Chevkinite-group minerals in selected intrusions of the Mazury Complex, North-Eastern Poland: insights into the formation of a titanite-like phase by hydrothermal alteration

Chevkinite-group minerals forming large and common (up to 0.03 vol%) accessory phases in monzodiorites and granodiorites from the Mesoproterozoic anorthosite-mangerite-charnockite-granite (AMCG) suite intrusions, Mazury Complex, north-eastern Poland, range from pristine magmatic types to hydrothermally altered varieties. The unaltered phase is perrierite-(Ce), with the uncommon feature of having Al dominant in the C site. Hydrothermal alteration of the perrierite-(Ce) followed two main trends: one shows depletion in Ca, Fe, Si, Al and Mg, and increasing Ti contents; the other shows increases in Si, Ti and Ca and decreases in light rare-earth elements (LREE), Y and Mg, at about constant Fe content. The second trend resulted in the formation of a phase compositionally similar to titanite. Result of chemical analyses show that the transition from perrierite-(Ce) to the titanite-like phase is sharp; an electron back-scatter diffraction (EBSD) study shows the titanite- like material to be amorphous. The hydrothermal alteration of a chevkinite-group mineral to titanite has commonly been reported in natural sequences but this is the first record of the identification of a titanite-like phase made on the basis of a structural analysis.


Introduction
The chevkinite-group minerals (CGM) are increasingly being recognized as important accessory phases in a wide range of igneous and metamorphic rocks. They are known from hundreds of localities and undoubtedly occur unreported in hundreds more . The group comprises twelve members, the generally accepted formula being A 4 BC 2 D 2 (Si 2 O 7 ) 2 O 8 , where A = REE, Ca, Sr, Th; B = Fe 2+ ; C = Fe 2+ , Fe 3+ , Ti, Al; D = Ti. However, varieties rich in Zr, Nb, Mn and W are also known. During our ongoing studies of Mesoproterozoic intrusions in NE Poland, we have found CGM which are of particular interest. First, the minerals are unusually large and common and are also texturally very variable. Second, they show the uncommon feature of Al dominance in the C site.
Like other REE-bearing accessory minerals, CGM are prone to alteration by hydrothermal fluids. There is a wide range of alteration mechanisms and products, mainly related to the nature of the ligands in the fluids, as reviewed by Macdonald et al. (2019). In the Mazury rocks, we report two modes of hydrothermal alteration, one of them representing an intermediate stage in the breakdown of a CGM to a titanite-like phase, which has been amenable to investigation by electron probe microanalysis (EPMA) and electron backscattered diffraction (EBSD).

Geological setting
The rocks studied here have been sampled from three boreholes, Krasnopol PIG-6, Pawłówka PIG-1A, and Filipów IG-1 (Fig. 1) located in the crystalline basement of northeastern Poland. The intrusion are built of rapakivitype granitoids (sensu Haapala et al. 2005) that is are the A-type granites of large batholiths, which include varieties with the rapakivi texture (sensu Vorma 1976;Happala and Rämö 1992;Rämӧ and Haapala 1995) and associated intermediate and mafic rocks. It belongs to a widespread set of rapakivi-type granitoid intrusions in northern Poland, which together with anorthositic massifs (Suwałki, Sejny and Kętrzyn) form the so-called Mazury Complex (Kubicki and Ryka 1982Ryka , 1993. The complex forms a belt of granitoid and associated mafic (anorthositic) and intermediate bodies intruded within the older metamorphic basement of the Eastern European platform and extends over 300 km from the Baltic Sea (near Elbląg in northern Poland) to the Kabeliai and Veisiejai complexes in eastern Lithuania (Skridlaite et al. 2003). The boundaries between individual intrusions in the Mazury Complex in Poland are poorly defined because their structure is based mainly on a few boreholes and geophysical data (Ryka and Sawicki 1980;Kubicki and Ryka 1982;Wybraniec 1999).
Granitoid intrusions of the Mazury Complex are examples of a widespread series of rapakivi granitoids (Happala and Rämö 1992;Rämӧ and Haapala 1995;Bagiński et al. 2001;Duchesne et al. 2010). These intrusions form massifs of varying sizes, often occurring with anorthosite intrusions, both with features typical of anorogenic magmatism. The crystallization ages of the rapakivi intrusions are in the range 1. 81 -1.47 Ga (Happala andRämö 1992, 1999;Rämӧ and Haapala 1995;Shumlyanskyy et al. 2017). The crystallization ages of rapakivi granitoids from the Mazury Complex range from 1548 to 1499 Ma, some 50 Ma (Dorr et al. 2002;Wiszniewska et al. 2007). Two samples of quartz monzonite from the Gołdap and Bartoszyce boreholes yielded U-Pb zircon ages of 1502 ± 2 Ma and 1499 ± 4 Ma, respectively (Claesson et al. 1995). The crystallization age of the rapakivi granitoids at Filipów (Fig. 1) is estimated at 1512 ± 17 Ma based on results of U-Th-Pb zircon SIMS (secondary ion mass spectrometry) analyses (Wiszniewska et al. 2016). The crystallization ages of granites from the Kabeliai and Veisiejai intrusions (in eastern Lithuania) fit the age range of the Mazury Complex, e.g. the 1505 Ma age of the Kabeliai granites in southeastern Lithuania (Sundblad et al. 1994).

