changelog.rst

Changelog

1.4.4

---

Functionality improvements

• Improved constrained optimisation (CRFOptimiser) and handling of multiple constraints
• Adds compatability with numpy v2.0
• Improved implementation of the RFO-TRM (QAOptimiser) optimiser that can handle constraints
• Added static internal back-transform and damping for faster and easier DIC to Cartesian coordinate transformation

Bug Fixes

• DIC to Cartesian transform will now always use PIC.close_to() to ensure steps along dihedral have the smallest change, even after back-transform is complete

Usability improvements/Changes

• Optimiser convergence criteria have been improved to consider energy change, RMS and max. gradient and step sizes.

1.4.3

---

Functionality improvements

• DHS and DHS-GS can now switch between two step sizes
• Peak detection in bracket methods now uses cubic polynomial fit with energies and gradients instead of only projecting gradients

Bug Fixes

• Fixes no solvent being added in QRC calculations
• Fixes cases where .xyz files are printed with no space between coordinates when the coordinate value is large.
• Fixes DHS and DHS-GS methods ignoring number of cores

Usability improvements/Changes

• Added consistent aliases for double dagger across all energies in autode.reaction.delta
• Optimiser trajectories are saved on disk instead of keeping completely in memory
• Hessian updates the refactored into OptCoordinates class

1.4.2

---

Functionality improvements

• Replaces hard-coded derivatives for primitive internal coordinates with automatic differentiation
• More comprehensive primitive internal coordinate generation
• Adds the capacity to handle linear molecules in internal coordinate system

Bug Fixes

• Fixes triangular rings being incorrectly treated as dihedral angles

Usability improvements/Changes

• Faster import of autode package by lazily loading matplotlib
• ORCA output files copied after a calculation are configurable

1.4.1

---

Functionality improvements

• Adds the thermochemistry method from A. Otlyotov, Y. Minenkov in https://doi.org/10.1002/jcc.27129
• Adds the improved Elastic Image Pair (i-EIP) method for double-ended transition state search
• Adds a autode.Species.solvent_name property and setter for setting solvents from a string
• Enables reloading molecules from xyz files with their previously defined charge/multiplicity/solvent

Bug Fixes

• Fixes Hessian extraction in some G16 output files
• Fixes large step sizes in DHSGS
• Fixes the ability to define both a SMILES string and a xyz file without raising an exception

1.4.0

---

Functionality improvements

• Adds temporary_config() context manager for temporary configuration changes
• Adds a ForceConstant value
• Adds a more robust hybrid RFO-TRM geometry optimiser with trust radius update and a feature to detect oscillation and damp it
• Adds Dewar-Healy-Stewart (DHS) method and its variant DHS-GS to find transition states from initial and final geometries

Usability improvements/Changes

• Adds full usability of autodE on Windows, including parallelisation with loky
• Optional timeout for graph isomorphism test in Windows, turned on by Config.use_experimental_timeout=True (default behaviour kept for Linux/macOS)
• The electronic temperature and the version of parameterisation for xTB calculations are made configurable
• A NEB Image now derives from a Species superclass
• Modifies NEB Image constructor to be formed from an image
• Defines named constructors (from_end_points(...), from_list(...)) for NEB
• Removes NEB().contains_peak() in favour of NEB().images.contains_peak
• Modifies the CImages constructor to ensure it's constructed from an Images instance
• Removes NEB.get_species_saddle_point() in favour of NEB.peak_species
• Defines named constructors (from_end_points(...), from_list(...)) for NEB
• Removes mol_graphs.get_atom_ids_sorted_type()

Bug Fixes

• Fixes pickling issue with autode.config.Config on Windows and in multiprocessing spawn for Linux/macOS
• Fixes RFO Hessian update bug
• Fixes QChem not using Config.max_core

1.3.5

---

Usability improvements/Changes

• autode.value.ValueArray.to() now defaults to copying the object rather than inplace modification

Functionality improvements

• Adds a to_ method to autode.value.ValueArray for explicit inplace modification of the array

Bug Fixes

• Fixes ERROR logging level being ignored from environment variable AUTODE_LOG_LEVEL
• Fixes autode.values.Value instances generating items with units on division, and throw a warning if multiplying
• Fixes the printing of cartesian constraints in XTB input files, meaning they are no longer ignored
• Fixes Hessian instances changing units when normal modes are calculated
• Fixes an incorrect alias for ev_per_ang

1.3.4

---

Feature additions.

