Pre-Pegmatite Stage in Peralkaline Magmatic Process: Insights from Poikilitic Syenites from the Lovozero Massif, Kola Peninsula, Russia

: The Lovozero peralkaline massif (Kola Peninsula, Russia) is widely known for its unique mineral diversity, and most of the rare metal minerals are found in pegmatites, which are spatially associated with poikilitic rocks (approximately 5% of the massif volume). In order to determine the reasons for this relationship, we have investigated petrography and the chemical composition of poikilitic rocks as well as the chemical composition of the rock-forming and accessory minerals in these rocks. The differentiation of magmatic melt during the formation of the rocks of the Lovozero massif followed the path: lujavrite → foyaite → urtite (magmatic stage) → pegmatite (hydrothermal stage). Yet, for peralkaline systems, the transition between magmatic melt and hydrothermal solution is gradual. In the case of the initially high content of volatiles in the melt, the differentiation path was probably as follows: lujavrite → foyaite (magmatic stage) → urtitization of foyaite → pegmatite (hydrothermal stage). Poikilitic rocks were formed at the stage of urtitization, and we called them pre-pegmatites. Indeed, the poikilitic rocks have a metasomatic texture and, in terms of chemical composition, correspond to magmatic urtite. The reason for the abundance of rare metal minerals in pegmatites associated with poikilitic rocks is that almost only one nepheline is deposited during urtitization, whereas during the magmatic crystallization of urtite, rare elements form accessory minerals in the rock and are less concentrated in the residual solution.


Introduction
Pegmatites form at the final stage of the magmatic process with the participation of late magmatic fluids. Most pegmatites are mineralogically simple, essentially very coarse rocks. Others contain a tremendous concentration of incompatible elements and a highly varied mineralogy, commonly displaying a concentric zonation (see e.g., [1][2][3][4]). As the late fluid segregation concentrates trace elements, pegmatites are important economic resources and are mined for Li, Be, the rare earth elements (REE), W, Zr, and a host of other elements that are rarely concentrated in other environments [5,6]. They are also a major source of gems.
Pegmatite genesis is a subject of controversy. One school of thought (see, e.g., [7,8]) proposes that they are the crystallization products of residual melts intruding fractures in country rocks. Such residual melts are relatively rich in volatiles (H 2 O, F, Cl, etc.). A different origin has been proposed by other researchers (see, e.g., [1,9,10]). Their interpretation is based on the frequent chaotic structure of the pegmatite veins, in which recrystallization phenomena are widespread, with resorption and chemical substitution of earlier-crystallized minerals. In this scenario, pegmatites result from profound transfor-
A large number of xenoliths of Devonian volcaniclastic rocks [30,42], both unaltered and intensely metasomatized (fenitized), are found among the rocks of the Layered and Eudialyte complexes. Unaltered xenoliths are composed of olivine basalt, basalt tuff, tuffite, and sandstone. Alkaline lamprophyre dikes up to 5 m thick are most common in the northwest and south of the Lovozero massif.
Poikilitic complex (or complex of poikilitic syenites) is a collective name for a group of rocks diverse in mineral and modal composition [28][29][30]53]. The rocks of this complex occupy approximately 5% of the massif volume; they form lenses, sheet-like or irregularly shaped bodies with sizes of 10 km × 0.5 km to 50 cm × 30 cm, which are located in both the Layered and Eudialyte complexes (Figure 1b). The main features of the rocks of the poikilitic complex are the presence of large (up to 15 cm in length) crystals of feldspar with numerous inclusions of feldspathoids (sodalite, nepheline, cancrinite, vishnevite), as well as metasomatic texture. The amount of poikilitic crystals of feldspar varies from single grains to 90% of the rock volume. In the poikilitic complex, the following main groups of rocks are recognized [29,30]: Uneven-grained nepheline syenite (poikilitic feldspar is rare, the main foid is nepheline)the name of this rock reflects that the size of the grains of the rock-forming nepheline varies greatly (from 50 µm to 2 cm across); For brevity's sake, in the following text, we will use the term poikilitic rocks instead of rocks of the poikilitic complex as a general name for these two types of rocks.
According to Bussen and Sakharov [29], the boundaries between uneven-grained nepheline syenite and poikilitic foid syenite are gradual. These rocks can either be found together in the volume of one body or form separate independent bodies.
The contacts of the poikilitic rocks with surrounded rocks of the Layered or Eudialyte complex are diverse. The contact of uneven-grained nepheline syenite with lujavrite is sharp and the contact with foyaite is either sharp or gradual. At the contact of poikilitic foid syenite with lujavrite, there is a transition zone consisting of a large (up to 10 cm in length) microcline crystals (without poikilitic inclusions) and monomineral aegirine segregations. This zone is enriched in EGM, lorenzenite, murmanite, lamprophyllite, and rinkite-(Ce). Lujavrite lacks a trachytoid texture near the contact (in a zone up to 0.5 m wide) with poikilitic rocks.
Pegmatites are found both at the contact of poikilitic rocks with surrounded rocks (Figure 2a,b) and within bodies of poikilitic rocks (Figure 2c). The transition between poikilitic rocks and pegmatites is gradual. Table 2 lists the rare metal minerals (for example see Figure 2d,e) characteristic to pegmatites spatially associated with poikilitic rocks according to Pekov [25] and Semenov [28].

Materials and Methods
We studied poikilitic rocks, the bodies of which are located in the Eudialyte complex of the Lovozero massif. Samples for the study were collected in two areas ( Figure 1): Alluive Mountain (drill cores, sampling profile 1 in Figure 1) and Sengischorr Mountain (outcrops, sampling area 2 in Figure 1). Figure 3a shows a schematic cross-section of sampling profile 1 (see Figure 1), well locations, and sampling intervals. On the Sengischorr Mountain, samples were taken from six outcrops. One such outcrop and sampling points are shown in Figure 3b.

