MeshPhysicalMaterial: Added clearcoat support for rect area lights.

[TobiasNoell]

Description

directRectArea method in PhysicalLightingModel did not consider a materials' clearcoat settings. Thus, a material appeared without clearcoat if being illuminated by rect area light sources.

[github-actions]

📦 Bundle size

Full ESM build, minified and gzipped.

	Before	After	Diff
WebGL	356.54 84.66	357.28 84.79	+749 B +132 B
WebGPU	621.76 172.66	622.03 172.75	+272 B +86 B
WebGPU Nodes	620.36 172.43	620.63 172.51	+272 B +86 B

🌳 Bundle size after tree-shaking

Minimal build including a renderer, camera, empty scene, and dependencies.

	Before	After	Diff
WebGL	489.35 119.68	490.1 119.81	+749 B +135 B
WebGPU	692.73 187.97	693 188.05	+272 B +79 B
WebGPU Nodes	642.4 175.15	642.4 175.15	+0 B +0 B

[Mugen87]

Can you also provide a fiddle that allows use to verify the change? Like a simple scene with a clear coat material and a rect area light?

[TobiasNoell]

Can you also provide a fiddle that allows use to verify the change? Like a simple scene with a clear coat material and a rect area light?

Sure, but is there any documentation how I can make the fiddle use the build attached to this PR?

[Mugen87]

It's sufficient to show how it's not working with the current release. I can then test this change locally.

[TobiasNoell]

I have added a fiddle here: https://jsfiddle.net/TobiasNoell/jprxayeb/47/
It is basically the rect area light source example, just the torus has a roughness of 1 plus a clearcoat added with roughness 0

[Mugen87]

Small advice: In general, when you file a PR that updates a material's shader, it's good to share before/after screenshots in the initial comment so the community immediately sees what the PR is about.

[TobiasNoell]

I have also generalized the LTC formula to cope with materials that have a specularF90 that is not one
d6f6a00

[Mugen87]

For comparison:

Before:

After:

[Mugen87]

I have also generalized the LTC formula to cope with materials that have a specularF90 that is not one

When I understand the code correctly, the F90 value was previously hard coded to vec3(1.0). But depending how specular and metalness are configured, the material's F90 value is different. So the proposed change looks like a correct extension.

[TobiasNoell]

When I understand the code correctly, the F90 value was previously hard coded to vec3(1.0). But depending how specular and metalness are configured, the material's F90 value is different. So the proposed change looks like a correct extension.

The rational behind this is that in the LTC approximation (https://blog.selfshadow.com/publications/s2016-advances/s2016_ltc_fresnel.pdf) the Schlick approximation was used but with a hardcoded setting of F90 = 1,

vec3 F_Schlick(float u, vec3 f0) {
    return f0 + (1.0 - f0) * pow(1.0 - u, 5.0);
}

The more general term with F90 would be

vec3 F_Schlick(float u, vec3 f0, float f90) {
    return f0 + (vec3(f90) - f0) * pow(1.0 - u, 5.0);
}

which would yield to this update

[mrdoob]

Nice!

@TobiasNoell I was working on this before the holidays: #32496

Need to find some time to get back to it.