a series of lava flows crops out in a small depression. Several samples have been collected from various lava levels and have been analysed for major and trace elements. Mineral chemistry and Sr-Nd radiogenic isotope analyses have been carried out for selected samples. tinctively higher than other rocks with similar silica contents [3,6]. They come from the "Laghetto volcanic group" [24,28]. This group has been sampled with the aim of providing further data and investigating the origin of high Sr-Ba rocks.
In the Laghetto area ( Figure 2), a series of lava flows crops out in a small depression. Several samples have been collected from various lava levels and have been analysed for major and trace elements. Mineral chemistry and Sr-Nd radiogenic isotope analyses have been carried out for selected samples.

Analytical Methods
The major element composition of selected mineral phases has been obtained on thin polished sections (30 µm) through the Electron Microprobe JEOL JXA-8600 (IGG-CNR of Firenze). The instrument is equipped with four WDS spectrometers and an EDS detector; it is controlled by the software XMas. The system operates at 10 nA beam current and 15 kV accelerating voltage, with counting times of 15 s on the peak (10 s for Na) and of 5 s on each background. The data acquired have been processed using the PAP correction system [29].
Major and trace element analyses of the collected rock samples were carried out at the Activation Laboratories, Ancaster, Ontario, Canada. For major oxides and selected trace elements (Ba,Sr,Y,Zr,Sc,V), the samples were prepared and analysed in a batch system. Each batch contained a method reagent blank, certified reference material and 17% replicates. Samples were mixed with a flux of lithium metaborate and lithium tetraborate and fused in an induction furnace. The molten melt was immediately poured into a solution of 5% nitric acid containing an internal standard and mixed continuously until completely dissolved (~30 min). The samples were run on a combination simultaneous/sequential Thermo Jarrell-Ash ENVIRO II ICP (inductively coupled plasma). Calibration was performed using 7 prepared USGS and CANMET certified reference materials. One of the 7 standards was used during the analysis for every group of ten samples. Detection

Analytical Methods
The major element composition of selected mineral phases has been obtained on thin polished sections (30 µm) through the Electron Microprobe JEOL JXA-8600 (IGG-CNR of Firenze). The instrument is equipped with four WDS spectrometers and an EDS detector; it is controlled by the software XMas. The system operates at 10 nA beam current and 15 kV accelerating voltage, with counting times of 15 s on the peak (10 s for Na) and of 5 s on each background. The data acquired have been processed using the PAP correction system [29].
Major and trace element analyses of the collected rock samples were carried out at the Activation Laboratories, Ancaster, Ontario, Canada. For major oxides and selected trace elements (Ba,Sr,Y,Zr,Sc,V), the samples were prepared and analysed in a batch system. Each batch contained a method reagent blank, certified reference material and 17% replicates. Samples were mixed with a flux of lithium metaborate and lithium tetraborate and fused in an induction furnace. The molten melt was immediately poured into a solution of 5% nitric acid containing an internal standard and mixed continuously until completely dissolved (~30 min). The samples were run on a combination simultaneous/sequential Thermo Jarrell-Ash ENVIRO II ICP (inductively coupled plasma). Calibration was performed using 7 prepared USGS and CANMET certified reference materials. One of the 7 standards was used during the analysis for every group of ten samples. Detection limit was in the range between 0.01 and 0.001 (%) for major elements and between 1 and 5 (ppm) for trace elements. The same sample solution was spiked with internal standards to cover the entire mass range and was further diluted and introduced into a Perkin Elmer SCIEX ELAN 6000, 6100 or 9000 ICP/MS (inductively coupled plasma/mass spectrometry) using a proprietary sample introduction methodology. The detection limits in ppm were: Pb = 5 ppm; Ni, Co, Cr, Rb = 1 ppm; Nb= 0.2; Cs, Hf, Ta = 0.1; La, Ce, Nd, Eu, Th = 0.05; Lu = 0.02; Pr, Sm, Gd, Tb, Dy, Ho, Eb, Yb, U = 0.01; Eu, Tm = 0.005.
