Proterozoic VanDieland in Central Victoria: ages, compositions and source depths for late devonian silicic magmas

Abstract The Proterozoic to Cambrian VanDieland microcontinent was accreted to mainland Australia at ca 400 Ma, and its northern tip, the Selwyn Block, forms the basement in central Victoria. Here, mainly Late Devonian, silicic magmas were derived from the Selwyn Block and intruded into the shallow crust. We use the phase petrology of Late Devonian, S-type rhyolitic ignimbrites and a xenolith of pelitic migmatite, together with Nd-model ages for the silicic magmatic rocks to constrain the lithological characteristics of the metasedimentary component of the Selwyn Block, to infer minimum depths and temperature conditions here in the Late Devonian, and the likely ages of the source rocks for the S-type magmas. The most abundant source rocks are inferred to be volcaniclastic metagreywackes, with minor metadacites, meta-andesites and metapelites. The metapelitic xenolith cannot have been the source for any of the silicic magmas but constrains the upper amphibolite-facies part of the Selwyn Block to depths around 17 km, where temperatures reached ∼775 °C. The older ignimbrite magma was formed by partial melting at perhaps 770 °C and a depth of at least 33 km, while the younger ignimbrite magma formed at ∼23 km and 900 °C. These depths suggest source rocks in the Paleoproterozoic to Mesoproterozoic lower parts of the Selwyn Block. Nd-model ages of the silicic magmatic rocks confirm a dominance of Mesoproterozoic to Paleoproterozoic sources. If the inferred rock types in the Mesoproterozoic formations were as current correlations suggest, the sources for the Late Devonian silicic magmas would have to lie in the undocumented Paleoproterozoic basement of the Selwyn Block. Rock types here must include andesitic to dacitic volcanic components as well as volcaniclastic greywackes and minor pelites, which suggests a continental arc setting. The Late Devonian magmatism in the region may record the progression from amphibolite- to granulite-facies conditions during post-orogenic extension, with heat advected to the crust by mantle-derived mafic magmas. These processes would have resulted in mafitisation of the deep Selwyn Block.


Introduction
At about 400 Ma, a continental fragment known as VanDieland (Cayley, 2011) collided with the southern Australian mainland. This microcontinent, with apparent geological affinities in western Tasmania, forms a large, wedge-shaped, seismic feature that, at its northern end, forms the basement to central Victoria and tapers northward to a point roughly coinciding with the present location of the Murray River ( Figure 1). In Victoria, this basement terrane is known as the Selwyn Block (Cayley, Taylor, VandenBerg, & Moore, 2002;Rossiter & Gray, 2008) and, for the most part, it lies structurally beneath the Paleozoic rocks of the Melbourne Zone and parts of the adjacent Bendigo and Tabberabbera zones. The Neoproterozoic to Cambrian upper parts of the Selwyn Block do have sparse outcrops in the Barabool Hills and Dog Rocks (near Geelong and Batesford), on Phillip Island, around Waratah Bay, in the Glen Creek Inlier (SE of Mt Strathbogie) and near Licola. However, interpretations of seismic sections through the Victorian crust (Moore, Betts, & Hall, 2016) indicate that a thick sequence of Mesoproterozoic to Paleoproterozoic rocks lies at the base of VanDieland, with the Moho at around 35 to 40 km ( Figure 2). The presence of the Selwyn Block in Victoria is a geophysical reality. The structural thickness of the deformed Paleozoic rocks is 14 km in the Melbourne Zone and the Moho lies at around 35 km. Deep seismic data show clearly that the basement to the Melbourne Zone has a different structure to that of the adjacent Bendigo Zone (e.g. Willman et al., 2010).
Paleozoic rocks of the Melbourne Zone nor the Neoproterozoic to Cambrian rocks, which are inferred to be in the upper parts of the Selwyn Block, form chemically or isotopically suitable sources for the granitic magmas; and (2) that the granite source terrane was most likely dominated by greywackes with variable original clay contents, together with minor pelites, andesites and dacites. This likely assemblage of source-rock types strongly suggests accumulation in the back-arc basin of an Andean-type margin. Given the generally high-T, H 2 O-undersaturated character of the granitic magmas that were produced here (references cited above), the source terrane must now consist mainly of granulite-facies rocks that are significantly depleted in their granitic melt components, i.e. they are strongly restitic.
