Origin of Triassic magmatism of the Southern Alps ( Italy ) : Constraints from geochemistry and Sr-Nd-Pb isotopic ratios

During Middle-early Late Triassic (~243-235 Ma) a diffuse igneous activity developed in the Southern Alps (Italy). Sparse lava flow and pyroclastic succession remnants of such Southern Alps Triassic Igneous Rocks (SATIR) crop out in the Brescian pre-Alps, the Vicentinian Alps (Recoaro-Schio-Posina), Non Valley, Dolomites and Julian Alps. Plutonic rocks are found in two main plutonic complexes (Monzoni and Predazzo) and a small stock (Cima di Pape). Coeval igneous products can be traced eastward to Austria and Dinarides, for a total length of ~450 km. The coeval formation of major late Anisian-late Ladinian carbonate platforms in the subsiding eastern Dolomites, and significant uplift and subaerial erosion in the western Dolomites suggest, for these areas, the occurrence of large-scale strike-slip and extensional tectonics with the development of horst-and-graben structures. This study reports the first complete review of the SATIR activity, including new mineral chemical data on 14 samples, 61 major and trace element whole rock analyses and 7 Sr-Nd-Pb isotopic ratios for volcanic and plutonic samples from Dolomites (lavas plus Monzoni and Predazzo plutonic rocks) and Vicentinian Alps lavas. Despite the variable post-magmatic modifications, the large areal distribution of the products and their wide spectrum of chemical compositions, these samples show rather common geochemical and mineralogical characteristics and define major and trace element trends that can be associated with nearly close-system ACCEPTED MANUSCRIPT


A C C E P T E D M A N U S C R I P T
(237.3 ± 1.0 Ma from mechanically abraded zircon; Brack et al., 1997 and indicate an equivalent age for intrusive and effusive products. Lower Carnian magmatism is recorded also in the western Southern Alps (Brescian Prealps), with both effusive and intrusive products (Cassinis et al., 2008). A possible occurrence of coeval volcanic edifices more to the south is suggested by the occurrence of very common Lower Carnian volcaniclastic sandstones (Val Sabbia Ss., Garzanti, 1985;Jadoul et al., 2004;Cassinis et al., 2011).

Tectonic evolution of the Dolomites
The main tectonic events recorded in the Dolomites area can be resumed as follows: 1) Permian-Triassic strike-slip and extensional tectonics (details below); 2) nearly amagmatic Jurassic rifting associated with the opening of the Alpine Tethys Ocean; 3) Paleogene ENE-WSW thin-skinned shortening related to the Europe-Adria plates convergence, which generated WSW-verging thrusts, N-S or NNW-SSE trending folds and conjugate strike-slip faults (Doglioni, 1987); 4) Late Oligocene to Recent NNW-SSE thick-skinned shortening, which produced SSE-verging thrusting (Caputo et al., 2010).

Analytical methods
More than two hundred rocks have been collected during three fieldwork campaigns from 2015 to 2017. All the samples were cut to produce thin sections and, after a detailed petrographic check, 61 of them were selected for whole-rock chemical analyses (28 lavas from the Dolomites, 19 plutonic and volcanic rocks from Predazzo, 8 plutonic rocks from Monzoni and 6 lavas from the Vicentinian Alps; Fig.

A C C E P T E D M A N U S C R I P T
2). Representative samples were cut with a diamond-disc saw to small chips to completely remove weathered parts. These were first crushed in a steel mill to about mm-scale fragments, then washed with distilled water and dried at 110 °C overnight. The dried and clean chips were powdered in a low-blank agate mortar to a size passing 200-mesh sieve. Whole-rock major and trace element analyses were performed at Activation laboratories (Ontario, Canada) using Code 4LITHO Major Elements Fusion ICPAES and trace Elements Fusion ICPMS. The detection limit ranges from 0.001 to 0.01 % for major elements, 0.1 to 30 ppm for trace element and 0.004 to 0.1 ppm for REE.
Mineral chemistry analyses were performed at the Istituto di Geologia Ambientale e Geoingegneria (CNR-IGAG) laboratory, Dipartimento di Scienze della Terra of Sapienza University of Rome, with a CAMECA SX50 electron microprobe. The analyses were carried out at 15 kV accelerating voltage, 15 nA beam current, 10 µm beam diameter for feldspar and 1 µm for pyroxene and opaque minerals. Counting times were 20 s at the peak position and 10 s for background for all elements. Calibration was made on synthetic and natural minerals.
Seven representative samples from the volcanic rocks of the Dolomites area have been analysed for Sr-Nd-Pb isotope ratios at the Istituto di Geoscienze e Georisorse, (CNR-IGG) in Pisa. Rocks powders ì were dissolved in HF+HNO3 and then completely dried. Sr and Nd were extracted by ion exchange chromatography. Lead was extracted from the matrix after eluting with HBr and HCl. The separated Sr, Nd and Pb aliquots were loaded on Re filaments and then loaded into a Finnigan Mat 262 multi-collector Thermal Ionization Mass Spectrometer (TIMS). A complete description of the analytical techniques is reported in Di Giuseppe et al. (2017).

