Metabasic rocks from the Variscan Schwarzwald (SW Germany): metamorphic evolution and igneous protoliths

In the Variscan Schwarzwald metabasic rocks form small bodies included within anatectic plagioclase-biotite gneisses. Many metabasites first underwent an eclogite-facies metamorphism at about 2.0 GPa and 670–700 °C, resulting in the assemblage garnet + omphacite + rutile + quartz ± epidote ± amphibole ± kyanite. Since these eclogites are nearly free of an OH-bearing phase, they underwent almost complete dehydration during subduction, suggesting formation along an average to warm top-of-the-slab geotherm of 10–13 °C/km. The age of the Variscan high-P/high-T metamorphism is > 333 Ma. After partial exhumation from ~ 65 to ~ 15 km depth, the eclogites were overprinted under increasing activity of H2O by a number of retrograde reactions. The degree of this overprint under amphibolite-facies conditions (0.4–0.5 GPa/675–690 °C) was very different. Up to now, only retrograde eclogites have been found, but some samples still contain omphacite. Kyanite is at least partially transformed to aggregates of plagioclase + spinel ± corundum ± sapphirine. On the other hand, there are amphibolites that are extensively recrystallized and show the assemblage amphibole + plagioclase + ilmenite/titanite ± biotite ± quartz ± sulphides. The last relic phase that can be found in such otherwise completely recrystallized amphibolites is rutile. After the amphibolite-facies metamorphism at ~ 333 Ma, the metabasites underwent a number of low-temperature transformations, such as sericitization of plagioclase, chloritization of amphibole, and formation of prehnite. The intimate association of metabasite bodies with gneisses of dominantly meta-greywacke compositions suggests derivation from an active plate margin. This view is corroborated by bulk-rock geochemical data. Excluding elements that were mobile during metamorphism (Cs, Rb, Ba, K, Pb, Sr, U), the concentrations of the remaining elements in most of the metabasites are compatible with a derivation from island-arc tholeiites, back-arc basin basalts or calc-alkaline basalts. Only some samples have MORB precursor rocks.


Introduction
Many collisional belts contain metabasic rocks that were metamorphosed under variable P-T conditions. Although such rocks may quantitatively be of minor importance, they often play a substantial role in the interpretation of orogenic belts. Their metamorphic parageneses and textures may throw light on their geodynamic evolution before and during continental collision. Furthermore, their bulk-rock chemistry may give valuable information on the tectonic environment in which their igneous precursor rocks were generated (island or continental arc versus rift, plume or mid-ocean ridge). Such information may help to get a better idea of the tectonostratigraphic evolution of an orogenic belt. However, due to manifold metasomatic processes at variable P-T conditions, bulk-rock chemical compositions may change significantly during orogenesis and may blur the primary igneous character of a metabasic rock. Hence, the major problem is to disentangle the effects of various metasomatic alteration events in order to understand and reconstruct the chemical changes that affected these rocks.
In the Moldanubian Zone of the European Variscides (Kossmat 1927) polymetamorphic metabasites are widespread (e.g. Fluck 1974;Klein and Wimmernauer 1984;Finger and Steyrer 1995;Faryad et al. 2013;Benmamar et al. 2020;de Hoÿm et al. 2020;Lotout et al. 2020). This paper deals with metabasites from two major tectonostratigraphic units of the Variscan Schwarzwald: the Central and the Southern Schwarzwald Gneiss complexes (CSGC, SSGC) (Kalt et al. 2000). While the amphibolite-facies gneisses of both complexes are derived from Ordovician to Devonian clastic sediments and granitoids (Hanel et al. 1999;Hegner et al. 2001;Kober et al. 2004), the age and origin of their small metabasic inclusions are still not exactly known, since up to now only very limited modern chemical and isotopic data are available (e.g. Klein and Wimmenauer 1984;Büsch et al. 1985;Kalt et al. 1994b;Hegner et al. 2001). Many of these metabasic bodies underwent an eclogite-facies metamorphism before they became metamorphosed under amphibolite-facies conditions at ~ 333 Ma (Suter 1924;Erdmannsdörffer 1938;Eigenfeld-Mende 1948;Klein and Wimmenauer 1984;Kalt et al. 1994b). Furthermore, the CSGC contains an area (located about 10 km to the East of Lahr, near to Hohengeroldseck Castle; Fig. 1) with metabasites that suffered a high-P/ultrahigh-T (HP-UHT) granulitefacies metamorphism prior to a LP-HT amphibolite-facies overprint (Hanel et al. 1993). In this study, we present geochemical data on a large number of metabasic bodies and we will try to interpret the bulk-rock chemical compositions of these bodies with respect to their primary igneous precursor rocks. Furthermore, we will discuss the geodynamic importance of these two types of metabasic rocks within the framework of the Central European Variscan orogen.

Geological setting
Located in the central part of the Variscan collisional belt of Europe (Fig. 1), the crystalline basement of the Schwarzwald consists of four main tectonometamorphic complexes. From north to south, these are the Baden-Baden Zone (BBZ), the CSGC, the Badenweiler-Lenzkirch Zone (BLZ), and the SSGC. All these complexes are older than granitoids with intrusion ages of 330-333 Ma (Altherr et al. 1999(Altherr et al. , 2019Schaltegger 2000).
Eclogitic parageneses in mafic rocks of the CSGC were formed at minimum pressures of 1.6 GPa and peak temperatures of 670-750 °C (Kalt et al. 1994b). A minimum age of this HP-HT event was given by three Sm-Nd (Cpx-)Grt-WR isochrons of 332 ± 13, 334 ± 11 and 337 ± 6 Ma (Kalt et al. 1994b). Moreover, Chen et al. (2003) gave two Sm-Nd Grt1-Grt2-WR isochrons of 340 ± 8 and 348 ± 3 Ma (Grt1 and Grt2 are mildly and strongly leached garnet fractions, respectively; WR = whole rock). Such Sm-Nd age values may, however, well be the result of retrograde Sm-Nd diffusion in garnet as was shown for eclogites from the Bixiling massif in China .
The assembled tectonic units of the CSGC experienced a common LP-HT metamorphism and anatexis at 0.4-0.5 GPa and 675-690 °C (Mehnert and Büsch 1982). The age of this metamorphism is between 330 and 333 Ma as indicated by concordant U-Pb ages on monazite and by Rb-Sr thin slab dating of migmatites (Kalt et al. 1994a). Note, however, that Weyer et al. (1999) reported a higher U-Pb age of 341 ± 2 Ma for monazite from a kinzigitic gneiss rich in garnet and pinitized cordierite.