Radiogenic isotope ratio measurements were performed at IGG-CNR of Pisa (Italy). Sr and REE fractions were separated from the matrix using conventional cation columns with AG 50W-X8 200-400 mesh resin. Strontium separates were then loaded onto 99.98% Re filaments, and isotope ratios were measured using a Finnigan MAT 262 multiple collector thermal ionisation mass spectrometer (TIMS) for running in dynamic mode. Instrumental mass bias correction was done by internal normalisation to 86 Sr/ 88 Sr = 0.1194. Replicate measurements of NIST SRM 987 (SrCO 3 ) standard gave an average value of 0.710229 ± 11 (2SD, n = 16). Throughout the full chemical process, the Sr blanks were approximately 0.3 ng, which were negligible for the analysed samples (0.1-0.2 g of sample). Nd vas extracted from REE eluates using Eichrom Ln resin. Nd eluates were dried, then redissolved in 2% HNO 3 solution and measured in static mode with a Thermo Neptune Plus, MC-ICP-MS equipped with 10 11 and 10 12 Ohm resistors. Blank was negligible; thus, no blank correction was performed. 144 Sm interference over 144 Nd was calculated using 147 Sm; however, this correction also had also a negligible effect on the final ratio. Instrumental mass bias correction was done by internal normalis ation to 146 Nd/ 144 Nd = 0.7219. Seventeen replicates of the standard J Nd Replicate analyses of JNdi-1 standard [30] gave an average value of 0.512098 ± 6 (2SD).

Petrography and Mineral Chemistry
The collected samples show hypocrystalline, moderately porphyritic (Porphyritic Index = 20-30), textures, with phenocrysts of plagioclase, clinopyroxene, and biotite; amphibole, K-feldspar, opaque minerals and apatite are present in accessory amounts. The micro-to crypto-crystalline groundmass consists of the same mineral phases as the phenocrysts, with abundant microlites of plagioclase and biotite.
The composition of the main mineral phases is reported in Table 1. Plagioclase is the most abundant mineral (40-65% of total phenocryst abundance) and consists of euhedral or subhedral twinned crystals; in a few cases, phenocrysts display a sieved mantle. Both phenocryst core (An [37][38][39][40][41][42][43][44][45][46][47][48][49] ) and rim (An [44][45][46][47] ) have andesinic composition. The microphenocrysts (An  ) and groundmass (An 40-64 ) plagioclase span a larger range of compositions, from poorer to richer in Anorthite. Clinopyroxene is ubiquitous; it appears as colourless or lightly brown-coloured euhedral to rounded crystals, showing at times reaction rims. Phenocryst core composition is in the range Wo 43-45 , En 35-50 , Fs 5-21 ; zoning is generally slightly normal (more Mg-rich cores), even though, in a few cases, crystals with reverse zoning are observed. Some microphenocrysts and groundmass clinopyroxene show disequilibrium evidence, having higher Enstatite content than phenocrysts, in agreement with the higher Anorthite content in plagioclase microliths. When plotted in the quadrilateral classification diagram [31], the analysed pyroxenes fall in the augite field or on the boundary between the augite and diopside fields; they form two slightly distinct groups (Figure 3), each including phenocrysts, microphenocrysts, and groundmass clinopyroxene. Biotite phenocryst may display a ragged habit and variable extent of resorption; both biotite and amphibole often display opacitic rims or are almost entirely transformed to opaque minerals.

Geochemistry
Major and trace element compositions of the analysed samples are reported i 2, along with a few representative analyses of other Capraia rocks [9,28]

Geochemistry
Major and trace element compositions of the analysed samples are reported in Table 2, along with a few representative analyses of other Capraia rocks [9,28]. The Total-Alkali versus Silica (TAS) [32] and K 2 O vs. SiO 2 [33] classification diagrams are shown in Figure 4. Major and trace element diagrams are shown in Figures 5-7. Other rocks occurring at Capraia, as dykes at Elba Island and as mafic enclaves in the San Vincenzo rhyolites, are also reported [15,16,34,35].
