The Mesozoic and Palaeozoic granitoids of north-western New Guinea

Abstract A large portion of the Bird's Head Peninsula of NW New Guinea is an inlier that reveals the pre-Cenozoic geological history of the northern margin of eastern Gondwana. The peninsula is dominated by a regional basement high exposing Gondwanan (‘Australian’) Palaeozoic metasediments intruded by Palaeozoic and Mesozoic granitoids. Here, we present the first comprehensive study of these granitoids, including field and petrographic descriptions, bulk rock geochemistry, and U–Pb zircon age data. We further revise and update previous subdivisions of granitoids in the area. Most granitoids were emplaced as small to medium-scale intrusions during two episodes in the Devonian–Carboniferous and the Late Permian–Triassic, separated by a period of apparent magmatic quiescence. The oldest rocks went unrecognised until this study, likely due to the younger intrusive events resetting the K–Ar isotopic system used in previous studies. Most of the Palaeozoic and Mesozoic granitoids are peraluminous and in large parts derived from partial melts of the country rock. This is corroborated by local migmatites and country rock xenoliths. Although rare, metaluminous and mafic rocks show that partial melts of mantle-derived material played a minor role in granitoid petrogenesis, especially during the Permian–Triassic. The Devonian–Carboniferous granitoids and associated volcanics are locally restricted, whereas the Permian–Triassic intrusions are found across NW New Guinea and further afield. The latter were likely part of an extensive active continental margin above a subduction system spanning the length of what is now New Guinea and potentially extending southward through eastern Australia and Antarctica.

Abstract: A large portion of the Bird's Head Peninsula of NW New Guinea is an inlier that reveals the pre-Cenozoic geological history of the northern margin of eastern Gondwana. The peninsula is dominated by a regional basement high exposing Gondwanan ('Australian') Palaeozoic metasediments intruded by Palaeozoic and Mesozoic granitoids. Here, we present the first comprehensive study of these granitoids, including field and petrographic descriptions, bulk rock geochemistry, and U-Pb zircon age data. We further revise and update previous subdivisions of granitoids in the area. Most granitoids were emplaced as small to mediumscale intrusions during two episodes in the Devonian-Carboniferous and the Late Permian-Triassic, separated by a period of apparent magmatic quiescence. The oldest rocks went unrecognised until this study, likely due to the younger intrusive events resetting the K¬-Ar isotopic system used in previous studies. Most of the Palaeozoic and Mesozoic granitoids are peraluminous and in large parts derived from partial melts of the country rock. This is corroborated by local migmatites and country rock xenoliths. Although rare, the metaluminous and mafic rocks show that partial melts of mantle-derived material played a minor role in granitoid petrogenesis, especially during the Permian-Triassic. The Devonian-Carboniferous granitoids and associated volcanics are locally restricted, whereas the Permian-Triassic intrusions are found across NW New Guinea and further afield. The latter were likely part of an extensive active continental margin above a subduction system spanning the length of what is now New Guinea and likely extending southward through eastern Australia and Antarctica.

INTRODUCTION 33
North-western New Guinea represents part of the northern boundary of the Australian Plate and has 34 experienced much Eocene to Recent tectonic activity ( Fig. 1A; e.g., Baldwin et al., 2012;Davies, 2012;Hall, 35 2012; Pigram and Davies, 1987). While many young tectonic features are reported from the region, such 36 as Miocene-Pliocene high-pressure metamorphic rocks (Bailly et al., 2009;François et al., 2016)  control on the timing of magmatism in this region are K-Ar ages from 19 samples (Bladon, 1988;Dow et 59 al., 1988). None of these K-Ar ages have been formally published, some lack associated uncertainties or pellets, using a PVP-MC (Polyvinylpyrrolidone-Methyl Cellulose) gluing agent. For this, matrix corrections 151 were calculated from the major element compositions and calibrated against up to 40 international 152 standards. Limits of detection for the major and trace elements were determined using long-term 153 reproducibility data. The software GCDkit 3.00 (Janoušek, 2006) was used for various calculations and 154 plotting. Iron was measured as total Fe2O3, but recalculated and plotted as total FeO (FeOt). 155  (Petrus and Kamber, 2012). In addition, TEMORA 2 (416.78 ± 0.33 Ma; Black et al., 169 2004) was measured as a secondary standard and treated as an unknown during data reduction. This 170 allowed the calculation of an excess variance for each measurement session, which was subsequently 171 propagated onto the internal uncertainties of each individual measurement of the respective session. 172 Measurements with discordance ≤10% (within 1s) were treated as concordant and only the 206

