221020 Meeting ideas & protocol

[ehhartmann]

Introduction

• what are people working on (not necessarily related to KIMMDY)?

New Features since March

• reaction plugin architecture
• dummy HAT reaction
• HAT NN reaction
• changemanager for HAT reaction
• changemanager tests, some general tests
• file history, easy file handling
• more general mdmanager
• relaxation by short MD
• automated documentation

Feature debate

• mdmanager: which command line options are always used, which optional?
• relaxation: parallel existence of multiple schemes desirable? maintaining detailed balance and physical behavior
• HAT NN reaction: if we decide for a reaction, should the geometry change end at exactly this position?
• MC theory: how do we get reaction rates for competing reactions? attempt frequency, multiple snapshots

General

• when to release KIMMDY v2?
• rename modified force field and remove deficient versions

[jmbuhr]

• more general changemanger not reliant on hydrogens i.e. safe handing of topology changes

[jmbuhr]

• Checkpoint files and restarts for kimmdy (and internal md runs?)

[jmbuhr]

[KRiedmiller]

Protocol

• How to get prefactor? QM/MM? pure QM?
• Checkpoints for testing of individual parts
• cmdline for gmx:
  • specify everything all the time
  • no defname
  • files not found are no issue, as long all necessary information are somewhere
• MC w/ rejection:
  • cutoff in time domain
• How to mix reactions?
  • option 1: Only do one kind at a time
  • option 2: Competing rates need to be evaluated at the same time (as we currently do)
  • Competing rates can have diff prefactors
• maybe recalculate prefactor for homolysis in QMMM together with HAT QMMM
• KIMMDY Checkpoints:
  • save task queue maybe as pickle
  • gmx restarts automatically if chk is found
  • option: save to new config, problem: objects in queue
  • save current state (gmx running, or predicting rates..)
  • save runmanager.latest_files
• How to jump:
  • morse does the job
  • teleport H to ideal position, copy v of connected atom
  • lambda dynamics of bonds, can be reversible