The Capraia rocks are mostly high-K calcalkaline andesites and dacites. The samples from Punta dello Zenobito are more mafic and fall at the boundary between shoshonitic and high-K basaltic andesites. The Zurletto pyroclastics display more evolved composition plotting between dacite and rhyolite fields ( Figure 4b).
The investigated samples from the Laghetto Series plot at the boundary between the trackyandesite and trachydacite fields in the TAS diagram ( Figure 4a) and mostly in the high-K calcalkaline andesite and dacite fields in the K 2 O vs. SiO 2 diagram (Figure 4b).
The variation diagrams for major elements (Figures 4 and 5) show a similarity of the analysed rocks with other andesite-dacites from Capraia for most elements, although they exhibit slightly higher K 2 O and lower TiO 2 and FeO t .
Trace  Open circles (first activity phase) and triangles (second activity phase): data from the literature [9,28]; full green circles: this work. Some Elba dykes [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. Data are plotted on water-free basis.  Open circles (first activity phase) and triangles (second activity phase): data from the literature [9,28]; full green circles: this work. Some Elba dykes [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. Data are plotted on water-free basis. Figure 4. (a) TAS and (b) K2O vs. SiO2 classification diagrams for Capraia rocks. Open circles (first activity phase) and triangles (second activity phase): data from the literature [9,28]; full green circles: this work. Some Elba dykes [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. Data are plotted on water-free basis.            [9,28], Elba dykes [15,35] and a San Vincenzo mafic enclave [16] are shown for comparison. Normalising values from Sun and McDonough [36].

Classification, Occurrence and Genetic Processes of Sr-Ba-LREE-Rich Rocks
The analysed rocks of the Laghetto Series differ from other intermediate calcalkaline rocks from Capraia and most of the other centres of the Tuscany Magmatic Province for showing anomalous enrichments in Sr and Ba [3]. Moreover, they also show enrichment in LREE, Th, U, and Pb, and strong REE fractionation (high La/Yb and La/Y). A few dykes from Elba Island (Table 2) also show similar characteristics to the Laghetto samples [15,35].
Intermediate to felsic high-Sr rocks are found in many recent and old subduction zones and have been named adakites. They were interpreted early on as representing magmas formed directly by melting of MOR basaltic crust at high pressures along subduction zones ( [17,18] and references therein).
The analysed Laghetto rocks share several compositional features of adakites, as defined by Richard and Kerrich [38] and Castillo [19]. These include high Sr, Sr/Y, Ba and La/Yb (Figures 7 and 10). On the other side, Y contents are also somewhat higher in the Capraia rocks with respect to typical adakites, although low Y and HREE are found at Elba (Figure 10). Sr (and Nd) isotopic ratios differ from typical island-arc adakite values ( 87 Sr/ 86 Sr ~ 0.704) [38]; however, they are close to some Yang Tse Block and Northern Tibet adakite values (Figure 9). Based on this evidence, the Laghetto rocks can be considered as  Figure 9. Nd vs. Sr isotope ratios for Laghetto rocks. In grey is shown the Sr and Nd isotope area of the Tuscan province. Data from other adakite localities (orange areas) are reported for comparison [19].

Classification, Occurrence and Genetic Processes of Sr-Ba-LREE-Rich Rocks
The analysed rocks of the Laghetto Series differ from other intermediate calcalkaline rocks from Capraia and most of the other centres of the Tuscany Magmatic Province for showing anomalous enrichments in Sr and Ba [3]. Moreover, they also show enrichment in LREE, Th, U, and Pb, and strong REE fractionation (high La/Yb and La/Y). A few dykes from Elba Island (Table 2) also show similar characteristics to the Laghetto samples [15,35].
Intermediate to felsic high-Sr rocks are found in many recent and old subduction zones and have been named adakites. They were interpreted early on as representing magmas formed directly by melting of MOR basaltic crust at high pressures along subduction zones ( [17,18] and references therein).
