Issue with volume-stoichiometry regularisation scheme

[nfragapane]

Using volume-stoichiometry regularisation scheme for a RSS workflow gives me this error. Haven't seen this before using linear-hull scheme?

[JaGeo]

@nfragapane Thank you!

what autoplex version are you running? Have you been using the branch in #307?

[nfragapane]

No, the original 0.0.7 used for the paper

[JaGeo]

Please talk to @YuanbinLiu about this as I haven't tested this specific version for producing data. We have only discovered a bug in subsequent versions.

[YuanbinLiu]

@nfragapane, the linear-hull scheme is applicable to all systems. As for the volume-stoichiometry regularization scheme, we have tested it on binary systems with different stoichiometries, and it has worked well. But we have not specifically tested it on non-stoichiometric ternary systems. Thank you for reporting this issue, and we will look into it.

[YuanbinLiu]

@nfragapane, the problem occurs because different structures return a different number of mole fraction components. For example, a two-element system returns a scalar while a three-element system returns an array of length 2. This discrepancy leads to errors when concatenating these values with volume and energy into a uniform array. In this context, one needs to pass element_order in label_stoichiometry_volume to return all element fractions. Currently, label_stoichiometry_volume only supports 3D convex hull calculations. Since we haven't tested whether regularization remains valid for three-component non-stoichiometric systems, more testing is needed. However, on the coding side, we can implement this using calculate_hull_nd in autoplex. A new PR will be initiated for this.

[JaGeo]

@YuanbinLiu Thank you! We are looking forward to the PR.

[YuanbinLiu]

The PR is pushed. I think the issue has been fixed. See #354