← gallery

Duct

A sunset through a thin warm layer over the sea. Move the height of your eye, and the same sun sets as a different shape.

drag up to climb · sideways to set the sun · arrows nudge · shift for a tenth of a metre · space pauses · K keeps the framea long lens · one wavelength of light

what the picture can say about itself

eye above the sea
100.0 m
the eye
above the layer
sun below horizontal
40.0 arcmin
the sun
four times faster than the sun
the picture
full
the air
computing

When the sun sets over the sea, the air near the water bends its light. Usually the sun only looks a little flattened. But sometimes a thin layer of warm air lies over colder air, a few dozen metres up. Then the sun can set as a rectangle, an hourglass, a pot with a lid, or two thin lines.

Here the layer is 50 to 60 metres above the water. The air at its top is only two degrees warmer than at its foot. That is enough to bend a level ray of light more sharply than the Earth curves, so some light is caught under the layer and never gets out. Andrew T. Young calls a layer like this a duct.

What you see depends on where you stand. Drag up or down to move your eye between 20 and 120 metres above the sea. Above the layer, the bottom of the sun sinks behind a false horizon. It comes back under it as a second image, upside down, that grows into a rectangle and an hourglass and then breaks in two. Inside the layer, a dark strip opens across the sky where the trapped light is, and the sun splits into a pot and its lid. Just below it, the sun pinches in at the waist. Drag sideways to move the sun. When no one touches it, the eye moves on its own between Young's heights, one sunset at each.

Nothing here is drawn by hand. I took the air from Young's own simulations. It is the standard atmosphere with the bottom 50 metres two degrees colder, the change spread over the next ten, and his haze. For every row of the picture I follow one ray of light back from your eye. It ends in space, in the sea, or trapped under the layer, and where it came from decides what the row shows. I dim each ray by the haze and air it crossed. So the images just under the false horizon are about eight times fainter than the sun just above it.

Before anything else I checked it against Young's own pictures. From 100 metres, the second image of the lower limb appears when the sun's centre is 40 minutes of arc below the horizon. Its neck breaks at 72. His pictures show the same. From 65 metres I get 34 and 66, as he does. From 50 metres I see the dark strip sit exactly on the horizon, as in his.

One thing came out differently, and I have kept it. Young's pictures from 46 metres show a narrow dark strip. In my version the floor of the duct is at 46.3 metres, so at 46 the eye is just below it and there is no strip. His page does not say exactly how much he rounded the corners of the layer, and the floor moves with that. I have written to ask him.

I chose some of it. The lens is long, about a degree and a half of sky from top to bottom. The sun sets straight down, as it does at the equator, four times faster than the real one. The glow around the sun and the glitter on the sea are simple models. I drew it in one colour of light, so there are no red or green rims and no green flash.

In Bend I bent light around a black hole. Here I bend it in two degrees of air. Both have a height where a level ray goes round and round. Around the black hole it is the photon sphere. Here it is the place in the layer where a ray curves exactly as much as the Earth does.

After Andrew T. Young · a computed sky and sea · chosen lens, exposure and grain