Descriptors API

calculate_descriptors

AtomicAI.descriptors.calculate_descriptors.calculate_descriptors()[source]

generate_force_descriptor

AtomicAI.descriptors.generate_force_descriptor.force_descriptor(selected_frames, fp_type='Split2b3b_ss', rc=10.5, n2b=20, n3b=10)[source]

Compute force descriptors for all atoms in selected_frames.

Parameters:

  • fp_type (str) – Fingerprint type — ‘BP2b’ (2-body Behler-Parrinello) or ‘Split2b3b_ss’ (split 2+3-body).

  • rc (float) – Cutoff radius in Angstrom applied to both 2-body and 3-body terms.

  • n2b (int) – Number of 2-body eta decay functions.

  • n3b (int) – Number of 3-body eta decay functions.

AtomicAI.descriptors.generate_force_descriptor.generate_force_descriptors()[source]
AtomicAI.descriptors.generate_force_descriptor.prepare_vforce(data_num_rand)[source]
AtomicAI.descriptors.generate_force_descriptor.writeout_fp(Xv, Fv, atomic_symbols, nfout)[source]

laaf

Locally averaged atomic fingerprints (LAAF) — G2, G3, G4, G5, SOAP, MBSF.

class AtomicAI.descriptors.laaf.AverageFingerprintCalculator(cutoff_descriptor=5.0, cutoff_average=4.0, traj_data='traj_data', selected_snapshots=':', number_of_eta=100, element_conversion=None, descriptor_type='custom', descriptor=None, descriptor_parameters=None)[source]

Bases: object

Main class for calculation of LAAF

compute_averaged_fingeprints_selection(output_file='average_fingerprint.csv', selected_atoms=None, append=False, selected_steps=None, target_element=0, target_neighbor_element=None)[source]
compute_averaged_fingerprints_random(output_file='average_fingerprint.csv', append=False, number_of_data=None, start_step=0, final_step=None, target_element=0, target_neighbor_element=None, seed=None)[source]
AtomicAI.descriptors.laaf.calculate_average_fingerprint(cutoff_descriptor=5, cutoff_average=4, traj_data='traj_data', selected_snapshots=':', output_file='average_fingerprint.csv', append=False, target_element=0, number_of_data=None, number_of_eta=100, start_step=0, final_step=None, element_conversion=None, seed=None, lattice_parameters=None, local_descriptor_type=None, descriptor=None, use_buffer=False, descriptor_parameters=None, target_neighbor_element=None)[source]

Calculate averaged fingerprints for randomly selected center atoms. Center atomic sites are selected by target_element, neighbor atomic sites are selected by target_neighbor_element. REMARK: Descriptor order is different between original and acsf g2: original is 0-0/0-1 while acsf g2 is 0-1/0-0

Parameters:

  • target_neighbor_element (int)

  • descriptor_parameters – specific parameters for descriptors. For advanced use!

  • selected_snapshots (str) – str compliant with ase.io.read function

  • descriptor – instance of descriptor # TODO: List all available descriptors

  • local_descriptor_type – name of descriptor # TODO: Can it be inferred from “descriptor”?

  • append (bool) – If True, will append at the end of the csv file. Else overwriting it.

  • lattice_parameters – For normal xyz files

  • cutoff_descriptor (float)

  • cutoff_average (float)

  • input_file

  • traj_data (str)

  • output_file (str)

  • target_element (int)

  • number_of_data (int)

  • number_of_eta (int)

  • start_step (int) – 0 = first selected snapshot

  • final_step (int)

  • element_conversion (dict)

  • use_buffer (bool) – store calculated fingerprint vectors for later random choice

  • seed (int)

Returns:

AtomicAI.descriptors.laaf.calculate_eta(r0=0.45, cutoff_descriptor=5.0, number_of_eta=100)[source]

Calculate the decay function [goes from R0 to cutoff]

Return type:

list

Returns:

list of eta

AtomicAI.descriptors.laaf.calculate_fingerprint_vector(x_ij, y_ij, z_ij, atomic_type_list, lattice_a, lattice_b, lattice_c, eta_all_list, atom_type, cutoff_descriptor)

Calculate the fingerprint vector for a single atom i. Original implementation of Botu descriptor.

Parameters:

  • x_ij (float64)

  • y_ij (float64)

  • z_ij (float64)

  • atomic_type_list (int64)

  • lattice_a (float64)

  • lattice_b (float64)

  • lattice_c (float64)

  • eta_all_list (float64)

  • atom_type (int)

  • cutoff_descriptor (float)

Returns:

acsf

Atom-Centered Symmetry Functions only G2 function[1]

Notice that the order of pair in descriptors follows the order of defined species

[1] Jörg Behler, “Atom-centered symmetry functions for constructing high-dimensional

neural network potentials”, The Journal of Chemical Physics, 134, 074106 (2011), https://doi.org/10.1063/1.3553717

class AtomicAI.descriptors.acsf.ACSF(cutoff_descriptor, params_g2=None, params_g3=None, params_g4=None, params_g5=None, species=None, periodic=True)[source]

Bases: object

create(system, positions=None)[source]

Return the ACSF output for the given systems and given positions.

Parameters:

  • system (ase.Atoms or list of ase.Atoms) – One or many atomic structures.

