RenderMan for Houdini exposes a powerful system of AOV creation called and light transport output called Light Path Expressions.  To simplify usage, many of the most common AOVs are available as presets in the ROP node under Displays.

To specify an AOV you should provide the following information to RenderMan:

• Variable - Specify the name and type of the AOV, e.g. varying normal Nn. These selections that are built in simply need you to choose them, we provide the correct name to the renderer.
• Source - Set the source string (expression) for the an LPE.  If using a built-in LPE (see below images), you can leave this field empty; otherwise, you must provide the appropriate LPE expression, e.g. the type: color and expression: lpe:C<RD>[<L.>O].
• Display - Set this to framebuffer or a specify a filename, e.g. and OpenEXR file.
• Display Device - Specify the type of output. If you want to see your AOVs interactively, you would set it to houdini Houdini (for Render View) or framebuffer (for 'it').  
• Quantize - Set the quantitze for your output file.   For openexr, keep the default 0 0 0 0The defaults are typically appropriate for floating point capable formats like OpenEXR and TIF.
• Dither - This is unnecessary for floating point outputs such as an EXROpenEXR.
• Filter - For color AOVs, the recommended filter is Gaussian, this typically should match the Beauty or Ci Display Channel. Data AOVs should be unfiltered.
• Filter Width - As above, when using color AOVs, a filter width of 2 2 using Gaussian matches the default render settings.
• Statistics - This field is only used if you are specifying AOVs manually for Denoising.

...

Instead of specifying a variable, you can select one of the following:

Default AOVs are the beauty and alpha channels.

Standard AOVs are data AOVs used for compositing operations that need a data driven pass like normals, depth, etc.

Diagnostic provides you with data about the render stored in a channel for later

Lighting passes are the typical passes used when making visual adjustments during compositing. We also provide per-lobe LPE to extract specific lighting information from Pxr materials.

Shadow AOVs require that we compute the Beauty render and are used for compositing shadows onto filmed plates/footage

Mattes can help you select and isolate objects or materials in post, we recommend PxrCryptomatte for this now

Inegrator AOVs are data AOVs built into the integrator being used (verify support for your chosen integrator). __depth is useful and is anti-aliased. Again, check the examples on the AOVs page.

Custom AOV

You may specify your own AOVs and Light Path Expressions here. This is useful for creating and storing LPE for per-light, per-object, or specific light paths you wish you capture.