Bimodal volcanism in the Hegau region (SW Germany): Differentiation of primitive melilititic to nephelinitic rocks produces evolved nosean phonolites

Peculiar bimodal volcanism in

The melilititic-nephelinitic rocks formed by low degrees of partial melting of a carbonated amphibole ± phlogopite-bearing garnet wehrlite in the uppermost asthenosphere or in the thermal boundary layer and occur also in neighbouring regions.However, their coexistence with evolved nosean phonolites is a unique and so far, unexplained feature in the southern Central European Volcanic Province.
Thermodynamic modelling implies that removal of 11-19% oxyspinel, 4-10% olivine, 42-57% clinopyroxene, <3% mica, <9% feldspathoids, <8% feldspar from melilititic-nephelinitic parental melts at upper crustal conditions (~200 MPa) results in significant amounts of phonolitic residues that are compositionally similar to the exposed nosean phonolites.Strongly negative P and Ti anomalies and a trough of MREE in primitive mantlenormalized trace element patterns of phonolites indicate additional fractionation of titanite and apatite, consistent with the mineralogy of coarse-grained (nepheline) syenitic cumulates, which are present as enclaves in both rock suites.The modelling results suggest crystallization of the (nepheline) syenite cumulates between 1050 and 800 • C and ascent and eruption of the phonolite residues at no less than ~900 • C with complete solidification of the observed assemblage at >750 • C. Significant crustal assimilation during fractionation appears unnecessary to explain the mineralogical, mineral chemical, and geochemical characteristics of the phonolites.Prolonged upper crustal differentiation of the magmas in one case (nosean phonolites) and fast ascent of primitive melts in the other (olivine melilitites and melilite-bearing olivine nephelinites) can be explained by stress field changes from an extensional to a more compressive regime, the magma ascent becoming thereby increasingly structurally controlled and supported by brittle deformation.
Many concepts explaining bimodal rock suites involve different initial magma sources that occur either coincidentally or causally related in the same region, for example if the same geodynamic process causes partial melting of different lithologies and/or if the melting degree strongly varied in time, both of which produce different parental melts (e.g., Chen et al., 2018;Hoernle and Schmincke, 1993;Rooney et al., 2007;Zou and Zindler, 1996).Other models involve the wellknown scenario of magmatic underplating, where ascending mafic magmas (mostly basalts or andesites) pond at the Moho or in the lower crust and provide heat and fluids that are transferred upward and facilitate partial melting of overlying felsic lithologies (producing mostly granites/rhyolites; e.g., Peccerillo et al., 2007;Tamura and Tatsumi, 2002;Trua et al., 1999).
In other cases, the felsic compositions are last products of a differentiation path (via fractional crystallization) of the primitive magma compositions of the same volcanic region (e.g., Berger et al., 2009;Kong et al., 2023;Peccerillo et al., 2007;Ulrych et al., 2003;Vaněčková et al., 1993;Wörner and Schmincke, 1984).Elsewhere, significant crustal assimilation is additionally required to derive evolved assemblages from primitive ones (Corti, 2009;Peccerillo et al., 2003;Rooney, 2020;Schmitt et al., 2017;Trua et al., 1999).The absence of exposed intermediate rocks in bimodal volcanic regions has been explained by high viscosity and/or density of intermediate magmas, preventing ascent and emplacement at or near the surface (Freundt-Malecha, 2001 and references therein).Other concepts assume a general suppression of formation of intermediate rocks, i.e., fractionation of predominantly mafic minerals first and mostly felsic minerals thereafter with only little overlap (Peccerillo et al., 2003 and references therein).Accordingly, the intermediate differentiation stage was exceeded within a narrow temperature interval and with a low crystallization rate for thermodynamic and geochemical reasons.However, plutonic enclaves in volcanic rocks support the existence of cumulates of intermediate composition in crustal magma chambers and sometimes minor volumes of intermediate rocks like trachyandesites, trachybasalts, phonotephrites, or tephriphonolites do occur (e.g., Anti-Atlas, East African Rift, Eger Graben, Canary Islands, Rhön, Siebengebirge; Berger et al., 2009;Corti, 2009;Freundt-Malecha, 2001;Jung et al., 2012;Jung et al., 2013;Sliwinski et al., 2015;Ulrych et al., 2005;Ulrych et al., 2011;Vaněčková et al., 1993;Zaitsev et al., 2012).Peccerillo et al. (2003) suggest that small quantities of intermediate, crystal-rich melts form a thin mushy interlayer between the mafic and felsic lithologies of a fractionating, continuously fed magma chamber and can be easily entrained, constituting plutonic enclaves in the melts while rarely erupting themselves.
Except for scarce whole-rock analyses (von Engelhardt and Weiskirchner, 1961;Keller et al., 1990;Mahfoud and Beck, 1989), comprehensive mineral chemical and whole-rock geochemical analyses for the phonolites are still lacking.In the present study, extensive petrographic and mineral chemical data for the volcanic to subvolcanic rocks of the Hegau region (except for the tuffs) and their coarse-grained enclaves are presented and combined with fractional crystallization modelling using MELTS to shed light on the genetic relationship between primitive and evolved rocks and on the volcanic evolution in this area as a textbook example of Central European bimodal rock suites.

