Compositional and Textural Variations in Hainite-( Y ) and Batievaite-( Y ) , Two Rinkite-Group Minerals from the Sakharjok Massif , Keivy Alkaline Province , NW Russia

Compositional and textural variations in the rinkite group, seidozerite supergroup minerals, batievaite-(Y), hainite-(Y) and close to them titanosilicates from the Sakharjok massif were studied. Statistical analysis allowed for defining two major substitution schemes leading to batievaite-(Y) and cation-deficient titanosilicates forming: Ca2+ + Na+ + F− ↔ + Y3+ + (OH)− and Ca2+ + Na+ ↔ + REE3+. Batievaite-(Y) and other cation-deficient titanosilicates are the earlier minerals formed by solid state transformation of the primary full-cation phase. Hainite-(Y) is a later mineral. It forms rims around earlier titanosilicates, or, less often, separate crystals.


Introduction
Some titanosilicates, such as the rinkite group seidozerite supergroup minerals, are known to concentrate REE.Exploring the geochemical and crystal chemical behavior of Y and REE, we expand our notions on rare elements and their compounds for material science.
Three-layer HOH-blocks are common structural elements of these minerals.They have an inner O-layer formed by Ti, Na, Ca, Mn, and other cations (M O sites, Table 1), with an octahedral coordination, as a rule.There are also two outer heteropolyhedral H-layers that host 6-(7-) coordinated cations of Ca, REE, Y, Zr (M H sites, Table 1) linked to Si 2 O 7 groups.The Ti (+Nb + Zr):Si 2 O 7 = 1:2 is stoichiometric ratio for the rinkite-group minerals [4].* Ideal structural formulae are from [5], except for rinkite-(Y) [6].
In schemes 3-5, the correlation between Na and Ca is positive, unlike the basic pattern.
Since relations between structurally similar hainite-(Y) and batievaite-(Y) are still unclear, the task to analyze compositions of titanosilicates rises.Numerous individuals of those minerals co-existing show that batievaite-(Y), cation-deficient, morphologically, and chemically heterogeneous mineral, occurs in the central part of an individual.Its rims are represented by transparent well-crystallized hainite-(Y).The calcite and analcime aggregates exist of along the boundary between batievaite-(Y) and hainite-(Y) [14] indicates a break in the crystallization of these phases.Also, the paragenesis of analcime and calcite indicates the CO 2 saturation of pegmatitic fluid due to temperature dropping down to 250-100 • C [14].
There are also individuals, where hainite-(Y) rims surround phases with an intermediate chemical composition, which is more or less close to batievaite-(Y) [14].These compositions were not published in [1] and [2].According to geochemical indicators-(La/Nd)n and Y/Dy ratios, the batievaite-(Y) crystallized from CO 2 -rich fluid with high Ca activity prior to hainite-(Y) that formed from relatively F-rich fluid [14].
But, the question is whether batievaite-(Y) could be formed in the same form as we observe?It looks more like a transformational mineral species that was initially formed as a full-cation phase with the same structural elements, and then it was leached by a hydrothermal fluid, according to the Minerals 2018, 8, 458 3 of 12 A.P. Khomyakov principle [15].We did not identify any sign of precursor mineral to batievaite-(Y).In this paper, we try to reconstruct the titanosilicates origin based on compositional and textural variations in the minerals.
Notably, there is the same problem with the rinkite-(Ce)-mosandrite-(Ce) series.Sokolova and Hawthorne give a cautious remark [11]: "It seems likely that mosandrite is a product of alteration of rinkite, in accord with Slepnev [16], but the rarity of mosandrite and lack of textural context do not allow elucidation of any detail of this process".Truly, though the minerals are quite common for the Khibiny and Lovozero alkaline massifs, they do not coexist.At least, such intergrowths are not thoroughly studied, but just mentioned, which is certainly not enough.It is difficult to reconstruct genetic relations for the rinkite-(Ce)-mosandrite-(Ce) series, since rinkite-(Ce) is poorly crystalline and mosandrite-(Ce) is commonly amorphous.Some titanosilicates of the seidozerite supergroup, e.g., murmanite and lomonosovite, have similar elements (HOH-blocks) in their structures.These minerals have cation-exchange properties [17,18] and they can be easily leached.According to structural analysis, during leaching, cations remove from O-layer easier, and Na does so almost completely.
The purpose of this work is to define isomorphic schemes for rinkite group minerals from Sakharjok massif, as well as factors that are responsible for compositional variations and relations between the hainite-(Y) and batievaite-(Y).For that, the authors statistically processed both new and published data on compositions of hainite-(Y) and batievaite-(Y) from the Sakharjok, using Principal Components Analysis (PCA).