Petrographic description and mineral chemical composition
The investigated samples are essentially massive rocks (44 samples; dykes, sills and lava flows) plus rarer pyroclastic facies rocks (4 samples) from the Dolomites. Seven volcanic rocks from Vicentinian Alps (hereafter referred to as Recoaro), 8 plutonic rocks from Monzoni and 16 plutonic rocks from Predazzo complete the dataset. This research focusses on the massive volcanic rocks of the Dolomites and consequentially only minor comments will be devoted to the other volcanic and plutonic rocks presented here, the readers being referred to the recent articles of Casetta et al. (2018aCasetta et al. ( , 2018b and to Bonadiman et al. (1994) for Predazzo and Monzoni plutonic rocks description, respectively. A detailed petrographic description of the two types of Dolomite volcanic rocks and their mineral chemical compositions is reported in the electronic appendix 1.

A C C E P T E D M A N U S C R I P T
Thirty-one representative samples of Dolomites volcanic rocks were analysed for whole-rock chemistry. The results of these chemical analyses and CIPW normative minerals are reported in Table 1. In the same table, new whole-rock analyses of Recoaro volcanic rocks as well as the Monzoni and Predazzo plutonic rocks are reported too. Selected major oxides and trace elements of the new and literature data (a total of 1120 whole-rock analyses, available as Electronic Appendix 3) are reported in Figs. 6 and 7. For sake of clarity, the same diagrams, with each single district shown at the time, are reported in the Electronic Appendix 4.
In terms of silica content, the massive volcanic rocks of the Dolomites are basic to intermediate (SiO2 = 49.2-58.1 wt%, LOI-free basis) with alkali content (Na2O+K2O) ranging from 4.0 to 7.9 wt%. The Loss-on-ignition (LOI) value is generally <3 wt%, although few samples record higher values (up to 7.3 wt%). Based on TAS classification scheme (Le Maitre, 2002), the analysed samples are classified mostly as trachybasalts and basaltic trachyandesites, with minor alkali basalts, basaltic andesites and one trachyandesite (Fig. 3a). The Alpe di Siusi hyaloclastite tuff (sample AS6) is characterised by fresh glassy shards. These show peculiar compositions, being characterised by higher SiO2 (55.1-64.0 wt%) and alkalis (Na2O+K2O = 10.4-13.1 wt%) than the massive rocks. The glassy shard compositions mostly fall in the trachyte field with rarer tephri-phonolite and one trachyandesite (Fig. 3a). Only the altered tuff sample shows a sodic alkaline affinity, falling in the mugearite field, possibly indicating K2O depletion.
The Monzoni plutonic rocks range from ultrabasic to intermediate compositions (SiO2 = 41.7-57.2 wt%), aligned along a rather sharp trend with mild enrichment of alkalis (Na2O + K2O = 1.6-7.6 wt%). The Predazzo plutonic rocks spread over a much larger range of SiO2 (44.0-75.9 wt%) and alkalis (Na2O+K2O = 1.5-9.6 wt%), completely overlapping the Monzoni rocks, straddling the subalkaline-alkaline division in the TAS diagram, reaching rhyolitic compositions, with a gap in the trachyte field. The Recoaro volcanic rocks mostly fill the Predazzo rock gap, plotting in the 59.3-69.6 wt% SiO2 range. As already discussed in the literature (De Vecchi and Sedea, 1983), and confirmed in this study, the Recoaro rocks are characterised by strong alkali mobility, rendering handling of geochemical data of these samples at least dubious (see the discussion section).
In the K2O vs. Na2O diagram (Fig. 3b), all the Dolomites samples fall into the field of the transitional and potassic series fields, with the exception of two outliers characterised by high LOI. The two outliers are the Lago di Alleghe mugearitic tuff (LOI = 5.8 wt%), characterised by anomalously high Na2O (6.3 wt% on LOI-free basis) and very high Na2O/ K2O (3.93), and the Alpe di Siusi shoshonite (LOI = 6.2 wt), which is characterised by high K2O (5.7 wt%) and low Na2O/K2O (0.38). Also in this case, the EMP compositions of sample AS6 glassy shards fall into the potassic and ultrapotassic series fields (Fig. 3b). The new samples from Predazzo, Monzoni and Recoaro districts completely overlap the Dolomites rock field.

