There are formulas for working out deviation and the resultant audio bandwidth - which I have long ago forgotten
But you can't get much audio bandwidth without substantial carrier deviation, and an ultrasonic transducer is most likely to only work on a very specific frequency, FM'ing it is likely to lose the sidebands.
You are perfectly correct about NOT using VHF radio techniques, just a simple voltage controlled audio oscillator is all that's needed, such as the PLL you suggested.