Analytical methods
A total of 64 fresh samples were taken from the Krasnopol PIG-6 drill core from a depth range 621 -1500 m. Chevkinite have also been identified in a few granitoid rock samples taken from the Filipów IG-1 and Pawłówka PIG-1A boreholes. The selection of CGM for microprobe study was performed using reflected-light optical microscopy (NIKON LV100) and electron microscopy with control of the qualitative chemical composition of minerals by means of an energy dispersive spectroscopy (Zeiss Sigma VP equipped with EDS Bruker XFlash 6|10 spectrometer). Identification of minerals under the reflected-light microscope was possible due to their large grain sizes (up to 0.5 mm), and by clear visible optical properties (reflectance slightly less than 20% and the presence of visible brown internal reflections).
Prior to EBSD examination, the sample was additionally polished (using a vibrating polisher) for 8 h in a diamond suspension with a grain diameter of 1/4 µm. The sample was covered with a 4.0 nm carbon layer for better surface discharge. EBSD patterns were collected with a Zeiss Auriga electron microscope equipped with a Brukere − FlashHR + detector with integrated ARGUS imaging device. The sample was tilted to 70° using the dedicated stage (tilt about sample X axis) for an optimal EBSD signal with working distance set to 19.15 mm, while the detector tilt angle was 2.42° and sample to detector distance of 15.02 mm. The experiments were carried out using an electron beam with an energy of 10 kV. Image tilt correction was used on the Zeiss SmartSEM software and no image rotation was applied. A single EBSD pattern of 320 × 240 resolution was recorded from the 0.86 × 0.86 µm area during 130 ms. The system was calibrated in Bruker ESPRIT 2. The pattern centre (PC), in Bruker fractional coordinates, was measured as: PC x = 0.506, PC y = 0.252 with a pattern aspect ratio of 1.33 (width/height). The hit rate of indexed pixels was 32.4%; the remaining part of the analysed sample was amorphous.