A C C E P T E D M A N U S C R I P T
According to the K2O vs. SiO2 diagram (Peccerillo and Taylor, 1976), most of the investigated rocks belong to the high-K calcalkaline and shoshonitic series, (Fig. 3c), as most of the samples reported in literature. Only moderately to strongly altered SiO2-rich samples from Vicentinian and Brescian areas tend to plot in the calcalkaline or arc-tholeiite fields, due to K2O depletion. On the AFM Diagram (Kuno, 1968), all the studied samples and the literature analyses plot in the calcalkaline field, lacking substantial iron enrichment during the evolutionary trend (Fig. 3d).
More than half of the Dolomites samples are SiO2-saturated with CIPW normative orthopyroxene and olivine. Six are characterised by slight normative quartz content (0.62-4.98%), and eight of them are SiO2-undersaturated with normative nepheline ranging from 1.92 to 7.42%. The glassy shards in the hyaloclastite sample range from slightly SiO2-oversaturated (with CIPW normative quartz = 1.7-3.3%) to slightly SiO2undersaturated (normative nepheline = 1.1-3.4%). More than 37% of Monzoni samples, more than 60% of the Predazzo samples and all the Recoaro rocks are quartz-normative.
The major oxides of the analysed samples were plotted vs. the Differentiation Index parameter (D.I.= sum of CIPW normative Q, Or, Ab, Ne, Lc and Kp; Thornton and Tuttle, 1960) in Fig. 4. When considering the entire dataset, MgO, TiO2, Fe2O3tot and CaO exhibit well defined negative correlation with D.I (Fig. 4). Al2O3 and P2O5 define bell-shaped patterns, with maximum at ~60 and ~40, respectively, whereas a correlation is observed for SiO2, K2O (well-defined) and Na2O (more scattered) vs. D.I.. The glassy shards usually overlap whole-rock data of the other Triassic igneous rocks from the Alps and Dinarides.
Predazzo plutonic rocks define a large and quite continuous spread of major oxides, showing always a good correlation with D.I., in nearly all the cases with R 2 >0.85 (up to >0.96 for CaO and Fe2O3tot). Monzoni and Recoaro rocks exhibit straight correlations with D.I. too, overlapping the trend of the previous two areas.
All the Dolomites samples are metaluminous with ASI [ASI = Alumina Saturation Index = Al/(Na+K+Ca)] ranging from 0.52 to 1.00, similarly to what observed for the Alpe di Siusi glassy shards (ASI = 0.75-0.95) and the other investigated Southern Alps igneous rocks. A good correlation exists between ASI and D.I., with ASI values increasing from 0.47 to 1.00 in the 19.2-66.8 D.I. range. Only the most evolved compositions (D.I. >78; SiO2 >66.5 wt%) have peraluminous compositions with ASI as high as 1.49. Mg number [Mg# = 100*Mg/(Mg+Fe 2+ ), assuming Fe 3+ = 0.15% Fe] of Dolomites rocks ranges from 35 to 63, with no correlation with D.I. and a poor correlation with SiO2 (R 2 = 0.39, excluding the two SiO2-richest samples). On the contrary, the Predazzo and Recoaro rocks show good negative correlations of Mg# with SiO2 (R 2 = 0.96 and 0.86, respectively). The low Mg# value, coupled with relatively low Cr (30-330 ppm) and Ni (20-80 ppm) indicate that the investigated rocks cannot be considered primary melts in equilibrium with mantle rocks and that they experienced fractionation during their ascent to the surface (see below).

A C C E P T E D M A N U S C R I P T
Selected trace elements are plotted vs. D.I. in Fig. 5. Dolomites rocks show no correlation between compatible/incompatible elements and D.I.. Light-REE (LREE; La, Ce, Pr, Nd and Sm) are roughly correlated with SiO2, except for sample PR23 (highly altered basaltic andesite). Middle-REE (MREE: Eu, Gd, Tb, Dy, and Ho) and heavy-REE (HREE; Er, Tm, Yb and Lu) do not show any correlation with D.I..
The other three investigated SATIR districts show recognizable and rather coherent trends. Among LILE, only Rb behaves as an incompatible element, showing enrichment with increasing D.I. (Fig. 5). Strontium and Ba reach a peak at D.I. ~50 and then decrease. Transition elements show well-defined negative trends with D.I., particularly evident for Sc, V and Co (Fig. 5). HFSE + Y behave as a coherent group, aligning along straight correlation with D.I., with the only exception of the most differentiated compositions of Predazzo (D.I. >87). What described for the HFSE is mimicked by the REE, with the exception of Eu, which is strongly depleted in the most differentiated Predazzo samples (Fig. 5). Thorium and U behave as strongly incompatible elements, and the most differentiated Predazzo samples are characterised by high Th (up to 70 ppm) and U (up to 14 ppm). Lead distribution is less defined for all the districts.
Primitive mantle (PM)-normalised multi-elemental patterns (Fig. 6) are similar with limited inter-elemental fractionation for the different analysed rock types. Key features of the patterns are medium to high enrichment of LILE (Rb ~39-315, Ba ~39-143, Th ~38-181, U ~40-168, K ~46-235 and Pb ~42-257 times PM), deep Nb-Ta troughs (~13-26 times PM), strong trough at Ti (~3-8 times PM) and very flat HREE (HoN/LuN = 1.02-1.29). The Dolomite rocks define a pattern that, with the only exception of P, perfectly overlaps the Global Subducting Sediment (GloSS) composition (Plank, 2014). When compared to a classical within-plate oceanic locality such as the St. Helena basalts in central Atlantic Ocean (the type locality of HiMu-OIB), the Dolomite rocks show remarkable differences, in particular much higher LILE, negative anomalies at Nb-Ta-Ti and strong positive peaks at K and Pb (Fig. 6a).
The rocks from the other three investigated districts (Predazzo and Monzoni plutons and Recoaro) show inter-elemental fractionation patterns similar to those of Dolomites volcanic rocks. The Predazzo plutonic rocks nearly overlap the incompatible elements of Dolomites, showing slightly higher Rb, peaking at U-Th. Predazzo clinopyroxenites show the lowest incompatible element content, but overall they have the same pattern of the bulk plutonic rocks. The only syenitic sample analysed here (PR25) is characterised by the highest enrichment in nearly all the elements (Fig. 6b). Predazzo granites have an even more spiked pattern, with deep troughs at Ba, Sr, P, Eu and Ti and the highest concentration for the other elements, as well as flat HREE clustering around 13-16 PM times (Fig. 6c). Monzoni plutonic rocks closely resemble the Dolomites lava pattern too, sharing strong similarities with the GloSS composition, albeit incompatible elements content is generally displaced towards lower abundances (Fig. 6d). Also in this case the cumulitic rocks (gabbros) show very low elemental content, overlapping the Predazzo pyroxenites. The Recoaro samples are usually displaced towards higher incompatible element contents, with the exceptions of deeper troughs at Sr, P and Ti. The sample with the highest LOI (HA9; LOI = 4.19 wt%) slightly depart from the main pattern; in particular it displays the lowest Sr and P and a high MREE-HREE content (Fig. 6e).