Our division of the igneous rocks of NW New Guinea into different lithostratigraphic units builds on 181
associations made in previous studies (e.g., Dow, 1988;Pieters et al., 1983;1990, Visser andHermes, 182 1962) but also accommodates the necessary changes required by new petrographic, geochemical, and 183 geochronological data. Our new subdivision therefore differs from that of previous studies in some 184 respects (and therefore differs from that shown in Table 1). This section describes our findings according 185 to the new subdivision, the reasoning behind which is discussed in Section 5. Kemum Formation (Figs. 2, 3A). The name is derived from the Ngemona River, which drains Lake Giji into 199 the Warjori River (Fig. 2). It is a medium-to coarse-grained leucogranite, consisting of quartz, plagioclase, 200 K-feldspar (mostly microcline), primary muscovite, and often garnet and tourmaline, but lacks biotite (Fig.  201 4A). The rocks are highly evolved and grade into medium to coarse-grained pegmatites. The pegmatites 202 can contain varying amounts of primary muscovite, tourmaline, and rare garnet, but are often purely 203 quartzofeldspathic. To distinguish primary from secondary muscovite the textural conditions introduced 204 by Miller et al. (1981) were applied: Primary muscovite must (1) have a relatively coarse grain size, 205 comparable to other phases; (2) show clear crystal terminations; (3) not enclose or not be enclosed by a 206 mineral from which muscovite may have formed from alteration; and (4) be a constituent of an unaltered 207 rock with clear igneous texture. Opaques, apatite, and zircon are common accessory phases; titanite is 208 rare. Plagioclase is preferentially sericitised compared to K-feldspar and often shows core-and-mantle 209 structures (Fig. 5A). Myrmekites are common where plagioclase has replaced K-feldspar (Fig. 5B). 210 Submitted to Lithos 8 211 The Wasiani Granite (new name) was the only igneous unit not observed in situ and, like Pieters et al., 212 (1990), we were restricted to a single alluvial sample (BJ92). The unit consists of leucocratic granite 213 containing abundant quartz, plagioclase, and K-feldspar, supplemented with characteristic biotite that is 214 partially retrogressed to chlorite (Fig. 4B). Zircon and apatite are abundant accessory minerals. Another 215 alluvial sample of a melanocratic monzonite was found (BJ93) at the same location. This rock contains 216 amphibole and partially retrogressed biotite next to abundant plagioclase and K-feldspar. Accessory 217 minerals include opaques, titanite, apatite, and zircon (Fig. 4C). Along the same river transect, 218 downstream (ENE) of where the two float samples were collected (Fig. 2), the country rock is crosscut by 219 leucocratic pegmatite dykes (~5 120 cm thick), which were sampled in situ (BJ98, BJ104A; Fig. 4D). 220 These pegmatites contain abundant muscovite and rare kyanite and tourmaline. One pegmatite dyke 221 generated a narrow zone of contact metamorphism (hornfels) in the country rock that was subsequently 222 folded and boudinaged (Fig. 3B). 223

224
The Kwok Granite (new name) is exposed in the easternmost part of the Kemum Basement High (Fig. 2)  225 where it intrudes the highest-grade metamorphic basement rocks as dykes or small stocks, which are 226 metres to tens of metres thick. The leucocratic granitoid contains fractured garnet, partially resorbed but 227 likely primary muscovite, and abundant sillimanite (fibrolite) next to partially resorbed biotite, abundant 228 retrogressed plagioclase, orthoclase, and quartz ( Fig. 4E, 5C). Opaques, apatite, and zircon are common 229 accessory phases. Fibrolite grows on biotite sheets (Fig. 5C) and seems to replace feldspars (Fig. 4E). The 230 rock also displays a weak gneissose texture and is locally associated with coarse-grained pegmatites. In 231 close proximity to the Kwok Granite (locality BH15-088), metapelitic metatexites are exposed, showing 232 pockets and layers of leucosome within volumetrically dominant cordierite-bearing melanosome ( Fig. 6A-233 B). 234