The host rocks
Petrographic descriptions of CMG-bearing samples from the Filipów IG-1 and Pawłówka PIG-1A cores can be found in Bagiński et al. (2001). The Krasnopol PIG-6 core monzodiorites and granodiorites hosting the CGM are coarseand medium-grained with a well-developed porphyritic texture with plagioclase phenocrysts reaching up to 5 cm. The main minerals are plagioclase (An46 -An29), potassium feldspar, quartz, biotite [classified as phlogopite, with Mg/(Mg + Fe tot ) ratios of 0.52 -0.57], amphibole [classified as hastingsite, with Mg/(Mg + Fe tot ) ratios of 0.57 -0.61], clinopyroxene (classified as diopside, with Wo: 47-51, En: 33-37, Fs:14-19), Fe-Ti oxides and Fe-Cu sulphides (pyrrhotite, pyrite and chalcopyrite). The Krasnopol core is dominated by rocks with high contents of biotite and amphibole. Clinopyroxene also occurs over the intervals 1002.8 -1308.7 mand 1380.5 -1500 m. The colour of the granitoids is variable; in the upper part of the section it is pink-grey, below the depth of 1112.8 mKrasnopol granitoids lack the pink colour, except at 1263.3 -1308.7 m, where the granitoids are again reddish.
A notable feature of the rocks is the presence of finegrained enclaves. These are usually dark in colour. The overall size and form of the enclaves are difficult to define due to the fact that the rocks come from drill core; however, they commonly occur with monzodiorite/granodiorite at various depths. Compositionally they are monzonite, monzodiorite or monzogabbro. Usually they show non-oriented textures, are equigranular and occasionally contain large (up to 3 cm) K-feldspar or/and plagioclase (up to 0.5 cm) megacrysts. Sometimes they show a magmatic foliation, defined by mafic minerals. Enclaves are composed of the same minerals as the surrounding rocks, differing only in their greater abundance of plagioclase and mafic minerals. The chemical composition of enclave minerals is also similar to that of the host rock. The rocks matrix consist of 35-45 vol% plagioclase (An 48-24 ), 20-30 vol% phlogopite, with Mg/ (Mg + Fe tot ) ratios of 0.53 -0.64, 0-30 vol% clinopyroxene diopside, with Wo: 43-51; En: 33-37; Fs: 12-23 and 0-25 vol% hastingsite, with Mg/(Mg + Fe tot ) ratios of 0.56 -0.63. Quartz and K-feldspar are interstitial and subordinate. Common accessory minerals are magnetite and ilmenite, apatite, zircon and titanite.
The contacts of the enclaves with granodiorite are partly diffuse, lobate to crenulate. Large feldspar megacrysts are observed to cross the contacts between enclave and granodiorite, in addition to occurring inside them. The plagioclase phenocrysts show poikilitic rims with mainly clinopyroxene and ilmenite inclusions. K-feldspar phenocrysts also have poikilitic rims with inclusions of plagioclase, clinopyroxene, biotite, apatite and ilmenite. The textural relationships are consistent with the mixing/mingling of two melts of different composition. Sample details are given in Table 1.

Chevkinite-group minerals
The CGM are common in samples of monzodiorites and granodiorites from Krasnopol PIG-6 drillcore (Nejbert et al. 2020). From 3 to 7 grains with a diameter of more than 100 μm were identified in a standard 6 cm 2 thin section. In addition, numerous small CGM inclusions, with sizes well below 10 µm, were observed within feldspars and biotite. The number of such inclusions is difficult to estimate; generally during the study of standard thin sections, from 20 to 30 such grains were identified using scanning electron microscopy-energy dispersive X-ray spectroscopy (SEM-EDS) methods. In samples identified as the richest in CMG, and/ or titanite-like phases formed after this mineral, they may constitute a maximum up to 0.03 vol%. The CGM are present in a wide variety of forms but the two most common are prismatic and platy. The prismatic types are usually between 50 and 100 μm long, which in our experience is unusually large for CGM (other than in pegmatites). They are bright [high average atomic number ( Z ) contrast] on backscattered electrons (BSE) images and fairly homogeneous (Fig. 2a,  b). The crystal in Fig. 2a is enclosed in magnetite showing ilmenite exsolution lamellae, those in Fig. 2b in alkali feldspar, evidence that they were of magmatic crystallization. The spear-shaped crystal shown in Fig. 2c is patchily zoned, with high-Z material forming a residual part of the core and an extreme tip. The platy types are larger, up to 500 μm across, and usually patchily zoned from bright to vary dark. The crystal shown in Fig. 2d is unusual in being bright, but with a small, low Z , core area and a dark rim which is associated with a patchily zoned grain. The core of the subhedral crystal in Fig. 2e is very complex and patchy, with a higher Z mantle where oscillatory zoning is just visible. The patchy core is more metamict than the rim due to a higher content of Th. The patchily zoned CGM at the left-hand edge of the crystal in Fig. 2f has been replaced towards the rims by an aggregate of a REE carbonate.
Textural features typical of magmatic crystallization are preserved in some crystals, such as the subhedral-euhedral form of certain prismatic crystals (Fig. 2a, b), crystals intergrown with magmatic minerals (Fig. 2a, b), and those which display oscillatory zoning (Fig. 2e). Mainly, however, magmatic zoning has been overprinted by hydrothermal alteration, recognizable by the development of patchy zones of lower Z , as opposed to unaltered crystals which are invariably bright (high Z ) on BSE images.
The question arises as to why the degree of alteration was so variable, even within single thin sections. It seems that the unaltered crystals were protected by being enclosed within plagioclase crystals, e.g. in Fig. 2b. In contrast, the altered grains tended to occur along intergranular boundaries and were presumably involved in extensive intergranular fluid flow.
The CGM were also identified in granitoid samples from the Filipów IG-1 and Pawłówka PIG-1A boreholes. These occurred as large grains (up to about 100 μm), associated with Fe-Ti oxide aggregates and pyrrhotite, and fine, up to 10 μm, lamellar perrierite-(Ce) crystals appearing as fine, regularly distributed inclusions in plagioclase. In several large plagioclase crystals, a parallel arrangement of fine perrierite grains was observed, which were also arranged parallel to the fine Fe-Ti oxides. In the samples from the Filipów IG-1 and Pawłówka PIG-1A boreholes, all the larger CGM grains were highly altered and broken down into titanite-like phases, allanite, REE-fluorocarbonates, and Ca-Fe carbonates. Only the small-sized lamellar perrierite-(Ce) inclusions within plagioclase were relatively poorly altered as shown by their high brightness on the BSE images.
Chondrite-normalized REE plots (Fig. 5) show decreases from La to Sm, with a significant variation in La N /Sm N values (6 -24), and strong positive Eu anomalies (Eu/Eu* 1.7 -5.1, with two outliers at 9.5 and 10.8). However, the ranges may partly reflect Sm, Eu and Gd values close to the EPM detection limits. Positive Eu anomalies might be a result of plagioclase breakdown during the observed sericitisation process (Zozulya et al. 2020).