The analysed Laghetto rocks share several compositional features of adakites, as defined by Richard and Kerrich [38] and Castillo [19]. These include high Sr, Sr/Y, Ba and La/Yb (Figures 7 and 10). On the other side, Y contents are also somewhat higher in the Capraia rocks with respect to typical adakites, although low Y and HREE are found at Elba (Figure 10). Sr (and Nd) isotopic ratios differ from typical island-arc adakite values ( 87 Sr/ 86 Sr~0.704) [38]; however, they are close to some Yang Tse Block and Northern Tibet adakite values (Figure 9). Based on this evidence, the Laghetto rocks can be considered as members of the adakitic rock clan or as transitional rocks between adakitic and arc calcalkaline rocks. They represent a new magma type for Central Italy, which had not been highlighted by previous studies. Similar rocks occur in at least one other place in Tuscany: at Elba, a few dykes from Chiessi (Orano dykes) have similar characteristics as Laghetto, in addition to lower abundances of Y and HREE [35]. Some mafic enclaves from San Vincenzo rhyolites, southern Tuscany, likely also belong to the same magma type, as will be discussed later.
Geosciences 2020, 10, x FOR PEER REVIEW in addition to lower abundances of Y and HREE [35]. Some mafic enclaves from S cenzo rhyolites, southern Tuscany, likely also belong to the same magma type, as discussed later.
There is a general agreement that the high Sr-Ba contents of adakitic rocks genetic processes that occurred at high pressure, outside the stability field of feld In addition to the early hypothesis of a direct derivation by partial meltin MORB-type crust along subduction zones [17], other processes include high-press phibole-dominated crystal fractionation of arc hydrous basalts, a derivation from lous mantle rocks contaminated by upper crustal material coming from the slab, ing of thickened lower continental crust in zones of continental collision (e.g., [18 44]). In all cases, Sr and Ba behave as incompatible elements and, therefore, are e in the melts. Figure 10. Y vs Sr/Y diagram to discriminate adakitic from typical arc rocks. Fields from R al. [20]. Dashed lines are two mixing curves between Elba dykes (Chiessi) and different Ca samples. Note that the samples with high Sr/Y plot on these curves, and the Laghetto rock the transition between adakite and calcalkaline rocks.

Petrogenesis of Laghetto Rocks
Sharp spikes of some elements in the Laghetto rocks might suggest a shallo magma chamber process, for instance, the accumulation of Sr-Ba-LREE-rich phas as feldspars, biotite, and accessory apatite. All feldspars contain high concentratio K-feldspars and biotite are enriched in Ba (e.g., [45,46]); apatite contains high amo all REE (e.g., [47]). The analysed samples contain apatite and biotite. There is a correlation between P2O5 and LREE (not shown), which would support accumul apatite. However, apatite is rich in both Light and Heavy REE (e.g., [48]) and shou duce an increase of all REE, which is not observed. Also, accumulation of phl should lead to exceedingly high Rb contents, which is not observed in our rocks. ver, the porphyritic index of Laghetto rocks is moderate (about 20%) and much low other intermediate rocks from the same island [28]. Therefore, mineral accumul unlikely to be responsible for anomalous enrichment in the investigated rocks, b petrographic and geochemical evidence. Their anomalous composition as comp other andesite-dacite rocks in Tuscany must reflect processes occurring during There is a general agreement that the high Sr-Ba contents of adakitic rocks point to genetic processes that occurred at high pressure, outside the stability field of feldspars.
In addition to the early hypothesis of a direct derivation by partial melting of the MORBtype crust along subduction zones [17], other processes include high-pressure amphiboledominated crystal fractionation of arc hydrous basalts, a derivation from anomalous mantle rocks contaminated by upper crustal material coming from the slab, or melting of thickened lower continental crust in zones of continental collision (e.g., [18][19][20][37][38][39][40][41][42][43][44]). In all cases, Sr and Ba behave as incompatible elements and, therefore, are enriched in the melts.

