High-aluminum orthopyroxene megacrysts (HAOM) in the Ahvenisto complex, SE Finland, and the polybaric crystallization of massif-type anorthosites

The occurrence of high-aluminum orthopyroxene megacrysts (HAOMs) in several massif-type Proterozoic anorthosite complexes has been used as evidence of their polybaric crystallization. Here, we report such petrographic and geochemical (XRF and EMPA) evidence from HAOMs discovered in the 1.64 Ga Ahvenisto rapakivi granite—massif-type anorthosite complex in southeastern Finland. Two different types of HAOMs were recognized: type 1 HAOMs are individual, euhedral-to-subhedral crystals, and up to 15 cm in diameter, and type 2 HAOMs occur in pegmatitic pockets closely associated with megacrystic (up to 30 cm long) plagioclase. The type 1 megacrysts in particular are surrounded by complex corona structures composed of plagioclase, low-Al orthopyroxene, iddingsite (after olivine), and sulfides. Orthopyroxene crystallization pressure estimates based on an Al-in-Opx geobarometer reveal a three-stage compositional evolution in both textural HAOM types. The Al content decreases significantly from the core regions of the HAOM (4.4–7.6 wt% Al2O3), through the rims (1.3–3.6 wt%), into the host rock (0.5–1.5 wt%). Enstatite compositions overlap, but are generally higher in the cores (En~60–70) and rims (En~50–70) of the HAOMs than in the host rock (En~45–60) orthopyroxenes. The highest recorded Al abundances in the HAOM cores correspond to crystallization pressures of up to ~ 1.1 GPa (~ 34 km depth), whereas the HAOM rims have crystallized at lower pressures (max. ~ 0.5 GPa, 20 km depth). The highest pressure estimates for the host rock orthopyroxene were ~ 0.2 GPa (< 7 km depth). These observations confirm the polybaric magmatic evolution of the Ahvenisto anorthosites and suggest that the entire 1.65–1.55 Ga Fennoscandian rapakivi suite was emplaced at a relatively shallow level (< 7 km depth) in the upper crust. Global comparison to similar rock types reveals remarkable similarities in the petrogenetic processes controlling HAOM composition and evolution of anorthosite parental magmas.

Several petrogenetic models imply a significant mantle contribution to the parental magmas of both the anorthosites and the associated ferroan granites. In the "tholeiitic model" (Frost and Frost 1997), anorthositic rocks are formed as cumulates from mantle-derived magmas, which intrude lower crustal levels and partially crystallize to form a mafic Communicated by Chris Ballhaus.

Electronic supplementary material
The online version of this article (https ://doi.org/10.1007/s0041 0-019-1648-5) contains supplementary material, which is available to authorized users. underplate. Continuing mantle heat production re-melts the underplate and generates partial melts, which in turn react with lower crustal material to produce the associated granitoids. The "all-crustal models" explain all rock types in AMCG suites as consanguineous fractionates of a common crustal parental melt of evolved jotunitic composition (e.g., Duchesne et al. 1999;Longhi et al. 1999). Balancing between these two extremes are the classical "two-source models" that distinguish a mantle origin for the anorthositic rocks and a dominant lower crustal source for the granitic rocks (e.g., Rämö 1991;Emslie et al. 1994). Isotope data also indicate that contamination of mantle-derived magmas by lower crust material is an important process in AMCG petrogenesis (e.g., Bybee et al. 2014;Heinonen et al. 2015).
High-aluminum orthopyroxene megacrysts (HAOMs) are a ubiquitous mineralogical feature of massif-type anorthosite complexes (e.g., Emslie et al. 1994;Owens and Dymek 1995;Charlier et al. 2010;Bybee et al. 2014). Their primitive geochemical character (high Mg# and juvenile isotope signature), high-aluminum content (up to 13 wt% Al 2 O 3 ), and plagioclase (± ilmenite, rutile, and garnet) exsolution lamellae due to low-pressure re-equilibration have been regarded as evidence of polybaric crystallization history of their parent magmas (e.g., Ashwal 1993;Emslie et al. 1994). Emslie et al. (1994) also introduced a widely adopted Alin-Opx geobarometer to assess crystallization pressures of HAOMs: Origin of HAOMs as a consequence of rapid low-pressure crystallization has also been suggested by some authors (e.g., Morse 1975;Dymek and Gromet 1984;Xue and Morse 1994;Owens and Dymek 1995). For example, the higher Fe/Mg of HAOMs compared to groundmass orthopyroxene of the Labrieville massif in Quebec, Canada, was used by Owens and Dymek (1995) to argue for in situ crystallization at emplacement level rather than in higher pressures. These contrasting views alongside observations outlined by Emslie (1975) suggest that there may be at least two different types of HAOMs: type 1, high-Al megacrysts that have a high-pressure origin, and type 2, low-Al megacrysts that crystallized at distinctly lower pressures in the upperto-middle crust.
In this study, we report new field and petrographic observations, mineral chemistry (bulk XRF and spot EMPA analyses), and Al-in-Opx geobarometry of high-aluminum orthopyroxene megacrysts found in the ~ 1.64 Ga Ahvenisto AMCG complex in southeastern Finland (Savolahti 1956(Savolahti , 1966Alviola et al. 1999;Heinonen 2012;Heinonen et al. 2010bHeinonen et al. , 2015. This is the first systematic survey of a relatively neglected, but important mineralogical feature of the Ahvenisto anorthosite and will enable comparisons (1) P(kbar) = −8.2 + 34.4 × Al IV + Al VI 0.5 .
to similar occurrences elsewhere. Our results provide new information on the intrusion and petrogenesis of the complex as well as broader constraints for the magmatic evolution of the Fennoscandian rapakivi suite and AMCG complexes in general.
The suite is geochemically bimodal (g ranitic-anorthositic) with only minor intermediate rock types (monzodiorites and quartz monzodiorites; e.g., Alviola et al. 1999;Fred et al. 2019). Compared to other similar occurrences, such as the Grenville Province of Canada (e.g., Ashwal 1993), anorthositic rocks are fairly uncommon at the current level of exposure (Heinonen et al. 2010b). This is most likely due to differences in erosional depth, as deeper levels of AMCG complexes seem to expose nearly equal proportions of granitic and anorthositic rocks (Emslie et al. 1994). A large anorthositic intrusion is inferred from geophysical data to underlie the Wiborg batholith, which is the largest rapakivi intrusion in southern Finland (Elo and Korja 1993;Rämö and Haapala 2005).
A persistent problem in the general petrogenetic framework of the entire Wiborg rapakivi batholith and the other rapakivi intrusions related to it, including Ahvenisto, is the depth of emplacement (e.g., Heinonen et al. 2016Heinonen et al. , 2017. Studies based on mineral chemistry and other mineralogical evidence (Eklund and Shebanov 1999;Elliott 2001) have suggested that the main rock types of the Wiborg batholith at the current erosional level crystallized at high pressures (up to 500 MPa) and considerable depths of approximately 15-20 km. However, Heinonen et al. (2017) questioned the validity of the Al-in-Hbl geobarometer results used to estimate the crystallization depths of rapakivi granites due to their well-documented calibration issues in Fe-rich ferroan magma systems (e.g., Anderson and Smith 1995). Also, the reported close association of sub-volcanic, volcanic, and coeval sedimentary sequences with the Wiborg rapakivi granites would place them and associated rocks in a somewhat shallower crystallization environment (e.g., Pokki et al. 2013;Kohonen and Rämö 2005;Kohonen 2016). This discrepancy between the results of geobarometric methods and field observations has persisted in part due to the lack of independent geobarometric studies. Al-in-Opx geobarometry conducted in this study seeks to remedy this situation.
Equigranular rapakivi granites cover ~ 75% of the mapped area of the Ahvenisto complex. They include the main batholith granites, which range from hornblende-biotite granite to more evolved biotite and topaz granites, and a more primitive hornblende granite that occurs as separate intrusions on the eastern flank of the complex (Alviola et al. 1999;Fred et al. 2019). Anorthositic rocks (~ 20% of mapped area) of the complex are mainly leucogabbronorites. In the northeastern part of the complex, some olivine-bearing gabbroic rocks are also known. Anorthosite (plag > 90 vol.%) is rare and occurs only as erratic rafts in the leucogabbronoritic rocks. Monzodioritic rocks (~ 5% of mapped area) occur as dikes flanking the gabbroic arc mainly on the eastern margin of the complex (Alviola et al. 1999;Heinonen et al. 2010b;Fred et al. 2019). Emplacement and crystallization of the complex took place during a relatively short interval from 1644 to 1629 Ma (Alviola et al. 1999;Heinonen et al. 2010b). Nd, Hf, and O isotope compositions suggest that the anorthositic and monzodioritic rocks have a common, possibly MORB-type depleted mantle progenitor whereas the rapakivi granites have a separate, crustal origin (Heinonen et al. 2010a(Heinonen et al. , b, 2015.

