Rare earth ore provides them all, but it's reportedly difficult to separate Sc and Y from the lanthanides. Centrifuges could do it thanks to the well spread molar masses: 45g, 89g and for La (138g) 139g with at least 44g difference, 15* easier than for uranium.
Volatile compounds are rare. ScCl3, YCl3 and RCl3 need some 1800K to boil under 1atm, so hopefully 1000K make a usable vapour pressure. Monoisotopic Cl brings a bit.
Aluminum, maraging and graphite-epoxy can't operate that hot, superalloys rotate slowly. But carbon-carbon could make the rotors, maybe with a thin hermetic metal liner. 2D tubes resist even at heat around 200MPa, varying a lot among the suppliers, so near-azimuthal winding shall exceed that. 2000kg/m3 let a tube rotate at 316m/s, so 1/2*Delta(m)*V2=0.26*RT at 1000K, needing around 27 steps.
More volatile compounds would reduce RT and enable faster maraging or graphite-epoxy. Salts of organic acids seem excellent, if finding data or measuring.
Marc Schaefer, aka Enthalpy