7.3.1.2. ReaxFF¶

The ReaxFFParameters class provides a parameterizable interface to the ReaxFF reactive force field. It can be initialized by passing a force field path to the constructor:

ff = ReaxFFParameters('path/to/ffield.ff')

By default, all relevant parameters will be marked as is_active==True (i.e. marked for parameterization). See the API for parameter naming conventions.

7.3.1.2.1. Reading Legacy params Files ¶

Outside of the ParAMS formalism, a params file is common to mark a set of paramters for optimization, and set their respective ranges. Support for these files is implemented in the ReaxFFParameters.read_paramsfile() method:

ff = ReaxFFParameters('path/to/ffield.ff')
ff.read_paramsfile('path/to/params')

The method will mark all matching parameters as is_active=True (and deactivate all remaining ones). Will raise an error if blocks that are not present in the force field are referenced in the params file. Note that the 4th column in the params file (delta value) will be ignored.

7.3.1.2.2. Addition of Parameters ¶

The ReaxFF interface allows for the addition of new parameters through the add_blocks() method.

7.3.1.2.3. Trim (remove elements) from a ReaxFF force field ¶

Use the trim() method.

7.3.1.2.4. Apply recommended parameter ranges (constraints)¶

Note: This is only about setting technically meaningful upper/lower bounds on parameter values. Even when you apply these recommendations you are likely to have to modify the ranges yourself for your particular problem.

GUI: Select the parameters in the parameters table for which you would like to apply the recommended ranges. Then select Parameters → Apply Recommended Constraints. The recommended constraints are documented in the apply_default_constraints() method.
Python: Use the apply_default_constraints() method.

7.3.1.2.5. Parameter Databases and Value Suggestions ¶

The class can suggest suitable parameter values for a number of queried parameters. See the ReaxFFParameters.params_from_db() method for more information. By default, all force fields in $AMSHOME/atomicdata/ForceFields/ReaxFF will be used to return suggestions. Users can create their own databases from fewer (or more) force fields through the ReaxFFParameters.generate_paramsdb() method.

7.3.1.2.6. ReaxFF Conversion Functions ¶

Outside of ParAMS, ReaxFF is parameterized with a training set defined in trainset.in and geo files. There are two functions available to help with the conversion to the ParAMS format:

trainset_to_params(trainsetfile, use_weights=True, default_sigmas=None) → scm.params.core.dataset.DataSet ¶

Converts an a ReaxFF trainset.in file to a DataSet object.

Parameters

trainsetfilestr

Path to trainset.in.

use_weightsbool

If False, all weights will be set to 1.0, and Sigma will be set to the value from trainset.in

If True, then new_weight/(new_sigma**2) == 1/(value_from_trainset_in**2).

new_sigma is taken from the standard Sigma value for the pertinent extractor, or from the default_sigmas dictionary.

default_sigmasdict

A dictionary of the type

{ ‘energy’: sigma_energy, ‘forces’: sigma_forces, ‘charges’: sigma_charges, ‘angles’: sigma_angles,

‘distance’: sigma_distance, ‘dihedral’: sigma_dihedral }

The sigma values should be given in the classic ReaxFF units, so kcal/mol for energy, angstrom for distance, etc.

The sigma values are taken from this dictionary only if use_weights == True

Returns

ds : DataSet instance

geo_to_params(geofile: str, normal_run_settings: Union[scm.plams.core.settings.Settings, str]) → scm.params.core.jobcollection.JobCollection ¶

Convert a ReaxFF geo file to a JobCollection instance.

Each entry’s ID will be based on the DESCRP line. Duplicates will not be added.
Will guess the settings for each job based on the RUTYPE line. Can either be SinglePoint or GeometryOptimization (optionally with MaxIterations).

Parameters

geofilestr: Path to geo file.
normal_run_settingsplams.Settings, str: A plams.Settings instance representing the NORMAL RUN or the path to a control file from which the settings will be extracted

Returns

jc : JobCollection

7.3.1.2.7. ReaxFF Interface API ¶

class ReaxFFParameters(ffieldfile, bounds_scale=1.2, settings=None)¶

Interface to the ReaxFF force field format.

Attributes

In addition to the ones already availabe in BaseParameters

headerdict: A dictionary with metadata. header[‘head’] contains the first line of the force field
pathstr: The path to the last written ffield file.
blocksList[str]: Strings that mark valid parameter blocks. See ReaxFF manual for more information.

