Contrib Mineral Petrol (1994) 117:1-14 9 Springer-Verlag 1994 B. O. Mysen 9 J. D. Frantz Silicate melts at magmatic temperatures: in-situ structure determination to 1651 ~ C and effect of temperature and bulk composition on the mixing behavior of structural units Received: 16 April 1993 / Accepted: 4 January 1994 Abstract The abundance of coexisting structural units in K-, Na-, and Li-silicate melts and glasses from 25 ~ to 1651~ has been determined with in-situ micro- Raman spectroscopy. From these data an equilibrium constant, Kx, for the disproportionation reaction among the structural units coexisting in the melts, Si2Os(2Q3)<=~SiO3(Q2)+SiO2(Q4), was calculated (Kx is the equilibrium constant derived by using mol fractions rather than activities of the structural units). From In Kx vs 1/T relationships the enthalpy (AHx) for the dispro- portionation reaction is in the range of -30 to 30 kJ/ mol with systematic compositional dependence. In the potassium and sodium systems, where the dispropor- tionation reaction shifts to the right with increasing temperature, the AHx increases with silica content (M/Si decreases, M =Na, K). For melts and supercooled liq- uids of composition Li20 - 2SiOa (Li/Si = 1), the AH~ is indistinguishable from 0. By decreasing the Li/Si to 0.667 (composition LS3) and beyond (e.g., LS4), the dis- proportionation reaction shifts to the left as the temper- ature is increased. For a given ratio of M/Si (M = K, Na, Li), there is a positive, near linear correlation between the AHx and the Z/r 2 of the metal cation. The slope of the AHx vs Z/r 2 regression lines increases as the system becomes more silica rich (i.e., M/Si is decreased). Activi- ty coefficients for the individual structural units, % were calculated from the structural data combined with liq- uidus phase relations. These coefficients are linear func- tions of their mol fraction of the form ln?i = a lnXi + b, where a is between 0.6 and 0.87, and X i is the mol frac- tion of the unit~ The value of the intercept, b, is near 0. The relationship between activity coefficients and abun- Bjorn O. Mysen1([~) 9 John D. Frantz Geophysical Laboratory, 5251 Broad Branch Rd., NW, Washington, DC 20015, USA Also member of NSF-sponsored Science and Technology Center for High Pressure Research (CHiPR) Editorial responsibility: Ti. Grove dance of individual structural units is not affected by temperature or the electronic properties of the alkali metal. The activity of the structural units, however, de- pend on their concentration, type of metal cation, and on temperature. Introduction Phase relations, thermodynamic, and physical proper- ties of melts, fluids and crystalline phases comprising magmatic systems provide data required to describe magmatic processes. These properties in turn can be characterized in terms of the structure of the magmatic component phases provided that the link between struc- ture and properties has been established at the tempera- ture and pressure of interest. For example, from liquidus phase equilibrium measurements in metal oxide-silica systems, the activity coefficient of the chemical compo- nent SiO2 in melts is found to be correlated positively with ionization potential of the network-modifying cations (Ryerson 1985). There also appears to be a pos- itive correlation, however, between the abundance of structural units of SiO2 (or Q4) type and ionization po- tential of the metal cation in silicate glasses (e.g., Mysen 1988; Stebbins 1988; Mysen et al. 1982a). The observed increase in activity coefficient of an SiO2 chemical com- ponent (not to be confused with a structural unit) with Z/r 2 of the metal cation might result from this increased abundance of structural units of SiO2 (or Q4) type.1 This 1The existence of discrete structural units in silicate melts appar- ently was first suggested by Morey and Bowen (1924). They used the stoichiometric designations, SiO3, Si205 and SiO2 to describe these units. Their presence was subsequently confirmed by Raman spectroscopy (e.g., Virgo et al. 1980; Mysen et al. 1980, 1982a; Furukawa et al. 1981). Later Raman spectroscopic studies (e.g., Matson et al. 1983) and NMR studies (e.g., Schramm et al. 1984; Murdoch et al. 1985; Stebbins 1987, 1988) referred to these units in terms of their number of bridging oxygens as Q-species (Q4 has 4 bridging oxygens for example, and would be equivalent to SiO2 units in the treatments of Morey and Bowen (1924); and Mysen and coworkers). In order to try to avoid confusion, in this paper we use both designations as a reminder that these are equivalent.