Geological setting
The Hegau volcanic region lies in SW Germany directly north of the High Rhine and northwest of Lake Constance with the NW-SE-trending, parallel Hegau and Überlinger See graben systems (Theilen-Willige, 2011).It is situated at the southeastern end of the Bonndorf Graben (Fig. 1a), which crosses the Schwarzwald and hits the Upper Rhine Graben (URG) in the west near Freiburg.The former three grabens are collectively known as Freiburg-Bonndorf-Bodensee Fault zone (FBBFZ; Egli et al., 2017).Additionally, the seismically active Albstadt shear zone (e.g., Ring and Bolhar, 2020) commences in the Hegau region and trends parallel to the URG, terminating immediately west of the Urach volcanic field, whose activity lasted from 19 to 12 Ma (Binder et al., 2023).Although major rifting processes occurred in the Miocene, post-Variscan, Paleozoic-Mesozoic precursor structures probably had a large impact on the recent fault systems in the ~15-25 km wide graben complex, possibly tapping the upper mantle by reactivation of deepseated faults (Egli et al., 2017;Ring and Bolhar, 2020).This is supported by the length of the FBBFZ (~100 km), indicating a major crustalscale deformation zone (Egli et al., 2017).Since the Early Miocene, graben formation has been intensified by NE-oriented extension and mainly NW-directed subvertical shortening (Ring and Bolhar, 2020).
The Hegau region is characterized by bimodal volcanism (Fig. 1a; Mahfoud and Beck, 1989;Mäussnest and Schreiner, 1982): In the north and west, ~25 isolated vents/necks and, more rarely, dykes of primitive olivine melilitites and melilite-bearing olivine-nephelinites (continuous transition, in the following referred to as melilitites-nephelinites) and tuffs of the same composition occur.In contrast, in the eastern central Hegau region, six prominent domes of evolved nosean phonolite dominate the landscape (Fig. 1b).Especially between these two volcanic fields and in the southeast, there are numerous mafic tuff vents and extensive tuff blankets, whose crystal load comprises mainly augite, hornblende, and phlogopite, and which are mostly lapilli tuffs.
As the southern CEVP predominantly comprises melilititic, foiditic, and basanitic rocks (Binder et al., 2023), the phonolites of the Hegau together with those from Katzenbuckel (Mann et al., 2006) and the Kaiserstuhl volcanic complex (KVC; e.g., Braunger et al., 2018) represent a peculiarity but are comparable to volcanic rocks from the French Massif Central (Wilson et al., 1995a and references therein).Based on geochronology (Binder et al., 2023), the olivine melilitites and melilitebearing olivine nephelinites, and ijolitic patches therein (12-9 Ma) are the youngest rocks (of the entire southern CEVP), whereas the nosean phonolites are slightly older (14-11 Ma).Syenite and nepheline syenite enclaves in the phonolites, in the melilititic-nephelinitic rocks, and in the mafic tuffs show the same age range .The primitive melilitites-nephelinites of the Hegau region were produced by lowdegree partial melting of an amphibole ±phlogopite-bearing garnet peridotite in the uppermost asthenosphere or the lower lithospheric thermal boundary layer, which underwent CO 2 -and Ca-rich subductionrelated metasomatism (Alibert et al., 1983;Binder et al., 2024;Blusztajn and Hegner, 2002;Dunworth and Wilson, 1998;Hegner and Vennemann, 1997).Mobilization, infiltration, and emplacement of T. Binder et al. metasomatic agents probably occurred in consequence of continuous dehydration and recycling of Variscan oceanic lithosphere, comprising the release of Ca-rich carbonates, further volatiles, and incompatible elements into the mantle (Binder et al., in press).
Graben and basin fillings of the Hegau region consist of Oligocene-Miocene Alpine molasse sediments, and are underlain by Upper Jurassic limestones (Schreiner, 2008).The deposition of conglomerates (17-15 Ma) and particularly the Juranagelfluh sediments (<15 Ma) reflect an important stage of erosion and unroofing in the Alpine foreland as a result of pronounced uplift of the Vosges-Swabian arc (Ring and Bolhar, 2020).Bedding inclines to the southeast at up to 5 • due to gradually increasing subsidence caused by the loading of the Alpine orogen (Egli et al., 2017).Distinct step faults on the southwestern flanks of the Hegau Graben system reach into the basement and dip subvertically to subhorizontally to the NE (e.g., Randen, Schienerberg faults; Fig. 1a), adding up to an offset of ~300 m, while northeastern graben faults are less well defined and dip SW, displaying a maximum offset of ~120 m (Mindelsee fault; Fig. 1a; Egli et al., 2017;Schreiner, 2008).The wedge-like sedimentary sequence thickens from the NW basin edge to 1600 m at the western end of Lake Constance (Schreiner, 2008).

Material and methods
For petrographic examination, 44 thin sections from 23 localities of the Hegau region were available (Table 1).The samples originate from collections of the University of Tübingen and the LGRB Freiburg (Germany); in addition, several localities were resampled to augment the inventory.All thin sections were studied by light microscopy and scanning electron microscopy (Phenom XL, Thermo Fisher Scientific) at the Department of Geosciences, University of Tübingen.A subset of the samples was used in the geochronological and petrological studies of Binder et al. (2023) and Binder et al. (in press).For the present study, 15 samples (ijolite, nosean phonolite, nepheline syenite, syenite) from seven localities (Table 1) were selected for electron probe microanalysis (EPMA).

Whole-rock geochemistry
Whole-rock analyses (major, minor, and trace elements) of twelve fresh to weakly altered samples from all six nosean phonolite domes of the Hegau region (Table 1) were performed at the Laboratory of Environmental and Raw Materials Analysis (LERA, Institute for Applied Geosciences), Karlsruhe Institute of Technology (Germany), using wavelength dispersive XRF (S4 Explorer, Bruker AXS) and ICP-MS (iCap RQ, Thermo Fisher Scientific).The data protocol given in the Supplementary data, File 1 comprises the analyses of the reference material, reference values for the standards, deviations from the reference values, acid blind analyses and their 1σ.To determine the mass loss on ignition, one gram of the powdered sample was heated to 950 • C for 3 h.For XRF, fused beads were prepared with a Li tetraborate/metaborate mixture (Spectroflux 110, Alfa Aesar) as flux.The certified standards AGV-1 (andesite), GS-N (granite), SY-2 and SY-3 (syenites) were measured intermittently as secondary reference materials during the same session as our samples (Govindaraju, 1994; CRM-TMDW-A, High-Purity standards, Inc.), yielding an accuracy in the range of 2.5% for most elements and 8% for Ti and Mn relative to sample composition.
For ICP-MS, 100 mg of powdered rock was treated by HNO 3 -HF-HClO 4 acid digestion.A sample was heated together with 65% HNO (subboiled), 40% HF (Suprapur®), and 65% HClO 4 (Normatom®) in a closed Teflon vessel for 16 h at 120 • C, assuring a total silicate decomposition.After evaporating the acids to incipient dryness, the residue was redissolved in 65% HNO 3 (subboiled) and evaporated again (three times) for purification purposes.The final residue was dissolved in mL of ultrapure H 2 O.The quality of the ICP-MS trace element analyses was ensured by including the two certified reference materials SY-2 and SY-3 (Govindaraju, 1994), an in-house phonolite standard, and two blank samples into the protocol.The accuracy relative to sample composition was 10% for most trace elements, except for Sc, Cr, Ni, Cs, Hf, Ta, W, and some HREEs (25%).
The contents of total C and S were determined by combusting powdered material in an O 2 flow with an induction furnace at 2000 • C (CS 2000, Eltra).The reference materials steel (n = 7) and cast iron (n = 2) were used for calibration and quality control.Accuracy relative to sample composition was in the range 1%; the 1σ was 0.7% for steel, 2.5% for cast iron.

