×

Recommended by

Indexed by

Predicting hot-electron free energies from ground-state data

Chiheb Ben Mahmoud1, Federico Grasselli1, Michele Ceriotti1*

1 Laboratory of Computational Science and Modeling, IMX, École Polytechnique Fédérale de Lausanne, 1015 Lausanne, Switzerland

* Corresponding authors emails: michele.ceriotti@epfl.ch
DOI10.24435/materialscloud:36-ff [version v1]

Publication date: Oct 03, 2022

How to cite this record

Chiheb Ben Mahmoud, Federico Grasselli, Michele Ceriotti, Predicting hot-electron free energies from ground-state data, Materials Cloud Archive 2022.125 (2022), doi: 10.24435/materialscloud:36-ff.

Description

Machine-learning potentials are usually trained on the ground-state, Born-Oppenheimer energy surface, which depends exclusively on the atomic positions and not on the simulation temperature. This disregards the effect of thermally excited electrons, that is important in metals, and essential to the description of warm dense matter. An accurate physical description of these effects requires that the nuclei move on a temperature-dependent electronic free energy. We propose a method to obtain machine-learning predictions of this free energy at an arbitrary electron temperature using exclusively training data from ground-state calculations, avoiding the need to train temperature-dependent potentials, and benchmark it on metallic liquid hydrogen at the conditions of the core of gas giants and brown dwarfs. This Letter demonstrates the advantages of hybrid schemes that use physical consideration to combine machine-learning predictions, providing a blueprint for the development of similar approaches that extend the reach of atomistic modeling by removing the barrier between physics and data-driven methodologies. This record contains the raw outputs of the DFT calculations done on the training set. These files are denoted by the "training set folder #*" description. The record also contains a minimal working example showing the ML workflow for training the data and how to run the MD simulations. We include the training data in the format of XYZ files for the structures and NumPy arrays for the DFT energies, forces and DOS. We also provide a Chemiscope visualisation file of the training set.

Materials Cloud sections using this data

No Explore or Discover sections associated with this archive record.

Files

File name Size Description
old-data-set.tar.gz
MD5md5:fa1e7ad39b52ad5a607b3ef5d5f6b907
7.4 GiB training set folder #1
high-press.tar.gz
MD5md5:515a638dfb08ac26b6e036a986fc1ead
2.1 GiB training set folder #2
640.tar.gz
MD5md5:2b4c1bc78470d1b1beea7b091d2948ed
392.1 MiB training set folder #3
1120.tar.gz
MD5md5:de75b9cff07187d4a1786b008bbab3d4
470.8 MiB training set folder #4
1120-bis-rec.tar.gz
MD5md5:3bb2775aae437c443e5d495f588faea2
309.9 MiB training set folder #5
1240-rec.tar.gz
MD5md5:0e4ea3ecce8ba930f99a891f33b94a7f
310.4 MiB training set folder #6
1360-rec.tar.gz
MD5md5:c3e39ab85aefd8c2261a68f110361e85
466.6 MiB training set folder #7
validation_restricted.tar.gz
MD5md5:2b0652ab2430f35773c0799ef9244c5d
1.6 GiB training set folder #8
validation_restricted_30k.tar.gz
MD5md5:076898248d90b4d766ff5006d5fbeca3
2.7 GiB training set folder #9
gap_restricted_15k.tar.gz
MD5md5:1e2f8f3a09638a9a8537392f16bb6472
3.8 GiB training set folder #10
gap_restricted_20k.tar.gz
MD5md5:33e68368361b9f6780994cfd34a2efe1
3.8 GiB training set folder #11
low_npt.tar.gz
MD5md5:a37b4cf69ac6a4decc3078676f060f2d
199.1 MiB training set folder #12
low_nvt.tar.gz
MD5md5:cd58040775659ece9499b0ec7fe3bfd8
363.6 MiB training set folder #13
low_npt_ml_T.tar.gz
MD5md5:860f9b19ec7c000bdc2021d990ddf011
273.5 MiB training set folder #14
low_npt_ml_0.tar.gz
MD5md5:93502c45d129d84056b7ec1ca6eb057d
2.0 GiB training set folder #15
npt_qe_0.tar.gz
MD5md5:8c2e9835b5fb30d0a58c1d578a4aa4c7
10.8 GiB training set folder #16
README-raw-training_set.md
MD5md5:ba09506bcf7cad2bb47aa36f15884d25
2.8 KiB description of the structures in the files with description "training set folder #*"
minimal_working_example.tar.gz
MD5md5:05f6f588bdcc884e6dd27ad06ff7ad52
817.5 MiB contains a reduced dataset to prepare the data, train the ML models and run MD simulations using the electron-finite-temperature MLIP
training_data.tar.gz
MD5md5:b72a4360db62e89076cb2b29e88abd8f
987.9 MiB all the necessary data to train the ML models (GAP and ML DOS) and compute the repulsive 2-body potential
PCA-map-liquidH-chemiscope.json.gz
MD5md5:a4cef3d6711ed2af5a0236ea1c3bb5d5
Visualize on Chemiscope
158.6 MiB Chemiscope visualization of the training set used to generate an electron-finite-temperature ML interatomic potential for liquid hydrogen at planetary conditions

License

Files and data are licensed under the terms of the following license: Creative Commons Attribution 4.0 International.
Metadata, except for email addresses, are licensed under the Creative Commons Attribution Share-Alike 4.0 International license.

External references

Journal reference
C. Ben Mahmoud, F. Grasselli, M. Ceriotti, Phys. Rev. B 106, L121116 (2022) doi:10.1103/PhysRevB.106.L121116

Keywords

SNSF MARVEL Marie Curie Fellowship liquid hydrogen warm dense matter hot electrons machine learning

Version history:

2022.125 (version v1) [This version] Oct 03, 2022 DOI10.24435/materialscloud:36-ff