XACS Workshop 2024

第三届化学键与AI分子模拟XACS研讨会在郑州成功举办

2024年7月1-5日，由郑州大学化学学院和福建省理论与计算化学重点实验室共同主办，XACS团队、郑州大学化学拔尖学生培养基地和厦门大学化学拔尖学生培养基地协办的 第三届化学键与AI分子模拟XACS研讨会 (XACSW-2024) 在郑州大学成功举办。 本次研讨会旨在满足广大理论与计算化学研究学者系统学习价键理论方法、能量分解分析方法和机器学习化学计算模拟的需求。 自2012年至今，该系列研讨会已成功举办了四届国际研讨班和七届国内研讨班。本次研讨会以化学键理论和AI分子模拟为主题，邀请国际知名理论化学专家授课，系统介绍价键理论、分子相互作用理论与分析方法以及AI加强计算化学，详细讲解XACS云平台功能，并提供上机实践及案例分析。此次研讨会吸引了来自全国各地38所高校和科研机构的100余名师生参加。

在本次研讨会上，厦门大学吴玮教授、苏培峰教授、Pavlo O. Dral教授、周晨副教授、应富鸣高级工程师系统讲授了价键理论、分子相互作用理论和机器学习基础知识并详细介绍了原子计算套件(XACS)云平台功能、以及该套件所包括的XMVB, XEDA和MLatom软件的应用实例。北卡罗来纳大学格林斯伯勒分校莫亦荣教授、陕西师范大学王长崴副教授、中南大学林旭辉副教授讲授了块定域波函数的基本原理和理论基础，并通过具体算例展示了其在实际计算中的应用。此外，研讨会还邀请了希伯来大学Sason Shaik教授、林奈大学Ran Friedman教授、厦门大学汪骋教授等国内外知名专家远程连线进行专题授课。

附1：入校证明下载: 入校证明

附2：参会证明下载: 参会证明

Introduction

• 1. HF and DFT Computation with XACS
• 2. Get started with MLatom@XACS
  • 2.1. Slides
  • 2.2. Video
  • 2.3. Geometry optimization

Intermolecular Interaction

• 1. Documentation of GAMESS-US Input
• 2. Practice in energy decomposition analysis with XEDA
  • 2.1. Slides
  • 2.2. Exercises
• 3. Practice with BLW
• 4. Practice in energy decomposition analysis with SAPT

Valence Bond Theory

• 1. Slides for Basic VB Exercises
• 2. The Understanding of H₂
  • 2.1. Introduction
  • 2.2. Computation of H₂ at Stationary Point
    • 2.2.1. Input File
    • 2.2.2. Output File
• 3. Computing of F₂
  • 3.1. Introduction
  • 3.2. Computations with 3 Structures at Stationary Point
    • 3.2.1. Input File
    • 3.2.2. Computational Results
  • 3.3. Computations with 3 Structures at Dissociation Limit
    • 3.3.1. Input File
    • 3.3.2. Computational Results
  • 3.4. Computations with Covalent Structure at Stationary Point
    • 3.4.1. Input File
    • 3.4.2. Computational Results
• 4. Resonance in C₆H₆
  • 4.1. Introduction
  • 4.2. Computations with 5 Covalent Structures
    • 4.2.1. Input File
    • 4.2.2. Computational Results
  • 4.3. Computations with 2 Kekulé Structures
    • 4.3.1. Input File
    • 4.3.2. Computational Results
  • 4.4. Computations with 1 Kekulé Structure
    • 4.4.1. Input File
    • 4.4.2. Computational Results
• 5. Computing of O₂
  • 5.1. Introduction
  • 5.2. Computations of the singlet state
    • 5.2.1. Input File
    • 5.2.2. Computational Results
  • 5.3. Computations of the triplet state
    • 5.3.1. Input File
    • 5.3.2. Computational Results
  • 5.4. Computations with selected structures
    • 5.4.1. Input File
    • 5.4.2. Computational Results
• 6. Slides for Advanced Exercises
• 7. Computing of post-VBSCF methods
  • 7.1. Introduction
  • 7.2. Computing with BOVB method
    • 7.2.1. Input File
    • 7.2.2. Computational Results
  • 7.3. Computing with VBPT2 method
    • 7.3.1. Input File
    • 7.3.2. Computational Results
  • 7.4. Computing with hc-DFVB method
    • 7.4.1. Input File
    • 7.4.2. Computational Results
• 8. Computing of diabatic states with VB theory
  • 8.1. Introduction
  • 8.2. Computations of adiabatic states
    • 8.2.1. Input file
    • 8.2.2. Computation of diabatic states with chemical insights
    • 8.2.3. Computation of diabatic states with VBCAD
• 9. Menshutkin Reaction NH₃ + CH₃Cl \(\rightarrow\) [NH₃CH₃]⁺ + Cl^-
  • 9.1. Chemical question to be addressed
  • 9.2. Active/inactive electrons and orbitals
  • 9.3. Fragmentation and important VB structures
  • 9.4. Basis set and its subsets based on orbital symmetry
  • 9.5. Structures
  • 9.6. Input File
  • 9.7. Computational Results
    • 9.7.1. The VB structures
    • 9.7.2. Energies
    • 9.7.3. VB wavefunction and weights
• 10. Charge-shift bonding in propellane
  • 10.1. Introduction
  • 10.2. Computations of the resonance energy
  • 10.3. Computations of the bonding energy