EPMA
The mineral chemistry was determined using a JEOL JXA 8230 electron probe microanalyzer in wave-length dispersive mode at the Department of Geosciences, University of Tübingen.Depending on the composition of the minerals, different configurations (acceleration voltage, probe current, probe diameter) were applied (Supplementary data, Table A1).The specific sensitivities of the mineral surface to the probe current were a critical factor when performing the adjustments.Peak count times were fixed at 10 s for light and volatile elements in clinopyroxene, perovskite, apatite, and amphibole (Na, F), 16 s for the other major elements, and 30 s for minor elements, with background count time being each half of peak count time.The calibration was carried out by using synthetic and natural reference materials, and by application of several peak overlap corrections (Cr-V, V-Ti, F-Fe, F -Ce, Ce -Ba, Pr -La, Ba -Ti).An internal φ(ρz) raw data correction was performed for all analyses except oxyspinel and ilmenite analyses, for which ZAF correction was applied.Specific parameters for the analytical protocols are given in Supplementary data, Tables A2-8, details of mineral formula and endmember calculations in Supplementary data, File 2, and all analytical results in File 3. The results of the mineral chemical analyses performed in this study are supplemented by data from nine localities of melilitites-nephelinites from the Hegau region (Binder et al., in press).

Fractionation modelling using MELTS
Using the easyMELTS graphical user interface (version 0.2.4) of the MELTS software platform (Ghiorso and Sack, 1995), fractional crystallization of the primitive melilititic-nephelinitic Hegau magmas was modelled based on the thermodynamic database and algorithms provided in the rhyolite-MELTS 1.2.0 version (Ghiorso and Gualda, 2015;Gualda et al., 2012).This version is optimised for crystallization in bulk compositions at 0-3 GPa that are not saturated with quartz, considering also the H 2 O-CO 2 fluid saturation model of Ghiorso and Gualda (2015).
It represents the best approximation of the MELTS software package for fractional crystallization of mafic alkaline SiO 2 -undersaturated magmas at crustal depths and was successfully tested for very similar rock types (Kong et al., 2023;Mourey et al., 2023;Sliwinski et al., 2015 and references therein).The isothermal computation mode for each individual step and a temperature increment of − 10 K were used for the modelling, while the pressure was set constant (cf.Supplementary data, File 4).Only solids were considered as fractionating phases and the fO 2 buffer was set to ΔFMQ +3, based on thermodynamic modelling for these rocks (Binder et al., in press) and because Braunger et al. (2018) determined a similarly high ΔFMQ of +2 to +3 for the KVC olivine nephelinites.All available whole-rock analyses of melilititic-nephelinitic rocks of the Hegau region (Supplementary data, File 1) were used as input data to largely cover the entire geochemical range of conceivable parental magma compositions.These span H 2 O contents from 0.0 to 3.9 wt% (Supplementary data, Fig. A1) and CO 2 contents from 0.0 to 0.2 wt%, extending beyond the ranges of other work dealing with MELTS modelling of alkaline rocks (Mourey et al., 2023;Sliwinski et al., 2015).Note, however, that modelling was not applicable to some whole-rock compositions and resulted in a software error (see Discussion for details).The calculations were performed assuming crustal pressures of 200, 500, and 1000 MPa during fractional crystallization (cf.Kong et al., 2023;Mourey et al., 2023).The modelled evolution of the residual melt composition was compared with whole-rock data for the Hegau phonolites (this study; Alibert et al., 1983;Dunworth and Wilson, 1998;von Engelhardt and Weiskirchner, 1961;Keller et al., 1990;Krause, 1969;Krause and Weiskirchner, 1981;Staesche, 1995;Stock, 1990;Wimmenauer, 1974).

Whole-rock geochemistry
Consistent with mineralogy, most melilitites-nephelinites plot in the lower part of the foidite/melilitite field in the TAS diagram (Fig. 5).However, some occurrences have slightly elevated SiO 2 contents classifying them as picrobasalts, rarely as tephrites or basanites.Nosean phonolites fall within the range of phonolites in the TAS diagram, except for four analyses with slightly lower Na 2 O + K 2 O contents defining them as tephriphonolite, trachyandesite, or trachyte.
The Cr and Ni concentrations are high in the primitive rocks and decrease with decreasing MgO contents, whereas they are very low in the phonolites (Fig. 7a-b).Moderate Nb and Zr and high V concentrations in the melilititic-nephelinitic rocks contrast high Nb and Zr and moderate V contents in the evolved rocks (Fig. 7b-d).Rb, Pb, and U concentrations are significantly higher in the phonolites than in the primitive rocks (Fig. 7f-h).
Compared to the primitive mantle (Fig. 8a; Palme and O'Neill, 2014), the melilititic-nephelinitic rocks are strongly enriched in LREE (~80-130 times) but only slightly enriched in HREE (~2.5-5 times).The phonolites also show high enrichment for LREE (~45-150 times) and only moderate enrichment for HREE (~3-7 times), but additionally exhibit a trough for MREE (~2.5-13 times), whereas the primitive volcanic rocks are more enriched in these elements (~7-20 times).The primitive mantle-normalized trace element patterns show decreasing enrichment with increasing mantle compatibility in all volcanic Hegau rocks, albeit there are prominent exceptions (Fig. 8b).The melilititic-nephelinitic rocks have negative Rb, K, Pb, and positive Nb and P anomalies.In contrast, the phonolites reveal prominent positive U, Nb, and Zr anomalies and negative Ta, P, and Ti anomalies, while the K anomaly is absent and a slight positive Pb anomaly present.
T. Binder et al. in phonolites exhibits variable and increased proportions of Na pyroxenes (7-70%) and up to 27% Tschermak's pyroxene at the expense of the quadrilateral components.Clinopyroxene in the groundmass of phonolites resembles clinopyroxene in (nepheline) syenites and is on average even more dominated by Na end-members (up to 80%) at lower tschermakitic components.
While diopsidic clinopyroxene in the melilitites-nephelinites contains a maximum of 25% hedenbergite component, the proportion in the crystal rims of the ijolite can increase up to 51%, with the aegirine component reaching up to 21% (Fig. 9b).The trend for clinopyroxene in the phonolites and (nepheline) syenites is evolving more strongly towards aegirine (8-81%) at hedenbergite components of 20-45%, although some pyroxene crystals show a lower hedenbergite content (5-20%) deviating from the general trend.
In the primitive rocks and the ijolitic patches, the Al content reaches up to 0.6 apfu and increases with decreasing Mg contents that range All data represent original contents not renormalized to a volatile-free composition.For comparison, the compositions of the Kaiserstuhl and Katzenbuckel phonolites are also shown.from 0.52 to 0.88 apfu (Fig. 9c).Clinopyroxene rims in ijolite deviate from this trend and exhibit remarkably low Al (<0.05 apfu) and Mg concentrations (0.28-0.58 apfu).Clinopyroxene in the phonolites and (nepheline) syenites has even lower Mg contents (0.03-0.58 apfu), with the Al content (0.02-0.39 apfu) varying at Mg concentrations above 0.45 apfu but decreasing continuously at lower Mg contents.The same is observed for the Ti content (Fig. 9d), which is generally lower in the clinopyroxene rims of ijolite (0.04-0.08 apfu) and in the clinopyroxene of phonolites and (nepheline) syenites (<0.08 apfu) compared to that in the melilitites-nephelinites and to the macrocryst cores in the ijolite (<0.22 apfu).Usually, Zr does not exceed 1500 ppm; only in groundmass pyroxenes in phonolites, and in (nepheline) syenites Zr reach up to 6000 ppm (Fig. 9e).

