From the paper-
It appears possible, at least in theory, to overcome at least some of the factor-of-3 loss due to
color filtration by finding a way to detect color without filters. Current sensor technology
detects photons through allowing them to penetrate into a silicon semiconductor and transfer
their energy into exciting electrons into the conduction band.8 The mean depth for absorption is
a function of photon energy, which is inversely proportional to wavelength. Light at the blue
end of the spectrum has a mean absorption depth an order of magnitude shorter than light at the
red end. Thus one could imagine a sensor that measures the depth at which photons were
absorbed rather than just the fact of absorption. Photon absorption is a statistical process, so
this method, even if feasible to implement in practice, might not provide sufficient
discrimination between colors to be practical.
But isn't that how the Sigma Foveon sensor works?
Two things that could be a help with understanding more of the article are explanations of how a digital camera adjusts for sensitivity and why noise is a statistical issue.
Otherwise it's well worth trying to grasp although he states that film and digital have the same dynamic range, I thought film was two to three stops better.
Justin.