Usability improvements/Changes

• Throw useful exception for invalid ade.Config.ts_template_folder_path

Functionality improvements

• Adds ade.transition_states.TransitionState.from_species method to construct transition states from a species or molecule
• Adds autode.Reaction.save() and autode.Reaction.load() to save and reload a reaction state
• Adds saving checkpoints of a reaction during autode.Reaction.calculate_reaction_profile

Bug Fixes

• Fixes calculation clean_up() failing with a null filename

1.3.3

---

Bugfix release.

Functionality improvements

• Adds skipping Hessian re-evaluation when using autodE optimisers if a molecules has a Hessian calculated at the same level of theory
• Adds a Hessian recalculation frequency to autode.optimisers.PRFOptimiser
• Improves the default step size for TS optimising to be consistent with the ORCA default

Bug Fixes

• Adds checking of SMILES-defined charge against the user-specified value
• Fixes autode.optimisers.CRFOptimiser building incomplete internal coordinates for partially or completely fragmented molecules

1.3.2

---

Bugfix release.

Usability improvements/Changes

• Removes autode.geom.get_distance_constraints in favour of a better named method TSBase().active_bond_constraints

Bug Fixes

• autode.transition_states.ts_guess.TSguess.from_species now inherits solvent from the species
• Fixes a possible race condition in I/O for XTB conformer optimisations

1.3.1

---

Bugfix release.

Bug Fixes

• Fixes behaviour of autode.utils.work_in_tmp_dir and autode.utils.work_in decorators
• Fixes an exception being raised when autode.Calculation.clean_up is called with a method that doesn't implement external I/O
• Fixes autodE driven optimisations skipping execution when the input but not name changes

1.3.0

---

Optimisation features, graph assignment improvements and bugfixes.

Usability improvements/Changes

• Defines dummy atoms to have zero covalent and vdW radii
• Renames Method().available to Method().is_available
• Removes autode.bonds.get_ideal_bond_length_matrix and autode.bonds.get_avg_bond_length
• Removes autode.geom.rotate_columns
• Modifies the names of most optimiser classes e.g. autode.opt.optimisers.PRFOOptimiser -> PRFOptimiser
• Simplifies initialising a autode.calculations.Calculation by:
  • Requiring constraints to be attributed to a molecule
  • Removing the bond_ids_to_add argument and using the labeled graph instead (active edges)
  • Removing the other_input_block argument and appending to the keywords instead
• Removes autode.calculations.Calculation.print_final_output_lines in favour of a method on calculation.output
• Makes many methods in autode.calculations.Calculation private
• Deprecates all autode.calculations.Calculation.get_<method> methods in favour of setting properties of the input molecule
• Returns None rather than rasies exceptions when calling the (deprecated) calculation get methods, to be consistent with ...get_energy
• Adds an autode.wrappers.keywords package to improve file structure
• Removes any exceptions on calling .run() on an optimiser instance where the system has no degrees of freedom
• Removes support for Python < v3.8
• Tweaks the default ORCA TS optimisation keywords to be more conservative, i.e. slower and more accurate

Functionality improvements

• Adds a autode.atoms.Atom.covalent_radius property
• Adds a autode.atoms.Atoms.eqm_bond_distance method for the equilibrium bonded distance between two atoms
• Adds vibrational frequency scaling through both autode.Config.freq_scale_factor and a default value in wrapped functional keywords
• Adds a much more robust constrained rational function constrained optimiser in delocalised internal coordinates (DIC)
• Adds bond angle and dihedral primitive coordinates which can form part of the DIC set
• Improves the back transformation
• Adds an optional callback argument to autode.opt.optimisers.base.Optimiser for running custom functions after every optimisation step
• Adds the ability to save/reload an autode.opt.optimisers.NDOptimiser instance to/from a file
• Adds a solvent attribute to a autode.transition_states.transition_state.TransitionState constructor
• Adds functionality to partition a nudged elastic band into images where the maximum atom-atom distance between images is below a threshold
• Adds a sequential adapt+NEB TS finding method where a pure adapt. path fails to generate a geometry close enough to the TS for a successful TS optimisation

Bug Fixes

• Fixes variable harmonic frequencies (<2 cm-1 differences) due to projection vectors becoming close to rotational axes
• Fixes the extraction of atomic partial charges from ORCA output files
• Fixes gradients and Hessians not being reset on a molecule where the coordinates change
• Fixes unhelpful exception when calculating thermochemistry with EST methods without implemented "get_hessian" methods

See the table below for a quick benchmark of constrained optimisations in autodE compared to ORCA. In all cases the structures were generated from SMILES strings (RDKit) and optimised with a single constraint on the (0,1) distance of +0.1 Å from its current value.