With the structure shown in Figure 2, and the constraints on the required rock types, it seems likely that the sources for the Late Devonian S-and I-type silicic magmas must lie at the depth of the unseen Mesoproterozoic and most probably the Paleoproterozoic parts of the Selwyn Block. From the constraints already mentioned, these rocks are likely to be volcanic arc-related metasediments and Melbourne N 20 km  Zone; TZ, Tabberabbera Zone; BaZ, Bassian Zone), their bounding faults, and the principal outcrops of the mainly Late Devonian silicic magmatic rocks (granitic plutons and silicic ignimbrite complexes). Bodies referred to in this paper are named, designated as either S-or I-type and their ages (where well constrained) are given. Red colours indicate S-type character and green I-type character. Some I-type volcanic rocks are present in the Marysville Igneous Complex and the Mount Dandenong Volcanic Group but are volumetrically insignificant. White dots indicate the locations of the samples (S9, 9446 and T20) for which phase relations have been calculated. The inset shows the location of the city of Melbourne within the continent. meta-igneous rocks. Additionally, Clemens and Phillips (2014) considered the temporal evolution of S-type magma compositions in the Melbourne Zone, to suggest that the original sedimentary component of the source for the granitic magmas may have been deposited in a basin that had been widening and deepening with timesuggesting more clay-rich sediments toward the upper stratigraphic levels. Following on from that, Clemens and Elburg (2016) used the Sr-and Nd-isotope characteristics of the silicic igneous rocks in central Victoria to suggest a lateral compositional and age structure in the magma source rocks. From west to east, the sources of the I-type magmas seem to become less evolved (younger?) while the sources of the S-type magmas show the opposite trend, becoming more evolved toward the east, possibly reflecting an eastward deepening of the original arc basin.

Additional constraints on the structure of the deep Selwyn Block
In this contribution, we explore three additional sources of information on the locations and compositions of the rocks in the Selwyn Block. Firstly, we have the high-grade (migmatitic) metapelitic xenoliths in some of the granitic batholiths here, notably in the S-type Mt Wombat pluton of the Strathbogie batholith (Clemens & Phillips, 2014) and the Itype Baynton and S-type Pyalong plutons of the Cobaw batholith (Anderson, 1997;Clemens et al., 2016a). These provide direct evidence of the physicochemical conditions in the mid-to deep-crustal rocks. Note that we do not propose that these xenoliths represent the actual sources of the granitic magmas. Indeed, the granitic magmas cannot have had dominantly metapelitic sources, and they were produced by biotite-breakdown reactions that had not occurred in the xenoliths (Clemens & Wall, 1981). We are simply proposing that these accidental xenoliths, picked up by the granitic magmas during ascent, will have sampled parts of the Selwyn Block, at higher levels than those where the magmas originated. Thus, these provide minimum source depths. Secondly, we reassess P-T constraints from some of the Late Devonian, S-type, silicic ignimbrite deposits in the Violet Town Volcanics (Clemens & Wall, 1984) and the Tolmie Igneous Complex (Clemens et al., 2011). These represent samples of magmas formed by partial melting of rocks that are inferred to lie beneath the Paleozoic cover in the region. Our approach is to use calculated pseudosections to constrain conditions of magma formation, and thereby locate the depths of the magma sources and those of the rocks in the Selwyn Block that are not the magma sources. Thirdly, we have information from the Nd-model ages calculated for the various granitic intrusions in the region. This information is, by far, the least diagnostic of the three lines of evidence. However, in combination with the others, it provides verification as well as provisional constraints on the ages of the magma sources.
We have not examined phase relations in I-type rocks because I-type volcanic rocks are rare in this region, there are no experimental studies on their phase relations and our own reconnaissance survey of their calculated phase relations suggest that they provide no useful pressure constraints on the origins of the magmas. Thus, although we infer that substantial volumes of meta-andesitic and metadacitic volcanic rocks are present in the deep crust here, at this stage we cannot meaningfully constrain their locations or depths.

Metapelitic xenoliths in granitic rocks
Schistose, high-grade, metasedimentary rocks are relatively commonly encountered as xenoliths in granitic plutons of the region. These are chemically dissimilar to the exposed Paleozoic metaturbidites (see below) and have mineral assemblages of cordierite-sillimanite-quartz-plagioclase, with K-feldspar and garnet being present in some examples (Anderson, 1997). The presence of quartzo-feldspathic leucosomes in some of these xenoliths is interpreted to indicate that the rocks had been partially melted (Anderson, Price, & Fleming, 1998).
