OK .... now you changed things...

If you are studying R
a then it is a unilateral tolerance (cannot be less than zero). Therefore only 1/2 of the normal curve applies. You can use the Cp and CPk calculations for unilateral tolerances as described elsewhere on this site.
"Half of a normal curve" does not apply.
First, it is an
inadequate model - as it does not fit the data adequately at all. Weibull or beta generally are a much better model of unilateral tolerance behavior.
Secondly, calculating Cpk based on only one of either Cpu or Cpl is generally meaningless. Cpk is specifically designed to tell if the
distribution is centered. So, first, it needs to make sense that it is centered, which it does not make sense with a unilateral tolerance with the target at zero. What it actually describes is the
distance the mean is from the tolerance. That is a
totally different question than if the mean is centered. So, just because you can calculate it, does not make it meaningful.
I used to use half normal, until I learned more about the applicable statistics. It is a common approach for people trying to meet a customer's rubber stamp request for Cpk on any and every kind of characteristic. Unfortunately, it is incorrect.