Note

This is a class attribute, meaning that an instance can contain fewer blocks.
allnamesDict[strList[str]]: A dictionary with BLOCK_NAME:List[PARAMETER_NAMES], where BLOCK_NAME is one of the elements from blocks. See also name_to_index() and names_in_block().

Note

This is a class attribute, meaning that an instance can contain fewer parameters.
shapeDict[strList[int]]: A dictionary of BLOCK_NAME:[NBLOCKS, N_PARAMS_IN_BLOCK], where BLOCK_NAME is one of the elements from blocks, NBLOCKS is the number of BLOCK_NAME blocks and N_PARAMS_IN_BLOCK is the number of parameters per block. See ReaxFF manual for more information.
flagsdict[str, str]: This attribute stores flags for special force field flavours. The data is read from the first line of the force field file, if it starts with [ X ], where X is a series of flag strings. Currently, the following flags are understood: ereaxff, ereaxff2, acks2, lgDispersion.

Parameter Naming Conventions:

Naming follows the tables as documented in the ReaxFF manual.
The name string is generated from the table’s Name in Eq column.
All atom-specific parameter names will start with the atom prefix, followed by the name: At1.At2.At3:NAME. For example, the name of the first parameter in the atoms block for a Hydrogen will be H:r_0^sigma.
See below for a full list of names.
Each parameter in the ReaxFF interface has the following additional attributes: block, block_index, equation, description, atoms_reversible. For easy filitering of parameters, see get(). For conversions between block indices and name strings, see the blocks, name_to_index() and names_in_block().

class Parameter(block, block_index, atoms, bounds_scale=0, *a, **kw)¶

__init__(block, block_index, atoms, bounds_scale=0, *a, **kw)¶: Initialize self. See help(type(self)) for accurate signature.

property expose¶

property atoms_reversible¶

classmethod from_parent(parent)¶

__init__(ffieldfile, bounds_scale=1.2, settings=None)¶

Initialize the class instance, reading a ffied from ffieldfile.

Parameters

ffieldfilestr: Path to the force field file.
bounds_scalefloat >= 0: Apply arbitrary bounds, by providing a factor which will determine lower/upper bounds by b = x0 ± (bounds_scale-1)*abs(x0).

Note

Bounds for any x0 == 0 will be set to +-1.
settingsoptional, Settings: A plams.Settings instance that will be added to the output of get_engine(). Must have the input.ReaxFF keys defined.

get_engine(parameters=None, path=None, write=True)¶

Note

When called without path, will store a ffield file in a temporary directory for as long as the returned Engine instance is referenced.

Returns: A ReaxFF Engine class matching the passed params, or the current set of parameters if None.

write(file=None, parameters=None)¶: Writes the current parameters to file.

Note

When called without file, will store the ffield file in a temporary directory.
Use the path attribute to get the file path.

get_constraints() → List[scm.params.parameterinterfaces.base.Constraint]¶

Returns a list of physically reasonable Constraints for this instance to be passed to the Optimization.

Will constrain the covalent radii and bond lengths $r_0$ for the $\sigma, \pi, \pi\pi$ bonds to $r_0^\sigma >= r_0^\pi >= r_0^\pi\pi$ and the Van der Waals radius $r_\mathrm{VdW}$ to $r_\mathrm{VdW}$ >= r_0^sigma`.

read_paramsfile(paramsfile)¶: Reads a legacy params file, sets all matching parameters to is_active=True (i.e. flags them as paramters to be fitted), and adjusts each parameter’s ranges to the ones from paramsfile.

add_blocks(params: List[str], allow_duplicates=False)¶

Adds new parameters to the current instance.
The addition happens block-wise by providing a list of strings, where each string represents the addition of one block. Strings must have the form At1.At2.At3.At4, where At represents the chemical symbol of the desired element (separated by dots) and the number of elements determines the block to be added (1 for the atomic block, 2 for bonds, 3 for angles, 4 for torsions). For example,

>>> ffield.add_blocks(['C.O', 'C.O.H'])

adds the C-O bonds block and the C-O-H angles block to the ffield instance. Off-diagonals and hydrogen-bonds can be added by prefixing a string with OFD: and HBD: respectively:

>>> ffield.add_blocks(['OFD:C.O', 'HBD:O.H.C'])

The above will add the C-O off-diagonal and O-H-C hydrogen-bond block.

