Skip to content

Latest commit

 

History

History
89 lines (75 loc) · 6.16 KB

README.md

File metadata and controls

89 lines (75 loc) · 6.16 KB
Raw

TNPgeneration

This repository contains code written for generation of AuPtPd trimetallic nanoparticles (TNPs) structural data set for machine learning applications.

Conducted by: Kaihan Lu assisted by Haotai Peng (Bill)

Supervised by: Jonathan Yik Chang Ting and Amanda Barnard

Institution: School of Computing Australian National University

Research course: SCNC2021 Science Research Project

Date Accomplished: 1/1/23

Contents of each directory

  • InitStruct: LAMMPS, Python, and Bash scripts written for generation of TNP initial structures (.lmp format)
  • EAM: Modified LAMMPS tools for the generation of relevant interatomic potential files required for LAMMPS scripts execution
  • MDsim: LAMMPS, Python, and Bash scripts written for tasks related to molecular dynamics (MD) simulations of the generated TNPs
  • FeatExtEng: Python and Bash scripts written for tasks related to feature extraction of TNPs, along with the source code, executable file, and input files of Network Characterisation Package (NCPac), a software developed by Dr George Opletal from CSIRO and extended by Jonathan Ting for structural/geometrical feature extraction of TNPs

Instructions to use the repository to generate more TNPs structural data

  • The current script is designed to:
    • only generate TNPs with different combinations of the listed degrees of freedom, but extension is possible by appropriate modification of the code.
    • be run on high performance computing cluster such as Gadi of National Computational Infrastructure or cluster1 of ANU College of Engineering and Computer Science.

Degrees of freedom of TNPs generated

  • Elemental composition: Au, Pt, Pd
  • Size: 30 Angstroms
  • Shape: Spherical
  • Ratio: i:j:k where i, j, k are from {2, 4, 6, 8}, with an additional constraint if i+j+k == 10
  • Atomic ordering: (chosen with reference to Figure 2 in the review by Crawley et al, Heterogenous Trimetallic Nanoparticles as Catalysts, Chem. Rev. 2022, 122, 6, 6795--6849)
    • L10R: ordered alloy with randomly distributed M3 (o-M1M2-M3)
    • CS: inner-core@core@shell (M1@M2@M3)
    • CL10S: ordered-core@shell (o-M1M2@M3)
    • CRALS: random-core@shell (M1M2@M3)
    • RRAL: random solid solution (M1M2M3)
    • CSRAL: core@random-shell (M1@M2M3)
    • CSL10: core@ordered-shell (M1@o-M2M3)
    • CRSR: core@shell with randomly distributed M3 (M1M2@M2M3)
    • LL10: ordered intermetallic solution (o-M1M2M3)

Generation of TNP initial structures

  1. Modify the variables in the files below under ./InitStruct/ to generate other combinations:
    • constants.py
      • {diameterList}: NP diameters (Angstrom)
      • {ratioList}: Percentage of each element (only ratio combinations that add up to 100% will be generated)
      • {RANDOM_DISTRIB_NO}: Number of replicas for atomic orderings involving random distribution
      • {VACUUM_THICKNESS}: Size of the box containing the NP (need to be >= greatest diameter in {diameterList})
    • genBNPCS.sh
      • {SIZE_ARR}: NP diameters (Angstrom)
    • genTNPCS*.sh
      • Last three arguments representing NP diameters (Angstrom) of each component when calling the functions in these scripts
      • {RATIO_LIST}: Percentage of each element (only ratio combinations that add up to 100% will be generated)
      • {RANDOM_DISTRIB_NO}: Number of replicas for atomic orderings involving random distribution
      • Note: Some NPs that are supposed to be TNPs turned out to be BNP due to overly large overlap cutoff in genTNPCS.sh.
  2. Run genMBTNPs.sh to generate the TNPs to be simulated.

Simulation of TNPs

Stages of simulations

  • S0: Short equilibration of TNPs
  • S1: Heating up of TNPs, saving configurations along the way
  • S2: Short equilibration of the saved TNP configurations at the saved temperature

Instructions:

  1. Go to ./MDsim/
  2. Modify the path in setupMDsim.sh as appropriate and run it. This places each .lmp file into a unique directory in the specified path to store the simulation data, while initialising a config.yml file that stores the simulation progress/status for each directory.
  3. Check that parameters that are not variables in annealS{0/1/2}.in LAMMPS script templates are appropriate according to your simulation goals.
  4. Modify the paths and parameters in genAnnealIn.sh as appropriate and run it, taking simulation stage number as argument {0, 1, 2}.
  5. Generate 3 files at the path specified in setupMDsim.sh {'jobList', 'queueList', 'runList'}
  6. Modify the path in jobList.sh as appropriate and run it, taking simulation stage number as argument {0, 1, 2}. This script:
    • places jobs (LAMMPS running of *S{0/1/2}.in) that are ready to be run by the HPC into 'jobList'.
    • needs to be rerun after the jobs are finished (until all jobs are done), to:
      • update the progress of the simulations by updating config.yml in the TNP directory (whether each stage is done), and
      • update the status of the jobs (whether it's submitted, queuing, or running) by updating 'jobList', 'queueList', and 'runList'
      • place the job for next stage of simulation into 'jobList'.
  7. Modify the parameters in runAnneal.sh and subAnneal.sh, and run subAnneal.sh. This script:
    • submits {maxQueueNum} of job scripts listed in 'jobList' to the HPC.
    • Once a job finishes, it will automatically submit the next job from 'jobList', keeping the number of jobs submitted the same (i.e. {maxQueueNum}) until there are no more jobs in 'jobList'.
    • delJobs.sh and rmOutFiles.sh are provided to ease the debugging/cleaning process.
    • Remember to backup your simulation data to elsewhere when they are done.

Feature extraction of TNPs

  1. Go to ./FeatExtEng/
  2. Modify the paths and parameters in genCSVs.py as appropriate.
  3. Submit runGenDAPdata.sh to the HPC. This will generate:
    • {MDout.csv}, which contains the output of MD simulations of all TNPs.
    • {features.csv}, which contains the features extracted by NCPac for all TNPs.
  4. Modify the parameters in mergeFeatures.py and run it. This will merge the information from the 2 csv files and generate a new {AuPtPd_nanoparticle_dataset.csv} following the format of dataset stored on CSIRO's Data Access Portal, such as https://data.csiro.au/collection/csiro:40669 (gold nanoparticle).