Despite its low abundance, water has a great influence on the geodynamics of the Earth's upper mantle. Indeed, water has the ability to modify the phase relations and to affect in a significant way the rheological properties of minerals and melts. However the mechanisms of water incorporation in silicate melts and the impact on the melt properties is still not fully understood. To improve our understanding of hydrous silicate melts, we have performed a series of molecular dynamics simulations to evaluate the H 2 O solubility, the liquid-vapour coexistence, the surface tension, the water speciation, the equation of state, the viscosity, the electrical conductivity, the diffusion of silicate elements and protonated species, as well as the melt structure of various magmatic liquids representative of the Earth's upper mantle (rhyolite, andesite, MORB, peridotite, and kimberlite). For that, we introduce a new force field for water, which is compatible with an accurate force field for silicates recently developed (Dufils et al., 2018). A comparison between MD calculations and experimental data (when they exist) shows that the MD simulations are reliable. Among all the results obtained in this study, the following points may be emphasized. (1) The solubility of water changes very little when the melt composition evolves from rhyolitic to andesitic and basaltic, but it is strongly enhanced in ultramafic melts. (2) When hydrous melt and aqueous fluid are coexisting with each other, the oxide content of the aqueous fluid increases rapidly with the pressure. (3) A consequence of point (2) is that water has a large influence on the surface tension, as the latter one drops by a factor of 2 ∼ 4 when the water pressure increases from 1 bar to a few kbar. (4) Concerning the water speciation, an important point is that the MD simulation probes the liquid phase, when most of the experimental studies are dealing with glasses. Thus at magmatic temperatures the concentration in hydroxyl groups and the one in molecular water are crossing for a water content of about 15 wt%, a value much higher than the one observed in glasses (∼ 3–4 wt%). (5) MD calculations show that the molar volume of the melt is a linear function of the water content, and so for all the chemical compositions investigated. Therefore the water partial molar volume (V H 2 O) is virtually independent of total water content and of water speciation. A by-product of this result is that an ideal mixing rule between water and the silicate component leads to an accurate estimate of the melt molar volume. (6) At fixed T and P, the melt viscosity decreases with water content, more depolymerized the melt the smaller the influence of water on the viscosity. However, at the high temperatures investigated in this study (T ≥ 1673 K), the decrease in viscosity induced by water does not exceed one or two orders of magnitude, as compared with many orders of magnitude near the glass transition temperature. (7) The diffusivity of ions increases exponentially with water content. As for the protonated species, it is found that, D O 2 - < D O H - < D H 2 O ≤ D H 3 O + , the lower the NBO/T ratio the smaller the ratio D O H - / D H 2 O. (8) A structural analysis shows that hydroxyl groups are more preferentially linked to metal cations than to structure makers. In contrast, H 2 O molecules (and H 3 O+ as well) are almost exclusively linked to metal cations. As for the melt polymerization, it decreases gradually with the water content in andesitic and basaltic melts, whereas it remains almost invariant in peridotitic melt. (9) O-H...O bonds (hydrogen bonding) taking place between the hydroxyl groups, the water molecules, and the oxygens of the silicate are characterized by O⋯O distances in the range 2.5 ∼ 3.2 A, and by O...H-O distances in the range 1.5 ∼ 2.2 A. But, because of the high temperature of investigation, these H-bonds are generally weak (weaker than in liquid water at ambient). [ABSTRACT FROM AUTHOR]