Molecule	autodE	ORCA
C7H12	6	7
C3H7	8	17
C4H6	3	5
CClH3	3	4
C2H3O2	3	7
C2FH5	3	5
C4H6O2S	6	11

1.2.3

---

Minor functionality improvements and bugfixes.

Usability improvements/Changes

• All exceptions now inherit from a base autode.exceptions.AutodeException
• Fixes a typo in autode.exceptions.UnsupoportedCalculationInput
• Adds documentation explaining the intention of each exception in autode.exceptions
• Molecular graphs are now 'laziliy-loaded' i.e. generated once when the property is accessed

Functionality improvements

• Adds the ability to define atom classes for molecules in turn allowing for identity reactions to be calculated

Bug Fixes

• Fixes clashing names for a reaction initialised explicitly from molecules without defined names

1.2.2

---

Bugfix release.

Bug Fixes

• Fixes output redirection from XTB calculations resulting in missed lines on Mac

1.2.1

---

Bugfix release.

Functionality improvements

• Adds autode.mol_graphs.MolecularGraph (subclass of networkx.Graph) with a expected_planar_geometry method
• Adds an are_planar method to autode.atoms.Atoms

Bug Fixes

• Fixes a bug in assigning reasonable geometries which, in turn, could lead to skipped single point energy evaluations

1.2.0

---

Adds optimisation algorithms experimental explicit solvation, improves potential energy surface module as well as an array of usability improvements.

Usability improvements/Changes

• Adds more type hints and documentation
• Updates the TS template saved in the default template library
• Adds a setter for autode.species.Species.solvent so mol.solvent = 'water' will assign a autode.solvent.Solvent
• Removes autode.calculation.CalculationInput.solvent as an attribute in favour of using the molecule's solvent
• Removes autode.calculation.get_solvent_name in favour of a molecule check
• Removes autode.species.molecule.reactant_to_product in favour of a to_product() method for autode.species.molecule.Reactant (and likewise with a Reactant)
• Removes partially implemented autode.species.molecule.SolvatedMolecule and autode.species.complex.SolvatedReactantComplex as the type of solvation (implicit/explicit) should be a property of the solvent and not the molecule
• Removes autode.reactions.Reaction.calc_deltaXXX in favour of autode.reactions.Reaction.delta()
• Refactors classes to place constructors at the top
• Removes autode.values.PlottedEnergy as an estimated attribute is useful for all energies, not just those that are plotted
• Removes autode.reactions.Reaction.find_lowest_energy_ts as the function is not well named and can be replaced by a autode.reactions.Reaction.ts property
• Adds autode.transition_states.TransitionStates as a wrapper for TSs, much like autode.conformers.Conformers
• Updates autode.solvent.solvents.get_solvent to require specifying either an implicit or explicit solvent
• Improves validation of distance constraints and adds invariance to the key order i.e. autode.constraints.distance[(0, 1)] == autode.constraints.distance[(1, 0)]
• Removes autode.KcalMol and KjMol and enables a reaction to be plotted using a string representation of the units.
• Allows for keywords to be set using just a list or a string, rather than requiring a specific type
• Changes autode.wrappers.keywords.Keyword.has_only_name to a property
• Modifies the constructor of autode.species.molecule.Molecule to allow for a name to be specified when initialising from a .xyz file
• Modifies autode.calculation.Calculation.get_energy to raise an exception if the energy cannot be extracted
• Adds a runtime error if e.g. autode.calculation.Calculation.get_energy is called on a calculation that has not been run
• Skips low-level adaptive path searching if the high and low-level methods are identical (when XTB or MOPAC are not installed)
• Adds a default set of low-level single point keywords
• Adds a flag to override exiting a reaction profile calculation when association complexes are generated
• Adds a check that a calculation isn't going to exceed the maximum amount of physical memory on the computer