HAOMs in the Ahvenisto complex
Thus far, only a few HAOM localities have been known from the northwestern part of the Ahvenisto complex. Savolahti (1956Savolahti ( , 1966 gave detailed but isolated descriptions of "pegmatoid gabbros" consisting chiefly of large (1-20 cm), euhedral plagioclase megacrysts (~ An 53 ) and mostly anhedral interstitial orthopyroxene crystals of similar size, which even envelope plagioclase crystals in places. Savolahti (1956) also reported large individual orthopyroxene and plagioclase phenocrysts embedded in finer grained groundmass alongside outcrops of pegmatoidal anorthosite, but was unable to trace the contacts of these rock types due to poor exposure. These same localities were also recorded during a detailed mapping effort that was conducted in the area of the Ahvenisto complex by the Geological Survey of Finland (GTK) during the late 1970s and early 1980s (Johanson 1984;Alviola et al. 1999).
In this study, we report 11 new HAOM outcrops in the northwestern part of the complex (Fig. 1c). The new observations reveal two different HAOM types similar in habit to the ones reported by Savolahti (1956) and, more importantly, to the high-Al type 1 and low-Al type 2 orthopyroxene megacrysts reported by Emslie (1975) from the Nain Complex in Canada. The Ahvenisto type 1 megacrysts occur as individual crystals or in groups in an equigranular anorthositic host rock. They are euhedral-to-subhedral and commonly up to 15 cm in diameter (Fig. 2b). Plagioclase megacrysts may or may not accompany the type 1 pyroxenes, but are never intergrown with them. Type 2 megacrysts occur in very coarsegrained "pockets" or irregular "rafts" always intergrown with megacrystic (up to ~ 30 cm long) plagioclase laths (Fig. 2e). There are also reports of pegmatitic pockets that host megacrystic pyroxene and plagioclase alongside large crystals of apatite, zircon, and amphibole in the Ahvenisto complex (Savolahti 1956;Alviola et al. 1999). These pockets are in some respects similar to the type 2 HAOMs reported in this study, but most likely represent late-stage residual liquids rather than early crystallized deep fractionates in the anorthositic magma system.

Methods and materials
The main rock types of the study area and a set of highaluminum orthopyroxene megacrysts and associated plagioclase megacrysts were sampled for basic petrography, major-element geochemistry by WD-XRF (wavelengthdispersive X-ray fluorescence), trace-element geochemistry by ICP-AES (inductively coupled plasma atomic emission spectrometry), and quantitative major-element mineral analysis by WDS-EMPA (wavelength-dispersive electron microprobe analysis). A total of 12 bulk samples (four whole-rock samples, six orthopyroxene megacrysts, and two plagioclase megacrysts) were analyzed for major-and trace-element compositions. Nine samples were chosen for WDS-EMPA mineral analysis. A basic petrographic study was conducted on all 21 samples. All samples were collected with a hammer, except for HK-I, -II, and -V, which were sampled with a rock saw. Sampling localities are shown in Fig. 1c and listed in Table 1 alongside with the basic information and analysis history. Sample preparation for all methods was carried out at the Mineralogical Laboratory of the University of Helsinki.