Sr-Nd-Pb isotopic ratios
Based on petrographic characteristics, geochemical signatures and geographic position, seven Dolomites volcanic rock samples were selected for Sr-Nd-Pb isotopic determination. These analyses are reported in Table 2 and shown in Fig. 7 The Dolomites lavas mostly plot in the enriched Sr-Nd isotopic quadrant (Fig. 7a), partially overlapping the plutonic rock compositions of Predazzo plutonic rocks (Casetta et al., 2018b), but being displaced towards slightly more radiogenic Nd compared to Monzoni plutonic rocks (Bonadiman et al., 1994) for a given 87 Sr/ 86 Sr. Recent data from Valsugana dykes (poorly age constrained at 227-260 Ma with K/Ar datings; Bianchini et al., 2018), sampled between Dolomites and Vicentinian Alps, ~10-15 km E of Trento city, are characterised by a much larger content of radiogenic Sr ( 87 Sr/ 86 Sr = 0.7082-0.7172) and much lower content of radiogenic Nd ( 143 Nd/ 144 Nd = 0.51197-0.51209; Fig. 7a).
To the best of the authors' knowledge, the only Pb isotopic data on SATIR are those reported for the sulphides in Maso Furli, Val di Non, Doss de la Grave and Sasso Negro areas (Nimis et al., 2012) and for Valsugana dykes (Bianchini et al., 2018). The new data presented here show a partial overlap with these literature data, being the Valsugana dykes characterised by sensibly higher content of radiogenic 206 Pb/ 204 Pb (18.57-19.12) and 207 Pb/ 204 Fig. 7b,c). No direct comparison with 208 Pb/ 204 Pb can be made, because Th elemental content to recalculate the initial values is not reported in Bianchini et al. (2018). Assuming a reasonable Th/Pb ratio ~0.5 for the Valsugana dykes, the 208 Pb/ 204 Pb(238 Ma) ratios of these samples ranges from 38.87 to 40.19, values much more radiogenic than those for Dolomites massive rocks presented here.

A C C E P T E D M A N U S C R I P T
The Dolomites samples analysed in this study completely overlap the Predazzo and Monzoni plutonic rocks in terms of 87 Sr/ 86 Sr and do not show any appreciable variation with SiO2 or MgO, similarly to what recorded by Predazzo and Monzoni plutonic samples (Fig. 7a,b). The Valsugana dykes are displaced towards very high radiogenic values, but again without any correlation with differentiation parameters. On the other hand, the Northern Karawanken (Austria) Triassic igneous rocks mimic increase of 87 Sr/ 86 Sr with increase of magma evolution. With the exception of the most evolved sample analysed here, no correlation between 143 Nd/ 144 Nd and SiO2 or MgO is observed (Fig. 7c,d). The samples analysed here are the most radiogenic among the SATIR in terms of 143 Nd/ 144 Nd, with higher ratios recorded in the Karawanken pluton only (Miller et al., 2011). As Pb isotopes are concerned, the bulk of the Dolomites samples does not show any correlation with SiO2 or MgO, even if the most evolved sample (PR3 latite; 1.8 wt% MgO) is characterised by the lowest 207 Pb/ 204 Pb (Fig. 7e,f).

Discussion
New and literature data on SATIR are discussed with three approaches, the first focussing on the origin of the inter-and intra-district geochemical variations, the second trying to constrain the composition of mantle source(s) and the third aiming to define a geotectonic model in a large scale geological framework.

