Growth of the upper crust in intra-oceanic island arcs by intrusion of basaltic magmas: the case of the Koloula Igneous Complex, Guadalcanal, Solomon Islands (SW Pacific)

The Pleistocene (2.2–1.5 Ma) Koloula Igneous Complex (KIC) on Guadalcanal in the Solomon island arc consists of a low-K calc-alkaline sequence of ultramafic to felsic plutonic rocks. We present whole-rock major and trace element and Sr–Nd-Pb isotope data, as well as mineral compositions that record the magmatic evolution of the complex. The intrusive sequence is grouped into two cycles, Cycle 1 and 2, comprising gabbroic or dioritic to granodioritic rocks. The major and trace element data of each cycle forms a single calc-alkaline fractional crystallisation trend. The distinct radiogenic isotope and incompatible element compositions of the Cycle 1 and 2 intrusions imply slightly different mantle sources. The KIC formed by shallow (0.1 GPa) fractional crystallisation of mantle-derived Al-rich basaltic parental magmas (6–8 wt.% MgO) that were formed by deeper-level (0.7 GPa) fractionation of olivine and pyroxene from Mg-rich (~ 11 wt.% MgO) primary magmas in the Solomon intra-oceanic island arc. Olivine, clinopyroxene, plagioclase, amphibole, biotite, apatite, and Fe–Ti oxides fractionated from the KIC’s high-Al basaltic parental magmas to form calc-alkaline magmas. Liquid line of descent trends calculated using mass balance calculations closely match major element trends observed in the KIC data. The KIC crystallised at shallow, upper crustal depths of ~ 2.0–3.0 km in ~ 20 km-thick island arc crust. This complex is typical of other Cenozoic calc-alkaline ultramafic to felsic plutons in Pacific intra-oceanic island arcs in terms of field relationships, petrology, mineral chemistry and whole-rock geochemistry. Hornblende fractionation played a significant role in the formation of the calc-alkaline felsic plutonic rocks in these Cenozoic arc plutons, causing an enrichment of SiO2 and light rare earth elements. These plutons represent the fossil magma systems of arc volcanoes; thus, the upper arc crust is probably generated by migration of magmatic centres.

Models of island arc crust formation suggest a stratification with an ultramafic to mafic lower crust and a felsic upper crust similar to that of the continental crust, and this stratification may form early in the magmatic evolution of island arcs (DeBari 1997;Jagoutz 2014). Whereas the mafic magmas may stagnate in the deep crust and fractionate Mg-Fe-rich minerals, intermediate to felsic melts may ascend into the middle to upper crust (Annen et al. 2006). The accretion of oceanic island arcs to continents is an important process in facilitating the growth of the continental crust (Rudnick 1995;Hawkesworth et al. 2010).
In this contribution, we present new petrologic, geochemical and isotopic data on a suite of ultramafic to felsic plutonic rocks from the Pleistocene Koloula Igneous Complex (KIC) on Guadalcanal in the Solomon island arc. The rocks follow a calc-alkaline fractionation trend, and two intrusive cycles are defined, each defining one liquid line of descent. Most of the early Koloula intrusion cycle consists of gabbroic rocks, representing a mafic intrusion in the upper crust with consecutive fractionation to less voluminous more evolved melts, whereas the later intrusive cycle is more evolved. In this study, we show that the KIC formed by shallow-level fractional crystallisation of mantle-derived basaltic magmas within the Solomon intra-oceanic island arc.

Geological setting
The island of Guadalcanal is part of the Central Province of the Solomon Islands, which are bounded to the north by the tectonically inactive Vitiaz Trench and to the south by the San Cristobal (or South Solomon) Trench where active subduction occurs (Fig. 1). Guadalcanal forms part of the Cretaceous to Quaternary South Solomon MORB terrain (Petterson et al. 1999). The crust beneath Guadalcanal is about 20 km-thick (Segev et al. 2012), and probably consists of Cretaceous mid-ocean ridge basalts (MORBs) and younger island arc rocks (Petterson et al. 1999). Zircon ages support a ~ 96 Ma age for the Cretaceous MORBs and ages of 71 to 33 Ma for the island arc rocks (Tapster et al. 2014). However, xenocrystic zircons with Paleozoic to Archean ages have been found in magmatic rocks from the Oligocene (26-24 Ma) Umasani Complex in western Guadalcanal, suggesting that the Solomon islands may be underlain by continental crust rifted from East Gondwana ( Fig. 1; Tapster et al. 2014). There are numerous Pliocene-Holocene volcanic centres in the northwestern part of Guadalcanal and one such volcanic centre at Mbalo near the southeastern coast of the island (Chivas 1978). The nearby island of Savo, located just off the northwestern coast of Guadalcanal, has numerous Pleistocene-Holocene volcanic centres and last erupted in the 1830s and 1840s (Chivas 1977(Chivas , 1978Hackman 1978).
The KIC constitutes a low-K calc-alkaline sequence of leucogabbros, gabbros, pyroxenites, diorites, quartz diorites, tonalites, granodiorites, trondhjemites, and aplites (Chivas 1977(Chivas , 1978Chivas and McDougall 1978;Chivas et al. 1982;Tapster et al. 2016). This plutonic complex formed by more than 30 intrusions that have been grouped into 2 cycles (Cycle 1 and 2), the first of which occurred in the early Pleistocene (2.2 Ma; Gelasian age) in the form of gabbroic and ultramafic cumulates that evolved to diorite, quartz diorite, and granodiorite (Fig. 2;Chivas 1977Chivas , 1978Chivas and McDougall 1978;Tapster et al. 2016). The second cycle consists of mid-Pleistocene (1.7-1.45 Ma; Calabrian age) diorite and tonalite bodies comprising the concentrically zoned Inamumu Pluton (Fig. 2), which was intruded by aplite and trondhjemite dykes and two porphyritic tonalites (Fig. 2;Chivas and McDougall 1978;Tapster et al. 2016). The emplacement and crystallisation ages of the KIC were previously determined by K-Ar dating to be between 4.5 and 1.5 Ma (Chivas and McDougall, 1978). Rapid uplift (> 2 km) since the early Pleistocene and significant erosion has exposed the KIC (Chivas et al. 1982). Both Cycles 1 and 2 are intrusive into the spatially associated Early Miocene Suta Volcanics, which comprise pillow and massive lavas of basaltic to andesitic composition (Chivas 1977(Chivas , 1978. The KIC contains a porphyry Cu mineral prospect, with an estimated 50 Mt of ore at a low grade of 0.17% Cu (Holm et al. 2019;Keith et al. 2022). This mineralisation formed ~ 250 ka after the initial intrusion of the Cycle 2 rocks, with the so-called A and B systems of hydrothermal alteration forming at 1.46 Ma (Chivas and McDougall 1978). Late magmatic activity is indicated by the emplacement of post-mineralisation andesitic dykes (Chivas and McDougall 1978). The constant Sr isotopic compositions ( 87 Sr/ 86 Sr = 0.70363-0.70378) of rocks with variable SiO 2 and MgO contents were interpreted by Chivas et al. (1982) to reflect a liquid line of descent for the entire KIC.

Sampling and sample preparation
Samples representing the variable lithologies of both intrusive cycles were collected during mapping field work on Guadalcanal in 1972-1975(Chivas 1977. Some samples were previously analysed (Chivas 1977) and share the same field sample numbers. For the more recent analyses, rock powders were prepared afresh from original remnant hand specimens. Unaltered pieces were cut with a rock saw for geochemical analysis and thin section preparation. The whole-rock pieces were washed in an ultrasonic bath with deionised water and dried for 12 h at 60 °C. The clean geochemistry pieces were crushed using a hydraulic press and reduced to powder in an agate ball mill. The powders were dried for 12 h at 104 °C prior to fusion for major element analysis.

