Geoffrey A. Landis <geoffrey.landis@lerc.nasa.gov>

When the sun sets, sometimes the last bit of light from the disk itself
is an emerald green.  The same is true of the first bit of light from
the rising sun.  This phenomenon is known as the "green flash" or "green
ray."  It is not an optical illusion.

The green flash is common and will be visible any time the sun is
rises or sets on a *clear*, *unobstructed*, and *low* horizon.  From
our observatory at Mt. Hopkins, I (SW) see the sunset green flash
probably 90% of the evenings that have no visible clouds on the
western horizon.  It typically lasts one or two seconds (by estimate,
not stopwatch) but on rare occasions much longer (5 seconds??).  I've
seen the dawn green flash only once, but a) I'm seldom outside
looking, b) the topography is much less favorable, and c) it takes
luck to be looking in exactly the right place.  If you'd like to see
the green flash, the higher you can go, the better (see below).

The explanation for the green flash involves refraction, scattering,
and absorption.  First, the most important of these processes,
refraction: light is bent in the atmosphere with the net effect that
the visible image of the sun at the horizon appears roughly a solar
diameter *above* the geometric position of the sun.  This refraction
is mildly wavelength dependent with blue light being refracted the
most.  Thus if refraction were the only effect, the red image of the
sun would be lowest in the sky, followed by yellow, green, and blue
highest.  If I've understood the refraction table properly, the
difference between red and blue (at the horizon) is about 1/40 of a
solar diameter.

Now scattering: the blue light is Rayleigh scattered away (not Compton
or Thomson scattering).

Now absorption: air has a very weak absorption band in the yellow.
When the sun is overhead, this absorption hardly matters, but near the
horizon, the light travels through something like 38 "air masses," so
even a weak absorption becomes significant.

The explanation for the green flash is thus, 1) refraction separates
the solar images by color; 2) at just the right instant, the red image
has set, 3) the yellow image is absorbed; and 4) the blue image is
scattered away.  We are left with the upper limb of the green image.

Because the green flash is primarily a refraction effect, it lasts
longer and is easier to see from a mountain top than from sea level.
The amount of refraction is proportional to the path length through
the atmosphere times the density gradient (in a linear approximation
for the atmosphere's index of refraction).  This product will scale
like 1+(h/a)^(0.5), where h is your height and a the scale height of
the atmosphere.  The density scale height averaged over the bottom 
10 km of the atmosphere is about 9.2 km, so for a 2 km mountain the
increase in refraction is about a factor 1.5; a 3 km mountain gives
1.6 and a 4.2 km mountain (e.g., Mauna Kea) gives 1.7.

More details can be found in _The Green Flash and Other Low Sun
Phenomena_, by D. J. K. O'Connell and the classic _Light and Color in
the Open Air_.  A refraction table appears in _Astrophysical
Quantities_, by C. W. Allen.  There's also an on-line resource at
<URL:http://mintaka.sdsu.edu/GF>.