3
The BLZ (Fig. 1) is a complex tectonic zone with rightlateral transpressive movement to the SE. It consists of several tectonic units that were either metamorphosed at low pressure and medium to very low temperature or are virtually non-metamorphic. The BLZ is cut by intrusions of deformed and undeformed granites with U-Pb ages on zircon between 331 and 333 Ma (Schaltegger 2000;Altherr et al. 2019).

Sampling strategy and analytical techniques
During or after metamorphism, the chemical composition of an igneous rock may change, in particular by the influence of aqueous fluids or melts, but also by solid-state diffusion. Primarily, such metasomatic changes are to be expected near the contacts of rock bodies that are chemically very different. In the given case of small metabasic rock bodies (< 800 m in diameter) that are enclosed within biotite-quartz-feldsparrich gneisses, this problem becomes highly significant. To reduce possible negative consequences, we only sampled metabasic bodies measuring at least ~ 10 m in diameter, and we tried to get samples from the inner parts of these bodies at some distance from their contacts to gneisses. However, it was almost inevitable to avoid samples showing petrographic indications of a late-stage hydrothermal overprint, such as sericitization of plagioclase or chloritization of amphibole, orthopyroxene, biotite or garnet. Moreover, we tried to avoid samples from metabasic rock bodies that are located near to late-and post-Variscan hydrothermal veins with Pb-Zn-Cu or Co-Ni-Ag-Bi-U mineralizations (e.g. Bons et al. 2014;Epp et al. 2019). Nevertheless, while 45 of our samples have Pb concentrations of 2.4-18.2 µg/g, we found three samples that contain unusually high concentrations of 58-190 µg/g Pb possibly indicating the influence of late hydrothermal fluids. Chemical compositions and coordinates (WGS84) of the analysed samples are given in Table S1.
For bulk-rock analysis, specimens were processed in a steel jaw crusher. After splitting, about 50 g aliquots were ground in an agate mill. The powders were dried at 105 °C for about 24 h. Major elements analyses were determined by wavelength dispersive X-ray fluorescence (WDXRF) spectrometry on lithiumtetraborate fusion disks (1.5 g rock powder + 7.5 g di-Lithiumtetraborate). International rock standards were used for calibration.
Trace element concentrations on 22 samples were determined by inductively coupled plasma mass spectrometry (ICP-MS) using a VG PlasmaQuad PQ1 equipped with a VG multi-channel-analyzer at the Institute of Geosciences, Kiel University, Germany. For a description of the instrument configuration and operating conditions see Garbe-Schönberg (1993). Another 26 samples were analyzed by LA-ICP-MS at the Geozentrum Nordbayern (GZN), the University of Erlangen-Nürnberg, with a single-collector quadrupole Agilent 7500c inductively coupled plasma mass spectrometry system equipped with an Analyte Excite 193 nm laser ablation system. Measurements were performed on lithium borate fusion disks using a spot diameter of 85 µm and a laser frequency of 20 Hz. Bulk-rock SiO 2 concentrations from XRF were used as internal standards. For external calibration, the glass reference material NIST SRM 612 with the values of Pearce et al. (1997) was used. Further details are given in Brätz and Klemd (2002).
Mineral analyses were carried out at the Institute of Earth Sciences at Heidelberg using a CAMECA SX51 electron microprobe equipped with five wavelength-dispersive spectrometers. Operating conditions were 15 kV accelerating voltage, 20 nA beam current and a beam diameter of ~ 1 µm. Counting times were 10 s for Si, Al, Fe, Mn, Ca, Na, and K, but 20 s for Mg, Ti, and Cr. Matrix corrections were done online with the PAP algorithm Pichoir 1984, 1985).
Back-scattered electron images were obtained using a LEO 440 scanning electron microscope with an attached Oxford Inca energy-dispersive analytical system at Heidelberg.