Materials and Methods
We studied poikilitic rocks, the bodies of which are located in the Eudialyte complex of the Lovozero massif. Samples for the study were collected in two areas ( Figure 1): Alluive Mountain (drill cores, sampling profile 1 in Figure 1) and Sengischorr Mountain (outcrops, sampling area 2 in Figure 1). Figure 3a shows a schematic cross-section of sampling profile 1 (see Figure 1), well locations, and sampling intervals. On the Sengischorr Mountain, samples were taken from six outcrops. One such outcrop and sampling points are shown in Figure 3b.  images and pre-analyze all detected minerals. The chemical composition of minerals was analyzed with the Cameca MS-46 electron microprobe (Cameca, Gennevilliers, France) operating in the WDS-mode at 22 kV with a beam diameter of 10 µm, beam current of 20-40 nA, and counting times of 20 s (for a peak) and 2-10 s (for background before and after the peak), with 5-10 counts for every element in each point. The analytical precision (reproducibility) of mineral analyses was 0.2-0.05 wt% (2 standard deviations) for the major element and approximately 0.01 wt% for impurities. The standards used, the detection limits, and the analytical accuracy values are given in Supplementary Table S1. The systematic errors were within the random errors.
Diagnostics of feldspathoids, as well as the study the structural state of the potassium feldspar, were carried out using X-ray diffraction (XRD). XRD measurements were taken with a DRON-2 diffractometer (Burevestnik, Saint-Peterburg, Russia) at the GI KSC RAS. The operating parameters were as follows: CuKα radiation, 20 mA, 30 kV. XRD data were identified using the RRUFF Project database [55].
In situ laser ablation inductively coupled plasma mass spectrometer (LA-ICP-MS) analyses of Li in amphiboles were performed at the Institute of Geology Karelian Research Centre Russian Academy of Sciences (Petrozavodsk, Russia) using a mass spectrometer X-Series 2 (Thermo Fisher Scientific, Waltham, MA, USA) equipped with a 266 nm Nd-YAG laser (UP-266 Macro). NIST 640 was used as a secondary standard. The laser beam diameter at the sample surface was approximately 70 µm.
Major elements in rocks were determined by a wet chemical analysis at the GI KSC RAS. The accuracy limits of the wet chemical analysis are given in Supplementary Table S1. Statistical analyses were carried out using the STATISTICA 13 [56]. For the statistics, resulting values of the analyses below the limit of accuracy (see Supplementary Table S1) were considered ten times lower than the limit.
Factor analysis was used to detect patterns of changes in the chemical composition of rocks and minerals. It is a statistical method used to describe variability among observed, correlated variables in terms of a lower number of unobserved variables called factors [57]. This method allows us to describe an object comprehensively and, at the same time, compactly.
Mineral abbreviations mentioned in this article are shown in Table 3.
Based on petrographic observations, a sequence of metasomatic events could be determined. In uneven-grained nepheline syenite, the intensity of metasomatism is relatively low. This rock contains numerous relics of microcline, which were replaced by albite along the grain edges and along fractures ( Figure 4a). Albitization was followed by intense nephelinization. Indeed, relics of microcline and albite are often present in the fine-grained mass of nepheline. We assume that large crystals (up to 3 cm across, Figure 4b) of nepheline are formed as a result of recrystallization (or coarsening [60]) of small grains of this mineral. grains of loparite-(Ce), zircon, and pyrochlore-group minerals are often found within ilmenite grains (Figure 5c). Phlogopite forms lamellar crystals with numerous inclusions of nepheline, titanite, ilmenite, fluorapatite, loparite-(Ce) and is surrounded by a ferri-katophorite rim (Figure 5d). Parakeldyshite forms rounded grains with numerous inclusions of loparite-(Ce) and is surrounded by an EGM rim ( Figure 5e).  Mafic minerals form polymineral segregations in the fine-grained nepheline mass (Figure 4a,c). The main mafic minerals are clinopyroxenes and amphiboles, which are usually arranged in an orderly manner: calcium end-members (e.g., diopside) are first surrounded by a rim of a calcium-sodium mineral (e.g., aegirine-augite) and outside, there is a rim of a sodium end-member (e.g., aegirine, Figure 5a). Other examples are as follows (from core to rim): magnesio-arfvedsonite-aegirine (Figures 4a and 5b), diopside-ferrikatophorite-aegirine, diopside-richterite-aegirine. In spatial association with clinopyroxenes and amphiboles, there are ilmenite grains (sometimes with magnetite lamellae) surrounded by a titanite rim (Figure 5b), wedge-shaped crystals of titanite, rounded fluorapatite and magnetite grains, cube crystals of loparite-(Ce) and pyrochlore group minerals, zircon grains with small inclusions of baddeleyite. Moreover, small crystals or rounded grains of loparite-(Ce), zircon, and pyrochlore-group minerals are often found within ilmenite grains ( Figure 5c). Phlogopite forms lamellar crystals with numerous inclusions of nepheline, titanite, ilmenite, fluorapatite, loparite-(Ce) and is surrounded by a ferri-katophorite rim (Figure 5d). Parakeldyshite forms rounded grains with numerous inclusions of loparite-(Ce) and is surrounded by an EGM rim (Figure 5e). In uneven-grained nepheline syenite, nepheline is replaced by vishnevite (Figure 4a), and feldspar is recrystallized with the formation of small crystals of orthoclase In uneven-grained nepheline syenite, nepheline is replaced by vishnevite (Figure 4a), and feldspar is recrystallized with the formation of small crystals of orthoclase (Figures 4a  and 5f). Some orthoclase crystals contain single poikilitic inclusions of nepheline and/or vishnevite. However, the amount of vishnevite in this rock is small, and poikilitic orthoclase is rare.
Poikilitic foid syenite is composed of large poikilitic orthoclase crystals with vishnevite inclusions (Figure 6a). In this rock, not only nepheline ( Figure 6b) but also orthoclase ( Figure 6c) is intensively replaced by vishnevite. Poikilitic foid syenite was potentially formed as a result of further development of orthoclasization and vishnevitization processes, the first signs of which were observed in uneven-grained nepheline syenite. The association of mafic minerals in poikilitic foid syenite is similar to that in uneven-grained nepheline syenites. Tabular crystals of phlogopite contain titanite inclusions and are surrounded by a rim of small crystals of aegirine (Figure 6d,f). Diopside grains are intensively resorbed by ferri-kataphorite, richterite, and aegirine ( Figure 6e). Titanite forms irregular grains, wedge-shaped crystals, and rims around magnetite and ilmenite (Figure 6f). Magnetite contains ilmenite or ulvöspinel lamellae. Fluorapatite, zircon, tausonite, rinkite-(Ce), wöhlerite, låvenite, EGM, pyrrhotite, and sphalerite are characteristic for accessory minerals.
Thus, in the process of the formation of poikilitic rocks, there are two main stages: albitization + nephelinization (as a result, uneven-grained nepheline syenite was formed) and orthoclasization + vishnevitization (at this stage, poikilitic foid syenite was formed).