Whole-rock element analyses
The major element concentrations of 42 whole-rock samples were analysed using a Spectro XEPOS He X-ray fluorescence spectrometer at the GZN. Further details of the analytical techniques are provided in Freund et al. (2013). The accuracy and precision of the measurements were determined by multiple measurements of the international rock standards BE-N, BR, AC-E and GA (Table S2). The accuracy was generally better than ~ 3% (2σ), except for P 2 O 5 , which was better than 7.5%. The precision was better than 0.1-0.2%. Trace element analyses of 42 samples from the KIC were determined at the GZN on a Thermo Fisher Scientific XSeries 2 Quadrupole Inductively Coupled Plasma Mass Spectrometer (ICP-MS) connected to an Aridus 2 membrane desolvating sample introduction system. Major and trace element whole-rock geochemical data from the KIC are presented in Table S3. The Eu/Eu* and Ti/Ti* anomalies in Table S3 were calculated using the methods set out in Taylor and McLennan (1985). The normalised ratios in Table S3 were calculated based on the chondrite values in Sun and McDonough (1989).

Whole-rock Sr, Nd and Pb isotope analyses
Strontium, Nd and Pb isotope analyses on 17 samples were performed at the GZN. Strontium and Nd isotopes were analysed using a Thermo-Fisher Triton thermal ionisation mass spectrometer (TIMS) in static mode following the chemical and analytical procedures outlined in Haase et al. (2017). Strontium isotope measurements were corrected for mass fractionation assuming 88 Sr/ 86 Sr = 0.1194, where mass 85 was monitored to correct for the contribution of 87 Rb to 87 Sr. Neodymium isotope data were corrected for mass fractionation using 146 Nd/ 144 Nd = 0.7219. Samarium interferences on masses 144, 148, 150 were corrected by measuring 147 Sm, but the correction was insignificant for all samples. The maximum internal uncertainty was 0.000007 for 87 Sr/ 86 Sr and 0.000005 for 143 Nd/ 144 Nd (2σ). During the measurements, the NBS987 standard yielded 87 Sr/ 86 Sr = 0.710278 (2σ, n = 6) and the Erlangen Nd standard gave 143 Nd/ 144 Nd = 0.511534 (2σ, n = 4), which corresponds to a value of 0.511850 for the La Jolla Nd isotope standard determined at the GZN.
Lead isotope measurements were carried out on a Thermo-Fisher Neptune MC-ICP-MS in static mode using a 207 Pb/ 204 Pb double spike to correct for instrumental mass fractionation. Details of the Pb isotope procedures and analytical techniques are discussed by Woelki et al. (2018). The Pb fraction was diluted with 2% HNO 3 to a concentration of approximately 20 ppb. One part of this solution was spiked to obtain a 208 Pb/ 204 Pb ratio of ~ 1. The double spike, with a 207 Pb/ 204 Pb ratio of 0.5, was calibrated against a solution of the NBS982 equal atom Pb standard. Spiked and unspiked sample solutions were introduced into the plasma via a Cetac Aridus desolvating nebuliser, and measured in static mode. Interference of 204 Hg on mass 204 was corrected by monitoring 202 Hg. An exponential fractionation correction was applied offline using the iterative method of Compston and Oversby (1969); the correction was typically 4.5‰ per amu.
The internal uncertainty was 0.0003, 0.0003 and 0.0008  Table S2. Radiogenic Sr, Nd and Pb isotope data from the KIC are shown in Table S4.
There is a progressive decrease in the anorthite content of plagioclase from the Cycle 1 to Cycle 2 intrusions of the KIC ( Fig. 3; Table S1). The anorthite contents of plagioclase decrease from Cycle 1 (An 12-98 ) to Cycle 2 (An 10-82 ). Moreover, there is a decrease in the anorthite content of plagioclase within both cycles over time (Fig. 3). The higher average An contents of plagioclase from the Cycle 1 intrusions are attributed to the greater occurrence of cumulate rocks within this cycle and a paucity of similar lithologies in the Cycle 2 intrusions. Plagioclase crystals from Cycle 1 lithologies younger than the Chakachaka Gabbro have overlapping An contents with those from the Cycle 2 intrusions. The most evolved lithologies in the Cycle 1 intrusions have plagioclase crystals with lower An contents than the Cycle 2 quartz diorites and tonalites.
It should be pointed out that the plagioclase An content variations in the Cycle 1 and 2 lithologies were from the analysis of only a few plagioclase grains per sample and that some of these grains exhibit oscillatory zoning. Profiles across plagioclase grains reveal either constant or decreasing An contents from core to rim in the majority of analysed samples (e.g. from An 75 to An 47 in the Charilava Quartz Diorite). However, some zoned plagioclase grains from the olivine pyroxenites (Cycle 1) show slightly increasing An contents during crystal growth (e.g. from An 82 to An 87 ). Normal zoning of plagioclase grains generally supports a magma evolution dominated by crystallisation without significant assimilation or magma mixing. Yet, weak inverse zoning of plagioclase in the pyroxenites indicates that recharge of mafic magmas affected the Koloula magmatic system during its early stages. The outlying plagioclase An contents (e.g. the high-An contents in the Cycle 1 diorite or the low-An contents in the Cycle 1 granodiorite) in some of the samples may be indicative of other processes. For instance, the high-An contents of plagioclase in the Cycle 1 diorite may represent high-An cores that crystallised at an earlier magmatic stage at deeper crustal levels or xenocrystals entrained from pre-existing magmatic lithologies. The low-An contents of plagioclase in the Cycle 1 granodiorite could be attributed to alteration.

Major element variations
The samples from the KIC vary from ultramafic to felsic in composition with a range from 0.1 to 20 wt.% MgO and 43 to 78 wt.% SiO 2 ( Fig. 4; Table S3 and Table S5). The Koloula plutonic samples show similar major element trends to volcanic rocks from the Solomon arc (Figs. 1 and 4; Tables S3-S6). A key feature of the Koloula plutonic samples is that they have decreasing CaO, MgO and FeO T contents with increasing SiO 2 ( Fig. 4; Table S3 and Table S5; König & Schuth, 2011). The MgO contents of the Koloula plutonic samples range from 20 to 6 wt.% at 46-50 wt.% SiO 2 before decreasing from 6 wt.% MgO at 50 wt.% SiO 2 to 0 wt.% MgO at 78 wt.% SiO 2 (Fig. 4a). The TiO 2 contents of the Koloula plutonic rocks vary between 0.15 and 1.0 wt.% in the mafic plutonic rocks (Table S3 and Table S5). The KIC plutonic rocks with > 52 wt.% SiO 2 show decreasing TiO 2 contents. The Al 2 O 3 contents decrease slightly from 18 to 17 wt.% between 49 and 62 wt.% SiO 2 and then decrease more steeply towards higher SiO 2 contents (Fig. 4b). The FeO T and CaO contents in the mafic rocks are ~ 10 and 12 wt.%, respectively, and decrease to 0.7 and 0.8 wt.%, respectively, in the aplites (Fig. 4c, d). The Na 2 O contents increase from 0.4 to 5.0 wt.% with increasing SiO 2 (Fig. 4e). The K 2 O contents increase from 0.2 wt.% in the mafic rocks to 1.8 wt.% in the tonalites, whereas two aplites contain 5.5 and 7.2 wt.% K 2 O (Table S3 and Table S5). The P 2 O 5 concentrations are < 0.1 wt.% in the mafic rocks but increase to 0.25 wt.% in rocks with ~ 53-56 wt.% SiO 2 before decreasing steeply to 0.08 wt.% in the aplites (Fig. 4f).

