29Si NMR and Raman spectroscopic studies demonstrate that fusion of crystalline orthosilicates and metasilicates produces melts more polymerized than their precursor crystals. Forsterite, for example, consists of 100% Q0 species, whereas its melt consists of ~50 mol% of Q1 species (Q = a Si tetrahedron and the superscript indicates the number of bridging oxygen atoms in the tetrahedron). Polymerization during melting can be rationalized from an energetics perspective. Si-NBO-M moieties of Q species are more susceptible to librational, rotational, and vibrational modes than are Si-BO-Si moieties (NBO = non-bridging oxygen; BO = bridging oxygen; M = counter cation). Thermal agitation activates these additional modes, thus increasing the CP and free energy of melts. The reaction of Qn to Qn+1 species during melting eliminates Si-NBO-M moieties and produces Si-O-Si moieties that are less susceptible to the additional modes, thereby minimizing the CP of melts. By decreasing the abundances of Q0, Q1, and Q2 species in favor of Q3 and Q4 species, melts become more stable. In the absence of polymerization, melting temperatures of minerals would be appreciably greater than observed.
Polymerization involves formation of Si-O bonds, which are strongly endothermic (Si-O bond dissociation is ~798 kJ/mol). The large heats of fusion (ΔHf) of orthosilicates result primarily from polymerization reactions during melting (ΔHf of forsterite, fayalite, and tephroite are ~142, ~92, and ~90 kJ/mol). The fusion of metasilicates and sorosilicates (e.g., pyroxenes and melilites) involves endothermic polymerization and exothermic depolymerization reactions, although the former dominates. These reactions tend to negate each other during melting, yielding less positive ΔHf values than observed for orthosilicate fusion (e.g., ΔHf of enstatite, diopside, pseudowollastonite, and åkermanite are ~73, ~69, ~57, and ~62 kJ/mol). Where polymerization and depolymerization reactions are absent ΔHf is low and is due mostly to disordering during melting (e.g., ΔHf of cristobalite iŝ8.9 kJ/mol).
Experimental evidence indicates that ferric iron is present as a negatively charged oxy-anionic complex in melts (e.g., [FeO2]1–) so that oxidation of Fe2+ should proceed according to: + 1O2 + .
Free oxygen (O2–), a by-product of polymerization reactions, drives the reaction to the right. Midocean ridge basalts (MORBs) consequently should be more oxidized than their source (e.g., lherzolites) or their residues (e.g., harzburgites). Extraction of melt from the upper mantle and deposition in the crust should produce a crust more oxidized than its upper mantle source. Production of O2– during melting and its presence in alkali-rich magmas also explains the alkali-ferric iron effect.