Contact between Poikilitic Rocks and Eudialyte Lujavrite
We studied the contact between poikilitic foid syenite and eudialyte lujavrite using samples from profile LV-377 (see Figure 3b). The following zones were distinguished: • −20-0 cm from contact: Pegmatite-like zone consisting of large crystals of microcline and monomineral lenses of aegirine.

Rock Chemistry
The chemical compositions of uneven-grained nepheline syenite and poikilitic foid syenite vary widely (Supplementary Table S2). According to the results of factor analysis ( Figure 8), uneven-grained nepheline syenite is relatively enriched in Si and K, as well as in Zr, Ti, Mn, Fe 3+ , Li, while poikilitic syenite is enriched in Na and Al, as well as in Ca, Sr, P, and volatiles (S, Cl, H 2 O).
Using factor analysis, we compared the chemical compositions of the poikilitic rocks with those of the main rocks of the massif [54]. First, it should be noted that the compositions of lujavrite and foyaite from the Eudialyte complex and similar rocks from the Layered complex are similar. Lujavrite is enriched in ferrous and ferric iron, calcium, magnesium (included in compositions of rock-forming Na-clinopyroxenes and Na-amphiboles), titanium, fluorine (included in the composition of lamprophyllite), zirconium, manganese, and chlorine (included in the composition of EGM). With the transition from lujavrite to foyaite, the content of above-mentioned elements decreases and the amount of potassium (i.e., feldspar) increases. The main rock-forming mineral of the Layered complex urtite is nepheline, and the characterizing accessory minerals are various silicophosphates and phosphates, mainly fluorapatite [61]. Therefore, urtite is characterized by enrichment in Al, Na, Ca, and P. There is no urtite in the Eudialyte complex (Table 2), but its chemical composition corresponds to poikilitic rocks. The difference between urtite and poikilitic rocks is that the latter contain more sulfur and strontium. grained nepheline syenites. Tabular crystals of phlogopite contain titanite inclusions and are surrounded by a rim of small crystals of aegirine (Figure 6d,f). Diopside grains are intensively resorbed by ferri-kataphorite, richterite, and aegirine ( Figure 6e). Titanite forms irregular grains, wedge-shaped crystals, and rims around magnetite and ilmenite (Figure 6f). Magnetite contains ilmenite or ulvöspinel lamellae. Fluorapatite, zircon, tausonite, rinkite-(Ce), wöhlerite, låvenite, EGM, pyrrhotite, and sphalerite are characteristic for accessory minerals. Thus, in the process of the formation of poikilitic rocks, there are two main stages: albitization + nephelinization (as a result, uneven-grained nepheline syenite was formed) and orthoclasization + vishnevitization (at this stage, poikilitic foid syenite was formed).

Contact between Poikilitic Rocks and Eudialyte Lujavrite
We studied the contact between poikilitic foid syenite and eudialyte lujavrite using samples from profile LV-377 (see Figure 3b). The following zones were distinguished:

Rock Chemistry
The chemical compositions of uneven-grained nepheline syenite and poikilitic foid syenite vary widely (Supplementary Table S2). According to the results of factor analysis (Figure 8), uneven-grained nepheline syenite is relatively enriched in Si and K, as well as in Zr, Ti, Mn, Fe 3+ , Li, while poikilitic syenite is enriched in Na and Al, as well as in Ca, Sr, P, and volatiles (S, Cl, H2O).
Using factor analysis, we compared the chemical compositions of the poikilitic rocks with those of the main rocks of the massif [54]. First, it should be noted that the compositions of lujavrite and foyaite from the Eudialyte complex and similar rocks from the Layered complex are similar. Lujavrite is enriched in ferrous and ferric iron, calcium, magnesium (included in compositions of rock-forming Na-clinopyroxenes and Na-amphiboles), titanium, fluorine (included in the composition of lamprophyllite), zirconium, manganese, and chlorine (included in the composition of EGM). With the transition from lujavrite to foyaite, the content of above-mentioned elements decreases and the amount of potassium (i.e., feldspar) increases. The main rock-forming mineral of the Layered complex urtite is nepheline, and the characterizing accessory minerals are various silicophosphates and phosphates, mainly fluorapatite [61]. Therefore, urtite is characterized by enrichment in Al, Na, Ca, and P. There is no urtite in the Eudialyte complex (Table 2), but its chemical In Figure 9, the points corresponding to the chemical compositions of the samples from the LV-377 profile are separately highlighted (see Figures 3b and 7). Closer to the contact with poikilitic rocks, the content of Ca, Mg, Fe, Ti, Zr, Mn, F, and Cl (main components of mafic minerals) in lujavrite decreases, while the concentration of potassium (i.e., feldspar) increases.  Supplementary Table S3. composition corresponds to poikilitic rocks. The difference between urtite and poikilitic rocks is that the latter contain more sulfur and strontium. In Figure 9, the points corresponding to the chemical compositions of the samples from the LV-377 profile are separately highlighted (see Figures 3b and 7). Closer to the contact with poikilitic rocks, the content of Ca, Mg, Fe, Ti, Zr, Mn, F, and Cl (main components of mafic minerals) in lujavrite decreases, while the concentration of potassium (i.e., feldspar) increases.

Nepheline
In the igneous rocks of the Lovozero (and Khibiny) massifs, nepheline is represented by two morphological types [62,63]. Nepheline-I occurs as euhedral crystals with numerous small inclusions of aegirine and microcline, and nepheline-II forms anhedral grains without inclusions. Nepheline-II is enriched in iron, which, as it was found in [54], is included in nepheline crystal structure according to the following scheme:  Only nepheline-II (without inclusions) was found in poikilitic rocks. Representative chemical analyses are presented in Supplementary Table S4. Trivalent iron (up to 0.14 Fe 3+ pfu) enters in nepheline from these rocks according to the above-mentioned substitution scheme, because there is a negative correlation between (Si + Fe 3+ ) apfu and Al pfu (Figure 10a) as well as a positive correlation Al pfu vs. K pfu (Figure 10b). This substitution does not allow for using the nepheline composition for temperature determination [64]. Differences in the composition of nepheline from uneven-grained nepheline syenite and poikilitic foid syenite were not found.

