Year: 2026
Type: preprint
Abstract: MaRs is an open-source Python library for the simulation and fitting of electron paramagnetic resonance (EPR) spectra. MaRs supports simulations of field-swept and frequency-swept continuous-wave EPR, accounting for various spin-sublevel polarizations. Moreover, MaRs is capable of simulating complete 2D time-resolved EPR spectra (in field-time domain). The latter can be computed in different levels of theoretical complexity. First, the time evolution can be computed either with kinetic rate equations, which describe population transfer, or in full Liouville space, which captures coherence dynamics. Second, the relaxation pathways in timeresolved simulations can be specified either through intuitive rate parameters that are automatically expanded into the Lindblad master-equation form, or derived microscopically using the Bloch-Redfield formalism. To handle physically complex relaxation scenarios, MaRs provides automatic thermal-balance enforcement and composition operations that allow multicomponent relaxation models to be assembled from simple building blocks. Built on PyTorch (popular platform for machine learning development), MaRs leverages operations on both CPU and CUDA (GPU) backends. We demonstrate MaRs time-resolved EPR capabilities as well the differences and limitations of each built-in approach to time-domain simulations on the examples of coupled triplet-doublet pairs and systems exhibiting the reverse intersystem crossing phenomenon. Separately, we validate stationary EPR simulations across diverse spin systems, including high-spin complexes, achieving excellent agreement with established packages. We also show the possibilities of parameter fitting framework, applicable to all spectrum types implemented in MaRs, that supports multiple optimization strategies and uncertainty quantification.
Source: ChemRxiv
Language: English
Subfield: Spectroscopy
Field: Chemistry
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7213896400