"...The more you are willing to look upward, the more you will see, including a variety of arcs that lie tangent to the 22-degree halo. Refraction through the 90-degree sides of the crystals can yield a rare 46-degree halo. Crystal alignments can then create sundog analogues in the form of an arc around the point overhead and a gently curving arc that lies far below the Sun parallel to the horizon, one that is so brightly colored that it is commonly confused with the rainbow.
Ice crystals also reflect sunlight from their flat surfaces. One outcome is a white ring through the Sun that extends through the sundogs and that can very rarely be seen to go all the way around the sky. Much more common, and often spectacular, are sun pillars. Go to a romantic movie scene: a couple on a ship look off into the sunset, the sunlight reflecting from a rippled oceanic surface that stretches out in a long line all the way to the horizon. Tip it upside down. For the ocean, substitute a flat layer of high, ice-crystal clouds. The rising or setting Sun now lies below the clouds and reflects off the bottom surface, forming a long line of sunlight that stretches toward the observer. Because of the great distance, we lose any sense of depth, and the line appears instead to soar upward as a seeming pillar of light whose color depends on the apparent color of the Sun and can range from deep red to white. Pillars, halos, and the like can also occur in blowing light snow.
The air itself is a refracting medium. When light from a star enters the atmosphere, it bends slightly downward. The night's stars thus always appear just a bit higher in the sky than they would without the air. The shallower the angle of entry (that is, the closer a star is to the horizon), the greater the effect. It is small at high elevations, toward the point overhead (the zenith), but not at all subtle near the horizon where the star is raised upward by half a degree, the angular diameter of the Sun and Moon. The effect is very noticeable in sunrises and sunsets. When you see the lower edge of the Sun sitting on the horizon, it is actually fully below it.
The rotation of the Earth causes the Sun to appear to move across the sky by its own angular diameter in two minutes. At the equator, where the Sun rises vertically, it comes up two minutes early and is delayed at sunset by the same interval, thereby extending the daylight hours. Advance and delay are longer at higher latitudes, where the Sun rises and sets at a sharp angle relative to the vertical. We tell our students that days and nights are of equal lengths at equinox passages, the first day of spring and fall when the Sun crosses the equator. But in reality, refraction and the fact that sunrise and sunset are counted at first and last glimpse of the solar disk cause the equinoctial day to be several minutes longer than the canonical twelve hours.
Because the degree of refraction increases toward the horizon, the lower edge of the Sun is refracted upward more than the upper edge. The rising or setting Sun is therefore squashed into an oval shape that is very evident to the naked eye (as long as the Sun is sufficiently dimmed to be comfortable to view). Add atmospheric layering -- sudden variations in temperature and density with altitude above the ground -- and you get stripes across the Sun and even pieces of the Sun that seem to hover above the main disk.
Because refraction means dispersion, the air also acts like a natural prism. Think of the Sun as consisting of completely overlapping colored disks, red through violet. Near sunset, the blue and violet disks must be refracted upward more than the red disk. However, violet and blue are fiercely absorbed by the air. The shortest wavelengths of light that come through are green. The narrow top edge of the setting Sun thus turns the color of new grass, which is not directly visible because of the Sun's brightness. Don't try to see it. But then the Sun actually sets, and the last thing to go down is this green edge. Its light, picked up by the air, produces the "green flash" which is particularly evident in Pacific Ocean sunsets. For the same reason, stars near the horizon observed telescopically will all be little vertical spectra, which can drive astronomers to distraction because they need to work with integrated starlight. ..."
http://stars.astro.illinois.edu/dtn.html