Vishnevite
Representative chemical analyses of vishnevite are given in Supplementary Table S5. The sums of oxides in all chemical analyses of vishnevite are very low (mean 92.5 wt%). This could be due to the presence of water or a CO 3 2− group. According to Semenov [28], vishnevite from the Lovozero massif does not contain carbonate groups. Although we did not make direct measurements of CO 2 in the studied samples, the calculation of the charge balance showed a lack of positive charges and not negative ones (associated with the presence of carbonate groups). Supposedly, the lack of the sum in the composition of vishnevite corresponds to the water content, and not only in the form of H 2 O but also in the form of OH-groups. OH-groups could be included in vishnevite according to the scheme Na + + O 2− → ( , H 2 O) 0 + (OH) − , i.e., vishnevite in poikilitic rocks is hydrated. Indeed, the studied samples are characterized by a low content of sodium (mean 7.52 apfu) as well as potassium (mean 0.11 apfu) and calcium (mean 0.02 apfu). The Al/Si ratio in vishnevite from uneven-grained nepheline syenite is lower (mean 0.89) than the one from poikilitic foid syenite (mean 0.94).

Amphiboles
Representative chemical analyses of amphiboles are shown in Supplementary Table  S6. The amphibole-group mineral formulae were calculated based on O + OH + F = 24 atoms per formula unit and OH = 2 -2Ti. The formula calculation was performed following the IMA 2012 recommendations [65] using the Excel spreadsheet of Locock [66]. This calculation resulted in slightly underoccupied C-cites and over-occupied A-sites (for the general formula AB2C5T8O22W2), which may be explained by significant Li content [67].

Amphiboles
Representative chemical analyses of amphiboles are shown in Supplementary Table S6. The amphibole-group mineral formulae were calculated based on O + OH + F = 24 atoms per formula unit and OH = 2 -2Ti. The formula calculation was performed following the IMA 2012 recommendations [65] using the Excel spreadsheet of Locock [66]. This calculation resulted in slightly underoccupied C-cites and over-occupied A-sites (for the general formula AB 2 C 5 T 8 O 22 W 2 ), which may be explained by significant Li content [67]. Indeed, the lithium content in amphiboles reaches 0.12 apfu (Table S6).
Unlike other rocks of the Eudialyte complex, where amphiboles are represented only by sodium end-members (magnesio-arfvedsonite and, less often, arfvedsonite), a wide variety of amphiboles was found in poikilitic rocks. Magnesio-arfvedsonite, usually forming symplectite intergrowths with nepheline, is typical of uneven-grained nepheline syenite (Figures 4a and 5a). Poikilitic foid syenite typically contain sodium-calcium amphiboles, such as ferri-katophorite, ferri-fluoro-katophorite, and richterite. Indeed, amphiboles from poikilitic rocks are characterized by higher concentrations of aluminum and calcium compared with amphiboles from other rocks of the Eudialyte complex (Figure 11a,b). In addition to lithium, titanium (max 0.30 apfu in poikilitic foid syenite) and zirconium (max 0.02 apfu in uneven-grained nepheline syenite) are typical impurities in amphiboles.

Clinopyroxenes
The compositions of clinopyroxenes in the Mg-(Fe 2+ +Mn)-Na system are shown in Figure 12. Representative microprobe analyses of clinopyroxenes are available in Supplementary Table S7. The relict clinopyroxene grains are enriched in magnesium and calcium and correspond in composition to diopside or augite. In addition, there are relict grains of aegirine-augite (64-78 mol% Quad, 17-31 mol% aegirine, 1.5-6 mol% jadeite). The com-

Other Minerals
Typical accessory minerals of poikilitic rocks are phlogopite, fluorapatite, titanite, ilmenite, and magnetite. Representative chemical analyses of phlogopite are presented in Supplementary Table S8. Phlogopite is enriched in titanium, and the titanium concentration in phlogopite from uneven-grained nepheline syenite is higher (up to 0.24 apfu) than in phlogopite from poikilitic foid syenite (up to 0.19 apfu).
Representative chemical analyses of fluorapatite are shown in Supplementary Table  S9. Fluorapatite is enriched in strontium (up to 1.31 apfu) and REE (up to 0.10 Ce pfu, 0.05 La pfu), and the Si content does not exceed 0.09 apfu. The main substitution scheme in fluorapatite is 2(Ca,Sr) 2+ ⇆ Na + + REE 3+ . In fluorapatite from uneven-grained nepheline  Supplementary Table S8. Phlogopite is enriched in titanium, and the titanium concentration in phlogopite from uneven-grained nepheline syenite is higher (up to 0.24 apfu) than in phlogopite from poikilitic foid syenite (up to 0.19 apfu).
Representative chemical analyses of fluorapatite are shown in Supplementary Table S9. Fluorapatite is enriched in strontium (up to 1.31 apfu) and REE (up to 0.10 Ce pfu, 0.05 La pfu), and the Si content does not exceed 0.09 apfu. The main substitution scheme in fluorapatite is 2(Ca,Sr) 2+ Na + + REE 3+ . In fluorapatite from uneven-grained nepheline syenite, this reaction is shifted to the right, while fluorapatite from poikilitic foid syenite, on the contrary, contains more calcium and strontium (Figure 13a).
Ilmenite usually forms homogeneous grains, but in uneven-grained nepheline syenite, it sometimes contains magnetite lamellae, while in poikilitic foid syenite, on the contrary, ilmenite lamellae are found in magnetite. Homogeneous ilmenite grains are enriched in manganese (up to 0.41 Mn pfu in ilmenite from uneven-grained nepheline syenite, Figure 13b). Representative chemical analyses for ilmenite are presented in Supplementary Table S10. Magnetite forms homogeneous grains, lamellae in ilmenite (in uneven-grained nepheline syenite), and grains with ilmenite lamellae (in poikilitic foid syenite). In homogeneous grains, the titanium content is maximal and reaches 0.24 apfu (e.g., a homogeneous magnetite grain in Figure 5d). Representative chemical analyses of magnetite, both homogeneous and with ilmenite lamellae, are shown in Supplementary Table S11.
Representative chemical analyses for titanite are shown in Supplementary Table S12. Titanite contains a large number of elements, such as Nb, Fe, Na, and REE, and several exchange mechanisms allowing different cations to enter the structure have been proposed. The most important chemical substitutions responsible for the chemical variations in titanite are illustrated in Figure 14a,b: 2Ca 2+ ⇆ Na + + REE 3+ ; 2Ti 4+ ⇆ Nb 5+ + (Al, Fe 3+ ) ; Ti 4+ ⇆ Zr 4+ Titanite from poikilitic foid syenite is enriched in calcium and elements that replace titanium (Nb, Zr, Al, Fe 3+ ). In titanite from uneven-grained nepheline syenite, on the contrary, the titanium content is maximum, and the calcium content is minimum. Magnetite forms homogeneous grains, lamellae in ilmenite (in uneven-grained nepheline syenite), and grains with ilmenite lamellae (in poikilitic foid syenite). In homogeneous grains, the titanium content is maximal and reaches 0.24 apfu (e.g., a homogeneous magnetite grain in Figure 5d). Representative chemical analyses of magnetite, both homogeneous and with ilmenite lamellae, are shown in Supplementary Table S11.
Representative chemical analyses for titanite are shown in Supplementary Table S12. Titanite contains a large number of elements, such as Nb, Fe, Na, and REE, and several exchange mechanisms allowing different cations to enter the structure have been proposed. The most important chemical substitutions responsible for the chemical variations in titanite are illustrated in Figure 14a, Titanite from poikilitic foid syenite is enriched in calcium and elements that replace titanium (Nb, Zr, Al, Fe 3+ ). In titanite from uneven-grained nepheline syenite, on the contrary, the titanium content is maximum, and the calcium content is minimum.