WD-XRF and ICP-AES
All bulk samples were crushed into 1-3 cm pieces with a jaw crusher (possible contamination of Fe, Cr, Ni, and V). In addition to fresh sample selection all megacryst sample pieces (plagioclase and orthopyroxene) were further carefully hand-picked clean, so that no other phases would be included in the analysis. Possible minute inclusions (e.g., plagioclase, ilmenite, or amphibole; Heinonen et al. 2014) would be involved in the analysis, because they could not be reliably removed by this method. The picked fractions were pulverized in a Fritsch Pullverisette 6 mill, in a tungsten carbide pan (running time 15 min, 350 rounds per minute). Contamination from the mill might include W, Co, and Ta. WD-XRF and ICP-AES analyses were carried out at the Central Mineral and Environmental Resources Science Center in Denver, Colorado (offices of the USGS). Analytical details are summarized in Taggart and Siems (2002).

Wds-empa
Major-element compositions of orthopyroxene, plagioclase, and olivine were determined at the Mineralogical Laboratory of the University of Helsinki with a JEOL JXA-8600 Electron Probe Microanalyzer (EMPA) equipped with four wavelength-dispersive (WDS) spectrometers, upgraded with SAMx analytical software and PointElectronic SAMx hardware. The analysis was performed on carbon-coated thin sections. Eleven elements (Si, Al, Ti, Cr, Fe, Mn, Mg, Ni, Ca, Na, and K) were analyzed from the orthopyroxenes, twelve (Si, Al, Ti, Cr, Fe, Mn, Mg, Ca, Na, Sr, Ba, and K) from plagioclase, and nine (Mg, Al, Si, Ca, Ti, Cr, Mn, Fe, and Ni) from olivine. Matrix corrections were performed with the SAMx analysis software using the PAP correction. Potential Si K β overlap over Sr K α has been calculated by the software after measuring the Si K β intensity at the Sr K α peak position. The microprobe was calibrated with a combination of natural and synthetic oxide and silicate standards. The acceleration voltage, sample current, and beam diameter for the analyses were 15 kV, 15 nA, and ca. 10 μm, respectively. More analytical information is reported in the Online resource 1.

Structural occurrence of HAOMs
The known HAOM occurrences in Ahvenisto are restricted to a relatively small area within the northwestern part of the gabbroic arc (Fig. 1b, c), which combined with rather poor exposure, preclude significant petrological conclusions about their structural position. The observations, however, seem to follow a pattern in which type 1 HAOMs are more abundant near to the contact with the Svecofennian country rock and type 2 HAOMs are more numerous close to the contact with the younger granite. Both types are observed on two outcrops that are structurally situated between the type 1 dominated outer and type 2 dominated inner edge. There is no significant correlation observed between the host rock composition and HAOM type.

Leucogabbronorite, HAOM host rock
The host rock of the HAOMs is the main rock type in the study area and consists of heterogeneous mesocumulatic leucogabbronorite. Subhedral-to-euhedral and strongly twinned plagioclase 5-15 mm in size (locally porphyritic, 2-15 cm) is the dominant mineral (ca. 80-90%). Some signs of local deformation (bent plagioclase crystals) and pressure solution (secondary quartz) are observed. Intercumulus minerals include orthopyroxene, clinopyroxene, olivine (mostly altered), biotite, minor apatite, and opaque minerals. Orthopyroxene is the main mafic phase, but, on some outcrops, signs of altered interstitial olivine grains commonly rimmed by a thin orthopyroxene corona or biotite are observed. Clinopyroxene is observed both as exsolution lamellae in orthopyroxene and as independent grains. Apatite occurs mainly as small inclusions in biotite. The texture of the rock varies significantly from clearly ophitic to a more irregular plagioclase-porphyritic type (Fig. 2a). This change is local, can be very abrupt, and the contacts are gradual between the different types.
Post-crystallization alteration of the leucogabbronorite is commonly observed. It mostly affects the interstitial mafic phases, but is also seen in the cracks observed in the larger plagioclase crystals. Chloritization is common and Fe-rich chlorite is a typical alteration product after interstitial pyroxenes. Where alteration of the host has been most intensive, plagioclase alteration (to epidote, saussurite, and sericite) is also common. On a few outcrops, irregular rafts of coarsegrained plagioclase and orthopyroxene are observed within the host leucogabbronorite. These rafts vary considerably in size (10-50 cm up to several meters in width) and shape, but commonly have rounded edges (Fig. 2a).

Hornblende-biotite granite
The other main rock type in the study area is a homogeneous, equigranular hornblende-biotite granite that belongs to the main part of the Ahvenisto batholith ( Fig. 1; Alviola et al. 1999). The granite is pale pink in color, and consists mainly (~ 60%) of small (1-3 cm), euhedral-to-subhedral, perthitic alkali feldspar crystals (Fig. 2d), quartz, and biotite with minor plagioclase and hornblende. The main accessory minerals are fluorite and apatite. Zircon (± allanite) is observed as inclusions in biotite. The granite has brecciated the main leucogabbronorite in several locations. The contacts are sharp and fine-grained chilled margins are observed, especially in the thinner granite veins. It should be noted that although no rapakivi texture (alkali feldspar ovoids with plagioclase rims) is observed, on geochemical and geochronological grounds, the Ahvenisto granite is associated with the rapakivi magmatism of southeastern Finland (Alviola et al. 1999;Heinonen et al. 2010b).

