And what is the complex light wave pattern (assuming it works the
smae way) of a given R, G and B value when combined
--
I don't think it's productive to think about the problem in this way.
When light is collected by the sensor, it excites an electron; when
the sensor is read, the signal corresponds to the number of electrons
that have been excited, which correlates to the amount of illumination
of that pixel. Each electron being excited time-averages over many, many
oscillations of the light wave, because its response time is much longer
than the oscillation period of visible light; this is completely the opposite
of the way things work for audio or radio frequency waveforms, where the
detection device conducts many samplings of the wave over one period
of oscillation (for instance a CD samples the signal at 44KHz, even if
you are listening to a 500Hz tone). These frequency ranges are much
slower than the response time of an electron.
Another point: Even if we can make structures the size of a wavelength
of visible light, doesn't mean that its characteristic response frequency
will be the same. In fact typically it will not be so. The nanotubes
Mike mentioned have the right width, but their natural vibrational
frequency will be many orders of magnitude slower than
the corresponding light wave, and so they will be a poor transducer
for light detection (it's similar to the way a tuning fork will only
sympathetically vibrate with sounds near its natural frequency of
vibration).
Just dreaming here, I wonder though if one could make use of
the fact that photons carry energy that is in direct proportion
to their frequency. That energy is transferred to the electron
when it is excited in a photodetector. If a sensor could detect
the individual electrons and measure their initial
excitation energies accurately, one might be able to do away
with the Bayer filter.
--
emil
--
http://theory.uchicago.edu/~ejm/pix/20d/ 
