I'm glad you are using your PhD to try to help solve this energy problem.  As someone who has been thinking about blockchain since 2010 and energy use nearly as long, I can assure you the place to be looking is "difficult" algorithms.  CPU only algorithms are one of the best places to go.  Also research CAPEX/OPEX ratio, you will find it discussed in oPoW whitepaper.

If we can limit only CPU's to mine it democratizes mining like you want and also limits large mining farms by the much higher capex/opex ratio than easy algorithms like sha256 which can be sped up by orders of magnitude with specialized hardware, throwing much more power at it than CPU's ever could utilize.

There has been a missing link, and that is memory hardness is the only way to reliably stop asics or gpus, but given enough time they will overcome the memory hardness due to moores law.  The simple yet non-obvious solution which I described previously, is to scale the algorithm's memory requirement with moores law.  So for example every 2 years worth of blocks the memory requirement for the hashing algo doubles.  This will preclude asic manufacturers keeping up as the consumer commodity market is the only thing that can keep up with moores law in fast memory.