Type 1 HAOM
The type 1 orthopyroxene megacrysts are individual grains, commonly anhedral-to-subhedral, and up to 15 cm in size (Fig. 2b). The orthopyroxene in the rim of the type 1 HAOMs has a slightly different optic orientation, which may indicate compositional difference or crystallization under different conditions. Some of the type 1 megacrysts are weakly pleochroic, which might be due to the large amount of dust-like Fe-Ti oxide inclusions. Most of the studied type 1 HAOMs are severely cracked. The cracks host alteration products such as amphibole, carbonate, and chlorite among larger grains of opaque minerals. Some of the larger cracks contain small grains of re-crystallized plagioclase or even altered olivine.
Preliminary inspection of back-scattered electron (BSE) images revealed lines of ~ 5 µm inclusions in the studied HAOMs ( Fig. 3e; Heinonen et al. 2014) oriented most likely in the orthopyroxene [210] planes. As macroscopic plagioclase exsolution lamellae have not been observed in the Ahvenisto HAOMs, these inclusions were first suspected to represent the onset of plagioclase exsolution. However, X-ray diffraction (XRD) results  suggest that they consist of secondary amphibole. The type 1 HAOMs are commonly surrounded by a complex rim structure ( Fig. 2c) up to 5 mm wide and comprised mainly of plagioclase, low-Al orthopyroxene, iddingsite (after olivine), and sulfides. Petrography of these coronas is described in more detail below.

Type 2 HAOM
The type 2 HAOMs are large (up to 30 cm in length), mostly anhedral orthopyroxene crystals intergrown with megacrystic plagioclase of similar grain size (Fig. 2e). Most of the studied crystals are slightly pleochroic from pale red to light green. Type 2 HAOMs are commonly broken and cracked in several places. The cracks host secondary alteration products such as amphibole. The megacrystic plagioclase associated with type 2 HAOMs is homogeneous and strong polysynthetic twinning is observed. The plagioclase crystals are commonly cracked and altered along the cracks.

Corona structures
The corona rims surrounding type 1 HAOMs are divided into three zones moving outwards from the HAOM towards the host rock: (1) the plagioclase inclusion zone, (2) the olivine zone, and (3) the sulfide zone overlapping with the outer part of the olivine zone (Figs. 2c and 3a). These zones occur in various stages of development where locally any one of the zones may be incomplete or even completely missing.
The plagioclase zone is typically 1-3 mm wide, consists of low-Al (Al 2 O 3 = 0.9-3.6 wt%) orthopyroxene, and hosts abundant plagioclase inclusions (with minute higher density phases such as ilmenite). This zone is present in all studied samples. The olivine zone covers typically most of the rim area, the main phases being yellowish-brown iddingsite (after olivine, which in some samples is partly still fresh) and low-Al orthopyroxene. Vermicular intergrowths of orthopyroxene and plagioclase are observed where olivine is in contact with plagioclase within the rim (Fig. 3c, d). Altered olivine is surrounded by a thin layer of orthopyroxene (and sometimes an additional layer of biotite), which is followed by fan-like symplectic structures that always grow outward into the plagioclase. The sulfide zone forms a relatively narrow band on the outer edge of the corona rim and hosts pyrite as well as ilmenite. A small amount of biotite is also present.
Many of the type 2 HAOMs display a corona as well, but it is not as well developed as in type 1 (Fig. 2f). Unlike in type 1 HAOMs, where the reaction rim reaches out towards the host rock in the form of symplectite, in type 2, the rim is mostly composed of low-Al orthopyroxene and plagioclase and restricted to a width of ca. 1-5 mm. Commonly, only a thin layer of dark mica separates the HAOM from plagioclase megacrysts in the host. Olivine strongly altered to iddingsite or serpentine does occur locally as small crystals embedded in the type 2 HAOM rim. In most cases, olivine is in close contact with plagioclase, but separated by a very thin orthopyroxene film or a small-scale symplectite. Tables 2 and 3 show the geochemical compositions for leucogabbronorite whole rock (4), orthopyroxene (6), and plagioclase (2) bulk samples. Selected major-and trace-element compositions and the chondrite-normalized REE compositions are illustrated in Fig. 4.

Bulk sample geochemistry
The leucogabbronorite whole-rock samples show a rather constant major-element composition (SiO 2 = 52.1-52.7 wt%; Mg# = 41.9-46.5). An exception is observed for sample HEKI-505.1 that has somewhat lower TiO 2 (0.44 wt%), FeO tot (3.51 wt%), and MgO (0.96 wt%) abundances but higher total H 2 O (2.0 wt%) content. Petrographic observations suggest that these differences are due to later alteration in sample HEKI-505.1 whereas the other samples with uniform composition represent the original HAOM host rock composition. In general, the fresh whole-rock compositions are comparable to the ones reported for similar rock types in previous studies of the Ahvenisto complex (Alviola et al. 1999;Heinonen et al. 2010a).