Petrography and metamorphic reactions
In the petrographic description of the investigated metabasites, mineral abbreviations as given in Whitney and Evans (2010) and Roman numerals and letters are used to describe different mineral generations and textural locations, respectively. Based on their petrographic characteristics, the metabasites can formally be classified into three groups. Group 1 metabasites show textural and/or mineralogical relics of a former HP-HT eclogite-facies stage overprinted to variable degrees by a later LP-HT amphibolite-facies metamorphism. Group 2 comprises equilibrated amphibolites without any eclogitic relics. These rocks either did not undergo a high 1 3 P/T stage or were completely recrystallized during the LP-HT event. Group 3 metabasites are former HP-UHT granulites that were also overprinted by a LP-HT amphibolitefacies metamorphism.
In group 1 metabasites, four sequential metamorphic stages can be recognized: (I) An eclogite stage at P-T conditions of > 1.6 GPa and 670-750 °C (Kalt et al. 1994b), (II) retrograde reactions in partial volumes of the rocks, caused by pressure release during near-isothermal conditions, (III) an amphibolite-facies stage at 0.4-0.5 GPa and 675-690 °C (Mehnert and Büsch 1982;Kalt et al. 1994a), and (IV) a multistage retrograde hydrothermal evolution at greenschistto zeolite-facies conditions (Klein and Wimmenauer 1984;Kalt et al. 1994b). The eclogite stage I is documented by relics of high-P/T minerals, such as garnet (Grt I), omphacite (Cpx I) and rutile (Rt I) that are eventually accompanied by kyanite (Ky I), quartz (Qtz I), pargasitic amphibole (Amp I), epidote (Ep I), and rare phengite (Phe I). Garnet I grains may contain inclusions of Cpx I, Rt I, Amp I, Ep I, Zrn I, Ap I, and Qtz I. Omphacite grains (X Jad = 0.32-0.47) are rarely preserved and may contain inclusions of Rt I and Grt I. Rutile I may be partially altered to ilmenite and/or titanite (Erdmannsdörffer 1938;Eigenfeld-Mende 1948;Klein and Wimmenauer 1984;Kalt et al. 1994b). Light-optical pictures of some of the most preserved eclogites are given in Fig. 2a During exhumation of the eclogites, a decrease in P and a variable increase in the amount of infiltrating H 2 O-rich fluids caused a number of metasomatic retrograde alterations (Klein and Wimmenauer 1984;Kalt et al. 1994b): 1.Omphacite (Cpx Ia1) is often partly or even completely replaced by diopsidic pyroxene (Cpx IIa) and oligoclase (Pl IIa). In samples, that still contain original omphacite (Cpx Ia) the beginning of this transformation is often documented by domains in which a secondary, less Na-rich omphacite (Cpx Ia2) contains tiny elongated crystals (< 10 µm) of nearly pure albite (Pl Ia) . In most cases, however, omphacitic Cpx Ia was completely transformed to a diablastic intergrowth of Cpx IIa and Pl IIa. Such transformations need the local addition of SiO 2 that was most probably introduced by an H 2 O-rich fluid phase: 2. Decomposition of garnet I started with two reactions. At places where Qtz I was in contact with Grt I, a twolayer corona was formed. Garnet relics are surrounded by intermediate plagioclase (Pl IIb), followed by an outer layer of orthopyroxene (Opx IIb) and intermediate plagioclase (Pl IIb), eventually accompanied by clinopyroxene (Cpx IIb) and amphibole (Amp IIb) ( Fig. 3a-d). This suggests the reaction: (1) Cpx Ia (Omp) + SiO 2 (aq) = Cpx Ib∕IIa + Pl Ib∕IIa.
Na 2 O was probably derived from decomposing omphacite, but could as well be introduced by an H 2 O-rich fluid phase.
The second garnet-consuming reaction took place at contacts of Grt I and Cpx Ia/Ib, where a kelyphite or symplectite consisting of amphibole, oligoclase/andesine and minor spinel, sometimes accompanied by ilmenite was formed ( Fig. 3e-f): 3. In more Al-rich retrograde eclogites, Ky I grains are present and may be partially mantled or entirely replaced by symplectites (S IId) of An-rich plagioclase (Pl IId) and spinel (Spl IId) which are encircled by an outer zone of Pl IId (Fig. 4a-c). These symplectites may contain smaller volumes, where corundum (Crn IId) or sapphirine (Spr IId) occur instead of spinel. While Crn IId occurs rarely along the contacts to relict Ky I, Spr IId was found in some samples near to the outer rim of the symplectites ( Fig. 4d-g). Thus, these three Al-rich phases reflect a variation in the local chemical potentials of SiO 2 , Al 2 O 3 , CaO, Na 2 O, MgO and FeO, producing the reaction The symplectite volume is always surrounded by an outer corona of Pl IId (Fig. 4c) that shows a strong outward decrease in An content. At places near to Ilm IId, the Ky-replacing symplectites may also contain minor amounts of relatively Mg-rich staurolite (St IId) (Fig. 4h).
4. In the more intensely retrograded eclogites, symplectites of amphibole + plagioclase ± ilmenite are formed, most probably by the reaction These symplectites may show increasing recrystallization to coarser aggregates of Amp III + Pl III ± Ilm III. 5.At a later stage, ilmenite may be replaced by titanite and magnetite according to the following oxidation and hydration reactions: and (2) Grt I + Qtz I + Na 2 O ± H 2 O ± K 2 O = Pl IIb + Opx IIb ± Cpx IIb ± Amp IIb.
(3) At places, rutile was also directly transformed to titanite, or it is surrounded by ilmenite that itself may be overgrown by titanite. It is obvious that these transformations of rutile were dependent on variable ƒO 2 and ƒH 2 O.
6.Biotite III was locally formed around Grt I and indicates the introduction of larger amounts of K 2 O. In more intensively overprinted eclogites, Bio III may form aggregates with Amp III and Pl III.
Group 2 metabasites are amphibolites without any mineralogical or textural relics of an earlier eclogitic stage. Either these rocks were never eclogites or they recrystallized completely during the amphibolite-facies event. Textures are foliated to massive. Besides Amp III and Pl III, these rocks may contain diopsidic pyroxene (Cpx III), Bio III, Qtz III, Tit III, Ilm III, Ap III and Zrn III. Amphibole III shows dark greenish to brownish colours and Pl III is polysynthetically twinned. Both phases contain abundant inclusions of all the other phases.
Granulite-facies assemblages in metabasic rocks of group 3 are relatively coarse grained and show a former paragenesis of Grt I + Cpx I + Pl I + Ky I ± Qz I ± Rt I ± Ilm I ± Ap I ± Zrn I (Hanel et al. 1993). At a later stage, during decompression, but still without H 2 O, the following reaction took place: Later hydration promoted the formation of symplectite rims around Grt I according to the allochemical reaction Furthermore, kyanite was allochemically transformed to An-rich plagioclase and spinel: Metabasites of all three groups may show an allochemical hydrothermal overprint. It is well known that in the Schwarzwald such low-temperature overprints of Variscan crystalline rocks and of Upper Carboniferous sediments took place in Late Carbonifereous, Permo-Triassic as well as in Late Jurassic to Cretaceous times (e.g. Lippolt and Kirsch 1994;Glodny and Grauert 2009;Brockamp et al. 2015). In particular, An-rich plagioclase is partially or completely replaced by sericite IV. Moreover, amphibole, orthopyroxene, biotite and garnet may be partially chlorititized (Chl IV). Rarely, veins of prehnite (Prh IV) have been observed in strongly amphibolitized samples.

Mineral chemistry of retrograde eclogites and amphibolites
Gar net I grains show var iable compositions (Alm 23-53 Prp 16-40 Grs 15-36 Sps 0.4-2.0 Adr 1-8 ), depending on bulk-rock chemistry and retrograde evolution (Table 1). Core-to-rim zoning patterns are characterized by nearly constant compositions of broad cores and narrow (< 100 µm) rims with increasing Prp and Alm, but decreasing Grs components, while Sps and Adr components are nearly constant or may vary unsystematically.
Clinopyroxene I is an omphacite (Table 1), whereby the highest NaO contents (4.0-6.9 wt %) are found in the cores of larger grains (Cpx Ia1) and in smaller Cpx Ia1 inclusions in garnet. Cpx Ia2 may contain small albite grains and may form rims around Cpx Ia1 or small grains in the matrix. It is characterized by significantly lower NaO contents (1.2-4.0 wt %). Diablastic Cpx IIa is diopside showing low Na 2 O contents (0.1-1.2 wt %). If present, Cpx III is diopside with very low Na 2 O contents (0.3-0.5 wt %). Rutile I is generally characterized by significant concentrations of Fe 2 O 3 (0.4-1.1 wt %).
O , whereby the ratio of Mg/(Mg + Fe 2+ ) may vary from sample to sample or from position to position within a thin section (Table 1).