Inter-and intra-district geochemical variations.
As a whole, our data define relatively coherent trends in Harker-type diagrams for SATIR. This is particularly surprising, considering their subaqueous emplacement, with associated variable degrees of element mobility, the large width of the study area (>4000 km 2 ; Figs. 1 and 2) and the thick and compositionally heterogeneous crust pierced by the various magma batches. When major elements are plotted against differentiation parameters such as SiO2, MgO or D.I., some important considerations can be made: 1) Figures 6 and 7 show the major oxide and trace element variation against D.I. for the four investigated districts (Dolomites volcanic rocks, Predazzo and Monzoni plutonic rocks and Recoaro volcanic rocks). With a few exceptions, the investigated rocks define positive (SiO2, Na2O+K2O) and negative correlation (TiO2, Fe2O3tot, MnO, MgO, CaO, CaO/Al2O3) with D.I. (Fig. 4). Only Al2O3 does not show any clear trend with this index, with a rough bell-shaped pattern, being mostly confined in the 14.5-19 wt% range with D.I. ranging from ~5 to ~60.

A C C E P T E D M A N U S C R I P T
volcanic rocks show a rather restricted compositional range compared to the other SATIR districts.
5) The most differentiated syenogranitic samples of Predazzo are not considered cogenetic with the rest of the plutonic rocks because of the impossibility to model fractional crystallization processes (Casetta et al., 2018a). Consequently, the existence of three distinct magma batches has been proposed in literature (Visonà, 1997;Casetta et al., 2018a). Given the large volume of granitic magma (~1 km 3 ), its derivation from less differentiated melts via prolonged fractional crystallization is quite improbable. Similar mechanisms perhaps operated during the Permian rift stages in the Oslo rift, where small differences in silica activity of mafic to intermediate melts evolving in variable conditions (pressure, PH2O/CO2, fO2) generated a series of magma chambers at various levels in the thinned continental crust. Relatively hot magma ponding may have caused local crust partial melting generating evolved liquids with D.I. up to 94, with limited genetic link with mafic and intermediate magmas (Neumann, 1980).
6) The Mg# and major oxide variation of Dolomites volcanic rocks suggest fractional crystallization of Mg-bearing phases such as olivine and pyroxene. No quantitative assessment can be made because of the absence of clear correlations of major oxides and trace elements with D.I. The rough correlation of Al2O3 vs. D.I. and the strong negative correlation of CaO/Al2O3 vs. D.I. of Ladinian volcanic rocks indicate only a minor role of plagioclase in the fractionating assemblage, despite its common presence among the phenocryst phases and the Eu/Eu* <1 (1.01-0.75). 7) The definition of the serial affinity of SATIR is not as straightforward as it could be assumed. Essentially, all the SATIR plot in the calcalkaline field in the AFM diagram (Fig. 3d). On the other hand, in the Miyashiro (1974) FeOtot/MgO vs. SiO2 diagram, modified by Arculus (2003), more than two thirds of the SATIR (with <70 wt% SiO2) show relatively high FeOtot/MgO, clustering in the tholeiitic field (mostly "Medium-Fe field" of Arculus, 2003).
The mildly alkaline composition of most of the SATIR (Fig. 3a), coupled with variable K2O/Na2O ratios (mostly in the 0.3-1.6 range; average 1.16 ± 0.58) and the common mobility of alkalis, with widespread leakage of Na2O and enrichment of K2O (e.g., De Vecchi and Sedea, 1983), render the definition of the magmatic affinity a difficult task. Despite these caveats, we agree with the common use of "shoshonitic affinity" for the SATIR, whose least differentiated compositions have ~1.2-3.9 wt% K2O in the MgO 6-8 wt% and SiO2 49.2-51.2 wt% interval. 7) A limited to moderate amount (2-10 wt%) of assimilation of Permian micaschist/dioritic/rhyolitic/carbonate country rocks can explain the inter-group Sr and Nd isotopic variation of Predazzo plutonic rocks (Casetta et al., 2018a). The 143 Nd/ 144 Nd-87 Sr/ 86 Sr correlation of silica-undersaturated shoshonitic group of Predazzo (albeit defined by three samples only) is at odd with a crustal contamination process, but Casetta et al. (2018a) interpreted the spread of Sr isotopic ratios with local fluid mobilization along the intrusion borders.
8) The high Rb/Sr of the most differentiated plutonic rocks of Predazzo does not allow precise recalculation of the initial 87 Sr/ 86 Sr isotopic ratios (Casetta et al., 2018b). Initial Nd isotopic ratios of Predazzo granites ( 143 Nd/ 144 Nd = 0.51222-0.51230) overlap with those of the less differentiated compositions (0.51219-0.51229). More in detail, on the basis of the differences in Nd isotopes, Casetta et al. (2018b) confirm an earlier hypothesis of the existence of two independent magma batches for the least differentiated rocks with silica-undersaturated ( 143 Nd/ 144 Ndi = 0.51226-0.51229) and silica-saturated shoshonitic groups (0.51219-0.51225. Our Predazzo granitic rocks have not been analysed for isotopic ratios, but they could be qualitatively interpreted as the most extreme differentiation products of the basic compositions (gabbros). The major oxide variation seen in Fig. 4 indicates the possible involvement of feldspars in the latest crystallization phases, causing an abrupt change in the pattern of Al2O3, Na2O and, less evident, K2O. The strong depletion of Sr and Ba in the most differentiated samples of Predazzo is compatible with such a process, as well as the strong depletion of Hf and Zr could involve zircon fractionation. In any case, volumetric considerations on the evolved and least differentiated igneous compositions speak against simple direct derivation via fractional crystallization processes.