Discussion
The rocks of the poikilitic complex we studied are formally divided into two groups: uneven-grained nepheline syenite and poikilitic foid syenite. These two groups of rocks are connected by gradual transitions, have a metasomatic texture and a similar mineral (but not modal) composition.
Uneven-grained nepheline syenite formed at the first stage of metasomatiс alteration. This rock contains numerous relics of microcline laths, which are intensively replaced by albite and nepheline (Figure 4a). Foyaite is a rock enriched in feldspar (microcline) in both Eudialyte and Layered complexes of the Lovozero massif [29,30,54]). In the cross section along line 1 (Figures 1b and 3a), it can be seen that in the Eudialyte complex, poikilitic rocks are located either among the foyaite or at the same depth. In uneven-grained nepheline syenites, the content of microcline relics reaches 40% of the rock volume; the laths of the relict microcline are located chaotically. All this suggests that the protolith was a massive leucocratic feldspar rock, namely, foyaite.
To determine which elements were added to or lost from the rock during the transformation of foyaite into uneven-grained nepheline syenite, we used the method of Grant [68], which involves a pairwise graphical comparison of the compositions of altered and unaltered rocks. Immobile elements should define a line (isocon) passing through the origin, the slope of which represents the ratio of the masses of the altered to the fresh rock. A slope of 1 indicates that there has been no mass change, whereas higher or lower slopes imply a gain or loss of mass (or loss or gain of volume). Figure 15a shows an isocon diagram for a pair of foyaite (unaltered rock) and uneven-grained nepheline syenite (altered rock). To construct this diagram, we used the mean rock compositions from Supplementary Table S2. The line passing closest to the greatest number of elements has a slope of 1, indicating that this alteration occurred without a change in the mass of the rock ( Figure  15a). Conclusions that can be drawn from this isocon diagram are that Na2О, Al2О3, CaО, MgО, P2О5, REE2О3, TiО2, and F, S, CO2 were added to the rock during metasomatic alteration, while K2О, ZrО2, Li2О, MnО, Cl, and H2O were removed. Na + + REE 3+ ; (b) binary diagram shows that Ti is involved in a number of substitutions according to the schemes 2Ti 4+ Nb 5+ + (Al, Fe 3+ ) and Ti 4+ Zr 4+ .

