gSOMOs 0.9.0

A Python package to analyze magnetic molecular orbitals (SOMOs) from Gaussian outputs

Version [0.9.0] - 2024-04-26 Changed

• logo is now gSOMOs instead of SOMOs
• in the projection scheme (proj.py), there are now two criteria to identify a SOMO, namely "SOMO P2v?" (formerly SOMO?) and "SOMO dom. β MO?" (see scientific documentation)
  • SOMOs identified according to the P^2_virtual criterion are highlighted in green
  • SOMOs identified only on the basis of a dominant virtual beta MO are highlighted in orange (weaker criterion)
• Scientific documentation renamed gSOMOs.pdf. And content updated Added
• new analyzis scheme in proj.py: bases on the diagonalization of projection matrices
• new clean_logfile_name() function in io.py (made to solve a prefix issue for X.log.gz files)

SOMOs

🔗 Available on PyPI

A Python library to identify and analyze Single Occupied Molecular Orbitals (SOMOs) from Gaussian 09 or Gaussian 16 .log files.

---

Installation

pip install SOMOs

---

Capabilities Overview

SOMOs is a Python toolkit for analyzing molecular orbitals (MOs) from Gaussian log files, with a focus on identifying SOMOs (Singly Occupied Molecular Orbitals) in open-shell systems.

---

🚀 Main Features

Load Gaussian Log Files

• Parses .log and .log.gz Gaussian output files
• Extracts orbital energies, coefficients, overlap matrices, and spin

from somos import io
alpha_df, beta_df, alpha_mat, beta_mat, nbasis, S, info = io.load_mos_from_cclib(logfolder, logfile)

Cosine Similarity & SOMO Detection

• Computes cosine similarities between alpha and beta orbitals
• Identifies SOMO candidates from orbital projections

from somos import cosim
listMOs, coeffMOs, nBasis, dfSOMOs, S = cosim.analyzeSimilarity(logfolder, logfile)

Projection-Based Analysis

• Projects occupied and virtual alpha MOs onto virtual beta MOs
• Decomposes projection matrix to extract leading contributions

from somos import proj
df_proj, info = proj.project_occupied_alpha_onto_beta(logfolder, logfile)
display(proj.show_alpha_to_homo(df_proj, logfolder, logfile))

---

📊 Visualization Tools

Heatmaps

• Interactive or static heatmaps of MO similarities
• Highlights SOMO-related regions and orbital clustering

t-SNE (Dimensionality Reduction)

• Projects high-dimensional orbital space to 2D for visual exploration
• Enables inspection of orbital families and similarity patterns

cosim.heatmap_MOs(listMOs, coeffMOs, nBasis, S, logfolder, logfile)          # Generates heatmap from cosine similarities
cosim.tsne(listMOs, coeffMOs, S, logfolder,logfile)                          # Generates 2D layout from cosine similarities
proj.projection_heatmap_from_df(df_proj, info["nbasis"], logfolder, logfile) # Generates heatmap from the projection scheme

---

📁 Output

• .xlsx tables of SOMO similarity and projections
• .png images of heatmaps and projections
• All results saved in the logs/ folder

---

✅ Example Used in Notebook

• H2CO_T1_g09_wOverlaps.log.gz (compressed Gaussian file)

---

Examples

see the SOMOs-examples.ipynb Jupyter notebook

---

Technical and scsientific documentation

This document describes two complementary methods to identify singly occupied molecular orbitals (SOMOs) in open-shell systems:

• Orbital projection analysis, where occupied α orbitals are projected onto the β orbital basis using the AO overlap matrix;
• Cosine similarity mapping, which computes the angular similarity between α and β orbitals and matches them using the Kuhn–Munkres (Hungarian) algorithm.

An example based on the triplet state (T₁) of formaldehyde (H₂CO) is included in the doc/ folder

---