CoChat
Sign Up
Cited by 58Advanced Chemical Physics StudiesMachine Learning in Materials ScienceComputational Drug Discovery Methods

Rapid Detection of Strong Correlation with Machine Learning for Transition-Metal Complex High-Throughput Screening

Open access
Liu, Fang, Duan, Chenru, Kulik, Heather J.·2020·The Journal of Physical Chemistry Letters
DOI 10.1021/acs.jpclett.0c02288

Published version of a preprint

Open this paper in CoChat — free account

Abstract

Despite its widespread use in chemical discovery, approximate density functional theory (DFT) is poorly suited to many targets, such as those containing open-shell, 3d transition metals that can be expected to have strong multireference (MR) character. For discovery workflows to be predictive, we need automated, low-cost methods that can distinguish the regions of chemical space where DFT should be applied from those where it should not. We curate more than 4800 open-shell transition-metal complexes up to hundreds of atoms in size from prior high-throughput DFT studies and evaluate affordable, finite-temperature DFT fractional occupation number (FON)-based MR diagnostics. We show that intuitive measures of strong correlation (i.e., the HOMO–LUMO gap) are not predictive of MR character as judged by FON-based diagnostics. Analysis of independently trained machine learning (ML) models to predict HOMO–LUMO gaps and FON-based diagnostics reveals differences in the metal and ligand sensitivity of the two quantities. We use our trained ML models to rapidly evaluate MR character over a space of ∼187000 theoretical complexes, identifying large-scale trends in spin-state-dependent MR character and finding small HOMO–LUMO gap complexes while ensuring low MR character.

Integrity signals

Registry

Crossref · 10.1021/acs.jpclett.0c02288

Cites

109 of 109 references verified0 citation errors · 0 not checkable

Statistics

Recomputing p-values needs the paper’s full text.

Claims the paper makes

We couldn’t retrieve this paper’s full text, so its claims and self-check weren’t run. The registry and citation checks still apply.

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

  • Liu et al. observed that although multireference character is expected to be higher in low-spin states, this trend is seen for only some metals (e.g., Cr or Mn).

    “Through this analysis, Liu et al. (44) observed that although multireference character is expected to be higher in low-spin states, this trend is seen for only some metals (e.g., Cr or Mn) (Figure 3).”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [44]

  • Liu et al. applied an ANN trained to predict the multireference diagnostic r_ND to 3% mean absolute percent error (i.e., 0.018) to a theoretical space of 187.2k new complexes, only 1% of which had been seen in training data.

    “Starting from a data set of 4,865 complexes, they applied an ANN trained to predict the multireference diagnostic r ND to 3% mean absolute percent error (i.e., 0.018) to a theoretical space of 187.2k new complexes, only 1% of which had been seen in training data.”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [44]

  • Conversely, other properties, such as the frontier orbital energies, band gap, or ionization/redox potential, depend much more on the global structure of the molecule.

    “Conversely, other properties, such as the frontier orbital energies (22), band gap (44, 105), or ionization/redox potential (32,52), depend much more on the global structure of the molecule (Figure 5).”

    From Machine Learning for the Discovery, Design, and Engineering of Materials. (2022) · cites this paper as [44]

Cites

Flagged first, then by citations

Cited by

Sorted by citations

Related

Authors
Liu, F., Duan, C., & Kulik, H. J.
Affiliations
Massachusetts Institute of Technology (US)
Corresponding author
Kulik, Heather J.
ORCID
Venue
The Journal of Physical Chemistry Letters
Publisher
American Chemical Society (ACS)
Published
2020-08-31
Volume
11
Issue
19
Pages
8067-8076
ISSN
1948-7185 (ISSN-L)
DOI
10.1021/acs.jpclett.0c02288
OpenAlex
W3082636798
Cited by
58
References
109
Citations by year
2026: 6 · 2025: 7 · 2024: 6 · 2023: 12 · 2022: 17 · 2021: 10
FWCI
3.97
Type
journal article
Language
en
Open access
Licence
Locations
  • The Journal of Physical Chemistry Letters (publishedVersion)
  • ChemRxiv (acceptedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • Royal College of Surgeons in Ireland (submittedVersion, cc-by-nc-nd)
Funders
  • Defense Advanced Research Projects Agency (D18AP00039)
  • American Association for the Advancement of Science (Marion Milligan Mason Award)
  • Basic Energy Sciences (DE-SC0018096)
  • Office of Naval Research (N00014-18-1-2434, N00014-20-1-2150)
  • Division of Advanced Cyberinfrastructure (ACI-1547580 (MolSSI postdoctoral fellowship))
  • Burroughs Wellcome Fund (Career Award at the Scientific Interface)
Topics
  • Physical Sciences › Physics and Astronomy › Atomic and Molecular Physics, and Optics › Advanced Chemical Physics Studies
  • Physical Sciences › Materials Science › Materials Chemistry › Machine Learning in Materials Science
  • Physical Sciences › Computer Science › Computational Theory and Mathematics › Computational Drug Discovery Methods
Keywords
multireference character, transition metal complexes, open-shell systems, strong correlation, fractional occupation numbers, HOMO-LUMO gap, machine learning prediction, high-throughput screening, 3d transition metals, chemical space

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review