Alkali feldspar
Alkali feldspar macrocrysts and microcrysts in the phonolites show no compositional differences and are Or 60-85 Ab 15-40 with up to 8% celsian (Ba component) and variations within macrocrysts expressed in concentric zoning (Fig. 10a).The anorthite component (<0.4%;Fig. 10a) and Sr contents (2300 ppm) are very low.Alkali feldspar in the (nepheline) syenites largely resembles this composition with few crystals reaching Or 93 .

Oxyspinel
Skeletal oxyspinel crystals in the ijolitic patches are magnetiteulvöspinel ss covering the same compositional range as in the melilititic-nephelinitic host rocks; exsolution lamellae in the skeletal crystals are Fe-rich ilmenite (Supplementary data, Fig. A2a).Mn (0.03-0.06 apfu), Zn (300-800 ppm), and V contents (2300-6000 ppm) largely correspond to the contents in most evolved oxyspinel of the primitive rocks and are comparatively high (Supplementary data, Fig. A2b-d).
Oxyspinel is almost absent in the nosean phonolites.

Perovskite
Skeletal perovskite in the ijolitic patches is generally more enriched in various trace elements than accessory perovskite in the melilititic-nephelinitic host rocks.However, the initial point of the mineral chemical evolution of the skeletal perovskite is always in the compositional range of the perovskite in the host rocks (Figs.11a & b).Nb reaches up to 0.10 apfu and correlates positively with Ta (<3000 ppm; Fig. 11c) and Sr increases up to 0.05 apfu, correlating negatively with Ti (Fig. 11d).In perovskite in the ijolite, Fe (~0.01-0.03apfu) decreases with advancing crystallization (Fig. 11e).Th (<6000 ppm) correlates positively with the REEs (<0.05 apfu), with the accessory perovskite in the melilitites-nephelinites sharing the same compositional range as the skeletal perovskite in the ijolite (Fig. 11f).

Titanite
Titanite in the phonolites contains up to 0.04 apfu Zr, 9300 ppm LREE+Y, 0.06 apfu Nb, and 0.08-0.12apfu Al + Fe (Supplementary Data, Fig. A3).Differences in the contents of these elements cause irregular to concentric zoning in the crystals, but without a distinct corerim trend (Supplementary Data, Fig. A3c).

Discussion
In the following, the genetic relationships between primitive olivine melilitites and melilite-bearing olivine nephelinites, rare ijolitic patches therein, (nepheline) syenite enclaves, and evolved nosean phonolites will be discussed and a comprehensive model for the magmatic evolution of the Hegau region will be presented.

Rare ijolite patches in melilititic-nephelinitic rocks: Evidence for rapid in-situ differentiation
In-situ differentiation in primitive olivine melilitites and melilitebearing olivine nephelinites from Hohenstoffeln and Hohenhewen (Fig. 1b) is implied by rare coarse-grained and irregularly shaped ijolitic patches (Figs.2c-g).Their geochemical composition is more evolved than that of the volcanic host as the contents of MgO, alkalis, Al 2 O 3 , and Fe 2 O 3 fall between those of the melilitites-nephelinites and the nosean phonolites.Other compounds are slightly lower (e.g., SiO 2 ) or slightly higher (e.g., CaO, TiO 2 , P 2 O 5 ) than in the primitive rocks and do not fall on a linear trajectory between primitive and evolved rock suite (Fig. 6).This is reflected by the lack of olivine in the ijolitic patches, while clinopyroxene, nepheline, titanomagnetite, perovskite, and apatite are abundant (Figs.2e-h).The crystallization trend of clinopyroxene evolves from diopsidic cores, which resemble the composition in the host rock, to aegirine-and hedenbergite-rich rims with similar Al and Ti contents as observed in phonolites and nepheline syenites (Fig. 9).Abundant skeletal perovskite and magnetite indicate rapid growth and a high degree of supersaturation of corresponding elements in the trapped residues due to fast undercooling (Figs. 2d, g & h;Gornitz, 1981).Compared to the composition in the host rock, perovskite shows continuous enrichment of Nb, Ta, Na, Sr, Th, and REE (Fig. 11), apatite is Cl-and H 2 O-rich (Fig. 12), and magnetite is Mn-and V-rich (Supplementary data, Fig. A2), which indicates relative enrichment of these elements in the residual melt pockets (Fig. 2c).An interpretation of such in-situ fractionation as missing link between melilitites-nephelinites and nosean phonolites is possible, although an expected increase in SiO 2 content and alkali feldspar saturation required for the evolution towards phonolites or nepheline syenites is not observed.It is suggested that insitu crystallization under close-to-equilibrium conditions without sufficient removal of solids (i.e., lack of fractionation) and rapid cooling of the rocks prevented more extensive differentiation towards SiO 2 -rich and CaO-, TiO 2 -, P 2 O 5 -poor phonolitic compositions, a shallow fractionation trend which is often observed in plutonic rocks (e.g., in the Fen complex, Norway ;Andersen, 1988).Binder et al. (in press).For comparison, the composition of clinopyroxene in the Kaiserstuhl phonolites is also shown (Braunger et al., 2018).(a) Triangular plot showing the distribution among the quadrilateral end-members (QUAD), Tschermak's clinopyroxenes (CaXXSiO 6 with X = Al, Fe 3+ , Cr, Ti), and Na pyroxenes (aegirine -NaFe 3+ Si 2 O 6 , Ti aegirine -NaFe 2+ 0.5 Ti 0.5 Si 2 O 6 , jadeite -NaAlSi 2 O 6 ).(b) Triangular plot showing the distribution among the end-members diopside (Di), aegirine (Aeg), and hedenbergite (Hd) normalized to 100%.Mg 2+ vs. (c) Al 3+ , (d) Ti 4+ , and (e) Zr. apfuatoms per formula unit; LLDlower limit of detection; ppmparts per million.

