-------------------------------------------------- LR-TDDFT Absorption Spectra for the QM7 dataset -------------------------------------------------- Prepared by Peter Elliott in collaboration with Siddharth Dhanpal, Paolo E. Trevisanutto, Alin Marin Elena, and Gilberto Teobaldi, Scientific Computing Department - STFC - UKRI. This work was supported by the Ada Lovelace Centre STFC Programme. Computing resources have been provided by STFC Scientific Computing Department’s SCARF cluster. -------------------------------------------------- Molecular geometries were sourced from the QM7b dataset using file qm7b_coords.xyz available at https://archive.materialscloud.org/records/rmp9p-1k216 , the data repository for: Yang et al, Sci. Data 6, 152 (2019). which was based on the QM7 dataset developed in Rupp et al, Phys Rev Lett 108, 058301 (2012). Montavon et al, New J. Phys. 15, 095003, (2013). from the GDB-13 dataset: Blum and Reymond, J. Am. Chem. Soc. 131, 8732 (2009). In this work we only include molecules consisting of C,O,N,H only, which reduces the QM7 dataset from 7211 to 6872 molecules. -------------------------------------------------- The local density approximation (LDA) for DFT and adiabatic LDA was used for the TDDFT calculations. All calculations were performing using the Octopus (https://octopus-code.org) electronic structure code with the following configuration: Version : 13.0 Configuration options : maxdim3 openmp mpi sse2 libxc5 libxc_fxc Optional libraries : metis Architecture : x86_64 C compiler : mpicc (gcc) C compiler flags : -g -O2 C++ compiler : g++ C++ compiler flags : -g -O2 Fortran compiler : mpif90 (gfortran) (GCC version 12.3.0) Fortran compiler flags : -O3 -fopenmp Norm-conserving, scalar relativistic, Kleinman-Bylander pseudopotentials, downloaded from https://www.pseudo-dojo.org/ , were used. The grid spacing and box size given in the example octopus inp file. -------------------------------------------------- FILES: -------------------------------------------------- * 7CON.tar.gz The main dataset consisting of 6872 molecules, each in a separate directory. Each directory contains the following files * GEO_M<N>.xyz The molecular geometry in xyz format. The second line gives the corresponding molecule number in the QM7b dataset. * density_M<N>.cube * elf_M<N>.cube The real-space electronic density and electron localization function (ELF) in the cube format. * ks_spectrum_gamma_50meV_M<N>.dat * ks_spectrum_gamma_150meV_M<N>.dat * ks_spectrum_gamma_500meV_M<N>.dat * tddft_spectrum_gamma_50meV_M<N>.dat * tddft_spectrum_gamma_150meV_M<N>.dat * tddft_spectrum_gamma_500meV_M<N>.dat The Kohn-Sham (KS) and TDDFT spectra as a function of frequency (from 0 to 0.9995 Hartrees in 2001 points) calculated using the oct-casida_spectrum tool provided with Octopus for several broadening energies 50meV,150meV,500meV. * tddft_spectrum_list_<N>.dat The first 20 TDDFT frequencies and oscillator strengths for transitions with oscillator strength greater than 1e-3 . --------- * 7CON_casida.tgz Contains the octopus output casida/casida file for each molecule, listing frequency, transition dipole moment, and oscillator strength, labeled casida_M<N>.txt . --------- * gaps.txt For each molecule, the KS gap labelled KS_GAP, the first LR-TDDFT frequency in the casida_M<N>.txt file labelled FIRST_GAP, the first LR-TDDFT frequency with oscillator strength above 1e-3 (the same as the first frequency in the tddft_spectrum_list_<N>.dat file) labelled OPT_GAP, the difference between FIRST_GAP and KS_GAP labelled FIRST_COR, and the difference between the OPT_GAP and KS_GAP labelled OPT_COR. --------- * default_octopus_input.inp The octopus input file used for most calculations. If the chebyshev_filter eigensolver fails to converge, the CG method was tried. If that didn't converge, the number of empty states was reduced. --------- * convergence_notes.txt Lists the molecules that required the number of empty states to be reduced in order to converge, and the number of empty states.