Thin-Film Interference

A soap bubble has no color of its own. What you see is light splitting at the front surface of the film, with part of it bouncing back immediately and part traveling on, reflecting off the back surface, and emerging to travel alongside the first wave. Those two reflected waves took different path lengths, so they're out of step by some fraction of a wavelength. Where that mismatch lines up crests with crests, you get a bright reflection; where it lines up crests with troughs, that wavelength cancels and you see the complementary color. As the film drains and thins unevenly, the thickness at each point changes, so the colors shift and swirl.

There's a hidden twist: reflecting off a surface where the light moves into a higher index flips the wave's phase by half a wavelength, like a pulse reflecting off a fixed string end. Light hitting the top of a soap film (going from air, n=1, into soap, n=1.33) gets this flip. Light reflecting off the bottom of the film (going from soap back into air) does not. So even with zero extra travel distance, the two reflections start out half a wavelength offset. That's why the constructive-reflection condition isn't simply 2nt = lambda, but 2nt = lambda/2, lambda(3/2), and so on — and why the thinnest bright film satisfies t = lambda/(4n).

Thickness is everything here. A film must be within a few wavelengths of visible light, a few hundred nanometers, for the two reflections to stay coherent enough to interfere cleanly. An oil puddle in a jar is thousands of times thicker than that, so the two reflections lose any fixed relationship and no color shows. This same accounting — path length plus any phase flips — is exactly what engineers exploit in antireflection coatings, choosing a coating thickness and index so the two reflections cancel instead of reinforce.