Equilibrium crystallization of massif-type anorthosite residual melts: a case study from the 1.64 Ga Ahvenisto complex, Southeastern Finland

Fe–Ti–P-rich mafic to intermediate rocks (monzodiorites and oxide–apatite–gabbronorites, OAGNs) are found as small intrusions in most AMCG (anorthosite–magnerite–charnokite–granite) suites. The origin of the monzodioritic rocks is still debated, but in many studies, they are presumed to represent residual liquid compositions after fractionation of anorthositic cumulates. In the 1.64 Ga Ahvenisto complex, SE Finland, monzodioritic rocks occur as minor dike-like lenses closely associated with anorthositic rocks. We report new field, petrographic, and geochemical (XRF, ICP-MS, EMPA) data complemented with crystallization modeling (rhyolite-MELTS, MAGFRAC) for the monzodioritic rocks, apatite–oxide–gabbronorite, and olivine-bearing anorthositic rocks of the Ahvenisto complex. The presented evidence suggest that the monzodioritic rocks closely represent melt compositions while the apatite–oxide–gabbronorite and olivine-bearing anorthositic rocks are cumulates. The monzodioritic rocks seem to form a liquid line of descent (LLD) from primitive olivine monzodiorites to more evolved monzodiorites. Petrological modeling suggests that the interpreted LLD closely corresponds to a residual melt trend left after fractional crystallization (FC) and formation of the cumulate anorthositic rocks and minor apatite–oxide–gabbronorite in shallow magma chambers. Consequent equilibrium crystallization (EC) of separate monzodioritic residual magma batches can produce the observed mineral assemblages and the low Mg numbers measured from olivine (Fo25–45) and pyroxenes (En48–63, Mg#cpx 60–69). The monzodioritic rocks and apatite–oxide–gabbronorites show similar petrological and geochemical characteristics to corresponding rock types in other AMCG suites, and the model described in this study could be applicable to them as well.

Regarding the source of the anorthosite parental magmas, there are two main schools of thought: the advocates of (1) mantle (e.g., Emslie 1978;Mitchell et al. 1995;Frost and Frost 1997) and (2) lower crustal origins (e.g., Duchesne et al. 1999). Regardless of the penultimate source, formation of anorthosite suites is presumed to have undergone at least two stages (Ashwal and Bybee 2017). First is the accumulation of buoyant plagioclase in deep magma reservoirs, which is followed by continuous, polybaric crystallization and ascent of the low-density plagioclase mushes to emplacement and final crystallization levels at 5-10 km depth in the upper crust (e.g., Ashwal 1993;Duchesne et al. 1999;Charlier et al. 2010;. Most studies of anorthosites also agree that their parental magmas have undergone significant assimilation of crustal material, which is evident in the isotopic signatures of the resultant rock types (e.g., Mitchell et al. 1995;Heinonen et al. 2015) and that their formation involved several magma pulses (Ashwal and Twist 1994). The granitic rocks are considered either to be consanguineous with the anorthositic rocks (e.g., Frost and Frost 1997) or to comprise a separate lithological group, usually associated with lower crustal origins Ashwal and Bybee 2017).
Despite the focus of most previous studies on anorthosite petrogenesis, the monzodioritic rocks have also received warranted research interest since they are likely to bear key evidence on the petrogenesis of AMCG suites (e.g., de Waard and Romey 1969; Duchesne et al. 1974;Demaiffe and Hertogen 1981;Ashwal 1982;Duchesne 1984;Owens and Dymek 1992;McLelland et al. 1994;Mitchell et al. 1996;Markl 2001;Fred et al. 2019). The amount of monzodioritic rocks is usually minor relative to the other rock types in AMCG suites, and they are observed as dikes or other small intrusions in the contacts of associated anorthositic rocks or between the anorthositic and granitic rocks (Duchesne 1984). Whereas the anorthositic group consist of coarsegrained plagioclase cumulates that are far from melt compositions, the monzodiorites are usually fine-grained, contain more mafic minerals, and are often thus presumed to represent near melt compositions (e.g., Emslie et al. 1994;Markl and Frost 1999). In addition to high Fe, Ti, and P, the monzodiorites have low Mg-numbers and low SiO 2 , and moderate to low Al 2 O 3 (Duchesne 1984).
Oxide-apatite-gabbronorites (OAGN) are also found as dikes or other small intrusions in many anorthosite suites and are grouped together with the monzodiorites in some studies (see McLelland et al. 1994). These rocks are low in SiO 2 (27-42 wt%), high in TiO 2 (4-10 wt%), FeO tot (17-38 wt%), and P 2 O 5 (3-6 wt%), but have low K 2 O (Owens and Dymek 1992). The Al 2 O 3 content varies in respect to the amount of plagioclase (e.g., Owens and Dymek 1992). As their name suggest, OAGNs contain significant amounts of Fe-Ti-oxides, apatite, and also mafic phases (pyroxenes ± olivine, e.g., McLelland et al. 1994;Dymek and Owens 2001). Based on the high apatite content and enrichment in incompatible elements, they are presumed to represent a late-stage, highly differentiated magma fraction in the evolution of massif-type anorthosite suites (e.g., McLelland et al. 1994;Mitchell et al. 1996), although Dymek and Owens (2001) have suggested that OAGNs can form at several stages during the magmatic evolution of AMCG suites. Texturally, OAGNs are diverse and range from cumulates to finer-grained rocks (e.g., McLelland et al. 1994;Mitchell et al. 1996). In this study, the oxide-apatite-rich gabbronorites (OAGNs) are considered separately from the monzodioritic rocks and since the sample of this study contains more oxide it is hereafter referred as apatite-oxide-gabbronorite.
The Ahvenisto complex in southeastern Finland (Fig. 1) is one of the key locations for Fennoscandian massif-type anorthosite studies (Alviola et al. 1999;Heinonen et al. 2010aHeinonen et al. , b, 2015Heinonen et al. , 2020Heinonen 2012;Fred et al. 2019), as it is the best representative of AMCG suite rocks at the current level of exposure in Finland (e.g., Rämö 1991). Recent studies in the Ahvenisto complex area have focused on the mantle vs. crustal sources of the complex (Heinonen et al. 2010b(Heinonen et al. , 2015, the physical and chemical interactions of the associated monzodioritic and granitic magmas (Fred et al. 2019), and the polybaric crystallization history of the Ahvenisto anorthosite (Heinonen et al. 2020). Heinonen et al. (2010b) and Fred et al. (2019) preliminarily suggested that the monzodioritic rocks may approximate melt compositions and form trends that could be interpreted as a liquid line of descent (LLD) of the residual melt left after fractional crystallization of the anorthositic cumulates. In this study, we test this hypothesis by examining further petrographic and geochemical evidence together with crystallization modeling to evaluate the role of different modes of crystallization (fractional vs. equilibrium) in the formation of various rock types of the Ahvenisto complex. We compare our data and models to a global compilation of monzodioritic rocks and OAGNs and discuss the various hypotheses for the origin of the monzodioritic rocks.