Orthopyroxene
Representative orthopyroxene major-element analyses from different HAOM types and textural positions are shown in Table 4 (full analytical results for all minerals are reported in Online Resource 1). Enstatite compositions for orthopyroxene analysis from different textural positions overlap, but are generally higher in the cores (En~6 0-70 ) and rims (En ~ 50-70 ) of the HAOMs than in the host rock groundmass orthopyroxene (En~4 5-60 ). The orthopyroxene aluminum content (Fig. 5a) decreases significantly from the core regions (4.4-7.6 wt% Al 2 O 3 ), through the rims (1.3-3.6 wt%), into the host rock (0.5-1.5 wt%). There is a clear negative correlation between Mg# and MnO ( Fig. 5b) with HAOM core analyses at lower MnO values and progressively higher values in the HAOM rims and highest in the host rock orthopyroxene. The HAOM cores have clearly higher Cr 2 O 3 values than most of the HAOM rims and all of the host rock orthopyroxene, which fall under the limit of detection (ca. 0.1 wt% Cr 2 O 3 , values not plotted in Fig. 5c). In general, the positive correlation of Mg# with Cr 2 O 3 is similar compared to Al 2 O 3 (Fig. 5c) but hampered by the low analytical precision at lower absolute Cr 2 O 3 values. XRF analyses of    Major elements were analyzed with WD-XRF and trace elements with ICP-AES <Analysis was below detection limit indicated by the numeral a Calculated based on 6 oxygen atoms per formula unit b Calculated based on 8 oxygen atoms per formula unit Page 13 of 25 10 all six bulk samples overlap with the HAOM core EMPA analyses for Mg#, Al 2 O 3 , MnO, and Cr 2 O 3 , but are slightly higher for TiO 2 , CaO, and Na 2 O, none of which define any significant correlation with textural analysis type (Fig. 5d-f). These small differences may result from minute inclusions of ilmenite and plagioclase undetected during the sample preparation process (see Analytical methods and material) for XRF analysis.
Plagioclase Table 5 shows results of representative analyses from different textural plagioclase types. Plagioclase compositions were measured from inclusions mostly within the HAOM rims (I for inclusion) and in the host rock groundmass from euhedral crystals (H for host) and euhedral megacrysts (10-30 cm in length; M for megacryst). The plagioclase composition in one orthopyroxene-plagioclase symplectite (S) was also analyzed (HEKI-508.1-S-01; An 65 ). Overall, the mean (± 1SD) composition of plagioclase in all samples is rather homogeneous (An 51±4.7 ), and on average, the megacrysts (An 52.1±3.0 ) seem to have slightly more calcic composition than groundmass (An 49.3±3.6 ) plagioclase (Fig. 6). One plagioclase inclusion within a type 2 HAOM (HEKI-506.1B-E-01) yielded an anomalously high anorthite content of An 71 . No other significant compositional differences were observed in the plagioclase analyses.

Olivine
Olivine is a very minor phase in the studied rocks and is mostly concentrated on the rims of the HAOMs. Most of the grains have altered into iddingsite (or, in rare cases, serpentine), but also some fresh olivine grains remain. Table 6 shows all conducted olivine analyses. The two analyses reported for sample APHE-01.1 represent olivine alteration products (mostly iddingsite), have very low totals, and should only be considered as a baseline for screening out alteration in other analyses. The two analyses reported for sample HEKI-506.1B (Fo 56 and Fo 55 ) come from a single olivine crystal in contact with a type 2 HAOM. Two different types of olivine crystals with slightly different compositions were observed in sample HEKI-508.4: an independent larger crystal (Fo 57-58 ; spots 01-03; Table 6) and several smaller crystals in contact with the larger olivine and smaller orthopyroxene grains (Fo 52-54 ; spots 05-08; Table 6). Analyses from sample HK-I are from olivine grains embedded in the rim of a type 1 HAOM and record a very homogeneous composition of Fo 50±0.2 (± 1SD; n = 6).

Crystallization pressure estimates and Al-in-opx barometry
The higher MgO contents of the Ahvenisto HAOM (20.22-23.99 wt% MgO) compared to the host rock orthopyroxene (14.35-23.40 wt%), general positive correlation of Cr 2 O 3 with Al 2 O 3 , and the mineralogical evidence support the polybaric crystallization hypothesis for Ahvenisto anorthosite and warrant a closer estimation of crystallization pressures. Crystallization pressures of the orthopyroxene megacrysts and the host rock orthopyroxene of the Ahvenisto complex were evaluated, using the Al-in-Opx geobarometer of Emslie et al. (1994; Formula 1). Mg# vs. Al 2 O 3 contents and calculated crystallization depths of all studied orthopyroxene samples are illustrated in Fig. 7 relative to their textural types and analysis spot positions. Both bulk and spot analyses of the type 1 and type 2 HAOM cores yield a population of overlapping high Al 2 O 3 abundances around 7 wt% with Mg# ~ 70. This population of highest recorded Al abundances correspond to crystallization pressures of up to 1.14 GPa and depths of ca. 38 km. In general, the core values vary from 0.67 GPa to 1.14 GPa, but one type 1 sample (HEKI-508.1) shows clearly lower pressure estimates from only 0.05 GPa to 0.29 GPa. These lower Al contents and pressure estimates are most likely due to the severe cracking and alteration observed in this sample. The HAOM rims crystallized at lower pressures (maximum of 0.53 GPa or ca. 20 km for type 2 sample HEKI-506.1) and values within the barometer calibration [estimated minimum threshold (Al IV + Al VI ) 0.5 ca. 0.3-0.4 or absolute Al IV + Al VI ca. 0.09] range from 0.06 GPa to 0.53 GPa. The pressure estimate from an orthopyroxene-plagioclase symplectite gives an estimate of 0.12 GPa, which matches well with the late stages of rim crystallization. This is to be expected, as the symplectites would most likely not be so well developed or preserved if they had undergone significant pressure changes. Most of the host rock orthopyroxene gives pressure estimates below the barometer calibration, yet the maximum value observed is 0.19 GPa (< 7 km).
The entire data set, therefore, suggests that anorthosite crystallization took place during three distinct stages: (1) at high pressures most likely at around 35 km depth (HAOM cores), (2) progressively during ascent of the anorthositic mushes to the upper crustal levels at less than 20 km depth (HAOM rims), and (3) at emplacement level at less than 10 km depth (groundmass opx).
Comparisons of Ahvenisto type 1 and type 2 HAOMs further reveal that, contrary to the observations made by Emslie (1975) in Eastern Canada, both types most likely share a common high-pressure origin. We suggest that all    HAOMs observed in Ahvenisto, originally started to crystallize at the crust-mantle interface, were consequently trapped by the anorthositic mushes, and then transported to upper crustal levels, while they continued to crystallize along with megacrystic plagioclase. The coarse-grained orthopyroxene and plagioclase rafts, thus represent parts of an earlier cumulate, which was broken into smaller pieces before it settled at emplacement depth as the remaining magma crystallized. The apparently type 1 HAOMs, therefore, would represent completely broken down parts of these original cumulates and type 2 HAOMs still intact crystal aggregates.