Occurrence
The Sakharjok is a pint-sized massif (1.5-2 × 8 km) located in the central part of the Kola Peninsula, NW Russia.The main rocks are alkaline syenites, nepheline syenites, and genetically related pegmatoid schlierens and veins.The geology and petrology of the Sakharjok massif are described in details in [19,20].Alkaline gabbro form minor bodies within the massif.Nepheline syenite affects the alkaline gabbro with the formation of fractures and veins that are filled by a pegmatite material.Pegmatite bodies on the contacts of such contrasting rocks have unusual mineral associations and a complex internal structure with development of aggregates of leucocratic and melanocratic minerals.Three types of mineralization are revealed: rare-earth, beryllium, and chalcophile [7,8,21].The rinkite group minerals occur in leucocratic (major zeolitic) aggregates.
Minerals of the hainite-(Y)-batievaite-(Y) series occur in nepheline syenite pegmatite, Sakharjok massif as euhedral, subhedral or anhedral separate crystals and touching crystals.The size of individuals is up to 2 mm.In hand specimens, the minerals have a milky-white, brownish, pinkish, creamy color, semi-, and non-transparent.They are colorless or transparent in thin chips.SEM analyses allow identifying heterogeneous structure and intergrowths of different mineral phases in one individual.

The study shows that titanosilicates with intermediate composition or/and batievaite-(Y) compose central parts in all mineral intergrowths. Outer rims of intergrowths are formed by hainite-(Y). The contact between batievaite-(Y) and hainite-(Y
) is often marked by calcite and analcime aggregate.We divided the titanosilicates into six types with the following morphological features (Figures 1 and 2): 1. Hainite-(Y) forms separate prismatic crystals or anhedral grains [1], but more often rims around batievaite-(Y) and/or titanosilicates with intermediate compositions (Figure 1a,c).Hainite-(Y) refers to points 1 on PCA plots (Figure 3).
6. "Porous" titanosilicates are a heterogenous material with numerous point defects and cracks.The latter can be filled by a "heavy" material marked light in BSE images (Figures 1b and 2e)."Porous" titanosilicates refer to points 6 on PCA plots (Figure 3). 3. "Homogenous" titanosilicates are rounded areas ("coins") with smooth surface that occur as individuals or merge into larger areas.They occur among "loose" or "porous" titanosilicates and look like relics (Figure 2c,d)."Homogenous" titanosilicates refer to points 3 on PCA plots (Figure 3). 4. Mica-like "layered" titanosilicates are formed by numerous thin parallel layers (Figure 2f)."Layered" titanosilicates refer to points 4 on PCA plots (Figure 3).
6. "Porous" titanosilicates are a heterogenous material with numerous point defects and cracks.The latter can be filled by a "heavy" material marked light in BSE images (Figures 1b and 2e)."Porous" titanosilicates refer to points 6 on PCA plots (Figure 3).
Table 2 provides chemical composition of the different textural types of titanosilicates shown on the Figures 1 and 2, variations of components and their formula coefficients (apfu) calculated based on Si + Al = 4.These analyzes should not be considered representative, although they have been selected in accordance with Figures 1 and 2. They do not fully characterize the texture type, the full pool of analyses for the type is given in Supplementary Materials Table S1.Published data on batievaite-(Y) [2] correspond to some layered and porous varieties.
According to Table 2, the hainite-(Y) has the highest oxide and cations sums in comparison with other textural types of titanosilicates.Also there are significant differences in the component contents for mineral phases with a structure similar to batievaite-(Y).Numerous components and their significant and multidirectional content variations do not allow to find correlation between the texture group and the chemical composition using the "standard methods", in other words, these correlations are latent.Therefore, the PCA was chosen to detect any correlations.

PCA
To identify the most significant factors, and, as a consequence, the factor structure, it is most (entirely) justified to use the principal components analysis (PCA).PCA was applied using the STATISTICA 12.0 program for data processing.Formula coefficients were taken as variables.The procedure of orthogonal rotation "varimax" was applied, the Kaiser's rule became the criterion for estimating the number of principal components (factors).Supplementary Materials Table S1 provides the full set of 79 samples used for PCA.

Results
Seventy-nine mineral samples of the hainite-(Y)-batievaite-(Y) series were divided into six material types, according to their morphology.Factor Analysis (PCA) of coefficients in their formulae provided the following results.According to the Kaiser criterion, any factor that displays dispersion with an eigenvalue less than 1.00 (λj < 1.0) is viewed as trivial and is not retained.These four factors satisfy this criterion (Table 3).