235
The Melaiurna Rhyolite (name modified to reflect the petrography) is exposed at the very western 236 termination of the basement outcrop (Fig. 1B). While Visser and Hermes (1962) originally described this 237 as a granite, we classify it as a volcanic rock due to its subvolcanic, porphyritic texture: phenocrysts of 238 rounded and embayed quartz, muscovite, plagioclase, and orthoclase up to 15 cm in length are 239 surrounded by a red, microcrystalline groundmass composed of quartz and feldspar (Fig. 4F). The feldspars are almost completely replaced by sericite or clay minerals. Muscovite is marginally resorbed. In 241 parts, veins of translucent white calcite cut the rocks (BH15-027). We did not observe any contacts with 242 other units but walked across the covered contact to the overlying Aifam Group. This contact is likely a 243 nonconformity, as the Aifam Group unit starts with a coarse-grained red arkose at its base that contains likely represents a weathered granodiorite characterised by chlorite after biotite and secondary, 284 interstitial muscovite. It is associated with quartzofeldspathic pegmatite or leucogranite dykes (Fig. 4G) 285 that intrude the country rocks that are different from the metapelites associated with Devonian-286 Carboniferous granitoids. The country rocks encompass steeply dipping, black slates and layered, 287 amphibole-bearing, and quartz-rich rocks that we tentatively interpret as low-grade metavolcanic rocks. 288 The granodiorites are pervasively brittlely deformed and cut by several faults of various orientations. In 289 thin section, brittle deformation (cataclasis, fracturing, kinking) is ubiquitous and predominates over 290 ductile deformation (BLG recrystallization in quartz, deformation twins in plagioclase) ( The Anggi Intrusive Complex (revised unit) is exposed in the SE of the Kemum Basement High (Fig. 2) and 295 encompasses a range of lithologies. Sample BJ138 is a granite containing <1 mm garnets within sheets of 296 biotite next to quartz, feldspar, zircon, and opaques (Fig. 4H). A diorite consisting of plagioclase and 297 chlorite as the alteration product of biotite, titanite, and opaques is exposed at locality BH15-173 (sample 298 BJ134; Fig. 4I). In two leucocratic diorites (BJ124, LW13-6D), hornblende reacting to chlorite, plus garnet 299 coronae growing on biotite and quartz occur next to plagioclase, quartz, opaques, and zircon (Fig. 4J).
Country rock xenoliths and roof pendants are abundant ( Fig. 3C-D), and garnet crystals preferentially 301 occur in the vicinity of these. At locality BH15-174, a granite intrudes metacalcareous metasediments but 302 contains xenoliths of a metapsammitic to metapelitic rock. Despite the apparent contrast in lithology, 303 some of the xenoliths show bedding that dips sub-parallel to that of the country rock. Thus, it cannot be 304 said with certainty whether the xenoliths were entrained within the magma from lower structural levels 305 or stoped from the roof or the wall of the intrusion. The xenoliths have been stretched into pinch-and-306 swell structures, which were later crosscut by normal-separation faults (Fig. 3D). The unit is further 307 associated with schlieric, metapelitic diatexites ( Fig. 6C-D). Compared to the migmatites associated with 308 the Kwok Granite, the proportion of leucosome is clearly higher and the volumetrically subordinate 309 melanosome lacks restitic anhydrous mineral phases (such as cordierite, garnet, or orthopyroxene). The Maransabadi Granite is exposed on the eponymous small island to the east of the Bird's Head 332 Peninsula (Fig. 1B). The unit consists of medium-grained granite and granodiorite composed of quartz, 333 partially sericitised plagioclase and K-feldspar, chlorite as the alteration product of biotite, and opaques, 334 apatite, zircon, and titanite as accessory phases (Supp. Data File 7). 335

336
The Kwatisore Granite is exposed over a large area in the south-eastern Bird's Neck ( Most of the zircons analysed for this study are euhedral crystals with magmatic zonation visualised by CL 417 imaging (Fig. 12A). A few samples, however, yielded zircons that have a different morphology and 418 appearance in CL: for example, (1) predominantly xenomorphic grains without visible zonation (BJ93, Fig.  419 12B), (2) broken grains that were subsequently welded together by thin metamorphic zircon (BJ02, Fig. 12C), and (3) eu-to anhedral grains that are noticeably CL-dark and lack apparent zonation (BJ08, Fig.  421 12D). For the latter group, which includes samples BJ08, BJ09, and BJ80, many analyses had to be rejected 422 during the calculation of a mean age due to a large spread in ages (Fig. 10 igneous rocks found in certain areas that could not be clearly subdivided due to limited field observations. 439 There are also several samples or localities that were not classified to a particular intrusive unit, mainly 440 because of insufficient data (this includes samples BJ10, BJ80, and BJ83 and the tentative allocation of 441 monzonite BJ93 to the Wasiani Granite). 442 443

