Reactionmechanismgenerator

- Accelerator tools:
- Dynamics criterion provides another method to expand the mechanism by adding reactions to the core
- Surface algorithm enables better control of species movement to the core when using the dynamics criterion
- Multiple sets of model parameters can now be specified in a input file to allow different stages of model generation
- A species number termination criterion can now be set to limit model size
- Multiple items can now be added per iteration to speed up model construction
- New ModelSettings and SimulatorSettings classes for storing input parameters

- New features:
- Kinetics libraries can now be automatically generated during RMG runs to be used as seeds for subsequent runs
- Loading automatically generated seed mechanisms recreates the original template reaction objects to allow restarting runs from the seed mechanism
- Carbene constraints can now be set in the species constraint block using maxSingletCarbenes and maxCarbeneRadicals
- Chirality is now considered for determining symmetry numbers
- Thermodynamic pruning has been added to allow removal of edge species with unfavorable free energy (beta)

- Other changes:
- RMG-Py exception classes have been consolidated in the rmgpy.exceptions module
- Species labels will now inherit the label from a matched thermo library entry
- Sensitivity analysis is now available for LiquidReactor

- Fixes:
- Fixed sensitivity analysis following changes to the simulate method
- Add memory handling when generating collision matrix for pressure dependence
- Improved error checking for MOPAC
- Prevent infinite loops when retrieving thermo groups

- Known issues:
- Seed mechanisms cannot be loaded if the database settings are different from the original ones used to generate the seed

- Thermo central database:
- Framework for tracking and submitting species to a central database have been added
- Following species submission, the central database will queue and submit quantum chemistry jobs for thermochemistry calculation
- This is an initial step towards self-improving thermochemistry prediction

- Rotor handling in Cantherm:
- Free rotors can now be specified
- Limit number of terms used when fitting hinder rotor scans
- Fixed bug with ZPE calculation when using hindered rotors

- New reaction degeneracy algorithm:
- Use atom ID's to distinguish degenerate reactions from duplicates due to other factors
- Degeneracy calculation now operates across all families rather than within each separately
- Multiple transition states are now identified based on template comparisons and kept as duplicate reactions

- Nodal distances:
- Distances can now be assigned to trees in reaction families
- This enables better rate averages with multiple trees
- Fixed bug with finding the closest rate rule in the tree

- New features:
- Added methods for automatically writing RMG-database files
- New symmetry algorithm improves symmetry number calculations for resonant and cyclic species
- Group additivity algorithm updated to apply new long distance corrections
- Specific colliders can now be specified for pressure-dependent rates
- Very short superminimal example added (hydrogen oxidation) for checking basic RMG operation
- Cantera now outputs a Chemkin file which can be directly imported into Chemkin

- Fixes:
- Fixed bug with negative activation energies when using Evans-Polanyi rates
- Fixed walltime specification from command line when running RMG
- Fixes and unit tests added for diffusionLimited module

- Known issues:
- The multiple transition state algorithm can result in undesired duplicate reactions for reactants with multiple resonance structures

- Improvements:
- New nitrogen atom types
- Kinetics libraries can now be specified as a list of strings in the input file
- New script to generate output HTML locally: generateChemkinHTML.py
- New kekulization module replaces RDKit for generating Kekule structures
- Benzene bonds can now be reacted in reaction families
- Removed cantherm.geometry module due to redundancy with statmech.conformer

- Fixes:
- Reaction direction is now more deterministic after accounting for floating point error
- Multiple bugs with resonance structure generation for aromatics have been addressed

- Uncertainty analysis:
- Local and global uncertainty analysis now available for RMG-generated models
- Global uncertainty analysis uses MIT Uncertainty Quantification library, currently only supported on Linux systems
- Examples for each module are available in localUncertainty.ipynb and globalUncertainty.ipynb

- Fixes:
- Clar structure generation no longer intercepts signals
- Fixes to SMILES generation
- Fix default spin state of [CH]

This release contains several new features and bug-fixes.

- Clar structure generation
- optimizes the aromatic isomer representations in RMG
- lays the foundations for future development of poly-aromatic kinetics reaction families

- Flux pathway analysis
- introduces an ipython notebook for post-generatation pathway analysis (``ipython.mechanism_analyzer.ipynb``)
- visualizes reactions and provides flux statistics in a more transparent way

- Cantera mechanism
- automatically writes cantera version of RMG-generated mechanism at the end of RMG jobs

- Fixes bugs
- upgrades ``pruning`` to fix new memory leaks introduced by recent functionalities
- fixes the bug of duplicated species creation caused by ``getThermoData`` removing isomers unexpectedly
- fixes restart file generation and parsing problems and users can choose restart mode again
- upgrades bicyclic decomposition method such that more deterministic behaviors are ensured
- change bond order type to float from string to improve RMG's symmetry calculation for species with multiple resonance structures

This release includes several milestones of RMG project (see installation [here](http://reactionmechanismgenerator.github.io/RMG-Py/users/rmg/installation/index.html)):
- Parallelization finally introduced in RMG:
- Generates reactions during `enlarge` step in parallel fashion (`rmgpy.rmg.react`)
- Enables concurrent computing for QMTP thermochemistry calculations (`rmgpy.thermo.thermoengine`)
- Instructions of running RMG parallel mode can be found [`here for SLURM scheduler`](https://github.com/ReactionMechanismGenerator/RMG-Py/wiki/Running-RMG-in-parallel-with-a-SLURM-scheduler) and [`here for SGE scheduler`](https://github.com/ReactionMechanismGenerator/RMG-Py/wiki/Running-RMG-in-parallel-with-a-SGE-scheduler).
- Polycyclic thermochemistry estimation improved:
- Extends group additivity method for polycyclics and estimates polycyclics of any large sizes by a heuristic method (bicyclics decomposition)
- New tree averaging for kinetics:
- Fixes previous issue of imcomplete generation of cross-level rate rules
- Implements Euclidean distance algorithm for the selection of the best rate rules to use in `estimateKinetics`
- Streamlines storage of kinetics comments for averaged rules, which can be analyzed by `extractSourceFromComments`
- Database entry accessibility tests:
- Adds entry accessibility tests for future entries (`testing.databaseTest`)
- Fixes bugs
- fluxdiagram generation is now fixed, one can use it to generate short video of fluxdigram evolution
- mac environment yml file is introduced to make sure smooth RMG-Py installation and jobs on mac
- fixes failure of `checkForExistingSpecies` for polyaromatics species
- fixes execution failure when both pruning and pDep are turned on
- fixes pDep irreversible reactions
- fixes issue of valency of `Cbf` atom by dynamic benzene bond order assignment