Trace element variations
The normal mid-ocean ridge basalt (N-MORB)-normalised trace element patterns of the KIC rocks exhibit negative Nb and positive Pb anomalies, whilst the Cycle 1 and 2 intrusions are distinctive (Fig. 5). The Cycle 2 intrusions have more evolved trace element patterns displaying higher incompatible element concentrations than the Cycle 1 intrusions. The Cycle 1 and 2 intrusions have trace element patterns that overlap with those of lavas from the Solomon arc (Fig. 5;König and Schuth 2011). Some Cycle 2 samples have higher Th, U, MREE and HREE normalised abundances than the Solomon Islands lavas, whereas other Cycle 2 samples have lower MREE and HREE normalised abundances than these lavas. One Cycle 1 sample has higher MREE and HREE normalised abundances than the Solomon arc lavas.
The trace element ratios of lavas from the Solomon arc mostly overlap with those of the KIC (Fig. 6). The (Dy/Yb) N ratios in Cycles 1 and 2 are almost constant at ~ 1 (Fig. 6a). There are higher (Ce/Yb) N and (La/Sm) N ratios in samples from Cycle 2 than those from Cycle 1 (Fig. 6b, d). Cycle 1 and 2 samples have constant Eu/Eu* anomaly ratios of ~ 1 (Fig. 6c). Moreover, the Ba/La ratios of Cycle 1 intrusions vary between 20 and 40; however, most Cycle 2 samples with > 60 wt.% SiO 2 have higher Ba/La ratios of 40-60 ( Fig. 6e). The Koloula plutonic samples have near-constant Nb/La ratios (0.1-0.3) for a given SiO 2 content; however, the Cycle 2 intrusions generally have lower Nb/La ratios than the Cycle 1 intrusions (Fig. 6f).

Isotope variations of the Koloula plutonic rocks
As pointed out by Chivas et al. (1982), the Koloula plutonic rocks show a narrow range of radiogenic isotope compositions and we find near-constant Sr, Nd and Pb isotope ratios for the entire ultramafic to felsic compositional range (Figs. 7 and 8). Samples from Cycle 2 have overlapping 143 Nd/ 144 Nd and 87 Sr/ 86 Sr isotope ratios to those from Cycle 1 (Fig. 7). Cycle 1 and 2 intrusions have distinct Pb radiogenic isotope compositions (Fig. 8). The 206 Pb/ 204 Pb isotope ratios of samples from Cycle 2 are less radiogenic than those of the Cycle 1 samples when plotted against 207 Pb/ 204 Pb, 208 Pb/ 204 Pb, 87 Sr/ 86 Sr and 143 Nd/ 144 Nd (Fig. 8).

Estimation of the emplacement depth of the KIC
Fluid inclusion evidence from the porphyry Cu mineralisation within the KIC has been used to estimate the lithostatic load (~ 3.3 km) of the complex (Chivas and Wilkins 1977). This initial depth of emplacement estimate (~ 3.3 km) of the IZP (Cycle 2) was based on fluid-inclusion geobarometric determinations using the phase equilibria of Sourirajan and Kennedy (1962) and using a rock density of 2.6 g/cm −3 . Subsequent experimental data from Bodnar et al. (1985), particularly for the relevant temperature range of 500-700 °C, permit a re-evaluation of this depth of emplacement estimate, wherein the calculated parameters, with assigned homogenisation errors of ± 10-15 °C, provide trapping pressures of 525-630 bar (0.053-0.065 GPa) which, using a rock density of 2.6 g/cm −3 , translates to a lithostatic load of 2.0-2.5 ± 0.25 km. Roedder and Bodnar (1980) caution that such calculations may represent minimum pressures for such systems if, as in this case, the inclusions trapped a highly saline albeit unsaturated solution. This depth estimate pertains only to the IZP and, by extension, given they are of the same age, the Cycle 2 satellite plutons. The older Cycle 1 lithologies are assumed to have intruded at a slightly greater depth (~ 3 km), as they can be expected to have been semicontinuously uplifted since their emplacement. The high anorthite contents of plagioclase (up to An 99 ) from the KIC suggest that its parental magmas likely formed by partial Page 9 of 21 107 melting of a hydrous sub-arc mantle source and crystallised at a shallow depth (e.g. Sisson and Grove 1993a, b;Takagi et al. 2005). Moreover, amphibole barometry after Ridolfi et al. (2010) suggests crystallisation pressures of between 0.04 and 0.1 GPa (2.7-4.0 km) for most of the analysed Cycle 1 and 2 phases. This is in good agreement with the formation depth of the porphyry Cu mineralisation in the KIC (Chivas and Wilkins 1977) and implies near-constant intrusion depths for both cycles. Higher estimated pressures (0.14-0.24 GPa) for amphibole from the Kolochoro Quartz Diorite (Cycle 2) may suggest that amphibole in the KIC started crystallising at a greater depth and implies temporal magma stagnation at depths of ca. 5-9 km. Finally, the emplacement of this complex into the pillow-bearing basaltic to andesitic lavas of the Suta Volcanics indicates that it formed in shallow oceanic arc crust (Chivas 1977(Chivas , 1978.