Functionality improvements

• Adds a selection of molecule optimisers to locate minima and transition states
• Refactors autode.smiles.angles to use unique class names (preventing overlap with e.g. autode.values.Angle)
• Adds a autode.solvent.Solvent.dielectric property for a solvent's dielectric constant
• Adds a autode.solvent.Solvent.is_implicit property
• Adds methods (e.g. translate and rotate) to autode.point_charges.PointCharge
• Adds checking that both high and low-level electronic structure methods are available before running autode.reaction.Reaction.calculate_reaction_profile or calculate_reaction_profile
• Adds a more robust explicit solvation generation (autode.species.molecule.Molecule.explicitly_solvate())
• Removes criteria on using a TS template with large distance differences between the structure and the template in favour of running sequential constrained optimisations to the required point
• Rewrites autode.pes into a consistent module while maintaining much of the functionality. Simplifies the interface
• Adds a QChem electronic structure method wrapper
• Adds autode.species.Species.calc_hessian to calculate either an analytic or numerical Hessian (in parallel)
• Adds image dependent pair potential (IDPP) relaxation improved interpolated geometries
• Adds autode.hessians.HybridHessianCalculator to calculate numerical Hessians at two levels of theory

Bug Fixes

• Updates the TS template saved in the default template library
• Reloads output file lines from a failed then re-run calculation
• Fixes Hessian extractions from some Gaussian output files

1.1.3

---

Usability improvements

Usability improvements/Changes

• Improves consistency and behaviour of calc_thermo method of a species, allowing for keywords and non-run calculations
• Allows for a non-fork multiprocessing 'start_method'

1.1.2

---

Bugfixes

Usability improvements/Changes

• Fixes typo in autode.exceptions.ReactionFormationFalied

Bug Fixes

• Fixes a bug where rings containing mostly double bonds failed to build with autode.smiles.builder.Builder
• Fixes using XTB as a high-level method with the xtb-gaussian wrapper (thanks @kjelljorner)

1.1.1

---

Documentation and typing hints

Usability improvements/Changes

• Adds typing to user-facing functions
• Adds autode.config.location to easily locate the core configuration file for permanent editing
• Updates documentation for readability
• Ensures units are kept if constructing a Value from a Value (i.e. Value(x), when x is a Value)

Functionality improvements

• Changes Keyword to an abstract base class
• Improves speed of Species rotation (numpy rather than a Python for loop)

Bug Fixes

• Fixes bug where NCI conformers were generated with the same name thus did not optimise uniquely (introduced in v.1.1.0)

1.1.0

---

API improvements that broadly maintain backwards compatibility.

Usability improvements/Changes

• Adds more argument and return types
• Changes AtomCollection.atoms to a property for more flexible sub-classing
• Changes ElectronicStructureMethod.doi_str and Keyword.doi_str to properties
• Adds interpretable repr(Species)
• Species.energies is zeroed when the Species.atoms are reset or change
• Species.energy is a property of the last computed energy on that species
• Species.is_linear now uses an angle tolerance to determine linearity, which is slightly tighter than the previous float-based tolerance
• Removes CalculationOutput.set_lines in favour of a cached file_lines property to avoid set_file_lines()
• Removes CalculationOutput.get_free_energy() in favour of Species.free_energy once a Hessian is set for a molecule and similarly with CalculationOutput.get_enthalpy()
• Removes CalculationOutput.get_imaginary_freqs() (now Species.imaginary_frequencies) and CalculationOutput.get_normal_mode_displacements() (now Species.normal_mode())
• Species.imaginary_frequencies now returns None rather than an empty list for a species without any imaginary frequencies, to be consistent with other properties
• Changes CalculationOutput.terminated_normally() to a property (CalculationOutput.terminated_normally)
• Removes Reaction.find_complexes in favour of setting the reactant and product complexes dynamically, unless Reaction.calculate_complexes is called to find association complexes
• Tweaks the default relative tolerance on bonds to account for M-X agostic interactions lengthening bonds
• Enables Species.atoms to be added, even if they are None
• Improved atom setting of Complex.atoms
• Changes Complex.get_atom_indexes() to Complex.atom_indexes()
• Changes Complex.molecules to a private attribute as the atoms/energy/gradient is not propagated
• Allows for Species.translate() and Species.rotate() to be called using vectors as lists or tuples rather than just numpy arrays
• Modifies get_truncated_complex() to get_truncated_species() and changes the return type to a species to reflect a possibly different molecular composition of the complex
• Improves peak checking in adaptive path TS guess generation
• Removes autode.atoms.get_thing() functions, in favour of Atom.thing
• Raises an exception if a single point energy evaluation fails to execute successfully
• Removes autode.conformers.conformer.get_conformer() in favour of a more flexible autode.conformer.Conformer constructor
• Adds Species.constraints that are used in optimisations (still available in Calculation initialisation)
• Adds Conformers to enable parallel electronic structure calculations across a set of conformers
• Improves readability of pruning of conformers based on RMSD and energy thresholds