The 1.64-Ga Ahvenisto complex
The 1.64-Ga Ahvenisto AMCG complex (Savolahti 1956(Savolahti , 1966Johanson 1984;Alviola et al. 1999;Heinonen et al. 2010bHeinonen et al. , 2015Heinonen et al. , 2020 belongs to the Fennoscandian rapakivi suite (e.g., Rämö and Haapala 2005;Rämö et al. 2014) and is located northwest of the Wiborg batholith in southeastern Finland (Fig. 1). At the current level of exposure, the granitic rocks dominate the rapakivi suite, and only a small amount of anorthositic and mafic rocks are present as dikes or other small intrusions (e.g., Eklund 1993;Alviola et al. 1999;Rämö and Haapala 2005;Heinonen et al. 2010a, b;Fred et al. 2019). The Ahvenisto complex intruded the Svecofennian (1.9-1.8 Ga) country rocks and the crystallization ages of the different rock types vary from 1643 to 1632 Ma (Heinonen et al. 2010b). The complex comprises a granitic intrusion (ca. 70% of areal extent) surrounded by an anorthositic arc (25%) that includes volumetrically minor monzodioritic rocks found as dike-like lenses (5% of the area, e.g. Savolahti 1956Savolahti , 1966Johanson 1984;Alviola et al. 1999;Heinonen et al. 2010b;Fred et al. 2019)-the usual lithological assemblage of AMCG complexes . Field relations suggest that the anorthositic rocks are the oldest of the mafic rocks and monzodioritic rocks are the youngest (Alviola et al. 1999). The granitic rocks are dominantly younger than all the mafic rocks (Alviola et al. 1999), but mingling structures have been reported as evidence of simultaneous emplacement of the monzodioritic and the earliest granitic melts (Fig. 2a;Fred et al. 2019).
Hafnium, Nd, Sr, and O isotope data indicate a depleted upper mantle origin for the parental magmas of the anorthositic rocks (e.g., Heinonen et al. 2010a, b). The anorthositic rocks themselves display a largely crustal isotope signature, likely resulting from assimilation of crustal Outcrop photos showing some typical features of the different rock types in the Ahvenisto complex: a fine-grained (< 1 mm) monzodioritic material mingled with granitic material in pillow and net-veined structures, b fine-to medium-grained (< 1-5 mm) olivine monzodiorite with granitic inclusion, c coarse-grained olivine-bearing anorthositic rock with cumulus plagioclase (~ 1 cm) and inter-stitial mafic material, and d coarse-grained anorthositic rock with a pocket of orthopyroxene (HAOM) and plagioclase megacrysts (up to 10 cm). a and b are from the Iso Kuoppalampi area, c from the north eastern area, and d is located southwest from the Tuuliniemi area (Fig. 1) materials (e.g., Heinonen et al. 2010aHeinonen et al. , b, 2015. The granitic magmas of the suite were produced from partial melts of the lower crust with a possible minor mantle component (e.g., Heinonen et al. 2010aHeinonen et al. , 2015. Nd-Sr isotopes suggest a comagmatic origin for the anorthositic (initial Ɛ Nd − 0.9 to − 0.5 and Sr 0.7037 to 0.7041) and monzodioritic rocks (initial Ɛ Nd − 1.1 to − 0.2 and Sr 0.7028 to 0.7040; Heinonen et al. 2010b). Al-in-opx geobarometry suggests that the anorthositic rocks crystallized polybarically at three different levels: (1) high-P conditions in lower crust (~ 1100 MPa), (2) mid crustal levels (~ 500 MPa), and (3) upper crustal levels (~ 200 MPa) (Heinonen et al. 2020).

Monzodioritic rocks
The monzodioritic rocks are usually exposed between the outer rim of the anorthositic arc and either the hornblende granites (southeastern part) or the Svecofennian country rocks (northwestern part; Fig. 1). A small lens of monzodioritic rocks is found inside the anorthositic arc in the northeastern corner of the complex ( Fig. 1; Pitkäjärvi area). The monzodioritic dike-like lenses mostly consist of (1) fine-to medium-grained (< 1-3 mm) massive monzodiorite with poikilitic hornblende phenocrysts or (2) fine-grained (< 0.5 mm) monzodiorite as pillow-like structures mingled with net-veined granitic material (Fig. 2a;Johanson 1984;Alviola et al. 1999;Fred et al. 2019). The granitic material involved in the mingling structures is presumed to be associated with the marginal hornblende granites (Fig. 1) and evidence of mixing of these two magma types has also been observed in hybrid rocks (Fred et al. 2019). Also, a small lens (~ 200 m wide) of (3) medium-grained (1-5 mm) and dark olivine-monzodiorite is found in the southern part of the complex (Figs. 1, 2b; Iso Kuoppalampi area; Fred et al. 2019).
The massive monzodiorite (Fig. 3b) consists of finegrained groundmass of plagioclase, pyroxenes, and K-feldspar with minor amounts of quartz and biotite (Fred et al. 2019). Poikilitic hornblende crystals stand out from the groundmass (Fig. 3b). Typical accessory phases are oxide, apatite, and zircon. The pillow-type monzodiorite (Fig. 3a) consists of fine-grained groundmass of plagioclase, amphibole, biotite, and K-feldspar, and minor amounts of oxide and apatite (Fred et al. 2019). Quartz is an accessory phase. Plagioclase phenocrysts and amphibole pseudomorphs (replacing pyroxene) stand out from the groundmass (Fig. 3a). A few fresh pyroxene phenocrysts are present (Fred et al. 2019).
Geochemically, the Ahvenisto monzodioritic rocks seem to form an evolutionary trend from olivine monzodiorite through pillow type-monzodiorite to massive monzodiorite (Fred et al. 2019). The contact between the monzodioritic rocks and the anorthositic cumulate rocks, where visible, is either (1) sharp and monzodiorite shows chilled margins (Alviola et al. 1999), (2) locally gradational in wide scale, or (3) complex with sheets of granitic, and country rock material brecciated by monzodioritic and anorthositic rocks (Fred et al. 2019). The monzodioritic rocks are hereafter referred to as monzodiorites (pillow-type monzodiorites and massive monzodiorites) and olivine monzodiorites.

Methods and materials
Much of the field observations that form the basis of this study were performed and reported in Fred et al. (2019), but also some complementary field work was conducted and is discussed in the results section below. During the field work, monzodioritic rocks (eight monzodiorite and eight olivine monzodiorite samples), olivine-bearing anorthositic rocks (three samples), and apatite-oxide-gabbronorite (one sample) were sampled with a hammer for petrography,  Fred et al. 2019) in cross-polarized light. The monzodioritic rocks have fine-grained groundmass (pillows < 0.1 mm, massive < 0.5 mm, olivine monzodiorite < 1 mm). The monzodiorite pillows and olivine monzodiorite contain plagio-clase phenocrysts (1-10 mm) and the former also amphibole pseudomorphs. Poikilitic amphibole is common in massive monzodiorite. In contrast, the anorthositic rocks (e, f) are coarse-grained (< 1 cm) and consist of cumulus plagioclase with usually interstitial mafic minerals (20-40 vol%). The apatite-oxide-gabbronorite is medium-grained (1-10 mm) and contains high amount of mafic minerals (~ 50 vol%), ilmenite, and apatite. Abbreviations are according to Whitney and Evans (2010) whole-rock major and trace element geochemistry (X-ray fluorescence, XRF, and inductively coupled plasma mass spectrometry, ICP-MS), and major element mineral chemistry (electron microprobe analyzer, EMPA). The 20 new samples and 15 samples from Fred et al. (2019) were analyzed for whole-rock geochemistry. A more detailed petrographic study of 12 new samples was conducted and nine olivine-bearing samples (six olivine monzodiorites, two olivine-bearing anorthositic rocks, and one apatite-oxide-gabbronorite) were selected for mineral chemical analysis. All sample preparation stages were conducted at the Mineralogical Laboratory of the University of Helsinki.