Discussion
The rocks of the poikilitic complex we studied are formally divided into two groups: uneven-grained nepheline syenite and poikilitic foid syenite. These two groups of rocks are connected by gradual transitions, have a metasomatic texture and a similar mineral (but not modal) composition.
Uneven-grained nepheline syenite formed at the first stage of metasomatic alteration. This rock contains numerous relics of microcline laths, which are intensively replaced by albite and nepheline (Figure 4a). Foyaite is a rock enriched in feldspar (microcline) in both Eudialyte and Layered complexes of the Lovozero massif [29,30,54]). In the cross section along line 1 (Figures 1b and 3a), it can be seen that in the Eudialyte complex, poikilitic rocks are located either among the foyaite or at the same depth. In uneven-grained nepheline syenites, the content of microcline relics reaches 40% of the rock volume; the laths of the relict microcline are located chaotically. All this suggests that the protolith was a massive leucocratic feldspar rock, namely, foyaite.
To determine which elements were added to or lost from the rock during the transformation of foyaite into uneven-grained nepheline syenite, we used the method of Grant [68], which involves a pairwise graphical comparison of the compositions of altered and unaltered rocks. Immobile elements should define a line (isocon) passing through the origin, the slope of which represents the ratio of the masses of the altered to the fresh rock. A slope of 1 indicates that there has been no mass change, whereas higher or lower slopes imply a gain or loss of mass (or loss or gain of volume). Figure 15a shows an isocon diagram for a pair of foyaite (unaltered rock) and uneven-grained nepheline syenite (altered rock). To construct this diagram, we used the mean rock compositions from Supplementary Table S2. The line passing closest to the greatest number of elements has a slope of 1, indicating that this alteration occurred without a change in the mass of the rock (Figure 15a). Conclusions that can be drawn from this isocon diagram are that Na 2 O, Al 2 O 3 , CaO, MgO, P 2 O 5 , REE 2 O 3 , TiO 2 , and F, S, CO 2 were added to the rock during metasomatic alteration, while K 2 O, ZrO 2 , Li 2 O, MnO, Cl, and H 2 O were removed.
As mass was conserved during alteration (i.e., no correction needs to be applied to the data), we can investigate the relative importance of metasomatic alteration in controlling the composition of rocks by comparing element abundances directly (Supplementary Table S2, Figure 15b,c). With the metasomatic replacement of foyaite, an increase in the contents of Na 2 O and Al 2 O 3 and a decrease in the content of K 2 O occur. This is likely manifested in the replacement of microcline by albite and/or nepheline (Figure 4a), in accordance with possible reactions: Because, in addition to the loss of potassium, a decrease in the SiO 2 content is also observed, the replacement of the microcline with nepheline can also occur in accordance with the reaction: When microcline is replaced with nepheline and/or albite, a significant amount of potassium is released. Some of the potassium is removed from the rock, and some, supposedly, is included in the composition of potassium-enriched minerals (phlogopite, amphiboles, orthoclase), which crystallize in the metasomatic process simultaneously with nepheline. We propose a metasomatic origin for these minerals because (1) phlogopite occurs only in poikilitic rocks [28]; (2) magnesio-arfvedsonite from poikilitic rocks contains more magnesium, calcium, aluminum, and fluorine than magnesio-arfvedsonite from foyaite [54]. Potassium is also part of orthoclase, which forms only small laths at the initial stage of metasomatic replacement of foyaite (Figure 5f) and then appears as large poikilitic crystals.
Furthermore, with metasomatic replacement of foyaite, a significant addition of TiO 2 is observed. Titanium is a part of ilmenite, ulvöspinel-rich magnetite, and titanite. The addition of magnesium and titanium is the reason for the crystallization of Ti-enriched phlogopite with thin plates of ilmenite or titanite located along the cleavage. Moreover, as a result of the addition of magnesium, calcium, and titanium, Quad pyroxenes (diopside and augite) and Ca-enriched amphiboles (ferri-katophorite, richterite) crystallized. The sequence of crystallization of pyroxenes and amphiboles from Ca end-members to Ca-Na end-members and Na end-members potentially occurred due to the fact that at the initial stage of metasomatism, almost all of the added sodium was used on crystallization of nepheline, and the calcium addition was very significant (Figure 15b,c). Sodium-rich clinopyroxenes and amphiboles formed later, with the appearance of excess sodium. As a result of the addition of P, Ca, Sr, F, and REE, SrREE-rich fluorapatite crystallized, as did loparite-(Ce), pyrochlore group minerals, and rinkite-(Ce).
Supposedly, the loss of lithium (its content decreases by 48%), manganese (by 42%), as well as water (in general, the content decreases by 46%) and a part of silicon, is associated with the resorption of amphiboles from foyaite. The relict NaSi-enriched amphiboles of the foyaite were probably unstable during metasomatism and were completely resorbed. Indeed, at the initial stages of nephelinization, the activity of silica and the sodium content decrease. This is evidenced by both the sequence of crystallization of pyroxenes and amphiboles (Ca → Ca-Na → Na end-members) and the pattern in the arrangement of titanium minerals (ilmenite is surrounded by a titanite rim).
The decrease in the zirconium content could be associated with the resorption of alkaline zirconosilicates, mainly minerals from the eudialyte and lovozerite groups, which are characterizing to unaltered foyaite [27,28,51,54]. Part of the zirconium is released, and part is localized first in the form of baddeleyite or zircon and then in the composition of parakeldyshite and minerals of the eudialyte group enriched in calcium and strontium [54]. The sequence of crystallization of zirconium minerals also indicates a low activity of silica and the absence of excess sodium at the initial stage of nephelinization. As mass was conserved during alteration (i.e., no correction needs to be applied to the data), we can investigate the relative importance of metasomatic alteration in controlling the composition of rocks by comparing element abundances directly (Supplementary  Table S2, Figure 15b,c). With the metasomatic replacement of foyaite, an increase in the contents of Na2O and Al2O3 and a decrease in the content of K2O occur. This is likely manifested in the replacement of microcline by albite and/or nepheline (Figure 4a), in accordance with possible reactions: KAlSi3O8 (microcline) + Na + → NaAlSi3O8 (albite) + K + ; With the metasomatic replacement of foyaite, the sulfur content increases most significantly (by 65%). Sulfur is localized in the composition of sulfides, as well as in vishnevite, which replaces nepheline (Figure 4a)  Progress in these reactions led to the formation of complete pseudomorphs of vishnevite over nepheline. At the same time, potassium-released in the reactions microcline → nepheline, microcline → vishnevite, and nepheline → vishnevite-was part of orthoclase. As a boundary between the first (formation of uneven-grained nepheline syenite) and the second (formation of poikilitic foid syenite) stages of metasomatism, we conventionally consider the appearance of orthoclase with single poikilitic inclusions of vishnevite and nepheline.