For this study, we chose sample S9, a xenolith of upper amphibolite-facies metapelitic, stromatic migmatite that was recovered from the Mt Wombat pluton of the Strathbogie batholith (366 Ma; Kemp et al., 2008). Clemens and Phillips (2014) published its major-, minor-and traceelement composition, as well as its initial 87 Sr/ 86 Sr and eNd values. Table 1 shows its major-oxide composition. This rock is far more Na-, Ca-, Fe-and Mg-rich than the Paleozoic quartz-rich turbidites in the Melbourne Zone, and indeed in the wider region (e.g. L. A. I. Wyborn, & Chappell, 1979;D. Wyborn, & Chappell, 1983), the inference being that it is likely to have been derived from the Selwyn Block. Supplementary papers 1 and 2 provide the Theriak/ Domino modelling parameters and the calculation of the composition used in this modelling, respectively. S9 ( Figure 4) is a schist with foliae rich in biotite and poikilitic cordierite alternating with quartzo-feldspathic domains that have igneous textures (low dihedral angles between grains, euhedral cordierite prisms and tabular plagioclase crystals). The cores of some cordierite anhedra in the mafic foliae contain clusters of fine fibrous sillimanite and tiny octahedra of hercynitic spinel. K-feldspar does not seem to have been a stable sub-solidus mineral but it crystallised from the former granitic melt in the rock. The presence of hercynite and sillimanite in the cores of some cordierite crystals suggests the possibility that a quartzabsent decompression reaction such as Grt þ2Sil ¼ Crd þ Hc may have occurred in these rocks (mineral abbreviations as recommended by Whitney & Evans, 2010). This possibility cannot be confirmed because there are no remnant garnets, but it is possible that the original pressure could have been higher than inferred from the present mineral assemblage in S9. Figures 5 and 6 show the results of the Theriak-Domino modelling; Figure 5 as T-M H2O pseudosections at 300, 400 and 500 MPa (3, 4 and 5 kbar), and Figure 6 as a P-T pseudosection with M H2O ¼ 0.27 (1.15 wt% H 2 O). Note that, in these and subsequent pseudosections, phase-saturation boundaries are plotted, as is the usual procedure in studies of igneous systems. For some mineral species, there are curves shown for phase-in and phase-out. For example, with isobaric cooling, orthopyroxene would begin to crystallise at Opx-in but, with further cooling, this mineral commonly becomes unstable and the cooling path will then pass through the Opx-out boundary, below which orthopyroxene is unstable.  In Figure 5a there is no field for garnet above the solidus and no field for sillimanite (Als for aluminium silicate) at all. Ilmenite, plagioclase and cordierite are stable throughout. The apparent instability of sillimanite suggests that the original P of formation of S9 was probably >300 MPa (3 kbar). The grey-shaded field shows the area in which granitic melt would coexist with biotite and cordierite, in the absence of orthopyroxene. This implies temperatures between 700 and 790 C and M H2O about 0.27. Figure  5b is quite similar to Figure 5a, with quartz, plagioclase, cordierite and ilmenite stable throughout and neither garnet nor sillimanite stable above the solidus. Again, the grey-shaded field illustrates the conditions under which the assemblage in S9 would be stable -T between 750 and 800 C at M H2O about 0.27. Finally, in Figure 5c, quartz, plagioclase, cordierite and ilmenite are stable throughout and garnet is predicted to be stable whereas S9 contains none. The inference here is that, during partial melting, S9 equilibrated at P < 500 MPa (5 kbar).