Hydrothermal alteration trends
Compositions of altered CGM are given in Table 3 and Tables S1-S3. Previous studies of the hydrothermal alteration of CGM, summarized in Bagiński et al. (2015) and Macdonald et al. (2019) have shown that invariable compositional features of the process are low analytical totals and non-stoichiometric formulae. These features also mark the Mazury Complex CGM. Analytical totals are always less than 95 wt% and as low as 85 wt%, the low values presumably reflecting water absorbed into the structure and/ or the presence of vacancies. Sums of cations are as low as 12.2 apfu. Data for altered crystals initially follow a trend of decreasing Fe contents at about constant Ca content until at 3 wt% CaO they diverge into two trends. One, an initial Fe-depletion trend, varies between specimens; in Kr-6-41, for example, a slight increase in CaO reverses at FeO* 4.5 wt%, followed by rapid Fe and Ca depletion to < 2 wt% CaO, while in Filipów-1, the earliest alteration phase resulted in Ca depletion before the trend involved strong depletion in Ca and Fe to ~ 2 wt% CaO (Fig. 6a). Other features of this part of the trend include decreases in Si, Al and Mg contents and increasing Ti. In Kr-6-41 REE   Tables S1-S3 abundances decrease along the trend, whereas in Filipów-2, REE increase along it. The main feature of compositions along the second trend is the increasing contents of Ti, Ca (Fig. 6b) and Si at about constant Fe. Other features are the increasing Al and decreasing contents of REE, Y, Mg, Sc and Zr. The REE 2 O 3 contents, for example, which total ~ 36 -38 wt% in the unaltered CGM, are < 5 wt% in the most altered samples: this is discussed in more detail below.
There are minor variants on the alteration schemes described above. For example, in Kr-6-19 two low Ti-Ca analyses (Table 3 and S1; analyses 53, 54) have high Fe (12.61, 11.21 wt% FeO*) and low REE (20.08 wt% REE 2 O 3 ). In Fig. 6a, b for unusually Fe-rich altered analyses plot with the unaltered types (arrowed). Two other low Ti-Ca analyses (30, 34) have low Fe (0.90, 0.93 wt% FeO*) and high REE (44.62, 48.96 wt% REE 2 O 3 ). All these analyses have high F contents (2 -5 wt%) and it is possible that this has resulted in some unusual element redistribution. Alternatively, the compositional features may be related to the nature of other accessory minerals forming during the alteration.
Extrapolation of Ti, Si and Ca values in the second trend coincide with the composition of titanite (Fig. 6b). The two analyses closest to titanite are from Kr-6-19 (Table S1; analyses 63, 64). The average formula, calculated on the basis of 5 oxygens, is [Ca 0.88 Σ(La-Nd) 0.04 Sr 0.02 ) 0.94 (Ti 0.75 Al 0.20 Fe 2+ 0.03 ) 0.98 Si 1.09 ] , with the sum of cations = 3.04, i.e. the phase is compositionally close to titanite, CaTiOSiO 4 , with some replacement of Ti by Al. The Al levels are significant (≤ 5 wt% Al 2 O 3 ) but not exceptional: Seifert and Kramer (2003), for example, recorded values of Al 2 O 3 up to ~ 15 wt% in titanite from lamprophyres of the Variscan orogeny in Germany. In comparison, Jiang et al. (2006) reported 3.99 wt% Al 2 O 3 in titanite in an allaniteilmenite-titanite mantle on altered chevkinite-(Ce) in the Shuiquangou syenite, China. Additionally Vlach et al.  Tables S1-S3. Perrierites from granulites from India and Antarctica plotted for comparison  ,  Table 1). Values are plotted in apfu  Table 2. Normalizing factors after Sun and McDonough (1989) (2014) suggested that allanite growth is favored by relatively high oxidizing conditions.