A C C E P T E D M A N U S C R I P T
9) Only a few SATIR samples have been analysed in literature for both Sr-Nd isotopes and whole-rock analyses (5 from Monzoni, 8 from Predazzo, 8 from Karawanken, and 8 from Valsugana). Our seven samples are the first covering the entire Dolomites area. The situation for Pb isotopes is even worst, being the data presented here the first for the entire SATIR, with the exception of the recent work of Bianchini et al. (2018) on Valsugana dykes. Excluding the few Monzoni data (Bonadiman et al., 1994) and two basaltic and andesitic 218 Myr-old lavas from Brescian Alps (Cassinis et al., 2008), all the SATIR plot along a classical hyperbolic trend in 143 Nd/ 144 Nd vs. 87 Sr/ 86 Sr diagram (Fig. 8a). This is classically interpreted as the result of interaction of mantle melts with ancient continental crust or the derivation from ancient mantle sources that have experienced variable degrees of melt extractions and/or modifications in the form of digestion of subducted material (e.g., Lustrino et al., 2011;Lustrino and Anderson, 2015, and references therein). In the first case, the isotopic ratios are expected not to vary or to vary without any correlation with evolutionary degrees, while in the case of continental crust clear contamination trends should emerge with major oxides. 87 Sr/ 86 Sr ratios of the Dolomites volcanic rocks presented in this study (0.70432-0.70577) do not show any correlation with D.I. (R 2 = 0.03), with SiO2 (R 2 = 0.001) or MgO (R 2 = 0.14). The spread of isotopic data can be interpreted, therefore, as characteristic of the mantle sources, only little modified by interaction with local crust. Interestingly, Predazzo and Monzoni rocks (with the exception of the SiO2richest sample) do not show any correlation with D.I. too (R 2 = 0.02 and 0.0004, respectively). The same conclusions can be obtained for the (fewer) 143 Nd/ 144 Nd vs. D.I. trends. In conclusion, most of the isotopic variation seen in the SATIR is here related do mantle source heterogeneities rather than to post-melting interaction with upper crustal lithologies. It is also worth of noting, however, that interaction of basaltic melt with distinct crustal rock types (e.g., metapelite-metagreywackemetacarbonate successions, meta-igneous rocks) can lead to the absence of correlation between isotopic ratios and major oxides.

A C C E P T E D M A N U S C R I P T
Several are the geochemical tools to infer derivation from a subduction-modified mantle source, and the Th/Yb vs. Ta/Yb diagram is one of the most widely used. It is based on the behaviour of Ta, which is substantially immobile in the restite during the metamorphic reactions in a titanate-bearing sinking slab, while Th is strongly partitioned into the fluid phase (e.g., Brenan et al., 1994) in focused fluid transfer through mantle wedges (Pirard and Hermann, 2015). Moreover, Ta and other HFSE are forced to remain in the subducting slab where titanates with very high Kd HFSE are easily stabilised (e.g., Foley et al., 2000;Kimura, 2017). The Yb at the denominator is used to buffer the absolute Th and Ta values to the same degree of evolution. In Fig.  9a, the Dolomites samples plot outside of the mantle array, displaced towards high Th/Yb ratios (1.3-4.9) for given Ta/Yb values (0.22-0.45). This can be interpreted as additional evidence for the derivation from subduction-modified mantle sources.
The Dolomites lavas are characterised by Eu/Eu* <1 also in the least differentiated terms. The absence of correlation of this parameter with SiO2, MgO or D.I. indicates source characteristics that could be ultimately related to the digestion of subduction-related recycling of continental crust material (Eu/Eu* = 0.70; Rudnick and Gao, 2003) in their mantle sources, confirming the previous hypothesis.
A quantitative approach to define the intensive and extensive variables that had a role in the genesis of the middle Triassic volcanic rocks of the Dolomites (and in arc magmas in general) is not possible. Only qualitative or semi-quantitative geochemical modelling can be drafted, hypothesizing fundamental parameters responsible for the chemical composition of mantle wedge partial melts.
The latest release of the Arc Basalt Simulator software (ABS5; Kimura, 2017) was used to infer the conditions of magma genesis, assuming for the Dolomites lava the derivation from a supra-subduction mantle wedge. As input parameters, we hypothesized an original mantle source (pristine mantle wedge) with a depleted mantle composition (Salters et al., 2011), with an additional 0.8% MORB extraction from DMM. Additional assumed paremeters are the P-T path of Solomon subduction zone (#32 in Kimura, 2017), slab-derived flux originating at P = 3 GPa, and a slab surface temperature = 792 °C (calculated by ABS5 on the basis of the selected P and P-T path of subduction zone). The slab liquid is assumed to derive 20% from the 0.3 km thick sedimentary cover, and 80% from 0.3 km thick altered oceanic crust in a reactive porous flow regime (%R slab = 90; Kimura 2017). Without the involvement of zone refining processes, open system partial melting is assumed at P = 2.1 GPa and T = 1280 °C, with 2% slab melting and 1.1% peridotite wedge melting. The calculated slab melt contains 1.9 wt% H2O and the basaltic melt contains 2.1 wt% H2O. All these parameters represent reasonable values acting in a subduction system.
The composition of the hypothetical liquid produced, together with the hypothetical DMM source and LAT5 potassic basalt are shown in Fig. 9b. Bearing in mind the highly speculative approach, the model results in a close match between

