RMG-Py Version 3.1.0

Date: April 23, 2021

We recommend creating a new conda environment using the latest environment.yml
as many dependencies have changed, and upgrading an existing environment is
always troublesome.

• RMG-Py

  • Added support for Bromine
  • Added improved method to calculate temperature dependent solvation free energy
  • Made Rank 1 accuracy correspond to 0.2 kcal/mol instead of 0 kcal/mol
  • Improvements to Group Additivity comments, in particular adding missing group comments
  • Added support for trimolecular units in ArrheniusBM fits
  • Improvements to profiling
  • Use kekulized structures for transport estimation
  • Automatic tree generation script improvements
  • Properly short circuit is_isomorphic when strict=False
  • Added block for specifying species tuples to react when starting an RMG run
  • Improve ArrheniusBM fitting to a single reaction
  • Improvements in bidentate thermochemistry estimation
  • Added new surface attributes for metals and facets
  • Added support for Phosphorus
  • Enable use LSRs to scale thermo from different metals and enable proper
    use of training reactions from different metals
  • Added maximumSurfaceSites constraint

• Arkane

  • Added frequency scaling factors for apfd/deef2tzvp and wb97xd/def2svp
  • Kinetics and pdep sensitivities additionally saved in YAML format
  • Enable automatic isodesmic reaction generation
  • AECs, BACs and frequency scaling factors moved from Arkane to RMG-database
  • Added functionality for Petersson and Melius BAC fitting using Arkane and
    the reference database
  • Enabled two parameter Arrhenius fit option
  • Added functionality for fitting AECs
  • Added classes to standardize model chemistry definitions
  • Use adjlists instead of smiles when saving

• Bugfixes

  • QMTP updated to work with g16 executable
  • Fixed various Sticking Coefficient bugs
  • Fixed issues with Surface Arrhenius reactions written in the reverse being converted
    to ArrheniusEP instead of SurfaceArrheniusBEP
  • Fixed NaN handling in the explorer tool's steady state solve
  • Fixed determine_qm_software for Orca
  • Fixed bug where elementary_high_p library reactions with more than the maximum number of atoms for pdep
    never entered the edge
  • Fixed bug related to pdep networks having sources not contained in the core
  • Fixed various profiling bugs
  • Fixed issue with indexing when merging models
  • Fixed bug with ranged liquid reactors
  • Fixed bug with loading of autogenerated trees in Arkane
  • Fixed bug related to collision limit violation checks in LiquidReactor
  • Fixed bug related to Pmin and Pmax definition in SurfaceReactor
  • Fixed bugs in global uncertainty analysis for LiquidReactor
  • Fixed bug related to the units of reverse rate constants for reactions involving surface species
  • Fixed bug in Molecule isomorphism where it would simply assume the given initial map was correct
  • Remove deprecated matplotlib warn keyword
  • Fixed bug related to reading Chebyshev forms in Chemkin files
  • Fixed reference concentration for surface species when calculating Kc
  • Fixed issue with the reaction generation using the reversee of Surface_ElleyRideal_Addition_MultipleBond
  • Fixed bug with adjlist multiplicitly line being mistaken as the species name
  • Fixed bug with the library to training notebook
  • Remove temporary seed mechanisms if they exist from a previous run

• Miscellaneous

  • Modified find_parameter_sources_and_assign_uncertainties to regenerate chem.inp as needed
  • Added option to save atom order when labeling template reactions
  • Added option to ignore atom type errors when creating molecule objects
  • Enable use of critical_distance_factor in from_xyz
  • Improved SIGINT handling when calling lpsolve
  • Enable H-bond drawing
  • Improvements to debug messages
  • Updated dependencies cclib and OpenBabel

Note that the upgrade to OpenBabel v3+ will change the interpretation
of some ambiguous SMILES strings that use the lower-case aromatics notation.
Although we think the new interpretation is not wrong, it might be different
from previous versions, so take care.

This release represents a major milestone in RMG development and includes many backwards-incompatible changes, most notably Python 3 compatibility and major API changes. Users switching to RMG 3 will need to create new conda environments and update any scripts which access the API. We recommend using the futurize script from python-future for updating scripts for Python 3 and the provided rmg2to3.py script for updating scripts for RMG 3.