Constraints on the compositions of the KIC's primary and parental magmas
The Cycle 1 intrusions from the KIC have a basaltic composition similar to the mafic lavas from the Solomon island arc (Fig. 4). This implies that the plutonic rocks of the KIC largely reflect melt compositions rather than cumulates, except for the pyroxenites, gabbros, leucogabbros and anorthositic segregations ( Fig. 4; Table S3 and Table S5).
The Mg-rich (> 10 wt.% MgO) plutonic rocks in the KIC likely represent cumulates rather than melt compositions on account of their coarse-grained cumulate textures (Fig. 4). There are no known cross-cutting high-Mg dykes within the KIC that are representative of the high-Mg primary magmas to this complex. The Fo contents (75-82) of olivine from the KIC are not in equilibrium with the Mg# of Mg-rich lavas from the Solomon (Mg#=72-82) arc (Table S6) due to fractionation of olivine and clinopyroxene from the KIC's parental magma. To test whether the KIC had Mg-rich primary magmas, the composition of the KIC's primary magma was modelled using the reverse crystallisation tab in the Petrolog3 software of Danyushevsky and Plechov (2011). The starting composition was Kolochoro Quartz Diorite sample A.498, which was selected because it represents the most primitive composition in the KIC that can still be considered as a melt. Fractionation of this primary magma was assumed to occur at a pressure of 0.7 GPa at the base (~ 20 km) of the Solomon island arc crust (Furumoto et al. 1970;Segev et al. 2012; Electronic Supplementary Material (S) 2). A temperature of 1280 °C and an oxygen fugacity of QFM + 2.2 were assumed on the basis of temperature and fO 2 values reported in Rohrbach et al. (2005) for picrites from the nearby New Georgia Group to the west of Guadalcanal. A starting H 2 O content of 1 wt.% was used based on the H 2 O content of sample A.498. Olivine and clinopyroxene mineral-melt models from Ariskin et al. (1987) were used in the modelling due to these models being applicable to upper to lower crustal crystallisation pressures, island arc settings and basaltic compositions. This modelling indicates that the primary magmas to the KIC were Mg-rich (ca. 11 wt.% MgO) and basaltic in composition (indicated by the green diamonds in Fig. 4; see Tables S7 and S2 for details on the modelling input parameters). These Mg-rich primary magmas (SiO 2 = 49.0 wt.%, Al 2 O 3 = 13.9 wt.%, MgO = 11.0 wt.%, CaO = 14.2 wt.%; T7) have lower MgO contents than primary arc magmas from the Sunda, Aleutian, Honshu, New Hebrides and Lesser Antilles arcs (Table S7). However, the KIC's primary magmas were likely generated by hydrous melting of depleted mantle sources in subduction zones (Figs. 4 and 5; Tables S3-S5; Pichavant et al. 2002;König and Schuth 2011;Schmidt and Jagoutz 2017;Müntener and Ulmer 2018).
The modelled primary magma of the KIC (the green diamond in Fig. 4) was used as a starting composition to test whether a high-Al basaltic magma can be generated by fractionation of olivine and pyroxene at depth (e.g. Kay et al. 2019). The composition of the KIC's parental magma was determined using the crystallisation tab in the Petrolog3 software. The starting H 2 O content was assumed to be 0.74 wt.% and was derived from the modelled H 2 O content of the KIC's primary magma (Table S7). This assumption was supported by the application of %) and f P 2 O 5 (wt.%) are plotted against SiO 2 (wt.%). These diagrams demonstrate that the MgO, CaO and FeO T contents of the KIC decrease with increasing SiO 2 , whereas the Na 2 O contents increase with increasing SiO 2 . The Al 2 O 3 contents of the KIC gently decrease between ~ 49 and ~ 62 wt.% SiO 2 before decreasing more steeply with increasing SiO 2 . The P 2 O 5 contents of the KIC increase up to 56 wt.% SiO 2 before decreasing due to apatite fractionation. The direction of fractionation and the involved fractionating minerals are highlighted by the fractionation arrows (Ol: olivine; Cpx: clinopyroxene; Plag; plagioclase; Apat: apatite; Amph: amphibole; Biot: biotite). The crystallisation of cumulate lithologies and their constituent cumulate minerals (e.g. olivine, clinopyroxene, plagioclase) are indicated by the accumulation trend arrows. Published data for the KIC are from Chivas (1977), Chivas et al. (1982) and Tapster et al. (2016). Data shown for the Solomon arc lavas are from Chivas (1977), Schuth et al. (2004Schuth et al. ( , 2009); König et al. (2007), and König and Schuth (2011). The average compositions of Cycle 1 and 2 are indicated by the enlarged navy-blue and grey squares, respectively. These average compositions were determined by averaging the compositions of the Cycle 1 and Cycle 2 samples. The modelled composition (Table S7) of the primary magma to the KIC is shown by the green diamonds. The modelled composition (Table S7) (2008) plagioclase hygrometer, which when applied to whole-rock and plagioclase compositions indicates that the KIC parental magma had a H 2 O content of ~ 0.6 wt.%. An oxygen fugacity value of Ni-NiO + 1.4 was used during the modelling calculations and was also obtained during the primary magma composition modelling (see S2). It is assumed that fractional crystallisation occurred at a pressure of 0.7 GPa at the base (~ 20 km) of the Solomon intra-oceanic island arc (Furumoto et al. 1970;Segev et al. 2012). The modelling was stopped at 100% of fractionation and the fractionation percentages for olivine and clinopyroxene were set at 35% each after being increased from 20% each based on Kay et al. (2019). The selection of olivine and clinopyroxene fractionation percentages was also guided by the compositions of the KIC's mafic plutonic lithologies. The resultant modelled basaltic parental magma to the KIC resembles a high-Al basalt (SiO 2 = 48.7 wt.%, Al 2 O 3 = 16.7 wt.%, MgO = 8.4 wt.%, CaO = 12.5 wt.%; Table S7). This modelled parental magma composition indicates that the MgO-rich primary magmas to the KIC underwent olivine and pyroxene fractionation (e.g. Kay et al. 2019) at the base of the Solomon arc crust (ca. 20 km depth) to form the high-Al basaltic parental magmas to the KIC. Chivas et al. (1982) proposed that the KIC formed by fractional crystallisation of mantle-derived basaltic magmas. This model is supported by linear trends exhibited by the KIC data in major element variation diagrams (Fig. 4). To test whether the KIC formed by fractional crystallisation, the mass balance equation in Jagoutz (2010) was used to calculate the compositions of melts that was formed by fractional crystallisation (see Table S7 and S2 for further details). This mass balance equation was applied to a starting melt composition (the modelled high-Al basaltic parental magma to the KIC) and averaged cumulate compositions from the KIC. Melt compositions were calculated for ten successive fractionation steps by successively changing the fractionating compositions from gabbro to quartz diorite to tonalite to aplite. It should be noted here that Jagoutz (2010) used lithologies ranging from dunites to tonalites as fractionating cumulate compositions, the more evolved of   (1977), Chivas et al. (1982), Schuth (2011), andTapster et al. (2016). These diagrams display how the trace element ratio trends of the Cycle 1 and Cycle 2 intrusions compare with one another as well as lavas from the Solomon arc, and plutonic rocks from Cenozoic ultramafic to felsic arc plutons. Data shown for the Solomon arc lavas are from Schuth et al. (2004Schuth et al. ( , 2009, König et al. (2007) and König and Schuth (2011). The Cenozoic arc pluton data sources are the same as those in Fig. 4. The large yellow squares represent the composition of average continental crust from Rudnick (1995). The trace element ratios are normalised to chondrite values from Sun and McDonough (1989) which are not usually regarded as cumulates in the normal sense but liquids. The resultant calculated melt compositions are shown in Fig. 9 and Table S7 and represent the KIC's liquid lines of descent. The calculated fractionation trends closely resemble the KIC major element (MgO, Al 2 O 3 , CaO and FeO T vs SiO 2 ) trends observed in Fig. 4. The calculated fractionation and observed KIC data trends decrease from ~ 8 wt.% MgO at ~ 50 wt.% SiO 2 to ~ 0.5 wt.% MgO at ~ 74 wt.% SiO 2 (Fig. 9). The Al 2 O 3 contents of the calculated and observed KIC trends slightly decrease from ~ 18 to 17 wt.% between 49 and 62 wt.% SiO 2 before decreasing more steeply to 13 wt.% at 74 wt.% SiO 2 (Fig. 9). The CaO and FeO T contents of the calculated and observed KIC trends decrease from ~ 10 wt.% and ~ 12 wt.% at 50 wt.% SiO 2 to 2 wt.% and 1 wt.% at 74 wt.% SiO 2 , respectively (Fig. 9). This provides further evidence that the KIC formed by fractional crystallisation of basaltic magmas. The most evolved calculated melt composition (SiO 2 = 74.3 wt.%, Al 2 O 3 = 13.0 wt.%, MgO = 0.5 wt.%, CaO = 1.5 wt.%; Na 2 O = 3.5 wt.%; Table S7) is very similar to the most evolved felsic lithologies in KIC. The fractionation trend calculated for the KIC and observed in the KIC data is very similar to the low-pressure (0.1 GPa) fractionation trends calculated for the evolution of basaltic arc magmas (Fig. 9) by Loucks (2021). This suggests that KIC magmas fractionated at a shallow depth of ~ 3 km because the KIC data and calculated fractionation trends plot more along the 0.1 GPa trends in Fig. 9 than the high-pressure (1 GPa) trends calculated by Loucks (2021). Experimental and calculated liquid of descent trends determined by Nandedkar et al. (2014) and Bucholz et al. (2014) for basaltic arc magmas fractionating at pressures of 0.7 Ga and 0.2-0.4 GPa do not match the calculated and observed KIC trends to the same extent as the 0.1 GPa fractionation trend of Louck (2021).