Note

The following is of note when adding new blocks:

The addition of multiatomic blocks to a force field that does not yet contain all respective elements will automatically add the atomic block for that element. E.g., the addition of C-O to a force field that does not yet contain parameters for oxygen also implies the addition of O
Parameters added by this method have a value of 0. and a range of (-1, 1)
Already present blocks will not be overwritten, however, a duplicate block can be added by setting allow_duplicates=True

classmethod generate_paramsdb(paths=['$AMSHOME'], dbfile='reaxffdb.gz')¶

Generates a parameter database based on all force field files found in paths and stores it in dbfile.

Parameters

pathsList[str]: When generating the database, each entry in paths can be either a directory, or a full path to a force field file. If a directory, will scan recursively and consider all files with the .ff ending.
dbfileoptional, str: Stores the generated database in dbfile. Defaults to reaxffdb.gz.

classmethod params_from_db(params: List[str], dbfile=None, **subst_atoms) → List[collections.namedtuple]¶

Given a list of n parameter queries in params, scans a force field database stored in dbfile and returns n named tuples, which hold all parameters matching the query.

Properties

paramsList[str]

Scans the data base for all entries in params. Each element is a string with either of the following syntaxes:

‘BLOCK:At1.At2.At3.At4:INDEX’ or
‘At1.At2.At3.At4:NAME’

where At1.At2.At3.At4 are chemical symbols (one to four). For the general block, this is replaced by None.

In the former case, BLOCK has to be any of {‘GEN’, ‘ATM’, ‘BND’, ‘OFD’, ‘ANG’, ‘TOR’, ‘HBD’} referring to the General, Atoms, Bonds, Off-Diagonals, Angles, Torsions, Hydrogen-Bonds blocks respectively. And INDEX is an integer specifying the parameter number for the block (starting with 0).

In the latter case, NAME refers to the string name of the desired parameter, following the same naming as in the names attribute (see Full List of Parameter Names).

dbfile: optional, str

Path to the database file to be used for the search. Defaults to a database generated from $AMSHOME/atomicdata/ForceFields/ReaxFF. See generate_paramsdb() for more information.

subst_atoms: optional keywords

Additional keywords will be interpreted as chemical elements (keys) that can be substitued by a number of other elements (values). This extends each query by matching all valid combinations of chemical elements for a specific parameter, rather than limiting it to one specific set of elements. Useful when queries should also include chemically similar elements.

Example:

>>> cls.params_from_db(['BND:C.O:0'], C=('Si', 'Ge'), O=('S',))

Will search all first entries of the bonds block considering the follwing combinaitons of elements: C.O, C.S, Ge.O, G.S, Si.O, Si.S.

Returns

Returns a list of named 4-tuples. Each element in the list corresponds to a query. Each named tuple unpacks to (query, parametername, values, atoms, sources). Elements within the tuple are stored in the respective attribute (e.g. tup.values).

querystr: String used for the query, Same as in params
parameternamestr: String with the requested parameter’s name (excluding the chemical symbols prefix)
valuesarray: Array of all parameter values matching the query
atomsarray of str: Array of chemical elements corresponding to each returned parameter value
sourcesarray of str: Array of force field paths corresponding to the origin of each parameter value

classmethod block_from_db(block: str, dbfile=None)¶

blockstr: Should include the atoms. ‘GEN’, ‘BND:C.Cl’, ‘ANG:C.C.H’, etc.
dbfile: optional, str: Path to the database file to be used for the search. Defaults to a database generated from $AMSHOME/atomicdata/ForceFields/ReaxFF. See generate_paramsdb() for more information.

Returns a 2-tuple containing a dictionary and a list of parameter names. The keys of the dictionary are the sources (paths to .ff files), and the values are lists of parameter values.

names_in_block(block: Union[str, int], atoms: str = '') → List[str]¶: Returns the names of the all parameters in block, optionally prepended by an {atoms}: string. block can either be a an int in the range of [0, 6] or any of {‘GEN’, ‘ATM’, ‘BND’, ‘OFD’, ‘ANG’, ‘TOR’, ‘HBD’} referring to the General, Atoms, Bonds, Off-Diagonals, Angles, Torsions, Hydrogen-Bonds blocks respectively.

