The first-melting temperature of haplogranite in the presence of H2O-NaCl-KCl solutions was determined experimentally at deep-crustal conditions of 6-14 kbar and 700-900 degrees C Minimum melting occurs at fluid K/(K+Na) of 0.2-0.25. Melting temperature rises strongly with increasing salinity: at 10 kbar, first melting occurs at 800 degrees C with a H2O mole fraction (X-H2O) of 0.62 +/- 0.02, similar to the salinity of fluid inclusions in minerals of some granites and migmatites. At 900 degrees C and 10 kbar, X-H2O is 033 +/- 0.01 for first melting. All minimum melts are granitic over the entire P-T-X-H2O range, with SiO2 melt concentrations of 74 +/- 2 wt% on a H2O-free basis and Al2O3 and alkalis typical for granites. Normative quartz is near 30% for all liquids. The K/(K+Na) ratios of minimum melts increase strongly with decreasing H2O activity: at 10 kbar and X-H2O = 1, K/(K+Na) in the melt is 0.25, whereas at X-H2O = 0.34 it is 0.55. This "brine trend" is similar to, but more pronounced than, the trend described by decreasing H2O activity in H2O-CO2 fluids, and it better explains the compositions of K-rich granites. Minimum-melting curves in the presence of brines of constant X-H2O have strongly positive dP/dT slopes at P > 2 kbar, in contrast to vertical or negative dP/dT slopes for those in the presence of H2O-CO2 fluid. There is therefore a wide P-T range over which migrating low-H2O-activity brines may generate subsolidus, deep-crustal metasomatism in the form of K2O and silica enrichment. Moreover, once melting occurs, rising accumulations of granitic magma may be fluid saturated and even increase their melting capacity with decreasing depth because of the strong pressure dependence of H2O activity in salt solutions. Our results offer an explanation for mid-crust migmatization and granite production: rising hot brines may provoke rock melting at some threshold of decreasing depth in the range 15-20 km. Because of their enhanced capacity for metasomatism, leading to eventual melting at granulite fades conditions of temperature, pressure and H2O activity, concentrated brines should be considered as possibly important agents in crustal evolution. (C) 2013 Elsevier B.V. All rights reserved.