Python 3 #1724

• RMG is now compatible with Python 3.7 and newer
• RMG v2.x versions will no longer be supported
• API changes
  • Method, function, and argument names have been standardized to use snake_case across RMG and Arkane
  • Input file related code was not changed, in order to continue support for existing syntax
  • Conversion script has been provided to aid transition (scripts/rmg2to3.py)
  • Standardized submodule names in the rmgpy.tools module #1794
• Accompanying changes
  • Reduction and scoop_framework modules have been removed
  • New/updated hash and comparison methods for Species/Molecule/Atom/Bond classes
  • DDE thermochemistry estimator has been replaced by chemprop
  • Update example IPython notebooks #1735
  • Update global uncertainty module to work with MUQ 2 and Python 3 #1738
• Miscellaneous clean up and bug fixes following transition #1741, #1744, #1752, #1759, #1785, #1802, #1798, #1799, #1808

Arkane

• Improvements and refactoring of job output file creation and content #1607
• Fix kinetics fitting bug #1672
• Improvements to automatic network exploration tool #1647
• Support for ND classical and semi-classical rotor calculations #1640, #1849
• Support for 2D quantum mechanical rotor calculations using Q2DTor #1640
• Support for providing absolute file paths #1685
• Output RMG-style libraries #1769
• Check for error termination in Gaussian log files #1766
• Support for parsing Orca log files #1749
• Support for parsing MP2, double hybrid DFT, CCSD, and CCSD(T) energies from Gaussian log files #1815
• Support for TeraChem log files #1788
• Miscellaneous bug fixes #1810

New features and other additions

• Additional options for heterocycles in MLEstimator #1621
• Automatic tree generation algorithm implementation completed #1486, #1675, #1848
• New simulation restart approach using seed mechanisms (old pickle-based method removed) #1641
• Added new MBSampledReactor type for simulating molecular beam experiments (does not support model generation) #1669
• Improvements to group additivity thermo estimates for aromatics and sulfur species #1731, #1751
• Improvements to solvation correction determination with multiple resonance structures #1832
• Add support for reading and writing extended element syntax in Chemkin NASA polynomials #1636
• Add support for fitting negative Arrhenius rates (found in MultiArrhenius data) #1834

Bug fixes

• Fix numpy rcond usage to restore support for older numpy versions #1670
• Fix bug with duplicate library reactions when using RMG generated seed mechanisms #1676
• Move parse_command_line_arguments to facilitate importing in binary package #1717
• Fix issues with is_identical_to methods of kinetics models #1705
• Fix cython issue with make_object definitions #1817
• Fix issue with estimating solvation corrections for radicals #1773
• Fix parsing of certain types of RMG generated reaction comments #1842
• Fix identifier generation for surface species using OpenBabel #1842
• Fix mole fraction normalization for SimpleReactor #1809
• Fix permissions error when writing seed mechanisms in WSL #1796
• Fix issue with restarting from job without reaction filters #1847

Other

• Improvements to mergeModels.py script #1649
• Miscellaneous performance improvements #1677, #1765,
• Raise errors when NaN is encountered in solver #1679
• Allow sulfur species to have valence 12 in resonance algorithm #1751
• Add support for maxproc argument to generate_reactions module #1780
• Display atom index when drawing groups #1758
• Update sensitivity example #1805
• Update commented input file #1806
• Generate reverse reaction recipes in reverse order of the forward recipe #1829
• Add iodine to Chemkin elements list #1825
• Remove unnecessary duplicate checking for seed mechanisms #1824
• Organize examples for running RMG scripts #1840
• Increase RDKit version requirement to avoid memory leak #1851
• Logging changes #1721, #1755
• Documentation updates #1680, #1709, #1767, #1781, #1784, #1807, #1845

Thanks to all contributors: ajocher, alongd, amarkpayne, cgrambow, dranasinghe, hwpang, kspieks, goldmanm, mazeau, mjohnson541, mliu49, oscarwumit, rwest, rgillis8, sarakha, sudoursa, xiaoruiDong, yunsiechung, zjburas

• Bugfixes

  • Improve error handling in NASA as_dict method #1630
  • Fixes to Fluorine atomtypes #1656
  • Fix pressure dependent network generation #1658
  • Add support for reversing PDepArrhenius with MultiArrhenius rates #1659

• Arkane

  • Implement ZPE scaling factor #1619
  • Refactor infrastructure for bond additivity corrections #1605
  • Add frequency scale factors for wb97xd/def2tzvp and apfd/def2tzvpp #1653
  • Fix frequency scale factors in example files #1657
  • Get element counts from conformers #1651

• Miscellaneous

  • Update conda environment files #1623, #1644
  • Output RMS (Reaction Mechanism Simulator) format mechanism files #1629
  • Properly clean up files after running tests #1645
  • Documentation fixes #1650
  • Improve as_dict and make_object by making them recursive #1643

