Unfortunately, the sound effectively interacts with a particle when resonant conditions are fulfilled. It means that [Ultrasound freq.] = [Sound velocity]/[Particle size]. For typical values C~10E3 m/s, D~10E-6 m it gives the resonance frequency in a GHz range. Surely, this frequency can not be done with traditional techniques. More likely any results may be obtained with USP laser-generated ultrasound.
Vladimir - I am not sure I agree with your evaluation of the relevant resonance. If you were correct, the same argument would apply to dirt particles in water, giving 1 GHz for 1 um particles and 1,000 m/s sound speed . Yet we know dirty water cavitates easily at kHz frequencies.
Instead, gas pockets (bubbles) trapped in or around the particles are the resonating structures, with freq given by Fres = Sqrt(3*gam*Po/rhol)/(2*Pi*R) where gam is the ratio of the gas specific heats (~1.4), Po the ambient pressure (100,000 Pa), rhol the liquid density (1000 kg/m3), Pi ~ 3.142, and R the bubble radius (m). For a 150 um bubble, Fres is around 20 kHz.
Furthermore, even if there are no bubbles of this size, a large enough acoustic amplitude will expand much smaller pockets to larger sizes, due to gas diffusion from surrounding liquid. This is the most common occurrence.
For the case of cavitating in liquid Al with embedded nanoparticles, the particle itself is not the nucleus but the pocket of gas surrounding it. Given the higher liquid density but larger (hydrostatic) Po compared to water, Fres should be of the same order as for water. The one factor that might change is the minimum acoustic pressure needed for cavitation (higher cavitation threshold) due to higher surface tension of liquid Al.