Nosean phonolites: products of fractional crystallization of melilititic-nephelinitic melts
Primitive melilitites-nephelinites and evolved phonolites show overlapping age ranges (12-9 Ma and 14-11 Ma, respectively), but phonolitic volcanism started earlier and the existence of evolved rocks in the crust at the time of eruption of the primitive magmas is indicated by the entrained coarse-grained (nepheline) syenitic enclaves (Fig. 4a).These can be interpreted as cumulate rocks with alkali feldspar ± aegirine-augite as dominant cumulus crystals and variable amounts of  Binder et al. (in press).For comparison, the composition of feldspar and haüyne-nosean-sodalite ss in the Kaiserstuhl phonolites is also shown (Braunger et al., 2018).nepheline, aegirine-augite, alkali feldspar, dark mica, and little carbonate representing the intercumulus melt (Figs.4b-g).The variable proportions of nepheline and clinopyroxene, partly within the same enclave (Fig. 4b), indicate a layered arrangement of mesocratic and leucocratic lithologies and point to an evolving and convecting magma chamber that may have experienced repeated magma replenishment.Since all recorded enclaves mirror an already highly fractionated and thus SiO 2 -and alkali-enriched magma composition, the question remains as to the primary magma composition from which the nosean phonolites evolved by differentiation, and whether this could be the same magma composition represented by the primitive melilitites-nephelinites.In this case, the older phonolitic volcanoes would have erupted only after significant magmatic differentiation in the crust, while the mostly younger melilititic-nephelinitic magmas would have been transported "directly" from mantle to surface.The evolution of clinopyroxene composition (Fig. 9) and the presence of sodalite group minerals (Fig. 10) in both the melilitites-nephelinites and phonolites support this assumption.The differences between primitive and evolved rocks in terms of clinopyroxene composition are consistent with the evolution of the whole-rock composition towards very low MgO and CaO content and increase in K 2 O content relative to the primitive rocks is also evident in the composition of the sodalite-group minerals: in the phonolites, these minerals lack Ca, but partially contain K, whereas for those in the melilitites-nephelinites, a high Ca content and hardly any K is observed (Fig. 10c).Higher Na 2 O, Al 2 O 3 , and SiO 2 contents in phonolites (Figs. 5 & 6c) are reflected by abundant nepheline and alkali feldspar and the presence of titanite instead of perovskite, as it was observed in the primitive rocks including their ijolitic patches.The primitive mantle-normalized REE patterns show a continuous decrease in enrichment towards the HREE, which for the primitive rocks can be explained by residual garnet in the mantle source that fractionates LREE from HREE (e.g., Kolb et al., 2012;Ulrych et al., 2011).The phonolites show a comparable enrichment factor for the REE, but with a marked trough of the MREE and higher contents of HREE relative to the primitive rocks (Fig. 8a).This can be explained by titanite fractionation, a mineral that enriches MREE over LREE and HREE (Prowatke and Klemme, 2005), and is supported by a remarkable negative Ti anomaly in the primitive mantle-normalized trace element pattern for phonolites (Fig. 8b), and by low Ca and Ti contents in the whole rock (Figs.6b & f).Likewise, the fractionation of Zr from Hf (Zr/Hf = 78-103) and Nb from Ta (Nb/Ta = 38-82) in the phonolites (Fig. 8b; Supplementary data, File 1) can be attributed to titanite fractionation, as this mineral is more compatible for the heavy than the lighter geochemical twins (e.g., Berger et al., 2014).In contrast, the Zr/Hf ratios in the primitive melilitites-nephelinites are lower and relatively constant (38-44; cf.primitive mantle: ~34), whereas the Nb/Ta ratios vary already considerably (18-42; cf.primitive mantle: ~14), which may be attributed to different amounts of residual phlogopite (Green et al., 2000) or prior carbonate metasomatism in their mantle source (e.g., Pfänder et al., 2012).
The general lack of strong REE enrichment in the phonolites relative to the melilitites-nephelinites (Fig. 8a) is caused by apatite and possibly pyrochlore fractionation, in accordance with a striking P trough in the trace element pattern of the phonolites (Fig. 8b), and with low Ca contents in the whole rock.Abundant apatite, titanite, and accessory pyrochlore (Figs.4h-k) in the entrained (nepheline) syenites corroborates fractionation of these minerals to generate the residual phonolitic melt.The remaining trace element patterns reflect relative enrichment of certain incompatible elements (e.g., Cs, Th, U, Ta, K, Nb, Hf, and Zr) in the evolved melt manifested by the accessories zircon, titanite, and pyrochlore, and elevated contents of Zr in clinopyroxene (Fig. 9e) as well as orthoclase-rich feldspar in the phonolites (Fig. 10a).The partly high Ba concentrations in alkali feldspar of the groundmass and in zoned phenocrysts (Fig. 10a) can be explained by high Ba contents in the parental melt (Fig. 8b) and further enrichment during fractionation, as is typical of nephelinite-derived phonolites (Le Bas, 1987).The lack of Cl in apatite in the nosean phonolites compared to ijolite and the primitive rocks (Fig. 12) may be due to the presence of abundant haüyne-noseansodalite ss in the evolved rocks incorporating all the Cl.The slightly positive Pb anomaly compared to the strongly negative one in the primitive rocks (Fig. 8b), low Ce/Pb and strongly varying Nb/ U ratios (Supplementary data, Fig. A4) could indicate assimilation of continental crust, as this would lower these ratios in mantle-derived magmas (e.g., Jung et al., 2012).In the CEVP and elsewhere, the formation of phonolites has been attributed mainly to fractional crystallization with subordinate to negligible crustal assimilation (Berger et al., 2009;Bourdon et al., 1994;Ulrych et al., 2003;Vaněčková et al., 1993;Wörner and Schmincke, 1984) or to significant and various amounts of crustal assimilation of e.g., mica schists, granites, paragneisses, metapelites, and mafic granulites (Jung et al., 2013;Kolb et al., 2012;Panina et al., 2000;Schleicher et al., 1990;Schmitt et al., 2017).However, verification and quantification of assimilation in the case of the Hegau phonolites would require isotopic data.