• Heterogeneous catalysis!
  • RMG-cat fork has been merged #1573
    • Introduce SurfaceReactor
    • Thermo estimation for adsorbed species
    • Surface reaction generation and kinetics estimation
  • Introduce Van der Waals bonds (order 0) and quadruple bonds (order 4) #1542
• Arkane
  • Automatically detect rotor symmetry #1526
  • Introduce new YAML files for storing and loading species statmech data #1402, #1551
  • Don't create species dictionary file if there are no structures #1528
  • Improvements to network explorer tool #1545
  • Improved class inheritance for quantum log file classes #1571
  • Automatic determination of optical isomers and symmetry using symmetry package #1571
  • Parse CCSD(T) energies from Molpro output #1592
  • Automatically determine molecule linearity #1601
  • Determine frequency scaling factor based on geom/freq method rather than sp method #1612
  • Improve logging related to energy barriers #1575
  • Ensure that translational mode is calculated for atoms #1620
• Miscellaneous features
  • New enumerate_bonds method of Molecule to generate dictionary of bond types #1525
  • Introduce RMGObject parent class to support YAML dumping and loading #1402, #1540
  • Add support for fluorine atomtypes #1543
  • Introduce ArrheniusBM class for Blower-Masel kinetics #1461
  • Allow defining and using co-solvents for solvent libraries #1558
  • Introduce strict option to perform isomorphism between species/molecules while ignoring electrons and bond orders #1329
  • Molecule and Species objects can be instantiated by providing SMILES or InChI argument directly, and the identifiers can be accessed via the SMILES and InChI attributes #1329
  • Parallelization has been completely refactored using Python multiprocessing module in replacement of scoop, currently supports parallel reaction generation and QMTP #1459
  • Improvements to usability of uncertainty analysis functionality #1593
• Bug fixes
  • Various fixes for supporting mono-atomic molecules in Arkane #1513, #1521
  • Ensure keras_backend is set consistently #1535
  • Fix handling of disconnected graphs in VF2 isomorphism algorithm #1538
  • Ignore hydrogen bonds when converting to RDKit molecule #1552
  • Other miscellaneous bugs #1546, #1556, #1593, #1600, #1622
• Backward incompatible changes
  • Hydrogen bonds are now order 0.1 (instead of 0) #1542
• New dependencies
  • pyyaml (required) #1402
  • scikit-learn (required) #1461
  • textgenrnn (optional) #1573
• Other
  • Windows binaries are no longer officially supported. The new recommended way to use RMG on Windows computers is via a virtual machine or through the Linux subsystem. See documentation for updated installation instructions. #1531, #1534
  • Documentation updates #1544, #1567
  • Logging/exception improvements #1538, #1562
  • PEP-8 improvements #1566, #1592, #1596
  • Solver output files (png/csv) now report moles instead of mole fractions #1542
  • Replace global RMGDatabase object if the database is reloaded #1565
  • Print ML generated quote upon completion of RMG jobs #1573
  • Infrastructure for automatically generated reaction rate trees #1461
  • Testing related changes #1597, #1599
  • Updates to example Jupyter notebooks #1541, #1593

• Arkane (formerly CanTherm):

  • CanTherm has been renamed to Arkane (Automated Reaction Kinetics And Network Exploration)
  • New network exploration functionality using RMG-database
  • Support for all elements has been added for reading quantum output files
  • New supporting information output file with rotational constants and frequencies
  • Known thermo and kinetics can be provided in addition to quantum information
  • Improve general user experience and error handling

• New machine learning thermo estimator

  • Estimate species thermochemistry using a graph convolutional neural network
  • Estimator trained on quantum calculations at B3LYP and CCSD(T)-F12 levels
  • Currently supports C/H/O/N, with an emphasis on cyclic molecules

• Resonance:

  • New pathways added for lone-pair multiple-bond resonance, replacing
    two pathways which were more specific
  • New pathways added for aryne resonance
  • Aromatic resonance pathways simplified and refactored to use filtration
  • Kekule structures are now considered unreactive structures

• Miscellaneous changes:

  • Isotope support added for reading and writing InChI strings
  • New branching algorithm for picking up feedback loops implemented (beta)
  • Global forbidden structure checking is now only done for core species for
    efficiency, which may lead to forbidden species existing in the edge
  • Minor improvements to symmetry algorithm to fix a few incorrect cases

• Bug fixes:

  • Fixed issue where react flags were being reset when filterReactions was
    used with multiple reactors, resulting in no reactions generated
  • File paths for collision violators log changed to output directory
  • Fixed bug in local uncertainty introduced by ranged reactor changes
  • Fixed bug with diffusion limitation calculations for multi-molecular reactions
  • Various other minor fixes

This release is minor patch which fixes a number of issues discovered after 2.2.0.

