.. DO NOT EDIT.
.. THIS FILE WAS AUTOMATICALLY GENERATED BY SPHINX-GALLERY.
.. TO MAKE CHANGES, EDIT THE SOURCE PYTHON FILE:
.. "generated_examples/1-ase/run_ase_relaxation.py"
.. LINE NUMBERS ARE GIVEN BELOW.

.. only:: html

    .. note::
        :class: sphx-glr-download-link-note

        :ref:`Go to the end <sphx_glr_download_generated_examples_1-ase_run_ase_relaxation.py>`
        to download the full example code.

.. rst-class:: sphx-glr-example-title

.. _sphx_glr_generated_examples_1-ase_run_ase_relaxation.py:


Geometry optimization (LBFGS)
=============================

Two-stage geometry optimization of a slightly perturbed silicon unit cell.
First, positions are relaxed at fixed cell with
:py:class:`ase.optimize.LBFGS`. Then the cell itself is relaxed jointly
with the atomic positions by wrapping the ``Atoms`` in
:py:class:`ase.filters.FrechetCellFilter`.

The script records the maximum force and total energy at every step and
plots them so that the convergence of the two stages is visible at a
glance.

.. GENERATED FROM PYTHON SOURCE LINES 15-81



.. image-sg:: /generated_examples/1-ase/images/sphx_glr_run_ase_relaxation_001.png
   :alt: Energy vs. step, Force convergence
   :srcset: /generated_examples/1-ase/images/sphx_glr_run_ase_relaxation_001.png
   :class: sphx-glr-single-img


.. rst-class:: sphx-glr-script-out

 .. code-block:: none

           Step     Time          Energy          fmax
    LBFGS:    0 07:24:17      -45.148899        3.371935
    LBFGS:    1 07:24:17      -45.593735        2.373622
    LBFGS:    2 07:24:17      -46.253155        0.850056
    LBFGS:    3 07:24:17      -46.289375        0.575552
    LBFGS:    4 07:24:17      -46.315495        0.521658
    LBFGS:    5 07:24:17      -46.337902        0.390084
    LBFGS:    6 07:24:17      -46.348885        0.287841
    LBFGS:    7 07:24:17      -46.356415        0.264728
    LBFGS:    8 07:24:17      -46.365788        0.258876
    LBFGS:    9 07:24:17      -46.373116        0.179410
    LBFGS:   10 07:24:17      -46.375755        0.094182
    LBFGS:   11 07:24:17      -46.376369        0.057819
    LBFGS:   12 07:24:17      -46.376656        0.037955
           Step     Time          Energy          fmax
    LBFGS:    0 07:24:17      -46.376656        1.457212
    LBFGS:    1 07:24:17      -46.462551        1.391114
    LBFGS:    2 07:24:17      -47.069176        0.695960
    LBFGS:    3 07:24:17      -47.233089        0.137011
    LBFGS:    4 07:24:17      -47.238922        0.059608
    LBFGS:    5 07:24:17      -47.239189        0.039993






|

.. code-block:: Python


    import matplotlib.pyplot as plt
    import numpy as np
    from ase.build import bulk
    from ase.filters import FrechetCellFilter
    from ase.optimize import LBFGS

    from upet.ase import UPETCalculator


    atoms = bulk("Si", cubic=True, a=5.43, crystalstructure="diamond")

    # perturb positions and cell so the optimizer has something to do
    atoms.rattle(0.1, seed=0)  # ASE's built-in random displacement method
    atoms.set_cell(atoms.cell * 1.05, scale_atoms=True)

    calculator = UPETCalculator(model="pet-mad-xs", version="1.6.0", device="cpu")
    atoms.calc = calculator

    history = {"stage": [], "energy": [], "fmax": []}  # type: ignore


    def record(stage_name):
        def _cb():
            results = calculator.results
            history["stage"].append(stage_name)
            history["energy"].append(float(results["energy"]))
            history["fmax"].append(float(np.linalg.norm(results["forces"], axis=1).max()))

        return _cb


    # stage 1: positions only
    opt_pos = LBFGS(atoms)
    opt_pos.attach(record("positions"), interval=1)
    opt_pos.run(fmax=0.05, steps=30)

    # stage 2: joint position + cell relaxation
    filtered = FrechetCellFilter(atoms)
    opt_cell = LBFGS(filtered)
    opt_cell.attach(record("cell"), interval=1)
    opt_cell.run(fmax=0.05, steps=30)

    steps = np.arange(len(history["energy"]))
    stages = np.array(history["stage"])
    boundary = (
        int(np.searchsorted(stages == "cell", True))
        if (stages == "cell").any()
        else len(stages)
    )

    fig, (ax_e, ax_f) = plt.subplots(1, 2, figsize=(9, 3.5))
    ax_e.plot(steps, history["energy"], "o-")
    ax_e.axvline(boundary - 0.5, color="k", ls="--", lw=0.8)
    ax_e.set_xlabel("optimization step")
    ax_e.set_ylabel("total energy [eV]")
    ax_e.set_title("Energy vs. step")

    ax_f.semilogy(steps, history["fmax"], "o-")
    ax_f.axvline(boundary - 0.5, color="k", ls="--", lw=0.8)
    ax_f.set_xlabel("optimization step")
    ax_f.set_ylabel("max |force| [eV/Å]")
    ax_f.set_title("Force convergence")

    fig.tight_layout()
    plt.show()


.. _sphx_glr_download_generated_examples_1-ase_run_ase_relaxation.py:

.. only:: html

  .. container:: sphx-glr-footer sphx-glr-footer-example

    .. container:: sphx-glr-download sphx-glr-download-jupyter

      :download:`Download Jupyter notebook: run_ase_relaxation.ipynb <run_ase_relaxation.ipynb>`

    .. container:: sphx-glr-download sphx-glr-download-python

      :download:`Download Python source code: run_ase_relaxation.py <run_ase_relaxation.py>`

    .. container:: sphx-glr-download sphx-glr-download-zip

      :download:`Download zipped: run_ase_relaxation.zip <run_ase_relaxation.zip>`


.. only:: html

 .. rst-class:: sphx-glr-signature

    `Gallery generated by Sphinx-Gallery <https://sphinx-gallery.github.io>`_