Structural breakdown of perrierite
It appears that the Mazury Complex CGM have recorded intermediate stages in the alteration of a CGM to a titanitelike phase. We have carried out detailed EPMA and EBSD analyses in an attempt to determine the compositional and structural nature of the titanite-like phase. Five compositional profiles were made across the crystal in sample Kr-6-41 shown in Fig. 7a; the data are given in Table S2. Compositional variations along one profile, as an example (profile E-F), are shown in Fig. 8. The crystal has been divided into three zones, the bright, transitional and dark zones. The bright zone is slightly altered perrierite-(Ce), brighter spots pointing to local concentrations of Ti (Fig. 9). Within the transitional zone, there are increases in the contents of SiO 2 , TiO 2 , and ThO 2 , and decreases in FeO*, SrO, BaO, which are continued into the dark zone. Most marked is the sharp drop into the dark zone of LREE (La -Eu), from ΣREE 2 O 3 ≥ 35 wt% to < 5 wt%. While the most Ti-rich analyses have a titanite-like composition, there is a notable gap between 5 and 14 wt% CaO, the gap coinciding precisely with the transition zone. Thus the transition from perrierite-(Ce) to titanite-like compositions was sharp. The gap does not exist in all crystals in the data set; for example, there is a continuous series of compositions in sample Kr-6-19 (Table S1). Judging from the patchiness of the partially altered perrierite-(Ce), the alteration reached different parts   Fig. 7b is of particular interest because it provides evidence of the nature of the hydrothermal fluids promoting the alteration to the titanite-like phase, the presence of REE-carbonates and pyrrhotite points to the presence of CO 2 and S in the fluids.
A representative electron backscatter pattern of perrierite-(Ce), crystallizing in P2 1 /c (a standard setting of commonly used, P2 1 /a, the difference is only in the order of the crystallographic axes, both choices are symmetrically equivalent) is presented together with indexed, simulated pattern in Fig. 12.
Compositional data in Fig. 8, between analyses 22 -28 correspond to the crystalline section of perrierite-(Ce) obtained from EBSD (red zone in Fig. 10a). Perrierite-(Ce) breaks down to an amorphous phase of titanite-like composition. Figure 10b shows the crystalline areas in brighter colour. The interior of the analyzed surface, i.e. the titanite-like phase, does not diffract electrons, meaning that it is amorphous.