A C C E P T E D M A N U S C R I P T
calculated and target (LAT5) melts The calculated degree of melting of the mantle wedge is low, likely due to the effect of non-primary composition of the target melt (LAT5). Indeed, during fractional crystallization, incompatible elements tend to be concentrated in the residual melts, and this process is modelled by ABS5 reducing the degree of melting. The overlap between the calculated melt and LAT5 sample is nearly total in terms of incompatible element budget, with slight differences only in very mobile elements such as Rb, Th and K. In conclusion, the results obtained with ABS5 software indicate, with the necessary caveats, that it is possible to obtain a basaltic liquid with incompatible element fractionation and absolute content closely resembling that of basalt sample LAT5 assuming typical subduction settings from a peridotitic mantle wedge source.

Geodynamic and tectonic setting.
Although paleotectonic environments are usually inferred interpreting geochemical characters of igneous rocks (e.g., Pearce and Cann, 1973;Meschede, 1986;Saccani, 2015), a warm caveat seems to emerge in recent literature in using tectonic discrimination diagrams (e.g., Li et al., 2015). In particular, it has been suggested that the geochemical signature needs always to be checked against y geological/geophysical data (e.g., Lustrino et al., 2011Lustrino et al., , 2016. When misfits arise between geochemical data and geological/geophysical evidence, a solution needs to be looked for. Several authors invoked ancient (i.e., not directly related with the coeval tectonic setting) subduction-related modifications of mantle sources to reconcile such misfits (e.g., Lustrino et al., 2011;Mazzeo et al., 2014;Gaeta et al., 2016;Di Giuseppe et al., 2018). In other cases, the geochemical message of igneous rocks has been explained with shallow rather than deep geological processes (i.e., contamination of mantle melts with upper crustal lithologies instead of subductionmodified mantle sources; e.g., Esperança et al., 1992).
The SATIR were emplaced during Pangea rifting in a mainly transtensional tectonic setting, as testified by Triassic tectonic structures widespread in the Southern Alps (Doglioni 1984(Doglioni , 1987Brandner et al., 2016a, Abbas et al., 2018. However, the geodynamic scenario that originated these tectonic structures is still matter of debate. Indeed, the recorded stratigraphic successions and extensional tectonic structures have been alternatively related to rifting processes of the Meliata-Maliac Ocean (Gawlick et al., 2012;Sudar et al., 2012;Brandner et al., 2016b;Ferriere et al., 2016). Conversely, other paleogeographic settings have been proposed for the Middle Triassic western Tethys, invoking active subduction processes located west (Brack et al., 1999) or south (Stampfli et al., 2013) of Adria, Dinarides and Southern Alps nappes.
In this study we advocate, for the SATIR, an origin from subduction-modified (not directly subduction-related) mantle sources. The existence of a direct volcanic arc settings, as instead proposed by several authors (a.o., Castellarin et al., 1988; A C C E P T E D M A N U S C R I P T Garzanti 1985;Zanetti et al. 2013;Bianchini et al., 2018) can be excluded because the Rheic Ocean was already completely exhausted in the Alpine realm during Middle Triassic times. A back-arc setting for the SATIR (e.g., Marinelli et al. 1980;Viel, 1982;Brack et al., 1999;Cassinis et al., 2008) can be similarly excluded, because of geochemical considerations on Eisenkappel magmatism (Northern Karawanken, eastern Alps), pointing to anorogenic magmatism in an extensional setting (e.g., Visonà and Zanferrari, 2000;Miller et al., 2011). Back-arc basins are characterised by progressive migration of extensional tectonics and magmatism. Typical examples of migrating magmatism are described for the Tyrrhenian Sea, associated with the Apennine subduction , for the Lau Basin, associated with the Tonga subduction (e.g., Hawkins, 1995) and for the Parece Vela Basin, associated with the Marianas subduction (Ishizuka et al., 2010). The existence of a Middle Triassic back-arc basin setting is in our view unlikely because of two major observations: 1) the SATIR activity does not show any agespace migration trend of magmatism, differently from what observed in present-day back-arc basins. Indeed, the oldest SATIR occurred in the central-eastern part of the 450 km wide volcanic area, and subsequent activity occurred to the east and to the west with no clear trends (Fig. 1); 2) Magmas generated in back-arc basins have both MORB and subduction affinities, differently from the SATIR case (Taylor and Martinez, 2003).
The disagreement between the geochemical message of the igneous rocks (i.e., derivation from mantle sources modified by subduction) and the geological evidence (i.e., development of continental rift with horst-and-graben structures) requires a specific solution. The most likely model in this case is the derivation of partial melts by mantle sources previously modified by the north-directed subduction of the Rheic oceanic lithosphere before the late Carboniferous Gondwana-Laurussia collision.