An appraisal of a fractional crystallisation origin for the KIC
Residual sum of squares calculations were conducted using the MINSQ spreadsheet (see S2; Herrmann and Berry, 2002) to establish fractionating mineral proportions for the KIC and further appraise whether this complex formed by fractional crystallisation. This spreadsheet uses whole-rock and mineral compositions to determine the residual sum of squares value between a chosen starting composition and an estimated starting composition and can be applied to various petrological questions. Sample A.498 was chosen as the starting composition because this sample is gabbroic in composition and still represents a melt. Sample A.6 was selected as the evolved sample composition because it has the most evolved composition in Table S3. Representative compositions of plagioclase (An 60 ), olivine (Fo 78 ), amphibole (high Al), clinopyroxene (Mg# = 81), apatite, orthopyroxene (Mg# = 72) and biotite (Mg# = 63) were selected from Table S1 and entered into the spreadsheet. A representative magnetite composition from Chivas (1977) was also incorporated into the spreadsheet. A residual sum of squares value of 1.79 was determined for sample A.498 and the estimated starting composition based on these whole-rock and mineral composition data and the mineral proportions and the remaining melt fraction that were manually entered. This low residual sum of squares value indicates that the geochemical trends observed in the KIC data resulted from fractional crystallisation. The selection of the mineral proportions was based on mineral proportions reported in Kay et al. (2019) for the formation of felsic magmas by fractional crystallisation in the Aleutian arc. Amphibole (41%), plagioclase (31%), clinopyroxene (9%), biotite (3%), magnetite  Table S4 and published data in Chivas et al. (1982). a 87 Sr/ 86 Sr versus SiO 2 (wt.%). b 143 Nd/ 144 Nd versus SiO 2 (wt.%). These plots show that the 87 Sr/ 86 Sr and 143 Nd/ 144 Nd of the KIC are near constant across a wide compositional range. The radiogenic isotope data in a and b for the Solomon Island arc lavas are from and König and Schuth (2011). Uncertainty error symbols are shown in this figure (3%), apatite (2%) and olivine (1%) fractionated from the KIC's parental magma to form the intermediate and felsic lithologies in this complex. These fractionating mineral proportions do not entirely match the modal mineralogy of the mafic to intermediate plutonic rocks in the KIC (e.g. Chivas 1977). However, this is also apparent in the modelling presented by Kay et al. (2019) for the petrogenesis of the Hidden Bay Pluton (Aleutian arc, Alaska) by fractional crystallisation of high-Al basaltic magmas. Kay et al. (2019) found that the basaltic parental magmas to the Hidden Bay Pluton evolved to rhyodacitic and rhyolitic compositions by fractionation of amphibole (44%), plagioclase (35%), clinopyroxene (10%), magnetite (5%), biotite (5%) and apatite (1%). These fractionation mineral proportions are very similar to those found for the KIC.

Evaluation of other differentiation mechanisms proposed for the formation of intermediate to felsic magmas
Tonalitic plutons are commonly found in intra-oceanic island arcs and these felsic magmas are thought to have either formed by shallow fractional crystallisation from mafic melts (e.g. Chivas et al. 1982;Whalen 1985;Haase et al. 2014;Müntener and Ulmer 2018), re-melting of hydrothermally altered mafic crustal rocks and possibly sediments (e.g. Petford and Atherton 1996;Haraguchi et al. 2003;Smith et al. 2003;Saito et al. 2011 Tonalites, trondhjemites and granodiorites derived by melting of hydrothermally altered subducting oceanic crust or lower arc crust or by lower crustal fractional crystallisation of hydrous basaltic arc magmas have high Al 2 O 3 (≥ 15 wt.%) contents and (La/Yb) N (≥ 30) and Sr/Y (≥ 30) ratios and low Y (≤ 18 ppm) and Yb (≤ 1.9 ppm) contents reminiscent of adakites (Martin 1986;Drummond and Defant 1990;Petford and Atherton 1996;Hansen et al. 2002;Martin et al. 2005;Loucks 2021). Moreover, the melting of mafic crust near the base of thick (> 40 km) continental arcs has been proposed to account for the high (La/Yb) N and Sr/Y ratios  Kepezhinskas et al. (1997) and Schoenhofen et al. (2020) and show that the radiogenic isotope compositions of the KIC samples can be explained by 1-2% mixing between a depleted MORB man-tle (DMM) source and sediment melts. The radiogenic isotope data for the Solomon lavas are shown for comparison and are from König and Schuth (2011). The compositions of the sediment end members are from pelagic sediment data in Ben Othman et al. (1989). The isotopic compositions of DMM are from Workman and Hart (2005). The MORB and sediment fluid compositions used in the generation of the mixing models are from Ayers (1998) and Spandler et al. (2007) and low Yb and Y concentrations observed in tonalites and granodiorites from continental arc batholiths, such as the Miocene Cordillera Blanca Batholith of Peru (Petford and Atherton 1996). The tonalites from the KIC have (La/Yb) N ratios of < 10 (2.9-5.8), and most of them have Sr/Y ratios of ≥ 30 (14-64). Moreover, most of the Koloula tonalites have Al 2 O 3 , Y and Yb contents of ≥ 15 wt.% (15.4-18.2 wt.%), ≤ 18 ppm (7.4-24.9 ppm) and ≤ 2 ppm (0.8-2.6), respectively (Table S3). Moreover, the lack of a correlation between the near-constant 143 Nd/ 144 Nd isotope ratios (Fig. 7b) and the variable (La/Sm) N ratios of the KIC indicates that slab-derived sediment melts played a minor role and that the positive correlation between (La/Sm) N and SiO 2 (Fig. 6c) resulted from amphibole fractionation. The relative thinness (20 km) of the Solomon intra-oceanic island arc crust beneath the KIC indicates that melting of hydrothermally altered subducting oceanic crust or lower arc crust or by lower crustal fractional crystallisation of arc basaltic magmas was unlikely. Furthermore, there is currently no existing field evidence to suggest that the tonalites, trondhjemites and granodiorites of the KIC were derived by in situ melting of mafic arc crust. The low (La/Yb) N and near-constant (Dy/Yb) N (~ 1) ratios of the KIC (Fig. 6) indicate that neither garnet nor hornblende was significant residual minerals in its source (e.g. Petford and Atherton 1996;Dessimoz et al. 2012).
Polybaric fractionation has been proposed as a mechanism for the generation of intermediate to felsic magmas in intra-oceanic island arcs and continental arcs (Almeev et al. 2013;Hamada et al. 2014;Melekhova et al. 2015;Lewis et al. 2021;Marxer et al. 2022). Our findings indicate that the lithologies of the KIC formed by fractionation of high-Al basaltic magmas at shallow (0.1 GPa) levels within the Solomon intra-oceanic island arc crust. The similarity of the Sr and Nd isotope ratios of the mafic to felsic rocks of Cycle 1 and Cycle 2 (Fig. 7) suggests that assimilation of sediments or hydrothermally altered crustal material was limited during their magmatic evolution, in agreement with Chivas et al.  Table S2 and Table 7 for more details regarding the mass balance calculations. Liquid lines of descents calculated by Loucks (2021) for basaltic melts crystallising at pressures of 0.1 GPa and 1 GPa are plotted for comparison. The KIC samples plot more along the 0.1 GPa fractionation trend compared to the 1 GPa fractionation trend. Experimental liquid line of descent trends determined by Nandedkar et al. (2014) for primitive, hydrous arc magmas crystallising at 0.7 GPa are also plotted for comparison. Liquid line of descent trends calculated using the mass balance equation in Jagoutz (2010) by Bucholz et al. (2014) for the 0.2-0.4 GPa fractional crystallisation of the Dariv Igneous Complex (Mongolia) are also shown for comparison. The published KIC data are from Table S5 and Table S8 (1982). Radiogenic Sr and unradiogenic Nd isotope compositions indicate that there is no old continental material present beneath the KIC, in contrast to the Oligocene Umasani plutonic complex further to the northwest ( Fig. 1; Tapster et al. 2014). Moreover, the δ 18 O standard mean ocean water (SMOW) isotope values of unaltered whole-rock ultramafic to felsic samples from the KIC range between 5.4 and 7.2‰ (Chivas et al. 1982), indicating that their parental magmas were likely derived from a mantle source and not contaminated by older continental crust (e.g. Valley et al. 1998). The linear trends exhibited by the KIC (Fig. 4) may have resulted from mixing of magmas of different compositions (e.g. Ichiyama et al. 2020;Rezeau et al. 2021); however, the constancy in the 143 Nd/ 144 Nd and 87 Sr/ 86 Sr ratios of the KIC ultramafic to felsic lithologies indicates that magma mixing was unlikely (Fig. 7). Additionally, the changing trend from increasing to decreasing P 2 O 5 contents with increasing SiO 2 at ca. 56 wt.% SiO 2 (Fig. 4f) indicates that the trends of mafic to felsic magmas cannot reflect binary mixing between these melts.