r t h o p y r oxe n e I I b t h a t wa s fo r m e d by an allochemical reaction between Gr t I and Q t z I h a s a n a p p rox i m a t e c o mp o s i t i o n o f
Plagioclase compositions are variable, depending on the textural position. While Pl Ib is nearly pure albite, Pl IIa is oligoclase and Pl IIb is labradorite. Plagioclase IIc, formed by the reaction between Grt I and Cpx I, is oligoclase or andesine. Plagioclase IId within the pseudomorphs after kyanite is very rich in anorthite component (80)(81)(82)(83)(84)(85)(86)(87)(88)(89)(90)(91)(92)(93)(94)(95)  Amphibole analyses are given in Table 1. While rare Amp I is a ferroan pargasite, the compositions of Amp IIb, IIc, and IIe are very similar and are characterized by relatively high contents of Al 2 O 3 (~ 12-15 wt %) combined with low contents in TiO 2 (< 0.35 wt %). In marked contrast, Amp III compositions show lower Al 2 O 3 (8-10 wt %), but higher TiO 2 concentrations (0.7-1.8 wt %). While concentrations of Cl are always < 0.10 wt %, those of F may reach 0.35 wt %, but in most samples are also below 0.10 wt%.
Epidote/clinozoisite I is relatively rare and mostly occurs as inclusion in Grt I, where it is relatively rich in the pistacite component (e.g. Fe 3+ = 0.29 cpfu).
Chlorite IV grains show variable compositions, depending on their textural position. Within a sample, the mg# values are highest for chlorites formed from Opx IIb, intermediate for those formed from Amp II/III, and lowest for those formed from Bio III.  Stage  I  I  I  I  Ia  IIa  IIb  III  IIb  I  IIb  IIc  III  III  III

Bulk-rock chemistry of metabasites
Bulk-rock analyses for selected samples are given in Table 3 and analyses of all samples and of their coordinates are given in Table S1. In a TAS diagram (Le Maitre 2002), the samples are located in or near to the fields of basalt and basaltic andesite (Fig. 5a). Some samples, however, are characterized by very low SiO 2 values of 43-47 wt % and plot into the field of picrobasalts. Values of Mg# [Mg# = 100 × molar (MgO/ (MgO + FeO tot ))] of samples from group 1 and 2 show a continuous range from 69.8 to 38.4 and many of these samples are therefore considerably fractionated. The sub-alkaline character of most of the samples also becomes apparent in the nomenclature diagram of SiO 2 vs Nb/Y (Winchester and Floyd 1977) (Fig. 5b). In the diagram of P 2 O 5 vs Zr after Winchester and Floyd (1976) (Fig. 5c), most of the samples are located in the field for subalkaline basalts. In the    Winchester and Floyd (1977) showing the subalkaline character of the investigated metabasites. Note that the boundary between subalkaline basalt and basaltic andesite is at 53 wt % SiO 2 , instead of 52 wt % as in a. c P 2 O 5 vs Zr diagram discriminationg alkaline and subalkaline basalts (Winchester and Floyd, 1976). d FeO tot /MgO vs SiO 2 for tholeiitic (high FeO tot /MgO) and calcalkaline (low FeO tot /MgO) basalts (Miyashiro 1975) 1 3 diagram of FeO tot /MgO vs SiO 2 (Miyashiro 1975) (Fig. 5d), a tholeiitic nature of the basaltic protoliths is indicated. Concerning group 1 metabasites that suffered an eclogite-facies metamorphism, it is evident that nearly all Ky-bearing eclogites have CIPW-normative corundum (C). In the diagram of Al 2 O 3 vs SiO 2 (Fig. 5e) the former Ky-bearing eclogites show the highest Al 2 O 3 contents and there is no difference between the other three types of metabasites. For Mg# vs SiO 2 (Fig. 5f) there is no difference between the four different types.
In N-MORB normalized element concentration diagrams (Fig. 6) all samples of group 1 and 2 are characterized by variable positive Pb anomalies and nearly all samples show negative Nb-Ta, but positive K anomalies. There are only a few samples that show low K contents (e.g. FE10 in Fig. 6b, SP11 in Fig. 6e or FU11 in Fig. 6d). In general, the sizes of the positive anomalies of K and Pb are independent of each other. Moreover, many samples show positive Sr anomalies, but there are also samples with negative anomalies (e.g. FE10 and FE11 in Fig. 6b or OP1 and OP2 in Fig. 6f). Titanium and phosphorus contents are variable, with existing negative or positive anomalies, while Zr-Hf concentrations are either normal or show negative anomalies (e.g. FE92 in Fig. 6a or TN38 in Fig. 6b).
Samples of mafic metagranulites (group 3) are characterized by strongly positive K, Pb and Sr anomalies and negative P anomalies in N-MORB normalized element concentration diagrams (Fig. 6h). Furthermore, two of these samples (LA36, LA39) show marked negative Nb-Ta anomalies. Chondrite-normalized REE patterns of all three samples are relatively flat at low concentrations and show small, but significant positive Eu anomalies (Fig. 7h). The three samples show high Mg# values between 68.6 and 74.4 and Ni concentrations of 124-154 µg/g.

Pressure-temperature conditions of granulite-facies metamorphism
The P-T evolution of the mafic granulites (group 3 metabasites) was first investigated in Hanel et al. (1993). Usage of Grt-Cpx, Grt-Opx and two-pyroxene thermometry and Grt-Opx-Pl-Qtz barometry resulted in assumed formation conditions of 1.0-1.1 GPa and 780-850 °C. These values are lower than those published for felsic granulites of the Schwarzwald. For these rocks, metamorphic conditions were estimated as 1.4-1.8 GPa and 950-1015 °C by Marschall et al. (2003). On the other hand, Zack et al. (2004) gave a Zrin-rutile temperature of 980 °C (at 1.4 GPa) on one granulite from the CSGC. A more detailed analysis of the petrologic evolution of these rocks is in preparation.