A thermodynamic modelling approach
To test the hypothesis that phonolites originate from fractional crystallization of olivine melilitites and melilite-bearing olivine nephelinites, the MELTS software package was used.Temperatures of up to 1350 • C for partial melting of the source lithology of the primitive melilitites-nephelinites of the Hegau region have been inferred by Binder et al. (in press) and are consistent with those for similar rocks from the CEVP (e.g., Jung et al., 2012;Mayer et al., 2013).However, Fig. 13.Results of fractional crystallization modelling of melilititic-nephelinitic Hegau magmas at 200, 500 and 1000 MPa using MELTS (Ghiorso and Sack, 1995).The whole-rock compositions of the primitive rocks served as input data.For all major oxides, including (a) SiO 2 , (b) TiO 2 , (c) Al 2 O 3 , (d) Fe 2 O 3 tot , (e) MgO, (f) CaO, (g) Na 2 O, and (h) K 2 O (each in wt%), phonolitic compositions can be realized by magmatic differentiation without significant crustal assimilation at temperatures of 750-1150 • C. The light red rectangles show the range of the whole-rock composition of the Hegau phonolites.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) T. Binder et al. actual modelling was performed and depicted with higher starting temperatures to identify and avoid potential artifacts (discussed below).Residual melt fractions of >95 mass% and thus no significant changes in melt composition are predicted for most model curves until plausible liquidus temperatures for melilitites-nephelinites are reached (Figs. 13 & 14).Therefore, the used model parameters and most whole-rock analyses of the melilitites-nephelinites are considered suitable for testing fractional crystallization towards phonolites.A corresponding composition should be achieved at realistic temperatures and with a quantifiable residual melt fraction to propose fractional crystallization of melilititic-nephelinitic magmas in a closed system without requiring significant assimilation.All the phonolites have porphyritic textures, which indicates continuous crystallization during cooling.Additionally, magma convection and replenishment events can result in cumulate remobilization and remelting and/or entrainment of cumulate crystals, resetting the melt towards less evolved compositions and rendering a wide temperature interval conceivable (e.g., Sliwinski et al., 2015;Wörner and Wright, 1984).Thus, different element contents would not necessarily reach the concentrations observed in the nosean phonolites at the same temperature.
For crustal pressures between 200 and 1000 MPa, the range of whole-rock compositions of the phonolites can be reproduced (except for Mn and P) at temperatures of 750-1150 • C (Figs. 13 & 14; Supplementary data, Fig. A1), with upper crustal pressures (i.e., 200 MPa) Fig. 14. Results of fractional crystallization modelling of melilititic-nephelinitic Hegau magmas at 200, 500 and 1000 MPa using MELTS (Ghiorso and Sack, 1995).The whole-rock compositions of the primitive rocks served as input data.(a) MnO, (b) P 2 O 5 , and (c) Cr 2 O 3 (each in wt%).(d) Melt residues of ~12-35 mass% of the initial magma are responsible for most calculated phonolite compositions, regardless of formation pressure.(e) TAS diagram showing compositions of the volcanic Hegau rocks.12 new analyses (black-rimmed triangles) were supplemented by literature data stated in Fig. 5.The trajectories show the melt evolution of melilititic-nephelinitic magmas predicted by MELTS modelling with whole-rock compositions of the primitive rocks as input data.For compositions from which no graphs emanate (pale blue squares), fractionation modelling was not successful.(For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.) T. Binder et al. tending to achieve the best-fit compositions.Geothermometric results for fractional crystallization towards phonolitic composition elsewhere reveal a similarly wide temperature range.Monchiquite-and tephritederived KVC phonolites started crystallizing at ~880-960 • C (Braunger et al., 2018) or at ~1040-1060 • C (Panina et al., 2000), while for the nephelinite-derived Katzenbuckel phonolite ~780-880 • C were estimated (Mann et al., 2006).The basanite-derived Laacher See mafic phonolite (East Eifel) reveals melt temperatures of ~860-1060 • C and the more evolved felsic one 720-860 • C (Berndt et al., 2001;Bourdon et al., 1994;Ginibre et al., 2004;Schmitt et al., 2010;Wörner and Wright, 1984).However, for sanidine and plagioclase (~950-1125 • C) and haüyne and apatite phenocrysts (1120-1180 • C) in the Upper Laacher See Tephra even higher crystallization temperatures were postulated (Sharygin et al., 2005).Alkaline rocks from Tenerife underwent a two-stage evolution from basanites via phonotephrites to phonolites with main crystallization intervals at ~1150-1110 • C and ~ 925-875 • C (Sliwinski et al., 2015).
No thermodynamic data for the Mn clinopyroxene end-member johannsenite and for fluorapatite are implemented in the MELTS database, which explains observed misfits concerning MnO and P 2 O 5 (Figs. 14a & b).Further, the modelled Cr content drops by fractionation of oxyspinel well before the assumed melting temperatures for the primitive rocks are reached (Fig. 14c).This may be due to the lack of a Cr-bearing pyroxene in the MELTS database and/or excessively high Cr contents in the rocks because of Cr-rich oxyspinel and Cr-diopside xenocrysts (see Binder et al., in press).For a few whole-rock compositions of the primitive rocks with particularly low or high alkali contents, the modelling attempt resulted in an abortion of the computation, so that no differentiation path could be determined (pale blue squares in Fig. 14e).This may indicate that those whole-rock compositions do not represent primary melt compositions due to high xenocryst and antecryst loads, hydrothermal alteration, or low-temperature weathering.
Overall, modelling implies that, depending on the assumed initial composition and depth of the magma chamber, ~12-35 mass% of the initial melilititic-nephelinitic melt remains as residue when phonolitic compositions are reached (Fig. 14d).This agrees with estimates for phonolite evolution from Laacher See, for which 30 mass% derivative melt of a basanitic magma were calculated (Wörner and Schmincke, 1984).During melt evolution, SiO 2 and Na 2 O + K 2 O enrichment is predicted, resulting in differentiation trajectories via basanites/tephrites ±(trachy)basalts, phonotephrites and tephriphonolites ±alkalirich foidites to phonolites (blue curves in Fig. 14e).In particular, the lower alkali melilititic-nephelinitic magmas undergo a melt evolution resulting in the observed SiO 2 and alkali contents of the nosean phonolites at ~850-1000 • C (Fig. 14e).
Early fractionation of olivine (4-10%) and oxyspinel (11-19%) is predicted for most initial compositions (Figs.15a & b).However, up to 3 mass% oxyspinel crystallization postulated at >1350 • C is likely caused by entrainment of mantle xenocrysts, leading to such modelling artifacts.These are only small quantities, so the effect should be negligible for the overall modelling, especially since this very early alleged fractionation partially corrects excessive Cr and Mg contents in whole-rock compositions towards actual melt compositions.Subsequently, extensive clinopyroxene fractionation (42-57%) is predicted (Fig. 15c), consistent with clinopyroxene in the mesocratic (nepheline) syenite enclaves (Table 2; Figs.4f & g).Considering variable amounts of macrocrysts, but also microcrysts of clinopyroxene in the phonolites (<10% in total; Table 2; Figs.3a-d), it can be assumed that its crystallization occurred over a long time and temperature interval and that fractionation and separation was not yet completed at the time of ascent of the phonolitic melt.This suggests temperatures of >900 • C at the time of melt extraction as clinopyroxene formation stops at lower temperatures (Fig. 15c).
It must be considered that the thermodynamic database of MELTS does not contain sodalite-group minerals, so nepheline is possibly overestimated and depending on the Na/K ratio and aSiO 2 , additional feldspathoids (leucite, kalsilite) are erroneously predicted.However, the total amount of feldspathoids formed is plausible, since their average stoichiometry is similar with respect to the Al-Si-alkali ratio, and they form under comparable physicochemical conditions.The onset of garnet crystallization, which occurs in neither the nosean phonolites nor the (nepheline) syenites, is predicted at 750-700 • C (Fig. 15g), most likely representing sub-solidus temperatures for the rocks (e.g., Braunger et al., 2018;Mann et al., 2006).However, MELTS only includes the pyralspite solid solution series and the grossular end-member, but not andradite, so interpretations regarding garnet formation must be questioned.Further, 1-2% apatite fractionation may have been slightly underestimated (Fig. 15h) as the software considers only hydroxyapatite, yet F-dominated apatite prevails in the primitive and evolved rocks (Fig. 12a).
In summary, we propose crystallization of the phonolites between 1050 and 750 • C from an evolved magma of initially melilititic-nephelinitic composition, resembling the geochemical characteristics of the primitive Hegau rocks.Fractional crystallization occurred mainly in the upper crust (~200 MPa) and formed (nepheline) syenitic cumulates at 1050-800 • C, fragments of which were entrained into the ascending phonolitic, melilititic-nephelinitic magmas, and into melts forming the mafic augite-hornblende-phlogopite tuffs, causing the contemporaneous and subsequent eruptions in the Hegau region.Fractionation of 11-19% oxyspinel, 4-10% olivine, 42-57% clinopyroxene, <3% mica, <9% feldspathoids, <8% feldspar in the order mentioned, and minor amounts of titanite and apatite (~2%) result in a phonolitic melt residue representing ~12-35 mass% of the initial melt.Given the lack of isotope data, crustal contamination (assimilation) during fractional crystallization cannot be ruled out, although the modelling results imply that it is not required to explain the observed phonolite compositions.

