Thoughts about light bending
I had a sudden moment of clarity recently when thinking about how to remove a graphical artifact from the Black Hole Simulator.
The problem
The current simulator has one ugly aspect of the rendered image. Exactly 90 degrees from the direction to the black hole a graphical artifact appears - a strip of smudged, incorrectly calculated pixels (see the picture to the right). The reason is hidden deeply in the rendering mechanism.
In short, it looks like this - you can't efficiently calculate the color of every pixel by raytracing. Taking advantage of the symmetry of the Schwarzschild black hole, I created a table of angles of light deflection. Thanks to the symmetry, I can describe each ray of light by only one parameter - simply put, the minimal distance from the black hole along its path (actually, the impact parameter). To calculate the deflection, I need one more thing, and that is the distance from the black hole at which I send/receive it - the greater part of the whole path the ray has to travel, the greater the deflection.
Such a table was being sent to the graphics card. Then, during rendering, a direction of the ray was calculated for each pixel, and this was converted into the impact parameter. The distance was known independently. Appropriate deflection was being read from the table and this was used to calculate the color of the pixel.
In theory everything is fine, but one problem appeared - the light rays sent in directions close to 90 degrees from the black hole have very similar impact parameters. This gives nearly identical deflection angles, which makes many pixels the same color. An ugly strip appears.