In Blender 2.92 Cycles, I'm trying to model a quartz prism with physically correct refraction of a beam of light cast by a narrow slit (not worried about the color dispersion yet). But when I render it, the beam of light doesn't bend, and it doesn't seem to exit the prism in any meaningful way at all. According to my calculations, it should go horizontal inside of the prism, and exit at a symmetric angle to how it entered.

I'm using a plain ol' glass material for the prism with an index of refraction of 1.6, although I get the same results using the principled shader. Any ideas why this doesn't work?

Basic setup enter image description here

Edit: Also, if I change the prism to 100% glossy metal, it doesn't cast a sharp reflection caustic of the beam either. I'd expect to see a discrete blue reflected beam on the floor here, instead of the soft glowy caustic I actually got. Maybe Cycles can't actually do these kinds of things... or is there a "use actual optics" setting I can turn on?

enter image description here

  • 3
    $\begingroup$ Cycles doesn't really support caustics that well if I'm correct, most people just find ways to fake it. Have you tried Luxcore? $\endgroup$ Commented Mar 29, 2021 at 8:00
  • 1
    $\begingroup$ I might be wrong, but although it's called PBR "physically based rendering" it doesn't have the capabilities of a real scientific simulation to recreate real world examples - or at least not in a considerable amount of time since you always limit samples, bounces etc. to keep render times in acceptable ranges. But as the name indicates, it is "physically based", not "physically correct". $\endgroup$ Commented Mar 29, 2021 at 8:32
  • 2
    $\begingroup$ @GordonBrinkmann (As a non-biased 'brute force' engine, Cycles is physically accurate with infinite amount of samples) :) $\endgroup$ Commented Mar 29, 2021 at 8:51
  • 1
    $\begingroup$ Also related: How to simulate glass dispersion with only one Principled BSDF node $\endgroup$ Commented Mar 29, 2021 at 8:56
  • 1
    $\begingroup$ As I said, I might be wrong. but since you never let Cycles render forever you'll never get an accurate result. I mean with usual settings it's hard to even get any light through this prism... $\endgroup$ Commented Mar 29, 2021 at 9:10


You must log in to answer this question.

Browse other questions tagged .