Given the inference of M H2O ¼ 0.27 (an H 2 O content of 1.15 wt%), the P-T section in Figure 6 was constructed, at this value. Both quartz and plagioclase are stable over the entire range of P and T covered by the figure. From the petrography, we know that S9 was partially melted, so the P-T conditions to the left of the effective solidus (thick black line) can be ignored. There is no sign of rutile having been stable in the rock, so P must have been below about 700 MPa (7 kbar). Although sillimanite (fibrolite) wisps occur, armoured within the cores of some of the cordierite crystals, Al 2 SiO 5 phases do not appear to have been stable at the P-T conditions of partial melting. Likewise, although garnet may have been present at some stage, there is no trace of this phase in the present mineral assemblage. These features constrain the conditions of formation to <550 MPa (5.5 kbar) at 800 C. K-feldspar seems to have formed only as a product of in-situ crystallisation of former granitic melt, so the absence of K-feldspar from the assemblage further limits the possible peak conditions to the small, grey, wedge-shaped region in Figure 6 essentially around 400 to 500 MPa (4 to 5 kbar) and 750 to 800 C. The stable phase assemblage here would have been Qz þ Pl þ Bt þ Crd þ Ilm þ melt, as observed/inferred in the rock. If garnet and sillimanite had once been stable, the P may have been originally as high as 600 MPa (6 kbar), provided that cordierite was stable throughout but, in any case, lower than 700 MPa (7 kbar), as stated previously. Thus, for this part of the Selwyn Block, at the time of granitic magma formation at greater depth, our best estimate would be that migmatisation of the S9 protolith took place at about 450 MPa (4.5 kbar) and 775 C. Since Late Devonian, subaerially-deposited ignimbritic rocks are widespread and voluminous in central Victoria, we suggest that erosion need not be factored into any calculation of the depth of origin of the S9 protolith. Using an average crustal density of 2700 kg/m 2 , we thus infer a depth of about 17 km, with the source rocks for the host Mt Wombat monzogranite magma at some depth below this, where the Selwyn Block would have been at higher metamorphic grade.

Phase relations among phenocrysts in S-type ignimbrites
Aside from the granitic intrusive rocks, central Victoria boasts several voluminous, silicic volcanic complexes, also Late Devonian in age, and inferred to have similar deepcrustal sources to the plutonic granitic magmasthe Mount Dandenong Volcanic Group (Edwards, 1956), the Marysville Igneous Complex (Birch & Gleadow, 1974;Clemens & Birch, 2012), the Tolmie Igneous Complex (Clemens et al., 2011), the Violet Town Volcanics (Clemens & Wall, 1984;White, 1954) and the Macedon Igneous Complex (Mikucki, 1991). We chose two already well characterised samples for this study. Sample 9446 is a low-silica rhyolite ignimbrite from the Violet Town Volcanics (Clemens, 1981;Clemens & Wall, 1984) and sample T20 is a high-Ba, strongly ferroan, low-silica rhyolite from the Ryans Creek Ignimbrite of the Tolmie Igneous Complex (Clemens et al., 2011). Both these rocks have S-type chemistry and mineralogy, and their major-oxide compositions are given in Table 1. The Supplementary papers show the calculation of the compositions used in the Theriak-Domino modelling Figure 6. P-T pseudosection for metapelitic xenolith S9, calculated using Theriak-Domino and the H 2 O content inferred from the plots in Figure 3 (M H2O ¼ 0.27 ¼ 1.15 wt% H 2 O). The effective solidus is marked as a thick black line. Each mineral saturation boundary is labelled on the side on which the mineral is stable; the colour coding is shown in the legend. Mineral abbreviations are as recommended by Whitney and Evans (2010). Quartz and plagioclase are stable throughout the plotted area. The grey-shaded area represents possible formation conditions. See text for further explanation. and information on the modelling parameters. Below, we provide some relevant petrographic information that we use to constrain the conditions of early crystallisation in these rhyolitic magmas (and therefore minimum P and T of magma formation). We then show the graphical results and explain the reasoning behind the interpretations of the pseudosections.

