At optical frequencies, the process is referred to as "parametric downconversion" whereby a higher energy (higher frequency [shorter wavelength] = higher energy) photon is 'converted' to two lower energy photons in a non-linear optical medium.
The challenge is finding a medium/material that exhibits sufficient non-linearity for the wavelength in question (and available power) to make the pitiful efficiencies of the process tolerable. For near-IR/visible wavelengths, crystalline samples of weird substances like lithium/potassium niobate/tantalate are often used. For the mm wavelengths though, I wouldn't have a clue! Something about nitrogen vacancies and quantum dots rings a bell of familiarity but I can't recall having heard of anything practical - but my experience in the field is nearly 10 years old... (so undoubtedly there's now a funky metamaterial available that might do the trick
Sounds fascinating!