Episodes of magmatism 444
Carboniferous (329-328 Ma) and Permian-Triassic (257-223 Ma) magmatism was recognised and 445 described by previous workers (e.g., Bladon, 1988; Tab. 1), but ages corresponding to the oldest rocks 446 reported here (~355 Ma) have not been reported previously. This might reflect a sampling bias: As 447 previous authors collected all but one of their samples from the alluvium (Bladon, 1988), we can never be 448 sure that we dated exactly the same granitoid units they did. The ages previously reported for the Anggi 449 Granite (as defined by Pieters et al. (1983;1990)) agree with the U-Pb ages of the Anggi Intrusive Complex (as defined here). However, for the Wariki Granodiorite, Bladon (1988) reported five Permian-451 Triassic ages from an area, where we also found Devonian-Carboniferous ages of the Mariam Granodiorite 452 next to one sample of the Late Permian Wariki Granodiorite (sample BJ02). Also, U-Pb ages of the 453 granitoids in the higher-grade Kemum Formation north of lakes Giji and Gida (cf. Fig. 2) are exclusively 454 Palaeozoic (with the exception of sample BJ58) and apparently contradict to the Permian-Triassic ages 455 reported in Dow et al. (1988). Such discrepancies between the ages of previous studies and those 456 presented here cannot be explained by a sampling bias. 457