Note

Index numbering starts from zero

See also

block_to_names()

classmethod name_to_index(name: str, block_as='str') → Tuple[Union[int, str], int]¶: Given a parameter name, returns the indices at which this name is located within the force field blocks. The name argument should be a string of At1.At2:NAME, where At are the atoms (one to four) the parameter affects.
If block_as==’str’, the returned value is a (str, int), otherwise (int, int), where the first element specifies the block and the second one the index at which this is parameter located within the block. Blocks are represented by an int in the range of [0, 6] or strings of {‘GEN’, ‘ATM’, ‘BND’, ‘OFD’, ‘ANG’, ‘TOR’, ‘HBD’} referring to the General, Atoms, Bonds, Off-Diagonals, Angles, Torsions, Hydrogen-Bonds blocks respectively.

Note

Index numbering starts from zero.

copy_block(other_interface: scm.params.parameterinterfaces.reaxff.ReaxFFParameters, from_block: str, to_block: Optional[str] = None, copy_range: bool = True, allow_duplicates=False)¶

Copies the values (and optionally ranges) of a whole parameter block from other_interface to this instance.

Parameters

other_ff: ReaxFFParameters: Copy the parameters in the block from this other instance
from_block: str: ‘BND:C.C’ or similar. It is also possible to copy the complete block, _i.e._ all values in ‘BND’ or ‘OFD’ by passing only the three-letter block name.
to_blockstr or None: ‘BND:C.O’ or similar. If None, set to the same value as from_block
copy_range: bool: Whether to copy the .range attribute from the other interface

get(name: Optional[str] = None, atoms: Optional[str] = None, block: Optional[Union[str, int]] = None, index: Optional[int] = None) → List[scm.params.parameterinterfaces.reaxff.ReaxFFParameters.Parameter]¶

Returns a list of parameters in self based on a set of filters. Filtering can be based on:

Parameters

namestr: Match a all parameters with name. Can be exact in the form of At1.At2:NAME, where At are the atoms for a parameter, or fuzzy in the form of NAME, meaning that all paraemters with NAME (regardless of atoms) will be matched
atomsstr: A string of the form ‘C.H’, ‘C.C.H’, etc
blockstr or int: Parameter block name (str) or index (int), matching :attr:blocks
indexint: The parameter at index within a block.

trim(keepatoms: Optional[List[str]] = None)¶

Returns a new ReaxFFParameters instance, only keeping parameters describing interactions between atoms in the keepatoms list.

keepatoms: list of str: Elements to keep, e.g. [‘C’, ‘H’] or [‘C’, ‘H’, ‘0’]

NOTE: to save wildcard parameters, you need to explicitly include ‘0’ as one of the keepatoms.

Example:

>>> old = ReaxFFParameters('CHON2017_weak_bb.ff')
>>> new_interf = old.trim(keepatoms=['C', 'H', 'O'])

will keep the GEN block, ATM:C, …, BND:C.H, …., ANG:C.C.H, …., etc.

apply_default_constraints(only_active: bool = False, verbose: bool = False, parameter_names: Optional[List[str]] = None)¶

Changes the parameter values and ranges to agree with the default constraints.

Example: Below the p_bo4 parameter is set to a value greater than or equal to 1.

>>> ff = ReaxFFParameters('CHO.ff')
>>> p = ff['H.H:p_bo4']
>>> p.is_active = True
>>> print(p.value, p.range, p.is_active)
0.0 [-1.0, 1.0] True
>>> ff.apply_default_constraints(only_active=True)
>>> print(p.value, p.range, p.is_active)
1.0 [1.0, 1.0] True

The following values/ranges will be altered:

('ATM', 'alpha_ij') >= 0
('OFD', 'alpha_ij') >= 0
('ATM', 'D_ij') >= 0
('OFD', 'D_ij', >= 0
('ATM', 'p_boc3') >= 0
('ATM', 'p_boc4') >= 0
('ATM', 'p_boc5') >= 0
('ATM', 'r_0^sigma') >= 0
('OFD', 'r_0^sigma') >= 0
('ATM', 'r_vdW') >= 0
('ATM', 'gamma_i') >= 0
('ATM', 'Val_i^e') >= 0
('BND', 'D_e^pi') >= 0
('BND', 'D_e^pipi') >= 0
('BND', 'D_e^sigma') >= 0
('BND', "Val'_i^boc") >= 0
('BND', 'p_be2') >= 0
('BND', 'p_ovun1') >= 0
('ANG', 'Theta_0,0') >= 0
('ANG', 'p_val2') >= 0
('ANG', 'p_val7') >= 0
('HBD', 'r_hb^0') >= 0