XRF and ICP-MS analyses
Whole-rock samples were prepared with a rock saw to avoid any weathering surfaces and crushed with a jaw crusher (minor contamination with Fe, Cr, Ni, and V is possible). Representative pieces were handpicked to avoid altered and weathered parts. The whole-rock XRF and ICP-MS analyses for the 20 new samples were conducted at the Geoscience Laboratories, Ontario, Canada. The samples were pulverized in a 99.8% Al 2 O 3 planetary ball mill (minor contamination with Al) and run for LOI in 105 °C under nitrogen atmosphere and 1000 °C under oxygen atmosphere. The calcined samples were fused with a borate flux to produce glass beads for the XRF analyses (Hargreaves 2015). The deviation from standard is < 3% for all major elements. The coefficient of variation is < 2%. The preparation of samples for ICP-MS was conducted using closed vessel multi-acid digestion method for complete dissolution (Hargreaves 2017). The deviation from standard is ≤ 5% for most of the trace elements, except Nb (6%), Hf (8%), Cs (15%), Pb (17%), Tb (24%), Er (9%), Tm (23%), Lu (12%), and Li (6%). The coefficient of variation is ≤ 5% for most of the trace elements, except Ni (8%) and Pb (6%).The 15 previously analyzed samples (Fred et al. 2019) were reanalyzed for major and trace element compositions at the USGS laboratories. Sample preparation procedures for these samples are described in Fred et al. (2019) and analytical details follow those described in Taggart and Siems (2002). The deviation from standard is ≤ 2% for all major elements analyzed with XRF, except for MnO (6%). The coefficient of variation for major elements is ≤ 2%. The deviation from standard is ≤ 5% for most of the trace elements analyzed with ICP-MS, except for Cr (13%), Ni (8%), Cu (11%), Ta (11%), Nb (16%), Y (9%), Tm (7%), and Yb (7%). The coefficient of variation for REE is ≤ 2% and other trace elements ≤ 15%. The wholerock samples were also measured for ferrous iron. The 20 new samples were dissolved in an aggressive, non-oxidizing acid mixture, and the solubilized ferrous iron was quantified by potentiometric titration with a standardized permanganate solution (Hargreaves 2019). The deviation from standard is 1% for FeO and the coefficient of variation is 0.5%. The whole dataset together with standard measurements is given in Online Resource 1.

EMPA analyses
Electron microprobe analyses were conducted at the Mineralogical Laboratory of the University of Helsinki using a JEOL JXA-8600 electron microprobe analyzer (EMPA) equipped with four wavelength-dispersive (WDS) spectrometers, upgraded with SAMx analytical software and Poin-tElectronic SAMx hardware. Six carbon-coated thin sections of olivine monzodiorite, two olivine-bearing anorthositic rocks, and one apatite-oxide-gabbronorite were selected for quantitative (WDS) major element analysis of the mineral phases. Additional qualitative (energy dispersive spectrometer, EDS) analyses were performed to identify accessory phases. An acceleration voltage of 15 kV, sample current of 15 nA, and beam size of ca. 10 µm were used for the WDS analyses. Calibration was performed using a set of natural and synthetic oxide and silicate standards. Matrix corrections for the analysis were done by the SAMx analysis software using the PAP correction. The potential Si Kβ overlap over Sr Kα has been calculated after the Si Kβ intensity at the Sr Kα peak position was measured. General precision is estimated to be about 2% for major elements (> 10 wt%) and below 5% for minor elements (1-10 wt%; Michallik et al. 2017). The whole dataset is given in Online Resource 1.

MELTS software
MELTS is a software package that facilitates thermodynamic modeling of phase equilibria in magmatic systems (Ghiorso and Sack 1995;Gualda et al. 2012). It can be used to simulate igneous processes, such as equilibrium and fractional crystallization. The most recent rhyolite-MELTS versions (Gualda et al. 2012) have been appended with H 2 O-CO 2 fluid saturation models (Ghiorso and Gualda 2015) and can be used for natural silica-dominated compositions over temperatures of 500-2000 °C and pressures of 0-2 GPa. The software predicts phase equilibria (solid phases, melt, and fluid, and their major element compositions) from bulk composition of the system at, for example, given pressure-temperature conditions. The conditions can vary according to user-defined steps: decreasing temperature steps can be utilized in modeling of equilibrium (all phases stay in equilibrium) or fractional (formed solids are removed every step) crystallization.

MAGFRAC program
The MAGFRAC program of Morris (1984) is based on the least squares approximation of best-fit compositions. The least squares approximation is based on the following: "computer estimated proportions of fractionating minerals are 'mixed' with the daughter liquid to produce the best-fit estimated parent composition. The suitability of this estimated parent composition is given by squaring the sum of differences (i.e. residuals) between the estimate and observed parent compositions" (Bryan et al. 1969). In MAGFRAC, the input includes user-defined oxide compositions of the observed parent melt, daughter melt, and presumed fractionating minerals (1-X phases; Morris 1984). The program gives estimated parent melt composition, sum of squared residuals, and mineral and daughter liquid proportions (Morris 1984).

Field work
The focus of the recent field work was to gather additional and new knowledge on the most primitive olivine-bearing anorthositic rocks and monzodioritic rocks of the Ahvenisto complex. Six representative mapping locations were selected from these lithological units (Fig. 1).
As already discussed, the Ahvenisto complex hosts three different monzodiorite types: massive monzodiorite, which is the most common, pillow-type monzodiorite, and rare olivine monzodiorite (See Fred et al. 2019). Due to scarce outcrops in the assumed contact zones, the contacts between the olivine-bearing or other anorthositic rocks and monzodioritic rocks that have been described as sharp in earlier studies (Johanson 1984;Alviola et al. 1999), were not observed. Instead we found that, locally, and near the assumed contacts, the grainsize of anorthositic rocks decreased and that of monzodioritic rocks increased. This phenomenon may be related to uneven cooling of the individual intrusions in a complex setting of cold basement and still warm AMCGrelated wallrocks.
The interaction structures (Johanson 1984;Alviola et al. 1999;Fred et al. 2019) were found to be much more widespread than previously reported. The mingling structures occur throughout the monzodioritic rocks and also hybrid rocks are observed in several locations in the southeastern part of the complex. The olivine monzodiorites seem to be limited to a small area near lake Iso Kuoppalampi (Fig. 1). Due to scarce outcrops, the mode of occurrence (i.e. dike or lens, etc.) of this olivine monzodiorite could not be determined, but it is presumed to represent a more mafic and coarser-grained part of the monzodioritic dike. The observed areal existence of the olivine monzodiorite is ~ 0.02 km 2 .
The Ahvenisto olivine-bearing anorthositic body was found to correspond to previous descriptions: plagioclase (grainsize ~ 1 cm) cumulate rock with interstitial mafic minerals (20-40% vol%). However, this body was found to also contain lenses of pegmatitic material with compositions varying from very coarse-grained anorthositic rock with orthopyroxene megacrysts to megacrystic plagioclase-pyroxene rock. In the southern and eastern margins of the olivine-bearing anorthositic body, a medium-to coarsegrained (0.5-1 cm) cumulate rock with high amount (up to 50 vol% or more) of mafic minerals and red alteration color (usually the anorthositic rocks show light gray alteration color) was observed.