Violet Town Volcanics
These 373 Ma (Kemp et al., 2008) low-silica rhyolite ignimbrites are relatively crystal-rich (perhaps 35 vol% in samples such as 9446) and contain commonly shattered (but formerly euhedral) phenocrysts of quartz with b-habit, plagioclase tablets with oscillatory and normal zoning, prismatic, single-crystal cordierite (mostly fresh), biotite plates, and minor K-feldspar tablets, orthopyroxene prisms and corroded almandine garnets (with cordierite-orthopyroxene decompression coronas). Accessory minerals include zircon, ilmenite, apatite, monazite and xenotime. The quartzo-feldspathic groundmass represents recrystallised glass and pumice fragments but, close to the base of the pile, fine eutaxitic fabrics are preserved. Figure 7a, b shows pseudosections for rhyolite ignimbrite sample 9446, calculated for H 2 O contents of 3 and 4 wt%, respectively. The choice of these values is in line with the findings of Wall (1981, 1984), Clemens and Birch (2012) and Clemens and Phillips (2014) for the initial melt H 2 O contents of compositionally very similar Late Devonian, S-type granitic and rhyolitic magmas of the Strathbogie batholith and the Cerberean Cauldron in the Marysville Igneous Complex. Clemens and Wall (1984) identified the earliest phenocryst assemblage in this rock as Qz þ Pl þ Grt þ Opx, with accessory Ilm. Cordierite, biotite and K-feldspar are present also, but form a later generation of phenocrysts, inferred to have formed prior to eruption, in the subvolcanic magma chamber. Decompression coronas of Opx þ Crd on the early garnets are compatible with this inference. It is clear from Figure 7 that, within the range of chosen H 2 O contents, earliest crystallisation of these rhyolitic magmas must have taken place at the conditions of the small grey-shaded areas, close to 600 MPa (6 kbar) and 900 C. Essentially the limits are prescribed by the presence of garnet and the absence of both rutile and cordierite in the near-liquidus mineral assemblage. Clemens and Wall (1981) showed that the most likely source rocks for the Late Devonian S-type granitic magmas are metagreywackes. Al 2 SiO 5 minerals are unstable in the presence of these particular magmatic liquids, so this effectively rules out any high abundance of aluminous metapelites in the magma source terrane. These results suggest two things about the metagreywacke-dominated source in the Selwyn Blockfirst that the depth to these rocks is at least 23 km and second that the crust at this depth, and below, must now consist of granulite-facies metamorphic rocks, at least part of which should be restitic in character, i.e. depleted in the granitic melt component.

Ryans Creek Ignimbrite
The Ryans Creek Ignimbrite (377 Ma; Clemens, Frei, & Finger, 2014) is crystal-poor but contains phenocrysts of  (Clemens & Birch, 2012;Clemens & Wall, 1981, 1984. The effective solidus is marked as a thick black line and, for clarity, some critical mineral saturation boundaries are labelled. The legend shows the colour coding used, and all mineral abbreviations are as recommended in Whitney and Evans (2010). In general, mineral phases become stable on cooling to T lower than the relevant phase boundary (¼ mineral saturation boundary). Mineral-in curves mean that the mineral appears here on cooling from higher T, and mineral-out curves mean that the mineral is no longer stable as T falls below the curve. Rutile and garnet are stable at pressures higher than their respective phase boundaries and cordierite at pressures below the marked boundary. The grey-shaded area represents inferred initial crystallisation conditions. See text for further explanation. embayed quartz with b-habit, K-rich alkali feldspar, minor plagioclase, ilmenite, trace amounts of euhedral almandinerich garnet and cordierite euhedra (mostly altered to clays) and very rare biotite plates (commonly altered to chlorite). The groundmass is composed of devitrified glass (microcrystalline quartz and feldspar) with well-preserved fiamme and eutaxitic fabrics. The garnet phenocrysts typically contain inclusions of euhedral to subhedral ilmenite, prismatic subhedral to anhedral rutile and euhedral apatite, monazite and zircon.
From the work of Clemens et al. (2011), a reasonable estimate of the initial (near-liquidus) H 2 O content of the T20 magma is 4 wt%. We have therefore calculated a P-T pseudosection for this rock with 4 wt% H 2 O in the system (Figure 8). The rock contains phenocrysts of biotite, quartz, plagioclase, alkali feldspar, cordierite and garnet (with rutile inclusions). The stability of the garnet-rutile assemblage constrains early-magmatic crystallisation to P > 840 MPa (8.4 kbar) at T < 810 C. Figure 8 is also contoured with values for the Alm content of the garnet (% 100 Â mol. Fe/ [Fe þ Mn þ Mg þ Ca]). The cores of the garnet phenocrysts in the rock have compositions in the range Alm 82-90 (Clemens et al., 2011), so this constrains early crystallisation to have occurred within the grey triangular area in the figure, at 910 to 1020 MPa (9.1 to 10.2 kbar) at 745 to 770 C. At these conditions, kyanite is predicted to be stable, but no Al 2 SiO 5 phase has been identified in the rocks of this complex. Our modelling suggests that kyanite would have been modally insignificant (< 0.5 vol%) and would have readily reacted out when the magma rose to the subvolcanic magma chamber (at P < 150 MPa or 1.5 kbar) from which the magmas were eventually erupted (Clemens et al., 2011). The cordierite phenocrysts and ilmenite are inferred to have crystallised later, at much lower P, because the maximum pressure stability for cordierite in Figure 8 is approximately 460 MPa (4.6 kbar).