Factor F1
Analysis of the factor loading matrix (Table 3, Figure 3a) shows that the first component (factor F1, 27% of dispersion) binds a group of variables (Ca, Na, F) with a close positive correlation.According to scores plot (Figure 3c), F1 clearly differentiates between compositions of hainite-(Y) (points 1) and decationized, cation-deficient samples (points 2-6).In the right part of the plot (Figure 3a, III quadrant), there is the (Y, Mg, P) group that has a negative correlation with the first group of variables.It reflects a relationship between the decationization on the one hand and the increased Y content and admixture elements (Mg, P) present here on the other hand: (1) Ca 2+ + Na + + F − ↔ + Y 3+ + (OH) − for the main components and Ca 2+ + Na + + Si 4+ ↔ + Mg 2+ + P 5+ or more strictly (2a) Ca 2+ + 2Na + ↔ + 2Mg 2+ or (2b) Ca 2+ + Na + + F − ↔ + Mg 2+ + H 2 O for the admixture elements (Mg).
This trend is best-observed in samples 5 that have loose structure, but neither visible layering, nor porosity (Figure 3c, points 5)."Layered» (points 4) and "porous" (points 6) are divided less clearly, but still distinctly.Thus, the higher the Y, Mg, and P content in a material, the better its destruction is marked.

Factor F2
Variable REE has great loadings both on F1, and on F2 (Figure 3a, II quadrant).It is the opposite of the group (Ca, Na, F) and forms diagonal relationship between the objects.The "homogenous" group is associated with this element on scores plot (Figure 3c, points 3).PCA of a dataset on rinkite-(Ce) (a rare-earth analogue of hainite-(Y)) from the Khibiny massif showed a similar correlation [12].REE are included in the crystal structure according to scheme (3): Ca 2+ + Na + ↔ + REE 3+ .The authors believe that this scheme reflects the processes, partly at least, when rinkite was decationized, as it was altered.Namely, areas with the maximum REE content being subject to alteration, just like areas with the high Y content.However, they are not destructed, but preserved as "homogenous" relics.The difference in the way the material changes, i.e., with destruction or without it, is attributed to admixture elements (Mg, P).When the admixture content is high, the mineral is decationized on both schemes (1) and (2a), (2b) and the material is destructed more intensively.When the admixture content is low, the process follows scheme (3) only, with "homogenous" areas being preserved.
Nb associated with Zr and Ti has the greatest loading on Factor F2 (Table 3, Figure 3a).This reflects similar chrystallochemical features of the elements, in the absence of competition in Ti-dominant M O site (=M O (1)) of the structure.In case of hainite-(Y), the Ti-dominant M O site can be incompletely filled by these elements.In case of graph points with the high Y and particularly REE content, the sum of Ti, Nb and Zr apfu can reach 1.5 (non-stoichiometric composition).

Factor F3
It is difficult to interpret Factor F3 (Table 3, Figure 3b) with great loadings of Mn and Cl.We may see that Mn and Cl are associated with some decationized samples from the groups (points 4-6).
Unlike the (Ca, Na, F) group with a load on F1, (Na, F) group has loads both on F3 and F4 (Figure 3b, IV quadrant), which corresponds to the hainite-(Y) composition on scores plot (Figure 3d).It means that we may clearly associate elements to certain positions in the structure: (Na, F) corresponds with positions in O-layer, and (Ca, Na, F) correspond with all cation sites summarily.Hence, group (Zr, Ti) in II quadrant that is the opposite of (Na, F) explains the location of excessive Ti and Zr.They occur in other sites of O-layer, M O (2), M O (3), according to scheme (4): 4Na + ↔ 3 + Ti 4+ .The greater the number of such mixed positions, the more intensively the original material (protophases) alters.