Petrography
A generalized overview of the petrography of the different rock types was provided in the "Geological background" section. Here we give more detailed descriptions that are appended with the new data and observations.
The massive monzodiorite consists of fine-grained groundmass of plagioclase, K-feldspar, quartz, and pyroxenes with poikilitic hornblende (Fig. 3a;Fred et al. 2019). The pillow-type monzodiorite consists of very fine-grained groundmass of plagioclase, K-feldspar, amphibole, and biotite with plagioclase phenocrysts and amphibole pseudomorphs (Fig. 3b;Fred et al. 2019). The olivine monzodiorite consists of euhedral and fresh plagioclase laths (0.1-1 mm) with ortho-and clinopyroxene (< 0.5 mm) and olivine (< 0.5 mm) as the other major phases (Fig. 3c). Olivine is usually subhedral and fresh. Pyroxenes are anhedral and show strong exsolution of clinopyroxene in orthopyroxene and vice versa. Plagioclase phenocrysts (up to 5 mm) are common and occasionally show normal zoning. Fe-Ti oxides (ilmenite + minor magnetite), apatite, and interstitial K-feldspar, and minor biotite and zircon are found as accessory phases. The modal amount of plagioclase in olivine monzodiorite is approximately similar to the amount of mafic phases (olivine + pyroxenes).
The olivine-bearing anorthositic rocks are dominated by cumulus plagioclase with interstitial ortho-and clinopyroxene and olivine is also found as a major phase (15-40 vol%; Fig. 3e). Fe-Ti oxides (ilmenite + minor magnetite) and biotite are common accessory phases. The grain size of plagioclase is usually ~ 1 cm but clusters of smaller (2-5 mm) grains are also observed. Some of the larger crystals show zoning. Olivine is fresh and sub-to anhedral and complex rims (consisting of, e.g., bitotite + orthopyroxene + symplectite) are common. Pyroxenes are altered and clinopyroxene exsolution lamellae in orthopyroxene and vice versa are common.

Whole-rock geochemistry
Representative major and trace element compositions of the whole-rock samples are illustrated in Figs. 4 and 5. The whole dataset is presented in Online Resource 1.

Major elements
The monzodioritic rocks have SiO 2 contents between 46 and 52 wt%: the olivine monzodiorites have usually < 50 wt% and the monzodiorites > 50 wt% with two exceptions that have SiO 2 of ~ 46 wt% (Online Resource 1). The MgO content of the monzodioritic rocks decreases with decreasing Mg# (Mg 2+ /Mg 2+ + Fe 2+ ), while the FeO tot content remains nearly constant (Fig. 4c,d). The olivine monzodiorites show strong increase in TiO 2 and slight increase in K 2 O and P 2 O 5 contents and decreased MgO with decreasing Mg# (Fig. 4a,d,g,h). The monzodiorites show decrease in MgO and increase in K 2 O and P 2 O 5 with decreasing Mg# (Fig. 4d,g,h). The two monzodiorite samples with low SiO 2 have anomalously high TiO 2 , FeO tot , CaO, and P 2 O 5 contents (Fig. 4a,c,e,g).

Trace elements
The amount of most of the incompatible trace elements increase linearly with decreasing Mg# from olivine monzodiorites to the most evolved monzodiorites. All the samples show slight enrichment in LREE relative to HREE (Fig. 5a). The apatite-oxide-gabbronorite has the highest REE contents with slightly flatter LREE and steeper HREE curves than the monzodioritic rocks (Fig. 5a). The olivine-bearing anorthositic rocks show strong positive Eu anomalies (Eu/ Eu* 1.58-1.88) and monzodiorites and apatite-oxide-gabbronorite negative Eu anomalies (Eu/Eu* 0.56-0.90 and 0.69, respectively). The olivine monzodiorites do not show notable Eu anomalies (Eu/Eu* 0.84-1.08). The olivinebearing anorthositic rocks and monzodiorites show similar incompatible trace element patterns, except that the olivinebearing anorthositic rocks are enriched in Sr while the monzodiorites and apatite-oxide-gabbronorite are depleted in Sr (Fig. 5b). The apatite-oxide-gabbronorite shows a more variable pattern. The olivine-bearing anorthositic rocks and monzodiorites are relatively enriched in Ba, K, and Zr and depleted in Ti. The monzodiorite sample with anomalously K 2 O at a given Mg# (Fig. 4) is the one with the high Rb content (Fig. 5b).

Mineral chemistry
The full dataset of Ahvenisto mineral chemical compositions (for six olivine monzodiorites, two olivine-bearing anorthositic rocks, and one apatite-oxide-gabbronorite) is available in Online Resource 1. Main observations are discussed below and illustrated in Figs. 6, 7 and 8.

Plagioclase
Plagioclase composition was measured from groundmass, phenocrysts, and inclusions in pyroxenes. The groundmass plagioclase composition in the olivine-bearing anorthositic rocks varies between An 50-66 , in olivine monzodiorites between An 48-58 , and in the apatite-oxide-gabbronorite between An 40-43 (Fig. 6; Online Resource 1). The plagioclase inclusions in olivine monzodiorites have similar compositions (An 50-53 ) as groundmass plagioclase. Some of the largest phenocrysts in olivine-bearing anorthositic rocks and olivine monzodiorites show normal zoning from An 53-62 (center) to An 39-51 (rim). All analyzed plagioclase grains contain small amounts of Fe and have low Or content (≤ 4 mol%).

Olivine
The olivine composition is homogenous within samples and notable differences in composition between finer and coarser grains were not observed. However, the Fo content of olivine decreases with decreasing Mg# of the corresponding host whole rock sample (Fig. 7). The Fo content of olivine varies between Fo 45-53 in olivine-bearing anorthositic rocks, between Fo 25-45 in olivine monzodiorites, and between Fo 19-20 in apatite-oxide-gabbronorite (Fig. 7; Online Resource 1). Olivine is not in Fe-Mg equilibrium with the measured host whole-rock compositions (Fig. 9a).

