This directory contains the core data from the QUEST database.
It provides highly detailed information for each excited state, including both physical characteristics and computational results.

S, T, D, and Q refer to singlet, triplet, doublet, and quartet states, respectively.
GD, PD, CT, and FL denote genuine double, partial double, charge transfer, and fluorescence, respectively.

To assist users in identifying excited states using their preferred methodology, the QUEST database provides the percentage of single excitation involved in the transition (%T₁), oscillator strength $f$, the extent of the electron cloud $\langle r^2 \rangle$, and dominant orbital contributions for all excited states, computed with reasonably high levels of theory. Additionally, $\langle S^2 \rangle$ values are provided for doublet and quartet states, along with the dominant orbital combinations.

Dedicated files for subsets (see below) also include:

• Additional computational methods and basis sets.
• Molecular orbitals involved in each transition (orbital index starts at 1).
• Spatial extent of the electron cloud $\langle r^2 \rangle$.
• Expectation value of the spin operator $\langle S^2 \rangle$ for radicals.
• Extra annotations for complex cases.

• Wavefunction: CIS(D), CC2, EOM-MP2, STEOM-CCSD, CCSD, CCSD(T)(a)*, CCSDR(3), CCSDT-3, CC3, CCSDT
• ADC: SOS-ADC(2)[TM], SOS-CC2, SCS-CC2, SOS-ADC(2)[QC], ADC(2), ADC(3), ADC(2.5)
• Multi-reference: CASSCF, CASPT2, CASPT2 (No IPEA), CASPT3, CASPT3 (No IPEA), SC-NEVPT2, PC-NEVPT2
• TD-DFT (if available): e.g., TDDFT-B3LYP, TDDFT-PBE0, etc.

  [{
    "Molecule": "Water ",
    "Size": 1,
    "Group": 12,
    "State": "^1B_1",
    "Spin": 1,
    "V/R": "R",
    "Type": "n3s",
    "%T1 [CC3/AVTZ]": 93.4,
    "f [LR-CC3/AVTZ]": 0.054,
    "TBE/AVTZ": 7.626,
    "Method": "exFCI/AVTZ",
    "Safe ? (~50 meV)": "Y",
    "TBE/AVQZ": 7.672,
    "Corr. Method": "CCSDT",
    "CIS(D)": 7.168,
    "CC2": 7.234,
    "EOM-MP2": 7.577,
    "STEOM-CCSD": 7.564,
    "CCSD": 7.597,
    "CCSD(T)(a)*": 7.596,
    "CCSDR(3)": 7.597,
    "CCSDT-3": 7.605,
    "CC3": 7.605,
    "CCSDT": 7.591,
    "SOS-ADC(2) [TM]": 7.401,
    "SOS-CC2": 7.452,
    "SCS-CC2": 7.379,
    "SOS-ADC(2) [QC]": 7.23,
    "ADC(2)": 7.181,
    "ADC(3)": 7.842,
    "ADC(2.5)": 7.5115
  }]

In addition, .json files, gathering the same information, are available for each molecule.

The folder cas contains detailed information on the active spaces used in all CASSCF-based calculations. Each entry specifies:

• The number of active orbitals per irreducible representation
• The state-averaging scheme, including the number of states per symmetry
• Systematic inclusion of the ground state, even across symmetries

Proper active space design is key for balancing electron correlation accuracy with computational efficiency. This metadata ensures full reproducibility and transparency in multiconfigurational treatments.

This directory also includes an example of diet subset (located in the diet folder), which is carefully designed to reproduce the key statistical features of larger datasets while restricting data size. These size constraints make the subset well-suited for use with computationally demanding or inefficient methods, particularly during the early stages of method development.

The following subset is available:

• 50 excitations across 20 molecules: diet_subset_50.json

This set is derived from the QUEST Main dataset, excluding unsafe and genuine double excitations (resulting in 824 transitions across 119 molecules). This filtered set of excitations can be found in filtered_main_set.json. The corresponding output file (diet_subset_50.out) generated using the quest-diet.py script is also provided, as well as the analysis of the excitations included in this diet subset (excitations_diet_subset_50.out and analysis_diet_subset_50.out).

Note 1: This diet subset is not unique. It is generated using a genetic algorithm, which is inherently stochastic — meaning that operations like selection, crossover, and mutation involve random choices. As a result, running the quest-diet.py script multiple times (even with the same parameters) can produce different subsets of excitations. This variability allows for exploration of multiple near-optimal solutions, but it also means that results are not guaranteed to be reproducible unless a fixed random seed is explicitly set in the script or passed as a parameter.

Note 2: This set was obtained using a genetic algorithm that minimizes a scoring function incorporating the MAE, MSE, and RMSE of single-reference wave function methods. Multireference methods (e.g., CASSCF, CASPT2, and NEVPT2) are excluded. The script can be easily modified to target a different set of methods.

• All data use atomic units unless otherwise specified. Notably, excitation energies are in eV.
• The database is continuously expanded and refined following the "mountaineering strategy" for reaching high-accuracy benchmarks.