Pressure-temperature conditions of eclogite-facies metamorphism and amphibolite-facies overprint
All metabasites of group 1, but possibly also those of group 2 that contain no relict eclogite-facies minerals, had an eclogitic history for which conditions of > 1.6 GPa and 670-750 °C have been proposed (Kalt et al. 1994b). However, former Al-rich eclogites contained kyanite I in coexistence with Cpx I that is characterized by jadeite contents of up to 47 mol % (Table 1). Using the reaction Pg = Jd + Ky + H 2 O as a geothermobarometer (Holland 1979) yields higher pressures of up to 2.0 GPa (at 670 °C) or 1.8 GPa (at 750 °C). Such pressure values correspond to depth values between 65 and 58 km, which then result in an average temperature-depth gradient of 10-13 °C/km. Such conditions plot between those of an average and a warm top-of-the-slab geotherm (Penniston-Dorland et al. 2015: Fig. 5f). It follows, that during subduction to a depth between 65 and 58 km, the metabasites were progressively dehydrated leading to a relatively dry eclogite assemblage of Grt I + Omp I + Rt I + Qtz I ± Ky I ± Ep I ± Amp I.
There exist a lot of calculated phase diagrams (pseudosections) for a range of MORB compositons (e.g. Rebay et al. 2010;Wei and Clarke 2011;Hernández-Uribe and Palin 2019;Wei and Duan 2019). These calculations show that resulting mineral assemblages are not only dependent on P and T, but also on H 2 O content and oxidation state of the system. Moreover, chemical parameters such as X MgO [= molar MgO/(MgO + FeO tot )] will also play an important role. In particular, at 2.0 GPa and 700 °C, the calculated mineral assemblage is strongly dependent on X MgO (Wei and Clarke 2011). With increasing X MgO , the calculated mineral assemblage changes from Grt + Omp + Ep (X MgO ≤ 0.27), via Grt + Omp + Ep + Ky During subduction, the more or less hydrated mafic rocks lose their porosity by compaction, and pore fluids will be expelled. As a consequence, nearly all of the remaining H 2 O will be bound in hydrous minerals. With the increasing metamorphic grade, possible H 2 O-bearing phases in metabasites are prehnite, pumpellyite, chlorite, epidote, amphibole, paragonite, phengite, and eventually also Mgchloritoid. Their modal fractions will, to a large degree, control the remaining H 2 O content of the rock. Most of the aqueous fluid produced by dehydration reactions could be lost along with localized fracture systems (e.g. Zack and John 2007;Klemd 2013;Plümper et al. 2017). At about 2.0 GPa and 700 °C, low H 2 O contents will favour the eclogite assemblage of Grt + Omp + Rt + Qtz ± Ep ± Amp (e.g. Rebay et al., 2010;Hernández-Uribe and Palin 2019;Wei and Duan 2019). Moreover, saline solutions, if present during subduction of the metabasites, will cause dehydration reactions to proceed at lower temperatures and shallower depth compared to pure H 2 O fluids.
After their formation, the eclogites with their country rocks were exhumed from a depth of about 65 km to the middle crust (~ 15 km depth). Relative to their ductile quartzrich country rocks, the eclogites were more brittle and this quality delayed their retrograde evolution. Increasing a H2O then caused the reaction to amphibolite-facies mineral assemblages at about 0.4-0.5 GPa and 675-690 °C (Mehnert and Büsch 1982;Kalt et al. 1994a, b).
The fact, that sapphirine was formed within the symplectites after kyanite, does not necessarily imply that temperatures were higher than ~ 700 °C. The lower thermal stability limit of Fe-free sapphirine of compositions 2:2:1 and 7:9:3 (MgO:Al 2 O 3 :SiO 2 ) at a pressure of 0.7 GPa was determined as ~ 750 °C by Seifert (1974). For Fe 2+ -bearing sapphirine this lower thermal stability limit should be located at even lower temperatures.
According to the experimental phase-equilibrium study of Al-and Ti-contents of calcic amphiboles in a MORB system by Ernst and Liu (1998), amphibole compositions may be used as a semiquantitative thermobarometer. While Al 2 O 3 contents in Ca-amphibole increase with both P and T, TiO 2 isopleths are nearly independent of pressure and show increasing TiO 2 contents with an increase in temperature. Amphibole II compositions are characterized by very low concentrations of TiO 2 (< 0.4 wt %) and high values (> 12 wt %) of Al 2 O 3 (Table 1). In equilibrium, such amphibole compositions would indicate unrealistic P-T conditions of ~ 2-3 GPa and ~ 500 °C. It is clear, however, that Amp II was formed in local environments with bulk compositions unlike that of MORB. Therefore, compositions of Amp II cannot be used to deduce the P-T conditions at which they were formed. In marked contrast, Amp III compositions show TiO 2 contents of ~ 0.9 wt % and Al 2 O 3 contents of ~ 8 wt % (Table 1) suggesting P-T conditions of about 0.5 GPa and 680 °C (Ernst and Liu 1998: Figs. 8 and 9) that are very similar to those suggested for the anatexis of the gneisses in which the amphibolites are hosted (0.4-0.5 GPa / 675-690 °C; Mehnert and Büsch 1982).