Pyroxene
Orthopyroxene and clinopyroxene are usually found together in the Ahvenisto samples. The pyroxenes exhibit clear  exsolution textures either as irregular blebs or as regular lamellae: in orthopyroxene the lamellae are clinopyroxene and vice versa. The lamellae are thin (< 10 µm) and their effects for the analyses are reflected in Fig. 8 as compositions intermediate between ortho-and clinopyroxenes. Ilmenite exsolution is also common. The En content of the orthopyroxenes in olivine monzodiorites varies between En 48-63 , in olivine-bearing anorthositic rocks between En 56-66 , and in apatite-oxide-gabbronorite between En 33-35 (Fig. 8; Online Resource 1). The Mg# of clinopyroxenes varies between 60 and 69 in olivine monzodiorites, 68 and 74 in olivine-bearing anorthositic rocks, and 45 and 51 in apatite-oxide-gabbronorite (Online Resource 1). The pyroxenes contain small amounts of MnO and Al 2 O 3 , and the clinopyroxene also contains minor amounts of Na 2 O and TiO 2 . As is the case with olivine, pyroxenes are not in Fe-Mg equilibrium with the measured host whole-rock composition in any of the studied rock types (Fig. 9b, c).

Discussion
This study mainly deals with the of the monzodioritic rocks in the Ahvenisto complex and their comparison to similar rocks in other anorthosite suites. Therefore, major part of the following inferences concentrate on their features and the discussion on the complementary cumulates remains more concise. The observation of apparent mineral-melt Fe-Mg disequilibrium in the olivine monzodiorites is discussed first to set the stage, followed by an outline of the overall evolution of the compositional monzodiorite evolution and their relationship with the cumulate rocks. Regardless of its relatively evolved composition, the apatite-oxide-gabbronorite of the Ahvenisto complex is not considered to belong to the monzodioritic rocks and is dealt with separately. The penultimate part of the discussion is devoted to the regional implications of the findings in context of Fennoscandian rapakivi magmatism, and finally a global comparison of the petrogenetic significance of monzodioritic rocks in AMCG complexes is briefly outlined.

Equilibrium crystallization of the monzodioritic rocks
Multiple lines of field, textural, and geochemical evidence support the hypothesis that most of the Ahvenisto monzodioritic rocks record evidence of melt compositions and that they most likely represent residual melts left after fractionation of plagioclase (± mafic minerals) during anorthositic cumulate formation-these are reviewed in detail below. The directly observed contacts between the monzodioritic and anorthositic rock are sharp and the former always cross cut the latter (Alviola et al. 1999), which establishes the later crystallization of the monzodioritic rocks relative to the cumulates. Majority of the monzodioritic rocks are fine-grained and show no textural or petrographic cumulate features. Furthermore, the major element compositions of most monzodioritic rocks show no evidence of phase accumulation (Fig. 4), except for some of the more evolved monzodiorites that deviate from the main Ahvenisto trend towards higher Al 2 O 3 and CaO and lower K 2 O, most likely reflecting minor accumulation of plagioclase in these samples (Figs. 4, 11). Some samples show enrichment in TiO 2 and P 2 O 5 due to ilmenite and apatite accumulation, respectively. Unlike the anorthositic rocks, the monzodioritic rocks either have no Eu-anomaly (some olivine monzodiorites) or have negative Eu-anomalies suggesting that they are not plagioclase cumulates and that the more evolved monzodiorites have gone through plagioclase fractionation. The REE abundances of the monzodioritic rocks are much higher than in the anorthositic cumulates and increase from olivine monzodiorites to the monzodiorites (Fig. 5), which suggest that they represent evolving residual liquid compositions. Being interpreted as approximate melt compositions, the primitive olivine monzodiorites represent a rock type that has previously not been reported in any other AMCG complex and as such open up a new perspective in the study of parental magma compositions of massif-type anorthosites. If the olivine monzodiorites closely represent melt compositions as we suggest, it is odd (at first sight) that neither the olivine nor pyroxenes found in them are in Fe-Mg equilibrium with the host rock (i.e., presumed melt) compositions (Fig. 9). Especially olivine shows remarkably low Mg (Fo 25-45 ) contents compared to what would be expected to be in equilibrium with melts that correspond to the monzodioritic rocks (Mg# 42-52, in equilibrium with Fo~7 0-80 ; Fig. 9).
To explain these abnormal, excessively evolved mineral compositions, several options were considered. Accumulation, mixing, or other xenocrystic origins seem implausible already based on the above-mentioned whole-rock geochemical evidence. There is no petrographic evidence of xenocrystic origin either: the olivine is compositionally homogenous at the scale of individual grains, does not contain reaction rims, and its crystal habit and textural features support in-situ crystallization Mass balance calculations (using MAGFRAC) using measured mineral chemistries give almost identical whole-rock compositions compared to the analyzed compositions, and the phase proportions yielded by the calculations are reasonable as well. The Fe 3+ / Fig. 4 Major element variation diagrams of Mg# [Mg 2+ / (Mg 2+ +Fe 2+ )*100] vs. a TiO 2 , b Al 2 O 3 , c FeO tot , d MgO, e CaO, f Na 2 O, g K 2 O, and h P 2 O 5 for the Ahvenisto complex monzodioritic rocks, olivine-bearing anorthositic rocks, and apatite-oxide-gabbronorite. The black solid arrow indicates plagioclase accumulation and black dashed arrow indicates accumulation of oxide, apatite, and olivine in 1:1:5, respectively ◂ Fe tot (0.08-0.22) of the samples is in the usual range of mafic rocks. Furthermore, although not in Fe-Mg equilibrium, evolution of the olivine compositions nevertheless seems to correlate with the evolution of the host rock compositions (Figs. 7,9), which suggests that they are still somehow linked to each other.
All these observations point to the initially counterintuitive conclusion that the apparent mineral-melt disequilibrium could actually be caused by a local equilibrium crystallization process. When magma crystallizes in equilibrium, the bulk Mg/Fe of the magma does not change during crystallization and the resulting whole-rocks preserve the initial Mg#. However, if the crystallizing phases equilibrate with the evolving residual melt, the last possible Mg/ Fe recorded by the crystals can be much lower compared to what it would be at liquidus temperatures. In contrast, in  In order to test this hypothesis, we used rhyolite-MELTS version 1.2.0 (Gualda et al. 2012) to produce equilibrium crystallization (EC) models for the olivine monzodiorites for which mineral chemical data are available (six samples, Table 1). The models were run in 100 MPa from liquidus temperatures (T l ) determined by MELTS in 5 °C temperature decrement steps with 0.6 wt% of H 2 O (Online Resource 1). The models usually yield the mineral assemblage (in the order of crystallization): olivine, plagioclase, spinel, clinopyroxene, apatite, orthopyroxene, oxide, biotite, and alkali feldspar ( Table 1). The olivine composition evolves throughout the crystallization model, and the measured olivine compositions are within the range of modeled olivine compositions, except for the most evolved olivine monzodiorite sample (Fig. 9). In the models, pyroxenes begin to crystallize later than olivine, and hence the composition of the remaining melt at that point is already more evolved. Therefore, the Mg# of the first pyroxene to crystallize in the models is already lower than what would be expected if the pyroxene started to crystallize at liquidus temperature. The measured pyroxene compositions are mainly within the range of the model compositions as well (Fig. 9). Note that the rhyolite-MELTS modeling discussed here concentrates on the compositions of the anhydrous Fe-Mg silicates in the most primitive olivine monzodiorites and does not suffer from the same issues of plagioclase stability than the fractional crystallization modeling of more evolved melts discussed in the next section.
Based on these modeling results, we suggest that the apparent Fe-Mg disequilibrium between the mafic minerals and the host olivine monzodiorite whole-rock compositions is the result of equilibrium crystallization. During equilibrium crystallization the forming crystals exchange elements with a continuously evolving melt composition, which in the end results in the minerals being out of equilibrium with the bulk magma system composition. The individual whole-rock samples would thus represent magma batches that crystallized in equilibrium conditions. The process that Measured mineral compositions of the Ahvenisto complex olivine monzodiorites, olivine-bearing anorthositic rocks, and apatiteoxide gabbronorites shown in a olivine-melt, b orthopyroxene-melt, and c clinopyroxene-melt equilibrium diagrams (Roeder and Emslie 1970, see also Putirka 2008) showing the Fe-Mg disequilibrium of the minerals with the measured whole-rock compositions. Black arrows show the range of evolving mineral compositions of modeled olivine-monzodiorites resulting from equilibrium crystallization models conducted with rhyolite-MELTS version 1.2.0 (Gualda et al. 2012). The red line is the calculated mineral-melt equilibrium trend. See text for more details ▸ produced the range of these initial whole-rock compositions is discussed in the following section.