Prior to eruption, cordierite had joined the phenocryst assemblage but there is no sign of either orthopyroxene or fayalite. The phenocryst assemblage Qz þ Pl þ Kfs þ Bt þ Grt (Alm 82-90 ) þ Crd þ Ilm would be stable somewhere in the lower-P region marked in grey in Figure 8, at P between 230 and 430 MPa (2.3 to 4.3 kbar) and 750 to 780 C. Thus, the crust was rather thick in this location (just inside the Tabberabbera Zone), with the magma source at a depth of at least 33 km. Also, the calculated minimum temperature for magma genesis (745 to 770 C) suggests that some parts of the deep Selwyn Block may not have reached granulite grade at the time of formation of the Ryans Creek Ignimbrite magmas. On the other hand, Clemens et al. (2011) found that the overlying Toombullup Ignimbrite, a lower-silica rhyolite formation in the same Tolmie Igneous Complex, seems to have been formed at T > 850 C and P > 400 MPa (4 kbar). This suggests granulite-facies conditions, as for the source rocks of the neighbouring Violet Town Volcanics in the Melbourne Zone. At 377 Ma, the magmas of the Tolmie Igneous Complex appear to be the earliest of the S-type magmas generated in the region, apart from the Wilsons Promontory batholith, which lies in the far south of the Australian mainland and in the Bassian Zone, where the existence of the Selwyn Block basement is doubtful (Figure 3). Thus, it seems possible that the Late Devonian (Frasnian), silicic magmas actually record a rise in the temperature of the magma source, as metamorphism progressed. The proximity of the Tolmie Igneous Complex to the Governor Fault (the eastern edge of the Melbourne Zone; Figure 3) suggests the possibility that, early during the late-Devonian crustal heating event, there may have been some deep fluid ingress associated with this fault, allowing relatively low-T partial melting of the source rocks for the Ryans Creek Ignimbrite magmas.

Nd-model ages
Data for this compilation are taken mostly Black et al. (2010) and Clemens and Elburg (2016). Nd-model ages (t 2DM ) for the protoliths (Table 2) range from late Figure 8. P-T pseudosection for Ryans Creek Ignimbrite (low-silica rhyolite) T20, for a bulk H 2 O content of 4 wt%, calculated using Theriak-Domino. The thick black curve in the lower left is part of the effective solidus, so essentially the entire plot has silicate melt coexisting with the crystalline phases. Phase-saturation boundaries are shown for numerous minerals and colourcoded or dashed as given in the legend, with mineral abbreviations from Whitney and Evans (2010). The thin black curves with values are isopleths for mol% almandine in the garnet, with the isopleth for Alm 80 slightly thicker. In general, a mineral becomes stable on cooling to the temperature of its saturation boundary and is then stable at all temperatures lower than this. Exceptions include cordierite, garnet, fayalite and orthopyroxene, which have phase-in and phase-out boundaries. At pressures and temperatures lower than the phase-out boundaries, the mineral is no longer stable. Some of these boundaries have been labelled for clarity. The dashed black lines represent Al 2 SiO 5 phase boundaries, which include Als-in and Als-out, as well as the Ky to Sil reaction. Note that Als-out corresponds exactly with Ms-in. Two P-T regions have been shaded in grey to indicate early, higher-P and later, lower-P magmatic crystallisation. See text for further details and discussion.
Paleoproterozoic (1.66 Ga) to late Mesoproterozoic (0.99 Ga). The average model ages for the I-and S-types are 1.20 and 1.57 Ga, respectively, the difference being statistically significant at the 99% level of confidence. This difference is expected because S-type source rocks would commonly have been through longer periods of crustal recycling. Also, among the I-type rocks, the Lysterfield and Tynong plutons (both parts of the Tynong batholith) show signs of potential young mantle components, although the nature of those is debated (e.g. Clemens et al., 2016b;Regmi et al., 2016). Clemens et al. (2016b) noted the presence of mingled intermediate magmas and the paucity of evidence for widespread magma mixing in the Tynong pluton. Thus, these authors preferred a model in which the mantle signal is due to the magma source rocks being unweathered intermediate arc igneous rocks.
