<p>Small portions of pristine melt with diameters of 2 to 50&#181;m are increasingly recognized as a rather common occurrence in high grade metamorphic terranes which experienced melting. Their study delivers crucial chemical information on partial melts at depth. But they are also unique "natural experimental charges" where the behaviour of the silicate melt can be investigated, directly in the natural rocks, under P-T-t conditions which cannot be completely reproduced in the laboratory.</p><p>Each nanogranitoid case study has consistently shown H<sub>2</sub>O-bearing, silica and alkali-rich melt. However, rather than a classic granitoid assemblage consisting mainly of quartz and feldspar(s), on cooling these isolated melt droplets produce a plethora of mineral phases identified via microRaman spectroscopy that are rarely &#8211;or never- observed as rock-forming minerals. Cristobalite (tetragonal) and tridymite (orthorhombic) are often present as SiO<sub>2</sub> polymorphs, and hexagonal kokchetavite as a polymorph of KAlSi<sub>3</sub>O<sub>8</sub>. NaAlSi<sub>3</sub>O<sub>8</sub> occurs as orthorhombic kumdykolite, whereas CaAl<sub>2</sub>Si<sub>2</sub>O<sub>8 </sub>may occur either as monoclinic svyatoslavite or trigonal dmisteinbergite. Two presently unidentified phases have been also recognized via Raman and analysed via electron microprobe. One has the main peak at 426-430 cm<sup>-1</sup> and has the composition of a granitic glass, whereas the second has a main peak at 412 cm<sup>-1</sup> and a variable composition depending on the inclusion in which it occurs. As their main peaks occur in the same region of most tectosilicates, it is likely that they are two new polymorphs of feldspar, to the best of our knowledge never reported before. These polymorphs have been so far identified in inclusions mainly hosted in garnet, zircon and, in one case, sapphirine and trapped under an extremely variable range of metamorphic conditions (from low P migmatites to UHP eclogites) in very different rock types (metagranitoids, metasediments, mafic and ultramafic rocks).</p><p>Microstructures confirm that all of these phases crystallize directly from the trapped melt on cooling, independently of the internal P of the inclusions or the original conditions of melt entrapment. They appear to be the result of metastability in the inclusions, possibly during rapid crystallization of a melt, not caused by rapid cooling but by the peculiar undercooled and supersaturated conditions achieved on cooling by a melt confined in a small cavity (Ferrero & Angel, 2018). According to this possibility, these polymorphs can be regarded as kinetically stabilized, yet possibly thermodynamically metastable, phases as recently proposed by Zolotarev et al. (2019) for dmisteinbergite. A preliminary crystallization experiment on a haplogranitic melt at undercooled conditions however failed to reproduce such phases. Another possibility is that under natural cooling the confined inclusions experience underpressurization, and the system (i.e. the trapped melt) reacts crystallizing phases, i.e. the polymorphs, less dense than their common counterparts. This would result in the decreasing of the P gradient between inclusions and surrounding rock, equivalent to reducing the free energy of the system.</p><p>References</p><p>Ferrero, S. & Angel, R. 2018. JPet 59, 1671&#8211;1700.</p><p>Zolotarev, A.A. et al. 2019. Minerals&#160; 9, 570.</p>
