Requirement

• jse (>= 4.0.0)

• C & C++ Compiler

  For Windows, MSVC is recommended

  For Linux, GCC is recommended

• MPI Development Environment

  For Windows, Microsoft MPI is recommended (both msmpisdk.msi and msmpisetup.exe are required)

  For Linux like Ubuntu, you can use sudo apt install libopenmpi-dev

Usage

1. Ensure that the environment requirements mentioned above are met and that jse has finished compiling the relevant JNI libraries:

   jse --jnibuild

2. Ensure that the lmp folder is in the jse Groovy classpath. You can choose one of the following methods:

   • jse adds the current working directory to the classpath, so you can simply ensure that the lmp folder exists in the current running directory:

     ├─your_working_dir
     │  └─lmp
     │    ├─FixSgcmcNNAP.groovy
     │    ├─FixFlipNNAP.groovy
     │    └─FixFlipNNAPScaled.groovy
     └─flipFe.lmpin
     

   • OR, place the lmp folder inside the lib/groovy directory of jse. jse automatically adds this directory to the classpath:

     ├─lib
     │ ├─groovy
     │ │ └─lmp
     │ │   ├─FixSgcmcNNAP.groovy
     │ │   ├─FixFlipNNAP.groovy
     │ │   └─FixFlipNNAPScaled.groovy
     │ └─jse-all.jar
     └─jse
     

   • OR, add the directory containing lmp to the JSE_GROOVY_EXLIB_DIRS environment variable:

     export JSE_GROOVY_EXLIB_DIRS="path/to/your/package/dir:$JSE_GROOVY_EXLIB_DIRS"

3. Run LAMMPS via jse -lmp. For example:

   mpiexec -np 8 jse -lmp -in flipFe.lmpin

---

As a many-body interaction, NNAP requires double the cutoff radius to strictly calculate the energy difference caused by species changes. Therefore, you need to set a larger communications cutoff radius using a command similar to:

comm_modify cutoff 14.0

($r_{c} = 6.0,~~6.0 \times 2 + 2.0 = 14.0$). A simpler method is to configure the fix and run it serially; the program will output a suggested setting prompt. For specific settings, please refer to the example file flipFe.lmpin and the comments in the source code.

Citation

Qing'an Li, Lilong Luo, Jian Li, Pengfei Guan, Magnetic Machine-Learning Potential Unveils a Barrierless Martensitic Pathway in Iron.

License

Scripts on this repository are licensed under the GNU GPL v3. See LICENSE for details.