Petrogenesis of Laghetto Rocks
Sharp spikes of some elements in the Laghetto rocks might suggest a shallow-level magma chamber process, for instance, the accumulation of Sr-Ba-LREE-rich phases such as feldspars, biotite, and accessory apatite. All feldspars contain high concentrations of Sr; K-feldspars and biotite are enriched in Ba (e.g., [45,46]); apatite contains high amounts of all REE (e.g., [47]). The analysed samples contain apatite and biotite. There is a positive correlation between P 2 O 5 and LREE (not shown), which would support accumulation of apatite. However, apatite is rich in both Light and Heavy REE (e.g., [48]) and should produce an increase of all REE, which is not observed. Also, accumulation of phlogopite should lead to exceedingly high Rb contents, which is not observed in our rocks. Moreover, the porphyritic index of Laghetto rocks is moderate (about 20%) and much lower than other intermediate rocks from the same island [28]. Therefore, mineral accumulation is unlikely to be responsible for anomalous enrichment in the investigated rocks, based on petrographic and geochemical evidence. Their anomalous composition as compared to other andesite-dacite rocks in Tuscany must reflect processes occurring during melting or/and during the ascent of magma to the surface. Clear evidence emerging from isotopic data is that the Laghetto rocks differ from typical MORB rocks. Moreover, the Sr-Nd isotope signature clearly requires the participation of both mantle and crustal components in its origin and/or evolution. This excludes any direct genetic relationship with mid-ocean ridge basalts (MORB), e.g., by melting of the subducted oceanic slab. These conclusions are valid for most, if not all, Tuscany magmatic rocks (e.g., [3]). Therefore, an origin from melting of the subducting oceanic basaltic slab is unlikely. Other possible processes include high-pressure fractional crystallisation of basaltic magmas, melting of mixtures of crustal and mantle rocks above subduction zones (subduction mélange bodies), and melting of the lower continental crust, thickened during collision. These processes will be tested by trace element modelling.
High-pressure fractional crystallisation. Such a process has been tested by assuming a calcalkaline-shoshonitic mafic magma with a composition as the rocks from Punta dello Zenobito. Models have been carried out for both compatible and incompatible trace elements. Fractionation of feldspar-free mineral assemblages dominated by mafic minerals amphibole and/or clinopyroxene has been assumed. This mineral assemblage has a low bulk partition coefficient for Sr and Ba and other incompatible elements during fractionation and can explain the enrichments in the Laghetto rocks. However, amphiboles and clinopyroxenes also incorporate significant amounts of ferromagnesian elements, such as Cr, Co, and Sc, which have partition coefficients much higher than unity (see GERM website [49]). For modelling, Sr and Rb have been used as incompatible elements, whereas Cr has been selected as representative of compatible elements. Partition coefficients have been set at D Sr = 0.1, D Rb = 0.01, D Cr = 2 and 20; a very low partition coefficient for Rb has been assumed because this element is one of the most incompatible in both mafic and felsic systems. Partition coefficients for Cr are variable, and two extreme values are used for modelling (see GERM website [49]). The model of Cr vs. Sr (Figure 11a) shows that the separation of a mafic mineral assemblage from basaltic Capraia magmas generates high-Sr residual liquid as the Laghetto rocks, after about 50 to 70% crystal fractionation. However, the Cr contents are matched only if the minimum value of D Cr = 2 is assumed.
For higher and more realistic values of D Cr (= 5-10), the decrease in Cr is extreme and produces residual melts that are away from Laghetto rocks. This argues against fractional crystallisation of mafic phases. The same conclusion is suggested by Rb-Sr modelling (Figure 11b), which is unable to generate high-Sr and moderately Rb-rich residual liquids starting from Zenobito basalts unless the unrealistic assumption is made of a nearly compatible behaviour for Rb (i.e, D Rb = 0.9-1). A compatible behaviour of Rb would imply fractionation of either amphibole and/or phlogopite, which would not explain the increase in Ba.  [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. For further explanation, see text.
