CrysFieldExplorer

Papers that used CrysFieldExplorer:

1. Magnetic properties of the quasi-XY Shastry-Sutherland magnet Er2⁢Be2⁢SiO7 (https://journals.aps.org/prmaterials/abstract/10.1103/PhysRevMaterials.8.094001).
2. Quantum entanglement of XY-type spin dimers in Shastry-Sutherland lattice (https://arxiv.org/abs/2412.17913).
3. Crystal field splittings and magnetic ground state of the square-lattice antiferromagnets YbBi2⁢ClO4 and YbBi2⁢IO4 with 𝐽eff=1/2 (https://journals.aps.org/prb/abstract/10.1103/PhysRevB.111.104430).

Please reach out at [email protected] for collaborations. I would be happy to help on new systems.

1.0 version is released. To install:

pip install CrysFieldExplorer

Import the main modules:

from CrysFieldExplorer import CrysFieldExplorer as crs

from CrysFieldExplorer import Optimization as opt

from CrysFieldExplorer import Visulization as vis

CrysFieldExplorer is a fast-converging Python package for optimizing crystal field parameters.

It supports calculation of a list of common rare earth ions. The program consists of three major modules: CrysfieldExplorer(main), Optimization and Visulization. Detailed tutorials will be uploaded soon.

The novalty of CrysFieldExplorer is it adopts a new loss function using theory of characteristic polynomials. By adopting this loss function it can globaly optimize the CEF hamiltonian with Neutron + any other experimental data and does not rely much on accurate starting value, which is usually estimated from point charge models.

A comparsion of the new Spectrum-Characteristic loss ($L_{spectrum}$) and traditional $\chi^2$ loss has been displayed below.

The details of this program can be found at https://scripts.iucr.org/cgi-bin/paper?S1600576723005897.

A comparison of the new loss function $L_{spectrum}$ vs traditional $\chi^2$ loss along a random line in a 15 dimensional parameter space.

This repo is organized with two types of examples corresponding to two types of optimization methods, Particle Swarm Optimization (PSO) and Covariance Matrix Adaptation Evolution Strategy (CMA-ES) used in conjunction with $L_{spectrum}. Generally speaking, when dealing with < 8 CEF parameters, PSO is a good choice for accuracy and for >=8 CEF parameters CMA-ES has shown significant gain in optimizing speed. In both types of examples the codes are desgined with being able to run parallel using mpi4py in mind.

Example 1 Yb2Ti2O7

The Yb2Ti2O7 is a classical example with 6 CEF parameters, traditional algorithms requires estimation of point charge model to provide insight. With CrysFieldExplorer, it can search large parameter phase space and provide a cluster of solutions of all 6 CEF parameters.

From these solutions it can produce excellent agreement between physical measured data and theoretical predictions.