A frozen oceanic crystal mush

The processes driving the evolution of crystal mushes are often documented in complex systems where crystallization, assimilation, magma replenishment and mixing occur concurrently and are generally overprinted by compaction and deformation. Documenting the characteristics of an undisturbed crystal mush is thus of upmost importance; it highlights the initial conditions with which complex crystal mush processes proceed. We here present the structure and composition of an oceanic crystal mush through detailed petro‐structural and chemical study of metre‐scale intrusions from the Mid‐Atlantic Ridge. Textures, bulk‐rock and mineral compositions indicate closed‐system crystallization of primitive melts, undisturbed by dissolution–precipitation reactions and subsequent deformation. These frozen crystal mushes record the simplest possible evolution of small‐scale intrusions and can be used as a baseline to pinpoint the impact of crystal mush processes on the evolution of complex systems. Any divergence from this reference results from processes occurring concomitantly to the progressive closure of the magmatic system.

Each of these processes has profound implications on the chemical evolution of magmatic systems and may control the style of melt migration, extraction, and eruption (e.g., Jackson et al., 2018;Lissenberg et al., 2019;Sparks et al., 2019).
This contribution focuses on mid-ocean ridge systems, where percolation of primitive mid-ocean ridge basaltic (MORB) melts within the mushy gabbroic crust is driven by melt buoyancy and possibly enhanced by compaction, either gravity-or deformationrelated (e.g., Basch, Sanfilippo, Vigliotti, et al., 2022;Lissenberg et al., 2019;McKenzie, 1984). Throughout these crystal mushes, the chemical composition of tholeiitic melts is ubiquitously modified by several magmatic processes, such as: (i) partial crystallization of primitive mineral phases, which leads to progressive enrichment in the melt trace element contents, (ii) dissolution-precipitation reactions between the melt and the percolated gabbroic mush, leading to specific enrichments in highly incompatible trace elements, as commonly evidenced by strong fractionation between Light (L-) and Heavy Rare Earth Elements (H-REE) from core to rim of clinopyroxene crystals (e.g., Basch, Sanfilippo, Skolotnev, et al., 2022;Lissenberg & MacLeod, 2016;Sanfilippo et al., 2020;Zhang et al., 2020), (iii) replenishment and mixing with migrating melts having different chemical compositions, in turn generating hybrid melts, as displayed by complex zoning patterns in crystallized phases (i.e., Coogan & Dosso, 2016;Coumans et al., 2016;Leuthold et al., 2018;Moore et al., 2014). However, because these processes can occur concomitantly, deciphering the impact of a single magmatic process within such complex systems is not trivial. To constrain the evolution of oceanic crystal mushes and properly document the contribution of each process on their chemical composition, we currently lack knowledge on the initial chemical signature of a primitive crystal mush. In this contribution, we document a "baseline" of basaltic crystal mush, that is, the composition and texture of an undeformed crystal mush that solely involved fractional crystallization within a closed magmatic system, and therefore did not suffer dissolutionprecipitation processes, magma replenishment and mixing, nor subsequent deformation. We study metre-scale oceanic dolerites recording crystallization of small-scale magmatic intrusions within the oceanic crust. This crystal mush ought to serve as reference to constrain the early evolution of the tholeiitic intrusions building the lower oceanic crust. The limited dimension of these basaltic intrusions impeded compaction and chemical interaction with exotic melts, thereby providing a proxy of an undeformed crystal mush that did not suffer any chemical modification.