Lower crust melting. Alternative processes could be melting of the lower continental crust or a derivation from mixtures between mantle peridotite and upper crustal rocks, i.e., subduction mélanges of peridotite and metapelites [50]. Both occur at high pressure outside the stability field of feldspars (i.e., thickened crust), making Sr behave as an incompatible element. Lower crust batch melting processes of Sr vs. the compatible element Cr are reported in Figure 11c. The lower crust composition is from Wedepohl [51]. Modelling shows that, assuming an incompatible behaviour (i.e., few or no feldspars in the residue, compatible with high-P melting in the deep crust), Sr increases with a decreasing degree of melting, whereas compatible elements are not strongly depleted. Overall, a compositional trend that resembles the Laghetto rocks can be obtained. However, the adakitelike rocks from Elba and the San Vincenzo mafic enclave have exceedingly high Cr contents to be derived from the lower crust. Such a problem does not necessarily rule out lower crust melting. However, a contribution of mantle material has to be assumed in the origin of these rocks.
Crustal melting has been also tested using couples of incompatible elements (Sr vs. Rb). If partition coefficients of 0.1 and 0.01 are respectively assumed for Sr and Rb, crustal melting can generate liquids that resemble some of the most Sr-rich among the analysed samples but does not explain the entire range of the Laghetto rocks. However, the compositional range of the latter rocks can be matched by assuming that Sr-rich magmas  [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. For further explanation, see text.
Lower crust melting. Alternative processes could be melting of the lower continental crust or a derivation from mixtures between mantle peridotite and upper crustal rocks, i.e., subduction mélanges of peridotite and metapelites [50]. Both occur at high pressure outside the stability field of feldspars (i.e., thickened crust), making Sr behave as an incompatible element. Lower crust batch melting processes of Sr vs. the compatible element Cr are reported in Figure 11c. The lower crust composition is from Wedepohl [51]. Modelling shows that, assuming an incompatible behaviour (i.e., few or no feldspars in the residue, compatible with high-P melting in the deep crust), Sr increases with a decreasing degree of melting, whereas compatible elements are not strongly depleted. Overall, a compositional trend that resembles the Laghetto rocks can be obtained. However, the adakite-like rocks from Elba and the San Vincenzo mafic enclave have exceedingly high Cr contents to be derived from the lower crust. Such a problem does not necessarily rule out lower crust melting. However, a contribution of mantle material has to be assumed in the origin of these rocks.
Crustal melting has been also tested using couples of incompatible elements (Sr vs. Rb). If partition coefficients of 0.1 and 0.01 are respectively assumed for Sr and Rb, crustal melting can generate liquids that resemble some of the most Sr-rich among the analysed samples but does not explain the entire range of the Laghetto rocks. However, the compositional range of the latter rocks can be matched by assuming that Sr-rich magmas mixed with other Capraia liquids to form a suite of hybrid magmas matching the trend of the analysed samples. In conclusion, the most Sr-rich rocks from Capraia and Elba can be formed by melting of the lower crust. However, a contribution from upper mantle peridotite is required to obtain the observed values of compatible elements such as Cr.
Melting of mantle-upper crust mixtures (subduction mélange). The occurrence of upper mantle rocks strongly contaminated by upper crustal material in the Roman Province has been suggested by many papers ( [8] and others). A hybridism of mantle and crustal rocks is probably common along the entire Alpine-Himalayan orogenic belt [52]. Mantlecrust mixing can occur by sediment transport along subduction zones, melting and veining by these melts inside the overlying mantle wedge ( [4] and references therein). Other models assume that basalts, sediments, and slices of continental crust are transported along the subduction channel and are added to mantle wedge to form hydrous mélange bodies at the slab-mantle interface. The low density of these bodies favours ascent as plumes toward the hotter corner of the mantle wedge, where they dehydrate and melt (e.g., [50,[53][54][55]). The melting of different portions of diapirs generates various types of magmas, with variable enrichments in incompatible elements; melting of dehydrated sectors of the diapirs can generate magmas enriched in poorly mobile elements such as Th and REE [56]. These magmas can also be rich in Ba-Sr if melting occurs at high pressure, outside the stability field of feldspars.
Quantitative modelling of hybrid mantle-crust rocks melting is very difficult to perform because of the large number of assumptions that have to be made on the composition of the crustal rocks, the amounts of crustal material entering into the mantle, and the role of fluid in modifying different portions of the mantle wedge. Because of these difficulties, only some aspects will be explored.