Machine learning

• 1. Machine learning basics
  • 1.1. Slides
  • 1.2. Power of ML
  • 1.3. Training your first ML model
  • 1.4. ML algorithms
• 2. ML for PES
  • 2.1. Slides
  • 2.2. Machine learning potentials
  • 2.3. Testing ML models
  • 2.4. Choosing ML model settings to make it work
  • 2.5. Importance of including forces in training MLPs
  • 2.6. Splitting
  • 2.7. Learning curves
• 3. Universal machine learning models
  • 3.1. Slides
  • 3.2. What are the universal ML models?
  • 3.3. Frequencies and thermochemistry
  • 3.4. Uncertainty quantification
  • 3.5. Qualitative problems with PES shape
• 4. Spectroscopy
  • 4.1. Slides
  • 4.2. (Ro)vibrational spectra
    • 4.2.1. IR from frequency calculations
    • 4.2.2. IR and power spectra from MD
  • 4.3. UV/vis absorption spectra
    • 4.3.1. UV/vis spectra from single-point calculations
    • 4.3.2. UV/vis spectra with nuclear-ensemble approach
  • 4.4. Emission spectra
  • 4.5. Two-photon absorption spectra
• 5. Transfer learning
  • 5.1. Slides
  • 5.2. Jupyter notebook
  • 5.3. Tutorial
  • 5.4. Transfer learning in MLatom
  • 5.5. ANI TL model
• 6. Molecular dynamics
  • 6.1. Slides
  • 6.2. What is MD and why is it important?
  • 6.3. Why is molecular dynamics expensive?
  • 6.4. How can ML accelerate MD?
    • 6.4.1. Accelerating calculation of forces
    • 6.4.2. Directly learning dynamics
  • 6.5. MD types
    • 6.5.1. Constant energy vs constant temperature
    • 6.5.2. Quasi-classical trajectory
    • 6.5.3. Other approaches you should know
  • 6.6. Further reading
• 7. Transition state search and analysis
  • 7.1. Slides
  • 7.2. TS optimization
  • 7.3. TS validation with normal modes
  • 7.4. Intrinsic reaction coordinate (IRC)
  • 7.5. TS analysis via downhill dynamics
  • 7.6. Quasiclassical trajectories analysis on the bifurcating pericyclic reaction
• 8. Active learning
  • 8.1. Slides
  • 8.2. Tutorial for physics-informed active learning
  • 8.3. Get started
  • 8.4. Initial data
  • 8.5. ML models
  • 8.6. Sampler
  • 8.7. Convergence options
  • 8.8. Distributing labeling on multiple nodes
  • 8.9. Putting it all together
  • 8.10. Restarting
  • 8.11. Using MLPs in simulations
• 9. Surface-hopping nonadiabatic dynamics
  • 9.1. Slides
  • 9.2. Machine learning
  • 9.3. Setup for calculations on your computer with AIQM1
  • 9.4. AIQM1
    • 9.4.1. Single-point excited-state calculations
    • 9.4.2. Dynamics
• 10. Delta-learning
  • 10.1. Slides
  • 10.2. Delta-learning operating principle
  • 10.3. Getting rid of a QM baseline: hierarchical ML

Videos

• 1. Videos for basic courses
• 2. Videos for advanced courses