Conclusions
During Middle-Late Triassic time, a large area in the Southern Alps experienced an intense and diffuse igneous activity. This magmatism (here referred to as SATIR = Southern Alps Triassic Igneous Rocks) followed the Devonian-Carboniferous subduction of the Rheic Ocean and the Carboniferous collision between the northern margin of Gondwana and the southern sectors of Laurussia with the development of Pangea. The last subduction stages and the collisional events were associated with abundant Permian calcalkaline to potassic magmatism in large sectors of the Variscan/Hercynian Chain.
During early Mesozoic, the Southern Alps portion of Pangea was affected by extensional tectonics with formation of intra-montane rifts and the development of horst-and-graben structures. This extensional tectonic phase ultimately evolved into the development, in Middle Triassic time, of huge carbonate platforms along subsiding area. During a narrow (<1 Myr) eruptive interval (Storck et al., 2019) huge

A C C E P T E D M A N U S C R I P T
amounts of magmas reached the surface in Southern Alps, now represented by volcaniclastic and epiclastic deposits up to 150 m thick, massive lava flows, dyke swarms and rare plutonic complexes. The igneous products are characterised by a wide compositional spectrum with compositions ranging from metaluminous basalt/gabbros to peraluminous rhyolites/leucogranites, including also cumulitic lithologies such as ultrabasic clinopyroxenites. The petrography, mineral chemistry, whole-rock chemistry and Sr-Nd-Pb isotopic ratios point to the derivation of the least differentiated melts from a subduction-modified mantle source. The shoshonitic affinities defined in terms of K2O vs. SiO2 are the most abundant, followed by high-K calcalkaline types. The mildly to strongly evolved compositions are generically linked with the least differentiated terms by closed system fractional crystallization of gabbroic to monzonitic assemblages, even if the SiO2-richest magmas likely require a separate origin.
At a large scale, no substantial differences can be recorded in the four districts investigated in this study (Dolomites lavas, Predazzo, Monzoni and Vicentinian Alps), all sharing similar geochemical affinity and likely similar mantle sources, identified in a depleted mantle metasomatised by subduction event(s). A close geochemical modelling and quantification of evolutionary processes is hindered by the bad preservation state of some of the samples (e.g., the Vicentinian Alps) and, in general, by the absence of true primitive magmas and the abundance of mildly to strongly evolved compositions.
The geochemical message of these products, that is derivation from a subductionmodified mantle source, is at odds with the geological, sedimentological and structural evidence indicating that magmatism occurred in association with tectonic and geodynamic processes (rifting and/or wrench tectonics) that will eventually lead to the development of the Adria passive margins. The apparent paradox can be solved requiring the activation of upper mantle sources, contaminated during the previous Variscan subduction, when local geotherms were raised due to the passive upwelling of asthenospheric mantle along the rifted margins.

Acknowledgements
H.A. thanks the Erasmus Mundus European Commission for providing financial support for 33 months in framework of EU METALIC II project (Erasmus Mundus Action 2) during his PhD program. This article benefitted of thorough and detailed reviews of Tom Andersen (Oslo, Norway), Stefan Jung (Hamburg, Germany), Jörn-Frederik Wotzlaw (Zurich, Switzerland) and an anonymous reviewer. To all of them our thanks for the work that helped to clarify many aspects of the article. We thank also the Editor Sebastian Tappe (Johannesburg, South Africa) for coordinating the review process and for precious hints to improve the text readability. Special thanks to Marcello Serracino (CNR-IGAG, Rome) for his assistance during the electron microprobe analyses, to Domenico Mannetta for his help during the thin section    Sampling sites (yellow stars) collected from volcanic (purple) and plutonic (red) outcrops in the Dolomites (a) and in the Vicentinian Alps (b). The regional position of the two regions is reported in Fig. 1. Samples outside the coloured areas are from not mappable outcrops (mainly dykes). Detailed geographic sample locations are reported in Table 1.      (Hart, 1984). Fig. 9. a) Th/Yb vs. Ta/Yb (Pearce, 1982;Pearce and Peate, 1995)

Highlights
The first complete review of the Middle Triassic magmatism of Southern Alps is presented The composition of the igneous rocks indicates subduction-modified mantle sources The geochemical message is at odds with the geological evidences of continental rifting stages.