As stated earlier in this paper, Sr-Nd isotopic signatures on the Laghetto rocks require contributions by both mantle and upper continental crust, as amply recognised for other mafic to intermediate magmas in the Tuscany province and along the Alpine-Himalayan belt. The amount of crustal material involved in the mixing with mantle is uncertain. If a composition as that reported for the average gneiss from Southern Tuscany is assumed (Sr = 180 ppm, 87 Sr/ 86 Sr = 0.723 [57]), the contribution of the continental crust end-members to the Laghetto source is around 5%, starting from an upper mantle composition of Sr = 20 ppm and 87 Sr/ 86 Sr = 0.703 (see GERM website [49]). A mixture of 95% mantle material and 5% crustal material is still ultramafic in composition. Its melting generates a liquid containing high contents of compatible elements Ni and Cr (Ni around 300-400 ppm and Cr around 1000-1100 ppm, assuming batch melting), which differ from the studied rocks.
Role of magma mixing. The Laghetto volcanics display evidence of textural and mineralogical disequilibrium, such as the sieve-texture in plagioclase, the reaction rim and variable composition of clinopyroxene and plagioclase (Table 1), and the resorption rims in biotite and amphibole, which suggest mixing between compositionally different magma batches. To highlight this process, plots of elements vs. element ratios (i.e., Rb vs. Rb/Sr) are reported in Figure 12. The analysed samples and other Capraia rocks define different trends that are typical of mixing processes. These trends are consistent with the hypothesis that a range of high-Sr and low Rb/Sr magmas was formed during lower crustal melting (dash-dot line). These subsequently mixed with low-Sr and high-Rb/Sr melts, which represent the bulk of Capraia rocks and might be derived from mafic parents following shallow-level fractional crystallisation processes. The circled symbols in Figure 12 represent the different end members. Figure 12. Element vs. element ratio diagrams for the Laghetto and other Capraia rocks. Some Elba dykes [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. The dash-dot line is the melting curve of the lower continental crust. The full lines are possible mixing trends between different end members (circled symbols).
In conclusion, the melting of the lower continental crust seems the most plausible genetic mechanism for the investigated rocks. These processes occurred at high pressure, probably at the base of a continental crust that was thickened during continental collision. Mixing between these magmas and other mantle-derived calcalkaline liquids, showing lower Sr contents, modified primary lower crustal melts, thus giving the range of compositions observed at Laghetto.

Regional Distribution and Geodynamic Significance
High-Sr-Ba andesites, similar to the analysed samples from Capraia, occur elsewhere in the Tuscany Magmatic Province. In particular, a group of dykes from Elba share many characteristics with the Laghetto rocks, but have lower Y and HREE contents [35] and can be also classified as adakitic rocks. The Elba and Capraia adakitic rocks have been erupted around 7-8 Ma and represent some of the oldest rocks of the Tuscany Magmatic Province. Other adakitic compositions might be represented by mafic enclaves in the San Vincenzo rhyolites. These also have high Sr contents, but the available data are insufficient to compare these enclaves with the Capraia and Elba high-Sr rocks [14,16,34]. The San Vincenzo rhyolites are sited on mainland Tuscany, east of Capraia and Elba. Their age is 4.5 Ma, younger to coeval to Elba and Capraia activity. Overall, the episode of adakitic-type magmatism seems to be restricted to the westernmost and oldest sector of the Tuscany Province.
There is a general agreement that the mafic-intermediate magmas in Tuscany are generated in a heterogeneous and anomalous upper mantle, modified by the addition of subduction-related upper crustal material such as pelitic sediments or their metamorphic equivalents. According to some authors [2,3,6,50], modification of the upper mantle occurred during southeast directed Alpine subduction of the European plate beneath the Adriatic-African margin.