In order to assess which elements were added or removed from the rock at the second stage of metasomatism, we also used the method of Grant [67]. We compared the mean compositions of the uneven-grained nepheline syenite (protolith) and poikilitic foid syenite ( Figure 16). The line passing closest to the greatest number of elements has a slope of 1, indicating that this alteration occurred without any change in the mass of the rock (Figure 16a). It can be seen that the results are the same as when comparing the unevengrained nepheline syenite-foyaite pair. Sodium oxide, Al 2 O 3 , CaO, MgO, P 2 O 5 , REE 2 O 3 , and F, S, CO 2 were added to the rock during metasomatic alteration, K 2 O, ZrO 2 , Li 2 O, MnO, SiO 2 , and Fe 2 O 3 were removed.
There is an important difference in the behavior of chlorine, water, titanium, and iron. At the first stage of metasomatic alteration, chlorine and water are lost from the rock, and at the second, on the contrary, they are added. As mentioned above, the probable cause of the loss of chlorine and water at the first stage of metasomatism is the complete resorption of amphiboles and EGM. The ionic radius of fluorine (1.33 Å) is very close to the ionic radii of oxygen and hydroxyl (1.40 and 1.37 Å, respectively; [69]). Fluorine can readily substitute for O on the vertices of aluminate or silicate tetrahedra, which promotes its solubility in the silicate melt [70,71], whereas incorporation of Cl (1.82 Å) is restricted to interstitial or network-modifier sites, and this limits their solubility in the melt. Therefore, fluorine is predominantly distributed in the silicate melt and chlorine in the fluid [72]. In alkaline systems, the transition between silicate melts and hydrothermal solution is gradual. Considering this fact and the behavior of halogens, we can draw the following conclusion. At the first stage of metasomatism, the foyaite was exposed to a fluid-saturated melt (or melt-solution?) and at the second, to a hydrothermal aqueous solution.
Intensive replacement of ilmenite and Ti-rich magnetite with titanite, phlogopite, and aegirine is the reason for the loss of titanium and iron at the second stage of metasomatism.
Thus, poikilitic rocks were formed during metasomatic alteration of foyaite as a result of the gain of Na 2 O, Al 2 O 3 , CaO, MgO, SrO, P 2 O 5 , REE 2 O 3 , and volatiles. In the Lovozero massif, this set of elements is characteristic for urtite, i.e., an almost monomineral nepheline rock enriched in SrREE-fluorapatite and other phosphates, as well as loparite-(Ce) [30]. Therefore, the process of formation of poikilitic rocks can be called urtitization of foyaite.
As shown by field investigations and as can be seen in the cross section ( Figure  3a), there is no urtite (or other monomineral feldspatoid rocks) near the contact with the poikilitic rocks. Poikilitic rocks are located either among foyaite or among eudialyte lujavrite. We have not established any signs of metasomatic influence from the lujavrite; on the contrary, the surrounded lujavrite exhibits signs of metasomatic alterations at the contact with poikilitic rocks (Figure 7).
In our previous works, we established that the differentiation of the peralkaline melt during the formation of the rocks of the Lovozero massif proceeded in the following path: lujavrite → foyaite → urtite. In this sequence, the concentrations of sodium, high field strength elements (HFSE), and volatiles increase (Figure 9). With the formation of the rhythm of the Layered complex, a portion of the peralkaline melt evolved from top to bottom. Furthermore, the residual melt could be separated and subsequently crystallized in the form of sheet-like bodies and lenses, as it is observed in the middle part of the Layered complex. Similarly, foyaite lenses in the Eudialyte complex were formed by the segregation of a residual, highly evolved melt [54]. Thus, poikilitic rocks were formed during metasomatic alteration of foyaite as a result of the gain of Na2O, Al2O3, CaO, MgO, SrO, P2O5, REE2O3, and volatiles. In the Lovozero massif, this set of elements is characteristic for urtite, i.e., an almost monomineral nepheline rock enriched in SrREE-fluorapatite and other phosphates, as well as loparite-(Ce) [30]. Therefore, the process of formation of poikilitic rocks can be called urtitization of foyaite.
As shown by field investigations and as can be seen in the cross section (Figure 3a), there is no urtite (or other monomineral feldspatoid rocks) near the contact with the poikilitic rocks. Poikilitic rocks are located either among foyaite or among eudialyte lujavrite. We have not established any signs of metasomatic influence from the lujavrite; on the In evolution of peralkaline magmas, the transition from magmatic melt (minerals crystallize from a silicate melt) to hydrothermal solution (minerals crystallize from an aqueous solution) occurs without critical events. Late magmatic hydrothermal solution did not separate during crystallization (for example, by migration of the solution externally to the crystallizing rock) but remained in the rock and caused extensive autometasomatic changes in previously crystallized minerals. This solution, as shown by experimental studies [73], can be " . . . strongly enriched in dissolved solids (SiO 2 + Al 2 O 3 + Na 2 O, in range 40-50 wt%)". In the sequence lujavrite → foyaite → urtite, the hydrothermal stage could have occurred after the crystallization of urtite. Indeed, in urtite of the Layered complex, almost all nepheline (and feldspar) is replaced by natrolite with very small pyrite inclusions. However, if the initial melt was significantly enriched in volatiles, the hydrothermal stage could begin after the formation of foyaite. It is known that the Eudialyte complex was formed later than the Layered one [29,30] from peralkaline melts, initially rich in volatiles and HFSE.
We propose the following two paths of differentiation of the peralkaline melt depending on the initial content of volatile components: (1) Low concentration of volatiles: lujavrite → foyaite → urtite → natrolitization m a g m a t i c s t a g e hydrothermal stage; (2) High concentration of volatiles: lujavrite → foyaite → urtitization → natrolitization magmatic stage h y d r o t h e r m a l s t a g e.
Based on the previous arguments, we can conclude that poikilitic rocks arose as a result of autometasomatic alteration of foyaite and call this process autometasomatic urtitization.
According to Khomyakov [15], pegmatites in the Layered complex are associated precisely with urtite. This author notes that pegmatites with rare metal mineralization are confined to HFSE-and REE-enriched urtite. Pegmatites form lenses at the contacts of rhythms, while the underlying lujavrite is almost completely unchanged, and the overlying urtite contain numerous apophyses of pegmatite. Khomyakov [15] points out that pegmatites " . . . commonly replace the horizons of such rocks with the formation of pegmatoid beds 0.2-2.5 m thick, which have been traced, with breaks, throughout the entire massif".
Pegmatites in the Layered complex are the end of the evolutionary path lujavrite → foyaite → urtite → (pegmatite) (Figure 17a). Yubileinaya pegmatite is shown in Figure 17b as an example of pegmatite formed with such a sequence. Similarly, the formation of pegmatites associated with poikilitic rocks occurs according to the scheme lujavrite → foyaite → urtitized foyaite → (pegmatite) (Figure 17c). Sirenevaya pegmatite is shown in Figure 17d as an example of pegmatite formed in such a sequence.
