MeshPhysicalMaterial: Added .ior property

[WestLangley]

This is appears to be compatible with where the glTF spec is heading, and it allows MeshPhysicalMaterial to retain its [0, 1] reflectivity property.

/ping @donmccurdy

todo: update docs, ts

[donmccurdy]

Thanks, I like this addition!

Is our .reflectivity similar to .specular in KHR_materials_specular? If so, could/should the interaction be multiplying derived F0, rather than one property overwriting the other? See KHR_materials_ior#interaction-with-other-extensions and KHR_materials_specular#interaction-with-other-extensions).

[WestLangley]

@donmccurdy Using glTF terminology, and ignoring the optional texture,

dielectricSpecularF0 = 0.04 * specularFactor * specularColorFactor;
dielectricSpecularF90 = 1.0 * specularFactor;

In three.js, .reflectivity, which is normalized to [0, 1], can be varied to effectively change the 0.04 constant to a value in [0, 0.16]. Setting .ior is an alternate way of doing the same thing.

specularFactor, in [0, 1], is for artistic control, and is used to dampen the specular contribution -- or remove it completely.

Is there anything here that does not seem right to you?

[WestLangley]

/ping @donmccurdy Friendly reminder. :-)

[donmccurdy]

Ok, thanks! I'd like to support both KHR_materials_ior and KHR_materials_specular in MeshPhysicalMaterial. But I actually don't have much of a preference whether we adopt their naming. The term IOR is also used in Blender and Houdini's principled shaders, and in Autodesk's Standard Surface, so I think it's likely to be more familiar to artists. Whether it's familiar to most three.js users is another question.

Given that GLTFLoader could easily read the KHR_materials_ior value and set .reflectivity without this, do you think it is worthwhile to support both? Should our long-term goal be to remove .reflectivity?

[WestLangley]

My intention was to support KHR_materials_ior via this PR. It allows both the loader and the user to set the IOR. It also allows the user to retrieve the IOR via the getter.

MeshPhysicalMaterial was designed to have [0, 1] properties. Consequently, dielectric specular reflectance was remapped and named reflectivity.

Personally, I think it is beneficial to retain both. Certainly for now.

//

My plan is to add support for KHR_materials_specular soon.

[donmccurdy]

Ok, sounds good!

[mrdoob]

Thanks!