Petrographic and mineralogical features of the Ahvenisto HAOMs
In the general interpretation of massif-type anorthosite petrogenesis, HAOMs most likely represent a small proportion of the mafic minerals that crystallized from the anorthosite parental magmas at the crust-mantle interface (e.g., Ashwal and Bybee 2017). They were fortuitously trapped by high-viscosity anorthositic magma mushes and transported to upper crustal levels (cf., Bybee et al. 2014). However, based on a number of observations from several massif-type anorthosite localities (Morin, Harp Lake, Lac St. Jean, Egersund-Ogna, Ana-Sira, Haaland-Helleren), Emslie (1975) defined two types of HAOMs with distinct modes of occurrence, chemical compositions, and suggested origins. The type 1 megacrysts (2-50 cm) of Emslie (1975) are orthopyroxenes (rarely Cpx) of a deep lustrous, bronze color and form sub-rounded to irregular masses in anorthositic host rocks. They include regular (100)-parallel lamellae of plagioclase and their Al 2 O 3 content is relatively high (up to 5.96 wt%). The type 2 orthopyroxene (or Cpx) crystals occur in pod-like or irregular pegmatoidal masses within anorthositic rocks (e.g., Bybee et al. 2014  n/a n/a n/a n/a Depth(km) 2.0 0.6 3.2 n/a 4.5 n/a 1.0 0.3 n/a n/a n/a n/a of plagioclase, have relatively low Al 2 O 3 content (1.48-2.78 wt%) and higher Fe/Mg than type 1 crystals. Based on these observations, Emslie (1975) suggested the type 2 HAOMs crystallized at lower pressures at or close to the emplacement levels of the anorthositic mushes and that only type 1 HAOMs had a high-pressure origin.
High-aluminum orthopyroxene megacryst types texturally similar to those described by Emslie (1975) have also been reported from Ahvenisto (Savolahti 1956; this study). The most significant differences to the types of Emslie (1975) are, however, the similar Al 2 O 3 contents, complete lack of plagioclase lamellae, and conspicuous corona structures observed in both textural types of the Ahvenisto HAOMs.  Table 2) are plotted as well for comparison     Also, all Ahvenisto samples, regardless of their textural type, display homogeneous major-and trace-element HAOM compositions for both the bulk sample and EMPA core region analyses. These petrographic and compositional observations suggest that the Ahvenisto type 1 and type 2 HAOMs share a common origin, parental magma, and P-T history, and differ essentially only in their habit.
In contrast, plagioclase megacrysts in the host rocks associated with the HAOMs have slightly higher An composition than the groundmass plagioclase and, therefore, most likely represent parts that have crystallized at an earlier stage in the magmatic evolution. Furthermore, the composition of both individual plagioclase megacrysts and the plagioclase intergrown with type 2 HAOMs seem to be similar, suggesting that they share a common origin.
A key mineralogical feature of HAOMs described from nearly all massif-type anorthosite occurrences are the thin (some tens of up to hundreds of µm) plagioclase lamellae (e.g., Ashwal 1993;Bybee et al. 2014) oriented in the (100) direction. The composition of the plagioclase in the lamellae is commonly reported to be somewhat more calcic than that of the host anorthosite (Emslie 1975) or even inversely zoned (Owens and Dymek 1995). The host orthopyroxene is typically depleted in Al and Si but has unchanged Fe and Mg adjacent to the plagioclase lamellae (Emslie 1975). Based on these observations, the lamellae have generally been interpreted to have exsolved from the HAOMs as a re-equilibration response to lowering external pressure in their magmatic environment (Ashwal 1993;Emslie et al. 1994). Furthermore, Emslie (1975) extrapolated the pressure-dependent jadeite (NaAlSi 2 O 6 ) and Ca-Tschermak's molecule (CaAl 2 SiO 6 ) solubility observed in clinopyroxene (Kushiro 1969) to orthopyroxene, and suggested the observed inverse zoning to be caused by the decreasing solubility of NaAlSi 2 O 6 over CaAl 2 SiO 6 in orthopyroxene as a function of pressure.
The lack of plagioclase exsolution lamellae in the Ahvenisto HAOMs is curious and may be related to the decompression history of the mushes. A potential explanation may be found from the shallower emplacement level (< 7 km) of the Ahvenisto complex compared to most other massif-type anorthosite occurrences (Heinonen 2012). Shallower emplacement level in the crust translates to higher temperature differences between the intruding mushes and their surroundings, and potentially faster cooling rates. Faster cooling may inhibit diffusion required for the exsolution process to take place and in turn prevent the generation of plagioclase lamellae altogether. The more interesting mineralogical feature present in the Ahvenisto HAOMs, however, is the ubiquitous multi-stage corona structures. We suggest that the presence of these coronas and the lack of plagioclase exsolution lamellae represent a previously unreported decompression response caused by the relatively shallower emplacement level of the Ahvenisto complex relative to the other studied AMCG occurrences.