Metasomatism during metamorphic evolution
The metabasites of the CSGC and SSGC form small mafic bodies included within plagioclase-biotite gneisses that are poor in K-feldspar and were probably formed from continent-derived flyschoid sediments such as greywackes. In such a tectonic environment (trench), different types of basaltic rocks can occur in intimate primary, but also in secondary association with flyschoid sediments. First of all, these are orogenic basalts, such as back-arc basin basalts (BABB), island-arc tholeiites (IAT) or calc-alkaline basalts (CAB). All these rocks are generated above subduction zones, but blocks of them may later be included in the trench sediments. Secondly, there is the subducting oceanic crust that may include a wide range of chemical compositions from normal or enriched mid-ocean ridge basalt (N-MORB, E-MORB) to ocean island basalt (OIB). All these rocks may have been affected by seafloor alteration and hydrothermal metamorphism.
If the metabasic rocks of group 1 and 2 as well as those of group 3 in the CSGC and SSGC are derived from oceanic basalts (e.g. MORB, BABB or IAT), their chemical compositions might have been successively influenced by (1) submarine hydrothermal alteration, (2) subduction-zone metamorphism and related fluid-rock interactions, (3) fluid activity during LP-HT metamorphism (amphibolitization), and (4) hydrothermal activity within the crust during late to post-Variscan times. During all these processes, the extent of element mobility will strongly depend on the predominant physicochemical conditions, that is to say, the P-T conditions and the compositions of the aqueous (solute-poor or solute-rich) fluids that are coexisting, but not necessarily  1995). Values of (La/Yb) cn and Eu/Eu* [= Eu cn /(Sm cn × Gd cn ) 0.5 ] are given. a Samples with Eu/Eu* of ~ 1 and (La/Yb) cn > 1. b Samples with (La/Yb) cn > 1 and a negative Eu anomaly. c and d Samples with (La/Yb) cn > 1 and a negative Eu anomaly. e Samples with (La/ Yb) cn > 1 and a positive Eu anomaly. f Samples with (La/Yb) cn < 1 suggesting a MORB heritage. g Samples from the SSGC. h Mafic metagranulites showing patterns that suggest a cumulate nature of the protoliths ◂ in equilibrium with the solid phases and possibly small amounts of melt.
In case that the investigated metabasites initially suffered complex multistage hydrothermal alteration and submarine weathering, these processes must have had variable chemical effects. Seawater-basalt reactions often start at high temperature during the cooling of the newly created lithosphere (e.g. Jarrard 2003; Alt et al. 2010) and then proceed at lower temperatures during the lifetime of the oceanic crust. These reactions depend on local physicochemical conditions, e.g. temperature, oxygen and sulfur fugacities, chemical compositions of involved fluids, and water/rock ratio. There are a lot of studies on different parts of altered oceanic crust, and the chemical picture of these processes is still complicated and far from being uniform (e.g. Staudigel et al. 1996;Bach et al. 2003;Kelley et al. 2003;Schramm et al. 2005;Staudigel 2014). Still, there is a broad consense that (1) the contents of Si and Ca are often decreased, while that of Mg is increased, and (2) LILE (Cs, Rb, Ba, K, Pb) and U are often enriched in the altered rocks, while other elements, such as Sr, will either be enriched or depleted. Such alterations are often spatially limited and chemically heterogeneous. For example, in a vertical profile of the oceanic crust (Hole 504B), K 2 O was found to increase irregularly from an original value of about 0.04 wt % below ~ 600 m sub-basement to a range of values (0.04-0.45 wt %) near to the surface of the oceanic crust (Jarrard 2003), and in each sample such an increase in K is not always accompanied by a similar percentage of increase in other LILE or U.
Partial dehydration of the subducted metabasic rocks of the Schwarzwald occurred below the P-T conditions at which they were transformed to eclogites, i.e. below ~ 2.0 GPa and 670-750 °C, corresponding to depth values of less than 65 km. At these conditions, that are far below those of the second critical endpoint in the system MORB + H 2 O (Kessel et al. 2005a, b;Mibe et al. 2011), the solubility of most elements in aqueous fluids is fairly low, unless the fluids contain relatively high amounts of saline components, such as chlorine or fluorine. Such components have experimentally been shown to strongly increase fluid/solid partition coefficients for LILE (Cs, Rb, Ba, K, Pb), Th, U, and LREE, but not those for HREE and HFSE (Kessel et al. 2005a, b;Bali et al. 2011;Kawamoto et al. 2014;Tsay et al. 2014Tsay et al. , 2017Keppler 2017;Rustioni et al. 2019). During subduction, pore fluids (with Cl, F) in the altered oceanic crust will be lost at shallow depths (less than ~ 5 km), i.e. before relevant (OH, Cl, F)-bearing solid phases such as chlorite, epidote, amphibole or apatite become unstable (Jarrard 2003;Barnes et al. 2018). During further subduction, a number of dehydration reactions will take place and will be combined with a further bulk-rock loss in Cl and F. The last dehydration reactions that probably took place in the investigated rocks were those by which the contents of amphibole and epidote were minimized. It is worth mentioning, that serpentine, that might have been in tectonic contact with the eclogites, has its maximum thermal stability at about 2.2 GPa/720 °C (Ulmer and Tromssdorff 1995), i.e. somewhat above the P-T conditions at which the eclogites were formed.
As a consequence, it follows that the devolatilization of the investigated metabasites during subduction was probably accompanied by a significant impoverishment in those mobile elements that had probably been gained during submarine hydrothermal alteration.
Stages II and III in the evolution of the investigated metabasites include their uplift to about 15 km depth and their retrograde partial or complete transformation from eclogite to amphibolite. The chemical effects of this amphibolitization can best be evaluated by looking first at eclogite samples that were only moderately altered and do not contain significant amounts of amphibole, plagioclase or even biotite. Only five of our samples (FE10, FU7, FU11, SP11, TN14) are in accordance with this condition. These samples show relatively low K 2 O contents (0.18-0.78 wt%) and have low K/La and K/Th ratios, but high N-MORB normalized Th/U ratios (Fig. 6b, d, e). Moreover, the size of the positive Pbanomaly is variable and independent of the behavior of K and U. We conclude that the LILE (Cs, Rb, Ba, K, Pb) and U were strongly, but variably re-enriched during amphibolitefacies retrogression of the eclogites. Most probably, this was also true during later hydrothermal alteration events, such as the widespread sericitization of plagioclase. We do, however, not assume that elements such as Th and the LREE were significantly mobile during these alterations. There is only one sample (OP2) that may have lost some La and Ce, and another sample (OP9) that shows a negative Ce anomaly (Fig. 7f).

