This is a large project that creates and analyzes Molecular Dynamics(MD) simulations using LAMMPS www.lammps.org. The primary system of study is the passivating interface of a TOPCon solar cell, i.e., a c-Si/a-SiOx stack, but these tools can be used on any similar system. This project includes lots of code for creating, testing, and analyzing these samples, but the bulk of the code is a large Nudged Elastic Band (https://doi.org/10.1063/1.1323224) data pipeline to study the diffusion of atoms in these samples. NEB is a method that can calculate minimum energy paths for a reaction of interest. In this project, the reaction is an atom or set of atoms moving from one location to another, which allows us to develop an atomic-level understanding of the diffusion of atoms or complexes in the material.

NEB Pipeline

The NEB method requires the initial and final states of the reaction, so this pipeline speeds up the procedure of creating these two data files so that more NEB calculations can be done. An example diffusion process whose energy barrier has been calculated using this NEB pipeline is shown on the right. Significant understanding of the initial and final states is needed to be able to develop the code that creates the correct data files but common diffusion processes in solids are already implemented in this codebase.

The important steps and files in the NEB pipeline;

1. Create a sample and minimize it to the proper energy tolerance of the pipeline(crucial for speeding up the pipeline). There are several example files in the lmp/ folder that create and minimize samples.
2. Create the 'pair list' that is passed into the pipeline. The pair list describes which atoms are being tested and what their final locations are. This file can be manually created or by using an algorithm like those found in py/CreatePairList.py to go through all the atoms in your sample to test which ones fit the criteria you would like to test.
3. Run the pipeline on your pair list. This is done using a bash file which can be run locally run-NEB.sh or on a HPC cluster farm-run-neb.sh. These scripts loop through the given pair list and run;
   • A Python script to set up the initial and final NEB data files - PrepNEB.py,
   • A LAMMPS neb procedure - NEB.lmp, more information can be found here and finally
   • A Python script to analyze the results, save the data, and create a plot/animation of the process that the NEB calculation was done on(example shown on right) - Process-NEB.py.
4. The polished images/gifs and csv file containing the pipeline results default to neb-out/ while the specific log and debug files are placed in the output/ folder for deeper analysis or bug fixing.

Types of NEB processes implemented

There are a number of styles that are currently available to create an NEB process from.

1. Single/Multi 'Zap' - This process simulates vacancy migration in a solid. It takes the ID of an initial atom and a final atom, the final atom is deleted and its position is used as the final location for the NEB process. The multi-version of this takes a number of atom IDs for the final atom and does an NEB process for each of these using the final image of the previous NEB jump as the initial image of the next jump. This was used to study oxygen atoms vacating a specific region.
2. Single/Multi jump - This process takes the ID of an initial atom and a location that the atom will move to, the multi version proceeds similarly to the above. This type of process needs a fair bit of analysis ahead of time during the creation of the pairlist. This can be used to simulate lots of different processes, depending on what is needed, and is the primary function you should start using.
3. Boomerang - is a procedure which moves the atom from initial to final location multiple times over and over again. Used initially to produce forward and reverse barriers for multijump processes, it can also be used to relax the initial and final configurations to a more significant degree
4. H to H2 - This is a test NEB process; it takes an H atom and a final location. PrepNEB will then create another H atom in the region near that final location. The final NEB image is the initial H atom and this new atom in a formed H2 bond.

System requirements

• The pipeline requires Python with a number of common libraries as well as the Ovito Python library, which can cause conflicts with other plotting libraries. A working conda environment with strict versions of each of the necessary libraries is included in current_conda_env.yml. You may need to further install matplotlib via pip and not conda, but I was able to manage the entire environment within conda on my home PC.

• An installation of LAMMPS which was built in shared library mode with Python support as well as the replica package. An example build script using CMake is provided in shell/buildlmp.sh.