Crystallization conditions
The perrierite-bearing rocks from the Mazury Complex crystallized from broadly mafic to trachyandesitic melts. This is consistent with the findings of Macdonald and Belkin b a Fig. 6 a Two Ca-Fe depletion trends exemplified by samples Kr-6-41 and Filipów-1. b The Ca-Ti enrichment trend as shown by Kr-6-19 and Kr-6-59. Data from Tables S1 and S2. Also plotted is the composition of titanite formed by breakdown of chevkinite-(Ce) in the Shuiquangou syenitic intrusion, China (Jiang 2006) Table S2). b BSE image of sample Kr-6-19, showing partially replaced perrierite-(Ce), the titanite-like phase (Ttn*), pyrrhotite (Pyh), REE carbonates (REEcarb) and britholite-(Ce) (Bri-Ce). Amp, amphibole: Bt, biotite: Cal, calcite: Qz, quartz. D -dark, T -transitional, L -light (2002), who showed that in igneous suites perrierite tends to form in rocks of intermediate and mafic composition (Fig. 11) while chevkinite forms in more evolved melts, such as granites, syenites and extrusive equivalents. Information on the P-T crystallization conditions comes from the associated charnockitic rocks. Based on pyroxene compositions, (Bagiński and Krzemińska 2005;Bagiński et al. 2006) estimated that the charnockites crystallized at a maximum range of ca. 850 -950 °C and under variable polybaric pressure conditions. This temperature range is consistent with the likely crystallization temperatures of intermediate melts.
Note, however, that the estimates assumed limited thermal influence from the coeval anorthosite bodies, intruded during the same AMCG magmatism stage.
Crystallization of the perrierite-(Ce) in the granulite facies is consistent with the occurrence of dominant Al in the C site. Al-dominant perrierite has been reported from several areas of granulitic facies rocks in Antarctica and India (Harley 1994;Hokada 2007;Belkin et al. 2009) (Fig. 4). There is also experimental evidence that high-pressure appears to favour the entry of Al into the CGM structure: Green and Pearson (1988) crystallized Aldominant perrierites-(Ce) from REE-enriched andesitic and rhyolitic melts at P = 20 kbar, T = 1050 °C.

Effects of hydrothermal activity
Studies of the hydrothermal alteration of chevkinite-(Ce) from Russia and Scotland (Bagiński et al. 2015) and Brazil   Fig. 9 Ti Kα x-ray intensity map for crystal shown in Fig. 7a. The colour range from blue to red indicates increasing concentrations of Ti  (Vlach and Gualda 2007) found that progressive alteration led initially to Ca enrichment at constant Ti and then to rapid Ca depletion with strong Ti enrichment. REE behaviour was more variable, being enriched in some crystals, depleted in others. The Ca, REE, Si and Fe were removed in the fluid phase. The types of alteration shown in Fig. 6a are, however, broadly similar to that recorded by Bagiński et al. (2015). The second trend involved enrichment in Ca, Ti and Si and strong depletion in REE and Y. The breakdown of CGM to a titanite-like phase in a wide range of natural rocks (e.g. gabbroids, syenites, pegmatites and metasomatites) has been recorded by, inter alios, Jiang (2006), Prol-Ledemsa et al. (2012), Macdonald et al. (2015a, b), Bagiński et al. (2016) and Spiridonov et al. (2019). In none of these studies was the identification of the titanite-like phase made on the basis of a structural analysis. The Krasnopol case raises the question as to whether the other examples are truly crystalline. The alteration sequences and products were highly variable between crystals; it may be inferred that the variations were at least partly related to variable compositions of the hydrothermal fluids. Those described by Macdonald et al. (2015b) and Bagiński et al. (2016) were alkaline and contained significant CO 2 and F, whereas those in Macdonald et al. (2015a) were thought to have been acidic, with significant CO 2 but poor in F. The crystal shown in Fig. 7b is of particular interest because it provides evidence of the nature of the hydrothermal fluids promoting the alteration at Krasnopol. The presence of REE-carbonates and calcite (Figs. 2f,7b) shows that the fluids were CO 2 -bearing, while the pyrrhotite points to a component of S. The high F contents (up to 5 wt%) of some altered CGM and its presence [wt%] Fig. 12 a Simulated, indexed electron backscatter pattern of perrierite-(Ce). b Corresponding, representative experimental pattern in the REE-carbonate further suggest that the fluids were F-bearing. We can speculate that the titanite-like phase would eventually have crystallized to titanite. Possibly the temperatures during hydrothermal alteration were simply too low to form it but we note that titanite can form at very low temperatures: in an extreme case, it has been recorded as a result of microbial alteration of volcanic glass (Izawa et al. 2019). Alternatively, it may have been a function of the duration of the alteration episode.

Conclusions
Perrierite-(Ce), a member of the chevkinite-group of minerals, is present in monzodiorites and granodiorites of the Mesoproterozoic Krasnopol intrusion, formed under granulite facies conditions. Consistent with a high-pressure origin, Al is dominant at the C site. The perrierite-(Ce) was extensively altered by hydrothermal fluids, which were CO 2 and F-bearing, with a component of S. At least two alteration schemes, involving different element mobilities, have been identified, perhaps indicating the presence of more than phase of alteration. One scheme resulted in an amorphous phase compositionally close to titanite.