A new fractional crystallisation model for the KIC
Here, we present a new fractional crystallisation model for the KIC that builds on the findings of Chivas et al. (1982) who proposed that the complex formed by fractional crystallisation. The KIC's primary magmas (11 wt.% MgO) evolved to Al-rich basaltic magmas (6-8 wt.% MgO) through olivine and pyroxene fractionation in the lower crust, most likely at a depth of ~ 20 km at the base of the Solomon intra-oceanic island arc in a manner similar to the model proposed by Kay et al. (2019) for the Hidden Bay and Kagalaska plutons in the Aleutian intra-oceanic island arc. These high-Al basaltic magmas then ascended to the shallow crust to form the KIC. The earliest formed lithology in the KIC constitutes a small proportion of the Cycle 1 intrusions and formed by the crystallisation of olivine and clinopyroxene from the KIC's high-Al basaltic parental magmas. This resulted in the generation of clinopyroxene-olivine cumulates (pyroxenites and olivine pyroxenites) containing 10.6-16.1 wt.% MgO, representing the earliest formed lithologies in the KIC and the crystallisation of pyroxenites in the shallow crust (Chivas and McDougall 1978;Chivas et al. 1982). The more primitive nature of the high-Al basaltic magmas intruding the shallow crust during Cycle 1 compared to Cycle 2 is supported by the An-rich composition (> An 90 ) of plagioclase in the Cycle 1 pyroxenites (Fig. 3). The decreasing An contents in each cycle likely reflect increasing magma evolution beginning with the most mafic composition. Lavas from Savo contain abundant xenoliths of clinopyroxenites (Smith 2014) that likely represent cumulates from shallow magma chambers comparable to that of the KIC. Moreover, amphibole in amphibole cumulates beneath arc volcanoes on Savo probably formed by reaction replacement of earlier formed clinopyroxene, orthopyroxene, olivine, plagioclase, and Fe-Ti oxides (Smith 2014). The olivine compositions of the Koloula pyroxenites (Fo 75-82 ) indicate that their parental magmas (Mg# = 50-60) were ~ 23-50% more evolved than primary magmas (Mg# = 74) with more primitive olivine (Fo 90 ) compositions (Table S1; e.g. Almeev et al. 2007). On the other hand, the olivine compositions of the Koloula pyroxenites suggest that melts with Mg# of 50-60 ascended into the upper crust of the Solomon island arc. Following the crystallisation of the KIC pyroxenites, plagioclase, pyroxene and, to a lesser extent, olivine fractionated from the KIC's high-Al basaltic magmas to form the anorthositic segregations, leucogabbros, gabbros and pyroxenites in the Chakachaka Gabbro.
Shallow (0.1 GPa) fractionation of olivine, clinopyroxene, plagioclase, amphibole, biotite, apatite and Fe-Ti oxides from the KIC high-Al basaltic magmas resulted in the generation of calc-alkaline magmas in the Solomon intra-oceanic island arc beneath Guadalcanal and the formation of the intermediate-felsic plutonic rocks of the KIC (Figs. 4, 6 and 9; Table S7). The crystallisation of olivine, clinopyroxene, plagioclase and Fe-Ti oxides led to a significant decrease in MgO, CaO and FeO T contents with increasing SiO 2 (Fig. 4). Similar trends are observed in rocks from other Cenozoic ultramafic to felsic arc plutons, such as the Tanzawa Plutonic Complex and Uasilau-Yau Yau Intrusive Complex, which formed in the comparable settings of the Izu-Bonin-Mariana and Bismarck island arcs, respectively (Fig. 4;Whalen 1985;Kawate and Arima 1998;Suzuki et al. 2015). The Al 2 O 3 contents of the KIC samples started to gently decrease at ~ 49 wt.% SiO 2 , marking the onset of plagioclase crystallisation. The crystallisation of plagioclase is reflected by the positive Eu/Eu* anomaly ratios exhibited by some Cycle 1 samples, which represent plagioclase-rich cumulates (Fig. 6). The Na 2 O contents of the KIC decreased with increasing SiO 2 due to the crystallisation of An-rich plagioclase and amphibole.
Hornblende fractionation is known to be important for the formation of calc-alkaline magmas (Sisson and Grove 1993a, b;Davidson et al. 2007) and has been shown to occur in the deep roots of thick (> 40 km) continental arcs, as observed in the Cretaceous (120-100 Ma) Chelan Complex of the Cascades arc in Washington, USA (Dessimoz et al. 2012). The increasing (La/Sm) N and (Ce/Yb) N ratios of the KIC with increasing SiO 2 (Fig. 6), and the occurrence of abundant amphibole (10-30%) in the mafic and intermediate lithologies of the KIC suggest that hornblende fractionation played a significant role in the formation of the complex's calc-alkaline intermediate to felsic plutonic rocks (Chivas 1977;Davidson et al. 2007;Smith 2014). This indicates that significant amphibole fractionation can occur in thin, ca. 20 km-thick intra-oceanic island arcs, such as the Solomon 107 Page 16 of 21 arc. The crystallisation of amphibole after the KIC magmas had attained SiO 2 contents of > 60 wt.% led to a decrease in MgO, Al 2 O 3 , CaO and FeO T contents with increasing SiO 2 . This amphibole fractionation likely occurred after olivine (MgO) and plagioclase (Al 2 O 3 ) fractionation yet overlapped with clinopyroxene and plagioclase (CaO), and biotite and Fe-Ti oxide (FeO T ) fractionation (e.g. Loucks, 2021). The occurrence of abundant biotite (4-16%) in the mafic to intermediate plutonic rocks of the KIC (Chivas 1977) imply that biotite fractionation played a role in the formation of this complex. Biotite fractionation led to a decrease in the Al 2 O 3, FeO T and MgO contents of the KIC with increasing SiO 2 (Fig. 4). Apatite fractionation commenced at ~ 60 wt.% SiO 2 , leading to a decrease in the P 2 O 5 contents of the KIC magmas and an upside-down V-shaped trend in the KIC data (Fig. 4).
The basaltic composition of the parental magmas of the KIC evolved to andesitic, dacitic and rhyolitic compositions by fractional crystallisation (Figs. 4 and 6). The crystallisation of successively more evolved lithologies (gabbro, diorite, quartz diorite, tonalite, granodiorite and aplite) drove the evolution of the KIC magmas to increasingly more evolved compositions. This was manifested by the successive crystallisation of the Kolochoro Quartz Diorite, Kolokemau Tonalite, Vurakara Quartz Diorite, Charilava Quartz Diorite, melanocratic tonalite, tonalite 1, tonalite 2, porphyritic tonalite, tonalite 3, Chikora Tonalite Porphyry and aplite dykes. The bulk basaltic and andesitic compositions of Cycle 1 and Cycle 2 (Figs. 4 and 6), respectively, were calculated by averaging the respective compositions of the Cycle 1 and Cycle 2 samples, the latter being similar to that of the andesitic composition of the average continental crust (Rudnick 1995;Hawkesworth et al. 2010). Moreover, the bulk andesitic composition of Cycle 2 is very similar to the compositions of the andesite dykes that are exposed in the KIC and were mostly emplaced after the crystallisation of the Inamumu Zoned Pluton (Table S3 and Table S5). These dykes provide further evidence for the evolution of the KIC magmas from basaltic to more evolved compositions.