Functionality improvements

• Adds angle and dihedral angle properties to an AtomCollection
• Improves and adds more Unit definitions
• Adds Value and ValueArray base classes for energies, gradients etc. These allow for implicit (1 Hartree == 617.509 kcal mol-1) comparisons and explicit conversion (1 Hartree).to('kcal')
• Adds further conversion factors to Constants
• Adds Species.energies as a container of all energies that have been calculated at a geometry
• Adds Keywords.bstring as a 'brief' summary of the keywords e.g. PBE0/def2-SVP and are associated with an Energy (a type of Value)
• Improves quick reaction coordinate characterisation of TSs by providing a maximum atomic displacement for improved initial structures
• Adds Hessian diagonalisation to obtain normal modes with and without translation and rotation projections for linear and non-linear molecules
• Adds Species.weight and Species.mass as equivalent properties for the molecular weight
• Improves dihedral sampling in molecule generation
• Adds atoms.remove_dummy() to remove all dummy atoms from a set
• Enables different force constants to be used in XTB constrained optimisations (Config.XTB.force_constant, which sets wrappers.XTB.XTB.force_constant)
• Adds Solvent.copy()
• Adds Species.reorder_atoms() to reorder the atoms in a species using a mapping
• Adds Config.ORCA.other_input_block to allow for a block of input to be printed in all ORCA input files
• Changes the loose optimisations to only use a maximum of 10 iterations. This is based on an analysis of 3500 ORCA

optimisations, which plateaus quickly:

suggesting a value of 10 is a appropriate. This will be system dependent and need increasing for large/flexible systems. For path optimisations loose optimisations use a maximum of 50 cycles.

Bug Fixes

• Skips conformers with no atoms in finding unique conformers
• Corrects benchmark TS location for the Grubbs metathesis example, where the reactant complex is bound
• Fixes possible zero distance constraint for a single atom
• Fixes spin state definition for XTB calculations
• Fixes possible override of a constructor-defined spin state by the SMILES parser

1.0.5

---

Bugfix release

Bug Fixes

• Saves transition state templates with correct atom labels

1.0.4

---

Bug fixes in SMILES parser and 3D geometry builder from 1.0.3.

Usability improvements

• Improves doc strings
• Throws interpretable error when calling find_tss without reaction.reactant set

Functionality improvements

• SMILES strings with >9 ring closures are parsed correctly
• cis-double bonds in rings no longer minimise with constraints, which is a little faster

Bug Fixes

• Tweaks repulsion parameters in minimisation to build fused rings
• Enables SMILES parsing with "X(...)1" branching
• Fixes spin multiplicity for odd numbers of hydrogens
• Improves ring closure 3D build
• Fixes incorrect implicit valency for aromatic heteroatoms
• Improves metal finding in SMILES strings with regex
• Corrects atom type for sp2 group 16 elements
• Fixes dihedral rotation with atoms not close to any other

1.0.3

---

A minor API revision from 1.0.2 but adds C++ extension which should be extensible to further developments of fast C-based code.

Usability improvements

• autode.Species() inherit from a AtomCollection() base class for more flexibility
• autode.Constants attributes have more readable names (while retaining backwards compatability)
• autode.geom.length() as an explicit alias of np.linalg.norm has been removed
• autode.input_output.xyz_file_to_atoms() throws more informative errors
• autode.mol_graphs.make_graph() throws NoAtomsInMolecule for a species with no atoms
• species.formula and species.is_explicitly_solvated are now a properties
• autode.smiles.parser has been rewritten & is (hopefully) a more robust SMILES parser

Functionality improvements

• Metal complex initial geometries can now be generated with the correct stereochemistry
• Macrocycles default to an autodE builder that conserves SMILES stereochemistry (RDKit#1852)
• species.coordinates can be set from either 3xN matrices or 3N length vectors
• autode.Atom()`s have :code:.group` .period and .tm_row properties referring to their location in the periodic table
• autode.atoms.PeriodicTable added
• species.bond_matrix added as a property and returns a boolean array for interactions between all atom pairs

