Specified and actual layer thickness in ParaStell stellarator builds: measurements on four equilibria and a clearance-targeting construction

ParaStell builds CAD models of a stellarator's in-vessel layers by offsetting points within each toroidal plane, along the normal of the wall's poloidal cross-section. To first order, the distance between layers is then the specified thickness times the cosine of the angle between the wall's normal and that plane, so layers can be thinner than specified. We measured the clearance of each layer, the shortest distance to the layer beneath, against its specified thickness on four published equilibria with the radial build of a recent ParaStell-based study. Clearance was sampled between ParaStell's construction points, where the surfaces are least constrained. With ParaStell as is, the thinnest sample is 0.659 of the specified thickness on ARIES-CS, 0.862 on W7-X, 0.619 on a quasi-helical and 0.367 on a quasi-isodynamic design; on the last, a 30 cm vacuum vessel is 11.0 cm at its thinnest sample. Median absolute relative errors are 1.0–11.1%. We also describe a clearance-targeting construction. It keeps ParaStell's placement directions and flat segment end faces, but moves each construction point along its direction until its distance to the layer beneath equals the specified thickness. On the sampled points its clearance is within 0.046% (4.6×10⁻⁴ in ratio) of the specified thickness on three designs; on ARIES-CS a local 2.8% shortfall remains between construction points. ParaStell's unmerged 3d-offset branch, evaluated with the same surface, is comparably accurate on these uniform builds but leaves layers about 18.4% thin beyond a layer of varying thickness. All results are sampled measurements, not certified bounds. No neutronics were run.