Parental igneous rocks of the metabasites from the Schwarzwald
Since we have to accept that the concentrations of LILE were drastically increased during and after the LP-HT metamorphism leading to a partial or complete amphibolitization of the metabasites, we will not use these strongly mobile Fig. 8 Primitive mantle-normalized element concentration patterns for nearly immobile elements. a and b Patterns with a clear negative Nb-Ta anomaly (relative to Th and La) and variable normalized element concentrations of P, Zr-Hf and Ti. c Patterns with a negative Nb-Ta and a positive Eu anomaly. d Samples with low Th contents and variable concentrations of P, Zr-Hf and Ti. e Samples with a negative Nb-Ta and Zr-Hf anomalies. f Samples that show a minor negative (OP) or a positive (SP) Nb-Ta anomaly. g Samples from the SSGC show negative Nb-Ta anomalies and variable contents of Zr-Hf, P and Ti. h Patterns of three metagranulite samples with different behavior of Nb-Ta. See text for further explanation ◂ elements in our attempt to identify parental volcanic rock types. Following Xia and Li (2019), we will instead restrict us to the abundances of the incompatible elements Th, Ta, Nb, La, Ce, Nd, P, Zr, Hf, Sm, Eu, Ti, Tb, Y, Yb, and Lu, normalized to the primitive mantle (McDonough and Sun 1995).
First of all, we have to interpret the chondrite-normalized REE spectra of the former kyanite eclogites (Fig. 7e). Except for sample HA19, these samples show positive Eu anomalies that are best explained by an accumulation of magmatic plagioclase. Therefore, these samples were possibly part of a gabbroic fractionation series. It is important to note, that within this group, there is one sample (EL4) that is completely recrystallized to an ordinary amphibolite.
Most samples show negative Ta-Nb anomalies relative to Th and La (Fig. 8). There are only a few exceptions to this 'rule'. Seven samples (Fig. 8d) are characterized by normalized Ta and Nb values that are higher than those of Th, but somewhat lower than those of La. Furthermore, two samples (SP52, SP99) show anomalously high Nb-Ta contents, i.e. they are depleted in La, Ce and Th, but have positive Nb-Ta anomalies (Fig. 8f). The other HFSE (Zr, Hf, Ti, P) behave somewhat unsystematic. Many samples show negative Zr-Hf anomalies (Fig. 8b, c, e), while Ti and P may have negative or positive anomalies. All these features, predominantly the negative Ta-Nb anomalies in combination with the variable behavior of P, Zr-Hf and Ti are typical features of orogenic basalts such as island-arc tholeiites (IAT) or calc-alkaline basalts (CAB). For more details, see Schmidt and Jagoutz (2017). We will now check this presumption by looking at more traditional discrimination diagrams.
In the diagram of Zr/Y vs Ti/Y (Fig. 9a), developed by Pearce and Gale (1977) to distinguish basalts from convergent plate margins from within-plate basalts, most of the investigated metabasites fall into the field of rocks from convergent plate margins. Some of the samples plot into the field of within-plate basalts and all these samples have positive Ti anomalies (relative to Sm) in primitive mantlenormalized incompatible trace element diagrams, for example, HA6 (Fig. 8a) or FRNE50 and SP11 (Fig. 8e).
The diagram Nb/La vs Nb (Fig. 9b) was successfully used by Xia and Li (2019) to discriminate between arc and anorogenic basalts. The data of most metabasites show Nb/La ratios smaller than unity suggesting an origin as arc basalts.
The frequently used diagram to distinguish arc from anorogenic basalts by looking at the relation of Th/Yb vs Ta/ Yb (Fig. 9c) was developed by Pearce (1982). Again, most samples of the investigated metabasalts plot into the area of arc basalts. However, samples SP52 and SP99 that are both characterized by positive Nb-Ta anomalies (Fig. 8f) plot far outside the fields of arc basalts (Fig. 9c).
The diagram of Ce/Nb vs Th/Nb (Fig. 9d) is thought to be nearly independent of fractional crystallization of basalts (Saunders et al. 1988) and was used to discriminate between different types of basalt. In this diagram, the investigated metabasites form an array from the fields for N-type MORB via the points of continental crust to the compositions of typical subduction-related island-arc and continental arc basalts. A similar message is send by the diagram of Th/ Nb vs La/Nb (Fig. 9e) from Plank (2005) that discriminates between anorogenic basalts that are derived from the depleted upper mantle (MORB) with low values of Th/Nb (mostly < 0.1) from those of arcs containing a continental sedimentary component (Th/La = 0.1-0.3). Almost all of the investigated metabasites have Th/La ratios that suggest an origin from arc basalts.
The diagram of Zr vs Y (Fig. 9f) was originally developed as a diagram of Zr/TiO 2 vs Y/TiO 2 by Lentz (1998) who used such a diagram to differentiate between metabasites that originiated from calc-alkaline basalts with Zr/Y > 7 from those with a tholeiitic heritage (Zr/Y < 4. Some authors have, however, slightly changed these boundaries (e.g. Ross and Bédard 2009), but we will use the original version. All investigated metabasites from the Schwarzwald show Zr/Y ratios between 1.05 and 8.57 and most samples have Zr/Y values below 4, suggesting a tholeiitic to transitional calcalkaline/tholeiitic character of their protoliths.
In addition, there exist some triangular diagrams for differentiating the protoliths of metabasalts. In the diagram Th-3 Tb-2Ta after Cabanis and Thiéblement (1988), most of the samples plot into the fields of island-arc calc-alkaline basalts (IAC), island-arc tholeiites (IAT) and back-arc basin basalts (BABB) with transitions to MORB (Fig. 9g). Only one sample (SP99) that has a positive peak for Nb-Ta in a Fig. 9 Various basalt discrimination diagrams applied for metabasaltic rocks from the Schwarzwald (red symbols = group 1 and group 2; green symbols = group 3). a Zr/Y vs Ti/Y diagram discriminating basalts from convergent plate margins and within-plate basalts (Pearce and Gale, 1977). b Nb/La vs Nb diagram discriminating anorogenic from arc basalts (Xia and Li, 2019). c Th/Yb vs Ta/Yb diagram (Pearce 1982). Most of the samples clearly plot above the trend of anorogenic oceanic basalts (MORB to WPD = within plate basalt) and are clearly orogenic. d Ce/Nb vs Th/Nb diagram (Saunders et al. 1988). Fields for N-MORB, E-MORB and OIB and points of depleted MORB mantle (DMM), primitive mantle (PM), continental crust (CC), global subducting sediment (GLOSS) and slab-derived component (SDC) are from Wang et al. (2019). e Th/Nb vs La/Nb diagram after Plank (2005). MORB basalt is characterized by very low Th/Nb ratios, while arc basalts generally show elevated ratios. f Diagram of Zr vs Y, modified after Lentz (1998), to distinguish between calc-alkaline, transitional and tholeiitc basalts. g Triangular Th-3 Tb-2Ta discrimination diagram after Cabanis and Thiéblement (1988) with compositional fields of back-arc basin basalts (BABB), continental alkali basalts (AB), continental tholeiites (CT), island arc tholeiites (IAT), island arc calc-alkaline basalts (IAC), and ocean island basalts (OIB). h Th-Hf/3-Ta discrimination diagram after Wood (1980) with compositional fields of island-arc basalts (IAB), continental arc basalts (CAB), within-plate basalts (WPB) and within-plate tholeiites ◂ primitive mantle-normalized concentration diagram (Fig. 8f) falls into the field of ocean-island basalts.
The diagram Th-Hf/3-Ta of Wood (1980) is perhaps the most used plot to distinguish basaltic rocks from different tectonic settings (Fig. 9h). Most of the Schwarzwald metabasites plot into the fields of arc basalts. Some samples, however, fall into the field of E-MORB and within-plate basalts (WPT). One sample (again SP99) falls exceptionally near the Ta corner.
Summarizing, we may state that the metabasic rocks are most probably derived from island-arc tholeiites and BABB. Some samples, however, might also have MORB precursors. It is important to note, that unlike many IAT and BABB, the metabasic rocks are characterized by very high Cs/Rb ratios, suggesting a metasomatic addition of aqueous fluid-mobile Cs to these rocks.

