INSPIRED: Inelastic Neutron Scattering Prediction for Instantaneous Results and Experimental Design

INSPIRED is a graphic user interface (GUI) that performs rapid prediction and calculation of phonons and inelastic neutron scattering (INS) spectra. It consists of three modules. The "Predictor" module uses a symmetry-aware neural network (coupled with an autoencoder) [1-3] to perform direct prediction of total/partial phonon density of states and powder 1D/2D INS spectra from a given structure. The "DFT database" module uses pre-calculated force constants from density functional theory (DFT) [4] to perform INS simulations for single crystals and powders (for the crystals available in the database). The "MLFF" module uses pre-trained universal force fields [8-12] to perform structural optimization, phonon calculation, and INS simulations for single crystals and powders for any given crystal. The predicted/calculated results are saved in CSV files and can be visualized with the GUI. INSPIRED is developed to be a convenient tool for INS experimental planning, steering, and quick data analysis. This repository contains two files as an update to the previous version: 1. A tarball file (dftdb.tar.gz) containing the DFT database (currently with 12734 crystals) 2. A VirtualBox appliance file (inspired_vm.ova) to run INSPIRED as a virtual machine. The ML model file (model.tar.gz) remains the same and can be obtained from the previous version. Instructions on how to use these files, as well as the rest part of the software, can be found on the GitHub page.INSPIRED是一款图形用户界面（GUI），可实现声子与非弹性中子散射（INS）谱的快速预测与计算，包含三大功能模块： 1. “预测器”模块：采用具备对称性感知能力的神经网络（结合自编码器）[1-3]，可针对给定晶体结构直接预测总/分态声子态密度，以及粉末样品的一维/二维非弹性中子散射谱。 2. “DFT数据库”模块：依托密度泛函理论（DFT）[4]预先计算得到的力常数，可为数据库内收录的晶体开展单晶与粉末样品的非弹性中子散射模拟。 3. “机器学习力场（MLFF）”模块：使用预训练的通用力场[8-12]，可针对任意给定晶体开展结构优化、声子计算，以及单晶与粉末样品的非弹性中子散射模拟。 预测或计算得到的结果将保存为CSV文件，并可通过该图形界面进行可视化。开发INSPIRED的初衷，是为非弹性中子散射实验的规划、调控与快速数据分析提供一款便捷工具。 本仓库包含两份用于更新旧版本的文件： 1. 压缩包文件dftdb.tar.gz，内含DFT数据库（当前收录12734种晶体） 2. VirtualBox虚拟设备文件inspired_vm.ova，可通过虚拟机运行INSPIRED。 机器学习模型文件model.tar.gz未作改动，可从既往版本中获取。 关于上述文件以及软件其余功能的使用说明，可查阅该项目的GitHub页面。