The Daly gap and implications for magma ascent and emplacement
Why did pronounced upper crustal differentiation processes in the central eastern Hegau region at ~15-11 Ma (ages of nepheline syenites and syenites) culminated in the emplacement of phonolite domes 14-11 Myr ago, whereas largely undifferentiated primitive melilititic-nephelinitic lavas were emplaced in the northern and western Hegau region during and after phonolitic activity from 12 to 9 Ma?We suggest that the compositional gap between primitive and evolved magmas is due to the largely separated crystallization and segregation of specific mineral phases and the thermodynamically and rheologically favoured timing of melt extraction.MELTS modelling shows that down to a temperature of ~1100-1050 • C, predominantly the mafic minerals oxyspinel, olivine, and clinopyroxene as well as small amounts of apatite crystallize from a fractionating primitive, initially melilitite-and nepheline-normative magma and consequently reduce the melt to ~30-50% of the bulk (Figs.14d & 15a-c).The lack of entrained enclaves or individual crystals of such wehrlitic to olivine-pyroxenitic cumulates in the phonolites is probably the result of effective gravitational separation of these dense mafic cumulus minerals from the light intermediate to felsic residual melt or complete dissolution because of strong disequilibrium with the continuously evolving residual melt.
Clinopyroxene and apatite formation continued down to ~900 • C, while mica, feldspar, and feldspathoids started crystallizing between 1050 and 950 • C, proceeding until the phonolites solidified.This period of fractional crystallization is well recorded by the macrocryst load of the nosean phonolites and by the differently composed (nepheline) syenite cumulate enclaves (Table 2).They are characterized by a continuous evolution of mesocratic (nepheline) syenites with up to 30 vol% clinopyroxene and 5 vol% dark mica ± apatite, titanite, and zircon to leucocratic (nepheline) syenites with alkali feldspar and dark mica ± nepheline, carbonate, titanite, pyrochlore, zircon, thorite, and thorianite (Fig. 4).Possibly, magma replenishment of intermediate composition triggered the eruption of the phonolites with entrainment of cumulate fragments and crystals at a temperature > 900 • C when clinopyroxene and apatite still crystallized.Related models to explain compositional gaps in bimodal alkaline volcanism (Canary Islands, Ethiopian Rift) likewise include substantial temporal separation between the crystallization of mafic and felsic minerals and preferential magma mobilization at specific times.This is explained by exceeding the intermediate differentiation stage within a narrow temperature interval with a respectively low crystallization rate (Peccerillo et al., 2003) or by optimal melt extraction from a crystal mush at intermediate crystallinity (i.e., 30-60% fractionation;Sliwinski et al., 2015).Accordingly, the latter is associated with favourable rheological conditions and released heat of crystallization, which lowers the crystallization rate, providing time for crystal-magma separation.Moreover, variations of density and plasticity of different basement lithologies and deflection along the interface of low-angle thrust faults may have affected the buoyancy of mafic melts, protracting ascent and allowing differentiation of the magmas in some cases (e.g., Lusatian Volcanic field; Büchner et al., 2015).In the Eger Graben, primitive olivine melilitites and olivine nephelinites are associated with phases of compressional stress and occur mainly in the pre-rift and subordinately in the late-rift period, while more evolved rocks dominate in the syn-rift phase under extensional stresses (Ulrych et al., 2011).Accordingly, volcanic rocks coeval with compressive stresses are associated with major faults, whereas volcanism coinciding with extensional stress shows a much weaker structural control, which may hinder rapid ascent and protract differentiation.Similarly, Przybyla et al. (2018) suggest a phase of intense rifting, i.e., a discrete event of lithosphere stretching and/or disruption to have caused a short phase of felsic volcanism in the Siebengebirge near Bonn, whereas undifferentiated mafic lavas erupted over a longer period before and after (>8 Myr).They propose a magmatic plumbing system, which emptied from the top down over more evolved to more primitive lavas within <1 Myr.
Analogously, we propose a structurally controlled mechanism for differentiation, ascent, eruption, and emplacement of the Hegau magmas governed by the regional geodynamic evolution.The latest possible onset of upper crustal differentiation processes is documented by the oldest (nepheline) syenite ages (~15 Ma, Hohenstoffeln; Fig. 1; Binder et al., 2023) and extends over wide areas of the Hegau basement.This is evidenced by corresponding enclaves in numerous occurrences of volcanic rocks scattered throughout the region and consistent with remote sensing and GIS studies, indicating an extensive sub-surface, ring-shaped magmatic structure (Theilen-Willige, 2011).Pronounced uplift in the Vosges-Swabian arc, i.e., at the southern end of the South German Block including the Hegau region, reflected by a phase of intensified erosion in the Swabian Alb (Ring and Bolhar, 2020), may have triggered this magmatic activity.Uplift and erosion proceeded during late Miocene to early Pliocene shortening and provoked thinskinned thrusting and folding in the Jura Mountains and Molasse basin (<12-10 Ma; Becker, 2000;von Hagke et al., 2012), overlapping with the age of the younger primitive volcanic rocks in the study area.Thus, intensified uplift and rifting may have firstly led to the development of larger upper crustal magma chambers and associated differentiation processes producing (nepheline) syenite cumulates and phonolitic melts.Subsequently, ongoing, progressive graben formation and disruption of the upper crust by emergence of local and/or (re) activation of deep-seated fault systems may have facilitated rapid magma ascent such that eruptions of melilititic-nephelinitic melts and mafic tuffs occurred without extensive fractionation.Obviously, this does not exclude a temporal overlap of the different manifestations of volcanism in the Hegau region, which can be explained by the spatial heterogeneity of the faulted crust.