• Collision limit checking:

  • RMG will now output a list of collision limit violations for the generated model

• Fixes:

  • Ambiguous chemical formulas in SMILES lookup leading to incorrect SMILES generation
  • Fixed issue with reading geometries from QChem output files
  • React flags for reaction filter were not properly updated on each iteration
  • Fixed issue with inconsistent symmetry number calculation

• New features:

  • New ring membership attribute added to atoms. Can be specified in group adjacency lists in order to enforce ring membership of atoms during subgraph matching.
  • Reactors now support specification of T, P, X ranges. Different conditions are sampled on each iteration to optimally capture the full parameter space.
  • New termination type! Termination rate ratio stops the simulation when the characteristic rate falls to the specified fraction of the maximum characteristic rate. Currently not recommended for systems with two-stage ignition.
  • New resonance transitions implemented for species with lone pairs (particularly N and S containing species). A filtration algorithm was also added to select only the most representative structures.
  • Formal support for trimolecular reaction families.
  • New isotopes module allows post-processing of RMG mechanisms to generate a mechanism with isotopic labeling.

• Changes:

  • Library reactions can now be integrated into RMG pdep networks if the new elementary_high_p attribute is True
  • Library reactions may be duplicated by pdep reactions if the new allow_pdep_route attribute is True
  • Jupyter notebook for adding new training reactions has been revamped and is now located at ipython/kinetics_library_to_training.ipynb
  • Syntax for recommended families has changed to set notation instead of dictionaries, old style still compatible
  • Ranking system for database entries expanded to new 0-11 system from the old 0-5 system
  • Collision limit checking has been added for database entries

• Cantherm:

  • Improved support for MolPro output files
  • Added iodine support
  • Automatically read spin multiplicity from quantum output
  • Automatically assign frequency scale factor for supported model chemistries
  • Plot calculated rates and thermo by default
  • New sensitivity analysis feature analyzes sensitivity of reaction rates to isomer/TS energies in pdep networks

• Fixes:

  • Properly update charges when creating product templates in reaction families
  • Excessive duplicate reactions from different resonance structures has been fixed (bug introduced in 2.1.3)
  • Fixed rate calculation for MultiPdepArrhenius objects when member rates have different plists

• A more formal deprecation process is now being trialed. Deprecation warnings have been added to functions to be removed in version 2.3.0:

  • All methods related to saving or reading RMG-Java databases and old-style adjacency lists
  • The group additivity method for kinetics estimation (unrelated to thermo group additivity)
  • The saveRestartPeriod option and the old method of saving restart files

• Cantherm:

  • Atom counts are no longer necessary in input files and are automatically determined from geometries
  • Custom atom energies can now be specified in input files
  • Removed atom energies for a few ambiguous model chemistries
  • Add atom energies for B3LYP/6-311+g(3df,2p)

• Changes:

  • Refactored molecule.parser and molecule.generator modules into molecule.converter and molecule.translator to improve code organization
  • SMILES generation now outputs canonical SMILES
  • Molecule.sortAtoms method restored for deterministic atom order
  • PDep reactions which match an existing library reaction are no longer added to the model

• Fixes:

  • Fix issue with reaction filter initiation when using seed mechanisms

• New features:

  • Chlorine and iodine atom types have been added, bringing support for these elements to RMG-database
  • Forbidden structures now support Molecule and Species definitions in addition to Group definitions

• Changes:

  • Reaction pair generation will now fall back to generic method instead of raising an exception
  • Removed sensitivity.py script since it was effectively a duplicate of simulate.py
  • Thermo jobs in Cantherm now output a species dictionary
  • Fitted atom energy corrections added for B3LYP/6-31g**
  • Initial framework added for hydrogen bonding
  • Renamed molepro module and associated classes to molpro (MolPro) to match actual spelling of the program
  • Chemkin module is now cythonized to improve performance

• Fixes:

  • Allow delocalization of triradicals to prevent hysteresis in resonance structure generation
  • Fix reaction comment parsing issue with uncertainty analysis
  • Fix numerical issue causing a number of pressure dependent RMG jobs to crash
  • Template reactions from seed mechanisms are now loaded as library reactions if the original family is not loaded
  • Fix issues with degeneracy calculation for identical reactants

• Charged atom types:

  • Atom types now have a charge attribute to cover a wider range of species
  • New atom types added for nitrogen and sulfur groups
  • Carbon and oxygen atom types renamed following new valence based naming scheme

• Ring perception:

  • Ring perception methods in the Graph class now use RingDecomposerLib
  • This includes the getSmallestSetOfSmallestRings methods and a newly added getRelevantCycles method
  • The set of relevant cycles is unique and generally more useful for chemical graphs
  • This also fixes inaccuracies with the original SSSR method

• Other changes:

  • Automatically load reaction libraries when using a seed mechanism
  • Default kinetics estimator has been changed to rate rules instead of group additivity
  • Kinetics families can now be set to be irreversible
  • Model enlargement now occurs after each reactor simulation rather than after all of them
  • Updated bond additivity corrections for CBS-QB3 in Cantherm

• Fixes:

  • Do not print SMILES when raising AtomTypeError to avoid further exceptions
  • Do not recalculate thermo if a species already has it
  • Fixes to parsing of family names in seed mechanisms