Fractional crystallization of the cumulate rocks and LLDs of the residual melts
While the monzodioritic rocks closely correspond to melt compositions, field relations, petrography, and geochemistry clearly indicate that the olivine-bearing anorthositic rocks and apatite-oxide-gabbronorite are cumulates. They are coarse-grained and contain cumulus plagioclase. The olivine-bearing anorthositic rocks show evidence of plagioclase accumulation by having high amount of plagioclase and high Al 2 O 3 , CaO, Sr, and Eu and the apatite-oxide-gabbronorite shows evidence of ilmenite and apatite accumulation by having high modal proportions of those minerals and high TiO 2, FeO tot , and P 2 O 5 contents (Figs. 2, 3, 4). On the other hand, the apatite-oxide-gabbronorite shows trace element geochemical characteristics of fairly evolved magma compositions: high total REE, and a strong negative Eu-anomaly (Fig. 5). This might suggest that it represents a rather late-stage cumulate from the monzodioritic series magmas and possibly contains some interstitial melt as well. On the other hand, its REE pattern reflects the REE partition coefficients for apatite (highest for MREE, except for Eu; e.g., Sano et al. 2002), which is found as a cumulus phase. The net effect of accumulation of, for example, apatite and oxides on the major and trace element compositions of the apatite-oxide-gabbronorite sample is difficult to constrain and thus the parental melt composition for this rock type remains uncertain. We suggest that it was more similar to the evolved monzodiorites than olivine monzodiorites, however.
To constrain the possible process producing the cumulates and presumed LLD trend of the monzodioritic rocks (Fig. 10), we utilized two modeling approaches: (1) Rhyolite-MELTS version 1.2.0 (Gualda et al. 2012) was used to conduct thermodynamically constrained fractional crystallization (FC) models for the most primitive olivine monzodiorite composition at varying pressures (100-1200 MPa, 0.6 wt% of H 2 O; Table 1). (2) Best-fit parental melt compositions were calculated with MAGFRAC (Morris 1984) software in order to replicate the LLD in two stages using selected monzodioritic compositions as parent and daughter compositions and mineral compositions measured from the anorthositic cumulates (Johanson 1984) as the fractionating phases.
The rhyolite-MELTS FC simulations at 300, 600, and 1200 MPa did not give meaningful results (the software failed to find a solution) and are not considered here further. In Fig. 10, LLDs of the models ran at 100, 400, 700, and 1100 MPa are shown. This set of models illustrates the effect of increasing pressure adequately and the other successful runs plot within or very close to a field bordered by them. None of the simulations show very good fit with the data, but the ones run at lower pressures (especially the simulation at 100 MPa) are better.  For input parameters of the MAGFRAC (Morris 1984) model, we selected a two-stage approach based on major changes in the modal and geochemical composition between the olivine monzodiorites and monzodiorites. The division of the stages was set at Mg# of ~ 40 because rocks more primitive than that contain olivine and lack amphibole and have no or negligible negative Eu-anomaly, whereas rocks with Mg# < 40 lack olivine but contain amphibole and exhibit a negative Eu-anomaly. We selected the most primitive olivine monzodiorite as Parent 1, one of the intermediate monzodiorites as Parent 2 and Daughter 1, and one of the most evolved monzodiorites as Daughter 2. Representative analyses of plagioclase, pyroxenes, olivine, and ilmenite from the anorthositic rocks (Johanson 1984) were used as the compositions of the fractionating phases ( Table 2). The estimated parent compositions are very close fits to the parent compositions used in the calculations and model gives very reasonable r 2 values of 0.284 for stage 1 and 0.375 for stage 2 (Table 2; Fig. 10). A fractionation trend from the estimated parent to daughter is plotted in Fig. 10.
The major problem in the MELTS models is the delayed crystallization of plagioclase, which has also been observed to be an issue in some other occasions (see Putirka 2005). Petrographic observations indicate that plagioclase is the first phase to crystallize, but in the models plagioclase starts to crystallize only after the magma temperature has decreased 165 °C causing delayed depletion of Al 2 O 3 in the residual melt. Neither MELTS nor rhyolite-MELTS are well suited for igneous systems with intermediate composition and phase assemblage dominated by amphibole and biotite (Ghiorso and Sack 1995;Gualda et al. 2012). Both amphibole and biotite are major phases in the monzodiorites (but not in olivine monzodiorites, cf. previous section), which might also play a role in the mismatch between the rhyolite-MELTS fractionation models and the observations. The strength of the MAGFRAC modeling lies in the fact that it is based on actual observed mineral parageneses and measured mineral compositions. The MAGFRAC model also supports the assumption that the poor fit of the rhyolite-MELTS models in Al 2 O 3 and CaO is due to the inability of MELTS to correctly stabilize plagioclase in the modeled compositions.
The close fit of the MAGFRAC models to the observations, thus, strongly supports the hypothesis that the compositions of the monzodioritic magmas evolved in a fractionally crystallizing system that produced the juxtaposed anorthositic cumulates and that the monzodiorites represent their residual liquids. Individual monzodiorite magma batches were then intermittently tapped from this evolving reservoir and crystallized in equilibrium conditions as described in the previous section.