The process of urtitization is essential in the concentration of rare elements, because during the crystallization of urtite (but not during urtitization) a large amount of these elements form accessory minerals in the rock itself, such as loparite-(Ce), lueshite, minerals of the eudialyte and lovozerite groups, fluorapatite and stronadelphite, lomonosovite, and murmanite. In the process of urtitization, mainly nepheline is precipitated, and the residual solution is significantly enriched in rare and trace elements (REE, Zr, Ti, P, Be). In addition, the dissolution of foyaite minerals presumably leads to additional enrichment of the residual hydrothermal solution with potassium, lithium, and manganese. Thus, K 2 O, ZrO 2 , Li 2 O, MnO, SiO 2 , and Fe 2 O 3 , which are released during the metasomatic alteration of foyaite, are part of the minerals of pegmatites spatially associated with poikilitic rocks. Table 2 provides a list of minerals most typical of pegmatites associated with poikilitic rocks. Indeed, these pegmatites contain a wide variety of titanosilicates of lithium, potassium, and manganese (e.g., lintisite, kupletskite, neptunite), silicates of lithium and potassium (e.g., tainiolite, polylithionite), phosphates and silicophosphates (e.g., belovite-(Ce) and steenstrupine-(Ce)), REE-rich minerals (e.g., sazhinite-(Ce)) and beryllium (beryllite, bertrandite).  Figure 17. (a) Schematic representation of the evolutionary path lujavrite → foyaite → urtite → (pegmatite), which takes place during the formation of the rhythms of the Layered complex; the color of the arrow corresponds to that in Figure 9; (b) example of pegmatite formed in a sequence lujavrite → foyaite → urtite → (pegmatite); scheme of Yubileinaya pegmatite adapted from [25]; (c) schematic representation of the evolutionary path lujavrite → foyaite → urtitized foyaite → Figure 17. (a) Schematic representation of the evolutionary path lujavrite → foyaite → urtite → (pegmatite), which takes place during the formation of the rhythms of the Layered complex; the color of the arrow corresponds to that in Figure 9; (b) example of pegmatite formed in a sequence lujavrite → foyaite → urtite → (pegmatite); scheme of Yubileinaya pegmatite adapted from [25]; (c) schematic representation of the evolutionary path lujavrite → foyaite → urtitized foyaite → (pegmatite), which takes place during the formation of the poikilitic rocks in the Eudialyte complex; the color of the arrow corresponds to that in Figure 9; (d) example of pegmatite formed in a sequence lujavrite → foyaite → urtitized foyaite → (pegmatite); scheme of Sirenevaya pegmatite adapted from [25].
It is difficult to assess the PT conditions for the formation of poikilitic rocks because the rocks are not in equilibrium. It is reasonable to assume that the upper temperature limit is the temperature of protolith (foyaite) crystallization. According to [54], the foyaite of the Eudialyte complex crystallized at a temperature of about 550 • C. Metasomatic reactions took place during further cooling after the foyaite crystallization; hence, the poikilitic rocks crystallized at temperatures below 550 • C. The Lovozero massif is considered, based on geological data, to be a subvolcanic complex [29], and the maximum total pressure should not exceed 1.5 kbar [14]. The abundance of sulfate-bearing feldspathoids and absence of primary pyrrhotite at poikilitic rock indicate oxidizing conditions above the NNO (nickel-nickel oxide) buffer [74].
Poikilitic rocks are important for studying the composition and evolution of fluids associated with peralkaline rocks. It is known that alkaline massifs, including Lovozero, are abnormally saturated with molecular hydrogen together with other hydrocarbons (HCs) of putative abiotic origin [75], but the reasons for the appearance of these gases are debatable [76]. During the formation of poikilitic rocks, numerous mineral reactions occur with the participation of elements of variable valence (e.g., the replacement of magnesio-arfvedsonite with aegirine). Such reactions have been found to be responsible for the generation of molecular hydrogen [77]. Therefore, further detailed studies of the mineralogy of poikilitic rocks are very important.
As stated in the Introduction, poikilitic rocks are widespread in other peralkaline massifs of the world. Khibiny massif consists dominantly of foyaite (about 70% of the outcrop area) and foidolite (8% of the outcrop area) that intruded into the foyaite massif along the Main Ring fault [32,78,79]. Poikilitic nepheline syenite, rischorrite (consist of poikilitic alkali feldspar with nepheline inclusions), commonly occurs between the rocks of the Main Ring and the foyaite. The researchers believe that rischorrite was formed under the metasomatic influence of foidolite intrusion on the surrounded foyaite [79,80]. Thus, the way of formation of rischorrite in the Khibiny massif is very similar to the manner of formation of the poikilitic rocks in the Lovozero massif. The difference is that in the Lovozero massif, poikilitic rocks were formed as a result of autometasomatic alteration of foyaite, and in the Khibiny massif, such rocks were of contact metasomatic origin.
The naujaite (rock consisting of poikilitic alkali feldspar with sodalite inclusions) of the Ilímaussaq massif has been found to be of magmatic origin. This massif formed from three melt batches, successively crystallizing the augite syenite, a peralkaline granite, and various nepheline syenites (roof series) [81]. The roof series crystallized from the top downwards, forming the succession of pulaskite (nepheline-bearing alkali feldspar syenite), foyaite, sodalite foyaite, and naujaite. The contacts of these rocks grade into each other, but blocks of the uppermost rocks were loosened from the temporary roof of the magma chamber and engulfed by the underlying crystallizing rocks [82].
The formation of a poikilitic texture in peralkaline rocks probably occurs both during (auto)metasomatism and during magmatic crystallization. However, such a texture appears at a late stage of magmatic differentiation, when a gradual transition from magmatic melt to hydrothermal solution occurs.

Conclusions
Poikilitic rocks in the Lovozero massif were formed as a result of autometasomatic alteration (urtitization) of foyaite. There are two stages in this process: nephelinization + albitization and vishnevitization + orthoclasization. At the first stage, uneven-grained nepheline syenite is formed, and at the second stage, poikilitic syenite is crystallized.
The reason for the spatial association of pegmatites and poikilitic rocks is that pegmatites are the final stage of differentiation of the peralkaline melt in path: lujavrite → foyaite → urtite (or urtitization of foyaite with formation of poikilitic rocks) → pegmatite.
The reason for the abundance of rare elements in pegmatites associated with poikilitic rocks is that almost only one nepheline is deposited during urtitization, whereas during the magmatic crystallization of urtite, rare and trace elements form accessory minerals in the rock and are less concentrated in the residual solution.
Thus, the formation of poikilitic rocks is the pre-pegmatite stage of the peralkaline process, and its significant role is the additional concentration of trace elements (rare metals) in the solution, from which pegmatite veins are later formed.