As is well known, Nd-model ages are subject to a raft of uncertainties, both in their calculation and interpretation. We have used two-stage model ages, after the scheme of DePaolo, Linn, and Schubert (1991). These ages are supposed to reflect the time at which the source rock was extracted from the depleted mantle. However, as discussed earlier, the granitic magmas are likely to contain material with ancient metasedimentary and much younger metavolcanic components (i.e. the sources are probably of mixed character). Thus, provided that the original mantle-derived volcanic rocks had mantle-like Nd isotope ratios, the most that can strictly be claimed is that the Nd-model ages are estimates of the minimum source ages. For the central Victorian granitic rocks, the petrological data support sources that are dominated by arc volcanic rocks and greywackes derived largely from them. If that is correct, the time at which these were separated from the mantle would probably not be greatly different from the model ages (always assuming that these are meaningfully calculated). Some support for this interpretation comes from consideration of the model ages calculated for the middle Cambrian to Lower Devonian metasediments that overly the Selwyn Block. These younger metasediments have Ndmodel ages that fall in the range 1.84 to 1.98 Ga (Adams, Pankhurst, Maas, & Millar, 2005). These are significantly earlier than any of the granite model ages and suggest that these Paleozoic rocks are not significant components in the sources of the Devonian silicic magmas. Table 2 also shows the average t 2DM values for I-and Stype granitic rocks in eastern and western Tasmania. Students t-tests on the data (Table 2; Black et al., 2010, table 2) indicate that, except in eastern Tasmania, the model ages of the I-type sources are younger than those of the S-type sources. Model ages for the I-type granitic rocks in the Melbourne Zone and both east and west Tasmania are the same, within error. However, the S-type granitic rocks in western Tasmania have significantly older model ages than those of the S-types magmas in eastern Tasmania, and only the western Tasmanian S-type rocks show similar model ages to the S-type rocks in the Selwyn Block. This confirms the inference that the Selwyn Block shows affinities with western Tasmaniathe VanDieland connexion (e.g. Moore et al., 2016;Moresi, Betts, Miller, & Cayley, 2014).
If Figure 2 is accurate, magma genesis occurred at depths between approximately 23 and 36 km (around 600 to 950 MPa or 6 to 9.5 kbar), depending on their location. However, some rocks (e.g. the S-type Warburton pluton, with an average model age of 1.65 Ga) may have somewhat deeper sources. The distillation of this is that the source rocks for the Late Devonian silicic magmas in central Victoria lie at depths that indicate that the Mesoproterozoic to Paleoproterozoic part of the Selwyn Block is the most probable source, and that, in terms of the sources for the granitic magmas, the Selwyn Block shows closest isotopic similarities with western Tasmania.

Conclusions
The meaning of the S9 xenolith Clemens and Phillips (2014) published the Nd and Sr isotope characteristics of this rock. It has 87 Sr/ 86 Sr 366 Ma ¼ 0.71990 (± 41), eNd 366 Ma ¼ -12.2 (± 0.2, assumed) and a Nd model age (t 2DM ) of 2.09 Ga. No Devonian S-type granitic magma in central Victoria has remotely similar isotopic characteristics, and the same applies to the xenoliths analysed by Anderson (1997). All the granitic rocks have signatures that indicate far less evolved rocks in their source terranes, with 87 Sr/ 86 Sr t in the range 0.70709 to 0.71685 (with two exceptional samples around 0.72), eNd t of -6.6 to 0.35 and Nd-model ages 1.65 Ga (Clemens & Elburg, 2016;. Additionally, the S-type magmas here were never saturated with an Al 2 SiO 5 mineral (Clemens & Wall, 1981) and so cannot have been derived as partial melts of an Al 2 SiO 5 -bearing metapelitic source. It thus seems clear that S9 cannot be representative of a major rock volume at the crustal level of the source of the S-type magmas. S9 also has isotopic characteristics that are dissimilar to those of the Paleozoic metasedimentary rocks in the region. These Paleozoic rocks have 87 Sr/ 86 Sr 370 Ma between 0.71475 and 0.73573, eNd 370 Ma of -12.2 to -9.4 and Ndmodel ages of 1.90 to 2.14 Ga (Anderson, 1997;Turner, Foden, Sandiford, & Bruce, 1993). Also, as mentioned earlier, S9 has a major-oxide composition that is quite distinct from those of the Paleozoic, quartz-rich metaturbidites. Thus, S9 is most plausibly derived from an upper amphibolite-facies part of the Selwyn Block. The P-T constraints on its equilibration conditions, when carried to the shallow crust in the Mt Wombat monzogranitic magma, suggest a mid-crustal origin at about 17 km depth, where the temperature reached about 775 C during the metamorphism that resulted in granitic magma formation at greater depths. From Figure 2 we would infer a protolith age of Mesoproterozoic or Neoproterozoic for S9.