| SAMPLE S ELEC TI ON AND RE SULTS
The selected oceanic dolerites are interpreted as metre-scale magmatic intrusions within the lower oceanic crust formed at the Mid-Atlantic ridge. We chose eight samples recovered from a lower crustal section exposed at an Oceanic Core Complex located at 8.1° N along the Mid-Atlantic Ridge within the Doldrums Fracture Zone (Basch, Sanfilippo, Skolotnev, et al., 2022;Skolotnev et al., 2020). Petrographic descriptions have been accompanied by Olivine (Table S1) shows strongly variable Forsterite (Fo = 63.7-88.6 mol%) and NiO contents (0.06-0.26 wt%) (Figure 2a; Figure S2).
Plagioclase (Table S2) Figure S4). Similar to olivine and plagioclase, clinopyroxene (Table S3) shows strong chemical variability, with Mg-numbers varying from 53.8 to 83.9 mol% and TiO 2 contents varying between 0.51 and 1.46 wt% ( Figure 2c). Notably, the core-to-rim chemical variability of poikilitic clinopyroxene is less systematic than that of plagioclase crystals. Taken as a whole, Bulk-rock compositions of the selected samples reveal a primitive character (Table S4)

Statement of significance
We here document the structure and geochemical composition of a frozen basaltic crystal mush, using oceanic metre-scale intrusions as a proxy of uncompacted crystal mushes that evolved exclusively by fractional crystallization in a closed system. This frozen crystal mush records the simplest possible evolution of an oceanic magmatic chamber and can be used as a reference to pinpoint the impact of crystal mush processes on the structural and chemical evolution of more complex systems. The main implication of this study is that structural and chemical divergence from this undisturbed oceanic crystal mush must be attributed to additional processes occurring concomitantly to the progressive closure of the magmatic system. This contribution is one of the missing pieces of the crystal mush puzzle and we believe that it will be of interest to the scientific community studying igneous petrology of the lower oceanic crust.  (Table S1) similar to primitive N-MORB compositions (Workman & Hart, 2005), with overall low REE contents (e.g., Yb N = 11.9-12.0; N = normalized to CI chondrite from Sun & McDonough, 1989) showing relatively flat patterns slightly depleted in LREE (Ce N /Yb N = 0.82-0.83) and no Eu anomalies ( Figure 2f).

| CLOS ED -SYS TEM E VOLUTI ON AND CRYS TAL G ROW TH
The texture characterizing the dolerites suggests that no compaction nor solid-state deformation affected the samples during and after their crystallization (Bertolett et al., 2019;Cheadle & Gee, 2017;. The random orientation of olivine and plagioclase testifies to crystal suspension within a static and initially melt-supported crystal mush (e.g., Ildefonse et al., 1992Ildefonse et al., , 1997. Furthermore, the occurrence of more primitive troctolitic areas embedded within a more evolved matrix indicates that initial crystal nucleation and growth were neither homogeneously distributed within the metre-sized intrusion, nor concentrated on the walls of the magmatic chamber (e.g., Tegner et al., 2009) In agreement with these textures, the bulk-rock compositions of all samples do not correspond to cumulate gabbros typically showing positive Eu anomalies, but instead to primitive melts that did not suffer any melt extraction (e.g., Godard et al., 2009) Sun and McDonough (1989). The mineral compositions from the Atlantis Massif Hole 1309D gabbros and diabases (Miller et al., 2009) and the global MORB database from Gale et al. (2013) are plotted for comparison. Also shown is the fractional crystallization model of the average composition of dolerites (after Workman & Hart, 2005) at 1 kbar, using the MELTS thermodynamic software (Ghiorso & Sack, 1995). The inset in B compares the MELTS models of fractional crystallization (FC) to the assimilation-fractional crystallization (AFC) paths involving 0.5 and 1 g/°C dissolution of a primitive olivine gabbro (ol:plg:cpx = 0.  Sun and McDonough (1989). The yellow line represents the computed chemical profile of crystal growth involving assimilation of a gabbroic crystal mush upon melt crystallization (AFC), while the red line corresponds to the chemical profile resulting from simple fractional crystallization (FC). [Colour figure can be viewed at wileyonlinelibrary.com] the dissolution of primitive gabbroic mushes would lead to a buffer of the major elements towards primitive compositions (see inset in Figure 2b; detail of the thermodynamic models is given in the Data S1), which is in contrast with the documented linear mineral chemical variations. Furthermore, no strong fractionation is observed between highly to less incompatible elements during progressive growth of plagioclase and clinopyroxene crystals (Figure 3f; Figure S4), as it would be expected during dissolution-precipitation reactions (Figure 3e-h; e.g., Leuthold et al., 2018;Lissenberg & MacLeod, 2016;Sanfilippo et al., 2020;Sparks et al., 2019). Conversely, the dolerite mineral chemical covariations follow fractional crystallization paths at 1 kbar from primitive to evolved compositions (MELTS models; Figure 2a,b; Ghiorso & Sack, 1995). The normal zoning denoted by major and trace element chemical profiles through plagioclase-clinopyroxene contacts (Figure 3; Figure S4; Table S5), together with the crystal growth patterns (e.g., Smith & Lofgren, 1983), developed during the progressive cooling and closure of the magmatic system. This is in line with a simple process of closed-system fractional crystallization (FC) of a primitive parental melt, with no involvement of replenishment, mixing or reactive processes. Ultimately, the growth patterns of poikilitic clinopyroxene correspond to the closure of the porosity, "locking" the crystal mush at temperatures close to the solidus.

