% IMPORTANT: The following is UTF-8 encoded.  This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.

@ARTICLE{Zguns:637337,
      author       = {Zguns, Pjotrs and Pudza, Inga and Kuzmin, Aleksejs},
      title        = {{B}enchmarking {CHGN}et {U}niversal {M}achine {L}earning
                      {I}nteratomic {P}otential against {DFT} and {EXAFS}: {T}he
                      {C}ase of {L}ayered {WS}$_2$ and {M}o{S}$_2$},
      journal      = {Journal of chemical theory and computation},
      volume       = {21},
      number       = {16},
      issn         = {1549-9618},
      address      = {Washington, DC},
      publisher    = {[Verlag nicht ermittelbar]},
      reportid     = {PUBDB-2025-03829},
      pages        = {8142 - 8150},
      year         = {2025},
      abstract     = {Universal machine learning interatomic potentials (uMLIPs)
                      deliver near $\emph{ab$ initio} accuracy in energy and force
                      calculations at a low computational cost, making them
                      invaluable for materials modeling. Although uMLIPs are
                      pretrained on vast $\emph{ab$ initio} data sets, rigorous
                      validation remains essential for their ongoing adoption. In
                      this study, we use the CHGNet uMLIP to model thermal
                      disorder in isostructural layered 2H$_c$-WS$_2$ and
                      2H$_c$-MoS$_2$, benchmarking it against \emph{ab initio}
                      data and extended X-ray absorption fine structure (EXAFS)
                      spectra, which capture thermal variations in bond lengths
                      and angles. Fine-tuning CHGNet with compound-specific
                      \emph{ab initio} (density functional theory (DFT)) data
                      mitigates the systematic softening (i.e., force
                      underestimation) typical of uMLIPs and simultaneously
                      improves the alignment between molecular dynamics-derived
                      and experimental EXAFS spectra. While fine-tuning with a
                      single DFT structure is viable, using $\sim$100 structures
                      is recommended to accurately reproduce EXAFS spectra and
                      achieve DFT-level accuracy. Benchmarking the CHGNet uMLIP
                      against both DFT and experimental EXAFS data reinforces
                      confidence in its performance and provides guidance for
                      determining optimal fine-tuning data set sizes.},
      cin          = {DOOR ; HAS-User},
      ddc          = {610},
      cid          = {I:(DE-H253)HAS-User-20120731},
      pnm          = {6G3 - PETRA III (DESY) (POF4-6G3) / FS-Proposal: I-20170739
                      EC (I-20170739-EC) / CALIPSOplus - Convenient Access to
                      Light Sources Open to Innovation, Science and to the World
                      (730872)},
      pid          = {G:(DE-HGF)POF4-6G3 / G:(DE-H253)I-20170739-EC /
                      G:(EU-Grant)730872},
      experiment   = {EXP:(DE-H253)P-P65-20150101},
      typ          = {PUB:(DE-HGF)16},
      pubmed       = {pmid:40801247},
      doi          = {10.1021/acs.jctc.5c00955},
      url          = {https://bib-pubdb1.desy.de/record/637337},
}