Discussion
Statistic analysis of composition of structurally close minerals enables us to detect any correlations between components, both obvious and latent.These correlations reflected the behavior of main and admixture elements and helped us to reconstruct the mineral genesis.
According to the PCA, the main isomorphism with a negative relation of Na and Ca (see "Introduction") suggested for the rinkite group minerals is insignificant for minerals of the hainite-(Y)-batievaite-(Y) series from Sakharjok.Contents of Y and/or REE increase not with the growth of sodium, but, on the contrary, due to its reduction in conjunction with calcium and the appearance (formation) of vacancies according to schemes Ca 2+ + Na + + F − ↔ + Y 3+ + (OH) − and Ca 2+ + Na + ↔ + REE 3+ (schemes (1) and (3) in Section 3).
Figure 4 shows the temporal relationships between minerals and mineral phases, in addition to the spatial relationships shown in Figures 1 and 2 Batievaite-(Y) and titanosilicates of intermediate compositions: "eye-like", "layered", "porous" or "loose" titanosilicates are likely to be products of postcrystallization alterations of an early full-cation phase (=protophase).The protophase crystallized in conditions of highly active CO 2 , while contents Ca, Na, and Y were high enough to produce a full-cation phase and a content of F was low.
Statistic analysis of composition of structurally close minerals enables us to detect any correlations between components, both obvious and latent.These correlations reflected the behavior of main and admixture elements and helped us to reconstruct the mineral genesis.
According to the PCA, the main isomorphism with a negative relation of Na and Ca (see "Introduction") suggested for the rinkite group minerals is insignificant for minerals of the hainite-(Y)-batievaite-(Y) series from Sakharjok.Contents of Y and/or REE increase not with the growth of sodium, but, on the contrary, due to its reduction in conjunction with calcium and the appearance (formation) of vacancies according to schemes Ca 2+ + Na + + F − ↔ □ + Y 3+ + (OH) − and Ca 2+ + Na + ↔ □ + REE 3+ (schemes (1) and (3) in Section 3).
Figure 4 shows the temporal relationships between minerals and mineral phases, in addition to the spatial relationships shown in Figures 1 and 2 Batievaite-(Y) and titanosilicates of intermediate compositions: "eye-like", "layered", "porous" or "loose" titanosilicates are likely to be products of postcrystallization alterations of an early full-cation phase (=protophase).The protophase crystallized in conditions of highly active СО2, while contents Са, Na, and Y were high enough to produce a full-cation phase and a content of F was low.2, an.455, 477).There were admixture elements (Mg, P) as well.2, an.455, 477).There were admixture elements (Mg, P) as well.Schemes (2a) and (2b) (Section 3) indicate that magnesium could be present in the cationic positions of the calcium-sodium in protophase in larger content than in hainite-(Y).
It is possible that the inhomogeneity of the protophase could make it unstable to decationization when fluid regimes change.As for Y, we may recall that, e.g., hainite-(Y) from the Sakharjok massif differs from hainite-(Y) from other deposits, not only by its increased Y content, but also by a deficient analysis sum [13].
Postcrystallization alterations in the protophase with various compositional deviations from stoichiometry are related to the different morphology of studied variations of titanosilicates.There are two distinctive main types of alterations; we call them destructive and non-destructive.The leaching of high yttrium areas should be considered as the first type.The products of such destructive alteration are "eye-like", "layered", "porous", or "loose" textural variations of titanosilicates.Due to protophase inhomogeneity different textural variations showed within same mineral grain.PCA of the chemical
. Statistics data are consistent with geochemical indicators in showing batievaite-(Y) and hainite-(Y) to be minerals of different generations (Figure 4) and batievaite-(Y) to be prior to hainite-(Y).
. Statistics data are consistent with geochemical indicators in showing batievaite-(Y) and hainite-(Y) to be minerals of different generations (Figure 4) and batievaite-(Y) to be prior to hainite-(Y).

Figure 4 .
Figure 4. Relationship scheme of the hainite-(Y)-batievaite-(Y) series.The protophase composition was highly inhomogeneous.It was rich in REE, Y, Ti, and Nb that sometimes increased stoichiometric values (1 apfu).For hainite-(Y) as well for batievaite-(Y) apfu Ti + Nb must be 1 in Ti-dominant M O site (=M O (1)) of the structure.But, for some studied samples, this value is more than 1 (Table2, an.455, 477).There were admixture elements (Mg, P) as well.Schemes (2a) and (2b) (Section 3) indicate that magnesium could be present in the cationic positions of the calcium-sodium in protophase in larger content than in hainite-(Y).It is possible that the inhomogeneity of the protophase could make it unstable to decationization when fluid regimes change.As for Y, we may recall that, e.g., hainite-(Y) from the Sakharjok massif differs from hainite-(Y) from other deposits, not only by its increased Y content, but also by a deficient analysis sum[13].Postcrystallization alterations in the protophase with various compositional deviations from stoichiometry are related to the different morphology of studied variations of titanosilicates.There are two distinctive main types of alterations; we call them destructive and non-destructive.The leaching of high yttrium areas should be considered as the first type.The products of such destructive alteration are "eye-like", "layered", "porous", or "loose" textural variations of titanosilicates.Due to protophase inhomogeneity different textural variations showed within same mineral grain.PCA of the chemical

Table 1 .
Ideal structural formulae for rinkite group minerals.

Table 2 .
Chemical compositions (wt %) and formula coefficients (apfu) for different morphological groups of titanosilicates presented in Figures1 and 2. Variations correspond to 79 analyses of the rinkite group minerals from the nepheline syenite pegmatite, Sakharjok massif, Kola Peninsula.

Table 3 .
Principal Components Analysis (PCA) results of the mineral composition.