Variation of slab input into the primary magmas of the KIC
The distinctive Nd, Pb and Sr isotope ratios of the Cycle 1 and 2 intrusions indicate that they may have formed by partial melting of a sub-arc mantle source (Chivas et al. 1982) that was variably enriched by slab-derived melts and/ or fluids. The more radiogenic Pb isotope compositions and higher Ba/La and (Ce/Yb) N and lower Nb/La ratios of the Cycle 2 intrusions compared to the Cycle 1 intrusions suggest that the former was derived by partial melting of a similar sub-arc mantle source that was more re-enriched by LILE and LREE-rich slab-derived melts and/or fluids (Hawkesworth et al. 1991;Kent and Elliott 2002).
The slab-derived melts may have been generated as a consequence of the partial melting of subducting basaltic oceanic crust and/or sediments. Mantle-sediment mixing modelling based on the Nd, Pb and Sr isotopic compositions of the KIC, pelagic marine sediments and depleted MORB mantle (DMM) indicates that the KIC's primary magmas were derived from a DMM-like mantle source with a minor (1-2%) contribution of slab-derived sediment melts to the sub-arc mantle source(s) of the KIC (Fig. 8; Kepezhinskas et al. 1997;Schoenhofen et al. 2020). There is no evidence for melting of subducting altered oceanic crust because the (Dy/Yb) N ratios are low and do not show input of melts from a source with residual garnet, e.g. from eclogite (Figs. 6 and 8). Based on their 206 Pb/ 204 Pb ratios, the Cycle 2 intrusions had a slightly greater contribution of slab-derived Pb via melts or fluids in their sub-arc mantle source than the Cycle 1 intrusions (Fig. 8). However, the overlapping 87 Sr/ 86 Sr and 143 Nd/ 144 Nd ratios of Cycles 1 and 2 suggest that the amount of sediment melt input did not change significantly between these cycles (Fig. 7). The Cycle 2 intrusions were more influenced by slab-derived sediment melts than the Cycle 1 intrusions (Fig. 8). The sub-arc mantle source of the Cycle 2 intrusions was more enriched in LILE and HFSE and had higher LREE contents, and 206 Pb/ 204 Pb and 208 Pb/ 204 Pb isotopic ratios compared to the sub-arc mantle source of the Cycle 1 intrusions (Figs. 5,6,7 and 8). This suggests that the parental magmas to the KIC were derived by partial melting of a heterogeneous depleted mantle source that may have been re-enriched with distinct slab-derived sediment melts to various extents. Slab-derived sediment melts also played a minor role (1-3%) in the genesis of lavas from the Solomon arc ( Fig. 8; e.g. Kepezhinskas et al. 1997;König et al. 2007). There are some minor differences between the N-MORBnormalised trace element patterns of the KIC and the lavas from the Solomon arc (Fig. 5). These differences may be due to the respective locations of the other Solomon Islands between the inactive Vitiaz and active San Cristobal trenches and the consequent varying interactions between both subduction zones beneath these islands (e.g. Smith et al. 2009).