Implications for the petrogenesis of the Ahvenisto complex and Fennoscandian rapakivi magmatism
The Ahvenisto complex belongs to the Fennoscandian rapakivi magmatic system (Rämö and Haapala 2005) which spans a time interval of over 100 my (1.65-1.52 Ga; Rämö et al. 2014). Most likely the complex represents a volumetrically small but early (ca. 1.64 Ga; Heinonen et al. 2010b) series of compositionally diverse magma pulses related to the Wiborg magma system (Heinonen et al. 2016 and intruded and crystallized in a shallow (1-5 km) upper crustal magma chamber. Compositionally, the rock types of the complex represent the complete AMCG array observed in the Wiborg magma system and present a unique setting to study the magmatic emplacement and crystallization history of the entire rapakivi suite from its lower crustal origins all the way to the final emplacement at upper crustal levels.
Orthopyroxene megacrysts (HAOM) and associated megacrystic plagioclase entrapped within massif-type anorthosite cumulates have been interpreted to represent the earliest, deep stage of crystallization in anorthosite massifs (e.g., Heinonen et al. 2020 and references therein). It has been suggested that the megacrysts with the highest Al 2 O 3 contents crystallized at high pressures from mantle-derived high-Al tholeiitic melts, which are presumed parental to massif-type anorthosites (e.g., Emslie et al. 1994;Charlier et al. 2010). Their isotope compositions also suggest that crustal assimilation occurred before plagioclase accumulation . The accumulation of buoyant plagioclase enabled the ascent of the anorthositic crystal mushes through the crust leaving denser mafic cumulates behind (e.g., Bybee et al. 2019). Some HAOMs were carried to shallower crustal levels by the ascending mushes and record mineralogical evidence of low-pressure re-equilibration (e.g., exsolved plagioclase lamellae; Emslie 1975). The compositional zoning of HAOMS suggest that continuous crystallization took place during ascent and composition of the matrix orthopyroxene in the host anorthositic rocks indicates that final crystallization happened at upper crustal levels (Heinonen et al. 2020). The continuous crystallization of plagioclase and mafic phases from the mushes left behind a residual magma, usually considered to be monzodioritic in composition (e.g., Emslie et al. 1994;Markl and Frost 1999;Markl 2001;Heinonen et al. 2010b;Charlier et al. 2010). Observations of HAOMs and Al-in-opx geobarometry suggest similar polybaric crystallization history for the Ahvenisto anorthosites and for the entire Wiborg magma system (Heinonen et al. 2020).
Based on the evidence presented above and the polybaric crystallization hypothesis (Heinonen et al. 2020), we interpret that the monzodioritic rocks of the Ahvenisto complex represent such residual melts after fractional crystallization of anorthositic cumulates in a shallow (< 200

FeO tot
St ag e 1 S ta g e 2 S ta g e 1 St ag e 2 S ta g e 1 Stage 2 S t a g e 1 S ta g e 2 S ta ge 1 S ta g e 2 St ag e 1 S ta g e 2 S ta g e 1 Stage 2 St ag e 1 S ta g e 2 chamber, i.e. during the latest stages in the overall crystallization history of the complex. The monzodioritic residual melt moved towards more evolved compositions (lower Mg#:s) due to continuous fractionation of plagioclase (+ olivine + pyroxene + oxide), which formed the anorthositic cumulates. Batches of the monzodioritic residual melts were extracted from the fractionating system and crystallized in equilibrium producing the observed mineral assemblage characterized by low-Mg# mafic phases (Fig. 9). This interpretation lends further support to the previous models (Fred et al. 2019) in which interaction (mixing or mingling) between granitic and monzodioritic magmas observed in the Ahvenisto complex were suggested to have been controlled by the continuous extraction of compositionally evolving monzodioritic material to shallower crustal levels. During the ultimate shallow-level fractionation stages, accumulation of oxide, apatite, Fe-rich mafic phases, and An-poor plagioclase led to the formation of the apatite-oxide-gabbronorite.
The melts that the Ahvenisto complex rocks crystallized from had already gone through prior differentiation and assimilation, which is recorded in the isotopic signature of the whole complex (Heinonen et al. 2010b(Heinonen et al. , 2015. Based on the relatively shallow erosional level in Fennoscandia and seismic studies (Luosto et al. 1990;Elo and Korja 1993), a presumption that the main anorthositic mass most likely still remains unexposed at mid-crustal levels has been made (see Rämö and Haapala 1996). The rock types observed in Ahvenisto indicate that this mid-crustal mass may contain larger amounts of anorthositic cumulates and their Fe-Ti-Prich residual liquids crystallized at deeper levels. Consequently, although the Ahvenisto complex is only a small upper crustal level intrusion, it can be used to study the processes that have affected the formation of AMCG assemblages even in larger suites consisting of several intrusions, such as the Nain Plutonic Suite in Labrador, Canada.

Significance of monzodioritic rocks in AMCG petrogenesis-a global comparison
Major element compositions of monzodiorites and OAGNs globally are plotted in Fig. 11 to compare them with the data and models (MAGFRAC and MELTS FC, 100 MPa) of this study. The samples of all rock types show wide variation in composition and the global monzodioritic group overlaps with some of the OAGNs (Fig. 11). This overlap may reflect not only the differing classification of these Fe-Ti-P-rich rocks in various studies but also gradation from melt-representative samples to cumulus-dominated samples.

Monzodioritic rocks
Most of the monzodioritic rocks from other AMCG suites show similar major element characteristic to the Ahvenisto monzodioritic rocks and form a trend similar to what we suggest to represent the LLD of anorthosite residual melt evolution. However, the Ahvenisto olivine monzodiorites seem to define a compositionally unique group and almost no monzodioritic rocks with such high Mg-contents have been reported in any other AMCG suite. Although the general monzodioritic trend is similar, overall the Ahvenisto monzodioritic rocks have lower CaO and marginally lower Na 2 O and P 2 O 5 content than those from most other suites.
There seems to be two trends in terms of SiO 2 vs. MgO: some of the monzodioritic rocks plot on similar high MgO trend with the Ahvenisto olivine monzodiorites and others diverge towards the OAGNs (Fig. 11c). The former trend may represent a melt-evolution trend and the latter an accumulation trend. Similar, but less clearly distinguishable branching trends are present in the SiO 2 vs. TiO 2 and P 2 O 5 plots (Fig. 11a,g). In general, the monzodioritic rocks from other suites having similar Mg# with the Ahvenisto olivine monzodiorites have elevated Al 2 O 3 and CaO similar to Ahvenisto olivine-bearing anorthositic rocks (Fig. 11b,f), however, indicating they are more likely cumulates than melt compositions. The relatively lower Ca, Na, and P of the Ahvenisto rocks, compared to other monzodioritic rocks, indicates that there were differences in the contents of these elements in the parental melts (or in the possible assimilated material) between the different suites (see Duchesne 1990;, regardless of the general fractionation trends being similar. In addition, the mode of fractionated material, that depended not only on composition but also on the prevailing P-T conditions, may have had an effect on the residual liquid trends. Despite the apparent compositional differences between the localities, the LLD models constructed in this study indeed fit the observed global trends rather well (Fig. 11). With local adjustments of the input parameters, our models could be applicable also to other AMCG suites. Only the olivine monzodiorites are somewhat deviant from the general trends of monzodioritic rocks from other suites. However, this comparison is hampered by the general rarity of such primitive compositions and inclusion of cumulate rocks (rocks comparable to the Ahvenisto olivine-bearing anorthositic rocks and OAGNs, which will be discussed below) within the monzodioritic rock group elsewhere.  (Table 1) are shown. Gray arrows present the LLDs between estimated parent and daughter magmas of the MAGFRAC models in two stages (Table 2): the estimated parents plot close to the used parent compositions and the stages give r 2 values of 0.284 (Stage 1) and 0.375 (Stage 2). Only samples of this study are shown for clarity ◂