Bug Fixes

• reaction.calculate_complexes() calls reaction.find_complexes() if needed thus can be called in isolation

1.0.2

---

Usability improvements

• Effective core potentials can now be specified in Keywords()
• ORCA fitting basis sets now default to def2/J, which should be smaller but as accurate as AutoAux
• Molecule initialisation from a .xyz file now checks for an odd number of electrons. For example, Molecule('H_atom.xyz') will raise a ValueError but Molecule('H_atom.xyz', charge=1) or Molecule('H_atom.xyz', mult=2) are acceptable

Functionality improvements

• atom.atomic_number has been added as an atom attribute
• atom.atomic_symbol is a more intuitive alias for atom.label

1.0.1

---

Usability improvements

• Molecular complexes can now be initialised with a reasonable geometry Complex(..., do_init_translation=True)

Functionality improvements

• species.radius has been added as an approximate molecular radius (in Angstroms, excluding VdW radii)

Bug Fixes

• Final breaking bond distances are now the minimum of the product X-Y distance if present in the product, or 2x the distance. This is required for breaking bonds that cross a ring.
• Neighbour lists for comparing possibly equivalent bond rearrangements are now compared using a sorted list

1.0.0

---

The first stable release! Mostly documentation updates from v.1.0.0b3 with the package now being conda-install-able.

Usability improvements

• More documentation

Functionality improvements

• XTB wrapper now supports v. 6.4 (and hopefully higher)

Thanks to Joe, Alistair, Matina, Kjell, Gabe, Cher-Tian amongst others for their invaluable contributions.

1.0.0b3

---

This version brings several major changes and in some instances breaks backwards compatibility, but does feature significant improvements in speed and accuracy for finding transition states.

Usability improvements

• species.get_distance(i, j) is now species.distance(i, j)
• species.set_atoms(new_atoms) is now properly handled with a setter so species.atoms = new_atoms will set the new atoms
• species.n_atoms is more robust
• species.get_coordinates() is now species.coordinates, returning a numpy array copy of the species coordinates (Nx3 in Å)
• species.centre() will translate a species so it's coordinate centroid lies at the origin
• PBE0/def2-SVP is now the default 'low opt' method (keywords.low_opt) with loose optimisation. Path exploration uses this method, thus it needs to be very close to the 'opt' level

Functionality improvements

• 1D, 2D potential energy surface scans and nudged elastic band (NEB) methods to generate TS guesses from reactants have been replaced by an adaptive path search which seems to be very efficient for generating initial paths

For the prototypical SN2 between fluoride and methyl chloride the relaxed PES (PBE0-D3BJ/ma-def2-SVP/CPCM(water)) is

where the previously employed linear path (red) is compared to the adaptive scheme (blue, purple) and the 'true' intrinsic reaction coordinate. With a small minimum step size a path very close to the MEP is traversed with a very small number of required constrained optimisations. This enables NEB relaxations to be skipped and the associated limitations (corner cutting, oscillating path, optimisation in Cartesian coordinates) avoided. This exploration is essential when a linear path over multiple bonds leads to rearrangements, e.g. an (E2) elimination reaction the comparison for the linear, adaptive and IRC paths are shown below

• (CI)-NEB with adaptive force constant has been added
• Initial path exploration from reactants is performed at the 'low_opt' level with a final breaking bond distance below.

Previous implementations made use of a 1.5 Å additional shift for uncharged reactions and 2.5 Å for charged, this however lead to possible final C-H distances of ~3.6 Å and steps into unphysical regions. 1.0.0b3 uses an estimate based on the distance where the bond is mostly broken, as below

where X-Y corresponds to a molecule e.g. C-C with the appropriate hydrogens added then the BDE curve calculated at the GFN2-XTB level of theory. A multiplier of ~2 affords a 'mostly broken bond' (i.e. the distance at 3/4 of energy of the broken bond).

• There is now a heuristic used to skip TSs that go via small rings (3, 4-membered) if there is a >4-membered equivalent (ade.Config.skip_small_ring_tss)

Bug Fixes

• Calculations are now unique based on constraints, so NEB calculations executed in the same directory are not skipped with different bond rearrangements