| AN UND IS TURB ED CRYS TAL MUS H
The texture and compositions characterizing the studied small-scale intrusions provide important constraints on the petrological evolution of a tholeiitic crystal mush during progressive cooling. Namely, crystallization initiated heterogeneously within the intrusion, forming primitive areas of euhedral olivine and plagioclase oriented randomly (Stage 1 in Figure 4). Upon cooling, crystal growth dominated within the primitive areas, while clinopyroxene started to nucleate together with plagioclase within the crystal-poor areas (Stage 2 in Figure 4). The composition of the melt progressively evolved with decreasing melt mass, until the crystallization of Fe-Ti oxides and the growth of large clinopyroxene oikocrysts locked the porosity (Stage 3 in Figure 4). This resulted in a random orientation of the crystals and in chemical compositions that follow fractional crystallization trends. Additional processes such as compaction, melt extraction, melt mixing or reactive migration of exotic melts would have caused structural and/or chemical perturbation from the textural and chemical patterns documented in our samples. For example, such complexity has been extensively documented within the lower gabbroic crust from the East Pacific Rise (Hess Deep; e.g., Leuthold et al., 2018;Natland & Dick, 1996) and Southwest Indian Ridge (Atlantis Bank; e.g., Boulanger et al., 2021;Sanfilippo et al., 2020), where a combination of compaction, reactive porous flow and melt replenishment F I G U R E 4 Representative evolution of a crystal mush during static crystallization. Stage 1: Heterogeneous crystallization of meltsuspended euhedral olivine and plagioclase. Stage 2: Crystallization of plagioclase and clinopyroxene cores and progressive decrease in porosity. Stage 3: Closure of the porosity at solidus temperatures and formation of plagioclase rims, clinopyroxene oikocrysts and Fe-Ti oxides. Upper right corner: Textural and chemical evolution during cooling and progressive closure of the magmatic system. [Colour figure can be viewed at wileyonlinelibrary.com] obscured the initial evolution of the crystal mush. We emphasize that such "disturbed" magmatic systems are the rule rather than the exception for oceanic gabbros and the uniqueness of the studied samples stand in their relatively fast and static cooling, which allowed for preservation of the primary texture and chemical composition.
This contribution therefore provides a Rosetta stone for deciphering the complexity of the processes occurring in oceanic crystal mushes and implies that any divergence from this reference can be attributed to processes debated in literature, such as dissolution-precipitation, magma mixing and compaction-driven melt extraction.

ACK N O WLE D G E M ENTS
We thank the captain, the officers, and the crew of R/V Akademik Nikolaj Strakhov. We thank G. Wörner for his work as editor, as well as M. Holness and an anonymous reviewer for constructive comments that helped to improve the quality of the manuscript. We also thank

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that supports the findings of this study are available in the supplementary material of this article