Comparison of the KIC with ultramafic to felsic plutons in Cenozoic intra-oceanic island arcs
Ultramafic to felsic plutons similar to the KIC occur in other Cenozoic intra-oceanic island arcs in the Pacific Ocean, such as the Izu-Bonin-Mariana, Aleutian, Vitiaz and Bismarck island arcs (Whalen 1985;Kawate and Arima 1998;Suzuki et al. 2015;Kay et al. 2019;Marien et al. 2022). These plutons show similar whole-rock compositional ranges from pyroxenites and gabbros to diorites, tonalites, and granodiorites. They also show similar major and incompatible element variations with SiO 2 contents to the KIC and comparable N-MORB-normalised trace element ratios (Figs. 4, 5 and 6; Table S8). Thus, we suggest that very similar processes dominated by fractional crystallisation formed intermediate and felsic magmas from mafic melts, which is in agreement with the majority of studies on these plutons (Whalen 1985;Kay et al. 2019;Marien et al. 2022). The oldest rocks in the KIC and Uasilau-Yau Yau Intrusive Complex on the island of New Britain in the Bismarck island arc are gabbros (Fig. 4) and it appears that the more evolved rocks in Cenozoic ultramafic to mafic plutons represent magmas that formed from these initial mafic melts. Thus, the evolved compositions are generally less abundant than the gabbros. Amphibole fractionation causes an increase in (La/Sm) N and decrease in (Dy/Yb) N with increasing SiO 2 contents (Davidson et al. 2007;Smith 2014), which is observed to a variable extent in all of the aforementioned Cenozoic ultramafic to felsic arc plutons ( Fig. 6; e.g. Whalen 1985;Kawate and Arima 1998;Suzuki et al. 2014Suzuki et al. , 2015Kay et al. 2019;Marien et al. 2022).
The Miocene-Pliocene (8.9-4.0 ± 0.2 Ma) Tanzawa Plutonic Complex in the active Izu-Bonin-Mariana intraoceanic arc of Japan has been proposed to have formed by partial melting of mafic lower arc crust (Kawate and Arima 1998;Suzuki et al. 2014Suzuki et al. , 2015. However, the mostly linear trends exhibited by the Tanzawa Plutonic Complex are very similar to those displayed by the KIC (Fig. 4) and are most easily explained by fractional crystallisation processes given that we have ruled out other processes for the latter. The great similarity in the petrology and major and trace element geochemistry of the Tanzawa Plutonic Complex and KIC indicates that these young intra-oceanic island arc-derived plutonic complexes most likely formed by the same process. Additionally, Suzuki et al. (2015) found that most δ 18 O zircon isotope values of the Tanzawa tonalites were in the range of 5.3 ± 0.3‰, indicating equilibrium with the mantle (Valley et al. 1998), which are similar to the mantle-like wholerock δ 18 O compositions (5.4 to 7.2‰) of the KIC (Chivas et al. 1982). The highest δ 18 O compositions (> 6.0‰) from the KIC can be attributed to temperature-dependent fractionation of oxygen isotopes (Chivas et al. 1982). Only a few Tanzawa samples have low δ 18 O zircon isotope values that could be due to melting of hydrothermally altered oceanic arc crust (Suzuki et al. 2015).
The Late Oligocene-Early Miocene (34.6-30.9 ± 0.4 Ma) Hidden Bay and Middle Miocene (14.7-13.9 ± 0.3 Ma) Kagalaska plutons are located in the active central Aleutian intra-oceanic island arc in Alaska, USA, and had initially basaltic melts, containing ~ 7.6 wt.% MgO, that fractionated olivine, clinopyroxene and pargasitic amphibole at ~ 12 km depth leading to the emplacement of predominantly granodioritic magma(s) into the shallow crust at depths of 2-4 km (Kay et al. 2019). The initial basaltic magma of Cycle 1 intruded the shallow crust at the KIC had > 5 wt.% MgO, thus resembling the processes observed in the Aleutian island arc. As is the case with the KIC, the MgO, CaO and FeO T contents of the Aleutian plutons decrease with increasing SiO 2 , signifying olivine, clinopyroxene, amphibole, plagioclase, biotite, and Fe-Ti oxide fractionation ( Fig. 4; Kay et al. 2019). The Al 2 O 3 contents of the Aleutian plutons also gently decrease from ~ 18 to ~ 17 wt.% between ~ 49 and ~ 62 wt.% SiO 2 (plagioclase fractionation) before decreasing more steeply (amphibole and biotite fractionation). Moreover, the Na 2 O contents of the Aleutian plutons increase with increasing SiO 2 , indicating plagioclase and amphibole fractionation. The P 2 O 5 contents of the KIC and Aleutian plutons increase between ~ 49 and ~ 56 wt.% SiO 2 before decreasing due to the onset of apatite fractionation (Fig. 4).
The (La/Sm) N and (Ce/Yb) N ratios of the KIC and Aleutian plutons increase with increasing SiO 2 , pointing to amphibole fractionation (Fig. 6). The Aleutian arc is approximately 37 km-thick beneath the Hidden Bay and Kagalaska plutons (Kay et al. 2019), almost twice the thickness of the 20 km-thick Solomon intra-oceanic island arc crust beneath the KIC; therefore, hornblende fractionation would have been more likely in the thicker arc crust of the former (Petford and Atherton 1996; Dessimoz et al. 2012). This is borne out by the more distinctive decrease in the (Dy/Yb) N ratios with increasing SiO 2 in the Aleutian plutons compared to the KIC (Fig. 6). Moreover, the Aleutian plutons have higher (Dy/Yb) N and (Ce/Yb) N ratios at a given SiO 2 content than the KIC (Fig. 6).
It has been proposed that island arcs exhibit chemical stratification; however, this has only been shown for arc crust thicker than ~ 35-40 km, such as the Triassic-Jurassic Talkeetna arc, Alaska, and the Cretaceous-Eocene Kohistan arc in Pakistan (Greene et al. 2006;Jagoutz 2014). Chemical stratification models suggest that mafic lower crust and felsic upper crust formed early in the magmatic evolution of island arcs through the fractionation of ferromagnesian minerals from mafic magmas in the deep crust followed by the ascent of intermediate to felsic melts to the middle to upper crust (Annen et al. 2006;Greene et al. 2006;Jagoutz 2014). The Talkeetna and Kohistan arcs display this chemical stratification, with SiO 2 , K 2 O and Th/La decreasing with depth and CaO and MgO increasing with increasing depth (Table S8; Rioux 2006;Greene et al. 2006;Rioux et al. 2010;Jagoutz 2014).
The aforementioned Cenozoic ultramafic-felsic arc plutons have similar mineral assemblages, mineral (Mg-rich olivine, pyroxene and calcic plagioclase) and whole-rock compositions and emplacement depths, and all intruded into shallow (< 10 km) mafic island arc crust (Figs. 4, 5 and 6; see Chivas et al. 1982;Whalen 1985;Kawate and Arima 1998;Suzuki et al. 2015;Tapster et al. 2016;Kay et al. 2019;Marien et al. 2022). The Fo contents and Mg# numbers of olivine and clinopyroxene from the KIC and other Cenozoic arc plutons indicate that Mg-rich basaltic magmas with Mg# > 50-60 were emplaced into shallow island arc crust. The decreased density of high-Al basaltic magmas resulting from prior olivine and clinopyroxene fractionation and the similar density of the mafic island arc crust into which the ultramafic-felsic plutons intruded would have made their emplacement at shallow depths (< 10 km) more likely (e.g. Herzberg et al. 1983). This implies that the thickness, composition and age of arc crust govern whether ultramafic-felsic magmas can ascend to the shallow island arc crust. This may explain why the ultramafic to felsic KIC, Tanzawa Plutonic Complex and Uasilau-Yau Yau Intrusive Complex and Hidden Bay, Kagalaska, Colo, Yavuna and Momi plutons, are found in shallow island arc crust (< 10 km), whereas the Kohistan and Talkeetna arcs are chemically stratified (e.g. Rasmussen et al. 2022). This also shows that shallow fractional crystallisation of mantlederived basaltic magmas is an important crustal growth process in intra-oceanic arcs.
The magma reservoirs beneath active island arc volcanoes have been seismically detected at depths of 3-11 km (Tsuruga et al. 2006;Allard et al. 2016;Paulatto et al. 2019;Janiszewski et al. 2020;Yukutake et al. 2021), i.e. comparable to the intrusion depths of the aforementioned plutons. We suggest that the plutons studied here represent the fossil magma systems of arc volcanoes, which is supported by the fact that the Solomon arc lavas closely resemble the KIC plutonic rocks (Figs. 4, 5 and 6), indicating a close relationship between volcanic and plutonic rocks in the Solomon island arc. Therefore, the shallow island arc crust grows by multiple magma intrusion cycles beneath volcanoes and by the migration of magmatism with time.

Conclusions
(1) The Koloula Igneous Complex is divided into Cycle 1 and 2 intrusions, each representing a calc-alkaline liquid line of descent caused by the fractionation of olivine, clinopyroxene, plagioclase, amphibole, biotite, apatite, and Fe-Ti oxides. The continuous fractionation towards more evolved compositions within each cycle is recorded by the bulk rock compositional data and is also supported by decreasing An contents in plagioclase. Shallow Cenozoic plutons in other young Pacific islands arcs also show evidence for fractional crystallisation as the main process of magma evolution, implying that this process dominates crustal growth in intra-oceanic island arcs.
(2) Abundant hornblende and increasing (La/Sm) N and (Ce/Yb) N ratios with increasing SiO 2 in the KIC and other Cenozoic arc plutons demonstrate that horn-blende fractionation played a variably significant role in the genesis of their calc-alkaline felsic plutonic rocks, generating felsic magmas with an enrichment of light REE.
(3) The KIC formed by fractional crystallisation of basaltic parental magmas with 6-8 wt.% MgO that intruded at a depth of ~ 2.0-3.0 km (~ 0.1 GPa), i.e. similar to the depths determined for other intrusions in young Pacific island arcs. (4) Magmas of the Koloula Igneous Complex neither show evidence for crustal assimilation during fractional crystallisation nor is there evidence for the formation of the magmas by partial melting of mafic lower arc crust. (5) The parental magmas of the Koloula Igneous Complex were derived through hydrous melting of depleted MORB mantle sources that was enriched by slabderived sediment melts (~ 1-2%) and fluids. The mantle source of the Cycle 2 magmas was slightly more affected by slab-derived sediment melts and fluids with a distinct Pb isotope signature compared to the Cycle 1 intrusions. (6) Ultramafic to felsic plutons in Cenozoic intra-oceanic island arcs in the Pacific resemble the composition of arc lavas and are thus the fossil roots of arc volcanic systems. The formation and emplacement of ultramafic to mafic plutonic rocks at shallow depths (< 5 km) imply a relatively mafic upper island arc crust in contrast to the chemical stratification of thicker arcs. This also permitted the ascent of basaltic magmas to shallow depths.
adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.