('ATM', '1/gamma_w') > 0.5

('BND', 'p_bo6') >= 1
('BND', 'p_bo4') >= 1
('BND', 'p_bo2') >= 1
('ANG', 'p_val4') >= 1

('ANG', 'Theta_0,0') < 180

only_active: bool: If True, make changes only to the active parameter subset
verbose: bool: Prints all changed parameters and the reason for changing them to stdout
parameter_names: list of str: Names of the parameters to apply default constraints to. If not set, all parameters will be considered. Note: the value of only_active is still considered.

get_default_constraints(only_active=True)¶

Returns a list of constraints that can be used in the input Constraints block in ParAMS (if each entry is separated by a newline).

only_activebool: Whether to only give constraints for the active subset.

classmethod yaml_load(yamlfile, *a, **kw)¶: Loads a parameter interface from a (compressed) human-readable yamlfile. Child classes should call super() in this method before implementing their own routines.

7.3.1.2.8. Full List of Parameter Names ¶

General block, no prefix

p_boc1
p_boc2
-p_coa2
p_trip4
p_trip3
k_c2
p_ovun6
p_trip2
p_ovun7
p_ovun8
p_trip1
nonb_low,swa
R_cut
p_fe1
p_val6
p_lp1
p_val9
p_val10
p_fe2
p_pen2
p_pen3
p_pen4
p_fe3
p_tor2
p_tor3
p_tor4
p_elho only if eReaxFF is enabled
p_cot2
p_vdW1
cutoff*100
p_coa4
p_ovun4
p_ovun3
p_val8
X_soft
n/a 1
p_val only if eReaxFF is enabled
n/a 2
p_coa3
n/a 3
n/a 4 only if eReaxFF is enabled

Atoms block, prefix: At1:

r_0^sigma
Val_i
n/a 1
r_vdW
D_ij
gamma_i
r_0^pi
Val_i^e
alpha_ij
1/gamma_w
Val_j^angle
p_ovun5
p_i^xel2 only if eReaxFF is enabled
chi_i
eta_i
n/a 2
r_0^pipi
p_lp2
n/a 3
p_boc4
p_boc3
p_boc5
C_i
alpha_i only if eReaxFF is enabled
p_ovun2
p_val3
beta_i only if eReaxFF is enabled
Val_i^'boc
p_val5
p_c1
p_c2
p_c3
C_i only if Lg dispersion is enabled
R_eij only if Lg dispersion is enabled

Bonds block, prefix: At1.At2: (equivalent to At2.At1:)

D_e^sigma
D_e^pi
D_e^pipi
p_be1
p_bo5
Val'_i^boc
p_bo6
p_ovun1
p_be2
p_bo3
p_bo4
n/a
p_bo1
p_bo2
delta'_i
p_ij^xel1 only if eReaxFF is enabled

Off-diagonals block, prefix: At1.At2: (equivalent to At2.At1:)

D_ij
r_vdW
alpha_ij
r_0^sigma
r_0^pi
r_0^pipi
C_i,C_lg,ij only if Lg dispersion is enabled

Angles block, prefix At1.At2.At3: (equivalent to At3.At2.At1, At2 is the central atom)

Theta_0,0
p_val1
p_val2
p_coa1
p_val7
p_pen1
p_val4

Torsions block, prefix At1.At2.At3.At4: (equivalent to At4.At3.At2.A1:)

V_1
V_2
V_3
p_tor1
p_cot1
n/a 1
n/a 2

Hydrogen Bonds block, prefix At1.H.At2: (NOT equivalent to At2.H.At1:)

r_hb^0
p_hb1
-p_hb2
-p_hb3

7.3.1.2.9. Parameter Categories ¶

To guide your selection we have split the ReaxFF parameters into three categories:

Standard: Parameters which are generally safe to optimize.
Expert: Parameters which require some expert knowledge/insight. Generally, these are parameters which should be handled with more care and should only be tuned at the end of the parametrization process, if at all.
DoNotOptimize: Parameters which should never be activated and changed through optimization. This final category generally represents parameters which act as boolean flags or switches, are not used by the force field, or contain physical constants like atomic mass.

Please note that these categorizations are entirely empirical and based only on in-house experience and expert knowledge. They are not definitive groupings, and serve only to better inform and guide your own decisions.

The category for each ReaxFF parameter is shown in a separate column of the GUI. When scripting it can be accessed via the metadata attribute of the Parameter class e.g. p.metadata.category or p.metadata['category'].