Webinar: From Project to Platform: New Resources on PSDI – Session 3

Joint Session 3: High-Quality Scientific Data Resources

Challenge

• • Accurate excited‑state prediction is critical for organic semiconductor design (e.g. OLEDs, OSCs)
  • Widely used single‑reference methods (e.g. TD‑DFT) often fail for: strong static correlation; double‑excitation character; inverted singlet–triplet gaps
  • Lack of reliable, large‑scale multireference benchmark data limits: method development; validation of excited‑state models; data‑driven and ML‑based discovery

Approach

• • Since 2023, Development of BenchmarkSet1500
    • a curated dataset of 1,500 organic molecules
    • excited‑state properties computed using multireference electronic structure methods
  • Systematic analysis of
    • molecular diversity
    • statistical distribution of excited‑state properties
  • Derivation of practical guidelines
    • selecting suitable levels of theory
    • based on molecular fragment type
  • Demonstration through targeted molecular screening
    • inverted singlet–triplet gaps
    • thermally activated delayed fluorescence (TADF)
    • deviations from Kasha’s rule
      

Innovation

• • Establishes a pathway toward transparent, comparable, and reproducible testing standards for single‑cell flow batteries
  • First large‑scale multireference benchmark dataset focused on organic excited states
  • Enables quantitative assessment of TD‑DFT failure regimes
  • Provides a foundation for systematic excited‑state photophysics exploration
  • Supports method development, benchmarking, and validation beyond single‑reference models
  • Establishes a high‑quality data resource for future machine‑learning‑driven materials discovery

Challenge

• • NMR spectra of paramagnetic metal complexes in solution are notoriously difficult to interpret, as unpaired electrons produce large chemical shifts, broadened lineshapes, and temperature-dependent behaviour that standard diamagnetic analysis tools cannot handle.
  • Extracting meaningful electronic structure information (magnetic susceptibility tensors, correlation times, spin-Hamiltonian parameters) from pNMR data requires combining experimental spectra with ab initio calculations, a workflow that today remains fragmented, manual, and inaccessible to many researchers.

Approach

• • SimpNMR is a Python package that streamlines pNMR analysis by directly incorporating outputs of ab initio calculations.
  • It provides an end-to-end workflow for paramagnetic complexes in solution, including:
    • prediction of 1D NMR spectra (e.g. 1H, 13C)
    • assignment of experimental peaks to molecular sites
    • fitting of the magnetic susceptibility tensor and correlation times to experimental pNMR data
  • With variable-temperature experiments, SimpNMR extracts spin-Hamiltonian parameters such as the g-tensor, and in certain cases the D-tensor, directly from solution pNMR data, information typically accessible only from EPR or SQUID magnetometry.

Innovation

• • SimpNMR transforms pNMR analysis from a bespoke, expert-only procedure into a reproducible, scriptable Python workflow that bridges computational and experimental chemistry.
  • SimpNMR_DB, a curated companion database, stores the input data required for SimpNMR analysis, enabling:
    • reuse and benchmarking of ab initio inputs across complexes
    • reproducible, shareable analysis pipelines
    • accelerated discovery by lowering the barrier to quantitative pNMR interpretation
  • Together, SimpNMR and SimpNMR_DB open up solution pNMR as a practical route to electronic-structure parameters of paramagnetic metal complexes.