Starting from the Oligocene, there was an inversion of the subduction, with the Adriatic-African plate immerging westward beneath the older collision zone. New subduction and back-arc extension induced melting of the previously contaminated mantle, which Capraia Isl. first activity phase San Vincenzo mafic enclave Capraia Isl. second activity phase Elba Isl. Orano porphyry Capraia Isl. Laghetto (this work) Elba Si. Chiessi dyke Figure 12. Element vs. element ratio diagrams for the Laghetto and other Capraia rocks. Some Elba dykes [15,35] and mafic enclaves from San Vincenzo [16,34] are also plotted for comparison. The dash-dot line is the melting curve of the lower continental crust. The full lines are possible mixing trends between different end members (circled symbols).
In conclusion, the melting of the lower continental crust seems the most plausible genetic mechanism for the investigated rocks. These processes occurred at high pressure, probably at the base of a continental crust that was thickened during continental collision. Mixing between these magmas and other mantle-derived calcalkaline liquids, showing lower Sr contents, modified primary lower crustal melts, thus giving the range of compositions observed at Laghetto.

Regional Distribution and Geodynamic Significance
High-Sr-Ba andesites, similar to the analysed samples from Capraia, occur elsewhere in the Tuscany Magmatic Province. In particular, a group of dykes from Elba share many characteristics with the Laghetto rocks, but have lower Y and HREE contents [35] and can be also classified as adakitic rocks. The Elba and Capraia adakitic rocks have been erupted around 7-8 Ma and represent some of the oldest rocks of the Tuscany Magmatic Province. Other adakitic compositions might be represented by mafic enclaves in the San Vincenzo rhyolites. These also have high Sr contents, but the available data are insufficient to compare these enclaves with the Capraia and Elba high-Sr rocks [14,16,34]. The San Vincenzo rhyolites are sited on mainland Tuscany, east of Capraia and Elba. Their age is 4.5 Ma, younger to coeval to Elba and Capraia activity. Overall, the episode of adakitic-type magmatism seems to be restricted to the westernmost and oldest sector of the Tuscany Province.
There is a general agreement that the mafic-intermediate magmas in Tuscany are generated in a heterogeneous and anomalous upper mantle, modified by the addition of subduction-related upper crustal material such as pelitic sediments or their metamorphic equivalents. According to some authors [2,3,6,50], modification of the upper mantle occurred during southeast directed Alpine subduction of the European plate beneath the Adriatic-African margin.
Starting from the Oligocene, there was an inversion of the subduction, with the Adriatic-African plate immerging westward beneath the older collision zone. New sub-duction and back-arc extension induced melting of the previously contaminated mantle, which gave several compositionally different magma types, ranging from calcalkaline to potassic lamproitic, according to the depth and degrees of partial melting.
Continental collision and inversion of subduction immersion were followed by eastward migration of the magmatic arc, along with back-arc extension. Collision probably generated a thickening of the continental crust, which was the site of melting and formation of adakitic-like magmas. Melting was favoured by the post-collisional extension, which migrated eastward, behind the subduction zone shifting in the same direction. Therefore, these episodes of magmatism occurred in the westernmost sectors where the crustal thickness was produced by the collision. By contrast, they are not present in other younger manifestations of the Tuscany magmatism, probably because back-arc extension reduced the crustal thickness of the continental crust.

Conclusions
• A series of high-Sr-Ba andesite-dacite rocks occur at Capraia in the Laghetto area, representing a particular stage of calcalkaline activity at~7 Ma. They have a composition that is intermediate between adakitic and calcalkaline rocks. • Geochemical data suggest that these rocks cannot be related to other calcalkaline rocks occurring on the island by fractional crystallisation but represent a distinct type of magma. • Geochemical modelling suggests that the most likely hypothesis is the generation at mantle pressure by melting of the lower continental crust, followed by mixing with other mantle-derived Capraia magmas. • A similar rock type also occurs at Elba Island and is almost coeval with those from Capraia and possibly at San Vincenzo, on the mainland Tuscany. In contrast, it seems to be absent in other sectors of the Tuscany magmatic province. • The geodynamic model that best explains the composition of the studied rocks is the thickening of the continental crust during a continental collision, followed by extension that favoured melting of the lower crust. Such a process was possible during the early stages of extension, shortly after the collision event. This explains why lower crustal melting and formation of adakitic-like magmas are restricted to older stages of magmatism in Tuscany.