Possible equivalents of the investigated metabasites in other parts of the European Variscides
The Moldanubian Zone of the Central European Variscides exposed in the French Massif Central (FMC), the Vosges Mts (VM) and the Bohemian Massif (BM) contain at least two major tectonic units, the names of which vary from region to region. The uppermost 'Granulite (Gföhl) Unit' or 'Upper Gneiss Unit (UGU)' in the FMC consists of relic HP-UHT granulites with intercalations of UHP-UHT garnetbearing peridotites and eclogites (e.g. Fuchs 1976;Thiele 1984;Rey et al. 1992;Gardien et al. 1997;Berger et al. 2010;Faryad et al. 2013;Altherr and Soder 2018;Sorger et al. 2020;Vanderhaeghe et al. 2020). Not only the garnet peridotites and eclogites but also the granulites of this unit locally contain relic UHP minerals such as coesite and diamond (Lardeaux et al. 2001;Naemura et al. 2011;Perraki and Faryad 2014;Jedlicka et al. 2015;Thiéry et al. 2015). It follows, that present-day HP-UHT granulites are most probably retrograded UHP-UHT felsic rocks.
In the Schwarzwald, the metabasites with a HP-UHT granulite heritage from north of Hohengeroldseck Castle can easily be connected with felsic HP-UHT metagranulites from the CSGC and SSGC described by Marschall et al. (2003). In the Vosges Mts, such rocks occur in close association with UHP-UHT garnet peridotites and such an association is typical for the 'Granulite (Gföhl) Unit' of the Moldanubian part of the BM (Altherr & Soder 2018). The fact that granulites from these two areas suffered severe retrogression has limited the interest to discover UHP phases in these rocks.
The lowermost 'Monotonous Unit' or 'Lower Gneiss Unit (LGU)' in the FMC consists of anatectic gneisses (mostly poor in K-feldspar) with intercalated small bodies of HP-HT eclogite and spinel peridotite. The rocks of this unit first suffered a HP-HT event that was followed by a LP-HT metamorphism, creating retrogressed eclogites surrounded by anatectic gneisses. At least all the metabasites from the Schwarzwald that contain eclogitogenic relics belong to this unit. The P-T conditions at which these eclogites were equilibrated can be narrowed down to ~ 2.0 GPa and 670-700 °C (this paper) and are thus nearly similar to those from the 'Monotonous Unit' of the BM (Faryad et al. 2013).
It is important to note that retrogressed eclogites also occur in the Böllsteiner Odenwald that forms part of the Mid-German Crystalline High. These eclogites were formed at P-T conditions (Will and Schmädicke 2001) that are similar to those at which the eclogites from the Schwarzwald were equilibrated. For these rocks Scherer et al. (2002) report Lu-Hf Grt-WR ages around the Famenian/Tournaisian boundary (~ 358 Ma). A paper with U-Pb ages (SIMS) of zircon from retrogressed eclogites of the Schwarzwald is in preparation (Hanel et al., unpubl. data).
Being aware that the two major units of the Moldanubian zone ('Monotonous Unit' and 'Granulite (Gföhl) Unit') are very similar in the Variscan chain of Central Europe (FMC, Vosges Mts, Schwarzwald, BM), it stays the problem of the plate tectonic setting by which these units were generated. Franke et al. (2017) argue for two subduction zones with opposite vergence underneath Bohemia (S-Armorica). These authors assume that the Saxothuringian Ocean was subducted to the SE, while the Galicia-Moldanubian Ocean was subducted to the NE as indicated by the Moravo-Silesian unit that was overthrust by the Moldanubian. Such a configuration, however, is hardly able to explain the presence of the two subduction-related units ('Monotonous Unit' with Eclogites metamorphosed at ~ 2 GPa and 670-700 °C, and 'Granulite (Gföhl) Unit' metamorphosed at ≥ 4 GPa and 950-1030 °C). Exhumation of such units normally occurs onto the lower plate, yet these units are part of the upper plate.

Conclusions
In the Moldanubian Schwarzwald, two large gneiss complexes are tectonically separated by the BLZ. In both gneiss complexes, metabasic rocks occur as small inclusions (< 800 m in diameter) within amphibolite-facies plagioclase-biotite gneisses of predominantly greywacke compositions. The rocks of these complexes suffered at least two major metamorphic events, the last of which was an allochemical amphibolitefacies overprint at 0.4-0.5 GPa / 675-690 °C and high water activities, which led to anatexis in gneisses with suitable compositions. The preceding high-pressure metamorphic event is predominantly documented in metabasic rocks. Most of these contain mineral relics (garnet, omphacite, rutile, ± kyanite) of an earlier eclogite-facies metamorphism at 2.0 GPa and 670-700 °C. The thermal gradient during high P/T metamorphism thus was 10-13 °C/km, corresponding to an average to warm top-of-the-slab geotherm. Locally, however, there are metabasites that were metamorphosed during a HP-UHT granulite-facies event (1.5-2.0 GPa/950-1000 °C), before they were overprinted by the later amphibolite-facies metamorphism. Finally, there are amphibolites without any relics of a former HP-HT or HP-UHT metamorphism.
In this study, we investigated the chemical compositions of all these metabasic rocks. We found no systematic differences between samples from the three groups. In general, the rocks underwent strong metasomatism during and after the second metamorphic event, the amphibolite-facies metamorphism with beginning anatexis. Elements such as LILE (Cs, Rb, Ba, K, Pb, ± Sr) and U, which are highly soluble in aqueous fluids, were introduced to the rocks in variable proportions. Elements with lesser mobility, such as REE and HFSE were not significantly changed. Thorium might have been mobile in some samples showing low Th/La ratios.
The first event that the metabasites might have undergone was submarine alteration and/or hydrothermal metamorphism, eventually leading to an increase in LILE and U as well as changes in the relative amounts of Si, Mg, Fe, Ca, and Na. We, therefore, did not use all these potentially mobile elements for the determination of the primary volcanic rocks.
During subduction (groups 1 and 2) or during granulitefacies metamorphism (group 3), the rocks became strongly dehydrated under elevated P-T conditions. Most of our samples were transformed to eclogites that were almost free of OH-bearing phases such as amphibole or epidote. We, therefore, assume that the chemical budgets of such nearly dry eclogite samples did not contain large amounts of LILE and U that are well solvable in aqueous fluids liberated during high P/T metamorphism leading to eclogitization.
Chemical discrimination diagrams based on elements such as Th, HFSE, REE, and Y, that are largely immobile in aqueous fluids at P-T conditions below 2 GPa and 800 °C, strongly suggest that the former basaltic/andesitic rocks (groups 1-3) were island arc tholeiites, back-arc basin basalts and MORB.