Stability of olivine and late low-pressure reactions
Corona rims of orthopyroxene and clinopyroxene + spinel on early crystallized olivine are relatively common and interpreted in several studies to reflect reaction of olivine with a magmatic liquid (e.g., Turner and Stüve 1992;Best 2003, p. 482). Olivine rims surrounding orthopyroxene are not nearly as common and most likely require fortuitous crystallization conditions in very specific magma compositions.
The mutual stabilities of olivine, plagioclase, and orthopyroxene as presented on the An-Fo-Qz join ( Fig. 8; Sen and Presnall 1984) are critically dependent on pressure. The stability fields are dictated by the passing of the isobaric quaternary Fo-En-Spl-An-Liq invariant point through the join at approximately 500 MPa from a tholeiitic system in    The geometry of this transition and its consequences on the expansion of anorthite and olivine stability fields at lower pressures in relation to compositions of monzodioritic rocks from the Ahvenisto complex (Fred et al. 2019) are illustrated in Fig. 8. The monzodioritic rocks are inferred to represent late residual liquid compositions after anorthositic fractionation (Heinonen et al. 2010a) and the majority of them plot very close to the passing point of the isobaric quaternary invariant on the join. Most of the monzodiorite compositions plot in the enstatite stability field in high pressures and near to the An-Fo curve and the An-Fo-En piercing point. This suggests that Ahvenisto magmas that crystallized (high-aluminum) orthopyroxene in higher pressures shifted to co-crystallization of olivine, plagioclase, and (low-aluminum) orthopyroxene as a response to decompression, thus creating the Ol + Pl + Opx corona rims on HAOMs.
In some cases, the olivine grains also have an orthopyroxene rim that appears to have invaded the margins of the olivine grains, which would also indicate that olivine crystallized at least partly prior to the rim orthopyroxene. This could be interpreted as a consequence of the crystallization sequence: (high-Al) Opx; decompression → Pl + Ol → Pl + Ol + (low-Al) Opx. The outer parts of the coronas in some cases include a second, hydrous, shell of biotite, which can further be mantled by a wider zone of a twophase symplectite (Fig. 3a). This vermicular shell clearly impinges on or grows into the plagioclase grains of the surrounding host rock. It consists of plagioclase containing elongate vermicules of orthopyroxene, which generally point outwards from the olivine (Fig. 3c, d). Such a transition from coronitic to vermicular texture could be interpreted as an indicator of the distance beyond which chemical diffusion became too slow for broad reaction front migration (e.g., Turner and Stüve 1992). The growth of the symplectite into plagioclase grains could also be taken as evidence of a subsolidus origin for the symplectitic part of the texture. The incomplete biotite coronas may have also been caused by a subsequent hydration of the system, which is supported by Fig. 7 Mg# v. Al 2 O 3 (wt%, left axis) and depth of crystallization (km, right axis) diagram of all (EMPA and XRF) orthopyroxene analyses reported in this study relative to textural type and analysis spot location. Numbers of samples and analysis populations discussed in the text are also shown Fig. 8 Liquidus phase relationships in the Di-projected Fo-An-Qz join of the CMAS system at 1 GPa and 1 atm pressures with compositions of monzodioritic rocks from the Ahvenisto complex (Fred et al. 2019). Compositions are plotted as recalculated phase (Pl-Ol-Qz) weight-% and stability fields are reproduced after Sen and Presnall (1984) the late, locally almost complete, alteration of the host rock leucogabbronorite.
Most of the HAOM rim orthopyroxene have lower Al 2 O 3 abundances than the corresponding HAOM cores indicating maximum crystallization pressures of ca. 0.53 GPa (discussed above) for the rim. This corresponds well with the suggested shift in liquidus phase relations in response to lowering crystallization pressure during magma evolution. The symplectite and biotite shells crystallized later and may have a late magmatic or post-magmatic origin. The limited melt contact caused by the shielding by adjacent plagioclase most likely inhibited the development of complete reaction rims around type 2 HAOMs.

Crystallization depth of the Wiborg magma system
Orthopyroxene crystallization pressures derived from the Al-in-Opx geobarometer provide evidence to constrain the approximate intrusion depth of the Ahvenisto anorthosite and, by extension, also the Wiborg batholith. The presentday upper crustal architecture in southeastern Finland estimated from geophysical data (Elo and Korja 1993;Rämö and Haapala 2005) is illustrated in the two suggested cases in Fig. 9. The present-day thickness of the Wiborg batholith is approximately 5-8 km and it is underlain by a high-velocity zone (HVZ) of similar thickness that has been interpreted to represent a slab of anorthositic rock material coeval with the rapakivi granites (Elo and Korja 1993). The scenario A of deep crystallization derives chiefly from the Al-in-Hbl geobarometry results reported by Elliott (2001) inferring an average crystallization depth of ca. 15 km for the Wiborg rapakivi granites exposed at the present-day surface. The shallow crystallization scenario B inferred from field observations is further supported by the new Al-in-Opx results obtained in this study. The deep HAOM core crystallization at ca. 35 km depth was followed by ascent of the crystal mushes to 15-20 km depths, which coincides with the suggested emplacement depth of the anorthositic (HVZ) slab. The low-Al groundmass orthopyroxene in the Ahvenisto complex has crystallized at significantly lower pressures than the HAOMs and, therefore, also at relatively shallow levels (< 7 km depth) in the upper crust, favoring scenario B for the emplacement of the Fennoscandian rapakivi suite.