Conclusions
Based on petrography, mineral chemistry, fractionation modelling, and previous U-Pb geochronology, the following model for the magmatic evolution in the Hegau region is proposed.
(2) The modelled compositions representing the observed phonolite compositions best were achieved for a pressure of 200 MPa, consistent with upper crustal formation conditions for phonolites in the CEVP and elsewhere.The predicted mineral assemblages agree largely with the petrography of the nosean phonolites and (nepheline) syenite cumulates, suggesting that the latter crystallized at temperatures of 1050-800 • C and that the phonolite melt was extracted at >900 • C. (3) A temporal progression from evolved rocks that experienced prolonged differentiation in upper crustal magma chambers (nosean phonolites) towards near-primary, fast ascending primitive melts (melilititic-nephelinitic rocks) could explain the sequence of eruptions in the Hegau region.The onset of magmatism corresponds temporally with a regional phase of intensified erosion in response to pronounced uplift in the Vosges-Swabian arc.This may have firstly led to the development of larger upper crustal magma chambers and associated differentiation processes that produced (nepheline) syenite cumulates and phonolitic melts.Increased uplift and erosion during late Miocene to early Pliocene crustal shortening coincide with the age of the younger, but more primitive volcanic rocks.Thus, ongoing, progressive graben formation may have facilitated magma ascent such that eruptions of melilititic-nephelinitic melts and mafic tuffs occurred without extensive prior fractionation.

Fig. 1 .
Fig. 1.(a) Simplified map of the Hegau region showing the different volcanic rock units and major fault structures (after Egli et al., 2017).(b) Enlarged section of the central Hegau region with its melilititic-nephelinitic vents in the west and phonolite domes in the east.Large areas in between are covered by mafic tuffs.

Fig. 8 .
Fig. 8. Primitive mantle-normalized (a) REE pattern and (b) incompatible element pattern of the igneous Hegau rocks (normalization values from Palme and O'Neill, 2014); elements sorted by increasing mantle compatibility.Analyses are from the literature stated in Fig. 5.For comparison, the compositions of the Kaiserstuhl and Katzenbuckel phonolites are also shown.

Fig. 12 .
Fig. 12. Apatite composition in the igneous Hegau rocks.Data of this work supplemented by Binder et al. (in press).(a) Triangular classification diagram based on the anion position (OH − , F − , Cl − ).(b) Triangular diagrams showing the molar Ca-Na-REE and Ca-Sr + Ba-REE distribution each normalized to 100%.

Table 1
Sample list including locality name, coordinates, and rock type.

Table 2
Sample list including locality name, coordinates, and rock type.Modal mineralogy of investigated volcanic rocks in the Hegau region.Distinctive values are in grey, characteristic accessories are in bold.