Depths and temperatures in the silicic magma sources
From the modelling, above, it appears that the sources of the granitic and silicic volcanic magmas that were emplaced within the area underlain by the Selwyn Block must lie at depths greater than 17 km ($ 23 to 33 km). During silicic magma genesis source temperatures rose to between 745 and 900 C and, in most places, > 850 C. Thus, the bulk of the Selwyn Block must have reached granulite-facies conditions. The loss of the silicic magma fractions would have served to densify and somewhat mafitise the deep Selwyn Block crust (e.g. Clemens, 1990). Since the likely heat source for the metamorphism and partial melting was under-and intra-plated mantle-derived magmas (e.g. Petford & Gallagher, 2001), the deep Selwyn Block must now be quite mafic in character.

Ages and lithological character of the silicic magma sources
Based on the inferred depths of origin of the granitic and silicic volcanic magmas, the geophysical interpretations of Moore et al. (2016) suggest sources of Mesoproterozoic to perhaps Paleoproterozoic ages. However, if the inferred Mesoproterozoic part of the Selwyn Block contains the rock types specified by Moore et al. (2016), we would conclude that the sources for the granitic magmas most probably lie in the deeper, undocumented, Paleoproterozoic parts of the block. The Nd-model ages of the granitic and silicic volcanic rocks also suggest sources that were at least as old as the Mesoproterozoic. Inferences on source rock type come from the phase relations and chemistry of the silicic magmatic rocks and require the source to consist of an assemblage of metavolcanic and metasedimentary materials that were originally formed in a continental back-arc setting. The basement rocks in the Melbourne Zone are between 250 and 450 km north of the inferred correlatives in Tasmania. Thus, it could still be that the formations present in this northern tip of VanDieland differ from those identified further south.
Late Devonian high-grade metamorphism As indicated above, the Late Devonian in central Victoria was characterised by the emplacement of large volumes of H 2 O-undersaturated, high-T silicic magmas. As shown by Brown and Fyfe (1970), Clemens (1990) and Clemens and Watkins (2001), this requires the source rocks to have reached fluid-absent, granulite-facies metamorphic conditions. Since the crust of the Selwyn Block only produced voluminous silicic magmas in the Late Devonian, this implies that, prior to the heating event, these rocks must all have been at lower metamorphic grade and still fertile (with respect to their potential to yield silicic partial melts). The critical crustal heating event occurred in a late, posttectonic, extensional regime, and the only credible source of heat would be the emplacement of mantle-derived magmas as an underplated sheet or, more likely, as numerous thinner sheets emplaced sequentially (e.g. Annen, 2011;Petford & Gallagher, 2001). Together with the data for S9, the Anderson (1997) data for the high-grade metamorphic xenoliths in the granitic rocks can be plotted on a Rb-Sr errorchron. Using IsoplotR (Vermeesch, 2018) with a model-3 regression (maximum likelihood with overdispersion), and excluding one outlying analysis, this yields an apparent date of 364 ± 1 Ma, with 13 data points. These xenoliths show no textural signs of having been hornfelsed, and still retain their regional metamorphic fabrics, complete with foliation-defining biotite crystals. Thus, it seems possible that this reflects the date of the thermal event responsible for high-grade metamorphism, partial melting and the production of the Late Devonian granitic magmas.
As a result of the granulite-facies event, the deep parts of the Selwyn block must now be restitic in character, with mineral assemblages involving garnet and orthopyroxene in the metasedimentary rocks and two pyroxenes in the intermediate to mafic rock types. There would also be substantial sheets of Late Devonian metabasaltic magmas, derived from the heat source and also now in the granulite facies. well as an incisive review. Roland Maas also reviewed the manuscript and made many useful comments and suggestions.