I think there needs to be much more innovation in the field of smog processing.
This is my guess.
I wonder if its possible to classify a filter into the following decision vector. Pass without action, ?reflect, store, convert, … That is. If you could divide the problem of an air filter into small more manageable parts, air filter cells. Then dependent on the air units ?unique dirtiness the filter could make smart decisions. Pass, store, convert , … I also wonder if the filter needs to manage different chemical conversions. Probably not for fuel generation but conversions to grab hold of more particles. The idea is to create a particle “heat” pump. Where you can increase the efficiency of the filter.
Just my thoughts.