Oxide-apatite-gabbroborites
Although some of the OAGNs show similar major element characteristics with the monzodiorites, majority of the OAGN samples define scattered compositional arrays and show clearly distinct features, which are not compatible with residual liquid origins. Compared to the monzodioritic rocks, OAGNs have higher TiO 2 , FeO tot , MgO, CaO, and P 2 O 5 and lower SiO 2 , Al 2 O 3 , Na 2 O, and K 2 O due to strong accumulation of mafic minerals (pyroxenes ± olivine), oxide, and apatite (Figs. 4, 11).
Although the OAGNs show cumulus characteristics, they are very different from the anorthositic cumulates that have formed by the accumulation of plagioclase (Fig. 11). The composition of the OAGNs and their spatial relations suggest that they are also genetically associated with the monzodioritic rocks, however, and thus most likely still crystalized and accumulated from the same magmatic lineage as the anorthositic cumulates. Although the one sample from Ahvenisto suggests that it has formed during late stages in the evolution of the complex, the global data tentatively suggest that the OAGNs are formed throughout the crystallization history of anorthosites (see also Dymek and Owens 2001). Although OAGNs and monzodioritic rocks appear consanguineous, our modeling and observations suggest that the OAGNs, which are of cumulate origin, should not be grouped together with the melt-representative monzodioritic rocks.

Alternative modes of origin for the monzodioritic rocks in AMCG suites
We are not the first to propose a residual liquid origin for monzodioritic rocks in AMCG complexes. For example,  suggested that fractionation of olivinebearing anorthositic rocks would leave behind Fe-, apatite-, and oxide-enriched residual melts with OAGN affinities and the anorthositic fractionation without olivine would produce SiO 2 -enriched residuals with monzodioritic compositions. Although many of the studies on monzodioritic rocks in AMCG complexes argue for their residual liquid nature (e.g., Ashwal 1982;Morse 1982;Duchesne 1984;Owens and Dymek 1992;McLelland et al. 1994;Mitchell et al. 1996;Vander Auwera et al. 1998;Dymek and Owens 2001;Markl 2001;Heinonen et al. 2010b), some fundamentally different models for their origin have also been suggested. One of them suggest that the monzodioritic rocks would represent the actual parental magmas of the anorthositic rocks, like in studies of the Hidra massif of the Neoproterozoic Rogaland anorthosite province in SW Norway (e.g., Duchesne et al. 1974;Duchesne and Demaiffe 1978;Demaiffe and Hertogen 1981). Based on trace element fractional crystallization models, major element mass balance calculations, trapped liquid compositions, and lack of mafic veins, the charnokitic rocks associated with the anorthosites have, instead of monzodiorites, been inferred to represent residual liquids (Demaiffe and Hertogen 1981). One of the key arguments for the parental nature of monzodioritic rocks has been the lack of a negative Eu-anomaly that would had inevitably been produced by extensive plagioclase fractionation, especially in the reducing conditions characteristics of AMCG magmatism. However, this generalization is somewhat thwarted by the early saturation of apatite in the monzodioritic magmas, which drastically affects the bulk partition coefficients between minerals and melts and their ability to enrich REEs (Watson 1979). Fractionation of apatite has the opposite effect to Eu relative to that of plagioclase in reducing conditions and, therefore, may lead to smaller or lack of negative Eu-anomalies in the monzodioritic rocks as also observed in the monzodioritic rocks of this study.
A relatively early study on anorthositic suites suggested that monzodioritic rocks would represent intermediate derivatives of magmas parental to the entire AMCG series (De Waard and Romey 1969). Regardless of the suggested parental magma composition and source, many studies have since then described the bimodal (mafic-felsic) origin of AMCG rock types (e.g., Emslie et al. 1994;Ashwal and Bybee 2017 and reference therein). In addition, in order to produce the large amounts of more evolved compositions from the monzodioritic melts, very high degree of fractionation and considerably larger amounts of monzodioritic melts would be required than what are observed in most AMCG suites (e.g., Kolker et al. 1991). Especially, this mode of origin is not viable in the granite-dominated AMCG localities such as the Wiborg rapakivi suite.
On the other hand, the low Cr content and peculiar Srisotope composition of some monzodioritic rocks has led to suggestions that the monzodiorites are coeval but not comagmatic with the anorthosites (e.g., Emslie 1978;Duchesne et al. 1989;Duchesne 1990). The Sr content of some monzodioritic rocks have been interpreted as evidence of origin from basic crustal melts (e.g., Emslie 1978;Duchesne et al. 1989Duchesne et al. , 2017. Such low-degree melts would be enriched in Fe, Ti, and P and have lower SiO 2 compared to starting material, and their composition would be similar to those of the monzodiorites. Based on observed textures and experimental work an immiscible liquid origin for the monzodiorites have been suggested as well (e.g., Philpotts 1981;Powell et al. 1982). Crystallization of plagioclase from basaltic to andesitic magma leads to immiscible silica-rich and Fe-rich liquids (e.g., Philpotts 1981;Powell et al. 1982) and continued fractionation of plagioclase from these liquids would result in magneritic and monzodioritic composition, respectively. Further fractionation of the Fe-rich melt would lead to accumulation of oxides and apatite (Philpotts 1981).
The Nd-Sr-isotope compositions of the Ahvenisto complex anorthositic rocks (Ɛ Nd − 0.9 to − 0.5, Sr i 0.7037 to 0.7041) and monzodioritic rocks (Ɛ Nd − 1.1 to − 0.2, Sr i 0.7028 to 0.7040) overlap suggesting a comagmatic relationship at least in their case (Heinonen et al. 2010b). Furthermore, Hf-O isotopes suggest a mantle origin for these rocks, while the granitic rocks of the suite show a clearly separate crustal signature (Heinonen et al. 2010a(Heinonen et al. , 2015. Thus, the isotopic evidence rules out the possibility that the Ahvenisto monzodioritic rocks could be intermediate derivatives between anorthositic and granitic rocks or that they would be consanguineous with the granitic rocks in any other way. Immiscibility into two contrasting liquid compositions (i.e. monzodioritic vs. granitic lineages) prior to their late-stage independent fractionation is also unlikely based on the isotopic evidence of the Ahvenisto rocks.
We tentatively suggest, based on our observations, that the most suitable global mode of origin for the monzodioritic rocks is that they represent residual melt compositions after anorthosite fractionation and that with some modification the fractional crystallization models outlined in this study could be applied to other AMCG suites as well. The equilibrium crystallization stage revealed by a detailed mineral chemical analysis that followed the main fractionation stage in Ahvenisto requires further verification from other AMCG localities. The olivine monzodiorites are the most primitive melts in Ahvenisto, and although scarcely found in other AMCG suites, have potential to serve to help to trace the origin and ultimate parental melts of massif-type anorthosites. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, 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/.