Sound as a use to separate molecular structures Sound can be a destructive force.  However, could it be used to separate say the Hydrogen atom from the Oxygen atoms?
 A: Molecules are broken apart when they collide with one another with enough force to break the bonds that hold them together. These collisions happen all the time and depend on the density of the gas. This is what defines the mean free path. 
The frequency of collisions increases with density. The force involved in the collisions increases with temperature (because the temperature of a gas is related to the average kinetic energy of an ensemble of particles -- hotter gas == faster molecules == more energy transferred in collisions). 
Okay, so that's the background. Very, very crudely explained. To the specific question, sound is just pressure waves. Where the pressure is high in the wave, the density and the temperature of the gas increase relative to the baseline, and where it is low in the wave, the density and temperature decrease. 
It is therefore possible for a pressure wave to be strong enough to break apart molecules by increasing the temperature of the gas from the adiabatic compression. This happens all the time in hypersonic (Mach Number > 5 or so) regimes. This is space re-entry bodies, missiles, meteors... 
The real question is when does a pressure wave stop being a "sound wave" and start being a "shock wave." I'm not an expert in acoustics but typically acoustics implies linear wave theory. This means that shocks are not "sound." Since shocks happen when the flow reaches Mach 1, and chemical dissociation starts at Mach 5 or so in air, that would seem to imply that no, sound cannot cause chemical dissociation by the definition of sound as a linear pressure wave. But shock waves certainly can. 
The destructive nature of sound is typically more related to exciting natural frequencies of a material than just obliterating it to pieces from the incident wave itself. The frequency of the sound wave is chosen to match the resonance of the material so it self-amplifies and the material destroys itself.
A: My best guess is that you would need to use a ultra high frequency played directly into a small body of water.
Theoretically it could be done but actually doing it would be an immense task that I doubt anywhere other than the best lab could do 
