In reply to @gabrielbramley:matrix.org

Hello, we recently implemented a generator for NEB initial paths based on the geodesic interpolation algorithm in this paper (https://pubs.aip.org/aip/jcp/article/150/16/164103/198363/Geodesic-interpolation-for-reaction-pathways) using the ASE framework. In addition to what the paper sets out, we've managed to get the algorithm working for periodic boundary conditions and fixed atom constraints. This implementation is advantageous in that it avoids the clashing atom problem that would sometimes appear in the linear interpolation initial guesses of the IDPP algorithm, and seems to reduce the number of steps required for convergence in NEB.

We currently have this algorithm implemented in our own script repository, but would it be helpful to add this algorithm to the ASE NEB module?

In reply to @gabrielbramley:matrix.org
Hello, we recently implemented a generator for NEB initial paths based on the geodesic interpolation algorithm in this paper (https://pubs.aip.org/aip/jcp/article/150/16/164103/198363/Geodesic-interpolation-for-reaction-pathways) using the ASE framework. In addition to what the paper sets out, we've managed to get the algorithm working for periodic boundary conditions and fixed atom constraints. This implementation is advantageous in that it avoids the clashing atom problem that would sometimes appear in the linear interpolation initial guesses of the IDPP algorithm, and seems to reduce the number of steps required for convergence in NEB.

We currently have this algorithm implemented in our own script repository, but would it be helpful to add this algorithm to the ASE NEB module?

In reply to @tgmaxson:matrix.org
Is there any nice function in ASE to rotate a orthogonal cell atoms object to be aligned along the x, y, z axes?

In reply to @gabrielbramley:matrix.org
Thank you for your interest (And alin m elena🍀 )! I discussed this with Ask and Anubhab yesterday and the best option we came up with is splitting the functionality off into it's own repository, and advertising the functionality in the TS search documentation. I also have another set of TS search related workflows that can be included in this separate repository as well which improves the robustness of TS searches i.e., an ASE implementation of the NEB-TS workflow in ORCA, which runs NEB with a loose convergence criteria and then applies the dimer method to the highest energy image. This can be done with minimal modification to the source code (only energy weighted springs are needed in the default NEB methods), but having a single function that calls this workflow might be helpful to prospective users. I will try and get this done before

Opinions: Would it desirable to have an option in ase.build.find_optimal_cell_shape() to exclude supercells with overly acute angles? Something like

angles = cell_to_cellpar(P @ cell)[3:]
target_angle = 40.0  # should be optional input?
if (angles < target_angle).any() or (angles > 180.0-target_angle).any():
    continue