Global comparison of HAOM compositions
The Ahvenisto HAOM compositions are compared to results reported from other massif-type occurrences around the world in Fig. 10. Analysis of bulk samples has been the method of choice in most comprehensive HAOM studies (e.g., Charlier et al. 2010;Bybee et al. 2014), but a number of papers also report the results of spot EMPA analyses. Also Emslie (1975) in his pioneering study discussed the importance of sample representativeness and relevance of sampling method in HAOM research. The major concern about the representativeness of spot analysis of HAOMs relates to the exsolved plagioclase lamellae and the Aldepleting effect of the exsolution process to the composition of the orthopyroxene directly adjacent to the lamellae. Therefore, spot analyses are prone to underestimate the original Al content of the HAOMs and may lead to systematically low estimates of crystallization pressures. Thus, HAOM samples with abundant plagioclase lamellae should preferably be analyzed by bulk methods to obtain a better estimate of the original composition before the exsolution took place. In the absence of exsolved plagioclase, as is the case in Ahvenisto, spot analysis naturally offers a better spatial resolution and tools to evaluate the pressure evolution during HAOM crystallization. These tools have already been utilized above to argue for the three stages of polybaric crystallization of Ahvenisto magmas, but the effects of plagioclase exsolution in the global comparison have to be taken into account as the results of bulk and spot analyses are compared to each other. Emslie (1975) evaluated the exsolution effect on HAOM composition by analyzing pristine orthopyroxene and portions of HAOMs immediately adjacent to the plagioclase lamellae by EMPA. Based on Emslie (1975), a set of exsolution vectors from a composition of plagioclase lamellae with composition of An 70 and an iron-bearing phase (e.g., ilmenite) with Mg/Fe = 0 (vector origin at Mg# = 0; An 70 ) is depicted in Fig. 10. Theoretically, simultaneous exsolution of plagioclase and a Fe phase should move the composition of a spot analysis according to these vectors away from the original high-aluminum composition towards lower Al 2 O 3 and higher Mg#, thus, providing a robust visual tool to evaluate the potential effect of exsolution on HAOM composition. In similar terms, purely isobaric orthopyroxene crystallization should produce sub-horizontal evolution vectors towards lower Mg# at almost constant Al 2 O 3 values and the unlikely exsolution of pure plagioclase without a Fe phase [i.e., Eskola component (Ca 0.5 □ 0.5 AlSi 2 O 6 )-type exsolution] would produce a completely vertical shift. In theory, HAOM crystallization during decompression should produce intermediate vectors reflecting simultaneous decrease in Mg# due to melt evolution and Al 2 O 3 due to diminishing Al solubility.
Most of the spot analyses included in the global data set do not reveal significant deviation from bulk values except for the Harp Lake spot values (Emslie 1980) most of which define a clear exsolution trend towards higher Mg#'s and lower Al 2 O 3 values. Another clear positive correlation is defined by the non-megacrystic orthopyroxene analyses from almost all HAOM-bearing localities that varies from ca. Mg# = 75; Al 2 O 3 = 2 wt% to Mg#'s even as low as 40 and well below 1 wt% Al 2 O 3 . This trend most likely represents crystallization in low-pressure conditions at emplacement levels in the upper crust.
In general, a considerable portion of the HAOM data set has Mg#'s of around 85-70 but variable Al 2 O 3 from ca. 2.5 to 8.5 wt%, which can be interpreted to reflect differences in parental magma composition and pressures at the onset of HAOM crystallization (e.g., Charlier et al. 2010). This is especially evident for the bulk analyses from Egersund Ogna (Charlier et al. 2010) and Rogaland (Bybee et al. 2014) anorthosites that define clear groups at relatively constant Al 2 O 3 values. Using the Ahvenisto data set as a guideline for a general rapid-ascent trend from high-pressure (Al 2 O 3 = 8 ± 1 wt%) to upper crustal environments, similar evolution trends (beginning at different Al contents but overlapping trends) can be traced at least for the Nain (Wiebe 1986;Bybee et al. 2014), Saint Urbain (Dymek and Gromet 1984), and Adirondack (Jaffe and Schumacher 1985;Bohlen and Essene 1978) anorthosites. Bulk HAOM analyses from the Nain Plutonic Suite (Bybee et al. 2014) define a rather gentle-sloped positive correlation which could be interpreted as a slow ascent trend. A similar, but in terms of ascent rates, a slightly faster trend can be traced for the Egersund Ogna center phase (Charlier et al. 2010).
In summary, HAOM compositions from global massif-type anorthosite occurrences are remarkably similar and define broadly overlapping evolution trends that can be interpreted to have been caused by comparable crystallization processes. If all features of the data set can robustly be shown to correspond to similar crystallization histories, the petrogenetic processes for most massif-type anorthosites thus should be fairly similar. Fig. 9 Two alternative emplacement scenarios for the Fennoscandian rapakivi suite. Scenario A is based on Al-in-Hbl geobarometry from rapakivi granites (Elliott 2001), which suggests relatively high crystallization pressures (~ 400 MPa) for the granites exposed at the present-day erosion level. Scenario B combines field observations from several previous studies and results of Al-in-Opx geobarometry from this study to infer a much shallower emplacement depth for the entire rapakivi suite. Intrusion architecture of the upper crust is based on Rämö and Haapala (2005)

Conclusions
Two texturally different but compositionally similar HAOM types were recognized from the Ahvenisto complex of SE Finland. Their compositional similarities suggest that they share a common, high-pressure origin and crystallization history. A clear three-stage compositional evolution is observed in the Ahvenisto HAOMs. Their Al content decreases significantly from the core regions (4.4-7.6 wt% Al 2 O 3 ), through the rims (1.3-3.6 wt%), into the host rock (0.5-1.5 wt%). Enstatite compositions overlap, but are generally higher in the cores (En~6 0-70 ) and rims (En~5 0-70 ) of the HAOMs than in the host (En~4 5-60 ) orthopyroxenes. This compositional evolution reflects a polybaric crystallization history of the parental anorthositic magmas. The highest recorded Al abundances in the HAOM cores correspond to crystallization pressures of up to ~ 1.1 GPa and depths of ~ 34 km. In contrast, the HAOM rims crystallized at much lower pressures (maximum of ~ 0.5 GPa/15-20 km). The highest pressure estimates within barometer calibration for the host rock orthopyroxene are ~ 0.2 GPa (< 7 km). All Ahvenisto HAOMs lack the characteristic plagioclase exsolution lamellae present in nearly every other HAOM locality, but have a conspicuous corona rim that consists  1980;g Emslie 1980;h Owens and Dymek 1995;i Jaffe and Schumacher 1985; Bohlen and Essene 1978;k Dymek and Gromet 1984) of olivine, plagioclase, and low-aluminum orthopyroxene. These reaction rims most likely formed as a response to decompression followed by a subsequent hydration of the system at either late-or post-magmatic stage. The Al-in-Opx geobarometry results set constraints on the emplacement depth of the entire Fennoscandian rapakivi suite, which most likely happened at less than 7 km depth in the interval from 1.65 to 1.55 Ga. A global comparison of HAOMs from major massif-type anorthosite localities suggests that HAOM compositions and evolution of their parental magmas were controlled by similar polybaric petrogenetic processes.