.. role:: raw-html-m2r(raw)
   :format: html

Available interatomic models
============================

The `'models/' <https://github.com/maise-guide/maise/tree/master/models>`_ directory in the GITHUB depository
stores a library of interatomic potentials that can be used in MAISE
simulations. Our focus has been on creating neural network (NN)
interacomic models that can be used in unconstrained evolutionary
structure searches [1_][2_][3_][4_]. For comparison, several
traditional potentials have been represented in MAISE format. The
library includes:

+ ``nn_*_vp`` are previous NN models developed in our published
  studies as reviewed in Ref. [1_]. They have been tested and used for
  identification of stable crystal structures and nanoparticles.

+ ``nn_*_v0`` are new-generation NN models constructed using our
  automated MAISE-NET platform [1_]. They are designed to be used in
  structure searches for stable crystal alloys in the 0-30 GPa
  pressure range (see Figure below).

+ ``gp_*`` and ``sc_*`` are Gupta and Sutton-Chen potentials,
  respectively. The parameters were taken from published studies
  acknowledged in the body of the model file.

The **d3** or **d0** label in the model file name signifies that the
model was constructured primarily to describe bulk structures or
nanoparticles, respectively. Note that the Gupta and Sutton-Chen
potentials developed for nanoparticles have not cutoff radius. It is
set to 10 A or 110 A for simulating bulk structures or
nanoparticles, respectively. Further information about the interatomic
model format can be found in `neural networks section <nnet.html#nn-training-output>`__.



.. figure:: ./figs/models.png
   :align: center

   The latest generation of NN models constructed with MAISE-NET. They
   are designed to be used in unconstrained structure searches for
   stable crystal alloys in the 0-30 GP range.

.. note:: 
   Construction of accurate and robust NN models is a rather
   complicated process that requires the generation of proper
   reference data and the careful fitting of the NN parameters. 

   * We encourange users to try the MAISE-NET platform [1_] for an
     automated development of NN models.

   * We will consider a collaborative work on the development of a NN
     for a particular alloy with up to 3 metals. Please contact `Alexey
     Kolmogorov  <http://maise.binghamton.edu/group>`_ for more information.

.. _github: https://github.com/maise-guide/maise/tree/master/models

.. _1: https://arxiv.org/abs/2005.12131

.. _2: https://journals.aps.org/prb/abstract/10.1103/PhysRevB.95.014114

.. _3: https://journals.aps.org/prl/abstract/10.1103/PhysRevLett.109.075501

.. _4: https://pubs.rsc.org/en/content/articlelanding/2019/cp/c9cp00837c#!divAbstract