458
The potential discrepancy between ages reported here and those of previous authors (Bladon, 1988;Dow 459 et al., 1988) for the same granitoid unit could be due to resetting the K-Ar system by (1) intense alteration 460 or (2) a thermal event. We assume that the K-bearing minerals previously dated from alluvial samples (as 461 summarised in Bladon, 1988) were likely less altered and led to reliable K-Ar ages, as alluvial samples are 462 always fresh and resistant to weathering, compared to granitoid outcrops, which are often intensely 463 weathered. The resetting of the K-Ar system by a thermal event on the other hand is more likely due to a 464 number of reasons: (1) the extensive Permian-Triassic magmatism (Figs. 1, 2, 11); (2) a regional HT/LP 465 metamorphic event  with mineral assemblages suggesting temperatures in excess of 466 500°C; (3) the fact that previous K-Ar ages were predominantly obtained from biotite, muscovite, and 467 plagioclase (n = 24), which have lower closure temperatures for Ar than hornblende (n = 6) (e.g., Reiners 468 et al., 2005 and references therein); and (4) that previous authors themselves assumed that some of their 469 samples had been thermally disturbed and thus likely reset (Bladon, 1988). We therefore propose that the  ., 2002a; 2002b). Also, chessboard subgrains in quartz only develop above ~570°C at 1 kbar, 481 and even higher temperatures at higher pressures (Kruhl, 1996). Bulbous myrmekites (Fig. 6F) also 482 indicate recrystallization of the granitoids at similar metamorphic conditions (e.g., Phillips, 1980). 483 Although recrystallization can be a response to syn-or post-intrusive deformation of a cooling granitoid 484 body (e.g., Pennacchioni and Zucchi, 2013), a thermal overprint of older intrusive rocks in the eastern 485 Kemum Basement High is likely as is indicated by the significant Permian-Triassic magmatism and 486 concomitant regional HT/LP metamorphism (e.g., Pieters et al., 1990). 487 488 As previous authors (e.g., Dow et al., 1988;Pieters et al., 1983;1990) underestimated the diversity of ages 489 of granitoids from western New Guinea, it is possible that we have unintentionally done the same. A case 490 in point is the apparent 'magmatic gap' during much of the Permian implied by our analyses (Fig. 13). This 491 potentially reflects a sampling bias, particularly since a relatively small area of the Kemum Basement High 492 was sampled. We must also consider that there has been uplift of the region since the Miocene 493 accompanied by high erosion rates , so there may be igneous bodies that have not yet 494 been exposed at the surface or have already mostly been eroded away. Readers should also note that 495 Permian ages were previously reported for the Warjori Granite (Bladon, 1988;Pieters et al., 1990) and for 496 igneous rocks in the south of mainland New Guinea (Fig. 13) The weakly to highly peraluminous mineralogy and chemistry of most granitoids of NW New Guinea 502 indicate that they are primarily derived from partial melts of the metapsammitic to metapelitic country 503 rock and can thus be considered S-type granitoids (Chappell and White, 1974;1992;2001). Partial 504 melting of continental crustal material is supported by migmatites associated with the Kwok Granite and 505 the Anggi Intrusive Complex, which indicate incipient and pervasive partial melting, respectively (Fig. 6). 506 As migmatisation is not confined to the contact with intrusions, they are likely the result of regional 507 HT/LP metamorphism as opposed to contact metamorphism. Also, abundant metasedimentary xenoliths 508 in the Anggi Intrusive Complex corroborate the assimilation of and contamination with continental crustal 509 material (Fig. 3C-D). The xenomorphic to skeletal appearance of garnets in the Anggi Intrusive Complex, their association with biotite (Fig. 4H, J), the abundance of biotite and quartz inclusions within them, and 511 their seemingly preferred occurrence around country rock xenoliths further indicates that these are 512 restitic xenocrysts resulting from mica dehydration reactions (i.e., peritectic phases) and originated from 513 the country rock. It is likely that the xenoliths were incorporated and contributed to the melt at greater 514 depths (as garnet and not cordierite formed as a peritectic phase). Lastly, the presence of rounded and 515 concordant Precambrian zircon cores (Fig. 12E-H) provides additional evidence that the petrogenesis of 516 many of the granitoids involved partial melting of (meta)sedimentary material. The geochemical data of the granitoids need to be interpreted with caution as some of the analysed 565 samples show evidence of chemical weathering or metasomatic alteration. For instance, the anomalously 566 high Na 2O coupled with very low K2O and CaO contents of sample BJ134 may be indicative of albitisation 567 (Fig. 7). The high K2O and low Na2O contents of the Melaiurna Rhyolite may also reflect potassic alteration. 568 that numerous deformation phases have occurred since the granitoids were emplaced, our best indication 585 for the emplacement depth comes from mineral assemblages in the metamorphic country rocks. For 586 example, mineral assemblages characterised by andalusite and sillimanite (Pieters et al., 1983;1990) 587 indicate relatively low pressures (<4 kbar), corresponding to a depth of ~15 km or less. The kyanite found 588 in pegmatite BJ104A suggests that higher pressures may have been attained, but this does not necessarily 589 apply, as pegmatitic kyanite can form via a variety of processes other than prograde metamorphism (e.g., 590 Woodland, 1963). Although the pegmatites containing large proportions of hydrous phases such as 591 muscovite and tourmaline are often associated with mid-to upper crustal levels, muscovite is 592 thermodynamically unstable in granitic magma at pressures below 3-4 kbar (Zen, 1988). This suggests 593 that the original melt formed at a greater depth and subsequently intruded at a shallower level. The 594 magma also likely shifted from the stability field of muscovite, as is indicated by its sub-solidus 595 replacement where in contact with feldspar (Fig. 5D). The presence of narrow zones of contact 596 metamorphism around many intrusions (Fig. 3B) provides further support that hot magma was injected 597 into shallower and cooler rocks, rather than these being derived from in-situ partial melting of and 598 segregation from metasedimentary country rocks. 599 The production of large amounts of metasedimentary partial melts and regional HT/LP (Abukuma-type) 601 metamorphism overprinting the surrounding country rocks imply a high geothermal gradient and an 602 anomalously hot continental crust. Such regional HT/LP conditions likely accompanied both the Permian-603 Triassic and Devonian-Carboniferous episodes of magmatism, although the younger metamorphic phase 604 partially overprinted the older phase. This is supported by the metapelitic migmatites associated with the 605 Kwok Granite and the Anggi Intrusive Complex. The heat required to produce regional metamorphism and 606 partial melting at low pressure was likely advected from the lower crust or mantle (e.g., DeYoreo et al., 607 1991). This potentially occurred over a relatively short-term (million-year) timescale (e.g., Viete and 608 Lister, 2017) rather than due to long-term steady state processes and heating driven by radioactive decay 609 (e.g., England and Thompson, 1984). This Abukuma-type metamorphism likely occurred when the region 610 was part of an active continental arc system and heat flux to the crust was high (Fig. 13A;  alkali-calcic to alkalic composition is also indicative of granitoids inboard of a Cordilleran-type arc (Frost 624 et al., 2001). We therefore interpret that the Permian-Triassic granitoids formed in an active continental 625 margin setting above a subduction zone (Fig. 13), while the Palaeozoic granitoids are tentatively 626 interpreted to represent post-orogenic magmatism or magmatism further inboard of an active margin. 627 628 The Palaeozoic granitoids described above are restricted to the Bird's Head Peninsula and represent the 629 oldest known episode of magmatism in New Guinea and eastern Indonesia (Fig. 13). These intrusives represent a collection of sparse discrete exposures within the Bird's Head. They are not coeval with the 631 Devonian to Carboniferous granitoids found in south-western New Guinea ( Fig. 13; Richards and  632 Willmott, 1970), but may potentially be part of a broader Devonian and Carboniferous orogenic belt and 633 associated granitoid and volcanic rocks found through parts of eastern Australia and New Zealand (e.g.,          Wariki Granodiorite (Bladon, 1988) Anggi Granite (Bladon, 1988) Melaiurna Granite (Bladon, 1988) Warjori Granite (Bladon, 1988) Figure 13