  • positions (list) – Positions where to calculate ACSF. Can be provided as cartesian positions or atomic indices. If no positions are defined, the output will be created for all atoms in the system. When calculating for multiple systems, provide the positions as a list for each system.

Returns:

np.ndarray

class AtomicAI.descriptors.acsf.ACSF_Force(rcut=6.0, params_v2b=None, params_v3b=None, species=None, periodic=True)[source]

Bases: object

create(system, positions=None)[source]

Return the ACSF output for the given systems and given positions.

Parameters:

  • system (ase.Atoms or list of ase.Atoms) – One or many atomic structures.

  • positions (list) – Positions where to calculate ACSF. Can be provided as cartesian positions or atomic indices. If no positions are defined, the output will be created for all atoms in the system. When calculating for multiple systems, provide the positions as a list for each system.

Returns:

np.ndarray

AtomicAI.descriptors.acsf.calculate_mirror_cubic(position, cell, Rc)[source]
AtomicAI.descriptors.acsf.get_g2(xij, yij, zij, atomic_type_list, lattice_a, lattice_b, lattice_c, params_2b_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • n_species (int)

  • params_2b_list (float64)

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • lattice_a (float64)

  • lattice_b (float64)

  • lattice_c (float64)

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.acsf.get_g3(xij, yij, zij, atomic_type_list, lattice_a, lattice_b, lattice_c, params_g3_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • params_g3_list (float64)

  • n_species (int)

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • lattice_a (float64)

  • lattice_b (float64)

  • lattice_c (float64)

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.acsf.get_g4(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_g4_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • n_species (int)

  • params_g4_list (float64)

  • m_z (float64)

  • m_x (float64)

  • m_y (float64)

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.acsf.get_g5(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_g5_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector of G5 for a single atom i

Parameters:

  • n_species (int)

  • params_g5_list (float64)

  • m_z (float64)

  • m_x (float64)

  • m_y (float64)

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.acsf.get_v2b(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_2b_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • m_x (float64)

  • m_y (float64)

  • m_z (float64)

  • params_2b_list (float64)

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.acsf.get_v3b(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_v3b_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • m_x (float64)

  • m_y (float64)

  • m_z (float64)

  • params_v3b_list (float64)

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.acsf.random_vforce(data_num_rand)[source]
AtomicAI.descriptors.acsf.set_eta(r0=0.45, cutoff_descriptor=5.0, number_of_eta=100)[source]

Calculate the decay function [goes from R0 to cutoff]

Returns:

list of eta

mbsf

Modified Behler-Parrinello symmetry functions (MBSF) [1], is a modified version of Behler-Parrinello atom-centered symmetry functions (ACSF)[2]. Support two-body and three-body fingerprints

Notice that the order of pair in descriptors follows the order of defined species

[1] Smith J S, Isayev O and Roitberg A E 2017 Chem. sci. 8 3192 [2] Jörg Behler, “Atom-centered symmetry functions for constructing high-dimensional

neural network potentials”, The Journal of Chemical Physics, 134, 074106 (2011), https://doi.org/10.1063/1.3553717

class AtomicAI.descriptors.mbsf.MBSF(cutoff_descriptor, params_gr=None, params_ga=None, species=None, periodic=True)[source]

Bases: object

create(system, positions=None)[source]

Return the ACSF output for the given systems and given positions.

Parameters:

  • system (ase.Atoms or list of ase.Atoms) – One or many atomic structures.

  • positions (list) – Positions where to calculate ACSF. Can be provided as cartesian positions or atomic indices. If no positions are defined, the output will be created for all atoms in the system. When calculating for multiple systems, provide the positions as a list for each system.

Returns:

np.ndarray

AtomicAI.descriptors.mbsf.calculate_mirror_cubic(position, cell, Rc)[source]
AtomicAI.descriptors.mbsf.get_ga_nblst(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_ga_list, cutoff_descriptor, n_species, nb_index, nb_index_shift, nb_r2_lst)

Calculate the fingerprint vector for a single atom i

Parameters:

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • lattice_a

  • lattice_b

  • lattice_c

  • eta_all_list

  • atom_type

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.mbsf.get_gr(xij, yij, zij, atomic_type_list, lattice_a, lattice_b, lattice_c, params_2b_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • lattice_a (float64)

  • lattice_b (float64)

  • lattice_c (float64)

  • eta_all_list

  • atom_type

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.mbsf.get_gr_nblst(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_gr_list, cutoff_descriptor, n_species, nb_index, nb_index_shift, nb_r2_lst)

Calculate the fingerprint vector for a single atom i

Parameters:

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • lattice_a

  • lattice_b

  • lattice_c

  • eta_all_list

  • atom_type

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.mbsf.get_grga(xij, yij, zij, atomic_type_list, m_x, m_y, m_z, params_g4_list, cutoff_descriptor, n_species)

Calculate the fingerprint vector for a single atom i

Parameters:

  • xij (float64)

  • yij (float64)

  • zij (float64)

  • atomic_type_list (int64)

  • lattice_a

  • lattice_b

  • lattice_c

  • eta_all_list

  • atom_type

  • cutoff_descriptor (float)

Returns:

AtomicAI.descriptors.mbsf.get_neighbour_lst(xij, yij, zij, m_x, m_y, m_z, radius_cutoff)